Datasets:

kilian-group
/

arxiv-classifier

Dataset card Data Studio Files Files and versions Community

Subset (4)

default · ~99.6k rows (showing the first 120k)

Split (2)

train · ~72.4k rows (showing the first 93.1k)

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
1809.10806	1	1809	2018-09-28T00:19:47	Direct observation of electron capture & emission processes by the time domain charge pumping measurement of MoS2 FET	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Understanding interface properties in MoS2 field effect transistors with a high-k gate insulator is critical for improving the performance of the device. Here, by applying the time domain charge pumping method, the elementary process for capture and emission of electrons to the interface states is monitored directly using a fast acquisition system. The main outcome is the detection of the clear difference in the capture and emission process of electrons to the interface states. In addition to the transient current response for gate capacitance, the current peak is observed during electron capture, while the broad tail is detected during electron emission. This different behavior is associated with the fact that the time constant for electron capture is much shorter than that for electron emission. Moreover, Dit is evaluated to be in the range of 1012 - 1013 cm-2 eV-1, which is comparable with that estimated from subthreshold swing.	physics.app-ph	physics	Direct observation of electron capture & emission processes by the time domain charge pumping measurement of MoS2 FET Koki Taniguchi,1 Nan Fang1 and Kosuke Nagashio1,* 1Department of Materials Engineering, The University of Tokyo, Tokyo 113-8656, Japan E-mail: [email protected] Abstract Understanding interface properties in MoS2 field effect transistors with a high-k gate insulator is critical for improving the performance of the device. Here, by applying the time domain charge pumping method, the elementary process for capture and emission of electrons to the interface states is monitored directly using a fast acquisition system. The main outcome is the detection of the clear difference in the capture and emission process of electrons to the interface states. In addition to the transient current response for gate capacitance, the current peak is observed during electron capture, while the broad tail is detected during electron emission. This different behavior is associated with the fact that the time constant for electron capture is much shorter than that for electron emission. Moreover, Dit is evaluated to be in the range of 1012 - 1013 cm-2 eV-1, which is comparable with that estimated from subthreshold swing. the by in capacitance-voltage The demonstration of a natural thin-body MoS2 field- effect transistor (FET) with an effective channel length of ~3.9 nm has facilitated research on 2-dimensional (2D) layered channels due to overcoming the scaling limit of ~5 nm for Si gate length.1 The dominant issue necessary for practical application is improvement in the high-k/MoS2 interface properties, even though there are many other issues to be solved due to the nascence of 2D research. Ideally, the dangling bond-free surface of the layered channel is expected to provide an electrically inert interface. (C-V) using However, the multilayer MoS2 metal-oxide- measurements semiconductor capacitor structure,2-5 frequency dispersion of the top gate capacitance is observed. The interface trap densities (Dit) in the range of ~1012 - 1013 cm- 2 eV-1 have been extracted and show band tail energy dispersion and often an additional hump. On the other hand, the recent C-V study on MoS2 FET structure6 has revealed that the frequency dispersion observed in previous literature7-9 resulted from the high channel resistance, thereby making in a straightforward manner. Based on these trial-and-error efforts, several physical origins, such as bond bending of the Mo d orbital,9 S vacancy,3,5,10 and trap sites in the oxide, have been proposed. To further understand the origin for the interface states, it is important to analyze the elementary process for capture and emission of electrons, which is obscured in the conventional C-V measurement because the time-averaged steady-state current is measured. Charge pumping (CP) is one of the most widely used techniques for analyzing the interface states of metal- oxide-semiconductor FETs (MOSFET).11-13 In CP, during the single gate voltage (VG) pulse between accumulation and inversion, electrons that flow into the conduction band are captured in interface states, and some of these electrons recombine with holes that flow into the valence band, as shown in Supplementary figure S1. That is, the current to extract Dit it difficult 1 this flows from the substrate to the source and drain through the interface states. Experimentally, by applying the VG pulse string using a pulse generator, time-averaged recombination current (CP current) is monitored by a direct current ammeter, and the interface defect density (Nit, cm- 2) can be evaluated. Recently, CP has been examined with respect to time domain measurement, where the transient currents are monitored directly during a single VG pulse with ramping and falling times of ~10 s using a fast acquisition system.14-16 The main outcome is the detection of the clear difference in the capture and emission process of electrons to the interface states. In addition to the transient current response for gate capacitance, the current peak is observed during electron capture, while the broad tail is detected during electron emission. This different behavior is associated with the fact that the time constant for electron capture is much shorter than that for electron emission. Here, conventional CP is not applicable to the monolayer MoS2 FET, because the source/drain electrodes are connected only to the conduction band due to Fermi level pinning.17,18 Therefore, the recombination current cannot be measured. On the other hand, time domain CP is possible since the behavior of electrons that flow in and out through the source and drain can be directly monitored. Although pulse I-V measurements, which is one of time resolved measurements, have been applied to Si,19,20 graphene,21-23 and MoS2 24,25 so far, time domain CP has not been studied in 2D material systems. In this study, time domain CP is applied to monolayer MoS2 FETs to elucidate the carrier response to the interface trap sites during a single VG pulse by analyzing the transient response of the top gate capacitor. Moreover, Dit is also evaluated to compare the data by different methods. Monolayer MoS2 films were mechanically exfoliated onto a quartz substrate from natural bulk MoS2 flakes.6 Although it is slightly difficult to find the monolayer using an optical microscope due to the transparent substrate, it is still possible. Ni/Au were deposited as the source/drain electrodes. For the top gate formation, a Y2O3 buffer layer was first deposited, and then, an Al2O3 oxide layer with a thickness of 10 nm was deposited via atomic layer deposition (ALD), followed by the Al top-gate electrode formation. The detailed procedure for the top gate formation can be found elsewhere.26,27 The device image can be seen in Figure 1. Alternatively, the monolayer MoS2 FET is also fabricated on a SiO2/n+-Si substrate for comparison. The I-V and C-V measurements were conducted using Keysight B1500 and 4980A LCR meters, respectively. For time domain CP, a Keysight B1530A WGFMU with minimum time resolution of 10 ns was used. All electrical measurements were performed in a vacuum prober at room temperature, where the measurement cable length from the acquisition system to the probe tip (~650 mm) is kept as short as possible to minimize signal reflection. First, time domain CP was carried out using the Si N- MOSFET test device with intentional electrical stress to confirm the measurement system. The time domain CP is conducted, where the sharp peak during the electron capture and broad tail during the electron emission are evident (see Supplementary Figure S2). This preliminary measurement is consistent with previous results,14,15 suggesting the validity of the present set up. Now, let us move to the time domain CP for the monolayer MoS2 FET. Figure 2(a) shows the transfer characteristics of MoS2 FET on the quartz substrate, where the typical n-channel behavior is observed. The two- terminal field effect mobility (FE), subthreshold swing (S.S.), and threshold voltage (VTH) are ~7 cm2 V-1 s-1, ~210 mVdec-1, and -2.2 V, respectively. For the FE extraction, the contribution of quantum capacitance is neglected, and CTG is determined to be 0.22 Fcm-2 as the capacitance at the accumulation in the C-V measurement.9 For time domain CP, a single top gate voltage (VTG) pulse is applied, while the current through the source and drain (IS/D) is measured using the fast acquisition set up, as shown in Figure 1(c). Figure 2(b) schematically illustrates the single VTG pulse and the IS/D as a function of time. The final voltage (Vfin) is always fixed to be 5 V, while the initial voltage (Vinit) is changed in the range from -5 V to 3 V. The area integral of the hatched region ideally indicates the charge (Q) accumulated in the MoS2 by the gate capacitor. Q is simply estimated here by Q = CTG  VTG, where VTG is defined as Vfin -- Vinit. Figure 2(c) shows Q estimated as a function of VTG for the VTG ramping case with tr = tw = tf = 10 s, where tr, tw, and tf are the ramping time, the waiting time, and the falling time, respectively. The red solid circle shows Q obtained for MoS2 FET on the quartz substrate, which agrees well with Q calculated using CTG = 0.22 Fcm-2. On the other hand, the blue rectangles show Q obtained for another MoS2 FET on the SiO2/n+-Si substrate, which largely deviates from the calculated line. Figure 1. (a) Optical micrograph of the device on the quartz substrate. (b) Magnified image of monolayer MoS2 FET with the Al2O3 top gate insulator. (c) Schematic drawing of MoS2 top gate transistor and measurement set up. The source and drain are tied together, and a pulse voltage is applied to the gate. It should be noted that conventional CP cannot be applied to the present device structure because hole conduction is prohibited by the n-type access region as well as the n-type contact. 2 FIG 1K. Taniguchi, et al.W×L = 23μm×16μm =261μm2STGD5 mm(a)SDTG20 μmQuartzsubstrate(b)Access region (c)Prober stageQuartz substrateMoS2FETAl2O3 (10 nm)Quartz (0.5 mm)Y2O3buffer layer(1 nm)AIS/DAccessregion This is due to the large parasitic capacitance (CPara) between the electrode pad and n+-Si substrate. These results suggest that utilization of the highly insulating quartz substrate is critical to extract the correct transient current during high-frequency measurement of a few microseconds. Since the quantitative validity of the macroscopic transient current for the gate capacitor is confirmed, the elementary processes of electron capture to the interface states and electron emission from the interface states, which are expected in the transient current response, are examined herein. Figure 3(a) shows IS/D as a function of time at different time conditions at the fixed VTG pulse of Vinit = -5 V and Vfin = 5 V. The transient current curves for the falling stage are inverted for comparison with the transient curves for the ramping stage. For tr = tw = tf = 1,000 s, the signal to noise ratio (S/N) is relatively poor, which slightly blinds the signature on the interaction between electrons and the interface states. It should be noted that the time domain charge pumping method requires large current level to get the high time resolution. Therefore, the large top gate area (> 250 m2) and short access region length are intentionally prepared, and the cables are connected to both the source and drain electrodes, as shown in Figures 1(b) and (c). Then, the adequately high S/N has been obtained. For tr = tw = tf = 100 s, a sharp peak in the transient charging current is clearly observed for electron capture as shown by an arrow, while the broad tail behavior is evident as shown by the dotted circle. For a shorter pulse of tr = tw = tf = 10 s, the current peak becomes stronger. This is because Dit increases close to the conduction band.9 Finally, it becomes very broad for tr = tw = tf = 1 s (not shown here) because the measurement current level is insufficient for this short pulse time due to the limited MoS2 channel area. The current peak is observed during the voltage ramping because some of the electrons accumulated in the conduction band are instantly trapped in the interface states, leading to the current peak, as shown schematically in Figure 2(b). On the other hand, the broad current tail during the voltage falling can be explained by the fact that the time constant for electron emission (e) is much larger than that for electron capture (c). It should be noted that this visible band tail is only a small part of the distribution of the electron emission times. This can easily be seen in Figure 3 for short rise/fall times since the area under both curves differs. Therefore, the major part of the electron emission most likely takes place Figure 2. (a) Transfer characteristics of monolayer MoS2 FET on the quartz substrate. (b) Schematic illustration for VTG and IS/D as a function of time, and for electron capture and emission. Schematic illustration for electron capture and emission behavior is also shown. (c) The charge Q stored for the top gate capacitor as a function of VTG. 3 10-1210-1110-1010-910-810-710-602x10-74x10-76x10-78x10-71x10-6-4-2024Drain current / ATop gate voltage / VDrain current / Aon quartzon SiO2/n+-Sion quartzon SiO2/n+-Si(b-i) Electron capture05101520250246810Q / pCVTG / VTimeIS/D(b-ii) Electron emissionVTGVinitVfinExp.Calc.Effect of Cpara(a)(c)(b)(fixed)(changed)@VD= 100 mV, RTVTH= -2.2 V trtwtftk0EFEcEvEcEvFrom S/DTo S/DEFFIG 2K. Taniguchi, et al. during Vint and is below the measurement resolution. These behaviors are consistent with those observed in Si- MOSFET. Even for the 2D layered channel "ideally" without the dangling bond, the elementary process for interaction between the electrons and the interface states is suggested to be similar to that for Si. It should be noted that the observation of the current peak in Si is possible only after the interface states are intentionally formed by electrical stress.15,16 Therefore, the present stage of the high-k/MoS2 interface quality is worse than that for Si. The proper time condition with relatively high S/N is tr = tw = tf = 100 s. Next, under this time condition, Vinit is altered from depletion to VTH to reveal how the transient current response changes. As shown in Figure 3(b), the current peak gradually disappears at around Vinit = -3.4 V (not shown here) since the trap sites at the interface have already been occupied at the large initial Fermi energy (EF). At the same time, the broad band tail also disappears because EF remains above the trap state. VTG at the current peak (Vpeak) is calculated and plotted in Figure 3(c). Although the time at the current peak position shifts for different Vinit, Vpeak is almost constant in the subthreshold region for all Vinit values. This is because the ramping time is held constant (tr = 100 s). Once tr (= tw = tf) is reduced from 103 to 101 s, Vpeak shifts closer in value to VTH, but remains more negative than VTH. This indicates that the electron capture to the interface states finishes before the channel is completely formed. Moreover, this is the origin for the gentle slope of S.S., which is often encountered in transfer curves obtained by DC measurement.9 Since the electron capture and emission processes are confirmed in the transient current response, the interface quality should be evaluated quantitatively. In the case of conventional CP,11,12,16 Nit is estimated, which is the interface trap density averaged through the "whole" energy gap. In case of MoS2 FET, however, the recombination current during a gate voltage pulse between the conduction and valence bands cannot be obtained because there is no electrical connection to the valence band. Therefore, let us extract Dit as a function of VTG. The key feature is that the area integral observed for the transient current during VTG ramping is larger than that observed during falling VTG, because most of the broadly distributed electron emission takes place during Vint and is therefore below the measurement range. Figure 4(a) shows a schematic illustration for Dit as a function of VTG, where the band tail distribution of Dit is adapted for simplicity.9 In the capture process, all the interface states swept from Vinit to Vfin capture electrons, as shown by the region (i). On the other hand, a deeper energy of the interface states in the emission Figure 3. (a) IS/D as a function of time at different ramping/waiting/falling times. Vinit and Vfin are fixed at -5 V and 5 V, respectively. The transient current curves for the falling stage are inverted for comparison with the transient curves for the ramping stage. (b) IS/D as a function of time at tr = tw = tf = 100 s for ramping (left) and falling (right). Vinit is changed from -5 V to -2 V, while Vfin is fixed at 5 V. The transient current curves for the falling stage are inverted for comparison with the transient curves for the ramping stage. (c) Vpeak as a function of Vinit and tr (= tw = tf). 4 FIG 3K. Taniguchi, et al.05105001000150020002500IS/D / nATime / s-2002040608010012050100150200250IS/D / nATime / s(a)tr= tw= tf= 1000 str= tw= tf= 100 str= tw= tf= 10 srampingfalling(c)020040060080010001200510152025IS/D / nATime / sVTH-20020406080100120100150200IS/D / nATime / s-20020406080100120100150200IS/D / nATime / s(b)tr= tw= tf= 100 s-5 -3-2-4.4-4-3.6rampingfalling@Vinit/ V-3.2-3-2.8-2.6-2.4-2.2-2-5.0-4.5-4.0-3.5-3.0100101102103Vpeak / VVinit / Vtr / s process leads to a longer e. Therefore, electrons captured already in the deep interface states cannot escape during falling VTG, while electrons with relatively small e near the band edge can escape from the interface states. This is the region (ii). Therefore, the difference in the area integral of IS/D for the ramping and falling of VTG provides the charge Q for electrons captured to Dit in the region (iii). When Vinit is changed to Vinit' and Vfin is fixed, Q is also changed to Q'. In this case, the average Dit in the voltage range of Vinit' - Vinit can be defined as (Q - Q') / (e2  (Vinit' - Vinit)), where e is elementary charge. It is assumed that the Fermi energy has a linear relation with VTG for simplicity. Before quantitatively estimating Dit, one important caution should be taken into consideration. At Vinit, the system should be in equilibrium, that is, the interface states above EF at Vinit should be open as shown in Figure 2(b-ii). However, MoS2 FET is normally on-state at VTG = 0 V. As shown in Figure 2(b), when VTG is set to Vinit from 0 V, the system may not be in equilibrium, because electrons must escape from the interface states. Therefore, a proper Vinit keep time, which is indicated by tk in Figure 2(b), should be found. Figure 4(b) shows Q as a function of tk at different Vinit. As expected, Q increases with increasing tk and saturates for Vinit = ~2.5 - 4 V because Q for region (i) increases, and Q for region (ii) is almost constant. On the other hand, Q continuously increases for Vinit = -5 V due to the interface states newly generated by the electrical stress. Therefore, Vinit > -3.5 V and tk' = 10 s are adapted as Figure 4. (a) Schematic to explain Dit extraction. (b) Q as a function of tk for different Vinit. (c) Q as a function of Vinit. The inset shows ID - VTG for round sweep measurement. (d) Dit as a function of VTG estimated from time domain CP. For comparison, Dit calculated from S.S. is also plotted. 5 FIG 4K. Taniguchi, et al.101010111012101310141015-3.5-3.0-2.5-2.0Dit / cm-2eV-1VTG / V00.20.40.60.811.21.410-510-310-1101103Q / pCtk / s(c)VTH= -2.2 V(b)(a)VTGVinitDitVfin(i)(ii)(iii)Vinit'-5 V-4.5 V-4 V-3.5 V-3 V-2.5 V@Vinittk'(d)Ditfrom S.S.Ditfromtime domain CP0100200300400500600-4-2024Q / fCVinit / V00.20.40.60.81-4-2024Drain current / AVTG / V the measurement condition for Dit. Figure 4(c) shows Q as a function of Vinit. Two different regions are evident, and the transition voltage is close to VTH. Since the electron capture has already been finished before VTH, Q observed at VTG > VTH might result from the same origin as the hysteresis in the I-V, as shown in the inset of Figure 4(c). Therefore, the hatched region is examined herein. Figure 4(d) shows Dit as a function of VTG estimated from the time domain CP. The Dit values are roughly constant in the range of 1012 - 1013 cm-2 eV-1. Although the Dit peak related to S vacancy is often reported for multilayer MoS2,3,5 it has not been observed in the present monolayer MoS2, which is consistent with the Dit calculated from S.S. It should be noted that the current peak for electron capture in Figure 3(a) is not related with the S vacancy because Vpeak shifts with ramping speed in Figure 3(c) and the energy level for the S vacancy is much deeper than VTH.28-30 Since the inherently high MoS2 channel resistance prevents the extraction of Dit from the frequency dispersion in C-V analysis on the MoS2 FET structure, the present time domain CP provides an alternative method. In summary, time domain CP analysis on monolayer MoS2 FET elucidated the elementary process for capture and emission of electrons. The electrons induced during the top gate voltage ramping are suddenly trapped to the interface states, resulting in the peak behavior. This occurs before the channel is completely formed. On the other hand, the emission process is much slower than the electron capture process, resulting in the broad tail behavior. Based on the differences between c and e, Dit is evaluated to be in the range of 1012 - 1013 cm-2 eV-1, which is comparable with that estimated from S.S. The present time domain CP can be used as an alternative method for Dit estimation. Supplementary material See the supplementary material for the time domain charge pumping data for Si-MOSFET. Acknowledgements This research was supported by the JSPS A3 Foresight program, Core-to-Core Program, A. Advanced Research Networks, JSPS KAKENHI Grant Numbers JP25107004, JP16H04343, JP16K14446, and JP26886003, Japan. References 1. S. B. Desai, S. R. Madhvapathy, A. B. Sachid, J. P. Llinas, Q. Wang, G. H. Ahn, G. Pitner, M. J. Kim, J. Bokor, C. Hu, H. -S. P. Wong, and A. Javey, Science 354, 99 (2016). 2. X. Chen, Z. Wu, S. Xu, L. Wang, R. Huang, Y. Han, W. Ye, W. Xiong, T. Han, G. Long, Y. Wang, Y. He, Y. Cai, P. Sheng, and N. Wang, Nature commun. 6, 6088 (2015). 3. M. Takenaka, Y. Ozawa, J. Han, and S, Takagi, IEDM Tech. Dig. pp. 139-142 (2016). 4. S. Park, S. Y. Kim, Y. Choi, M. Kim, H. Shin, J. Kim, and W. Choi, ACS appl. mater. interfaces 8, 11189 (2016). 5. P. Xia, X. Feng, R. J. Ng, S. Wang, D. Chi, C. Li, Z. He, X. Liu, and K. -W. Ang, Sci. Rep. 7, 40669 (2017) 6. N. Fang, and K. Nagashio, ACS appl. mater. interfaces (2018), in press. 7. W. Zhu, T. Low, Y. -H. Lee, H. Wang, D. B. Farmer, J. Kong, F. Xia, and P. Avouris, Nature commun. 5, 3087 (2014). 8. H. Bae, C. -K. Kim, and Y. -K. Choi, AIP adv. 7, 075304 (2017). 9. N. Fang, and K. Nagashio, J. Phys. D 51, 065110 (2018). 10. T. Mori, T. Kubo, N. Uchida, E. Watanabe, D. Tsuya, S. Moriyama, M. Tanaka, and A. Ando, IEEE Nanotechnol. 15, 651 (2016). 11. J. S. Brugler, and P. G. A. Jespers, IEEE electron devices ED-16, 297 (1969). 12. G. Groeseneken, H. E. Maes, N. Beltran, and R. F. D. Keersmaecker, IEEE electron devices ED-31, 42 (1984). 13. T. Tsuchiya, and P. M. Lenahan, Jpn. J. Appl. Phys. 56, 031301 (2017). 14. L. Lin, Z. Ji, J. F. Zhang, W. D. Zhang, B. Kaczer, S. D. Gendt, and G. Groeseneken, IEEE electron devices 58, 1490 (2011). 15. M. Hori, T. Watanabe, T. Tsuchiya, and Y. Ono, Appl. Phys. Lett. 105, 261602 (2014). 16. T. Watanabe, M. Hori, T. Tsuchiya, A. Fujiwara, and Y. Ono, Jpn. J. Appl. Phys. 56, 011303 (2017). 17. S. Das, H. -Y. Chen, A. V. Penumatcha, and J. Appenzeller, Nano lett. 13, 100 (2013). 18. P. Bampoulis, R. van Bremen, Q. Yao, B. Poelsema, H. J. W. Zandviliet, and K. Sotthewes, ACS appl. mater. interfaces 9, 19278 (2017). 19. K. A. Jenkins, J. Y. -C. Sun, and J. Gautier, IEEE Electron Devices 44, 1923 (1997). 20. C. D. Young, Y. Zhao, M. Pendley, B. H. Lee, K. Matthews, J. H. Sim, R. Choi, G. A. Brown, R. W. Muroto, and G. Bersuker, Jpn. J. Appl. Phys. 44, 2437 (2005). 21. Y. G. Lee, C. G. Kang, U. J. Jung, J. J. Kim, H. J. Hwang, H. -J. Chung, S. Seo, R. Choi, and B. H. Lee, Appl. Phys. Lett. 98, 183508 (2011). 22. E. A. Carrion, A. Y. Serov, S. Islam, A. Behnam, A. Malik, F. Xiong, M. Bianchi, R. Sordan, and E. Pop, IEEE Electron Devices 61, 1583 (2014). 23. H. Ramamoorthy, R. Somphonsane, J. Radice, G. He, C. -P. Kwan, and J. P. Bird, Nano Lett. 16, 399 (2016). 24. X. Li, L. Yang, M. Si, S. Li, M. Huang, P. Ye, and Y. Wu, Adv. Mater. 27, 1547 (2015). 25. J. Park, J. H. Hur, and S. Jeon, Nanotechnol. 29, 175704 (2018). 26. N. Takahashi, and K. Nagashio, Appl. Phys. Express 9, 125101 (2016). 27. S. Kurabayashi, and K. Nagashio, Nanoscale 9, 13264 (2017). 28. H. Qiu, T. Xu, Z. Wang, W. Ren, H. Nan, Z. Ni, Q. Chen, S. Yuan, F. Miao, F. Song, G. Long, Y. Shi, L. Sun, J. Wang, and X. Wang, Nature Commun. 4, 2642 (2013). 29. J. Hong, Z. Hu, M. Probert, K. Li, D. Lv, X. Yang, L. Gu, N. Mao, Q. Feng, L. Xie, J. Zhang, D. Wu, Z. Zhang, C. Jin, W. Ji, X. Zhang, J. Yuan, and Z. Zhang, Nature Commun. 6, 6293 (2015). 30. D. Liu, Y. Guo, L. Fang, and J. Robertson, Appl. Phys. Lett. 103, 183113 (2013). 6 Supplementary information Direct observation of electron capture & emission processes by the time domain charge pumping measurement of MoS2 FET Koki Taniguchi,1 Nan Fang1 and Kosuke Nagashio1, 1Department of Materials Engineering, The University of Tokyo, Tokyo 113-8656, Japan E-mail: [email protected] Corresponding author, [email protected] Figure S1: Schematic of Si-N-MOSFET and the measurement set up for conventional CP. There are four processes in the one pulse of VG between the accumulation and the inversion. (i) electrons flow-in, where the electrons are captured in interface states. It should be noted that electrons are also captured during the "high state" of the gate pulse (constant voltage region), depending on the time constant of the traps. (ii) electrons flow-out, where some of electrons captured in the interface states cannot flow out. Therefore, the number of electrons which flow out to S/D is reduced. (iii) holes flow-in, where the recombination of captured electrons and free holes takes place. Again, free holes also capture the electrons during the "low state" of the gate pulse. (iv) holes flow-out, where the number of holes which flow out to the substrate is reduced. In this single CP process, the current flows from the substrate to the S/D through the interface 1 FIG S1K. Taniguchi, et al.EcEvInterface states(i) Electrons flow-in(ii) Electrons flow-out(iii) Holesflow-in(iv) Holesflow-outinv.acc.recombinationFrom S/DtoS/DFrom sub.tosub.VGMoS2FET n+SiO2TGp-SiAPulse stringIsubn+ states. Experimentally, by applying the VG pulse string using the pulse generator, this substrate current (CP current) is monitored by the DC ammeter. Therefore, the conventional CP does not require the fast acquisition set up. However, this conventional CP is not applicable to MoS2 FET, because source/drain electrodes are connected only to the conduction band due to the Fermi level pinning. At this moment, it is difficult to connect to the valence band. Therefore, the recombination current cannot be measured. Instead of that, the time domain CP measurement is possible only for the (i) and (ii) processes, as shown by the dashed rectangular because the elemental process of electron flow-in (capture) and electron flow-out (emission) can be extracted by fast measurement set up. Figure S2: (a) Schematic of Si N-MOSFET and measurement set up. Optical micrograph of Si N- MOSFET. All the measurement was carried out at the room temperature. (b) ID-VD transfer curve at VD = 0.1 V. The threshold voltage is VTH = 0.035 V. (c) C-V curves at different frequency. The flat band voltage is VFB = -1.4 V. (d) Time domain CP measurement for Si N-MOSFET. Top: Top gate voltage pulse as a function of time. Bottom: Electron current detected through the source/drain (shown by red curve) and hole current detected through the substrate (shown by blue curve). (e) Comparison of electron current for capture and emission, where electron capture current is "inverted" to compare two data. (f) QCP,hole as a function of VH. QCP,hole is calculated from the difference in the time-integral of IS/D (the area of (e)) for capture and emission. 2 -2-1012050100150200250300350400Current / ATime / sFIG S1K. Taniguchi, et al.0123456-2-1012Qcp, hole / pCVH / V010203040-3-2-10123Capacitance / pFTop gate voltage / V10-1210-1110-1010-910-810-710-610-510-4-3-2-1012345Drain current / ATop gate voltage / V(b)(c)N-MOSFET1 mmn+SiO2TGp-SiAPulse voltageIsubAIS/Dn+(a)0.00.51.01.52.02.5050100150IS/D / ATime / s(e)Conventional CPTime domain CP(d)1 kHz1 MHzVTH= 0.035 VVFB= -1.4 V-2-1012Gate voltage / VelectronholecaptureemissionElectron(f)VL= -2 V: fixedVL= -2 VVH= 2 Ve-emissione-captureh-captureh-emissionVTH
1812.00095	1	1812	2018-11-30T23:35:40	Nonlinear Wave Propagation in 3D-Printed Graded Lattices of Hollow Elliptical Cylinders	[ "physics.app-ph" ]	We propose a 3D-printed graded lattice made of hollow elliptical cylinders (HECs) as a new way to design impact mitigation systems. We observe asymmetric dynamics in the graded HEC chains with increasing and decreasing stiffness. Specifically, the increasing stiffness chain shows an acceleration of the propagating waves, while the decreasing stiffness chain shows the opposite. From the standpoint of impact mitigation, the decreasing stiffness chain combined with the strain-softening behavior of HECs results in an order-of-magnitude improvement in force attenuation compared to the increasing stiffness chain. We extend this finding to the graded 2D arrays and demonstrate a similar trend of wave transmission efficiency contrast between the increasing and decreasing stiffness lattices. The 3D-printed HEC lattices shown in this study can lead to the development of a new type of impact mitigating and shock absorbing structures.	physics.app-ph	physics	Nonlinear Wave Propagation in 3D-Printed Graded Lattices of Hollow Elliptical Cylinders Hyunryung Kima, Eunho Kimb, Jinkyu Yanga,∗ aAeronautics and Astronautics, University of Washington, Seattle, WA, USA, bDivision of Mechanical System Engineering & Automotive Hi-Technology Research Center, Chonbuk National University, 567 Baekje-daero, Deokjin-gu, Jeonju-si, Jeollabuk-do 54896, Republic of Korea 98195-2400 Abstract We propose a 3D-printed graded lattice made of hollow elliptical cylinders (HECs) as a new way to design impact mitigation systems. We observe asym- metric dynamics in the graded HEC chains with increasing and decreasing stiffness. Specifically, the increasing stiffness chain shows an acceleration of the propagating waves, while the decreasing stiffness chain shows the op- posite. From the standpoint of impact mitigation, the decreasing stiffness chain combined with the strain-softening behavior of HECs results in an order-of-magnitude improvement in force attenuation compared to the in- creasing stiffness chain. We extend this finding to the graded 2D arrays and demonstrate a similar trend of wave transmission efficiency contrast between the increasing and decreasing stiffness lattices. The 3D-printed HEC lattices shown in this study can lead to the development of a new type of impact mitigating and shock absorbing structures. Keywords: Nonlinear, 3D printing, Stress wave, Tunability, Impact mitigation 1. Introduction Tapered or graded granular chains have been of great interest to re- searchers for their exceptional impact absorbing ability (Sen et al., 2008; ∗Author to which all correspondence should be addressed: [email protected] Preprint submitted to Journal of the Mechanics and Physics of Solids July 1, 2021 Rosas and Lindenberg, 2018). First proposed by Sen et al. (2001), long ta- pered granular chains (total number of granular particles, or chain length, N = 100) have been shown to reduce their leading pulse's kinetic energy up to 90% compared to monodispersed chains in numerical analysis. Later on, analytic models have been developed (Wu, 2002; Doney and Sen, 2005; Harbola, Rosas, Esposito and Lindenberg, 2009; Machado et al., 2013) to de- scribe the physics of the remarkable energy attenuation. Numerical (Sokolow et al., 2005) and experimental demonstration (Nakagawa et al., 2003; Melo et al., 2006) have also been reported to claim the feasibility in shorter chains (N ≈ 20). Doney and Sen (2006) introduced a superb shock absorbing system, so- called decorated tapered chain (DTC), which mitigates the impact energy significantly in a very short chain length. Supporting experimental results (Doney et al., 2009) and analytic interpretations (Harbola, Rosas, Romero, Esposito and Lindenberg, 2009) have been reported. The challenge with the DTC is that DTC itself is not feasible to be assembled in higher dimensions due to their geometrical constraints. Machado et al. (2014) found an alter- native to DTC by surrounding the regular tapered chain with smaller DTCs. Impulse wave propagation is more attenuated in this quasi-one-dimensional (1D) tapered chain than the regular tapered chain. Ideas of stacking ta- pered granular chains into two-dimensional (2D) (Tiwari et al., 2016) and three-dimensional (3D) (Sen et al., 2017) space have been proposed recently. However, these studies on tapered granular structures in higher dimensions are limited to numerical investigation, and their experimental demonstration is yet elusive, mainly due to the challenges in their assembly. 3D-printing techniques have been recently gaining their popularity in var- ious fields as a versatile fabrication tool to create complicated shapes. There have been a number of studies to design and 3D-print shock-absorbing struc- tures. Tsouknidas et al. (2016) have evaluated impact absorption of 3D- printed porous polylactic acid (PLA) structures. Bates et al. (2016) have investigated energy absorption of 3D-printed honeycomb structures made of thermoplastic polyurethane (TPU). Recently, Chen et al. (2018) have ana- lyzed crushing behavior of graded lattice cylinders, 3D-printed using Acry- lonitrile Butadiene Styrene (ABS) plastic, subject to axial impact. In this study, we evaluate nonlinear wave dynamics in 3D-printed, graded lattices composed of hollow elliptical cylinders (HECs). The gradient in the geometry of the HECs along the chain results in the asymmetric dynamics of stress wave propagation. That is, the impact mitigating behavior in the 2 positive gradient chain and that in the negative gradient chain are highly distinctive. The nonreciprocal wave propagation in nonlinear systems has been studied extensively, mainly relying on the multi-stability of their unit cells. For example, Nadkarni et al. (2016) have found a mechanical diode using bistable lattices. Fang et al. (2017) have used multi-stable stacked- origami to realize a static mechanical diode effect. Wu et al. (2018); Wu and Wang (2017) have recently reported unidirectional transmission of energy in metastable structures. However, the asymmetric dynamics in a nonlin- ear mechanical system without relying on multi-stability has been relatively unexplored. Here, we numerically and experimentally verify the impact absorbing mechanism of graded HEC structures in 1D as well as 2D architectures. In particular, we leverage the strain-softening nature of the HECs to demon- strate the asymmetric wave dynamics. The stiffness and the mass gradient combined with the strain-softening nonlinearity create the unique dynamics. These HEC lattices can be easily fabricated using 3D-printing and assembled by simple bonding, in sharp contrast to granular crystals which need delicate mechanical contact. This provides us the following advantages: First, we can avoid the issues regarding the energy dissipation due to the rotational and shear friction of granular beads in experiment. Second, we have the flexibility to design and the convenience to fabricate the unit cells. These advantages enable us to customize the unit cells to show various behaviors and investigate their effect on wave dynamics. Finally, we can achieve a low weight-to-volume ratio structure by using generic polymer materials as well as by optimally designing the structures. The rest of the manuscript is arranged in the following manner. In Sec- tion 2, we provide details on the technical approach. Section 3.1 describes the asymmetric nonlinear dynamics in the HEC chain with a positive and a negative gradient direction. We experimentally demonstrate the simulated dynamics in the graded HEC chains. In Section 3.2, we conduct rigorous comparison of the impact mitigation performance between the two opposite gradient chains. Section 3.3 suggests a potential extension of the 1D graded HEC chains to the 2D graded HEC array. We finish our manuscript with conclusion in Section 4. 3 2. Methods We first perform 1D experiment as follows: we 3D-print 26 HECs using PLA material with their wall thicknesses linearly varying from 0.4 mm to 3 mm. See Appendix A for the detailed geometry and force-displacement relationship of the HECs under the variation of their wall thicknesses. We align these HECs into two different chain configurations: (i) a chain with the increasing HEC thicknesses (called a thickening chain for the rest of this article) and (ii) a chain with the decreasing HEC thicknesses (i.e., thinning chain). Figure 1(a) shows an experimental setup of the thinning chain. Here, the HECs are printed with two small holes (diameter of 2.50 mm) along the major axis, such that they are allowed to slide along the horizontal direction by using a pair of stainless steel rods (diameter of 2.38 mm) that penetrate through all HECs (inset of Fig. 1(a)). These rods are held by 3D-printed supports to prevent them from bending. To ensure their mechanical contact, the HECs are bonded together using glue (Loctite 431). For the application of mechanical impact, we send an impulse signal to a shaker (LDS V406, B&K) such that it launches a rectangular striker mass towards the one end of the HEC chain at a desired velocity. The other end of the chain is bounded by a piezoelectric sensor (208C02, PCB Piezotronics), which is mounted on the rigid chassis and is used to measure the transmitted forces through the HEC chain. The striker mass (ms = 4.3 g) is 10 times larger than the mass (m) of the thinnest cylinder. When the striker hits the chain, it triggers the high speed camera (Phantom v1211) to start recording. The high speed camera moves along the linear stage, capturing three cylinders at a time at 40000 fps. The measurements are made at every other cylinder location, and we iterate the experiment five times for statistical treatments. To facilitate the tracking of the HECs' translational motions, we print bosses on the top and bottom of each HEC (see the inset of Fig. 1(a) or Fig. A.1(a) in Appendix A). We use the digital image correlation technique to measure the displacements of HECs (Kim et al., 2018). The data collected are stitched together, and the aligned data after reconstruction in the space domain are presented in the results hereafter. To compare with the experimental results, we perform numerical analysis using commercial finite element analysis (FEA) software (ABAQUS). We ap- ply the Timoshenko beam model (B22 element) to capture the behavior of the HECs. To take into account the hyperelastic behavior of PLA, we implement the Neo-Hookean model. We find its shear modulus and material damping 4 Figure 1: (a) A schematic diagram of the 1D HEC setup for dynamic testing. Shaker jig is attached to the shaker head (not shown but indicated by the red arrow) to slide the striker. The graded HEC chain is impacted by the striker on the left end at the input velocity, vs = 2.95± 0.09 m/s. The right end is in contact with a piezoelectric sensor. The inset is a close-up view of the single HEC inside the dotted box. (b) A schematic diagram of experiment setup for the 2D graded HEC array. factor by fitting the experimental data based on uniaxial compression tests. We also compose a 2D graded HEC array as shown in Fig. 1(b). We print 11 rows of HECs whose thicknesses vary linearly from 0.4 mm to 3 mm along the horizontal direction. Then we carefully align them and glue between the rows into a 6 × 11 array. Similar to the 1D setup, the HECs in the middle row are guided by a pair of stainless steel shafts. We cut two 5-mm-thick aluminum plates to a size of 200 mm by 12 mm with two small holes (diameter of 2.5 mm) and glue them to the left and the right end of the graded HEC array. These plates and HEC arrays are equivalent to the facesheet and core of sandwich structures, respectively, in 2D representation. The right plate is in contact with the piezoelectric sensor to record the transmitted forces through this 2D HEC array. To prevent any rotational dynamics of the array, we firmly hold the horizontal guiding rods by using a 3D-printed support. The way we apply striker input is the same as that in the 1D chain experiment. 5 AluminumplateForce sensorStriker(a)(b)2D graded HEC arrayShakerjigHigh speed camera1D graded HEC chainxy3. Results and Discussion 3.1 Asymmetric impact response Figure 2 shows experimental (top row) and numerical (middle) results of HECs' velocity profiles for thinning (left) and thickening (right) chains. It is notable that the thinning and the thickening chain show completely different dynamic behaviors. The first feature to notice is the accelera- tion/deceleration of the propagating waves depending on the gradient di- rection. This phenomenon is closely related to the mass and stiffness of the unit cells composing the chain. In case of the thinning chain, the contact stiffness decreases towards the end of the chain. Lower stiffness results in the decrease in wave speed (Fig. 2(a) and (c)). Furthermore, the negative mass gradient will amplify the propagating wave amplitude, slowing down the wave due to the effective strain-softening nature of the HECs. In the same manner, acceleration occurs in the opposite direction (Fig. 2(b) and (d)). The experimental data (Fig. 2(a) and (b)) corroborate the FEA results (Fig. 2(c) and (d)) for both thinning and thickening chains. The accelerating and decelerating waves in graded chains have been also explored in previous studies (Sen et al., 2001; Melo et al., 2006; Doney et al., 2009; Harbola, Rosas, Esposito and Lindenberg, 2009; Machado et al., 2013; Chaunsali et al., 2018). However, our system is different from those in pre- vious studies in that (1) the unit cells are bonded together supporting both compressive and tensile motions, and (2) they are deformed modestly in a weakly nonlinear regime. As a result, we observe unique wave dynamics in the HEC system, such as the formation of weakly nonlinear pulses followed by oscillatory tails due to tensile motions (see the ripples in Fig. 2). Another feature to notice is the amplitude change. The amplitude sig- nificantly decreases as the wave approaches the end of the thickening chain (Fig. 2(b) and (d)). However, in the thinning case, the amplitude increases minutely as the wave propagates along the chain (Fig. 2(a) and (c)). This amplitude change is obvious if we look at the individual wave profiles in Fig. 2(e) and (f). This attenuation/amplification of the propagating wave can be also explained by the change in mass and stiffness within the graded chain. In the thickening chain, the mass of the HEC increases, thereby decreasing the particle velocity to satisfy the conservation of momentum principle. More- over, the increasing stiffness results in a faster group velocity, which extends the width of the leading wave and contributes to the attenuation of the wave amplitude. On the contrary, in the thinning chain, the HEC masses become 6 lighter towards the end, resulting in higher particle velocities. Contact stiff- ness also decreases in this case, which contributes to narrowing width and thus amplifying the wave. More details on the effect of the contact stiffness and the mass are described in Appendix B. The amplitude change also is affected by the energy transfer from kinetic energy to strain energy. See Appendix C for the energy analysis. The asymmetric wave propagation in opposite directions is in reminis- cence of the study by Wu et al. (2018); Wu and Wang (2017). They take ad- vantage of the asymmetry of their system to realize non-reciprocal wave prop- agation by selectively triggering nonlinear instability in a single direction. The transmission induced by overcoming the threshold with high amplitude input is called supra-transmission. Similarly, for the thinning chain in our system, the striker impact contains a wide range of frequencies (Fig. D.1(a) in Appendix D) and the output signal contains most of the input frequency components, despite the significant transfer to the low frequency modes. The thickening chain does not transfer much energy to the end of the chain, as seen in Fig. D.1(b) in Appendix D. As a result, we confirm a similar trend of asymmetric supra-transmission in our graded HEC chain, by leveraging not the bistability of unit cells, but their arrangements in gradient. More details on the spectral analysis can be found in Appendix D. 3.2 Impact mitigation We evaluate the shock absorption performance of the graded HEC chains by investigating their transmitted force profiles (Fig. 3). In the thinning chain, the initially localized force peak disperses slowly as the wave prop- agates along the chain (Fig. 3(a)). In the thickening chain, the initially dispersed force profile tends to become highly localized towards the end of the chain (Fig. 3(b)). This difference is evident if we compare the force pro- files measured at the end of the chain by using a force sensor (Fig. 3(c)). Here, we also include experimental results for comparison (circular markers). We observe a solitary wave-like localized pulse in the thickening chain (see the dashed red curve based on the FEM result, which show no oscillatory tails). On the other hand, the thinning chain exhibits a small-amplitude leading pulse followed by oscillations. While there exist noticeable discrep- ancies between the experimental and computational results possibly due to friction and material damping, we still witness the qualitative difference in the empirical force packets between the thinning and thickening chains. 7 Figure 2: Surface map of HECs' velocity profiles in space and time domain for (a) the thinning and (b) the thickening chain, based on the experimental data. The FEA results for the corresponding configurations are shown in (c) and (d). The cross sections along the dotted lines in (c) and (d) are plotted in (e) and (f) in dashed curves, respectively. The solid curves represent the corresponding velocity profiles from the experimental data. 8 510152025Particle Number0246810Time (ms)510152025Particle Number0246810Time (ms)510152025Particle Number0246810Time (ms)510152025Particle Number0246810Time (ms)ExperimentFEM2.5-0.525Time (ms)10Particle Number1.5152001250.5Velocity (m/s)11.52.52Time (ms)051.510Particle Number15201Velocity (m/s)1252-2-1012Velocity (m/s)ExperimentExperimentFEMFEM(a)(b)(c)(d)(e)(f)Thick → ThinThin → ThickFigure 3: Contact force between HECs in (a) the thinning chain and (b) the thickening chain. The contact force is calculated in ABAQUS simulations. 0 index means the striker. Negative force means tensile direction. The dashed lines at n = 26 indicate the output force profiles which are plotted in (c). (c) Contact force output at the right boundary. The purple and red dashed curves show the output force in the thinning and the thickening HEC chain from FEA results, respectively. We run 10 experiments and plot the average values in circular markers and the standard deviations in shaded areas (yellow for the thinning and green for the thickening HEC chain). The striker velocity vs = 3.00 ± 0.07 m/s). 9 0510152025Particle Number0246810Time (ms)-1001020304050Force (N)0510152025Particle Number0246810Time (ms)051015Force (N)Force (N)Time(ms)010246802051015Thinning - FEMThinning - ExperimentThickening - FEMThickening - Experiment(a)(b)(c)Thick → ThinThin → ThickFigure 4: Kinetic energy (blue curve) and strain energy (orange curve) of the entire model in (a) the thinning chain and (b) the thickening chain based on numerical simulations without damping. These force profiles may look contradictory to the velocity profiles pre- sented in Section 3.1. In the velocity profiles, the thinning chain showed the focusing behavior of wave packets (left column in Fig. 2), while the thick- ening chain exhibited the attenuation of those packets (right column in Fig. 2). This seemingly contradictory behavior between velocity and force profiles can be explained by the energy analysis. We calculate the evolution of kinetic and strain energy stored in the HEC system over the span of 5 ms, which covers a travel of the initial impact to the end of the chain (Fig. 4). In the thinning chain, we observe the kinetic energy and the strain energy reach an equilibrium point where their levels are comparable (Fig. 4(a)). In the thick- ening chain, however, most kinetic energy is converted into the strain energy. In other words, in the thinning chain, the impulse facilitates the rattling of the HECs after it passes by the chain (thus causing the oscillatory tails as witnessed in Fig. 3(c)), while the thickening chain rather piles up its energy into the strain energy instead of shedding it into dynamic energy. This is the key difference in energy transfer between the thinning and thickening chains. Since the velocity profiles are directly related to the kinetic energy, the trend looked contradictory to the force profiles that are associated with the strain energy. Further details on energy analysis can be found in Appendix C. 3.3 Two-dimensional expansion Now we assess the feasibility of extending the 1D HEC system into a 2D one, such that it can be potentially used as a core material for impact 10 012345Time (ms)00.0020.0040.0060.0080.01Energy (J)012345Time (ms)00.0020.0040.0060.0080.01Energy (J)TotalKineticEnergyTotal Strain Energy(a)(b)Thick → ThinThin → ThickFigure 5: Contact forces at the output boundary are plotted for (a) the 2D thinning array and (b) the 2D thickening array. The array size is 6 × 11. We run 10 experiments and plot the average in the yellow solid line and the standard deviations in the yellow shaded areas (vs = 2.93 ± 0.05 m/s). mitigating structures. To this end, we build a sandwich structure as shown in Fig. 1(b). Here, the arrangement is in a pseudo-2D manner, graded in the x direction and homogeneous in the y direction. Given the shallow thickness of the layer (12 mm), we impose the plane stress condition in the out-of-plane direction. We measure the transmitted force profiles at the right end of the prototype and plot them in yellow curves for the thinning (Fig. 5(a)) and the thickening (Fig. 5(b)) array. The corresponding numerical data are also plotted in red curves in Fig. 5. Consistent with the 1D chain result, the maximum force at the wall in the thinning array is lower than that of the thickening chain. However, the performance of the thinning chain is better than the thickening chain only by twice, which is smaller than the 7-fold improvement observed in the 1D system. This is likely due to the transverse dispersion of energy, as well as the short chain length in 2D. The dispersive effect can be clearly seen in numerical simulations (Fig. 6), where the thickening lattice immediately radiates energy in the transverse di- rection (y-direction) as the striker hits the plate, while the thinning lattice does not. Despite this transverse dispersion effect, we still verify the asym- metric wave propagation in the 2D graded HEC array. Particularly, the amplifying velocity towards the end of the thinning array is consistent with the dynamics we observe in the 1D thinning chain (Fig. 2(a)). Likewise, the localization of the velocity at the input boundary of the thickening array is 11 00.511.522.53Time (ms)05101520Force (N)00.511.522.53Time (ms)01020304050Force (N)FEAExperiment(a)(b)Thick → ThinThin → ThickFigure 6: Velocity contour of the 2D model from ABAQUS for (a) thinning HEC and (b) thickening HEC array at t = 0.4 ms. in agreement with the 1D thickening chain (Fig. 2(b)). To investigate the chain size effect, we model a larger array, 120 × 11, as shown in Fig. 7(a) and examine how the impact absorbing level changes. A striker with the mass of 30 g is applied at an initial velocity of 3 m/s. We observe an order-of-magnitude difference in the maximum output force between the thinning and the thickening arrays (Fig. 7(b)). The oscillating envelope of the force profile is due to the vibrations of the HECs. In fact, the dominant frequency component of the wave profile is 1.2 kHz, which is around the 1st vibration mode of the thickest HECs and the 2nd vibration mode of the thinnest HECs. This numerical result hints that the proposed 2D HEC lattices can be potentially used as a core material in sandwich structures for incurring asymmetric wave propagation and thus efficient impact mitigation. 4. Conclusion We investigated nonlinear dynamic behavior of the graded lattice made of hollow elliptical cylinders (HECs). We experimentally and numerically demonstrated the asymmetric wave dynamics under the condition of external impact applied in opposite directions. This includes acceleration/deceleration and attenuation/amplification of the mechanical wave in terms of velocities, depending on whether the system has a positive or negative gradient of mass and stiffness imposed on the HEC chain. We also investigated the force transmission through the one-dimensional chain, whose trend is seemingly 12 (a)(b)Thick → ThinThin → ThickFigure 7: (a) The schematic of the 120×11 HEC model. The thickness gradient is assigned such that it varies from 3 mm to 0.4 mm along the impact direction. The 2D HEC core is bonded and sandwiched between aluminum plates. The plate thickness is determined such that its mass is the same as the thickest HEC column's mass. (b) FEA analysis on contact force at the right end for 2D thinning array (turquoise line) and 2D thickening array (purple line). The inset shows the thinning chain response in a magnified view. Damping is not considered in these simulations. flipped with respect to the velocity profiles. We explained this contradictory behavior by energy analysis. We extended our findings from the 1D system to the 2D array, and demonstrated the efficacy of this asymmetric wave dy- namics for impact mitigation purposes. This expansion in 2D space sheds light on a new way to design sandwich core structures for controllable stress wave management. Assigning stiffness gradients in both directions of the 2D array is an on-going study and will potentially lead to the manipulation of wave directionality. Acknowledgments J.Y. and H.K. thank the support of the National Science Foundation under Grant No. CAREER-1553202. E.K acknowledges the support from the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) under Grant No.2017R1C1B5018136 13 051015202530Time (ms)-40-20020406080Contact Force (N)0102030-4048ThickeningThinning...StrikerAl plate120(b)(a)////////////////////////////////Appendix A. Force-displacement relationship of HECs In this study, we achieve the variations of the HECs' stiffness by modifying their wall thicknesses. To assess this thickness effect, we numerically calculate the force-displacement relationship between adjacent cylinders in the graded HEC chain using FEA (Fig. A.1(a) for the geometrical configuration of the gradient chain composed of HECs). We see from Fig. A.1(b) that the HECs show effective strain-softening behavior, i.e., decreasing stiffness (the slope of the force-displacement curve) in compression. This holds true regardless of the thickness of the HECs, as seen from the force-displacement curve of thickest pair (the orange curve in Fig. A.1(b)) to the thinnest pair of HECs (the yellow curve in the inset of Fig. A.1(b)). In fact, the force-displacement curve of the HEC nearly scales up as its thickness increases. As a result, the graded HEC chain exhibits a gradual change of force-displacement re- lationship, leading to a universal power-law relationship (Kim et al., 2018). Figure A.1: (a) An illustration of the graded HEC chain and its dimensions. The chain is composed of N = 26 HECs with linearly graded thickness from 3 mm to 0.4 mm. The out-of-plane width is 12 mm. (b) The force-displacement curve between adjacent HECs in the graded chain. The compressive deformation is plotted positive. The force- displacement relationship of cylinder 1 and 2 is plotted in the orange curve. Likewise, the force-displacement relationship between cylinder 2 and 3, 3 and 4, and so on (refer to (a) for cylinder numbers) are plotted respectively from the top to the bottom. The inset shows an enlarged view of the force-displacement curves for the last few cylinder pairs. 14 -20246Displacement (mm)-400-20002004006008001000Force (N)-20246-606121-22-3...24-2525-26...CompressionTension...18 mm3 mm 0.4 mm21252632430 mm(b)(a)Appendix B. Amplitude-dependent wave speed In this section, we qualitatively explain why the amplitude-dependent wave speed deforms the shape of the leading wave. We assume the wave speed is proportional to the stiffness and inversely-proportional to the mass, i.e., cg ∝(cid:112)k/m where cg is the wave speed, k is the stiffness, and m is the mass. Since both the stiffness and the mass vary together in our graded HEC chain, we need to consider both to estimate the wave speed. Figure B.1 shows how deformation wave profile changes as the wave propagates through the chain. For the thinning chain, as shown in Fig. B.1(a), the leading wave stiffens. Although the later part of the chain has smaller masses, the stiffness drops much more drastically. As a result, the thinning chain has positive relation between the amplitude and the wave speed (wave peak travels faster than the wave front). On the contrary, we observe expansion of the leading wave in the thickening chain (Fig. B.1(b)). The increase in stiffness overshadows the increase in mass towards the chain end, resulting in negative relationship between the amplitude and the wave speed. Figure B.1: Snapshot of deformation profiles at two time points in (a) the thinning and (b) the thickening HEC chain. The blue curve is at time 1 (t1) and the yellow curve is at time 2 (t2) where t1 < t2. The arrows indicate wave speed of the corresponding points. The length of the arrow is relative scale of the wave speed. 15 510152025Particle Number-0.2-0.100.10.20.3Deformation (mm)510152025Particle Number00.511.52Deformation (mm)Thick → ThinThin → Thick(a)(b)t1t2Wave frontAppendix C. Energy analysis In this section, we investigate the evolution of kinetic and strain energy during the propagation of waves in thinning and thickening HEC chains. We export kinetic and strain energy data from ABAQUS simulation results and plot the contour map in both time and space domain in Fig. C.1. For the thinning chain, we see a large portion of the kinetic energy (Fig. C.1(a)) is transmitted towards the end of the chain, even showing clear reflection from the boundary. An equivalent amount of the strain energy (Fig. C.1(c)) is also observed, but it diminishes noticeably as the wave propagates through the chain. The thickening chain shows a different trend of kinetic and strain energy transmission. We see a drastic reduction of the kinetic energy in the first few HECs in the chain (Fig. C.1(b)). This kinetic energy lost is trans- ferred to the strain energy as shown by the highlighted area in Fig. C.1(d). This energy transfer phenomenon is again due to the large deformation of the thin HECs in the early part of the chain, caused by the striker impact. This corroborates the results reported in the velocity and force profiles as shown in Fig. 2 and 3. Appendix D. Spectrum interpretation We perform Fast Fourier Transform (FFT) on the velocity data to in- vestigate the frequency components in the propagating wave in the graded HEC chain, as shown in Fig. D.1. The red curves represent cutoff frequencies of the band structures, also denoted in Fig. D.1(c) and (d) for the thinning and thickening chains, respectively. From Fig. D.1(a) and (b), we observe that the power spectral density tends to follow the first and second passing bands of the dispersion curves (compare Fig. D.1(a) and (c) for thinning and Fig. D.1(b) and (d) for thickening HEC chain). This means that the propa- gating wave is mostly composed of the first and second vibrational modes of the HEC. The higher modes are not as evident as the lower ones, especially in the thickening chain, partly due to material damping of PLA. Figure D.1(e) and (f) show the spatial spectrum in the thinning and the thickening chain, respectively. The spatial spectrum in the thinning HEC chain gradually includes high wavenumbers as time passes, which is consistent with the stiffening wave front as shown in Fig. 2(e). The widely expanded wave front with barely moving wave peak in the thickening HEC chain in Fig. 2(f) is well represented in Fig. D.1(f). The wavenumber is more 16 Figure C.1: The evolution of kinetic energy (KE) and strain energy (SE) is plotted for thinning chain in (a) and (c) and for thickening chain in (b) and (d), respectively. The energy is calculated in ABAQUS simulations without damping. 0 index means the striker. localized to low region as the wave width expands while the wide spectrum still exists even when the wave reaches the end for the localized peak. 17 0510152025Particle Number0246810Time (ms)0510152025Particle Number0246810Time (ms)0510152025Particle Number0246810Time (ms)0510152025Particle Number0246810Time (ms)(a)(b)(c)(d)KEKESESEThick → ThinThin → Thick0 0.2 0.4 0.6 0.8 1Energy (x10-3J)Figure D.1: Spectral contour of the velocity data from FEA without damping is calculated using Fast Fourier Transform (FFT) in time domain for (a) the thinning and (b) the thickening chain. The red dashed boxes represent the pass bands obtained from the band structures in (c) and (d) for an infinite thinning and thickening chain, respectively. Spatial spectrum of the velocity data is also calculated using FFT in space domain for (e) the thinning and (f) the thickening chain up to 5 ms around when the wave peak reaches the end of the chain. 18 510152025Particle Number0246810Frequency (kHz)-70-60-50-40-30-20-10(dB)0.511.522.53Thickness (mm)0246810Frequency (kHz)0.511.522.53Thickness (mm)0246810Frequency (kHz)510152025Particle Number0246810Frequency (kHz)Thin → ThickThick → ThinPass bandStop band012345Time (ms)00.10.20.30.40.5Wavenumber (Particle )-1012345Time (ms)00.10.20.30.40.5Wavenumber (Particle )-1-40-35-30-25-20-15-10(dB)(a)(b)(c)(d)(e)(f)References Bates, S. R., Farrow, I. R. and Trask, R. S. (2016), '3d printed polyurethane honeycombs for repeated tailored energy absorption', Materials and Design 112, 172 -- 183. URL: http://www.sciencedirect.com/science/article/pii/S026412751631125X Chaunsali, R., Kim, E. and Yang, J. (2018), 'Demonstration of accelerating and decelerating nonlinear impulse waves in functionally graded granu- lar chains', Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences . Chen, L., Zhang, J., Du, B., Zhou, H., Liu, H., Guo, Y., Li, W. and Fang, D. (2018), 'Dynamic crushing behavior and energy absorption of graded lattice cylindrical structure under axial impact load', Thin-Walled Structures 127, 333 -- 343. URL: http://www.sciencedirect.com/science/article/pii/S0263823117309680 Doney, R. L., Agui, J. H. and Sen, S. (2009), 'Energy partitioning and im- pulse dispersion in the decorated, tapered, strongly nonlinear granular alignment: A system with many potential applications', Journal of Ap- plied Physics 106(6), 064905. URL: https://doi.org/10.1063/1.3190485 Doney, R. L. and Sen, S. (2005), 'Impulse absorption by tapered horizontal alignments of elastic spheres', Phys. Rev. E 72, 041304. URL: https://link.aps.org/doi/10.1103/PhysRevE.72.041304 Doney, R. and Sen, S. (2006), 'Decorated, tapered, and highly nonlinear granular chain', Phys. Rev. Lett. 97, 155502. URL: https://link.aps.org/doi/10.1103/PhysRevLett.97.155502 Fang, H., Wang, K. and Li, S. (2017), 'Asymmetric energy barrier and me- chanical diode effect from folding multi-stable stacked-origami', Extreme Mechanics Letters 17, 7 -- 15. URL: http://www.sciencedirect.com/science/article/pii/S2352431617301414 Harbola, U., Rosas, A., Esposito, M. and Lindenberg, K. (2009), 'Pulse propagation in tapered granular chains: An analytic study', Phys. Rev. E 80, 031303. URL: https://link.aps.org/doi/10.1103/PhysRevE.80.031303 19 Harbola, U., Rosas, A., Romero, A. H., Esposito, M. and Lindenberg, K. (2009), 'Pulse propagation in decorated granular chains: An analytical approach', Phys. Rev. E 80, 051302. URL: https://link.aps.org/doi/10.1103/PhysRevE.80.051302 Kim, H., Kim, E., Chong, C., Kevrekidis, P. G. and Yang, J. (2018), 'Demon- stration of dispersive rarefaction shocks in hollow elliptical cylinder chains', Phys. Rev. Lett. 120, 194101. URL: https://link.aps.org/doi/10.1103/PhysRevLett.120.194101 Machado, L. P., Rosas, A. and Lindenberg, K. (2013), 'Momentum and en- ergy propagation in tapered granular chains', Granular Matter 15(6), 735 -- 746. URL: https://doi.org/10.1007/s10035-013-0444-6 Machado, L. P., Rosas, A. and Lindenberg, K. 'A quasi- unidimensional granular chain to attenuate impact', The European Physi- cal Journal E 37(11), 119. URL: https://doi.org/10.1140/epje/i2014-14119-9 (2014), Melo, F., Job, S., Santibanez, F. and Tapia, F. (2006), 'Experimental ev- idence of shock mitigation in a hertzian tapered chain', Phys. Rev. E 73, 041305. URL: https://link.aps.org/doi/10.1103/PhysRevE.73.041305 Nadkarni, N., Arrieta, A. F., Chong, C., Kochmann, D. M. and Daraio, C. (2016), 'Unidirectional transition waves in bistable lattices', Phys. Rev. Lett. 116, 244501. URL: https://link.aps.org/doi/10.1103/PhysRevLett.116.244501 Nakagawa, M., Agui, J. H., Wu, D. T. and Extramiana, D. V. (2003), 'Im- pulse dispersion in a tapered granular chain', Granular Matter 4(4), 167 -- 174. URL: https://doi.org/10.1007/s10035-002-0119-1 Rosas, A. and Lindenberg, K. (2018), 'Pulse propagation in granular chains', Physics Reports 735, 1 -- 37. URL: http://www.sciencedirect.com/science/article/pii/S0370157318300231 20 Sen, S., Hong, J., Bang, J., Avalos, E. and Doney, R. (2008), 'Solitary waves in the granular chain', Physics Reports 462(2), 21 -- 66. URL: http://linkinghub.elsevier.com/retrieve/pii/S0370157308000586 Sen, S., Manciu, F. S. and Manciu, M. (2001), 'Thermalizing an impulse', Physica A: Statistical Mechanics and its Applications 299(3), 551 -- 558. URL: http://www.sciencedirect.com/science/article/pii/S0378437101003405 Sen, S., Mohan, T. R. K. and Tiwari, M. (2017), 'Impact dispersion using 2d and 3d composite granular packing', KONA Powder and Particle Journal 34, 248 -- 257. Senn, M. (2015), 'This open source code is used for dic tracking https://www.mathworks.com/matlabcentral/fileexchange/50994-digital- image-correlation-and-tracking'. Sokolow, A., Pfannes, J. M. M., Doney, R. L., Nakagawa, M., Agui, J. H. and Sen, S. (2005), 'Absorption of short duration pulses by small, scalable, tapered granular chains', Applied Physics Letters 87(25), 254104. URL: https://doi.org/10.1063/1.2149218 Tiwari, M., Krishna Mohan, T. R. and Sen, S. (2016), 'Decorated granular layers for impact decimation', Granular Matter 18(3), 45. URL: https://doi.org/10.1007/s10035-016-0652-y Tsouknidas, A., Pantazopoulos, M., Katsoulis, I., Fasnakis, D., Maropoulos, S. and Michailidis, N. (2016), 'Impact absorption capacity of 3d-printed components fabricated by fused deposition modelling', Materials and Design 102, 41 -- 44. URL: http://www.sciencedirect.com/science/article/pii/S0264127516304282 Wu, D. T. (2002), 'Conservation principles in solitary impulse propagation through granular chains', Physica A: Statistical Mechanics and its Appli- cations 315(1), 194 -- 202. Slow Dynamical Processes in Nature. URL: http://www.sciencedirect.com/science/article/pii/S0378437102012402 Wu, Z. and Wang, K.-W. (2017), 'On the wave propagation analysis and supratransmission prediction of a metastable modular metastructure for adaptive non-reciprocal energy transmission', ArXiv e-prints . 21 Wu, Z., Zheng, Y. and Wang, K. W. (2018), 'Metastable modular metastruc- tures for on-demand reconfiguration of band structures and nonreciprocal wave propagation', Phys. Rev. E 97, 022209. URL: https://link.aps.org/doi/10.1103/PhysRevE.97.022209 22
2001.02642	1	2001	2019-12-10T17:34:29	A verified equivalent-circuit model for slotwaveguide modulators	[ "physics.app-ph", "physics.optics" ]	We formulate and experimentally validate an equivalent-circuit model based on distributed elements to describe the electric and electro-optic (EO) properties of travellingwave silicon-organic hybrid (SOH) slot-waveguide modulators. The model allows to reliably predict the small-signal EO frequency response of the modulators exploiting purely electrical measurements of the frequency-dependent RF transmission characteristics. We experimentally verify the validity of our model, and we formulate design guidelines for an optimum trade-off between optical loss due to free-carrier absorption (FCA), electro-optic bandwidth, and {\pi}-voltage of SOH slot-waveguide modulators.	physics.app-ph	physics	A verified equivalent-circuit model for slot- waveguide modulators Heiner Zwickel,1 Stefan Singer,1 Clemens Kieninger,1,2 Yasar Kutuvantavida,1,2 Narek Muradyan,1 Thorsten Wahlbrink,3 Shiyoshi Yokoyama,4 Sebastian Randel,1 Wolfgang Freude,1 and Christian Koos1,2,* 1Institute of Photonics and Quantum Electronics (IPQ), Karlsruhe Institute of Technology (KIT), Germany 2Institute of Microstructure Technology (IMT), Karlsruhe Institute of Technology (KIT), Germany 3AMO GmbH, Aachen, Germany 4Institute for Materials Chemistry and Engineering, Kyushu University, Japan [email protected] Abstract: We formulate and experimentally validate an equivalent-circuit model based on distributed elements to describe the electric and electro-optic (EO) properties of travelling- wave silicon-organic hybrid (SOH) slot-waveguide modulators. The model allows to reliably predict the small-signal EO frequency response of the modulators exploiting purely electrical measurements of transmission characteristics. We experimentally verify the validity of our model, and we formulate design guidelines for an optimum trade-off between optical loss due to free-carrier absorption (FCA), electro-optic bandwidth, and π-voltage of SOH slot-waveguide modulators. frequency-dependent RF the Introduction 1. Broadband efficient electro-optic (EO) modulators are key elements for a wide variety of applications, ranging from high-speed optical communications [1 -- 6] to ultra-fast signal processing [7,8] and further to optical metrology and sensing [9,10]. Among the various technology platforms that are available today, silicon photonics (SiP) represents a particularly promising option, exploiting mature CMOS fabrication processes for cost-efficient mass production of densely integrated photonic circuits [11 -- 13]. When it comes to EO modulators, hybrid combinations of SiP slot waveguides with highly efficient organic EO cladding materials are an attractive approach to overcome the intrinsic lack of second-order optical nonlinearities in bulk silicon [14]. Exploiting theory-guided optimization of organic EO materials on a molecular level [15], these silicon-organic hybrid (SOH) devices primarily stand out due to small voltage-length products down to 0.32 Vmm [16] -- more than an order of magnitude below that of conventional depletion-type SiP modulators [17]. SOH devices have been demonstrated to offer attractive performance parameters such as EO bandwidths of 100 GHz [18] as well as line rates of up to 120 Gbit/s for intensity modulation [5,19] and up to 400 Gbit/s for coherent 16QAM signaling [6], while supporting driver-less operation with sub-1 V drive signals obtained from binary outputs of a standard CMOS FPGA [20]. However, the EO bandwidth of these devices is still subject to large uncertainties. In addition, most of the aforementioned experiments [5,6,18,19,21] relied on a so-called gate voltage that is applied across the buried oxide (BOX) of the underlying silicon-on-insulator (SOI) substrate to induce a highly conductive electron accumulation layer that reduces the RC time constant of the slot waveguide [22]. Moreover, an experimentally verified model describing the EO frequency response of SOH modulators is still lacking, which represents a major obstacle towards a targeted design of the modulator structures and of the associated drive circuitry. 1 In this paper, we formulate and experimentally verify a reliable quantitative model of slot- waveguide modulators that accounts for the various bandwidth-limiting effects. We build upon an equivalent-circuit approach proposed by Witzens et al. [23], which relies on a distributed-element description of the slot waveguide structure and of the associated radio- frequency (RF) transmission line. We refine this model and verify it quantitatively using experimental data obtained from SOH devices. We find that the model accurately predicts the EO frequency response of a slot-waveguide modulator based on purely electrical RF characteristics of the device, which are easily accessible by measurements. Our model allows to identify non-optimum doping profiles as the main reason for a limited EO bandwidth of our current device generation. Based on these findings, we extract design rules that lead to an optimum trade-off between optical loss due to free-carrier absorption (FCA), EO bandwidth, and π-voltage. Optimized device designs should lead to sub-1 mm Mach-Zehnder modulators (MZM) with π-voltages smaller than 1 V, electro-optic-electric (EOE) bandwidths larger than 100 GHz, and FCA-induced optical losses below 0.1 dB. To the best of our knowledge, this work represents the first experimental validation of a general model of slot-waveguide modulators. We believe that quantitatively reliable models represent a key step towards fast and reliable design and efficient wafer-level characterization of slot-waveguide modulators which are embedded into complex optical systems and RF drive circuits. The paper is structured as follows: In Section 2, we introduce the concept of SOH modulators and formulate a model based on distributed elements. In Section 3, we discuss S- parameter measurements of SOH modulators, extract the characteristic impedance and the RF propagation parameter, and determine the parameters of the equivalent RF circuit by a least- squares fit. In Section 4, we validate the model by comparing the measured EO frequency response to the model prediction using parameters extracted from purely electrical measurements. In Section 5, we summarize the findings and discuss design guidelines and achievable performance parameters of SOH EO modulators. 2. Device principle and model based on distributed-element equivalent circuit The concept of a slot-waveguide SOH Mach-Zehnder modulator (MZM) is depicted in Figure 1 [24 -- 26]. The MZM consists of two parallel phase shifter sections having a length  of typically 0.5 mm … 1.5 mm, see Fig. 1(a). The electric drive signal is applied via a coplanar RF transmission line in ground-signal-ground (GSG) configuration. The device is realized as a travelling-wave structure, in which the modulating electric mode co-propagates with the optical mode. Each phase shifter comprises a silicon slot waveguide with a slot width of typically 60 … 160 nm, which is filled with a highly efficient organic electro-optic (EO) material, that may offer in-device EO coefficients in excess of 300 pm/V [16], see cross- section in Fig. 1(b). To activate the macroscopic EO activity in the slot, chromophores in the organic EO material have to be aligned in a one-time poling process, see [24,27] for details. The rails of the slot waveguide are connected to the metal RF transmission line via thin, doped silicon slabs, such that the externally applied drive voltage predominantly drops across the narrow slot. At the same time, the optical mode is highly confined to the slot region, leading to a very strong interaction with the modulating electric field. The EO frequency response of SOH modulators strongly depends on the RC time constant that is associated with the slot capacitance and the slab resistance. A two-level doping profile as illustrated in Fig. 1(b) can help to increase the bandwidth while maintaining acceptable optical loss [28,29]: In the rails and near the slot, the doping concentration is kept low (light blue) to avoid excessive free-carrier absorption, while a higher doping concentration (dark blue) is used further away from the optical waveguide to increase the electric conductivity. The impact of increased slab conductivity can also be studied by applying a DC "gate" voltage Ugate between the substrate and the slabs. This induces a highly conductive electron accumulation layer at the interface between the Si slabs and the BOX [22]. 2 Fig. 1. Device concept and equivalent-circuit model. (a) Schematic top view of the Mach- Zehnder modulator (MZM) with metal electrodes (yellow) that form a coplanar ground-signal- ground electrical transmission line of length . The device is equipped with contact pads at both ends. The optical waveguide depicted in blue is split into two arms to realize the Mach- Zehnder interferometer. (b) Cross-section of the MZM along the line A-A' indicated in Subfigure (a). Each arm comprises a SiP slot waveguide filled with an organic electro-optic (EO) material. The metal electrodes are connected to the rails of the slot waveguide via thin, doped Si slabs. A modulating voltage Umod, applied to the signal electrode, drops mainly across the narrow slot. This leads to a strong electric field in the slot that changes the refractive index of the organic EO material and thus modulates the phase of the optical wave. The Si slabs are doped to increase the conductivity. A low doping concentration (light blue) is used near the slot to keep optical losses low. A higher doping concentration (dark blue) is used further away from the slot to achieve a low resistance of the slabs and to thus decrease the RC time constant of the modulator. In addition, a DC "gate" voltage Ugate can be applied between the substrate and the ground electrodes. This induces a highly conductive electron accumulation layer at the interface between the Si slabs and the buried oxide layer (BOX) and thus further decreases the RC time constant. (c) Enlarged cross-sectional view of the MZM along with the associated elements of an equivalent-circuit representation of the modulator. The capacitance of the slot is represented by a capacitor with capacitance CS/2. The finite conductivity of each slab is represented by two parallel conductances GS,bulk and GS,acc that account for the conductance of the bulk Si slab and of the accumulation layer formed by the gate voltage, respectively. The total capacitance CTL. (d) Equivalent-circuit model of an metal electrodes have a infinitesimally short element of the electrical transmission line formed by the MZM. The distributed elements shown in yellow correspond to the metal electrodes. The distributed elements shown in blue represent the additional RC shunt load formed by the slot waveguide. (e) Telegrapher-type equivalent-circuit representation of an MZM transmission-line element [23]. This representation is obtained by transforming the circuit in Subfigure (d), using SG and a frequency-dependent shunt capacitance a frequency-dependent shunt conductance S,C see Eq. (2) and Eq. (3). 3 TL,L For a quantitative description of the electric behavior of the SOH modulator, we adapt the model suggested in [23], which has also been adapted to model depletion-type pn- modulators [30 -- 36]. Reference [23] is primarily focused on the design of phase-matched travelling-wave transmission lines, which is key to achieve broadband operation of MZM with rather large interaction lengths  of several millimeters. Over the previous years, however, substantial in-device EO efficiencies [16] have allowed to significantly reduce the interaction length. This renders phase-matching less crucial and requires adapted design considerations that focus on the RC time constant of the slab/slot configuration and on the associated trade-off with respect to optical loss. improvements of organic EO materials and R  , a series inductance TL z TLC In our analysis, we rely on a distributed-element model of the underlying RF transmission line, see Fig. 1(c) for a cross-sectional view with elements of an equivalent circuit representing an infinitesimally short transmission-line section of length z . The full equivalent circuit is shown in Fig. 1(d). First, we consider the circuit elements depicted in yellow in Fig. 1(d) that describe the coplanar GSG transmission line formed by the metal electrodes only, without the optical slot-waveguide structure. These circuits elements L  , and a shunt capacitance comprise a series resistance TL z TL zC  , where TL,R are differential quantities that describe the resistance, and inductance, and capacitance of the GSG transmission line per length in /m, /m, and F/m, respectively. This corresponds to the classical telegrapher circuit with negligible shunt conductance. In a second step, we expand the equivalent circuit by the electrical representation of the Si structure (blue) forming the optical slot waveguide. For practical doping concentrations, the conductivity of the slabs is much smaller than that of the metal electrodes. As a consequence, the longitudinal currents can be assumed to be confined to the metal electrodes, whereas the currents in the silicon slabs flow predominantly in the transverse (x) direction between the electrodes and the slot. These assumptions are in accordance with findings from finite-element simulations [23] and allow us to model the electrical behavior of the optical slot waveguide by simply adding an RC shunt to the equivalent circuit [23,30,37]. The associated elements are the differential capacitance SC of of the resistive Si slabs to both sides of the pair of slots, the differential conductance S,bulkG each of the slot waveguides, and the differential conductance , which accounts for the increased conductance of an accumulation layer that may be induced by applying a gate voltage Ugate. Note that the capacitance between center signal electrode and , and that the overall shunt capacitance of the each of the ground electrodes amounts to metal electrodes forming the coplanar GSG transmission line is hence  . z  represents the Similarly, zC S combination of both slots. The quantities  refer to the z G  slab conductance to either side of each slot, and the equivalent circuit-element in Fig. 1(d) represents the combined effect of four slabs. The total differential conductance is SY  G G  S,bulk given by is the capacitance of an individual slot, and  and G z The two slot waveguides thus create a differential shunt admittance  g U S,acc TL 2C  S,acc. G  S 2C C TL C  TL G  S,bulk G S,bulk C S S,acc G  S,acc gate S,acc  S  1  Y S  G  S,bulk 1   G  S,acc 1 j C  S R F . (1) For completeness, we have also indicated additional elements by dashed lines in Fig. 1(c) that seem to be reasonable to include. These elements model an additional RC shunt that accounts for the finite resistance of the handle wafer and the capacitance created by the buried oxide (BOX) layer. It turned out, however, that in our case the impact of this additional RC circuit 4 is negligible despite the rather low substrate resistivity of about 20 cm as specified by the wafer manufacturer. This additional RC shunt load is therefore disregarded in the following, but could be easily included if modified device or material choices would require an adapted model [30]. The overall circuit shown in Fig. 1(d) can be transformed into the circuit in Fig. 1(e) by introducing the frequency-dependent differential conductance  ) and the frequency-dependent differential capacitance   Y S (  RF   G  S C   S RF ( )  1  RF   Y S   . (2) (3) This circuit resembles that of a classical telegrapher-type transmission-line model, and the transmission line can thus be fully described by the complex propagation parameter  =  + j together with the complex characteristic impedance Z0 obtained from the well- known transmission-line equations [38]       j ( R  TL  j  L G   S RF TL )(  j  RF ( C C   TL S )), Z 0  R   TL j   RF j L   RF TL ( C C   S TL ) . G  S (4) (5) Note that in these relations denotes an amplitude attenuation coefficient. The corresponding power attenuation coefficient amounts to 2 . For analyzing the EO interaction, we now consider a travelling-wave EO modulator with characteristic impedance Z0, complex RF propagation parameter  , optical group refractive index ng,opt in the phase shifter, and active phase-shifter length  . The device is connected to an RF source with impedance Zsrc and terminated by an impedance Ztrm. Due to the Pockels effect, the local optical phase shift increases in proportion to the local modulation voltage. If a RF is applied to the modulator's signal electrode, sinusoidal voltage with angular frequency RF t the total phase shift in each arm can then be obtained from an integral of the modulation voltage seen by the optical signal along the phase shifter length  , see ( ) contains both the frequency- ref. [39], Eq. (2). In this relation, the complex amplitude dependent amplitude and the phase of the sinusoidally varying phase shift )(t . The complex m  of such a phase modulator is given by the phase amplitude frequency response ( 0  for a ) , normalized to its value that would be obtained for an angular frequency RF source and a termination impedance that are both perfectly matched to the modulator's characteristic transmission-line impedance 0Z ( [39], Eq. (3)) . This leads to   RF ( t  e )  j  PH RF RF RF ) ( ( ) where m P H (  R F )  ) (  RF (0) Z  Z  src trm  HA R C  Z 0       q e q 1  B   A  2  Z 0 Z  0 Z B  1    src c r s   exp( 2 )  1  exp( 2 )  m r t  ,  trm ,   q e   q , 1     (6) (7) (8) 5 and where the reflection factors at the source and the termination are given by . g,opt RF c      and  jn trm src   , q 0 0   src trm (9) (10) Z Z src   Z Z 0 Z Z trm   Z Z 0 ) RC R( F The quantity H  in Eq. (6) accounts for the fact that the voltage that becomes effective for modulation in the slot waveguide is reduced with respect to the voltage on the metal transmission lines through a voltage divider formed by the slab conductance and the slot capacitance [30]. This reduction becomes more pronounced as the frequency increases and can be written as 1 C G    S F S R / 1 j  H (  RF )  R C . (11) The frequency response of EO modulators is usually obtained from a frequency-dependent decay of an intensity modulation rather than from a phase modulation. To connect the two quantities, we consider an MZM in push-pull configuration, for which the phase shifts 1 and  . The phase difference between the two 2 in the two arms fulfill the relation 1 arms  is then given by of the optical output signal is related to the phase difference  by 2   , and the intensity opt 12      I I 2 cos ( / 2) co 2 (s ).  1   opt      / 2   (   (12) If the push-pull MZM is biased at the 3 dB point (quadrature point) and operated under small- signal conditions, i.e., with ), Eq. (12) can be linearized and the change in optical power optP is in proportion to . In characterization experiments, the optical output of the modulator is usually coupled to a photodetector, the photocurrent pI of which is proportional to the received optical power such that . I  The small-signal intensity frequency response conversion is therefore identical to the phase modulator frequency response given in Eq. (6), (13)  p m  of the electro-optic-electric (EOE) EOE P   opt (  ) / 2 R F To quantify f  3dB,EOE 3dB,EOE 2 the bandwidth of  PH RF EOE( m m RF the modulator, (14) frequency of the intensity modulation may be used. The 3 dB corner frequency m  is reduced by a ( the 3 dB RF 2 , for which 3dB,EOE 2 is defined as the frequency corner  2 EOE   RF . ) )   2 f  EOE(0) m factor of two compared to , which corresponds to a power decay of the associated spectral component of the photocurrent by a factor of two. Alternatively, also the 6 dB corner m  can be considered, which corresponds to frequency 6dB,EOE  m  and of the associated photocurrent power by a factor of four and is a decay of hence equivalent to a reduction of the photocurrent amplitude or, equivalently, the phase modulation amplitude by a factor of two. Both values are used in the literature to refer to the bandwidth of a modulator [40 -- 42], sometimes without explicitly stating which of the two values is specified. In the following, we either specify both values or reside to the 6 dB corner 6dB,EOE 2  2  EOE of EOE    R R F F 2 6 frequency better modulators [23,26,30,42]. for comparability to similar investigations of SiP EO To apply the model developed here to a given device design, the complex propagation parameter  and the characteristic impedance Z0 can be directly determined from S-parameter measurements or with numerical mode solvers. The equivalent-circuit parameters SC can then be obtained by fitting Eq. (4) and Eq. (5) to the frequency- R L C G   S TL dependence of  and Z0, see Section 3. If, in addition, ng,opt is known, Eq. (6) allows to predict the EO response of the modulator.  TL  TL , , , , ( ) R F and  2  1 500 µm 750 µm 3. Electrical characterization and fitting procedure To experimentally verify the model introduced in Section 2, we determine the frequency- dependent characteristic impedance Z0 and propagation parameter  of a typical SOH modulator. These parameters fully describe the electrical behavior of the device and build the m  of the electro-optic-electric (EOE) base for determining the frequency response EOE conversion, see Section 4. For the devices used here, the thickness of the BOX layer was 3 µm, the device layer thickness was 220 nm, and the slot width was 120 nm. To de-embed the characteristics of the devices-under-test (DUT) from those of the RF contact-pad parasitics, we investigate two MZM with nominally identical cross sections and different active lengths of the phase shifters. In a first step, we measure S-parameters of the embedded devices using a vector network analyzer (VNA) and microwave contact probes. Figure 2(c) shows the measured (embedded) transmission S21 and reflection S11 of the two modulators for a gate voltage Ugate = 0 V. The reference plane was moved to the tip of the probes using a short-open-load-through (SOLT) calibration routine with an impedance-standard substrate. The phase-shifter sections of both devices are thus only embedded into identical on-chip contact circuits that are formed by the contact pads and subsequent transitions on both ends of the modulator, see Fig. 2(a), and that are referred to as contact fixtures in the following. Each of these fixtures is modeled by a series impedance Z and a shunt admittance Y as shown in Fig. 2(b), which can be represented by electrical transfer matrices (T-matrices [38]) TL and TR for the left and the right fixture, respectively, see Appendix A, Eq. (A1) and Eq. (A2). These T-matrices can be calculated from the measured S-parameters S21 and S11 using standard microwave theory, see Appendix A. To de-embed the T-matrices TDUT1 and TDUT2 of the phase shifters with lengths 1 and 2 , we first extract the measured T-matrix of the embedded modulator Tm1 = TLTDUT1TR and Tm2 = TLTDUT2TR from the measured S-parameters. Introducing the T-matrix TDUT0 of an 0 -long de-  we can rewrite TDUT1 03 , embedded modulator section and exploiting that and TDUT2 as TDUT1 = TDUT0TDUT0 and TDUT2 = TDUT0TDUT0TDUT0. The T-matrix of the virtual through connection shown in Fig. 2(b) is defined as Tthru = TLTR and can then be calculated from the measured matrices as . This allows to determine TL and TR directly from the measured transfer matrices Tm1 and Tm2 and to de-embed the modulator response by calculating . Using Eq. (A7) of Appendix A, we can then extract the frequency-dependent characteristic impedance Z0 and the frequency-dependent propagation parameter  from the transfer matrices TDUT1 and TDUT2 of the de-embedded modulator, see black traces in Figures 2(d-g). We find that the results obtained from TDUT1 and TDUT2 show very good agreement, thereby confirming the validity of our de-embedding approach. For comparison, we also extract Z0 and  by applying Eqs. (A3- A7) directly to the measured T-matrices Tm1 and Tm2 of the embedded modulators, thereby ignoring the impact of the contact fixtures. The resulting frequency characteristics of Z0 and  T T T T T 1 T T T  1 R 1 1 T T T   L 2 R 02 1  and 2 m1 m2 m1 m2 m1 250 µm D TT U 1 D TT U 2   T thru  and  1  L m 1  1   m 7 are depicted as blue and red traces in Figures 2(d-g) and show clear differences, thereby underlining the importance of proper de-embedding. Our results indicated that Z0 has a negligible imaginary part and amounts to approximately 50  -- except for the low-frequency region, where Z0 deviates from its ideal low-loss and high-frequency approximation, see Eq. (B1) in Appendix B. This ensures good impedance matching to typical feeding networks and drivers. The real part  of , associated with RF propagation loss, increases with frequency due to loss in the metal stripes and in the resistive slabs, see the following paragraphs for a detailed discussion of the loss mechanisms. The imaginary part  of  shows a linear increase indicating a low dispersion of the RF transmission line. Fig. 2. (a) Simplified electrical structure of the SOH modulators. The phase-shifter sections with lengths 1 or 2 and T-matrices TDUT1 or TDUT2 are electrically embedded into contact fixtures, modeled by T-matrices TL and TR. (b) The contact fixtures are modelled by a series impedance Z and a shunt admittance Y. (c) Amplitude and phase of measured S-parameters for (blue). (d) Real part and (e) embedded MZM with lengths 1 500µm imaginary part of the characteristic impedance Z0, along with (f) real part and (g) imaginary part of the propagation constant  Black curves indicate values as extracted from de- embedded modulators of lengths 1  and 2  . These results show good agreement and hence confirm the validity of our de-embedding approach. In contrast to that, the red and blue curves correspond to values extracted from the embedded modulators by ignoring the impact of the contact fixtures. Consequently, the results obtained for the different device lengths are not in agreement. (red) and 2 750µm   8 ,  TL ,  Z and L  TL   S,bulk R  TL , , G  S,acc L C G  TL The measurement of S-parameters and subsequent de-embedding and extraction of  and Z0 was repeated for a range of gate voltages Ugate = 0…300 V. Figure 3 shows the numerical values obtained from measurements in blue. To determine the six equivalent-circuit SC , the following approach was used: First, parameters , and  0 Eq. (4) and Eq. (5) were used to calculate directly TLR Z     from  and Z0. Using the fact that both and are independent of frequency and of the T LL  applied gate voltage, the two values were calculated for all frequencies and gate voltages and then averaged. In a second step, a multi-dimensional multi-parameter least-squares fit was used to obtain SC by fitting Eq. (4) and Eq. (5) to the measured frequency-dependent and gate-voltage-dependent values of  and Z0, see Table 1 for the resulting numerical values. The fitted curves are shown in red in Fig. 3 and show good agreement with the measured data. For increasing frequencies, the characteristic impedance / ( ) 50 C C Z L     , which  quickly approaches its ideal high-frequency real value 0 TL S TL can be obtained from Eq. (5) using TL The characteristic  ). S( C C   TL impedance Z0 as well as the imaginary part of  are rather insensitive to the gate voltage.  G  RF 0   RF R   and S C G  TL ,  S,bulk RF TLL  , G  S,acc and T LR  Fig. 3. Measured values (blue) and fitted curves (red) according to Eq. (4) and (5) as a function of both RF frequency and gate voltage Ugate for (a) the real part of the characteristic impedance Z0, (b) the imaginary part of the characteristic impedance Z0, (c) the real part of the complex propagation parameter  associated with loss, and (d) the imaginary part of the complex propagation parameter  Table 1 Circuit parameters obtained by fitting Eq. (4) and (5) to the measured frequency-dependent and gate- voltage-dependent values of  and Z0, see Fig. 3. 9 Fit 13.00 /mm pH/mm 414.00 130.00 fF/mm 2.76 mS/mm 145.00 µS/(Vmm) 160.00 fF/mm In contrast to this, the losses show a strong dependence on the conductivity of the slabs and thus on Ugate. To explain this finding, we first consider the losses in the metal transmission line represented by separately from the losses in the RC element which represents the slab and the slot and which is formed by SC . The different loss contributions are plotted as a function of frequency in Fig. 4 for three different gate voltages. Note that the separation and the independent analysis of the two loss contributions represents an approximation and might thus lead to small deviations from the mathematically correct overall loss that can be obtained from the real part  of the complex propagation parameter  according to Eq. (4).  and TLC TL ,R L TL G  S,bulk G  S,acc and , For the contribution of the bare transmission line, we neglect Eq. (4), such that the loss can be written as G  S,bulk , G  S,acc and SC in For low frequencies RF RF ,   TL  TL , TL    R  TL   R  TL L  TL j   L   RF TL , the loss  . C  RF TL j (15)  TL increases in proportion to For higher frequencies, RF low-loss, high-frequency approximation [38] and is constant with frequency, (16) TL of the bare transmission line approaches its for RF   TL   TL  RF  TL  R C  TL  TL 2 , the loss (17) The frequency dependence of TL according to Eq. (15) is illustrated in Fig. 4 by the dotted black lines. for RF    TL TL. For the loss contribution of the RC-element that is formed by the slot capacitance SC and  ( L R   TL TL C  TL / 2) / the slab conductance S  G G  S,acc G  S,bulk , we derive a relationship in Appendix C, C G 2 2     S RF S ) G C 1 2     S RF S For frequencies well below the RC corner frequency RC  can approximate Eq. (18) by 1 (   RC  1 1   1 2 Z   0 /G C    S S . (18) of the RC element, we 2   R RC  F 2 C  S G   S 2 1  Z 1  0  for RF   RC, (19) leading to a quadratic increase of the losses with frequency. For frequencies well above the RC corner frequency RC , we obtain  RC G  S 1  1  2 Z 0  for RF   RC, (20) i.e., the losses are independent of frequency. The frequency dependence of RC according to Eq. (18) is illustrated in Fig. 4 by the dashed black lines. 10 Fig. 4. Detailed representation of RF loss mechanisms in the slot-waveguide modulator. Blue data points represent measured data, and red traces correspond to the loss obtained by calculating the real part of the complex propagation parameter , Eq. (4), for the circuit parameters obtained by the fit. The three panels correspond to three values of the gate voltage Ugate and represent slices of the 3D plot in Fig. 3(c). Besides the gate voltage Ugate, the S'G is specified. The black dotted traces depict the loss associated slab conductance contribution TL of the transmission line formed by the metal electrodes only, see Eq. (15), and the black dashed traces show the loss contribution RC of the RC element formed by the slab and the slot, see Eq. (18). The solid black line shows the sum of TL and RC, which is in good agreement with the loss obtained from the full model represented by the red solid trace. (a) Ugate = 0 V: Except for very low frequencies, see inset, the loss is constant with frequency. (b) Ugate = 150 V: The quadratic frequency-dependence of RC for frequencies below the corner frequency RC of the RC element, see Eq. (19), leads to a strong increase of the total loss with frequency. At frequencies the total loss approaches a constant value, see Eq. (17) and Eq. (20). (c) Ugate = 300 V: The corner frequency fRC is further increased compared to (b), and the total loss no longer reaches its constant high- frequency value in the considered frequency range. 2 ) f    / ( RC  )2   ( RF f RF  RC f 11 The overall loss can be estimated by adding TL and RC , leading to the solid black lines in Fig. 4. These estimated losses are in good agreement with the loss obtained from directly using    according to Eq. (4), see red lines in Fig. 4.   1 SG  The three panels for Fig. 4(a, b, c) correspond to different gate voltages of 0 V, 150 V, and 300 V, respectively. For the samples used in our experiments, the doping of the slabs is rather low, and hence the resistance of the slabs is rather high in case no gate voltage is applied. This leads to a small corner frequency RC , illustrated by a dashed vertical line in Fig. 4(a), and the loss contribution RC of the RC-element quickly reaches the frequency- independent value given by Eq. (20). For low frequencies, the total loss is dominated by the contribution TL of the bare transmission line, Eq. (16), and increases in proportion to RF . If Ugate is increased, Fig. 4(b) and 4(c), RC becomes larger and exceeds TL , which leads to a strongly frequency-dependent increase of the overall loss dominated by for frequencies RF between TL and RC . At even higher frequencies ( RF ), the total losses are again frequency-independent, see Eq. (17) and Eq. (20). For a gate voltage of 300 V, Fig. 4(c), the plateau, where the overall loss becomes frequency-independent, is no longer reached in the considered range of RF frequencies. For the discussion of design trade- offs (Section 5) it is important to note that in the 5…50 GHz range, increasing the slab conductivity, in the range considered here by changing Ugate, brings the modulator from a regime with weakly frequency-dependent loss to a regime with strongly frequency-dependent loss and may hence limit the bandwidth of the device. 2   RF   RC RC , , , , , G  TL  TL  S,acc  S,bulk TLR L C G  4. Measured and modelled electro-optic response With the knowledge of the six equivalent-circuit parameters SC and of the optical group velocity vg,opt in the slot waveguide, Eq. (4 - 11) can be used to predict the EO frequency response of the SOH modulator. The optical group velocity vg,opt = c/ng,opt is obtained by calculating the group refractive index ng,opt = 3.2 by means of a numerical mode solver. In Fig. 5(a), the red lines show the EO frequency response as predicted by the model for a 750 µm-long modulator terminated with a 50  impedance under the influence of gate voltages between 0 V to 300 V. These predictions are based on the circuit parameters listed in Table 1, as obtained from the purely electrical measurements of the frequency-dependent RF scattering parameters S21 and S11. Increasing the gate voltage increases the bandwidth, which is mainly due to a reduction of the RC time-constant of the slab-loaded slot waveguides. The predicted frequency response is compared against measured values obtained by using a VNA and a calibrated optical receiver [43], see blue traces in Fig. 5(a). In these measurements, the swept-source microwave stimulus of the VNA is applied to the modulator via microwave probes. The modulator is biased at its 3 dB point and the microwave signals amplitude is chosen small enough for the modulator to be operated under small-signal condition. The optical signal is subsequently amplified using an erbium-doped fiber amplifier (EDFA), filtered by a band-pass filter to remove out-of-band amplified- spontaneous-emission noise (ASE), and converted back to the electrical domain using a photodiode that is connected to the receiver port of the VNA. The frequency-dependent decay m  according to Eq. (14) and Eq. (6) can be of the EOE frequency response derived from the VNA measurement by the relation EOE   RF 10 log  m EOE (  RF ) 2   S 21,VNA,dB (  RF )  S 21,VNA,dB (  RF,0 ). (21) 12 21,VNA,dB 40 MHz In this relation, denotes the scattering parameter measured by the VNA in dB after de-embedding from the frequency response of the photodiode and of the microwave probe, and RF,0 is a small reference frequency that is chosen as small as allowed 2  by the VNA. The different shades of blue in Fig. 5(a) indicate three independent measurements with small statistical variations. Overall, we find a good agreement of the measurements with the model predictions (red). In Fig. 5(b), both the 3 dB and the 6 dB EOE bandwidths of the modulator are shown for different gate voltages. For the 6 dB EOE bandwidth f6dB,EOE, blue diamonds correspond to the measured values, and the dotted red line represents the bandwidths predicted by the model. Similarly, blue circles and the solid red line indicated measurements and model predictions of the 3 dB EOE bandwidth f3dB,EOE. The good agreement of the predicted bandwidths with their measured counterparts validates the model introduced in Section 2, i.e., the proposed method allows to accurately predict the EOE frequency response of an SOH modulator based on the knowledge of purely electrical characteristics along with the group velocity g,opt of the optical signal in the slot waveguide. The electrical characteristics are easily accessible by either measurements in the RF domain, by simulations of the electrical RF device characteristics, or by estimating the distributed- element parameters from the modulator geometry and material properties. We believe that high-throughput wafer-level testing protocols might greatly benefit from a quantitatively verified model-based prediction of the EOE frequency response based on easily accessible electrical scattering parameters. v S   ) RF EOE( m  according to Eq. (6) and Fig. 5. Electro-optic-electric (EOE) frequency response Eq. (14) of a 750 µm-long SOH modulator terminated with a 50  impedance. (a) EOE response for different gate voltages Ugate of 300 V, 200 V, 150 V, 100 V, 75 V, 50 V, 40 V, 30 V, 20 V, 10 V, and 0 V (from top to bottom). The red curve represents the prediction of the model according to Eq. (4) to Eq. (11), and the blue line corresponds to the measurement results, where the different shades indicate three independent measurements. With increasing gate voltage, the frequency response flattens. (b) 3dB EOE bandwidth f3dB,EOE (red solid line: model; blue circles: measurement) and 6 dB EOE bandwidth f6dB,EOE (red broken line: model; blue diamonds: measurement) as extracted from the curves in (a) as a function of applied gate voltage Ugate. The predicted bandwidths agree well with their measured counterparts, thus confirming that the proposed method allows to accurately predict the bandwidth of an SOH modulator based on the knowledge of purely electrical characteristics. 5. Design guidelines 5.1 Overcoming bandwidth limitations Having validated the model of the SOH modulator, various design parameters can be tested for their impact on device performance. In the following, we assume a device with 13  TL , and TLC transmission line parameters TL as specified in Table 1 and investigate the ,R L impact of variations of the slot capacitance SC and of the slab conductance SG on the EO bandwidth. This leads to the following considerations:  The slab conductance SG can be adjusted by modifying the doping profile in the slab. Increased slab conductivity will affect the electrical behavior and thus opens a way to increase the modulators bandwidth as discussed in detail below. As a trade-off, when increasing the slab conductivity by an increased doping concentration, the higher concentration of free carriers will also increase optical losses due to free carrier absorption (FCA).  The slot capacitance SC is mainly affected by the geometry of the slot waveguide and can be reduced by increasing the slot width. Reduced slot capacitance will affect the electrical behavior and thus the bandwidth as discussed in detail below. As a trade-off, increasing the slot width reduces the modulating electric RF field in the slot as well as the confinement factor of the optical field in the slot region. This can lead to a reduced modulation efficiency [23,44]. However, the poling efficiency may also be impaired by too narrow slot widths [45]. As an example, previous investigations of SOH modulators based on the EO material JRD1 [16] have shown that increasing the slot width to, e.g., 200 nm does not decrease the modulation efficiency of these devices. The frequency response expressed by Eq. (6) considers all bandwidth-limiting effects, namely RC limitations of the slot waveguide together with the slabs, RF loss, impedance mismatch, and velocity mismatch. To investigate the impact of RC limitations, we first consider a theoretical modulator only limited by the RC-low-pass formed by the slab resistance and the slot capacitance, explicitly neglecting impedance mismatch, velocity mismatch and RF loss. The response of such a device is given by H  , Eq. (11). Figure 6 shows the 6 dB for a device that is only limited by the RC-low-pass as a function of EOE bandwidth 6dB,EOE slab conductance SG and slot capacitance SC . The parameters of the devices used for experimental verification indicated by a green this paper, see Table 1, are RC ( in f ) RF SC and the slab conductance Fig. 6. 6 dB EOE bandwidth f6dB,EOE of a modulator only limited by the RC low-pass formed by SC and an increase the slot capacitance SG linearly increases the bandwidth. The parameters of the devices used in this paper for of experimental verification of the model, see Table 1, are indicated by a green star for the case of a gate voltage of 0 V and by a magenta star for a gate voltage of 300 V . For the parameters indicated by Ⓐ , Ⓑ , and Ⓒ , the frequency response of a 750 µm-long modulator is exemplarily shown in Fig. 8. SG . Both a reduction of 14 S f star for the case of 0 V gate voltage applied and by a magenta star for 300 V gate voltage applied. The bandwidth increases linearly with both SG and 1/ C . Note that the RC limitation is independent of the device length. Using the current devices as a reference, an increase of SG , e.g., by a factor of 20, possibly combined with a slight reduction of SC by, e.g., 40 %, in excess of 100 GHz if the device was would permit RC-related EOE bandwidths 6dB,EOE only RC-limited. Increasing SG by a factor of 20 can be realistically achieved without introducing excessive optical loss by using optimized doping profiles, see Section 5.2 for details. For all slot capacitances Next, we consider the impact of RF loss, impedance mismatch, and velocity mismatch on the modulator bandwidth. Figure 7(a) shows f6dB,EOE of a 750 µm-long modulator limited only by RF loss, impedance mismatch, and velocity mismatch, explicitly neglecting RC is given by Eq. (6) for limitations. The response of such a SG tends H   to reduce the bandwidth, see Fig. 7(a). This can be understood by considering that increasing the slab conductance SG initially brings the modulator from a regime with low and frequency- independent RF loss to a regime with high RF loss that increases with frequency, see Fig. 4. If the slab conductance SG is increased further, the RF loss and its frequency dependence are reduced again, and the impact on the bandwidth decreases. In practical devices, however, SG is constrained due to optical loss caused by FCA in the highly doped slab regions. This trade- off can be mitigated by optimized doping profiles, see Section 5.2. theoretical device SC, an increase in the slab conductance const. RC( RF )   50GHz We further investigate the overall RF loss  of the modulator, including both the loss in the metal electrodes and in the RC element formed by the slot waveguide and the slabs, as well as the associated real-part Z0 of the characteristic impedance Z0 and the effective RF index neff,RF. The numerical values for  Z0, and neff,RF are shown exemplarily for an RF frequency of 50 GHz in Fig. 7(b), Fig. 7(c), and Fig 7(d), respectively. While the RF loss, Fig. 7(b), initially increases substantially with the slab conductance SG , the characteristic impedance Z0, Fig. 7(c), stays close to 50 . For the effective microwave index neff,RF, we SG , Fig. 7(d). This can be observe a strong initial increase with the slab conductance understood by considering that, for a low slab conductivity SG , the RC element is operated above its corner-frequency RC . In this regime, the slot capacitance SC is only partially charged during one cycle. If the slab conductance SG is increased, the slot capacitance becomes increasingly effective, and the RF wave is slowed down. For larger values of the slab conductance SG , the RC element is operated above its corner frequency and the capacitor is fully charged and de-charged during each cycle, leading to an RF index neff,RF that is essentially independent of the slab conductance. Following the same argument, it can also be understood why an increase of the slot capacitance SC therefore leads to an increase of neff,RF. To investigate the frequency dependence of the RF loss, of the characteristic impedance, of the effective RF index, and of the RC low-pass as well as the associated impact on the modulator bandwidth, we consider three exemplary SG -value-pairs for a 750 µm-long device. The chosen value pairs are marked by Ⓐ,Ⓑ, and Ⓒ in Fig. 6 and Fig. 7. The slot capacitance is fixed to SG is increased from (Ⓐ), which corresponds to the value extracted from the device used in our 2.76 S m G  S experiment, to (Ⓒ). Figure 8(a) shows, for the G  S , while the slab conductance (Ⓑ) and further to 100 pF m 40 S m 80 S m C  S G  S SC - 15 S'C and S'C and a low SG . For a large SG , an increase of SG , see Fig. 4(a) and Fig. 4(b), but decreases again for very large values of Fig. 7. (a) 6 dB EOE bandwidth f6dB,EOE of a 750 µm-long modulator limited only by RF loss, impedance mismatch, and velocity mismatch, explicitly neglecting the impact of the RC-low- SG leads to a pass formed by strong frequency-dependent increase of the RF loss and thus to a reduced bandwidth. For the parameters indicated by Ⓐ,Ⓑ, and Ⓒ the frequency response is exemplarily shown in Fig. 8. (b) RF loss for a 750 µm-long modulator at 50 GHz. The RF loss initially increases with increasing SG , see Eq. (19). (c) Characteristic impedance Z0 at 50 GHz. The impedance remains close to the system impedance of 50 . (d) Effective RF refractive index neff,RF . For values of S/m and pF/m, neff,RF is close to the effective optical group refractive index of 3.2 Increasing the slot capacitance slows down the RF wave and hence increases the RF refractive index. The detailed analysis in Fig. 8 shows that the bandwidth is not limited by velocity S'C for the considered length of the modulator. The mismatch for practical values of parameters of the devices (see Table 1) used in this paper for experimental verification of the model are indicated by a green star for the case of 0 V gate voltage applied and by a magenta star for 300 V gate voltage applied. S 130 C  G  S 50 S 2.76 G  S/m, the EOE frequency response m  , the weakly conducting slab with EOE RF loss , the characteristic impedance Z0, and the effective RF refractive index neff,RF, all as a function of RF frequency up to 150 GHz. The solid red line depicts the total frequency response. Broken lines in blue, magenta, green, and black depict the frequency response by the sole contribution of the RC low-pass, the RF loss, the impedance mismatch, and the velocity mismatch, respectively. For a weakly conducting slab with S/m, the total bandwidth (red line) is limited almost exclusively by the RC low-pass (blue broken line), whereas the impact of RF loss (red magenta broken line), impedance-mismatch (green broken line) and of velocity mismatch (black broken line) are negligible, because , Z0, and neff,RF G  2.76 RF S ( ) 16 SG  2.76 ) RF and (a) SG  40 SG  80 S/m, (b) S/m, and (c) Fig. 8. Electro-optic bandwidth and RF properties for a 750 µm-long modulator with S 100 pF m C  S/m, corresponding to the points marked by Ⓐ,Ⓑ, and Ⓒ, respectively, in Fig. 6, Fig. 7(a), and Fig. 9(b). The first EOE( m  of the modulator (solid red line), as row shows the overall frequency response well as the frequency responses that correspond to the sole impact of RC limitations (blue dashed line), RF loss (magenta dashed line), impedance mismatch (green dashed line), and velocity mismatch (black dashed line). The second row shows the total RF loss  In (a) the RF loss is low and frequency-independent, see Eq. (17). In (b) and (c) losses in the resistive slab are increased and the losses becomes strongly frequency-dependent, see Eq. (19), thereby reducing the bandwidth to values below the RC limit. The third row shows the characteristic impedance Z0, which stays close to the system impedance of 50  except for low frequencies, where the Z0 deviates from its ideal high frequency, low loss value, see Appendix B. The last row shows the effective RF refractive index neff,RF. In (a), the RC corner frequency R C is small and the slot capacitance is not fully charged and de-charged as the RF wave propagates along the modulator. If the resistance of the slabs is reduced, see (b) and (c), RC increases and the slot capacitance has a stronger impact on the RF wave, leading to a reduced velocity. The effective RF refractive index neff,RF then becomes close to the optical group index ng,opt = 3.2. Overall, impedance and velocity-mismatch are negligible for all of the three considered cases. In (a), the device is limited by the RC time constant of the slab-slot configuration. In (b) and (c) the RC time constant is reduced and the associated bandwidth is increased accordingly. At the same time, the increased RC corner frequency leads to an increased frequency-dependence of the RF loss, which causes additional bandwidth limitations. 17 and S show little frequency dependence. If the slab conductivity is increased to, e.g., 40 S/m, Fig. 8(b), or 80 S/m, Fig. 8(c), the RC corner frequency increases. However, frequency- dependent RF loss starts to become significant and limits the overall bandwidth of the device to values smaller than the RC corner frequency. For the increased slab conductance impedance mismatch and velocity mismatch do not G  S significantly affect the bandwidth either. This confirms that RF loss is the most important bandwidth-limiting effect besides the RC time constant of the slab-slot configuration. 80 S m, 40 S m G  f Finally, we consider the 6 dB EOE bandwidth 6dB,EOE obtained from the full model as a function of the slab conductance SG and the slot capacitance SC , taking into account all bandwidth-limiting effects, i.e., the RF loss, the impedance mismatch, the velocity mismatch and the RC low-pass. Figure 9 shows 6dB,EOE as obtained by the full model for four different lengths  of the modulator. For long devices, the frequency dependence of the RF loss is more significant, and increasing the slab conductivity SG has a smaller impact on the overall f Fig. 9. 6 dB EOE bandwidth f6dB,EOE for modulators of different lengths  based on the full model considering all bandwidth-limiting effects. The color scale is the same as in Fig. 6. If the length is reduced, the bandwidth of the modulator approaches the RC-limited case depicted SC - in Fig. 6. The frequency response of the 750-µm-long modulator is shown in Fig. 8 for the S'G -values indicated by Ⓐ,Ⓑ, and Ⓒ in (b). The parameters of the devices used in this paper for experimental verification of the model, see Table 1, are indicated by a green star for the case of a gate voltage of 0 V and by a magenta star for a gate voltage of 300 V. 18 bandwidth than for short devices, see Fig. 9(a). Specifically, if  is reduced to 500 µm, Fig. 9(c), and further to 200 µm, Fig. 9(d), the bandwidth approaches the RC-limited case shown in Fig. 6. A combination of reduced device length, reduced slab capacitance SC, and increased slab conductance SG thus enables bandwidths well beyond 100 GHz. 5.2 Optimized doping profiles for low-loss modulators To increase the bandwidth of the modulators, highly conductive slab regions are desirable. Increasing the doping concentration near the optical slot waveguide, however, leads to optical loss due to free-carrier absorption (FCA). This effect can be mitigated by a two-step doping profile, see Fig. 1(b) [23,28,29]. Based on this approach, we suggest the following design procedure for SOH modulators: (i) (ii) (iii) Given a certain EO material and a practical slot width, the targeted -voltage U will determine the length  of the modulator. Based on the slot width, which determines the slot capacitance SC , and the length  of the modulator, the targeted bandwidth determines the slab conductance SG , see, e.g., Eq. 6 and Fig. 9. For a given slab conductance SG , the optical loss can be minimized by optimizing the two-step doping profile [28,29]. This optimization is described in the following paragraph.  TL 102 S G  We exemplarily demonstrate the optimization of the doping profile for a device with a S/m and a fixed slot width of 200 nm. For this device, targeted slab conductance of a slot capacitance SC of about 100 pF/m can be estimated by assuming a parallel-plate approximation. This assumption has to be taken with caution since fringing fields can SG contribute to a significant portion of the slot capacitance [23]. For the chosen values of and , and ,R SC , a device with length  = 1 mm, and transmission line parameters TL L as specified in Table 1, the model according to Eq. (6) would predict a bandwidth of TLC f6dB,EOE = 114 GHz. The geometry of the slot waveguide is fixed with a slab height of hslab = 70 nm, a rail width wrail = 240 nm, and a rail height of hrail = 240 nm as shown in Fig. 10(a). The weakly doped region of the slab near the slot is depicted in green, and the more heavily doped region further away from the slot is depicted in blue in Fig. 10(a). As free parameters, we can vary the resistivity heavy of the heavily doped region as well as the resistivity weak and the width wweak of the weakly doped part. For simplicity, we choose a fixed resistivity weak of the weakly doped section in a first step, here weak = 5.5×10−4 m. For a target conductance S/m of the slab, Fig. 10(b) shows the required bulk resistivity heavy of the heavily doped region as a function of its width wheavy . To reach the overall target conductance of S/m, the width wweak of the heavily doped region must stay below 1.16 µm. The bulk resistivity in the weakly and heavily doped regions can be translated into optical loss parameters using the experimentally obtained relations published in [46]. Figure 10(c) shows the bulk power absorption coefficient Si in the heavily and weakly doped silicon. Note that Si refers to the optical power -- in contrast to the RF attenuation coefficient  , specified in Eq. (4), which refers to the amplitude. The underlying model for the optical power absorption coefficient assumes a linear increase of the optical loss with the electrical conductivity in Si [46]. This assumption is only valid as long as the electrical conductivity is not limited by electron-impurity- scattering [47]. To translate the bulk absorption coefficient into a waveguide loss parameter, G  S 102 G  S 102 19 we compute the optical mode in the slot waveguide and the overlap with the doped silicon region [24] using a vectorial mode solver. The resulting power absorption coefficient WG for the optical waveguide mode is shown in Fig. 10(d). If the weakly doped region is narrow, the heavily doped section comes close to the slot region leading to high optical loss. If the heavily doped section is far away from the slot waveguide, the conductivity has to be increased to maintain a given overall conductance SG , which again increases the optical loss. A minimum excess loss of approximately 0.048 mm-1 (0.2 dB/mm) is found for a weakly doped region with an intermediate width of wweak = 1 µm and a medium resistivity of the heavily doped section of heavy = 1.43×10−4 m. Note that this doping-related excess loss does not significantly change the overall propagation loss of practically used slot waveguides, which is of the order of 0.5 dB/mm [48]. Fig. 10. Optimization of the doping profile in the slab-waveguide region to minimize optical loss. (a) Cross section of the slot waveguide and the adjacent slab: A low doping concentration is chosen for the rails of the slot waveguide and for the adjacent region of the slab (depicted in green), which interact strongly with the guided light. The doping is increased in the slab region that is further away from the rails (depicted in blue). We consider a waveguide with rails having a width of wrail = 240 nm and a height of hrail = 220 nm. The slabs have a height of hslab = 70 nm and a width of wslab = 1.66 µm. The width of the slot is 200 nm. For a fixed resistivity of the weakly doped silicon (here weak = 5.5×10−4 m), wweak and heavy can be S/m). (b) Bulk resistivity varied to realize the target conductivity of the slab (here heavy in the heavily doped slab regions that is required to reach the target conductivity, specified as a function of wweak for fixed values of weak = 5.5×10−4 m and wslab = 1.66 µm. (c) Corresponding material absorption coefficient in the doped silicon. Note that Si refers to the power attenuation coefficient for the optical wave, whereas the symbol  in Eq. (4) refers to the amplitude attenuation coefficient of the corresponding RF wave. (d) Power loss coefficient experienced by the guided optical mode, obtained from the numerically calculated optical mode field and its overlap with the doped silicon regions. An optimum in terms of optical loss is found for an intermediate extension of the heavily doped section in combination with a medium resistivity. S 102 G  20 U 0.200 mm, So far, we have only considered a fixed resistivity weak = 5.5×10−4 m of the weakly doped section. Varying this resistivity allows to further increase the bandwidth of the device or decrease its optical loss. This degree of freedom becomes particularly important when the length of the modulator is varied. For a quantitative analysis, we have repeated the optimization exemplarily described in the previous paragraph for devices of different lengths  0.5 mm, and 1.0 mm and for different resistivities weak = 5.5×10−4 m, 16.5×10−4 m, and 49.5×10−4 m of the weakly doped region. Figure 11 shows the resulting optical excess loss caused by absorption in the doped silicon regions for modulators of different lengths  as a function of their total bandwidth f6dB,EOE. In this plot, the trade-off   , bandwidth and insertion loss becomes between the modulator figures of merits evident. The numerical values presented in Fig. 11 suggest that it will be possible to realize SOH devices with a length of, e.g., 0.5 mm, having a bandwidth of f6dB,EOE = 100 GHz while keeping the doping-related optical excess loss below 0.1 dB. Using highly efficient EO polymers [16], such a device would have a -voltage of less than 1 V. The detailed device parameters leading to the results shown in Fig. 11 are depicted in Fig. 12. Figure 12(a) shows SG for the the relationship between the device bandwidth and the overall slab conductance three different modulator lengths  considered in Fig. 11. The graphs are obtained from the full model according to Eq. (6) in combination with a slot capacitance of S and ,R the transmission line parameters TL , and TLC as specified in Table 1. Figure 12(b) L specifies the optimized resistivity heavy of the heavily doped slab region along with the associated width wweak of the weakly doped region, see Fig. 12(c), and the associated power absorption coefficient WG for the optical waveguide mode, see Fig. 12(d), considering three fixed values of weak = 5.5×10−4 m, 16.5×10−4 m, and 49.5×10−4 m. 1  TL C  100 pF m Fig. 11. Optical excess loss caused by absorption in the doped silicon regions for modulators of length  = 0.2 mm, 0.5 mm, and 1 mm as a function of the total 6 dB electro-optic-electric (EOE) bandwidth f6dB,EOE. For each length, we consider three different resistivities weak of 5.5×10−4 m, 16.5×10−4 m, and 49.5×10−4 m. The parameters heavy and wweak are then optimized according to the procedure described in Fig. 9. For a given length of the modulator, there is a trade-off between electro-optic bandwidth and optical insertion loss. For a given bandwidth, optimizing the doping profile can lead to a reduction of optical loss. Data points are connected by interpolated lines as guide to the eye. 21 Fig. 12. Details of the optimized SOH modulator design for the modulators of length  = 0.2 mm, 0.5 mm, and 1 mm considered in Fig. 11. For each length, we consider three different resistivities weak of 5.5×10−4 m, 16.5×10−4 m, and 49.5×10−4 m. The parameters heavy and wweak are then optimized according to the procedure described in Fig. 9. (a) Relationship between device bandwidth f6dB,EOE and overall slab conductance SG for three different modulator lengths. (b) Optimized width wweak of the weakly doped region, (c) the associated resistivity heavy of the heavily doped slab region, and (d) the associated power absorption coefficient for the optical mode in the doped region of the silicon, all as a function SG , considering the three fixed values of weak. Data points in all of slab conductance subfigures are connected by interpolated lines as guide to the eye. 6. Summary We formulate and experimentally validate an equivalent-circuit model of silicon-organic hybrid (SOH) slot-waveguide modulators based on distributed elements that describe the electric properties of the device. The model parameters can be extracted from purely electrical measurements, that provide the frequency-dependent scattering parameters of the travelling- wave device. With these parameters, we can accurately predict the small-signal electro-optic (EO) frequency response of the modulators, as confirmed by direct measurements of the EO frequency response. We further formulate design guidelines that lead to an optimum trade-off between EO bandwidth and optical insertion loss of SOH slot-waveguide modulators. We find that proper choice of doping concentrations and geometrical device parameters can 22 enable SOH modulators with lengths of 0.5 mm and 6 dB electro-optic-electric (EOE) bandwidths of more than f6dB,EOE = 100 GHz while maintaining π-voltages of less than 1 V and while keeping the optical excess loss caused by free-carrier absorption in the doped silicon regions below 0.1 dB. Appendix A: Mathematical relations for de-embedding the RF characteristics of on-chip modulators Transfer matrices of contact pads The transfer matrix (T-matrix) of the left-hand-side contact pad including the subsequent transition to the modulator electrodes, see Fig. 2(a), is modelled by a series impedance Z and a shunt admittance Y T L  1 Y    0 1 1 0     Z 1     1 Z   1 Y YZ     . (A1) Similarly, the T-matrix of the right-hand-side contact pad including the subsequent transition to the modulator electrodes is modelled by the same series impedance Z and a the same shunt admittance Y T R  1 0    Z 1    1 0 1 Y     1 Y Z Y Z 1          . (A2) Conversion between S-parameters and T-matrix parameters For conversion between S-parameters and electrical T-matrix parameters, we use the following relation [38,49]: 1  S 11 T 11  S   22 , S  2 S 1  S 11 T 12  Z ref S   22 , 21 S  2 S 21 S  2 S 21 22  S 11   S , (A3) (A4) (A5) 11 ref 1  T 21  Z T 22  S 11 S 22   S  2 S 21 is the reference impedance of the refZ (A6) ,    , and S S 21 12 S S 11 22 In these relations, S measurement system. Transfer matrix of a transmission line The transfer matrix of a transmission line of length  with propagation parameter  and characteristic impedance 0Z is given by [38,49]  cosh(  1 sinh(  )  Z 0 cosh( sinh( (A7)   T   ) . Z )       0 )      23 Appendix B: Low-frequency characteristic impedance of a transmission line The Z   and RFC lines at high frequencies only, where RFL     impedance can then be approximated by a real-valued constant by [38] is oftentimes considered for low-loss transmission  holds. The characteristic characteristic ) / ( R   impedance j   RF standard given form j  RF its in G G C R )  L  (  0 L  0  Z C  . (B1) Note that this approximation is not valid for the low-frequency region where Z0 can be complex and frequency-dependent and the real and imaginary part may diverge for RF 0.   Appendix C: Loss parameter of the RC element In this section we derive the RF amplitude attenuation parameter RC associated with the loss of the RC element formed by the slab resistance S S'C following the approach in [23]. The active power dissipated in the RC element of an infinitesimally short section of the transmission line is linked to the real part of the complex line admittance RCY  dz 1G  and the slot capacitance (C1) where U denotes the voltage phasor and where the admittance RCY  of the RC element is given by U Y  R   d , z P d  2  C Y   RC     1 G  S   1   j C   RF S 1   1 2 2   RF S   1  j C G   S  21  C G   S RF S  . (C2) The active power that is transported on the entire transmission line is given by , such that the ratio of the power lost in the RC element and the total power P        * UU * Z 0      entering the infinitesimally short section is given by 2 1 2  C G    S RF S   1  G C     S RF S exp( 2 z P P RC 0  This differential equation is solved by , where the additional factor of two in the argument arises from the fact that RC refers to the amplitude rather than the power attenuation. The amplitude attenuation parameter RC can thus be written as d P P 1  (C3) 1 Z   d . z 1  0   ) 2  RC  2 1 2  C G    S RF S  1  C G   S  RF S 2  1  1  Z 1  0  . 2  (C4) Funding Deutsche Forschungsgemeinschaft (DFG) projects HIPES (383043731), PACE (403188360), and GOSPEL (403187440); Federal Ministry of Education and Research (BMBF) project SPIDER (01DLR18014A) and Advanced-BAM (05K19VKB); European Research Council 24 (ERC) Consolidator Grant "TeraSHAPE" (773248); IARPA SuperCables Program ICENET (W911NF1920114), Alfried Krupp von Bohlen und Halbach Foundation; Helmholtz International Research School for Teratronics (HIRST); Karlsruhe School of Optics and Photonics (KSOP); Karlsruhe Nano-Micro Facility (KNMF) References 1. S. S. Azadeh, F. Merget, S. Romero-García, A. Moscoso-Mártir, N. von den Driesch, J. Müller, S. Mantl, D. Buca, and J. Witzens, "Low Vπ Silicon photonics modulators with highly linear epitaxially grown phase shifters," Opt. Express 23, 23526 (2015). D. Patel, A. Samani, V. Veerasubramanian, S. Ghosh, and D. V. Plant, "Silicon Photonic Segmented Modulator-Based Electro-Optic DAC for 100 Gb/s PAM-4 Generation," IEEE Photon. Technol. Lett. 27, 2433 -- 2436 (2015). A. Samani, D. Patel, M. Chagnon, E. El-Fiky, R. Li, M. Jacques, N. Abadía, V. Veerasubramanian, and D. V. Plant, "Experimental parametric study of 128 Gb/s PAM-4 transmission system using a multi-electrode silicon photonic Mach Zehnder modulator," Opt. Express 25, 13252 (2017). S. Lange, S. Wolf, J. Lutz, L. Altenhain, R. Schmid, R. Kaiser, M. Schell, C. Koos, and S. Randel, "100 GBd Intensity Modulation and Direct Detection With an InP-Based Monolithic DFB Laser Mach -- Zehnder Modulator," J. Lightw. Technol. 36, 97 -- 102 (2018). H. Zwickel, S. Wolf, C. Kieninger, Y. Kutuvantavida, M. Lauermann, T. de Keulenaer, A. Vyncke, R. Vaernewyck, J. Luo, A. K.-Y. A. K.-Y. Jen, W. Freude, J. Bauwelinck, S. Randel, C. Koos, Y. Ku- Tuvantavida, M. Lauermann, T. de Keulenaer, A. Vyncke, R. Vaernewyck, J. Luo, A. K.-Y. A. K.-Y. Jen, W. Freude, J. Bauwelinck, S. Randel, and C. Koos, "Silicon-organic hybrid (SOH) modulators for intensity-modulation / direct-detection links with line rates of up to 120 Gbit/s," Opt. Express 25, 23784 (2017). S. Wolf, H. Zwickel, C. Kieninger, M. Lauermann, W. Hartmann, Y. Kutuvantavida, W. Freude, S. Randel, and C. Koos, "Coherent modulation up to 100 GBd 16QAM using silicon-organic hybrid (SOH) devices," Opt. Express 26, 220 (2018). M. Burla, C. Hoessbacher, W. Heni, C. Haffner, Y. Fedoryshyn, D. Werner, T. Watanabe, H. Massler, D. L. Elder, L. R. Dalton, and J. Leuthold, "500 GHz plasmonic Mach-Zehnder modulator enabling sub-THz microwave photonics," APL Photonics 4, 056106 (2019). S. Ummethala, T. Harter, K. Koehnle, Z. Li, S. Muehlbrandt, Y. Kutuvantavida, J. Kemal, P. Marin- Palomo, J. Schaefer, A. Tessmann, S. K. Garlapati, A. Bacher, L. Hahn, M. Walther, T. Zwick, S. Randel, W. Freude, and C. Koos, "THz-to-optical conversion in wireless communications using an ultra-broadband plasmonic modulator," Nat. Photon. 13, 519 -- 524 (2019). M. Lauermann, C. Weimann, A. Knopf, W. Heni, R. Palmer, S. Koeber, D. L. Elder, W. Bogaerts, J. Leuthold, L. R. Dalton, C. Rembe, W. Freude, and C. Koos, "Integrated optical frequency shifter in silicon- organic hybrid (SOH) technology," Opt. Express 24, 11694 (2016). S. Shimotsu, S. Oikawa, T. Saitou, N. Mitsugi, K. Kubodera, T. Kawanishi, and M. Izutsu, "Single side- band modulation performance of a LiNbO 3 integrated modulator consisting of four-phase modulator waveguides," IEEE Photon. Technol. Lett. 13, 364 -- 366 (2001). X. Chen, M. M. Milosevic, S. Stankovic, S. Reynolds, T. D. Bucio, K. Li, D. J. Thomson, F. Gardes, and G. T. Reed, "The Emergence of Silicon Photonics as a Flexible Technology Platform," Proceedings of the IEEE 106, 2101 -- 2116 (2018). A. Rickman, "The commercialization of silicon photonics," Nat. Photon. 8, 579 -- 582 (2014). A. H. Atabaki, S. Moazeni, F. Pavanello, H. Gevorgyan, J. Notaros, L. Alloatti, M. T. Wade, C. Sun, S. A. Kruger, H. Meng, K. Al Qubaisi, I. Wang, B. Zhang, A. Khilo, C. V. Baiocco, M. A. Popović, V. M. Stojanović, and R. J. Ram, "Integrating photonics with silicon nanoelectronics for the next generation of systems on a chip," Nature 556, 349 -- 354 (2018). O. Marshall, M. Hsu, Z. Wang, B. Kunert, C. Koos, and D. Van Thourhout, "Heterogeneous Integration on Silicon Photonics," Proceedings of the IEEE 106, 2258 -- 2269 (2018). L. Dalton and S. Benight, "Theory-Guided Design of Organic Electro-Optic Materials and Devices," Polymers 3, 1325 -- 1351 (2011). C. Kieninger, Y. Kutuvantavida, D. L. Elder, S. Wolf, H. Zwickel, M. Blaicher, J. N. Kemal, M. Lauermann, S. Randel, W. Freude, L. R. Dalton, and C. Koos, "Ultra-high electro-optic activity demonstrated in a silicon-organic hybrid modulator," Optica 5, 739 (2018). J. Witzens, "High-Speed Silicon Photonics Modulators," Proceedings of the IEEE 106, 2158 -- 2182 (2018). L. Alloatti, R. Palmer, S. Diebold, K. P. Pahl, B. Chen, R. Dinu, M. Fournier, J.-M. Fedeli, T. Zwick, W. Freude, C. Koos, and J. Leuthold, "100 GHz silicon -- organic hybrid modulator," Light: Science & Applications 3, e173 (2014). S. Wolf, H. Zwickel, W. Hartmann, M. Lauermann, Y. Kutuvantavida, C. Kieninger, L. Altenhain, R. Schmid, J. Luo, A. K.-Y.-Y. Y.-Y. Jen, S. Randel, W. Freude, and C. Koos, "Silicon-Organic Hybrid (SOH) Mach-Zehnder Modulators for 100 Gbit/s on-off Keying," Scientific Reports 8, 2598 (2017). S. Wolf, M. Lauermann, P. Schindler, G. Ronniger, K. Geistert, R. Palmer, S. Kober, W. Bogaerts, J. Leuthold, W. Freude, and C. Koos, "DAC-Less Amplifier-Less Generation and Transmission of QAM 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 25 Signals Using Sub-Volt Silicon-Organic Hybrid Modulators," J. Lightw. Technol. 33, 1425 -- 1432 (2015). M. Lauermann, S. Wolf, P. C. Schindler, R. Palmer, S. Koeber, D. Korn, L. Alloatti, T. Wahlbrink, J. Bolten, M. Waldow, M. Koenigsmann, M. Kohler, D. Malsam, D. L. Elder, P. V. Johnston, N. Phillips- Sylvain, P. A. Sullivan, L. R. Dalton, J. Leuthold, W. Freude, and C. Koos, "40 GBd 16QAM Signaling at 160 Gb/s in a Silicon-Organic Hybrid Modulator," J. Lightw. Technol. 33, 1210 -- 1216 (2015). L. Alloatti, D. Korn, R. Palmer, D. Hillerkuss, J. Li, A. Barklund, R. Dinu, J. Wieland, M. Fournier, J. Fedeli, H. Yu, W. Bogaerts, P. Dumon, R. Baets, C. Koos, W. Freude, and J. Leuthold, "42.7 Gbit/s electro-optic modulator in silicon technology," Opt. Express 19, 11841 (2011). J. Witzens, T. Baehr-Jones, and M. Hochberg, "Design of transmission line driven slot waveguide Mach- Zehnder interferometers and application to analog optical links," Opt. Express 18, 16902 (2010). C. Koos, J. Leuthold, W. Freude, M. Kohl, L. Dalton, W. Bogaerts, A. L. Giesecke, M. Lauermann, A. Melikyan, S. Koeber, S. Wolf, C. Weimann, S. Muehlbrandt, K. Koehnle, J. Pfeifle, W. Hartmann, Y. Kutuvantavida, S. Ummethala, R. Palmer, D. Korn, L. Alloatti, P. C. Schindler, D. L. Elder, T. Wahlbrink, and J. Bolten, "Silicon-Organic Hybrid (SOH) and Plasmonic-Organic Hybrid (POH) Integration," J. Lightw. Technol. 34, 256 -- 268 (2016). C. Koos, J. Brosi, M. Waldow, W. Freude, and J. Leuthold, "Silicon-on-insulator modulators for next- generation 100 Gbit/s-Ethernet," in European Conference on Optical Communication (ECOC) (2007), Vol. 2007, pp. P056 -- P056. R. Ding, T. Baehr-Jones, Y. Liu, R. Bojko, J. Witzens, S. Huang, J. Luo, S. Benight, P. Sullivan, J.-M. Fedeli, M. Fournier, L. Dalton, A. Jen, and M. Hochberg, "Demonstration of a low VπL modulator with GHz bandwidth based on electro-optic polymer-clad silicon slot waveguides," Opt. Express 18, 15618 (2010). H. Zwickel, J. N. Kemal, C. Kieninger, Y. Kutuvantavida, J. Rittershofer, M. Lauermann, W. Freude, S. Randel, and C. Koos, "Electrically packaged silicon-organic hybrid (SOH) I/Q-modulator for 64 GBd operation," Opt. Express 26, 34580 (2018). R. Ding, Y. Liu, Y. Ma, Y. Yang, Q. Li, A. E.-J. Lim, G.-Q. Lo, K. Bergman, T. Baehr-Jones, and M. Hochberg, "High-Speed Silicon Modulator With Slow-Wave Electrodes and Fully Independent Differential Drive," J. Lightw. Technol. 32, 2240 -- 2247 (2014). M. Streshinsky, R. Ding, Y. Liu, A. Novack, Y. Yang, Y. Ma, X. Tu, E. K. S. Chee, A. E.-J. Lim, P. G.-Q. Lo, T. Baehr-Jones, and M. Hochberg, "Low power 50 Gb/s silicon traveling wave Mach-Zehnder modulator near 1300 nm," Opt. Express 21, 30350 (2013). H. Yu and W. Bogaerts, "An Equivalent Circuit Model of the Traveling Wave Electrode for Carrier- Depletion-Based Silicon Optical Modulators," J. Lightw. Technol. 30, 1602 -- 1609 (2012). R. Ding, Y. Liu, Q. Li, Y. Yang, Y. Ma, K. Padmaraju, A. E.-J. Lim, G.-Q. Lo, K. Bergman, T. Baehr- Jones, and M. Hochberg, "Design and characterization of a 30-GHz bandwidth low-power silicon traveling-wave modulator," Optics Communications 321, 124 -- 133 (2014). Y. Zhou, L. Zhou, H. Zhu, C. Wong, Y. Wen, L. Liu, X. Li, and J. Chen, "Modeling and optimization of a single-drive push -- pull silicon Mach -- Zehnder modulator," Photonics Research 4, 153 (2016). J. Zhou, Q. Zhang, J. Wang, L. Zhu, and J. Hong, "Model and design of silicon photonic carrier-depletion Mach-Zehnder modulators for 400Gb/s and beyond PAM and QAM applications," in Silicon Photonics XIV, G. T. Reed and A. P. Knights, eds. (SPIE, 2019), Vol. 1092318, p. 41. J. Zhou, J. Wang, L. Zhu, and Q. Zhang, "High Baud Rate All-Silicon Photonics Carrier Depletion Modulators," J. Lightw. Technol. 1 -- 1 (2019). D. Patel, S. Ghosh, M. Chagnon, A. Samani, V. Veerasubramanian, M. Osman, and D. V. Plant, "Design, analysis, and transmission system performance of a 41 GHz silicon photonic modulator," Opt. Express 23, 14263 (2015). D. Petousi, L. Zimmermann, A. Gajda, M. Kroh, K. Voigt, G. Winzer, B. Tillack, and K. Petermann, "Analysis of Optical and Electrical Tradeoffs of Traveling-Wave Depletion-Type Si Mach -- Zehnder Modulators for High-Speed Operation," IEEE J. Sel. Topics Quantum Electron. 21, 199 -- 206 (2015). Young Rack Kwon, V. M. Hietala, and K. S. CHamplin, "Quasi-TEM Analysis of "Slow-Wave" Mode Propagation on Coplanar Microstructure MIS Transmission Lines," IEEE Transactions on Microwave Theory and Techniques 35, 545 -- 551 (1987). D. M. Pozar, Microwave Engineering, 4. ed. (Wiley, 2012). N. J. Parsons, A. C. O'Donnell, and K. K. Wong, "Design Of Efficient And Wideband Travelling-Wave Modulators," in Proc. SPIE 0651, Integrated Optical Circuit Engineering III (1986). M. Bass and Optical Society of America., Handbook of Optics (McGraw-Hill, 1995). H. Chung, W. S. C. Chang, and E. L. Adler, "Modeling and optimization of traveling-wave LiNbO3 interferometric modulators," IEEE Journal of Quantum Electron. 27, 608 -- 617 (1991). F. Merget, S. S. Azadeh, J. Mueller, B. Shen, M. P. Nezhad, J. Hauck, and J. Witzens, "Silicon photonics plasma-modulators with advanced transmission line design," Opt. Express 21, 19593 (2013). P. D. Hale and D. F. Williams, "Calibrated measurement of optoelectronic frequency response," IEEE Transactions on Microwave Theory and Techniques 51, 1422 -- 1429 (2003). P. Steglich, C. Villringer, S. Pulwer, M. Casalboni, and S. Schrader, "Design Optimization of Silicon-on- Insulator Slot-Waveguides for Electro-optical Modulators and Biosensors," in Photoptics 2015. Springer Proceedings in Physics, P. Ribero and M. Raposo, eds. (Springer, Cham, 2016). 26 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. W. Heni, Y. Kutuvantavida, C. Haffner, H. Zwickel, C. Kieninger, S. Wolf, M. Lauermann, Y. Fedoryshyn, A. F. Tillack, L. E. Johnson, D. L. Elder, B. H. Robinson, W. Freude, C. Koos, J. Leuthold, and L. R. Dalton, "Silicon-Organic and Plasmonic-Organic Hybrid Photonics," ACS Photonics (2017). J. Degallaix, R. Flaminio, D. Forest, M. Granata, C. Michel, L. Pinard, T. Bertrand, and G. Cagnoli, "Bulk optical absorption of high resistivity silicon at 1550 nm," Optics Letters 38, 2047 (2013). L. Alloatti, M. Lauermann, C. Sürgers, C. Koos, W. Freude, and J. Leuthold, "Optical absorption in silicon layers in the presence of charge inversion/accumulation or ion implantation," Applied Physics Letters 103, 051104 (2013). R. Ding, T. Baehr-Jones, W.-J. Kim, X. Xiong, R. Bojko, J.-M. Fedeli, M. Fournier, and M. Hochberg, "Low-loss strip-loaded slot waveguides in Silicon-on-Insulator," Opt. Express 18, 25061 (2010). W. R. Eisenstadt and Y. Eo, "S-parameter-based IC interconnect transmission line characterization," IEEE Transactions on Components, Hybrids, and Manufacturing Technology 15, 483 -- 490 (1992). 27
1903.02339	1	1903	2019-03-06T12:33:46	Efficient Interfacial Solar Steam Generator with Controlled Macromorphology Derived from Flour via "Dough Figurine" Technology	[ "physics.app-ph" ]	Solar-driven interface steam generator (SISG) is a most promising technology for seawater desalination and wastewater purification. A shape- and size-controlled, low-cost, eco-friendly solar-absorber material is urgently desired for practical application of SISG. Herein, we proposed a facile, sustainable and scalable approach to produce tailored SISG with controlled macromorphology derived from flour via "dough figurine" technology which is originated from the China Han Dynasty. Three kinds of self-floated flour-based absorbers i.e. near-cylindrical (integrated), near-spherical (loose packing) and powdery (dense packing) absorber used as SISGs were discussed, we found that the macromorphology significantly influences water transport and interfacial thermal management of SISG, the integrated absorber has an overwhelming advantage, which possesses a high evaporation efficiency 71.9% at normal solar illumination. The proposed "dough figurine" technology breaks the limitations of the inherent geometry of reported biomass based SISG, which provides an important guidance for SISG use in remote and impoverished areas.	physics.app-ph	physics	Efficient Interfacial Solar Steam Generator with Controlled Macromorphology Derived from Flour via "Dough Figurine" Technology Zhengtong Li1, Chengbing Wang1,, Zeyu Li1, Lin Deng1, Jinbu Su1, Jing Shi2, Meng An2, 1School of Materials Science and Engineering, Shaanxi Key Laboratory of Green Preparation and Functionalization for Inorganic Material, Shaanxi University of Science and Technology, Xi'an, Shaanxi 710021, China 2College of Mechanical and Electrical Engineering, Shaanxi University of Science and Technology, Xi'an 710021, China * To whom correspondence should be addressed. E-mail: [email protected] (C. Wang), [email protected] (M. An) 1 Abstract Solar-driven interface steam generator (SISG) is a most promising technology for seawater desalination and wastewater purification. A shape- and size-controlled, low- cost, eco-friendly solar-absorber material is urgently desired for practical application of SISG. Herein, we proposed a facile, sustainable and scalable approach to produce tailored SISG with controlled macromorphology derived from flour via "dough figurine" technology which is originated from the China Han Dynasty. Three kinds of self-floated flour-based absorbers i.e. near-cylindrical (integrated), near-spherical (loose packing) and powdery (dense packing) absorber used as SISGs were discussed, we found that the macromorphology significantly influences water transport and interfacial thermal management of SISG, the integrated absorber has an overwhelming advantage, which possesses a high evaporation efficiency 71.9% at normal solar illumination. The proposed "dough figurine" technology breaks the limitations of the inherent geometry of reported biomass based SISG, which provides an important guidance for SISG use in remote and impoverished areas. Keyword: flour, interfacial solar steam generation, stacking forms, dough figurine 2 1. Introduction The discovery of solar-driven interfacial steam generation (SISG) with the ultrahigh performance evaporation has sparked a wave of vigor and excitement due to its critical role in ubiquitous related applications, such as seawater desalination,1, 2 power generation3-5 and steam sterilization6 and so on. Indeed, SISG overcomes many inherent obstacles in conventional solar-driven steam generation, i.e. the efficiency of the solar energy receiver, thermal management capabilities, high production costs and complex preparation processes.7-9 Recently, various artificial materials of SISG equipment have been designed based on different novel mechanisms, such as plasmonic metal particles absorber,10-12 carbon-based materials absorber,13-15 and heat confinement layers16, 17. For example, in such plasmonic absorber, the plasmonic noble metal particles are deposited on the alumina template (AAO) or wood blocks, where the plasmon effect of plasmon particle and light traps induced by porous substrate (e.g. AAO, wood blocks) are utilized to enhance the absorption of solar energy.10, 11, 18, 19 Such porous substrates can transport bulk water to the photo-thermal conversion interface by capillary force as well as provide the channel of steam overflow. Meantime, their low thermal conductivity effectively can reduce the heat conduction loss. In analogy, large amounts of porous substrates combined with traditional synthetic carbon materials are also developed as efficient solar distillation equipment, such as graphene,20 reduced graphene oxide,21, 22 graphene alkyne,23 carbon nanotubes,24 carbon power.25 However, previous SISG materials with high efficiency steam evaporation often entail complex fabrication processes, high cost and difficult to recycle, which limits their application domain. Therefore, the challenge of developing a cost-effective, robust, environmentally friendly, SISG material remains to obtain a high photo-thermal conversion efficiency under ambient solar illumination. 3 Recently, biomass materials with natural ubiquitous micro- and macrostructures, such as mushroom,26 lotus 27, cotton,28 wood,29 daikon,30 louts leaf,31 are developed as efficient solar steam generation device by a series of simple processing i.e. cutting, freeze drying and high-temperature carbonization (Figure S1). Some simple treatments for biomass materials have produced unparalleled advantages for synthetic materials (Table S1 and S2).32 For example, mushroom with an umbrella-shaped structure (i.e. positive solid cone geometric structure) can possess a high evaporation efficiency 78% under 1 sun irradiation.26 Natural wood is constructed as a high performance of SISG system. The evaporation efficiency of the carbonization poplar wood-based SISG device can reach 86.7% at 10 kW m-2, which is a fascinating discover for the biomass materials application in SISG system.33 Due to different natural mesoporous structures in natural wood, it is also used to investigate the effect of water transport and heat transfer on evaporation efficiency. The carbonized lotus seedpods with a high a evaporation efficiency (86.5% @1 SUN) mainly relied on their unique macroscopic cone shape and hierarchical mesopores and macropore structures.27 Moldy bread, carbonized by a barbecue-like method, achieves 71.4% evaporation efficiency in a high humidity level (70%).34 In our recent study, a kind of arched structure bamboo charcoal is developed as a high-performance absorber, which possess a larger area for absorbing sunlight and a smaller area for water absorption.1 These studies of biomass materials- based SISG can provide a marvelous opportunity to lower cost and high evaporation efficiency. However, it is quite difficult to make a large-scale fabrication due to the inherent geometry structures of biomass materials. Therefore, a shape- and size- controlled, low-cost, eco-friendly biomass-based absorber is urgently desired for practical application of SISG Flour, prepared by mechanical grinding of wheat, is the main raw material for 4 cooking steamed bread and noodles. The production of wheat all over the world is very huge. Plenty of expired flour (> 10 million tons in China every year) is generated due to excessive production.35-38 Therefore, it is desirable to explore novel value-added application. After two basic processes i.e. shaping and dehydration, four can be made as several typical macrostructures (Figure S2) with the large surface area beneficial for solar light absorption, as well as inner porous structures to achieve a better performance of water transport. After carbonization, the solar absorption capacity of flour-based absorber with near-columnar is greatly improved to 92.4% across the entire solar spectrum without affecting its mechanical properties. During preparation, only commercial deionized water without any expensive and toxic chemicals is added as a good binder/templating agent. Inspired by Chinese traditional craft, called "Dough Figurine (DF)" in the Han Dynasty, the flour mixing with water can form dough. Through several simple operations such as pinching, rubbing, pushing, etc., various macro geometries can be prepared. More importantly, the macroscopic morphology of the flour-based materials can preserve without any significant changes after the dehydration and carbonization. Therefore, these advantages of flour indicate its promising as a candidate material for solar absorber and searching for a high- performance evaporation structure. In this study, we utilized flour to create solar-driven interface steam conversion equipment. Herein, inspired by the craft of DF, we prepared flour block with different shapes (near-columnar, cup-shaped, umbrella-shaped and near-spherical) which were freeze- dried and carbonized at a lower temperature (300 °C). Porous carbon foam can be prepared easily, which satisfies all essential elements for interfacial steam generation system. In details, it can float freely on the water surface due to its lower mass density and the self-floating time can keep for a long time (more than 720 hours in seawater). 5 Meanwhile, it has high light absorbance as high as 92.4% from 200 nm to 2500 nm. The efficient water transport (carbonized flour with a rich pore structure) can be achieved by its rich pores and hydrophilic. We firstly discussed in detail the impact of different stacking forms of the same materials on optical absorption, water transport, and thermal management of SISG, we found that the macromorphology significantly influences the evaporation efficiency of SISG, the integrated absorber has an overwhelming advantage, which possesses a high evaporation efficiency 71.9% at normal solar illumination. The proposed "dough figurine" technology breaks the limitations of the inherent geometry of reported biomass based SISG, which provides an important guidance for SISG use in remote and impoverished areas. 2. Result and discussion Similar with the common morphology of solar absorber in SISG system, the carbonized flour-based absorber with near-cylindrical (NC) geometry is fabricated and used to study the evaporation performance of SISG system.10, 14, 15, 23, 39 Figure 1 shows schematic diagram of making flour-based solar absorber. As shown in Figure 1(a) and 1(b), flour is a kind of biomass material obtained by mechanical grinding of wheat. The short growth cycle (one or two times a year) and large yield (128.8106 t in China 2016)40 of wheat show an overwhelming advantage over that of other biomass materials, which suggests a large-scale flour can be produced in a factory with low cost. In addition, plenty of flour can't keep for a long time due to some environmental factors, such as moisture, pests, etc. Herein, we utilized flour, an excellent candidate for biomass carbon sources, to create SISG equipment. Propelled by the Chinese traditional method called "DF" craft, flour-based absorbers with different geometries can be prepared (The details of preparation are shown in the experimental section). Figure 1(c) 6 and 1(d) show macro model of the flour block sample with NC geometry before (FNC) and after (FFNC) freeze-drying process, where large amounts of nanopores form by adding and removing of water. It is obviously found that tight starch granules piled together on the surface of FFNC. The FFNC is carbonized under an air environment at 300 °C, which is more available compared with that of previously reported studies (Table S1). The easy access to prepare FFNC sample (low carbonization temperature and no gas protection during carbonization) is more conducive to large-scale fabrication of SISG equipment. Figure 1. Schematic of preparation process of near cylindrical solar absorber based flour block. (a) Natural wheat. (b) Commercially available flour. (c) Preparation of flour block with near-cylindrical (FNC) structure using dough figurine technology. (d) Freeze-dried (FFNC) sample. (e) The carbonization process of freeze-dried (C-FFNC) sample. (f) The schematic picture of solar-driven interface steam generation (SISG) 7 device. The inset picture: the localized evaporation of C-FFNC can be enhanced. (g) The infrared photo of SISG device under one sun irradiation at 1min. The carbonized freeze-drying flour block with near-cylinder structure, simplified as C- FFNC, is shown Figure 1e. During the process of fabrication, C-FFNC possesses a large mechanical strength (Figure S3) due to the absence of activator (such as Na2CO3, K2CO3, and KOH). Without plenty of other gas (CO2 or H2) formation, the inner wall of the tunnel is kept thick35, 36 (Figure 2c). The C-FFNC sample as SISG sample can form localized evaporation (Figure 1f and inset picture), where water can be pumped spontaneously through its micro-scale channels under the synergistic assistance of capillary force and the absorptive nature of the hydrophilic starch and the vapor for fresh water can be generated under the nature solar light illumination.27 As shown in Figure 1g, the infrared image of the SISG device under one sun radiation in one minute, the localized hot region can be obviously observed. The microstructure plays a critical role on the water transport and solar absorption of the SISG system.41 Figure 2a shows the outer surface of FFNC sample. It consists of starch granules with different sizes. After carbonization, multi-scale holes are formed on inner and outer surface of C-FFNC samples shown in Figure 2b and 2c. In addition, the porosity of samples is evaluated before and after carbonization using a mercury intrusion instrument. Based on the measured diameter of pores, the average values of the pore diameters are analyzed (Figure S4). The pores diameter of FFNC samples are mainly micro and mesopores (≤50nm) which are induced by dehydration in the freeze- drying process. In carbonization, the pore structures of C-FFNC samples collapses and reorganizes due to the burning of organics (Figure S5). Therefore, the pore structures 8 with meso- and macrospores (≥50 nm) are dominated. The porosity is decreased from 80.3% to 59.8% while the corresponding average pore diameter is increased from 5.4 nm to 173.0 nm. The macro-pores surface of C-FFNC (shown in Figure 1c) provide the channel for the generated steam to escape. This prevents the generated steam from cooling in the original position, thereby reducing the evaporation efficiency. The inset structure of C-FFNC (Figure 1d) showed channels with different pore diameters, which can transport water from the bulk water to the hot region by capillary force. Further, these internal tunnel structures provide a rich heat exchange area that effectively reduces heat loss to the bulk water. The thermal conductivity of FFNC and C-FFNC are 0.12 W/m-K and 0.09 W/m-K (Figure S6), respectively. Figure 2. (a) SEM image of the surface of freeze-dried samples. (b) SEM image of the surface of C-FFNC sample. (c) SEM image of the inner structure of C-FFNC sample. (d) Porosity data statistics for FFNC sample and C-FFNC sample. (e) The XRD 9 diffraction pattern of C-FFNC. (f) Raman spectra of C-FFNC fitted by Gaussian functions. (g) The XPS spectra of FFNC and C-FFNC. Inset picture: contact angle test of C-FFNC. (h) The FTIR spectra of FFNC and C-FFNC. (i) Solar irradiance (AM 1.5 G) (blue, left side axis) and reflection (right side axis) of FFNC and C-FFNC. The water transport of C-FFNC can be enhanced due to its hydrophilicity originating from its special components and carbonization condition. Flour, 17% protein and 25% carbohydrate as the main components, it is easily bonded by stirring after adding deionized water. Also, both carbohydrates and proteins have good hydrophilicity, so the formed dough also has good hydrophilicity which is conducive to the transfer of water from bulk water to air-liquid interface. The degree of carbonization is measured by X- ray diffractometry (XRD). As shown in Figure 2e, two broad diffractions were observed at the 2-theta of the degree of 21° and 40°, which suggests that the main structure in the C-FFNC sample is amorphous carbon. The stronger peak of 21° is a typical graphene reflection, indicating the presence of a majority graphene-like structure in the C-FFNC sample. And it can be proved by Raman spectra (Figure 2f). In the Raman spectra of C-FFNC, the D-band with disorder amorphous carbon and G- band with sp2 vibration of graphite crystal can be observed at 1361 cm-1 and 1560cm-1. In order to accurately analyze the data, we performed a sub-peak fitting of the measured Raman spectra, the ID/IG=2.4 what can prove that the C-FFNC sample have a higher degree of amorphous carbon.27, 42, 43 In addition, we further explored surface chemical composition of C-FFNC by X-ray photoelectron spectroscopy (XPS) and Fourier transformed infrared spectroscopy (FTIR). The XPS spectrum of the sample before (FFNC) and after (C-FFNC) carbonization is shown in the Figure 2g, the main elements of the two including carbon, nitrogen, oxygen. The relative contents of C1s, N1s and 10 O1s of FFNC are 55.39%, 2.49%, 42.15%, respectively. After carbonization, the relative content of N 1s of C-FFNC is increased to 2.67%, and corresponding to C-N/N- H. As the FTIR (Figure 2h) show, the C-N/N-H functional group can still be detected at the 1135cm-1 and 3435cm-1. The C-FFNC has excellent hydrophilicity and can be further confirmed by the contact angle test. The contact angle of its surface is about 34.6° (inset picture of Figure 2g) which contribute its surface chemistry elements and its rough texture. As it is mentioned above, C-FFNC can ensure that it floats on the water surface to transport a sufficient amount of water from the bulk water to the upper surface of the absorber due to its rich pore structures and hydrophilicity.32 Besides, excellent optical absorption and effective heat management are critical to enhance evaporation efficiency. Therefore, it is very meaningful to further examine optical capabilities and thermal management. As shown in Figure 2i, the sample before carbonization, the FFNC has a very high reflection in the visible light band. After carbonization, the color of the sample turned black and the reflection was greatly reduced. Based on the formula: α= ∞ ∫ 𝐴(𝜆)[1−𝑅(𝜆)]𝑑𝜆 0 ∞ ∫ 𝐴 0 (𝜆)𝑑𝜆 where A(λ) is the solar radiant energy density at different wavelengths of AM 1.5 at atmospheric mass. R is the reflectance of the solar absorber at different wavelengths that is usually taken from 0.2 to 2.5μm. Therefore, the absorbance of C-FFNC sample is calculated about 92.4% and even it can reach 95.5% in the visible range (400-760 nm). As mentioned earlier, the stacking way of SISG devices is also an important topic that has to be considered (i.e. under the same conditions: whether an integrated material works well compared with the stacked multiple materials in a certain way.). So we carbonized samples of different shapes (i.e. near-cylindrical (integrated), near-spherical 11 (loose packing) and powdery (dense packing)) were prepared to examine the placement effect of evaporation efficiency. Integrated samples, loosely packed samples, and densely packed samples are three common and important ways of stacking (Figure 4a). The C-FFNC is a sample of the monolithic material. Then we prepared some near spherical (NS) samples with a diameter of ~1 cm, and obtained loosely packed samples after freeze-drying, carbonization (i.e. called C-FFNS). Further, we pulverize some C- FFNS samples and pass the standard filtration to obtain a powder particle size of ~450μm as dense packed samples (i.e. called C-FFP). Unlike nanograde samples stacks, these three kinds of macro samples stacking are similar in terms of optical absorption, hydrophilicity and mechanical strength.44, 45 We conclude that important factors affecting the evaporation efficiency of these three different stacking materials are their different thermal management. The mass changes of C-FFNC, C-FFNS and C-FFP have large differences (Figure 3a). To systematically evaluate the evaporation performance of C-FFNC, the mass change of water under normal solar illumination (1 sun) was recorded (Figure 3a). The evaporation rate and efficiency are 1.0 kg m-2 h-1 and 71.9%, respectively, which is much higher than that of pure water. Such high evaporation efficiency of C-FFNC mainly is explored based on the formula of evaporation rate, the net evaporation rate 𝑚 can be expressed as 𝑚 h𝑓𝑔 = Aαq𝑠𝑜𝑙𝑎𝑟 − Aεσ(𝑇4 − 𝑇𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡 4 ) −Ah(𝑇 − 𝑇𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡) − Aq𝑤𝑎𝑡𝑒𝑟 (1) where h𝑓𝑔 is the latent heat, A is surface area of the absorber facing the sun, α is the solar absorption, q𝑠𝑜𝑙𝑎𝑟 is the solar flux, ε is the emittance of the absorbing surface, σ is the Stefan-Boltzmann constant (i.e., 5.67 × 10-8 Wm-2K-4),46 T is the surface temperature of the absorber, Tenvironment is the temperature of the adjacent environment, h is the convection heat transfer coefficient, and q𝑤𝑎𝑡𝑒𝑟 is the heat flux to the 12 underlying water, including conduction and radiation. The second and the third term on the right side of Eq. (1) denotes the heat loss to ambient by heat radiation and convection. According to Eq. (1), the evaporation rate can be enhanced by increasing heat energy absorption and decreasing heat loss (heat convection, heat radiation, and heat conduction). For C-FFNC absorber, the solar light trap formed by the surface pores and the carbonized nature with black color can form a high optical absorption (95.5% in the visible range), which provide sufficient solar energy source. The thermal conductivity of flour-based carbon foam is 0.09 W/m-K, one order of magnitude lower than that of pure water (0.60 W/m-K at room temperature).47, 48 The low thermal conductivity of C- FFNC enables the captured solar energy to be confined to the air-liquid surface, which reduces the heat conduction loss to the bulk water and promote the formation of localized hot area as shown in Figure 1(g). Moreover, the rich pore structures (porosity 59.79 %) and hydrophilicity of C-FFNC facilitates the water transport from bulk water transport to air-liquid interface. In addition, the previous studies suggested that the enthalpy of confined water is reduced compared with that of pure water. In C-FFNC absorber, the rich micro-scale pores of C-FFNC facilitates the reduced vaporization enthalpy, further enhancing evaporation rate. Admittedly, precisely controlling the size and shape of SISG device, especially at a harsh environment, is challenging and subject to large uncertainties. Interestingly, our proposed flour-based absorber can be prepared into different geometries (shown in Figure S2), which promises to expand applications of SISG device and making existing applications more advantageous and cost-effective. Moreover, such advantage of flour- based absorber make flour as an optimal candidate material to explore the placement effect of absorber on evaporation rate. To systematically analyze the evaporation 13 performance of flour-based absorber with stack ways, the mass change of water under normal solar illumination (1 sun) was recorded (Figure 3a). The mass change of C- FFNC, C-FFNS, and C-FFP are 0.54 kg/m2, 0.41 kg/m2 and 0.34 kg/m2 in half an hour, respectively, which are 2.3, 1.7, and 1.5 times as high as that of pure water (Figure 3a). The corresponding evaporation rates are 1.0 kg m-2 h-1, 0.74 kg m-2 h-1 and 0.62 kg m-2 h-1, respectively. The evaporation efficiency (η) is defined as below: η= ℎ𝐿𝑉𝑚 𝐶𝑜𝑝𝑡𝑞𝑖 (2) where m represents the evaporation rate (kg m-2h-1) after the subtraction of the evaporation rate without light illumination, hLV is the total enthalpy of the liquid-vapor phase change (sensible heat + phase-change enthalpy), Copt represents the optical concentration and qi represents the nominal solar illumination (1 kW m-2). Figure 3. (a) Mass change of C-FFNC, C-FFNS, and C-FFP as well as water in the dark field and under 1 sun illumination. (b) The evaporation rate (left side axis) and evaporation efficiency (right side axis) of C-FFNC, C-FFNS and C-FFP. The calculated energy efficiencies of evaporation for C-FFNC, C-FFNS, and C- FFP are presented in Figure 3b. It is obviously observed that the energy efficiency of C-FFNC can reach 71.9%, which are much higher than those of C-FFNS (~ 53.1%) and C-FFP (~ 44.7%). In other words, the C-FFNC absorber is the most effective for enhancing the energy efficiency among three kinds of samples. We will explore the 14 potential mechanism for this trend from three main factors: heat localization, heat loss and water transport. Figure 4. (a) The optical image and the infrared image of the C-FFNC/ C-FFNS/ C- FFP sample surface in 1 min under one sun irradiation. The radius of three samples are mainly about 5 cm, 1 cm, 425 μm for C-FFNC, C-FFNS and C-FFP, respectively. (b) Surface temperature statistical point distribution image of the (a) infrared images. 15 The heat transfer behavior of flour-based absorber with different sample diameters is systematically evaluated (Figure 4). The surface temperatures of C-FFNC, C-FFNS and C-FFP under 1 sun illumination are carefully measured via an IR camera. Figure 4 (a) plots the surface temperature image from IR camera, from which we calculated the temperature distribution of sample surfaces. Obviously, the average steady-state temperatures of the C-FFNC, C-FFNC, C-FFNS and C-FFP under 1 sun illumination reach 49.9 ℃, 41.2 ℃ and 44.4 ℃, respectively (Figure 4b). These surface temperature are much higher that of pure water (23 ℃). Compared with thermal conductivity (0.09 W/m-K) of carbonized flour-based materials (the details are shown in Figure S4), the larger thermal conductivity of bulk water results in a considerable proportion of the absorbed energy for increasing the underlying water temperature instead of evaporation, which reduces the energy efficiency. More interestingly, the surface temperature of C- FFNC is higher than those of C-FFNS and C-FFP samples. Although the intrinsic porosity among samples with different diameters are close, such stacking with different samples diameters leads to this decreasing trend of water content inside carbonized floured-based absorber: C-FFP, C-FFNS, C-FFNC. The more water, the more the absorbed energy is used to heat underlying water. Therefore, the heat localization by for adiabatic material (C-FFNC), only the water on C-FFNC sample surface is heated and the underlying water will not cause large heat conduction loss. That is, almost all the heat is localized at the air-liquid interfacial, which results in the high surface temperature and fast evaporation. 16 Figure 5. (a) The schematic diagram of thermal resistance between solid-liquid interfaces for samples floating on the bulk water. (b) The wetting process of C-FFNC, C-FFNS and C-FFP at 0s and 120s. To further understand the high performance of evaporation of C-FFNC, the heat energy dissipated by interfacial thermal resistance is qualitatively analyzed.49-51 The total thermal resistance of the floured-based absorber including flour-based materials and inside water is defined by introduce the interfacial thermal resistance Ri shown in Figure 5(a), 𝑅𝑡 = 𝑅𝑤 + 𝑅𝑠 + 𝑅𝑖 (3) Here, Rw and Rs are the intrinsic thermal resistance of bulk water and flour-based sample. Ri is the interfacial thermal resistance between water and flour-based sample, which is related to the interfacial properties of interfacial component materials and 17 interface area. According to Eq. (3), among three samples (i.e., C-FFNC, C-FFNS and C-FFP), the interfacial properties are almost the same while the total interface area are obviously different due to different diameters of flour-based absorber. The diameters of C-FFNC, C-FFNS and C-FFP are about 5.0 cm, 1.2 cm, and 1.5 × 10-4 cm, respectively. Specifically, the ratio of their interface areas is 1: 4.16: 3.32 × 104. Thus, the heat energy inside C-FFP absorber is severely dissipated by interfacial thermal resistance, leading to a low evaporation rate. As shown in figure 5b, we explore the water transporting capability of samples with different stacking ways. C-FFNC, C-FFNS, and C-FFP samples are heated in oven and then are put into water directly under the lab environment (detail in supporting information). The cooling of the sample is mainly cooled by water nature evaporation. The faster the cooling rate, the faster the water transport capacity. Experiments show that water transport capacity decreases from C-FFP to C-FFNS to C-FFNC in turn. As aforementioned, the evaporation efficiency (E. E.), thermal resistance (H. R.), water transportation (W. T.) relationships are: E.E.C-FFNC >E.E.C-FFNS >E.E.C-FFP. H.R.C-FFNC >H.R.C-FFNS >H.R.C-FFP. W.T.C-FFNC <W.T.C-FFNS <W.T.C-FFP. Since the three samples are of the same material, the wettability, porosity, and optical absorption capacity (Figure S7) are very close. Their difference in evaporation efficiency is mainly caused by the different heat resistance and water transport. The relationship between the speed trend of water transport (Figure 5b, details in Figure S8) and the evaporation efficiency is linearly negatively correlated, which further indicates that the thermal resistance is the main factor determining the different placement modes of the samples. 18 3. Conclusion Inspired by the technique of dough figurine, we developed a new method for designing biomass-based SISG with controlled micromorphology derived from flour. The self- floated carbonized flour-based absorber with excellent mechanical properties, high efficiency of solar absorption, low thermal conductivity and rich pore structure, which fulfill all the require of SISG for efficient localized evaporation. This flour-based SISG with the common structure (near-cylindrical) can achieve a high evaporation efficiency about 71.93% at normal solar illumination. Furthermore, we first investigated the effects of different macroscopic accumulations of uniform absorber on evaporation efficiency, and explored in detail the effects of different cumulative modes of optical absorption, water transport, and thermal management. This study provides an important guidance for the large-scale use of SISG device in remote and impoverished areas. 4. Experiment Section The preparation process of flour dough with specific shape: The water content and mixing time play a critical role for ensuring the similar micro- structure of the samples. The flour (purchased from local supermarket) and deionized water is mixed in a ratio of 2:1. The time of mixing-process is about 15 minutes. According to the craft of DF, these samples with different geometric shapes (near- columnar, umbrella-shaped and near-spherical) were prepared (the details are shown in Figure S2) with hand. The preparation process of absorber material: Here, we take flour dough with near-columnar (FNC) as an example to introduce the process of carbonization. First, FNC sample is frozen at 0 °C in the refrigerator for 10 hours. Then, the frozen FNC sample is dehydrated in a freeze dryer for 24 hours. Finally, the freeze-dried sample (FFNC) is carbonized in a muffle furnace. The carbonization is 19 conducted under the condition of heating to 300 °C for two hours at 5 °C/min. The carbonized samples (C-FFNC) can be used as SISG device directly. Materials Characterization: The flour-based sample morphology was characterized via a high-resolution field- emission scanning electron microscope (SEM, FEI Verios 460 USA), Environmental scanning electron microscope (FEI Q45+EDAX Octane Prime), and the contact angle measurement through video optical contact angle measurement with measurement accuracy of 0.1°. The characteristic spectral reflection peak of sample was mainly measured by an ultraviolet-visible-near-infrared spectrophotometer (Cary 5000). The sample surface temperature was measured via an IR camera (Fotric) with model LS5.5B4−0038 lens and FTIR spectra were obtained on a VERTEX 70 spectrometer. The porosity was measured by fully automatic mercury porosimeter (AutoPore IV 9510). The Phase testing was performed via an X-ray diffractometer (XRD, D8 advance, Bruker, Germany). The Raman spectra were measured via the micro-Raman spectrometer (Renishaw inVia Reflex, UK) with a CCD detector at room temperature. The elemental valence states were determined by X-ray photoelectron spectroscopy (XPS, ThermoFisher ESCALAB 250Xi, monochromatic Al Kα. Thermal Decomposition and Metamorphic Test Method Synchronous Thermal Analysis TG- DSC (TA, NETZSCH). Thermal conductivity of samples are measured using hot disk method (TPS 2500S), Standard sieve for powder filtration (US standard). Solar steam generation experiments. The C-FFNC was placed in a beaker rolled by commercial expanded polyethylene in order to reduce the ambient effect on experimental results. The salt water (3.5 wt% of salt, which is the same salt concentration as seawater) is placed in the beaker. The samples were illuminated under one-sun intensity with a xenon lamp (CEL-HXF300, 20 AM1.5 filter). The mass during the evaporation process was measured by an electron microbalance (AR224CN) with an accuracy of 0.0001 g and the data were transmitted to the personal computer (PC). The room temperature and humidity during the tests were controlled at 24°C and 30%, respectively. Supporting Information. The Supporting Information is available free of charge on the ACS Publication Website. Notes The authors declare no competing financial interest. Acknowledgement This work is financially supported by the National Natural Science Foundation of China (Nos. 51562020 and 51575253) and Natural Science Research Start-up Fund of Shaanxi University of Science and Technology (2018GBJ-10). We are grateful to Guilong Peng, Xiaoxiang Yu for useful discussions. 21 Reference 1 Li, Z.; Wang, C.; Lei, T.; Ma, H.; Su, J.; Ling, S.; Wang, W., Arched Bamboo Charcoal as Interfacial Solar Steam Generation Integrative Device with Enhanced Water Purification Capacity. Advanced Sustainable Systems 2019, 3 (1), 1800144. 2 Peng, G.; Ding, H.; Sharshir, S. W.; Li, X.; Liu, H.; Ma, D.; Wu, L.; Zang, J.; Liu, H.; Yu, W. J. A. T. E., Low-cost high-efficiency solar steam generator by combining thin film evaporation and heat localization: Both experimental and theoretical study. Applied Thermal Engineering 2018, 143, 1079-1084. 3 Zhu, L.; Gao, M.; Peh, C. K. N.; Wang, X.; Ho, G. W., Self‐Contained Monolithic Carbon Sponges for Solar ‐ Driven Interfacial Water Evaporation Distillation and Electricity Generation. Advanced Energy Materials 2018, 8 (16), 1702149. 4 Xue, G.; Xu, Y.; Ding, T.; Li, J.; Yin, J.; Fei, W.; Cao, Y.; Yu, J.; Yuan, L.; Gong, L.; Chen, J.; Deng, S.; Zhou, J.; Guo, W., Water-evaporation-induced electricity with nanostructured carbon materials. Nature Nanotechnology 2017, 12 (4), 317-321. 5 Hou, B.; Cui, Z.; Zhu, X.; Liu, X.; Wang, G.; Wang, J.; Mei, T.; Li, J.; Wang, X., Functionalized carbon materials for efficient solar steam and electricity generation. Materials Chemistry and Physics 2019, 222, 159-164. 6 Li, J.; Du, M.; Lv, G.; Zhou, L.; Li, X.; Bertoluzzi, L.; Liu, C.; Zhu, S.; Zhu, J., Interfacial Solar Steam Generation Enables Fast-Responsive, Energy-Efficient, and Low-Cost Off-Grid Sterilization. Advanced Materials 2018, 30 (49), e1805159. 7 Deng, Z.; Zhou, J.; Miao, L.; Liu, C.; Peng, Y.; Sun, L.; Tanemura, S., The emergence of solar thermal utilization: solar-driven steam generation. Journal of Materials 22 Chemistry A 2017, 5 (17), 7691-7709. 8 Wang, Z.; Liu, Y.; Tao, P.; Shen, Q.; Yi, N.; Zhang, F.; Liu, Q.; Song, C.; Zhang, D.; Shang, W.; Deng, T., Bio-inspired evaporation through plasmonic film of nanoparticles at the air-water interface. Small 2014, 10 (16), 3234-9. 9 Sharshir, S. W.; Peng, G.; Elsheikh, A.; Edreis, E. M.; Eltawil, M. A.; Abdelhamid, T.; Kabeel, A.; Zang, J.; Yang, N. J. E. C.; Management, Energy and exergy analysis of solar stills with micro/nano particles: a comparative study. Energy Conversion and Management 2018, 177, 363-375. 10 Zhu, M.; Li, Y.; Chen, F.; Zhu, X.; Dai, J.; Li, Y.; Yang, Z.; Yan, X.; Song, J.; Wang, Y.; Hitz, E.; Luo, W.; Lu, M.; Yang, B.; Hu, L., Plasmonic Wood for High-Efficiency Solar Steam Generation. Advanced Energy Materials 2018, 8 (4), 1701028. 11 Zhou, L.; Tan, Y.; Wang, J.; Xu, W.; Yuan, Y.; Cai, W.; Zhu, S.; Zhu, J., 3D self- assembly of aluminium nanoparticles for plasmon-enhanced solar desalination. Nature Photonics 2016, 10 (6), 393-398. 12 Wang, X.; He, Y.; Liu, X.; Shi, L.; Zhu, J., Investigation of photothermal heating enabled by plasmonic nanofluids for direct solar steam generation. Solar Energy 2017, 157, 35-46. 13 Wang, Y.; Wang, C.; Song, X.; Megarajan, S. K.; Jiang, H., A facile nanocomposite strategy to fabricate a rGO -- MWCNT photothermal layer for efficient water evaporation. Journal of Materials Chemistry A 2018, 6 (3), 963-971. 14 Li, X.; Xu, W.; Tang, M.; Zhou, L.; Zhu, B.; Zhu, S.; Zhu, J., Graphene oxide-based efficient and scalable solar desalination under one sun with a confined 2D water path. 23 Proceedings of the National Academy of Sciences of the United States of America 2016, 113 (49), 13953-13958. 15 Hu, X.; Xu, W.; Zhou, L.; Tan, Y.; Wang, Y.; Zhu, S.; Zhu, J., Tailoring Graphene Oxide-Based Aerogels for Efficient Solar Steam Generation under One Sun. Advanced Materials 2017, 29 (5). 16 Ghasemi, H.; Ni, G.; Marconnet, A. M.; Loomis, J.; Yerci, S.; Miljkovic, N.; Chen, G., Solar steam generation by heat localization. Nature Communication 2014, 5, 4449. 17 Cooper, T. A.; Zandavi, S. H.; Ni, G. W.; Tsurimaki, Y.; Huang, Y.; Boriskina, S. V.; Chen, G., Contactless steam generation and superheating under one sun illumination. Nature Communication 2018, 9 (1), 5086. 18 Zhou, L.; Zhuang, S.; He, C.; Tan, Y.; Wang, Z.; Zhu, J., Self-assembled spectrum selective plasmonic absorbers with tunable bandwidth for solar energy conversion. Nano Energy 2017, 32, 195-200. 19 Chen, C.; Zhou, L.; Yu, J.; Wang, Y.; Nie, S.; Zhu, S.; Zhu, J., Dual functional asymmetric plasmonic structures for solar water purification and pollution detection. Nano Energy 2018, 51, 451-456. 20 Ito, Y.; Tanabe, Y.; Han, J.; Fujita, T.; Tanigaki, K.; Chen, M., Multifunctional Porous Graphene for High-Efficiency Steam Generation by Heat Localization. Advanced Materials 2015, 27 (29), 4302-7. 21 Zhang, Y.; Zhao, D.; Yu, F.; Yang, C.; Lou, J.; Liu, Y.; Chen, Y.; Wang, Z.; Tao, P.; Shang, W.; Wu, J.; Song, C.; Deng, T., Floating rGO-based black membranes for solar driven sterilization. Nanoscale 2017, 9 (48), 19384-19389. 24 22 Wang, G.; Fu, Y.; Guo, A.; Mei, T.; Wang, J.; Li, J.; Wang, X., Reduced Graphene Oxide -- Polyurethane Nanocomposite Foam as a Reusable Photoreceiver for Efficient Solar Steam Generation. Chemistry of Materials 2017, 29 (13), 5629-5635. 23 Gao, X.; Ren, H.; Zhou, J.; Du, R.; Yin, C.; Liu, R.; Peng, H.; Tong, L.; Liu, Z.; Zhang, J., Synthesis of Hierarchical Graphdiyne-Based Architecture for Efficient Solar Steam Generation. Chemistry of Materials 2017, 29 (14), 5777-5781. 24 Jiang, F.; Liu, H.; Li, Y.; Kuang, Y.; Xu, X.; Chen, C.; Huang, H.; Jia, C.; Zhao, X.; Hitz, E.; Zhou, Y.; Yang, R.; Cui, L.; Hu, L., Lightweight, Mesoporous, and Highly Absorptive All-Nanofiber Aerogel for Efficient Solar Steam Generation. ACS Applied Material Interfaces 2018, 10 (1), 1104-1112. 25 Liu, Z.; Song, H.; Ji, D.; Li, C.; Cheney, A.; Liu, Y.; Zhang, N.; Zeng, X.; Chen, B.; Gao, J.; Li, Y.; Liu, X.; Aga, D.; Jiang, S.; Yu, Z.; Gan, Q., Extremely Cost-Effective and Efficient Solar Vapor Generation under Nonconcentrated Illumination Using Thermally Isolated Black Paper. Global Challenge 2017, 1 (2), 1600003. 26 Xu, N.; Hu, X.; Xu, W.; Li, X.; Zhou, L.; Zhu, S.; Zhu, J., Mushrooms as Efficient Solar Steam-Generation Devices. Advanced Materials 2017, 29 (28). 27 Fang, J.; Liu, J.; Gu, J.; Liu, Q.; Zhang, W.; Su, H.; Zhang, D., Hierarchical Porous Carbonized Lotus Seedpods for Highly Efficient Solar Steam Generation. Chemistry of Materials 2018, 30 (18), 6217-6221. 28 Wu, X.; Wu, L.; Tan, J.; Chen, G. Y.; Owens, G.; Xu, H., Evaporation above a bulk water surface using an oil lamp inspired highly efficient solar-steam generation strategy. Journal of Materials Chemistry A 2018, 6 (26), 12267-12274. 25 29 Xue, G.; Liu, K.; Chen, Q.; Yang, P.; Li, J.; Ding, T.; Duan, J.; Qi, B.; Zhou, J., Robust and Low-Cost Flame-Treated Wood for High-Performance Solar Steam Generation. ACS Applied Material Interfaces 2017, 9 (17), 15052-15057. 30 Zhu, M.; Yu, J.; Ma, C.; Zhang, C.; Wu, D.; Zhu, H., Carbonized daikon for high efficient solar steam generation. Solar Energy Materials and Solar Cells 2019, 191, 83- 90. 31 Liao, Y.; Chen, J.; Zhang, D.; Wang, X.; Yuan, B.; Deng, P.; Li, F.; Zhang, H., Lotus leaf as solar water evaporation devices. Materials Letters 2019, 240, 92-95. 32 Li, Z.; Wang, C.; Su, J.; Ling, S.; Wang, W.; An, M., Fast-Growing Field of Interfacial Solar Steam Generation: Evolutional Materials, Engineered Architectures, and Synergistic Applications. Solar RRL 2019, 3 (1), 1800206. 33 Jia, C.; Li, Y.; Yang, Z.; Chen, G.; Yao, Y.; Jiang, F.; Kuang, Y.; Pastel, G.; Xie, H.; Yang, B.; Das, S.; Hu, L., Rich Mesostructures Derived from Natural Woods for Solar Steam Generation. Joule 2017, 1 (3), 588-599. 34 Zhang, Y.; Ravi, S. K.; Vaghasiya, J. V.; Tan, S. C., A Barbeque-Analog Route to Carbonize Moldy Bread for Efficient Steam Generation. iScience 2018, 3, 31-39. 35 Tian, W.; Zhang, H.; Sun, H.; Tadé, M. O.; Wang, S., One-step synthesis of flour- derived functional nanocarbons with hierarchical pores for versatile environmental applications. Chemical Engineering Journal 2018, 347, 432-439. 36 Wu, X.; Jiang, L.; Long, C.; Fan, Z., From flour to honeycomb-like carbon foam: Carbon makes room for high energy density supercapacitors. Nano Energy 2015, 13, 527-536. 26 37 Yu, P.; Zhang, Z.; Zheng, L.; Teng, F.; Hu, L.; Fang, X., A Novel Sustainable Flour Derived Hierarchical Nitrogen-Doped Porous Carbon/Polyaniline Electrode for Advanced Asymmetric Supercapacitors. Advanced Energy Materials 2016, 6 (20), 1601111. 38 Yuan, Y.; Ding, Y.; Wang, C.; Xu, F.; Lin, Z.; Qin, Y.; Li, Y.; Yang, M.; He, X.; Peng, Q.; Li, Y., Multifunctional Stiff Carbon Foam Derived from Bread. ACS Appl Mater Interfaces 2016, 8 (26), 16852-61. 39 Ma, S.; Chiu, C. P.; Zhu, Y.; Tang, C. Y.; Long, H.; Qarony, W.; Zhao, X.; Zhang, X.; Lo, W. H.; Tsang, Y. H., Recycled waste black polyurethane sponges for solar vapor generation and distillation. Applied Energy 2017, 206, 63-69. 40 Tan, Z.; Gao, X., 1Comparative advantage and spatial distribution of wheat in China from 1997 to 2016. Journal of Henan Agricultural University. 2018, 52 (05), 825-838. 41 Zhu, L.; Gao, M.; Peh, C. K. N.; Ho, G. W., Recent progress in solar-driven interfacial water evaporation: Advanced designs and applications. Nano Energy 2019, 57, 507-518. 42 Zhang, Q.; Xu, W.; Wang, X., Carbon nanocomposites with high photothermal conversion efficiency. Science China Materials 2018, 61 (7), 905-914. 43 Zhao, Q.; Fan, T.; Ding, J.; Zhang, D.; Guo, Q.; Kamada, M., Super black and ultrathin amorphous carbon film inspired by anti-reflection architecture in butterfly wing. Carbon 2011, 49 (3), 877-883. 44 Wang, H.; Miao, L.; Tanemura, S., Morphology Control of Ag Polyhedron 27 Nanoparticles for Cost-Effective and Fast Solar Steam Generation. Solar RRL 2017, 1 (3-4), 1600023. 45 Guo, A.; Fu, Y.; Wang, G.; Wang, X., Diameter effect of gold nanoparticles on photothermal conversion for solar steam generation. RSC Advances 2017, 7 (8), 4815- 4824. 46 Jia, C.; Li, Y.; Yang, Z.; Chen, G.; Yao, Y.; Jiang, F.; Kuang, Y.; Pastel, G.; Xie, H.; Yang, B.; Hu, L., Rich Mesostructures Derived from Natural Woods for Solar Steam Generation. Joule 2017, 1 (3), 588-599. 47 Tang, N.; Peng, Z.; Guo, R.; An, M.; Chen, X.; Li, X.; Yang, N.; Zang, J. J. P., Thermal transport in soft PAAm hydrogels. 2017, 9 (12), 688. 48 An, M.; Demir, B.; Wan, X.; Meng, H.; Yang, N.; Walsh, T. R., Predictions of Thermo-Mechanical Properties of Cross-Linked Polyacrylamide Hydrogels Using Molecular Simulations. Advanced Theory and Simulations 2019, 2 (1), 1800153. 49 Chen, G., Nanoscale energy transport and conversion: a parallel treatment of electrons, molecules, phonons, and photons. Oxford University Press: 2005. 50 Cahill, D. G.; Braun, P. V.; Chen, G.; Clarke, D. R.; Fan, S.; Goodson, K. E.; Keblinski, P.; King, W. P.; Mahan, G. D.; Majumdar, A.; Maris, H. J.; Phillpot, S. R.; Pop, E.; Shi, L., Nanoscale thermal transport. II. 2003 -- 2012. Applied Physics Reviews 2014, 1 (1), 011305. 51 Wang, Y.; Qin, Z.; Buehler, M. J.; Xu, Z., Intercalated water layers promote thermal dissipation at bio -- nano interfaces. Nature Communications 2016, 7, 12854. 28 Table of Contents Efficient Interfacial Solar Steam Generator with Controlled Macromorphology Derived from Flour via "Dough Figurine" Technology 29 Supporting information Efficient Interfacial Solar Steam Generator with Controlled Macromorphology Derived from Flour via "Dough Figurine" Technology Zhengtong Li1, Chengbing Wang1,, Zeyu Li1, Lin Deng1, Jinbu Su1, Jing Shi2, Meng An2,* 1School of Materials Science and Engineering, Shaanxi Key Laboratory of Green Preparation and Functionalization for Inorganic Material, Shaanxi University of Science and Technology, Xi'an, Shaanxi 710021, China 2College of Mechanical and Electrical Engineering, Shaanxi University of Science and Technology, Xi'an 710021, China * To whom correspondence should be addressed. E-mail: [email protected] (C. Wang), [email protected] (M. An) 30 Table S1 Comparison of important parameters of commercially available materials for use as SISG equipment Carbon ization Carbonizat Experiment Evaporation ment Experi Material temper ion humidity Ref. ature condition (°C) efficiency tempera (%) ture(°C) mushrooms 500 Ar@12h 78%@1sun 28 41 basswood 500 Hot plate 57.3%@1sun constant constant Flame basswood * 72%@1sun 26 40 treatment Sponge 500/70 0/900 N2/2h (2.5-fold) 22.5 Hot wood 500 86.7%@10sun * plate/60s wood 500°C * 74%@1sun * * * melamine 400/55 foams 0/ 700 N2/2h 87.3%@1sun 25 50 daikon 750 N2/2h 85.9%@1sun 28 louts 500 N2 86.5%@1sun 28 * * 1 2 3 4 5 6 7 8 9 bread 400 71.4%@1sun 21 70 10 Homemad e furnace 31 flour 300 Air/2h 71.4%@1sun 24 30 Our work Ps. *On behalf of the reference does not specifically mention the relevant conditions 32 Figure S1. Several typical special structure macroscopic absorber models. The detail of mode is show in Table S2. The models shown in Figure S1 are currently typical solar absorbers with special macro geometry. We summarize the corresponding geometric model structure (see Table S2) and its unique advantages. And we have made some structures using flour-based materials (see Figure S2 for details). This section mainly shows that the flour-based material is a solar absorber preparation material with great potential. 33 Table S2. Advantage Analysis of Several Typical Special Structure Macroscopic Absorber Models N . Ref. Style Characteristics Hollow cylinder Solid cylinder Multiple reflection hit inner wall long surface(minimizing the energy loss from the top surface and maximizing the energy gain from the side surfaces) Inverted Large surface(light), small path(water) a 11, 12 b 13 c 9, 14 solid cone Inverted d 15,16 Multiple reflection hit inner wall; small path(water) hollow cone Positive large surface-projected area ratio(smaller temperature e 1 solid cone increase)---suppressed convection and radiation losses Positive f 16, 17 large surface-projected area ratio hollow cone 34 Figure S2. Flour block preparation of different shapes and volumes by dough figurine. (a) Solid cone, hollow cylinder, solid cylinder (from left to right). (b) Different size of samples. Figure S3. The mechanical performance test of C-FFNC sample (a) The sample is placed on the leaves of a campus plant. (b) 1 kg of weight is placed on top of the sample 35 Here we use the traditional carbon foam mechanical performance test method to express its mechanical properties. The carbon foam is placed on the leaves of the campus plants without obvious deformation, indicating that the quality is lighter. Furthermore, it is placed under the 1kg standard weight without any mechanical deformation or breakage, so it shows that its mechanical strength is excellent. Figure S4. Pulverized mercury instrument raw data mapping analysis of FFNC and C- FFNC. 36 Figure S5. TG-DSC curve analysis of FFC samples. Thermogravimetric (TG)-differential scanning calorimetry (DSC) analysis of freeze- dried samples can explained the substance change during the carbonization process. As shown in the figure, in the 0-60 minutes, the temperature is raised at 5 °C min-1, then it is kept at 300 °C for 60 minutes, and then naturally cooled to room temperature. The TG curve showed two processes of mass reduction, which reduced the total mass by 7.6% and 58.36% respectively. There are two corresponding peaks in the DSC curve. The first peak of DSC curve is an endothermic process with an onset temperature of 41.1°C and a peak temperature of 83.6°C. It can therefore be inferred that the first mass reduction of the TG curve is the evaporation of water inside. The other peak of DSC curve is an exothermic process with an onset temperature of 275.1°C and a peak temperature of 301.1°C. The main starch and protein were incompletely carbonized during this process. 37 Figure S6 The specific heat and thermal conductivity of samples before and after carbonization. Figure S7. The reflection curve of C-FFP, C-FFNS, C-FFNC sample. 38 To further confirm that the optical properties of the different samples are the same, here we test the reflectance of the C-FFNC, C-FFP, and C-FFNS with four different angles, respectively. In the band (400-760 nm), there is no obvious fluctuation for all samples. However, through detailed calculations, the absorption rate of C-FFNC is slighter lower than C-FFNS and C-FFP (about 1%) due to C-FFNS and C-FFP with higher surface area. \ Figure S8.The wetting process of C-FFNC, C-FFNS, C-FFP sample at 0 s, 60 s, 120 s, 180 s. We preheated all samples in an oven (set temperature 110 °C) for 1 min, then placed them inside the chamber and observed a temperature drop of ~4 °C in three minutes. The above heating step was repeated, and the sample was placed in water, and the surface temperature was changed within three minutes as shown in Figure S2. In the absence of sunlight, effective water transport and natural evaporation mainly promote the cooling process of the sample. The surface temperature of the three samples is close to the water temperature at three minutes. This can again prove that the transport of water is nearly close for the three samples. 39 Reference 1 Xu, N.; Hu, X. Z.; Xu, W. C.; Li, X. Q.; Zhou, L.; Zhu, S. N.; Zhu, J., Mushrooms as Efficient Solar Steam-Generation Devices. Advanced Materials 2017, 29 (28), 1606762. 2 Zhu, M.; Li, Y.; Chen, G.; Jiang, F.; Yang, Z.; Luo, X.; Wang, Y.; Lacey, S. D.; Dai, J.; Wang, C.; Jia, C.; Wan, J.; Yao, Y.; Gong, A.; Yang, B.; Yu, Z.; Das, S.; Hu, L., Tree- Inspired Design for High-Efficiency Water Extraction. Advanced Materials 2017, 29 (44). 3 Xue, G.; Liu, K.; Chen, Q.; Yang, P.; Li, J.; Ding, T.; Duan, J.; Qi, B.; Zhou, J., Robust and Low-Cost Flame-Treated Wood for High-Performance Solar Steam Generation. ACS Applied Materials Interfaces 2017, 9 (17), 15052-15057. 4 Zhu, L.; Gao, M.; Peh, C. K. N.; Wang, X.; Ho, G. W., Self-Contained Monolithic Carbon Sponges for Solar-Driven Interfacial Water Evaporation Distillation and Electricity Generation. Advanced Energy Materials 2018, 8 (16). 5 Jia, C.; Li, Y.; Yang, Z.; Chen, G.; Yao, Y.; Jiang, F.; Kuang, Y.; Pastel, G.; Xie, H.; Yang, B.; Hu, L., Rich Mesostructures Derived from Natural Woods for Solar Steam Generation. Joule 2017, 1 (3), 588-599. 6 Liu, H.; Chen, C.; Chen, G.; Kuang, Y.; Zhao, X.; Song, J.; Jia, C.; Xu, X.; Hitz, E.; Xie, H.; Wang, S.; Jiang, F.; Li, T.; Li, Y.; Gong, A.; Yang, R.; Das, S.; Hu, L., High- Performance Solar Steam Device with Layered Channels: Artificial Tree with a Reversed Design. Advanced Energy Materials 2018, 8 (8), 1701616. 7 Lin, X.; Chen, J.; Yuan, Z.; Yang, M.; Chen, G.; Yu, D.; Zhang, M.; Hong, W.; Chen, X., Integrative solar absorbers for highly efficient solar steam generation. Journal of Materials Chemistry A 2018, 6 (11), 4642-4648. 40 8 Zhu, M.; Yu, J.; Ma, C.; Zhang, C.; Wu, D.; Zhu, H., Carbonized daikon for high efficient solar steam generation. Solar Energy Materials and Solar Cells 2019, 191, 83- 90. 9 Fang, J.; Liu, J.; Gu, J.; Liu, Q.; Zhang, W.; Su, H.; Zhang, D., Hierarchical Porous Carbonized Lotus Seedpods for Highly Efficient Solar Steam Generation. Chemistry of Materials 2018, 30 (18), 6217-6221. 10 Zhang, Y.; Ravi, S. K.; Vaghasiya, J. V.; Tan, S. C., A Barbeque-Analog Route to Carbonize Moldy Bread for Efficient Steam Generation. iScience 2018, 3, 31-39. 11 Shi, Y.; Li, R.; Jin, Y.; Zhuo, S.; Shi, L.; Chang, J.; Hong, S.; Ng, K.-C.; Wang, P., A 3D Photothermal Structure toward Improved Energy Efficiency in Solar Steam Generation. Joule 2018, 2 (6), 1171-1186. 12 Shi, Y.; Zhang, C.; Li, R.; Zhuo, S.; Jin, Y.; Shi, L.; Hong, S.; Chang, J.; Ong, C.; Wang, P., Solar Evaporator with Controlled Salt Precipitation for Zero Liquid Discharge Desalination. Environmental Science & Technology 2018, 52 (20), 11822-11830. 13 Li, X.; Li, J.; Lu, J.; Xu, N.; Chen, C.; Min, X.; Zhu, B.; Li, H.; Zhou, L.; Zhu, S.; Zhang, T.; Zhu, J., Enhancement of Interfacial Solar Vapor Generation by Environmental Energy. Joule 2018, 2 (7), 1331-1338. 14 Wang, G.; Fu, Y.; Guo, A.; Mei, T.; Wang, J.; Li, J.; Wang, X., Reduced graphene oxide-polyurethane nanocomposite foams as a reusable photo-receiver for efficient solar steam generation. Chemistry of Materials 2017, 29 (13), 5629. 15 Wang, Y.; Wang, C.; Song, X.; Huang, M.; Megarajan, S. K.; Shaukat, S. F.; Jiang, H., Improved light-harvesting and thermal management for efficient solar-driven water 41 evaporation using 3D photothermal cones. Journal of Materials Chemistry A 2018, 6 (21), 9874-9881. 16 Hong, S.; Shi, Y.; Li, R.; Zhang, C.; Jin, Y.; Wang, P., Nature-Inspired, 3D Origami Solar Steam Generator toward Near Full Utilization of Solar Energy. ACS Appl Mater Interfaces 2018, 10 (34), 28517-28524. 17 Zhang, P.; Liao, Q.; Yao, H.; Cheng, H.; Huang, Y.; Yang, C.; Jiang, L.; Qu, L., Three-dimensional water evaporation on a macroporous vertically aligned graphene pillar array under one sun. Journal of Materials Chemistry A 2018, 6 (31), 15303-15309. 42
1802.07815	1	1802	2018-02-20T02:22:32	A Compact Low-Cost Low-Maintenance Open Architecture Mask Aligner for Fabrication of Multilayer Microfluidics Devices	[ "physics.app-ph", "physics.ins-det" ]	A custom-built mask aligner (CBMA), which fundamentally covers all the key features of a commercial mask aligner, while being low cost, light weight, and having low power consumption and high accuracy is constructed. The CBMA is comprised of a custom high fidelity LED light source, vacuum chuck and mask holder, high-precision translation and rotation stages, and high resolution digital microscopes. The total cost of the system is under $7,500, which is over ten times cheaper than a comparable commercial system. It produces a collimated ultraviolet illumination of 1.8-2.0 mW cm-2 over an area of a standard 4-inch wafer, at the plane of the photoresist exposure; and the alignment accuracy is characterized to be < 3 microns, which is sufficient for most microfluidic applications. Moreover, this manuscript provides detailed descriptions of the procedures needed to fabricate multilayered master molds using our CBMA. Finally, the capabilities of the CBMA are demonstrated by fabricating two and three-layer masters for micro-scale devices, commonly encountered in biomicrofluidics applications. The former is a flow-free chemical gradient generator and the latter is an addressable microfluidic stencil. Scanning electron microscopy is used to confirm that the master molds contain the intended features of different heights.	physics.app-ph	physics	A Compact Low-Cost Low-Maintenance Open Architecture Mask Aligner for Fabrication of Multilayer Microfluidics Devices Quang Long Pham,a Nhat-Anh N. Tong,a Austin Mathew,b Roman S. Voronova,† aOtto H. York Department of Chemical, Biological and Pharmaceutical Engineering, New Jersey Institute of Technology, Newark, NJ 07102, USA bDepartment of Biomedical Engineering, New Jersey Institute of Technology, Newark, NJ 07102, USA. †Corresponding author. Email: [email protected] ABSTRACT: A custom-built mask aligner (CBMA), which fundamentally covers all the key features of a commercial mask aligner, while being low cost, light weight, and having low power consumption and high accuracy is constructed. The CBMA is comprised of a custom high fidelity LED light source, vacuum chuck and mask holder, high-precision translation and rotation stages, and high resolution digital microscopes. The total cost of the system is under $7,500, which is over ten times cheaper than a comparable commercial system. It produces a collimated ultraviolet illumination of 1.8-2.0 mW cm-2 over an area of a standard 4-inch wafer, at the plane of the photoresist exposure; and the alignment accuracy is characterized to be < 3 µm, which is sufficient for most microfluidic applications. Moreover, this manuscript provides detailed descriptions of the procedures needed to fabricate multilayered master molds using our CBMA. Finally, the capabilities of the CBMA are demonstrated by fabricating two and three-layer masters for micro-scale devices, commonly encountered in biomicrofluidics applications. The former is a flow-free chemical gradient generator and the latter is an addressable microfluidic stencil. Scanning electron microscopy is used to confirm that the master molds contain the intended features of different heights. Keywords: mask aligner, photolithography, microfluidics, microfabrication, multi-height 1. Introduction Multilayer photolithography has played a central role in the microfabrication of multi-height photoresist master molds for polydimethylsiloxane (PDMS) microfluidic devices, which is widely used in various fields of biological, micro electro-mechanical systems (MEMs), micro total analysis systems (TAS), sensors and other applications.[1-6] Moreover, with the current explosion of the lab-on-a-chip technology in diagnosis and fundamental medical research, the need for the miniaturization of systems via deposition and etching procedures is greater than ever before.[7-9] However, the high cost of the equipment and its maintenance, combined with the expensive and space-consuming installation requirements associated with commercial systems, limit the access of many small laboratories and companies to manufacturing microfluidics devices in-house. For example, at the center of the photolithography fabrications is a mask aligner, which simultaneously provides precise mask-to-wafer alignment and generates a uniform ultra violet (UV) illumination over an exposed photoresist surface. However, current commercial systems (i.e. Karl Suss, OAI, and EVG) are very costly (e.g., $60,000-100,000 capital investment, plus $1,000-3,000 annual maintenance), heavy (> 250 lbs.), and relatively bulky (> 700 in2). Therefore, there is an unmet need for photolithography systems that are fully 1 functional, yet compact and affordable. To that end, this manuscript aims to present a custom- built mask aligner (CBMA), that can be constructed easily and for as little as $7,500 (and potentially even lower, if an in-house machine shop is used). A collimated UV light source is the key component of any mask aligner, as it is directly responsible for creating the master molds for the microfluidics devices by crosslinking photoresist coated on wafers in a masked pattern. Typically, a high-pressure mercury lamp is used for this purpose by commercial systems. However, the UV light alone requires a significant capital investment (~$10,000-20,000), and accounts for a dominant portion of a mask aligner's annual maintenance costs. Moreover, the mercury lamps tend to provide a broadband illumination (bad for fabricating tall features) with an intensity that often drifts with time. Consequently, several efforts have been made to produce low-cost alternatives to the mercury lamp. For example, Huntingtan et al[10] introduced a portable light source composed of an array of hundreds of light emitting diodes (LEDs), which could be powered by AA batteries. Though such sources can provide a relatively uniform illumination that yields sub-micron features, they are not suitable for thick photoresist masters (i.e. over 100 µm) due to the large divergence of the illumination. this high. Consequently, a more recent light source addressed this problem by improving the light collimation using optical lenses that help to reduce the divergence of the light beam.[11] As a result, it was able to pattern thick masters of up to several hundred microns, while using an array of just 9 high power 365nm LEDs. Therefore, we chose this light source for our CBMA for its superior quality and minimalistic design. However, many microfluidics applications require features stages translation In addition to the light source, a complete mask aligner system should have the capability for a mask-to-wafer alignment, which is critical for multi-layer soft lithography. Alignment systems have been made by other researchers. For example, Li et al[12] introduced a desktop aligner that combines digital microscopes with high-precision for aligning polydimethylsiloxane (PDMS) slabs. However, photolithography is not the main purpose of this system, and thus it is not equipped with a UV light source for photoresist exposure. Therefore, there is a need for an alignment system with an integrated high fidelity light source, capable of deep (i.e. > 100 m) photoresist exposure. To that end, we designed a low cost alternative to current commercial systems that is well equipped with the fundamental functions of a complete mask aligner: a uniform UV light with deep photoresist penetration, high-resolution rotation and translation stages, vacuum chucks and mask holder, and digital microscopes for high-precision alignment, etc. Along with the instructions for how to build the system, this manuscript also provides the detailed procedures for how to optimize its performance and manufacture multi-layer master molds with a high accuracy. Given that this system is affordable and low maintenance, it is expected that it will have widespread applications, especially in small and medium laboratories that do not have access to an in-house microfabrication facility. 2. Key Features & System Description Key features of our CBMA include cost-savings, light weight, space savings, and an alignment accuracy sufficient for most typical microfluidics applications. A comparison of some of these the key features between the CBMA and a typical commercial system is given in Table 1. Relative to a commercial system (i.e. Karl Suss MA100, EVG 620, OAI 200), our system is an order of magnitude cheaper, weighs 5 times less, occupies 5 times less space, and consumes 50 times less power. Moreover, it has an open adjustable architecture and a negligible annual 2 Table 1. Key feature comparison between the CBMA and a typical commercial mask aligner. Features Custom-built Commercial Cost/Price, $ Weight, lbs Light source power, W Width x Depth, in2 XY Alignment Accuracy, µm < 7,500 < 50  22 15 x 10 2.96 60,000 - 120,000 > 250 > 1000 32x23 < 0.5 Table 2. Cost breakdown showing the approximate costs of the main components of our CBMA. Item Part Numbers (quantity) Company Translation & rotation stages 423-MIC (3), 461-XZ-M(1) 360-90 (2), UTR80 (1), 39 (1) Newport, Irvine, CA Cost ($) 4,100 1,000 900 390 300 280 140 100 100 70 40 7,420 - Machined adapters Digital microscopes UV LED Machined wafer chuck Dovetail optical Rail & carrier UV-collecting lens 0-50 V power supply Laser cut & engraving Analog relay timer T-slotted aluminum frame Total AD4113T (2) UV LED 897-LZ110U600 (9) - 10R300 (2), 20C (2) FCN12592_LE1-D-COP (9) 29612 PS (1) Mask holder (1), Lens mount (1) H3CR-A8-AC100-240 (1) 1010 (160 in) Zera Development, S. Clara, CA Dino-lite, Torrance, CA Mouser, Mansfield, TX Emachineshop, Mahwah, NJ Optic Focus Solution, China Mouser, Mansfield, TX MPJA, West Palm Beach, FL Ponoko, Oakland, CA Mouser, Mansfield, TX Knotts Co, Berkeley Heights, NJ maintenance cost. Finally, although the resolution of CBMA is not as high as that of a commercial system, it is still sufficient for most MEMS and microfluidics applications. A detailed cost breakdown of all the main components of the CBMA is provided in Table 2. The major cost comes from the use of translation and rotation stages, which provide the high precision, stable, and smooth motion of the mask, wafer, and microscopes, thus allowing for a fine and accurate alignment. Other components include a collimated UV light source, custom- machined wafer chuck and mask holder, digital microscopes, and other minor mechanical and electronic components. All of these components are discussed in detail in the subsequent sections of this manuscript. The entire unit can be mounted on an optical breadboard with imperial threaded holes in order to dampen vibrations. 2.1. Light Source The LED-based light source for the CBMA is reproduced from a system published by Erickstad et al.[11] This system was selected because it offers several advantages such as excellent illumination uniformity, low cost, short response time, and stable and low power consumption. Hence, the LED light source is significantly cheaper and more energy efficient when compared to a standard UV mercury lamp, which requires significant capital investment and annual maintenance, costs. 3 Figure 1. (A) An "exploded" view of the custom UV light source constructed from a squared array of 9 LEDs. Compared to original design,[11] the tip/tilt adjustment is added in order to further to aid the accurate alignment of the light source components. (B) Photograph of the light source assembly viewed from the top and from the bottom of the mask aligner. (C) & (D) Heat maps showing the distribution of light intensity before and after LED alignment, respectively. The dashed circles represent the location of a 4-in Si wafer, which is aligned co-centrically with the LED array. The complete technical details of the LED light source have been discussed elsewhere.[11] For the sake of brevity, we describe some its main components, and the modifications we made to the original design. Typically, the light source consists of a 3x3 square array of 9 LEDs mounted on an aluminum plate, 50 mm apart from each other. The LEDs were soldered to 26G electric wires using low temperature solder cores (Indium Corporation, Clinton, NY). A UV-collecting lens was glued to each LED to reduce the divergence angle of the emitted light from 80 to 12. The divergence angle was further brought down to 4.7 using a squared array of credit card- sized plastic Fresnel lenses (with the same pitch as the LEDs). Light passing through the lens array is projected down to create an overlapping area of uniform illumination at the wafer plane. The plane is located 906 mm away from the aluminum plate that serves as a base for the LEDs. 4 In the original design, the light source was fixed to a ceiling of a room. This does not provide flexibility to the system, especially when it needs to be moved around the lab space. Instead, we mounted the light source on a supporting frame made from T-slotted aluminum 80/20 extrusions (Knotts Co, Berkeley Heights, NJ). This way, the frame and the light source can be moved together easily. We further modified the system by introducing a tip/tilt translation to the light source using an acrylic plate with adjustment knobs (Figure 1A, B). This adjustment ensures the perpendicularity of the light beam. The tip/tilt translation and a linear translation were also added to the Fresnel lens plate to make sure the lenses were co-aligned with the light source. Five cooling fans were provided to dissipate heat generated from long-exposure LEDs. However, they were usually turned off to avoid vibration when working with patterns of small feature sizes, since they require a short exposure time. It was found that doing this would not significantly affect the performance of the CBMA. The adjustment of the illumination is demonstrated in Figure 1C, D. Here, it can be seen that without the adjustment, the illumination was not well centered, and is thus unable to provide a uniform photoresist exposure over a wafer surface (Figure 1C). However, by tuning the tip/tilt adjustment knobs as well as the translation stage, we were able to center the light, thus providing a more uniform illumination over the wafer area (Figure 1D). In order to generate Figure 1C and D, the spatial distribution of the light intensity at the theoretical wafer plane (i.e., 906 mm from the light source) was measured using a Traceable UV meter (#06-662-65, Fisher, Waltham, MA) at 1-cm increment in the X and Y directions. The light source was supplied with 34-V, 0.6-A power, and the heat maps were generated using the "Color Scales" function of Microsoft Excel (Microsoft, Belleville, WA). 2.2. Vacuum Wafer Chuck and Mask Holder In order to securely fasten the wafer and the photomask during alignment and exposure, an aluminum chuck (machined by Emachineshop, Mahwah, NJ) and an acrylic holder (laser cut and engraved by Ponoko, Oakland, CA) were designed (Figure 2). Both the wafer chuck and the mask holder operate via vacuum. The wafer chuck was patterned with 0.8-mm wide, 1-mm deep concentric troughs, which help to hold the wafer tightly and minimize any vibration due to rotation (Figure 2B). The chuck is compatible with a wafer size of 4 in or smaller. The vacuum was supplied to the innermost trough through 4 vacuum holes, which were connected to a 15- Torr house vacuum via a bearing adapter (machined by Zera Development Co, Santa Clara, CA) attached to the bottom of the chuck (Figure 2D). This setup allows the chuck to be rotated freely in 360, thus offering a high-flexibility alignment. The holder for the photomask was laser-cut from a 10-mm thick acrylic plate in order to create a window with 4 in square aperture for the UV light. Additionally, it was engraved with a 1-mm deep, 3.25-mm wide vacuum channel for holding the mask. The holder was also connected to the house vacuum via tubing, as shown in Figure 2C. Once the vacuum was activated, the chuck could firmly hold a 5-inch square 90 mil- thick quartz mask (Photomaskportal, Richardson, TX). Both the wafer chuck and the mask holder were levelled using a bull's eye level to ensure that the mask and the wafer were in full contact. The 3D drawings for the wafer chuck and mask holder assemblies can be found in the supplemental materials. 2.3. Alignment Microscopes The registration of multiple masked patterns on a wafer requires placement of "alignment" marks on the masks and "reference" marks on the wafer. This process relies heavily on the visibility of the micron-size marks, often located diametrically opposite to each other on the wafer / mask. 5 Figure 2. (A) A 3D illustration view showing the design of the alignment assembly comprised of a wafer chuck, mask holder, microscopes, and high-precision positioning stages. (B) A photograph of a wafer- mounting assembly showing the wafer chuck placed on top of a rotational and tip/tilt translational station. (C) An illustration showing different parts of vacuum adapters that help to connect the wafer chuck to the house vacuum. (D) A photograph showing an acrylic mask holder connected to the house vacuum via tubing. (E) A 3D view of the mask holder showing a vacuum channel engraved on the holder's surface, and holes supplying vacuum to the channel. Herein, we used two digital microscopes (AD4113T, Dino-lite Digital Microscope, Taiwan) positioned on both sides of the CBMA to provide microscopic views of the marks (Figure 2A). These microscopes were used since they are cheap, compact, and simple to operate. Moreover, they can be connected to a laptop via a Universal Serial Bus (USB) interface, and controlled using the Dino Capture 2.0 software provided by the manufacturer. During alignment, both microscopes were used simultaneously to acquire a live view of the marks at various magnifications (Figure S1). The UV content of the microscopes' light sources is negligible, as measured by a UV meter; however, to ensure the photoresist was completely intact during alignment, the microscopes were covered with amber UV filter films (#F007-006, UVPS, Chicago, IL), which absorbed any UV radiation emanated from the microscopes' light sources. 2.4. Kinematic Mask Alignment System In order to ensure of a robust mask-to-wafer alignment with a high fidelity, high-precision positioning stages were used. The wafer chuck was mounted on a rotary stage (#UTR80, Newport) and a tip/tilt stage (#39, Newport), which offer high sensitivities of 0.001 and 0.002, respectively (see Figure 2B). The mask holder was mounted on a translation complex, which consisted of a XZ translation stage (#461-XZ-M, Newport) and a Y translation stage (#423, Newport) combined. This allows the mask to travel in all three directions, with travel distances of 1 inch in X and Z, and 2 inches in Y (see Figure 2A). Digital microscopes were also mounted on linear translation stages (#423-MIC, Newport, Irvine, CA) configured in XYZ directions, thus allowing the microscopes to be accurately positioned and focused in a highly repetitive manner. Fast and coarse vertical motion of the microscopes was facilitated using tail optical rail and carrier (10R300 and 20C, respectively; Optics Focus Instruments Co, China). 6 2.5. Electronics A bench-top variable 0-50 V power supply (#29612 PS, MPJA, West Palm Beach, FL) was used to power the light source using a voltage setting of 34 V and a current of 0.6 A. This provides an average illumination power of 1.85-2.00 mW/cm2 at the wafer's surface (see Figure 1C, D). The exposure time was controlled by a mechanical relay timer from Omron (#H3CR-A8-AC100- 240, Mouser, Mansfield, TX) which automatically disconnects the light source from the power supply when the set time is reached. The cooling fans (#SY124020L, Scythe Co., Germany) were powered by a 12V AC power supply (#1670, Current USA, Vista, CA). 2.6. Supporting Frame The frame of the CBMA was constructed from T-slotted 8020 aluminum extrusions (Knotts Co, Berkeley Heights, NJ) which offers stable mounting of the hardware components, and at the same time provides for the light-weight nature of the CBMA (Figure S1). The light source was attached directly to the frame, while the wafer chuck, the mask holder, and the microscopes were not connected to it. This frame can be conveniently mounted on top of an optical table with imperial threaded holes. Alternatively, it can also be used as a standalone setup, without the optical table. However, one is recommended in order to reduce fabrication defects caused by vibration. This setup offers the flexibility often desired in laboratories with a limited space. 3. Fabrication Procedure The fabrication of a master mold of a microfluidic device (shown in Figure 6 and discussed in detail in the "Mask Aligner Application" section) with two different heights is used to illustrate a typical alignment procedure for the CBMA. Generally, this type of fabrication follows a standard photolithography technique, which includes three basic elements: spin coating, UV exposure, and development. In order to generate a master mold with two different heights, we used three photomasks: a "reference" mask containing only reference marks to be imprinted on the wafer, and two "device" masks containing templates for the different photoresist heights. These masks can be either chrome coated on a glass substrate or printed on transparencies at a high resolution (> 10,000 dots per inch). Even though the transparency masks are not likely to offer feature sizes smaller than 10 µm, they are inexpensive and suitable for most standard microfluidic applications. Therefore, we went with this option. The transparency masks were taped onto 5-in square bare quartz substrates (90-mil thickness) (Photomaskportal, TX) with the printed side facing outward, using a ¼-in polyimide film tape (#5413, 3M, Mapplewood, MN). A 3-step procedure for the fabrication of a multi-height master is outlined in Figure 3. In the first step of this procedure, the two reference mark arrays were fabricated diametrically opposite to each other on a 4-in Si wafer (#1196, University wafer, South Boston, MA), using the "reference" mask. The marks were 70-80 mm apart from each other. Positive photoresist (# AZ P4620, Microchem, Westborough, MA) was diluted prior to spin-coating with an edge bead removal solvent (#EBR PG, MicroChem). A resist:EBR ratio of 1:4 was used to achieve a thin layer of 5 µm. This photoresist was selected due to its amber shade when developed, which provides alignment marks that are highly contrasted relative to the color of the wafer. The alignment accuracy is significantly enhanced as a result. In the second step of the procedure in Figure 3, the 20 µm-high feature of the master mold was generated from a negative photoresist SU-8 2015 (Microchem). The negative photoresist was spin-coated on the "marked" wafer generated in step 1. After the coating, the first layer "device" mask is aligned with the wafer's reference masks using the alignment procedure discussed in the subsequent "Mask Alignment Procedure" section of this manuscript. After 7 Figure 3. Fabrication of a multi-height master mold on a 4-in Si wafer using the CBMA in a 3-step procedure. In step 1, reference marks are imprinted on the wafer using a diluted positive photoresist (#AZ P4620, Microchem) via a "reference" mask. In step 2, a 20-µm high SU-8 photoresist layer is patterned on the wafer, which is aligned with the mask using the CBMA. In step 3, a 60-µm thick SU-8 layer is patterned on the same wafer following a similar procedure mentioned in step 2, but with a different mask. Abbreviations: FT = Film Thickness. alignment, the wafer was exposed to UV (140 mJ), and developed to create a 20-µm high resist pattern. The third step of the procedure in Figure 3 is needed in order to generate the 60-µm high feature of the master mold. It essentially repeats the second step, but with a different "device" mask. Photoresist SU-8 2035 (Microchem) was again spin-coated on the same wafer to achieve a 60-µm thick film. This was then followed by alignment of the second layer "device" mask using the CBMA. The resist was then exposed to UV (210 mJ) and developed to attain the final master mold. 4. Mask Alignment Procedure A detailed aligning procedure is illustrated in Figure 4A. First, a "device" photomask is placed on top of a vacuum wafer chuck with the pattern side facing downward. The mask holder is lowered in the Z direction by a XYZ translation stage, until it is in contact with the mask. Then vacuum is applied in order to seal the mask to the mask holder. Once the mask is securely 8 Figure 4. (A) Diagram showing a step-wise alignment procedure using the CBMA. (B) Two alignment microscopes in operation. (C) Visualization of the fine alignment process captured by the alignment microscopes at 245x. Black hollow and red solid crosses are the mask "alignment" and the wafer "reference" marks, respectively. Scale bar is 200 µm. attached to the holder, the assembly is raised up in order to vacuum-seal a marked wafer (described in Step 1 of Figure 3) on to the wafer chuck. The mask is then brought in close proximity to the wafer (i.e. less than 100 µm, assuming the resist film thickness is smaller than that). Note that the distance between the mask and the wafer can be judged by observing the reflection of the mask on the wafer surface. Excessively close proximity is not necessary, as it would restrict the movement of the mask and the wafer during alignment. The coarse alignment begins by focusing two digital microscopes on the two alignment mark regions on the "device" masks, as shown in Figure 4B. The reference marks on the wafer are brought into the microscopes' views using the translation and rotation stages. This is first done at a 30x magnification of the two microscopes in order to achieve coarse alignment (Figure S2A). Fine alignment then follows at a 245x magnification. At this time, the marks viewed by Microscope I should be symmetrically opposite to those viewed by Microscope II with respect to the center of the wafer. This helps to minimize the rotational error of the alignment. In addition, it should also be noted that at the 245x magnification, the Dino-Lite microscopes have a working distance of  10 mm. This allows both the wafer's reference marks and the mask's alignment marks to stay in focus at the same time (Figure 4C), which also greatly enhances the alignment accuracy. The fine adjustment is first started by bringing the two masks' alignment marks to the same Y position on the observation monitor (see coordinate axis in Figure S2B). Rotation of the wafer is then performed in order to bring the wafer reference marks to the same Y position on the view 9 Figure 5. (A) Two photomasks containing different alignment patterns, "bracket" and "dot", were used to generate a photoresist profile for the alignment accuracy measurement. (B) Micrograph of a defective SU- 8 print caused by misalignment of the two patterns. (C) Micrograph of a SU-print obtained from well- aligned patterns. (D) Illustration of how translational x-y and rotational  error values are determined. (E) & (F) Alignment errors quantitatively determined from the SU-8 prints with respect to the photoresist's thickness. screen. After that, the rotation stage is locked to restrict the rotation movement of the wafer. This is followed by bringing the mask closer to the wafer by adjusting the Z stage attached to the mask holder. The actual proximity depends on the operation type that is most appropriate for the user's purpose: either contact-free or hard contact. Following that, the mask is then translated in the XY plane in order to center the wafer reference marks inside of the mask's alignment marks, as shown in Figure 4C. Once the alignment is finished, the two microscopes are moved out of the way, and the light source is turned "on" in order to start the exposure. 5. Characterization of Mask Alignment Accuracy The accuracy of the alignment was quantitatively determined from SU-8 prints which are constructed using two different mask patterns, as shown in Figure 5A. The masks with feature size of 50 µm were printed on plastic transparencies at 10,160 DPI (Fineline Imaging, Colorado Springs, CO). The first mask contains "bracket" patterns and the second mask has "dot" patterns. SU-8 resist films of various thicknesses, ranging from 5 µm to 120 µm, were coated on a 4-in Si wafer containing positive photoresist alignment marks. The "bracket" mask was then aligned with the wafer following the protocol mentioned in the previous section. The photoresist layer was then exposed to UV radiation for a duration of 40 to 120 seconds, depending on the film's thickness. After that, the "bracket" mask was replaced by the "dot" mask, and followed by 10 alignment. The resists were then exposed at the same UV dose, baked at 95 C for 5-10 min, and developed to generate the measurement patterns (see Figure 5B, C). Images of the patterns were acquired using a CCD camera Guppy Pro F-146 (Allied Vision, Germany) coupled to a reflected microscope Reichert Zetopan Trinocular (Austria). The acquisition was performed using a 10X objective (PLN10X, Olympus, Japan). Measurement of the patterns was conducted using open source ImageJ software.[13] Upon successful alignment, the final SU-8 replica should contain the "dot" pattern centered inside of the "bracket" pattern, without offsets in any direction. Severe misalignment leads to the overlapping of the "dot" on the "bracket", as shown Figure 5B. In contrast, careful alignment results in a well-defined print that looks closer to what is expected from the superposition of the two masks (see Figure 5C). The aligned patterns were then used for the characterization of the alignment accuracy. Figure 5D demonstrates how the translational and rotational alignment errors are determined quantitatively.[12] Namely, the translational errors x and y were calculated from the offset of the "dot" pattern with respect to the "bracket" pattern in the x and y directions, respectively. In addition, the mean translational error L was calculated by averaging the translational errors in the two individual directions. On the other hand, the rotational error  was determined from the angular amount that the former pattern was offset from the latter one. In practice, errorless alignment (i.e., zero delta values), is not possible even with a state-of- the art mask aligner, which can achieve sub-micron alignment accuracy. In essence, the alignment accuracy depends strongly on the user's ability to observe the alignment marks on the mask and the wafer. This, in turn, is dictated by the resolution and by the image quality of the microscopes used for the alignment. Thus, in order to attain optimal alignment accuracy, the digital microscopes were used at their maximum magnification of 245X (corresponding to a resolution of 1.25 µm per pixel). However, other factors such as human error and film thickness[14] may contribute to the imperfection of the alignment. Therefore, the accuracy of 1.25 µm is more or less theoretical for our system, especially considering that the resolution of the stages is also ~1µm. It has been reported elsewhere that the film thickness in particular affects the alignment accuracy by increasing the distance between the mask and the wafer.[14] This makes it progressively more difficult to keep both the mask's marks and the wafer's within the microscope's depth of field. Consequently, the inability to focus on both of the marks simultaneously creates more room for error. We therefore characterized how the alignment accuracy of our CBMA varies with respect to the different photoresist thicknesses: ranging from 5 to 120 µm. Given that the alignment accuracy is also dependent on the operator's skill and experience, we performed each of the thickness experiments in triplicate in order to ensure that the human error was minimized. Surprisingly, both the translational and the rotational accuracies were not found to be highly dependent on the film's thickness (Figure 5E, F); however, the thinner ones resulted in a higher consistency (Figure 5E). With film thickness of 50 µm or thinner, the translational errors appears to increase with the resist's thickness. Typically, the 5-µm film thickness resulted in average translational and rotational errors of 2.96  0.92 µm and 0.062  0.012, respectively. Alignment errors of 20 and 50-µm thick prints were 4.29  0.74 µm, 0.018  0.0007 and 5.36  4.04 µm, 0.045  0.012 respectively. On the other hand, the rotational errors did not exhibit any distinguishable pattern, with values ranging well below 0.1 for all thickness values. This is likely due to the high resolution of the rotation stage, which is about an order of magnitude higher than the alignment accuracy of the translational stages. Finally, to our surprise we were 11 Figure 6. (A) A photograph of the chemotaxis microfluidic device used as a demonstrative application of our CBMA. The device consists of an array of 15 x 20 (W x H) µm channels connecting two media reservoirs of 60-µm in height. Inset is a bright-field microscopy view showing how the microfluidic channels connect the two media reservoirs. Scale bar is 4 mm for the photograph and 400 µm for the inset. (B) Two photolithography masks used to fabricate the master mold for the device: one for the reservoir and one for the channel. Bottom is a 3D drawing of the resulting mold. (C) SEM image of the fabricated two-height SU-8 master. Scale bar is 100 µm (D) Interpretation of how the alignment of the device is characterized. The quantities x, y, and  are calculated from the difference in x, y, and  between the device and the reference mask. (E) Fluorescent microscopy image of fluorescein gradient formed inside the micro channel of a PDMS device. Scale bar is 50 µm. able to achieve a high accuracy with the 120-µm thick film (3.12  2.65 µm, 0.036  0.025), suggesting that film's thickness is not the only factor that hampers the alignment. For example, other factors like vibration, glare, unevenness in the transparency mask, and others could have also contributed to the variation in the alignment results. 12 6. Mask Aligner Application In order to demonstrate an application of our CBMA, we fabricated a SU-8 master mold for a PDMS device similar to ones commonly used to study cell chemotaxis in response to a concentration gradient of a chemoattractant (see Figure 6A).[4, 15-17]. The device consists of an array of 20 µm-high diffusion "maze" channels that connect two identical reservoirs, each of which are 60 µm in height. In the chemotaxis application, one of the reservoirs is meant to hold the cells, while the other typically serves a source of the chemoattractant. In order to generate the device, a master of two different heights (20 and 60 µm) was fabricated from SU-8 photoresist using two different masks: one for the reservoir and one for the diffusion channel (see Figure 6B). The distance between the two reservoirs was 462 µm and the distance between the two ends of the channel was 482 µm. Since, carrying out experiments in parallel requires that the mazes are identical, it is critical to align the two mask patterns accurately. Moreover, any mismatch between the diffusion channel and the reservoir would result in the malfunction of the device. Scanning electron microscopy (SEM) image (Figure 6C) shows that a two-thickness master was successfully generated. SEM was acquired at 45 tilt angle using Field Emission Scanning Electron Microscope (FESEM, LEO 1530VP, Zyvex, Richardson, TX). Prior to imaging, the sample was sputter coated with 5-nm thin gold layer using a sputter coater (EMS150T ES, Quorumtech, Lewes, UK). The alignment accuracy of the device was characterized in the x, y, and  directions (Figure 6D), using the same procedure as in Figure 5D. The quantities x and y were determined to be  3.9 µm, while  was 0.02. This result suggests a successful alignment of the device. We further investigated the performance of the device by testing chemical gradient generation within the diffusion channel using fluorescence tracer Alexa Fluor 488 (Thermofisher, Waltham, MA). The top of the reservoirs was punctured by a 5-mm biopsy punch in order to form a circular inlet port for the liquid media input. The fluorescent solution (Alexa Fluor 488 (1 µg ml-1) in deionized water) was added to one of the reservoirs, while deionized water was added to the other reservoir. Figure 6E show that the resulting concentration profile of the tracer established inside the channels, between the source and the sink of the fluorescein. This result suggests that the device functioned successfully without any defects (i.e. no mismatch between the channels and the reservoirs) and was able to generate a stable gradient, which is suitable for a chemotaxis study of cells. Fluorescence microscopic images were acquired using an inverted microscope (IX83, Olympus, Japan) coupled to a 488-nm laser light source. In a second application, we use our CBMA to fabricate something even more complex: a 3- height replica of an addressable stencil device, which was inspired by a work published by others.[6] This type of device is commonly used to deliver varied concentrations of chemicals to different "addresses" in the chip. Typically, the stencil device is comprised of three PDMS layers: a "control" layer, a "flow" layer, and a "display" layer. However, since the "control" and the "display" layers require just a single-height master mold (which could be easily fabricated using just a UV lamp without the need for a mask aligner), we focus exclusively on the fabrication of the 3-height "flow" layer master-mold. This fabrication includes the use of 3 different mask patterns: a 24-µm high main flow channel to transport liquid from a media source, a 25-µm high bypass channel, and a 90-µm high hole pattern to deliver media to the "display" layer underneath (Figure 7A). In order to generate this master mold, a 3-exposure process was conducted using 3 different photomasks, each of which corresponds to a featured pattern. Firstly, the flow layers was fabricated by double coating the positive photoresist AZ P4620 to obtain 24-µm thick film on an 13 Figure 7. (A) Design of a master mold for the "flow" layer of an addressable stencil device that consists of three photoresist patterns of various heights: 24-µm high main flow channel made out of positive photoresist AZ P4620 (green), 25-µm high bypass channel made out of negative photoresist SU-8 (red), and 90 µm-high delivery hole made out of SU-8 (pink). Colors indicate individual photoresist patterns achieved via different photomasks. (B) SEM images showing different feature heights of the resulting master mold. Scale bars are 200 µm. "reference" mark-containing 4-in Si wafer, following a 3-min soft bake at 110 C. Then the wafer was aligned with a photomask containing the pattern for the flow channel, and exposed to UV (2 mW cm-2) for 960 seconds using the CBMA. The resists were developed and heated to 150 C on a hot plate for 20 min to generate a round channel profile, as well as for improving the adhesion of the resist to the wafer for subsequent fabrication. After that, the wafer was coated with a 25-µm thick layer of the negative photoresist SU-8 2015, aligned with the second mask containing the pattern of the bypass channels, and exposed to 75 seconds of UV (2 mW cm-2) . It was then developed and heated on a hot plate at 150 C for 5 min. Next, the wafer was coated with 90-µm thick SU-8 2035, aligned with the third mask containing the pattern for the holes, and exposed to UV (2 mW cm-2) for 110 seconds. Finally, the wafer was developed, and hard baked at 150 C for 5 min to generate the final master. The master was then examined using SEM. The SEM images reveal that the master was successfully generated, with all the patterns were well aligned with each other (Figure 7B). 14 7. Conclusion & Future Work In this paper, we demonstrated a cost-effective, lightweight, and accurate custom-built CBMA for the fabrication of multi-height photoresist patterns which are subsequently used as master molds for microfluidic applications. The CBMA was approximately tenfold cheaper than analogous commercial systems, thus making it suitable for medium and small laboratories in need of in-house microfluidic device fabrication. Owing to its lightweight and space-saving feature, the CBMA can be integrated into any lab space as a bench top unit. In addition, it is suitable for thick resist layers (> 100 µm) with an alignment accuracy of <3 µm, making it a potential replacement for the high-cost, high-maintenance commercial systems typically used in microfabrication. Although our CBMA, is capable of achieving a wide spectrum of layered features that cover most of the microfluidics applications in its presented form, several major and minor improvements can be made to it based on the user's needs. For example, a simple enhancement can be achieved via supplementing the two alignment microscopes with a dedicated tablet PC each (simple ones can cost as low as ~$100). This can save the user time involved with connecting/disconnecting the microscopes to computers and monitors every time the CBMA is used. Another possible upgrade is automating the alignment stages with electronic controllers. For example, the manual rotation stage can be replaced by a motorized version (#URS75BPP, Newport); while the linear translation stages can be actuated by motorized actuators (#TRA25, Newport). This can both make the resulting products more accurate and repeatable by eliminating the possibility of inaccuracy introduced by manual operation error. However, the increased price associated with the automation and the need for custom software are, of course, the trade-off with this option. Additionally, the use of Grating Light Waves (GLV), Digital Micro-mirror Devices (DMD), Spatial Light Modulators (SLM), and Liquid Crystal Displays (LCD) as electrically controllable active photomasks has recently become increasingly common because it can significantly simplify the 3D structuring of the photoresist.[18, 19] In this case, a computer-generated pattern serves as a reconfigurable mask in the lithography system, rendering a mask alignment procedure unnecessary because the position of the photomask is always fixed. Although alignment of the wafer with the mask may still be necessary (e.g., if multiple coatings of different photoresists are required), the active-mask photolithography makes the pattern alignment precise at the exposure plane and allows for rapid fabrication of 3D microstructures. Among these, the LCD is an especially promising for high-resolution lithography, because of its small pixel size: 8.5m vs 17m for DMD.[18, 19] Moreover, the LCD mask enables "gray-tone" photolithography (useful for tapered devices)[20] by allowing the user to adjust the percent transmittance of each individual pixel in order to produce advanced 3D features. Other potential design modifications could include exposing the photoresist from the bottom by defining the mask as a part of the substrate.[21] By spinning the photoresist directly on to the mask, perfect contact alignment is achieved enabling the fabrication of tall high aspect ratio structures. Finally, recent exciting developments in LED technology such as the Nano-LED single photon lithography, could potentially be implemented in our CBMA once the technology matures.[22] The advantage of this LED type is that it can significantly improve the resolution (down to molecular scale) while at the same time eliminating the need for a physical photomask if arranged in an electronically driven individually addressable array. 15 S. Ma; D. N. Loufakis; Z. Cao; Y. Chang; L. E. K. Achenie; C. Lu, Lab Chip 2014, 14 Supporting Information Supporting Information is provided as an appendix to this manuscript. Acknowledgements We acknowledge the financial support from the Gustavus and Louise Pfeiffer Research Foundation and the National Science Foundation I-Corps Site Award #: 1450182. We also thank the Otto H. York Department of Chemical, Biological, and Pharmaceutical Engineering at New Jersey Institute of Technology for the help with machining custom parts for the CBMA. References [1]. (16), 2905. [2]. M. Junkin; Alicia J. Kaestli; Z. Cheng; C. Jordi; C. Albayrak; A. Hoffmann; S. Tay, Cell. Rep. 2016, 15 (2), 411. [3]. N. Ghorashian; S. K. Gokce; S. X. Guo; W. N. Everett; A. Ben-Yakar, PLoS One 2013, 8 (9), e74480. [4]. C. Scherber; A. J. Aranyosi; B. Kulemann; S. P. Thayer; M. Toner; O. Iliopoulos; D. Irimia, Integr. Biol. 2012, 4 (3), 259. [5]. R. Gómez-Sjöberg; A. A. Leyrat; D. M. Pirone; C. S. Chen; S. R. Quake, Anal. Chem. 2007, 79 (22), 8557. [6]. Y. Gao; J. Tian; J. Wu; W. Cao; B. Zhou; R. Shen; W. Wen, RSC Adv. 2016, 6 (104), 101760. [7]. K. L. Berkowski; K. N. Plunkett; Q. Yu; J. S. Moore, J. Chem. Educ. 2005, 82 (9), 1365. [8]. J. C. McDonald; D. C. Duffy; J. R. Anderson; D. T. Chiu; H. Wu; O. J. Schueller; G. M. Whitesides, Electrophoresis 2000, 21 (1), 27. [9]. 96. [10]. M. D. Huntington; T. W. Odom, Small 2011, 7 (22), 3144. [11]. M. Erickstad; E. Gutierrez; A. Groisman, Lab Chip 2015, 15 (1), 57. [12]. X. Li; Z. T. Yu; D. Geraldo; S. Weng; N. Alve; W. Dun; A. Kini; K. Patel; R. Shu; F. Zhang; G. Li; Q. Jin; J. Fu, Rev. Sci. Instrum. 2015, 86 (7), 075008. [13]. C. A. Schneider; W. S. Rasband; K. W. Eliceiri, Nat. Methods 2012, 9 (7), 671. [14]. M. Heymann; S. Fraden; D. Kim, J. Microelectromech. Syst. 2014, 23 (2), 424. [15]. N. G. Jain; E. A. Wong; A. J. Aranyosi; L. Boneschansker; J. F. Markmann; D. M. Briscoe; D. Irimia, Integr. Biol. 2015, 7 (11), 1423. [16]. Mater. 2014, 13 (11), 1063. [17]. D. Irimia; M. Toner, Integr. Biol. 2009, 1 (0), 506. [18]. T. Hayashi; T. Shibata; T. Kawashima; E. Makino; T. Mineta; T. Masuzawa, Sensors and actuators A: Physical 2008, 144 (2), 381. [19]. T. Horiuchi; S. Koyama; H. Kobayashi, Microelectronic Engineering 2015, 141, 37. [20]. V. Nock; R. J. Blaikie, Microelectronic Engineering 2008, 85 (5), 1077. [21]. C. P. Mark; H. Philip; D. M. Bloom; A. F. Harvey, J. Micromech. Microeng. 2003, 13 (3), 380. [22]. M. Mikulics; H. Hardtdegen, Nanotechnology 2015, 26 (18), 185302. J. P. Urbanski; W. Thies; C. Rhodes; S. Amarasinghe; T. Thorsen, Lab Chip 2006, 6 (1), I. Y. Wong; S. Javaid; E. A. Wong; S. Perk; D. A. Haber; M. Toner; D. Irimia, Nat. 16 Supporting Information A Compact Low-Cost Low-Maintenance Open Architecture Mask Aligner for Fabrication of Multilayer Microfluidics Devices Quang Long Pham, Austin Mathew, Nhat-Anh N. Tong, Roman S. Voronov* Figure S1. Photograph of the CBMA showing all of its major components: the light source, the microscope, the mask holder, and the wafer-chuck assembly. The light source and the Fresnel lens array are mounted on a supporting frame assembled from T-slotted 8020 aluminum extrusions. 17 Figure S2. Screenshots from two monitors, showing the live-view microscopy acquisition windows corresponding to the two alignment microscopes. Alignment and reference marks are positioned at the same Y location on each respective screen. (A) Two alignment/reference mark arrays being focused upon at a 30x magnification. (B) Individual alignment/reference marks being focused upon at a 245x magnification. Alignment marks (black color) of the on the left side and on the right side of the "device" mask were brought into same Y positions on each of the view screens. 18 Table S1. List of computer-aided design (CAD) parts; 3D files are provided upon request. 19
1908.03906	1	1908	2019-08-11T13:22:44	Spin Hall Magnetoresistance in Metallic Bilayers with In-plane Magnetized Ferromagnets	[ "physics.app-ph" ]	We revisit the theory and experiment on spin Hall magnetoresistance (SMR) in bilayers consisting of a heavy metal (H) coupled to in-plane magnetized ferromagnetic metal (F), and determine contributions to the magnetoresistance due to SMR and anisotropic magnetoresistance (AMR) in four different bilayer systems: W/$\text{Co}_{20}\text{Fe}_{60}\text{B}_{20}$, W/Co, $\text{Co}_{20}\text{Fe}_{60}\text{B}_{20}$/Pt, and Co/Pt. To do this, the AMR is explicitly included in the diffusion transport equations in the ferromagnet. The results allow precise determination of different contributions to the magnetoresistance, which can play an important role in optimizing prospective magnetic stray field sensors. They also may be useful in the determination of spin transport properties of metallic magnetic heterostructures in other experiments based on magnetoresistance measurements.	physics.app-ph	physics	Spin Hall Magnetoresistance in Metallic Bilayers with In-plane Magnetized Ferromagnets Łukasz Karwacki,1, 2, ∗ Krzysztof Grochot,1, 3, † Stanisław Łazarski,1 Witold Skowroński,1 Jarosław Kanak,1 Wiesław Powroźnik,1 Józef Barnaś,2, 4 Feliks Stobiecki,2 and Tomasz Stobiecki1, 3 1AGH University of Science and Technology, Department of Electronics, al. Mickiewicza 30, 30-059 Kraków, Poland 2Institute of Molecular Physics, Polish Academy of Sciences, ul. M. Smoluchowskiego 17, 60-179 Poznań, Poland 3Faculty of Physics and Applied Computer Science, AGH University of Science and Technology, al. Mickiewicza 30, 30-059 KrakÃşw, Poland 4Faculty of Physics, Adam Mickiewicz University, ul. Uniwersytetu Poznańskiego 2, 61-614 Poznań, Poland We revisit the theory and experiment on spin Hall magnetoresistance (SMR) in bilayers consist- ing of a heavy metal (H) coupled to in-plane magnetized ferromagnetic metal (F), and determine contributions to the magnetoresistance due to SMR and anisotropic magnetoresistance (AMR) in four different bilayer systems: W/Co20Fe60B20, W/Co, Co20Fe60B20/Pt, and Co/Pt. To do this, the AMR is explicitly included in the diffusion transport equations in the ferromagnet. The results allow precise determination of different contributions to the magnetoresistance, which can play an important role in optimizing prospective magnetic stray field sensors. They also may be useful in the determination of spin transport properties of metallic magnetic heterostructures in other experiments based on magnetoresistance measurements. I. INTRODUCTION Spin Hall magnetoresistance (SMR) is a phenomenon that consists in resistance dependence on the relative ori- entation of magnetization and spin accumulation at the interface of ferromagnet and strong spin-orbit material (such as 5d metals [1 -- 6], topological insulators [7], or some 2D systems [8]). In such transition metals as W and Pt, the spin accumulation results from spin current driven by the spin Hall effect (SHE) [9 -- 12]. The spin current diffuses then into the ferromagnet or exerts a torque on the magnetization while being backscattered. Due to the inverse spin Hall effect (ISHE), the backscat- tered spin current is converted into a charge current that flows parallel to the bare charge current driven by ex- ternal electric field, which effectively reduces the resis- tance [3]. One of the most important advantages of driv- ing spin currents by SHE is that the spin currents can be induced by a charge current flowing in the plane of the sample [13]. This may remedy some obstacles on the road to further miniaturization of prospective electronic components, which have been encountered in spin-valves and magnetic tunnel junctions when the electric field is applied perpendicularly to interfaces. One of the draw- backs, however, is that the strength and effectiveness of such subtle effects depend strongly on the quality and spin properties of interfaces [14 -- 21]. Although early SMR experiments were performed on heavy-metal/ferromagnetic-insulator bilayers [1], recent efforts are focused on the bilayers with ferromagnetic metallic layers, such as Co or Co20Fe60B20 ones [4, 6], ∗ [email protected] † [email protected] which are currently more relevant for applications. When the magnetization is parallel to the spin accumulation, the spin current from the heavy-metal can easily diffuse into the ferromagnetic metal (influencing its spin trans- port properties and spin accumulation on the ferromag- netic metal side) [4, 22 -- 28]. This is especially impor- tant when an additional spin sink (another heavy-metal layer or an antiferromagnet) is on the other side of the ferromagnetic layer, where effects such as spin current interference might take place [5]. Moreover, as charge current flows in plane of the sample, additional phenomena may occur, such as anisotropic magnetoresistance (AMR) or anomalous Hall effect (AHE) [29 -- 34]. These effects can obscure determi- nation of spin transport parameters and make evalua- tion of the SMR contribution to the measured magne- toresistance more difficult. Since the determination of such transport properties as the spin Hall angle (which parameterizes strength of the spin Hall effect) and spin diffusion length in different experimental schemes, for instance in spin-orbit torque ferromagnetic resonance (SOT-FMR) [35, 36], relies heavily on the magnetore- sistive properties of a system, it is important to properly determine all the contributions to magnetoresistance. Here, we revisit the theory of spin Hall magnetore- sistance in metallic bilayers by explicitly including the contributions from AMR and AHE into the spin drift- diffusion theory for the ferromagnetic metal layer. The expressions for magnetoresistance are then fitted to the data obtained from resistance meansurements on heavy- metal (H)/ferromagnet (F) bilayers, where H: W, Pt, while F: Co, Co20Fe60B20. This allows us to deter- mine more accurately contributions from various mag- netotransport phenomena occuring in metallic bilayers where the spin Hall effect is the driving source. Such analysis may also be useful in the efforts to optimize x ^ m z Capping layer (C) Ferromagnetic metal (F) Heavy metal (H) Substrate tF t H y Ex FIG. 1. Schematic representation of the system considered in this paper. The system consists of heavy-metal (H) and ferromagnetic-metal (F) layers. The bilayer is deposited on a substrate and capped with a non-conductive layer (C). Thick- nesses tH and tF of the H and F layers are varied, in general. Electric field, Ex, is applied along the x axis, while the unit vector m along magnetization of the ferromagnet is oriented in the plane and at an angle ϕ to the x axis. prospective devices for information technology. The paper is organized as follows: Sec. II contains the- oretical derivation of the formulas for multilayer magne- toresistance with explicit AMR and AHE contributions from metallic ferromagnet and SMR from H/F interface. Section III contains experimental details on the resis- tance and resistivity of the samples studied in this paper. Section IV contains results and discussion of the to ex- perimental data on magnetoresistance. Finally, in Sec. V we briefly summarize the paper. II. THEORY Figure 1 shows schematic representation of the system examined here. Note that the electric field in this figure is oriented along the axis x, but in order to properly capture the planar Hall effect in calculations we consider a more general orientation of the field, E = Ex ex + Ey ey, (1) where ex and ey are unit vectors along the axes x and y, respectively. The spin current density tensor (measured in the units of charge current density) in the heavy-metal (H) layer can be written as follows: 2e q H s = Xi eijH si , (2) + where i = x, y, z, and in the r.h.s. we used the convention for dyadic vector products according to which the first vector denotes the flow direction, and the second vector Note, in the above equations for the current densities in both H and F layers we assumed linear response to 2 describes the spin polarization and magnitude of the spin current. Only the component flowing along the normal to interfaces is relevant and will be taken into account in the following, i.e. ez · 2e s (z), where H s ≡ jH q jH s (z) = − θSH ρH 0 ez × E + 1 2eρH 0 ∂µH s (z) ∂z . (3) Here θSH is the spin Hall angle, ρH of the heavy metal, and µH that is generally z-dependent. 0 is the bare resistivity s (z) is the spin accumulation The charge current density in the heavy-metal (H) layer, in turn, can be written in the form jH c (z) = 1 ρH 0 E + θSH 2eρH 0 ez × ∂µH s (z) ∂z , (4) and contains the bare charge current density and the cur- rent due to inverse spin Hall effect. Note, that the spin current in general can induce charge current also flow- ing along the axes x and y. However, due to lateral dimensions of the samples much larger than the layer thicknesses and spin diffusion lengths, those additional components can be neglected. Analogously to the above, we define the spin current density tensor (in the units of charge current density) in the ferromagnetic layer (F) as 2e q F s = jF s m . (5) However, now the first vector on r.h.s. describes direction flow and magnitude of spin current while the second one spin polarization, which now is along the magnetization. Thus, one can write 2e s is given by the equation [26, 27]: F s · m = jF s , where jF q jF s (z) = + 1 2eρF 0 θAH 2eρF 0 ∇µF s (z) + β 2eρF 0 ∇µF c (r) m × ∇µF s (z) − θAMR 2eρF 0 m(cid:2) m · ∇µF s (z)(cid:3) , (6) in which θAH and θAMR are the anomalous Hall angle and AMR angle, defined as θAH = σAHρF 0 and θAMR = σAMRρF c (z) is the electrochemical potential. 0 , respectively, while µF c (r) = 2eE · r + µF Charge current density in the ferromagnetic layer (F) can be written as [26, 27], jF c (z) = 1 2eρF 0 θAH 2eρF 0 ∇µF c (r) + β 2eρF 0 ∇µF s (z) m × ∇µF c (r) − θAMR 2eρF 0 m(cid:2) m · ∇µF c (r)(cid:3) . (7) electric field, i.e. we neglected the so-called unidirectional spin Hall magnetoresistance effect [22 -- 25]. The interfacial spin current density tensor can be writ- ten as 2e q HF s = ezjHF s , (8) where the spin current jHF flowing through the heavy- metal/ferromagnet interface is given by the following ex- pression [37]: s jHF s = GF(cid:20)(cid:0)µF s − µH s (cid:1) · m(cid:21) m + Gi m × µH s (0) + Gr m × m × µH s (0) . (9) Here GF = (1 − γ2)(G↑ + G↓)/2 with γ defined as γ = (G↑ − G↓)/(G↑ + G↓) and G↑ and G↓ denoting the interface conductance for spin-↑ and spin-↓. Further- more, Gr ≡ Re Gmix and Gi ≡ Im Gmix, where Gmix is the so-called spin-mixing conductance. Note, that we neglect explicitly a contribution from the interfacial Rashba-Edelstein spin polarization [36]. A strong inter- facial spin-orbit contribution which induces spin-flip pro- cesses can also be combined with the interfacial spin con- ductance GF as a spin-conductance reducing parameter 1 − η, with η = 0 for no interfacial spin-orbit coupling, and η = 1 for maximal spin-orbit coupling. Note, that this reduction could also be attributed to the magnetic proximity effect, especially in the case of Pt-based het- erostructures [11], however recent studies suggest its ir- relevance for spin-orbit-torque -- related experiments [20]. In the following discussion we assume η = 0 and treat GF as an effective parameter. To find charge and spin currents we need to find first the spin accumulation at the H/F interface and also at external surface/interfaces. This can be found from the following boundary conditions: jH s (z = −tH ) = 0 , jF s (z = tH + tF ) = 0 , jH s (z = 0) = jHF s,z(z = 0) = jHF jF · m . , s s (10a) (10b) (10c) (10d) Having found spin accumulation and also electrochemical potential, one can find the longitudinal (l) and transver- sal (t) in-plane components of the averaged charge cur- rent j( m) from the formula: jl(t)( m) = dz ex(y) · jH c (z) 1 tH + tF (cid:20)ZtH +ZtF dz ex(y) · jF c (z)(cid:21) . (11) The total charge current can be written down in the Ohm's-law form, j( m) = σ( m)E , (12) where the conductivity (resistivity) tensor takes the form: 3 σ( m) ≡ (cid:2)ρ( m)(cid:3)−1 = (cid:20)σ0 + σxm2 mxmyσxy x + σym2 y mxmyσxy σ0 + σym2 x + σxm2 y(cid:21) , with (13) tanh(cid:18) tH 2λH(cid:19) + ρF 0 tH + ρH ρF 0 ρH 0 tF + ρF 0 tF 0 ρH 0 tH tF + tH 0 tF 0 ρH tanh(cid:18) tH 0 tH 2λH(cid:19) − θAMR tF ρF 0 tF + tH , (14) (15) (16) (17) σ0 = −2 σx = σy = ≈ λH θ2 SH ρH 0 ρF 0 tH + ρH 0 tF + ρF 0 ρH ρF θ2 gDLλH SH ρH tF + tH 0 = σSH x + σAMR θ2 gF H λH SH ρH tF + tH 0 θ2 gH F λF AH ρF tF + tH 0 θ2 AH ρF 0 tF + tH λF + 2 − x , + tF θ2 AH ρF tF + tH 0 y + σAH = σSH , y σxy = σx + σy . tanh(cid:18) tH 2λH(cid:19) 2λF(cid:19) tanh(cid:18) tF 2λF(cid:19) tanh(cid:18) tF (1 − β2) β2 β2 (1 − β2) In the above expressions the following dimensionless co- efficients have been introduced to simplify the notation: λH(cid:19)(cid:21) gr(1 + gr) + g2 i (1 + gr)2 + g2 i gDL = (cid:20)1 − sech(cid:18) tH gr,i = 2Gr,iρH H = (cid:20)1 − sech(cid:18) tH F = (cid:20)1 − sech(cid:18) tF 0 λH coth(cid:18) tH λH(cid:19)(cid:21) λF(cid:19)(cid:21) gH gF λH(cid:19) , , 1 2GF λHρH 0 2GF λFρF 0 1 1 1 +(cid:16) 1 +(cid:16) coth(cid:18) tF coth(cid:18) tH λF(cid:19) , λH(cid:19) . , + γF 1 + γH λH(cid:17) H(cid:17) tanh(cid:16) tH λF(cid:17) F(cid:17) tanh(cid:16) tF , (18) (19) γF H = γH F = λHρH λHρH λFρF 0 0 (1 − β2) 0 (cid:0)1 − β2(cid:1) λFρF 0 With the resistivity defined in Eq. (13) we can now define magnetoresistance, M R = ρxx( m k ex) − ρxx( m k ey) ρxx( m k ex) , (20) (a) W/F tF (nm) 4 6 0 2 8 10 (b) 0 2.0 ) k ( R (c) ) k ( R 2.0 1.5 1.0 0.5 0.0 0 0 0.4 0.3 0.2 0.1 0.0 0 F: Co20Fe60B20 2.0 1.5 1.5 1.0 1.0 2 2 4 6 tH (nm) tF (nm) 4 6 8 8 F: Co 0.5 0.5 0.0 10 0.0 0 (d) 10 1.6 0 0.4 1.2 0.3 0.8 0.4 0.2 0.1 0.0 10 0.0 0 2 4 6 tH (nm) 8 F/Pt tF (nm) 6 4 2 8 10 F: Co20Fe60B20 6 4 2 2 2 4 6 tH (nm) tF (nm) 6 4 0 10 8 8 10 0.8 F: Co 0.6 0.4 0.2 0.0 10 8 2 4 6 tH (nm) FIG. 2. Resistance R as a function of thickness tH for in- dicated heavy metal layers (black squares): W [(a),(c)], Pt [(b),(d)], and as a function of thickness tF for ferromagnets (blue dots): Co20Fe60B20 [(a),(b)] and Co [(c),(d)]. where ρxx( m) ≡ ρ( m) · ex. Taking into account Eqs. (14)-(16), the above formula can be written as, M R ≈ σy − σx σ0 . (21) In order to compare the models with and without AMR and AHE, we define SMR as: SM R = −M R(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)θAMR→0 θAH→0 , (22) which simplifies our model to that introduced by Kim et. al [4]. 4 TABLE I. Composition of samples, and resistivities of heavy metal and ferromagnetic layers. Numbers in parentheses de- note thickness (in nm) of the corresponding layer. No. Sample ρH 0 ρF 0 (µΩcm) (µΩcm) W(5)/Co(tF )/Ta(1) W(tH)/Co(5)/Ta(1) W1 W(5)/Co20Fe60B20(tF )/Ta(1) W2 W(tH)/Co20Fe60B20(5)/Ta(1) W3 W4 P1 P2 P3 P4 Co20Fe60B20(tF )/Pt(3) Co20Fe60B20(5)/Pt(tH ) Co(tF )/Pt(4) Co(5)/Pt(tH ) 185 166 120 120 95 151 55 24 144 144 22 30 102 161 18 57 After deposition, multilayered systems were nanostruc- tured using either electron-beam lithography or optical lithography, ion etching and lift-off. The result was a ma- trix of Hall bars and strip nanodevices for further elec- trical measurements. The sizes of produced structures were: 100 µm x 10 µm or 100 µm x 20 µm. In order to ensure good electrical contact with the Hall bars and strips, Al(20)/Au(30) contact pads with dimensions of 100 µm x 100 µm were produced. Appropriate placement of the pads allows rotation of the investigated sample and its examination at any angle with respect to the exter- nal magnetic field in a dedicated rotating probe station using a four-points probe. The constant magnetic field, controlled by a gaussmeter exceeded magnetization satu- ration in plane of the sample and the sample was rotated in an azimuthal plane from -120◦ to +100◦. The resistance of the system was measured with a two- and four-point technique using Keithley 2400 sourceme- ters and Agilent 34401A multimeter. As shown in Fig. 2, resistances of bilayers with amorphous ferromag- net Co20Fe60B20 are about one order higher than these with polycrystalline Co. The same results were obtained using both techniques. The thickness-dependent resis- tivity of individual layers was determined by method described in Ref. [6], and by a parallel resistors model. For more details on resistivity measurements we refer the reader to Supplemental Material [39]. III. EXPERIMENT IV. RESULTS AND DISCUSSION Table I shows the multilayer systems that were pro- duced for SMR studies. The magnetron sputtering tech- nique was used to deposit multilayers on the Si/SiO2 thermally oxidized substrates. Thickness of wedged lay- ers were precisely calibrated by X-ray reflectivity (XRR) measurements. The details of sputtering deposition pa- rameters as well as structural phase analysis of highly resistive W and Pt layers can be found in our recent pa- pers [36, 38]. In turn, structure analysis of the Co crys- tal phases grown on disoriented β-W can be found in the Supplemental Material [39]. Table II shows parameters used for fitting the model to the experimental data on magnetoresistance. In order to simplify the analysis, we assumed thickness-independent interfacial spin-mixing conductances. However, it is well- known from the spin-pumping theory [40] that the in- terfacial spin-mixing conductance is strongly dependent on Gilbert damping and saturation magnetization of the ferromagnet, which, in turn, depend on thickness of the layer. These parameters vary strongly mostly for thin layers, while they saturate for thicker ferromagnets. Moreover we assumed transparent contacts for parallel 0 4 2 0 0 (a) ) ( R (e) ) % ( R M (i) ) % ( R M 0.6 0.4 0.2 0.0 0.8 0.6 0.4 0.2 0.0 W/Co20Fe60B20 tF (nm) 4 6 2 (b) 8 10 W/Co tF (nm) 4 6 2 (c) 8 10 6 0 2 1 4 2 10 8 6 4 2 2 4 6 tH (nm) 0 10 8 (f) 0 0 2 4 6 tH (nm) 8 0 10 (g) 0.9 0.6 0.3 0 2 4 6 tH (nm) 8 10 (j) 0 2 4 6 tF (nm) 8 10 0.0 1.2 0.9 0.6 0.3 0.0 0 2 4 6 tH (nm) 8 10 (k) x(-1) 0 2 4 6 tF (nm) 8 10 Co20Fe60B20/Pt tF (nm) 4 6 0 2 (d) 8 10 0 4 2 6 4 2 0 0.4 0.3 0.2 0.1 0.0 0.6 0.4 0.2 0.0 8 6 4 2 0 2 4 6 tH (nm) 8 0 10 (h) 2 0 0 2.5 2.0 1.5 1.0 0.5 0.0 0.6 0.4 0.2 0.0 0 2 4 6 tH (nm) 8 10 (l) 0 2 4 6 tF (nm) 8 10 Co/Pt tF (nm) 4 6 8 10 5 1.0 0.8 0.6 0.4 2 4 6 tH (nm) 8 10 , Experiment Theory -SMR AMR 0 2 4 6 tH (nm) 8 10 x(-1) 0 2 4 6 tF (nm) 8 10 FIG. 3. Magnetoresistance ∆R as a function of both H (black squares) and F (blue dots) layers thicknesses [(a)-(d)]; relative magnetoresistance M R for: W/Co20Fe60B20 [(e),(i)], W/Co [(f),(j)], Co20Fe60B20/Pt [(g),(k)], and Co/Pt [(h),(l)]. TABLE II. Parameters used for fitting the model to experi- mental data for samples defined in Tab. I. θSH θAMR (%) Gr (Ω−1m−2) λH (nm) λF (nm) No. W1 0.32 W2 0.28 W3 0.34 W4 0.47 P1 0.25 0.09 P2 0.09 P3 P4 0.28 0.14 0.13 1.1 0.5 0.2 0.35 2.4 0.7 1013 1014 1017 1013 1011 1011 1013 1014 1.3 1.3 1.3 1.3 2.2 2.2 2.2 2.2 1 2 5 5 2.5 2.5 5 5 spin transport, i.e. GF → ∞, and also assumed Gi to be neglible. Both assumptions are valid for metallic in- terfaces and the fitted parameters should be understood as upper limits. Although the theory predicts influence of AHE on the magnetoresistance, we neglect it in fur- ther discussion, as the so-called anomalous Hall angle is small in ferromagnets considered here, but should play an important role in out-of-plane magnetized systems or in- plane magnetized systems with more significant anoma- lous Hall angle. Finally, we assumed spin polarization β = 0.3 for both Co and Co20Fe60B20. Figures 3(a) -- (d) show magnetoresistance as a function of both heavy metal and ferromagnetic metal layer thick- nesses, while Figs. 3(e) -- (l) show relative magnetoresis- tance, on which we will focus our further discussion. As expected, SMR is larger in heterostructure with W as a heavy metal layer, than in the heterostructure with Pt, due to generally larger spin Hall angle of W, −θSH ≈ 0.3 -- 0.4, compared to Pt, θSH ≈ 0.1 -- 0.3. On the other hand, AMR is stronger in Co, θAMR ≈ 0.5 -- 2.4%, than in Co20Fe60B20, for which θAMR ≈ 0.13 -- 0.35%. Note, that we assume θAMR and θSH to be independent of layer thickness, which may result in overestimated parameters for very thin ferromagnetic layers. Due to relatively high spin Hall angle in W, magne- toresistance as a function of heavy-metal layer thickness, shown in Figs. 3(e) and 3(f), has a large SMR compo- nent which is qualitatively and quantitatively similar for both W/Co20Fe60B20 and W/Co bilayers. In W/Co, however, the total magnetoresistance is larger, due to larger AMR contribution of Co. In Pt bilayers, on the other hand, small spin Hall angle results in magnetoresis- tance dominated mostly by AMR contribution, as shown in Figs. 3(g) and 3(h). The dependence of magnetoresistance on ferromag- netic layer thickness for fixed thickness of heavy-metal is shown in Figs. 3(i) -- (l). For bilayers with layers' Co20Fe60B20, shown in Figs. 3(i) and 3(k), AMR con- tribution is weaker than SMR for both W and Pt. In the case of Co-based heterostructures, shown in Figs. 3(j) and 3(l), however, we observe negative SMR, which is domi- nated by parallel spin transport through the H/F inter- face. In this case large AMR might lead to strong over- estimation of the relevant transport parameters. More- over, the discrepancies between the model and the data can be attributed, as mentioned before, to varying spin- mixing conductance for small thickness of the ferromag- netic metal layer, possible magnetic dead layer, and mag- netization that does not lie completely in-plane of the sample. [1] M. Althammer, S. Meyer, H. Nakayama, M. Schreier, S. Altmannshofer, M. Weiler, H. Huebl, S. Geprägs, M. Opel, R. Gross, D. Meier, C. Klewe, T. Kuschel, J.-M. Schmalhorst, G. Reiss, L. Shen, A. Gupta, Y.-T. Chen, G. E. W. Bauer, E. Saitoh, and S. T. B. Goennen- wein, Quantitative study of the spin Hall magnetoresis- tance in ferromagnetic insulator/normal metal hybrids, Phys. Rev. B 87, 224401 (2013). [2] H. Nakayama, M. Althammer, Y.-T. Chen, K. Uchida, Y. Kajiwara, D. Kikuchi, T. Ohtani, S. Geprägs, M. Opel, S. Takahashi, R. Gross, G. E. W. Bauer, S. T. B. Goennenwein, and E. Saitoh, Spin Hall magne- toresistance induced by a nonequilibrium proximity ef- fect, Phys. Rev. Lett. 110, 206601 (2013). [3] Y.-T. Chen, S. Takahashi, H. Nakayama, M. Al- thammer, S. T. B. Goennenwein, E. Saitoh, and G. E. W. Bauer, Theory of spin Hall magnetoresis- tance, Phys. Rev. B 87, 144411 (2013); Theory of spin Hall magnetoresistance (SMR) and related phenomena, J. Phys.: Condens. Matt. 28, 103004 (2016). [4] J. Kim, P. Sheng, S. Takahashi, S. Mitani, and M. Hayashi, Spin Hall magnetoresistance in metallic bi- layers, Phys. Rev. Lett. 116, 097201 (2016). [5] J.-G. Choi, J. W. Lee, and B.-G. Park, Spin Hall magne- toresistance in heavy-metal/metallic-ferromagnet multi- layer structures, Phys. Rev. B 96, 174412 (2017). V. SUMMARY 6 In conclusion, we have developed an extended model of magnetoresistance for magnetic metallic bilayers with in-plane magnetized ferromagnets, which explicitly treats both SMR and AMR contributions. The model was then fitted to experimental data on magnetoresistance in W/Co20Fe60B20, W/Co, Co20Fe60B20/Pt, and Co/Pt heterostructures to estimate the strength of SMR and AMR effects. These results allow for a more accurate es- timation of different contributions to magnetoresistance in magnetic metallic systems, which is important for ap- plications for instance in spintronic memory read-heads or in other experimental schemes that rely on magnetore- sistance measurements in evaluation of the spin transport properties. ACKNOWLEDGMENTS Science Centre This work is supported by the National Science Cen- in Poland Grant UMO-2016/23/B/ST3/01430 tre Na- (SPINORBITRONICS). WS in Poland Grant No. tional 2015/17/D/ST3/00500. The nanofabrication process was performed at the Academic Centre for Materials and Nanotechnology (ACMiN) of AGH University of Science and Technology. We would like to thank M. Schmidt and J. Aleksiejew for technical support. acknowledges [6] M. Kawaguchi, D. Towa, Y. C. Lau, S. Takahashi, and M. Hayashi, Anomalous spin Hall magnetoresistance in Pt/Co bilayers, Appl. Phys. Lett. 112, 202405 (2018). [7] Y. Lv, J. Kally, D. Zhang, J. S. Lee, M. Ja- mali, N. Samarth, and J.-P. Wang, Unidirectional spin-hall and rashba-edelstein magnetoresistance in topological insulator-ferromagnet layer heterostructures, Nature Communications 9, 111 (2018). [8] K. Narayanapillai, G. Go, R. Ramaswamy, K. Gopinad- han, D. Go, H.-W. Lee, T. Venkatesan, K.-J. Lee, and H. Yang, Interfacial rashba magnetoresistance of the two- dimensional electron gas at the LaAlO3 / SrTiO3 inter- face, Physical Review B 96, 064401 (2017). Hall effect, [9] J. Spin Phys. Rev. Lett. 83, 1834 (1999). Hirsch, E. [10] M. I. Dyakonov, Magnetoresistance due to edge spin ac- cumulation, Phys. Rev. Lett. 99, 126601 (2007). [11] S. Y. Huang, X. Fan, D. Qu, Y. P. Chen, W. G. Wang, J. Wu, T. Y. Chen, J. Q. Xiao, and C. L. Chien, Transport magnetic proximity effects in platinum, Phys. Rev. Lett. 109, 107204 (2012). S. O. Valenzuela, J. Wunderlich, C. H. Back, and T. Jungwirth, Spin Hall effects, Rev. Mod. Phys. 87, 1213 (2015). Sinova, [12] J. [13] Y. Liu, B. Zhou, and J.-G. Zhu, Field-free mag- netization switching by utilizing the spin hall ef- fect and interlayer iridium, coupling of exchange Scientific Reports 9, 325 (2019). IEEE Trans. Magn. 11, 1018 (1975). 7 [31] J. Bass and W. P. Pratt, Spin-diffusion lengths in metals and alloys, and spin-flipping at metal/metal interfaces: review, J. Phys.: Condens. Mat. 19, 183201 (2007). experimentalist's critical an [32] P. S. Bechthold, from Galvanomagnetic effect to Hall port: Spintronics -- From GMR to Quantum Information, edited by S. Blügel, D. Bürgler, M. Morgenstern, C. M. Schneider, and R. Waser, Institute of Solid State Re- search, Forschungszentrum Jülich (Forschungszentrum Jülich GmbH, Jülich, 2009). AMR, trans- in [33] N. Nagaosa, Donald, and N. P. Ong, Anomalous Hall Rev. Mod. Phys. 82, 1539 (2010). J. Sinova, S. Onoda, A. H. Mac- effect, [34] S. Iihama, T. Taniguchi, K. Yakushiji, A. Fukushima, Y. Shiota, S. Tsunegi, R. Hiramatsu, S. Yuasa, Y. Suzuki, and H. Kubota, Spin-transfer torque induced by the spin anomalous hall effect, Nature Electronics 1, 120 (2018). [35] L. 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). [36] W. Skowroński, Ł. Karwacki, S. Ziętek, J. Kanak, S. Łazarski, K. Grochot, T. Stobiecki, P. Kuświk, F. Stobiecki, of heavy-metal/Co-Fe-B-based spin Hall interfacial heterostructures with fields, Phys. Rev. Appl. 11, 024039 (2019). and J. Barnaś, Determination spin-orbit angle in [37] A. Brataas, G. Bauer, and P. Kelly, Non-collinear mag- netoelectronics, Phys. Rep. 427, 157 (2006). [38] S. Łazarski, W. Skowroński, J. Kanak, L. Karwacki, S. Ziętek, K. Grochot, T. Stobiecki, and F. Sto- biecki, in Co/Pt/Co multilayer with mixed magnetic anisotropies, Phys. Rev. Appl. 12, 014006 (2019). spin-orbit-torque Field-free switching [39] See Supplemental Material at [URL will be inserted by publisher] for more details on structural characterization of the samples and on resistivity measurements. Supple- mental Material contains Refs. [6, 36, 38]. [40] Y. Tserkovnyak, A. Brataas, and G. E. W. Bauer, Spin pumping and magnetization dynamics in metallic multi- layers, Phys. Rev. B 66, 224403 (2002). [14] A. Kobs, S. Hesse, W. Kreuzpaintner, G. Winkler, D. Lott, P. Weinberger, A. Schreyer, and H. P. Oepen, Anisotropic interface magnetoresistance in Pt/Co/Pt sandwiches, Phys. Rev. Lett. 106, 217207 (2011). [15] C.-F. Pai, Y. Ou, L. H. Vilela-Leão, D. C. Ralph, and R. A. Buhrman, Dependence of the efficiency of spin Hall torque on the transparency of Pt/ferromagnetic layer in- terfaces, Phys. Rev. B 92, 064426 (2015). [16] W. Zhang, W. Han, X. Jiang, S.-H. Yang, and S. S. P. Parkin, Role of transparency of platinum -- ferromagnet interfaces in determining the intrinsic magnitude of the spin Hall effect, Nat. Phys. 11, 496 (2015). [17] S. S.-L. Zhang, G. Vignale, and S. Zhang, Anisotropic magnetoresistance driven by surface spin-orbit scatter- ing, Phys. Rev. B 92, 024412 (2015). [18] M. Tokaç, M. Wang, S. Jaiswal, A. W. Rushforth, B. L. Gallagher, D. Atkinson, and A. T. Hindmarch, Interfacial contribution to thickness dependent in-plane anisotropic magnetoresistance, AIP Advances 5, 127108 (2015). [19] M. Tokaç, S. A. Bunyaev, G. N. Kakazei, D. S. Schmool, D. Atkinson, and A. T. Hindmarch, Interfacial structure dependent spin mixing conductance in cobalt thin films, Phys. Rev. Lett. 115, 056601 (2015). [20] L. J. Zhu, D. C. Ralph, and R. A. Buhrman, to spin- in heavy-metal/ferromagnet bilayers, Irrelevance of magnetic proximity effect orbit Phys. Rev. B 98, 134406 (2018). torques [21] V. P. Amin, Interface-generated J. Stiles, Phys. Rev. Lett. 121, 136805 (2018). spin Zemen, and M. D. currents, [22] C. O. Avci, K. Garello, A. Ghosh, M. Gabureac, S. F. Alvarado, and P. Gambardella, Unidirectional spin Hall magnetoresistance in ferromagnet/normal metal bilayers, Nat. Phys. 11, 570 (2015). [23] C. O. Avci, K. Garello, J. Mendil, A. Ghosh, N. Blasakis, M. Gabureac, M. Trassin, M. Fiebig, of and heavy bilayers, Appl. Phys. Lett. 107, 192405 (2015). metal/ferromagnet Magnetoresistance Gambardella, light and P. [24] S. S.-L. Zhang and G. Vignale, Theory unidirectional in Phys. Rev. B 94, 140411(R) (2016). heavy-metal/ferromagnetic-metal Hall spin of magnetoresistance bilayers, [25] C. O. Avci, P. Gambardella, Origins spin Hall magnetoresistance Phys. Rev. Lett. 121, 087207 (2018). J. Mendil, G. S. D. Beach, and unidirectional in metallic bilayers, the of [26] T. Taniguchi, J. Grollier, Stiles, Spin-transfer anoma- lous Hall effect and anisotropic magnetoresistance, Phys. Rev. Appl. 3, 044001 (2015). and M. D. generated torques the by [27] T. Taniguchi, Magnetoresistance charge-spin conversion by in metallic Phys. Rev. B 94, 174440 (2016). ferromagnetic/nonmagnetic generated anomalous Hall from effect bilayers, [28] Y. Yang, Z. Luo, H. Wu, Y. Xu, R. W. Li, S. J. Penny- cook, S. Zhang, and Y. Wu, Anomalous Hall magnetore- sistance in a ferromagnet, Nat. Commun. 9, 2255 (2018). [29] J.-P. Jan, Galvamomagnetic and thermomagnetic effects in metals, in Solid State Physics (Elsevier, 1957) pp. 1 -- 96. [30] T. McGuire netoresistance and R. Potter, Anisotropic mag- alloys, ferromagnetic 3d in
1802.00207	1	1802	2018-02-01T09:31:20	Highly Deformable and Mobile Palladium Nanocrystals as Efficient Carbon Scavengers	[ "physics.app-ph", "cond-mat.mes-hall" ]	Fouling of surfaces leads to performance degradation in many energy-intensive industrial processes, but the present solutions are either too complicated to be routinely used or incomplete for eradication. Here we propose and demonstrate that carbon-containing deposits can be catalytically wiped out in an efficient way by roaming palladium nanoparticles with extreme shape flexibility at relatively low temperatures. Surprisingly, during their dramatic liquid-like migrations, these particles could still maintain crystalline interior and conserve their initial crystal orientations through self-surface diffusion. Moreover, these catalytic particles were even able to become regenerated by other roaming particles after occasionally deactivated by surface coking or multiple-particle sintering. These findings shed light on metabolically driven, "living" nanocrystals, and also open a new avenue for efficient catalysis.	physics.app-ph	physics	Highly Deformable and Mobile Palladium Nanocrystals as Efficient Carbon Scavengers Peng-Han Lu1†, De-Gang Xie1, Bo-Yu Liu1, Fei Ai1, Zhao-Rui Zhang2, Ming-Shang Jin2, Xiao Feng Zhang3, Evan Ma1,4, Ju Li1,2,5* and Zhi-Wei Shan1* 1 Center for Advancing Materials Performance from the Nanoscale (CAMP-Nano) and XJTU- HHT Research & Development Center (XHRDC), State Key Laboratory for Mechanical Behavior of Materials, Xi'an Jiaotong University, Xi'an 710049, China 2 Frontier Institute of Science and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi 710054, China 3 Hitachi High Technologies America, Pleasanton, CA 94588, USA 4 Department of Materials Science and Engineering, Johns Hopkins University, Baltimore, MD 21218, USA 5 Department of Nuclear Science and Engineering and Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA † Present address: Ernst Ruska-Centre for Microscopy and Spectroscopy with Electrons and Peter Grünberg Institute, Forschungszentrum Jülich, 52425 Jülich, Germany E-mail: [email protected]; [email protected] KEYWORDS: Metallic nanoparticles, heterogeneous catalysis, carbon removal, surface diffusion, environmental transmission electron microscopy. 1 Fouling of surfaces leads to performance degradation in many energy-intensive industrial processes, but the present solutions are either too complicated to be routinely used or incomplete for eradication. Here we propose and demonstrate that carbon-containing deposits can be catalytically wiped out in an efficient way by roaming palladium nanoparticles with extreme shape flexibility at relatively low temperatures. Surprisingly, during their dramatic liquid-like migrations, these particles could still maintain crystalline interior and conserve their initial crystal orientations through self-surface diffusion. Moreover, these catalytic particles were even able to become regenerated by other roaming particles after occasionally deactivated by surface coking or multiple-particle sintering. These findings shed light on metabolically driven, "living" nanocrystals, and also open a new avenue for efficient catalysis. Carbon deposition on hot surfaces (e.g. in internal combustion engine, solid oxide fuel cell, electrochemical fuel production, etc.) usually leads to a measurable drop of energy efficiency and performance1,2. These carbon deposits originate from heat-induced decomposition of unconsumed vaporized fuel into solid fouling deposits. In a motor engine, components such as injectors, pistons, valves, spark plugs, throttle body, intake manifold, catalytic converter, and oxygen sensors etc., may all be potentially fouled by carbon2. Thus to bring the engine back to optimum combustion efficiency, routine maintenance is needed to clean up the fouling deposits, by adding chemical fuel additives3,4 into the fuel or by immersing the components into ultrasonic baths. Adopting new engineering designs might also be helpful, but these are still incomplete solutions5. 2 Catalytic oxidation could be an alternative way to eradicate fouling. However, distinct from many reactions where all reactant gases flow over anchored nanoparticles6,7, catalytic defouling would involve a solid carbonaceous reactant besides gaseous oxygen. Given that carbon adheres to the surface and does not flow like gases, mobile catalytic nanoparticles would be necessary. Moreover, high shape flexibility is also essential for particles to stay adhered to the fouled surface while eating their ways forward. Finally, these particles must be able to maintain their catalytically active sites, such as certain crystalline facets, as they move. The past decade has witnessed an explosion of elegantly synthesized metal nanocrystals with controlled sizes, shapes, compositions and structures8,9. Indeed, metal nanocatalysts often experience dynamic evolution in response to the surrounding gaseous environment, including sintering/redispersion10,11, shape transition12,13, surface reconstruction14-17, surface segregation18- 20, phase transformation21,22, etc. These dynamic changes usually correlate with their catalytic reactivity. For example, the morphology change of metal nanocrystals was found to bring about a transient23 or oscillatory24 reaction kinetics. Inspired by the latest findings that metal nanostructures can experience fast geometry transformation through surface diffusion25,26, one may alternatively take advantage of these dynamic changes to mobilize catalytic nanoparticles and regenerate them after catalyst deactivation (e.g. coking and sintering)27 so as to efficiently catalyze carbon removal. Here we use palladium nanoparticles (Pd NPs) due to its demonstrated ability to catalyze the oxidation of graphite28. The solution containing Pd NPs of diameter ~10 nm were dispersed in ethanol under sonication and then dropped onto an amorphous SiNx support membrane. After 3 drying up, a carbonaceous film, originating from the nonvolatile organic residues (e.g. surfactants, see more details in Supplementary Materials and Fig. S2), was left atop the surface of the support and some particles. This model system was then tested in a differentially pumped gas-environmental transmission electron microscope (Hitachi H-9500 ETEM) with a microchip- based heating stage, which can achieve direct visualization of the particle dynamics under oxygen atmosphere at elevated temperature. The schematic illustration of the experimental setup is shown in Supplementary Fig. S1. The particles were firstly heated up to ~300 oC in vacuum. They stayed immobile even after half an hour except that the shape of some particles became a little round. Surprisingly, after a pure oxygen atmosphere was established around the sample with a partial pressure of ~0.2 Pa, some particles became "nanomotors" roaming around, but would become stationary again if the oxygen flow was cut off (see more details in Supplementary Materials Table S1). One typical example is illustrated in Fig. 1a-d. Fig. 1a, b displays the beginning (oxygen injection) and final (oxygen cut off) state of the particles, respectively. It can be seen that all of the six particles moved away from their initial positions. From the in situ movie data (Supplementary Movie S1), the moving trajectory of each of them was mapped out in Fig. 1c. These particles, however, did not show a universal moving direction, but seemed to move around semi-randomly like in Brownian motion. A check of the final state in the under-focus imaging mode demonstrated several etched channels on the substrate. Interestingly, those channels (Fig. 1d) coincided exactly with the trajectories of the particles extracted from the recorded movie (Fig. 1c), while in other areas where particles had not passed by, no such channels could be found. The channels were seen to be self-avoiding. That is, a particle would not cross the previous track of any other 4 particles. Therefore, the random walks of these particles are not history-independent but has interesting non-Markovian (history-dependent) and multi-particle correlated features. These dynamics of the nanoscale particle ensemble cannot be rationalized by equilibrium thermodynamics, but is driven by active chemical energy dissipation (energy metabolism). In order to unveil the mechanism of aforementioned phenomena, local chemical characterization using high angle annular dark field-scanning TEM (HAADF-STEM) imaging and electron energy loss spectroscopy (EELS) was performed around the channels. Fig. 1e presents a series of core-loss spectra collected across a channel shown in the inset image. The nitrogen (N) K peak deriving from the SiNx membrane existed in all the spectra, while the carbon (C) K peak could only be detected in spectra 1&4 (outside the channel). This demonstrated that the carbonaceous layer has been consumed during the particle migration. On the other hand, the nanobeam diffraction pattern of Pd NPs demonstrated that they remained face-centered cubic palladium structure after the reaction, instead of forming carbides or oxides (Supplementary Fig. S3). Therefore, it can be inferred that the carbonaceous layer over the SiNx substrate was oxidized into volatile carbon monoxide (CO) or carbon dioxide (CO2) through the palladium-catalyzed reaction with oxygen at relatively low temperatures, and hence the carbon-deficient channels were etched out accompanying the migration of Pd NPs. Apart from their flexible spatial mobility, we also found an elongation-contraction shape oscillation of these Pd NPs. Fig. 2a-d, are four typical snapshots from an in situ movie (Supplementary Movie S3) demonstrating the contraction (Figs. 2a and 2c) and elongation (Figs. 2b and 2d) in particle shape. In order to quantitatively describe the extent of deformation, the 5 circumscribed ellipse of the particle was used to mimic its profile. The evolution of the eccentricity of the ellipse is plotted in Fig. 2e. It shows that the eccentricity quasi-periodically fluctuates between 0.65 (relatively round) and 0.93 (much more elongated), indicating a morphological oscillation of the particle. This thereupon gave rise to a peristalsis-fashion migration of the particle, which is akin to the movement of an earthworm by alternately shortening and lengthening of its body. Furthermore, when imaged with electron beam out of focus, some shadows with bright contrast (red crosses in Fig. 3a-f) were seen displaced from the particle (red spots in Fig. 3a-f), and appeared intermittently at some fixed angles (Fig. 3g) during the particle migration (Supplementary Movie S6). The formation of these shadows resulted from the image delocalization of the corresponding particle at large defocus setting. The shifts of the delocalization relative to the particle are in the direction of the vectors of the strongly excited diffraction beams, and the distances in between are inversely proportional to the interspacing of the corresponding crystal plane29. It is striking to confirm that the polar coordinates of the shadows (red crosses in Fig. 3h) coincided well with the diffraction pattern of a face-centered cubic structure under [100] zone axis (blue dots in Fig. 3h). It presents a persuasive evidence to prove that the active particle, although behaving like liquids morphologically, always maintained a crystalline interior with unchanging/conserved crystal orientation (always travelling around the fixed zone axis with negligible off-axis angle during its migration). This suggests that the locomotion did not involve obvious rigid-body rotation, but was dominated by surface diffusion, which is also affirmed in a bi-grained particle with a pinned grain boundary (see Supplementary Fig. S5 and Movie S5). We have thus established a high-throughput technique to simultaneously 6 monitor the morphology and crystallography of the nanocrystals during their locomotion by performing in situ TEM with under-focus imaging condition. Physically, the particle movements are accomplished by Coble creep, that is, surface diffusion of Pd atoms, which happens so rapidly at this length scale that it is sufficient to sustain the mass transport needed for the shape changes. The key difference with traditional Coble creep is that it is driven by chemical energy release instead of mechanical energy release or capillary energy reduction25; that is, the Pd atoms diffuse to where C+OCO or CO2 can happen. Consequently, the Pd particle is then motivated to spread part of itself towards the neighboring carbon, protruding in a way that allows it to meet carbon to mediate further catalytic reaction. This results in the observed "elongation", as the ends appear as if they were anchored and "stretched" by the carbon borderline that is receding away from the particle. However, the catalytic etching perpendicular to the carbon borderline might experience different speeds at each point, especially at both ends of the particle due to the perturbation of the surface diffusion of Pd atoms parallel to the borderline. As a result, the slope of the carbon borderline and thus the elongation direction of the particle were modified constantly, as indicated by the angles between the horizontal line and one fixed end of the long axis of the ellipse which are plotted in Fig. 2f. On the other hand, the particles do not elongate indefinitely, and would sometime shrink towards a more spherical shape driven by surface energy minimization. This competition between active catalytic metabolism of the surroundings and self-energy minimization provokes the shape oscillation and "peristaltic" migration of Pd NPs. Especially, the shape oscillations of the particles enabled themselves to eat out carbon-free passages much longer and wider than their equilibrium sizes, which enhanced the efficiency of the catalytic carbon removal. 7 Two main effects that could lead to the deactivation of the catalysts needs to be discussed further. The first one is surface coking30, which refers to the blocking of the metal surface by the accumulation of carbon. Since the catalytic oxidation is initiated by the chemisorption of oxygen on the Pd surface, a dense surface carbonaceous layer over the NPs would serve as a kinetic barrier to prevent the oxygen from directly contacting the catalyst, thus poisoning the catalytic reaction. However, these barricades are not all uniform or dense. Some particles with bald spots on the surface barrier layer could adsorb the oxygen, facilitate the metabolism of carbon over the particle surface (see the in situ high resolution ETEM results in Supplementary Fig. S4) as well as around the particle, and thus become the "pioneers" moving around (e.g. Fig. 1a-d). Moreover, a mobile NP will also, upon contact, help to remove the protective carbon shell of an initially-dormant particle that it touches, as illustrated in Fig. 4e-h. As a result of this "contagious mobilization", more particles will be activated upon encountering mobile particles, which will then start to move themselves (Fig. 4a-d, Supplementary Movie S2). This is thus a self- propelling chain reaction of mobilization and decoking. The other effect is particle sintering. In fact, we found that particle coalescence can be counteracted statistically by particle splitting, analogous to the cell fusion and fission in biology. One typical example is shown in Fig. 5 (Supplementary Movie S4). Three particles coalesced and then underwent shape elongation. However, soon after, this coalesced particle split into two, like in biological cell division. This splitting is an alternative process to the particle contraction: there seems to be a critical neck thickness below which the polarizing influence of the carbon on 8 the two ends is too great to maintain unity. The split particles were then able to move around as independent entities. As either the density of the NPs increases, or the carbon resources nearby get exhausted, the NPs will migrate less, split off less and eventually turn into immobile particles that slowly coarsen, and the system will demonstrate less "living" but more near-equilibrium features. We conjecture based on our understanding of the dynamics, however, that if one were able to replenish carbon on the substrate continuously as in a fuel injection engines or many actual industrial processes, then fusion and fission events of NPs might balance statistically, and we should achieve a NP population at quasi-steady state, because the principle of maximum entropy production rate in non-equilibrium thermodynamics calls for many small NPs with long total carbon-Pd contact lines to metabolize carbon, rather than a gigantic fused particle favored by equilibrium thermodynamics. In summary, we have discovered an efficient catalytic carbon removal system with coking- and sintering-resistant nanocatalysts that maintains crystalline interior with strongly conserved crystallographic orientation, working via peristalsis-like migration with high shape flexibility and spatial mobility through surface diffusion. This lifelike dynamics of inorganic matter represents a startling emergent behavior that can arise out of an energy-metabolizing nano- system, and may inspire more efficient use of nanostructures that are coarsening-resistant in energy-intensive industrial processes. 9 Figure 1. "Self-propelled" migration. a-d, "Self-propelled" migration of Pd NPs under 0.2 Pa pure oxygen at 300 oC. The moving trajectories (c) of particles from the initial (a, also red circles in b and d) to the final (b) state coincided with the channels that were found in the final state at under-focused imaging mode (d). Scale bar, 50 nm. See Supplementary Movie S1. e, The electron energy loss spectra collected across a carved channel showed that the carbon inside the channel was consumed after the particle passing through. The collecting positions of each spectrum were marked in the inset HAADF-STEM image. Scale bar, 50 nm. 10 Figure 2. Shape oscillation and peristalsis-like locomotion. a-d, In situ TEM snapshots illustrating an elongation-contraction shape oscillation and thus peristalsis-like locomotion of a Pd NP. Scale bar, 20 nm. See Supplementary Movie 3. e, Time-dependent evolution of the eccentricity of the particle geometry as delineated by the circumscribed ellipses in (a-d). f, Time- dependent elongation direction of the particle. The direction is denoted by the angle between the horizontal line and one fixed end of the long axis of the ellipse. The four insets correspond to (a- d). 11 Figure 3. Crystalline orientation-conservation. a-f, Six example snapshots from an in situ movie in the under-focus imaging mode. The shadow images with bright contrast could be found near the particles. Scale bar, 50 nm. g, Time-dependent azimuthal angle of the shadow image marked with red "+" symbol in a-f. h, Polar coordination diagram of the azimuth angle and relative distance between the shadow image and the particle from each snapshot (red crosses). They coincided very well with the diffraction pattern of a face-centered cubic structure under [100] zone axis (blue dots). See Supplementary Movie S6. 12 Figure 4. Surface decoking and contagious mobilization. a-b, Time-lapsed TEM images of Pd NPs showing a "contagious mobilization" behavior. After the contact with the mobile Particle 1, Particle 2 (delineated by dashed red polygon in a and d) started migrating away from its initial position (c, d). Scale bar, 20 nm. See Supplementary Movie S2. e-h, Schematic illustration of the chain-reaction resulted from the surface decoking by contiguous particles. 13 Figure 5. Particle fission after fusion. a-c, Fusion of the particles. Particle 1 moved around and coalesced with Particles 2&3 into a larger one. d-f, Fission of a larger particle and the split particles were then able to move around as independent entities. All scale bar, 20 nm. 14 References 1 Güralp, O., Najt, P. & Filipi, Z. S. Method for Determining Instantaneous Temperature at the Surface of Combustion Chamber Deposits in an HCCI Engine. Journal of Engineering for Gas Turbines and Power 135, 081501-081501, doi:10.1115/1.4024180 (2013). 2 Kalghatgi, G. T. Deposits in gasoline engines-a literature review. Report No. 0148-7191, (SAE Technical Paper, 1990). 3 Zerda, T., Yuan, X. & Moore, S. Effects of fuel additives on the microstructure of combustion engine deposits. Carbon 39, 1589-1597 (2001). 4 Kasper, M., Sattler, K., Siegmann, K., Matter, U. & Siegmann, H. The influence of fuel additives on the formation of carbon during combustion. J. Aerosol Sci 30, 217-225 (1999). 5 Kalghatgi, G. T. Combustion chamber deposits in spark-ignition engines: a literature review. Report No. 0148-7191, (SAE Technical Paper, 1995). 6 Liu, J. J. Advanced electron microscopy of metal–support interactions in supported metal catalysts. ChemCatChem 3, 934-948 (2011). 7 Schauermann, S., Nilius, N., Shaikhutdinov, S. & Freund, H. J. Nanoparticles for heterogeneous catalysis: new mechanistic insights. Acc Chem Res 46, 1673-1681, doi:10.1021/ar300225s (2013). 8 Xia, Y., Xiong, Y., Lim, B. & Skrabalak, S. E. Shape‐Controlled Synthesis of Metal Nanocrystals: Simple Chemistry Meets Complex Physics? Angewandte Chemie International Edition 48, 60-103 (2009). 9 Lu, J., Elam, J. W. & Stair, P. C. Synthesis and stabilization of supported metal catalysts by atomic layer deposition. Acc Chem Res 46, 1806-1815, doi:10.1021/ar300229c (2013). 15 10 Newton, M. A., Belver-Coldeira, C., Martinez-Arias, A. & Fernandez-Garcia, M. Dynamic in situ observation of rapid size and shape change of supported Pd nanoparticles during CO/NO cycling. Nat Mater 6, 528-532, doi:10.1038/nmat1924 (2007). 11 Li, Y. et al. Complex structural dynamics of nanocatalysts revealed in Operando conditions by correlated imaging and spectroscopy probes. Nat Commun 6, 7583, doi:10.1038/ncomms8583 (2015). 12 Hansen, P. L. et al. Atom-Resolved Imaging of Dynamic Shape Changes in Supported Copper Nanocrystals. Science 295, 2053-2055, doi:10.1126/science.1069325 (2002). 13 Hofmann, S. et al. In situ observations of catalyst dynamics during surface-bound carbon nanotube nucleation. Nano Lett 7, 602-608, doi:10.1021/nl0624824 (2007). 14 Tao, F. et al. Break-Up of Stepped Platinum Catalyst Surfaces by High CO Coverage. Science 327, 850-853, doi:10.1126/science.1182122 (2010). 15 Yoshida, H. et al. Visualizing Gas Molecules Interacting with Supported Nanoparticulate Catalysts at Reaction Conditions. Science 335, 317-319, doi:10.1126/science.1213194 (2012). 16 Gustafson, J. et al. High-energy surface X-ray diffraction for fast surface structure determination. Science 343, 758-761, doi:10.1126/science.1246834 (2014). 17 Zhang, S. et al. Dynamical Observation and Detailed Description of Catalysts under Strong Metal–Support Interaction. Nano Letters 16, 4528-4534, doi:10.1021/acs.nanolett.6b01769 (2016). 18 Hansen, T. W. et al. Atomic-resolution in situ transmission electron microscopy of a promoter of a heterogeneous catalyst. Science 294, 1508-1510, doi:10.1126/science.1064399 (2001). 16 19 Tao, F. et al. Reaction-driven restructuring of Rh-Pd and Pt-Pd core-shell nanoparticles. Science 322, 932-934, doi:10.1126/science.1164170 (2008). 20 Xin, H. L. et al. Revealing the Atomic Restructuring of Pt–Co Nanoparticles. Nano Letters 14, 3203-3207, doi:10.1021/nl500553a (2014). 21 de Smit, E. et al. Nanoscale chemical imaging of a working catalyst by scanning transmission X-ray microscopy. Nature 456, 222-225, doi:10.1038/nature07516 (2008). 22 Chenna, S., Banerjee, R. & Crozier, P. A. Atomic-Scale Observation of the Ni Activation Process for Partial Oxidation of Methane Using In Situ Environmental TEM. ChemCatChem 3, 1051-1059, doi:10.1002/cctc.201000238 (2011). 23 Vesborg, P. C. et al. Transient behavior of Cu/ZnO-based methanol synthesis catalysts. Journal of Catalysis 262, 65-72 (2009). 24 Vendelbo, S. B. et al. Visualization of oscillatory behaviour of Pt nanoparticles catalysing CO oxidation. Nat Mater 13, 884-890, doi:10.1038/nmat4033 (2014). 25 Sun, J. et al. Liquid-like pseudoelasticity of sub-10-nm crystalline silver particles. Nat. Mater. 13, 1007-1012, doi:10.1038/nmat4105 (2014). 26 Xie, D. G. et al. In situ study of the initiation of hydrogen bubbles at the aluminium metal/oxide interface. Nat Mater 14, 899-903, doi:10.1038/nmat4336 (2015). 27 Lu, J. L. et al. Coking- and Sintering-Resistant Palladium Catalysts Achieved Through Atomic Layer Deposition. Science 335, 1205-1208, doi:10.1126/science.1212906 (2012). 28 Fryer, J. R. Oxidation of Graphite catalysed by Palladium. Nature 220, 1121-1122 (1968). 29 Zandbergen, H., Tang, D. & Van Dyck, D. Non-linear interference in relation to strong delocalisation. Ultramicroscopy 64, 185-198 (1996). 17 30 Moulijn, J. A., Van Diepen, A. & Kapteijn, F. Catalyst deactivation: is it predictable?: What to do? Applied Catalysis A: General 212, 3-16 (2001). Supporting Information Supporting information of nanoparticle synthesis, TEM experimental setup, Table S1, Figure S1- S5 and in situ TEM movies S1-S6. Acknowledgment The authors acknowledge support from the Natural Science Foundation of China (51231005, 51401159 and 51321003), and 973 Programs of China (2012CB619402). We also appreciate the support from the 111 project (B06025). J.L. acknowledges support by NSF CBET-1240696 and DMR-1410636. E.M. acknowledges support from US DoE-BES-DMSE, under Contract No. DE- FG02-13ER46056. We also thank Dr. Chao Wu (Xi'an Jiaotong University, Xi'an, China) and Prof. Kaili Jiang (Tsinghua University, Beijing, China) for helpful discussions. Author Contributions Z.-W.S. and J.L. supervised the project. P.-H.L. performed the in situ ETEM experiments. B.- Y.L. and P.-H.L. conducted the STEM-EELS characterization. P.-H.L., D.-G.X. and F.A. analyzed the data with input from Z.-W.S., J.L., E.M. and X.F.Z. Z.-R.Z. and M.-S.J. synthesized the nanoparticle samples. P.-H.L., J.L., E.M. and Z.-W.S. wrote the manuscript. All the authors discussed the results and commented on the manuscript. 18 Supporting Information for Highly Deformable and Mobile Palladium Nanocrystals as Efficient Carbon Scavengers Peng-Han Lu1†, De-Gang Xie1, Bo-Yu Liu1, Fei Ai1, Zhao-Rui Zhang2, Ming-Shang Jin2, Xiao Feng Zhang3, Evan Ma1,4, Ju Li1,2,5* and Zhi-Wei Shan1* 1 Center for Advancing Materials Performance from the Nanoscale (CAMP-Nano) and XJTU-HHT Research & Development Center (XHRDC), State Key Laboratory for Mechanical Behavior of Materials, Xi'an Jiaotong University, Xi'an 710049, China 2 Frontier Institute of Science and Technology, Xi'an Jiaotong University, Xi'an, Shaanxi 710054, China 3 Hitachi High Technologies America, Pleasanton, CA 94588, USA 4 Department of Materials Science and Engineering, Johns Hopkins University, Baltimore, MD 21218, USA 5 Department of Nuclear Science and Engineering and Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA † Present address: Ernst Ruska-Centre for Microscopy and Spectroscopy with Electrons and Peter Grünberg Institute, Forschungszentrum Jülich, 52425 Jülich, Germany *E-mail: [email protected]; [email protected] 1 Synthesis of Pd nanoparticles. Pd cubic seeds with an average edge length of 10 nm were synthesized following a previously reported protocol1. In a typical synthesis experiment, 8.0 mL of deionized (DI) water (18.2 MΩ·cm), 105 mg of Poly(vinyl pyrrolidone) (PVP, MW≈55,000, Aldrich), 60 mg of L-ascorbic acid (AA, Aldrich), and 300 mg of KBr (Aldrich) were mixed together in a 20 mL vial and pre-heated in air under magnetic stirring at 80 oC for 10 min. 57 mg of Na2PdCl4 (Aldrich) were dissolved in 3 mL of DI water and then introduced into the pre-heated solution. After the vial had been capped, the reaction was allowed to proceed at 80 oC for 3 h. The final product was collected by centrifugation. In situ heating ETEM experiments. The in situ heating experiments were carried out on a micro-electro-mechanical system (MEMS) thermal chip (Protochips Aduro heating holder). Pd NPs were dispersed in ethanol under sonication and dropped onto a MEMS chip for the in situ TEM experiments. They were directly heated up to 300 oC at a heating rate of 2 oC/s and kept at that temperature for ~30 min. Pure oxygen gas up to 0.2 Pa was then injected into the TEM specimen chamber. The dynamics of Pd NPs under oxygen atmosphere were investigated with Hitachi H-9500 environmental transmission electron microscope (ETEM) operated at 300 keV. The microscope is featured by a differentially pumped system and thus can accommodate reactive gases within the specimen chamber. The real-time TEM movies were recorded with Gatan Orius 832 high-speed charge coupled device (CCD) camera at a time resolution of 0.2 s. 2 Supplementary Figure S1 Pd NPs of diameter ~10 nm were dropped onto a ~50 nm- thick, amorphous SiNx membrane on a MEMS thermal chip for the in situ TEM experiments. Pure oxygen gas was then injected into the TEM specimen chamber. Using a differentially pumped gas-environmental transmission electron microscope (Hitachi H- 9500 ETEM) and a microchip-based Joule-heating stage, a direct visualization of the particle dynamics under oxygen atmosphere and at elevated temperature were obtained. HAADF-STEM, NBED and EELS characterization. Further characterization was performed using High Angle Annular Dark Field Scanning TEM (HAADF-STEM) imaging, Nano-Beam Electron Diffraction (NBED) and Electron Energy-Loss Spectroscopy (EELS) after the in situ reaction inside ETEM. The samples on the microchip were reloaded to another TEM (JEOL JEM-2100F) with compatible holder for this characterization. 3 Critical conditions for the self-propelled motion of Pd particles Supplementary Table S1 Vacuum (2e-4 Pa) O2 (2e-1 Pa) N2 (2e-1 Pa) RT F F F 100 oC 200 oC F F F F F F 300 oC F either w/ or w/o e-beam T F F: No obvious motion T: Rapid dramatic motion From Table S1, it can be determined that the oxygen atmosphere and an elevated temperature (much lower than the melting point of Pd but still a little bit higher than the room temperature) are necessary for this reaction, while e-beam irradiation is not required. Source of the carbonaceous layer Supplementary Figure S2 HAADF-STEM images of (a) a fresh SiNx support without any samples and (b) the same position after dispersing Pd NP samples. Electron energy loss spectra (c) showed that notable carbon signal could only be detected after depositing the samples on the substrate. All scale bar, 100 nm. 4 We examined several possible origins of the source of the carbonaceous layer. First, electron beam irradiation-induced carbon contamination can be ruled out, because in the beam-off condition the particles still experienced similar dynamic evolution (Table S1). Second, EELS characterization of a fresh SiNx support without any samples exhibited little carbon signal. However, after dropping Pd NP samples, notable carbon signal could be detected at the same position (Fig. S2). Therefore, the carbon source should come from the sample solution. Third, it has been reported that most of the surfactant polymers, which are often used in the shape-controlled synthesis of nanomaterials, could hardly be washed away completely through centrifugation2, especially when the size of the NPs decreases down to ~10 nm. The residue surfactant polymers after air drying and vacuum annealing will transform into a carbonaceous layer. This was further confirmed by in situ high resolution TEM (HRTEM) study of an individual Pd NP. As presented in Supplementary Fig. S4, it was obvious that there was an amorphous layer over the surface of the particle, which should come from the surface coating surfactants. Similar to what we found about the carbonaceous layer over the substrate, the amorphous layer on NP surface could also be removed under oxygen atmosphere. Structure of the Pd NP after the reaction Nano-beam diffraction characterization of the Pd NP after the reaction showed that the particles remained face-centered cubic structure, instead of forming carbide or oxide. One typical example is given in Fig. S3. 5 Supplementary Figure S3 HAADF-STEM image (a) and nano-beam diffraction pattern (b) of a Pd NP after the reaction, showing that the particle remained a face- centered cubic Pd crystal, instead of forming carbide or oxide. Scale bar, 20 nm. 6 In situ HRTEM observation of the carbon gasification Supplementary Figure S4 TEM snapshots from an in situ HRTEM movie. The Pd NP was immersed in 0.2 Pa oxygen atmosphere after being heated up to 300 oC. The lattice fringes correspond to the Pd (200) planes. The surface carbon layer was finally oxidized and removed. Scale bar, 2 nm. 7 Surface diffusion-dominated migration of a bi-grained particle The liquid-like, orientation-conserved crystal migration through profuse diffusion can be best displayed in a bi-grained particle in Fig. S5. As shown in Fig. S5a, a particle, possibly fused by smaller ones, included two grains with a grain boundary (GB) along the red dashed line, which was inferred from the distinct diffraction contrast of the two halves (one darker, one brighter). Instead of undergoing grain rotation or grain boundary migration that are often found at high temperature, or under mechanical stress, the particle retained its grain boundary along the red dashed line with constant slope and abscissa, swinging around it and catalyzing the oxidation of the carbon on both sides (Supplementary Movie S5). This process was realized through surface self-diffusion of the Pd atoms, that is, the mass on one side travelled through the surface of the particle to the other side, so the grain boundary must remain pinned to that line in both slope and abscissa. The projected area of both sides underwent an oscillatory change alternately (Fig. S5e). When and only when the mass of either side was completely transported to the other one, the particle would be depinned from that GB line (since GB has disappeared) and be able to freely move in two dimensions again as a single-grained NP (Fig. S5d). 8 Supplementary Figure S5 Surface diffusion-dominated migration of a bi-grained particle. a-c, The bi-grained particle can only migrate along the red dashed line, i.e. its grain boundary, with its mass swung around it. d, Once all the mass was transported to one side, the bi-grained particle became a single-grained particle and was able to move away from the red dashed line. Scale bar, 20 nm. e, The projected area of both sides underwent an oscillatory evolution alternately. See Supplementary Movie S5. References Lim, B. et al. New insights into the growth mechanism and surface structure of palladium nanocrystals. Nano Research 3, 180-188 (2010). Zhang, H., Jin, M., Xiong, Y., Lim, B. & Xia, Y. Shape-controlled synthesis of Pd nanocrystals and their catalytic applications. Accounts of chemical research 46, 1783-1794 (2012). 9 1 2
1906.12096	1	1906	2019-06-28T08:45:43	Spin Logic Devices via Electric Field Controlled Magnetization Reversal by Spin-Orbit Torque	[ "physics.app-ph" ]	We describe a spin logic device with controllable magnetization switching of perpendicularly magnetized ferromagnet / heavy metal structures on a ferroelectric (1-x)[Pb(Mg1/3Nb2/3)O3]-x[PbTiO3] (PMN-PT) substrate using current-induced spin-orbit torque. The devices were operated without an external magnetic field and controlled by voltages as low as 10 V applied across the PMN-PT substrate, which is much lower compared to previous reports (500 V). The deterministic switching with smaller voltage was realized from the virgin state of the PMN-PT. Ferroelectric simulation shows the unsaturated minor loop exhibits obvious asymmetries in the polarizations. Larger polarization can be induced from the initial ferroelectric state, while it is difficult for opposite polarization. The XNOR, AND, NAND and NOT logic functions were demonstrated by the deterministic magnetization switching from the interaction between the spin-orbit torque and electric field at the PMN-PT/Pt interface. The nonvolatile spin logic scheme in this work is simple, scalable, programmable, which are favorable in the logic-in-memory design with low energy consumption.	physics.app-ph	physics	> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1 Spin Logic Devices via Electric Field Controlled Magnetization Reversal by Spin-Orbit Torque Meiyin Yang, Yongcheng Deng, Zhenhua Wu, Kaiming Cai, Kevin William Edmonds, Yucai Li, Yu Sheng, Sumei Wang, Yan Cui, Jun Luo, Yang Ji , Hou-Zhi Zheng and Kaiyou Wang  local magnetization switching [4]， [5] . This offers a way to switching Abstract -- We describe a spin logic device with controllable of perpendicularly magnetized magnetization ferromagnet / heavy metal structures on a ferroelectric (1-x)[Pb(Mg1/3Nb2/3)O3]-x[PbTiO3] (PMN-PT) substrate using current-induced spin-orbit torque. The devices were operated without an external magnetic field and controlled by voltages as low as 10 V applied across the PMN-PT substrate, which is much lower compared to previous reports (500 V). The deterministic switching with smaller voltage was realized from the virgin state of the PMN-PT. Ferroelectric simulation shows the unsaturated minor loop exhibits obvious asymmetries in the polarizations. Larger polarization can be induced from the initial ferroelectric state, while it is difficult for opposite polarization. The XNOR, AND, NAND and NOT logic functions were demonstrated by the deterministic magnetization switching from the interaction between the spin-orbit torque and electric field at the PMN-PT/Pt interface. The nonvolatile spin logic scheme in this work is simple, favorable scalable, programmable, which the logic-in-memory design with low energy consumption. are in Index Terms -- Electric field, logic, spin-orbit torque, I. INTRODUCTION L OGIC and programmable memory devices are of great significance for information technology today. A spin logic function by spin-orbit torque (SOT) is a promising candidate to realize such devices with high integration and low power consumption. In heavy metal/ferromagnet systems, spin-orbit torques can be used to switch the perpendicularly magnetized ferromagnets [1]-[3]. The spin polarization due to SOT is directed perpendicular to the out-of-plane ferromagnetic easy axis, providing an additional degree of freedom to control the This work was supported by National Key R&D Program of China No. 2017YFB0405700 and 2017YFA0303400. This work was supported also by the NSFC Grant No. 11604325, 11474272 and 61774144. The project was sponsored by Chinese Academy of Sciences, grant No. QYZDY-SSW-JSC020, XDA18000000, XDPB12 and XDB28000000. M. Yang, Y. Deng, Y. Li, K. Cai, Y. Sheng, Y. Ji, Z. Zhi and K. Wang are with SKLSM, Institute of Semiconductors, CAS, P. O. Box 912, Beijing 100083, China (e-mail: kywang@ semi.ac.cn). M. Yang, Z. Wu, S. Wang, Y.Cui and J. Luo are with Key Laboratory of Institute of Microelectronic Devices Microelectronics, Chinese Academy of Sciences, Beijing 100029, China. Integrated Technology, and K.W.Edmonds is with School of Physics and Astronomy, University of Nottingham, Nottingham NG7 2RD, United Kingdom. K. Wang is with Center of Materials Science & Optoelectronics Engineering , Center for Excellence in Topological Quantum Computation, University of Chinese Academy of Sciences, Beijing 100049 , P. R. China, and Beijing Academy of Quantum Information Sciences, Beijing 100193, China. lost switching was construct logic functions in a single memory cell [6] and greatly favors the scaling of integration [7]. The SOT enables the separation of reading and writing channels in magnetic tunneling junctions, which will greatly increase the endurance of magnetic random access memory (MRAM) arrays. However, generally an in-plane magnetic field is required to realize the deterministic magnetization switching induced by SOT [8]-[11], and the direction of the in-plane field determines the chirality of the magnetization loop, i.e. whether a positive current favors "up" or "down" orientations of the magnetization. The requirement for an external field impedes high-density integration. Although lots of efforts have been made to realize magnetic field-free switching by SOT, including using wedged structures [12], [13], tilted anisotropy [14]-[16], exchange bias [17]-[20], interface engineering [21], stray fields [22], [23] or exchange coupling [24]-[27], the controllable chirality of the magnetization reports. All-electrical manipulation of magnetization while maintaining the chirality control, which is crucial for programmable storage, still requires work. Recently, deterministic current-induced magnetization switching was demonstrated in a hybrid [28]. By polarizing the PMN-PT substrate with an applied voltage, deterministic switching was achieved by electrical current in the absence of an external magnetic field. The chirality of the magnetic switching loop can also be controlled by changing the sign of the voltage applied to the PMN-PT substrate. However, the 500 V required to polarize the PMN-PT substrate is too large for practical applications. Also, logic functionality was not demonstrated in Ref. [28] due to the requirement of high voltage. Here, we describe devices in which the distance between the electrodes on the PMN-PT is reduced from 1 mm to 100 μm, so that the voltage required is greatly reduced. By applying voltage pulses to the PMN-PT and current pulses to the heavy metal/ferromagnet device, we demonstrate multiple logic functions using a single magnetic unit. ferromagnetic-ferroelectric structure for most electric control field of II. EXPERIMENTS Pt (3 nm)/Co (0.9 nm)/AlOx (2 nm) films were deposited on top of (001)-oriented PMN-PT ferroelectric substrates with thickness of 0.5 mm by magnetron sputtering at room temperature. Pt, Co, and Ni were sputtered at a pressure of 0.8 mTorr with deposition rates of 1.3 nm/min, 0.7 nm/min, and > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2 substrate. We found that the deterministic switching could be achieved under a small voltage from the virgin state of the PMN-PT. Then it requires larger voltage to reverse the deterministic switching loop, and only partial reversal is achieved. To explain this phenomenon, we conducted the ferroelectric hysteresis loop simulation as is shown in Fig.1 (c). The saturated major loop is very symmetric, while the initial minor loop was obviously asymmetric. The initial remanence polarization (point (1)) is larger than that of point (4) after applied with the same opposite electric field. Thus, due to the surface local electrodes on the bulk PMN-PT substrate, the devices are operated at the minor zone so that the magnetization loop exhibits asymmetric between 10 V and -30 V voltages on the PMN-PT substrate. The asymmetric behaviors could be improved by using patterned ferroelectric thin film with low coercivity field in future. A XNOR logic function was demonstrated as illustrated in Fig. 1. (a) The schematic device structure for Pt/Co/AlOx sample. The current and voltage channels of the Pt/Co/AlOx film are 2 um wide (light blue), while the voltage contacts to the PMN-PT substrate are separated by 100 um (dark blue) (b) The anomalous Hall resistance loop of the device under a perpendicular magnetic field. respectively. The Pt/Co/AlOx 2.52 nm/min film was subsequently patterned into an array of Hall bar devices by electron beam lithography (EBL) and etching techniques. A Keithley 2602 current source and a Keithley 2182A nanovoltmeter were used the d.c. Hall voltage measurements. The ferroelectric simulation was done by the technology computer aided design (TCAD) software using the package of global TCAD solutions (GTS). for III. RESULTS AND DISCUSSIONS A. Basic magnetic properties The device structure of the Pt/Co/AlOx film is shown in Fig. 1(a). The channels for current injection and Hall voltage detection are 2 μm wide. The separation of the voltage electrodes on the PMN-PT substrate is 100 μm, which is one-tenth of the distance in the previous work [28]. The hysteresis loop of the anomalous Hall resistance (RH) with the magnetic field applied in the out-of-plane direction is presented in Fig. 1(b). The square hysteresis loop shows that the Pt/Co/AlOx sample has perpendicular anisotropy with a switching field of 2 Oe. B. The Logic functions The electrical current switching behavior of the Pt/Co/AlOx sample after applying a voltage to the PMN-PT substrate is shown in Fig. 2. The minimum voltage to observe the deterministic switching without a magnetic field is 10 V, which was applied first on the PMN-PT substrate, and then removed before the electrical measurements. Figure 2(a) presents the RH-I loop after applying the voltage of 10 V. Consistent with our previous work [28], the deterministic switching was achieved but with much less voltages. The decrease of the voltage is an important step towards the application of SOT-induced programmable memory [29]. For ‒10 V applied to the PMN-PT, no deterministic switching was observed, indicating no preferred polarization under the device. The reversed deterministic switching was observed after applying ‒30 V (Fig. 2(b)). The size of the Hall resistance signal indicates that nearly 80% of the magnetic domains exhibits deterministic switching. Fully deterministic switching did not show up when further raising the voltage. The switching behavior of the Pt/Co/AlOx device is closely related to the sequence of the applied voltage on the PMN-PT Fig. 2. (a,b) Current-induced magnetization switching after a voltage was applied across the PMN-PT substrate for the Pt/Co/AlOx sample, (a) 10 V, (b) -30 V. (c) The simulated ferroelectric Polarization vs. electric field hysteresis loop. (d) Demonstration of the XNOR logic function, showing the Hall signal RH after current pulses I applied to the Pt/Co/AlOx device and voltage pulses U applied to the PMN-PT. Fig. 2(d). The voltage applied to the current channel and the PMN-PT electrodes are defined as the two inputs, and the Hall resistance signals are the outputs. Positive and zero Hall resistance are defined as "1" and "0", respectively. Current pulses of ±0.1 mA with the duration of 0.5 s were injected and then 0.02 mA constant current was applied to read the Hall resistance. Voltages of 10 V and ‒30 V were applied to the PMN-PT electrodes and removed before the current pulses were injected. First, after the 10 V pulse (input "1"), the magnetization followed the switching loop in Fig. 2(a), where the positive current (input "1") resulted in the negative Hall resistance (output "0") and negative current led to positive Hall resistance (output "1"). After the ‒30 V pulse (input "0"), the magnetization switching was reversed, so that the positive current (input "1") and negative current (input "0") generated positive resistance (output "1") and nearly zero resistance (output "0"). The switching behavior acts as the function of > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3 Table I summarized some averaged features of logics functions by CMOS, other spintronics logics and our work. The energy consumption of this work was calculated with the speed of 5 ns using 500 nm Hall bar. Due to the low switching current, the energy consumption of our logics is lower than previous spintronic devices and then comparable with the CMOS logic gates. Further more, this device is also a memory, so that no static energy are consumed. Also, logic gates(AND, NAND, XNOR and NOT) can be achieved in the same cell with less transistors needed, which is favorable in the future logic-in-memory devices. the different [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] REFERENCES L. Liu, C. -F. Pai, Y. Li, H. W. Tseng, D. C. Ralph and R. A. Buhrman, "Spin-torque switching with the giant spin Hall effect of tantalum," Science, vol. 336, no. 6081, pp. 555-558, May 2012. doi:10.1126/science.1218197. I. M. Miron, K. Gaudin, S. Auffret, B. Rodmacq, A. Schuhl, S. Piz zini, J. Vogel and P. Gambardella, "Perpendicular switching of a si ngle ferromagnetic layer inuced by in-plane current injection," Nat ure, vol. 476, no. 7359, pp. 189-193, Jul. 2011. doi: 10.1038/natur e10309. Y. Li, K. W. Edmonds, X. Liu, H. Zheng and K. Wang, "Manipulat ion of magnetization by spin -- orbit torque," Adv. Quantum Technol., vol. 2, Oct. 2018, Art. no. 1800052. doi:10.1002/qute.201800052. L. Liu, T. Moriyama, D. C. Ralph and R. A. Buhrman, "Spin-torque ferromagnetic resonance induced by the spin Hall effect," Phys. Re v. Lett., vol. 106, no. 3, Jan. 2011, Art. no. 036601. doi: 10.1103/Ph ysRevLett.106.036601. M. Yang, K. Cai, H. Ju, K. W. Edmonds, G. Yang, S. Liu, B. Li, B. Zhang, Y. Sheng, S. Wang, Y. Ji and K. Wang, "Spin-orbit torque in Pt/CoNiCo/Pt symmetric devices," Sci. Rep., vol. 6, Feb. 2016, Art. no. 20778. doi: 10.1038/srep20778. C. Wan, S. Zhang, Z. Yuan, C.Fang, W.Kong, Q. Zhang, H. Wu, U. Khan and H. Feng, "Programmable Spin Logic Based on Spin Hall Effect in a Single Device," Adv. Electron. Mater., vol. 3, no. 3, Jun. 2017, Art. no. 1600282. doi:10.1002/aelm.201600282 . X. Wang, C. Wan, W. Kong, X. Zhang, C. Fang, B. Tao, W. Yang, H. Li, H. Wu, M. Irfan and X. Han, "Field-free programmable spin logics via chirality-reversible spin -- orbit torque switching," Adv. Ma ter., vol. 30, no. 0, Feb. 2018, Art. no. 1801318, doi:10.1002/adma. 201801318. J. Kim, H. Sinha, M. Hayashi, M. Yamanouchi, S. Fukami, T. Suzu ki, S. Mitani and H. Ohno, "Layer thickness dependence of the curr ent-induced effective field vector in TaCoFeBMgO," Nat. Mater., vol. 12, no. 3, pp. 240-245, Dec. 2013. doi: 10.1038/nmat3522. X. Qiu, K. Narayanapillai, Y. Wu, P. Deorani, D. -- H. Yang, W. -- S. Noh, J. -H. Park, K. -- J. Lee, H. -- W. Lee and H. Yang, "Spin -- orbit-torque engineering via oxygen manipulation," Nat. Nano., vol. 10, no. 4, pp. 333-338, Mar. 2015. doi: 10.1038/nnano.2015.18. C. H. Sim, J. C. Huang, M. Tran, and K. Eason, "Asymmetry in eff ective fields of spin-orbit torques in Pt/Co/Pt stacks," Appl. Phys. L ett., vol. 104, no. 1, Jan. 2014, Art. no. 012408. doi: 10.1063/1.486 1459. Y. Sheng, Y. C. Li, X. Q. Ma, and K. Y. Wang, "Current-induced four-state magnetization switching by spin-orbit in perpendicular ferromagnetic trilayers," Appl. Phys. Lett., vol. 113, no. 11, Sep. 2018, Art. no. 112406. doi: 10.1063/1.5034380. G. Yu, P. Upadhyaya, Y. Fan, J. G. Alzate, W. Jiang, K. L. Wong, S. Takei, S. A. Bender, L. -T. Chang, Y. Jiang, M. Lang, J. Tang, Y. Wang, Y. Tserkovnyak, P. K. Amiri and K. L. Wang, "Switching of perpendicular magnetization by spin-orbit torques in the absence of external magnetic fields," Nat. Nano., vol. 9, no. 7, pp. 548-554, May 2014. doi: 10.1038/nnano.2014.94. G. Yu, L. -- T. Chang, M. Akyol, P. Upadhyaya, C. He, X. Li, K. L. Wong, P. K. Amiri and K. L. Wang, "Current-driven perpendicular magnetization switching in Ta/CoFeB/[TaOx or MgO/TaOx] films with lateral structural asymmetry," Appl. Phys. Lett., vol. 105, no. 1 0, Sep. 2014, Art. no 102411. doi: 10.1063/1.4895735. torques Fig. 3(a) The schematic device structure for Logic gates, (b) the truth table of Logic gates XNOR gate, which is realized using only one memory unit. Besides the XNOR logic function, AND, NAND and NOT logic gates are proposed. The logic gates setup and true table were shown in Fig. 3. The two inputs are the currents that applied to the Hall channel. The "0" represents 0 mA and the "1" represents "0.03 mA". The magnetization are initialized before every operation by injecting a large current (e.g. -0.1 mA). NAND logic function gate can be achieved when applying 10 V to the electrodes on PMN-PT. Only on one condition that the two inputs are "1" which exceeds the critical switching current can reverse the magnetization, resulting a lower voltage output ( Uout=0), which is a NAND logic function. Besides, if we fixed I1 to "1" every time (which can be regarded as the operating current) and varied the input I2, the outputs are always opposite to I2, resulting a NOT gate functions. If we applied -30 V to the electrodes on PMN-PT, only two "1" inputs can set the output to "1", which is a AND logic. Thus, we can achieve three different logic functions using the same structure. C. The features of the new logic scheme The proposed AND, NAND and NOT logic gates are concatenable since the input and output bits are encoded with the same physical quantities: current amplitude. For Hall devices, the output current are realized by anomalous Hall voltage, which transferred into current to the next logic gates. The gain of the logic by Hall bar is small but can be improved using a magnetic junction. The controllable deterministic switching methods in this paper is compatible to tunnel TABLE I COMPARISON OF LOGIC DEVICES BY DIFFERENT METHODS E This work Ref. [30] Ref.[31] Ref.[7] ~2fJ ~2fJ ~1fJ ~1.5 pJ Non-volatile Transistors C/M yes 3 M no 6~12 C no 1 M yes 3 M E represents the energy consumption and C/M symbolizes the CMOS or ferromagnetic logic gate. MTJ, in which case the output can be directly as the input to the other devices without amplifier. The logic output are almost linear before the switching of the magnets, since the magneto-resistance hardly changes with the applied voltage or current. Since the logic gate only use one memory cell and the magnets are stable below the Curie temperature, the necessity of feedback elimination is greatly reduced. > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4 Wang, "Piezo voltage controlled planar Hall effect devices," Sci. Rep., Vol. 6, Jun. 2016, Art. no. 28458. doi:10.1038/srep28458. [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] by in switching L. You, O. Lee, D. Bhowmik, D. Labanowski, J. Hong, J. Bokor and S. Salahuddin, "Switching of perpendicularly polarized nanomagnets with spin orbit torque without an external magnetic field by engineering a tilted anisotropy," PNAS, vol. 112, no. 33, pp. 10310-10315, Aug. 2015. doi: 10.1073/pnas.1507474112. J. Torrejon, F. Garcia-Sanchez, T. Taniguchi, J. Sinha, S. Mitani, J. -- V. Kim and M. Hayashi, "Current-driven asymmetric magnetizati on switching in perpendicularly magnetized CoFeB/MgO heterostr uctures," Phys. Rev. B, vol. 91, no. 21, Jun. 2015, Art. no, 214434. doi: 10.1103/PhysRevB.91.214434. C.-F. Pai, M. Mann, A. J. Tan, and G. S. D. Beach, "Determination of spin torque efficiencies in heterostructures with perpendicular magnetic anisotropy," Phys. Rev. B, vol. 93, no. 14, Apr. 2016, Art. no. 144409. doi: 10.1103/PhysRevB.93.144409. Y.-W. Oh, S. h. Chris Baek, Y. M. Kim, H. Y. Lee, K. -- D. Lee, C. -- G. Yang, E.-S. Park, K. -- S. Lee, K. -- W. Kim, G. Go, J. -- R. Jeong, B. -- C. Min, H. -- W. Lee, K. -- J. Lee and B. -- G. Park, "Field-free switching of perpendicular magnetization through spin -- orbit torque in antiferromagnet/ferromagnet/oxide structures," Nat. Nano., vol. 11, pp. 878-884, Jul. 2016. doi: 10.1038/nnano.2016.109. S. Fukami, C. Zhang, S. DuttaGupta, A. Kurenkov and H. Ohno, "Magnetization an antiferromagnet -- ferromagnet bilayer system," Nat. Mater., vol. 15, pp. 535-541, Feb. 2016. doi: 10.1038/nmat4566. W. J. Kong, Y. R. Ji, X. Zhang, H. Wu, Q. T. Zhang, Z. H. Yuan, C. H. Wan, X. F. Han, F. Kenji, N. Hiroshi and T. Mean-Jue, "Field-free spin Hall effect driven magnetization switching in Pd/Co/IrMn exchange coupling system," Appl. Phys. Lett., vol. 109, no. 13, Sep. 2016, Art. no. 132402. doi: 10.1063/1.4963235. A. van den Brink, G. Vermijs, A. Solignac, J. Koo, J. T. Kohlhepp, H. J. M. Swagten and B. Koopmans, "Field-free magnetization reversal by spin-Hall effect and exchange bias," Nat. Commun., vol. 7, Mar. 2016, Art. no. 10854. doi: 10.1038/ncomms10854. S.-h. C. Baek V. P. Amin, Y. -- W. Oh, G. Go, S. -- J. Lee, G. -- H. Lee, K. -- J. Kim, M. D. Stiles, B. -G. Park and K. J. Lee, "Spin currents and spin -- orbit torques in ferromagnetic trilayers," Nat. Mater., vol. 17, no. 6, pp. 509-513, Jun. 2018. doi: 10.1038/s41563-018-0041-5 S. Wang, M. Yang, and C. Zhao, "Deterministic switching of perpendicularly magnetic layers by spin-orbit torque through stray field engineering," IEEE Magn. Lett., vol. 9, Feb. 2018. Art. no. 3101804. doi: 10.1109/lmag.2018.2801834. Z. Zhao, A. K. Smith, M. Jamali and J.-P. Wang, "External-Field-Fr ee Spin Hall Switching of Perpendicular Magnetic Nanopillar with a Dipole-Coupled Composite Structurear, " arXiv:1603.09624. Y.-C. Lau, D. Betto, K. Rode, J. M. D. Coey and P. Stamenov, "Spi n -- orbit torque switching without an external field using interlayer e xchange coupling," Nat. Nano., vol. 11, pp. 758-762, May 2016. do spin -- orbit torque i： 10.1038/nnano.2016.84. W. Y. Kwak, J. H. Kwon, P. Grünberg, S. H. Han and B. K. Cho, "Current-induced magnetic switching with spin-orbit torque in an interlayer-coupled junction with a Ta spacer layer," Sci. Rep., vol. 8, no. 1, Feb. 2018, Art. no. 3826. doi: 10.1038/s41598-018-22122-1. Y. Sheng, K. W. Edmonds, X. Ma, H. Zheng and K. Wang, "Adjustable Current-Induced Magnetization Switching Utilizing Interlayer Exchange Coupling," Adv. Electron. Mater., vol. 4, no. 0, Jun. 2018. Art. no. 1800224. doi:10.1002/aelm.201800224. Y. Cao, A. Rushforth, Y. Sheng, H. Zheng, and K. Wang, "Tuning a Binary Ferromagnet into a Multistate Synapse with Spin -- Orbit-Torque-Induced Plasticity," Adv. Funct. Mater., Mar. 2019. Art. no. 1808104. doi:10.1002/adfm.201808104. K. Cai, M. Yang, H. Ju, S. Wang, Y. Ji, B. Li, K. W. Edmonds, Y. Sheng, B. Zhang, N. Zhang, S. Liu, Z. Zheng and K. Wang, "Electric current-induced magnetization switching in a hybrid ferromagnetic/ferroelectric structure," Nat. Mater., vol. 16, no. 7, pp. 712-716, Apr. 2017. doi: 10.1038/nmat4886. K. Wang, "Complementary logic with a spin," Nat. Electron., vol. 1, no. 7, pp. 378-379, Jul. 2018. doi: 10.1038/s41928-018-0105-1. R. Zimmermann and W. Fichtner, "Low-power logic styles: CMOS versus pass-transistor logic," IEEE J. SOLID-ST. Circ., Vol. 32, No. 7, pp.1070-1090, Jul. 1997. doi: 10.1109/4.597298. B. Zhang, K. K. Meng , M. Y. Yang, K. W. Edmonds, H. Zhang , K. M. Cai, Y. Sheng, N. Zhang. Y. Ji, J. -- H. Zhao, H. Z. Zheng and K. deterministic control field of
1909.04952	3	1909	2019-12-04T15:40:57	Time of Flight Modulation of Intensity by Zero Effort on Larmor	[ "physics.app-ph", "quant-ph" ]	A time of flight Modulation of IntEnsity by Zero Effort (MIEZE) spectrometer mode has been developed for the Larmor instrument at the ISIS pulsed neutron source. The instrument utilizes resonant neutron spin flippers which employ electromagnets with pole shoes, allowing the flippers to operate at frequencies of up to 3 MHz. Tests were conducted at modulation frequencies of 103 kHz, 413 kHz, 826 kHz and 1.03 MHz, resulting in a Fourier time range of ~0.1 ns to 30 ns using wavelength band of 4 {\AA}-11 {\AA}.	physics.app-ph	physics	Time of Flight Modulation of Intensity by Zero Effort on Larmor N. Geeritsa,b, S. R. Parnella, M.A. Thijsa, A.A. van Wella, C. Franzc,d, A.L. Washingtone, D. Raspinoe, R.M. Dalglieshe and J. Plompa aFaculty of Applied Sciences, Delft University of Technology, Mekelweg 15, Delft, 2629JB, The Netherlands bAtominstitut, TU Wien, Stadionallee 2, Vienna, 1020, Austria cPhysik Department, Technische Universität München, Garching, D-85748, Germany dHeinz Maier-Leibnitz Zentrum (MLZ), Technische Universität München, Garching, D-85748, Germany eISIS, Rutherford Appleton Laboratory, Chilton, Oxfordshire, OX11 0QX, UK Correspondence email: [email protected] Abstract A time of flight Modulation of Intensity by Zero Effort spectrometer mode has been developed for the Larmor instrument at the ISIS pulsed neutron source. The instrument utilizes resonant neutron spin flippers which employ electromagnets with pole shoes, allowing the flippers to operate at frequencies of up to 3 MHz. Tests were conducted at modulation frequencies of 103 kHz, 413 kHz, 826 kHz and 1.03 MHz, resulting in a Fourier time range of ~0.1 ns to 30 ns using a wavelength band of 4 Å to 11 Å. 1. Introduction Inelastic neutron scattering measures dynamics in materials by determining the energy transfer between the sample and the neutron beam. Resolution limitations occur in traditional techniques when the energy transfer is much smaller than the accuracy with which neutron energy can be determined (~µeV). To overcome this limitation, interferometric techniques based on the neutron spin-echo principle [1] can be used. Neutron spin-echo instruments (NSE) [1, 2] employ polarised neutrons and two magnetic field regions. According to the semi-classical description, the first region will separate the two neutron spin states and the second region will recombine them again. Inelastic scattering from a sample occurs between the two field regions and induces a phase shift of the recombined states depending on the energy exchange of the scattering process. The distribution of the energy transfer will lead to a distribution of phases and will reduce the spin-echo amplitude for the neutron beam [3]. This principle leads to a high energy resolution in the order of neV [4]. Measurements of magnetic samples are possible [5,6], however it is challenging in cases where a sample depolarises the beam. An alternative method to NSE, that is not affected by sample induced depolarisation, is the Modulation of IntEnsity by Zero Effort (MIEZE) technique (shown in figure 1) introduced by Gähler [8], which utilizes only one magnetic field region created by resonant spin flippers [9]. However, the energy resolution of MIEZE spectrometers (~300-600 neV or 1-2 ns [10-12] and in the case of longitudinal MIEZE [13,14] 30 neV or 20 ns) is significantly lower than that of NSE instruments (~0.2 neV). For several systems an energy resolution on the order of ~120 neV corresponding to a Fourier time of 5 ns is sufficient [15, 16] and in other cases longer Fourier time is required [17, 18]. To achieve the highest possible resolution in MIEZE with a suitable beam size, large and homogeneous magnetic fields are required in the resonant spin flippers, furthermore the RF field must oscillate at several MHz. These fields and frequencies are available on the Larmor instrument at the ISIS pulsed neutron source, a versatile Small Angle Neutron Scattering (SANS) instrument which is also capable of employing a variety of spin echo techniques such as Spin Echo SANS (SESANS) [19] and Spin Echo Modulated SANS (SEMSANS) [20]. This instrument utilizes spin flippers, which employ soft iron pole shoe magnets. The pole shoes enable the generation of homogeneous high magnetic fields at relatively low current. Furthermore, unlike most resonant spin flippers these flippers use longitudinal solenoids to generate an RF field parallel to the beam. Hence, they do not employ any material in the beam, which could induce scattering, reducing the intensity and q (scattering vector) resolution. In addition, due to the pulsed nature of the ISIS neutron source, the setup studied here is a time of flight MIEZE spectrometer [21], which yields additional advantages by utilising simultaneously the large wavelength band of neutrons. 2 2. Theory The MIEZE arrangement is shown schematically in figure 1. The resonant spin flippers, like the magnetic field regions in NSE, introduce a longitudinal Stern-Gerlach effect. The first spin flipper introduces a kinetic energy shift between the two spin states, which causes the neutron spin states to move out of phase as they move towards the second spin flipper. The second flipper reverses this effect by overcompensating the kinetic energy shift, hence the phase difference will decrease towards the detector plane. Finally, once the neutrons reach the detector, the phase difference should be zero for every individual neutron. This is called the focusing plane. The energy difference between the two interfering spin states manifests itself as a beating of the polarisation in time [22, 23]. When an inelastic scattering sample is placed downstream of the second spin flipper the inelastic scattering process will influence both spin states in the same way. However, as the total kinetic energy has changed, the position of the focusing plane will move away from the detector plane. The distribution of scattering energies will yield a distribution in the focusing plane position and can be measured as a reduction of the depth of the modulation pattern. The amplitude of the MIEZE signal is therefore a measure of the energy transfer between the neutrons and the sample. The sample can be placed anywhere in the instrument after the first RF flipper, even downstream of the analyser, as its position will only affect the sensitivity, since the path length from sample to the detector dictates the magnitude of the shift of the focusing plane. Hence, with this technique all inelastic scattering, both incoherent and magnetic, is determined in one measurement. This also allows MIEZE to measure samples that depolarise the beam, [24] since sample induced spin rotations are not encoded when the sample is place downstream of the analyser. 3 Figure 1 Schematic of the MIEZE spectrometer as implemented on the Larmor instrument (top). The 𝜋 flippers 𝐹1 and 𝐹2(thickness d) are positioned, a distance D apart, within a guide field. 2 flippers are placed outside of the guide field to project the polarised neutrons on the y-axis. A sample can be placed downstream of the analyser. The semi-classical explanation (middle) shows the separation of kinetic energy introduced by the spin flippers. This splitting causes the two spin states, represented by the arrows, to move apart. The second resonant spin flipper F2 overcompensates this energy splitting and ensures that the phase difference between the two spin states is zero at the detector position. Note that the arrows indicate the locations of the respective wavepacket and not the locations of equal phase. Inelastic scattering from a sample placed in the beam (a distance 𝐿𝑠 from the detector) will result in a reduction of the modulation depth seen at the detector. The corresponding Fourier time is shown at the bottom depending on the sample position. Adapted from reference 7. 4 2.1. MIEZE Condition In a conventional MIEZE setup the RF flippers are placed in a zero field chamber so no spurious fields will distort the encoding. As shown in figure 1, the time of flight MIEZE setup described in this work utilizes a guide field that needs to be taken into account. The MIEZE focus, derived in [22, 25], which indicates the detector position where the modulation of intensity is maximal, is shifted and the modified MIEZE condition is given by 𝐿 = 2𝜔1𝐷+𝜔𝑔(𝐺1+𝐺2−𝐷) 2Δ𝜔𝑚 + 𝑑 2 (1) Where Δ𝜔𝑚 = 𝜔2 − 𝜔1, 𝜔𝑔 = 𝛾𝐵𝑔 with 𝐵𝑔 the guide field strength. All other parameters are defined in figure 1. If this modified condition is fulfilled, the intensity at the detector will oscillate in time with a frequency equal to 2Δ𝜔𝑚. This frequency is often referred to as the modulation frequency or MIEZE frequency. In the special case, where 𝐷 = 𝐺1 + 𝐺2 and the guide field is homogenous, this modified expression is reduced to the classical MIEZE condition, given in [22]. This special case was used in our TOF Larmor setup as the position of the flippers can easily be tuned to this condition. 2.2. Intermediate Scattering Function In the low energy transfer limit the ratio of the MIEZE amplitude with (Psample) and without (P0) sample yields the intermediate scattering function [1], which is the cosine Fourier transform of the scattering function: 𝑃𝑠𝑎𝑚𝑝𝑙𝑒 𝑃0 = 𝑆(𝑞, 𝜏) = ∫ +∞ −∞ 𝑆(𝑞, 𝜔) cos(𝜔𝜏) 𝑑𝜔 (2) Where 𝜔 is the energy transfer and the Fourier time, is given by 𝜏 = Δ𝜔𝑚𝑚2𝜆3 𝜋ℎ2 𝐿𝑠 (3) Where m is the neutron mass, 𝐿𝑠 the sample to detector distance, 𝜆 the neutron wavelength and ℎ is Plancks constant. Thus the Fourier time can be selected by tuning the instrument parameters such as the modulation frequency, the sample - detector distance or the neutron wavelength. In time-of-flight (TOF) MIEZE - a wavelength band is scanned in each TOF pulse; therefore a Fourier time range is scanned in a single pulse. Since the Fourier time is proportional to the reciprocal of the energy transfer, a large Fourier time corresponds to a good energy resolution. 2.3. TOF MIEZE Frequency Shift TOF MIEZE has a rather unique property at its disposal, namely that the modulation frequency is shifted if the MIEZE condition (eq. 1) is not satisfied [25, 26]. In a monochromatic approach the violation of the MIEZE condition introduces a wavelength dependent phase shift to the modulation. In time-of-flight the neutron wavelength is time dependent, thus the wavelength dependent phase shift, 5 manifests as a shift of the MIEZE frequency. Furthermore, due to the better wavelength resolution on time of flight setups compared to most monochromatic setups the modulation amplitude drops less quickly when the detector is moved out of focus. In the case of a misaligned setup, where the detector is displaced by a distance ΔL from the focal point the expectation value of the neutron spin in the x-direction is given by the following expression < 𝜎𝑥 >= ∫ R(λ, λ0, Δλ)cos (2𝛥𝜔𝑚 [𝑡 − ∞ −∞ 𝑚𝜆𝛥𝐿 ℎ ]) 𝑑𝜆 = 𝑅𝑒[𝐹(𝑅(𝜆, 𝜆0, Δ𝜆))] cos (2𝛥𝜔𝑚 [1 − 𝛥𝐿 𝐿 ] 𝑡) (4) Where 𝑅(𝜆, 𝜆0, Δ𝜆) is the normalized wavelength resolution function, which is averaged over and 𝑅𝑒[𝐹(𝑅(𝜆, 𝜆0, Δ𝜆))] indicates the real part of the Fourier transform. This is directly related to the envelope of the wavefunction and therefore the coherence length. The shifted frequency is thus given by 𝜔′ = 2Δ𝜔𝑚 (1 − Δ𝐿 𝐿 ) (5) This frequency shift simplifies the alignment procedure of the setup. Furthermore, it can yield additional information on non-symmetric inelastic scattering as the observed MIEZE frequency shift will have a different sign for energy gain or loss. 3. Experimental Setup Figure 2 shows a graphical representation of the experimental setup where a double v-cavity polarizer and a single mirror analyser are used to polarise and analyse the polarisation of the beam. The incident beam has a circular shape at the sample position (diameter 20 mm). A gradient RF flipper (not depicted) is placed between the polarizer and the first π/2 flippers allowing sequential measurement of both spin states to determine the beam polarisation. The π/2 flippers (v-coils) [27, 28] are employed at the entrance and exit of the guide field, which create a superposition of the spin up and down state. A guide field (~0.6 mT) spanning from the edges of both v-coils surrounds the setup. The instrument is setup such that 𝐷 = 𝐺1 + 𝐺2 (1.209m) and 𝐺1 = 𝐺2 = 𝐷 2 . As a result once the MIEZE focus is found using the frequency shift technique, one can double, triple etc. the frequencies of the spin flippers, without displacing the focus. 6 Figure 2 Schematic of the Larmor instrument in MIEZE mode, showing the most essential components: (1) double v-cavity polarizer, (2) v-coils, (3) guide field coil, (4) pole shoe magnets, (5) RF coils, (6) Flight tube, (7) Single mirror analyser, (8) Sample position (no samples were measured in these experiments) and (9) Detector. The spin flippers used on Larmor are shown in figure 3, these consist of a copper shielded longitudinal solenoid which generates the RF field and an electromagnet with soft iron pole shoes which allow static magnetic fields up to ~120 mT. The design of these RF flippers is very flexible as they can be operated in a gradient (or adiabatic) flipper mode [29, 27] and also a resonant flipper mode [9, 23]. The latter is used for the experiments described in this paper due to the lower power consumption and the large band of useable flipper frequencies using direct RF amplifier drive from 35 kHz to 3MHz. The low power consumption makes a resonant matching circuit obsolete, since the RF amplifiers are able to handle the reflected power. Thus, the amplifier outputs are connected directly to the RF solenoids. When used in time of flight mode the RF amplitude of a resonant flipper must be modulated using a 1/t function to match the exact π-flip for all wavelengths, where t is the time-of- flight from the source to the respective flipper. This ensures a high flipping efficiency [30] (>99%), over a wavelength range from 2.5 Å to 13.5 Å. 7 Figure 3 Picture of an RF spin flipper used by the Larmor instrument. This spin flipper has soft iron pole shoes to generate the required static magnetic field and a longitudinal RF coil to create the oscillating field. Furthermore the flipper has a gradient coil (not used in this experiment), which can be used to produce an adiabatic RF spin flip. The yoke pole shoe configuration can be translated and rotated on the table, thus allowing for both tilted and straight field regions. The latter is required for MIEZE, while the former is used in SESANS and SEMSANS. A vacuum exists within the flight tube to minimize air scattering. Each RF flipper is operated at four different frequencies (c.f. table I), giving rise to four different modulation frequencies. This results in a large Fourier time range at the sample position, which is 3.3 m from the detector, though no samples were used for these experiments. The wavelength range that could be used with reasonable statistics was 4 Å - 11 Å, with the peak of the spectrum being at roughly 4 Å. Table I RF flipper frequencies and the resulting modulation frequency (2Δf). 𝒇𝟏 [𝐌𝐇𝐳] 0.2484 0.9936 1.9872 2.4840 𝒇𝟐 [𝐌𝐇𝐳] 0.3000 1.2000 2.4000 3.0000 𝟐𝚫𝒇 [𝐌𝐇𝐳] 0.1032 0.4128 0.8256 1.0320 8 The detector is positioned 5.82 m from the second RF flipper. High frequency MIEZE setups require a detector with a fast response time and a thin detection path as this determines the time resolution. A Gas Electron Multiplier (GEM) type detector [31] was used, which employs a thin boron sheet where neutrons that are absorbed emit α particles [32, 33]. Such a detector has a sampling rate up to 100 MHz. Furthermore it has a 2D spatial sensitivity with a pixel size of 0.8 mm2, allowing for a spatial analysis of the MIEZE signals, which is useful for determining the spatial field homogeneity of the spectrometer. Note that it is not possible to measure the instantaneous polarisation as it varies in time, rather the detector measures the average polarisation over the sampling time interval and the detector area. This averaged polarisation is equal to the visibility, which is defined as 𝑉 = 𝐼+−𝐼− 𝐼++𝐼− (6) with 𝐼+ the intensity of the first spin state and 𝐼− the intensity of the second spin state [34]. Alternatively it can also be useful to look at the absolute value of the Fourier transform of the visibility, which is termed, spectral amplitude in this paper. 9 4. Results Figure 4 shows the MIEZE signals for various modulation frequencies. Figure 4 MIEZE signals at various frequencies with sinusoidal fits. The MIEZE frequencies are 103kHz (a), 413kHz (b), 826kHz (c) and 1.03MHz (d). The visibility shown here is the average visibility over a circular detector area with a diameter of 28 mm. Here, the modulation is shown in a limited wavelength range. The wavelength-dependent modulation can be found in figure 4. The 103 kHz signal was measured for a total of 60 minutes, the 413 kHz signal for 90 minutes, the 826 kHz signal for 105 minutes and the 1.03 MHz signal for 270 minutes. Measurement times are long due to the efficiency (~16% at 4 Å) of the GEM detector used which results from the thin detection volume. The respective sampling rates were 1 MHz for the 103 kHz signal and 10 MHz for the other signals. Further analysis of individual pixels demonstrated that there are no averaging effects, which negatively affect the visibility when averaging over the entire detector surface. Table II shows the measured modulation frequency against the expected modulation frequency. 10 Table II Expected modulation frequency, 𝑓 = Δ𝜔𝑚/𝜋 in MHz and the measured modulation frequency in MHz. The frequencies and their respective uncertainties are estimated from least squares fits to a sine function Expected (MHz) Measured (MHz) 0.1032 0.4128 0.8256 1.0320 0.1031±0.0014 0.4119±0.0005 0.8229±0.0067 1.0193±0.0337 Using a least squares regression the amplitude of the MIEZE signal is determined at various wavelengths (4 Å-11 Å for 1.03 MHz and 4 Å-8 Å for the others). This allows an extraction of the visibility as a function of the Fourier time at our Larmor sample position shown in Figure 5 for all four modulation frequencies. Figure 5. Visibility as a function of Fourier time plotted for various MIEZE frequencies. The visibility shown here is the average visibility over the detector, with the frequency given in the legend. Figure 5 shows that the setup has a Fourier time range from ~100 ps to 30 ns. This resolution is comparable with other MIEZE setups previously reported. However, it is unique as it combines high energy resolution and time of flight, while not employing any material in the beam, thus providing good q resolution for SANS applications. The MIEZE frequency shift (figure 6) due to the detector displacement was measured using a CASCADE detector and can be described well (within error) using eq. 5. 11 Figure 6. MIEZE frequency shift due to a displacement of the detector (~ 1 cm). Since the distance between the last flipper and the detector is 5.82 m, the expected frequency shift according to eq. 5 is 710 Hz. The observed frequency shift is 650 Hz, with an uncertainty of 100 Hz. 5. Discussion The results demonstrate the feasibility of time of flight MIEZE on the Larmor instrument, with an energy resolution that is comparable to that of other MIEZE instruments. MIEZE on Larmor is unique as it offers a high energy resolution, time of flight and employs minimal material in the beam. The time of flight option yields a broad range of Fourier times, which are scanned, during each TOF pulse. However, this means that the scattering vector, q, is also scanned during each pulse. Thus for a single instrument setting the measured intermediate scattering function is on a certain contour of q-τ. This is not optimal for weak signals that occur at specific points in q space. The detector area and its efficiency are too small to measure dynamics of isotropic scatterers, thus for the time being anisotropic samples (i.e. coherent scattering) would be better candidates for TOF MIEZE on Larmor. Depolarising samples are also good candidates for MIEZE, as it is more challenging to characterize them using classical neutron spin echo. A resolution of 5 ns is sufficient for most magnetic samples [18], hence lower MIEZE frequencies (i.e. 100 kHz-400 kHz) are most applicable to these situations. Furthermore the 103 kHz (and lower) mode could also be used with the SANS 3He based detector on Larmor, which would enable MISANS measurements [35]. This 3He based detector would also yield a larger efficiency enabling measurements of quasi-elastic scattering from samples such as water. Due to the lower time resolution of this detector the modulation frequency would have to be below 100 kHz. The q range covered by this detector would be 0.003 -- 0.7 Å−1. 12 Lower Fourier times could be achieved by reducing the modulation frequency further, by switching to π/2 MIEZE/MICE [36,37] or by placing the sample position closer to the detector. A combination of these methods would increase the Fourier time range of the instrument by another order of magnitude to the 10 ps domain. For high Fourier times the range cannot be increased significantly due to poor visibility. Figures 4 and 5 demonstrate that the visibility drops as the Fourier time is increased, the most plausible explanation for this is that the detector position did not perfectly coincide with the MIEZE focal point, as a result the two partial wavefunctions of the neutron would be slightly separated at the detector position. Due to the finite coherence length of the neutron, this separation results in decreased overlap and therefore interference between the two partial waves, leading to a reduction in visibility (see equation 4). Separation between the partial wavefunctions increases with Fourier time. This premise is supported by table 3, which shows that the measured frequency differs from the expected frequency, indicating that the detector position did not coincide with the focal point. The coherence length, which determines the sensitivity of the MIEZE signal to defocusing, is dictated by the wavelength resolution Δ𝜆/𝜆, or the TOF distribution (the distribution of neutron flight times). At Larmor, this resolution is roughly 1% and is primarily governed by the size of the moderator and the distance between the moderator and the instrument. Further optimisation of the experimental configuration is still possible through improvements of the analyser (reduce background) and detector system (efficiency) as well as the installation of background suppression elements such as evacuated flight paths and additional shielding. 6. Conclusion Time of flight MIEZE for the Larmor instrument has successfully been tested. Modulation frequencies up to 1MHz have been achieved and a Fourier time range from ~0.1 ns to 30 ns was measured using a wavelength band from 4 Å to 11 Å. The visibility remains above 0.5 up to 1 ns Fourier time. Thus the Larmor instrument offers a unique time of flight MIEZE option, which combines high energy resolution, comparable to other MIEZE instruments. Acknowledgements We thank C. Pappas for the discussion and input regarding this publication. Experiments at the ISIS Neutron and Muon Source were supported by a beam time allocation RB1900030 [38] from the Science and Technology Facilities Council. This research was funded by a NWO groot grant No. LARMOR 721.012.102. References [1] F. Mezei, Neutron spin echo: A new concept in polarized thermal neutron techniques, Zeitschrift für Physik A - Hadrons & Nuclei 255 (2) (1972) 146 -- 160. [2] P. Schleger, G. Ehlers, A. Kollmar, B. Alefeld, J.F. Barthelemy, H. Casalta, B. Farago, P. Giraud, C. Hayes, C. Lartigue, F. Mezei and D. Richter, Physica B 266 (1999) 49-55. 13 [3] R. Gähler, R. Golub, K. Habicht, T. Keller and J. Felber, Physica B 229 (1996) 1-17. [4] M. Monkenbusch and D. Richter, C. R. Physique 8 (2007) 845 -- 864. [5] B. Farago and F. Mezei, Physica 136B (1986) 100-102. [6] C. Pappas, E. Lelièvre-Berna, P. Bentley, E. Bourgeat-Lami, E. Moskvin, M. Thomas, S. Grigoriev and V. Dyadkin, Nucl. Instr. And Meth. A 592 (2008) 420 -- 427. [7] G. Brandl, R. Georgii, W. Häussler, S. Mühlbauer and P. Böni, Nucl. Instr. And Meth. A 654, (2011) 394-398. [8] R. Gaehler, R. Golub and T. Keller, Physica B 180 & 181, (1992) 899-902. [9] L.W. Alvarez and F. Bloch, Phys. Rev. 57 (1940) 111. [10] H. Hayashida, M. Kitaguchi, M. Hino, Y. Kawabata and T. Ebisawa, Physica B 397 (2007) 144 -- 146. [11] M. Hino, M. Kitaguchi, H. Hayashida, Y. Kawabata, S. Tasaki, T. Ebisawa, D. Yamazaki, R. Maruyama, K. Tanaka, N. Torikai, R. Inoue and T. Kanaya Physica B 385 -- 386 (2006) 1125 -- 1127. [12] Y. Kawabata, M. Hino, M. Kitaguchi, H. Hayashida, S. Tasaki, T. Ebisawa, D. Yamazaki, R. Maruyama, H. Seto, M. Nagao and T. Kanaya, Physica B 385 -- 386 (2006) 1122 -- 1124. [13] C. Franz, O. Soltwedel, C. Fuchs, S. Säubert, F. Haslbeck, A. Wendl, J.K. Jochum, P. Böni and C. Pfleiderer, Nucl. Instrum. Meth. Phys. Res., Sect. A 939 (2019) 22-29. [14] C. Franz and T. Schröder, Journal of large-scale research facilities, 1, A 14 (2015). [15] R. Georgii, G. Brandl, N. Arend, W. Häussler, A. Tischendorf, C. Pfleiderer, P. Böni and J. Lai, Applied Physics Letters 98, (2011) 073505. [16] F. Haslbeck, S. Säubert, M. Seifert, C. Franz, M. Schulz, A. Heinemann, T. Keller, P. Das, J.D. Thompson, E.D. Bauer, C. Pfleiderer and M. Janoschek, Physical Review B 99, (2019) 014429. [17] C. Pappas, L.J. Bannenberg, E. Lelièvre-Berna, F. Qian, C.D. Dewhurst, R.M. Dalgliesh, D.L. Schlagel, T.A. Lograsso and P. Falus, PRL 119, (2017) 047203. [18] M. Ohl, M. Monkenbusch, D. Richter, C. Pappas, K. Lieutenant, Th. Krist, G. Zsigmond and F. Mezei, Physica B 350 (2004) 147 -- 150. [19] M.T. Rekveldt Nucl. Instrum. Methods Phys. Res., Sect. B, 1996, 114 , 366 -370 [20] W.G. Bouwman, C.P. Duif, J. Plomp, A. Wiedenmann and R. Gähler, Physica B 406, (2011) 2357-2360. [21] M. Bleuel, M. Bröll, E. Lang, K. Littrell, R. Gähler and J. Lal Physica B 371 (2006) 297. [22] R. Golub, R. Gähler and T. Keller, et al. Am. J. Phys. 62 (9), (1994) 779-788. 14 [23] V.K. Ignatovich and F.V. Ignatovich, Am. J. Phys. 71, (2003) 1013-1024. [24] H. Hayashida, M. Hino, M. Kitaguchi, N. Achiwa and Y. Kawabata, Nucl. Instr. And Meth. A 600 (2009) 56 -- 59. [25] N. Geerits, S.R. Parnell, M.A. Thijs, W.G. Bouwman and J. Plomp, Applying Resonant Spin Flippers with Poleshoes and Longitudinal Radio Frequency Fields to Time of Flight MIEZE, arXiv:1911.02815 (2019). [26] T. Oda, M. Hino, M. Kitaguchi, P. Geltenbort and Y. Kawabata, Rev. Sci. Instrum. 87 (2016) 105124. [27] W.H. Kraan, S.V. Grigoriev, M.Th. Rekveldt, H. Fredrikze, C.F. de Vroege and J. Plomp Nucl. Instr. And Meth. A 510, (2003) 334-345. [28] W.H. Kraan, M.Th. Rekveldt and P.T. Por Nucl. Instr. And Meth. A 300 (1991) 35-42. [29] A.N. Bazhenov, V.M. Lobashev, A.N. Pirozhkov and V.N. Slusar Nucl. Instr. and Meth. A 332 (1993) 534. [30] H. Hayashida, M. Kitaguchi, M. Hino, Y. Kawabata, R. Maruyama and T. Ebisawa, Nucl. Instr. And Meth. A 574, (2007) 292-296. [31] https://www.bbtech.co.jp/en/products/ngem/ [32] M. Köhli, M. Klein, F. Allmendinger, A-K. Perrevoort, T. Schröder, N. Martin, C.J. Schmidt and U. Schmidt, J. Phys.: Conf. Ser. 746 (2016) 012003. [33] S. Uno, T. Uchida, M. Sekimoto, T. Murakami, K. Miyama, M. Shoji, E. Nakano, T. Koike, K. Morita, H. Satoh, T. Kamiyama and Y. Kiyanagi, Physics Procedia 26 (2012) 142 -- 152. [34] M. Strobl, M. Sales, J. Plomp, W. G. Bouwman, A. S. Tremsin, A. Kaestner, C. Pappas and Klaus Habicht, Sci. Rep. 5, (2015) 16576. [35] M. Bleuel, K. Littrell, R. Gähler and J. Lal, Physica B 356 (2005) 213 -- 217. [36] D. Yamazaki Nucl. Instr. and Meth. A 488 (2002) 623 -- 633. [37] J. Zhao, W.A. Hamilton, S. Lee, J.L. Robertson, L. Crow and Y.W. Kang Appl. Phys. Lett. 107 (2015) 113508. [38] DOI: 10.5286/ISIS.E.RB1900030. 15
1904.00430	1	1904	2019-03-31T15:07:24	Field emission from diameter-defined single-walled carbon nanotubes	[ "physics.app-ph" ]	Field electron emission (FE) from a structure-defined single-walled carbon nanotube (SWNT) was observed experimentally. A series of observations of the same SWNT by scanning electron microscopy, field emission microscopy and transmission electron microscopy were performed for the characterization of the SWNT emitter. This characterization work was very difficult because the observations affected the sample. Additionally, different sample setups were needed in each observation. We improved measurement conditions to minimize the influence to SWNT and developed techniques of sample setup for these observations. Combining these techniques, FE from an individual diameter-defined SWNT could be observed. The diameter of SWNT was $\sim 1.73\ \mathrm{nm}$ and the FE current of $2.2 \times 10^{-7}\ \mathrm{A}$ was obtained at a low applied voltage of $270\ \mathrm{V}$. We observed a dynamic behaviors of the SWNT emitter during FE.	physics.app-ph	physics	Field emission from diameter-defined single-walled carbon nanotubes 1Department of Physics, Tokyo University of Science, Shinjuku, Tokyo 162-8601, Japan Masaru Irita1, ∗ and Yoshikazu Homma1 Field electron emission (FE) from a structure-defined single-walled carbon nanotube (SWNT) was observed experimentally. A series of observations of the same SWNT by scanning electron microscopy, field emission microscopy and transmission electron microscopy were performed for the characterization of the SWNT emitter. This characterization work was very difficult because the observations affected the sample. Additionally, different sample setups were needed in each obser- vation. We improved measurement conditions to minimize the influence to SWNT and developed techniques of sample setup for these observations. Combining these techniques, FE from an indi- vidual diameter-defined SWNT could be observed. The diameter of SWNT was ∼ 1.73 nm and the FE current of 2.2 × 10−7 A was obtained at a low applied voltage of 270 V. We observed a dynamic behaviors of the SWNT emitter during FE. I. INTRODUCTION Field emission microscopy (FEM) was invented by Er- win Muller in 1936 [1]. The FEM image shows surface atomic structures directly and analysis of field electron emission (FE) characteristics reveal electronic properties of the surface of tip-shaped samples. FEM is used to study electron emitters of scanning electron microscopy (SEM), transmission electron microscopy (TEM), etc. In recent years, carbon nanotubes (CNT) have been stud- ied intensively because the unusually high aspect ratio as well as the chemical stability of CNT is the main advantage over the conventional metallic tips [2]. Par- ticularly, single-walled carbon nanotubes (SWNTs) are attractive nanomaterials having excellent properties the- oretically predicted [3]. Most of the properties have been confirmed experimentally. In the field of FEM, multi- walled carbon nanotubes (MWNTs) were mostly used as CNT emitters [4 -- 6]. Dean et al. demonstrated use of in- dividual SWNTs for extracting FE [7, 8]. They reported the effect of adsorption of molecules on the FE current from SWNTs [7]. At a high current regime exceeding a Fowler-Nordheim behavior, they showed field evapora- tion of the atoms on the end of SWNT, resulting in re- duced nanotube length [8]. They reported a spinning mo- tion of FEM images and current degradation due to the field evaporation. Marchand et al. reported the FE from growing individual SWNTs [9]. They discovered axial ro- tation of SWNTs during growth. Those works revealed extraordinary behaviors of SWNTs as field electron emit- ters. However, the FE properties from clean SWNTs be- low the evaporation regime have not been understood yet. Furthermore, although FEM images of SWNT caps have been studied [10], no systematic relationships between FE images and cap structures were found. So far, FE from a structure defined SWNT has not been explored yet because of the difficulty in the sample treatment. It is important to observe FE from structure-defined SWNTs to understand the FE characteristics from SWNTs. The ∗ [email protected] purpose of this study is to establish the observation tech- niques of SWNT emitter and to observe the FE from a structure-defined SWNT. We have developed technologies for fabrication of SWNT tips and could produce many SWNT tips stably at the yield rate of 25%. In the previous work, SWNT tips was examined as the probe of ultrahigh vacuum scan- ning tunneling microscopy (STM) [11]. We found that the length of SWNT was a crucial factor for the appli- cation to STM [11]. For FE, the length of SWNT is important too. In the case of a SWNT tip with length of ≥ 1000 nm, the FE current was not stable. We could not observe an FEM image because of the instability, and found disappearance of SWNT on the tip by SEM. Therefore, we needed to fabricate an isolated SWNT as short as possible on a W tip. The fabrication of SWNT tip required SEM obser- vation of the products to find an individually standing SWNT on the W tip. We selected only good samples for FEM observation among the products. Furthermore, the SWNT tip was observed by TEM and the diameter was determined. We need a series of observations of the same SWNT by SEM, FEM and TEM to understand the relationship between FEM images and the SWNT cap structure. This characterization work was very difficult because the observations affected the sample. Addition- ally, the sample setup was different in each observation. Success of observations depends greatly on the sample setup method. We improved techniques of sample setup for these observations. We show a few successful experi- mental results. II. EXPERIMENTAL In this study, we produced many W tips (length of 10 mm and 0.3 mm in diameter) simultaneously with an automated electropolishing device that had paralleled circuits of nanosecond-order switching. On the W tip, SWNT was directly synthesized by vacuum deposition of catalyst metals and chemical vapor deposition of carbon [11]. Fabricated SWNT tips were observed by SEM, FEM and TEM with minimized influence to the sample. The sample setup is different in each observation as shown in Fig. 1. SWNT tip samples were very likely to break with clumsy treatments. In particular, TEM sample prepara- tion was very difficult and SWNT disappeared often. We developed techniques for the observations as shown in the following section. Combining these techniques, FE from an individual diameter-defined SWNT could be observed. SEM observation First, the fabricated SWNT tips were observed by SEM and the length and position of SWNT on the W tip were determined. Ten W tips were mounted on each sample stage for SEM observation as shown in Fig. 1 (a). The fabricated SWNT tips were classified into three types as shown in Fig. 2. About 43% of products had standing SWNTs, but about 30% of SWNT tips were not usable for FEM because of off-axis SWNT growth or growth out of the tip apex. About 10% of SWNT tips were longer than 1000 nm on the tip apex. Only 3% of SWNT tips were shorter than 500 nm and good for this study. The primary electron energy for SEM observation was 0.5 keV. The electron beam did not break SWNT sam- ples. However, the SEM observation caused deposition of electron-irradiation-induced contamination and made the diameter determination by TEM difficult. We needed to reduce the electron influence. We tried to minimize the SEM observation time to ≤ 30 s with a total electron dose of ≤ 2 × 1017 cm−2. In this case, the SWNT diame- ter could be determined by TEM observation even after SEM observation. FEM observation Second, the SWNT tip was installed in a sample cham- ber of FEM. The W wire with SWNT tip was welded to a Ta filament as shown in Fig. 1 (b). Figure 3 shows a schematic of the FEM apparatus used in this study. The sample chamber was evacuated down to ultra high vacuum of 4.0 × 10−8 Pa. To clean the surface of SWNT tips, heating was carried out to the tip up to < 400◦C for 5 min. In this treatment, we took special care not to disappear the SWNT by watching gas species and their partial pressures in the mass spectra measured by a quadrupole mass spectrometer (QMS). During FEM ob- servation, we measured the FE current I, applied voltage V to extraction electrode, movie of FEM image and mass spectrum in the chamber at the same time. These quan- tities permit detailed analyses of the FE from SWNT. 2 FIG. 1. Photographs of sample setups for (a) SEM, (b) FEM and (c) TEM observations. FIG. 2. SEM images showing typical types of SWNT tips and their abundance. (a) Off-axis growth: 30%. (b) Long on-axis growth: 10%. (c) Short on-axis growth: 3%. FIG. 3. Schematic diagram of the FEM apparatus. The SWNT tip is mounted inside a vacuum chamber. The volt- age applied to the extraction electrode and the load current of the SWNT emitter are monitored during FE experiment. Emitted electrons are amplified through a micro-channel plate (MCP) and detected with a phosphor screen in front of the tip. FEM image on the screen is recorded with a CCD cam- era (the minimum illumination is 2 × 10−5 Lux and the image acquisition rate is ∼ 1/25 s/flame). Gas species are analyzed with a QMS. TEM observation Finally, after FEM observation, the SWNT tip was ob- served by TEM and the diameter was determined. Fig- ure 1 (c) shows the sample setup. In TEM observation, it is necessary to process the sample to the size of about 1 mm as shown in Fig. 1 (c). This was a very difficult task and SWNT disappeared often because of impact by mechanical processing. Nevertheless, we have obtained a few successful experimental results as follows, although we need to further improve the processing method. The primary electron energy for TEM was 200 keV. The electron beam can break the SWNT sample easily. Therefore, we need to observe the sample after all of other measurements. III. RESULTS AND DISCUSSION Figure 4 shows SEM and TEM images of the same SWNT tip sample. The length of SWNT was 800 nm as shown in Fig. 4 (a). The SWNT tip barely changed the shape before and after FEM observation as shown in Figs. 4 (a) and (b). After FE measurements, TEM observation was performed. However, the sample pro- cessing caused the SWNT tip bent and adhesion to the the W tip as shown in Fig. 4 (c). Accidentally, this made stable TEM observation possible, otherwise vibra- tion of the tip hindered observation of the tip at a high magnification. Although the very end of the tip could not be observed, the TEM image in Fig. 4 (d) clearly shows that the SWNT tip is a four-SWNT bundle with the average SWNT diameter of 1.73 nm. The isolated SWNT kept being extruded on the W tip until the SEM observation performed after FEM observation as shown in Fig. 4 (b). We believe that an isolated SWNT would be extruded from the bundle of SWNTs. If the SWNT tip was processed well, the apex of SWNT tip could not have been observed easily because of vibration of the can- tilever structure. Figure 5 (a) shows the FEM images. The FE from SWNT behaved like Figs. 5 (b)-(c). We observed these data at the same time. Each snapshot of FEM image in Fig. 5 (a) corresponds to the circles indicated on the I-t curve in Fig. 5 (b). From those results, the following three characteristics are drawn. First, the FE occurs with a low applied volt- age of 150 V as shown in Fig. 5 (b). Second, at the instant the extraction voltage was applied, the FE cur- rent was increased suddenly and then decreased. The FE current exhibits sawtooth shapes and depends on the applied voltage as shown in Fig. 5 (b). When the ap- plied voltage is fixed at 270 V, a stable FE current of 2.2 × 10−7 A could be obtained as shown in Fig. 5 (b). Third, the FEM image of SWNT sometimes twist back and forth slightly. We observed the twisting of FEM im- ages ∼ 1◦ between 20.8 s and 20.9 s in Fig. 5 (a). How- ever, the twisting cannot be recognized easily from the snapshot images because of a small twisting angle. It can be recognized easier in the movie (see supporting movie). This phenomenon was reflected to the FE current. The FE current change with twisting was 0.1 × 10−7 A at 270 V as shown in Fig. 5 (b). This phenomenon is dif- ferent from adsorption of gas molecule between 33.0 s and 35.0 s in Fig. 5 (a). The FE current change with gas adsorption was 0.2 × 10−7 A at 270 V, larger than 3 FIG. 4. Multifaceted observation results of the SWNT tip. (a, b) SEM images of the SWNT tip: (a) before FEM and (b) af- ter FEM. (c, d) TEM images of the SWNT tip observed after FEM: (c) low magnification image and (d) high magnification image. that of twisting. The twisting angle was small and the period was ∼ 1/25 s. Furthermore, we did not observe any gas peaks of carbon-bearing molecules induced by FE as shown in Fig. 5 (c). We have observed various SWNT tips, and observed similar twisting. At present, the exact reason of the FEM image twisting is not clear. The observed motion is different from the spinning mo- tion of FEM images due to the field evaporation [8] or "screw-dislocation-like" growth mechanism [9]. We spec- ulate that a twist motion of the SWNT bundle could be the origin of it. Here, we should comment on the transient behavior of the FE current upon increase in the applied voltage (Fig. 5 (b)). Dean et al. reported rapid FE current decays co- incided with the field evaporation of SWNT at high FE currents [8] and adsorption in non-ideal vacuum condi- tions [7]. However, our FEM observation condition was different from their environments. The FE current and applied voltage were much lower and did not induce the evaporation of SWNT. The FEM images did not change in the shape during the FE processes as shown in Fig. 5 (a). Observed FE current change was reproducible un- 4 FIG. 5. FEM observation results of the SWNT tip in 1.3 × 10−7 Pa. (a) Snapshots of the FEM image. The arrow shows a defect of MCP. The red squares show the same position on the screen. The FEM image is strongly related to the FE current. (b) Time variation of I and V . (c) Mass spectra during FEM experiment. The time axis is common in (b) and (c). The arrows in (b) correspond to each snapshot of FEM image. less the SWNT emitter was broken. We assume that the phenomenon is related to transient change in the tip tem- perature due to Joule heating by FE current, but details are under investigation. IV. CONCLUSIONS We have developed a multifaceted characterization method for SWNT tips, and succeeded in measuring field electron emission properties and observing FEM images from a diameter-defined SWNT. We showed dynamic be- haviors of the SWNT tip during FE and confirmed the advantage of low applied voltage to obtain FE. We hope to determine the cap structure of an SWNT from FEM images, by accumulating FEM images from diameter- defined SWNTs. [1] E. W. Muller, Zeitschrift fur Phys. 102, 734 (1936). [2] Y. Saito, Carbon Nanotube and Related Field Emitters: [7] K. A. B. Appl. Phys. Lett. 76, 375 (2000). Dean and R. Chalamala, Fundamentals and Applications (WILEY-VCH, 2010). [8] K. A. Dean, T. P. Burgin, and B. R. Chalamala, [3] S. Han and J. Ihm, Phys. Rev. B 66, 241402 (2002). [4] Y. T. Murata, Saito, Hata, and K. Jpn. J. Appl. Phys. 39, L271 (2000). [5] C. Oshima, K. Mastuda, T. Kona, Y. Mogami, and T. Yamashita, M. Komaki, Y. Murata, Phys. Rev. Lett. 88, 038301 (2002). [6] E. C. Heeres, T. H. Oosterkamp, Appl. Phys. Lett. 79, 1873 (2001). [9] M. Marchand, C. Journet, D. Guillot, J.-M. Benoit, B. I. Yakobson, and S. T. Purcell, Nano Lett. 9, 2961 (2009). Chalamala, Dean and [10] K. A. B. R. J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. 21, 868 (2003). [11] M. Irita, Y. Homma, and T. Miura, Phys. Rev. Lett. 108, 036804 (2012). and N. de Jonge, e-Journal Surf. Sci. Nanotechnol. 11, 105 (2013).
1805.02706	1	1805	2018-04-30T10:43:26	Generation of tunable, high repetition rate optical frequency combs using on-chip silicon modulators	[ "physics.app-ph", "physics.optics" ]	We experimentally demonstrate tunable, highly-stable frequency combs with high repetition-rates using a single, charge injection based silicon PN modulator. In this work, we demonstrate combs in the C-band with over 8 lines in a 20-dB bandwidth. We demonstrate continuous tuning of the center frequency in the C-band and tuning of the repetition-rate from 7.5GHz to 12.5GHz. We also demonstrate through simulations the potential for bandwidth scaling using an optimized silicon PIN modulator. We find that, the time varying free carrier absorption due to carrier injection, an undesirable effect in data modulators, assists here in enhancing flatness in the generated combs.	physics.app-ph	physics	Generation of tunable, high repetition rate optical frequency combs using on-chip silicon modulators KP NAGARJUN,1 VADIVUKARASSI JEYASELVAN,1 SHANKAR KUMAR SELVARAJA,1,2 AND V R SUPRADEEPA1,* 1Centre for Nano Science and Engineering, Indian Institute of Science, Bangalore 560012, India [email protected] *[email protected] Abstract: We experimentally demonstrate tunable, highly-stable frequency combs with high repetition-rates using a single, charge injection based silicon PN modulator. In this work, we demonstrate combs in the C-band with over 8 lines in a 20-dB bandwidth. We demonstrate continuous tuning of the center frequency in the C-band and tuning of the repetition-rate from 7.5GHz to 12.5GHz. We also demonstrate through simulations the potential for bandwidth scaling using an optimized silicon PIN modulator. We find that, the time varying free carrier absorption due to carrier injection, an undesirable effect in data modulators, assists here in enhancing flatness in the generated combs. References and links T. Steinmetz, T. Wilken, C. Araujo-hauck, R. Holzwarth, W. Hänsch, L. Pasquini, A. Manescau, S. D. M. T. Murphy, T. Udem, R. Holzwarth, A. Sizmann, L. Pasquini, C. Araujo-Hauck, H. Dekker, S. S. A. Diddams, L. Hollberg, and V. Mbele, "Molecular fingerprinting with the resolved modes of a S. Coen, H. G. Randle, T. Sylvestre, and M. Erkintalo, "Modeling of octave-spanning Kerr frequency S. T. Cundiff and J. Ye, "Colloquium: Femtosecond optical frequency combs," Rev. Mod. Phys. 75(1), 1. Odorico, M. T. Murphy, T. Kentischer, W. Schmidt, and T. Udem, "Laser Frequency Combs for Astronomical Observations," Science 321(5894), 1335–7 (2008). 2. D'Odorico, M. Fischer, T. W. Hänsch, and A. Manescau, "High-precision wavelength calibration of astronomical spectrographs with laser frequency combs," Mon. Not. R. Astron. Soc. 380(2), 839–847 (2007). 3. T. Udem, R. Holzwarth, and T. W. Hänsch, "Optical frequency metrology," Nature 416(6877), 233–237 (2002). 4. 325–342 (2003). 5. femtosecond laser frequency comb," Nature 445(7128), 627–630 (2007). 6. C.-B. Huang, Z. Jiang, D. E. Leaird, and A. M. Weiner, "High-rate femtosecond pulse generation via line-by-line processing of phase-modulated CW laser frequency comb," Electron. Lett. 42(19), 1114 (2006). 7. P. Del'Haye, A. Schliesser, O. Arcizet, T. Wilken, R. Holzwarth, and T. J. Kippenberg, "Optical frequency comb generation from a monolithic microresonator," Nature 450(7173), 1214–1217 (2007). 8. combs using a generalized mean-field Lugiato–Lefever model," Opt. Lett. 38(1), 37-39 (2013). 9. Science 332(6029), 555–559 (2011). 10. J. S. Levy, A. Gondarenko, M. A. Foster, A. C. Turner-Foster, A. L. Gaeta, and M. Lipson, "CMOS- compatible multiple-wavelength oscillator for on-chip optical interconnects," Nat. Photonics 4(1), 37–40 (2010). 11. V. Brasch, M. Geiselmann, T. Herr, G. Lihachev, M. H. P. Pfeiffer, M. L. Gorodetsky, and T. J. Kippenberg, "Photonic chip-based optical frequency comb using soliton Cherenkov radiation," Science 351(6271), 357–360 (2016). 12. H. Murata, A. Morimoto, T. Kobayashi, and S. Yamamoto, "Optical pulse generation by electrooptic- modulation method and its application to integrated ultrashort pulse generators," IEEE J. Sel. Top. Quantum Electron. 6(6), 1325–1331 (2000). 13. R. Wu, V. R. Supradeepa, C. M. Long, D. E. Leaird, and A. M. Weiner, "Generation of very flat optical frequency combs from continuous-wave lasers using cascaded intensity and phase modulators driven by tailored radio frequency waveforms.," Opt. Lett. 35(19), 3234–3236 (2010). 14. Y. Dou, H. Zhang, and M. Yao, "Generation of flat optical-frequency comb using cascaded intensity and phase modulators," IEEE Photonics Technol. Lett. 24(9), 727–729 (2012). T. J. Kippenberg, R. Holzwarth, and S. A. Diddams, "Microresonator-based optical frequency combs," 15. 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). 16. G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson, "Silicon optical modulators," Nat. Photonics 4(8), 518–526 (2010). 17. R. A. Soref and B. R. Bennett, "Electrooptical effects in silicon," IEEE J. Quantum Electron. 23(1), 123– 129 (1987). 18. P. P. Absil, P. De Heyn, H. Chen, P. Verheyen, G. Lepage, M. Pantouvaki, J. De Coster, A. Khanna, Y. Drissi, D. Van Thourhout, and J. Van Campenhout, "Imec iSiPP25G silicon photonics: a robust CMOS-based photonics technology platform," Proc. SPIE 9367, 93670V (2015). 19. "SiGe:C BiCMOS Technologies for MPW & Prototyping," https://www.ihp- microelectronics.com/en/services/mpw-prototyping/sigec-bicmos-technologies.html. 20. V. Torres-Company, J. Lancis, and P. Andres, "Lossless equalization of frequency combs," Opt. Lett. 33(16), 1822–1824 (2008). 1. Introduction in applications Frequency combs have been extensively used like WDM optical communications [1], astronomy [2,3], spectroscopy and metrology [3, 4] and sensing [5]. Conventionally, frequency combs are generated using mode-locked lasers or bulk optical modulators. Combs generated in this manner utilize bulky setups, have fixed, low repetition rates and require expensive feedback electronics [6], making them unsuitable for several applications including optical communications. Recently, several on-chip frequency comb generation techniques have been proposed to overcome these drawbacks [7]. These on-chip schemes have the key advantage of being compact, portable and can be mass-produced. On-chip comb generation techniques in the C-band have typically been investigated in Silicon Nitride resonators, where cascaded four-wave mixing is used to generate Kerr combs [7–11] . These combs, though having a large number of lines [11] depending on the quality factor of the resonator, typically have fixed repetition rates. Further, the input laser has to be locked to a mode of the resonator and any drift of the laser or the resonator over time has to be compensated. In the absence of such stabilization, it is known that the comb ceases to exist. It is strongly desired to have an integrated, high repetition rate, optical frequency comb which is very stable over an extended period of time with the ability to easily change the center frequency as well as the repetition rate. In this work, we demonstrate such a source. The comb generation mechanism is based on Strong phase modulation. Phase modulating an external continuous wave (CW) laser with an RF sinusoid can be used to generate multiple comb lines which results from the generation of harmonic side-bands [12]. Comb lines generated this way can have tunable repetition rates based on the frequency of the driving RF sinusoid and a tunable center frequency based on the frequency of the CW laser. This technique of comb generation along with envelope shaping has been extensively demonstrated with bulk Lithium-Niobate modulators. However, with phase modulation alone, frequency combs generated this way result in combs with missing lines and poor flatness and require cascading of a series of phase and intensity modulators with complex driving schemes to achieve flat combs [13,14]. Recently, this idea of using strong phase modulation for comb generation was investigated using Silicon-Organic Hybrid (SOH) modulators in an intensity modulator configuration [15]. Since the electro-optic effect in Silicon is weak, an organic cladding was utilized to obtain the required phase modulation using SOH modulator configuration. A flexible and tunable property of the optical material could potentially increase the frequency of operation beyond some of the non-linear crystal. However, the following two disadvantages could weight against SOH. Firstly, the polymer should be polled at very high-voltage at an elevated temperature followed by controlled cooling, which is not feasible in a highly integrated circuit. Secondly, the back-end needs to be opened to access the Silicon waveguide, which could lead to long-term ion-diffusion and subsequent degradation of the Si device layer. In this work, we utilize a charge-injection based Silicon phase modulator fabricated using a standard Silicon photonics foundry process for the comb generation. Unlike, the SOH platform the devices require no post-processing and can accommodate multiple devices with varying specifications. Since the fabrication is done using mature technology, it allows seamless integration with driver and control electronics, unlike other platforms. Silicon based charge injection modulators, have become a popular means by which high speed data (several 10's of Gbps) is modulated onto an optical carrier [16]. These devices work by changing the free carrier dispersion in a Silicon optical waveguide through charge injection from doped regions and the change in refractive and absorptive indices is governed by the Soref-Bennett relation [17]. Apart from having a small device footprint it is compatible with standard CMOS fabrication technology and is thus easily integrated with other Silicon photonic devices. In typical micro-ring based comb generators based on four wave mixing which is the more common technique for generation of integrated combs [7–11], combs with bandwidths exceeding 650 nm have been demonstrated [11] . However, only coarse tuning of the center frequency is possible due to the periodic resonances of the ring. Further, they have a fixed and very high repetition rate (for example 189.2 GHz in [11]). This is dictated by requirement of fixed and short lengths of the micro-ring resonators necessary to obtain the high Q factors required for nonlinear generation. This makes such comb generation systems unsuitable for applications which require agility in the form of tunable center frequency and tunable, reasonable repetition rates. An example would be comb generation for super-channel based optical communications. Here, we demonstrate combs whose center frequency can be both coarse (~nm) and fine (~pm) tuned, which makes the combs generated in this manner very attractive on-chip DWDM and super-channel sources. The combs generated here also have the additional advantage of a tunable repetition rate. In this work, we demonstrate repetition rate tuning between 7.5 to 12.5 GHz. However, it is to be noted that the tunability of the repetition rate is only limited by the bandwidth of the modulator and can be further reduced as desired. In each case, we demonstrate ~8 lines in a 20dB bandwidth. Further, owing to the fact that this technique does not employ resonant structures, this system is very robust to thermal drift. In our experiments, we observed highly stable combs that remained unaltered even during continuous operation of over 12 hours. In this work, we also investigate bandwidth scaling through electro-optic simulations of an optimal Silicon PIN modulator. A PIN modulator is chosen for the simulations since lower linear loss in comparison to PN modulators enable longer modulators which provide greater phase modulation and thus more lines. We find that free carrier absorption due to carrier injection shapes the amplitude to enable high quality combs from a single modulator instead of a cascade of intensity and phase modulators as needed in Lithium-Niobate equivalent [13]. We demonstrate that over 37 lines is possible in a 1cm device with a flatness of the spectrum better than 5dB. We further compare our simulation to a pure phase modulation and observe that our Si PIN-based design produces comb lines with a flattened spectrum without missing lines in contrast to the pure phase modulation case. 2. Experimental results Fig. 1. (a) Side view and top view of the designed device (b) Microscope image of device used for comb generation. (c) Setup for comb generation. (d) Simulated combs from the setup at 10GHz repetition rate. The devices were fabrication through Europractice multi-project wafer run [18]. We designed and customized a 4.5 mm long travelling-wave PN-phase modulator in a standard 220 nm -SOI wafer. The Silicon rib waveguide width is 450 nm with an etch-depth of 160nm and shoulder height of 60 nm (Fig. 1(a), Fig. 1(b)). An average symmetric doping concentration of 1018/cm3in the N and P regions was chosen for the entire length of the device to prevent excessive absorption due to high doping and simultaneously provide sufficient refractive index modulation. The modulator employs a co-planar travelling wave electrode design. An RF-drive from a tunable microwave source (upto 20GHz) and ~7.8Vpp is used to drive one end of the phase modulators while the other end is terminated with a high frequency 50-ohm load. This method of comb generation does not require a DC bias since it's based on a pure phase modulation based process. Figure 1(c) shows the schematic of the comb generation setup. A tunable C-band laser with100 kHz linewidth is used as the input. On-chip grating couplers are used in conjunction with single mode fibers to couple light into/out of the modulator. The fiber to waveguide loss in this setup is found to be 2.5dB. The polarization of the input light is optimized using a polarization controller such that maximum power is coupled through the grating coupler. Figure 1(d) shows the simulated comb at 10 GHz repetition rate. Figures 2(a)-(d) shows the experimental results with increasing RF power inputs for a center frequency of ~1557.84 nm. As expected, we observe an increase in the number of comb lines with the RF power. For an RF power of 27dBm corresponding to ~7.8Vpp, we observe 8 comb lines in a 20dB bandwidth. OSATunable laserSilicon PM50Ω(b)(c)(d)15561556.515571557.5-80-60-40-200Wavelength (in nm)Power (in dB)N+(a)450nmP+N+P+Fig. 2. Frequency Comb generation in a travelling-wave Silicon PN-modulator for various RF- powers at 10 GHz (a) +14dBm (b) +20 dBm (c) +23 dBm (d) +27 dBm. The center frequency of the comb can be continuously tuned, with the tunability only being limited by the bandwidth of the input-output grating couplers which can routinely have 1dB bandwidths spanning the C-band. Here, we show a representative plot of center frequency shifting by ~1.2 nm (Fig. 3). Fig. 3. Tuning of the center frequency of the comb from ~1556.6 nm (blue) to ~1557.84nm (red) We also performed experiments of tuning of repetition rate of the comb from 5GHz to 15GHzand representative plots of the comb at 7.5GHz and 12.5GHz are shown (Fig. 4). Beyond 12.5 GHz, we observe a limitation that the number of comb lines decrease with increasing frequency. The number of lines reduces to 4 in a 20dB bandwidth at 15GHz. (a)(b)(c)(d)1557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm)1557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm)1557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm)1557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm)(b)1555.751556.251556.751557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm) Fig. 4. Comb Repetition rate tuning(a) 7.5GHz (b) 12.5GHz. We also observed that these combs are stable over very long durations. In our experiments, the combs were allowed to run freely for over 12 hours and were unaltered in their shape. The measured linear absorption loss for the modulator is ~16dB. It is observed that the linear absorption loss does not depend on RF drive frequency. We also observe that the maximum CW power that can be applied on the device is 75mW beyond which the net output power saturates due to nonlinear absorption. There is also significant contribution to this from the passive input/output coupling waveguides which together is ~3.3 mm. The method for comb generation used in this work, namely strong phase modulation in a single pass modulator is anticipated to generate a coherent comb as has been observed in Lithium-Niobate modulator based systems. In order to evaluate any degradation of coherence or linewidth across the generated comb, we performed the following experiment with the comb at 10GHz repetition rate. Starting from the center line, we choose neighboring pairs of comb- lines using a programmable spectral filter and generated their respective beat notes using a high-speed photodetector and RF spectrum analyzer. Figure 5 shows the beat signal spectra as we move across the comb to one side (right). Fig. 5. Beat notes between successive comb lines. Frequency offset of 0 set to the repetition rate of the comb. (a) center and 1st line to the right (b) 1st and 2nd line (c) 2nd and 3rd line (d) 3rd and 4th line (e) 4th and 5th line (f) 5th and 6th line to the right respectively. 1557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm)(a)(b)1557.251557.751558.251558.75-80-60-40-200Wavelength (in nm)Power (in dBm)-40-2002040-110-90-70-50-30Frequency offset (in kHz)Power (in dBm)-40-2002040-110-90-70-50-30Frequency offset (in kHz)Power (in dBm)-40-2002040-80-60-40-200Frequency offset (in kHz)Power (in dBm)-40-2002040-90-70-50-30-10Frequency offset (in kHz)Power (in dBm)-40-2002040-80-60-40-200Frequency offset (in kHz)Power (in dBm)(a)(b)(c)(d)(e)(f)-40-2002040-80-60-40-200Frequency offset (in kHz)Power (in dBm)Here, the frequency offset of '0' is fixed to be the comb repetition rate. We consistently measure very similar beat notes across the spectrum with a sub-kHz bandwidth. This is substantially smaller than the natural laser linewidth of 100 kHz. At these bandwidth levels, we see that there is no degradation of the coherence of the additionally generated comb lines and the input laser linewidth dominates the coherence properties. 3. Bandwidth scaling simulations Further bandwidth scaling of frequency combs in Silicon is possible at the systems level by increasing the electro-optical interaction length or by cascading more phase modulators. At the device level, improved RF-coupling (i.e. by use of an appropriate impedance matching network) leads to generation of more phase modulated sidebands and thus more comb lines. This could also be achieved by increasing the doping concentration of the PN junction, the tradeoff being higher free carrier absorption leading to increased absorption losses. Thus, to increase the number of lines, it is preferable to move to a PIN configuration which can support much longer lengths due to reduced absorption losses. The integration of high frequency, high power RF platforms with CMOS based photonics is available for example in MPW [19], however an alternative approach involves the use of 3D integration of the two platforms using Through Silicon Vias (TSV's) or by wire-bonding the two platforms following a system in a package approach. To investigate the comb generation process, we demonstrate the generation of mostly flat frequency combs by electro-optic simulations on an optimized 10mm long, 450nm wide, 250 nm high Silicon rib waveguide with a shoulder height of 50nm, having abrupt highly doped P++/I/N++ junctions with average doping concentrations of 1019/cm3 that is routinely used for ohmic contacts with current implantation technology and an intrinsic doping concentration of 1015/cm3 respectively in a PIN configuration. The device is biased through idealized contacts placed above the highly doped regions that start 50 nm from the rib edge and has a planarizing oxide cladding. These device parameters characterize a narrow-base PIN diode. We model the change in the effective refractive and absorption indices of the intrinsic region due to a sinusoidal modulating input (See Fig. 6) using the Soref-Bennett relations [17] as We study the large-signal, electro-optical, AC response of the device to a 4Vpp, 10 GHz sinusoidal input by simulation on a drift-diffusion equation simulator, Silvaco Atlas. Concentration dependent Shockley-Read-Hall recombination, concentration dependent mobility transport, band gap narrowing effects along with Auger recombination and Fermi- Dirac carrier statistics is incorporated in the simulation model. 0.8221818188.8108.5108.510610ehehnNNNN Fig. 6. Simulated change ineffective refractive index and absorption generated from an electro- optic drift diffusion simulation for a strongly sinusoidal drive. The change in refractive index and absorption profiles derived from the simulation (Fig. 6) show mutually out-of-phase saturation regions. The maximum change in refractive index is ~6.7x10-4. These saturation regions arise from reduction of charge injection due to the series resistance of the long diode considered in the simulation. The schematic for mostly-flat frequency comb generation is shown in Fig 7(a). We consider a 10mm long modulator which would correspond to ~ 4VπLπ, chosen to ensure a high modulation index required for strong phase modulation, whose Lπ=2.3mm. This modulator is driven by a 10 GHz microwave source. The loss of our idealized device is found to be ~23 dB. A ring-resonator based notch filter is used for enhancement of the flatness of the comb. Fig 7(b) shows the spectrum of the generated 10GHz comb. A broad, flat spectrum is seen, however, there is substantial power around the carrier. The reason for that can be seen from the plot of refractive index variation vs time as shown in Fig 6. Since the response to the sinusoid is not strictly sinusoidal, in regions where the refractive index profile is flat, there is no modulation and thus substantially higher fraction of the power rests in and around the carrier. If the response of the modulator was strictly sinusoidal, this effect would not have been observed. The flatness however can be recovered using a simple ring based notch filter by optimally tuning the parameters. The parameters of the notch filter are obtained by solving a simple three-dimensional grid search to determine the coupling coefficient, loss coefficient and ring radius that maximize a quadratic metric of spectral flatness. The additional loss due to the shaping by the ring based notch filter is ~5dB. Figure 7(b) shows the filter transmission (red) for getting best flatness of the total system. The frequency comb smoothed by the notch filter is observed to have enhanced spectral flatness (Fig. 7(c)). Fig. 7(a) Silicon PIN phase-modulator followed by a ring-resonator based notch filter. (b) Comb lines (blue) from the modulator with notch filter response (red) (c) Flattened Comb lines generated from optimal PIN-modulator device post notch filtering. With 37 usable lines in a 5dB band. (d) Equivalent frequency comb with missing lines generated by pure phase modulation. Unlike field-effect based modulators that have small and constant time-dependent linear absorption, the time varying free carrier absorption in Silicon charge-injection phase modulators has important consequences for the comb envelope. Fig 7(d) shows the anticipated comb from pure phase modulation with a modulator which has the same level of phase modulation ( ). As is seen, multiple lines are missing, and the flatness is poor. This is expected in the case of pure sinusoidal modulation. However, from our simulations, and from the experimental results (Fig 2) the frequency comb produced by the Silicon phase modulator yields a flat spectrum with no missing lines. The origin of this crucial improvement lies in the amplitude shaping which accompanies phase modulation in these modulators. It has been rigorously shown in theory and experiment [13] [20] that when the amplitude of the time- domain waveform is shaped into a more pulse like structure during sinusoidal phase modulation, the phase modulation itself is equivalent to a time-to-frequency mapping which creates an optical spectrum similar to the time domain waveform. Thus, the shape of the spectrum (Fig 7(c)) is anticipated to be similar to a scaled compliment of the amplitude shaping function in Fig 6. Further details on the effects of amplitude shaping in comb generation can be found in [13] [20]. The cascaded intensity and phase modulator based comb generator systems based on Lithium-Niobate typically have a net loss of around 11dB in conventional operating conditions [14]. Thus, though the experimentally demonstrated system is lossier, it is of comparable performance. However, we now have the advantage of on-chip operation. In addition, we believe that this is not a fundamental issue and it is possible to tune the absorption induced loss by tweaking the doping profile and length of the device appropriately. 4. Summary and future work 46.7210maxn37 lines(a)(b)(c)(d)In this paper, we demonstrated the generation of flat, tunable, high-repetition rate frequency combs in Silicon charge injection based PN-modulators. The modulators were generated in a fully CMOS compatible process. These combs have the unique advantage of continuously tunable center frequency across the C-band and an independently tunable repetition rate (upto 12.5 GHz). We believe that the envelope of the combs can be further enhanced by tuning the doping parameters and introducing asymmetry in both doping concentration and alignment with the optical mode. We also demonstrated through simulations the potential for bandwidth scaling using an optimized Silicon PIN modulator based on similar principles. We demonstrated that the time varying free carrier absorption due to carrier injection, an undesirable effect in data modulators, provides substantial benefits here for improving the spectral flatness and overcoming missing lines present in combs generated using pure phase modulation. Funding Office of the Principal Scientific Advisor, Govt of India (Prn.SA/ADV/Photonics/2015-16). Acknowledgments The authors would like to thank Prof SV Raghavan for extensive discussions and initiation of the program. KP Nagarjun would like to thank MHRD for financial support, Sushobhan Avashthi for extensive discussions regarding device behavior and would also like to acknowledge BS Vikram for help with the measurement setup, Abhishek Mishra and Chethan Kumar of DESE, IISc for their help with the simulations, and Nikhil R Kumar for his help with GDS preparation. V R Supradeepa and Shankar Kumar Selvaraja thank MeitY for support through Sir Visvesvaraya young faculty research fellowships.
1804.06927	1	1804	2018-04-18T21:52:02	Tomographic imaging of inhomogeneous non-local media using fractional order models	[ "physics.app-ph", "physics.comp-ph" ]	We investigate a generalized tomographic imaging framework applicable to a class of inhomogeneous media characterized by non-local diffusive energy transport. Under these conditions, the transport mechanism is well described by fractional-order continuum models capable of capturing anomalous diffusion that would otherwise remain undetected when using traditional integer-order models. Although the underlying idea of the proposed framework is applicable to any transport mechanism, the case of fractional heat conduction is presented as a specific example to illustrate the methodology. By using numerical simulations, we show how complex inhomogeneous media involving non-local transport, can be successfully imaged if fractional order models are used. In particular, results will show that by properly recognizing and accounting for the fractional character of the host medium not only allows achieving increased resolution but, in case of strong and spatially distributed non-locality, it represents the only viable approach to achieve a successful reconstruction.	physics.app-ph	physics	Tomographic imaging of inhomogeneous non-local media using fractional order models University of Notre Dame, Notre Dame, Indiana, 46556. Salvatore Buonocore Fabio Semperlotti∗ Purdue University, Ray W. Herrick Laboratories, West Lafayette, Indiana, 47907 (Dated: March 15, 2021) We investigate a generalized tomographic imaging framework applicable to a class of inhomo- geneous media characterized by non-local diffusive energy transport. Under these conditions, the transport mechanism is well described by fractional-order continuum models capable of capturing anomalous diffusion that would otherwise remain undetected when using traditional integer-order models. Although the underlying idea of the proposed framework is applicable to any transport mechanism, the case of fractional heat conduction is presented as a specific example to illustrate the methodology. By using numerical simulations, we show how complex inhomogeneous media involv- ing non-local transport, can be successfully imaged if fractional order models are used. In particular, results will show that by properly recognizing and accounting for the fractional character of the host medium not only allows achieving increased resolution but, in case of strong and spatially distributed non-locality, it represents the only viable approach to achieve a successful reconstruction. I. INTRODUCTION Tomography, or tomographic imaging, is a particu- lar class of inverse problems that enables reconstruct- ing an image of the internal structure of a body (or of specific material properties) without requiring intrusive access to it. Depending on the specific method, differ- ent types of penetrating waves or field transport phe- nomena can be selected to probe the body. Typical ex- amples include electrostatic and electromagnetic waves (employed in Electrical Impedance Tomography-EIT and Magnetic Resonance Imaging-MRI, respectively)1, acoustic waves (acoustic tomography)2, and thermal fields (thermography)3. In the past few decades, tomo- graphic techniques have found many engineering applica- tions including biomedics, atmospheric science, geology, oceanography, plasma physics, materials science, astro- physics, and acoustics4 -- 6. More recently, tomography was also successfully applied to structural health moni- toring and non-destructive testing for the inspection and characterization of structures and materials3,7 -- 12. inverse methodologies fit, As many inverse problems, tomographic methods are heavily dependent on the accuracy of the underlying mathematical model used to predict the response of the medium. Typically, in a least-squares sense, a mathematical model to experi- mentally measured data by iteratively tuning the model parameters1,13,14. It is evident that the ability to recon- struct accurate images and to improve the resolution is strongly dependent on the ability of the model to capture and properly simulate the underlying physical response of the system. Typical applications of tomographic meth- ods deal with domains of analysis that are highly inho- mogeneous (e.g. including geometric or material inho- mogeneities), have complex material properties (e.g. vis- coelastic materials), and are eventually characterized by hybrid and non-local transport behavior. If not properly accounted for, these properties limit both the sensitivity and resolution of the methodology and can even prevent entirely the possibility to reconstruct a meaningful im- age. To-date, several numerical and experimental studies have demonstrated that field transport processes in com- plex inhomogeneous media (e.g. random and/or porous materials) can occur according to hybrid modalities15 -- 20 and anomalous behavior21 -- 24. The hybrid transport typ- ically involves a combination of propagating and diffusive mechanisms25 -- 29. The diffusion processes can be either standard, that is characterized by a typical Gaussian distribution of the field quantities, or anomalous where instead an α-stable distribution emerges. In general, diffusive processes are characterized based on the functional relationship that describes the evolution of the variance of the particle displacement with time. This relationship is expressed as (cid:104)x(cid:105)2 ∼ tγ where the normal diffusion is characterized by a linear scaling γ = 1, while the anomalous diffusion occurs for γ (cid:54)= 1. In particular, processes described by γ > 1 correspond to diffusion phenomena that evolve in time faster than the standard (Gaussian) process and are therefore denoted as superdiffusive. On the contrary, pro- cesses that are characterized by 0 < γ < 1, are denoted as subdiffusive, because the variance increases more slowly than what predicted by the classical (Gaussian) diffusion model. Subdiffusive processes can be described at a mi- croscopic level by continuous time random walk (CTRW) models employing a probability density function for the waiting time of a particle jump characterized by heavy tailed distribution30,31. On the other hand, superdiffu- sive phenomena can be described by L´evy flights or L´evy walks where the probability density function of the step length is a heavy-tailed distribution (that emerges as a consequence of the long-range interactions among parti- cles). These unconventional transport mechanisms are such as long- typically driven by non-local effects, range hereditary mechanisms34,35, and have been shown to be well interactions32,33, memory and described by fractional order models. At macro- scopic scales, the CTRW subdiffusive processes can be mapped to a time-fractional diffusion equation while the superdiffusive processes map into a space-fractional diffusion equation. Examples of these unusual transport behavior can be observed in different media and fields of application including, but not limited to, porous soils36 -- 40, heterogeneous aquifers21,41, and underground fluid flow24,42. Meerschaert et al.42,43 performed an extensive research campaign focused on the transport of contaminants in heterogeneous soils and porous materials and showed that the underlying advective- diffusive behavior exhibits indeed a fractional nature. Other studies28,31,44 -- 47 have highlighted the occurrence of regular and anomalous diffusion mechanisms in the propagation of electromagnetic and acoustic waves in random and periodic materials. Propagation of light through fog48 or murky media49, of sound through a forest or a urban environment50 -- 52, of seismic waves through the ground53,54 are all examples of practical applications where hybrid and non-local transport mech- anisms can take place. In all these situations, fractional calculus was shown to be a powerful tool to properly capture and simulate complex transport mechanisms in continuum media. From the discussion above, it appears that the integra- tion of fractional order models within tomographic meth- ods could improve the imaging performance and even allow capturing details and inhomogeneities that would otherwise go undetected when using conventional imag- ing approaches based on integer order models. In this paper, we present a numerical investigation into the application of fractional order continuum mod- els to tomographic imaging techniques. A few applica- tions of fractional calculus to inverse problems have al- ready been explored in the literature. Cheng et al.55 successfully solved a one-dimensional fractional diffusion inverse problem to determine the order of the temporal fractional derivative and the spatially varying diffusion coefficients. Miller et al.56 formulated a parameter iden- tification problem based on the fractional diffusion equa- tion. Battaglia et al.57 proposed the solution of an in- verse heat conduction problem for parameter estimation based on non-integer forward models. The precision of the inverse procedure proposed in their method heavily depended on the accuracy of the identified model that had to be determined by experiments. Murio58,59 pre- sented the numerical solution of an ill-posed problem con- sisting in determining analytical functions of the bound- ary temperatures and the heat fluxes based on transient temperature measurements at some internal point of a 1D conductor. Jin and Rundell60 investigated the degree of ill-posedness of a series of theoretical and numerical in- verse problems based on fractional differential equations involving the Caputo definition in both time and space. They found out that the fractional character of the op- erator can either improve or worsen the conditioning of the inverse problem based on the type of input data and quantities to be reconstructed. Kirane et al.61 solved a 2 2D source reconstruction problem for a time fractional diffusion equation using biorthogonal sets of functions. In this study, we explore the capabilities and per- formance of fractional tomography applied to two- dimensional domains. More specifically, we refer to frac- tional tomography as a tomographic approach relying on fractional-order continuum models for the solution of the forward problem. The use of these models within a to- mographic imaging framework is expected to impact the reconstruction performance at multiple levels: 1. It was previously discussed that anomalous diffu- sion processes, such as those arising in imaging of highly aberrating and scattering media, are asso- ciated with non-Gaussian distributions of the field quantities having noticeable heavy-tails. Note that a distribution of a random variable is said to be heavy-tailed if the moment generating function of the distribution function always diverges62. Typi- cally, the content of the tails is discarded because not considered as a primary source of information for the imaging process. However, while this could be a reasonable assumption when the domain is dominated by classical diffusion (in this case the tails of the Gaussian distribution have a low con- tent of information), in presence of anomalous dif- fusion the tails contain a non-negligible amount of information about the interior structure of the do- main to be imaged. Hence, neglecting the heavy tails results in considerable and irremediable loss of information. 2. Recognizing and exploiting the fractional nature of the host medium and the information contained in the heavy tails can represent a turning point in the development of accurate imaging technologies ca- pable of sensing through highly inhomogeneous and scattering media. 3. In order to extract meaningful information from the non-Gaussian distribution of the field quanti- ties, the underlying mathematical model should be able to simulate these hybrid and anomalous phys- ical mechanisms and to relate them to the physical parameters of the medium. In order to clearly illustrate the methodology and with- out limiting the generality of our approach, we focus the following study on diffusive transport processes. An ex- ample of a classical tomographic method that falls un- der this category is Thermal Tomography (TT)3,8,63,64, where diffusive heat transfer is used as a probing mecha- nism to sense the medium. Note that the results pre- sented below could be easily generalized to other ap- proaches whose transport mechanism can be effectively described by fractional models. II. FRACTIONAL TOMOGRAPHY: IMAGING BASED ON FRACTIONAL ORDER MODELS Tomographic imaging techniques are usually formu- lated as iterative inverse problems in which measured data are fit, in a least-squares sense, to a mathematical model that simulates the response of the host medium1. Although many inversion methodologies have been pre- sented over the years, the most common approach to solve complex geometries is based on iterative numeri- cal techniques. These methods typically consist of two main steps: the forward and the inverse problem. The forward problem simulates the response of the system assuming that the excitation conditions and the system parameters (e.g. material properties, etc.) are known. The inverse problem relies on an optimization approach designed to minimize the least-square error between the measured and the predicted response at selected mea- surement points. The system parameters, or a subset of them, are typically selected as design variables. In this work, we use fractional order models in the forward problem in order to describe the heat transport process at the basis of the thermographic technique. We will show that the introduction of fractional models has two main effects on the general formulation of the to- mographic problem: 1) if the behavior of the real physi- cal system is fractional, the model will capture informa- tion that would otherwise go undetected when using an integer order model, however 2) if the behavior is not fractional but the system is simply inhomogeneous, the fractional formulation will provide an enlarged parame- ter space (which now includes the order of the differential operator) in which the reconstruction can be performed. From a general perspective, thermal tomography re- constructs the internal (thermal) properties of a solid medium based on the measurements of boundary (or sur- face) temperature fields. The data acquisition procedure is performed by heating the sample at a specific location and measuring the resulting surface temperature at mul- tiple locations. The same procedure is repeated for sev- eral source locations in order to improve the conditioning of the inverse problem and increase the accuracy of the reconstruction. In the numerical analyses performed in this study, the stationary heat sources were placed near the boundaries as shown in Fig. 1. The response of the system was estimated in terms of absolute temperature measurements performed at the sensor locations (Fig. 1). The thermographic technique is typically formulated as an optimization process based on iterative solution of the forward and inverse problem. The thermal properties, such as the conductivity or the heat capacity, are typi- cally used as design variables. In the forward problem, the thermal properties are as- sumed known (either due to an initial guess or to val- ues available from the previous iteration) so that the mathematical model allows estimating the thermal field in the entire domain. In the inverse problem, the mea- sured temperature values at selected locations are com- pared with numerical predictions in order to find an up- 3 dated distribution of thermal properties that minimizes the least-square error. The reconstruction of different thermal properties is an indicator either of material in- homogeneities or of possible defects, depending on the specific application of the tomographic technique. 1. Forward problem: fractional heat transport The fractional heat diffusion process in a 2D domain is described by the following fractional partial differential equation65: ∂γT (x, y, t) ∂tγ c1− (x, y) · ∂αT (x, y, t) = c1+(x, y) · ∂αT (x, y, t) ∂+xα + + c2+(x, y) · ∂βT (x, y, t) + ∂+yβ (1) ∂−xα c2− (x, y) · ∂βT (x, y, t) ∂−yβ + s(x, y, t) where T (x, y, t) is the temperature field at time t in the finite domain a < x < b, c < y < d, the terms c1± (x, y), c2± (x, y) are spatially dependent functions representing the thermal diffusivity and are associated with the left- and right-handed fractional operators as indicated by the directional sign notation. s(x, y, t) represents the heat source. In the following simulations, we will assume that c1+(x, y) = c1− (x, y) = c2+(x, y) = c2− (x, y) = c(x, y) which means that we consider a symmetric non-local dif- fusion process. The thermal diffusivity in Eq.(1) has di- mension Lαt−γ 40. The notation ∂± in Eq. (1) refers to the left- and right-handed Riemann-Liouville fractional derivative defined as: (cid:90) x (cid:90) b a x (Dα a+f )(x) = (Dα b−f )(x) = ∂αf (x) ∂+xα = ∂αf (x) ∂−xα = 1 Γ(n − α) 1 Γ(n − α) dn dxn dn dxn f (ξ)dξ (x − ξ)α+1−n (2) (ξ − x)α+1−n f (ξ)dξ where Γ is the Euler gamma function and n is an integer such that n − 1 ≤ α ≤ n. Equation (1) represents a general form of the fractional wave-diffusion equation capable of representing several forms of transport mechanisms66. When 0 < γ ≤ 2 and α = β = 2, the equation describes a time-fractional pro- cess having diffusion-like characteristics when 0 < γ ≤ 1 and wave-like characteristics when 1 < γ ≤ 2. When γ = 1 and α = β = 2 we recover the classical diffu- sion equation. For γ = 1 and 0 < [α, β] ≤ 2, Eq.(1) describes a space-fractional diffusion process whose so- lutions belong to the superdiffusion regime67. The su- perdiffusive behavior is the results of dynamics dom- inated by long-range interactions (the so called L´evy flights), where the step length distribution decays asymp- totically as x−(α+1). This is also the interval of interest for the present study in which we want to explore the regime characterized by heavy-tailed distributions of the field quantities due to L´evy flights dominated dynamics. In the following numerical analysis, we will consider the case of γ = 1 and α = β. Once complemented with proper boundary conditions, this space-fractional diffu- sion equation (1) can be solved numerically. The thermographic problem considered below is solved using the following initial conditions: T (x, y, t = 0) = T0(x, y) (x, y) ∈ Ω ∪ ∂Ω and boundary conditions: T (x, y, t) = 0(x, y) ∈ ∂Ω t > 0, (3) (4) where Ω and ∂Ω indicate the 2D domain and its bound- ary (Fig. 1(b)), respectively. The initial value problem defined by the set of equations (1,3,4) can be solved numerically using a shifted Grunwald-Letnikov (GL) finite difference (FD) scheme67. More specifically, the two-sided shifted Grunwald-Letnikov scheme65 was selected due to its abil- ity to respect the local symmetry of the solution at ev- ery node of the domain. Fig. 1(b) shows a uniform discretization grid used for the numerical solution of equation Eq.(1). The grid has a uniform spacing with ∆x = (b − a)/I and ∆y = (d − c)/J, where I and J are the total number of discretization nodes in the x and y directions, respectively. The finite difference approx- imation to T (xi, yj, tn) will be denoted with T n ij where xi = i∆x, yj = j∆y and tn = n∆t. FIG. 1. (a) Schematics of the thermal tomographic compu- tational setup indicating the location of the transducers used for both excitation and sensing. (b) The numerical uniform grid used to discretize the fractional diffusion equations. Following the above definition, the general form of the shifted Grunwald-Letnikov formulas are67: dαf (x) d+xα = lim M+→∞ 1 hα dαf (x) d−xα = lim M−→∞ 1 hα gk · f [x − (k − 1)h] gk · f [x + (k − 1)h] (5) where M+, M− are positive integers, h+ = (x − a)/M+, h− = (b− x)/M− and k = 1, 2, 3...M(±). The coefficients gk are the Grunwald weights expressed as: M+(cid:88) M−(cid:88) k=0 k=0 4 (cid:40) gk = 1 (−1)k (α)(α−1)...(α−k+1) k! k = 0 k = 1, 2, ...M(±) (6) Using the shifted GL relations (5) to discretize Eq.(1), we obtain65: ij − T n T n+1 ij ∆t (cid:20) i+1(cid:88) k=0 = cn ij hα j+1(cid:88) gk · T n i−k+1 + gk · T n I−i+1(cid:88) k=0 J−j+1(cid:88) gk · T n j−k+1 + gk · T n j+k−1 i+k−1+ (7) (cid:21) + sn ij k=0 k=0 that can be explicitly solved for T n+1 The stability condition of this scheme for the case 1 < ij . α < 2 is given by65: β = ∆t hα ≤ 1 2 · α · c 2. Inverse problem (8) The solution of the inverse problem is performed ac- cording to a classical iterative minimization approach fol- lowing the Levenberg-Marquardt (LM) method. Accord- ing to the LM method: P k+1 = P k +(cid:2)µkΩk + (J k)T J k(cid:3)−1 (J k)T [Y − T (P k)](9) where P is the vector of design variables consisting in the system parameters (e.g. the coefficients c or the frac- tional order α), T is the vector of numerically estimated temperatures, Y is the vector of measured temperatures, µk is a regularization parameter1, Ωk is a regularization matrix chosen in our simulations as the identity matrix, J is the Jacobian matrix, and the superscript ()T indi- cates the transpose. The term µkΩk is a regularization term used to damp the instabilities introduced by the ill- conditioned Hessian matrix J T J 13. The generic element of the Jacobian, or sensitivity matrix, is defined as: Jlm = ∂Tl ∂Pm (10) with l = 1, 2, ..., L and m = 1, 2, ..., M , where M is the total number of unknown parameters, L is the total num- ber of measurements (L ≥ M ), Tl is the lth estimated temperature, and Pm is the mth unknown parameter. The iterative problem described by Eq. (9) is repeated until a prescribed error threshold is satisfied. In our sim- ulations, the exit conditions was chosen to be a threshold value of the residual error Y − T (P k) < 10−5. III. PROBLEM DESCRIPTION AND NUMERICAL EXPERIMENTS In this section, we apply the methodology discussed above to reconstruct the spatial distribution of the ther- mal diffusivity of a 2D domain. The main difference, compared to traditional thermography, is that we let the domain have a non-homogeneous distribution of the order of the underlying differential operator. In other terms, we consider media in which the transport pro- cess is non-homogeneous and can result from a combi- nation of standard and anomalous diffusion. The follow- ing three scenarios will be considered: 1) the reference domain exhibits a fractional-order behavior with local- ized integer-order sub-domains, 2) the reference domain exhibits integer-order behavior with localized fractional- order sub-domains, 3) the reference domain exhibits integer-order behavior with local sub-domains having ei- ther different fractional order or diffusivity coefficients with respect to the background. The selection of these case of studies has been made to illustrate some specific characteristics of the fractional operator within the con- text of an inverse problem. A. Case 1: domain with fractional-order behavior and integer-order inhomogeneities This case addresses a situation in which the thermal field transport within the physical domain is described by a space-fractional diffusion equation. The domain in- cludes two localized inhomogeneities consisting of areas where the transport is of integer order. The reference map showing the distribution of the order of the spatial differential operator is shown in Fig. 2(a). Note that this condition can be mapped to a physical situation where the domain of interest is characterized by long-range in- teractions (i.e. an anomalous diffusion process) other than for small localized areas that are controlled instead by short-range interaction (hence resulting in a standard diffusion process). FIG. 2. Case 1: (a) Reference map of the 2D domain showing the distribution of the spatial order of the differential opera- tor. (b) Reconstructed map of the thermal diffusivity distri- bution when an integer-order model is used for the forward problem. (c) Reconstructed map of the operator order when a fractional model is used for the forward problem. In order to illustrate the need for fractional order mod- els in inverse problems involving anomalous transport behavior, we compared the results of the reconstruction obtained using both a fractional- and an integer-order forward model. In the case of the integer order model the only unknown parameter that can be reconstructed is the thermal diffusivity. The reconstructed spatial dis- tribution of the thermal diffusivity c(x, y) is shown in Fig. 2(b). A direct visual inspection of the results in- dicates that the integer-order model is completely un- 5 able to capture the inhomogeneities in the domain due to the inability to simulate the underlying anomalous trans- port process. Conversely when a fractional-order model is used, the process successfully identifies (both in am- plitude and location) the inhomogeneities present in the domain (Fig. 2(c)). In this case, the order of the frac- tional operator was selected as the unknown parameter describing the characteristics of the domain. The results above is likely the most significant among the three cases and clearly highlights the necessity of us- ing a fractional-order model when the imaging is sought in domains characterized by anomalous field transport. They also highlight another important feature of the frac- tional inverse problem. The fractional order can be used as an additional parameter to track both the presence, the magnitude, and the location of the inhomogeneities. In other terms, the fractional order expands the parame- ter space that can be used to characterize inhomogeneous domains (which could be relevant in applications such as non-intrusive sensing and non-destructive evaluation). The numerical experiments presented in the next two cases are intended to address and characterize the role of the fractional order as an additional indicator of inhomo- geneities, that is beyond the traditional parameters such as the transport coefficients c(x, y). B. Case 2: domain with integer-order behavior and fractional-order inhomogeneities This case explores the ability to image a domain char- acterized by local fractional-order inhomogeneities dis- tributed in an otherwise integer-order background. Such situation physically corresponds to a domain in which the heat transfer in the background occurs according to a standard (Fourier) diffusion process but it experiences localized areas in which the transport is non-local. The reference case to be reconstructed is shown in Fig. 3(a) where the background of the 2D domain has order α = 2 other than for a small area in the center where α = 1.8. The reconstruction was performed by us- ing a fractional-order model where the thermal diffusivity coefficients and the fractional order were both treated as unknown variables. To solve the inverse problem, the iterative approach was initialized using the true values of both c and α in the background. The inverse procedure exited with a residual error of 5.44e − 10 well below the set threshold. Fig. 3(b) and Fig. 3(c) show the results of the recon- structed maps. In general, both parameter distributions show clear traces of the location of the inhomogeneity. However, a marked difference is visible in their corre- sponding amplitude. More specifically, the order of the differential operator is well captured by returning a value of α = 1.835 which is within 2% error with respect to the reference. This was expected because the original inhomogeneity was set in terms of an anomaly in the order of the op- erator. The thermal diffusivity distribution (Fig. 3(b)) 6 which ultimately is connected to changes in the proper- ties of the host medium. Fig. 5 summarizes the results of the numerical recon- struction. The background values of c and α are the same of the previous case, and the two inhomogeneities are rep- resented by an anomaly in the thermal diffusivity of in- tensity c = 0.02 [Lα/t] (Fig. 5(b)) and an anomaly in the fractional order distribution equal to α = 1.2 (Fig. 5(a)). The iterative inverse procedure was initialized using as initial guess the true values of the background distribu- tions of c and α. The inverse procedure exited with a residual error of 1e − 5. FIG. 3. Case 2: (a) Reference map for case 2 expressed in terms of the order of the operator. Reconstructed maps of (b) the diffusion coefficients and of (c) the fractional order. The inhomogeneity is represented by a discontinuity in the distri- bution of the fractional order α while the inverse procedure is set to reconstruct simultaneously the spatial distribution of the diffusion coefficients and of the fractional order. instead shows a variation of 16% with respect to the background in correspondence to the inhomogeneity de- spite the original configuration was uniform in terms of c(x, y). This result was not unexpected because previ- ous studies68 had shown that there is a direct correspon- dence between spatially inhomogeneous coefficients of an integer order differential equation and the corresponding order of a matched fractional equation. We note that, although the location and extent of the anomaly is still well captured, the absolute value of the reconstructed co- efficient is not necessarily meaningful and does not allow quantifying of the inhomogeneity. FIG. 4. Case 2 : Error maps. Percentage error maps of the (a) reconstructed thermal diffusivity and of the (b) reconstructed fractional order. C. Case 3: domain with combined inhomogeneities of the order and diffusivity In this scenario, we explore the possibility to simulta- neously reconstruct two distinct types of inhomogeneities affecting either the fractional order or the thermal diffu- sivity. This case is intended to underline the fundamental difference in the role played by both the thermo-physical properties (captured in the coefficients of the differential equation) and by the physical trasport mechanisms (cap- tured by the fractional operator). In fact, variations of the thermal diffusivity do not affect the functional form of the solution of the heat conduction equation, while perturbations in the order of the operator do affect the transport process by introducing non-local effects. The underlying idea is that a change in the nonlocal parame- ter α indicates the occurrence of non-local heat transfer FIG. 5. Case 3: (a-b) Reference and reconstructed maps of the (c) fractional order and of the (d) diffusion coefficients. The damages are modeled by a discontinuity in the distribu- tion of fractional order α and a discontinuity in the distribu- tion of the thermal diffusivity c. The inverse procedure is set to reconstruct simultaneously the spatial distribution of the diffusion coefficients and of the fractional order. The reconstructed distributions could correctly iden- tify the location and intensity of both the inhomo- geneities in terms of fractional order (Fig. 5(c)) and ther- mal diffusivity (Fig. 5(d)). A marked inhomogeneity emerged in the reconstructed map of the diffusion co- efficients. This result should have been expected given that, as shown in case 2, there is a strict correspondence between the spatially inhomogeneous coefficients of an integer order differential equation and the correspond- ing order of a matched fractional equation. This link between the system parameters also tend to reduce the accuracy of the identification of the fractional order, as seen in Fig. 5(c). The indeterminacy that occurs when both parameters are perturbed cannot be resolved math- ematically unless more information on the host system was provided. As an example, in cases where a priori in- formation on the inhomogeneity was available, then the inverse problem could be solved using constraints on the optimization problem so the facilitate resolving indeter- minacy. Such situations might correspond to cases in which, as an example, a knowledge of the materials in- volved or of possible structural defects was available from other measurements. Fig. 6(a)−(b) shows the percentage error maps for the thermal diffusivity and the fractional order. FIG. 6. Case 3: Error maps. Percentage error maps of the (a) reconstructed fractional order and of the (b) reconstructed thermal diffusivity. The results of the numerical experiments suggest that the augmented parameter space resulting from the use of fractional order models has the potential to improve sig- nificantly the accuracy of the reconstruction procedure when compared to the conventional tomographic tech- niques. In fact, the nonlocal parameter introduced by the frac- tional formulation (i.e. the order of the operator) serves as a new design variable able to capture the underlying nature of the heat conduction process as well as of de- scribing the anomalous behavior observed in complex and inhomogeneous media. In addition, fractional tomogra- phy opens the way to exploit hybrid field transport (i.e. a combination of propagating and diffusive mechanisms) as a probing mechanism, whereas classical tomography is constrained to the use of a single transport process. Overall, the above results suggest that the enrichment of the parameter space due to the integration of frac- tional order models in tomographic methods, may have important effects on the sensitivity of the reconstruction process with respect to classical tomography. IV. CONCLUSIONS 7 framework for the implementation of This study presented the general theoretical and math- ematical frac- tional tomographic imaging. This general class of tech- niques was conceived to achieve non-intrusive and non- destructive imaging in a variety of applications where the host medium is characterized by anomalous and hybrid field transport. Among these applications, we can in- clude imaging of viscous, nonlocal, and highly scattering media. The proposed fractional tomographic method was tested via numerical experiments. In this study, the fractional nature of the model was limited to the spa- tial derivative in order to target specifically nonlocal and anomalous diffusion processes. Numerical results showed that the order of the operator can be used as a sensi- tive and effective parameter to probe and image the in- terior of a medium. This approach provides, not only a tomographic framework with an augmented parameter space, but also a powerful method for sensing and imag- ing in nonlocal complex media that traditionally pose ex- tremely challenging conditions. Even though further re- search is required to establish constitutive relations able to connect the inhomogeneity of the system parameters to the fractional order, the present study shows clear evidence of the potential of the fractional tomographic framework and its ability to considerably enhance sensi- tivity and resolution in remote sensing applications. V. ACKNOWLEDGMENTS The authors gratefully acknowledge the financial sup- port of the National Science Foundation under the grant CMMI CAREER #1621909. ∗ [email protected] 1 D. S. Holder, Electrical impedance tomography: methods, history and applications (CRC Press, 2004). 2 W. Munk, P. Worcester, and C. Wunsch, Ocean acoustic tomography (Cambridge University Press, 2009). 3 V. P. Vavilov, D. A. Nesteruk, V. V. Shiryaev, A. Ivanov, and W. Swiderski, Russian Journal of Nondestructive Test- ing 46, 151 (2010). 4 H. Ammari, An introduction to mathematics of emerging biomedical imaging, vol. 62 (Springer, 2008). 5 S. J. Norton, Geophysical Journal International 97, 161 (1989). 6 R. Bayford, Annu. Rev. Biomed. Eng. 8, 63 (2006). 7 E. Verstrynge, H. Pfeiffer, and M. Wevers, Smart Materials and Structures 23, 065022 (2014). 8 W. Swiderski, in Proceedings of the Tenth Annual Confer- ence of the Materials Research Society of Serbia (2009), vol. 115, pp. 800 -- 804. 9 L. Zhao, J. Yang, K. Wang, and F. Semperlotti, Smart Materials and Structures 24, 065044 (12 pp.) (2015), ISSN 0964-1726. 10 S. Buonocore, M. Sen, and F. Semperlotti (USA, 2016), vol. 9803, p. 98030Y (8 pp.), ISSN 0277-786X. 11 T. Tallman, F. Semperlotti, and K. Wang (2012), vol. 8348, pp. The Society of Photo -- Optical Instrumentation Engi- neers (SPIE); American Society of Mechanical Engineers -- , ISSN 0277786X. 12 T. N. Tallman, F. Semperlotti, and K. Wang, Journal of In- telligent Material Systems and Structures 26, 2565 (2015), ISSN 1045389X. 13 M. N. Ozisik, Inverse heat transfer: fundamentals and ap- plications (CRC Press, 2000). 14 M. P. Bendsøe, O. Sigmund, M. P. Bendsøe, and O. Sig- mund, Topology optimization by distribution of isotropic material (Springer, 2004). 15 F. Mainardi, Chaos, Solitons & Fractals 7, 1461 (1996). 16 F. Mainardi, Applied Mathematics Letters 9, 23 (1996). 17 F. Mainardi and M. Tomirotti, Transform methods and special functions, Sofia 171 (1994). 18 W. Chen and S. Holm, The Journal of the Acoustical So- ciety of America 114, 2570 (2003). 19 W. Chen and S. Holm, The Journal of the Acoustical So- ciety of America 115, 1424 (2004). 20 F. Mainardi and M. Tomirotti, arXiv preprint arXiv:1008.1341 (2010). 21 D. A. Benson, S. W. Wheatcraft, and M. M. Meerschaert, Water Resources Research 36, 1403 (2000). 22 D. A. Benson, R. Schumer, M. M. Meerschaert, and S. W. Wheatcraft, in Dispersion in heterogeneous geological for- mations (Springer, 2001), pp. 211 -- 240. 23 J. H. Cushman and T. R. Ginn, Water resources research 36, 3763 (2000). 24 S. Fomin, V. Chugunov, and T. Hashida, in Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences (The Royal Society, 2005), vol. 461, pp. 2923 -- 2939. 25 A. Asatryan, P. Robinson, R. McPhedran, L. Botten, C. M. de Sterke, T. Langtry, and N. Nicorovici, Physical Review E 67, 036605 (2003). 26 P. Sheng, Introduction to wave scattering, localization and mesoscopic phenomena, vol. 88 (Springer Science & Busi- ness Media, 2006). 27 P. Sheng, Scattering and localization of classical waves in random media, vol. 8 (World Scientific, 1990). logical Formations (Springer, 2002), pp. 211 -- 240. 8 Schumer, D. Benson, M. Meerschaert, and Journal of Contaminant Hydrology 43 R. S. Wheatcraft, 48, 69 (2001). 44 L. A. Cobus (2016). 45 M. Burresi, V. Radhalakshmi, R. Savo, J. Bertolotti, K. Vynck, and D. S. Wiersma, Physical review letters 108, 110604 (2012). 46 S. Stutzer, U. Naether, T. Kottos, R. Vicencio, M. Molina, A. Tunnermann, S. Nolte, D. Christodoulides, and A. Sza- meit, in The European Conference on Lasers and Electro- Optics (Optical Society of America, 2013), p. JSIII 1 4. 47 Z. Fellah and C. Depollier, The Journal of the Acoustical Society of America 107, 683 (2000). 48 E. Belin, F. Christnecher, F. Taillade, and M. Laurenzis, in Optical Engineering+ Applications (International Society for Optics and Photonics, 2008), pp. 70880O -- 70880O. 49 M. E. Zevallos L, S. Gayen, M. Alrubaiee, and R. Alfano, Applied Physics Letters 86, 011115 (2005). 50 P. Chobeau, Ph.D. thesis, Le Mans (2014). 51 M. Hornikx, Building and Environment 106, 409 (2016). 52 J. Kang, The Journal of the Acoustical Society of America 107, 1394 (2000). 53 Z. E. A. Fellah, S. Berger, W. Lauriks, C. Depollier, and M. Fellah, Journal of Applied Physics 93, 296 (2003). 54 Z. E. A. Fellah, M. Fellah, W. Lauriks, C. Depollier, J.-Y. Chapelon, and Y. Angel, Wave motion 38, 151 (2003). 55 J. Cheng, J. Nakagawa, M. Yamamoto, and T. Yamazaki, 28 P. Barthelemy, J. Bertolotti, and D. S. Wiersma, Nature Inverse problems 25, 115002 (2009). 453, 495 (2008). 56 L. Miller and M. Yamamoto, Inverse Problems 29, 075013 29 G. Bayer and T. Niederdrank, Physical review letters 70, (2013). 3884 (1993). 30 R. Metzler and J. Klafter, Physics reports 339, 1 (2000). 31 J.-P. Bouchaud and A. Georges, Physics reports 195, 127 (1990). 32 V. E. Tarasov and G. M. Zaslavsky, Communications in Nonlinear Science and Numerical Simulation 11, 885 (2006). 33 V. E. Tarasov, International Journal of Solids and Struc- tures 51, 2900 (2014). 34 V. E. Tarasov and G. M. Zaslavsky, Communications in Nonlinear Science and Numerical Simulation 13, 1860 (2008). 35 V. E. Tarasov and E. C. Aifantis, Communications in Non- linear Science and Numerical Simulation 22, 197 (2015). 36 M. Caputo and F. Mainardi, Pure and Applied Geophysics 91, 134 (1971). 37 M. Caputo and F. Mainardi, La Rivista del Nuovo Cimento (1971-1977) 1, 161 (1971). 57 J.-L. Battaglia, O. Cois, L. Puigsegur, and A. Oustaloup, International Journal of Heat and Mass Transfer 44, 2671 (2001). 58 D. Murio, Computers & Mathematics with Applications 56, 2371 (2008). 59 D. Murio, Computers & Mathematics with Applications 53, 1492 (2007). 60 B. Jin and W. Rundell, Inverse Problems 31, 035003 (2015). 61 M. Kirane, S. Malik, and M. Al-Gwaiz, Mathematical Methods in the Applied Sciences 36, 1056 (2013). 62 J. Nolan, Stable distributions: models for heavy-tailed data (Birkhauser New York, 2003). 63 V. Vavilov, P. Bison, C. Bressan, E. Grinzato, and S. Marinetti, in Proc. Eurotherm Seminar (1992), vol. 27, pp. 7 -- 9. 64 R. Kline, W. Winfree, and V. Bakirov, in AIP Conference Proceedings (2003), vol. 657, pp. 682 -- 688. 38 R. L. Bagley and P. J. Torvik, Journal of Rheology 30, 133 65 M. M. Meerschaert and C. Tadjeran, Applied Numerical (1986). 39 F. Meral, T. Royston, and R. Magin, Communications in Nonlinear Science and Numerical Simulation 15, 939 (2010). 40 F. Mainardi, Fractional calculus and waves in linear viscoelasticity: an introduction to mathematical models (World Scientific, 2010). 41 D. A. Benson, M. M. Meerschaert, and J. Revielle, Ad- vances in water resources 51, 479 (2013). Mathematics 56, 80 (2006). 66 F. Mainardi, Y. Luchko, and G. Pagnini, arXiv preprint cond-mat/0702419 (2007). 67 M. M. Meerschaert and C. Tadjeran, Journal of Computa- tional and Applied Mathematics 172, 65 (2004). 68 J. P. Hollkamp, M. Sen, and F. Semperlotti, Journal of Sound and Vibration 419 (2016). 42 D. Benson, R. S. Wheatcraft, Schumer, M. Meerschaert, and in Dispersion in Heterogeneous Geo-
1908.08230	1	1908	2019-08-22T07:14:34	Molecular characterization of macroscopic aerogels of single-walled carbon nanotubes	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Single-walled carbon nanotubes (SWCNT) can be assembled into various macroscopic architectures, most notably continuous fibers and films, produced currently on a kilometer per day scale by floating catalyst chemical vapor depositionand spinning from an aerogel of CNTs. An attractive challenge is to produce continuous fibers with controlled molecular structure with respect to the diameter, chiral angle and ultimately(n,m)indices of the constituent SWCNT molecules. This work presents an extensive Raman spectroscopy and high resolution transmission electron microscopy study of SWCNT aerogels produced by the direct spinning method. By retaining the open structure of the SWCNT aerogel, we reveal the presence of both semiconducting and metallic SWCNTs and determine a full distribution of families of SWCNT grouped by optical transitions. The resulting distribution matches the chiral angle distribution obtained by electron microscopy and electron diffraction. The effect of SWCNT bundling on the Raman spectra, such as the G line shape due to plasmons activated in the far-infrared and semiconductor quenching, are also discussed. By avoiding full aggregation of the aerogel and applying the methodology introduced, rapid screening of molecular features can be achieved in large samples, making this protocol a useful analysis tool for engineered SWCNT fibers and related systems.	physics.app-ph	physics	Molecular characterization of macroscopic aerogels of single-walled carbon nanotubes Bel´en Alem´ana, Juan J. Vilatelaa, a IMDEA Materials Institute. Eric Kandel, 2, Tecnogetafe, 28906, Getafe, Madrid (Spain) 9 1 0 2 g u A 2 2 ] h p - p p a . s c i s y h p [ 1 v 0 3 2 8 0 . 8 0 9 1 : v i X r a Abstract Single-walled carbon nanotubes (SWCNT) can be assembled into various macroscopic architectures, most notably con- tinuous fibers and films, produced currently on a kilometer-per-day scale by floating catalyst chemical vapor deposition and spinning from an aerogel of CNTs. An attractive challenge is to produce continuous fibers with controlled molecular structure with respect to the diameter, chiral angle and ultimately (n,m) indices of the constituent SWCNT "molecules". This work presents an extensive Raman spectroscopy and high-resolution transmission electron microscopy study of SWCNT aerogels produced by the direct spinning method. By retaining the open structure of the SWCNT aerogel, we reveal the presence of both semiconducting and metallic SWCNTs and determine a full distribution of families of SWCNT grouped by optical transitions. The resulting distribution matches the chiral angle distribution obtained by electron microscopy and electron diffraction. The effect of SWCNT bundling on the Raman spectra, such as the G− line shape due to plasmons activated in the far-infrared and semiconductor quenching, are also discussed. By avoiding full aggregation of the aerogel and applying the methodology introduced, rapid screening of molecular features can be achieved in large samples, making this protocol a useful analysis tool for engineered SWCNT fibers and related systems. Keywords: SWCNT, chiral angle, fiber, metallic, semiconducting, Raman 1. Introduction The two main challenges to exploit the exceptional me- chanical, electrical, thermal and optical properties of car- bon nanotubes (CNT) on a macroscopic scale, are the syn- thesis of building blocks with molecular control, and their organized assembly into larger structures [1, 2]. Macro- scopic fibers of aligned CNT are a particularly attractive format that naturally maximizes the contribution from longitudinal properties of CNTs. Extensive efforts to im- prove synthetic methods and fiber spinning processes have led to continuous CNT fibers with tensile mechanical prop- erties on par with carbon fibers [3, 4, 5], thermal conduc- tivity above copper [6, 7] and mass-normalized electrical conductivity above that of metals [7, 8]. These proper- ties are encouraging, but still represent a small fraction of those of the constituent CNTs and thus leave large room for improvement. Key for the realization of CNT fibers with superior properties to those currently available is the control over the morphology of the constituent CNT in terms of num- ber of layers, diameter and chiral angle. These parame- ters affect the packing of CNTs in bundles and the stress transfer between them [9], hence the accepted view in the field that few-layer, large-diameter CNTs lead to superior tensile properties. Better packing of CNTs is also likely ∗Corresponding author Email address: [email protected] (Juan J. Vilatela ) to favour charge transfer, but fiber longitudinal electri- cal conductivity depends very strongly on the content of metallic versus semiconducting CNTs. Fibers of single- walled carbon nanotubes (SWCNT) of specific metallicity are yet to be produced, but work on transparent conduc- tors, for example, shows a conductivity enhancement of around 5.6 for samples of metallic SWCNTs compared to samples with unsorted SWCNTs [10]. A similar enhance- ment applied to existing CNT fibers currently made up of mixtures of SWCNTs would lead to a longitudinal conduc- tivity of around 40·106 S/m, or about 4 times higher than copper on a mass basis. There are two hypothetical routes towards producing a continuous, macroscopic fiber of single chirality SWC- NTs: wet spinning from a dispersion of pre-synthesized and sorted SWCNTs [11, 12], or by direct spinning of a fiber from an aerogel of chirallity-controlled SWCNTs synthesized by floating catalyst chemical vapor deposition (FCCVD). The direct FCCVD spinning method offers es- tablished control over the CNT number of layers [13, 14], diameter [15], and degree of alignment. But the deter- mination of further molecular features, particularly the fraction of metallic SWCNTs has proven elusive. This is partly a consequence of the fundamental difficulty in char- acterising individual molecular features (i.e. chiral angle, diameter, etc) in a system with around 108 SWCNTs per cross section and 1011 per metre of fiber in a highly ag- gregated format. Resonant Raman spectroscopy has been an ideal method to probe (relatively) large sample vol- umes while resolving fine molecular futures. Yet, studies Preprint submitted to Journal of LATEX Templates August 23, 2019 on SWCNT fibers produced by FCCVD typically extract limited information beyond identification of CNT number of layers and D/G ratio [16, 17, 18, 19]. An approximate chiral angle distribution can be obtained from extensive electron diffraction measurements on SWCNT bundles in a CNT fiber [20], but the method should be complemented by other techniques, especially those enabling faster char- acterization. In this work, we introduce a method for the screening of fine molecular features of CNT fibers made of SWCNTs by coupling high-resolution electron microscopy analysis and Raman spectroscopy mapping of individual SWCNT bundles of macroscopic aerogels. This strategy unveils a large abundance of previously masked semiconducting SWCNTs, and ultimately leads to the determination of the full distribution of (n,m) indices for the SWCNT fiber samples and a relative ratio of metallic to semiconducting tubes. It is intended to accelerate improvement of SWCNT fibers by enabling determination of SWCNT polydisper- sity (electronic and geometric) in continuous fibers, and ultimately assisting in the quest towards producing single chiral angle SWCNT fibers. 2. Experimental section 2.1. Sample preparation SWCNT macroscopic material was synthesized by di- rect spinning of a SWCNT aerogel produced by FCCVD, using butanol as carbon source, ferrocene as Fe catalyst and thiophene as sulfur catalyst promoter. Synthesis of predominantly SWCNTs was achieved by adjusting the precursor weight fraction at 99.1:0.8:0.1 [13]. Two sample formats were produced: SWCNT fiber and an aerogel film (Fig. 1(a-b)). The low-density aerogel films correspond to samples of SWCNT aerogel that were directly collected at the exit of the reactor. A SWCNT fiber is produced by densifying the free-standing SWCNT aerogel through capillary forces exerted by a volatile liquid applied at the exit of the FCCVD reactor. These two types of samples differ in their degree of aggregation, but they are otherwise chemically identical. 2.2. Characterization Morphological characterization of the aerogel film was carried out using a dual beam with field-emission scanning electron micro- scope (FEGSEM) Helios NanoLab 600i (FEI) at 10 keV. For extract- ing the SWCNT and SWCNT bundle diameter distributions, over 100 high-resolution transmission electron microscope (HRTEM) im- ages were obtained using a Talos F200X (FEI) TEM. SWCNT and bundle diameters were determined by image analysis of these mi- crographs using Image J software (examples and processing in ESI). Samples for TEM analysis were directly deposited onto TEM grids at the exit of the reactor, thus avoiding degradation by sonication or other undesired effects induced during dispersion in solvents. Ra- man spectroscopy was performed with a Renishaw PLC with 532 nm wavelength laser and a Bruker Senterra equipped with 532 nm, 633 nm and 785 nm waveleght lasers. The measurements of directly performed on SWCNT bundles were carried out with 100x/0.85 and 50x/0.75 Leica microscope objectives using a low power configura- tion in order to keep the laser focused and avoid heating effects. Analizing the SWCNT aerogel under these conditions enables the characterization of individual bundles (ESI). Indeed, the largest lat- eral distance between bundles in the film is around 1.5-2 µm, slightly larger that the calculated laser beam spot of 0.8-1 µm. However, not that all spectra collected when mapping the sample correspond to individual bundles (Examples of spectra for multiple bundles, as well as an example showing the loss of semiconductor features when using a large laser spot size are included in ESI). 3. Results and discussion 3.1. Exposing molecular features in SWCNT aerogels The strategy to study the molecular structure of SWCNT fibers consists in collecting a SWCNT aerogel at its exit from the FCCVD reactor (Fig.1(a)). By avoiding the densification into a SWCNT fiber, we retain the open structure of the aerogel, thus providing access to individualized bundles and other structural features that are inaccessible in a fully densified fiber. The resulting sample is similar to a transparent conductive film[21] (Fig.1(b)). The SWCNT bundles form a continuous network by virtue of the exceptionally long length of the constituent SWCNTs (1mm), but form an open structure whose projected area covers less than 35 % of the substrate (Fig. 1(c) and ESI). Figure 1: SWCNT aerogel samples used for molecular characteriza- tion and their correspondance to fibres (a) Schematic of the process to produce a densified fiber or an aerogel film by avoiding densifica- tion into a fiber. (b) Optical micrograph of a typical samples and (c) electron micrograph showing the the aerogel structure as a continous network of bundles . A general overview of the morphology of the bundle network and constituent SWCNTs is gained by analysis of the aerogel directly de- posited on a TEM grid. We obtain full diameter distributions of the bundles and individual SWCNTs from analysis of over 100 HRTEM micrographs. Some examples of micrographs are included in Figure 2. As expected, the SWCNT are of relatively large diameter (> 1 nm) (Fig.2(c)), at least compared to those produced by standard substrate-based chemical vapor deposition (CVD) in the absense of promotors, but not sufficiently large to collapse into ribbons. The 2 quency G− band in the SWCNT aerogel film, which originate from semiconducting nanotubes, to our knowledge not previously observed in SWCNT fibers. In contrast, the characteristic G band of the macroscopic SWCNT fiber consistently presents only the G− band related to the metallic nanotubes, as reported in similar SWCNT fibers [16, 18, 20] and related SWCNT bundle materials [17, 19]. Figure 3: Emergence of fine Raman features in SWCNTs aerogels compared with densified fibers. (a) Raman spectrum showing semi- conducting componets in the G band and well resolved RBM peaks, compared to (b) the spectrum for a densified fiber of the same com- position showing only a metallic G− band lineshape. 3.2. Resolving metallic and semiconducting SWCNT bun- dles A closer look at the G band lineshape provides further details in about the presence of semiconducting and metallic SWCNTs: semiconducting (SC) SWCNTs the G− band has Lorentzian line- shape, whereas in metallic (M) SWCNTs it has Breit-Wigner-Fano (BWF) lineshape. Fig.4(a) shows and example of a bundle that clearly exhibits intense semiconducting features with Lorentzian line- shapes at 1571, 1553, 1536 and 1522 cm−1, as well as contribution from metallic SWCNTs at 1585 cm−1 (more spectra showing SC- SWCNT contributions in ESI). This is indicative of a bundle of mainly SC-SWCNTs. Bundles of M-SWCNTs are also present in the SWCNT aerogel, as in Fig. 4(b), with a predominant BWF G− band caused by a gapless plasmon mode (0.2 eV) along the axial CNT direction, which is enhanced in nanotube bundles through the for- mation of a plasmon band [23] due to coulomb interactions between CNTs[24]. The plasmon band position depends on the SWCNT di- ameter [25], hence the relatively wide range observed for the SWCNT aerogel. In commercial samples of sorted M-SWCNTs the intensity ratio between G− BW F and G+ bands strongly increases with bundle /IG+ > 1 for bundles as thickness [26, 27], reaching values of I small as 6nm. The presence of semiconducting nanotubes in a bundle of predominantly M-SWCNTs produces a reduction of this ratio. In- deed, in the SWCNT aerogels analyzed in this work, we find that even /IG+ ≤ 0.5 bundles without Lorentzian G− peaks, the ratio I which is an indirect proof of the presence of SC-SWNCTs. But as a rapid screening method, the most direct evidence of SC-SWCNTs in the aerogel film is the presence of more than two peaks in the G band region: G+, collective metallic (BWF) G− and at least one semiconducting (Lorentzian) G−. Using this criteria, we obtain an abundance of metallic and semiconducting SWCNTs throughout the aerogel film, presented in Fig. 4(c). The abundance of semicon- ducting features in the aerogel contrasts with their apparent absence − BW F G − BW F G Figure 2: SWCNTs and bundles observed in aerogels produced by FCCVD: (a-b) Typical HRTEM images showing SWCNT and their association into bundles and (c-d) statistical analysis of SWCNTs and bundle diameters obtained from over 100 HRTEM images. bundle diameter distribution observed by TEM spans from 5-60 nm and peaks at around 20 nm (Fig.2(d)), which is comparable to the average bundle lateral size determined by small-angle X-ray scatter- ing (9 nm) [22]. A comparison of Raman spectra for the SWCNT aerogel film and SWCNT fiber (Fig. 3(a-b)) shows the presence of RBM peaks at the lower frequency range (100-300 cm−1) and the highly graphitic char- acter of the bundles manifested by a low ID/IG ratio of 0.06±0.02. But there are large differences between spectra. The most striking is the presence of the two well-resolved components of the low fre- 3 in the SWCNT fibers made up of the same building blocks. Over- all, our analysis of around 30 spectra indicates that semiconducting SWCNTs are present in 20% of the bundles probed. importantly, their correspondence with the G− lineshape of the bun- dle. We assign RBM modes to (n,m) indices by means of SWCNT family branches (2n+m=constant) and using the procedure devel- oped by Maultzsch and co-workers [28]. The method is based on the relationship between optical transitions and SWCNT diameter, rep- resented in the Kataura plot [29], taking into account that character- istic branches of families of SWCNTs show a downshift in transition energies arising from Van der Waals interactions. Figure 5 shows examples of the most common RBM peaks ob- served in the SWCNT aerogel film probed with three different energy lasers (785, 633 and 532 nm). Superimposing the Kataura plot en- ables identification of the SWCNT families (f ). Key for this step though, is the metallicity character observed in the G− lineshape, as it enables the accurate location of the spectra onto the Kataura plot. Only by analizing separated bundles (instead of the dense fiber) are SC-SWCNTs exposed and thus the family assignation can be made. Using a 785 nm laser, for example, there are strong RBMs at 140 and 240 cm−1; together with evidence of mixed BWF and Lorentzian G−, they can thus be assigned to transitions M11 of f36 and S22 of f25, respectively. Further assignation of RMBs from larger-diameter SWCNTs that are less separated in the Kataura plot can then pro- ceed. A summary of families observed are included in ESI. Figure 4: Raman G peaks for λ=532 nm obtained from individ- ualized bundles in the SWCNT aerogel film. (a) Well-resolved Lorentzian G− components together with metallic G− contribu- tion in a predominantly semiconducting bundle. (b) G− peak with BWF lineshape showing predominance of M-SWCNTs in the bundle probed. Figure 5: Kataura plot of transition energies vs inverse diameter for metallic (black) and semiconducting (red) SWCNTs and RBMs of a bundle for 785, 633 and 532 nm laser wavelengths with the indexed metallic or semiconducting branches (Mii or SMii) as well as the family assignation. 3.3. SWCNT assignation Retaining the bundle structure of SWCNT fibers by collapsing them into an open network enables an analysis of SWCNT chiral angle distribution through better resolved RBM peaks, and more Next, after identification of families on the Kataura plot, specific SWCNTs can be assigned to RBM peaks by identifying those whose optical transition fall within ±50 meV of the laser excitation energy, and are thus in resonance. For the spectra using a 785 nm laser, 4 for example, the most likely SWCNT of f25 with an RBMs at ∼236 cm−1 is the semiconducting (12,1) with a diameter of 0.98 nm. The assignation of all observed RBMs in the sample is included in ESI. Its accuracy is confirmed by the expected reciprocal dependence between RBM Raman shift and SWCNT diameter obtained experimentally (Figure 6), leading to the relationship ωRBM (cm−1) = (1) The values for the coefficients A= (214±5) cm−1·nm and B= (17±5) cm−1 are in the range reported for individualized, surfactant-wrapped SWCNTs in aqueous suspensions [28, 30], as well as for aligned SWC- NTS grown on a Si wafer and including both individualized and bun- dled tubes [31]. + 17 214 d Figure 6: Graphic representation of the experimental RBM Raman shifts and their corresponding assigned nanotube diameter. Overall, the results after full assignation show compelling evi- dence of the presence of semiconducting nanotube families together with the metallic ones, and which emerge in the open network aero- gel by avoiding masking of semiconducting features in the spectra by metallic SWCNTs. Analysed in this manner, the data from Raman spectroscopy is in full agreement with TEM observations. Figure 7 shows a map of (n,m) indices based on the nanotube diameter data extracted from the HRTEM images and the chiral angle distribution previously determined from electron diffraction measurements [20]. The families of SWCNTs determined by Raman from measurements on the open network and accessible with the three laser lines used in the present study, are also marked on the map. The agreement between them is clear, providing a consistent interpretation of the molecular structure of SWCNT fibers obtained using the two char- acterization techniques. Finally, we highlight some challenges for SWCNT molecular con- trol that stem from the large diameter and wide distribution observed in SWCNT fibers produced by FCCVD. The most immediate one is the increasing difficulty for (n,m) indices assignation due to the over- lap in possible CNT families when the measurements are conducted with standard 3-4 laser lines. Moreover, for the predominant diame- ters observed in these SWCNT aerogels (1.5 - 2 nm), the correspond- ing energy transitions are of third and fourth order, S33 and S44, and thus more likely to present energy deviations from Kataura plot due to larger delocalization of electrons in higher lying levels compared with lower order transitions [31]. This characterization challenge can be partially solved using a tunable laser Raman spectrometer to broaden the selection of incident energies to resolve the optical tran- sitions of a larger number of SWCNTs types. However the ultimate benefit of synthesizing a narrower distribution of smaller-diameter SWCNTs is the reduction in possible (n,m) indices. The number of geometrically possible chiral indices for the diameter range of 0.75 - 5 2.25 nm observed by HRTEM in the samples in this study, is ∼240. Even if the chiral angle distribution is biased towards armchair (20◦- 30◦) [20], this landscape appears too broad at present, as illustrated in the comparison with conventional substrate-grown SWCNTs in ESI. Future work aimed at molecular control will requires a chain of characterisation methods of increasing resolution and decreasing probe size, the first of which would be the Raman technique intro- duced here. We envisage the characterization protocol to follow this sequence: i) determination of SWCNT family distribution by Raman using 3-4 laser lines (resolution of ±50 meV); if the distribution is narrow, for example, less than ≈ 3 families observed per laser energy, then ii) perform HRTEM electron diffraction and diameter measure- ments (requiring resolving diameter and chiral angle differences of < 0.1 nm and 1◦ respectively) and iii) carry out absorption and emission measurements (example of optical absorption spectrum for an aerogel ESI). An interesting possibility is to combine these tools with high-throughput synthesis and analysis methods [32]. 4. Conclusions This work presents a study on the molecular structure of SWC- NTs in macroscopic fibers produced by FCCVD. By retaining the open structure of the aerogel directly produced during the synthe- sis stage, we produce samples consisting of a continuous open net- work of bundles sufficiently separated to be analyzed individually by HRTEM and Raman spectroscopy. Avoiding aggregation of bun- dles in the aerogel unveils the abundant presence of SC-SWCNTs in these fibers, manifested clearly by a Lorentzian lineshape of the low frequency G− bands. In the fully densified SWCNT fiber or other aggregated ensembles, these SC-SWCNTs are masked by M- SWCNTs, resulting in a BWF lineshape due to the formation of a plasmon band. Equipped with a clearly resolved G− band, a full map of SWCNT families and chiral indices can be produced from assignment of RBMs to optical transitions according to the Kataura plot. The empirical relationship found ωRBM =214/d+17 indicates good agreement in the assignation, and gives empirical coefficients in the range observed for SWCNTs with strong Van der Waals inter- action with the medium, as is expected for bundles. The SWCNTs observed with the three laser lines employed, is in agreement with the distribution obtained by combining HRTEM diameter measure- ments and chiral angle determination from electon diffraction. The method presented offers a tool to characterize the molecular structure of SWCNT fibers (in the aerogel state) probing much larger areas and much faster than alternative techniques. Future work is directed at exploiting it for rapid SWCNT fiber optimization, and more impor- tantly, to establish clear correlations between constituent SWCNTs and bulk properties. The open structure format would lend itself, for example, to transport measurements or surface potential charac- terisation methods that could resolve the conduction mechanism in bundles with different constituent SWCNTs. Acknowledgements The authors are grateful to Professor E. P´erez for access to Raman spectrometers, and acknowledge generous financial support from the European Union Seventh Framework Program under grant agreement 678565 (ERC-STEM) and from the Air Force Office of Scientific Research of the US (NANOYARN FA9550-18-1-7016). Appendix A. Supplementary data Supplementary data to this article can be found online at .... Figure 7: Distribution of chiral indices (n,m) for SWCNTs in aerogels produced by FCCVD, determined by combining HRTEM diameter measurements, chiral angle determination by electron diffraction, and confirmed by Raman spectroscopy measurements over large areas of SWCNT aerogel samples. References References [1] L. Liu, W. Ma, Z. Zhang, Macroscopic Carbon Nanotube As- semblies: Preparation, Properties, and Potential Applications, Small 7 (11) (2011) 1504 -- 1520. [2] R. Rao, C. L. Pint, A. E. Islam, R. S. Weatherup, S. Hof- mann, E. R. Meshot, F. Wu, C. Zhou, N. Dee, P. B. Amama, J. Carpena-Nunez, W. Shi, D. L. Plata, E. S. Penev, B. I. Yakobson, P. B. Balbuena, C. Bichara, D. N. Futaba, S. Noda, H. Shin, K. S. Kim, B. Simard, F. Mirri, M. Pasquali, F. For- nasiero, E. I. Kauppinen, M. Arnold, B. A. Cola, P. Niko- laev, S. Arepalli, H.-M. Cheng, D. N. Zakharov, E. A. Stach, J. Zhang, F. Wei, M. Terrones, D. B. Geohegan, B. Maruyama, S. Maruyama, Y. Li, W. W. Adams, A. J. Hart, Carbon nan- otubes and related nanomaterials: Critical advances and chal- lenges for synthesis toward mainstream commercial applica- tions, ACS Nano 12 (12) (2018) 11756 -- 11784. [3] K. Koziol, J. Vilatela, A. Moisala, M. Motta, P. Cunniff, M. Sen- nett, A. Windle, High-performance carbon nanotube fiber, Sci- ence 318 (5858) (2007) 1892 -- 1895. [4] H. Cho, H. Lee, E. Oh, S.-H. Lee, J. Park, H. J. Park, S.-B. Yoon, C.-H. Lee, G.-H. Kwak, W. J. Lee, J. Kim, J. E. Kim, K.-H. Lee, Hierarchical structure of carbon nanotube fibers, and the change of structure during densification by wet stretching, Carbon 136 (2018) 409 -- 416. [5] T. W. . Y. C. J. N. Wang, X. G. Luo, High-strength carbon nanotube fibre-like ribbon with high ductility and high electrical conductivity, Nature Communications 5 (2014) 3848. [6] T. S. Gspann, S. M. Juckes, J. F. Niven, M. B. Johnson, J. A. Elliott, M. A. White, A. H. Windle, High thermal conductivities of carbon nanotube films and micro-fibres and their dependence on morphology, Carbon 114 (2017) 160 -- 168. 6 [7] N. Behabtu, C. C. Young, D. E. Tsentalovich, O. Kleinerman, X. Wang, A. W. K. Ma, E. A. Bengio, R. F. ter Waarbeek, J. J. de Jong, R. E. Hoogerwerf, S. B. Fairchild, J. B. Fergu- son, B. Maruyama, J. Kono, Y. Talmon, Y. Cohen, M. J. Otto, M. Pasquali, Strong, light, multifunctional fibers of carbon nan- otubes with ultrahigh conductivity, Science 339 (6116) (2013) 182 -- 186. [8] R. J. Headrick, D. E. Tsentalovich, J. Berdegu´e, E. A. Ben- gio, L. Liberman, O. Kleinerman, M. S. Lucas, Y. Talmon, M. Pasquali, Structure -- property relations in carbon nanotube fibers by downscaling solution processing, Advanced Materials 30 (9) (2018) 1704482. [9] J. J. Vilatela, J. A. Elliott, A. H. Windle, A model for the strength of yarn-like carbon nanotube fibers, ACS Nano 5 (3) (2011) 1921 -- 1927. [10] A. A. Green, M. C. Hersam, Colored semitransparent conduc- tive coatings consisting of monodisperse metallic single-walled carbon nanotubes, Nano Letters 8 (5) (2008) 1417 -- 1422. [11] B. Vigolo, A. P´enicaud, C. Coulon, C. Sauder, R. Pailler, C. Journet, P. Bernier, P. Poulin, Macroscopic fibers and rib- bons of oriented carbon nanotubes, Science 290 (5495) (2000) 1331 -- 1334. [12] L. M. Ericson, H. Fan, H. Peng, V. A. Davis, W. Zhou, J. Sulpizio, Y. Wang, R. Booker, J. Vavro, C. Guthy, A. N. G. Parra-Vasquez, M. J. Kim, S. Ramesh, R. K. Saini, C. Kit- trell, G. Lavin, H. Schmidt, W. W. Adams, W. E. Billups, M. Pasquali, W.-F. Hwang, R. H. Hauge, J. E. Fischer, R. E. Smalley, Macroscopic, neat, single-walled carbon nanotube fibers, Science 305 (5689) (2004) 1447 -- 1450. [13] V. Reguero, B. Alem´an, B. Mas, J. J. Vilatela, Controlling Car- bon Nanotube Type in Macroscopic Fibers Synthesized by the Direct Spinning Process, Chemistry of Materials 26 (11) (2014) 3550 -- 3557. [14] S.-H. Lee, J. Park, H.-R. Kim, J. Lee, K.-H. Lee, Synthesis of M. Krein, J. Poleski, R. Barto, B. Maruyama, Autonomy in materials research: a case study in carbon nanotube growth, Npj Computational Materials 2 (2016) 16031. high-quality carbon nanotube fibers by controlling the effects of sulfur on the catalyst agglomeration during the direct spinning process, RSC Adv. 5 (2015) 41894 -- 41900. [15] B. Mas, B. Alem´an, I. Dopico, I. Martin-Bragado, T. Naranjo, E. M. P´erez, J. J. Vilatela, Group 16 elements control the syn- thesis of continuous fibers of carbon nanotubes, Carbon 101 (2016) 458 -- 464. [16] R. M. Sundaram, K. K. K. Koziol, A. H. Windle, Continu- ous direct spinning of fibers of single-walled carbon nanotubes with metallic chirality, Advanced Materials 23 (43) (2011) 5064 -- 5068. arXiv:https://onlinelibrary.wiley.com/doi/pdf/10. 1002/adma.201102754. [17] Q. Liu, W. Ren, Z.-G. Chen, D.-W. Wang, B. Liu, B. Yu, F. Li, H. Cong, H.-M. Cheng, Diameter-Selective Growth of Single- Walled Carbon Nanotubes with High Quality by Floating Cat- alyst Method, ACS Nano 2 (8) (2008) 1722 -- 1728. [18] J. S. Barnard, C. Paukner, K. K. Koziol, The role of carbon pre- cursor on carbon nanotube chirality in floating catalyst chemical vapour deposition, Nanoscale 8 (39) (2016) 17262 -- 17270. [19] E.-X. Ding, H. Jiang, Q. Zhang, Y. Tian, P. Laiho, A. Hussain, Y. Liao, N. Wei, E. I. Kauppinen, Highly conductive and trans- parent single-walled carbon nanotube thin films from ethanol by floating catalyst chemical vapor deposition, Nanoscale 9 (44) (2017) 17601 -- 17609. [20] B. Alem´an, M. M. Bernal, B. Mas, E. M. P´erez, V. Reguero, G. Xu, Y. Cui, J. J. Vilatela, Inherent predominance of high chi- ral angle metallic carbon nanotubes in continuous fibers grown from a molten catalyst, Nanoscale 8 (7) (2016) 4236 -- 4244. [21] L. Yu, C. Shearer, J. Shapter, Recent Development of Carbon Nanotube Transparent Conductive Films, Chemical Reviews 116 (22) (2016) 13413 -- 13453. [22] C. Santos, E. Senokos, J. C. Fernandez Toribio, A. Ridruejo, R. Marcilla, J. J. Vilatela, Pore structure and electrochemical properties of cnt based electrodes studied by in situ small wide angle x ray scattering, J. Mater. Chem. A (2019) 1. [23] K. Kempa, Gapless plasmons in carbon nanotubes and their interactions with phonons, Physical Review B 66 (19) (2002) 195406. [24] M. F. Lin, D. S. Chuu, pi-plasmons in carbon nanotube bundles, Physical Review B 57 (16) (1998) 10183 -- 10187. [25] A. Jorio, A. G. Souza Filho, G. Dresselhaus, M. S. Dresselhaus, A. K. Swan, M. S. Unlu, B. B. Goldberg, M. A. Pimenta, J. H. Hafner, C. M. Lieber, R. Saito, G-band resonant Raman study of 62 isolated single-wall carbon nanotubes, Physical Review B 65 (15) (2002) 155412. [26] C. Jiang, K. Kempa, J. Zhao, U. Schlecht, U. Kolb, T. Basch´e, M. Burghard, A. Mews, Strong enhancement of the Breit- Wigner-Fano Raman line in carbon nanotube bundles caused by plasmon band formation, Physical Review B 66 (16) (2002) 161404. [27] M. Paillet, P. Poncharal, A. Zahab, J.-L. Sauvajol, J. C. Meyer, S. Roth, Vanishing of the Breit-Wigner-Fano Component in In- dividual Single-Wall Carbon Nanotubes, Physical Review Let- ters 94 (23) (2005) 237401. [28] J. Maultzsch, H. Telg, S. Reich, C. Thomsen, Radial breathing mode of single-walled carbon nanotubes: Optical transition en- ergies and chiral-index assignment, Physical Review B 72 (20) (2005) 205438. [29] H. Kataura, Y. Kumazawa, Y. Maniwa, I. Umezu, S. Suzuki, Y. Ohtsuka, Y. Achiba, Optical properties of single-wall carbon nanotubes, Synthetic Metals 103 (1) (1999) 2555 -- 2558. [30] C. Fantini, A. Jorio, M. Souza, M. S. Strano, M. S. Dresselhaus, M. A. Pimenta, Optical Transition Energies for Carbon Nan- otubes from Resonant Raman Spectroscopy: Environment and Temperature Effects, Physical Review Letters 93 (14) (2004) 147406. [31] P. T. Araujo, S. K. Doorn, S. Kilina, S. Tretiak, E. Einarsson, S. Maruyama, H. Chacham, M. A. Pimenta, A. Jorio, Third and Fourth Optical Transitions in Semiconducting Carbon Nan- otubes, Physical Review Letters 98 (6) (2007) 067401. [32] P. Nikolaev, D. Hooper, F. Webber, R. Rao, K. Decker, 7
1805.05601	1	1805	2018-05-15T07:16:31	A distributed electrical model for superconducting nanowire single photon detectors	[ "physics.app-ph" ]	To analyze the switching dynamics and output performance of a superconducting nanowire single photon detector (SNSPD), the nanowire is usually modelled as an inductor in series with a time-varying resistor induced by absorption of a photon. Our recent experimental results show that, due to the effect of kinetic inductance, for a SNSPD made of a nanowire of sufficient length, its geometry length can be comparable to or even longer than the effective wavelength of frequencies contained in the output pulse. In other words, a superconducting nanowire can behave as a distributed transmission line so that the readout pulse depends on the photon detection location and the transmission line properties of the nanowire. Here, we develop a distributed model for a superconducting nanowire and apply it to simulate the output performance of a long nanowire designed into a coplanar waveguide. We compare this coplanar waveguide geometry to a conventional meander nanowire geometry. The simulation results agree well with our experimental observations. With this distributed model, we discussed the importance of microwave design of a nanowire and how impedance matching can affect the output pulse shape. We also discuss how the distributed model affects the growth and decay of the photon-triggered resistive hotspot.	physics.app-ph	physics	A distributed electrical model for superconducting nanowire single photon detectors Qing-Yuan Zhao1,2,, Daniel F. Santavicca3, Di Zhu2, Brian Noble3, and Karl K. Berggren2,† 1Research Institute of Superconductor Electronics (RISE), School of Electronic Science and Engineering, Nanjing University, Nanjing, Jiangsu 210093, China 2Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA 3Department of Physics, University of North Florida, Jacksonville, Florida 32224, USA [email protected] ; †[email protected] Abstract: To analyze the switching dynamics and output performance of a superconducting nanowire single photon detector (SNSPD), the nanowire is usually modelled as an inductor in series with a time-varying resistor induced by absorption of a photon. Our recent experimental results show that, due to the effect of kinetic inductance, for a SNSPD made of a nanowire of sufficient length, its geometry length can be comparable to or even longer than the effective wavelength of frequencies contained in the output pulse. In other words, a superconducting nanowire can behave as a distributed transmission line so that the readout pulse depends on the photon detection location and the transmission line properties of the nanowire. Here, we develop a distributed model for a superconducting nanowire and apply it to simulate the output performance of a long nanowire designed into a coplanar waveguide. We compare this coplanar waveguide geometry to a conventional meander nanowire geometry. The simulation results agree well with our experimental observations. With this distributed model, we discussed the importance of microwave design of a nanowire Page 1 of 14 and how impedance matching can affect the output pulse shape. We also discuss how the distributed model affects the growth and decay of the photon-triggered resistive hotspot. Article: Nanostructured superconducting wires are used as a platform for constructing single-photon detectors. A typical superconducting nanowire single-photon detector (SNSPD) is a single long nanowire meandered into a compact two-dimensional shape for efficiently collecting incident photons. Alternative nanowire architectures have been developed to extend the functions of a SNSPD, such as nanowires arranged in parallel for resolving photon numbers[1] or increasing detector output amplitude[2,3] and serial nanowires for photon number resolution[4]. By integrating multiple SNSPDs with on-chip multiplexing circuits, a SNSPD array can also realize spatial sensitivity[5,6]. To understand the detector performance of a particular nanowire geometry, it is important to have an electrical model of the superconducting nanowire. Conventionally, a superconducting nanowire is modelled as an inductor Lk in series with a time-dependent photon-induced resistor Rn, as shown in Figure 1a. A numerical electro-thermal model was developed to model the evolution of the resistance growth after a photon detection [7], and a similar phenomenological model was developed based on fitting to experimental observations[8]. These models can simulate the growth of Rn, the waveform of the output pulse, and the current recovery in the nanowire. Recently, a model has been developed that integrates the lumped-element model with the SPICE circuit simulation software for convenient implementation in complex circuit systems [9]. However, the lumped element model of a SNSPD is only valid when its geometric length is much shorter than the effective wavelengths in the nanowire. We recently found that effective wavelengths of microwave signals propagating in thin superconducting nanowires are reduced by about two orders of magnitudes from their free space values [10]. This reduction results because a thin superconducting nanowire has significantly higher kinetic inductance than its Faraday inductance. Therefore, a millimeter-long nanowire Page 2 of 14 would introduce ~100 ps propagation delay, which is no longer negligible. For determining the photon detection time, such variations of signal propagation adds additional timing jitter to a SNSPD[11]. These experimental observations suggest that, for nanowire lengths ≈1 mm or longer, it is more accurate to consider a highly inductive superconducting nanowire to be a distributed transmission line and model photon detection as a position-dependent phenomenon. Figure 1 (a) The lumped element model of a highly inductive superconducting nanowire, which is taken as an inductor Lk and a photon-induced resistor Rn in series. The same nanowire can also be modelled as a transmission line, as shown in (b). The total length is l. A photon absorption at x destroys the continuity of a superconducting nanowire transmission line and introduces a resistor Rn at x. (c) Simulated characteristic impedance and velocity for signal propagation along the superconducting transmission line. The simulated transmission line is a 200 μm long straight line. The simulated material stack is air (500 μm)/NbN nanowire (with a sheet inductance of 50 pH per square)/SiO2(300 nm)/Si(500 μm)/perfect conductor. Therefore, we report a distributed electrical model of a SNSPD and use this model to reproduce output pulses of a long superconducting nanowire patterned into a coplanar waveguide (CPW) structure. As shown in Fig. 1c, the superconducting nanowire is modeled as a transmission line, where a photon detected at Page 3 of 14 location x will trigger the corresponding superconducting nanowire into a resistor Rn. As the nanowire is pre-biased with a constant current, such sudden change of resistance will generate a positive pulse propagating from x to 0 and a negative pulse propagating from x to l. Compared to the total length of the nanowire, the length of the resistive domain is relatively short. Therefore, the electrothermal model we used for a lumped circuit is still valid for the distributed circuit model to calculate the growth of Rn, but the calculation of local current and voltage at x has to consider microwave propagations and reflections based on transmission line theory. In the distributed model, it is important to extract the characteristic impedance Zn (which determines the transmission and reflection coefficients) and the signal propagation velocity vn (which determines the propagation delay) of the equivalent transmission line from the geometry and material properties of a superconducting nanowire and its electromagnetic environment. For a lossless transmission line, the characteristic impedance is 𝑍n = √𝐿n/𝐶n and the phase velocity is 𝑣n = 1 √𝐿n𝐶n ⁄ (where 𝐿n and Cn are the inductance and capacitance per unit length, respectively). The kinetic inductance LK of the nanowire is much larger than its geometric inductance LG, resulting in 𝐿n = (𝐿K+𝐿G) 𝑙 ≅ 𝐿K 𝑙 , where l is the total length of the nanowire. Cn is the capacitance per unit length of the nanowire in its surrounding dielectric materials. We estimated Zn and vn by simulation using the Sonnet software package, defining a material of sheet inductance of Ln×w. The results are shown in Fig.1c. For a typical fiber-coupled SNSPD (100 nm wide nanowire, 50% fill factor, 10 μm × 10 μm area), the total length is 0.5 mm, which gives a total maximum delay of 64 ps if the nanowire is considered as a transmission line of vn = 0.026c0. This delay is close to the measured jitter of a SNSPD and thus is not easily observed with a single-ended readout [11]. For a longer nanowire, however, transmission line effects become more significant. In figure 2, we show the effect of the meander spacing on the propagation velocity on the nanowire. These simulations were performed using the AXIEM tool in the NI AWR Design Environment software platform. We varied the distance between adjacent lines in the meander (the fill Page 4 of 14 factor) and the total nanowire length. A straight wire with a length of 572 m was also simulated. The nanowire length was chosen to give approximately the same electrical length for each geometry. Figure 2. (a) Geometry of the nanowire meander (100 nm wide with an inductance of 50 pH/square) with 100 nm meander spacing embedded in the center conductor of a 50  CPW transmission line. Blue is the NbN and white is the sapphire substrate (300 μm thick). (b) Simulated phase velocity as a function of frequency for nanowire meanders with spacings of 100 nm (1.50 mm total length), 400 nm (1.17 mm total length), and 1.0 m (948 μm total length), along with a straight nanowire (572 μm total length). In all devices, the nanowire width is 100 nm. The nanowire length for each geometry was chosen to have a first-order self-resonance at approximately 5 GHz. We see that the phase velocity, and hence also the characteristic impedance, vary significantly as a function of frequency for the meander geometry. The more tightly packed the meander, the faster the phase velocity. For a given nanowire length, this effect results in a decreased ability to determine the photon absorption location based on a comparison of the timing of the output pulses from each end of the nanowire. The Page 5 of 14 periodic structure seen in the phase velocity as a function of frequency results from standing wave resonances along the length of the nanowire[10]; each peak in the propagation velocity corresponds to a half-wave resonance, with the first order (/2) resonance at approximately 5 GHz. The frequency-dependent phase velocity exhibited by the meander geometry results in pulse dispersion. To illustrate this, we show in figure 3a a comparison of the rising edge of output pulses due to photon-induced hotspots in three different locations along the nanowire, one at each end and a third in the middle. A clear time separation of the pulses is only observed at the very beginning of the rise, and this time separation is relatively small. In a real device, noise would likely prevent triggering at a sufficiently low voltage to capture this time separation. We conclude that accurately resolving the photon detection location based on pulse timing would not be practical for this nanowire geometry. This pulse dispersion will also result in the geometrical contribution to the measured device jitter varying based on the trigger threshold voltage. Figure 3b shows the initial pulse rise for a straight nanowire with a length of 0.5 mm, which corresponds to an equivalent electrical length to the longer meander. The structure in the pulse rise shown in figure 3b is due to the large impedance mismatch between the nanowire and the 50  readout. The pulse timing, and hence the location of photon absorption, can still be accurately determined from the initial sharp, well- defined portion of the pulse rising edge, which is about 50% of the total pulse height. An impedance taper could be used to impedance-match the higher-frequency pulse components, eliminating most of this structure from the rising edge [12]. Page 6 of 14 Figure 3 Simulation of rising edge of pulses measured at one end of the nanowire that are produced by a 6 k resistive hotspot of 300 ps duration. The sheet inductance is set to 50 pH/square and the bias current is 20 A. Pulses are simulated for hotspots at each end of the nanowire, as well as in the center. Pulses are passed through a 5 GHz low pass filter to match the bandwidth of a typical SNSPD readout. (a) Results for a 100 nm wide nanowire patterned in a meander with 100 nm meander spacing (50% fill factor) and total length of 1.5 mm. (b) Results for a straight 100 nm wide nanowire with a total length of 0.5 mm, which is approximately the same electrical length as the 1.5 mm long meander. Previous simulation results of a SNSPD suggest that if a clear dependence of detection pulses on the photon absorption locations is wanted, we should design the device with minimal dispersion and read it with an impedance matched readout. Such a device was recently demonstrated as a single-photon imager [12]. In this device, the nanowire was patterned into a CPW structure with a characteristic impedance of 1.0 kΩ and Page 7 of 14 a signal velocity of 5.56 μm/ps. The total length of the nanowire was 19.7 mm. This resulted in a maximum propagation delay of ≈ 3.2 ns, a significant delay that we can measure precisely with a 6 GHz bandwidth oscilloscope. To minimize impedance mismatch between the nanowire and the readout circuit, we designed the ends of the nanowire into Klopfenstein tapers to transform the impedance from 1.0 k to 50 Ω [13]. The taper had a passband starting from 700 MHz, ensuring the rising edge of the photon detection pulse would be efficiently transmitted. As a photon detection triggered two electrical pulses of opposite propagation directions toward to each end of the nanowire, we used a timing-differential readout to collect both of them. As the nanowire was designed into a CPW, dispersion was assumed to be negligible. Thus, a lossless transmission line model in SPICE can be used for representing the pulse propagation along the nanowire. As shown in Fig. 4a, we inserted a hotspot at position x in a transmission line that has a length of l. To simulate the dynamics of the hotspot resistor Rn, we used a SPICE model of a superconducting nanowire [9]. This model simplified the calculation of Rn with adequate accuracy for estimating the switching effects in an SNSPD. We simulated a distributed circuit of a conventional meandered SNSPD in SPICE, which gave similar shapes of the rising edges shown in Figure 3 but reduced the simulation time to a few seconds. To model the tapered nanowire for impedance matching, we divided the taper into 100 transmission lines of incrementally varying impedance and connected them in series. Page 8 of 14 Figure 4. Simulated output pulses from a nanowire designed into a CPW with an impedance matched readout. (a) Schematic diagram of the simulated circuit. Each end of the nanowire connects to a tapered nanowire for transforming the impedance. The readout circuit reads out the output pulses from the two ends of the nanowire, which are shown in (b) and (c). In (b) and (c), the blue traces are from experimental acquisition while the red traces are from the simulation. The firing location is at x = 8278 μm. The amplitudes of these pulses are normalized to the first maximum. As shown in Fig. 3b and c, the distributed model can reproduce similar pulse shapes to those seen experimentally. In particular, the pulses have similar reflection ripples on the falling edge. Because the impedance transform taper has a low-cutoff frequency of 700 MHz, a lumped description is still valid for low frequency signals. Thus, the pulse envelope follows an exponential decay. In the simulation, the total inductance of the nanowire was set to Lt = 3.2 μH, giving a decay constant of Lt/(2 × 50 Ω) = 32 ns. The more prominent features of the distributed nanowire model were exhibited at the first transmitted pulses, which are shown in Fig. 4. We simulated two different firing locations (x = 1668 μm and x = 8278 μm) and compared the output traces to our experimental observations. Both the simulation pulses and experimental pulses showed that the rising edges of the output pulses were well preserved so that the pulse arrival times were linearly proportional to x/v. Although the nanowire had a total inductance as large as 3.2 μH, the rising time of the edge was about 100 ps for the simulation pulses and ~250 ps for the experimental pulses. It indicates that the rising edge of the pulses from the distributed detector reflected the transition time from a superconducting state to a normal resistor, but not the overall lifetime of the normal resistor. The simulated pulses succeeded in producing ripples of large amplitude. However, some small ripples did not match to our experimental results, which was probably because the distributed model did not include impedance mismatch from bonding wires and connections and other parasitic parameters. Page 9 of 14 Figure 5. Details of the simulated (red) and experimental (blue) pulses at two firing locations. (a) and (b) are the two output pulses for x = 1668 μm, while (c) and (d) are the two output pulses for x = 8278 μm. (e) Simulated growth of the normal resistance Rn for the two firing cases (dashed line for x = 1668 μm and solid line for x = 8278 μm). The distributed model indicates that the growth of the normal region Rn depends on the photon detection location. As shown in Fig. 5e, when a photon was detected in the middle (x = 8278 μm), Rn reached a maximum value of 17 kΩ within a lifetime of 4.1 ns. For the detection event at the end close to the readout amplifier (x = 1668 μm), Rn increased to a maximum value of 13 kΩ after oscillating a few times. The lifetime of the normal domain increased to 6.7 ns. The difference of growth of Rn can be explained by analyzing the pulse reflections in the distributed model. If the photon detection happened at the middle of the nanowire, the hotspot kept growing until the two reflected pulses returned to the detection region. Because the nanowire impedance was larger than the load impedance, the pulses flipped into reversed pulses (in direction and amplitude) and reflected back to Rn. Since x was at the middle of the nanowire, these reflected pulses arrived at Rn at almost the same time. This event resulted in a sudden reduction of the local current through Rn to below the level for sustaining a normal domain. Consequently, the normal domain was reset to the superconducting state. For the case where x = 1668 μm, the pulse propagating to the near end reflected back to the normal domain earlier than the other pulse propagating to the far end. The Page 10 of 14 first reflected pulse decreased the local current to reduce Rn, but this reduction was not enough to bring Rn to zero. A portion of the pulse then returned to the near-end, repeating the process. The red curve in Fig. 5c shows two dips, corresponding to the reduction of current causing by two reflected pulses. Finally, the original detection pulse propagating to the far-end of nanowire reflected to Rn, adding another reduction of the local current, the normal domain can reset to the superconducting state. In conclusion, we created a distributed model for a superconducting nanowire and applied this model to study the photon response pulses from a superconducting nanowire with different geometries and readout schemes. Our simulation results suggest that, to observe a clear position dependence of the output pulses on photon detection locations, the microwave design of the nanowire should be considered. The distributed model also showed that the growth of the hotspot resistance depended on its locations, which could be explained by considering pulse propagations and reflections. With this model, we can design a distributed nanowire detector and use the microwave readout to extract additional detection information, e.g. to map the photon detection locations from the pulse arrival times[12] or to resolve multi-photon detection events in a serial detector array[14]. The distributed model can also be used in frequency multiplexed detector arrays to simulate the transient pulses[15]. We envision this distributed model will be a useful electrical tool for developing new nanowire-based devices and microwave circuits. Acknowledgments This research was supported by the National Science Foundation (NSF) grants under contact No. ECCS-1509486 (MIT) and No. ECCS-1509253 (UNF), and the Air Force Office of Scientific Research (AFOSR) grant under contract No. FA9550-14-1-0052. Di Zhu is supported by National Science Scholarship from A*STAR, Singapore. Qing-Yuan Zhao is partially supported by the Fundamental Research Funds for the Central Universities No.021014380100. Page 11 of 14 References [1] Divochiy A, Marsili F, Bitauld D, Gaggero A, Leoni R, Mattioli F, Korneev A, Seleznev V, Kaurova N, Minaeva O, Gol'tsman G, Lagoudakis K G, Benkahoul M, Lévy F and Fiore A 2008 Superconducting nanowire photon-number-resolving detector at telecommunication wavelengths Nat. Photonics 2 302–6 [2] Marsili F, Najafi F, Dauler E, Bellei F, Hu X, Csete M, Molnar R J and Berggren K K 2011 Single-photon detectors based on ultranarrow superconducting nanowires Nano Lett. 11 2048–53 [3] Zhao Q, McCaughan A N, Dane A E, Najafi F, Bellei F, De Fazio D, Sunter K a., Ivry Y and Berggren K K 2014 Eight-fold signal amplification of a superconducting nanowire single-photon detector using a multiple-avalanche architecture Opt. Express 22 24574 [4] Jahanmirinejad S, Frucci G, Mattioli F, Sahin D, Gaggero a., Leoni R and Fiore a. 2012 Photon-number resolving detector based on a series array of superconducting nanowires Appl. Phys. Lett. 101 72602 [5] Zhao Q, McCaughan A, Bellei F, Najafi F, De Fazio D, Dane A, Ivry Y and Berggren K K 2013 Superconducting-nanowire single-photon-detector linear array Appl. Phys. Lett. 103 [6] Verma V B, Horansky R, Marsili F, Stern J A, Shaw M D, Lita A E, Mirin R P and Nam S W 2014 A four-pixel single-photon pulse-position array fabricated from Page 12 of 14 WSi superconducting nanowire single-photon detectors Appl. Phys. Lett. 104 51115 [7] Yang J K W, Kerman A J, Dauler E A, Anant V, Rosfjord K M and Berggren K K 2007 Modeling the Electrical and Thermal Response of Superconducting Nanowire Single-Photon Detectors IEEE Trans. Appl. Supercond. 17 581–5 [8] Kerman A J, Yang J K W, Molnar R J, Dauler E A and Berggren K K 2009 Electrothermal feedback in superconducting nanowire single-photon detectors Phys. Rev. B 79 100509 [9] Berggren K K, Zhao Q-Y, Abebe N S, Chen M, Ravindran P, McCaughan A N and Bardin J 2018 A Superconducting Nanowire can be Modeled by Using SPICE Supercond. Sci. Technol. 31 055010 [10] Santavicca D F, Adams J K, Grant L E, McCaughan A N and Berggren K K 2016 Microwave dynamics of high aspect ratio superconducting nanowires studied using self-resonance J. Appl. Phys. 119 234302 [11] Calandri N, Zhao Q Y, Zhu D, Dane A and Berggren K K 2016 Superconducting nanowire detector jitter limited by detector geometry Appl. Phys. Lett. 109 0–8 [12] Zhao Q-Y, Zhu D, Calandri N, Dane A E, McCaughan A N, Bellei F, Wang H-Z, Santavicca D F and Berggren K K 2017 Single-photon imager based on a superconducting nanowire delay line Nat. Photonics 11 247–51 Page 13 of 14 [13] Klopfenstein R 1956 A Transmission Line Taper of Improved Design Proc. IRE 44 31–5 [14] Zhu D, Zhao Q-Y, Choi H, Lu T-J, Dane A E, Englund D R and Berggren K K 2017 A scalable multi-photon coincidence detector based on superconducting nanowires arXiv:1711.10546 [15] Doerner S, Kuzmin A, Wuensch S, Charaev I, Boes F, Zwick T and Siegel M 2017 Frequency-multiplexed bias and readout of a 16-pixel superconducting nanowire single-photon detector array Appl. Phys. Lett. 111 1–5 Page 14 of 14
1808.06589	1	1808	2018-06-30T10:00:23	Modelling a conductive-capacitive medium using the boundary element method	[ "physics.app-ph", "physics.comp-ph" ]	In this paper a generalized fundamental solution using the boundary element method to solve the Helmholtz equation is proposed. It is observed that the commonly used fundamental solution is only valid for good conductors since the capacitive effect of the considered medium is always neglected. By the use of the well-known Lorentz gauge condition a fundamental solution which incorporates the phase as well as the attenuation transmission coefficients is derived by the authors. Next, a model is developed using this proposed fundamental solution for modelling a coating layer of a buried pipeline. Subsequently, a model of a buried coated pipeline in close proximity to a high voltage power line is developed and numerically implemented. Finally, the model is used to simulate two configurations in order the verify the proposed general fundamental solution of the boundary element method for electromagnetic field problems. Hereby, its validity is proven and it is shown that the generalized solution and related model can be used for industrial applications.	physics.app-ph	physics	Modelling a conductive-capacitive medium using the boundary element method Bram Schoonjans, Johan Deconinck Vrije Universiteit Brussel, Faculteit Ingenieurswetenschappen, Elektrotechniek en Energietechniek, Pleinlaan 2, 1050 Brussel, Belgium, e-mail address: [email protected] Abstract In this paper a generalized fundamental solution using the boundary element method to solve the Helmholtz equation is proposed. It is observed that the commonly used fundamental solution is only valid for good conductors since the capacitive effect of the considered medium is always neglected. By the use of the well-known Lorentz gauge condition a fundamental so- lution which incorporates the phase as well as the attenuation transmission coefficients is derived by the authors. Next, a model is developed using this proposed fundamental solution for modelling a coating layer of a buried pipeline. Subsequently, a model of a buried coated pipeline in close proximity to a high voltage power line is developed and numerically implemented. Finally, the model is used to simulate two configurations in order the verify the proposed general fundamental solution of the boundary element method for electromagnetic field problems. Hereby, its validity is proven and it is shown that the generalized solution and related model can be used for industrial applications. Keywords: boundary element method, fundamental solutions, electromagnetics, pipelines, coating layer Introduction In order to transport natural gas or oil, steel pipes are used. These pipes are constructed in order to supply resources at different locations located at Preprint submitted to ArXiv.org, August 21, 2018 great distances apart from each other. To avoid visual inconvenience, these pipelines are often buried at specific locations. Besides these pipelines, there is also the necessity to transport electrical energy over long distances and to supply at the same locations. Due to the scarcity of land in the developed world, there is the tendency to allocate the same utility to these means of energy transport. As a result, pipelines and power lines come into close proximity and an electromagnetic coupling occurs between both. The electromagnetic field originating from the power line induces a voltage on the buried pipeline which leads to risks for human safety and structural damage. Since the calculation of the level of induced voltages is dependent of a multitude of parameters, it is obvious an analytical solution of a real life situation is excluded. Therefore, there is the need to model the problem using numerical calculations methods. In 1926, Carson [Carson (1926)] was the first who modelled electromag- netic field problems involving overhead powerlines. Using equivalent circuits this lead to his well-known formulae. In 1949 Sunde extended the model towards a multilayered earth [Sunde (1966)]. Later on simulation software was developed using a transmission line model (TLM), like ECCAPP [Dawalibi (1986, 1989)], that was further improved by Bortels in [Bortels (2006)]. Munteanu [Munteanu (2012)] however didn't use a TLM to tackle to model, but developed a method that combined a finite element method (FEM) with a boundary element method (BEM). This model consists of a one dimensional discretization of the pipeconductor using FEM and whereby the surrounding earth is modelled using a three dimensional BEM. In fact this was an extension of the DC model of Brichau [Brichau (1994)] to AC. However, it was observed that a hiatus was present in the model, namely the way of dealing with unbounded regions, in this case the soil surrounding the pipeline. In such a situation, the BEM requires a specific solving method, namely to calculate the geometry dependent coefficient ci in the general ex- 2 pression in an indirect way. This problem was solved by the authors through the proposal of an innovative yet general method based on the skin depth [Schoonjans (2017)]. Next, without anyhow devaluating the excellent work performed by Munteanu, another limitation of his model is the absence of a coating layer. Obviously, this is a drawback when real life situations need to be investigated. Since a coating layer cannot be perceived as a good conductor and because the fun- damental solutions described in the literature are only valid when modelling conductive media, a more general expression of the fundamental solution for the BEM when modelling high-voltage low frequency electromagnetic field problems is derived in this paper. This general expression, which can be approximated for either good conductors or for dielectric materials, is used to model the coating layer. In the first part a briefly overview is given of electromagnetic waves in order to define the attenuation as well as the phase coefficients. In the second part these coefficients are used to derive the most general fundamental solu- tion for the BEM needed to model the coating layer. In the following part the model is discussed and in the final part of this paper simulation results are discussed and compared with those of a TLM. 1. Description of the electromagnetic waves 1.1. Wave equations In homogeneous, linear and isotropic media, electromagnetic waves can be described in a direct way by the electric field (cid:126)E and the magnetic field (cid:126)B [Griffiths (1999)] ∆ (cid:126)E − µσ ∆ (cid:126)B − µσ ∂ (cid:126)E ∂t ∂ (cid:126)B ∂t − µ − µ ∂2 (cid:126)E ∂t2 = 0, ∂2 (cid:126)B ∂t2 = 0, (1a) (1b) where µ, σ and represents respectively the permeability, conductivity and permittivity of the medium wherein the wave propagates. As generally known, equations (2a) and (2b) admit plane-wave solutions that are expressed by [Reitz (1979)] (cid:126)E((cid:126)r, t) = E0e−i(k(cid:126)r−ωt), (2a) 3 (2b) in which E0 and B0 are functions of the applied boundary conditions and k is the complex wave number. The complex wave number equals (cid:126)B((cid:126)r, t) = B0e−i(k(cid:126)r−ωt), k = α + iβ, (3) where the attenuation transmission coefficient α and the phase transmission coefficient β are [Stratton (2007)] (cid:32)(cid:114) (cid:32)(cid:114) (cid:34) (cid:34) µ 2 µ 2 α ≡ ω β ≡ ω (cid:33)(cid:35)1/2 (cid:33)(cid:35)1/2 (cid:16) σ (cid:16) σ ω + 1 (cid:17)2 (cid:17)2 − 1 1 + 1 + ω , . (4a) (4b) 1.2. Potential functions In numerical electromagnetics it is more common to use potential functions in order to execute the computations rather than to calculate the electric field (cid:126)E and the magnetic field (cid:126)B in a direct way. Therefore the magnetic vector potential (cid:126)A and the electric scalar potential Φ are introduced and defined as (cid:126)B = ∇ × (cid:126)A, (cid:126)E = −∇Φ − ∂ (cid:126)A ∂t . (5a) (5b) Field equations are invariant to gauge transformations, allowing prescribing any convenient gauge condition. The aim is to model a coating layer of a buried pipeline. For this medium, the capacitive characteristics cannot be neglected in comparison to the conductive properties. Therefore, the Lorentz gauge condition used in the following form ∇ · (cid:126)A = −µ (σ + iω) Φ. (6) Using this condition, Maxwell's equations can be reduced to the following Helmholtz equations that describe electromagnetic field problems without loss of generality, (7a) ∆ (cid:126)A − λ2 (cid:126)A = 0, 4 The wavenumber equals ∆Φ − λ2Φ = 0. λ =(cid:112)−iωµ (σ + iω) (7b) (8) and characterizes the medium where the Helmholtz equation holds. 2. Fundamental solutions in the BEM If a Helmholtz equation of an arbitrary function f in a region Ω with boundary Γ as a governing equation is considered, the BEM can be used to calculate any function value of any arbitrary point pi within the region by use of the following general expression of the method GdΓ, (9) with G the so-called fundamental solution or Green function of the governing equation and n the unit outward normal to the boundary Γ [Brebbia (1989)]. cif (pi) + ∂G ∂n ∂f ∂n Γ dΓ = f Γ (cid:90) (cid:90) Generally, the fundamental solution listed in BEM reference books like [Brebbia (1989); Shen (1995); Sadiku (2001); Sykulski (1995)] related to electromagnetic field problems neglects the capacitive effect of the modelled medium. In order to take this effect into account, the fundamental solution takes the form G1 = −β(cid:126)r(1+i α β ) e 4π(cid:126)r , (10) with again α and β the coefficients in equations (4a-b). Note that if the capacitive effect is neglected - i.e. a good conductor is considered - it is stated that [Balanis (2012)] (cid:18) σp ωp (cid:19)2 (cid:29) 1, (cid:114)ωµσ 2 5 thus the attenuation as well as the phase transmission coefficients converges to the following value [Balanis (2012)] α (cid:39) β (cid:39) , (12) (11) which leads to the well-known fundamental solution for good conductors, G2 = e−β(cid:126)r(1+i) 4π(cid:126)r . (13) If dielectric materials are considered the attenuation transmission coefficient converges to [Balanis (2012)], (14) (15) (16) α (cid:39) ω √ µ, and the phase transmission coefficient to [Balanis (2012)], (cid:114)µ . β (cid:39) σ 2 The fundamental solution for a dielectric material is therefore, −(cid:126)r( 1 e 2 G3 = √ µσ 4π(cid:126)r +iω √ µ) . 3. Combined BEM-FEM-model for coated buried pipelines In what follows the model described by [Munteanu (2012)] is completed by incorporating the coating layer between the pipeline and earth that is always present. The buried pipeline is modelled using the FEM, while the coating layer is modelled using the BEM by making use of the general fundamental solution in (10). The surrounding earth is modelled using the BEM whereby the innovative method as described in [Schoonjans (2017)] is implemented. Each medium is, in general, characterized by its corresponding permeability µ, conductivity σ and permittivity , see figure 1. However, for the steel pipe as well as for the surrounding soil p respectively s can be neglected. Due to the inductive coupling between buried pipelines and neighbouring high voltage AC power lines, a voltage Φ is induced inside the steel medium. 6 Because of the high conductivity of this medium it is assumed that Φ is radial independent, meaning that the potential on the inside of the coating layer over the whole boundary equals Φ. Through the coating layer a radial current density Jρ will flow because of the voltage difference between the potential Φ in the steel pipe and the potential V of the surrounding earth. The potential on the outside of the coating layer is defined as V and a current density from the coating layer to the surrounding earth will flow as well and is called Qρ. These different physical quantities are schematically presented in figure 2. Figure 1: Electromagnetic characterization of the three media. Figure 2: Voltages and current densities in the different media. 3.1. The steel pipe modelled using the FEM In the steel pipe the material properties are noted with the subscript p. Since steel pipes are good conductors, condition (11) holds and the 7 δcμc, σc, εcμp, σp, εpearth surfaceRpRinμe, σe, εeearthcoatingsteel pipeinner pipeQρΦJρVwavenumber in (7b) reduces to λ =(cid:112)−iωµpσp and the Helmholtz equation that holds in the steel pipe is ∆Φ − iωµpσpΦ = 0. (17) This so-called internal Helmholtz equation is solved using an axisymmetric coordinate system that is always aligned with the centre of the pipeline. Since the dimensions in axial z-direction are orders of magnitude higher compared to those in the radial θ-direction any variance of the potential in the θ-direction is neglected. The Helmholtz equation (17) in an axiymmetric coordinate system and neglecting the θ-direction is (cid:18) (cid:19) ∂2Φ ∂z2 + 1 ρ ∂ ∂ρ ρ Φ ∂ρ = iωµpσpΦ. (18) In the gauge condition (6) the capacitive effect is neglected and substituted into (18). The magnetic vector potential (cid:126)A originates from the high voltage power line - see section 3.3 - and since the wall thickness of the pipe is negligible compared to its distance to the power line any variation of the radial component of (cid:126)A in that direction will also not be taken into account. Equation (18) becomes ∂2Φ ∂z2 + 1 ρ ∂ ∂ρ (cid:19) (cid:18) ρ Φ ∂ρ = −iω ∂Az ∂z . (19) The latter equation is reduced to an equation along only the z-direction by integrating it over the pipe surface, (cid:18) (cid:19) (cid:19) (cid:90) 2π (cid:90) Rp 0 Rin (cid:18) ∂2Φ σp ∂z2 + 1 ρ ∂ ∂ρ ρ Φ ∂ρ + iω ∂Az ∂z ρdθdρ = 0, (20) and assuming that no radial current flows into the inner side of the pipeline, the governing equation for the inner problem becomes ∂2Φ ∂z2 − 2πRp σpSp Jρ + iω ∂Az ∂z = 0, 8 (21) wherein Jρ = σp ∂Φ ∂ρ Rp becomes a source term. The standard FEM with linear elements is used to discretize (21) yield- ing for each element of length (cid:96) with nodes i and j the following set of equations:  (cid:19) (cid:19) + Jρj 6 + Jρj 3 3 (cid:18) Jρi (cid:19) (cid:18) Azi + Azj (cid:18) Jρi (cid:19) (cid:18) Azi + Azj 2 , 6 , 2 Ii = (Φi − Φj) · SP + k · −Ij = (Φj − Φi) · SP + k · − jω(cid:96) · SP · + jω(cid:96) · SP · (22) (23) with I the current in the axial direction, k = 2πRp(cid:96) the surface of revolution per element and SP = σpSp/(cid:96) the axial conductivity. In this approach it is assumed that Jρ varies linearly within each element between the values Jρi and Jρj . The same reasoning holds for Az. Writing these two equations for all elements yields the system of equations representing the internal potential problem in the pipes noted with superscript p using the FEM: (cid:2)Sp K p.(cid:3)(cid:26) Φ (cid:27) J ρ =(cid:2)Ap(cid:3)(cid:8) Az (cid:9) . In this system of equations the axial component of (cid:126)A is an imposed source term but it is obvious that since two unknowns remain, this system of equations still needs to be extended. 3.2. Modelling the coating layer and the earth using the BEM Two boundaries are distinguished, namely that of the steel pipe and that of the coating layer. The coating layer, bounded by the steel pipe and surrounding earth, as well as the surrounding earth, are modelled using the BEM. 9 3.2.1. The coating layer A purpose of the coating layer is to protect the pipeline against corrosion. Hence, it is obvious that a coating layer has a very high electrical resistance compared to the steel pipeline and the surrounding earth. Thus, one could argue to introduce the coating layer in the model as an impedance with a high ohmic resistance between the soil and the steel pipe. This approach however does not work properly since it only takes into account the voltage drop over the coating layer seen at each node. This was observed after numerous calculations. Note that the strength of the BEM is that the influence of each neigh- bouring node is taken into account. Therefore the coating layer is modelled using the BEM. A solution with the well-known fundamental solution (13) with a high resistance leads to numerical errors. Again, also this observation was made after numerous computations. Thus, to be able to use the BEM for modelling conductive-capacitive media, the proposed general expression of the fundamental solution (10) needs to be used. This fundamental solution may be approximated when dealing with dielectric materials, see (16), yet due to the specific characteristics of a coating layer, it cannot be perceived as a dielectric material for the following reasons. In general, a coating layer is, from an electromagnetic point of view, char- acterized by its permeability µ, conductivity σ and permittivity . These characteristics combined with the angular frequency ω can be used to calcu- late the attenuation transmission coefficient α and the phase transmission coefficient β, see (4a) respectively (4b). For a polyethylene coating layer with the characteristics listed in Table 1, the attenuation transmission coefficient α and the phase transmission coefficient β becomes 2.36957 × 10−6 Np/m respectively 3.33211× 10−6 rad/m. When on the one hand the approximation formula for good conductors is used, see (12), a relative error of 18.6% respec- tively 15.7% is made. On the other hand, when the approximation formulas for good dielectrics are used, see (14) and (15), the relative errors are 42.2% respectively 29.7%. Thus, the coating layer is modelled by use of the most general form of the fundamental solution G1, see (10). The subscript c is used to distinguish the material properties. The coating layer around the steel pipe is discretized 10 Table 1: Characteristics of polyethylene coating layer [CIGRE (1995)]. Thickness of the coating δc Relative permittivity of the coating r Relative permeability of the coating µr Resistance of the coating ρc Angular frequency ω 4mm 5 5 25 × 106 Ωm 100π rad/s in the z direction in elements with length (cid:96) that coincide with the elements used to discretize the steel pipe. It is assumed that the current density Jρ that enters or leaves the coating at the steel surface as well as the current density Qρ that enters or leaves the coating at the earth surface are only varying along the z-direction and in a linear dependency of z. These assumptions are complying with linear pipe elements, see [Munteanu (2012)], and the FEM discretization of the pipe. Further, for the potentials Φ and V the same assumptions are made. Using the BEM, one has to integrate in each element over the outer surfaces of the coating: once at the outer radius of the pipeline, and once at the outer radius plus the coating thickness δc, see figure 3. Figure 3: Modelling the coating layer using boundary elements. As a result, if N elements are implemented to model the coating layer, the following BEM matrix system of dimension 2N × 2N is obtained (cid:2)H c(cid:3)(cid:26) Φ V (cid:27) = −(cid:2)Gc/(σc + iωc)(cid:3)(cid:26) J ρ (cid:27) Q ρ , (24) with H c and Gc being the classical but complex BEM H and G matrices for 11 z,ξ1φj−1,Vj−1φj,Vjδc−1'1the coating, hence the superscript c. Furthermore, it is obvious that the coating layer is a closed system: the incoming current density summed over the whole inner surface equals the total outcoming current density over the exterior one, or (cid:90) (cid:88) (cid:90) (cid:88) JρidS + QρidS = 0. (25) elements Selint elements Selext with again N the number of elements. Observe that this is not necessarily true within each element. As the coating is a closed system, the diagonal terms of the H c-matrix may be calculated as follows, ii = − 2N(cid:88) H c H c ij. (26) j=1 3.2.2. The surrounding earth The sub- and superscript s annotates the material properties of the soil. In general the capacitive effect of the soil surrounding the buried pipeline is not taken into account since it can be considered to be a good conductor [Gary (1985)]. Therefore the fundamental solution, G2, see (13), is used. The function f in the general expression of the BEM, see (9), is now the potential V on the outside of the coating layer. Combined with (5b), the Lorentz gauge condition for good conductors and the constitutive equation of conductivity [Munteanu (2012)], one has, (cid:90) ciVi + ∂G2 ∂n V dΓ = Γ (cid:90) (cid:18) Qρ σs Γ (cid:19) + iωAρ G2dΓ, (27) with σs the conductivity of the surrounding soil and Aρ the radial component of the magnetic vector potential. The equation describing the influence of the surrounding earth becomes (cid:2)H s −Gs/σs (cid:27) (cid:3)(cid:26) V Q ρ =(cid:2)iωGs(cid:3)(cid:8) Aρ (cid:9) . (28) 12 The effect of the discontinuity between the air and the ground is taken into account using the principle of the mirror technique as discussed in [Munteanu (2012)]. ii = − N(cid:88) H s Following the same reasoning as in [Schoonjans (2017)] the diagonal terms in the latter equation are calculated as follows ij + e−2π (1 + 2π (1 + i)) . H s (29) 3.3. The magnetic vector potential j=1 The applied source conditions in equations (23), (20) and (23) are the z and ρ components of the magnetic vector potential (cid:126)A originating from the high voltage power line. The general formula to compute (cid:126)A is [Shen (1995)] ∆ (cid:126)A − iωµσ (cid:126)A = −µ (cid:126)Js, (30) whereby (cid:126)Js is the current density of one phase of the power line. Since the current source is located in the air and the observation point is in the soil, the modified image theory needs to be applied, see [L.D. (1996)]. In the case of power line frequencies (50 and 60 Hz), this implies that the modified current source is doubled compared to the initial current source [L.D. (1996)]. Thus if the conductor is considered to be filiform and if again the fundamental solution (13) is used, for a 3 phase system the total magnetic vector potential at distance (cid:126)rf per phase, seen from the power line to the buried pipeline, becomes (cid:90) 3(cid:88) f =1 (cid:126)A = µ 2π e−β(cid:126)rf(1+i) 4π(cid:126)rf (cid:126)1s,f dΓ, (31) I s,f C where the vector (cid:126)1s,f is the unit vector in the direction of the power line per phase and I s,f is the phase current and normally shifted 120 degrees per phase. Using (31) the contribution of each point of the power line to the buried pipeline of the magnetic vector potential can be computed. This is executed 13 by discretization of the power line into M elements of length (cid:96) whereby each element contributes to the total vector potential in each node of the discretized buried pipeline. Note that in the general case the number of elements for discretization of the power line does not need to be equal to the number of elements for the discretization of the buried pipeline nor these elements need to be equidistant. The complex magnetic vector potential (cid:126)A computed in each point through equation (31) is further split per element j into its axial Azj respectively radial Aρj component by taking into account the angle between the pipeline and the power line. If (cid:126)1pj is the direction vector of the pipeline in element j and (cid:126)1HVi is that of the i-th element of the power line from which the contribution is calculated, the partition is [Bortels (2006)] (cid:17) Azj = (cid:126)A(cid:16)(cid:126)1pj · (cid:126)1HVi (cid:113) (cid:126)A2 − A2 Aρj = , zj , (32a) (32b) Hence, successively computing (31), (32a) and (32b) considering the 3 phases yields the applied source conditions of the total system matrix. 3.4. The total system matrix Combining (23), (20) and (23) yields the total system matrix Sp K p H c 0 0 H s Φ J ρ V Q ρ     Az 0 Aρ   . 0 0 Gc/(σc + iωc) Ap 0 Gs/σs 0 0 iωGs = 14 (33) By following the procedure described in section 3.3 the applied boundary conditions in the RHS of the total system matrix is calculated resulting in a solvable system of the form A X = B. (34) The one dimensional column X represents the unknowns i.e. the voltage Φ in the pipe, the voltage V in the earth at the coating surface, the inner radial current density Jρ and the outer radial current density Qρ. Due to the use of (29) the potential at infinity is fixed at 0V . 4. Validation of the model 4.1. Electromagnetic parameters The total system matrix is numerically implemented in a C-environment of XCode Version 6.3 (6D570) through which any arbitrary configuration of a buried pipeline in the presence of a high voltage power line can be calculated. The used electromagnetic simulation parameters are listed in table 2. Table 2: Electromagnetic simulation parameters. Pipeline outer radius Rp Wall thickness Steel pipe conductivity σp Soil conductivity Coating conductivity σc Coating thickness δc Relative permittivity of the coating r,c Magnetic permeability of free space HV-line type Phase current Is,f Phase 1 Phase 2 Phase 3 Frequency 0.25m 2cm 5.88 106(Ωm)−1 0.01(Ωm)−1 4 10−6(Ωm)−1 4mm 5 4π10−7 N/A2 SH345kV NS 500A 0 rad 2π/3 rad −2π/3 rad 50Hz 15 Figure 4: BEM model with coating layer: simuation results for configuration 1. 4.2. Overview of the configurations Two specific configurations are simulated. The first configuration is one of a pipeline that parallels a high voltage line for a few kilometers and ends at both ends. The second configuration is a real life situation where a pipeline approaches a high voltage line at a certain angle, runs parallel with it for a few kilometers and diverges away from it. The coordinates of both configurations as well as that of the high voltage line are listed in table 3. 4.3. Simulation Results In order to verify the model, the results are compared with those of a transmission line model discussed in Bortels (2006). Since this latter model already has proven its validity, this comparison is justified. Furthermore, it is also observed by the authors that the coating layer could not be modelled as an impedance between the internal problem using the FEM and the external problem using the BEM, since this approach would assume that Jρi = Qρi in each node. It was observed after numerous calculations that such an assumption in this application leads to numerical errors. The simulation results of the first configuration together with those of the transmission line model are presented in figure 4, while those of the second configuration are presented in figure 5. 16 BEMTLM050010001500200025003000length(m)0246810ϕ(V)Figure 5: BEM model with coating layer: simuation results for configuration 2. Observe that analytical solutions do not exist. Hence, both models, the transmission line model and the proposed one in this paper are approxima- tions. However, since the transmission line model demands characteristic impedances at both ends of the configuration in order to simulate a pipeline that diverges away from a power line, it can be argued that the proposed model contains a better approach, since this condition is implemented in an intrinsic way. It is observed that the simulation results of the proposed model using a fundamental solution which takes into account the capacitive effect of the considered medium are in good agreement with those of a transmission line model. This good agreement is valid for a simple as well as a complex con- figuration. Yet, the proposed method includes all different electromagnetic characteristics in a directly implemented way including the coating thickness and conditions at both ends of a pipeline. 5. Conclusion In this paper a generalized fundamental solution when using the boundary element method for electromagnetic field problems is proposed. This gener- alized fundamental solution incorporates the capacitive effect when dealing with media wherein this effect is not negligible, such as coating layers. 17 BEMTLM020004000600080001000012000length(m)024681012ϕ(V)Using this generalized fundamental solution, a model was developed which combined a finite element method with the boundary element method of a buried coated pipeline in close proximity to a high voltage line, where as source conditions the magnetic vector potential originating from this high voltage line is calculated. The model is implemented by the authors in a C-environment, and any configuration can be calculated. Two configurations were implemented and the results were compared with the results of a transmission line model. There is a slight difference between the two models that we assign to unavoid- able approximations of both methods. This requires additional research and eventually comparison with field measurements. One can conclude that the proposed model with the generalized fundamental solution using the boundary element method is valid and can be used for industrial applications. 18 Bibliography References Balanis, C., 2012. Advanced Engineering Electromagnetics, 2nd Edition. John Wiley & Sons, Inc., USA. Bortels, L., D. J. M. C. . T. V., 2006. A general applicable model for ac predictive and mitigation techniques for pipeline networks influenced by hv power lines. IEEE Transactions on Power Delivery 21 (1), 210 -- 217. Brichau F., 1994. A Mathematical Model for the Cathodic Protection of Under- ground Pipelines including Stray Currents. Ph.D. Thesis, Vrije Universiteit Brussel, Belgium. Brebbia, C.A. & Dominguez, J., 1989. Boundary Elements: An Introduc- tory Course. Computational Mechanics Publications, McGraw-Hill Book Company, New York, USA. Carson, J., 1926. Wave propagation in overhead wires with ground return. Bell System Technical Journal Vol.5, 339 -- 35. CIGRE, Working Group 36.02 (1995). Guide to the influence of High Voltage AC Power Systems on Metallic Pipelines. Dawalibi, F.P. & Pinhi, A., 1986. Computerized analysis of power systems and pipelines proximity effects. IEEE Transactions on Power Delivery 4 (2), 40 -- 48. Dawalibi, F.P. & Southey, R., 1989. Analysis of electrical interference from power lines to gas pipelines, part i: Computation methods. IEEE Transac- tions on Power Delivery 4 (3), 1840 -- 1846. Gary, C., 1985. Nouvelle methode de calcul des inductances propres et mutuelles des lignes en presence du sol. CIGRE Symposium 06-85. Griffiths, D., 1999. Introduction to Electrodynamics: Third Edition. Prentice Hall International, Inc., Upper Saddle River, New Jersey, USA. L.D., G., 1996. Computer analysis of transient voltages in large grounding system. IEEE Transactions on Power Delivery 11 (2), 815 -- 823. 19 Munteanu, C., M. G. P. M. T. V. P. I. G. L. . R. C., June 2012. Electromag- netic field model for numerical computation of voltages induced on buried pipelines by high voltage overhead power lines. The European Physical Journal Applied Physics 58 (3), 30902 (9). Reitz, J.R., M. F. . C. R., 1979. Foundation of Electromagnetic Theory: Third Edition. Jaddison-Wesley Publishing Company, Inc., USA. Sadiku, M., 2001. Numerical Techniques in Electromagnetics, Second Edition. CRC Press LLC, Florida, USA. Schoonjans, B. & Deconinck, J., 2017. Calculation of HVAC inductive coupling using a generalized BEM for helmholtz equations in unbounded regions. International Journal of Electrical Power & Energy Systems 84 (2017) 242-251. Shen, J., 1995. Topics in Engineering Vol.24: Computational Electromag- netics using Boundary Elements - Advances in Modelling Eddy Currents. Computional Mechanics Publications, Southampton UK and Boston, USA. Stratton, J., 2007. A Classic Reissue: Electromagnetic Theory. IEE Press Series on Electromagnetic Wave Theory, Wiley-Interscience. Sunde, E., 1966. Earth Conduction Effects in Transmission Systems, Chapter 5. Van Nostrand, New York, USA. Sykulski, J., 1995. Computational Magnetics. James & James (Science Pub- lishers) Ltd and Chapman & Hall, London, UK. 20 Nomenclature (cid:126)A, Aρ, Az Magnetic vector potential Ap FEM matrix related with the axial magnetic vector potential [V s/m] [T ] [V /m] [V /m] [Hz] [A/m2] [m2] [m] Magnetic flux density BEM correction factor Electric field Electromotive force induced per unit length Frequency Fundamental solution BEM matrix related with the current density and the radial magnetic vector potential BEM matrix related the potential Imaginary unit Current density Surface of revolution FEM matrix related with the current densitiy Complex wave number Length of pipe element Normal unit vector Number of elements Vector to source point Vector to image point FEM matrix related with the potential (cid:126)B ci (cid:126)E E f G Gp H p i (cid:126)J, Jρ k K p k l (cid:126)n N (cid:126)r (cid:126)r(cid:48) Sp Sp α β δ 0 µr ν σc σs Φ ω . . . Axial conductivity of the pipe [1/Ωm] Attenuation transmission coefficient Phase transmission coefficient Skin depth Electric constant Relative permittivity Relative permeability Coefficient expressing a scaled radius Conductor electrical conductivity Soil electrical conductivity Electrical potential Angular frequency Magnitude of vector 21 [N p/m] [rad/m] [m] [F/m] [1/Ωm] [1/Ωm] [V ] [rad/s] Table 3: Coordinates of the power line and buried pipeline. x [km] y [km] z[m] HV-line point 1 point 2 Pipeline (configuration 1) point 1 point 2 Number of elements Length of a pipe element Pipeline (configuration 2) point 1 point 2 point 3 point 4 point 5 point 6 point 7 point 8 point 9 point 10 point 11 point 12 point 13 point 14 point 15 point 16 point 17 point 18 point 19 point 20 point 21 Number of elements Length of a pipe element -1 12 4.5 7.5 0 2 3.1 3.75 4 5 6 7 7.057 7.094 7.114 7.17 7.28 7.556 7.9 8 8.844 9.79 10 11 12 0 0 0.025 0.025 2 1.025 0.495 0.18 0.057 0.057 0.057 0.057 0.029 0.010 -0.010 -0.020 -0.080 -0.230 -0.400 -0.450 -0.872 -1.345 -1.450 -1.950 -2.450 18,29 18,29 -5 -5 45 66.7m -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 -5 261 51.2m 22
1802.01401	1	1802	2018-01-12T16:04:51	Transition Metal Chalcogenide Tin Sulfide Nanodimensional Films Align Liquid Crystals	[ "physics.app-ph" ]	Transition metal chalcogenide tin sulfide (SnS) films as alternative noncontact alignment layer for liquid crystals, have been demonstrated and investigated. The SnS has an anisotropic atomic chain structure similar to black Phosphorous which causes the liquid crystal molecules to align without the need for any additional surface treatments. The high anisotropic nature of SnS promotes the alignment of the easy axis of liquid crystal molecules along the periodic atomic grooves of the SnS layer. The atomically thin SnS layers were deposited on indium tin oxide films on glass substrates, at room temperature by chemical vapor deposition. The device characteristics are comparable to those commercially available, which use photo-aligning polymer materials. We measured threshold voltage of 0.92V, anchoring energy of 1.573x10^(-6) J/m^2, contrast ratio better than 71:1 and electro-optical rise/fall times of 80/390ms, respectively for ~11 micron thick liquid crystal device as expected.	physics.app-ph	physics	Transition Metal Chalcogenide Tin Sulfide Nanodimensional Films Align Liquid Crystals Asi Solodar1, Ghadah AlZaidy2, Chung-Che Huang2, Daniel W. Hewak2, Ibrahim Abdulhalim1,3 1Ben Gurion University of the Negev, Department of Electro-Optics Engineering and The Ilse Katz Institute for Nanoscale Science and Technology, Beer Sheva 84105, Israel 2Optoelectronics Research Centre, University of Southampton, Southampton, SO17 1BJ, U.K. 3Singapore-HUJ Alliance for Research and Enterprise (SHARE), Nanomaterials for Energy and Energy-Water Nexus (NEW), Campus for Research Excellence and Technological Enterprise (CREATE), Singapore 138602 KEYWORDS: Transition Metal Chalcogenide, Tin Sulphide, Liquid crystal, 2D materials, Anisotropy. 1 ABSTRACT Transition metal chalcogenide tin sulfide (SnS) films as alternative noncontact alignment layer for liquid crystals, have been demonstrated and investigated. The SnS has an anisotropic atomic chain structure similar to black Phosphorous which causes the liquid crystal molecules to align without the need for any additional surface treatments. The high anisotropic nature of SnS promotes the alignment of the easy axis of liquid crystal molecules along the periodic atomic grooves of the SnS layer. The atomically thin SnS layers were deposited on indium tin oxide films on glass substrates, at room temperature by chemical vapor deposition. The device characteristics are comparable to those commercially available, which use photo-aligning polymer materials. We measured threshold voltage of 0.92V, anchoring energy of 1.573 × 10−6 𝐽/𝑚2, contrast ratio better than 71:1 and electro-optical rise/fall times of 80/390ms, respectively for ~11 micron thick liquid crystal device as expected. 2 1. INTRODUCTION Over the past several years, 2D materials with diverse properties are under investigation for their great potential in photonic and electronic applications. Examples of such newly rising materials include Graphene, transition metal (di)-chalcogenides and black-phoshorous.[1] The bulk black phosphorus (BP) is the most stable allotrope of the phosphorus and has a high hole mobility, strong in-plane anisotropy and tunable direct bandgap energy of about ~0.3 eV to ∼2 eV, covering the visible to infrared spectrum.[2,3] Recently, alignment of rod-like LC molecules, such as thermotropic and lyotropic LCs on black phosphorus surfaces were demonstrated.[4] However, the preparation of BP is complicated and involves high temperature and pressure. In addition the BP films are very sensitive to water vapor and oxygen, which leads to their rapid degradation, when exposed to air .[5,6] On the other hand, the study of Chao et al.,[7] assessed the Tin sulfide (SnS) material as a compound analogue of elemental black phosphorus with higher chemical stability than the latter one. The SnS material is a IV-VI compound semiconductor with direct band gap energy 1.2 eV to ∼1.5 eV, high carrier mobility and high in-plane anisotropy.[8] In addition SnS exhibits a chemical inert surface without dangling bonds and surface density states, which makes it chemically and environmentally stable. Furthermore, the mere fact of high in-plane anisotropy, can teach on possible alignment abilities of LC molecules. Liquid crystals (LC) are of a great importance for many electro-optical applications. They have found wide use in a large variety of electro-optical modules, such as flat panel displays,[9] linear polarization rotators,[10] dynamical wave plate retarders,11 spatial light modulators and tunable filters. [12-15] A typical LC device has a configuration of a sandwich cell in which LC is filled between a pair of glass substrates coated with thin alignment layer on top of an indium tin oxide (ITO) films, used as transparent conducting electrodes. The alignment layers determine the 3 orientation of LC molecules on the surface in the absence of an external electric field, and therefore the quality of the device and its response to an applied electric field. As a result, the interaction between the LC molecules and the alignment layer is of great importance in determining the device performance. Obtaining high contrast LC devices is immensely important for certain imaging and photonic applications; hence better alignment layers that can improve the devices quality is still in demand. In addition the fundamental science behind the interfacial interactions of liquid crystals and alignment layers is of high interest. Several surface treatment methods have been developed during the last century to form the surface alignment films for LCs. Such well known alignment techniques as unidirectional mechanical rubbing of thin polymer or polyimide spin coated on glass substrate,[16] photo- alignment by illumination with ultraviolet polarized light of photosensitive polymers and chemically by surface active agents (surfactants). [17] As a result, LC molecules are aligned through the surface anisotropy caused by either the micro/nano-grooves, which had been formed on the rubbed surface, or by the interaction with polymer arranged molecular chains in photo activated polymers or interaction with vertically aligned surfactant molecules on the surface. Although widely used in the industry, the mechanical rubbing method has some shortcomings, as it involves physical contact interaction between the cloth and the surface, which can introduce particle dust generation and surface damage. On the other hand, noncontact methods such as photo–alignment, suffer from insufficient time stability and lower anchoring energy. In order to overcome the imperfections of traditional aforementioned techniques, some alternative surface alignment methods have been investigated based on noncontact ideology. New concepts for photo alignment were investigated by our group, where nano dimensional chalcogenide glass films evaporated in vacuum and utilized for LC alignment as well as photo- 4 sensing film for optically addressed spatial light modulators.[18,19] Analogous to conventional photoalignment materials, this method exhibits photo-induced anisotropy, stimulated by polarized blue light irradiation (436 nm). Another method was proposed by Kitson et al.,[20] where they demonstrated that tilted micron-scale posts prepared by photolithographic processes, can be used to generate uniform liquid crystal alignment. In addition, our group recently reported a direct writing of nano structured patterns on ITO coated glass surface by ultrafast femtosecond laser and LC alignment on lead sulphide (PbS) nano-sculptured thin films with glancing angle deposition technique as an alternative method for LC alignment layers with different alignment orientations and better transparency.[21-24] Apart from photo-induced optical anisotropy behavior, other types of alignment are manifested as an artificial generation of anisotropy on top of the surface and are not a result of intermolecular bonds. Therefore, those alignment methods at this stage cannot guarantee that LC molecules will orient unequivocally in preferable alignment direction and thus, at the same time can exhibit anti-parallel and parallel configurations as well. In this work, we report on a novel alignment method of LC molecules based on the anisotropic nature of the SnS crystal structure without any additional surface treatments. The atomically thin SnS layers were deposited on indium tin oxide (ITO) films on top of the glass substrates, at room temperature by chemical vapor deposition (CVD) techniques. High contrast LC devices are shown to be achievable with this methodology. 2. EXPERIMENTAL SECTION In this work, we have used two ITO coated glass substrates, each of dimensions 17x15x2.5 mm3 coated with 145±10 nm ITO film (purchased from Thin Film Devices Inc. California, USA) and assembled on either side of the LC device. Prior to use, the substrates were cleaned by ultrasonic 5 bath with deionized (DI) water, isopropanol (IPA) and at last with acetone, where each cycle lasts about 20 minutes at a temperature of 50 oC. (b) (a) Figure 1. LC device assembly. (a) Deposition of SnS layer on top of the ITO coated glass substrate and SnS molecular structure with LC molecules aligned along the periodic atomic grooves. (b) Illustration of two possible alignment configurations, which can be achieved by this method. Hereafter, SnS thin films were deposited on the ITO substrates (Figure 1a) by CVD with SnCl4 precursor to react with H2S gas at room temperature and then annealed at 450 oC for 2 hours.[25] Then, the liquid crystal mixture E-7 (Merck) was sandwiched between SnS coated substrates, separated by 10 micron spacers. Generally, at this stage two possible configurations of anti- parallel and parallel alignment modes can be achieved as depicted in Figure 1b. The extinction and alignment states of the devices were examined under polarization microscope (Olympus mx50) between crossed polarizers in reflection mode. The linearly polarized light (P) was focused by x5 objective onto the device as demonstrated in Figure 2. Upon reflection from the surface, the linearly polarized light beam is then split by the beam splitter, while part of it is passing through the analyzer (A), which is perpendicularly oriented with respect to polarizer axis. 6 First, the liquid crystal device was placed with its optical axis parallel to the axis of one of the polarizers in order to examine the extinction level. Then the LC alignment axis was oriented at 45o towards the polarizer axis, which will provide maximum reflected light intensity. Figure 2. Illustration of LC device between crossed polarizer (P) and analyzer (A), under the polarization microscope in reflection operational mode. Demonstration of two working configurations: (1) Extinction (zero degrees) configuration with LC alignment axis (𝑃𝐿𝐶) being parallel to the polarizer axis and (2) maximum reflected light intensity, where 𝑃𝐿𝐶 axis oriented at 450 with respect to the polarizers axis. The transmission spectrum characteristics of the device were studied by the optical setup, illustrated in Figure 3. The device was placed at 45o between crossed polarizers, with respect to the polarizer axis. White collimated linearly polarized light passed through the LC device and focused onto the fiber spectrometer by the x5 objective. 7 Figure 3. Schematic representation of the electro optical setup to determine transmission spectrum of the LC device. The setup comprises a fiber light source collimated using a lens with focal length f=8cm, polarizer, aperture, the LC device attached to the 6-axis kinematic mount, an analyzer, a x5 focusing microscope objective, an output fiber connected to spectrometer (StellarNet Inc.). 3. RESULTS AND DISCUSSION First, in order to illustrate the LC molecules alignment along the periodic grooves of SnS layer, we used polarizing microscope as illustrated in Figure 2. The device placed between crossed polarizers and the alignment direction (top surface easy axis) oriented parallel to the polarizer axis as illustrated in zero degrees configuration in Figure 2. As a result, a high extinction level was achieved as depicted in the inset of Figure 4. It is clearly evident that the LC molecular alignment axis (easy axis) coincident with the SnS molecular grooves direction. Then, the device was oriented 45° with respect to the polarizers as illustrated in the 45° configuration in Figure 2, A 2kHz sinusoidal wave with variable amplitude voltage values was applied across the device in order to reorient the liquid crystal director. 8 Figure 4. Camera read out of LC reflection intensities under polarized microscope between crossed polarizers as illustrated in Figure 2. The LC device rotated at 45 degrees with respect to the polarizer axis, while various amplitude AC voltages were applied. Inset black image in the 0V case shows extinction position when the LC easy axis oriented parallel to the axis of the polarizer. As demonstrated in Figure 4 the device exhibits a uniform reflection change and tunable color at various bias voltages. In addition, a uniform extinction is demonstrated, when the LC device alignment axis is parallel to the axis of one of the polarizers as depicted in Figure 2a. Characterization of the LC device as a function of voltage was achieved quantitatively, also by spectral transmission measurements. The device was placed between crossed polarizers at an angle of 45𝑜 between the easy axis of the LC (𝑃𝐿𝐶) and the polarizer as depicted in Figure 3. The measured transmission spectra for various applied voltages are shown in Figure 5. 9 Figure 5. Spectrometer readout from electro optical setup demonstrated in Figure 3. Transmission spectra at different applied voltages (sinusoidal wave at 1 kHz) between crossed polarizers. Upon reaching the threshold voltage, the transmission spectrum considerably changed (Figure 5). As can be seen, the transmission spectrum can be effectively controlled by the applied voltage, which changes the orientation of the LC director. As expected, for high enough bias voltages the alignment in the bulk is practically parallel to the field, thus the incoming linear polarized light feels a very weak optical anisotropy, resulting from residual influence of fixed boundary conditions. Thus, any further voltage increase will reduce the transmission until it will be completely blocked by the analyzer. In order to obtain maximum accuracy during further electro optical measurements, a photodetector was used conjugated to lock-in amplifier SR830 (Stanford Research Systems) as illustrated in Figure 6a. The LC device was placed at 45o between crossed polarizers in order to receive maximum transmission, while a He-Ne laser (633 nm) was used as a light source. 10 Figure 6. Illustration of electro optical setup for transmission measurement through LC device (a). Modulated by the chopper HeNe laser source was transmitted through the LC device, which was placed at 450 between crossed polarizers. (b) Transmitted signal detected by the photo-detector. (c) Accumulated retardation of the device, deduced from the transmitted signal. (d) Measured threshold voltage. The applied voltage to the device was swept in steps of 0.1V, while the transmission signal was detected by a photodiode connected to the lock-in amplifier. Based on the transmitted intensity signal (Figure 6b), we can deduce the accumulated phase retardation of the device as a function of voltage. As shown in Figure 6c, the retardation tends to decrease at higher applied voltages as expected for homogeneously aligned nematic LCs with positive dielectric anisotropy. In addition, we can also utilize the transmitted signal obtained by the electro optical setup shown in Figure 6, in order to determine the threshold voltage of the LC device. As can be seen from the 11 linear fit in Figure 6d, the measured threshold voltage is 0.92V, which is in very close agreement to the value calculated based on the equation for splayed cell: 𝑠𝑝𝑙𝑎𝑦 = 𝜋√ 𝑉𝑡ℎ 𝐾11 𝜀0Δ𝜀 (1) Here 𝐾11 is a splay elastic constant, Δ𝜀 is a dielectric anisotropy and 𝜀0 is the vacuum permittivity. For LC E7: Δ𝜀 = 14, 𝐾11 = 10.8 𝑝𝑁.[26] In order to measure the switching speed of the device, a modulated AC signal was applied. The light from a collimated laser diode (633nm) transmitted through the device and detected by a photodiode, which is connected to a digitizing oscilloscope. The switching speed is defined as the time required for a change from 10% to 90% of the total transmission change. The rise and fall times of the LC were approximately 80ms and 390ms, respectively as shown in Figure 7. (a) (b) Figure 7. Measured electro-optical time response of the LC device (a). The transient oscillations are a result of the relatively thick LC cell. The measured decay time is about 390ms and the rise time is around 80ms. (b) Measured and simulated transmission signal for normal incident light through the LC device, between crossed polarizers, oriented at 45-degree. 12 The relaxation time (𝑡𝑟𝑒𝑙𝑎𝑥) and the rise time (𝑡𝑟𝑖𝑠𝑒) of a LC device can be evaluated by the dynamics of the Freedericksz transition under the one splay elastic constant approximation and can be calculated by: 𝑡𝑟𝑒𝑙𝑎𝑥 = 2 ) , 𝑡𝑟𝑖𝑠𝑒 = 𝛾𝜃 𝐾11 ( 𝑑 𝜋 𝑡𝑟𝑒𝑙𝑎𝑥 𝑠𝑝𝑙𝑎𝑦 (𝑉 𝑉𝑐 ⁄ 2 − 1 (2) ) Here 𝛾𝜃 is the rotational viscosity of the LC, d is the thickness of the LC layer and V is the RMS value of an applied voltage supply. In order to calculate the theoretical value of the response time we need an accurate measurement of the thickness. The thickness of the cell was determined by fitting the experimentally measured transmission spectrum through the device for normal incidence light to the theoretical simulated results (Figure 7b), given by Equation 3: 𝑇 = 𝑐𝑜𝑠2(𝑃 − 𝐴) − sin (2𝑃) ∙ sin (2𝐴) ∙ 𝑠𝑖𝑛2 (Γ𝐿𝐶 2⁄ ) (3) Here P and A are the angles between the LC device alignment axis and the polarizer / analyzer, respectively and Γ is the LC phase retardation, defined as: Γ𝐿𝐶 = 2𝜋 𝜆 𝑑 ∫ (𝑛𝑒𝑓𝑓(𝑧) − 𝑛𝑜)𝑑𝑧 0 (4) Here 𝜆 is the wavelength, 𝑧 is the depths coordinate normal to the cell, 𝑛𝑜 and 𝑛𝑒𝑓𝑓(𝑧) are the ordinary and the extraordinary refractive index of the LC, respectively and given by: 𝑛𝑒𝑓𝑓(𝑧) = 𝑛⊥𝑛∥ , 𝑛𝑜 = 𝑛⊥ √𝑛⊥ 2 𝑐𝑜𝑠2𝜃(z) + 𝑛∥ 2𝑠𝑖𝑛2𝜃(z) (5) For LC E-7: 𝑛∥ = 1.693 and 𝑛⊥ = 1.499. The extraordinary refractive index depends on the LC angle 𝜃(z) and in the absence of external electric field can be interpreted as the LC pre-tilt angle. As apparent from Equations 2-5 and 13 Figure 7b, the thickness of LC device is 𝑑 = 11.52𝜇𝑚 with a pre-tilt angle of 𝜃(0, 𝑑) = 2𝑜. Thus, according to Equation 2, the fall and rise times of the LC should be about 302ms and 11ms (for V=5V), respectively. Despite the deviation between the measured and calculated switching times, the results are still in a plausible range of thick time response LC devices. Part of the disagreement is due to the fact that the expressions in equation 2 are approximations for the small deformations, very small pretilt angle and strong anchoring cases while here we used 5V which is more than five folds the threshold voltage and the anchoring as will be shown below can be considered intermediate. Another reason is the fact that we are measuring the electrooptic response times while the expressions in equation 2 refer to the molecules rotational motion. For example the relaxation time under the intermediate anchoring condition can be written as[22]: (6) Using the values of the elastic constant, the thickness and the anchoring energy of 1.573 × 10−6 𝐽/𝑚2 as estimated below we get that the relaxation time when considering the anchoring energy effect should be larger by a factor of x3 bringing it to around 900 msec and the rise time to around 30msec. The measure fall and rise times are 390 msec and 80 msec. The relaxation time is in good agreement if we consider the fact that the electrooptical relaxation time is nearly half the molecular relaxation time.[27] The rise time measured electro-optically can be larger than the molecular rise time depending on the applied voltage. An increase by factor of x2.5 can be explained as originating from the fact that the pretilt angle is 2 degrees, the use of large deformation and the effect of the electrooptical transfer function.[27] As mentioned slower switching times may also indicate intermediate surface anchoring energy and defect domains. However; no defects observed over the entire active area, under the polarized microscope with and without applied voltages. The strength of the anchoring (W𝐴) between the 14 dWKrelaxrelax41'nematic LC molecules and the SnS alignment layer can be determined by improved high electric field method, proposed by Nastishin et al.[28] This method is based on linear fit of the voltage 0⁄ (𝑉 − 𝑉′) and phase retardation normalized by the zero voltage phase Γ𝐿𝐶(𝑉 − 𝑉) Γ𝐿𝐶 . Here V is an applied voltage and 𝑉′ = 𝛼(Δ𝜀 𝜀∥⁄ )𝑉𝑐 𝑠𝑝𝑙𝑎𝑦 , where 𝛼 coefficient varies between 1 − 2 𝜋⁄ . Additional discussions and various aspects of this method are described elsewhere.[28] The value of measured anchoring energy was on the order of 1.573 × 10−6 𝐽/𝑚2 as deduced from Figure 8. Figure 8. Plot of 𝛤𝐿𝐶(𝑉 − 𝑉)/𝛤𝐿𝐶 𝑜 as function of (𝑉 − 𝑉) and linear fit used for anchoring energy evaluation (W𝐴), measured by electric field techniques,[28] yielding the value of 1.573 × 10−6 𝐽/𝑚2. This anchoring energy is of the same magnitude range as that measured using commercial photo- aligning polymer materials, which is usually limited to between 10−5 − 10−6 𝐽𝑚−2 .[29,30] Another important measure of the performance of a LC based devices is the contrast ratio. From measurement of the transmitted intensity signal in Figure 6b, we found the contrast ratio of: 𝑇𝑚𝑎𝑥 𝑇𝑚𝑖𝑛 ⁄ =44.487 0.622 ⁄ = 71: 1. We believe that this value can be increased after further optimization process such as better cleaned surfaces prior to SnS coating. 15 4. CONCLUSIONS The control of molecular alignment in liquid-crystal based applications has a high importance in the overall performance of the devices, thus new alignment techniques are still extensively investigated. Here we demonstrated an alternative noncontact alignment method based on the naturally highly anisotropic tin sulfide crystal structure. The molecular characteristics of di-chalcogenide SnS material is very similar to the structure of black phosphors, however in the former case the material is much more stable against atmospheric interactions. The LC molecules clearly reveal the preferred orientation along the molecular chains of the SnS films and successfully reorients as a result of electrical supplied voltage. The device exhibits threshold voltage of 0.92V, which is typical to nematic phase LC devices. In addition the measured anchoring energy is about 1.573 × 10−6 𝐽/𝑚2 comparable with the anchoring energy of nematic LC on photosensitive polymers. The rise and fall switching times of 80ms/390ms, respectively also demonstrate plausible switching times for thick LC devices. The contrast ratio at this stage is about 71:1, but can be increased by optimization process. We believe that this alignment technique is of high potential due to its simplicity, limited in scale by only the size of the CVD chamber and the high contrast that can be achieved. Corresponding Author (IL). Tel: (+972)08-6428599 E-mail: [email protected] ACKNOWLEDGMENT This research is partially supported by grants from the National Research Foundation, Prime Minister's Office, Singapore under its Campus of Research Excellence and Technological Enterprise (CREATE) programme. The 2D materials work at the University of Southampton is funded in part through the Future Photonics Manufacturing Hub (EPSRC EP/N00762X/1). 16 REFERENCES [1] X. Ling, H. Wang, S. X. Huang, F. N. Xia, M. S. Dresselhaus, The renaissance of black phosphorus. P Natl Acad Sci USA 2015, 112 (15), 4523-4530. [2] F. N. Xia, H.Wang, Y. C. Jia, Rediscovering black phosphorus as an anisotropic layered material for optoelectronics and electronics. Nat Commun 2014, 5, 4458. [3] H. Liu, Y. C. Du, Y. X. Deng, P. D. Ye, Semiconducting black phosphorus: synthesis, transport properties and electronic applications. Chem Soc Rev 2015, 44 (9), 2732-2743. [4] D. W. Kim, H. S. Jeong, K. O. Kwon, J. M. Ok, S. M. Kim, H. T. Jung, Ultrastrong Anchoring on the Periodic Atomic Grooves of Black Phosphorus. Adv Mater Interfaces 2016, 3 (22), 1600534. [5] J. O. Island, G. A. Steele, H. S. J. van der Zant, A. C. Gomez, Environmental instability of few-layer black phosphorus. 2d Mater 2015, 2 (1), 011002. [6] A. Ziletti, A. Carvalho, D. K. Campbell, D. F. Coker, A. H. C. Neto, Oxygen Defects in Phosphorene. Phys Rev Lett 2015, 114 (4), 046801. [7] C. Xin, J. X. Zheng, Y. T. Su, S. K. Li, B. K. Zhang, Y. C. Feng, F. Pan, Few-Layer Tin Sulfide: A New Black-Phosphorus-Analogue 2D Material with a Sizeable Band Gap, Odd- Even Quantum Confinement Effect, and High Carrier Mobility. J Phys Chem C 2016, 120 (39), 22663-22669. [8] N. K. Reddy, M. Devika, E. S. R. Gopal, Review on Tin (II) Sulfide (SnS) Material: Synthesis, Properties, and Applications. Crit Rev Solid State 2015, 40 (6), 359-398. [9] Khoo, I.-C. Liquid crystals: physical properties and nonlinear optical phenomena, Wiley: New York, N.Y., 1995. [10] I. Abdulhalim, Liquid crystal active nanophotonics and plasmonics: from science to devices. J Nanophotonics 2012, 6, 061001. [11] D.-K. Yang, S.-T. Wu, Fundamentals of liquid crystal devices, 2 ed.; John Wiley: Chichester ; Hoboken, NJ, 2006. [12] P. K. Shrestha, Y. T. Chun, D. P. Chu, high-resolution optically addressed spatial light modulator based on ZnO nanoparticles. Light-Sci Appl 2015, 4,e259. [13] A. Solodar, T. A. Kumar, G. Sarusi, I. Abdulhalim, Infrared to visible image up-conversion using optically addressed spatial light modulator utilizing liquid crystal and InGaAs photodiodes. Appl Phys Lett 2016, 108 (2), 021103. 17 [14] S. T. Wu, Design of a Liquid-Crystal Based Tunable Electrooptic Filter. Appl Optics 1989, 28 (1), 48-52. [15] L. Vicari, Optical applications of liquid crystals, Institute of Physics Pub.: Bristol, 2003. [16] J. M. Geary, J. W. Goodby, A. R. Kmetz, J. S. Patel, The Mechanism of Polymer Alignment of Liquid-Crystal Materials. J Appl Phys 1987, 62 (10), 4100-4108. [17] N. Kawatsuki, H. Ono, H. Takatsuka, T. Yamamoto, O. Sangen, Liquid crystal alignment on photoreactive side-chain liquid-crystalline polymer generated by linearly polarized UV light. Macromolecules 1997, 30 (21), 6680-6682. [18] I. Abdulhalim, M. G. Kirzhner, Y. Kurioz, M. Klebanov, V. Lyubin, Y. Reznikov, N. Sheremet, K. Slyusarenko, Integration of chalcogenide glassy films and liquid crystals for photoalignment and optically addressed modulators. Phys Status Solidi B 2012, 249 (10), 2040-2046. [19] M. Gelbaor, M. Klebanov, V. Lyubin, I. Abdulhalim, Permanent photoalignment of liquid crystals on nanostructured chalcogenide glassy thin films. Appl Phys Lett 2011, 98 (7), 071909. [20] S. C. Kitson, E. G. Edwards, A. D. Geisow, Designing liquid crystal alignment surfaces. Appl Phys Lett 2008, 92 (7), 073503. [21] A. Solodar, A. Cerkauskaite, R. Drevinskas, P. G. Kazansky, I. Abdulhalim, Ultrafast Laser Nanostructured ITO Acts as Liquid Crystal Alignment Layer and Higher Transparency Electrode. To be published. [22] A. Chaudhary, M. Klebanov, I. Abdulhalim, Liquid crystals alignment with PbS nanosculptured thin films. Liquid Crystals 2017, 1-8. [23] A. Solodar, A. Cerkauskaite, R. Drevinskas, P. G. Kazansky, I. Abdulhalim, Liquid crystal alignment on ultrafast laser nanostructured ITO coated glass, CLEO/Europe-EQEC Munich, 2017. https://eprints.soton.ac.uk/410668/ [24] A. Solodar, A. Cerkauskaite, R. Drevinskas, P. G. Kazansky, I. Abdulhalim, Nanogratings Written by a Femto-second Laser on ITO Act as Higher Transparency Electrode and Alignment Layer for Liquid Crystals, ICMAT, Singapore, 2017. https://www.conftool.pro/icmat2017/index.php?page=browseSessions&form_session=351 18 [25] G. Alzaidy, C. C. Huang, D. W. Hewak, CVD-grown tin sulphide for thin film solar cell devices. In 11th Photovoltaic Science Applications and Technology (PVSAT-11), United Kingdom, 2015. [26] H. W. Chen, R. D. Zhu, J. X. Zhu, S. T. Wu, A simple method to measure the twist elastic constant of a nematic liquid crystal. Liquid Crystals 2015, 42 (12), 1738-1742. [27] H. Wang, T. X. Wu, X. Zhu, S.T. Wu, Correlations between liquid crystal director reorientation and optical response time of a homeotropic cell. J. Appl. Phys. 2004, 95 (10), 5502. [28] Y. A. Nastishin, R. D. Polak, S. V. Shiyanovskii, V. H. Bodnar, O. D. Lavrentovich, Nematic polar anchoring strength measured by electric field techniques. J Appl Phys 1999, 86 (8), 4199-4213. [29] V. G. Chigrinov, V. M. Kozenkov, H.-S.Kwok, Photoalignment of liquid crystalline materials : physics and applications, John Wiley & Sons: Chichester, 2008. [30] O. Yaroshchuk, Y. Reznikov, Photoalignment of liquid crystals: basics and current trends. J Mater Chem 2012, 22 (2), 286-300. 19
1905.07838	3	1905	2019-09-02T04:40:25	Substrate Integrated Bragg Waveguide: an Octave-bandwidth Single-mode Functional Transmission-Line for Millimeter-Wave and Terahertz Applications	[ "physics.app-ph", "physics.optics" ]	We demonstrate an air-core single-mode hollow waveguide that uses Bragg reflector structures in place of the vertical metal walls of the standard rectangular waveguide or via holes of the so-called substrate integrated waveguide. The high-order modes in the waveguide are substantially suppressed by a modal-filtering effect, making the waveguide operate in the fundamental mode over more than one octave. Numerical simulations show that the propagation loss of the proposed waveguide can be lower than that of classic hollow metallic rectangular waveguides at terahertz frequencies, benefiting from a significant reduction in Ohmic loss. To facilitate fabrication and characterization, a proof-of-concept 20 to 45 GHz waveguide is demonstrated, which verifies the properties and advantages of the proposed waveguide. A zero group-velocity dispersion point is observed at near the middle of the operating band. This work offers a step towards a novel hybrid transmission-line medium that can be used in a variety of functional components for broadband millimeter-wave and terahertz applications.	physics.app-ph	physics	Substrate Integrated Bragg Waveguide: an Octave-bandwidth Single-mode Functional Transmission-Line for Millimeter-Wave and Terahertz Applications Binbin Hong,1,2 Naixing Feng,1 Jing Chen, 1 Guo Ping Wang,1,* Viktor Doychinov,2 Roland Clarke,2 Nutapong Somjit,2 John Cunningham,2 and Ian Robertson2,† 1 Institute of Microscale Optoelectronics, Shenzhen University, Shenzhen 518060, China 2School of Electronic and Electrical Engineering, University of Leeds, Leeds LS2 9JT, UK We demonstrate an air-core single-mode hollow waveguide that uses Bragg reflector structures in place of the vertical metal walls of the standard rectangular waveguide or via holes of the so-called substrate integrated waveguide. The high-order modes in the waveguide are substantially suppressed by a modal- filtering effect, making the waveguide operate in the fundamental mode over more than one octave. Numerical simulations show that the propagation loss of the proposed waveguide can be lower than that of classic hollow metallic rectangular waveguides at terahertz frequencies, benefiting from a significant reduction in Ohmic loss. To facilitate fabrication and characterization, a proof-of-concept 20 to 45 GHz waveguide is demonstrated, which verifies the properties and advantages of the proposed waveguide. A zero group-velocity dispersion point is observed at near the middle of the operating band. This work offers a step towards a novel hybrid transmission-line medium that can be used in a variety of functional components for broadband millimeter-wave and terahertz applications. I. INTRODUCTION [1], sensing Recently, the millimeter-wave (mmW) and terahertz (THz) frequency bands have attracted much attention owing to their unique applications, such as ultra-broadband spectroscopic high-data-rate wireless communication [2], imaging and non-destructive testing [3], security screening [4] and radio astronomy [5]. However, the development of practical mmW to THz systems has proved to be challenging due to the lack of high-power radiation sources and high-sensitivity detectors, along with the high loss of planar transmission-line structures. Traditional hollow metallic rectangular waveguides are surprisingly difficult to replace as the main electromagnetic waveguiding medium for the integration of subsystems. There is great interest in finding an alternative transmission-line medium that is low-loss, single-mode, broadband, compact, and has good isolation but is much easier to fabricate and suitable for functional components and multichip modules. Planar transmission lines are widely used in functional components and multichip modules for realizing passive components and interconnects, but they generally suffer from high substrate and radiation losses at 1 THz and above, as well as mode competition with the substrate modes [6]. Reported propagation loss results for microstrip, coplanar waveguide, coplanar stripline, and Goubau line at 1 THz are approximately 𝛼ms=43.3 dB/cm, 𝛼cpw=65.1 dB/cm, 𝛼cs=26 dB/cm, and 𝛼gl =30.4 dB/cm, respectively [7-10]. The dielectric substrate of planar transmission lines supports unwanted TM modes which limit the single-mode operating frequency band. An on-chip THz liquid sensor based on a Goubau line has been reported operating up to 0.8 THz, the * E-mail: [email protected] † E-mail: [email protected] in highest operating frequency limit of which is mainly determined by the mode competition between the guided mode and surrounding modes the substrates or superstrates [11]. The propagation loss of rectangular dielectric waveguide has been reported to be less than 0.54 dB/cm between 0.44 and 0.5 THz with a minimum of 0.18 dB/cm occurring at 0.481 THz [12]. The loss is mainly limited by the absorption of the host material, but it still suffers from poor isolation [12, 13]. Photonic crystal slab waveguide has been shown to have a loss of about 4 dB/cm between 0.54 and 0.63 THz, but likewise, it suffers from poor isolation and limited bandwidth which is undesirable for multilayer integration and broadband applications [14, 15]. Substrate-integrated waveguide (SIW) has been reported to be broadband and operate to 0.5 THz with loss less than 3.5 dB/cm by selecting different operating modes at different bands, but it is difficult to metalize the vias in SIW - especially when the vias become very small at short wavelengths. Microstructured fibers, including photonic crystal fibers [16-18], antiresonant fibers [19], porous step- index fibers [20], hollow dielectric-lined fibers [21], etc., have all been reported as promising solutions for low-loss quasi-single-mode THz wave-guiding, but they are as yet relatively bulky and hard to integrate with active devices. In this paper, we show that millimeter and THz waves can be tightly confined in the air core of a hybrid waveguide that is essentially a hollow rectangular waveguide with Bragg reflectors instead of vertical side walls. By avoiding the vertical walls, the waveguide can be fabricated as a photonic crystal structure on a single planar substrate sandwiched between two parallel ground planes, with no metalized via holes. It is, in essence, a flat form of the well-known Bragg fiber [22] that can be realized on a single substrate and so we call it a Substrate Integrated Bragg Waveguide (SIBW). Since it can be realized on a single substrate it is versatile for designing a variety of functional components based on the highly flexible and expandable photonic crystal structure, such as sensors, filters, H-plan horn antennas, lasers and frequency combs, which has already been successfully demonstrated in optical frequencies [23-27]. Based on our theoretical analysis, the SIBW operates in the fundamental HE10 mode while suppressing the high-order competing modes using a modal-filtering effect, which allows the waveguide to operate in single-mode fashion over more than one octave. Compared with planar transmission line modes, the vertically polarized HE10 mode of the SIBW can be more easily coupled with the free-space Gaussian mode or the TE10 mode of a rectangular waveguide, which are the most widely used interfaces for THz-TDS or THz vector network analyzer (VNA) systems, respectively. Besides, the SIBW uses dielectric rather than metal in the vertical wall resulting in a lower propagation loss than that of the classic hollow metallic rectangular waveguide. A proof-of-concept millimeter-wave SIBW was designed and tested in order to verify the transmission properties of the proposed structure. Although it has a relatively low operating frequency, it verifies the physical mechanisms for octave bandwidth single-mode operation and so demonstrates that this is an important candidate for an octave-bandwidth single-mode transmission line for mmW and THz functional components, such as filters, power dividers, and H-plane horn antennas. II. PRINCIPLE The schematic diagram of the proposed SIBW is illustrated in Fig. 1(a). The top and bottom layers are parallel metal ground planes, and the middle layer is a substrate that realizes an air-core line-defect 1D photonic crystal Bragg reflector structure. The Bragg reflector consists of periodically alternating layers of high- and low-refractive- index materials. Air is chosen as the low-refractive-index material for its low refractive index and low loss. The other material that forms the Bragg reflector is desired to be high refractive index and low loss, to create a large photonic bandgap and reduce the material absorption loss. High- resistivity silicon is one of the most commonly used materials which has high refractive index and low absorption loss in the THz frequency range [28], and thus it is chosen as Top Metal Plate (Semi-transparent) 𝑏 𝑎 𝑤 ℎ Air 𝑦 𝑧 𝑥 Dielectric Slab Bottom Metal Plate （a）（b） FIG 1. (a) Schematic diagram of the proposed SIBW with the key design parameters indicated. 𝛬 = 𝑎 + 𝑏 is the period of the photonic crystal structure. 𝑤 = 1.89𝛬 is the width of the line defect in the core region. ℎ < 𝑤 2Τ is the thickness of the dielectric slab. (b) Bandgap diagram and dispersion curves for the vertically polarized modes in the SIBW. Here, the dielectric slab is silicon, 𝑎 = 0.2276𝛬, and 𝑏 = 0.7724𝛬. The orange dashed line is the light line. The green dashed lines indicate the operating band which is wider than one octave. The black region is the bandpass region for the 1D photonic crystal in which the electromagnetic wave can pass through the periodic structure and leak out. The white region above the light line is the bandgap. The red solid line and the purple dotted line represents the dispersion curves for the fundamental HE10 mode and the second-order competing HE20 mode, respectively. The blue curves overlaid upon the black region are cladding modes or lossy defect modes whose electric field are mainly distributed in the bulk crystal. The black dot-dashed line overlaid upon the red curve represents the simulated dispersion curve of the fundamental mode using COMSOL. y x (a) Even mode (b) Odd mode FIG. 2. Mode pattern (Ez) for points A and B in Fig. 1(b). (a) Point A. (𝛽𝑛, 𝜔𝑛) = (0.015, 0.2828) ; (b) Point B. ⁡(𝛽𝑛, 𝜔𝑛) = (0.015, 0.4431) . Periodic boundaries are applied. Here, 𝛽𝑛 = 𝛽𝛬 2𝜋Τ and 𝜔𝑛 = 𝜔𝛬 2𝜋𝑐 Τ . the high-refractive-index component of the photonic crystal in our theoretical analysis. The fundamental mode of the proposed SIBW is vertically polarized HE10 mode, and its electric field is evenly distributed along the 𝑧 -axis. The height of the SIBW, ℎ, does not contribute to the dispersion relations for vertically polarized modes. Therefore, the dispersion relation problem for vertically polarized modes can be simplified into a two-dimensional (2D) problem in the 𝑥 - 𝑦 plane assuming the third dimension is infinite and uniform. To avoid higher-order competing modes, ℎ should be not greater than 𝑤 2Τ . To maximize the first-order photonic bandgap which supports operating HE10 mode, 𝑎 and 𝑏 are chosen based on the following quarter-wavelength condition [29]: 𝜆𝑐 4 , 𝑛𝑎𝑎 = 𝑛𝑏𝑏 = (1) where 𝑛𝑎 and 𝑛𝑏 are the refractive index of the high- and low-refractive-index materials and 𝜆𝑐 is the center wavelength of the first-order bandgap when the longitudinal propagation constant 𝛽 = 0. As we can see from Eqn. (1), by properly choosing a target operating wavelength (𝜆𝑐) and the materials for the high- and low-refractive-index layers, the geometric parameters of the Bragg reflector structure can change accordingly and the operating band can thereby be scaled to any target operating wavelength. The high- resistivity silicon is dispersionless, and the real part of its complex refractive index is 3.4175 ± 0.00005 between 0.5 and 4.5 THz [28]. Thus, in our theoretical design, 𝑛𝑎 = 3.4175 , 𝑛𝑏 = 1 , 𝑎 = 𝛬 = 0.2264𝛬 and 𝑏 = 𝑛𝑏 𝑛𝑎+𝑛𝑏 𝑛𝑎 𝛬 = 0.7736𝛬 . The width of the line defect in the 𝑛𝑎+𝑛𝑏 middle of the Bragg reflector waveguide, 𝑤, greatly affects the dispersion curves of the defect modes, but it barely affects the dispersion curves of the cladding modes. The value of 𝑤 should be carefully chosen to push the dispersion curve of the competing second-order odd mode to the edge of the photonic bandgap, so the competing mode becomes lossy which stops it from being excited in the SIBW. This Τ ( 𝑣𝑔 = 𝜕𝜔 𝜕𝛽Τ FIG. 3. Group-velocity dispersion ( 𝜕2𝛽 𝜕𝜔2 normalized group velocity fundamental mode calculated using MPB and COMSOL. creates the well-known modal-filtering effect [30] and allows the SIBW operates over more than one octave. ) and the ) of Dispersion curves for vertically (z-axis) linearly polarized modes in the line-defect Bragg reflector waveguide overlaid upon the bandgap diagram are shown in Fig. 1(b). The dispersion curves are numerically calculated using the MIT Photonic-Bands (MPB) package [31], while the photonic bandgap is calculated separately and analytically according to Bloch's theorem [32]. The photonic bandgap obeys the following condition, which provides the constraint which stops the EM wave from propagating through into the periodic claddings from the line defect [32]: Re(𝑋𝑠) < 1, (2) where 𝑋𝑠 = [𝑐𝑜𝑠(𝑘𝑏𝑏) − 𝑖 2 𝜉𝑏𝑘𝑏 ( 𝜉𝑎𝑘𝑎 + 𝜉𝑎𝑘𝑎 𝜉𝑏𝑘𝑏 ) 𝑠𝑖𝑛⁡(𝑘𝑏𝑏)] 𝑒𝑥𝑝⁡(−𝑖𝑘𝑎𝑎) (3) is the constant; 𝑘0 = 𝜔 𝑐Τ Here, 𝑘𝑖 = √(𝑛𝑖𝑘0)2 − 𝛽2⁡(𝑖 = 𝑎, 𝑏) lateral propagation vacuum wavenumber; 𝛽 is the longitudinal propagation constant; 𝑛𝑒𝑓𝑓 = 𝛽 𝑘0Τ is the effective refractive index of the mode; 𝜔 is the vacuum angular frequency; 𝑐 is the speed of light in vacuum; and 𝜉𝑖 is 1 or 1 𝑛𝑖 corresponding to 𝑆 = TE or TM modes, respectively. 2⁡(𝑖 = 𝑎, 𝑏) the is Τ MPB calculates both the cladding modes, which are mainly distributed in the bulk photonic crystal and the defect modes which are mainly distributed in the central line defect region. As we can see from Fig. 1(b), the dispersion curves for the cladding modes calculated by MPB are coincident with the bandpass region calculated based on Bloch's theorem, validating both methods. The red solid curve corresponds to the desired fundamental HE10 mode in SIBW, HR-Si Copper (f) h a w (a) b (b) Point A (c) Point B (d) Point C (e) FIG. 4 Structure and characterization of the proposed HCPW. (a) Structure of the SIBW. (b)-(d) Normalized electric field distribution of the desired HE10 mode at points A, B, and C shown in Fig. 4(f). The field linearly decreases from red to blue. (e) Normalized electric field along the red dotted line shown in Fig. 4(a) at selected frequency points. (f) Propagation losses of SIBW and hollow metallic rectangular waveguide. The frequencies of points A, B and C are 1.532 THz, 1.843 THz and 2.154 THz, respectively. In the simulation (COMSOL), 𝑎 = 0.2264𝛬 ≈ 21.7⁡μm, 𝑏 = 0.7736𝛬 ≈ 72⁡μm, 𝑤 = 1.89𝛬 ≈ 176⁡𝜇m, and ℎ = 𝑤 2Τ ≈ 88⁡𝜇m. The width and height of the hollow metallic rectangular waveguide are 164 μm and 82 μm, respectively. while the purple dotted line lying at the edge of the photonic bandgap corresponds to the competing and lossy HE20 mode. Regarding the 𝑦 = 0 plane, the HE10 and HE20 modes correspond to the symmetric even mode and antisymmetric odd mode, respectively. The mode pattern of point A for the fundamental HE10 mode and point B for the lossy HE20 mode in Fig. 1(b) are shown in Fig. 2. As we can see, only the desired fundamental HE10 mode can be tightly confined in the line defect region and propagate along the waveguide, while the HE20 mode lying at the edge of the photonic bandgap is very lossy and attenuates in the waveguide. This results in the modal-filtering effect [30], which is similar to the method we used in a THz Bragg fiber to suppress unwanted competing modes [33]. The frequency range between the two horizontal green dashed lines in Fig. 1(b) gives the operating band of this design. The parts of the dispersion curve for the even mode 𝑙 = 0.295) and above the below the lower dashed line (𝜔𝑛 ℎ = 0.64) are not selected as operating upper dashed line (𝜔𝑛 bands since in these frequencies, the guided mode can be relatively lossy and potentially couple to unwanted cladding modes. Here, 𝛽𝑛 = 𝛽𝛬 2𝜋Τ . The and 𝜔𝑛 = 𝜔𝛬 2𝜋𝑐 Τ ℎ 𝜔𝑛 Τ 𝑙 ≈ 2.17 selected operating band gives a ratio bandwidth of 𝜔𝑛 , which is wider than one octave. The group- velocity dispersion (GVD) and the normalized group velocity of the fundamental HE10 mode are plotted vs. normalized frequency in Fig. 3. Fig. 3 shows that as the frequency increases, the group velocity first increases and then decreases, and the turning point happens to be at 𝜔𝑛 = 0.5442 where the GVD is zero. III. SIMULATION 𝑙 𝑐 𝛬Τ , ⁡𝜔𝑛 By choosing 𝛬 = 93.1⁡μm , the operating band of the above-mentioned SIBW is settled around the frequency ℎ𝑐 𝛬Τ ) = (0.95⁡THz, 2.06⁡THz). The range of (𝜔𝑛 eigenvalue problem of SIBW is solved using COMSOL, which is a commercial package based on the finite-element method. The cross-section of the SIBW is used in COMSOL to find the eigenvalues and eigenmodes, as shown in Fig. 4(a). For simplicity, the conductivity of the copper layer is set as 5.8×107 S/m over the frequency of interest and its surface roughness is not considered in the simulation. The dispersion curve of the fundamental HE10 mode simulated using COMSOL is plotted in Fig. 1(b) using blue circles, and FIG. 5 Scattering Parameters for SIBW when different modes are excited in the hollow metallic rectangular waveguide. The length of the SIBW is 3 mm. the GVD and group velocity can be accordingly calculated based on the simulated dispersion relation, as presented using dot-dashed lines in Fig. 3. It can be seen that the COMSOL simulated results agree very well with the MPB calculation which is based on the plane wave expansion method. The propagation loss of the SIBW is calculated using the following equation: 1 𝜔 5ln10 𝑐⁡ copper hollow Im(𝑛eff), α⁡(dB/cm) = (4) where 𝑛eff is the effective mode index of the operating fundamental mode obtained from the COMSOL simulation. Fig. 4(f) presents the simulated propagation loss of the proposed SIBW, as well as the propagation loss of a standard-size WM-164 rectangular waveguide for comparison [34]. The propagation loss of the proposed SIBW is below 2 dB/cm over the frequency range wider than one octave from 0.982 to 2.133 THz, which is much lower than the losses of planar transmission lines [7- 10], and even lower than that of the standard-size WM-164 hollow copper rectangular waveguide over a broad band wider than the frequency range from 1 to 2 THz. The propagation loss of SIBW consists of Ohmic loss from metal parts, absorption loss from dielectrics, and radiative loss, which are illustrated as the light yellow, purple, and red regions below the total propagation loss line in Fig. 4(f), respectively. It can be seen that the Ohmic loss in the top and bottom copper layers still plays the most significant role. The radiative loss is almost negligible within the operating band, but it increases gradually towards the edges of the operating band, and eventually, the photonic crystal structure loses the ability to confine the electromagnetic wave outside the operating band. The dielectric absorption loss is mainly determined by the absorption coefficient of the dielectric material that constitutes the photonic crystal structure, which is low in this design since the absorption coefficient of high- resistivity silicon is very low over the target frequency range. The dielectric absorption loss is also influenced by the confinement ability of the photonic structure. It can be seen from Fig. 4(f) that, at the high-frequency end of the operating band, the dielectric absorption loss increases with the increase of the radiative loss. This is because when the radiative loss is high, the electromagnetic field penetrates the photonic crystal structure more deeply, and thereby attenuates more significantly by interacting with more dielectric materials. Figs. 4(b) to 4(d) present the field distributions of the desired HE10 mode at several representative frequencies, namely the points A, B and C showed in Fig. 4(f). For the points A and B, the field is mainly confined in the air core and the first period of the photonic crystals structure, while for the point C, which is located at the edge of the operating band, the electromagnetic field penetrates further into the photonic crystal structure. Fig. 4(e) presents the normalized electric field along the red dotted line shown in Fig. 4(a) at the representative frequencies, which quantitatively confirms the above results. Wave propagation in the SIBW was studied using CST Microwave Studio. A hollow metallic rectangular waveguide, whose width (176 μm) and height (88 μm) are equal to those of the defect core of the SIBW, was chosen to feed the SIBW. The single-mode operating frequency range of the hollow metallic rectangular waveguide is between 0.85 THz and 1.70 THz, so at frequencies above 1.70 THz, high order modes, such as TE20, TE01, etc., can be excited by the hollow metallic rectangular waveguide. Ez (V/m) (a) Point A: 1.532 THz Ez (V/m) (b) Point C: 1.843 THz Ez (V/m) (c) Point C: 2.154 THz FIG. 6 The z-component of the electric field of SIBW at different representative frequencies which corresponds to the points A, B, and C in Fig. 4, respectively. The color bar is in logarithmic scale. The scattering parameters for the SIBW when different modes in the hollow metallic rectangular waveguide are excited are shown in Fig. 5. From the S21 frequency response, when TE10 mode is excited we observe that a transmission window wider than an octave centered at about 1.5 THz is supported in the SIBW. The S11 response when the TE10 mode is excited indicates that the impedance mismatch between the TE10 mode in the hollow metallic rectangular waveguide and the HE10 mode in the SIBW is low, since the reflection coefficient is below -14 dB from 1 to 2 THz. When the second-order TE20 mode is excited in the hollow metallic rectangular waveguide at around 1.7 THz, the S21 rises but it is still below -9 dB owing to the aforementioned modal- filtering effect, and therefore the high-order mode is suppressed. Under the high-order mode (TE20) excitation condition, the S11 is high at around 1.7 THz which means the impedance mismatch is large at the waveguide transition, and at around 2 THz, both S11 and S21 are relatively small indicating that the second-order mode is leaky in the SIBW because the electromagnetic field is neither reflected nor transmitted. In addition to the TE20 mode, other high-order modes operate at higher frequencies, which is beyond the frequency range of interest from 1 to 2 THz. The mode patterns of the guided mode in SIBW at several representative frequencies, namely the points A, B, and C shown in Fig. 4, are shown in Fig. 6. It can be seen that the electric field for both frequency point A and point B can be tightly confined in the central defect region, and the operating mode HE10 mode is well preserved along the waveguide for both cases. However, for the frequency point C, the electric field penetrates the cladding layers strongly and the guided mode is no longer the single HE10 mode. Comparing (a) with (b) in Fig. 6, the electric field at the point B is slightly leakier than that at the point A, and thus the propagation loss at the point B is larger than that at the point A, as shown in Fig. 4. It is common for photonic crystals that the confinement loss is usually smaller at the middle of a photonic bandgap than that at the edge [33]. 𝐽𝑠 (A/m) 𝑑ห (A/m2) ห𝐽𝑦 (a) (b) FIG. 7 The current distribution of the SIBW. (a) The magnitude of the surface current at 1.532 THz calculated using CST. (b) The magnitude of the y-component of the displacement current density in the cross-section plane of the SIBW calculated using COMSOL. The distribution of the surface current density in the SIBW at point A is shown in Fig. 7(a). The surface current density is mainly distributed along the central defect region and the first cladding layer, with the remaining small portion being distributed in the outer cladding layers. There is no free current flow between the top and bottom metal plates, in contrast with both rectangular waveguide and substrate integrated waveguide. Instead, there is displacement current in the dielectric structure that closes the current loop, so generating the magnetic field as shown in Fig. 7(b). the hollow metallic IV. SCALED EXPERIMENTAL VERIFICATION To facilitate the fabrication and measurement, a proof-of- concept experimental demonstration of SIBW was performed at millimeter-wave frequencies (ranging from 20 to 45 GHz). Two SIBW samples with different lengths and fed by microstrip and coaxial launchers were fabricated and tested, as shown in Fig. 8. The design principles are the same as those discussed in the principle section. The dielectric material used for the HCPW is Rogers RO3010, which has a dielectric constant of 10.2 and loss tangent of 0.0022 over the frequency range from 8 to 40 GHz. The reason why we use the Rogers RO3010 substrate rather than high-resistivity silicon is that the PCB board is compatible with standard PCB fabrication techniques. The similarities between the RO3010 and the high-resistivity silicon are that they both have relatively high dielectric constant and low absorption loss, which are desirable for constructing the designed photonic crystal structure. In the two samples shown in Fig. 8, both the microstrip feed lines and the photonic crystal structures were machined using laser direct writing technology with the LPKF ProtoLaser U3. The edges of the laser cutting trajectory were slightly burnt and the burnt dielectric material was left on the structure. The substrates for the microstrip and photonic crystal structure are double-sided copper-clad Rogers RT/Duroid® 5880LZ™ ( 𝜀𝑟 = 2 ) and RO3010™ ( 𝜀𝑟 = 10.2 ), respectively. The thicknesses of the dielectric and copper claddings of both substrates were 0.635 mm and 17.5 µm, respectively. The copper claddings of the RO3010™, which are far thinner than the operating wavelength, were not removed from the dielectric in order to provide extra mechanical support for the delicate photonic crystal structure, and should have negligible impact on the dispersion relation of the SIBW, and only very slightly increase the overall propagation loss, as confirmed by our simulations. The top and bottom copper plates are made of C101 copper. The copper plates were first machined with a DMU 40 CNC milling machine and then manually polished Top Copper Plate (Transparent) M1.6 Screw Holes 𝑧 𝑦 𝑥 2.4mm Female End Launch Connectors (Connected to VNA via coaxial cables) 1 mm Microstrip I II SIBW Microstrip III Bottom Copper Plate （a）（b）（c） FIG. 8 CAD design and fabricated samples (before assembling) of the SIBW. (a) CAD design. The SIBW in region II is fed by microstrips in regions I and III. (b) Short SIBW with the length between the two microstrip lines 𝑙 = 46.54 mm. (c) Long SIBW with 𝑙 = 150.56 mm. The inserted picture in (c) shows the zoomed view of the photonic crystal structure. In the design, 𝑎 = 1⁡mm, 𝑏 = 3.28⁡mm, 𝑤 = 8.14⁡mm, and ℎ = 0.671⁡mm by using 500-grade wet-and-dry silicon carbide sandpaper. The bottom plates still had visible residual CNC milling tool marks on the surface, the surface finish of the top plates was smooth. A two-tier calibration technique was used to characterize the SIBW. The first-tier SOLT calibration was performed using a 2.4mm mechanical calibration kit to place the reference planes to the input interface of the end launch connectors. A second-tier calibration used multiline calibration [35] to de-embed using the complex propagation constant of the SIBW. The two SIBW samples were tested using a vector network analyzer and the scattering parameters are shown in Fig. 9(a). The ripples on the S- parameter responses are mainly due to the impedance mismatches at the interfaces of waveguide transitions and inevitable reflection coefficients (S11) of the two SIBWs are lower than -6 dB (<25%) over the frequency of interest between 20 and 45 GHz, which means that a significant amount of the power was fed into the SIBWs. Besides, the magnitude of the S11 parameters for the short and long SIBWs are at the same level, which means that the repeatability of the microstrip to SIBW transitions for both SIBWs are acceptable for multiline calibration. The transmission coefficients (S21) for both SIBWs show wideband transmission windows ranging from 20 to 45 GHz which are wider than one octave. imperfections. The fabrication The complex propagation constant of the SIBW can be extracted from the S-parameters shown in Fig. 9(a) using the multiline calibration technique [35]: 𝛾 = ln(𝜒) , ∆𝑙 (5) where 𝛾 = 𝛼 + 𝑖𝛽 is the complex propagation constant, 𝛼 is the attenuation factor, 𝛽 is the phase factor, ∆𝑙 is the length difference of the two SIBWs, and 𝜒 is the first eigenvalue of the matrix T𝑙(T𝑠)−1. Here, T𝑠 and T𝑙 are the T-parameters of the short and long SIBWs including the waveguide transition parts, respectively, which can be obtained from the measured S-parameters of the two SIBWs. The theoretically analyzed and experimentally measured normalized group velocity (𝑣𝑔 𝑐Τ = 𝜕𝜔 𝜕𝛽Τ /𝑐) of the SIBW can be extracted from the dispersion curve calculated using MPB and the measured phase factor based on the Eqn. (5), as shown in Fig. 9(b). Since small ripples on the spectrum of phase factor owing to the small impedance mismatches at discontinuities of the SIBWs or the calibration errors due to the inconsistencies of the two SIBWs have significant impacts on its first-order derivative (𝑣𝑔), reduced frequency points are used when calculating the group velocity so as to analyze the overall trend of the group velocity. The ripples on the curve of the normalized group velocity, which also makes the normalized group velocity larger than one at several frequencies, attributes to the residual and inevitable small impedance mismatches and calibration errors. The 4th- order polynomial fitted measured curve shows the average trends of the normalized group velocity, which agrees with the theoretically analyzed results according to the dispersion relations of SIBW calculated by using MPB. The reason of using 4th-order polynomial curve fitting to the normalized group velocity is due to the factor that the group-velocity should be one order higher than the GVD which follows the trends of a 3rd-order polynomial fitted curve (as shown in Fig. 3). Both the experimentally measured normalized group velocity rise firstly with increasing frequency and decrease after reaching a maximum. The maximum points for the theoretical and experimental results are 39.73 GHz and 41.05 GHz, respectively, which are close to each other. The turning points in the normalized group velocity respond correspond to the zero GVD point. theoretically analyzed and the To demonstrate the SIBW operates in the desired fundamental HE10 mode over the frequency range from 20 GHz to 45 GHz, we measured the GVD and compared it with （a）（b）（c）（d） FIG. 9 Characterization of the SIBW. (a) Measured S-parameters of the two SIBW samples. One percentage moving window smoothing, a built-in feature of the VNA, is applied to reduce the very high-frequency ripples due to minor impedance mismatches at fabrication imperfections. (b) Normalized group velocity (𝑣𝑔 𝑐Τ ). Fourth-order polynomial curve fitting is applied to fit the measured normalized group velocity. (c) Group-velocity dispersion (𝜕2𝛽 𝜕𝑤2 ). The measurement result is extracted based on the fitted curve for the normalized group velocity shown in Fig. 9(b). (d) Simulated and measured propagation loss. Second-order polynomial curve fitting is applied to fit the measured propagation loss. Τ Τ the theoretical predictions. Fig. 9(c) shows the GVD of the SIBW. The measurement results are calculated using 𝜕2𝛽 𝜕𝑤2 based on the fitted measurement results of group velocity in Fig. 9(b). The theoretically analyzed results are obtained according to the dispersion relations of SIBW calculated by using MPB, which has also been adopted in Fig. 3. It can be seen from Fig. 9(c) that the theoretically analyzed and experimentally measured GVDs are at the same level and follow similar trends. A zero GVD point is observed for both theoretical and experimental results, occurring at 39.73 GHz and 41.05 GHz, respectively. The simulated and measured propagation losses of the SIBW are shown in Fig. 9(d). The measured propagation loss 𝑒 𝛼 according to the Eqn. (5). The is calculated using 20log10 simulated results are obtained based on a COMSOL simulation similar to that used in Fig. 4(f). In general, the the propagation measured and fitted simulated results follow similar trends, namely, losses decrease firstly with increasing frequency and then increases after reaching a minimum. On average, the measured propagation loss of the fabricated SIBW is less than 0.66 dB/cm (compared with 0.15 dB/cm simulated) over the frequency range from 20 to 45 GHz, which is wider than one octave. The measured loss is observed to be greater than the simulated results, which may due to several factors that increase the loss such as surface roughness, oxidation of the copper layer, residual burnt dielectric material, and fabrication or assembly imperfections. At around 42.8 GHz the propagation loss is below zero due to a dip in the response which is a measurement artifact. Comparing with Fig. 9(a), we can see that at the same frequency point, the S21 of the short SIBW declines and the S11 of the short SIBW rises abnormally, while for the long SIBW, it does not behave in the same way at this frequency. This implies that there is a physical inconsistency between the two SIBWs resulting in multiline calibration errors at around 42.8 GHz, which could be avoided by the fabrication and assembly precision. improving In summary, we performed a proof-of-concept experimental verification of the guidance mechanism of the proposed SIBW at millimeter-wave frequencies. We found that the measurement results agree with theoretical analysis, which means that the SIBW can operate in single-HE10- mode fashion over a bandwidth wider than one-octave, with a GVD passing through zero close to the middle of the operating band. Though the measured propagation loss is slightly higher than the simulated one, as presented in Fig. 9(d), they both follow similar trends. The propagation loss can be further improved and a THz design implemented by using more precise fabrication and assembly techniques, such as deep reactive ion etching (RIE) and wafer-bonding techniques. THz time-domain spectroscopy is a useful tool for wideband characterization of such a waveguide, which will be further discussed in our future work. V. CONCLUSIONS This paper presents a novel substrate integrated Bragg waveguide that can operate in single-HE10-mode over more than one octave and in zero group-velocity dispersion fashion around the middle of the operating band. First, we theoretically investigated the SIBW between 1 to 2 THz, and the propagation loss of the proposed SIBW is found to be even lower than that of a classic hollow metallic rectangular waveguide, benefiting from the reduction of Ohmic loss. To facilitate the fabrication and measurement, we performed a proof-of-concept experimental demonstration at millimeter- wave frequencies ranging from 20 to 45 GHz to verify the guidance properties of the proposed SIBW, such as the operating bandwidth, GVD, operating mode, and propagation loss, with consistency between the theoretical analyses (both MPB and COMSOL) and the experimental measurements being observed. Showing many advantages over conventional THz planar waveguides, the proposed SIBW is a promising transmission line for broadband submillimeter-wave and THz systems using functional components or multilayer multichip module technology. ACKNOWLEDGMENTS The author would like to acknowledge Dominic Platt and Lei Shi for their helpful discussion, and also thank Graham Brown and Andrew Pickering from the Engineering and Physical Sciences Research Council (EPSRC) National Facility for Innovative Robotic System for their timely help in fabrication. Besides, the authors also thank financial supports from the National Natural Science Foundation of China (Grant No. 11734012, 51601119, 11574218), the Natural Science Foundation of Shenzhen University (Grant No. 2019006), the EPSRC (Grant No. EP/R00501X/1, EP/P021859/1), and the National Physical Laboratory. ------------------------------- [1] P. U. Jepsen, D. G. Cooke, and M. Koch, Terahertz spectroscopy and imaging -- Modern techniques and applications, Laser Photonics Rev., 5, 124 (2011). [2] H. J. Song and T. Nagatsuma, Present and future of terahertz communications, IEEE Trans. Terahertz Sci. Technol., 1, 256 (2011). [3] D. M. Mittleman, Twenty years of terahertz imaging, Opt. Express, 26, 9417 (2018). [4] S. S. Dhillon, et al. The 2017 terahertz science and technology roadmap, J. Phys. D Appl. Phys., 50, 043001 (2017). [5] C. Kulesa, Terahertz spectroscopy for astronomy: From comets to cosmology, IEEE Trans. Terahertz Sci. Technol., 1, 232 (2011). [6] I. Somjit, and N. Robertson, M. and millimetre- Chongcheawchamnan, Microwave wave design for wireless communications. John Wiley & Sons (2016). [7] H. M. Heiliger, M. Nagel, H. G. Roskos, H. Kurz, F. Schnieder, W. Heinrich, R. Hey, and K. Ploog, Low- dispersion thin-film microstrip lines with cyclotene (benzocyclobutene) as dielectric medium, Appl. Phys. Lett., 70, 2233 (1997). [8] M. Y. Frankel, S. Gupta, J. A. Valdmanis, and G. A. Mourou, Terahertz dispersion characteristics of coplanar transmission lines, IEEE Trans. Terahertz Sci. Technol., 39, 910 (1991). attenuation and [9] D. R. Grischkowsky, Optoelectronic characterization of transmission lines and waveguides by terahertz time- domain spectroscopy, IEEE J. Sel. Top. Quantum Electron., 6, 1122 (2000). [10] J. Cabello-Sánchez, H. Rodilla, V. Drakinskiy, and J. Stake, Transmission Loss in Coplanar Waveguide and Planar Goubau Line between 0.75 THz and 1.1 THz, In 2018 43rd International Conference on Infrared, Millimeter, and Terahertz Waves (IRMMW-THz) (pp. 1-2). IEEE (2018). [11] M. Swithenbank, A. D. Burnett, C. Russell, L. H. Li, A. G. Davies, E. H. Linfield, J. E. Cunningham, and C. D. Wood, On-Chip Terahertz-Frequency Measurements of Liquids, Anal. Chem., 89, 7981 (2017). [12] A. Malekabadi, S. A. Charlebois, D. Deslandes, and F. Boone, High-resistivity ribbon waveguide for single-mode low-loss propagation at F/G-bands, IEEE Trans. Terahertz Sci. Technol., 4, 447 (2014). silicon dielectric [13] N. Ranjkesh, M. Basha, A. Taeb, and S. Safavi-Naeini, Silicon-on-glass dielectric waveguide -- Part II: For THz applications, IEEE Trans. Terahertz Sci. Technol., 5, 280 (2015). on asymmetric Bragg sidewall gratings in a multimode waveguide. Optics letters, 41, 2450 (2016). [14] K. Tsuruda, M. Fujita, and T. Nagatsuma, Extremely low-loss terahertz waveguide based on silicon photonic- crystal slab, Opt. Express, 23, 31977 (2015). [15] H. Amarloo and S. Safavi-Naeini, Terahertz Line Defect Waveguide Silicon-on-Glass Technology, IEEE Trans. Terahertz Sci. Technol., 7, 433 (2017). Based on [16] K. Nielsen, H. K. Rasmussen, A. J. Adam, P. C. Planken, O. Bang, and P. U. Jepsen, Bendable, low-loss Topas fibers for the terahertz frequency range, Opt. Express, 17, 8592 (2009). [17] J. Yang, J. Zhao, C. Gong, H. Tian, L. Sun, P. Chen, L. Lin, and W. Liu, 3D printed low-loss THz waveguide based on Kagome photonic crystal structure, Opt. Express, 24, 22454 (2016). [18] B. Hong, M. Swithenbank, N. Greenall, R. G. Clarke, N. Chudpooti, P. Akkaraekthalin, N. Somjit, J. E. Cunningham, I. D. Robertson, Low-loss asymptotically single-mode THz Bragg fiber fabricated by digital light processing rapid prototyping, IEEE Trans. Terahertz Sci. Technol., 8, 90 (2018). and [19] H. Bao, K. Nielsen, O. Bang, and P. U. Jepsen, Dielectric tube waveguides with absorptive cladding for broadband, low-dispersion and loss THz guiding, Sci. Rep., 5, 7620 (2015). low [20] H. Bao, K. Nielsen, H. K. Rasmussen, P. U. Jepsen, and O. Bang, Fabrication and characterization of porous- core honeycomb bandgap THz fibers, Opt. Express, 20, 29507 (2012). [21] M. Navarro-Cía, J. E. Melzer, J. A. Harrington, and O. Mitrofanov, Silver-coated Teflon tubes for waveguiding at 1 -- 2 THz, J. Infrared Millim. Terahertz Waves, 36, 542 (2015). [22] P. Yeh, A. Yariv, and E. Marom, Theory of Bragg fiber, J. Opt. Soc. Am. A., 68, 1196 (1978). [23] P. Lova, G. Manfredi, L. Boarino, A. Comite, M. Laus, M. Patrini, F. Marabelli, C. Soci, and D. Comoretto, Polymer distributed bragg reflectors for vapor sensing. Acs. Photonics, 2, 537 (2015). [24] H. Qiu, J. Jiang, P. Yu, T. Dai, J. Yang, H. Yu, and X. Jiang, Silicon band-rejection and band-pass filter based [25] A. R. Weily, K. P. Esselle, and B. C. Sanders, Layer- by-layer photonic crystal horn antenna. Physical Review E, 70, 037602 (2004). [26] G. Singh, J. D. B. Bradley, N. Li, E. S. Magden, M. Moresco, T. N. Adam, G. Leake, D. Coolbaugh, and M. R. Watts, Resonant pumped erbium-doped waveguide lasers using distributed Bragg reflector cavities. Optics letters, 41, 1189 (2016). [27] A. Abdolvand, A. M. Walser, M. Ziemienczuk, T. Nguyen, and P. S. T. Russell, Generation of a phase- locked Raman frequency comb in gas-filled hollow-core photonic crystal fiber. Optics letters, 37, 4362 (2012). [28] J. Dai, J. Zhang, W. Zhang, and D. Grischkowsky, Terahertz time-domain spectroscopy characterization of the far-infrared absorption and index of refraction of high-resistivity, float-zone silicon, J. Opt. Soc. Am. B., 21, 1379 (2004). [29] Y. Xu, R. K. Lee, and A. Yariv, Asymptotic analysis of Bragg fibers, Opt. Lett., 25, 1756 (2000). [30] Y. Zhang, and I. D. Robertson, Analysis and design of Bragg fibers using a novel confinement loss diagram approach, J. Light. Technol., 28, 3197 (2010). [31] S. G. Johnson, and J. D. Joannopoulos, Block-iterative frequency-domain methods for Maxwell's equations in a planewave basis, Opt. Express, 8, 173 (2001). [32] A. Kitagawa, and J. I. Sakai, Bloch theorem in cylindrical coordinates and its application to a Bragg fiber, Phys. Rev. A, 80, 033802 (2009). [33] B. Hong, M. Swithenbank, N. Somjit, J. Cunningham, and I. Robertson, Asymptotically single-mode small- core terahertz Bragg fibre with low loss and low dispersion, J. Phys. D Appl. Phys., 50, 045104 (2016). [34] IEEE Std 1785.1-2012, "IEEE Standard for Rectangular for Metallic Waveguides and Their Frequencies of 110 GHz and Above -- Part 1: Frequency Bands and Waveguide Dimensions". Interfaces [35] F. Xu, and K. Wu, Guided-wave and leakage characteristics of substrate integrated waveguide, IEEE Trans. Terahertz Sci. Technol., 53, 66 (2005).
1811.12864	1	1811	2018-11-30T16:14:36	Large enhancement of sensitivity in NiFe/Pt/IrMn-based planar Hall sensors by modifying interface and sensor architecture	[ "physics.app-ph" ]	The planar Hall sensitivity of obliquely deposited NiFe(10)/Pt(tPt) /IrMn(8)/Pt(3) (nm) trilayer structures has been investigated by introducing interfacial modification and altering sensor geometry. The peak-to-peak PHE voltage and AMR ratio of the sensors exhibit an oscillatory increase as a function of Pt thickness. This behaviour was attributed to the strong electron spin-orbit scattering at the NiFe/Pt interface of the trilayers. The temperature-dependent PHE signal profiles reveal that the Pt-inserted PHE sensors are stable even at 390 K with a high signal-to-noise ratio and an increased sensitivity due to reduction of exchange bias. In order to further increase the sensitivity, we have fabricated PHE sensors for a fixed Pt thickness of 8 {\AA} by using sensor architectures of a cross, tilted-cross, one-ring and five-ring junctions. We have obtained a sensitivity of 3.82 {\mu}V/Oe.mA for the cross junction, while it considerably increased to 298.5 {\mu}V/Oe.mA for five-ring sensor geometry. The real-time voltage profile of the PHE sensors demonstrate that the sensor states are very stable under various magnetic fields and sensor output voltages turn back to their initial offset values. This provides a great potential for the NiFe/Pt/IrMn-based planar Hall sensors in many sensing applications.	physics.app-ph	physics	Large enhancement of sensitivity in NiFe/Pt/IrMn-based planar Hall sensors by modifying interface and sensor architecture H. Pişkin* and N. Akdoğan Gebze Technical University, Department of Physics, 41400 Gebze, Kocaeli, Turkey *Corresponding author: [email protected] Keywords: Magnetoresistive sensors, interface effects, planar Hall effect, exchange bias. Abstract The planar Hall sensitivity of obliquely deposited NiFe(10)/Pt(tPt) /IrMn(8)/Pt(3) (nm) trilayer structures has been investigated by introducing interfacial modification and altering sensor geometry. The peak-to-peak PHE voltage (𝛥𝑉𝑃𝐻𝐸) and AMR ratio of the sensors exhibit an oscillatory increase as a function of Pt thickness. This behaviour was attributed to the strong electron spin-orbit scattering at the NiFe/Pt interface of the trilayers. The temperature- dependent PHE signal profiles reveal that the Pt-inserted PHE sensors are stable even at 390 K with a high signal-to-noise ratio and an increased sensitivity due to reduction of exchange bias. In order to further increase the sensitivity, we have fabricated PHE sensors for a fixed Pt thickness of 8 Å by using sensor architectures of a cross, tilted-cross, one-ring and five-ring junctions. We have obtained a sensitivity of 3.82 μV/Oe.mA for the cross junction, while it considerably increased to 298.5 μV/Oe.mA for five-ring sensor geometry. The real-time voltage profile of the PHE sensors demonstrate that the sensor states are very stable under various magnetic fields and sensor output voltages turn back to their initial offset values. This provides a great potential for the NiFe/Pt/IrMn-based planar Hall sensors in many sensing applications. 1. Introduction Among the magnetoresistive sensors, the planar Hall effect (PHE) based systems have attracted much attention due to their high signal to noise ratio, linear response at low magnetic fields, small offset voltage in the absence of magnetic field, low power consumption and insensitiveness to small thermal fluctuations[1]. With these unique properties, the PHE sensors have been investigated for their potential applications in micro-crack detectors[2], high accuracy electronic micro-compasses[3], magnetoresistive biosensors[4], on-chip magnetometers[5], heart beat counters and blood pressure detectors (tactile sensing)[6]. Recent efforts in this field have been concentrated to increase the sensor sensitivity and to integrate them with lab-on-chip platforms. The PHE sensor signal have a quite linear zone that describes the sensor working range. The slope of this linear zone defines the sensor's magnetic field sensitivity which can be manipulated by the geometry and structure of the sensor. Up to now, several sensor geometries such as cross junction[7], tilted cross junction[8], Wheatstone bridges in ring [9,10] and diamond[11,12] shapes have been studied. Furthermore, each layer of the sensor structure has influence on the sensor signal by creating shunt current and/or modifying the anisotropy energy of the sensing layer. In the literature, Ni thin film[13], NiFe/IrMn bilayer [14,15][15] NiFe/X/IrMn trilayer (X: Cu[16], Au[17]) and NiFe/Cu/NiFe/IrMn spin-valve[18,19] systems have extensively been studied. Among these structures, the trilayer system provides the highest sensor sensitivity due to its reduced exchange bias compared to the bilayers and lower shunt current compared to the spin-valves structures [16]. In addition, the type and thickness of the spacer layer in trilayer structures strongly affect the PHE sensor sensitivity. Except one group used Au[17], in all trilayer-based PHE sensors studies thin Cu layer is used as a spacer[12,16,20 -- 22]. Although the Cu spacer magnetically softens the NiFe sensing layer[23][24], it results in a decreased sensor output voltage due to its low resistivity[16]. This negative effect of Cu on the sensor sensitivity becomes dominant especially for thicker spacer layers. In order to avoid these effects, a spacer material with a higher resistivity must be used by optimizing its thickness for PHE sensor applications. In this study, considering that Pt has strong spin -- orbit coupling and much higher resistivity compared to the Cu spacer layer, we have investigated the PHE sensor sensitivity as a function of Pt spacer layer thickness in NiFe/Pt/IrMn trilayers. The sensor elements were deposited by magnetron sputtering and photolithography methods. Magnetization and transport measurements reveal that the sensor sensitivity increases as a function of Pt spacer thickness due to decreased exchange bias in trilayers. Notably, we have observed an oscillatory enhancement of PHE sensor output voltage and AMR ratio by inserting thicker Pt layers. Temperature-dependent measurements indicate that the produced PHE sensors are very stable even at 390 K. Furthermore, we have prepared different sensor geometries to increase the sensor sensitivity for a fixed sensor structure. We have obtained 298.5 μV/Oe.mA sensor sensitivity for five-ring geometry. We have also carried out timescan experiments under different magnetic fields in order to investigate the magnetic bead detection capability and the stability of the produced PHE sensors. 2. Sensor fabrication The PHE sensors with architectures of Hall bar, cross, tilted-cross, one-ring and five-ring were fabricated by the lift-off of a trilayer structure of NiFe(10)/Pt(tPt) /IrMn(8)/Pt(3) (nm) deposited on Si substrates by magnetron sputtering. The nominal thickness of the Pt spacer layer was varied from 0 nm to 1 nm with a step of 1 Å. Ni80Fe20, Pt and Ir22Mn78 layers were sequentially deposited at room temperature by using 10W DC, 10W DC and 20W DC powers, respectively. Since the PHE sensors were fabricated in zero applied field, the easy axis of the magnetization was formed parallel to the current axis by growth-induced magnetic anisotropy. Fig. 1(a) shows a picture of the sample on a printed circuit board (PCB) with a Hall bar geometry and simultaneously grown continuous film. The six-terminal Hall bars have a 25 μm×25 μm active area. Fig. 1(b) presents microscopic images of cross, tilted-cross, one-ring and five-ring junctions. The width of the cross and tilted-cross junctions was 5 μm and the tilt angle was 45°. The outer radii of the ring sensors were kept 150 μm with a ring width of 5μm. Figure 1. (a) A photograph of the sample mounted on a PCB which contains both Hall bar and continuous film parts. (b) Microscopic images of the PHE sensors with the architectures of cross, tilted-cross, one-ring and five- rings. (c) Schematic representation of the current, magnetization and applied field directions, easy axis of the sensors, exchange bias field and the angle () between the magnetization and easy axis. 3. Experimental results Magnetic hysteresis curves, exchange bias and coercive field of the sensors were measured by using a magneto-optical Kerr effect (MOKE) setup which is equipped with a flow cryostat for temperature-dependent experiments[25]. The easy axis hysteresis loops of three samples were shown in Fig. 2(a). When the Pt spacer layer were not inserted into the NiFe/IrMn interface, the values of coercive field (HC) and exchange bias (HEB) are very large due to strong pinning. The insertion of Pt weakens the pinning and shifts the hysteresis loops to lower negative fields. This systematic decrease in both HEB and HC as a function Pt spacer thickness can be seen in Fig. 2(b). It is also important to note that exchange bias does not disappear in trilayer system even with a Pt spacer thickness of 1 nm. We also measured PHE and AMR voltage profiles of the sensors using a Keithley 2002 multimeter. A sensor current of 5 mA was applied for Hall bar geometries, whereas 1mA was applied for all other sensor architectures using a Keithley 2400 sourcemeter. The measurement geometry is presented in Fig. 1(c). The PHE sensitivity of the sensors were calculated from the linear region of the PHE voltage profiles and given in Fig. 2(c) as a function Pt spacer thickness. Fig. 2(c) indicates that the PHE sensitivity of the Pt inserted trilayers were increased from 0.86 μV/Oe.mA to 5.32 μV/Oe.mA due to decrease in HEB. Figure 2. (a) Easy axis hysteresis loops of trilayers taken by MOKE setup for certain thicknesses of Pt spacer layer. (b) HC and HEB values of the samples as a function of Pt thickness. (c) Pt thickness dependence of the PHE sensor sensitivities calculated from the linear regions of PHE voltage profiles. Inset shows the PHE measurement of the Hall bar with a Pt spacer of 1 nm. Solid lines are guide for the eye. We have also plotted peak-to-peak PHE voltage (𝛥𝑉𝑃𝐻𝐸) and AMR ratio of the sensors in Fig. 3 as a function of Pt spacer thickness. In principle, the addition of each conductive layer to the sensor structure reduces the current that passing through the sensing layer by creating a shunting effect. Thus, the 𝛥𝑉𝑃𝐻𝐸 and AMR ratio of trilayers were expected to decrease as a function Pt thickness. On the contrary, we have observed an oscillatory increase of 𝛥𝑉𝑃𝐻𝐸 and AMR ratio in trilayers by addition of Pt spacer. This is attributed to the enhancement of resistivity difference (𝜌∥ − 𝜌⊥) in the sensing layer due to strong electron spin-orbit scattering at the NiFe/Pt interfaces of the trilayers. The similar behaviour of the resistivity difference at the NiFe/Pt interface were also reported in Ref. [26]. However, the oscillatory behaviour of this -300-200-1000-1.0-0.50.00.51.00.00.20.40.60.81.012345(c)(b) tPt = 0 nm tPt = 0.5 nm tPt = 1 nmNormalized MagnetizationMagnetic Field (Oe)(a)0.00.20.40.60.81.020406080100120140160 tPt (nm) HEB tPt (nm)HEB (Oe)102030405060 HCHC (Oe) Sensitivity (mV/Oe.mA)-200-1000100200-0.6-0.4-0.20.00.20.40.6VPHE(mV)Magnetic field (Oe)enhancement observed in the present work requires further studies and analysis to understand its origin. Figure 3. The peak-to-peak PHE voltage (𝛥𝑉𝑃𝐻𝐸) and AMR ratio of the sensors as a function of Pt spacer thickness. Solid lines are guide for the eye. In order to investigate the high temperature characteristics of the PHE sensors, we have performed temperature-dependent measurements for a fixed spacer layer thickness of 3 Å. Fig. 4 shows the PHE sensitivity of the sensor as a function of temperature scanned from 300 K to 390 K with a step of 10 K. It is clear from the data that the sensitivity was enhanced from 2.0 μV/Oe.mA to 5.1 μV/Oe.mA due to the reduced exchange bias as a function of temperature. It is worthy of note that the sensor is still very stable even at 390 K with a good sensitivity and high signal-to-noise ratio. The PHE voltage profiles given in the inset of Fig. 4 indicate that the peak-to-peak PHE voltage value (𝛥𝑉𝑃𝐻𝐸) decreases at high temperatures due to decrease of the resistivity difference (𝜌∥ − 𝜌⊥). However, the maximum of the PHE voltage shifts to the lower magnetic fields as a function of temperature. This shift results in an increase in the slope of the PHE signals and it is attributed to the decrease of the HEB at high temperatures. 0.00.20.40.60.81.00.400.440.480.520.56 AMR Ratio VPHE (mV)tPt (nm)0.900.951.001.051.101.15 VPHEAMR Ratio (%) Figure 4. Temperature-dependence of the PHE sensor sensitivity for a trilayer structure with a Pt spacer of 3 Å. Solid line is a guide for the eye. Inset presents PHE voltage profiles taken at different temperatures from 300 K to 390 K. Furthermore, we have investigated the geometry dependence of the sensitivity in Pt inserted PHE sensors for a fixed Pt thickness of 8 Å. The sensor output signals of four different geometries were given in Fig. 5. We have observed a sensitivity of 3.82 μV/Oe.mA for the cross junction, and it slightly increased to 5.80 μV/Oe.mA for the tilted-cross architecture. However, we have recorded huge increase in the sensitivities of the ring junctions. The obtained sensor sensitivities were 96.08 μV/Oe.mA and 298.46 μV/Oe.mA for one-ring and five-ring junctions, respectively. Indeed, it was expected to observe higher sensitivities in tilted-cross and ring junctions compared to the cross junction for a fixed sensor structure. Because, the AMR and Wheatstone bridge contributions were added to the conventional PHE voltage in the tilted-cross and ring geometries [8,10]. These contributions did not change the position of the maximum PHE signal in the field axis for the ring sensors. However, the peak-to-peak PHE voltages (𝛥𝑉𝑃𝐻𝐸) were increased. Thus, the slope of the linear zone, namely the sensitivity, was increased. To further increase the PHE sensitivity, the number of the rings can be increased or the diamond shaped PHE sensors can be fabricated[10]. -200-1000100200-0.4-0.20.00.20.40.60.83003103203303403503603703803902.02.53.03.54.04.55.05.5VPHE (mV)Magnetic Field (Oe)T=300 KSensitivity (mV/Oe.mA)Temperature (K)tPt = 0.3 nmT=390 K Figure 5. PHE sensor signals for the cross (a), tilted-cross (b), one-ring (c) and five-ring (d) junctions. All sensors were simultaneously grown on a Si substrate with a trilayer structure of NiFe(10)/Pt(0.8)/IrMn(8)/Pt(3) (nm). In addition, the stability of the sensor states under different magnetic fields and the reproducibility of the sensor signal are crucial for real applications of the PHE sensors. With this motivation, we have carried out time-profile experiments presented in Figure 6. Firstly, the sensor output voltage was recorded for one minute under zero magnetic field. Then, a constant in-plane magnetic field was applied perpendicular to the current line for one minute. This process was repeated three times to see whether the sensor voltage goes back to the initial offset value. Subsequently, we have increased the magnetic field from 4 Oe to 20 Oe with a step of 4 Oe and repeated the same measurement cycle. The real-time profiles of the PHE sensors indicate that the sensor output voltages completely turned to their initial values, except with a small offset in one-ring sensor presented by blue colour. We have also observed very stable sensor states at different magnetic fields. -400-2000200400-0.2-0.10.00.10.2-400-200020040095.495.595.695.795.895.996.0-400-2000200400-2-1012-400-2000200400-6-4-20246(c)(a)(b)Cross junctionS= 3.82 mV / Oe.mAVPHE (mV)Magnetic Field (Oe)(d)Tilted cross junctionS= 5.80 mV / Oe.mAVPHE (mV)Magnetic Field (Oe)Ring junctionS= 96.08 mV / Oe.mAVPHE (mV)Magnetic Field (Oe)Multi ring junctionS= 298.46 mV / Oe.mAVPHE (mV)Magnetic Field (Oe)Figure 6. Real-time profile measurements of different PHE sensors recorded at 300 K. The figure at the bottom shows a magnified version of the real-time signals for a magnetic field of 4 Oe. 4. Discussion The origin of the hysteretic-like behaviour observed in the PHE signals was attributed to an incoherent magnetization rotation during the positive and negative field sweepings[17,27]. In order to further understand this behaviour, we have performed PHE, perpendicular AMR and L-MOKE measurements at the hard axis geometry for a sample of 5 Å Pt inserted cross junction. Figure 7 shows that the coercive fields of the hard axis L-MOKE curve and the peak positions of the perpendicular AMR signal overlap with the hysteresis of the PHE signal. During the fabrication of the PHE sensors, the external magnetic field was zero. Thus, the growth induced anisotropy was achieved by oblique deposition. However, the data presented in Fig. 7 indicates that few portions of the exchange-biased regions slightly deviate from the main easy axis of the system during the magnetization reversal process. The observed remanent magnetization in the hard axis L-MOKE curve support this conclusion. 250500750100012501500012345501001502002503003500.00.20.40.60.81.01.21.4H=20 Oe H=16 Oe H=12 Oe H=8 Oe 5 Ring, tPt = 0.8 nm, 1 mA 1 Ring, tPt = 0.8 nm, 1 mA Cross, tPt = 1.0 nm, 5 mA Cross, tPt = 0.5 nm, 5 mAVPHE (mV)Time (Sec.)H=4 Oe H=4 Oe H=4 Oe H=4 Oe VPHE (mV)Time (Sec.)Figure 7. The PHE, perpendicular AMR and hard axis L-MOKE measurements taken from a cross junction with a Pt spacer thickness of 5 Å. 5. Conclusion We have performed a systematic study of PHE sensor sensitivity in NiFe/Pt/IrMn-based trilayer structures as a function of Pt spacer thickness and sensor geometry. The PHE voltage profile measurements indicate that the sensitivity of the sensors with a cross junction architecture gradually increased up to 5.32 μV/Oe.mA by inserting thicker Pt spacers. It is worthy of note that the peak-to-peak PHE voltage (𝛥𝑉𝑃𝐻𝐸) and AMR ratio of the sensors were increased with an oscillatory behaviour as a function of Pt thickness. Temperature-dependent experiments reveal that the PHE sensors kept working even at 390 K with a good stability, high sensitivity and high signal-to-noise ratio. In addition, we have obtained very large PHE sensitivity of 298.5 μV/Oe.mA for a five-ring sensor architecture. The real-time profile measurements indicate that the PHE sensor output voltages were reproducible and reversible at different field conditions. This makes NiFe/Pt/IrMn-based PHE sensors very promising for practical detection of low magnetic fields. Acknowledgements This work was supported by TÜBİTAK (The Scientific and Technological Research Council of Turkey) through the project number 116F083. -600-400-2000200400600L - MOKEAMR (H^i)Normalized Signal (arb. units)Magnetic Field (Oe)PHE (H^i)tPt = 0.5 nmT = 300 KReferences [1] A. Schuhl, F.N. Van Dau, J.R. Childress, Low‐field magnetic sensors based on the planar Hall effect, Appl. Phys. Lett. 66 (1995) 2751 -- 2753. doi:10.1063/1.113697. [2] H.Q. Pham, B.V. Tran, D.T. Doan, V.S. Le, Q.N. Pham, K. Kim, Highly Sensitive Planar Hall Magnetoresistive Sensor for Magnetic Flux Leakage Pipeline Inspection, (2018) 1 -- 5. doi:10.1109/TMAG.2018.2816075. [3] [4] F. Montaigne, A. Schuhl, F.N. Van Dau, A. Encinas, Development of magnetoresistive sensors based on planar Hall effect for applications to microcompass, Sensors Actuators, A Phys. 81 (2000) 324 -- 327. doi:10.1016/S0924-4247(99)00102-8. F.W. Østerberg, G. Rizzi, M. Donolato, R.S. Bejhed, A. Mezger, M. Strömberg, M. Nilsson, M. Strømme, P. Svedlindh, M.F. Hansen, On-Chip Detection of Rolling Circle Amplifi ed DNA Molecules from Bacillus Globigii Spores and Vibrio Cholerae, (2014) 2877 -- 2882. doi:10.1002/smll.201303325. [5] K.W. Kim, V. Reddy, S.R. Torati, X.H. Hu, A. Sandhu, C.G. Kim, On-chip magnetometer for characterization of superparamagnetic nanoparticles, Lab Chip. 15 (2015) 696 -- 703. doi:10.1039/C4LC01076K. [6] S. Oh, Y. Jung, S. Kim, S. Kim, X. Hu, H. Lim, C. Kim, Remote tactile sensing system integrated with magnetic synapse, Sci. Rep. 7 (2017) 1 -- 8. doi:10.1038/s41598-017- 17277-2. [7] C. Kim, N.T. Thanh, L.T. Tu, N.D. Ha, C.O. Kim, K.H. Shin, B. Parvatheeswara Rao, Thickness dependence of parallel and perpendicular anisotropic resistivity in Ta/NiFe/IrMn/Ta multilayer studied by anisotropic magnetoresistance and planar Hall effect, J. Appl. Phys. 101 (2007). doi:10.1063/1.2435816. [8] [9] T.Q. Hung, J.R. Jeong, D.Y. Kim, N.H. Duc, C. Kim, Hybrid planar Hall- magnetoresistance sensor based on tilted cross-junction, J. Phys. D. Appl. Phys. 42 (2009). doi:10.1088/0022-3727/42/5/055007. S. Oh, P.B. Patil, T.Q. Hung, B. Lim, M. Takahashi, D.Y. Kim, C. Kim, Hybrid AMR/PHR ring sensor, Solid State Commun. 151 (2011) 1248 -- 1251. doi:10.1016/j.ssc.2011.05.049. [10] A.D. Henriksen, G. Rizzi, M.F. Hansen, Experimental comparison of ring and diamond shaped planar Hall effect bridge magnetic field sensors, J. Appl. Phys. 118 (2015). doi:10.1063/1.4930068. [11] A.D. Henriksen, B.T. Dalslet, D.H. Skieller, K.H. Lee, F. Okkels, M.F. Hansen, Planar Hall effect bridge magnetic field sensors, Appl. Phys. Lett. 97 (2010) 2008 -- 2011. doi:10.1063/1.3460290. [12] M.F. Hansen, G. Rizzi, Exchange-Biased AMR Bridges for Magnetic Field Sensing and Biosensing, IEEE Trans. Magn. 53 (2017). doi:10.1109/TMAG.2016.2614012. [13] L. Ejsing, M.F. Hansen, A.K. Menon, Planar Hall Effect Magnetic Sensor for Micro- Bead Detection, (2014) 1095 -- 1098. [14] A.D. Henriksen, G. Rizzi, F.W. Østerberg, M.F. Hansen, Optimization of magnetoresistive sensor current for on-chip magnetic bead detection using the sensor self-field, J. Magn. Magn. Mater. 380 (2015) 209 -- 214. doi:10.1016/j.jmmm.2014.09.056. [15] L. Ejsing, M.F. Hansen, A.K. Menon, H.A. Ferreira, D.L. Graham, P.P. Freitas, Magnetic microbead detection using the planar Hall effect, J. Magn. Magn. Mater. 293 (2005) 677 -- 684. doi:10.1016/j.jmmm.2005.02.071. [16] T.Q. Hung, S. oh, B. Sinha, J.R. Jeong, D.Y. Kim, C. Kim, High field-sensitivity planar Hall sensor based on NiFe/Cu/IrMn trilayer structure, J. Appl. Phys. 107 (2010) 2008 -- 2011. doi:10.1063/1.3337739. [17] X.J. Li, C. Feng, X. Chen, J.Y. Zhang, Y.W. Liu, S.L. Jiang, Y. Liu, M.H. Li, G.H. Yu, Enhanced planar hall sensitivity with better thermal stability by introducing interfacial modification of Au spacer, J. Magn. Magn. Mater. 381 (2015) 386 -- 389. doi:10.1016/j.jmmm.2015.01.022. [18] B.D. Tu, L.V. Cuong, T.Q. Hung, D.T.H. Giang, T.M. Danh, N.H. Duc, C. Kim, Optimization of spin-valve structure NiFe/Cu/NiFe/IrMn for planar hall effect based biochips, IEEE Trans. Magn. 45 (2009) 2378 -- 2382. doi:10.1109/TMAG.2009.2018580. [19] S. Oh, M. Jadhav, J. Lim, V. Reddy, C.G. Kim, An organic substrate based magnetoresistive sensor for rapid bacteria detection, Biosens. Bioelectron. 41 (2013) 758 -- 763. doi:10.1016/j.bios.2012.09.069. [20] A.D. Henriksen, G. Rizzi, M.F. Hansen, A.D. Henriksen, G. Rizzi, M. Fougt, Planar Hall effect bridge sensors with NiFe / Cu / IrMn stack optimized for self-field magnetic bead detection, J. Appl. Phys. 119 (2016). doi:10.1063/1.4943033. [21] Z.D. Zhao, M.H. Li, P. Kang, C.J. Zhao, J.Y. Zhang, L.J. Zhou, Y.C. Zhao, S.L. Jiang, G.H. Yu, The influence of ultrathin Cu interlayer in NiFe/IrMn interface on rotation of the magnetic moments, Appl. Surf. Sci. 332 (2015) 710 -- 715. doi:10.1016/j.apsusc.2015.01.111. [22] T.Q. Hung, F. Terki, S. Kamara, K. Kim, S. Charar, C. Kim, Planar Hall ring sensor for ultra-low magnetic moment sensing, J. Appl. Phys. 117 (2015). doi:10.1063/1.4918534. [23] S. Mizukami, Y. Ando, T. Miyazaki, The Study on Ferromagnetic Resonance Linewidth for NM/80NiFe/NM (NM=Cu, Ta, Pd and Pt) Films, Jpn. J. Appl. Phys. 40 (2001) 580 -- 585. doi:10.1143/JJAP.40.580. [24] S. Mizukami, Y. Ando, T. Miyazaki, Effect of spin diffusion on Gilbert damping for a very thin permalloy layer in Cu/permalloy/Cu/Pt films, Phys. Rev. B - Condens. Matter Mater. Phys. 66 (2002) 1044131 -- 1044139. doi:10.1103/PhysRevB.66.104413. [25] E. Demirci, Investigation of perpendicular exchange bias effect in magnetoelectric Cr2O3 based thin films, PhD Thesis, Gebze Technical University, 2016. [26] Y.F. Liu, J.W. Cai, L. Sun, Large enhancement of anisotropic magnetoresistance and thermal stability in Ta/NiFe/Ta trilayers with interfacial Pt addition, Appl. Phys. Lett. 96 (2010). doi:10.1063/1.3334720. [27] B. Sinha, T. Quang Hung, T. Sri Ramulu, S. Oh, K. Kim, D.Y. Kim, F. Terki, C. Kim, Planar Hall resistance ring sensor based on NiFe/Cu/IrMn trilayer structure, J. Appl. Phys. 113 (2013). doi:10.1063/1.4790139.
1709.01898	1	1709	2017-09-06T17:06:45	Room temperature 9 $\mu$m photodetectors and GHz heterodyne receivers	[ "physics.app-ph", "physics.ins-det", "physics.optics" ]	Room temperature operation is mandatory for any optoelectronics technology which aims to provide low-cost compact systems for widespread applications. In recent years, an important technological effort in this direction has been made in bolometric detection for thermal imaging$^1$, which has delivered relatively high sensitivity and video rate performance ($\sim$ 60 Hz). However, room temperature operation is still beyond reach for semiconductor photodetectors in the 8-12 $\mu$m wavelength band$^2$, and all developments for applications such as imaging, environmental remote sensing and laser-based free-space communication$^{3-5}$ have therefore had to be realised at low temperatures. For these devices, high sensitivity and high speed have never been compatible with high temperature operation$^{6, 7}$. Here, we show that a 9 $\mu$m quantum well infrared photodetector$^8$, implemented in a metamaterial made of subwavelength metallic resonators$^{9-12}$, has strongly enhanced performances up to room temperature. This occurs because the photonic collection area is increased with respect to the electrical area for each resonator, thus significantly reducing the dark current of the device$^{13}$. Furthermore, we show that our photonic architecture overcomes intrinsic limitations of the material, such as the drop of the electronic drift velocity with temperature$^{14, 15}$, which constrains conventional geometries at cryogenic operation$^6$. Finally, the reduced physical area of the device and its increased responsivity allows us, for the first time, to take advantage of the intrinsic high frequency response of the quantum detector$^7$ at room temperature. By beating two quantum cascade lasers$^{16}$ we have measured the heterodyne signal at high frequencies up to 4 GHz.	physics.app-ph	physics	Room temperature 9m photodetectors and GHz heterodyne receivers  Daniele Palaferri1, Yanko Todorov1, Azzurra Bigioli1, Alireza Mottaghizadeh1, Djamal Gacemi1, Allegra Calabrese1, Angela Vasanelli1, Lianhe Li2, A. Giles Davies2, Edmund H. Linfield2, Filippos Kapsalidis3, Mattias Beck3, Jérôme Faist3 and Carlo Sirtori1 1 Laboratoire Matériaux et Phénomènes Quantiques, Université Paris Diderot, Sorbonne Paris Cité, CNRS- UMS 7162, 75013 Paris, France 2 School of Electronic and Electrical Engineering, University of Leeds, Leeds LS2 9JT, United Kingdom 3 ETH Zurich, Institute of Quantum Electronics, Auguste-Piccard-Hof 1, Zurich 8093, Switzerland Room temperature operation is mandatory for any optoelectronics technology which aims to provide low-cost compact systems for widespread applications. In recent years, an important technological effort in this direction has been made in bolometric detection for thermal imaging1, which has delivered relatively high sensitivity and video rate performance (60 Hz). However, room temperature operation is still beyond reach for semiconductor photodetectors in the 8–12 µm wavelength band2, and all developments for applications such as imaging, environmental remote sensing and laser-based free-space communication3-5 have therefore had to be realised at low temperatures. For these devices, high sensitivity and high speed have never been compatible with high temperature operation6, 7. Here, we show that a 9 µm quantum well infrared photodetector8, implemented in a metamaterial made of subwavelength metallic resonators9-12, has strongly enhanced performances up to room temperature. This occurs because the photonic collection area is increased with respect to the electrical area for each resonator, thus significantly reducing the dark current of the device13. Furthermore, we show that our photonic architecture overcomes intrinsic limitations of the material, such as the drop of the electronic drift velocity with temperature14, 15, which constrains conventional geometries at cryogenic operation6. Finally, the reduced physical area of the device and its increased responsivity allows us, for the first time, to take advantage of the intrinsic high frequency response of the quantum detector7 at room temperature. By beating two quantum cascade lasers16 we have measured the heterodyne signal at high frequencies up to 4 GHz. 1 An important intrinsic property of inter-subband (ISB) quantum well infrared photodetectors (QWIPs) based on III-V semiconductor materials that has not yet been exploited is the very short lifetime of the excited carriers. The typical lifetime is of the order of one picosecond7, which leads to two important consequences: the detector frequency response can reach up to 100 GHz, and the saturation intensity is extremely high (107 W/cm2)17. These figures are ideal for a heterodyne detection scheme where a powerful local oscillator (LO) can drive a strong photocurrent, higher than the detector dark current, that can coherently mix with a signal shifted in frequency with respect to the LO. Notably, these unique properties are unobtainable in infrared inter-band detectors based on mercury-cadmium-telluride (MCT) alloys, which have a much longer carrier lifetime and therefore an intrinsic lower speed response2,18. Yet, the performance of all photonic detectors is limited by the high dark current that originates from thermal emission of electrons from the wells, and rises exponentially with temperature, thus imposing cryogenic operation ( 80 K) for high sensitivity measurements. Although highly doped (1x1012 cm-2) 10 µm QWIPs have been observed to operate up room temperature, tens of mW incident power from a CO2 laser was required to measure the signal19,20. In the present work, we show that this intrinsic limitation in QWIP detectors can be overcome through use of a photonic metamaterial. We are able to calibrate our detector at room temperature using a black body emitting only hundreds of nW, orders of magnitude smaller than that required previously. To date, room temperature performance with values comparable to those that we report here has only been demonstrated in the 3–5 µm wavelength range, using quantum cascade detectors (QCDs)21-23 and MCT standard detectors24. The photonic metamaterial structure is shown in Fig. 1a. The GaAs/AlGaAs QWIP8 contains Nqw = 5 quantum wells absorbing at 8.9 µm wavelength (139 meV) that has been designed according to an optimized bound-to-continuum structure from ref. 7. The absorbing region is inserted in an array of double-metal patch resonators9-12, which provide sub-wavelength electric field confinement and act as antennas. The resonant wavelength is fixed by the patch size s through the expression = 2sneff, where neff = 3.3 is the effective index9. As shown in the reflectivity measurement in Fig. 1b, the cavity mode is in close resonance with the peak responsivity of the detector. In our structure, the microcavity increases the device responsivity by a local field enhancement in the thin semiconductor absorber10, while the antenna effect extends the photon collection area of the detector, Acoll, making it much larger than the electrical area  = s2 of the device13. As the detector photocurrent is proportional to Acoll, while the dark current is proportional to , for the same number of collected photons there is therefore a substantial reduction of the dark current that results in a net increase of the detector operating temperature. 2 Besides the collection area Acoll, which defines the absorption cross section per patch resonator, another crucial parameter is the contrast C of the reflectivity resonance shown in Fig. 1b. This parameter quantifies the fraction of the incident photon flux absorbed collectively by the array and should be as close as possible to 1. As shown in Fig. 1c, the contrast can be adjusted by changing the array periodicity p10. Optimal detector responsivity is obtained at the critical coupling point, C = 1, where all incident radiation is coupled into the array. The collection area per patch is related to the contrast according to the expression Acoll = Cp², where the factor = 0.7 takes into account the polarizing effect of the connecting wires (see S.I.) 13. From the data in Fig. 1c, the critical coupling is obtained with a period p = 3.3 µm, which corresponds to a collection area Acoll = 7.5 µm², four times larger than the electrical area = 1.7 µm² of the patch. Fig. 1 Device concept: a, Double-metal antenna-coupled microcavity realized by e-beam lithography with ohmic alloy (PdGeTiAu) contacts and connected by 150 nm thin wires isolated by Schottky barriers (Ti/Au). The active region contains a QWIP structure (386 nm) with five QWs Si-doped at n=7×1011 cm-2 . The pixel size of the array is 50 µm. This metamaterial structure allows photons to be collected from a collection area Acoll that is much larger than the electrical surface area σ. b, reflectivity spectrum (blue curve) of an array with s=1.30 µm and p=3.30 µm polarized along the TM100 mode (perpendicular to the thin wires): the dashed line is a Lorentzian fit yielding a contrast C = 1–Rdip ~ 90%. The size of the patch cavity has been chosen to resonate with the ISB electronic transition E12 ~ 139 meV. c, contrast and collection area Acoll=CΣ as function of the array unit cell area Σ=p2. The collection area saturates for large unit cell periods, as expected from the theory in Ref. 13. The contrast is optimal for arrays with a period p =3.3 µm. Notice that the device processing has been optimized in order to generate current solely under the metallic square patches and not below the 150 nm wide leads connecting them. To this end we have realised ohmic contacts between the patches and the underlying semiconductor layers using PdGeTiAu annealed alloy, while a Schottky barrier, made by depositing TiAu, prevents 3 vertical current between the metallic wire and the semiconductor. Moreover, all cavities are connected to an external wire-bonding pad insulated by an 800-nm-thick Si3N4 layer (S.I.). Thanks to all these precautions the conductive area is reduced to the sum of the areas of all the patch resonators, which prevents additional dark current from flowing across the device. Fig. 2 Responsivities & Spectra. a, peak responsivity of QWIP devices fabricated in 200 µm diameter mesa (circles), and into patch resonator arrays with s=1.35 µm (squares) and s=1.30 µm (triangles), both with pixel size of 50 µm. The responsivities were measured with a calibrated 1000°C blackbody source as a function of detector heatsink temperature. b, normalized photocurrent spectra of the array with s=1.30 µm at 78 K, 200 K and 295 K. c, photoconductive gain and electronic drift of the three devices presented in 2a as a function of temperature: the data are shown at a bias voltage 0.5 V corresponding to a field of 21 kV/cm. The drift velocity is obtained using a QW capture time of 5 ps (see ref. 7 and S.I.). In order to quantify the detector performance, we have compared the detector array with a reference device, here referred to as "mesa", where the same absorbing region is processed into 200 µm diameter circular mesa and light is coupled in through the 45°-polished substrate edge25. The mesa reference provides the intrinsic photo-response of the detector (see S. I.). In Fig. 2a we compare the peak responsivities for the two configurations, obtained with a calibrated black body source at 1000°C (see Methods for more details). The mesa device could be characterized only up to 150 K, as the photo-current becomes undetectable at higher temperatures. The array detectors show a seven-fold enhancement of the responsivity at low temperatures. Most remarkably, the responsivity could be characterized up to room temperature, where the measured responsivity (0.2 A/W) is comparable with the best responsivity for the mesa device measured at around 50 K. We were thus able to record photo- current spectra up to room temperature, Fig. 2b, which is, to our knowledge, the first type of such measurement with a QWIP operating in the 9 µm band using a thermal source. 4 identical By quantifying carefully the number of photons absorbed in each geometry (Methods), we were also able to extract the photoconductive gain g for each structure (Fig. 2c). We recall that the gain provides the number of electrons circulating per photon absorbed in the QWs26,27, and is an intrinsic property of the detector absorbing region. All our devices show the same values of the gain as a function of temperature, irrespective of their fabrication geometry, which proves that the material properties are for the two structures. Following Ref.7, the photoconductive gain is proportional to the electron drift velocity in the AlGaAs barriers and its temperature dependence is linked to microscopic scattering processes in polar materials14,15. Our results fit well the temperature dependence of the drift velocity described on ref. 14. The derived low temperature value of the drift velocity is of the order of 6×106 cm/s as expected at an electric field of 20 kV/cm for an Al concentration in the range 20–30%28. These results account for the temperature drop of the responsivity observed in Fig. 2a. Above 200 K, the gain acquires an almost constant value g = 0.25 – 0.2, of the order of 1/Nqw. This implies that photoexcited electrons can only travel from one well to the next adjacent well, as the mean free path of the electrons is now shorter than the distance between two wells. Very interestingly, in this limit, it clearly appears that a detector based on a single QW would be advantageous at high temperatures. These results illustrate how our devices give access to the high temperature physics of quantum detectors, a unique regime unexplored so far. The best assessment of detector performance is the background-limited specific detectivity26 ∗ = 𝐷𝐵 𝑅√𝐴𝑑𝑒𝑡 √4𝑒𝑔𝐼 plotted in Fig. 3a for the mesa reference and for the patch devices. The experimental results are compared with our model that describes the impact of the photonic design on the detectivity as a function of the temperature13. For clarity, in Fig. 3b we provide the ratio between the detectivities. At low temperature, we observe an enhancement of only a factor of two. Here, the dark current is negligible and the main source of noise is the background photocurrent induced by the 300 K black body of the environment. In this regime higher responsivity means also higher background noise, and the detectivity enhancement scales with the square root of the responsivities ratio i.e. (Rarray/Rmesa)1/2= 2.6. The situation is totally different at high temperature, where the dark current is the dominant contribution to the noise. In this case the detectivity enhancement is Rarray/Rmesa (Acoll/)1/2 14, and the actual performance of the arrays at 300 K is equivalent to the performance of the mesa reference at 150 K, doubling the temperature of operation. This is a significant improvement, well beyond that is predictable from the low temperature operation. Our device concept therefore takes advantage of both the responsivity enhancement, and the strong suppression of the dark current owing to the antenna effect, as expressed by the factor (Acoll/)1/2. As explained in Ref. 13, the combination of the microcavity and the antenna effect thus slows down the 5 decrease of the detectivity with temperature, pushing the detector operation to much higher temperatures than expected. Fig. 3 Background-limited detectivity. a, background-limited specific detectivity (2π field of view) as a function of the temperature, in the case of the mesa (circles) and the arrays with s=1.30 µm (triangles) and s=1.35 µm (squares) at 0.5V. The red is a fit of the mesa experimental data to d(T)=d0/[1+d1Texp(-Eact/kBT)]1/2 where d0 and d0 are fit parameters, Eact = 120meV is the activation energy and kB is the Boltzmann constant. The blue curve is the model described in ref. 13, expressing the performances of quantum detectors embedded in patch resonators. b, ratio between the detectivities in mesa =70 K the two different detector geometries. Dots show the corresponding BLIP temperatures: TBLIP for the mesa and TBLIP cavity =83 K for the patch cavity arrays. line By exploiting our photonic concepts we have achieved high temperature operation with relative high sensitivities. We now seek to benefit from the inherent very high frequency response together with the reduced electrical capacitance of our devices in order to use them as heterodyne receivers. In this case, by increasing the power of the local oscillator one may achieve the ultimate heterodyne sensitivity set only by the detector absorption coefficient. The heterodyne scheme allows the very fast frequency response of the QWIP detectors at room 6 temperature, as well as their very high saturation intensity, to be fully exploited, paving the way to a new class of coherent quantum devices in the mid- and far-infrared spectral ranges. Fig. 4 Tunable heterodyne experiment and results. a, heterodyne arrangement involving two DFB QCLs and the room temperature QWIP in the cavity array geometry. A 40 dB heterodyne power spectrum is shown, acquired with a 1MHz resolution bandwidth of the spectrum analyzer. b, normalized heterodyne power signal (in linear scale) acquired using a spectrum analyzer. By slightly changing the (see S.I.). c, log-log plot of the signal- current on the DFB lasers we can tune the positions of ωS and ωLO to-noise ratio as function of the QCL power. The noise of the QWIP is calculated using the measured gain and dark current values at room temperature. This realization is depicted in Fig. 4a, where we show schematically the heterodyne arrangement that we used to probe our detector at room temperature. It consists of two DFB quantum cascade lasers (QCLs)16 operating at = 8.36 m that are made collinear by a beam splitter before they impinge on the detector. The latter is connected via wire bonding to a high frequency coaxial cable that is connected to a spectrum analyser. Each laser has a linewidth of the order of one MHz when current and temperature are stabilised. In order to ensure that their individual frequencies are very close, one laser is kept at 293 K, while the second is kept at 254 K. When the detector is illuminated by both lasers a clear heterodyne signal appears on the spectrum analyser. In Fig. 4a we show a measurement at 1.06 GHz, with a 40 dB signal-to-noise ratio. We have measured heterodyne signals up to 4.2 GHz as it is illustrated in Fig. 4b. Our bandwidth is 7 presently limited by a strong impendence mismatch between the detector and the external circuit. In Fig. 4c we report our first characterisation of the sensitivity of the heterodyne receiver at room temperature. The blue dots correspond to the D.C. saturation curve for the LO, while the orange curve is the heterodyne signal at 1 GHz as a function of the signal power. The straight line is a linear fit for the LO saturation curve. The saturation experiment shows that the detector responds linearly up to 78 mW ( 3.1 kW/cm2) of incident power. Moreover, the linear fit intercepts the 1 Hz integration band for a power of 0.5 nW, in very good agreement with the measured detectivity at room temperature. As can be observed from Fig. 4c, the heterodyne data are very well fitted with a square root dependence (dashed line) and can reach a signal-to-noise ratio of a few pW, for an integration time of the order of 10 ms. This clearly shows the strength of the heterodyne technique that let us envision sensitivity in the thermal region at  = 9 m which is unreachable with any other technique at room temperature. It has to be mentioned that the power of the LO in our experiment is still far from being the source of the highest current circulating into the device. We are in fact dominated by the dark current of the detector, Idark  3.5 mA, while the local oscillator photocurrent is ILO 0.5mA. By increasing the LO power and/or decreasing the temperature of the detector by few tens of degrees using thermo-cooled elements, these detectors could reach the ultimate heterodyne detection limit, set by their absorption efficiency7,13 and the relative intensity noise of the local oscillator29. In conclusion, we have demonstrated coherent heterodyne detection at 9 m wavelength. By beating two single mode QC lasers with close frequencies we have produced a heterodyne signal up to 4.2 GHz which allows detecting phase and amplitude with unmatched sensitivity at room temperature. Moreover, this scheme could be also very efficient for the generation and synthesis of microwaves. References 1. Wood, R. A. Uncooled microbolometer infrared sensor arrays. In Infrared detectors and emitters: materials and devices 8 149-175, edited by Springer, Boston https://doi.org/10.1007/978-1-4615-1607-1_6 (2001) 2. Rogalski, A. Infrared detectors: status and trends. Progress in quantum electronics 27.2, 59-210 http://dx.doi.org/10.1016/S0079-6727(02)00024-1 (2003) 8 3. Gunapala, S. D. & Bandara, S. V. in Intersubband Transition in Quantum Wells: Physics and Device Applications I, Semiconductors and Semimetals 62(4) 197–282, edited by H. C. Liu and F. Capasso Academic Press, San Diego (2000) 4. Mizaikoff, B. Peer reviewed: Mid-ir fiber-optic sensors Am. Chem. Soc. 75 258A http://pubs.acs.org/doi/pdf/10.1021/ac031340g (2003) 5. Martini, R., & E. A. Whittaker. Quantum cascade laser-based free space optical communications. Journal of Optical and Fiber Communications Reports 2(4) 279-292 http://dx.doi.org/10.1007/s10297-005-0052-2 (2005) 6. Henini, M. & Razeghi, M. Handbook of infrared detection technologies Elsevier, Oxford (2002) 7. Schneider, H. & Liu, H.C. Quantum Well Infrared Photodetectors Physics and Applications, Springer, New York United States (2007) 8. Levine, B. F., Choi, K. K., Bethea, C. G., Walker, J., & Malik, R. J. New 10 μm infrared detector using intersubband absorption in resonant tunneling GaAlAs superlattices. Appl. Phys. Lett. 50(16) 1092-1094 http://dx.doi.org/10.1063/1.97928 (1987) 9. Todorov, Y. et al. Optical properties of metal-dielectric-metal microcavities in the THz frequency range Opt. Express 18 13886 https://doi.org/10.1364/OE.18.013886 (2010) 10. C. Feuillet-Palma, Y. Todorov, A. Vasanelli, and C. Sirtori. Strong near field enhancement in THz nano-antenna arrays. Sci. Rep. 3 1361 https://doi.org/10.1038/srep01361 (2013). 11. Nga Chen, Y. et al. Antenna-coupled microcavities for enhanced infrared photo- detection. Appl. Phys. Lett. 104 031113 http://dx.doi.org/10.1063/1.4862750 (2014) 12. Palaferri, D. et al. Patch antenna terahertz photodetectors. Appl. Phys. Lett. 106 161102 http://dx.doi.org/10.1063/1.4918983 (2015) 13. Palaferri, D. et al. Ultra-subwavelength resonators for high temperature high performance quantum detectors. New J. Phys. 18(11) 113016 http://dx.doi.org/10.1088/1367- 2630/18/11/113016 (2016) 14. Grundmann, M. Physics of Semiconductors 11 Springer Berlin (2010). 9 15. Howarth, D. J., & Sondheimer, E. H. The theory of electronic conduction in polar semi- conductors. Proc. R. Soc. A: Mathematical, Physical and Engineering Sciences 219(1136) 53-74 http://dx.doi.org/10.1098/rspa.1953.0130 (1953) 16. Faist, J. et al. Distributed feedback quantum cascade lasers. Appl. Phys. Lett. 70(20) 2670-2672 http://dx.doi.org/10.1063/1.119208 (1997) 17. Vodopyanov, K. L., Chazapis, V., Phillips, C. C., Sung, B., & Harris Jr, J. S. Intersubband absorption saturation study of narrow III-V multiple quantum wells in the spectral range Semicond. Sci. Technol. 12(6) 708 https://doi.org/10.1088/0268-1242/12/6/011 (1997) 18. Theocharous, E., Ishii, J. & Fox, N. P. A comparison of the performance of a photovoltaic HgCdTe detector with that of large area single pixel QWIPs for infrared radiometric applications Infrared physics & technology 46(4) 309-322. http://doi.org/10.1016/j.infrared.2004.05.002 (2005) 19. Grant, P. D., Dudek, R., Buchanan, M., & Liu, H. C. Room-temperature heterodyne detection up to 110 GHz with a quantum-well infrared photodetector. IEEE Photon. Technol. Lett. 18(21) 2218-2220 https://doi.org/10.1109/LPT.2006.884267 (2006) 20. Hao, M. R., Yang, Y., Zhang, S., Shen, W. Z., Schneider, H., & Liu, H. C. (2014). Near‐room‐ temperature photon‐noise‐limited quantum well infrared photodetector. Laser & Photonics Reviews, 8(2), 297-302. https://doi.org/10.1002/lpor.201300147 (2014) 21. Graf, M., Hoyler, N., Giovannini, M., Faist, J., & Hofstetter, D. InP-based quantum cascade detectors in the mid-infrared. Appl. Phys. Lett. 88(24), 241118 http://dx.doi.org/10.1063/1.2210088 (2006) 22. Hofstetter, D. et al. Mid-infrared quantum cascade detectors for applications in spectroscopy and pyrometry. Appl. Phys. B 100(2) 313-320 http://dx.doi.org/10.1007/s00340- 010-3965-2 (2010) 23. Hinds, S. et al. Near-Room-Temperature Mid-Infrared Quantum Well Photodetector. Adv. Mater. 23(46) 5536-5539 http://dx.doi.org/10.1002/adma.201103372 (2011) 24. Piotrowski, J., Galus, W., & Grudzien, M. Near room-temperature IR photo-detectors. Infrared Physics 31(1) 1-48 https://doi.org/10.1016/0020-0891(91)90037-G (1991) 10 25. Helm, M. The basic physics of intersubband transitions. Semiconductors and semimetals 62 1-99 edited by H. C. Liu and F. Capasso, Academic Press, San Diego (2000). 26. 27. Rosencher, E. & Vinter, B. Optoelectronics Cambridge University Press (2004). Liu, H.C. Photoconductive gain mechanism of quantum‐well intersubband infrared detectors Appl. Phys. Lett. 60 1507 http://dx.doi.org/10.1063/1.107286 (1992) 28. Hava, S., & Auslender, M. Velocity‐field relation in GaAlAs versus alloy composition. J. Appl. Phys. 73(11) 7431-7434 http://dx.doi.org/10.1063/1.353985 (1993) 29. Gensty, T., Elsässer, W., & Mann, C. Intensity noise properties of quantum cascade lasers. Opt. Express 13(6) 2032-2039 https://doi.org/10.1364/OPEX.13.002032 (2005) Methods QWIP fabrication. The QWIP structure is grown by MBE (molecular beam epitaxy). It consists of five GaAs QWs, each with a thickness LQW = 5.2 nm and each n-doped across the central 4 nm region with Si at a density of Nd = 1.75x1018 cm-3. The QWs are separated by Al25Ga75As barriers of thickness Lb = 35 nm. At the top and bottom of this periodic structure GaAs contact layers are grown, with thicknesses Lc,top = 100.0 nm and Lc,bottom = 50.0 nm and doping Nd,top = 4.0x1018 cm-3 and Nd,bottom = 3.0x1018 cm-3, respectively. The double-metal structures are obtained through wafer-bonding on a GaAs host substrate using 500 nm gold layers, and by selectively etching down to an etch-stop Al65Ga35As layer grown before the bottom contact. As shown in Fig. 1a, the patch-antennae are connected by 150 nm thin metallic wires which are realized using electron-beam lithography (consecutive alignments allow different metallic alloy contacts). The final structure is obtained by ICP etching of the semiconductor region between the antennae. The 45° facet substrate-coupled geometry consists of a 200 µm diameter circular mesa, with annealed Pd/Ge/Ti/Au as a top contact and annealed Ni/Ge/Au/Ni/Au as a diffused bottom contact. Reflectivity and photocurrent analysis. Reflectivity spectra and photocurrent spectra were obtained using a Bruker Vertex interferometer. Reflectivity measurements were performed at a 15° incident angle and at room temperature. For the photocurrent spectra, QWIP devices were mounted in a cryostat with an internal cooled metallic shield and a ZnSe optical window. Photocurrent and responsivity were measured using a blackbody source at 1000 °C, which was calibrated with an MCT detector. The source is focused onto the detector by two gold parabolic mirrors (f/1 and f/3), providing typical field of view of 60°. The photocurrent is measured with a lock-in technique using an optical chopper at 1059 Hz and a shunt resistance connected to the voltage input of a lock-in amplifier Stanford Research SR1830, without using pre-amplifiers. 11 Heterodyne measurement. The two beams from the QCLs are made collinear using f/0.5 germanium lenses and a beam splitter, and then focused onto the detector by a f/1.5 lens and a λ/4 waveplate to avoid optical feedback (Fig. 3a). The two lasers are DC biased with a voltage supply and are mounted in two Janis cryostats to stabilize their temperatures using liquid nitrogen flow. The QWIP is polarized by a Keythley 2450 sourcemeter and the heterodyne signal is sent to the spectrum analyser Agilent E4407B using a bias tee. In this arrangement the QWIP detector is at room temperature, without using any cooling system. Acknowledgements We acknowledge financial support from the FP7 ITN NOTEDEV project (Grant. No. 607521), the ERC grant "ADEQUATE", the French National Research Agency (ANR-16-CE24-0020 Project "hoUDINi"), and the EPSRC (UK) projects "COTS" and "HYPERTERAHERTZ" (EP/J017671/1, EP/P021859/1). EHL and AGD acknowledge the Royal Society and the Wolfson Foundation, and thank Dr L. Chen for skilled support with the device processing. Author contributions D.P., Y.T. and C.S. conceived the experiments, designed the QWIP structure, analysed the data and wrote the manuscript. D.P. fabricated the QWIP devices and performed measurements and data analysis together with A.B. A.M. and D.G. helped with the heterodyne measurements. A.C. calibrated the blackbody for the responsivity measurements and helped with the characterization of the mesa device. A.V. helped with data analysis. L.L., A.G.D. and E.H.L. grew the QWIP structure and provided the wafer-bonding for the double-metal processing. F.K., M.B. and J.F. provided the DFB QCLs for the heterodyne experiment. All the work has been realised under the supervision of C.S. Competing financial interests The authors declare no competing financial interests. 12 Supporting Information Room temperature 9m photodetectors and GHz heterodyne receivers Device architecture of the QWIP patch antenna array Fig. S1 shows a SEM image of the quantum detector device made of our metamaterial photonic concept. The pixel of the device is 50x50µm2. The external pad is connected to the array by the 150nm wires and is insulated from the bottom ground plane by a 800nm thick Si3N4 layer. The TiAu pad connects the device to the external circuit by wire bonding. Fig. S1. Mid-infrared QWIP structure embedded into an array of patch resonators, pixel size 50µm. Light polarization dependence Our structures support two fundamental modes, TM100 and TM010, which are represented in Fig. S2a. This figure shows the vertical electric field Ez in the plane of the resonator, obtained through finite elements simulations. The electric field distribution follows a standing wave pattern, with a node in the center of the square and maxima at the edges. The connecting wires perturb the TM010 mode slightly, which results in a lower coupling efficiency for this mode. As a result, the total photoresponse of the antenna-coupled device has a co-sinusoidal dependence with the light polarization of the normally incident wave. 13 2𝜋 0 In Fig. S2b, we plot the peak value of the photocurrent for a s = 1.30 µm structure as a function of the polarization of a plane wave incident on the array (open circles), with the 90° direction corresponding to the direction of the connecting wires. The angular integral of Fig. S2b gives a polarization coupling coefficient 𝜉𝑎𝑟𝑟𝑎𝑦 = ∫ 𝐼𝑝ℎ𝑜𝑡𝑜(𝜃)𝑑𝜃 = 71%. The product 𝐶𝜉𝑎𝑟𝑟𝑎𝑦 quantifies the percentage of incident photons that are gathered by the structure and this allows one to define a photon collection area of each single patch resonator of the array: Acoll = 𝐶𝜉𝑎𝑟𝑟𝑎𝑦𝑝2. The contrast value C of the TM100 polarized light is obtained from the measurement of Fig. 1b. For comparison, in the same graph we also plot the polarization dependence of the photoresponse measured for the mesa geometry (open squares). Here the 0° direction corresponds to the growth direction of the QWs, and the incident wave propagates normally to the 45° polished facet. This polar plot therefore recovers the inter-subband selection rule, as expected7,25. Fig. S2. a, Finite element simulation of the Ez field component coupled with the patch cavity QWIP, for the TM100 and the TM010 modes. b, polar graph of the cavity photocurrent peak as function of the wire grid polarization angle. The photocurrent is normalized at its maximum at 0. The open circles are the results for the cavity array, where the 90° direction corresponds to the connecting wires. The open squares are the results for the mesa geometry, where the 0 direction corresponds to the growth direction of the QWs. Responsivity, gain and background-limited detectivity In fig. S3a we show the responsivity curves as function of voltage for both the mesa and the patch cavity with s = 1.35 µm. The decrease of the responsivity with temperature is attributed to the thermal dependence of the charge carrier drift velocity and to an increased phonon- electron interaction14,15 (see Fig. 2c). Note that QWIP devices show the typical negative differential photoconductivity, identified as the Gunn effect, which consists of a photocurrent 14 decrease as function of voltage at specific critical fields, at which inter-valley electron scattering is induced in GaAs7. Fig. S3 a, responsivity of the mesa and the s=1.35 µm antenna-coupled devices as function of applied voltage. The temperature in K of the QWIP is indicated for each measured curve. b, background-limited specific detectivity for the mesa (up to 150 K) and the microcavity (up to room temperature) devices. The responsivities of the mesa can be expressed by considering the voltage dependent photoconductive gain g(T,V) of the detector active region and the peak inter-subband energy E21 = 143 meV (taking into account many-body effects) : 𝑅𝑚𝑒𝑠𝑎(𝐸21, T, 𝑉) = 𝜂𝑖𝑠𝑏(𝐸21) 𝑒𝑔(T, 𝑉) 𝑡𝐺𝑎𝐴𝑠𝜉𝑚𝑒𝑠𝑎/𝐸21 (S1) where 𝜂𝑖𝑠𝑏= 5.0% is the absorption coefficient for the five QW system in the 45o facet geometry, e is the electron charge, 𝑡𝐺𝑎𝐴𝑠 = 0.67 is the substrate transmission coefficient at 8.6 µm and 𝜉𝑚𝑒𝑠𝑎= 0.5 is the polarization factor (only one polarization of the incident light is coupled with the 45° facet). Analogously to Eq. (S1), we can define21: 𝑅𝑎𝑟𝑟𝑎𝑦(𝐸21, T, 𝑉) = 𝐵𝑖𝑠𝑏(𝐸21) 𝐵𝑖𝑠𝑏(𝐸21)+𝑄𝑜ℎ𝑚 −1 +𝑄𝑟𝑎𝑑 −1 𝑒𝑔(T, 𝑉) 𝐶𝜉𝑎𝑟𝑟𝑎𝑦/𝐸21 (S2) where Qohm = 4 and Qrad = 22 represent the ohmic and radiative dissipation of the double metal cavity, respectively, obtained by reflectivity measurements. The dimensionless parameter Bisb quantifies the energy dissipation through inter-subband absorption and is expressed by a lorentzian lineshape: 𝐵𝑖𝑠𝑏(𝐸) = 𝑓𝑤 2 𝐸𝑃 4𝐸21 ℏΓ (𝐸−𝐸21)2+ (ℏΓ)2 4 (S3) 15 where fw = NQWLQW/L=0.067 is the filling factor of the absorbing QWs on the overall thickness, Ep = 47.2 meV is the inter-subband plasma energy, and = 15.0 meV is the full-width-at-half- maximum of the mesa photo-response, obtained by a fit to the experimental data. We obtain a similar value Bisb=0.07 for the two resonant cavities s = 1.30 µm and s = 1.35 µm. The absorption coefficient in the antenna-coupled QWIPs is described by the branching ratio 𝜂𝑎𝑟𝑟𝑎𝑦 = = 18.9%. Using Eq. (S1) and Eq. (S2) with the measurement data in Fig. 2a, we 𝐵𝑖𝑠𝑏 −1 +𝑄𝑟𝑎𝑑 −1 𝐵𝑖𝑠𝑏+𝑄𝑜ℎ𝑚 obtain very similar values for the photoconductive gain for the mesa and the array, as shown for the data at 0.5 V (21 kV/cm) in Fig. 2a. This confirms that the absorbing regions for the two geometries are identical. Furthermore, the data shows an exponential decrease of the gain as a function of temperature. Following Ref. 7 the photoconductive gain can be defined as: 𝑔 = 𝜏𝑐𝑎𝑝𝑡𝑣𝑑 𝑁QW𝐿𝑝 (S4) where 𝜏𝑐𝑎𝑝𝑡 = 5 ps is the capture time, 𝑣𝑑 is the drift velocity, NQW = 5 is the number of quantum wells and Lp = 40.2 nm is the length of a period in the structure. The thermal dependence of the gain is related directly to the drift velocity and therefore to the electron mobility. Following Ref. 14 we can express the temperature dependence as: 𝑔(𝑇) = 1 1 𝑔0 + 𝐵 𝐸𝐿𝑂 𝑘𝐵𝑇 exp( +( 𝐸𝐴𝐶 𝑘𝐵𝑇 ) 3/2 ) (S5) Here ELO=36 meV is the longitudinal optical phonon energy in GaAs, and the fit parameter g0=1.25±0.03 expresses the value of the gain at equilibrium (without thermal scattering dependence). The second term in the denominator represents the polar optical scattering (see Ref. 15) where the parameter B=24.4±1.6 is a dimensionless polar constant and the third term represents the deformation potential scattering caused by interaction of carriers with acoustic phonons, with a corresponding parameter EAC=0.07±0.01 meV which characterizes the acoustic deformation potential. Eq. (S5) provides very good fits of the experimental data, confirming the model. The values of photoconductive gain obtained in this way are used to calculate the background limited detectivity as function of applied voltage, at different temperatures, as illustrated in Fig. S3b. Linearity and Heterodyne Measurement In Fig. S4 we show the spectra of the two QCLs compared to the room temperature response of the QWIP in the microcavity array geometry. We notice that the lasers are detuned from the maximum inter-subband absorption, resulting in a detector photoresponse that is half of the 16 maximum achievable. This is an important remark because the responsivity and detectivity values we report in Figs. 2 and 3 correspond to the peak values of detector photoresponse. The background-limited NEP (noise equivalent power) is defined as NEP=√𝐴𝑑𝑒𝑡/D. Using our measured value of detectivity at 295 K for the cavity with s = 1.30 µm at 0.5 V (Fig. 3) we have D=2.8×107 cmHz0.5/W and NEP = 0.2 nW/Hz0.5. Taking into account the 50% spectral overlap, we obtain NEP = 0.4 nW/Hz0.5, which agrees with that observed from the linearity measurement in Fig. 4c. Therefore the data presented in the main text are perfectly consistent. Fig. S4 Emission spectra of the QC lasers compared to the room temperature response of the microcavity QWIP For the heterodyne measurement the QC laser used as the LO is kept at a temperature 254 K while the QC laser used for the signal is kept at 293 K. With the temperature stabilized, it is possible to tune the spectral position of the two DFBs by slightly changing the applied DC current, according to the tuning coefficients βLO=378 MHz/mA and βS=413 MHz/mA (extracted from a linear fit to the emission frequency of the lasers as a function of temperature and bias). In the case of a high power LO, the NEP of the heterodyne can be written7 NEPhet=E21Δf/𝜂 where 𝜂 is the absorption coefficient of the QWIP. For our device in the microcavity array we have a theoretical limit of NEPhet of less than 1 aW for an integration time of 1 s at 300 K. In the experiment shown in Fig. 4, the signal-to-noise ratio is still mainly limited by the dark current. The square root fit of the signal-to-noise ratio can be extrapolated to a S/N ratio equal to 1, which provides NEPhet 10 fW for an integration time of 1 s (NEPhet 1 pW for an integration time of 10 ms), that is still four orders of magnitude higher than the theoretical limit. These estimations indicate that a high power LO could achieve sensitivities at the single photon level at room temperature. 17
1907.06505	1	1907	2019-07-12T12:20:31	Hybrid MoS$_2$-gap-mode metasurfaces photodetectors	[ "physics.app-ph", "physics.optics" ]	Two-dimensional molybdenum disulfide (MoS$_2$) featuring atomically thin thickness and unique electronic structure with favorable bandgap has been widely recognized as an attractive new material for the development of the next generation of ultra-compact, light-weight optoelectronic components. In parallel, the recently emerged metasurfaces have demonstrated exceptional controllability over electromagnetic field within ultra-compact subwavelength dimension offering an unprecedented approach to improve the performance of optoelectronic devices. In this work, we are proposing an integration of metasurfaces with 2D semiconductor materials to achieve polarization sensitive, fast-response photodetectors. The reported devices are among the most compact hybrid MoS2-gap-plasmon metasurface detectors. Relying on the significant electromagnetic field confinement provided by the metasurfaces to enhance light absorption and to reduce the surface states, which generally limit the photo-generated carriers lifetime, we measured enhanced photocurrent and a fast detection speed. Moreover, the strong optical anisotropy introduced by the metasurfaces is used to efficiently control the polarization sensitivity of the photodetector. This work provides a feasible and effective solution to improve the performance of two-dimensional materials based photodetectors.	physics.app-ph	physics	Hybrid MoS2-gap-mode metasurfaces photodetectors Peinan Ni1, Andrés De Luna Bugallo2,, Xun Yang3, Victor M. Arellano Arreola2, Mario Flores Salazar2, Elodie Strupiechonski2, Blandine Alloing1, Chongxin Shan3, and Patrice Genevet1, 1Université Cote d'Azur, CNRS, CRHEA, rue Bernard Gregory, Sophia Antipolis 06560 Valbonne, France 2 CONACYT- Cinvestav Unidad, Querétaro, Querétaro, Qro 76230, Mexico 3School of Physics and Engineering, Zhengzhou University, Zhengzhou 450052, China * corresponding authors: [email protected] and [email protected] Two-dimensional molybdenum disulfide (MoS2) featuring atomically thin thickness and unique electronic structure with favorable bandgap has been widely recognized as an attractive new material for the development of the next generation of ultra-compact, light-weight optoelectronic components. In parallel, the recently emerged metasurfaces have demonstrated exceptional controllability over electromagnetic field within ultra-compact subwavelength dimension offering an unprecedented approach to improve the performance of optoelectronic devices. In this work, we are proposing an integration of metasurfaces with 2D semiconductor materials to achieve polarization sensitive, fast-response photodetectors. The reported devices are among the most compact hybrid MoS2-gap-plasmon metasurface detectors. Relying on the significant electromagnetic field confinement provided by the metasurfaces to enhance light absorption and to reduce the surface states, which generally limit the photo-generated carriers lifetime, we measured enhanced photocurrent and a fast detection speed. Moreover, the strong optical anisotropy introduced by the metasurfaces is used to efficiently control the polarization sensitivity of the photodetector. This work provides a feasible and effective solution to improve the performance of two-dimensional materials based photodetectors. KEYWORDS: Gap plasmon polariton; metasurfaces; 2D materials; MoS2; photodetector 1 Introduction The unique optical and optoelectronic properties of atomically thin and layered two dimensional (2D) transition metal dichalcogenides (TMDCs) have been exploited to study and demonstrate a great variety of emerging optoelectronic nano-devices in the visible spectral range. 1-3 Molybdenum disulfide (MoS2), a typical 2D TMDC, has attracted extensive attention as a promising candidate for the next-generation ultra- compact photodetectors due to its excellent electronic and optical properties, i.e. a direct bandgap of 1.8 eV in monolayer and an indirect bandgap of 1.3 eV for bulk or multilayer, high carrier mobility of about 200 cm2V-1s-1 for monolayer and about 500 cm2V-1s-1 for few layers, wide wavelength absorption (350- 1000 nm) and high power conversion efficiencies.4-6 However, currently the performances of the MoS2 based photodetectors are mainly limited, essentially due to their extreme thinness, which leads to a decrease of light-matter interaction and therefore weak light absorption, and also because of their large surface to volume ratio, which not only makes them very sensitive to the environment but also largely limits their response speed.7,8 Indeed, the large amount of surface trapping states of MoS2 can capture photo-generated electrons and results in the long lifetimes of photo-generated holes, leading to slow photoresponse generally ranging from several microseconds to seconds.9-10 Great efforts have thus been devoted to improve the performance of MoS2 based photodetectors by solving those problems. For example, the localized surface plasmon of metallic nanostructures has been employed to enhance the light absorption of MoS2.11 In addition, both vacuum annealing and surface encapsulation approaches have been utilized to improve the surface quality of MoS2-based devices by reducing the trapping states.9,10 Emerging as a rapidly developing technique , metasurfaces, a new class of 2D optical components have been demonstrated, showing extraordinary capabilities for molding the light properties in a very compact and 2 efficient way. Metasurfaces are thus giving rise to a large variety of novel optical components including frequency selective surface (FSS), polarization converter, wavefront shaping, hologram, hybrid and free- form metaoptics.12-17This compelling features characterize metasurfaces as versatile and indispensable platforms for spectral and spatial manipulation of electromagnetic waves within subwavelength regime. Extending the remarkable controllability of metasurfaces into 2D materials-based optoelectronic devices through hybridization and integrations will open up great opportunities to realize the high performance ultra-compact optoelectronic devices. In this work, high performance MoS2 photodetectors with polarization sensitivity and fast response are demonstrated through the integration of a gap-surface-plasmon (GSP) metasurface. The resonance mode of the GSP metasurface is designed to couple with the neutral A-exciton resonance of monolayer MoS2. By doing this, the light-matter interactions inside the monolayer MoS2 are selectively enhanced due to the strong light confinement of the GSP metasurface. Meanwhile, placing GSP in close proximity to form a metasurface, i.e. by placing nanostripes parallel and close to each other's, we form an ensemble of adjacent nanostripes that serve as antenna-assisted electrodes to collect the photo-generated carriers very efficiently.18 On the other hand, since the electromagnetic field inside this structure is concentrated within the small volume of the metasurface, the proposed photodetectors are less affected by the large amount of surface states located on the entire surface and, as a result, show a fast photo-response. Furthermore, we show that the strong optical anisotropy introduced by the metasurface can be used to effectively control the polarization sensitivity of the photodetectors. 3 Results and discussion Since 2D materials present no dangling bonds on the surfaces they are perfect candidates to implement integrated hybrid structure with other materials beyond the limitation of lattice matching. In particular, the integration of a plasmonic structure can remarkably enhance the light-matter interactions inside a 2D material19-21. To this end, the hybrid MoS2-gap-mode metasurface photodetector is designed and realized by integrating metallic nanostripe-arrays onto the MoS2 flake, as shown in Fig. 1.This integration can greatly improve the performance of MoS 2 based photodetectors by making advantage of the extraordinary capabilities of metasurface in manipulating light-matter interactions at subwavelength dimensions. In the proposed structure, each metallic Au nanostripe acts as individual gap plasmon cavities. The gap plasmon resonance established inside this structure is determined by the standard Fabry-Pérot resonance formula: Eq. 1 where w is the width of the nanostripe, k 0 is the vacuum wave number, ngsp is effective index of the GSP, m is an integer defining the mode order, and φ is an additional phase shift acquired by the gap plasmon upon reflection at the edges of the metallic top strip. It can be seen that the width of each nanostripe plays a critical role in the design to determine the resonance conditions of the GSP metasurface. Surface plasmon excited and propagating back and forth across the nanostripe acquires propagation and reflection phase delays leading to a resonant behavior, as discussed in Eq.1. The additional reflection phase shift depends on the structural and materials parameters and increases with increasing gap layer to 𝜙~ 3 5 𝜋 for a relatively thick cladding layer of about 100 nm in this structure. Using Eq1, we estimate that the effective mode index of the GSP mode is 𝑛𝑔𝑠𝑝~1.8. 4 0gspwknmThe GSP resonance of the metasurface integrated in this structure is designed to match the exciton A resonance of the MoS2 monolayer at around 650 nm. For such purpose, finite difference time domain (FDTD) simulations are performed to investigate the resonance of the bare GSP metasurface to determine the geometry of the nanostripe arrays. Figure 2(a) shows the simulated reflectance of the designed GSP metasurface as a function of the width of the Au nanostripe while the period of the stripe array is fixed at 445 nm. At resonance, the optical energy can be effectively coupled into the GSP mode. Meanwhile, due to the close proximity of the nanostripes and bottom metallic mirror, the electromagnetic field is tightly confined, mostly within the gap region, in a mode volume in the order of 𝑉𝑒𝑓𝑓 ~3.5 10−4 𝜇𝑚3, almost two order of magnitude smaller than the diffraction limited mode volume ( 𝜆 2𝑛𝑔𝑠𝑝 3 ) considering 𝜆 as the free space PL wavelength, as confirmed by the full electromagnetic wave simulations shown in Fig. 2(b). The large confinement factor provided by the plasmonic nanocavities can greatly enhance the photocurrent response of the 2D materials-based photodetectors by improving the light absorption. 18 To benefit from the electromagnetic field confinement enabled by the proposed GSP metasurface in this work, the MoS2 monolayer is placed in close contact with the metallic nanostripes within the large electromagnetic field enhancement regions, as illustrated in Fig. 2(b). Figure 3 shows the photoluminescence (PL) spectra of the MoS2 flakes both inside and outside the GSP metasurface. PL enhancement can be clearly observed from the MoS2 flake inside the metasurface as a result of the tight electromagnetic confinement of the GSP metasurfaces, which is in good agreement with the design of the GSP mode coupling with the exciton resonance of MoS2. It is also worth pointing out that the large PL enhancement is observed even if the PL emission of the MoS2 is partially quenched by the absorption of the Au stripes (considering that the MoS2 flake is in direct contact with top Au stripes). In addition to the 5 strong enhancement of light-matter interactions, the GSP metasurface will also act as electrodes which greatly increases the collection efficiency of the photo-generated carriers, further improving the performance of the MoS2 based photodetectors. Due to the relatively small electron and holes mobility (𝜇) in MoS2, around 𝜇𝑒 = 150 cm2 V−1 S−1 and 𝜇ℎ = 80 cm2 V−1 S−1 respectively, and the short carrier lifetime (𝜏) due to various recombination and phonon excitations 𝜏 (𝑀𝑜𝑆2) ≈ 300 𝑝𝑠, the diffusion lengths ( 𝐿 = √𝜇ℎ𝑘𝑏𝑇 𝑒 𝜏 ) with of photogenerated electrons and holes are in the order of 𝐿𝑒 (𝑀𝑜𝑠2) ≈ 300𝑛𝑚 𝑎𝑛𝑑 𝐿ℎ (𝑀𝑜𝑠2) = 200𝑛𝑚 .21 We thus designed nanosize gap between adjacent collecting antennas of 370nm, in agreement with the photogenerated carrier diffusion length to maximize photoresponse. To further investigate the impact on the crystal structure of the MoS2 flakes once the photodetector was fabricated, we performed Raman spectroscopy inside and outside the metallic nanostripes , as shown in Fig. 3(b). The A1g peak is located at 405 cm-1 for both cases while the E12g is centered at 384.54 cm-1 outside and 383.3 cm-1 inside the gold grating. The slight Raman downshift of the E12g peak suggest an induced strain by the metallic structures, however, since A1g peak position remains unchanged one can dismiss charge transfer from the plasmonic cavity to the underlying MoS2 flakes. These results are in good agreement with the high photoluminescence emission intensity observed in Fig. 3(a). Furthermore, the remarkable near-field electromagnetic enhancement of plasmonic nanostructures has demonstrated exceptional polarization manipulation capability. For example, the combination of chiral plasmonic fields with the valley-selective response of TMD materials has been successfully employed to increase the valley polarization and direct valley-selective exciton emissions.22-24 In this work, we show that the strong anisotropy introduced by the GSP metasurface design provides a feasible and effective approach to control the polarization sensitivity of the photodetector. It is found that the photocurrent of 6 the structure is at a maximum when the electric field component of the incident light is perpendicular to the Au stripes (defined as 0o polarization) and at a minimum when the electric field component is parallel to the Au stripes (defined as 90o polarization), as shown in Fig. 4(a), which gives rise to an about 30-fold polarization sensitivity. The polarization sensitivity of the proposed structure can be further characterized by the polarization ratio , leading to a polarization ratio of ~0.94. Note that the anisotropic crystal structure of 2D materials can also result in directional dependence of the light absorption and the corresponding polarization-sensitive photocurrent generation, which has been demonstrated successfully to realize polarization-sensitive photodetectors. However, the reported polarization sensitive photodetectors relying on this effect only exhibit relatively small polarization ratio. For example, a ratio about 0.5 is obtained from a photodetector using a black phosphorus vertical p -- n junction,25 0.655 from a photodetector based on a MoS2/GaAs heterojunction26 and 0.709 from a photodetector based on black phosphorus on WSe2 photogate vertical heterostructure.27 On the other hand, since the orientation of the Au stripes has not been deliberately aligned with respect to the MoS2 in our structure, the polarization sensitivity is mainly due to the large anisotropy introduced by the GSP metasurface. In contrast, the larger polarization ratio observed from this structure compared to the previous works demonstrates its superior advantage to manipulate the polarization sensitivity of photodetectors. Moreover, our work also suggests that other metasurface designs such as zigzag wires, asymmetric metallic rods or chiral structures could also be employed for polarization sensitive photodetections. Finally, the photocurrent of the proposed hybrid structure shows a linear dependence on the applied voltage, as shown in Fig. 4(b). In addition, the response speed is another important figure of merit of a photodetector, which determines the ability of the detector to follow the fast varying optical signals. The 7 TMTETMTE()/()IIIIMoS2 based photodetectors usually show a slow response in the range from several microseconds to seconds due to the large amount of surface trapping states, which capture the photo-generated carriers and release them slowly when the incident light is turned off. The tightly confined electromagnetic fields inside the hybrid structure within a small mode volume greatly reduce the influence of surface states on the generation of photocurrent and as a result increases its photodetection response speed. To investigate the response speed of the hybrid MoS2 gap-plasmon metasurface photodetector, a 650 nm laser modulated by an optical chopper with tunable frequencies was used as the incident light source. The photocurrent measured at different chopper frequency is shown in Fig. 4(c). It can be seen that the structure can maintain large photocurrent at higher light switching frequency with only a slight loss in performance, indicating that it is capable of detecting fast varying optical signals. Experimental methods MoS2 flakes were grown on SiO2/Si substrates in a quartz-tube-in-a-furnace vapor-phased deposition system at ambient pressure in a quartz tube with argon as a carrier gas. MoO3 (99% Sigma Aldrich) and Sulfur powders (99% Sigma Aldrich) are used as precursors, while MoO 3 is placed at the center of the furnace in an alumina boat, sulfur is situated upstream at the edge of the furnace. The temperature of the system is first increased from room temperature to 300°C with a rate of 10°C/min, after 30 minutes the temperature is raised up to 700°C with a ramp of 30°C/min, the system is immediately cooled down naturally once it reaches the growth temperature (700°C). To fabricate the hybrid MoS2-gap plasmon metasurface structure, 5 nm Ti/ 200 nm Au were first deposited onto 1cm x 1cm sapphire substrates by electron beam evaporator acting as the metallic mirror. After that, a 100 nm SiO2 dielectric layer was deposited by a conventional sputtering system. Then, an optical lithography step is performed to define 8 large metallic contacts. MoS2 flakes are transferred from SiO2/Si substrates using PMMA and KOH solution onto the Au/SiO2 substrate. Finally, metallic nanostripe-array metasurfaces were defined by e- beam lithography followed by a 50 nm Au film deposition and lift-off. The optical properties of the MoS2 crystals were characterized by micro-photoluminescence (PL) spectroscopy, all the spectra were measured at room temperature (RT) using an inverted microscope coupled with an Andor spectrometer equipped with a intensified CCD camera. The excitation was provided by a temperature stabilized continuous laser (405 nm). The photocurrent was measurement using a semiconductor characterization system (Keithley 4200 SCS) under the excitation of a diode laser (650 nm). Conclusion Polarization sensitive, fast-response photodetectors are designed and demonstrated in an ultra-compact way by integrating MoS2 monolayer with gap-plasmon mode metasurfaces. This structure is capable of significantly improving the photodetection characteristics of 2D materials-based photodetectors such as light absorption, polarization control, and response speed. We show that these advanced performances are controlled by the strong confinement of the electromagnetic field and the resonant coupling between the excitonic resonance and the gap plasmon resonance. We show that the integration of optical metasurfaces opens up new ways of modifying and improving the performance of 2D materials-based photodetectors for a variety of applications. 9 Fig. 1 The schematic (a) and the scanning electron micrograph images (b) and (c) of the fabricated hybrid MoS2 gap plasmon metasurface photodetectors which comprise a monolayer MoS2 flakes on the top of a 100 nm thick insulating SiO2 dielectric layer sandwiched between an array of Au metallic stripe and a Au substrate. In this example, the width of the nanostripe is 75nm. 10 Fig. 2 The simulated reflectance of the bare gap plasmon metasurface without MoS2 as a function of the width of the nanostripe (a) The calculated electric field intensity in the plane transverse to the metallic stripes showing that the electric field is tightly confined in the gap layer (b), the dash line indicates the position of active MoS2 in the proposed photodetector. The period of the nanostripe array is fixed at 445 nm, i.e. sufficiently small to match roughly with the diffusion length of the photogenerated carriers. 11 Fig. 3 (a) Photoluminescence of MoS2 both with and without the GSP metasurface under the same pumping conditions. The large confinement of electromagnetic field inside the GSP metasurface leads to strong PL enhancement; (b) the Raman spectra of MoS2 monolayer inside and outside the plasmonic cavity show only slight difference likely due to the induced strain. Note that the A1g peak position remains unchanged, indicating no charge transfer from the plasmonic cavity to the underlying MoS2 flakes 12 Fig. 4 (a) the photocurrents measured at different polarization angles in polar coordinates demonstrate a large polarization sensitivity of the proposed hybrid photodetector structure; (b) the measured photocurrent and the dark current as a function of applied voltage exhibit a linear dependence on the applied voltage, revealing the photoconductive characteristic of the proposed structure; (c) the photocurrent slightly decreases when increasing the switching frequency of incident light, indicating its capacity of fast response for photodetection applications. Acknowledgements and Funding: We acknowledge financial support from the European Union's Horizon 2020 under the European Research Council (ERC) grant agreement No. 639109 (project Flatlight). ADLB acknowledges financial support from SEP-CONACYT Ciencia Basica grant No 258674 and CONACYT-ERC grant 291826. X. Yang acknowledges financial support from Natural Science Foundation of China (61804136) and China postdoctoral science foundation (2018M630829). REFERENCES (1) Hill H M, Rigosi A F, Roquelet C, Chernikov A, Berkelbach T C, Reichman D R, Hybertsen M S, Brus L E, Heinz T F, Nano Lett. 2015, 15, 2992. (2) Hu F, Luan Y, Scott M, Yan J,Mandrus D, Xu X, Fei Z, Nat. Photon. 2017, 11, 356. 13 (3) Splendiani, A, Sun L, Zhang Y, Li T, Kim J, Chim C Y, Galli G, Wang F, Nano Lett. 2010, 10, 1271. (4) Radisavljevic B, Radenovic A, Brivio J, Giacometti V, Kis A, Nat. Nanotechnol. 2011, 6 , 147. (5) Yoon Y, Ganapathi K, Salahuddin S, Nano Lett. 2011, 11, 3768. (6) Shanmugam M, Durcan C A, Yu B, Nanoscale 2012, 4, 7399. (7) Yu W J, Liu Y, Zhou H L, Yin A X, Li Z, Huang Y and Duan X F, Nat. Nanotechnol 2013, 8, 952. (8) Gong F, Fang H, Wang P, Su M, Li Q, Ho J C, Chen X, Lu W, Liao L, Wang J, Hu W, Nanotecchnol 2017, 28, 484002. (9) Kufer D, Konstantatos G, Nano Lett., 2015, 15, 7307-7313. (10) Zhang W, Huang J K, Chen C H, Chang Y H, Cheng Y J, Li L J, Adv. Mater. 2013, 25, 3456. (11) Miao J, Hu W, Jing Y, Luo W, Liao L, Pan A, Wu S, Cheng J, Chen X, Lu W, Small 2015, 11, 2392. (12) Yu N, Genevet P, Kats M, Aieta F, Tetienne J, Capasso F, Gaburro Z, Science, 2011,334, 333-337. (13) Genevet P, Capasso F, Aieta F, Khorasaninejad M, Devlin R, Optica 2017, 4, 139-152. (14) Aieta F, Kats M, Genevet P, Capasso F, Science, 2015, 347, 1342-1345. (15) Shaltout A M, Kim J, Boltasseva A, Shalaev V M, Kildishev A V, Nature Communicaitons, 2018, 9,2673. (16) Wang S, Wu P C, Su V C, Lai Y C, Chen M K, Kuo H Y, Chen B H, Chen Y H, Huang T T, Wang J H, Lin R M, Kuan C H, Li T, Wang Z, Zhu S Tsai D P, Nat. Nanotechnol. 2018, 13, 227. (17) Sawant R, Bhumkar P, Zhu AZ, Ni P, Capasso F, and Genevet P, Advanced Materials 2019, 31, 1805555 14 (18) Yao Y, Shankar R, Rauter P, Song Y, Kong, J, Loncar M, Capasso F, Nano Lett., 2014, 14, 3749. (19) Li Z, Li Y, Han T, Wang X, Yu Y, Tay B, Liu Z, Fang Z, ACS Nano, 2017, 11, 1165. (20) Liu W, Lee B, Naylor C H, Ee H S, Park J, Charlie Johnson A T, Agarwal R, Nano Lett. 2016, 16, 1262-1269 (21) Kim YC, Nguyen VT, Lee S, Park JY, Ahn YH, ACS Applied Mat. & interfaces (2018), 10, 5771- 5778 (22) Li Z, Liu C, Rong X, Luo Y, Cheng H, Zheng L, Lin F, Shen B, Gong Y, Zhang S, Fang Z, Adv Mater. 2018, 30, 1801908. (23) Hu G, Hong X, Wang K, Wu J, Xu H X, Zhao W, Liu W, Zhang S, Garcia-Vidal F, Wang B, Lu P, Qiu C W, Nature Photonics 2019 (24) Sun L, Wang C Y, Krasnok A, Choi J, Shi, Gomez-Diaz J S, Zepeda A, Gwo S, Shih C K, Alu A, Li X, .Nature Photonics 2019, 13, 180. (25) 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, Nat. Nanotechnol. 2015, 10, 707-713. (26) Jia C, Wu D, Wu E, Guo J, Zhao Z, Shi Z, Xu T, Huang X, Tiam Y, Li X, J. Mater. Chem. C 2019,7, 3817-3821. (27) Ye L, Wang P, Luo W, Gong F, Liao L, Zang J, Xu J, Hu W, Nano Energy 2017, 37, 53. 15
1812.07708	1	1812	2018-12-19T00:35:13	Blue (In,Ga)N Light-Emitting Diodes with Buried n+-p+ Tunnel Junctions by Plasma-Assisted Molecular Beam Epitaxy	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Blue light-emitting diodes (LEDs) consisting of a buried n+-p+ GaN tunnel junction, (In,Ga)N multiple quantum wells (MQWs) and a n+-GaN top layer are grown on single-crystal Ga-polar n+-GaN bulk wafers by plasma-assisted molecular beam epitaxy. The (In,Ga)N MQW active regions overgrown on the p+-GaN show chemically abrupt and sharp interfaces in a wide range of compositions and are seen to have high structural and optical properties as verified by X-ray diffraction and spatially resolved cathodoluminescence measurements. The processed LEDs reveal clear rectifying behavior with a low contact and buried tunnel junction resistivity. By virtue of the top n+-GaN layer with a low resistance, excellent current spreading in the LEDs is observed at low currents in this device structure. A few of new device possibilities based on this unique design are discussed.	physics.app-ph	physics	Blue (In,Ga)N Light-Emitting Diodes with Buried n+-p+ Tunnel Junctions by Plasma-Assisted Molecular Beam Epitaxy YongJin Cho,1,a) Shyam Bharadwaj,1 Zongyang Hu,1 Kazuki Nomoto,1 Uwe Jahn,2 Huili Grace Xing,1,3 and Debdeep Jena1,3,b) 1School of Electrical and Computer Engineering, Cornell University, Ithaca, New York 14853, USA 2Paul-Drude-Institut für Festkörkperelektronik, Hausvogteiplatz 5 -- 7, 10117 Berlin, Germany 3Department of Materials Science and Engineering and Kavli Institute for Nanoscale Science, Cornell University, Ithaca, New York 14853, USA ABSTRACT Blue light-emitting diodes (LEDs) consisting of a buried n+-p+ GaN tunnel junction, (In,Ga)N multiple quantum wells (MQWs) and a n+-GaN top layer are grown on single-crystal Ga-polar n+- GaN bulk wafers by plasma-assisted molecular beam epitaxy. The (In,Ga)N MQW active regions overgrown on the p+-GaN show chemically abrupt and sharp interfaces in a wide range of compositions and are seen to have high structural and optical properties as verified by X-ray diffraction and spatially resolved cathodoluminescence measurements. The processed LEDs reveal clear rectifying behavior with a low contact and buried tunnel junction resistivity. By virtue of the top n+-GaN layer with a low resistance, excellent current spreading in the LEDs is observed at low currents in this device structure. A few of new device possibilities based on this unique design are discussed. a) Electronic mail: [email protected] b) Electronic mail: [email protected] 1 Compared to narrower bandgap compound semiconductors such as the group III-Arsenides, the group III-Nitrides possess relatively large ionic bonding nature stemming from large differences in electronegativity between the cations and N, which results in the superior optical properties of the material system. Combined with wide span of direct bandgaps, the group III-Nitrides are thus attractive for and have been actively utilized in light emitting devices.1 For technical reasons, most nitride light emitting diodes (LEDs) so far have focused on heterostructures with p-GaN layer as the top layer, grown on top of the optically active layers. Especially metal organic chemical vapor deposition (MOCVD)-grown nitride LEDs need a post-growth-annealing step in order to break Mg-H bonds and activate p-GaN:Mg layers, requiring the p-GaN layer to be located on the surface. Due to the relatively high resistivity of p-GaN, however, current spreading is problematic in such structures. Contrarily, MBE does not require a post-growth-annealing-step for activating p- GaN:Mg since the growth is performed under H-free-ultra-high-vacuum environment, and therefore there is no penalty for positioning buried p-GaN:Mg layer in MBE structures. On the other hand, wurtzite III-Nitride semiconductor heterostructures exhibit strong spontaneous and piezoelectric polarization fields of the order of a few MV/cm along the polar c- axis. 2 These polarization fields cause quantum-confined Stark effect in the active regions of quantum-well LEDs. The polarization-induced reduction of the oscillator strength due to poor electron-hole overlap reduces the efficiency of LEDs. However, polarization engineering in such heterostructures offers several creative opportunities for photonic and electronic devices such as tunnel junctions, including ultra-low power tunneling transistors.3,4,5 The N-polar direction has recently drawn attention for unique device properties such as buried-barrier HEMTs and interband tunnel junctions.6,7,8,9,10 Although epitaxial growth along the N-polar direction presents certain 2 fundamental advantages stemming from the polarity-dependent decomposition temperatures of the materials, 11 it has been reported that N-polar nitrides show considerably low luminous efficiency and are also vulnerable to wet chemical etching process, preventing the realization of the full advantage of the favorable polarization fields in N-polar nitrides.12,13 In this paper, Ga-polar (In,Ga)N LED structures with a buried p+-GaN and a top n+-GaN were grown on GaN(0001) bulk wafers. This is structurally equivalent to the conventional (In,Ga)N LED structures, i.e., (substrate) / n-GaN / (In,Ga)N / p-GaN (surface), grown along the N-polar direction. Thus, the structure in this study enables to not only keep the high luminescence properties of Ga- polar (In,Ga)N but also exactly mimic the polarization properties of N-polar nitrides, while showing excellent current spreading on the top surface due to the n+-GaN cap layer. This novel idea was introduced in a recent work with single (In,Ga)N emitter regions, and fixed doping densities.5 In this work, we introduce multiple (In,Ga)N quantum wells, and explore the dependence of the behavior of the buried tunnel junctions with doping. Furthermore, we explore by spatially resolved cathodoluminescence the optical emission in the active regions, and demonstrate uniform light emission over mm-scale diodes at low current densities using the new device geometry. The (In,Ga)N LED structures were grown on single-crystal Ammono Ga-face GaN(0001) bulk wafers with a dislocation density of 5×104 cm-2 in a Veeco Gen10 MBE reactor equipped with standard effusion cells for elemental Ga, In, Mg and Si, and a radio-frequency plasma source for the active N species. Si was used as the n-type donor, and Mg as the p-type acceptor for doping of the GaN layers. The base pressure of the growth chamber was in the range of 10-10 Torr under idle conditions, and 2×10-5 Torr during the growth runs. The MBE-grown (In,Ga)N LED structures starting from the nucleation surface is 200 nm GaN:Si / 100 nm GaN:Mg / 5 period (In,Ga)N 3 multiple quantum wells (MQWs) / 100 nm GaN:Si. The details of the layer structures are shown in Fig. 1(a). The p+-GaN layer is located below the active region and forms a tunnel junction (TJ) with the heavily doped n+-GaN layer underneath it. Two LED samples differing in Mg concentrations [5×1018 cm-3 (sample A) and 3×1019 cm-3 (sample B)] of the p+-GaN layers were prepared in order to study the impact of the Mg concentration on the performance of the LED. All the GaN layers (ΦGa > ΦN) and the (In,Ga)N MQWs (ΦIn + ΦGa > ΦN; ΦGa < ΦN) were grown under metal-rich conditions at 710 ⷪC and 660 ⷪC, where ΦGa, ΦIn and ΦN are Ga, In and active N fluxes, respectively. The growth rate, which is limited by ΦN, was 7 nm/min. Two Ga effusion cells were used for the growth of the (In,Ga)N MQWs in order to obtain abrupt changes in the Ga fluxes in the well (ΦGa=5.5 nm/min) and the barrier layers (ΦGa=6 nm/min), which is known to be critical for the MBE growth of high-quality (In,Ga)N QWs.14 The excess Ga droplets after the growth were first removed in HCl before ex situ characterization and device fabrication. The (In,Ga)N LED samples were fabricated by optical contact-lithography followed by two-step mesa etching in Cl2- and BCl3- based inductively-coupled plasma (ICP) etching. For alloyed ohmic contacts on the top and bottom n+-GaN surface, the same Ti (25 nm) / Al (100 nm) metal stack was deposited in N2 ambient at 550 ⷪC for 1 minute by DC-sputtering. Structural properties and surface morphology of the samples were characterized by in situ reflection high energy electron diffraction (RHEED), atomic force microscopy (AFM) and x-ray diffraction (XRD) measurements. Spatially resolved cathodoluminescence (CL) spectroscopy using a beam energy of 7 keV in a scanning electron microscope (SEM) was used to probe the local variations of the optical properties of the samples. 4 Electrical transport and electroluminescence (EL) measurements were used to evaluate the device performance. All the ex situ characterization steps were performed at room temperature. Figure 1(b) displays a RHEED pattern of sample A taken at low temperature (< 300 ⷪC) after growth. The formation of a laterally contracted metallic Ga bilayer structure is confirmed by the observation of the diffused satellite streaks ("1×1" reconstruction) in the RHEED pattern as indicated by the two arrows in Fig. 1(b). These streaks guarantee that there was no polarity inversion during the growth.15,16 The AFM micrograph shown in Fig. 1(c) reveals the smooth surface morphology exhibiting clear atomic steps. Note that there are no spiral hillocks on the surface, which are commonly observed on group III-Nitride layers grown by MBE on substrates with high dislocation densities,17 indicating that dislocation density in this sample is low and no strain relaxation occurred during the growth of the (In,Ga)N active region. Instead the surface morphology is characterized by wide step terraces with widths of 300 -- 500 nm. Figure 1(d) shows the symmetric XRD ω-2θ scan of sample A. Excellent agreement between the simulated curve based on the layer structure in Fig. 1(a) and the experimental data can be seen, implying that the interfaces are sharp in a wider range and each layer in the structure is chemically abrupt. The indium distribution and luminescence characteristics of the (In,Ga)N in a micro-scale was studied by using spatially resolved cathodoluminescence (CL). Figure 2(a) shows the spatially averaged CL spectrum of sample A. Two clear CL peaks, one from GaN at ~365 nm and the other from the (In,Ga)N MQWs at 444 nm are clearly seen. The single CL peak with a narrow full-width- at-half-maximum (FWHM) of 120 meV from the (In,Ga)N MQWs implies that there is no phase separation in the MQWs and each (In,Ga)N well and barrier layer is compositionally homogeneous. 5 This can also be inferred from the excellent match between the XRD data and the simulation in Fig. 1(d). Figures 2(b) -- (f) show secondary electron (SE) [Fig. 2(b)] and CL images [Figs. 2(c) -- (f)] for different values of the CL detection wavelength on the same region of sample A. The SE image [Fig. 2(b)] is linked to the morphological features of the AFM image, [Fig. 1(c)] i.e., the surface is characterized by large step terraces and edge structure. It should be noted that the CL images do not show any clear dark spots which are easily found for (In,Ga)N grown on substrates with a high dislocation density.18 Such dark spots in CL can readily be connected to threading dislocations (TDs) with screw component (c-type dislocation: Burgers vector of c) which typically induce spiral hillocks on the surface and V-pits for (In,Ga)N.18 Such TDs with screw component are especially more detrimental than other defects for (In,Ga)N-based optical devices as they not only act as strong non-radiative centers but also induce lateral fluctuations of In incorporation around them resulting in broadening the optical transition energy.18 Figures 2(d) -- (f) show the CL images for different values of detection wavelength corresponding to the high-energy [Fig. 2(d)], the center [Fig. 2(e)], and the low-energy [Fig. 2(f)] side of the (In,Ga)N MQW CL spectrum of Fig. 2(a). Here the spectral resolution of the maps amounted to 5 nm. Within this resolution, overall the CL intensity distribution does not vary clearly with the detection wavelength but with the morphology, not showing any TD-related features (dark spots). This is indeed supported by the fact that not a single spiral hillock on the surface over an area of 20 × 20 µm2 could be observed by AFM (not shown here). Instead, the CL characteristics of this sample is seen to be correlated with the surface structure, i.e., the CL tends to be brighter on the terraces compared to that on the edges regardless of the CL detection wavelength. 6 In order to study this correlation between the CL and the morphology in more detail, a CL line scan with a higher spectral resolution of 0.5 nm was performed along a line passing through a few atomic step terraces [Fig. 2(g)]. Figure 2(h) shows the corresponding CL map as functions of scan position and wavelength taken along the scan line in Fig. 2(g). Two CL bands from the GaN and the (In,Ga)N MQWs are clearly seen. As can be expected from Figs. 2(c) -- (f), the intensities of these CL bands tend to be modulated in phase along the scan position, i.e., the CL intensities of both the GaN and the MQWs are seen to be high on the terraces and low at the edges. Figure 2(i) shows the central CL wavelength and the maximum CL intensity of the (In,Ga)N MQWs as a function of the scan position. One can see that both the CL wavelength and the maximum CL intensities of the (In,Ga)N MQWs tend to be higher on the terraces than at the edges, e.g., see the three terrace regions divided by the three dashed lines in Figs. 2(g) -- (i). At the edges, the wavelength tends to be a bit (~1 nm) lower than on the terraces. Thus carriers excited at the edges will drift towards the terraces before radiative recombination. Since for CL only the excitation is spatially resolved and not the detection, high-energy regions (the edges) with neighboring low- energy regions (terraces) show a smaller intensity in CL. The same behavior holds for the GaN [Fig. 2(h)]. Now we turn to the effects of Mg doping density on the LED performance by studying samples A and B which contain different Mg concentrations in the buried p+-GaN layers. We first check the top contacts by using transmission line model (TLM) analysis [Fig. 3(a)]. As expected, the top n+- GaN layers show linear current vs voltage relations with clear scaling effects of the TLM pad spacings [e.g., see the inset in Fig. 3(a) for sample A]. The extracted contact resistivity and sheet resistance are 1.35×10-5 Ωcm2 (9.13×10-6 Ωcm2) and 1.69×102 Ω/sq (1.18×102 Ω/sq) for sample A 7 (B), respectively. Thus, the two samples show similar low contact resistivities. Figure 3(b) and the inset depict the logarithmic and linear current density vs voltage (J-V) relation of the LEDs, respectively, exhibiting clear rectification for both samples. Here the positive voltage was applied on the bottom contact and the top contact was grounded. It is seen that sample B, which has a higher Mg doping density than sample A, reveals a lower diode turn-on voltage than sample A, indicating that carrier tunneling indeed depends on the doping densities at the n+-p+ TJ. The apparent turn-on voltages of ~5 V and ~12 V in the linear J-V curves indicate the additive voltage drop incurred at the buried tunnel junction. This turn-on voltage can be lowered further by heavier doping (without compromising the active layer growth), and/or by inserting a thin (Al,Ga)N layer between the n+-p+ junction for polarization-boost for interband tunneling, and/or by decreasing the number of MQWs.19 The significantly higher turn-on voltage for Sample A is due to a much higher reverse-bias voltage that must drop across the wider depletion region of the buried tunnel junction (BTJ) of the lighter doped Sample A, to drive the same current as the heavier doped Sample B. In other words, while ~2 V drops across the BTJ of sample B, ~8 V is needed in sample A. This indicates a strong dependence of the voltage drop on the Mg doping of the BTJ. In addition, the turn-on at TJs may become more sensitive to dopant distribution with lower doping density, which may lead to less current spreading on the device. The total series resistance of the device in sample B in forward bias has been extracted by fitting the linear region of the forward bias J-V curve to be 8.35×10-3 Ωcm2, indicating that the contact resistance (9.13×10- 6 Ωcm2) of the top n-type contact is negligible in this total resistance. The specific resistivity of the n+-p+ junction is therefore lower than 8.35×10-3 Ωcm2, which can be comparable with the lowest 8 resistances observed for nitride-based tunnel junctions, the difference being the samples reported here have little or no dislocations that can provide additional leakage paths.20,21,22 Figure 3(c) displays the normalized electroluminescence (EL) intensity spectra of the two devices in Fig. 3(b). The EL spectra consist of a single peak located around 444 nm. It is seen that sample B shows a slightly broader EL peak compared to sample A, implying that higher Mg incorporation induces a slight crystal degradation of the overgrown diode. It can also be judged from the higher leakage current of sample B than sample A in reverse bias [Fig. 3(b)]. The insets in Fig. 3(c) show optical micrographs taken from a large device with a size of 0.5×0.1 mm2 of sample B at different injection currents. As is expected for the buried-tunnel junction geometry, one can see that current spreading in the n+-GaN layer is excellent, enabling light emission from the whole device area even at a low injection current density. For the higher resistance sample A, the current spreading is not as efficient (not shown here) as sample B, underlining the need for higher Mg doping in the buried TJs. To conclude, the buried-tunnel junction (In,Ga)N LED structures reported in this study not only show excellent current spreading, but also has several other advantages due to the inherent properties where polarization and p-n junction fields are aligned parallel to each other. Wavefunctions of carriers are more concentrated in the well regions and electron injection and blocking in the active region are more facilitated, making emission efficiency much higher.5 In addition, this study suggests that by using buried n+-p+ TJs, the Ga-polar structure which is structurally equivalent to N-polar ones can be grown, enabling the use of the attractive properties of N-polar structures, without paying the penalty of the poor optical properties of N-polar (In,Ga)N and sensitivity to wet chemical etching, thus enabling nanostructure fabrication. Buried 9 n+-p+ TJs also make it possible to integrate and stack multiple light emitters such as monolithic multiple color LEDs.23 The authors thank Kevin Lee and Henryk Turski for useful discussions. This work was supported by National Science Foundation (NSF) grants: NSF DMREF award #1534303 monitored by Dr. J. Schluter, NSF Award #1710298 monitored by Dr. T. Paskova, NSF CCMR MRSEC Award #1719875, and NSF RAISE TAQs Award #1839196 monitored by Dr. D. Dagenais. The cleanroom fabrication at the Cornell Nanofabrication Facility (CNF) was supported in part by the NSF National Nanotechnology Coordinated Infrastructure (Grant ECCS-1542081). 1 S. Nakamura and M. R. Krames, Proc. IEEE 101, 2211 (2013). 2 F. Bernardini, V. Fiorentini, and D. Vanderbilt, Phys. Rev. B 56, R10024 (1997). 3 D. Jena, J. Simon, A. Wang, Y. Cao, K. Goodman, J. Verma, S. Ganguly, G. Li, K. Karda, V. Protasenko, C. Lian, T. Kosel, P. Fay, and H. Xing, Phys. Status Solidi A 208, 1511 (2011). 4 W. Li, S. Sharmin, H. Ilatikhameneh, R. Rahman, Y. Lu, J. Wang, X. Yan, A. Seabaugh, G. Klimeck, D. Jena, and P. Fay, IEEE J. Expl. CDC 1, 28 (2015). 5 H. Turski, S. Bharadwaj, H. Xing, and D. Jena, arXiv:1810.01897 (2018). 10 6 M. H. Wong, S. Keller, N. S. Dasgupta, D. J. Denninghoff, S. Kolluri, D. F. Brown, J. Lu, N. A. Fichtenbaum, E. Ahmadi, U. Singisetti, A. Chini, S. Rajan, S. P. DenBaars, J. S. Speck, and U. K. Mishra, Semicond. Sci. Technol. 28, 074009 (2013). 7 S. Krishnamoorthy, D. N. Nath, F. Akyol, P. S. Park, M. Esposto, and S. Rajan, Appl. Phys. Lett. 97, 203502 (2010). 8 X. Yan, W. Li, S. M. Islam, K. Pourang, H. Xing, P. Fay, and D. Jena, Appl. Phys. Lett. 107, 163504 (2015). 9 S. Wienecke, B. Romanczyk, M. Guidry, H. Li, X. Zheng, E. Ahmadi, K. Hestroffer, L. Megalini, S. Keller, and U. K. Mishra, IEEE Electron Device Lett. 37, 713 (2016). 10 Y. -J. Cho, Z. Hu, K. Nomoto, H. G. Xing, and D. Jena, Appl. Phys. Lett. 110, 253506 (2017). 11 K. Xu and A. Yoshikawa, Appl. Phys. Lett. 83, 251 (2003). 12 S. Fernández-Garrido, J. Lähnemann, C. Hauswald, M. Korytov, M. Albrecht, C. Chèze, C. Skierbiszewski, and O. Brandt, Phys. Rev, Appl. 6 034017 (2016). 13 D. Zhuang and J. H. Edgar, Mater. Sci. Eng. R Rep. 48, 1 (2005). 14 O. Brandt, Y. J. Sun, H.-P. Schönherr, K. H. Ploog, P. Waltereit, S.-H. Lim, and J. S. Speck, Appl. Phys. Lett. 83, 90 (2003). 15 C. Adelmann, J. Brault, G. Mula, B. Daudin, L. Lymperakis, and J. Neugebauer, Phys. Rev. B 67, 165419 (2003). 16 A. R. Smith, R. M. Feenstra, D. W. Greve, M. -S. Shin, M. Skowronski, J. Neugebauer, and J. E. Northrup, Appl. Phys. Lett. 72, 2114 (1998). 17 B. Heying, E. J. Tarsa, C. R. Elsass, P. Fini, S. P. DenBaars, and J. S. Speck, J. Appl. Phys. 85, 6470 (1999). 18 U. Jahn, O. Brandt, E. Luna, X. Sun, H. Wang, D. S. Jiang, L. F. Bian, and H. Yang, Phys. Rev. B 81, 125314 (2010). 19 J. Simon, Z. Zhang, K. Goodman, H. Xing, T. Kosel, P. Fay, and D. Jena, Phys. Rev. Lett. 103, 026801 (2009). 11 20 J. K. Sheu, J. M. Tsai, S. C. Shei, W. C. Lai, T. C. Wen, C. H. Kou, Y. K. Su, S. J. Chang, and G. C. Chi, IEEE Electron Device Lett. 22, 460 (2001). 21 S. Krishnamoorthy, T. F. Kent, J. Yang, P. S. Park, R. Myers, and S. Rajan, Nano Lett. 13, 2570 (2013). 22 S. Krishnamoorthy, F. Akyol, P. S. Park, and S. Rajan, Appl. Phys. Lett. 102, 113503 (2013). 23 F. Akyol, S. Krishnamoorthy, Y. Zhang, and S. Rajan, Appl. Phys. Express 8, 082103 (2015). 12 Fig. 1 (a) Schematic layer structure of MBE-grown (In,Ga)N LEDs with a buried p+-GaN layer. (b) The RHEED pattern, (c) 2 × 2 µm2 AFM micrograph and (d) symmetric XRD ω-2θ scan of sample A. The RHEED pattern has been taken below 300 C along the <11-20> azimuth after growth. The two arrows in (b) indicate the RHEED pattern from a Ga-bilayer on the surface. The root-mean- square roughness measured by AFM on the surface in (c) is 0.24 nm. 13 Fig. 2 (a) Cathodoluminescence (CL) spectrum averaged over an area of 88 µm2 of sample A. (b) Secondary electron (SE) and (c) -- (f) CL images of the same surface region of sample A. For the CL images, the detection wavelength is (c) near-band gap of GaN, (d) on the short-wavelength side, (e) at the center, and (f) on the long-wavelength side of the (In,Ga)N QW peak as indicated by the dashed line in (a). The lengths of the scale bars on (b) -- (f) are 1 µm. (g) SE image of a surface piece of sample A. (h) CL line-scan map acquired along the yellow line in (g). Central wavelengths (triangles) and maximum CL intensities (circles) of the (In,Ga)N QW signal in (h) are depicted in (i). The three dashed lines in (g) -- (i) indicate the positions of the step terrace edges crossing the scan line in (g). 14 Fig. 3 (a) Resistance as a function of spacing of TLM pads of samples A and B. The lines in (a) are linear fits for extracting top contact resistivities. The inset in (a) is, as a representative, the current vs voltage characteristics of sample A measured on two top n+-GaN contacts separated by distance d. Current density vs voltage (J-V) characteristics of samples A and B in semilog scale (b) and in linear scale [the inset in (b)], where the voltage was applied to the backside of the substrates and the top contacts were grounded. (c) Normalized EL spectra of samples A and B at a constant injection current of 1 mA. The upper three insets in (c) show optical micrographs taken when light is emitted from a device in sample B in a 0.5 × 0.1 mm2 mesa at currents of 0.5, 1 and 2 mA, respectively. The dashed line in the left inset outlines the probe for the top contact, which is much smaller than the mesa, most of the mesa has no metal on top. The lower inset is a color photo at an injection current of 1 mA showing blue emission. The sizes of the devices for the J-V in (b) and the EL spectrum in (c) are 80 x 80 µm2. 15
1706.05998	1	1706	2017-06-19T15:01:33	Injection locking at 2f of spin torque oscillators under influence of thermal noise	[ "physics.app-ph", "cond-mat.mes-hall" ]	Experiments, numerical simulations and an analytic model were developed to elucidate the effects of noise in the synchronized state of a tunnel junction based spin torque nano oscillator (STNO). It is demonstrated that in the in plane magnetized structure, while the frequency is locked, much higher reference currents are needed to reduce the noise by phase locking. Our analysis shows that it is possible to control the phase noise by the reference microwave current (IRF) and that it can be further reduced by increasing the bias current (IDC) of the oscillator, keeping the reference current in feasible limits for applications.	physics.app-ph	physics	Injection locking at 2f of spin torque oscillators under influence of thermal noise M. Tortarolo1, B. Lacoste1, J. Hem1, C. Dieudonné1, M.-C. Cyrille2, J. A. Katine3, D. Mauri3, A. Zeltser3, L.D. Buda-Prejbeanu1, U. Ebels1 1Univ. Grenoble Alpes, CEA, INAC-SPINTEC, CNRS, SPINTEC, 38000 Grenoble, France 2Univ. Grenoble Alpes, CEA-LETI MINATEC-CAMPUS, 38000 Grenoble, France 3HGST, 3403 Yerba Buena Road, San Jose, California 95135, USA Experiments, numerical simulations and an analytic model were developed to elucidate the effects of noise in the synchronized state of a tunnel junction based spin torque nano oscillator (STNO). It is demonstrated that in the in- plane magnetized structure, while the frequency is locked, much higher reference currents are needed to reduce the noise by phase locking. Our analysis shows that it is possible to control the phase noise by the reference microwave current (IRF) and that it can be further reduced by increasing the bias current (IDC) of the oscillator, keeping the reference current in feasible limits for applications. I. INTRODUCTION a magnetic microwave A spin polarized current passing through multi-layered nanosystem can drive its magnetization into large amplitude periodic oscillations 1,2,3 when the spin polarized current is large enough to compensate the natural damping. These spin transfer driven magnetization oscillations, together with their particular nonlinear properties4 spurred the interest in STNO's for several applications in current devices5. controlled Nevertheless, one of the main issues that remains to be addressed for these spin STNO's is their relative large linewidth. One possibility to reduce the linewidth is to couple either different layers within an oscillator6, or to couple several oscillators. For this second case, several options were proposed, experimentally and theoretically: current mediated dipolar coupling9,10 coupled nanocontacts11,12. In order to understand the conditions for electric synchronization of coupling7,8, spin wave or 1 layer, IP), for which an several oscillators by their own emitted RF current, we studied the synchronization of an STNO to a reference current source, with known spectral specifications. Here we focus on standard uniform in plane magnetized oscillators (in-plane polarizer and in-plane free in-plane precession (IPP) mode is stabilized. The injection locking of such an STNO to a reference current at two times the generated frequency (2f) was demonstrated both numerically and by experiments13. However, the linewidth in the locked regime was reduced only by a factor of seven, while a reduction to the linewidth of the microwave source (several Hz) was expected. These large linewidths are associated to the thermal noise that induces fluctuations which can drive the phase from an equilibrium state to a neighbouring one, with an associated phase slip of  2 which can be envisaged as non- syncronization re-synchronization events. Zhou et al14 demonstrated that the particular way the phase approaches its and in slips role of the phase synchronized value has consequences in the transients that may limit the modulation of an STO. Recent works investigated the mechanisms of the so called pure phase locking state in double vortex based STNO: Robust synchronization was experimentally shown, with a 105 linewidth reduction15 and the the synchronized state was investigated16. In this work we study the injection locking at 2f to an external reference current of an uniform IP magnetized STNO under the influence of thermal both experiments and numerical simulations, together with an analytic model to describe the transients to the locked regime in the IPP geometry. Our results show the key features of electric synchronization of a uniform in plane magnetized STNO. noise. We performed II. ANALYTIC MODEL The effect of thermal fluctuations on the transient behaviour of the synchronized state of an STNO is analyzed in the frame of a generic model of a non-linear auto oscillator4 that we extended IPP mode synchronized by an RF current at 2f (details in Appendix). Since STNO´s are non-linear (non-isochronous) oscillators, the power and the phase of the oscillator are not independent, leading to a system of coupled equations. the for d  dt pd  dt  pN  2  1 op Im[ ]  (1) 2 p   p  2 p F 0   cos   2 p o ]Re[  (2) Here ψ(t) = 2 - ωextt is the phase difference between the STNO phase  and the phase of 2 to the external source ωextt, N is the coefficient of is a real non-linear frequency shift, F parameter proportional the reference current, Γp is the damping rate of the power fluctuations and  has statistical properties of the Gaussian thermal noise17. Linearizing the equations (1) and (2) around a stable solution po (without considering thermal noise) allows us the transient behaviour of the synchronized state, and analytically calculate the decay rate and the power spectral density (PSD) of the phase fluctuations S of the synchronized state. In the limit of strongly non-linear oscillator  >> 1 there are two solutions for the decay rate : to study  2,1  p  1   1     c    (3) = c and IRF/IDC  = Here, 2/(2Np0 2sinψsPxΓJB/A ) (see Apendix for Γp the definition of the parameters). When >c, is complex with a real part given by p, that is the decay rate to the phase locked state and an imaginary part that describes an oscillatory approach to the phase-locked state with a frequency given by:  p /  c  1 (4) This is in agreement with Zhou et al14, where they found for out of plane (OP) magnetized STNO's that the phase approaches its locked state exponentially and oscillating above a certain critical reference current. Before discussing in more detail the oscillatory transient, we first will provide an expression for the synchronized state. By taking into account the thermal noise, we can calculate from eq. 1,2 the power spectral density (or single the phase noise in sideband) S of the phase fluctuations of the synchronized state:    p      2    p   2    S f   2 o      2 1( 2   )  f 2   c  2 f 2     4  2 p f 2 (5) is Here  = Npo/p the dimensionless nonlinear frequency shift and fo is the free running linewidth. Eq. 5 is plotted in Fig. 1a with the parameters calculated from the analytical model (see Appendix) for the free running state with a bias current IDC= 50 x 1010A/m2, which leads to an IPP stable precession mode around 4.7 GHz and a fo = 50 MHz. In this configuration the system has a coefficient of nonlinear frequency shift N = -3.161010 rad/sec, a damping rate of the power fluctuations Γp = 666 Mrad/sec, a normalized nonlinear frequency shift parameter ν = 16, and c= 0.025. Since this value of c for these uniform IP STNO's is small compared to  (~0.1 or higher) the synchronized state always takes place via an oscillatory transient. dimensionless locking the to The characterization of the phase noise properties by the PSD in the Fourier space has the advantage that its inverse power law dependence on frequency PSD ~1/fx provides information about underlying noise processes. The model predicts a crossover from 1/f2 to 1/fo with increasing reference current, with the two limit cases: 3 S    f  f roll  off 2: f   f 2 o f  f roll  off 2:   f 0           f   14   p 2 1 4   2        2  p  c  2 (6) As already shown experimentally18 for IPP STNO´s, the free running oscillator ( = 0) shows a 1/f2 dependence associated to a random walk of the phase (blue line, Fig. 1a). This behaviour is modified when applying a reference at 2f: Even for a low external force ( = 0.1, yellow line, Fig.1a) below the roll of frequency froll off~1/p down to the lowest (calculated) offset frequency the phase noise is constant. This corresponds to fluctuations of the phase around its locked value. The Eq. (6) shows that the phase noise level in this region can be decreased upon increasing the reference current . Above  = 0.1 there is a peak around froll off that is related with relaxation mechanism14,20,21. This will be discussed in more detail in the next section. the oscillatory III. MACROSPIN ANALYSIS We performed macrospin simulations for the in-plane precession (IPP) mode of an in-plane magnetized polarizer and free layer MTJ, using a solver for the Landau-Lifshitz- Gilbert equation and taking into account the damping like torque term (the field like term was neglected in this work, see Appendix). The simulation parameters are as follows: size 90x80x3.9 nm3; free spontaneous magnetization Ms=1000 kA/m, damping parameter α=0.02 zero magneto-crystalline The polarizer is aligned in the plane at 165° from the free layer magnetization equilibrium position and a spin-polarization =0.37 is supposed. A static magnetic field of 40 mT anisotropy. layer of and was applied along the in plane easy axis (Ox). The continuous current was set to JDC = -501010 A/m2, leading to an IPP stable precession mode with f ~ 4.7 GHz. A white Gaussian thermal noise field was added, corresponding to 10 K, 20 K, 50 K, and 100 K13,21. We present here the results for 50 K, similar behavior was obtained for all the studied temperatures. The frequency of the RF current was set to two times the free running STNO frequency 2fo=9,5396 GHZ, which corresponds to the centre of the locking range. FIG.1. Phase noise from analytic (a) and numerical simulations at 50K (d). Notice the peaks appearing ~200 - 300 MHz. (b) The PSD of the signal as a function ofand schematics of the oscillator (inset) for JDC = -50x1010 A/m2. c) FWHM vs. for low current JDC= -40x1010 A/m2 (red dots) and medium current JDC = -50x1010 A/m2 (blue dots). For > 2 (low JDC regime), the linewidth falls below the resolution of the technique.(f) Sideband frequency for both low current regime (red set) and medium current regime (blue set) extracted from the PSD (stars), and from the peaks on the phase noise(Fig. 1d) (open circle). The solid line corresponds to the analytical model developed for the IPP geometry (eq. 4). The phase and amplitude noise in the synchronized state as a function of the reference current was extracted from the simulated the my component of the magnetization (in-plane magnetization along the short axis of the transform pillar) the Hilbert traces of temporal using method22,23 which allows the reconstruction of an analytic signal from the voltage output: V  V  10  a   cos   2 ft  (7) The S at 50K extracted from the the LLG in Fig. 1d. The integration of time is numerical equation shown 4 corresponding evolution of the power spectral density of the my component of the magnetization with  and its FWHM are displayed in Fig. 1b and 1c respectively. Both phase and amplitude noise (Fig. 1d and e) decrease with the reference current and a clear crossover from a 1/f2 to a 1/f0 (white noise) is seen upon increasing it. Before addressing the phase noise level in comparison to the analytic results and the linewidth, we now discuss the PSD of the my component of the magnetization, Fig. 1b. The peak of the free running state becomes very narrow and two symmetric sidebands appear upon increasing the reference current (Fig.1b). These sidebands are also visible on S, that shows a peak around froll–off whose frequency depends on . All these features are consistent with the analytic model, where S calculated with the parameters listed above (eq.5, Fig. 1a) shows the peaks associated to the sideband frequencies and predicts transient with frequency given by eq. (4). Figure 1f shows the frequency of these sidebands extracted from the PSD of the my component of the magnetization and from the numeric S for two different bias currents IDC= -40x1010 A/m2 (red symbols) and IDC= -50x1010 A/m2 (blue symbols). The full line represents the model (eq.4) for both bias currents. This comparison confirms that the peaks of the phase noise and the sidebands have the same physical origin arising from the oscillatory approach of the transient. Furthermore the comparison supports the analytic model. the oscillatory In the following we discuss the phase noise level of the numerical results. There are two contributions to the phase noise in the synchronised state. The first one, as was discussed for the analytical description are phase fluctuations around the stable phase that is given by the external source plus a constant phase shift. The second contribution are phase slips24, not considered in the analytical model but that are present in the numerical calculation. To understand their contribution to the phase noise and linewidth we extracted the phase from simulated time traces for different reference current values (Fig. 2c). The phase trace shows a drift in time, together with the appearance of the phase slips, which become well defined upon increasing the reference current (Fig.2a). As can be seen the number of phase slips per sampling period (~40 μs) decreases. These phase slips are responsible for the 1/f2 contribution of the phase noise at low offset frequencies. To demonstrate this, we compare the phase noise extracted from different 10 μs segments of the total 40 µs phase trace that contain respectively none, one or two phase slips. In Fig. 2b it is clearly seen that in presence of phase slips the phase noise has a 1/f2 dependence at low frequency f <froll off, while in absence of phase slips the phase noise level is constant. The increase of the phase noise level due to phase slips is also expressed in an enhanced linewidth, see Fig. 1c. 5 FIG.2. Simulated (a) and experimental (c) phase temporal traces. Inset: detail of 5s segments of the temporal trace. The phase slips decrease in number with increasing  disappearing at  = 2.75. (b) Phase noise analysis on the temporal trace segments (inset) corresponding to no phase slips (black), 1 phase slip (red) and 2 phase slips (green) . (b) Phase noise analysis from the experimental time trace from 3s segments with 0, 1,2,10 phase slips. From the numerical analysis we can see that the drastic decrease in the linewidth with  (Fig. 1c) can be related to the decreasing number of phase slips. Particularly, when the phase noise flattens for>0.3 (> 2.75) for IDC= -50x1010 A/m2 (IDC= -40x1010 A/m2), the phase slips are absent in the phase trace and the linewidth falls under the resolution limit of the numerical calculation (20 kHz). This fact raises the question whether the so called "pure" synchronization15,16 is due the absence of phase slips, where the STNO would reduce its linewidth ideally to the one of the reference source. In the case discussed here , this means that for values larger than  = 0.3 ( = 2.75) torques from the reference current on the magnetization are strong enough to stabilize the phase around a single value and the remaining noise is given by the one discussed within the analytical model, 6 describing damped oscillations around the stable phase, for noise frequencies f <froll off . We point out here that the absence of the phase slips depends on the observation time, i.e. the length of the temporal trace: longer observation times increase the probability of phase slips. These results evidence that even if the system is in the frequency locking regime, higher values of reference current are needed linewidth reduction by phase locking. This observation is in agreement with Lebrun et al16 that highlighted the reported "frequency locked state"13,26,27 and pure phase locked states in absence of phase slips for vortex oscillators with a free running frequency of ~200 MHz and ~100 kHz linewidth. the difference between to achieve full IV. EXPERIMENT the qualitatively are tunnel experiment was The analytical and numerical results explain experimental observations on the synchronisation by a reference current. The studies were realized on the same type of devices presented in in Ref.[13,18,28], which plane magnetized magnetic junctions (MTJ), having a stack composition of IrMn/CoFeB/Ru/CoFeB/MgO/CoFe/CoFeB and nominal resistance area RA=1 Ωμm2. While results were obtained on different devices, we present here results on for one device. Its autonomous, i.e. free running regime is characterized by a free running frequency of f0= 7.5 GHz for a bias current IDC= -1.6 mA and an applied in plane field of 350 Oe, with a linewidth of 55 MHz. The synchronization done varying the reference current frequency around two times the free running frequency (2f) of the oscillator, from 14 GHz to 16 GHz, and the source power was varied from -15 dBm to -5 dBm (corresponding to a reference current of ~0.3 to 1.3 mA), just before the sample starts to show signs of degradation. A detailed description of the experiment is available in Ref. [13]. The temporal traces were measured using a single shot oscilloscope13,29,30, and amplitude and phase noise were extracted using the same protocol as for the simulated data. The PSD map of the output powerfor the STNO frequency f as a function of the source frequency fext is shown in the Fig. 3a for ε=JRF /JDC=0.7. In Figure 3b it is clearly seen that for increasing reference current ε, the frequency locking range is wider, and that the linewidth decreases until its minimal value. Notice that the linewidth reaches a 10x reduction (8 MHz with a 1 MHz resolution bandwidth). The amplitude noise shows a 1/f0 behavior both for synchronized (Fig.4c, grey dashed line) and the free running state (Fig.3c, black full line), 7 to the external source, characteristic of white noise fluctuations of the amplitude around its stable value. The experimental plots of the phase noise in the synchronized (Fig. 4c, red dashed line) and the free running state (red full line) show that the synchronization mechanism is efficient to reduce the phase noise by 20 dBm with respect the non-synchronized state. corresponing to the observed relatively large linewidth of ~ 8 MHz, instead of the expected complete reduction to the linewidth of the reference source (few Hz). Both plots display a 1/f2 behaviour characteristic. However the origin is different. In the free running state it results from a random walk of the phase, while in the synchronised state it is due to the phase slips as explained in section III. Our experiments show that for the maximum applied reference current Irf, even though the oscillator is synchronized with the emission linewidth remains broad. The phase noise decreases but it does not reach the constant level at which the linewidth is expected to be reduced to the source noise. This was not observed in our experiments because the voltage breakdown of the samples did not allow to continue increasing the reference current preventing the STNO from achieving a pure phase locked state. For the same reason, we were not able to observe sidebands in the experimental voltage output. Nevertheless, we have extracted the phase noise from shorter 3 s segments of the temporal trace25 of the phase (Fig2.c), that include either 0, 1, 2 or 10 phase slips (Fig.2d). As can be seen in absence of phase slips the phase noise is flat in a certain range of offset frequencies ~5-100MHz in Fig. 2d. This demonstrates the lowest phase noise level that can be reached for the in-plane STNO, when phase slips would be completely suppressed. Note that this level is with -100dBc/Hz the the same as in simulations, Fig. 2b and also the same as has been reported for vortex devices16. The relative large linewidth of ~ 50 MHz of our STNO's could be a drawback to linewidth as discussed by Hamadeth et al15 where by decreasing 7 times the linewidth, reduction running free the linewidth reduction in the locked regime goes from only 10 to 105. Indeed, this is also witnessed in our numerical simulations, where the smaller free running linewidth at higher bias current (JDC= -50x1010A/m2) leads reduction for significantly lower reference currents. linewidth to a FIG.3:PSD map of the output voltage at IRF = 1.12 mA (a). Linewidth vs = IRF/IDC(b) and amplitude and phase noise from the experiment (c) for the non-synchronized state, (continuous line,=0) and synchronized state (dashed line, =0.7). Notice that the synchronization mechanism is efficient to reduce the noise level (green arrows). V. CONCLUSION in under noise. thermal We have studied the synchronization mechanism of an uniform IP magnetized STNO The these devices was synchronization of demonstrated several experiments, however no more than a 10 times reduction of linewidth was achieved. This is explained by numerical simulations including thermal noise. While the STNO can be synchronized by moderate rf currents, higher rf currents are needed for full linewidth reduction. Full linewidth reduction is achieved when phase slips are increasing reference current the number of phase slips is reduced resulting in a crossover from 1/f2 to1/f0 behaviour in the phase noise when the phase slips are suppressed. Experiment and suppressed. With 8 simulations indicate that the lowest phase noise level achievable under synchronisation is on the order of -100 dBc. The simulations also shows that it is possible to achieve linewidth reduction for lower reference currents by increasing the bias current of the oscillator. This study will be important for designing of appropriate performances for microwave applications in the gigahertz range. configurations STNO ACKNOWLEDGMENT This work was supported in part by the French National Research Agency (ANR) under contract N° 2011 Nano 01607 (SPINNOVA) and in part by the EC under the FP7 program N° 317950 MOSAIC. M.T. is now at CONICET and Universidad Nacional de San Martin, Buenos Aires, Argentina. *B. L. Iberian Nanotechnology Laboratory, Braga, Portugal International at is replaced by the canonical variables (a, a) such as: a  m  y 1(2  jm z m x ) (A2) It is convenient to express the reduced Gibbs free energy in reduced units: APPENDIX A: Model for 2f synchronization in IPP mode  0 ' VMµ 0 s E (a,a)  A in The transient behavior the synchronized state of a STNO is analyzed in the frame of a generic model of a nonlinear auto-oscillator. The model proposed by ref. [4] is extended for the synchronization by an RF spin-current at 2f. The configuration selected here has both the free-layer and the polarizer The magnetization of the free layer is supposed to be uniform, thus the Gibbs free energy associated of the ferromagnetic free layer of the nanopillar is: in-plane magnetized. E ( m ) K u 1 2  (A1) 2 x    1  VmNmNmNMµ VmHMµVm x    0 0 s 2 s 2 z 2 x 2 y X Z Y 0 Where m=M/Ms is unitary magnetization vector, V the volume of the sample, Ms the spontaneous magnetization. A static magnetic field of amplitude H0 is applied in the plane of the sample along the Ox direction. The demagnetizing effects are accounted by tensor N=(NX, NY, NZ) and a uniaxial magneto- crystalline anisotropy along Ox direction is considered of (Ku>0). Following Holstein-Primakoff transformation the variables mx, my, mz are the demagnetizing amplitude Ku the 9 2 1  2  * aa  aa 3 B  a aa 3   *  U  V a * 2  aa 2 (A3) 2 * µ  0 1 2   µ   0 0 Where  0 '  with is the damping constant, is the gyromagnetic ratio of the free electron multiplied by the vacuum permeability µ0.The notations are similar to that of Ref. 4.  2/    A A M   B M    U M   V M   0   0 H 2/  A 4/  ' NM ' H 2' K  u  Mµ  0  0  M H A Z 0 S s 2/  (A4)  N Y    NM s Y  N X    is used transform the A second diagonalization of the quadratic part of the reduced free energy (Hamiltonian) :b=ua+va* where Gibbs for u    A 0 2  0 , v    A 0 2  0 (A5) with  0 2 A  B 2 . The last transformation is simply a normalization: b  0 A c . Once a spin-polarized current of polarization p=(Px,Py, 0) is injected in the sample the modified Gilbert equation is used to describe the magnetization dynamics considering the damping-like term of the spin-transfer:  Hm  γ 0 eff   α mm d    dt     J Jaγ 0 app m d  dt (A6) pmm    The effective field is given by the functional derivative of the Gibbs free energy with respect magnetization: the . The injected current H eff  to 1 E m  VMµ 0 s density is time dependent and given by:  cos  cos  ext  ext )( t  1 app     DC DC J J J J RF  t t  The spin–torque amplitude coefficient is where t is the thickness of the a  2 tMµe s 0 J free layer and  is the spin-polarization. The numerical analysis is carried out on the equivalent Landau-Lifshitz equation: modified m d dt  γ  (A7) '   Hm 0 ' Jaγ 0 J app '    Hmm eff  pmm  αγ  0  eff Jαa J  γ    ' 0 app    pm    , aa   ,  m , cc bb  resonant the potential Applying the three transformations presented above , keeping only the terms following equation for the complex variable c is obtained:  dc dt   0 j   cN 2  c  J P x  1  cos  e 0  1    t   J P y  1  (A8) cos  e    t 2 cQ 1  cQ 2 4  c  1   1 2 2  c   c 2 B 2 A  c 2 c      u v  B  0 The expressions for the coefficients are the following: N  2  M  uuv  u vu 2  4       v v A 4 2 4 2 2    0 A  0  3  M  A    Q 1  3 uv  M  u 2   2 v     M  3   H A 2  Q   3 uuv ' Jaγ  0 J J (A9) 2  2 v   M  u 4   4 vu 22  v DC        0 2 A 3 2  2  M  A 4 2       0 A    In Eq. A8 it is important to keep the term involving c2+c2 since it mediates the 2f synchronization. This has been neglected until now in the literature. The analysis is continued in terms of for power p and phase  such as which two coupled equations are obtained: jep   c dp dt            (A10) d  dt    1  2 0  ppQ 1  2 P x  J  1  t  e cos     1    p B A p cos  2  p     2 P y  1  J t  e cos    1 2 B  0  vu   p cos    0  Np   J P y  1  t  e cos    1 2 B  0  vu   sin  p regime In DC these generic equations allow to extract the power p0 in the free-running state: (JRF=0) p 0  P  0 x J PQ   0 x J 1 (A11) and to define the amplitude relaxation rate: 10  PQ  0 1 x p  J  0 p . (A12)   2 the STNO and In the 2f synchronized state the phase difference between the is constant and external signal text the stationary power ps is shifted by . Keeping the resonant terms at p  2f, it is possible to obtain the coupled equations: s  0p p       d  dt pd  dt  pN  2 (A13) 2  p  p  2 0Fp cos     0 Np and with   02   . p 0 ext B 2 A F   P J  x The stable stationary solution corresponding to the 2f synchronized state has the phase with difference given by  arccos  s          0 F anda stationary shift in power 2  Np  p given by ps     2 N . J. Magn. J. C. Slonczewsky, [1] Magn.Mater195, L1 (1996). [2] L. Berger, Phys. Rev. B 54, 9353 (1996). [3] W. Bebber. S. Riesen and H. C. Siegman, Science 29, 2015 (2001). [4] A. Slavin and V. Tiberkevich, IEEE Trans. Magn. 45, 4 (2009). and [5] N. Locatelli, V. Cross Grollier.Nature Materials 13, 11 (2014). [6] D. Gusakova, M. Quinsat, J. F. Sierra, U. Ebels, B. Dieny, L. D. Buda-Prejbeanu, M.- C. Cyrille, V. Tiberkevich, and A. N. Slavin, Appl. Phys. Lett.99, 052501 (2011). [7] B. Georges, J. Grollier, M. Darques, V. Cross, C. Dernalot, B. Marcillhac, G. Faini and A. Fert. Phys. Rev. Lett. 101, 017201 (2008). J. 11 [8] J. Grollier, V. Cros, and A. Fert, Phys. Rev. B 73, 060409(R) (2006). [9] S. Urazhdin, V. S. Tiberkevich, and A. N. Slavin, Phys. Rev. Lett.105, 237204 (2010). [10] A.D. Belanovsky, N. Locatelli, P.N. Skirdkov, F. Abreu Araujo, K.A. Zvezdin, J. Grollier, V. Cros, A.K. Zvezdin, Appl. Phys. Lett. 103, 122405 (2013). [11] S. Sani, J. Persson, S.M. Mohseni, Y. Pogoryelov,P.K. Muduli, A. Eklund, G. Malm, M. Kall, A. Dmitriev, J. Akerman, Nat Commun.4, 2731 (2013). [12] W. H. Rippard, M. R. Pufall, S. Kaka, T. J. Silva, S. E.Russek, and J. A. Katine, Phys. Rev. Lett 95, 067203 (2005). [13] M. Quinsat, J. F. Sierra,I. Firastrau, V. Tiberkevich, A. Slavin, D. Gusakova, L. D. Buda-Prejbeanu, M. Zarudniev, J.-P. Michel, U. Ebels, B. Dieny, M.-C. Cyrille, J. A. Katine, D. Mauri, and A. Zeltser, Appl. Phys. Lett. 98, 182503 (2011). [14] Y. Zhou, V. Tiberkevich, G. Consolo, E. Iacocca, B. Azzerboni, A. Slavin, and J. Akerman, Phys. Rev. B 82, 012408 (2010). [15] A.Hamadeh, N. Locatelli, V.V. Naletov, R. Lebrun, G. de Loubens, J. Grollier, O. Klein and V. Cross, Appl. Phys. Lett. 104, 022408 (2014). [16] R. Lebrun, A. Jenkins, A. Dussaux, N. Locatelli, S. Tsunegui, E. Grimaldi, H. Kubota, P. Bortolotti, K. Yakushiji, J. Grollier, A. Fukushima, S. Yuasa and V. Cross, Phys. Rev. Lett. 115, 017201 (2015) [17] <(t)> = 0, <(t)(t)> = 0, <(t)(t)> = (t-t´) [18] M. Quinsat, D. Gusakova, J. F. Sierra, J. P. Michel, D. Houssameddine, B. Delaet, M.-C. Cyrille, U. Ebels, B. Dieny, L. D. Buda-Prejbeanu, J. A. Katine, D. Mauri, A. Zeltser, M. Prigent, J.-C. Nallatamby, and R. Sommet, Appl. Phys. Lett. 97, 182507 (2010). [19] L. Bianchini, S. Cornelissen, Joo-Von Kim, T. Devolder,W. van Roy, L. Lagae, and C.Chappert, Appl. Phys.Lett.97, 032502 (2010). Sciences Nonlinear [21] M. d'Aquino, C. Serpico, G. Bertotti, R. Bonin, and I. D. Mayergoyz, Phys. B 84, 064415 (2010). [22] G. Finocchio, M. Carpentieri, A. Giordano and B. Azzerboni,Phys. Rev. B 86, 014438 (2012). [23] B. Picinbono, IEEE Trans. Signal Process 45, 552, (1997). [24] S. E. Russek, S. Kaka, W. H. Rippard, M. R. Puffall, and T. Silva, Phys. Rev. B 71 104425 (2005). [25] A. Pikovsky, M. Rosenblum, and J. Kurths, Synchronization:A universal concept (Cambridge in University Press, Cambridge 2001). [26] Not all the jumps on the experimental phase temporal trace (Fig. 3a) are phase slips. Only the  2 jumps are compatible with phase slips, larger ones are related to extinctions in the oscillator, as seen in D. Houssammedineet al, Phys. Rev. Lett 102, 257202 (2009). [27] V. E. Demidov, H. Ulrichs, S. V. Gurevich, S. O. Demokritov, V. S. Tiberkevich, A. N. Slavin. A. Zholud, and S. Urashdin, Nat. Commun. 5, 3179 (2014) [28] A. Dussaux, A. V. Khvalkovskiy, P. Bortolotti, J. Grollier, V. Cross, and A. Fert, Phys. Rev. B 86, 014402 (2012). [28] D. Houssameddine, S. H. Florez, J. A. Katine, J.-P. Michel, U. Ebels, D. Mauri, O. Ozatay, B. Delaet, B. Viala, L. Folks, B. D. Terris, and M.-C. Cyrille, Appl. Phys. Lett. 93, 022505 (2008). [29] D. Houssameddine, U. Ebels and B. Deny, Phys. Rev. Lett 102, 257202 (2009). [30] Mark W. Keller, A. B. Kos, T. J. Silva, W. H. Rippard, and M. R. Pufall , Appl. Phys. Lett. 94, 193105 (2008). 12
1908.06330	1	1908	2019-08-17T18:37:14	Screening Current-Induced Field and Field Drift Study in HTS coils using T-A homogenous model	[ "physics.app-ph" ]	The emergence of second generation (2G) high-temperature superconductor (HTS) tapes has favored the development of HTS magnets for their applications in areas such as NMR, MRI and high field magnets. The screening current-induced field and the field drift are two major problems hindering the use of HTS tapes in the mentioned areas. Both problems are caused by the screening current, then it is necessary to have a modeling strategy capable to estimate such phenomena. Thus far, the H formulation has been the most successful and used approach to model medium-size systems (hundreds of tapes). However, its application to large-scale systems is still impaired by excessive computation times and memory requirements. Homogenization and multi-scaling strategies have been successfully implemented to increase the computational efficiency. In this contribution, we show that using the homogenization technique with the recently developed T-A formulation allows reducing the computation time and the amount of memory up to the point that real-time simulations of slow ramping cycles of large-scale systems are possible. The T-A homogeneous model also allows systematically investigating the screening current using numerical simulations.	physics.app-ph	physics	Screening Current-Induced Field and Field Drift Study in HTS coils using T-A homogenous model Edgar Berrospe-Juarez1, Frederic Trillaud2, Víctor M R Zermeño3, and Francesco Grilli4 1Postgraduate School of Engineering, National Autonomous University of Mexico, Mexico 2Institute of Engineering, National Autonomous University of Mexico, Mexico 3NKT, Germany 4Karlsruhe Institute of Technology, Germany Email: [email protected] Abstract: The emergence of second generation (2G) high-temperature superconductor (HTS) tapes has favored the development of HTS magnets for their applications in areas such as NMR, MRI and high field magnets. The screening current-induced field and the field drift are two major problems hindering the use of HTS tapes in the mentioned areas. Both problems are caused by the screening current, then it is necessary to have a modeling strategy capable to estimate such phenomena. Thus far, the H formulation has been the most successful and used approach to model medium-size systems (hundreds of tapes). However, its application to large-scale systems is still impaired by excessive computation times and memory requirements. Homogenization and multi-scaling strategies have been successfully implemented to increase the computational efficiency. In this contribution, we show that using the homogenization technique with the recently developed T-A formulation allows reducing the computation time and the amount of memory up to the point that real-time simulations of slow ramping cycles of large-scale systems are possible. The T-A homogeneous model also allows systematically investigating the screening current using numerical simulations Keywords: large-scale superconductor systems, HTS coils, screening current, screening current- induced field, field drift. 1. Introduction The quality of second-generation high-temperature superconductor (HTS) tapes has improved during recent years. Now that HTS tapes are commercially available in sufficient length, there is an increasing interest in the development of devices based on this technology. The applications include power system devices [1], NMR and MRI [2-4], accelerators for medical applications [5], scientific magnets [6-8]. The HTS materials are performing so well that they preserve their high critical current even under the effect of high magnetic fields that exceed the critical fields of low-temperature superconductors. For this reason, they are good candidates to produce magnets above 20 T. Some projects around the world are moving in this direction [7, 9, 10]. In coils wound with HTS tapes, the large aspect ratio (width/thickness) of the tapes favours the presence of screening currents. These currents are induced by the gradual penetration of the radial component of the magnetic field. The magnetic field generated by screening currents affects the quality of the overall magnetic field in terms of field reduction by the shielding current and temporal stability of the central field. These two issues are known as screening current-induced field (SCIF) and central field drift [2, 4, 11]. Some of these applications, like accelerator, NMR and MRI magnets, require a high-accuracy of the magnetic field in a fiducial volume located around the center of the magnet [3, 5]. Therefore, phenomena like the SCIF and the field drift need to be understood and remedied. For such purpose, numerical models are an irreplaceable tool. Finite element models based on the H formulation have been successfully used during the last years to model systems made with HTS tapes [12]. The H formulation models are limited to the study of medium-size systems, because the study of large-scale systems requires an excessive amount of computational resources. This limitation has favored the emergence of the homogenization [13] and multi-scale methods [14, 15]. A major step towards the reduction of the required computational resources was undergone with the use of the T-A formulation [16, 17]. More recently, T-A homogenous and T-A multi-scale strategies have further reduced the computation time and the amount of memory up to the point that it is possible to achieve real-time simulations of slow ramping cycles of large-scale superconducting systems [18]. In this manuscript, we present the T-A homogenous model of an HTS solenoidal coil. This coil model is used to qualitatively investigate the dependence of the SCIF and field drift on the critical current density (𝐽𝑐) and the 𝑛 power law index. The effect of 𝐽𝑐 has already been addressed in the literature [2, 19]. Here, the effects of 𝑛 are included in the discussion. Indeed, 𝑛, similarly to 𝐽𝑐, depends on quantities such as the magnetic field, the temperature and the stress [20, 21], and, therefore, its variation could have an important impact as well. The T-A homogenous model also allows to study the effectiveness of two remedies for the SCIF and field drift: the striation of the tapes [11] and the current sweep reversal [2, 22]. 2. T-A homogenous approach In this section, a brief description of the T-A homogenous approach is presented. Comprehensive descriptions of the T-A formulation can be found in [16, 17], while the reference of the T-A homogenous approach is given in [18]. The so-called T-A formulation is the combination of the T and the A formulations. The HTS layer of the tapes are considered to be one dimensional (1D) lines. The current vector potential (𝐓) is exclusively defined in the HTS 1D layers, and the magnetic vector potential (𝐀) is defined over the entire system. The two vector potentials are defined by the following relations, 𝐉 = 𝛻 × 𝐓 𝐁 = 𝛻 × 𝐀 (1) (2) where 𝐁 and 𝐉 are the magnetic flux density and the current density, respectively [12]. The governing equations of the T and A formulation are, 𝛻 × 𝜌𝛻 × 𝐓 = − ∂𝐁 𝜕𝑡 , ∇ × ∇ × 𝐀 = 𝜇𝐉 , (3) (4) where 𝜌 is the resistivity and 𝜇 is the magnetic permeability. The study of solenoidal coils can be addressed by means of a 2D axisymmetric model, see figure 1, where 𝐓 and 𝐀 have only one component. Consequently, equations (3) and (4) are simplified to, ∇2𝐴𝜑 = 𝜇𝐽𝜑 𝜕 𝜕𝑧 (𝜌𝐻𝑇𝑆 𝜕𝑇𝑟 𝜕𝑧 ) = 𝜕𝐵𝑟 𝜕𝑡 (5) (6) where 𝜌𝐻𝑇𝑆 is the resistivity of the HTS material. The transport current flowing through the tapes is imposed by the Dirichlet boundary condition on 𝑇𝑟, that must satisfy the following relation, 𝐼 = (𝑇1 − 𝑇2)𝛿 (7) where 𝐼 is the transport current, 𝑇1 and 𝑇2 are the 𝑇𝑟 values at the edges of the superconducting line, and 𝛿 is the thickness of the HTS layer. The latter being infinitely thin, equation (7) imposes a sheet current density 𝐼/𝛿 in practice. Figure 1. 2D axisymmetric representation of a solenoidal coil where the HTS layers are considered to be 1D lines. The T-A formulation combines the T and A formulations, T is defined just along the HTS lines. The homogenization process transforms the stacks of lines into homogeneous bulks. The homogeneous model assumes that a stack of HTS tapes can be represented by a homogeneous bulk, such process is depicted in figure 1. The bulk can be understood as the limit case of a stack where the spaces between adjacent HTS layers disappear. Analogous to the non-homogenous case, 𝐓 is now exclusively defined inside the bulks. Equation (6) is valid again, because the influence of the magnetic field component parallel to the surface of the tapes on the computation of 𝐓 is neglected. Such component, however, influences the local value of 𝐽𝐶 (see equation (9)). The HTS bulks are considered to be surrounded with a medium of zero electrical conductivity, as proposed in [18]. 3. Case study The case study used in this manuscript is a solenoidal coil made of 10 HTS pancakes, already described in [15]. Each pancake coil have 80 turns and each turn has same dimensions. Tables 1 and 2 contain the relevant parameters of the coil and of the HTS tape. Table 1. Coil parameters. Parameter Pancakes Turns/pancake Inner radius Outer radius Height Value 10 80 20 mm 40 mm 44.5 mm Table 2. HTS tape parameters. Parameter Width HTS layer thickness Ec n Jc0 B0 k  Value 4 mm 1 m 1e-4 Vm-1 25 4.5e10 Am-2 0.03 T 0.2 0.6 The electrical resistivity of the HTS material is modeled with the so-called E-J power law, 𝜌𝐻𝑇𝑆 = 𝐸𝑐 𝐽𝑐(𝑩) 𝑱 𝑛−1 𝐽𝑐(𝑩) . (8) It is considered that the critical current density (Jc ) has an anisotropic dependence on the magnetic field, described by the following elliptic relation, 𝐽𝐶(𝑩) = 𝐽𝑐0 2 √𝑘2𝐵𝑧 2 + 𝐵𝑟 𝐵0 𝛼 , ) (1 + (9) where 𝐵𝑧 and 𝐵𝑟 are the magnetic flux density components. The values of the parameters of (8) and (9) are listed in Tables 1 and 2. Since there are no magnetic materials present in the system, its permeability is the one of the vacuum 0. The pancakes, once converted into bulks, is meshed with 60 equally-spaced rectangular elements, along the tape's width. The number of elements along the bulk's thickness is 11, the distribution of these elements is graded with increasing number of elements at the extremities of the bulk. The symmetries of the coil allow employing adequate boundary conditions to model just the right upper quadrant of the full coil. The model was implemented in COMSOL Multiphysics 5.4, with the help of the PDE and AC/DC modules. 4. Screening current-induced field The screening current is induced by the gradual penetration of the magnetic field into the tapes. The magnetic field penetrates from the extremities of the tapes producing current fronts. The screening current affects the magnetic field distribution all over the coil and particularly attenuates the central magnetic field. This attenuation is referred to as screening current-induced field (SCIF) [2], and it is quantified by the following relation, 𝐵𝑆𝐶 = 𝐵𝑠𝑖𝑚 − 𝐵𝑛 (10) where 𝐵𝑠𝑖𝑚 is the central magnetic flux density estimated by means of the model, here by the T-A formulation homogeneous model described above, and 𝐵𝑛 is the central nominal magnetic flux density produced by a uniform distribution of current density. For this case study the central point has the coordinates (0, 0). When 𝐵𝑆𝐶 is plotted as a function of 𝐵𝑛, a hysteresis loop is formed. The simulations in this section consider that the transport current is given by the figure 2. The simulations are extended to one and a half cycles (6 h), this lapse is enough to capture the full hysteretic behavior of the SCIF. The amplitude of the transport current is 26.5 A. This charge cycle represents the real conditions of operations of high-field magnets [9]. Figure 2. Transport current. Triangular shape with a 4 h period, and 26.5 A amplitude. The magnitude of the magnetic flux density produced when the current density is uniform, at the first current peak (𝑡 = 1 h), is shown in figure 3 a). The magnetic flux density estimated with the T-A homogeneous model, i.e. including the screening current effects, at 𝑡 = 1 h, is shown in figure 3 b). For these conditions 𝐵𝑛 = 0.36 T and 𝐵𝑠𝑖𝑚 = 0.345 T, then 𝐵𝑆𝐶 = 0.015 T ( 4.1% ). Even if it is not easy to appreciate the attenuation of the central field, figure 3 clearly shows how the screening current distorts the field all over the coil. The SCIF is known to depend on the coil's shape and values as large as 18% have been reported in [23]. Figure 3. a) Magnetic flux density magnitude at peak current produced by a uniform 𝐽. b) Magnetic flux density magnitude estimated with the T-A homogeneous model in the base conditions {𝐽𝑐 = 𝐽𝑏, 𝑛 = 25}. The black lines show the position of the pancakes. The white stream lines show the field distortions produced by the screening current. 0123456Time [h]-26.5-1301326.5Transport Current [A]The effects of 𝐽𝑐 and 𝑛 in the form and size of the SCIF hysteresis loop are studied hereinafter. The 𝐽𝑐0 and 𝑛 provided in Table 2 are considered to be the reference values for the rest of the study. Two batches of simulations are performed modifying the base conditions. The first batch of simulations modifies the critical current density by a factor at a constant n whereas the second batch modifies n at a constant 𝐽𝑐0. The first batch considers 𝑛 = 25 and 𝐽𝑐0𝑛 = [1.5, 1.0, 0.5, 0.25] ∗ 𝐽𝑐0; the critical current density 𝐽𝑐 remains as defined in (9) replacing 𝐽𝑐0 by 𝐽𝑐0𝑛. Figure 4 shows the normalized current density (𝐽𝑛 = 𝐽 𝐽𝑐⁄ ) at 𝑡 = 1 h for the four new 𝐽𝑐0 values. Figure 4 b) corresponds to the field shown in figure 3 b). It is possible to observe how 𝐽𝑐 influences the current density distribution, when 𝐽𝑐0𝑛 = 1.5 ∗ 𝐽𝑐0, the lower pancakes have non-penetrated zones, while when 𝐽𝑐0𝑛 = 0.25 ∗ 𝐽𝑐0 the current density is almost uniform, but reaching fully saturated values for the upper pancakes. Figure 4. Normalized current density (𝐽𝑛) at peak current, for the cases 𝑛 = 25, 𝐽𝑐0𝑛 = [1.5, 1.0, 0.5, 0.25] ∗ 𝐽𝑐0. The screening current is clearly observed when 𝐽𝑐0𝑛 = 1.5 ∗ 𝐽𝑐0, while an almost uniform 𝐽 is observed when 𝐽𝑐0𝑛 = 0.25 ∗ 𝐽𝑐0. The SCIF hysteresis loops corresponding to the first simulation batch are presented in figure 5 a). The evolution of the current density can explain the hysteresis loops. At the beginning of the cycle, the current gradually penetrates the tapes incrementing the BSC magnitude. Once the tapes are fully penetrated, BSC tends to saturate. When the magnet is discharged, new current fronts penetrate the tapes and the BSC magnitude decreases, but the declining rate is so that a remnant magnetic field remains at the moment of the zero-crossing point of the cycle. The maximum values for 𝐵𝑆𝐶 as well as the 𝐵𝑆𝐶 values at the current peak (𝑚𝑎𝑥(𝐵𝑛)) are presented in figure 5 b). As it could have already been anticipated from figure 4, figure 5 shows that lower 𝐽𝑐 produce lower screening currents and lower 𝐵𝑆𝐶. Figure 5. a) SCIF hysteresis loop for the cases 𝑛 = 25, 𝐽𝑐0𝑛 = [1.5, 1.0, 0.5, 0.25] ∗ 𝐽𝑐0. b) Maximum values of the cases in the left. second batch of and considers simulations 𝐽𝑐0 = 4.5 × 1010Am-2 𝑛 = The [100, 25, 10, 5], including thereby the reference case. The SCIF hysteresis loops corresponding to this second batch are presented in figure 6 a). The corresponding maximum values are presented in figure 6 b). It can be seen that, compared to the impact of the 𝐽𝑐0, the n has a tendency to round the extremities of the hysteresis loops when the magnetic field changes direction. Similarly to the previous batch, as the n is decreased, the current distribution tends to be more uniform and the hysteresis loops flattens. When n and 𝐽𝑐0 are decreased, the currents are induced in a larger portion of the HTS layers and the pancakes can saturate at a lower transport current. Figure 6. a) SCIF hysteresis loop for the cases 𝐽𝑐0 = 4.5 × 1010A. m2, 𝑛 = [100, 25, 10, 5]. b) Maximum values of the cases in the left. 5. Field drift Once the coil is charged and the transport current is fixed, the flux creep inside the HTS material causes the relaxation of the screening current and the SCIF, which in turn produces a drift in the central field. In this section, the effect of 𝐽𝑐 and 𝑛 on central field drift is investigated. As in the previous section, the two batches of simulations are considered anew. However, the transport current is a single ramp that reaches a maximum value equal to 26.5 A at 𝑡 = 1 h, followed by a plateau that extends for 30 h to observe the field drift. Figure 7. Central field drift for the different cases, the plot starts at 𝑡 = 1 h once the maximum current is reached. Figure 7 shows the estimated 𝐵𝑠𝑖𝑚 for the tested conditions. The x-axis has a logarithmic scale which starts at 𝑡 = 1 h and extends to 30 h. The larger drifts occur during the first hours, and a linear growth with logarithmic time is observed during the last hours. The case {𝐽𝑐0, 𝑛 = 10} (orange dashed line) exhibits the larger drift, from 0.352 T to 0.36 T ( 2.2% ). The case {1.5 ∗ 𝐽𝑐0, 𝑛 = 25} and { 𝐽𝑐0, 𝑛 = 25} (blue and black solid lines) have the same 1.5% drift. The cases represented by the red solid, cyan solid and black dashed lines have a small drift because they are close to the nominal 𝐵𝑛 = 0.36 𝑇, and as can be seen in figure 4 d) the 𝐽 is almost uniform. The case {𝐽𝑐0, 𝑛 = 100} green dashed line has a small 0.4% drift, even when it is away from 𝐵𝑛. The conclusion drawn from these simulations is that the size of the drift is a function of the 𝑛 index, the smaller 𝑛, the larger the drift. The case with 𝑛 = 100 is closer to the critical state where no flux creep is modeled. 6. Striation and current sweep reversal There are some remedies for the SCIF and the field drift. For example, the SCIF can be reduced by applying an external AC magnetic field forcing the "vortex shaking effect" [24]. A method to reduce the field drift is by lowering the temperature once the operating current is reached, then the critical current is increased and the flux creep diminishes [2]. In this section, two of the common remedies are explored for the SCIF and the field drift problems: i) the striation of the HTS material, and ii) the current sweep. The striation process is the subdivision of the tapes into narrow strips, this process is known to reduce the SCIF [3]. The current sweep reversal is the overshooting of the transport current, this process reduces the field drift [2, 22] In order to implement the striation in the simulations, the homogeneous bulks, representing the pancakes, are subdivided into narrow pancakes. The boundary conditions for T are modified to force that each new strip forming the narrow pancakes transport a fraction of the original transport current. These conditions correspond to having the strips perfectly uncoupled. The T-A homogenous model for the striation case was simulated considering the triangular shape transport current and the reference values used in section 4 and found in Table 2. Figure 8 shows the normalized current density (𝐽𝑛) at 𝑡 = 1 h for the tapes with no striation, taken as a reference case, and for the cases of tapes subdivided into 2, 4 and 8 strips. Figure 8. Normalized current density (𝐽𝑛) at peak current, for the cases with and without striations. The striation process causes the formation of current fronts in each strip, and the penetration of the magnetic field from the extremities of the narrow strips produce the overall effect of the reduction of the SCIF. The SCIF hysteresis loops corresponding to the cases with striations are presented in figure 9 a). The maximum values for 𝐵𝑆𝐶 are presented in figure 9 b). Figure 9. a) SCIF hysteresis loop for the cases with striations. b) Maximum values of the cases in the left. The ramp transport current used in section 5 is modified to implemented the current sweep reversal, the ramp rate is increased, current is overshot and subsequently reduced to reach 26.5 A at 𝑡 = 1 h. Three simulations were performed, one with 5%, 10% and 20% overshoots. These ramps are shown in figure. The reduction of the current after the overshoot produces the appearance of new current fronts at the extremities of the tapes, thus the flux creep and the field drift are reduced. Figure 10 shows the estimated 𝐵𝑠𝑖𝑚 for the two overshot ramps and the ramp without overshoot. The case with 5% overshoot reduces the drift just during the first minutes. The case with 10% overshoot successfully "eliminates" the drift during the first hours, but around 𝑡 = 5.5 h the drift starts again due to a relaxation of the screening current according to the proposed model. Originally the reappearance of the drift was the motivation to test the 20% overshoot case. It is possible to appreciate that the current sweep reversal with 20% overshoot successfully eliminates the field drift, but during the first minutes a small negative field drift can be appreciated. These results demonstrate that the 1% overshoot reported in [2, 25] is insufficient for our proposed coil and the choice of the overshoot should be determined on a coil basis to address effectively the temporal drift of the central magnetic field. From the lines in figure 10, we can also conclude that the current sweep reversal is an appropriate method to reduce the SCIF. Figure 10. a) Ramps with overshoot. b) Estimated central field for the ramps with and without overshoot. The field drift is partially eliminated with 5% and 10% overshoot, and successfully "eliminated" with a 20% overshoot. The computer used to run the simulations is a desktop computer (6 cores, Intel (R) Xeon(R) ES-2630, 2.2 GHz, 64 GB RAM). The time steps taken by the solver were adjusted to optimize the computation time, this is important because the simulated lapse in the simulations conducted to study the field drift is 3 h, but the time steps can be enlarged when the current is constant. The average computation time required to complete the simulations is 1 h, for both triangular and ramp transport currents. Both simulated lapses, 6 h and 30 h, are larger than the 1 h computation time, therefore, the T-A homogeneous models, in the following case studies, could have been performed in real-time during the ramping of the magnet. 7. Conclusions The penetration of the magnetic field into the HTS tapes causes screening currents. The coils wound with HTS tapes face two major problems caused by the screening current, the SCIF and the field drift. These problems, amongst others, must be understood and solved to pave the way for the use of HTS coils in applications like NMR, MRI, high field magnets, etc. The effects of 𝐽𝑐 and 𝑛 in both problems (SCIF and field drift) were studied by means of several numerical simulations. It was demonstrated that higher 𝐽𝑐 values cause higher screening current, while higher 𝑛 values hinder the relaxation of such screening current. The main conclusion is that the reduction of both quantities 𝐽𝑐 and 𝑛 causes the attenuation of the mentioned problems but they go against the use of high performant HTS tape technologies for which greater critical currents at larger n are sought to the generation of higher magnetic field. Fortunately, the striation of the HTS tapes and the current sweep reversal are effective methods to mitigate the SCIF and the field drift. However, simulations showed that the field drift could reappear for current sweep reversal with 5 and 10% overshoot. This effect could be avoided if the overshoot is increased to 20%. It is likely that the magnitude of the overshoot will depend on the coil configuration as well as the ramping cycle and studies should be performed for each magnet for different operating cycles. The efficiency of the T- A homogenous models was exploited in this paper to run dozens of simulations allowing studying the behavior of the magnet during its operation. It should be mentioned that the use of other modeling strategies such as the ones based on the H formulation would have required computation times in the order of months to obtain the same conclusion. References [1] Melhem Z 2012 High temperature superconductors (HTS) for energy applications (Cambridge, U.K.: Woodhead Publishing). [2] Maeda, H. and Yanagisawa, Y. (2014). Recent Developments in High-Temperature (Review). IEEE Transactions on Applied Superconducting Magnet Technology Superconductivity, 24(3), pp.1-12. [3] Amemiya, N. and Akachi, K. (2008). Magnetic field generated by shielding current in high Tc superconducting coils for NMR magnets. Superconductor Science and Technology, 21(9), p.095001. [4] Yanagisawa, Y., Kominato, Y., Nakagome, H., Hu, R., Takematsu, T., Takao, T., Uglietti, D., Kiyoshi, T., Takahashi, M. and Maeda, H. (2011). Magnitude of the Screening Field for YBCO Coils. IEEE Transactions on Applied Superconductivity, 21(3), pp.1640-1643. [5] Ueda, H., Fukuda, M., Hatanaka, K., Tao Wang, Ishiyama, A. and Noguchi, S. (2013). Spatial and Temporal Behavior of Magnetic Field Distribution Due to Shielding Current in HTS Coil for Cyclotron Application. IEEE Transactions on Applied Superconductivity, 23(3), pp.4100805-4100805. [6] Weijers, H., Trociewitz, U., Markiewicz, W., Jiang, J., Myers, D., Hellstrom, E., Xu, A., Jaroszynski, J., Noyes, P., Viouchkov, Y. and Larbalestier, D. (2010). High Field Magnets With HTS Conductors. IEEE Transactions on Applied Superconductivity, 20(3), pp.576-582. [7] Benkel, T., Miyoshi, Y., Escamez, G., Gonzales, D., Chaud, X., Badel, A. and Tixador, P. (2016). REBCO Performance at High Field With Low Incident Angle and Preliminary Tests for a 10-T Insert. IEEE Transactions on Applied Superconductivity, 26(3), pp.1-5. [8] Benkel, T., Richel, N., Badel, A., Chaud, X., Lecrevisse, T., Borgnolutti, F., Fazilleau, P., Takahashi, K., Awaji, S. and Tixador, P. (2017). Preliminary Tests and Margin Estimate for a REBCO Insulated 10 T Insert Under High Magnetic Field. IEEE Transactions on Applied Superconductivity, 27(4), pp.1-5. [9] Weijers, H., Markiewicz, W., Gavrilin, A., Voran, A., Viouchkov, Y., Gundlach, S., Noyes, P., Abraimov, D., Bai, H., Hannahs, S. and Murphy, T. (2016). Progress in the Development and Construction of a 32-T Superconducting Magnet. IEEE Transactions on Applied Superconductivity, 26(4), pp.1-7. [10] Berrospe-Juarez, E., Zermeno, V., Trillaud, F., Gavrilin, A., Grilli, F., Abraimov, D., Hilton, D. and Weijers, H. (2018). Estimation of Losses in the (RE)BCO Two-Coil Insert of the NHMFL 32 T All-Superconducting Magnet. IEEE Transactions on Applied Superconductivity, 28(3), pp.1-5. [11] Yanagisawa, Y., Xu, Y., Jin, X., Nakagome, H. and Maeda, H. (2015). Reduction of Screening Current-Induced Magnetic Field of REBCO Coils by the Use of Multi-Filamentary Tapes. IEEE Transactions on Applied Superconductivity, 25(3), pp.1-5. [12] Brambilla, R., Grilli, F. and Martini, L. (2006). Development of an edge-element model for AC loss computation of high-temperature superconductors. Superconductor Science and Technology, 20(1), pp.16-24. [13] Zermeno, V., Abrahamsen, A., Mijatovic, N., Jensen, B. and Sørensen, M. (2013). Calculation of alternating current losses in stacks and coils made of second generation high temperature superconducting tapes for large scale applications. Journal of Applied Physics, 114(17), p.173901. [14] Quéval, L., Zermeño, V. and Grilli, F. (2016). Numerical models for ac loss calculation in large-scale applications of HTS coated conductors. Superconductor Science and Technology, 29(2), p.024007. [15] Berrospe-Juarez, E., Zermeño, V., Trillaud, F. and Grilli, F. (2018). Iterative multi-scale method for estimation of hysteresis losses and current density in large-scale HTS systems. Superconductor Science and Technology, 31(9), p.095002. [16] Zhang, H., Zhang, M. and Yuan, W. (2016). An efficient 3D finite element method model based on the T -- A formulation for superconducting coated conductors. Superconductor Science and Technology, 30(2), p.024005. [17] Liang, F., Venuturumilli, S., Zhang, H., Zhang, M., Kvitkovic, J., Pamidi, S., Wang, Y. and Yuan, W. (2017). A finite element model for simulating second generation high temperature superconducting coils/stacks with large number of turns. Journal of Applied Physics, 122(4), p.043903. [18] Berrospe-Juarez, E., Zermeño, V., Trillaud, F. and Grilli, F. (2019). Real-time simulation of the T -- A large-scale HTS systems: multi-scale and homogeneous models using formulation. Superconductor Science and Technology, 32(6), p.065003. [19] Wang, M., Li, Z., Sheng, J., Dong, F., Zhong, L., Duan, X., Song, M., Hong, Z. and Jin, Z. (2019). Study on the Screening Current Induced Field of a Novel Narrow HTS Tape with 1 mm Width. IEEE Transactions on Applied Superconductivity, 29(5), pp.1-5. [20] Shin, H., Kim, K., Dizon, J., Kim, T., Ko, R. and Oh, S. (2005). The strain effect on critical current in YBCO coated conductors with different stabilizing layers. Superconductor Science and Technology, 18(12), pp.S364-S368. [21] Sass, F., Sotelo, G., Junior, R. and Sirois, F. (2015). H-formulation for simulating levitation forces acting on HTS bulks and stacks of 2G coated conductors. Superconductor Science and Technology, 28(12), p.125012. [22] Uglietti, D., Yanagisawa, Y., Maeda, H. and Kiyoshi, T. (2010). Measurements of magnetic field induced by screening currents in YBCO solenoid coils. Superconductor Science and Technology, 23(11), p.115002. [23] Yanagisawa, Y., Nakagome, H., Uglietti, D., Kiyoshi, T., Ruixin Hu, Takematsu, T., Takao, T., Takahashi, M. and Maeda, H. (2010). Effect of YBCO-Coil Shape on the Screening Current-Induced Magnetic Field Intensity. IEEE Transactions on Applied Superconductivity, 20(3), pp.744-747. [24] Kim, K., Song, J., Yang, D., Kim, Y., Kim, T., Kim, S., Park, M. and Lee, H. (2016). Study on in pancake-type non-insulated HTS elimination of screening-current-induced coil. Superconductor Science and Technology, 29(3), p.035009. field [25] Zhang, M., Yuan, W., Hilton, D., Canassy, M. and Trociewitz, U. (2014). Study of second- screening high-temperature generation effect. Superconductor Science and Technology, 27(9), p.095010. superconducting magnets: the self-field
1910.07028	1	1910	2019-10-01T07:32:38	Techno-economic analysis of solar PV power-to-heat-to-power storage and trigeneration in the residential sector	[ "physics.app-ph" ]	This article assesses whether it is profitable to store solar PV electricity in the form of heat and convert it back to electricity on demand. The impact of a number of technical and economic parameters on the profitability of a self-consumption residential system located in Madrid is assessed. The proposed solution comprises two kinds of heat stores: a low- or medium-grade heat store for domestic hot water and space heating, and a high-grade heat store for combined heat and power generation. Two cases are considered where the energy that is wasted during the conversion of heat into electricity is employed to satisfy either the heating demand, or both heating and cooling demands by using a thermally-driven heat pump. We compare these solutions against a reference case that relies on the consumption of grid electricity and natural gas and uses an electrically-driven heat pump for cooling. The results show that, under relatively favourable conditions, the proposed solution that uses an electrically-driven heat pump could provide electricity savings in the range of 70-90% with a payback period of 12-15 years, plus an additional 10-20% reduction in the fuel consumption. Shorter payback periods, lower than 10 years, could be attained by using a highly efficient thermally driven heat pump, at the expense of increasing the fuel consumption and the greenhouse gas emissions. Hybridising this solution with solar thermal heating could enable significant savings on the global emissions, whilst keeping a high amount of savings in grid electricity (> 70 %) and a reasonably short payback period (< 12 years).	physics.app-ph	physics	Techno-economic analysis of solar PV power-to- heat-to-power storage and trigeneration in the residential sector A. Datas1,2, A. Ramos1,2, and C. del Cañizo1 1Instituto de Energía Solar, Universidad Politécnica de Madrid, 28040 Madrid, Spain. 2Universitat Politècnica de Catalunya, Jordi Girona 1-3, Barcelona 08034, Spain. Keywords: PHPS, CCHP, TES, trigeneration, cogeneration, energy storage, heat, thermal, photovoltaics, solar, self-consumption, thermophotovoltaics, Stirling, ultra-high temperature. Abstract This article assesses whether it is profitable to store solar PV electricity in the form of heat and convert it back to electricity on demand. The impact of a number of technical and economic parameters on the profitability of a self-consumption residential system located in Madrid is assessed. The proposed solution comprises two kinds of heat stores: a low- or medium-grade heat store for domestic hot water and space heating, and a high-grade heat store for combined heat and power generation. Two cases are considered where the energy that is wasted during the conversion of heat into electricity is employed to satisfy either the heating demand, or both heating and cooling demands by using a thermally-driven heat pump. We compare these solutions against a reference case that relies on the consumption of grid electricity and natural gas and uses an electrically-driven heat pump for cooling. The results show that, under relatively favourable conditions, the proposed solution that uses an electrically-driven heat pump could provide electricity savings in the range of 70 -- 90% with a payback period of 12 -- 15 years, plus an additional 10 -- 20% reduction in the fuel consumption. Shorter payback periods, lower than 10 years, could be attained by using a highly efficient thermally driven heat pump, at the expense of increasing the fuel consumption and the greenhouse gas emissions. Hybridising this solution with solar thermal heating could enable significant savings on the global emissions, whilst keeping a high amount of savings in grid electricity (> 70 %) and a reasonably short payback period (< 12 years). 1 1. Introduction The energy consumption in buildings represents around 40% of the world energy consumption and one-third of global CO2 emissions [1]. Thus, during the last decade a number of technological solutions have been proposed targeting the improvement of the energy efficiency and the reduction of the CO2 emissions associated to energy consumption in the building sector. Among them, the so-called combined cooling, heating and power (CCHP), or trigeneration systems [2] -- [6] are of especial interest, as they are intended to reduce the emissions associated to the three main kinds of energy demands in a building: electricity, heating and cooling. A typical CCHP system comprises a power generation unit (PGU), e.g. a microturbine, and a thermally-driven heat pump (THP) [7], e.g. an absorption chiller. The PGU is typically powered by natural gas and produces electricity and heat as by-product. This heat can be directly used for heating in winter or transferred to a THP for cooling in summer. Compared with the solution based on large centralized power plants and local air-conditioning systems, distributed CCHP provides a significant improvement of fuel utilization, in the range of 70 to 90% with respect to 30 -- 45% of centralized power plants [2]. Further reduction of the emissions is possible by fully or partially replacing natural gas by solar thermal heating [8]. Most of the studies consider organic Rankine cycles (ORC) [9], [10] in this case, as they enable operation temperatures below 220 ºC, meeting the requirements of non-concentrating solar collectors that could be used in residential and commercial environments. However, the low exergy content of such a low- grade heat results in very low PGU conversion efficiencies, typically below 10 -- 15% [9], [11], [12], which precludes the achievement of a clear economic advantage. In the last decade, the dramatic cost reductions of solar PV technology have triggered the interest on self-consumption of PV electricity in both commercial and residential buildings [13]. With an average growth rate of 51% [14], new solar PV power additions in 2017 accounted up to 98 GW, out-stripping the 70 GW of net fossil fuel generating capacity added the same year. Small-scale solar PV systems in distributed applications, mostly building integration, accounts for a significant share of these new PV additions (~ 38% [15]). Only in China, distributed capacity additions in 2017 summed up to 19.4 GW, and new rooftop systems saw a three-fold increase relative to 2016 [16]. Solar PV installations are simple, reliable, and do not require high maintenance. Thus, they are very appealing to produce electricity in small distributed applications. In the residential context, many solutions have been proposed to integrate solar PV systems with CCHP systems, including the hybridization of solar PV with gas-powered CCHP [17] -- [20], the use of hybrid PV/thermal (PVT) solar collectors, using both non-concentrating [10], [21] -- [24] and concentrating technologies [25], the integration of PV systems with electrically-driven heat pumps (EHP) [26] -- [28], as well as the direct use of PV electricity for heat production [29]. 2 In all these cases, the self-consumption of PV electricity is limited due to the lack of synchronization between solar irradiance and local consumption. For systems where PV electricity is used exclusively to satisfy the electric power demand, self-consumption ratio (defined here as the fraction of the PV generated electricity that is used to supply the loads) is usually limited to 25 -- 30% [30]. Self-consumption can be increased in systems where solar PV electricity is also used to provide heating and cooling [13], [26], [31], [32]. However, in this case, the attainable self-consumption ratio strongly depends on the climate conditions and the building standardized heating load. For instance, in old building standards located in Stuttgart (Germany), the combination of a heat pump and a hot water storage enable self-consumption increments of ~ 35% [26]. However, for next generation zero-energy buildings the same solution provides self-consumption potential increments of only ~ 10%, as energy demand of appliances dominates compared to the electrical energy demand to drive the heat pump [26]. To increase the self-consumption levels, solutions must integrate a system that stores the excess of PV electricity, such as electrochemical batteries [17] -- [19], [33]. However, current prices for stationary residential battery storage are prohibitive, exceeding 1000 US$2015 per kWh of electricity storage capacity [34]. Even in the case of reaching global cumulative storage capacities of 1000 GWh (from a current cumulative capacity of ~ 1 GWh), empirical learning curves project that the future cost of stationary residential electricity storage, regardless of the technology type, will be in the range of ~ 340 $2015/kWhel [34]. Additionally, in most CCHP systems lifetime (> 20 years), the batteries require replacement (4 -15 years) and this can have a significant impact on the lifetime cost of the full installation [17]. Consequently, when looking at the profitability of the CCHP solution, the high capital cost of batteries results in optimal systems with a relatively small storage capacity, consequently providing small self-consumption improvements in the range of 13 -- 24% [13]. For this reason, finding low-cost alternatives for electricity storage in the residential sector is an important field of research today [35], [36]. Although it might sound counterintuitive, among the many possible energy storage options, those that are particularly interesting for CCHP applications are those with low round-trip (electric-to-electric) efficiency, as they may deliver part of the stored energy in a form of heat. Some examples are compressed air [37] or hydrogen storage combined with fuel cells [38]. In these cases, the comparatively low round-trip efficiency can be compensated by the (eventual) lower cost of the technology and the profitable use of the exhaust heat. In this regard, a potentially low-cost alternative for electricity storage that has not received much attention is the power-to-heat-to-power storage (PHPS) concept. PHPS involves the conversion of electricity into heat, which is then stored and later converted back to electricity on demand. The dramatic cost reduction of solar PV electricity along with the potentially lower capital costs of PHPS might result in a profitable PV+PHPS solution in the context of CCHP applications, where the 3 low-grade heat produced during the heat-to-power conversion process may be used for satisfying both heating and cooling demands. Today, the use of thermal storage for power generation is virtually limited to concentrated solar power (CSP) plants, where the store temperature rarely surpasses ~ 500 ºC [39]. On the contrary PHPS systems could theoretically reach ultra-high temperatures (> 1000 ºC), subsequently enabling very high heat-to-power (H2P) conversion efficiencies of ~ 40% or beyond. This is around 3 - 4 times greater than that of low-temperature solar thermal ORC systems discussed above [9], [11], [12]. A hypothetical system comprising a 40% efficient solar thermal collector and a 10% efficient ORC [40], [9] would produce an overall solar-to-electric conversion efficiency of 4%, which is half the efficiency than that of a PHPS system comprising 20% efficient solar PV modules and 40% efficient PGU. Besides, the use of high grade heat store would potentially enable much more efficient THP cooling, as well as higher stored energy densities [41], [42], which is a remarkable advantage in space constrained residential applications. Recent studies have established a few conceptual PHPS embodiments, which differentiate in the way that heat is produced, stored, and converted back into electricity. A particularly promising concept is pumped heat electricity storage (PHES), in which a high temperature heat pump cycle transforms electricity into heat, which is stored inside two large regenerators, and a thermal engine cycle transforms the stored heat back into electricity [43] -- [45]. PHES has a high theoretical round-trip efficiency (RTE) in the range of 40 -- 70%, depending on the operational temperature range [44]. Predominately conceived for large grid-electricity storage applications, the potential viability of PHES in the residential sector has not been assessed yet. Other conceptually simpler approaches consider the use of ultra-high temperature (> 1000 ºC) joule heating for sensible- [46] and latent- [41] heat storage combined with a thermophotovoltaic (TPV) power generation. Despite having lower RTE potential (less than ~ 40%), these designs might bring some advantages, as the modularity and the lack of moving parts. Solutions based on the use of high temperature heat pumps have been also recently proposed to mitigate the high thermal losses that could be eventually derived from operating at ultra-high temperatures; theoretically enabling an increment of the round-trip conversion efficiency up to ~ 50% [47], [48]. Regardless of the particular system implementation, it is still unclear under what circumstances a PHPS system could be profitable. In [46], [49] the minimum tolerable RTE of an energy storage system used for grid-electricity storage is estimated at 36% for the case of Pennsylvania- New-Jersey-Maryland grid in 2017. However, this study assumes that the electricity is purchased from the grid, and therefore, the minimum RTE, is entirely determined by the ratio between on-peak and off-peak electrical prices in a very specific case. Besides, the exhaust heat produced by the PGU is not used, which is detrimental for the profitability of a PHPS solution. 4 In this study we assess the integration of a PHPS system in a CCHP solution for the self- consumption of solar PV electricity in the residential environment. To the best of our knowledge, no comprehensive techno-economical assessment of PV generation coupled with a PHPS system has been evaluated so far. Thus, we will answer some of the most fundamental questions regarding the profitability of this solution, such as the maximum cost, the minimum PGU conversion efficiency, or the maximum heat insulation losses that are tolerable in order to provide reasonably low payback periods and significant energy savings. 2. System model and methodology Figure 1 shows the three kinds of system configurations that are analysed in this work. The reference case (Figure 1-a) comprises a conventional boiler (for heating) and an EHP (for cooling). In this case, all the energy consumption, either electricity (𝐶𝑒) or heat (𝐶ℎ), is obtained from the retail markets. This reference case will be used to evaluate the relative improvements of the proposed solutions incorporating a PHPS system. Figure 1-b illustrates a PHPS configuration comprising a solar PV system, high-grade thermal energy store (HTES), a PGU, a low- or mid-grade thermal energy store (LTES), and an EHP. Grid electricity and fuel are used as backups to ensure reliability of energy supply. In what follows, this configuration will be named PHPS-E. Figure 1-c illustrates a very similar configuration but using a thermally-driven heat pump (THP), instead of an EHP. This configuration will be named PHPS-T. In both configurations the PV electricity (𝐺𝑒) is either i) used directly to satisfy the electric demand, ii) stored as high-grade heat in the HTES (𝑃𝑖𝑛,𝐻𝑇𝐸𝑆), iii) stored as low-grade heat in the LTES (𝑃𝑖𝑛,𝐿𝑇𝐸𝑆), or iv) lost (𝑃𝑙𝑜𝑠𝑠,𝑃𝑉) if there is no electricity consumption and both HTES and LTES stores are at their maximum capacities (𝐸𝐻𝑇𝐸𝑆,𝑚𝑎𝑥 and 𝐸𝐿𝑇𝐸𝑆,𝑚𝑎𝑥, respectively). The possibility of selling the excesses of PV electricity to the grid is not considered in this study. The high- grade heat stored in the HTES can be supplied on demand to the PGU (𝑄𝑖𝑛,𝑃𝐺𝑈) to produce electricity with a conversion efficiency 𝜂𝑃𝐺𝑈 (𝑃𝑜𝑢𝑡,𝑃𝐺𝑈 = 𝜂𝑃𝐺𝑈𝑄𝑖𝑛,𝑃𝐺𝑈). During this process, the produced low-grade exhaust heat 𝑄𝑜𝑢𝑡,𝑃𝐺𝑈 = (1 − 𝜂𝑃𝐺𝑈)𝑄𝑖𝑛,𝑃𝐺𝑈 can be either stored in the LTES (𝑄𝑖𝑛,𝐿𝑇𝐸𝑆), or lost (𝑄𝑙𝑜𝑠𝑠,𝑃𝐺𝑈) if the LTES is at maximum capacity. Additional losses take place due to the non-ideal thermal insulation of the LTES ( 𝑄𝑙𝑜𝑠𝑠,𝐿𝑇𝐸𝑆 ) and the HTES (𝑄𝑙𝑜𝑠𝑠,𝐻𝑇𝐸𝑆). All these kinds of losses (𝑄𝑙𝑜𝑠𝑠 = 𝑄𝑙𝑜𝑠𝑠,𝐻𝑇𝐸𝑆 + 𝑄𝑙𝑜𝑠𝑠,𝐿𝑇𝐸𝑆 + 𝑄𝑙𝑜𝑠𝑠,𝑃𝐺𝑈 + 𝑃𝑃𝑉,𝑙𝑜𝑠𝑠) contribute to reduce the self-consumption ratio, defined in this work as 𝑆𝐶 = 1 − ∑ 8760𝑇 𝑡=1 𝑄𝑙𝑜𝑠𝑠(𝑡) ⁄ ∑ 8760𝑇 𝑡=1 𝐺𝑒(𝑡) system lifetime 𝑇 (in years). , being 8760 × 𝑇 the total number of hours during the entire The cooling power (𝐶𝑐) is satisfied by either an EHP (case PHPS-E, Figure 1-b) or a THP (case PHPS-T, Figure 1-c). The former case implies an extra consumption of electricity (𝑃𝐸𝐻𝑃 = 5 𝐶𝑐 𝐶𝑂𝑃𝐸𝐻𝑃 ⁄ ), whereas the latter implies an extra consumption of heat (𝐶ℎ𝑐 = 𝐶𝑐 𝐶𝑂𝑃𝑇𝐻𝑃 ⁄ ). The coefficient of performance (COP) of each device indicates the ratio between the useful cooling provided to the input energy, either heat (THP) or electricity (EHP). The low-grade heat in the LTES can be used either to satisfy the space heating and domestic hot water demands (𝐶ℎ = 𝐶ℎℎ) in the PHPS-E configuration (Figure 1-b) or both heating and cooling demands (𝐶ℎ = 𝐶ℎℎ + 𝐶ℎ𝑐) in the PHPS-T configuration (Figure 1-c). In both cases, additional heating from the external boiler (𝑄𝑒𝑥𝑡) might be necessary to ensure supply reliability. Notice that in the PHPS-T configuration (Figure 1-c) an additional heat coming from the boiler may be needed for satisfying the cooling demand. Hybrid absorption-compression heat pumps [50] enabling both heat and electricity inputs might be interesting in this application, but they are not considered in this study for the sake of simplicity. In this study we use simplified models for each device of the PHPS system. The detailed model equations are shown in Figure 2, which also illustrates the energy management algorithm. At every time step (𝛥𝑡 = 1 hour), the energy rates (in kWel and kWth) shown in Figure 1, and the stored energy (in kWhth) in the HTES and LTES (𝐸𝐻𝑇𝐸𝑆 and 𝐸𝐿𝑇𝐸𝑆, respectively) are calculated following the procedure illustrated in Figure 2. This algorithm first evaluates whether there is an excess or defect of generated PV electricity, i.e. whether the net consumed electrical power 𝑃𝑛𝑒𝑡 = 𝐶𝑒 − 𝐺𝑒 is negative or positive, respectively. If demand exceeds the PV generation (𝐶𝑒 < 𝐺𝑒), all the PV electricity is directly used to satisfy that demand. The additional electricity ( 𝑃𝑛𝑒𝑡 ) is supplied by either the PGU, if the HTES has enough stored heat ( 𝐸𝐻𝑇𝐸𝑆 > 𝑃𝑛𝑒𝑡𝛥𝑡 𝜂𝑃𝐺𝑈 ⁄ ), or by the electrical grid, in the opposite case. If the electrical power demand is higher than the maximum power capacity of the PGU (𝑃𝑛𝑒𝑡 > 𝑃𝑚𝑎𝑥,𝑃𝐺𝑈), both the PGU and electrical grid contribute to satisfy such demand. The low-grade heat produced during the operation of the PGU (𝑄𝑜𝑢𝑡,𝑃𝐺𝑈 ) may be stored in the LTES, if this is not already at its maximum capacity. In the case the PV generation exceeds the consumption of electricity (𝐺𝑒 > 𝐶𝑒), such excess may be stored as high-grade heat in the HTES or as low-grade heat in the LTES. In principle, the system will prioritize the charge of the HTES, as it stores a higher-grade heat that can be later converted into electricity by the PGU. There is only one scenario where this excess of PV electricity is stored in the LTES instead of HTES. This is the case that there is heat consumption (𝐶ℎ = 𝐶ℎℎ + 𝐶ℎ𝑐 > 0), the HTES charge is high (𝐸𝐻𝑇𝐸𝑆 > 𝑥𝐸𝐻𝑇𝐸𝑆,𝑚𝑎𝑥), and the LTES charge is low (𝐸𝐿𝑇𝐸𝑆 < 𝑦𝐸𝐿𝑇𝐸𝑆,𝑚𝑎𝑥). In this study we have fixed 𝑥 = 0.99 and 𝑦 = 0.1, which means that this situation is very improbable, and the vast majority of the excesses of PV electricity is stored in the HTES rather than in the LTES. The optimization of the values of the parameters 𝑥 and 𝑦 is out of the scope of this work. In the case that both LTES and HTES are at their maximum capacities, the PV electricity is inevitably lost, contributing to the increase of the total energy losses of the system (𝑄𝑙𝑜𝑠𝑠). Finally, the total heat consumption (𝐶ℎ = 𝐶ℎℎ + 6 𝐶ℎ𝑐) is satisfied either by the LTES, the boiler, or both combined, depending on whether the amount of stored heat in the LTES is enough to fully satisfy the heat demand. The amount of heat stored in the HTES and LTES is updated at every time step by evaluating the corresponding energy balance equations, as shown at the bottom of Figure 1. The model described above assumes no exergy degradation of stored heat in both HTES and LTES. A non-degraded heat flowing out of the high-grade HTES enables assuming a constant PGU conversion efficiency during the full discharge cycle. This assumption is reasonable especially for thermal stores based on latent heat, where exergy losses are minimum. The energy losses in the LTES are approximated using the standard EN 12977-3 for hot water stores, according to which the heat loss rate in W/K is 𝑈𝐴𝐿𝑇𝐸𝑆 = 𝑎𝐿𝑇𝐸𝑆√𝑉𝐿𝑇𝐸𝑆, being 𝑉𝐿𝑇𝐸𝑆 the water volume in litres, and 𝑎𝐿𝑇𝐸𝑆 a constant that usually takes a value of ~ 0.1 for good insulated hot water storage stores. The overall heat losses (in W) are approximated by 𝑄𝑙𝑜𝑠𝑠,𝐿𝑇𝐸𝑆 = 𝑎𝐿𝑇𝐸𝑆√𝑉𝐿𝑇𝐸𝑆∆𝑇𝐿𝑇𝐸𝑆 , being ∆𝑇𝐿𝑇𝐸𝑆 the mean difference between LTES and ambient temperatures (assumed ∆𝑇𝐿𝑇𝐸𝑆 = 70 ºC for the PHPS-E system and ∆𝑇𝐿𝑇𝐸𝑆 = 130 ºC for the PHPS-T one, as the later requires higher temperature to operate a highly efficient thermally driven heat pump). The heat losses in the HTES are calculated using a similar approach, so that 𝑄𝑙𝑜𝑠𝑠,𝐻𝑇𝐸𝑆 = 𝑎𝐻𝑇𝐸𝑆√𝑉𝐻𝑇𝐸𝑆∆𝑇𝐻𝑇𝐸𝑆 , being 𝑎𝐻𝑇𝐸𝑆 an unknown parameter that should be eventually determined for high temperature heat stores.To relate the volume (𝑉𝐻𝑇𝐸𝑆 and 𝑉𝐿𝑇𝐸𝑆) with the storage maximum capacity (𝐸𝐻𝑇𝐸𝑆,𝑚𝑎𝑥 and 𝐸𝐿𝑇𝐸𝑆,𝑚𝑎𝑥) we assume an energy storage density of 0.08 kWhth/l for the LTES, corresponding to that of a hot water store, while for the HTES, the energy density strongly depends on the storage media and its operational temperature. In this study we assume a latent heat store with an energy density that depends on the operational following equation: 𝐸𝑑,𝑇𝐸𝑆 = 6.74 × 10−7(𝑇𝑇𝐸𝑆)2 − 1.72 × 10−4𝑇𝑇𝐸𝑆 + 0.140 that has been obtained from fitting the latent heat of temperature according to the fusion of a few pure metals with different melting points in the range of 230 -- 2070 ºC (Sn, Zn, Al, Si and B) [41]. This equation captures the fact that higher temperatures enable higher stored energy densities. For instance, a 20 kWhth HTES operating at 1414 ºC (which will be the case if silicon is used as phase-change material) would have an energy density of 1.24 kWhth/l (silicon latent heat). Assuming an optimistic value for 𝑎𝐻𝑇𝐸𝑆 of 0.1, such HTES system would lose about 2.8% of its maximum stored energy in one hour. As a reference, a hot-water store of the same capacity and with 𝑎𝐿𝑇𝐸𝑆 = 0.1 would lose only 0.24% per hour. For a HTES at 800 ºC, heat losses would be only very slightly reduced to 2.7% (assuming the same value of 𝑎𝐻𝑇𝐸𝑆). This is because the lower temperature also brings a lower energy density, counteracting to keep a similar amount of heat losses. The evident way to reduce the losses is the use of advanced thermal insulation designs with a very small 𝑎𝐻𝑇𝐸𝑆 value. Possible options include the use of 7 vacuum insulation or other advance concepts, such as the use of heat pumps [47]. Whether such low 𝑎𝐻𝑇𝐸𝑆 values are attainable by practical thermal insulation systems is out of the scope of this work. Our aim is to provide its bounds in order to reach profitability in a PHPS application. An additional possibility to reduce heat losses in the HTES is to use eutectic alloys, instead of pure elements, with enhanced latent heats at lower melting temperatures, such as Al-12Si (549 kWh/m3 at 577 ºC) [51] or Fe-26.3Si-9.3B (~ 1240 kWh/m3 at ~ 1200 ºC) [42]. Time-dependent profiles of heating (space and hot water), cooling, and electricity consumptions (𝐶ℎℎ and 𝐶𝑐, 𝐶𝑒) are obtained by the Energy Plus software [52] for a detached household with two floors with an area of 60 m2 each, 30% of openings (glazing), the U-value of the façade is 0.26 W/(m2·K) and the U-value of the roof is 0.18 W/(m2·K). These U-values match with the guidelines provided in the Spanish Building Code [53], which in turn are also in agreement with the corresponding guidelines of most of the European countries [54], [55]. For the energy simulations, typical occupancy profiles of a 4-inhabitant house (2 adults and 2 children) are considered. The simulations differentiate between working and non-working days, and provide loads, schedules for lighting and home appliances, as well as for occupancy, and the air renovations in the different months. The primary and secondary set-points for air conditioning in summer (i.e. from June to September) are 25 ºC and 27 °C, respectively, while primary and secondary temperature set-points for space heating (i.e. from January to October) are 20 °C and 17 °C. DHW inlet/delivery temperature is 15/60 ºC and the normalized lighting power density in W/m2 -100 lux is 5. The Energy Plus software calculates the energy consumption (for heating, cooling and DHW) of a given household for a user-defined HVAC and DHW systems, thus resultant energy consumption will depend on the conversion efficiency of the selected systems. In the particular case of the electricity, electricity demand will be equal to electricity consumption if no losses are considered. Therefore, in order to obtain energy demand profiles which will be later inputs in our system model, conversion efficiency of the HVAC and DHW systems is set to 1. Energy simulations of the reference building consider 1-hour step weather data from Energy PlusTM database, which includes: direct and diffuse solar irradiance, solar height and azimuth, atmospheric pressure, ambient temperature and wind direction and velocity. Finally, from the energy simulations, hourly global energy demand of the reference building over a year is calculated. Thus, 1-hour step data of electricity for lighting and other appliances, cooling, domestic hot water (DHW) and space heating demands is obtained. The hourly PV electrical power generation per kW of installed PV capacity (𝐺𝑒 ∗, in equivalent hours) in Madrid is calculated by means of the PVsyst software [56]; and considering the same weather data from Energy PlusTM database for coherence with the time-dependent energy consumption profiles. For this calculation monocrystalline silicon PV modules of 318 Wp from the manufacturer SunPower and a Sunny Boy series inverter of the manufacturer SMA have been considered. PV 8 modules are orientated to the South and their tilt angle is 34º, which is the optimum value that maximises the annual PV generation in Madrid. For the comprehensive system model, energy output after the inverter is considered. The total PV electricity generated during one hour (𝐺𝑒, in ∗𝑃𝑛𝑜𝑚,𝑃𝑉, being 𝑃𝑛𝑜𝑚,𝑃𝑉 the nominal PV kWh) is estimated from these simulations as 𝐺𝑒 = 𝐺𝑒 installed power (in kW). Figure 3-a shows an example of the resultant electrical (top) and thermal (bottom) energy rates along with the amount of energy stored in the HTES (top, solid line) and LTES (bottom, solid line) for a system with 𝜂𝑃𝐺𝑈 = 30 %, 𝑃𝑛𝑜𝑚,𝑃𝑉 = 10.6 kWel, 𝑃𝑚𝑎𝑥,𝑃𝐺𝑈 = 1.35 kWel, and 𝐸𝑚𝑎𝑥,𝐻𝑇𝐸𝑆 = 33.7 kWhth. As it will be seen in the discussion section, the sizing of the three key elements of this system, i.e. the PV, HTES and PGU, has been optimized to minimize the cost of the consumed electricity all through the lifetime of the installation. The proper selection of the merit function that must be minimized is of extreme relevance, as different merit functions provide completely different optimum solutions. For instance, a key performance indicator of the system could be its levelized cost of total (consumed) energy or LCOE, which is defined in this work as: 𝑇 𝐶𝐴𝑃𝐸𝑋 + ∑ 𝑡′=1 [ 𝑂𝑃𝐸𝑋(𝑡′) (1 + 𝑊𝐴𝐶𝐶𝑛𝑜𝑚)𝑡′] 𝐿𝐶𝑂𝐸 = 𝑇 ∑ 𝑡′=1 ∑ [ (𝐶𝑒(𝑡) + 𝐶ℎ(𝑡))𝛥𝑡 8760 𝑡=1 (1 + 𝑊𝐴𝐶𝐶𝑟𝑒𝑎𝑙)𝑡′ ] (1) where 𝐶𝐴𝑃𝐸𝑋 is the total capital expenditure (in €), 𝑡′ is the time variable (in years), 𝑇 is the total lifetime of the installation (in years), 𝑂𝑃𝐸𝑋(𝑡′) are the total operational expenditures in year 𝑡′ (in €), and ∑ 8760 𝑡=1 𝐶𝑒(𝑡) + 𝐶ℎ(𝑡)𝛥𝑡 is the total energy consumption (in kWh) during a year, evaluated at time intervals of 𝛥𝑡 = 1 hour. 𝑊𝐴𝐶𝐶𝑛𝑜𝑚 and 𝑊𝐴𝐶𝐶𝑟𝑒𝑎𝑙 are the nominal and real weighted average cost of capital, respectively, and are mutually related through 𝑊𝐴𝐶𝐶𝑟𝑒𝑎𝑙 = (1 + 𝑊𝐴𝐶𝐶𝑛𝑜𝑚) (1 + 𝐼𝑛𝑓𝑙) ⁄ − 1, being 𝐼𝑛𝑓𝑙 the annual inflation rate. The 𝐶𝐴𝑃𝐸𝑋of the solution is the addition of the individual 𝐶𝐴𝑃𝐸𝑋of each element (PV, HTES, LTES, PGU, Boiler, and THP or EHP), each of which are assumed proportional to the element's nominal capacity. For instance, the 𝐶𝐴𝑃𝐸𝑋 of the PV system is estimated as 𝐶𝐴𝑃𝐸𝑋𝑃𝑉 = 𝐶𝐴𝑃𝐸𝑋𝑃𝑉 ∗ 𝑃𝑛𝑜𝑚,𝑃𝑉, being 𝐶𝐴𝑃𝐸𝑋𝑃𝑉 ∗ the capital expenditure per kW of a PV installation. A key difference of equation (1) with respect to conventional definitions of LCOE is that in the denominator we put the total energy consumption, rather than the electricity generated by the PV system. Thus, the LCOE defined in equation (1) refers to the cost of the total amount of energy that is consumed (heat plus electricity), including the cost of the electricity and the fuel that are purchased from the grid. The 𝑂𝑃𝐸𝑋(𝑡′) is calculated as 9 𝑂𝑃𝐸𝑋(𝑡′) = (1 + 𝐼𝑛𝑓𝑙𝑒𝑙)𝑡′ {𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑓𝑖𝑥 ∗ max[𝑃𝑔𝑟𝑖𝑑(𝑡)] + 𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑣𝑎𝑟 ∗ ∑ 8760 𝑡=1 𝑃𝑔𝑟𝑖𝑑(𝑡)𝛥𝑡 } + (1 + 𝐼𝑛𝑓𝑙𝑓𝑢𝑒𝑙) 𝑡′ {𝑂𝑃𝐸𝑋𝑓𝑢𝑒𝑙,𝑓𝑖𝑥 + (2) 𝑂𝑃𝐸𝑋𝑓𝑢𝑒𝑙,𝑣𝑎𝑟 ∗ ∑ 8760 𝑡=1 𝑄𝑒𝑥𝑡(𝑡)𝛥𝑡 } where 𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑓𝑖𝑥 ∗ is the fixed annual cost per installed electric-grid power capacity in current money, and max[𝑃𝑔𝑟𝑖𝑑(𝑡)] is the maximum peak-power demanded to the electrical grid, assumed here to be equal to the maximum grid power capacity. This is an important assumption, as the incorporation of storage in the system enables a significant reduction of max[𝑃𝑔𝑟𝑖𝑑(𝑡)], subsequently providing a noticeable reduction in the fixed costs of external electric power supply. 𝑂𝑃𝐸𝑋𝑓𝑢𝑒𝑙,𝑓𝑖𝑥 is the annual fixed cost of fuel, while 𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑣𝑎𝑟 ∗ and 𝑂𝑃𝐸𝑋𝑓𝑢𝑒𝑙,𝑣𝑎𝑟 ∗ are the annual costs of externally-purchased electricity and heat (in €/kWhel and €/kWhth) expressed in current money. Equation (2) implicitly assumes that electricity and fuel prices increase with constant annual energy inflation rates 𝐼𝑛𝑓𝑙𝑒𝑙 and 𝐼𝑛𝑓𝑙𝑓𝑢𝑒𝑙 , respectively (which are not necessarily equal to the overall economy inflation rate 𝐼𝑛𝑓𝑙), and that all years are identical in terms of energy generation and consumption. Other kinds of operational expenditures, such as taxes and the maintenance costs, are neglected. Preliminary attempts to minimize the total LCOE, as defined in equation (1), resulted in "optimal" solutions tending to maximize the amount of PV electricity dedicated to produce heat. This subsequently resulted in large PV, HTES and LTES systems. Despite the fact that this brings some economical savings after the entire lifetime of the installation, the very high CAPEX results in an intolerable increase of the discounted payback period. As a consequence, we opted for using the levelized cost of consumed electricity (LCOEel) as merit function, which is defined in this work as: 𝑇 𝐶𝐴𝑃𝐸𝑋𝑡 + ∑ 𝑡′=1 [ 𝑂𝑃𝐸𝑋𝑒𝑙(𝑡′) (1 + 𝑊𝐴𝐶𝐶𝑛𝑜𝑚)𝑡′] (3) 𝐿𝐶𝑂𝐸𝑒𝑙 = 𝑇 ∑ 𝑡′=1 8760 𝑡=1 𝐶𝑒(𝑡)𝛥𝑡 ∑ [ (1 + 𝑊𝐴𝐶𝐶𝑟𝑒𝑎𝑙)𝑡′] and differentiates from equation (1) in that it only considers electricity consumption, i.e.: 𝑂𝑃𝐸𝑋𝑒𝑙(𝑡′) = (1 + 𝐼𝑛𝑓𝑙𝑒𝑙)𝑡′ {𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑓𝑖𝑥 ∗ max[𝑃𝑔𝑟𝑖𝑑(𝑡)] + 𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑣𝑎𝑟 ∗ 8760 ∑ 𝑃𝑔𝑟𝑖𝑑(𝑡)𝛥𝑡 𝑡=1 } (4) 10 The minimization of 𝐿𝐶𝑂𝐸𝑒𝑙 will guide us towards the best system configuration in terms of maximum savings of grid-electricity, which is the most relevant contributor to the total OPEX. Minimizing 𝐿𝐶𝑂𝐸𝑒𝑙 appears to be a more reasonable approach that produces shorter payback periods at the expense of obtaining slightly higher LCOE. The search for the minimum 𝐿𝐶𝑂𝐸𝑒𝑙 is performed in this work by means of the multi-variable direct search (Nelder-Mead) algorithm [57] evaluated over a matrix of different initial conditions to avoid local minimums. Finally, the discounted payback period is calculated in this study by computing the annual discounted saves (in current money) in the OPEX due to the reduced consumption of external electricity and fuel, with respect to the reference case in Figure 1-a. The discounted payback period represents the time needed for these cumulative savings to pay-off the higher CAPEX of the PV+PHPS installation. All the variables used in this study to describe the PHPS+PV solution are summarized in Table 1. A selected number of them have been used to define the four different economic scenarios that are summarized in Table 2. These scenarios are ordered from more favourable (Scenario 1) to less favourable (Scenario 4). The most favourable scenario assumes a PV CAPEX of 900 €/kW, while the rest of scenarios assume a price of 1200 €/kW (taxes included). These data are selected based on the estimations of the European JCR, according to which the worldwide average price of a residential PV systems without tax was 1150 €/kW[58] in 2018, being the minimum prices found in Australia (950 €/kW).Taking into account that the average learning curve of CAPEX for residential PV installations is in the range of 80-90% [59], it is expectable that CAPEX values below 900 € (taxes included) could be reached in the near future. Concerning the CO2-equivalent emissions reported in Table 1 it must be noticed that the value for solar PV (20 gCO2eq/kWhth) is taken from the harmonization of a number of published life- cycle assessments conducted by Louwen et al. that assumes a performance ratio of 0.75 and insolation conditions of 1700 kWh/m2-year [60], which are similar to those existing in Madrid. In the case of natural gas, the CO2-equivalent emissions are typically reported in the range of 220 -- 280 gCO2eq/kWhth [61]; thus, we set a value of 250 gCO2eq/kWhth. However, it is worth mentioning that a controversy exists on the determination of the emissions for natural gas in CO2-equivalent units. Natural gas is largely composed of methane, which has a lower atmospheric lifetime (~ 12 years) than CO2 (> 100 years), but a much higher greenhouse potential. Thus, the amount of emissions of natural gas in CO2-equivalent units depends on the time-frame considered for the calculation. Howarth [62] estimated emissions in the range of 550 -- 750 gCO2eq/kWhth for natural gas when considering a shorter timeframe of 20 years. These considerations are not taken into account in this study to keep the coherence among the data 11 obtained from different sources. Finally, the emissions per kWhel of consumed grid electricity is set to 340 gCO2eq/kWhel, which is the one obtained for Spain in 2013 from a Well-To-Wheels (WTW) analysis that takes into account not only the direct emissions in the generation site, but also the upstream emissions associated with the fuel extraction and transport, and the power losses along the grid [63].It is worth mentioning that the average emissions for EU grid electricity is 447 gCO2eq/kWhel, meaning that a PHPS system installed in Spain has a comparatively lower potential to reduce the CO2 emissions than in other countries such as Germany (615 gCO2eq/kWhel) or Italy (431 gCO2eq/kWhel). The investigation of the impact of this solution on different emplacements is not the aim of this study. It is also worth commenting on the assumption of an identical CAPEX for both EHP and THP. Some studies have reported air conditioning prices ranging from 500 to 700 €/kWcool [64]. These values are in agreement with the empiric relation between the capital cost and the heating capacity (kWheat) for EHPs reported by Staffell in 2012 as £2012/kWheat = 200 + 1.25 4750 kWheat ⁄ [7]. This equation results in CAPEX of 395 and 590 €2012/kWheat for systems with large (20 kWheat) and medium (10 kWheat) heating capacities, respectively. Unfortunately, the price of THPs is not as well reported in the literature. In 2002, Grossman [65] estimated the cost of LiBr-water absorption chiller technology in the rage of 165 -- 200 $/kWcool. However, in 2008, Kim and Infante Ferreira [66] reported significantly higher prices (300 - 400 €/kWcool) for the same kind of technology. In both studies, it is estimated that the higher COP of double- and triple-stage chillers would result in lower prices per unit of cooling capacity, despite the higher complexity. Although a lower CAPEX has been reported for THPs, in this study we set an identical price for both EHP and THP of 500 €/kWcool in order to cope with the uncertainty in the published data. A constant COP is assumed for the EHP all through the year. A value of 4 has been used in most of the cases, as this is the average seasonal energy efficiency ratio (SEER) for residential air conditioners installed worldwide, as reported by the IEA in 2019. However, the best SEER available in the market are up to ~ 12. According to the IEA, it is unlikely that these efficient devices reach the market in many countries, but it illustrates the huge potential for improvement in the near future. For the THP, current absorption technology's COPs range from 0.4 -- 1.7 depending on the operation temperature and number of stages. Low temperature (75 -- 90 ºC) single-stage chillers have COPs in the range of 0.4 to 0.7; double-stage chillers operating at intermediate temperatures of ~ 150 ºC have COPs in between 1.2 and 1.4; and high temperature (200 -- 250 ºC) triple-stage chillers have COPs of up to ~ 1.7. [35], [66], [67]. The COP for THP will be set to 1.3; thus, representing the case of a double-stage chiller operating at intermediate temperatures of ~ 150 ºC. This temperature is compatible with typical indirect pressurized hot water storage systems. Thus, for the case of PHPS-T systems we will consider the combination 12 of COP = 1.3 and a pressurized hot water store at 150 ºC. As it will be seen in the discussion section, lower COPs do not provide a significant advantage with respect to the PHPS solution comprising an EHP. Concerning the energy prices, two main components have been assumed: energy and network. Typically, energy and taxes are charged per kWh of consumed energy, while the network component is charged annually per contracted kWel. The share of the fixed and variable components in the price of natural gas and electricity varies significantly among countries . The increasing amount of grid-connected distributed renewable power systems is triggering a trend to increase the fixed components of the price [68], which contributes to increase the uncertainty of this variable in the future. In this study we assume a price of electrical energy and power of 0.17 €/kWhel and 50 €/kWel-year, respectively, which results in an average annual price of electricity of 22.41 c€/kWhel for the specific consumption profile of the reference case considered in this study (Figure 1-a). This value is only slightly below the average price of electricity for households' consumers in Spain, as reported by Eurostat, of 24.77 c€/kWhel. The resultant share of fixed component in the grid electricity price is 24%. For the natural gas, we assume a price for heating energy and power of 0.07 €/kWhth and 60 €/year, respectively, which results in an average price of 7.39 c€/kWhth for the specific consumption profile of the reference case in Figure 1-a. This price is slightly below the average value reported for Spain in Eurostat of 8.75 c€/kWhth, and the fixed term represents only 5% of the total price. 3. Results and Discussion This work aims at determining under which circumstances a PHPS system could be profitable for the self-consumption of solar PV electricity in the residential sector. To afford this analysis, we will search for the optimal sizing of the three main elements of the system, i.e. the storage capacity of HTES in kWhth, the peak power output of the PGU in kWel, and the nominal power of the solar-PV installation in kWel, which result in the minimum costs of the consumed electricity after the entire lifetime of the installation. This optimal sizing depends on a number of parameters, such as the cost of grid-electricity and fuel, or the cost and productivity of the PV system, among many others. In this study, we will pay special attention to those parameters related to the PGU and HTES components, as they are not typically used in residential applications and subsequently, there are no reliable data on their cost and performance. Such elements are described in this study by the following five parameters: the PGU conversion efficiency (𝜂𝑃𝐺𝑈), the HTES CAPEX (𝐶𝐴𝑃𝐸𝑋𝑇𝐸𝑆 ∗ in €/kWhth), the HTES thermal losses (𝑎𝐻𝑇𝐸𝑆 and ∆𝑇𝐻𝑇𝐸𝑆), and the PGU CAPEX (𝐶𝐴𝑃𝐸𝑋𝐻2𝑃 ∗ in €/kWel). Performing a bottom-up estimation of these five parameters would be very unprecise, provided the immaturity of these technologies; thus, our first analysis focuses on conducting an up-bottom analysis, according to 13 which the minimum value of 𝜂𝑃𝐺𝑈 , and the maximum values of 𝐶𝐴𝑃𝐸𝑋𝐻𝑇𝐸𝑆 ∗ , 𝐶𝐴𝑃𝐸𝑋𝑃𝐺𝑈 ∗ , 𝑎𝐻𝑇𝐸𝑆 and ∆𝑇𝐻𝑇𝐸𝑆that leads to profitability will be determined assuming a favourable economic scenario. Thus, the outcome of this analysis will be the technological requirements for each element of the solution in order to reach profitability. This information will be useful to assess the candidate technologies. Nevertheless, despite the fact that we will mention a few possible specific options, it is not the aim of this study to provide a thorough assessment of the most suitable technological implementation. To quantify the "profitability" of the solution we will look at the amount of electricity savings enabled by the PHPS system. Electricity savings can only be attributed to the presence of significantly large PV, HTES and PGU systems (optimized variables) that provide a significant reduction of the cost of consumed electricity (merit function for the optimization). The first part of the discussion focuses on the PHPS-E configuration (Figure 1-b), which is characterized by using an electrically driven heat pump for cooling; thus, the rejected heat from the PGU is only used to satisfy the DHW and space heating needs. In the second part of the discussion, we assess the PHPS-T configuration (Figure 1-c), and we discuss under which conditions a PHPS- T system would be preferable than a PHPS-E one. Some of the most relevant results are summarized in Table 3 in order to facilitate the readability of the analysis. Table 4 shows identical results but for the ideal case of loss-less (adiabatic) HTES and LTES systems; thus, they represent the upper bounds of performance for the PHPS solution, which are unattainable in practice. 3.1. PHPS-E system Figure 3 (b-d) shows the savings on grid electricity resulting from an optimized PHPS-E system as a function of different variables. Every dot in these figures represents an optimized system, meaning that the optimal sizing of the three main elements of the system (PV, HTES and PGU) are set to minimize the cost of the electricity consumed during the system lifetime. The rest of parameters are set to the values reported in Table 1, and to those corresponding to the most favourable scenario in Table 2 (Scenario 1). In the case of Figure 3-b, the two independent variables are the PGU efficiency and the HTES and PGU CAPEX. As it could be expected, the largest savings are obtained for high PGU efficiency and small HTES and PGU CAPEX. Besides, the minimum attainable savings of electricity are about 30%, and correspond to the case of direct self-consumption of PV electricity, without storage. The results shown in Figure 3-b enable determining the maximum cost targets for HTES and PGU technologies if the PGU conversion efficiency is known. Similarly, it is possible to estimate the minimum PGU conversion efficiency that is needed to reach profitability if the costs of HTES and PGU devices are provided. For instance, a CAPEX of ~ 300 €/kWhth and ~ 4000 €/kWel for the HTES and 14 PGU devices are tolerable (enable a significant savings on electricity consumption) if the PGU conversion efficiency is higher than ~ 40%. However, if the conversion efficiency is 20%, the maximum tolerable CAPEX is reduced to ~ 100 €/kWhth and ~ 2000 €/kWel, respectively. It is worth noticing that high thermal-to-electric conversion efficiencies (> 40%) and low costs (< 1000 €/kW) are attainable by current state of the art Rankine and open-cycle Brayton engines of large (> 1 MWel) [39], [69], [70] and medium (30 - 300 kWel) [71] sizes. Unfortunately, current state of the art small-scale (< 10 kWel) closed-cycle engines, which have been developed for both terrestrial and space-power applications, have either low conversion efficiency (e.g. Rankine/ORC) or high cost (Brayton) [72], [73]. Probably, among all the current dynamic- engine options, the best choice is a Stirling engine, which has a conversion efficiency in the range of 30 - 40% at power outputs ranging from 1 to 30 kWe [73]. However, low power costs (~ 2,000 €/kWel) are only attainable by the largest units (~ 30 kWel), whilst the smallest units (~ 1-2 kWel) could reach power costs more than 10,000 €/kWel. This price lies outside the profitability limit, as seen in Figure 3-b. Solid-state converters are better suited for power generation at small-scale and could lead, in principle, to low power costs. In this regard, thermoelectric generators are the most mature technology, but they lack of high conversion efficiency (typically below ~ 10%). A particularly interesting highly-efficient alternative is thermophotovoltaics (TPV), which has been already assessed theoretically for PHPS applications [41], [46]. Despite its much lower degree of development, TPV has already demonstrated significantly higher conversion efficiencies (~ 24%) [74], becoming the most efficient solid-state thermal-to-electric converter to date. TPV is particularly well suited for ultra-high temperature (> 1000 ºC) heat conversion; thus, enabling the use of ultra-dense heat stores, such as silicon or boron latent heats [41]. TPV technology has also potential to reach very low power costs, even below 300 €/kWel [46]. Thus, in the opinion of the authors, TPV should be regarded as a promising choice for future developments in residential (small-scale) PHPS solutions. Figure 3-c illustrates how heat losses in the HTES, i.e. the parameters 𝑎𝐻𝑇𝐸𝑆 and ∆𝑇𝐻𝑇𝐸𝑆 , impact on the profitability of the PHPHS-E solution. As it could be expected, small 𝑎𝐻𝑇𝐸𝑆 and ∆𝑇𝐻𝑇𝐸𝑆 values (small heat losses in the HTES) result in larger savings in electricity. For instance, reducing the HTES temperature below ~ 500 ºC allows a significant reduction of the heat losses, or alternatively enables the use of higher 𝑎𝐻𝑇𝐸𝑆 values; thus, simplifying the thermal insulation design, and eventually enabling a reduction of its cost. However, lower HTES temperatures could result in lower PGU conversion efficiencies and negatively impact on the profitability of the solution. For instance, a HTES operating at 1500 ºC with 40% conversion efficiency requires 𝑎𝐻𝑇𝐸𝑆 < 0.2 for reaching electricity savings more than ~ 80%. Reducing the HTES temperature to 250 ºC while keeping a conversion efficiency of 40% could enable similar 15 savings for less a restrictive 𝑎𝐻𝑇𝐸𝑆 < 0.4 . However, if the efficiency drops to 20%, such amounts of savings would not be reachable even using an ideal thermal insulation system (𝑎𝐻𝑇𝐸𝑆 = 0). This illustrates an existing trade-off on the optimal HTES temperature between the PGU conversion efficiency and the HTES's thermal insulation losses. High conversion efficiencies and low thermal losses are needed to reach profitability, but both of them increase with temperature. Nevertheless, a very interesting result illustrated in Figure 3-c is that this trade-off disappears for temperatures more than ~ 500 ºC. For such high temperatures, 𝑎𝐻𝑇𝐸𝑆 values near or below ~ 0.1, and PGU efficiency more than ~ 20% result in a profitable solution, independently of the HTES temperature. This is partially attributed to the higher storage energy density that is attainable at very high temperatures, which enables a more compact and smaller HTES that compensates the larger amounts of heat losses per unit of store area at higher temperatures. According to this result, the HTES temperature could be as high as required to reach high energy density and PGU conversion efficiency, provided that a low-cost thermal insulation system with 𝑎𝐻𝑇𝐸𝑆 values near or below ~ 0.1 are attainable at such temperatures. Therefore, the selection of the optimal operation temperature should be made based on an overall techno-economic analysis of the entire solution, being the main objective the achievement of high PGU efficiency and low thermal insulation losses at low cost. In this regard, the results shown in Figure 3-c could guide the design of an optimal thermal insulation system for PHPS applications. The results in Table 4 show to the ideal case of adiabatic HTES and LTES systems (i.e. 𝑎𝐻𝑇𝐸𝑆 = 𝑎𝐿𝑇𝐸𝑆 = 0). Despite the fact that this ideal situation is unattainable in practice, these results are valuable to set the upper bounds of the technology. If compared with the results in Table 3, (which assume 𝑎𝐻𝑇𝐸𝑆 = 𝑎𝐿𝑇𝐸𝑆 = 0.1 ) the self-consumption ratio is drastically increased due to the reduced amount of heat losses. This results in a smaller PV system and CAPEX, which leads to significantly shorter payback periods. These results highlight the relevance of thermal insulation in the PHPS solution, at least for small-scale residential applications. Figure 3-d assess the sensibility of the PHPS-E solution to variations in the economic conditions. The results shown in Figure 3-c and d assume the most favourable case, which is characterized by low PV CAPEX (900 €/kWel), low WACCnom (2%), and a high energy price inflation rate (Infle = 4%). This optimistic scenario allows estimating the minimum requirements for the PHPS-E solution (in terms of cost and heat losses of the HTES, and efficiency of the PGU) to reach profitability in the most favourable economic conditions. Figure 3-d illustrates how these requirements are modified for more unfavourable economic scenarios. By calculating the savings on consumed grid electricity resulting from an optimized PHPS-E system as a function of the Infle and the WACCnom we observed that the scenarios having the same 16 difference between the WACCnom and the energy price inflation rate result in identical optimal systems and savings of electricity. This means that the profitability can be analyzed through a single variable: the difference between the WACCnom and Infle. Figure 3-d illustrates the combined effect on the electricity savings of the PV CAPEX and (WACCnom - Infle). As expected, having a low PV CAPEX and an energy inflation rate higher than the WACCnom is very important to reach profitability, especially when the PGU conversion efficiency is low. Higher PGU conversion efficiencies enable reaching profitability under less favourable economic conditions. However, even in the case of high PGU conversion efficiency (e.g. 40%), a low PV CAPEX (below 1000 €/kWel) and an energy price inflation rate greater than WACCnom are needed to reach electricity savings of 80% or beyond. A less favourable scenario, e.g. PV CAPEX of 1300 €/kWel and an energy price inflation rate 1% smaller than the WACCnom, would enable maximum electricity savings of ~ 55% if the PGU conversion efficiency is 40%, and of ~ 30% if the PGU conversion efficiency is 20%. The latter represents a solution without HTES nor PGU devices, in which the PV electricity is directly consumed without being stored. This illustrates how important are the economic boundary conditions on the potential of the PHPS-E concept. All the results shown in Figure 3(b-d) show deviations of a selected number of parameters (i.e. CAPEX of HTES, PGU and PV, PGU efficiency, WACCnom and energy price inflation rate) from the most favourable economic scenario indicated in Table 2. Besides, they illustrate the electricity savings only, as they represent a quantification of the profitability of a PHPS solution. But no observations have been made so far on other relevant parameters such as the discounted payback period, the savings of fuel and CO2 emissions, or the self-consumption ratio, among others. In the next discussion, we will pay attention to these parameters. To that end, Figure 4 shows a selection of eight system's parameters as a function of the PGU conversion efficiency, for an optimized PHPS-E system under the four different economic scenarios indicated in Table 2. The selected parameters for the representation are: the optimal PGU maximum power capacity (a), the optimal HTES storage capacity (b), the optimal PV nominal power (c), the discounted payback period (d), the savings in fuel (e), and CO2 emissions (f), the self-consumption ratio (g), and the savings in electricity (h). In these graphs, each dot represents the best solution obtained by the direct-search algorithm after being evaluated with 78 different initial simplexes to avoid local minimums. The first important observation is that there exists a threshold value for the PGU conversion efficiency beyond which the PHPS-E system becomes profitable, i.e. the optimal sizes for the PGU (Figure 4-a) and the HTES (Figure 4-b) are significant. Besides, this threshold efficiency is higher when the economic conditions are less favourable. For instance, the most favourable scenario (Scenario 1) enables profitability at PGU conversion efficiencies more than ~ 10%, 17 while the most unfavourable one (Scenario 4) requires PGU conversion efficiencies more than ~ 80%. For efficiencies below those thresholds, the optimal sizes for the PGU and HTES devices tend to zero, and the only component that remains in the solution is the PV system. In this case, the PV electricity is consumed instantaneously (without storage in the HTES); thus, resulting in a small optimal PV system that enables very high self-consumption ratios (~ 80%, Figure 4-g). Notice that such high self-consumption ratios are attributed to the use of solar PV electricity to produce low-grade heat that is stored in the LTES. This explains why such high self- consumption ratios are combined with small electricity savings (~ 30%, Figure 4-h). When the PGU efficiency is higher than the profitability threshold, the optimal PHPS-E system comprises significantly large HTES, PGU and PV systems (Figure 4-a, b, c), subsequently enabling electricity savings in the range of 70 - 95% (Figure 4-h), depending on the specific economic boundary conditions. The most favourable scenarios bring particularly large optimal HTES (up to 80 kWhth), PGU (up to 3 kWel) and solar PV (up to 12 kWel) systems, along with higher electricity savings (up to 95%). On the other hand, less favourable economic conditions bring smaller optimal HTES (10 -- 20 kWhth), PGU (~ 0.8 kWel), and solar PV (5 -- 7 kWel) systems, as well as lower electricity savings of 70 -- 75%. A particularly important observation is that increasing the efficiency beyond a certain value does not improve the savings in electricity. Instead, it leads to lower savings of fuel and subsequently higher CO2 emissions. This critical efficiency is ~ 40% for the scenarios 1 and 2, and ~ 60% (~ 80%) for the scenario 3 (4). At these specific efficiencies, the CO2 emissions are minimum, as higher PGU conversion efficiencies result in the production of smaller amounts of heat that bring a larger consumption of fuel. Unfortunately, the same efficiency that minimizes the CO2 emissions maximizes the payback period and minimizes the self-consumption ratio, as it implies the use of significantly large PV, HTES and PGU systems. However, the potential reduction in the payback period due to an increment of the conversion efficiency beyond such values might not be very significant. For instance, in the Scenario 1, the payback period is reduced only 3 years (from 14 to 11 years) when the PGU conversion efficiency increases from 40% to 80%, while the electricity savings are not improved. This illustrates that increasing the PGU conversion efficiency beyond a certain critical value (which depends on the boundary economic conditions) might not be as important as it could be intuitively expected. Other interesting observation concerns the very small self-consumption ratios (35 -- 60%) that are obtained, meaning that a high amount of PV electricity is finally wasted. To understand the source of these losses, Figure 5 shows the four main contributions to the energy lost in the system as a function of the PGU conversion efficiency for each of the four economic scenarios indicated in Table 2. For low conversion efficiencies, i.e. when the solution lacks of both HTES and PGU devices, most of the losses (> 80%) are attributed to heat losses in the LTES. 18 However, these losses are not very significant in absolute value, as they enable very high self- consumption ratios of ~ 80% (Figure 4-g). On the other hand, when the PGU conversion efficiency is large enough, and subsequently the optimal HTES and PGU sizes are significant, the HTES thermal insulation heat losses account for most of the energy losses of the PHPS-E solution (70 -- 90%). Thus, heat losses in the HTES are the main reason for the low self- consumption levels observed in Figure 4-h. The next contribution is the heat lost in the LTES (10 -- 20%). Both the PV electricity that is neither consumed nor stored, and the exhaust heat from the PGU converter that is wasted, represent a negligible contribution to the total energy losses in the system. Only in Scenario 1 the PV electricity that is directly wasted could represent a significant amount of losses (up to 20%) due to the oversized PV installation, which is possible due to its low cost. But even in this case the heat losses in the HTES represent the highest contributor to the overall energy losses of the system. Therefore, improving thermal insulation of the HTES is key to improve the self-consumption ratio. However, it is important to notice that, even with such low self-consumption ratios, the PHPS-E system can provide significant electricity savings (> 70%) and reach profitability with reasonably short payback periods. The heat losses through the HTES thermal insulation system could be recovered as low-grade heat to further reduce the amount of fuel consumption, as proposed in [47], [48]. Analysing this and other possible improvements could be the aim of a future work. 3.2. PHPS-T system This last part of the article focuses on the analysis of the PHPS-T system (Figure 1-c), where a thermally-driven heat pump (THP) is used instead of an electrically-driven one to satisfy the cooling demand. In this case, the heat generated in the PGU is not only used for DHW and space heating, but it is also used for powering the THP and satisfying the cooling needs in summer season. Thus, one could expect a more efficient use of the generated heat all through the year. However, one could also argue that an increment in the COP of the EHP could have a similar effect on reducing the electricity consumption both in the reference case (Figure 1-a) and in the PHPS-E solution (Figure 1-b); thus, hindering the profitability of the PHPS-T solution. Thus, the payback period of the PHPS-T system should be analysed as a function of the difference between the COPs of the THP and the EHP. In this regard, Figure 6 shows the payback period of an optimized PHPS-T (top), and the difference between the payback periods of PHPS-E and PHPS-T solutions (bottom) as a function of the THP and EHP COPs for four different PGU conversion efficiencies. Payback periods are calculated with respect to the reference case (Figure 1-a). As expected, increasing the EHP COP produces a significant increment on the payback period of the PHPS-T solution, hindering its profitability (Figure 6-top). This increment could result in intolerable payback 19 periods (> 20 years) if the THP COP is low (< 0.7) and the EHP COP is reasonably high (> 4). In this case, the PHPS-E solution would be the preferable solution, with payback periods significantly shorter than those of PHPS-T. Very high THP COPs (more than ~ 1.3) would be needed for PHPS-T to reach reasonably low payback periods (< 12 years) when the EHP COP is considerably high (~ 6 or beyond). In this case, the PHPS-T solution could provide significantly shorter payback periods than PHPS-E. It is worth noticing that the improvement of the EHP COP is beneficial not only for the PHPS-E system, but also for the reference case. Thus, improving the EHP COP does not bring a significant reduction of the payback period for the PHPS-E system with respect to the reference case. For this reason, the only way to drastically reduce the payback period is to adopt the PHPS-T solution with a very highly efficient THP. High THP COPs of 1.2 -- 1.7 can be attained with current state of the art double-stage or triple- stage absorption chillers operating at temperatures in the range of 150 - 250 ºC [35], [66], [67]. This implies that a PHPS-T solution should rely on the use of PGU with a high-grade rejected heat (> 150 ºC) able to power an efficient double- or triple-stage absorption chiller. Otherwise, the PHPS-E system would be a preferable solution. This will impact on the selection of the PGU. High rejection temperature dynamic engines or solid-state devices are preferable. In this regard, the low rejection temperature of TPV devices is detrimental. High TPV cell temperatures lead to a significant reduction of the conversion efficiency. Thus, the requirement of higher temperature rejected heat points in the direction of an interesting research line for TPV, which is the development of highly efficient TPV cells able to operate at temperatures of ~ 200ºC [72]. Other high-rejection temperature solid-state alternatives, such as thermionic generators [72], could also be regarded as an interesting option for the future. Figure 7 shows the optimization results for a PHPS-T system with COPEHP = 4 and COPTHP = 1.3; thus, representing a case where the PHPS-T system provides shorter payback periods than PHPS-E, as shown in Figure 6 (bottom). The results are presented in a similar way than it was done in Figure 4 for the PHPS-E system, evaluating the four scenarios in Table 2. Despite the fact that the general tendencies are similar, there are some remarkable differences. First, the optimal PHPS-T system requires smaller components, ultimately resulting in significantly shorter payback periods (8 -- 14 years). This is mostly attributed to the reduction in the electricity consumption when removing the EHP, which results in a significantly lower amount of electricity that needs to be generated and stored by the PV+PHPS system; thus, leading to smaller HTES and PGU devices. It is worth mentioning that the smaller HTES and PGU devices also result in a slightly lower potential of the PHPS-T solution to reduce the grid- electricity consumption. This illustrates an existing trade-off between obtaining either small payback periods or high energetic savings. 20 The other main difference observed when comparing PHPS-T (Figure 7) with PHPS-E (Figure 4) is the increment in the consumption of fuel (up to ~ 20%). The THP consumes large amounts of heat, especially in summer, which cannot be fully provided by the rejected heat from the PGU. This implies that a larger amount of external fuel is consumed. This contributes to the increase of the CO2 equivalent emissions, which are barely compensated by the savings in emissions due to the self-consumption of solar-PV electricity. As a result, the PHPS-T solution produce significantly lower savings in CO2 emissions (up to ~ 0.75-ton eqCO2/year, from a total emission of 5.2 ton/year). If the COPTHP would be lower, a PHPS-T solution could even produce an increment of the emissions with respect to the reference case. It must be noticed that these results are obtained for a system that aims at minimizing the lifetime cost of electricity. Choosing other kinds of merit functions (e.g. the total LCOE in eq. 1) would lead to larger HTES and PGU systems that bring higher savings in CO2 emissions at the expenses of increasing the payback period. Another obvious strategy to reduce the CO2 emissions could be hybridizing a PHPS-T solution with a kind of renewable heat source, such as solar thermal collectors. The resultant hybrid PHPS-T solution would enable a significant reduction in the CO2 emissions of up to 4.5 ton/year (from the total reference emissions of 5.2 ton/year), as well as a low payback period of 12 years or below. 4 Conclusions In this article we have evaluated the implementation of a power-to-heat-to-power storage solution for the self-consumption of solar PV electricity in a dwelling in Madrid. Our results indicate that the solution has potential to provide a significant amount of savings in the consumption of grid electricity (> 70%) with reasonably short discounted payback periods (< 15 years), if compared to a solution that uses fuel for heating and grid-electricity. This holds true even when the heat-to-power conversion efficiencies is moderately low (20 - 30%), provided that the economic conditions are favourable. If the cost of the technology is sufficiently low, there exists a certain thermal-to-electric conversion efficiency (~ 40% in the most favourable case analysed in this study) beyond which the power generation system can be made slightly smaller, but the amount of savings in electricity does not increase, i.e. increasing the efficiency might not be as important as it could be intuitively expected. Besides, the inefficient power conversion process results in a large amount of rejected heat that can be used to satisfy other kinds of energetic demands. For instance, if an electrically driven heat pump is used for cooling, the rejected heat could be used for providing space heating and domestic hot water, bringing additional savings in the fuel consumption in the range of 10 -- 20%, and global savings in CO2 equivalent emissions in the range of 1.2 -- 1.6 ton/year (24 -- 31%). If a thermally driven heat 21 pump is used instead, the generated heat could be also used for satisfying the cooling demands. This leads to shorter payback periods if the COP of the thermally driven heat pump is significantly high (> 1.3). This condition is attainable by current state of the art double- or triple-stage absorption chillers operating at temperatures in the range of 150 -- 250 ºC, meaning that a high rejection temperature would be necessary in the conversion of the stored heat into electricity. However, the very large cooling needs during summer in Madrid may not be fully satisfied by the rejected heat in the system, in which case an extra consumption of fuel is needed, subsequently leading to a small reduction (or even an increment) in the greenhouse gas emissions. A possible solution consists of replacing the use of fossil fuels by a renewable heat source, such as solar thermal. In this case, the entire power-to-heat-to-power solution comprising a highly efficient thermally driven heat pump and solar thermal collectors could provide a drastic reduction of the emissions (~ 4.5 ton/year, or 86%) while keeping reasonably short payback periods (< 12 years). Notice that a clearer economic advantage would be obtained if we compared the proposed PHPS solution with a reference case that uses an electric-powered boiler, instead of a fuel-powered one. This most favourable scenario is not analysed in this study, and should be assessed in future works. The main drawback of the proposed solution, independently of using a thermally- or an electrically-driven heat pump, concerns the small self-consumption ratios of PV electricity (40 -- 60%), which are mainly attributed to the large amount of heat losses in the high temperature thermal store. Possible ways of minimizing these losses include the development of novel ultra- dense heat stores at moderately low temperatures, or more advanced thermal insulation systems. Another disadvantage concerns the low readiness level of heat-to-power generation technologies that are highly efficient at small scales (~ 1 kW). Some solid-state power generators, such as thermophotovoltaics, show potential to reach the required conversion efficiency and low cost. However, they are not readily available in the market. Next works should assess the use of PHPS systems in larger residential buildings, where the use of market-available dynamic engines, such as Stirling generators, could be profitable. Future analyses should also assess the possibility of selling the excess of PV electricity to the grid, which has been disregarded in the current work. This could negatively impact on the profitability of the solution as it is considered in this study. Finally, regardless of the particular system implementation, the system model and methodology presented in this article becomes a quite powerful tool which will allow to answer fundamental questions regarding the profitability of different energy solutions (including novel energy storage technologies) by evaluating payback periods and energy saving as well as optimal sizing that results in minimum costs of energy, among many other parameters. 22 5 Acknowledgements This work has been partially funded by the projects MADRID-PV2-CM (P2018/EMT-4308), funded by the government of the Comunidad de Madrid with the support from FEDER Funds, TORMES (ENE2015-72843-EXP), and AMADEUS (737054), funded by the European Commission through the Horizon 2020 programme, FET-OPEN action. The sole responsibility for the content of this publication lies with the authors. It does not necessarily reflect the opinion of the European Union. Neither the REA nor the European Commission are responsible for any use that may be made of the information contained therein. A. Datas acknowledges postdoctoral fellowship support from the Spanish "Juan de la Cierva-Incorporación" program (IJCI-2015-23747). A. Ramos acknowledges the Universitat Politècnica de Catalunya for her Serra Hunter Tenure Track professor post. The authors acknowledge the fruitful discussions to Dra. Marta Victoria Pérez, especially concerning the methodology to assess the economic aspects of the system, and to Dr. Stephan Lang for his advices concerning the thermal losses analysis in the heat stores. References [1] P. Nejat, F. Jomehzadeh, M. M. Taheri, M. Gohari, and M. Z. Abd. Majid, "A global review of energy consumption, CO2 emissions and policy in the residential sector (with an overview of the top ten CO2 emitting countries)," Renewable and Sustainable Energy Reviews, vol. 43, pp. 843 -- 862, Mar. 2015. [2] D. W. Wu and R. Z. Wang, "Combined cooling, heating and power: A review," Progress in Energy and Combustion Science, vol. 32, no. 5, pp. 459 -- 495, Sep. 2006. [3] H. I. Onovwiona and V. I. Ugursal, "Residential cogeneration systems: review of the current technology," Renewable and Sustainable Energy Reviews, vol. 10, no. 5, pp. 389 -- 431, Oct. 2006. [4] H. Cho, A. D. Smith, and P. Mago, "Combined cooling, heating and power: A review of performance improvement and optimization," Applied Energy, vol. 136, pp. 168 -- 185, Dec. 2014. [5] D. Sonar, S. L. Soni, and D. Sharma, "Micro-trigeneration for energy sustainability: Technologies, tools and trends," Applied Thermal Engineering, vol. 71, no. 2, pp. 790 -- 796, Oct. 2014. [6] L. Trygg and S. Amiri, "European perspective on absorption cooling in a combined heat and power system -- A case study of energy utility and industries in Sweden," Applied Energy, vol. 84, no. 12, pp. 1319 -- 1337, Dec. 2007. I. Staffell, D. Brett, N. Brandon, and A. Hawkes, "A review of domestic heat pumps," Energy Environ. Sci., vol. 5, no. 11, pp. 9291 -- 9306, Oct. 2012. [7] [8] E. Bellos and C. Tzivanidis, "Multi-objective optimization of a solar driven trigeneration system," Energy, vol. 149, pp. 47 -- 62, Apr. 2018. [9] A. Ramos, M. A. Chatzopoulou, J. Freeman, and C. N. Markides, "Optimisation of a high- efficiency solar-driven organic Rankine cycle for applications in the built environment," Applied Energy, vol. 228, pp. 755 -- 765, Oct. 2018. 23 [10] A. Kasaeian, G. Nouri, P. Ranjbaran, and D. Wen, "Solar collectors and photovoltaics as combined heat and power systems: A critical review," Energy Conversion and Management, vol. 156, pp. 688 -- 705, Jan. 2018. [11] K. E. N'Tsoukpoe et al., "Integrated design and construction of a micro-central tower power plant," Energy for Sustainable Development, vol. 31, pp. 1 -- 13, Apr. 2016. [12] Y. M. Seshie, K. E. N'Tsoukpoe, P. Neveu, Y. Coulibaly, and Y. K. Azoumah, "Small scale concentrating solar plants for rural electrification," Renewable and Sustainable Energy Reviews, vol. 90, pp. 195 -- 209, Jul. 2018. [13] R. Luthander, J. Widén, D. Nilsson, and J. Palm, "Photovoltaic self-consumption in buildings: A review," Applied Energy, vol. 142, pp. 80 -- 94, Mar. 2015. [14] "REN21. 2018. Renewables 2018 Global Status Report," ISBN 978-3-9818107-0-7, 2018. [15] "IEA SNAPSHOT OF GLOBAL PHOTOVOLTAIC MARKETS," IEA - International Energy Agency, IEA PVPS T1-33:2018, 2018. [16] REN21, "RENEWABLES 2018 GLOBAL STATUS REPORT." [17] A. H. Nosrat, L. G. Swan, and J. M. Pearce, "Improved performance of hybrid photovoltaic-trigeneration systems over photovoltaic-cogen systems including effects of battery storage," Energy, vol. 49, pp. 366 -- 374, Jan. 2013. [18] A. S. Mundada, K. K. Shah, and J. M. Pearce, "Levelized cost of electricity for solar photovoltaic, battery and cogen hybrid systems," Renewable and Sustainable Energy Reviews, vol. 57, pp. 692 -- 703, May 2016. [19] S. Acha, A. Mariaud, N. Shah, and C. N. Markides, "Optimal design and operation of distributed low-carbon energy technologies in commercial buildings," Energy, vol. 142, pp. 578 -- 591, Jan. 2018. [20] Y.-Y. Jing, H. Bai, J.-J. Wang, and L. Liu, "Life cycle assessment of a solar combined cooling heating and power system in different operation strategies," Applied Energy, vol. 92, pp. 843 -- 853, Apr. 2012. [21] A. H. A. Al-Waeli, K. Sopian, H. A. Kazem, and M. T. Chaichan, "Photovoltaic/Thermal (PV/T) systems: Status and future prospects," Renewable and Sustainable Energy Reviews, vol. 77, pp. 109 -- 130, Sep. 2017. [22] A. Ramos, M. A. Chatzopoulou, I. Guarracino, J. Freeman, and C. N. Markides, "Hybrid photovoltaic-thermal solar systems for combined heating, cooling and power provision in the urban environment," Energy Conversion and Management, vol. 150, pp. 838 -- 850, Oct. 2017. [23] M. Herrando, A. Ramos, I. Zabalza, and christos N. Markides, "Energy Performance of a Solar Trigeneration System Based on a Novel Hybrid PVT Panel for Residential Applications," in Proc. of the Solar World Conference, 2017. [24] M. Herrando and C. N. Markides, "Hybrid PV and solar-thermal systems for domestic heat and power provision in the UK: Techno-economic considerations," Applied Energy, vol. 161, pp. 512 -- 532, Jan. 2016. [25] A. Selviaridis, B. R. Burg, A. S. Wallerand, F. Maréchal, and B. Michel, "Thermo- economic analysis of a trigeneration HCPVT power plant," AIP Conference Proceedings, vol. 1679, no. 1, p. 100004, Sep. 2015. [26] C. J. C. Williams, J. O. Binder, and T. Kelm, "Demand side management through heat pumps, thermal storage and battery storage to increase local self-consumption and grid compatibility of PV systems," in 2012 3rd IEEE PES Innovative Smart Grid Technologies Europe (ISGT Europe), 2012, pp. 1 -- 6. [27] J. Tong, Z. Quan, Y. Zhao, G. Wang, J. Cai, and Y. Chi, "The Study on the Performance of Solar Photovoltaic and Air Source Heat Pump Composite Building Energy Supply System," Procedia Engineering, vol. 205, pp. 4082 -- 4089, Jan. 2017. [28] C. Roselli, M. Sasso, and F. Tariello, "Integration between electric heat pump and PV system to increase self-consumption of an office application," Renew. Energy Environ. Sustain., vol. 2, p. 28, 2017. [29] A. Bloess, W.-P. Schill, and A. Zerrahn, "Power-to-heat for renewable energy integration: A review of technologies, modeling approaches, and flexibility potentials," Applied Energy, vol. 212, pp. 1611 -- 1626, Feb. 2018. 24 [30] J. Li and M. A. Danzer, "Optimal charge control strategies for stationary photovoltaic battery systems," Journal of Power Sources, vol. 258, pp. 365 -- 373, Jul. 2014. [31] R. Thygesen and B. Karlsson, "Simulation and analysis of a solar assisted heat pump system with two different storage types for high levels of PV electricity self- consumption," Solar Energy, vol. 103, pp. 19 -- 27, May 2014. [32] D. Parra, G. S. Walker, and M. Gillott, "Are batteries the optimum PV-coupled energy storage for dwellings? Techno-economic comparison with hot water tanks in the UK," Energy and Buildings, vol. 116, pp. 614 -- 621, Mar. 2016. [33] A. Mariaud, S. Acha, N. Ekins-Daukes, N. Shah, and C. N. Markides, "Integrated optimisation of photovoltaic and battery storage systems for UK commercial buildings," Applied Energy, vol. 199, pp. 466 -- 478, Aug. 2017. [34] O. Schmidt, A. Hawkes, A. Gambhir, and I. Staffell, "The future cost of electrical energy storage based on experience rates," Nature Energy, vol. 2, no. 8, p. 17110, Aug. 2017. [35] R. Z. Wang, X. Yu, T. S. Ge, and T. X. Li, "The present and future of residential refrigeration, power generation and energy storage," Applied Thermal Engineering, vol. 53, no. 2, pp. 256 -- 270, May 2013. [36] M. Zheng, C. J. Meinrenken, and K. S. Lackner, "Smart households: Dispatch strategies and economic analysis of distributed energy storage for residential peak shaving," Applied Energy, vol. 147, pp. 246 -- 257, Jun. 2015. [37] Y. Li, X. Wang, D. Li, and Y. Ding, "A trigeneration system based on compressed air and thermal energy storage," Applied Energy, vol. 99, pp. 316 -- 323, Nov. 2012. [38] D. Teichmann, K. Stark, K. Müller, G. Zöttl, P. Wasserscheid, and W. Arlt, "Energy storage in residential and commercial buildings via Liquid Organic Hydrogen Carriers (LOHC)," Energy Environ. Sci., vol. 5, no. 10, pp. 9044 -- 9054, Sep. 2012. [39] L. A. Weinstein, J. Loomis, B. Bhatia, D. M. Bierman, E. N. Wang, and G. Chen, "Concentrating Solar Power," Chem. Rev., vol. 115, no. 23, pp. 12797 -- 12838, Dec. 2015. [40] M. A. Sabiha, R. Saidur, S. Mekhilef, and O. Mahian, "Progress and latest developments of evacuated tube solar collectors," Renewable and Sustainable Energy Reviews, vol. 51, pp. 1038 -- 1054, Nov. 2015. [41] A. Datas, A. Ramos, A. Martí, C. del Cañizo, and A. Luque, "Ultra high temperature latent heat energy storage and thermophotovoltaic energy conversion," Energy, vol. 107, pp. 542 -- 549, Jul. 2016. [42] A. Datas et al., "AMADEUS: Next generation materials and solid state devices for ultra high temperature energy storage and conversion," in AIP Conference Proceedings, 2018, vol. 2033, p. 170004. [43] T. Desrues, J. Ruer, P. Marty, and J. F. Fourmigué, "A thermal energy storage process for large scale electric applications," Applied Thermal Engineering, vol. 30, no. 5, pp. 425 -- 432, Apr. 2010. [44] A. Thess, "Thermodynamic Efficiency of Pumped Heat Electricity Storage," Phys. Rev. Lett., vol. 111, no. 11, p. 110602, Sep. 2013. [45] R. B. Laughlin, "Pumped thermal grid storage with heat exchange," Journal of Renewable and Sustainable Energy, vol. 9, no. 4, p. 044103, Jul. 2017. [46] C. Amy, H. R. Seyf, M. A. Steiner, D. J. Friedman, and A. Henry, "Thermal energy grid storage using multi-junction photovoltaics," Energy Environ. Sci., Nov. 2018. [47] A. Robinson, "Ultra-high temperature thermal energy storage. part 1: concepts," Journal of Energy Storage, vol. 13, pp. 277 -- 286, Oct. 2017. [48] A. Robinson, "Ultra-high temperature thermal energy storage. Part 2: Engineering and operation," Journal of Energy Storage, vol. 18, pp. 333 -- 339, Aug. 2018. [49] R. Sioshansi, P. Denholm, T. Jenkin, and J. Weiss, "Estimating the value of electricity storage in PJM: Arbitrage and some welfare effects," Energy Economics, vol. 31, no. 2, pp. 269 -- 277, Mar. 2009. [50] P. Srikhirin, S. Aphornratana, and S. Chungpaibulpatana, "A review of absorption refrigeration technologies," Renewable and Sustainable Energy Reviews, vol. 5, no. 4, pp. 343 -- 372, Dec. 2001. 25 [51] J. P. Kotzé, T. W. von Backström, and P. J. Erens, "High Temperature Thermal Energy Storage Utilizing Metallic Phase Change Materials and Metallic Heat Transfer Fluids," Journal of Solar Energy Engineering, vol. 135, no. 3, pp. 035001 -- 035001, May 2013. [52] EnergyPlusTM. . [53] Spanish Government, "Spanish Building Code (Código técnico de la edificación): Orden FOM/1635/2013, de 10 de septiembre, por la que se actualiza el Documento Básico DB- HE «Ahorro de Energía», del Código Técnico de la Edificación, aprobado por Real Decreto 314/2006, de 17 de marzo, Boletín Oficial del Estado.," 2013. [54] B. Atanasiu, J. Maio, D. Staniaszek, I. Kouloumpi, and T. Kenkmann, "ENTRANZE Project, "Overview of the EU-27 Building Policies and Programs. Factsheets on the nine Entranze Target Countries," 2014. [55] JRC IPTS, 2012, "Green Public Procurement: Windows Technical Background Report. Windows, Glazed Doors and Skylights," 2012. [56] "PVsyst Software." [Online]. Available: http://www.pvsyst.com/en/software. [Accessed: 11-Sep-2018]. [57] J. Lagarias, J. Reeds, M. Wright, and P. Wright, "Convergence properties of the Nelder- Mead simplex method in low dimensions," Siam Journal on Optimization, vol. 9, no. 1, pp. 112 -- 147, Dec. 1998. [58] A. JÄGER-WALDAU, "PV Status Report 2018," EUR 29463 EN, 2018. [59] A. M. Elshurafa, S. R. Albardi, S. Bigerna, and C. A. Bollino, "Estimating the learning curve of solar PV balance -- of -- system for over 20 countries: Implications and policy recommendations," Journal of Cleaner Production, vol. 196, pp. 122 -- 134, Sep. 2018. [60] A. Louwen, W. G. J. H. M. van Sark, A. P. C. Faaij, and R. E. I. Schropp, "Re-assessment of net energy production and greenhouse gas emissions avoidance after 40 years of photovoltaics development," Nature Communications, vol. 7, p. 13728, Dec. 2016. [61] R. Dones, T. Heck, and S. Hirschberg, "Greenhouse Gas Emissions From Energy Systems: Comparison And Overview," 2004. [62] R. W. Howarth, "A bridge to nowhere: methane emissions and the greenhouse gas footprint of natural gas," Energy Science & Engineering, vol. 2, no. 2, pp. 47 -- 60, 2014. [63] A. Moro and L. Lonza, "Electricity carbon intensity in European Member States: Impacts on GHG emissions of electric vehicles," Transportation Research Part D: Transport and Environment, vol. 64, pp. 5 -- 14, Oct. 2018. [64] N. Pardo Garcia, K. Vatopoulos, A. Krook Riekkola, A. Perez Lopez, and L. Olsen, "Best available technologies for the heat and cooling market in the European Union," Joint Reserach Center, EUR 25407 EN, 2012. [65] G. Grossman, "Solar-powered systems for cooling, dehumidification and air- conditioning," Solar Energy, vol. 72, no. 1, pp. 53 -- 62, Jan. 2002. [66] D. S. Kim and C. A. Infante Ferreira, "Solar refrigeration options -- a state-of-the-art review," International Journal of Refrigeration, vol. 31, no. 1, pp. 3 -- 15, Jan. 2008. [67] Y. Fan, L. Luo, and B. Souyri, "Review of solar sorption refrigeration technologies: Development and applications," Renewable and Sustainable Energy Reviews, vol. 11, no. 8, pp. 1758 -- 1775, Oct. 2007. [68] E. Vartiainen, G. Masson, and C. Breyer, "The True Competitiveness of Solar PV. A European Case Study," European PV Technology and Innovation Platform (ETIP PV), 2017. [69] C. K. Ho, M. Carlson, P. Garg, and P. Kumar, "Cost and Performance Tradeoffs of Alternative Solar-Driven S-CO2 Brayton Cycle Configurations," presented at the ASME 2015 9th International Conference on Energy Sustainability collocated with the ASME 2015 Power Conference, the ASME 2015 13th International Conference on Fuel Cell Science, Engineering and Technology, and the ASME 2015 Nuclear Forum, 2015, p. V001T05A016-V001T05A016. [70] S. Quoilin, M. V. D. Broek, S. Declaye, P. Dewallef, and V. Lemort, "Techno-economic survey of Organic Rankine Cycle (ORC) systems," Renewable and Sustainable Energy Reviews, vol. 22, pp. 168 -- 186, Jun. 2013. 26 [71] P. A. Pilavachi, "Mini- and micro-gas turbines for combined heat and power," Applied Thermal Engineering, vol. 22, no. 18, pp. 2003 -- 2014, Dec. 2002. [72] A. Datas and A. Martí, "Thermophotovoltaic energy in space applications: Review and future potential," Solar Energy Materials and Solar Cells, vol. 161, pp. 285 -- 296, Mar. 2017. [73] A. Mazzetti, M. Gianotti Pret, G. Pinarello, L. Celotti, M. Piskacev, and A. Cowley, "Heat to electricity conversion systems for moon exploration scenarios: A review of space and ground technologies," Acta Astronautica, vol. 156, pp. 162 -- 186, Mar. 2019. [74] D. N. Woolf et al., "High-efficiency thermophotovoltaic energy conversion enabled by a metamaterial selective emitter," Optica, OPTICA, vol. 5, no. 2, pp. 213 -- 218, Feb. 2018. [75] EASE (European Association for Storage of Energy), "Thermal Hot Water Storage." . [76] L. F. Cabeza, E. Galindo, C. Prieto, C. Barreneche, and A. Inés Fernández, "Key performance indicators in thermal energy storage: Survey and assessment," Renewable Energy, vol. 83, pp. 820 -- 827, Nov. 2015. 27 Figure 1. Illustration of the electrical (grey dashed lines) and thermal (red solid lines) energy flows in the system configurations analysed in this study: (a) reference case comprising a boiler for heating and an electrically driven heat pump (EHP) for cooling, (b) PHPS-E configuration comprising an EHP for cooling, a solar PV system, a high- and low- grade thermal stores (HTES and LTES, respectively), and a power generation unit (PGU), c) PHPS-T configuration similar to the PHPS-E but using a thermally-driven heat pump (THP) instead of an EHP. Notice that in each case, the variables take different values (e.g. the electricity consumption Ce is lower in the PHPS-T system than in the other two cases). The same notation is used to avoid an excessively complicated notation. 28 Figure 2. Energy management algorithm and system model equations. 29 Figure 3. Simulation results for a PHPHS-E system (Figure 1) under Scenario 1 (Table 1). Other system parameters as set to the values in Table 2. (a) Hourly variation of both grid- and self-consumed electricity along with the thermal energy stored in the HTES (top) and the hourly variation of both grid- and self-consumed heat along with the thermal energy stored in the LTES (bottom) for an optimized system with 𝜂𝑃𝐺𝑈 = 30%, 𝑃𝑛𝑜𝑚,𝑃𝑉 = 10.6 kWel, 𝑃𝑚𝑎𝑥,𝑃𝐺𝑈 = 1.35 kWel, and 𝐸𝑚𝑎𝑥,𝐻𝑇𝐸𝑆 = 33.7 kWth; (b-d) Contour plots representing the savings of electricity, in percentage with respect to the reference case (Figure 1), for an optimized system and two PGU conversion efficiencies (20 % and 40 %) as a function of (b) the CAPEX of PGU and HTES, (c) the HTES temperature and 𝑎𝐻𝑇𝐸𝑆, (d) the CAPEX of PV system and the difference between WACC and the energy price inflation rate. 30 Figure 4. Performance parameters of an optimized PHPS-E system (Figure 1-b) installed in Madrid as a function of the PGU conversion efficiency, evaluated for the four different economic scenarios indicated in Table 2. 31 Figure 5. Contribution of each kind of energy loss in a PHPS-E system optimized for each scenario indicated in Table 2: (a) Thermal insulation losses in the HTES, (b) thermal insulation losses in the LTES, (c) PV electricity that is not used nor stored, (d) waste heat from the PGU that is not stored in the LTES. 32 Figure 6. (top) Contour plots representing the discounted payback period resulting from the installation of an optimized PHPS-T system as a function of the COP of THP (used in the PHPS-T system) and the COP of the EHP (used in the reference system); (bottom) contour plots representing the difference between the discounted payback period of a PHPS-E and a PHPS-T solution as a function of the COP of THP (used in the PHPS-T system) and the COP of the EHP (used in the PHPS-E system). Positive values (in blue) represent a PHPS-E solution with a shorter payback period than the PHPS-T one. Results are shown for four different PGU efficiencies. The rest of the parameters are taken from Table 1 and the Scenario 1 in Table 2. 33 Figure 7. Performance parameters of an optimized PHPS-T system (Figure 1-c) installed in Madrid as a function of the PGU conversion efficiency, evaluated for the four different economic scenarios indicated in Table 2. 34 Table 1. Summary of the techno-economic variables used to describe the PHPS+PV system. Parameter Value Units % % % ºC ºC ∗ ) ) Variable (Table 2) Variable (optimized) Variable (optimized) kWel kWhth kWel Variable (optimized) Independent variable ∗ = 𝐶𝐴𝑃𝐸𝑋𝑇𝐻𝑃 ∗ ) 30 [75], [76] 500 [7][64] 70 (for PHPS-E) 130 (for PHPS-T) €/kWhth €/kWel €/kWel €/kWhth €/kWcool Variable (Table 2) Variable (Table 2) Variable (Table 2) Variable (Table 2) PGU conversion efficiency (𝜂𝑃𝐺𝑈) PV nominal installed power (𝑃𝑛𝑜𝑚,𝑃𝑉) HTES maximum capacity (𝐸𝐻𝑇𝐸𝑆,𝑚𝑎𝑥) PGU maximum generation capacity (𝑃𝑚𝑎𝑥,𝑃𝐺𝑈) Inflation for electricity and fuel (𝐼𝑛𝑓𝑙𝑒 = 𝐼𝑛𝑓𝑙𝑒𝑙 = 𝐼𝑛𝑓𝑙𝑓𝑢𝑒𝑙) Weighted average cost of capital (𝑊𝐴𝐶𝐶𝑛𝑜𝑚) CAPEX of HTES (𝐶𝐴𝑃𝐸𝑋𝐻𝑇𝐸𝑆 ∗ CAPEX of PGU (𝐶𝐴𝑃𝐸𝑋𝑃𝐺𝑈 ) ∗ ) CAPEX of PV (𝐶𝐴𝑃𝐸𝑋𝑃𝑉 ∗ CAPEX of LTES (𝐶𝐴𝑃𝐸𝑋𝐿𝑇𝐸𝑆 CAPEX of EHP and THP (𝐶𝐴𝑃𝐸𝑋𝐸𝐻𝑃 LTES temperature (∆𝑇𝐿𝑇𝐸𝑆) HTES temperature (∆𝑇𝐻𝑇𝐸𝑆) Parameter "aLTES" for heat losses Parameter "aHTES" for heat losses Overall inflation (𝐼𝑛𝑓𝑙) LTES maximum capacity (𝐸𝐿𝑇𝐸𝑆,𝑚𝑎𝑥) Installation lifetime (𝑇) CO2 emissions of external fuel heating CO2 emissions of external grid electricity CO2 life-cycle equivalent emissions of PV electricity CO2 life-cycle equivalent emissions of HTES, LTES and PGU COP of EHP (𝐶𝑂𝑃𝐸𝐻𝑃) COP of THP (𝐶𝑂𝑃𝑇𝐻𝑃) Variable cost of grid electricity (𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑣𝑎𝑟 Fixed cost of grid electricity (𝑂𝑃𝐸𝑋𝑒𝑙𝑒𝑐,𝑓𝑖𝑥 Variable cost of fuel (𝑂𝑃𝐸𝑋𝑓𝑢𝑒𝑙,𝑣𝑎𝑟 Fixed cost of fuel (𝑂𝑃𝐸𝑋𝑓𝑢𝑒𝑙,𝑓𝑖𝑥) () Due to the difficulty on estimating the life cycle emissions for the HTES, LTES and PGU devices, we neglect their contribution. Thus, the CO2 equivalent emissions will need to be corrected accordingly, when confident data become available. 250 [61] 340 [63] 20 [60] 1200 0.1 0.1 2 1.3 [35], [66], [67] 0.17 50 0.07 % kWhth years €/kWhth €/year €/kWhel €/kWel-year Neglected () 4 [7] W·K-1dm-3/2 W·K-1dm-3/2 ) ) ∗ ∗ ) gCO2eq/kWhth gCO2eq/kWhel gCO2eq/kWhel gCO2eq/kWhel - - 20 25 60 ∗ 35 Table 2. Economic scenarios considered in the study Scenario Scenario 1 Scenario 2 Scenario 3 Scenario 4 WACCnom (%) Infle (%) CAPEXHTES CAPEXPGU (€/kWhth) (€/kWel) CAPEXPV (€/kWel) 2 3 4 2 30 100 200 300 900 1000 2000 1200 36 Table 3. Summary of the results for the two kinds of systems (PHPS-E and PHPS-T) indicated in Figure 1, and the four scenarios described in Table 2. The rest of parameters are set to the values indicated in Table 1. Optimized variables Merit figures Scenario PGU Eff (%) PGU size HTES size PV size (kWel) (kWhth) (kWel) Total CAPE X () (k€) Discounte d payback period (years) PV self- consumptio n ratio (%) Electricit y saves (%) Fuel saves (%) CO2 emission saves (ton/year) PHPS-E Scenario 1 PHPS-E Scenario 2 PHPS-E Scenario 3 PHPS-E Scenario 4 PHPS-T Scenario 1 PHPS-T Scenario 2 PHPS-T Scenario 3 PHPS-T Scenario 4 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 3.11 2.15 1.25 0.94 0.82 0.16 0.79 0.04 0.02 0.04 0.04 0.02 0.80 0.80 0.80 0.80 0.79 0.04 0.71 0.04 0.02 0.05 0.04 0.02 49.02 11.48 12.74 62.87 11.62 12.99 29.47 16.67 22.20 3.85 8.53 0.01 0.01 0.01 0.00 0.01 14.18 17.19 19.37 7.73 10.24 0.01 6.82 0.00 0.01 0.00 0.00 0.01 8.37 6.20 7.63 2.29 3.47 1.33 1.45 1.34 1.34 1.40 4.68 5.50 6.02 3.36 3.82 1.75 2.85 1.32 1.23 1.14 1.18 1.19 8.79 10.05 12.20 3.28 5.81 1.64 1.76 1.68 1.69 1.71 4.88 5.71 6.24 5.60 6.39 2.14 4.81 1.62 1.50 1.46 1.49 1.46 11.82 12.99 12.22 14.14 15.51 14.70 17.11 13.45 13.54 13.53 13.66 13.84 9.14 9.59 9.52 11.38 12.17 7.95 13.36 8.47 8.51 8.33 8.44 8.55 41.0 38.6 48.7 50.2 48.1 49.7 56.2 70.4 70.4 71.6 70.6 69.0 42.3 45.9 56.6 52.6 53.8 66.4 51.3 63.2 59.7 63.9 61.6 58.5 95.3 91.7 69.3 75.6 75.0 28.1 54.3 20.5 20.7 20.9 20.6 20.2 85.6 84.8 77.9 76.4 75.0 49.3 71.8 47.4 45.8 46.5 46.2 45.4 19.0 18.0 19.8 9.6 14.8 3.7 4.4 3.8 4.5 3.8 3.9 4.1 -17.6 -12.5 -3.2 -17.5 -14.6 -16.2 -19.2 -18.8 -19.3 -19.7 -19.5 -19.6 () The addition of the CAPEX of PGU, HTES and PV systems 1.69 1.59 1.46 1.20 1.36 0.45 0.80 0.39 0.41 0.39 0.39 0.39 0.32 0.49 0.74 0.24 0.32 -0.02 0.13 -0.14 -0.18 -0.18 -0.18 -0.19 37 Table 4. Identical results than in Table 3, but for the case of loss-less (adiabatic) HTES and LTES (i.e. 𝑎𝐻𝑇𝐸𝑆 = 𝑎𝐿𝑇𝐸𝑆 = 0). The values in this table represent the upper bound for performance (unattainable in practice) of PHPS concept under the selected scenarios. Optimized variables Merit figures Scenario PGU Eff (%) PGU size HTES size PV size (kWel) (kWhth) (kWel) PHPS-E Scenario 1 PHPS-E Scenario 2 PHPS-E Scenario 3 PHPS-E Scenario 4 PHPS-T Scenario 1 PHPS-T Scenario 2 PHPS-T Scenario 3 PHPS-T Scenario 4 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 60 40 20 3.35 2.78 1.32 2.13 0.87 0.71 0.78 0.78 0.17 0.67 0.30 0.03 0.80 0.80 0.80 0.79 0.80 0.47 0.79 0.72 0.13 0.61 0.20 0.02 51.92 52.84 32.09 31.25 20.47 20.14 13.74 13.53 6.76 12.24 6.29 0.27 17.69 19.62 24.01 9.55 12.62 14.78 8.83 11.13 5.56 6.98 4.77 0.18 6.77 8.67 7.31 5.93 5.86 5.20 4.05 3.77 2.01 3.65 2.32 1.42 3.11 3.94 5.34 2.94 3.42 3.68 2.48 2.84 1.54 2.18 1.69 1.17 Total CAPE X() (k€) 8.65 10.22 7.93 12.37 9.95 8.97 7.01 6.66 3.25 8.16 4.64 1.81 3.57 4.38 5.77 5.27 6.16 6.36 4.64 5.24 2.53 5.23 3.39 1.49 Discounte d payback period (years) 7.78 8.80 9.26 11.69 11.25 12.65 11.27 12.69 12.81 14.28 14.16 11.55 5.86 6.25 6.98 8.39 9.06 9.67 9.51 10.30 9.12 11.78 10.44 7.48 PV self- consumptio n ratio (%) 81.3 70.4 73.5 85.9 81.6 80.9 93.4 92.5 98.8 95.2 99.0 100 100 100 97.0 100 100 99.7 100 100 100 100 100 100 Electricit y saves (%) 99.0 97.5 78.5 95.4 81.5 63.3 78.4 68.7 32.4 73.0 44.1 21.1 93.3 91.8 86.6 90.4 87.7 74.4 86.8 82.4 54.3 80.6 61.4 45.8 Fuel saves (%) CO2 emission saves (ton/year) 24.8 30.7 28.8 22.1 22.7 22.1 14.4 14.2 9.8 13.1 9.7 7.5 -10.7 -2.8 9.6 -11.6 -6.5 -0.9 -14.8 -10.4 -15.1 -15.9 -15.5 -16.3 2.09 2.25 1.96 1.96 1.80 1.56 1.49 1.36 0.75 1.38 0.90 0.53 0.74 1.00 1.38 0.67 0.82 0.85 0.51 0.61 0.09 0.39 0.17 -0.06 () The addition of the CAPEX of PGU, HTES and PV systems 38
1908.00197	1	1908	2019-08-01T03:46:38	Solution Processability and Thermally Activated Delayed Fluorescence: Two Steps towards Low-Cost Organic Light Emitting Diodes	[ "physics.app-ph" ]	Organic Light Emitting Diodes (OLED) have experienced vast attention in the scientific literature as the leading contender for the next generation of electronic lighting technology. However, while research continues to push the limit for OLED performance and efficiencies, minimal attention is paid to the processability and suitability for mass production. In this regard, Solution Processability (SP) and Thermally Activated Delayed Fluorescence (TADF) in OLEDs are two recent discoveries which have significantly improved the capacity for low-cost processing without compromise to their performance efficiencies. This meta-analysis is focused on these discoveries and their application to OLEDs. Recent peer-reviewed studies of OLEDs employing both SP and TADF are reviewed in order to highlight physical fundamentals and discover synergistic effects when incorporating both factors into a single OLED device. Additionally, some other recent trends targeting efficient manufacturing of OLEDs are identified, and logical future directions for interested researchers are discussed.	physics.app-ph	physics	Solution Processability and Thermally Activated Delayed Fluorescence: Two Steps towards Low-Cost Organic Light Emitting Diodes Sahadev Somasundaram School of Chemical Engineering, The University of Queensland St. Lucia, QLD 4072, Australia [email protected] Abstract Organic Light Emitting Diodes (OLED) have experienced vast attention in the scientific literature as the leading contender for the next generation of electronic lighting technology. However, while research continues to push the limit for OLED performance and efficiencies, minimal attention is paid to the processability and suitability for mass production. In this regard, Solution Processability (SP) and Thermally Activated Delayed Fluorescence (TADF) in OLEDs are two recent discoveries which have significantly improved the capacity for low-cost processing without compromise to their performance efficiencies. This meta-analysis is focused on these discoveries and their application to OLEDs. Recent peer-reviewed studies of OLEDs employing both SP and TADF are reviewed in order to highlight physical fundamentals and discover synergistic effects when incorporating both factors into a single OLED device. Additionally, some other recent trends targeting efficient manufacturing of OLEDs are identified, and logical future directions for interested researchers are discussed. Keywords: Organic light emitting diode; solution processable; thermally activated delayed fluorescence; OLED; TADF; organic semiconductor. Introduction Organic light emitting diodes (OLED) are expected to serve as the next generation of technology for electronic lighting and display applications. While not yet sufficiently developed for broad commercialisation, OLEDs are beginning to show superior performance to their inorganic counterparts -- in fact, several global electronics developers have opted for OLED-based displays in their product lines. The organic materials used in OLEDs carry added versatility in that they can maintain complete opto-electronic function when deposited onto flexible plastics or even fabric- type substrates. Furthermore, the underlying organic molecular structure allows for endless possibilities in tailoring production and performance efficiencies through unique molecular designs. Finally, recent research has focused on two inherent properties that dramatically improve production efficiency, which are only available to organic materials: Thermally Activated Delayed Fluorescence (TADF) and Solution Processability (SP). The first notable demonstration of TADF by Chihaya Adachi and collaborators in 2012 [1] essentially quadrupled the possible performance efficiencies of conventional fluorescent OLEDs. By utilising a phenomena formerly termed "E-type delayed fluorescence" [2], the group discovered that purely organic materials can harness light emission from triplet-state excitons otherwise only accessible to materials containing expensive heavy metals. Thus, TADF materials not only improve the efficiency of conventional OLEDs but reduce the production cost by eliminating the need for expensive materials. Since discovery, most TADF studies have focused on achieving greater device performance through innovative molecular design. Over the previous few years, several reports have demonstrated achieving near 100% Internal Quantum Efficiency (IQE) [3,4], as well as External Quantum Efficiency (EQE) higher than the most promising phosphorescent OLEDs [5,6]. As a consequence, the production efficiency, or the ability to be manufactured at the commercial scale with minimal cost, receives considerably less attention. Studies of TADF-based materials and devices are widespread throughout the scientific literature; the mechanisms and significance of TADF are explained in numerous reviews, books and articles [1,7-14], and are thus not discussed here. Solution processing is significantly cheaper than the typical vacuum-based deposition techniques, which require expensive equipment and a high resource cost per unit area of deposited material. Solution processing techniques also require less technical expertise for operation, are more suitable for large-scale depositions and are compatible with the existing roll-to-roll (R2R) processes used for mass production of thin-film semiconductors. Solution processing occurs from a solution of the deposition materials dissolved in a solvent, which is to be evaporated through the deposition process. It should be noted that while solubility is considered a major advantage of organic semiconductor materials, not all such materials are equally, or even sufficiently, soluble in suitable solvents; thus, solubility is a critical property when evaluating materials for solution processing. High solubilities are often associated with a greater presence of alkyl and alkoxyl moieties and a longer aliphatic chain [15]; this is typically seen in polymers. Higher solubility will i) result in a more uniform and homogeneous film morphology, removing the potential for charge trapping; and ii) allow a more uniform distribution of dopants in host:dopant complex films, resulting in a lower chance of concentration quenching [16] and aggregation-caused quenching [17,18]. Many industrial reports and popular science articles recognise that despite the exceptional performance of modern OLEDs, the cost involved with manufacturing is preventing them from dominating the electrical lighting and electronic display markets. It is believed that the key to these manufacturing issues lies in breakthroughs of the fundamental physics and chemistry of materials processing and OLED design, which is likely to be found in scientific literature and academic publishing (as opposed to grey literature). Thus, the aim of this review is to: and TADF EML; o Review all recent peer-reviewed papers examining OLED devices using a solution-processed o Examine the concepts of TADF and solution processing in organic semiconductors, and identify any issues uniquely associated with OLED devices employing both phenomena, incl. synergistic effects; and o Highlight significant findings and identify a direction for future studies regarding solution processability, TADF and economic design of OLEDs. Given the relatively low number of studies with exact relevance (i.e. development or discussion of OLED devices with both SP and TADF), articles are reviewed on a case-by-case basis and categorised with similar papers. Then, the findings are summarised and discussed with respect to the aims listed above. Articles not showing significant findings or discussions relevant to the advancement of SP TADF OLEDs or low-cost processing are not discussed here. Whereas other recent reviews on SP TADF OLEDs focus on novel molecular designs and structure-property relationships [19,20], this review retains focus on factors that contribute towards low cost manufacturing and the ability to be translated from research to commercialisation. 2 Review Table 1. Performance factors of solution processed, thermally activated delayed fluorescence OLEDs found in recent literature Reference Photoluminescence Quantum Yield (PLQY) (%) Solution Film Singlet- Triplet Splitting ∆EST (eV)a Photolumiscence Wavelength, λPL (nm) Solution Film Maximum External Quantum Efficiency (EQEmax) (%) 26 94 100 94 92 95 96 95 76 70 34 37 11 20 65 55 16 94 67 78 45 98 44 68 48 [21] [22] [23] [24] [25] [26] [27] [16] [28] [29] [30] [24] [31] [32] 26 52 31 8.5 52 40 45 49 12 47 95 0.1 0.03 0.06 0.06 0.13 0.07 0.08 0.07 0.07 0.09 0.058 0.048 0.12 0.05 0.01 0.07 0.31 0.18 0.13 0.07 0.09 0.20 550 540 575 608 478 484 509 536 429 560 424 498 484 492 464 9.3 2.4 3.4 1.5 6.0 9.4 9.5 8.2 16.1 16.0 17.5 15.2 1.0 8.9 13.9 8.1 8.2 18.1 4.43 12.9 12.4 7.0 12.2 2.2 9.4 9.9 3 [26] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45,46] [47] [48] [49] [50] [51] 11 20 65 71 44 56 71 42 61 46 2 78 12 3.8 65 76 81 77 75 74 32 58 76 60 64 75 39 28 52 25 65 92 74 1.0 42 58 0.12 0.05 0.01 0.16 0.18 0.17 0.20 0.24 0.23 0.21 0.05 0.24 0.22 0.04 0.11 0.05 0.15 0.08 0.13 0.06 0.06 0.11 0.15 0.04 0.28 0.20 0.20 0.33 0.25 478 484 509 478 492 498 506 590 594 591 490 510 608 439 452 407 483 510 443 453 520 499 606 502 490 487 421 438 553 545 541 460 477 522 445 475 549 572 612 473 505 507 15 16 1.0 8.9 13.9 1.2 1.1 4.0 4.3 18.7 ~24 2.3 11.05 1.2 5.5 11.8 1.56 12.6 2.4 3.7 23.8 23.5 8.1 5.8 7.5 4.3 14.9 17.8 20.0 12.0 5.2 9.0 1.09 6.5 10.1 4 [52] [53] [54] [55] [56] [57] [58] [59] [60] 52 2.1 45 30 22 63 71 64 0.8 56 (powder) 87 (powder) 75 (powder) 61 94 88 50 67 66 60 52 44 39 74 70 59 0.17 0.1 0.18 0.17 0.18 0.06 0.05 0.17 0.2 0.35 0.46 0.42 0.40 451 444 444 448 550 590 536 540 465 476 550 533 536 539 556 507 511 512 513 5.7 3.6 1.4 2.0 1.2 13.9 3.18 3.72 8.47 19.1 8.0 15.5 17.1 10.0 20.1 15.2 1.8 1.4 16.1 a Calculated via the singlet (S1) and triplet (T1) energies determined via onset of fluorescence and phosphorescence. Table 1 presents a comparison of literature on OLED devices employing both SP and TADF within the emissive layer. To maintain concision, only primary performance figures are included to allow for general comparison of device performances across studies. Photoluminescence quantum yield, or PLQY (%), describes the ratio of photons emitted to photons absorbed during controlled photoluminescence in either solution or neat film states. Singlet-triplet splitting, or ∆EST (eV), represents the energetic difference between the lowest singlet (S1) and triplet (T1) electronic states, which can be conceptualised as the energetic gap electrons must overcome through Reverse Intersystem Crossing (RISC) and emit via the delayed fluorescence pathway. Photoluminescence wavelength, or λPL (nm), describes the wavelength of emission during photoluminescence in the solution or neat film states, which is indicative of colour. External Quantum Efficiency, or EQE (%), describes the ratio of photons emitted to electrons injected during controlled electroluminescence, which reflects the overall device efficiency. Detailed analyses and discussions of each study are included in the following sections. 5 Polymers The foremost benefit of polymers is their high solubility in organic solvents due to their long aliphatic chains. On the other hand, their large size and variability in molecular structure between production batches results in difficulties controlling molecular design factors that affect TADF emission performance. An example of such a design factor is the distances and torsions between donor and acceptor moieties; both must be separated in order to reduce overlap of the HOMO and LUMO (typically localised around the donor and acceptor, respectively), which is directly related to ∆EST [61]. In 2015, Nikolaenko, et al. [58] developed an SP TADF Emissive Layer (EML) made of a polymeric material. In this case, the donor and acceptor moieties served as distinct monomers in the greater polymer chain, and polymerisation amounts were controlled to vary charge transport properties and buffer the recombination zone. Other monomers were selectively added to control the electronic, optical and physico-chemical properties, such as solubility and viscosity, and a block- polymer structure was used to ensure tight control of polymer sequence. Interestingly, the experimental values of the containing devices showed a relatively low PLQY (in air) of 41% with estimated RISC yields of 83%, suggesting that light conversion, rather than triplet conversion, was limiting the efficiency of the device. Finally, the authors noted that the long-chain nature of the polymer EML may have resulted in horizontal alignment of the dipole moment of the EML polymer with the substrate, which is known to improve efficiency [62,63]. Nobuyasu, et al. [36] continued the foray into TADF polymers. Devices were made using two copolymers consisting of a known small molecule emitter. Both solution-processed polymer devices showed significantly lower efficiencies than vacuum-deposited small molecule devices and exhibited noticeable roll-off. Regardless, the authors noted that their polymer device efficiencies were comparable to those of the major polymer TADF effort by Nikolaenko, et al [58] the previous year. The low polymer performance was attributed to weak electronic properties as a result of unrefined solution-processing methods, without any further discussion given. The authors presented two primary challenges of TADF polymers: i) high molecular weight compounds cause high ∆EST and non-radiative internal conversion, effectively reducing the efficiency; and ii) Triplet- Triplet Annihilation (TTA) is frequent in polymers, creating an alternative non-radiative pathway for triplet excitons besides RISC. While the solubility of polymers makes them attractive for OLED devices, these issues need to be rectified before seeing commercial use. Zhu, et al. [46] developed a conjugated polymer for use in a SP TADF OLED. In contrast to the typical alternating D-A backbone structure seen in most emissive polymers, acceptors were perpendicularly grafted to a backbone of donors to increase torsion, thus increasing HOMO-LUMO separation and reducing ∆EST. Despite the unique electronic and optical properties of conjugated polymers for optoelectronics [64,65], the characteristic effects of conjugated polymers on the current device were not discussed. The same polymer was used in an SP TADF OLED by Lin, et al. [45]. They observed high rates of triplet exciton quenching and attributed it to intra- and inter-molecular accumulation quenching between polymer chains. This accumulation quenching in TADF, which was recently studied, occurs via the close-range, electron-exchange Dexter energy transfer mechanism, as opposed to the long-range, dipole-dipole Forster mechanism [66]. To circumvent the accumulation quenching, they added an interfacial exciplex host to improve RISC rates and bolster the delayed fluorescence contribution. Indeed, quenching and roll-off were significantly reduced as evidenced by L50 and J50 (i.e. luminance and current density after EQE has reduced to 50% of maximum over prolonged operation), which were each increased nearly three-fold as compared to the pure polymer OLED. 6 A study by Xie, et al. [60] developed an SP TADF polymer using the same side-chain grafting strategy mentioned above and a TADF small molecule reported earlier [67,68] as an emissive repeating unit. The devices utilised a novel transfer mechanism where an additional TADF small molecule was included as an "assistant dopant", which mediates charge transfer between the host and TADF dopant and contributes towards a portion of the RISC processing of singlet excitons. Devices with optimised structure showed remarkable EQE of 16% with near 100% IQE. While the assistant dopant strategy is likely to improve device efficiencies, a three-component-EML is likely to require extra design considerations and, consequently, production costs for commercial production. Interestingly, the results also indicated that RISC rates are higher in polymers than small molecules. Ren, et al. [59] developed a series of copolymer and homopolymer SP TADF OLEDs to illustrate the effects of polymer structure on TADF performance. In this case, emissive pendant moieties consisting of a phenothiazine donor and dibenzothiophene-dioxide acceptor were partitioned by inert styrene moieties in the greater polymer chain. This design was intended to boost emissive efficiency by reducing intermolecular interactions between pendant units. Indeed, the copolymers showed much lower ∆EST and significantly higher EQE, Current Efficiency (CE) and Power Efficiency (PE) than the homopolymer, and copolymers showed increasing performance along with increasing degree of copolymerisation (i.e. ratio of styrene units to emissive units). Partitioning the emissive units was believed to reduce quenching via internal conversion and triplet-triplet annihilation. However, no discussion was provided on the effect of molecular structure on solubility. Small Molecules Small molecules are relatively low molecular weight (as compared to polymers) tightly-packaged molecules that form discrete particular neat films. As opposed to polymers, small molecules are not always soluble in ideal solvents, and thus present an extra complication if solution processing is desired. On the other hand, the tightly-packed, discrete particle film structure of small molecular neat films are more electrically conductive and thermally stable than polymers, and are thus more resistant to dissociation after formation of the neat film. A review of solution-processable small molecules for OLEDs can be found here [15]. In what might be the first of its kind, Cho, et al. [16] produced three separate solution processed TADF devices with EQEmax values of up to 18.1%. One emitter -- t4CzIPN -- showed higher device efficiencies when fabricated via solution processing instead of vacuum deposition. In this case, it was believed that the t-butyl moiety improved solubility (0.4%) as compared to other designed emitters (0.1%) and resulted in a more stable film morphology. Further discussion purported that greater solubility resulted in a more uniform distribution in the final film which: i) subdues exciton quenching; and ii) minimises interstitials and defects, which lead to charge trapping -- both of these factors will reduce overall quantum efficiency. Furthermore, when deposited in higher doping concentrations (3% and 5%), the resultant film morphology was noticeably more disrupted, which was believed to cause more charge trapping. Indeed, these devices showed lower current densities and quantum efficiencies. As noted earlier, the higher solubility of the t4CzIPN emitter might well be a result of the longer and more planar carbon backbone. Cho et al. [47] developed highly efficient solution processed TADF emitters by modifying 4CzIPN, replacing either 1 or 2 CN groups with F atoms (3CzFCN and 4CzFCN). The rationale behind the design was to use the hydrophobic nature of F to counteract the hydrophilicity of CN and improve solubility in organic solvents. Indeed, the novel molecules showed higher solubility in toluene, and the solubility increased according to number of F moieties (1.5 wt% for 3CzFCN, 1.0 wt% for 4CzFCN and 0.1 wt% for 4CzIPN [69]). It is important to note that toluene is commonly required for TADF studies for its ability to prevent PL quenching of TADF molecules in solution. AFM revealed a much 7 greater film morphology in 3CzFCN as a result of its higher dissolution in toluene. Furthermore, it was also believed that the weak electron withdrawing nature of the F moiety, relative to the parting CN, caused a slight bathochromic shift in the PL emission (443 nm and 453 nm for 3CzFPN and 4CzFPN, respectively, from 470 nm of 4CzIPN [69]). The study indicates that while functionalisation may be a simple yet effective method to improve solubility, PL or electrodynamic characteristics are likely to change -- this might be particularly more pronounced in TADF molecules due to their delicate electronic nature. The same group [34] later tried to improve the unstable emissive lifetimes of their previous efforts through a benzonitrile core substituted with 5 carbazole units (5CzCN). The aim of the design was to increase solubility by agglomerating as many donors as possible (while retaining the CN unit) in order to increase both the donor-acceptor dihedral torsion and the available surface area for solvent penetration. Indeed, the additional carbazole increased solubility in toluene from 0.1% for 4CzIPN, as per their previous study [69], to 0.5% for 5CzCN. However, it should be noted this figure is lower than the fluorinated benzonitriles of the group's previous effort mentioned above. Furthermore, while the lifetime was aimed to improve through extending conjugation of the benzonitrile acceptor and increase aromatic units through carbazole functionalisation, notable lifetime improvements were only detected in vacuum-deposited devices. Kim, et al. [35] focused on improving production costs not only through soluble molecules, but through simplifying device structures. The aim was to create a device with no transport layers by using a "self-organised buffer hole injection layer" (Buf-HIL). 4CzIPN dopant and CBP host were used as the EML. The Buf-HIL consisted of PEDOT:PSS and a perfluoric acid copolymer (PFI), which resulted in a neat film with graded PFI concentration - the lower surface energy of PFI allows for natural self-organisation during spin coating. This gradient caused the work function to change across the Buf-HIL, allowing each end to closely match the electronic levels of its neighbouring layer and improve internal hole transport and reduce hole accumulation at the interface (i.e. work function of the bottom end was similar to the anode, and that of the top matched with the HOMO of the host material). Additionally, the Buf-HIL also improved efficiency by reducing contact between excitons and the lower energy levels of PEDOT:PSS, which has been reported to result in exciton quenching [70,71]. A polar aprotic solvent, THF, was selected to dissolve the highly polar 4CzIPN -- indeed, solubilities were much higher in THF (>10%) than in toluene (<1%). THF not only improved film uniformity, it significantly reduced aggregation of 4CzIPN, resulting in a two-fold increase in PL intensity from a film spun from toluene as a result of aggregation-caused quenching. The authors found that solvents with higher dipole moments resulted in greater morphology and PL intensity. Ultimately, the study demonstrates a resourceful method to simultaneously improve both device efficiency and manufacturability. Adachi's group [70] performed a comparative study of a series of 4CzIPN devices with either CBP or CPCB host, solution or vacuum processed layers and varying Electron Transport Layers (ETL). Interestingly, CBP devices showed no notable variation in EQEmax between solution or vacuum processed devices, while those of CPCB showed up to a 31% decrease. Lifetimes were significantly lower in most devices when solution processed. However, solution processed CPCB devices had much improved lifetimes than solution processed CBP devices, which was attributed to a higher molecular weight, greater solubility and better film morphology. Xie et al. [25] developed two small-molecule emitters that were both vacuum and solution deposited and compared. The solution deposited devices of each molecule showed comparable EQEs to their vacuum-deposited counterparts -- this was even more notable considering that the vacuum deposited devices did not include hole transport layers. 8 Not long after the worldwide recognition of TADF, Chen, et al. [55] developed one of the first and most notable devices utilising both TADF and solution processability. The idea was to utilise the inherent narrow ∆EST found in transition metal complexes for TADF without the natural instability of rare phosphorescent transition metals like iridium or platinum. The same group [28] later developed a cuprous-complex small molecule which showed clear TADF in solution processed devices. The device was made by using a carbazole-based molecule, czpzpy, as both a ligand and host material for a separate emissive dopant, [Cu(CH3CN)2(POP)]BF4 -- both materials were simply spin deposited in tandem without prior preparation. The ligand was used in excess as an attempt to reduce dissociation of the emissive dopant and thus improve PLQY in the device. Verma, et al. [54] developed a copper iodide complex for use in an inkjet-printed EML. A Copper- Iodide core was substituted with alkyl phosphites to improve solubility in nonpolar solvents suitable for inkjet deposition. The molecular design rationale was to include asymmetrically-substituted ligands in order to reduce the conformational possibilities, thereby increasing the entropy contribution and solubility. Solubilities in polar solvents such as ethanol were between 10 -- 30 mg mL-1 but were higher than 100 mg mL-1 for non-polar solvents. The devices with a printed EML and evaporated organic layers showed an EQEmax of 13.9%. With the broad solubility behaviour, the authors noted that the molecule was suitable for various types of high-perfomance inks and print heads -- the molecule could be dissolved in solvents with tuneable viscosity or surface interactive behaviour. Sun, et al. [53] developed solution processed TADF EMLs with thermally crosslinkable host and dopant. The desire for crosslinking EML materials was not made completely clear, but it was believed to contribute to a greater thermal stability and more homogenous film morphology. The device with a film of host:dopant mass ratio of 1:0.09 showed the best performance compared to those of 1:0.06 and 1:0.12, but the reason was not elaborated on. Chen, et al. [56] developed a solution processed, non-doped, TADF emitter showing an EQEmax of 19.1% - reportedly the highest as at publication (Oct 2017). This breakthrough was attributed to a dual-charge-transfer mechanism, in which charge was transferred both through conventional D-A transport as well as through-space electron transfer -- in this case, electrons from the π-orbitals of donor and acceptor moieties in close proximity transfer through space [72,73]. Having an alternate pathway for charge transfer was believed to result in higher dipole moment and lower ∆EST. Huang, et al. [43] developed a TADF molecule with mechanochromic luminescence - colour variation depending on the solution with which the molecule was deposited. The molecule, Cz-AQ, emitted as a deep red at 680 nm when deposited through dichloroethane and yellow at 600 nm through a mixture of dichloroethane and ethanol (equivalent volume ratio). Crystals of the molecule were grown through different solutions and the crystallographies were studied to examine the luminescent behaviour. Between the two crystals, one showed stacking with an H-type aggregation along with b axis, leading to strong π-π interactions induced by significant orbital superimposition -- this was believed to cause the red shift in emission of the deposited films. TGA and DSC also revealed transition of emission colour from red to yellow in the same crystal after thermal decomposition. Unfortunately, the specific solvatochromic effect on the molecule (i.e. the mechanism by which different solvents affect the molecular orientation and emission wavelength) was not discussed. It is also worth noting that due to the closely stacked molecular orientation, Cz-AQ showed aggregation-induced emission (AIE) properties -- a mechanism in which luminescence is enhanced in aggregated states such as neat films [74]. Dendrimers 9 Dendrimers serve as a new class of molecules with sizes in between those of small molecules and polymers. Dendrimers are identified by their tree-like structure, with "branch" molecules surrounding a single core in three dimensions. These branches then attach to subsequent branch molecules, thus giving the dendrimer a tree-like structure. Branch molecules are referred to as dendrons, while the number of branches in succession is termed "generation". Dendrimers are often designed using some existing small molecule as either dendrons or the core molecule. A typically high molecular weight ensures a good degree of solubility in common solvents, but also avoids forming inhomogeneous, impure or poorly distributed thin films that often occur in polymers with indefinite molecular weight [15]. While review articles of emissive dendrimers do exist [75,76], functional dendrimer OLED devices have only appeared in the literature within the last decade, and no relevant review articles have appeared since. Albrecht, et al. [22] developed the first solution processable TADF OLED using a dendrimer-based emissive layer. The emissive dendrimers consisted of a triazine core unit attached to various generations of carbazole dendrons (up to 4 generations). Perhaps since Adachi's novel 4CzIPN molecule [1], Carbazole has become well-known moiety for functionalising TADF molecules because of its hole-transport velocity, high triplet energy, stable thermal resistance and its ability to improve solubility and resultant film morphology [31,77]. It was found that higher generation number resulted in worse performance in regard to ∆EST and PLQY (film and solution), but the highest EQE came from 3 generations. The increased intermolecular interactions resulting from higher generation numbers was believed to be the cause of lower PLQY. Regardless, each dendrimer showed clear TADF, indicating that dendrimers may well be the ideal solution to achieving both TADF and solution processability in future molecular designs. The authors also noted that carbazole- based dendrimers have been previously shown to become insoluble after photo-induced crosslinking [78] -- this is hugely beneficial for solution processed devices, as it will allow the deposited EML to remain solid after contact with solution of neighbouring layers. This orthogonal dendrimer and a device with fully solution-processed organic layers was later developed by the same group through the use of a similar triazine- and carbazole-based dendrimer [23]. Part of the strategy was to include terminating functional groups on the final generation of carbazole dendrons to increase initial solubility. Terminating groups included methyl, t-butyl, phenyl and dibenzocarbazole substituents. As expected, all funtionalised dendrimers showed good solubility in common solvents, such as chloroform, THF and toluene, and the dibenzocarbazole- functionalised dendrimer dissolved in hexane. However, no quantitative analysis nor qualitative speculation on the degree and change of solubility among variants was given. It was also shown that the functionalised dendrimers showed much better lifetimes than the non-functionalised -- this was attributed to the electrochemical stability induced by the terminating groups, and the belief that the non-functionalised dendrimer might undergo electro-polymerisation in the powered device, which generated carrier traps. The same group then continued their work and developed devices with t-butyl-substituted dendrimer as a dopant and a carbazole-dendrimer-based hosts [24]. The motivation was to select a host that could improve the PLQY of previously developed devices while maintaining orthogonality or resistance to dissolution in alcohol used to deposit ETLs. Hosts with generation numbers higher than 2 showed near-complete resistance to alcohol (likely due to higher molecular weight [79]), while that of 2nd generation was largely dissolved. Furthermore, a separate control experiment that even if the host is sufficiently resistant, alcohol can permeate the solid film of a host:dopant complex and dissolve the dopant, even if the dopant is intrinsically resistant to alcohol. The most recent effort of Albrecht's group [52] involved a imparting AIE to a solution processable TADF dendrimer. Indeed, AIE was only noticed in dendrimers, as opposed to small molecules of similar composition, due to the intermolecular interactions occurring in dendritic structures -- these 10 interactions inhibit intramolecular rotations and vibrations, which create non-radiative pathways. As noted in previous studies, the dendrimers showed good resistance to alcohols, which allows for solution deposition of neighbouring ETLs in alcohol without degradation of the EML. While the EQEs of these dendrimers were comparatively lower than the authors' previous work, the synergy of AIE with TADF and solution processability serves as an innovative method to improve both performance and cost effectiveness of devices. Hu, et al. [39] and Huang, et al. [42] recently reported molecules with AIE, TADF and solution processability, albeit with much less notable results. Li et al. [48] developed a new molecule by dendronising a DMAC-DPS core with up to two generations of t-butyl-substituted carbazole. Both dendrimers exhibited good solubility in dichloromethane, tetrahydrofuran, toluene and chlorobenzene, likely boosted by the t-butyl groups, but no quantitative solubility analysis was performed. High decomposition temperatures (471oC and 507oC for 1 generation and 2 generations, respectively) and post-annealing AFM revealed very high thermal stability and morphological homogeneity after spin coating for both dendrimers. However, different devices showed minor changes in efficiency according to varying annealing temperatures -- EQEs for the 1-generation-dendrimer device showed fluctuations up to nearly 4 percentage points through annealing at 25oC to 160oC. Regardless, all devices showed relatively low roll-offs. Ban, et al. [50] designed a dendrimer TADF emitter using alkyl, instead of π-conjugated or aromatic, chains to link dendrons to the emissive core. The rationale of this design was threefold: i) to improve solubility and film morphology by reducing molecular rigidity and steric hindrance; ii) to improve charge transport through non-conjugated linkers; and iii) to reduce emission quenching by using non-conjugated spacers to isolate the emissive core [80]. While the improvement to solubility and morphology was not verified, the alkyl-linked dendrimers indeed showed improved device performance, which was attributed to the effects predicted above. Electronic properties between dendrimers showed little variance (for example, relative standard deviation of ∆EST between dendrimers was only 13%) -- this is interesting to note as steric hindrance and rigidity are known to be greatly influential on electronic states in TADF molecules. Thus, aliphatic linkers may well be one method to reconcile solution processability and TADF. The same group later witnessed similar effects through yellow- [37] and blue-emitting [38] dendrimers. Ban et al. [57] later developed the first TADF OLED using fully solution processable organic layers. The dendrimer emitter, a cyanobenzene with two generations of carbazole dendrons (Cz-CzCN), was used as a non-doped (self-host) EML. The high molecular weight dendrimer was designed to provide resistance to alcohol solvents in order to avoid orthogonal dissolution when depositing the ETL. Thus, neat films of Cz-CzCN and a single-generation molecule (5CzCN) were subjected to absorption spectroscopy before and after rinsing in isopropyl alcohol - as expected, 5CzCN's absorption intensity reduced drastically, while that of Cz-CzCN only reduced by around 5%. Unfortunately, the mechanism behind the dendrimer's high alcohol resistance was not expounded on beyond the high molecular weight. Alkyl chains were used to link the 2nd generation dendrons, which were believed to enhance overall solubility and help maintain the electronic properties of the emissive core separate from dendrons. Furthermore, the encapsulating structure of dendrons was believed to limit exciton quenching via intermolecular interactions in the emissive core. Soon after, the same group [40] reported a new OLED device using similarly designed molecules (with 3 and 4 dendrons per generation instead of 5) as a host and guest EML, respectively. These devices were believed to use the same phenomena (i.e. inhibition of exciton quenching via molecular encapsulation and enhanced solubility via alkyl chain linkers) to achieve higher EQEs than the previous effort (EQEmax of 23.8% and 23.5% for 3 dendrons and 4 dendrons, respectively). It should be noted that this appears to be the highest EQE reported for a solution processed TADF device to date. 11 Summary Synergistic effects There do not appear to be any beneficial synergistic effects between TADF and SP; in other words, the effects of TADF and SP do not combine to boost performance to levels unexpected by each factor individually. There are, however, notable detrimental effects - soluble materials have inherent properties that can affect the performance of TADFs and thus must be considered during design stages (and vice versa): o Deposition via solvent alters atomistic morphology and molecular orientation (as compared to solid crystalline structure), and thus effects the electron orbital configuration. Consequently, this has significant effects on the excitonic and electronic pathways, which will alter photo- and electroluminescent efficiencies; to the film structure, molecular orientation or both; o Different types of solvents will alter emission characteristics, whether through alterations o Both TADF contribution and solubility are currently precisely controlled by molecular structure -- altering structure to tune one property is likely to affect the other. For example, increasing alkyl chain length to increase solubility may separate donor and acceptor moieties, disrupting the charge transfer mechanism or ∆EST; and o Solution processed neat films are expected to have greater decay lifetimes [70]. This may disrupt the contributions to fluorescence from different sources (i.e. delayed or prompt). It appears any breakthroughs in SP-TADF OLEDs will occur through trial and error and empirical observations of new molecular designs (as opposed to mechanistic analysis). Molecular Type Polymers have higher solubilities as a result of the longer aliphatic chain. However, the electronic performance of TADF polymers is initially at a disadvantage as the length, complexity and variability of polymer structures introduces difficulties in tuning the structure-related properties of TADF (e.g. ∆EST and consequent IQE, which are related to D-A separation). On the other hand, small molecules are smaller and more precise in nature, allowing one to tune structure-related properties and performances through minor alterations to molecular structure or composition. Furthermore, they are also associated with simpler and more controllable synthesis methods, which is likely to reduce the costs of large-scale production. Finally, assuming a decent dissolution, small molecules generally produce well-ordered, neatly-packed thin films, resulting in greater electronic performance and lifetime. Dendrimers are a great middle ground for low-cost design, capturing the solubility of polymers with the facile synthesis methods and the ability to precisely control molecular structure of small molecules. Dendrimers are suitable for TADF emission as the electronic structure and exciton transmission can be tuned through addition and modification of dendrons. Additionally, molecular weight can be largely varied by tuning the number and structure of dendrons; this allows one to selectively tune solubility and/or resistance to orthogonal solvents. Ken Albrecht's group has performed a number of studies into the characteristics of dendrimers and development of suitable dendrimer-based SP TADF OLEDs. Future work could focus on finding novel dendrimer and dendron structures for optimal performance along with suitable solvents and device structures. 12 Future Research Solubility While recognised as a critical factor for OLED design, few experimental studies actually discuss solubility or the consideration thereof before design. Since solubility is not easily quantified, many researchers simply rely on qualitative descriptors such as length of alkyl chains or molecular size to evaluate solubility. Glass transition temperature is one parameter that is associated with solubility [15,81]. While many of the studies reviewed herein do determine Tg as part of experimental procedure, none have discussed it with regards to solubility. Instead, more effort should be given to quantify solubility and develop generic correlations between solubility and device stability or performance. In a review on SP small molecules for OLEDs, Duan, et al. [15] noted that the effect of solvent on solubility and film quality needs further investigation. The Solubility Parameter [82,83] attempts to numerically predict interactive dissolution via the intermolecular and diffusive forces. Derivations of the solubility parameter method have been used to study ink-based deposition of organic semiconductors in some theoretical studies [84-87], albeit with unpredictable results. Sanders, et al. [88] recently developed a methodical design process for OLED materials -- determination of solubility parameters through computational molecular dynamics (MD) simulations was included in one step. Numerical correlations between solubility and resultant film properties would assist future researchers in SP OLED design. SP-electrodes While solution processability is important in organic emissive materials, each device layer, including the electrode material, must be compatible with R2R processes in order to experience complete low-cost, large-scale manufacturing; this includes the electrode, which is typically metallic. Several articles have been devoted to SP metallic electrodes via deposition of metallic ink -- a suspension of metallic nanoparticles in a colloidal system [89-93]. Unfortunately, as is common in bulk electrodes, this research focuses on the noble metals Au or Ag, which are not economically suitable for large- scale manufacture. As an alternative, Lee, et al. [94,95] have studied Al-based electrodes deposited via metallic ink and roll-to-roll process, and have recently reported a functional touch-sensitive panel with apparent resistance to oxidation. Cu, with its high inherent conductivity, may be another suitable candidate -- as discussed earlier, Cu-based TADF molecules have already been solution processed. Widespread attention is given to carbon nanotube- or graphene- based cathodes, which have given promising electrical performances and are easily adapted to wet-deposition [89,96,97]. Polymer cathodes may be a viable alternative as they have been studied for use in batteries [98-103] and, to a lesser extent, nanoscale batteries [104] and solar cells [105]. Designers should also consider the solution processability of organic transport layers for fully SP OLEDs. However, transport layers have fewer design criteria than EMLs, and will not experience the same number of detrimental synergistic effects when depositing via solution as discussed earlier. Thus, there are fewer considerations when incorporating SP transport layers. Solvent orthogonality is perhaps the most significant consideration. Inkjet Deposition The majority of SP-OLED studies use spin coating as the chosen method for deposition. While spin coating is indeed a wet process and allows one to analyse film structure, it does not necessarily translate into large-scale manufacture. Not only is the physical process of spin coating not applicable to the large scale, it cannot be incorporated into R2R processes. On the other hand, inkjet deposition 13 is compatible and thus more suitable for commercial manufacture. In fact, electronic developers have recently begun using inkjet deposition in production lines or unveiled inkjet-deposited OLED prototypes. Thus, by using inkjet deposition in initial studies, researchers can ensure any developed materials are suitable for manufacture immediately and skip additional steps involved in translating their research to the market. Single-layer and non-doped devices More efforts should be paid towards reducing production costs through single-layer devices and non-doped (i.e. no host) EMLs. Not only will this minimise production costs through reducing materials consumption, it will lessen the number of factors needing to be considered during design stages, thus simplifying the overall design procedure. As an example, doped EMLs require additional consideration of the entire electronic structure of the dopant along with that of the host. Similarly, in a fully-solution processed device, each layer needs a high level of orthogonal resistance to certain solvents to avoid dissolution when depositing neighbouring layers; single-layer devices will not require such consideration. In recent years, non-doped EMLs have made substantial progress in the literature, boasting greater stability and somewhat similar performance efficiency when compared to their doped counterparts [106]. On the other hand, single-layer OLEDs are considerably less common in research; the few notable recent studies of single-layer OLEDs in literature show [107,108]). Light emitting significantly electrochemical cells, which operate by a slightly different mechanism, are a perfect example of solid-state lighting through minimalistic single-layer design [109]. The Zysman-Colman group [110- 113] have produced several light-emitting electrochemical cells using TADF EMLs, albeit at reduced efficiencies. lower performance efficiencies (for example, see Conflicts of interest The author declares no conflicts of interest. 14 References 1. Uoyama, H.; Goushi, K.; Shizu, K.; Nomura, H.; Adachi, C. Highly efficient organic light-emitting diodes from delayed fluorescence. Nature 2012, 492, 234, doi:10.1038/nature11687 https://www.nature.com/articles/nature11687#supplementary-information. 2. Triplet 2017, 29, Quenching. Advanced Materials Parker, C.A.; Hatchard, C.G. Triplet-singlet emission in fluid solutions. Phosphorescence of eosin. Transactions of the Faraday Society 1961, 57, 1894-1904, doi:10.1039/TF9615701894. Kaji, H.; Suzuki, H.; Fukushima, T.; Shizu, K.; Suzuki, K.; Kubo, S.; Komino, T.; Oiwa, H.; Suzuki, F.; Wakamiya, A.J.N.c. Purely organic electroluminescent material realizing 100% conversion from electricity to light. 2015, 6, 8476. Zhang, Q.; Tsang, D.; Kuwabara, H.; Hatae, Y.; Li, B.; Takahashi, T.; Lee, S.Y.; Yasuda, T.; Adachi, C.J.A.M. Nearly 100% internal quantum efficiency in undoped electroluminescent devices employing pure organic emitters. 2015, 27, 2096-2100. Lin, T.A.; Chatterjee, T.; Tsai, W.L.; Lee, W.K.; Wu, M.J.; Jiao, M.; Pan, K.C.; Yi, C.L.; Chung, C.L.; Wong, K.T.J.A.M. Sky-blue organic light emitting diode with 37% external quantum efficiency using thermally activated delayed fluorescence from spiroacridine-triazine hybrid. 2016, 28, 6976-6983. Moon, C.-K.; Suzuki, K.; Shizu, K.; Adachi, C.; Kaji, H.; Kim, J.-J. Combined Inter- and Intramolecular Charge-Transfer Processes for Highly Efficient Fluorescent Organic Light-Emitting Diodes with Exciton Reduced 1606448, doi:doi:10.1002/adma.201606448. Yersin, H. Highly Efficient OLEDs: Materials Based on Thermally Activated Delayed Fluorescence; Wiley: 2018. Koden, M. OLED Displays and Lighting; Wiley: 2016. Volz, D. Review of organic light-emitting diodes with thermally activated delayed fluorescence emitters for energy-efficient sustainable light sources and displays; SPIE: 2016; Vol. 6, pp. 13. Im, Y.; Kim, M.; Cho, Y.J.; Seo, J.-A.; Yook, K.S.; Lee, J.Y. Molecular Design Strategy of Organic Thermally Activated Delayed Fluorescence Emitters. Chemistry of Materials 2017, 29, 1946-1963, doi:10.1021/acs.chemmater.6b05324. Santos, P.L.; Ward, J.S.; Data, P.; Batsanov, A.S.; Bryce, M.R.; Dias, F.B.; Monkman, A.P.J.J.o.M.C.C. Engineering the singlet -- triplet energy splitting in a TADF molecule. 2016, 4, 3815-3824. Bergmann, L.; Zink, D.M.; Bräse, S.; Baumann, T.; Volz, D. Metal -- Organic and Organic TADF- Materials: Status, Challenges and Characterization. In Photoluminescent Materials and Electroluminescent Devices, Springer: 2017; pp. 201-239. Yersin, H.; Czerwieniec, R.; Shafikov, M.Z.; Suleymanova, A.F.J.C. TADF Material Design: Photophysical Background and Case Studies Focusing on CuI and AgI Complexes. 2017, 18, 3508-3535. Dias, F.B.; Penfold, T.J.; Monkman, A.P. Photophysics of thermally activated delayed fluorescence molecules. Methods and applications in fluorescence 2017, 5, 012001. Duan, L.; Hou, L.; Lee, T.-W.; Qiao, J.; Zhang, D.; Dong, G.; Wang, L.; Qiu, Y. Solution processable small molecules for organic light-emitting diodes. Journal of Materials Chemistry 2010, 20, 6392-6407. Cho, Y.J.; Yook, K.S.; Lee, J.Y. High Efficiency in a Solution-Processed Thermally Activated Delayed- Fluorescence Device Using a Delayed-Fluorescence Emitting Material with Improved Solubility. Advanced Materials 2014, 26, 6642-6646, doi:doi:10.1002/adma.201402188. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 15 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. external quantum 53, efficiency. Chemical Communications Nie, H.; Hu, K.; Cai, Y.; Peng, Q.; Zhao, Z.; Hu, R.; Chen, J.; Su, S.-J.; Qin, A.; Tang, B.Z. Tetraphenylfuran: aggregation-induced emission or aggregation-caused quenching? Materials Chemistry Frontiers 2017, 1, 1125-1129, doi:10.1039/C6QM00343E. Jenekhe, S.A.; Osaheni, J.A. Excimers and Exciplexes of Conjugated Polymers. Science 1994, 265, 765- 768, doi:10.1126/science.265.5173.765. Huang, T.; Jiang, W.; Duan, L. Recent progress in solution processable TADF materials for organic light- emitting diodes. Journal of Materials Chemistry C 2018, 6, 5577-5596, doi:10.1039/C8TC01139G. Zou, Y.; Gong, S.; Xie, G.; Yang, C. Design Strategy for Solution-Processable Thermally Activated Delayed Fluorescence Emitters and Their Applications in Organic Light-Emitting Diodes. Advanced Optical Materials 2018, 6, 1800568, doi:doi:10.1002/adom.201800568. Lee, S.Y.; Yasuda, T.; Komiyama, H.; Lee, J.; Adachi, C. Thermally Activated Delayed Fluorescence Polymers for Efficient Solution-Processed Organic Light-Emitting Diodes. Advanced Materials 2016, 28, 4019-4024, doi:doi:10.1002/adma.201505026. Albrecht, K.; Matsuoka, K.; Fujita, K.; Yamamoto, K. Carbazole Dendrimers as Solution-Processable Thermally Activated Delayed-Fluorescence Materials. Angewandte Chemie International Edition 2015, 54, 5677-5682, doi:doi:10.1002/anie.201500203. Albrecht, K.; Matsuoka, K.; Yokoyama, D.; Sakai, Y.; Nakayama, A.; Fujita, K.; Yamamoto, K. Thermally activated delayed fluorescence OLEDs with fully solution processed organic layers exhibiting nearly 10% 2439-2442, doi:10.1039/C6CC09275F. Albrecht, K.; Matsuoka, K.; Fujita, K.; Yamamoto, K. A dendrimer emitter doped in a dendrimer host: efficient thermally activated delayed fluorescence OLEDs with fully-solution processed organic-layers. Materials Chemistry Frontiers 2018, 2, 1097-1103. Xie, G.; Li, X.; Chen, D.; Wang, Z.; Cai, X.; Chen, D.; Li, Y.; Liu, K.; Cao, Y.; Su, S.-J. Evaporation- and Solution-Process-Feasible Highly Efficient Thianthrene-9,9′,10,10′-Tetraoxide-Based Thermally Activated Delayed Fluorescence Emitters with Reduced Efficiency Roll-Off. Advanced Materials 2016, 28, 181-187, doi:doi:10.1002/adma.201503225. Liu, Y.; Xie, G.; Wu, K.; Luo, Z.; Zhou, T.; Zeng, X.; Yu, J.; Gong, S.; Yang, C. Boosting reverse intersystem crossing by increasing donors in triarylboron/phenoxazine hybrids: TADF emitters for high-performance solution-processed OLEDs. Journal of Materials Chemistry C 2016, 4, 4402-4407. Tang, C.; Yang, T.; Cao, X.; Tao, Y.; Wang, F.; Zhong, C.; Qian, Y.; Zhang, X.; Huang, W. Tuning a Weak Emissive Blue Host to Highly Efficient Green Dopant by a CN in Tetracarbazolepyridines for Solution- Processed Thermally Activated Delayed Fluorescence Devices. Advanced Optical Materials 2015, 3, 786- 790, doi:doi:10.1002/adom.201500016. Chen, X.-L.; Lin, C.-S.; Wu, X.-Y.; Yu, R.; Teng, T.; Zhang, Q.-K.; Zhang, Q.; Yang, W.-B.; Lu, C.-Z. Highly efficient cuprous complexes with thermally activated delayed fluorescence and simplified solution process OLEDs using the ligand as host. Journal of Materials Chemistry C 2015, 3, 1187-1195, doi:10.1039/C4TC02255F. Joo, C.W.; Cho, H.; Kwon, B.-H.; Cho, N.S.; Kim, Y.; Kim, Y.-H.; Lee, J. Development of solution- processable blue/hybrid-white OLEDs based on thermally activated delayed fluorescence. Journal of Industrial and Engineering Chemistry 2018, 65, 35-39, doi:https://doi.org/10.1016/j.jiec.2018.03.030. 2017, 16 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. thermally activated delayed to enhance exciton utilization efficiency Li, W.; Zhao, J.; Li, L.; Du, X.; Fan, C.; Zheng, C.; Tao, S. Efficient solution-processed blue and white OLEDs based on a high-triplet bipolar host and a blue TADF emitter. Organic Electronics 2018, 58, 276- 282, doi:https://doi.org/10.1016/j.orgel.2018.04.027. Luo, J.; Gong, S.; Gu, Y.; Chen, T.; Li, Y.; Zhong, C.; Xie, G.; Yang, C. Multi-carbazole encapsulation as a simple strategy for the construction of solution-processed, non-doped thermally activated delayed fluorescence emitters. Journal of Materials Chemistry C 2016, 4, 2442-2446, doi:10.1039/C6TC00418K. Komatsu, R.; Sasabe, H.; Inomata, S.; Pu, Y.-J.; Kido, J. High efficiency solution processed OLEDs using fluorescence emitter. Synthetic Metals 2015, 202, 165-168, a doi:https://doi.org/10.1016/j.synthmet.2015.02.009. Luo, J.; Xie, G.; Gong, S.; Chen, T.; Yang, C. Creating a thermally activated delayed fluorescence channel in a single polymer system for bluish-green electroluminescence. Chemical Communications 2016, 52, 2292-2295, doi:10.1039/C5CC09797E. Cho, Y.J.; Jeon, S.K.; Lee, J.Y. Molecular Engineering of High Efficiency and Long Lifetime Blue Thermally Activated Delayed Fluorescent Emitters for Vacuum and Solution Processed Organic Light- Emitting Diodes. Advanced Optical Materials 2016, 4, 688-693, doi:doi:10.1002/adom.201500634. Kim, Y.-H.; Wolf, C.; Cho, H.; Jeong, S.-H.; Lee, T.-W. Highly Efficient, Simplified, Solution-Processed Thermally Activated Delayed-Fluorescence Organic Light-Emitting Diodes. Advanced Materials 2016, 28, 734-741, doi:doi:10.1002/adma.201504490. Nobuyasu, R.S.; Ren, Z.; Griffiths, G.C.; Batsanov, A.S.; Data, P.; Yan, S.; Monkman, A.P.; Bryce, M.R.; Dias, F.B. Rational Design of TADF Polymers Using a Donor -- Acceptor Monomer with Enhanced TADF Efficiency Induced by the Energy Alignment of Charge Transfer and Local Triplet Excited States. Advanced Optical Materials 2016, 4, 597-607, doi:doi:10.1002/adom.201500689. Sun, K.; Sun, Y.; Huang, T.; Luo, J.; Jiang, W.; Sun, Y. Design strategy of yellow thermally activated delayed fluorescent dendrimers and their highly efficient non-doped solution-processed OLEDs with low driving voltage. Organic Electronics 2017, 42, 123-130, doi:https://doi.org/10.1016/j.orgel.2016.12.026. Ban, X.; Jiang, W.; Sun, K.; Lin, B.; Sun, Y. Self-Host Blue Dendrimer Comprised of Thermally Activated Delayed Fluorescence Core and Bipolar Dendrons for Efficient Solution-Processable Nondoped Electroluminescence. ACS Applied Materials & Interfaces 2017, 9, 7339-7346, doi:10.1021/acsami.6b14922. Hu, J.; Zhang, X.; Zhang, D.; Cao, X.; Jiang, T.; Zhang, X.; Tao, Y. Linkage modes on phthaloyl/triphenylamine hybrid compounds: Multi-functional AIE luminogens, non-doped emitters and organic hosts for highly efficient solution-processed delayed fluorescence OLEDs. Dyes and Pigments 2017, 137, 480-489, doi:https://doi.org/10.1016/j.dyepig.2016.10.029. Ban, X.; Zhu, A.; Zhang, T.; Tong, Z.; Jiang, W.; Sun, Y. Design of encapsulated hosts and guests for highly efficient blue and green thermally activated delayed fluorescence OLEDs based on a solution- process. Chemical Communications 2017, 53, 11834-11837. Gong, S.; Luo, J.; Wang, Z.; Li, Y.; Chen, T.; Xie, G.; Yang, C. Tuning emissive characteristics and singlet- triplet energy splitting of fluorescent emitters by encapsulation group modification: Yellow TADF emitter for solution-processed OLEDs with high luminance and ultraslow efficiency roll-off. Dyes and Pigments 2017, 139, 593-600, doi:https://doi.org/10.1016/j.dyepig.2016.12.058. Bin, H.; Ji, Y.; Li, Z.; Zhou, N.; Jiang, W.; Feng, Y.; Lin, B.; Sun, Y. Simple aggregation -- induced delayed fluorescence materials based on anthraquinone derivatives for highly efficient solution -- processed red OLEDs. Journal of Luminescence 2017, 187, 414-420, doi:https://doi.org/10.1016/j.jlumin.2017.03.038. 17 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. Dyes and Pigments 133, 2016, green OLED. Huang, B.; Li, Z.; Yang, H.; Hu, D.; Wu, W.; Feng, Y.; Sun, Y.; Lin, B.; Jiang, W. Bicolour electroluminescence of 2-(carbazol-9-yl)anthraquinone based on a solution process. Journal of Materials Chemistry C 2017, 5, 12031-12034, doi:10.1039/C7TC04444E. Huang, B.; Ban, X.; Sun, K.; Ma, Z.; Mei, Y.; Jiang, W.; Lin, B.; Sun, Y. Thermally activated delayed fluorescence materials based on benzophenone derivative as emitter for efficient solution-processed non-doped 380-386, doi:https://doi.org/10.1016/j.dyepig.2016.06.025. Lin, X.; Zhu, Y.; Zhang, B.; Zhao, X.; Yao, B.; Cheng, Y.; Li, Z.; Qu, Y.; Xie, Z. Highly Efficient TADF Polymer Electroluminescence with Reduced Efficiency Roll-off via Interfacial Exciplex Host Strategy. ACS Applied Materials & Interfaces 2018, 10, 47-52, doi:10.1021/acsami.7b16887. Zhu, Y.; Zhang, Y.; Yao, B.; Wang, Y.; Zhang, Z.; Zhan, H.; Zhang, B.; Xie, Z.; Wang, Y.; Cheng, Y. Synthesis and Electroluminescence of a Conjugated Polymer with Thermally Activated Delayed Fluorescence. Macromolecules 2016, 49, 4373-4377, doi:10.1021/acs.macromol.6b00430. Cho, Y.J.; Chin, B.D.; Jeon, S.K.; Lee, J.Y. 20% External Quantum Efficiency in Solution-Processed Blue Thermally Activated Delayed Fluorescent Devices. Advanced Functional Materials 2015, 25, 6786-6792, doi:doi:10.1002/adfm.201502995. Li, Y.; Xie, G.; Gong, S.; Wu, K.; Yang, C. Dendronized delayed fluorescence emitters for non-doped, solution-processed organic light-emitting diodes with high efficiency and low efficiency roll-off channels. Chemical Science 2016, 7, 5441-5447, simultaneously: doi:10.1039/C6SC00943C. Yu, L.; Wu, Z.; Xie, G.; Zhong, C.; Zhu, Z.; Cong, H.; Ma, D.; Yang, C. Achieving a balance between small singlet -- triplet energy splitting and high fluorescence radiative rate in a quinoxaline-based orange-red thermally activated delayed fluorescence emitter. Chemical Communications 2016, 52, 11012- 11015, doi:10.1039/C6CC05203G. Ban, X.; Jiang, W.; Lu, T.; Jing, X.; Tang, Q.; Huang, S.; Sun, K.; Huang, B.; Lin, B.; Sun, Y. Self-host thermally activated delayed fluorescent dendrimers with flexible chains: an effective strategy for non- doped electroluminescent devices based on solution processing. Journal of Materials Chemistry C 2016, 4, 8810-8816, doi:10.1039/C6TC03063G. Li, J.; Liao, X.; Xu, H.; Li, L.; Zhang, J.; Wang, H.; Xu, B. Deep-blue thermally activated delayed fluorescence dendrimers with reduced singlet-triplet energy gap for low roll-off non-doped solution- processed 79-86, doi:https://doi.org/10.1016/j.dyepig.2017.01.036. Matsuoka, K.; Albrecht, K.; Yamamoto, K.; Fujita, K. Mulifunctional Dendritic Emitter: Aggregation- Induced Emission Enhanced, Thermally Activated Delayed Fluorescent Material for Solution- Processed Multilayered Organic Light-Emitting Diodes. Scientific Reports 2017, 7, 41780, doi:10.1038/srep41780 two parallel emissive diodes. Dyes organic light-emitting and Pigments 2017, 140, https://www.nature.com/articles/srep41780#supplementary-information. 53. Sun, K.; Xie, X.; Liu, Y.; Jiang, W.; Ban, X.; Huang, B.; Sun, Y. Thermally cross-linkable thermally activated delayed fluorescent materials for efficient blue solution-processed organic light-emitting diodes. Journal of Materials Chemistry C 2016, 4, 8973-8979, doi:10.1039/C6TC02634F. Verma, A.; Zink, D.M.; Fléchon, C.; Leganés Carballo, J.; Flügge, H.; Navarro, J.M.; Baumann, T.; Volz, D. Efficient, inkjet-printed TADF-OLEDs with an ultra-soluble NHetPHOS complex. Applied Physics A 2016, 122, 191, doi:10.1007/s00339-016-9726-2. 54. 18 55. 56. 57. 58. 59. 60. 61. 62. 2017, 9, Interfaces Emitter. ACS Applied Materials & Chen, X.-L.; Yu, R.; Zhang, Q.-K.; Zhou, L.-J.; Wu, X.-Y.; Zhang, Q.; Lu, C.-Z. Rational Design of Strongly Blue-Emitting Cuprous Complexes with Thermally Activated Delayed Fluorescence and Application in Solution-Processed OLEDs. Chemistry of Materials 2013, 25, 3910-3920, doi:10.1021/cm4024309. Chen, X.-L.; Jia, J.-H.; Yu, R.; Liao, J.-Z.; Yang, M.-X.; Lu, C.-Z. Combining Charge-Transfer Pathways to Achieve Unique Thermally Activated Delayed Fluorescence Emitters for High-Performance Solution-Processed, Non-doped Blue OLEDs. 2017, 56, 15006-15009, doi:doi:10.1002/anie.201709125. Ban, X.; Zhu, A.; Zhang, T.; Tong, Z.; Jiang, W.; Sun, Y. Highly Efficient All-Solution-Processed Fluorescent Organic Light-Emitting Diodes Based on a Novel Self-Host Thermally Activated Delayed Fluorescence 21900-21908, doi:10.1021/acsami.7b04146. Nikolaenko, A.E.; Cass, M.; Bourcet, F.; Mohamad, D.; Roberts, M. Thermally Activated Delayed Fluorescence in Polymers: A New Route toward Highly Efficient Solution Processable OLEDs. Advanced Materials 2015, 27, 7236-7240, doi:doi:10.1002/adma.201501090. Ren, Z.; Nobuyasu, R.S.; Dias, F.B.; Monkman, A.P.; Yan, S.; Bryce, M.R. Pendant Homopolymer and Copolymers as Solution-Processable Thermally Activated Delayed Fluorescence Materials for Organic Light-Emitting Diodes. Macromolecules 2016, 49, 5452-5460, doi:10.1021/acs.macromol.6b01216. Xie, G.; Luo, J.; Huang, M.; Chen, T.; Wu, K.; Gong, S.; Yang, C. Inheriting the Characteristics of TADF Small Molecule by Side-Chain Engineering Strategy to Enable Bluish-Green Polymers with High PLQYs up to 74% and External Quantum Efficiency over 16% in Light-Emitting Diodes. Advanced Materials 2017, 29, 1604223, doi:doi:10.1002/adma.201604223. Dias, F.B.; Bourdakos, K.N.; Jankus, V.; Moss, K.C.; Kamtekar, K.T.; Bhalla, V.; Santos, J.; Bryce, M.R.; Monkman, A.P. Triplet Harvesting with 100% Efficiency by Way of Thermally Activated Delayed in Charge Transfer OLED Emitters. Advanced Materials 2013, 25, 3707-3714, Fluorescence doi:doi:10.1002/adma.201300753. Kim, K.-H.; Lee, S.; Moon, C.-K.; Kim, S.-Y.; Park, Y.-S.; Lee, J.-H.; Woo Lee, J.; Huh, J.; You, Y.; Kim, J.- J. Phosphorescent dye-based supramolecules for high-efficiency organic light-emitting diodes. 2014, 5, 4769, doi:10.1038/ncomms5769 https://www.nature.com/articles/ncomms5769#supplementary-information. 63. Flämmich, M.; Frischeisen, J.; Setz, D.S.; Michaelis, D.; Krummacher, B.C.; Schmidt, T.D.; Brütting, W.; Danz, N. Oriented phosphorescent emitters boost OLED efficiency. Organic Electronics 2011, 12, 1663- 1668, doi:https://doi.org/10.1016/j.orgel.2011.06.011. Facchetti, A. π-Conjugated polymers for organic electronics and photovoltaic cell applications. Chemistry of Materials 2010, 23, 733-758. Salaneck, W.R.; Stafstrom, S.; Brédas, J.L. Conjugated polymer surfaces and interfaces: electronic and chemical structure of interfaces for polymer light emitting devices; Cambridge University Press: 2003. Lee, J.; Aizawa, N.; Numata, M.; Adachi, C.; Yasuda, T. Versatile molecular functionalization for inhibiting concentration quenching of thermally activated delayed fluorescence. Advanced Materials 2017, 29, 1604856. Xie, Z.; Chen, C.; Xu, S.; Li, J.; Zhang, Y.; Liu, S.; Xu, J.; Chi, Z. White-Light Emission Strategy of a Single Organic Compound with Aggregation-Induced Emission and Delayed Fluorescence Properties. Angewandte Chemie International Edition 2015, 54, 7181-7184. Song, W.; Lee, I.; Lee, J.Y. Host Engineering for High Quantum Efficiency Blue and White Fluorescent Organic Light-Emitting Diodes. Advanced Materials 2015, 27, 4358-4363. 19 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. Cho, Y.J.; Yook, K.S.; Lee, J.Y. High Efficiency in a Solution-Processed Thermally Activated Delayed- Fluorescence Device Using a Delayed-Fluorescence Emitting Material with Improved Solubility. Advanced Materials 2014, 26, 6642-6646. Suzuki, Y.; Zhang, Q.; Adachi, C. A solution-processable host material of 1, 3-bis {3-[3-(9-carbazolyl) phenyl]-9-carbazolyl} benzene and its application in organic light-emitting diodes employing thermally activated delayed fluorescence. Journal of Materials Chemistry C 2015, 3, 1700-1706. Kim, J.-S.; Friend, R.H.; Grizzi, I.; Burroughes, J.H. Spin-cast thin semiconducting polymer interlayer for improving device efficiency of polymer light-emitting diodes. Applied Physics Letters 2005, 87, 023506. Tsujimoto, H.; Ha, D.-G.; Markopoulos, G.; Chae, H.S.; Baldo, M.A.; Swager, T.M. Thermally Activated Delayed Fluorescence and Aggregation Induced Emission with Through-Space Charge Transfer. Journal of the American Chemical Society 2017, 139, 4894-4900, doi:10.1021/jacs.7b00873. Feuillastre, S.; Pauton, M.; Gao, L.; Desmarchelier, A.; Riives, A.J.; Prim, D.; Tondelier, D.; Geffroy, B.; Muller, G.; Clavier, G., et al. Design and Synthesis of New Circularly Polarized Thermally Activated Delayed Fluorescence Emitters. J Am Chem Soc 2016, 138, 3990-3993, doi:10.1021/jacs.6b00850. Mei, J.; Leung, N.L.C.; Kwok, R.T.K.; Lam, J.W.Y.; Tang, B.Z. Aggregation-Induced Emission: Together We Shine, United We Soar! Chemical Reviews 2015, 115, 11718-11940, doi:10.1021/acs.chemrev.5b00263. Li, J.; Liu, D. Dendrimers for organic light-emitting diodes. Journal of materials Chemistry 2009, 19, 7584- 7591. Lo, S.-C.; Burn, P.L. Development of dendrimers: macromolecules for use in organic light-emitting diodes and solar cells. Chemical Reviews 2007, 107, 1097-1116. Albrecht, K.; Pernites, R.; Felipe, M.J.; Advincula, R.C.; Yamamoto, K. Patterning Carbazole -- Phenylazomethine Dendrimer Films. Macromolecules 2012, 45, 1288-1295, doi:10.1021/ma202485h. Kimoto, A.; Cho, J.-S.; Ito, K.; Aoki, D.; Miyake, T.; Yamamoto, K. Novel Hole-Transport Material for Efficient Polymer Light-Emitting Diodes by Photoreaction. Macromolecular Rapid Communications 2005, 26, 597-601, doi:doi:10.1002/marc.200400657. Aizawa, N.; Pu, Y.-J.; Watanabe, M.; Chiba, T.; Ideta, K.; Toyota, N.; Igarashi, M.; Suzuri, Y.; Sasabe, H.; Kido, J. Solution-processed multilayer small-molecule light-emitting devices with high-efficiency white-light emission. Nature Communications 2014, 5, 5756, doi:10.1038/ncomms6756 https://www.nature.com/articles/ncomms6756#supplementary-information. 80. Chen, L.; Ma, Z.; Ding, J.; Wang, L.; Jing, X.; Wang, F. Self-host heteroleptic green iridium dendrimers: achieving efficient non-doped device performance based on a simple molecular structure. Chemical Communications 2011, 47, 9519-9521, doi:10.1039/C1CC13276H. Leijtens, T.; Ding, I.-K.; Giovenzana, T.; Bloking, J.T.; McGehee, M.D.; Sellinger, A. Hole transport materials with low glass transition temperatures and high solubility for application in solid-state dye- sensitized solar cells. ACS nano 2012, 6, 1455-1462. Hansen, C.M. The three dimensional solubility parameter. Danish Technical: Copenhagen 1967, 14. Barton, A.F. Solubility parameters. Chemical Reviews 1975, 75, 731-753. Machui, F.; Abbott, S.; Waller, D.; Koppe, M.; Brabec, C.J. Determination of solubility parameters for organic semiconductor formulations. Macromolecular chemistry and Physics 2011, 212, 2159-2165. Burgués-Ceballos, I.; Machui, F.; Min, J.; Ameri, T.; Voigt, M.M.; Luponosov, Y.N.; Ponomarenko, S.A.; Lacharmoise, P.D.; Campoy-Quiles, M.; Brabec, C.J. Solubility based identification of green solvents for small molecule organic solar cells. Advanced Functional Materials 2014, 24, 1449-1457. 81. 82. 83. 84. 85. 20 86. 87. 88. 89. 90. 91. 92. Machui, F.; Langner, S.; Zhu, X.; Abbott, S.; Brabec, C.J. Determination of the P3HT:PCBM solubility parameters via a binary solvent gradient method: Impact of solubility on the photovoltaic performance. Solar Energy Materials and Solar Cells 2012, 100, 138-146, doi:https://doi.org/10.1016/j.solmat.2012.01.005. Duong, D.T.; Walker, B.; Lin, J.; Kim, C.; Love, J.; Purushothaman, B.; Anthony, J.E.; Nguyen, T.Q. Molecular solubility and hansen solubility parameters for the analysis of phase separation in bulk heterojunctions. Journal of Polymer Science Part B: Polymer Physics 2012, 50, 1405-1413. Sanders, J.M.; Kwak, H.S.; Mustard, T.J.; Browning, A.R.; Halls, M.D. Atomistic simulations of mechanical and thermophysical properties of OLED materials; SPIE: 2018; Vol. 10736, pp. 8. Kamyshny, A.; Magdassi, S. Conductive Nanomaterials for Printed Electronics. Small 2014, 10, 3515- 3535, doi:doi:10.1002/smll.201303000. Ankireddy, K.; Vunnam, S.; Kellar, J.; Cross, W. Highly conductive short chain carboxylic acid encapsulated silver nanoparticle based inks for direct write technology applications. Journal of Materials Chemistry C 2013, 1, 572-579, doi:10.1039/C2TC00336H. Vaseem, M.; Lee, K.M.; Hong, A.R.; Hahn, Y.-B. Inkjet Printed Fractal-Connected Electrodes with Silver Nanoparticle Ink. ACS Applied Materials & Interfaces 2012, 4, 3300-3307, doi:10.1021/am300689d. Yamada, T.; Fukuhara, K.; Matsuoka, K.; Minemawari, H.; Tsutsumi, J.y.; Fukuda, N.; Aoshima, K.; Arai, S.; Makita, Y.; Kubo, H., et al. Nanoparticle chemisorption printing technique for conductive silver patterning with submicron resolution. Nature Communications 2016, 7, 11402, doi:10.1038/ncomms11402 https://www.nature.com/articles/ncomms11402#supplementary-information. 93. 94. 95. 96. 97. 98. 99. 100. 101. Wu, Y.; Li, Y.; Ong, B.S.; Liu, P.; Gardner, S.; Chiang, B. High-Performance Organic Thin-Film Transistors with Solution-Printed Gold Contacts. Advanced Materials 2005, 17, 184-187, doi:doi:10.1002/adma.200400690. Jung, S.-h.; Choi, D.Y.; Lee, H.M. Roll-to-roll processed, highly conductive, and flexible aluminum (Al) electrodes based on Al precursor inks. RSC Advances 2018, 8, 19950-19957, doi:10.1039/C8RA03303J. Lee, H.M.; Seo, J.Y.; Jung, A.; Choi, S.-Y.; Ko, S.H.; Jo, J.; Park, S.B.; Park, D. Long-Term Sustainable Aluminum Precursor Solution for Highly Conductive Thin Films on Rigid and Flexible Substrates. ACS Applied Materials & Interfaces 2014, 6, 15480-15487, doi:10.1021/am504134f. Chang, J.-H.; Lin, W.-H.; Wang, P.-C.; Taur, J.-I.; Ku, T.-A.; Chen, W.-T.; Yan, S.-J.; Wu, C.-I. Solution- processed transparent blue organic light-emitting diodes with graphene as the top cathode. Scientific reports 2015, 5, 9693. Matyba, P.; Yamaguchi, H.; Eda, G.; Chhowalla, M.; Edman, L.; Robinson, N.D. Graphene and mobile ions: the key to all-plastic, solution-processed light-emitting devices. Acs Nano 2010, 4, 637-642. Wang, J.; Liu, L.; Ling, Z.; Yang, J.; Wan, C.; Jiang, C. Polymer lithium cells with sulfur composites as cathode materials. Electrochimica Acta 2003, 48, 1861-1867, doi:https://doi.org/10.1016/S0013- 4686(03)00258-5. Kim, J.-K.; Cheruvally, G.; Choi, J.-W.; Ahn, J.-H.; Lee, S.H.; Choi, D.S.; Song, C.E. Effect of radical polymer cathode thickness on the electrochemical performance of organic radical battery. Solid State Ionics 2007, 178, 1546-1551, doi:https://doi.org/10.1016/j.ssi.2007.09.009. Novák, P.; Müller, K.; Santhanam, K.S.V.; Haas, O. Electrochemically Active Polymers for Rechargeable Batteries. Chemical Reviews 1997, 97, 207-282, doi:10.1021/cr941181o. Song, Z.; Zhan, H.; Zhou, Y. Polyimides: Promising Energy-Storage Materials. Angewandte Chemie International Edition 2010, 49, 8444-8448, doi:doi:10.1002/anie.201002439. 21 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. Oyama, N.; Tatsuma, T.; Sato, T.; Sotomura, T. Dimercaptan -- polyaniline composite electrodes for lithium batteries with high energy density. Nature 1995, 373, 598-600, doi:10.1038/373598a0. Peterson, B.M.; Ren, D.; Shen, L.; Wu, Y.-C.M.; Ulgut, B.; Coates, G.W.; Abruña, H.D.; Fors, B.P. Phenothiazine-Based Polymer Cathode Materials with Ultrahigh Power Densities for Lithium Ion Batteries. ACS Applied Energy Materials 2018, 1, 3560-3564. Suga, T.; Konishi, H.; Nishide, H. Photocrosslinked nitroxide polymer cathode-active materials for application in an organic-based paper battery. Chemical Communications 2007, 10.1039/B618710B, 1730- 1732, doi:10.1039/B618710B. Gadisa, A.; Tvingstedt, K.; Admassie, S.; Lindell, L.; Crispin, X.; Andersson, M.R.; Salaneck, W.R.; Inganäs, O. Transparent polymer cathode for organic photovoltaic devices. Synthetic Metals 2006, 156, 1102-1107. Godumala, M.; Choi, S.; Cho, M.J.; Choi, D.H. Recent breakthroughs in thermally activated delayed fluorescence organic light emitting diodes containing non-doped emitting layers. Journal of Materials Chemistry C 2019, 10.1039/C8TC06293E, doi:10.1039/C8TC06293E. Angioni, E.; Chapran, M.; Ivaniuk, K.; Kostiv, N.; Cherpak, V.; Stakhira, P.; Lazauskas, A.; Tamulevičius, S.; Volyniuk, D.; Findlay, N. A single emitting layer white OLED based on exciplex interface emission. Journal of Materials Chemistry C 2016, 4, 3851-3856. Feng, X.J.; Chen, S.F.; Ni, Y.; Wong, M.S.; Lam, M.M.; Cheah, K.W.; Lai, G.Q. Fluorene derivatives for highly efficient non-doped single-layer blue organic light-emitting diodes. Organic Electronics 2014, 15, 57-64. Meier, S.B.; Tordera, D.; Pertegas, A.; Roldan-Carmona, C.; Ortí, E.; Bolink, H.J. Light-emitting electrochemical cells: recent progress and future prospects. Materials Today 2014, 17, 217-223. Pertegás, A.; Wong, M.Y.; Sessolo, M.; Zysman-Colman, E.; Bolink, H.J. Efficient Light-Emitting Electrochemical Cells Using Small Molecular Weight, Ionic, Host-Guest Systems. ECS Journal of Solid State Science and Technology 2016, 5, R3160-R3163, doi:10.1149/2.0201601jss. Wong, M.Y.; La-Placa, M.-G.; Pertegas, A.; Bolink, H.J.; Zysman-Colman, E. Deep-blue thermally activated delayed fluorescence (TADF) emitters for light-emitting electrochemical cells (LEECs). Journal of Materials Chemistry C 2017, 5, 1699-1705. Wong, M.Y.; Hedley, G.J.; Xie, G.; Kolln, L.S.; Samuel, I.D.; Pertegas, A.; Bolink, H.J.; Zysman-Colman, E. Light-emitting electrochemical cells and solution-processed organic light-emitting diodes using small molecule organic thermally activated delayed fluorescence emitters. Chemistry of Materials 2015, 27, 6535-6542. Wong, M.Y.; Zysman-Colman, E. Thermally Activated Delayed Fluorescence Emitters in Light- Emitting Electrochemical Cells. In Light-Emitting Electrochemical Cells: Concepts, Advances and Challenges, Costa, R.D., Ed. Springer International Publishing: Cham, 2017; 10.1007/978-3-319-58613-7_9pp. 237- 266. 22
2002.00706	1	2002	2019-12-09T06:48:37	Redirection of a crack driven by viscous fluid taking into account plastic effects in the process zone	[ "physics.app-ph", "cond-mat.soft", "physics.class-ph" ]	In this paper the problem of redirection of a crack driven by viscous fluid under mixed mode loading is considered. The loading includes the classical Modes I - III and the hydraulically induced tangential traction on the fracture walls. The effect of the plastic deformation of the near tip zone is accounted for. Different criteria to determine the fracture deflection angle are examined and compared.	physics.app-ph	physics	Redirection of a crack driven by viscous fluid taking into account plastic effects in the process zone M. Wrobel(1), A. Piccolroaz(2), P. Papanastasiou(1) and G. Mishuris(3) (1) Department of Civil and Environmental Engineering, University of Cyprus, 75 Kallipoleos Street, 1678 Nicosia, Cyprus (2)Dipartimento di Ingegneria Civile, Ambientale e Meccanica, Universita di Trento, via Mesiano, 77 I-38123 Trento, Italia (3)Department of Mathematics, Aberystwyth University, Ceredigion SY23 3BZ, Wales, UK February 4, 2020 Abstract In this paper the problem of redirection of a crack driven by viscous fluid under mixed mode loading is considered.The loading includes the classical Modes I - III and the hydrauli- cally induced tangential traction on the fracture walls. The effect of the plastic deformation of the near tip zone is accounted for. Different criteria to determine the fracture deflection angle are examined and compared. 1 Introduction Although the hydraulic fracturing (HF) technique was developed 70 years ago for enhancing the recovery of hydrocarbons from low permeability reservoirs, it became widely known this century as the fracking technology for the exploitation of shale gas and oil resources, with huge success in the United States. Another application of HF, relevant to this study, is in weak and unconsolidated rock formations for combined production stimulation and sand control. This method is known as 'frac-pack' wellbore completion. In all applications, HF involves pumping of a viscous fluid at high rates from a well into the rock formation under high pressure sufficient to fracture the reservoir. The initiated fracture propagates in a complex stress field near the wellbore and re-orients itself to extend further in the direction of the least resistance which is always perpendicular to the minimum in situ compressive stress. During the pumping process, a sand-like material called 'proppant' is mixed with the fracturing fluid. The proppant prevents the fracture from closing after the fluid injection is stopped. Hence, a permeable channel of high conductivity is formed for oil or gas to flow from the reservoir to the well [9]. The 'frac-pack' wellbore completion involves additionally, after fracturing the formation, a screen placement and gravel packing of the annulus between screen and rock face for sand control. Most of the models used by the petroleum industry to simulate and design hydraulic fractur- ing assume linear-elastic mechanism of the rock deformation and the classic fracture mechanics. 1 The near tip processes include viscous flow in the fracture and formation of a dry zone (fluid- lag), cohesive zone ahead of the crack tip and the surrounding area that is dominated by shear plastic deformation, labeled here a 'process zone'. In weakly consolidated formations growth of short hydraulic fractures is very much determined by the near-fracture-tip plastic deformation of rock coupled with the interaction between the fracturing fluid and fractured layer. Thus, the required propagation pressure and geometry of created fracture depend critically on the tip screening mechanism caused by the plastic rock deformation. An attempt to increase the reliability of numerical simulations of HF and explain the discrepancies between model pre- dictions and field measurements was made by [12, 13, 7] who investigated numerically with a fully coupled elastoplastic FEM HF models the effect of non-linear rock behaviour on the pressure needed for propagation and on the dimensions of the created fractures. Sarris and Pa- panastasiou [17, 18] and Wang [21] included the pressure diffusion and porous behavior of the rock in the framework of the cohesive zone model and plastic deformation. All these numerical studies found that plastic yielding provides a shielding mechanism near the tip resulting in an increase of the effective fracture toughness and the resistance to fracturing [14]. Higher pressure is needed to propagate an elastoplastic fracture that appears to be shorter and wider than an elastic fracture. These studies made also clear that the near tip mechanisms are coupled and interact with each other, thus eliminating the relative importance of one mechanism, or may result in an elevated effective fracture toughness at the macroscopic level. One of the mechanisms that so far has not been studied is the influence of shear stresses that are transferred by the viscous fluid on the crack surfaces near the tip. Wrobel et al. [24, 25] showed that, due to the order of the tip singularity of the hydraulic shear stress in elastic material, this component of the load cannot be omitted when computing the Energy Release Rate (ERR). They introduced a new parameter, called a hydraulic shear stress intensity factor (Kf ), and proved that it plays an important role in the HF process. In a recent study Papanastasiou and Durban [15] studied the near tip singular plastic fields in a Drucker-Prager power law material and found that the shear stress does not influence the level of singularity but it changes the shape of the developed plastic zones with the emergence of a boundary layer near the fracture surface. Thus, the shear loading and the emergence of a boundary plastic layer influences the 'roughness' of the fracture surface which is very important for the hydraulic conductivity of the fracture [6]. In all the studies mentioned above it was assumed that the fracture was initiated and prop- agated to the preferential direction, perpendicular to the minimum in situ stress. In practice, however, a hydraulic fracture initiates from perforations that are created with shape charges that penetrate the steel casing, cement and near wellbore rock, approximately 30 cm deep. The directions of perforations are not necessarily aligned with the fracture preferential direction as the stress field is unknown and the perforation gun is not usually oriented. Hence the fracture will initiate from perforations and reorient itself gradually as propagating to get aligned far away from the wellbore to the preferential direction. In this study we will examine the influence of plasticity in the re-orientation of the hydraulic fracture within an analytical solution framework, considering the role of viscous shear stresses. In a recent study [16] the authors considered the effect of the shear stress induced by the fluid on the crack surfaces to determine the direction of the HF crack propagation in elastic rocks and showed that it may play an essential role in the case of the mixed mode when the 2 total contribution of the classical Stress Intensity Factors (SIF) leads to ERR value close to its critical one. They used the two most popular criteria: the maximum circumferential stress (MCS) criterion [8] and the minimum strain energy density (MSED) criterion [19]. It was shown that hydraulic fracture has its own features when the shear traction induced by the fluid on the crack surface originates the local shear stress intensity factor Kf and this leads to significant difference in terms of the crack redirection if the propagation regime is close to viscosity dominated one. The analysis was based on the recent paper by Wrobel et al. [24] where the authors introduced a new parameter, Kf , called the shear stress intensity factor, which plays an important role in accuracy and efficiency of numerical computations. In particular, it was shown that the tip singularity of hydraulically induced shear traction on the crack surfaces is of order r−1/2 as r → 0 and thus has to be taken into account when ERR is computed. However, it is well known that inelastic rock deformation near the crack tip may play an important role in hydraulic fracturing [12, 13, 14, 15]. Thus, the analysys of a combined effect of the plastic deformation and the mixed mode loading (including the shear traction induced by the fluid) on the re-orientation of the fracture is important. There are two ways to tackle the problem of plastic deformation in the near-tip zone: a) to solve a complete elastic-plastic problem or b) to implement a simplified analysis that assumes that the plastic region is much smaller than the zone in which the elastic solution with a dominant singular term holds (i.e. the solution can be fully represented by this singular term). In the latter case the external boundary of the plastic zone is determined from a respective yield criterion. The first approach, although naturally much more adequate, is simultaneously extremely difficult to conduct from the technical point of view (see [1] and references therein). In [1] the authors considered a Mode I problem with the Drucker-Prager yield criterion. They mentioned that implementation of similar analysis for the mixed mode loading is rather impossible due to its complexity (to the best of our knowledge, no analytical solution has been delivered so far for the full elastic-plastic problem for a crack under mixed mode loading). Clearly, one can find many numerical research in the area [3, 10] but their application to parametric analysis is questionable. Thus, following e.g. [20, 26], we implement the simplified approach assuming that the plastic deformation zone is small enough in the aforementioned sense and it is enough to consider only the 'external' elastic problem, combined with pertinent yield criterion to find the boundary of the plastic zone. We start from the maximum dilatational strain energy density (MDSED) criterion intro- duced in [20, 26] to show that, in contrast to the classic fracture mechanics, this criterion in the case of hydraulic fracture does not always allow for predicting the direction of the crack prop- agation due to the problems with uniqueness and stability of solution. In particular, MDSED criterion fails to produce reasonable results in the proximity of the so-called viscosity domi- nated regime [16]. To eliminate this drawback, we propose a concept of a modified maximum circumferential stress criterion (MMCS) that accounts for local plastic effects by implementing, similarly to MDSED, the plastic zone described by the respective yield condition (von Mises, Drucker-Prager, Tresca or Mohr-Coulomb). It has been recognized that the applicability of the known criteria is not well justified when accounting for the impact of severe Mode III [11]. Recently, an attempt has been made to tackle such case [4, 5]. Note that the MMCS intro- duced in this paper, provides meaningful results even for a mixed mode with a severe Mode III component. It is clear that the new criterion should be validated by experiments. 3 The structure of the paper is as follows. In Section 2 we introduce necessary notations with reference to [16] where all details can be found. In Section 3 we consider the classic MDSED criterion. Then, in Section 4, we introduce a new criterion for the direction of crack propagation: the Modified Maximum Circumferential Stress (MMCS) criterion. Four variants of the criterion are analyzed, each of them utilizing different yield condition). In Subsection 5 all analysed criteria are compared and the impact of the shear stress on the crack propagation direction is discussed. Finally, the conclusions are summed up in Section 6. 2 Preliminary results As discussed in [16], the stresses around the crack tip in elastic region are distributed in the following manner: σ(r, θ, z) = 1√ 2πr [KI ΨI (θ) + KII ΨII (θ) + KIII ΨIII (θ) + Kf Ψτ (θ)] + O (log r) , (1) where {r, θ, z} is a local polar coordinate system traditionally associated with the crack tip, KI , KII and KIII are the classical stress intensity factors (SIFs) and Kf is the shear stress intensity factor related to the hydraulic shear traction acting on the crack surfaces. The functions Ψj(θ) define the polar angle dependence and are given by the formulae: (cid:20) (cid:21) Ψrr I (θ) = 1 4 5 cos θ 2 − cos 3θ 2 , Ψθθ 1 2 cos Ψrθ I (θ) = θ 2 II (θ) = −3 sin Ψθθ I (θ) = cos3 θ (cid:20) 2 − 3 sin (cid:20) 5 sin θ 2 + 3 cos (cid:21) (cid:21) , 3θ 2 3θ 2 , , sin θ, Ψrr II (θ) = − 1 4 θ 2 cos2 θ 2 , Ψrθ cos II (θ) = 1 4 τ (θ) = −Ψθθ τ (θ) = − cos θ 2 Ψrz III (θ) = sin For the plane strain Ψzz = ν(cid:0)Ψrr + Ψθθ(cid:1). III (θ) = cos , Ψθz θ 2 θ 2 , Ψrr 3θ 2 , Ψrθ τ (θ) = sin 3θ 2 . To define the critical fracture state, we use the Energy Release Rate (ERR) criterion as evaluated in [16] that accounts for the shear traction induced by the fluid on the crack surfaces: K2 I + K2 II + 4(1 − ν)KI Kf + (2) where ν is the Poisson's ratio, E is the Young's modulus, while EC and KIC are the critical values determining ERR and material toughness, respectively, and need to be found experimentally. To decrease the number of unknown parameters during the parametric study below, the 1 1 − ν III = EC ≡ K2 K2 IC, problem is normalised by introducing the following natural scaling: KI = KI KIC , KII = KII KIC , KIII = KIII KIC , Kf = Kf KIC . (3) 4 The fracture criterion (2) then becomes: K2 I + K2 II + 4(1 − ν) KI Kf + 1 1 − ν K2 III = 1. (4) Note that in such formulation the value of the material toughness is completely hidden and the large and small toughness regimes cannot be recognised or identified immediately. To make this possible, we introduce a dimensionless parameter p0 = 2πp0(1− ν2)/E, where p0 is a multiplier of the leading (logarithmic) term in the asymptotic expansion of the fluid pressure near the crack tip (see [24] or [23]). It can be used to produce the following interrelation between KI and Kf (compare with equation (71) from [24]): = p0 π(1 − ν) − p0 , Kf = KI , 0 < p0 < π(1 − ν). (5) The values of the parameter p0 and the stress intensity factors are not independent. As it was shown in [24] for the Mode I deformation (KII = KIII = 0), the value of the parameter p0 determines the propagation regime (p0 → 0 corresponds to toughness dominated one while p0 → π(1 − ν) determines the viscosity dominated regime). Following [16] we conclude that: KIC · Keff IC → ∞ ⇔ KI → 1 and p0 → 0, and KIC · K eff IC → 0 ⇔ KI → 0 and p0 → π(1 − ν), where the normalised effective toughness is defined as follows: Combining (4) and (5) provides, after rearrangement, a formula for KI when KII , KIII and p0 are known: KI = π(1 − ν) − p0 π(1 − ν) + p0(3 − 4ν) K eff IC . After substitution of (5) into (9) and some algebra we obtain a relation for Kf : (cid:113)(cid:2)π(1 − ν) − p0 p0 (cid:3)(cid:2)π(1 − ν) + p0(3 − 4ν)(cid:3) K eff IC . Kf = It can be easily seen that for any fixed values of KII and KIII one has (compare (9) and (10)): Finally, note that for KIII = 0 and p0 = 0 (classic mixed Mode I and II), both normalised stress intensity factors, KI = II and Kf = 0, are independent of ν. 1 − K2 (6) (7) (8) (9) (10) (11) K eff IC = 1 − K2 II − 1 1 − ν K2 III . (cid:114) (cid:115) (cid:16) K eff IC (cid:1)2, lim p0→π(1−ν) Kf KI = 1 4(1 − ν) (cid:113) 5 Equations (9) and (10) provide a relationship between the normalised symmetric SIFs, KI and Kf , and the normalised anti-symmetric SIFs, KII and KIII , taking into account also the influence of the hydraulically induced shear stresses through the pressure parameter p0. This allows for a parametric study of the fracture propagation angle, where the independent parameters are KII , KIII and p0. This analysis is given in the next sections. 3 Maximum Dilatational Strain Energy Density (MDSED) cri- terion Let us begin by analysing the maximum dilatational strain energy density (MDSED) criterion, proposed by Theocaris and Andrianopoulos in [20]. Noting that the total strain energy density, W , can be divided into two components at any given point -- the dilatational (volumetric) strain energy density, Wv, and the distortional strain energy density, Wd -- we have that: W = Wv + Wd, Wv = 1 18K (tr σ)2 = 1 − 2ν 6E [σrr + σθθ + σzz]2 , (12) (13) for: Wd = 1 4G dev σ · dev σ = (cid:0)σ2 1 + ν 3E θz)(cid:1) , (14) where σzz = ν (σrr + σθθ) for plane strain and K = E/[3(1 − 2ν)] is the volumetric modulus of elasticity (bulk modulus). zz − σrrσθθ − σθθσzz − σzzσrr + 3(σ2 rθ + σ2 rz + σ2 rr + σ2 θθ + σ2 The criterion assumes that the radius of the elastic-plastic boundary, rb, is determined from the von Mises yield condition to produce a surface of constant distortional strain energy density. Then a maximum of the dilatational strain energy density is searched along this surface. This maximum determines the angle of crack propagation. According to the von Mises yield condition, the critical value of the distortional strain energy density, Wd, obtained for the elastic-plastic boundary, is: Wd = 1 + ν 3E σ2 t , with σt being the uniaxial yield strength. The corresponding radius of the plastic deformation zone, rb, was derived as: where rb = Rbrb(θ), θ ∈ (−π, π), Rb = K2 IC σ2 t , has a dimension of length [m] and the dimensionless part 6 (15) (16) (17) (cid:110) K2 1 8π rb = I c1(θ) + K2 II c2(θ) + KI KII c3(θ) + K2 f c4(θ) + KI Kf c5(θ) + KII Kf c6(θ) + 12 K2 III (18) describes the shape of the zone and simultaneously affects its size by the values of the parameters (SIFs and the Poison's ratio ν). Here by cj(θ) we have defined the following dimensionless trigonometrical factors: (cid:111) , (5 − 8ν (1 − ν) − 3 cos θ) , c1(θ) = 2 cos2 θ 2 c2(θ) = 5 − 8ν (1 − ν) − 2 (1 − 2ν)2 cos θ + 9 cos2 θ, c4(θ) = 48 (1 − ν)2 , c5(θ) = 24(1 − ν) sin2 θ, c3(θ) = 4 sin θ c6(θ) = 12(1 − ν) sin 2θ, (cid:16) 3 cos θ − (1 − 2ν)2(cid:17) , (19) The relative sizes and shapes of the plastic zones are depicted in Fig. 1 for three values of KII = {0, 0.5, 0.9} and ν = {0, 0.3, 0.5} by means of a normalized radius: rb = rb π(1 − ν) − p0 π(1 − ν) . (20) This additional scaling has been introduced to make the curves that refer to different crack propagation regimes distinguishable in one picture. Indeed, under this scaling, the blue lines reflect the actual values of rb, the red ones utilize the scaling factor 0.5, while the black curves are multiplied by 0.1. Clearly, a symmetry of the plastic zone with respect to the plane of crack propagation (horizontal axis) is obtained only for the symmetrical loading ( KII = 0). When approaching the viscosity dominated regime the shape of the zone becomes elliptic regardless of the magnitude of KII and a value of the Poisson ratio. Conversely, as moving towards the toughness dominated mode, the plastic zone contour tends to be more irregular, which is especially pronounced for the severe antisymmetric loading ( KI (cid:28) KII ). When analyzing the size of plastic yield area we can see two different trends for the limiting regimes. Namely, for the viscosity dominated mode the plastic zone shrinks with growing KII , while for the toughness dominated regime a reverse mechanism is observed. Obviously, the character of the zone evolution in the interim is governed by those two counteracting tendencies. It is always the viscosity dominated regime which provides the greatest size of the plastic area, and the toughness dominated mode which results in the smallest extent of it. For all considered cases an increase of ν reduces the area of plastic zone. The criterion described in [20] states that the maximum value of the dilatational strain energy density, Wv(r, θ), has to be found along the elastic-plastic boundary. In other words, one needs to analyse an auxiliary function: F1(θ) = 8(1 + ν)2(1 − 2ν) Wv(rb(θ), θ), where the crack propagation direction, θf , is found from the condition: 3E (cid:12)(cid:12)(cid:12){F1=F max 1 7 θf = θ }∧{σθθ>0}. (21) (22) KII = 0 ν = 0 ν = 0.3 ν = 0.5 KII = 0.5 KII = 0.9 Figure 1: MDSED: The shapes of of the plastic zones described by the normalized radius rb (20) for various values of p0 and fixed KII and ν. The blue lines reflect the actual sizes of the plastic zones, the red ones utilize the scaling factor 0.5, while the black curves are multiplied by 0.1. As it was in the case of the Minimum Strain Energy Density criterion (see [16]), to preserve the physical sense of solution, the maximum should be sought for under additional condition that the circumferential stress is positive. According to [16] the expression for σθθ has the following form: (cid:20) σθθ(r, θ) = KIc√ 2πr KI cos3 θ 2 − 3 KII sin θ 2 cos2 θ 2 + 2(1 − ν) Kf cos 3θ 2 In Fig. 2 values of the function F1(θ) normalised by its maximum are given for p0 = 0, p0 = 0.5π(1 − ν) and p0 = 0.9π(1 − ν). We do not show the results for the viscosity dominated 8 (cid:21) . (23) -0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6regime, because when approaching p0 = π(1−ν) there is either no local maximum for θ < 0 that would provide continuity of the solution, or there is no value of θf that satisfies the condition of positive circumferential stress. Even for the analyzed liming case p0 = 0.9π(1 − ν) one can observe that, depending on the values of KII and ν, there may not exist a local maximum of F1 inside the interval of θ that corresponds to σθθ(r, θ) > 0. For this reason we will consider in the following two variants of the criterion for the case when the condition of local maximum of F1 is not satisfied: • variant I - we assume that there is locally no solution to the problem (or in other words, the criterion does not predict the propagation angle over certain range of the loading parameters); • variant II - we take the value of θ which corresponds to the global maximum of F1 in the domain σθθ ≥ 0; x a m 1 F / ) θ ( 1 F x a m 1 F / ) θ ( 1 F x a m 1 F / ) θ ( 1 F KII = 0.1 p0 π(1 − ν) = 0 p0 π(1 − ν) = 0.5 p0 π(1 − ν) = 0.9 KII = 0.5 KII = 0.9 θ θ θ Figure 2: MDSED: Value of F1(θ)/F max p0. The grey regions on the graphs correspond to the areas where σθθ < 0. for various values of Poisson's ratio and fixed KII and 1 Respective results for ν = 0.3 and all admissible values of KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] are presented in Fig.3 (variant I) and Fig.4 (variant II). In Fig.3 (variant I) one can clearly see the region in the vicinity of the viscosity dominated regime (edge BC), over which the solution for θf does not exist. Even if this drawback is eliminated in the variant II (Fig.4), the 9 -180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0-180°-120°-60°060°120°180°0.00.20.40.60.81.0solution in this area behaves in a rather peculiar way. Pertinent trends can be easily identified in Fig.5c). It shows that when employing variant II, the angle of crack propagation is only a continuous (but locally not smooth) function of KII and any small perturbation to the value of KII around the locations of θ(cid:48) f discontinuities in the considered parametric space can result in a serious change of θf . These facts cast doubts on the applicability of the MDSED criterion in the case when hydraulically induced tangential tractions are accounted for. Note that outside the discussed area both considered variants yield the same results. Clearly, for KII = 0 (edge AB) one has θf = 0. Moreover, for KII = 1 (edge CD) the solution can be found analytically: (cid:18) (cid:19) θf = − arccos 1 − 2√ 3 . (24) Notably, in the corners B and C ( KII = 0 and KII = 1 for the viscosity dominated mode) θf has no limit, which effectively means that the angle of crack propagation depends here crucially on the load history. D A C B Figure 3: MDSED: Predicted propagation angle θf for KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] for ν = 0.3. The local maximum of F1 is accepted (variant I). To complement the analysis of the MDSED criterion we show in Fig.5 the sensitivity of the crack propagation angle to the value of the Poisson's ratio. It turns out that for the toughness dominated regime one obtains the same results regardless of the value of ν (the same trend was reported in [16] for the Minimum Strain Energy Density Criterion). Some differences between the results for various ν can be observed when moving towards the viscosity dominated mode, however the influence of the Poisson's ratio on θf is rather limited. The maximal disparities between the crack propagation angles for set value of KII amount to 10◦ (that is around 20 % of the absolute value of θf ). 10 -85°-80°-75°-70°-65°-60°-55°-50°-45°-40°-35°-30°-25°-20°-15°-10°-5°0.00.20.40.60.81.00.00.20.40.60.81.0D A C B Figure 4: MDSED: Predicted propagation angle θf for KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] for ν = 0.3. The global maximum of F1 is accepted (variant II). θf p0 π(1 − ν) = 0 θf p0 π(1 − ν) = 0.5 θf p0 π(1 − ν) = 0.9 KII KII KII Figure 5: MDSED: Value of θf for various values of Poisson's ratio. 4 Modified Maximum Circumferential Stress (MMCS) crite- rion In order to avoid the mentioned above drawbacks of MDSED criterion we propose here a new one. It is based on the assumption that the direction of fracture propagation is defined by the maximal value of the circumferential stress (compare [16]) obtained along the elastic-plastic boundary in the elastic domain. In this way the new approach incorporates the plastic yield stress criterion, depends on the Mode III component and preserves all the advantages of the MCS criterion [16]. From now on it will be referred to as the Modified Maximum Circumferential Stress (MMCS) criterion. 11 -85°-80°-75°-70°-65°-60°-55°-50°-45°-40°-35°-30°-25°-20°-15°-10°-5°0.00.20.40.60.81.00.00.20.40.60.81.00.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°Below, we analyze four variants of the criterion, each of them based on a different yield condition i.e.: the von Mises condition, the Drucker-Prager condition, the Tresca condition or the Mohr-Coulomb condition. 4.1 MMCS criterion: von Mises yield criterion (MMCS-vM) Let us first analyse the MMCS under von Mises yield criterion. As it was in the case of the MDSED approach, the elastic-plastic boundary, rb, is found from (30).The shapes and sizes of plastic zones are shown in Fig. 1 with the respective comments included in Subsection 3. In order to determine the angle of fracture deflection one needs to find the maximum of the following function: and hence: F2(θ) = σθθ (rb(θ), θ) , θf = θ (cid:12)(cid:12)(cid:12)F2=F max 2 . (25) (26) 2 The graphs of function F2(θ) normalised by its maximum value, F max where the circumferential stress, σθθ is computed according to (23). , are presented in Fig. 6 for three values of KII = {0.1, 0.5, 0.9} and p0/(π(1 − ν)) = {0, 0.5, 0.9}. One can clearly see that this time, unlike the MDSED criterion, there is no problem with existence or uniqueness of the solution. There is only one maximum of F2 in the interval [−π, π]. Moreover, this maximum corresponds by nature to a positive value of the circumferential stress, σθθ. Relatively small influence of the Poisson's ratio on the crack propagation angle can be seen in the analyzed examples, however this issue will be discussed in more detail later on. In Fig. 7 the solution, θf , is presented for all admissible values of KII and p0 for a fixed Poisson's ratio, ν = 0.3. As can be seen, a continuous solution is obtained in almost the entire range of considered parameters (except for the point C). Naturally, for KII = 0 one has θf = 0, however this time no fracture deflection is observed over the whole span of the viscosity dominated regime (edge BC). For KII = 1 one has an invariable value of the crack propagation angle which amounts to -82.3◦. A peculiar behaviour of θf manifests in the corner C (viscosity dominated regime under severe antisymmetric load) where the solution has no limit. Likewise in the MDSED criterion, the actual angle of crack propagation depends here on the history of loading. As compared to the MDSED case, the MMCS-vM criterion provides lower values of θf . Lastly, an impact of the Poisson's ratio on the propagation angle is given in Fig. 8. The discrepancies between results obtained for three analysed values of ν are comparable to the ones for the MCS criterion for p0 > 0 (compare with Fig. 4b and Fig. 4c in [16]) and can reach over 6◦. However, in the toughness dominated regime (p0 = 0, Kf = 0, Fig. 8a) a dependence of the solution on the Poisson's ratio can be noticed, which was not the case for the MCS approach (Fig. 4a in [16]). For KII = 0 and KII = 1 the solution does not depend on the Poisson's ratio. 12 x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F KII = 0.1 p0 π(1 − ν) = 0 p0 π(1 − ν) = 0.5 p0 π(1 − ν) = 0.9 KII = 0.5 KII = 0.9 θ θ θ Figure 6: MMCS-vM: Value of F2(θ)/F max and p0. 2 for various values of Poisson's ratio and fixed KII 4.2 MMCS criterion: Drucker-Prager yield criterion (MMCS-DP) The second variant of the MMCS criterion utilizes the the Drucker-Prager yield theory in which the following relation holds: (27) where I1 = tr σ, J2 = 1/2 dev σ · dev σ, and α > 0 is a material parameter called pressure- sensitivity index and assumed to be a constant. It can be computed from experimental data by the following formula: αI1 + σt, J2 = α + α = , ξ = , (28) where σt and σc are the yield stresses in uniaxial tension and compression, respectively. Relation √ (28) can be easily obtained by evaluating the yield function (27) for a uniaxial compression state, I1 = −σc, J2 = σc/ 3, which yields (cid:112) (cid:18) ξ − 1 √ 3(ξ + 1) (cid:19) 1√ 3 σc σt (cid:19) 1√ 3 (cid:18) = α + 13 − ασc + σc√ 3 and then solving for α. σt, (29) -180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0D A C B Figure 7: MMCS-vM: Predicted propagation angle θf for KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] for ν = 0.3. θf p0 π(1 − ν) = 0 θf p0 π(1 − ν) = 0.5 θf p0 π(1 − ν) = 0.9 KII KII KII Figure 8: MMCS-vM: Value of θf for various values of Poisson's ratio. From (27) one can derive an expression for rb: rb = Rbrb(θ), θ ∈ (−π, π), where Rb = K2 IC σ2 t , 14 (30) (31) -75°-70°-65°-60°-55°-50°-45°-40°-35°-30°-25°-20°-15°-10°-5°-2°0.00.20.40.60.81.00.00.20.40.60.81.00.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°has a dimension of length [m] and the dimensionless part is rb = 3 3 + 3α)2 √ 8π( (cid:113) (cid:20) √ 3α(1 + ν) 4 (cid:18) (cid:19) KI cos θ 2 − KII sin θ 2 + (32) (cid:21)2 , K2 I c1(θ) + K2 II c2(θ) + KI KII c3(θ) + K2 f c4(θ) + KI Kf c5(θ) + KII Kf c6(θ) + 12 K2 III (33) where the functions cj(θ) are defined in (19). Note in the case of α = 0 the Drucker-Prager yield condition (27) coincides with the von Mises condition (15). As for many geomaterials the tensile strength is 8-10 times smaller than their compressive strength [22], we assume that ξ = 10 which yields α = 0.4724. To compute the angle of crack propagation, one needs to find the maximum of an auxiliary function F2 defined in (25) (where rb is calculated from (32)) and apply condition (26). The graphs illustrating rb(θ) computed according to equation (32) are presented in Fig. 9 (note that the scaling (20) is used in the figure). This time we see that, contrary to the previous criteria based on the von Mises theory, the regular elliptic shape of the plastic zone when approaching the viscosity dominated regime is retained only for the symmetric loading. Indeed, it is the only case when any symmetry of the yield area (here with respect to the plane of crack propagation) is observed. When increasing the value of KII the plastic zone shape becomes increasingly more irregular with a characteristic cut-off along the negative part of the horizontal axis. As for the size of the yield area, it is always the greatest for the viscosity dominated regime and the smallest for the toughness dominated mode. Just as it was in the case of von Mises criterion, two different trends can be observed for the limiting regimes. In the viscosity dominated one the size of the plastic zone decreases with growing KII , while for the toughness dominated mode an inverse tendency holds. In the figure we can also see that for constant KII an increase in p0 reduces the angle of crack propagation. In Fig. 10 the values F2(θ)/ max(F2) over the interval θ ∈ [−π, π] are plotted. As before for the MMCS-VM criterion, no problems with existence or uniqueness of solution are reported. There always exists a single value of θf for which the maximum of F2 (that corresponds to positive circumferential stress) is obtained. Again, one can observe a relatively low sensitivity of θf to the Poisson's ratio. In Fig. 11 the solution, θf , for all admissible values of KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] is presented for ν = 0.3. Similarly at it was in the MMCS variant utilizing von Mises yield criterion, a smooth solution exists over the whole range of the analyzed parameters. It is again only the point C (p0 = π(1 − ν), KII = 1) where θf has no limit and its actual value depends on the loading history. Conventionally θf = 0 for KII = 0. Moreover, no crack reorientation is observed in the viscosity dominated regime (p0 = π(1 − ν)). For a severe antisymmetric load ( KII = 1) the angle of crack propagation remains constant θf = −67.63◦. In general, when using MMCS-DP criterion one obtains smaller values of θf than those from MMCS-vM. Finally, to analyse how the value of Poisson's ratio affects the solution, graphs for three values of ν are plotted in Fig. 12. The solution for p0 = 0 and p0 = 0.9π(1− ν) depends on the value of ν to a very little extent. An impact of the Poisson's ratio was more noticeable for the 15 KII = 0 ν = 0 ν = 0.3 ν = 0.5 KII = 0.5 KII = 0.9 Figure 9: MMCS-DP: The shapes of of the plastic zones described by the normalized radius rb (20) for various values of p0 and fixed KII and ν. The blue lines reflect the actual sizes of the plastic zones, the red ones utilize the scaling factor 0.5, while the black curves are multiplied by 0.1. The angle of crack propagation, θf , is marked by a circle. von Mises yield criterion. Here, as can be seen in the figure, the maximum deviation of results for the neighbouring values of ν does not exceed 4◦. 16 -0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F KII = 0.1 p0 π(1 − ν) = 0 p0 π(1 − ν) = 0.5 p0 π(1 − ν) = 0.9 KII = 0.5 KII = 0.9 θ θ θ Figure 10: MMCS-DP: Value of F2(θ)/F max and p0. 2 for various values of Poisson's ratio and fixed KII 4.3 MMCS criterion: Tresca yield criterion (MMCS-TR) Let us analyze now the problem of fracture deflection on the assumption that the Tresca criterion is used to define the yield stress. The criterion itself reads [2]: (cid:34) (cid:112) 2 J2 cos cos−1 1 3 (cid:32) √ 3 3J3 2J 3/2 2 (cid:33) (cid:35) − π 6 = σt, (34) where J3 = 1/3 tr(dev σ)3 and σt is the uniaxial yield stress. Conventionally, the angle of crack propagation is found by maximizing the function F2 (25) along the elastic-plastic boundary. This time we do not write an expression for rb as it is to complex. The resulting areas of the plastic yield are depicted in Fig. 13. As can be seen, the obtained shapes are very similar to those for the MDSED criterion (compare Fig. 1). The general trends remain the same as in the recalled case, however the zones of plastic yield are larger now (which is quite obvious if one notes that the von Mises yield stress is always greater or equal to the Tresca yield stress). Some irregularities of the boundary are observed in the viscosity dominated regime for KII = 0.5 and KII = 0.9, which shall be explained later on. In Fig. 14 the qualitative behaviour of function F2(θ) is depicted. It resembles to a large degree that obtained for the MMCS-vM criterion (compare Fi.g 6). Again, no problem with 17 -180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0D A C B Figure 11: MMCS-DP: Predicted propagation angle θf for KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] for ν = 0.3. θf p0 π(1 − ν) = 0 θf p0 π(1 − ν) = 0.5 θf p0 π(1 − ν) = 0.9 KII KII KII Figure 12: MMCS-DP: Value of θf for various values of Poisson's ratio and: a) π(1−ν) = 0.5, c) π(1−ν) = 0.9. p0 p0 p0 π(1−ν) = 0, b) existence or uniqueness of solution appears. Fig. 15 presents a distribution of θf over the permissible range of p0 and KII for ν = 0.3. The general behaviour and values of solution are very similar to those reported for the MMCS- vM criterion (compare Fig. 7). Again no limit of θf is observed for p0 → π(1 − ν) and KII = 1 (corner C). For the extreme antisymmetric load ( KII = 1) θf = −83.4◦ was obtained. This time however we see some pecularity of the crack propagation angle. Namely there exists a line in the analyzed parametric space along which the solution smoothness is broken (it can be easily identified in the Fig. 15b)). After a careful analysis it turned out that this line 18 -60°-55°-50°-45°-40°-35°-30°-25°-20°-15°-10°-5°-2°0.00.20.40.60.81.00.00.20.40.60.81.00.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°KII = 0 ν = 0 ν = 0.3 ν = 0.5 KII = 0.5 KII = 0.9 Figure 13: MMCS-TR: The shapes of of the plastic zones described by the normalized radius rb (20) for various values of p0 and fixed KII and ν. The blue lines reflect the actual sizes of the plastic zones, the red ones utilize the scaling factor 0.5, while the black curves are multiplied by 0.1. The angle of crack propagation, θf , is marked by a circle. corresponds to a situation when for the predefined mixed mode loading the edge of the Tresca yield surface is reached. Respective transition points can be also identified in Fig. 16a) and Fig. 16b). Moreover, the aforementioned mechanism manifests itself in the irregularities of the plastic zone shapes that were noted when discussing Fig. 13. Finally, in Fig. 16 the influence of the Poisson's ratio on the crack propagation angle is presented. It shows that the level of deviations between the results for the neighbouring values of ν is similar to that in the previous criteria. However, unlike the previous cases, one obtains coincidence of the respective results for KII = 1 only when approaching the viscosity dominated 19 -0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F KII = 0.1 p0 π(1 − ν) = 0 p0 π(1 − ν) = 0.5 p0 π(1 − ν) = 0.9 KII = 0.5 KII = 0.9 θ θ θ Figure 14: MMCS-TR: Value of F2(θ)/F max and p0. 2 for various values of Poisson's ratio and fixed KII regime. It is the variant of ν = 0 which increasingly diverges from the two remaining curves for declining p0. 4.4 MMCS criterion: Mohr-Coulomb yield criterion (MMCS-MC) The last analyzed case assumes that the yield stress follows from the Mohr-Coulomb criterion [2]: ξ − 1 (cid:112)ξ2 + ξ + 1 I1√ 3 + 2 (cid:112) (cid:34) (cid:33) (cid:32) √ 3 3J3 2J 3/2 2 J2 cos cos−1 1 3 − tan−1 3 2ξ + 1 (cid:32) √ (cid:33)(cid:35) √ (cid:112)ξ2 + ξ + 1 3σc = , (35) where σt and σc are the yield stresses in uniaxial tension and compression, respectively, and ξ = σc/σt. Throughout this subsection we use ξ = 10, which is consistent with the value α = 0.4724 in the Drucker-Prager criterion in Sec. 4.2. The angle of crack propagation is found as previously by maximizing the function F2 (25). Again, as respective expression for the radius of the plastic zone is very complex, we do no present it here. In Fig. 17 the shapes of the plastic zones for different values of the analyzed parameters are shown. One can see a resemblance of the contours of the zones to those obtained for the 20 -180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0D A C B Figure 15: MMCS-TR: Predicted propagation angle θf for KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] for ν = 0.3. θf p0 π(1 − ν) = 0 θf p0 π(1 − ν) = 0.5 θf p0 π(1 − ν) = 0.9 KII KII KII Figure 16: MMCS-TR: Value of θf for various values of Poisson's ratio and: a) π(1−ν) = 0.5, c) π(1−ν) = 0.9. p0 p0 p0 π(1−ν) = 0, b) MMCS-DP criterion (compare Fig. 9). Their sizes are always greater for the Drucker-Prager criterion, which is especially pronounced in the toughness dominated regime (p0 = 0). The general trends remain the same as in the MMCS-DP variant. Fig. 18 shows the qualitative distribution of the F2 function. We can see that the results are hardly distinguishable from those obtained for the MMCS-DP. For every analyzed combination of parameters there exists a unique solution to the problem. A distribution of the crack propagation angle, θf , over KII ∈ [0, 1] and p0 ∈ [0, π(1− ν)] for ν = 0.3 is shown in Fig. 19. The values of θf are close to those computed for the MMCS-DP 21 -75°-70°-65°-60°-55°-50°-45°-40°-35°-30°-25°-20°-15°-10°-5°-2°0.00.20.40.60.81.00.00.20.40.60.81.00.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°KII = 0 ν = 0 ν = 0.3 ν = 0.5 KII = 0.5 KII = 0.9 Figure 17: MMCS-MC: The shapes of of the plastic zones described by the normalized radius rb (20) for various values of p0 and fixed KII and ν. The blue lines reflect the actual sizes of the plastic zones, the red ones utilize the scaling factor 0.5, while the black curves are multiplied by 0.1. The angle of crack propagation, θf , is marked by a circle. criterion. The general tendencies are also very similar. For KII = 1 θf = −72.45◦ is obtained. Again there is a line in the analyzed parametric space (Fig. 19a)) along which the smoothness of fracture deflection angle is broken. It is due to the edge of the Mohr-Coulomb yield surface, however the trend itself is much less pronounced than it was in the MMCS-DP. Finally, in Fig. 20 the impact of the Poisson's ratio on θf is analyzed. Respective trends remain the same as in the MMCS-DP criterion. In the toughness dominated regime (p0 = 0) the influence of ν is infinitesimal, respective curves are hardly distinguishable. 22 -0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6-0.6-0.4-0.20.00.20.40.6x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F x a m 2 F / ) θ ( 2 F KII = 0.1 p0 π(1 − ν) = 0 p0 π(1 − ν) = 0.5 p0 π(1 − ν) = 0.9 KII = 0.5 KII = 0.9 θ θ θ Figure 18: MMCS-MC: Value of F2(θ)/F max and p0. 2 for various values of Poisson's ratio and fixed KII 5 Comparison between criteria In this section we shall provide a brief comparison of the results obtained for respective criteria. We decided to omit the MDSED criterion. This is because it either provides no solution in a relatively wide area near the viscosity dominated regime (MDSED - variant I) or the solution in this region exhibits some instability (MDSED - variant II). For this reason we deem the MDSED criterion questionable, at least in the recalled zone. As for the remaining criteria we notice very strong similarities between the angles of crack propagation computed for the following pairs: i) MMCS-vM and MMCS-TR, ii) MMCS-DP and MMCS-MC. Its is not a surprise if one recalls that the von Mises yield surface can be considered a smooth approximation of the one for the Tresca criterion. Similarly, the Drucker- Prager yield surface is a smooth extension of that defined by the Mohr-Coulomb theory. Clearly, in those cases when the yield surface is represented by the quadratic in the space of principal stresses (MMCS-vM and MMCS-DP) the solution (θf ) is always a smooth function. Otherwise (MMCS-TR and MMCS-MC), if for a certain configuration of the loading the edge of the yield surface is transgressed, the angle of crack propagation is no longer smooth. This constitutes an evident drawback of the latter criteria. In Fig. 21 - Fig. 22 we have collated the results (the angles of crack propagation) obtained 23 -180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0-180°-120°-60°060°120°180°-1.5-1.0-0.50.00.51.0D A C B Figure 19: MMCS-MC: Predicted propagation angle θf for KII ∈ [0, 1] and p0 ∈ [0, π(1 − ν)] for ν = 0.3. θf p0 π(1 − ν) = 0 θf p0 π(1 − ν) = 0.5 θf p0 π(1 − ν) = 0.9 KII KII KII Figure 20: MMCS-MC: Value of θf for various values of Poisson's ratio and: a) π(1−ν) = 0.5, c) π(1−ν) = 0.9. p0 p0 p0 π(1−ν) = 0, b) for the discussed criteria under different values of KII , po and ν. It shows that there is a good convergence between the respective curves representing θf for two groups of criteria: i) MMCS-vM and MMCS-TR, ii) MMCS-DP and MMCS-MC (pressure sensitive materials), which was explained above. Indeed, in the cases when the edge of the yield surface is not reached, the MMCS-TR criterion mimics almost identically that of MMCS-vM (corresponding plots are hardly distinguishable from each other). For the pair: MMCS-DP and MMCS-MC, a worse resemblance is observed. Surprisingly, when the aforementioned condition is not met (the transition through the edge of yield surface occurs), the data for MMCS-TR and MMCS- 24 -65°-60°-55°-50°-45°-40°-35°-30°-25°-20°-15°-10°-5°-2°0.00.20.40.60.81.00.00.20.40.60.81.00.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°MC match almost perfectly starting from the viscosity dominated regime (p0 = 0) up to the moment when the solution kink takes place (see Fig. 22). Clearly, when approaching the viscosity dominated regime (p0 → π(1 − ν)) all the results become identical. Moreover, from Fig. 21 and Fig. 22 it can be seen that although the Poisson's ratio does affect the solution, generally in most cases the behaviour of θf and discrepancies between analysed criteria are very similar for each value of ν. Finally, we recall the results from [16] obtained for the Maximum Circumferential Stress (MCS) criterion which does not account for the plastic deformation effect. By comparing the graphs given in the cited paper with those presented here we conclude that the MCS criterion produces the angles of crack propagation very similar to those given by MMCS-DP and MMCS- MC over the entire range of analyzed loadings and parameters. 25 θf p0 π(1−ν) = 0 ν = 0 ν = 0.3 ν = 0.5 θf p0 π(1−ν) = 0.5 θf p0 π(1−ν) = 0.9 θf p0 π(1−ν) = 0.999 KII KII KII Figure 21: Predicted propagation angle θf for various KII and fixed p0 and ν. All analysed criteria are presented in each graph. 26 0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°0.00.20.40.60.81.0-100°-80°-60°-40°-20°020°θf KII = 0.1 ν = 0 ν = 0.3 ν = 0.5 θf KII = 0.5 θf KII = 0.75 θf KII = 0.9 p0 π(1−ν) p0 π(1−ν) p0 π(1−ν) Figure 22: Predicted propagation angle θf for various criteria are presented in each graph. p0 π(1−ν) and fixed KII and ν. All analysed 27 0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°0.00.20.40.60.81.0-80°-60°-40°-20°020°6 Conclusions In the paper a problem of redirection of a fluid driven crack was considered for a mixed mode loading that includes the classical Modes I-III and hydraulically induced tangential tractions on the fracture walls. The effect of plastic deformation in the near-tip zone was taken into account. Different criteria based on various yield conditions were employed to define the angle of fracture deflection. The following conclusions can be drawn from the conducted analysis: • The component of loading related to the hydraulically induced shear stress has a substan- tial influence on the crack orientation. Its magnification under a fixed Mode II loading reduces the angle of fracture deflection. High sensitivity of the angle to the magnitude of loading is observed for a combination of severe antisymmetric shear (Mode II) and the so called viscosity dominated regime of fracture propagation. In such a case the actual angle of fracture deflection depends on the loading history. • Among the analyzed criteria it is the one based on the Maximum Dilatational Strain Energy Density (MDSED) that seems to be the least credible. It stems from the fact that in a certain range of loading parameters near the viscosity dominated regime it either produces no solution or yields unstable results. • Strong similarities between the results are observed for those criteria that use related i) MMCS-vM and MMCS-TR, ii) MMCS-DP and for the pairs: yield conditions, i. e. MMCS-MC (pressure sensitive materials). • The angle of crack propagation is always a smooth function of the loading components (p0 and KII ) only for those criteria which employ yield surfaces that are quadratics in the principal stresses space (MMCS-VM and MMCS-DP). For MMCS-TR and MMCS-MC, depending on the configuration of loading, the angle of fracture deflection can be only a continuous function (locally not smooth). • The Maximum Circumferential Stress (MCS) criterion presented in [16] (which does not account for the plastic deformation effect) produces similar results to those obtained for MMCS-DP ans MMCS-MC over the entire range of analyzed parameters. This suggests that, at least in some cases, it may be sufficient to base the analysis on the Linear Elastic Fracture Mechanics without a need to account for the plastic deformation. • Even though the investigated criteria produce very similar results it needs experimental verification to decide which one is the most credible for a predefined type of fractured material and loading components. Special attention should be devoted here to establish whether the applied solid material model has to be pressure sensitive or not. Compliance with Ethical Standards Funding: This work was funded by European Regional Development Fund and the Republic of Cyprus through the Research Promotion Foundation (RESTART 2016 - 2020 PROGRAMMES, 28 Excellence Hubs, Project EXCELLENCE/1216/0481) (PP, MW) and by the ERC Advanced Grant Instabilities and nonlocal multiscale modelling of materials under number ERC-2013- ADG-340561-INSTABILITIES (AP, GM). The authors declare that these publicly funded ar- rangements have not created a conflict of interest. Acknowledgments GM gratefully acknowledges support from a grant No. 14.581.21.0027 unique identifier: RFMEFI58117X0027 by Ministry of Education and Science of the Russian Federation and is thankful to the Royal Society for the Wolfson Research Merit Award. The authors are also thankful to Prof. Davide Bigoni and Dr Monika Perkowska for their useful comments and discussions. References [1] Bigoni, D. and Radi, E. (1993) Mode I crack propagation in elastic-plastic pressure-sensitive materials. International Journal of Solids and Structures, 30(7): 899-919 [2] Bigoni, D. and Piccolroaz, A. (2004) Yield criteria for quasibrittle and frictional materials. International Journal of Solids and Structures, 41: 2855-2878. [3] Camas D., Hiraldo I., Lopez-Crespo P., Gonzalez-Herrera A. (2011) Numerical and exper- imental study of mixed-mode cracks in non-uniform stress field. Procedia Engineering, 10: 1691-1696. [4] Cherny S., Lapin V., Esipov D., Kuranakov D., Avdyushenko A., Lyutov A., Karnakov P. (2016) Simulating fully 3D non-planar evolution of hydraulic fractures. International Journal of Fracture, 201: 181-211. [5] Cherny S., Esipov D., Kuranakov D., Lapin V., Chirkov D., Astrakova A. (2017) Prediction of fracture initiation zones on the surface of three-dimensional structure using the surface curvature. Engineering Fracture Mechanics, 172: 196-214. [6] van Dam D.B and de Pater C.J. (1999). Roughness of Hydraulic Fractures: The Importance of In-Situ Stress and Tip Processes, SPE, 56596. [7] van Dam D.B, Papanastasiou P, De Pater C.J. (2002) Impact of rock plasticity on hydraulic fracture propagation and closure, J. SPE Production & Facilities, 17 (3): 149-159. [8] Erdogan F., Sih G.C. (1963) On the crack extension in plates under plane loading and transverse shear. Journal of Basic Engineering, 85(4): 519-525. [9] Economides M., Nolte K. (2000) Reservoir Stimulation, 3rd edn. Wiley, Chichester. [10] Hallback N., Nillson F. (1994) Mixed-mode I/II fracture behavior of an aluminum alloy. Journal of the Mechanics and Physics of Solids ,42(9): 1345-1374. 29 [11] Lazarus V., Buchholz F.-G., Fulland M., Wiebesiek J. (2008) Comparison of predictions by mode II or mode III criteria on crack front twisting in three or four point bending experiments. International Journal of Fracture, 153(2): 141-151. [12] Papanastasiou P., Thiercelin M. (1993) Influence of inelastic rock behaviour in hydraulic fracturing. International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 30: 1241-1247. [13] Papanastasiou P. (1997) The influence of plasticity in hydraulic fracturing. International Journal of Fracture, 84: 61-97. [14] Papanastasiou P. (1999) The effective fracture toughness in hydraulic fracturing. Interna- tional Journal of Fracture, 96: 127-147. [15] Papanastasiou P., Durban D. (2018) The influence of normal and shear stress loading on hydraulic fracture-tip singular plastic fields. Rock Mechanics and Rock Engineering, 51(10): 31913203. [16] Perkowska M., Piccolroaz A., Wrobel M., Mishuris G. (2017) Redirection of a crack driven by viscous fluid, International Journal of Engineering Science , 121: 182-193. [17] Sarris E., Papanastasiou P. (2011) The influence of the cohesive process zone in hydraulic fracturing, International Journal of Fracture, 167: 33-45. [18] Sarris E., Papanastasiou P. (2013) Numerical Modelling of Fluid-Driven Fractures in Co- hesive Poro-elastoplastic Continuum, International Journal for Numerical and Analytical Methods in Geomechanics, 37(12): 1822-1846. [19] Sih G.C. (1974) Strain-energy-density factor applied to mixed mode crack problems. In- ternational Journal of Fracture, 10(3): 305-321. [20] Theocaris P.S., Andrianopoulos N.P. (1982) The T-Criterion applied to ductile fracture. International Journal of Fracture, 20: R125-R130. [21] Wang H. (2015) Numerical Modeling of Non-Planar Hydraulic Fracture Propagation in Brittle and Ductile Rocks using XFEM with Cohesive Zone Method, Journal of Petroleum Science and Engineering, 135: 127-140. [22] Weinberger R., Reches 2.. Eidelman A., Scott T.S. (1994) Tensile properties of rocks in four-point beam tests under confining pressure, in Proceedings First North American Rock Mechanics Symposium, Austin, Texas, eds Nelson, P. & Laubach, S.E., pp. 435-442 [23] Wrobel M., Mishuris G. (2015) Hydraulic fracture revisited: Particle velocity based simu- lation. International Journal of Engineering Science, 94: 23-58. [24] Wrobel M., Mishuris G., Piccolroaz A. (2017) Energy Release Rate in hydraulic fracture: can we neglect an impact of the hydraulically induced shear stress? International Journal of Engineering Science, 111: 28-51. 30 [25] Wrobel M., Mishuris G., Piccolroaz A. (2018) On the impact of tangential traction on the crack surfaces induced by fluid in hydraulic fracture: Response to the letter of A.M. Linkov. Int. J. Eng. Sci. (2018) 127, 217219. International Journal of Engineering Science, 127: 220-224. [26] Yehia N.A.B. (1991) Distortional strain energy density criterion: the Y-Criterion. Engi- neering Fracture Mechanics, 39(3): 477-485. 31
1907.09996	1	1907	2019-05-02T12:28:11	Wireless and Batteryless Surface Acoustic Wave Sensors for High Temperature Environments	[ "physics.app-ph", "physics.ins-det" ]	Surface acoustic wave (SAW) devices are widely used as filter, resonator or delay line in electronic systems in a wide range of applications: mobile communication, TVs, radar, stable resonator for clock generation, etc. The resonance frequency and the delay line of SAW devices are depending on the properties of materials forming the device and could be very sensitive to the physical parameters of the environment. Since SAW devices are more and more used as sensor for a large variety of area: gas, pressure, force, temperature, strain, radiation, etc. The sensors based SAW present the advantage to be passive (batteryless) and/or wireless. These interesting properties combined with a small size, a low cost radio request system and a small antennas when operating at high frequency, offer new and exiting perspectives for wireless measurement processes and IDTAG applications. When the materials constituting the devices are properly selected, it becomes possible to use those sensors without embedded electronic in hostile environments (as high temperature, nuclear site, ...) where no solutions are currently used. General principle of the SAW sensor in wired and wireless configurations will be developed and a review of recent works concerning the field of high temperature applications will be presented with specific attention given to the choice of materials constituting the SAW device.	physics.app-ph	physics	WIRELESS AND BATTERYLESS SURFACE ACOUSTIC WAVE SENSORS FOR HIGH TEMPERATURE ENVIRONMENTS T. Aubert1, O. Elmazria1,2, M.B. Assouar1 1Institut Jean Lamour (IJL), UMR 7198 CNRS-Nancy University 54506 Vandoeuvre lès Nancy, France 2Ecole Supérieur des Sciences et Techniques d'Ingénieurs de Nancy, 54506 Vandoeuvre-lès-Nancy, France e-mail : [email protected] Abstract and/or wireless. These Surface acoustic wave (SAW) devices are widely used as filter, resonator or delay line in electronic systems in a wide range of applications: mobile communication, TVs, radar, stable resonator for clock generation, etc. The resonance frequency and the delay line of SAW devices are depending on the properties of materials forming the device and could be very sensitive to the physical parameters of the environment. Since SAW devices are more and more used as sensor for a large variety of area: gas, pressure, force, temperature, strain, radiation, etc. The sensors based SAW present the advantage to be passive (batteryless) interesting properties combined with a small size, a low cost radio request system and a small antennas when operating at high frequency, offer new and exiting perspectives for wireless measurement IDTAG applications. When the materials constituting the devices are properly selected, it becomes possible to use those sensors without embedded electronic in hostile environments (as high temperature, nuclear site, …) where no solutions are currently used. General principle of the SAW sensor in wired and wireless configurations will be developed and a review of recent works concerning the field of high temperature applications will be presented with specific attention given to the choice of materials constituting the SAW device. processes and Key words: SAW sensors, Wireless, High temperature, LGS, AlN I. INTRODUCTION be could catalysis Surface acoustic wave (SAW) devices are used for several years as components for signal processing in communication systems. SAW devices are for example widely used as bandpass filter and resonator in mobile phones [1,2]. Far from being confined to this single use, SAW are or may find applications in many other areas. The SAW can be used to generate movement in microfluidics leading to mixe move, and heat very low quantities of liquid in the range of nanoliter [3]. They can also be used in chemistry, where some properties of the SAW in terms of heterogeneous identified. Applications in the cleaning and decontamination surface (desorption controlled ejection of drops, dust and fragments of coatings ...) are also possible. Finally, SAW offer very interesting prospects in the field of sensors. It is indeed quite possible to achieve with these devices small size and very robust sensors for deformation [4,5], temperature and gas [6,7]. Also note that in addition to being small, simple and robust, these devices have the advantage of being passive requestable (wireless) [8,9] and are inexpensive if fabricated on a large scale. The use of SAW devices as passive and wireless sensors allows them to operate in extreme conditions such as those with high levels of radiation, high temperatures, or electromagnetic in which no other sensors can operate. This is obviously conditioned by the fact that the materials constituting the device permit. In this paper we focus our interest on SAW sensor for high temperature environment by (batteryless) and interference, remotely given a review of main works performed in this field. The study will be started by explaining the operating principle of SAW sensor and particularly the wireless SAW sensors. Note that the SAW sensors capable of operating in high-temperature environments will find applications in aerospace, power, nuclear, steel industry, automotive, chemical, and petrochemical processes plants [10,14]. II. SAW SENSOR PRINCIPLE The SAW devices bring into use the electro- acoustic properties of piezoelectric substrates such as quartz, lithium niobate, zinc oxide, to generate surface acoustic waves. The mechanical vibrations generated propagate at velocity depending on the crystalographic direction of material and its elastic, piezoelectric and dielectric constants [1,2]. The generation of acoustic wave is done with two electrodes, also called Interdigital transducers (IDT). The IDT consist on a structure of overlapping metal fingers that is fabricated on the piezoelectric substrate by a photolithographic process (Fig. 1). When an alternative voltage is applied to the IDT input, it follows alternating cuts and horizontal and vertical expansion of underlying material, which can generate surface acoustic waves. Different propagation modes can be generated, in this case and the commonly used is the Rayleigh wave that is similar to waves that spread on the surface of the material. Figure 1: Optcal Photograph of SAW device (Aluuminum IDT on Quartz substrate. In insert, a scanning electron micrograph shows a zoom of the IDTs. The acoustic wave devices are sensitive to any disturbance that may affect the velocity, distance travel or even the mode of wave propagation. A Imput IDT Output IDT Acoustic wave in gas The disturbance resulting in a variation of the electrical response of the device (frequency, amplitude ...). SAW systems are no exception to this rule and are sensitive to three major types of disturbances: the change in temperature, deformation and in gaseous, liquid or solid species deposited on acoustic wave travel surface. The change temperature and deformation induces an effect on both a variation of speed (alteration coefficients elastic and piezoelectric) and a change of length to cross. species deposition (adsorption/ absorption) varies the speed of waves. This variation may result from a change of inertia mechanical surface (increase in a move masse), a modification of elastic coefficients (resulting from the spread of chemical species adsorbed in the propagation media) or a disruption of the electric field surface. In the case of the liquid or solid species their deposition affect and modify mode of wave propagation. Adding a layer of liquid or viscoelastic film on the surface may cause the emergence of guided modes at frequencies. The velocity of these modes depends on the physical parameters of the guide layer and it is therefore possible to measure them by monitoring the evolution of the frequency response of the devices. These parameters are the viscosity, density, stiffness, thickness, etc. All SAW sensors on the market operate one or a combination of these three sensitivities. III. WIRLESS SENSOR PRINCIPLE Two principles could be considered to use SAW devices as wireless sensor, delay lines or resonators. Both principles are described below: III-1 Delay line The operating principle of such a sensor system is sketched in figure 2. The reader unit (local radar transceiver) sends out a radio frequency (RF) electromagnetic read-out signal. This read-out signal is picked up by the antenna of the passive SAW transponder and conducted to an IDT. The IDT converts the received signal into a SAW signal by the converse piezoelectric effect. The SAW propagates towards a characteristic pattern. A small part of the wave is reflected at each reflector. The micro acoustic wave packets now returning to the IDT are re-converted distributed reflectors several in into electrical signals by the IDT and retransmitted to Radio waves Reader Unit the radar [10]. IDT Reflected SAW Antenna Reflector s Figure 2: SAW wirless sensor in delay line configuration. Same principle is used when SAW device is considered as RFID Tag. The mesurand (temperature, pressure, strain, …) affects the velocity of the micro acoustic wave and thereby also the time distances of the RF transponder response. The evaluation of this response signal in the radar unit thus allows the determination of the environmental temperature or pressure of the passive SAW transponder. III-2 Resonator The technological advances in the field of SAW have allowed the achievement resonator with a high quality factor (Q) which allows design sensor based SAW resonator (SAWR) with high sensitivity, accuracy, long-term stability and the possibility of storing electromagnetic energy, A SAW resonator consists of the piezoelectric substrate, an interdigital transducer (IDT), and two reflectors in the direction of the propagating wave (Fig. 3). The IDT is connected to an antenna. It receives energy for the SAW by an electromagnetic wave coming from the interrogation unit. The to mechanical energy of the surface acoustic wave. The two reflector gratings form a resonating cavity in which a standing wave is generated in the case of resonance. A portion of the stimulating electromagnetic energy is stored in this standing wave. After the stimulating IDT converts electrical energy the excitation of signal is switched off, energy still is present in the form of the SAW. The IDT converts a portion of the mechanical energy back to electrical energy now because the process of energy conversion is partially reversible. The electrical energy is transmitted as an electromagnetic wave back to the interrogation unit and can be analyzed. Typical response obtained at the input of interrogation unit is shown in figure 4. Reflector 2 Reflector 1 IDT Figure 3: One port SAW resonator architecture Duration of oscillations is of course depending on quality factor of the resonator. Resonance frequency of the sensor, directly linked to the mesurand, could this signal using Fourrier from be extracted Transform Analysis for example [11-13]. Figure 4: Typical response signal obtained at the input of interrogation unit. VI. SAW SENSOR FOR HIGH TEMPERATURE APPLICATIONS As mentioned above, the use of SAW devices as passive and wireless sensors allows them to operate in extreme conditions including high temperature where no other electronic based system can operate. This unique capability makes SAW sensor very attractive for such applications. Several groups are working on achieving SAW devices capable of operating in high temperatures conditions up to 1000 °C studied for applications in aerospace, power, nuclear, chemical, and petrochemical processes plants. Knowing that the conventional substrates as piezoelectric quartz or lithium niobate are limited in temperature, R&D are focused on new generation of piezoelectric materials stable in these conditions, as Langasite (LGS) and gallium phosphate (GaPO4) [10,14] or layered structures such as GaN/Sapphire, AlN/Sapphire or AlN/diamond [15]. The metal constituting the electrodes must also be reistant to such treatments and platinum (Pt, melting point 1763◦C) is preferred to, the commonly used metal aluminium (Al, melting point 660◦C), which cannot be used above 200◦C due to problems with softening and electromigration. Additional layer is required for good adhesion between platinum and substrate. Zirconium [16] and tantalum are investigated as adhesion layer and are preferred on commonly titanium used to It that semiconductor application. Antenna should be also optimized for high temperature or harsh environment reached. VI-1 Piezoelectric material is now well demonstrated the conventional SAW substrates, such as quartz or lithium niobate (LiNbO3) are not suitable for high temperature applications. For example, quartz knows a phase transition at 573°C, which lets him almost un- piezoelectric. However, Hornsteiner et. al. have even shown that SAW signals on quartz decrease dramatically above about 500°C [17]. They attribute this phenomenon to twins formation in the crystal. In the case of lithium niobate (or lithium tantalate), the limiting factor is the chemical decomposition that starts at about 300°C [9]. Hauser et. al. have demonstrated that the kinetic of this phenomenon follows an Arrhenius law. Thereby, the lifetime of a LiNbO3-based SAW device is about 10 days at 400°C, 1 day at 425°C, but only 2 hours at 450°C [18]. Above this temperature, they also noticed a strong deterioration of IDTs , that could be attributed to the pyroelectricity of LiNbO3 which leads to the formation of sparks between the electrodes. To overcome these limitations, new piezoelectric materials, such as langasite (La3Ga5SiO14), gallium phosphate (GaPO4) and aluminium nitride (AlN) have been proposed about ten years ago [19]. Since then, as we will discuss below, if langasite has paid much more attention and gave better results than gallium phosphate, aluminium nitride have not been yet really investigated in high temperature SAW applications. • Langasite (LGS) Langasite belongs to the same crystal class as quartz but does not undergo a phase transition up to its melting Its electroacoustic properties at room temperature are enough good to consider it as a candidate to replace quartz in some applications. Indeed, its lower SAW velocity allowing the fabrication of smaller devices for a given frequency, the existence of temperature- compensated cuts with zero power flow angle and minimal diffraction, and especially its coupling coefficient k2 which is about four times higher are strong arguments [21] (see table 1). Moreover, wafers of good quality are commercially available. At last, elastic, piezoelectric, dielectric constants of LGS and their temperature variations up to the second temperature 1470°C [20]. at order are available, allowing precise simulation [22,23]. It has to be noted that these sets are relevant in the range from 0°C to about 250°C [24,25]. There is currently no LGS constants set available for higher temperatures. Tab. 1 : comparison between electroacoustic properties of quartz and LGS vSAW (m/s) [26] Quartz ST-X [9] LGS 3200 2700 k2 (%) 0,12 0,44 Prop. losses at 1 GHz (dB/cm) 10 15 (0◦, 138.5◦, 26.6◦) Langasite have been intensively investigated for high temperature SAW applications in the last decade [9,17, 20-21, 25, 27-35]. The most promising results are : - No deterioration of LGS substrates was mentioned in any of these papers - SAW signals on langasite were measured up to 1085°C [17]. The authors pointed out this limit was that of the used package, not of LGS. - LGS-based SAW device have been operated at 800°C for more than 5½ months [35], showing very good stability. - LGS crystals have shown great resistance to thermal shock from ambient temperature to 700°C [35]. All these results demonstrate that langasite is well adapted for high temperature applications, even with extreme variations of temperature. Nonetheless, it is necessary to slightly qualify this idyllic description. In fact, Fachberger et. al. have highlighted a dramatic the LGS propagation losses with both the frequency and the temperature, which led them to define 1 GHz as a limit frequency for high temperature RF sensor applications [29]. The table below illustrates this problem. Table 2 : Maximum reached temperature for different increase of operating frequencies Operating frequency 100 MHz 167 MHz 434 MHz 1 GHz Maximum reached temperature 1085 °C 850°C 750°C 500°C Referen ce [17] [35] [27] [29] for high temperature • Gallium phosphate Gallium orthophosphate (GaPO4) is also a promising material electroacoustic applications. Though, its properties are a notch below those of LGS. As the latter, it has the same crystallographic structure as quartz, but it undergoes a phase transition at 930°C [36]. We can therefore expect some problems starting at about 850 °C from the experience conducted with quartz, which suffer the same type of crystalline phase transition. To our best knowledge, such temperatures have not been reached yet in acoustic applications. Buff et. al. have conducted experiments with GaPO4-based SAW resonators up to 700°C [28], whereas some other studies on BAW and SAW devices have reached 600°C [14, 37, 38]. Anyway, electroacoustic properties of GaPO4 are slightly lower than those of LGS (see table 3). Note that this problem (strong propagation losses) is largely due to the fact that currently there are no good quality crystals available on the market [9]. Table 3 : comparison between electroacoustic properties of GaPO4 and LGS [9] GaPO4 (90◦, 5◦, 0°) LGS (0◦, 138.5◦, 26.6◦) vSAW (m/s) 2500 2700 K2 (%) 0,29 0,44 Prop. losses at 1 GHz (dB/cm) 25 15 • Aluminium nitride (AlN) Aluminium nitride brings many qualities that make it a material of great interest for electronic and opto- electronic applications. It has for instance a large bandgap of 6,2 eV, high resistivity of 1013 Ω.cm and high thermal conductivity of 320 W.mK-1 [39, 40]. In the SAW field, it has attracted considerable attention because of its high phase velocity, close to 5700 m.s-1 [41], allowing the fabrication of high frequency devices up to 5 GHz [42, 43] when combined with high velocity substrate as diamond or sapphire. Unlike piezoelectric materials cited previously, it cannot be obtained by crystal growth methods, but must be deposited on a substrate. Several methods have been applied to obtain thin films of AlN: MOCVD [44, 45], pulsed laser deposition [39], magnetron sputtering [46], etc In order to use in high temperature conditions, sapphire could be a reasonable choice of substrate. Indeed, it is commercially available and not very expensive unlike diamond for example, can withstand very high temperature (at least 1600°C) [47], has a reasonable lattice mismatch with AlN allowing an epitaxial growth [48], and its surface acoustic wave velocity is very close to that of AlN [49], ensuring operating at high frequencies for AlN/Sapphire SAW devices. The AlN/Sapphire layered structure has been intensively investigated for room temperature SAW applications because of the possibility to obtain zero TCF SAW devices [44, 45, 49]. Compared to LGS, AlN/Sapphire structure exhibits higher properties concerning electromechanical coupling or acoustic propagation losses (see table 4). velocity, SAW Tab. 4 : comparison between electroacoustic properties of AlN/Sapphire and LGS AlN (0001)/ Sapphire (0001) LGS (0◦, 138.5◦, 26.6◦) vSAW (m/s) 5700 [44] k2 (%) 0,65 [49] Prop. losses at 1 Ghz (mdB/λ) 0,7 [45] 2700 [9] 0,44 [9] 4 [9] Aluminium nitride can be used in high temperature conditions. It melts at 3214°C but starts to decompose chemically in vacuum at 1040°C following the reaction 2AlN(s) = 2Al(s) + N2(g) [50]. In an N2-rich atmosphere (like air), the equilibrium will be shifted towards AlN formation, and the decomposition temperature will be, of course, higher. Thus, Pattel and Nicholson have measured a piezoelectric activity for AlN up to 1150°C [51]. Moreover, the bilayer structure offers also an outstanding advantage for harsh conditions. Indeed, it is possible to sandwich the IDTs between the thin film and the substrate, offering them a natural protection against external aggression. Nonetheless, there are two potential limitations to the use of AlN at high is is no report pyroelectric [19]. Currently, pointing out problems occurring at high temperature with AlN-based SAW devices due this temperatures. First, there to it temperature the maximum phenomenon, and for a good reason : to our best of knowledge, investigation for this kind of SAW device is only 350°C [52] ! Recently, Kishi et. al. have successfully investigated AlN thin films as pressure sensor for engine combustion chamber up to 700°C without mentioning problems due to pyroelectricity [53]. The second problem is related to the oxidation of AlN by the reaction 4AlN(s) + 3O2(g) = 2Al2O3(s) + 2N2(g). In air, it starts at 950°C and films of 500 nm thickness are completely oxidised at 1000°C in a time of minutes [54]. To summarize, conventional SAW substrates cannot be used for high temperature applications. Langasite have proved its efficiency in harsh conditions up to 1000°C, but suffers from large propagation losses at high frequency, which limit its use at frequencies below 1 GHz. This may be a problem for achieving wireless sensor with reasonable size of antenna. An alternative solution could come from the layered structure AlN/Sapphire that can theoretically operate at frequencies up to the ISM band at 2,45 GHz and up to 950°C in air. VI-2 Metallic material for electrodes The 3 main requirements concerning the material constituting the IDT are: • High electrical conductivity; • Elevated melting temperature, largely higher than the aimed operating temperature of the device (as we will discuss further) • Good resistance to oxidation at high temperature, and more generally good chemical inertness even at high temperature Thus, noble metals (gold, palladium, platinum, rhodium, ruthenium, iridium …), well known for their chemical inertness, are serious candidates. Melting temperature and electrical resistivity of those materials are given in table 5. Tab. 5 : Properties of noble metals Metal Gold Palladium Platinum Melting temperature (°C) [55] Resistivity (µΩ.cm) [17] 1063 1550 1773 2,3 10,5 10,6 Rhodium Ruthenium Iridium 1966 2334 2440 4,7 7,7 5,3 [56]. temperature Nonetheless, except platinum, all these metals have some weaknesses. Thus, gold has a melting temperature too close from the aimed range (400- 1000°C). The solubility of oxygen is very high in palladium at high Iridium, ruthenium and osmium form oxide phases at temperatures close to 700-800°C, and these oxides become volatile in the range 900-1000°C [56]. At last, Rhodium form also an oxide phase, but protective. Platinum is an exception, because the very thin (some nm) surface oxide existing at room temperature disappears at temperatures above 400°C [57]. So, platinum is the only noble metal for which no mass variation is detected up to 1000°C when heated in air [56]. This aspect is very important because any mass variation would interfere with temperature sensing, both mass and temperature variation resulting in frequency shifting of the device [28, 37] These properties, but also the greater experience in deposition and processing of platinum over other noble metals, can explain that currently, all the studies on high temperature SAW devices were conducted quasi-exclusively with platinum electrodes [17, 20-21, 25, 27-28, 30-35]. To our best knowledge, the only exception can be found in [30] where palladium IDTs were investigated. But, the aim of the authors was to use the high H2-absorption capacity of palladium to realize a SAW gas sensor. • Adhesion layer lithium niobate, However, the chemical inertness of platinum (and other noble metals) have a major drawback : it can't form chemical bonds with nitride or oxide materials such as quartz, langasite or aluminium nitride. As a result, the adhesion of platinum on these substrates is very poor [58]. A classical solution consists in placing underneath the platinum film a more reactive metal as adhesion layer [59, 60]. Bonding between platinum and this metal will be assured by interface alloying. Detailed studies on the adhesive strength of metals show that the optimal thickness of the adhesion layer is 10-20 nm [61]. Of course, for high temperature applications, this metal must be refractory. So, good candidates temperature. Diffusion of oxygen could be titanium, zirconium, tantalum, tungsten, cobalt, chromium ... In practice, titanium [14, 17, 27, 37, 55], zirconium [14, 21, 30-33, 35, 37] and tantalum [25, 34, 58, 59] are commonly used for intermediate-to-high temperature applications. Maeder et. al. have compared the behaviour of these 3 adhesion layer materials after an annealing of 10 minutes in a low-pressure (10 Pa) O2 atmosphere at 620°C [62] . They showed the superiority of tantalum and zirconium over titanium. Indeed, Zr and Ta form dense oxide (ZrO2 and Ta2O5) layers below the platinum during the annealing. It is due to the fact that O2 diffuses more quickly than Zr and Ta in platinum. The important fact is that this oxidation is not a real problem, the adhesion of the platinum film remaining very good after annealing. Titanium in turn raises serious problems already at this through platinum film leads also to the formation of TiO2 under the latter, but unlike Zr and Ta, Ti diffuses through its oxide phase and through platinum, leading to the formation of oxide precipitates in grain boundaries inside the platinum film or near to the surface. The result is a loss of adhesion for the platinum film, the modification of its electrical properties and eventually its destruction. Several studies have confirmed the unsuitability of titanium as adhesion layer above 600°C, as well as the good performance of tantalum and zirconium up to 700°C [14, 16, 21, 27, 30, 32, 37, 58, 59]. Thereby, AES measurements have highlighted that Zr have not diffused at all through Pt after an 8h 700°C annealing [16]. Tiggelaar et.al. have obtained the same result with tantalum after a 20 min annealing at the same temperature [58]. • Limitations of platinum thin film Nonetheless, a progressive deterioration of thin films of platinum is systematically observed when they are investigated at temperatures above 700°C, even when Zr or Ta adhesion layers are used [32, 35, 58, 59]. This phenomenon starts by the apparition of crystallites at the surface of the film as shown by SEM micrograph (Fig. 5). If the experience goes on, voids appear and grow, resulting in the formation of separate metal islands (Fig. 6) and leading to the loss of the electrical continuity of the film [58, 59] and the failure of the is essential SAW device. Understanding the origin of these crystallites to consider alternative solutions. The most probable assumptions are : (1) melting of a Pt-Ta or Pt-Zr eutectic (2) alloying between Pt and Ta (or Zr) (3) oxidation of the Pt thin film (4) recrystallisation of the Pt thin film layer [58]. Now, without adhesion the same phenomena of deterioration were observed with these films during annealing. Assertion removed because experiments conducted in the vacuum, nitrogen or argon have leaded to the same results [32, 33, 58, 59]. Furthermore, platinum oxide PtO2 cannot exist above 400°C [57]. (3) can also be The only remaining possibility (recrystallisation of platinum) is confirmed by XRD measurements. Indeed, the Pt peak width after annealing is about twice smaller than before. This indicates a strong increase of the Pt grain size, i.e. an important recrystallisation [58, 59]. Figure 5 : Crystallites at the surface of Pt thin film • Alternative solutions The mobility of atoms at the surface of a metal (which is essential to the recrystallisation in the case of thin films) depends on the temperature. As a rule of thumb, it is known that the latter becomes high above a "critical temperature" close to half the melting temperature of the metal (in K) [58, 59]. In is presented the this critical temperature for the noble metals family. following table Table. 6 : Melting and Critical temperatures of noble metals Metal Gold Palladium Platinum Rhodium Ruthenium Iridium Melting temperature (°C) Critical temperature (°C) 1063 1550 1773 1966 2334 2440 395 638 748 847 1031 1084 Figure 6 : Voids at the surface of Pt thin film Assertions (1) and (2) can be excluded for two reasons : first, the lowest eutectic point found in the Pt-Ta diagram is at 1760°C (1190°C for the Pt-Zr case) [63]. Otherwise, Tiggelaar et.al. have developed a method to obtain highly adhesive Pt thin film the experimental The correlation between the critical temperature of platinum and temperature of recrystallisation (about 700°C) is striking. Anyway, it seems likely that with gold, recrystallisation problems would occur at temperature close to 400°C. That is why gold, despite a melting temperature above 1000°C, is not used for high temperature applications. Therefore, it appears that two alternative solutions are possible. The first one would consist to use ruthenium or case recrystallisation problems should not occur below instead platinum. iridium this In 1000°C. This assumption is confirmed by [55] where an Ir thin film annealed for two hours at 1000°C in nitrogen showed excellent stability. Nonetheless, the use of ruthenium or iridium in air at high temperature will be difficult. Indeed, these two materials are known to form volatile oxide phases at about 900°C, this phenomenon being much faster in the case of ruthenium [56]. Thus, Lisker et.al. have shown by XRD measurements that 100 nm-thick iridium thin film is almost completely oxidized after a 30 min annealing in oxygen at 800°C [64]. The situation is even worse with ruthenium for which quasi-complete oxidation occurs from 700°C, and high losses of the volatile RuO2 phase are observed at 900°C. This effect is not observed in this work for IrO2 but Krier et.al. have measured a rate of weight loss for iridium at 1000°C equal to 3.1 mg/cm2/h, which corresponds to a recession rate of about 300 nm/h [56] ! Therefore, it is clear that it is not possible to consider iridium (or ruthenium) IDTs without a protective layer. Actually, Iridium IDTs sandwiched structure discussed in the piezoelectric material section could be a good solution to achieve high frequency, long-term operation wireless sensor up to 900°C. The second alternative solution could come from the use of Pt-Rh alloy in replacement of platinum. Indeed, rhodium, as platinum, does not undergo weight loss due to oxidation up to 1000°C [56]. Moreover, in alloys, precipitates forming in grain boundaries contribute to prevent grain boundary diffusion, and thus recrystallisation [59]. This way have been recently explored by Da Cunha et. al., who have highlighted the excellent stability of a Pt/10%Rh alloy up to 950°C [31]. the AlN/Sapphire in V. CONCLUSION Wireless SAW sensors are a promising solution for applications in harsh environment, such high temperature, in which no other wireless sensors can operate. However, conventional SAW substrates cannot be used for high temperature applications. Langasite have proved in harsh conditions up to 1000°C and shows the optimal solution for wired solution. However, LGS suffers from large propagation losses at high frequency, which limit its use at frequencies below 1 GHz. This may be a problem for achieving wireless sensor with its efficiency reasonable size of antenna. An alternative solution could come from the layered structure AlN/Sapphire that can theoretically operate at frequencies up to the ISM band at 2,45 GHz and up to 950°C in air. Regarding the metallic electrodes, it has been proved that platinum thin films cannot be used above 700°C for long-time applications because of recrystallisation phenomena. Interesting alternative solutions could come from the use of iridium or platinum/rhodium alloys. REFERENCES [1] Surface Acoustic Wave Filters With application to communications and signal processing; David Morgan; 2d Edition, 2007; Elsevier, Academic Press. [2] Surface Acoustic Wave for Mobile and Wireless Communications; C. K. Campbell; 1998 Academic Press. [3] Microfluidic device based on surface acoustic wave Beyssen, D., Le Brizoual, L., Elmazria, O., Alnot, P. Sensors and Actuators, B: Chemical 118 (1-2), pp. 380-385 [4] D. Cullen and T. Reeder, Measurement of SAW Velocity versus Strain for YX and ST Quartz, Proc. IEEE Ultrasonics Symp., pp. 519-522, 1975 [5] A. Lonsdale and B. Lonsdale, Method and apparatus for measuring strain, U.S. Patent US5585571, 1996 [6] H. Wohltjen and R. Dessy, Surface Acoustic Wave Probe for Chemical Analysis - Introduction and Instrument Design, Ana. Chem., Vol. 51 (9), pp. 1458-1475, 1979 [7] E. Staples, Electronic Nose Simulation of Olfactory Response Containing 500 Orthogonal Sensors in 10 Seconds, Proc. IEEE Ultrasonics Symp., pp. 417-423, 1999 [8] Passive Radio Requestable SAWWater Content Sensor, L. Reindl, C.C.W Ruppel, A. Kirmayr, N. Stockhausen and M.A. Hilhorst, Proc. IEEE Ultrasonics Symp., pp. 461-466, 1999 [9] Applicability of LiNbO3, Langasite and GaPO4 in high temperature SAW sensors operating at radio frequencies R. Fachberger, G. Bruckner, G. Knoll, R. Hauser, L. Reindl; IEEE Trans. Ultrason. Ferroel. Freq. Control, 51, No.11, pp. 1427-1431, 2004. [10] Wireless SAW based high-temperature measurement systems; R. Fachberger, G. Bruckner, R. Hauser , L. Reindl Proc. of the IEEE Inter. Freq. Contr. Symp. pp. 358-367. [11] An Interrogation Unit for Passive Wireless SAW Sensors Based on Fourier Transform M. Hamsch, R. Hoffmann, W. Buff, M. Binhack, and S. Klett; IEEE Trans.Ultras.Ferroel.& Frequ. Cont., vol. 51, pp 1449 (2004) [12] A second generation in-car tire pressure monitoring system based on wireless passive SAW sensors Dixon, B., Kalinin, V., Beckley, J., Lohr, R. Proceedings of the IEEE International Frequency Control Symposium and Exposition, art. no. 4053792, pp. 374-380 IEEE [13] Application of passive SAW resonant sensors to contactless measurement of the output engine torque in passenger cars Kalinin, V., Lohr, R., Leigh, A., Bown, G. Proceedings of the IEEE International Frequency Control Symposium and Exposition, art. no. 4319124, pp. 499-504 [14] High-Temperature 434 MHz Surface Acoustic Wave Devices Based on GaPO4 M.N. Hamidon, V. Skarda, N.M. White,F. Krispel, P. Krempl, M. Binhack, and W. Buff, IEEE Trans.Ultras.Ferroel.& Frequ. Cont., vol. 53, pp 2465 (2006) [15] Theoretical and experimental study of the differential thermal expansion effect on the TCD of layered SAW temperature sensors. Application to aluminum nitride based layered structures. P. Nicolay, O. Elmazria, M.B. 1ssouar, F. Sarry, L. LeBrizoual; Proceedings of IEEE Ultrasonics Symposium, art. no. 4409652, pp. 272-275 (2007) [16] G. Bernhardt, C. Silvestre, N. LeCursi, S.C. Moulzolf, D.J. Frankel and R.J. Lad "Performance of Zr and Ti adhesion layers for bonding of platinum metallization to sapphire substrates", Sensors and Actuators B 77 (2001), p368-374 [17] Surface acoustic wave sensors for high temperature applications, J. Hornsteiner, E. Born, G. Fischerauer and E. Riha 1998 International Frequency Control Symposium [18] High Temperature Stability of LiNbO3 Based SAW Devices; R. Hauser, L. Reindl and J. Biniasch, 2003 IEEE Ultrasonics Symposium, p192-195 [19] Materials temperature piezoelectric transducers; D. Damjanovic ", Current Opinion in Solid State and Materials Science" 1998,3, p 469-473 [20] Langasite for high temperature surface acoustic wave applications; J. Hornsteiner, E. Born and E. Riha Phys. Stat. Sol. (a) 163, R3 (1997) [21] High Temperature LGS SAW devices with Pt/WO3 and Pd Sensing Films; J. Thiele and M.P. Da Cunha "2003 IEEE Ultrasonics Symposium, p 1750-1753 [22] Experimental and theoretical analysis of SAW properties of the Langasite Substrate with Euler Angle (0°,140°, f); A. Bungo, C. Jian, K. Yamaguchi, Y. Sawada, R. Kimura, S. Uda, Proc. IEEE Ultrasonics Symp., 1999, pp 231-234 [23] Recent measurements of material constants versus temperature for langatate, langanite and langasite; D.C. Malocha, M.P. Da Cunha, E. Adler, R.C. Smythe, S. Frederick, M. Chou, R. Helmbold, Y.S. Zhou, "Proc. of the Annual IEEE International Frequency Control Symposium, 2000, pp 200-205 [24] Investigation on SAW Properties of LGS and Optimal Cuts for High-Temperature Applications ; XiaoJun Ji, Tao IEEE Han, WenKang Shi, and GuoWei Zhang for high Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, Vol. 52, NO.11, November 2005, p 2075-2080 [25] Extraction of COM parameters on Pt/LGS for high temperature SAW sensor; T. Aubert, F. Sarry, O. Elmazria, L. Bouvot, B. Assouar, P. Nicolay and M. Hehn Proc. of 2008 IEEE Ultrasonics International Sympo; p 820-823 [26] Surface-Wave Devices for Signal Processing; D. P. Morgan, Elsevier Editions, 1991 [27] Langasite Surface Acoustic Wave Sensors For High Temperatures; M. Honal, R. Fachberger, T. Holzheu, E. Riha, E. Born, P. Pongratz and A. Bausewein 2000 IEEE/EIA International Frequency Control Symposium and Exhibition, p 113-118 [28] SAW Resonators at High Temperatures; W. Buff, M. Binhack, S. Klett, M. Hamsch, R. Hoffmann, F. Krispel, W. Wallnöfer, 2003 IEEE Ultrasonics Symposium, p187- 191 [29] Properties of radio frequency Rayleigh waves on Langasite at elevated temperatures; R. Fachberger, G. Bruckner, R. Hauser, C. Ruppel, J. Biniasch and L. Reindl 2004 IEEE Ultrasonics Symposium, p 1223-1226 [30] Platinum and Palladium High Temperature Transducers on Langasite; J. Thiele and M.P. Da Cunha IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, Vol. 52, NO. 4, April 2005, p545-549 [31] Enabling very High température acoustic wave devices for sensor and frequency control applications; M. Pereira da Cunha, T. Moonlight, R. J. Lad, G. Bernhardt and D.J. Frankel Proc. of 2007 IEEE Ultrasonics International Symposium; p 2107-2110 [32] SAW-Relevant material properties of Langasite in the temperature range from 25 to 750 °C : new experimental results; I. Shrena, D. Eisele, E. Mayer, L. Reindl, J. Bardong and M. Schmitt, Proc. of 2008 IEEE Ultrasonics International Symposium, p 209-212 [33] Precise measurements of BAW and SAW properties of Langasite in the range from 25°C to 1000°C; J. Bardong, M. Schulz, M. Schmitt, I. Shrena, D. Eisele, E. Mayer, L. Reindl and H. Fritze Proc. of 2008 IEEE Ultrasonics International Symposium ; p 326-331 [34] New measurement method characterise piezoelectric SAW substrates at very high temperature; P. Nicolay, O. Elmazria, F. Sarry, T. Aubert, L. Bouvot and M. Hehn ", Proc. of 2008 IEEE Ultrasonics International Sympo; p 1877-1880 [35] High langasite surface acoustic wave devices; M. Pereira da Cunha, R. J. Lad, T. Moonlight, G. Bernhardt and D.J. Frankel "Proc. of 2008 IEEE Ultrasonics International Sympo. ; p 205-208 [36] Gallium Phosphate, GaPO4 : a new piezoelectric crystal material for high-temperature sensorics; P. W. Krempl, G. Schleinzer and W. Wallnöfer "Sensors Actuators A 1997, 61, p 361-363 [37] M. N. Hamidon, V. Skarda, N. M. White, F. Krispel, P. Krempl, M. Binhack and W. Buff "Fabrication of high temperature stability of to temperature surface acoustic wave devices for sensor applications", Sensors and Actuators A 123-124 (2005), p 403-407 [38] Temperature sensors based on GaPO4; P. W. Krempl, C. Reiter, W. Wallnöffer Proc. of 2002 IEEE Ultrasonics International Symposium, p 949-952 [39] Distinct orientation of AlN thin films deposited on sapphire substrates by laser ablation; J. Meinschien, F. Falk, R. Hergt and H. Stafast "Appl. Phys. A70 (2000), p 215-218 [40] Electrical properties of Piezoelectric AlN Films Deposited by DC Magnetron Sputtering; O. Elmazria, M.B. Asouar, P. Renard and P. Alnot, Physica Status Solidi (a) 196, No. 2, 416 -- 421 (2003); z [41] Sound velocity of AlxGa1-xN thin films obtained by surface acoustic-wave measurements ; C. Deger, E. Born, H. Angerer, O. Ambacher, M. Stutzmann, J. Hornsteiner, E. Riha, and G. Fischerauer, Appl. Phys. Lett., Vol. 72, No. 19, 11 May 1998, p 2400-2402 [42] 5GHz SAW devices based on AlN/diamond layered structure realized using e-beam lithography , P. Kirsch, M. B. Assouar, O. Elmazria, V. Mortet, and P. Alnot Appl. Phys. Lett. 88, (2006) p 223504-1 223504-3 [43] Influence of nitrogen incorporation on the electrical properties of MPCVD diamond films growth in CH4-CO2- N-2 and CH4-H-2-N-2 gas mixtures; Elmazria O, Bougdira J, Chatei H, M. Remy, Alnot, P; Thin Solid Films; Vol. 374, (2002) pp 27-33 [44] Zero-Temperature Coefficient SAW Devices on AlN Epitaxial Films; K. Tsubouchi and N. Mikoshiba "IEEE Transactions on Sonics and Ultrasonics, Vol. SU-32, NO. 5, September 1985, p 634-644 [45] Fabrication of 5-GHz-Band SAW Filter with Atomically-Flat-Surface AlN on Sapphire; K. Uehara, C. M. Yang, T. Shibata, S. K. Kim, S. Kameda, H. Nakase and K. Tsubouchi ", 2004 IEEE Ultrasonics Symposium, p 203-206 [46] Low Temperature AlN Thin Films Growth for Layered Stucture SAW and BAW Devices; M. B. Assouar, O. Elmazria, M. El Hakiki, P. Alnot and C. Tiusan "2004 IEEE International Symposium on Applications of Ferroelectrics, p 43-46 [47] High-temperature coatings; National academy of sciences/National academy of engineering, 1970 [48] A crystallographic model of (001) aluminium nitride expitaxial thin film growth on (001) sapphire substrate; C. J. Sun, P. Kung, A. Saxler, H. Ohsato, K. Haritos and M. Razeghi J. Appl. Phys. 75 (8), 15 April 1994, p 3964-3967 [49] Theoretical and experimental investigation of gigahertz-band, temperature-compensated electromechanical coupling configurations based on AlN films; C. Caliendo, Appl. Phys. Lett. 92, 033505 (2008) oxidation-resistant for frequency, high temperature materials [50] Thermal stability and desorption of Group III nitrides prepared by metal organic chemical vapor deposition; O. Ambacher, M. S. Brandt, R. Dimitrov, T. Metzger, M. Stutzmann, R. A. Fisher, A. Miehr, A. Bergmaier and G. Dollinger, J. Vac. Sci. Technol. B 14(6), Nov/Dec 1996, p 3532-3542 [51] High temperature ultrasonic transducers; N. D. Patel and P. S. Nicholson; Ndt International Volume 23 N° 5 Oct. (1990), p 262-266 [52] High temperature assessment of nitride-based devices; R. Cuerdo, J. Pedrós, A. Navarro, A. F. Braña, J. L. Pau, E. Munõz and F. Calle ; J. Mater. Sci : Mater. Electron. (2008) 19, p189-193 [53] Measurement of output voltage of aluminium nitride thin film as a pressure sensor at high temperature; K. Kishi, Y. Ooishi, H. Noma, E. Ushijima, N. Ueno, M. Akiyama and T. Tabaru, Journal of the European Ceramic Society 26 (2006), p 3425-3430 [54] Oxidation behaviour of AlN films at high temperature under controlled atmosphere; C-Y. Lin and F-H. Lu "J. Eur. Ceram. Soc., 2008, 28, 691-698 [55] High thin-film thermocouples on silicon wafers; K. G. Kreider, G. Gillen; Thin Solid Films 376 (2000), p32-37 [56] Oxidation of the platinum-group metals; C. A. Krier and R. I. Jaffee J. Less-Common Metals, 5 (1963), p 411- 431 [57] Reactions of oxygen with the platinum metals -- II; J. C. Chaston; Platinum Metals Rev., 1965, 9, (2), p 51-56 [58] Stability of thin platinum films implemented in high- temperature microdevices; R. M. Tiggelaar, R. G. P. Sanders, A. W. Groenland, J. G. E. Gardeniers Sens. Actuators A: Phys. (2009), doi : 10.1016/j.sna.2009.03.017 [59] silicon microreactors; C. Alépée; thesis, École polytechnique fédérale de Lausanne (EPFL), Lausanne, Switzerland, 2000 [60] Acoustic Wave Sensors D. S. Ballantine, R. M. White, S. J. Martin, A. J. Ricco, E. T. Zellers, G.C. Frye and H. Wohltjen, Academic Press, 1997 [61] Handbook of thin film technology; L. I. Maissel and R. Glang, McGraw-Hill, New York, USA, 1970 [62] Stabilized Platinum Electrodes for Ferroelectric Film Deposition using Ti, Ta and Zr Adhesion Layers; T. Maeder, L. Sagalowicz an; d P. Muralt "Jpn. J. Appl. Phys. Vol. 37 (1998), p 2007-2012 [63] Binary alloy phase diagrams; T. B. Massalski 1st edi., vol. 2, 1986, American Society for Metals [64] "Effect of annealing in oxygen atmosphere on morphological and electrical properties of iridium and ruthenium thin films prepared by liquid delivery MOCVD; M. Lisker, T. Hur'yeva, Y. Ritterhaus and E. P. Burte Surface and Coatings Technology 201 (2007) high-temperature Technologies for Ph.D.
1811.11735	1	1811	2018-11-28T18:37:11	Biomimetic potassium selective nanopores	[ "physics.app-ph", "cond-mat.soft" ]	Reproducing the exquisite ion selectivity displayed by biological ion channels in artificial nanopore systems has proven to be one of the most challenging tasks undertaken by the nanopore community, yet a successful achievement of this goal offers immense technological potential. Here we show a strategy to design solid-state nanopores that selectively transport potassium ions and show negligible conductance for sodium ions. The nanopores contain walls decorated with 4'-aminobenzo-18-crown-6 ether and ssDNA molecules located at one pore entrance. The ionic selectivity stems from facilitated transport of potassium ions in the pore region containing crown ethers, while the highly charged ssDNA plays the role of a cation filter. Achieving potassium selectivity in solid-state nanopores opens new avenues toward advanced separation processes, more efficient biosensing technologies and novel biomimetic nanopore systems.	physics.app-ph	physics	Biomimetic potassium selective nanopores Elif Turker Acar, 1,2,#,* Steven F. Buchsbaum,3,#,* Cody Combs,1 Francesco Fornasiero,3* Zuzanna S. Siwy1,4,5,* 1Department of Physics and Astronomy, University of California, Irvine 2Department of Chemistry, Faculty of Engineering, Istanbul University Cerrahpasa, Avcılar-Istanbul, Turkey 3 Lawrence Livermore National Laboratory, Livermore, CA 94550 4Department of Chemistry, 5Department of Biomedical Engineering, University of California, Irvine, CA 92697 Abstract: Reproducing the exquisite ion selectivity displayed by biological ion channels in artificial nanopore systems has proven to be one of the most challenging tasks undertaken by the nanopore community, yet a successful achievement of this goal offers immense technological potential. Here we show a strategy to design solid-state nanopores that selectively transport potassium ions and show negligible conductance for sodium ions. The nanopores contain walls decorated with 4′-aminobenzo-18-crown-6 ether and ssDNA molecules located at one pore entrance. The ionic selectivity stems from facilitated transport of potassium ions in the pore region containing crown ethers, while the highly charged ssDNA plays the role of a cation filter. Achieving potassium selectivity in solid-state nanopores opens new avenues toward advanced separation processes, more efficient biosensing technologies and novel biomimetic nanopore systems. # These authors contributed equally to this work; * [email protected], [email protected], [email protected], [email protected] 1 Since the discovery of biological channels and their importance in physiological processes, scientists have attempted to create robust man-made structures that exhibit transport properties mimicking those of their biological counterparts.(1-5) Responsiveness to external stimuli and exquisite ionic selectivity are two of the most exciting properties for which efficiency remains unmatched by solid-state nanopores.(6) Stimuli in the form of electrical potential modulation, chemical interactions, or mechanical stress can induce so called gated channels to switch between ion conductive and closed states. Biological channels are also frequently able to differentiate between ions of the same charge so that, for example, potassium-selective channels can transport potassium ions thousands times faster than sodium ions.(7-9) Matching biological gating and ion selectivity capabilities in a synthetic nanopore platform could not only enable new sensing technologies but also lay the groundwork for a deeper understanding of ionic and molecular transport at the nanoscale with simple and robust model systems. Gating has been successfully achieved in a number of man-made systems. Current rectification(10, 11) and voltage-responsive pore opening(12-14) were demonstrated in polymer and solid state nanopores having walls modified, for example, with single- stranded DNA (ssDNA), or polymer brushes. Transport properties of nanopores can also be made responsive to pressure(15-17) or molecules present in a solution, thus mimicking the behavior of ligand-gated channels.(1, 18-21) 2 The ability to efficiently transport one monovalent cation over another monovalent cation underpins key cellular processes such as signal propagation in neurons but is very challenging to achieve in artificial nanopores. The literature has shown a few examples of lipid-bilayer-supported synthetic constructs prepared by supramolecular self- assembly to stack crown-ether molecules on top of each other and create a channel.(22-24) These synthetic pores, however, display only modest cation/cation selectivities. In a similar approach, self-assembly of rigid macrocycles formed hydrophobic nanopores in a lipid bilayer, which exhibited excellent selectivity towards protons over potassium ions and no measurable conductance in NaCl or LiCl.(25) Due to their hydrophobic character, however, the macrocycle-based channels exhibited only very low, pS conductivities in ionic strength as high as 4 M. To our knowledge, cation/cation selectivity has not been reproduced yet in solid-state nanopores. This type of nanopore offers the advantages of tunable geometry and surface properties and are far more robust than lipid-bilayer inserted channels. Thus, solid-state nanopores permit exploration into a wider range of electrochemical conditions and easier integration into fluidic devices. In this manuscript, we demonstrate a solid-state nanopore that preferentially conducts potassium ions over sodium ions at concentrations up to 1 M and with selectivities far surpassing those previously reported in any other man-made nanopore platform. The mechanism of ionic selectivity is based on facilitated transport(26) of potassium ions through a nanopore with an opening less than 2 nm, whose walls are decorated with 18- crown-6 ether molecules. In polar solvents such as water, the crown ethers are known 3 to selectively bind and release potassium ions quickly, allowing for their transport.(27, 28) We chose 4′-aminobenzo-18-crown-6 ether due to the presence of an amino group that permits easy attachment to a carboxylated surface.(29) Potassium selectivity is achieved via attachment of crown ether Figure 1a shows the fabrication of single nanopores and the types of chemical modifications the structures were subjected to. The pores used in this study were formed in 30 nm thick films of silicon nitride by dielectric breakdown.(30-32) Transport characteristics of the as-prepared pores were measured in 1 M and 100 mM solutions of KCl and NaCl. The pore walls and the membrane surfaces were then modified with triethoxysilylpropylmaleamic acid (TESPMA), which led to the attachment of carboxyl groups and the reduction of the pore diameter by a few nanometers.(33) Carboxylated nanopores were again tested in 1 M KCl, 100 mM KCl and NaCl solutions, and the recorded I-V curves were used to size the pore diameter by relating the pore resistance with its geometry (Supplementary Information). All reported pore diameters in this work are therefore calculated after TESPMA functionalization. Following carboxylation, nanopores were subjected to one of two modification strategies. (i) The first sub-set of nanopores (n=3) was decorated with 4′-aminobenzo-18-crown-6 using 1-ethyl-3-(3- dimethylaminopropyl) carbodiimide (EDC) coupling chemistry(34) (Figure 1b). (ii) The second sub-set of nanopores (n=6) was modified from one side with ssDNA oligomer, and the other side with the crown ether (Figure 1c). This asymmetric functionalization scheme was motivated by the goal of combining voltage-gated transport and cation/cation selectivity in a single synthetic pore that mimics the structure and double 4 functionality of potassium gated channels.(7, 35) Reported selectivity of our pores towards potassium ions are defined here as the ratio of currents in KCl and NaCl solutions measured at 1 M, 100 mM and 10 mM concentrations with a voltage of +1 V across the membrane. The positive sign of electric potential difference corresponds to a working electrode on the side of the membrane with ssDNA; thus, for positive voltages, cations enter the pore from the DNA side and move towards the pore opening containing crown ethers. Figure 1b shows recordings for a 1 nm wide nanopore modified with crown ether only. These include the ion selectivities before and after each modification step for all studied concentrations and the current-voltage curves in 1 M KCl and 1 M NaCl solutions. Selectivity towards potassium ions is evident only after attachment of crown ether so that the ionic current at +1 V in KCl solutions becomes at least 10 times higher than in NaCl solutions. The pore shows current rectification, thus the selectivity calculated at positive and negative voltages is different (Figure S1). Before implementing the asymmetric functionalization scheme (ii), we tested also the ionic selectivity of a pore subjected to crown modification only from one side. Current recordings for this pore (Figure S2) revealed that a partly modified pore still preferentially conducts potassium ions. DNA plays a role of a cation filter In our pore design inspired by voltage-gated ion channels, grafting of ssDNA was localized to the membrane surface and pore mouth through the selection of a 30-mer ssDNA, which is too large to diffuse inside the nanopore.(36, 37) We expected the high 5 density of negative charges on the DNA to increase the cation concentrations at the pore entrance, thus causing the process of binding/releasing of ions from the crown ether to be the limiting step in the ion transport process. Note that potassium channels in a cell membrane also feature negative surface charges at one entrance, which are believed to increase local ionic concentrations and pore conductance.(35) The highly charged DNA functions as a cation filter, preventing anions from passing through, (38) but does not contribute to the pore selectivity towards potassium ions (Figure S3). Example recordings for a 1 nm wide pore subjected to ssDNA/crown ether modification is shown in Figure 1c. Potassium selectivity is based on facilitated transport We hypothesized that the selectivity observed in the two nanopores shown in Figure 1 is based on facilitated transport of K+ ions,(26) which undergo binding/unbinding to crown ethers on the pore walls. In addition, we speculated that the pore opening after TESPMA (1 nm) is narrow enough to be fully occupied by the crown ether groups after EDC chemistry, thus preventing non-selective 'bulk transport' of sodium ions through the middle section of the pore. Pore diameter is therefore predicted to play a very important role in determining the magnitude of K+ selectivity. Toward validating our hypothesis, we considered six ssDNA/crown-ether modified nanopores as well as three nanopores that contained only crown ether. In Figure 2, we plotted the ratios of ionic currents recorded in KCl and NaCl solutions as a function of pore size determined after the carboxylation step. Clearly, K+ ion selectivity decreases 6 exponentially with the increase of the pore diameter and disappears for pore sizes larger than 3 nm. Importantly, thanks to the combined action of facilitated transport and negligible bulk transport, we could achieve K+/Na+ selectivities up to 84, which surpass those reported in other synthetic nanopores by about one order of magnitude. Another striking finding is the increase of K+ selectivity with salt concentration for the majority of nanopores with diameter below 2 nm, a dependence that contradicts expectations from an electrostatics-based selectivity mechanism.(39) In order to understand this observation, we looked in detail at the concentration dependence of sodium and potassium currents (Figure S4). Sodium currents in < 2 nm nanopores are nearly concentration independent, which suggests that the nanopores are indeed too small to allow for significant non-selective transport of sodium through the pore middle region. Potassium currents on the other hand do depend on the bulk concentration, thus the ratio of currents in potassium and sodium is lower at lower concentrations. Figure 2 also offers comparison between pores modified only with crown ethers and those with both DNA and crown ether functionalities. Even though both types of nanopores can exhibit ratios of currents in KCl and NaCl solutions > 78, the crown- ether-only modified pores lose their selectivity more rapidly with the increase of pore size. This observation supports our prediction that DNA would act as a cation selectivity filter and/or contribute to further narrow the size of any non-selective transport pathway at the pore center. Overall, our finding suggests a larger robustness of the potassium selectivity in the presence of DNA. 7 Like the pores modified only with crown ether, pores functionalized with both ssDNA and crown ether displayed current rectification and different selectivities with voltage sign. The difference in K+/Na+ selectivity between positive and negative voltages can be explained considering that crown ethers acquire a positive charge in the presence of cations.38 Consequently, a DNA/crown ether nanopore system features oppositely charged membrane surfaces, which can give rise to a diode behavior.(40) Continuum modeling with the Poisson-Nernst-Planck equations shows that for positive voltages ionic concentrations in the pore are higher, and that the electric potential decays nearly linearly across the entire pore length (Figures, S5, S6). Negative voltages on the other hand cause the formation of a depletion zone with a localized voltage drop. Thus, the resulting weak voltage gradient in the crown-ether-modified region induces a smaller selectivity. Finally, we looked at the voltage dependence of potassium ions selectivity. All pores examined exhibited a selectivity increase at larger positive voltages, the magnitude of which is strongly dependent on pore size (Figure 3). The sensitivity of selectivity to voltage was maximized at a pore diameter of ~1 nm and decreased for both larger and smaller pores. Phenomenological model with voltage-dependent kon and koff help explain potassium selectivity of nanopores. 8 Toward validating the claim that the pore behavior discussed thus far can be explained by crown-ether mediated transport, we developed a simple model, which neglects the presence of DNA and any surface charge effects. We assumed a pore with walls decorated by crown ethers arranged in a ring, which interact with cations through binding/unbinding events governed by kon and koff rate constants. For simplicity, we focused only on transport through a single crown-ether layer (Figure 3a). Both kon and koff constants depend exponentially on voltage, V, according to: 𝑘(cid:3036) = 𝑘(cid:3036)(cid:2868)𝑒 (cid:2879) (cid:3279)(cid:3278)(cid:3280)(cid:3280)(cid:3271) (cid:3261)(cid:3291)(cid:3286)(cid:3251)(cid:3269), i=on, off, where (cid:3031)(cid:3278)(cid:3280) (cid:3013)(cid:3291) corresponds to the fractional potential drop an ion experiences as it approaches and binds to a crown ether, Lp is the pore length, e the unit charge, and kB the Boltzmann constant.(20, 41, 42) The total current in the crown ether region can then be calculated from the total time,  an ion takes to pass through the pore, 𝜏 = (cid:2869) (cid:3038)(cid:3290)(cid:3289)(cid:3004) + (cid:2869) (cid:3038)(cid:3290)(cid:3281)(cid:3281) , the charge density calculated from the percent of bound cation/crown ether complexes, the ion concentration C, and the pore radius, Rp. To approximate a leakage current, the model assumes bulk transport in the center of the pore if Rp > Rmod, where Rmod is the constant radial thickness of the crown-ether-modified region. We set the pore length Lp equal to 30 nm to match the experimental system. The model was then fit to the combined experimental dataset of current versus voltage for the nanopore shown in Figure 1c, and selectivity versus pore diameter for all ssDNA/crown ether modified pores (Figure 2a) while keeping Rmod constant at 0.25 nm. Best fit values of parameters are summarized in Table 1. Fitted dce suggests a reasonable spacing between crown ethers of one to a few nm, while the binding constants, especially for K+, differ significantly from those seen in a bulk system.(27) While at first unexpected, this result 9 appears to be supported by previously reported theories that binding/unbinding kinetics of close packed macrocycles under nanoscale confinement can deviate substantially from bulk behavior.(23, 24) Table 1. Values of parameters used to fit the experimental data of potassium selectivity and ion current values (Figures 2,3) 𝒌𝒐𝒇𝒇𝟎,𝑲 𝒌𝒐𝒏𝟎,𝑲 𝒌𝒐𝒇𝒇𝟎,𝑵𝒂 𝒌𝒐𝒏𝟎,𝑵𝒂 dce 9.9 × 10(cid:2876) s-1 5.1 × 10(cid:2877) s-1M-1 9.2 × 10(cid:2874) s-1 1.1 × 10(cid:2876) s-1M-1 1.9 nm The model is able to successfully capture several key trends observed in the experimental data, the first of which is the diameter dependence of selectivity versus pore diameter (Figure 2a). As expected, the selectivity decreases for large pores due to leakage current while plateauing at a maximum for small pore diameters where the leakage current becomes negligible. Next, the model reproduces well the exponential current dependence on voltage along with reasonably accurate current magnitudes (Figure 3c,d). Our set of equations also replicate the experimentally observed weaker concentration dependence for NaCl versus KCl current, which we attribute to Na+ transport being limited by the concentration independent koff. Finally, the model clearly portrays how the voltage dependence of the K+/Na+ selectivity changes with pore diameter (Figure 3d). At small enough pore size, the leakage current drops to zero and the selectivity becomes voltage independent since both K+ and Na+ binding rates respond to voltage similarly. As the pore diameter gets too large, the leakage current begins to dominate and, thus, the selectivity becomes less sensitive to voltage. Overall 10 the ability of this simple model to reproduce all observed trends in the data strongly supports the proposed facilitated transport mechanism, and this simple set of equations may prove to be useful in the design of future biomimetic pores. Conclusions In conclusion, we presented a solid-state nanopore system decorated with crown ethers, which render the pores highly selective toward potassium ions when the pore diameter is sufficiently small. We found that placement of ssDNA at one pore entrance made the system selectivity more robust by concentrating cations at one entrance. Future studies will focus on equipping such a potassium selective pore with a voltage- gated component to enable voltage-regulated pore opening. Acknowledgments This work was funded by NSF (CBET-1803262), and by The Scientific and Technological Research Council of Turkey (TUBITAK), 2219-International Postdoctoral Research Fellowship Program (App. No: 1059B191600613). Elif T. Acar also acknowledges the Istanbul University Cerrahpasa, Engineering Faculty Chemistry Department for additional financial support. Steven F. Buchsbaum and Francesco Fornasiero acknowledge support from the Laboratory Directed Research and Development Program at LLNL under project tracking code 18-LW-057. Work at LLNL was performed under the auspices of the US Department of Energy under contract DE- AC52-07NA27344. We are grateful to Prof. Mark S.P. Sansom and Prof. Stephen J. Tucker from the University of Oxford for discussions. 11 Figure 1. Designing potassium selective solid-state nanopores. (a) Single nanopores with tunable opening diameter were created in 30 nm thick silicon nitride films by the process of dielectric breakdown. The first modification step led to the attachment of carboxyl groups. The second modification involved either symmetric attachment of 4′-aminobenzo-18-crown-6 ether (b), or asymmetric modification with the crown ether and ssDNA (c). (b) Current-voltage curves in 1 M KCl and 1 M NaCl recorded for a 1 nm in diameter pore whose walls were decorated with crown ether, as shown in the scheme. The graph on the right summarizes ratios of currents in KCl and NaCl at 1 V before and after each modification step for the same nanopore. Ratios of currents for the nanopore before and after carboxylation are calculated based on the recordings in 100 mM of the salts. (c) Current-voltage curves in 1 M KCl and 1 M NaCl for a 0.6 nm wide 12 nanopore modified with crown ether and ssDNA. Selectivity of the nanopore is shown as ratios of ionic currents in KCl and NaCl solutions measured at the same conditions as in (b). 13 Figure 2. Selectivity of nanopores towards potassium. (a) Experimental ratios of ion currents in KCl and NaCl solutions for 6 independently prepared nanopores subjected to chemical modification with crown ethers and ssDNA. Data for three different bulk concentrations of the salts are shown. The model fit is shown as dashed lines. (b) Experimental data of potassium selectivity for three nanopores modified only with crown ether. Standard deviations of currents for individual voltages are shown in I-V curves in Figures 1, S4. 14 Figure 3. Phenomenological model of the potassium selectivity of solid-state nanopores. (a) Scheme of the modeled system with geometrical parameters used in the model. (b) Diameter dependence of the selectivity sensitivity (mS) to voltage. mS is defined here as the slope of a linear fit of Log(Ik/INa) versus voltage for 1 M KCl and NaCl solution concentrations. (c) Current versus voltage curves at three different bulk KCl concentrations for the same pore shown in Figure 1c. (d) Current versus voltage curves at three different bulk NaCl concentraitons for the same pore as (c). Symbols are for experimental data, while dashed lines represent model predictions using the parameters listed in Table 1. Standard deviations of currents for individual voltages are shown in I-V curves in Figure S4. 15 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. W. Guo, Y. Tian, L. Jiang, Asymmetric Ion Transport through Ion-Channel-Mimetic Solid-State Nanopores. Accounts Chem Res 46, 2834-2846 (2013). M. Barboiu, A. Gilles, From Natural to Bioassisted and Biomimetic Artificial Water Channel Systems. Accounts Chem Res 46, 2814-2823 (2013). X. Hou, W. Guo, L. Jiang, Biomimetic smart nanopores and nanochannels. Chem Soc Rev 40, 2385-2401 (2011). S. W. Kowalczyk, T. R. Blosser, C. Dekker, Biomimetic nanopores: learning from and about nature. Trends Biotechnol 29, 607-614 (2011). M. Tagliazucchi, I. Szleifer, Transport mechanisms in nanopores and nanochannels: can we mimic nature? Mater Today 18, 131-142 (2015). B. Hille, Ion Channels of Excitable Membranes. (2001). R. MacKinnon, Potassium channels and the atomic basis of selective ion conduction (Nobel lecture). Angew Chem Int Edit 43, 4265-4277 (2004). Y. Zhou, J. H. Morais-Cabral, A. Kaufman, R. MacKinnon, Chemistry of ion coordination and hydration revealed by a K+ channel -- Fab complex at 2.0 Å resolution. Nature 414, 43 (2001). E. Gouaux, R. MacKinnon, Principles of selective ion transport in channels and pumps. Science 310, 1461-1465 (2005). Z. S. Siwy, S. Howorka, Engineered voltage-responsive nanopores (vol 39, pg 1115, 2010). Chem Soc Rev 39, 5067-5067 (2010). Z. S. Siwy, Ion-current rectification in nanopores and nanotubes with broken symmetry. Adv Funct Mater 16, 735-746 (2006). S. F. Buchsbaum, G. Nguyen, S. Howorka, Z. S. Siwy, DNA-Modified Polymer Pores Allow pH- and Voltage-Gated Control of Channel Flux. J Am Chem Soc 136, 9902-9905 (2014). C. C. Harrell, P. Kohli, Z. Siwy, C. R. Martin, DNA−Nanotube Ar(cid:415)ficial Ion Channels. J Am Chem Soc 126, 15646-15647 (2004). F. Xia et al., Gating of Single Synthetic Nanopores by Proton-Driven DNA Molecular Motors. J Am Chem Soc 130, 8345-8350 (2008). S. N. Smirnov, I. V. Vlassiouk, N. V. Lavrik, Voltage-Gated Hydrophobic Nanopores. Acs Nano 5, 7453-7461 (2011). M. R. Powell, L. Cleary, M. Davenport, K. J. Shea, Z. S. Siwy, Electric-field-induced wetting and dewetting in single hydrophobic nanopores. Nat Nanotechnol 6, 798-802 (2011). J. L. Trick, P. Aryal, S. J. Tucker, M. S. P. Sansom, Molecular simulation studies of hydrophobic gating in nanopores and ion channels. Biochem Soc T 43, 146-150 (2015). Z. Siwy et al., Protein biosensors based on biofunctionalized conical gold nanotubes. J Am Chem Soc 127, 5000-5001 (2005). H. C. Zhang, Y. Tian, L. Jiang, Fundamental studies and practical applications of bio-inspired smart solid-state nanopores and nanochannels. Nano Today 11, 61-81 (2016). R. Wei, V. Gatterdam, R. Wieneke, R. Tampé, U. Rant, Stochastic sensing of proteins with receptor-modified solid-state nanopores. Nat Nanotechnol 7, 257 (2012). S. Howorka, S. Cheley, H. Bayley, Sequence-specific detection of individual DNA strands using engineered nanopores. Nature Biotechnology 19, 636 (2001). A. J. Wright, S. E. Matthews, W. B. Fischer, P. D. Beer, Novel resorcin[4]arenes as potassium- selective ion-channel and transporter mimics. Chem-Eur J 7, 3474-3481 (2001). 16 Z. H. Sun, M. Barboiu, Y. M. Legrand, E. Petit, A. Rotaru, Highly Selective Artificial Cholesteryl Crown Ether K+- Channels. Angew Chem Int Edit 54, 14473-14477 (2015). C. H. Ren, J. Shen, H. Q. Zeng, Combinatorial Evolution of Fast-Conducting Highly Selective K+- Channels via Modularly Tunable Directional Assembly of Crown Ethers. J Am Chem Soc 139, 12338-12341 (2017). X. B. Zhou et al., Self-assembling subnanometer pores with unusual mass-transport properties. Nat Commun 3, (2012). P. H. Vonhippel, O. G. Berg, Facilitated Target Location in Biological-Systems. J Biol Chem 264, 675-678 (1989). R. M. Izatt, J. S. Bradshaw, S. A. Nielsen, J. D. Lamb, J. J. Christensen, Thermodynamic and Kinetic Data for Cation Macrocycle Interaction. Chem Rev 85, 271-339 (1985). R. M. Izatt, K. Pawlak, J. S. Bradshaw, R. L. Bruening, Thermodynamic and Kinetic Data for Macrocycle Interaction with Cations, Anions, and Neutral Molecules. Chem Rev 95, 2529-2586 (1995). G. T. Hermanson, Bioconjugate Techniques, 3rd Edition. Bioconjugate Techniques, 3rd Edition, 1- 1146 (2013). H. Kwok, K. Briggs, V. Tabard-Cossa, Nanopore Fabrication by Controlled Dielectric Breakdown. Plos One 9, (2014). I. Yanagi, K. Fujisaki, H. Hamamura, K. Takeda, Thickness-dependent dielectric breakdown and nanopore creation on sub-10-nm-thick SiN membranes in solution. J Appl Phys 121, (2017). E. T. Acar, P. Hinkle, Z. S. Siwy, Concentration-Polarization-Induced Precipitation and Ionic Current Oscillations with Tunable Frequency. J Phys Chem C 122, 3648-3654 (2018). Y. Kurihara et al., Fabrication of Carboxylated Silicon Nitride Sensor Chips for Detection of Antigen-Antibody Reaction Using Microfluidic Reflectometric Interference Spectroscopy. Langmuir 28, 13609-13615 (2012). Bioconjugate Techniques 2nd Edition Preface to the First Edition. Bioconjugate Techniques, 2nd Edition, Xxvi-+ (2008). D. A. Doyle et al., The structure of the potassium channel: Molecular basis of K+ conduction and selectivity. Science 280, 69-77 (1998). A. Y. L. Sim, J. Lipfert, D. Herschlag, S. Doniach, Salt dependence of the radius of gyration and flexibility of single-stranded DNA in solution probed by small-angle x-ray scattering. Phys Rev E 86, (2012). Scaling Concepts in Polymer Physics. (P.G. de Gennes). G. Perez-Mitta et al., Host-guest supramolecular chemistry in solid-state nanopores: potassium- driven modulation of ionic transport in nanofluidic diodes. Nanoscale 7, 15594-15598 (2015). R. B. Schoch, J. Y. Han, P. Renaud, Transport phenomena in nanofluidics. Rev Mod Phys 80, 839- 883 (2008). I. Vlassiouk, S. Smirnov, Z. Siwy, Nanofluidic Ionic Diodes. Comparison of Analytical and Numerical Solutions. Acs Nano 2, 1589-1602 (2008). B. Hornblower et al., Single-molecule analysis of DNA-protein complexes using nanopores. Nat Methods 4, 315 (2007). A. M. Woodhull, Ionic Blockage of Sodium Channels in Nerve. J Gen Physiol 61, 687-708 (1973). 17 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42.
1906.05487	2	1906	2019-08-05T05:26:37	Full spatiotemporal control of laser-excited periodic surface deformations	[ "physics.app-ph", "cond-mat.mtrl-sci", "physics.optics" ]	We demonstrate full control of acoustic and thermal periodic deformations at solid surfaces down to sub-nanosecond time scales and few-micrometer length scales via independent variation of the temporal and spatial phase of two optical transient grating (TG) excitations. For this purpose, we introduce an experimental setup that exerts control of the spatial phase of subsequent time-delayed TG excitations depending on their polarization state. Specific exemplary coherent control cases are discussed theoretically and corresponding experimental data are presented in which time-resolved x-ray reflectivity measures the spatiotemporal surface distortion of nanolayered heterostructures. Finally, we discuss examples where the application of our method may enable the control of functional material properties via tailored spatiotemporal strain fields.	physics.app-ph	physics	Full spatiotemporal control of laser-excited periodic surface deformations J.-E. Pudell,1 M. Sander,2 R. Bauer,3 M. Bargheer,1, 4 M. Herzog,1 and P. Gaal3 1Institut fur Physik und Astronomie, Universitat Potsdam, 14476 Potsdam, Germany 2European Synchrotron Radiation Facility ESRF, 23800 Grenoble, France 3Institut fur Nanostruktur und Festkorperphysik, Universitat Hamburg, 20355 Hamburg, Germany 4Helmholtz-Zentrum Berlin fur Materialien und Energie GmbH, Wilhelm-Conrad-Rontgen Campus, BESSY II, 12489 Berlin, Germany (Dated: August 6, 2019) We demonstrate full control of acoustic and thermal periodic deformations at solid surfaces down to sub-nanosecond time scales and few-micrometer length scales via independent variation of the temporal and spatial phase of two optical transient grating (TG) excitations. For this purpose, we introduce an experimental setup that exerts control of the spatial phase of subsequent time-delayed TG excitations depending on their polarization state. Specific exemplary coherent control cases are discussed theoretically and corresponding experimental data are presented in which time-resolved x-ray reflectivity measures the spatiotemporal surface distortion of nanolayered heterostructures. Fi- nally, we discuss examples where the application of our method may enable the control of functional material properties via tailored spatiotemporal strain fields. I. INTRODUCTION Ultrashort strain pulses can be generated by absorp- tion of femtosecond or picosecond optical light pulses in solids and nanostructures [1, 2]. This photoacoustic gen- eration is employed to study properties of phonons in solids [3, 4] or the interaction of lattice strain with op- tic [5], electronic [6] or magnetic [7, 8] degrees of free- dom. The various interaction channels suggest that lat- tice strain may be used as a functional tool to control and trigger specific processes and functions in materi- als. Recently, the control and enhancement of quan- tum entanglement using tailored surface acoustic waves was suggested [9]. In fact, thanks to their customizable short length and time scales down to few nanometers and picoseconds, respectively, optically generated strain pulses may be particularly suited for selective excita- tion of nanostructures. However, a high level of con- trol of the shape, frequency, lifetime, etc. of lattice strain is necessary before strain pulses can be used as func- tional tools for device operation. The control is typically gained by tailoring the temporal excitation sequence [10 -- 13]. We recently demonstrated strain control on sub- nanosecond timescales by exploiting spatial variation of transient grating excitation sequences [14]. This method is not only limited to the control of coherent strain pulses but also applies to thermal deformations. In particular, the control of thermal deformations can be applied on the same timescale as the control of coherent excitations. Hence, thermal strain, which is often regarded as unde- sired side effect to optical excitations, can now be used to trigger specific material functions. In this article we present a comprehensive discussion of spatiotemporal control of acoustic and thermal excita- tions in solids. Our method relies on shaping the tempo- ral and spatial sequence of optical excitations using the so-called transient grating (TG) technique. In particular, our experimental TG setup allows for easy tuning of the relative spatial phase of subsequent TG excitations. The periodic surface deformation (PSD) of the sample upon optical excitation is detected using time-resolved x-ray reflectivity (TR-XRR) [15]. The article provides a de- tailed discussion of spatiotemporal coherent control with particular focus on important limiting cases in section II. Also the quantitative probing of PSDs by diffraction of x-rays in x-ray reflection geometry is briefly explained. The optical setup for generating and controlling PSDs is presented in section III. In section IV we discuss re- sults of spatiotemporal coherent control measurements on thermal and acoustic PSDs in nanoscopic heterostruc- tures. We analyze and decompose the experimental data by comparison to an empirical modeling. Finally, sec- tion VI summarizes and emphasizes the main results. II. COHERENT CONTROL OF PERIODIC SURFACE DEFORMATIONS Classical coherent control can be performed on any oscillator or wave-like harmonic excitation due to the superposition principle. Without restriction of the gen- eral validity, we restrict our considerations to an impul- sive and displacive excitation of modes by an ultrashort excitation pulse [16], i.e., the dynamics of the involved modes is much slower than the excitation pulse duration and the coherent oscillation occurs around a displaced equilibrium. The oscillation amplitude of such an im- pulsively excited harmonic oscillator can either be sup- pressed or doubled by a second identical excitation with a relative time delay τ of a half-period (relative phase φ = π) or a full period (φ = 2π), respectively. This type of coherent control has been successfully applied to2 var- ious phenomena such as phonons [17 -- 20], magnons [21 -- 23], phonon-polaritons [24], and surface acoustic waves (SAWs) [14, 25 -- 27]. In displacive excitations, coherent control is limited to the oscillatory motion, i.e., the co- herent part of the system response. The displacement from the equilibrium ground state, which for optical ex- citations corresponds to a heating of the sample due to the absorbed optical energy, is typically ignored in the co- herent control experiments although it may contain most of the absorbed energy[12]. In addition to the temporal coherence exploited in ear- lier coherent control experiments, an impulsive TG exci- tation also possesses a spatial coherence in form of a si- nusoidal intensity variation with spatial period Λ which is typically oriented parallel to the sample surface. As- suming a linear response of the sample, the absorption of optical energy thus generates a spatially periodic energy density along the sample surface with periodicity Λ. The depth profile of the absorbed energy density is dictated by either the optical properties of the sample or the sam- ple dimensions. In the following we will consider a ther- moelastic excitation, i.e., the absorbed energy density re- sults in a mechanical stress that eventually gives rise to a thermal transient grating (thTG). Given the impulsive excitation with ultrashort laser pulses, the thermoelastic stress also launches coherent counter-propagating surface acoustic waves (SAW) resulting in a standing SAW in the optically excited area. Both excitations can be as- sociated with the characteristic wavevector (cid:126)q(cid:107) = 2π/Λ. The spatial coherence introduces an additional coherent control coordinate given by the spatial phase of the TG excitation [14]. Hence, a second TG excitation with a spatial phase ϕx relative to the first TG excitation can be employed to control the relative spatial phase φth of the corresponding thTGs. Choosing φth = 2nπ or φth = (2n − 1)π either amplifies or suppresses the thTG, respectively. Here, the relative spatial phase of the TG excitation patterns directly determines the relative spa- tial phase of the thTGs, i.e., φth = ϕx. The coherent SAW, however, is controlled via the spatiotemporal phase φSAW = ϕx + ϕt = ϕx + vSAW(cid:126)q(cid:107)τ , where vSAW is the phase velocity of the excited SAW and τ is the temporal delay between the two TG excitation pulses. Similar to the thTG, the SAW can be amplified (φSAW = 2nπ) or suppressed (φSAW = (2n − 1)π). According to the defi- nition of φSAW, a change in the relative spatial phase ϕx implies an adopted time delay τ if the interference of the SAWs is to be kept unchanged. In summary, we have in- troduced two experimental coherent control coordinates ϕx and ϕt given by the relative spatial phase and the relative time delay of two consecutive TG excitations in order to control the thermal and coherent PSD via the spatiotemporal phases φth and φSAW. In this article, we employ these coherent control co- ordinates to disentangle four extreme cases of coherent control, which are depicted in Fig. 1: A. Constructive interference of thTGs and construc- tive interference of two standing SAWs: ϕx = 0, ϕt = 0 → φth = 0, φSAW = 0. (Fig. 1a)) B. Destructive interference of thTGs and destructive interference of two standing SAWs: ϕx = π, ϕt = 0 → φth = π, φSAW = π. (Fig. 1b)) C. Constructive interference of thTGs and destructive 2 FIG. 1. Spatiotemporal coherent surface control with two tailored excitations: yellow and red dashed lines de- pict quasi-static thTGs of the individual excitations I and II respectively. The gradient of the colored shaded areas repre- sent the temporal evolution of the standing SAW on the thTG over a half-period. Blue lines and blue colored areas repre- sent the combined excitation (excitation I & II): a) case A: constructive interference of two thTGs and constructive inter- ference of two standing SAWs. Note: both excitations have the same spatiotemporal phase, therefore the shaded area is orange. b) case B: destructive interference of two thTGs and destructive interference of two standing SAWs. c) case C: constructive interference of two thTGs and destructive inter- ference of two standing SAWs. Note: both excitations have the same spatial phase and opposite temporal phase, there- fore the shaded area is orange without a color gradient. d) case D: destructive interference of two thTGs and construc- tive interference of two standing SAWs. e) square root of the diffracted intensity from the combined excitation for case A and C; f) for case B and D. Note: for x-rays the square root of the intensity is proportional to the surface modulation am- plitude. interference of two standing SAWs: ϕx = 0, ϕt = π → φth = 0, φSAW = π. (Fig. 1c)) D. Destructive interference of thTGs and constructive interference of two standing SAWs: ϕx = π, ϕt = π → φth = π, φSAW = 2π ≡ 0. (Fig. 1d)) Note that while there is no SAW oscillation present in case B and C, the SAW modulates the surface deforma- tion of a thTG or a flat surface in case A and D, respec- tively, as indicated by the blue shading in Fig. 1a-d). As will be discussed below, TR-XRR can unambiguously dis- tinguish these situations as they will manifest differently excitationIuthuth+uSAWexcitationIIuthuth+uSAWexcitationI&IIuthuth+uSAWa)caseA:ffith=0,ffiSAW=0012surfaceamplitudeu(nm)b)caseB:ffith=ı,ffiSAW=ıc)caseC:ffith=0,ffiSAW=ı051015012surfacex(µm)d)caseD:ffith=ı,ffiSAW=2ı051015e)01230123delayt(ns)pintensity(arb.units)detection:caseAdetection:caseCf)0123detection:caseBdetection:caseDin the experimental data. Note that we assume two iden- tical excitations in accordance with a time-independent thTG. The amplitude of the second TG excitation may of course be adapted to compensate for a potential decay of the first excitation during the time delay τ . In the following we briefly discuss the detection of dy- namics triggered by the TG excitation. Typically, exper- iments employ an optical laser pulse which is diffracted from the excited modes via photothermal or photoelas- tic effects [28, 29]. Alternatively, PSDs can be detected using x-ray diffraction and photoemission electron mi- croscopy techniques [30 -- 32]. We have recently shown that the PSD associated with the excited quasi-static and transient modes may also be probed by TR-XRR [14, 15]. This method is exclusively sensitive to the sur- face displacement and can detect deformations of only few nanometers [14]. Thermally induced expansion can be unambiguously disentangled from coherent elastic ef- fects via the characteristic timescale of the surface de- formation [15], thus yielding a complete picture of the surface dynamics. For the TR-XRR probe the Laue con- dition, i.e., momentum conservation, must be fulfilled (cid:126)k±1 = (cid:126)kin + (cid:126)q⊥ ± (cid:126)q(cid:107) (1) where (cid:126)kin and (cid:126)q⊥ are the wavevector of the incident probe photons and the recoil momentum due to reflec- tion at the surface, respectively. We restrict our con- siderations to first-order diffraction from the PSD al- though diffraction to higher orders is also present even if a perfect sinusoidal PSD is monitored [15]. Note that Eq. (1) is independent of the spatial phase of the TG, i.e., the phase of the PSD cannot be inferred from the diffracted intensity per se. The intensity of the diffracted x-ray probe pulse is proportional to ∆u2(t), i.e., to the squared difference between the minimum and maximum surface deformation[15]. Thus, the square root of the diffracted intensity measures the magnitude of the total ∆uth + ∆uSAW cos(ωt), where ω = vSAWq(cid:107) is the fre- quency of the SAW. surface modulation amplitude, i.e., (cid:112)I(t) = ∆u(t) = First consider the case where a strong thTG is mod- ulated by a SAW with relatively small amplitude, i.e., ∆uSAW < ∆uth. Excluding thermal diffusion within the thTG, ∆u(t) > 0 holds for all times and the mea- sured signal directly reveals the SAW's frequency and its relative phase with respect to the thTG as depicted in Fig. 1e). However, if ∆uSAW > ∆uth the detec- tion signal is altered. The extreme case is shown in Fig. 1d), where the SAW modulates a flat surface (case i.e., ∆uth = 0 (note that uth (cid:54)= 0 may hold). D), Here, the time-dependent intensity is proportional to ∆u2(t) = [∆uSAW cos(ωt)]2 = ∆u2 SAW [1 + cos(2ωt)] /2. Therefore, the diffracted probe intensity shows twice the frequency of the SAW as presented in Fig. 1f). The ef- fect of the thTG is analogous to a spatial local oscillator, which allows inferring the spatiotemporal phase of the coherent SAW from the diffracted probe pulse after TG excitation. 3 III. EXPERIMENTAL METHODS A detailed view of our optical experimental setup is shown in Fig 2a-e). The general layout of the TG setup is described elsewhere[14, 33]. Our particular setup was de- signed for optical pump -- TR-XRR probe measurements at the ID09 beamline at the European Synchrotron Ra- diation Facility (ESRF). Taking into account the spe- cific conditions at the beamline, we optimized the optical setup for small size, stability and tunability. Therefore, we use two 50 mm wide acylindrical lenses with a focal length of 40 mm to image the +1st and −1st diffraction order from a transmission phase mask onto the sample surface. Interference of the +1 and −1 beam at the sam- ple surface results in a sinusoidal modulation of the opti- cal intensity. Grating, lenses and sample were mounted in 4f-geometry. We optionally introduce an additional cylindrical lens with a focal length of 75.6 mm in the per- pendicular plane. This lens is mounted with variable distance to the sample to generate higher excitation flu- ences. To generate two replica of an ultrashort optical pump pulse with perpendicular polarization for coherent con- trol we use a Michelson interferometer [cf. Fig. 2b)] with a polarizing beamsplitter (PBS). The relative intensity of the pulses can be tuned by a half-wave plate (HWP) in front of the PBS. Each arm of the interferometer in- cludes a quarter-wave plate (M-QWP) which, due to the double-passage of the beams, effectively rotates the lin- ear polarization by 90◦. Thus, the output of the inter- ferometer yields one s- and one p-polarized optical pump pulse. The pulses have a variable relative time delay τ that is defined by the difference of the path lengths of the Michelson interferometer arms. Both pulses are sub- sequently coupled into the TG setup where they essen- tially are diffracted into ±1st order beams by the phase mask. Zero-order and higher-order diffraction intensities are minimized by the specific design of the transmission phase mask. In addition to simple TG excitation, our setup allows for selecting the spatial phase of the TG to perform spa- tiotemporal coherent control. We use the fact that the spatial phase of the TG at the sample surface depends on the relative temporal phase of the interfering opti- cal beams. For example, interference gratings generated by two optical beams of either parallel or anti-parallel polarization have opposite spatial phase with respect to one another[34]. Note that in the anti-parallel case the electric fields of +1st and −1st order beams have a rela- tive temporal phase of π, i.e., exactly the same value as the relative spatial phase shift of the generated TG. As explained below, we employ quarter-wave plates (QWP) and different polarizations to impose a relative spatial phase shift between two consecutive TG excitations. After collimation by the first acylindrical lens, each beam propagates through a QWP. The QWPs are ori- ented either with the fast or slow axis aligned with the polarization of the laser pulses. If the QWPs have identi- 4 FIG. 2. Optical setup: a) The output of a 1 kHz laser system is coupled to a Michelson interferometer [b)], where s- and p-polarized replica of the pump pulse are generated. c) shows a side view of the TG-setup. The front view d) shows the beam path of the +1st and −1st diffraction order. A 3D-view d) of the TG-setup shows the opto-mechanical components. In f) and g) the created relative spatial phase of the interference grating for parallel and perpendicular setting of the QWPs are shown. cal orientation, there is no relative temporal phase offset between the ±1st order beams as they both traverse the QWPs at either the fast or the slow axis. Note that both consecutive s- and p-polarized laser pulses generate TGs with identical spatial phase, i.e. ϕx = 0. Hence, with a parallel setting of the QWPs, one can generate cases A and C discussed in section II. If the QWPs are oriented perpendicular to each other, the +1st and −1st order beam experience a relative tem- poral phase shift of ±π/2 which directly translates into a spatial phase offset of ±π/2 of the TG excitations. The opposite sign of the temporal phase shift holds for the s- and p-polarized beams, respectively. The magni- tude of the spatial phase difference between the s- and p- polarized TG is therefore equal to π, i.e. ϕx = π. Hence, with a perpendicular setting of the QWPs, one readily obtains cases B and D discussed in section II where the PSD due to the thTG is relieved by the second TG exci- tation. The experimental results presented and discussed in the next section have been obtained on a 30 nm thick metallic SrRuO3 (SRO) film epitaxially grown by pulsed laser deposition on a (110)-oriented DyScO3 (DSO) sub- strate. The sample was excited with TG excitations each having a spatial period Λ = 2.4 µm and an incident pump fluence of 18 mJ/cm2 for the central fringes of the TG[35]. We employ a commercial Ti:Sapphire laser amplifier (Co- herent Legend Elite) which delivers 800 nm pulses with a duration of 1.2 ps and a pulse energy of 3.6 mJ. The laser repetition rate is 1 kHz synchronized to the synchrotron. The shortest grating period Λ inscribed in the sample is ultimately limited by the laser wavelength. By frequency doubling or tripling of the fundamental frequency one can reduce the period to less then 300 nm. The generation of transient gratings with periods less than 100 nm has been demonstrated by using high-energy radiation from free electron laser sources [36, 37]. Thus, our method can be employed truly on nanometer length scales. Monochromatized 15 keV x-ray probe pulses were se- lected from the synchrotron pulse train by a high-speed chopper at the same frequency. The pump-probe delay can be changed electronically by the laser synchroniza- tion unit. In the experiment presented here, the total temporal resolution was limited to 75 ps mainly due to the duration of the x-ray probe pulses. In principle, the experimental time resolution is also limited by the rather large wavefront tilt between exciting laser and probing x-ray pulses, however, in the present case this is only a minor limitation (approx. 10 ps). Diffracted x-ray photons where captured on an area detector (Ray- onics MX170-HS).[38, 39] For the evaluation the inten- sity I−1st of the −1st diffraction order is integrated in a region of interest on the area detector. The recorded intensity is normalized to a static diffraction background In to reduce influences of beam instabilities and thermal drifts of sample and setup. In order to extract the sur- face modulation amplitude ∆u, we take the square root of the diffracted intensity after subtracting a scattering background Ibg by averaging all unpumped detected in- filasersystem:1kHz1:2ps3:6mJsamplesurfacexIlaserexcitationIIexcitationIsamplesurfacexIlaserexcitationIIexcitationIgratingacylindricallens2xQWPcylindricallensacylindricallenssamplex-rayp-BS2xM-QWPHWPdelaystage5cm4cm3cm2cm1cm0cmspecular−1.order1.ordera)b)c)d)e)f)g)5 The rise time τrise of the thTG is dictated by the ratio of the thickness (or the optical penetration depth if the latter is much shorter) and sound velocity of the laser- excited film. Typical time scales of thin-film expansion are a few tens of picoseconds or even down to a few pi- coseconds for very thin films [40]. In the present case τrise is much shorter than all other involved dynamics and thus approximated by the Heaviside function Θ(t) in Eq. 3. In fact, τrise defines the fundamental limit for coherent control of the thTG (case B and D) which can thus be truly applied down to picosecond time scales as demonstrated in Ref. 15. The concept of spatiotempo- ral coherent control is generally applicable to an arbi- trary number of coherent modes [14, 15, 41], but in the present case the data only exhibit a single Rayleigh-like SAW mode. We can thus restrict our model to only in- clude this single coherent mode. Note that the first TG excitation always defines the zero phase ϕx and ϕt, re- spectively. In order to mimic the sensitivity of the x-ray probe beam to only the modulation of the PSD, we ex- tract the transient wavevector-dependent surface modu- lation amplitude ∆u(q, t) = (Fu)(q, t) by Fourier trans- formation of the spatiotemporal surface deformation field u(x, t) depicted in Fig. 3b). We than evaluate the sur- face modulation amplitude ∆u(q(cid:107), t) at the characteristic wavevector q(cid:107). The dark blue solid line in Fig. 3a) shows the temporal behavior convoluted with the experimental temporal resolution of 75 ps. The slowly decaying signal shown in Fig. 3a) is caused by the PSD associated with the thTG that slowly de- cays due to thermal diffusion. This decay clearly ex- hibits a fast component decaying within the first 2 ns and a much slower component. The latter is due to in- plane thermal diffusion between hot and cold areas of the thTG. An analytical solution of the in-plane ther- mal diffusion for a sinusoidal thermal grating yields the exponential term in Eq. (4) which implies a decay time of (αxq2(cid:107))−1 = 210 ns [42, 43]. Here, the homogeneous in-plane diffusivity αx = 0.8 mm2/s is used for a sub- strate temperature of 323 K [44]. The initial fast decay originates from different thermal expansion coefficients of the metallic SRO film and the insulating DSO sub- strate. In fact, SRO expands stronger upon heating than DSO [45, 46]. Thus, as heat diffuses along the out-of- plane direction from the excited areas in the SRO film into the substrate, the total surface excursion is reduced. In order to verify this, we model the out-of-plane heat transport by solving the one-dimensional heat diffusion equation with a finite-element method [47, 48] by ac- counting for all relevant thermophysical properties of the materials. The experimental data is reproduced with- out including additional thermal resistance at the inter- face due to the nearly perfect acoustic impedance match between the involved materials and the high structural quality of the sample. The simulation yields the sur- face distortion uth,0(t) used as input for Eq. (4). The surface modulation amplitude due to combined out-of- plane and in-plane thermal diffusion is shown as a gray Single-pulse excitation: FIG. 3. a) Time-dependent surface modulation amplitude ∆u(t) measured with time- resolved x-ray reflectivity (TR-XRR). The data (blue bullets) is modeled by calculating the spatiotemporal surface deforma- tion field u(x, t) [c.f. Eq. (3)]. The dark blue solid line shows the time dependence of the Fourier component (Fu)(q(cid:107), t) at the characteristic wavevector q(cid:107). The surface modulation am- plitude of only the thermal distortion (Futh)(q(cid:107), t) due to the thTG is shown by the gray lines with (solid) and without (dashed) in-plane heat diffusion. b) Illustration of the spa- tiotemporal surface deformation field u(x, t) calculated using Eq. (3). tensities of the −1st diffraction order. This results in (cid:115)(cid:12)(cid:12)(cid:12)(cid:12) I−1st − Ibg In (cid:12)(cid:12)(cid:12)(cid:12)sgn (I−1st − Ibg) ∆u ∝ (2) where the absolute function circumvents imaginary re- sults and sgn function projects these values on the nega- tive axis for the surface modulation amplitude ∆u. IV. RESULTS AND DISCUSSION First, we briefly discuss the transient response of the sample surface to a single TG excitation. As derived in earlier studies [15], the surface modulation amplitude ∆u is proportional to the square root of the diffracted intensity. The transient amplitude of the laser-generated PSD inferred from the the x-ray intensity diffracted into −1st order is shown by the blue bullets in Fig. 3a). It features a step-like rise followed by oscillations on top of a slowly decaying thermal offset. The surface excursion field can thus be precisely modeled by u = Θ(t) uth(x, t, ϕx) + uSAW(x, t, ϕx, ϕt) (3) (cid:104) (cid:105) where uth is the slowly decaying amplitude of the thTG and uSAW is the amplitude of the coherent surface acous- tic mode defined as −αxq2(cid:107)t(cid:2)1 + sin(cid:0)q(cid:107)x − ϕx (cid:1)(cid:3) (cid:1) cos(ωt − ϕt) sin(cid:0)q(cid:107)x − ϕx (4) (5) uth,0(t) uth = e uSAW = − uSAW,0 2 2 0.511.52surfaceamplitudeu(arb.units)a)0.511.52∆u(arb.units)singlepulsedatathTGwithoutin-planeheatthTGsinglepulsefitb)01234567810Λdelayt(ns)surfacex6 FIG. 4. Double-pulse excitation: a,e) case A; b,f) case B; c,g) case C; d,h) case D. The single pulse response as modeled in Fig. 3 are shown in gray for comparison in a-d). Symbols show the TR-XRR measurements and dark blue solid lines show the transient surface modulation amplitude ∆u(q(cid:107), t) for the characteristic wavevector q(cid:107) of the spatiotemporal surface deformation field shown in e-h) derived from Eq. (3). All calculated traces are convoluted with a 75 ps Gaussian to match the temporal resolution of the experiment. The orange line in d) represents the best fit to the double pulse data by allowing for an amplitude and phase variation of the second excitation pulses. solid line in Fig. 3a). If in-plane thermal diffusion is ne- glected (αx = 0), the grey dashed line is obtained which proves that the initial fast decay is indeed governed by the out-of-plane thermal transport. The very good agree- ment between experiment and calculation evidences that the presented method can be a powerful tool to inves- tigate multidirectional thermal transport in nanoscopic heterostructures. We do not observe deviations from lin- ear behaviour of our sample even up to very large excita- tion densities [49]. This aspect is particularly important for the multipulse excitation discussed in the next para- graph. With the well-calibrated single-pulse excitation we fi- nally demonstrate full spatiotemporal control of transient and quasi-static PSDs via double-pulse TG excitation. By employing two consecutive TG excitations, we set the spatiotemporal phase of the excitation to the four differ- ent cases introduced in Sec. II. The corresponding dy- namic surface excursion measured by TR-XRR for these four cases is shown in Fig. 4a-d), respectively. As pre- dicted above, the spatial phase setting ϕx = 0, selected by a parallel alignment of the QWPs in the optical setup, results in an increase of the PSD (case A and C) due to an enhancement of the thTG. In contrast, the 90◦ rotation of only one QWP suppresses the PSD and relieves the thTG completely (case B and D). The phase of the SAW is controlled by both the spatial and temporal phase ϕx and ϕt, respectively. For any setting of the spatial phase, one can either suppress (case C and B) or enhance (case A and D) the SAW mode by choosing the correct time delay τ of the second TG excitation. Note in particular the comparison of cases A and B, where τ = 0.73 ns is identical. Still, not only the timing τ alone determines the amplitude of the SAW after the second excitation, as explained in Sec. II. The possibility of suppressing the coherent mode (case B and C) offers a precise tool for investigations of the multidirectional thermal transport in nanoscale het- erostructures without undesired coherent signals yet en- suring sufficient time resolution given by the ultrashort laser and x-ray pulses. In case C we clearly observe the multicomponent relaxation due to in-plane and out-of- plane heat diffusion which was discussed above for single- pulse excitation. However, here, the signal of thermal origin is not masked by the coherent signal. The finite decaying intensity after the second TG excitation in case B (Fig. 4b)) evidences that the thTG was not immedi- ately suppressed by the second excitation. This is caused by the partial decay of the first thTG between the two excitations due to fast out-of-plane heat diffusion which results in the observed imbalance of both thTGs. All observations are accurately reproduced by our modeling 0.511.52surfaceamplitudeu(arb.units)a)caseA0123∆u(arb.units)singlepulsefitdoublepulsedatadoublepulsemodel0:73nse)0Λsurfacexb)caseB0.511.52∆u(arb.units)singlepulsefitdoublepulsedatadoublepulsemodel0:73nsf)0Λsurfacexc)caseC0.511.522.5∆u(arb.units)singlepulsefitdoublepulsedatadoublepulsemodel1:10nsg)0123456780Λdelayt(ns)surfacexd)caseD0.511.52∆u(arb.units)singlepulsefitdoublepulsedatadoublepulsemodeldoublepulsefit1:10nsh)0123456780Λdelayt(ns)surfacexintroduced above. Altogether, the sensitivity to thermal transport in layered heterostructures can be greatly en- hanced with spatiotemporal coherent control using TG excitations. A qualitative difference from the other recorded signals is observed in case D shown in Fig. 4d) where the thTG is suppressed and the SAW is enhanced. Here, we ob- serve a coherent oscillation exhibiting twice the frequency of the excited SAW. Recall that the data represent the variations of the 1st-order diffracted x-ray intensity. If a true second harmonic of the fundamental SAW was present, Eq. (1) implies that the corresponding 1st order of the second harmonic would be diffracted towards larger angles. In other words, the 1st-order diffraction angle cannot contain signatures of a second harmonic SAW. Again, we model the transient diffracted x-ray intensity caused by the spatiotemporal sample surface dynamics in analogy to the single-pulse excitation data analysis [c.f. Eqs. (3-5)]. Note that our model is a purely lin- ear response model and thus does not support higher harmonics of the employed modes. Assuming the time delay τ chosen in the experiment and two identical TG excitations we obtain the dark blue curve in Fig. 4d). Obviously, our modeling correctly yields the doubled fre- quency. As indicated above this is due to the detection process which measures different oscillation frequencies from the same acoustic mode with and without the addi- tional thTG that acts as a spatial local oscillator. Note that the amplitudes of the even and odd oscillation max- ima are different which can be traced back to the heat diffusion dynamics of the first thTG between the two TG excitations breaking the symmetry. As case B re- vealed, the two thTGs converge in amplitude after a few nanoseconds. In case D (dark blue line) this is mani- fested in the equilibration of the even and odd oscillation maxima on the same timescale. However, this equilibra- tion of oscillation amplitudes is not fully featured in the experimental data which indicates an asymmetry in the TG excitation strengths. Also, the level around which the signal oscillates is larger. Indeed, the data are accu- rately reproduced (orange line in Fig. 4d)) if we assume a 9% larger amplitude and a slight detuning of the spatial phase of 3% for the second excitation pulse. Such errors may stem from slight deviations from optimal laser beam and QWP alignments. V. STRAIN CONTROL IN FUNCTIONAL MATERIALS In the following paragraph we outline possible interac- tion channels between the excited deformation and func- tional properties of the crystal. In particular, we de- scribe strain-induced changes of the free energy density via magnetoelastic effects and changes of the electronic band energy via deformation potential coupling. Finally, we will briefly introduce active optical elements which use dynamic strain fields to manipulate x-ray pulses emitted 7 by synchrotron storage rings. If strain is used as a functional tool, it is important to recall, that the interaction of lattice deformations with a material strongly depends on the symmetry, i.e., on the specific component of the strain tensor ε. The strain fields corresponding to thTG and coherent Rayleigh-like SAWs are composed of both compressive/tensile (e.g., εxx, εzz) and shear (e.g., εxz) components. The TR-XRR method detects the absolute surface deformation, i.e., the integrated out-of-plane expansion of the excited volume. However, knowing one component of the strain tensor of a Rayleigh wave allows inferring all other components as well.[50] In our coherent control scheme, all strain com- ponents of the coherent mode are customizable as well as the in-plane and shear components of the thTG. Only the out-of-plane component of the thermal grating is given by the initially absorbed energy density profile. In multiferroic materials a dynamic strain wave modi- fies the free energy density due to elastic deformations of the lattice. As an example, we discuss ferromagnetic ma- terials, where the magnetoelastic interaction modulates the free magnetic energy density fmag.[51] In a static case, fmag is composed of the Zeeman energy, which depends on an external magnetic field, and of static anisotropy components such as magneto-crystalline, shape and mag- netoelastic anisotropy.[52] The interplay of these terms results in a direction and magnitude of the macro- scopic magnetization (cid:126)M . Their dynamics can be induced through time-dependent changes of free magnetic energy fmag. Prominent examples are ferromagnetic resonance (FMR) measurements[53], which act on the Zeeman energy or all-optical switching[54], where laser-induced heating leads to changes of the shape anisotropy. In complete analogy, an acoustic waves dynamically changes fmag via the magnetoelastic energy term[51]. Although magnetoelastic interaction is well-known, magnetoacoustics has only been investigated quite re- cently [8, 55, 56]. Since then, strain-induced magen- tization dynamics of nanoparticles excited specifically by Rayleigh waves, has gained strong interesst [57, 58]. These efforts are driven by the potential of strain-induced dynamics, i.e., energy efficiency, mode selectivity and the ability to tailor the excitation to nanosize dimensions. Several of these recent experiments use optical generation of strain waves, thus pushing magnetoelastic excitations to picosecond timescales [33, 41, 59]. The strain control scheme described in this article not only allows to selec- tively excite magnetization dynamics, but also enables control of theses excitations on picosecond timescales. In particular, this is not only limited to the coherent strain but rather extends to thermal strain, while maintaining the high temporal resolution. The second interaction channel we discuss is defor- mation potential coupling of electrons with acoustic phonons. The deformation of the crystal lattice by an acoustic lattice distortion leads to an energy shift ∆E = aε of the extremal points of the electron bands where a is the deformation potential which typically has a value of about 10 eV at the Γ-point of tetrahedral semiconductors such as Si or GaAs.[60] Hence, already a small dynamic strain of the order of 10−3 up to 10−2 leads to changes of the conduction and valence bands of 10 to 100 meV. Strain-induced changes of the electronic structure affect charge transport and optical properties [61, 62] and allow for control of recombination dynamics in nanostructures [6, 63]. Strain-control of optical properties of nanostruc- tures is a promising candidate for applications in quan- tum computation and quantum information technology [64, 65]. The realization of such applications depends on the ability to control the lattice strain, ideally on short to ultrashort timescales. While the two examples given above may require probing mechanisms other than TR- XRR (e.g. magnetooptical probing, optical and/or x-ray dichroism, or valence spectroscopies), the method pre- sented in this paper may pave the way for these future applications. Finally, we discuss a specific application developed by our group, where strain-induced deformations are used to realize active ultrafast x-ray optics. The devices are optimized for installation at synchrotron beamlines. A prominent example is the picosecond Bragg switch (Pi- coSwitch), which shortens an incident synchrotron x-ray pulse to a duration of few picoseconds [66]. The coherent control of thTGs similar to case B [c.f. section II] allows for controlling diffraction of an incident x-ray pulse into the ±1st diffraction order of the thTG. In particular, our approach allows to turn the diffraction on and off on sub- nanosecond timescales. Thus, thTGs could be employed to pick individual x-ray pulses from a synchrotron pulse train for subsequent pump-probe experiments. Further- more, the device may also be employed as variable beam splitter in order to, e.g., distribute x-ray pulses among multiple beamlines. This may be particularly interesting at x-ray free electron laser (XFEL) facilities, where cur- rently only one experimental station is operational at a time. With such an approach several instruments could be supplied with XFEL pulses in parallel. The main chal- lenge for this device is to achieve high diffraction effi- ciencies. Our previous studies suggest that a maximum 8 efficiency of more than 30% could be reached [14, 15, 67]. VI. CONCLUSION In conclusion, we demonstrated spatiotemporal control of acoustic and thermal deformations of solid surfaces. The optical setup allows for generation of transient sur- face gratings with variable spatial phase. Hence, a ther- mal deformation can either be enhanced or suppressed by a temporal sequence of excitation pulses on timescales much shorter than the deformation lifetime. In addi- tion, we showed that the suppression of the coherent sig- nal facilitates investigations of multidirectional thermal transport in nanolayered heterostructures with high time resolution. We believe that our method presents an im- portant step towards developing strain as functional tool for solids and nanostructures. As examples we discuss the magnetoelastic interaction in ferromagnetic materi- als. While numerous recent studies have demonstrated the ability to manipulate the macroscopic magnetization with coherent strain pulses, our new scheme paves the way for controlled strain-induced preparation of a ferro- magnetic state. Strain control may also be applied to manipulate electronic states in bulk and low-dimensional semiconductors. Finally, we discuss active optical ele- ments, which are a new kind of strain-based devices for ultrafast x-ray beam manipulation at synchrotrons. ACKNOWLEDGEMENTS The TR-XRR experiments were performed at the beamline ID09 of the European Synchrotron Radiation Facility (ESRF), Grenoble, France. We are grateful to Michael Wulff and Norman Kretzschmar for pro- viding assistance in using beamline ID09. We also gratefully acknowledge technical support of Christine Fischer and Elko Hannemann. Finally, we thank Jutta Schwarzkopf from Leibniz-Institut fur Kristallzuchtung, Berlin for providing the sample. We acknowledge the Deutsche Forschungsgemeinschaft for the financial sup- port via BA2281/8-1 and funding from the BMBF via FK05K16GU3. [1] C. Thomsen, H. T. Grahn, H. J. Maris, and J. Tauc, Sur- face generation and detection of phonons by picosecond light pulses, Phys. Rev. B 34, 4129 (1986). [2] P. Ruello and V. E. Gusev, Physical mechanisms of coher- ent acoustic phonons generation by ultrafast laser action, Ultrasonics 56, 21 (2015). [3] R. Shayduk, M. Herzog, A. Bojahr, D. Schick, P. Gaal, W. Leitenberger, H. Navirian, M. Sander, J. Goldshteyn, I. Vrejoiu, and M. Bargheer, Direct time-domain sam- pling of subterahertz coherent acoustic phonon spectra in SrTiO3 using ultrafast x-ray diffraction, Phys. Rev. B 87, 184301 (2013). [4] A. Bojahr, M. Gohlke, W. Leitenberger, J. Pudell, M. Reinhardt, A. von Reppert, M. Roessle, M. Sander, P. Gaal, and M. Bargheer, Second harmonic generation of nanoscale phonon wave packets, Phys. Rev. Lett. 115, 195502 (2015). [5] D. J. Singh, Q. Xu, and K. P. Ong, Strain effects on the band gap and optical properties of perovskite SrSnO3 and BaSnO3, Appl. Phys. Lett. 104, 011910 (2014). [6] M. Weiss, J. B. Kinzel, F. J. R. Schulein, M. Heigl, D. Rudolph, S. Morkotter, M. Doblinger, M. Bichler, G. Abstreiter, J. J. Finley, G. Koblmuller, A. Wixforth, and H. J. Krenner, Dynamic acoustic control of individ- ual optically active quantum dot-like emission centers in heterostructure nanowires, Nano Lett. 14, 2256 (2014). [7] A. V. Scherbakov, A. S. Salasyuk, A. V. Akimov, X. Liu, M. Bombeck, C. Bruggemann, D. R. Yakovlev, V. F. Sapega, J. K. Furdyna, and M. Bayer, Coherent mag- netization precession in ferromagnetic (Ga, Mn)As in- duced by picosecond acoustic pulses, Phys. Rev. Lett. 105, 117204 (2010). [8] J.-W. Kim, M. Vomir, and J.-Y. Bigot, Ultrafast magne- toacoustics in nickel films, Phys. Rev. Lett. 109, 166601 (2012). [9] R. Blattmann, H. J. Krenner, S. Kohler, and P. Hanggi, Entanglement creation in a quantum-dot -- nanocavity sys- tem by fourier-synthesized acoustic pulses, Phys. Rev. A 89, 012327 (2014). [10] K. A. Nelson, R. J. D. Miller, D. R. Lutz, and M. D. Fayer, Optical generation of tunable ultrasonic waves, J. Appl. Phys. 53, 1144 (1982). [11] C. Klieber, E. Peronne, K. Katayama, J. Choi, M. Yam- aguchi, T. Pezeril, and K. A. Nelson, Narrow-band acous- tic attenuation measurements in vitreous silica at fre- quencies between 20 and 40 GHz, Appl. Phys. Lett. 98, 211908 (2011). [12] M. Herzog, A. Bojahr, J. Goldshteyn, W. Leitenberger, I. Vrejoiu, D. Khakhulin, M. Wulff, R. Shayduk, P. Gaal, and M. Bargheer, Detecting optically synthesized quasi- monochromatic sub-terahertz phonon wavepackets by ul- trafast x-ray diffraction, Appl. Phys. Lett. 100, 094101 (2012). [13] F. J. R. Schulein, E. Zallo, P. Atkinson, O. G. Schmidt, R. Trotta, A. Rastelli, A. Wixforth, and H. J. Kren- ner, Fourier synthesis of radiofrequency nanomechanical pulses with different shapes, Nat. Nanotechnol. 10, 512 EP (2015). [14] M. Sander, M. Herzog, J. E. Pudell, M. Bargheer, N. Weinkauf, M. Pedersen, G. Newby, J. Sellmann, J. Schwarzkopf, V. Besse, V. V. Temnov, and P. Gaal, Spatiotemporal coherent control of thermal excitations in solids, Phys. Rev. Lett. 119, 075901 (2017). [15] M. Sander, J.-E. Pudell, M. Herzog, M. Bargheer, R. Bauer, V. Besse, V. Temnov, and P. Gaal, Quanti- tative disentanglement of coherent and incoherent laser- induced surface deformations by time-resolved x-ray re- flectivity, Appl. Phys. Lett. 111, 261903 (2017). [16] M. Bargheer, N. Zhavoronkov, Y. Gritsai, J. C. Woo, D. S. Kim, M. Woerner, and T. Elsaesser, Coherent atomic motions in a nanostructure studied by femtosec- ond x-ray diffraction., Science 306, 1771 (2004). [17] A. M. Lindenberg, I. Kang, S. L. Johnson, R. W. Falcone, P. A. Heimann, Z. Chang, R. W. Lee, and J. S. Wark, Coherent control of phonons probed by time-resolved x- ray diffraction., Opt. Lett. 27, 869 (2002). [18] O. Synnergren, T. N. Hansen, S. Canton, H. Enquist, P. Sondhauss, A. Srivastava, and J. Larsson, Coherent phonon control, Appl. Phys. Lett. 90, 171929 (2007). [19] P. Beaud, S. L. Johnson, A. Streun, R. Abela, D. Abram- sohn, D. Grolimund, F. Krasniqi, T. Schmidt, V. Schlott, and G. Ingold, Spatiotemporal stability of a femtosecond hardx-ray undulator source studied by control of coher- ent optical phonons, Phys. Rev. Lett. 99, 174801 (2007). [20] Y.-H. Cheng, F. Y. Gao, S. W. Teitelbaum, and K. A. Nelson, Coherent control of optical phonons in bismuth, 9 Phys. Rev. B 96, 134302 (2017). [21] Q. Zhang, A. V. Nurmikko, A. Anguelouch, G. Xiao, and A. Gupta, Coherent magnetization rotation and phase control by ultrashort optical pulses in CrO2 thin films, Phys. Rev. Lett. 89, 177402 (2002). [22] T. Kampfrath, A. Sell, G. Klatt, A. Pashkin, S. Mahrlein, T. Dekorsy, M. Wolf, M. Fiebig, A. Leitenstorfer, and R. Huber, Coherent terahertz control of antiferromag- netic spin waves, Nat. Photonics 5, 31 (2010). [23] J. Nishitani, T. Nagashima, and M. Hangyo, Terahertz radiation from antiferromagnetic MnO excited by optical laser pulses, Appl. Phys. Lett. 103, 081907 (2013). [24] D. W. Ward, J. D. Beers, T. Feurer, E. R. Statz, N. S. Stoyanov, and K. A. Nelson, Coherent control of phonon- polaritons in a terahertz resonator fabricated with fem- tosecond laser machining., Opt. Lett. 29, 2671 (2004). [25] D. H. Hurley, R. Lewis, O. B. Wright, and O. Matsuda, Coherent control of gigahertz surface acoustic and bulk phonons using ultrafast optical pulses, Appl. Phys. Lett. 93, 113101 (2008). [26] Q. Li, K. Hoogeboom-Pot, D. Nardi, M. M. Murnane, H. C. Kapteyn, M. E. Siemens, E. H. Anderson, O. Hell- wig, E. Dobisz, B. Gurney, R. Yang, and K. A. Nelson, Generation and control of ultrashort-wavelength two- dimensional surface acoustic waves at nanoscale inter- faces, Phys. Rev. B 85, 195431 (2012). [27] H. F. Yang, F. Garcia-Sanchez, X. K. Hu, S. Sievers, T. Bohnert, J. D. Costa, M. Tarequzzaman, R. Ferreira, M. Bieler, and H. W. Schumacher, Excitation and co- herent control of magnetization dynamics in magnetic tunnel junctions using acoustic pulses, Appl. Phys. Lett. 113, 072403 (2018). [28] A. A. Maznev, A. Mazurenko, L. Zhuoyun, and M. Gostein, Laser-based surface acoustic wave spectrom- eter for industrial applications, Rev. Sci. Instrum. 74, 667 (2003). [29] A. Vega-Flick, J. K. Eliason, A. A. Maznev, A. Khanolkar, M. Abi Ghanem, N. Boechler, J. J. Alvarado-Gil, and K. A. Nelson, Laser-induced transient grating setup with continuously tunable period, Rev. Sci. Instrum. 86, 123101 (2015). [30] W. Sauer, M. Streibl, T. H. Metzger, A. G. C. Haubrich, S. Manus, A. Wixforth, J. Peisl, A. Mazuelas, J. Hartwig, and J. Baruchel, x-ray imaging and diffraction from sur- face phonons on GaAs, Appl. Phys. Lett. 75, 1709 (1999). [31] J.-D. Nicolas, T. Reusch, M. Osterhoff, M. Sprung, F. J. R. Schulein, H. J. Krenner, A. Wixforth, and T. Salditt, Time-resolved coherent x-ray diffraction imag- ing of surface acoustic waves, J. Appl. Crystallogr. 47, 1596 (2014). [32] M. Foerster, N. Statuto, B. Casals, A. Hern´andez- M´ınguez, S. Finizio, A. Mandziak, L. Aballe, J. M. Hern`andez Ferr`as, and F. Maci`a, Quantification of prop- agating and standing surface acoustic waves by strobo- scopic x-ray photoemission electron microscopy, J. Syn- chrotron Radiat. 26, 184 (2019). [33] J. Janusonis, T. Jansma, C. L. Chang, Q. Liu, A. Gatilova, A. M. Lomonosov, V. Shalagatskyi, T. Pez- eril, V. V. Temnov, and R. I. Tobey, Transient grating spectroscopy in magnetic thin films: Simultaneous detec- tion of elastic and magnetic dynamics, Sci. Rep. 6, 29143 (2016). [34] F. Lagugn Labarthet, T. Buffeteau, and C. Sourisseau, Azopolymer holographic diffraction gratings: time depen- dent analyses of the diffraction efficiency, birefringence, and surface modulation induced by two linearly polarized interfering beams, J. Phys. Chem. B 103, 6690 (1999). [35] The incident laser power per TG excitation was 970 mW at 1 kHz repetition rate. The footprint of the non- interfering laser beams at the sample surface (ellipti- cal Gaussian) was 4.0 × 3.4 mm2 (major and minor 1/e- diameter). Assuming a homogeneous distribution of the laser pulse energy over an ellipse having the above size yields a pump fluence of 9 mJ/cm2. Due to the interfer- ence of the crossing laser beams the TG fringes at the center of the elliptical Gaussian will have twice that flu- ence in the maximum. [36] A. A. Maznev, F. Bencivenga, A. Cannizzo, F. Capo- tondi, R. Cucini, R. A. Duncan, T. Feurer, T. D. Frazer, L. Foglia, H.-M. Frey, H. Kapteyn, J. Knobloch, G. Knopp, C. Masciovecchio, R. Mincigrucci, G. Monaco, M. Murnane, I. Nikolov, E. Pedersoli, A. Simoncig, A. Vega-Flick, and K. A. Nelson, Generation of coher- ent phonons by coherent extreme ultraviolet radiation in a transient grating experiment, Appl. Phys. Lett. 113, 221905 (2018). [37] C. Svetina, R. Mankowsky, G. Knopp, F. Koch, G. Se- niutinas, B. Rosner, A. Kubec, M. Lebugle, I. Mochi, M. Beck, C. Cirelli, J. Krempasky, C. Pradervand, J. Rouxel, G. F. Mancini, S. Zerdane, B. Pedrini, V. Es- posito, G. Ingold, U. Wagner, U. Flechsig, R. Follath, M. Chergui, C. Milne, H. T. Lemke, C. David, and P. Beaud, Towards x-ray transient grating spectroscopy, Opt. Lett. 44, 574 (2019). [38] M. Wulff, A. Plech, L. Eybert, R. Randler, F. Schotte, and P. Anfinrud, The realization of sub-nanosecond pump and probe experiments at the esrf, Faraday Dis- cuss. 122, 13 (2003). [39] M. Cammarata, L. Eybert, F. Ewald, W. Reichenbach, M. Wulff, P. Anfinrud, F. Schotte, A. Plech, Q. Kong, M. Lorenc, B. Lindenau, J. Rabiger, and S. Polachowski, Chopper system for time resolved experiments with syn- chrotron radiation, Rev. Sci. Instrum. 80, 015101 (2009). [40] D. Schick, M. Herzog, A. Bojahr, W. Leitenberger, A. Hertwig, R. Shayduk, and M. Bargheer, Ultrafast lat- tice response of photoexcited thin films studied by x-ray diffraction, Struct. Dyn. 1, 064501 (2014). [41] C. L. Chang, A. M. Lomonosov, J. Janusonis, V. S. Vlasov, V. V. Temnov, and R. I. Tobey, Parametric frequency mixing in a magnetoelastically driven lin- ear ferromagnetic-resonance oscillator, Phys. Rev. B 95, 060409(R) (2017). [42] O. W. Kading, H. Skurk, A. A. Maznev, and E. Matthias, Transient thermal gratings at surfaces for thermal char- acterization of bulk materials and thin films, Applied Physics A 61, 253 (1995). [43] J. A. Johnson, A. A. Maznev, M. T. Bulsara, E. A. Fitzgerald, T. C. Harman, S. Calawa, C. J. Vineis, G. Turner, and K. A. Nelson, Phase-controlled, hetero- dyne laser-induced transient grating measurements of thermal transport properties in opaque material, J. Appl. Phys. 111, 023503 (2012). [44] J. Hidde, C. Guguschev, S. Ganschow, and D. Klimm, Thermal conductivity of rare-earth scandates in compar- ison to other oxidic substrate crystals, J. Alloys Compd. 738, 415 (2018). [45] S. Yamanaka, T. Maekawa, H. Muta, T. Matsuda, S.-i. Kobayashi, and K. Kurosaki, Thermophysical properties 10 of SrHfO3 and SrRuO3, J. Solid State Chem. 177, 3484 (2004). [46] R. Uecker, B. Velickov, D. Klimm, R. Bertram, M. Bernhagen, M. Rabe, M. Albrecht, R. Fornari, and D. Schlom, Properties of rare-earth scandate single crys- tals (Re=NdDy), J. Cryst. Growth 310, 2649 (2008). [47] R. Shayduk, H. Navirian, W. Leitenberger, J. Gold- shteyn, I. Vrejoiu, M. Weinelt, P. Gaal, M. Herzog, C. V. K. Schmising, and M. Bargheer, Nanoscale heat transport studied by high-resolution time-resolved x-ray diffraction, New J. Phys. 13, 093032 (2011). [48] D. Schick, A. Bojahr, M. Herzog, R. Shayduk, C. von Ko- rff Schmising, and M. Bargheer, udkm1Dsim -- A simula- tion toolkit for 1D ultrafast dynamics in condensed mat- ter, Computer Physics Communications 185, 651 (2014). [49] M. Herzog, D. Schick, W. Leitenberger, R. Shayduk, R. M. van der Veen, C. J. Milne, S. L. Johnson, I. Vre- joiu, and M. Bargheer, Tailoring interference and non- linear manipulation of femtosecond x-rays, New J. Phys. 14, 013004 (2012). [50] P. Haas, Surface acoustic waves in materials science, Phys. Today 55, 42 (2002). [51] D. Sander, A. Enders, and J. Kirschner, Stress and mag- netic properties of surfaces and ultrathin films, J. Magn. Magn. Mater. 200, 439 (1999). [52] R. C. O'Handley, Modern magnetic materials : principles and applications (John Wiley, New York, 2000). [53] M. Farle, Ferromagnetic resonance of ultrathin metallic layers, Rep. Prog. Phys. 61, 755 (1998). [54] A. Kirilyuk, A. V. Kimel, and T. Rasing, Ultrafast optical manipulation of magnetic order, Rev. Mod. Phys. 82, 2731 (2010). [55] M. Deb, E. Popova, M. Hehn, N. Keller, S. Man- gin, and G. Malinowski, Picosecond acoustic-excitation- driven ultrafast magnetization dynamics in dielectric bi- substituted yttrium iron garnet, Phys. Rev. B 98, 174407 (2018). [56] S. P. Zeuschner, T. Parpiiev, T. Pezeril, A. Hillion, K. Dumesnil, A. Anane, J. Pudell, L. Willig, M. Rossle, M. Herzog, A. von Reppert, and M. Bargheer, Tracking picosecond strain pulses in heterostructures that exhibit giant magnetostriction, Structural Dynamics 6, 024302 (2019). [57] Y. Yahagi, B. Harteneck, S. Cabrini, and H. Schmidt, Controlling nanomagnet magnetization dynamics via magnetoelastic coupling, Phys. Rev. B 90, 140405(R) (2014). [58] J. Tejada, E. M. Chudnovsky, R. Zarzuela, N. Statuto, J. C. de la Rosa, P. V. Santos, and A. Hern´andez- M´ınguez, Switching of magnetic moments of nanopar- ticles by surface acoustic waves, Europhys. Lett. 118, 37005 (2017). [59] C. L. Chang, R. R. Tamming, T. J. Broomhall, J. Janu- sonis, P. W. Fry, R. I. Tobey, and T. J. Hayward, Selec- tive excitation of localized spin-wave modes by optically pumped surface acoustic waves, Phys. Rev. Applied 10, 034068 (2018). [60] A. Blacha, H. Presting, and M. Cardona, Deformation potentials of k = 0 states of tetrahedral semiconductors, Phys. Stat. Sol. (b) 126, 11 (1984). [61] Y. Wang, Y. Chen, H. Li, X. Li, H. Chen, H. Su, Y. Lin, Y. Xu, G. Song, and X. Feng, Buckling-based method for measuring the strainphotonic coupling effect of gaas nanoribbons, ACS Nano 10, 8199 (2016). 11 [62] O. D. D. Couto Jr, S. Lazic, F. Iikawa, J. A. H. Stotz, U. Jahn, R. Hey, and P. V. Santos, Photon anti-bunching in acoustically pumped quantum dots, Nat. Photonics 3, 645 EP (2009). [63] M. M. de Lima and P. V. Santos, Modulation of photonic structures by surface acoustic waves, Rep. Prog. Phys. 68, 1639 (2005). [64] C. H. W. Barnes, J. M. Shilton, and A. M. Robinson, Quantum computation using electrons trapped by sur- face acoustic waves, Phys. Rev. B 62, 8410 (2000). [65] 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). [66] M. Sander, R. Bauer, V. Kabanova, M. Levantino, M. Wulff, D. Pfuetzenreuter, J. Schwarzkopf, and P. Gaal, Demonstration of a picosecond bragg switch for hard x-rays in a synchrotron-based pump -- probe experi- ment, J. Synchrotron Radiat. 26, 1253 (2019). [67] S. Vadilonga, I. Zizak, D. Roshchupkin, A. Petsiuk, I. Dolbnya, K. Sawhney, and A. Erko, Pulse picker for synchrotron radiation driven by a surface acoustic wave, Opt. Lett. 42, 1915 (2017).
1801.10333	1	1801	2018-01-31T08:04:19	Root Cause Analysis and Correction of Single Metal Contact Open-Induced Scan Chain Failure in 90nm node VLSI	[ "physics.app-ph" ]	In this paper, the localization of open metal contact for 90nm node SOC is reported based on Electron Beam Absorbed Current (EBAC) technique and scan diagnosis for the first time. According to the detected excess carbon, silicon and oxygen signals obtained from X-ray energy dispersive spectroscopy (EDX), the failure was deemed to be caused by the incomplete removal of silicate photoresist polymer formed during the O2 plasma dry clean before copper plating. Based on this, we proposed to replace the dry clean with diluted HF clean prior to the copper plating, which can significantly remove the silicate polymers and increase the yield.	physics.app-ph	physics	Root Cause Analysis and Correction of Single Metal Contact Open-Induced Scan Chain Failure in 90nm node VLSI Chao-Cheng Tingab, Ya-Chi, Liuc, Hsuan-Hsien Chena, Chung-Ching Tsaia and Liwen Shihd aDivision of IC Technology, Faraday Technology Corporation, No.5, Li-Hsin Rd III, Hsin Chu Science park, Hsin Chu , bDepartment of Materials science and Engineering, National Chiao-Tung University, 1001 Ta-Hsueh Road, Hsinchu, Taiwan 300, R.O.C 30010 Taiwan, R.O.C Corresponding author: Taiwan 30010, R.O.C. cOptmic Lab, 340 S. Lemon Ave. #8365, Walnut, CA 91789, U.S.A. dDepartment of Computational Engineering, University of Houston-Clear Lake, Houston, U.S.A. Department of Materials Science and Engineering, National Chiao-Tung University, 1001 Ta Hsueh Road, Hsinchu, *Corresponding author: E-mail: [email protected] loop. Focused Ion Beam (FIB) cross section image reveals the copper void causing the open contact. Finally, Transmission Electron Microscope (TEM) and X-ray energy dispersive spectroscopy (EDX) analysis are conducted to identify the failure mechanism. The investigation results show that the failure was possibly due to the incomplete removal of photo- resist residues and solvent at the via bottom that results in the formation of silicate polymer during the plasma dry clean prior to copper plating. To substantiate this hypothesis, the fabrication flow was modified with an extra diluted HF cleaning prior to copper plating to replace the dry cleaning. With the modified flow, it was found that the silicate polymer is significantly removed and the yield greatly enhanced from approximately 34% to 90%. Discussion Background of the problem A 90 nm node VLSI based on standard performance process suffered significant low yield of approximately 34%. The failing wafer map is shown in Figure 1. With reference to Figure 1, the two top failures are identified as scan chain test and ATPG (Automatic Test Pattern Generation) test. Since the scan chain failure dominates, this paper studies the cause and corrective action to recover the scan chain failure. Abstract In this paper, the localization of open metal contact for 90nm node SOC is reported based on Electron Beam Absorbed Current (EBAC) technique and scan diagnosis for the first time. According to the detected excess carbon, silicon and oxygen signals obtained from X-ray energy dispersive spectroscopy (EDX), the failure was deemed to be caused by the incomplete removal of silicate photoresist polymer formed during the O2 plasma dry clean before copper plating. Based on this, we proposed to replace the dry clean with diluted HF clean prior to the copper plating, which can significantly remove the silicate polymers and increase the yield. Key words: Metal Void; 90nm VLSI; EBAC; Wet Clean Introduction As transistors shrink in size, the localization of integrated circuits (IC) failures is facing a great challenge to detect weak leakage-current-induced thermal emissions [1]. The challenge is aggravated when metal voids that occurred in very-large- scale-integration (VLSI) do not give rise to thermal emissions, hindering emission microscopy (EMMI) detection. Electron Beam Absorbed Current (EBAC) technique uses an electron beam as a probing element to achieve nano-meter scale resolution in this isolation of open failures in VLSI [1, 2] based on the repulsion nature of the electron signal at the open spot. In this paper, we propose a combination of scan diagnosis and EBAC to locate open defects with faster turnaround and precision. First, scan diagnosis is used to narrow down the suspected failure loop before EBAC locates the failing path from the Figure 1. The failing wafer map. (Green) Passing dies. (Purple) Dies failing scan chain. (Blue)Dies failing ATPG test. Scan Diagnosis and Sample preparation First, scan diagnosis was performed to localize the suspected failure net as shown in Figure 2 (top). Based on the scan diagnosis result, the sample was polished to expose via 5 for probing. As shown in Figure 2 (bottom), the power was supplied to via 5 and the ground was applied to the substrate for further EBAC analysis. Fault isolation using EBAC EBAC technique was performed on that suspected trace to identify the failure location. Figure 3 shows the EBAC image indicating the failure location to be around via 4 as shown by the disconnected metal trace after via 4 as compared to the scan diagnosis suspected net; this technique provides a quick way to assess the suspected failing net without physically inspecting the entire net which can be time- consuming. Figure 3. The scan diagnosis result of failure trace (left) and the EBAC image (right). Material and Root Cause analysis Figure 4 shows the SEM image of the FIB cross section revealing a sub-surface void at the bottom of via 4, which is of the nature as reported by Richard L., et al [3]. In addition, there is no deformed via as compared to the adjacent ones. This indicates that the open failure does not belong to the via etch or former lithography process. In order to uncover the actual cause of the open failure, we performed EDX element analysis. As shown in Figure 5(c) and (f), the extra carbon and oxygen signals were discovered at the bottom of the metal contact. Based on the above information and the presence of silicon signal, as shown in Figure 5(b), we reasonably assumed that the Si, O and C signals come from silicate polymer as it cannot be removed by wet solvent cleaning as depicted in Figure 6. Figure 2. The scan diagnosis indicating the suspected failure loop (top) and the illustration of probing for EBAC imaging (bottom). Figure 4. SEM overview image of the FIB cross section (left) and the high magnitude SEM image indicating via 4 open (right). plasma dry clean is capable of ionizing the photoresist and solvent residue, leading to the formation of Si, C and O radicals. Third, after the plasma dry clean, the Si, C and O radicals recombined to form silicate polymer which give rise to C, Si, O signals in the EDX images. Finally, the unwanted silicate polymer along with the solvent residue evaporate during the later pre-barrrier baking process leading to a poor adhesion in electrochemical plating process. This explains the subsequent copper void. Figure 6. The fabrication flow and the suspected photoresist O2 plasma dry clean step. Figure 5. TEM and EDX mapping image. (a) The HAADF (High Angle Annular. Dark Field) TEM image. (b) The EDX mapping of Si (light blue). (c) The EDX mapping of O (pink). (d) The EDX mapping of N (green). (e) The EDX mapping of Cu (dark blue). (f) The EDX mapping of C (yellow). (g) The EDX mapping of Ta (brown). Process flow and partition check analysis A partition check analysis was conducted using defect performance inspection at each process step. It was found that the failure likely occur in the steps after the first solvent clean, as indicated in Figure 6. Based on this insight, we postulate that the formation of silicate polymer was due to the reaction of the silicon and photoresist radicals generated during the plasma dry clean process, as indicated by the red spot in Figure 6. This is because unlike the organic compounds (photoresist), the silicate polymer cannot be easily removed during the wet clean. Also, based on Figure 6, the solvent may remain in the silicate polymer and outgas during the pre- barrier baking process, which can lead to poor copper adhesion in the following electrochemical plating process (Koh, L. T., et al argued the carbon containment may affect the bottom fill up ability and He, Q. Y et al noted the residual carbon contamination will increase the occurrences of copper void during the ECP. [4,-6]). To substantiate our hypothesis, we devised a revised fabrication flow with ultra-dilute hydrogen fluoride (UDHF) clean and N2 purge to replace the dry clean and second solvent clean to avoid radical recombination and solvent outgassing during pre-barrier bake respectively. As shown in Figure 7(a), the UDHF clean was applied just after the solvent clean and before the DI wafer flush to ensure the total removal of the expected silicate polymer. The scan chain fail rate was significantly reduced with the revised fabrication flow. The wafermap with the improved yield is presented in Figure 7(b). The yield improvement from approximately 34% to 90% as a result of the corrective action directly supports our assumption on the formation of silicate polymer. For a complete understanding on the failure mechanism, Figure 8 illustrates the formation of silicate polymer in three stages. First, after the etching process, the photoresist and solvent residue that remain at the bottom of via are hard to be completely removed by the wet clean process. Second, the O2 Figure 7. (a) The revised fabrication flow. (b) The test result of the revised flow. (Green) Passing dies on all test items. Figure 8. The schematic of the failure mechanism. (a) solvent and photo-resist residue (carbonaceous compound) adhere to the via bottom. (b) The O2 plasma dry clean.generate O, Si and C radicals. (c) The unwanted silicate polymer is formed at the via bottom based on the O, Si and C radical recombination . Conclusions This study demonstrates the effectiveness of using a combination of scan diagnosis and EBAC to locate the metal contact failure for complex VLSI in 90nm technology node. Subsequent EDX analysis at the failing location reveal the root cause of the open metal contact to be due to the deposition of silicate polymer formed by the recombination of O, Si and C radicals. To fix this problem, we modified the fabrication process by implementing an extra UDHF cleaning and reducing the oxygen plasma processes to prevent possible formation of the undesirable O, Si and C radicals during the plasma dry clean. The yield enhancement successfully proves the failure mechanism. Acknowledgments We acknowledge the failure analysis support of United Microelectronics Corporation and Materials Analysis Technology Inc. References [1] Kleindiek, et al. Stephan, "Current imaging, EBIC/EBAC, and electrical probing combined for fast and reliable in situ electrical fault isolation." Microelectronics Reliability 64 (2016): 313-316. [2] Chang, Link, et al. "The investigation of Active VC and EBAC analysis utilization on test structure." Physical and Failure Analysis of Integrated Circuits (IPFA), 2015 IEEE 22nd International Symposium on the. IEEE, 2015. [3] Guldi, Richard L., et al. "Characterization of copper voids in dual damascene processes." Advanced Semiconductor Manufacturing 2002 IEEE/SEMI Conference and Workshop. IEEE, 2002. [4] Koh, L. T., et al. "Investigation of the effects of byproduct components in Cu plating for advanced interconnect metallization." Microelectronics journal 33.3 (2002): 229-234. [5] Horii, Yuichi, and Kurunthachalam Kannan. "Survey of organosilicone compounds, including cyclic and linear siloxanes, in personal-care and household products." Archives of environmental contamination and toxicology 55.4 (2008): 701. He, Qi-yang, Jun-qing Zhou, and Hai-Yang Zhang. "Study of BEoL dual damascene ULK etch and the correlation to Cu gapfill performance." Semiconductor Technology International Conference (CSTIC), 2016 China. IEEE, 2016. [6]
1911.05007	1	1911	2019-11-12T16:48:23	Iridium Enabled Field-free Spin-orbit Torque Switching of Perpendicular Magnetic Tunnel Junction Device	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Writing magnetic bits by spin-orbit torques (SOTs) arising from spin Hall effect creates new possibilities for ultrafast and low-power magnetoresistive random access memory (MRAM). For perpendicular MRAM, an extra in-plane field is required to break the symmetry for the deterministic SOT writing of the perpendicular storage layer. Although schemes have been demonstrated in external-field-free SOT switching of a perpendicular layer, practically integrating them with perpendicular MTJs still appears to be challenging. Here, we present experimental demonstration of spin-orbit torques (SOTs) switching a perpendicular magnetic tunnel junction (MTJ) device without applying an external magnetic field. An Ir layer is used to serve dual-purpose of both injecting the pure spin current via spin Hall effect and mediating an in-plane exchange field to the perpendicular free layer of the MTJ. Robust field-free SOT switching with pulsed write path current is demonstrated for various MTJ sizes ranging from 50 nm to 500 nm. The effect of MTJ size and pulse width on the critical switching current is studied. Combined micromagnetic simulations are carried out to provide in-depth analysis of the switching dynamics as well as the thermal effect on the switching.	physics.app-ph	physics	Iridium Enabled Field-free Spin-orbit Torque Switching of Perpendicular Magnetic Tunnel Junction Device Yang Liu1,3*, Bing Zhou1,3, Zhengkun Dai2,3, Enbo Zhang2,3, Jian-Gang (Jimmy) Zhu2,3 1 Department of Materials Science and Engineering, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, Pennsylvania, 15213, USA 2 Department of Electrical and Computer Engineering, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, Pennsylvania, 15213, USA 3 Data Storage Systems Center, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, Pennsylvania, 15213, USA Abstract Writing magnetic bits by spin-orbit torques (SOTs) arising from spin Hall effect creates new possibilities for ultrafast and low-power magnetoresistive random access memory (MRAM). For perpendicular MRAM, an extra in-plane field is required to break the symmetry for the deterministic SOT writing of the perpendicular storage layer. Although schemes have been demonstrated in external-field-free SOT switching of a perpendicular layer, practically integrating them with perpendicular MTJs still appears to be challenging. Here, we present experimental demonstration of spin-orbit torques (SOTs) switching a perpendicular magnetic tunnel junction (MTJ) device without applying an external magnetic field. An Ir layer is used to serve dual-purpose of both injecting the pure spin current via spin Hall effect and mediating an in-plane exchange field to the perpendicular free layer of the MTJ. Robust field-free SOT switching with pulsed write path current is demonstrated for various MTJ sizes ranging from 50 nm to 500 nm. The effect of MTJ size and pulse width on the critical switching current is studied. Combined micromagnetic simulations are carried out to provide in-depth analysis of the switching dynamics as well as the thermal effect on the switching. Introduction Over the last decades, magnetoresistive random access memory (MRAM) has come to the point where it's considered as a strongly competitive candidate for cache memory replacement1 -- 4. In MRAM, storage units with perpendicular magnetic anisotropy (PMA) are preferred for that they render good thermal stability as well as high storage density5 -- 7. Moreover, it has been shown that magnetization reversal via spin-orbit torques (SOT) arising from spin Hall effect can allow MRAM to be capable of ultra-fast and low-power writing operations8 -- 11. To realize SOT induced magnetization switching in perpendicular MRAM, however, a static in-plane magnetic field is necessary for breaking the symmetry9,12. But applying a field externally is impractical for technological implementation. Successful field-free perpendicular SOT switching has been realized with either utilizing antiferromagnetic materials13 -- 15, or incorporating an additional exchange coupling layer16, or manipulating device geometry17,18. In those schemes, although field-free switching of a single Co/ CoFe layer or Co/Ni multilayers is demonstrated, the integration with perpendicular MTJs doesn't seem to be so straightforward. The difficulties involve weak PMA and/or small exchange bias with FeCoB, inappropriate device design preventing MTJ integration and incapability to scale down. In this work, we experimentally demonstrate the field-free SOT switching of perpendicular MTJs via utilizing Ir as the material of dual functions: generation of SOT along with mediating an in-plane field. The device adopts the usual three-terminal device design and doesn't require special engineering in device shape or inserting other functional layers. We show that the Ir-enabled field- free device exhibits reliable writing and reading operations, moving a step closer to the practical applications of SOT-related magnetoresistive devices. In our approach, the key components for enabling zero-field switching is the [in-plane magnet/ Ir/ free perpendicular magnet] trilayer structure. The mechanism for the field-free switching is attributed to the interlayer exchange coupling and spin Hall effect of Ir19. Here, a preferable thickness of Ir is used so that the Ir layer is able to provide sufficient interlayer exchange coupling between adjacent ferromagnetic layers. As such, the free perpendicular layer (mp // z) above Ir experiences a local in-plane exchange coupling field (Hx // x) from the lower in-plane magnetized layer (mi // x). During current-induced switching, charge current along x flows into the Ir layer, generating pure spin current with spin polarization 𝝈 // y due to spin Hall effect. The can be decomposed into a Slonczewski-like torque mp x (𝝈 x mp) and field-like torque mp x 𝝈. spin current travelling along z is absorbed by the free layer mp and subsequently exert SOTs that Facilitated by the local in-plane coupling field Hx, the SOTs are capable of driving the perpendicular magnetization reversal of mp in absence of an external magnetic field19. We also show that Ir fits well with FeCoB/MgO-based perpendicular MTJ and so the SOT driven magnetization reversal can be efficiently detected by the magnetoresistance change. In addition, the switching current threshold is characterized in terms of varying device size as well as pulse width, and in-depth study is carried out via micromagnetic simulations. Results × 10-8 Torr. The film structure, as shown in Figure 1(a), is substrate/ IrMn(5)/ Co(2)/ Ru(0.85)/ We deposit the film at room temperature by magnetron sputtering with base pressure < 2 Co(2)/ Ir(1.35)/ FeCoB(1.2)/ MgO(1.2)/ FeCoB(1.3)/ Ta(0.5)/ [Co(0.4)/ Pt(1)]3.5/ Ru(0.85)/ [Co(0.4)/ Pt(1)]6.5/ Pt(7) (in nm). The purpose of adopting an in-plane SAF instead of a single in- plane Co layer below Ir is to minimize the effect of the in-plane stray field on the perpendicular SOT switching. The thickness of the Ir layer corresponds to the second antiferromagnetic coupling peak in the Ruderman -- Kittel -- Kasuya -- Yosida (RKKY) thickness dependence curve (see Supplementary Information). After deposition, the film is post-annealed at 300 0C for 10 min with a 5000 Oe magnetic field applied along -x direction. Figure 1(a) also shows the hysteresis loops of the film, exhibiting the well-defined PMA of the MTJ. The deposited film is processed into three-terminal magnetoresistive devices (Figure 1(b)) (more details in Method). The tunnel magnetoresistance ratio (TMR) is evaluated by measuring the resistance of MTJ with various perpendicular magnetic field. Figure 1(b) shows the TMR minor loop of 1.2-nm-thick free FeCoB layer of the device with a 150 nm-diameter MTJ. Bi-stable states of the low and high resistance of the MTJ is observed at zero field. The TMR is about 60%, with the abrupt magnetization reversal of the free layer. In current-induced switching measurements, 200-ns current pulses are applied into the write path consisting of the Ir layer and the bottom pinned in-plane SAF structure. We connect a 10 MΩ resistor in series with the MTJ to reduce the bias voltage across the MgO layer. The resistance change of the MTJ is monitored by the lock-in technique. Figure 2(a) shows the well- defined current-induced switching of the MTJ at zero external field. Starting with a low-resistance state of MTJ and sweeping the current from positive to negative, a sharp increase in the MTJ resistance is observed when the current is below a negative threshold value, indicating the parallel (P) to antiparallel (AP) magnetization switching. Next, as the current is swept back to a sufficiently large positive value, the switching is reversed (AP to P switching), as shown by the abrupt drop in MTJ resistance. By applying a sequence of switching pulses (see Supplementary Information), the devices exhibit reliable zero-field switching capability, where the MTJ can be switched repeatedly between low and high resistance states with applied current of opposite polarities. To further study the switching behaviors, an external in-plane field Hx is applied during the current switching measurements. In Figure 2(b), when Hx is applied along -x direction, we obtain almost identical switching behavior as the zero-field switching, i.e., the switching is clockwise. It's as expected since the exchange coupling field acting on the free FeCoB layer also lies in the -x direction due to the antiferromagnetic coupling characteristic of the 1.35 nm Ir. When the direction of Hx is reversed, the switching loop becomes counterclockwise. It's due to that the anti-aligned Hx now overcomes the local exchange coupling field (~100 Oe, see Supplementary Information) and thus the net field acting on the free FeCoB layer is along +x direction, leading to the opposite switching outcome. Next, we compare the switching current for the devices with different sizes. In Figure 2(c), the results show that the switching current is linearly proportional to the width of the write path. The linear relationship can be interpreted by the fact that the SOT switching condition is determined by the current density instead of absolute current in the write path8,12. Furthermore, we study the effect of MTJ size on the switching current threshold. As shown in Figure 2(d), no significant change in the switching current is observed with MTJ sizes ranging from 500 nm to 150 nm while the switching current starts to increase as the MTJ size becomes smaller than 80 nm. To explain such trend in the switching current, we conduct micromagnetic simulations to compare the switching processes in devices with different sizes. Fig.3(a) illustrates the switching dynamics of the free FeCoB layer in a relatively large (240 nm x 240 nm) device. It's found that the switching starts with nucleation and formation of dense strips of reverse domains followed by rather complex expansion and merge of these reverse strip domains till the full reversal of the magnetic layer is reached. Note that such SOT resulted nucleation and expansion of reverse strip domains has been observed with utilization of Kerr microscopy in our previous experiments19. In Fig. 3(b), we show the initially formed reverse domains for five difference device sizes ranging from 300 x 300 nm2 to 30 x 30 nm2. Note the size of each reverse strip domain is essentially independent of device size. As the size of the free layer goes down, we observe the reducing number of domains. When the size is reduced below 60 nm, only a single reverse strip domain would nucleate under the SOT in the presence of the in-plane exchange field and thus near single domain switching is observed. Such transition from multi-domain switching in large devices to single domain switching in smaller devices causes a slight increase in the calculated switching current, as shown in Fig. 3(c). The simulated trend of the switching current as a function of MTJ size agrees well with our experimental data (Fig. 2(d)). We then study the effect of pulse width on the switching current threshold. The duration of current pulse is varied from 20 ns to 5 𝜇s and the effect on the critical switching current is linear function of inverse √𝜏. When current pulse width is beyond 100 ns, the switching current plotted in Fig. 4. At current pulse width below 100 ns, the switching current value is essentially a levels off to almost a constant value. Combined micromagnetic simulation using all measured properties of the fabricated MTJ stack implies that the SOT switching in our devices should occur within 1 ns. Furthermore, the simulation shows that when the pulse width becomes longer than 1 ns, the switching current threshold should become a constant value, no longer depending on the pulse width. However, the measured switching current dependence on pulse width ranging from 20 ns and 100 ns evidently disagree with the micromagnetic simulations. The most plausible reason should be that the perpendicular anisotropy of the free layer reduces due to the joule heating generated during the current pulse injection, further causing the change in the switching current threshold. Our simulation results shown in Fig. 5(a) indicate a linear relationship between the critical switching current and the perpendicular anisotropy of the free layer. Previous study has found that the perpendicular anisotropy of FeCoB/MgO decreases linearly with temperature, with a measured Blocking temperature at TB = 450K at which the anisotropy vanishes20. Assuming the same Blocking temperature for the free layer fabricated here, the temperature dependence of the interfacial perpendicular anisotropy of the free layer is: ( ) TK = K ae 10 çç è - TT - RT T T - B RT ö ÷÷ ø Eq.(1) where K0 is the perpendicular anisotropy constant at room temperature TRT. With a current pulse of duration t, the temperature rise of the write wire underneath the free layer of the MTJ can be written as [ kRI 2 - TT - is the power of heat dissipation, CVeffective is the heat capacitance, 𝐼 is the current and 𝑅 is the resistance of the write path. Combining the two above equations with this Eq(2) - ( TT dissp VC × Where ( TT - )RT k dissp = RT ] ) × t RT linear dependence in Fig. 5(a), we have I = I o 2 M - p ) 2 S ( K 0 K 0 ì ïï í ï ï î ae ç ç ç ç è 1 - RI 2 ö ×÷ ø dissp kVC × + t ae ç è ( T B - T RT ) - 2 M p 2 S ö ÷ ÷ ÷ ÷ ø ü ïï ý ï ï þ Eq.(3) where I0 is the level-off switching current threshold at room temperature. The above equation should describe the switching current threshold as a function of pulse width t, with considering the anisotropy reduction of the free layer. Fig. 5(b) combines micromagnetic simulation results (blue curve) and measurement data (red circles) as a function of inverse pulse width. The black curves are simulation results for the free layers with varying perpendicular anisotropy due to different joule heating, showing how the heat effect curve (blue curve) is generated. The results shown in Fig. 5(b) provides a good understanding on the current driven perpendicular SOT switching in our devices. If the current pulse width is shorter than 1 ns, the temperature of the write wire doesn't rise, and the free layer anisotropy remains unchanged. The switching current threshold follows the micromagnetic simulation result for the case with K0. In this case, the switching occurs within a fraction of nanosecond and the threshold current will level off to a constant value for any pulse width longer than 1 ns. For the cases with current pulse width longer than 1 ns, the temperature elevates, and the perpendicular anisotropy of the free layer decreases accordingly. The measured data is actually the level-off current threshold at different anisotropy. A longer duration pulse (for t < 100 ns) leads to higher temperature, hence lower threshold current, until the heat generated becomes balanced by the heat dissipation. This lowest threshold current corresponding to the maximum temperature increase (from room temperature), which is 57 K in the case based on the calculation. Conclusion In summary, we have demonstrated reliable and repeatable field-free SOT switching of perpendicular MTJs, facilitated by a thin Ir layer which performs dual-function of injection of pure spin current via spin Hall effect and providing an in-plane field via exchange coupling. The switching dynamics has been investigated by both experiments and micromagnetic simulation. We found that the non-thermal impact SOT switching occurs over sub-nanosecond range since it is governed by magnetization precession. At nanosecond scale or longer, thermally induced anisotropy field degradation in the free layer of the MTJ becomes significant, resulting in lowering switching threshold current. Our approach for field-free switching is based on well-understood phenomena and doesn't require special works in order to integrate with magnetoresistive stacks. The devices also exhibit potentials in ultra-fast switching speed and scaling down to dimensions of domain wall width. We believe such Ir-enabled manipulation of perpendicular MTJs can move forward the practical realizations of applications involving SOT memory, sensors and logic devices. Method All films are deposited by magnetron sputtering on Si/SiO2 substrate at room temperature. We use DC sputtering to deposit metal layers and RF sputtering to deposit the MgO layer. The composition of IrMn is Ir20Mn80 (at%) and of FeCoB is Fe17.5Co52.5B30 (at%). The films are post- annealed at 3000C for 10 min with a 5000 Oe magnetic field applied along -x direction as shown in Figure 1. To pattern three-terminal devices, we first deposit a hard mask of 60 nm C /5 nm SiN. Then we perform e-beam lithography with HSQ negative resist followed by reactive ion etching and ion beam etching to fabricate MTJ pillars. Next, we pattern the write path by photolithography with AZ4210 positive resist and etch away all layers using ion beam etching. After striping the resist, a layer of 140 nm SiN passivation layer is then deposited on the chip. We then do a low-angle ion beam etching to planarize the chip. Top and bottom lead channels are opened sequentially by photolithography with AZ4110 positive resist and reactive ion etching. At last, we fabricate the leads by photolithography, 5 nm Ti/ 80 nm Au deposition and lift-off process. All electrical measurements are performed at room temperature. Before measurements, the devices are saturated with -10 kOe perpendicular field and relaxed back to zero field. This is to set the reference FeCoB layer as up magnetization state in all devices. The TMR loop is measured by detecting device resistance with sweeping perpendicular field. The current-induced switching loop is obtained by applying current pulses (with pulse width ranging from 5 𝜇s to 20 ns) along the writing channel and measuring the device resistance after each pulse, with or without the presence of an in-plane field. In the micromagnetic simulation, the Landau-Lifshitz-Glibert equation adding Slonczewski spin transfer torque f md dt " Hm g ´ f -= + m a f ´ f md dt + ! 2 e × × g J q SH M d s m f ´ p SH ´ m f is used to model the magnetization, unit vector fm , of the perpendicular magnetic free layer (for the MTJ) on top of the Ir layer. In the above equation, SHp is the unit vector of injected spin polarization, dsM is the magnetic moment density of the free layer, J is the current density in the Ir layer and SHq is the spin Hall angle. ! H is the effective magnetic field defined as ! H -= 1 M S E ¶ m ¶ where the energy here includes the perpendicular anisotropy energy, the exchange energy arising from spatial inhomogeneity of the magnetization, and the coupling energy between the free layer and the in-plane magnetic layer in the SAF underneath the Ir layer: KE = ( mk ×- 2 ) [ ù ( mA Ñ+úû x é 1 êë 2 ) ( Ñ+ m y 2 ) ( Ñ+ m z 2 ) ] + mm s 1 × f where =A 1.0 ´ 10 -6 erg/cm is the exchange stiffness constant used and 0.014 erg/cm2 is used for the interlayer exchange coupling between the magnetic layers on either side of the Ir layer. The two in-plane magnetic layers in the synthetic antiferromagnet, unit magnetization vectors 1m and 2m , are also modeled simultaneously with the following LLG equations: md 2,1 dt -= m g 2,1 ´ ! H + m a 2,1 ´ 2,1 md 2,1 dt Each of the three magnetic layers is meshed into a 2D array of mesh elements of 2 nm x 2 nm thickness in size. The magnetization vectors of the mesh cells for all the elements are solved simultaneously with numerical integrations of the above coupled dynamic equations. Reference 1. Wolf, S. A. and Awschalom, D. D. and Buhrman, R. A. and Daughton, J. M. and von Molnr, S. and Roukes, M. L. and Chtchelkanova, A. Y. and Treger, D. M. Spintronics: a spin-based electronics vision for the future. Science 294, 1488 -- 95 (2001). 2. Huai, Y. Spin-transfer torque MRAM (STT-MRAM): Challenges and prospects. AAPPS Bull. 18, 33 -- 40 (2008). 3. Chen, E. et al. Advances and future prospects of spin-transfer torque random access memory. IEEE Trans. Magn. 46, 1873 -- 1878 (2010). 4. Kent, A. D. & Worledge, D. C. A new spin on magnetic memories. Nat. Nanotechnol. 10, 187 -- 191 (2015). 5. Ikeda, S. et al. A perpendicular-anisotropy CoFeB-MgO magnetic tunnel junction. Nat. Mater. 9, 721 -- 724 (2010). 6. 7. Zhu, J.-G. G. J. & Park, C. D. Magnetic tunnel junctions. Mater. Today 9, 36 (2006). Sato, H. et al. Perpendicular-anisotropy CoFeB-MgO magnetic tunnel junctions with a MgO/CoFeB/Ta/CoFeB/MgO recording structure. Appl. Phys. Lett. 101, (2012). 8. Liu, L. Spin-torque switching with the giant spin Hall effect of tantalum. Science (80-. ). 336, 555 -- 558 (2012). 9. Liu, L., Lee, O. J., Gudmundsen, T. J., Ralph, D. C. & Buhrman, R. A. Current-induced switching of perpendicularly magnetized magnetic layers using spin torque from the spin Hall effect. Phys. Rev. Lett. 109, 96602 (2012). 10. Pai, C. F. et al. Spin transfer torque devices utilizing the giant spin Hall effect of tungsten. Appl. Phys. Lett. 101, 1 -- 5 (2012). 11. Cubukcu, M. et al. Spin-orbit torque magnetization switching of a three-terminal perpendicular magnetic tunnel junction. Appl. Phys. Lett. 104, (2014). 12. Fukami, S., Anekawa, T., Zhang, C. & Ohno, H. A spin -- orbit torque switching scheme with collinear magnetic easy axis and current configuration. Nat. Nanotechnol. 1 -- 27 (2016). doi:10.1038/nnano.2016.29 13. Fukami, S., Zhang, C., DuttaGupta, S., Kurenkov, A. & Ohno, H. Magnetization switching by spin -- orbit torque in an antiferromagnet -- ferromagnet bilayer system. Nat. Mater. 15, 535 (2016). 14. van den Brink, A. et al. Field-free magnetization reversal by spin-Hall effect and exchange bias. Nat. Commun. 7, 10854 (2016). 15. Oh, Y.-W. et al. Field-free switching of perpendicular magnetization through spin -- orbit torque in antiferromagnet/ferromagnet/oxide structures. Nat. Nanotechnol. 11, 878 (2016). 16. Lau, Y.-C., Betto, D., Rode, K., Coey, J. M. D. & Stamenov, P. Spin -- orbit torque switching without an external field using interlayer exchange coupling. Nat. Nanotechnol. 11, nnano-2016 (2016). 17. Yu, G. et al. spin -- orbit torques in the absence of external magnetic fields. 9, (2014). 18. You, L. et al. Switching of perpendicularly polarized nanomagnets with spin orbit torque without an external magnetic field by engineering a tilted anisotropy. Proc. Natl. Acad. Sci. 112, 10310 -- 10315 (2015). 19. Liu, Y., Zhou, B. & Zhu, J. J. Field-free Magnetization Switching by Utilizing the Spin Hall Effect and Interlayer Exchange Coupling of Iridium. Sci. Rep. 1 -- 7 (2019). doi:10.1038/s41598-018-37586-4 20. Lee, K.-M., Choi, J. W., Sok, J. & Min, B.-C. Temperature dependence of the interfacial magnetic anisotropy in W/CoFeB/MgO. AIP Adv. 7, 65107 (2017). Figure 1. (a) Left: film stack, unit in nanometer. Right: M-H loops of the film. (b) Left: schematic of the device and testing set-up. Right: Perpendicular-field-driven TMR minor loop of a device with a 150-nm-diameter MTJ. externally applied field along current direction. The device has a 2.5 𝜇m-width write path and a Figure 2. Upper: Current-induced switching results (a) at zero external field and (b) with an 150-nm-diameter MTJ. Lower: Switching current as a function of (c) the width of the write path and (d) the MTJ size. dynamics of the free FeCoB layer in a 240 nm x 240 nm device. Time: 𝜏+<𝜏-<𝜏.<𝜏/. (b) Figure 3. Micromagnetic simulation of the zero-field switching dynamics. (a) Switching Comparison of the switching in devices with various sizes. (c) Simulated trend of the switching current as a function of MTJ size. Figure 4. Switching current with different pulse width 𝜏. Upper right: Switching current is plotted as a function of inverse √τ. Figure 5. (a) Simulated critical current density as a function of the perpendicular anisotropy of the free layer. (b) Simulated switching current as function of the pulse width with different perpendicular anisotropy of the free layer (black curves). The blue curve fits the level-off current of each black curve and shows how the switching current changes when considering the heat effect. The red dots correspond to the experimental data. Supplementary Information 1. Interlayer exchange coupling of Iridium (Ir) Film stacks for characterize the interlayer exchange coupling of Ir are substrate/ Co (2 nm)/ Ir (tIr)/ Co (2 nm)/ Ta (2 nm) where tIr ranges from 0.6 nm to 1.65 nm. In-plane hysteresis loops are measured by alternating gradient field magnetometer (AGFM) and used to determine the strength of interlayer exchange coupling. Figure S1 shows the exchange coupling field as a function of Ir layer thickness. Note that only antiferromagnetic coupling is shown in Figure S1. As can be seen, the first antiferromagnetic coupling peak locates at 0.6 nm and the second at 1.35 nm. In the field-free switching experiments, we choose Ir layer thickess to be 1.35 nm because the exchange coupling via 0.6 nm Ir is so strong that the perpendicular layer can be pulled into in- plane direction. Figure S1. Antiferromagnetic coupling strength via an Ir spacer with various Ir layer thickness 2. Field-free SOT switching of a perpendicular FeCoB layer Here we show the field-free SOT switching of a perpendicular FeCoB layer. The film stack is substrate/ Ir20Mn80(5 nm)/ Co(2 nm)/ Ru(0.85 nm)/ Co(2 nm)/ Ir (1.35 nm)/ FeCoB (1.2 nm)/ MgO (2 nm)/ Ta capping. The film is post-annealed at 3000C for 10 min with a 5000 Oe in-plane anisotropy of the FeCoB layer. field. The perpendicular M-H loop in Figure S2(b) exhibits the strong perpendicular magnetic Then we pattern the film into cross-shape Hall-bar devices with 1 𝜇𝑚-wide current-channel and 1 𝜇𝑚-wide voltage-channel. During the switching measurements, an external in-plane field is applied along the current direction with the magnitude ranging from 500 Oe to -500 Oe. In Figure S2(c), we observe well-defined and complete SOT switching of the perpendicular FeCoB layer at zero field. Such field-free switching, as discussed in the main manuscript, is attributed to the interlayer exchange coupling as well as the spin Hall effect of Ir. It's also observed that the switching loop collapses at a 100 Oe external in-plane field. No switching happens because the anti-aligned external field cancels with the local exchange coupling field, indicating the coupling field acting on the FeCoB layer has the magnitude of about 100 Oe. Figure S2. (a) Film stack (unit in nm). (b) M-H loops of the film. (c) Current-induced magnetization switching w/o an external field. 3. Zero-field switching of perpendicular MTJs with switching pulse cycles Figure S3 shows the detected signals from MTJ resistance change with a sequence of positive and negative switching-current pulses. As can be seen, the MTJ can be switched back and forth between low and high resistance states by applying current of opposite polarities. This is to demonstrate that the Ir-enabled field-free switching devices exhibit reliable writing and reading performance. Figure S3. Field-free switching of perpendicular MTJs with a sequence of positive and negative switching pulses.
1905.11242	1	1905	2019-05-27T14:05:15	Time-resolved imaging of non-diffusive carrier transport in long-lifetime halide perovskite thin films	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Owing to their exceptional semiconducting properties, hybrid inorganic-organic perovskites show great promise as photovoltaic absorbers. In these materials, long-range diffusion of charge carriers allows for most of the photogenerated carriers to contribute to the photovoltaic efficiency. Here, time-resolved photoluminescence (PL) microscopy is used to directly probe ambipolar carrier diffusion and recombination kinetics in hybrid perovskites. This technique is applied to thin films of methylammonium lead tri-iodide MAPbI$_3$ obtained with two different fabrication routes, methylammonium lead tribromide (MAPbBr$_3$), and an alloy of formamidinium lead tri-iodide (FAPbI$_3$) and methylammonium lead bromide FA$_{0.85}$MA$_{0.15}$Pb(I$_{0.85}$Br_${0.15}$)$_3$. Average diffusion coefficients in the films leading to the highest device efficiencies and longest lifetimes, i.e., in FA$_{0.85}$MA$_{0.15}$Pb(I$_{0.85}$Br$_{0.15}$)$_3$ and acetonitrile-processed MAPbI$_3$, are found to be several orders of magnitude lower than in the other films. Further examination of the time-dependence shows strong evidence for non-diffusive transport. In particular, acetonitrile-processed MAPbI$_3$ shows distinct diffusion regimes on short and long timescales with an effective diffusion constant varying over 2 orders of magnitude. Our results also highlight the fact that increases in carrier lifetime in this class of materials are not necessarily concomitant with increased diffusion lengths and that the PL quantum efficiency under solar cell operating conditions is a greater indication of material, and ultimately device, quality.	physics.app-ph	physics	Time-resolved imaging of non-diffusive carrier transport in long-lifetime halide perovskite thin films Aravindan Sridharan, Nakita K. Noel, Hyeon Hwang, Soroush Hafezian, Barry P. Rand, Stéphane Kéna-Cohen* A. Sridharan, H. Hwang, S. Hafezian, Prof. S. Kéna-Cohen Department of Engineering Physics Polytechnique Montréal 2900 Édouard-Montpetit Blvd, Montreal, QC H3T 1J4, Canada E-mail: [email protected] Dr. N. K. Noel, Prof. B. P. Rand† Princeton Research Institute for the Science and Technology of Materials † Andlinger Center for Energy and the Environment Department of Electrical Engineering Princeton University Princeton, New Jersey 08544, USA Keywords: hybrid organic-inorganic perovskites, diffusion, photoluminescence, time-resolved spectroscopy, streak camera imaging 1 Abstract Owing to their exceptional semiconducting properties, hybrid inorganic-organic perovskites show great promise as photovoltaic absorbers. In these materials, long-range diffusion of charge carriers allows for most of the photogenerated carriers to contribute to the photovoltaic efficiency. Here, time-resolved photoluminescence (PL) microscopy is used to directly probe ambipolar carrier diffusion and recombination kinetics in hybrid perovskites. This technique is applied to thin films of methylammonium lead tri-iodide (MAPbI3) obtained with two different fabrication routes, methylammonium lead tribromide (MAPbBr3), and an alloy of formamidinium lead tri-iodide (FAPbI3) and methylammonium lead bromide FA0.85MA0.15Pb(I0.85Br0.15)3. Average diffusion coefficients in the films leading to the highest device efficiencies and longest lifetimes, i.e., in FA0.85MA0.15Pb(I0.85Br0.15)3 and acetonitrile- processed MAPbI3, are found to be several orders of magnitude lower than in the other films. Further examination of the time-dependence shows strong evidence for non-diffusive transport. In particular, acetonitrile-processed MAPbI3 shows distinct diffusion regimes on short and long timescales with an effective diffusion constant varying over 2 orders of magnitude. Our results also highlight the fact that increases in carrier lifetime in this class of materials are not necessarily concomitant with increased diffusion lengths and that the PL quantum efficiency under solar cell operating conditions is a greater indication of material, and ultimately device, quality. 2 1. Introduction Solar energy, which draws upon our nearly infinite supply of solar irradiation, is widely considered to be one of our most important resources for a sustainable energy future. Significant efforts have recently been dedicated to improving a class of solution-processable, metal-halide perovskite materials that when incorporated into solar cells, can yield certified power conversion efficiencies above 23%[1]. Perovskite solar cells surpass their organic and dye- sensitized predecessors in terms of both performance and material cost.[2 -- 6] The prototypical 3D iodide-based perovskites possess low exciton binding energies, small surface and bulk non- radiative recombination rates, a bandgap close to the ideal Shockley-Queisser value for single junction solar cells, and composition dependent bandgap tunability.[7] The rapid increase in the power conversion efficiencies (PCEs) of perovskite solar cells has been unparalleled compared to other photovoltaic technologies.[8] However, there remain many questions regarding the physicochemical processes and optoelectronic properties leading to such high efficiencies. The nature of the photogenerated species, binding energies, as well as recombination pathways are being continuously re-examined.[9 -- 25] Long carrier diffusion lengths have been identified as an important aspect enabling high efficiencies in planar heterojunction architectures, but reports differ greatly with respect to reported values. In methylammonium lead iodide (MAPbI3) and MAPbI3-xClx thin films, i.e. MAPbI3 processed from a PbCl2 precursor to improve film quality, photoluminescence (PL) quenching and transient absorption measurements have been used to extract respective diffusion lengths of ~100 nm and ~1 µm in the direction perpendicular to the substrate. This corresponds to diffusion coefficients D = 0.01 -- 0.05 cm2s-1.[26,27] In such measurements, the reduced lifetime in the presence of a (perfectly) quenching interface for one carrier type is fit to the diffusion equation. Alternatively, the diffusion coefficient can be obtained via Einstein's relation, using measurements of the carrier mobility, µ. In MAPbI3 single crystals, time-of- flight and transport measurements have been used to extract diffusion lengths exceeding 175 3 µm and up to 3 cm under low fluence excitation.[28] In other single crystal perovskites, mobilities ranging from µ = 105 to 2000 cm2V-1s-1 have been measured using the Hall effect, corresponding to 𝐷 = 3 -- 50 cm2s-1.[29,30] In MAPbI3 thin films, time-resolved THz cm2V-1s-1,[33] corresponding to 𝐷 = 0.8 cm2s-1. Optical conductivity measurements tend to give conductivity[31] and microwave conductivity measurements[32] have found mobilities of ~30 slightly higher values than the other techniques given that they yield the sum of both carrier mobilities and probe the motion of charge carriers on the few-nanometer length scale. Although the direction of most importance to photovoltaic efficiencies is the perpendicular direction, lateral diffusion has been shown to be critical in perovskite films due to the impact of grain boundary recombination on device performance.[34] The study of lateral diffusion also lends itself well to spatially-resolved measurement techniques. For example, spatially-resolved photocurrent measurements have been used to extract diffusion lengths of 𝐿& = 10 -- 28 µm (electrons) and 27 -- 65 µm (holes) in MAPbI3 single crystals and to demonstrate unbalanced charge transport in MAPbBr3.[35] All-optical methods such as four-wave mixing,[36] the related light-induced transient grating technique[37] and transient absorption microscopy[38] have been used to analyze in-plane carrier diffusion on thin films without the need for quenching layers. These measurements have yielded values of 𝐿'(𝐷) of 1 µm (2 cm2s-1), 0.1 -- ambipolar diffusion coefficient, which under high-level injection is described as 𝐷= ⁄ 2𝐷+𝐷, 𝐷++𝐷, 0.3 µm (0.02 -- 0.07 cm2s-1), and 0.2 µm (0.05 -- 0.08 cm2s-1). These values of D correspond to the , where Dn and Dp are the electron and hole diffusion coefficients. The broad spread of values suggests both a need for further study, and for techniques which can unambiguously identify diffusion coefficients under a limited number of assumptions. In this work, we directly image the spreading of charge carriers in both time and space and extract the diffusion length from a fit to the 3-dimensional (3D) diffusion equation. Surprisingly, we find that samples with long, monomolecular lifetimes tend to be characterized by small diffusion 4 coefficients, resulting in negligible increases in the diffusion length as compared to the prototypical MAPbI3. Further examination of acetonitrile-processed MAPbI3 shows that in these films non-diffusive transport occurs, where a very large effective diffusion coefficient can be observed only in the first few nanoseconds, on a scale much shorter than the µs lifetime. In contrast, on a scale comparable to the lifetime, the effective diffusion coefficient is more than 2 orders of magnitude smaller. Further examination of the studied samples suggest that this effect is present in most of the films to a varying extent. We also highlight various effects and assumptions that can lead to artifacts in diffusion length measurements. A notable example is the apparent spatial spreading of charge carriers that can occur in the complete absence of diffusion when monomolecular and bimolecular recombination are both present. This effect is analogous to one recently observed in a study of exciton diffusion in monolayer transition metal dichalcogenides in the presence of exciton-exciton annihilation.[39] 2. Experiment To probe carrier diffusion, we record the spreading of luminescence in both space and time, after excitation with a diffraction-limited pump spot, which is used to generate the initial carrier distribution. The technique is depicted schematically in Figure 1a. This approach was originally introduced to study exciton diffusion in tetracene single crystals.[40] We implement a slight variant that uses a streak camera to avoid the need for any moving parts.[39] By fitting the time and space-dependent luminescence to the diffusion equation, we directly obtain the recombination and ambipolar diffusion coefficients of free carriers for a range of hybrid organic-inorganic perovskite thin films. spreading of the gaussian luminescence width 𝜎(𝑡), such that σ2(t) = σ1(0) + 2Dt. However, If only linear recombination is present, in-plane diffusion leads to a time-dependent a full solution to the diffusion equation becomes necessary when bimolecular recombination 5 occurs. In this case, the spreading is no longer determined by this simple relationship and apparent spreading occurs even in the case D = 0.[39] This is the regime typical of free carrier materials and that where our measurements and all previous in-plane measurements of perovskite thin films have been performed. Prior to performing measurements with a tightly focused pump, we first estimate the linear and bimolecular recombination coefficients by performing time-resolved measurements with a large excitation spot. Under these conditions, diffusion plays a reduced role which allows us to decouple diffusion from the apparent lifetime observed when measuring photoluminescence with a diffraction-limited pump. We then perform diffusion measurements by focusing a l = 532 nm laser on the sample surface, through the glass coverslip substrates using an oil-immersion objective to a nearly diffraction-limited spot of width σ = 120 nm (see Experimental Section for details). In this configuration, a slice through the center of the pump spot is imaged through the same objective onto the streak camera entrance slit. This allows us to obtain the spreading of the PL in both space and time as shown schematically on the right- hand side of Figure 1a. Note that the use of immersion oil also prevents any artificial spreading of the spatial profile due to the re-absorption of photons that are totally internally reflected in the glass. Here, we present results for MAPbI3 (MA) processed via antisolvent quenching, MAPbI3 processed using an acetonitrile/methylamine compound solvent (MA-ACN), the mixed-cation, mixed-halide perovskite FA0.85MA0.15Pb(I0.85Br0.15)3 (FAMA) and MAPbBr3 (BR). Scanning electron micrographs of the studied samples are shown in Figure 2. We note that the different processing routes and compositions lead to drastically different surface morphology and apparent grain size. For the samples showing clear grain boundaries, the excitation laser was centered within individual domains to minimize any scattering of the gaussian pump beam. Figure 1b shows the gaussian half-width (standard deviation) obtained by fitting the photoluminescence (PL) spatial profile to a gaussian profile at each time step. For all of the 6 samples, we observe a clear increase in the PL spot size as a function of time. This increase, however, is due to an interplay between bimolecular recombination and diffusion as will be shown below. The difference between the initial PL spot size and the pump size is due to photon recycling which occurs on the ps timescale.[41] Note that the degree of time-dependent spreading varies significantly amongst the various films. One strength of the streak-camera based technique as compared to pump-probe, which is highlighted in Figure 1b, is its adaptability to a wide range of lifetimes, ranging from sub-ns up to ~10 µs. 3. Model and Results From the data, the diffusion and recombination coefficients can be extracted by simultaneously fitting the entire data set (space and time) to the three-dimensional diffusion equation: ¶D 2 ¶ D - D ù ú û 2 ¶ D é ê ë , ) k n r z t 1 ( , , ) k n r z t 2 ( , 2 =- G r z t D , ) + ( , ( , , ) n r z t z 2 ¶ 1 ¶D - D r ( , , ) n r z t r ¶ ( , , ) n r z t + r 2 ¶ ( , , ) n r z t t ¶ (1) (band-to-band) recombination coefficients, respectively. Auger recombination can be safely ignored given the low initial carrier densities in our experiments (< 5 × 1017 cm-3). We assume Here, 𝐺(𝑟,𝑧,𝑡) is the carrier generation rate determined from the pump, 𝐷 is the (isotropic) diffusion constant; and 𝑘8 and 𝑘1 the monomolecular (Shockley-Read-Hall) and bimolecular high-level injection, i.e. Δ𝑛= Δ𝑝, where Δ𝑛 and Δ𝑝 correspond to the excess minority carrier density. This assumption is justified when Δ𝑛 exceeds the majority carrier density 𝑛< (or 𝑝<). Previously reported values are 𝑝< ~ 1015 cm-3 in MAPbI3 and MAPbI3-xClx thin films, and values ranging from 𝑝< = 109 -- 1012 cm-3 in MAPbI3 single crystals.[28,30,42] Samples from At the carrier densities relevant for most of our experiments, this is a reasonable assumption. synthetic routes and compositions which yield longer monomolecular lifetimes (fewer traps), such as MA-ACN and FAMA, are expected to possess even lower majority carrier densities. Alternatively, a quadratic dependence of the PL on the pump power, such as that observed BR, 7 can also be used to demonstrate the validity of the high-level injection assumption.[43] Because of the low exciton binding energy (<16 meV) in the materials under study, any excitonic contribution to the photoluminescence at the excitation densities utilised is considered negligible.[44] The reported diffusion coefficients correspond to the ambipolar D, which is dominated by the least diffusing species. Under the assumption of high-level injection, the PL can be calculated from the simulated carrier density using PL ∝Δ𝑛1 . The dependence of the recombination dominates at initial times, i.e. 𝑘1Δ𝑛(𝑡=0)>𝑘8. Given that all of the samples initial carrier density on the pump profile depends on the relative contributions of monomolecular vs bimolecular recombination. In all samples except BR, bimolecular are fabricated on glass and capped with PMMA thin films, we assume reflecting boundary conditions at the top and bottom interfaces. The fit obtained by a least-squares minimization of the error between our data and the calculated PL is shown in Figure 3 for MA at a peak excitation density, Δn(0) ≡ Δn(r = 0, z = 0, t = 0) = 2.5 × 1016 cm-3, and in Figure S2 for the remaining samples. Figure 3a shows a cross- sectional fit to the time-resolved data taken from the center of the spot, while Figure 3b shows spatially-resolved fits taken at various times. The time dependence and spatial spreading are both well-reproduced by the calculation. The full data sets and fits are shown for all the films as Videos S1-S5. The fits show excellent agreement with the model, which confirms that invoking the explicit presence of an exciton population or trap states is unnecessary for explaining the results under our measurement conditions (although they may be present). Some small discrepancies arise in the tails at longer times that appear to be due to deviations between the gaussian initial condition (t = 0) for the simulation and the real experimental profile (see Experimental Section). A summary of the diffusion and recombination coefficients obtained for the different samples is given in Table 1. The 1D monomolecular diffusion length, defined 8 as 𝐿'= ?𝐷/𝑘8, is also indicated. Peak carrier density ∆𝑛< for each measurement is given in For MA, we obtain a value of 𝐷 = (1.25 ± 0.01) × 10-1 cm2s-1 and a monomolecular diffusion length of 𝐿'= (1.68 ± 0.03) µm. Processing perovskite films from an Table S1. acetonitrile/methylamine compound solvent system leads to solar cell efficiencies of up to 19%.[45] Our calculations indicate that the diffusion coefficient in the MA-ACN sample is D = (6.9 ± 0.1) × 10-3 cm2s-1, which is two orders of magnitude lower than that of MA. The much smaller value k1, however, leads to a comparable monomolecular diffusion length 𝐿'= (1.73 ± 0.04) µm. The FAMA composition was engineered as a thermodynamically stable substitute to the FAPbI3 α-phase perovskite which was found to transition into a non-perovskite δ-phase at room temperature (Figure S3).[46] For FAMA, which is the composition of the highest- reported device efficiencies (~22%),[8] the extracted diffusion coefficient is even smaller, 𝐷 = (5.6 ± 0.1) × 10-4 cm2s-1 with a corresponding diffusion length of 𝐿'= (0.78 ± 0.04) µm. We coefficient of 𝐷= (5.08 ± 0.05) × 10-1 cm2s-1. Despite the high value of 𝐷, the short diffusion length 𝐿'= (0.96 ± 0.04) µm obtained for BR is a consequence of its very short lifetime. For have also performed measurements on BR (MAPbBr3) and our results show a large diffusion MA, which is the most studied film, the recombination coefficients agree with previously reported values,[33] with the caveat that we ignore photon recycling in the transverse direction, but implicitly account for it in the lateral direction. The diffusion coefficient we observe for MA is comparable to that obtained in several previous reports (~0.04 cm2s-1). [26,27,38] Our results point to an inverse relation between long lifetimes and D for both MA-ACN and FAMA which suggests weakly mobile, localized minority carriers as the source of the PL in long-lifetime films. When MAPbI3 is processed from the acetonitrile/methylamine solvent, we observe a much longer monomolecular lifetime, suggesting a reduced trap density, but at the same time, 9 a diffusion coefficient that is reduced by one order of magnitude compared to MA. The same trend is observed in FAMA, but with an even longer lifetime and smaller diffusion coefficient. 4. Potential Artifacts in Time-Resolved PL Measurements Two potential artifacts need to be considered when analyzing the kinetics. First, we note that diffusion can strongly modify the time-dependent luminescence, particularly in geometries with a tightly focused pump such as in confocal fluorescence lifetime measurements.[47] Figure 4a shows a series of linear (monomolecular) PL transients calculated for increasing values of the diffusion coefficient. Diffusion outside of the excitation spot leads to apparent non-linear behavior, which is simply an artefact due to the spatial dynamics. For this reason, initial conditions for our fits were obtained by first measuring k1 and k2 using measurements performed with a large pump spot (with D = 0). These values were then refined using the tightly focused data. Second, bimolecular recombination leads to an artificial spreading of the PL spatial profile in time. To illustrate this, Figure 4b shows the calculated spreading of the gaussian width of the spatial profile for our samples using recombination coefficients from Table 1, but in the absence of diffusion (D = 0). The figure shows a large increase of the gaussian width that is purely due to k2. This is analogous to a recent demonstration that exciton-exciton annihilation leads to increases in the apparent D for excitonic materials.[39] This effect becomes more important for increasing carrier densities. At the densities relevant for our measurements (1016- 1017 cm-3), both D and k2 play important roles in dictating the time-dependent spreading. This is always the case under high-level injection for D and k2 values typical of perovskite thin films. As a result, one cannot directly use the gaussian width as a function of time to extract the diffusion constant. Note also that the traditional transient grating decay formula, which has previously been used for measuring in-plane diffusion in perovskite films, has not been properly modified to account for bimolecular recombination.[36,48,49] 10 5. Photon Recycling Finally, photon recycling also plays an important role in dictating the time-dependent behavior in the directions perpendicular and parallel to the substrate. Recent studies show that in films thicker than the absorption length, the measured 𝑘1 can differ from the intrinsic 𝑘1 by up to one order of magnitude due to photon recycling in the direction perpendicular to the substrate.[41,42] In the simplest picture, emission and re-absorption in this redistributes the initial level injection and when bimolecular recombination dominates, the initial carrier density in- carrier density over the entire thickness of the film on a sub-nanosecond time scale. Here, we have only reported the apparent 𝑘1, without correcting for this effect. In principle, under high- plane should agree with that obtained from the laser profile using Δn(0) ∝?𝑃(𝑟,0), where 𝑃(𝑟,𝑡) is the laser pump profile. However, the mismatch between the laser (𝜎C≈120 nm) and the photoluminescence (𝜎FC,GH<≈300−580 nm) gaussian widths is evident from Figure 1c. To understand the difference in initial widths, we have measured the spreading of the spatial profile of MA-ACN at the highest resolution possible in our system (~50 ps) (Figure S2 - insets g and h). The spreading of the spatial profiles is also shown side by side for the short and long timescales in Figure 5. Even at this time resolution, there is a large mismatch between 𝜎C= 120 nm and 𝜎FC,GH<=444 nm. This is indicative of very fast spreading on a < 50 ps timescale. One can show that this agrees well with the anticipated spreading of ~365 nm from a single photon recycling event (see Supplementary). When exciting at energies high above the bandgap, Guo et al. have recently observed fast spreading of carriers over ~100 nm in only 1.2 ps in MA. Although this effect may contribute to the spreading, the spectral redshift that we observe towards the edge of the PL spot (not shown) and the absence of a strong dependence on the pump wavelength is more indicative of photon recycling. In any case, we implicitly correct for the initial in-plane spreading by using the t = 0 PL width to determine the initial carrier density. This leads to much lower initial carrier densities than those that would be 11 obtained directly from the pump size. One could appropriately include photon recycling in the theoretical model or modify the initial transverse density to account for redistribution. However, we have verified that although such changes can affect 𝑘1 by up to an order of magnitude, the effect on D is negligible (Table S2). The values for MA in Table 1 coincide well with apparent values of k2 measured for films of similar thickness although one should be wary of the various assumptions made when comparing different values.[41,42] Note, for example, the importance of correctly calculating the carrier density from the pump profile when measuring the bimolecular recombination coefficient. Under high-level injection and at carrier densities where bimolecular recombination dominates, Δn(0) is given by the square root of the pump profile in space (or of the t = 0 PL profile) . The spatial width of Δn(0) thus differs by a factor √2 from that of the pump (or PL). 6. Non-Diffusive Transport 𝐷= (1.559 ± 0.006) cm2s-1 for MA-ACN. This high value of 𝐷, which corresponds to a For the shorter time window in Figure 5, our modelling gives a diffusion coefficient of mobility of ~60 cm2V-1s-1 is more characteristic of intrinsic values measured on single crystals or over short distances using THz conductivity measurements. Our results thus point to two distinct diffusive regimes in the longer lifetime film. Very fast initial diffusion over a time window nearly 1000 times shorter than the monomolecular lifetime, followed by very slow diffusion on longer timescales. One can systematically investigate this effect for all of the film compositions. Figure 6 shows best fits to the diffusion coefficient over distinct time slices from the data (dashed lines), with k1 and k2 fixed to the values in Table 1. Although the weighted diffusion coefficients are similar to those obtained using the global fit (Table 1), we observe a clear time-dependent decrease of the diffusion coefficient for all of the films except FAMA. The latter is characterized by a very small diffusion constant (undistinguishable from D~0 cm2s- 1) for all times other than the one shown in the figure. We suspect that these trends are due to 12 very efficient trapping of the initial rapidly diffusing carriers on defects and at grain boundaries, followed by slower de-trapping/re-trapping. This hypothesis is supported by recent measurements of the microscopic surface potential that have found very large potential fluctuations for MA-ACN (~80 meV) as compared to antisolvent-quenched MA (~20 meV).[50] Ščajev et al. have also observed a strong dependence of D on the initial carrier density and suggested the presence of band-like and localized diffusion mechanisms. The latter was found to be dominant in mixed-halide films and was attributed to a diffusion mechanism that is dominated by carrier hopping and delocalization in localized states due to alloy disorder.[37] Localization due to alloy disorder in the mixed-halide, mixed-cation FAMA film might thus be another source of the reduced D. 8. Relevance to Photovoltaic Efficiencies The diffusion length and the photoluminescence quantum efficiency (PLQE) as a function of the photogenerated carrier density, can also be calculated from the recombination validates the assumption of high-level injection. We thus estimate the PLQE as the ratio of shown in Figure 7b ignores Auger recombination and any possible change in the coefficients themselves that can occur due to phase-space filling at high carrier densities.[37] We note that MA-ACN and FAMA show the highest PLQE values at carrier densities typical of solar cell operating conditions (~1015 cm-3). Given that these film compositions show the best device 13 MA, MA-ACN and BR show the highest diffusion lengths. By performing PLQE coefficients shown in Table 1. In Figure 7a, the diffusion length is calculated as a function of the photogenerated carriers via 𝐿',OPP= ?𝐷/𝑘OPP, where 𝑘OPP=𝑘8+𝑘1Δ𝑛, and as in Table 1, that the radiative (low-level injection) component of 𝑘8 is negligible (≲1%), which further bimolecular recombination to total recombination rate, i.e. 𝑘1Δ𝑛/(𝑘8+𝑘1Δ𝑛). The result measurements in an integrating sphere at low carrier density (~10 µW/cm2), we have confirmed performance, the low non-radiative loss appears to be the key factor to attaining high- performance optoelectronic devices.[51] 8. Conclusions In summary, we have directly imaged the diffusion of charge carriers in a range of hybrid organic-inorganic perovskite thin films. Our 3D diffusion model shows excellent agreement with experimental data and is used to extract the diffusion and recombination coefficients. Surprisingly, the application of this technique to film compositions that show the best device performance such as FA0.85MA0.15Pb(I0.85Br0.15)3 and ACN processed MAPbI3 reveal the smallest diffusion coefficients. Furthermore, we have observed that in the latter, very efficient diffusion does occur, but only on a scale much shorter than the lifetime. Time-resolved fitting of the diffusion coefficient shows strong evidence for non-diffusive transport. In all of the films, except for FAMA, we observe a notable decrease of D as a function of time, on a scale shorter than the carrier lifetime, In general, we find an inverse relationship between carrier lifetime and diffusion coefficient, which may be due to the presence of weakly localized minority carriers in the longest lifetime films. In contrast, we find a strong correlation between the PLQE at carrier densities typical of solar cell operating conditions and reported device performances,[8,45] which indicates that reduced non-radiative loss plays a much greater role than the diffusion length in determining solar cell efficiencies. Experimental Section Film fabrication: For the mixed-cation, mixed halide perovskite, a 1.25 M solution of FA0.85 MA0.15Pb(I0.85Br0.15)3 (FAMA) was prepared using a 4:1 v:v ratio of DMF:DMSO. The solution was spin-coated onto the desired substrate at 1000 rpm for 10 s, followed by 6000 rpm for 35 s. 100 µL of anisole was dropped onto the substrate 35 s after the beginning of the spin-coating. The films were then annealed at 100 °C for 60 min. The same protocol was repeated for the 14 MAPbI3 (MA) and MAPbBr3 (BR) samples, where the concentration of the precursor solutions was 1 M. In the case of the acetonitrile processed MAPbI3 (MA-ACN), a solution of methylamine in ethanol (Sigma Aldrich, 33 wt%) was placed into an aerator which was kept in an ice bath. A carrier gas, N2, was then bubbled into the solution, thus degassing the solution of methylamine. The produced methylamine gas was then passed through a drying tube filled with a desiccant (Drierite and CaO) before being bubbled directly into the acetonitrile (Sigma Aldrich) which contained the perovskite precursors (MAI:PbI2 ratio of 1:1.06 M) at a concentration of 0.5 M. The gas was bubbled into the black dispersion until the perovskite particles were wholly dissolved resulting in a clear, light yellow solution. Streak camera measurement: A schematic of the optical setup alongside explanations is shown in Figure S4. The Fianium supercontinuum laser has a tunable repetition rate of 0.1 to 40 MHz with a pulse width of ~48 ps. The repetition rate used for each sample is shown in Table S1. The Hamamatsu C10910 streak camera is equipped with a M10913-11 triggering unit which has a < 20 ps resolution in the smallest time window of 1 ns. The experiments are done in photon counting mode at 10 ms exposure and maximum MCP gain. An electronically delayed signal from the laser system, obtained using a Stanford DG645 delay unit, is used to trigger the streak camera. Fitting procedure: The initial carrier density profile is obtained via a CCD camera image of the pump laser profile. The peak density of excess photogenerated charge carriers ∆𝑛< is obtained from: 0) dV n D=D = total ( , , n r z t =Dò Where Δ𝑛GRGST=𝑃U 𝐸,W⁄ is the total excess charge carriers obtained via the pulse energy 𝑃U and the photon energy 𝐸,W=ℎ𝑐< 𝜆C⁄ . Assuming high level injection. i.e. PL ∝Δ𝑛1 , the (2) ( ) ( ) f r g z rdrdz V n 0 2 p¥ st ò ò ò 0 0 0 15 is the gaussian , where 𝜎FC,GH< 1 is given by the square root of the Beer-Lambert equation for light width. Here, since the gaussian width of the measured spot is much larger than that of the imaging system PSF at the emitted wavelength, we safely ignore its contribution. The profile horizontal profile of Δ𝑛(𝑟,𝑧,𝑡=0), derived from square root of the gaussian fit of the radial photoluminescence profile at t =0, is given by 𝑓(𝑟)=𝑒−𝑟2 4𝜎𝑃𝐿,𝑡=0 2] along z-axis 𝑔(𝑧)=𝑒_' 1Ga⁄ attenuation. Here, 𝑡b is the absorption length at laser wavelength 𝜆C, whereas 𝑡c in equation (2) in the following expression for ∆𝑛<: Parameters 𝜆C, 𝑡b, 𝑡c, and ∆𝑛< are shown in Table S1 for each sample. The obtained ∆𝑛< value (r-z axes) data at time 𝑡 = 0. Following the forward Euler discretization scheme, this array is in equation (3) and spatial profiles presented in equation (2) are subsequently used to fit the 2D is the physical thickness of the sample. Resolving the integral with the substituted terms results (3) n D total ( 1 t - , 0 a = - t s /2 t a e ) n D 0 = 8 2 ps PL t then used as an initial condition to evaluate the subsequent time steps of the diffusion equation in the deconvoluted domain. The resulting 3D array from the simulation, integrated along the z-axis, produces a 2D (r, t) intensity matrix for the same spatiotemporal coordinates as the streak camera data. The difference between the normalized model and experimental profiles produces an error value for a given set of recombination and diffusion coefficients. The fit coefficients are then found by least squares minimization of the error function. Acknowledgements Stéphane Kéna-Cohen acknowledges support from the NSERC Discovery Grant Program and the Canada Research Chair in Hybrid and Molecular Photonics. Aravindan Sridharan acknowledges support from NSERC PGS-D as well as Institut d'Énergie Trottier (IET) grant programs. Nakita K. Noel acknowledges funding via a fellowship from the Princeton Center 16 for Complex Materials (PCCM). Barry P. Rand acknowledges support from the Office of Naval Research (ONR) Young Investigator Program (Award #N00014-17-1-2005). 17 References [1] "NREL: National Renewable Energy Laboratory Efficiency Chart," 2018. [2] Y. Deng, E. Peng, Y. Shao, Z. Xiao, Q. Dong, J. Huang, Energy Environ. Sci. 2015, 8, 1544. [3] M. Hambsch, Q. Lin, A. Armin, P. L. Burn, P. Meredith, J. Mater. Chem. A 2016, 4, 13830. [4] S. Razza, F. Di Giacomo, F. Matteocci, L. Cinà, A. L. Palma, S. Casaluci, P. Cameron, A. D'Epifanio, S. Licoccia, A. Reale, T. M. Brown, A. Di Carlo, J. Power Sources 2015, 277, 286. [5] J. H. Kim, S. T. Williams, N. Cho, C. C. Chueh, A. K. Y. Jen, Adv. Energy Mater. 2015, 5, 2. [6] A. T. Barrows, A. J. Pearson, C. K. Kwak, A. D. F. Dunbar, A. R. Buckley, D. G. Lidzey, Energy Environ. Sci. 2014, 7, 2944. [7] J. H. Noh, S. H. Im, J. H. Heo, T. N. Mandal, S. Il Seok, Nano Lett. 2013, 13, 1764. [8] W. S. Yang, B.-W. Park, E. H. Jung, N. J. Jeon, Y. C. Kim, D. U. Lee, S. S. Shin, J. Seo, E. K. Kim, J. H. Noh, S. Il Seok, Science 2017, 356, 1376. [9] A. R. Srimath Kandada, A. Petrozza, Acc. Chem. Res. 2016, 49, 536. [10] Y. Yamada, T. Nakamura, M. Endo, A. Wakamiya, Y. Kanemitsu, J. Am. Chem. Soc. 2014, 136, 11610. [11] V. D'Innocenzo, G. Grancini, M. J. P. Alcocer, A. R. S. Kandada, S. D. Stranks, M. M. Lee, G. Lanzani, H. J. Snaith, A. Petrozza, Nat. Commun. 2014, 5, 3586. [12] P. Piatkowski, B. Cohen, J. F. Ramos, D. Nunzio, M. R. Di Nunzio, N. Mohammad K., M. Grätzel, S. Ahmad, D. Nunzio, Phys. Chem. Chem. Phys. 2015, 17, 14674. [13] M. Saba, M. Cadelano, D. Marongiu, F. Chen, V. Sarritzu, N. Sestu, C. Figus, M. Aresti, R. Piras, A. Geddo Lehmann, C. Cannas, A. Musinu, F. Quochi, A. Mura, G. Bongiovanni, Nat. Commun. 2014, 5, 5049. 18 [14] V. Sarritzu, N. Sestu, D. Marongiu, X. Chang, S. Masi, A. Rizzo, S. Colella, F. Quochi, M. Saba, A. Mura, G. Bongiovanni, Sci. Rep. 2017, 7, 44629. [15] E. Edri, S. Kirmayer, S. Mukhopadhyay, K. Gartsman, G. Hodes, D. Cahen, Nat. Commun. 2014, 5, 1. [16] C. Wehrenfennig, M. Liu, H. J. Snaith, M. B. Johnston, L. M. Herz, Energy Environ. Sci. 2014, 7, 2269. [17] C. Wehrenfennig, M. Liu, H. J. Snaith, M. B. Johnston, L. M. Herz, APL Mater. 2014, 2, DOI 10.1063/1.4891595. [18] K. T. Munson, C. Grieco, E. Kennehan, R. J. Stewart, J. B. Asbury, ACS Energy Lett. 2017, 2, 651. [19] M. Saba, F. Quochi, A. Mura, G. Bongiovanni, Acc. Chem. Res. 2016, 49, 166. [20] T. C. Sum, S. Chen, G. Xing, X. Liu, B. Wu, Nanotechnology 2015, 26, 342001. [21] L. M. Herz, Annu. Rev. Phys. Chem. 2016, 67, 65. [22] S. D. Stranks, V. M. Burlakov, T. Leijtens, J. M. Ball, A. Goriely, H. J. Snaith, Phys. Rev. Appl. 2014, 2, 034007. [23] G.-J. A. H. Wetzelaer, M. Scheepers, A. M. Sempere, C. Momblona, J. Ávila, H. J. Bolink, Adv. Mater. 2015, 27, 1837. [24] A. Baumann, S. Väth, P. Rieder, M. C. Heiber, K. Tvingstedt, V. Dyakonov, J. Phys. Chem. Lett. 2015, 6, 2350. [25] G. Giorgi, J. I. Fujisawa, H. Segawa, K. Yamashita, J. Phys. Chem. Lett. 2013, 4, 4213. [26] S. D. Stranks, S. D. Stranks, G. E. Eperon, G. Grancini, C. Menelaou, M. J. P. Alcocer, T. Leijtens, L. M. Herz, A. Petrozza, H. J. Snaith, Science 2014, 342, 341. [27] G. Xing, N. Mathews, S. S. Lim, Y. M. Lam, S. Mhaisalkar, T. C. Sum, Science 2013, 6960, 498. [28] Q. Dong, Y. Fang, Y. Shao, P. Mulligan, J. Qiu, L. Cao, J. Huang, Science 2015, 347, 967. 19 [29] C. C. Stoumpos, C. D. Malliakas, M. G. Kanatzidis, Inorg. Chem. 2013, 52, 9019. [30] Y. Chen, H. T. Yi, X. Wu, R. Haroldson, Y. N. Gartstein, Y. I. Rodionov, K. S. Tikhonov, A. Zakhidov, X. Y. Zhu, V. Podzorov, Nat. Commun. 2016, 7, 12253. [31] C. Wehrenfennig, G. E. Eperon, M. B. Johnston, H. J. Snaith, L. M. Herz, Adv. Mater. 2014, 26, 1584. [32] C. S. Ponseca, Y. Tian, V. Sundström, I. G. Scheblykin, Nanotechnology 2016, 27, 082001. [33] M. B. Johnston, L. M. Herz, Acc. Chem. Res. 2016, 49, 146. [34] Y. Yang, Y. Yan, M. Yang, S. Choi, K. Zhu, J. M. Luther, M. C. Beard, Nat. Commun. 2015, 6, 7961. [35] G. A. Elbaz, D. B. Straus, O. E. Semonin, T. D. Hull, D. W. Paley, P. Kim, J. S. Owen, C. R. Kagan, X. Roy, Nano Lett. 2017, 17, 1727. [36] D. Webber, C. Clegg, A. W. Mason, S. A. March, I. G. Hill, K. C. Hall, Appl. Phys. Lett. 2017, 111, 121905. [37] P. Ščajev, R. Aleksiejunas, S. Miasojedovas, S. Nargelas, M. Inoue, C. Qin, T. Matsushima, C. Adachi, S. Juršenas, J. Phys. Chem. C 2017, 121, 21600. [38] Z. Guo, J. S. Manser, Y. Wan, P. V. Kamat, L. Huang, Nat. Commun. 2015, 6, 7471. [39] M. Kulig, J. Zipfel, P. Nagler, S. Blanter, C. Schüller, T. Korn, N. Paradiso, M. M. Glazov, A. Chernikov, Phys. Rev. Lett. 2018, 120, 1. [40] G. M. Akselrod, P. B. Deotare, N. J. Thompson, J. Lee, W. a Tisdale, M. a Baldo, V. M. Menon, V. Bulović, Nat. Commun. 2014, 5, 3646. [41] T. W. Crothers, R. L. Milot, J. B. Patel, E. S. Parrott, J. Schlipf, P. Muller-Buschbaum, M. B. Johnston, L. M. Herz, Nano Lett. 2017, 17, 5782. [42] F. Staub, H. Hempel, J. C. Hebig, J. Mock, U. W. Paetzold, U. Rau, T. Unold, T. Kirchartz, Phys. Rev. Appl. 2016, 6, 1. [43] C. M. Sutter-Fella, Y. Li, M. Amani, J. W. Ager, F. M. Toma, E. Yablonovitch, I. D. 20 Sharp, A. Javey, Nano Lett. 2016, 16, 800. [44] A. Miyata, A. Mitioglu, P. Plochocka, O. Portugall, J. T.-W. Wang, S. D. Stranks, H. J. Snaith, R. J. Nicholas, Nat Phys 2015, 11, 582. [45] N. K. Noel, S. N. Habisreutinger, B. Wenger, M. T. Klug, M. T. Hörantner, M. B. Johnston, R. J. Nicholas, D. T. Moore, H. J. Snaith, Energy Environ. Sci. 2017, 10, 145. [46] N. Pellet, P. Gao, G. Gregori, T. Y. Yang, M. K. Nazeeruddin, J. Maier, M. Grätzel, Angew. Chemie - Int. Ed. 2014, 53, 3151. [47] D. W. DeQuilettes, S. Jariwala, S. Burke, M. E. Ziffer, J. T. W. Wang, H. J. Snaith, D. S. Ginger, ACS Nano 2017, 11, 11488. [48] P. Ščajev, C. Qin, R. Aleksieju nas, P. Baronas, S. Miasojedovas, T. Fujihara, T. Matsushima, C. Adachi, S. Juršėnas, J. Phys. Chem. Lett. 2018, 9, 3167. [49] J. R. Salcedo, A. E. Siegman, D. D. Dlott, M. D. Fayer, Phys. Rev. Lett. 1978, 41, 131. [50] A. J. Ramadan, N. K. Noel, S. Fearn, N. Young, M. Walker, L. A. Rochford, H. J. Snaith, Chem. Mater. 2018, 30, 7737. [51] R. T. Ross, J. Chem. Phys. 1967, 46, 4590. [52] J. M. Richter, M. Abdi-Jalebi, A. Sadhanala, M. Tabachnyk, J. P. H. Rivett, L. M. Pazos-Outón, K. C. Gödel, M. Price, F. Deschler, R. H. Friend, Nat. Commun. 2016, 7, 506. 21 Figure 1. a) Left: Schematic of the experiment showing the incoming laser pulse generating electron-hole pairs which diffuse and subsequently recombine, emitting photons that contain information on the spatial distribution of carriers. Right: A streak camera measurement showing spreading of charge carriers in time and space for a film of MAPbI3 processed using an acetonitrile/methylamine (MA-ACN) compound solvent. Data are normalized at each time step to highlight the spreading. b) Log-log plot showing spreading of the standard deviation obtained from a gaussian fit to the spatial profiles as a function of time for samples MA (MAPbI3), MA- ACN on both long and short timescales, FAMA (FA0.85MA0.15Pb(I0.85Br0.15)3), and BR (MAPbBr3). The circles and solid lines correspond to raw streak camera data and modelled results respectively. Figure 2. SEM images of the studied samples. a) Antisolvent-processed MAPbI3, b) Acetonitrile-processed MAPbI3, c) FA0.85MA0.15Pb(I0.85Br0.15)3 and d) MAPbBr3. 22 Figure 3. a) Photoluminescence transient measured in the center of the spatial profile, i.e. spatial position of 0 µm, and b) Spatial profiles (horizontal slices) for different times extracted from 3D diffusion model-based fitting and experimental data for the MA sample. Figure 4. a) Impact of the diffusion coefficient on the linear lifetime i.e. with monomolecular recombination only (k1 = 1.02 × 105 s-1), in the tightly-focused regime. The rates for bimolecular and Auger recombination are neglected to highlight the significant impact of diffusion alone on the initial part of the lifetime curve. b) Modelled spreading of the gaussian width due to bimolecular recombination coefficient k2 in the absence of diffusion (D = 0). 23 Figure 5. Spatial profiles of the PL for the MA-ACN sample at the times indicated in the legend. These are shown a) over a timescale comparable to the monomolecular lifetime and b) at very short times. Figure 6. Diffusion coefficient, D, as a function of time obtained by fitting the data over different time sections, whose ranges are shown as horizontal dashed lines. The full lines are guides to the eye. For MA-ACN the result of the short-time data has been combined with that of the longer time data set. Only one data point is shown for FAMA since D ≈ 0 cm2s-1 reproduces the data very well for all later time sections. 24 Figure 7. a) Diffusion length and b) PLQE as a function of photogenerated carrier density calculated from the coefficients in Table 1 for various metal-halide perovskite films. In both graphs, the lines for BR were calculated using the previously reported value k2 = 7 × 10-11 cm3s- 1.[52] 25 Table 1. Diffusion and recombination coefficients obtained through iterative fitting of streak camera data for various perovskite films. Error values correspond to 95% confidence intervals obtained from the fits. The value of k2 is not corrected for photon recycling in the transverse direction and is strongly dependent on the approach used to determine the initial carrier density (see text). For MA, MA-ACN and FAMA, the values of k1 were obtained using separate measurements with larger spot sizes. Sample k2 D k1 [s-1] -- (4.4 ± 0.1) × 106 (2.331 ± 0.003) × 105 (7.38 ± 0.06) × 10-10 (8.58 ± 0.08) × 10-10 LD [µm] 1.68 ± 0.03 1.73 ± 0.04 -- [cm3s-1] -- -- MA MA-ACN MA-ACNa) [cm2s-1] (1.25 ± 0.01) × 10-1 (6.9 ± 0.1) × 10-3 (1.56 ± 0.01) (5.6 ± 0.1) × 10-4 (5.08 ± 0.05) × 10-1 FAMA BRb) 0.78 ± 0.01 0.96 ± 0.01 a) Measured in the short 1 ns time window; b) In BR, k2 could not be obtained due to the dominance of the monomolecular regime k1 >> k2n under the experimental conditions. (9.25 ± 0.08) × 104 (5.55 ± 0.05) × 107 (1.23 ± 0.06) × 10-9 26 Supporting Information Derivation of gaussian spreading due to a single photon recycling event The photon diffusion coefficient is given by: Where 𝑐 is speed of light, 𝑛c is the refractive index of the perovskite film and 𝛼e the absorption at wavelength 𝜆. The average photon lifetime is given by: c 3 s n a D l = l t ph = n s c l a The average diffusion length of a re-emitted photon prior to reabsorption can then be obtained as: l = D t = L D l 1 1 3 a l n c s 3 c n a a s l Taking 𝛼e=2.24 ×10h cm-1 for MA-ACN at the peak emission wavelength, i.e. 𝜆=754.2 nm (Figure S1), we obtain 𝐿'j=2.58 ×10_k cm. We can estimate the temporal spreading of the gaussian width via σ2(t) = σ1(0) + 2D𝜏𝑝ℎ= σ1(0) + 2𝐿𝐷𝜆2 , or: Substituting the above value of 𝐿'j, we obtain Δ𝜎,W.mOnonTp+q=3.65×10_k cm or 365 nm. s s s D=-= ph recycling L D l 2 ( ) t 2 (0) 2 = . 27 Figure S1. Absorption and emission profiles used to calculate the spatial profile at starting time step of simulation. 28 Figure S2. Shown for samples MA-ACN (long and short timescales), FAMA, and BR are the time decay profile at the central vertical slice, i.e. spatial position of 0 µm, in a), c), e) and g); spatial profiles (horizontal slices) for different times extracted from 3D diffusion model-based fitting and experimental data in b), d), f) and h). 29 Figure S3. XRD data showing FAPbI3 in pure α-phase (experimental condition), and with appearance of δ-phase peaks (near ~12º). 30 Figure S4. Schematic of experimental setup. The optical components for beam shaping are the following: L1 -- 150 mm planoconvex lens, L2 -- -75 mm planoconcave lens, -- 50 mm aspheric lens, L4 -- 200 mm biconvex lens, L5 -- 100 mm achromatic doublet lens, L6 -- 150 mm achromatic doublet lens. L1 and L2 first reducing the beam diameter to half of the original size before focusing onto the pinhole using L3. A 25 µm high-intensity pinhole is used as a spatial filter, which produces a smooth beam profile, between L3 and L4. Subsequently, L4 then re- collimates the beam. Magnification values of 4 and 1.5 are induced by the lens pairs L3-L4 and L5-L6, resulting in a total magnification of 3 from the laser output. Lens L7 is used to focus the light reflected from the first beam splitter onto the power meter. The remaining laser power alongside white light is coupled into the microscopy setup for spatiotemporal imaging through the second beam splitter. An Olympus coverslip-corrected plan apochromatic 60X oil immersion objective of 1.42 NA and 0.15 mm working distance serves the dual purpose of firstly focusing the beam on the sample and secondly, relaying back the collected image from the surface onto the streak/CCD camera through a 200 mm EFL tube lens. 31 Table S1. Experimental conditions and calculated initial parameters for fitting procedure described in experimental section shown for the perovskite thin film samples. Laser wavelength Sample thickness Absorption length Electron density 𝝀𝑳 [nm] 532 532 532 450 𝒕𝒔 [nm] 398 353 486 290 Sample name MA MA-ACN FAMA BR 𝒕𝜶 [nm] 172 161 192 251 [1016 cm-3] 𝒏𝟎 2.49 4.29 1.25 0.76 Laser repetition rate (60x objective) [MHz] 2 0.5 0.1 20 32 excess carrier densities Δ𝑛<. These changes in the initial charge carrier densities simulate Table S2. Simulation results for Sample MA showing fitted coefficient D for varying initial photon recycling conditions. Δn0 [1017 cm-3] 1.5 2.0 2.5 3.0 3.5 D [10-1 cm2s-1] (1.29 ± 0.01) (1.27 ± 0.01) (1.25 ± 0.01) (1.23 ± 0.01) (1.21 ± 0.01) 33
1907.11453	1	1907	2019-07-26T09:32:38	Switchable Organic Plasmonics with Conductive Polymer Nanoantennas	[ "physics.app-ph", "cond-mat.mes-hall", "physics.optics" ]	Metal nanostructures are key elements in nanooptics owing to their strong resonant interaction with light through local plasmonic charge oscillations. Their ability to shape light at the nanoscale have made them important across a multitude of areas, including biosensing, energy conversion and ultrathin flat metaoptics. Yet another dimension of avenues is foreseen for dynamic nanoantennas, ranging from tuneable metalenses for miniaturized medical devices to adaptable windows that control radiation flows in and out of buildings. However, enabling nano-optical antennas to be dynamically controllable remains highly challenging and particularly so for traditional metals with fixed permittivity. Here we present state-of-the-art conductive polymers as a new class of organic plasmonic materials for redox-tuneable nano-optics. Through experiments and simulations, we show that nanodisks of highly conductive polymers can provide clear optical extinction peaks via excitation of dipolar localised surface plasmon resonances. Resonance frequencies redshift with increasing nanodisk aspect ratio, in agreement with analytical calculations based on dipolar polarizability theory. We furthermore demonstrate complete switching of the optical response of the organic nanoantennas by chemical tuning of the polymer's redox state, which effectively modulates the material permittivity between plasmonic and non-plasmonic regimes. Our results thereby show that conductive polymer nanostructures can act as redox-tuneable plasmonic nanoantennas, based on bipolaronic charge carriers rather than electrons as in conventional metals. Future directions may investigate different polymers and geometries to further widen the plasmonic spectral range (here around 0.8 to 3.6 {\mu}m) as well as different ways of tuning.	physics.app-ph	physics	Switchable Organic Plasmonics with Conductive Polymer Nanoantennas Shangzhi Chen1, Evan S. H. Kang1, Mina S. Chaharsoughi1, Vallery Stanishev2, Philipp Kühne2, Hengda Sun1, Vanya Darakchieva2 and Magnus P. Jonsson1★ 1Laboratory of Organic Electronics, Department of Science and Technology (ITN), Linköping University, SE-601 74 Norrköping, Sweden. 2Terahertz Materials Analysis Center (THeMAC) and Center for III-N Technology, C3NiT - Janzèn, Department of Physics, Chemistry and Biology (IFM), Linköping University, SE-581 83 Linköping, Sweden. ★e-mail: [email protected] Metal nanostructures are key elements in nanooptics owing to their strong resonant interaction with light through local plasmonic charge oscillations. Their ability to shape light at the nanoscale have made them important across a multitude of areas, including biosensing1, energy conversion2 and ultrathin flat metaoptics3. Yet another dimension of avenues is foreseen for dynamic nanoantennas4, ranging from tuneable metalenses for miniaturized medical devices to adaptable windows that control radiation flows in and out of buildings. However, enabling nano-optical antennas to be dynamically controllable remains highly challenging and particularly so for traditional metals with fixed permittivity4. Here we present state-of-the-art conductive polymers as a new class of organic plasmonic materials for redox-tuneable nano-optics. Through experiments and simulations, we show that nanodisks of highly conductive polymers can provide clear optical extinction peaks via excitation of dipolar localised surface plasmon resonances. Resonance frequencies redshift with increasing nanodisk aspect ratio, in agreement with analytical calculations based on dipolar polarizability theory. We furthermore demonstrate complete switching of the optical response of the organic nanoantennas by chemical tuning of the polymer's redox state, which effectively modulates the material permittivity between plasmonic and non- plasmonic regimes. Our results thereby show that conductive polymer nanostructures can act as redox-tuneable plasmonic nanoantennas, based on bipolaronic charge carriers rather than electrons as in conventional metals. Future directions may investigate different polymers and geometries to further widen the plasmonic spectral range (here around 0.8 to 3.6 μm) as well as different ways of tuning. 1 Fig. 1 Material properties and calculated plasmonic resonances for PEDOT:OTf in its high-conductivity oxidized state. a, Chemical structure of PEDOT:OTf. b, In-plane permittivity dispersion of PEDOT:OTf in its oxidized state (blue curve: real part; red curve: imaginary part). The shaded spectral range between 0.8 to 3.6 μm is defined as plasmonic regime where the real permittivity is below zero and its magnitude is larger than the imaginary component (inset: absolute magnitude of real and imaginary permittivity). c, Simulated extinction spectrum for a PEDOT:OTf nanodisk array (blue curve), with disk thickness of 30 nm, diameter of 500 nm, and periodicity of 1000 nm. The red curve shows the extinction for a non-structured thin PEDOT:OTf film scaled to the same material coverage as the disks (scaled by π/16). Inset: a schematic illustration of a PEDOT:OTf nanodisk (blue) on glass substrate (grey) with indicated x-, y-, and z-axes. d, e, Calculated nearfield profiles at the wavelength of the extinction maximum (2.9 μm) for one of the PEDOT:OTf nanodisks of the array in c: d, x-y in- plane direction 2 nm above the nanodisk; e, x-z cross-section through the center of the nanodisk. The color scale bars show the electric field strength relative to the incident light. Conductive polymers are conjugated materials with electrical conduction originating from (bi-) polaronic charge carriers situated along their backbones.5 Materials based on poly[3,4- ethylenedioxythiophene] (PEDOT), especially PEDOT doped with trifluoromethanesulfonate 2 (PEDOT:OTf), have shown particularly high electrical conductivity and metallic character6,7. We base our study on PEDOT:OTf (chemical formula in Fig. 1a) thin films produced by vapour phase polymerization (see Methods) and treated by sulphuric acid to further enhance the conductivity beyond 5000 S/cm (see Table. S1). The possibility for resonant light-matter interaction in these materials is governed by their complex permittivity 𝜀(𝜆), which we determined by ultrawide range spectroscopic ellipsometry, employing an anisotropic Drude- Lorentz model as described previously8. Fig. 1b shows the resulting in-plane permittivity of thin PEDOT:OTf film with thickness of 32 nm (Fig. S1 presents the raw data). The shaded area highlights a spectral region (0.8 to 3.6 μm) with negative permittivity ε1 (blue curve) that is also larger in magnitude than that of the imaginary component ε2 (red curve), which we refer to as plasmonic regime. The optically metallic and plasmonic character is related to the high conductivity within the thin film, in turn, due to high concentration (2.6 ×1021 cm-3) of mobile bipolaronic charge carriers. We also note that the mobility is highly anisotropic8,9 and the out- of-plane permittivity (Fig. S2a) is primarily positive throughout the measured range, making the conductive polymer thin film a natural hyperbolic material10 (Fig. S3). The measured optical properties of the thin PEDOT:OTf film imply that nanostructures of the material should be able to sustain plasmonic resonances. Indeed, the calculated optical extinction (see Methods for details) for a PEDOT:OTf nanodisk array (thickness of 30 nm, nanodisk diameter of 500 nm and periodicity of 1000 nm) shows a clear resonance peak at around 2.9 μm (Fig. 1c), which is absent for the non-structured thin film. The small extinction kink at about 1.4 μm disappears when examining single nanodisks instead of arrays (Fig. S4) and hence, arises from lattice scattering of the array (see Fig. S5a). Examining the optical nearfield profile at resonance (2.9 μm) for one of the nanodisks reveals that the extinction peak originates from a dipolar plasmonic mode (Fig. 1d and e), with enhanced fields on the opposite edges of the nanodisk in the polarization direction of the incident light. The optical nearfield patterns slightly above (Fig. 1d) and through the nanodisk (Fig. 1e) both resemble that of traditional gold nanodisk antennas (comparison in Fig. S6). Varying periodicity for fixed nanodisk dimensions did not significantly shift the resonance wavelength (Fig. S5b), further confirming that the extinction peak originates from localized surface plasmons rather than grating effects. In fact, also single nanodisks (Fig. S4a) show the same nanooptical behaviour, with almost identical resonance positions as the periodic arrays (Fig. S4b). To experimentally verify excitation of plasmons in conductive polymer nanostructures, we fabricated PEDOT:OTf nanodisks on sapphire substrates, using a modified version of colloidal lithography11 (see Methods and Fig. S7 for details). The protocol could provide large areas of nanodisks of desired diameters, visualized by atomic force microscopy (AFM) for nanodisk diameters of 120 nm, 280 nm, and 710 nm in Fig. 2a, b, and c, respectively (more images are provided in Fig. S8). The nanodisks all originate from 30 nm thick PEDOT:OTf films, while the final thickness of the disks varied somewhat due to residual PMMA remaining on top of the disks after fabrication (Fig. S9). Importantly, the fabricated polymer nanodisk samples exhibit clear extinction peaks (Fig. 2d, e, and f), verifying the simulated nanooptical behaviour. As expected for plasmonic nanoantennas, the resonance positions increase with disk diameter. The experimental results largely match the simulated predictions (Fig. 2g, h, and i) in terms of spectral shapes, peak widths and resonance wavelengths. Small differences in peak positions 3 are attributed to geometrical differences and imperfections of the fabricated nanodisks. The experimental peaks also show somewhat larger broadening, as expected for measured ensembles compared with simulated arrays composed of identical nanostructures12. We also note a tendency of slightly more asymmetric peak shape for the fabricated nanodisks compared with the simulated spectra, manifested by a smaller slope on the red side of the peak. As indicated by simulations (Fig. S10), this feature may be related to some few nanometers of remaining PEDOT:OTf film between nanodisks in some samples, although this needs further investigation. The experimental features between 2.7 μm and 3.3 μm (with multiple closely-packed sharp peaks) and at 4.3 μm are due to absorption by water vapour and carbon dioxide13, respectively, and therefore absent in the simulated spectra. Fig. 2 Extinction spectra of PEDOT:OTf nanodisk arrays. Three different sizes of nanodisk arrays were made on sapphire substrates: a, AFM image of a 120 nm diameter nanodisk; b, AFM image of a 280 nm diameter nanodisk; c, AFM image of a 710 nm diameter nanodisk. The diameter and height measurements are in Fig. S11. d, e, and f, Experimental measured extinction spectra of 120 nm, 280 nm, and 710 nm diameter nanodisks. UV-Vis-NIR measurements are plotted in light blue and FTIR measurements are in dark blue. g, h, and i, Simulated extinction spectra of 120 nm, 280 nm, and 710 nm diameter nanodisk arrays. In the simulation, the PEDOT:OTf thickness was 30 nm and the excessive thickness (8 nm, 36 nm, and 25 nm respectively) comes from remaining unremoved PMMA layer. The results above indicate that the resonance position of the polymer nanodisk antennas can be tuned by geometry, which we here investigate in more detail. Fig. 3a presents the simulated extinction for 30 nm thick single PEDOT:OTf nanodisks of varying diameter on a 4 substrate with refractive index of 1.6. Normalized extinction versus diameter is presented in Fig. 3c as colour maps. It is clear that the resonance position redshifts with increasing diameter, enabling tuning in a large spectral range from around 2 μm to around 4 μm for disks with sizes ranging from 200 nm and 700 nm in diameter. The spectral tunability can likely be extended further by other geometries. While the nanodisk resonances redshift with increasing disk diameter, they instead blueshift with increasing thickness, as presented in Fig. 3d and 3f for nanodisks with fixed diameter of 500 nm. Both these geometrical dependencies match expectations based on plasmonic nanodisk resonances14-16. oblate spheroids with diameter D and thickness t, which in the quasi-static limit 𝐷≪𝜆 gives We use the in-plane permittivity of PEDOT:OTf as 𝜀(𝜆) and set 𝜀s = 1.69 as the effective (1) where V is the volume of the spheroid and εs is the permittivity of the surrounding medium. 𝜀(𝜆)−𝜀s 𝜀s+𝐿[𝜀(𝜆)−𝜀s] To enable analytical calculation of the optical response, we approximate the nanodisks as the dipolar polarizability α as17 𝛼(𝜆)=𝑉 surrounding permittivity for disks in air on a substrate with refractive index 1.6 (see Methods). L is a geometrical factor that equals 1/3 for a sphere (D = t) and decreases for increasing nanodisk ratio (D > t, see Fig. S12). To fulfil the resonance condition of maximum polarizability when L decreases, the magnitude of the negative permittivity needs to increase. Because the permittivity of the conductive polymer increases in magnitude with wavelength (see Fig. 1b), the resonance position therefore redshifts with increasing aspect ratio (D/t). This illustrates why the resonance of the PEDOT:OTf nanoantennas redshifts with increasing disk diameter and blueshifts with increasing disk thickness. Larger disks require corrections for finite wavelength effects, which gives the corrected polarizability as18,19 where k is the wave number of the incident light. The extinction cross-section 𝜎(𝜆) can now 𝛼1(𝜆)=𝛼(𝜆)21− 𝑘52𝜋𝐷𝛼(𝜆)−𝑖𝑘96𝜋𝛼(𝜆); (2) be calculated via17 (3) Fig. 3b and 3e show the final calculated extinction cross sections of single PEDOT:OTf oblate spheroids, with sizes corresponding to the simulated nanodisks in Fig. 3a and 3d, respectively. The calculated extinction based on the dipolar polarizability match the results from the full simulations quite well, both in terms of extinction magnitude and its increase with aspect ratio, and in terms of peak positions and their redshift with increasing nanodisk aspect ratio. The results thereby further corroborate that the observed extinction peaks of the PEDOT:OTf nanodisks originate from dipolar plasmonic excitations. 𝜎(𝜆)=𝑘Im[𝛼′(𝜆)] 5 Fig. 3 Geometry dependence of single PEDOT:OTf nanodisk localized plasmons. a, Simulated extinction cross- section of 30 nm thick single nanodisks of different diameters (diameter step size = 100 nm, substrate refractive index = 1.6). b, Analytically calculated extinction cross-section for oblate spheroids corresponding to the nanodisk sizes in a. c, Colour map of simulated normalized extinction versus diameter for single 30 nm thick nanodisks. d, Simulated extinction cross-section of 500 nm in diameter single nanodisks with different thickness (thickness step size = 10 nm, substrate refractive index = 1.6). e, Analytically calculated extinction cross section for oblate spheroids corresponding to the nanodisk sizes in d. c, Colour map of simulated normalized extinction versus diameter for single 30 nm thick nanodisks. f, Colour map of simulated normalized extinction versus thickness for single 500 nm in diameter nanodisks. The white dashed lines in c and f indicate the resonance peak position and its shift with changes in diameter and thickness, respectively. The color scale bars in c and f present the normalized extinction. See Fig. S13 for normalized extinction spectra corresponding to a and d. Finally, we demonstrate that the conductive polymer nanoantennas can be switched on and off. Such dynamic control of optical nanoantennas is a topic of intense research, not least motivated by the enormous potential for important applications enabled by active plasmonic devices and metasurfaces4. Among various approaches to tune nanophotonic systems, recent research has explored tuning by modulating the free charge carriers in plasmonic systems, including electrical gating20 and photo-carrier excitation21. While this approach is rather limited for traditional plasmonic materials, conductive polymers hold great promise since their bipolaron charge carrier concentration can be modulated by several orders of magnitude via their redox state5. Here, we control the redox state chemically, utilizing that exposure of PEDOT:OTf to highly branched PEI (poly[ethylene imine], see chemical structure in left panel of Fig. 4a) vapour (volatile impurities of ethyleneimine dimers and trimers) can reduce the material22. Optical extinction spectroscopy of a (non-structured) thin PEDOT:OTf films visualizes the process via almost complete reduction of the free charge carrier absorption in the IR and the emergence of a neutral state peak at around 600 nm (see Fig. 4b)22. For the 6 reduced polymer, the reaction between PEI impurities and counterions reduces the bipolaronic charge carrier concentration, resulting in a material with largely reduced electrical conductivity (schematic mechanism in Fig. 4a right panel)22. The process is reversible and we can recover the original optical properties of the PEDOT film via acid treatment of the reduced film (see Methods). This process re-oxidizes the material, for which the neutral state disappears and the absorption returns to that of the initial pristine film (Fig. 4b). Knowing that the optical material properties of PEDOT:OTf can be reversibly modulated, we utilize this feature to actively tune our polymer nanodisk metasurfaces. The black curve in Fig. 4c shows the extinction spectra of a PEDOT:OTf nanodisk array in its oxidized pristine state, with plasmonic resonance peak at around 1900 nm. This peak completely disappears upon PEI vapour treatment, for which the material in the nanodisks is no longer plasmonic, due to drastic reduction of the bipolaronic charge carrier concentration. Indeed, the neutral state material absorption emerges at 600 nm for the PEI vapour treated metasurfaces. Importantly, the optical properties are not volatile, but stable over time and we observe only minimal extinction changes of the sample after one week (see Fig. S14). By re-oxidizing the sample with sulfuric acid, the plasmonic resonance peak recovers to its initial state, with both similar intensity and width as for the original plasmonic metasurface (Fig. 4c). The increase in extinction below 800 nm for the re-oxidized sample is likely due to different probe areas combined with some polystyrene beads remaining after fabrication (similar effects were observed for samples before and after bead removal, Fig. S15). Fig. 4 Redox state tunability of PEDOT:OTf nanodisk arrays. a, Chemical structure of highly branched PEI (left panel) and the redox state tunability mechanism for PEDOT-based materials (right panel) where M- stands for the counterion of PEDOT (e.g. PSS, Tos, or OTf). b, Measured extinction for thin PEDOT:OTf film on glass at different redox states (black: oxidized state, blue: reduced state, and red: re-oxidized state). Shaded area indicates the neutral state absorption peak. c, Measured extinction for PEDOT:OTf nanodisk arrays on glass (thickness of 43 nm and nanodisk diameter of 140 nm) at different redox states (black: oxidized state, blue: reduced state, and red: re-oxidized state). 7 We have demonstrated that nanodisks made of highly conductive polymers can be used as optical nanoantennas to form active plasmonic metasurfaces. While previous research has showed excitonic resonances in organic nanomaterials made from small molecules and dyes,23,24 our system is fundamentally different in that the resonances originate from localized plasmons formed by bipolaronic charge oscillations. Our observations open up new avenues for dynamically controllable plasmonics based on redox-tunable conductive polymers, where future work may explore dynamic control also by other means, including electrochemical modulation of the PEDOT:OTf redox state25. The future also holds promise for conductive polymers with yet further improved plasmonic properties, based on strategies to improve electrical conductivity and lower defect density, including effective chain alignment26, and sequential doping27. We hope that our study of redox-tunable conductive polymer plasmonics will inspire research in this interdisciplinary field of manipulating light-matter interaction with organic materials at the nanoscale. Acknowledgements The authors thankfully acknowledge financial support from the Swedish Research Council, the Swedish Foundation for Strategic Research, the Wenner-Gren Foundations, and the Swedish Government Strategic Research Area in Materials Science on Functional Materials at Linköping University (Faculty Grant SFO-Mat-LiU No. 2009 00971). Author contributions M.P.J. conceived and supervised the project. S.C, V.S., P.K., and V.D. performed ellipsometry measurements and data analysis. S.C. and M.S.C. fabricated the nanostructures. S.C., M.P.J. and E.S.K. performed numerical simulations. S.C. and H.S. performed PEI vapour treatments. S.C performed all the other characterizations. S.C. and M.P.J. organized the data and wrote the manuscript. All authors reviewed and commented on the manuscript. Competing interests The authors declare no conflicts of interest. References 1 Willets, K. A. & Van Duyne, R. P. Localized surface plasmon resonance spectroscopy and sensing. Annu. Rev. Phys. Chem. 58, 267-297 (2007). Atwater, H. A. & Polman, A. Plasmonics for improved photovoltaic devices. Nature Materials 9, 205 (2010). Yu, N. et al. Light propagation with phase discontinuities: generalized laws of reflection and refraction. Science 334, 333-337 (2011). Shaltout, A. M., Shalaev, V. M. & Brongersma, M. L. Spatiotemporal light control with active metasurfaces. Science 364, eaat3100 (2019). Kim, N. et al. in Conjugated Polymers: Properties, Processing, and Applications (CRC Press, 2019). Massonnet, N., Carella, A., de Geyer, A., Faure-Vincent, J. & Simonato, J.-P. Metallic behaviour of acid doped highly conductive polymers. Chemical Science 6, 412-417 (2015). 2 3 4 5 6 8 Gueye, M. N. et al. Structure and dopant engineering in PEDOT thin films: Practical tools for a dramatic conductivity enhancement. Chemistry of Materials 28, 3462-3468 (2016). Chen, S. et al. On the anomalous optical conductivity dispersion of electrically conducting polymers: ultra-wide spectral range ellipsometry combined with a Drude -- Lorentz model. Journal of Materials Chemistry C 7, 4350-4362 (2019). Aasmundtveit, K. E. et al. Structure of thin films of poly (3, 4-ethylenedioxythiophene). Synthetic Metals 101, 561-564 (1999). Poddubny, A., Iorsh, I., Belov, P. & Kivshar, Y. Hyperbolic metamaterials. Nature Photonics 7, 948 (2013). Hanarp, P., Käll, M. & Sutherland, D. S. Optical properties of short range ordered arrays of nanometer gold disks prepared by colloidal lithography. The Journal of Physical Chemistry B 107, 5768-5772 (2003). Stete, F., Koopman, W. & Bargheer, M. Signatures of strong coupling on nanoparticles: revealing absorption anticrossing by tuning the dielectric environment. ACS Photonics 4, 1669- 1676 (2017). Larkin, P. Infrared and Raman spectroscopy: principles and spectral interpretation. (Elsevier, 2017). Chang, W.-S. et al. Tuning the acoustic frequency of a gold nanodisk through its adhesion layer. Nature Communications 6, 7022 (2015). Fang, Z. et al. Active tunable absorption enhancement with graphene nanodisk arrays. Nano Letters 14, 299-304 (2013). Knight, M. W. et al. Aluminum for plasmonics. ACS Nano 8, 834-840 (2013). Maier, S. A. Plasmonics: fundamentals and applications. (Springer Science & Business Media, 2007). Langhammer, C., Yuan, Z., Zorić, I. & Kasemo, B. Plasmonic properties of supported Pt and Pd nanostructures. Nano Letters 6, 833-838 (2006). Wokaun, A., Gordon, J. P. & Liao, P. F. Radiation damping in surface-enhanced Raman scattering. Physical Review Letters 48, 957 (1982). Lee, S. H. et al. Switching terahertz waves with gate-controlled active graphene metamaterials. Nature Materials 11, 936 (2012). Alam, M. Z., De Leon, I. & Boyd, R. W. Large optical nonlinearity of indium tin oxide in its epsilon-near-zero region. Science 352, 795-797 (2016). Fabiano, S. et al. Poly (ethylene imine) Impurities Induce n-doping Reaction in Organic (Semi) Conductors. Advanced Materials 26, 6000-6006 (2014). Gentile, M. J., Núñez-Sánchez, S. & Barnes, W. L. Optical field-enhancement and subwavelength field-confinement using excitonic nanostructures. Nano Letters 14, 2339-2344 (2014). Gu, L., Livenere, J., Zhu, G., Narimanov, E. E. & Noginov, M. A. Quest for organic plasmonics. Applied Physics Letters 103, 021104 (2013). Mitraka, E. et al. Oxygen-induced doping on reduced PEDOT. Journal of Materials Chemistry A 5, 4404-4412 (2017). Xu, J. et al. Multi-scale ordering in highly stretchable polymer semiconducting films. Nature Materials 18, 594 (2019). Vijayakumar, V. et al. Bringing Conducting Polymers to High Order: Toward Conductivities beyond 105 S cm− 1 and Thermoelectric Power Factors of 2 mW m− 1 K− 2. Advanced Energy Materials 9, 1900266 (2019). 9 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 2- from the sulfuric acid. Methods Thin film deposition. PEDOT:OTf thin film were prepared via vapour phase polymerization (VPP) as reported in the literature.8,28 The oxidant solution for EDOT polymerization was prepared by mixing 0.03 g of iron (III) trifluoromethanesulfonate (Fe[OTf]3, from Alfa Aesar), 0.2 g of tri-block co-polymer poly(ethylene glycol)-block-poly(propylene glycol)-block-poly(ethylene glycol) (PEG-PPG-PEG or P-123, average Mn ~5,800, from Sigma-Aldrich) and 0.8 g of 99.5% ethanol (from Solveco). Oxidant films were deposited by spin-coating at 1500 rpm for 30 s onto pre-cleaned sapphire or glass substrates (sonicated in cleaning detergent, de-ionized water, acetone, and isopropanol each for 10 min respectively and treated with oxygen-plasma at 200 W for 5 min before use). After 30 s baking on a hotplate at 70 °C, the samples were transferred into a heated vacuum desiccator (Vacuo-temp, from SELECTA). EDOT (142.18 g mol-1, from Sigma-Aldrich) droplets were drop-casted onto a glass substrate on a hot plate at 30 °C in the desiccator to ensure its evaporation. After 30 min of polymerization at a pressure of 70 mBar, the samples were taken out from the desiccator and washed with ethanol multiple times to remove byproducts and unreacted residues, followed by air-drying with nitrogen. Acid treatment by soaking in 1 M sulfuric acid (H2SO4) for 10 min at 100 °C was applied to the samples for further enhancement of their electronic properties.29 With acid treatment, parts of counterions (OTf) in PEDOT:OTf thin film are replaced by SO4 Nanoantenna fabrication. The detailed process flow for nanodisk array fabrication is shown in Fig. S7, which is a modified version of colloidal lithography.11 Briefly, a 4 wt% PMMA (Mw ~996,000, from Sigma-Aldrich) solution in anisole (from Sigma Aldrich) was spin-coated onto the as-prepared PEDOT:OTf thin films. Soft baking at 140 °C for 10 min was then applied. The samples were treated with reactive oxygen plasma (50 W, 250 mTorr) for 5 s to increase the hydrophilicity of the surface. In order 2 wt% poly(diallyldimethylammonium chloride) (PDDA, 522376 from Sigma-Aldrich) in DI water was dropped on the samples. After 1 min, the samples were rinsed with deionized water for 40 s and dried with nitrogen stream. Negatively charged polystyrene nanoparticles (PS beads with different diameters, 0.2-0.3 wt% in deionized water, from Microparticles GmbH) were then dropped on the samples. After 10-30 min, the samples coated with PS beads were rinsed with DI water and dried with nitrogen stream resulting in a sparse monolayer of PS beads on the PMMA/PEDOT:OTf thin films. A heat treatment at 100 °C for 2 min were applied to the samples to improve the adhesion of PS beads on the samples. Reactive oxygen plasma etching (250 mTorr, 50 W) for 3-5 min were applied to the samples, using the PS beads monolayer as mask. Depending on the size of PS beads and thickness of PMMA and PEDOT:OTf thin films, the time interval of etching can be varied to ensure a complete removal of PMMA and PEDOT:OTf parts that are not covered by the mask. The samples were then placed into an acetone bath and soaked for 10-30 min followed by a mild sonication for 3 min and nitrogen stream drying to remove PMMA and PS beads and finally the PEDOT:OTf nanodisks were obtained. In this study, three different diameters of PS beads were used: 239 nm (PS-ST-0.25, Microparticles GmbH), 497 nm (PS-ST-0.50, Microparticles GmbH), and 1046 nm (PS-ST-1.0, Microparticles GmbH). Vapour treatment of thin films and nanoantennas. The vapour treatment was conducted inside a N2- filled glovebox by exposing the samples to the vapour of ethyleneimine dimers and trimers by heating a vial containing highly branched poly(ethylene imine) liquid (PEI, Mw ~ 800, from Sigma-Aldrich) at 120 °C for 5 min.22 After the vapour treatment, the samples were annealed at 120 °C for another 5 min. For re-oxidizing of the samples, they were put into 1 M sulfuric acid bath for 10 min followed by a drying process of 10 min at 100 °C on a hot plate. functionalize positively charged, to the PMMA surface to be 10 Ellipsometry. PEDOT:OTf thin film samples were measured at normal ambient conditions at room temperature. The films were deposited on 2-inch single side polished c-plane sapphire wafers (from Semiconductor Wafer Inc.). Ellipsometric data for PEDOT:OTf thin films were collected using three different ellipsometers covering a wide spectral range from 0.0028 eV to 5.9 eV. UV-Vis-NIR measurements were performed on a J. A. Woollam Co. RC2® spectroscopic ellipsometer for five incident angles (40°, 50°, 60°, 70°, and 80°) and spectral range from 0.73 eV (1690 nm) to 5.90 eV (210 nm). Infrared measurements were performed on a J. A. Woollam Co. IR-VASE® spectroscopic ellipsometer for two incident angles (50° and 70°) and spectral range from 28.0 meV (230 cm-1) to 1.0 eV (7813 cm-1). THz measurements were performed on the THz ellipsometer at the Terahertz Materials Analysis Center (THeMAC) at Linköping University.30 Three incident angles (40°, 50°, and 60°) were used for THz measurements, in the spectral range between 2.8 meV (0.67 THz) and 4.0 meV (0.97 THz). The typical ellipsometer measures the complex reflectance ratio ρ at different frequencies, as obtained from ρ = rp/rs = tan(Ψ)eiΔ, where rp and rs are the complex Fresnel reflection coefficients for p- and s- polarized light; Ψ shows the amplitude ratio change of the two polarizations; and Δ indicates the phase difference between them.31 WVASE® (J. A. Woollam Co.) was used for data analysis and a Drude- Lorentz model were employed for model fitting and optical parameter extraction for the PEDOT:OTf thin films.8 Electrical and structural characterization. Sheet resistance, Rs, of the thin film was measured using a 4-point probe set-up using a Signatone Pro4 S-302 resistivity stand and a Keithley 2400. Film thickness t was determined by a surface profiler (Dektak 3st, Veeco). The thickness of the VPP PEDOT:OTf films varied in the range from 30 to 40 nm. The electrical conductivity can then be calculated by σ = 1/(Rst). Atomic force microscopy (AFM) was employed for surface morphology characterization, in tapping mode using a Veeco Dimension 3100. The morphological images were analyzed using Nanoscope Analysis software (Bruker). Optical characterization. The extinction spectra in the Vis-NIR range (400 nm to 3300 nm) were measured using a UV-Vis-NIR spectrometer (Lambda 900, Perkin Elmer Instruments). The extinction spectra include transmission losses due to both absorption and scattering. Fourier-transform infrared spectroscopy (FTIR) measurements were performed in the spectral range from 1333 nm (7500 cm-1) to 5000 nm (2000 cm-1) or 6667 nm (1500 cm-1) using an Equinox 55 spectrometer (Bruker). FTIR spectra were acquired in absorbance mode using a resolution 4 cm-1 and 100 scans. Samples deposited on 20×20×0.5 mm double-side polished sapphire substrates (from Semiconductor Wafer Inc.) were made for FTIR and UV-Vis-NIR measurements. Optical numerical simulations. Numerical simulations (electric nearfield intensity and farfield spectra) of the electromagnetic response of PEDOT:OTf nanoantennas were performed via the finite-difference time-domain the commercial software Lumerical FDTD Solutions (http://www.lumerical.com/fdtd.php). The optical parameters for the PEDOT:OTf thin film were taken as the anisotropic complex permittivity obtained from the ellipsometry measurements. For periodic nanodisk arrays and thin films, the spectra and nearfield profiles were recorded via field and power monitors. Periodic PEDOT:OTf nanodisk arrays (or thin film) were placed on top of glass or sapphire substrates. The structures were illuminated by a planewave light source at normal incidence. Anti- symmetrical and symmetric boundaries were used for the x-axis (parallel to polarization) and y-axis (normal to polarization) and perfectly matched layer (PML) were used for the z-axis (parallel to light incident direction). For single nanodisks, spectra were obtained using a total field/scattered field and by extracting the extinction cross-section of isolated PEDOT:OTf nanodisks on a sapphire substrate. Geometry parameters are indicated in each graph (diameter, thickness, and periodicity) and the mesh size was typically 3 × 3 × 3 nm3, or 2 × 2 × 2 nm3 for the smaller size disks. The optical parameters for (FDTD) method using 11 gold32, glass33 and PMMA34,35 were taken from literature while the permittivities of sapphire substrate and PEDOT:OTf were determined by ellipsometry. In the analytical calculations, the effective permittivity of the surroundings was calculated based on an average refractive index of air and sapphire (εs = [(nair + nsapphire)/2]2). The refractive index of sapphire is 1.75 at about 1 μm and 1.6 at about 5 μm and for simplicity we fix nsapphire=1.6 which gives εs=1.69. References 28 Brooke, R. et al. Vapor phase synthesized poly (3, 4-ethylenedioxythiophene)- trifluoromethanesulfonate as a transparent conductor material. Journal of Materials Chemistry A 6, 21304-21312 (2018). Kim, N. et al. Highly conductive PEDOT: PSS nanofibrils induced by solution - processed crystallization. Advanced Materials 26, 2268-2272 (2014). Kühne, P. et al. Advanced Terahertz Frequency-Domain Ellipsometry Instrumentation forIn SituandEx SituApplications. IEEE Transactions on Terahertz Science and Technology 8, 257-270 (2018). Tompkins, H. & Irene, E. A. Handbook of ellipsometry. (William Andrew, 2005). Weaver, J. H. & Frederikse, H. P. R. Optical properties of metals and semiconductors. CRC Handbook of Chemistry and Physics 74, 1993-1994 (1993). Philipp, H. R. in Handbook of optical constants of solids 749-763 (Elsevier, 1997). Tsuda, S., Yamaguchi, S., Kanamori, Y. & Yugami, H. Spectral and angular shaping of infrared radiation in a polymer resonator with molecular vibrational modes. Optics Express 26, 6899- 6915 (2018). Beadie, G., Brindza, M., Flynn, R. A., Rosenberg, A. & Shirk, J. S. Refractive index measurements of poly (methyl methacrylate)(PMMA) from 0.4 -- 1.6 μm. Applied Optics 54, F139-F143 (2015). 29 30 31 32 33 34 35 12 Supporting Information Fig. S1 Spectroscopic ellipsometry for thin PEDOT:OTf film. Ellipsometric raw data a, ψ and b, ∆ of the film are exhibited with the spectral range from 2.8 meV (0.67 THz) to 5.90 eV (210 nm), including THz, MIR, and NIR-Vis-UV three regions. For each spectral window, two, three or five incident angles were employed. The experimental data are in grey dashed lines and Drude-Lorentz model8 generated data are in solid lines. Details of the model can be found in Table. S1. Fig. S2 Permittivity dispersion for thin PEDOT:OTf thin film. Both a, real permittivity, and b, imaginary permittivity dispersion curves are plotted. Red curves are for the in-plane direction (x-y plane) and blue curves for the out-of-plane direction (z-axis). The extracted data are based on the Drude-Lorentz model in Fig. S1 and using the fitted parameters in Table. S1. 13 Fig. S3 Natural hyperbolic properties of a PEDOT:OTf thin film. In-plane (ε⊥, ordinary axis or x-y plane) and out-of-plane (ε∥, extraordinary or z-axis) real permittivity comparison. Three spectral regions are indicated: elliptical regime (permittivity for both directions are positive), hyperbolic regime (the sign of permittivity for the two directions are opposite), and ENZ regime (epsilon-near-zero, one direction has permittivity close to zero). The results show that the PEDOT:OTf thin film has a natural hyperbolic permittivity in the range 0.7 μm and 6 μm. Fig. S4 Simulated optical extinction spectra for single nanodisks. a, The extinction spectra for single 30 nm thick nanodisks with different diameters from 200 nm to 700 nm on sapphire substrates. b, Comparison between normalized single nanodisk extinction (solid curves) and normalized periodic nanodisk arrays extinction (dashed curves). The resonance peak positions are similar for single and periodic nanodisks. 14 Fig. S5 Grating effects in periodic PEDOT:OTf nanodisk arrays. PEDOT:OTf nanodisk arrays with thickness of 40 nm on glass substrates, with 100 nm PMMA on top of the nanodisks. a, Extinction of nanodisks with different diameters (from 320 nm to 480 nm) at a fixed periodicity of 1000 nm. The kink feature at about 1.4 μm comes from grating effects. b, Extinction of nanodisks with different periodicity (from 1000 nm to 1400 nm) at a fixed disk diameter of 320 nm. Fig. S6 Optical extinction spectra and nearfield profile for gold (Au) nanodisk antennas. a, Simulated extinction spectra of a Au nanodisk array (thickness of 30 nm, disk diameter of 500 nm, and periodicity of 1000 nm) on a glass substrate. A sharp resonance peak located at 1.52 μm can be observed. x-, y-, and z-axes are indicated in the inset. Calculated nearfield profiles at the wavelength of extinction maximum (1.52 μm) for one of the Au nanodisks of the array in a: b, x-y in-plane direction 2 nm above the nanodisk; c, x-z out-of-plane direction, cross- section through the center of the nanodisk. The color scale bars show the electric field strength relative to the incident light. 15 Fig. S7 Fabrication procedure of large scale PEDOT:OTf nanodisk arrays. The process flow is based on a modified version of colloidal lithography11. a, Deposition of thin PEDOT:OTf film by VPP. b, Spin coating of sacrificial PMMA layer on top of PEDOT:OTf film. c, Deposition of colloidal PS microbeads on oxygen-plasma treated PMMA layer by colloidal lithography. d, Reactive ion etching of PMMA and PEDOT:OTf thin film. The PS microbeads also shrink somewhat. e, Tape stripping of PS microbeads. f, PMMA layer removal by acetone soaking and sonication. Fig. S8 AFM topography images of PEDOT:OTf nanodisk array for three different disk sizes. a, Sample with 120 nm diameter nanodisks, corresponding to Fig. 2a. b, Sample with 280 nm diameter nanodisks, corresponding to Fig. 2b. c, Sample with 710 nm diameter nanodisks, corresponding to Fig. 2c. Fig. S9 Effects from PMMA layer on PEDOT:OTf nanodisks. In the simulation, PMMA nanodisk layers with different thickness (10 nm to 120 nm) were added to study its influence on the resonance position. The PEDOT:OTf nanodisk arrays have a thickness (T) of 40 nm, diameter (D) of 500 nm, and periodicity (P) of 1500 nm. With the increase of the PMMA layer thickness, the resonance peak red shifts with a reduced maximal extinction. 16 Fig. S10 Influence of ultrathin PEDOT:OTf bottom layer for nanodisk arrays. a, Simulated extinction of 3 nm thick bottom PEDOT:OTf layer on glass substrate. b, Simulated extinction of 4 nm thick bottom PEDOT:OTf layer on glass substrate. c, Simulated extinction of 5 nm thick bottom PEDOT:OTf layer on glass substrate. For the simulated nanodisk arrays, the thickness is 30 nm, the diameter is 500 nm, and the periodicity is 1000 nm. Samples with bottom layer are plotted in dashed curves and samples without bottom layer are in solid curves. Extinction (Ext, in black), absorption (Abs, in red), and reflection (Ref, in blue) are indicated. For samples with bottom layer thickness less than 3 nm, the resonance peak blue-shift is negligible. Fig. S11 Height profile of single nanodisks. Height and diameter of PEDOT:OTf single nanodisks measured by AFM. a, From Fig. 2a (height: 38 nm, diameter: 120 nm); b, From Fig. 2b (height: 66 nm, diameter: 280 nm); and c, From Fig. 2c (height: 55nm, diameter: 710 nm). 17 Fig. S12 Geometrical factor L for different aspect ratios of nanodisks when approximated as oblate spheroids. The geometrical factor L, for excitation at normal incidence, was calculated via17 𝐿=𝐷5𝑡16C 𝑑𝑞 J F𝐷54+𝑞H𝑓(𝑞) K 𝑓(𝑞)=LM𝑞+𝐷54N5M𝑞+𝑡54N (E1) (E2) where D and t are the diameter and thickness of the nanodisks. With the increase of aspect ratio (ratio between diameter and thickness of the disk, D/t), the geometrical factor L decreases from 0.5 for long rods to 1/3 for D = t, and further towards 0 for ultrathin disks. Fig. S13 Normalized extinction cross-section spectra for single PEDOT:OTf nanodisks of different dimensions. a, Spectra for varying diameter show increase in resonance wavelength with increasing diameter (from 200 to 700 nm, with thickness fixed at 30 nm). b, Spectra for varying disk thickness show decrease of resonance wavelength with increasing thickness (from 20 to 70 nm, with diameter fixed at 500 nm. The spectra were simulated based on single PEDOT:OTf nanodisks on a substrate with fixed refractive index of 1.6. 18 Fig. S14 Extinction of reduced PEDOT:OTf nanodisks. The optical properties of as-prepared nanodisks (blue curve) is stable and changes little after one week in air (red curve). Fig. S15 Measured extinction spectra for PEDOT:OTf nanodisks with and without PMMA and polystyrene beads on top of the nanodisks. The two different colors correspond to two different samples, both with nanodisk diameter of 120 nm. a, Extinction spectra for nanodisks with polystyrene beads remaining on top of the disks. A strong absorption can be observed below 700 nm due to the polystyrene beads. b, Extinction spectra for PEDOT:OTf nanodisks after removing the polystyrene beads, for which the absorption feature of the polystyrene below 700 nm has disappeared. 19 Table. S1 Measured electrical properties of thin PEDOT:OTf film. The parameters were determined by both ellipsometry and conventional electrical measurements (profiler and 4- point-probe). The predictions from ellipsometry are based on the anisotropic Drude-Lorentz model8. Thickness (nm) Electrical conductivity (S/cm) Charge density (1021 cm-3) In-plane mobility (cm2 V-1 s-1) Ellipsometry predictions 31.802 ± 0.394 5603.8 ± 760.0 2.61 ± 0.07 13.419 ± 1.46 Experimental measurements 30 ± 5 5200 ± 500 20
1905.01268	1	1905	2019-05-03T16:54:00	Effect of Imbalanced Charge Transport on the Interplay of Surface and Bulk Recombination in Organic Solar Cells	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Surface recombination has a major impact on the open-circuit voltage ($V_\text{oc}$) of organic photovoltaics. Here, we study how this loss mechanism is influenced by imbalanced charge transport in the photoactive layer. As a model system, we use organic solar cells with a two orders of magnitude higher electron than hole mobility. We find that small variations in the work function of the anode have a strong effect on the light intensity dependence of $V_\text{oc}$. Transient measurements and drift-diffusion simulations reveal that this is due to a change in the surface recombination rather than the bulk recombination. We use our numerical model to generalize these findings and determine under which circumstances the effect of contacts is stronger or weaker compared to the idealized case of balanced charge transport. Finally, we derive analytical expressions for $V_\text{oc}$ in the case that a pile-up of space charge is present due to highly imbalanced mobilities.	physics.app-ph	physics	Effect of Imbalanced Charge Transport on the Interplay of Surface and Bulk Recombination in Organic Solar Cells Dorothea Scheunemann∗ and Sebastian Wilken Physics, Faculty of Science and Engineering and Center for Functional Materials, Abo Akademi University, Porthansgatan 3, 20500 Turku, Finland and Institute of Physics, Energy and Semiconductor Research Laboratory, Carl von Ossietzky University of Oldenburg, 26111 Oldenburg, Germany Department of Physics, Swansea University, Singleton Park, Swansea, SA2 8PP, Wales, United Kingdom Oskar J. Sandberg Physics, Faculty of Science and Engineering and Center for Functional Materials, Abo Akademi University, Porthansgatan 3, 20500 Turku, Finland Ronald Osterbacka Institute of Physics, Energy and Semiconductor Research Laboratory, Carl von Ossietzky University of Oldenburg, 26111 Oldenburg, Germany Manuela Schiek Surface recombination has a major impact on the open-circuit voltage (Voc) of organic photo- voltaics. Here, we study how this loss mechanism is influenced by imbalanced charge transport in the photoactive layer. As a model system, we use organic solar cells with a two orders of magnitude higher electron than hole mobility. We find that small variations in the work function of the anode have a strong effect on the light intensity dependence of Voc. Transient measurements and drift- diffusion simulations reveal that this is due to a change in the surface recombination rather than the bulk recombination. We use our numerical model to generalize these findings and determine under which circumstances the effect of contacts is stronger or weaker compared to the idealized case of balanced charge transport. Finally, we derive analytical expressions for Voc in the case that a pile-up of space charge is present due to highly imbalanced mobilities. I. INTRODUCTION Organic solar cells typically consist of a blend of an electron donor and an electron acceptor sandwiched be- tween two electrical contacts [1 -- 3]. Ideally, the contacts act as semipermeable membranes for electrons (cathode) and holes (anode). In this case, the open-circuit voltage Voc is solely determined by the splitting of the quasi- Fermi levels EF,n − EF,p in the blend [4, 5]. If only bi- molecular recombination is present, G (cid:19) qVoc = Eg − kT ln(cid:18) βN 2 (1) has been suggested, where q is the elementary charge, Eg the band gap, k the Boltzmann constant, T the temper- ature, β the recombination coefficient, N the density of states and G the generation rate [6, 7]. The latter part of Eq. (1) predicts a slope of kT /q when plotting Voc versus the logarithm of the light intensity. However, many contacts like metals or doped polymers are non-selective contacts. This means they have the ability to exchange both minority and majority charge carriers with the photoactive layer. The extraction of ∗ [email protected] minority carriers (electrons at the anode, holes at the cathode) leads to a reduction of Voc [8]. Here, we call this loss mechanism surface recombination, with a corre- sponding recombination current jS = qSnmin, (2) where S denotes the surface recombination velocity and nmin the concentration of minority carriers close to the contact under consideration [9, 10]. In the case of Ohmic contacts, jS is strongly reduced by charge carrier injection. Because of the high concentra- tion of majority carriers at the interface, minority carri- ers are much more likely to recombine in the bulk, rather than leaving the device via the "wrong" electrode. As a result, Voc is still determined solely by properties of the bulk and follows Eq. (1). The situation changes when one of the contacts is non- Ohmic. For instance, if an injection barrier ϕan is present at the anode, less holes are injected into the blend, so that nmin and jS are effectively increased. Solak et al. [11] showed that the open-circuit voltage at high light inten- sities may then be described by qVoc = Eg − ϕan − kT 2 ln(cid:18) βN 2 G (cid:19) . (3) Compared to Eq. (1), there are two differences: First, the constant energetic part is reduced by the barrier height. arXiv:1905.01268v1 [physics.app-ph] 3 May 2019 Second, because of the factor 1/2 in front of the loga- rithm, the intensity dependence of Voc is now given by a slope of kT /2q (instead of kT /q). Such a reduction of the slope has been demonstrated both in experiment and simulation [11 -- 13]. The transition between Eq. (1) and Eq. (3) has been assumed to take place when Voc equals the built-in voltage Vbi [11]. For large S, surface recombination is not limited by the interface kinetics, but the transport of carriers to- wards the contact [12]. Hence, the question arises, how the open-circuit voltage depends on the charge carrier mobility µ. Numerical studies have indicated a decrease of Voc(µ) with increasing mobility if the contacts are non- selective [7, 14, 15]. The result is a finite optimum value of µ in terms of the overall device efficiency, independent of the recombination mechanism in the bulk. In the above considerations, the mobilities of elec- trons (µn) and holes (µp) are considered balanced. How- ever, this condition is often not fulfilled in practice. Many polymer-fullerene solar cells, for instance, exhibit a higher electron than hole mobility [16, 17], while it is the other way round for devices based on recent non-fullerene acceptors [3, 18, 19]. It is well known that the imbalanced mobilities lead to a pile-up of space charge close to one contact, which may reduce both the fill factor and short- circuit current [20, 21]. In contrast, little attention has been paid on how this affects the open-circuit voltage. Recently, Spies et al. [22] suggested that the additional charge will further reduce the built-in potential and, thus, severely affect the magnitude of Voc. In this work, we use an experimental system with a strong mobility mismatch of µn/µp = 100 and a well calibrated numerical model to discuss the effect of im- balanced transport on the open-circuit voltage in more detail. We show that the ratio between electron and hole mobility critically determines whether Voc is dominated by surface recombination or bimolecular recombination in the bulk. With the help of the numerical simulations we expand the analytical framework given by Eq. (3) to the case of imbalanced mobilities. II. EXPERIMENTAL AND NUMERICAL FRAMEWORK A. Experiment We fabricated solar cells based on a bulk hetero- junction of the small molecule donor 2,4-bis[4-(N,N - diisobutylamino)-2,6-dihydroxyphenyl] squaraine (SQIB) and the fullerene acceptor [6,6]phenyl-C61-butyric acid methyl ester (PCBM). This blend is known for a strong contrast between the mobility of electrons (10−4 cm2/Vs) and holes (10−6 cm2/Vs) [23]. Our devices had the struc- ture indium tin oxide/HTL/SQIB:PCBM/LiF/Al, where HTL denotes the hole transport layer. Details regarding the used materials and the device preparation can be found elsewhere [23 -- 25]. 2 FIG. 1. Current-voltage curves for MoOx (squares) and PE- DOT:PSS (circles) devices under simulated AM1.5G illumina- tion (100 mW/cm2). Solid lines are the result of drift-diffusion simulations using the parameter set in Tab. I. The only pa- rameter varied was the injection barrier ϕan at the anode. Inset: Schematic energy level diagram at V = Voc. To realize both devices with an Ohmic and a non- Ohmic contact, we changed the HTL from molybdenum suboxide (MoOx) to the doped polymer PEDOT:PSS. The energy level of PEDOT:PSS lies within the band gap of the photoactive blend, effectively reducing Vbi by 280 mV (see Supplemental Material [26]). As shown in Fig. 1, this reduction results primarily in a significant drop of Voc from 920 to 800 mV, which is in agreement with previous reports [27 -- 29]. Conversely, the HTL had little effect on the short circuit current and the fill fac- tor; both current-voltage curves exhibit the typical shape of space-charge-limited collection [20]. Averaged photo- voltaic characteristics for both types of devices can be found in the Supplemental Material [26]. B. Numerical Model We aimed to understand these findings using a nu- merical drift-diffusion model [30]. The model treats the bulk heterojunction as an effective semiconductor sand- wiched between two electrical contacts. The alignment of the work function of the contacts and the transport levels of the effective semiconductor is given by the in- jection barriers ϕan (anode) and ϕcat (cathode). The injection of charge carriers is then assumed to occur via thermionic emission. Surface recombination at the con- tacts is treated according to Eq. (2). We then assumed that excess charges are generated by illumination through the transparent anode. To take into account the spatial distribution G = G(x) of the pho- togeneration, we coupled the drift-diffusion model with transfer-matrix calculations [31, 32]. The recombination of mobile carriers in the photoactive layer is considered 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 - 8 - 6 - 4 - 2 0 2 M o O x P E D O T : P S S S i m u l a t i o n C u r r e n t D e n s i t y ( m A / c m 2 ) V o l t a g e ( V ) E g to be solely bimolecular, R = β(np − n2 i ), (4) where n and p is the density of electrons and holes, re- spectively, and ni the intrinsic carrier density. This is motivated by a recent study [23], where we show that non-geminate recombination SQIB:PCBM blends resem- bles a second-order process with a prefactor β indepen- dent on the carrier density. However, we note that herein, Eq. (4) is used only as an empirical rate equation, with- out making any assumptions on the details of the actual recombination mechanism (e.g., whether it is radiative or non-radiative). All relevant input parameters for the simulation are listed in Tab. I. TABLE I. Input parameters used for the numerical drift- diffusion simulations. Parameter Value Description T Eg εr d N β S ϕcat ϕan µn µp Temperature Effective band gap Dielectric constant Active-layer thickness Effective density of states Recombination coefficent Surface recombination velocity Injection barrier, cathode Injection barrier, anode 300 K 1.36 eV 4 100 nm 1026 m−3 10−17 m3s−1 105 ms−1 0 eV varied 2 × 10−8 m2(Vs)−1 Electron mobility 2 × 10−10 m2(Vs)−1 Hole mobility With this model we were able to describe the experi- mental data only by varying the injection barrier height at the anode, while keeping all other parameters con- stant (see solid lines in Fig. 1). This proves that the vari- ation of the HTL only affects the energy level alignment at the anode, but not the bulk properties of the active layer. Thus, we have at hand a suitable model system to study the effect of imbalanced mobilities on Voc. III. RESULTS AND DISCUSSION A. Impact of an injection barrier Figure 2(a) shows the experimental light intensity (I) dependence of the open circuit voltage. For the MoOx device, Voc versus ln(I) has nearly a slope of kT /q, as predicted by Eq. (1) for Ohmic contacts. Hence, we can assume that Voc is limited by bimolecular recombination in the bulk only. In contrast, the PEDOT:PSS device shows a transition towards a lower slope at mid to high light intensity. The reduced slope agrees qualitatively well with Eq. (3). This kind of behavior, with the slope going from ∼ kT /q at low intensity to kT /2q at higher 3 FIG. 2. Light-intensity dependence of Voc under AM1.5G illumination for (a) experimental and (b) simulated devices based on MoOx (squares) and PEDOT:PSS (circles). Error bars in (a) are the standard deviation for 5 individual sam- ples. Dotted lines indicate a scaling of kT /q ("slope 1") and kT /2q ("slope 1/2"). intensity (while Voc remains lower than for the MoOx device over the entire intensity regime), suggests that surface recombination at one non-Ohmic contact is dom- inating in the PEDOT:PSS device [12]. As can be seen in Fig. 2(b), both the absolute value and the intensity dependence of Voc are well captured by our numerical model if we only change the magnitude of ϕan. It was not possible to reproduce the experimental data by varying the surface recombination velocity at the anode instead (see Supplemental Material [26]). A signif- icant reduction of S would give rise to an extraction bar- rier, which would then result in S-shaped current-voltage curves [9, 10, 33, 34]. Because such an S-kink is not present in the data shown in Fig. 1, we expect the sur- face recombination current to be mainly determined by the carrier concentrations at the anode and the transport properties of the bulk. Figure 3 shows the effect of the anode work function on the energy band diagrams, as well as the electron and 1 e p o l s slope 1/2  a  a 1 1 0 1 0 0 0 . 7 0 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 0 . 9 5 1 1 0 1 0 0 0 . 7 0 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 0 . 9 5 L i g h t I n t e n s i t y I ( m W / c m 2 ) O p e n - C i r c u i t V o l t a g e V o c ( V ) M o O x P E D O T : P S S ( a ) E x p e r i m e n t ( b ) S i m u l a t i o n n = 0 n = 2 5 0 m e V O p e n - C i r c u i t V o l t a g e V o c ( V ) L i g h t I n t e n s i t y I ( m W / c m 2 ) 4 FIG. 3. Simulated band diagrams and local carrier concentrations for (a,c) ϕan = 0 and (b,d) ϕan = 250 meV. The anode is positioned at x = 0. Solid lines in panels (a) and (b) denote the transport bands and dashed lines the quasi-Fermi levels under AM1.5G illumination (100 mW/cm2). Panels (c) and (d) show the total electron and hole concentration for incident light intensities of 1, 10 and 100 mW/cm2. The result of the injection barrier is a gradient of EF,p at the anode, which reduces Voc relative to its optimum value Voc,max given by the quasi-Fermi level splitting in the bulk. hole concentration. The solid lines in panels (a) and (b) denote the transport levels under 1-sun illumination and the dashed lines the quasi-Fermi levels. The MoOx de- vice shows significant band bending at both electrodes caused by injection of majority carriers into the semicon- ductor. In case of the anode, there is a high concentra- tion of holes, exceeding the concentration of photogen- erated carriers in the bulk by several orders of magni- tude. Because of the high hole concentration, electrons are likely to recombine within the bulk, rather than be- ing extracted. The quasi-Fermi levels at both electrodes are flat, so that the open-circuit voltage represents the splitting of the quasi-Fermi levels in the bulk. The situa- tion remains relatively unchanged with increasing photo- generation. Hence, the light intensity dependence of Voc shows a constant slope over the intensity range studied herein and can be described by bimolecular recombina- tion in the bulk. For the PEDOT:PSS device with a non-Ohmic contact, the concentration of injected holes is much lower, leading to a reduced band bending at the anode. Consequently, the electron concentration close to the anode is higher than for the case with an Ohmic contact. According to Eq. (2), this non-negligible concentration of minority carriers induces a surface recombination current jS. To ensure open-circuit conditions (no net current), it must be compensated by a hole current jp = pµp dEF,p dx . (5) Because the magnitude of p close the anode is fixed by the barrier height ϕan, and µp is considered constant, an increase of jS due to increasing photogeneration can only be compensated by a gradient of the quasi-Fermi level for holes. At 1-sun illumination, the gradient in EF,p is clearly visible. Consequently, the open-circuit voltage is reduced and no longer a measure of the quasi-Fermi level splitting in the bulk. Notably, the drift-diffusion model predicts that beyond a thin region of approximately 15 nm close to the an- ode, both the quasi-Fermi level splitting and the carrier concentrations remain unchanged regardless of the anode work function. To check this prediction, we measured the  a  a - 0 . 5 0 . 0 0 . 5 1 . 0 1 . 5 - 0 . 5 0 . 0 0 . 5 1 . 0 1 . 5 0 2 0 4 0 6 0 8 0 1 0 0 1 0 1 1 1 0 1 3 1 0 1 5 1 0 1 7 1 0 1 9 0 2 0 4 0 6 0 8 0 1 0 0 1 0 1 1 1 0 1 3 1 0 1 5 1 0 1 7 1 0 1 9 ( b ) E F , n n = 2 5 0 m e V E F , n E n e r g y ( e V ) q V o c q V o c , m a x E F , p n = 0 E F , p E n e r g y ( e V ) ( a ) C a r r i e r D e n s i t y n , p ( c m - 3 ) D i s t a n c e f r o m A n o d e x ( n m ) n p ( d ) ( c ) C a r r i e r D e n s i t y n , p ( c m - 3 ) D i s t a n c e f r o m A n o d e x ( n m ) n p L i g h t I n t e n s i t y 5 of 250 meV has such a strong effect on both the magni- tude and light intensity dependence of Voc. In the fol- lowing we show that this is a direct consequence of the highly imbalanced mobilities. B. Effect of imbalanced charge transport on the open-circuit voltage Having shown that our numerical model describes the experimental data well, we will now use it to discuss the effect of charge transport in more detail. Figure 5(a) demonstrates how an injection barrier at the anode (sim- ilar to the PEDOT:PSS device) affects the hole quasi- Fermi level for different ratios between µn and µp. For balanced mobilities (µn = µp), the injection barrier in- duces a certain gradient dEF,p/dx, which leads to a volt- age loss ∆Voc,1 compared to the case with an Ohmic hole contact. If we now lower the hole mobility by one or two orders of magnitude, the gradient of the quasi-Fermi level increases significantly. This can be reflected by introduc- ing a second loss component ∆Voc,2 due to the imbal- anced charge transport. Hence, the total loss in Voc can be expressed as ∆Voc = ∆Voc,1 + ∆Voc,2. (6) Another possible loss mechanism would be a reduc- tion of the quasi-Fermi level splitting (and, thus, the car- rier concentration) in the bulk due to very strong surface recombination [22]. However, such a reduction is not present here, which is both evident from the bulk recom- bination measurements (see Fig. 4) and the additional band diagrams shown in the Supplemental Material [26]. Figure 5(b) and 5(c) illustrate the effect of imbalanced charge transport in more detail. In Fig. 5(b), an accu- mulation of holes close to the anode for µn/µp (cid:29) 1 is clearly seen. However, at the same time, the absolute value of µp is decreased. Hence, it is worthwhile to take a look at the conductivity σp = qµpp. Figure 5(c) shows that the increase of the hole concentration is not large enough to compensate the decrease of the hole mobil- ity. At the same time, the conductivity σn for electrons is nearly unaffected. Hence, assuming more imbalanced charge transport effectively decreases the difference be- tween σp and σn close to the anode. This can be under- stood in terms of a further loss of selectivity or an virtual increase of the injection barrier height [8, 22]. Figure 6 shows Voc as a function of µp for a fixed elec- tron mobility of µn = 10−4 cm2/Vs and different barrier heights ϕan. First, we discuss the case of a constant β, which means the mobility does not affect the bulk re- combination [Fig. 6(a)]. Only a weak variation of Voc with µp is then visible for an Ohmic contact (ϕan = 0). The slight decline of Voc is due to the fact that also the cathode is assumed to be non-selective. When µp is very large (µp (cid:29) µn), the device is dominated by hole trans- port, so that surface recombination at the cathode be- comes limiting. If we now introduce a significant bar- FIG. 4. Results of charge extraction and transient photovolt- age measurements for MoOx (squares) and PEDOT:PSS (cir- cles) devices. (a) Extracted charge carrier concentration ver- sus open-circuit voltage. (b) Recombination rate constant β versus carrier concentration. carrier concentration under open-circuit conditions us- ing bias-assisted charge extraction [23, 35]. Figure 4(a) shows indeed only a constant voltage shift ∆Voc between the data points for the MoOx and the PEDOT:PSS sam- ple, while the carrier concentration at a given light in- tensity remains unchanged. Furthermore, we performed transient photovoltage ex- periments to determine the carrier lifetime. By plotting the lifetime versus the carrier concentration (see Supple- mental Material [26]) we obtained a reaction order close to 2 in both cases. This indicates that the recombina- tion in the bulk is bimolecular, independent of whether an Ohmic or an non-Ohmic hole contact is present. That the nature of the contact does not affect the recombina- tion in the bulk is also evident from Fig. 4(b), where we plot the recombination rate constant β as a function of the carrier concentration. For both devices we find fairly similar values of β ∼ 10−11 cm3s−1. Hence, we can conclude that the variation in Voc be- tween the MoOx and PEDOT:PSS devices is solely re- lated to a gradient in the hole quasi-Fermi level at the anode, while the recombination in the bulk is largely un- affected. This is in line with previous studies on material combinations with balanced mobilities [13, 22]. However, it seems surprising that a relatively low injection barrier  0 . 7 0 . 8 0 . 9 1 . 0 1 0 1 4 1 0 1 5 1 0 1 6 1 0 1 4 1 0 1 5 1 0 1 6 1 0 - 1 2 1 0 - 1 1 1 0 - 1 0 C a r r i e r C o n c e n t r a t i o n ( c m - 3 ) O p e n - C i r c u i t V o l t a g e V o c ( V ) ( a ) D V o c ( c m 3 s - 1 ) C a r r i e r C o n c e n t r a t i o n ( c m - 3 ) ( b ) M o O x P E D O T : P S S 6 FIG. 5. (a) Quasi-Fermi level EF,p close to the anode (lo- cated at x = 0) for an ideal device with Ohmic contacts and balanced transport (dashed line), and for a device with an injection barrier and varied µn/µp (solid lines). The arrows indicate the voltage loss due to the barrier height (1) and due to the imbalanced mobilities (2), respectively. Panels (b) and (c) show the corresponding hole concentration p and hole conductivity σp = qµpp. rier at the anode (ϕan (cid:29) kT ), it is clearly seen that the voltage loss ∆Voc becomes determined by the mo- bility ratio. For µp = µn, the reduction of Voc is solely caused by ∆Voc,1, which is proportional to ϕan. This no longer holds true for imbalanced mobilities. In the case µp (cid:28) µn, the voltage loss is further increased by the mobility-dependent ∆Voc,2, and a logarithmic depen- dence of Voc on µp can be seen. In contrast, for µp (cid:29) µn, surface recombination is partly compensated, as charges accumulate now at the (Ohmic) cathode. As a result, the total voltage loss is effectively reduced (∆Voc,2 < 0). We also did simulations for d = 70 and 140 nm (see Sup- plemental Material [26]). Previously, we have shown that this thickness range produces clear differences in the com- petition between charge extraction and bimolecular re- combination [23]. However, we find here that the mobil- ity dependence of Voc is fairly unaffected by the active- layer thickness. This shows that our results are indepen- dent on the collection of majority carriers. FIG. 6. Simulated Voc for a varied hole mobility but fixed electron mobility of µn = 10−4 cm2/Vs and different bar- rier heights ϕan at the anode. Panel (a) shows the case of a constant β, while for (b) we assumed reduced Langevin re- combination with β = ζβL. Vertical dashed lines indicate the case µn = µp. The arrows in (a) exemplify the voltage loss ∆Voc,1 for a barrier of 300 meV and ∆Voc,2 for a mobility mismatch of µn/µp = 100, respectively. Next, we consider the case that also the bulk recombi- nation is limited by diffusion [Fig. 6(b)]. Such a process is commonly described by the Langevin model, predicting a mobility-dependent recombination coefficient βL = q εε0 (µn + µp). (7) However, it is known that the recombination in phase- separated organic blends is reduced compared to the Langevin model, β = ζβL, where ζ is a reduction fac- tor [36 -- 38]. Here, we chose ζ = 0.1, so that with the given mobilities of the SQIB:PCBM system, the experi- mental value of β is reproduced. For µp (cid:28) µn, no dif- ference in the mobility dependence of Voc can be seen between Fig. 6(a) and Fig. 6(b), as β is largely deter- mined by the fixed µn. If the mobilities are similar, there is a significant contribution of µp to the magni- ) µ p - 0 . 2 - 0 . 1 0 . 0 1 0 1 6 1 0 1 7 0 1 0 2 0 3 0 4 0 1 0 - 9 1 0 - 8 1 0 - 7 1 0 - 6 1 0 - 5 µ n / µ p = 1 0 0 µ n / µ p = 1 0 Q u a s i - F e r m i L e v e l E F , p ( e V ) ( 2 ) ( a ) ( 1 ) O h m i c µ n / µ p = 1 p ( c m - 3 ) ( b ) µ n / µ p ( S / c m n / µ p ( c ) D i s t a n c e f r o m A n o d e x ( n m )   a    L 1 0 - 7 1 0 - 6 1 0 - 5 1 0 - 4 1 0 - 3 1 0 - 2 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 1 0 - 7 1 0 - 6 1 0 - 5 1 0 - 4 1 0 - 3 1 0 - 2 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 ( a ) H o l e M o b i l i t y µ p ( c m 2 / V s ) O p e n - C i r c u i t V o l t a g e V o c ( V ) n ( e V ) 0 0 . 1 0 0 . 2 0 0 . 2 5 0 . 3 0 D V o c , 2 D V o c , 1 = c o n s t ( b ) = H o l e M o b i l i t y µ p ( c m 2 / V s ) O p e n - C i r c u i t V o l t a g e V o c ( V ) In the case µp (cid:29) µn, the coefficient β be- tude of β. comes so large that the device is entirely dominated by bulk recombination, and the barrier at the anode is no longer relevant. Altogether, this results in a maximum of Voc(µp), which shifts towards larger values of µp with increasing ϕan. Such an optimum value of Voc is unique for imbalanced charge transport and was not observed in previous studies, where µn and µp were varied simulta- neously [7, 14, 15]. Our numerical results show that independent of the bulk recombination mechanism, surface recombination at the anode critically determines Voc for imbalanced mo- In contrast, for µp (cid:29) µn, the bilities with µp (cid:28) µn. quality of the anode is less important. We note that our conclusions are directly transferable to the case of a non- Ohmic cathode. A significant barrier ϕcat would severely limit Voc for µn (cid:28) µp, which is a relevant scenario for non-fullerene solar cells [3, 18, 19]. C. Analytical expression for Voc in the case of imbalanced mobilities Recently, Sandberg et al. [12] provided analytical means to describe Voc for different cases related to sur- face recombination. If one contact is non-Ohmic (here the anode) and surface recombination is limited by the diffusion of minority carriers (here electrons; effective ve- locity vd,n) rather than the interface kinetics (S (cid:29) vd,n), the authors derived the expression qVoc = Eg − ϕan − kT ln(cid:18) vd,nN Gd (cid:19) (8) for low light intensities, where bulk recombination is neg- ligible. When Voc is far from flat-band conditions, we have vd,n ≈ µnF (0), where F (0) is the electric field close to the anode. Equation (8) is valid as long as L∗p (cid:29) d, with the effective diffusion length 7 FIG. 7. Simulated Voc (solid lines) versus light intensity for a device having a non-Ohmic anode (ϕan = 0.3 eV) at different ratios between µn and µp. Equations (3) and (11) are indi- cated by dashed and dotted lines, respectively. Our extended analytical expression Eq. (11) shows striking agreement with the simulations for highly imbalanced mobilities. The dash- dotted line indicates Eq. (8), valid in the low-intensity regime, with vd,n approximated according to Ref. 12. For µp (cid:28) µn, a considerable pile-up of holes is taking place close to the anode at high light intensities, as is clearly visible in Fig. 3(d). As evident from Fig. 3(b), the resulting space charge region of holes gives rise to a photo-induced upward energy-level bending near the anode. In the hole-dominated space charge region, the bulk recombination of holes is negligibly small. Instead, at open-circuit conditions, the hole current within this re- gion is balanced by an equal, but opposite surface recom- bination current of electrons, diffusing against the band bending at the anode. After accounting for the hole- induced energy-level bending and the associated electron diffusion, Eq. (10) modifies to L∗p ≈(cid:18) µeffkT q√βG(cid:19)1/2 , (9) qVoc = Eg − ϕan − kT 2 nN 2 ln(cid:18) qµ2 εε0µpG(cid:19) , (11) where µeff = 2√µnµp is an effective mobility. At high enough light intensities, so that L∗p (cid:28) d and flat-band conditions prevail on the anode side of the active layer, the surface recombination is restricted to a region given by the effective diffusion length. Under these conditions, Voc can be approximated by [10, 12] qVoc ≈ Eg − ϕan − kT 2 ln(cid:18) µnβN 2 µpG (cid:19) , (10) which in the limit µn → µp is equivalent to the result of Solak et al. [11]. We note that Eq. (10) already pro- vides a framework to predict Voc for moderate mobil- ity contrasts; however, its derivation assumes that flat- band conditions prevail close to the surface recombina- tion dominated region near the anode. This is no longer valid for for highly imbalanced mobilities. as shown in Supplemental Material [26]. The main fea- ture of Eq. (11) is that it no longer depends on the bi- molecular recombination strength β. Hence, in the limit of highly imbalanced mobilities and at high enough light intensities, Voc becomes independent of bulk recombina- tion and is solely given by the contacts and the charge transport in the active layer. Equation (11) also ex- plains the logarithmic µp dependence at high mobility contrasts µn/µp (cid:29) 1 seen in Fig. 6. The voltage loss due to the imbalanced mobilities is then given by ∆Voc,2 = kT 2 ln(cid:18) qµ2 εε0µpβ(cid:19) . n (12) As shown in Fig. 7 for mobility contrasts of one and two orders of magnitude, the modified analytical expres- sion in Eq. (11) describes the numerical data well in the 1 0 - 4 1 0 - 3 1 0 - 2 1 0 - 1 1 0 0 1 0 1 1 0 2 1 0 3 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 E q . ( 8 ) E q . ( 3 ) E q . ( 1 1 ) L i g h t I n t e n s i t y I ( m W / c m 2 ) O p e n - C i r c u i t V o l t a g e V o c ( V ) µ n / µ p 1 1 0 1 0 0 high-intensity regime. We note that as long as the anode can be considered as non-Ohmic (ϕan (cid:29) kT ), this holds true also for other injection barrier heights (see Supple- mental Material [26]). In the low-intensity regime, Voc becomes independent of the mobility contrast and can be described by Eq. (8) instead. Our analysis demonstrates that the transition is shifted towards lower photogenera- tion for increasing µn/µp. Furthermore, we checked the validity of our analytical framework, as detailed in the Supplemental Material [26]. We find that Eq. (11) predicts the open-circuit voltage for mobility contrasts µn/µp ≥ 5 with an relative error below 1% (for the parameters in Tab. I). It is worth not- ing that Eq. (10), which neglects band bending in the hole-induced space charge region, reaches a similar accu- racy only for a mobility imbalance of less than a factor of 2. Finally, we point out that Eq. (1) is in none of the cases presented in Fig. 7 suitable to describe the data, even in the low-intensity regime and even though only pure bimolecular recombination was assumed in the sim- ulation. Hence, special care has to be taken when trying to assess information about the bulk recombination from the slope of Voc versus ln(I), also called the light ideality factor [39]. If the contacts are not sufficiently selective, the ideality factor will always be affected by surface re- combination. IV. CONCLUSION In summary, we have studied how imbalanced charge transport affects the interplay of bulk and surface re- combination in organic solar cells. Combining experi- ments and simulations for a blend system with a strong mobility mismatch, we have identified two cases with re- spect to the energy level alignment at the electrodes: For 8 Ohmic contacts, the open-circuit voltage Voc still is rep- resentative of the quasi-Fermi level splitting in the bulk, even though the mobilities µn and µp are highly imbal- anced. However, if one contact is non-Ohmic, Voc be- comes critically determined by the mobility ratio. For the devices studied herein (µn/µp (cid:29) 1), we find that surface recombination at the anode reduces Voc more strongly than it would be the case with balanced mobilities. The reason is that with decreasing µp, a larger gradient of the quasi-Fermi level is required to cancel out the sur- face recombination current of electrons. An analogous situation occurs for a device dominated by hole trans- port (µn/µp (cid:28) 1) at a the cathode. Hence, it is prop- erties of the photoactive blend that decide whether an electrode can be considered appropriate. We have also derived analytical equations for Voc that take into account the pile-up of space charge due to highly imbalanced mobilities. In particular, Eq. (11) shows ex- cellent agreement with the data from our experimentally validated numerical device model. With this we hope to provide a framework that helps researchers in designing efficient organic photovoltaics from materials with imbal- anced charge transport. ACKNOWLEDGMENTS The authors thank Matthias Schulz and Arne Lutzen (University of Bonn, Germany) for providing the squaraine dye, as well as Mathias Nyman and Staffan Dahlstrom (Abo Akademi University) for fruitful discus- sions. D. S., S. W. and M. S. thank Jurgen Parisi for con- stant support of their research in Oldenburg. Financial support by the Magnus Ehrnrooth Foundation and the Research Mobility Programme of Abo Akademi Univer- sity is gratefully acknowledged. [1] C. Deibel and V. Dyakonov, "Polymer-Fullerene Bulk Heterojunction Solar Cells," Rep. Prog. Phys. 73, 96401 (2010). [2] A. Mishra and P. Bauerle, "Small Molecule Organic Semiconductors on the Move: Promises for Future Solar Energy Technology," Angew. Chem. Int. Ed. 51, 2020 -- 2067 (2012). [3] J. Hou, O. Inganas, R. H. Friend, and F. Gao, "Or- ganic Solar Cells Based on Non-Fullerene Acceptors," Nat. Mater. 17, 119 -- 128 (2018). [4] Peter Wurfel, Physics of Solar Cells, 2nd ed. (Wiley- VCH, Weinheim, Germany, 2009). [5] K. Vandewal, K. Tvingstedt, A. Gadisa, O. Inganas, and J. V. Manca, "On the Origin of the Open-Circuit Voltage of Polymer-Fullerene Solar Cells," Nat. Mater. 8, 904 -- 909 (2009). [6] L. J. A. Koster, V. D. Mihailetchi, R. Ramaker, and P. W. M. Blom, "Light Intensity Dependence of Open- Circuit Voltage of Polymer:Fullerene Solar Cells," Appl. Phys. Lett. 86, 123509 (2005). [7] W. Tress, K. Leo, and M. Riede, "Optimum Mobility, Contact Properties, and Open-Circuit Voltage of Organic Solar Cells: A Drift-Diffusion Simulation Study," Phys. Rev. B 85, 155201 (2012). [8] J. Reinhardt, M. Grein, Buhler C., M. Schubert, and U. Wurfel, "Identifying the Impact of Surface Recombi- nation at Electrodes in Organic Solar Cells by Means of Electroluminescence and Modeling," Adv. Energy Mater. 4, 1400081 (2014). [9] A. Wagenpfahl, D. Rauh, M. Binder, C. Deibel, and V. Dyakonov, "S-Shaped Current-Voltage Characteris- tics of Organic Solar Devices," Phys. Rev. B 82, 115306 (2010). [10] O. J. Sandberg, M. Nyman, and R. Osterbacka, "Effect of Contacts in Organic Bulk Heterojunction Solar Cells," Phys. Rev. Appl. 1, 24003 (2014). [11] S. Solak, P. W. M. Blom, and G. A. H. Wetzelaer, "Ef- fect of Non-Ohmic Contacts on the Light-Intensity De- pendence of the Open-Circuit Voltage in Organic Solar Cells," Appl. Phys. Lett. 109, 053302 (2016). [12] O. J. Sandberg, A. Sundqvist, M. Nyman, and R. Osterbacka, "Relating Charge Transport, Contact Properties, and Recombination to Open-Circuit Voltage in Sandwich-Type Thin-Film Solar Cells," Phys. Rev. Appl. 5, 44005 (2016). [13] S. Wheeler, F. Deledalle, N. Tokmoldin, T. Kirchartz, J. Nelson, and J. R. Durrant, "Influence of Surface Re- combination on Charge-Carrier Kinetics in Organic Bulk Heterojunction Solar Cells with Nickel Oxide Interlay- ers," Phys. Rev. Appl. 4, 24020 (2015). [14] T. Kirchartz, B. E. Pieters, K. Taretto, and U. Rau, "Mobility Dependent Efficiencies of Organic Bulk Het- erojunction Solar Cells: Surface Recombination and Charge Transfer State Distribution," Phys. Rev. B 80, 035334 (2009). [15] A. Wagenpfahl, C. Deibel, and V. Dyakonov, "Organic Solar Cell Efficiencies Under the Aspect of Reduced Sur- face Recombination Velocities," IEEE J. Sel. Top. Quan- tum Electron. 16, 1759 -- 1763 (2010). [16] J. A. Bartelt, D. Lam, T. M. Burke, S. M. Sweetnam, and M. D. McGehee, "Charge-Carrier Mobility Require- ments for Bulk Heterojunction Solar Cells with High Fill Factor and External Quantum Efficiency >90%," Adv. Energy Mater. 5, 1500577 (2015). [17] M. Stolterfoht, A. Armin, S. Shoaee, I. Kassal, P. Burn, and P. Meredith, "Slower Carriers Limit Charge Gen- eration in Organic Semiconductor Light-Harvesting Sys- tems," Nat. Commun. 7, 11944 (2016). [18] S. Holliday, R. S. Ashraf, A. Wadsworth, D. Baran, S. A. Yousaf, C. B. Nielsen, C.-H. Tan, S. D. Dimitrov, Z. Shang, N. Gasparini, M. Alamoudi, F. Laquai, C. J. Brabec, A. Salleo, J. R. Durrant, and I. McCulloch, "High-Efficiency and Air-Stable P3HT-Based Polymer Solar Cells with a New Non-Fullerene Acceptor," Nat. Commun. 7, 11585 (2016). [19] C. Yan, S. Barlow, Z. Wang, H. Yan, A. K. Y. Jen, S. R. Marder, and X. Zhan, "Non-Fullerene Acceptors for Or- ganic Solar Cells," Nat. Rev. Mater. 3, 18003 (2018). [20] V. D. Mihailetchi, J. Wildeman, and P. W. M. Blom, "Space-Charge Limited Photocurrent," Phys. Rev. Lett. 94, 126602 (2005). [21] M. Stolterfoht, A. Armin, B. Philippa, R. D. White, P. L. Burn, P. Meredith, G. Juska, and A. Pivrikas, "Photo- carrier Drift Distance in Organic Solar Cells and Pho- todetectors," Sci. Rep. 5, 9949 (2015). [22] A. Spies, M. List, T. Sarkar, and U. Wurfel, "On the Impact of Contact Selectivity and Charge Transport on the Open-Circuit Voltage of Organic Solar Cells," Adv. Energy Mater. 7, 201601750 (2017). [23] D. Scheunemann, O. Kolloge, S. Wilken, M. Mack, J. Parisi, M. Schulz, A. Lutzen, and M. Schiek, "Re- vealing the Recombination Dynamics in Squaraine-Based Bulk Heterojunction Solar Cells," Appl. Phys. Lett. 111, 183502 (2017). [24] S. Bruck, C. Krause, R. Turrisi, L. Beverina, S. Wilken, W. Saak, A. Lutzen, H. Borchert, M. Schiek, and J. Parisi, "Structure-Property Relationship of Anilino- Squaraines in Organic Solar Cells," Phys. Chem. Chem. Phys. 16, 1067 -- 1077 (2014). [25] O. S. Abdullaeva, M. Schulz, F. Balzer, J. Parisi, A Lutzen, K. Dedek, and M. Schiek, "Photo- electrical Stimulation of Neuronal Cells by an Or- 9 ganic Semiconductor-Electrolyte Interface," Langmuir 32, 8533 (2016). [26] See Supplemental Material for a determination of the built-in voltage and the reaction order, additional nu- merical simulations and a derivation of the analytical ex- pression for highly imbalanced mobilities. [27] Y. Liu, F. Zhang, H. Dai, W. Tang, Z. Wang, J. Wang, Aiwei Tang, Hongsheng Peng, Zheng Xu, and Y. Wang, "Interfacial Layer for Efficiency Improvement of Solution- Processed Small Molecular Solar Cells," Sol. Energy Mater. Sol. Cells 118, 135 -- 140 (2013). [28] Q. Yang, D. Yang, S. Zhao, Y. Huang, Z. Xu, X. Liu, Xing Fan, Qingyu Huang, and X. Xu, "The Improved Performance of Solution-Processed SQ:PC71BM Photo- voltaic Devices via MoO3 as the Anode Modification Layer," Appl. Surf. Sci. 284, 849 -- 854 (2013). [29] G. Chen, H. Sasabe, Z. Wang, X. Wang, Z. Hong, J. Ki- doa, and Y. Yang, "Solution-Processed Organic Photo- voltaic Cells Based on a Squaraine Dye," Phys. Chem. Chem. Phys. 14, 14661 -- 14666 (2012). [30] M. Burgelman, P. Nollet, and S. Degrave, "Modelling Polycrystalline Semiconductor Solar Cells," Thin Solid Films 361-362, 527 -- 532 (2000). [31] L. A. A. Pettersson, L. S. Roman, and O. Inganas, "Mod- eling Photocurrent Action Spectra of Photovoltaic De- vices Based on Organic Thin Films," J. Appl. Phys. 86, 487 -- 496 (1999). [32] G. F. Burkhard, E. T. Hoke, and M. D. McGehee, "Ac- counting for Interference, Scattering, and Electrode Ab- sorption to Make Accurate Internal Quantum Efficiency Measurements in Organic and Other Thin Solar Cells," Adv. Mater. 22, 3293 -- 3297 (2010). [33] S. Wilken, J. Parisi, and H. Borchert, "Role of Oxygen Adsorption in Nanocrystalline ZnO Interfacial Layers for Polymer -- Fullerene Bulk Heterojunction Solar Cells," J. Phys. Chem. C 118, 19672 -- 19682 (2014). [34] A. Sundqvist, O. J. Sandberg, M. Nyman, J.-H. Smatt, and R. Osterbacka, "Origin of the S-Shaped JV Curve and the Light-Soaking Issue in Inverted Organic Solar Cells," Adv. Energy Mater. 6, 1502265 (2016). [35] J. Kniepert, I. Lange, N. J. Van Der Kaap, L. J. A. Koster, and D. Neher, "A Conclusive View on Charge Generation, Recombination, and Extraction in As- Prepared and Annealed P3HT:PCBM Blends: Com- bined Experimental and Simulation Work," Adv. Energy Mater. 4, 1301401 (2014). [36] A. Pivrikas, G. Juska, A. J. Mozer, M. Scharber, K. Ar- lauskas, N. S. Sariciftci, H. Stubb, and R. Osterbacka, "Bimolecular Recombination Coefficient as a Sensitive Testing Parameter for Low-Mobility Solar-Cell Materi- als," Phys. Rev. Lett. 94, 176806 (2005). [37] D. H. K. Murthy, A. Melianas, Z. Tang, G. Juska, K. Arlauskas, F. Zhang, L. D. A. Siebbeles, O. In- ganas, and T. J. Savenije, "Origin of Reduced Bimolecu- lar Recombination in Blends of Conjugated Polymers and Fullerenes," Adv. Funct. Mater. 23, 4262 -- 4268 (2013). [38] C. Gohler, A. Wagenpfahl, and C. Deibel, "Nongeminate Recombination in Organic Solar Cells," Adv. Electron. Mater. 4, 1700505 (2018). [39] T. Kirchartz, F. Deledalle, P. S. Tuladhar, J. R. Durrant, and J. Nelson, "On the Differences between Dark and Light Ideality Factor in Polymer:Fullerene Solar Cells," J. Phys. Chem. Lett. 4, 2371 -- 2376 (2013). Supplemental Material for "Effect of Imbalanced Charge Transport on the Interplay of Surface and Bulk Recombination in Organic Solar Cells" Dorothea Scheunemann∗ and Sebastian Wilken Physics, Faculty of Science and Engineering and Center for Functional Materials, Abo Akademi University, Porthansgatan 3, 20500 Turku, Finland and Institute of Physics, Energy and Semiconductor Research Laboratory, Carl von Ossietzky University of Oldenburg, 26111 Oldenburg, Germany Oskar J. Sandberg Department of Physics, Swansea University, Singleton Park, Swansea, SA2 8PP, Wales, United Kingdom Ronald Osterbacka Physics, Faculty of Science and Engineering and Center for Functional Materials, Abo Akademi University, Porthansgatan 3, 20500 Turku, Finland Manuela Schiek Institute of Physics, Energy and Semiconductor Research Laboratory, Carl von Ossietzky University of Oldenburg, 26111 Oldenburg, Germany arXiv:1905.01268v1 [physics.app-ph] 3 May 2019 ∗ [email protected] S1 CONTENTS Determination of the Build-In Voltage Photovoltaic Characteristics Surface Recombination Velocity Determination of the Reaction Order Additional Band Diagrams Variation of the Active-Layer Thickness Analytical Model Validity of Analytical Model for Different Injection Barrier Heights Validity and Limits of Analytical Equations References S2 S4 S5 S6 S7 S8 S9 S11 S12 S12 DETERMINATION OF THE BUILD-IN VOLTAGE To estimate the built-in voltage Vbi, we followed a method that has recently been devel- oped by Dahlstrom et al. [S1]. Therefore, we determined the maximum extraction current time tmax from charge carrier extraction by linearly increasing voltage (CELIV) measure- ments in the dark and at different offset voltages Voff. As shown in Ref. S1, tmax depends on the voltage according to t−2 max = qAµ kT d2 (Vbi − Voff), (S1) where A is the slope of the linearly increasing voltage pulse, µ the charge carrier mobility, k the Boltzmann constant, T the temperature, d the active layer thickness and q the elementary charge. Hence, Vbi can be estimated from a linear fit of t−2 max versus Voff, see Fig. S1. Prior to the analysis, we corrected the data for RC time effects by replacing tmax with tmax−3RsCgeo, where Rs is the sum of series and load resistance and Cgeo the geometrical capacitance of S2 the device. The intersection of the linear fits with the voltage axis yields Vbi ≈ 0.94 V for the PEDOT:PSS device and Vbi ≈ 1.22 V for the MoOx device. FIG. S1. Determination of Vbi for devices with a hole transport layer of MoOx (squares) and PEDOT:PSS (circles) from the results of dark CELIV measurements. S3 0 . 6 0 . 8 1 . 0 1 . 2 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 M o O x P E D O T : P S S t - 2 m a x ( µ s - 2 ) O f f s e t V o l t a g e ( V ) PHOTOVOLTAIC CHARACTERISTICS Current -- voltage characteristics under AM1.5G illumination were recorded with a semi- conductor characterization system (Keithley 4200) and a solar simulator (Photo Emission Tech., class AAA) at ambient conditions. The light intensity was adjusted to 100 mW/cm2 using a calibrated silicon solar cell. Spectral mismatch was not taken into account. TABLE S1. Average photovoltaic characteristics of the SQIB:PCBM solar cells with varied HTL under simulated AM1.5G illumination. HTL (nm) MoOx PEDOT:PSS Jsc (mA/cm2) 6.7 ± 0.1 6.1 ± 0.2 Voc (V) 0.92 ± 0.01 0.80 ± 0.01 FF (%) 38 ± 1 34 ± 1 PCE (%) 2.30 ± 0.03 1.67 ± 0.04 S4 SURFACE RECOMBINATION VELOCITY We also attempted to model the differences between the MoOx and the PEDOT:PSS devices by varying the surface recombination velocity S at the anode. As can be seen in Fig. S2, a reduction of S predominantly results in a deformation of the current-voltage curves around V = Voc ("S-shape"), in accordance with literature [S2, S3]. However, such an S-shape behavior was not present in our experimental data. FIG. S2. Simulated current-voltage curves for a SQIB:PCBM device with varying surface recom- bination velocity at the anode, but otherwise ideal contact properties (ϕan = ϕcat = 0). S5 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 - 1 0 - 8 - 6 - 4 - 2 0 2 4 C u r r e n t D e n s i t y ( m A / c m 2 ) V o l t a g e ( V ) D e c r e a s i n g S DETERMINATION OF THE REACTION ORDER To estimate the reaction order of the bulk recombination, we plotted the small-perturbation lifetime τ∆n from transient photovoltage (TPV) measurement as a function of the carrier concentration n from bias-extracted charge extraction (BACE) measurements under the same experimental conditions (Figure S3). According to previous studies [S4, S5], we ap- plied a power-law fit according to τ∆n ∝ n1−δ to these data, which yields an empirical reaction order of δ = 2.1 (MoOx) and δ = 2.3 (PEDOT:PSS), respectively. FIG. S3. Small-perturbation lifetime τ∆n versus carrier concentration n as obtained from combined TPV and BACE measurements. Dashed lines are fits according to τ∆n ∝ n1−δ. S6 1 0 2 0 1 0 2 1 1 0 2 2 1 0 - 6 1 0 - 5 1 0 - 4 M o O x P E D O T : P S S d ~ 2 . 3 d ~ 2 . 1 S m a l l P e r t u r b a t i o n L i f e t i m e ( s ) C a r r i e r C o n c e n t r a t i o n ( m - 3 ) ADDITIONAL BAND DIAGRAMS Figure S4 shows band diagrams for different ratios between the electron and hole mobility. It is clearly seen that, independent of the barrier height ϕan, the mobility contrast has virtually no effect on the quasi-Fermi level splitting within the photoactive layer. FIG. S4. Simulated band diagrams at open circuit and 1-sun illumination for an injection barrier height of (a) ϕan = 0 mV and (b) ϕan = 300 mV and for mobility contrasts of µn/µp = 100 (thin lines), µn/µp = 1 (medium lines) and µn/µp = 0.01 (thick lines), respectively. The electron mobility was fixed to µn = 10−4 cm2/Vs. S7 0 2 0 4 0 6 0 8 0 1 0 0 - 0 . 5 0 . 0 0 . 5 1 . 0 1 . 5 E F , p E n e r g y ( e V ) T h i c k n e s s d ( n m ) j a n = 0 m e V E F , n a ) 0 2 0 4 0 6 0 8 0 1 0 0 - 0 . 5 0 . 0 0 . 5 1 . 0 1 . 5 E F . p E n e r g y ( e V ) T h i c k n e s s d ( n m ) j a n = 3 0 0 m e V E F , n M o b i l i t y C o n t r a s t b ) VARIATION OF THE ACTIVE-LAYER THICKNESS FIG. S5. Simulated Voc for a varied hole mobility but fixed electron mobility of µn = 10−4 cm2/Vs and different barrier heights ϕan at the anode. Panel (a) shows the result for an active-layer thickness of d = 70 nm, and panel (b) for an active-layer thickness of d = 140 nm. S8  a 1 0 - 7 1 0 - 6 1 0 - 5 1 0 - 4 1 0 - 3 1 0 - 2 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 ( a ) H o l e M o b i l i t y µ p ( c m 2 / V s ) O p e n - C i r c u i t V o l t a g e V o c ( V ) n ( e V ) 0 0 . 1 0 0 . 2 0 0 . 2 5 0 . 3 0 d = 7 0 n m 1 0 - 7 1 0 - 6 1 0 - 5 1 0 - 4 1 0 - 3 1 0 - 2 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 ( b ) d = 1 4 0 n m O p e n - C i r c u i t V o l t a g e V o c ( V ) H o l e M o b i l i t y µ p ( c m 2 / V s ) ANALYTICAL MODEL We assume the electron and hole mobilities to be highly imbalanced, with holes being the slower charge carriers (µn/µp (cid:29) 1). Under these conditions, provided that the light intensity is high enough, holes will pile up within the active layer; as a result, a hole- dominated mobility-induced space charge region (0 < x < ws) is formed adjacent to the anode, where the thickness of the space charge region is given by ws < d. The space charge will screen the electric field in the region (x > ws) immediately outside the space charge region. In the region x > ws (not too close to the cathode), the electric field is close to zero and the generation rate is closely balanced by recombination. Since the space charge region is dominated by holes (p (cid:29) n), the bulk recombination of holes is negligible within this region (0 < x < ws). Assuming the hole current to be drift-dominated (i.e., the electric field is strong enough), the continuity equation, the hole current equation and the Poisson equation, respectively, in the region 0 < x < ws read: (S2) (S3) (S4) = G, djp dx 1 q jp(x) = qµpp(x)F (x), dF dx = q εε0 p(x). Then, assuming jp(ws) = F (ws) = 0, and taking the photogeneration rate G to be relatively homogenous within the space charge region, Eqs. (S2) to (S4) can be combined as jp(x) = −qG(ws − x) = µpεε0F (x) dF dx for x < ws and solved for the electric field, F (x) =s qG µpεε0 (x − ws). (S5) (S6) We note that this analysis yields the correct Goodman and Rose [S6] type voltage and generation dependence expected for space-charge-limited photocurrents, as discussed in Ref. S7. jph ≈ qGws ∝ G3/4pV0 − V , (S7) As the bulk recombination is assumed negligibly small within the space charge region, at open-circuit conditions the hole current is exactly balanced by an equal, but opposite S9 (S11) (S12) (S13) surface recombination current of electrons, so that j = jp(x) + jn(x) = 0. Subsequently, jp(0) = −jn(0) must hold true at Voc. A general expression for the surface recombination current jn(0) of electrons at the anode was derived by Sandberg et al. [S8] from the drift- diffusion equations. Equation (A4) from Ref. S8 can be readily rewritten as Z d 0 jn(x) exp(cid:20) Ec(x) − Ec(0) kT (cid:21) dx = µnkT nan(cid:20)exp(cid:18) qV kT(cid:19) − 1(cid:21) , (S8) whereby it is assumed that the electron transport is limited by diffusion rather than by the interface kinetics at the anode. Here, nan = N exp [−(Eg − ϕan)/kT ] is the equilibrium electron density at the anode. Provided that the bulk recombination within the space charge region remains negligible, we may write the electron current as (S9) (S10) = G djn dx 1 q − jn(x) = jn(0) − qGx. for x < ws, which then yields Since the main contribution to the integral in Eq. (S8) is from the region close to x = 0, we can approximate Ec(x) ≈ Ec(0) + qF (0)x and rewrite Eq. (S8) as jn(0) ≈ −qµnF (0)nan(cid:20)exp(cid:18) qV kT(cid:19) − 1(cid:21) under the assumption qF (0)ws (cid:29) kT . At open circuit (V = Voc), where jp(0) = −jn(0) = −qGws, it then follows directly form Eq. (S6) and Eq. (S11) that exp(cid:18) qVoc kT (cid:19) ≈ − Gws µnF (0)nan = 1 nansεε0µpG qµ2 n assuming qVoc (cid:29) kT . Using the definition of nan, we finally arrive at qVoc = Eg − ϕan − kT 2 nN 2 ln(cid:18) qµ2 εε0µpG(cid:19) . Comparing this expression with Eq. (3) in the main text, the enhanced surface recombination loss in the open-circuit voltage due to the mobility imbalance is given by ∆Voc,2 = kT 2 ln(cid:18) qµ2 εε0µpβ(cid:19) . n (S14) S10 VALIDITY OF ANALYTICAL MODEL FOR DIFFERENT INJECTION BAR- RIER HEIGHTS FIG. S6. Simulated open-circuit voltage (solid lines) versus light intensity for a device with a mobility contrast of µn/µp = 100 and different injection barrier heights ϕan at the anode. Dotted lines are the prediction of Eq. (11) in the main text. S11  a 1 0 - 4 1 0 - 3 1 0 - 2 1 0 - 1 1 0 0 1 0 1 1 0 2 1 0 3 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 . 0 E q . ( 1 1 ) L i g h t I n t e n s i t y I ( m W / c m 2 ) O p e n - C i r c u i t V o l t a g e V o c ( V ) n ( e V ) 0 . 5 0 . 4 0 . 3 0 . 2 0 . 1 VALIDITY AND LIMITS OF ANALYTICAL EQUATIONS FIG. S7. (a) Simulated dependence of Voc on the mobility contrast µn/µp for different injection barrier heights ϕan at the anode. Solid lines are the prediction of Eq. (11) in the main text. (b) Relative error between numerical and analytical data. Shadowed area corresponds the the region with < 1% error. (c) Numerical data for an injection barrier of ϕan = 0.3 eV together with the prediction of Eq. (10) in the main text. Shadowed area indicates the region with < 1% error between simulation and analytical model. (d) Comparison of the relative error produced by Eq. (10) and Eq. (11), respectively, for a device with ϕan = 0.3 eV. S1 S. Dahlstrom, O. J. Sandberg, M. Nyman, and R. Osterbacka, Phys. Rev. Appl. 10, 054019 (2018). S12  a  a 1 0 0 1 0 1 0 . 1 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 0 . 9 5 1 0 0 1 0 1 0 . 1 0 1 2 3 4 5 6 L i n e s : E q . ( 1 1 ) n ( e V ) 0 . 2 0 . 2 5 0 . 3 O p e n - C i r c u i t V o l t a g e V o c ( V ) M o b i l i t y C o n t r a s t µ n / µ p ( b ) ( a ) n ( e V ) 0 . 2 0 . 2 5 0 . 3 R e l a t i v e E r r o r ( % ) M o b i l i t y C o n t r a s t µ n / µ p  a 1 0 1 0 . 1 0 . 7 5 0 . 8 0 0 . 8 5 0 . 9 0 1 0 0 1 0 1 0 1 2 3 ( c ) E q . ( 1 0 ) O p e n - C i r c u i t V o l t a g e V o c ( V ) M o b i l i t y C o n t r a s t µ n / µ p n = 0 . 3 e V ( d ) R e l a t i v e E r r o r ( % ) M o b i l i t y C o n t r a s t µ n / µ p E q . ( 1 0 ) E q . ( 1 1 ) S2 S. Wilken, J. Parisi, and H. Borchert, J. Phys. Chem. C 118, 19672 (2014). S3 A. Sundqvist, O. J. Sandberg, M. Nyman, J.-H. Smatt, and R. Osterbacka, Adv. Energy Mater. 6, 1502265 (2016). S4 A. Foertig, J. Kniepert, M. Gluecker, T. Brenner, V. Dyakonov, D. Neher, and C. Deibel, Adv. Funct. Mater. 24, 1306 (2014). S5 S. Wheeler, F. Deledalle, N. Tokmoldin, T. Kirchartz, J. Nelson, and J. R. Durrant, Phys. Rev. Appl. 4, 24020 (2015). S6 A. M. Goodman and A. Rose, J. Appl. Phys. 42, 2823 (1971). S7 O. J. Sandberg, "Charge Collection in Thin-Film Devices Based on Low-Mobility Semiconduc- tors: Theory, Simulation, and Applications to Organic Solar Cells," (PhD Thesis, Abo Akademi University, Turku, Finland, 2018), pp. 101 -- 103. S8 O. J. Sandberg, A. Sundqvist, M. Nyman, and R. Osterbacka, Phys. Rev. Appl. 5, 44005 (2016). S13
1801.00750	1	1801	2017-11-17T17:54:30	Integrated RF-photonic Filters via Photonic-Phononic Emit-Receive Operations	[ "physics.app-ph", "physics.optics" ]	The creation of high-performance narrowband filters is of great interest for many RF-signal processing applications. To this end, numerous schemes for electronic, MEMS-based, and microwave photonic filters have been demonstrated. Filtering schemes based on microwave photonic systems offer superior flexibility and tunability to traditional RF filters. However, these optical-based filters are typically limited to GHz-widths and often have large RF insertion losses, posing challenges for integration into high-fidelity radiofrequency circuits. In this article, we demonstrate a novel type of microwave filter that combines the attractive features of microwave photonic filters with high-Q phononic signal processing using a photonic-phononic emit-receive process. Through this process, a RF signal encoded on a guided optical wave is transduced onto a GHz-frequency acoustic wave, where it may be filtered through shaping of acoustic transfer functions before being re-encoded onto a spatially separate optical probe. This emit-receive functionality, realized in an integrated silicon waveguide, produces MHz-bandwidth band-pass filtering while supporting low RF insertion losses necessary for high dynamic range in a microwave photonic link. We also demonstrate record-high internal efficiency for emit-receive operations of this type, and show that the emit-receive operation is uniquely suitable for the creation of serial filter banks with minimal loss of fidelity. This photonic-phononic emitter-receiver represents a new method for low-distortion signal-processing in an integrated all-silicon device.	physics.app-ph	physics	Integrated RF-photonic Filters via Photonic-Phononic Emit-Receive Operations Eric A. Kittlaus1, Prashanta Kharel,1 Nils T. Otterstrom,1 Zheng Wang,2 and Peter T. Rakich1 1Department of Applied Physics, Yale University, New Haven, CT 06520 USA. 2Microelectronics Research Center, Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX 78758 USA. (Dated: January 3, 2018) The creation of high-performance narrowband filters is of great interest for many RF-signal pro- cessing applications. To this end, numerous schemes for electronic, MEMS-based, and microwave photonic filters have been demonstrated. Filtering schemes based on microwave photonic systems offer superior flexibility and tunability to traditional RF filters. However, these optical-based filters are typically limited to GHz-widths and often have large RF insertion losses, posing challenges for integration into high-fidelity radiofrequency circuits. In this article, we demonstrate a novel type of microwave filter that combines the attractive features of microwave photonic filters with high-Q phononic signal processing using a photonic-phononic emit-receive process. Through this process, a RF signal encoded on a guided optical wave is transduced onto a GHz-frequency acoustic wave, where it may be filtered through shaping of acoustic transfer functions before being re-encoded onto a spatially separate optical probe. This emit-receive functionality, realized in an integrated silicon waveguide, produces MHz-bandwidth band-pass filtering while supporting low RF insertion losses necessary for high dynamic range in a microwave photonic link. We also demonstrate record-high internal efficiency for emit-receive operations of this type, and show that the emit-receive opera- tion is uniquely suitable for the creation of serial filter banks with minimal loss of fidelity. This photonic-phononic emitter-receiver represents a new method for low-distortion signal-processing in an integrated all-silicon device. INTRODUCTION Rapid growth in wireless communications has created a pressing demand for high-performance and reconfigurable schemes for radio-frequency (RF) signal processing, spec- tral awareness, and frequency-agile filtering of microwave signals [1–3]. As a result, considerable work has been di- rected toward creating high-performance, frequency-agile RF filters with narrow spectral bandwidths. While tradi- tional electronic filters can support narrow bandwidths, it is challenging to design such filters with rapid wideband tunability [4]. By contrast, filters based on microwave (or RF-) photonic platforms offer superior tunability and the potential for integration in complex photonic circuits [2, 4, 5]. However, integrated all-optical RF-photonic fil- ters are typically limited to spectral resolutions above 100- 1000 MHz and often have trade-offs between performance metrics such as bandwidth, rejection, and nonlinearity [2, 5, 6]. Moreover, many of these systems are plagued by high RF insertion losses due to inefficiencies in RF-optical conversion and low power handling in photonic integrated circuits [3]. On-chip stimulated Brillouin scattering (SBS) has re- cently emerged as a powerful tool for hybrid photonic- phononic signal processing operations, many of which have no analogue in all-optical systems [6–9]. SBS, which re- sults from the interaction between optical fields and acous- tic phonons in the MHz-GHz range, has been used for op- tical amplification [10–14], RF filtering [15–18] and signal processing [19–22], nonreciprocity [23–26] and time delay [27, 28]. The recent realization of SBS in silicon waveg- uides [29–31] has been used to create all-silicon amplifiers [12, 13, 32], a silicon-based Brillouin laser oscillator [33], and new architectures which permit MHz-bandwidth fil- tering operations [6, 8, 9] with the potential for large-scale integration into silicon photonic circuits. A promising method to develop Brillouin-based RF fil- ters that has recently been explored harnesses phonon- mediated emit-and-receive coupling between distinct nanophotonic waveguides [8]. Through this emit-receive process, forces produced by an intensity-modulated op- tical signal in an emitter waveguide transduce a coher- ent travelling-wave phonon. The device structure is de- signed to shape the transfer function of this phononic sig- nal to a nearby receiver waveguide, which converts the signal back to the optical domain as phase modulation through photoelastic coupling. The emit-receive process was used to create filter functions with Q-factors in excess of 1500 at GHz frequencies with tailorable, ultrasharp fre- quency rolloffs [8]. Because this filtering process is me- diated by GHz-frequency acoustic phonons, rather than light, this narrowband operation is supported without the use of high-Q optical resonators which experience nonlin- ear distortion at low optical powers. To adapt this nascent technology to RF-photonic filtering applications, however, it is necessary to understand the performance capabilities and limitations of such emit-receive processes within prac- tical RF-photonic systems. In this article, we report a new all-silicon photonic- phononic emitter-receiver (PPER) that produces record- high internal efficiency and use this device to demonstrate high-fidelity narrowband filtering within an RF-photonic link. We leverage a dual-core silicon waveguide that guides both light and sound to produce efficient photon-phonon coupling with an interaction strength sufficient to pre- serve or amplify modulation of optical signals through the PPER process. This structure produces a narrowband (5 MHz) filter response with high optical power handling (>100 mW) as a basis for low-loss integrated RF-photonic filters. Through systematic study of this device inside of an RF-photonic link, we demonstrate high-fidelity signal transfer with low insertion loss; a net RF link gain of G 7 1 0 2 v o N 7 1 ] h p - p p a . s c i s y h p [ 1 v 0 5 7 0 0 . 1 0 8 1 : v i X r a = -2.3 dB and spur-free dynamic range SFDR3 = 99 dB Hz−2/3 are measured. We show that this architecture is uniquely suited for operations that require cascaded fil- ter banks with minimal loss of signal fidelity. Building on these results, we discuss the potential for further improve- ments in the performance of RF-photonic PPER systems. RESULTS Operation Scheme The basic operation scheme of the photonic-phononic emitter-receiver system is diagrammed in Fig. 1a. The device consists of two Brillouin-active optical waveguides coupled to a common acoustic phonon mode. In typi- cal operation, an intensity-modulated light signal (Fig. 1b.ii) is incident in port 1 of the device (labeled the 'emit' waveguide). When the modulation frequency of this signal (Ωs) is equal to the Brillouin frequency (ΩB) of the device, the phonon is excited with signal frequency Ωs through a forward-SBS process (Fig. 1b.iii). This acoustic wave then phase-modulates a probe wave incident in port 2 of the device (the 'receive' waveguide) through a linear acousto- optic scattering process (Fig. 1b.iv). Phase-matching per- mits phonon-mediated coupling for this process even when signal and probe waves are disparate in wavelength (>200 nm) [8]. The phase-modulated signal then exits port 4 of the device, where it can be converted back to intensity modulated light through use of a frequency discrimina- tor (e.g. interferometer, filter, phase-shifting element, etc. [3]) and the modulated light can then be used for RF- over-fiber, integrated signal processing, or converted into an RF signal via a photodiode. The frequency response of the PPER is determined by the engineered acoustic re- sponse of the system, which can be tailored for specific applications. Depending on the desired filter bandwidth, the acoustic phonon that mediates the emit-receive pro- cess can be designed to be travelling-wave, resonant, or a super-modal excitation of a collection of acoustic res- onators [8]. In summary, the PPER can be understood as a device that (1) converts intensity modulation to phase modula- tion and (2) acts as an RF filter with a transfer function set by the device's acoustic response. At the same time, this process can be used to implement wavelength con- version due to the large phase-matching bandwidth of the receive-waveguide scattering process. Silicon Waveguide Photonic-Phononic Emitter-Receiver We study the photonic-phononic emit-receive operation using the optomechanical waveguide diagrammed in Fig. 2a. This PPER structure is fabricated from a single- crystal silicon layer using an SOI fabrication process. The device consists of a silicon membrane which guides optical waves in the fundamental optical modes of two separate ridge waveguides (Fig 2d-f). The emit-receive functional- 2 FIG. 1: Operation scheme of the photonic-phononic emitter- receiver. (a) depicts a diagrammatic representation of the PPER process. An optical signal wave (blue) is intensity mod- ulated at microwave frequencies and injected into port 1 of the PPER device. The modulation is encoded on an acous- tic phonon (green) which transduces information onto light in- jected into port 2 as phase modulation. Signal light exits port 3 of the device where it can then be used for further operations. Probe light exits port 4 of the device where it can be converted back into a microwave signal. (b) depicts (i) an example inci- dent broadband RF spectrum (ii) the optical spectrum of the signal wave resulting from intensity modulation (iii) the acous- tic power spectrum resulting from a combination of the original broadband RF modulation and the device's engineered acous- tic transfer/filter function, and (iv) the filtered signal imparted on the probe beam as phase modulation. ity in this structure is mediated by a resonant Lamb-like acoustic mode (Fig 2g) confined to the device by the large acoustic impedance mismatch between silicon and air. In this process, the acoustic wave is driven through a forward- SBS process in the emit waveguide. This optically-driven transverse acoustic wave time-modulates the effective in- dex of the structure, imparting phase modulation on probe light in the receive waveguide. The core widths of these two waveguides are designed to be different by 100 nm. This changes the relative propagation constants of light to inhibit evanescent coupling between the emit and re- ceive waveguides. As a result, the two channels are opti- cally isolated and information is transferred only through phononic transduction. To quantify the performance of this system, we begin by performing basic measurements of the modulation ef- ficiency and frequency-dependent response of the PPER device. In this experiment, two strong optical tones of equal amplitude separated by around the Brillouin fre- quency of the device are injected into the emit waveguide. These tones drive a single-frequency acoustic field which imparts phase modulation on a probe beam incident in the receive waveguide, with a phase shift related to the total incident pump power. In the absence of optical losses, this phase shift is simply φ = GB(Ω)PpL, where GB(Ω) is the frequency-dependent Brillouin gain coefficient, Pp is the power in each drive tone, and L is the device length. This experiment is carried out using the characteriza- tion apparatus diagrammed in Fig. 2h. Two strong pump LaserRFLaserEmitReceive (φM)1234Intensity Mod.OutPPER DeviceFrequencyPowerRFFrequencyPowerOpticalFrequencyPowerAcousticFrequencyPowerOpticalabiiiiiiivPhonon3 FIG. 2: Silicon photonic-phononic emitter-receiver. (a) Artistic representation of device; a dual-ridge structure is suspended by a series of nanoscale tethers. (b) scanning electron micrograph (SEM) of device cross-section. The scale bar represents 1000 nm in length. (c) Top-down SEM of suspended device, with a scale bar representing 20 µm. (d) Diagram of device cross-section with dimensions listed. (e) and (f) are the Ex fields of the two guided optical modes which couple through a ∼4.3 GHz Brillouin-active acoustic mode whose x-displacement field is plotted in (g). (h) depicts an experiment for measuring the PPER response. Pump light is intensity modulated around the zero-bias point and amplified to create two strong pump waves separated by a frequency around the Brillouin resonance frequency. Probe light is passed through the device and its output spectrum characterized by heterodyne detection using a frequency-shifted local oscillator. (i) plots the normalized frequency response of the device as a function of modulation frequency. (j) plots the sideband powers normalized to the incident probe wave power as a function of incident pump power. waves with frequencies ωp±Ω/2 are synthesized by passing light from a pump laser through an intensity modulator operating in the null-bias regime. The amplitude of these waves is controlled using an erbium-doped fiber amplifier and variable attenuator. Pump light is coupled in and out of the emit waveguide of the PPER device using inte- grated grating couplers. A probe wave with frequency ωpr from a different laser is split into two paths; one is coupled on-chip while the other is passed through an acousto-optic modulator to produce a frequency-shifted local oscillator at frequency ωp + ∆ = ωp + 2π × 44 MHz. After probe light passes through the receive waveguide of the PPER device, it is recombined with the local oscillator arm to perform heterodyne spectral analysis for the transmitted spectrum in the microwave domain using a fast photodi- ode and radio-frequency spectrum analyzer. As the pump-wave detuning, and hence modulation fre- quency, is swept through the Brillouin frequency Ω = ΩB, efficient phase modulation of the probe wave is observed, resulting in optical energy transfer to frequency-detuned sidebands at ωpr ± Ω. Fig 2i plots the amplitude of the first phase-modulated sideband at frequency ωpr + Ω as a function of input modulation frequency. These data re- veal a narrowband (Q ∼ 820) resonance corresponding to a Brillouin-active phonon at ΩB/2π = 4.33 GHz. As the total pump power is increased to a maximum of 100 mW, the relative sideband amplitude increases to a maximum of 12%, corresponding to a phase shift of 0.8 rad. These metrics represent a 10−fold improvement in efficiency over previous devices of this type [8]. While at low powers this efficiency follows the expected quadratic dependence (∝ φ2), at high powers it saturates due to a combination of the Bessel-like phase modulation response and nonlinear absorption in the emit waveguide. These data correspond to a Brillouin gain coefficient GB = 820 ± 20 W−1m−1 for this L = 2.53 cm long PPER device. This effective nonlin- ear coefficient results from a combination of the Brillouin couplings in the emit and receive waveguides. In reality, these waveguides have different single-waveguide gain co- efficients due to the engineered asymmetry of the PPER device. The individual gain coefficients for each waveguide were found through standard Brillouin gain measurements [12] to be GB = 880 ± 50 W−1m−1 for the emit waveg- uide and GB = 740 ± 50 W−1m−1 in the receive waveg- uide. These values agree well with simulated values of GB = 860 ± 130 W−1m−1 and GB = 790 ± 120 W−1m−1, respectively. RF-Photonic Link Performance We next show that the performance of this device is sufficient to support efficient filter architectures in an RF- photonic link. First, we characterize the input/output RF xy1.0 μm130 nm80 nm0.9 μm1.05 μmxyxyxymax-maxmax0bcde5.0 μmProbe1545 nm AOMΩRFSA PDDUTPump1550 nmBCIM EDFAΔ+Ω Sideband–Ω SidebandTheoryω4.24.34.4-30-20-100Normalized Signal (dB)Modulation Frequency (GHz)0204060801000.000.050.100.15Normalized Sideband PowerIncident Power (mW)0.00.20.40.60.8Phase (rad)afghij5 MHz @ -3 dB1000 nm20 μmCharacterization Apparatus4 FIG. 3: RF-photonic link performance and noise analysis. (a) Diagram of the RF-photonic link with PPER device. Pump light is intensity modulated about the quadrature point by an input RF signal and amplified before passing into the emit port of the device. Probe light passes through the receive port and is amplified after coupling off-chip to compensate for coupling losses. A fiber Bragg grating is used as a narrowband filter for frequency discrimination, and a photodiode converts the signal back into the RF domain. (b) RF input/output data for the fundamental (top) and third harmonic (bottom) tones of the RF-photonic link. These are used to find the third-prder intercept point OIP3 via extrapolation (dashed grey lines) (c) Characterization of the Brillouin noise floor at Ω/2π = 4.33 GHz as a function of probe power. An inset shows the Brillouin noise spectrum and background due to amplified spontaneous emission from the EDFA. response of this system. We then identify the dominant sources of distortion and noise through this link. to the device's phononic bandwidth of 5 MHz measured in Fig. 2i. We explore RF-photonic performance of the PPER de- vice using the RF-photonic link diagrammed in Fig 3a. An RF signal is encoded on a pump beam using a commer- cial intensity modulator (Optilab IM-1550-20, Vπ = 9.9 V) biased at the quadrature point. Light is amplified using an erbium-doped fiber amplifier and coupled into the emit waveguide using an integrated grating coupler. Probe light from a separate laser is coupled into the receive waveguide, where the signal is filtered through the PPER device's phononic response and encoded onto the probe wave as phase modulation. After passing through the ac- tive device region, this light is coupled off-chip, amplified through an erbium-doped fiber amplifier to offset fiber- chip coupling losses, and converted to intensity modula- tion by filtering out one of the ±Ω modulation sidebands. This signal is then converted back to the RF-domain using a commercial high-power photodiode (Discovery Semicon- ductors, Inc. DSC100S, responsivity R = 0.75 A/W, bias voltage Vb = 7 V). Link gain is measured using a single RF input tone at Ω = ΩB/2π = 4.33 GHz and with incident on-chip optical pump power Pp = 105 mW and probe power at detector Pdet = 75 mW. The corresponding RF signal input/output curve is plotted as the top curve in Fig. 3b. With these specifications, the net RF-photonic link gain is -2.3 dB. The 3 dB bandwidth of the RF filter response is identical Independent of the external RF components, the per- formance of the PPER device can be understood by com- paring the output phase modulation strength to the in- put intensity modulation strength. Through these exper- iments, the effective phase modulation produced through the PPER device is φout PM = 0.559 rad/V, compared to an input intensity modulation of φin IM = 0.317 rad/V. We define a figure of merit which is the ratio between these two modulation strengths which can be applied to any modulation-conversion operations of this type. In this system, φout IM = 1.76, indicating a net enhancement in modulation through the silicon PPER device. PM/φin We next seek to quantify the dominant forms of dis- tortion in the PPER RF-photonic link. The first form of distortion in RF-photonic links is compression of the lin- ear response at high input RF powers. This is quantified through measuring the RF output power at 1 dB com- pression from a linear response. The power output at 1 dB compression P1dB = 7.7 dBm through the PPER filter. The second form of distortion we consider is that due to third-order spurious tones generated through the RF- photonic link. Due to the narrowband filter response of the PPER system, out-of-band frequency components are rejected by the phononic filter. Therefore spurious tones must be either (1) generated at harmonics of ΩB due to phonon-mediated phase modulation in the receive waveg- Probe1550 nmRFDUTPump1545 nmIM EDFA1 EDFA2FBGRFSA PD1020304050607080-145-140-135-130-125Brillouin Noise Floor (dBm/Hz)Probe Power (mW)4.254.304.354.40-130-128-126RF Power (dBm/Hz)Frequency (GHz)BrillouinASE-20020-60-40-20020RF Power Out (dBm)RF Power In (dBm)ExperimentTheoryExtrapolationExperimentTheoryOIP3bcaLink Gain = -2.3 dBSignal3rd-order SpurRF-Photonic Link Experimentuide or (2) unwanted frequency components around ΩB created by the intensity modulator at the frontend of the link. For most systems and device parameters, the latter is the dominant source of spurious tones, as is the case here. Through this process, frequency mixing during modula- tion between input frequencies not within the PPER filter bandwidth can result in new frequency components that can pass through the filter. We quantify the strength of this effect by driving the system at a frequency Ω = ΩB/3 and examining the strength of the third-order spurious signal passing through the PPER at Ω = ΩB, plotted as the bottom curve in Fig. 3b. These spurs correspond to an output third-order intercept point OIP3 of 23 dBm through the PPER system (black circle in Fig. 3b). To characterize the dynamic range of this RF system, we finally carry out careful measurements of the sources of noise through the PPER device. The frequency-dependent noise floor around the Brillouin frequency is plotted as an inset in Fig. 3c. At the filter frequency (ΩB/2π = 4.33 GHz), the noise spectrum is dominated by spontaneous scattering from thermal phonons (population at room tem- perature kBT /¯hΩB ≈ 1390) [34, 35], with an out-of band noise floor interpreted as amplified spontaneous emission from the output EDFA (labeled EDFA2 in Fig. 3a) in the link. The probe power dependence of the Brillouin noise is quantified in Fig 3c. For detector powers of 75 mW used in this measurement, the Brillouin noise power spectral density N = -126 dBm/Hz. This noise power corresponds to a receive waveguide Brillouin gain coeffi- cient of GB = 730 ± 40 W−1m−1, in good agreement with the numbers obtained from stimulated gain measurements. This fundamental noise source can be mitigated by design- ing PPER devices which couple through higher-frequency phonons with lower thermal occupations at room temper- ature. Combining the measurements of total system nonlin- earities and noise, we find a spur-free dynamic range SFDR3 = 2(OIP3 - N )/3 = 99.3 dB Hz−2/3 normalized to a 1 MHz bandwidth. This corresponds to the dynamic range over which a signal is detectable through the link while third-order spurs are not. The linear dynamic range is CDR1dB = P1dB − N = 135 dB Hz−1. In the next sec- tion we discuss use cases for the narrowband PPER filter as well as potential improvements to device performance. DISCUSSION We have characterized the performance of the integrated PPER as both an active optical device and as a narrow- band RF-photonic filter. This system supports a record- high nonlinear figure of merit for emit-receive systems of this type in an all-silicon device structure. Next, we dis- cuss the potential for further improvement of the system as well as architectures for scalability and integration. We have shown that new photonic-phononic emit- receive operations provide a promising path towards high- performance narrowband RF-photonic filters. Looking forward, several potential refinements of the PPER link and device are possible to produce higher performance in 5 a significantly smaller device footprint. First, the perfor- mance of the PPER filter system can be readily enhanced by improving the modulation and detection schemes exter- nal to the filter. The intensity modulator used in the fron- tend of the link demonstrated here has a relatively high half-wave voltage Vπ = 9.9 V. Decreasing Vπ would allow the reduction of device length by a corresponding factor without affecting link gain or dynamic range. In addi- tion to shrinking the device footprint, this reduced length would lead to a lower spontaneous noise floor, allowing the PPER system to operate on smaller RF signals as may be desirable in practical applications. Many modulator tech- nologies have been already demonstrated in silicon with Vπ < 1 V [36, 37], demonstrating the feasibility of this ap- proach. In addition, the current method used to convert phase to intensity modulation is sub-optimal since filter- ing out one of the two signal sidebands discards half of the signal power. The link gain of the system could therefore be improved by replacing this filter with interferometric balanced detection or a phase-shifting element. Significant improvements in device performance can also be achieved by increasing optical power in the emit waveg- uide, which directly enhances the modulation efficiency of the PPER device. The pump power in the current device was limited to ∼ 100 mW due to nonlinear absorption [12]. This can be mitigated through electrical removal of free carriers generated due to two-photon absorption (TPA) [38] or by changing the pump wavelength to be above the TPA threshold for silicon (λp > 2.2 µm). These would allow pump powers >2 W to propagate through the emit waveguide with negligible nonlinear losses, allowing dramatic reduction in device length and probe power to enable lower-noise operation with 5-10 dB improvements in dynamic range. Shorter devices are also known to have a smaller degree of inhomogeneous broadening of the Bril- louin resonance [39]. Therefore shorter device lengths would decrease the filter width and increase device per- formance, which depends directly on acoustic quality fac- tor. A modified silicon PPER system with the parameters (Pp = 2 W, λp = 2.75 µm, Ppr = 10 mW, λpr = 1.55 µm, Vπ = 1 V, Q = 1500, L = 200 µm) would have superior linearity (link gain G = 0 dB, SFDR3 = 105 dB Hz−2/3, CDR1dB = 146 dB Hz−1) in a much smaller footprint. The dynamic range of this system could be further improved by using a linearized modulation scheme, which have been studied extensively for RF-photonic applications [40–42]. Unlike many conventional RF filters, the PPER filter allows for serial cascading without significant signal degra- dation, making it particularly attractive for signal chan- nelization. Once an RF signal is encoded on the pump wave, it can pass through several successive PPER devices to select out separate spectral channels without impact- ing the fidelity of the original signal, as diagrammed in Fig 4a. By contrast, traditional channelization techniques based on band-pass filters require power to be split evenly between each channel (Fig 4b), resulting in noise increases due to prior amplification stages that grow with the num- ber of channels. As an example application, a cascaded array of 200 PPER devices with the improved parameters discussed above could span a 500 MHz bandwidth with 2.5 nects with applications in communications, sensing, and signal analysis. 6 METHODS Device Fabrication and Optical Performance The silicon waveguides were written on a silicon-on- insulator chip with a 3 µm oxide layer using electron beam lithography on hydrogen silsesquioxane photoresist. After development, a Cl2 reactive ion etch (RIE) was employed to etch the ridge waveguides and integrated grating cou- plers. After a solvent cleaning step, slots were written to expose the oxide layer with electron beam lithography of ZEP520A photoresist and Cl2 RIE. The device was then wet-released in 49% hydrofluoric acid to remove the oxide undercladding. The device under test is comprised of 498 suspended segments. These waveguides have an estimated linear optical prop- agation loss of 0.2 dB/cm from cutback measurements. The nonlinear absorption coefficients in units of guided- wave power are βTPA ≈ 50 W−1m−1 for two-photon absorption and γFCA ≈ 2400 W−2m−1 for two-photon absorption-induced free-carrier absorption as measured through nonlinear power transmssion experiments. Experiment The following abbreviations are used in the experimen- tal diagrams: IM Mach-Zehnder intensity modulator, BC bias controller, EDFA erbium-doped fiber amplifier, DUT device under test, AOM acousto-optic frequency shifter, FBG fiber Bragg grating, PD photodetector, RFSA radio- frequency spectrum analyzer. [1] Seeds, A.J. & Williams, K.J. Microwave photonics. J. Lightwave Technol. 24, 4628–4641 (2006). [2] Marpaung, D. et al. Integrated microwave photonics. Laser Photon. Rev. 7, 506–538 (2013). [3] Urick, V., Williams, K. & McKinney, J. Fundamentals of Microwave Photonics. Wiley Series in Microwave and Optical Engineering (Wiley, 2015). [4] Capmany, J. & Novak, D. Microwave photonics combines two worlds. Nat Photon 1, 319–330 (2007). [5] Capmany, J., Ortega, B., Pastor, D. & Sales, S. Discrete- time optical processing of microwave signals. Journal of Lightwave Technology 23, 702–723 (2005). [6] Marpaung, D. et al. Low-power, chip-based stimulated Brillouin scattering microwave photonic filter with ultra- high selectivity. Optica 2, 76–83 (2015). [7] Pant, R. et al. On-chip stimulated Brillouin scattering for microwave signal processing and generation. Laser Photon. Rev. 8, 653–666 (2014). [8] Shin, H. et al. Control of coherent information via on- chip photonic-phononic emitter-receivers. Nat. Commun. 6, 6427 (2015). [9] Casas-Bedoya, A., Morrison, B., Pagani, M., Marpaung, D. & Eggleton, B.J. Tunable narrowband microwave pho- FIG. 4: Comparison of serial and parallel filter architectures for channelization. (a) Serial cascaded architecture for a PPER channelization filter. Because the same pump light can be used for each device (i.e. because the PPER device makes a copy of a narrowband slice of the input RF spectrum), minimal sig- nal degradation is experienced as the number of channels is increased. (b) Architecture based on parallel RF filters. Note that because the signal must be split equally among the chan- nels, amplification is necessary to preserve signal amplitude. MHz bandwidth per filter channel, while maintaining the same dynamic range and sensitivity as a basis for real-time spectral analysis. Emit-receive operations can be realized in many sys- tems where optical fields guided in distinct modes or po- larizations couple through the same phonon mode, pro- vided that linear and nonlinear optical crosstalk between channels is negligible. Current and potential avenues of research include multi-core fibers [43], optomechani- cal waveguides coupled by acoustic supermodes [8], and light fields of disparate wavelengths [30, 31] or spatial modes [13] within the same Brillouin-active waveguide core. Building on the present work, PPER interactions may be harnessed in a variety of systems for applications in optical and RF signal processing and spectral analysis. The integrated silicon-based PPER device demon- strated here can be readily interfaced with existing on- chip technologies for silicon-based modulators [44, 45] and integrated photodetectors [46–48]. Such on-chip integra- tion would allow low-cost implementation of such filters in complex RF-photonic circuits. Like other acousto-optic based RF-photonic filters, the PPER can also be used for frequency-agile operation; while the PPER device imple- mented in the RF-photonic link here operates at a fixed frequency, it is straightforward to adapt Brillouin-based RF-photonic filter systems to frequency agile operation by changing the modulation scheme [2, 6, 8, 9]. In conclusion, we have demonstrated an all-silicon hy- brid photonic-phononic filter and characterized its per- formance within an RF-photonic link. This device sup- ports record-high modulation efficiency for phononic emit- receive devices and robust performance as an RF-photonic filter. This work demonstrates the potential for large-scale integration of these agile filters into RF-photonic intercon- Disc./PDPPER 1Ω=Ω1LaserDisc./PDPPER 2Ω=Ω2LaserRFLaserMod....RFRF Amp.BP FilterBP FilterBP FilterBP FilterBP FilterBP FilterBP FilterBP Filterbatonic filter created by stimulated Brillouin scattering from a silicon nanowire. Opt. Lett. 40, 4154–4157 (2015). [10] Olsson, N. & van der Ziel, J. Fibre Brillouin amplifier with electronically controlled bandwidth. Electron. Lett. 22, 488–490 (1986). [11] Pant, R. et al. On-chip stimulated Brillouin scattering. Opt. Express 19, 8285–8290 (2011). [12] Kittlaus, E.A., Shin, H. & Rakich, P.T. Large Brillouin amplification in silicon. Nat. Photon. 10, 463–467 (2016). [13] Kittlaus, E.A., Otterstrom, N.T. & Rakich, P.T. On-chip inter-modal Brillouin scattering. Nat. Commun. 8, 15819 (2017). Article. [14] Morrison, B. et al. Compact Brillouin devices through hybrid integration on silicon. Optica 4, 847–854 (2017). [15] Tanemura, T., Takushima, Y. & Kikuchi, K. Narrowband optical filter, with a variable transmission spectrum, using stimulated Brillouin scattering in optical fiber. Opt. Lett. 27, 1552–1554 (2002). [16] Zadok, A., Eyal, A. & Tur, M. Gigahertz-wide optically reconfigurable filters using stimulated Brillouin scattering. J. Lightwave Technol. 25, 2168–2174 (2007). [17] Vidal, B., Piqueras, M.A. & Mart´ı, J. Tunable and recon- figurable photonic microwave filter based on stimulated Brillouin scattering. Opt. Lett. 32, 23–25 (2007). [18] Zhang, W. & Minasian, R. Ultrawide tunable microwave photonic notch filter based on stimulated Brillouin scatter- ing. IEEE Photon. Technol. Lett. 24, 1182–1184 (2012). [19] Yao, X. Brillouin selective sideband amplification of mi- crowave photonic signals. IEEE Photon. Technol. Lett. 10, 138–140 (1998). [20] Loayssa, A., Benito, D. & Garde, M.J. Optical carrier Brillouin processing of microwave photonic signals. Opt. Lett. 25, 1234–1236 (2000). [21] Li, J., Lee, H. & Vahala, K.J. Microwave synthesizer using an on-chip Brillouin oscillator. Nat. Commun. 4, 2097 (2013). [22] Merklein, M. et al. Widely tunable, low phase noise mi- crowave source based on a photonic chip. Opt. Lett. 41, 4633–4636 (2016). [23] Huang, X. & Fan, S. Complete all-optical silica fiber isola- tor via stimulated Brillouin scattering. J. Lightwave Tech- nol. 29, 2267–2275 (2011). [24] Kang, M.S., Butsch, A. & Russell, P.St.J. Reconfigurable light-driven opto-acoustic isolators in photonic crystal fi- bre. Nat. Photon. 5, 549–553 (2011). [25] Kim, J., Kuzyk, M.C., Han, K., Wang, H. & Bahl, G. Non-reciprocal Brillouin scattering induced transparency. Nat. Phys. 11, 275–280 (2015). [26] Dong, C. et al. Brillouin-scattering-induced transparency and non-reciprocal light storage. Nat. Commun. 6, 6193 (2015). Article. [27] Okawachi, Y. et al. Tunable all-optical delays via Brillouin slow light in an optical fiber. Phys. Rev. Lett. 94, 153902 (2005). [28] Zhu, Z., Gauthier, D.J. & Boyd, R.W. Stored light in an optical fiber via stimulated Brillouin scattering. Science 318, 1748–1750 (2007). [29] Rakich, P.T., Reinke, C., Camacho, R., Davids, P. & Wang, Z. Giant enhancement of stimulated Brillouin scat- tering in the subwavelength limit. Phys. Rev. X 2, 011008 (2012). 7 [30] Shin, H. et al. Tailorable stimulated Brillouin scattering in nanoscale silicon waveguides. Nat. Commun. 4, 1944 (2013). [31] Van Laer, R., Kuyken, B., Van Thourhout, D. & Baets, R. Interaction between light and highly confined hypersound in a silicon photonic nanowire. Nat. Photon. 9, 199–203 (2015). [32] Van Laer, R., Bazin, A., Kuyken, B., Baets, R. & Thourhout, D.V. Net on-chip Brillouin gain based on sus- pended silicon nanowires. New J. Phys. 17, 115005 (2015). [33] Otterstrom, N.T., Behunin, R.O., Kittlaus, E.A., Wang, Z. & Rakich, P.T. A silicon Brillouin laser. arXiv preprint arXiv:1510.00079 (2017). [34] Kharel, P., Behunin, R.O., Renninger, W.H. & Rakich, P.T. Noise and dynamics in forward Brillouin interactions. Phys. Rev. A 93, 063806 (2016). [35] Van Laer, R., Sarabalis, C.J., Baets, R., Van Thourhout, D. & Safavi-Naeini, A.H. Thermal Brillouin noise observed in silicon optomechanical waveguide. Journal of Optics 19, 044002 (2017). [36] Van Campenhout, J., Green, W.M.J., Assefa, S. & Vlasov, Y.A. Low-power, 2x2 silicon electro-optic switch with 110- nm bandwidth for broadband reconfigurable optical net- works. Opt. Express 17, 24020–24029 (2009). [37] Van Camp, M.A. et al. Demonstration of electrooptic modulation at 2165nm using a silicon Mach-Zehnder in- terferometer. Opt. Express 20, 28009–28016 (2012). [38] Rong, H. et al. A continuous-wave Raman silicon laser. Nature 433, 725–728 (2005). [39] Wolff, C., Van Laer, R., Steel, M.J., Eggleton, B.J. & Poul- ton, C.G. Brillouin resonance broadening due to struc- tural variations in nanoscale waveguides. New Journal of Physics 18, 025006 (2016). [40] Johnson, L.M. & Roussell, H.V. Reduction of intermod- ulation distortion in interferometric optical modulators. Opt. Lett. 13, 928–930 (1988). [41] Betts, G.E. Linearized modulator for suboctave-bandpass IEEE Transactions on Microwave optical analog links. Theory and Techniques 42, 2642–2649 (1994). [42] Bridges, W.B. & Schaffner, J.H. Distortion in lin- earized electrooptic modulators. IEEE Transactions on Microwave Theory and Techniques 43, 2184–2197 (1995). Opto- mechanical inter-core cross-talk in multi-core fibers. Op- tica 4, 289–297 (2017). [43] Diamandi, H.H., London, Y. & Zadok, A. [44] Xu, Q., Schmidt, B., Pradhan, S. & Lipson, M. Micrometre-scale silicon electro-optic modulator. Nature 435, 325–327 (2005). [45] Reed, G.T., Mashanovich, G., Gardes, F.Y. & Thomson, D.J. Silicon optical modulators. Nat Photon 4, 518–526 (2010). [46] Chen, L., Preston, K., Manipatruni, S. & Lipson, M. Integrated GHz silicon photonic interconnect with micrometer-scale modulators and detectors. Opt. Express 17, 15248–15256 (2009). [47] Baehr-Jones, T. et al. A 25 Gb/s silicon photonics plat- form (2012). arXiv:1203.0767. [48] Xie, X. et al. High-power and high-speed heterogeneously integrated waveguide-coupled photodiodes on silicon-on- insulator. J. Lightwave Technol. 34, 73–78 (2016).
1711.09938	1	1711	2017-11-27T19:22:04	Infeasibility of the nonlocal strain gradient theory for applied Physics	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	In this communication, the feasibility of the nonlocal strain gradient theory for fields of applied mechanics is investigated. It is demonstrated that the nonlocal strain gradient theory is physically incorrect. It is proved that each of the nonlocal theory and the strain gradient theory can model, both, hardening and softening behaviors of materials. Moreover, it is proved that the nonlocal theory and the strain gradient theory describe the same physical phenomena, and hence the strain gradient theory cannot be merged with the nonlocal theory in a unified model. This short communication comments on a series of papers in which the nonlocal strain gradient theory was utilized in different areas of applied mechanics.	physics.app-ph	physics	Infeasibility of the nonlocal strain gradient theory for applied Physics Mohamed Shaat* Engineering and Manufacturing Technologies Department, DACC, New Mexico State University, Las Cruces, NM 88003, USA Mechanical Engineering Department, Zagazig University, Zagazig 44511, Egypt Abstract In this communication, the feasibility of the nonlocal strain gradient theory for fields of applied mechanics is investigated. It is demonstrated that the nonlocal strain gradient theory is physically incorrect. It is proved that each of the nonlocal theory and the strain gradient theory can model, both, hardening and softening behaviors of materials. Moreover, it is proved that the nonlocal theory and the strain gradient theory describe the same physical phenomena, and hence the strain gradient theory cannot be merged with the nonlocal theory in a unified model. This short communication comments on a series of papers in which the nonlocal strain gradient theory was utilized in different areas of applied mechanics. Keywords: nonlocal theory; strain gradient theory; elasticity; wave propagation; dispersion curves. 1. Introduction Eringen's nonlocal theory has been proposed with the flavor of accounting for the non-neighbor interactions between the building units of the material microstructure [Eringen, 1972a; 1972b; Eringen and Edelen, 1972; Eringen and Kim, 1974; Eringen et al., 1977; Eringen, 1978, 1979]. Eringen's nonlocal theory of linear elasticity emulates the classical theory in considering the infinitesimal strain as the fundamental measure of deformation. However, Eringen's model incorporates an integral operator to sum the nonlocal residuals inside the material; this form is known as integral nonlocal theory [Eringen and Kim, 1974; Eringen et al., 1977; Eringen, 1978, 1979]. In 1983, the integral operator was replaced by a differential operator introducing the differential nonlocal theory [Eringen, 1983]. Eringen's two forms of the nonlocal theory have been used in various studies. Integral nonlocal theory was used to investigate the dispersions of plane waves [Eringen, 1992], stress concentrations near a crack tip [Eringen and Kim, 1974], plasticity and damage of materials [Bazant and Jirasek, 2002], softening plasticity [Jirasek and Rolshoven, 2003], and mechanics of nanostructures [Pisano and Fuschi, 2003; Norouzzadeh and Ansari, 2017; Fernández-Sáez et al., 2016; Tuna and Kirca, 2016a; Tuna and Kirca, 2016b]. Uncountable list of studies utilized the differential nonlocal theory in investigating the mechanics ------------------------------------------------------------------------------------------------ ∗Corresponding author. Tel.: +15756215929. E-mail addresses: [email protected]; [email protected] (M. Shaat). of nanostructures [e.g. Lu et al., 2006; Reddy, 2007; Shaat, 2015; Lu et al., 2007; Zhang et al., 2017; Liew et al., 2017]. Another group of nonclassical theories is the strain gradient and couple stress theories. Mindlin [Mindlin 1964, 1965; Mindlin and Eshel, 1968] has proposed the strain gradient theory for linear elastic continua in which the strain energy depends on the infinitesimal strain as well as its gradients. Moreover, the couple stress theory [Mindlin and Tierestan, 1962; Toupin, 1962] has been developed such that the internal energy depends on the strain and the first-order rotation gradient. An extension of the couple stress theory by developing the second rotation gradient theory was proposed by incorporating the 2nd-order rotation gradient [Shaat and Abdelkefi, 2016]. In fact, the field of applied nonlocal mechanics suffers from misleading information. For example, it is thought that nonlocal models only account for softening behaviors of materials. The strain gradient and the couple stress theory, on the contrary, only account for hardening behaviors of materials. Thus, it is thought that the nonlocal parameter of the nonlocal theory and the length scales of the strain gradient and couple stress theories describe different material characteristics [e.g. Lim et al., 2015]. This motivated some researchers [e.g. Lim et al., 2015; Attia and Mahmoud, 2016] to combine the nonlocal theory with the strain gradient and/or the couple stress theory in a unified model. It was thought that these unified models give the flavor of combining the softening and hardening mechanisms. However, these attempts should be revised for their feasibilities. It was thought that Eringen's nonlocal model of linear elasticity can be generalized by the attempt proposed by Lim et al. [2015] by developing a nonlocal strain gradient theory in which the strain energy function depends on two independent integral operators for the infinitesimal strain and its first-order gradient. In fact, this attempt needs to be revised for its feasibility. Lim, Zhang and Reddy [2015] claimed, "nonlocal elastic models can only account for softening stiffness with [an] increasing nanoscale parameter." Then, they added, "the stiffness enhancement effect, noticed from experimental observation[s] and as well as the gradient elasticity (or modified couple stress) theories cannot be characterized," by the nonlocal theory, they meant. Moreover, they mentioned, "Many related research works have been performed to model the static and dynamic problems based on various gradient theories and all the results showed a stiffness enhancement effect with an increase of the gradient coefficients." Accordingly, they concluded, "the length scales present in Erinegn's nonlocal elasticity model and the strain gradient theories describe two entirely different physical characteristics of materials and structures at nanoscale." Therefore, they literally mentioned, as the main motivation behind their proposed theory, "There is a need to bring both of the length scales into a single theory so that the true effect of the two length scales on the structural response can be assessed." It follows from these quotations that the nonlocal strain gradient theory was proposed with the aim of developing a unified model that can account for the softening and hardening behaviors of materials. However, in fact, the hardening and the softening behaviors of materials can be captured by either the nonlocal theory or the strain gradient theory and no need to merge them in a single model. In this study, a proof for this fact is presented. In this communication, the fact that each of the nonlocal elasticity theory and the strain gradient theory can reflect, both, softening and hardening behaviors of materials is demonstrated. Moreover, it is revealed that these two theories describe the same physical phenomena and cannot be merged in a unified theory. This argues against the claims by Lim et al. [2015] and Attia and Mahmoud [2016] and demolishes the concept of the nonlocal strain gradient. 2 2. The nonlocal theory models softening and hardening behaviors Referring to Eringen's book, "Nonlocal continuum field theories," [Eringen, 2002, p. 99], the follows: nondimensional wave frequency, (cid:2033)(cid:3365)/(cid:1855)(cid:2868), can be written in the context of the nonlocal elasticity theory as (cid:2033)(cid:3365)(cid:1855)(cid:2868)=(cid:1863)(cid:3364)(cid:3496) 1 1+(cid:2010)(cid:1863)(cid:3364)(cid:2870)+(cid:2012)(cid:1863)(cid:3364)(cid:2872) where (cid:1863)(cid:3364) denotes the nondimensional wave number. (cid:2010) and (cid:2012) are two nondimensional nonlocal parameters. When (cid:2012)=0 and (cid:2010)=(cid:2028)(cid:2870), the dispersion relation (1) recovers the one presented in Lim et al. [2015] for of materials, (cid:2010) and (cid:2012) may possess negative or positive values. However, researchers have confined the use of these nonlocal parameters to (cid:2012)=0 and (cid:2010)=(cid:2028)(cid:2870)>0 . In fact, utilizing these two parameters and behavior (the decrease in the nondimensional wave frequency) when (cid:2010)>0 and (cid:2012)=0, (cid:2010)≥0 and (cid:2012)>0, or (cid:2012)+(cid:2010)≥0 and (cid:2012)<0. On the other hand, the nonlocal theory depicts the hardening behavior (the increase in the nondimensional wave frequency) when (cid:2010)<0 and (cid:2012)=0, (cid:2012)+(cid:2010)≤0 and (cid:2012)>0, or (cid:2010)≤0 and (cid:2012)<0. Indeed, Figure 1 demonstrates that the nonlocal theory can give all the possible configurations considering their possible negative and positive values, the nonlocal theory can reflect, both, hardening and softening behaviors of materials, as shown in Figure 1. The nonlocal theory can reflect the softening Eringen's nonlocal theory. As stated by Eringen [2002, p. 99], for better fitting with the dispersion curves (1) of the dispersion curves. This contradicts with the claims that the nonlocal theory can only reflect softening behaviors of materials and nanostructures. Figure 1: The nonlocal theory can reflect softening and hardening behaviors of materials. This theory depicts all the possible configurations of dispersion curves. 3 strain gradient theory (up to the second-order strain gradient) in the form: (2) Figure 2: The strain gradient theory can reflect softening and hardening behaviors of materials. This theory depicts all the possible configurations of dispersion curves. 3. The strain gradient theory models softening and hardening behaviors recovers the one presented in Lim et al. [2015] for the strain gradient theory. It should be mentioned that Mindlin has clearly stated that the length scales may possess real or complex Referring to Mindlin [1965] and Shaat [2017], the dispersion relation can be written according to the (cid:2033)(cid:3365)(cid:1855)(cid:2868)=(cid:1863)(cid:3364)(cid:3495)1+(cid:1864)(cid:2869)(cid:2870)(cid:1863)(cid:3364)(cid:2870)+(cid:1864)(cid:2870)(cid:2872)(cid:1863)(cid:3364)(cid:2872) where (cid:1864)(cid:2869) and (cid:1864)(cid:2870) are two nondimensional length scales. When (cid:1864)(cid:2870)=0 and (cid:1864)(cid:2869)=(cid:2022), the dispersion relation (2) values (i.e. (cid:1864)(cid:2869)(cid:2870),(cid:1864)(cid:2870)(cid:2872) ≥0 or (cid:1864)(cid:2869)(cid:2870),(cid:1864)(cid:2870)(cid:2872)≤0) [Mindlin, 1965, p. 424; Mindlin and Tiersten, 1962, p. 416]. However, reflects a softening behavior when (cid:1864)(cid:2869)(cid:2870)<0 and (cid:1864)(cid:2870)(cid:2872)=0, (cid:1864)(cid:2869)(cid:2870)+(cid:1864)(cid:2870)(cid:2872)≤0 and (cid:1864)(cid:2870)(cid:2872)>0, or (cid:1864)(cid:2869)(cid:2870)≤0 and (cid:1864)(cid:2870)(cid:2872)<0. However, the strain gradient theory gives a hardening behavior when (cid:1864)(cid:2869)(cid:2870)>0 and (cid:1864)(cid:2870)(cid:2872)=0, (cid:1864)(cid:2869)(cid:2870)≥0 and (cid:1864)(cid:2870)(cid:2872)> 0, or (cid:1864)(cid:2869)(cid:2870)+(cid:1864)(cid:2870)(cid:2872)≥0 and (cid:1864)(cid:2870)(cid:2872)<0. As depicted in Figure 2, the strain gradient theory can give all the possible the use of these length scales in the literature was only confined for the real values. Shaat [2017] demonstrated that the real and complex forms of these length scales are needed in order to give the strain gradient theory the ability to model the softening and hardening behaviors of materials and nanostructures. As shown in Figure 2, utilizing all the possible forms of the length scales, the strain gradient theory can reflect both hardening and softening behaviors of materials and nanostructures. The strain gradient theory configurations of the dispersion curves. This, as well, argues against the claims that the strain gradient theory can only give the hardening behaviors of materials and nanostructures. 4 4. The strain gradient theory is a nonlocal theory To demonstrate the fact that the nonlocal theory and the strain gradient theory describe the same physical phenomena, different dispersion curves as obtained by these two theories are plotted in Figure 3. As shown in the figure, the strain gradient theory can exactly fit the nonlocal theory when the length scales, (cid:1864)(cid:2869) and (cid:1864)(cid:2870), are properly related to the nonlocal parameters, (cid:2010) and (cid:2012), [Shaat, 2017]. Indeed, the two theories can reflect the dispersive nature of the acoustic dispersion curves of materials, as shown in Figure 3. This demonstrates the fact that the strain gradient theory is a nonlocal theory where it accounts for the non-neighbor interactions exactly the same as the nonlocal theory [Shaat, 2017]. Hence, the strain gradient theory cannot be merged with the nonlocal theory in a unified model. Figure 3: The nonlocal theory and the strain gradient theory describe the same physical phenomena. These two theories give the same dispersion curves. (Dispersions as obtained by the nonlocal theory are presented by dark solid curves while the dispersions as obtained by the strain gradient theory are depicted by open circles). 5. Conclusions It follows from the derived results and the presented discussions that the nonlocal strain gradient theory [Lim et al., 2015] and the nonlocal couple stress theory [Attia and Mahmoud, 2016] are physically incorrect. It was demonstrated that the strain gradient theory can present the same phenomena exactly the same as the nonlocal theory. Moreover, both theories can reflect the softening and hardening behaviors of materials. The nonlocal strain gradient theory has been utilized in various studies including [e.g. Li et al., 2015; 2016a; 2016b; Li and Hu, 2015; Ebrahimi and Barati, 2016; 2017; Farajpour et al., 2016]. Although of the wide spread of this theory among researchers, it is not physically possible to combine the nonlocal theory with the strain gradient theory nor the couple stress theory in unified models. Therefore, these studies should be revised for their feasibility. 5 References Attia, M.A., Mahmoud, F.F., 2016. Modeling and analysis of nanobeams based on nonlocal couple-stress elasticity and surface energy theories. International Journal of Mechanical Sciences 105, 126–134. Bazant, Z.P., Jirasek, M., 2002. Nonlocal Integral Formulations of Plasticity and Damage: Survey of Progress. Journal of Engineering Mechanics, 128(11), 1119-1149. Ebrahimi, F. Barati, M. R., 2016. Wave propagation analysis of quasi-3D FG nanobeams in thermal environment based on nonlocal strain gradient theory. Applied Physics A 122, 843. Ebrahimi, F., Barati, M., 2017. Hygrothermal effects on vibration characteristics of viscoelastic FG nanobeams based on nonlocal strain gradient theory. Composite Structures 159, 433-444. Eringen, A.C., 1972a. Nonlocal polar elastic continua. Int. J. Eng. Sci., 10, 1–16. Eringen, A.C., 1972b. Linear theory of nonlocal elasticity and dispersion of plane waves. Int. J. Eng. Sci., 10, 425–435. Eringen, A.C., Edelen, D.G.B., 1972. On nonlocal elasticity. Int J Eng Sci, 10(3), 233–48. Eringen, A.C., Kim, B.S., 1974. Stress concentration at the tip of a crack. Mechanics Research Eringen, A.C., Speziale, C.G., Kim, B.S., 1977. Crack-tip problem in non-local elasticity. Journal of Communications, 1, 233–237. Mechanics Physics and Solids, 25, 339–355. Eringen, A.C., 1978. Line crack subjected to shear. International Journal of Fracture, 14, 367–379. Eringen, A.C., 1979. Line crack subjected to anti-plane shear. Engineering Fracture Mechanics, 12, 211– 219. Eringen, A. C., 1983. On differential equations of nonlocal elasticity and solutions of screw dislocation and surface waves. J. Appl. Phys., 54 (9), 4703-4710. Eringen, A.C., 1992. Vistas of nonlocal continuum physics. Int J Eng Sci, 30(10), 1551–65. Eringen AC. Nonlocal continuum field theories. NewYork: Springer-Verlag; 2002. Fernández-Sáez, J. , Zaera, R. , Loya, J. A. , Reddy, J. N., 2016. Bending of Euler–Bernoulli beams using Eringen's integral formulation: A paradox resolved. International Journal of Engineering Science, 99, 107–116. Farajpour, A, Haeri Yazdi, M. R., Rastgoo, A., Mohammadi, M., 2016. A higher-order nonlocal strain gradient plate model for buckling of orthotropic nanoplates in thermal environment. Acta Mechanica 227(7), 1849–1867. Jirasek, M., Rolshoven, S., 2003. Comparison of integral-type nonlocal plasticity models for strain- softening materials. International Journal of Engineering Science, 41, 1553–1602. Lu, P., Lee, H.P., Lu, C., 2006. Dynamic properties of flexural beams using a nonlocal elasticity model. J Appl Phys, 99, 073510. Li, L., Hu, Y., Ling, L., 2015. Flexural wave propagation in small-scaled functionally graded beams via a nonlocal strain gradient theory. Composite Structures 133, 1079–1092. Li, L., Li, X., Hu, Y. 2016a. Free vibration analysis of nonlocal strain gradient beams made of functionally graded material. International Journal of Engineering Science 102, 77–92. Li, L., Hu, Y., 2015. Buckling analysis of size-dependent nonlinear beams based on a nonlocal strain gradient theory. International Journal of Engineering Science 97 (2015) 84–94. Li, L., Hu, Y., Ling, L., 2016b. Wave propagation in viscoelastic single-walled carbon nanotubes with surface effect under magnetic field based on nonlocal strain gradient theory. Physica E 75, 118–124. 6 Liew, K.M., Zhang, Y., Zhang, L. W., 2017. Nonlocal elasticity theory for graphene modeling and simulation: prospects and challenges. Journal of Modeling in Mechanics and Materials, 1(1), doi:10.1515/jmmm-2016-0159. Lim, C.W., Zhang, G., Reddy, J.N., 2015. A Higher-order nonlocal elasticity and strain gradient theory and its Applications in wave propagation. Journal of the Mechanics and Physics of Solids. 78, 298–313. Lu, P., Zhang, P.Q., Lee, H.P., Wang, C.M., Reddy, J.N., 2007. Non-local elastic plate theories. Proc R Soc A, 463, 3225–40. Mindlin, R.D., 1964. Microstructure in linear elasticity, Arch. Rational Mech. Anal. 16, 51–78. Mindlin, R.D., 1965. Second gradient of strain and surface-tension in linear elasticity. International Journal of Solids and Structure 1, 417–438. Mindlin, R. D. and Tiersten, H. F., 1962. Effects of couple stresses in linear elasticity, Arch. Mindlin, R.D., Eshel, N.N., 1968. On first strain-gradient theories in linear elasticity, Int. J. Solids Rational Mech. Anal. 11, 415–448. Struct. 4, 109-124. Norouzzadeh, A., Ansari, R., 2017. Finite element analysis of nano-scale Timoshenko beams using the integral model of nonlocal elasticity. Physica E, 88, 194–200. Pisano, A.A., Fuschi, P., 2003. Closed form solution for a nonlocal elastic bar in tension. Int J Solids Struct, 41, 13–23. 307. Reddy, J.N., 2007. Nonlocal theories for bending, buckling and vibration of beams. Int J Eng Sci, 45, 288– Shaat, M., 2015. Iterative nonlocal elasticity for Kirchhoff plates. Int J Mech Sci, 90, 162–70. Shaat, M. and Abdelkefi, A., 2016. On a second-order rotation gradient theory for linear elastic continua, International Journal of Engineering Science 100, 74–98. Shaat M., 2017. A general nonlocal theory and its approximations for slowly varying acoustic waves. Int J Mech Sci 130, 52–63. Tuna, M., Kirca, M., 2016a. Exact solution of Eringen's nonlocal integral model for bending of Euler- Bernoulli and Timoshenko beams. International Journal of Engineering Science 105, 80–92. Tuna, M., Kirca, M., 2016b. Exact solution of Eringen's nonlocal integral model for vibration and buckling of Euler-Bernoulli beam. International Journal of Engineering Science 107, 54-67. Toupin, R.A., 1962. Elastic materials with couple-stresses, Arch. Rational Mech. Anal. 11, 385– 414. Zhang, L.W., Zhang, Y., Liew, K.M., 2017. Vibration analysis of quadrilateral graphene sheets subjected to an inplane magnetic field based on nonlocal elasticity theory. Composites Part B, 118, 96-103. 7
1907.03062	2	1907	2019-07-09T07:26:49	Space-time Variant Self-growing Bandgap in Nonlinear Acoustic Metamaterial	[ "physics.app-ph" ]	Material band structure is key foundation for various modern technologies, but it was regarded as a space-time invariant feature. Acoustic metamaterials show extraordinary properties for processing elastic waves, but conventional realizations suffer from narrow bandgaps. Here we first report a nonlinear acoustic metamaterial whose band structure self-adapts to the propagation distance/time and the bandgap exhibits a self-growing behaviour stemming from giant nonlinear interaction. This space-time self-modulating characteristic highlights an unconventional understanding of the band structure, and the self-growth generates an ultralow and ultrabroad bandgap that breaks through the limitation of the mass law for linear locally resonant bandgaps. We also elucidate the self-adaptive mechanisms. This first demonstration sheds light on conceiving advanced devices and metamaterials with broadband, space-time variant bandgaps for wave self-manipulation.	physics.app-ph	physics	Space-time Variant Self-growing Bandgap in Nonlinear Acoustic Metamaterial Xin Fang, Jihong Wen†, Dianlong Yu Laboratory of Science and Technology on Integrated Logistics Support, College of Intelligent Science, National University of Defense Technology, Changsha, Hunan 410073, China. Material band structure is key foundation for various modern technologies, but it was regarded as a space-time invariant feature. Acoustic metamaterials show extraordinary properties for processing elastic waves, but conventional realizations suffer from narrow bandgaps. Here we first report a nonlinear acoustic metamaterial whose band structure self-adapts to the propagation distance/time and the bandgap exhibits a self-growing behaviour stemming from giant nonlinear interaction. This space-time self-modulating characteristic highlights an unconventional understanding of the band structure, and the self-growth generates an ultralow and ultrabroad bandgap that breaks through the limitation of the mass law for linear locally resonant bandgaps. We also elucidate the self-adaptive mechanisms. This first demonstration sheds light on conceiving advanced devices and metamaterials with broadband, space-time variant bandgaps for wave self-manipulation. Modern technologies rely on the band structures of materials to control electrons, photons and phonons [1, 2]. Advanced functions can be realized by tuning the bands of crystals [3-5]. However, band structures and bandgaps of linear periodic media are independent of space and time with respect to wave propagation, i.e., they are unique and invariable -- the conventional opinion over the century since the band theory was proposed [6,7]. The uncontrolled adaptive modulations of wave propagation (i.e., self-modulation) realized by nonlinearity shows its great reliability in many applications [8]. Plenty of novel lasers and sensors have been created by the self-focusing [9, 10] or self-trapping [11] in nonlinear optical and acoustic media. Although the bandgaps of nonlinear media are amplitude-dependent, they are deemed to be invariable for specified parameters [12-14]. Actually, if the bandgaps could be self-adaptive, the devices relying on them would present interesting adaptive functions, but the feasibility remains unclear. Moreover, low-frequency elastic waves low-frequency or broad bandgaps are in expansive applications [15,16]. Acoustic desired metamaterials [17-19] (AMs) offer unusual functions in manipulating [20-22]. However, in locally resonant (LR)-type linear AMs (LAMs) [23,24] widely studied since 2000, LR bandgaps are narrow in nature [24]. Obtaining broad LR bandgaps whose generalized width γ>1 remains a great challenge [25] due to the limitation of the mass law [26]. Nonlinearity helps make great advances [27-30], and nonlinear AMs (NAMs) [12,13,31] can boost the exploration of new features such as the harmonic generation [32], broadband chaotic mechanisms [33,34] and bridging coupling [35]. Bringing giant nonlinearity in NAM may generate new underlying physics for band [email protected] † [email protected] 1 structures and bandgaps, especially the self-modulating feature and broadband mechanism. In this Letter, we report a NAM with giant nonlinear interactions which presents the first theoretical and experimental demonstration of the space-time variant self-growing bandgap and self-adaptive band structure. Its band structure self-adapts to the propagation distance/ time, and increases spontaneously. self-grown ultralow and ultrabroad Moreover, a (double-ultra) bandgap that breaks through the limitation of the mass law evolves within a short distance. We elucidate its mechanisms with an equivalent approach. the bandwidth Model. -- The unusual phenomenon occurs in the NAM contains strongly coupled resonators featuring giant nonlinearity. A typical fundamental NAM prototype we fabricated is shown in FIG. 1. In the metacell, the primary oscillator m0 is a rectangular bulk, and two neighbouring m0 are coupled through a pair of springs whose entire stiffness is k0. The hollow cylinder m2 has a hole whose radius is 4.045 mm. m2 couples to m0 by two springs with total stiffness k2. A steel sphere m1 is set at the centre point of the cylindroid cavity of m2, and the two are connected by a curved spring whose stiffness is k1. There is a symmetrical clearance, δ0=45±15 μm, between the sphere and the cylinder wall at rest. When p(t)>δ0, m1 will contact m2, which leads to collisions -- a strong interaction. Therefore, m1 and m2 are nonlinear vibro-impact oscillators, and the force between them, FN(t), is a piecewise function. The tiny clearance, δ0, is paramount to generating giant nonlinearity [36]. The NAM prototype consists of 30 metacells with the lattice constant a=27 mm. Its left terminal is connected to a vibration exciter through k0, and the other end is fixed (FIG. 1c). 1d. The motion equations for the the nth cell are (1) Here, un, yn, and zn denote the barycentre displacements of m0, m1, and m2 in the nth cell, respectively, and p=y-z. We specify FN(t)=k1p+kNp3 in theories for generality, where k1 (kN) denotes the linear (nonlinear) stiffness coefficient. Natural frequencies of individual oscillators are ωi=2πfi= i=0, 1, 2. If necessary, the damping coefficient c0=d0·k0 in the primary oscillator is considered. As shown in FIG. 1e, the numerical NAM model consists of 120 periodic metacells. An optimized perfect match layer (PML) is connected to its terminal [37]. The incident wave u0(t) is launched from the other terminal. Parameters in simulations and experiments are: m0=5.8, m1=2.1, m2=2 g; f0=322, f1=100 and f2=390.6 Hz. For the LAM with kN=0, there is a Bragg bandgap and two LR bandgaps, LR1 and LR2 (FIG. 2a,c). Their ranges are LR1 (96.1, 110.4) Hz, with a generalized width γ1=0.15, and LR2 (373.4, 462) Hz, with γ2=0.24. Thus, LR1 and LR2 are narrow bandgaps. Moreover, if k1→∞, the triatomic AM degenerates to a diatomic AM characterized by (254.4, 356.4) Hz near =272.8 Hz, mc=m1+m2, γ=0.4 (FIG. 2b,c). a bandgap LRc FIG. 1 Prototype. (a) Metacell structure. (b) Experimentally measured transmission of a single cell by driving m0 with white-noise signals. Herein, the voltages, v V, represent different driving levels; the result for 0.1 V denotes the linear state; ωpLi and ωpN symbolize peak frequencies for linear and nonlinear models, respectively, i=1, 2. (c) NAM chain consists of 30 meta-cells (only 20 cells are shown here). More experimental images are provided in Ref.[37]. (d) Theoretical triatomic metacell. (e) Numerical NAM model. If only the first linear resonances of m0, m1 and m2 in the x direction are considered, the entire metacell is equivalent to the triatomic configuration shown in FIG. FIG. 2 (a) Dispersive curves of the triatomic LAM. (b) Dispersive curves of the limiting diatomic LAM. The shading regions represent bandgaps. (c) Transmission of local resonances, i.e, the value P/A solved with Eq. (4) below, where the black (blue dashed) line corresponds to the linear triatomic (diatomic) model; other curves are relevant to the cubic nonlinear triatomic model for A=5 µm. (d) Transmissions at the 100th cell, T(100). (e) Wave transmission T(x=na, fe). A0=30 µm and kN=1×1013 N/m3are specified in (c-e). in Phenomena. -- We take kN=1×1013 N/m3 to show the nonlinear phenomena theory. When nonlinearity appears, LR1 and LR2 become nonlinear LR bandgaps, and the coupling between them is visually defined as the bridging coupling [35]. The nonlinear strength for the cubic nonlinear system is σ=3kNA2/k1. σ=32.6 for A=30 μm that indicates giant nonlinearity [36]. In experiments, the transmission spectrum, H(f), of a single cell is measured by inputting different levels of white noises, as illustrated in FIG. 1b. By increasing the excitation, the first resonance notably shifts upwards at first, and the two resonances ultimately merge into a single resonance near 270 Hz. This the giant nonlinearity generated by the clearance. tremendous shift verifies To show the wave transmission in the NAM, the sinusoidal wave packets, u0(t)=A0{[1-cos(ωt/10)]sinωt} for t≤20π/ω, are adopted as input signals, where ω=2πfe is the central frequency and V0≈2πfeA0. Wave transmission at the coordinate x for fe is where Amax=max[u(t)]. At the nth cell, T(n)(fe)=T(x=na, fe). 2 x0()ut(,)uxtLaser(c)(a)ExcitationAmplifierComputer(b)Primary structure0mHollow cylinder1mSphere0k2ma = 27 mmSpringSpring1k(d)2k0k0d2m1m0m31NkpkpInputPML… thnMetamaterial (e)x(,)uxt0()ut0011001121N22N(2)(2)()(),()()nnnnnnnnnnnnnmukuuudkuuukzumyFtmzkzuFtiikmc2c2fkmemaxe0e,,/TxfAxfAf quasi-linear diatomic model that leads to the attenuation in LRc. As the propagation distance/time increases, waves the near-field passbands, 200-fB, are attenuated in gradually due to the bandgap effect. The entire bandgap becomes broader at longer propagation distances, successfully demonstrating space-time variant adaptive band structure and self-growth. The nonlinear bandgap at the 13th cell has reached 200-500 for levels 8 V (180-500 Hz for level 4 V [37]). Our experiments show that the bandgap grows into a double-ultra one within 9~13 cells that is much faster than the numerical prediction in FIG. 2e. the For an infinite LAM, T≈1 in passbands in theory, and bands for T≤0.2 (but not T→0) are determined as bandgaps due to the broadband energy in the input sinusoidal packet. As shown in FIG. 2d, a surprising phenomenon is that the far-field transmission of the NAM is greatly reduced (T(100)≈0.2) in the range 150-550 Hz. It is the bandgap at the 100th cell and its γ reaches 2.67, a double-ultra bandgap. Over 550-670 Hz, T(100)≈0.3 is much smaller than that of LAM. Regulars remain consistent for stronger nonlinearity. However, present studies fail to reveal the mechanism for this phenomenon. The transmission in FIG. 2e illustrates that the near-field (x<20a) bandgap is not LR1 or LR2. Instead, it is LRc, indicating that m1 and m2 in the near-field meta-cells behave as an integrated resonator, mc, due to the giant nonlinearity. Moreover, as x increases, the initial frequency for the greatly attenuated band, fst, shifts downwards gradually, and the transmission above LRc decreases much faster. Thus, the total band for wave suppression broadens until the entire range from LR1 to LR2 (even all above LR1) obtains bandgap effects. This process shows that the bandgap in this NAM adaptively grows from a narrow one, LRc, into a double-ultra one covering the range LR1-LR2, γ=3.5 (at least), which far exceeds the bandwidth of the LAM with mc/m0 =0.7 [26]. We term this phenomenon a self-growing bandgap. As x=c·t, where c and t denote the wave velocity and propagation time, a space-dependent property is also time-dependent. the band structure determines Self-growth highlights a space-time self-modulating band structure, which modifies the conventional concept that the bandgap is invariable for specified condition. Furthermore, as the material properties, refraction, focusing, and attenuation that depend on the properties may be adaptive. Therefore, the new fundamental law enlightens us the extensive potential applications for manipulating elastic waves with broadband self-adaptive properties. the wave velocity, transmission, scattering, insulation fe in Experiments. -- NAM responses under three driving levels, 2, 4 and 8 V, are measured by a Doppler laser vibrometer (FIG. 3). A0 decreases with the experiments (FIG. 3a). At a certain propagation distance, a greater transmission loss in 200-350 Hz is induced by a larger A0 (FIG. 3b). A comparison between the numerical and experimental results of a diatomic LAM shows that there is an unavoidable weak damping, d0=4×10-5 s, in the experiments. Moreover, to reproduce the experimental phenomena, we establish a numerical model consists of 30 clearance-nonlinearity, which confirms the credibility of experiments [37]. As shown in FIG. 3c, the near-field transmission exhibits a jump at a frequency, fB. It is the grazing bifurcation [38] frequency of the vibro-impact oscillation [37]. containing cells indicate Figures 3c,d (n<3) transmission near 273 Hz is greatly reduced, but waves in 0-255 Hz and 350-fB can still propagate. This result demonstrates that the near-field NAM behaves as a the near-field that 3 FIG. 3 Experimental demonstration. Panels (a-d) are results for inputting sinusoidal packets. (a) A0≈V0/ω, where V0 denotes the directly measured velocity amplitude. (b) Experimentally measured transmissions at 8th cell. Here, 2 V, 4 V and 8 V are the three driving levels for the NAM; 'L' represents the result for the diatomic LAM prototype. (c, d) Transmissions under 8 V. (e, f) Transmissions obtained in sweep-frequency experiments. In (c, e), the number, n, in legends represents the nth cell; 'L' represents the numerical result of the 5th cell in the linearized model. (d h) Transmission T(x=na, fe), n=x/a. Moreover, this experiment still prove that opening the bandgap by observing the residual amplitude below 180 Hz requires a longer propagation distance, which is attributed to the challenge in attenuating all broadband energy of packets. The fast sweep-frequency experiment [37] (see FIG. 3e,f) shows a faster self-growing speed, especially for f<250 Hz. Herein the entire bandgap successfully expands to 130 Hz at the 13th cell, which is close to LR1. Certain peaks at high frequencies correspond to the standing waves arising from the boundary finite NAM. Results demonstrate that the total bandgap can adaptively grow to reflection of the at least 130~500 Hz (γ=2.85) within 9~13 cells. For mc/m0=0.7, this unusual double-ultra bandgap extends far beyond the mass law of a LR bandgap. Mechanisms. -- We propose an equivalent linearized approach based on the bifurcation of the coupled nonlinear local resonance to clarify the self-growing mechanisms [37]. A plane wave in the 1D medium is , where A0 is the source amplitude; the wave number κ=κr+iκI, and μ=κa=μr+iμI. The dispersion equation of this NAM is , and the bandgaps shift downwards; if a shifted bandgap covers ω, a significant reduction, , is induced. (ii) All bandgaps are active to reflect the broadband energy. Once the attenuation begins, band structure varies and broader bands are swept by bandgaps, then more energy is reflected, thereby larger reductions are induced. This is a self-strengthening process. Finally, the observed far-field bandgap is significantly broadened, i.e., the self-adaptive band structure and the self-growing bandgap are realized. (2) where ω1eL denotes the equivalent nature frequency of m1, (3) here, ωJ1 is the first bifurcation frequency of the nonlinear local resonances as labelled in FIG. 2c, and we determine it by the algebraic equations: (4) where Y and P denote the amplitudes of y(t) and p(t), respectively. Therefore, the band structure is determined by the wave amplitude A. When considering damping, we replace k0 in Eq. (2) by k0(1+iωd0). The bandgap can be described by κI=μI/a>0 solved with Eq. (2). The amplitude reduction in propagation distance Δx denotes . Bandgaps lead to great attenuation attributing to the large κI. The amplitude-dependent regime in FIG. 4a shows that the band structure approaches the state of the diatomic LAM if giant nonlinearity is generated when A is large, which agrees with the regulars in FIG. 2e. Furthermore, if A decreases from a large value, LRc approaches LR1, and LR2 first decouples from the Bragg gap and then shifts to lower frequencies. A narrow band referred as leakage band is never swept by the adaptive bandgap. It results in the transmission peak near 370 Hz in FIG. 2d. Moreover, we establish a large numerical model to separate the incident and reflected waves in the NAM [37]. Wave inside NAM is a superposition of the forwardly transmitting and backward reflected waves. If (A0, ω) first appears in LRc, as represented by the spectra of 270 Hz in FIG. 4b, most input energy is reflected by several metacells near the incident boundary, thereby the amplitude decreases rapidly. If (A0, ω) first appears in passbands, κI=0 for d0=0 in linear regimes, thereby the amplitude never attenuates, and the band structure never varies in linear case. However, as represented by 200 Hz in FIG. 4c,d, waves transmitting into NAM undergo the harmonic generation [32] and chaotic responses that pump the fundamental energy to a broad band [37]. This process also reduces the total amplitude. Then after a certain distance, two effects appear. (i) The amplitude-dependent band structure varies 4 FIG. 4 (a) Amplitude-dependent bandgap. The first and third insets are dispersion curves of the diatomic and triatomic LAMs. The second shading image is the distribution of the imaginary wave vector, μI=κIa, in which the white regions represent μI=0 for passbands. (b, c) Velocity spectra for 270 and 200 Hz of the input wave, reflected wave, waves at 1th (X1) and 100th (X100) triatomic cells, respectively. (d) Distance-dependent velocity spectra for 200 Hz. (e, f) T(x=na, fe) solved with the equivalent method, kdN=2.5×10-5. (e) d0=0; (f) Damped case for d0=4×10-5. In (c, d), kN=1×1013 N/m3, A0 = 30 μm. As analysed, without the damping, the self-growth is initiated by harmonic generation and chaotic responses. Defining dN as the entire attenuation rate induced by them, we have . Simulations and experiments indicate that dN ω. We adopt dN=kdNω to approximate this effect. By specifying kdN=2.5×10-5, the analytical transmissions (FIG. 4e) agrees very well with the numerical illustration in FIG. 2e, which confirms the initializing regimes and verifies that the amplitude- distance-dependent bandgap is the main mechanism for the adaptive property. However, as kdN is small, opening the double-ultra bandgap needs 40 cells. Fortunately, a weak damping, d0=4×10-5 s, can greatly accelerate the growth, as shown in FIG. 4e, and it agrees with the experiments. Thereby, damping the reason for is r()(,)eixtuxtAI0()exAxA2222220211eL21eL2222220021eL211eL[()]22[()()1]cosmmmkkmm422J1J121eL2212J12(1)mm231122221234()()NmYkPkPkmYPmYkAIexIexNI0()eedxxAxAaccelerating the self-growth. Natl Sci Rev 5, 159 (2018). Conclusions. -- This letter presents the first theoretical and experimental demonstration of the time-space variant adaptive band structure and self-growing bandgap in a AM featuring giant nonlinear coupling. This finding changes the conventional opinion on band structures, and the bandwidth breaks through the limitation of the mass law. Due to the analogous properties of elastic, optical and electromagnetic waves, expanding the scope of this study is desirable. [1] J. Kim, S. S. Baik, S. H. Ryu, Y. Sohn, S. Park, B. G. Park, J. Denlinger, Y. Yi, H. J. Choi, and K. S. Kim, Science 349, 723 (2015). [2] M. Abdi-Jalebi, Z. Andaji-Garmaroudi, S. Cacovich, C. Stavrakas, B. Philippe, J. M. Richter, M. Alsari, E. P. Booker, E. M. Hutter, and A. J. Pearson et al., Nature 555, 497 (2018). [3] M. Schwarze, W. Tress, B. Beyer, F. Gao, R. Scholz, C. Poelking, K. Ortstein, A. A. Gunther, D. Kasemann, and D. Andrienko et al., Science 352, 1446 (2016). [4] Z. Chen, X. Zhang, S. Lin, L. Chen, and Y. Pei, Natl Sci Rev (2018). [5] R. Bogue, Sensor Rev 37, 305 (2017). [6] M. Born, and V. K. T. T, Z Phys, 297 (1912). [7] R. Hofstadter, Felix Bloch (1905-1983) (National Academy of Sciences, Washington D.C., 1994). [8] Z. Wu, Y. Zheng, and K. W. Wang, Phys. Rev. E 97, 22209 (2018). [9] X. Wang, H. Chen, H. Liu, L. Xu, C. Sheng, and S. Zhu, Phys. Rev. Lett. 119, 33902 (2017). [10] P. L. Kelley, Phys. Rev. Lett. 15, 1005 (1965). [11] S. Autti, Y. M. Bunkov, V. B. Eltsov, P. J. Heikkinen, J. J. Hosio, P. Hunger, M. Krusius, and G. E. Volovik, Phys. Rev. Lett. 108, 145303 (2012). [12] X. Fang, J. Wen, J. Yin, D. Yu, and Y. Xiao, Phys. Rev. E 94, 52206 (2016). [13] X. Fang, J. Wen, J. Yin, and D. Yu, Aip Adv 6, 121706 (2016). [14] R. K. Narisetti, M. J. Leamy, and M. Ruzzene, Journal of Vibration & Acoustics 132, 31001 (2010). [15] Y. Y. Chen, M. V. Barnhart, J. K. Chen, G. K. Hu, C. T. Sun, and G. L. Huang, Compos. Struct. 136, 358 (2016). [16] J. Y. Tsao, S. Chowdhury, M. A. Hollis, D. Jena, N. M. Johnson, K. A. Jones, R. J. Kaplar, S. Rajan, C. G. Van de Walle, and E. Bellotti et al., Adv Electron Mater 4, 1600501 (2018). [17] Z. Liu, X. Zhang, Y. Mao, Y. Y. Zhu, Z. Yang, C. T. Chan, and P. Sheng, Science 289, 1734 (2000). [18] S. Zhang, C. Xia, and N. Fang, Phys. Rev. Lett. 106, 24301 (2011). [19] H. Abbaszadeh, A. Souslov, J. Paulose, H. Schomerus, and V. Vitelli, Phys. Rev. Lett. 119, 195502 (2017). [20] S. H. Lee, C. M. Park, Y. M. Seo, Z. G. Wang, and C. K. Kim, Phys. Rev. Lett. 104, 54301 (2010). [21] N. Kaina, F. Lemoult, M. Fink, and G. Lerosey, Nature 525, 77 (2015). [22] M. Yang, and P. Sheng, Annu Rev Mater Res 47, 83 (2017). [23] H. Ge, M. Yang, C. Ma, M. Lu, Y. Chen, N. Fang, and P. Sheng, [24] G. Ma, and P. Sheng, Sci Adv 2, e1501595 (2016). [25] V. Romero-García, A. Krynkin, L. M. Garcia-Raffi, O. Umnova, and J. V. Sánchez-Pérez, J. Sound Vib. 332, 184 (2013). [26] The generalized width of a bandgap is γ=(fcut-fst)/fst, where fst (fcut) is its initial (cutoff) frequency. A LAM's LR bandgap obeys the mass law , where mc (m0) denotes the mass of the local resonator and primary oscillator. A smaller mass ratio mc/m0 is better for most applications. For example, γ ≈ 0.3 for mc/m0=0.7, and a huge value, mc/m0=3, is required to obtain γ=1 in LAM. Therefore, γ>1 can be regarded as a double-ultra bandgap because there is γ<<1 generally. [27] G. Wehmeyer, T. Yabuki, C. Monachon, J. Wu, and C. Dames, Appl Phys Rev 4, 41304 (2017). [28] V. F. Nesterenko, C. Daraio, E. B. Herbold, and S. Jin, Phys. Rev. Lett. 95, 158702 (2005). [29] B. Liang, X. S. Guo, J. Tu, D. Zhang, and J. C. Cheng, Nat. Mater. 9, 989 (2010). [30] B. Deng, P. Wang, Q. He, V. Tournat, and K. Bertoldi, Nat Commun 9, 3410 (2018). [31] Y. Li, J. Lan, B. Li, X. Liu, and J. Zhang, J. Appl. Phys. 120, 145105 (2016). [32] X. Fang, J. Wen, D. Yu, G. Huang, and J. Yin, New J. Phys. 20, 123028 (2018). [33] X. Fang, J. Wen, B. Bonello, J. Yin, and D. Yu, Nat Commun 8, 1288 (2017). [34] X. Fang, J. Wen, B. Bonello, J. Yin, and D. Yu, New J. Phys. 19, 53007 (2017). [35] X. Fang, J. Wen, D. Yu, and J. Yin, Phys Rev Appl 10, 54049 (2018). [36] Nonlinear strength: The nonlinear strength σ<0.1 is weak nonlinearity and σ=0.3 can generate strongly nonlinear phenomena [12]. In our experiments, if we approximate the clearance nonlinearity with the smooth cubic nonlinearity, we obtain kN=3.3×1012 N/m3 [37]. The amplitude under 8 V reaches 533 and 53 μm for 100 and 300 Hz, respectively, and their σ=3392 and 33.5. They are huge numbers relative to σ=0.3. Therefore, it can generate giant nonlinear responses. [37] see Supplementary Material. [38] M. di Bernardo, C. J. Budd, A. R. Champneys, P. Kowalczyk, A. B. Nordmark, G. O. Tost, and P. T. Piiroinen, Siam Rev. 50, 629 (2008). 5 c011mm
1711.03603	1	1711	2017-11-09T20:58:14	Angle-dependent reflectivity of twill-weave carbon fibre reinforced polymer for millimetre waves	[ "physics.app-ph" ]	Twill-weave carbon fibre reinforced polymer is measured as reflector material for electromagnetic waves at 60 GHz. The reflectivity at millimetre wavelength shows a predominant direction, which coincides with the diagonal pattern of the twill's top layer. In the remaining directions the investigated 2/2 twill-weave composite's reflectivity is almost isotropic.	physics.app-ph	physics	Angle-dependent reflectivity of twill-weave carbon fibre reinforced polymer for millimetre waves G. Artner, E. Zochmann, S. Pratschner, M. Lerch, M. Rupp and C. F. Mecklenbrauker November 13, 2017 Abstract Twill-weave carbon fibre reinforced polymer is measured as reflector material for electromagnetic waves at 60 GHz. The reflectivity at millimetre wavelength shows a predominant direction, which coincides with the diagonal pattern of the twill's top layer. In the remaining directions the investigated 2/2 twill-weave composite's reflectivity is almost isotropic. 1 Introduction Carbon fibre reinforced polymer (CFRP) are composites, with carbon fibres embedded in a polymer matrix. Nowadays, they are widely used in the con- struction of lightweight vehicles and antennas. Typical applications of CFRP are reflectors for spacecraft antennas [1]. A millimetre wave (mmW) antenna with a CFRP reflector for radar applications is presented in [2, 3]. Measurements of CFRP reflectivity and conductivity consider mostly unidi- rectional fibres and few material orientations [4–6]. High precision reflectivity measurements in the range of 110-200 GHz of unidirectional CFRP measured in the fibre direction and perpendicular to the fibre direction are conducted with a Fabry-Perot resonator in [4]. There, the reflector is mounted in fixed orientations and material orientation is not considered. Riley et al. present measurements of the radar cross section of twill-weave CFRP in [5], but only in a mono-static set-up and sample orientation was not considered. Artner et al. estimate the conductivity of twill-CFRP from wave-guide measurements at 4-6 GHz with the Nicolson-Ross-Weir method in steps of 10◦ [6]. Contribution: In this letter, the anisotropy of twill-weave CFRP reflectivity at millimeter wavelength is observed in a bi-static set-up. Measurements are performed with millimetre waves at 60 GHz. Several widely used CFRP fabrics were investigated: unidirectional fibres, shredded fibres, plain weave and twill weave. A sharp angular dependency in reflectivity of twill-weave CFRP is found around angles, where the direction of the twill pattern aligns with the incident 1 Figure 1: Photograph of the measured 2/2 twill-weave CFRP. Ruler dimensions are in centimetre. The photograph is sketched with various degrees of abstrac- tions: It reveals fibre strand orientation, twill-weave pattern and the distinctive diagonal twill-direction (dashed lines). The blue dotted square marks the twill weave pattern. wave's electric field vector. The sharp angular dependency is only found in bi-static set-ups and only for the twill-weave's reflectivity. Other investigated CFRPs (plain-weave, unidirectional, shredded-fibres) are omitted, as measure- ments do not show the observed effect. To the best of the authors' knowledge this effect has not been measured or described before. 2 Twill-weave CFRP The material under test (MUT) is a CFRP laminate of 1.6 mm total thickness, produced from 2/2 twill-weave plies with the press method. A circular disk with a diameter of dMUT = 280 mm was cut by a water-jet. Sheets are commer- cially available. A photograph of the MUT is displayed in Fig. 1. Sketches are superimposed, which reveal fibre-, strand- and twill-pattern-directions. On the surface, the fibre strands have a visible size of about 2 × 4 mm2, which is within the same order of magnitude as the free space wavelength at 60 GHz of λ ≈ 5 mm. The blue dotted square in Fig. 1 shows the twill-weave pattern (Bindungspatrone). This quadratic pattern of size dtwill× dtwill has a side length of 8 mm. Theoretically the twill-weave CFRP is an example of a finitely con- ducting bigrating with λ/dtwill ≈ 5/8 = 0.625, see [8] Section 7.6, pp. 258ff. The lower layers' geometry is unimportant due to the electromagnetic waves' low penetration depth resulting from the skin effect. 2 3 Measurement set-up Two identical, horizontally polarised conical horn antennas are used in a bi- static set-up as shown in Fig. 2. This polarisation corresponds to the transversal magnetic (TM) reflection case. Electromagnetic waves with a center frequency of 60 GHz are transmitted by one antenna, reflected by the MUT and received by a second antenna. The baseband transmit signal is a 2 GHz wide multi-tone waveform with Newman phases to ensure a low crest factor [7]. It is generated with an arbitrary waveform generator and up-converted with an external mmW- module. At the receiver side, the 60 GHz signal is directly measured with a spectrum analyser. The received power over the whole bandwidth is recorded and subsequently averaged. The MUT disk is placed on a rotary plate to adjust the turn angle ϕ au- tomatically in steps of 1◦. This allows automated measurement of material anisotropy, by changing the MUT's alignment towards the incident polariza- tion. The reflection angle θ is adjusted by changing the distance between the antennas. The normal distance R is kept constant, due to the mounting of the antennas as shown in Fig. 2. For each reflection angle θ, the antennas are aligned to point towards the MUT's center. The horn antennas have an aperture of 27.41 mm and a 3 dB opening angle of α3dB = 18◦. The distance R was chosen to be 55 cm, such that the MUT is in the far-field and most of the energy is focused on the MUT. Absorbers are placed around the MUT to eliminate undesired reflections. Figure 2: Geometry of the measurement set-up, top view (left). Photograph of the two horn antennas and the MUT in front of absorbers (right). 4 Measurement results We observe a sharp change in reflectivity at angles corresponding to the diagonal twill pattern on the top layer. Measured results of the twill-CFRP's reflectivity are presented in Fig. 3. Due to the constant distance R, the average line-of-sight distance to the MUT depends on the reflection angle θ. To allow comparison of the angular dependency, receive power is normalised, i.e., P (ϕ, θ)/Eϕ {P (ϕ, θ)}, where Eϕ denotes the expectation operator with respect to ϕ. 3 The estimated temperature coefficient of our measurement set-up is −0.14 dB/K. During our measurements, the temperature varied by ±0.7 K. The receive power is corrected by the estimated temperature coefficient. Measurements are performed for θ ∈ {0◦, 16◦, 33◦}. At θ ≈ 0◦ the antennas are placed above each other. For each elevation angle θ the MUT is rotated two full turns in azimuth ϕ. Both curves are shown in Fig. 3 in order to visualize the influence of MUT rotation precision. trate estimation inaccuracies. Aluminium's measured isotropic reflectivity is shown as comparison to illus- The curves show reflectivity dependency on the MUT orientation ϕ. A 180◦ periodicity is evident due to the MUT twill-weave. Sharp changes in reflectivity are observed at certain angles of MUT orien- tation ϕ, close to the diagonal twill-direction. Depending on reflection angle θ, one or two notches appear. The notches are not observed for the mono- static case, i.e., θ ≈ 0◦. The notches appear around the diagonal twill-direction and also have a periodicity of 180◦. We therefore conclude, that these sharp angular changes in reflectivity are caused by the CFRP twill pattern. Sur- prisingly, no significant influence is found in the directions of warp and weft ϕ ∈ {0◦, 90◦, 180◦, 270◦}, where fibre directions coincide with field vectors. 5 Conclusion Twill-CFRP's electromagnetic reflectivity is anisotropic at millimetre wave- lengths. Sharp notches in reflectivity are observed, that appear around the predominant direction of the diagonal twill pattern. These notches do not ap- pear for mono-static set-ups. In the directions of warp and weft ϕ ∈ {0◦, 90◦, 180◦, 270◦} no significant influence on reflectivity is observed. A quantitative numerical electromagnetic simulation for the twill-weave CFRP is still open as well as full elevation scans of reflected power. G. Artner, E. Zochmann, S. Pratschner, M. Lerch, M. Rupp and C. F. Meck- lenbrauker (Institute of Telecommunications, Technische Universitat Wien, 1040 Vienna, Austria) E-mail: [email protected] References [1] Keen, K. M.: 'Gain-loss measurements on a carbon-fibre composite reflector antenna', Electronics Letters, 1975, vol. 11, no. 11 [2] Futatsumori, S., Kohmura, A. and Yonemoto, N.: 'Performance measure- ment of compact and high-range resolution 76 GHz millimeter-wave radar 4 Figure 3: Normalised measured reflected power of twill-CFRP for two full turns of the MUT. The second turn is displayed opaque. 5 0 dB-1-2+10°90°180°270°315°45°225°135°352°301°172°121°323°142°θ≈0°θ=33°θ=16°aluminiumφsystems for autonomous unmanned helicopters', IEICE Transactions on Elec- tronics, 2013, vol. E96-C, no. 4, pp. 586-594 [3] Futatsumori, S., Morioka, K., Kohmura, A., Shioji, M. and Yonemoto, N.: 'Fundamental applicability evaluation of carbon fiber reinforced plastic mate- rials utilized in millimeter-wave antennas', in IEEE Conference on Antenna Measurements & Applications (CAMA), 2014, Antibes Juan-les-Pins [4] van't Klooster, C. G. M., Parshin, V. V., and Myasnikova, S. E.: 'Reflectivity of antenna reflectors: measurements at frequencies between 110 and 200 GHz', in IEEE Antennas and Propagation Society International Symposium Digest, 2003, Columbus, OH, USA, vol. 3, pp. 528-531 [5] Riley, E. J., Lenzing, E. H., and Narayanan, R. M.: 'Characterization of radar cross section of carbon fiber composite materials', in Proceedings SPIE Defense+ Security, 2015, 9461 [6] Artner, A., Gentner, P. K., Nicolics, J., and Mecklenbrauker, C. F.: 'Carbon Fiber Reinforced Polymer with Shredded Fibers: Quasi-Isotropic Material Properties and Antenna Performance', International Journal of Antennas and Propagation, 2017, vol. 2017, [7] Newman, D. J.: 'An L1 extremal problem for polynomials', Proceedings of the American Mathematical Society, 1965, vol. 16.6, pp. 1287-1290 [8] Petit, R.(Ed.): 'Electromagnetic Theory of Gratings' (Springer, Berlin- Heidelberg-New York, 1980, 1st edition) 6
1904.08192	1	1904	2019-04-17T11:21:16	Grid Inadequacy Assessment against Power Injection Diversity from Intermittent Generation, Dynamic Loads, and Energy Storage	[ "physics.app-ph", "eess.SY" ]	The integration of more intermittent generation, energy storage, and dynamic loads on top of a competitive market environment requires future grids to handle increasing diversity of power injection states. Grid planners need new tools and metrics that can assess how vulnerable grids are against this future. To this end, we propose grid inadequacy metrics that expose grid inability to accommodate power injection diversity from such sources. We define the metrics based on a previously unexplored characterization of grid inadequacy, that is, the size of the DC power flow infeasible set relative to the size of the power injection set is indicative of inherent grid inadequacy to accommodate power injection diversity without intervention. We circumvent the difficulty of characterizing the high-dimensional sets involved using three approaches: one sampling-based approach and two approaches that project the sets in lower dimensions. Illustrative examples show how the metrics can reveal useful insights about a grid. As with other metrics, the proposed metrics are only valid relative to the assumptions used and cannot capture all intricacies of assessing grid inadequacy. Nevertheless, the metrics provide a new way of quantifying grid inadequacy that is potentially useful in future research and practice. We present possible use-cases where the proposed metrics can be used.	physics.app-ph	physics	Grid Inadequacy Assessment against Power Injection Diversity from Intermittent Generation, Dynamic Loads, and Energy Storage 1 Adonis E. Tio, David J. Hill, Life Fellow, IEEE and Jin Ma This work looks at a new approach to defining grid inade- quacy in future grids with increased diversity in power injec- tion states. We characterize grid inadequacy as a grid's inher- ent inability to serve all possible power injection states with- out any intervention. The proposed inadequacy metrics are based on the following idea that we believe is not yet consid- ered in existing literature. That is, given all possible power injection scenarios, the proportion of infeasible scenarios that lead to network congestion provide an indicative measure of inherent grid inadequacy against power injection diversity. This work provides an initial working framework that explores this characterization of grid inadequacy that researchers and practitioners may find useful in the future. In the proposed framework, we represent the uncertainty in bus power injections from renewables, dynamic loads, and energy storage together with injections from more convention- al generation and loads using interval sets that span the whole power injection space. We then use the constrained DC power flow model to identify infeasible subspaces. Since it is gener- ally difficult to make sense of the high-dimensional sets in- volved, we use three approaches to develop numeric quantities or metrics that describe the infeasible set. While single numer- ic values cannot fully capture the complex relationships be- tween the sets involved, let alone capture the intricacies of defining grid inadequacy, the metrics can still provide valua- ble insights on network vulnerabilities and bottlenecks that can motivate detailed analyses later on. We organize the paper as follows. Section II reviews the available literature on power injection diversity and grid ade- quacy assessment, identifies shortcomings, and motivates the need for new dedicated measures of grid inadequacy in the future grid setting. Section III presents a framework for meas- uring grid inadequacy using the infeasible set of the con- strained DC power flow model. Section IV defines tractable metrics that embody the framework presented in Section III. Section V show examples for the 6- and 118-bus test systems. Section VI comments on the framework assumptions, identi- fies areas for future work, and presents a potential use-case. Section VII concludes. Abstract1 -- The integration of more intermittent generation, energy storage, and dynamic loads on top of a competitive mar- ket environment requires future grids to handle increasing diver- sity of power injection states. Grid planners need new tools and metrics that can assess how vulnerable grids are against this fu- ture. To this end, we propose grid inadequacy metrics that expose a grid's inability to accommodate power injection diversity from such sources. We define the metrics based on a previously unex- plored characterization of grid inadequacy, that is, the size of the DC power flow infeasible set relative to the size of the power in- jection set is indicative of inherent grid inadequacy to accommo- date power injection diversity without intervention. We circum- vent the difficulty of characterizing the high-dimensional sets involved using three approaches: one sampling-based approach and two approaches that project the sets in lower dimensions. Illustrative examples show how the metrics can reveal useful insights about a grid. As with other metrics, the proposed metrics are only valid relative to the assumptions used and cannot cap- ture all intricacies of assessing grid inadequacy. Nevertheless, the metrics provide a new way of quantifying grid inadequacy that is potentially useful in future research and practice. We present possible use-cases where the proposed metrics can be used. Keywords -- power system planning, future grids, inadequacy metrics, power injection diversity, transmission network expansion I. INTRODUCTION The integration of bulk and distributed intermittent renew- able generation, dynamic loads, and energy storage systems on top of a competitive market environment is expected to intro- duce more diversity in power injection states in future grids. Coupled with long-term uncertainty in user demand patterns, demographic trends, climate change, economic activity, and policies around new technologies, it will increasingly become more difficult to design adequate, secure, reliable, and cost- effective grids. There is a need to develop new measures of grid adequacy, inadequacy, and risks that captures the interre- lated uncertainties in both operational and strategic timescales. A. E. Tio acknowledges1the DOST-ERDT Faculty Development Fund of the Republic of the Philippines for sponsoring his degree leading to this research. A. E. Tio is a PhD student at the School of Electrical and Information Engi- neering, University of Sydney, Sydney, NSW 2006, Australia and a faculty (on-leave) of the Electrical and Electronics Engineering Institute, University of the Philippines, Diliman 1101, Quezon City, Philippines (e-mail: ado- [email protected]). D. J. Hill is with the Department of Electrical and Electronics Engineering, The University of Hong Kong, Hong Kong, and the School of Electrical and Information Engineering, University of Sydney, Sydney, NSW 2006, Austral- ia (e-mail: [email protected]). J. Ma is with the School of Electrical and Information Engineering, Univer- sity of Sydney, Sydney, NSW 2006, Australia (e-mail: [email protected]). II. POWER INJECTION DIVERSITY AND GRID INADEQUACY A. Grid adequacy assessment Ref. [1] defines power system adequacy as the existence of enough facilities to meet customer needs and operational constraints at static conditions under a set of probabilistic sys- tem states. Assessing the adequacy of the transmission net- work is generally complicated because of complexities in both (a) network and operations modeling and (2) system state se- lection, see Fig. 1. Network and operations modeling involves the choice of a power flow model, whether contingencies are considered, and whether power flow control such as power injection, topology, impedance, or voltage angle control are modeled. System state selection depends on the application and can span both probable and possible system states hours or days ahead in an operational planning setting and years or decades ahead in a strategic planning setting. System states are much more predictable in the operational setting with future system states more likely to resemble system states from the immediate past. On the other hand, system states in the distant future are much harder to predict because of the dependency on extraneous factors such as climate change, technological development, and legal and market restructuring. Because of this complexities, measures of grid adequacy are highly de- pendent on the assumptions used and applicable only within the intended use-case. Fig. 1 Basic components of grid adequacy assessment. With the integration of more intermittent renewable gen- eration, dynamic loads, and energy storage on top of existing deregulated electricity grids, the number and diversity of sys- tem states are expected to increase. Buses can both become a source and a sink as energy storage becomes embedded in generator buses with either dispatchable or intermittent gener- ation as well as in load buses with distributed generation, tra- ditional loads, and dynamic loads. Combined with new mar- ket, data, and communications infrastructure, market players will have more and more flexibility in controlling when and how much energy they will draw or inject from the grid de- pending on prevailing and forecasted market conditions. These trends will require a rethink of what grid adequacy means. Ideally, future grids must be able to facilitate all possible en- ergy exchanges between future market players without dis- crimination, while keeping the service secure, reliable, and low-cost. New tools and metrics must be explored to study the adequacy of existing grids and proposed reinforcements against such a future. This work looks at the antithesis of grid adequacy which is grid inadequacy. As with characterizing adequacy, charac- terizing grid inadequacy is complicated and highly depends on 2 the modeling assumptions used and system states modeled. It is equally important to quantify grid inadequacy (what a grid cannot do) but most work tends to focus on characterizing grid adequacy (what a grid can do). Since doing grid adequacy assessment does not necessarily provide grid inadequacy in- formation, there is a need to study and develop dedicated as- sessment frameworks and measures of grid inadequacy. B. Existing metrics We first review the generic adequacy metrics presented in [1] that can be used with any combination of modeling as- sumptions used and system states chosen. We then limit our literature review to more application-specific measures of grid adequacy under non-contingency situations against a defined set of power injection states or scenarios. The papers use the DC power flow model and most consider power flow control interventions in the form of generator dispatch flexibility or load shedding but not topology, impedance, or voltage angle control. The range of power injection states reflect the intend- ed applications and spans hours, days, years, or decades ahead into the future. Ref. [1] presents seven basic grid adequacy indices and five derived indices. The indices revolve around characteriz- ing load curtailments given a discrete pool of system states. Some of the more notable ones include the Probability of Load Curtailments (PLC), Expected Number of Load Curtail- ments (ENLC), and Expected Energy Not Supplied (EENS). PLC is the sum of probabilities of all states that lead to cur- tailment while ENLC is the total number of load curtailment states. EENS gives the expected energy curtailment within a given period. It is more commonly encountered in generation adequacy assessment but can also be used to incorporate the effects of transmission adequacy. These metrics are defined broadly enough to accommodate any level of network and operations modeling fidelity over a range of pre-selected sys- tem states. The basic framework is to enumerate discrete sys- tem states and evaluate which states lead to curtailment. After evaluating the enumerated states, adequacy metrics can be used to describe the properties of the curtailment states. The obvious drawback of this framework is that the metrics are only as good as the system states identified. As such, proper context and caveats must be established when using these met- rics in decision-making. Other works define measures of grid adequacy more spe- cifically and tailored to an application. In works related to operations planning, measures of grid adequacy are described in terms of security, flexibility, loadability, or robustness. De- spite the different terminologies used, these measures pertain to a grid's ability to accommodate power injection states under static conditions, in contrast to other metrics evaluated under dynamic conditions. Some work characterize adequacy in terms of sets while others define metrics. In [2] -- [4], security regions are used to define allowable variations in power injec- tions about an operating point. To compute the allowable ad- justments, Ref. [2] uses a linear program based on the con- strained DC power flow model to maximize the sum of allow- able adjustment range of all generators. A performance index is then defined using the proportion of the allowable genera- tion adjustment relative to the maximum generation adjust- ment for the whole system. Likewise, Ref. [5] uses the allow- able generator adjustments to define grid flexibility but solves for the ranges using a different approach: by taking the differ- ence between a reference dispatch and a new dispatch that minimizes the line loading margins. A flexibility index is de- fined as the maximum generation variation in each generator bus or the whole system. Ref. [6] uses the range of allowable load variations to define grid robustness under load uncertain- ty. The authors solve a linear program repeatedly to approxi- mate the extent of allowable load changes about a loading condition. A robustness index is then defined in terms of the probability of a grid working correctly even if it operates out- side a specified set of operating conditions. And in [7], a set of constraints is used to define loadability sets. The paper notes that metrics derived from the regions bounded by the loadabil- ity sets, like inner-area approximations and centroid calcula- tions, would provide useful indicators of grid flexibility and are important areas of future research. In works related to infrastructure planning, Refs. [8] and [9] use an ad hoc sampling-based approach to compare the robustness of grid expansion options to both load and renewa- ble generation uncertainty. K scenarios are drawn from the distribution of uncertain load and generation variables and the number of scenarios that do not result in curtailment, K1, are counted. The larger the ratio of K1 and K, the more robust a grid is to power injection variability. This approach follows the framework behind ENLC tailored for a planning problem under load and renewable generation uncertainty. Ref. [10] uses the maximum load shedding under uncertainty as metric for comparing grid alternatives. Ref. [11] defines a framework to quantify the benefits of grid-side flexibility options provid- ed by line switches and FACTS devices using sets defined from the constrained DC power flow model. The idea is to compare the size of feasible sets with and without flexibility options installed. Instead of defining discrete system states, spaces in the power injection space are used to represent sys- tem states in a continuum. And in a completely radical per- spective, Ref. [12] uses a graph theoretic approach to grid ad- equacy assessment. The authors propose to use the volume of feasible bus injections to measure grid flexibility to absorb and deliver power. The authors claim that line additions that max- imize the determinant of the DC power flow susceptance ma- trix will also maximize the volume of feasible bus injections under certain assumptions. However, it remains unclear whether improvements in underlying graph structures of pow- er grids will indeed translate to actual improvements in grid adequacy. C. Need for new metrics As we mentioned in Section II-A, adequacy metrics are on- ly as good as the assumptions and methodologies used. As such, each has its own limitations and intended use-cases. In this section, we comment on the limitations of the reviewed works to motivate the need for new dedicated measures of grid inadequacy against increased power injection diversity in fu- ture grids. The basic metrics presented in [1] rely on the enumeration of discrete power injection states. In the future, scenario sam- pling and scenario reduction techniques can be used to derive a tractable but representative number of power injection states. However, planners must also explore grid vulnerabilities out- 3 side of the pre-selected states and quantify risk measures on being in these states. The methods in [2] -- [6] are defined for a specific operating point and are limited to modeling either generation variability only or load variability only. While use- ful in their respective intended applications, overall measures of grid performance that are independent of a specific operat- ing point and can simultaneously capture load and generation variability will be useful. To address these issues, the ideas in [8]-[12] provide inspi- ration, but these works have limitations as well. The sampling- based approach in [8] and [9] suffer the same blindness to unsampled scenarios as with the metrics in [1]. Moreover, without appropriate scenario selection and scenario reduction methods, sampling-based approaches can easily become com- putationally prohibitive if the number of uncertain variables increase, which is likely in the future as power injections be- come much more controllable and flexible. Using the maxi- mum load shedding as indicator of grid adequacy is also use- ful when used in the proper context but it fails to capture the extent of network vulnerability. Moreover, the model assumes flexibility in generation dispatch and as such, masks a grid's inherent limitations. While [11] provides a useful framework for quantifying adequacy in terms of spaces instead of discrete scenarios, it is unclear how to operationalize this framework especially for a large number of uncertain variables. A visual comparison of feasible sets used as an illustration in [11] is possible in 2D space for two uncertain variables but is gener- ally difficult for more variables. And while the exploration of the relationship between graph properties and adequacy of power grids is interesting, the work is still in its seminal stag- es. In [12], the determinant of the susceptance matrix is used as a proxy for the volume of feasible power injection and no proposal was given to actually quantify the latter parameter. This makes it difficult to know how much grid adequacy im- proved and how much risk of congestion remains. In addition to the limitations identified, all works reviewed focus on characterizing and measuring adequacy As such, explicit measures of grid inadequacy, are lacking. Since char- acterizing grid adequacy does not necessarily give information about grid inadequacy, it is difficult to estimate the extent of risk of congestion and assess the efficacy of risk mitigation solutions without dedicated grid inadequacy metrics. The lack of grid inadequacy metrics also hinders the development of complementary optimization models that provide solutions that minimize this problem. To address the gaps identified, this work offers the follow- ing novel contributions: a. We present a framework that is specifically designed to characterize and measure inherent grid inadequacy to ac- commodate power injection diversity in future grids. b. The framework uses a previously unexplored way of characterizing grid inadequacy that is potentially useful in future research and practice. That is, given all system states within the whole power injection space, the relative size of the infeasible subspace is indicative of inherent grid inadequacy to accommodate power injection diversi- ty without intervention. c. Since the sets involved are generally high-dimensional and difficult to analyze, we present three approaches to circumvent this difficulty. One is a sampling-based ap- proach while the other two are novel dimension-reduction approaches. d. We present three inadequacy metrics that result from the the multi- in characterizing three approaches used dimensional sets. The advantages of the proposed framework and metrics complement the limitations of the different approaches re- viewed as follows. The metrics can capture the spatiotemporal diversity of both positive and negative power injections simul- taneously and can be computed using minimal network data including (1) bus connectivity, (2) line susceptances, (3) line loading limits, and (4) bus power injection limits. The metrics provide an overall measure of grid inadequacy that is not de- pendent on a given operating condition. The first set of metrics generalizes the sampling-based approach in [8] and [9] under the presented framework. As such, the first metric is very much dependent on the quality of the sample and can be blind outside of the identified states. The other two sets of metrics are original and does not have this dependency. The underly- ing assessment against the whole power injection space fits well with an ideal future grid that provides non-discriminatory access to all participants that accommodates a diversity of power exchanges. Moreover, we decouple contingency analy- sis and power flow control interventions in the operations modeling in order to highlight the inherent properties of the network. However, as with other metrics, the presented met- rics are only as good as the models and assumptions used. We will comment on these and identify potential applications. III. PROPOSED FRAMEWORK TO MEASURE GRID INADEQUACY A. Infeasible power system operation Power grids facilitate the exchange of electrical energy be- tween generators and loads through interconnecting transmis- sion lines. Feasible power exchanges (in the static, non- contingency, and non-intervention sense) must satisfy not only strict physical laws governing the power flows but also equipment and operational constraints including bus power injection, line power flow, and voltage magnitude constraints. A large infeasible power injection set means that constraints are violated for many power exchange schedules and is indica- tive of the need to deploy preventive of corrective measures. If possible, a network with a small infeasible power injec- tion set is preferred because this means that the network can accommodate diverse power injection scenarios on its own even without preventive or corrective interventions. However, characterizing the infeasible set using the non-linear AC pow- er flow model is generally complicated and computationally intensive [12]. The DC power flow model is a linear approxi- mation of the full AC model [13] and makes the analysis of feasible and infeasible sets simpler at the expense of model fidelity and accuracy. This model is ubiquitous in research and practice, in both the operations and infrastructure planning applications, evident in the literature reviewed in Section II. B. Infeasible set of the constrained DC power flow model In the DC power flow model, infeasible operation is main- ly characterized by line overloading. That is, a vector of bus 𝑠]T representing scenario s power injections Ps = [𝑃1 𝑠, … , 𝑃𝑛 𝑠, 𝑃2 is infeasible in grid ℊ with l lines and n buses if the following relations are satisfied except (1): 4 𝑓𝑎𝑏 𝑠,ℊ = 𝐵𝑎𝑏 𝑓𝑎𝑏 𝑠,ℊ ≤ 𝐶𝑎𝑏 ℊ (𝜃𝑎 𝑠,ℊ − 𝜃𝑏 𝑠,ℊ) ℊ 𝑠 = ∑ 𝑓𝑖𝑗 𝑃𝑖 𝑠,ℊ 𝑗∈𝒱ℊ,𝑗≠𝑖 = 0 𝑠 ∑ 𝑃𝑖 𝑖∈𝒱ℊ 𝑚𝑖𝑛 ≤ 𝑃𝑖 𝑃𝑖 𝑠 ≤ 𝑃𝑖 𝑚𝑎𝑥 ∀𝑎𝑏 ∈ ℰℊ ∀𝑎𝑏 ∈ ℰℊ ∀𝑖 ∈ 𝒱ℊ ∀𝑖 ∈ 𝒱ℊ (1) (2) (3) (4) (5) min and Pi where fab is the power flow and Cab is the capacity of line ab with end buses a and b, ℰ is the set of lines with cardinality l, Bab is the susceptance of line ab, θa is the voltage angle at bus a, Pi is the power injection at bus i, 𝒱 is the set of buses la- max are the power injection lim- belled from 1 to n, Pi its at bus i, and the superscripts indicate dependence in s and ℊ. Eq. (1) limits the power flow in a line within its capacity. Eqs. (2) and (3) are the DC power flow analog of Ohm's law and Kirchhoff's Current Law (KCL) respectively. Eq. (4) is the power balance equation and (5) keeps power injection var- iability within limits. Depending on the type of analysis, the power injection limits can represent various sources of uncertainty. They can capture long-term uncertainty in spatiotemporal load and gen- eration growth as well as short-term uncertainty in market dispatch, load flexibility and variability, energy storage charge and discharge capabilities, and renewable energy intermitten- cy. Detailed models can capture the interrelation between bus power injections and limit the spanned power injection space if data is available, observed trends are deemed to hold in the planning horizon, or market simulation tools are deemed relia- ble. Otherwise, interval sets provide a viable alternative in addition to giving information on infeasible operation that are not captured by historical or market-based models. C. Framework to measure grid inadequacy Any power injection vector satisfying (4)-(5) represents a valid power exchange. However, some vectors can be infeasi- ble due to network congestion, i.e. when at least one line is overloaded. Power flow control via topology, susceptance, or voltage angle control or power injection control via generation rescheduling or load curtailment can mitigate some congestion scenarios. The prevalence of grid congestion and the need to use preventive or corrective control strategies highlight inher- ent grid inadequacy to accommodate power injection diversity on its own without intervention. A way to quantify this inherent inadequacy is to compare the size of the following sets: • Set ℱ of valid power injections satisfying (4)-(5) and • Set ℐ of infeasible power injections satisfying (2)-(5) but fail to satisfy (1) for at least one line. The larger the size of Set ℐ relative to Set ℱ, the more inade- quate a grid inherently is. However, solving for the size of these sets is complicated especially for large n and l. Interpret- ing the constrained DC power flow model as a vector mapping from one space to another and projecting the resulting surfaces in lower dimension can circumvent this challenge. D. Visualizing the constrained DC power flow model The constrained DC power flow model can also be formu- lated using matrix operations as derived in detail in [14]. That is, a power injection scenario is feasible for a given grid if the following relations are met including (4): 𝒇𝑠,ℊ ≤ 𝑪ℊ 𝒇𝑠,ℊ = 𝑿ℊ𝑲ℊ𝜽𝑠,ℊ 𝑷𝑠 = 𝑩ℊ𝜽𝑠,ℊ 𝑷𝑚𝑖𝑛 ≤ 𝑷𝑠 ≤ 𝑷𝑚𝑎𝑥 (6) (7) (8) (9) where f is the vector of line power flows, C is the vector of line capacities, X is the diagonal matrix of line susceptances, K is the edge-to-bus incidence matrix, θ is the vector of bus voltage angles, P is the vector of bus power injections, B is the full DC power flow susceptance matrix, Pmin and Pmax are the vectors of bus power injection limits, and the superscripts de- note dependence in s and ℊ. Eqs. (6) and (9) are the matrix representation of (1) and (5) respectively while (7) and (8) are the matrix equivalent of (2) and (3) respectively. We can com- bine (7) and (8) to eliminate θ as follows: 𝒇𝑠,ℊ = 𝑿ℊ𝑲ℊ(𝑩ℊ)+𝑷𝑠 (10) where (B)+ is the pseudoinverse of B. Eq. (10) highlights the cause-and-effect relationship between P and f in terms of the grid topology and network parameters. This also allows us to conveniently interpret the constrained DC power flow model as a vector mapping as follows. Eq. (10) can be interpreted as a mapping of vector P in the n-dimensional power injection space, P-space, to the l- dimensional line power flow space, f-space, through the linear mapping function XKB+. This becomes a constrained mapping when combined with Eqs. (4), (6), and (9). Fig. 2 visualizes this mapping for a three-bus three-line grid. The susceptances are 1.0 p.u. for Line 1-2 and 0.5 p.u. for Lines 1-3 and 2-3. The power injection limits are 0.0 -- 2.0 p.u. for Bus 1 and -1.0 -- 0.0 p.u. for Buses 2 and 3. The line capacities are all 1.0 p.u. In Fig. 2, an n-dimensional hypercube defined by (9) bounds all possible power injection combinations in P-space. A hypersurface defined by (4) lies within the hypercube and characterizes Set ℱ of valid power exchanges. Linear mapping using (10) projects this hypersurface into f-space. Portions of the projection fall inside or outside an l-dimensional bounding hypercube defined by (6) in f-space. Power exchanges project- ed outside the bounding hypercube have one or more congest- ed lines and hence are infeasible. This characterizes Set ℐ. The larger the projection outside the f-space bounding hypercube, the larger the size of the infeasible power injection set and the more inadequate a grid inherently is to accommodate power injection diversity on its own. Traditional methods in designing adequate grids ensure that the point representing the peak load dispatch or points representing a number of power injection scenarios lie inside the f-space bounding hypercube, with or without employing power flow or power injection control. However, increased power injection diversity and increased demand for non- discriminatory access in future grids may require a paradigm shift. It will become increasingly important not just to ac- commodate a number of power injection points but instead accommodate a larger area of the power injection hypersur- face as economically as possible. The ideas in this paper help motivate this paradigm shift and will be useful future devel- opments in this field. 5 (a) (b) Fig. 2: Visualization of the constrained DC power flow model as a vector mapping: (a) a hypersurface represents the set of valid power injection scenarios in the power injection space, P-space, and (b) the vector projection of the hypersurface within the bounding hypercube in the line power flow space, f-space, determine feasibility. E. Visualizing the f-space surfaces in 1D Usually, projections of feasible spaces are made in P-space as was done in [2]-[7], [10]. This is generally complicated because the feasible range of power injection on a bus is de- pendent on the power injections at other buses. We present below a complementary approach that projects instead in f- space, a non-standard practice as far as the authors are aware. This is relatively much less convoluted and provides new in- sights on grid vulnerabilities and bottlenecks by identifying lines that are prone to congestion when power injections are not adequately scheduled. We return to the problem of project- ing in P-space in the next section. We can project the feasible and infeasible spaces in f-space along the f-space basis vectors to be able to visualize and ana- lyze the high-dimensional surfaces involved. The projection of the feasible space can be obtained by solving for the maxi- mum feasible line flows for each line in both directions while considering all line capacity limits. The projection of the whole power injection hypersurface can be obtained by solv- ing for the maximum line flows for each line in both directions while ignoring all line capacity limits. Superimposing these two sets of parameters will show the extent of the infeasible space, projected in terms of individual line power flows. The following mathematical models can be used to solve these parameters. The projection of the power injection hypersurface is spanned by the maximum unconstrained line flows in a line in both directions, 𝑓max_𝑢,𝑎𝑏 and 𝑓max_𝑢,𝑏𝑎, with all line capacity constraints ignored. The difference determines a line's uncon- strained loading range. The following linear program in θ, P, and f solves for 𝑓max_𝑢,𝑎𝑏. A similar linear program can be used to solve for 𝑓max_𝑢,𝑏𝑎. Objective Subject to 𝑓max_𝑢,𝑎𝑏 = max 𝑓𝑎𝑏 Eqs. (2)-(5) (11a) (11b) On the other hand, the projection of the feasible subspace of the power injection hypersurface is spanned by the maxi- mum feasible line flows in a line in both directions, 𝑓max_𝑓,𝑎𝑏 and 𝑓max_𝑓,𝑏𝑎, with all line capacity constraints considered. The difference determines a line's feasible loading range. The following linear program, along with a similar program, can be used to solve for these parameters. Objective Subject to 𝑓max_𝑓,𝑎𝑏 = max 𝑓𝑎𝑏 Eqs. (1)-(5) (12a) (12b) After solving for the maximum unconstrained and feasible line flows, we can superimpose the loading ranges and line capacities in a line loading range diagram. Fig. 3 shows this for the three-bus network in Section II-D. Red outlines repre- sent the line capacities. White areas represent excess line ca- pacity and remains unused unless the power injection limits change, the network is modified by design or by accident, or non-injection power flow control strategies are implemented. The black areas identify infeasible operation. Fig. 3: A line loading range diagram showing the one-dimensional visualization of the f-space bounding hypercube (outlined areas) and hypersurface (shaded areas) for the three-bus three-line network. The presence of black areas outside the outlined areas indicate infeasible operating states. The presence of infeasible areas does not necessarily mean that the network cannot satisfy peak or maximum loading conditions. Rather, the infeasible areas indicate that some loading states served with a given set of generator dispatch, which may or may not correspond to system peak conditions, may overload a given line. There may or may not exist a set of dispatch able to serve peak demand. As such, the proposed procedure in this section is not meant to measure adequacy to meet peak demand but rather to highlight grid inability to serve a diversity of dispatch scenarios. The procedures in the next section can answer the question of adequacy during peak loading and other loading conditions. In the three-bus example, there is ample capacity in Lines 1-3 and 2-3 to cover all possible diversity of power injection scenarios within bus power injection limits. However, capacity constraints in Line 1-2 prohibit the dispatch of some power injection scenarios. The prevalence of the infeasible projec- tions is indicative of network inability to accommodate power injection diversity and the need to deploy preventive or correc- tive interventions. Moreover, the infeasible areas indicated by 6 this procedure can be used to motivate more detailed analyses. For example, the families of power injection scenarios that lead to infeasible states and the likelihood of these states oc- curring merit further exploration. Since it is the first time that this approach is presented in the context of grid inadequacy assessment, the authors hope that future work will find other suitable uses for it. F. Visualizing the P-space surfaces in 1D As we mentioned in Section III-E, projecting the P-space surfaces to a lower dimension to show the infeasible set is challenging. Previous works such as [2]-[7], [10] have tried various approaches along this direction and have their own set of capabilities and limitations. One limitation is the depend- ence on an operating point. This is as expected because the feasible range of power injection in one bus is dependent on the power injections in other buses. In this section, we present an approach that projects in a space related to P-space. The choice of a different space removes the complications of de- pendence on individual bus power injection. This is not meant to replace existing methods however and is intended to pro- vide a complementary approach that provides useful infor- mation in a different form. The approach we use is to project along the total network loading vector, Lt, given by the sum of the loads at each bus. The idea is to find thresholds of total network loading that trigger congestion instead of finding allowable power injection ranges as is done in [2]-[7], [10]. Under the DC power flow model assumptions, we can find Lt -- and Lt+ such that Lt -- is a network loading threshold below which congestion is not pos- sible regardless of the generation schedule and Lt+ is the max- imum network loading that the grid can accommodate without congestion through at least one generation schedule. We can solve for Lt -- in two steps: (1) solving for Lt,ab, the minimum load that triggers congestion of line ab in direction ab and (2) solving for the minimum of the values obtained in Step 1 for all lines in both the forward and reverse power flow directions. That is, 𝐿𝑡− = min{𝐿𝑡−,𝑎𝑏 ∀𝑎𝑏 ∈ ℰ ∗} (13a) where ℰ ∗ is the set of line indices for both forward and reverse power flow directions. Here, each 𝐿𝑡−,𝑎𝑏 is calculated by solv- ing the following linear program: Objective Subject to 𝐿𝑡−,𝑎𝑏 = min 𝐿𝑡 𝑓𝑎𝑏 ≥ 𝐶𝑎𝑏 Eqs. (2)-(5) Likewise, we can solve for Lt+ as follows: Objective Subject to 𝐿𝑡+ = max 𝐿𝑡 Eqs. (1)-(5) (13b) (13c) (13d) (14a) (14b) After solving for the relevant points, we can make the one- dimensional visualization in Fig. 4 where we can also super- impose the maximum total load, Lmax. This diagram visualizes the network loading ranges that are congestion-free wherein Lt < Lt -- and congestion-prone wherein Lt > Lt -- . A subset of the latter set is congestion-positive for Lt > Lt+ if Lt+ < Lmax. This approach provides grid adequacy and inadequacy in- formation in terms of congestion-free, congestion-prone, and congestion-positive network loading ranges. For the three-bus example, operating below Lt -- = 1.5 p.u. guarantees no conges- tion while operating above Lt+ = 1.75 p.u. guarantees conges- tion regardless of generation schedule under DC power flow model assumptions. Since Lmax = 2.0 p.u. exceeds Lt+, there is no feasible generation schedule for the maximum load. These information can motivate further studies that explore the cause of these thresholds as well as help quantify inherent grid inad- equacy to serve loads. For example, a low value of Lt -- relative to Lmax is indicative of (1) a grid that needs a lot of operational intervention to serve demand or (2) a selective grid that only allows non-discriminatory access up to a certain loading level. Fig. 4: One-dimensional visualization of the P-space surfaces using the total network loading. IV. PROPOSED MEASURES OF GRID INADEQUACY This section extends the framework developed in Section III to define metrics of grid inadequacy. We present one sam- pling-based metric and two metrics derived from the projec- tion of the feasible and infeasible sets in lower dimension. A. Scenario Pool Infeasibility Ratio (SPIR) We can define a representative scenario pool ΩS of T pow- er injection vectors 𝑷𝑠1, 𝑷𝑠2, … , 𝑷𝑠𝑇 to represent the set of valid power exchanges. Each vector represents a power injection scenario that satisfy (4) and (5). System state selection re- quires a balance between adequate space representation and computational tractability. It is a challenging problem on its own and hence is placed outside the scope of this work. Some of the common approaches include the following. If data is available and trends are assumed to hold in the planning hori- zon, probability distribution and correlation models may be used to generate scenarios. Market simulation can also be used if there are reliable datasets and models. Otherwise, combina- toric sampling or worst-case sampling become viable options. After defining the scenario pool, we can evaluate each scenario using DC power flow for a given network topology ℊ and solve for the number of line overloading instances per scenario NLs, the number of line overloading instances in the scenario pool NL, and the number of scenarios with congestion NS as follows: 𝑠,ℊ = ∑ {𝑢(𝑓𝑎𝑏 𝑠,ℊ − 𝐶𝑎𝑏 ℊ ) + 𝑢(−𝑓𝑎𝑏 𝑠,ℊ − 𝐶𝑎𝑏 ℊ )} 𝑁𝐿𝑠 𝑎𝑏∈ℰℊ 𝛺𝑆,ℊ = ∑ 𝑁𝐿𝑠 𝑠,ℊ 𝑁𝐿 𝑠∈𝛺𝑆 𝛺𝑆,ℊ = ∑ 𝑢(𝑁𝐿𝑠 𝑠,ℊ) 𝑁𝑆 𝑠∈𝛺𝑆 1 (15) 1 (16) 1 (17) where u(x) is a function that returns 1 if x > 0 and returns zero otherwise and the superscripts show the dependence on s, ℊ, and ΩS. Using NS and T, we can then define the Scenario Pool Infeasibility Ratio (SPIR) for a given grid and scenario pool as follows: 7 𝑆𝑃𝐼𝑅𝛺𝑆,ℊ = 𝛺𝑆,ℊ 𝑁𝑆 𝑇𝛺𝑆 1 (18) This metric captures the proportion of power exchanges in the scenario pool that is infeasible. Using the sets defined in Sec- tion III-C, we use NS to measure Set ℐ and T to measure Set ℱ. The more representative the scenario pool is of the actual P-space hypersurface, the better the approximation given by SPIR of the relative sizes of Sets ℱ and ℐ. If the assessment is made using the same scenario pool and between grid topologies with the same buses, we can use NS to compare grid alternatives directly. Furthermore, since NL is related to NS as implied in (15)-(17), NL is also useful to com- pare options. Dividing NL with the number of line flow as- sessments given by T·l gives the Scenario Pool Line Over- loading Ratio (SPLOR) as follows: 𝑆𝑃𝐿𝑂𝑅𝛺𝑆,ℊ = 𝛺𝑆,ℊ 𝑁𝐿 𝑇𝛺𝑆 · 𝑙ℊ 1 (19) SPIR is a generalization of the ad hoc approaches in [8] and [9] to the whole power injection space. It shares the same underlying framework with the ENLC metric from [1] but counts grid congestion states instead of load curtailment states under the grid inadequacy assessment framework outlined in Section III. As such, it is a flexible metric such that power flow control and contingency analysis can also be integrated in the computation by replacing the basic DC power flow model with more appropriate model. However, as with all other sam- pling-based metrics, SPIR is highly dependent on the quality of the selected states and is potentially blind to other states outside this sample. The next two metrics do not require the identification of discrete system states and can be used to complement SPIR. B. Congestion-prone Network Loading Ratio (CPNLR) Using the notations in Section III-F, we can use Lt -- and Lmax to define the Congestion-free Network Loading Ratio (CFNLR) as follows: 𝐶𝐹𝑁𝐿𝑅ℊ = ℊ 𝐿𝑡− 𝐿𝑚𝑎𝑥 1 (20) This metric gives the proportion of the maximum network loading that is always free from congestion under the DC power flow model assumptions. Similarly, the ratio of Lmax -- Lt -- and Lmax gives the Congestion-prone Network Loading Ratio (CPNLR) as follows: 𝐶𝑃𝑁𝐿𝑅ℊ = 1 − 𝐶𝐹𝑁𝐿𝑅ℊ = ℊ 𝐿𝑚𝑎𝑥 − 𝐿𝑡− 𝐿𝑚𝑎𝑥 1 (21) This metric gives the proportion of maximum network loading that may result in congestion if the power injections are not properly scheduled. Using the sets defined in Section III-C, we use Lmax to give a measure for Set ℱ and Lmax -- Lt -- for Set ℐ in P-space. A limitation of CPNLR is that it overestimates grid inade- quacy while CFNLR underestimates adequacy. Since operat- ing the grid beyond Lt -- may or may not result in network con- gestion depending on the chosen generation schedule, Lmax -- Lt -- gives an inflated measure of Set ℐ. As such, CPNLR also gives an inflated measure of grid inadequacy. Regardless, CPNLR is still useful in revealing the extent of grid neediness for careful monitoring and potential interventions as a result of not being able to serve power exchanges on its own. An ideal value of zero means that all loading levels can be served by any combination of generator dispatches without necessary intervention. However, this only guarantees adequacy from a non-contingency perspective under DC power flow assump- tions. It must be used with its limitations in mind and only in an appropriate context with other decision criteria considered. C. Total Line Overloading Ratio (TLOR) The visualization in f-space in Section III-E estimates the extent of infeasible operating ranges of each line given the variability in power injection in the buses. From the visualiza- tion in Fig. 3, we can define the Infeasible Line Loading Mar- gin (ILLM), Unconstrained Line Loading Range (ULLR), and Infeasible Line Loading Ratio (ILLR) for each line as follows: 𝐼𝐿𝐿𝑀𝑎𝑏 ℊ = 𝑓max_𝑢,𝑎𝑏 ℊ 𝑈𝐿𝐿𝑅𝑎𝑏 ℊ − 𝑓max_𝑓,𝑎𝑏 ℊ = 𝑓max_𝑢,𝑎𝑏 ℊ ℊ + 𝑓max_𝑓,𝑏𝑎 ℊ − 𝑓max_𝑢,𝑏𝑎 ℊ − 𝑓max_𝑢,𝑏𝑎 𝐼𝐿𝐿𝑅𝑎𝑏 ℊ = ℊ 𝐼𝐿𝐿𝑀𝑎𝑏 ℊ 𝑈𝐿𝐿𝑅𝑎𝑏 1 (22) 1 (23) 1 (24) For a given line ab, ILLMab gives the difference between the maximum unconstrained line flow and feasible line flow, ULLRab gives the difference between the maximum uncon- strained line flows in both directions, and ILLRab gives the ratio of the two. We can then define the Total Infeasible Line Loading Ratio (TILLR) as a measure of overall grid inadequacy. It uses the ILLM and ULLR of all the lines to compute a metric like ILLR for the whole system as follows: 𝑇𝐼𝐿𝐿𝑅𝑎𝑏 ℊ = ℊ ∑ 𝐼𝐿𝐿𝑀𝑎𝑏 𝑎𝑏∈ℰℊ ℊ ∑ 𝑈𝐿𝐿𝑅𝑎𝑏 𝑎𝑏∈ℰℊ 1 (25) Using the sets defined in Section III-E, we use the total ULLR as a measure of Set ℱ and the total ILLM as a measure of Set ℐ. As with CPNLR, this representation is not exact be- cause of the limitations in the low-dimensional projections and the redundancy in accounting for the projected spaces. Never- theless, TILLR gives a measure of grid inadequacy that can be useful indicators of the extent of network vulnerabilities. The individual ILLRs give the proportion of the infeasible loading range relative to the total unconstrained loading range of a line. This can be used as an indicator of line adequacy if the value is zero or an indicator for overloading and network bottleneck otherwise. While a similar physical interpretation cannot be made for TILLR, it can be seen as the extent of addi- tional power injection diversity that the grid can accommodate 8 only if there are no line capacity constraints. As with CPNLR, the ideal value of TILLR is zero. If this is the case, no lines in the network will ever get congested even without intervention if power injection limits are satisfied. This guarantee, likewise, is limited to the non-contingency case under DC power flow assumptions. Additional checks are required to assess adequa- cy and inadequacy outside the scope of the framework pre- sented in this paper. V. ILLUSTRATIVE EXAMPLES A. 6-bus test system The relevant data for the 6-bus test system is available in [15]. Of the six buses, three have generators and five have loads. There are a total of thirteen installed lines. To construct the scenario pool for SPIR evaluation, we use a combinatoric sampling approach since it is tractable for this test system. We sample T = 3,139,831 power injection scenar- ios that represent eleven negative power injection levels at 0%, 10%, …, 100% of maximum and eleven positive power injection levels at 0%, 5%, 15%, …, 95% of maximum ±5% to balance out the demand. The scenarios represent power injec- tions from traditional loads and generators as well as power injections from renewables, dynamic loads, and energy stor- age. We specifically chose not to exceed the 100% threshold to illustrate the potential for network congestion even if the power injections remain within limits. Fig. 5 visualizes the distribution of the six-dimensional power injection scenario pool in reduced 3D space obtained using principal component analysis (PCA) [15]. The axes are the top three principal components. Points corresponding to the scenarios with and without congestion are interspersed with one another, making it difficult to gauge the relative sizes of the infeasible and feasible sets. This challenge reinforces the utility of the proposed inadequacy metrics in quantifying the relative sizes of these sets. Fig. 5: Three-dimensional visualization of the 6-bus test system infeasible set using a scenario pool (T = 3,139,831) and PCA. SPIR is equal to 33.1% implying that a third of the sampled scenarios resulted in network congestion. The infeasible sce- narios can be analyzed further to determine the family of pow- er injection scenarios that congest the network, explore the likelihood of such scenarios, and then devise appropriate pre- ventive or corrective interventions. CPNLR is equal to 82.1% with Lt -- = 135.7 MW and Lmax = Lt+ = 760 MW. This means that operating the grid above Lt = 135.7 MW has potential for congestion and requires coordinated scheduling of bus power injections. Since Lmax = Pt+, there is at least one generation schedule that can serve the maximum total load. In a way, the grid is "needy" since a large proportion of network loading requires careful monitoring and potential operational interven- tion. Likewise, without adequate power flow control interven- tions, the grid can also be seen as inherently selective from a market perspective because it discriminates against some schedules in serving a particular loading level. Finally, TILLR is equal to 27.1%. Fig. 6 shows how each line contributes to this value. The diagram shows that only Line 2-4 will never become overloaded regardless of the pow- er injection distribution. The diagram also identifies lines that limit a grid's ability to accommodate diverse power injection states. This can serve as basis for later analyses on what kinds of power injection distributions are affected by grid bottle- necks and what remedial action can be made. Fig. 6: Line loading range diagram of the 6-bus test system. B. 118-bus test system We perform a similar analyses using the 118-bus test sys- tem using the data in [16]. The grid has 118 buses and 186 lines, with generation at 54 buses and loads at 91 buses. Since defining discrete load and generation levels to define the scenario pool is computationally prohibitive, we sample from the line loading space, including low and intermediate loading scenarios as well as worst case scenarios that result in maximum line loading. We define ten loading levels for each line by dividing each line's unconstrained loading range into equal parts. We then solve for the power injection scenarios that result in each line loading level Fab at minimum network loading as follows: Objective Subject to min 𝐿𝑡 𝑓𝑎𝑏 = 𝐹𝑎𝑏 Eqs. (2)-(5) (26a) (26b) (26c) This process yielded T = 1,860 scenarios that represent a spec- trum of network loading conditions. Around forty percent of the sampled scenarios result in congestion with SPIR = 43.4%. CPNLR is equal to 94% with Lt -- = 222.9 MW and Lmax = Lt+ = 3,733 MW. A low value of Lt -- relative to Lmax indicates that congestion is expected even at very low loading levels if there is no coordinated generation 9 scheduling. This is indicative of the grid's inherent selectivity in serving demand unless power flow control options are available. Since Lt = Lt+, there should be at least one feasible generation schedule able to meet the maximum demand. How- ever, this schedule may not necessarily have the least cost or the most renewable energy utilization rate. TILLR is equal to 31.5%. Fig. 7 shows the top ten lines with the largest ILLMs. These results provide indicators of inherent grid inadequa- cies, vulnerabilities, and bottlenecks that merit further explora- tion and the subsequent design of preventive and corrective interventions, if deemed necessary. Fig. 7: Line loading range diagram of the 118-bus test system show- ing the top ten lines with the largest ILLM. VI. REMARKS, FUTURE WORK, AND POTENTIAL USE-CASES As we discussed in Section II-A, grid adequacy assessment requires assumptions on network and operations modeling and system state selection. The level of fidelity of the models used and the comprehensiveness of system states represented de- pends on the accuracy and tractability requirements in the de- sired use-case. The proposed approach chooses a specific set of assumptions that defines its capabilities and limitations. We comment on these in this section and identify interesting fu- ture work and potential use cases. A. Remarks on grid and operations modeling assumptions In the proposed framework, the DC power flow model is used to assess grid inadequacy in a non-contingency and strict- ly non-intervention setting. As with the other works reviewed that uses the DC power flow model, the results are subject to the model's inherent limitations. Despite its limitations, how- ever, this model remains widely-used in recent work such as in [7] in either operations planning or infrastructure planning setting. It will be interesting future work to look at counter- parts to the metrics proposed using other power flow models and then compare the results. Likewise, using a non-contingency and non-intervention approach means that the resulting inadequacy metrics does not reflect the full capabilities of a grid to manage power ex- changes in both normal and contingency conditions. What the metrics provide are indicators of inherent grid inadequacy, vulnerabilities, and bottlenecks to accommodate power injec- tion diversity on its own. These indicators are useful in gaug- ing the extent of necessary intervention the grid requires and the inherent readiness of a grid for non-discriminatory access. As such, the proposed metrics are not meant to replace other adequacy and inadequacy metrics but rather provide comple- mentary information. Adequacy and inadequacy assessment considering operational interventions and contingency scenar- ios will still be necessary. Future work can look at ways to incorporate operational interventions and contingency model- ing in the proposed framework. Possible approaches for the latter include deterministic N-1 analysis or probabilistic risk assessment. B. Remarks on system state selection approach In terms of system state selection, the proposed metrics us- es two approaches. The first metric uses a sampling-based approach to identify discrete system states. This is similar to the approaches used in [1], [8], and [9] but generalized to the whole power injection space. As with other approaches that use this framework, it is a challenge to identify systems states that best represent future states while keeping the size tracta- ble. Scenario selection and scenario reduction methods can be used for these types of cases and will remain relevant areas of research in the future. However, planners must remember that such an approach can be blind to system states not included in the samples. As such, additional analysis must be made to explore these blind spots, assess their likelihood, and design appropriate interventions. The approach used in the two other metrics helps in this regard. The other two metrics implicitly considers all system states using interval sets that represent the full power injection space. The approach used is similar to those in [7] and [11] using loadability and feasibility sets but further developed into tractable metrics tailored specifically for inadequacy assess- ment. While this approach is conservative, it complements sampling-based approaches by providing insights on potential blind spots not captured in sampling-based approaches. Using interval sets also becomes particularly useful when historical power injection distribution data is not readily available, his- torical trends are deemed inapplicable in the planning horizon, market simulation models are deemed unreliable, or when planners would like to explore grid inadequacy against previ- ously unobserved trends. These conditions are especially ap- plicable in an expansion planning setting with increased un- certainty that we comment on next. C. Potential use cases Adequacy assessment using sampling-based metrics such as SPIR are very flexible and can reflect use-case requirements by changing the network and operation model used as well as the system state selection process. As such, it can easily find potential applications in adequacy assessment in either opera- tions planning or expansion planning provided that the appro- priate set of assumptions are applied. The other two metrics, CPNLR and TILLR, are more spe- cialized because of the underlying models used. These metrics tests for inadequacy against all possible power injection sce- narios which may be impractical in some cases. In the opera- tions setting for example, hour-ahead and day-ahead system states are much more likely to resemble previously observed system states from the recent past. As such, the level of varia- bility of future states against which adequacy or inadequacy must be assessed is much more limited. In this case, adequacy assessment against detailed distribution and correlation mod- els, market simulation models, or tighter interval sets are more appropriate. Testing against potential blind spots is still neces- 10 sary but not in the scale of exploring the whole power injec- tion space. On the other hand, power injections years or decades ahead in an expansion planning setting are much more uncer- tain and diverse. This is especially true today when data and models on how future market players owning one or a combi- nation of renewable generation, dynamic loads, and energy storage are not yet available. In this application, the assump- tions underlying the CPNLR and TILLR metrics become viable preliminary options until more detailed data and simulation models become available. Even if detailed models become available, these can have blind spots and it will still be neces- sary to scan the whole power injection space for potential vul- nerabilities. Since no model will be accurate enough to predict the long-term future with certainty, information on potential grid vulnerabilities that the proposed framework and metrics provide will allow planners to see the network's limitations and subsequently devise appropriate interventions to hedge against likely problematic future scenarios. In this regard, future work can look at the development of decision support tools that use the metrics presented in identi- fying grid expansion designs that are robust against power injection diversity in future grids. A suitable application is in initial stages of expansion planning where a small subset of interesting grid expansion plans needs to be identified from a large combinatoric pool. One approach is to use a greedy heu- ristic where the metrics are minimized by adding one line at a time. More sophisticated approaches would involve optimiza- tion using mathematical or metaheuristic models that try to find local or global minima. In such applications, other screen- ing tools can be used, each with its own set of capabilities and limitations, to generate a subset of interesting grid designs. The shortlisted options can then be subject to more rigorous multi-criteria assessment including PQ balancing studies, con- gestion management, contingency response, cost-benefit anal- ysis, market simulations, stability analysis, and robustness assessment against different realizations of uncertain parame- ters. And as new information becomes available, existing plans can be reassessed or the whole process repeated alto- gether. VII. CONCLUSION The integration of more intermittent renewable sources, dynamic loads, and energy storage on top of a competitive market environment will require a non-discriminatory grid that can accommodate diverse power injection states. This work presents a new framework and set of metrics that measure grid inadequacy to accommodate increasing levels of power injec- tion diversity. Such a framework and set of metrics comprise fundamental building blocks in assessing and designing future grids. The framework uses the infeasible set of the constrained DC power flow model to quantify grid inadequacy against power injection diversity. Since it is complicated to compute the actual size of the high-dimensional sets, we present three approaches to give a measure to the sets involved. One uses a sampling-based approach while the other two uses novel di- mension reduction techniques. We then use these ideas to formulate three inadequacy metrics that describe different but complementary grid inadequacy properties. Illustrative exam- ples using the 6- and 118-bus test systems show that the met- rics can be used as indicators of grid readiness to accommo- date diverse operating states as well as indicators of grid vul- nerabilities and bottlenecks. Like other metrics, the proposed metrics capture a limited aspect of grid inadequacy assessment defined by the network and operations model used as well as the system state selec- tion process chosen. We discuss the implications of the choice of assumptions made and made a case for the use of the pro- posed framework and metrics in an expansion planning set- ting. We identified interesting areas for future work including the use of more detailed power flow models in developing similar metrics, appropriate scenario selection and scenario reduction techniques to balance space representation and mod- el tractability, integration of operational interventions and con- tingency analysis in the proposed framework, and develop- ment of decision-support tools for identifying candidate grid expansion plans. We close by noting that the problem of grid adequacy as- sessment considering power injection diversity against more intermittent renewable generation, dynamic loads, and energy storage is timely and important. The framework and metrics presented provide a new and different approach to assessing grid inadequacy that is potentially useful in future research and practice. We identified a lot of interesting areas for future work that we hope can inspire other researchers to study this problem. REFERENCES R. Billinton and W. Li, Reliability Assessment of Electric Power Systems Using Monte Carlo Methods. New York: Springer Science + Business Media, 1994. J. Zhu, R. Fan, G. Xu, and C. S. Chang, "Construction of maximal steady-state security regions of power systems using optimization method," Electr. Power Syst. Res., vol. 44, pp. 101 -- 105, 1998. C. C. Liu, "A new method for the construction of maximal steady- state security regions of power systems," IEEE Trans. Power Syst., vol. 1, no. 4, pp. 19 -- 26, 1986. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] 11 Z. Jizhong, Optimization of power system operation, vol. 53, no. 9. 2013. A. Capasso, A. Cervone, M. C. Falvo, R. Lamedica, G. M. Giannuzzi, and R. Zaottini, "Bulk indices for transmission grids flexibility assessment in electricity market: A real application," Int. J. Electr. Power Energy Syst., vol. 56, pp. 332 -- 339, 2014. J. Wu, U. Schlichtmann, and Y. Shi, "On the Measurement of Power Grid Robustness Under Load Uncertainties," in 2016 IEEE International Conference on Smart Grid Communications (SmartGridComm), 2016, pp. 218 -- 223. A. A. Jahromi and S. Member, "On the Loadability Sets of Power Systems -- Part I :," IEEE Trans. Power Syst., vol. 32, no. 1, pp. 137 -- 145, 2017. H. Yu, C. Y. Chung, and K. P. Wong, "Robust transmission network expansion planning method with Taguchi's orthogonal array testing," IEEE Trans. Power Syst., vol. 26, no. 3, pp. 1573 -- 1580, 2011. R. A. Jabr, "Robust transmission network expansion planning with uncertain renewable generation and loads," IEEE Trans. Power Syst., vol. 28, no. 4, pp. 4558 -- 4567, 2013. P. Wu, H. Cheng, and J. Xing, "The Interval Minimum Load Cutting Problem in the Process of Transmission Network Expansion Planning Considering Uncertainty in Demand," vol. 23, no. 3, pp. 1497 -- 1506, 2008. J. Li, F. Liu, Z. Li, C. Shao, and X. Liu, "Grid-side flexibility of power systems in integrating large-scale renewable generations : A critical review on concepts , formulations and solution approaches," Renew. Sustain. Energy Rev., vol. 93, no. March, pp. 272 -- 284, 2018. F. B. Thiam and C. L. Demarco, "Transmission expansion via maximization of the volume of feasible bus injections," Electr. Power Syst. Res., vol. 116, pp. 36 -- 44, 2014. A. J. Wood and B. F. Wollenberg, Power Generation, Operation, and Control, 2nd ed. John Wiley and Sons Inc., 2012. H. Cetinay, F. A. Kuipers, and P. Van Mieghem, "A Topological Investigation of Power Flow," IEEE Syst. J., vol. 12, no. 3, pp. 2524 -- 2532, 2018. J. Lever, M. Krzywinski, and N. Altman, "Principal component analysis," Nat. Publ. Gr., vol. 14, no. 7, pp. 641 -- 642, 2017. "Index of data Illinois Institute of Technology." [Online]. Available: http://motor.ece.iit.edu/data/JEAS_IEEE118.doc. [Accessed: 20- Jun-2018].
1911.05919	1	1911	2019-11-14T03:23:21	Complementary metasurfaces for guiding electromagnetic waves	[ "physics.app-ph" ]	Waveguides are critically important components in microwave, THz, and optical technologies. Due to recent progress in two-dimensional materials, metasurfaces can be efficiently used to design novel waveguide structures which confine the electromagnetic energy while the structure is open. Here, we introduce a special type of such structures formed by two penetrable metasurfaces which have complementary isotropic surface impedances. We theoretically study guided modes supported by the proposed structure and discuss the corresponding dispersion properties. Furthermore, we show the results for different scenarios in which the surface impedances possess non-resonant or resonant characteristics, and the distance between the metasurfaces changes from large values to the extreme limit of zero. As an implication of this work, we demonstrate that there is a possibility to excite two modes with orthogonal polarizations having the same phase velocity within a broad frequency range. This property is promising for applications in leaky-wave antennas and field focusing.	physics.app-ph	physics	Complementary metasurfaces for guiding electromagnetic waves 1School of Electronics and Information, Northwestern Polytechnical University, 710129 Xi'an, China 2Laboratory of Wave Engineering, ´Ecole Polytechnique F´ed´erale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland X. Ma1, M. S. Mirmoosa2,3, and S. A. Tretyakov3 9 1 0 2 v o N 4 1 ] h p - p p a . s c i s y h p [ 1 v 9 1 9 5 0 . 1 1 9 1 : v i X r a 3Department of Electronics and Nanoengineering, Aalto University, P.O. Box 15500, FI-00076 Aalto, Finland (Dated: November 15, 2019) Waveguides are critically important components in microwave, THz, and optical technologies. Due to recent progress in two-dimensional materials, metasurfaces can be efficiently used to design novel waveguide structures which confine the electromagnetic energy while the structure is open. Here, we introduce a special type of such structures formed by two penetrable metasurfaces which have complementary isotropic surface impedances. We theoretically study guided modes supported by the proposed structure and discuss the corresponding dispersion properties. Furthermore, we show the results for different scenarios in which the surface impedances possess non-resonant or resonant characteristics, and the distance between the metasurfaces changes from large values to the extreme limit of zero. As an implication of this work, we demonstrate that there is a possibility to excite two modes with orthogonal polarizations having the same phase velocity within a broad frequency range. This property is promising for applications in leaky-wave antennas and field focusing. I. INTRODUCTION Planar structures with periodically arranged subwave- length elements, coined as metasurfaces [1 -- 5], have at- tracted considerable attention in the last decade (see e.g. [6 -- 15]). One important application of such artificial surfaces is in controlling surface waves [16 -- 19], including a single metasurface [1] or a metasurface over a grounded slab [20, 21]. This guiding property occurs since we can engineer the surface impedance of such structures [22 -- 25]. Analogous to these planar waveguides, new designs for open waveguide structures can be introduced, which are formed by two parallel and penetrable metasurfaces separated by a finite distance [26]. Such open waveguides confine the electromagnetic energy while the correspond- ing fields are attenuated away from the structure (note that these guiding structures can be intriguingly invisible under plane-wave illumination [13]). An interesting spe- cial case of those waveguides is the set of two penetrable metasurfaces which are Babinet-complementary. This structure is intriguing since the product of the surface impedances of two complementary sheets must be nondis- persive and equal to η2 0/4, where η0 is the impedance of the background isotropic medium (free space, in this work). Moreover, in the limit of zero distance between the sheets, they combine into a continuous PEC sheet. In this paper, we study this special case of two comple- mentary sheets and investigate the corresponding guiding waves. For microwave waveguides, complementary inclusions are realized by interchanging metal and vacuum regions for a given planar structure, assuming the perfect electric conductor model for metals. Incorporation of such inclu- sions have been studied and applied also in planar waveg- uide structures [27 -- 30]. However, earlier works studied planar waveguide structures which contain complemen- tary inclusions only in one single layer [31 -- 35]. Here, we instead impose the condition that the two metasurfaces are complementary with respect to each other. It is worth noting that some attention has been paid to the scatter- ing characteristics of two parallel complementary sheets placed in close proximity of each other [36 -- 39]. However, to the best of our knowledge, there is not enough knowl- edge on eigenmodes of coupled complementary metasur- faces operating as a single waveguiding structure. The proposed guiding structure is shown in Fig. 1(a). Two complementary metasurfaces are separated by the distance d, and the space between them is filled by air (vacuum). The surface impedances of the two metasur- faces are denoted as Zs1 and Zs2, respectively, and we assume that these values do not depend on the spatial coordinates in the sheet planes. Based on the Babinet principle, we have Zs1 · Zs2 = η2 0 4 . (1) We orient the z-axis along the direction of the wave prop- agation. Here, we classify the proposed structure into two different categories: non-resonant and resonant struc- tures. For each category, we analyze the corresponding guided modes and study the extreme case when the dis- tance between the two metasurfaces tends to zero. Fur- thermore, we also reveal a possibility of exciting two modes of different polarizations propagating with the same phase velocity (degeneracy state). The paper is organized as follows: Sections II and III study two non-resonant and resonant structures. Sec- tion IV illustrates the polarization insensitivity, and fi- nally Section V concludes the paper. II. NON-RESONANT DISPERSIVE IMPEDANCE SHEETS In this category, the surface reactances are far from res- onances (we assume a lossless structure). In other words, 2 (b) (c) (a) FIG. 1. (a) -- The structure under study: Two complementary impedance sheets. (b) -- Equivalent circuit model for two non- resonant dispersive impedance sheets. (c) -- Equivalent circuit model for two resonant dispersive impedance sheets. they are corresponding to the reactance of a single capac- itance or inductance. If the two metasurfaces, shown in Fig. 1(a), are complementary to each other, one of the metasurfaces should be inductive and the other meta- surface should be capacitive. Under this condition, we keep the product of the two surface impedances constant (canceling out the frequency ω) and satisfy the Babinet principle. As a consequence, there are two modes propa- gating along the waveguide and having transverse mag- netic (TM) and transverse electric (TE) polarizations. Note that one single metasurface can support guided waves (surface waves) that possess only transverse elec- tric or transverse magnetic polarization [1, 40]. How- ever, in waveguides which consist of two parallel meta- surfaces [26], there are always two simultaneous modes whose polarizations depend on the surface impedances. Two inductive (capacitive) sheets correspond to guiding of two TM modes (TE modes), and one inductive and one capacitive sheet correspond to guiding of co-existing TM and TE modes. The equivalent circuit model of two non-resonant dispersive impedance sheets is shown in Fig. 1(b). Practical realization of such structure can be in form of dense meshes of metallic strips for the in- ductive sheet and arrays of small metallic patches for the capacitive sheet. Let us denote the effective inductance of the inductive sheet by L, and the effective capacitance of the other sheet as C. From Eq. 1, we can immediately see that those effective parameters are related to each other as (see also Eq. (A4) of Appendix): C = 4L η2 0 . (2) Based on the above expression, the dispersion relation of the guided modes can be written in terms of L only: α2 0 (1 − e−2αd) + α 2L η2 0 = ω2(µ0 + 2αL), (3) for the TM-polarized wave and 0 = ω2(cid:104) η2 α2 0 + α 2L (cid:105) µ0(1 − e−2αd) + 2αL , (4) for the TE-polarized wave. Here α denotes the field at- tenuation constant for fields outside the waveguide. Con- templating Eq. 4, we see that there is a cut-off frequency for the TE wave, which is associated with the distance between the two metasurfaces. Studying the limit of α approaching zero, by using l'Hopital's rule we find the cut-off frequency: fcut−off = η0 4π 1(cid:112)L(L + µ0d) . (5) As an example, taking L = 2.5 nH, we can read- ily find the corresponding value of C in according to Eq. (2). Suppose that β is the phase constant along the wave propagation direction. Figure 2 shows the an- alytically calculated dispersion curves (frequency versus β) for different distances between the two non-resonant sheets. Concurring with expectations, both TM and TE modes can simultaneously propagate along such waveg- uide. When d = 10 mm, the TE and TM modes have ap- proximately the same phase velocity within a certain fre- quency range, which we will explain later in Section IV. 3 (a) (b) (c) (d) FIG. 4. The distribution of the electric field of two non- resonant sheets. Here, (a) TM polarization, d = 1 mm, β = 117.4 1/m, (b) TM polarization, d = 0.00001 mm, β = 274 1/m, (c) TE polarization, d = 1 mm, β = 167.7 1/m, and (d) TE polarization, d = 0.00001 mm, β = 153.1 1/m. The operational frequency is f = 2.7 GHz for TM modes and f = 5 GHz for TE modes. (a) and (b) are the longitudinal component of the electric field. The vertical and horizontal axes are y- and z-axes, respectively. The metasurface posi- tions are shown by solid black lines. Figs. 5 and 6 show the surface current density. For the TM mode, the operating frequency is assumed to be 2.7 GHz and two different distances are considered. For the TE modes, we assume that the operating frequency is 5 GHz. In Fig. 4, note that the top metasurface is induc- tive and the bottom one is capacitive. Also, the two black solid lines show the positions of the two metasurfaces. One can see that for the TM polarization the field is strongly tied to the inductive sheet and attenuates as the vertical distance from that sheet increases, while the field is strongly confined to the capacitive sheet for the TE polarization. When the two sheets get close to each other, strong coupling can be observed from the phe- nomenon that the magnitude of electric field does not change too much between the two sheets because of the small distance. The surface current corresponding to the TE-polarized mode flows along the x-axis, while it flows along the z-axis for the TM polarization. It is perceived that the unequal magnitude of surface current density on the two sheets can be attained due to the asymmet- ric impedances. Also, as Fig. 6 shows, the TM-polarized mode is characterized by out-of-phase surface currents. In contrast, for the TE-polarized mode, the surface cur- rents are in phase. FIG. 2. Dispersion curves of two non-resonant sheets, when the distance between the sheets is d = 10 mm, d = 1 mm, and d = 0.00001 mm. FIG. 3. Dispersion curves for two non-resonant sheets in which the distance between two sheets tends to zero. Decreasing the distance causes that the dispersion curve corresponding to the TE mode (red line) moves coun- terclockwise towards the light line at high frequencies and the curve corresponding to the TM mode (blue line) moves clockwise becoming far away from the light line. Interestingly, the limit of d → 0 leads to the appearance of a new resonance frequency for TM mode, which can be analytically found as fmix ≈ 1 2π 1√ LC . (6) III. RESONANT DISPERSIVE IMPEDANCE SHEETS This new resonance is exactly the same as the cut-off frequency of the TE-polarized wave (Eq. (5)) when d = 0. The dispersion curve for this extreme case is illustrated in Fig. 3. In order to get better physical insight into the elec- tromagnetic behavior of the waveguide structure, Fig. 4 plots the electric field distribution, and the following In this category, the waveguide under study consists of metasurfaces having resonant properties. In other words, the surface impedance of each metasurface can be expressed as an equivalent impedance of a series or parallel connection of an inductance and capacitance. The unit-cell size of the structure is much smaller than 0500100015002000 [1/m]05101520/2 [GHz]TM(d=10mm)TE(d=10mm)TM(d=1mm)TE(d=1mm)TM(d=0.00001mm)TE(d=0.00001mm)Light Line4 (a) (b) (c) (d) FIG. 5. Magnitude of surface currents for two non-resonant sheets. Here, (a) TM polarization, d = 1 mm, β = 117.4 1/m, (b) TM polarization, d = 0.00001 mm, β = 274 1/m, (c) TE polarization, d = 1 mm, β = 167.7 1/m, and (d) TE polarization, d = 0.00001 mm, β = 153.1 1/m. The operational frequency is f = 2.7 GHz for TM modes and f = 5 GHz for TE modes. (a) (b) (c) (d) FIG. 6. Phase of surface currents for two non-resonant sheets. Here, (a) TM polarization, d = 1 mm, β = 117.4 1/m, (b) TM polarization, d = 0.00001 mm, β = 274 1/m, (c) TE polarization, d = 1 mm, β = 167.7 1/m, and (d) TE polarization, d = 0.00001 mm, β = 153.1 1/m. The operational frequency is f = 2.7 GHz for TM modes and f = 5 GHz for TE modes. the free-space wavelength (at the resonant frequency), which allows the structure to be modeled by homoge- nized impedances. From the circuit theory point of view, we explicitly know that the equivalent impedance of the series connection of an inductance and capacitance is ca- pacitive and inductive below and above the resonance frequency, respectively. On the contrary, the equivalent impedance of the parallel connection is inductive below the resonance frequency and capacitive above it. Hence, following these considerations, one of the metasurfaces must be realized as a series connection and the other one as a parallel connection of two reactances, as shown in Fig. 1(c). Here, the surface impedances are characterized by effective inductances and capacitances denoted as L1, C1 and L2, C2 for each metasurface. Let us assume that L1 and C1 are in series connection and L2, C2 are in parallel connection. These values are obtained from the full-wave solution of a plane-wave reflection problem in the quasi-static limit [41, 42] for self-resonant structures. In Fig. 1(c), we should notice that the corresponding res- onance frequencies of two complementary metasurfaces must be the same in order to realize the opposite reac- tances at all frequencies. For the two complementary metasurfaces, the impedance of one metasurface is determined by the other metasurface impedance using the Babinet principle (the relations can be found in Eq. (B4) of Appendix). Ac- cordingly, the dispersion relation for two complementary -0.500.5z [mm]2345678910Magnitude of surface current density [A/m]10-31st layer2nd layer-0.500.5z [mm]2345678910Magnitude of surface current density [A/m]10-31st layer2nd layer-0.500.5z [mm]2345678910Magnitude of surface current density [A/m]10-31st layer2nd layer-0.500.5z [mm]2345678910Magnitude of surface current density [A/m]10-31st layer2nd layer-0.500.5z [mm]-100-50050100Phase of surface current density [Deg]1st layer2nd layer-0.500.5z [mm]-100-50050100Phase of surface current density [Deg]1st layer2nd layer-0.500.5z [mm]-100-50050100Phase of surface current density [Deg]1st layer2nd layer-0.500.5z [mm]-100-50050100Phase of surface current density [Deg]1st layer2nd layer5 (a) (b) (c) FIG. 7. Dispersion curves of two resonant sheets, when the distances between the sheets is (a) d = λ6GHz, (b) d = λ6GHz/50, and (c) d = λ6GHz/5000. resonant sheets can be written in terms of L1 and C1 as (cid:16) 1 − e−2αd(cid:17) C1α2 + + (4 + η2 0C1 2L1 )0α + (cid:16) 1 − e−2αd(cid:17) 20α + (cid:17)(cid:16) ω (cid:17)4− (cid:17)2 C1α2(cid:105)(cid:16) ω ω0 0µ0 L1 (cid:16) 20α + (cid:104) 0µ0 L1 ω0 = 0, (7) for the TM polarization. Analogously, the dispersion re- lation for the TE polarization reads (cid:104) (cid:16) (cid:17)4− 1 − e−2αd(cid:17)(cid:105)(cid:16) ω (cid:105)(cid:16) ω (cid:17)2 ω0 )µ0α + η2 0C1α2) ω0 µ2 0 L1 η2 0C1 2L1 2µ0α + η2 (cid:16) 0C1α2 + 1 − e−2αd(cid:17) (cid:104) µ2 0 L1 2µ0α + + (4 + = 0, (8) FIG. 8. Dispersion curves for two resonant sheets when the distance between the two sheets tends to zero. where ω0 is the resonant angular frequency of both se- ries and parallel connections. Below/above the resonance frequency, there are always two modes with TE and TM polarizations. Regarding the TM polarization, one of them suffers from a cut-off frequency due to the reso- nance. For the TE polarization, both of them have cut- off frequencies which can be calculated by (see Eq. (B5) of Appendix) fcut−off ≈ f0 2 ± √(cid:52) L1 + η2 0 C1 8 + µ0d L1 + µ0d , (9) (cid:115) 2ω0 )2 + ( η2 0 C1 8 + µ0d where (cid:52) = ( η0 2 )2. As an example, we assume f0 = 6 GHz as the resonance frequency of resonant metasurfaces, and additionally L1 = 10 nH and C1 = 0.07 pF. The values of L2 and C2 are ob- tained by employing Eq. (B4) of Appendix, which gives L2 = 2.5 nH and C2 = 0.28 pF. Figure 7 presents the dispersion curves for different distances between the two resonant sheets. As expected, one TE-polarized mode and one TM-polarized mode exist simultaneously be- low/above the resonance frequency. When d = λ6GHz, the TE and TM modes have approximately the same phase velocity. However, as the distance decreases, the modes are being separated. The limit of d → 0 brings about two new resonance frequencies fmix1 and fmix2 which are given by (cid:118)(cid:117)(cid:117)(cid:116) η2 0C1 + 8L1 −(cid:113)(cid:0)η2 (cid:118)(cid:117)(cid:117)(cid:116) η2 (cid:113)(cid:0)η2 0C1 + 8L1 + 0C1 8C1L2 1 (cid:1)2 (cid:1)2 0C1 8C1L2 1 fmix1 = fmix2 = 1 2π 1 2π + 16η2 0C1L1 , (10) + 16η2 0C1L1 . (11) The dispersion curve of the extreme case is classified by fmix1, f0 and fmix2, which are illustrated in Fig. 8. IV. POLARIZATION INSENSITIVITY Metasurface-based waveguides have been exploited for the applications of leaky-wave radiation and field focus- 02004006008001000 [1/m]024681012141618/2 [GHz]1st TM mode2nd TM mode1st TE mode2nd TE modeLight Line02004006008001000 [1/m]024681012141618/2 [GHz]1st TM mode2nd TM mode1st TE mode2nd TE modeLight Line02004006008001000 [1/m]024681012141618/2 [GHz]1st TM mode2nd TM mode1st TE mode2nd TE modeLight Line6 However, according to the first equation of the paper, the multiplication of the surface impedances must be equal to η2 0/4. Therefore, in the above equation, the expres- sion inside the bracket should be unity meaning that the distance between the two metasurfaces must be consid- erable compared to the operating wavelength. This was explicitly shown in Figs. 2 and 7. When the distance be- tween the two metasurfaces is not small, the dispersion curves, corresponding to both modes, match with each other (they have the same phase velocity meaning that the power is transferred regardless of the polarization). For practical realization, small square metal patches and their complementary grids are chosen as metasur- faces topology. For different distances, numerical sim- ulations are done using CST Microwave Studio. The simulation setting is similar to the simulation setting for electromagnetic band gap structures (in which after many numerical simulations and comparing the obtained results with analytical and experimental considerations, a practical rule for choosing background material and boundary setting is found for the correct computation of surface wave dispersion diagram [46]). Regarding the simulations, the key parameter in the computer model is the height of the air space outside the waveguide which emulates infinite free space outside of the structure (in "Background Setting" this space is "Upper Z-distance" and "Lower Z-distance"). Based on numerous computer simulations, an airbox with the height of more than 10 times of the distance between two metasurfaces has to be placed over the unit cell to maintain high accuracy. Figures 9 and 10 show the dispersion curves of com- plementary metasurfaces with 10 mm and 1 mm unit- cell sizes, respectively. In order to examine the accuracy of the results, in those figures the simulated results are compared with the analytical results. As it is seen, they are in good agreement. One can also see that the dis- persion curves of the TM and TE modes overlap in a broad frequency range for larger distances between the two metasurfaces. V. CONCLUSION In this paper, we have investigated a waveguiding structure formed by two complementary metasurfaces. We have studied supported guided modes and derived dispersion equations for non-resonant and resonant dis- persive impedance sheets. Furthermore, we have inves- tigated the extreme scenarios when the two complemen- tary sheets get close to each other and provided physical insight into the behavior of surface modes. In the limit of zero distance between the two sheets, two new resonance frequencies emerge and four modes are classified by three resonance frequencies. In addition, we have theoretically studied the condition on the sheet impedance to achieve the property of supporting both TM and TE modes with the same phase velocity. Waveguides with such charac- teristic have great potential to be used in many applica- (a) (b) FIG. 9. Dispersion curves of practical realization for the square metal patches and their complementary grid, when the distance between two sheets is: (a) d = 10 mm and (b) d = 50 mm. Here the length of square patch is 9.8 mm, the width of grid is 0.2 mm and period of unit cell is 10 mm. ing [43, 44]. However, the limitations of applications ex- ist due to polarization sensitivities. Designing a surface- wave waveguide which is insensitive to polarization be- comes a significant challenge. Different kinds of waveg- uide topologies have been explored to propagate waves regardless of polarization at certain frequencies [45]. Let us define the polarization insensitivity as following: Both TM-polarized and TE-polarized waves can propagate in- side the waveguide having the same phase velocity (de- generacy condition). Regarding our proposed waveguide structure, we can find this interesting property based on the dispersion relations written for TE and TM modes. For each value of α, the value of ω is calculated from the general dispersion relations. Subsequently, it is enough to ensure that the solutions are equal to each other (see Appendix C) in order to find the condition which pro- vides us with the property of polarization insensitivity. After some algebraic manipulations, we find the following condition: Zs1 · Zs2 = η2 0 4 (1 − e−2αd). (12) 020406080100120140160 [1/m]0123456/2 [GHz]TM(Analytical)TE(Analytical)TM(Simulated)TE(Simulated)Light Line020406080100120140160 [1/m]0123456/2 [GHz]TM(Analytical)TE(Analytical)TM(Simulated)TE(Simulated)Light Line(a) (b) FIG. 10. Dispersion curves of practical realization for the square metal patches and their complementary grid, when the distance between two sheets is: (a) d = 1 mm and (b) d = 5 mm. Here the length of square patch is 0.9 mm, the width of grid is 0.1 mm and the period is 1 mm. Appendix A: Capacitance and inductance of non-resonant sheets tions like holographic surfaces or leaky-wave antennas. µ2 0 C1 ω2 + 4α2L2 + 2µ0α = 0, 7 in terms of the sheet impedances (Zs1 and Zs2) as [1] Zs1 · Zs2 = η2 0 4 . (A3) From the above equation, the relation between L and C for complementary non-resonant sheets is written as C = 4L η2 0 , (A4) Substituting Eq. (A4) into Eqs. (A1) and (A2), we can readily find the dispersion relations for waves along two complementary sheets. Appendix B: Sheet capacitance and inductance of resonant metasurfaces The dispersion equation of a waveguide formed by one series-connection metasurface and one parallel- connection metasurface can be expressed as [26] (cid:104) (cid:104) ω4− 0C1L1L2 20αC1C2L1L2 + 42 ω2 + 20α + α2(cid:16) (cid:105) 20α(C1L1 + C2L2 + C1L2) + α2(cid:16) (cid:105) 1 − e−2αd(cid:17) 1 − e−2αd(cid:17) (cid:16) 2µ0αC1C2L1L2 + µ2 0 for TM-mode waves and 42 0L2 (cid:104) (cid:104) (cid:16) 1 − e−2αd(cid:17) (cid:105) 1 − e−2αd(cid:17) C1 = 0, (cid:105) ω4− L2C1C2 4α2C1L1L2 + 2µ0α(C1L1 + C2L2 + C1L2)+ C1C2L2+ (B1) (B2) for TE-mode waves. The series-connection impedance can be represented as Zs1 = jωL1 + 1/jωC1 and parallel- connection impedance is written as Zs2 = jωL2/(1 − ω2L2C2). Two complementary resonant sheets must have the same resonant angular frequency denoted as ω0 = 1√ L1C1 = 1√ L2C2 . (B3) In our previous work [26], the dispersion relations of a parallel-metasurface waveguide formed by one inductive grid and one capacitive grid are given as + α2(1 − e−2αd) = ω2(cid:104) 42 20α C for the TM polarization and = ω2(cid:104) 4α2L C + 2µ0α C 2µ0αL + µ2 , (A2) 0L C , + 20αL (cid:105) (cid:105) 0(1 − e−2αd) for the TE polarization. In this case, the impedances of the two sheets read Zs1 = jωL and Zs2 = 1/jωC, respectively. Since the two sheets are complementary to each other, we can use the Babinet principle which reads (A1) According to the Babinet principle (Eq. (A3)), the effective inductance and capacitance of the parallel- connection impedance are related to the inductance and capacitance of series-connection impedance as L2 = η2 0 4 C1, C2 = 4L1 η2 0 (B4) Substituting Eq. (B4) into Eqs. (B1) and (B2), the dis- persion relations for complementary sheets can be ob- tained. 0200400600800100012001400 [1/m]0102030405060/2 [GHz]TM(Analytical)TE(Analytical)TM(Simulated)TE(Simulated)Light Line0200400600800100012001400 [1/m]0102030405060/2 [GHz]TM(Analytical)TE(Analytical)TM(Simulated)TE(Simulated)Light Line(cid:115) Interestingly, both TE-polarized modes suffer from cut-off frequencies when the attenuation constant α ap- proaches zero, which can be approximately evaluated as (cid:104) + 2µ0d ±(cid:112)(cid:52)(cid:105) fcut−off ≈ f0 1 4(L1 + µ0d) 4L1 + η2 0C1 2 0 C1 (B5) where (cid:52) = (4L1 + η2 1−16µ0dL1, which can be simplified as (cid:52) = 4η2 0 C1 2 + 2µ0d)2. Ob- viously, the expression of Eq. (B5) under the two square roots on the right-hand side are always non-negative, which results in two real solutions. 2 +2µ0d)2−16L2 0L1C1 + ( η2 Appendix C: Condition for achieving the same phase velocity of both modes Based on the definition of phase velocity, the ratio of the angular frequency and the phase constant should be the same at each value of α for TE- and TM-polarized waves when the phase velocities of TE- and TM-polarized waves are same. We denote this ratio by ωTE βTE = ωTM βTM = A(α). (C1) Here, ωTE(TM) and βTE(TM) are the angular frequencies and the phase constants for the TE and TM polariza- tions, respectively. Based on the relation between ω, α, [1] S. A. Tretyakov, Analytical Modeling in Applied Electro- magnetics (Artech House, Massachusetts, 2003). [2] C. L. Holloway, E. F. Kuester, J. A. Gordon, J. O'Hara, J. Booth, and D. R. Smith, An overview of the theory and applications of metasurfaces: The two- dimensional equivalents of metamaterials, IEEE Anten- nas Propag. Mag. 54, 10 (2012). [3] S. B. Glybovski, S. A. Tretyakov, P. A. Belov, Y. S. Kivshar, and C. R. Simovski, Metasurfaces: From microwaves to visible, Phys. Rep. 634, 1 (2016). [4] H.-T. Chen, A. J. Taylor, and N. Yu, A review of meta- surfaces: Physics and applications, Rep. Prog. Phys. 79, 076401 (2016). [5] V. S. Asadchy, A. D´ıaz-Rubio, and S. A. Tretyakov, Bianisotropic metasurfaces: Physics and applications, Nanophotonics 7, 1069 (2018). [6] N. Yu, P. Genevet, M. A. Kats, F. Aieta, J.-P. Tetienne, F. Capasso, Z. Gaburro, Light propagation with phase discontinuities: Generalized laws of reflection and refrac- tion, Science 334, 333 (2011). [7] Y. Yang, W. Wang, P. Moitra, I. Kravchenko, D. P. Briggs, and J. Valentine, Dielectric meta- reflectarray for broadband linear polarization conver- sion and optical vortex generation, Nano Lett. 14, 1394 (2014). I. [8] M. Decker, D. N. Neshev, I. Staude, M. Falkner, J. Dominguez, and I. Brener, T. Pertsch, (cid:3), (cid:3), 1 A(α)2 − 1 c2 1 A(α)2 − 1 c2 α2 = ω2 TE α2 = ω2 TM 8 (C2) (C3) (cid:2) (cid:2) (cid:16) and β for TE and TM polarizations, we can find that where c is the speed of light. Comparing the above two equations, the expressions in square brackets on the right- hand side are the same for each value of α. Therefore, it can be concluded that the angular frequencies of waves of TE and TM polarizations at each value of α should be the same. In [26], the dispersion relations for the waveguides which have arbitrary sheet impedance have been derived: α2(cid:16) (cid:17) e−2αd − 1 = j2ω0 Zs1 + Zs2 (cid:17) α − 4ω22 0Zs1Zs2, (C4) for TM polarization and 4Zs1Zs2α2 + j2ωµ0(Zs1 + Zs2)α + ω2µ2 0(e−2αd − 1) = 0, (C5) for TE polarization. At each value of α, the value of ω calculated from Eqs. (C4) and (C5) should be the same. Using this condition, after some algebraic manipulations, we come to Eq. (12). Y. S. Kivshar, High-efficiency dielectric Huygens' surfaces, Adv. Opt. Mater. 3, 813 (2015). [9] N. M. Estakhri and A. transformation Phys. Rev. X 6, 041008 (2016). with gradient Al´u, Wave-front metasurfaces, [10] V. S. Asadchy, M. Albooyeh, S. N. Tcvetkova, A. D´ıaz- Rubio, Y. Ra'di, and S. A. Tretyakov, Perfect control of reflection and refraction using spatially dispersive meta- surfaces, Phys. Rev. B 94, 075142 (2016). [11] X. Wang, A. D´ıaz-Rubio, V. S. Asadchy, G. Ptitcyn, A. A. Generalov, J. Ala-Laurinaho, and S. A. Tretyakov, Extreme asymmetry in metasurfaces via evanescent fields Angular-asymmetric absorption, Phys. Rev. Lett. 121, 256802 (2018). engineering: [12] A. M. H. Wong and G. V. Eleftheriades, Perfect anoma- lous reflection with a bipartite huygens' metasurface, Phys. Rev. X 8, 011036 (2018). [13] F. S. Cuesta, V. S. Asadchy, A. D. Sayanskiy, V. A. Lenets, M. S. Mirmoosa, X. Ma, S. B. Glybovski, and S. A. Tretyakov, Non-scattering metasurface-bound cavities for field localization, enhancement, and suppres- sion, arXiv:1802.00070, 2018. [14] M. S. Mirmoosa, G. A. Ptitcyn, V. S. Asadchy, and S. A. Tretyakov, Time-varying reactive elements for extreme accumulation of electromagnetic energy, Phys. Rev. Appl. 11, 014024 (2019). [15] F. Liu, O. Tsilipakos, A. Pitilakis, A. C. Tasolam- prou, M. S. Mirmoosa, N. V. Kantartzis, D.-H. Kwon, J. Georgiou, K. Kossifos, M. A. Antoniades, M. Kafe- saki, C. M. Soukoulis, and S. A. Tretyakov, Intelli- gent metasurfaces with continuously tunable local sur- face impedance for multiple reconfigurable functions, Phys. Rev. Appl. 11, 044024 (2019). [16] S. N. Tcvetkova, D.-H. Kwon, A. D´ıaz-Rubio, and S. A. Tretyakov, Near-perfect conversion of a propagat- ing plane wave into a surface wave using metasurfaces, Phys. Rev. B 97, 115447 (2018). [17] S. Sun, Q. He, S. Xiao, Q. Xu, X. Li, and L. Zhou, Gradient-index meta-surfaces as a bridge linking propagating waves and surface waves, Nature Mater. 11, 426 (2012). [18] D. R. Smith, O. Yurduseven, L. P. Mancera, P. Bowen, and N. B. Kundtz, Analysis of a waveguide-fed metasur- face antenna, Phys. Rev. Applied 8, 054048 (2017). [19] A. Vakil and N. Engheta, Transformation optics using graphene, Science 332, 1291 (2011). [20] A. M. Patel and A. Grbic, A printed leaky-wave an- tenna based on a sinusoidally-modulated reactance sur- face, IEEE Trans. Antennas Propag. 59, 2087 (2011). [21] F. Elek, B. B. Tierney, and A. Grbic, Synthesis of tensor impedance surfaces to control phase and power flow of guided waves, IEEE Trans. Antennas Propag. 63, 3956 (2015). [22] H. J. Bilow, Guided waves on a planar tensor impedance surface, IEEE Trans. Antennas Propag. 51, 2788 (2003). [23] A. A. Oliner and A. Hessel, Guided waves on sinusoidally- IRE Trans. Anten- surfaces, modulated reactance nas Propag. 7, 201 (1959). [24] R. G. Quarfoth and D. F. Sievenpiper, Nonscattering waveguides based on tensor impedance surfaces, IEEE Trans. Antennas Propag. 63, 1746 (2015). [25] R. Quarfoth and D. F. Sievenpiper, Artificial tensor impedance surface waveguides, IEEE Trans. Antennas Propag. 61, 3597 (2013). [26] X. Ma, M. S. Mirmoosa, and S. A. Tretyakov, Parallel- plate waveguides formed by penetrable metasurfaces, arXiv:1901.07940, 2019. [27] T. Zentgraf, T. P. Meyrath, A. Seidel, S. Kaiser, H. Giessen, C. Rockstuhl, and F. Lederer, Babinet's prin- ciple for optical frequency metamaterials and nanoanten- nas, Phys. Rev. B 76, 033407 (2007). [28] Q. Cheng, H. F. Ma, and T. J. Cui, Broadband planar Luneburg lens based on complementary metamaterials, Appl. Phys. Lett. 95, 181901 (2009). [29] G. Kumar, A. Cui, S. Pandey, and A. Nahata, Planar terahertz waveguides based on complementary split ring resonators, Opt. Express 19, 1072 (2011). [30] A. Bitzer, A. Ortner, H. Merbold, T. Feurer, and M. Walther, Terahertz near-field microscopy of com- plementary planar metamaterials: Babinet's principle, Opt. Express 19, 2537 (2011). [31] Y. D. Dong, T. Yang, and T. Itoh, Substrate integrated waveguide loaded by complementary split-ring resonators 9 and its applications to miniaturized waveguide filters, IEEE Trans. Microwave Theory Tech. 57, 2211 (2009). [32] N. Landy, J. Hunt, and D. R. Smith, Homogeniza- tion analysis of complementary waveguide metamate- rials, Photonics and Nanostructures-Fundamentals and Applications 11, 453 (2013). [33] L. Pulido-Mancera, P. T. Bowen, M. F. Imani, N. Kundtz, and D. Smith, Polarizability extraction of complementary metamaterial elements in waveguides for aperture modeling, Phys. Rev. B 96, 235402 (2017). [34] D. Gonz´alez-Ovejero, E. Martini, and S. Maci, Sur- supported by metasurfaces with self- IEEE Trans. Antennas face waves complementary geometries, Propag. 63, 250 (2015). [35] D. J. Bisharat and D. F. Sievenpiper, Guiding waves along an infinitesimal line between impedance surfaces, Phys. Rev. Lett. 119, 106802 (2017). [36] D. S. Lockyer, J. C. Vardaxoglou, and R. A. Simpkin, Complementary frequency selective surfaces, IEE Proc.- Microw. Antennas Propag. 147, 501 (2000). [37] S. S. Bukhari, W. G. Whittow, J. C. Vardaxoglou, and S. Maci, Dipole-slot-dipole metasurfaces, IET Microw. Antennas Propag. 10, 1384 (2016). [38] S. S. Bukhari, W. G. Whittow, J. C. Vardaxoglou, for coupled IEEE Trans. Antennas and S. Maci, Equivalent circuit model complementary metasurfaces, Propag. 66, 5308 (2018). [39] A. D´ıaz-Rubio, X. Ma, V. Asadchy, and S. Tretyakov, Resonances and embedded eigenstates in shifted double arrays of strips, 13th European Conference on Antennas and Propagation (EuCAP 2019), pp. 141-143, Krakow, Poland, 1-5 April, 2019. [40] R. E. Collin and F. J. Zucker, Antenna Theory Part 2 (McGraw-Hill, New York, 1969). [41] I. Anderson, On the theory of self-resonant grids, Bell Syst. Tech. J. 54, 1725 (1975). [42] D. Ferreira, R. F. S. Caldeirinha, I. Cuinas, and T. R. Fernandes, Square loop and slot frequency selective surfaces study for equivalent circuit model optimization, IEEE Trans. Antennas Propag. 63, 3947 (2015). [43] A. M. Patel and A. Grbic, Effective surface impedance of a printed-circuit tensor impedance surface (PCTIS), IEEE Trans. Microwave Theory Tech. 61, 1403 (2013). [44] C. Rusch, J. Schafer, H. Gulan, P. Pahl, and T. Zwick, Holographic mmW-antennas with TE0 and TM0 surface wave launchers for frequency-scanning FMCW-radars, IEEE Trans. Antennas Propag. 63, 1603 (2015). [45] M. Li, S. Xiao, J. Long, and D. F. Sievenpiper, Surface waveguides supporting both TM mode and TE mode with the same phase velocity, IEEE Trans. Antennas Propag. 64, 3811 (2016). [46] F. Yang and Y. Rahmat-Samii, Electromagnetic Band Gap Structures in Antenna Engineering (Cambridge Uni- versity, New York, 2009).
1811.00259	1	1811	2018-11-01T06:52:23	Effect of sputter deposited ZnO thin layers on magnetoimpedance response and field sensitivity of Co-based micro-wires	[ "physics.app-ph", "cond-mat.mes-hall" ]	Semiconductive ZnO has shown to be a fascinating element for application in high-performance sensors. ZnO with various thicknesses (87-500 nm) was deposited on the surface of Co-based amorphous micro-wire (Co68.15Fe4.35Si12.5B15) using magnetron sputtering technique and the magnetoimpedance (MI) response was evaluated. The MI% increases monotonously up to the 240 nm thickness of the ZnO layer. By further increase in the ZnO layer thickness the MI% decreases. MI response and field sensitivity increased from 227% and 1.7 %/Oe for bare micro-wire up to 406% and 4.99 %/Oe for ZnO deposited micro-wire with a thickness of 240 nm. The x-ray diffraction (XRD) and scanning electron microscopy (SEM) were used to characterize the structural grown ZnO layers. Vibrating sample magnetometry (VSM) was used to reveal the role of the ZnO thin layer in the MI response. The same trend of increase and decrease in the MI% by thickness was observed for the transverse permeability of the samples determined by VSM. The obtained results address a simple way to achieve high MI response and sensitivity with selective surface sensing.	physics.app-ph	physics	Effect of sputter deposited ZnO thin layers on magnetoimpedance response and field sensitivity of Co-based micro-wires A. Dadsetan 1, M. Almasi Kashi 1, 2, , S.M. Mohseni 3, M. R. Hajiali3 1 Institute of Nanoscience and Nanotechnology 87317, University of Kashan, Kashan, Iran 2 Department of physics, University of Kashan, Kashan, Iran 3 Faculty of physics, Shahid Beheshti University, Tehran 19839, Iran Abstract Semiconductive ZnO has shown to be a fascinating element for application in high performance sensors. ZnO with various thicknesses (87-500 nm) was deposited on the surface of Co-based amorphous micro-wire (Co68.15Fe4.35Si12.5B15) using magnetron sputtering technique and the magnetoimpedance (MI) response was evaluated. The MI% increases monotonously up to the 240 nm thickness of ZnO layer. By further increase in the ZnO layer thickness the MI% decreases. MI response and field sensitivity increased from 227% and 1.7 %/Oe for bare micro-wire up to 406% and 4.99 %/Oe for ZnO deposited micro-wire with a thickness of 240 nm. The x-ray diffraction (XRD) and scanning electron microscopy (SEM) were used to characterize the structural grown ZnO layers. Vibrating sample magnetometry (VSM) was used to reveal the role of the ZnO thin layer in the MI response. The same trend of increase and decrease in the MI% by thickness was observed for the transverse permeability of the samples determined by VSM. The obtained results address a simple way to achieve high MI response and sensitivity with selective surface sensing. Corresponding author's email address: [email protected] 1. Introduction Magnetoimpedance (MI) effect consists of a significant change in the impedance (both real and imaginary parts) of a soft magnetic conductor under the influence of an external magnetic field [1,2]. This effect is very important due to its capabilities to be used for technological applications such as magnetic field sensors [3]. The MI effect has been studied in different forms of soft magnetic materials such as ribbons, films, micro-wires, and their nanocrystalline counterparts [4,5]. MI is correlated with the skin depth (δ), δ = (ρ/πμtf)1/2 of the high frequency f current and the transverse magnetic permeability μt of metallic ferromagnet with electric resistivity ρ. μt changes by applying external field and this results in a new current skin depth and so varying the MI response [2]. Magnetic properties of soft magnetic materials and MI effect will be influenced by coating of a magnetic or conducting or insulating layer on bare samples [6 -- 12]. A controlled engineering of surface of a soft ferromagnetic ribbon has proved its useful controllability in enhancing the MI effect and so its practical application in magnetic sensors. Toward this, we can refer to some examples of dielectric materials, which are Zinc oxides [8], Copper oxide [9] diamagnetic organic thin film [10], CoFe2O4 [11] and very recently graphene oxide [12]. These papers have discussed the influence of the surface structure on the magnetic field and surface domain structure and so tuning the MI response. Also the reports of the conductive magnetic layers coated on the surface of MI sensors [13] have discussed on the magnetoimpedance exchange coupling. Very recently we have proposed that impedance spectroscopy can be used for the detection of spin orbit torque resulting from the spin hall effect in ribbon/Pt [14] and ribbon/IrMn heterostructures [15]. Therefore, surface sensitivity of the MI effect especially when the skin effect is sensitive against any small variations is significant. Besides, annealing and applying tensile stress to ribbons towards improvement of sensitivity is another interesting subject discussed by many authors [16 -- 18]. The studied heterostructure in this article is made of a ferromagnetic Co-based micro-wires and a thin layer of semiconductor ZnO. Among all oxide semiconductor materials, ZnO has been considered as an important element in technological devices due to its wide applications in photodetectors [19], solar cells [20], biosensors or gas sensors [21,22], and spintronic devices [23]. Furthermore, room temperature ferromagnetism has been found in undoped ZnO [24 -- 26] in the forms of thin films, which is attributed to oxygen defects, especially singly ionized oxygen vacancies. There are different deposition techniques used for the preparation of ZnO thin films such as laser deposition [27], sol-gel [28], magnetron sputtering [29], chemical vapor deposition [30] and electrodeposition [31]. Among these, sputtering offers advantages such as low temperature of fabrication while yielding preferred crystalline orientation and uniform properties. Based on only one study, ZnO thin film was deposited on a magnetic ribbon using SILAR method that resulted in 95.37% improvement (8 to 15.63%) in the MI effect [8]. The thickness for the deposition of ZnO layer was within micrometer range which resulted in a low MI effect and further accomplishment in this study is required. In this paper, we systematically study the effect of different thicknesses of sputter deposited ZnO thin films on the MI response and field sensitivity of co-based micro-wires. The highest MI ratio of 406% and the highest field sensitivity of 4.99 %/Oe are obtained at ZnO thickness of 240 nm. Vibrating sample magnetometry (VSM) hysteretic plots and consequently derived permeabilities are successively used to interpret the effect of ZnO thin film layers on MI response. Results are considerable in development of highly sensitive MI sensors. 2. Experimental Amorphous micro-wires with the nominal composition of Co68.15Fe4.35Si12.5B15 (42 mm in length, effective length of 40 mm, and 180 μm in diameter) were used the samples were initially cutting into suitable dimension and then were immersed in Dimethyl sulfoxide polishing solution and sonicated for 30 minutes. After that, they were immersed in a beaker containing distilled water for 15 minutes. Finally, the samples were baked in an oven for 20 minutes in a temperature of 60 °C to dry and be ready for the deposition process. ZnO layer was deposited on the cleaned Co-based micro-wire samples using magnetron sputtering technique. The base pressure and argon pressure was 1.5×10-7 and 1.5×10-2 mbar, respectively. ZnO layer was deposited using a ZnO target (99.999%) with thicknesses ranging from 87 to 500 nm, as shown schematically in Fig.1. X-ray diffraction (XRD) was performed using a STADI STOE diffractometer with CuKα (λ = 1.54 Å) for the angle (2θ) range from 10 -- 80°. Transverse magnetization curves of the samples were measured using a VSM (500 MDK). To measure the MI response of samples, an external magnetic field produced by a solenoid applied along the micro-wire axis and the impedance was measured by means of the four-point probe method. An ac current passed through the longitudinal direction of the micro-wire with different frequencies supplied by function generator (GPS-2125), with constant amplitude of current. The impedance was evaluated by measuring the voltage and current across the sample using a digital oscilloscope (GPS-1102B). The MI ratio can be defined as 𝑀𝐼% = 𝑍(𝐻)−𝑍(𝐻𝑚𝑎𝑥) 𝑍(𝐻𝑚𝑎𝑥) × 100 (1) In which Z is the impedance as a function of external field H, and Hmax is the maximum field applied to the samples in the MI measurement. Magnetic field sensitivity is defined as the slope of the MI curve of the sample 𝑆 = 𝑑 ( 𝛥𝑍 𝑍 ) /𝑑𝐻 . (2) Fig. 1. Schematic diagram of CoFeSiB micro-wire coated with ZnO thin film. 3. Results and discussion Fig. 2 shows the XRD pattern of bare micro-wire and 240 nm ZnO deposited micro-wire. The XRD pattern of bare micro-wire shows an amorphous nature. For 240 nm ZnO deposited micro-wire, the ZnO layer is crystalline in nature and the coexistence of two peaks at 2𝜃 = 34.39° and 36.18° shows the formation of (002) and (101) crystalline planes. To investigate the effect of the deposition of the ZnO layer on the magnetic properties of the micro- wires, the magnetic hysteresis loops of the undeposited and ZnO deposited layer with various thicknesses were recorded at room temperature. Figure 3(a-d) shows the magnetic hysteresis loops of undeposited and ZnO-deposited (87, 215 and 240 nm) micro-wires normalized to the saturation value (MS). The loops are all thin and narrow, and magnetization is saturated at a small applied field, indicating their soft ferromagnetic characteristics. There is just a little difference between deposited and undeposited micro-wires. This is because of the large volume of micro-wires has a dominant contribution in the measured magnetization of the samples. There are indeed changes in coercivity (Hc) and slope at zero point (permeability) which were determined for all the samples as shown in the Fig 4(e,f). Fig. 2. X-ray diffraction pattern of bare micro-wire and 240 nm ZnO deposited 1020304050607080 micro-wire/ZnO (nm)ZnO (101)ZnO (002)Intensity (a.u.)2degree bare micro-wire Fig. 3. Normalized magnetization curves for ZnO deposited layer with thickness of (a) 0, (b) 87, (c) 215, (d) 240 nm and effect of thickness of ZnO layer on (e) the slope at zero point (permeability) (f) coercivity (Hc). From the hysteresis loops of Fig. 3(a-d) it is observed that the coercivity (Hc) decreases by increasing the thickness of ZnO layer. Nevertheless, the decisive parameter determining the MI% of the samples is the μt. Therefore, the μt of the samples were derived (Fig. 3e). It is observed that the μt increases by increasing the thickness of the ZnO layer up to 240 nm. By further increase in the thickness the μt decreases too. This result can be used in interpretation of the MI results. Scanning electron microscopy (SEM) images of the bare micro-wire and 240 nm ZnO-deposited micro-wires are presented in figures 4(a,b), respectively. As can be seen, ZnO is formed in a granular structure atop the micro-wire surface. The surface if fully covered by the layer, a common characteristic of the layers fabricated by the sputtering method. As known, surface coverage of MI sensors without voids is vital for an effectively enhanced MI% with surface selectivity. Fig. 4. SEM images of (a) the bare micro-wire and (b) 240 nm ZnO thin layer deposited on micro-wire. -100-50050100-1.0-0.50.00.51.0 M (emu/g)0 nm(a)-100-50050100-1.0-0.50.00.51.0 87 nm(b)-100-50050100-1.0-0.50.00.51.0 M (emu/g)H (Oe)215 nm(c)-100-50050100-1.0-0.50.00.51.0 H (Oe)240 nm(d)0.040.060.080.100.120501001502002503001.01.52.02.53.0 Permeability(e) Hc (Oe)Thickness of ZnO (nm)(f) In order to understand the impact of ZnO deposition on the MI response of the micro-wire, magnetic field and frequency dependences of MI ratio of all the samples were measured up to 300 Oe, magnetic field strength and at different frequencies (1-10 MHz). Figure 5 (a) shows the MI ratio of the bare and deposited micro-wire with different thickness of ZnO at f = 2 MHz. As can be seen in Fig. 5 (a), MI response for ZnO deposited samples changes significantly. Increasing the thickness of the deposited layers from 87 to 240 nm, MI response increases considerably and then decreases by further increase in thickness. In figure Fig. 5 (c,d), the maximum of MI ratio and field sensitivity for different thicknesses of ZnO layers at f = 2 MHz can be seen. Maximum MI ratio of ~406.64% is obtained for the sample with the 240 nm ZnO layer. To better illustrate the effect of ZnO deposition, Maximum of the MI response for the bare and deposited micro-wire with 240 nm at various frequencies (1 -- 10 MHz) are presented in Fig. 4(b). It is noted that with increasing frequency, the maximum MI ratio initially increases, reaches a maximum at a particular frequency (2 MHz), and then decreases for higher frequencies. Relative contributions of domain wall motion and magnetization rotation to the μt should be considered in interpreting this trend [3]. The reduction of MI ratio at high frequencies is due to presence of eddy currents that causes damping of domain wall displacements and only rotation of magnetic moments takes place. In turn the μt diminishes, and the MI ratio decreases. The increase of MI response in the magnetic layer deposited on ribbon was already explained according to modifications of the ribbon surface and the closure of magnetic flux paths of deposited ribbons with magnetic materials having low thicknesses and decreasing MI for higher thicknesses [13]. We see the same trend and the MI ratio for ZnO deposited changes accordingly. These effects as well as the physical aspects of thickness and layer dependence of MI response according to roughness, and phase of ferromagnetic turns into diamagnetic by increasing the thickness of deposited layers is discussed. As ZnO is a well-known oxide semiconductor, we can conclude that there is less current passing through it and therefore no change in the skin depth of the micro-wire does occur. In the case of our samples, the ZnO oxide semiconductor layer diminishes the magnetic field passing through the surface. Therefore, the magnetic field penetrates better inside the magnetic micro-wire relative to the bare micro-wire. Also, the measured improvement in the MI response of the deposited micro-wire by ZnO can be considered as a result of the stress on the domain structure of the micro-wire and changing its magnetization dynamics [9]. According to the Fig. 3, another reason for enhancement of MI response is increase of magnetic permeability of the ZnO deposited sample which is proportional to the magnetization slope. Since the MI effect strongly depend on μt, by increasing μt the MI effect increases. As mentioned above, there are many reports on the increase of MI% by coating ferromagnetic layer on the surface of MI sensors [6-12]. Therefore, for the explanation of the evolution of the MI response we take look at the magnetic characteristics of ZnO reported by others [29]. As can be seen in fig. 4(a) the MI response significantly decreases with an increase in thicknesses of ZnO layer more than 240 nm. Kapilashrami et al. reported a systematic study of the film thickness dependence (0.1-1 𝜇𝑚) of room-temperature ferromagnetism in pure magnetron-sputtered ZnO thin films [29]. They observed that the ZnO films exhibit a sequential transition from ferromagnetism to paramagnetism and diamagnetism as a function of film thickness. We therefore speculate that by increasing those thin layer thicknesses above the 240 nm, ZnO thin layer evolved from ferromagnetism to paramagnetism and thereby MI effect decreases. It is worth mentioning that the thicknesses of the layers are well above the thresholds which their magnetic properties could be affected by the substrate. Therefore, our speculation for the similar magnetic evolution is reasonable. Fig. 5. (a) MI response for bare micro-wire and ZnO deposited with various thicknesses at f= 2 MHz. (b) Maximum MI response for bare micro-wire and ZnO (240 nm) deposited versus frequency. (c,d) Maximum MI response and field sensitivity versus thickness for ZnO deposited samples at f = 2 MHz. 4. Conclusion ZnO thin layer deposited by magnetron sputtering technique (87-500 nm), improved MI response and field sensitivity of Co-based micro-wire up to 406% and 4.99%/Oe respectively. The evolution of MI% was well understood by the magnetic properties of the ZnO layer at different thicknesses. The transverse permeability determined by VSM results, was used to reveal the role of the ZnO thin layer in the MI response and field sensitivity. XRD and SEM showed the granular structure of ZnO layer with (101) and (200) crystalline aligned directions. The result of this study show that, thin dielectric layers is one of the easiest and at the same time the most efficient methods for MI response optimization; and sensing elements made by this method can be a suitable candidate for development of high-performance MI sensors. -40-2002040100200300400 MI (%)H(Oe) As-cast 87nm 179nm 215nm 240nm 275nm 500nm(a)0246810100200300400 MI(%)Frequency(MHz) micro-wire micro-wire/ZnO (240 nm)(b)020040060012345100200300400 S (%/oe)thickness (nm)(d) MI (%) (c)References [1] L. V. Panina, K. Mohri, Magneto‐ impedance effect in amorphous wires, Appl. Phys. Lett. 65 (1994) 1189 -- 1191. doi:10.1063/1.112104. [2] M. Knobel, M. Vázquez, L. Kraus, Giant Magnetoimpedance, in: Handb. Magn. Mater., 2003: pp. 497 -- 563. doi:10.1016/S1567-2719(03)15005-6. [3] M.-H. Phan, H.-X. Peng, Giant magnetoimpedance materials: Fundamentals and applications, Prog. Mater. Sci. 53 (2008) 323 -- 420. doi:10.1016/j.pmatsci.2007.05.003. [4] F. Qin, H.-X. Peng, Ferromagnetic microwires enabled multifunctional composite materials, Prog. Mater. Sci. 58 (2013) 183 -- 259. doi:10.1016/j.pmatsci.2012.06.001. [5] M. Hajiali, S.M. Mohseni, S.E. Roozmeh, M. Moradi, Asymmetric magnetoimpedance effect in CoFeSiB amorphous ribbons by combination of field and current annealing for sensor applications, Superlattices Microstruct. 96 (2016) 191 -- 197. doi:10.1016/j.spmi.2016.05.028. [6] N. Laurita, A. Chaturvedi, C. Bauer, P. Jayathilaka, A. Leary, C. Miller, M.-H. Phan, M.E. McHenry, H. Srikanth, Enhanced giant magnetoimpedance effect and field sensitivity in Co- coated soft ferromagnetic amorphous ribbons, J. Appl. Phys. 109 (2011) 07C706. doi:10.1063/1.3548857. [7] A. Chaturvedi, K. Stojak, N. Laurita, P. Mukherjee, H. Srikanth, M.-H. Phan, Enhanced magnetoimpedance effect in Co-based amorphous ribbons coated with carbon nanotubes, J. Appl. Phys. 111 (2012) 07E507. doi:10.1063/1.3676214. [8] A.A. Taysioglu, Y. Kaya, A. Peksoz, S.K. Akay, N. Derebasi, G. Irez, G. Kaynak, Giant Magneto -- Impedance Effect in Thin Zinc Oxide Coated on Co-Based ( 2705 X ) Amorphous Ribbons, IEEE Trans. Magn. 46 (2010) 405 -- 407. [9] A.A. Taysioglu, A. Peksoz, Y. Kaya, N. Derebasi, G. Irez, G. Kaynak, GMI effect in CuO coated Co-based amorphous ribbons, J. Alloys Compd. 487 (2009) 38 -- 41. doi:10.1016/j.jallcom.2009.08.041. [10] A. Peksoz, Y. Kaya, A.A. Taysioglu, N. Derebasi, G. Kaynak, Giant magneto-impedance effect in diamagnetic organic thin film coated amorphous ribbons, Sensors Actuators A Phys. 159 (2010) 69 -- 72. doi:10.1016/j.sna.2010.03.002. [11] A. Ruiz, D. Mukherjee, J. Devkota, M. Hordagoda, S. Witanachchi, P. Mukherjee, H. Srikanth, M.H. Phan, Enhanced giant magneto-impedance effect in soft ferromagnetic amorphous ribbons with pulsed laser deposition of cobalt ferrite, J. Appl. Phys. 113 (2013) 17A323. doi:10.1063/1.4795802. [12] L. Jamilpanah, S. Azadian, J. Shoa e Gharehbagh, S. Haghniaz Jahromi, Z. Sheykhifard, S. Hosseinzadeh, S. Erfanifam, M.R. Hajiali, M.M. Tehranchi, S.M. Mohseni, Electrophoretic deposition of graphene oxide on magnetic ribbon: Toward high sensitive and selectable magnetoimpedance response, Appl. Surf. Sci. 447 (2018) 423 -- 429. doi:10.1016/J.APSUSC.2018.04.002. [13] L. Jamilpanah, M.R. Hajiali, S.M. Mohseni, S. Erfanifam, S.M. Mohseni, M. Houshiar, S.E. Roozmeh, Magnetoimpedance exchange coupling in different magnetic strength thin layers electrodeposited on Co-based magnetic ribbons, J. Phys. D. Appl. Phys. 50 (2017) 155001. doi:10.1088/1361-6463/aa6098. [14] M.R. Hajiali, S.M. Mohseni, L. Jamilpanah, M. Hamdi, S.E. Roozmeh, S.M. Mohseni, Spin- orbit-torque driven magnetoimpedance in Pt-layer/magnetic-ribbon heterostructures, Appl. Phys. Lett. 111 (2017) 192405. doi:10.1063/1.5000712. [15] M.R. Hajiali et al, Independence of spin-orbit-torque from exchange-bias probed via training effect in IrMn-layer/ferromagnetic-ribbon heterostructures, (unpublished). [16] M.R. Haj Ali, M. Khezri, S.E. Roozmeh, M. Moradi, S.M. Mohseni, The effect of mechanical polishing on current annealed Co 67 Fe 5 Si 15 B 13 amorphous ribbons: magnetoimpedance response, Eur. Phys. J. Appl. Phys. 65 (2014) 10601. doi:10.1051/epjap/2013130264. [17] M. Moradi, M. Hajiali, M. Khezri, S.E. Roozmeh, S.M. Mohseni, Structural Characterization and Magnetoimpedance Effect of Current Annealed Co-Based Amorphous Ribbons at Different Ambient, J. Supercond. Nov. Magn. 28 (2014) 265 -- 269. doi:10.1007/s10948-014- 2857-9. [18] M. Moradi, A. Dadsetan, S.M. Mohseni, A.J. Gharehbagh, Magnetoimpedance and Field Sensitivity of CoFeSiB Amorphous Ribbons under Applied Tensile Stress, J. Supercond. Nov. Magn. 28 (2015) 2059 -- 2062. doi:10.1007/s10948-015-2983-z. [19] Y. He, W. Zhang, S. Zhang, X. Kang, W. Peng, Y. Xu, Study of the photoconductive ZnO UV detector based on the electrically floated nanowire array, Sensors Actuators A Phys. 181 (2012) 6 -- 12. doi:10.1016/j.sna.2012.04.020. [20] L.E. Greene, M. Law, D.H. Tan, M. Montano, J. Goldberger, G. Somorjai, P. Yang, General Route to Vertical ZnO Nanowire Arrays Using Textured ZnO Seeds, Nano Lett. 5 (2005) 1231 -- 1236. doi:10.1021/nl050788p. [21] J.J. Hassan, M.A. Mahdi, C.W. Chin, H. Abu-Hassan, Z. Hassan, A high-sensitivity room- temperature hydrogen gas sensor based on oblique and vertical ZnO nanorod arrays, Sensors Actuators B Chem. 176 (2013) 360 -- 367. doi:10.1016/j.snb.2012.09.081. [22] J. Zhou, N.S. Xu, Z.L. Wang, Dissolving Behavior and Stability of ZnO Wires in Biofluids: A Study on Biodegradability and Biocompatibility of ZnO Nanostructures, Adv. Mater. 18 (2006) 2432 -- 2435. doi:10.1002/adma.200600200. [23] S. Deng, H. Fan, M. Wang, M. Zheng, Ќ.J. Yi, R. Wu, Ќ.H. Tan, C. Sow, Ќ.J. Ding, Y. Feng, K. Loh, Thiol-Capped ZnO Nanowire / Nanotube Arrays with Tunable Magnetic Properties at Room Temperature, 4 (2010) 495 -- 505. [24] W. Liu, W. Li, Z. Hu, Z. Tang, X. Tang, Effect of oxygen defects on ferromagnetic of undoped ZnO, J. Appl. Phys. 110 (2011) 13901. doi:10.1063/1.3601107. [25] G. Xing, D. Wang, J. Yi, L. Yang, M. Gao, M. He, J. Yang, J. Ding, T.C. Sum, T. Wu, Correlated d0 ferromagnetism and photoluminescence in undoped ZnO nanowires, Appl. Phys. Lett. 96 (2010) 112511. doi:10.1063/1.3340930. [26] Q. Xu, H. Schmidt, S. Zhou, K. Potzger, M. Helm, H. Hochmuth, M. Lorenz, A. Setzer, P. Esquinazi, C. Meinecke, M. Grundmann, Room temperature ferromagnetism in ZnO films due to defects, Appl. Phys. Lett. 92 (2008) 82508. doi:10.1063/1.2885730. [27] Y.Y. Villanueva, D. Liu, P.T. Cheng, Pulsed laser deposition of zinc oxide, Thin Solid Films. 501 (2006) 366 -- 369. doi:10.1016/j.tsf.2005.07.152. [28] L. Znaidi, Sol -- gel-deposited ZnO thin films: A review, Mater. Sci. Eng. B. 174 (2010) 18 -- 30. doi:10.1016/j.mseb.2010.07.001. [29] M. Kapilashrami, J. Xu, V. Ström, K. V. Rao, L. Belova, Transition from ferromagnetism to diamagnetism in undoped ZnO thin films, Appl. Phys. Lett. 95 (2009) 33104. doi:10.1063/1.3180708. [30] J.-J. Wu, S.-C. Liu, Low-Temperature Growth of Well-Aligned ZnO Nanorods by Chemical Vapor Deposition, Adv. Mater. 14 (2002) 215 -- 218. doi:10.1002/1521- 4095(20020205)14:3<215::AID-ADMA215>3.0.CO;2-J. [31] S. Peulon, D. Lincot, Cathodic electrodeposition from aqueous solution of dense or open- structured zinc oxide films, Adv. Mater. 8 (1996) 166 -- 170. doi:10.1002/adma.19960080216.
1907.10177	1	1907	2019-07-23T23:50:43	Hybrid Integrated Photonics Using Bulk Acoustic Resonators	[ "physics.app-ph", "physics.optics" ]	Microwave frequency acousto-optic modulation is realized by exciting high overtone bulk acoustic wave resonances (HBAR resonances) in the photonic stack. These confined mechanical stress waves transmit exhibit vertically transmitting, high quality factor (Q) acoustic Fabry Perot resonances that extend into the Gigahertz domain, and offer stress-optical interaction with the optical modes of the microresonator. Although HBAR are ubiquitously used in modern communication, and often exploited in superconducting circuits, this is the first time they have been incorporated on a photonic circuit based chip. The electro-acousto-optical interaction observed within the optical modes exhibits high actuation linearity, low actuation power and negligible crosstalk. Using the electro-acousto-optic interaction, fast optical resonance tuning is achieved with sub-nanosecond transduction time. By removing the silicon backreflection, broadband acoustic modulation at 4.1 and 8.7 GHz is realized with a 3 dB bandwidth of 250 MHz each. The novel hybrid HBAR nanophotonic platform demonstrated here, allowing on chip integration of micron-scale acoustic and photonic resonators, can find immediate applications in tunable microwave photonics, high bandwidth soliton microcomb stabilization, compact opto-electronic oscillators, and in microwave to optical conversion schemes. Moreover the hybrid platform allows implementation of momentum biasing, which allows realization of on chip non-reciprocal devices such as isolators or circulators and topological photonic bandstructures.	physics.app-ph	physics	Hybrid Integrated Photonics Using Bulk Acoustic Res- onators Hao Tian1, Junqiu Liu2, Bin Dong1, J Connor Skehan2, Michael Zervas2, Tobias J. Kippenberg2,†, Sunil A. Bhave1,† 1OxideMEMS Lab, Purdue University, 47907 West Lafayette, IN, USA 2Institute of Physics, Swiss Federal Institute of Technology Lausanne (EPFL), 1015 Lausanne, Switzerland † E-mail: [email protected], [email protected] Integrated photonic devices based on Silicon Nitride (Si3N4) waveguides allow for the ex- ploitation of nonlinear frequency conversion, exhibit low propagation loss, and have led to advances in compact atomic clocks, supercontinuum generation, optical sensing, ultrafast ranging, coherent communication, and spectroscopy. Yet, in contrast to silicon, the amor- phous nature of (Si3N4) requires combination with other materials to achieve active tuning or modulation. Approaches so far range from thermal tuning, integration with Graphene and atomically thin semiconductors to solution based processing of piezoelectric materials. Here, microwave frequency acousto-optic modulation is realized by exciting high overtone bulk acoustic wave resonances (HBAR resonances) in the photonic stack. These confined me- chanical stress waves exhibit vertically transmitting, high quality factor (Q) acoustic Fabry- Perot resonances that extend into the Gigahertz domain, and offer stress-optical interaction with the optical modes of the microresonator. Although HBAR are ubiquitously used in mod- ern communication, and often exploited in superconducting circuits, this is the first time they have been incorporated on a photonic circuit based chip. The electro-acousto-optical interac- tion observed within the optical modes exhibits high actuation linearity, low actuation power and negligible crosstalk. Using the electro-acousto-optic interaction, fast optical resonance tuning is achieved with sub-nanosecond transduction time. By removing the silicon backre- flection, broadband acoustic modulation at 4.1 and 8.7 GHz is realized with a 3 dB bandwidth 1 of 250 MHz each. The novel hybrid HBAR nanophotonic platform demonstrated here, allow- ing on chip integration of micron-scale acoustic and photonic resonators, can find immediate applications in tunable microwave photonics, high bandwidth soliton microcomb stabiliza- tion, compact opto-electronic oscillators, and in microwave to optical conversion schemes. Moreover the hybrid platform allows implementation of momentum biasing, which allows realization of on chip non-reciprocal devices such as isolators or circulators and topological photonic bandstructures. Integrated photonics has drawn increasing attention in recent years. While within the last decade, silicon photonics has transitioned from laboratory based research into commercial prod- ucts used in data-centers, major efforts are still underway with regards to silicon nitride (Si3N4) photonic devices 1. The platform has attracted intense efforts due to its wide transparency window from the visible to mid-infrared, ultralow linear losses, absence of two photon absortion in the telecommunication band, space-compatibility 2, large Kerr nonlinearity (χ(3)), and wide geometric flexibility for waveguide dispersion engineering 3. The material has been the material of choice for soliton microcomb generation 4, and supercontinuum generation 3, optical filters 5, gyroscopes 6, and optical interconnects 7 have been demonstrated. Recent advances of Si3N4-based dissipa- tive Kerr soliton (DKS) microcombs, have included octave-spanning comb spectra 8, 9, ultralow initiation power 10 -- 12, and microcomb repetition rates in the microwave domain 13. Reliable and fast resonance tuning of Si3N4 microring resonators is becoming an important asset and requirement for a number of applications in integrated nonlinear photonics. For exam- ple, high bandwith tuning allows microcomb repitition rate stabilization 14, or resonance tuning for filters. Likewise, recently emerged platforms such as spatio-temporal modulation based opti- cal non-reciprocity 15 and topological optical bandstructures both require GHz speed modulation within optical microresonators, which poses stringent requirements on the cross-talk and size. However, due to the inversion symmetry, and thus the lack of χ(2) nonlinearity, it is difficult to electrically modulate the refractive index of Si3N4. Traditionally, the thermo-optical effect is employed to fulfill the tuning requirement 16, 17 which, however, presents low tuning speed (∼1 2 ms), high power consumption (∼1 mW), and thermal cross-talk. These drawbacks make it incom- patible with large-scale integration and cryogenic applications 18. Although hybrid integration with various electro-optical materials, e.g., graphene 19, lead zirconate titanate (PZT) 20, lithium niobate (LiNbO3) 21, and monolayer WS2 22 has made significant progresses, there are still remaining challenges related to CMOS-compatibility, fabrication complexity, optical losses, and dispersion engineering. To fully utilize the maturity and advantages of Si3N4 photonics, new tuning mecha- nisms which retain the original optical properties are needed. The stress-optical effect, discovered over a hundred years ago, has recently gained attention for its role in the modulation of Si3N4 waveguides and microring resonators both theoretically23, 24 and experimentally 25, 26, thanks to the advances in Micro-Electro-Mechanical Systems (MEMS) 27. In this work, by integrating aluminium nitride (AlN) piezoelectric actuators onto Si3N4 pho- tonic devices, we demonstrate, to the best of our knowledge, the first acousto-optic modulation of Si3N4 microring resonators using High-overtone Bulk Acoustic wave Resonances (HBAR) 28. Sub-micron wavelength acoustic waves are excited by macroscopic actuators, which transmit ver- tically into the substrate and perpendicular to optical paths (see Fig. 1a). Trapped inside a Fabry- Perot-like acoustic cavity that is formed by the top and bottom surfaces of the entire substrate, a rich family of acoustic resonant modes is efficiently excited at the microwave frequencies up to 6 GHz. The coupling between vertical acoustic waves and in-plane optical circuits makes it possible for independent optimization of the actuator and optical components. The high lateral acoustic mode confinement further enables low cross-talk and compact integration. These features are in stark contrast with conventional surface acoustic wave (SAW) acousto-optic modulation (AOM) 24, 29 which requires an inter-digital transducer (IDT) with sub-micron electrode fingers, coplanar integration of IDT and optical waveguides, and perfect termination of acoustic waves to reduce cross-talk. The same structure platform is capable of linear bi-directional tuning of Si3N4 microring resonators, which shows a high power efficiency of 5 nW per picometer resonant wavelength tuning (compared to thermal tuning on the order of 1 mW/pm 16). Compared with PZT 25, 26, AlN has 3 Figure 1: Hybrid nanophotonic high-overtone bulk acoustic resonator (HBAR) platform. a 3D schematic illustrating excitation of bulk acoustic wave resonances via a macroscopic piezoelec- tric actuator, which transmit vertically into the stack and form acoustic standing waves inside the various acoustic Fabry-Perot cavities. The resonance enhanced mechanical stress changes the waveguide's effective index via the stress-optical effect, and thereby modulate the output optical intensity. b Cross-section of the device along the black dashed line in c, with critical dimensions labeled. The AlN piezoelectric actuator is placed directly on top of the Si3N4 microring resonator which is embedded in 5.4-µm of SiO2 cladding. c Optical microscope image of the fabricated device. the advantage of of high actuation linearity, no hysteresis, high power handling (breakdown field >100 V/µm), and low current leakage. Here, the tuning speed is fully explored, demonstrating sub-nanosecond switching capability, which is mainly limited by intrinsic acoustic resonances. Broadband (250 MHz) AOMs around 4.1 and 8.7 GHz are realized by damping acoustic resonances inside the Si substrate. The simple but efficient stress-optical platform demonstrated in this work may find widespread applications in Si3N4 microwave photonics 30, such as the repetition rate stabilization and tuning of soliton microcombs via injection-locking 31, on-chip optomechanical frequency comb generation 32, and comb-assisted microwave photonic filters 33. As a well-known piezoelectric thin film, c-axis oriented polycrystalline AlN (with a piezo- 4 𝜆acousticSiO2AlN 1µmAl 100nmMo 100nmSiN5.4µm3µm3µmabc100 µmwaveguideSi230µmelectric coefficient e33 ∼1.55 C/m2 34) is utilized to form the piezoelectric actuator which will be deformed and generate stress around the Si3N4 waveguide when a vertical electric field is ap- plied. As shown in Fig. 1a, a disk-shaped actuator is placed directly on top of the Si3N4 ring resonator, with the microring resonator positioned at the outer edge of the disk actuator. As we drive the actuator harmonically, vibration of the AlN disk will launch an acoustic wave vertically into the substrate (Fig. 1a). Since the bottom surface of the substrate is smooth and flat, the acous- tic wave will be reflected and subsequently bounce back and forth between the top and bottom surfaces. Working as an acoustic Fabry-Perot cavity, the counter-propagating acoustic waves will constructively interfere when the cavity length is an integer number of the acoustic wavelength, with acoustic energy trapped inside the cavity. These bulk acoustic standing waves will enhance the stress field around optical waveguides and modulate the effective index through stress-optical effect 23. The stack's cross-section is illustrated in higher detail in Fig. 1b. To apply the vertical electric field, an AlN film of 1 µm thickness is sandwiched between top Al (100 nm thickness) and bottom Mo (100 nm thickness) metal layers. The Si3N4 waveguides are fabricated using a subtractive process 35, having a height of 800 nm and width of 1.8 µm, and which are fully buried in SiO2 cladding 5.4 µm thick (2.4 µm SiO2 from the Mo layer to prevent metal absorption). The entire device sits on a 230-µm-thick Si substrate. The radius of the microring resonator is 118 µm, and is placed 3 µm within the edge of disk actuator. The final fabricated device is shown in Fig. 1c (see Methods for more fabrication details). The area around waveguide-to-microring coupling region is opened to prevent any actuation-induced perturbation of light coupling between bus waveguide and microring resonator. Experimentally, the electromechanical reflection parameter S11 is first measured using port 1 of a network analyzer (Fig. 2g), which characterizes the energy conversion from electrical to mechanical vibration as a dip. As shown in Fig. 2a, a series of resonance dips is found to evenly distribute over multiple octaves in the microwave regime. A zoom-in of the spectrum is illustrated in Fig. 2d which highlights the resonance shape, linewidth (∼3 MHz), and acoustic free spectral 5 range (FSR) of around 17.5 MHz. The narrow linewidth demonstrates high mechanical Q (∼1000) which is mainly limited by the intrinsic acoustic loss in the substrate and scattering at interfaces. It is interesting to note that the envelope of these sharp resonances varies slowly and smoothly with a period of ∼ 490 MHz. This is caused by the resonance inside the 5.4 µm SiO2 cladding layer, which stems from acoustic wave reflections at the Si-SiO2 interface due to an acoustic impedance mismatch. The position of SiO2 resonance is located at the node of the envelope, whereas anti- resonance is located at the anti-node. Intuitively speaking, at the anti-resonance of the SiO2 layer, it works as an acoustic anti-reflection coating such that more acoustic energy will transmit into the Si substrate, which has larger acoustic impedance, and thus better electromechanical conversion. When the acoustic half wavelength matches the 1 µm AlN thickness, the AlN layer reaches its fundamental resonance mode and becomes more efficient in excitation of acoustic waves around 4 -- 4.3 GHz. It can be seen that, by optimizing AlN and SiO2 thicknesses, the resonances of these two cavities can be misaligned to further improve acoustic wave excitation. Additionally, the coupling between the Si substrate cavity and the SiO2 and AlN cavities causes periodic fluctuations of the FSR and higher order dispersion, as will be discussed later. The acousto-optic modulation of the microring resonator can be characterized by an optome- chanical S21 measurement as shown in Fig. 2g, where by biasing the input laser (∼1550 nm) at the slope of the optical resonance, its output intensity is modulated as we launch a -5 dBm RF signal from port 1 of the VNA. The intensity modulation of the optical signal is measured using a photodiode and sent back to port 2 for the S21 measurement. Note that no optical and electrical amplifiers are employed in an effort to preserve direct electro-opto-mechanical transduction. The optomechanical S21 measurements are performed for both the transverse electric (TE) and trans- verse magnetic (TM) optical modes, as shown in Fig. 2b and c respectively. As expected, a broad range of periodic peaks is observed which correspond to each HBAR mode in S11. Due to the dif- ferent optical mode profiles of the TE and TM modes (and thus acousto-optic mode overlaps) and optical Q factors, they respond differently, with the TE mode response strongest around 2 GHz and the TM mode responding most strongly around 4 GHz. Since the TE mode shows higher optical Q, it will enter the resolved sideband regime at frequencies far beyond its linewidth (1 -- 2 GHz), 6 Figure 2: Microwave frequency electro-acousto-optic modulation. a Electromechanical S11 spectrum from 1 to 6 GHz. A range of equidistant bulk acoustic resonances is found to exist over a broad frequency range. The inset schematic illustrates the acoustic wave reflection at interfaces. b, c Optomechanical S21 responses of TE and TM modes demonstrate acousto-optic modulation covering multiple octave-spanning microwave frequencies. Due to different optical mode shapes (insets in b and c) and thus acousto-optic mode overlap, TE and TM modes show dissimilar S21 spectra. d, e The zoom-in of S11, and TE mode's S21 within the window (green shaded area in b) around 2 GHz. f The zoom-in of TM mode's S21 around 4 GHz in c. The resonances distribute evenly with an FSR of 17.5 MHz. g Schematic of the setup for measuring electromechanical and optomechanical response. ECDL: external cavity diode laser, PC: polarization controller, DUT: device under test, PD: photo-diode, VNA: vector network analyzer. h Numerical simulation of vertical stress σz distribution for one typical acoustic resonant mode at 2.041 GHz under 1 V driving field, with a zoom-in around the optical waveguide (red box in h) shown in i. At several GHz, the acoustic wavelength is similar in scale to optical wavelength and waveguide structure. 7 -4.4-4.0-3.6-100-80-60123456-120-100-80-60 S11 (dB)S21 (dB) S21 (dB)Frequency (GHz)ECDLPD2FPCVNA1DUTadbcef2041 MHzrzhig500𝜎𝑧(MPa)2500-250-50060300-30-60TETM-3.75-3.70-3.652.102.122.142.162.182.20-90-80-70Frequency (GHz)S21 (dB)S11 (dB) 4.064.084.104.124.14-80-70Frequency (GHz)S21 (dB)where the modulation sidebands are suppressed when biasing at the resonance slope. The zoom-in of highlighted regions (green shaded areas) are as shown in Fig. 2e and f, illustrating clear peaks with high contrast (>20 dB) between resonance and anti-resonance. Interestingly, the same mea- surements are performed for devices with a different actuator lateral shape, which show similar results for S11 and S21 of the TE and TM modes (see Supplementary Material). This indicates that the HBAR mode distribution and optomechanical spectra are less related to the shape of the actuator, but more to the vertical stack. This relaxes the requirements on actuator shape and size, which lends itself with high design freedom and small footprint. A numerical (COMSOL) simulation of one typical acoustic mode at 2.041 GHz is shown in Fig. 2h, and a zoom-in around the optical waveguide is in Fig. 2i. The acoustic standing wave distributes uniformly over the entire substrate, which indicates that the optical circuits can be buried deeply inside the SiO2 cladding, free from the trade-off between actuation efficiency and absorption losses due to metal as found in traditional optical modulators. Also, an acoustic wave- length that is comparable to optical wavelength and waveguide structure at microwave frequencies is achieved with macroscopic actuators. As is evident, the simplicity of our structure presents advantages such as simple fabrication, high fabrication tolerance, high rigidity, and high power handling. Additionally, it can be seen from Fig. 2h that the acoustic mode is largely confined be- neath the actuator which guarantees low electromechanical cross-talk (-60 dB) between adjacent devices (see Supplementary Material). These features may supplement conventional SAW based AOM for future microwave photonics applications with ultra-high optical Q, low cost, and dense integration. An analytical electromechanical model is established to help us get deep understanding of the mechanical performance of the device (see Supplementary Material for details). The S11 re- sponse is first calculated as shown in Fig. 3a, illustrating high accuracy of the model as compared to the experiment (Fig. 3b). As mentioned above, the coupling from SiO2 and AlN cavities not only modulates the resonances' magnitude envelope, but also the dispersion of mechanical modes (deviation from equidistant spectrum). Fig. 3c clearly shows the variation of frequency difference 8 Figure 3: Mechanical dispersion analysis of HBAR modes. a Calculated and b measured S11 spectrum showing good agreement between the electromechanical model and experiment. Each green dashed line denotes the location of a resonance from the SiO2 cavity. The inset in b illustrates the coupling between Si, SiO2, and AlN cavities. c The measured (circle) and calculated (solid line) frequency difference between each pair of adjacent Si resonances, showing a periodic variation of local FSR around an average value of 17.5 MHz. Note that the maxima of FSR align with green dashed lines where the SiO2 resonances are located. d Measured (circle) and calculated (solid line) higher order dispersion represented by the frequency deviation from an equidistant frequency grid (with average FSR = 17.515 MHz), referencing to mode ν0 (= 3.0145 GHz, labeled as yellow dot). µ is the mode number difference relative to the mode at 3.0145 GHz. 9 1.01.52.02.53.03.54.04.55.0-10-5017.017.518.0-4.5-4.0-3.5 Frequency (GHz)- (MHz)-(+FSR) (MHz)S11 (dB)-3-2-10S11 (dB)acbAlNSiO2SidCalc.Meas.(local FSR) between each pair of adjacent resonances. The local FSR varies nearly periodically around an average value of 17.5 MHz with the same period as the envelope in Fig. 3b. By com- paring Fig. 3b and c, it can be found that at each node of the envelope, the FSR reaches maximum value, which means the spacing of Si resonances increases near the SiO2 resonance. From the standpoint of the Si cavity, the variation of FSR can be understood intuitively that, the wave re- flected back into the Si substrate from the Si-SiO2 interface experiences varied phases relative to the SiO2 resonance. For example, around SiO2 resonance, the Si-SiO2 interface locates at the node of an acoustic stress wave with near-zero stress (maximum displacement), which presents a free boundary condition. Far beyond SiO2 resonances, the interface is at an anti-node corresponding to fixed boundary condition (zero displacement). These various boundary conditions each impose a particular phase for a given reflected wave, and thus change the effective cavity length of Si. However, for the AlN and SiO2 cavities, the green dashed lines that denote the location of each SiO2 resonance bunch together around 4 GHz where the AlN cavity resonance is located. Also, it can be seen that the average FSR of acoustic modes decreases around the AlN resonance. Based on these observations and the fact that the acoustic impedance of SiO2 is smaller than Si and then AlN, we can conclude that, for two coupled acoustic cavities, the small cavity (e.g., SiO2) with smaller acoustic impedance tends to decrease (increase) the effective cavity length (FSR) of the big cavity (e.g., Si) when it's on resonance compared to off resonance, and vice versa 36. This is, to some extent, similar to coupled optical cavities by treating acoustic impedance as the effective refractive index. The higher order dispersion is presented in Fig. 3d, which shows the frequency deviation of each resonance from the ideally even distribution with reference to mode at 3.0145 GHz and period of 17.515 MHz. Mathematically, it can be interpreted as the integral of Fig. 3c (that is, the accumulation of FSR deviation relative to 17.515 MHz) with respect to the origin at 3.0145 GHz. The higher order dispersion also shows periodic variation caused by the coupling between the Si and SiO2 cavities, which varies between normal and anomalous group velocity dispersion. The roll-off starting around 3.5 GHz is caused by reduced FSR due to coupling of the AlN cavity. This 10 study of mechanical dispersion could benefit the growing field of mechanical dispersion engineer- ing for future applications (e.g., mechanical solitons) and future devices by optimizing Si, SiO2, and AlN thicknesses or by choosing materials with different acoustic impedance. In this sense, further theoretical and numerical studies are necessary for a more complete understanding of the acoustic wave propagation and mechanical cavity coupling in such a platform. Besides the resonant microwave frequency acousto-optic modulation, the proposed stress- optical platform is capable of ultra-fast tuning of optical resonances with ultra-low power con- sumption (see Supplementary Material for detailed results). With the same device, bi-directional tuning can be achieved with linear tunability of -0.2 pm/V and power efficiency of 5 nW/pm. Here, the tuning speed is fully explored, showing sub-ns response time. To do this, the time do- main dynamic response is recorded while applying a modulated signal. A small-signal (Vpp = 2 V) square wave (1 MHz) is first applied in Fig. 4a, and the transmitted light intensity modulation is measured. Although with only 0.4 pm relative resonant wavelength shifting, the output signal shows clear separation between two switching states which is larger than the noise level. We also observe sharp peaks at tuning edges when rapidly switching from one state to the other, which is caused by the excitation of acoustic resonances. Periodic (5 MHz) short pulses with 20 ns pulse widths are also applied as seen in Fig. 4b. To quantitatively show the similarity between input and output signals, the normalized cross-correlation between them is calculated as shown in the inset of 4b. Indeed, the similarity between the auto-correlation (red) of the input signal itself and the cross-correlation (black) demonstrates ultra-fast actuation beyond the nanosecond. Despite the high actuation speed, the existence of HBAR modes prevents digital modulation at repetition rates beyond 1 GHz, where a flat and broadband response is usually required as in traditional optical communication. In this sense, pre-conditioning of the input signal or data post- processing is necessary to eliminate the distortion from mechanical resonances. However, due to the constant f · Q product in a general mechanical resonant systems, the acoustic resonances at high frequency gradually disappear due to low mechanical Q. If we apply an ultra-high repetition rate (frep) signal in the mechanical resonance damping-out region, we can still retrieve a clear out- 11 Figure 4: High speed actuation and broadband electro-acousto-optic modulation. a Time domain response (orange) to a small-signal square wave (gray) with 2 V Vpp, 1 MHz repetition rate, and a 50% duty cycle. The sharp peaks at tuning edges are caused by mechanical ringing of acoustic resonances. b Time domain response to short pulses with 20 V Vpp, 5 MHz repetition rate, and a 20 ns pulse width, demonstrating ultra-fast (sub-ns) tuning speed. The inset shows the normalized cross-correlation (black) between input and output signals and the auto-correlation (red) of the input signal. The right Y axis in a, b denotes the resonant wavelength shifting relative to 0 V voltage, according to a linear tuning of -0.2 pm/V. c 6 GHz square wave driving at the frequency where mechanical resonances disappear due to low mechanical Q. The electrical signal (gray) is measured by an oscilloscope after 20 dB attenuation of the original signal. The optical output shows clear oscillations with a frequency equal to the driving field, illustrating GHz level piezoelectric actuation. d Cross-section of the device after roughing and then pasting polyurethane (PU) epoxy (mixed with 3 µm Nickle powders) on the backside of the Si substrate. e S11 and f S21 of the TM mode after weakening acoustic resonances. The VNA responses become smoother with only wide range envelope variation. Enabled only by AlN fundamental and second harmonic resonances (green shaded regions), broadband modulation can be achieved with 3 dB bandwidth of 250 MHz for each. 12 -400-300-200-1000100200300400Electrical input (mV)-400-2000200400-6-4-20246Time (ps)Optical output (mV)-2.2-2.0-1.8246810-120-100-80-60-40-20 Frequency (GHz)S21 (dB)S11 (dB) -500501002.02.22.4-10010Voltage (V)Time (ns) 20-2 (pm)-5005001 (ns)0.00.20.40.60.81.01.451.501.55-101Voltage (V)Time (s) 0.20.0-0.2 (pm)InputOutput20nsInputOutputSiO2AlNMoAlSiNPU + Ni particleacbdefput, getting rid of resonance induced signal distortion. This is demonstrated in Fig. 4c, where a 6 GHz square wave (gray curve) is applied using a programmable pattern generator and amplified by an optical modulator driver to Vpp of 7 V. The output optical modulation (orange) shows dis- tinguishable oscillation at the same repetition rate, which suggests that the piezoelectric actuator itself can perform ultra-fast actuation, despite the fact that broadband modulation is mostly limited by acoustic resonances. Because of the lower responsivity to higher harmonic Fourier components (e.g., 12 GHz, 24 GHz) of the input square wave, the output behaves more like a 6 GHz sinusoidal wave. Note that, even at ultra-high frep, the pseudo-random Non-Return-to-Zero (NRZ) signal also contains broad frequency components besides harmonics of frep, which still suffer from acoustic resonances. For applications where a broad bandwidth is required, these acoustic modes should be effectively damped or even eliminated. To do this, the bottom surface of the Si substrate is first roughed by XeF2 isotropic etching to diffract acoustic waves, and then polyurethane (PU) epoxy (mixed with 3 µm Nickle powder) is pasted onto the backside for damping acoustic vibrations 37, as shown in Fig. 4d. After post-processing the fabricated device, its electromechnical and optomechanical performances are recorded in Fig. 4e and f, respectively. To increase the signal to noise ratio of S21 at high frequencies, the RF signal from the VNA is amplified before being applied to the actuator. From the S11 response, one can see that nearly all resonances from the Si substrate cavity are diminished. However, the smoothly varying envelope from resonances in SiO2 and AlN cavities still exists, since they remain unaffected by the post-processes. The S21 measurement also demonstrates smoothing of the modulation spectrum, but with broad range variations. Some exceptionally small resonances are still visible below 2.5 GHz in S21, because the acoustic wavelength at low frequencies is comparable or larger than the Si roughing scale (∼5 µm). It's worth noting here that the optical measurement is more sensitive than its electrical counterpart due to high optical Q and its signal to noise ratio. The fast roll-off of S21 starting around 9.5 GHz is mainly limited by the optical quality factor of the TM mode, and the actual acousto-optic interaction may potentially extend to frequencies beyond 10 GHz. 13 The fundamental resonance from the AlN thin film enhances the electromechanical conver- sion efficiency and thus optical modulation around 4.13 GHz, and thanks to its low mechanical Q, we observe broadband modulation with a 3 dB bandwidth of 250 MHz, as shown in Fig. 4f. The second harmonic resonance of AlN cavity is also found in S21 around 8.7 GHz, here with a 260 MHz bandwidth. These broad bands of modulation can potentially be used to connect suppercon- ducting circuits with optical interfaces for low-loss quantum information communication 27, 38. Of note, the positions of these bands can be engineered by modifying the AlN and SiO2 film thickness for specific applications. For future works, dedicated design and patterning of the bottom sur- face can be done for better suppression of acoustic resonances from Si cavity. To further increase modulation bandwidth, more effort is needed to eliminate resonances from SiO2 and AlN layers, which can be realized by matching mechanical impedance at interfaces, such as using an acoustic anti-reflection layer 39, or through phononic crystal band gap engineering to suppress undesired acoustic modes 40. In conclusion, we demonstrate for the first time the integration of HBAR resonators within a nano-photonic platform. We realize a hybrid multi-functional stress-optical platform based on bulk acoustic standing and traveling waves. In the device, acoustic resonances are effectively generated by a macroscopic AlN actuator, and are used for microwave frequency acousto-optic modulation. These acoustic modes are evenly distributed over a broad frequency range up to 6 GHz. To put this work into perspective, it is worth pointing out several key differences between our platform and conventional SAW-based AOMs 29. Firstly, in our scheme, optical structures can be buried deep inside the cladding material, which preserves the high optical Q. Secondly, the critical size of the actuator does not need to be scaled down for exciting high frequency (sub-micron acoustic wavelength) acoustic waves, since acoustic resonances primarily rely on the dimensions of the vertical stack. This enables optical lithography, high power handling, and a small footprint (due to the overlay of mechanical and optical components). Finally, the confinement of acoustic modes vertically inside the substrate guarantees low cross-talk between adjacent devices, and thus compact integration. Altogether, these advantages serve to produce a device which is not only novel, but has potential for widespread applications across a variety of fields. 14 Methods Device fabrication The Si3N4 waveguides are fabricated using a subtractive process 35. The fab- rication process of AlN piezoelectric actuators is described here in detail. 100 nm Mo and 1 µm polycrystalline AlN films are sputtered on SiO2 cladding through foundry services (OEM Group). The AlN disk is patterned by AZ1518 photoresist, and dry etched using Cl2 and BCl3 in a Pana- sonic E620 Etcher. The dry etching of the bottom electrode (Mo) is performed using Cl2 in the same Panasonic E620 Etcher. Finally, the top 100 nm of Al is evaporated by a PVD E-beam evap- orator, and patterned using a standard lift-off process. The whole wafer is diced into individual chips by deep reactive ion etching (RIE) followed by chemical mechanical polishing (CMP) of the Si substrate. This enables smooth chip facets for efficient coupling of light from lensed fiber to inverse waveguide taper at the edge. This three-mask, photolithographic-only fabrication leads to a low cost and high fabrication tolerance. Damping bulk acoustic resonances To damp and eliminate the intrinsic bulk acoustic wave res- onances, the backside surface of the Si substrate is first isotropically etched by XeF2 in a Xactix Xenon Difluroide E1 system, with the top surface protected by the photoresist AZ1518. The rough- ing will diffract acoustic waves to random directions and thus weaken constructive interference. Next, as suggested by previous work 37, a layer of polyurethane (PU) epoxy mixed with 3 µm nickle powder is pasted at the bottom, which can damp acoustic vibration at the boundaries and absorb acoustic energy. Experimental setup The electromechanical S11 reflection response is measured by port 1 of a network analyzer (Agilent E8364B), where the electrical signal is applied to the device through an RF GS probe (GGB 40A-GS-150). For the optomechanical S21 measurment, around 100 µW continuous wave (CW) light from a diode laser (Velocity Tunable Laser 6328) is edge coupled to the device using a lensed fiber via inverse taper. A -5 dBm RF electrical signal is applied from port 1 of the network analyzer to driving the piezoelectric actuator, and the light intensity modulation is detected by a 12 GHz photodiode (New Focus 1544), which is sent back to port 2 of network analyzer. 15 In the high repetition rate time domain modulation measurement, a 6 GHz square wave is supplied by a programmable pattern generator (Tektronix PPG1251), which is amplified by an optical modulator driver (WJ communication SA1137-2) to 7 V Vpp before applying to the device. The intensity modulation is detected by the same photodiode (New Focus 1544) and then recorded by a broad bandwidth oscilloscope (Tektronix DSA8200). Acknowledgements This work was supported by Swiss National Science Foundation under grant agree- ment No. 176563 (BRIDGE), and by the Defense Advanced Research Projects Agency (DARPA), Mi- crosystems Technology Office (MTO) under contract No. HR0011-15-C-0055 (DODOS), and by funding from the European Unions H2020 research and innovation programme under FET Proactive grant agreement No. 732894 (HOT). The samples were fabricated in the EPFL center of MicroNanoTechnology (CMi), and Birck Nanotechnology Center at Purdue University. The authors would like to thank Dr. Daniel E. Leaird for helping testing high frequency time domain modulation, and Mohammad Bereyhi and Dr. Ben Yu for valuable discussion. Additional information See Supplementary Material for supporting content. The data that support the plots are available from the corresponding authors upon reasonable request. Competing interests The authors declare no competing interests. 16 Supplementary Material: Hybrid Integrated Photonics Using Bulk Acoustic Resonators Hao Tian1, Junqiu Liu2, Bin Dong1, J Connor Skehan2, Michael Zervas2, Tobias J. Kippenberg2,†, Sunil A. Bhave1,† 1OxideMEMS Lab, Purdue University, 47907 West Lafayette, IN, USA 2Institute of Physics, Swiss Federal Institute of Technology Lausanne (EPFL), 1015 Lausanne, Switzerland † E-mail: [email protected], [email protected] 1 Analytic analysis of electro-opto-mechanical response Electromechanical model of HBAR mode To better explain the electromechanical S11 response and mechanical dispersion shown in the main text, a one-dimensional (1D) analytic electrome- chanical model is established by combining the well-known Mason model 36, 41 and the transfer matrix method 42. As shown in Fig. 5(a), the acoustic wave is assumed to propagate bidirection- ally along the z axis due to acoustic reflection at interfaces, such that the mechanical displacement u(z, t) can be expressed as 41: u(z, t) = A+e−j(kz−ωt) + A−ej(kz+ωt) (1) where, ω and k = ω/vac (vac is acoustic velocity) are the frequency and acoustic wave number respectively, and A+ and A− are amplitudes for the bidirectional propagating waves. The wave is then related to the velocities v and forces F (or stress σ) at the two surfaces at z1 and z2, working as boundary conditions 41: v1 = v2 = F1 = Sc F2 = Sc du(z1) dt du(z2) dt du(z1) dz du(z2) dz = jω(A+e−jkz1 + A−ejkz1) = jω(A+e−jkz2 + A−ejkz2) = −jSck(A+e−jkz1 − A−ejkz1) = −jSck(A+e−jkz2 − A−ejkz2) 17 (2) (3) (4) (5) Figure 5: Model of acoustic wave propagation in a non-piezoelectric material. (a) Acoustic wave propagates in forward and backward directions in a layer of non-piezoelectric material. The layer thickness is d with two boundaries located at z1 and z2. The wave distribution is solely determined (b) Equivalent circuit model describing by the boundary conditions of force F and velocity v. acoustic wave transmission. (c) The two boundaries can be correlated by a transfer matrix M which is a function of the acoustic impedance and the propagation length. S is the surface area, and c is the stiffness coefficient of the material . The prefactor related with time is omitted for simplicity. After some brief algebra, the forces can be expressed as the combination of velocities as: F1 = F2 = Zac jsin(kd) Zac jsin(kd) (v1 − v2) + jZactan( (v1 − v2) − jZactan( kd 2 kd 2 )v1 )v2 (6) (7) where Zac (=Sρvac, where ρ is the material density) is the acoustic impedance of the material, and d is the thickness. Interestingly, if we treat force and velocity as voltage and current, an equivalent circuit model can be built which satisfies Eq. (6-7) according to Kirchhoff's law, as shown in Fig. 5(b). The circuit consists of three resistors with impedance as labeled in Fig. 5(b). Since the force and velocity must be continuous at the boundary between two different layers, the circuit model makes it easy to cascade different layers by connecting their corresponding circuits. 18 𝒆−𝒋(𝒌𝒛−𝝎𝒕)𝒆𝒋(𝒌𝒛+𝝎𝒕)𝒛𝑧1𝑧2𝑣1𝐹1𝑣2𝐹2dc+-+-𝐹1𝐹2𝑣1𝑣2𝑧𝑎𝑐𝑗sin𝑘𝑑𝑗𝑧𝑎𝑐tan𝑘𝑑2𝑣1−𝑣2(a)(c)(b)c𝑧𝑎𝑐M𝐹1𝐹2𝑣1𝑣2𝑗𝑧𝑎𝑐tan𝑘𝑑2From Eq. (6-7), two adjecent boundaries can be related using a transfer matrix M as 42: F2  = M v2 F2  v2 (8) F1  =  cos(kd) jZacsin(kd) v1 jsin(kd)/Zac cos(kd) In this way, each layer can be represented by its characteristic transfer matrix M [Fig. 5(c)], and the relation between any two boundaries can be connected by multiplying the transfer matrix of each layer in between. The boundary condition at each interface can thus be determined from the very end boundaries of the entire stack structure which, for a general mechanical structure, satisfy the free boundary condition where the force is zero (F = 0), or the fixed boundary condition where the velocity (or equivalently the displacement) is zero (v = 0). After knowing the boundary conditions, the acoustic wave distribution in each layer can be determined from Eq. (2-5) by solving for A+ and A−. This is known as transfer matrix method for solving one-dimensional propagation of acoustic waves in multiple layer structures, which is suitable for our vertical stack structure of HBAR mode. After the derivation of acoustic wave propagation, we are now ready for the model of acoustic wave excitation through a piezoelectric actuator. As we apply voltage to the piezoelectric material, the electric field will generate stress inside the film, which in turn builds up extra charges at the surfaces and change the electric field accordingly. The interplay between stress and electric field can be related as 41: σ = cEε + eE D = eε + E (9) (10) where cE is the stiffness coefficient under constant E, e is piezoelectric coefficient, is dielectric constant, ε is strain, E is electric field, and D is electric displacement. In general, the coefficients are matrices which correlate the mechanical and electric field in different directions. In our case, we consider only the terms related to the z direction. By performing a similar procedure as before, the velocities and forces generated at the bound- 19 Figure 6: Model of acoustic wave propagation in a piezoelectric material. (a) Equivalent circuit Mason model that describes the excitation and propagation of acoustic waves in the piezoelec- tric layer. The three resistors represent the propagation of acoustic waves, while the transformer represents energy conversion between electrical and mechanical domain. (c) Three ports repre- sentation of the piezoelectric actuator. aries can be calculated from the applied external voltage and current through 41: F1 = F2 = V = Zac jsin(kd) Zac jsin(kd) (v1 − v2) + jZactan( (v1 − v2) − jZactan( kd 2 kd 2 )v1 + )v2 + h jω h jω I I 1 jωC0 [I + hC0(v1 − v2)] (11) (12) (13) where C0 (= S/d) is the intrinsic capacitance of the piezoelectric actuator, and h = e/ is a constant related with the material properties. Based on these equations, an equivalent circuit model can be established as shown in Fig. 6(a), which is the so-called Mason model 41. Compared with the previous circuit in Fig. 5(b) which does not consider the piezoelectric effect, a transformer is added to the middle branch with a ratio of 1 : hC0, which then connects to the external power source through series and parallel capacitances C0. The series capacitance has a negative sign which indicates that its current will combine with the external current I and go through the parallel capacitance. This is to be consistent with Eq. (13). The other resistors describe the acoustic wave propagation in the piezoelectric layer as before. The piezoelectric layer can be treated as a three port component as shown in Fig. 6(b), where the mechanical ports are dependent on I-V port. In 20 𝐼c𝑧𝑎𝑐𝐹1𝐹2𝑣1𝑣2𝑉(a)(b)c+-+-𝐹1𝐹2𝑣1𝑣2𝑧𝑎𝑐𝑗sin𝑘𝑑𝑗𝑧𝑎𝑐tan𝑘𝑑2𝑣1−𝑣2-C0C01 : hC0𝑉𝐼+-𝑗𝑧𝑎𝑐tan𝑘𝑑2transfer matrix method, this active component introduces additional boundary conditions through the electric port. Figure 7: Electromechanical model of the actual device in this work. (a) Vertical stacking structure of the whole device. (b) Equivalent circuit model by connecting adjacent layers. The end ports are shorted as required by the free boundary condition. (c) Transfer matrix chain that connects each interface. The input impedance at each port can be correlated and calculated by the multiplication of matrices in between. As we have the circuit model and transfer matrix of each layer, the actual device as described in the main text and shown in Fig. 7(a) can be modeled by simply connecting each adjacent layer. Fig. 7(b) shows the equivalent circuit of the whole device. Free boundary conditions at the top and bottom surfaces are employed such that the forces F (cid:48) 2 and F4 are zero, which correspond to an electric short in the circuit. The impedance looking into one interface can be defined as Z = F/v. 2 = 0 and F4 = 0, the impedance at port 1(cid:48) and 3 from the top By utilizing Eq. (8), and assuming F (cid:48) Al electrode and bottom Si substrate can be calculated easily as: Ztop = jZAltan(kAldAl) Z3 = jZSitan(kSidSi) (14) (15) where ZAl and ZSi are the acoustic impedance of Al and Si respectively. Similarly, by multiplying 21 𝐼𝑉𝐹1′𝐹1𝐹2𝑣1𝑣2𝐹3𝑣3𝐹4𝑣4𝑉𝐼𝐹2′𝑣1′𝑣2′ccccc𝑧Al𝑧AlN𝑧Mo𝑧ox𝑧Si𝐹2′𝑣1′𝑣2′𝐹1′𝐹1𝐹2𝑣1𝑣2𝐹3𝑣3𝐹4𝑣4𝑧top𝑧bot𝑧2𝑧3SiSiO2AlNAl(a)(c)(b)SiSiO2MoAlNAl𝑧inMomatrices of cascaded layers, the impedance at ports 2 and 1 can be calculated: Z2 = j Zbot = ZSitan(kSidSi) + Zoxtan(koxdox) 1 − (ZSi/Zox)tan(kSidSi)tan(koxdox) Z2 + jZMotan(kModMo) 1 + j(Z2/ZMo)tan(kModMo) (16) (17) where Zox and ZMo are the acoustic impedance of SiO2 and Mo respectively. From Eq. (11-13), the electrical impedance Zin (= V /I) can be obtained by 36, 41: (ztop + zbot)sin(kAlNdAlN) + j2[1 − cos(kAlNdAlN)] (cid:26) Zin = 1 1 − k2 t (cid:27) (18) kAlNdAlN (ztop + zbot)cos(kAlNdAlN) + j(1 + ztopzbot)sin(kAlNdAlN) jωC0 where, k2 is the intrinsic electromechanical coupling coefficient of AlN which is 6.5%, ztop (= t Ztop/ZAlN) and zbot (= Zbot/ZAlN) are the impedance from the top and bottom side of AlN which are normalized by the AlN acoustic impedance ZAlN. Electromechanical S11 reflection parameter By applying actual material properties as summa- rized in Table 1, the electrical input impedance can be calculated, from which the S11 response can be calculated as: S11 = Z0 − Zin Z0 + Zin (19) where Z0 (50 Ω) is the standard normalized impedance of the network analyzer. The results are shown in Fig. 8 which demonstrates similarity between measurements and calculations from the electromechanical model. The difference in magnitude is mainly introduced by the calibration and parasitic capacitance from the probe landing during the electrical experiments. The variation of envelope caused by the coupling between Si, SiO2 and AlN cavities is well captured by the model. As mentioned in the main text, the node of the envelope corresponds to a SiO2 resonance. This is because, at the SiO2 resonance, more acoustic energy is confined in SiO2 which is softer and has smaller acoustic impedance as compared to Si. Mechanical dispersion analysis Provided the precision of the model demonstrated in last sec- tion, we can rely on the model and analyze the mechanical dispersion by calculating resonant frequencies. By letting the denominator in Eq. (18) equal zero, we retrieve the parallel resonant frequencies corresponding to maximum resistance 36: (ztop + zbot)cos(kAlNdAlN) + j(1 + ztopzbot)sin(kAlNdAlN) = 0 (20) 22 Table 1: Material properties of each layer employed in the analytic model Material Al AlN Mo SiO2 Si Density ρ (kg/m3) 2700 3300 10200 2200 2329 Velocity v (m/s) 6300 11050 6636 5640 8430 Thickness d (µm) 0.1 0.92 0.1 5.44 231.5 Figure 8: Measured (a) and calculated (b) S11 reflection parameter, showing much similarity in terms of varied envelope and its period. 23 (a)(b)MeasurementCalculation123456-4-20-4.4-4.0-3.6 S11 (dB)Frequency (GHz) S11 (dB)The dispersion equation is transcendental and can be solved numerically. The FSR variation and higher order dispersion can thus be calculated from resonant frequencies as shown in the main text. To get a feeling of why and how the FSR varies, an analytic expression of FSR can be derived under simplified assumptions. Specifically, we can assume the metal thickness is much smaller than the acoustic wavelength, such that their impedance is nearly zero. Under this assumption, the dispersion equation Eq. (20) can be simplified to: ZSitan(kSidSi) + Zoxtan(koxdox) + ZAlNtan(kAlNdAlN) 1 − ZAlN tan(kAlNdAlN)tan(koxdox) Zox = 0 (21) The equation is arranged in the way that SiO2 and AlN are combined and work together as an external cavity coupled to the Si cavity. The FSR can be divided into four regions in terms of resonance and anti-resonance of SiO2 and AlN cavities (see Table 2 and Fig. 9). Firstly, let's consider frequencies around AlN resonance, where kAlNdAlN = pπ + δAlN, with p an integer number and δAlN denoting a small deviation. When the SiO2 is at resonance such that the second term in Eq. (21) is near zero, koxdox = nπ + δox. In this case, Si cavity must satisfy kSidSi = mπ + δSi, such that Eq. (21) becomes 36: ZSiδSi + Zoxδox + ZAlNδAlN = 0 From Eq. (22) we derive the relation between the three cavities as: kSidSi = mπ − Zox ZSi (koxdox − nπ) − ZAlN ZSi (kAlNdAlN − pπ) (22) (23) where kSi = 2πfm/vSi, kox = 2πfm/vox, kAlN = 2πfm/vAlN. This is also true for the m − 1 mode. Note that the mode order p and n of AlN and SiO2 will not change for the m and m− 1 modes. By taking the frequency difference between the m and m − 1 modes, we can get the local FSR after some algebra as: ∆f = ∆f0 dSi dox + ρAlN ρSi dAlN dSi + ρox ρSi (24) where ∆f0 = vSi/(2dSi) is the original FSR of the Si cavity. This new FSR is smaller than the original FSR since the SiO2 and AlN extends the effective cavity length. It is interesting to note 24 that this extension is not simply the physical length of each layer but the effective length weighted by its density relative to Si. Following a similar procedure, the FSR at SiO2's anti-resonance and around AlN's resonance can be obtained. In this time, the three cavities satisfy: kAlNdAlN = pπ + δAlN koxdox = nπ + π/2 + δox kSidSi = mπ + π/2 + δSi By inserting them into Eq. (21), it becomes 36: − Zox −ZSi δSi + + ZAlNδAlN δox 1 + ZAlNδAlN Zoxδox = 0 Since ZAlNδAlN is a very small term, we can ignore it and get: δSi = − ZSi Zox δox − ZSiZAlN Z 2 ox δAlN We can then retrieve the relation between the three cavities as: (25) (26) (27) (28) (29) kSidSi = mπ + π/2 − ZSi Zox (koxdox − nπ − π/2) − ZSiZAlN Z 2 ox (kAlNdAlN − pπ) (30) By taking the frequency difference between fm and fm−1, we can find the local FSR as: (31) ∆f = ∆f0 dSi + Z2 Si Z2 ox (cid:16) ρox ρSi dSi dox + ρAlN ρSi dAlN (cid:17) By comparing Eq. (24) and (31), it can be seen that the extra effective length due to SiO2 and AlN is now multiplied by the square of the ratio between the acoustic impedance of Si and SiO2. Since SiO2 has a smaller impedance than Si, the effective length is longer than before, and the FSR is thus smaller . The cases where the AlN is near its anti-resonance condition kAlNdAlN = pπ + π/2 + δAlN can be calculated in a similar manner which are summarized in Table 2. Note the SiO2 resonance 25 is defined as when there is no Si, and now its resonant condition becomes koxdox = nπ + π/2 + δox due to the π/2 phase introduced from the AlN -- SiO2 interface. Similar as before, the length (FSR) is longer (smaller) around SiO2's anti-resonance as compared to its resonance. By comparing cases between AlN resonance and anti-resonance regions, there is an additional factor of the ratio square between the impedance of oxide and AlN around AlN anti-resonance. Since AlN has a higher impedance than SiO2, the effective length (FSR) around AlN anti-resonance regions is shorter (larger) than the corresponding region around AlN resonance. Based on the effective length expressions, we can find that, for two coupled mechanical cavities, if the impedance of the small cavity is smaller (e.g., Si and SiO2), the effective length (FSR) is shorter (larger) around the small cavity resonance than its anti-resonance. This is also true for the reverse case (e.g., SiO2 and AlN), where the small cavity has larger impedance, and the effective length (FSR) is longer (smaller) around AlN resonance 36. The boundary conditions at the interfaces in each case are also summarized in Table 2. The AlN -- SiO2 interface is free at the AlN resonance and fixed at anti-resonance. This is also true for the SiO2 -- Si interface. The four cases are numbered as 1-4 from top to bottom in Table 2 and labeled and compared with experiments as shown in Fig. 9. Around the AlN resonance, the FSR oscillates between 1 and 2, while at AlN's anti-resonance it varies between 3 and 4, which is consistent with experiments. Finally, as a rule of thumb, since the envelope and FSR are both related with the matching of acoustic impedance, the node (anti-node) of the S11 envelope corresponds to larger (smaller) local FSR. Electromechanical coupling and acousto-optic overlap The model can also be used to estimate the optomechanical S21 response which is dependent on both electromechanical coupling efficiency and acousto-optic overlap. The stress field distribution in AlN will strongly influence the effective electromechanical coupling, which can be estimated by following the method, as found in the literature 36: k2 t,ef f = π2 4 fs fp (1 − fs fp ) (32) 26 Table 2: Effective cavity length at different resonant conditions. The effective boundary condition at each interface is compared. kAlNdAlN pπ pπ kSidSi koxdox AlN -- SiO2 SiO2 -- Si Free nπ nπ+π/2 mπ + π/2 Free Free Fixed mπ pπ +π/2 nπ+π/2 mπ Fixed pπ +π/2 nπ mπ + π/2 Fixed Free Fixed Effective length (cid:16) ρox (cid:16) ρox ρSi ρSi dSi + ρox ρSi dSi + Z2 Si Z2 ox dSi + ρox ρSi dSi + Z2 Si Z2 ox dox + ρAlN ρSi dAlN dox + ρAlN ρSi ρAlN ρSi dox + Z2 ox Z2 AlN AlN dox + Z2 ox Z2 dAlN (cid:17) dAlN (cid:17) ρAlN ρSi dAlN Figure 9: Four different regions are labeled by number in the order from top to bottom correspond- ing to Table 2. The green dashed lines denote the FSR of each region and the corresponding effective length deff is labeled on the right. The bottom insets show the schematics of the acoustic stress wave distribution for each region, illustrating the locations of interfaces relative to the stress wave. 27 1.01.52.02.53.03.54.04.55.017.017.518.0-4.5-4.0-3.5 Frequency (GHz) - (MHz)S11 (dB)𝑑effshortlong3412𝑑Si+𝜌ox𝜌Si𝑑ox+𝜌AlN𝜌Si𝑑AlN𝑑Si+𝑍Si2𝑍ox2(𝜌ox𝜌Si𝑑ox+𝜌AlN𝜌Si𝑑AlN)𝑑Si+𝜌ox𝜌Si𝑑ox+𝑍ox2𝑍AlN2𝜌AlN𝜌Si𝑑AlN𝑑Si+𝑍Si2𝑍ox2(𝜌ox𝜌Si𝑑ox+𝑍ox2𝑍AlN2𝜌AlN𝜌Si𝑑AlN)3412Figure 10: (a) Measured S11 response (top) and calculated effective electromechanical coupling coefficient k2 t,ef f (bottom). The coupling reaches a maximum value of 0.15% around 4.3 GHz. (b- c) Measured S21 response of the TE and TM mode (top) and corresponding normalized product of t,ef f (bottom). When the node of the acoustic stress wave is located at acousto-optic overlap and k2 the center of the waveguide, the acousto-optic overlap integral Γ becomes zero and thus causes notches in the S21 response. where fp is the parallel resonant frequency when the denominator of Eq. (18) equals zero, and fs is t,ef f is thus calculated the series resonant frequency when the numerator of Eq. (18) equals zero. k2 as shown in Fig. 10(a). A maximum value of 0.15% is reached around 4.3 GHz where the AlN resonance is located. It varies with a envelope similar as the S11 measurement. In addition to elecromechanical conversion efficiency, the S21 also depends on acousto-optic overlap which determines the modulation of the optical resonant frequency. According to the perturbation theory, the relative change of resonant frequency can be related with the modulation of refractive index distribution as 29: (cid:82)(cid:82) ∆n(x, y)E(x, y)2dxdy (cid:82)(cid:82) n(x, y)E(x, y)2dxdy ≈ − ∆ω ω (33) where the perturbation of refractive index is caused by the induced stress through the stress-optical effect, and is proportional to the stress via the stress-optical coefficient. In our specific case, the stress is dominated by the vertical stress σz. Therefore, the normalized acousto-optic overlap can 28 123456-30-20-100-120-100-80-60 Frequency (GHz)  x k2t,eff S21 (dB)(a)(b)(c)12345610-310-210-1-4.5-4.2-3.9-3.6k2t,eff (%) Frequency (GHz) S11 (dB)123456-30-20-100-120-100-80-60  x k2t,eff Frequency (GHz) S21 (dB)be approximately estimated as 29: (cid:82)(cid:82) σz(z)E(r, z)2drdz (cid:82)(cid:82) E(r, z)2drdz Γ = (34) where σz is assumed to be dependent only on z and uniform in the r direction. In this way, the distribution of σz can be calculated from the aforementioned 1-D acoustic model. The electric field takes only the dominant component, that is Er for TE mode and Ez for TM mode. To better compare with the measured optomechanical S21 response, the normalized product t,ef f is plotted in Fig. 10 (b-c) for TE and TM modes. Their product can help us to of Γ and k2 explain the variation of S21 qualitatively. For instance, when the node of a stress wave locates at the center of the waveguide, the overlap integral approaches zero due to the vertical symmetry of the optical mode. This causes notches in the S21 response, such as the decreasing of S21 near 2 GHz of the TE mode. The nearly periodic variation of the envelope of S21 is due to the modulation t,ef f. Due to the high confinement of the optical mode in the waveguide, there exhibits only a of k2 small difference of the acousto-optic overlap between the TE and TM modes. On the other hand, the difference of the measured S21 response is mainly caused by the optical Q. Since TE mode shows nearly two times larger Q than TM mode, its response beyond 4 GHz is suppressed by entering into the resolved sideband regime. The perturbation of the local stress field due to the Si2N4 waveguide is not taken into account which requires 2D numerical simulation. This may lead to the difference between measurement and calculation shown in Fig. 10(b-c). Despite this, the analytic model can still give us a good qualitative estimation of the electro-optomechanical response. 2 Investigation of different actuator shapes Devices with different actuator shape are fully explored, and its influence on DC optical resonance tuning and high frequency modulation is investigated. The original design with a disk shaped actuator as described in the main text is shown in Fig. 11(a), which is refered as the Disk device in the following text. Since the stress mainly originates from the corner of the actuator, the microring resonator is positioned at the outer edge of the disk actuator. Under positive DC biasing on the 29 Figure 11: Schematics for the designed devices with (a) Disk shape and (b) Donut shape actu- ators with the silicon nitride ring resonator (blue) having different relative positions. (c) and (d) are the cross-sections for the Disk and Donut devices along black dashed lines in (a) and (b), re- spectively. (e-f) COMSOL simulation of horizontal stress distribution around the optical waveguide under +60 V DC biasing for the Disk and Donut devices respectively. The overlaid gray arrows denote the local mechanical displacement with the biggest arrow scaled to 1 nm. (g-h) Optical microscope images of fabricated devices. (i-j) False color scanning electron micrograph (SEM) images around the actuator corner as labeled in white dashed box in (g) and (h) respectively with color similar to the cross-sections in (c-d). 30 AlSiSiO2AlNMoSiN5µm52.50-2.5-5𝜎𝑟(MPa)SiSiO2AlN 1µmAl 100nmMo 100nmSiN5.4µm3µm3µmDiskDonut(a)(c)(b)(d)(f)(g)(h)10 µm(i)10 µm(j)(e)rztop electrode (while the bottom electrode is always grounded), the AlN film expands and pushes the ring outwards (i.e., ring expanding) [Fig. 11(e)]. Intuitively, the other design is to use a donut-shape actuator which squeezes the microring resonator if it is placed at the inner edge of the actuator, as shown in Fig. 11(b). This is referred as the Donut device in the following. These two designs will show different stress distribution and static resonance tuning as demonstrated later. The cross-sections for the two designs are illustrated in Fig. 11(c-d). The ring is placed 3 µm within the edge of the Disk actuator, while 5 µm for the Donut actuator. The final fabricated devices are shown in Fig. 11(g-h) for the two designs. False color SEMs zoom-in around actuator corners in Fig. 11(i-j) show the relative position between the optical microring resonator (blue) and the AlN actuator (green). The static mechanical simulation is conducted using the finite element method (COMSOL) as shown in Fig. 11(e-f), in which +60 V DC biasing is applied on the top Al layer while the bottom Mo layer is grounded. In this case, a negative electric field Ez forms (points downwards) across the AlN thin film, and the AlN film will be squeezed in the z-direction and expand horizontally (positive Poisson ratio). Fig. 11(e) illustrates the horizontal stress distribution and mechanical dis- placement around the optical waveguide at the upper right corner of Fig. 11(c). As AlN expands, starting from the corner, the stress splits into two parts: extensional stress under the actuator and compressing stress at outside, and the mechanical displacement orients mainly horizontally and points outside the actuator. Similar results can be drawn for Donut device as in Fig. 11(f). De- pending on the position of the waveguide, it experiences extensional, compressing, or the interplay between these two. For the Disk device, the horizontal stress inside the waveguide is a mixture of extension and compression, while for the Donut device, it is mainly extension. Besides the dominant horizontal stress, stresses in other directions will also affect the mod- ulation on the refractive index. As shown in Fig. 12(a) and (b), originating from the corner of top metal, σz exhibit two main lobes with different sign. Inside the actuator, the waveguides in both cases experience extensional stress around 2 MPa. Under positive biasing, the ring of the Disk device will be pushed outwards as the actuator expands, which builds up extensional stress 31 Figure 12: Static stress distribution in z and φ directions. (a) and (b) show the numerical sim- ulation of vertical (z) stress distribution under +60 V DC biasing for the Disk and Donut devices, respectively. Both of them show extensional stress around the waveguide. (c) and (d) are out of plane (tangential to optical ring) stress distributions for Disk and Donut devices. They present different sign inside the waveguide, since the optical ring of the Disk device expands whereas the Donut device squeezes. in the waveguide as in Fig. 12(c). On the other hand, at the inner edge of the Donut actuator, the optical ring is squeezed, generating compressing stress [Fig. 12(d)]. Since shear stress is found to play a less role compared with normal stress23, it is not taken into account in this study. Addi- tionally, when applying negative voltages, the AlN actuator changes from expanding (horizontally) to shrinking, so that all the stresses in the above analysis change sign. In this way, bi-directional tuning can be achieved by reversing the applied voltage's sign, as demonstrated in the following section. The presence of stress will change the refractive index of optical material, and affect differ- ently for distinct polarization of the electric field of light waves, causing so-called birefringence. 32 DiskDonut52.50-2.5-5(MPa)𝜎𝑧𝜎𝜙rz(a)(d)(b)(c)Numerically, they are correlated by stress-optical coefficients by 23: nr = n0 − C1σr − C2(σφ + σz) nφ = n0 − C1σφ − C2(σz + σr) nz = n0 − C1σz − C2(σr + σφ) (35) (36) (37) where, n0 is original refractive index of the material, C1 relates the refractive index and stress that are in the same direction while C2 relates the two that are orthogonal. These equations are appli- cable to isotropic material such as amorphous Si3N4 from low-pressure chemical vapor deposition (LPCVD) used in this work. The lack of the stress-optical coefficient for Si3N4 in the literature makes it difficult to predict precisely the response of the optical ring resonator. However, it would be possible to extract the coefficients by comparing experimental tuning of optical ring resonator with simulation, which is under investigation. DC optical resonance tuning Working as a Si3N4 ring resonator tuner, the static optical reso- nance tuning is performed by applying DC biasing. Fig. 13(a) shows the transmission spectrum of one resonance under different voltages for both the TE and TM modes of the Disk device. One can see that as we apply a positive 60 V the resonance shifts to shorter wavelength (blue curve) relative to the original resonance (black curve), and the tuning changes direction after reversing the voltage (red curve), demonstrating bi-directional tuning ability. Also, it can be observed that the resonance dip only shifts horizontally with little changes of vertical depth, since the waveguide coupling region is unaffected by the opening section of the actuator. The dependence of resonant wavelength on voltages is summarized in Fig. 13(c), showing high linearity. Both TE and TM modes demonstrate similar tuning performances with nearly -12 pm under positive 60 V. This tun- ing range is on a similar order or larger than the linewidth of the high optical Q resonances, which is applicable for Si3N4 microcomb applications such as Kerr comb generation and stabilization. More interestingly, due to different relative positions of ring resonator, the Donut device shows opposite behaviour: the resonant wavelength increases for positive voltages and vice versa, as can be seen in Fig. 13(b) and (d). Here, the slope of tuning with respect to voltage changes from negative to positive. On the other hand, it shows much smaller tuning range with 5.66 pm 33 Figure 13: Transmission spectrum of one resonance of TE and TM polarization mode for (a) Disk and (b) Donut device under +60 V (blue), 0 V (black), and -60 V (red). The x axis represents wavelength shifts relative to resonant wavelength λ0 (∼1550 nm) of each mode. The resonant wavelength decreases for the Disk, while it increases for the Donut under +60 V. The tuning di- rection reverses for opposite voltages, demonstrating bi-directional tuning. (c-d) Dependence of resonant wavelength detune on voltages for the Disk and Donut devices respectively. Experimen- tal results (squares) show high linearity for both TE (pink) and TM (cyan) modes, with R2 > 99% linear fitting (straight lines). The two actuator designs show different tuning directions and tuning ranges. (e-f) Influence of piezoelectric actuation on optical quality factor Q for TE and TM modes of both designs, which verifies that the actuation will not influence the optical Q. 34 0.20.40.60.81.0-40-20020400.00.20.40.60.81.0 Transmission (a.u.) (pm) 0V 60V -60V0.60.81.0-40-20020400.20.40.60.81.0 Transmission (a.u.) (pm) 0V 60V -60V-60-40-200204060-6-4-20246Voltage (V) TE TM (pm)-60-40-200204060-10-50510Voltage (V) TE TM (pm)DiskDonut(a)(b)(c)(d)TETMTETM-60-40-200204060Voltage (V) TE TM-60-40-20020406050.0k100.0k150.0k200.0k250.0k300.0kQuality factor (Q)Voltage (V) TE TM(e)(f)Figure 14: The same S11 and S21 measurements on the Donut device in GHz range. (a) Elec- tromechanical S11 spectrum from 1 to 6 GHz. Optomechanical S21 responses of (b) TE and (c) TM modes demonstrate effective stress-optical modulation spanning a broad range of microwave frequencies. The Donut device shows similar results as reported for the Disk device in the main text. (d) and (e) show the zoom-in of S11 and S21 responses of TE mode within the window (green shaded area) around 2 GHz in (b), while (f) shows the zoom-in of TM mode's S21 response around 4 GHz in (c). (g) Numerical simulation of σz distribution for one typical acoustic resonant mode at 2.125 GHz, with a zoom-in around the optical waveguide (red box) shown in (h). for the TM mode and 1.63 pm for the TE mode under +60 V. The different tuning range of two orthogonal polarization modes, TE and TM, demonstrates tunable birefringence in an otherwise isotropic material by controlling stress, which can be utilized for polarization control 43 or tuning the mode spacing and coupling between a pair of TE and TM modes in a ring resonator. The effect of mechanical actuation on optical Q is plotted in Fig. 13(e-f), which is found to be insignificant. The TE mode shows much higher Q than the TM mode, since the TM mode extends further in the vertical direction which is prone to the absorption of the bottom metal. RF frequency modulation of the Donut-shape device The same measurements described in the main text are also done for the Donut device, including S11 and S21 responses as shown in Fig. 14. No big differences can be found between the Disk and Donut devices in terms of mode distribution, envelope of resonances, and signal to noise ratio. In the zoom-in around 2 GHz in 35 -3.52-3.48-3.442.052.102.152.202.25-110-99-88-77 S11 (dB)S21 (dB)Frequency (GHz)-4.0-3.6-3.2-140-120-100-80123456-100-80-60 S21 (dB)Frequency (GHz)S21 (dB)S11 (dB)(a)(d)(b)(c)(e)(f)2125 MHz50𝜎𝑧(MPa)rz(g)(h)20100-10-20250-25-504.004.054.104.154.20-70-60-50S21 (dB)Frequency (GHz)Fig. 14(e), there are multiple peaks inside each resonance due to existence of higher order acoustic modes. Numerical simulations of one of the fundamental modes at 2.125 GHz is shown in Fig. 14(g), and the zoom-in around waveguide is in Fig. 14(h). From the mode distribution, we can see the acoustic wave is effectively excited and confined under the actuator vertically with little diffraction angle. The good mode confinement guarantees low cross-talk between closely placed actuators. Figure 15: Demonstration of low electromechanical cross-talk between adjacent actuators. (a) Optical microscope image of the device with three closely placed actuators. (b) (top) Two port electromechanical S21 measurement by driving actuator 1 and sensing from actuator 2 as labeled in (a). (bottom) S11 reflection for device 1. The cross-talk between the two devices is as low as -60 dB which guarantees compact integration. (c) Zoom-in of the measured S21 and S11 responses in the green shaded region in (b). 3 Electromechanical cross-talk between adjacent actuators As mentioned in the previous section, the high confinement of the acoustic mode beneath the actuator guarantees low cross-talk between adjacent actuators. To demonstrate this, three actuators are closely placed on the same optical ring resonator, which cover the whole ring in an interval of 120◦, as shown in Fig. 15(a). The one port reflection parameter S11 of actuator 1 is first measured as illustrated in Fig. 15(b). The difference from the S11 shown in the main text is caused by the thicker Si substrate (500 µm) and thicker SiO2 cladding (7 µm). These lead to smaller FSR and period of the envelope. 36 -64-62-60-583.463.483.503.523.54-6.00-5.25-4.50-3.75 S21 (dB) S11 (dB)Frequency (GHz)-80-70-60-5012345-6.0-4.5-3.0-1.5 S21 (dB)S11 (dB)Frequency (GHz)12(a)(b)(c)More importantly, the cross-talk is measured by performing a two-port electromechanical S21 measurement, where we drive actuator 1 and sense the electrical signal out from the adja- cent actuator 2. The cross-talk mainly comes from the leaking of acoustic waves from actuator 1 which can be sensed out by actuator 2 via the piezoelectric effect. The leakage of electric field will also be sensed by device 2 and cause cross-talk between electrical signals of device 1 and 2. As demonstrated in Fig. 15(b), S21 as low as -60 dB of cross-talk is achieved, which illustrates the electrical and mechanical isolation between the two adjacent devices. This low cross-talk en- ables us to fabricate several actuators on the same optical ring, which may realize optical isolation through spatial-temporal modulation 15, 44, or dispersion engineering of Si3N4 microring resonator 19 by engineering stress distribution. 37 References 1. Moss, D. J., Morandotti, R., Gaeta, A. L. & Lipson, M. New cmos-compatible platforms based on silicon nitride and hydex for nonlinear optics. Nature Photonics 7, 597 -- 607 (2013). URL http://dx.doi.org/10.1038/nphoton.2013.183. 2. Brasch, V., Chen, Q.-F., Schiller, S. & Kippenberg, T. J. Radiation hardness of high-q silicon nitride microresonators for space compatible integrated optics. Optics express 22, 30786 -- 30794 (2014). 3. Gaeta, A. L., Lipson, M. & Kippenberg, T. J. Photonic-chip-based frequency combs. Nature Photonics 13, 158 -- 169 (2019). URL https://doi.org/10.1038/ s41566-019-0358-x. 4. Kippenberg, T. J., Holzwarth, R. & Diddams, S. A. Microresonator-based optical frequency combs. Science 332, 555 -- 559 (2011). URL http://science.sciencemag.org/ content/332/6029/555. 5. Huffman, T., Baney, D. & Blumenthal, D. J. High extinction ratio widely tunable low-loss integrated si3n4 third-order filter. arXiv preprint arXiv:1708.06344 (2017). 6. Gundavarapu, S. et al. Interferometric optical gyroscope based on an integrated si 3 n 4 low- loss waveguide coil. Journal of Lightwave Technology 36, 1185 -- 1191 (2018). 7. Yao, Z. et al. Integrated silicon photonic microresonators: Emerging technologies. Journal of Selected Topics in Quantum Electronics 24, 1 -- 24 (2018). IEEE 8. Li, Q. et al. Stably accessing octave-spanning microresonator frequency combs in the soli- ton regime. Optica 4, 193 -- 203 (2017). URL http://www.osapublishing.org/ optica/abstract.cfm?URI=optica-4-2-193. 9. Pfeiffer, M. H. P. et al. Octave-spanning dissipative kerr soliton frequency combs in si3n4 microresonators. Optica 4, 684 -- 691 (2017). URL http://www.osapublishing.org/ optica/abstract.cfm?URI=optica-4-7-684. 10. Liu, J. et al. Ultralow-power chip-based soliton microcombs for photonic integration. Optica 5, 1347 -- 1353 (2018). URL http://www.osapublishing.org/optica/ abstract.cfm?URI=optica-5-10-1347. 11. Stern, B., Ji, X., Okawachi, Y., Gaeta, A. L. & Lipson, M. Battery-operated integrated fre- quency comb generator. Nature 562, 401 -- 405 (2018). URL https://doi.org/10. 1038/s41586-018-0598-9. 12. Raja, A. S. et al. Nature Communications 10, 680 (2019). s41467-019-08498-2. Electrically pumped photonic integrated soliton microcomb. URL https://doi.org/10.1038/ 38 13. Liu, J. et al. Photonic microwave oscillators based on integrated soliton microcombs. arXiv preprint arXiv:1901.10372 (2019). 14. Spencer, D. T. et al. An optical-frequency synthesizer using integrated photonics. Nature 557, 81 -- 85 (2018). 15. Shi, Y., Lin, Q., Minkov, M. & Fan, S. Nonreciprocal optical dissipation based on direction- IEEE Journal of Selected Topics in Quantum Electronics 24, 1 -- 7 dependent rabi splitting. (2018). 16. Xue, X. et al. Thermal tuning of kerr frequency combs in silicon nitride microring res- onators. Opt. Express 24, 687 -- 698 (2016). URL http://www.opticsexpress.org/ abstract.cfm?URI=oe-24-1-687. 17. Joshi, C. et al. Thermally controlled comb generation and soliton modelocking in microres- onators. Opt. Lett. 41, 2565 -- 2568 (2016). URL http://ol.osa.org/abstract. cfm?URI=ol-41-11-2565. 18. Moille, G. et al. Kerr microresonator soliton frequency combs at cryogenic temperatures. arXiv 1906.06554 (2019). 19. Yao, B. et al. Gate-tunable frequency combs in graphene -- nitride microresonators. Nature 558, 410 (2018). 20. Alexander, K. et al. Nanophotonic pockels modulators on a silicon nitride platform. Nature communications 9, 3444 (2018). 21. Ahmed, A. N. R., Shi, S., Zablocki, M., Yao, P. & Prather, D. W. Tunable hybrid silicon nitride and thin-film lithium niobate electro-optic microresonator. Optics letters 44, 618 -- 621 (2019). 22. Datta, I. et al. Low-loss composite photonic platform based on 2d semiconductor monolayers. arXiv preprint arXiv:1906.00459 (2019). 23. Huang, M. Stress effects on the performance of optical waveguides. International Journal of Solids and Structures 40, 1615 -- 1632 (2003). 24. Van Der Slot, P. J., Porcel, M. A. & Boller, K.-J. Surface acoustic waves for acousto-optic modulation in buried silicon nitride waveguides. Optics express 27, 1433 -- 1452 (2019). 25. Hosseini, N. et al. Stress-optic modulator in triplex platform using a piezoelectric lead zir- conate titanate (pzt) thin film. Optics express 23, 14018 -- 14026 (2015). 26. Jin, W., Polcawich, R. G., Morton, P. A. & Bowers, J. E. Piezoelectrically tuned silicon nitride ring resonator. Optics Express 26, 3174 -- 3187 (2018). 27. Midolo, L., Schliesser, A. & Fiore, A. Nano-opto-electro-mechanical systems. Nature nan- otechnology 13, 11 (2018). 39 28. MacQuarrie, E., Gosavi, T., Jungwirth, N., Bhave, S. & Fuchs, G. Mechanical spin control of nitrogen-vacancy centers in diamond. Physical review letters 111, 227602 (2013). 29. Tadesse, S. A. & Li, M. Sub-optical wavelength acoustic wave modulation of integrated photonic resonators at microwave frequencies. Nature communications 5, 5402 (2014). 30. Marpaung, D., Yao, J. & Capmany, J. Integrated microwave photonics. Nature photonics 13, 80 (2019). 31. Obrzud, E., Lecomte, S. & Herr, T. Temporal solitons in microresonators driven by optical pulses. Nature Photonics 11, 600 (2017). 32. Fan, L., Zou, C.-L., Zhu, N. & Tang, H. X. Spectrotemporal shaping of itinerant photons via distributed nanomechanics. Nature Photonics 1 (2019). 33. Supradeepa, V. et al. Comb-based radiofrequency photonic filters with rapid tunability and high selectivity. Nature Photonics 6, 186 (2012). 34. Tsubouchi, K. & Mikoshiba, N. Zero-temperature-coefficient saw devices on aln epitaxial films. IEEE Transactions on Sonics Ultrasonics 32, 634 -- 644 (1985). 35. Luke, K., Dutt, A., Poitras, C. B. & Lipson, M. Overcoming si3n4 film stress limitations for high quality factor ring resonators. Opt. Express 21, 22829 -- 22833 (2013). URL http: //www.opticsexpress.org/abstract.cfm?URI=oe-21-19-22829. 36. Zhang, Y., Wang, Z. & Cheeke, J. D. N. Resonant spectrum method to characterize piezo- ieee transactions on ultrasonics, ferroelectrics, and electric films in composite resonators. frequency control 50, 321 -- 333 (2003). 37. Pinrod, V., Davaji, B. & Lal, A. Coexisting surface and bulk gyroscopic effects. In 2018 IEEE International Ultrasonics Symposium (IUS), 1 -- 4 (IEEE, 2018). 38. Kurizki, G. et al. Quantum technologies with hybrid systems. Proceedings of the National Academy of Sciences 112, 3866 -- 3873 (2015). 39. Zhu, X., Wei, Q., Cheng, Y., Wu, D. & Liu, X. Perfect monochromatic acoustic anti-reflection: A first-principles study. Journal of Applied Physics 121, 094504 (2017). 40. Lu, M.-H., Feng, L. & Chen, Y.-F. Phononic crystals and acoustic metamaterials. Materials today 12, 34 -- 42 (2009). 41. Tirado, J. A. V. Bulk acoustic wave resonators and their application to microwave devices. Ph.D. thesis, Universitat Aut`onoma de Barcelona (2010). 42. Chen, Q. & Wang, Q.-m. Characterization of mechanical and piezoelectric properties of the aln thin film in a composite resonator stuctrue. In 2006 IEEE International Frequency Control Symposium and Exposition, 104 -- 110 (IEEE, 2006). 40 43. Xu, D.-X. Polarization control in silicon photonic waveguide components using cladding stress engineering. In Silicon Photonics II, 31 -- 70 (Springer, 2011). 44. Sohn, D. B., Kim, S. & Bahl, G. Time-reversal symmetry breaking with acoustic pumping of nanophotonic circuits. Nature Photonics 12, 91 (2018). 41
1812.10164	2	1812	2019-04-19T12:26:04	Designing Metagratings Via Local Periodic Approximation: From Microwaves to Infrared	[ "physics.app-ph", "physics.optics" ]	Recently, metamaterials-inspired diffraction gratings (or metagratings) have demonstrated unprecedented efficiency in wavefront manipulation by means of relatively simple structures. Conventional one-dimensional (1D) gratings have a profile modulation in one direction and a translation symmetry in the other. In 1D metagratings, the translation invariant direction is engineered at a subwavelength scale, which allows one to accurately control polarization line currents and, consequently, the scattering pattern. In bright contrast to metasurfaces, metagratings cannot be described by means of surface impedances (or local reflection and transmission coefficients). In this paper, we present a simulation-based design approach to construct metagratings in the "unit cell by unit cell" manner. It represents an analog of the local periodic approximation that has been used to design space-modulated metasurfaces and allows one to overcome the limitations of straightforward numerical optimization and semianalytical procedures that have been used to date to design metagratings. Electric and magnetic metagrating structures responding to, respectively, transverse electric and transverse magnetic incident plane waves are presented to validate the proposed design approach.	physics.app-ph	physics	Designing Metagratings Via Local Periodic Approximation: From Microwaves to Infrared SONDRA, CentraleSup´elec, Universit´e Paris-Saclay, F-91190, Gif-sur-Yvette, France Vladislav Popov∗ Marina Yakovleva Centre de Nanosciences et de Nanotechnologies (C2N), CNRS, Univ. Paris-Sud, Universit´e Paris-Saclay, 10 bvd Thomas Gobert, 91120 Palaiseau, France and The Institute for Nuclear Problems, Belarusian State University, 11 Bobruiskaya Street, 220030, Minsk, Belarus SONDRA, CentraleSup´elec, Universit´e Paris-Saclay, F-91190, Gif-sur-Yvette, France and DEMR, ONERA, Universit´e Paris-Saclay, F-91123, Palaiseau, France Fabrice Boust Jean-Luc Pelouard and Fabrice Pardo Centre de Nanosciences et de Nanotechnologies (C2N), CNRS, Univ. Paris-Sud, Universit´e Paris-Saclay, 10 bvd Thomas Gobert, 91120 Palaiseau, France LEME, UPL, Univ Paris Nanterre, F92410, Ville d'Avray, France Shah Nawaz Burokur Recently, metamaterials-inspired diffraction gratings (or metagratings) have demonstrated un- precedented efficiency in wavefront manipulation by means of relatively simple structures. Conven- tional one-dimensional (1D) gratings have a profile modulation in one direction and a translation symmetry in the other. In 1D metagratings, the translation invariant direction is engineered at a subwavelength scale what allows one to accurately control polarization line currents and, conse- quently, the scattering pattern. In bright contrast to metasurfaces, metagratings cannot be described by means of surface impedance densities (or local reflection and transmission coefficients). In this paper, we present a simulation-based design approach to construct metagratings in the "unit cell by unit cell" manner. It represents an analog of the local periodic approximation (LPA) that has been used to design space modulated metasurfaces and allows one to overcome the limitations of straightforward numerical optimization and semi-analytical procedures that have been used up to date to design metagratings. Electric and magnetic metagrating structures responding to respec- tively transverse electric (TE) and transverse magnetic (TM) incident plane-waves are presented to validate the proposed design approach. I. INTRODUCTION In the last decade there has been tremendous interests in metasurfaces due to their amazing capabilities in ma- nipulating electromagnetic fields [1 -- 3] and broad range of potential applications [3 -- 6]. Metasurfaces are rep- resented by dense distributions of engineered subwave- length scatterers on a surface being planar analogs of metamaterials. When the characteristics of a metasur- face are spatially modulated, it can perform wavefront transformations [7, 8]. The local periodic approximation (LPA) plays a crucial role in designing such metasur- faces and has been already used for long time [9, 10]. Essentially, it serves to estimate scattering properties of a unit cell embedded in a nonuniform array. To that end, a unit cell is placed in the corresponding uniform array whose reflection and transmission coefficients are ∗ [email protected] then attributed to the unit cell in a nonuniform array. Scattering parameters of a uniform array are usually cal- culated from full-wave numerical simulations. However, there are particularly simple cases (e.g. metallic patches) that can be treated analytically [11, 12]. Recently, metamaterials-inspired diffraction gratings have demonstrated unprecedented efficiency in manipu- lating scattering patterns with relatively simple struc- tures [13 -- 18]. Reflecting configuration of a metagrating represented by a one-dimensional periodic array of thin "wires" placed on top of a metal-backed dielectric sub- strate is illustrated in Fig. 1 (a). Generally, each period of a metagrating consists of N different "wires". Note- worthy, the distance d between adjacent "wires" always remains of the order of operating wavelength λ, which does not allow one to introduce neither averaged sur- face impedances nor local reflection coefficient in contrast to metasurfaces. Incident wave excites electric or mag- netic polarization line currents in "wires" that can be controlled by judiciously adjusting the electromagnetic response of the "wires". Consequently, it becomes possi- ble to manipulate diffraction orders. It has been demon- strated that it is possible to achieve perfect nonspecular reflection and beam splitting with a single "wire" per period [14, 15, 17]. In a more general manner, it was shown that in order to perfectly control diffraction pat- terns one needs two degrees of freedom (represented by reactively loaded "wires") per each propagating diffrac- tion order [18]. However, even having the number of re- active "wires" per period equal to the number of propa- gating diffraction orders enables to perform efficient mul- tichannel reflection, as demonstrated in [16]. Practically, a "wire" is constituted from subwavelength meta-atoms arranged in a dense uniform 1D array . Ge- ometrical parameters of meta-atoms determine electro- magnetic response of a "wire". Up to date metagratings have been designed either by performing 3D full-wave nu- merical optimization of a whole metagrating's period [14] or semi-analytically [15, 16, 18 -- 20]. While the first ap- proach can be very time consuming when it comes to designing metagratings having many "wires" per period, the second one allows one to consider only very simple meta-atoms such as printed capacitors [15] or dielectric cylinders [19]. In this paper, we develop the local peri- odic approximation for designing metagratings with the help of 3D full-wave numerical simulations. In compar- ison to a straightforward numerical optimization, it sig- nificantly reduces the time spent on the design of meta- gratings since within the LPA one deals with a single unit cell at a time [meaning that we consider a uniform array formed by a given unit cell as illustrated by Fig. 1(b)]. Contrary to simple analytical models, the LPA also al- lows one to deal with complex meta-atom's geometries and accurately account for interactions between adjacent "wires". The outline of the paper is as follows. In the second section we describe a retrieval technique which is used to extract characteristics of a "wire" from scattering param- eters of the corresponding uniform array. We consider "wires" possessing either electric or magnetic responses which allows one to deal with both TE and TM polar- izations. In the same section, we outline a model that can be used in a full-wave simulation software to con- struct a look-up table connecting retrieved characteristics of "wires" with corresponding parameters of meta-atoms. The third section is devoted to validation examples where we demonstrate metagratings' designs operating in mi- crowave and infrared frequency domains. In the fourth section we discuss remained challenges and conclude the paper. II. LOCAL PERIODIC APPROXIMATION FOR METAGRATINGS We consider reflecting metagratings operating either under TE or TM incident wave polarization. There- fore, each "wire" composing a metagrating can be char- 2 FIG. 1. (a) Schematic diagram of a metagrating: a periodic array of thin "wires" placed on a dielectric substrate backed by a metal plate and having relative permittivity εs, perme- ability µs and thickness h. The array is excited by a plane- wave incident at an angle θ. (b) Schematic diagram of a uni- form array of "wires" characterized by the same impedance density Zq. The inset represents the different meta-atoms composing a "wire". (c) Principal model used in numerical simulations to calculate the reflection coefficient from a uni- form array of "wires" implemented with meta-atoms. acterized by a scalar electric impedance density Zq (or scalar magnetic admittance density Yq in the case of a "wire" possessing magnetic response). However, the ap- proach can be readily generalized to transmitting meta- gratings as it is discussed in the last section, polarization insensitive metagratings can be developed by designing "wires" having both electric and magnetic responses and applying the method developed further for each polar- ization separately. Previously, one has distinguished be- tween load- and input-impedance densities [15, 16, 18] which we do not separate in the present study dealing only with impedance density as the principal charac- teristic of a "wire". To be more accurate, we assume that the impedance density represents the sum of the load-impedance density and reactive part of the input- impedance density. According to the LPA, in order to find impedance density, a "wire" is placed in the corre- sponding uniform array (period d < λ) illuminated by a plane-wave incident at angle θ as illustrated by Fig. 1 (b). We start by describing a way to retrieve electric impedance Zq and magnetic admittance Yq densities from scattering parameters. Since Zq (Yq) represents itself as a complex number and an electric (magnetic) line current radiates TE (TM) wave, it is sufficient to measure only the complex amplitude of the specularly reflected TE (TM) plane-wave. Noteworthy, time dependence exp[iωt] is assumed throughout the paper. 3 FIG. 2. Examples and characteristics of unit cells that can be used to compose "wires" in a metagrating. The top row represents the schematics of printed capacitor (a) and inductances (c) and (e). The bottom row demonstrates impedance densities found by means of the LPA as functions of geometrical parameters of unit cells and corresponding (from (b) to (f)), respectively, to "wires" built up from printed capacitor shown in figure (a) and printed inductances in figures (c) and (e). Other parameters are fixed: dielectric substrate is F4BM220 of permittivity εs = 2.2(1 − 10−3i) and thickness h = 5 mm, B = 3.75 mm, d = 15 mm, w = 0.25 mm, (B/C = 3 in figure (d) and B/C = 5 in figure (f)). Working frequency is set to 10 GHz, corresponding to the vacuum wavelength of 30 mm. Normally incident plane-wave is assumed where electric field is oriented along the B dimension. A. Electric response, TE polarization An electric polarization line current I (excited by TE plane-wave in a "wire" composing the uniform array) is linked to the complex amplitude ST E of the electric field of the specularly reflected wave via the following formula (see Appendix A) 0 I = − 2d kη (ST E 0 − RT E (1 + RT E 0 0 e2iβ0h)β0 )eiβ0h . (1) 0 Here, k and η are the wavenumber and characteristic impedance outside the substrate, RT E is the Fresnel's reflection coefficient from the metal-backed substrate of a TE plane-wave at incidence angle θ and β0 = k cos(θ). Since all "wires" are assumed to be very thin, they are modeled as lines represented mathematically by the Dirac delta function δ(y, z). Consequently, the interaction with the substrate and between adjacent "wires" can be taken into consideration analytically by means of the mutual- impedance density Zm (see Appendix A). It allows one to obtain the characteristic of a "wire" itself and not of the array. "Wire's" electric impedance density is found by means of Ohm's law leading to the following expression for Zq Zq = E0 I − kη 4 − Zm, (2) 0 where E0 = (1 + RT E ) exp[iβ0h] represents the value of the external electric field (incident wave plus its reflec- tion from the metal-backed substrate) at the the "wire" located at y = 0 and z = −h. The radiation resistance of a "wire" is equal to kη/4 being independent on its particular implementation as it follows from power con- servation conditions [21]. It is important to note that mutual-impedance density depends only on the period of the uniform array and parameters of the metal-backed substrate, but does not depend on the current I. It means that the impedance density given by Eq. (2) accurately represents the characteristic of the corresponding "wire" in a nonuniform array as long as the distance between adjacent "wires" and the metal-backed substrate remain the same as those of the uniform array. However, we leave open the questions of accuracy when considering "wires" built up from finite size meta-atoms as infinitely thin and estimating mutual interactions by means of analytically calculated mutual-impedance density. B. Magnetic response, TM polarization The case of TM polarization and "wires" possessing magnetic response can be treated with the help of duality relations [22]: E → H, H → −E, I → V and η → 1/η. Since the metal-backed dielectric substrate is not replaced by the corresponding dual equivalent, we have to additionally make the following substitution RT E 0 → RT M . Thus, from Eq. (1) one can arrive at the formula for retrieving the magnetic current V from the complex amplitude of the magnetic field of the specularly reflected plane-wave (see Appendix B) 0 V = − 2dη k (ST M 0 − RT M (1 + RT M 0 0 e2iβ0h)β0 )eiβ0h , (3) 0 where RT M is the Fresnel's reflection coefficient from the metal-backed substrate of a TM-polarized plane-wave at incidence angle θ. As previously, the interaction with the substrate and between adjacent "wires" can be taken 4 obtained by modelling wires as infinitesimally thin, all practical features of unit cells (like dielectric and con- duction losses, finite thickness of metallic traces, etc.) are taken into account by means of numerical simula- tions. The geometry of the model consists of two princi- pal parts: a tested unit cell (illustrated by a printed in- ductance on a metal-backed dielectric substrate in Fig. 1 (c)) and air region. Periodic boundary conditions are imposed on the side faces (as shown in Fig. 1 (c)). The model is excited by a periodic port assigned to the face of the air region in opposite to the unit cell, as highlighted in red color in Fig. 1 (c). The periodic Port creates a plane- wave incident at angle θ. It is important to take into account θ since meta-atoms are usually spatially disper- sive (impedance density depends on the incidence angle, see for instance Ref. [10]). The thickness of the air region equals operating vacuum wavelength λ which is normally enough to eliminate higher order evanescent modes. The periodic port is also used as a listening port to calculate the scattering parameter S11. It is related to the complex 0 e−2iβ0(h+λ). In the amplitude S0 in Eq. (1) as S11 = ST E case of TM polarization S11 = −ST M 0 e−2iβ0(h+λ). FIG. 3. (a) Schematic of a unit cell based on two split ring resonators and having an inversion center. (b) Zoom view of the magnetic meta-atom with defined parameters. (c) Calcu- lated by means of the LPA, admittance density of the "wire" built up from the magnetic meta-atoms versus the outer ra- dius Rout of the split ring resonator. Other parameters are fixed: no substrate (air is considered as spacer), h = 3.75 mm, B = 3.75 mm, d = 15 mm, s = 0.15 mm, g = 0.20 mm. Working frequency is set to 10 GHz (vacuum wavelength of 30 mm). Normally incident plane-wave is assumed where mag- netic field is oriented along the B dimension. III. EXAMPLES In order to validate the developed approach, we first employ it to construct look-up tables that are further used to implement metagratings for controlling propagat- ing diffraction orders. In order to find impedance den- sities necessary to establish a given diffraction pattern we use the theoretical approach developed in Ref. [16]. In what follows, we focus on two frequency ranges: mi- crowave (operating vacuum wavelength 30 mm) and in- frared (operating vacuum wavelength 4 µm). into account by means of the mutual-admittance density Ym calculated analytically (see Appendix B). Then, the magnetic admittance density Yq can be found as Yq = H0 V − k 4η − Ym. (4) Here H0 = (1 + RT M ) exp[iβ0h] is the value of the ex- ternal magnetic field at the "wire" located at y = 0 and z = −h, k/(4η) represents the radiation conductance. 0 C. Look-up table As it is stated above, in practice a "wire" would be implemented with subwavelength meta-atoms arranged in a line. The ultimate goal of the developing approach is to construct a look-up table linking geometrical pa- rameters of meta-atoms with corresponding impedance (admittance) densities. To that end, we harness 3D full- wave numerical simulations (in our examples COMSOL MULTIPHYSICS is used). Although, Eqs. (1) -- (4) are A. Microwave frequency range We start by considering simple meta-atoms repre- sented by printed capacitance and inductance (illus- trated in Fig. 2) that have already been used to imple- ment metagratings at microwave frequencies by means of printed circuit board (PCB) technology [17, 18]. In the simulations, we take into account practical aspects of the design such as finite thickness of the copper traces (tm = 35 µm) and dielectric losses introduced by the substrate (F4BM220 in our examples, εs = 2.2 with loss tangent 10−3). Figure 2 depicts the impedance densities calculated by means of the developed LPA at 10 GHz (vacuum wavelength λ is 30 mm). It is seen that having printed capacitance and inductance one is able to cover a broad range of impedance densities (imaginary part) that is normally enough to realize any diffraction pattern for TE polarization. It is important to note, that "wires" can exhibit significant resistive response (see Fig. 2 (f) at the resonance) which should be kept in mind while designing metagratings. When comparing the numerical 5 illustrates the schematics of the unit cell which at close look consists of two SRRs separated by a short distance as seen from Fig. 3 (b). The unit cell has an inversion center allowing to eliminate the bianisotropic response attributed to single and double SRRs [23, 24]. In order to adjust the response of a "wire" represented by a 1D array of SRRs, we use the outer radius Rout as a free parameter. The result of applying the LPA to find an equivalent admittance density is depicted in Fig. 3 (c). Due to the small separation distance between the two SRRs we are able to achieve strong magnetic response with the outer radius being of the order of λ/20. As in the case of TE polarization discussed above we see that when approaching the resonance there is an increase of the real part of the "wire's" admittance density resulting in enhanced absorption. In order to validate calculated impedance and admit- tance densities we develop designs of metagratings based only on the data from Figs. 2 and 3 to construct a certain diffraction pattern. Particularly, we demonstrate a split- ting of normally incident plane-wave equally between four propagating diffraction orders (-3rd, -1st, +2nd and +3rd) while suppressing the rest three (-2nd, 0th and +1st), as illustrated by the schematic in Fig. 4 (a). In accordance with Ref. [16], to that end we need the number of reactive "wires" per period equal to the number of propagating diffraction orders, i.e. seven. Although Ref. [16] deals only with electric line currents and TE polarization, it is straightforward to generalize the approach to magnetic currents and TM polarization (See Appendix B). Fig- ures 4 (b) and (c) demonstrate the frequency response of electric and magnetic metagratings designed for 10 GHz. Overall, one can see that despite all practical limitations, the designed metagratings almost perfectly perform the desired splitting of the incident wave. It is important to note that the response of the metagrating operating under TE polarization (Fig. 4 (b)) shows a broader re- sponse than the one operating under TM polarization (Fig. 4 (c)). It is naturally explained by the resonant be- havior of the considered SRR-based unit cell (see Fig. 3 (b)). Indeed, printed capacitance and inductance (illus- trated, respectively, by Figs. 2 (a) and (c)) used for the metagrating dealing with TE polarization do not exhibit resonances (see Figs. 2 (b) and (d)). The other feature of the magnetic metagrating is the significant absorption (compared to the case of TE polarization) which, how- ever, does not deteriorate the overall performance. absorption FIG. 4. (a) Schematic of a metagrating having period 7λ/2 (λ is the operating vacuum wavelength) and exciting seven propagating diffraction orders under normally incident plane- wave. The red and green beams represent suppressed and equally excited orders, respectively. (b), (c) Simulated fre- quency response (normalized power scattered in propagating diffraction orders versus frequency) of the metagrating oper- ating under (b) TE and (c) TM polarizations and establishing the diffraction pattern illustrated by figure (a). Both meta- gratings are designed to operate at 10 GHz. Impedance and admittance densities as well as geometrical parameters of the "wires" composing the metagratings are given in Appendix C. results with analytical models used in Refs. [15, 18], one would see very good agreement for the imaginary part of the impedance density at small values of parameter A. The resonance observed in Fig. 2 (f) appears when one decreases the parameter C and it cannot be found by simple analytical formula used in Ref. [18]. In order to deal with TM polarization, we harness split ring resonators (SRRs) excited by the magnetic field and, thus, having effective magnetic response. Figure 3 (a) B. Infrared frequency range In this subsection we give an example of possible de- signs of unit cells that can be used as building blocks for metagratings operating at infrared frequencies. In what follows, the operation frequency is set to 75 THz corre- sponding to the vacuum wavelength of 4 µm. In order to implement capacitive and inductive unit cells for in- frared domain we consider metallic (gold) patches and 6 of the impedance density remains very small due to non- resonant response of the unit cells even though we take into account absorption in gold and silicon dioxide. It is worthwhile to note that a single straight metal- lic wire may not be enough in case strong inductive re- sponse is necessary (large positive imaginary part of the impedance density). Cross section area is usually re- stricted by fabrication tolerances that does not allow one to infinitely reduce it. Meandering can be a solution (see Figs. 2 (c) and (e)) though it might complicate the fab- rication. To validate calculated impedance densities we demon- strate a metagrating equally splitting a normally incident plane-wave between three propagating diffraction orders (-2nd, 0th and +3rd). The rest four propagating orders are cancelled, as presented by the schematic in Fig. 6 (a). Simulated frequency response of the metagrating is demonstrated in Fig. 6 (b). The incident wave equally (with an accuracy of a few percentages) excites chosen diffraction orders. One can see that the scattering in the -3rd, -1st, +1st and +2nd orders remains suppressed in a wide frequency range due to the nonresonant nature of the unit cells and that absorption constitutes only 2% of total power. The 5% level of scattering losses is rather a drawback of the approach presented in Ref. [16], but can still be reduced as proposed by the design methodology of Ref. [18]. To conclude this subsection, let us make a few prac- tical comments. In the numerical simulations, dielectric permittivities of gold and silicon dioxide were taken from Refs. [25] and [26], respectively. Silicon dioxide was cho- sen as a substrate due to low losses and low refractive index at 4 µm. It allows one using a thick substrate while avoiding excitation of waveguide modes. On the other hand, if the dielectric substrate is thin, it may not be possible to model "wires" as having infinitely small cross section area and accurately account for the interac- tion with the substrate. Another practical feature is that between gold parts and a silicon dioxide substrate there is usually a thin chromium (or titanium) adhesion layer of a few nm thickness. Although all practical aspects should be taken into account while designing an exper- imental sample, we do not expect that the influence of such a thin intermediate layer can lead to a qualitative change and, therefore, did not consider it. FIG. 5. (a), (b) Schematic diagrams of (a) two gold patches and (b) a gold "wire" exhibiting capacitive and inductive re- sponses, respectively. The gold elements are placed on a sili- con dioxide layer backed by a gold plating. (c), (d) Impedance densities calculated by means of the LPA as functions of ge- ometrical parameters of the unit cells and corresponding, re- spectively, to "wires" built up from gold patches shown in figure (a) and gold wires in figure (b). Other parameters are fixed: silicon dioxide layer has permittivity εs ≈ 1.93 and thickness h = 700 nm, d = 2 µm, w = 200 nm, tm = 200 nm (tm = w in case of figure (d)), g = 350 nm. Working frequency is set to 75 THz (vacuum wavelength of 4 µm). Normally incident plane-wave is assumed where electric field is oriented along the B dimension. wires. Gold elements are placed on a dielectric substrate (silicon dioxide) backed with gold as illustrated in Figs. 5 (a) and (b). The capacitive response is attributed to the gap between two patches. By changing the size of the gap or the size of patches one is able to adjust the capac- itance. The inductance of a metallic wire is determined only by the cross section area. Although the design of the unit cells is relatively simple, it allows one to obtain impedance densities in a quite wide range of values as shown in Figs. 5 (c) and (d). Interestingly, the real part IV. DISCUSSION AND CONCLUSION In this work we have presented a simulation-based de- sign approach to construct metagratings in the "unit cell by unit cell" manner. It represents an analog of the lo- cal periodic approximation that has been used to design space modulated metasurfaces and in comparison to a brute force numerical optimization (that deals straight with a whole supercell of a metagrating) can consider- ably reduce the time spent on the design. Indeed, as- sume that each unit cell constituting a N -cells period of 7 count the impact of the metal-backed dielectric substrate and the interaction between "wires" composing the meta- gratings. It makes the local periodic approximation not only fast but also a rigorous approach to design meta- gratings represented by nonuniform arrays of "wires" (in bright contrast to metasurfaces). In this work, we have considered a reflecting configu- ration of a 1D metagrating but the developed design ap- proach can be generalized to deal also with transmitting and 2D metagratings. Up to date there are two meth- ods to control transmission with metagratings: either by means of an asymmetric three-layer array [27] of electric- only "wires" or a single-layer array of bianisotropic par- ticles [28]. In order to deal with 2D metagratings one would have to consider a 2D periodic array of point scat- terers instead of a 1D array of line currents. In light of recently demonstrated acoustic metagratings [29, 30] we also expect that the LPA can be adjusted to design acoustic wavefront manipulation devices. To conclude, wavefront control by means of metagrat- ings is not only an innovative and efficient way to re- alize the transformation between incident and reflected and/or transmitted waves, but it also allows one to further reduce the fabrication complexity while signifi- cantly improving the performance and functionality of metagratings-based integrated optics components and re- configurable antennas in microwave communication sys- tems. In this respect, the developed approach can be an essential tool when designing complex passive or active metagratings across all electromagnetic spectrum. ACKNOWLEDGEMENT M.Y., J-L.P. and F.P. acknowledge financial support from ANR (grant mEtaNiZo). Appendix A: Retrieval of electric current and impedance density from specular reflection In this section, we derive Eqs. (1) and (2). To that end we use analytical formulas for the electric field scattered by a N -cells metagrating given in Ref. [16]. The system is excited by an incident plane wave with the electric field e−ik sin θy−ik cos θz+iωt being along the x-direction. Since in the LPA we simulate only a single "wire" per period, the electric field of the wave reflected from the periodic array depicted in Fig. 1 (b) can be represented via plane- wave expansion as follows +∞ kη 2d Ex(y, z) = −I Xm=−∞ 0 e−iξ0y+iβ0(z+2h), +RT E (1 + RT E m ) βm e−iξmy+iβm(z+h) (A1) where ξm = k sin θ + 2πm/d and βm = pk2 − ξ2 m are, respectively, tangential and normal components of the wave vector of the mth diffraction order. The Fresnel's absorption FIG. 6. (a) Schematic diagram of a metagrating having pe- riod 7λ/2 (λ is the operating vacuum wavelength) and excit- ing seven propagating diffraction orders under normally inci- dent plane-wave. The red and green beams represent sup- pressed and equally excited orders, respectively. (b) Sim- ulated frequency response (normalized power scattered in different propagating diffraction orders versus frequency) of the metagratings operating under TE polarization and es- tablishing the diffraction pattern illustrated by figure (a). The metagrating is designed for 75 THz operating frequency. Impedance densities as well as geometrical parameters of the "wires" composing the metagrating are given in Appendix C. a metagrating has only one parameter to be adjusted and this parameter takes in total P different values during a parametric sweep. Then, one would have to perform P N full-wave simulations in order to cover all possible com- binations of parameters of unit cells and find optimal configuration of a metagrating's supercell [assuming no "smart" algorithms (like genetic) are harnessed]. Clearly, this number strongly depends on the number of unit cells in a supercell N . Meanwhile, the LPA deals with one unit cell at a time what makes the number of required full- wave simulations can be as small as P and it does not depend on the number N . We have validated the devel- oped approach via 3D full-wave numerical simulations by demonstrating designs of metagratings controlling scat- tering patterns at microwave and infrared frequencies for both TE and TM polarizations. The both electric meta- gratings (TE polarization) required two different types of unit cells (capacitive and inductive) and overall we have performed only 2P simulations, while the magnetic meta- grating required one type of a unit cell and, thus, only P simulations. Finally, simple and accurate analytical model describing metagratings allows one to take into ac- reflection coefficient from the substrate backed by per- fect electric conductor (PEC) is given by the following formula RT E m = iγT E iγT E m tan[βs m tan[βs mh] − 1 mh] + 1 , γT E m = ksηsβm kηβs m , (A2) where η = pµ/ε and ηs = pµs/εs. When a metal backing the dielectric substrate cannot be modeled as PEC, one has to correspondingly modify the reflection coefficient. The period d and incidence angle θ are such that the only propagating diffraction order corresponds to the specular reflection (m = 0). Then, the amplitude ST E of the zeroth diffraction order can be found from Eq. (A1) to be 0 ST E 0 = RT E 0 e2iβ0h − I kη 2d 1 + RT E 0 β0 eiβ0h. (A3) This formula is then used to express the current I leading to Eq. (1). In order to calculate the impedance density of a "wire" one needs to find the ratio between the total electric field E(loc) at the position of a "wire" and the current I in the "wire" (and then subtract the radiation resistance). E(loc) can be represented by the sum of the external elec- tric field E0, the electric field created by all other line cur- rents and the field resulted from the reflection from the metal-backed substrate. The last two can be united and expressed as −ZmI with the mutual-impedance density Zm given by the following formula Zm = + kη 2d +∞ kη 2 Xn=1 Xm=−∞ +∞ RT E m βm cos[k sin[θ]nL]H (2) 0 [knd] . (A4) Thus one arrives at the Eq. (2). 8 0 hkpy2 + z2i is the Hankel function of the where H (2) second kind and zeroth order. Consequently, the mag- netic field radiated by a periodic array of N magnetic line currents per period Mnq(r) = Vq exp[−ik sin[θ]nL]δ(y − ynq, z)x0 (see Fig. 1 (a)) is given by the series of Hankel functions Hx(y, z) = − k 4η N +∞ Xq=1 Xn=−∞ Vqe−ik sin[θ]nL H (2) 0 [kp(y − nL − (q − 1)d)2 + z2], (B2) where ynq = nL + (q − 1)d, n and q take integer values from −∞ to +∞ and from 1 to N , respectively. Since the array is periodic the field given by Eq. (B2) can be expressed via the series of plane-waves by means of the Poisson's formula (see Ref. [16]) as follows Hx(y, z) = − k 2ηL +∞ Xm=−∞ ρ(V ) m βm e−iξmye−iβmz, (B3) where ρ(V ) m = PN q=1 Vq exp[iξm(q − 1)d]. The effect of the metal-backed dielectric substrate on the field radiated by the array Mnq(r) = Vq exp[−ik sin[θ]nL]δ(y − ynq, z + h)x0 can be derived fol- lowing Ref. [20]. After some algebra, one would arrive at the following expressions for the magnetic field profile outside the substrate (z < −h) Hx(y, z < −h) = ρ(V ) m (1 + RT M m ) +∞ k − 2ηL Xm=−∞ βm e−iξmy+iβm(z+h). (B4) Here RT M m is the Fresnel's reflection coefficient of a plane- wave (having tangential component of wave vector equal to ξm) from the metal-backed dielectric substrate Appendix B: Magnetic metagratings RT M m = − iγT M iγT M m tan[βs m tan[βs mh] − 1 mh] + 1 , γT M m = kηsβs m ksηβm . (B5) A reflecting metagrating possessing magnetic-only re- sponse is modeled as a periodic array of magnetic line currents placed on top of a metal-backed dielectric sub- strate. In the first two subsections we derive analytical formulas (following Ref. [16]) for the field radiated by such a system and describe the way towards control of diffraction patterns with magnetic metagratings. In the last subsection we derive Eqs. (3) and (4). 1. Radiation of an array of magnetic line currents A single magnetic line current M(r) = V δ(y, z)x0 ra- diates a TM-polarized wave with the magnetic field being along the x-direction (see Ref. [22]) Hx(y, z) = − k 4η V H (2) 0 hkpy2 + z2i , (B1) 2. Controlling diffraction patterns with magnetic metagratings When a magnetic metagrating is illuminated by a TM- polarized plane-wave incident at angle θ, the scattered magnetic field can be represented as a superposition of m e−iξmy+iβmz. Amplitudes of the plane-waves are found analytically (see previous sub- section) and can be expressed as follows plane-waves P+∞ m=−∞ ST M ST M m = − k 2ηL (1 + RT M m )eiβmh βm ρ(V ) m + δm0RT M 0 e2iβ0h (B6) where δm0 is the Kronecker delta representing the re- flection of the incident wave from the metal-backed sub- strate. As in case of TE polarization described in Impedance density (η/λ) Microwaves: TE Infrared: TE Admittance density (ηλ) Microwaves: TM Geometrical parameters (mm) Z1 −i4.87 −i0.30 Y1 −i3.61 A1 Z2 −i1.16 −i3.06 Y2 i4.42 A2 Z3 i6.88 −i3.20 Y3 i0.89 A3 Z4 i17.16 i3.61 Y4 i3.05 A4 Z5 −i0.33 −i1.63 Y5 i0.31 A5 Z6 i3.97 −i1.04 Y6 i0.71 A6 Z7 −i0.81 −i11.59 Y7 i4.51 A7 Microwaves: TE 1.48 (cap. uc) 2.74 (cap. uc) 3.08 (ind. uc) 7.88 (ind. uc) 3.26 (cap. uc) 10.33 (cap. uc) 2.94 (cap. uc) Geometrical parameters (mm) Microwaves: TM Rout,1 1.348 Rout,2 1.4065 Rout,3 1.382 Rout,4 1.397 Rout,5 1.378 Rout,6 1.381 Rout,7 1.4071 9 Geometrical parameters (nm) Infrared: TE w1, B1, g1 w2, B2, g2 175, 1600, 350 59, 800, 100 w3, B3, g3 w4, tm,4 = w4 w5, B5, g5 53, 800, 100 73 (ind. uc) 157, 800, 100 260, 800, 100 137, 800, 350 w6, B6, g6 w7, B7, g7 TABLE I. Parameters of the metagratings demonstrated as examples in Section III. The indexes correspond to the numbered "wires" in Figs. 4 (a) and 6 (a). The rest parameters are fixed and given in the captions to Figs. 2, 3 and 5. Where it is necessary the type of a used unit cell (uc) is specified in the brackets. In the example demonstrated by Fig.4(b), inductive unit cells are represented only by meanders with B/C = 3, as illustrated in Fig. 2(c). Ref. [16], Eq. (B6) demonstrates that magnetic line cur- rents contribute to the scattered plane-waves via the pa- rameter ρ(V ) m . Magnetic line currents Vq can be used to control plane-waves scattered in the far-field (ξm < k). Since in practice it is more straightforward to deal with passive structures and also the excitation of magnetic line currents can be of particular difficulty, we are in- terested in the case when magnetic line currents are po- larization currents excited by the incident plane wave in thin "wires" characterized by admittance densities Yq. Then, necessary currents Vq can be obtained by choosing admittance densities Yq from the following equation by the following series − k 2η Vq +∞ Xn=1 cos[k sin[θ]nL]H (2) 0 [knL]. (B8) The magnetic field created by all other line currents can be accounted for as follows − k 4η N Xp=1,p6=q Vp +∞ Xn=−∞ e−ik sin[θ]nLH (2) 0 [k(q − p)d − nL]. (B9) Eventually, the waves reflected from the metal-backed substrate create the following magnetic field at the loca- tion of the qth "wire" (zeroth period) (cid:18)Yq + k 4η(cid:19) Vq = Hq − N Xp=1 Y (m) qp Vp. (B7) − k 2ηL N +∞ 0 The right-hand side of Eq. (B7) represents the total magnetic field at the location of the qth "wire", where Hq = (1 + RT M ) exp[iβ0h − iξ0(q − 1)d] is the external magnetic field, Y (m) are the mutual-impedance densities which account for the interaction with the substrate and adjacent "wires". The magnetic field created by qth line current from all periods (except the zeroth one) is given qp Vp Xp=1 Xm=−∞ eiξm(p−q)d RT M m βm . (B10) In contrast to the case of TE polarization discussed in Refs. [16, 20], the series in Eq. (B10) does not converge. Indeed, RT M m has (εs −1)/(εs+1) as limit when m goes to infinity and βm ∼ −im for large m. Meanwhile, it is well known that the harmonic series is divergent. However, the divergence in Eq. (B10) is artificial and was brought when using the Poisson's formula (see Eq. (B3)). Thus, we can avoid divergence by using the Poisson's formula backwards. To that end, we perform the following trans- formation of the series in Eq. (B10) − k 2ηL +∞ Xm=−∞ eiξm(p−q)d RT M m βm = − k 2ηL +∞ Xm=−∞ eiξm(p−q)d 1 βm (cid:18)RT M m − εs − 1 εs + 1(cid:19) − k 2ηL εs − 1 εs + 1 +∞ Xm=−∞ eiξm(p−q)d 1 βm (B11) . 10 The first series on the right hand side of Eq. (B11) is now converging while the second one contains the singularity and should be transformed by means of the Poisson's formula in the following way − k 2ηL εs − 1 εs + 1 − k 2ηL εs − 1 εs + 1 +∞ Xm=−∞ Xm=−∞ +∞ eiξm(p−q)d 1 βm = − k 4η εs − 1 εs + 1 +∞ Xn=−∞ e−ik sin[θ]nLH (2) 0 [k(q − p)d − nL], q 6= p, 1 βm = − k 2η εs − 1 εs + 1 +∞ Xn=1 e−ik sin[θ]nLH (2) 0 [knL] − k 4η εs − 1 εs + 1 , q = p. (B12) Summarizing Eqs. (B8), (B9), (B10) and (B12) one arrives at the explicit expression for the mutual-admittance density Y (m) qp = (cid:18)1 + Y (m) qq = εs − 1 εs + 1 +∞ 4η εs − 1 εs + 1(cid:19) k +(cid:18)1 + k 4η Xn=−∞ εs + 1(cid:19) k εs − 1 2η e−ik sin[θ]nLH (2) 0 [k(q − p)d − nL] + +∞ Xn=1 cos[k sin[θ]nL]H (2) 0 [knL] + k 2ηL k 2ηL +∞ Xm=−∞ Xm=−∞ +∞ eiξm(p−q)d 1 βm (cid:18)RT M m − εs − 1 εs + 1(cid:19) , q 6= p, 1 βm (cid:18)RT M m − εs − 1 εs + 1(cid:19) . (B13) It is worth to note that from the mathematical point of view, it is not strictly correct to represent a diverging se- ries as a sum of two terms as it is done in Eq. (B11). However, when comparing estimations of the mutual- admittance densities given by Eq. (B13) with the ones obtained from 2D full-wave numerical simulations (see Ref. [18]), we observed a good agreement between the two in case of low permittivity substrates (e.g., εs = 2.2). A more rigorous treatment is necessary for dealing ana- lytically with high permittivity substrates. The amplitude of specularly reflected plane-wave (the only propagating diffraction order, d < λ) is then found to be (see also Eq. (B6)) ST M 0 = − k 2dη (1 + RT M 0 β0 )eiβ0h V + RT M 0 e2iβ0h. (B15) From this equation it is straightforward to arrive at Eq. (3). The other equation (Eq. (4)) is readily obtained from Eq. (B7) when N equals 1. The mutual-admittance den- sity in this case is given by the following expression (see Eq. (B13)) 3. Retrieval of magnetic current and admittance density from specular reflection Ym = (cid:18)1 + In this subsection, we derive Eqs. (3) and (4) of the main text following the same order as in Appendix A. The magnetic field of the wave reflected from the uni- form array of magnetic line currents illustrated in Fig. 1 (b) can be represented via the plane-wave expansion as follows 2η εs − 1 εs + 1(cid:19) k βm (cid:18)RT M 1 m − +∞ Xn=1 cos[k sin[θ]nL]H (2) 0 [knd] εs − 1 εs + 1(cid:19) + εs − 1 εs + 1 k 4η . (B16) + k 2dη +∞ Xm=−∞ Appendix C: Parameters of metagratings shown in examples Hx(y, z) = − k 2ηd V +∞ Xm=−∞ (1 + RT M m ) βm +RT E 0 e−iξ0y+iβ0(z+2h). e−iξmy+iβm(z+h) (B14) Table I provides impedance (admittance) densities as well as geometrical parameters of the "wires" composing the metagratings demonstrated as examples in Section III. [1] A. V. Kildishev, A. Boltasseva, and V. M. Shalaev, "Pla- nar photonics with metasurfaces," Science 339, 1232009 (2013). [2] S. B. Glybovski, S. A. Tretyakov, P. A. Belov, Y. S. Kivshar, and C. R. Simovski, "Metasurfaces: From mi- crowaves to visible," Physics Reports 634, 1 -- 72 (2016). [3] V. S. Asadchy, A. D´ıaz-Rubio, and S. A. Tretyakov, "Bianisotropic metasurfaces: physics and applications," Nanophotonics 7, 1069 -- 1094 (2018). [4] N. Yu and F. Capasso, "Flat optics with designer meta- surfaces," Nature materials 13, 139 -- 150 (2014). [5] H.-T. Chen, A. J. Taylor, view of metasurfaces: Reports on Progress in Physics 79, 076401 (2016). and N. Yu, "A re- physics and applications," [6] F. Capasso, "The future and promise of flat optics: a personal perspective," Nanophotonics 7, 953 -- 957 (2018). [7] 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 334, 333 -- 337 (2011). [8] C. Pfeiffer and A. Grbic, "Metamaterial Huygens' sur- faces: Tailoring wave fronts with reflectionless sheets," Physical Review Letters 110, 1 -- 5 (2013). 11 [18] V. Popov, F. Boust, of a reflectionless bianisotropic huygens' metasurface for wide-angle refraction," Phys. Rev. B 97, 125433 (2018). and S. N. Burokur, "Con- structing the near field and far field with reactive metagratings: freedom," Phys. Rev. Applied 11, 024074 (2019). Study on the degrees of [19] H. Chalabi, Y. Ra'di, D. L. Sounas, and A. Al`u, "Effi- cient anomalous reflection through near-field interactions in metasurfaces," Phys. Rev. B 96, 075432 (2017). [20] O. Rabinovich of printed design gratings IEEE Transactions on Antennas and Propagation 66, 4086 -- 4095 (2018). anomalous perfect for and A. circuit Epstein, board "Analytical (pcb) meta- reflection," [9] D. M. Pozar, S. D. Targonski, and H. D. Syrigos, "Design of millimeter wave microstrip reflectarrays," IEEE Transactions on Antennas and Propagation 45, 287 -- 296 (1997). [10] A. Epstein and G. V. metasurfaces "Huygens' lence Journal of the Optical Society of America B 33, A31 -- A50 (2016). principle: design and the via Eleftheriades, equiva- applications," [21] S. Tretyakov, Analytical modeling in applied electromag- netics (Artech House, Norwood, MA, 2003). [22] L. B. Felsen and N. Marcuvitz, Radiation and scattering of waves, Vol. 31 (John Wiley & Sons, New York, 1994). [23] R. Marqu´es, F. Medina, and R. Rafii-El-Idrissi, "Role of bianisotropy in negative permeability and left-handed metamaterials," Phys. Rev. B 65, 144440 (2002). [11] O. Luukkonen, C. Simovski, G. Granet, G. Gous- and analyt- and high-impedance strips patches," setis, D. Lioubtchenko, A. V. Raisanen, Tretyakov, S. A. Tretyakov, and accurate alytical ical model tions," surfaces IEEE Transactions on Antennas and Propagation 56, 1624 -- 1632 (2008). https://doi.org/10.1080/02726340490457890. S. A. An- simula- Electromagnetics 24, 317 -- 338 (2004), and resonators: "Double modeling comprising metal of planar numerical Simovski, split-ring Sauviac, "Simple C. R. [24] B. grids and or [12] X. C. Wang, A. D´ıaz-Rubio, A. Sneck, A. Alastalo, T. Makela, J. Ala-Laurinaho, J. F. Zheng, A. V. Raisanen, and S. A. Tretyakov, "Systematic de- sign of printable metasurfaces: Validation through reverse-offset absorbers," IEEE Transactions on Antennas and Propagation 66, 1340 -- 1351 (2018). printed millimeter-wave [13] David Sell, Jianji Yang, Sage Doshay, Rui Yang, and Jonathan A. Fan, "Large-angle, multifunctional meta- gratings based on freeform multimode geometries," Nano Letters 17, 3752 -- 3757 (2017), pMID: 28459583, https://doi.org/10.1021/acs.nanolett.7b01082. [14] Y. Ra'di, D. L. Sounas, and A. Al`u, "Metagratings: Be- yond the limits of graded metasurfaces for wave front control," Phys. Rev. Lett. 119, 067404 (2017). [15] A. Epstein and O. Rabinovich, of metagratings properties alytical model Phys. Rev. Applied 8, 054037 (2017). via synthesis for "Unveiling detailed the an- analysis," a and [16] V. Popov, F. Boust, trolling Phys. Rev. Applied 10, 011002 (2018). diffraction and S. N. Burokur, "Con- patterns with metagratings," [25] S. Babar and J. H. Weaver, "Optical constants of Cu, Ag, and Au revisited," Appl. Opt. 54, 477 -- 481 (2015). [26] J. Kischkat, S. Peters, B. Gruska, M. Semtsiv, M. Chash- nikova, M. Klinkmuller, O. Fedosenko, S. Machulik, A. Aleksandrova, G. Monastyrskyi, Y. Flores, and W. T. Masselink, "Mid-infrared optical properties of thin films of aluminum oxide, titanium dioxide, sil- icon dioxide, aluminum nitride, and silicon nitride," Appl. Opt. 51, 6789 -- 6798 (2012). [27] A. Epstein and O. anomalous IET Conference Proceedings , 479 (5 pp.) (2018). refraction Rabinovich, with "Perfect metagratings," [28] Z. Fan, M. R. Shcherbakov, M. Allen, J. Allen, B. Wen- ner, and G. Shvets, "Perfect diffraction with multires- onant bianisotropic metagratings," ACS Photonics 5, 4303 -- 4311 (2018). [29] P. Packo, A. N. Norris, and D. Torrent, "In- verse grating problem: Efficient design of anoma- lous refractors," Phys. Rev. Applied 11, 014023 (2019). flexural wave reflectors and [30] D. Torrent, "Acoustic anomalous grating reflectors engineering," [17] M. Chen, E. Abdo-S´anchez, A. Epstein, and G. V. Eleft- heriades, "Theory, design, and experimental verification based Phys. Rev. B 98, 060101 (2018). diffraction on
1711.03236	1	1711	2017-11-09T02:28:33	Band structure and giant Stark effect in two-dimensional transition-metal dichalcogenides	[ "physics.app-ph", "cond-mat.mes-hall" ]	We present a comprehensive study of the electronic structures of 192 configurations of 39 stable, layered, transition-metal dichalcogenides using density-functional theory. We show detailed investigations of their monolayer, bilayer, and trilayer structures' valence-band maxima, conduction-band minima, and band gap responses to transverse electric fields. We also report the critical fields where semiconductor-to-metal phase transitions occur. Our results show that band gap engineering by applying electric fields can be an effective strategy to modulate the electronic properties of transition-metal dichalcogenides for next-generation device applications.	physics.app-ph	physics	Band structure and giant Stark effect in two-dimensional transition-metal dichalcogenides M. Javaid,1, 2, ∗ Salvy P. Russo,1, 3 K. Kalantar-Zadeh,4 Andrew D. Greentree,1, 2 and Daniel W. Drumm2 1Chemical and Quantum Physics, School of Science, RMIT University, Melbourne VIC 3001, Australia 2The Australian Research Council Centre of Excellence for Nanoscale BioPhotonics, School of Science, RMIT University, Melbourne, VIC 3001, Australia 3ARC Centre of Excellence in Exciton Science, School of Science, RMIT University, Melbourne, VIC 3001, Australia 4School of Engineering, RMIT University, Melbourne, VIC 3001, Australia (Dated: November 10, 2017) We present a comprehensive study of the electronic structures of 192 configurations of 39 stable, layered, transition-metal dichalcogenides using density-functional theory. We show detailed investi- gations of their monolayer, bilayer, and trilayer structures' valence-band maxima, conduction-band minima, and band gap responses to transverse electric fields. We also report the critical fields where semiconductor-to-metal phase transitions occur. Our results show that band gap engineer- ing by applying electric fields can be an effective strategy to modulate the electronic properties of transition-metal dichalcogenides for next-generation device applications. I. INTRODUCTION fields [17–19]. Recent advances in the development of atomically- thin materials have opened up new possibilities to ex- plore a two-dimensional (2D) semiconducting era [1–8]. Among 2D materials, transition-metal dichalcogenides (TMDCs) are an interesting family with a diverse range of material properties varying from metals to insula- tors [9]. TMDCs have general formula MX2, where M ∈ {Mo, W, Hf, Zr, Sc, etc.} is a transition metal and X ∈ {O, S, Se, Te} is a chalcogen. (In this article, we in- clude oxygen in the chalcogens for brevity of expression.) In MX2, there is a layer of metal atoms sandwiched be- tween two layers of chalcogen atoms in a X–M–X pattern [10, 11]. Many of the semiconducting, layered TMDCs have similar general features. Their band gaps widen with decreasing number of layers. For example, by reducing the number of layers from bulk down to a single layer, the band gap of MoS2 switches from an indirect band gap of 1.28 eV to a direct band gap of 1.80 eV [2, 12] due to the stronger quantum confinement in the vertical direction. This band gap tunability via the layer thick- ness provides further avenues for novel nanoelectronics and nanophotonics applications [2, 12]. A tunable band gap is highly desirable to allow design flexibility and to control the properties of electronic de- vices, e.g., single-electron transistors [13–15], to achieve tunable fluorescence, display technology, and in changing electrical conductivity applications etc. [16]. However, after fabrication the layer thickness cannot be varied to further tune the band gap. One possible way to control the device properties after the fabrication process is to tune the band structure using external transverse electric ∗ [email protected] Two-dimensional semiconductors are also good candi- dates for photocatalytic water applications due to their large specific surface area, excellent light absorption, and tunable electronic properties [20, 21]. It has been re- ported in [22] that many of the TMDCs qualify for water- splitting applications. In the present work, our goal is to explore the band structure modification via electric field of as many sta- ble, layered, predominately semiconducting TMDCs as possible. Around 40 of the layered TMDCs were re- ported by Wilson et al. in the 1960s [23] and reviewed recently by Kuc et al. [24]. They reported the bulk structures and electrical characteristics of the H and/or T phases of ScS2, ScSe2, ScTe2, TiS2, TiSe2, TiTe2, ZrS2, HfS2, HfSe2, VS2, VSe2, VTe2, NbS2, NbSe2, NbTe2, TaS2, TaSe2, TaTe2, CrS2, CrSe2, CrTe2, MoS2, MoSe2, MoTe2, WS2, WSe2, WTe2, MnS2, MnSe2, MnTe2, ReS2, ReSe2, ReTe2, FeS2, FeSe2, FeTe2, NiS2, NiSe2, NiTe2, and PdS2. There have been significant efforts to synthe- size single and multilayer nanosheets from bulk struc- tures, e.g., TiS2, ZrS2, NbSe2, TaS2, TaSe2, MoS2, MoSe2, MoTe2, WS2, WSe2, WTe2, ReS2, ReSe2, and NiTe2 have been grown by various methods [25–49] and discussed in recent reviews of TMDCs [50–52]. Our cur- rent work also explores those TMDC monolayers and multilayers which have not yet been synthesized but have been predicted to be stable, layered, and semiconducting. Molybdenum- and tungsten-based dichalcogenides have been intensively investigated. The band gap varia- tion of a few layers of these materials under electric field have been widely studied [24, 53–59]. Leb`egue et al. [60] have reported 2D materials by data filtering from their known bulk structures. The band gaps of monolayers of the stable, layered TMDCs have been reported by Ataca et al. in [61] and later by Rasmussen et al. in [22]. How- ever, to date there has been no comprehensive study of the band structure responses of the stable, semiconduct- 2 II. COMPUTATIONAL DETAILS We studied TMDCs constructed by combining tran- sition metals (brown-highlighted) and chalcogens (red- highlighted) in the periodic table as shown in Fig. 1(e). We studied two different structures: honeycomb (H) crys- tal structures with hexagonal space group P63/mmc and centred honeycomb (T) crystal structures with trigonal space group P3m1. Both of these structures and their unit cells are shown in Fig. 1. All the calculations were performed using zero-Kelvin, density-functional theory (DFT) in the crystal09 code [63, 64]. The lattice constants of the bulk structures were determined by structure relaxation using the PBEsol functional [65] for both exchange and correlation terms. The PBEsol functional generally predicts lattice con- stants more accurately than PBE and LSDA (thus better approximating the equilibrium properties of solids), and also handles the electronic response to potentials better than most GGA functionals [65]. We compared our com- puted c lattice parameter with the experimental values where available. We did not find a difference larger than 5% between the experimental and our computed values, indicating that the PBEsol functional is reasonably calcu- lating this parameter. As there are no published van der Waals correction factors for third-row transition metals [66], we ignored these corrections to keep our calculations consistent for all studied materials. less electrons plus triple-zeta sets; polarization We used Gaussian basis a The geometries were optimized to the default crys- tal09 convergence criteria of than 4.5×10−4 Hartree/Bohr maximum force, 3×10−4 Hartree/Bohr RMS force, 1.8×10−3 Hartree maximum displacement, and 1.2×10−3 Hartree RMS displacement. for valence function (TZVP) for the lighter elements (third-period met- als and all chalcogens but Te) and pseudopoten- tial basis follows: Sc pob TZVP 2012 [67] for Sc, Ti pob TZVP 2012 for Ti [67], Zr ECP HAYWSC 311d31G dovesi 1998 [68] for Zr, Hf ECP Stevens 411d31G munoz 2007 [69] for Hf, Cr pob TZVP 2012 [67] for Cr, Mo SC HAYWSC- 311(d31)G cora 1997 [70] for Mo, W cora 1996 [71] for W, Ni pob TZVP 2012 [67] for Ni, Pd HAYWSC- 2111d31 kokalj 1998 unpub [72] for Pd, Pt doll 2004 [73] for Pt, O pob TZVP 2012 [67] for O, S pob TZVP 2012 [67] for S, Se pob TZVP 2012 [67] for Se, and Te m- pVDZ-PP Heyd 2005 [74] for Te. sets for the heavy elements as We created the monolayer, bilayer, and trilayer TMDC unit slabs by defining (001) planes from bulk models, and including vacuum to a total cell height of c = 500 A. We set an 8×16×1 Monkhorst-Pack [75] k-point mesh. We optimized the geometries of monolayer, bilayer, and tri- layer TMDC unit slabs under zero electric field. We used these zero-field-optimized geometries to study the effects of transverse electric fields on their band structures as the influence of field-based geometric disturbances on the band structures is negligible [54]. We also checked the FIG. 1. Bulk TMDC structures; (a) top view of the H phase, (b) top view of the T phase, (c) side view of the H phase, (d) side view of the T phase. Large, brown circles are metal atoms and small, red circles are chalcogens. The unit cells have lengths a = b and c along the directions shown by axes labels between the corresponding sub-figures. The interlayer sepa- ration IS (from centre to centre of the metal atoms) between the two TMDC layers is shown by dashed, arrowed lines. (e) Periodic table (without lanthanides and actinides) showing the brown-highlighted transition metals and red-highlighted chalcogens which we have studied in this article. ing, few-layer TMDCs to electric fields. In this article, we study the role of transverse electric field in engineering band gaps in the known, stable, pre- dominately semiconducting, few-layer TMDCs. We also study responses of their valence-band maxima (VBM) and the conduction-band minima (CBM) with electric field. This is of especial significance for device design and the optimization of atomically thin optoelectronic systems [13, 16, 62]. This paper is organized as follows: In Section II, we describe the computational details. Then we discuss the electronic structures in Sec. III, followed by the lattice pa- rameters and the band gap analysis at zero field in Sec. III A. Then we discuss the variations of the band gaps un- der field for monolayer, bilayer, and trilayer TMDCs in Sec. III B followed by the discussion of the responses of the VBM and CBM to the electric field in Sec. III C. HHeLiBeBCNOFNeNaMgAlSiPSClArKCaScTiVCrMnFeCoNiCuZnGaGeAsSeBrKrRbSrYZrNbMoTcRuRhPdAgCdInSnSbTeIXeCsBaHfTaWReOsIrPtAuHgTlPbBiPoAtRnFrRa*RfDbSgBhHsMtDsRgCnUutFlUupLvUusUuo(a)(b)(c)(d)(e)HphaseTphaseISISacab3 FIG. 2. Bulk structures' in-plane lattice parameters; a for (a) H-phase and (b) T-phase, perpendicular to the plane, c for (c) H-phase and (d) T-phase. Bulk structure band gaps for (e) H-phase and (f) T-phase materials. The colors of the rectangles represent the chalcogens; blue for oxides, red for sulphides, green for selenides, and black for tellurides. Transition metal group numbers are indicated below the braces. effects of electric field on the interlayer separation in H- TiO2 and H-MoS2 bilayer structures and found that the electric fields used do not modify the optimal interlayer separations of these structures. The applied field strength was consistently varied from 0 to ± 0.2 V/A as discussed later in the relevant sections. To compute the critical field (where the semiconductor- to-metal phase transition occurs), we further increased the field strength above 0.2 V/A until we reached the field value where the band gap closed, except for those materials that have critical fields smaller than 0.2 V/A. We calculated the band structures along the high- symmetry path Γ-M-K-Γ. Band structures were cal- culated for uniformly varying electric fields applied per- pendicular to the TMDC slabs. III. RESULTS AND DISCUSSION In this section, we present our electronic structure cal- culations. For all the TMDC materials under investiga- tion, we computed the relaxed bulk (three-dimensional) structure parameters (the in-plane lattice parameters or unit-cell lengths (a) and the unit-cell lengths (c) per- pendicular to the plane), the interlayer separations IS, and band gaps of the bulk structures without electric field. We then computed the band gaps of the relaxed monolayer, bilayer, and trilayer structures with and with- out electric field along the ±c directions. We discuss the band gaps' modulation, the critical fields where the semiconductor-to-metal phase transition occurs and the responses of the VBMs and CBMs under electric field. We report our computed parameters in Table III (which is at the end of the document due to its considerable length). We also report the published values of the mono- layer, bilayer, and trilayer structures' band gaps at zero field, band gap variations under field, and the critical field where available. Throughout the discussion, we fo- cus more on the trends observed across the TMDC fam- ily, in various ways, instead of the in-depth analysis of individual materials. A. Properties under zero field For several of the materials (H-ScX2, T-ZrX2, T-NiX2, T-PdX2, and T-PtX2), we observe increasing bulk struc- ture lattice parameters a and c as we move down the chalcogen group from oxides to tellurides, accompanied by decreasing band gaps (Fig. 2), as reported in Table III. The other materials depart from this trend in vari- ous ways, to various extents. For example, H-MoX2 and H-WX2 follow this trend for both their lattice parameters but their bulk band gaps from oxides to tellurides do not. H-CrX2 follow an increasing trend in the lattice param- eter a but not in c. Also their bulk band gaps appear non-linear with respect to chalcogen element. T-HfX2 follow the trend in a and the bulk band gaps but T-HfS2 deviates for c. H-ZrX2, H-HfX2 and H-TiX2 show devia- tion from this general trend across all of the parameters, i.e., a, c, and the bulk band gaps. A similar oxide to telluride trend in TMDC monolayer band gaps has been reported by Rasmussen et al. [22]. No major trends in the lattice parameters or bulk band gaps are observed either across the periods or down the groups in the transition metals (Fig. 2). Companion analyses have been carried out by instead changing the transition metal period (3, 4, and 5), transi- tion metal group (III, IV, VI, and X), the material phase (H and T), or the number of layers in the model (1, 2, or 3), while holding all other dimensions constant. The number of non-singleton, non-zero-valued, subset classes for each dimension are: 12(×4) varying chalco- gen, 12(×3) + 4(×2) varying transition-metal period, 24a(A)241015c(A)46ScTiZrHfCrMoW02Eg(eV)TiZrHfNiPdPt05H-phasematerialsT-phasematerials(a)(b)(c)(d)(e)(f)IIIIVVIIVX4 FIG. 3. H-TiO2 bilayer sample band structures with zero and finite field. Band structures of: (a) H-TiO2 bilayer at zero field with an indirect band gap of 1.09 eV (red-arrowed lines) from the CBM to the VBM, (b) H-TiO2 bilayer at an electric field of 0.1 V/A with an indirect band gap of 0.76 eV, (c) H-TiO2 bilayer at an electric field of 0.2 V/A with an indirect band gap of 0.41 eV. (d) Modulation of the VBM and CBM of the H-TiO2 bilayer for various values of absolute electric field (Note, the energy zero here is the energy of the vacuum.). (e) Band gap of the H-TiO2 bilayer for various values of the absolute electric field showing a decrease in the band gap with the field. 4(×3) + 17(×2) varying transition-metal group, 9(×2) varying material phase, and 46(×3) varying the number of layers. Note: the numbers in parentheses indicate the number of members of each class. Examining combinations The results are summarised in Table I, which shows the dominant behavior (or the two, most-prevalent, non- dominant behaviors) across all subset classes with more than one element. The behaviors considered are: flat (F), monotonically increasing (↑), monotonically decreasing (↓), or other non-monotonic behavior (X) for classes with more than two members. (multi- parameter patterns), we found that with a change of phase (from H to T) 5 of 9 subsets showed a combined increase to a, decrease to c, and increasing band gap. With increasing transition-metal group, 10 of 21 subsets showed decreases to both a and c (but a 4/6 split be- tween increasing and decreasing band gap). All other combinations accounted for smaller proportions of the subset classes and are therefore regarded as insignificant. For most of the materials we find an increase in the band gap with the decreasing number of layers from bulk of the trends to monolayer which is consistent with the properties of the Mo- and W-based dichalcogenides [76]. (Note there are too many values to show clearly in a figure, but their values are available in Table III.) We observe minor devi- ations from this trend by H-ZrO2, H-ZrSe2, H-HfO2, T- HfO2, H-HfSe2, H-CrS2, H-CrSe2, H-CrTe2, H-MoO2, H- MoSe2, H-MoTe2, H-WO2, H-WTe2, T-NiO2, T-PdO2, and T-PtO2. For example, the band gap of H-ZrO2 in- creases from 1.54 eV to 2.08 eV from bulk to its bilayer structure but the monolayer has a smaller gap of 1.62 eV. B. Band gap variations under electric field An electric field can potentially be used to tailor the band gaps of layered materials. The band structure vari- ations with electric field arise due to the well-known Stark effect. The Stark effect induces a potential difference be- tween the layers which causes splitting of energy bands belonging to different layers as well as shifting of the VBM and CBM [54]. This splitting and shifting of the MK5.442.7202.725.44E-EF(eV)Electricfield=0V/AE=1.09eVMKE=0.76eVElectricfield=0.1V/AMKElectricfield=0.2V/AE=0.41eV00.10.28.487.67.2Abs.electricfield(V/A)Energy(eV)00.10.20.40.60.811.2Abs.electricfield(V/A)Bandgap(eV)(a)(b)(c)(d)(e)CBMVBM5 FIG. 4. Band gaps as functions of absolute electric field strength for all studied H- and T-phase, bilayer and trilayer TMDCs (except the Sc family, which are all metallic - see Table 1). The subfigures separate the gamut into families of common phase and metal element, and are arranged preserving period and group order. In each subfigure, the colors denote the chalcogens; blue for oxides, red for sulphides, green for selenides, and black for tellurides. Bilayers are represented by circular markers (and = symbols in the legend for brevity), and trilayers by square markers (and ≡ symbols in the legend). The lines (solid for H phase and dashed for T phase) are linear fits to each data set, excluding any inconsistent points (as discussed in the text). Most materials exhibit monotonic band gap decrease with increasing electric field, at least to their critical fields. Note that zero-band-gap and zero-band-gap-modulation values are not shown for clarity, and that the vertical scales differ between subfigures. 00.20.000.400.801.20TiO2=TiO2⌘TiS2=TiS2⌘TiSe2=TiTe2=TiTe2⌘00.21.902.202.50TiO2=TiO2⌘00.20.000.501.001.502.00ZrO2=ZrO2⌘ZrS2=ZrS2⌘ZrSe2=ZrTe2⌘00.20.001.002.003.004.00ZrO2=ZrO2⌘ZrS2=ZrS2⌘ZrSe2=ZrSe2⌘00.20.000.501.001.502.002.50HfO2=HfO2⌘HfS2=HfS2⌘HfSe2=HfTe2=HfTe2⌘00.20.001.002.003.004.005.00HfO2=HfO2⌘HfS2=HfS2⌘HfSe2=HfSe2⌘00.20.000.300.600.90CrO2=CrO2⌘CrS2=CrS2⌘CrSe2=CrSe2⌘CrTe2=00.20.000.300.600.901.20MoO2=MoO2⌘MoS2=MoS2⌘MoSe2=MoSe2⌘MoTe2=MoTe2⌘00.20.000.300.600.901.201.50WO2=WO2⌘WS2=WS2⌘WSe2=WSe2⌘WTe2=WTe2⌘00.20.720.760.800.840.88NiO2=NiO2⌘00.20.850.951.051.15PdO2=PdO2⌘00.20.800.901.001.10PtO2⌘H-TiX2T-TiX2H-ZrX2T-ZrX2H-HfX2T-HfX2H-CrX2H-MoX2H-WX2T-NiX2T-PdX2T-PtX2Abs.electricfield(V/A)Bandgap(eV)GroupIVGroupIVGroupVIGroupXVariable Chalcogen TM period TM group Phase Layers a ↑ ↑/Xc ↓ ↑ F Parameter Bulk Eg X/↓b ↑ ↑/↓e ↑ F c ↑/Xa X/↑d ↓ ↓ F Eg(E = 0) ↑ / ↓f ↓ ↑ ↑ ↓ a 50/50%; b 50/42%; c 50/38%; d 44/38%; e 43/38%; f 47/42% TABLE I. Multidimensional analysis of dominant be- haviour(s) within the TMDC family for the lattice param- eters a and c, the bulk band gap Eg, and the layer-dependent band gap Eg(E = 0). F, ↑, ↓, and X respectively indicate flat, monotonically increasing, monotonically decreasing, and other non-monotonic behaviour with an increase to the rele- vant dimension. The dimensions (with the ordering used in the analysis) are: chalcogen (period 2, 3, 4, or 5), transition- metal period (3, 4, or 5), transition-metal group (III, IV, VI, or X), material phase (from H to T), and number of layers (1, 2, or 3). Single indicators are shown when >50% of the sub- set classes exhibit the behavior; pairs of indicators are shown when there are behaviors that capture most subsets (their relative percentages are shown as footnotes). FIG. 5. Histogram of the trilayer to bilayer GSE coefficient ratios showing most of the materials clustered around a value of 2 which might be due to the two interlayer spacings of the trilayer as compared to the bilayer structures. energy bands can increase [77–79] or reduce [53, 80] the band gaps. In the case of multilayered TMDCs, this splitting and shifting pushes both the CBM and VBM towards the Fermi level. This results in the reduction of the band gap with applied electric field in multilayered TMDCs. For each material under investigation, we computed band gaps of its monolayer, bilayer, and trilayer struc- tures with an external electric field applied perpendicu- lar to the layers along the ±c directions. We consistently varied the field strength from 0 to ± 0.2 V/A, in steps 6 no larger than 0.02 V/A, except for materials (H-TiO2 trilayer, H-ZrSe2 bilayer, T-ZrSe2 trilayer, H-CrO2 bi- layer and trilayer, H-CrTe2 bilayer, and H-WO2 trilayer) whose band gaps close before 0.2 V/A. This maximum field strength is stronger than usual device fields and is shown to more clearly illustrate the band structure vari- ations under electric field. For all structures, we find that monolayer TMDCs do not respond to electric field up to a field strength of 0.2 V/A. However the band structures of most of the bilayer and trilayer TMDCs do vary with electric field offering applications in band engineering. Such results are con- sistent with previously reported studies of Mo- and W- based dichalcongenides. Figure 3 shows sample band structures of H-phase TiO2 bilayers at zero field and for two different finite values of the electric field. Increasing the electric field causes splittings of the energy bands resulting in a reduc- tion of the band gap from 1.09 eV (at zero field) to 0.41 eV (at 0.2 V/A). The responses of the VBM and CBM are shown in Fig. 3(d) for electric fields varying from 0 V/A to ±0.2 V/A. There is significant variation of both the VBM and the CBM with the electric field pushing them towards the Fermi level, EF for the H-TiO2 bilayer resulting in the reduction of the band gap under field. The band gap as a function of the absolute electric field is shown in Fig. 3(e) where the solid line is a linear fit to the data. For an electric field of 0.2 V/A, the band gap is reduced by 686 meV. Thus band gap variation in H-TiO2 bilayers is achievable via electric field at a rate of -3.43 eV/(V/A) while the corresponding VBM and CBM bendings are 1.69 and -1.73 eV/(V/A) respectively. We obtain qualitatively similar band structure re- sponses to electric field for most of the studied materials. The band gaps as functions of electric field are shown in Fig. 4 for all materials. Most band gaps decrease mono- tonically with electric field strength. The Ni, Pd, and Pt oxides seem to exhibit slightly quadratic behavior at low field which is similar to the band structure modulation via electric field reported for few layer black phosphorus [81]. The linear fits applied across all band gap data as functions of electric field (Fig. 4) are shown by solid lines for all H-phase materials and by dashed lines for all T- phase materials. Circular and square markers represent the bilayer and trilayer data in all the sub-figures while the blue, red, green, and black colors represent the ox- ides, sulphides, selenides, and tellurides respectively. All the materials show symmetric response to fields oriented in the ±c directions which is due to the symmetry of the chalcogen layers around the metal layer, except a few odd data points (Fig. 4). We have excluded these odd data points from the linear fits as we believe that they do not represent the underlying physics. The change in the band gap with the electric field, dEg dEF is given by [53] , dEg dEF = −e S. (1) 01234502468GSEcoecientratioCountsCropCrop7 FIG. 6. Band gap variations with electric field strength in H- and T-phase bilayer and trilayer TMDCs. The top of each rectangular bar shows the band gap at zero field while its bottom shows the band gap at a maximum field of 0.2 V/A or less for materials whose band gaps close before 0.2 V/A as discussed in the main body and in Table III. These are shown as unfilled rectangles crossing the zero-band-gap line as far as has been calculated. To allow comparisons with non-metallic TMDCs, dotted rectangles extrapolate the pre-critical-field data to 0.2 V/A field to allow rapid, direct comparison with the other (solid) rectangles as regards the pre-critical-field behavior. This is directly related to the giant Stark effect coefficient, S, as explained in the main body. The color of the bars represents the associated chalcogen; blue for oxides, red for sulphides, green for selenides, and black for tellurides. Metallic or semimetallic TMDCs, which have zero band gap, are denoted by asterisks. Materials whose non-zero band gaps do not change with electric field are denoted by triangles. Here Eg is the band gap in eV, EF is the electric field strength in V/A, e is the fundamental charge, and S is the GSE coefficient, calculated from the slope of the lin- ear fits to our data shown in Fig. 4. We report this GSE coefficient ( S) in Table III along with 95% confidence in- tervals. We compare our computed GSE coefficient with the literature where available in Table III. We observe an interesting trend in S for the bilayer and trilayer structures of several materials. We find that the S values for trilayer structures are approximately twice those for bilayer structures, with a few anomalies. For example, H-TiO2 trilayers has S = 6.87 A which is twice that of 3.43 A for H-TiO2 bilayers. Similarly, H-CrO2 and H-CrS2 show a factor of two in S between their trilayer and bilayer structures. We report this trilayer- bilayer S ratio ( S≡/ S=) in Table III and display it in histogram form in Fig. 5. The symbols ≡ and = repre- sent trilayer and bilayer structures respectively. A group TiZrHfCrMoW1012H-phasebilayerTiZrHfCrMoW1012H-phasetrilayerIVVIIVVITiZrHfNiPdPt0246T-phasebilayerTiZrHfNiPdPt0246T-phasetrilayerBandgap(eV)CropIVXIVXof the materials cluster around a S≡/ S= ratio of two with a few outliers. T-ZrSe2, H-HfO2, H-CrSe2, H-MoO2, H- MoSe2, H-WSe2, and T-NiO2 are outliers. The tellurides show anomalous behavior where they have non-zero band gaps in both bilayer and trilayer structures. Their S≡/ S= ratios are smaller than the other dichalcogenides suggest- ing that telluride trilayers are less sensitive to field. The near-double modulation of trilayer TMDCs com- pared to their associated bilayers might be explained by the number of interlayer separations (IS) they have. Bi- layers have one IS, while trilayers have two. Monolayers have none, and it is well established that they do not exhibit band gap variation (and therefore their S = 0). All studied materials' band gap variations with field are shown in Fig. 6. The length of each bar shows the change in the band gap under field. The top of each rectangular bar shows the material's band gap at zero field and the bottom of each rectangular bar shows the band gap at the maximum field of 0.2 V/A or less for materials whose band gaps close before 0.2 V/A. Blue, red, green, and black colors represent the oxides, sul- phides, selenides, and tellurides. Rectangular bars cross- ing the zero-band-gap line indicate materials where the CBM switches below VBM at maximum field. Table II details a multidimensional trend analysis for S (and several other properties which are discussed be- low), similar to those for bulk properties presented in Table I. The numbers of subset classes change due to the separation of materials by number of layers and sub- sequent exclusion of monolayer data or others whose S are uniformly zero, along with any completely metallic or semi-metallic classes. For S (and EFc, ∂EVBM/∂E, and ∂ECBM/∂E, discussed below), there are 19(×4) + 2(×3) subset classes varying chalcogen, 19(×3) + 2(×2) classes varying transition-metal period, 8(×3) + 30(×2) vary- ing transition-metal group, 17(×2) varying phase, and 34(×2) varying layer number. We notice an overall decrease in band gap variation under field, S, down the chalcogen group from the oxides to tellurides with a few anomalies. This trend is also illustrated in Fig. 6 where we can see a decrease in the bar lengths from oxides (blue) to tellurides (black) for most of the transition metals, which again suggest that tellurides are less sensitive to electric fields than other dichalcogenides. S behaves non-monotonically with transition-metal pe- riod, but decreases with transition-metal group. The group X TMDCs exhibit the least band gap variation under field when compared as a whole to other groups. For example, the PtO2 bilayer, the PtS2 and the PtSe2 bilayer and trilayer do not respond to the field within the range of 0.2 V/A as compared to the Hf- and W-based dichalcogenides (Table III). The GSE coefficients increase from the H- to T-phase materials as shown in Fig. 6 for those materials which are stable in both H and T phases. This is mainly due to the distinct intralayer stacking of the two phases leading to a difference in the interplanar X–X dispersion interac- 8 tions [82]. Thus selection of the material phase is another potential pre-fabrication lever for controlling band gaps and their variations under field. We further reveal the critical fields (EFc), where clo- sures of band gaps occur and the bilayer or trilayer TMDCs undergo semiconductor-to-metal phase transi- tions. To achieve this, we increased the field strengths beyond 0.2 V/A for those materials whose band gaps had not yet closed. We report these computed critical fields in Table III, and summarise their trends in Table II. The critical field strengths vary non-monotonically down the chalcogen group from oxides to tellurides. Sim- ilarly, no dominant trends in the critical field strengths are predicted across the transition metal periods or down the transition metal groups for the same phase, chalco- gen, and number of layers. However, EFc does increase from H- to T-phase materials, and and is generally lower for trilayer materials than the related bilayer ones. The converse dependencies of S and EFc on transition- metal group or layer number are understandable; since S corresponds to the slope of the band gap with field, it stands to reason that more responsive materials (tri- layers) should become metallic or semi-metallic at lower fields (EFc ), particularly once account is taken of any discrepancy in their zero-field band gaps. However, the behavior with phase is aligned, with both parameters in- creasing. Comparison with Table I shows that this is explained by an evident accompanying increase in the zero-field band gaps with phase. We defer discussing compound trends until the rest of Table II has been explored in Sec. III C. Our general findings are that: trilayer band gaps re- spond more to field than bilayer gaps; T-phase materials respond more than H-phase ones; band gap modulations decrease from oxides to tellurides; and the response to the electric field decreases from left to right across the tran- sition metals when compared within the same period and for the same chalcogens. These findings reveal the whole range of TMDC band gap responses under field. They en- able one to select the appropriate material(s) to engineer devices in according to one's application's requirements. C. Valence and conduction band absolute positions and variations under field For many applications, the energies of the valence and conduction bands with respect to the vacuum level are important. For example, in the construction of 2D het- erostructure materials, knowledge of absolute VBM and CBM energies is required to predict behavior. For device design, such as a single-electron transistor [13, 14], where we require control over its transport properties, study of the VBMs' and CBMs' responses to electric fields is es- sential. In Fig. 7, we show the computed energies of all studied TMDCs VBMs, and CBMs with respect to the vacuum – Variable Chalcogen TM period TM group Phase Layers S ↓ X ↓ ↑ ↑ EFc X ↑/Xb ↑/↓e ↑ ↓ EVBM(E = 0) EVBM(Emax) ECBM(E = 0) ECBM(Emax) Parameter X ↑/Xc ↑ ↑ ↑ X X/↑d ↑ ↑ ↑ X ↑ ↑ ↑ ↓ X ↑ ↑ ↑ ↓ a 48/48%; b 48/39%; c 44/34%; d 41/38%; e 47/37%. ∂EVBM ∂E ↓ ↑ ↓ ↑ ↑ 9 ∂ECBM ∂E ↑/Xa X ↑ ↑ ↓ TABLE II. Multidimensional analysis of dominant behaviour(s) within the TMDC family for the calculated parameters: giant Stark coefficient S, critical field for insulator-to-metal/semi-metal transition EFc , VBM and CBM energies at zero and maximum (as defined in the main text) fields, and the VBM and CBM bending rates with field. F (not exhibited), ↑, ↓, and X respectively indicate flat, monotonically increasing, monotonically decreasing, and other non-monotonic behaviour with an increase to the relevant dimension. The dimensions (with the ordering used in the analysis) are: chalcogen (period 2, 3, 4, or 5), transition- metal period (3, 4, or 5), transition-metal group (III, IV, VI, or X), material phase (H or T), and the number of atomic layers (2 or 3). Single indicators are shown when >50% of the subset classes exhibit the behavior; pairs of indicators are shown when there are two behaviors that capture most subsets (their relative percentages are shown as footnotes). at zero field (left rectangular bar) and for the maximum field of 0.2 V/A or less for those materials whose band gaps close before 0.2 V/A (right rectangular bar). The lengths of the rectangular bars show the band gaps. The blue, red, green, and black colors show oxides, sulphides, selenides, and tellurides respectively. We also summarise any trends in Table II. Here, for the absolute energies, we have 20(×4) + 2(×3) subset classes when varying chalcogen, 24(×3) + 8(×2) varying transition-metal period, 42(×2) varying transition-metal group, 18(×2) varying phase, and 44(×2) varying layer number. The rates of change of the VBM and CBM have the same numbers and types of subset classes as S and EFc, described in Sec. III B. The absolute VBM and CBM energies behave non- monotonically as the chalcogen period increases. How- ever, they largely increase with transition-metal period and group, and with changing from H to T phase. Whilst the zero-field extrema energies increase with the number of layers, the maximum-field extrema energies both de- crease with layer number, consistent with the increased S also exhibited by trilayer materials. When changing transition metal period, 11 of 32 sub- set classes show an overall increases in all four reported VBM and CBM energies (Fig. 7 top panels). This behav- ior is particularly prominent in the oxides. 28 of 42 subset classes show the same behaviour when transition-metal group is changed (i.e., comparing same-colored bars be- tween left and right groups in subfigures of Fig. 7). How- ever, while increasing the number of layers 16 of 44 sub- set classes show increases to the zero-field extrema with decreases to the high-field extrema. For almost all of these materials, we find significant bending in the CBM under field (as reported in Table III and shown in Fig. 7) except for a few materials, e.g., WSe2 bilayers and PdO2 trilayers show anomalous be- havior; their CBMs increase with the field. The VBMs for all the materials show increasing energy shifts un- der electric field except TiTe2 bilayers, H-HfTe2 bilayers, CrTe2 bilayers, and MoS2 bilayers, which show a negative bending in their VBMs (Table III). Looking broadly across all parameters reported in Ta- ble II, we identify the following large-scale compound behaviours: when varying transition-metal group, 10 of 42 subset classes show consistent behavior across all eight parameters – increases to all absolute energies plus ∂ECBM/∂E combined with decreases to the other three parameters; when increasing the number of layers, 19 of 44 subsets show increasing S and ∂EVBM/∂E with de- creasing EFc and ∂ECBM/∂E , 15 of 44 show increased S and VBM energies (both fields) with decreased EFc, and 23 of 44 show decreased EFc with increased VBM ener- gies (both fields). Other compound behaviours are only exhibited by small fractions of the subset classes. It has been reported in [22] that materials with CBMs above the standard hydrogen electrode (SHE), i.e. -4.03 eV relative to vacuum at pH 7, can be used at the cathode of photocatalytic water splitting devices to evolve hydro- gen. Similarly materials with VBMs below the oxygen evolution potential (1.23 eV below the SHE) can be used as photoanodes in water splitting devices. It has been suggested that the CBM/VBM should lie a few tenths of an electron volt above/below the redox potentials [83] to account for the intrinsic energy barriers of the water splitting reactions [84, 85]. Any material with a CBM a few tenths of an electron volt above -4.03 eV or VBM a few tenths of an electron volt below -5.26 eV is desirable for water splitting applications. In Fig. 7, the CBMs of Mo- and W-based, bilayer and trilayer sulphides lie ≥ 0.1 eV above -4.03 eV at zero and low fields. For example, MoS2 bilayers have a use- ful CBM for water splitting applications in electric fields ranging from 0 to 0.14 V/A. Similarly, MoS2 trilayers, WS2 bilayers, and WS2 trilayers are useful for this ap- plication in fields ranging from 0 to 0.04, 0.1, and 0.04 V/A, respectively. The T-phase, Zr and Hf, bilayer and trilayer oxides have CBMs well above the SHE potential for both zero and maximum field and could be useful for water splitting applications under any field strength from 0 to 0.2 V/A. 10 FIG. 7. Positions of the valence band maxima (VBMs) and conduction band minima (CBMs) with respect to the vacuum level (0 eV) in H- and T-phase, bilayer and trilayer TMDCs as labeled. The bottom of each rectangular bar shows the VBM and the top the CBM; its length shows the band gap. The color of the bars represent the chalcogens; blue for oxides, red for sulphides, green for selenides, and black for tellurides. The double bars show the band positions at zero field (left) and at the maximum field of 0.2 V/A or less (as discussed in the main text and in Table III) for the materials where the band gap closes before 0.2 V/A (right). Asterisks denote the Fermi levels of zero-band gap materials at their critical fields (or at zero field for materials which are naturally metallic or semimetallic). Note: the critical fields can be found in Table III. Dashed lines show the energies 0.1 eV above the hydrogen evolution potential, i.e., -4.03 eV and 0.1 eV below the oxygen evolution potential, i.e., -5.23 eV. Most of the materials have their VBM below the oxy- gen redox potential except for the bilayers and trilayers of CrS2, CrTe2, MoS2, MoTe2, WS2, WTe2, PtS2, and PtSe2, and the zero band gap materials (Fig. 7). The TMDCs therefore appear to be a highly useful class of materials for water splitting. IV. CONCLUSIONS We have presented a comprehensive density-functional theory study of the electronic structures of 192 config- urations of 39 stable, two-dimensional, transition-metal dichalcogenides. Our calculations show the band gaps of few-layer TMDCs along with variation of the band struc- tures by electric fields. The data has been analysed across five dimensions: chalcogen period, transition-metal pe- riod, transition-metal group, phase, and number of lay- ers. TiZrHfCrMoW8.575.54H-phasebilayerTiZrHfCrMoW8.575.54H-phasetrilayerIVVIIVVITiZrHfNiPdPt9753T-phasebilayerTiZrHfNiPdPt9753T-phasetrilayerEnergyrelativetovacuum(eV)CropIVXIVXBand gaps generally decrease down the chalcogen group from oxides to tellurides, increase across the tran- sition metals from left to right, are larger for T-phase materials than corresponding H-phase ones, and decrease with more layers. The responses to the electric field de- crease down the chalcogens and across transition met- als in the same period, are larger for T-phase materials than H-phase ones, and increase with increasing number of layers. We generally found that the CBMs decrease with higher fields while the VBMs increase, narrowing the band gap from both sides. We also have suggested materials which could be useful for water splitting appli- cations under zero and low fields. By presenting the field-modulated behavior of mono- layer, bilayer, and trilayer structures of 39 different ma- terials, this work supports the advance of 2D materials from fundamental research to real applications. In partic- ular, it will aid band-gap and opto-electronic engineers to select the optimal TMDC for their device requirements. 11 ACKNOWLEDGEMENTS The authors acknowledge financial support from the Australian Research Council (Project Nos. DP130104381, CE140100003, FT160100357, LE160100051, and CE170100026). This work was supported by computational resources provided by the Australian Government through the National Com- putational Infrastructure (NCI) under the National Computational Merit Allocation Scheme. [1] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, Nature 438, 197 (2005). [21] A. K. Singh, K. Mathew, H. L. Zhuang, and R. G. Hen- nig, J. Phys. Chem. Lett. 6, 1087 (2015). [22] F. A. Rasmussen and K. S. Thygesen, J. Phys. Chem. C [2] A. Splendiani, L. Sun, Y. Zhang, T. Li, J. Kim, C.-Y. and F. Wang, Nano Lett. 10, 1271 Chim, G. Galli, (2010). 119, 13169 (2015). [23] J. Wilson and A. Yoffe, Adv. Phys. 18, 193 (1969). [24] A. Kuc, T. Heine, and A. Kis, MRS Bulletin 40, 577 [3] A. K. Geim and K. S. Novoselov, Nat. Mater. 6, 183 (2015). (2007). [4] X. Qian, Y. Wang, W. Li, J. Lu, and J. Li, 2D Mater. 2, 032003 (2015). [5] D. Golberg, Y. Bando, Y. Huang, T. Terao, M. Mitome, C. Tang, and C. Zhi, ACS Nano 4, 2979 (2010). [6] M. Osada and T. Sasaki, J. Mater. Chem. 19, 2503 (2009). [7] H. Wang, L. Yu, Y.-H. Lee, Y. Shi, A. Hsu, M. L. Chin, L.-J. Li, M. Dubey, J. Kong, and T. Palacios, Nano Lett. 12, 4674 (2012). [8] D. Akinwande, N. Petrone, and J. Hone, Nat. Commun. 5, 5678 (2014). [9] K. F. Mak and J. Shan, Nat. Photon 10, 216 (2016). [10] Q. H. Wang, K. Kalantar-Zadeh, A. Kis, J. N. Coleman, and M. S. Strano, Nat. Nanotechnol. 7, 699 (2012). [11] D. Xiao, G.-B. Liu, W. Feng, X. Xu, and W. Yao, Phys. Rev. Lett. 108, 196802 (2012). [12] K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, Phys. Rev. Lett. 105, 136805 (2010). [13] S. J. Angus, A. J. Ferguson, A. S. Dzurak, and R. G. Clark, Nano Lett. 7, 2051 (2007). [14] M. Javaid, D. W. Drumm, S. P. Russo, and A. D. Green- tree, Nanotechnology 28, 125203 (2017). [15] J. Sun, M. Larsson, I. Maximov, H. Hardtdegen, and H. Q. Xu, Appl. Phys. Lett. 94, 042114 (2009). [16] K. K. N. Simon M. Sze, Physics of Semiconductor De- vices, Edition 3 (John Wiley & Sons, 2006). [17] N. Lu, H. Guo, L. Li, J. Dai, L. Wang, W.-N. Mei, X. Wu, and X. C. Zeng, Nanoscale 6, 2879 (2014). [18] D. Xu, H. He, R. Pandey, and S. P. Karna, J. Phys.: Condens. Matter 25, 345302 (2013). [19] R. Balu, X. Zhong, R. Pandey, and S. P. Karna, Appl. Phys. Lett. 100, 052104 (2012). [25] M. M. Ugeda, A. J. Bradley, Y. Zhang, S. Onishi, Y. Chen, W. Ruan, C. Ojeda-Aristizabal, H. Ryu, M. T. Edmonds, H.-Z. Tsai, A. Riss, S.-K. Mo, D. Lee, A. Zettl, Z. Hussain, Z.-X. Shen, and M. F. Crommie, Nat. Phys. 12, 92 (2016). [26] M. M. Ugeda, A. J. Bradley, S.-F. Shi, F. H. da Jornada, Y. Zhang, D. Y. Qiu, W. Ruan, S.-K. Mo, Z. Hussain, Z.-X. Shen, F. Wang, S. G. Louie, and M. F. Crommie, Nat. Mater. 13, 1091 (2014). [27] S. Barja, S. Wickenburg, Z.-F. Liu, Y. Zhang, H. Ryu, M. M. Ugeda, Z. Hussain, Z.-X. Shen, S.-K. Mo, E. Wong, M. B. Salmeron, F. Wang, M. F. Crommie, D. F. Ogletree, J. B. Neaton, and A. Weber-Bargioni, Nat. Phys. 12, 751 (2016). [28] D. Dumcenco, D. Ovchinnikov, K. Marinov, P. Lazi, M. Gibertini, N. Marzari, O. L. Sanchez, Y.-C. Kung, D. Krasnozhon, M.-W. Chen, S. Bertolazzi, P. Gillet, A. Fontcuberta i Morral, A. Radenovic, and A. Kis, ACS Nano 9, 4611 (2015). [29] X. Wang, Y. Gong, G. Shi, W. L. Chow, K. Keyshar, G. Ye, R. Vajtai, J. Lou, Z. Liu, E. Ringe, B. K. Tay, and P. M. Ajayan, ACS Nano 8, 5125 (2014). [30] Y.-H. Chang, W. Zhang, Y. Zhu, Y. Han, J. Pu, J.-K. Chang, W.-T. Hsu, J.-K. Huang, C.-L. Hsu, M.-H. Chiu, T. Takenobu, H. Li, C.-I. Wu, W.-H. Chang, A. T. S. Wee, and L.-J. Li, ACS Nano 8, 8582 (2014). [31] J. Chen, B. Liu, Y. Liu, W. Tang, C. T. Nai, L. Li, J. Zheng, L. Gao, Y. Zheng, H. S. Shin, H. Y. Jeong, and K. P. Loh, Adv. Mater. 27, 6722 (2015). [32] M. Yoshida, T. Iizuka, Y. Saito, M. Onga, R. Suzuki, Y. Zhang, Y. Iwasa, and S. Shimizu, Nano Lett. 16, 2061 (2016). [33] J.-W. Chung, Z. Dai, and F. Ohuchi, J. Cryst. Growth [20] T.-F. Yeh, C.-Y. Teng, S.-J. Chen, and H. Teng, Adv. 186, 137 (1998). Mater. 26, 3297 (2014). 12 [34] K. Keyshar, Y. Gong, G. Ye, G. Brunetto, W. Zhou, D. P. Cole, K. Hackenberg, Y. He, L. Machado, M. Kabbani, A. H. C. Hart, B. Li, D. S. Galvao, A. George, R. Vajtai, C. S. Tiwary, and P. M. Ajayan, Adv. Mater. 27, 4640 (2015). [53] A. Ramasubramaniam, D. Naveh, and E. Towe, Phys. Rev. B 84, 205325 (2011). [54] Q. Liu, L. Li, Y. Li, Z. Gao, Z. Chen, and J. Lu, J. Phys. Chem. C 116, 21556 (2012). [55] C. V. Nguyen, N. N. Hieu, and V. V. Ilyasov, J. Electron. [35] M. Hafeez, L. Gan, H. Li, Y. Ma, and T. Zhai, Adv. Mater. 45, 4038 (2016). Funct. Mater. 26, 4551 (2016). [56] N. A. Lanzillo, T. P. ORegan, and S. K. Nayak, Comput. [36] C. H. Naylor, W. M. Parkin, J. Ping, Z. Gao, Y. R. Zhou, Y. Kim, F. Streller, R. W. Carpick, A. M. Rappe, M. Drndi´c, J. M. Kikkawa, and A. T. C. Johnson, Nano Lett. 16, 4297 (2016). [37] J. Zhou, F. Liu, J. Lin, X. Huang, J. Xia, B. Zhang, Q. Zeng, H. Wang, C. Zhu, L. Niu, X. Wang, W. Fu, P. Yu, T.-R. Chang, C.-H. Hsu, D. Wu, H.-T. Jeng, Y. Huang, H. Lin, Z. Shen, C. Yang, L. Lu, K. Sue- naga, W. Zhou, S. T. Pantelides, G. Liu, and Z. Liu, Adv. Mater. 29, 1603471 (1 (2017). [38] D. Dumcenco, D. Ovchinnikov, O. L. Sanchez, P. Gillet, D. T. L. Alexander, S. Lazar, A. Radenovic, and A. Kis, 2D Mater. 2, 044005 (1 (2015). [39] S. M. Eichfeld, L. Hossain, Y.-C. Lin, A. F. Piasecki, B. Kupp, A. G. Birdwell, R. A. Burke, N. Lu, X. Peng, J. Li, A. Azcatl, S. McDonnell, R. M. Wallace, M. J. Kim, T. S. Mayer, J. M. Redwing, and J. A. Robinson, ACS Nano 9, 2080 (2015). [40] Y. Gao, Z. Liu, D.-M. Sun, L. Huang, L.-P. Ma, L.-C. Yin, T. Ma, Z. Zhang, X.-L. Ma, L.-M. Peng, H.-M. Cheng, and W. Ren, Nat. Commun. 6, 8569 (1 (2015). [41] J. N. Coleman, M. Lotya, A. O'Neill, S. D. Bergin, P. J. King, U. Khan, K. Young, A. Gaucher, S. De, R. J. Smith, I. V. Shvets, S. K. Arora, G. Stanton, H.-Y. Kim, K. Lee, G. T. Kim, G. S. Duesberg, T. Hallam, J. J. Boland, J. J. Wang, J. F. Donegan, J. C. Grunlan, G. Mo- riarty, A. Shmeliov, R. J. Nicholls, J. M. Perkins, E. M. Grieveson, K. Theuwissen, D. W. McComb, P. D. Nellist, and V. Nicolosi, Science 331, 568 (2011). Mater. Sci. 112, Part A, 377 (2016). [57] W. Li, T. Wang, X. Dai, X. Wang, C. Zhai, Y. Ma, and S. Chang., Solid State Commun. 225, 32 (2016). [58] H. Yuan, M. S. Bahramy, K. Morimoto, S. Wu, K. No- mura, B.-J. Yang, H. Shimotani, R. Suzuki, M. Toh, C. Kloc, X. Xu, R. Arita, N. Nagaosa, and Y. Iwasa, Nat. Phys. 9, 563 (2013). [59] X. Dai, W. Li, T. Wang, X. Wang, and C. Zhai, J. Appl. Phys. 117, 084310 (2015). [60] S. Leb`egue, T. Bjorkman, M. Klintenberg, R. M. Niemi- nen, and O. Eriksson, Phys. Rev. X 3, 031002 (2013). [61] C. Ataca, H. ahin, and S. Ciraci, J. Phys. Chem. C 116, 8983 (2012). [62] H. Linke, W. D. Sheng, A. Svensson, A. Lofgren, L. Christensson, H. Q. Xu, P. Omling, and P. E. Linde- lof, Phys. Rev. B 61, 15914 (2000). [63] R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V. R. and C. M. Zicovich-Wilson, Zeitschrift fur Saunders, Kristallographie 220, 571 (2005). [64] R. Dovesi, V. R. Saunders, C. Roetti, R. Orlando, F. Pas- cale, B. Civalleri, K. Doll, N. M. Harrison, I. J. Bush, M. Llunell, C. Science, and A. Technologies, Crystal09 User's Manual , University of Torino: Torino, 2nd ed. (2013). [65] J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E. Scuseria, L. A. Constantin, X. Zhou, and K. Burke, Phys. Rev. Lett. 100, 136406 (2008). [66] S. Grimme, J. Comput. Chem. 27, 1787 (2006). [67] M. F. Peintinger, D. V. Oliveira, and T. Bredow, J. [42] Z. Zeng, Z. Yin, X. Huang, H. Li, Q. He, G. Lu, F. Boey, Comput. Chem. 34, 451 (2013). and H. Zhang, Angew. Chem. Int. Ed. 50, 11093. [68] T. Bredow and M. Lerch, Zeitschrift fr anorganische und [43] Y. Zhan, Z. Liu, S. Najmaei, P. M. Ajayan, and J. Lou, allgemeine Chemie 630, 2262 (2004). Small 8, 966. [44] Y.-C. Lin, W. Zhang, J.-K. Huang, K.-K. Liu, Y.-H. Lee, C.-T. Liang, C.-W. Chu, and L.-J. Li, Nanoscale 4, 6637 (2012). [69] D. Munoz Ramo, J. L. Gavartin, A. L. Shluger, and G. Bersuker, Phys. Rev. B 75, 205336 (2007). [70] F. Cor`a, A. Patel, N. M. Harrison, C. Roetti, C. R. A. Catlow, J. Mater. Chem. 7, 959 (1997). and [45] A. Ennaoui, S. Fiechter, K. Ellmer, R. Scheer, and [71] F. Cor, A. Patel, N. M. Harrison, R. Dovesi, and C. R. A. K. Diesner, Thin Solid Films 261, 124 (1995). Catlow, J. Am. Chem. Soc. 118, 12174 (1996). [46] K. Kang, S. Xie, L. Huang, Y. Han, P. Y. Huang, K. F. Mak, C.-J. Kim, D. Muller, and J. Park, Nature 520, 656 (2015). [47] Y. Zhang, Y. Zhang, Q. Ji, J. Ju, H. Yuan, J. Shi, T. Gao, D. Ma, M. Liu, Y. Chen, X. Song, H. Y. Hwang, Y. Cui, and Z. Liu, ACS Nano 7, 8963 (2013). [72] A. Kokalj, I. Kobal, H. Horino, Y. Ohno, and T. Mat- sushima, Surf. Sci. 506, 196 (2002). [73] K. Doll, Surf. Sci. 573, 464 (2004). [74] J. Heyd, J. E. Peralta, G. E. Scuseria, and R. L. Martin, J. Chem. Phys. 123, 174101 (2005). [75] H. J. Monkhorst and J. D. Pack, Phys. Rev. B 13, 5188 [48] J. C. Shaw, H. Zhou, Y. Chen, N. O. Weiss, Y. Liu, (1976). Y. Huang, and X. Duan, Nano Res. 7, 511 (2014). [76] A. Kuc, N. Zibouche, and T. Heine, Phys. Rev. B 83, [49] Y.-H. Lee, X.-Q. Zhang, W. Zhang, M.-T. Chang, C.- T. Lin, K.-D. Chang, Y.-C. Yu, J. T.-W. Wang, C.-S. Chang, L.-J. Li, and T.-W. Lin, Adv. Mater. 24, 2320 (2012). [50] M. Chhowalla, H. S. Shin, G. Eda, L.-J. Li, K. P. Loh, and H. Zhang, Nat Chem 5, 563 (2013). 245213 (2011). [77] K. F. Mak, C. H. Lui, J. Shan, and T. F. Heinz, Phys. Rev. Lett. 102, 256405 (1 (2009). [78] Y. Zhang, T.-T. Tang, C. Girit, Z. Hao, M. C. Martin, A. Zettl, M. F. Crommie, Y. R. Shen, and F. Wang, Nature 459, 820 (2009). [51] S. Manzeli, D. Ovchinnikov, D. Pasquier, O. V. Yazyev, [79] Z. Ni, Q. Liu, K. Tang, J. Zheng, J. Zhou, R. Qin, Z. Gao, and A. Kis, Nat. Rev. Mater. 2, 17033 (2017). D. Yu, and J. Lu, Nano Lett. 12, 113 (2012). [52] X. Duan, C. Wang, A. Pan, R. Yu, and X. Duan, Chem. [80] C.-W. Chen, M.-H. Lee, and S. J. Clark, Nanotechnology Soc. Rev. 44, 8859 (2015). 15, 1837 (2004). 13 [81] Y. Liu, Z. Qiu, A. Carvalho, Y. Bao, H. Xu, S. J. R. Tan, W. Liu, A. H. Castro Neto, K. P. Loh, and J. Lu, Nano Lett. 17, 1970 (2017). [82] A. Kumar, H. He, R. Pandey, P. K. Ahluwalia, and K. Tankeshwar, Phys. Chem. Chem. Phys. 17, 19215 (2015). [83] S. Trasatti, Pure & Appl. Chem. 58, 955 (1986). [84] J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T. Bligaard, and H. J´onsson, J. Phys. Chem. B 108, 17886 (2004). [86] J. A. Reyes-Retana and F. Cervantes-Sodi, Sci. Rep. 6, 24093 (2016). [87] G. C. Loh and R. Pandey, J. Mater. Chem. C 3, 6627 (2015). [88] H. Zhang, C.-J. Tong, J. Wu, W.-J. Yin, and Y.-N. Zhang, J. Phys.: Condens. Matter 28, 015004 (2016). [89] H. Zhang, X. Lin, and Z.-K. Tang, Solid State Commun. 220, 12 (2015). [90] Y. H. Lu, B. Xu, A. H. Zhang, M. Yang, and Y. P. Feng, J. Phys. Chem. C 115, 18042 (2011). [85] M. A. Butler and D. S. Ginley, J. Electrochem. Soc. 125, [91] H. Guo, N. Lu, L. Wang, X. Wu, and X. C. Zeng, J. 228 (1978). Phys. Chem. C 118, 7242 (2014). [92] P. Mir, M. Ghorbani-Asl, and T. Heine, Angew. Chem. Int. Ed. 53, 3015 (2014). e p y T c F E ) A / V ( - - - - - - - - - - - - - M S M S - M M - M S M - M S - - M S M S - M S M - M S M - - - - - - - - - - - - - 2 3 . 0 6 1 . 0 - 0 6 . 1 3 7 . 0 - 0 1 . 1 2 4 . 0 - 0 7 . 1 - - 0 4 . 2 0 1 . 1 - 8 5 . 0 0 3 . 0 - 0 4 . 2 8 2 . 1 14 d e h s i l b u P c F E ) A / V ( M B C M B V g n i d n e b ) A / V / V e ( g n i d n e b ) A / V / V e ( - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 3 7 . 1 - 1 4 . 3 - - 6 9 . 0 - 7 8 . 1 - - 7 5 . 0 - 3 . 1 - - 0 1 . 0 - - - 8 3 . 0 - 4 2 . 0 - - 3 5 . 1 - 0 1 . 3 - - 8 4 . 0 - 5 0 . 1 - - - - - - - - - - - - - - 9 6 . 1 5 4 . 3 - 6 6 . 0 1 5 . 1 - 3 0 . 0 2 0 . 0 - 2 0 . 0 - - 8 0 . 0 - 0 0 . 0 - 0 5 . 1 0 8 . 2 - 4 8 . 0 0 8 . 1 o i t a r S ) i b / i r t ( - - - - 0 0 . 2 8 0 . 2 5 1 . 2 0 0 . 0 2 0 . 0 ± 3 4 . 3 4 0 . 0 ± 7 8 . 6 0 0 . 0 5 0 . 0 ± 2 6 . 1 4 0 . 0 ± 7 3 . 3 0 0 . 0 5 0 . 0 ± 0 6 . 0 8 0 . 0 ± 9 2 . 1 - 0 0 . 0 1 0 . 0 ± 2 1 . 0 - 3 8 . 0 1 9 . 1 4 1 . 2 0 0 . 0 2 0 . 0 ± 9 2 . 0 3 0 . 0 ± 4 2 . 0 0 0 . 0 2 0 . 0 ± 3 0 . 3 5 5 . 0 ± 1 9 . 5 0 0 . 0 9 0 . 0 ± 3 3 . 1 7 1 . 0 ± 5 8 . 2 S ) A ( 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 d e h s i l b u P ) A ( S t a g E d l e fi o r e z ) V e ( d e h s i l b u P t a g E d l e fi o r e z ) V e ( ) V e ( s r e y a L g E k l u B ) A ( S I ) A ( c ) A ( a l a i r e t a M - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 9 5 . 2 0 0 . 0 0 0 . 0 8 5 . 1 0 0 . 0 0 0 . 0 7 5 . 1 0 0 . 0 0 0 . 0 3 4 . 1 0 0 . 0 0 0 . 0 2 1 . 1 9 0 . 1 8 0 . 1 8 6 . 2 1 6 . 2 6 5 . 2 4 0 . 1 2 7 . 0 6 5 . 0 2 4 . 0 5 2 . 0 0 0 . 0 1 1 . 1 8 7 . 0 1 4 . 0 2 6 . 1 8 0 . 2 4 0 . 2 1 4 . 4 6 1 . 4 5 0 . 4 , ] 6 8 [ 1 2 5 . 1 , ] 1 6 [ 5 0 . 1 o n o M ] 8 8 [ 0 3 . 1 , ] 7 8 [ 0 9 . 2 0 0 . 0 6 1 . 4 2 3 . 8 1 2 . 3 2 O c S - H , ] 6 8 [ 1 2 7 . 0 , ] 1 6 [ 4 4 . 0 o n o M ] 7 8 [ 0 - i B i r T ] 9 8 [ 4 7 . 0 - - i B i r T ] 6 8 [ 6 5 4 . 0 , ] 1 6 [ 7 2 . 0 o n o M - - ] 1 6 [ 0 0 . 0 - - i B i r T o n o M i B i r T ] 0 9 [ 1 . 2 , ] 2 2 [ 0 1 . 1 o n o M - - ] 2 2 [ 5 6 . 2 - - ] 2 2 [ 2 6 . 0 - - ] 2 2 [ 2 4 . 0 - - i B i r T o n o M i B i r T o n o M i B i r T o n o M i B i r T ] 1 6 [ 0 , ] 2 2 [ 0 o n o M - - ] 2 2 [ 9 5 . 1 - - ] 2 2 [ 7 3 . 4 - - i B i r T o n o M i B i r T o n o M i B i r T 0 0 . 0 5 2 . 5 5 . 0 1 3 7 . 3 2 S c S - H 0 0 . 0 8 8 . 5 5 7 . 1 1 8 8 . 3 2 e S c S - H 0 0 . 0 5 9 . 5 1 9 . 1 1 5 1 . 4 2 e T c S - H 8 0 . 1 5 6 . 6 0 3 . 3 1 0 9 . 2 4 5 . 2 9 7 . 4 9 7 . 4 6 9 . 2 1 5 . 0 7 3 . 6 3 7 . 2 1 0 3 . 3 0 0 . 0 0 7 . 6 0 4 . 3 1 0 8 . 3 6 2 . 0 2 9 . 6 4 8 . 3 1 2 7 . 3 i 2 O T H - i 2 O T T - i 2 S T H - i 2 e S T H - i 2 e T T H - 4 5 . 1 2 1 . 6 4 2 . 2 1 2 1 . 3 2 O r Z - H 6 9 . 3 1 8 . 4 1 8 . 4 6 2 . 3 2 O r Z - T a , ) g E ( d l e fi c i r t c e l e t u o h t i w s e r u t c u r t s k l u b f o s p a g d n a b , S I n o i t a r a p e s r e y a l r e t n i , ) c r a l u c i d n e p r e p - e n a l p d n a a e n a l p - n i ( s r e t e m a r a p e c i t t a l d e x a l e R . I I I E L B A T s u o i r a v r o f d l e fi o r e z t a g E s p a g d n a b d e t u p m o c r u o , s r e y a l f o s r e b m u n s u o i r a v s e i r a d n u o b r e w o l d n a r e p p u e h t h t i w g n o l a S d e t u p m o c r u o , d l e fi c i r t c e l e r e d n u r o f ) S ( d l e fi o r e z t a s p a g d n a b l a c i t e r o e h t r o / d n a l a t n e m i r e p x e d e h s i l b u p e h t f o w e i v e r s t n e i c ffi e o c ) E S G ( t c e ff e - k r a t S - t n a i g d e h s i l b u p e h t f o w e i v e r a , s r e y a l f o s r e b m u n e r e h w c F E s d l e fi l a c i t i r c d e t u p m o c r u o d n a d e h s i l b u p , d l e fi r e d n u s M B C d n a s M B V e h t . ) l a t e m m e s i r o f M S , l a t e m r o f M ( n o i t i s n a r t e s a h p e h t r e t f a l a i r e t a m f o f o s e t a r g n i d n e b , o i t a r S r e y a l i b o t r e y a l i r t , s l a v r e t n i e c n e d fi n o c % 5 9 e h t f o s e p y t e h t d n a , r u c c o s n o i t i s n a r t e s a h p i l a t e m m e s / l a t e m - o t - r o t c u d n o c i m e s e p y T c F E ) A / V ( - M S M S - M S M - M S - - M S M - M S - - - - - M S M S - M M - M S M - M S M S - M S - - M S M S - 4 0 . 1 4 3 . 0 - 6 9 . 0 6 4 . 0 - 8 0 . 0 - - 4 2 . 0 6 0 . 0 - 0 7 . 0 - - - - - 0 6 . 1 2 3 . 1 - 8 5 . 1 0 8 . 0 - 5 0 . 1 0 3 . 0 - 5 1 . 1 5 5 . 0 - 0 8 . 1 - - 8 8 . 0 0 4 . 0 15 d e h s i l b u P c F E ) A / V ( M B C M B V g n i d n e b ) A / V / V e ( g n i d n e b ) A / V / V e ( o i t a r S ) i b / i r t ( S ) A ( d e h s i l b u P ) A ( S - ] 2 8 [ a 2 . 0 - - ] 2 8 [ a 7 2 . 0 - - ] 2 8 [ a 5 1 . 0 - - ] 2 8 [ a 7 1 . 0 - - ] 2 8 [ a 7 0 . 0 - - ] 2 8 [ a 5 0 . 0 - - - - - - - - - - - - - - - - - - - - 2 5 . 0 - 0 9 . 0 - - 1 3 . 0 - 7 8 . 0 - - 6 0 . 2 - - - 0 2 . 0 - 5 7 . 0 - - 2 0 . 0 - - - - - - 0 1 . 1 - 3 8 . 0 - - 5 6 . 1 - 5 2 . 3 - - 7 5 . 0 - 7 1 . 1 - - 5 6 . 0 - 4 3 . 1 - - 5 1 . 0 - - - 6 2 . 0 - 4 5 . 0 - - 5 1 . 0 0 2 . 0 - 7 5 . 0 8 0 . 1 - 6 3 . 1 - - 0 2 . 0 4 1 . 1 - 9 0 . 0 - - - - - 7 7 . 0 2 1 . 0 - 0 5 . 1 4 1 . 3 - 9 1 . 0 4 4 . 0 - 2 7 . 0 5 5 . 1 - 5 0 . 0 - - 7 3 . 0 1 5 . 0 6 6 . 1 0 2 . 2 0 0 . 0 4 0 . 0 ± 7 6 . 0 6 0 . 0 ± 1 1 . 1 0 0 . 0 7 0 . 0 ± 9 8 . 0 0 1 . 0 ± 6 9 . 1 - 0 0 . 0 6 0 . 0 ± 3 4 . 3 - 3 6 . 4 - - 0 5 . 0 3 0 . 2 0 1 . 2 0 1 . 2 0 0 . 0 9 0 . 0 ± 1 4 . 0 0 9 . 1 ± 0 9 . 1 0 0 . 0 8 0 . 0 ± 2 1 . 0 - - - - 0 0 . 0 5 0 . 0 ± 1 9 . 1 2 0 . 0 ± 6 9 . 0 0 0 . 0 3 0 . 0 ± 4 1 . 3 3 0 . 0 ± 9 3 . 6 0 0 . 0 6 0 . 0 ± 6 7 . 0 1 1 . 0 ± 0 6 . 1 0 0 . 0 6 0 . 0 ± 8 3 . 1 8 3 . 0 ± 9 8 . 2 - 0 0 . 0 3 0 . 0 ± 0 2 . 0 - 7 6 . 1 0 0 . 0 3 3 . 0 ± 4 6 . 0 1 6 . 0 ± 6 0 . 1 - ] 2 8 [ 1 6 . 4 - - ] 2 8 [ 3 5 . 5 - - ] 2 8 [ 2 . 5 - - ] 2 8 [ 8 2 . 5 - - ] 2 8 [ 0 . 4 - - ] 2 8 [ 6 . 0 - - - - - - - - - - - - - - - - - - - t a g E d l e fi o r e z ) V e ( 5 8 . 0 3 6 . 0 7 4 . 0 8 0 . 1 1 0 . 1 6 9 . 0 5 3 . 0 8 2 . 0 0 0 . 0 4 2 . 0 2 1 . 0 6 0 . 0 9 3 . 0 0 2 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 1 1 . 2 9 3 . 2 0 5 . 1 0 0 . 5 2 0 . 5 5 0 . 5 7 0 . 1 4 8 . 0 9 6 . 0 1 7 . 1 5 6 . 1 2 6 . 1 5 8 . 0 5 5 . 0 0 0 . 0 4 7 . 0 5 6 . 0 1 6 . 0 d e h s i l b u P t a g E d l e fi o r e z ) V e ( ] 2 2 [ ] 2 8 [ 5 8 . 0 0 9 . 0 - ] 2 8 [ 3 4 . 1 - ] 2 2 [ ] 2 8 [ 4 6 . 0 5 7 . 0 o n o M i B i r T i B i r T o n o M i B i r T ] 1 9 [ 0 1 . 1 , ] 2 2 [ 3 0 . 1 o n o M ] 1 9 [ 5 4 . 0 - , ] 2 8 [ 9 . 0 - ] 2 2 [ ] 2 8 [ 8 1 . 0 0 3 . 0 - ] 1 9 [ 0 , ] 2 2 [ 0 ] 2 8 [ 3 0 . 0 - ] 2 2 [ 0 8 . 1 - - ] 2 2 [ 3 6 . 4 - - ] 2 2 [ 3 9 . 0 - - i B i r T o n o M i B i r T o n o M i B i r T o n o M i B i r T o n o M i B i r T o n o M i B i r T ] 1 9 [ 7 2 . 1 , ] 2 2 [ 6 0 . 1 o n o M ] 2 2 [ 5 2 . 0 o n o M i B i r T o n o M i B i r T ] 1 9 [ 1 6 . 0 - - , - - - - ] 2 2 [ 0 7 . 0 ) V e ( s r e y a L g E k l u B ) A ( S I ) A ( c ) A ( a l a i r e t a M 0 0 . 0 7 6 . 5 3 3 . 1 1 1 5 . 3 2 S r Z - H 4 7 . 0 5 8 . 5 5 8 . 5 2 6 . 3 2 S r Z - T 5 1 . 0 5 7 . 7 0 5 . 5 1 0 1 . 4 2 e S r Z - H 1 0 . 0 7 0 . 6 7 0 . 6 4 7 . 3 2 e S r Z - T 0 0 . 0 4 1 . 7 8 2 . 4 1 4 8 . 3 2 e T r Z - H 0 0 . 0 2 5 . 6 2 5 . 6 7 8 . 3 2 e T r Z - T 0 0 . 2 0 0 . 6 0 0 . 2 1 0 1 . 3 7 6 . 4 0 9 . 5 0 9 . 5 5 2 . 3 f 2 O H H - f 2 O H T - 3 4 . 0 2 4 . 6 3 8 . 2 1 8 4 . 3 2 S f H H - 0 3 . 1 5 8 . 5 5 8 . 5 3 6 . 3 2 S f H T - 7 5 . 0 5 4 . 7 0 9 . 4 1 0 0 . 4 2 e S f H H - ] 2 2 [ 0 3 . 0 o n o M i B i r T 1 5 . 0 5 1 . 6 5 1 . 6 4 7 . 3 2 e S f H T - . s e r u g fi e h t m o r f d e t a m i t s e s e u l a V a e p y T c F E ) A / V ( - M M S - 9 1 . 1 8 3 . 0 - - - - M S M S - M S M - M M - - - - - 2 1 . 0 6 0 . 0 - 8 2 . 1 2 6 . 0 - 8 2 . 1 4 5 . 0 - M S 4 1 . 0 - - M S M S - - - 0 5 . 0 4 2 . 0 - M S 5 4 . 1 M S - M M 7 6 . 0 - 0 5 . 2 0 5 . 1 16 - - - - - - - - - - - - - - - - - - - - - - - 5 5 . 0 - 5 1 . 0 - - - - - 8 8 . 1 - 0 0 . 4 - - 7 5 . 0 - 4 2 . 1 - - 6 3 . 0 - 0 9 . 0 - - 2 5 . 0 - - - 3 0 . 1 - 7 2 . 2 - - - 7 3 . 0 - 5 0 . 0 - - - - 3 0 . 2 3 8 . 3 - 9 0 . 0 8 0 . 0 - 0 0 . 0 6 3 . 0 - 1 1 . 1 - 0 0 . 2 0 0 . 2 7 5 . 3 4 2 . 0 - - - - 2 7 . 0 5 0 . 2 - 0 0 . 0 3 0 . 0 ± 8 1 . 0 3 0 . 0 ± 0 2 . 0 - - - 0 0 . 0 1 0 . 0 ± 1 9 . 3 7 0 . 0 ± 2 8 . 7 0 0 . 0 1 0 . 0 ± 6 6 . 0 5 0 . 0 ± 3 3 . 1 0 0 . 0 0 0 . 0 ± 6 3 . 0 6 0 . 0 ± 7 2 . 1 0 0 . 0 2 0 . 0 ± 8 2 . 0 - 0 0 . 0 8 5 . 2 5 0 . 0 ± 5 7 . 1 3 1 . 0 ± 1 5 . 4 0 0 . 0 ] 3 5 [ 0 - - - - - - - - - - - - - - - - - - - - - , ] 4 2 [ a 5 . 1 ] 3 5 [ 3 . 0 , ] 4 5 [ a 5 . 1 ] 5 5 [ ] 6 5 [ 5 5 . 0 5 1 . 0 - , ] 4 2 [ a 5 . 1 ] 3 5 [ 5 2 . 0 - 8 8 . 0 - 8 0 . 0 - 0 9 . 1 2 5 . 0 ± 0 8 . 0 9 2 . 1 - - 6 0 . 0 - 6 0 . 0 - 3 2 . 0 - 7 1 . 0 0 0 . 0 7 6 . 3 1 2 . 0 ± 2 5 . 1 0 0 . 0 1 0 . 0 ± 6 0 . 0 4 0 . 0 ± 3 2 . 0 , ] 3 5 [ 0 5 . 5 , ] 4 5 [ a 6 6 . 0 , ] 5 5 [ ] 4 2 [ a 5 2 . 2 a 3 8 . 0 ] 6 5 [ 2 . 6 - ] 3 5 [ 0 ] 4 2 [ , ] 3 5 [ 5 2 . 6 - t a g E d l e fi o r e z ) V e ( 7 6 . 0 5 4 . 0 2 1 . 0 0 0 . 0 0 0 . 0 0 0 . 0 8 4 . 0 7 4 . 0 7 4 . 0 4 3 . 1 4 0 . 1 6 9 . 0 3 8 . 0 4 5 . 0 4 7 . 0 4 3 . 0 4 0 . 0 0 0 . 0 2 0 . 1 0 9 . 0 2 9 . 0 0 8 . 1 0 3 . 1 8 1 . 1 3 4 . 1 8 9 . 0 8 7 . 0 d e h s i l b u P c F E ) A / V ( M B C M B V i g n d n e b ) A / V / V e ( i g n d n e b ) A / V / V e ( ) i S o i t a r b / i r t ( S ) A ( d e h s i l b u P ) A ( S s r e y a L g E k l u B ) V e ( S I ) A ( ) A ( c ) A ( a l a i r e t a M d e h s i l b u P t a g E d l e fi o r e z ) V e ( ] 2 2 [ 6 0 . 0 - - , ] 2 2 [ 3 4 . 0 , ] 1 6 [ 0 5 . 0 o n o M ] 1 9 [ 0 , ] 2 2 [ 0 o n o M - - i B i r T o n o M i B i r T , ] 4 2 [ ] 6 7 [ 9 . 1 , ] 6 8 [ 6 0 7 . 1 a 5 2 . 1 , ] 2 1 [ 8 5 . 1 i B ] 1 9 [ 2 9 . 0 , ] 6 8 [ 9 2 9 . 0 , ] 2 2 [ 0 9 . 0 , ] 1 6 [ 7 0 . 1 o n o M - - i B i r T ] 1 9 [ 4 7 . 0 , ] 6 8 [ 6 5 7 . 0 , ] 2 2 [ 0 7 . 0 , ] 1 6 [ 6 8 . 0 o n o M - - i B i r T ] 1 9 [ 0 6 . 0 , ] 6 8 [ 4 3 5 . 0 , ] 2 2 [ 5 4 . 0 , ] 1 6 [ 0 6 . 0 o n o M - - i B i r T , ] 2 2 [ 1 9 . 0 , ] 1 6 [ 7 9 . 0 o n o M - - i B i r T 0 1 . 1 5 1 . 6 1 3 . 2 1 0 9 . 2 2 S r C H - 7 6 . 0 9 3 . 6 7 7 . 2 1 8 0 . 3 2 e S r C H - 5 0 . 0 8 4 . 6 6 9 . 2 1 6 4 . 3 2 e T r C H - ] 6 8 [ 1 8 3 . 0 5 4 . 0 9 1 . 7 7 3 . 4 1 8 5 . 2 2 O r C H - , ] 2 1 [ 8 8 . 1 , ] 2 [ 5 8 . 1 o n o M - - i B i r T ] 6 8 [ 8 9 8 . 0 1 9 . 0 7 0 . 6 3 1 . 2 1 0 8 . 2 2 O o M H - , ] 2 2 [ 8 5 . 1 , ] 1 6 [ 7 8 . 1 6 1 . 1 6 1 . 6 2 3 . 2 1 5 1 . 3 2 S o M H - ] 6 7 [ 5 . 1 , ] 3 5 [ 5 2 . 1 ] 6 5 [ a 7 . 0 , ] 2 1 [ 5 4 . 1 i r T , ] 1 6 [ 2 6 . 1 , ] 2 2 [ 2 3 . 1 o n o M ] 6 8 [ 8 3 4 . 1 ] 3 5 [ 4 1 . 1 - i B i r T 1 9 . 0 5 4 . 6 0 9 . 2 1 8 2 . 3 2 e S o M H - . s e r u g fi e h t m o r f d e t a m i t s e s e u l a V a 0 0 . 0 5 9 . 6 9 8 . 3 1 4 8 . 3 0 0 . 0 7 6 . 6 7 6 . 6 5 9 . 3 f 2 e T H H - f 2 e T H T - e p y T c F E ) A / V ( - M M - M S M S - - 9 7 . 1 5 8 . 0 - 1 2 . 0 8 1 . 0 - d e h s i l b u P c F E ) A / V ( - ] 3 5 [ 2 . 0 - - - - - - 5 0 . 0 - 7 0 . 0 - - 0 2 . 1 - 2 2 . 2 - - M 0 9 . 1 a 2 . 1 , ] 3 5 [ , ] 4 2 [ 7 2 . 0 0 6 . 0 - M S - 2 8 . 0 - ] 7 5 [ 3 3 2 . 0 - - 8 2 . 1 - - M S 0 7 . 2 ] 9 5 [ 2 . 0 , ] 4 2 [ 4 2 . 1 0 0 . 0 M S - M M - 4 0 . 1 - 0 8 . 1 0 9 . 0 - M S M 0 . 4 b 0 2 . 0 > - - - - - - 17 - - - - - - ] 9 5 [ 7 9 0 . 0 - - - - - - - - - - - - 8 5 . 0 - - 2 2 . 0 - 5 0 . 0 - - 3 1 . 0 - 8 2 . 0 - - - - - - - g n i d n e b ) A / V / V e ( M B C M B V g n i d n e b ) A / V / V e ( - 4 0 . 0 3 2 . 0 - 4 0 . 1 5 5 . 1 - 1 0 . 0 7 2 . 0 - 4 2 . 0 9 1 . 0 - 5 3 . 0 9 3 . 0 - 1 0 . 0 2 1 . 0 - - - - - - o i t a r S ) i b / i r t ( S ) A ( d e h s i l b u P ) A ( S 0 0 . 0 - 1 3 . 3 8 6 . 1 9 0 . 0 ± 9 0 . 0 3 0 . 0 ± 1 3 . 0 0 0 . 0 6 0 . 0 ± 4 2 . 2 0 2 . 0 ± 7 7 . 3 0 0 . 0 0 5 . 2 7 0 . 0 ± 2 6 . 0 0 3 . 3 6 7 . 0 8 8 . 2 - - 5 6 . 0 ± 5 5 . 1 0 0 . 0 7 0 . 0 ± 3 2 . 0 7 0 . 0 ± 7 7 . 0 0 0 . 0 6 0 . 0 ± 8 5 . 0 1 1 . 0 ± 4 4 . 0 0 0 . 0 2 0 . 0 ± 5 1 . 0 4 0 . 0 ± 2 4 . 0 0 0 . 0 - - 0 0 . 0 - - ] 3 5 [ 2 6 . 6 - - - - - , ] 3 5 [ 1 9 . 5 , ] 4 2 [ a 0 9 . 0 ] 7 5 [ a 3 8 . 5 - - , ] 4 2 [ a 4 8 . 0 ] 9 5 [ a 6 ] 9 5 [ a 4 3 . 1 1 - - - - - - - - - - - - t a g E d l e fi o r e z ) V e ( d e h s i l b u P t a g E d l e fi o r e z ) V e ( s r e y a L g E k l u B ) V e ( S I ) A ( ) A ( c ) A ( a l a i r e t a M 9 0 . 1 0 9 . 0 6 7 . 0 9 7 . 0 7 4 . 0 7 6 . 0 0 9 . 1 4 4 . 1 8 3 . 1 0 5 . 1 0 4 . 1 7 9 . 0 8 2 . 1 0 2 . 1 0 0 . 1 8 1 . 1 8 8 . 0 0 8 . 0 4 5 . 0 0 0 . 0 0 0 . 0 5 2 . 0 0 0 . 0 0 0 . 0 , ] 2 2 [ 3 9 . 0 , ] 1 6 [ 5 2 . 1 o n o M ] 6 8 [ 6 1 1 . 1 ] 3 5 [ 0 9 . 0 - i B i r T , ] 2 2 [ 2 3 . 1 , ] 1 6 [ 7 3 . 1 o n o M 0 8 . 0 9 9 . 6 7 9 . 3 1 1 5 . 3 2 e T o M H - ] 6 8 [ 9 4 3 . 1 0 7 . 0 8 0 . 6 5 1 . 2 1 0 8 . 2 2 O W H - , ] 1 6 [ 8 9 . 1 , ] 2 2 [ 1 5 . 1 o n o M - - i B i r T ] 6 8 [ 1 7 7 . 1 , ] 6 7 [ 1 . 2 ] 6 7 [ 5 . 1 , ] 3 5 [ 6 3 . 1 - i B i r T , ] 1 6 [ 8 6 . 1 , ] 2 2 [ 2 2 . 1 o n o M ] 6 8 [ ] 9 5 [ 5 3 5 . 1 a 2 . 1 ] 9 5 [ a 1 . 1 ] 1 6 [ 4 2 . 1 - - i B i r T o n o M i B i r T 0 3 . 1 6 1 . 6 2 3 . 2 1 5 1 . 3 2 S W H - 2 9 . 0 8 4 . 6 6 9 . 2 1 8 2 . 3 2 e S W H - 0 1 . 1 4 1 . 7 8 2 . 4 1 6 4 . 3 2 e T W H - , ] 1 6 [ 0 1 . 0 , ] 2 2 [ 0 o n o M ] 2 9 [ 0 - i B i r T , ] 1 6 [ 1 5 . 0 , ] 2 2 [ 1 5 . 0 o n o M - - i B i r T , ] 1 6 [ 8 3 . 1 , ] 2 2 [ 7 1 . 1 o n o M ] 6 8 [ 5 6 2 . 1 9 8 . 0 2 4 . 4 2 4 . 4 0 8 . 2 ] 2 9 [ 1 5 . 0 , ] 6 8 [ 1 6 5 . 0 0 0 . 0 6 4 . 4 6 4 . 4 8 2 . 3 - - i B i r T ] 6 8 [ 4 9 0 . 0 0 0 . 0 6 1 . 5 6 1 . 5 5 5 . 3 i 2 O N T - i 2 S N T - i 2 e S N T - . s d l e fi h g i h t a e g r e v n o c t o n s e o d e r u t c u r t S . s e r u g fi e h t m o r f d e t a m i t s e s e u l a V a b e p y T c F E ) A / V ( - M S M - 0 5 . 2 8 9 . 0 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - M 0 8 . 0 18 d e h s i l b u P c F E ) A / V ( M B C M B V g n i d n e b ) A / V / V e ( g n i d n e b ) A / V / V e ( o i t a r S ) i b / i r t ( S ) A ( d e h s i l b u P ) A ( S - - - - - - - - - - - - - - - - - - - - - - - - - 5 0 . 0 - 8 2 . 0 - 0 2 . 0 2 9 . 0 2 5 . 2 0 0 . 0 3 0 . 0 ± 5 2 . 0 6 0 . 0 ± 3 6 . 0 - - - - - - - - - - - - - - - - - - - - - - 0 3 . 0 - 0 0 . 1 - - - - - - - - - - - - - - - - - - - - - - - - - 0 0 . 0 0 0 . 0 - 0 0 . 0 - - 0 0 . 0 - - 0 0 . 0 0 0 . 0 7 0 . 0 ± 0 3 . 1 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 - - - - - - - - - - - - - - - - - - - - - - - - - - t a g E d l e fi o r e z ) V e ( 9 3 . 1 0 1 . 1 0 0 . 1 6 3 . 1 1 0 . 0 0 0 . 0 4 8 . 0 0 0 . 0 0 0 . 0 0 3 . 0 0 0 . 0 0 0 . 0 9 2 . 1 2 1 . 1 6 0 . 1 7 7 . 1 1 8 . 0 1 6 . 0 3 4 . 1 0 3 . 0 1 0 . 0 3 8 . 0 0 0 . 0 0 0 . 0 ] 2 9 [ 1 1 . 1 , ] 2 2 [ 1 1 . 1 o n o M ] 2 9 [ 0 0 . 0 - ] 2 2 [ 8 4 . 0 - - ] 2 2 [ 0 - - ] 2 2 [ 0 6 . 1 - - i B i r T o n o M i B i r T o n o M i B i r T o n o M i B i r T ] 2 9 [ 5 7 . 1 , ] 2 2 [ 1 6 . 1 o n o M ] 2 9 [ 4 6 . 0 - ] 2 2 [ 7 0 . 1 - - ] 2 2 [ 3 2 . 0 - - i B i r T o n o M i B i r T o n o M i B i r T o n o M i B i r T 2 1 . 1 6 2 . 4 6 2 . 4 7 0 . 3 2 O d P T - 0 0 . 0 2 5 . 4 2 5 . 4 2 5 . 3 2 S d P T - 0 0 . 0 7 5 . 4 7 5 . 4 5 7 . 3 2 e S d P T - 0 0 . 0 9 1 . 5 9 1 . 5 2 0 . 4 2 e T d P T - 7 1 . 1 1 2 . 4 1 2 . 4 5 2 . 3 2 O t P T - 6 1 . 0 4 0 . 5 4 0 . 5 4 5 . 3 2 S t P T - 0 0 . 0 8 0 . 5 8 0 . 5 2 7 . 3 2 e S t P T - 0 0 . 0 2 2 . 5 2 2 . 5 2 0 . 4 2 e T t P T - d e h s i l b u P t a g E d l e fi o r e z ) V e ( ] 2 2 [ 0 3 . 1 - - ) V e ( s r e y a L g E k l u B ) A ( S I ) A ( c ) A ( a l a i r e t a M
1705.07720	1	1705	2017-05-07T20:00:48	Enhanced Second-harmonic Generation Using Nonlinear Metamaterials	[ "physics.app-ph" ]	We demonstrate a nonlinear metamaterial which enhance higher order harmonics in microwave frequency regime. Nonlinearity in the structure is introduced by adding a varactor diode in the common slit of the double split ring resonator (DSRR) design. By engineering the structure such that inner ring resonance frequency of the DSRR is twice as the outer ring resonance frequency, we have demonstrated that the second harmonic of the outer ring can be enhanced significantly. By comparing with a single ring (SRR) unit cell structure, DSRR has an enhancement factor of 70. Furthermore, we elucidate that the second harmonic signals generated by two identical double split rings interfere and we can use its constructive interference positions to construct an array of DSRRs that gives a maximum second harmonic signal with phase matching condition. When the phase matching condition occurs the enhancement factor of an array is more than the contribution of individual unit cells which proves that the enhancement is due to not only constructive interference of the second harmonic generated by individual unit cells but also due to cavity mechanism occurred in the array structure.	physics.app-ph	physics	Enhanced Second-harmonic Generation Using Nonlinear Metamaterials Sinhara R Silva, Khagendra Battarai, Alexander Shields, Jiangfeng Zhou University of South Florida Abstract We demonstrate a nonlinear metamaterial which enhance higher order harmonics in microwave frequency regime. Nonlinearity in the structure is introduced by adding a varactor diode in the common slit of the double split ring resonator (DSRR) design. By engineering the structure such that inner ring resonance frequency of the DSRR is twice as the outer ring resonance frequency, we have demonstrated that the second harmonic of the outer ring can be enhanced significantly. By comparing with a single ring (SRR) unit cell structure, DSRR has an enhancement factor of 70. Furthermore, we elucidate that the second harmonic signals generated by two identical double split rings interfere and we can use its constructive interference positions to construct an array of DSRRs that gives a maximum second harmonic signal with phase matching condition. When the phase matching condition occurs the enhancement factor of an array is more than the contribution of individual unit cells which proves that the enhancement is due to not only constructive interference of the second harmonic generated by individual unit cells but also due to cavity mechanism occurred in the array structure. Introduction A great deal of research in the field of metamaterial was conducted in a linear regime, where the electromagnetic responses are independent of the external electric or magnetic fields. Unfortunately, in linear regime the desired properties of metamaterials have only been achieved within a narrow bandwidth, around a fixed frequency. Therefore, nonlinearity is introduced into metamaterials by merging meta-atoms with well-known nonlinear materials. Among different ways to introduce nonlinearity to the metamaterial, embedding an additional controllable media[1, 2] or inserting nonlinear elements[3-5] to the structure has led to extensive studies on nonlinear metamaterials and demonstrated novel ways to obtain nonlinear responses in metamaterials. Studies have shown nonlinear metamaterials show fascinating features such as resonant frequency tuning[6-8], bistability[9, 10], harmonic generation[11-13], parametric amplification[14] , modulation[15] and backward phase matching[16] . Among basic metamaterial structures, split ring resonator (SRR) as shown in fig 1(a) is known as the most common, best characterized[9, 17-20] metamaterial with geometrically scalable meta-atoms that can be translated to operability in many regimes of frequencies. Recent studies[21] have shown that nonlinear resonating structures such as SRRs can be advantageous to use in nonlinear metamaterial designs since it has been observed that Split ring resonator produces a huge electric field at its gap at the resonance even it is exposed to an low incident electromagtic waves. In order to increase the nonlinear polarization generated by a non linear medium we can either increase suseptibility χ(n) or increase the applied electric field in the medium which explain by the nonlinear polarization equation (1). Wave equation for Propagation of Electromagnetic waves in a nonlinear medium (cid:1487)(cid:2870)(cid:2035)(cid:3398) 1(cid:1855)(cid:2870) (cid:2034)(cid:2870)(cid:2035)(cid:2034)(cid:1872)(cid:2870) (cid:3404) (cid:2020)(cid:2868)(cid:2034)(cid:2870)(cid:1842)(cid:2034)(cid:1872)(cid:2870) Polarization density (cid:1842)(cid:3404)(cid:2013)(cid:2868)(cid:3435)(cid:2031)(cid:4666)(cid:2869)(cid:4667)(cid:2035)(cid:3397)(cid:2031)(cid:4666)(cid:2870)(cid:4667)(cid:2035)(cid:2870)(cid:3397)(cid:2031)(cid:4666)(cid:2871)(cid:4667)(cid:2035)(cid:2871)(cid:3397)⋯(cid:3439) (cid:1842)(cid:3404)(cid:2013)(cid:2868)(cid:2031)(cid:4666)(cid:2869)(cid:4667)(cid:2035)(cid:3397)(cid:1842)(cid:3015)(cid:3013) (cid:1842)(cid:3015)(cid:3013)(cid:3404) (cid:2013)(cid:2868)(cid:3435)(cid:2031)(cid:4666)(cid:2870)(cid:4667)(cid:2035)(cid:2870)(cid:3397)(cid:2031)(cid:4666)(cid:2871)(cid:4667)(cid:2035)(cid:2871)(cid:3397)⋯(cid:3439) …………………….(1) Induced polarization is given by Where (cid:2035) is the applied external electric field. In natural materials, higher order nonlinear susceptibility χ(2) , χ(3) , χ(4) .. are extemely small[22] and hence nonlinear polarization is typically very weak. In this letter, we present a nonlinear metamaterial with enhanced second harmonic signal using a modified Split ring resonator (SRR) design. The modification of the split ring resonator is created by adding another ring as shown in Fig 1(c) which is a Double Split Ring Resonator (DSRR) with a shared varactor diode. Inserting a varactor diode to the design provides the nonlinearity to the unit cell. Fig 1(a) shows the transmission through SRR and DSRR where SRR has a fundamental resonance frequency at (cid:2188)(cid:2778) = 822MHz and DSRR has two distinct peaks with fundamental resonance frequencies at (cid:2188)(cid:2778) = 825MHz and (cid:2188)(cid:2779) = 1670MHz corresponding to two rings. The current density of the DSRR unit cell at the resonance frequencies are shown near each resonance frequency, which shows a LC resonance mode for the fundamental frequencies. Fig 1(b) shows the relationship between the inner ring frequency and outer ring frequency by plotting outer ring frequency with half of the inner ring frequency. 1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 0.4 (a) (cid:2188)(cid:2778) 0.8 i i n o s s m s n a r T DSR SRR (cid:2188)(cid:2779) (cid:2188)(cid:2778) (cid:2188)(cid:2779) (cid:2779)⁄ (b) 1.1 1.0 0.9 0.8 0.7 0.6 i i n o s s m s n a r T 0.5 0.750 0.775 0.800 0.825 0.850 0.875 0.900 Frequency(GHz) (c) 1.0 1.2 1.4 Frequency (GHz) 1.6 1.8 SRR DSRR Figure 1 (a) Transmission for DSRR and SRR. (b) Alignment of inner ring and outer ring resonances for DSRR. (c) SRR and DSRR design A varactor loaded split ring resonator can be moddeled as a RLC circuit as shown in fig 2(a) and fig 2(b) and in DSRR there is an extra capacitance added to the circuit in parallel to the exsiting capacitance due to the second ring in the structure. Figure 2 (a) Equivalent circuit of nonlinear SRR with diode. (b)&(c) Equivalent circuit of SRR (b) and DSRR (c) with effective circuit of diode. In varactor loaded DSRR, the external signal provides the driven source which can be modeled as a virtual voltage source in the equavalent cicuit model. Where L is the total inductance of the DSRR, including self-inductance and mutual (cid:1848)(cid:3032)(cid:3051)(cid:3047)(cid:3032)(cid:3045)(cid:3041)(cid:3028)(cid:3039)(cid:4666)(cid:1872)(cid:4667)(cid:3404) (cid:1848)(cid:2868) cos(cid:2033)(cid:1872) (cid:1838)(cid:3031)(cid:3010)(cid:3031)(cid:3047)(cid:3397)(cid:1844)(cid:1835)(cid:3397)(cid:1848)(cid:3004)(cid:3397)(cid:1848)(cid:3005)(cid:3404) (cid:1848)(cid:2868) cos(cid:2033)(cid:1872) inductance, R is the effective resistance of the unit cell and (cid:1848)(cid:3005) is the voltage across the corresponding to its fundamental frequency (cid:2033)(cid:2868)(cid:3404) 1√(cid:1838)(cid:1829)⁄ varactor diode. Varactor loaded split ring resonator structure produces a second harmonic signal . Since the capacitance of the varactor diode depends on the reverse bias voltage across the diode is given by the following expression, (cid:1848)(cid:3005)(cid:3404) 1(cid:1829) (cid:3505)(cid:1835) (cid:1856)(cid:1872)(cid:3404) (cid:3047) (cid:2868) (cid:1848)(cid:3017) (cid:3429) 1 (cid:3398) (cid:3436)1(cid:3398)(cid:1843)(cid:1829)° 1(cid:3398)(cid:1839)(cid:1848)(cid:3017) (cid:3440)(cid:2869)(cid:4666)(cid:2869)(cid:2879)(cid:3014)(cid:4667) (cid:3415) (cid:3433) (cid:3404)(cid:1869)(cid:3397)(cid:1853)(cid:1869)(cid:2870)(cid:3397)(cid:1854)(cid:1869)(cid:2871)(cid:3397)(cid:1855)(cid:1869)(cid:2872)(cid:3397)⋯ Where higher order terms contributes to the harmonics generated by the structure. In the letter we only consider second harmonic generated by the varactor loaded SRR and DSRR structures. The amplitude of the second harmonic signal can be obtained using the circuit model of the SRR with the use of perturbation method under weak nonlinearity[23] and by considering only fundamental and second order terms. (cid:1869)(cid:3556)(cid:4666)2(cid:2033)(cid:4667)(cid:3404) The amplitude of the oscillations are given by (cid:3080)(cid:3023)(cid:3116)(cid:3118) (cid:3104)(cid:3119)(cid:3027)(cid:4666)(cid:2870)(cid:3104)(cid:4667)(cid:3027)(cid:4666)(cid:3104)(cid:4667)(cid:3118) ………………………………. (2) When the circuit resonates at (cid:2033)(cid:3404)(cid:2033)(cid:2868) and the impedance (cid:1852)(cid:4666)(cid:2033)(cid:4667) will be minimized resulting an increase of amplitude of the oscillations. In similar manner by minimizing (cid:1852)(cid:4666)2(cid:2033)(cid:4667) will further enhance the amplitude of oscillations and enhance the effect of second harmonic generation in the metamaterial. In order to maximize the second harmonic and hence maximize other higher order harmonics, DSRR structure was fabricated such that outer ring size is 40mm x 40mm and inner ring size is 20mm x 20mm. Both gaps in two rings have a gap size of 1mm and the ring width is 1mm. When the inner ring size changes, it effects the second harmonic signal as shown in fig 3b and we observe maximum second harmonic power when the outer ring frequency ((cid:1858)(cid:2869)) is twice as the inner ring frequency ((cid:1858)(cid:2870)) as illustrated in Fig. 3a. However when the condition (cid:1858)(cid:2870) = 2 (cid:1858)(cid:2869) does not satisfy, for 20mm and 22mm inner ring sizes we see Fig. 2b. In the case of satisfied condition of (cid:1858)(cid:2870) = 2 (cid:1858)(cid:2869) , we have an enhancement of 70 two peaks in second harmonic signal and the power of the signal gets weaker as shown in when we compare a single split ring with a double split ring as. Furthermore, it can be seen that the probe signal power also can be used to increase second harmonic signal power and throughout this experiment we have used 0dBm(1mW) as the probe signal power. ) B d ( r e w o p i c n o m r a h d n o c e S -40 -50 -60 -70 -80 -90 22mm inner ring 21mm inner ring 20mm inner ring 1.45 1.50 1.55 1.60 1.65 1.70 Frequency(GHz) Figure 3 (a) SHG by SRR and DSRR. (b) SHG by DSRR with different inner ring size. These varactor loaded nonlinear metamaterial unit cells can be considered as point sources of second harmonic signals which we demonstrate that these point sources has constructive and destructive interference when we keep them at specific positions. Y X Figure 4 Measured SHG by two DSRRs. Two unit cells of similar DSRR structures were arranged as shown in fig 3a and excited by a horn antenna from the direction indicated by x-axis and the second harmonic signal was collected from the direction of y using another horn antenna. Fig 3b shows the interference pattern when we change the relative distance in the y direction of the two unit cells. Second harmonic signal emits from the DSRR structure is 1668MHz, which corresponds to a wavelength of 17.98cm. In Fig3b solid line indicates the situation where the separation of the unit cells in x direction is 0cm and dashed line indicates that the separation of the unit cells in x direction is 36cm. in both cases relative change is in the y direction, and the dashed line represents the x distance between the unit cells is 36cm. The figure shows two different orientations with respect to varactor diode. One with the same orientation of the varactor diode in both unit cells and the other is 180 differences between the varactor diodes. Equation 3 gives the path difference of the two coherent second harmonic signals generated by 2 DSRR unit cells ∆(cid:1864)(cid:3404) (cid:4666)(cid:2870)(cid:3039)(cid:3300)(cid:2879) (cid:3039)(cid:3299) (cid:4667) (cid:3090)(cid:3255) Where ∆(cid:1864)the path difference of the two signals is generated by DSRR cells, (cid:1864)(cid:3052) is the distance between the unit cells in y direction, (cid:1864)(cid:3051) is the distance between the unit cells in x direction and (cid:2019)(cid:3007) is the wavelength of the fundamental frequency. 2π ……………………………………. (3) The fundamental wavelength of the signal is around 36cm and the second harmonic wavelength is 18cm. when separation of unit cells in x direction is either 0cm or 36cm, we can obtain the same interference pattern for both orientations of the diode as shown in Fig 3b (red solid line and dashed line indicates both cases). When the diodes are in the same orientation for λ/2 (9cm) we observe destructive interference and for λ (18cm) we observe a constructive interference. Similarly, for the opposite orientation of the diodes we observe constructive interference at a distance of λ/2 (9cm) and destructive interference at λ (18cm) which agrees with the Eq 3. Y X Figure 5 Array of DSRRs. Studies have shown that for non-symmetric structures, second harmonic signal originates from the side arms and they are in phase. The SRR and DSRR structures are also non- symmetric and its second harmonic field is produced in the side arms of the structures are radiates to the far field. By arranging unit cells in an array as shown in Fig 5a, can increase the total second harmonic signal with constructive interference of second harmonic waves from individual unit cells. Fig 5b shows interference of second harmonic signals from 6 unit cells of SRR and DSRR compared with their respective single unit cells. Distance between the unit cells in y direction (cid:1864)(cid:3052)= 9cm and distance between the unit cells in x direction (cid:1864)(cid:3051) = 36cm and the direction of the diodes of the two unit cells in the middle is kept opposite to the other unit cells. Array is constructed such that in x direction unit cell orientation is in the same direction and in y direction unit cell direction is in opposite direction. When we compare the second harmonic signal power generated by six DSRR unit cells to the power of 1 DSRR unit cell the enhancement factor is 8. It indicates that not only interference of the second harmonic generated by DSRR unit cells is effecting the enhancement in the array structure but also these unit cells act as mirrors when they are at resonance and form a cavity inside the array. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] M. Gorkunov and M. Lapine, "Tuning of a nonlinear metamaterial band gap by an external magnetic field," Physical Review B, vol. 70, Dec 2004. T. Driscoll, S. Palit, M. M. Qazilbash, M. Brehm, F. Keilmann, B. G. Chae, et al., "Dynamic tuning of an infrared hybrid‐metamaterial resonance using vanadium dioxide," Applied Physics Letters, vol. 93, Jul 14 2008. K. Aydin and E. Ozbay, "Capacitor‐loaded split ring resonators as tunable metamaterial components," Journal of Applied Physics, vol. 101, Jan 15 2007. J. Carbonell, V. E. Boria, and D. Lippens, "Nonlinear effects in split ring resonators loaded with heterostructure barrier varactors," Microwave and Optical Technology Letters, vol. 50, pp. 474‐ 479, Feb 2008. M. Lapine, M. Gorkunov, and K. H. Ringhofer, "Nonlinearity of a metamaterial arising from diode insertions into resonant conductive elements," Physical Review E, vol. 67, Jun 2003. Y. X. He, P. He, S. D. Yoon, P. V. Parimi, F. J. Rachford, V. G. Harris, et al., "Tunable negative index metamaterial using yttrium iron garnet," Journal of Magnetism and Magnetic Materials, vol. 313, pp. 187‐191, Jun 2007. Q. Zhao, B. Du, L. Kang, H. J. Zhao, Q. Xie, B. Li, et al., "Tunable negative permeability in an isotropic dielectric composite," Applied Physics Letters, vol. 92, Feb 4 2008. D. A. Powell, I. V. Shadrivov, Y. S. Kivshar, and M. V. Gorkunov, "Self‐tuning mechanisms of nonlinear split‐ring resonators," Applied Physics Letters, vol. 91, Oct 1 2007. B. N. Wang, J. F. Zhou, T. Koschny, and C. M. Soukoulis, "Nonlinear properties of split‐ring resonators," Optics Express, vol. 16, pp. 16058‐16063, Sep 29 2008. T. Driscoll, H. T. Kim, B. G. Chae, B. J. Kim, Y. W. Lee, N. M. Jokerst, et al., "Memory Metamaterials," Science, vol. 325, pp. 1518‐1521, Sep 18 2009. I. V. Shadrivov, A. A. Zharov, and Y. S. Kivshar, "Second‐harmonic generation in nonlinear left‐ handed metamaterials," Journal of the Optical Society of America B‐Optical Physics, vol. 23, pp. 529‐534, Mar 2006. T. Nakanishi, Y. Tamayama, and M. Kitano, "Efficient second harmonic generation in a metamaterial with two resonant modes coupled through two varactor diodes," Applied Physics Letters, vol. 100, p. 044103, 2012. C. Ciraci, E. Poutrina, M. Scalora, and D. R. Smith, "Origin of second‐harmonic generation enhancement in optical split‐ring resonators," Physical Review B, vol. 85, May 15 2012. A. B. Kozyrev, K. Hongjoon, and D. W. v. d. Weide, "Observation of the Parametric Amplification in Nonlinear Left‐Handed Transmission Line Metamaterials," in 2006 IEEE Antennas and Propagation Society International Symposium, 2006, pp. 4129‐4132. A. Degiron, J. J. Mock, and D. R. Smith, "Modulating and tuning the response of metamaterials at the unit cell level," Optics Express, vol. 15, pp. 1115‐1127, Feb 5 2007. A. Rose, D. Huang, and D. R. Smith, "Nonlinear Interference and Unidirectional Wave Mixing in Metamaterials," Physical Review Letters, vol. 110, Feb 5 2013. I. V. Shadrivov, S. K. Morrison, and Y. S. Kivshar, "Tunable split‐ring resonators for nonlinear negative‐index metamaterials," Optics Express, vol. 14, pp. 9344‐9349, Oct 2 2006. L. Kang, Q. Zhao, H. J. Zhao, and J. Zhou, "Ferrite‐based magnetically tunable left‐handed metamaterial composed of SRRs and wires," Optics Express, vol. 16, pp. 17269‐17275, Oct 27 2008. K. Aydin, I. Bulu, K. Guven, M. Kafesaki, C. M. Soukoulis, and E. Ozbay, "Investigation of magnetic resonances for different split‐ring resonator parameters and designs," New Journal of Physics, vol. 7, Aug 8 2005. P. Markos and C. M. Soukoulis, "Numerical studies of left‐handed materials and arrays of split ring resonators," Physical Review E, vol. 65, Mar 2002. E. Poutrina, D. Huang, and D. R. Smith, "Analysis of nonlinear electromagnetic metamaterials," New Journal of Physics, vol. 12, Sep 6 2010. R. W. Boyd, "Nonlinear Optics, 3rd Edition," Nonlinear Optics, 3rd Edition, pp. 1‐613, 2008. E. Poutrina, D. Huang, Y. Urzhumov, and D. R. Smith, "Nonlinear oscillator metamaterial model: numerical and experimental verification," Optics Express, vol. 19, pp. 8312‐8319, Apr 25 2011. [18] [19] [20] [21] [22] [23]
1811.09987	1	1811	2018-11-25T11:01:41	MOSFET Characterization and Modeling at Cryogenic Temperatures	[ "physics.app-ph" ]	Cryogenic CMOS technology (cryo-CMOS) offers a scalable solution for quantum device interface fabrication. Several previous works have studied the characterization of CMOS technology at cryogenic temperatures for various process nodes. However, CMOS characteristics for various width/length (W/L) ratios and under different bias conditions still require further research. In addition, no previous works have produced an integrated modeling process for cryo-CMOS technology. In this paper, the results of characterization of Semiconductor Manufacturing International Corporation (SMIC) 0.18 {\mu}m CMOS technology at cryogenic temperatures (varying from 300 K to 4.2 K) are presented. Measurements of thin- and thick-oxide NMOS and PMOS devices with different W/L ratios are taken under four distinct bias conditions and at different temperatures. The temperature-dependent parameters are revised and an advanced CMOS model is proposed based on BSIM3v3 at the liquid nitrogen temperature (LNT). The proposed model ensures precision at the LNT and is valid for use in an industrial tape-out process. The proposed method presents a calibration approach for BSIM3v3 that is available at different temperature intervals.	physics.app-ph	physics	MOSFET Characterization and Modeling at Cryogenic Temperatures Chao Luo,1,2 Zhen Li, 1,2 Teng-Teng Lu,1,2 Jun Xu,2Guo-Ping Guo1,a) 1 Key Laboratory of Quantum Information, University of Science and Technology of China, Hefei, Anhui 230026, China 2Department of Physics, University of Science and Technology of China, Hefei, Anhui 230026, China (Received XXXXX; accepted XXXXX; published online XXXXX) (Dates appearing here are provided by the Editorial Office) Abstract -- Cryogenic CMOS technology (cryo-CMOS) offers a scalable solution for quantum device interface fabrication. Several previous works have studied the characterization of CMOS technology at cryogenic temperatures for various process nodes. However, CMOS characteristics for various width/length (W/L) ratios and under different bias conditions still require further research. In addition, no previous works have produced an integrated modeling process for cryo-CMOS technology. In this paper, the results of characterization of Semiconductor Manufacturing International Corporation (SMIC) 0.18 μm CMOS technology at cryogenic temperatures (varying from 300 K to 4.2 K) are presented. Measurements of thin- and thick-oxide NMOS and PMOS devices with different W/L ratios are taken under four distinct bias conditions and at different temperatures. The temperature-dependent parameters are revised and an advanced CMOS model is proposed based on BSIM3v3 at the liquid nitrogen temperature (LNT). The proposed model ensures precision at the LNT and is valid for use in an industrial tape-out process. The proposed method presents a calibration approach for BSIM3v3 that is available at different temperature intervals. ⅠINTRODUCTION Quantum computers take advantage of two fundamental properties of quantum mechanics, i.e., superposition and quantum entanglement, and are based on qubits. A minimum of a few thousand and possibly millions of qubits are required for quantum computers and these qubits operate at deep cryogenic temperatures. High- precision control electronics are required on a massive scale for the manipulation and readout of single qubits in the so-called fault-tolerant mode1. A quantum computer comprises a quantum processor, in which the computations are performed, and a classical-type electronic 1 controller, which is implemented using room- temperature (RT) laboratory instrumentation. The wiring requirements between the cryogenic quantum processor and the RT read-out controller mean that the electronic interface required for control of the quantum processor can be both expensive and unreliable. As an alternative, an electronic interface operating at cryogenic temperatures has been proposed. Cryo-CMOS enables scalability in quantum processors that will allow improved system compactness and cryogenic operation, but a few challenges, including the power limitations of the refrigerators, the interconnections, the packaging and device modeling, have still to be faced2. Therefore, CMOS modeling is a priority subject also has good application prospects in fields for cryogenic studies. including space exploration, cryogenic electronics and quantum computing. However, BSIM3v3 is an industry-standard model that modeling at the liquid helium temperature (LHT) is valid over the temperature range from 230 K to will require further research. 430 K. However, the characteristics of metal- oxide-semiconductor field-effect transistors Ⅱ CRYOGENIC MEASUREMENTS (MOSFETs) change because of the freeze-out effect, which has stimulated a requirement for a Measurements of the CMOS transistors suitable SPICE model to be developed for use at were performed using both the thin (3.87 nm) and cryogenic temperatures. Cryogenic temperature thick oxide (11.9 nm) SMIC 0.18 μm operation of MOSFETs offers advantages that technologies and for a wide range of feature sizes, include improved turn-on capabilities, high gain, as shown in Table I. All the electrical latch-up immunity, improved driving capabilities, measurements were performed using the Agilent leakage current reduction, lower thermal noise, B1500A semiconductor device analyzer. higher integration density, and low power Measurements were taken at different cryogenic consumption. temperatures ranging from 298 K to 4.2 K. For the thin-oxide N-type MOS (NMOS) devices, CMOS technologies ranging from the 3 μm drain-source current (IDS) versus drain-source node down to 14 nm have been characterized over voltage (VDS) curves were measured under two the temperature range from 100 K down to 4.2 K substrate bias voltages (VBS=0 V and −1.8 V), 1,3-16. A brief review, which is illustrated in Fig. 1, while sweeping VDS from 0 to 1.8 V in 0.05 V compares the results of this work with those of the steps, and for values of VGS =0 V to 1.8 V in 0.3 previous publications. V steps; IDS versus gate-source voltage (VGS) curves were also measured for two drain-source In this paper, characterization of bias values (VDS=0.05 V and 1.8 V), while Semiconductor Manufacturing International sweeping VGS from 0 to 1.8 V in 0.05 V steps, and Corporation (SMIC) 0.18 μm CMOS transistors for values of VBS=0 V to −1.8 V in 0.3 V steps. is presented at various temperatures over the For the thick-oxide NMOS/P-type MOS (PMOS) range from 300 K to 4.2 K. In addition, an devices, the measurement bias conditions were advanced CMOS model based on BSIM3v3 is similar to those used for the thin-oxide devices, as proposed for use at the LNT. Measurements are indicated in the captions of Fig. 3, Fig. 4 and Fig. performed using the B1500A semiconductor 5. device analyzer and modeling is performed using BSIMProPlus. The temperature-dependent The drain current (ID) increases with parameters are revised and the model shows good decreasing temperature (Fig. 2) because of the agreement with the measurement results. The increased carrier mobility. An obvious kink is method provides a calibration approach for the shown for a large VDS. However, no kinks show BSIM3v3 parameters that is available for a up around 77 K for the maximum operating value variety of different process nodes and of VDS. The threshold voltage (Vth) increases at temperature intervals. A comparison with cryogenic temperatures for both NMOS and previous works on integrated modeling under PMOS devices, as shown in Fig. 6(a). As shown various W/L and bias conditions is also provided. in Fig. 4(c) and (d), the threshold slope (SS) at The model presented is well matched with the LHT for both the thin- and thick-oxide NMOS industrial process. The research described here devices is steeper with respect to the RT. The SS 2 of the thick-oxide device is 1.51 times steeper at where 𝑉𝐷𝑆∗ is the critical value for kink the LNT and 3.43 times steeper at the LHT with effect. respect to RT. In addition, the transconductance (Gm) tends to increase in proportion to decreasing No kink occurs around 77K for the T (Fig. 6(c)). maximum operating value of VDS. The deviation between BSIM3v3 model and characteristics at Table I SUMMARY OF CHARACTERIZED LNT is mainly ascribed to the freeze-out effect DEVICES of source-drain parasitic resistance19. In CMOS process node, lightly doped drain (LDD) Device Size Thin-oxide Thick-oxide structure is employed to reduce the hot carrier NMOS/PMOS NMOS/PMOS effect and short channel effect. The ionization of W(μm) L(μm) W(μm) L(μm) impurities will decrease as temperature SMIC 0.18 μm 100 0.18 100 0.6/0.5 decreases, especially in lightly doping regions 10 10 10 10 10 10 0.6 0.2 0.18 0.16 10 10 10 10 10 near source and drain. This results in a large 0.65/0.55 series resistor in the LDD region at low 0.5/0.45 temperatures. When the source-drain voltage 0.2 becomes very high, the ionization of impurities 0.3 0.6/0.5 will be activated to restrain the freeze-out effect under strong electric field. In summary, Ⅲ DISCUSSION ON CRYOGENIC MOSFET characteristics are manifested as BEHAVIORS suppressed leakage current in deep linear regions and normal-state at high source-drain voltages. It The most noticeable irregularity in the IDS- should be pointed out that the performances of VDS characteristics (Fig. 2(b)) is the kink that devices with higher doping are not typically occurs in the near mid-VDS range at 4.2 K. This degraded by freeze-out, as was previously the phenomenon is ascribed to the self-polarization of case for lightly doped devices20. the bulk at cryogenic temperatures17. The high electric field results in the presence of impact- When the temperature decreases, Vth ionized carriers along the channel. The self- increases as expected, as shown in Fig. 6(a). The polarization leads to a significant bulk current that mobility is strongly enhanced because of the raises the internal body potential and causes a reduction in carrier scattering due to lattice reduction in the apparent Vth above a certain VDS vibrations21,22. The effective mobility (μeff) is value. Therefore, a jump occurs in ID. This dependent on the threshold voltage and the gate irregular behavior can be modeled using a double and substrate voltages. At a low gate voltage, the Vth model18, where ID is given by the following Coulomb scattering is dominant and μeff formula. 𝐼𝐷 = 𝐾𝑃 𝑊 𝐿 (𝑉𝐺𝑆 − 𝑉𝑇)2 (1) increases with increasing gate voltage, which then results in an increase in the transconductance. At higher gate voltages, strong electric field effects play a major role, leading to a reduction in the mobility and a corresponding reduction in the transconductance value. Figure 6(b) illustrates this behavior for a with 𝑉𝑇 = 𝑉𝑇𝑀𝐴𝑋 𝑓𝑜𝑟 𝑉𝐷𝑆 < 𝑉𝐷𝑆∗ and thick-oxide NMOS device at various temperatures. The SS is an important parameter 𝑉𝑇 = 𝑉𝑇𝑀𝐼𝑁 𝑓𝑜𝑟 𝑉𝐷𝑆 > 𝑉𝐷𝑆∗. (2) 3 for MOSFETs operating as digital logic switches The basic idea consisted of construction of a in the sub-threshold region. SS is given by function that can eliminate the effects of the 𝑆𝑆(𝑇) = [ 𝜕log (𝐼𝐷) 𝜕𝑉𝐺𝑆 −1 ] = ln(10) 𝑛𝑘𝑇 𝑞 (3) mobility's gate voltage dependence. In this work, the extraction procedure is performed using BSIMProPlus. Before the where k is Boltzmann constant, q the temperature-dependent parameter extraction electron charge, T the Kelvin temperature and n process, the temperature of the simulator must be is the sub-threshold slope factor. Cryogenic performance at RT, LNT and LHT are listed and compared in Table II. The parameters that are known to affect the performance of analog circuits strongly all show improvements. An increase in Gm will result in a wider bandwidth being available for the same power budget. Additionally, the concomitant increase in Gds leads to a reduction of the set to 77 K to match the SS. The parameters with the greatest influence on ID are extracted first, i.e., VT (vt0, K1, K2, dvt0 for BSIM) and μ (μ0, μa, μb, μc). Additional modeling was performed by taking the drain/source charge sharing behavior (dvt0, dvt1, dvt2), the carrier drift velocity (vsat) and the bulk charge effect (a0, ags, keta) into consideration. The remaining parameters would lead to further enhancement of the fitting but have only a minor influence on the device transistor's intrinsic DC gain. Any increase in SS characteristics. is beneficial for the digital logic performance, leading to increased speed in the sub-threshold The root-mean-square (RMS) error is then region. introduced to estimate the deviation between the results from the measurements and the Table II COMPARISON OF PERFORMANCE AT RT, LNT AND LHT simulations.The RMS is given by Technology NMOS,W/L=10μm/10μm Temperature 4.2K 77K 298K Vth [V] 1.164 1.153 0.799 SS [mV/dec] 26.14 59.31 89.86 RMSerr = √ 1 𝑁 ∑ ( 𝑛 𝑖=1 𝐼𝑚𝑒𝑎𝑠𝑖−𝐼𝑠𝑖𝑚𝑢𝑖 𝐼𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 ) 2 × 100 (4) Gm(VGS=1.5V)[mS] 0.114 0.071 0.011 where N is the number of data, Imeasi is Gds(VDS=1.5V)[mS] 0.623 0.600 0.225 measured data and Isimui is simulated data. The Ⅳ MODELING value of the threshold current Ithreshold can be set appropriately to obtain meaningful results. In this case, Ithreshold has been set to the maximum Several prior works modified their developed measured value according to BSIMProPlus. models by adjusting the temperature-dependent parameters based on the results from the LNT Table III RMS ERROR of MEDOL WITH measurements and augmented the sub-circuit DEFAULT AND REVISED PARAMETERS. model using additional electrical components23. A cryogenic modeling methodology is presented (Thick-oxide) NMOS PMOS here that allows temperature-dependent W/L=10μm/10μm default revised default revised parameters to be adjusted by inserting a ID(VGS)(VDS=0.1V) 35.31% 1.53% 20.07% 1.63% correction coefficient into the BSIM equations ID(VDS)(VBS=0V) 33.29% 1.57% 20.97% 1.36% and then solving the modified equations24. Another parameter extraction method for use at low temperatures was introduced elsewhere25. 4 ⅤSIMULATION VERSUS MEASUREMENT Good agreement with the DC measurements approach using temperature-dependent was achieved for devices operating at the LNT. parameters that can be extended to any Measurements and simulations of the IDS-VDS and temperature region. Improvements in the IDS-VGS curves for the NMOS (Fig. 7(a) -- (f)) and proposed model will allow precise description of PMOS (Fig. 7(g) -- (h)) devices are plotted at the MOSFETs to aid in the design of more complex LNT. The RMS errors were calculated for the cryogenic circuits and systems. Further research measured data with respect to the simulations into modeling at the LHT will continue as part of using both the default and revised parameters, as our future work. summarized in Table III. A significant optimization of the BSIM3v3 model has been Ⅶ ACKNOWLEDGMENT achieved. Ⅵ CONCLUSION The authors would like to thank SMIC for devices fabrication and software support. This work was supported by the National Key A study of the performance of SMIC 0.18 Research and Development Program of China μm/1.8 V CMOS technology at cryogenic (Grant No.2016YFA0301700), the National temperatures has been presented in this work. An Natural Science Foundation of China (Grants No. advanced model based on BSIM3v3 has been 11625419), the Anhui initiative in Quantum proposed to optimize the deviations between the information Technologies (Grants No. measurement results and the results of AHY080000) and this work was partially carried simulations using the default parameters. The out at the USTC Center for Micro and Nanoscale proposed method presents an optimization Research and Fabrication. REFERENCES 1. E.Charbon, F.Sebastiano, "Cryo-CMOS for Quantum Computing," 978-7-5090-3902- 10. A.Akturk et al., "Compact and distributed modeling of cryogenic bulk MOSFET operation," 9/16/$31.00, 2016, IEEE. IEEE Transactions on Electron Devices, vol. 57, no. 6, pp. 1334 -- 1342, 2010. 2. B.patra, E.Charbon, "Cryo-CMOS Circuits and Systems for Quantum Computing Applications," 11. S. R. Ekanayake et al., "Characterization of SOS-CMOS FETs at low temperatures for the IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL.53, NO.1, JANUARY 2018 design of integrated circuits for quantum bit control and readout," IEEE Transactions on Electron 3. M. Shin et al., "Low temperature characterization of 14nm FDSOI CMOS devices," in WOLTE, Devices, vol. 57, no. 2, pp.539 -- 547, 2010. 2014, pp. 29 -- 32. 12. G. S. Fonseca et al., "Extraction of static parameters to extend the EKV model to cryogenic 4. A. H. Coskun and J. C. Bardin, "Cryogenic small-signal and noise performance of 32nm SOI temperatures," in SPIE Defense+ Security. SPIE, 2016, pp. 98 192B -- 98 192B. CMOS," in IMS. IEEE, 2014, pp. 1 -- 4. 13. Y. Creten et al., "A cryogenic ADC operating down to 4.2 K," in ISSCC. IEEE, 2007, pp. 468 -- 5. L.Song, F.Sebastiano, E.Charbon, "Nanometer CMOS characterization and compact modeling 616. at deep-cryogenic temperatures,"DOI:10.1109/ESSDERC.2017.8066591,Sep,2017. 14. H. Hanamura et al., "Operation of bulk CMOS devices at very low temperatures," IEEE Journal 6. Z. Chen et al., "Temperature dependences of threshold voltage and drain-induced barrier of Solid-state Circuits, vol. 21, no. 3, pp. 484 -- 490, 1986. lowering in 60nm gate length MOS transistors," Microelectronics Reliability, vol. 54, no. 6, pp. 15. E. Simoen, B. Dierickx, and C. L. Claeys, "Low-frequency noise behavior of Si NMOSTs 1109 -- 1114, 2014. stressed at 4.2K," IEEE Transactions on Electron Devices, vol. 40, no. 7, pp. 1296 -- 1299, 1993. 7. A. Siligaris et al., "High-frequency and noise performances of 65-nm mosfet at liquid nitrogen 16. F. Balestra, L. Audaire, and C. Lucas, "Influence of substrate freeze-out on the characteristics temperature," IEEE Transactions on Electron Devices, vol. 53, no. 8, pp. 1902 -- 1908, 2006. of mos transistors at very low temperatures,"Solid-state electronics, vol. 30, no. 3, pp. 321 -- 327, 8. G. D. Geronimo et al., "Front-end ASIC for a liquid argon TPC," in NSS. IEEE, 2010, pp. 1658 -- 1987. 1666. 17. G.Ghibaudo and F.Balestra, "Low temperature characterization of silicon CMOS devices", in 9. Z. Zhu et al., "Design applications of compact MOSFET model for extended temperature range Proc.20th.Conf.Microelectronics, Sep.1995, pp.613-622. (60 -- 400K)," Electronics letters, vol. 47,no. 2, pp. 141 -- 142, 2011. 5 18. O.Claeye, "Design of read-out electronics in a standard CMOS process for operation below 22. C.Huang, S.Gildenblat,. "Measurements and modeling of the n-channel MOSFET inversion 30K", in Proc.4th.European Workshop on Low Temperature Electronics(WOLTE-4), ESTEC, layer mobility and device characteristics in the temperature range 60-300K", IEEE Trans Electron Noordwijk, The Netherlands, Jun.2000.. Dev 1999 ;46 :1912-3. 19. Mazure C., et al., Impact of LDD spacer reduction on MOSFET performance for sub-μm 23. Hongliang Zhao, Xinghui Liu, "Modeling of a standard 0.35μm CMOS technology operating gate/space pitches, Electron Devices Meeting, 1992: 893-896. from 77K to 300K", Cryoagenics 59 (2014) 49-59. 20.A.Akturk, et al, "Device Modeling at Cryogenic Temperatures : Effects of Incomplete 24. A.Kabaoglu, M.B.Yelten, "A Cryogenic Modeling Methodology of MOSFET I-V Ionization", IEEE Transactions on Electron Devices, Vol.54, No.11, Nov.2007,pp.2984-2990. Characteristics in BSIM3", 978-1-5090-5052-9/17/$31.00,2017,IEEE. 21. A.Emrani, F.Balestra, G.Ghibaudo, "On the understanding of electron and hole mobiliyt models 25. G.Ghibaudo and F.BAlestra,"A Method for MOSFET Parameter Extraction at Very Low from room to liquid helium temperatures", Solid State Electron 1994 ;37 :1723-30. Temperature", Solid-State Electronics Vol.32, No.3, pp.221-223,1989 Fig. 1 A brief review of characterization of CMOS transistors at cryogenic temperatures. Prior works are indicated with black dot and devices presented in this paper with red dots. Fig. 2 IDS-VGS and IDS-VDS of thin-oxide NMOS at different temperatures varying from 298K to 4.2K. W/L in μm. (a) W/L =10/10. (b) W/L =10/0.2. 6 Fig. 3 IDS-VGS and IDS-VDS of NMOS in thin- (a,b) and thick-oxide (e,f) , and thick-oxide PMOS (g,h) at LNT (dashed lines). Gm-VGS (c) and Gds-VDS (d) of thin-oxide NMOS. The solid lines are simulations of BSIM with non-revised parameters. W/L = 10μm /10μm for all deveices. (a,c) VDS=0.05V, VBS= 0 → -1.8V; (b,d) VBS=0V, VGS= 0 → 1.8V; (e) VDS=0.10V, VBS= 0 → -4V; (f) VBS=0V, VGS= 0 → 5V; (g) VDS=-0.10V, VBS= 0 → 4V; (h) VBS=0V, VGS= 0 → -5V; 7 Fig. 4 IDS-VDS and IDS-VGS of NMOS in thin- (a,c) and thick-oxide (b,d) at RT (dashed lines )and LHT (solid lines). W/L in μm. (a) W/L=10/10, VBS=0V, VGS= 0 → 1.8V; (b) W/L=10/10, VBS=0V, VGS= 0 → 5V; (c) W/L=10/10, VDS=0.05V, VBS= 0 → -1.8V; (d) W/L=10/10, VDS=0.10V, VBS= 0 → -4V. Fig. 5 IDS-VGS and IDS-VDS measurements of thick-oxide PMOS at RT(dashed lines) and LHT (solid lines). (a) VDS=-0.10V, VBS= 0 → 4V; (b) VBS=0V, VGS= 0 → -5V; Fig. 6 (a) Variations and fitting curves of extracted threshold voltage at zero substrate bias for different 8 temperatures; (b) Variations of the trans-conductance with gate voltage on a 10μm /10μm thick-oxide NMOS for different temperatures; (c) SS versus temperature of thick-oxide NMOS, W/L=10μm /10μm. Fig. 7 Revised models (solid lines) and measurements (dashed lines) at LNT. W/L = 10μm /10μm for all deveices. (a,c) VDS=0.05V, VBS= 0 → -1.8V; (b,d) VBS=0V, VGS= 0 → 1.8V; (e) VDS=0.10V, VBS= 0 → -4V; (f) VBS=0V, VGS= 0 → 5V; (g) VDS=-0.10V, VBS= 0 → 4V; (h) VBS=0V, VGS= 0 → -5V; 9
1711.04193	1	1711	2017-11-11T20:44:32	Elastic metasurfaces for splitting SV- and P-waves in elastic solids	[ "physics.app-ph" ]	Although recent advances have made it possible to manipulate electromagnetic and acoustic wavefronts with sub-wavelength metasurface slabs, the design of elastodynamic counterparts remains challenging. We introduce a novel but simple design approach to control SV-waves in elastic solids. The proposed metasurface can be fabricated by cutting an array of aligned parallel cracks in a solid such that the materials between the cracks act as plate-like waveguides in the background medium. The plate array is capable of modulating the phase change of SV-wave while keeping the phase of P-wave unchanged. An analytical model for SV-wave incidence is established to calculate the transmission coefficient and the transmitted phase through the plate-like waveguide explicitly. A complete $2\pi$ range of phase delay is achieved by selecting different thicknesses for the plates. An elastic metasurface for splitting SV- and P-waves is designed and demonstrated using full wave finite element (FEM) simulations. Two metasurfaces for focusing plane and cylindrical SV-waves are also presented.	physics.app-ph	physics	Elastic metasurfaces for splitting SV- and P-waves in elastic solids Xiaoshi Su,1, a) Zhaocheng Lu,1 and Andrew N. Norris1 Mechanical and Aerospace Engineering, Rutgers University, Piscataway, NJ 08854 (Dated: 11 November 2018) Although recent advances have made it possible to manipulate electromagnetic and acoustic wavefronts with sub-wavelength metasurface slabs, the design of elastodynamic counterparts remains challenging. We intro- duce a novel but simple design approach to control SV-waves in elastic solids. The proposed metasurface can be fabricated by cutting an array of aligned parallel cracks in a solid such that the materials between the cracks act as plate-like waveguides in the background medium. The plate array is capable of modulating the phase change of SV-wave while keeping the phase of P-wave unchanged. An analytical model for SV-wave incidence is established to calculate the transmission coefficient and the transmitted phase through the plate- like waveguide explicitly. A complete 2π range of phase delay is achieved by selecting different thicknesses for the plates. An elastic metasurface for splitting SV- and P-waves is designed and demonstrated using full wave finite element (FEM) simulations. Two metasurfaces for focusing plane and cylindrical SV-waves are also presented. I. INTRODUCTION Achieving full control of wave propagation with ultra- thin material slabs is of particular interest in engineering applications. In the past decade, the emerging area of metasurface research has made it possible to manipulate optical and electromagnetic waves in an almost arbitrary way by tuning the phase gradient at the sub-wavelength scale.1 -- 5 This concept has also found applications in acoustic designs such as focal lenses,6 -- 8 anomalous re- flection and refraction,9 -- 14, and generation of acoustic orbital angular momentum.15,16 Elastic metasurfaces17,18 are relatively unexplored; they present specific challenges due to the mode conversion at the material interface which makes the phase modulation more complicated. Recently, Zhu and Semperlotti 17 designed and experi- mentally demonstrated a few metasurfaces for controlling mode converted and unconverted lamb waves in plates. However, their design approach involves mode conver- sion, as a result, the transmitted field contains unwanted wave types. We are interested in controlling different types of bulk waves individually without introducing oth- ers. We propose a new metasurface design to split SV- and P-waves in elastic solids into different propagation directions without involving mode conversion. Achieving full control of these types of waves may have applications in ultrasonics and nondestructive evaluations. The meta- surfaces are designed by introducing an array of aligned parallel cracks in a bulk elastic medium. The materi- als separated by these cracks act as plate-like waveguides connecting two elastic half-spaces. For a normally in- cident SV-wave, the transverse vibration couples to the flexural waves in each plate without mode conversion. Then each of the plate-like waveguide serves as a phase modulator to achieve certain phase gradient for the trans- mitted wavefronts. The main idea is that the phase speed of the SV-wave in the background material only depends a)Electronic mail: [email protected] on the material properties, while the flexural wave speed in the plate is sensitive to the thickness which makes it possible to achieve desired phase shift for the metasur- face design. For instance, we can design a metasurface to change the propagation direction of the transmitted SV-wave by tuning the phase gradient according to the generalized Snell's law1 (sin θt − sin θi)kT = dφ/dy, (1) where kT denotes the wavenumber of the SV-wave in the solid; φ is the phase of the transmitted wave; θi and θt are the incident and transmitted angle, respectively. The mechanism for normally incident P-wave is different in that the longitudinal wave speeds in the plate array are the same. Due to this feature, there is no phase difference in the transmitted wavefronts and therefore the P-wave travels along the incident direction. In the applications proposed in this paper, the phase gradients are small so that the phase modulation is more critical than the amplitude modulation. In order to pre- dict the transmitted phase accurately, we first establish an analytic model to calculate the transmission coeffi- cient of the unit cell. Then we take advantage of this model to select the thicknesses of the plates for the meta- surface designs. The same transmission problem has been considered by Su and Norris 19, but the model was based on a thin plate assumption which is only valid for a low frequency range, i.e. kh ≪ 1. In this paper, we im- prove the analytic model using Mindlin plate theory20 which introduces two high frequency correction factors and therefore works for thick plates at higher frequency range. The explicit expressions for the transmission co- efficient and the transmitted phase are obtained. Note that the analytic model for P-wave incidence in Ref. 19 works very well, we will use this model in the discussion of the transmission properties for P-wave incidence. Sev- eral metasurface devices for different purposes, includ- ing mode splitting of SV- and P-waves and focusing of plane and cylindrical SV-waves, are designed using the analytic model and demonstrated using full wave FEM simulations. The paper is organized as follows. The unit cell design and the transmission properties of the metasurface are introduced and discussed in Sec. II. The phase modula- tion by thickness variations of the plates is also proposed in Sec. II. Several metasurfaces are designed and demon- strated using full wave FEM simulations in Sec. III. Sec- tion IV concludes the paper. II. UNIT CELL DESIGN AND TRANSMISSION PROPERTIES A. Description of the transmission problem We first derive the transmission coefficient for a nor- mally incident SV-wave propagating through an array of uniform parallel plate separated by equally spaced rect- angular cracks. The Young's modulus of the solid is de- noted by E, shear modulus by µ, Poisson's ratio by ν and density by ρ. The configuration is shown in Fig. 1. Each individual plate has thickness h and length L; the (a) (b) FIG. 1. 2D schematic view; (a) shows two half-spaces con- nected by a uniform plate array separated by parallel cracks (white slits). The parallel black arrows indicate the propa- gation direction of the normally incident SV-wave; the red arrows indicate the direction of particle motion; (b) shows a unit cell of the slab. crack has width a and length L. The transmission prob- lem can be understood as follows: the normally incident plane SV-wave in a half-space impinges on an array of aligned parallel plates and couples with the flexural waves in plates; flexural waves then transmit through the plate array into another half-space producing the transmitted SV-wave. The critical physics underlying the tranmission is that the phase speed of the SV-wave only depends on the material properties, while the flexural wave speed is 2 sensitive to the thickness of each plate. This combina- tion of wave properties makes it possible to achieve spe- cific phase delay through thickness variations. In order to quantify the metasurface design, it is useful to de- rive explicit expressions for the transmission coefficient and transmitted phase. Due to the periodicity in the y- direction, the whole transmission problem is equivalent to that outlined in the dashed box in Fig. 1(b). A de- tailed description regarding the notations in this figure will be given in Sec. II C. The derivation procedure of the transmission and re- flection coefficients for SV-wave incidence is similar to that in Ref. 19 in which Kirchhoff plate theory was used under thin plate assumptions. However, the Kirchhoff theory does not hold at the high frequency range even for thin plates. In this section, we will develop a more sophisticated model to better predict the transmission coefficient at the high frequency range. Mei and Mace 21 studied wave reflection and transmission in beams with discontinuities using Timoshenko beam theory. We are dealing with transverse wave reflection and transmission at the junction of bulk material and an array of plates where the governing equations are different. Here we use the Mindlin plate theory20 with two high frequency correction factors and consider similar boundary condi- tions to establish the analytic model and calculate the transmission and reflection coefficients accurately. More detailed description and derivation will be given in the following sections. B. Governing equations in the plate-like waveguides The plate-like waveguides connecting the two half- spaces act as phase modulators. We assume the plate array is uniform and due to the periodicity in the y- direction only consider the transmission problem in one plate, see Fig. 1(b). According to the Mindlin plate the- ory, the governing equations of the transverse waves in the absence of external force are ∂2w κµ(cid:16) ∂ψ ∂x2(cid:17) + ρ ∂x − ∂x2 + κµA(cid:16) ∂w ∂x − ψ(cid:17) − λρI ∂2ψ ∂2w ∂t2 = 0, ∂2ψ ∂t2 = 0, (2) EI where w(x, t) is the displacement in the y-direction, ψ(x, t) is the bending angle, I = bh3/12 is the area mo- ment of inertia; A = bh is the cross section area of the plate; κ and λ are the shear and inertia correction fac- tor, respectively. In the absence of external force, the free wave propagation solutions are w(x, t) = W ei(kx−ωt) and ψ(x, t) = Ψei(kx−ωt). Plugging the solutions into Eq. (2) leads to (cid:20)ρω2 − κµk2 iκµAk λρIω2 − κµA − EIk2(cid:21)(cid:18)W −iκµk Ψ(cid:19) = 0. (3) The equation for the wavenumber k is therefore k4 − (λ2 P + k2 T κ )k2 + λk2 T k2 κ − k4 P F = 0, (4) where kT = ωpρ/µ, kP = ωpρ(1 − ν2)/E and kF = (cid:0)12ω2ρ(1 − ν2)/Eh2(cid:1)1/4 are the wavenumbers of trans- verse, longitudinal and flexural waves, respectively. From Eq. (4), the wavenumbers in the plate are 2 k1 = ±(cid:26) 1 k2 = ±(cid:26) 1 2 (λk2 P + (λk2 P + k2 T κ k2 T κ 4 ) +r 1 ) −r 1 4 (λk2 P − (λk2 P − k2 T κ k2 T κ )2 + k4 )2 + k4 F(cid:27)1/2 F(cid:27)1/2 , . (5) Note that within the frequency range of interest (ω < p12µκ/λρh2) k1 is real and corresponds to a propagat- ing wave, while k2 is imaginary and corresponds to an evanescent wave. C. Transmission and reflection coefficients Considering the transmission problem for a normally incident SV-wave, in Fig. 1(b), we assume the amplitude of the incident SV-wave as 1, reflected SV-wave as R, transmitted SV-wave as T , and flexural waves in the plate array as Aj, Bj, j = 1, 2 so that eikT x + Re−ikT x, A1eik1x + B1e−ik1x +A2eik2x + B2e−ik2x, T eikT x, x < − L 2 , x < L 2 , x > L 2 , (6) w =  with time harmonic dependence e−iωt understood. The relations between the shear and inertia correction fac- tors and the slope in the plate can be obtained from the equations of motion as20 Ψ = ikjβjW, βj = 1 − k2 T κk2 j , j = 1 or 2. (7) The deflection angles in different parts of the structure based on Eq. (6) are ikT(cid:0)eikT x − Re−ikT x(cid:1), ik1β1(cid:0)A1eik1x − B1e−ik1x(cid:1) +ik2β2(cid:0)A2eik2x − B2e−ik2x(cid:1), ikT T eikT x, x < − L 2 , x < L 2 , x > L 2 , (8) ψ =  Equations (6) and (8) involve six unknowns where the transmitted amplitude T and the reflected amplitude R 3 are of particular interest. To solve for the six unknown parameters, we need six boundary conditions, i.e., dis- placement, deflection angle and shear force continuity at both ends of the plate. The displacement and deflection angle continuity can be easily established using Eqs. (6) and (8). The average shear forces along the z-direction in the half-space and the plate are Q = µh′(∂w/∂x) and Q = κµh(∂w/∂x− ψ), respectively. Using the six bound- ary conditions we can establish a 6 × 6 system with six unknowns to solve for the transmission and reflection co- efficients. This procedure involves too many undesired long equations, so as an alternative we split the solutions into symmetric and antisymmetric modes, which reduces the problem to two 3 × 3 systems. spaces and the plate are rewritten as For the symmetric mode, the displacements in the half- 1 2 (eikT x + RSe−ikT x), CS1 cos k1x + CS2 cos k2x, 1 2 (e−ikT x + RSeikT x), x < − L 2 , x < L 2 , x > L 2 , (9) wS =  and the deflection angles are ikT ψS =  2 (eikT x − RSe−ikT x), −k1β1CS1 sin k1x − k2β2CS2 sin k2x, − ikT 2 (e−ikT x − RSeikT x), x < − L 2 , x < L 2 , x > L 2 , (10) where CS1 and CS2 denote the amplitude of symmetric modes in the plate, RS corresponds to the amplitude of the reflected wave in the half-spaces. Similarly, consid- ering the antisymmetric mode we can write the displace- ments and deflection angles as 1 2 (eikT x + RAe−ikT x), CA1 sin k1x + CA2 sin k2x, − 1 2 (e−ikT x + RAeikT x), x < − L 2 , x < L 2 , x > L 2 , (11) wA =  and ikT ψA =  2 (eikT x − RAe−ikT x), k1β1CA1 cos k1x + k2β2CA2 cos k2x, ikT 2 (e−ikT x − RAeikT x), x < − L 2 , x < L 2 , x > L 2 , (12) where CA1 and CA2 denote the amplitude of antisymmet- ric modes in the plate, RA corresponds to the amplitude of the reflected wave in the half-spaces. The boundary conditions for each problem now reduce to displacement, deflection angle and shear force conti- nuity at one end of the plate. Applying the boundary conditions leads to cos(k2L/2) cos(k1L/2) k2β2 sin(k2L/2) k1β1 sin(k1L/2)   κhk1(1 − β1) sin(k1L/2) κhk2(1 − β2) sin(k2L/2) 1   κhk1(1 − β1) cos(k1L/2) κhk2(1 − β2) cos(k2L/2) 1 − sin(k2L/2) k2β2 cos(k2L/2) − sin(k1L/2) k1β1 cos(k1L/2) − 1 2 z 1 2 ikT z 2 ikT h′z  2 ikT h′z  − 1 2 z 1 2 ikT z     CS1 CS2 RS CA1 CA2 RA   =    =  1 1 1 2 z−1 2 ikT z−1 2 ikT h′z−1  , 2 ikT h′z−1  , 2 ikT z−1 2 z−1 1 1 1 4 (13) (14) where z = eikT L/2. Solving Eqs. (13) and (14) yields the reflection coefficients for the symmetric and antisymmet- ric modes as RS = RA = α2 k1 α2 k1 α2 k1 α2 k1 cot k1L cot k1L cot k1L cot k1L k2 k2 2 + α1 2 + α1 2 − α1 2 − α1 k2 k2 cot k2L cot k2L cot k2L cot k2L kT kT 2 + i (β1−β2) 2 − i (β1−β2) 2 + i (β1−β2) 2 − i (β1−β2) kT kT e−ikT L, e−ikT L, (15) 2 T where αj = βj − 1 + βjh′/(κh) for j = 1, 2. The trans- mission and reflection coefficients for the full problem are then T = R = 1 2 1 2 (RS − RA), (RS + RA). (16) The main result here is the transmission coefficient T which not only shows the amplitude but also contains information about the transmitted phase. D. Numerical validation and phase modulation We now show that the theoretical model accurately predicts the transmission coefficient using a numerical ex- ample. Consider an array of uniform plates in aluminum as described in Fig. 1(a). The material properties are Young's modulus E = 70 GPa, Poisson's ratio ν = 0.33 and density ρ = 2700 kg/m3. All the plates have length L = 5 cm and thickness h = 0.5 cm and are separated by cracks of constant width a = 1 mm. A plane SV-wave is normally incident from the left side of the metasurface. Figure 2 compares the present model and the theoretical model developed in Ref. 19 to FEM simulation results using COMSOL Multiphysics. The Bloch-Floquet peri- odic condition was prescribed on the boundaries of the unit cell to mimic an infinite metasurface slab. The red curve in Fig. 2 is calculated using the displacements ex- tracted from the FEM simulation results. We can clearly see that the transmission curve calculated using the an- alytic model agrees well with the FEM results at higher frequencies. It is also remarkable that the present model can accurately predict the total transmission frequencies, which correspond to the flexural resonances. This charac- teristic provides strong evidence that the predicted phase 1 0.8 0.6 0.4 0.2 0 0 Present model FEM simulation results Analytic model in Ref. 19 1 2 3 4 5 6 Frequency [Hz] 7 104 FIG. 2. Comparison between analytical models and FEM sim- ulation results. The blue curve corresponds to the frequency dependence of the transmission coefficient calculated using the model developed by Su and Norris 19 ; the black curve cor- responds to the transmission calculated using Eq. (16) in this paper; the red circles represent the FEM simulation results. change is close to the simulation results. Though the an- alytic model shows certain mismatch of amplitudes with the simulation results, the present model is still a useful tool since phase modulation is more crucial in metasur- face design. This issue could potentially be addressed by using more complicated boundary conditions,22 but is beyond the scope of this paper. The explicit expression for the transmitted phase can be easily extracted from Eq. (16) as (17) φ = tan−1(cid:0)imag(T )/real(T )(cid:1), −π < φ < π. According to the generalized Snell's law, i.e. Eq. (1), the metasurface design requires the transmitted wave- fronts to cover the full 2π span. This can be easily sat- isfied using our design elements. The objective of this paper is to achieve full control of SV-wave using meta- surface slabs with sub-wavelength thickness. In all the designs presented in Sec. III, the length of each plate is chosen as L = 5 cm; the width of the gaps between plates is constant a = 1 mm; the operation frequencies for all the metasurfaces are identical 60 kHz, at which the wavelength of the transverse wave in the bulk material is larger than the slab thickness, i.e. L < λT . Though it is possible to design for a lower frequency range, 60 kHz is selected to maintain relatively high transmission as shown in Fig. 2. Consider a metasurface slab com- prised of plates with different thickness in an aluminum background. The transmitted phases at 60 kHz corre- sponding to different thicknesses of plates are calculated using Eq. (17) and plotted in Fig. 3. It is clear that the can now design metasurface devices for various purposes such as mode splitting of SV- and P-waves and focusing of plane and cylindrical SV-waves. 5 III. APPLICATIONS IN METASURFACE DESIGN A. Metasurface for splitting SV- and P-waves The material used in all the metasurface designs is alu- minum with Young's modulus E = 70 GPa, Poisson's ratio ν = 0.33 and density ρ = 2700 kg/m3. In the design for anomalous refraction of normally incident SV- wave, we choose a linear phase gradient dφ/dy = 40π/√3 rad/m. This phase gradient results in a transmitted an- gle of θt = 30◦ according to Eq. (1). The schematic view of the metasurface is illustrated in Fig. 4(a). The width SV-wave ) ( 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 P-wave (a) 2 3 4 5 6 7 h [mm] 8 9 10 11 FIG. 3. Transmitted phase at 60 kHz through plates of differ- ent thicknesses. The solid line corresponds to the transmit- ted phase for a normally incident SV-wave; the dashed line is computed using the model developed by Su and Norris 19 and corresponds to the transmitted phase for P-wave incidence. phase shifts through the internal plates, with thickness varying from 2.4 mm to 10.4 mm, cover a range of 2π for SV-wave incidence. Related to a point made earlier, all the plates are thinner than 11 mm in the metasurface designs presented in this paper such that the minimum cutoff frequency for the k2 mode is above 141.9 kHz. The dashed line in Fig. 3 is computed using Eq. (13) in Ref. 19 and corresponds to the transmitted phase of a normally incident P-wave. It indicates nearly identi- cal phase changes for different plate thickness. This can be understood from the fact that the longitudinal wave speed in plates and P-wave speed in bulk material, cL =s E ρ(1 − ν2) and cP =s E(1 − ν) ρ(1 + ν)(1 − 2ν) , (18) are both functions of the material properties only. Phase modulation in metasurface designs is usually done by re- ducing the wave speed in each waveguide to achieve the desired phase delay. Thickness variation of the plate-like waveguides does not provide such a mechanism to modu- late longitudinal waves in plates. The sppeds cL and cP are very close so that the effective P-wave impedance of the metasurface is similar to that of the background ma- terial. Therefore the normally incident P-wave will travel straight through the slab. Based on these properties, we One period of the metasurface (b) 1 ) ( t f i h s e s a h P 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 e d u t i l p m A One period of the metasurface FIG. 4. Metasurface design: (a) outline of the design; (b) transmission properties of each plate-like waveguide in one period of the metasurface. The blue circles and the black tri- angles represent the phase and amplitude of the transmitted wave through each plate, respectively. of the metasurface slab, i.e. length of each plate, is cho- sen as L = 5 cm; the width of all the gaps are identical, a = 1 mm. The thicknesses of the plates satisfying the constant phase gradient are selected from Fig. 3. Sixteen plates with thickness covering a complete phase change of 2π at 60 kHz form one period of the metasurface as shown in the dashed box in Fig. 4(a). The metasurface of infinite extent is formed by repeating the structure in the dashed box, such that the length of one period is 8.59 cm. The transmitted phases and amplitudes through all these plates are listed in Table I, while the transmitted ampli- TABLE I. Transmitted phase and amplitude through each L=5 cm long plate-like waveguide at 60 kHz. Plate h (mm) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 2.483 2.586 2.744 2.953 3.165 3.351 3.516 3.676 3.850 4.069 4.384 4.864 5.527 6.345 7.393 9.000 φ (π rad) -0.9599 -0.8789 -0.7946 -0.7058 -0.6121 -0.5138 -0.4113 -0.3048 -0.1950 -0.0807 0.0398 0.1698 0.3129 0.4730 0.6547 0.8672 T (a.u.) 0.7944 0.6886 0.6169 0.6253 0.7106 0.8268 0.9303 0.9901 0.9937 0.9410 0.8605 0.8103 0.8372 0.9154 0.9828 0.9988 tudes and phases in one period of the metasurface are plotted in Fig. 4(b). Note that the plate array here is ro- tated 90◦ counterclockwise, and this figure does not show the full length of each plate. It is clear that the transmit- ted phase strictly follows the constant spatial gradient. The amplitudes are not modulated in our design, but this does not affect the performance of the metasurface. Full wave simulations were performed in COMSOL to demonstrate the functionality of the metasurface at 60 kHz. Displacements in the y-direction were applied along the vertical line at the left side of the simulation domain to generate an in-plane shear wave. Simulation result for SV-wave incidence is shown in Fig. 5. The curl of the displacement field is plotted to show the distortional field of the wave, i.e. SV-wave. The black arrow indi- cates the propagation direction of the SV-wave, and the red arrow indicates the direction of particle motion of the incident wave. As indicated by the black arrow, a planar SV-wave is normally incident from the left side of the metasurface and is refracted at 32.5◦ which is very close to the designed refraction angle. The metasurface can maintain this steering angle of the uniform transmitted beam over a wide frequency range from 55 to 70 kHz. The metasurface does not work as effectively when the frequency is decreased or increased since the modulation is based on a single frequency. It is worthwhile to point out that there is no mode converted wave, i.e. P-wave, in the far-field of the transmitted region. This can be 6 ) . u . a ( s t n e m e c a p s i d f o l r u C l 1 - 5 cm x position [m] FIG. 5. Anomalous refraction of a normally incident SV-wave at 60 kHz. The curl of the displacement field is plotted to show the SV-wave. explained as the modes in the transmitted field mainly come from the waveguides, i.e. plates. In this design, the SV-wave impinges normally on the interface so that there is only a flexural wave in the plate array, which does not induce any mode converted wave in the transmitted field. It is noted that there are some longitudinal components in the interface wave at the right boundary of the meta- surface, but they do not effect the far-field and hence do not influence the functionality of the metasurface. Near- field wave motion at the interface of the metasurface and the bulk material are not well studied and remain to be further investigated. Another feature of the metasurface is that it does not alter the propagation direction of a normally incident P- wave; it can therefore be used as a mode splitter to sep- arate SV- and P-waves. Simulation results for normal incidence of P-wave are shown in Fig. 6. In this case, displacements in the x-direction were applied along the vertical line at the left side of the simulation domain to generate a longitudinal wave, and the trace of the strain 1 -1 ) . u . a ( r o s n e t n a r t s f o e c a r T i 5 cm x position [m] FIG. 6. Unaffected normally incident P-wave at 60 kHz. The trace of the strain tensor is plotted to show the P-wave. tensor is plotted to show the dilatational field. The black 7 Full wave FEM simulations using COMSOL demon- strate the focusing effect of the metasurface. Displace- ments in the y-direction were applied along the vertical line at the left side of the simulation domain to generate an in-plane shear wave. Figure 8 shows the simulated field at 60 kHz. The curl of the displacement field is B. Metasurface for focusing plane SV-waves m 2 arrow indicates the propagation direction of the P-wave, and the red arrow indicates the direction of particle mo- tion of the incident wave. It is clear that the transmitted P-wave still travels along the normal direction. Similar to the SV-wave incidence case, there is no mode con- verted wave in the transmitted region. Since there are no mode conversions for both SV- and P-wave incidence, this metasurface is capable of steering these two types of waves into different directions without introducing un- wanted wave types. Other than the metasurface for splitting SV- and P- waves, this approach can also be adopted in the design of a focal metasurface. The physics behind the focal meta- surface is different from the gradient index lenses23 -- 26 which are designed using ray theory and by varying the refractive indices of bulk materials. Here the focal meta- surface is based on the constructive and destructive in- terferences of the diffracted waves through all the waveg- uides. The required phase profile along the metasurface can be written as φ(y) = kT(cid:0)p(y − y0)2 + d2−d(cid:1)+φ0, −π < φ < π, (19) where y0 denotes the location where the metasurface is symmetric about and φ0 is the transmitted phase through the plate at y0. We choose y0 = 0 and set φ0=0 for convenience. The metasurface is chosen to have slab thickness, i.e. plate length, L = 5 cm; the constant gap width is a = 1 mm. The focal distance is selected to be d = 5L = 25 cm from the lens. The required phase profile at 60 kHz is calculated using Eq. (19) and plotted in Fig. 7. A total 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 ) ( -1.2 -0.3 -0.2 -0.1 0 y [m] 0.1 0.2 0.3 FIG. 7. Phase profile of the metasurface for focusing a plane SV-wave at 60 kHz. number of 97 plates are selected from Fig. 3 to form a metasurface. 1 -1 ) . u . a ( s t n e m e c a p s i d f o l r u C l 5 cm x position [m] FIG. 8. Focusing a normally incident plane SV-wave at 60 kHz. The curl of the displacement field is plotted to show the SV-wave. plotted to represent the distortional wave. The focus- ing effect can be clearly observed at the right side of the metasurface. It is interesting that a focal spot can be ob- served over a wide range of frequency from 40 to 75 kHz, however, the focal distance is varying with frequency due to the dispersive nature of the design elements. By com- paring the energy density along the x-direction across the center of the metasurface, the focal point at 60 kHz was found to be approximately 25.7 cm away from the interface, which agrees with the designed distance to a remarkable degree. C. Metasurface for focusing a cylindrical SV-wave We now design a metasurface for focusing a cylindrical SV-wave. Due to the cylindrical spreading of the wave- front, the phases of the incident wave along the metasur- face is different, therefore Eq. (19) needs to be revised. The profile requires a more rapid phase change as com- pared to the phase change for focusing a plane SV-wave. The modified phase profile can be written as14 φ(y) = kT(cid:0)p(y − y0)2 + d2 − d(cid:1) + φc(y), −π < φ < π, (20) where φc(y) is a phase correction term which compen- sates the phase difference of the incident wave. In this design the y0 location, the slab thickness and the gap width are set the same as for the plane wave incidence design. The focal distance is selected to be d = 5L = 25 cm. The only difference is that the cylindrical wave source location, distance dS from the metasurface, needs to be taken into account. For instance, if we choose dS = d in this design, then the phase correction term is simply φc(y) = kT(cid:0)p(y − y0)2 + d2 − d(cid:1). The required phase profile at 60 kHz is calculated using Eq. (20) and plotted in Fig. 9. Then a total number of 93 plates are 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 ) ( 8 of the power flux of the transmitted waves. Nearly 38.7% of the transmitted energy is converted to P-waves. The focal distance evaluated from the simulation results at 60 kHz is 28.3 cm, which is 3.3 cm longer than the design distance. Given that the focal distance in the previous design is only 0.7 cm longer than the designed distance, our model does not predict the transmitted phase ac- curately for oblique incidence. The main reason is that the current model does not consider the mode conversion which occurs for oblique incidence. Improvement of the model to include such effects will be considered later. As a comparison, the COMSOL-generated simulated P-wave field at 60 kHz is plotted in Fig. 11. The trace 1 r ra T -1 -1.2 -0.3 -0.2 -0.1 0 y [m] 0.1 0.2 0.3 FIG. 9. Phase profile of the metasurface for focusing a cylin- drical SV-wave at 60 kHz. selected from Fig. 3 to form a metasurface. The COMSOL-generated simulated SV-wave field at 60 kHz is plotted in Fig. 10. The curl field is shown to 5 cm on [m] x 1 FIG. 11. Cylindrical P-wave incidence at 60 kHz. The trace of the strain tensor is plotted to show the P-wave. 28.3 cm ) . u . a ( s t n e m e c a p s i d f o l r u C l -1 of the strain tensor is plotted to show the dilatational wave. It is clear that the phase change of the transmitted P-wave is almost negligible, and the wavefront is still cylindrical on the transmitted side. This is not surprising since the impedance of the slab is close to that of the background medium. Moreover, the slab width is sub- wavelength so that the phase shifts are small when the waves transmit through the plate array. IV. CONCLUSION 5 cm x position [m] FIG. 10. Focusing a cylindrical SV-wave source at 60 kHz. The curl of the displacement field is plotted to show the SV- wave. represent the distortional wave. The focal spot is evi- dent at the right side of the metasurface. This suggests that our design elements are suitable for rapidly chang- ing phase profiles. Similar to the previous focal meta- surface for a line source, this design also works over a broadband frequency range from 45 to 65 kHz with focal distances varying with frequency. It is also interesting to see the percentage of energy carried by the mode con- verted P-wave in the transmitted waves. Integration of the curl field and the strain were performed along the y- direction near the metasurface to estimate time-average In conclusion, we have presented a novel metasurface design approach for controlling SV-wave motion in elas- tic solids using plate-like waveguides of varying thickness. A theoretical model based on the Mindlin plate theory is developed and compared with the FEM simulation re- sults. The model works well for thick plates in the high frequency range and is therefore well suted to metasur- face design. It is also found that the transmission prop- erties for normally incident P- and SV-waves are distinct. The transmitted phase of a normally incident SV-wave can cover a full span of 2π. However, the transmitted phase of a normally incident P-wave is nearly constant. By taking advantage of these properties, we designed and numerically demonstrated several metasurfaces that are capable of steering SV-waves while remaining transpar- ent to P-wave. The fundamental mode of this type of wave in a thick plate is nondispersive and can travel in a planar manner over long distances thus is of particular interest in nondestructive evaluations.27 -- 29 ACKNOWLEDGMENTS This work was supported by Office of Naval Research through MURI Grant No. N00014-13-1-0631. REFERENCES 1N. Yu, P. Genevet, M. A. Kats, F. Aieta, J.-P. Tetienne, F. Ca- passo, and Z. Gaburro. Light propagation with phase disconti- nuities: Generalized laws of reflection and refraction. Science, 334(6054):333 -- 337, 2011. doi:10.1126/science.1210713. 2S. Sun, Q. He, S. Xiao, Q. Xu, X. Li, and L. Zhou. Gradient-index meta-surfaces as a bridge linking propagating waves and surface waves. Nat. Mater., 11(5):426 -- 431, 2012. doi:10.1038/nmat3292. 3A. V. Kildishev, A. Boltasseva, and V. M. Shalaev. Planar pho- tonics with metasurfaces. Science, 339(6125):1232009, 2013. doi: 10.1126/science.1232009. 4C. Pfeiffer and A. Grbic. Metamaterial Huygens' surfaces: Tai- loring wave fronts with reflectionless sheets. Phys. Rev. Lett., 110(19):197401, 2013. doi:10.1103/physrevlett.110.197401. 5N. M. Estakhri and A. Al`u. Wave-front transformation with gradient metasurfaces. Phys. Rev. X, 6(4):041008, 2016. doi: 10.1103/physrevx.6.041008. 6Y. Li, B. Liang, X. Tao, X. Zhu, X. Zou, and J. Cheng. Acoustic focusing by coiling up space. Appl. Phys. Lett., 101(23):233508, 2012. doi:10.1063/1.4769984. 7W. Wang, Y. Xie, A. Konneker, B.-I. Popa, and S. A. Cummer. Design and demonstration of broadband thin planar diffractive acoustic lenses. Appl. Phys. Lett., 105(10):101904, 2014. doi: 10.1063/1.4895619. 8Y. Li, G. Yu, B. Liang, X. Zou, G. Li, S. Cheng, and J. Cheng. Three-dimensional ultrathin planar lenses by acoustic metama- terials. Sci. Rep., 4:6830, 2014. doi:10.1038/srep06830. 9J. Zhao, B. Li, Z. Chen, and C.-W. Qiu. Redirection of sound waves using acoustic metasurface. Appl. Phys. Lett., 103(15): 151604, 2013. doi:10.1063/1.4824758. 10J. Zhao, B. Li, Z. Chen, and C.-W. Qiu. Manipulating acoustic wavefront by inhomogeneous impedance and steerable extraordi- nary reflection. Sci. Rep., 3:2537, 2013. doi:10.1038/srep02537. 11M. Dubois, C. Shi, Y. Wang, and X. Zhang. A thin and conformal metasurface for illusion acoustics of rapidly changing profiles. Appl. Phys. Lett., 110(15):151902, 2017. doi:10.1063/1.4979978. 12Y. Xie, W. Wang, H. Chen, A. Konneker, B.-I. Popa, and S. A. Cummer. Wavefront modulation and subwavelength diffractive acoustics with an acoustic metasurface. Nat. Commun., 5:5553, 2014. doi:10.1038/ncomms6553. 13Y. Li, X. Jiang, B. Liang, J. Cheng, and L. Zhang. Metascreen- based acoustic passive phased array. Phys. Rev. Applied, 4(2), 2015. doi:10.1103/physrevapplied.4.024003. 9 14Y. Li, S. Qi, and M. B. Assouar. Theory of metascreen-based acoustic passive phased array. New J. Phys., 18(4):043024, 2016. doi:10.1088/1367-2630/18/4/043024. 15X. Jiang, Y. Li, B. Liang, J. Cheng, and L. Zhang. Convert acous- tic resonances to orbital angular momentum. Phys. Rev. Lett., 117(3):034301, 2016. doi:10.1103/PhysRevLett.117.034301. 16C. Shi, M. Dubois, Y. Wang, and X. Zhang. High-speed acoustic communication by multiplexing orbital angular momen- tum. Proc. Natl. Acad. Sci., 114(28):7250 -- 7253, 2017. doi: 10.1073/pnas.1704450114. 17H. Zhu and F. Semperlotti. Anomalous refraction of acous- tic guided waves in solids with geometrically tapered meta- surfaces. doi: 10.1103/physrevlett.117.034302. Phys. Rev. Lett., 117(3):034302, 2016. 18Y. Liu, Z. Liang, F. Liu, O. Diba, A. Lamb, and J. Li. lamb waves using elastic doi: Source illusion devices for flexural metasurfaces. Phys. Rev. Lett., 119(3):034301, 2017. 10.1103/PhysRevLett.119.034301. 19X. Su and A. N. Norris. Focusing, refraction, and asymmetric transmission of elastic waves in solid metamaterials with aligned parallel gaps. J. Acoust. Soc. Am., 139(6):3386 -- 3394, 2016. doi: 10.1121/1.4950770. 20R. D. Mindlin. Influence of rotary inertia and shear on flexural motions of isotropic, elastic plates. J. Appl. Mech. ASME, 18: 31 -- 38, 1951. 21C. Mei and B. Mace. Wave reflection and transmission in timo- shenko beams and wave analysis of timoshenko beam structures. J. Vib. Acoust., 127(4):382 -- 394, 2005. doi:10.1115/1.1924647. 22R. D. Gregory and F. Y. M. Wan. Boundary conditions at the edge of a thin or thick plate bonded to an elastic support. J. Elasticity, 36(2):155 -- 182, 1994. doi:10.1007/bf00040963. 23A. Climente, D. Torrent, and J. Sanchez-Dehesa. Gradient index lenses for flexural waves based on thickness variations. Appl. Phys. Lett., 105(6):064101, 2014. doi:10.1063/1.4893153. 24S. Tol, F. Degertekin, and A. Erturk. Phononic crystal luneb- urg lens for omnidirectional elastic wave focusing and energy harvesting. doi: 10.1063/1.4991684. Appl. Phys. Lett., 111(1):013503, 2017. 25X. Su, A. N. Norris, C. W. Cushing, M. R. Haberman, and P. S. Wilson. Broadband focusing of underwater sound using a trans- parent pentamode lens. J. Acoust. Soc. Am., 141(6):4408 -- 4417, 2017. doi:10.1121/1.4985195. 26X. Su and A. N. Norris. Isotropic transformation acoustics and applications. Proc. of SPIE, 10170:101700M, 2017 doi: 10.1117/12.2259793. 27M. Hirao and H. Ogi. An SH-wave EMAT technique for gas pipeline inspection. NDT E Int., 32(3):127 -- 132, 1999. doi: 10.1016/S0963-8695(98)00062-0. 28J. S. Lee, Y. Y. Kim, and S. H. Cho. Beam-focused shear- horizontal wave generation in a plate by a circular magne- tostrictive patch transducer employing a planar solenoid array. Smart Mater. Struct., 18(1):015009, 2008. doi:10.1088/0964- 1726/18/1/015009. 29P. Petcher, S. E. Burrows, and S. Dixon. Shear horizontal (SH) ultrasound wave propagation around smooth corners. Ultrason- ics, 54(4):997 -- 1004, 2014. doi:10.1016/j.ultras.2013.11.011.
1712.02279	4	1712	2018-01-21T06:26:51	A novel superconducting magnetic levitation method to support the laser fusion capsule by using permanent magnets	[ "physics.app-ph", "physics.plasm-ph" ]	A novel magnetic levitation support method is proposed, which can relieve the perturbation caused by traditional support methods and provide more accurate position control of the capsule. This method can keep the perfect symmetry of the octahedral spherical hohlraum and has the characteristics in stability, tunability and simplicity. It is also favorable that all the results, such as supporting forces acting on the superconducting capsule, are calculated analytically, and numerical simulations are performed to verify these results. A typical realistic design is proposed and discussed in detail. The superconducting coating material is suggested, and the required superconducting properties are listed. Damped oscillation of the floating capsule in thin helium gas is discussed, and the restoring time is estimated.	physics.app-ph	physics	A novel superconducting magnetic levitation method to support the laser fusion capsule by using permanent magnets Xiaojia Lia,∗, Tingting Xiaoa, Fengwei Chena, Yingjuan Zhanga, Weidong Wua,b,c,∗∗ aResearch Center of Laser Fusion, China Academy of Engineering Physics, Mianyang 621900, China bScience and Technology on Plasma Physics Laboratory, Mianyang 621900, China cIFSA Collaborative Innovation Center, Shanghai Jiao Tong University, Shanghai 200240, China 8 1 0 2 n a J 1 2 ] h p - p p a . s c i s y h p [ 4 v 9 7 2 2 0 . 2 1 7 1 : v i X r a Abstract A novel magnetic levitation support method is proposed, which can relieve the perturbation caused by traditional support methods and provide more accurate position control of the capsule. This method can keep the perfect symmetry of the octahedral spherical hohlraum and has the characteristics in stability, tunability and simplicity. It is also favorable that all the results, such as supporting forces acting on the superconducting capsule, are calculated analytically, and numerical simulations are performed to verify these results. A typical realistic design is proposed and discussed in detail. The superconducting coating material is suggested, and the required superconducting properties are listed. Damped oscillation of the floating capsule in thin helium gas is discussed, and the restoring time is estimated. PACS codes: 52.57.Fg; 74.70.Ad; 74.78.-W Keywords: ICF capsule support, Magnetic levitation, Symmetry 1. Introduction Symmetry plays a crucial role in inertial confinement fu- sion (ICF) experiments, because any perturbation to the spher- ical symmetry will be strongly amplified through ablative hy- drodynamic instabilities[1, 2]. To improve the radiation sym- metry, indirect drive has been developed decades ago, spher- ical hohlraum has been reintroduced in recent years[3, 4, 5]. Capsule support is another important link in ICF experiments, which influences the symmetry around the capsule directly. Tra- ditionally, people use thin wires, tents, low density foams or their hybrid systems[6] to support the deuterium-tritium fuel capsule. However, in all the traditional methods, the supporting materials require direct contact with the capsule. Experimen- tal results of National Ignition Facility (NIF) show that large perturbation exists where the tent departs from the capsule[6]. Researchers are trying to make the wires thinner and the foam density lower, sacrificing their mechanical strength, but pertur- bation to the spherical symmetry is still inevitable, no matter how small it could be. It is then natural to think that only non- contact method can hopefully solve this problem and preserve the spherical symmetry around the capsule maximally. Dif- ferent kinds of levitation methods has been pursued since the early years of ICF research since the 1970s[7]. To our knowl- edge, the idea of using superconductors to support ICF capsule ∗corresponding author ∗∗corresponding author [email protected] [email protected] was first presented by David Glocker[8], whose design was to support a permanently magnetized fuel pellet within a spher- ical superconducting shell. According to Meissner effect, su- perconductors are diamagnets, so that they will feel a repulsive magnetic force which points at the direction of the magnetic induction gradient. This magnetic force could be used to coun- teract gravity with proper design of the magnetic field. How- ever, Glocker's design can no longer cooperate with nowadays ICF experiment. Y. Ishigaki et al. developed an accurate po- sition control system for superconducting ICF capsule using three electromagnet coils[9, 10]. But technically, it does not seem possible to put three coils around the capsule without in- terfering with the laser paths, especially in the octahedral spher- ical hohlraums scheme where there are six laser entrance holes (LEHs)[4]. There are also relative ideas such as using negative feedback digital circuits to control the capsule position[11], us- ing magnetic levitation to build a capsule transport system[12], which we do not intend to comment too much on. With symmetry preserved naturally, there are several issues we need to consider during schematic designing, such as sta- bility, tunability and simplicity. First, the floating pellet must be in stable equilibrium in all degrees of freedom, both trans- lational and rotational. Second, the equilibrium point should be tunable in a certain range and can be accurately controlled, because there are always slight mass differences between dif- ferent pellets. Third, the supporting system should be compact enough to avoid all the laser paths and cooperate with other ac- cessories around the hohlraum. Meanwhile, it should also be Preprint submitted to Matter and Radiation at Extremes June 24, 2021 simple enough and easy to be installed. In this paper, we in- troduce a novel non-contact support method, in which the cap- sule is coated with superconducting thin film, and a potential well is designed right at the center of the hohlraum. Instead of coils[9, 10], we use the mature Nd-Fe-B permanent magnets to generate the magnetic field. The typical magnetic flux density at the surface of a single Nd-Fe-B magnet is about 1.2 tesla, then we have from Ampere's law that the corresponding effec- tive surface current density is as high as M ∼ 106A/m, so that the magnets could be much smaller than coils due to their in- tense magnetization. The cryogenic cooling temperature of the capsule is about 20 Kelvin, giving us strict restriction in choos- ing material of the superconducting film. Another restriction comes from the ablation process, which demands that only low- Z atoms are acceptable[13]. MgB2 becomes the ideal choice due to the above restrictions, whose average Z-value is rela- tively low and superconducting critical temperature is around 39 Kelvin[14]. According to our initial settings, the film-coated pellet is supposed to behave like a solid perfect diamagnet. MgB2 is a type II superconducting material[15], so that the pel- let can be treated as a perfect diamagnetic sphere if and only if the outer magnetic flux density is below the lower critical field Bc1, and the film thickness is over several times the London pen- etration depth λ. Fortunately, the above conditions are not strin- gent. It is reviewed and concluded in Ref.[15] that Bc1 is around 125mT, and the corresponding λ is about 40nm. Meanwhile, our following calculations show that the required flux density in the vicinity of the capsule is only about 10mT, and films of 100nm to 200nm are acceptable for the ablation[16, 17]. Tak- ing these wonderful properties of MgB2 into consideration, we believe that the E − J complexities and flux pinning will not play a central role in the levitation process. The above discus- sions about MgB2 have to be verified experimentally. While performing the simulations, we simply set the capsule's rela- tive permeability µr = 0, and the capsule will behave similar enough to a superconductor (a perfect diamagnet). However, the levitating force may be weakened if the MgB2 film is not as perfect as expected, so that we need to study this issue carefully in the future. This article is organized as follows. In section 2, the ele- mentary model is analyzed thoroughly. Using small pellet ap- proximation, both near field and far field cases are discussed. The proposed supporting method is described in detail in sec- tion 3. In section 4, time needed for the capsule to restore equilibrium is estimated using perturbation method. Finally, we give a summary in section 5. 2. Analysis of the elementary model The simplest model in generating magnetic field must be a closed electric circle with radius a, carrying current I. As for permanent magnet, a uniformly magnetized wafer (a thin cylin- der whose height h is much smaller than the radius of under- surface a) is equivalent to such an electric circle, if its magne- tization direction is parallel to its generatrix. This equivalence relation can be expressed quantitatively as I = h · M. (1) M represents the norm of the magnetization vector M, similar notations will be widely used in the following text. Specifi- cally, a thin cylindrical magnet with thickness h = 0.5mm is equivalent to a circle with 500 ampere current. We would keep M = 106A/m a constant for simplicity. First, we need to cal- culate the magnetic field generated by the circle; second, we have to evaluate the magnetic force acting on the superconduct- ing capsule. The above two steps are not difficult to work out separately, and most of the working procedures can be found in textbooks of electromagnetic dynamics. So we tend to list some useful results only, and focus our attention on the discussions and applications of these results. Suppose a circle (cylinder) is placed in the z = 0 plane of the standard cylindrical coordinates, and its center is coincide with the original point. The current I runs counter clockwise seeing from above. According to sym- metry, the magnetic vector potential A has only one non-zero component Aφ, which can be written directly as: (cid:90) 2π 0 Aφ(r, z) = µ0Ia 4π (cid:112) (cid:48) r2 + z2 + a2 − 2arcosφ(cid:48) dφ cosφ(cid:48) . (2) The result of Eq. (2) can be expressed by the secondary elliptic integration. These analytical calculations are performed using Wolfram Mathematica, and all figures except Fig.4 are drawn by this software. After A is known, we can derive the magnetic field by B(r, z) = ∇ × A(r, z). Vector B has two non-zero components, Br and Bz. We do not introduce any approximations here, but in the next step, we have to introduce the so-called small pellet approximation, in order to calculate the magnetic forces perturbatively. First, assume that the pellet radius is much smaller than the curvature radius of the contours of the magnetic flux density, which allows us to ignore the field variance when deriving the magnetization of the pellet. Considering a superconducting sphere with ra- dius r0 placed into uniform magnetic field B, the sphere will be magnetized as a small dipole. Using the standard coordinates separation method in the spherical coordinates, together with proper boundary conditions, we can solve the magnetic scalar potential, which is expressed as the summation of two terms. The first term is just the original outer potential, and the second term represents a potential generated by a magnetic dipole: m0 = −2πr3 0 µ0 B. (3) This should be the dipole moment of the sphere itself, which is induced by the outer field. The second meaning of small pellet approximation is that we ignore the radius of the pellet and treat it as a point dipole placed in curved magnetic field, then the corresponding potential energy is: Um = −1 2 m0 · B. (4) 2 Figure 1: Fz as a function of z at r = 0. Figure 3: Distribution of total potential Ut. Red lines represent the contour surfaces of Ut Figure 2: Fr as a function of r at different height. One may notice that there is an extra factor 1/2 compared with the normal expression. This factor arises because half of the energy is used to magnetize the pellet. After these preparations we can finally get the expression of the magnetic force: Fm = −∇Um. (5) We test the effectiveness of these equations by comparing the results with numerically simulated ones. Considering a near field case, if we set a = 9mm, r0 = 0.8mm and I = 355A a pri- ori, the pellet with m = 10mg can be floated at the destination height of z0 = 5mm, right above the center of the circle. Sim- ilarly, we set h = 0.355mm while doing the simulation using finite element method software. The results are shown in Fig.1 and Fig.2, in which the blue lines are calculated forces and the simulated forces are shown by red points. Results obtained by different methods show good agree- ment with each other, which demonstrates that the small pel- let approximation works excellent in such a near field example. We can see from Fig.1 that the simulated supporting force Fz is always slightly smaller than the calculated one. This is a systematic error caused by the measurement of distance. Ap- parently, Fz(z) is a convex function in the vicinity of the equi- librium point, so that the supporting force decreases faster than linear with the increase of z. But when performing the simula- tions, the distances are measured from the geometry centers of the objects. Now we discuss the stability problem we put forward be- fore. Translational equilibrium is acquired naturally by putting the capsule into the equilibrium point z = 5mm, r = 0mm, where Fz = mg and Fr = 0. In the vicinity of the equilibrium point, stability in z direction is obvious. In r direction there is a critical height (approximately z/a = 0.63), above which the equilibrium will be unstable because Fr will become positive. The stable range can be found by defining the total potential energy: Ut = Um + mgz, (6) where g = 9.8N/kg is the gravitational acceleration. The struc- ture of Ut is shown in Fig.3, which confirms the existence of the axially symmetric three dimensional potential well of about 4mm×1mm in size. Such a well is large enough to trap the cap- sule. In the rotational degrees of freedom, stability is naturally guaranteed by the magnetization mechanism of superconduc- tors. According to Eq.(3), the direction of the dipole is just op- posite to the outer field B, so that the rotational moment acting on the capsule is L = m0 × B ∝ B × B = 0, (7) (8) at any case. We now turn to the far field case, which means the distance from the dipole to the circle R is much greater than both r0 and a. The small pellet approximation is then automatically satis- fied. The circle then can also be treated as a magnetic dipole with moment mc = πr2 0hM, whose magnetic scalar potential at a field point R is already well known as: mc · R 4πR3 . ϕm = Eq.(8) comes from the leading term of Taylor expansion of the exact expression of ϕm, which tells us that in far field case, all other components of the magnetic force F become insignificant compared with Fz along z axis. We briefly summarize this section as follows. The magnetic field of a thin cylindrical permanent magnet is studied both an- alytically and numerically. A stable equilibrium point exists in near field case, with proper adjustment of relevant parameters. In far field case, the magnet wafer acts as a rod, which provides only a force perpendicular to its undersurface. These conclu- sions will be used in the next section. 3 2345670.60.81.01.21.4Fz/10-4Nz/mmmg123456-2.5-2.0-1.5-1.0-0.50.5Fr/10-5Nr/mmz=4mmz=5mmz=6mming a detailed design with typical geometry parameters as fol- lows. We put the gasket like magnet on the z = 0 plane as before, whose inner and outer radius is r1 = 4.4mm and r2 = 6.6mm respectively. A spherical hohlraum of r3 = 5.5mm can be put tightly in the gasket, with its center at z0 = 3.3mm. The destination height above the gasket is then 3.3mm because the capsule must be put at the center of the hohlraum. Four small wafers with radius r4 = 0.88mm are stick on the hohlraum un- der the gasket right below the x and y axes. Their undersurfaces are set θ = 35 degrees away from horizontal, facing the des- tination point, as shown in Fig.4. The thickness of the gasket and the wafers are d1 = d4 = 0.4mm(with equivalent currents 400A). If the radius of the LEHs is set to be 1.2mm, the largest allowed laser entrance angle will be θi = 67 degrees, which is large enough for experiments[5]. Without loss of generality, we set the wafers to be magnetized toward the capsule. To enhance (other than weaken) the field of the wafers, the gasket should be magnetized downward. Using Eqs.(2)-(5), together with some trivial translational and rotational coordinates transformations, we can obtain the supporting force for a r0 = 1.1mm superconducting pellet is Fz = 2.093 × 10−4N. The simulated result is 2.092 × 10−4N, which is also slightly smaller as expected. After inputting the corresponding equilibrium mass (m = 21.36mg) to Eq.(6), we can plot the total potential of x − z plane in Fig.5. A poten- tial well of 10mm × 3.5mm is shown clearly, which should be satisfactory for practical application. The potential well on the y = x cut plane looks similar but slightly larger, because the wafers are placed under the x and y axes, which breaks the ax- ial symmetry of the original system. With stability guaranteed by the potential well, tunability is another problem to tend to. This supporting system is highly flexible. Parameters as r1, d1, r4, d4 and θ can be tuned to adjust different hohlraum sizes and different capsule masses, which we can call coarse tuning. In traditional methods, the position error of the capsule is mainly caused by the supporting wires or foams, whose mechanical strength is limited by geometry. This problem no longer exists in our method because Nd-Fe-B mag- nets are solids with excellent mechanical strength. Accurate position control of the capsule can be achieved by fine tuning the position of the gasket through mechanical methods. As for simplicity, this system is compact enough to keep away from all laser paths. 4. Estimation of restoring time of the floating capsule in thin helium This section is relatively independent, but essential for the application of this supporting method. Recall that our initial purpose is to fix the capsule to a exact point with minimal er- ror. The potential well ensures the specific equilibrium point, but it is difficult to find this very point at beginning. Suppose the capsule is released statically at some random point within 1mm from the equilibrium point, then the capsule will become a harmonic oscillator approximately. Taking the actual situa- tion into consideration that the hohlraum is filled with helium gas, the amplitude of the oscillation must decrease with time. 4 Figure 4: Typical design of the magnets' placements. Figure 5: Cut plane (x-z plane) of the total potential well. 3. Realistic design of the supporting scheme In the six LEHs spherical hohlraums scheme[4], one LEH is set right under the capsule, making it impossible to put a solid magnet underneath. However, a thin cylinder with a coaxial hole is doable if the radius of the hole is large enough. Such a gasket like magnet is equivalent to two electric loops placed at the inner and outer circle of the magnet. The two loops share the same current intensity but opposite current directions. We can see from common sense that near field distribution is dom- inated mainly by the inner loop, but the outer loop will weaken the field and the supporting force, making the stable equilib- rium range lower than the elementary one-circle case. Up to some specific height, the magnetic field will switch direction. This critical height is determined by the ratio of the inner and outer loop diameters. Consequently, we can not increase the levitating height by simply increasing the effective current. For- tunately, the gasket like magnet is placed out of the spherical hohlraum and the existence of the hole allows us to put the magnet higher, but the outer diameter must be set smaller at the same time to avoid laser paths. To achieve satisfactory desti- nation height and stability, small wafers of permanent magnets can be used as rods to counteract part of the capsule gravity and refine the shape of the potential well. These small wafers can be placed on the outer surface of the hohlraum. We demonstrate the effectiveness of this scheme by propos- This system is just a typical three dimensional damped oscilla- tor. How much time it takes to achieve equilibrium is one thing that engineers care about. Generally, the equation of motion for a one dimensional damped oscillator can be written as: q(cid:48)(cid:48)(t) + q(cid:48)(t) + α m κ m q(t) = 0. (9) In our case, α = 6πr0η is the friction coefficient caused by helium viscosity, and κ = −F(q)/q is the linear restoring co- efficient determined by the total potential Ut. Before solving the above equation, we evaluate the order of magnitude of both terms. From material data base we find that at the typical tem- perature (T = 20K) and pressure (p = 0.1atm) of the hohlraum, helium viscosity η = 3.55 × 10−6Pa · s, leading to α/m = 3.44× 10−3s−1. Meanwhile, even at the weakest direction of the potential well, κ/m ∼ 700s−2. That is to say, the weak damping condition is perfectly satisfied, which allows us to solve Eq.(9) perturbatively. Taking the second term as a perturbation, to- gether with the initial values q(0) = q0 ∼ 1mm and q(cid:48)(0) = 0, the solution up to O( α m) is then written as: (cid:114) q(t) = q0e − t t0 cos( κ m t), (10) √ where we defined the eigen time t0 = 2m/α = 580s. We can see from Eq.(10) that the oscillation magnitude decreases expo- nentially with time, and will be one order of magnitude smaller after every (ln 10 · t0 = 22) minutes. For the realistic three dimensional case, the position error is magnified only by a fac- tor of 3. We conclude that the position error of the capsule should be smaller than 10µm within 50 minutes, which is al- ready more accurate than traditional methods. This estimation is rather conservative, because the weak damping approxima- tion no longer works if the amplitude is small enough. As a re- sult, the real restoring time should be much shorter. In addition, the fill tube also plays a role in damping the oscillation, even ad- ditional damping mechanism could be introduced if necessary. As far as we know, this result is acceptable for experiments. 5. Summary In this paper, we propose a novel non-contact method to support the capsule in octahedral spherical hohlraum, in which only superconducting film and permanent magnets are needed. The feasibility of the method is theoretically verified by both analytical and numerical methods. The high symmetry around the capsule can be maximally preserved because of the non- contact nature of this method. Since there is only one stable equilibrium point in the potential well, and the position of the magnets can be tuned more precisely, the capsule position can be controlled more accurately. This system is adequately small in geometry size, thus applicable in engineering. In the cylin- drical hohlraum case, even the wafers are not necessary because the gasket is almost free in z direction. Time required to restore equilibrium is estimated conservatively, which we believe is not a problem to worry about. 5 The main challenge of this project might be the fabrication of the high quality MgB2 thin film on a small sphere, which affects the diamagnetism property of the capsule directly. Cor- responding experimental research is being carried out simulta- neously by our team. References [1] R. Betti, V. N. Goncharov, R. L. McCrory, P. Sorotokin, and C. P. Verdon, Self-consistent stability analysis of ablation fronts in inertial confinement fusion, Phys. Plasmas 3 (1996) 2122. [2] V. N. Goncharov, R. Betti, R. L. McCrory, P. Sorotokin, and C. P. Ver- don, Self-consistent stability analysis of ablation fronts with large Froude numbers, Phys. Plasmas 3 (1996) 1402. [3] K. Lan, J. Liu, X. T. He, W. D. Zheng, D. X. Lai, High flux symmetry of the spherical hohlraum with octahedral 6 LEHs at the hohlraum-tocapsule radius ratio of 5.14, Phys. Plasmas 21 (2014) 010704. [4] W. Huo, Z. Li, Y. Chen, X. Xie, K. Lan, J. Liu , G. Ren, Y. Li, Y. Liu, X. Jiang, D. Yang, S. Li, L. Guo, H. Zhang, L. Hou, H. Du, X. Peng, T. Xu, C. Li, X. Zhan, G. Yuan, H. Zhang, B. Jiang, L. Huang, K. Du, R. Zhao, P. Li, W. Wang, J. Su, Y. Ding, X. He and W. Zhang, First Investigation on the Radiation Field of the Spherical Hohlraum, Phys. Rev. Lett. 117 (2016) 025002. [5] K. Lan, J. Liu, Z. Li, X. Xie, W. Huo, Y. Chen, G. Ren, C. Zheng, D. Yang, S. Li, Z. Yang, L. Guo, S. Li, M. Zhang, X. Han, C. Zhai, L. Hou, Y. Li, K. Deng, Z. Yuan, X. Zhan, F. Wang, G. Yuan, H. Zhang, B. Jiang, L. Huang, W. Zhang, K. Du, R. Zhao, P. Li, W. Wang, J. Su, X. Deng, D. Hu, W. Zhou, H. Jia, Y. Ding, W. Zheng, X. He, Progress in octahedral spherical hohlraum study, Matter Radiat. Extrem. 1 (2016) 8. [6] C. R. Weber, D. T. Casey, D. S. Clark, B. A. Hammel, A. MacPhee, J. Milovich, D. Martinez, H. F. Robey, V. A. Smalyuk, M. Stadermann, P. Amendt, S. Bhandarkar, B. Chang, C. Choate, J. Crippen, S. J. Felker, J. E. Field, S. W. Haan, S. Johnson, J. J. Kroll, O. L. Landen, M. Marinak, M. Mcinnis, A. Nikroo, N. Rice, and S. M. Sepke, Improving ICF implo- sion performance with alternative capsule supports, Physics of Plasmas 24 (2017) 056302. [7] Editor: P. E. Coyle, Laser Program Annual Report, 4 (1976) 181. [8] D. A. Glocker, A Proposed Design for Multishell Cryogenic Laser Fusion Targets using Superconducting Levitation, Appl. Phys. Lett. 39 (1981) 478. [9] Y. Ishigaki, H. Ueda, K. Agatsuma, and A. Ishiyama, Accurate Position Control of Active Magnetic Levitation Using Sphere-Shaped HTS Bulk for Inertial Nuclear Fusion, IEEE. T. Appl. Supercon. 19 (2009) 3. [10] T. Wang, H. Ueda, K. Agatsuma, and A. Ishiyama, Evaluation of Posi- tional Stability in Active Magnetic Levitation Using Spherical HTS Bulk for Inertial Nuclear Fusion, IEEE. T. Appl. Supercon. 21 (2011) 3. [11] H. Yoshida, K. Katakami, Y. Sakagami, Magnetic Suspension of a Pellet for Inertial Confinement Fusion, Laser and Particle Beam, 11 (1993) 455. [12] E. R. Koresheva, I. V. Aleksandrova, O. M. Ivanenko, V. A. Kalabukhov, E. L. Koshelev, A. I. Kupriashin, K. V. Mitsen, M. Klenov, I. E. Osipov, L. V. Panina, HTSC Maglev Systems for IFE Target Transport Applications, Journal of Russian Laser Research, 35 (2014) 151. [13] G. Meng, J. Wang, X. Wang, J. Li, W. Zhang, Generation of a sharp density increase in radiation transport between high-Z and low-Z plasmas, Matter Radiat. Extrem. 1 (2016) 249. [14] J. Nagamatsu, N. Nakagawa, T. Muranaka, Y. Zenitani and J. Akimitsu, Superconductivity at 39 K in magnesium diboride, Nature 410 (2001) 63. [15] M. Eisterer, Magnetic properties and critical currents of MgB2, Super- cond. Sci. Technol. 20 (2007) 47, Topical Review. [16] J. Gu, Z. Dai, S. Zou, W. Ye, W. Zheng, P. Gu, and S. Zhu, Effects of mode coupling between low-mode radiation flux asymmetry and intermediate- mode ablator roughness on ignition capsule implosions, Matter Radiat. Extrem. 2 (2017) 9. [17] S. P. Regan, R. Epstein, B. A. Hammel, L. J. Suter, H. A. Scott, M. A. Barrios, D. K. Bradley, D. A. Callahan, C. Cerjan, G. W. Collins, S. N. Dixit, T. Doppner, M. J. Edwards, D. R. Farley, K. B. Fournier, S. Glenn, S. H. Glenzer, I. E. Golovkin, S. W. Haan, A. Hamza, D. G. Hicks, N. Izumi, O. S. Jones, J. D. Kilkenny, J. L. Kline, G. A. Kyrala, O. L. Lan- den, T. Ma, J. J. MacFarlane, A. J. MacKinnon, R. C. Mancini, R. L. McCrory, N. B. Meezan, D. D. Meyerhofer, A. Nikroo, H.-S. Park, J. Ralph, B. A. Remington, T. C. Sangster, V. A. Smalyuk, P. T. Springer, and R. P. J. Town, Hot-Spot Mix in Ignition-Scale Inertial Confinement Fusion Targets, Phys. Rev. Lett. 111 (2013) 045001. 6
1901.07664	1	1901	2019-01-23T00:47:22	Efficient Er/O Doped Silicon Light-Emitting Diodes at Communication Wavelength by Deep Cooling	[ "physics.app-ph" ]	A silicon light source at communication wavelength is the bottleneck for developing monolithically integrated silicon photonics. Doping silicon with erbium ions was believed to be one of the most promising approaches but suffers from the aggregation of erbium ions that are efficient non-radiative centers, formed during the standard rapid thermal treatment. Here, we apply a deep cooling process following the high-temperature annealing to suppress the aggregation of erbium ions by flushing with Helium gas cooled in liquid nitrogen. The resultant light emitting efficiency is increased to a record 14% at room temperature, two orders of magnitude higher than the sample treated by the standard rapid thermal annealing. The deep-cooling-processed Si samples were further made into light-emitting diodes. Bright electroluminescence with a spectral peak at 1.54 um from the silicon-based diodes was also observed at room temperature. With these results, it is promising to develop efficient silicon lasers at communication wavelength for the monolithically integrated silicon photonics.	physics.app-ph	physics	Efficient Er/O Doped Silicon Light-Emitting Diodes at Communication Wavelength by Deep Cooling Huimin Wen1,#, Jiajing He1,#, Jin Hong2,#, Fangyu Yue2* & Yaping Dan1* 1University of Michigan-Shanghai Jiao Tong University Joint Institute, Shanghai Jiao Tong University, Shanghai, 200240, China. 2Key Laboratory of Polar Materials and Devices, Ministry of Education, East China Normal University, Shanghai, 200241, China. #These authors contributed equally * To whom correspondence should be addressed: [email protected], [email protected] Abstract A silicon light source at communication wavelength is the bottleneck for developing monolithically integrated silicon photonics. Doping silicon with erbium ions was believed to be one of the most promising approaches but suffers from the aggregation of erbium ions that are efficient non-radiative centers, formed during the standard rapid thermal treatment. Here, we apply a deep cooling process following the high- temperature annealing to suppress the aggregation of erbium ions by flushing with Helium gas cooled in liquid nitrogen. The resultant light emitting efficiency is increased to a record 14% at room temperature, two orders of magnitude higher than the sample treated by the standard rapid thermal annealing. The deep-cooling-processed Si samples were further made into light-emitting diodes. Bright electroluminescence with a spectral peak at 1.54 µm from the silicon-based diodes was also observed at room temperature. With these results, it is promising to develop efficient silicon lasers at communication wavelength for the monolithically integrated silicon photonics. The monolithic integration of fiber optics with complementary metal-oxide- semiconductor (CMOS) integrated circuits will significantly speed up the computing and data transmission of current communication networks.1-3 This technology requires efficient silicon-based lasers at communication wavelengths.4, 5 Unfortunately, silicon is an indirect bandgap semiconductor and cannot emit light at communication wavelengths. Erbium ions (Er3+) have radiative emissions at 1.54 µm which is in the telecommunication wavelength band for long haul optic fibers.5-8 Introducing erbium ions (often with oxygen) into silicon by ion implantation was believed to be one of most promising approaches8-10 among others11-16 to create silicon-based lasers. However, during the standard rapid thermal treatment, erbium dopants aggregate to form clusters that are efficient nonradiative recombination centers.17, 18 As a result, the emission from erbium doped silicon is extremely weak at room temperature.5-10, 17, 18 Here, we applied a deep cooling (DC) process following the high-temperature thermal annealing on the erbium and oxygen co-doped silicon wafer at a cooling rate up to 1000 °C/s by flushing with Helium gas cooled in liquid nitrogen (77 K). The dramatic cooling process suppresses the aggregation of Er dopants that occurs during the slow cooling process of the high-temperature thermal annealing, allowing for a strong room-temperature emission at 1.54 µm from the Er/O co-doped silicon wafer. The internal quantum efficiency of the emission reaches a record of 14%, approximately 2 orders of magnitude higher than the samples treated by the standard rapid thermal annealing (RTA). The DC-processed Si samples were further made into light-emitting diodes (LED). Bright electroluminescence with a spectral peak at 1.54 µm from the silicon- based diodes was also observed at room temperature. Given the single crystalline nature of the silicon substrate, efficient silicon lasers may be readily developed for the full integration of fiber optics with CMOS circuitry. Figure 1 │ Temperature-dependent PL spectra of the Er/O-Si samples. a, PL spectra of the deep cooling (DC) processed samples measured at different temperatures. b, Peak intensity as a function of temperature for the DC- (red squares) and RTA- processed (blue dots) samples, respectively. c, Ambient room-temperature PL spectrum for the DC- (red line) and RTA-processed (blue line) samples. d, Quantum efficiency of the DC-processed sample as a function of the excitation laser power at 300 K (black dots) and 4 K (red squares), respectively. The intrinsic single-crystalline silicon wafer was first doped with erbium and oxygen by ion implantation. The Er/O-implanted silicon wafer was cut into two sets of samples that separately went through the DC and RTA procedure (see Methods). The secondary ion mass spectrometric (SIMS) measurements in Supplementary Fig. 1a show that the Er and O dopants are mostly located at ~80 nm below the surface with a maximum concentration of 2.8 × 1020 cm-3 and 1.4 × 1021 cm-3, respectively. Fig.1a shows the temperature-dependent photoluminescence (PL) of the DC-processed samples. The sharp luminescence at 1.54 μm with a narrow full width at half maximum (FWHM, ~25 nm) represents the characteristic 4f transition of Er3+.5-10 The temperature-dependent PL of the RTA-processed samples is shown in Supplementary Fig. 2a. For better comparison, the normalized peak amplitude as a function of temperature for these two sets of samples is plotted in Fig. 1b. As the temperature increases from 4 K to 300 K, the radiative emission is thermally quenched by nearly 3 orders of magnitude for the samples that were treated by the standard RTA method. In contrast, the radiative emission is only quenched by less than 1 order of magnitude for the samples that went through the DC process. This results in a two-orders-of-magnitude increase of PL at room temperature for the DC-processed samples in comparison with the RTA- processed ones, as shown in Fig. 1c. It is known that the photoexcited electron-hole pairs in silicon can recombine through three competing paths. The first path is radiative recombination and the other two paths are the nonradiative Shockley-Read-Hall (SRH) and Auger recombination.19 The SRH recombination can be suppressed by lowering the operating temperature. The Auger recombination is highly dependent on the concentration of photogenerated electron- hole pairs. As the excitation laser power is reduced, the Auger recombination will be minimized. In Fig. 1d, we recorded the relative quantum efficiency (QE) of the DC- processed sample at 4 K to suppress the SRH recombination (PL shown in the Supplementary Fig. 2b). When the excitation laser power is reduced to < 1 mW, the Auger recombination is minimized and the excited electron-hole pairs can only recombine radiatively via the 4f transition of the Er ions. In this case, the internal quantum efficiency (IQE) must saturate to 100% (red squares in Fig. 1d).20 At 300 K, the IQE also saturates as the excitation laser power is reduced (PL shown in the Supplementary Fig. 2c). But the saturated IQE is ~7 times smaller than the value at 4 K. It indicates that the IQE of our DC-processed samples can reach a record of ~14% at room temperature, approximately 2 orders of magnitude higher than the samples treated by the standard RTA process. It is expected that the IQE will be further improved by tuning the Er and O concentration and their ratio. Since the Er/O impurities are donor-type dopants in silicon, a PN junction diode can be formed simply by doping the sample with boron via ion implantation. The peak concentration of boron dopants is ~1019 cm-3 and the peak center is located at ~240 nm below the surface, approximately ~160 nm below the Er/O distribution peak (see Supplementary Fig. 1b). The DC process was applied to activate the Er/O and B dopants at the same time. A vertical PN junction diode was formed when a metal electrode was in contact with the underneath p-type silicon layer after the Er/O doped silicon layer is removed. The three-dimensional schematic and the cross-section of the device are depicted in Fig. 2a and b. An optical microscopic top view of the device is shown in the inset of Fig. 2c. The diode exhibits a rectifying current density v.s. voltage (J-V) curve (Fig. 2c). When it is forward biased, the diode has a strong emission at room temperature with the spectral characteristics of Er ions (Fig. 2d). As the injecting current ramps up, the emission peak at 1.54 μm increases rapidly. A near-infrared optical microscope (0.9 - 1.7 μm) was employed to record the device lightening-up process (see the setup in Supplementary Fig. 3 and the video in the Supplementary Video). The emission images of the device at the driving current of ~35 mA, ~50 mA and ~65 mA are shown in Fig. 2e, f and g, respectively. Figure 2 │ Optoelectronic properties of Er/O-Si LED devices treated by the deep cooling (DC) process. Three-dimensional schematic (a) and cross-sectional illustration (b) of the device. c, J-V curve and optical microscope image (inset) of the device. Room-temperature EL spectra (d) and near infrared microscopic images (e-g) of the device emission under electrical pumping. To understand the mechanism of the strong emission from the DC-processed sample, we used the high resolution transmission electron microscopy (HR-TEM) to examine the microstructure of the Er/O co-doped layer as shown in Fig. 3. It was previously reported that Er ions tend to aggregate during the slow cooling process of the standard RTA treatment.17, 18 The aggregation of Er ions forms efficient nonradiative clusters that strongly suppress the emission of Er ions at room temperature.5-10 The dramatic cooling of the DC process can freeze the dispersed Er dopants at high temperature (950 ℃) in an ultrashort time (possibly micro seconds), which helps prohibit the aggregation of Er ions. The cross-sectional TEM images show that Er nanocrystals as large as ~5 nm in diameter are clearly visible in the RTA-treated samples (Fig. 3a). The nanocrystals are more interconnected and aggregated. In the DC processed sample, Er nanocrystals are clearly smaller (< 1 nm) and more sparsely distributed as shown in Fig. 3b. The TEM image of a DC-processed intrinsic Si sample is shown in Fig. 3c for reference. This contrast is consistent with our expectation that the aggregation of Er ions is suppressed by the DC process, as a result of which the emission efficiency of Er/O co-doped Si samples is significantly increased at room temperature. The side effect of the DC process is that the thermal shock may create dislocations and cracks in the wafer surface.21 The X-ray diffraction (XRD) pattern (Supplementary Fig. 4) shows that the silicon wafer remains a nearly perfect crystalline structure after the wafer is treated with the DC process. No cracks were observed by visual inspection. Indeed, TEM images show that some dislocations were formed in the silicon substrate 150 nm below the surface (Supplementary Fig. 5). The impact of these dislocations on CMOS transistors and the Si LED will be further evaluated. If needed, new measures, for example, a long time of thermal treatment at relatively low temperature may be able to remove these dislocations. Figure 3 │ Microstructure characterization of the fabricated Er/O-Si samples. Cross-sectional HRTEM micrographs of (a) RTA- and (b) DC-processed Er/O-Si samples. The Er/O nanocrystal sizes distributed in a silicon matrix after RTA and DC treatments are approximately 5 nm (a) and 1 nm (b), respectively. c, Cross-sectional HRTEM micrograph of intrinsic single-crystalline silicon sample by the DC treatment was also manifest for comparison. Inset: Selected area electron diffraction (SAED) pattern of the intrinsic Si sample. In short, we presented a DC process to treat the Er/O co-doped single crystalline silicon wafers. After the treatment, the aggregation of Er ions is suppressed, as a result of which strong room-temperature photoluminescence at telecommunication wavelength was observed from the silicon samples. We further doped the samples with boron dopants and eventually realized an efficient room-temperature silicon LED device at 1.54 μm. Given the fact that these results were realized on single-crystalline silicon, it is promising to develop room-temperature silicon lasers for fully integrated silicon photonics in future.22 METHODS Preparation and characterization. FZ intrinsic Si (100) wafers (Resistivity: ≥ 10 kΩ·cm; Thickness: 500 ± 20 μm) were purchased from Suzhou Resemi Semiconductor Co., Ltd, China. The silicon wafers were first cleaned with ethanol, deionized water and then immersed in Piranha solution (Sulfuric acid : 30% Hydrogen peroxide = 3 : 1) for 30 min at 90 °C, followed by 2.5% HF etching. Erbium and Oxygen ions were implanted at the Institute of Semiconductors, Chinese Academy of Sciences in Beijing, China. The injection energy and dose of Er and O ions were 200 keV, 4 × 1015 cm-2, and 32 keV and 1016 cm-2, respectively. After the implantation, the Er/O- implanted silicon samples went through the aforementioned cleaning procedure again before the deposition of a 200-nm-thick SiO2 films by reactive magnetron sputtering (Delton multi-target magnetic control sputtering system, AEMD, SJTU). The deep cooling and standard RTA processes were applied to two Er/O co-doped silicon samples separately. After the capping SiO2 film was removed in buffered oxide etchant (BOE), PL characterizations were firstly performed. As shown in Supplementary Fig. 1a, SIMS technique (Evans Analytical Group, NJ, USA) was subsequently utilized to characterize Er and O dopant distribution profiles at the top 300 nm of silicon substrate. For Er element, the SIMS data are the summation of three Er isotopes (166, 167, and 168). For O element, no other isotopes were detected. The bump near the surface (blue line) likely comes from the contamination of native silicon oxide. To form a PN junction, boron dopants were later implanted into the Si substrate (see the section of LED fabrication and imaging). The SIMS profile of boron dopants is shown in Supplementary Fig. 1b. Deep cooling and RTA processes. The deep cooling process was performed on the Er/O-Si samples in an upgraded dilatometer (DIL 805A, TA Instruments) in which the samples were annealed at 950 °C for 5 min by means of copper coil-based electromagnetic heating, followed by a flush of high purity He (99.999%) gas cooled in liquid N2 (77 K). The RTA procedure for Er/O-Si samples was firstly conducted in a rapid thermal processing furnace (RTP-300) at 1050 ℃ for 5 min. After the high temperature treatment, the silicon samples were gradually cooled to room temperature with a high purity N2 gas. PL measurements. The PL measurements were performed using a Fourier transform infrared (FTIR) spectrometer (Vertex 80V, Bruker) equipped with a liquid-nitrogen-cooled Ge detector. A CW diode laser of λ = 405 nm (MLL-III-405, CNI, Changchun, China) providing a maximum excitation power of ~150 mW was employed as the excitation source. During the measurements, the Er/O-Si samples were mounted on the cold head of a Helium closed cycle cryostat (DE-202S, ARS), which allows for temperature adjustment from 4 K to ambient. (Fig. 1a and Supplementary Fig. 2a). For the excitation power dependent PL measurements, different optical filters including Notch filters, neutral density filters, and long-pass filters were utilized for avoiding the influence of the excitation source. The results are shown in Supplementary Fig. 2b and c. LED fabrication and imaging. To form a PN junction on the Er/O co-doped silicon wafers, boron ions were implanted (80 keV and 5.2 × 1014 cm-2) into the silicon wafer with a peak concentration located ~240 nm below the surface (Supplementary Fig. 1b). All dopants were simultaneously activated in the subsequent deep cooling process. Since the Er/O dopants (donors) are located in a shallower layer, a vertical PN junction is formed between the n-type Er/O layer and the p-type B doping layer. A pair of co-axial electrodes was formed by UV photolithography (MDA-400M, MIDAS), deep silicon etching (SPTS ICP), and metal film deposition (Nexdep, Angstrom Engineering Inc.). As shown in Fig.2a and b, the internal electrode is in contact with n-type Er/O doping region and the external electrode in contact with the p-type region after the top Er/O layer is removed by deep silicon etching. All the microfabrication processes were performed at the Center for Advanced Electronic Materials and Devices, Shanghai Jiao Tong University. After Al metal wire bonding (7476D, West Bond), the devices were integrated on a PCB board. The J-V curves of the Er/O-Si LED devices were measured using a digital sourcemeter (Keithley 2400) controlled by a Labview script. For EL measurements, the devices were electrically pumped by the sourcemeter and the emission spectra were collected using the same FTIR system for PL measurements. Near Infrared imaging of the emission from the Er/O-Si LED devices was obtained at room temperature under optical microscope (BX53M, Olympus) equipped with a NIR camera (C12471- 03, Hamamatsu). The setup is shown in the Supplementary Fig. 3. The recorded photographs and videos are shown in Fig. 2e, f, and g and Supplementary Video. Microstructure characterization. The microstructure of Er/O-Si samples were extensively investigated with TEM and XRD techniques. XRD patterns were acquired with a poly-functional X-ray diffractometer (D8 advance, Bruker) shown in Supplementary Fig. 4. In comparison, XRD patterns of the intrinsic and Er/O-implanted Si samples without any thermal treatment were also collected. Talos F200X field-emission TEM was employed to take the EDX and TEM images (Supplementary Fig. 5) of a sample slide that was cut from the samples using focused ion beam (FIB, GAIA3 GMU Model 2016). For EDX mapping measurements, the acceleration voltage was 200 kV, and other parameters were set as default. High resolution TEM images were further performed with JEOL JEM-ARM200F, as shown in Fig. 3. Data availability. The data that support the findings of this study are available from the authors on reasonable requests, see author contribution for specific data sets. 1. Atabaki, A. H. et al. Integrating photonics with silicon nanoelectronics for the next generation of systems on a chip. Nature 556, 349−354 (2018). 2. Subbaraman, H. et al. Recent advances in silicon-based passive and active optical interconncets. Opt. Express 23, 2487−2510 (2015). 3. Miller, D. A. B. Device requirements for optical interconnects to silicon chips. Proc. IEEE 97, 1166−1185 (2009). 4. Liang, D. & Bowers, J. E. Recent progress in lasers on silicon. Nat. Photonics 4, 511−517 (2010). 5. Ennen, H., Schneider, J., Pomrenke, G. & Axmann, A., 1.54-μm luminescence of erbium- implanted III-V semiconductors and silicon. Appl. Phys. Lett. 43, 943-945 (1983). 6. Lourenço, M. A., Milošević, M. M., Gorin, A., Gwilliam, R. M. & Homewood, K. P. Super- enhancement of 1.54 μm emission from erbium codoped with oxygen in silicon-on-insulator. Sci. Rep. 6, 37501 (2016). 7. Feklistov, K. V., Abramkin, D. S., Obodnikov, V. I. & Popov, V. P. Doping silicon with erbium by recoil implantation. Tech. Phys. Lett. 41, 788792 (2015). 8. Zhou, Z., Yin, B. & Michel, J. On-chip sources for silicon photonics. Light-Sci. Appl. 4, e358 (2015) 9. Kenyon, A. J. Erbium in silicon. Semicond. Sci. Technol. 20, R65−R84 (2005). 10. Polman, A. Erbium implanted thin film photonic materials. J. Appl. Phys. 82, 1−39 (1997). 11. Takahashi, Y. et al. A micrometre-scale Raman silicon laser with a microwatt threshold. Nature 498, 470−474 (2013). 12. Rong, H. et al. An all-silicon Raman laser. Nature 433, 292−294 (2005). 13. Sun, Y. et al. Room-temperature continuous-wave electrically injected InGaN-based laser directly grown on Si. Nat. Photonics 10, 595−599 (2016). 14. Chen, S. et al. Electrically pumped continuous-wave III-V quantum dot lasers on silicon. Nat. Photonics 10, 307−311 (2016). 15 Liu, J., Sun, X., Camacho-Aguilera, R., Kimerling, L. C. & Michel, J. Ge-on-Si laser operating at room temperature. Opt. Lett. 35, 679−681 (2010). 16. Wang, D.-C. et al. An all-silicon laser based on silicon nanocrystals with high optical gains. Sci. Bull. 63, 75−77 (2018). 17. Kuznetsov, V. P. et al. Crystal lattice defects and hall mobility of electrons in Si:Er/Si layers grown by sublimation molecular-beam epitaxy. Phys. Solid State 47, 102−105 (2005). 18. Scalese, S. et al. Effect of O:Er concentration ratio on the structural, electrical, and optical properties of Si:Er:O layers grown by molecular beam epitaxy. J. Appl. Phys. 88, 4091−4096 (2000). 19. Schroder, D. K. Carrier lifetimes in silicon. IEEE Trans. Electron Devices 44, 160−170 (1997). 20. Watanabe, S. et al. Internal quantum efficiency of highly-efficient InXGa1-XN-based near- ultraviolet light-emitting diodes. Appl. Phys. Lett. 83, 4906−4908 (2003). 21．Ng, W. L. et al. An efficient room-temperature silicon-based light-emitting diode. Nature 410, 192−194 (2001). 22. Thomson, D. et al. Roadmap on silicon photonics. J. Opt. 18, 073003 (2016). Acknowledgements This work was supported by the National Science Foundation of China (61376001, 21703140 and 61874072), the China Postdoctoral Science Foundation (2016M601582), and the special-key project of the "Innovative Research Plan", Shanghai Municipality Bureau of Education. We thank Jianming Li (Institute of Semiconductors, CAS) for the ion implantation processing and Yihua Chen (Shanghai Jiao Tong Univ.) for the deep cooling treatment. We also thank Xuecheng Fu (Shanghai Jiao Tong Univ.) for the SiO2 film deposition, Weiwei Wang (Spark Electro-Optics Co., Ltd.) and Lin Zhu (Cinv Co., Ltd) for the support in quantum efficiency and NIR imaging measurements. Author Contributions Y.D. conceived the deep cooling idea and designed the experiments. H.W., J.J.H., J.H., and F.Y. conducted the experiments. H.W. and Y.D. analysed the data and wrote the manuscript. All authors reviewed and commented on the manuscript. Competing interests The authors declare no competing interests. Corresponding authors Correspondence and requests for materials should be addressed to Yaping Dan ([email protected]) and Fangyu Ye ([email protected]). Author Information Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper. Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary Figures Supplementary Figure 1 SIMS characterizations of Er/O-Si samples. a, Erbium (red line) and Oxygen (blue line) ion distribution profiles of the DC-processed Er/O-Si samples (to 300 nm below surface) by SIMS measurements. b, Boron ion (black line) distribution profile of Er/O-Si LED devices. Supplementary Figure 2 Temperature- and excitation power-dependent PL spectra of Er/O- Si samples. a, Temperature-dependent PL spectra of the RTA-processed samples, which were measured as a comparison with the results in Fig. 1a. b, Cryogenic- (4 K), and c, room-temperature (300 K) PL spectra as a function of excitation power for the DC-processed samples. The insets show the corresponding relationships between peak intensity (black squares) or integral area (red dots) and the excitation power, derived from b and c, respectively. Supplementary Figure 3 Room-temperature NIR imaging setup. A NIR Olympus optical microscope equipped with a Hamamatsu InGaAs camera (0.9 - 1.7µm) was used to image the electrically-pumped emission (Keithley sourcemeter) from the Er/O-Si LED device that is treated with the deep cooling process. The optical image of PCB-integrated Er/O-Si LED devices was also displayed. Supplementary Figure 4 XRD characterization. XRD patterns of DC- (pink line) and RTA- processed (blue line) Er/O-Si samples. As comparison, the XRD results of intrinsic (red line) and Er/O-implanted (black line) silicon samples were also included in this figure. Supplementary Figure 5 TEM characterization. a, Er EDX mapping, and b, cross-sectional TEM image of the RTA-processed samples. c, Er EDX mapping, and d, cross-sectional TEM image of the DC-processed samples.
1704.06579	1	1704	2017-04-20T08:35:29	Self-supporting structure design in additive manufacturing through explicit topology optimization	[ "physics.app-ph" ]	One of the challenging issues in additive manufacturing (AM) oriented topology optimization is how to design structures that are self-supportive in a manufacture process without introducing additional supporting materials. In the present contribution, it is intended to resolve this problem under an explicit topology optimization framework where optimal structural topology can be found by optimizing a set of explicit geometry parameters. Two solution approaches established based on the Moving Morphable Components (MMC) and Moving Morphable Voids (MMV) frameworks, respectively, are proposed and some theoretical issues associated with AM oriented topology optimization are also analyzed. Numerical examples provided demonstrate the effectiveness of the proposed methods.	physics.app-ph	physics	Self-supporting structure design in additive manufacturing through explicit topology optimization Xu Guo1, Jianhua Zhou, Weisheng Zhang2, Zongliang Du, Chang Liu and Ying Liu State Key Laboratory of Structural Analysis for Industrial Equipment, Department of Engineering Mechanics, International Research Center for Computational Mechanics, Dalian University of Technology, Dalian, 116023, P.R. China Abstract One of the challenging issues in additive manufacturing (AM) oriented topology optimization is how to design structures that are self-supportive in a manufacture process without introducing additional supporting materials. In the present contribution, it is intended to resolve this problem under an explicit topology optimization framework where optimal structural topology can be found by optimizing a set of explicit geometry parameters. Two solution approaches established based on the Moving Morphable Components (MMC) and Moving Morphable Voids (MMV) frameworks, respectively, are proposed and some theoretical issues associated with AM oriented topology optimization are also analyzed. Numerical examples provided demonstrate the effectiveness of the proposed methods. Keywords: Additive manufacturing; Self-supporting structures; Moving Morphable Component (MMC); Moving Morphable Void (MMV); Topology optimization. 1Corresponding author. E-mail: [email protected] Tel: +86-411-84707807 2Corresponding author. E-mail: [email protected] Tel: +86-411-84707807 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 1 1. Introduction Additive manufacturing (AM, also known as 3D printing) is a relatively new manufacture technique which enables the fabrication of components in an additive (layer-by-layer) way. In contrast to traditional subtractive and formative manufacturing techniques (e.g., machining and casting), AM has intrinsic ability to build components with very complex structural geometries and highly optimized mechanical/physical functionalities. On the other hand, topology optimization, which aims at designing innovative and lightweight products by distributing material within a prescribed domain in an optimal way, has reached a certain level of maturity and becomes a well-established research area [1-3]. Up to now, many methods have been proposed for topology optimization and some of them have already been applied successfully in various application fields. Although topology optimization has great potential to become a perfect design tool that can fully exploit the tremendous design freedom provided by AM, it must be admitted that existing topology optimization approaches cannot be fully adapted to the current AM techniques. This is because although seemingly a free-form manufacturing technique, AM does have some design limitations which must be taken into account when AM oriented topology optimization approaches are devised for making topology optimization and AM an ideal fit. These limitations can be summarized briefly as follows (see also Fig. 1 for a schematic illustration). Firstly, since 3D printing is achieved either by depositing material (in a Fused Deposition Modeling, FDM), applying focused laser to cure powder (in Selective Laser Melting, SLM), spraying liquid binding onto particles (in Inkjet Printing, IP) or by using some combinations of these approaches, the achievable smallest print resolution will inevitably be influenced by the parameters such as the nozzle diameter (in FDM) or the beam width/offset in laser sintering [4]. Therefore, optimal designs with too small feature sizes (e.g., the wall thickness or the diameter of a void) may not be fabricated even with AM technique. Secondly, since in AM products are actually being built in a layer by layer way, each part of the products must be sufficiently supported from below during the manufacturing process otherwise the quality of the product cannot be fully guaranteed. Usually, support material must be introduced to manufacture certain structural topologies in order to prevent the structural material from being distorted too much or even fail (due to high bending stresses) during the build process. This Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 2 treatment, however, will waste print time, increase material usage and/or require chemical processing to remove the support material. It is therefore highly desirable to design structures that are self-supporting during the course of AM. Thirdly, the inclination of downward facing and non-supported (overhang) surfaces cannot be at too large an angle with respect to the build direction. The component with a high overhang angle may deform, droop or warp during the printing process. The upper bound of the overhang angle is obvious material and process-dependent and has been investigated extensively in literature. A commonly accepted value of the maximum overhang angle amounts to 40°-50° [5, 6]. Lastly, there should be no enclosed voids existing in the structure otherwise it will be very difficult to get the unmelted powder (in SLM) and support material (in FDM) out of the void once the structure is built up by AM. Based on the above discussions, it can be concluded that if topology optimization is expected to be used as a tool for innovative design purpose, the above issues must be taken into consideration in order to guarantee that the corresponding products/structures can be successfully manufactured with AM. The readers are also referred to Brackeet et al. [7, 8] for an overview of the corresponding issues in AM oriented topology optimization. Recent years witnessed an increasing interest to develop AM oriented topology optimization approaches by taking the aforementioned manufacturing constraints into consideration. In order to control the minimum length scale in topology optimization, both global implicit [9-11] and local explicit [12-17] methods are developed. We refer the readers to the above works and the references therein for more recent advances on this aspect. By introducing a virtual temperature field, Li et al. developed an approach which can deal with connectivity constraint in topology optimization and produce optimal designs without enclosed voids [18]. In order to tackle the issues of designing of self-supporting structures and controlling overhang features, Leary et al. [19] and Hu et al. [20] proposed the post-processing methods to ensure the printability without introducing additive support materials. This is achieved by modifying the theoretically optimal topologies to respect the overhang angle constraint. In [7], the authors suggested first identifying the angles of overhang parts in structures obtained by topology optimization and then achieving the overhang angle control by incorporating the corresponding constraints in optimization problem explicitly. No subsequent research work following this idea, however, can be found in literature. Langelaar [21, 22] proposed Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 3 an AM oriented topology optimization formulation to design self-supporting optimal structures without additional supporting materials. To this end, the author developed a filter scheme that can incorporate the main characteristics of a typical AM process and implemented it in a density based topology optimization approach. It was reported that fully self-supporting can be obtained with use of this approach. More recently, Gaynor and Guest [23] developed a topology optimization approach to produce self-supporting optimal designs. In this approach, a series of projection operators that can enforce the minimum length scale constrains and restrict the overhang angles are introduced under the variable density solution framework. Numerical examples showed that self-supporting optimal designs satisfying minimum length scale, overhang angle and volume constraints do can be obtained with use of this approach. Although numerous efforts have been made to resolve the aforementioned printable issues associated with AM processes, there is still room for further improvement. For example, the post-processing methods suggested in [19] and [20] will inevitably introduce extra computational efforts and more importantly destroy the optimality of the original optimized designs. Although the strategy of introducing additional structures to a previously optimized design [19, 20] does can make it self-supporting, the mass of the structure, however, will increase substantially and the structure after this treatment may deviate significantly from the originally optimized geometry [21]. The problem associated with the filter approach invented in [21] and [22] is that the filter functions introduced are highly nonlinear and dependent on the regular mesh used for finite element analysis. It is also found that there exist a relatively large number of elements with intermediate densities in some optimized designs provided. This phenomenon is obvious the side effect of the corresponding filter operation. It is also unclear how to optimize the build orientation in the proposed solution approach. For the projection approach developed in [23], as pointed out by the authors, since the determination of structural topology and the corresponding sensitivity analysis must be carried out in a layer-by-layer manner, the proposed approach is in general computationally inefficient. Furthermore, for some problems, the involvement of multiple embedded nonlinear functions in the solution scheme may also lead to convergence issues. Compared to adding extra material or carrying out post-processing to make an optimal structure printable, it is generally believed that designing self-supporting printable structures through topology Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 4 optimization method directly is more preferable since it can simplify the post-processing and reduce the manufacture cost. In the present work, we intend to discuss how to design self-supporting structures produced by AM without introducing support materials in a more explicit and geometrical way. Our motivation is from the consideration that since the constraints associated with the designing of self-supporting structures (e.g., overhang angle, minimum length scale) are actually geometrical in nature, it seems more appropriate to include more geometry information in the mathematical formulation of the considered problem and perform topology optimization in an geometrically explicit way (it is worth noting that topological design is usually achieved in an implicit way in tradition solution approaches such as variable density method and level set method, see [24-29] for more detailed discussions on this aspect). Recent years witnessed a growing interest in solving topology optimization problems by optimizing a set of geometrical parameters explicitly (i.e., a revival of shape optimization) [17, 24-29]. In particular, the so-called Moving Morphable Components (MMC) and Moving Morphable Voids (MMV) where a set of components (in MMC) or voids (in MMV) are used as basic building blocks of optimization have been developed. As demonstrated in the following sections, the MMC and MMV approaches can deal with the problem of designing self-supporting structures in a more natural way since more geometry information (e.g., the outward normal vector of structural boundary, the inclined angle of a structural component) is embedded in the corresponding problem formulations. In the present work, two approaches for designing AM oriented self-supporting structures established in the MMC and MMV solution framework are developed, respectively. The rest of the paper is organized as follows. In Section 2, the proposed two solution approaches are presented and analyzed in detail. Some theoretical issues associated with AM oriented topology optimization are discussed in Section 3. Numerical solution aspects are addressed in Section 4. In Section 5, some numerical examples are presented to demonstrate the effectiveness of the proposed methods. Finally some concluding remarks are provided in Section 6. 2. Two approaches for designing self-supporting structures In this section, two approaches for designing self-supporting structures established based on the Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 5 Moving Morphable Components (MMC) and Moving Morphable Voids (MMV), respectively, will be presented. As a preliminary attempt to address this difficulty problem, only two dimensional problems are considered in the present work. 2.1 MMC-based approach As discussed in the previous section, a critical issue in topology optimization of self-supporting structures is to control the inclined angles of structural components in the structure. Under this circumstance, the MMC-based explicit topology optimization approach first initialized in [24] where the inclined angles of each component are adopted as design variables, is a natural choice to serve this purpose. In the following, a brief introduction of the basic idea of MMC will be given first. As pointed out in [24], unlike in traditional topology optimization approaches where structural topologies are represented either by element densities (in variable density approach) or by nodal values of a level set function (in level set approach), in the MMC-based approach, a set of moving morphable components are adopted as basic building blocks of topology optimization (see Fig. 2 for a schematic illustration). These components are allowed to move, deform, overlap and merge in the design domain freely, and optimal structural topology can be obtained by optimizing the positions, inclined angles, lengths, widths and the layout of these components. This treatment provides a new paradigm for topology optimization and has big potential to resolve some challenging issues that cannot be dealt with easily by traditional methods. We refer the readers to [24, 25, 27-29] for more details on the generalization and variants of this approach. In the present work, as a preliminary attempt to generate self-supporting structures through topology optimization, we propose to use structural components with hyperelliptic shapes as building blocks of topology optimization [24]. Under this circumstance, the corresponding topology optimization problem can be formulated as: Find 𝑫𝑫=((𝑫𝑫1)⊤,…,(𝑫𝑫𝑛𝑛𝑛𝑛)⊤)⊤, 𝛼𝛼 and 𝒖𝒖 Minimize 𝐼𝐼=𝐼𝐼(𝑫𝑫,𝒖𝒖) S.t. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 6 In Eq. (2.1) and with =�𝐻𝐻(𝜒𝜒𝑠𝑠)𝒇𝒇∙𝒗𝒗dV �𝐻𝐻(𝜒𝜒𝑠𝑠)𝔼𝔼𝑠𝑠:𝜺𝜺(𝒖𝒖):𝜺𝜺(𝒗𝒗)dV D D +�𝒕𝒕∙𝒗𝒗dS, ∀𝒗𝒗∈𝒰𝒰ad , Γ𝑡𝑡 ≤𝑉𝑉�meas(D), 𝑉𝑉=�𝐻𝐻(𝜒𝜒𝑠𝑠)dV D (sin(𝜃𝜃𝑘𝑘+𝛼𝛼))2≥(sin(𝜃𝜃))2, 𝑘𝑘=1,…,𝑛𝑛𝑛𝑛, 𝑫𝑫⊂𝒰𝒰𝑫𝑫, 0≤𝛼𝛼≤𝛼𝛼≤𝛼𝛼≤𝜋𝜋/2, 𝒖𝒖=𝒖𝒖, on Γu. (2.1) 𝜒𝜒𝑠𝑠(𝒙𝒙)=max�𝜒𝜒1(𝒙𝒙),𝜒𝜒2(𝒙𝒙),⋯,𝜒𝜒𝑛𝑛𝑛𝑛(𝒙𝒙)� (2.2) is the topology description function (TDF) of the region occupied by the components and 𝜒𝜒𝑘𝑘 (𝑘𝑘= 1,⋯,𝑛𝑛𝑛𝑛) denotes the TDF of the region occupied by the 𝑘𝑘-th component (i.e., Ω𝑘𝑘), that is, �𝜒𝜒𝑘𝑘(𝒙𝒙)>0, if 𝒙𝒙∈Ω𝑘𝑘, if 𝒙𝒙∈𝐷𝐷\Ω𝑘𝑘. (2.3) 𝜒𝜒𝑘𝑘(𝒙𝒙)=0, if 𝒙𝒙∈𝜕𝜕Ω𝑘𝑘, 𝜒𝜒𝑘𝑘(𝒙𝒙)<0, 𝜒𝜒𝑘𝑘(𝑥𝑥,𝑦𝑦)=1−�𝑥𝑥′𝐿𝐿𝑘𝑘�𝑝𝑝−�𝑦𝑦′𝑡𝑡𝑘𝑘�𝑝𝑝, (2.4) cos𝜃𝜃𝑘𝑘��𝑥𝑥−𝑥𝑥0𝑘𝑘 sin𝜃𝜃𝑘𝑘 �𝑥𝑥′𝑦𝑦′�=� cos𝜃𝜃𝑘𝑘 𝑦𝑦−𝑦𝑦0𝑘𝑘�, (2.5) −sin𝜃𝜃𝑘𝑘 where 𝑝𝑝 is a relatively large even number (we take 𝑝𝑝=6 in the present study). Actually Eq. (2.4) represents a hyperelliptic shape component centered at point �𝑥𝑥0𝑘𝑘,𝑦𝑦0𝑘𝑘� with a half-length 𝐿𝐿𝑘𝑘, a half-thickness 𝑡𝑡𝑘𝑘 and a inclined angle 𝜃𝜃𝑘𝑘(with respect to the horizontal axis). The vector of design variables associated with the 𝑘𝑘 -th component is 𝑫𝑫𝑘𝑘=�𝑥𝑥0𝑘𝑘,𝑦𝑦0𝑘𝑘,𝐿𝐿𝑘𝑘,𝑡𝑡𝑘𝑘,𝜃𝜃𝑘𝑘�⊤ (see Fig. 3 for if 𝒙𝒙∈Ω𝑠𝑠, if 𝒙𝒙∈D\Ω𝑠𝑠, (2.6) if 𝒙𝒙∈𝜕𝜕Ω𝑠𝑠, �𝜒𝜒𝑠𝑠(𝒙𝒙)>0, 𝜒𝜒𝑠𝑠(𝒙𝒙)=0, 𝜒𝜒𝑠𝑠(𝒙𝒙)<0, Computer Methods in Applied Mechanics and Engineering, under review reference). Obviously, 2016-11-12 7 represents the region occupied by the structural components. In Eq. (2.1), Ω𝑘𝑘 corresponding test represent the symmetrized fourth and the second order identity tensor, respectively) is the fourth function defined on Ω=⋃ 𝑛𝑛𝑛𝑛𝑘𝑘=1 Ω𝑘𝑘 where Ω𝑠𝑠=⋃ 𝑛𝑛𝑛𝑛𝑘𝑘=1 the symbol 𝑛𝑛𝑛𝑛 denotes the total number of components in the design domain. 𝒰𝒰𝑫𝑫 is the admissible set that 𝑫𝑫 belongs to. The symbols 𝒖𝒖 and 𝒗𝒗 are the displacement field and the with 𝒰𝒰ad={𝒗𝒗 𝒗𝒗∈𝐇𝐇1(Ω),𝒗𝒗= 𝟎𝟎 on Γu}. The symbol meas(D) stands for the measure of the design domain D. The symbols 𝒇𝒇 and 𝒕𝒕 denote the body force density and the surface traction on Neumann boundary Γt , respectively. 𝒖𝒖� is the prescribed displacement on Dirichlet boundary Γu. The symbol 𝜺𝜺 denotes the second order linear strain tensor. 𝔼𝔼𝑆𝑆=𝐸𝐸𝑠𝑠/(1+𝜈𝜈𝑠𝑠)[𝕀𝕀+𝜈𝜈𝑠𝑠/(1−2𝜈𝜈𝑠𝑠)𝛅𝛅⨂𝛅𝛅] (𝕀𝕀 and 𝛅𝛅 order isotropic elasticity tensor of the solid material with 𝐸𝐸𝑠𝑠 and 𝜈𝜈𝑠𝑠 denoting the corresponding Young's modulus and Poisson's ratio, respectively. The symbol 𝑉𝑉� denotes the upper bound of the relative available volume of solid material and 𝐻𝐻=𝐻𝐻(𝑥𝑥) in Eq. (2.1) is the Heaviside function. (sin(𝜃𝜃𝑘𝑘+𝛼𝛼))2≥(sin(𝜃𝜃))2,𝑘𝑘=1,…,𝑛𝑛𝑛𝑛 where 𝛼𝛼 is the rotation angle of the work plane (see Fig. 4 for a schematic illustration) and 𝜃𝜃 is the lower bound of the overhang angle (as discussed before usually 𝜃𝜃∈[40°𝜋𝜋/180°,50°𝜋𝜋/180°]), respectively. These constraints are included to ensure that the inclined angles of components will not be less than the critical value (i.e., 𝜃𝜃) which can make the It is worth noting that compared to MMC-based topology optimization formulations where no self-supporting requirement is considered, the only extra constraints in the present formulation is components self-supportable. It is also worth noting that including the angle of working plane as a design variable increases the design freedom substantially. As shown in Fig. 4 and the examples in the following section, some optimal structures which are not printable for a specific working plane can be produced without any difficulty if the angle of working plane can be selected appropriately. As can be seen from Eq. (2.1), the advantage of the present formulation is that the self-supporting requirement can be dealt with by introducing a set of geometry constraints explicitly. Furthermore, both the structural topology and the angle of working plane can be optimized in a simultaneous way. It is, however, should be pointed out that this formulation has a deficiency such that it cannot totally exclude the unprintable case of V-shape material distribution as shown in Fig. 5. As plotted in Fig. 5, the two components actually have no supports from below and therefore cannot be printed even though their inclined angles are far beyond the threshold value. An accompanying Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 8 case is also shown in Fig. 5 where a set of overlapping components with high inclined angles may constitute an unprintable shallow overhang part of a structure. Although these cases are hypercritical and may be avoided to a large extent by optimizing the rotation angle of the working plane, a theoretically complete way to eliminate these unpleasant cases is to introduce the following pointwise supportable constraint and inclined angle constraint into the problem formulation: meas �Ωs∩C𝜖𝜖(𝒙𝒙)�>𝛿𝛿, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠, (2.7𝑎𝑎) 𝒏𝒏∙𝒃𝒃𝑝𝑝≤cos𝜃𝜃, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠, (2.7𝑏𝑏) where C𝜖𝜖(𝒙𝒙) represents a semicircle of radius 𝜖𝜖 center on 𝒙𝒙 and 𝜖𝜖 as well as 𝛿𝛿 are two small positive values. The symbols 𝒏𝒏 is the inward normal vector of 𝜕𝜕Ω𝑠𝑠 and 𝒃𝒃𝑝𝑝 is the unit vector meas �Ω𝑠𝑠∩C𝜖𝜖(𝒙𝒙)�=∫𝐻𝐻�min�𝜒𝜒𝑠𝑠(𝒙𝒙),𝜒𝜒C𝜖𝜖(𝒙𝒙)��dV, where 𝜒𝜒C𝜖𝜖 is the TDF of the region occupied by C𝜖𝜖(𝒙𝒙)), we do not intend to discuss this treatment in detail in the present study. In representing the print direction. We refer the authors to Fig. 6 for a schematic illustration of the implication of these geometrical constraints. order to circumvent the aforementioned difficulties, another approach to generate self-supporting printable structures based on the MMV framework will be described in the next subsection. Although the constraint in Eq. (2.7) is indeed numerically implementable (actually and 𝐷𝐷 2.2 MMV-based approach In this section, we shall discuss how to design self-supporting structures suitable for AM with use of topology optimization under the so-called Moving Morphable Voids (MMV) framework through explicit boundary evolution. The central idea to introduce printable features (voids), whose boundaries can be described explicitly by a set of B-spline curves, into problem formulation and transfer the corresponding topology optimization into a shape optimization problem. As shown in Fig. 7, in the MMV-based approach, the topological change of a two dimensional structure is achieved by the deformation, intersection and merging of a set of closed parametric curves (i.e., 𝐶𝐶𝑖𝑖,𝑖𝑖=1,…,𝑛𝑛𝑣𝑣) which represent the interior boundary of the structure (i.e., boundaries of a set of voids). Unlike the traditional level set approach (which is also a boundary-based approach for topology optimization), Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 9 in the MMV-based approach, there is no need to introduce an extra level set function defined in a higher dimensional space to represent the structural boundary implicitly. The design variables involved in the MMV-based approach are only the coordinates of the control points (or the related parameters) of the parametric curves. The readers are referred to [29] for more discussions on technical details of this approach. 𝑛𝑛 𝑘𝑘=0 with and In order to circumvent the problems associated with the MMC-based approach discussed in the previous subsection, we suggest introducing a set of printable voids with explicit boundary representation as the basic building blocks of topology optimization. In the present approach, B-spline curve expressed in Eq. (2.8) is used to describe the shape of each structural component (see Fig. 8 for reference): 𝑪𝑪(𝑢𝑢)=�𝑁𝑁𝑘𝑘,𝑝𝑝(𝑢𝑢)𝑷𝑷𝑘𝑘 , 𝑎𝑎≤𝑢𝑢≤𝑏𝑏, (2.8𝑎𝑎) 𝑷𝑷𝑖𝑖=�𝑥𝑥𝑛𝑛+sin�(𝑖𝑖−1)𝜃𝜃+𝜋𝜋2�𝑑𝑑𝑖𝑖,𝑦𝑦𝑛𝑛+cos�(𝑖𝑖−1)𝜃𝜃+𝜋𝜋2�𝑑𝑑𝑖𝑖�⊤, 𝜃𝜃= 𝜋𝜋𝑛𝑛−2, 𝑖𝑖=1,…,𝑛𝑛−1, (2.8𝑏𝑏) 𝑷𝑷0=𝑷𝑷𝑛𝑛=(𝑥𝑥𝑛𝑛,𝑦𝑦𝑛𝑛+𝑑𝑑0)⊤. (2.8𝑛𝑛) The meanings of 𝑥𝑥𝑛𝑛,𝑦𝑦𝑛𝑛 and 𝑑𝑑𝑖𝑖≥0,𝑖𝑖=0,…,𝑛𝑛 are self-evident from Fig. 8. As pointed out in [29], definitely be respected once the condition 𝑑𝑑0/𝑑𝑑1≥tan𝜃𝜃 and 𝑑𝑑0/𝑑𝑑3≥tan𝜃𝜃 are satisfied. Fig. 9 under this treatment, it can always guarantee that there is no cusp on the B-spline curve and the curve cannot be self-intersected. It can be seen clearly from Fig. 8 that the printability requirement can also plots some other printable features whose shape can be represented by the B-spline curved expressed in Eq. (2.8). It can be seen from this figure that printable features with fairly complex shapes do can be represented by the proposed geometry representation scheme. Furthermore, it is also worth noting that the V-shape issue mentioned previously, which cannot be account for by only imposing the inclined angle constraint (i.e., 𝒏𝒏∙𝒃𝒃𝑝𝑝≤cos𝜃𝜃) can be dealt with in an easy way by introducing the printable features. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 10 In the present study, in order to preserve the printability of each the void in the structure, it is fully respected as shown in Fig. 10. Furthermore, we also need a B-spline curve to describe the required that every two voids cannot be intersected (i.e., ΩV𝑖𝑖∩ΩV𝑗𝑗=∅ where ΩV𝑖𝑖 and ΩV𝑗𝑗 are the regions enclosed by 𝐶𝐶𝑖𝑖 and 𝐶𝐶𝑗𝑗, respectively) otherwise the printability requirement may not be exterior boundary of the structure (i.e., 𝐶𝐶0∩D in Fig. 10). In the present study, it is assumed that the central point of 𝐶𝐶0 (i.e., 𝑥𝑥𝑛𝑛0,𝑦𝑦𝑛𝑛0) is fixed, only 𝑑𝑑10,…,𝑑𝑑𝑛𝑛00 are adopted as design variables. Obviously, the printability condition associated with the exterior boundary is 𝒏𝒏𝑛𝑛0∙𝒃𝒃𝑝𝑝≤ cos𝜃𝜃,∀ 𝒙𝒙∈𝐶𝐶0∩𝐷𝐷, where 𝒏𝒏𝑛𝑛0 is the inward normal vector of 𝐶𝐶0 and 𝒃𝒃𝑝𝑝 is the unit vector boundary curves 𝐶𝐶𝑖𝑖 and 𝐶𝐶𝑗𝑗 (𝑖𝑖≠𝑗𝑗) cannot be intersected with each other, they, however, all can be intersected with the exterior boundary curve 𝐶𝐶0 as shown in Fig. 10. At this position, it is worth noting that if only one control point at the lower part of a B-spline curve is used to construct an interior boundary, it can always guarantee that the corresponding void is printable. If, however, more control points are introduced to describe the shape of the interior boundary, as shown in Fig. 11a, it is possible that the corresponding enclosed void is not printable. representing the print direction. It is also worth noting that even though the every pair of interior This problem can be resolved by requiring that 𝑥𝑥𝑖𝑖≥𝑥𝑥𝑖𝑖+1, and 𝑦𝑦𝑖𝑖≤𝑦𝑦1−𝑦𝑦0 𝑥𝑥1−𝑥𝑥0(𝑥𝑥𝑖𝑖−𝑥𝑥0)+𝑦𝑦0 as well 𝑥𝑥𝑛𝑛−1−𝑥𝑥0(𝑥𝑥𝑖𝑖−𝑥𝑥0)+𝑦𝑦0, for all 𝑖𝑖=2,…,𝑛𝑛−2, respectively (see Fig. 11b for reference). In as 𝑦𝑦𝑖𝑖≤𝑦𝑦𝑛𝑛−1−𝑦𝑦0 the above statement 𝑥𝑥𝑖𝑖 and 𝑦𝑦𝑖𝑖,𝑖𝑖=2,…,𝑛𝑛−1 are the horizontal and vertical coordinates of the control points of the corresponding B-spline curve. These constraints are linear in nature and thus easy to be dealt with in numerical implementation. Based on the above discussions, the problem formulation for design self-supporting structures under MMV framework can be written as Find 𝑫𝑫=((𝑫𝑫0)⊤,(𝑫𝑫1)⊤,…,(𝑫𝑫𝑛𝑛𝑣𝑣)⊤)⊤ Minimize 𝐼𝐼=𝐼𝐼(𝑫𝑫) S.t. meas�(D∩ΩV0)\∪𝑖𝑖=1𝑛𝑛𝑣𝑣 ΩV𝑖𝑖�≤𝑉𝑉�meas(D), ΩV𝑖𝑖∩ΩV𝑗𝑗=∅, 𝑖𝑖≠𝑗𝑗,𝑖𝑖,𝑗𝑗=1,⋯,𝑛𝑛𝑛𝑛, Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 11 𝑖𝑖=1,…𝑛𝑛𝑛𝑛, 𝑖𝑖=1,…𝑛𝑛𝑛𝑛, ∀ 𝒙𝒙∈𝐶𝐶0∩D, 𝑑𝑑0𝑖𝑖/𝑑𝑑1𝑖𝑖≥tan𝜃𝜃, 𝑑𝑑0𝑖𝑖/𝑑𝑑𝑛𝑛−1𝑖𝑖 ≥tan𝜃𝜃, 𝒏𝒏𝑛𝑛0∙𝒃𝒃𝑝𝑝≤cos𝜃𝜃, 𝑫𝑫⊂𝒰𝒰𝑫𝑫, (2.9) where 𝑫𝑫𝑖𝑖=�𝑥𝑥𝑛𝑛𝑖𝑖,𝑦𝑦𝑛𝑛𝑖𝑖,𝑑𝑑0𝑖𝑖,…,𝑑𝑑𝑛𝑛𝑖𝑖−1 �⊤,𝑖𝑖=0,…,𝑛𝑛𝑛𝑛 denote the vector of design variables associated 𝑖𝑖 with 𝐶𝐶𝑖𝑖, resepectively. The symbol meas�(D∩ΩV0)\∪𝑖𝑖=1𝑛𝑛𝑣𝑣 ΩV𝑖𝑖� represents the measure of the region (D∩ΩV0)\∪𝑖𝑖=1𝑛𝑛𝑣𝑣 ΩV𝑖𝑖 (i.e., the solid part of the structure), where ΩV0 represents the solid region occupied by 𝐶𝐶0. 3．The optimality of a self-supporting structure-some theoretical considerations Although optimal design of self-supporting structures has been investigated intensively in literature, it is still an open question what the optimal structure should like when self-supporting requirement is taken into consideration. In this section, the optimality of a self-supporting structure will be discussed from a theoretical point of view. Here the objective and constraint functionals are taken as the compliance and the volume of solid material, respectively. Assuming that an optimal structure obtained with topology optimization without considering self-supporting requirement takes the form shown in Fig. 12a. It can be conjectured that the corresponding optimal self-supporting solution can be constructed from it as the way shown in Fig. 12b. The key point of this treatment is that we can introduce supporting materials with infinitesimal small amount of volume and stiffness to satisfy the self-supporting requirement without degrading the stiffness of the structure. This can be explained as follows. rectangles as shown in Fig. 12c. If we reinforce each void using supporting material as the way shown in Fig. 12d, it can be confirmed that the resulting structure is obvious self-supporting. Under this circumstance, it can be estimated that the total volume of the support material in each void is Without loss of generality, assuming that all void parts of an optimal structure are 𝑙𝑙×1 V𝑠𝑠=𝑙𝑙𝛿𝛿�12𝛿𝛿2+𝛿𝛿2(1−𝛿𝛿)�=32𝛿𝛿𝑙𝑙−𝛿𝛿2𝑙𝑙∝𝑂𝑂(𝛿𝛿). (3.1𝑎𝑎) Furthermore, the Voigt upper bound of the modulus of the equivalent elasticity tensor of the Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 12 reinforced void can be estimated as ‖ℂ�𝑠𝑠‖≅V𝑠𝑠VE𝑠𝑠=�32𝛿𝛿𝑙𝑙−𝛿𝛿2𝑙𝑙� 𝑙𝑙×1 (original structure), respectively. Since the structural compliance is weakly continuous with respect E𝑠𝑠=�32𝛿𝛿−𝛿𝛿2�E𝑠𝑠∝𝑂𝑂(𝛿𝛿E𝑠𝑠). (3.1𝑏𝑏) From Eq. (3.1), it yields that V𝑠𝑠→0 and ‖ℂ�𝑠𝑠‖→0 as 𝛿𝛿→0. Therefore it can be concluded that V𝛿𝛿→V0,𝒖𝒖𝛿𝛿 𝑤𝑤→𝒖𝒖0 (weak convergence) as 𝛿𝛿→0 (in an appropriate function space, e.g., H1(D), where V𝛿𝛿(V0) and 𝒖𝒖𝛿𝛿(𝒖𝒖0) denote the total volume and displacement field of the reinforced structure to the displacement field, then it yields that 𝐼𝐼𝛿𝛿→𝐼𝐼0 as 𝛿𝛿→0, where 𝐼𝐼𝛿𝛿 and 𝐼𝐼0 denote the compliance of the reinforced structure and original structure, respectively. Although the above conclusions are obtained under the assumption that the voids in the structure are of rectangle shapes, they still hold when the structure has curved interior and exterior boundaries although the corresponding rigorous mathematical proof needs more technical treatments. It is also worth noting that the above treatment for obtaining self-supporting structures is not unique, other treatments (for example distributing the support material in a hierarchical way as shown in Fig. 12e) are also applicable and the essential feature of the corresponding mathematical analysis is the same as that of the analysis presented above. From the above analysis, it seems appropriate to conjecture that the theoretically optimal self-supporting structure can be constructed from the corresponding optimal structure without considering the self-supporting requirement by introducing infinitely many rods with infinitely small cross sections as additional supporting structures. If no regularization technique is introduced (e.g., imposing minimum length scale constraint or total perimeter constraint, etc.), it can be expected that the numerical solution results may be highly mesh-dependent if the solution algorithms are smart and robust enough to find global optimal solutions under prescribed finite element mesh. This may also explain why regions with intermediate densities are prone to exist when variable density approach are employed to design self-supporting structures [21, 23]. Of course, if regularization formulations/techniques are introduced in prior in problem formulation or employed in the numerical solution process, these "chattering designs" can definitely be suppressed. Although the above theoretical consideration is not mathematically rigorous, it provides useful insight into the problem Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 13 under consideration and may give an estimation on the lower bound of the optimal value of the objective functional. 4. Numerical solution aspects In this section, we shall discuss several relevant issues for the numerical implementation of the proposed approaches. 4.1 Finite element analysis In the present study, the X-FEM method proposed in [30] is employed to carry out structural response analysis based on fixed finite element (FE) mesh. Under this circumstance, it is necessary in the following way (see Fig. 13 for reference). Firstly, the MATLAB function Inpolygon and the explicit expressions of the B-spline boundary curves can be used to determine whether a FE node to construct a topological description function (TDF) 𝜒𝜒𝑠𝑠=𝜒𝜒𝑠𝑠(𝒙𝒙) characterizing solid part of the structure implicitly. For the MMC-based approach, 𝜒𝜒𝑠𝑠 can be constructed easily since the TDF of each component (i.e., 𝜒𝜒𝑘𝑘,𝑘𝑘=1,…,𝑛𝑛𝑛𝑛 ) has closed form analytical expression and 𝜒𝜒𝑠𝑠(𝒙𝒙)= max�𝜒𝜒1(𝒙𝒙),𝜒𝜒2(𝒙𝒙),⋯,𝜒𝜒𝑛𝑛𝑛𝑛(𝒙𝒙)�. For the MMV-based approach, the TDF of 𝜒𝜒𝑠𝑠 can be constructed (e.g., 𝑁𝑁𝑖𝑖) is inside or outside the region Ω𝑗𝑗 enclosed by a specific B-spline boundary curve 𝐶𝐶𝑗𝑗. If 𝑁𝑁𝑖𝑖 is inside Ω𝑗𝑗 then let IN(𝑁𝑁𝑖𝑖)=1 otherwise let IN(𝑁𝑁𝑖𝑖)=0. Secondly, calculating the minimum distance 𝑑𝑑𝑖𝑖 of 𝑁𝑁𝑖𝑖 to 𝐶𝐶𝑗𝑗 and define 𝜒𝜒𝑗𝑗(𝑁𝑁𝑖𝑖)=𝑑𝑑𝑖𝑖 if IN(𝑁𝑁𝑖𝑖)=1 otherwise define 𝜒𝜒𝑗𝑗(𝑁𝑁𝑖𝑖)=−𝑑𝑑𝑖𝑖. Repeating this procedure until all 𝜒𝜒𝑗𝑗(𝑁𝑁𝑖𝑖),𝑗𝑗=0,…,𝑛𝑛𝑛𝑛 are obtained. Finally, it can be calculated that 𝜒𝜒𝑠𝑠(𝑁𝑁𝑖𝑖)=max�𝜒𝜒1(𝑁𝑁𝑖𝑖),…,𝜒𝜒𝑛𝑛𝑣𝑣(𝑁𝑁𝑖𝑖)�. Once the value of 𝜒𝜒𝑠𝑠 on every FE node is calculated, then one can carry out the X-FEM analysis following the same way as in [27, 28]. It is also worth noting that if structural components with more complex shapes are introduced in problem formulation, the more general approach developed in [28] can be employed to construct the corresponding TDFs. 4.2 Sensitivity analysis Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 14 For both that the compliance of the assumption the MMC and MMV-based approaches, under 𝜕𝜕𝐼𝐼𝜕𝜕𝑑𝑑𝑖𝑖=� 𝑓𝑓(𝒖𝒖,𝒘𝒘)𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖d𝑆𝑆 𝜕𝜕Ω𝑠𝑠 objective/constraint functional 𝐼𝐼 is differentiable with respect to design variables, the general form of the partial derivative of 𝐼𝐼 with respect to a design variable 𝑑𝑑𝑖𝑖 can be expressed as follow , (4.1) where 𝒖𝒖 and 𝒘𝒘 are the so-called primary and adjoint fields, respectively. In Eq. (4.1), the integration is performed along the movable part of the boundary of the structure. When 𝐼𝐼 is the the structure we have 𝒖𝒖=𝒘𝒘 and 𝑓𝑓(𝒖𝒖,𝒘𝒘)=𝑓𝑓(𝒖𝒖,𝒘𝒘)=−𝔼𝔼𝑠𝑠:𝜺𝜺(𝒖𝒖):𝜺𝜺(𝒖𝒖)= −𝐸𝐸𝑖𝑖𝑗𝑗𝑘𝑘𝑘𝑘𝑠𝑠 𝑢𝑢𝑖𝑖,𝑗𝑗𝑢𝑢𝑘𝑘,𝑘𝑘 where 𝔼𝔼𝑠𝑠,𝒖𝒖 and 𝜺𝜺(𝒖𝒖)=sym∇(𝒖𝒖) denote the fourth-order elasticity tensor of the solid material, the displacement field and the strain tensor field, respectively. When 𝐼𝐼 represents the volume of the solid material, we have 𝑓𝑓(𝒖𝒖,𝒘𝒘)=1. In Eq. (4.1), 𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖 denotes the variation of the boundary along the outward normal direction due to the variation of 𝑑𝑑𝑖𝑖 (i.e., 𝛿𝛿𝑑𝑑𝑖𝑖). The expression of 𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖 can be found in the Appendix. We also refer the readers to [24, 27-29] for more details on calculating 𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖 from the variation of 𝑑𝑑𝑖𝑖 in the MMC and MMV-based approaches, respectively. In the MMV-based approach, the constraint 𝒏𝒏𝑛𝑛0∙𝒃𝒃𝑝𝑝≤cos𝜃𝜃,∀ 𝒙𝒙∈𝐶𝐶0∩D is actually pointwise in nature. In order to deal with this constraint, we discretize 𝐶𝐶0 with a set of points 𝒙𝒙𝑗𝑗∈ 𝐶𝐶0∩D,𝑗𝑗=1,…,𝑛𝑛 and impose the constraints on these points (i.e., 𝑔𝑔𝑖𝑖=𝒏𝒏𝑖𝑖𝑛𝑛0∙𝒃𝒃𝑝𝑝−cos𝜃𝜃≤0,𝑖𝑖= 1,…,𝑛𝑛). For the sake of reducing the computational effort, these constraints are aggregate into an equivalent single global constraint function as in [17]. Actually, 𝑔𝑔𝑖𝑖=𝒏𝒏𝑖𝑖𝑛𝑛0∙𝒃𝒃𝑝𝑝−cos𝜃𝜃≤0,𝑖𝑖= 1,…,𝑛𝑛 is fully equivalent to ℋ(𝑔𝑔1,…,𝑔𝑔𝑛𝑛)≤0, (4.2) 𝑛𝑛 ℋ(𝑔𝑔1,…,𝑔𝑔𝑛𝑛)=��𝑔𝑔𝑖𝑖2, (4.3) 0, 𝑖𝑖=1 if 𝑔𝑔𝑖𝑖>0, otherwise. 4.3 Treatment of geometry constraints to other approximation based Compared and Kreisselmeier-Steinhauser (K-S) function approach), the advantage of this treatment is that there is no artificial parameter involved in this aggregation scheme and thus can make the original (e.g., P-norm approaches approach where Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 15 constraints being satisfied exactly. Another type of constraints in the MMV-based approach is the non-intersection constraint for further refined in [32], the aforementioned non-intersection constraint can be represented in a single mathematical expression analytically as illustration). This treatment can not only enhance the robustness of the solution process but also control the minimum length scale of the structure in an implicit way. In the present work, we take any two different printable features. It is require that ΩV𝑖𝑖∩ΩV𝑗𝑗=∅,𝑖𝑖≠𝑗𝑗,𝑖𝑖,𝑗𝑗=1,⋯,𝑛𝑛𝑛𝑛 where ΩV𝑖𝑖 and ΩV𝑗𝑗 are the regions enclosed by 𝐶𝐶𝑖𝑖 and 𝐶𝐶𝑗𝑗, respectively. As first suggested in [31] and �𝐻𝐻(min (𝜒𝜒𝑖𝑖(𝒙𝒙),𝜒𝜒𝑗𝑗(𝒙𝒙))dV =0, (4.4) 𝐷𝐷 where 𝜒𝜒𝑖𝑖(𝒙𝒙) and 𝜒𝜒𝑗𝑗(𝒙𝒙) denote the TDF of ΩV𝑖𝑖 and ΩV𝑗𝑗, respectively. In numerical implementation, 𝜒𝜒𝑖𝑖(𝒙𝒙) and 𝜒𝜒𝑗𝑗(𝒙𝒙) in Eq. (4.4) are replaced by 𝜒𝜒𝑖𝑖−𝛿𝛿(𝒙𝒙) and 𝜒𝜒𝑗𝑗−𝛿𝛿(𝒙𝒙) which represent the TDF of a δ-expension of ΩV𝑖𝑖 and ΩV𝑗𝑗, respectively (see Fig. 14 for a schematic 𝛿𝛿=min(Δ𝑥𝑥,Δ𝑦𝑦) with Δ𝑥𝑥 and Δ𝑦𝑦 denoting the mesh size along two coordinate directions. Furthermore, the derivative of 𝜒𝜒𝑖𝑖(𝒙𝒙) with respect to design variables can be obtained by a finite For the MMV-based approach, the admissible set 𝒰𝒰𝑫𝑫 that geometry design variables belongs 𝑥𝑥𝑛𝑛𝑖𝑖≤0 if ΩV𝑖𝑖∩D𝐿𝐿≠∅ and 𝑥𝑥𝑛𝑛𝑖𝑖≥𝐿𝐿 if ΩV𝑖𝑖∩D𝑅𝑅≠∅,∀ 𝑖𝑖=1,…,𝑛𝑛𝑛𝑛, (4.5) It is also worth noting that the conditions ΩV𝑖𝑖∩D𝐿𝐿≠∅ and ΩV𝑖𝑖∩D𝑅𝑅≠∅ can also be expressed in the following analytical form � >0 (4.6a) 𝐷𝐷∪𝐷𝐷𝐿𝐿∪𝐷𝐷𝑅𝑅 � >0, (4.6b) 𝐷𝐷∪𝐷𝐷𝐿𝐿∪𝐷𝐷𝑅𝑅 respectively. In Eq. (4.6), 𝜒𝜒𝑖𝑖(𝒙𝒙),𝜒𝜒D𝐿𝐿(𝒙𝒙) and 𝜒𝜒D𝑅𝑅(𝒙𝒙) denote the TDF of ΩV𝑖𝑖, D𝐿𝐿 and D𝑅𝑅, to should also be determined appropriately. Actually in order to ensure the self-supporting property of the structure, we also require that respectively. The meanings of the symbols appeared in the Eq. (4.5) can be understood from Fig. 15. difference scheme suggested in [25, 27]. 𝐻𝐻(min (𝜒𝜒𝑖𝑖(𝒙𝒙),𝜒𝜒D𝐿𝐿(𝒙𝒙))dV 𝐻𝐻(min (𝜒𝜒𝑖𝑖(𝒙𝒙),𝜒𝜒D𝑅𝑅(𝒙𝒙))dV and respectively. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 16 In the present work, the exterior boundary of the structure is also represented by a B-spline curve (i.e., 𝐶𝐶0) shown in Fig. 16 where 10 control points are adopted. The only constraint imposed on the coordinates of these control points is 𝑦𝑦5≤−𝑅𝑅, where 𝑅𝑅 is a sufficiently large positive number. This treatment guarantees that the exterior boundary cannot be of the wavy shape and is totally separated from the baseplate since these two situations will inevitably lead to the existence of unprintable structures (see Fig. 16 for reference). 5. Numerical examples In this section, several numerical examples are provided to demonstrate the effectiveness of the proposed approaches for designing self-supporting structures through topology optimization. Since the main purpose of the present section is to test the numerical performance of the suggested approach, the material, load and geometric data are all chosen as dimensionless unless otherwise stated. The Young's modulus and Poisson's ratio of the solid material are taken as 𝐸𝐸=1 and 𝜈𝜈= 0.3, respectively. Plane stress state (with unit thickness) is assumed and four-node bilinear square elements are adopted for finite element analysis in all presented examples. The Method of Moving Asymptotes (MMA) [33] is adopted as numerical optimizer to solve the optimization problems. Structural compliance is taken as the objective functional and the printability constraints as well as volume constraint on available solid material are always considered in all examples. For all examples, in Eq. (2.8) and among of them only five control points are independent) and B-splines with ten control points are used to characterize the exterior boundary of the structures, respectively, in the B-splines with six control points are adopted to describe the shape of each printable void (i.e., 𝑛𝑛=6 MMV-based approach. Furthermore, the lower bound of the overhang angle is set to 𝜃𝜃=45°. proposed approaches. In this example, a rectangular design domain which is discretized by a 160× 80 FE mesh is shown in Fig. 17. The left side of the design domain is fixed and distributed horizontal loads with magnitude of 1 are imposed on the right top side of the design domain. It is A simple tensile beam example [21] is first considered in order to test the effectiveness of the 5.1 Tensile beam example Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 17 Firstly, this example is tested by employing MMC-based approach formulated in Eq. (2.1). The considering the self-supporting constraint (a horizontal strait beam). The final optimized rotation assumed that the print direction is 𝒃𝒃𝑝𝑝=(0,1)⊤. initial design composed of 16 components is shown in Fig. 18 and the initial rotation angles (i.e., 𝛼𝛼) of the working plane is set to 45°. It is assumed that the upper bound of the solid material is 𝑉𝑉�= 20%. The optimized structure is shown in Fig. 19, which is almost the same as that without angles of the working plane is 𝛼𝛼=75.93°. Here the total number of design variables equals 16× 5+1=81. Some intermediate optimization results obtained during the optimization process is upper bound of the relative available volume constraint is set to 𝑉𝑉=40% since part of solid totally 10×(4+2)+10=70 (i.e., 4 independent control points for each printable void and 1 center point as well as 10 control points for the exterior structural boundary) and 15×(4+2)+ 10=100 design variables for these two initial designs, respectively. Optimized structures obtained material will be used to form support structure. Two different initial designs containing 10 and 15 non-overlap printable voids (shown in Fig. 21a and Fig. 21b, respectively) are considered. There are from the two initial designs are shown in Fig. 22, respectively. It can be seen from these figures that besides a horizontal solid beam at the top side of the structure which constitutes the main load transition path, some of the solid material (about 15%) has been used to form the supporting structure between the horizontal beam and baseplate in the optimized designs. These two optimized designs are obviously satisfied the self-supporting requirement since every part of the structures is sufficiently supported along print direction and there is actually no component whose inclined angle given in Fig. 20. The problem is also solved by the MMV-based approach formulated in Eq. (2.9). This time, the is larger than the prescribed critical value (i.e., 𝜃𝜃=45°). Numerical results also indicate that as the number of printable voids included in the initial design increases, the performance (measure in terms of the value of the objective functional) of the structure improves. This is quite consistent with our theoretical analysis made in Section 3 which states that the optimal support structure may be constructed by many thin components forming a hierarchical network. It is also obvious that the creation of the support structure is purely to satisfy the printability requirement. Furthermore, Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 18 compared to the optimized designs show in [21], the optimized designs obtained by the proposed approach are totally free from the existence of undesirable grey elements. Fig. 23 shows some intermediate optimization results obtained during the course of optimization where the initial design shown in Fig. 21b is adopted. From this figure it can be observed clearly how printable topologies are reached by changing the positions and shapes of the voids. Corresponding convergence iteration histories are also provided in Fig. 24. It can be observed from this figure (and corresponding figures in the following examples), in the MMV approach, the objective function can be reduced to a stable value within 100 iterations and the subsequent iteration steps are used to satisfy the AM related geometry constraints. 5.2 Short cantilever beam example In this example, the well-known short beam problem is examined. The design domain, external load, and boundary conditions are all shown in Fig. 25. A rectangular design domain has the width of 𝑊𝑊=2 and length 𝐿𝐿=1 is discretized by a 120×60 FE mesh. A unit vertical load is imposed on the middle point of the right side of the design domain. For this problem it is assumed that 𝑉𝑉=50% and 𝒃𝒃𝑝𝑝=(0,1)⊤. shown in Fig. 26. The corresponding optimal value of the objective functional is 𝐼𝐼=61.59. This obviously less than the prescribed critical overhang angle 𝜃𝜃=45°. rotation angles of the working plane are set to 45° and 90°, respectively. The total number of design variables is actually only 16×5+1=81 in this MMC-based approach. Corresponding the external load but unprintable when 𝛼𝛼=0°, can now be printed by rotating the working plane to optimization results are shown in Fig. 28a and Fig. 28b, respectively. It can be observed from these figures that the proposed approach does have the capability to find the optimal value of the rotation angle of the working plane. The horizontal components in Fig. 26 which are very effective to transfer Secondly, the problem is solved by employing MMC-based approach formulated in Eq. (2.1). The initial designs shown in Fig. 27a and Fig. 27b are composed of 16 components and the initial Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 19 Firstly, the optimized structure obtained without considering the self-supporting constraint is structure is, however, not self-supportive since the overhang angles of some structural parts are during the process of optimization, the symmetry property of the problem is lost and the final optimized structure is not symmetric anymore. It is also worth noting that the optimal value of the variables in the problem formulation. Some intermediate steps of the optimization process starting from the initial design shown in Fig. 27a are shown in Fig. 29. It can be seen from this figure that 𝛼𝛼=90°. This demonstrates the advantage of adopting the rotation angle as one of the design compared to the case where no self-supporting requirement is considered and 𝛼𝛼=0° is fixed objective functional is 𝐼𝐼=62.8 which is very close to the value of the structure shown in Fig. 26. a design variable. In addition, if we start the optimization from 𝛼𝛼=90°, it can be observed from value of the objective functional for this structure is 𝐼𝐼=62.55. The values of the inclined angles of some components in the optimized structure shown in Fig. 28 are listed in Table 1. Corresponding iteration history is also plotted in Fig. 31. Compared with the MMV approach, convergence is more rapid in the MMC approach since less number of geometry constraints are included in the problem formulation. For this example, optimized structure is obtained within 300 iterations. Fig. 30 that the optimized value of the rotation angle of the working plane keeps the same value. The This demonstrates once again the necessity of introducing the rotation angle of the working plane as Lastly, the considered problem is solved with the MMV-based approach formulated in Eq. are distributed in the design domain as the initial design. Under this circumstance, the total number (2.9). As shown in Fig. 32 (where 𝒃𝒃𝑝𝑝=(0,1)⊤ ) and Fig. 33 (𝒃𝒃𝑝𝑝=(1,0)⊤), twelve printable voids of design variables is 12×(4+2)+10=82. Fig. 34 and Fig. 35 plot the optimized structures obtained under different print directions. It can be observed that the self-supporting requirement does have been satisfied by these structures. The obtained structures are totally black-and-white and have crisp boundaries. These features cannot be achieved easily with use of traditional methods. It is also worth noting that the optimal structural topology is highly dependent on the print direction when self-supporting requirement is considered. When 𝒃𝒃𝑝𝑝=(0,1)⊤, the corresponding optimized structure is not symmetric and the value of the objective functional for this structure is 𝐼𝐼=72.164. print direction is changed to 𝒃𝒃𝑝𝑝=(1,0)⊤, the obtained optimized structure is very similar to the one shown in Fig. 26 and the compliance of this structure is 𝐼𝐼=63.408 which is also very close to the value associated with the structure obtained without considering the self-support constraint (i.e., 𝐼𝐼= This non-symmetric behavior is also consistent with the observation made in [23]. If, however, the Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 20 61.59). Fig. 36 provides some intermediate results during the process of optimization where 𝒃𝒃𝑝𝑝= (1,0)⊤. It can be observed from this figure that although the voids are restricted to be non-overlap, significant topology changes can still be achieved during the course of optimization. Twelve voids in the initial design finally reduce to only three ones through shrinking or moving outside. This demonstrates clearly the capability of the MMV-based approach to deal with topology changes. Fig. 37 plots the histories of the values of the objective functional and constraint functions during the process of numerical optimization. 5.3 MBB-beam example Fig. 39 plots the optimization results obtained without considering self-supporting requirement In this example, a MBB example will be investigated. The setting of this problem is described schematically in Fig. 38. A vertical load is imposed on the middle point of the top side of a rectangular design domain，which is discretized by a 360×60 FE mesh. The upper bound on the relative available volume of the solid material (for half of the structure) is 𝑉𝑉�=50%. for comparison purpose. The corresponding value of the objective functional is 𝐼𝐼=381.80. Since there exist several horizontal structural members in this structure, it is not printable when 𝒃𝒃𝑝𝑝= (0,1)⊤. Next, the same problem is solved by employing the MMC-based approach formulated in Eq. (2.1) where the layout of the structural components and the rotational angle of the working plane are all taken as design variables. The initial design shown in Fig. 40 contains 48 straight components (the corresponding number of design variables is 121) and the initial rotation angle of the working plane is set to 𝛼𝛼=45°. The optimized structure is shown in Fig. 41 and the corresponding value of the objective functional is 𝐼𝐼=386.36. Besides, the optimized value of the rotation angle is 𝛼𝛼= 84.87°. For this optimized solution, the compliance is only increased 1% compared to the solution without considering the self-supporting requirement (𝐼𝐼=381.80). This demonstrates clearly that it is necessary to include the print direction as a design variable in AM oriented topology optimization problems. Finally, the MMV-based approach formulated in Eq. (2.9) is also used to solve this problem. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 21 Since the problem under consideration is symmetric in nature, only half of the structure is considered. As shown Fig. 42, there are totally 14 printable voids in the initial design and the total number of design variables is 94. Fig. 43a plots the optimized design where printable voids with different sizes are distributed separately to achieve high structural efficiency. The corresponding compliance is 𝐼𝐼= 436.08, which is only approximately 14% higher than that of the unconstrained optimal solution (𝐼𝐼=381.80). This is of course the price that should be paid for considering the self-supporting constraint. The obtained structure is obviously self-supporting and quite reasonable from mechanics point of view. From Fig. 43b, it is also noteworthy that the optimized structure inherits most the optimal feature of the unconstrained design, for example, three printable voids (but now with more large inclined angles) and a relatively long horizontal top chord. It is also interesting to observe that two unprintable voids in the unconstrained structure have been replaced by six printable voids (which also provide sufficient support on the top chord) in the optimized self-supporting structure. In some sense, this optimality mechanism is also consistent with the theoretical analysis made in first segment: Section 3. Moreover, the explicit piecewise expression of the 𝑪𝑪1 boundary curve in Fig. 43a is: the 𝑪𝑪1(𝑢𝑢)=�𝑥𝑥𝑦𝑦�=�−10.72𝑢𝑢2+4.76𝑢𝑢+2.05 0≤𝑢𝑢≤0.33, (5.1𝑎𝑎) 7.29𝑢𝑢2−4.89𝑢𝑢+0.87 �, the second segment: 𝑪𝑪1(𝑢𝑢)=�𝑥𝑥𝑦𝑦�=�0.09𝑢𝑢2−2.44𝑢𝑢+3.25 0.33≤𝑢𝑢≤0.66 (5.1𝑏𝑏) 0.07𝑢𝑢2−0.07𝑢𝑢+0.07�, and the third segment 𝑪𝑪1(𝑢𝑢)=�𝑥𝑥𝑦𝑦�=�10.45𝑢𝑢2−16.25𝑢𝑢+7.85 0.66≤𝑢𝑢≤1, (5.1𝑛𝑛) 7.29𝑢𝑢2−9.70𝑢𝑢+3.28 �, respectively. Fig. 44 plots some intermediate optimization results. 6. Concluding remarks In the present paper, AM oriented topology optimization methods that can generate self-supporting structures are developed under the MMC and MMV solution frameworks. Numerical examples show that the proposed approaches (especially the MMV-based approach) do have the capability of finding optimized designs where overhang angle constraints can be fully respected. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 22 Besides optimal structural topology, the build orientation of AM can also be optimized with use of the proposed approaches in a straightforward way. Compared with existing approaches, the distinctive feature of the present approach is that it solves the corresponding problem through a more explicit and geometrical treatment. As for other AM related manufacture constraints, it is worth noting that as shown in [17], the proposed MMC-based approach also has potential to deal with minimum length scale constraint. Furthermore, for two dimensional problems the issue of enclosed voids can be dealt with easily using the proposed MMV-based approach by imposing the constraint such that there is no printable features existing in the interior of a design domain. Extending the proposed methods to three dimensional (3D) problems can be achieved by introducing some 3D printable features like the one shown in Fig. 45. Corresponding results will be reported elsewhere. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 23 Appendix and In Eq. (A2a), In this appendix, sensitivity analysis associated with the MMV approach will be outlined briefly. Actually, with use of the B-spline curve description scheme, the variation of the boundary corresponding to 𝑪𝑪𝑖𝑖(𝑢𝑢) along the outward normal direction in Eq. (4.1) can be calculated as: 𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖=𝛿𝛿𝑪𝑪𝑖𝑖(𝑢𝑢)⋅𝒏𝒏𝑖𝑖, (A1) where 𝒏𝒏𝑖𝑖 is the outward normal vector of the boundary and it can be calculated from the relations 𝒕𝒕𝑖𝑖∙𝒏𝒏𝑖𝑖=0 (A2a) ‖𝒏𝒏𝑖𝑖‖=1. (A2b) 𝒕𝒕𝑖𝑖=𝑑𝑑𝑪𝑪𝑖𝑖(𝑢𝑢)𝑑𝑑𝑢𝑢 =�𝑑𝑑𝑁𝑁𝑘𝑘,𝑝𝑝(𝑢𝑢) 𝑛𝑛 𝑷𝑷𝑘𝑘𝑖𝑖 , 𝑎𝑎≤𝑢𝑢≤𝑏𝑏. (A3) 𝑑𝑑𝑢𝑢 𝑘𝑘=0 𝛿𝛿𝑪𝑪𝑖𝑖(𝑢𝑢)=�𝑁𝑁𝑘𝑘,𝑝𝑝(𝑢𝑢)𝛿𝛿𝑷𝑷𝑘𝑘𝑖𝑖 , 𝑎𝑎≤𝑢𝑢≤𝑏𝑏, (A4) 𝜃𝜃= 𝜋𝜋𝑛𝑛−2, 𝑘𝑘=1,…,𝑛𝑛−1, (A5a) 𝛿𝛿𝑷𝑷0𝑖𝑖=𝛿𝛿𝑷𝑷𝑛𝑛𝑖𝑖=�𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖,𝛿𝛿𝑦𝑦𝑛𝑛𝑖𝑖+𝛿𝛿𝑑𝑑0𝑖𝑖�⊤. (A5b) The above derivation immediately leads to the conclusion that 𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖=𝛿𝛿𝑪𝑪𝑖𝑖(𝑢𝑢)⋅𝒏𝒏𝑖𝑖=∑ (𝐴𝐴𝑗𝑗𝑖𝑖𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖+ 𝐵𝐵𝑗𝑗𝑖𝑖𝛿𝛿𝑦𝑦𝑛𝑛𝑖𝑖+𝐶𝐶𝑗𝑗𝑖𝑖𝛿𝛿𝑑𝑑𝑗𝑗𝑖𝑖) . The detailed expressions of 𝐴𝐴𝑗𝑗𝑖𝑖,𝐵𝐵𝑗𝑗𝑖𝑖 and 𝐶𝐶𝑗𝑗𝑖𝑖 are omitted here since the The expression of 𝛿𝛿𝑪𝑪𝑖𝑖(𝑢𝑢) can be calculated from Eq. (2.8a) in the main text as 𝛿𝛿𝑷𝑷𝑘𝑘𝑖𝑖=�𝛿𝛿𝑥𝑥𝑛𝑛𝑖𝑖+sin�(𝑘𝑘−1)𝜃𝜃𝑘𝑘𝑖𝑖+𝜋𝜋2�𝛿𝛿𝑑𝑑𝑘𝑘𝑖𝑖, δ𝑦𝑦𝑛𝑛𝑖𝑖+cos�(𝑘𝑘−1)𝜃𝜃𝑘𝑘𝑖𝑖+𝜋𝜋2�𝛿𝛿𝑑𝑑𝑘𝑘𝑖𝑖�⊤, with and 𝑛𝑛 𝑘𝑘=0 𝑛𝑛𝑗𝑗=0 corresponding derivation is a trivial task. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 24 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 25 Acknowledgements The financial supports from the National Key Research and Development Plan (2016YFB0201600), the National Natural Science Foundation (11402048, 11372004), Program for Changjiang Scholars, Innovative Research Team in University (PCSIRT) and 111 Project (B14013) are gratefully acknowledged. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 26 References [1] K.T. Cheng, N. Olhoff, An investigation concerning optimal design of solid elastic plate, Int. J. Solids Struct. 17 (1981) 305–323. [2] M.P. Bendsøe, N. Kikuchi, Generating optimal topologies in structural design using a homogenization method, Comput. Methods Appl. Mech. Engrg. 71 (1988) 197–224. [3] G.I.N. Rozvany, A critical review of established methods of structural topology optimization, Struct. Multidiscip. Optim. 37 (2009) 217–237. [4] S. Nelaturi, W. Kim, T. Kurtoglu, Manufacturability feedback and model correction for additive manufacturing, ASME J. Manuf. Sci. Engrg. 137 (2015) 21015-1-021015-9. [5] R. Mertens, S. Clijsters, K. Kempen, J.P. Kruth, Optimization of scan strategies in selective laser melting of aluminum parts with downfacing areas, ASME J. Manuf. Sci. Engrg. 136 (2014) 61012-1-061012-7. [6] J. Krantz, D. Herzog, C. Emmelmann, Design guidelines for laser additive manufacturing of light weight structures in TiAl6V4, J. Laser Appl. 27 (2015) S14001-1- S14001-16. [7] D. Brackett, I. Ashcroft, R. Hague, Topology optimization for additive manufacturing, Proceedings of the 24th Solid Freeform Fabrication Symposium, Austin, TX, 2011, 348–362. [8] J.K. Liu, Y.S. Ma, A survey of manufacturing oriented topology optimization methods, Adv. Eng. Softw. 100 (2016) 161-175. [9] T.A. Poulsen, A new scheme for imposing minimum length scale in topology optimization, Int. J. Numer. Methods Engrg. 57 (2003) 741–760. [10] J.K. Guest, Imposing maximum length scale in topology optimization, Struct. Multidiscip. Optim. 37 (2009) 463–473. [11] S.K. Chen, M.Y. Wang, A.Q. Liu, Shape feature control in structural topology optimization, Comput. Aided Des. 40 (2008) 951–962. [12] X. Guo, W.S. Zhang, W.L. Zhong, Explicit feature control in structural topology optimization via level set method, Comput. Methods Appl. Mech. Engrg. 272 (2014) 354–378. [13] W.S. Zhang, W.L. Zhong, X. Guo, An explicit length scale control approach in SIMP-based topology optimization, Comput. Methods Appl. Mech. Engrg. 282 (2014) 71–86. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 27 [14] Q. Xia, T.L. Shi, Constraints of distance from boundary to skeleton: For the control of length scale in level set based structural topology optimization, Comput. Methods Appl. Mech. Engrg. 295 (2015) 525-542. [15] M.D. Zhou, B.S. Lazarov, F.W. Wang, O. Sigmund, Minimum length scale in topology optimization by geometric constraints, Comput. Methods Appl. Mech. Engrg. 293 (2015) 266-282. [16] G. Allaire, F. Jouve, G. Michailidis, Thickness control in structural optimization via a level set method, Struct. Multidiscip. Optim. 53 (2016) 1349–1382. [17] W.S. Zhang, D. Li, J. Zhang, X. Guo, Minimum length scale control in structural topology optimization based on the Moving Morphable Components (MMC) approach, Comput. Methods Appl. Mech. Engrg. 311 (2016) 327–355. [18] Q.H. Li, W.J. Chen, S. Liu, L.Y. Tong, Structural topology optimization considering connectivity constraint, Struct. Multidiscip. Optim. 54 (2016) 971-984. [19] M. Leary, L. Merli, F. Torti, M. Mazur, M. Brandt, Optimal topology for additive manufacture: A method enabling additive manufacture of support-free optimal structures, Mater. Des. 63 (2014) 678-690. [20] K. Hu, S. Jin, C.C. Wang, Support slimming for single material based additive manufacturing, Comput. Aided Des. 65 (2015) 1-10. [21] M. Langelaar, An additive manufacturing filter for topology optimization of print-ready designs, Struct. Multidiscip. Optim. doi:10.1007/s00158-016-1522-2. [22] M. Langelaar, Topology optimization of 3D self-supporting structures for additive manufacturing, Addit. Manuf. 12 (2016) 60-70. [23] A.T. Gaynor, J.K. Guest, Topology optimization considering overhang constraints: Eliminating sacrificial support material in additive manufacturing through design, Struct. Multidiscip. Optim. doi:10.1007/s00158-016-1551-x. [24] X. Guo, W.S. Zhang, W.L. Zhong, Doing topology optimization explicitly and geometrically-a new moving morphable components based framework, ASME Trans. J. Appl. Mech. 81 (2014) 081009-1–081009-12. [25] W.S. Zhang, J. Yuan, J. Zhang, X. Guo, A new topology optimization approach based on Moving Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 28 Morphable Components (MMC) and the ersatz material model, Struct. Multidiscip. Optim. 53 (2016) 1243–1260. [26] J. A. Norato, B.K. Bell, D.A. Tortorelli, A geometry projection method for continuum-based topology optimization with discrete elements, Comput. Methods Appl. Mech. Engrg. 293 (2015) 306–327. [27] W.S. Zhang, J. Zhang, X. Guo, Lagrangian description based topology optimization-A revival of shape optimization, ASME Trans. J. Appl. Mech. 83 (2016) 041010-1–041010-16. [28] X. Guo, W.S. Zhang, J. Zhang, Explicit structural topology optimization based on moving morphable components (MMC) with curved skeletons, Comput. Methods Appl. Mech. Engrg. 310 (2016) 711-748. [29] W.S. Zhang, W.Y. Yang, J.H. Zhou, D. Li, X. Guo, Structural optimization through explicit boundary evolution, ASME Trans. J. Appl. Mech. 84 (2017) 011011-1–011011-10. [30] P. Wei, M.Y. Wang, X.H. Xing, A study on X-FEM in continuum structural optimization using a level set model, Comput. Aided Des. 42 (2010) 708-719. [31] P. Shan, Optimal embedding objects in the topology design of structure (Master thesis) Dalian University of Technology (2008) http://d.wanfangdata.com.cn/Thesis_Y1247462.aspx [32] W.S. Zhang, W.Z. Zhong, X. Guo, Explicit layout control in optimal design of structural systems with multiple embedding components, Comput. Methods Appl. Mech. Engrg. 290 (2015) 290-313. [33] K. Svanberg, The method of moving asymptotes-a new method for structural optimization, Int. J. Numer. Methods Engrg. 24 (1987) 359–373. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 29 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 30 Too small feature sizes (a). Unprintable structural features. n o i t c e r i d t n i r P Overhang part (b). An unprintable structural part without sufficient support from below. Print direction (c). A non-self-supporting structure. Fig. 1 A schematic illustration of the design limitations in AM. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 31 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 32 Moving morphable components Fig. 2 The basic idea of the MMC-based topology optimization approach. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 33 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 34 2𝐿𝐿𝑘𝑘 𝜃𝜃𝑘𝑘 (𝑥𝑥0𝑘𝑘,𝑦𝑦0𝑘𝑘) 90°≤𝜃𝜃𝑘𝑘≤180° x 2𝑡𝑡𝑘𝑘 2𝐿𝐿𝑘𝑘 𝜃𝜃𝑘𝑘 (𝑥𝑥0𝑘𝑘,𝑦𝑦0𝑘𝑘) 0°≤𝜃𝜃𝑘𝑘≤90° 2𝑡𝑡𝑘𝑘 x Fig. 3 Geometry description of a structural component in the MMC approach. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 35 y o y o Print direction Working plane 𝜃𝜃𝑘𝑘 sin𝜃𝜃𝑘𝑘<sin𝜃𝜃 , unprintable before rotation x Working plane x Print direction 𝜃𝜃𝑘𝑘 𝛼𝛼 𝛼𝛼 sin(𝜃𝜃𝑘𝑘+𝛼𝛼)≥sin𝜃𝜃 , printable after rotation Fig. 4 The rotation angle of the working plane. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 36 i j 𝜃𝜃𝑖𝑖 𝜃𝜃𝑗𝑗 sin𝜃𝜃𝑗𝑗≥sin𝜃𝜃 sin𝜃𝜃𝑖𝑖≥sin𝜃𝜃 (a). An unprintable V-shape material distribution case. Component 1 Component k (𝑥𝑥0𝑘𝑘,𝑦𝑦0𝑘𝑘) 2𝐿𝐿𝑘𝑘 2𝑡𝑡𝑘𝑘 sin𝜃𝜃𝑘𝑘≥sin𝜃𝜃, 𝑘𝑘=1,…,𝑛𝑛 Component n 𝐹𝐹 (b). Another unprintable components distribution case. Fig. 5 An illustration of unprintable components distribution cases. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 37 𝒃𝒃𝑝𝑝 𝒏𝒏 𝒃𝒃𝑝𝑝 𝒏𝒏 Ω𝑠𝑠 𝒙𝒙 𝜀𝜀 𝐶𝐶𝜀𝜀(𝒙𝒙) An unprintable case Ω𝑠𝑠 meas�Ω𝑠𝑠∩𝐶𝐶𝜀𝜀(𝒙𝒙)�>0, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 𝒏𝒏∙𝒃𝒃𝑝𝑝≥cos𝜃𝜃, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 𝒃𝒃𝑝𝑝 𝒏𝒏 𝒃𝒃𝑝𝑝 𝒏𝒏 Ω𝑠𝑠 Ω𝑠𝑠 𝜀𝜀 𝒙𝒙 𝐶𝐶𝜀𝜀(𝒙𝒙) 𝒃𝒃𝑝𝑝 Ω𝑠𝑠 𝒙𝒙 𝜀𝜀 𝐶𝐶𝜀𝜀(𝒙𝒙) 𝒃𝒃𝑝𝑝 𝒏𝒏 𝒏𝒏 Ω𝑠𝑠 meas�Ω𝑠𝑠∩𝐶𝐶𝜀𝜀(𝒙𝒙)�>0, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 𝒏𝒏∙𝒃𝒃𝑝𝑝≤cos𝜃𝜃, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 𝒃𝒃𝑝𝑝 𝒃𝒃𝑝𝑝 A printable case 𝒏𝒏 Ω𝑠𝑠 𝒏𝒏 Ω𝑠𝑠 𝜀𝜀 𝒙𝒙 𝐶𝐶𝜀𝜀(𝒙𝒙) An unprintable case meas�Ω𝑠𝑠∩𝐶𝐶𝜀𝜀(𝒙𝒙)�≤0, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 𝒏𝒏∙𝒃𝒃𝑝𝑝≥cos𝜃𝜃, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 An unprintable case meas�Ω𝑠𝑠∩𝐶𝐶𝜀𝜀(𝒙𝒙)�≤0, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 𝒏𝒏∙𝒃𝒃𝑝𝑝≤cos𝜃𝜃, ∀𝒙𝒙∈𝜕𝜕Ω𝑠𝑠 Fig. 6 A schematic illustration of the meanings of the constraints in Eq. (2.7a) and Eq. (2.7b). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 38 Moving morphable voids Fig. 7 The basic idea of the MMV-based topology optimization approach. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 39 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 40 𝑷𝑷4=𝑷𝑷0(𝑥𝑥0,𝑦𝑦0) 𝑑𝑑0 (𝑥𝑥𝑛𝑛,𝑦𝑦𝑛𝑛) 𝑑𝑑1 𝑑𝑑3 𝜃𝜃 2𝜃𝜃 𝑑𝑑2 𝑷𝑷2(𝑥𝑥2,𝑦𝑦2) 𝑷𝑷1(𝑥𝑥1,𝑦𝑦1) 𝑷𝑷3(𝑥𝑥3,𝑦𝑦3) 𝑦𝑦 𝑂𝑂 Fig. 8 The construction of a typical B-spline curve 𝑪𝑪(𝑢𝑢) described in Eq. (2.8). 𝑥𝑥 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 41 Fig. 9 Some printable features with complex shapes. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 42 D (𝑥𝑥𝑛𝑛0,y𝑛𝑛0) 𝐶𝐶𝑗𝑗 𝐶𝐶0 𝐶𝐶𝑖𝑖 (a). ΩV𝑖𝑖⋂ΩV𝑗𝑗=∅,�ΩV𝑖𝑖⋃ΩV𝑗𝑗�⋂𝐶𝐶0=∅ (printable case). 𝐶𝐶𝑖𝑖 𝐶𝐶𝑗𝑗 (b). ΩV𝑖𝑖⋂ΩV𝑗𝑗=∅,ΩV𝑖𝑖⋂𝐶𝐶0≠∅,ΩV𝑗𝑗⋂𝐶𝐶0≠∅ (printable case). 𝐶𝐶𝑗𝑗 𝐶𝐶𝑖𝑖 (c). ΩV𝑖𝑖⋂ΩV𝑗𝑗≠∅ (unprintable case). Fig. 10 Printable and unprintable cases in the MMV-based approach. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 43 𝐶𝐶0 𝐶𝐶0 D (𝑥𝑥𝑛𝑛0,y𝑛𝑛0) D (𝑥𝑥𝑛𝑛0,y𝑛𝑛0) 𝑦𝑦=𝑦𝑦7−𝑦𝑦0 𝑥𝑥7−𝑥𝑥0(𝑥𝑥−𝑥𝑥0)+𝑦𝑦0 𝑷𝑷7(𝑥𝑥7,𝑦𝑦7) 𝑷𝑷5(𝑥𝑥5,𝑦𝑦5) 𝑦𝑦=𝑦𝑦1−𝑦𝑦0 𝑥𝑥1−𝑥𝑥0(𝑥𝑥−𝑥𝑥0)+𝑦𝑦0 𝑷𝑷1(𝑥𝑥1,𝑦𝑦1) 𝑷𝑷3(𝑥𝑥3,𝑦𝑦3) 𝑷𝑷8=𝑷𝑷0(𝑥𝑥0,𝑦𝑦0) 𝑷𝑷2(𝑥𝑥2,𝑦𝑦2) 𝑷𝑷6(𝑥𝑥6,𝑦𝑦6) 𝑷𝑷4(𝑥𝑥4,𝑦𝑦4) 𝑥𝑥2<𝑥𝑥3 , 𝑥𝑥5<𝑥𝑥6 𝑷𝑷8=𝑷𝑷0(𝑥𝑥0,𝑦𝑦0) (a). An unprintable case. 𝑦𝑦=𝑦𝑦7−𝑦𝑦0 𝑥𝑥7−𝑥𝑥0(𝑥𝑥−𝑥𝑥0)+𝑦𝑦0 𝑷𝑷7(𝑥𝑥7,𝑦𝑦7) 𝑷𝑷6(𝑥𝑥6,𝑦𝑦6) 𝑦𝑦=𝑦𝑦1−𝑦𝑦0 𝑥𝑥1−𝑥𝑥0(𝑥𝑥−𝑥𝑥0)+𝑦𝑦0 𝑷𝑷1(𝑥𝑥1,𝑦𝑦1) 𝑷𝑷2(𝑥𝑥2,𝑦𝑦2) 𝑷𝑷3(𝑥𝑥3,𝑦𝑦3) 𝑷𝑷5(𝑥𝑥5,𝑦𝑦5) 𝑥𝑥𝑖𝑖≥𝑥𝑥𝑖𝑖+1 , 𝑖𝑖=2,⋯,n−2. 𝑷𝑷4(𝑥𝑥4,𝑦𝑦4) (b). A printable case. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 44 Fig. 11 A schematic illustration of the construction of a printable void. (a). Optimal structure without considering self-supporting requirement. (b). A conjectured optimal self-supporting structure. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 45 𝑙𝑙 1 (c). An assumed non-self-supporting optimal structure with rectangular holes. (d). Making the structure shown in Fig. (12c) self-supporting by adding support material. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 46 (e). A hierarchical support structure. Fig. 12 Construction of an optimal self-supporting structure. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 47 FE node IN=0 IN=1 Fig. 13 A schematic illustration of the construction of 𝜒𝜒𝑠𝑠. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 48 Void1 Void2 Void3 Void4 2δ Void5 Void6 Fig. 14 A schematic illustration of the non-intersection constraint with a δ-expsnsion treatment. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 49 D𝐿𝐿 Ω𝑉𝑉𝑖𝑖 �𝑥𝑥𝑛𝑛𝑖𝑖,𝑦𝑦𝑛𝑛𝑖𝑖� Design domain Printable case Unprintable case Void Void Unprintable D𝑅𝑅 Ω𝑉𝑉𝑗𝑗 �𝑥𝑥𝑛𝑛𝑗𝑗,𝑦𝑦𝑛𝑛𝑗𝑗� Fig. 15 A schematic illustration of the meanings of the symbols appeared in the Eq. (4.5). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 50 𝑷𝑷8(𝑥𝑥8,𝑦𝑦8) 𝑷𝑷9(𝑥𝑥9,𝑦𝑦9) 𝑷𝑷7(𝑥𝑥7,𝑦𝑦7) 𝑷𝑷6(𝑥𝑥6,𝑦𝑦6) 𝑷𝑷8(𝑥𝑥8,𝑦𝑦8) 𝑷𝑷7(𝑥𝑥7,𝑦𝑦7) 𝑷𝑷6(𝑥𝑥6,𝑦𝑦6) 𝑷𝑷9(𝑥𝑥9,𝑦𝑦9) 𝑷𝑷2(𝑥𝑥2,𝑦𝑦2) D 𝑷𝑷3(𝑥𝑥3,𝑦𝑦3) 𝑷𝑷4(𝑥𝑥4,𝑦𝑦4) 𝑷𝑷1(𝑥𝑥1,𝑦𝑦1) 𝑷𝑷10=𝑷𝑷0(𝑥𝑥0,𝑦𝑦0) 𝜃𝜃 𝑷𝑷5(𝑥𝑥5,𝑦𝑦5) Shape change 𝑷𝑷10=𝑷𝑷0(𝑥𝑥0,𝑦𝑦0) 𝜃𝜃 𝑷𝑷1(𝑥𝑥1,𝑦𝑦1) 𝑷𝑷2(𝑥𝑥2,𝑦𝑦2) D 𝑷𝑷3(𝑥𝑥3,𝑦𝑦3) 𝑷𝑷5(𝑥𝑥5,𝑦𝑦5) 𝑷𝑷4(𝑥𝑥4,𝑦𝑦4) Fig. 16 A schematic illustration of the B-spline representation of the exterior boundary of a structure. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 51 1 160×80 mesh 2 Baseplate 0.2 𝑞𝑞=1 n o i t c e r i d t n i r P Fig. 17 The tensile beam example. 𝑦𝑦 𝑂𝑂 𝑥𝑥 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 52 𝛼𝛼=45° Fig. 18 The initial design of the tensile beam problem under the MMC-based formulation. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 53 𝛼𝛼=75.93° (a). The optimized structure (contour plot). 𝛼𝛼=75.93° (b). The optimized structure (component plot). Fig. 19 The optimized structure obtained with the MMC-based formulation. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 54 (a). step 6. (b). step 46. (c). step 96. (d). step 151. Fig. 20 Some intermediate steps of the optimization process. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 55 (a). A initial design with 10 voids. (b). A initial design with 15 voids. Fig. 21 The initial designs of the tensile beam problem (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 56 (a). The optimized structure obtained with initial design shown in Fig. 21a. (b). The optimized structure obtained with initial design shown in Fig. 21b. Fig. 22 The optimized structures of the tensile beam problem (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 57 (a). step 2. (b). step 20. (c). step 95. (d). step 465. Fig. 23 Some intermediate steps of the optimization process with the initial design shown in Fig. 21b. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 58 l a n o i t c n u f e v i t c e b O j 2x108 1x108 0 0 Objective functional Volume constraint fucntional 2.0x104 1.5x104 1.0x104 5.0x103 0.0 1000 1100 1200 1300 200 400 600 800 Iteration step 1000 1200 l a n o i t n c u f t i n a r t s n o c e m u o V l 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 1400 (a). Convergence history of the optimized structure with initial design shown in Fig. 21a. l a n o i t c n u f e v i t c e b O j 2x108 1x108 0 0 Objective functional Volume constraint fucntional 2.0x104 1.5x104 1.0x104 5.0x103 0.0 900 1000 1100 1200 200 400 600 800 Iteration step 1000 1200 l a n o i t n c u f t i n a r t s n o c e m u o V l 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 1400 (b). Convergence history of the optimized structure with initial design shown in Fig. 21b. Fig. 24 Convergence history of the tensile beam problem (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 59 𝑦𝑦 𝑂𝑂 𝑥𝑥 120×60 mesh 2 Baseplate 1 𝑡𝑡=1 n o i t c e r i d t n i r P Fig. 25 The short beam example. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 60 Fig. 26 The optimized structure without considering self-supporting requirement. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 61 𝛼𝛼=45° (a). A initial design with working plane 𝛼𝛼=45°. 𝛼𝛼=90° (b). A initial design with working plane 𝛼𝛼=90°. Fig. 27 The initial designs of the short beam problem under the MMC formulation. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 62 (a). The optimized structure (contour plot). (b). The optimized structure (component plot). Fig. 28 The optimized structure obtained with initial design shown in Fig. 27a Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 63 under the MMC formulation. 𝛼𝛼=77° (a). step 12. (b). step 70. 𝛼𝛼=81° 𝛼𝛼=82° (c). step 100. 𝛼𝛼=86° (d). step 240. Fig. 29 Some intermediate steps of the optimization process with the initial design shown in Fig. 27a. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 64 𝛼𝛼=90° Fig. 30 The optimized structure obtained with the initial design shown in Fig. 27b under the MMC formulation. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 65 l a n o i t c n u f e v i t c e b O j 1000 800 600 400 200 0 0 Objective functional Volume constraint fucntional Inclined angle constraint functional 0.3 0.2 0.1 0.0 l a n o i t c n u f t i n a r t s n o c l e g n a d e n i l l a n o i t n c u f t i n a r t s n o c e m u o V l 50 100 150 200 250 Iteration step c n I -0.1 300 Fig. 31 Convergence history of the short beam example with the initial design shown in Fig. 27a under the MMC formulation. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 66 Fig. 32 The initial design of the short beam problem under the MMV formulation (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 67 Fig. 33 The initial design of the short beam problem under the MMV formulation (𝒃𝒃𝑝𝑝=(1,0)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 68 (a). The optimized structure (contour plot). (b). The optimized structure (B-spline plot). Fig. 34 The optimized structures obtained with the initial design shown in Fig. 32 under the MMV formulation (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 69 (a). The optimized structure (contour plot). (b). The optimized structure (B-spline plot). Fig. 35 The optimized structure obtained with the initial design shown in Fig. 33 under the MMV formulation (𝒃𝒃𝑝𝑝=(1,0)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 70 (a). step 7. (b). step 103. (c). step 1021. (d). step 1327. Fig. 36 Some intermediate steps of the optimization process with the initial design shown in Fig. 33 (𝒃𝒃𝑝𝑝=(1,0)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 71 l a n o i t c n u f e v i t c e b O j Objective functional Volume constraint fucntional 1000 800 600 400 200 0 0 200 400 800 600 Iteration step 1000 1200 l a n o i t 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 n c u f t i n a r t s n o c e m u o V l 1400 Fig. 37 Convergence history of the short beam example with initial design shown in Fig. 33 under the MMV formulation.. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 72 𝑡𝑡=2 360×60 mesh 6 Baseplate Fig. 38 The MBB example. 1 n o i t c e r i d t n i r P 𝑦𝑦 o 𝑥𝑥 Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 73 Fig. 39 The optimized structure without considering self-supporting requirement. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 74 𝛼𝛼=45° Fig. 40 The initial design of the MBB problem under the MMC-based formulation. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 75 𝛼𝛼=84.87° (a). The optimized structure (contour plot). 𝛼𝛼=84.87° (b). The optimized structure (component plot). Fig. 41 The optimized structure obtained with the MMC-based approach. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 76 Fig. 42 The initial design of the MBB problem under the MMV-based formulation (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 77 46.41° 45.74° 46.48° 𝑪𝑪1 (a). Optimized solution considering self-supporting constraint. (b). A comparison with the unconstrained optimized solution. Fig. 43 The optimized structure obtained with the MMV-based approach (𝒃𝒃𝑝𝑝=(0,1)⊤). Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 78 (a). step 2. (b). step 43. (c). step 337. (d). step 1072. (e). step 1555. Fig. 44 Some intermediate steps of the optimization process. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 79 𝑧𝑧 𝑦𝑦 𝑥𝑥 Fig. 45 A schematic illustration of a 3D printable feature. Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 80 Tab. 1 Optimal solution of the short beam example. The number of active component Inclined angle (local coordinate system ) Inclined angle (global coordinate system ) 1 2 4 6 8 9 12 16 39.18° -14.02° -18.49° 5.13° -7.97° 27.74° -44.23° -19.04° 129.18° 75.98° 71.51° 95.13° 82.03° 117.74° 45.77° 70.96° Computer Methods in Applied Mechanics and Engineering, under review 2016-11-12 81
1712.04218	1	1712	2017-12-12T10:38:08	Metal - TiO2 contacts: An electrical characterization study	[ "physics.app-ph" ]	The electrical properties of thin TiO2 films have recently been extensively exploited towards enabling a variety for metal-oxide electron devices: unipolar/bipolar semiconductor devices and/or memristors. In such efforts, investigations on the role of TiO2 as active material have been the main focus, however, electrode materials are equally important. In this work, we address this need by presenting a systematic quantitative electrical characterisation study on the interface characteristics of Metal-TiO2-Metal structures. Our study employs typical contact materials that are used both as top and bottom electrodes in a Metal-TiO2-Metal setting. This allows investigating the effect of Metal-TiO2 contacts as well as holistically studying an electrode's influence to the opposite interface, referred to as top/bottom electrodes coupling. Our methodology comprises the recording of current-voltage (I-V) characteristics from a variety of solid-state prototypes in the temperature range of 300-350 K and by analysing them through appropriate modelling. Clear field and temperature dependent signature plots were also obtained towards shinning more light on the role of each material as top/bottom electrodes in Metal-TiO2-Metal structures. Our results provide a useful database for selecting appropriate electrode materials in Metal-TiO2 devices, offering new insights in metal-oxide electronics applications.	physics.app-ph	physics	Metal – TiO2 contacts: An electrical characterization study L. Michalas, A. Khiat, S. Stathopoulos, T. Prodromakis Nanotechnology Research Group, Electronics and Computer Science Department, University of Southampton, Southampton, SO17 1BJ, UK. Corresponding Author's email:[email protected] Abstract: The electrical properties of thin TiO2 films have recently been extensively exploited towards enabling a variety of metal-oxide electron devices: unipolar/bipolar semiconductor devices and/or memristors. In such efforts, investigations on the role of TiO2 as active material have been the main focus, however, electrode materials are equally important. In this work, we address this need by presenting a systematic quantitative electrical characterization study on the interface characteristics of Metal-TiO2-Metal structures. Our study employs typical contact materials that are used both as top and bottom electrodes in a Metal-TiO2- Metal setting. This allows investigating the effect of Metal-TiO2 contacts as well as holistically studying an electrode's influence to the opposite interface, referred to as top/bottom electrodes coupling. Our methodology comprises the recording of current-voltage (I-V) characteristics from a variety of solid-state prototypes in the temperature range of 300-350 K and by analysing them through appropriate modelling. Clear field and temperature dependent signature plots were also obtained towards shinning more light on the role of each material as top/bottom electrodes in Metal-TiO2-Metal structures. Our results provide a useful database for selecting appropriate electrode materials in Metal-TiO2 devices, offering new insights in metal-oxide electronics applications. Keywords: TiO2, Metal Electrodes, Schottky Barrier, Interface Coupling, RRAM, TFT, Selectors 1 introducing a new era for to oxide-based photovoltaics9 and sensors10, Introduction Metal-oxides (MOs) combine a unique ensemble of properties presenting great potential to meet the diverse requirements of modern electronics technologies. Specifically they offer low temperature (<100oC) manufacturability1, thus compliance with large-scale uniformity and deployment on alternative materials (paper, flexible); adequate device level mobility2 that can be further optimized by engineering appropriate gate dielectrics; high transparency due to their wide band gap; capacity for post fabrication tuneable resistance (memristive effects)3,4 and good chemical stability. The features enabled the use of MOs in a variety of applications ranging from Resistive Random Access Memories (RRAMs)5,6 and Thin Film Transistors (TFTs)7,8 large area transparent/stretchable electronics11,12 and neuromorphic systems13,14. To this end, a variety of materials have been scrutinised over the recent years with distinct MO and thus properties being exploited according to the application needs, i.e neuromorphic applications favour MO devices that show large memory capacity4. TiO2 is without a doubt one of the most celebrated materials through practical implementations of memristors15, TFTs,16,17, sensors18 and corollary applications of these. The ability of TiO2 to obtain different microstructures (i.e. amorphous, micro/nano crystalline, rutile, anatase etc.) and thus a plethora of electronic properties that can be determined/controlled by the fabrication and/or biasing conditions augmented its use in practical applications. This was further enhanced by the incorporation of foreign metal elements (doping) in TiO2 thin films that were shown to improve RRAM switching characteristics19 but also enable both n- and p- type functionalities20,21. The latter area is still in its infancy, yet, offers substantial prospects for low- temperature manufacturable electronic systems. Research efforts are thus targeted on addressing outstanding challenges22, with first proof-of-concept results on the development of bipolar components entirely based on TiO2 p-n homo-junctions23,24 showcasing the exciting technological potential. Notwithstanding the importance of TiO2 as active layer, identifying appropriate metal contacts and deciphering their interfacial role is now of paramount importance to a device's electrical behaviour. This role becomes even more important in highly scalable thin-film devices, where device properties strongly rely on interfacial effects. These essentially need to be fully understood for optimising a device's electrical response. Up to date, such studies are mainly restricted to room temperature semi-quantitative approaches, focusing only in studying electrodes formed atop of TiO2 films. More specifically, it was reported that recording the rectification ratio measured at read out-voltage of ± 1V for a variety of top electrodes (deposited by RF magnetron sputtering on Metal-TiO2-(Ohmic-Pt) configuration) can reveal the role of a metal's electronegativity on the formation of the interface barrier25. In the same study, barriers evaluated by the forward diode current, suggested partial Fermi level pinning. In addition, the role of the top electrode (evaporated through shadow mask) on a Metal-TiO2-(Ohmic-Pt) memory cell, evaluated with a read-out voltage of ± 1V, was correlated the symmetry/asymmetry of Metal-TiO2-(Ohmic-Al) current-voltage characteristics attributed also to the top metal fabrication details such as the thermal annealing27. More recently, the source and drain electrodes of TiO2 based TFTs were found to play a major role on performance parameters such as the ON/OFF ratio and the field effect mobility17. All the above assessments suggest that this is not a conventional Metal- 2 top metal electrode oxide26. Moreover, to the formation free energy for Semiconductor contact where the metal work function dominates the process and therefore additional studies are still required for an in-depth investigation. In this paper, we present a detailed quantitative electrical characterization study of Metal-TiO2 interface characteristics. The work involves the commonly utilized contact metals acting both as top and bottom contacts in Metal-TiO2-Metal configuration, whilst it also reports the influence of the top electrode (TE) material to the bottom electrode (BE) interface. This is performed by recording the current-voltage (I-V) characteristics at different temperatures and through appropriate modelling and analysis that includes field and temperature dependent signature plots. Considering also the potential applications, the design and fabrication of all prototyped samples, on which this study was based, remained aligned to standard low temperature microelectronic processes, while the electrodes have been patterned through optical lithography and deposited via electron-beam evaporation. Results Device Modelling and Parameters Extraction: The studied Metal-TiO2-Metal devices are presented in Fig. 1(a). In order to obtain quantitative results the tested devices have been modelled by the equivalent circuit presented in Fig. 1 (b). This consists of two series resistances accounting for the top (RTE) and the bottom (RBE) access electrodes and the device under test (DUT). As TiO2 exhibits an intrinsic n-type character, the DUT can be considered as two inverse polarized Schottky diodes, both emulating the TE and BE contacts, connected through a resistance that corresponds to the TiO2 core. This equivalent circuit was further used for calculating the effective applied bias (Veff)/electric field across the DUT, with respect to the bias applied through the voltage source (V) and as a function of the measured current (I), summarised by Eq.1 𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒=𝑉𝑉−𝐼𝐼(𝑅𝑅𝑇𝑇𝑇𝑇+𝑅𝑅𝐵𝐵𝑇𝑇) (1) In the absence of an interface barrier, the Schottky diodes (TE/BE) can be considered as short-circuit. In this case the I-V characteristic should be symmetric with respect to the applied bias polarity and the dominant transport mechanism is purely determined by the properties of the active layer film. For all other cases, where an interface barrier is formed, any positive bias applied to the TE, will result in a forward biased TE/TiO2 and reversed biased TiO2/BE contact. In this case, the current flow will be determined by the reversed biased BE, given that this constitutes the most resistive element in this series configuration. The situation is the opposite when a negative bias is used at the TE; the TE determining the current flow. The interface-controlled transport is typically identified by an asymmetric I-V characteristic with respect to the applied bias polarity. The transport is then dominated by either tunnelling through the barrier formed at the interface or by thermionic emission over it. Tunnelling is thus expected in cases where carriers do not possess sufficient energy to overcome the potential barrier. The two mechanisms can be distinguished via the temperature dependence of the I-V28. Tunnelling currents obey the following Eq. 2: 𝐼𝐼∝𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒2 𝑒𝑒𝑒𝑒𝑒𝑒�− 𝑏𝑏𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒� (2) 3 Figure 1: The studied Metal-TiO2-Metal devices (a). The equivalent circuit utilized to model the devices and the electrodes for our (a) (b) quantitative analysis. with "b" being just a constant. Consequently, an asymmetric, temperature independent I-V, supporting also a linear relation on a ln(I/Veff 2) vs 1/Veff signature plot serves as a strong indication of tunnelling dominated transport. On the contrary thermionic emission over the interface barrier is the most commonly thermally activated mechanism for interface controlled transport and can be described by Eq. 3: 𝐼𝐼=𝐴𝐴𝑇𝑇2𝑒𝑒𝑒𝑒𝑒𝑒�−Φ𝐵𝐵0−𝑎𝑎�𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒 𝐾𝐾𝑇𝑇 � (3) where K is the Boltzmann constant, T is the absolute temperature, ΦΒ0 is the zero bias potential barrier height at Metal/MO interface and A=(Area x A), with A being the Richardson constant and "α" the barrier lowering factor. A straight line on a ln(I/T2) vs 1000/T plot for any applied electric field serves as strong evidence of thermionic emission. The slope of this line corresponds to the apparent interface barrier (ΦApp) and should decrease by increasing the applied electric field/bias as: Φ𝐴𝐴𝐴𝐴𝐴𝐴=Φ𝐵𝐵0−𝛼𝛼�𝑉𝑉𝑒𝑒𝑒𝑒𝑒𝑒 (4) 1/2 signature plot is also supported by the extracted results, then the intercept extrapolated to If this ΦApp vs Veff Veff =0 Volts, provides an experimental determination of ΦΒ0, whilst the slope corresponds to "α". By following the aforementioned methodology we can obtain clear signature plots that allow deciphering the transport mechanism and provide quantitative measures of interfaces in Metal-TiO2-Metal structures. The Bottom Electrode: It is important to note that it is possible for the bottom electrode-TiO2 interface to show a distinct behaviour to an identical interface lying at the top of a device, due to processing induced effects. The up to date published works have typically utilized ohmic junctions for the bottom electrode, solely targeting the influence on the top interface characteristics and therefore assuming similar signatures for both interfaces due to symmetry. Nonetheless, we recently demonstrated that even symmetric structures may overall render an asymmetric performance29. 4 (b) (d) (f) (a) (c) (e) Figure 2: Devices having identical TiN TE and different BE were assessed (a). Tunnelling identified in case of Pt BE signature plots (b),(c) and Schottky emission when Ni is the BE (e), (f). The I-V characteristics of structures having Au, Ti, Ni and Pt as BE are presented in Fig. 2(a). In all cases the TE is TiN deposited by sputtering. By observation to the acquired symmetric I-Vs, it follows that no interface barrier is formed in the cases where Au and Ti were utilized as bottom electrodes, denoting that the overall conductivity is controlled by the TiO2 thin-film. Besides, the two structures exhibit different resistance measured at read-out voltage of 1V (Table 1), suggesting that the BE material plays a major role on the deposited film conductivity. In cases where Pt and Ni were utilized as BE, strongly asymmetric I-V characteristics where obtained, denoting an interface controlled transport. In these cases, the I-V characteristics 5 have been further assessed in the temperature range of 300 K to 350 K in order to obtain the corresponding signature plots. The applied biases have been limited up to 1.5 V and 1 V respectively to avoid the devices undergoing either a soft or irreversible (hard) breakdown. For the sample with Pt BE (Fig. 2(b)), the part of I- V plot corresponding to the BE (positive biases) presents a temperature independent behaviour (highlighted in the red square). This along with the signature plot presented in Fig. 2(c), supports that tunnelling is the major mechanism responsible for the conductivity, indicating the presence of a very high barrier at this interface that is however immeasurable via Eq. 2. On the contrary for the device formed atop of Ni BE (Fig. 2(d)), clear signature plots have been obtained that confirm thermionic emission over the interface barrier as the dominant transport mechanism. The data analysis allowed by the clear signature plots, led to the estimation of the zero bias potential barrier at the interface as ΦΒ0 = 0.69 eV. The results for the BE study are summarized in table 1 along with the metal work function and electronegativity as those parameters may determine the interface barrier. We note however that no correlation can be obtained for the BE. BE material Wf (eV) Au Ti Ni Pt 5.1 4.1 5.15 5.65 Xm 2.54 1.54 1.91 2.28 Interface Barrier No Barrier No Barrier 0.69 eV Very high Additional Information R (at 1V) = 8.5 KΩ R (at 1V) = 150 KΩ α = 0.511 eV1/2 Tunnelling Table: 1: Quantitative results extracted from the experimental data regarding the BE materials The Top Electrode: The TE/TiO2 contacts have been assessed to date mainly through evaluating the DUT's resistance at a specific read-out voltage, typically of 1V. Here we are aiming to present an approach focusing more on quantitative extraction of parameters with physical meaning i.e. the zero bias potential barrier at the interface, ΦΒ0. For this, several configurations having different TE have been fabricated on the same wafer, with the TiO2 active layer films deposited on top of Au BE during the same process step. Therefore, with the exception to the TE interface, identical characteristics are expected. On the contrary, the results depicted in Fig. 3 provide evidence that the TE deposition plays indeed a major role on the device characteristics. Ti and Al TE, result in no barrier formation (Fig. 3(a)), while an interface barrier is formed by Au, Pt and Ni (Fig. 3(b)). For these materials, detailed temperature characterization was performed and clear signature plots were extracted by analysing the corresponding I-V curves (Fig. 4). This led to the calculation of several interface quantities (Fig. 5(a)) that are summarized along with resistance of the "ohmic" contacts in Table 2. Considering these experimental results, several interesting features can be noted. Starting from Al and Ti that do not form any interface barrier, it was found that they both affect the conductivity of the structure resulting in a different resistance measured at 1V. 6 (a) (b) Figure 3: Al and Ti as TE result in symmetric I-V denoting no formation of interface barrier (a). Asymmetric I-Vs obtained in cases of Au, Pt and Ni suggesting interface controlled conduction mechanism (b). TE material Wf (eV) Al Ti Pt Ni Au 4.28 4.1 5.65 5.15 5.1 Xm 1.61 1.54 2.28 1.91 2.54 Interface Barrier at TE Additional Information No Barrier No Barrier 0.61 eV 0.50 eV 0.70 eV R (at 1V) = 1 KΩ R (at 1V) = 430 KΩ α = 0.265 eV1/2 α = 0.208 eV1/2 α = 0.203 eV1/2 Table 2: Quantitative results extracted from the experimental data regarding the TE materials This behaviour maybe supported by the mechanism proposed by in Ref. 26 and the Ellingham diagram, where the argument is that electrode metals interact with TiO2 generating oxygen vacancies. Oxygen vacancies act as n-type dopants, providing free electrons in the MO conduction band. Al should induce more oxygen vacancies with respect to Ti and thus results in a more conductive behaviour. Indeed, in cases of "ohmic" contacts and when the transport in the film is dominated by band conduction due to free electrons, the average conductivity is proportional to the free carriers concentration (i.e. σ = qnµ, for n-type material, where n denotes the free electron density, q the electronic charge and µ the band mobility) and thus to the amount of the oxygen vacancies induced by the electrodes. However, this might not be straightforward in cases where the film conductivity is not dominated by the free carriers. In such cases, that also hold for wide band gap materials (which is the expected native condition for TiO2), additional mechanisms should be considered, such as hopping or Frenkel-Poole, depending on the band-gap states density and energy distribution, the operation temperature and the applied field intensity. The situation becomes more complicated in the case of Ni, Au Pt, where the conductivity is controlled by thermionic emission over the interface barriers, as confirmed by the signature plots. The presence of oxygen vacancies may affect the depletion layer width and thus indirectly the interface barrier. However, the barrier height is also affected by the fermi level position, which is mainly determined by the trap charge at the interface. Moreover, according to Eq. 3 the current conduction depends exponentially on the apparent barrier, 7 including the major role of its lowering due to the applied bias, determined by the interface controlled factor "α". Thus in the present study, the non-ohmic contacts are assessed through the estimation of the zero bias potential barrier, ΦΒ0, extracted in Fig. 5 (a) through Eq. 4, providing straightforward quantitative information on the interface characteristics. (a) (c) (e) Figure 4: I-V curves vs temperature for Ni (a), Pt (b) and Au (c), TE and the corresponding signature plots (b), (d) and (f) respectively, confirming Schottky emission as the dominant transport mechanism. Considering the interface barrier, this should be typically proportional to the metal work function. In cases, however, where large amount of defect states are present at the interface, these could significantly affect this dependence. Pinned Fermi position resulting in barrier heights independent of the metal work function are 8 (b) (d) (f) commonly obtained, as for example in the case of III-V compounds. Nonetheless, this does not apply to our work as while the barrier is formed it is not proportional to the metal work function; significant variations exist for distinct materials. The correlation of the formed barrier to the electronegativity has been reported for ionic semiconductors28, where the strength of the correlation is typically assessed by the index S: 𝑆𝑆=𝑑𝑑Φ𝐵𝐵0𝑑𝑑𝑋𝑋𝑚𝑚 (5) A strong correlation to electronegativity is considered as the S index approaches 1. For example, Silicon results in S=0.05 while SiO2 to S=128. A clear linear relation is supported by our results depicted in Fig. 5 (b). The calculated index is S= 0.32. Similar results extracted in a completely different way are presented in Ref. 25 that resulted to S= 0.55 including materials, such as Al and Ti, which in our case did not form a barrier. This is referred to as partial Fermi pinning that showcases the important role of interface states. (a) (b) Figure 5: Experimental verification for the conduction mechanism and extraction of the zero bias potential barrier at the top interface (a). The barrier is found to increase linearly with the electronegativity of the TE material (b). It is worth noting that while this feature was identified for the TE, it is not the case for the BE (Table 1), highlighting the differences in the formation mechanism for the two interfaces. Furthermore, an issue that has not been pointed out to date, is the obtained coupling between the two interfaces. Throughout this study, we employ this term for highlighting that the deposition of the TE also results in modifications to the BE interfaces (positive biases in the I-V). Considering that the BE and the TiO2 should be identical from the fabrication point of view, we consider this an interesting finding and therefore it is discussed in more detail in the following section. Coupling Issues: The TE deposition is considered to affect the bottom Au/TiO2 contacts (Fig. 3, positive biases) that are nominally identical for all employed prototype devices. In cases of Al and Ti, no barrier is formed and therefore the conductivity is mainly determined by the film properties. In contrast, depositing Ni, Pt or Au as TE, appears to also instigate the formation of a barrier at the bottom interface, offering the opportunity to perform a quantitative study regarding the role of the TE to the BE interface. Therefore, the BE/TiO2 interfaces have been characterized in detail following the previously employed temperature dependent methodology. The obtained results for the Au/TiO2 BE with respect to the TE material are 9 summarized in table 3, while the corresponding signature plots are presented in Fig. 6. A linear relation to the electronegativity was also obtained for the BE interface barriers. In this case, the barriers formed are lower than the TE ones but follow the same trend. The data presented in Fig. 7 support an S= 0.48. This clearly depicts the influence of the TE material throughout the TiO2 all the way to the BE, but also signifies the major role of the interface states. TE material Wf (eV) Al Ti Pt Ni Au 4.28 4.1 5.65 5.15 5.1 Xm 1.61 1.54 2.28 1.91 2.54 Interface Barrier at BE Additional Information No Barrier No Barrier 0.54 eV 0.37 eV 0.67 eV R (at 1V) = 1 KΩ R (at 1V) = 430 KΩ α = 0.279 eV1/2 α = 0.161 eV1/2 α = 0.236 eV1/2 Table 3: Quantitative results extracted from the experimental data regarding the bottom interface with respect to the TE materials (a) (b) Figure 6: Signature plots (I-Vs at Fig. 4) correspond to the bottom interface, for Ni (a), Pt (b) and Au (c) TE, confirming Schottky emission as the dominant transport mechanism. (c) 10 (a) (b) Figure 7: Further experimental verification for the conduction mechanism and extraction of the zero bias potential barrier at the interface for the bottom interface (a). The barrier is found to increase linearly with the electronegativity of the TE material (b). Discussion This study has provided several new insights on the electrical properties of Metal-TiO2 interfaces. Evidently, these cannot be considered as conventional Schottky contacts, where the conduction is mainly determined by the metal work function. In the case of Au BE, two different regimes were identified for TE. When no interface barrier is formed (Al and Ti) the film conductivity is enhanced for higher formation free energy of the metal oxide, possibly due to an increase of oxygen vacancies, as proposed in Ref. 26. In cases where an interface barrier exists, its height is found to be proportional to the metal electrode electronegativity. The proportionality factor (S index) is far from 1, thus denoting the important role of the interface states. This study also came across an unexpected observation (interface coupling), where the BE interface barrier was found to be determined by the TE material, following accordingly ohmic or electronegativity dependent barrier. This requires that the metal electrodes are selected in pairs, as the TE defines the properties of the bottom interface as well. The opposite trend, i.e. the BE barrier determined by the electronegativity is not however true. It is therefore concluded that different mechanism are responsible for the formation of the interfaces in cases of BE, where the TiO2 film is deposited on the metal, in contrast to the TE interfaces where the metal is deposited atop the TiO2 film. As a result, each material should be carefully considered with respect to its utilization as TE or BE; although in some case like Ti, no interface barrier is formed for either configuration. Furthermore, the device conductivity and the top interface characteristics are also found to be affected by the BE material, supporting our argument that electrode materials should always be selected in pairs. Considering the growth of TiO2-based electronic devices and the necessity for metal electrodes on top or below their active area, we proceed by sharing some thoughts on the usefulness of our results across the diverse MO applications. In the field of RRAM, where typically the devices are utilized after an electroforming step, Schottky barriers at the TE are considered essential for supporting resistive switching30. It was also reported that high barriers at TE support both Unipolar and Bipolar resistive switching, in contrast to lower barriers that typically favour bipolar operation31. The proper selection of metal electrodes can be utilized for engineering the electroforming process, in terms of polarity and applied field intensity, and perhaps even developing forming-free RRAM technologies. The electrodes' characteristics are also of major importance to 11 semiconductor devices. For example, when considering TiO2-based bipolar devices it is essential to avoid any rectification stemming from the metal electrodes. Moreover, in the case of TFTs, different configurations may require the Source and Drain electrodes atop or below the deposited MO films defining the TFTs' active area and thus should be accordingly selected. Finally, several applications require the monolithic integration of such technologies. A prominent example is the use of diodes/transistors as selecting elements32 for RRAM, rendering 1D1R and/or 1T1R configurations. Such cases typically exploit common metal electrodes acting both as BE and TE for different devices and thus should be accordingly considered, bearing also in mind the coupling issues. Methods Device Fabrication: The tested devices were fabricated on two separate 6-inch Si wafers having a 200 nm thick SiO2 layer that was grown by dry oxidation. All electrodes and active areas where patterned via standard optical lithography. The first wafer was dedicated for studying the BE influence, with four different areas utilised for depositing 15 nm of Ti, Ni, Pt and Au. These metals were deposited via electron beam evaporation, on top of 5 nm thick Ti adhesion layer. The second wafer enabled the study of the TE and thus Au was utilized as BE. 25 nm of amorphous TiO2 films were deposited by reactive sputtering (Helios XP, Leybold optics) using a Ti target in oxygen plasma environment. 15nm thick TiN was sputtered everywhere across the first wafer, serving as TE. For the second wafer five separate areas were defined and 15nm thick Al, Ti, Ni, Pt and Au films were respectively deposited via electron beam evaporation. I-V measurements: The current vs voltage (I-V) characteristics were obtained on 30 x 30 μm2 devices using our in-house memristor characterization platform ArC ONETM 33 . The voltage sweeping was carried out in staircase mode towards positive biases, while both positive and negative polarities were always applied to the top electrode (TE) with respect to the bottom electrode (BE) that was continuously kept grounded. All experiments were performed on a Cascade SUMMIT 12000B semi-automatic probe station that incorporates a thermal chuck, whose temperature can be controlled by an ESPEC ETC-200L unit. Measurements were performed in the temperature range of 300 K to 350 K. References 1. Yu, X., Marks, T. J. & Facchetti, A. Metal oxides for optoelectronic applications. Nat. Mater. 15, 383– 396 (2016). Brox-Nilsen, C., Jidong, J., Yi, L., Peng, B. & Song, A. M. Sputtered ZnO Thin-Film Transistors With Carrier Mobility Over 50 cm2/ Vs. Electron Devices, IEEE Trans. 60, 3424–3429 (2013). Kim, W. et al. Multistate Memristive Tantalum Oxide Devices for Ternary Arithmetic. Sci. Rep. 6, 36652 (2016). Stathopoulos, S. et al. Multibit memory operation of metal-oxide bi-layer memristors - Supplementary information. ArXiV 1704.03313, 2–8 (2017). Ielmini, D. Resistive switching memories based on metal oxides: mechanisms, reliability and scaling. 12 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. Semicond. Sci. Technol. 31, 063002 (2016). Sawa, A. Resistive switching in transition metal oxides. Mater. Today 11, 28–36 (2008). Park, J. S., Maeng, W. J., Kim, H. S. & Park, J. S. Review of recent developments in amorphous oxide semiconductor thin-film transistor devices. Thin Solid Films 520, 1679–1693 (2012). Kamiya, T., Nomura, K. & Hosono, H. Origins of High Mobility and Low Operation Voltage of Amorphous Oxide TFTs: Electronic Structure, Electron Transport, Defects and Doping. J. Disp. Technol. 5, 273–288 (2009). Rühle, S. et al. All-oxide photovoltaics. J. Phys. Chem. Lett. 3, 3755–3764 (2012). Fine, G. F., Cavanagh, L. M., Afonja, A. & Binions, R. Metal oxide semi-conductor gas sensors in environmental monitoring. Sensors 10, 5469–5502 (2010). Petti, L. et al. Metal oxide semiconductor thin-film transistors for flexible electronics. Appl. Phys. Rev. 3, (2016). Liu, Y., Pharr, M. & Salvatore, G. A. A Lab-on-Skin: A Review of Flexible and Stretchable Electronics for Wearable Health Monitoring. ACS Nano acsnano.7b04898 (2017). doi:10.1021/acsnano.7b04898 Serb, A. et al. Unsupervised learning in probabilistic neural networks with multi-state metal-oxide memristive synapses. Nat. Commun. 7, 12611 (2016). 14. Mehonic, A. & Kenyon, A. J. Emulating the electrical activity of the neuron using a silicon oxide 15. 16. 17. RRAM cell. Front. Neurosci. 10, (2016). Strukov, D. B., Snider, G. S., Stewart, D. R. & Williams, R. S. The missing memristor found. Nature 459, 1154–1154 (2009). Shih, W. S., Young, S. J., Ji, L. W., Water, W. & Shiu, H. W. TiO2-Based Thin Film Transistors with Amorphous and Anatase Channel Layer. J. Electrochem. Soc. 158, H609 (2011). Choi, H., Shin, J. & Shin, C. Impact of Source/Drain Metal Work Function on the Electrical Characteristics of Anatase TiO2 -Based Thin Film Transistors. ECS J. Solid State Sci. Technol. 6, 379– 382 (2017). Bai, J. & Zhou, B. Titanium dioxide nanomaterials for sensor applications. Chem. Rev. 114, 10131– 10176 (2014). Trapatseli, M. et al. Engineering the switching dynamics of TiOx-based RRAM with Al doping. J. Appl. Phys. 120, (2016). Zhao, L., Park, S. G., Magyari-Köpe, B. & Nishi, Y. Dopant selection rules for desired electronic structure and vacancy formation characteristics of TiO2 resistive memory. Appl. Phys. Lett. 102, (2013). 21. Anitha, V. C., Banerjee, A. N. & Joo, S. W. Recent developments in TiO2 as n- and p-type transparent semiconductors: synthesis, modification, properties, and energy-related applications. J. Mater. Sci. 50, 7495–7536 (2015). 20. 18. 19. 22. Wang, Z., Nayak, P. K., Caraveo-Frescas, J. A. & Alshareef, H. N. Recent Developments in p-Type Oxide Semiconductor Materials and Devices. Adv. Mater. 28, 3831–3892 (2016). 23. Hazra, A., Chattopadhyay, P. P. & Bhattacharyya, P. Hybrid Fabrication of Highly Rectifying p - n Homojunction Based on Nanostructured TiO2. IEEE Electron Device Lett. 36, 505–507 (2015). 13 24. Vasu, K., Sreedhara, M. B., Ghatak, J. & Rao, C. N. R. Atomic Layer Deposition of p-Type Epitaxial Thin Films of Undoped and N-Doped Anatase TiO2. ACS Appl. Mater. Interfaces 8, 7897–7901 (2016). Zhong, N., Shima, H. & Akinaga, H. Rectifying characteristic of Pt/TiOx/metal/Pt controlled by electronegativity. Appl. Phys. Lett. 96, (2010). 25. 26. Yang, J. J. et al. Metal/TiO2 interfaces for memristive switches. Appl. Phys. A Mater. Sci. Process. 102, 785–789 (2011). 27. Hossein-Babaei, F., Lajvardi, M. M. & Alaei-Sheini, N. The energy barrier at noble metal/TiO2 junctions. Appl. Phys. Lett. 106, (2015). S.M., Sze & K.K., Ng Physics of Semiconductor Devices. (John Wiley & Sons, 2006). 28. 29. Michalas, L. et al. Interface asymmetry induced by symmetric electrodes on metal-Al:TiOx-metal structures. Nanotechnology, IEEE Trans. - accepted 30. Hernández-Rodríguez, E. et al. Effect of electrode type in the resistive switching behaviour of TiO2 thin films. J. Phys. D. Appl. Phys. 46, (2013). 31. Kim, W.-G. & Rhee, S.-W. Effect of the top electrode material on the resistive switching of TiO2 thin 32. 33. film. Microelectron. Eng. 87, 98–103 (2010). Cortese, S., Khiat, A., Carta, D., Light, M. E. & Prodromakis, T. An amorphous titanium dioxide metal insulator metal selector device for resistive random access memory crossbar arrays with tunable voltage margin. Appl. Phys. Lett. 108, (2016). Berdan, R. et al. A μ-Controller-Based System for Interfacing Selectorless RRAM Crossbar Arrays. IEEE Trans. Electron Devices 62, 2190–2196 (2015). Acknowledgements This work was supported by the EPSRC Grant EP/K017829/1. Author Contribution L.M and T.P planned the experiments, L.M. performed the electrical characterization and analysed the results. A.K and S.S. optimised the processes and fabricated the samples. L.M. and T.P. wrote the manuscript. A.K. and S.S. provided useful suggestions for improving it. Data Availability All data supporting this study are openly available from the University of Southampton repository at: (to be added in the final form) Additional Information Competing Interests: The authors declare that they have no competing interests 14
1909.01556	1	1909	2019-09-04T05:17:03	Interfacial Thermal Conductance across Room-Temperature Bonded GaN-Diamond Interfaces for GaN-on-Diamond Devices	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	The wide bandgap, high-breakdown electric field, and high carrier mobility makes GaN an ideal material for high-power and high-frequency electronics applications such as wireless communication and radar systems. However, the performance and reliability of GaN-based HEMTs are limited by the high channel temperature induced by Joule-heating in the device channel. High thermal conductivity substrates integrated with GaN can improve the extraction of heat from GaN based HEMTs and lower the device operating temperature. However, heterogeneous integration of GaN with diamond substrates is not trivial and presents technical challenges to maximize the heat dissipation potential brought by the diamond substrate. In this work, two modified room temperature surface activated bonding techniques are used to bond GaN and single crystal diamond with different interlayer thicknesses. TDTR is used to measure the thermal properties from room temperature to 480 K. A relatively large TBC of the GaN-diamond interfaces with a 4nm interlayer was observed and material characterization was performed to link the structure of the interface to the TBC. Device modeling shows that the measured GaN-diamond TBC values obtained from bonding can enable high power GaN devices by taking the full advantage of the high thermal conductivity of single crystal diamond and achieve excellent cooling effect. Furthermore, the room-temperature bonding process in this work do not induce stress problem due to different coefficient of thermal expansion in other high temperature integration processes in previous studies. Our work sheds light on the potential for room-temperature heterogeneous integration of semiconductors with diamond for applications of electronics cooling especially for GaN-on-diamond devices.	physics.app-ph	physics	Interfacial Thermal Conductance across Room-Temperature Bonded GaN- Diamond Interfaces for GaN-on-Diamond Devices Zhe Cheng,1,a) Fengwen Mu,2,3,a),* Luke Yates,1 Tadatomo Suga,2 Samuel Graham1,4,* 1 George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332, USA 2 Collaborative Research Center, Meisei University, Hino-shi, Tokyo 191-8506, Japan 3 Kagami Memorial Research Institute for Materials Science and Technology, Waseda University, Shinjuku, Tokyo 169-0051, Japan 4 School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332, USA a) These authors contribute equally * Corresponding authors: [email protected]; [email protected] Abstract The wide bandgap, high-breakdown electric field, and high carrier mobility makes GaN an ideal material for high-power and high-frequency electronics applications such as wireless communication and radar systems. However, the performance and reliability of GaN-based high electron mobility transistors (HEMTs) are limited by the high channel temperature induced by Joule-heating in the device channel. High thermal conductivity substrates (e.g., diamond) integrated with GaN can improve the extraction of heat from GaN-based HEMTs and lower the device operating temperature. However, heterogeneous integration of GaN with diamond substrates is not trivial and presents technical challenges to maximize the heat dissipation potential brought by the diamond substrate. In this work, two modified room-temperature surface-activated bonding (SAB) techniques are used to bond GaN and single crystal diamond with different interlayer thicknesses. Time-domain thermoreflectance (TDTR) is used to measure the thermal properties from room temperature to 480 K. A relatively large thermal boundary conductance (TBC) of the GaN-diamond interfaces with a ~4-nm interlayer (~90 MW/m2-K) was observed and material characterization was performed to link the structure of the interface to the TBC. Device modeling shows that the measured GaN-diamond TBC values obtained from bonding can enable high power GaN devices by taking the full advantage of the high thermal conductivity of single crystal diamond and achieve excellent cooling effect. Furthermore, the room-temperature bonding process in this work do not induce stress problem due to different coefficient of thermal expansion in other high temperature integration processes in previous studies. Our work sheds light on the potential for room-temperature heterogeneous integration of semiconductors with diamond for applications of electronics cooling especially for GaN-on-diamond devices. Introduction With wide bandgap, high-breakdown electric field, and high carrier mobility, GaN has been used for high-power and high-frequency electronics applications such as wireless communication, satellite communication, and radar systems.1 The maximum output power of GaN-based HEMTs is limited by the high channel temperature induced by localized Joule-heating, which degrades device performance and reliability.2,3 Diamond has the highest thermal conductivity among natural materials and is of interest for integration with GaN to help dissipate the generated heat from the channel of GaN-based HEMTs.2,4-6 Current techniques involve two ways to integrate GaN with diamond. One is direct growth of chemical vapor deposited (CVD) diamond on GaN with a transition layer of dielectric material.7 The nanocrystalline diamond near the nucleation interface has reduced thermal conductivity (tens of W/m-K) which could contribute to an additional thermal resistance of 10 m2K/GW.8 Combining with the GaN-diamond thermal boundary resistance, these near-hotspot thermal resistances have been shown to have a large impact on impeding the flow of heat from the device channels, especially for high frequency applications in which the thermal penetration depth is small.9-12 The high growth temperature of the diamond also induces large residual stress in the GaN because of the mismatch of the coefficients of thermal expansion.13,14 Stresses generated during the direct growth of diamond on GaN have been shown to vary, dependent upon GaN thickness, diamond growth temperature, and the sacrificial carrier wafer15,16. Residual stresses greater than 1 GPa at the free surface of the GaN have been reported.17 The elevated stress conditions in the GaN ultimately limit the total thickness and material quality of the GaN by inducing layer cracking and wafer bow, and impact the electrical performance of the device.18-21 The aforementioned effects of elevated stress cause a significant reliability concern when considering the function and lifetime of a GaN device. Another method is high temperature bonding of GaN with diamond.22,23 The GaN device is firstly grown on silicon substrates. Then the GaN device is transferred and bonded with a CVD diamond with an adhesion layer. The transfer and bonding processes are performed at elevated temperatures exceeding 700 oC.23 The adhesion layer increases the thermal resistance of GaN-diamond interface, which offsets the effect of the high thermal conductivity of diamond substrates. The stress due to the different coefficients of thermal expansion results in wafer bow and warp, even fractures.23 Even though some attempts have been made to bond GaN with diamond at lower temperatures4,24, additional techniques to integrate GaN with diamond substrates are in demand to be developed to take full advantage of the high thermal conductivity of diamond without inducing additional stress resulting from high temperature processes. In this work, we use two modified SAB techniques (one with Si nanolayer sputtering deposition and the other with Si-containing Ar ion beam) to bond GaN with diamond substrates with different interlayers at room temperature. TDTR is used to measure the thermal properties. Materials characterization such as high-resolution scanning transmission electron microscopy (HR-STEM) and electron energy loss spectroscopy (EELS) are used to study the interface structure and chemistry to help elucidate the measured thermal properties. An analytical modeling for devices is performed to estimate the device cooling performance of these bonded interfaces. Results and Discussion In this work, GaN was bonded to diamond using a Si interlayer to aid in the chemical adhesion at the interface. The first sample, Samp1, is comprised of a thin layer of GaN (~700 nm) bonded on a commercial single crystal diamond substrate (grown by CVD and purchased from EDP Corporation) with ~10-nm-thick Si interlayer. The Si interlayer can lower TBC, so a different bonding method is applied to Samp2 which has a ~1.88-μm-thick GaN bonded onto a commercial single crystal diamond substrate grown by a high-pressure high-temperature (HPHT) method and purchased from Sumitomo Electric Industries, Ltd. In this sample, the Si-containing Ar ion beam is employed to introduce a ~2-nm-thick interlayer to enhance the interfacial chemical interaction between GaN and diamond. More details about the bonding process can be found in Experimental Section and references25,26. To create the sample for measuring the TBC between the GaN and diamond by TDTR as shown in Figure 1(a), a layer of Al was deposited on the sample surface as a TDTR transducer. With this sample structure, the thermal conductivity of the single crystal diamond substrates were measured first on the area without GaN. Then TDTR measurements were performed on the area with the GaN layer to measure the GaN-diamond TBC. The measured thermal conductivity of the diamond substrates was used as a known parameter in the TDTR data fitting to extract the TBC when measuring over the GaN layer. Overall, there are three unknown parameters: Al-GaN TBC, GaN thermal conductivity, and GaN-diamond TBC. As shown in Figure 1 (b), the TDTR sensitivity of these three unknown parameters are large which is good for accurate thermal measurements. TDTR is a pump-probe technique which can measure thermal properties of both nanostructured and bulk materials.27,28 A modulated pump laser beam heats the sample surface while a delayed probe beam detects the surface temperature variation via thermoreflectance.11 The signal which is picked up by a photodetector and a lock-in amplifier is fitted with an analytical heat transfer solution to infer unknown parameters. More details about TDTR measurements on similar sample structures can be found in the literature.10,26,28,29 An example of the agreement between experimental data (Exp) and the analytical heat transfer solution (Theo) in TDTR data fitting is shown in Figure 1 (c). (a) Al-GaN TBC GaN κ GaN-C TBC TDTR laser Measure diamond κ Al GaN Diamond (b) (c) Figure 1. (a) TDTR measurements on the diamond and bonded GaN-diamond samples. (b) TDTR sensitivity of the three unknown parameters of Samp2. (c) TDTR data fitting of Samp2 with modulation frequency of 2.2 MHz at room temperature. Figure 2(a-b) show the temperature dependence of the measured thermal conductivity of the diamond and the GaN layer and are compared with literature values. Our measured diamond and GaN thermal conductivity match with literature values.30-34 The light yellow color of HPHT sample shows relatively high concentration of impurities while the CVD diamond sample used in this work is transparent. The CVD diamond sample has a higher thermal conductivity than the HPHT sample as expected and matches well with literature values of high-purity diamonds. 30,31 The measured thermal conductivity of the GaN layer (~1.88 μm) in Samp2 is close to experimentally measured bulk values and lower than density-function-theory calculated (DFT) values because of impurities in these samples. The slightly larger thermal conductivity difference between experimental values and DFT values at high temperatures is due to higher order of phonon scatterings such as four phonon scattering.32 Phonon-phonon scattering dominates in thermal transport at high temperatures. Because of the large phonon bandgap of GaN (the optical phonons have much larger energy than the acoustic phonons), the three-phonon scattering process among acoustic and optical phonons are limited (energy conservation during phonon scattering process). Four-phonon scattering process which is not included in the DFT calculation in the reference33 becomes relatively important, leading to an overestimated thermal conductivity. The thermal conductivity of the GaN thin film (~700 nm) in Samp1 does not have good TDTR sensitivity so literature values are used in the TDTR data fitting.26,35 (a) (b) (c) (d) Figure 2. (a) temperature dependence of the measured thermal conductivity of two diamond substrates: Samp1 (CVD) and Samp2 (HPHT). The DFT values and the measured thermal conductivity of Ref. CVD1, CVD2, and natural are from literature.30,31 (b) temperature dependence of the measured thermal conductivity of GaN layer. "DFT", "Exp-1", and "Exp-2" are DFT- calculated thermal conductivity and two experimentally measured thermal conductivity of bulk GaN from literature.32-34 (c) temperature dependence of the measured TBC of bonded GaN- diamond interfaces. (d) phonon density of state of GaN, Si, and diamond.10,36 The temperature dependence of the measured TBC of room-temperature bonded GaN-diamond interfaces are shown in Figure 2(c). Samp2 has a much higher TBC (92 MW/m2-K) than Samp1 (53 MW/m2-K) at room temperature because of the thinner interlayer. For the data fitting of the TDTR measurements in this work, the Si interlayer is very thin, so its thermal resistance is added to the total thermal resistance of the interface. The measured TBC is the effective TBC of the GaN- Si-diamond architecture at the interface (GaN-Si interface + Si layer + Si-diamond interface). The measured TBC of both samples show weak temperature dependence in the measured temperature range (295 K to 480 K). The Debye temperatures of GaN, Si, and diamond are higher than 480 K.10,37 For perfect GaN-Si-diamond interfaces, effective TBC should increase with temperature in the range of 295 K to 480 K because an increasing number of phonons are involved in the thermal transport across the GaN-diamond interfaces as temperature increases. However, for the bonded GaN-Si-diamond interfaces in this work, the disorder at the interface and Ar atoms trapped at the interface increases phonon scattering, which possibly explains the weak temperature dependence of the measured TBC. More details about the interfacial structure-thermal property relationship will be discussed later. Figure 2(d) shows the phonon density of states (DOS) of GaN, Si, and diamond.10,36 Generally speaking, large phonon DOS overlap leads to large TBC. A thin interlayer with a max phonon frequency between two adjacent materials has been reported to cause an increase in TBC.38,39 As shown in Figure 2(d), the max phonon frequency of Si is lower than those of both GaN and diamond. The Si interlayer would reduce the TBC of GaN-diamond interfaces. Therefore, the TBC of bonded GaN-diamond interfaces still have the potential to be improved by using other interfacial layers such as SiC, AlN, or SiNx, even though our measured TBC for Samp2 is already among the high TBC for GaN-diamond interfaces.6,40 Figure 3. (a-b) Cross-section bright-field (BF) and high-angle annular dark-field (HAADF) HR- STEM images of GaN-diamond interfaces of Samp1. (c-d) Cross-section bright-field (BF) and high-angle annular dark-field (HAADF) HR-STEM images of GaN-diamond interfaces of Samp2. To help elucidate the measured TBC and its relationship to the sample architecture, high-resolution scanning transmission electron microscopy (HR-STEM) and electron energy loss spectroscopy (EELS) are used to study the structure of the GaN-diamond interfaces. As shown in Figure 3(a-b), the GaN-diamond interfaces of Samp1 are composed as two layers of amorphous deposited silicon and one amorphous diamond layer. Similar to a previous study in the literature25, no amorphous HAADF-STEM~5.0 nm~5.0 nm~3.0 nm5 nm5 nmSi layerSi layerAmorphous diamondGaNDiamond(a)BF-STEM(b)GaNDiamondAmorphous diamond~4.2 nm~2.2 nmGaNDiamondGaNDiamond5 nm5 nm(c)(d)HAADF-STEMBF-STEM~2.0 nm GaN is observed. The bonding interface is the Si-Si interface, marked by two triangles in Figure 3(a-b). The thicknesses of the Si interlayer and the amorphous diamond induced by the surface activation with an Ar ion beam are ~10 nm and ~3.0 nm, respectively. Figure 3(c-d) shows the BF and HAADF HR-STEM images of the GaN-diamond interface of Samp2. Only a ~4-nm-thick overall amorphous layer is observed. The bonded interface is not sharp and the thickness of amorphous diamond is approximately 2 nm. The 2-nm-thick amorphous layer between GaN and diamond is supposed to be amorphous silicon deposited by the Si-containing Ar ion beam during surface activation. TEM images of the interfaces with a low magnification are show in Figure S2, showing that both interfaces are very uniform. Figure 4. STEM images of the GaN-diamond interface of Samp1 (a) and Samp2 (b), followed by their high-resolution EELS mappings: Ga map in purple, N map in red, Si map in pink, C map in blue, O map in Cyan, Fe map in green, and Ar map in white. Additionally, high-resolution EELS analysis is used to study the chemical composition at the interfaces. As shown in Figure 4(a), the EELS element mapping of the interface confirms the interlayer thicknesses in the TEM images. Si atoms are implanted into diamond and GaN, and Ar atoms are also implanted into diamond. Some ion elements also show up at the interfaces, originating from the ion beam source which is made of stainless steel. The three-layer distribution of Fe corresponds to the activated GaN surface, the activated diamond surface, and the bonding interface. The bonding interface has the highest atomic composition of Fe (~6%). Please note these Fe contamination could be removed after further improvement of the bonding environment. The O maps in Figure 4(a) is oxygen-contaminated after sample preparation, which is supposed to be no oxygen at the interface. Different from the interface bonded by modified SAB with sputtering- deposited Si nanolayer, Si-containing Ar ion beam does not cause much implantation of Si at the interface. As shown in Figure 4(b), the EELS element mapping of the interface of Samp2 indicates that the interface layer is composed of Si, Ar, O, and Fe. No implantation of Si is observed near the interface. The disorder at the interface confirms our discussion above about the weak temperature dependence of GaN-diamond TBC. Due to the complicated nature of interfacial thermal transport, it is still unclear that how these imperfections such as Ar, O, and Fe defects, and amorphous Si, affect the TBC of the bonded interfaces. Further processing refinements are necessary to change or control the distribution of defects, impurities, and the amorphous layers to further improve TBC and elucidate their effects. 10 fingers and Power Density = 10 W/mm (b) (a) GaN Substrate (c) (d) Figure 5. (a) A 800-nm GaN device with 10 fingers seated on a substrate is modeled with a power density of 10 W/mm.41 (b) the max temperature of the device is calculated with different GaN- substrate TBC and substrates. (c) the effect of gate-gate spacing (10-50 μm) on the max temperature with different GaN-diamond TBC. (d) the effect of gate width (50-1000 μm) on the max temperature with different GaN-diamond TBC. To estimate the potential effect of room-temperature bonded GaN-Diamond interfaces on the thermal response of GaN HEMTs, a 800-nm GaN device with 10 fingers seated on different substrates is modeled under fully open channel condition with a power density of 10 W/mm.41 The thermal conductivity of 800 nm layer GaN is 150 W/m-K.26,35 Figure 5(b) shows the max temperature of the device with a heating source width of 4 μm, a heating source length of 500 μm, and gate-gate spacing of 50 μm. The thermal conductivity of SiC and diamond used in the modeling are 380 W/m-K and 2000 W/m-K, respectively.32,42 The max temperature of GaN devices on a diamond substrate is much lower than that on a SiC substrate, indicating the advantage of using diamond substrates. Max temperature decreases sharply with increasing GaN-substrate TBC when GaN-substrate TBC is small but saturates for large GaN-substrate TBC. Figure 5(c) shows the effect of gate-gate spacing on the max temperature with different GaN-diamond TBC. The power density, heating source width, gate width of the device are 10 W/mm, 4 μm, and 500 μm, respectively. The gate-gate spacing ranges from 10 μm to 50 μm. The max temperature of devices increases with decreasing gate-gate spacing. When the GaN-diamond TBC is small (<50 MW/m2-K), the max temperature increases more sharply with decreasing GaN-diamond TBC for small gate-gate spacing. Increasing GaN-diamond TBC is very important for devices with reduced gate-gate spacing. Figure 5(d) shows the effect of gate width on the max temperature with different GaN-diamond TBC. The power density, heating source width, gate-gate spacing of the device are 10 W/mm, 4 μm, and 20 μm, respectively. The gate width ranges from 50 μm to 1000 μm. The max temperature increases with gate width while keeping power density constant. The GaN- substrate TBC shows similar trend with Figure 5(b). For all cases modelled in Figure 5(b-d), GaN- diamond TBC is the key to minimize the max temperature especially for small GaN-diamond TBC values. Additionally, the measured TBC of the GaN-diamond interfaces in this work are about 50 and 90 MW/m2-K. The max temperatures of GaN-on-diamond devices with these TBC values are shown as solid gray dots in Figure 5(b). The max temperature does not decrease significantly with TBC if the TBC is larger than 50 MW/m2-K. The max temperature of GaN devices bonded on diamond substrates is much lower than that on SiC substrates, showing the great potential of cooling GaN devices bonded with single crystal diamond substrates. To compare the cooling performance of GaN devices with state-of-the-art GaN-diamond/SiC/Si TBC, we model the max temperature of 800 nm GaN devices with heating source width of 4 μm, gate-gate spacing of 20 μm, and gate width of 500 μm. Even if we use the highest GaN-SiC TBC (230 MW/m2-K) and GaN-Si TBC (143 MW/m2-K) in the literature,26,43 the power density of GaN-on-diamond can reach 20.3 W/mm, ~2.5 times as that of GaN-on-SiC (8.2 W/mm), and ~5.4 times as that of GaN- on-Si (3.78 W/mm) with a max temperature of 250 oC of the devices. Table 1. Summary of GaN-diamond TBC in the literature and this work. Method Conditions Interlayer Method TBC Ref.44 Ref.44 CVD growth of diamond on GaN CVD growth of diamond on GaN Ref.45 CVD growth of diamond on >600 oC Ref.46 Ref.47 GaN CVD growth of diamond on GaN CVD growth of diamond on GaN (MW/m2-K) Raman Raman ~37 ~28 Raman 56 TDTR ~35 TTR 83 ~25 nm dielectric ~50 nm dielectric ~50 nm dielectric ~30 nm SiNx 28 nm SiNx Ref.22 High temperature bonding >700 oC Adhesion TDTR 21-28 Ref.48 High temperature bonding >700 oC Ref.6 CVD growth of diamond on GaN Ref.40 CVD growth of diamond on This work SAB bonding GaN This work SAB bonding Room Temp. Room Temp. layer 22 nm SiNx ~ 5 nm SiNx ~ 5 nm SiNx TDTR ~58 TDTR ~100 TTR ~150 ~ 10 nm Si TDTR ~ 2 nm Si TDTR 53 92 Table 1 summarizes the GaN-diamond TBC measured in the literature and this work. "TTR" is transient thermoreflectance. Our measured GaN-diamond TBC for Samp2 is among the high TBC values reported ever. An advantage of the bonding technique in this work is the room-temperature processing temperature so no additional stress remains due to different coefficient of thermal expansion after bonding.23 It is notable that the diamond used in this work is single crystal diamond with ultra-high thermal conductivity. However, for CVD growth of diamond on GaN with a dielectric layer, the nanoscrystalline diamond near the GaN-diamond interfaces has significantly reduced thermal conductivity (tens of W/m-K).9-11,49 Moreover, the thermal conductivity of nanoscrystalline CVD diamond is anisotropic and nonhomogeneous which offsets the high thermal conductivity of diamond for cooling GaN devices. The single crystal diamond substrates used in this work do not have these disadvantages and will pave the way for thermal management of GaN- on-diamond devices. Conclusions This work reported the heterogeneously integration of GaN with single crystalline diamond substrates with two modified room-temperature surface-activated bonding techniques for thermal management of GaN-on-diamond applications. The measured TBC of the bonded GaN-diamond interfaces is among the high values reported in the literatures and is affected by the thickness of the interlayer. Due to the disorder and defects at the interfaces, a weak temperature dependence of GaN-diamond TBC was observed. HR-STEM and EELS results show the presence of interfacial amorphous layers and their compositions at the bonded interfaces. The thermal conductivity of two single crystal diamond substrates and the GaN film were also measured and matches reasonably with literature values. Device modeling shows a relatively large GaN-diamond TBC value (>50 MW/m2-K) achieved by surface activated bonding for GaN-on-diamond devices could enable to take full advantage of the high thermal conductivity of single crystalline diamond. This work paves the way for room-temperature heterogeneous integration of GaN with diamond and will impact applications such as electronics cooling especially for GaN-on-diamond devices. Experimental Sections Sample Preparation Two single crystalline diamond (CVD diamond and HPHT diamond with size of 10 mm×10 mm and 3 mm×3 mm) are bonded to templated GaN films at room temperature. The GaN films are Ga- face ~2-μm GaN layer grown on sapphire substrates (~430 μm thick). The root-mean-square (RMS) surface roughness of the GaN and diamond surface is ~0.4 nm and ~0.3 nm, respectively. A modified SAB method with a sputtering-deposited Si nano-layer is used to bond the CVD diamond (Samp1) and a modified SAB method with Si-containing Ar ion beam is used to bond the HPHT diamond (Samp2). The detailed bonding process are similar to literature25,50. After bonding, the sapphire substrate was removed by a laser lift-off process. The GaN layer was polished to be thinner to obtain good TDTR sensitivity of the buried GaN-diamond interface. The GaN layer of Samp1 is thinned to ~700 nm, while the GaN layer of Samp2 is thinned to ~1.8 μm. After that, a ~70 nm Al layer was deposited on the samples by sputtering as TDTR transducer. TDTR Measurements The measured thermal conductivity of the diamond substrates are used as input in the data fitting of TDTR measurements on GaN-diamond interfaces. A 10X objective (pump radius: 9.7 μm; probe radius: 5.8 μm) is used with a modulation frequency of 2.2 MHz (Samp2) or 3.6 MHz (Samp1). The GaN layer of Samp2 is thicker than that of Samp1 so we need large thermal penetration depth to penetrate through the GaN layer to obtain large TDTR sensitivity of the GaN- diamond TBC. Therefore, lower modulation frequency (2.2 MHz) is used for Samp2 to get larger thermal penetration depth. The heat capacity of GaN is from literature.32 Because of the thin thickness of the GaN layer of Samp1, the GaN thermal conductivity is not sensitive so we fixed the GaN thermal conductivity in the data fitting using literature values of similar GaN layer bonded on SiC.26 The thermal conductivity and heat capacity of the Al transducer is from literature with similar Al films.26 The picosecond acoustic technique is used to measure the local Al and GaN thicknesses.26,51 Materials Characterization Cross-section TEM samples were prepared with a FEI Helios dual beam focused ion beam (FIB) system. The interface structures were characterized by a HR-STEM (Probe-corrected FEI Titan) and the interface composition was measured by EELS (Gatan Enfinium) with a step size of 0.2 nm. The observation in this study is along <11-20> axis of GaN. Supplementary Materials The supplementary materials include the TEM images of the bonded interface with low magnification. Acknowledgements Z. C., L. Y., and S. G. would like to acknowledge the financial support from U.S. ONR MURI Grant No. N00014-18-1-2429. F.M., and T.S. would like to acknowledge the financial support from R&D for Expansion of Radio Wave Resources, organized by the Ministry of Internal Affairs and Communications, Japan and JSPS KAKENHI Grant Number 19K15298. Completing Financial Interest The authors claim no completing financial interest. References 1 Amano, H. et al. The 2018 GaN power electronics roadmap. Journal of Physics D: Applied Physics 51, 163001 (2018). 2 Yates, L. et al. in Compound Semiconductor Integrated Circuit Symposium (CSICS), 2016 IEEE. 1-4 (IEEE). 3 Won, Y., Cho, J., Agonafer, D., Asheghi, M. & Goodson, K. E. Fundamental cooling limits for high power density gallium nitride electronics. IEEE Transactions on Components, Packaging and Manufacturing Technology 5, 737-744 (2015). 4 Liu, T. et al. 3-inch GaN-on-diamond HEMTs with device-first transfer technology. IEEE Electron Device Letters 38, 1417-1420 (2017). 5 Ohki, T. et al. An Over 20-W/mm S-Band InAlGaN/GaN HEMT With SiC/Diamond-Bonded Heat Spreader. IEEE Electron Device Letters 40, 287-290 (2018). 6 Yates, L. et al. Low Thermal Boundary Resistance Interfaces for GaN-on-Diamond Devices. ACS applied materials & interfaces 10, 24302-24309 (2018). 7 Tadjer, M. J. et al. GaN-On-Diamond HEMT Technology With T AVG= 176° C at P DC, max= 56 W/mm Measured by Transient Thermoreflectance Imaging. IEEE Electron Device Letters 40, 881-884 (2019). 8 Sun, H. et al. Temperature-dependent thermal resistance of GaN-on-diamond HEMT wafers. IEEE Electron Device Letters 37, 621-624 (2016). 9 Anaya, J. et al. Simultaneous determination of the lattice thermal conductivity and grain/grain thermal resistance in polycrystalline diamond. Acta Materialia (2017). 10 Cheng, Z. et al. Tunable Thermal Energy Transport across Diamond Membranes and Diamond-Si Interfaces by Nanoscale Graphoepitaxy. ACS applied materials & interfaces (2019). 11 Cheng, Z. et al. Probing Growth-Induced Anisotropic Thermal Transport in High-Quality CVD Diamond Membranes by Multi-frequency and Multi-spot-size Time-Domain Thermoreflectance. ACS applied materials & interfaces (2018). 12 Cheaito, R. et al. in Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm), 2017 16th IEEE Intersociety Conference on. 706-710 (IEEE). 13 Moelle, C. et al. Measurement and calculation of the thermal expansion coefficient of diamond. Diamond and Related Materials 6, 839-842 (1997). 14 Reeber, R. R. & Wang, K. Lattice parameters and thermal expansion of GaN. Journal of Materials Research 15, 40-44 (2000). 15 Hancock, B. L. et al. Ultraviolet micro-Raman spectroscopy stress mapping of a 75-mm GaN- on-diamond wafer. Applied Physics Letters 108, 211901 (2016). 16 Siddique, A. et al. Structure and Interface Analysis of Diamond on an AlGaN/GaN HEMT Utilizing an in Situ SiN x Interlayer Grown by MOCVD. ACS Applied Electronic Materials (2019). 17 Hines, N. J. Advanced Solutions for Thermal Management and Reliability of Gallium Nitride- Based High Electron Mobility Transistors on Diamond Substrates, Georgia Institute of Technology, (2019). 18 Tripathy, S., Chua, S., Chen, P. & Miao, Z. Micro-Raman investigation of strain in GaN and Al x Ga 1− x N/GaN heterostructures grown on Si (111). Journal of Applied Physics 92, 3503- 3510 (2002). 19 Ahmad, I., Holtz, M., Faleev, N. & Temkin, H. Dependence of the stress -- temperature coefficient on dislocation density in epitaxial GaN grown on α-Al 2 O 3 and 6H -- SiC substrates. Journal of applied physics 95, 1692-1697 (2004). 20 Choi, S., Heller, E., Dorsey, D., Vetury, R. & Graham, S. The impact of mechanical stress on the degradation of AlGaN/GaN high electron mobility transistors. Journal of Applied Physics 114, 164501 (2013). 21 Kang, B. et al. Effect of external strain on the conductivity of AlGaN/GaN high-electron- mobility transistors. Applied physics letters 83, 4845-4847 (2003). 22 Cho, J. et al. Improved thermal interfaces of GaN -- diamond composite substrates for HEMT applications. IEEE Transactions on Components, Packaging and Manufacturing Technology 3, 79-85 (2012). 23 Francis, D. et al. Formation and characterization of 4-inch GaN-on-diamond substrates. Diamond and Related Materials 19, 229-233 (2010). 24 Chao, P.-C. et al. Low-temperature bonded GaN-on-diamond HEMTs with 11 W/mm output power at 10 GHz. IEEE Transactions on Electron Devices 62, 3658-3664 (2015). 25 Mu, F., He, R. & Suga, T. Room temperature GaN-diamond bonding for high-power GaN-on- diamond devices. Scripta Materialia 150, 148-151 (2018). 26 Mu, F. et al. High Thermal Boundary Conductance across Bonded Heterogeneous GaN-SiC Interfaces. arXiv preprint arXiv:1905.04171 (2019). 27 Cahill, D. G. Analysis of heat flow in layered structures for time-domain thermoreflectance. Review of scientific instruments 75, 5119-5122 (2004). 28 Gaskins, J. T. et al. Thermal Boundary Conductance Across Heteroepitaxial ZnO/GaN Interfaces: Assessment of the Phonon Gas Model. Nano letters 18, 7469-7477 (2018). 29 Cheng, Z. et al. Thermal conductance across β-Ga2O3-diamond van der Waals heterogeneous interfaces. APL Materials 7, 031118 (2019). 30 Ward, A., Broido, D., Stewart, D. A. & Deinzer, G. Ab initio theory of the lattice thermal conductivity in diamond. Physical Review B 80, 125203 (2009). 31 Inyushkin, A. et al. Thermal conductivity of high purity synthetic single crystal diamonds. Physical Review B 97, 144305 (2018). 32 Zheng, Q. et al. Thermal conductivity of GaN, GaN 71, and SiC from 150 K to 850 K. Physical Review Materials 3, 014601 (2019). 33 Lindsay, L., Broido, D. & Reinecke, T. Thermal conductivity and large isotope effect in GaN from first principles. Physical review letters 109, 095901 (2012). 34 Simon, R. B., Anaya, J. & Kuball, M. Thermal conductivity of bulk GaN -- Effects of oxygen, magnesium doping, and strain field compensation. Applied Physics Letters 105, 202105 (2014). 35 Ziade, E. et al. Thickness dependent thermal conductivity of gallium nitride. Applied Physics Letters 110, 031903 (2017). 36 Kazan, M. Interpolation between the acoustic mismatch model and the diffuse mismatch model for the interface thermal conductance: Application to inn/gan superlattice. Journal of heat transfer 133, 112401 (2011). 37 Roder, C., Einfeldt, S., Figge, S. & Hommel, D. Temperature dependence of the thermal expansion of GaN. Physical Review B 72, 085218 (2005). 38 Lee, E. & Luo, T. The role of optical phonons in intermediate layer-mediated thermal transport across solid interfaces. Physical Chemistry Chemical Physics 19, 18407-18415 (2017). 39 Sun, F. et al. Molecular Bridge Enables Anomalous Enhancement in Thermal Transport across Hard‐Soft Material Interfaces. Advanced Materials 26, 6093-6099 (2014). 40 Zhou, Y. et al. Barrier-Layer Optimization for Enhanced GaN-on-Diamond Device Cooling. ACS Applied Materials & Interfaces 9, 34416-34422 (2017). 41 Bagnall, K. R., Muzychka, Y. S. & Wang, E. N. Analytical solution for temperature rise in complex multilayer structures with discrete heat sources. IEEE Transactions on Components, Packaging and Manufacturing Technology 4, 817-830 (2014). 42 Pearton, S. et al. A review of Ga2O3 materials, processing, and devices. Applied Physics Reviews 5, 011301 (2018). 43 Cho, J., Li, Z., Asheghi, M. & Goodson, K. E. Near-junction thermal management: Thermal conduction in gallium nitride composite substrates. Annual Review of Heat Transfer 18 (2015). 44 Pomeroy, J. W., Bernardoni, M., Dumka, D., Fanning, D. & Kuball, M. Low thermal resistance GaN-on-diamond transistors characterized by three-dimensional Raman thermography mapping. Applied Physics Letters 104, 083513 (2014). 45 Dumka, D. et al. in 2013 IEEE Compound Semiconductor Integrated Circuit Symposium (CSICS). 1-4 (IEEE). 46 Cho, J., Won, Y., Francis, D., Asheghi, M. & Goodson, K. E. in 2014 IEEE Compound Semiconductor Integrated Circuit Symposium (CSICS). 1-4 (IEEE). 47 Sun, H. et al. Reducing GaN-on-diamond interfacial thermal resistance for high power transistor applications. Applied Physics Letters 106, 111906 (2015). 48 Cho, J., Francis, D., Altman, D. H., Asheghi, M. & Goodson, K. E. Phonon conduction in GaN- diamond composite substrates. Journal of Applied Physics 121, 055105 (2017). 49 Rougher, T. L. et al. in Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm), 2017 16th IEEE Intersociety Conference on. 30-38 (IEEE). 50 Mu, F. et al. A comparison study: Direct wafer bonding of SiC -- SiC by standard surface- activated bonding and modified surface-activated bonding with Si-containing Ar ion beam. Applied Physics Express 9, 081302 (2016). 51 Hohensee, G. T., Hsieh, W.-P., Losego, M. D. & Cahill, D. G. Interpreting picosecond acoustics in the case of low interface stiffness. Review of Scientific Instruments 83, 114902 (2012).
1710.00046	2	1710	2017-10-10T20:27:08	Roadmap on Atto-Joule per Bit Modulators	[ "physics.app-ph", "physics.optics" ]	Electrooptic modulation performs the conversion between the electrical and optical domain with applications in data communication for optical interconnects, but also for novel optical compute algorithms such as providing nonlinearity at the output stage of optical perceptrons in neuromorphic analogue optical computing. Since the clock frequency for photonics on chip has a power overhead sweet slot around 10s of GHz, ultrafast modulation may only be required in long distance communication, but not for short onchip links. Here we show a roadmap towards atto Joule per bit efficient modulators on chip as well as some experimental demonstrations of novel plasmon modulators with sub 1fJ per bit efficiencies. We then discuss the first experimental demonstration of a photon plasmon-hybrid Graphene-based electroabsorption modulator on silicon. The device shows a sub 1V steep switching enabled by near ideal electrostatics delivering a high 0.05dB per V um performance requiring only 110 aJ per bit. Improving on this design, we discuss a plasmonic slot based Graphene modulator design, where the polarization of the plasmonic mode matches with Graphenes inplane dimension. Here a push pull dual gating scheme enables 2dB per V um efficient modulation allowing the device to be just 770 nm short for 3dB small signal modulation. This in turn allows for a device-enabled two orders of magnitude improvement of electrical optical co integrated network on chips over electronic only architectures. The latter opens technological opportunities in in cognitive computing, dynamic data-driven applications system, and optical analogue compute engines to include neuromorphic photonic computing.	physics.app-ph	physics	Roadmap on Atto-Joule per Bit Modulators Volker J. Sorger1,, Rubab Amin1, Jacob B. Khurgin2 Zhizhen Ma1, Hamed Dalir3, Sikandar Khan1 1Department of Electrical and Computer Engineering, George Washington University 800 22nd St., Science & Engineering Hall, Washington, DC 20052, USA 2Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, Maryland 21218, USA 3Omega Optics, Inc., 8500 Shoal Creek Blvd., Bldg. 4, Suite 200, Austin, TX 78757, USA corresponding author, E-mail: [email protected] Abstract Electro-optic modulation performs the conversion between the electrical and optical domain with applications in data communication for optical interconnects, but also for novel optical compute algorithms such as providing non-linearity at the output stage of optical perceptrons in neuromorphic analogue optical computing. While resembling an optical transistor, the weak light-matter-interaction makes modulators 105 times larger compared to their electronic counterparts. Since the clock frequency for photonics on-chip has a power-overhead sweet-slot around 10's of GHz, ultrafast modulation may only be required in long-distance communication, but not for short on-chip links. Hence the search is open for power-efficient on-chip modulators beyond the solutions offered by foundries to date. Here we show a roadmap towards atto-Joule per bit efficient modulators on-chip as well as some experimental demonstrations of novel plasmon modulators with sub-1fJ/bit efficiencies. Our parametric study of placing different actively modulated materials into plasmonic vs. photonic optical modes shows that 2D materials overcompensate their miniscule modal overlap by their unity-high index change. Furthermore, we reveal that the metal used in plasmonic-based modulators not only serves as an electrical contact, but also enables low electrical series resistances leading to near-ideal capacitors. We then discuss the first experimental demonstration of a photon-plasmon-hybrid Graphene-based electro- absorption modulator on silicon. The device shows a sub-1V steep switching enabled by near-ideal electrostatics delivering a high 0.05dB/V-µm performance requiring only 110 aJ/bit. Improving on this design, we discuss a plasmonic slot-based Graphene modulator design, where the polarization of the plasmonic mode matches with Graphene's in-plane dimension. Here a push-pull dual-gating scheme enables 2dB/V-µm efficient modulation allowing the device to be just 770 nm short for 3dB small signal modulation. Lastly, comparing the switching energy of transistors to modulators shows that modulators based on emerging material-based, plasmonic-Silicon hybrid integration perform on-par relative to their electronic counter parts. This in turn allows for a device-enabled two orders-of- magnitude improvement of electrical-optical co-integrated network-on-chips over electronic-only architectures. The latter opens technological opportunities in cognitive computing, dynamic data-driven applications system, and optical analogue compute engines to include neuromorphic photonic computing. Keywords: electrooptic modulation, plasmonics, integrated optics, Graphene, energy efficiency, DDDAS, optical computing. Main Body Electro-optic modulation is a key function in modern data communication as it performs the conversion between the electronic data originating from compute cores, to the optical domain of low-loss data routing. While this function is universally used around the globe in long-haul, metro, and short-haul communications [1], such as data centers [2,3], the case for on-chip optical interconnects was made [4] mainly to address the widening discrepancy between the data handling capability of electronic cores vs. delays and power overheads in the communication- handling network-on-chip [5-8]. The parallels between an electro-optic modulator (EOM) and a field-effect-transistor (FET) however are noticeable; both control a channel via an electrostatic gate. The discrepancy of the physical device lengths are, however, significant, as state-of-the-art silicon photonics modulators are of millimeter dimensions [9], while FETs are just 10 nanometer. The reason for this is known, and lies in the weak light-matter-interaction, and inefficient material ability to change its optical refractive index upon applying a gate bias [11-14]. Given the resemblance of an EOM to FETs, we interchangeably use the words 'switching ' and 'modulation', while accepting the slight discrepancy – switching technically refers to a strict 2-level system, while modulation is an analogue function. Of course, in reality both EOMs and FETs have analogue transfer functions, hence justifying the wording definition in this work. This manuscript focuses on charge-driven electro-absorption modulators only, as suppose to electric field-driven designs, such as those based on Franz-Keldysh, or Pockel's effect [16-18]. We also mainly discuss three actively materials only, namely Silicon, Indium-Tin-Oxide (ITO) and Graphene, but briefly mention some results regarding expected absorption of other 2D materials [19-21], quantum dots [22], and quantum wells [23,24]. The discussion thence includes a parametric study of achievable optical effective index changes a function of the optical mode overlap factor with the active material, the materials own index change potential, and the effective mode's group index change with bias. We then show a experimental results of a hybrid- photon-plasmon modulator based on Graphene on Silicon photonics, and a mode-overlap improved dual-gated Graphene plasmon-slot waveguide modulator design allowing for sub 1 micrometer shot device lengths, allowing for a ~4dB/µm strong a signal modulation. We close by comparing our plasmonic silicon EOMs with FETs, which shows an energy-convergence of these two classes of devices. Optoelectronic devices are significantly more bulky compared to electronic counterparts. For instance electrooptic modulators based on Silicon's plasma dispersion in photonic integrated waveguide modes are several millimeter long in order to obtain the desired phase shift, leading to amplitude modulation in Mach-Zehnder-Interferometers (MZI) [9]. Three-terminal switching devices consist of a 'source', a 'drain' and some form of a control 'gate'. In field effect transistors (FET) all, the channel and the gate are electrically controlled. The reason why this is possible over only a few to ten's of nanometers is based on the fact that both the channel (electron current) and the gate (electron- loading a capacitive gate) have the same length-scale, namely their spatial extend are bound by their fermionic wave function which is on the order of nanometers. Photons being boson, on the other hand, do not interact with one- another. The only option for them to interact is via matter. This means, that the opto-electronic response of a device is fundamentally governed by the ability and efficiency of the photon (or plasmon) to interact with the electronic wave functions of matter. The fundamental inefficiency of optoelectronic effects, however, is that the spatial length scales of the electronic- versus the photonic wavefunctions are 3-orders of magnitude apart, considering wavelengths in the visible or near IR range. This work focuses on telecomm wavelengths only, namely 1550 nm or 0.8 eV. This large mismatch is the physical cause of the weak interaction between light with matter, and has technologically led to bulky opto-electronics with low chip integration densities [25]. Thus, in order to shrink down (i.e. 'scale') device lengths and footprints, one has two general two options to increase the weak light-matter-interactions (LMI, Fig. 1); a) one can increase the photon lifetime using resonators, aiming to increase their cavity quality-Q factor, or, b) one can enhance the electric field density non-resonantly by shrinking the optical mode possibly beyond the diffraction limit by deploying polaritonic modes. This is possible for instance by exploiting discontinuities in the permittivity leading to plasmonics, metal optics, or slot-waveguides [26-24]. High-Q cavities however introduce several technological disadvantages, but their spectral sensitivity allows reducing the applied voltage needed to shift the effective mode's index in and out of resonance of the cavity, thus lowering the required drive voltage [35,36]. However, the spectral sensitivity gains, are overshadowed by spectral tuning overheads in form of thermal tuning heaters, thus increasing the devices' dynamic the operation wavelength of the device, nm the cavity's modal refractive index, and c the speed of light in vacuum) in cavities leads to slow electro-optic modulation timescales. For example, a ring resonator with a Q-factor of 40,000 results in modulation frequencies of only 12 GHz at telecom wavelengths, which is already below industry standard. However, it is interesting to ask what net-benefits resonant photonic structures do enable. For instance, the footprint of modulator can be lowered by folding the linear length of a MZI-based modulator into a cavity. Here, the footprint of a cavity-based EOM can be lowered by a factor proportional to the resonators finesse compared [19, 20]. In brief, the lengths of the MZI and ring quality factor, respectively. Thus, the ratio of the lengths of the two EOMs is proportional to the ring EOM, which is somewhat surprising, however can be seen by the following arguments; the resonator given by !!"#$%!!"=!∙! 2!!, and !!"#$=2!", where R and ! are the ring's radius and finesse ℱ of the resonator. Secondly, the sensitivity of the MZI is actually equal to that of the Δ!≈Δ!!"#$= !!!!!! with!!!!! ≈!!!! requires ∆!=!!! , where Δ!!"#$ is linewidth, !! the group index, Δ! the modulated index (i.e. a function of applied voltage), !!! Δ!=!/2!, extending the MZI length to be !=!∙! 2!!, which equals the same sensitivity as !!!"#$%≈ given by; !!=!!!! and hence the transit-time limited bandwidth is !!!" !"=0.44∙ power consumption. Moreover, the long photon lifetimes (!!!!"!#=!!!!!/2!", where λ0 is the photon lifetime in the resonator. Since a single arm MZI requires an index change of the ring design. Thirdly, the question is whether a long linear device (i.e. larger capacitance) or a more compact due to a cavity (but potentially photon lifetime limited) design has a higher modulation speed. The answer is that they are in fact equal, provided that the MZI is not parasitic capacitance limited as shown next. The MZI transit time (treat the MZI as a lumped-element) is !!!!"#$%= !!!!!=!!. The cavity's photon-limited bandwidth due to the long photon lifetime can be estimated via !!!" !"#$≈ !!!!!!=!!; closing the above argument that the MZI and ring (or tuning, Δ! corresponds to the change in the effective refractive index (real part), Δ!!"" from the loss, Δ!, as a direct result from the Kramers–Kronig relations. Thus, the design challenge is to figure of merit is FOMEOM-cavity = Δ!/Δ!, i.e. the change in the loss, Δ! is a function of the modal effective extinction coefficient change, Δ!!"". A timely example of a deployment of this FOM is Fabry Perot) cavity EOM have about the same modulation speed. For example a modulator with a Q of 10,000 has a cut-off speed of about 34GHz (for an electro-optic coefficient of 300 pmV-1, and extinction ratio (ER) = 3dB). We recently investigated the effect of a cavity on both the shift in resonance wavelength, i.e. modal tuning of the underlying waveguide mode [19, 20]. However, any tuning also increases optimize the ratio of obtainable tuning which improves the modulators extinction ratio, ER, (i.e. modulation depth) relative to these incurred losses; hence an appropriate cavity-based EOM the target EOM values of the AIM Photonics consortium's device roadmap [37]. With respect to cavity-enhanced FOM performance, one can show that if optical losses are minimal, that longer devices perform higher [19, 20], as known from millimeter-long foundry designs [9]. However, their sizable footprints lead to not insignificant insertion losses, IL, and obtaining high-speed circuit designs are challenging given that travelling-wave designs must be adhered to. While, sub-volt driving voltages switching at 10's of GHz have been demonstrated [24], the device capacitance of these sizable modulators, limits their potential for sub fJ/bit efficient devices required for next generation modulators [25]. That is why in this work, we focus on LMI enhancements using field enhancements in sub-diffraction limited waveguides without resonance enhancement. The latter also enables spectrally broadband devices allowing these modulators to be used in wavelength-division-multiplexing (WDM) photonic circuitry (Fig. 1). Interestingly, the ratio of both, namely a high Q-factor and a small mode-volume, Vmode, are proportional to the Purcell factor – a merit that can be regarded as an optical concentration factor [38]. Next, we discuss how this factor allows predicting nanophotonic scaling laws. This hypothesis of "smaller-is-better" has motivated optical engineers to build various nanophotonic devices in an ad-hoc manner thus far. That is, an understanding leading to fundamental scaling behavior for this new class of devices is yet outstanding. Here we analyze scaling laws of EOMs with a focus interest in the micrometer to sub-micrometer scale. With technology options such as plasmonics [11-14], nanoscale dielectric resonators [39, 40] and slot-waveguides [41], we are able to surpass the diffraction limit of light by engineering the effective refractive index. Still, decreasing the optical mode volume, Vm, introduces adverse effects, such as bending- and ohmic losses for polaritonic modes. It is therefore not straightforward to predict modulator performance scaled into sub-micron size [42], leading to a rigorous analysis of fundamental scaling laws for nanophotonics as a function of critical device length [38]. In this scaling analysis, we assume three types of optical cavities cavities, a) a traveling-wave ring resonator (RR) [43], b) a metal-mirror based Fabry-Perot (FP) cavity [39], and c) a plasmonic metal nano-particle (MNP) [44] (Fig. 2b), that enhance the fundamentally weak interaction between light and matter via the ratio of !/Vm, where ! is the cavity quality mode volume !! for the RR and FP cavities are given in ref [38]. The resulting Purcell factor, defined as18!!!= !!!! !!! ! !!! , where !! is the resonant wavelength of the cavity, and ! is the factor, Vm is the effective volume of electromagnetic energy of a resonant mode. An interesting, although expected result is, that all cavity types do not perform equally well for vanishing critical dimension due to their respective non-monotonic Purcell factor scaling. The critical length for these cavities are the radius for the RR and the MNP, and the physical distance between two mirrors for the FP. Analytical expressions for both the cavity quality factor Q and the optical cavity material refractive index shows a significant influence on the EOMs power consumption, E/bit ~ (FpQ)-1 [38]. This can be understood from the dimensional schematic of an EOM (Fig. 2a); the required electric field, E-Field = Vbias/h, to obtain a desired bit-error-rate (BER) at the detector downstream and the device capacitance, C = εrε0A/d determine the energy efficiency, E/bit = ½ CV2 of an EOM (Fig. 2a). Once substituted, we find that E/bit is proportional to the physical volumetric dimensions of the modulator divided by the cavity Q2, or E/bit ~ (FpQ)-1. Q for the three selected cavities overall decreases with scaling as expected since light is either less confined (i.e. bending losses), or polaritonically (i.e. matter-like) bound to the metal's electron see, thus increasing the loss. The mode volume on the other hand, scales linearly for the one- dimensional scaling for RRs but cubically for the MNP plasmon cavity. The FP cavity exhibits an inverse scaling, due to the mode character changing from travelling wave to an MIM plasmon mode one both mirrors start to couple. The scaling for Q and Vm, thus show optima for the Purcell factor, which quantify the point of maximum feedback (highest Q, low losses), while offering high optical confinement. Beyond Fp's maxima, the parasitic losses of the cavity become dominant. Then, the lowest E/bit match the Fp maxima well, where discrepancies originate from the strong ~Q-2 dependency. The results show that devices (optimized for energy consumption only) are high-Q EOMs, which are, however spectrally sensitive, 10's of micrometer in footprint, slow in response time, and require thermal tuning, which was not taking into account in this analysis here. Based on these limitations, plasmonic modulators offer an interesting alternative as they can approach 100's of aJ/bit efficiencies while allowing for sub- micron short device lengths, which supports small electrical capacitances enabling fast switching, provided the active material allows for such. A modulator's key performance is the ability to change the waveguide mode's effective index most efficiently, i.e. with the lowest voltage bias. Thus, the aim is interestingly similar to that of FETs, where the steepness of the I-V transfer function is quantified by the sub-threshold swing, i.e. the amount of voltage change required to induce a 10-fold current change. For EOMs this translates into an effective modal index change for applied voltage bias. Similar to optical gain building devices like lasers [26,28,30,31,46-48], EOMs require optimization with respect to the modal overlap, Γ, with the active material [45, 49,50]. A high extinction ratio (ER) does critically depend on the obtainable index change upon biasing the device. For instance for electrooptic (phase) modulation the effective change of the k–vector of the light is given by !"=!!!!"!"=!"!!!!"!"=!"!!!!!"!"=!"!!!!! , where !! is the group Δϕ=!"!Δ!!""!=ϵ!Γ!!"!!!!!!!, where ! is the optical confinement factor and !!= !!!!""!!" waveguide mode. becomes . !! index in the change then The phase For discrete modulation states this relationship can be expressed as the ratio of the active The index change inside the modulator (Δ!!"") is then given by Δ!!""=ϵ!Γ!!"!!! [45, 49,50]. material index relative to its initial condition (Δ!!"#/!!"#) multiplied by its modal confinement factor (!), relative modal permittivity enhancement (!!) and effective group index (!!), i.e. [45]!"∝∆!!""=!!Γ∆!!"# !!"#∆!! , where the group index !! corresponds to dispersive propagation in the longitudinal direction given by !!=!!""−!!!!""!" , which applies to isotropic index materials such as Silicon and ITO based structures. Due to the unique electro- optic nature of graphene and anisotropy of the indices (tensor), this simple equation insufficiently describes modulation performances in the graphene-based modulators. Note, that Graphene's propagating energy index and group index need to be represented by directional tensor terms and solved for each component, which is however beyond the scope of this work. Here, we follow a similar approach for the Graphene based modes to the bulk cases in order to associate modulation effects relating to the modal illumination pattern and effective index change. Next, we discuss the various optical modes considered here, then focus on the confinement factor first, and finally discuss obtainable effective index changes governing modulation performance [49,50]. We study three different mode structures for each of the three active materials introduced above (Fig. 3), while our aim is to explore modulator-suitable material/mode combinations for electro- absorption modulation mechanisms. The target is to increase the LMIs towards ultra-compact modulators while preserving ER, and we consider plasmonics as a spatial mode compression tool towards increasing the LMI and compare two distinct plasmonic modes with a bulk-case for comparison. The two plasmonic modes analyzed are the slot waveguide in a metal-insulator- metal (MIM) configuration, and a hybrid plasmonic polariton (HPP) design in a metal-insulator- semiconductor (MIS) configuration [26-34]. In order to understand the LMI enhancement effect from modal compression, we compare each active material with a bulk case where the waveguide consists of the active material only. The resulting design space is a 3×3 matrix, where we capture both the modal overlap factor and the group index as a function of carrier concentration for Silicon and ITO, whereas for Graphene as a function of chemical potential (Fig. 4 and 5). As expected the overlap factor for a bulk silicon-based modulator is rather high (~80%) but does not change significantly with carrier concentration. Hybridizing Silicon as active material with plasmonics worsens the overlap since field is partly confined inside the plasmonic slot and not in the actively tuned Silicon layer underneath (Fig. 4a). Photonic hybridization provides performance between these two extreme cases. Changing from Silicon to ITO shows a bias (i.e. carrier-sensitive) overlap factor approaching 70% near ITOs epsilon-near-zero (ENZ) point close to 7x10-20 cm-3 [51,52]. Such change helps in EAMs where absorption in the lossy OFF state (high carrier concentration) should have a high overlap factor, while the light ON state, losses should be reduced. Since the ITO's capacitive-gated index change occurs only at a thin 5-10 nm layer corresponding to a graded-index accumulation layer, the bulk modes do not provide any advantage for ITO modulators (Fig. 4a). Graphene's atomistic cross-section naturally leads to low overlap factors about 2 orders lower compared to ITO (Fig. 4b), but can reach almost 1% for a single Graphene layer, in slot-waveguide structures (Fig. 5c) [49]. Generally for all modulators, the effective group index change with voltage swing, Δng, should be maximized. For bulk modes this dispersion, or slow light-effect while present, is relatively weak (Fig. 4c). Similarly, when Silicon is used for modulation the weak electrooptic plasma tuning of Silicon does not lead to strong dispersion, which limits the obtainable group index change. ITO's Drude model and strong carrier-dependent index change, however, lead to significant (~200%) group index change in particular when used in conjunction with plasmonic modes (Fig. 4c). The latter gives precedence for synergistic use of physical effects to design high-performance EOMs. Next we discuss obtainable modulation performance and focusing on electro-absorption modulators (EAM). We emphases on obtainable absorption per unit effective thickness (i.e. a quantifier for the optical mode overlap with the active medium), voltage efficiency, and modulation strength vs. overlap factor. An integrated modulator's task is to change the power flow of electromagnetic energy inside a waveguide, which relates to the Poynting vector. Combining Maxwell equations with the Poynting vector gives n x E 2 (1) S x ( ) 2 y 2 y = = ( ) ω ε 0 2 β n x E ( ) 2 n 2 η eff 0 where the effective index is effn cβ ω= / . The total power flow is then Wt n x E dx 2 ( ) ∞ = 2 y eff W n η eff 0 2 ∫ −∞ n E 2 a a 2 η 0 P W S x dx ( ) = = ∞ ∫ −∞ (2) where an is the refractive index and aE is the transverse electric field in the active layer. The effective thickness a value proportional to the modal overlap factor of the active material with the waveguide mode is given by: ! !! teff = 1 neff na 1 Ea 2 ∞ ∫ −∞ n2(x)Ey 2 dx (1) where Ea and na is the maximum field in the waveguide cross-section and the index at that location respectively. Note the former does not necessarily need to be the center of the waveguide, thus allowing for a wide variety of designs. Here the x- and y-directions are the cross-section plane of an in z-direction propagating wave. From here one can ask what is the absorption of a given active material, and find a universal dependency of the latter on the effective thickness, teff, and the absorption cross-section, s (w). The explicit expression of s (w), however depends on the material dependent absorption mechanism. Next is a discussion of Graphene as per example; the intensity and constant like the free space impedance, h0, and the modulated intensity change as: (2) where Ngr is the carrier density of the active material, here Graphene, and fc the Fermi-Dirac function. Then the absorption coefficient inside a waveguide is given as: depends on the field dN p D ,2 dt e E 2 2 a 8 E η / 2a 2 0 e 2 η 0 4 (5) h h h ⊥ = α = I δ ω α [1 [1 [1 ] I N f c ] N f c ⊥ ⊥ − = − = = = I I gr gr ⊥ ⊥ − f c ] πα 0 n a gr N t eff gr where 0α is the fine structure constant. The absorption results as a function of carrier concentration (Fig. 5a) and explicit voltage (tox = 100 nm was used, Fig. 5b) show that various materials not only have (trivially) different absorption values for a given voltage (or carrier concentration), but more importantly that they exhibit inverse scaling trends; for instance free- carriers in Silicon or ITO increase absorption, since more carriers simple increase the loss in the Drude formula. For quantum dots (QD), two dimensional (2D) materials and quantum wells (QW) the trend is inverse since more carriers occupy states elsewise available for electron-hole pairs upon absorption. For these state-filling materials, the trend for absorption vs. voltage is similar, however appears with different magnitude and voltage scaling. The latter modulation 'steepness' is, however, relevant from an efficiency point of view similar to Landau limit of 60mV/dec in transistors; that is the less voltage for switching, the lower the dynamic power consumption of the modulator. Here quantum dots- wells, and Graphene perform particularly steep, while the highest absorption is found for WSe2, a 2D material of the class of transition metal dichalcogenides (TMD). The graphene modulation mechanism is well established, and referred to as Pauli-Blocking [53-55]; here a photon can only be absorbed when an electron transition occurs which requires an empty state in the conduction band. If the Fermi level is above the energy level given by the sum of the electron's initial state and the photon energy, all states are filled, and the photon is not absorbed, 'blocked'. The absorption steepness, thus depends on the energy-sharpness of the density of states, which at first order is ~ kBT = 10's – 100 meV, here T is the temperature, and kB the Boltzmann constant. The weak index modulation of Silicon is exemplified in Figure 5c, by requiring close to unity- high overlap factors in order to obtain a reasonable modulation strength, defined as the absorption change upon modulation [49]. However, for realistic on-chip waveguide designs, the modal overlap can never reach 100% even in bulk modes, since some amount of field is always leaking into the cladding (see forbidden region, Fig. 5c). Turning to emerging active materials such as ITO and Graphene for modulation shows that the strong index modulation of unity is able to over compensate their small overlap factor. Graphene in particular is remarkable that a 0.3 nanometer thin sheet of material with an overlap factor if 0.5x10-4 provides a stronger modulation than the best silicon mode ever can [53]. A single 2D material like Graphene is able to show an overlap factor approaching 1% when optimized for in-plane polaritons such as in slot-waveguides (blue squares, Fig. 5c). In fact, stacking multiple layers of 2D materials may be an interesting approach to increase the modulation efficiency further, enabled by overlap factors of up to 40%. This temperature dependence could be an option to reduce the modulators switching energy, which could be interesting in quantum applications where the system operates a-priori at cryogenic temperatures [56]. A discussion on this temperature dependency for Graphene-based modulators is given below (Fig. 6). Since there has been much interest in 2D materials beyond Graphene for opto-electronics due to their unique excitonic properties originating from their high anisotropy given by their contrasting dimensionality [57, 58]. The high excitons binding energy in 2D materials is due to the reduced amount of coulomb screening leading to high absorption. Therefore it was suggested that such 'robust' excitons can be used for efficient light modulation. The 2D material exciton is characterized by its 2D Bohr radius h 2 a ex = 2 πε ε eff 0 e m 2 r (6) effε which is 50% as large as the radius of 3D-material excitons. The effective dielectric constant in 2D materials approaches unity while the effective mass m is somewhat larger than in III-V semiconductors, thus given a TMD exciton Bohr radius just a few nanometer small leading to a high corresponding exciton binding energy of 0.5-0.7 eV [59]. ~ 0.25 (7) rm eV 0 2 ~ 0.5 E ex h h where we used 2 m a r ex 2 = = ' =h ' ( / ≈ = r 2 0 12 / 2 m γ c ) 4 = ω γ 2 πα 0 σ ω πα ( ) 4 ( ) / a t 2 ' αω σ ω π ex eff r 2 12 The absorption of the exciton can be obtained by using the value of exciton envelope where and operating on the excitonic wavefunction at the origin resulting in resonance . Modulating 2D P m 2 2 2 ω cv 0 materials results in (i) state filling, (ii) bandgap renormalization, and (iii) screening. As a result the exciton bleaches due to Mott transition with a screening radius comparable to the exciton radius. Therefore, exciton bleaching most likely takes place because of state filling. The exciton wavefunction can be considered a coherent superposition of the electron hole pair states with wavevectors ranging from 0 to ~(αex)-1 density of these states. Here the exciton radius and binding energy do not impact the switching charge significantly. Since the effective mass in TMDs is typically larger than in III-V semiconductors, using excitons does not change the fundamental fact that each time a single electron is injected inside the active layer a single transition is being blocked, and, given the fact that the oscillator strength for each allowed transition is about the same, the expected obtainable absorption change is roughly the similar. m ω c Given the high modulation potential of Graphene, we next consider more modulation physics relating to Graphene's Pauli blocking electro-optic modulation mechanism. Graphene is an anisotropic material given its dimensions: in its honeycomb like lattice plane, the in-plane permittivity (ε) can be tuned by varying its chemical potential !!, whereas the out-of-plane permittivity is reported to remain constant around 2.5 [19,20]. We model graphene with two different temperatures by Kubo model at T = 0K and T = 300K (Fig. 6). At higher temperature, the imaginary refractive index vs. chemical potential is smeared due to the natural temperature dependency of the Fermi-Dirac distribution function, leading to a sharp transition upon cooling. Doping (i.e. biasing) graphene to a chemical potential near half of the photon's energy, a small switching energy is needed for of only >30meV for T = 0K versus ~100meV at T = 300K (Fig. 6). This difference in the minimum voltage of about 3x is equivalents to an energy saving of about 10-fold improvement of devices are operated at cry-temperatures. However, the voltages required in devices for actual devices is modulated from zero-chemical potential to the Pauli- blocking regime of about Δµc = 0.4-0.5eV. However, in a real device, one would only modulate around the point of half the photon energy, here ½ hv = 0.4eV, given the telecom wavelength. The latter introduces a voltage drop across the contacts, lowering the actual applied voltage range across the device capacitor. The metal contact from a plasmonic modulator offers here a unique advantage over photonic devices, since the drive voltage suffers no degradation in the contacts. Graphene's optical absorption (α) for telecom wavelength (λ = 1550 nm) relates correlates to an optical transmission change with bias (Fig. 7a,b). Graphene placed on photonic modes separated by a thin oxide creates a capacitor (Fig. 7d). The voltage steepness required to exploit the Pauli Blocking modulation is however relatively high, <0.1 dB/V when weak LMI photonic modes are used (Fig. 7b) [60]. With the aim of creating a steepest electro-optic 'switching' behavior, obtainable performance depends on the quality of both the electrostatics and the ability to deliver a voltage change to the optical mode without resistive losses, i.e. a low contact resistance is desired. This is in fact identical, from an electrostatics point of view, to FET device physics, where the switching steepness is quantified by the sub-threshold swing. Improving the electrostatics can be achieved in two ways; (i) improving the gate's control of the electro-optic effect, and (ii) reducing the series resistance. Thus, a higher capacitor while hurting the RC-delay of the modulator, is actually desired, since it enables a steeper switching transfer function. Practically this can be realized by reducing the capacitor's oxide thickness, and/or introducing high-k dielectrics. Secondly, the applied bias voltage drop should only appear at the capacitor (the device itself), and not at the contacts, or channel leading to the device. This means that photonic modulators are fundamentally challenged by the required doping to reduce series resistance near the contact regions (Fig. 7d). Plasmonic devices, in contrast, use their optically field-confining metal synergistically as a low-resistive contact (Fig. 7e). Simply put, plasmonics enables to obtain a capacitor with perfect spatial overlap relative to the device region, thus minimizing series- and contact resistances. While photonic modulator designs do have the flexibility of spatial selective doping, such as increasing the doping level near the contacts, the accompanying effect of the higher carrier concentration introduces parasitic optical losses due to increased absorption. Indeed, the main difference between photonic and plasmonic modulators is that in plasmonics losses are high but tolerable due to the increases LMI allowing λ-size short devices, but in photonics losses are to be avoided at all costs, since the required device lengths are 100-1000 times λ. Focusing on Graphene-based modulators, experimentally obtained power transmission changes per voltage-length, ΔPopt, are 0.013 dB/V-µm [53], which is then improved by about 2x in a double layer Graphene push-pull configuration resulting in 0.025 dB/V-µm [54]. Here we present first results of a plasmonic Graphene-based modulator in a hybrid plasmon-silicon integration configuration improving this performance by another 2x to 0.050 dB/V-µm (see next section for device details). This novel device utilizes the plasmonic LMI benefits resulting in device length shrinkage of about 5x (from 40 um down to 8 um) when comparing photonic designs [53,54]. We note, that the focus of this work is to explore devices that are optimized for power consumption and not for speed. However, when optimized for speed, this power metric drops to 0.003dB/V-um, but enables 35 GHz fast modulation [60]. Similarly, the contact resistances for experimental photonic devices is about 1,000Ω, which is one order of magnitude lower compared for plasmonic modulators; 50-200Ω for the Graphene contact and channel (depending how close this contact is to the device), and almost no resistance at the plasmonic contact ~10's Ω (Fig. 7e). The latter indeed provides a unique opportunity to design highly- energy efficient devices. A goal should be to design and demonstrate >3 dB/V-µm devices, which would enable sub-λ long and sub-1V efficient devices if a minimum of ER > -3 dB are required. This however necessitates an improvement of over 10x compared to our latest plasmon-photonic hybrid modulator. A possible roadmap towards this is to consider the optical polarization to match the in-plane component of 2D materials thus increasing the optical overlap factor as discussed in the next section. The schematic of this first Graphene-based hybrid-photonic-plasmon EAM on a silicon photonics platform uses the in Figure 7 discussed gating scheme. Based on the above considerations, we here show experimental results of a hybrid-plasmon Graphene-based electro- absorption modulator operating in the telecom C-band. The device is a single graphene layer sandwiched inside a Silicon-based HPP mode (Fig. 8a,b). This design allows for strong field confinement and decent modal overlaps (Γ ~ 5x10-4), when surface roughness of the ALD and metal gate deposition are considered (Fig. 8c). The latter is important since the plasmonics dictates that the electrical field lines are always perpendicular to the metal and hence to the graphene, which would result in a vanishing in-plane (Ex) field component in Graphene, thus zeroing out the modal overlap. However, the natural process roughness (~10 nm, mainly from the poly-crystalline grain boundaries of the metal deposition with respect to a 'flat' ALD process) actually helps in this work to provide for a small amount of in-plane graphene fields (Fig. 8c). Tuning the Fermi level of Graphene sandwiched inside this electrical MOS capacitor we achieve an extinction ratio of 0.4 dB/µm resulting in an ER efficiency per unit device-length of 0.05 dB/V-µm (Fig. 8d). This is enabled by a multitude of device improvements; (i) the index change of the active material is high (unity), (ii) the group index is relatively large (10), (iii) the overlap factor which not high is improved by the intrinsic roughness at the graphene-metal interface. Considering other device performance-related factors there are other fundamental benefits of this design to include the contact resistance (Rc = 210 Ω) can be fundamentally lower compared to any photonic (non-plasmonic) mode and cavity structures where any placement of the metallic contact close to the optical mode will introduce intolerable losses. This is different for our plasmonic mode, which is inherently lossy, but the polaritonic (matter-like) mode allows to scale-down the device into a few micrometer small device (a reduction of a factor of 100) compared to traveling-wave Silicon-based modulators. We refer to this design as an 'in-the- device-basing', as suppose to biasing the device few to 10's of micrometer away from the active region. As such, the overall design allows for a more compact overall footprint. Lastly, reducing the dielectric thickness (tox = 5±1 nm), improves the electrostatics enabling a sub-1 Volt modulation performance. In determining the power consumption of the device, here a circuit designed has options with respect to bias conditions; For instance, (i) a bias voltage of 0.75V enables 0.5dB/µm of ER, while (ii) a bias of 0.1V just 0.2dB/µm. Assuming small signal modulation requiring a minimum ER of -3dB a device length required are 6 µm and 15 µm, respectively. This results thence, in an E/bit = ½ CV2 of 2.6fJ/bit and 110 aJ/bit, respectively. Here the device areas are just 3.6- and 8.8 µm2, taking the waveguide width of 600nm as a lateral capacitor dimension. The latter, however, could be reduced by another factor of ~3 approaching the silicon waveguide cutoff, entering the ten's of aJ/bit regime. It is interesting to ask what the fundamental lower limit for modulator energies are given a desired BER and operating temperature, which is however not part of the discussion provided here. Suffice to say, for any charge-driven devices such as the Graphene ones considered here, the ultimate limit is likely set by broadening, γ, of the Fermi-level or any other relevant absorption states (depending on the modulation mechanism). Thus, the minimum required drive voltage is therefore expected to be Boltzmann approximation smeared' on the order of a few kBT. An improvement from the HPP-based Graphene EAM discussed in Figure 8 is a device that ensured that the optical field density is in-plane with the lateral dimension of the 2D material. This can be achieved, for instance, when the 2D material is combined with slot-waveguides; here Graphene could be either above the slot [61] or below as discussed in this work. Results show a high modulation performance of 1.2dB/µm for thin and narrow metal slots, given a ER metric of 2 dB/V-µm for 0.5V of bias change (Fig. 5c) [49]. This is indeed close to the set grand challenge of 3 dB/V-µm. To provide an outlook, further device improvements should consider dual gating in a push-pull configuration similar to ref [54], but with two pairs of Graphene layers above and below the slot (Fig. 9). The latter could result in about 4dB/µm of switching, enabling a just 770 nm long, or about ½λ. While not as compact as the atomistic switch from ref [62] where just a few atoms (possible one single atom) control the modulation by cutting off a surface plasmon gap mode, a Graphene modulator is expected to switch faster than metal migration-based switching mechanisms. An estimated energy consumption for the 4-Graphene layer modulator assuming the same lateral and gate-oxide dimension as for our actual device from Figure 8, gives a miniscule power consumption of 170 aJ/bit for 0.75V of bias, respectively (taking ER = 3dB, Fig. 9). We note, that the latter is as efficient as a single FET without driving connecting via (Fig. 10). Coming back to drawing a parallel between the energy-per-bit function of FETs and modulators, one can show a declining trend with time (Fig. 10) [63]. Based on IBMs device scaling, where physical dimensions of the FET and other parameters such as doping level and footprint scale as a function of a universal scaling factor, K [64]. As the critical dimension of the FET scales down, its energy consumption also declines simply from reduced capacitance mainly driven by improved electrostatics (blue dots, Fig. 10). Now, at the end of Moore's law, sub 10 nm feature sizes allow energy functions <104 kBT corresponding to 10's of aJ/bit for the FET gate alone. However, adding the wires to control this MOS capacitor adds about 2-orders of magnitude to the power consumption. The key question for electronics is therefore who long does the connection to the device be, and the answer lie in both circuit and interconnect design. Still, the charging and discharging of wires is a fundamental challenge in electronics [4], impacting emerging technologies such as neuromorphic memristors devices and crossbar architectures [65]. It is therefore interesting to ask how the energy consumption at the device level compares between electronics and photonics; where the promise in the latter is based on the 'wires' in optics being only limited by light propagation delay and power consumption of the laser determined by the detectors sensitivity and desired BER or SNR. Thus, if an equivalent energy consumption per switch is possible, photonics would see an performance improvement over electronic links of up to about 250x using WDM and medium fast EOM drivers (10's GHz) [66], high higher improvements depending on link details such as length [5]. Mapping the recent advances of modulators into this 'other Moore's Law' (Fig. 10), shows that consistent improvements in active material selection, photonic-plasmonic hybridization (mode and plasmon/photon devices for active/passive light control), allows approaching energy-per-bit functions comparable to FETs (i.e. 10's of aJ/bit). Yet, 1kBT would be the ultimate limit and be about 5 zJ/bit given T = 300K. Thus a technological gap of 1000x (between 1 and 1,000 kBT) exists, where it is unclear how this gap could be bridged. Energy levels below 1 kBT are fundamentally unattainable given broadening such as from thermal and other effects. Conclusions In conclusion, we have discussed a roadmap for atto-Joule efficient electro-optic modulators. We discussed light matter interactions required to approach on this goal leading to a required optical concentration factor proportional to the Purcell factor. Selecting three material-modulation mechanisms utilizing charge, namely silicon, ITO, and Graphene, we showed how optical mode designs impact obtainable effective index changes as a function of the optical overlap factor, the obtainable material index change, and the mode's effective group index. Results show that the weak plasma dispersion of Silicon limits achievable extinction ratio per nominal device length, whereas the strong index modulation of (and above) unity of TCO and 2D materials such as Graphene overcompensates the low optical overlap factor (10-4-10-3). Cooling a Graphene-based modulator enables a about 10x lower energy-per-bit function by improving the Fermi-Dirac function-based switching steepness of Graphene, based on Pauli blocking. Furthermore, we showed that improving the capacitive electrostatics of any modulator improves achievable extinction ratio per applied voltage, a value similar to the sub-threshold swing in transistors. We experimentally demonstrate a reduction of 10x in the switching steepness in a hybrid-photon- plasmon mode Graphene-based electro-absorption modulator on Silicon with a modulation efficiency of 0.05 dB/V-µm requiring sub-1V of drive voltages. The plasmonic metal serves here synergistically as a gate to drive the capacitor, by lowering the series resistance (~200Ω) leading up to the device – a fundamental advantage of plasmonics over photonic-based devices. The latter performance can be further improved by ensuring that the polarization of the optical or plasmonic mode are in-plane with the 2D material, such as by a proposed metal slot-based Graphene modulator. Deploying two pairs of Graphene, one above and one below the slot, we showed a design for a 4dB/µm efficient modulator, resulting in a 770 nm short modulator device length for -3dB of signal modulation. We explored two drive voltage options of 0.75V or 0.1V, where the latter increases the device length to achieve the required 3dB modulation depth. The experimental hybrid photon plasmon modulator requires just 2.6fJ/bit or 110 aJ/bit depending on which bias voltage was used, which shows the expected benefits for treading-in voltage for length, provided the RC-delay still supports the desired driver speeds. The latter however, seems not to bear any advantages beyond 10 GHz given circuit (driver) power consumption and thermal instabilities. Lastly, we show that a double dual-gate plasmon slot Graphene modulators enables in-plane optical polarization with the Graphene film, resulting in quite efficient modulation of ~4dB/µm, a resulting 770 nm short device length, and just 170aJ/bit for a bias voltage of 0.75V. We note, however, that the latter is below the required signal to noise ratio of detectors for bit- error-rates anticipated for on-chip communication. Acknowledgements V.S. is supported by ARO (W911NF-16-2-0194), by AFOSR (FA9550-14-1-0215) and (FA9559-15-1-0447), which is part of the data-driven applications system (DDDAS) program. H.D. and V.S. are supported by AFOSR (FA9550-17-P-0014) of the small business innovation research (SBIR) program. Figures Figure 1. Enhancing Light-matter-interactions to enable atto-Joule efficient modulation. The orders-of-magnitude large device size of optoelectronic components compared to electronic counterparts can be addressed by either increasing the interaction time leading to high-Q devices, or by matching the optical-to-electronic wavelength such as in polaritonic (matter-like) modes. If millimeter large interferometer-based modulators are to be avoided, cavity-based modulators increase the interaction time by folding light spatially in space. Their spectral sensitivity, however, requires active thermal control, which is energy costly. Polaritonic modes, in contrast, performance an impedance matching realizing compact, energy efficient, and non-photon lifetime-limited fast switching. It is the latter that are discussed in this work, and we show experimental proof that the concept leads in 1st-generation devices switching at 100's aJ/bit. Interestingly, the ratio of Q/Vmode (Vmode = optical mode volume) is the optical concentration factor, which fundamentally increases the light-matter-interactions, and is proportional to the Purcell factor. a Fabry-Pérot (FP) cavity comprised of a dielectric material sandwiched by a pair of highly of scaling. c quality (Q) factor, d optical mode volume, Vm, and e Purcell factor, Fp. While the general trend shows a reduced Q upon scaling, significant differences between the three cavity Figure 2. Scaling Laws of electro-optic modulators, answering the question whether scaling improves modulator performance similarly to transistor scaling. a, Electro-optic modulator schematic showing spatial dimensions and light propagation. b, We explore three device- underlying cavity types; a ring resonator (RR) cavity with the waveguide width, !, and radius, !; reflecting metal mirrors (reflectivities !! and !!); and a plasmon cavity formed by metal nanoparticle (MNP) embedded in a dielectric, where ! is the particle radius. The modulator scaling parameters are r (RR), l (FP), and ! (MNP) cavity. c$e, Cavity performance as a function types exist. Parameters: propagation loss of a diffraction limited beam, !!=1.0 dB/cm used in the RR; Silver metal mirrors, !!"= 0.41+10.05i, the dielectric refractive index !!=3.0- i0.001, Silver conductivity !!"=6.3×107 mho/m, and the damping rate for the MNP to be !!=2.0×1015 rad/s. Each cavity type shows a maximum Purcell factor given by the ideal optical confinement- to-loss point. f, The energy efficiency of the modulator (E/bit) scales with ~(QFp)-1. Thus, improving the optical confinement helps to reduce the energy efficiency, while a higher Q offers further support. 100's of atto-Joule efficient modulators are possible with pure plasmonic modes allowing for high-density optical circuits due to sub-micron compact critical device lengths [38]. Figure 3. Optical mode exploration for studying the material-field interactions such as the modal overlap factor and the effective mode's group index upon modulation. Materials considers are Silicon and merging materials (ITO, Graphene) a$i, Schematic of the mode structures and FEM simulated mode profiles for all the structures at their respective starting point from the material dispersion at !=1550!!". The relevant parameters are !/!!"= 451!!" , !/!!"#=800!!" , ℎ!"=ℎ!"#=200!!" , ℎ!"!!"#=600!!" , ℎ!"#$=100!!" , !!"#$=300!!" , ℎ!!!"#$=30!!" , !=20!!" , !!"#$%=550!!" , !!"#=300!!" , ℎ!"#$%=20!!", ℎ!"!!!"=10!!" and !!"!!!"=250!!". The simulated results are shown in thickness ℎ!"=5!!" to ensure similar electrostatics [19,20]. a'$i', respective optical field log scale due to their largely varying electric field strengths. All gate oxides in this work have profiles. Figure 4. a, Optical overlap (confinement) factors andb, effective mode group index for waveguide options from Figure 3 as a function of carrier concentration, i.e. capacitive gating (charge modulation). a, Confinement factors corresponding to the Si and ITO modes vs. carrier concentration and b, confinement factors corresponding to the Graphene modes vs. chemical potential b. (i, ii) ITO slot at 1019cm-3 and 6×1020 cm-3, (iii, iv) ITO hybrid at 1019 cm-3 and 6×1020 cm-3; respectively [19,20]. Results show a relatively strong change in the optical mode overlap and effective group index for ITO slot and hybrid designs near the ENZ point of ITO. The weak index tuning of the Silicon's plasma dispersion, on the other hand, keeps both the parameters almost flat, indication a low modulation potential. Graphene's slot and hybrid modulator devices show a medium-strong modulation for Pauli blocking with changing chemical potential, µc. Figure 5. Material impact on absorption modulation. a and b, Material-based optical absorption for electro-absorption modulators vs. charge (i.e. carrier concentration, n2D). The free- carrier-based absorption (Silicon, ITO) scales with carrier concentration (and bias, b) due to the Drude model. All other modulation mechanisms rely on some form of 'band-filling' leading to absorption blocking, such as the Pauli blocking in Graphene. 2D material TMDs show in general the highest absorption (see main text), but quantum wells and quantum dots the steepest switching. c, The modulation performance of an electro-absorption modulator improves with overlap factor. However, the weak effective mode index change of Silicon is outperformed by emerging materials such as ITO or Graphene despite the low optical overlap factor [49]. λ = 1550 nm. Figure 6. Graphene's real (a) and imaginary (b) refractive index tuning as function of chemical potential and temperature. The steepest index change is, as expected, near the Pauli blocking edge of 0.4eV for a telecom beam at 0.8eV (λ = 1550nm, Kubo formulism used) [49]. An energy efficient modulator would change its ON-OFF states around the 0.4eV point. Here cooling the device could allow a 10x energy reduction, since E/bit scales with ~V2. the Graphene extinction coefficient at impact of plasmonics. Figure 7. Electro-absorption modulator electrostatics and Considering telecom wavelengths (a), device performances depend on the electrostatics of the device; i.e. low contact and series resistances to drive the device capacitor with minimal voltage drops at the contacts leading to the device region. Transmission changes using photonic modes are fundamentally challenged by balancing contact resistance (i.e. doping Silicon higher), vs. minimizing optical losses (b,d). In plasmonics, The metal contact not only firms the optical light-matter-interaction enhanced mode positively impacting the effective group index and modal overlap factor, but also enables to place the capacitor precisely on top of the device region (e). The effect is that the required modulation voltage drops from, for example 6V [54] to <1V (c). This 10x reduction is driven by a lower contact resistance (e), and a lower oxide thickness (see Fig. 8 for details shown in panel c). Figure 8.a, Schematic of a hybrid-photon-plasmon graphene-based electro-absorption modulator. The modulation mechanism is based on Pauli-blocking upon gating the Fermi level of graphene. b, Silicon waveguide-integrated modulator. A cw laser (λ = 1.55 µm) is fiber coupled into the SOI waveguide via grating couplers. Device length, L = 8 µm, tox = 5 nm. c, Electric field density across the active MOS region of the modulator showing an enhanced field strength coinciding with the active graphene layer. This improves the optical overlap factor by about 25%. Taking into consideration the grain boundaries introduced during the metal deposited creates in-plane field vectors inside the graphene layer. d, Modulator transfer function; normalized modulation depth at different drive voltages (VD). The modulator performance yields a high extinction ratio of 0.4-0.5 dB/µm, due to the combination of the plasmonic MOS mode enhancing the electroabsorption in the active region. Figure 9. Push-pull dual gating of graphene plasmon slot modulators. a, Schematic of the plasmonic slot EAM using multiple Graphene layers. tox = 5 nm oxide layer separates neighboring graphene layers. b, Increasing the number of Graphene layers ER increases the extinction ratio (ER) about linearly. For the 4-layer graphene case, the device length is about ½λ (770 nm), requiring 170 aJ/bit. This could be further lowered by a lower drive voltage, and but longer device length. Figure 10. The other Moore's Law: Energy/bit vs. time for FETs and modulators. FETs are able to switch at the aJ/bit at the device level, but require high parasitic electrical connectors, which increase the energy-per-bit function by about 100x [62]. Optoelectronics is limited to 10- 100+ fJ/bit efficiencies, dramatically lagging electronic transistors performance, which illustratively highlights the weak light-matter-interaction. Applying the design physics criteria discussed in this work shows that 10's of aJ/bit efficient devices are possible, when combining highly index-changing emerging materials with polaritonic modes while optimizing polarization, mode overlap, effective index, contact resistance, and capacitance. Our ITO and Graphene-based plasmon Silicon hybrid integrated devices already perform at the 100's of aJ/bit level. Going beyond this level, one can introduce low Q-cavities and multi-gating schemes as discussed in Figure 9, or explore other stronger (steeper) switching materials such as quantum dots in conjunction with plasmonics or polaritonic modes. Having almost reached the long-standing goal of merging lengths scales of photonics and electronics, one can estimate the fundamental performance benefits of optics over electronics in data communication; the faster device speed enabled by only driving the sub-micrometer small photonic device capacitor (vs. the device and wire in electronics) allows for shorter RC delays, but also for lower dissipative energy consumption. In addition WDM offers data processing parallelism over electronics, thus resulting in an improvement factor of about 250x. This is just a device-level analysis, and a proper benchmark should include the link, and circuit level. The minimal fundamental switching energy is given by broadening effects such as temperature 'smearing' of the Fermi level of about a few kBT. References [1] E. L. Wooten et al., "A review of lithium niobate modulators for fiber-optic communications systems," in IEEE Journal of Selected Topics in Quantum Electronics, vol. 6, no. 1, pp. 69-82, Jan.-Feb. 2000. [2] Y. Shi, C. Zhang, et al. "Low (Sub-1-Volt) Halfwave Voltage Polymeric Electro-optic Modulators Achieved by Controlling Chromophore Shape" Science, 119-122 (2000). [3] Y. A. Vlasov, "Silicon CMOS-integrated nano-photonics for computer and data communications beyond 100G," in IEEE Communications Magazine, vol. 50, no. 2, pp. s67-72 (2012). [4] D. A. B. Miller, "Attojoule Optoelectronics for Low-Energy Information Processing and Communication" J. LIghtwave Tech., 35, 3 (2017) [5] S. Sun, A. Badaway, V. Narayana, T. El-Ghazawi, V. J. Sorger, "Photonic-Plasmonic Hybrid Interconnects: Efficient Links with Low latency, Energy and Footprint", IEEE Photonics Journal, 7, 6 (2015). [6] S. Sun, V. Narayana, T. El-Ghazawi, V. J. Sorger, "A Universal Multi-Hierarchy Figure-of- Merit for On-chip Computing and Communications", arXiv:1612.02486 (2016). [7] W. T. Wu and A. Louri, "A Methodology for Cognitive NoC Design," in IEEE Computer Architecture Letters, vol. 15, no. 1, pp. 1-4, (2016). [8] V. K. Narayana, S. Sun, A.-H. Badawya, V. J. Sorger, T. El-Ghazawi, "MorphoNoC: MorphoNoC: Exploring the Design Space of a Configurable Hybrid NoC using Nanophotonics," Microprocessors and Microsystems, (2017). [9] IMEC ePIXfab SiPhotonics: ISIPP50G, http://www.europractice- ic.com/SiPhotonics_technology_imec_ISIPP25G.php, accessed 9/14/2017. [10] Liu, Ke, et al. "Review and perspective on ultrafast wavelength-size electro-optic modulators." Laser & Photonics Reviews 9.2 (2015): 172-194. [11] C. Ye, S. Khan, Z.R. Li, E. Simsek, V. J. Sorger, "l-Size ITO and Graphene-based Electro- optic Modulators on SOI" IEEE Selected Topics in Quantum Electronics, 4, 20 (2014). [12] S. K. Pickus, S. Khan, C. Ye, Z. Li, and V. J. Sorger, "Silicon Plasmon Modulators: Breaking Photonic Limits" IEEE Photonic Society, 27, 6 (2013). [13] C. Huang, S. Pickus, R. Lamond, Z. Li, V. J. Sorger, "A Sub-λ Size Modulator Beyond the Efficiency-Loss Limit" IEEE Photonics Journal 5, 4 (2013). [14] Z. Ma, Z. Li, K. Liu, C. Ye, V. J. Sorger, "Indium-Tin-Oxide for High-performance Electro- optic Modulation", Nanophotonics, 4, 1 (2015). [15] J. Leuthold et al., "Silicon-Organic Hybrid Electro-Optical Devices," in IEEE Journal of Selected Topics in Quantum Electronics, vol. 19, no. 6, pp. 114-126, Nov.-Dec. 2013. [16] M. Lauermann, et al. "40 GBd 16QAM Signaling at 160 Gb/s in a Silicon-Organic Hybrid Modulator", Lightwave Technology Journal of, vol. 33, pp. 1210-1216, ISSN 0733-8724 ( 2015). [17] S. Wolf, et al. "DAC-Less Amplifier-Less Generation and Transmission of QAM Signals Using Sub-Volt Silicon-Organic Hybrid Modulators", Lightwave Technology Journal of, vol. 33, pp. 1425-1432, ISSN 0733-8724 (2015) [18] R. Palmer, et al. "Low power mach-zehnder modulator in silicon-organic hybrid technology", IEEE Photonics Technol. Lett., vol. 25, no. 13, pp. 1226-1229, (2013). [19] R. Amin, C. Suer, Z. Ma, J. Khurgin, R. Agarwal, V. J. Sorger, "Active Material, Optical Mode and Cavity Impact on electro-optic Modulation Performance" arXiv:1612.02494 (2016) (accepted at Nanophotonics, 8-2017). [20] R. Amin, C. Suer, Z. Maet al., "A Deterministic Guide for Material and Mode Dependence of On-Chip Electro-Optic Modulator Performance", Solid-State Electronics, Special Issue, DOI: 10.1016/j.sse.2017.06.024 (2017). [21] Z. Ma, R. Hemnani, L. Bartels. R. Agarwal, et al. "2D Materials in Electro-optic Modulation: energy efficiency, electrostatics, mode overlap, material transfer and integration" arXiv:1708.05986 (2017). [22] Midolo, Leonardo, et al. "Electro-optic routing of photons from single quantum dots in photonic integrated circuits." arXiv preprint arXiv:1707.06522 (2017). [23] Edwards, Elizabeth H., et al. "Low-voltage broad-band electroabsorption from thin Ge/SiGe quantum wells epitaxially grown on silicon." Optics express 21.1 (2013): 867-876. [24] J. H. Shin and N. Dagli, "Ultralow Drive Voltage Substrate Removed GaAs/AlGaAs Electro-Optic Modulators at 1550 nm," in IEEE Journal of Selected Topics in Quantum Electronics, vol. 19, no. 6, pp. 150-157 (2013). [25] A. Fratalocchi, C. M Dodson, R. Zia, P. Genevet, E. Verhagen, H. Altug, V. J. Sorger, "Nano-optics gets practical: Plasmon Modulators", Nature Nanotechnology, 10, 11-15 (2015). [26] V. J. Sorger, R. F. Oulton, R.-M. Ma, X. Zhang "Toward integrated plasmon circuits" MRS Bulletin 37, 8, 728-738 (2012). [27] R.-M. Ma, et al. "Plasmon lasers: coherent light source at molecular scales", Laser & Photonics Reviews 7, 1-21 (2012). [28] V. J. Sorger, N. Pholchai, E. Cubukcu, R. F. Oulton, P. Kolchin, C. Borschel, M. Gnauck et. al "Strongly Enhanced Molecular Fluorescence inside a Nanoscale Waveguide Gap" Nano Letters 10.1021/nl202825s (2011). [29] V. J. Sorger, Z. Ye, R. F. Oulton, G. Bartal, Y. Wang and X. Zhang "Experimental demonstration of low-loss optical waveguiding at deep sub-wavelength scales" Nature Communication 2, 331 (2011). [30] R.-M. Ma, R. F. Oulton, V. J. Sorger and X. Zhang "Room-temperature sub-diffraction- limited plasmon laser by total internal reflection" Nature Materials 10, 110-113 (2010). [31] R. F. Oulton, V. J. Sorger, T. Zentgraf, R.-M. Ma, C. Gladden, L. Dai, G. Bartal and X. Zhang "Plasmon Lasers at Deep Subwavelength Scale" Nature 461, 629-631 (2009). (equal author contribution) [32] V. J. Sorger, R. F. Oulton, J. Yao, G. Bartal and X. Zhang "Fabry-Perot Plasmonic Nanocavity" Nano Letters 9, 3489-3493 (2009). [33] B. Min, E. Ostby, V. J. Sorger, E. Ulin-Avila, L. Yang, X. Zhang, K. Valhalla "High-Q surface plasmon-polariton whispering-gallery microcavity" Nature 457, 455-458 (2009). [34] R. F. Oulton, V. J. Sorger, D. F. B. Pile, D. Genov, X. Zhang "Nano-photonic confinement and transport in a hybrid semiconductor-surface plasmon waveguide" Nature Photonics 2, 496- 500 (2008). [35] Phare, Christopher T., et al. "Graphene electro-optic modulator with 30 GHz bandwidth." Nature Photonics 9.8 (2015). [36] Lin, Hongtao, et al. "Breaking the energy-bandwidth limit of electrooptic modulators: theory and a device proposal." Journal of Lightwave Technology31.24: 4029-4036 (2013). [37] AIM Photonics, http://www.aimphotonics.com/ accessed 9/14/17. [38] Liu, Ke, et al. "Fundamental scaling laws in nanophotonics." Scientific Reports 6 (2016). [39] Sorger, V.J., Oulton, R.F., Yao, J., Bartal, G. & Zhang, X. Plasmonic fabry- pérot $nanocavity. Nano Lett. 9, 3489-3493 (2009). $ [40] Liu, K. & Sorger, V.J. Enhanced interaction strength for a square plasmon resonator $embedded in a photonic crystal nanobeam cavity. J. Nanophotonics. 9, 093790- $093790 (2015). $ [41] Anderson, P.A., Schmidt, B.S. & Lipson, M. High confinement in silicon slot waveguides with sharp bends. Opt. Express. 14, 9197-9202 (2006).$ [42] Zanotto, S., Morichetti, F. & Melloni, A. Fundamental limits on the losses of phase and amplitude optical actuators. Laser Photon. Rev. 9, 666-673 (2015). 14.Manipatruni, S., Lipson, M. & Young, I.A. Device scaling considerations for nanophotonic CMOS global interconnects. IEEE J. Sel. Top. Quantum Electron. 19, 8200109 (2013).$ [43] Yariv, A. Critical coupling and its control in optical waveguide-ring resonator systems, IEEE Photon. Tech., Lett. 14, 483-485(2002).$ [44] Sun, G., Khurgin, J.B. & Soref, R.A. Plasmonic light-emission enhancement with $isolated metal nanoparticles and their coupled arrays. J. Opt. Soc. Am. B 25, 1748- $1755 (2008). $ [45] Haffner, C., et al. "All-plasmonic Mach–Zehnder modulator enabling optical high-speed communication at the microscale." Nature Photonics 9.8 (2015): 525-528. [46] K. Liu, N. Li, D. K. Sadana, V. J. Sorger, "Integrated nano-cavity plasmon light-sources for on-chip optical interconnects" ACS Photonics, 3. 233-242 (2016). [47] N. Li, K. Liu, D. K. Sadana, V. J. Sorger, "Nano III-V Plasmonic Light-Sources for Monolithic Integration on Silicon", Nature: Scientific Reports, 5, 14067 (2015). [48] K. Liu, V. J. Sorger, "An electrically-driven Carbon nanotube-based plasmonic laser on Silicon" Optical Materials Express, 5, 1910-1919 (2015). [49] Z. Ma, M. H. Tahersima, S. Khan and V. J. Sorger, "Two-Dimensional Material-Based Mode Confinement Engineering in Electro-Optic Modulators," IEEE Journal of Selected Topics in Quantum Electronics, vol. 23, no. 1, 1-8 (2017). [50] Z. Ma, M. Tahersima, R. Amin, S. Khan, V. J. Sorger, "Sub-wavelength Plasmonic Graphene-based Slot Electro-optic Modulator" OSA Frontiers in Optics (FiO) (2017) (accepted 7-2017). [51] V. J. Sorger, D. Kimura, R.-M. Ma and X. Zhang „Ultra-compact Silicon nanophotonic Modulator with broadband Response" Nanophotonics 1, 1, 17-22 (2012) [52] C. Ye, K. Liu, R. Soref, and V. J. Sorger, "3-Waveguide 2x2 Plasmonic Electro-optic Switch" Nanophotonics, 4, 1, pp. 261-268 (2015). [53] Liu, Ming, et al. "A graphene-based broadband optical modulator." Nature 474.7349 (2011): 64-67. [54] Liu, Ming, Xiaobo Yin, and Xiang Zhang. "Double-layer graphene optical modulator." Nano letters 12.3 (2012): 1482-1485. [55] Ansell, D., et al. "Hybrid graphene plasmonic waveguide modulators." Nature Communications 6 (2015). [56] Shainline, Jeffrey M., et al. "Superconducting optoelectronic circuits for neuromorphic computing." Physical Review Applied 7.3 (2017): 034013. [57] Liu, Chang-Hua, et al. "Nanocavity integrated van der Waals heterostructure light-emitting tunneling diode." Nano letters 17.1 (2016): 200-205. [58] Wan, Zhenni, et al. "A forming-free bipolar resistive switching behavior based on ITO/V2O5/ITO structure." Applied Physics Letters 111.4 (2017): 041601. [59] Ye, Yu, et al. "Electrical generation and control of the valley carriers in a monolayer transition metal dichalcogenide." Nature nanotechnology 11.7 (2016): 598-602. [60] Dalir, Hamed, et al. "Athermal broadband graphene optical modulator with 35 GHz speed." ACS Photonics 3.9 (2016): 1564-1568. [61] Ding, Yunhong, et al. "Efficient graphene based electro-optical modulator enabled by interfacing plasmonic slot and silicon waveguides." arXiv preprint arXiv:1610.05352 (2016). [62] Emboras, Alexandros, et al. "Atomic scale plasmonic switch." Nano letters16.1 (2015): 709-714. [63] Class notes from Eli Yablonovitch, (2009). [64] Wong, H-S. Philip. "Silicon Devices at the "End of Scaling"–Opportunities and Challenges." [65] Adam, Gina C., et al. "3-D memristor crossbars for analog and neuromorphic computing applications." IEEE Transactions on Electron Devices 64.1 (2017): 312-318. [66] Sun, Chen, et al. "Microprocessor chip with photonic I/O." Optical Fiber Communications Conference and Exhibition (OFC), 2017. IEEE, 2017.
1902.07394	1	1902	2019-02-20T03:58:32	Nanostructuring of LNOI for efficient edge coupling	[ "physics.app-ph", "physics.optics", "quant-ph" ]	We present the design, fabrication and characterization of LNOI fiber-to-chip inverse tapers for efficient edge coupling. The etching characteristics of various LNOI crystal cuts are investigated for the realization of butt-coupling devices. We experimentally demonstrate that the crystal cut limits the performance of mode matching tapers. We report a butt-coupling loss of 2.5 dB/facet and 6 dB/facet by implementing 200 nm tip mode matching tapers in $+Z$-cut LNOI and $X$-cut MgO:LNOI waveguides with low propagation loss. We anticipate that these results will provide insight into the nanostructuring of LNOI and into the further development of efficient butt-coupling in this platform.	physics.app-ph	physics	Nanostructuring of LNOI for efficient edge coupling Inna Krasnokutska,1 Jean-Luc J. Tambasco,1 and Alberto Peruzzo1, ∗ 1Quantum Photonics Laboratory and Centre for Quantum Computation and Communication Technology, School of Engineering, RMIT University, Melbourne, Victoria 3000, Australia We present the design, fabrication and characterization of LNOI fiber-to-chip inverse tapers for efficient edge coupling. The etching characteristics of various LNOI crystal cuts are investigated for the realization of butt-coupling devices. We experimentally demonstrate that the crystal cut limits the performance of mode matching tapers. We report a butt-coupling loss of 2.5 dB/facet and 6 dB/facet by implementing 200 nm tip mode matching tapers in +Z-cut LNOI and X-cut MgO:LNOI waveguides with low propagation loss. We anticipate that these results will provide insight into the nanostructuring of LNOI and into the further development of efficient butt-coupling in this platform. I. INTRODUCTION Lithium niobate on insulator (LNOI) and magnesium oxide doped LNOI (MgO:LNOI) have recently emerged as promising photonic platforms. Many key components have been demonstrated in LNOI including low-loss waveg- uides [1], electro-optical modulators [2, 3], and wavelength converters [4], while MgO:LNOI has been minimally in- vestigated despite its superior high power properties over LNOI [5 -- 7]. Efficient coupling of light into LNOI waveg- uides is a major challenge yet to be overcome before this platform can be competitive with high-index contrast platforms including silicon on insulator and silicon nitride [8 -- 10]. Broadband, efficient and polarization insensitive coupling of light into photonic chips is essential for de- vices including Mach-Zehnder modulators and wavelength converters, as well as for reliable packaging of photonic chips [11 -- 13]. Several approaches including the use of grating cou- plers, lensed fibers, high numerical aperture fibers and inverse tapers have been used to improve coupling effi- ciency [8, 14 -- 16]. Grating couplers, which are commonly used for vertical coupling into the chip, allow good cou- pling efficiency and the ability to access circuits inside a chip with alignment tolerances; however, compared to butt-coupling, limit the wavelength and polarization per- formance of devices and are prone to fabrication errors [8, 17, 18]. Butt-coupling is insensitive in wavelength and polarization; however, suffers from high mode-mismatch losses unless mitigated, for example, with spot size con- verters via inverse tapering [15, 16, 19, 20]. It has been demonstrated that LNOI can be nanos- tructured [1, 2, 17, 21], though the fabrication of small features required for inverse tapers is challenging due to the highly isotropic etching of LN in argon (Ar) plasma, dependent on crystal cut and MgO doping. The non- vertical sidewall angle limits the minimum feature size down to which LNOI may be structured. In this paper, we analyze the material limitations in nanostructuring various LNOI and MgO:LNOI faces. We ∗ [email protected] then design, fabricate and characterize mode-matching tapers on +Z-cut LNOI and X-cut MgO:LNOI with cou- pling loss of 2.5 dB/facet and 6 dB/facet respectively with a 200 nm tip. II. NANOSTRUCTURING LNOI Recently, we demonstrated that by mixing argon and fluorine ions during plasma reactive ion etching (RIE) it is possible to achieve low loss rib waveguides with sidewall angle of 75° on +Z-cut LNOI [21]. In this paper, we apply the developed fabrication process to a range of com- mercially available LNOI types, reporting their etching characteristics in Tab. I. The nanostructuring charac- terization enables to determine the minimum achievable feature size afforded by the particular cut and MgO dop- ing of the LNOI platform. Figure 1 shows the dependence of minimum feature size on sidewall angle. We calculate the the minimum bottom width of a feature that can be fabricated for different etching depths up to 700 nm. It can be seen that sidewall angle strongly limits nanostuc- turing of LN films. For example, the 50° angles resulting from Ar milling will, in the best case, allow 500 nm wide features etched 300 nm deep, making the design and fabrication of efficient mode matching inverse tapers and gratings challenging. On the other hand, +Z-cut LNOI has a 75◦ etch angle, which allows feature widths down to ∼ 180 nm for the same 300 nm etch depth. In order to investigate limitations of the fabrication pro- cess, we choose five LN samples: +Z-cut LNOI and X-cut LNOI with film thickness 500 nm, +Z-cut MgO:LNOI and X-cut MgO:LNOI with LN thin film thickness of 600 nm, and −Z LN bulk substrate 500 µm thick. All samples are then patterned according to the process discussed in the [21] to achieve a ∼1 µm wide metal mask for rib waveg- uides. The samples are then RIE etched with the same recipe and their etching characteristics are deduced based on SEM imaging of the waveguide cross section as well as atomic force microscopy (AFM) measurements. Finally, 2 FIG. 1. Calculated minimum feature size versus sidewall angle for several etch heights. The minimum feature size that can be achieved for +Z-cut LNOI and X-cut MgO:LNOI are shown as inset. The schematic representation of minimum feature size definition is shown on the side of the picture. A thin LN film (blue) is seated atop SiO2 (pink), where f is the width of the minimum achievable feature and is dependent on the sidewall angle, θ, and the etch depth, H. TABLE I. Etching characteristics of different LNOI types. Material Etching rate [nm/min] of Degree Anisotropy, D Sidewall an- gle [◦] Propagation loss [dB/cm] Z-cut LNOI 18 (+Z) LN Z-cut (−Z) 18.6 14.6 LNOI X-cut MgO:LNOI Z-cut MgO:LNOI X- cut 14 16.6 0.86 -- 0.76 0.25 0.45 75 78 70 60 71 less than 0.1 not on insulator less than 0.1 1 less than 0.1 the waveguides fabricated on LNOI and MgO:LNOI are cladded with 3 µm of plasma enhanced chemical vapor deposition (PECVD) SiO2, diced on a Disco DAD-321 using optical grade dicing, and their optical propagation losses measured. The propagation loss of −Z LN bulk was not measured due to the absence of bottom cladding. We analyze various etching properties of LNOI includ- ing etch rate, sidewall angle, degree of anisotropy and propagation loss. The degree of anisotropy, D, is defined as D = 1 − B/2H, where B = Wmask-Wf inal is the etch bias, defining the amount of lateral etching, Wmask and Wf inal are the pre-etching and post-etching widths, and H is the etch depth. Z-cut LN is more chemically active in fluorine plasma than X-cut, as a result, it is possible to achieve faster etching rates, better anisotropy and near-vertical sidewall angles with Z-cut; this is reflected in our tests -- we observe slower etching rates and a smaller degree of anisotropy for X-cut LNOI and we found that it possesses more shallow sidewall angle, however we did not observe effect of this on the optical propagation loss Tab. I, measured by using the Fabry-Perot technique and discussed in our previous papers [21, 22]. It is well known that ±X faces of LN have similar etching characteristics, meanwhile large differences can be observed for +Z and −Z faces, that was confirmed in our experiments and reported in the Tab. I. It was also shown that MgO:LNOI possesses different characteristics to not doped LNOI. It is more resistant to the etching process, moreover exhibits less anisotropy and Z-cut MgO:LNOI possesses the slowest etching rate and 3 FIG. 2. Design of inverse taper: (a) schematic representations of light coupling into the waveguide via an inverse taper; (b) simulated dependence of the overall transmission through waveguide via taper length for different taper dimensions; (c) simulated mode profiles for tapers with different base widths for X-cut MgO:LNOI and Z-cut LNOI illustrating the increase in MFD; (d) the simulated mode profiles for the untapered waveguides in X-cut MgO:LNOI and Z-cut LNOI discussed in this paper. the most shallow angle compared to the other investigated LN crystal cuts. These results can be explained by its faster etching rate in lateral direction, which reduces the waveguide width resulting in increased propagation loss due to the enhanced light interaction with the waveguide sidewalls causing increased scattering losses. III. LNOI INVERSE TAPER The typical base width of a C-band single mode LNOI waveguide is < 1.2 µm in Fig. 2(d), resulting in a mode field diameter (MFD) of ∼1.5 µm; meanwhile, the typical MFD of an optical single mode (SM) fiber is ∼10 µm at 1550 nm wavelength light, resulting in a significant mode mismatch between the optical fiber and the waveguide, increasing the butt-coupling loss. By exchanging the SM fiber for a lensed fiber, which has a ∼2.5 µm MFD across the C-band, improved mode matching can be achieved; however, there is still significant mode mismatch. By combining lensed fibers with inverse tapers, a drastic improvement to the waveguide MFD can be obtained, enabling good mode matching. The inverse tapers also have the added benefit of matching the effective index of the waveguide to that of the fibre, minimizing back reflections. Fig. 2(a) shows a schematic representation of the de- signed spot size converter. The device consists of a linear inverse taper on LNOI, which gradually becomes thinner towards the chip facets. The light is coupled in and out of the chip through the tapered regions using lensed fiber, and the width of the tip, typically 100 -- 300 nm, is designed to achieve a good overlap between the fiber mode and the waveguide mode; a narrower tip width leads to a larger MFD, as can be seen from the simulated mode profiles shown in Fig. 2(c). The nominal mode field profile of the LNOI waveguide discussed in this paper is shown in Fig 2(d). To adiabatically match the waveguide width at the taper tip to the nominal waveguide width, the taper length must be sufficiently long. The length of the taper depends on the tip width and the nominal waveguide width -- for smaller tip widths, longer tapers are required. Simulations of the inverse taper length are performed using the eigenmode expansion (EME) solver provided by the commercially available software Lumerical Mode, and are presented in Fig 2(b); it can be seen that for lengths greater than 250 µm, all tapers, regardless of tip width, have negligible insertion loss. We target tip widths ∼ 200 nm in +Z-cut LNOI and X-cut MgO:LNOI to ensure a sufficiently large MFD. It is more challenging to increase the MFD in the thicker 600 nm X-cut MgO:LNOI film than the 500 nm for +Z- 4 FIG. 3. Scanning electron microscope pictures of an etched single mode MgO:LNOI waveguide: (a) waveguide sidewall imaged at a 40◦ tilt; (b) cross-section taken using FIB slicing and SEM imaging; the yellow false coloring highlights the waveguide outline. Optical characterization of fabricated MgO:LNOI mode-matching tapers and single mode waveguide: (c) Fabry-Perot measurements of propagation loss performed on 200 nm inverse taper (the input laser power is 0.5 mW); (d) measured optical power distribution at the output of the untapered waveguide; (e) measured optical power distribution at the output of 200 nm base width taper and (f) < 100 nm base width taper illustrating the increase in MFD, where black lines schematically show the fabricated waveguide dimensions. cut LNOI, so we expect increased coupling loss. Based on the minimum feature size calculations presented in Fig. 1, we expect to be able to achieve taper tips down to 170 nm in +Z-cut LNOI and 210 nm for a 300 nm etch in +Z-cut LNOI and X-cut MgO:LNOI respectively. To compensate for etch bias, the taper tip widths are increased in the mask layout to achieve wider metal lift-off features that then etch down to approximately the target width. We lifted off 400 nm and 500 nm taper tips on X-cut MgO:LNOI resulting in etched base width tips <100 nm and 200 nm. The fabricated MgO:LNOI waveguides side- wall is shown on the Fig. 3(a) taken prior to SiO2 cladding; the sidewall roughness is estimated to be ∼ 2 nm RMS as reported in our previous work [21] and the the etched waveguide profile is cross-sectioned using focused ion beam (FIB) milling and then SEM imaged at a 50◦ angle as shown in Fig. 3(b). The transmission spectrum of the 200 nm taper is shown in the Fig. 3(c) corresponding to a propagation loss of < 0.1 dB/cm based on the Fabry- Perot technique. The laser light is coupled in and out of the chip via a pair of polarization maintaining (PM) lensed fibers. The output transmission of the waveguide is monitored over a 3 × 3 µm window by sweeping the fiber position and recording the transmitted power at each point -- this provides an indication of the waveguide MFD. The coupling loss of a nominal width (untapered) MgO:LNOI waveguide is 10 dB/facet and the power distri- bution at the output is shown at the Fig. 3(d), meanwhile the 200 nm taper achieves a transmission of 6 dB/facet width the power distribution shown in Fig. 3(e). In gen- eral, a smaller taper tip leads to a larger waveguide MFD resulting in higher butt-coupling efficiency; however, the overall transmission through the <100 nm taper drops as the tip becomes increasingly narrow due to the etch bias. Towards the tip of the inverse taper, the MgO:LNOI film has excessively thinned and does not support a guided mode. Figure 3(f) shows that butt-coupling is still possi- ble as the inverse taper eventually becomes wide enough to pick up the leaky mode of the <100 nm tip; however, the structure exhibits a high loss of ∼13 dB/facet. Due to the significantly larger degree of anisotropy of the +Z-cut LNOI (over the X-cut MgO:LNOI), narrower taper tips of width 200 nm and 300 nm were lifted off and etched. The 300 nm lift-off mask resulted in the target ∼200 nm etched tip width as shown in Fig. 4(a). We, again, record the power distribution at the taper out- put in Fig. 4(b) and measure an overall coupling loss of 2.5 dB/facet. The 200 nm taper suffered similar film thin- ning issues as those seen with the narrower MgO:LNOI taper, resulting in a coupling loss of ∼12.5 dB/facet. The quality of the mode matching tapers is strongly affected by the etching properties of the (MgO:)LNOI. 5 FIG. 4. (a) Scanning electron microscope of an etched taper in Z-cut LNOI tip; the false blue-coloring highlights the base and top edges of the taper tip. (b) Measured optical power distribution at the output of 200 nm taper, where black lines schematically show the fabricated waveguide dimensions. We observe improvements in coupling efficiency for both X-cut MgO:LNOI and +Z-cut LNOI at ∼200 nm base taper tips widths. Due to the less anisotropic etching of X-cut MgO:LNOI, the performance of the spot size converters is strongly compromised. Although etch bias may be compensated by increasing mask dimensions, the minimum feature size is limited by the etch angle and depth. IV. CONCLUSION The fiber-waveguide mode matching that can be achieved using the spot size converter is restricted by the etching properties of LN thin film. We have pro- vided a detailed study of many fabrication limitations in LNOI platform. Even nanostructuring with a 75 ◦ waveguide sidewall angle, the minimum achievable fea- ture size hinders the fabrication of ∼200 nm taper tips needed for efficient butt-coupling. We demonstrated de- sign and fabrication of LNOI inverse taper for improved mode matching between mode of an optical fiber and LNOI photonic components. We achieve butt-coupling ef- ficiencies of 2.5 dB/facet and 6 dB/facet for +Z-cut LNOI and X-cut MgO:LNOI respectively, while preserving low propagation losses of ¡0.1 dB/cm in both cases. These results play a critical role in the understanding of LNOI nanostructuring for photonics, and towards the develop- ment of efficient butt-coupling devices -- a major obstacle for LNOI to become competitive photonics platform. FUNDING Australian Research Council Centre for Quantum Com- putation and Communication Technology CE170100012; Australian Research Council Discovery Early Career Re- searcher Award, Project No. DE140101700; RMIT Uni- versity Vice-Chancellors Senior Research Fellowship. ACKNOWLEDGMENTS This work was performed in part at the Melbourne Centre for Nanofabrication in the Victorian Node of the Australian National Fabrication Facility (ANFF) and the Nanolab at Swinburne University of Technology. The authors acknowledge the facilities, and the scientific and technical assistance, of the Australian Microscopy & Mi- croanalysis Research Facility at RMIT University. [1] M. Zhang, C. Wang, R. Cheng, A. Shams-Ansari, and M. Loncar, "Monolithic ultra-high-Q lithium niobate mi- croring resonator," Optica 4, 1536 -- 1537 (2017). [2] A. J. Mercante, S. Shi, P. Yao, L. Xie, R. M. Weikle, and D. W. Prather, "Thin film lithium niobate electro-optic modulator with terahertz operating bandwidth," Opt. Express 26, 14810 -- 14816 (2018). [3] C. Wang, M. Zhang, X. Chen, M. Bertrand, A. Shams- Ansari, S. Chandrasekhar, P. Winzer, and M. Loncar, "Integrated lithium niobate electro-optic modulators oper- ating at CMOS-compatible voltages," Nature 562, 101 -- 104 (2018). [4] Y. He, H. Liang, R. Luo, M. Li, and Q. Lin, "Disper- sion engineered high quality lithium niobate microring resonators," Opt. Express 26, 16315 -- 16322 (2018). [5] P. Gunter and J.-P. Huignard, Photorefractive effects and materials (Springer Berlin Heidelberg, Berlin, Heidelberg, 1988), pp. 7 -- 73. [6] J. Sun and C. Xu, "466 mw green light generation using an- nealed proton-exchanged periodically poled MgO:LiNbO3 ridge waveguides," Opt. Lett. 37, 2028 -- 2030 (2012). [7] C. Wang, C. Langrock, A. Marandi, M. Jankowski, M. Zhang, B. Desiatov, M. M. Fejer, and M. Loncar, "Ultrahigh-efficiency wavelength conversion in nanopho- tonic periodically poled lithium niobate waveguides," Op- tica 5, 1438 -- 1441 (2018). [8] D. Taillaert, F. V. Laere, M. Ayre, W. Bogaerts, D. V. Thourhout, P. Bienstman, and R. Baets, "Grating cou- plers for coupling between optical fibers and nanophotonic waveguides," Japanese Journal of Applied Physics 45, 6071 (2006). [9] R. Orobtchouk, A. Layadi, H. Gualous, D. Pascal, A. Koster, and S. Laval, "High-efficiency light coupling in a submicrometric silicon-on-insulator waveguide," Appl. Opt. 39, 5773 -- 5777 (2000). [10] "High efficiency silicon nitride surface grating couplers," Opt. Express 16, 328 -- 333 (2008). [11] W. D. Sacher, Y. Huang, L. Ding, B. J. F. Taylor, H. Jay- atilleka, G.-Q. Lo, and J. K. S. Poon, "Wide bandwidth and high coupling efficiency Si3N4-on-SOI dual-level grat- ing coupler," Opt. Express 22, 10938 -- 10947 (2014). [12] J. Tremblay, Y.-H. Lin, P.-K. Hsu, M. Malinowski, S. No- vak, P. Qiao, G. F. Camacho-Gonz´alez, C. J. Chang- Hasnain, K. A. Richardson, S. Fathpour, and M. C. Wu, "Large bandwidth silicon nitride spot-size converter for effi- cient supercontinuum coupling to chalcogenide waveguide," 2017 Conference on Lasers and Electro-Optics (CLEO) pp. 1 -- 2 (2017). [13] L. W. Liao, D. Samara-Rubio, M. Morse, A. Liu, D. Hodge, D. Rubin, U. Keil, and T. Franck, "High speed silicon Mach-Zehnder modulator." Optics express 13 8, 3129 -- 35 (2005). 6 [14] M. Papes, P. Cheben, D. Benedikovic, J. H. Schmid, J. Pond, R. Halir, A. O.-M. nux, G. Wanguemert-P´erez, W. N. Ye, D.-X. Xu, S. Janz, M. Dado, and V. Vasinek, "Fiber-chip edge coupler with large mode size for silicon photonic wire waveguides," Opt. Express 24, 5026 -- 5038 (2016). [15] Y. Maegami, R. Takei, E. Omoda, T. Amano, M. Okano, M. Mori, T. Kamei, and Y. Sakakibara, "Spot-size con- verter with a SiO2 spacer layer between tapered Si and SiON waveguides for fiber-to-chip coupling," Opt. Express 23, 21287 -- 21295 (2015). [16] J. Cardenas, C. B. Poitras, K. Luke, L. Luo, P. A. Morton, and M. Lipson, "High coupling efficiency etched facet tapers in silicon waveguides," IEEE Photonics Technology Letters 26, 2380 -- 2382 (2014). [17] Z. Chen, R. Peng, Y. Wang, H. Zhu, and H. Hu, "Grating coupler on lithium niobate thin film waveguide with a metal bottom reflector," Opt. Mater. Express 7, 4010 -- 4017 (2017). [18] J. Jian, P. Xu, H. Chen, M. He, Z. Wu, L. Zhou, L. Liu, C. Yang, and S. Yu, "High-efficiency hybrid amorphous sil- icon grating couplers for sub-micron-sized lithium niobate waveguides," Opt. Express 26, 29651 -- 29658 (2018). [19] Y. Fu, T. Ye, W. Tang, and T. Chu, "Efficient adiabatic silicon-on-insulator waveguide taper," Photon. Res. 2, A41 -- A44 (2014). [20] M. Pu, L. Liu, H. Ou, K. Yvind, and J. M. Hvam, "Ultra- low-loss inverted taper coupler for silicon-on-insulator ridge waveguide," Optics Communications 283, 3678 -- 3682 (2010). [21] I. Krasnokutska, J.-L. J. Tambasco, X. Li, and A. Peruzzo, "Ultra-low loss photonic circuits in lithium niobate on insulator," Opt. Express 26, 897 -- 904 (2018). [22] I. Krasnokutska, J.-L. J. Tambasco, and A. Peruzzo, "Large free spectral range microring resonators in lithium niobate on insulator," arXiv preprint arXiv:1807.06531 (2018).
1903.07998	1	1903	2019-03-19T13:46:12	Near-Field UHF RFID Transponder with a Screen-Printed Graphene Antenna	[ "physics.app-ph" ]	As a method of producing RFID tags, printed graphene provides a low-cost and eco-friendly alternative to the etching of aluminum or copper. The high resistivity of graphene, however, sets a challenge for the antenna design. In practice, it has led to using very large antennas in the UHF RFID far field tags demonstrated before. Using inductive near field as the coupling method between the reader and the tag is an alternative to the radiating far field also at UHF. The read range of such a near field tag is very short, but, on the other hand, the tag is extremely simple and small. In this paper, near field UHF RFID transponders with screen-printed graphene antennas are presented and the effect of the dimensions of the tag and the attachment method of the microchip studied. The attachment of the microchip is an important step of the fabrication process of a tag that has its impact on the final cost of a tag. Of the tags demonstrated, even the smallest one with the outer dimensions of 21 mm * 18 mm and the chip attached with isotropic conductive adhesive (ICA) was readable from a distance of 10 mm with an RF power marginal of 19 dB, which demonstrates that an operational and small graphene-based UHF RFID tag can be fabricated with low-cost industrial processes.	physics.app-ph	physics	Near-Field UHF RFID Transponder with a Screen-Printed Graphene Antenna Kaarle Jaakkola, Henrik Sandberg, Markku Lahti and Vladimir Ermolov Abstract -- As a method of producing RFID tags, printed graphene provides a low-cost and eco-friendly alternative to the etching of aluminum or copper. The high resistivity of graphene, however, sets a challenge for the antenna design. In practice, it has led to using very large antennas in the UHF RFID far field tags demonstrated before. Using inductive near field as the coupling method between the reader and the tag is an alternative to the radiating far field also at UHF. The read range of such a near field tag is very short, but, on the other hand, the tag is extremely simple and small. In this paper, near field UHF RFID transponders with screen- printed graphene antennas are presented and the effect of the dimensions of the tag and the attachment method of the microchip studied. The attachment of the microchip is an important step of the fabrication process of a tag that has its impact on the final cost of a tag. Of the tags demonstrated, even the smallest one with the outer dimensions of 21 mm * 18 mm and the chip attached with isotropic conductive adhesive (ICA) was readable from a distance of 10 mm with an RF power marginal of 19 dB, which demonstrates that an operational and small graphene-based UHF RFID tag can be fabricated with low-cost industrial processes. Index Terms -- antenna, graphene, inductive near field, ink, integration, isotropic conductive adhesive (ICA), near field UHF RFID, printing, RFID I. INTRODUCTION R to fields ADIO-FREQUENCY identification (RFID) is based on transfer data exploiting electromagnetic wirelessly, for the purposes of automatically identifying and tracking tags attached to objects. The technology finds applications in many areas: supply chain management, inventory tracking, contactless payment etc. [1] [2] [3]. The elements of an RFID systems are transponders and a reader, which exchange information wirelessly. A passive RFID transponder consists of a microchip and an antenna. Ultra-high frequency (UHF, 860-960 MHz) RFID transponders with printed graphene antennas and read distances comparable with commercial RFID transponders have been demonstrated recently [4] [5]. Printed graphene-based antennas have several advantages such as low cost, chemical stability, mechanical flexibility, resistance against fatigue and eco-friendliness. For comparison, etching of metal is an environmentally unsafe process and the cost of silver ink strongly depends on the price level of bulk silver metal. The price of silver is very stable and is not expected to decrease in the foreseen future. The cost of graphene ink raw material is very low and the manufacturing process is simple and scalable. The cost can be estimated based on the complexity and scalability of the process and is expected to be much lower than that of commercially available silver inks. UHF RFID transponders operate typically in the far field mode with electromagnetic waves propagating between reader and transponder antennas. This operation mode is optimal for applications for which long read distances (up to 10 meters) are required. However, there are applications in which RFID transponders do not need a long read range, for example tracking of small objects such as gadgets, pharmaceutical packages, bottles, cartridges or battery packs. For such applications, small size of a transponder is expected and the read range of under 0.5 meters can be accepted. Near field low frequency (LF, 125-134 KHz) and high frequency (HF, 13.56 MHz) RFID systems based on inductive coupling between reader and transponder antennas are commonly used in such applications today. It has been demonstrated that the near field coupling can be utilised in the UHF frequency range as well [6]. In this case, magnetic (inductive) coupling or electric (capacitive) coupling can be employed between the antennas of a reader and a transponder. The utilization of near field UHF transponders allows the usage of a single UHF infrastructure for all needed RFID applications instead of the combination of UHF, LF and HF RFID systems. It saves the investments in tracking infrastructures considerably. In this work, we present operational near field UHF RFID transponders with printed graphene antennas. The transponder is based on inductive coupling between the reader and the tag antennas. Because of the stronger inductive coupling at higher frequencies, magnetic UHF transponder has a single-loop antenna, which is much simpler and cheaper compared to LF/HF multi-turn antenna coils with a bridge between the first and the last turn. For mass-produced transponders in disposable applications, direct printing of carbon-based conducting materials is a sustainable and low-cost alternative, and the graphene-based printing ink offers a robust system with a reasonable conductivity. Manuscript submitted Oct 20, 2018; revised Jan 10, 2019; accepted Feb 21, 2019. This work has been part of the Graphene Flagship project, which has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 785219. The partial funding from VTT Technical Research Centre of Finland Ltd is also gratefully acknowledged. The authors are with the VTT Technical Research Centre of Finland Ltd., FI-02044 Espoo, Finland (e-mail: [email protected]). II. DESIGN In practice, the inductive near field transponders are readable from very close distance also with far field antennas, which is due to the magnetic near field component of any practical antenna, but in order to achieve an ideal read range, special near field antennas should be used [6]. Even though referred commonly as "antennas", the near field antennas are not, in fact, antennas, but inductors. Instead of radiating, their purpose is to form a high-frequency magnetic field that is, at least close to the antenna, as uniform as possible. In practice, both the reader and transponder inductors at UHF frequencies are single-turn planar coils and therefore referred to as "coils" in the following. Electromagnetically, the reader coil and the transponder coil form a transformer with the power coupling efficiency Γ. If we assume that the impedance of the reader coil is conjugately matched to the RF output of the reader and that the transponder antenna is reactively matched to a chip, the coupling efficiency can be calculated [7] [8]: (cid:1989) = (cid:2872)(cid:3038)(cid:3278)(cid:3118)(cid:3018)(cid:3293)(cid:3018)(cid:3295)(cid:3267)(cid:3258)(cid:3252)(cid:3267)(cid:3295) (cid:3428)(cid:2870)(cid:3436)(cid:2869)(cid:2878)(cid:3267)(cid:3258)(cid:3252)(cid:3267)(cid:3295)(cid:3440)(cid:2878)(cid:3038)(cid:3278)(cid:3118)(cid:3018)(cid:3293)(cid:3018)(cid:3295)(cid:3432)(cid:3118) , (1) where kc is the coupling factor between the coils, Qr and Qt are the Q values of the reader and the transponder coils, respectively. RIC is the series resistance of the microchip and Rt is the series resistance of the transponder coil. The coupling factor kc can be determined by the mutual inductance between the coils (M) and the self-inductances of the reader and the transponder coils Lr and Lt:(cid:1863)(cid:3030)= (cid:3014)(cid:3493)(cid:3013)(cid:3293)(cid:3013)(cid:3295). By simulations or measurements, kc can also be determined by using the inductance value of the primary coil in two cases: with the secondary coil shorted and the secondary coil left open. Using this method, kc becomes [7]: (2) (cid:1863)(cid:3030)=(cid:3495)1−(cid:3013)(cid:3294)(cid:3013)(cid:3290), (3) where Ls is the inductance of the primary coil with the secondary coil shorted and Lo the inductance of the primary coil with the secondary coil left open. As the transponder is a simple single-turn coil that is small compared to the wavelength, its Q value is equal to its inductive reactance divided by its series resistance: (cid:1843)(cid:3047)=(cid:3025)(cid:3295)(cid:3019)(cid:3295)=(cid:2870)(cid:3095)(cid:3033)(cid:3013)(cid:3295)(cid:3019)(cid:3295) , (4) where Xt is the inductive reactance of the coil, Rt is the serial resistance of the coil and f is frequency. Even though (1) defines the exact coupling efficiency, a more simple formula that is also a term of (1) can be used to define the figure of merit U for an inductively coupled system [8] [9]: 2 (5) (cid:1847)=(cid:1863)(cid:3030)(cid:3493)(cid:1843)(cid:3045)(cid:1843)(cid:3047), Additionally to the coupling factor and the Q values, the third term affecting the coupling efficiency is the impedance match of both the reader coil and the transponder. The reader coil is designed to provide a good match to the 50-Ohm RF output of the reader. The coil on the transponder side is designed to match the reactance of the microchip, but due to its simple structure, the real parts of the impedances are typically not matched. The reactive matching on the transponder side is done by adjusting the circumference of the loop until its reactance has the same but opposite value as that of the microchip. For near field transponders with an aluminum antenna, the real part of the impedance remains lower than that of the microchip, but in the case of materials with higher resistivity, such as graphene, the real part can be significantly higher. This means that the Q value of the transponder as a coil remains low, which reduces the coupling efficiency as defined by (1). The transponder was designed with Ansys HFSS electromagnetic simulation tool. The simulation model shown in Fig. 1 comprises the transponder (gray, on top) and the reader coil. Impinj Matchbox™ was selected as the reader coil [10]. Matchbox contains a single-turn coil that is divided into four segments that are connected to each other with series capacitors. The capacitors shift the phase of the signal so that the structure, the circumference of which is already significant compared to the wavelength, produces a uniform magnetic field without radiating [11]. The simulation model shown in Fig. 1 contains two 50-Ohm lumped ports, of which port 1 is connected to the input of the transponder coil and port 2 is connected to the input of the reader coil. The power coupling F between the reader coil and the RFID microchip can then be calculated using the simulated S parameters of the ports: S11 and S21: ((cid:2869)(cid:2879)(cid:3020)(cid:3117)(cid:3117)(cid:3118))(1−Γ(cid:2870)), (cid:1832)= (cid:3020)(cid:3118)(cid:3117)(cid:3118) (6) where Γ is the reflection coefficient at the RF input of the transponder and is calculated from the input impedance of the transponder coil Zt (= Rt + jXt) and the RF impedance of the microchip ZIC (= RIC + jXIC): In order to evaluate only the power coupling efficiency between the coils (A), omitting the effect of the impedance of the microchip, a simplified form of (6) can be used: Γ=(cid:3027)(cid:3295)(cid:2879)(cid:3027)(cid:3258)(cid:3252)∗ (cid:3027)(cid:3295)(cid:2878)(cid:3027)(cid:3258)(cid:3252), (cid:1827)= (cid:3020)(cid:3118)(cid:3117)(cid:3118) ((cid:2869)(cid:2879)(cid:3020)(cid:3117)(cid:3117)(cid:3118)). (7) (8) curve of the highest-Q case (5 Ohms 7 mm) has a mild form of resonance. 3 TABLE I COILS AT 867 MHZ. PARAMETER VALUES AND THE SIMULATION RESULTS OF THE TRANSPONDER R/□ (ς) 5 10 5 10 w (mm) 4 4 7 7 dx (mm) 12.05 14.3 11.8 12.8 dy (mm) 10.1 10.1 10.1 10.1 Zt (ς) 86 + j129 185+j129 70 + j129 136+j127 1-Φ2 (dB) -3.2 -5.9 -2.6 -4.8 F (dB) -29.4 -34.4 -25.7 -30.2 Fig. 1. Simulation model of the transponder and the reader coil. The design guideline of the inductive near field transponder is based on the round shape that is optimal in terms of generating magnetic flux that protrudes out of the coil. As stated above, the circumference of the transponder coil is determined by the required reactance, which, in turn, is defined by the input impedance of the microchip. The input impedance of the selected microchip, Monza R6 by Impinj, is (13 - j126) Ohms at 867 MHz, making the ideal transponder input impedance (13 + j126) Ohms. The input impedance of Monza R6 includes the effect of the estimated stray capacitance due to adhesive bonding, as specified by Impinj [12]. The width of the conductor affects the resistance of the coil and gets in this case relatively large - several millimeters - due to the low conductivity of graphene. In order to find the right reactive tuning and to study the effect of the conductor width, prototypes with different dimensioning were fabricated. The effect of the square resistance was also studied. Figure 2 shows the critical dimensions of the transponder and the position of the microchip (IC). The reactance of the coil was adjusted to be close to 126 Ohms at 867 MHz by varying the dimension dx. The simulation results of the transponders with different parameter values are listed in Table 1; the input impedance of the transponder Zt calculated from S11, the effect of the impedance mismatch as described by the term 1-Φ2 of (6) and the overall coupling factor F as defined by (6) and (7). Fig. 3. Simulated reflection coefficient of the transponders of Table I as a function of frequency. As the Matchbox reader antenna is, instead of a single inductor, a set of distributed inductors in series, unambiguous single inductance or respective Q value cannot be defined for it. However, the coupling factor kc between the coils can be determined using (3). In the simulations made to determine the values of kc, the transponder coil is used as the measured primary coil and the input of Matchbox is shorted and left open, as explained above. Table II lists the values of kc obtained this way for the same cases as introduced in Table I. Additionally to kc, the changing Q value of the transponder coil has its impact on the coupling efficiency. Therefore, the figure of merit as defined by (5) is studied here as well. As the Q value of Matchbox Qr is undefined but constant, Table II lists the relative figure of merit(cid:1863)(cid:3030)(cid:3493)(cid:1843)(cid:3047). Table II lists also Qt as defined by (4) and A defined by (8). The results of Table II verify that the relative figure of merit gives the same order of coupling efficiency between the cases as A. kc, instead, is mostly affected by the size of the transponder coil. PARAMETER VALUES AND THE SIMULATION RESULTS OF THE TRANSPONDER AND READER COILS AT 867 MHZ. Fig. 2. Critical dimensions of the transponder prototypes. Fig. 3 illustrates the simulated reflection coefficient Φ as defined by (7) for the cases of Table I as a function of frequency. Fig. 3 shows that, even though the reactances are matched, due to the poor match between the real parts of the impedance, the reflection coefficient remains high and only the R/□ (ς) 5 10 5 10 w (mm) 4 4 7 7 TABLE II kc 0.10 0.12 0.13 0.15 Qt 1.50 0.70 1.84 0.93 (cid:1863)(cid:3030)(cid:3493)(cid:1843)(cid:3047) 0.12 0.10 0.18 0.15 A (dB) -26.2 -28.5 -23.1 -25.4 III. FABRICATION Fabrication of the transponder includes two steps: printing of the graphene antenna and integration of the antenna and a commercial RF chip Monza R6. A. Antenna fabrication (PET), polyethylene naphthalate Antenna structures were deposited on flexible substrates by screen printing. Screen printing is a very high volume printing method that can produce thick structured layers with a resolution linewidth of about 100 µm that is sufficient for integration of even small discrete SMD components. Screen printing inks have a high viscosity and solid content, resulting in a thick film after drying. There are many graphene ink providers and the typical achieved conductivity for screen printed graphene-based conductors is 1-10 Ω/(cid:0) with 1-3 screen printing passes. The graphene ink used was prepared by TU/e (Eindhoven University of Technology) and is printed as received. This type of printing paste is based on graphene platelets produced by intercalation and thermal expansion in an optimized process for producing a few-layer thick graphene particles suitable for formulation with a variable binder into a viscous dispersion by gelation [13] [14] [15]. Only the graphene platelets contribute to the conductivity and thus the operation of the antenna, and the electrical conductivity of the printed structures used in this work is used as an input for the design modelling. The ink is compatible with plastic substrates and thus the antenna structures were printed on polyethylene terephthalate (PEN), polyimide (PI) and paper (Lumisilk) substrates. A laboratory scale printing device (EKRA E2) shown in Fig. 4 was used for printing. The device is equipped with a camera, which allows multilayer printing. The screen used has a mesh of 78 stainless steel threads per cm (200 threads per inch), open area of 47% and a theoretical ink transfer volume of 54 ml/m2. Three layers of graphene were printed with a drying step (100 °C, 10 minutes) in a hot air oven after each printing repetition. For high-throughput printing, a single print layer is desired. Furthermore, multilayer printing increases the risk of shorts in particular where small components are attached, due to mismatch between the layers. The lateral rigidity of the substrate also determines how small structures can be printed in multilayers, since thermal relaxation occurs in most of the flexible thin film materials. The printing layout contains a range of dimensional variation for the antenna structures, with a varying gap for the chip integration and varying antenna structure dimensions, the most sensitive being the conctact pads where the chip is attached; the contact dimensions and the gap between the contacts is less than 200 µm and the chip size less than 500 µm. In order to further increase the conductivity of the printed graphene, the PI samples were annealed at up to 320 °C for 36 minutes. Due to the good conductivity results of thermal annealing, the PI samples were used for the final chip integration. For the rest of the substrates that are more temperature-sensitive, photonic annealing could be used [5]. The annealing removes some of the electrically insulating binder from the ink, leaving a higher fraction of conductive graphene in the printed structure. However, after annealing, the 4 layer is porous due to the binder removal and the sample was compressed using a compression roller system (4 passes, 80 bar, 2 m/min) [15]. The conductivity increases significantly with each process step; the triple-layer printed structure has a sheet resistance of 135 Ω/(cid:0) while the annealed and compressed structure has a sheet resistance of < 10 Ω/(cid:0), which is a suitable conductivity level for the antenna modelling, thus justifying the extra processing steps. The fabricated prototypes of graphene near field UHF RFID antennas with different dimensioning are shown in Fig. 5. Fig. 4. EKRA S2 screen printing device. Fig. 5. Printed antenna structures (right) and the attached bare die chip (left). The SEM picture shows the integration of the chip using isotropic conductive adhesive. B. Antenna and RF chip integration A challenge of combining any graphene-based elements, in our case antennas, with conventional electronics is about forming an electrical connection between two dissimilar materials: printed graphene and metal. In former publications about implementing RFID transponders with graphene-based antennas, this has been typically solved by using an aluminum- or copper-based strap that is glued with conductive silver-based adhesive to the graphene antenna. [4] [5]. The problem with such an implementation is that the process needs more steps and individual parts than the conventional way of producing an RFID transponder. Consequently, it does not bring the whole potential of cost-savings and eco-friendliness of a graphene- based transponder into action. Therefore, in this work, the goal was to develop a process based on conductive adhesives for attaching the RFID microchip into a graphene-based antenna. For the bonding process, rough and non-uniform surfaces of printed graphene is a challenge. High and anisotropic resistivity of graphene also increases the contact resistance as the contact area of the microchip is small. The assembly of chips for RFID transponders in high volumes is usually based on the use of adhesives: anisotropic conductive adhesive (ACA) or Isotropic conductive adhesives (ICA). The ACA contains small amounts of conductive sphere particles dispersed in a polymer matrix. Therefore, the ACA is not conductive in the horizontal plane, eliminating the risk of short-circuits. The processing is started by dispensing the adhesive onto the chip area. Then the chip is aligned with the pads of the substrate and pressed down at a certain temperature. The temperature depends on the substrate and the adhesive. The curing time is also short, typically ~10 s. The ACA suits well to mass production and fine-pitch applications. The thermocompression of chips is typically done in a batch process, which further speeds up the assembly. ICAs, made up of a composition of polymer resin and conductive silver fillers, provide electrical conductivity in X, Y and Z directions when the silver flakes pack together due to the shrinkage of the ICA after being cured [16]. ICA allows higher conductivity due to higher filling load of highly conductive Ag flake-shaped particles, but they request wider pitching for chips. ICAs are preferable for applications in which pitches are not critical but higher conductivity of the contact is required. for is than lower temperature The electrical resistance of the adhesives is in a range of 4 µΩ * m. For comparison, the electrical resistance of typical solder materials, such as PbSn or SnAgCu with different compositions, is in the range of ~0.1 µΩ * m. Although the electrical resistance is higher in the case of adhesives, they provide other benefits over solder materials. For example, the processing that of solders. Additionally, the maximum curing temperature can be tuned, which enables the use of variety of substrate materials. For example, the melting temperatures of eutectic PbSn and SnAgCu are 183 and 217-220 °C, respectively. In practice, the maximum temperature during soldering is typically 30-50 °C higher than the melting temperature. The final properties of adhesives are achieved by curing (i.e. by cross-linking of polymer chains). Different adhesives are cured in different ways, although heat is the most common one. For example, the curing temperature for H20E can vary from 80 to 175 °C. The duration is then varied from 3 h to 45 s, respectively. Due to these benefits, conductive adhesive was selected for this case. Today, chip assembly with ACA on PET substrate is the standard technology for mass production of RFID tags and the use of ACA on e.g. paper substrate has also been successfully demonstrated [16] [17]. Monza R6 chip has a rectangular shape with the outer dimensions of 464.1 μm x 442 μm. The chip has two metal pads for antenna connection. The size of the pads is 166 μm x 422 μm with a spacing of 112 μm between them. The bonding pads 5 are large enough, allowing the use of ICA instead of ACA for better electrical contact between a chip and a graphene antenna. The isotropically conductive adhesive Epotek H20-E was manually dispensed onto the pads of the chip. The chip was then flip-chipped onto the graphene antenna with Finetech Fineplacer flip-chip bonder. The antenna printed on polyimide (PI) was kept at the room temperature. The chip was aligned to the proper area of the antenna using a camera alignment system. The alignment accuracy with the bonder is in the range of a few micrometers, which was adequate in this case. The assembled transponder was then placed into a drying oven at the temperature of 100 °C for 30 min. The adhesive changes into electrically conductive during this drying step. The obtained connection between the chip and the graphene antenna is mechanically strong. A photograph of the assembled module is shown in Fig. 6. Fig. 6. Monza R6 chip bonded by ICA glue to the printed graphene antenna. In order to evaluate the feasibility and the electrical characteristics of the different ways of attaching the microchip to the antenna, prototypes were made with three attachment methods: an aluminum strap glued with silver lacquer, the microchip attached directly with silver lacquer and the microchip attached with ICA glue. IV. RESULTS AND DISCUSSION The fabricated tag prototypes were measured using Voyantic Tagformance™ UHF RFID characterization system with an Impinj Matchbox near field reader antenna [10] [11]. The measurement setup inside the anechoic cabinet of the system is shown in Fig. 7. Even though echoes are not that critical for a near-field system, the Faraday-cage shielding from external signals is important. On the Impinj Matchbox antenna (black) there is a 10 mm spacer of Styrofoam (blue) to maintain a constant distance between the measurement antenna and the tag under test. The measurement setup is similar to the simulation model shown in Fig. 1. In the measurement, the activation level of the transponder is measured as a function of frequency [18] [19]. Although in an inductive near field system the maximum reader power is not restricted similarly to radiating systems, the maximum power specification of the reader antenna sets the limit to +30 dBm (1 W) of conducted power [10]. Therefore, in the following, the sensitivity of the prototypes is evaluated in terms of the power marginal that is the difference between the needed activation level and the maximum allowed +30 dBm. In order to compare the simulation and measurement results, the power needs to be defined similarly for the two. This is achieved by defining the 50-Ohm input port of the antenna as the reference point of the RF power. In the simulations, the power marginal AM is then defined in dB form: (cid:1827)(cid:3014)((cid:1856)(cid:1828))=(cid:1832)+(cid:1842)(cid:3047)(cid:3051) (cid:3040)(cid:3028)(cid:3051)−(cid:1842)(cid:3046)(cid:3032)(cid:3041)(cid:3046) (cid:3010)(cid:3004), (9) (10) where F is the power coupling coefficient as defined by (6) and (7), Ptx max is the maximum RF power (+30 dBm) and Psens IC is the wakeup (read) sensitivity of the microchip, -20 dBm for Monza R6 [12]. The corresponding power marginal can be calculated from the measurement results: (cid:1827)(cid:3014)((cid:1856)(cid:1828))=(cid:1842)(cid:3047)(cid:3051) (cid:3040)(cid:3028)(cid:3051)+(cid:1838)(cid:3030)(cid:3036)(cid:3045)(cid:3030)+(cid:1838)(cid:3030)(cid:3028)(cid:3029)(cid:3039)(cid:3032) −(cid:1842)(cid:3047)(cid:3051), where Lcirc is the attenuation of the circulator between the RF output of the Tagformance device and the Matchbox antenna and Lcable is the attenuation of the coaxial cables and connectors between the circulator and the Matchbox antenna. Ptx is the RF output power of the Tagformance device that is recorded as a function of frequency during the measurement. The value of Lcirc, as given by the specification of the device, is -0.2 dB and that of Lcable, obtained by a measurement, is -1.4 dB. The different prototypes fabricated and measured are shown in Fig. 8. Two conductor widths, 7 mm and 4 mm, and three methods of attaching the microchip to the coil: strap with silver lacquer, chip directly attached with the same lacquer and chip directly attached with ICA, have been studied. The prototype on the top left has a 7 mm wide conductor and a strap-attached microchip (series "7 mm strap"). The prototype on the top right has a 7 mm wide conductor and a microchip attached directly with silver lacquer ("7 mm silver"). The one on the bottom left has a 4 mm wide conductor and a microchip attached with a strap ("4 mm strap"). The one on the bottom right has a 4 mm wide conductor and a chip attached with ICA ("4 mm ICA"). The outer dimensions of the tags in mm are 27 * 24.5 ("7 mm strap"), 29.5 * 24.5 ("7 mm silver"), 21 * 18.5 ("4 mm strap") and 21 * 18.1 ("4 mm ICA"). Fig. 9 shows the measured power marginal AM as a function of frequency for the transponders of Fig. 8 at the distance of 10 mm from the surface of Matchbox reader coil. The operation band specified for Matchbox is 865…956 MHz, which should be taken into account when analyzing the results. The measurement result of a commercial aluminum-based near field UHF RFID transponder, "UPM Trap" is also included for comparison [20]. The dashed lines represent the simulation results for the 4 mm wide conductor with two values of square resistance of graphene: 5 Ohms and 10 Ohms. The challenge of the electromagnetic modelling of graphene is about printed graphene being an anisotropic material, the conductivity of which the horizontal direction. The DC is higher in 6 measurement made on the surface of the printed graphene has showed square resistance values slightly lower than 10 Ohms, but as the material defined in the simulation model is homogeneous and isotropic, the two values have been studied by simulations. Another thing that is difficult to model is the contact resistance that depends on the attachment method of the microchip. A strap with silver lacquer forms a big contact area (2 * 3.5 mm2) with a lower series resistance, whereas the direct gluing with ICA forces the RF current to flow through very small area (166 μm x 422 μm) of high-resistivity graphene, resulting in high contact resistance. In terms of the contact resistance, the direct attachment with silver lacquer (case "7 mm silver") lies between these two. The anisotropic nature of graphene relates also to the contact resistance, since, depending on the attachment method, the portion of the current flowing in the vertical direction varies. Geometrically, the simulation model is closest to the ICA bonded case. Based on the maximum variation in a repeatability test with ten measurements for each case, with the transponder repositioned between the measurements, the repeatability is +/- 0.5 dB over the whole frequency range. According to measurements, moving the tag horizontally 5 mm from the middle produces a maximum attenuation of 1.0 dB (+/- 0.5 dB) within the 865…956 MHz operation band of Matchbox. Fig. 7. Measurement setup with an Impinj Matchbox near field reader antenna, Styrofoam spacer and a tag under test inside an anechoic cabinet. Fig. 8. Near field UHF RFID tag prototypes on graphene: with a 7 mm wide conductor and the microchip attached by a strap of aluminum (top left), with a 7 mm wide conductor and the microchip attached with silver lacquer (top right), with a 4 mm wide conductor and the microchip attached with a strap of aluminum (bottom left) and with a 4 mm wide conductor and the microchip attached with isotropic conductive adhesive (ICA) (bottom right). 7 area of a near field transponder as well as the absolute size remain moderate. Even though the read range of a near field transponder is clearly shorter than that of a far field tag, a short and well-defined read area is a benefit in many applications. The combination of a short read range and a small, low-cost transponder makes the graphene-based near field transponder an ideal solution for item-level tagging of small objects. V. CONCLUSION Near field UHF RFID transponders with a screen-printed graphene antenna were designed, fabricated and tested. All of the transponder prototypes fabricated were proven operational with a fair power marginal. The fabrication method, particularly the bonding of a chip to a printed graphene antenna is based on already-existing industrial processes, which makes the concept viable also commercially. The approach demonstrated here encompasses an industrially relevant printing technique, post- processing methods compatible with roll-to-roll manufacturing paradigms and temperature sensitive flexible substrates. ACKNOWLEDGMENT Prof. Heiner Friedrich and Dr Artur Pinto are acknowledged for providing the graphene ink and for discussions regarding the printing and the post processing parameters. The authors would like to thank Impinj for providing the Monza R6 microchips. The research leading to these results has received funding from the EC. The authors would like to thank the Directorate-General for Science, Research and Development of the European Commission for support of the research. Authors would like to acknowledge the Graphene Flagship for the support of the research. REFERENCES [1] [2] [3] K. Finkenzeller, "RFID Handbook: Radio-Frequency Identification Fundamentals and Applications," 2nd ed. New York, NY, USA, Wiley, 2004. R. Want, "RFID explained: A primer on radio frequency identification technologies", Synthesis Lectures on Mobile and Pervasive Computing, vol. 1, Jan. 2006, pp. 1-94 J. Landt, "The History of RFID", IEEE Potentials, vol. 24, no. 4, Oct.- Nov. 2005, pp. 8 - 11, [4] M. Akbari, M. Waqas, A. Khan, M. Hasani, T. Björninen, L. Sydänheimo, L. Ukkonen, "Fabrication and characterization of graphene antenna for low-cost and environmentally friendly RFID tags," IEEE Antennas Wireless Propag. Lett. vol. 15, pp. 1569‒1572, (2016). [5] K. Arapov, K. Jaakkola, V. Ermolov, G. Bex, E. Rubingh, S. Haque, H. Sandberg, R. Abbel, G. de With, H. Friedrich, "Graphene Screen-printed Radio-frequency Identification Devices on Flexible Substrates", Physica status solidi (RRL) - Rapid Research Letters, vol. 10, issue 11, pp. 812- 818 (2016) [6] P. Nikitin, K. V. S. Rao and S. Lazar "An Overview of Near Field UHF RFID," 2007 IEEE International Conference on RFID, Grapevine, TX, USA, March 26-28, (2007). [7] R. M. Duarte and G. K. Felic, "Analysis of the Coupling Coefficient in Inductive Energy Transfer Systems," Active and Passive Electronic Components, Vol. 2014, Article ID 951624. [8] M. Kesler, "Highly Resonant Wireless Power Transfer: Safe, Efficient, and over Distance," WiTricity Corporation, 2017, accessed on Jan 6, 2019. http://witricity.com/wp- content/uploads/2016/12/White_Paper_20161218.pdf Available: [Online]. Fig. 9. Power marginal of the transponders of Fig. 8 and a commercial near field tag "UPM Trap" as a function of frequency at the distance of 10 mm from the surface of Matchbox reader coil. One can conclude from Fig. 9 that there is a sufficient power marginal available for all of the tags measured. Comparing the graphs "7 mm strap" and "4 mm strap" shows that the wide conductor gives better sensitivity due to lower resistive losses, as expected. The comparison of "7 mm strap" with "7 mm silver" shows the difference between the attachment methods, as does the comparison between "4 mm strap" and "4 mm ICA". The results confirm that, in terms of the increasing contact resistance, the order of the attachment methods is strap, silver lacquer and ICA. However, there is still a good power marginal of 19 dB at the frequency of 867 MHz with the ICA-bonded prototype. The curves of the magnetic field intensity as a function of distance for Matchbox predict this marginal to equal the read range of about 40 mm from the surface of Matchbox, whereas for a typical patch antenna the equivalent read range is about 90 mm [10]. Of the prototypes with a 4 mm wide conductor, the one with the ICA contact is between the lines of the simulation results, showing that if the effective square resistance would be adjusted to match the measurement result, the value would be between 5 Ohms and 10 Ohms, but closer to 5 Ohms. The measurement curves bending differently to the simulated ones close to the 800 MHz end may be related to the limited bandwidth of Matchbox. The commercial transponder, "UPM Trap" shows larger marginal that is better sensitivity than the graphene prototypes, even though it uses a less sensitive microchip (Monza 3). The shape of its curve also shows higher Q value than any of the graphene transponders, which was expected due to its aluminum conductor. Additionally to a simple and straightforward structure, the size of an antenna is an important parameter when aiming at low-cost production. A UHF near field transponder combines these two properties. The size benefit of a near field UHF transponder is especially true for graphene-based transponders, for which implementing an efficient far field antenna requires a very large conductor area. [4] [5] This is due to the low conductivity of graphene, which also affects the performance and size of a near field transponder. However, the conductor [9] N. Nitta, N. Li, K. Fujimori and Y. Sugimoto, "A Q-enhancement method for resonant inductive coupling wireless power transfer system," 2017 IEEE Wireless Power Transfer Conference (WPTC), 2017. [10] Impinj Matchbox™ Near Field UHF RFID Antenna, accessed on Oct 14, https://support.impinj.com/hc/en- 2018. us/articles/202755658-MatchBox-Antenna-Datasheet Available: [Online]. [11] Y. S. Ong, X. Qing, C. K. Goh, Z. N. Chen, "A segmented loop antenna for UHF near-field RFID", 2010 IEEE Antennas and Propagation So- ciety doi: 10.1109/APS.2010.5561018 International Symposium, [12] Datasheet of Impinj Monza R6 UHF RFID Microchip, accessed on Oct 11-17 2010, July 14, 2018. [Online]. Available: https://support.impinj.com/hc/article_attachments/115001963950/Monz a%20R6%20Tag%20Chip%20Datasheet%20R5%2020170901.pdf [13] K. Arapov, A. Goryachev, G. de With, and H. Friedrich, "A simple and flexible route to large-area conductive transparent graphene thin-films," Synthetic Metals, vol. 201, pp. 67 -- 75, 2015. [14] K. Arapov, E. Rubingh, R. Abbel, J. Laven, G. de With, and H. Friedrich, "Conductive screen printing inks by gelation of graphene dispersions," Advanced Functional Materials, vol. 26, no. 4, pp. 586 -- 593, 2016. [15] K. Arapov, G. Bex, R. Hendriks, E. Rubingh, R. Abbel, G. de With, and H. Friedrich, "Conductivity Enhancement of Binder-Based Graphene Inks by Photonic Annealing and Subsequent Compression Rolling," Advanced Engineering Materials, vol. 18, no. 7, pp. 1234 -- 1239, 2016. [16] H. Chu, B. An, F. Wu and Y. Wu, "RFID Tag Packaging with Anisotropically Conductive Adhesive," Electronic Packaging Technology, 2006. ICEPT '06. 7th International Conference on , pp. 1-4, 26-29 Aug. 2006. [17] J. S. Rasul, "Chip on paper technology utilizing anisotropically conductive adhesive for smart label applications," Microelectronics Reliability, Vol. 44, No. 1, Jan. 2004, pp. 135-140. [18] Tagformance by Voyantic, accessed on Oct 14, 2018. [Online]. Available: http://voyantic.com/tagformance [19] P. Nikitin, K. V. S. Rao, S. Lam. UHF RFID Tag Characterization: Overview and State-of-the-Art, accessed on Oct 14, 2018. [Online]. Available: https://pdfs.semanticscholar.org/4c92/ad48e34cf7ef6a11e7652819cb6c b293d9e2.pdf [20] UPM Trap near field UHF RFID tag, datasheet available via Yumpu.com, accessed on Oct 14, 2018. [Online]. Available: https://www.yumpu.com/en/document/view/21832041/upm-trap-rfid Technology (Tech.) degree Kaarle Jaakkola received the Master of Science in electrical engineering from the Helsinki University of (currently Aalto University), Espoo, Finland, in 2003. Since 2000 he has been working at the VTT Technical Research Centre of Finland, currently as a Senior Scientist. His research interests and expertise include RFID systems, electronics, wireless and applied sensors, antennas, electromagnetic modelling and RF electronics. He has e.g. developed RF parts for RFID systems and designed antennas for both scientific use and commercial products. Antennas designed by him can be found in several commercial RFID transponders. He has authored and coauthored 20 peer-reviewed journal and conference articles and he holds 10 patents as an inventor or a co-inventor. 8 Dr. Henrik Sandberg is a principal scientist at VTT's centre for Printable and Hybrid Functionalities. His work is currently focused on printed electronic devices and circuits, polymer device physics, printing technology and printing ink formulation, as well as heterogeneous and monolithic integration for flexible electronics. He coordinates work on printed graphene based materials and hybrid integration, specifically targeting flexible applications such as wearable devices. He specializes in device and circuit development as well as the development of analog printing compatible processing techniques from the lab to the R2R pilot scale and on related ink development, in particular on the topics of thin film polymer transistors and circuits, organic photovoltaics and graphene and other 2D material applications. Markku Lahti received his M.Sc. and D.Sc. degrees in Electrical Engineering from University of Oulu in 1993 and 2008, respectively. He has been with VTT Technical Research Centre of Finland, Ltd. since 2001, where he is currently working as a Senior Scientist. His research interests are related to the manufacturing of low-temperature co-fired ceramic (LTCC) substrates, and module-level integration and packaging of components with ceramic and polymer platforms. He has been working in several EU and ESA projects, and is author or co-author of over 20 peer-reviewed publications and over 70 conference papers. Vladimir Ermolov graduated with honours in 1981 and received the Ph.D in 1986 from the Moscow Engineering Physics University (MEPhI). Since 1981, he had been as Senior Research Associate with the laboratory of Dielectric Devices (MEPhI). He worked many times as a visiting researcher in the Department of Physics, Helsinki University and Fraunhofer Institute of Non- destructive testing, Germany. He had been a principal research scientist with Nokia research center from 1998 till 2011, where he worked as a project manager for many projects in areas of MEMS, nanotechnology, radiotechnology, mass memory technology and sensors. He was involved in commercialization of several technologies. Since 2011 he has been with VTT Technical Research Center of Finland as a principal research scientist, where he had made project management and research in areas of radiotechnology, MEMS, nanotechnology, and technology commercialization. He is the author and co-author of 54 in referred journals and the holder and co-holder of 47 patents and patents pending. Vladimir Ermolov is a recipient of the 47th European Microwave Conference Microwave Prize.
1908.06914	1	1908	2019-08-19T16:27:23	Low temperature growth and optical properties of {\alpha}-Ga2O3 deposited on sapphire by plasma enhanced atomic layer deposition	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Plasma enhanced atomic layer deposition was used to deposit thin films of Ga2O3 on to c-plane sapphire substrates using triethylgallium and O2 plasma. The influence of substrate temperature and plasma processing parameters on the resultant crystallinity and optical properties of the Ga2O3 films were investigated. The deposition temperature was found to have a significant effect on the film crystallinity. At temperatures below 200{\deg}C amorphous Ga2O3 films were deposited. Between 250{\deg}C and 350{\deg}C the films became predominantly {\alpha}-Ga2O3. Above 350{\deg}C the deposited films showed a mixture of {\alpha}-Ga2O3 and {\epsilon}-Ga2O3 phases. Plasma power and O2 flow rate were observed to have less influence over the resultant phases present in the films. However, both parameters could be tuned to alter the strain of the film. Ultraviolet transmittance measurements on the Ga2O3 films showed that the bandgaps ranges from 5.0 eV to 5.2 eV with the largest bandgap of 5.2 eV occurring for the {\alpha}-Ga2O3 phase deposited at 250{\deg}C.	physics.app-ph	physics	Low temperature growth and optical properties of α-Ga2O3 deposited on sapphire by plasma enhanced atomic layer deposition. J.W. Roberts (1), P.R. Chalker (1), B. Ding (2), R. A. Oliver (2), J.T. Gibbon (3), L.A.H. Jones (3), V.R. Dhanak (3), L. J. Phillips (3), J. D. Major (3), F. C-P. Massabuau (2) (1) University of Liverpool Centre for Materials and Structures (2) University of Cambridge Department of Materials Science and Metallurgy (3) University of Liverpool Stevenson Institute for Renewable Energy Plasma enhanced atomic layer deposition was used to deposit thin films of Ga2O3 on to c-plane sapphire substrates using triethylgallium and O2 plasma. The influence of substrate temperature and plasma processing parameters on the resultant crystallinity and optical properties of the Ga2O3 films were investigated. The deposition temperature was found to have a significant effect on the film crystallinity. At temperatures below 200°C amorphous Ga2O3 films were deposited. Between 250°C and 350°C the films became predominantly α-Ga2O3. Above 350°C the deposited films showed a mixture of α-Ga2O3 and ε-Ga2O3 phases. Plasma power and O2 flow rate were observed to have less influence over the resultant phases present in the films. However, both parameters could be tuned to alter the strain of the film. Ultraviolet transmittance measurements on the Ga2O3 films showed that the bandgaps ranges from 5.0 eV to 5.2 eV with the largest bandgap of 5.2 eV occurring for the α- Ga2O3 phase deposited at 250°C. INTRODUCTION Recent interest has been focused on gallium oxide (Ga2O3) because of its wide bandgap, which has been reported to be in the range from ca. 4.5 to 5.3 eV 1 -- 6, making it a promising candidate for deep ultraviolet (UV) sensing applications 7,8. Ga2O3 can exist as different phases (α, β ,γ ,ε ,κ ,δ ), some of which are shared with In2O3 and Al2O3 9 -- 11, which have bandgaps of approximately 3.7 eV and 8.8 eV respectively 12,13. Consequently, growth of In- or Al- doped Ga2O3 opens a viable route to bandgap tuning over a wide range of energies 14. α- and β-Ga2O3 have also been identified as next generation candidates for the current carrying channels in power transistors due to their higher critical field strength compared to GaN and SiC 7. To date, monoclinic β-Ga2O3 has been the most extensively studied phase, due to its stability films formation polycrystalline compared to the other polymorphs and the ease with which it can be grown 10. High quality crystals, and nanostructures of β-Ga2O3, with bandgaps of ca. 4.9 eV 5, have been grown directly from 7, by chemical vapour melt deposition (CVD) 15,16, pulsed laser deposition (PLD) 3,17 and metal-organic chemical vapour deposition (MOCVD) 4,18,19, to name just a few examples. α-Ga2O3, a metastable hexagonal phase of Ga2O3, has a wider bandgap than β-Ga2O3, around 5.1 eV - 5.3 eV 5,8 and is structurally analogous to α-Al2O3 (corundum)11. However, unlike the monoclinic β-phase, which is thermodynamically favourable, there are few processes which can reliably form α-Ga2O3, and all of them require temperatures above ca. 430°C 5 or high pressures of several GPa 20. These requirements have limited the use of this polymorph in temperature and pressure sensitive processes and devices. Currently, the most prevalent method used for the growth of α-Ga2O3 is the "mist-CVD" technique that has been used to directly grow α-Ga2O3 between ca. 430°C and ca. 470°C 5,21,22. Other synthesis methods include RF sputtering 23, precipitation 24, halide vapor phase epitaxy (HVPE) 25and molecular beam epitaxy (MBE) 8. Most Ga2O3 atomic layer deposition (ALD) papers using a variety of deposition temperatures, chemical precursors and substrates report deposition of amorphous material 26 -- 32. However, in a recent study we reported that α-Ga2O3 could be deposited by plasma enhanced ALD (PEALD) onto sapphire (α-Al2O3) substrates 33. Here, we investigate the impact of several growth parameters (substrate temperature, plasma power and O2 flow) on Ga2O3 thin films deposited by PEALD on sapphire. GROWTH METHODS to Deposition of Ga2O3 films was carried out using an Oxford Instruments OpAL PEALD reactor with the baffle plate removed from above the chamber, giving a direct line-of-sight from the remote plasma the substrates. The depositions utilised an inductively coupled plasma system located approximately 60 cm above the deposition chamber. Adduct grade triethylgallium (TEGa) from Epichem was used as the gallium source and dry O2 from BOC was used as the oxygen source. Argon from BOC was used for chamber purges and as the precursor carrier gas. The TEGa source was maintained at 30°C, with line temperatures into the reactor chamber held at 80°C and 90°C. For the lowest temperature deposition (120°C substrate) the chamber walls were held at 125°C, while the chamber walls were set at 150°C for all other growths. For each growth run, several c-plane sapphire samples with a miscut of 0.25±0.10° towards (1120) were positioned centrally in the reaction chamber along with Si(100) pieces. 500 PEALD cycles were used for the growth of each film. Initial ALD growth parameters were taken from Shih et al 31 in which they found a saturative growth regime and linear growth per cycle using 0.1 s TEGa dose times and 5 s O2 plasma exposure. For each experimental set, the following conditions were kept constant between growths: 0.1 s TEGa dose, 5 s TEGa purge, 5 s O2 plasma duration, 5 s O2 plasma purge. 100 sccm Ar was used as a carrier gas during the TEGa dose and as the purge gas to remove unreacted precursors from the chamber during the purge steps. Table 1 lists experimental conditions that were changed between growths. The base pressure in the chamber (with no process gases flowing) was ca. 10 mTorr. During the deposition processes the chamber pressure varied between ca. 80 mTorr (during the plasma steps) and 160 mTorr (during the TEGa dose). Sample analysis was conducted on the as-grown films with no further annealing steps. Table 1: Summary of the growth conditions for each of the experimental sets. Experimental O2 plasma flow O2 plasma power (W) Substrate temperature (°C) 120, 150, 200, 250, 300, 350, 400, 450 250 250 set Temperature O2 flow O2 power (sccm) 20 10, 20, 40, 60, 100 300 300 20 25, 50, 100, 200, 300 ANALYSIS METHODS MM-16 After deposition, the ellipsometric parameters, psi () and delta (), were measured over a range of 430-850 nm for each sample using a Horiba-Yvon spectroscopic ellipsometer. Film thicknesses and refractive indices were computed using a Ga2O3 Cauchy model based on optical values reported by Rebien et al 34. Whilst  and  measurements yield accurate values of refractive index (the error bar is smaller than the datapoints in Figures 1(A), 6(A) and 7(A)), the film thickness obtained using this approach are estimated to be within ~10% accuracy, owing to the model itself. Measurements were carried out on both c-plane sapphire and Si(100) substrates. Characterisation by X-ray diffraction (XRD) was carried out using a PANalytical Empyrean diffractometer with a Cu Kα1 X-ray source (λ = 1.5405974 Å 35), a hybrid monochromator, and either a two-bounce Ge crystal analyzer (for 2θ-ω and ω scans) or a PIXcel detector (for reciprocal space maps (RSMs)). The strain state of the α-Ga2O3 films was obtained by measuring the peak position from RSMs recorded around the symmetric 0006 and asymmetric 10110 reflections. The surface topography of the samples was investigated by atomic force microscopy (AFM), using a Bruker Dimension Icon Pro microscope operated in peak force tapping mode. A SCANASYST-AIR tip, with a nominal radius of 2 nm, was used. Transmission electron microscopy (TEM) was used to observe the structure of the films in cross-section. The samples were prepared using standard sample preparation that entails mechanical polishing followed by Ar+ ion milling at 5 kV and 0.1-1 kV. Annular dark field scanning TEM (ADF-STEM) was conducted in a Tecnai Osiris operated at 200 kV. Scanning electron diffraction (SED), which involves the acquisition of an electron diffraction pattern at each probe position 36 as a convergent electron probe is scanned across the sample, was performed using a Tecnai F20 operated at 200 kV and retrofitted with a NanoMegas Digistar system. This system enables the simultaneous scan and acquisition of diffraction patterns with an external optical charge coupled device imaging the phosphor viewing screen of the microscope. The SED data was inspected and analysed using for crystallographic electron microscopy 37,38. A series of diffraction contrast images were formed by plotting the intensity within a selected subset of pixels in the diffraction pattern as a function of probe position to form so-called 'virtual dark-field' images. a Python library The optical transmittance and reflectance of the films grown on sapphire were measured using a Shimadzu SolidSpec-3700 dual beam UV-Vis spectrophotometer (Shimadzu Corp., Kyoto, Japan) and the bandgap 𝐸𝑔 was extracted using the Tauc method. RESULTS AND DISCUSSION Effect of growth temperature substrate temperature, Figure 1(A) shows the variation in thickness of the Ga2O3 films grown on sapphire with increasing as determined from fitting of the spectroscopic ellipsometry data. Average and standard error values across three samples used in each growth run are shown. At 120°C and 150oC, the Ga2O3 films thickness and refractive index are ca. 30 nm and ca. 1.87, respectively. XRD (Figure 1(B)) shows that these films are amorphous, with no visible peaks between 2θ values of 37° to 44° (apart from the 0006 reflection from the α-Al2O3 substrate). At 200°C an increase in growth rate is observed, which coincides with the appearance of a partially strained α-Ga2O3 0006 peak in the XRD scan. Due to the broad assumptions used in the Ga2O3 Cauchy ellipsometry model, this could be attributed to either a real increase in growth rate due to the change in crystallinity Figure 1: (A) Ga2O3 film thicknesses and refractive indices (at 632 nm) against substrate temperature for growth on c-plane sapphire. (B) 2θ-ω scans recorded around the symmetric α-Al2O3 0006 reflection showing the effect of growth temperature on the crystal structure of the Ga2O3 films. In inset, strain state of α-Ga2O3 with regard to the sapphire substrate at different growth temperatures. (see XRD in Figure 1(B)), a deviation in the refractive index beyond that shown in Figure 1(A), or a change in the surface morphology which the model fails to account for. Between 200°C and 300°C the film thickness decreases to ca. 31 nm (corresponding to a growth rate of ca. 0.6 Å/cycle) and the α-Ga2O3 film approaches a more strain-free state (Figure 1(B) inset). We also note in Figure 1(B) the presence of fringes on the α-Ga2O3 0006 peak with increasing growth temperature. Such fringes are often interpreted as improved film quality or greater uniformity of film thickness. However, as can be seen in the AFM results (Figure 2), the samples exhibit a similar roughness between 200 and 350°C, indicating a comparable thickness uniformity between these samples. Therefore, the fringes in the XRD scan are probably more indicative of the crystal quality. At 350°C and above, XRD shows a decrease in the relative intensity of the α-Ga2O3 peak and the appearance of peaks at lower 2θ values, which may be assigned to one or several of the following reflections: ε- Ga2O3 0004, ε-Ga2O3 2111 or β-Ga2O3 311 (the low intensity of the peaks does not allow us to discern the phase with certainty). The thickness of the films also appears to increase (Figure 1(A)), reaching a maximum growth rate of 0.84 Å/cycle at 450°C. However, considering the AFM (Figure 2(C)) and TEM (Figure 3(C)) film data of the sample grown at 400°C, which shows a rough surface with the film thickness varying between 28 and 33 nm, it is possible that the ellipsometry model's assumptions break down for the samples deposited at 350oC and above. The XRD peaks seen in Figure 1(B) also correlate well with the refractive indices (at 632.8 nm) calculated from the ellipsometry data, with all crystalline films showing values between 1.9 and 2.0 26,31,1,39,40 and the relaxed α-Ga2O3 films (250°C to 350°C) showing higher refractive indices between 1.95 and 2 in agreement with the literature 41. Figure 2 shows 500 nm x 500 nm AFM images of Ga2O3 films grown at 120°C, 250°C and 400°C, selected to represent each crystal phase observed, and a summary plot of the root mean squared (RMS) roughness for this sample set. At 150°C and below, the topography of the samples seems to indicate that the surface is amorphous and the RMS roughness values below 0.5 nm indicate a very uniform, conformal deposition. Between 200°C and 350°C, corresponding to the α phase, the films show a slight increase in roughness (between 0.5 nm and 0.7 nm) and the presence of grain- like structure with lateral size of around 10 nm (accurate measurement is impeded by the radius of the AFM tip). Figure 2: 500 nm x 500 nm AFM images of the Ga2O3 films grown at (A) 150°C (amorphous Ga2O3), (B) grown at 250°C (α-Ga2O3), and (C) 400°C (mixed phases Ga2O3). (D) Plot of RMS roughness versus growth temperature. In our previous study of 130 nm thick Ga2O3 films, we observed by SED that the ca. 70 nm of Ga2O3 material adjacent to the substrate was dominantly α phase columns with a diameter of 2-23 nm33. Since the films shown here are approximately 30 nm thick, we expect that the grain-like structure observed here by AFM corresponds to the termination of the α phase columns. Above 350°C, the surface roughness increases by an order of magnitude up to a maximum of 3 nm at 450°C. The surface of the films at these temperatures features larger, less densely packed equiaxed features approximately 30 nm across and 5 nm high for the sample grown at 400°C, and 45 nm across and 8 nm high for the sample grown at 450°C. Given that these larger features appear in the samples where multiple phases cohabit (according to the XRD results (Figure 1(B))), it is reasonable to suggest that these large features correspond to grains of a different phase than the remainder of the film. Figure 3 shows cross-sectional TEM images of Ga2O3 films grown on sapphire at 150oC, 250oC and 400oC, selected to represent each crystal phase observed. As can be seen in Figure 3(A), the film grown at 150 °C shows a uniform contrast and a smooth upper surface. The absence of features in this image support the XRD results, that is, that the Ga2O3 film is amorphous. The image also agrees with the AFM data which shows an average surface roughness of below 0.5 nm and the ca. 30 nm thickness determined by ellipsometry. Figure 3: Cross-sectional ADF-STEM images of the Ga2O3 films grown at (A) 150°C (amorphous Ga2O3), (B) grown at 250°C (α- Ga2O3), and (C) 400°C (mixed phases Ga2O3). The Ga2O3 film grown at 250°C is shown in Figure 3(B). Distinct columns of Ga2O3 can be seen running for the most part through the full thickness of the films. From our XRD results (Figure 1(B)), the film is dominantly α-Ga2O3. The structure of this film, that is, α phase columns, agrees well with our earlier report 33. It can be seen that in comparison with Figure 3(A), there is a slight increase in surface roughness related to the grain boundaries where α-Ga2O3 columns touch. This is also in agreement with the AFM data (Figure 2). The film thickness (29 nm approximately) also agrees with the ellipsometry measurements (Figure 1(A)). Figure 3(C) shows the Ga2O3 film grown at 400°C. The structure of this film is distinctly different from the first two images with large inverted triangular features visible. The space between these triangles is filled with material with a different contrast. Given that the whole film is expected to have the same average atomic number, contrast here arises from change of crystallinity in the film (crystal structure, orientation, etc.). The dimensions of the triangular features when they meet the free surface of the film agree very well with the dimensions of the equiaxed features we observed by AFM (Figure 2(C)). In light of the XRD and ADF-STEM results, we can safely say that these triangular features and the space between them relate to different phases of Ga2O3. Phase identification was obtained using SED, as shown in Figure 4(A). In this virtual dark field image, we can clearly attribute the inverted triangular features to α phase crystallites typical diffraction pattern recorded in such region is shown in Figure 4(B)) while the film between such features is mainly in the ε phase (a typical diffraction pattern recorded in such region is shown in Figure 4(C)). It should be noted that while the α-phase triangular be monocrystalline, the ε-phase material seem to be made from several small crystallites -- which could explain why the corresponding XRD peak (around 38-39o in Figure 1(B)) is broad and weak. SED also confirms that the α-Ga2O3 features are grown along (0001) -- which is the same orientation as the α-Al2O3 substrate. The ε-Ga2O3 crystallites are also dominantly oriented along (0001) although in places we observed signs of misorientation. features (a appear to Figure 4: (A) Virtual dark-field image obtained by SED of the Ga2O3 film grown at 400°C, seen along the [1120] zone axis for α-Al2O3. In inset, corresponding bright field image. The blue colouring is obtained from the α phase -- a typical diffraction pattern is shown in (B), and red colouring represents the ε phase -- a typical diffraction pattern is shown in (C). Figure 5(A) shows the % transmittance against wavelength for the Ga2O3 films grown on sapphire. One data set for each morphology is shown for clarity. In the visible range, all films show similar transmittance at around 80%. functional to density The fundamental electronic bandgap of β- Ga2O3 (located on the I-L line in reciprocal- space) is indirect with a magnitude of 4.84 eV, according theories simulations and experimental results obtained by other groups. However, several studies have also shown that the direct bandgap (located at Γ) is only 0.04 eV larger, with a bandgap of 4.88 eV 42 -- 44. The majority of papers available in which UV-Vis spectroscopy has been used for the study of Ga2O3 films use the direct bandgap assumption when calculating Tauc plots 1,4,26,28,30,32,45, i.e: 𝛼ℎ𝑣 ∝ (ℎ𝑣 − 𝐸𝑔) 1 2⁄ where 𝛼 is the optical absorption coefficient, ℎ𝑣 is the photon energy and 𝐸𝑔 is the optical bandgap. Thus a plot of (𝛼ℎ𝑣)2 against ℎ𝑣 can be used to extract 𝐸𝑔. Studies on the band structure of α-Ga2O3 are less readily available than those on β-Ga2O3, presumably due to the difficulty in synthesising high quality single crystals of the material. Choi et al. simulated doping of various materials into α-Ga2O3 and found that for the undoped oxide, the electronic bandgap was indirect with a value of 4.70 eV, whilst the nearest direct bandgap was 0.21 eV higher at 4.91 eV 14. To conform with the existing literature, in the present work, we also use the direct bandgap assumption. The inset in Figure 5(A) shows a Tauc plot for the same films assuming direct bandgaps 14,42 -- 44. Figure 5(B) shows the calculated optical bandgaps of the deposited films against deposition temperature. We observe an optical bandgap of ca. 5.05 eV for the amorphous Ga2O3 films (150-200 °C), while the α-Ga2O3 films (250-350 °C) exhibit a greater optical bandgap, nearer 5.15-5.20 eV. The bandgap of the mixed phase films (400-450 °C) is slightly lower than that of the purely α-Ga2O3 films, presumably due to the contribution of the ε phase. These bandgaps are in agreement with those found by other groups using a variety of different growth methods 5,19,21,40,46. We observe that α-Ga2O3 exhibits the widest bandgap of all polymorphs, in agreement with the literature 19. Effect of O2 flow and plasma power Figure 5: (A) UV-vis transmittance for selected Ga2O3 films grown on sapphire (Inset -- Tauc plot) and (B) Optical bandgaps for Ga2O3 films calculated from (A), assuming direct bandgap for each phase. Figure 6: (A) Spectroscopic ellipsometry model results add (B) XRD 2θ-ω scans of the samples grown under varying O2 plasma gas flow at 300 W and 250°C substrate. In inset of (B), strain state of α- Ga2O3 with regard to the sapphire substrate at different O2 flow. Figure 7: (A) Spectroscopic ellipsometry model results add (B) XRD 2θ-ω scans of the samples grown under varying plasma power at 20 sccm O2 flow and 250°C substrate. In inset of (B), strain state of α- Ga2O3 with regard to the sapphire substrate at different plasma power. Figures 6 and 7 show the spectroscopic ellipsometry and XRD results for varying O2 flow and plasma power at a fixed growth temperature of 250°C for films grown on sapphire. The XRD results show that these two parameters have very limited impact on the crystal structure of the films -- all films are in the α phase. The main noticeable difference was in terms of relaxation of the film (see insets Figures 6(B),7(B)). For the O2 flow sample set, we observe that the lowest O2 flows result in higher refractive indices and lower growth rates (Figure 6(A)) while also showing the highest strain relaxation close to 100% (Figure 6(B) inset). In comparison, films grown with higher O2 flows show relaxation at around 50% -- at the moment the mechanisms leading to strain relaxation in these films remain unclear. A higher refractive index and lower growth rate both imply that a denser film is being grown under the low O2 flow conditions. AFM was also carried out on the films (not shown) but the change in these deposition parameters had no observable effect on the surface morphology or roughness -- all the samples look similar to that presented in Figure 3(B). CONCLUSIONS PEALD can be used to deposit α-Ga2O3 onto c- plane sapphire substrates at low temperatures and without the need for further annealing processes. A growth window for the α phase has been determined between 250°C and 350°C, with temperatures below the growth window resulting in amorphous Ga2O3 and temperatures above the growth window yielding mixed α and ε phases films. UV-vis transmittance measurements of α-Ga2O3 films on sapphire showed that the α-Ga2O3 film exhibits the highest optical bandgaps (up to 5.2 eV) amongst the samples studied. For α-Ga2O3 grown at 250°C, variation in the O2 flow rate and O2 plasma power growth parameters allow for further tuning of the strain relaxation of the α-Ga2O3 films with respect to the sapphire substrate. ACKNOWLEDGEMENTS This project is funded by the Engineering and Physical Sciences Research Council (Grants No. EP/P00945X/1, EP/M010589/1, EP/N014057/1 and EP/K014471/1). REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. Ghose, S. et al. Growth and characterization of β-Ga2O3thin films by molecular beam epitaxy for deep- UV photodetectors. J. Appl. Phys. (2017). doi:10.1063/1.4985855 Alema, F. et al. Fast growth rate of epitaxial β -- Ga2O3by close coupled showerhead MOCVD. J. Cryst. Growth (2017). doi:10.1016/j.jcrysgro.2017.06.001 Wakabayashi, R. et al. Oxygen-radical- assisted pulsed-laser deposition of β- Ga2O3and β-(AlxGa1-x)2O3films. J. Cryst. Growth (2015). doi:10.1016/j.jcrysgro.2015.05.005 Kong, L., Ma, J., Luan, C. & Zhu, Z. Structural and optical properties of Ga2O3:In films deposited on MgO (1 0 0) substrates by MOCVD. J. Solid State Chem. (2011). doi:10.1016/j.jssc.2011.05.048 Shinohara, D. & Fujita, S. Heteroepitaxy of corundum-structured α-Ga2O3thin films on α- Al2O3substrates by ultrasonic mist chemical vapor deposition. Jpn. J. Appl. Phys. (2008). doi:10.1143/JJAP.47.7311 Pearton, S. J. et al. A review of Ga2O3materials, processing, and devices. Applied Physics Reviews (2018). doi:10.1063/1.5006941 Baldini, M., Galazka, Z. & Wagner, G. Recent progress in the growth of β- Ga2O3for power electronics applications. Materials Science in Semiconductor Processing (2018). doi:10.1016/j.mssp.2017.10.040 Chen, X. et al. Solar-Blind Photodetector with High Avalanche Gains and Bias-Tunable Detecting Functionality Based on Metastable Phase α-Ga2O3/ZnO Isotype Heterostructures. ACS Appl. Mater. Interfaces (2017). doi:10.1021/acsami.7b09812 9. 10. 11. 12. 13. 14. 15. 16. 17. Roy, R., Hill, V. G. & Osborn, E. F. Polymorphism of Ga2O3and the System Ga2O3−H2O. J. Am. Chem. Soc. (1952). doi:10.1021/ja01123a039 Yoshioka, S., Hayashi, H. & Kuwabara, A. Structures and energetics of Ga2O3 polymorphs - Abstract - Journal of Physics: Condensed Matter - IOPscience. J. Phys. … (2007). Playford, H. Y., Hannon, A. C., Barney, E. R. & Walton, R. I. Structures of uncharacterised polymorphs of gallium oxide from total neutron diffraction. Chem. - A Eur. J. (2013). doi:10.1002/chem.201203359 Fujita, S. & Kaneko, K. Epitaxial growth of corundum-structured wide band gap III-oxide semiconductor thin films. J. Cryst. Growth (2014). doi:10.1016/j.jcrysgro.2014.02.032 French, R. H. Electronic Band Structure of Al2O3, with Comparison to Alon and AIN. J. Am. Ceram. Soc. (1990). doi:10.1111/j.1151- 2916.1990.tb06541.x Choi, M. & Son, J. Doping-induced bandgap tuning of α-Ga2O3for ultraviolet lighting. Curr. Appl. Phys. (2017). doi:10.1016/j.cap.2017.02.019 Terasako, T., Ichinotani, H. & Yagi, M. Growth of β-gallium oxide films and nanostructures by atmospheric- pressure CVD using gallium and water as source materials. Phys. Status Solidi Curr. Top. Solid State Phys. (2015). doi:10.1002/pssc.201510012 Zhong, M., Wei, Z., Meng, X., Wu, F. & Li, J. High-performance single crystalline UV photodetectors of β- Ga2O3. J. Alloys Compd. (2015). doi:10.1016/j.jallcom.2014.09.070 Orita, M., Hiramatsu, H., Ohta, H., Hirano, M. & Hosono, H. Preparation of highly conductive, deep ultraviolet transparent β-Ga2O3 thin film at low deposition temperatures. in Thin Solid Films (2002). doi:10.1016/S0040- 18. 19. 6090(02)00202-X Chen, Y. et al. Effect of growth pressure on the characteristics of β- Ga2O3films grown on GaAs (100) substrates by MOCVD method. Appl. Surf. Sci. (2015). doi:10.1016/j.apsusc.2014.11.074 Sun, H. et al. HCl Flow-Induced Phase Change of α-, β-, and ε-Ga 2 O 3 Films Grown by MOCVD. Cryst. Growth Des. (2018). doi:10.1021/acs.cgd.7b01791 20. Machon, D., McMillan, P. F., Xu, B. & Dong, J. High-pressure study of the beta-to-alpha transition in Ga2O3. Phys. Rev. B - Condens. Matter Mater. Phys. (2006). doi:10.1103/PhysRevB.73.094125 Lee, S. D., Akaiwa, K. & Fujita, S. Thermal stability of single crystalline alpha gallium oxide films on sapphire substrates. Phys. Status Solidi Curr. Top. Solid State Phys. (2013). doi:10.1002/pssc.201300259 Kaneko, K., Suzuki, K., Ito, Y. & Fujita, S. Growth characteristics of corundum- structured α-(AlxGa1- x)2O3/Ga2O3heterostructures on sapphire substrates. J. Cryst. Growth (2016). doi:10.1016/j.jcrysgro.2015.12.013 Akazawa, H. Formation of various phases of gallium oxide films depending on substrate planes and deposition gases. Vacuum (2016). doi:10.1016/j.vacuum.2015.10.009 Sakata, Y. et al. Photocatalytic properties of gallium oxides prepared by precipitation methods toward the overall splitting of H2O. J. Catal. (2014). doi:10.1016/j.jcat.2013.06.025 Yao, Y. et al. Growth and characterization of α-, β-, and ϵ-phases of Ga2O3 using MOCVD and HVPE techniques. Mater. Res. Lett. (2018). doi:10.1080/21663831.2018.1443978 21. 22. 23. 24. 25. 27. 28. 29. 30. 31. 32. 33. layer deposition of Ga 2O 3 films using trimethylgallium and ozone. Chem. Mater. (2012). doi:10.1021/cm300712x Donmez, I., Ozgit-Akgun, C. & Biyikli, N. Low temperature deposition of Ga 2 O 3 thin films using trimethylgallium and oxygen plasma. J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. (2013). doi:10.1116/1.4758782 Choi, D. W., Chung, K. B. & Park, J. S. Low temperature Ga2O3atomic layer deposition using gallium tri- isopropoxide and water. in Thin Solid Films (2013). doi:10.1016/j.tsf.2013.03.066 Altuntas, H., Donmez, I., Ozgit-Akgun, C. & Biyikli, N. Effect of postdeposition annealing on the electrical properties of β-Ga 2 O 3 thin films grown on p- Si by plasma-enhanced atomic layer deposition. J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. (2014). doi:10.1116/1.4875935 O'Donoghue, R. et al. Low temperature growth of gallium oxide thin films via plasma enhanced atomic layer deposition. Dalt. Trans. (2017). doi:10.1039/c7dt03427j Shih, H. Y. et al. Atomic Layer Deposition of Gallium Oxide Films as Gate Dielectrics in AlGaN/GaN Metal -- Oxide -- Semiconductor High-Electron- Mobility Transistors. Nanoscale Res. Lett. (2016). doi:10.1186/s11671-016- 1448-z Ramachandran, R. K. et al. Plasma enhanced atomic layer deposition of Ga2O3thin films. J. Mater. Chem. A (2014). doi:10.1039/c4ta05007j Roberts, J. W. et al. α-Ga2O3grown by low temperature atomic layer deposition on sapphire. J. Cryst. Growth (2018). doi:10.1016/j.jcrysgro.2018.02.014 26. Comstock, D. J. & Elam, J. W. Atomic 34. Rebien, M., Henrion, W., Hong, M., Mannaerts, J. P. & Fleischer, M. Optical properties of gallium oxide thin films. Appl. Phys. Lett. (2002). doi:10.1063/1.1491613 35. Deslattes, R. D. et al. X-rays. in International Tables for Crystallography (2006) (ed. Prince, E.) 200 -- 212 (Kluwer Academic Publishers, 2006). 36. Moeck, P. et al. High spatial resolution semi-automatic crystallite orientation and phase mapping of nanocrystals in transmission electron microscopes. in Crystal Research and Technology (2011). doi:10.1002/crat.201000676 45. 46. structure of In2O3 and Ga2O3. Sensors (2014). doi:10.1063/1.1427137 Kumar, S. S. et al. Structure, morphology, and optical properties of amorphous and nanocrystalline gallium oxide thin films. J. Phys. Chem. C (2013). doi:10.1021/jp311300e Ghose, S. et al. Structural and optical properties of β-Ga 2 O 3 thin films grown by plasma-assisted molecular beam epitaxy. J. Vac. Sci. Technol. B, Nanotechnol. Microelectron. Mater. Process. Meas. Phenom. (2016). doi:10.1116/1.4942045 37. 38. Pyxem, version 0.7, www.github.com/pyxem/pyxem. Peña, F. de la et al. Hyperspy/Hyperspy: Hyperspy 1.3. hyperspy/hyperspy: HyperSpy 1.3 (2017). doi:10.5281/ZENODO.583693 39. Mu, W. et al. High quality crystal growth and anisotropic physical characterization of β-Ga2O3single crystals grown by EFG method. J. Alloys Compd. (2017). doi:10.1016/j.jallcom.2017.04.185 40. 41. 42. Choi, K. H. & Kang, H. C. Structural and optical evolution of Ga2O3/glass thin films deposited by radio frequency magnetron sputtering. Mater. Lett. (2014). doi:10.1016/j.matlet.2014.03.038 He, H. et al. First-principles study of the structural, electronic, and optical properties of Ga2 O3 in its monoclinic and hexagonal phases. Phys. Rev. B - Condens. Matter Mater. Phys. (2006). doi:10.1103/PhysRevB.74.195123 Peelaers, H. & Van de Walle, C. G. Brillouin zone and band structure of β- Ga2O3. Phys. Status Solidi Basic Res. (2015). doi:10.1002/pssb.201451551 43. Mohamed, M. et al. The electronic structure of β-Ga2O3. Appl. Phys. Lett. (2010). doi:10.1063/1.3521255 44. Chen, C. et al. Experimental electronic
1804.05896	1	1804	2018-03-23T21:06:34	High-throughput high-resolution multi-scale three dimensional laser printing	[ "physics.app-ph", "physics.optics" ]	Throughput and resolution are both of critical importance for modern manufacturing, however, they are usually contradictory with each other. Here, a high-throughput high-resolution three dimensional (3D) printing is demonstrated by incorporating a simultaneous spatiotemporal focusing (SSTF) scheme in two-photon polymerization (TPP). Remarkably, the SSTF can ensure generation of a spherical focal spot of which the size depends linearly on the laser power. Thus, the resolution can be continuously adjusted from sub-10 {\mu}m to ~40 {\mu}m by only increasing the laser power. A multi-scale 3D structure is fabricated using this technique, for which the regions of coarse and fine feature sizes are produced at high and low resolutions, respectively.	physics.app-ph	physics	High-throughput high-resolution multi-scale three dimensional laser printing Yuanxin Tan,1,2 Wei Chu,1, * Peng Wang,1,3 Wenbo Li,1,2,4 Jia Qi,1,2 Jian Xu,5,6 Ya Cheng1,5,6,7, * *Corresponding Author: E-mail: [email protected] [email protected] 1State Key Laboratory of High Field Laser Physics, Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Shanghai 201800, China 2University of Chinese Academy of Sciences, Beijing 100049, China 3School of Physics Science and Engineering, Tongji University, Shanghai 200092, China 4School of Physical Science and Technology, Shanghai Tech University, Shanghai 200031, China. 5State Key Laboratory of Precision Spectroscopy, School of Physics and Materials Science, East China Normal University, Shanghai 200062, China 6XXL - The Extreme Optoelectromechanics Laboratory, School of Physics and Materials Science, East China Normal University, Shanghai 200241, China 7Collaborative Innovation Center of Extreme Optics, Shanxi University, Taiyuan, Shanxi 030006, China Abstract: Throughput and resolution are both of critical importance for modern manufacturing, however, they are usually contradictory with each other. Here, a high- throughput high-resolution three dimensional (3D) printing is demonstrated by incorporating a simultaneous spatiotemporal focusing (SSTF) scheme in two-photon polymerization (TPP). Remarkably, the SSTF can ensure generation of a spherical focal spot of which the size depends linearly on the laser power. Thus, the resolution can be continuously adjusted from sub-10 μm to ~40 μm by only increasing the laser power. A multi-scale 3D structure is fabricated using this technique, for which the regions of coarse and fine feature sizes are produced at high and low resolutions, respectively. 1 1. Introduction Femtosecond laser two-photon polymerization has enabled 3D laser nanofabrication at resolutions as high as tens of nanometers.[1-7] The high resolutions achieved by TPP result from a combined contribution of the use of high numerical aperture (NA) objectives and existence of a threshold intensity in initiating the photopolymerization process. The choice of high NA objectives of short working distances limits the heights of end products of TPP, which has recently been circumvented by replacing the conventional focusing scheme with a simultaneous spatiotemporal focusing (SSTF) scheme.[8,9] The general concept of SSTF was proposed by Zhu et al for demonstrating wide-field 3D bioimaging based on two-photon fluorescence excitation.[10] However, for femtosecond laser 3D direct writing, the optical layout of SSTF must be changed for creating a tight spherical focal spot rather than a thin focal plane as shown by He et al.[11] The working mechanism of generation of a tight SSTF spot is briefly given as follow. In the SSTF, the femtosecond pulse first passes through a pair of gratings, thus the different frequency components in the pulse can be spatially separated from each other before they incident upon the back aperture of the focal lens. This reduces the local bandwidth of the pulse and leads to a dramatic extension of the pulse width. Temporal focusing, or focusing in the time domain, occurs after the focal lens during propagation of the pulse toward the geometric focal point because all the frequency components tend to recombine at the focus. In this way, the initial transform-limited pulse transiently restores itself at the focus while being stretched either before or after the focus, enabling shortening of the focal depth in terms of the peak intensity.[12,13] Here, we show that when combining the SSTF with TPP, the nonlinear threshold effect in TPP uniquely allows for producing multi-scale 3D structures with a variable fabrication 2 resolution in an alignment-free manner. This is benefitted from the fact that the size of laser affected volume with the STTF spot linearly depends on the power of the laser beam, making it extremely easy to dynamically tune the fabrication resolution during the continuous writing process. This provides an efficient means for boosting the throughput in fabricating multi- scale 3D microstructures, as otherwise the entire structures must be fabricated at the highest resolution (i.e., corresponding to a low fabrication throughput) as required by the regions of finest features. 2. Experiment Section Our experimental setup is schematically illustrated in Figure 1. The positively pre-chirped femtosecond laser pulses delivered by a commercial laser amplifier (Libra, Coherent, Inc.) were of a bandwidth of ~27 nm and a pulse duration of ~ 200 ps. The pulse energy was tuned by a half waveplate combined with a polarizer. After reducing the beam width by a telescope system, the spectral components of the femtosecond laser pulses were first spatially separated by passing through a single-pass grating compressor which consists of two 1500 grooves/mm gratings blazed for the incident angle of 53°. The distance between the grating pair was adjusted to ~730 mm to compensate for the initial positive temporal dispersion of the beam, and the spatial chirp of the spatially separated frequency components was held constant at ~0.75 nm/mm after the grating pair. Then the spatially chirped pulses were focused into a SU8 resin by an objective lens (Leica 2 ×, NA =0.35). In combination with the threshold effect of the TPP, SSTF have the capability to generate the spherical voxel within the photosensitive resins. 3. Dependence of fabrication resolution on the laser power 3 To demonstrate the 3D isotropic fabrication resolution offered by the SSTF-TPP, we first fabricated two arrays of rods, one of which oriented along X direction while the other along Y direction, both embedded in a cube of SU-8 resin as shown in Figure 2a. The cube was intentionally fabricated to mechanically stabilize the logpile structures for the subsequent examination. The inset of Figure 2a shows the scanning electron microscopy (SEM) of the fabricated structures. All the rods were written at a fixed scan speed of 400 μm/s, whilst the laser power was varied to write the rods of different cross sectional sizes in the two arrays. It should be noted that the effective TPP printing speed depends on various parameters, such as the repetition rate of the femtosecond laser, the laser power and the resolution chosen for the fabrication. In our work, the major limit of the scanning speed was the low repetition rate of our femtosecond laser which is only 1 kHz. With this low repetition rate, the scan speed was kept below 400 μm/s to ensure sufficient overlap between consecutive laser shots. However, because of the enlarged focal spot with SSTF, the writing speed was 40 times higher than that in previous work which also employed a 1 kHz femtosecond laser.[14] Figure 2b-2g shows the optical micrographs of cross sections of the rods fabricated at different femtosecond laser powers. The laser powers below were measured before the grating pair. Considering a total loss of 35% due to the two reflections at the gratings, the laser powers would drop to 65% of the measured powers at the back aperture of the objective lens. In Figure 2b, the diameters of the rods written with a laser power of 1.5 mW were measured to be ~9.4 μm for both X and Y scan directions, providing a direct evidence on the isotropic fabrication resolution. Then, we gradually raised the laser power to 2 mW, 3 mW and 4 mW in writing each pair of the rods oriented perpendicular to each other, i.e., one in X direction and the other in Y direction. The cross sectional micrographs of the rod pairs fabricated at the increasing laser power are respectively presented in Figure 2c to 2e. For each pair of the rods, 4 a nearly circular cross sectional shape has been achieved, resulting in various 3D isotropic resolutions of ~13.5 μm, ~20.5 μm and ~27 μm. When the laser power further increased to 5 mW and 6 mW, the cross sections of the fabricated rods slightly deviated from the perfect circular shape, i.e., the rods exhibit diamond-shaped cross sections as shown in Figure 2f and 2g. At the laser powers of 5 mW and 6 mW, the respective lateral sizes of the rods were measured to be 31.6 μm and 37.7 μm, while the longitudinal sizes were measured to be slightly larger, i.e., 34.5 μm and 44 μm, respectively. To quantitatively determine the dependence of the fabrication resolution on the power of femtosecond laser, the lateral and longitudinal cross sectional sizes of the rods are plotted as functions of laser power in Figure 2h. We observe that in the range of 1.5 mW and 6 mW, the lateral and longitudinal resolutions are well balanced, and the resolutions in both lateral and longitudinal directions show nearly linear dependence on the power of femtosecond laser. The mechanism behind the power dependence of the fabrication resolution is the threshold effect in the interaction of femtosecond laser pulses with transparent materials.[5] 4. Application of the SSTF-TPP for 3D laser printing Freestanding 3D structures are ideal model structures to examine the alignment-free tunability in the fabrication resolution provided by the SSTF-TPP in a straightforward way. Therefore, we fabricated a series of coils in SU-8 resin at different laser powers, as shown in Figure 3. In the writing process as illustrated in Figure 3a, we scanned the SSTF focal spot along a circular trajectory in XY plane while the motion stage was translated along Z direction. Figure 3b to 3d show the optical micrographs of the coils fabricated at various laser power of 3 mW, 4 mW, and 5 mW, respectively. The coils all have a same diameter of 200 μm in XY plane, but they are of different pitches in Z direction. The pitch of the coils in Figure 3b and 3C is 250 5 μm, while the pitch of the coil in Figure 3d is 150 μm. The coils show smooth surfaces and circular cross sections by examining at different angles of view. This is consistent with the results in Figure 2 and provides another unambiguous evidence on the isotropic 3D fabrication resolution achieved with SSTF-TTP. Furthermore, we fabricated a multi-scale 3D structures in a single continuous writing process. Figure 4a shows the design of the structure, which is composed of a boat of a height of 8 mm, an elephant of a height of 3 mm standing on the deck of the boat, and a dog of a height of 1 mm sitting on top of the cabin. Because of the different feature sizes of the boat, the elephant, and the dog, these parts can be fabricated at different fabrication resolutions. In the experiment, we chose the contour scan method to fabricate the 3D structure for its relatively high fabrication efficiency. The boat was sliced into 200 layers with a layer height of 40 μm and fabricated at the laser power of 6 mW. Meanwhile, the elephant was sliced into 167 layers with a layer height of 18 μm and fabricated at the laser power of 3 mW, and the dog was sliced into 125 layers with a layer height of 8 μm and fabricated at the laser power of 1 mW. The parameters chosen above give rise to the fabrication resolutions of 40 μm, 18 μm, and 8 μm in writing of the boat, the elephant, and the dog, respectively. Figure 4b, 4c, and 4d present the digital-camera-captured images of a top view of the fabricated structure, one side view with the face of the elephant, and the opposite side view with the back of the elephant, respectively. The quality of the fabricated sculptures of the dog and elephant was further examined with a scanning electron microscope (SEM), as shown in Figure 5. Figure 5a presents the overview of both the dog and elephant, showing their different feature sizes. Figure 5b shows that the details in the fabricated elephant such as the ear, the tusk, the trunk, the tail, and even the 6 backbone can all be clearly seen. Likewise, Figure 5c shows the details in the fabricated dog, which are of a high quality as well thanks to the 8 μm resolution chosen for writing it. 5. Conclusion To conclude, we demonstrate fabrication of multi-scale 3D structures using the SSTF-based TPP. We show that the fabrication resolution, which is isotropic in all the three dimensions in space, can be dynamically tuned only by varying the laser power. This is important for real- world applications in which it is preferable to avoid the cumbersome and time-consuming processes of precise realignment of the optics for resetting the fabrication resolution. In the current experiment which employed a Ti:Sapphire laser of a low repetition rate of 1 kHz, it took around 10 hrs to complete the whole fabrication process for the structure shown in Fig. 4(a). However, the fabrication duration can be significantly shortened by replacing the 1 kHz laser source with high repetition rate, high average power laser systems. Moreover, the demonstrated tuning range of the fabrication resolution achievable by merely using the threshold effect is between sub-10 μm and ~40 μm. Meanwhile, the tuning range can be easily expanded to reach either higher or lower resolutions by modifying the focusing paramters, i.e., synergetically adjusting the width of the incident femtosecond laser beam and the distance between the two gratings, or replacing the focal lenses of different NAs. The wide tuning range of fabrication resolution renders the SSTF-TPP attractive for multi-scale 3D printing with unprecedented flexibility. Thus, the sophisticated capability of balancing the fabrication throughput and the fabrication resolution in the SSTF-TPP laser printing will benefit a number of applications such as bioengineering, microelectromechanics (MEMS), photonics, microfluidics, and manufacturing, to name a few. 7 Acknowledgements This work is supported by the National Basic Research Program of China (Grant 2014CB921303), National Natural Science Foundation of China (Grants 11734009, 11674340, 61327902, 61590934, 61675220, 61505231), Strategic Priority Research Program of the Chinese Academy of Sciences (Grant XDB16030300), Key Research Program of Frontier Sciences, Chinese Academy of Sciences (Grant QYZDJ-SSW-SLH010), Project of Shanghai Committee of Science and Technology (Grant 17JC1400400) and Shanghai Rising- Star Program (Grant 17QA1404600). 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)) Keywords: (femtosecond laser two photon polymerization, multiscale 3D printing, simultaneous spatiotemporal focusing) 8 References [1] M. Malinauskas, M. Farsari, A. Piskarskas, S. Juodkazis, Phys. Rep. 2013, 533(1), 1- 31. [2] [3] Y. L. Zhang, Q. D. Chen, H. Xia, H. B. Sun, Nano Today 2010, 5(5), 435-448. F. Korte, J. Serbin, J. Koch, A. Egbert, C. Fallnich, A. Ostendorf, B. N. Chichkov, Appl. Phys. A 2003, 77(2), 229-235. [4] M. Farsari, B. N. Chichkov, Nat. Photonics 2009, 3(8): 450. [5] [6] [7] [8] R. R. Gattass, E. Mazur, Nat. Photonics 2008, 2(4): 219. K. Sugioka, Y. Cheng, Light Sci. Appl. 2014, 3(4), e149. S. Kawata, H. B. Sun, T. Tanaka, K. Takada, Nature 2001, 412(6848), 697-698. Y. Tan, Z. Wang, W. Chu, Y. Liao, L. Qiao, and Y. Cheng, Opt. Mater. Express 2016, 6(12), 3787-3793. [9] W. Chu, Y. Tan, P. Wang, J. Xu, W. Li, J. Qi, Y. Cheng, Adv. Mater. Technol 2018. [10] G. Zhu, J. Van Howe, M. Durst, W. Zipfel, C. Xu, Opt. Express 2005, 13(6), 2153- 2159. [11] F. He, H. Xu, Y. Cheng, J. Ni, H. Xiong, Z. Xu, K. Sugioka, and K. Midorikawa, Opt. Lett. 2010, 35(7), 1106–1108. [12] C. G. Durfee, M. Greco, E. Block, D. Vitek, and J. A. Squier, Opt. Express 2012, 20(13), 14244–14259. [13] C. Jing, Z. Wang, and Y. Cheng, Appl. Sci. 2016, 6(12), 428. [14] S. Juodkazis, V. Mizeikis, K. Seet, M. Miwa, H. Misawa, Nanotechnol. 2005, 16(6), 846. 9 Figure 1. Schematic of the experimental setup. HWP, half waveplate; PBS, polarizing beam- splitter; L1, L2, telescope system; M, reflective mirrors; G1, G2, gratings; DM, dichroic mirror. Inset: The evolution of the pulse duration when the light propagated in focal volume. 10 Figure 2. Control of 3D isotropic resolution. (a) Illustration of the designed model structure for measuring the fabrication resolutions at different femtosecond laser powers. Inset: SEM images of the fabricated rods with laser power of 3 mW (upper) and 5 mW (lower). (b-g): the end- sectional-view optical micrographs in XZ (left column) and YZ (right column) with the laser power of 1.5 mW, 2 mW, 3 mW, 4 mW, 5 mW, and 6 mW, respectively. (h): the lateral and longitudinal diameters of the rods plotted as functions of laser power. 11 Figure 3. (a) Illustration of the coils fabrication, and the optical micrographs of the fabricated coils produced at different laser powers of (b) 3 mW, (c) 4 mW, and (d) 5 mW. Inset: close-up views of the cross sections at the end of the coils. 12 Figure 4. (a) Illustration of a model 3D structure, and fabricated structure shown by (b) the top view, (c) the side view with the face of the elephant, and (d) the side view with the back of the elephant. 13 Figure 5. SEM images of the animals on the boat. (a) The overview of both the dog and the elephant. (b) The top view of the elephant. (c) The side view of the dog. 14
1709.06937	1	1709	2017-07-24T14:42:07	Neutronic Analysis on Potential Accident Tolerant Fuel-Cladding Combination U$_3$Si$_2$-FeCrAl	[ "physics.app-ph" ]	Neutronic performance is investigated for a potential accident tolerant fuel (ATF),which consists of U$_3$Si$_2$ fuel and FeCrAl cladding. In comparison with current UO$_2$-Zr system, FeCrAl has a better oxidation resistance but a larger thermal neutron absorption cross section. U$_3$Si$_2$ has a higher thermal conductivity and a higher uranium density, which can compensate the reactivity suppressed by FeCrAl. Based on neutronic investigations, a possible U$_3$Si$_2$-FeCrAl fuel-cladding systemis taken into consideration. Fundamental properties of the suggested fuel-cladding combination are investigated in a fuel assembly.These properties include moderator and fuel temperature coefficients, control rods worth, radial power distribution (in a fuel rod), and different void reactivity coefficients. The present work proves that the new combination has less reactivity variation during its service lifetime. Although, compared with the current system, it has a little larger deviation on power distribution and a little less negative temperature coefficient and void reactivity coefficient and its control rods worth is less important, variations of these parameters are less important during the service lifetime of fuel. Hence, U$_3$Si$_2$-FeCrAl system is a potential ATF candidate from a neutronic view.	physics.app-ph	physics	Neutronic analysis on potential accident tolerant fuel-cladding combination U3Si2-FeCrAl CHEN Shengli1 YUAN Cenxi1,* 1 Sino-French Institute of Nuclear Engineering and Technology, Sun Yat-sen University, Zhuhai, Guangdong 519082, China Corresponding author. E-mail address: [email protected] Abstract Neutronic performance is investigated for a potential accident tolerant fuel (ATF), which consists of U3Si2 fuel and FeCrAl cladding. In comparison with current UO2-Zr system, FeCrAl has a better oxidation resistance but a larger thermal neutron absorption cross section. U3Si2 has a higher thermal conductivity and a higher uranium density, which can compensate the reactivity suppressed by FeCrAl. Based on neutronic investigations, a possible U3Si2-FeCrAl fuel-cladding system is taken into consideration. Fundamental properties of the suggested fuel-cladding combination are investigated in a fuel assembly. These properties include moderator and fuel temperature coefficients, control rods worth, radial power distribution (in a fuel rod), and different void reactivity coefficients. The present work proves that the new combination has less reactivity variation during its service lifetime. Although compared with the current system, it has a little larger deviation on power distribution, a little less negative temperature coefficient and void reactivity coefficient, and the control rods worth of it is less importance, variations of these parameters are less important during the service lifetime of fuel. Hence, U3Si2-FeCrAl system is a potential ATF candidate from a neutronic view. Keywords: Neutronic analysis, Accident Tolerant Fuel, FeCrAl, U3Si2 1. Introduction After the Fukushima nuclear disaster in 2011, extensive focuses on the accident tolerant fuel (ATF) have been developed into seeking of advanced nuclear fuel and cladding options. Many materials for fuel and cladding have been investigated over the past years. Three potential approaches have been proposed for the development of the fuel and cladding with enhanced accident tolerance [1]: 1. Modifications of current zircaloy alloy in order to improve the oxidation resistance, including the coating layer design. 2. Replacement of zircaloy cladding by an alternative high-performance oxidation resistant cladding. 3. Improvement or replacement of the ceramic oxide fuel. The cladding material should have good oxidation resistance, proper delay of the ballooning and burst [2], stress resistance, and small thermal neutron absorption cross section. Among a large number of candidates, stainless steels have better mechanic performance than the current zircaloy-4 alloy. FeCrAl is a potentially promising excellent cladding material [1][3][4]. For example, its oxidation rate is at least two orders of magnitude lower than that of zircaloy [1]. On the other hand, FeCrAl is better to be applied as monolithic cladding than coating, in consideration of matching the thermal expansion coefficient [2], the diametrical compression [2], the volumetric and microstructural evolution [2], the high temperature oxidation protection [5], and the inter diffusion with zirconium [6], which is the reason that the material, Cr coated zircaloy, which is of better mechanic performance has not been selected [7]. In the present work, monolithic FeCrAl is chosen as the potential ATF cladding material. Many properties have to be considered for the investigation of fuel materials, such as the heavy metal density, the melting point, and the thermal conductivity. Uranium mononitride (UN) based composite fuels may have potential benefits when applied in light water reactor because of its enhanced thermal conductivity and large fuel density. However, UN chemically reacts with water [8], especially at high temperature. Additional shielding material UN/U3Si5 has been studied to overcome the defect [9]. But a problem still exists, which is the determination of the percentage of U3Si5 to prevent the fuel from reacting with water in an accident condition. One of the possible solutions is the Tristructural-isotropic (TRISO) fuel design [10][11]. Uranium-silicon binary system, which is thermodynamically stable, is another potential fuel [12]. Among the multiple compounds, U3Si and U3Si2 are the best candidates due to their high uranium densities. However, U3Si swells considerably under irradiation and dissociates into U3Si2 and solid solution U above 900°C, which is below some possible temperatures in uranium silicide fueled pins [13]. U3Si2 has promising records under irradiation in research reactor fuels and maintains several advantageous properties compared with UO2. In consideration of neutronic performance, the thermal neutron absorption cross section of FeCrAl is 2.43 barns, while that of Zircaloy is 0.20 barns [14]. In order to compensate the larger cross section, one needs to decrease the thickness of cladding and/or increase the quantity of fissile nuclides in the fuel. Under economical and safety considerations, the present work chooses nuclear fuel U3Si2 which has higher uranium density than the current UO2 fuel. Since the neutronic performance of FeCrAl is different from current Zircaloy and U3Si2 has different uranium percentage from UO2, analysis on neutronic performance is necessary for the new fuel-cladding system. Neutronic analyses on U3Si2-FeCrAl, U3Si2-SiC, and UO2-Zr are performed based on an I2S-LWR [15]. In addition, in order to study the core performance under normal and accident conditions, it is also necessary to investigate the fundamental properties of nuclear reactor core besides the neutron economy analysis, such as moderator and fuel temperature coefficients, control rods worth, radial power distribution, and different void reactivity coefficients. 2. Methodology and input parameters The neutronic behavior is performed using TRITON and KENO-VI modules from SCALE 6.1 [16][17][18]. SCALE provides a "plug-and-play" framework with 89 computational modules, including three Monte Carlo and three deterministic radiation transport solvers. It includes state of the art nuclear data libraries and problem- dependent processing tools for both continuous-energy and multigroup neutronic calculations, multigroup coupled neutron-gamma calculations, as well as activation and decay calculations. TRITION is a module for isotopic depletion calculation. KENO-VI is used for critical calculation. The nuclear data used in our simulation is based on ENDF/B-VII.0 238-group neutron library [19]. Monte Carlo codes are more reliable for radial distribution calculation than the deterministic codes due to the self-shielding. The advantage of Monte Carlo codes is especially evident in such calculation for a new fuel-cladding combination considering the less correction formula existing in deterministic codes. The radial power distribution is calculated by the Monte Carlo based code RMC [20]. RMC is a 3-D Monte Carlo neutron transport code developed by Tsinghua University. The code RMC intends to solve comprehensive problems in reactor, especially the problems on reactor physics. It is able to deal with complex geometry, using continuous-energy pointwise cross sections ENFF/B-VII.0 for different materials and at different temperatures. It can carry out both criticality and burnup calculations, which help to obtain the effective multiplication factor and the isotopic concentrations at different burnup level. Monte Carlo method is also used in fundamental nuclear physics. Nuclear structure problem can be solved with traditional nuclear shell model [21][22] or Monte Carlo shell model [23]. 2.1 Geometric parameters and model description The Westinghouse 17×17 assembly design is used, as shown in Fig. 1. Larger rings placed within the lattice represent the guide tubes and instrumentation tube. Control rods are inserted in the 24 tubes except the center tube for instrumentation. When no control rods or instrument are inserted, these tubes are filled with moderator. All the simulations are based on one assembly unit, except the calculation on radial power distribution. The infinite lattice cell is used to calculate the effective resonant cross section. Fig. 1. Westinghouse 17×17 PWR lattice model In order to maintain the power transfer, pitch-to-rod-diameter (P/D) is kept constant at 1.326. The pellet-cladding gap is also kept constant at 82.55 µm to maintain the thermal conductivity of the gap. The change of cladding thickness is achieved by changing the pellet radius and the inner diameter of cladding. An average value of 630 ppm boron, which is the equivalent concentration at Middle of Cycle (MOC), is placed in the coolant. The other parameters are all presented in Table 1. Property Assembly fuel height cm Reference Cladding composition wt% mm Fuel pellet radius Gap thickness µm Cladding Inner Radius mm Cladding thickness µm Cladding Outer Radius mm Fuel enrichment P/D Cladding IR of guide tube Cladding OR of guide tube Number of guide tubes mm Fuel density Specific power density for reference UO2-Zr Coolant density g/cm3 25 U3Si2: 11.57 (94.7% theoretical density) UO2: 10.47 (95.5% theoretical density) MW/ 38.33 MtU g/cm3 0.7119 Unit Value 365.76 Zr-4: Fe/Cr/Zr/Sn = 0.15/0.1/98.26/1.49 FeCrAl: Fe/Cr/Al = 75/20/5 4.09575 82.55 4.1783* 571.5* 4.7498 4.9* 1.326 5.624 % mm 6.032 [4] [16] [4] Helium density g/L g/cm3 Cladding density 1.625 (2.0 MPa) Zircaloy-4 cladding: 6.56 FeCrAl cladding: 7.10 580 Coolant temperature K 900 K Fuel temperature Cladding and gap 600 K temperature EFPD 1500 Simulation time 630 ppm Boron concentration Boundary conditions Reflective Assembly design Westinghouse 17×17 PWR fuel rod * Values only for the reference case Table 1. Model specification [24] [4] In Table 1, the density of helium gas is roughly calculated by the ideal gas equation with a pressure of 2.0 MPa [24]. The density of helium gas has little influence on neutronic analysis. It should be noted that both fuel enrichment and cladding thickness are changed in the following discussions in order to study the sensitivity of the two parameters. 2.2 EOC reactivity calculations The batch-specific parameters of a typical Westinghouse PWR core are given in Table 2 [25]. The core fraction volume is defined by the number of assemblies of the total 193 assemblies presented in a PWR core for each depletion cycle. The relative assembly power is the ratio between power per batch and the average power of the core. The Effective Full Power Days (EFPDs) achieved at the EOC represent the cycle EFPDs for each batch. Batch 1 2 3 Total Number of assemblies 73 68 52 193 Core fraction vol% (Vb) 38% 35% 27% 100% Relative assembly power (Pb) 1.25 1.19 0.40 - EFPDs achieved at EOC (eb) 627 1221 1420 - Table 2. Distribution in Population and Power per Fuel Cycle Batch in a Typical Westinghouse PWR The infinite multiplication factor (k-infinity) at 627, 1221, and 1420 EFPDs are , the difference of reactivity in the reactor used to estimate the EOC reactivity core level compared with that of the current PWR core, for each set of fuel geometry. According to the equivalent reactivity method described in Ref. [26], the EOC reactivity for each case is compared with that of a reference case (standard PWR fuel corek∆ rod containing 4.9% enriched UO2 pellets). The average eigenvalue difference of the core can be estimated by the formula below: ) k Σ ∆ ( e PV b b b b − PV b b b inf Σ , (1) k ∆ core = b k −∆ inf b where is the difference of k-infinity between the fuel design under consideration and the reference case for batch b as a function of exposure (eb). The EOC EFPD values in Table 2 are used to quantify the level of exposure for each batch. The power weighting factor (Pb) approximates the power distribution in the core to provide the contribution of each batch. The number of assemblies per batch is denoted as Vb. The EOC reactivity calculation is a simple but accurate method for the estimation of the cycle length. Through the reactivity analysis at the EOC, the U3Si2-FeCrAl fuel-cladding combination is suggested in the following section to replace the current combination in the condition of no reduction in the cycle length. 2.3 Moderator temperature coefficient calculation For a new fuel-cladding combination, the Moderator Temperature Coefficient (MTC) is the safety parameter to be firstly analyzed because it reflects the feedback of reactivity during incident and accident conditions in reactor core. To analyze the MTC, only the temperature of moderator is to be suddenly changed to compare the reactivity at different temperatures. In the present study, the MTC is analyzed with 630 ppm constant boron concentration for feasibility. The density variation with temperature should be considered for liquid phase. The thermal expansion coefficient of the moderator is β𝑀=0.00329871/℃ [27]. 2.4 Control rod worth as a function of EFPDs In the design and the analysis of a nuclear reactor core, the reactivity worth of control rods (i.e. their efficiency in neutron absorption) is another important parameter. The control rods worth is affected by the fuel burnup due to the isotopes variation during operation. For global critical safety consideration, the present work explores the integral worth of traditional and currently used Ag(80%)-In(15%)- Cd(5%) control rods [28] in an assembly as a function of EFPDs because the sub- critical condition must be ensured in the case of full insertion of control rods. 2.5 Radial distribution of power Many investigations show a non-uniformed radial power distribution in a fuel rod due to the spatial self-shielding. The strong neutron absorption of 238U at certain energies induces the significant plutonium production, and local power and burnup near the surface of the fuel rod. Therefore, it is of great importance to investigate the rim effect when considering the new fuel-cladding system. The fuel region is divided into 9 rings, while the gap and cladding are located outside, as shown in Fig. 2. Fig. 2. Radial profile of a fuel-cladding system with fuel region divided into 9 rings 3. Results and discussion 3.1 Depletion k-infinity results Fig. 3 shows the infinite multiplication factor k-infinity of a standard assembly as a function of EFPDs. The reference scenario uses the UO2-Zr fuel rod with parameters shown in Table 1. Results of U3Si2-FeCrAl fuel rod are also shown with 4.9% enriched U3Si2 and different cladding thicknesses. For U3Si2-FeCrAl cases, k-infinity decreases when the cladding thickness increases because of the smaller volume of U3Si2 and the more neutrons absorbed by FeCrAl. At the beginning, k-infinity in the reference scenario is larger than that in U3Si2- FeCrAl cases, which may be caused by the hardened spectrum (presented in section 3.2). The k-infinity of the reference scenario decreases more quickly than in U3Si2- FeCrAl cases with the increment of EFPDs. One reason is that the burnup in U3Si2- FeCrAl cases is less than that in the reference case. Another reason is the larger production of the fissile nuclide 239Pu in U3Si2-FeCrAl scenarios than that in the reference scenarios, as shown in Fig. 4. The larger concentrations of fissile nuclides give positive contribution to the k-infinity, which makes the k-infinity reduces less quickly in the new system. More 239Pu are produced because FeCrAl cladding absorbs more thermal neutrons and 238U nuclei absorb more fast neutrons, especially of which the energy is in the resonance region, to produce 239Pu at the same EFPDs. Fig. 5 shows that Nf concentration (equivalent 235U concentration considering fission) of the new combination (with 4.9% enriched U3Si2-FeCrAl system and the same cycle length as the reference case) varies less than that of the reference scenario. The reason is that more 239Pu, of which the fission cross section is almost 2.5 times that of 235U [29], are produced. Although U3Si2-FeCrAl system has less variation in reactivity from low to high burnup at constant boron concentration, the boron coefficient is also less negative due to the hardened spectrum, which is also presented in UO2-FeCrAl combination [30] and UN-U3Si2 combination [31]. Higher boron concentration is needed at the BOC for critical condition, which is presented in an I2S-LWR [15]. Fig. 3. k-infinity versus EFPDs for 4.9% enriched fuel Remark: in all figures, ref represents the reference case and new represents the 4.9% enriched U3Si2 fuel and 450 𝜇 m thickness of FeCrAl cladding. Fig. 4. Concentration of 239Pu versus EFPDs Fig. 5. Concentration of 235U and equivalent fissile isotopic versus EFPDs Table 3 displays the corresponding to each cladding thickness and uranium enrichment. In this work, one of the most important objectives is to find the relationship between the uranium enrichment and the cladding thickness without reduction in cycle length. It is thus not necessary to analyze the reactivity of combinations with far away from zero. The specific power per metric tonne of uranium (MW/MtU) corresponds to the constant power of 18.0 MW/assembly, which corek∆ corek∆ corek∆ Cladding thickness t (µm) means 38.33 MW/MtU specific power density for reference 4.9% enriched UO2-Zr case. The positive (negative) values of signify longer (shorter) cycle length than the reference scenario. 235U enrichment e (%) 4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2 540 - - -0.02456 -0.01933 -0.01490 -0.00929 -0.00456 0.00001 Table 3. Cycle reactivity difference 400 - -0.00669 -0.00213 0.00377 0.00827 0.01242 0.01748 - - 450 - -0.01465 -0.00935 -0.00386 0.00019 0.00528 0.00999 0.01538 - 571.5 - - -0.03060 -0.02499 -0.02007 -0.01460 -0.00968 -0.00452 500 - - -0.01778 -0.01304 -0.00806 -0.00238 0.00156 0.00712 350 -0.00484 0.00078 0.00522 0.01097 0.01565 0.02018 corek∆ between the U3Si2-FeCrAl and reference With these values, by linear fitting of thickness, one can obtain: corek∆ system as function of fuel enrichment and cladding corek ∆ = ae bt + + , (2) c , b × ± ± = − a = 0.17234 0.00190 0.04980 0.00042% ± with the coefficient of determination where e and t represent uranium enrichment (%) and cladding thickness (µm), and respectively, c = − . Positive value of the coefficient a and negative value of the coefficient b are logical because the increment of uranium enrichment increases the cycle length, while the increment of cladding thickness has a converse effect. The coefficient c is negative because when the uranium enrichment is too low, the cycle length in the reference case can never be achieved. 1.59488 0.01050 10 0.99875 4 mm− 2 R = Based on the above results, it can be concluded that 1% uranium enrichment at the EOC, while 1 µm more in cladding thickness contributes +4980 pcm to induces -16 pcm. In other words, a 1% increment in uranium enrichment can compensate the negative reactivity induced by an increment of 312 µm in the thickness FeCrAl cladding. corek∆ According to equation (2), the cladding thickness needs to be 450µm to keep the same cycle length and uranium enrichment as in the reference case, 4.9%. The 4.9% enriched U3Si2 fuel and 450µm thickness of FeCrAl cladding is the new fuel-cladding combination which will be discussed hereafter. In general, for U3Si2-FeCrAl fuel-cladding system, in order to achieve the same cycle length as the current 4.9% enriched UO2-Zr system, the relationship between uranium enrichment and cladding thickness is taken as: bt e = − c or e = 3.2 10 × 3 t− + 3.5 , (3) + a In practice, the uranium enrichment and cladding thickness for U3Si2-FeCrAl system should be located in the white area in Fig. 6 to achieve no shorter cycle length compared with the current UO2-Zr system. Fig. 6. Feasible combination in consideration of cycle length Stainless steel of a cladding thickness of 350 µm was used as nuclear fuel cladding [32]. For the same FeCrAl cladding with a thickness of 350 µm, only 4.58% enriched U3Si2 fuel is needed to achieve the cycle length in the reference scenario, while 5.06% uranium enrichment of UO2 fuel is needed, as mentioned in Table 6 in Ref. [3]. It costs less for the fabrication of enriched nuclear fuel due to the lower uranium enrichment. 3.2 Spectral hardening Spectral hardening is a phenomenon of FeCrAl cladding because of its higher thermal neutron absorption cross section than that of the zirconium alloy. Another reason is that the moderator-to-fuel ratio is smaller in the new combination, which reduces the resonance escape probability and the percentage of thermal neutrons. The hardening of the neutron spectrum is also found in the transition from low burnup to high burnup. This phenomenon is due to the accumulation of fission products and actinides. A hardened neutron spectrum includes less thermal neutrons, which mainly induce the fission. As a consequence, the reactivity is reduced in the case of spectral hardening. A possible solution to compensate such effect is the increment of the moderator-to-fuel ratio. Fig. 7. Spectrum of neutron flux 3.3 Temperature coefficients The MTC and the Fuel Temperature Coefficient (FTC, also known as Doppler coefficient) are two of the most important temperature feedback parameters in nuclear safety because they automatically control the reactivity of the core when temperature changes. It is thus of great interest to study the two parameters for a new fuel-cladding combination. From the definition of the reactivity = ρ , the difference of two reactivity can 1k − k be calculated using the corresponding multiplication factor k as: ρ ρ ρ ∆ = 1 − 2 = k k 1 − 2 k k 1 2 , (4) In the calculation of temperature coefficients, all concentrations of nuclides in the fuel are extracted at each burnup depth in normal condition. The moderator temperature and the corresponding density (fuel temperature respectively) are changed through the MTC (FTC respectively). The corresponding multiplication factors are also calculated using the concentrations of nuclides extracted in normal condition calculation. The difference of reactivity can be obtained through Eq. (4). The temperature coefficients are determined by dividing the difference of reactivity by the corresponding difference of temperature. the definition of The MTC is shown in Fig. 8 for the new combination and the reference scenario. The MTC values are determined by the reactivity calculation in normal condition and at 610 K moderator temperature. At the beginning, the MTC value is more negative for new combination due to the hardened spectrum, which reduces the number of thermal neutrons and enhances the importance of the moderator. As shown in Fig. 10, the FTC is always negative because the increment of fuel temperature leads to the broadening of resonance absorption of 238U. The FTC values become more and more negative with burnup due to the increment of 238U to Nf ratio, which reflects the impact of 238U. Moreover, the less negative FTC values of the new combination are due to the smaller 238U to Nf ratio. After the beginning of the operation, the MTC values become more and more negative. The reason is that the increment of 238U to Nf ratio (as shown in Fig. 9) enhances the resonance absorption of 238U, which leads to the decrement of resonance escape probability. Such effect emphasizes the importance of moderator and results in enlarged negative MTC values with burnup. The less negative MTC value of the new combination can also be explained by the lower 238U to Nf ratio (as shown in Fig. 9) due to the hardened spectrum with FeCrAl cladding, as explained in section 3.1. It should be remarked that the MTC is calculated at 630 ppm boron concentration. In practice, the MTC is less negative at the BOC because of the actually higher boron concentration, and more negative at EOC due to the lower boron concentration. Fig. 8. MTC vs EFPDs Fig. 9. 238U to Nf ratio for 630 ppm boron concentration Fig. 10. FTC vs EFPDs 3.4 Moderator temperature sensitivity The reactivity of a reactor core depends on the moderator temperature. The MTC parameter represents the reactivity induced by sudden change of moderator temperature. It is also important to study the Moderator Temperature Sensitivity (MTS) at different moderator temperatures, especially for a new fuel-cladding The MTS increases with boron concentration as shown in Fig. 11. The reason is that the decrement of moderator density also reduces the boron density, which has an opposite effect. In the range of 0 to 500EFPDs, MTS decreases with EFPDs for the same reason as MTC because there is no great difference between different scenarios with the same boron concentration. combination. The MTS represents the reactivity sensitivity induced by different moderator temperatures design, which is different from current moderator temperature (580 K) in core design. The moderator temperature keeps constant during a cycle length for the MTS calculations because the MTS reflects the reactivity change in normal operation with different temperatures design. Nonetheless, in the range of 500 to 1500 EFPDs, the MTS increases with burnup because of the larger accumulated number of 239Pu in less reactivity cases. This is caused by the higher moderator temperature during operation, as shown in Fig. 12. Accumulation of 239Pu has an opposite effect on the increment of moderator temperature due to the large fission cross section. When burnup is larger than 900 EFPDs, the MTS of the new combination is more negative. This is due to lower burnup than the reference scenario caused by higher uranium quantity in U3Si2. The MTS value is about -20 pcm/K at 1500 EFPDs for the new combination. In spite of less thermal efficiency with lower moderator temperature, the decrement of moderator temperature is considerable for the new combination in consideration of the prolongation of the service life of fuel, while it is much less effective for the reference scenario. On the contrary, the increment of moderator temperature in the new design largely reduces the fuel life. Fig. 11. MTS vs EFPDs Fig. 12. Concentration of 239Pu for 580 and 610 K moderator temperature 3.5 Control rods worth In a similar method to that in the determination of temperature coefficients, calculations are performed to obtain the control rods worth. The multiplication factor is firstly calculated as a function of burnup. The concentrations of all nuclides in the fuel are extracted for each burnup depth. The multiplication factor is obtained at each burnup through the same concentrations and the insertion of all the control rods in 24 tubes. The control rods worth is calculated directly using Eq. (4). It is remarkable that the control rods worth is often expressed by its absolute value. The integral control rods worth in an assembly is shown in Fig. 13. Control rods worth increases with burnup due to the depletion of 235U and the decrement of total fissile nuclides. In other words, the ratio of absorbent isotopes in control rods to fissile isotopes in fuel increases with burnup. In the case of new combination, the control rods worth is smaller than in the reference case. One reason is the higher uranium quantity in the new combination as the control rods are the same. Another reason is that FeCrAl cladding has larger macroscopic thermal neutron absorption cross section, which also plays a similar role as control rods. In other words, spectrum hardening reduces the effect of control rods. It will be of great importance to verify that the control rods worth meets the safety requirement for the new fuel-cladding system. Lower burnup and lower moderator-to-fuel ratio lead to less variation of the control rods worth for the new combination than in the reference case, which is another advantage of the new fuel-cladding combination. Fig. 13. Control rod worth vs EFPDs 3.6 Radial power distribution The normalized power distribution versus relative radius is shown in Fig. 14. Because of the spatial self-shielding, some resonance neutrons are shielded from the center of the fuel due to the absorption by the outer fuel. The relative fission power is largest near the surface of the fuel rod and smallest at the center of the pellet. At the BOL, although there is certain difference in neutron absorption ability between two cladding materials, there is no significant difference in the relative power in a fuel rod. During the operation, more plutonium is produced near the surface of the fuel pellet than at the center because of the spatial self-shielding. Additional fissile nuclides cause a sharp increment in fission reactions near the surface of the fuel rod. As shown in Fig. 14, fission profiles in inner regions are flat, while a sharp increment in power is found close to bound. The new combination has a little larger relative power near the surface because more 239Pu are produced in the outer ring, as shown in Fig. 15. Fig. 14. Relative radial power distribution Fig. 15. Radial distribution of 239Pu Fig. 15 shows the concentration of 239Pu at the EOL. 239Pu is chosen because it is the most important fissile nuclide besides 235U. More 239Pu are accumulated in the new combination, which corresponds to the results in Fig. 5. Such fact further emphasizes that cladding materials with higher absorbing ability induce less reactivity in early life due to the hardened neutron spectrum but larger reactivity near the EOL due to the more considerable accumulation of plutonium. 3.7 Void reactivity coefficient The Void Reactivity Coefficient (VRC) is calculated in a similar way to the calculation of the temperature coefficients and the control rods worth, with the change of the moderator void. The VRC calculation is of great interest for a new fuel-cladding combination because it describes the variation of reactivity of a reactor in Loss of Coolant Accident (LOCA) scenario. Similar to the MTC, a more negative VRC value is better in the consideration of nuclear safety. It is remarkable that the shape of the VRC as function of EFPDS is similar to that of MTC because the increment of the moderator void has a similar effect to that of the increment of the moderator temperature, which reduces the moderator density. The reason that the VRC is less negative for the new combination is thus the same as that explained in MTC analysis. At the very beginning, the VRC value increases with burnup. This is not caused by the softened neutron spectrum (it is actually hardened as shown in Fig. 17) but by the abrupt presence of 239Pu, which has a positive contribution to the VRC [33]. Because of the less negative value of the VRC, the feedback of anti-reactivity in LOCA scenario is less important for the new combination. Therefore, it is another parameter which must be considered in a new core design. Fig. 16. VRC for 5% and 50% void Fig. 17. Neutron flux spectrum 3.8 Void reactivity sensitivity The VRC is a parameter that represents the reactivity change caused by sudden change of moderator quantity. In the primary circuit of a PWR, a little void in moderator may exist. The increment of the percentage of void in moderator is also used to study the neutronic performance of decreased moderator-to-fuel ratio when the geometry is kept the same as the current core design. Similar to the study of MTS, the Void Reactivity Sensitivity (VRS) is investigated by setting different void percentages in moderator but keeping them constant during the cycle length. As shown in Fig. 18, the shapes of the VRS as function of EFPDs are similar to that of the MTS. The reason is that the VRS and the MTS are both parameters that measure the reactivity change by modifying the moderator density. The increment of VRS at high burnup is from the large number of accumulation of 239Pu (as shown in Fig. 19) due to the hardened spectrum. The dependence of k-infinity on the percentage of void of moderator is shown in Fig. 20. Fig. 18. VRS for 5% and 50% void Fig. 19. Concentration of 239Pu with different percentage of void of moderator Fig. 20. k-infinity with different percentage of void of moderator 4 Conclusion FeCrAl is a potential cladding material due to its excellent oxidation resistance. But FeCrAl has the neutronic penalty compared with Zr-4 due to its larger thermal neutron absorption cross section. Better neutron economy performance is found for U3Si2 fuel than UO2 fuel, which provides a potential solution for all alternative claddings with larger thermal neutron absorption cross section than zirconium alloys. The relationship between uranium enrichment and cladding thickness of U3Si2- FeCrAl fuel-cladding combination is found by reactivity analysis in the condition of no reduction in the cycle length compared with that of UO2-Zr system. The critical cladding thickness is determined by keeping the uranium enrichment the same as in the reference scenario. The investigation proves that the new combination has less reactivity variation during its lifetime. The new combination provides less negative feedback for both fuel and moderator temperature reactivity than the reference case. Other disadvantages of the new combination are that it has a little larger deviation in power distribution, a less important control rods worth, and a less negative VRC than the current system. Nonetheless, the MTC, the FTC, the control rods worth and the VRC of the new combination are better in practice in consideration of the stability during the lifetime. Furthermore, it is possible to prolong the service time of the fuel by reducing the moderator temperature for the new combination. It should be noted that the sensitive and uncertainty analyses are also very important for a model calculation, such as the uncertainty analysis on a simple nuclear mass model which has been performed recently [34]. U3Si2-FeCrAl system is a potential ATF candidate with many advantages compared with potential UO2-FeCrAl system. For example, lower uranium enrichment is required and larger cladding thickness is permitted in this system. Moreover, the thermophysical properties of U3Si2 have been studied up to a temperature of 1773 K [35], which provides a complementary knowledge for this fuel material. The present work is also one of the possible new potential ATF fuel-cladding combinations proposed by the U.S. DOE NE Advanced Fuels Campaign [36], which shown that the U3Si2 and stainless steel can be considered as potential fuel and cladding, respectively. Nevertheless, chemical reactions between water and U3Si2-FeCrAl system should be taken into the consideration in a reactor with water as the moderator. The possible chemical reaction between U3Si2 and Al [37] should also be taken into account. Moreover, attention should be paid to different fuel densities due to the different porosities caused during the fabrication of the fuel. Conflicts of Interest The authors declare that there is no conflict of interest regarding the publication of this manuscript. Acknowledgments The authors acknowledge the authorized usage of the RMC code from Tsinghua University for this study. The authors acknowledge also the useful discussions with Dr. XIAO Sicong. This work has been supported by the National Natural Science Foundation of China under Grant No. 11305272, the Special Program for Applied Research on Super Computation of the NSFC Guangdong Joint Fund (the second phase), the Specialized Research Fund for the Doctoral Program of Higher Education under Grant No. 20130171120014, the Fundamental Research Funds for the Central Universities under Grant No. 14lgpy29, the Guangdong Natural Science Foundation under Grant No. 2014A030313217, and the Pearl River S&T Nova Program of Guangzhou under Grant No. 201506010060. References [1] Ott L J, Robb K, Wang D. Preliminary assessment of accident-tolerant fuels on LWR performance during normal operation and under DB and BDB accident conditions. 520–533. Doi:10.1016/j.jnucmat.2013.09.052 Mater, 2014, Nucl 448: J [2] Zinkle S J, Terrani K, Gehin J et al. Accident tolerant fuels for LWRs: A 374-379. Mater, 2014, 448: perspective. Doi:10.1016/j.jnucmat.2013.12.005 Nucl J [3] George N M, Terrani K, Powers J et al. Neutronic analysis of candidate accident-tolerant cladding concepts in pressurized water reactors. Ann Nucl Energy, 2015, 35: 703-712. Doi:10.1016/j.anucene.2014.09.005 [4] Wu X, Kozlowski T, Hales J. Neutronic and fuel performance evaluation of accident tolerant FeCrAl cladding under normal operation conditions. Ann Nucl Energy, 2015, 85: 763-775. Doi:10.1016/j.anucene.2015.06.032 [5] Terrani K A, Zinkle S J, Snead L. Advanced oxidation-resistant iron-based alloys for LWR fuel cladding. J Nucl Mater, 2014, 448: 420-435. Doi:10.1016/j.jnucmat.2013.06.041 [6] Terrani K A, Parish C, Shin D et al. Protection of zirconium by alumina and chomia-forming iron alloys under high-temperature steam exposure. J Nucl Mater, 2013, 438: 64-71. Doi:10.1016/j.jnucmat.2013.03.006 [7] Brache J C, Lorrette C, Michaux A et al. CEA studies on advanced nuclear fuel claddings for enhanced Accident Tolerant LWRs Fuel (LOCA and beyond LOCA conditions). Fontevraud 8 - Contribution of Materials Investigations and Operating Experience to LWRs' Safety, Performance and Reliability, France, Avignon – 2014, September 15 │18 [8] Ferris L M. Reactions of uranium mononitride, thorium monocarbide and uranium monocarbide with nitric acid and other aqueous reagents. J Inorg Nucl Chem, 1968, 30(10):2661-2669. Doi:10.1016/0022-1902(68)80393-8 [9] Brown N R, Todosow M, Cuadra A. Screening of advanced cladding materials and UN–U3Si5 fuel. Doi:10.1016/j.jnucmat.2015.03.016 J Nucl Mater, 2015, 462: 26–42. [10] Powers J J, Lee W J, Venneri F et al. Fullu ceramic microencapsulated (FCM) replacement fuel for LWRs. ORNL/TM-2013/173, KAERI/TR-5136/2013, 2013 [11] Brown N R, Ludewig H, Aronson A et al. Neutronic evaluation of a PWR with fully ceramic microencapsulated fuel. Part I: Lattice benchmarking, cycle length, and reactivity coefficients. Ann Nucl Energy, 2013, 62: 538-547. Doi:10.1016/j.anucene.2013.05.025 [12] Nelson A T, White J T, Byler D D et al. Overview of Properties and Performance of Uranium-Silicide Compounds for Light Water Reactor Applications. T Am Nucl Soc, 2014, 110, Jun 15–19. [13] Jason M H, Paul A L, Rita E H. Uranium Silicide Fabrication for use in LWR Accident Tolerant Fuel. Idaho National Laboratory, PO Box 1625, MS 6188, Idaho Falls, ID 83415-6188 [14] George N M, Terrani K, Powers J. Neutronic analysis of candidate accident- tolerant iron alloy cladding concepts. Oak Ridge National Laboratory, ORNL/TM-2013/121, 2013 [15] Salazar, D., Franceshini F, Ferroni P et al. "I2S-LWR Fuel Management Option for an 18-Month Cycle Length." In Proceedings of ANFM V (2015). http://www.casl.gov/docs/CASL-U-2015-0154-000.pdf [16] DeHart M D, Bowman S M. Reactor physics methods and analysis capabilities in SCALE. Nucl Technol. 2011, 174 (2): 196–213 DOI: 10.13182/NT174-196 [17] Bowman S M. SCALE 6: comprehensive nuclear safety analysis code system. Nucl Technol. 2011, 174 (2), 126–148. DOI: 10.13182/NT10-163 [18] SCALE: A Comprehensive Modeling and Simulation Suite for Nuclear Safety Analysis and Design, ORNL/TM-2005/39, Version 6.1. Available from Radiation Safety Information Computational Center at Oak Ridge National Laboratory as CCC-785. June 2011, Oak Ridge National Laboratory, Oak Ridge, Tenn. [19] Chadwick M, Obložinský P, Herman M et al. ENDF/B-VII. 0: next generation evaluated nuclear data library for nuclear science and technology. Nucl Data Sheets 2006:107 (12), 2931–3060. Doi:10.1016/j.nds.2006.11.001 [20] K. Wang, Z. Li, D. She, et al., "RMC – A Monte Carlo Code for Reactor International Conference on Mathematics and Analysis," Physics Computational Methods Applied to Nuclear Science and Engineering, M and C 2013, v 1, p 89-104, Sun Valley, ID, USA, May 5-9, 2013. [21] C. Yuan, C. Qi, F. Xu, et al., "Mirror energy difference and the structure of loosely bound proton-rich nuclei around A=20", Physical Review C, 89, 044327 (2014). [22] C. Yuan, Z. Liu, F. Xu, et al., "Isomerism in the "south-east" of 132Sn and a predicted neutron-decaying isomer in 129Pd", Physics Letters B, 762 (2016) 237–242. [23] T. Otsuka, M. Honma, T. Mizusaki, et al., "Monte Carlo shell model for atomic nuclei", Progress in Particle and Nuclear Physics, 47, 319-400, (2001). [24] YU Ji-yang , WANG Song-tao , ZHANG Xiang-yu et al. Numerical simulation on scatheless measurement of helium pressure in fuel rod. Chinese Journal of Nuclear Science and Engineering, 2007: Vol. 27 No. 3 [25] George N M, Maldonado I, Terrani K et al. Neutronics Studies of Uranium- Bearing Fully Ceramic Microencapsulated Fuel for Pressurized Water Reactors. Nucl Technol, 2014, 188(3): 238-251. Dx.doi.org/10.13182/NT14-3 [26] Driscoll, M. J., Downar, T. J., & Pilat, E. E. The linear reactivity model for nuclear fuel management. Amer Nuclear Society (1990) [27] Memoli V, Cammi A. Evaluation of the moderator temperature coefficient of reactivity in a PWR. Grenoble, France, 2007 [28] http://www.nek.si/en/about_nuclear_technology/technical_data/ [29] Yuan C, Wang X and Chen S. A Simple Formula for Local Burnup and Isotope Distributions Based on Approximately Constant Relative Reaction Rate. Sci Technol Nucl Ins, 2016. Dx.doi.org/10.1155/2016/6980547 [30] Nicholas R. Brown, Aaron J. Wysocki, Kurt A. Terrani et al. The potential impact of enhanced accident tolerant cladding materials on reactivity initiated accidents in light water reactors. Ann. Nucl. Energy 99 (2017) 353–365. Dx.doi.org/10.1016/j.anucene.2016.09.033 [31] Nicholas R. Brown, Arnold Aronson, Michael Todosow et al. "Neutronic performance of uranium nitride composite fuels in a PWR." Nucl Eng Des 275 (2014): 393-407. Dx.doi.org/10.1016/j.nucengdes.2014.04.040 [32] Strasser A, Santucci J, Lindquist K et al. Evaluation of stainless steel cladding for use in current design LWRs, New York, USA, NP2642, 1982 [33] Slavickas A, Pabarcius R, Tonkunas A et al. Decomposition Analysis of Void Reactivity Coefficient for Innovative and Modified BWR Assemblies. Sci Technol Nucl Ins, 2014. Dx.doi.org/10.1155/2014/132737 [34] Cenxi Yuan, "Uncertainty decomposition method and its application to the liquid drop model", Physical Review C, 93, 034310 (2016) [35] J.T. White, A.T. Nelson, J.T. Dunwoody et al. Thermophysical properties of 275–280. J Nucl Mater (2015) 464 U3Si2 Dx.doi.org/10.1016/j.jnucmat.2015.04.031 1773 K. to [36] Shannon M. Bragg-Sitton (Deputy National Technical Director, FCRD AFC). Update on Accident Tolerant Fuel and IAEA Technical Meeting on ATF. Idaho National Laboratory, April 22, 2015 [37] V. P. Sinha, D. Kohli, and R. Rakesh et al. Study on Chemical Interaction of U3Si2 Dispersoid and Aluminum Matrix in Plate Fuel Elements and Its Influence on the Mechanical Properties. Nucl Technol 192(1) · October 2015 DOI: 10.13182/NT14-59
1707.06731	1	1707	2017-07-21T01:02:22	Mn$_2$VAl Heusler alloy thin films: Appearance of antiferromagnetism and an exchange bias in a layered structure with Fe	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Mn$_2$VAl Heusler alloy films were epitaxially grown on MgO(100) single crystal substrates by means of ultra-high-vacuum magnetron sputtering. A2 and L2$_1$ type Mn$_2$VAl order was controlled by the deposition temperatures. A2-type Mn$_2$VAl films showed no spontaneous magnetization and L2$_1$-type Mn$_2$VAl films showed ferrimagnetic behavior with a maximum saturation magnetization of 220 emu/cm$^3$ at room temperature. An antiferromagnetic reflection was observed with neutron diffraction at room temperature for an A2-type Mn$_2$VAl film deposited at 400$^\circ$C. A bilayer sample of the antiferromagnetic A2 Mn$_2$VAl and Fe showed an exchange bias of 120 Oe at 10 K.	physics.app-ph	physics	(cid:1) Mn2VAl Heusler alloy thin films: Appearance of antiferromagnetism and an exchange bias in a layered structure with Fe Tomoki Tsuchiya1,a), Ryota Kobayashi2, Takahide Kubota3,4,b) , Kotaro Saito5, Kanta Ono5, Takashi Ohhara6, Akiko Nakao7, Koki Takanashi3,4. 1 Department of Material Science, Graduate School of Engineering, Tohoku University, Sendai 980-8579, Japan 2 Department of Physics, Graduate School of Science, Tohoku University, Sendai 980- 8578, Japan 3 Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan 4 Center for Spintronics Research Network, Tohoku University, Sendai 980-8577, Japan 5 High Energy Accelerator Research Organization (KEK), 1-1 Oho, Tsukuba 305-0801, Japan 6 J-PARC Center, Japan Atomic Energy Agency, Tokai 319-1195, Japan, 7 Research Center for Neutron Science and Technology, Comprehensive Research Organization for Science and Society, Tokai 319-1106, Japan a) Author to whom correspondence should be addressed. Electronic mail: [email protected] b) Author to whom correspondence should be addressed. Electronic mail: [email protected] (cid:1) (cid:1) (Abstract) Mn2VAl Heusler alloy films were epitaxially grown on MgO(100) single crystal substrates by means of ultra-high-vacuum magnetron sputtering. A2 and L21 type Mn2VAl order was controlled by the deposition temperatures. A2-type Mn2VAl films showed no spontaneous magnetization and L21-type Mn2VAl films showed ferrimagnetic behavior with a maximum saturation magnetization of 220 emu/cm3 at room temperature. An antiferromagnetic reflection was observed with neutron diffraction at room temperature for an A2-type Mn2VAl film deposited at 400ºC. A bilayer sample of the antiferromagnetic A2-type Mn2VAl and Fe showed an exchange bias of 120 Oe at 10 K. (cid:1) (cid:1) I. INTRODUCTION Exchange bias (EB) is the shift of a hysteresis loop caused by the interaction between an antiferromagnet and a ferromagnet, which was discovered by Meiklejohn and Bean in 1956 [1]. Since B. Dieny et al. applied EB for a spin valve structure [2], EB has been utilized in electronics such as magnetoresistive devices. To date, many researchers have focused on Ir-contained antiferromagnetic alloys which exhibit a large EB shift with high thermal stability [3-8]. However, the scarcity of iridium, which is the rarest element in the earth's crust [9], will be a critical issue for the stable supply of devices containing iridium in near future. Therefore, it is needed to find thermally stable antiferromagnets without iridium and also other scarce elements. We focus on Mn2VAl Heusler alloy for the replacement of Ir-containing antiferromagnets. L21-type Mn2VAl is known as a ferrimagnetic full Heusler alloy [10-15]. Recently, however, magnetization and neutron diffraction measurements of a bulk sample have demonstrated that A2-type Mn2VAl is an antiferromagnet with a Néel temperature above 600 K [16]. There is an advantage for Heusler-type antiferromagnets from a view point of the lattice matching with ferromagnetic materials for spintronics, such as Co-Fe alloys, and cobalt-based Heusler alloys in which highly spin polarized transports are realized for the layered film samples with a (001)-texture [17-20]. In this paper, we present the antiferromagnetism in A2-type Mn2VAl thin films and an EB shift in a Mn2VAl/Fe layered sample. The chemical phases of Mn2VAl films were successfully controlled by the deposition temperatures, and the magnetic properties were investigated by using laboratory facilities and neutron diffraction technique. (cid:1) (cid:1) II. EXPERIMENTAL PROCEDURES All samples were deposited on 0.5 mm-thick MgO(100) single crystal substrates by using ultra-high-vacuum (UHV) magnetron sputtering technique. Surfaces of MgO substrates were cleaned by flushing at 700ºC in the UHV-chamber before sputtering. Subsequently, a Mn2VAl layer was deposited by using an alloy target with the composition of Mn51V25Al24 (at.%) determined by inductively coupled plasma optical emission spectroscopy technique. The deposition temperatures (Tsub) were set at room temperature (RT) and temperatures from 300ºC to 800ºC in 100ºC increments. The thicknesses of Mn2VAl were 100 nm for structure determination and magnetization measurements and 1 µm for neutron diffraction measurements. A Ta capping layer with a thickness of 3 nm was deposited on the Mn2VAl layer at room temperature. The crystalline structures of Mn2VAl films were characterized by x-ray diffraction (XRD) and neutron diffraction. Mn2VAl/Fe bilayers with a stacking structure of Ta (3 nm)/Fe (3 nm)/Mn2VAl (100 nm)//MgO(100) substrate were also prepared for EB measurements. Tsub for the Mn2VAl layer was RT or 400ºC and that for Fe and Ta layers was RT. The bilayer samples were annealed at 200ºC with a magnetic field of 10 kOe in a vacuum furnace to induce an exchange anisotropy. Magnetization curves (M– H curves) were measured at RT and 10 K by using a vibrating sample magnetometer and a superconducting quantum interference device, respectively. Magnetic field was applied parallel to the <100> direction of the MgO substrate which corresponds to the <110> direction of epitaxially grown Mn2VAl and Fe layers. Neutron diffraction experiments were performed at BL18 SENJU, installed at the Materials and Life Science Experimental Facility in the Japan Accelerator Research Complex (MLF/J- PARC) [21]. The measurements were made in two sample settings where the epitaxial (cid:1) (cid:1) axis was parallel and perpendicular to the incident neutron beam to observe reflections in the (hhl) plane and the (hk0) plane of Mn2VAl, respectively. III. RESULTS AND DISCUSSION A. Structural properties Out-of-plane XRD patterns of the Mn2VAl films are shown in Figure 1(a). All samples show 004 fundamental reflection and the samples deposited at 500 and 600ºC show 002 superlattice reflections. The Φ scan results for Tsub = 500 and 600ºC are shown in Figure 1(b) and 1(c), respectively. Both samples exhibit the (111) superlattice reflections with four-fold symmetry. These results indicate that the samples deposited at 500 and 600ºC are L21-type Mn2VAl and the other samples are A2-type Mn2VAl. Note that all samples show the fundamental (220) reflections with four-fold symmetry representing an epitaxial growth of the Mn2VAl layer. Lattice parameters along a- and c-axes calculated from 004 and 022 reflections are plotted in Figure 1(d) as a function of Tsub. The value of a bulk sample [10] is also shown for comparison. The in-plane lattice parameter of Mn2VAl is expanded probably due to the lattice mismatch of 1.2% with the MgO(100) substrate. For Tsub = 700ºC and 800℃, both lattice parameters are larger than those of the bulk sample, which may be attributed to a change of the film composition: The compositions of the Mn2VAl layer for Tsub = 700 and 800ºC are Mn35.5V32.3Al32.2 and Mn21.9V39.1Al39.0, respectively, whereas the stoichiometry of the Mn2VAl layer is maintained for Tsub ≤ 600ºC. This indicates that a high Tsub leads to Mn sublimation, and such a composition change was also reported in a previous paper by Meinert et al. [14]. Long-range order parameters of Mn2VAl for B2 and L21 phases, SB2 and SL21, are (cid:1) (cid:1) plotted in Figure 1(e). SB2 and SL21 are estimated using the following equations: SB2 ={[I(002) exp /I(004) exp ]/[I(002) sim /I(004) sim ]} 1/2, (1) SL21 ={[I(111) exp /I(022) exp ]/[I(111) sim /I(022) sim ]} 1/2, (2) where I(hkl)exp(sim) represents the experimental (simulated) integrated intensity of a (hkl) reflection. For Tsub ≤ 400ºC, the samples are A2 phase. Both SB2 and SL21 are more than 0.6 for Tsub = 500 and 600ºC. For Tsub over ≥ 700ºC, the samples become another phase due to the Mn sublimation. FIG. 1. Structural properties of 100 nm-thick Mn2VAl films. (a) Out-of-plane XRD (cid:1) (cid:1) patterns, Φ scan results for (111) reflections of the Mn2VAl layer deposited at (b) 500 and (c) 600ºC, (d) the lattice parameters for in-plane (a-axis) and perpendicular-to-plane (c-axis) directions, and (d) order parameters for B2 phase (SB2) and L21 phase (SL21) as a function of the deposition temperature, Tsub. B. Magnetic properties of Mn2VAl films M–H curves of the Mn2VAl films measured at RT are shown in Figure 2. The maximum applied magnetic field was 20 kOe. No hysteresis is observed for Tsub less than 400ºC and more than 700ºC showing that A2-type Mn2VAl is a paramagnet or an antiferromagnet. On the other hand, the samples with Tsub = 500 and 600ºC exhibit hysteresis. While the trends in the Tsub dependence of the crystal structure and the saturation magnetization are similar to those reported in the previous paper [13], our samples show larger SB2, SL21, and Ms. This may be explained by the improved stoichiometry of the film composition. FIG. 2. M-H curves of 100 nm-thick Mn2VAl films measured at RT (cid:1) (cid:1) C. Neutron diffraction experiments In order to identify the magnetism of A2-type Mn2VAl, neutron diffraction experiments at RT were performed using 1-µm-thick Mn2VAl films with Tsub = RT, 400ºC, and 600ºC. Figures 3(a) – 3(c) show reflections in the (hhl) plane of the Mn2VAl films from which one can determine the chemical phases of the samples. Schematic illustrations of the expected reflections from A2 and L21 phases are shown in fig. 3(d), together with those from MgO. Fig 3(e) depicts the epitaxial relationship between Mn2VAl and MgO. Note that the broken lines in fig. 3(a)-3(c) and the solid lines in fig. 3(d) illustrate integer indices of MgO, while the broken lines in fig. 3(d) illustrate those of L21 phase. For Tsub = 600ºC, apart from strong spots from MgO, reflections with all-even and all- odd Miller indices with L21 notation were observed (blue arrows in fig. 3(a) and 3(d)), showing that the sample has L21 structure. For Tsub = 400ºC and RT, on the other hand, only reflections from the A2 phase (red arrows in fig. 3(b) and 3(c)) were confirmed, showing that these samples have the A2 structure. (cid:1) (cid:1) FIG. 3 Bragg reflections in the (hhl) planes of A2 and L21 phases. (a)-(c) Observed reflections for Mn2VAl films with Tsub = 600ºC, 400ºC, and RT. (d) Schematic illustration of the observed reflections indicated by blue and red arrows for L21 and A2 phases, respectively. (e) Epitaxial relationship between Mn2VAl and MgO. Figures 4(a) – 4(c) show reflections in the (hk0) plane of the Mn2VAl films. Schematic illustrations of the expected reflections are shown in fig. 4(d) and 4(e) in which blue and red arrows indicate observed reflections. A reflection corresponding to 100 in A2 phase or 200 in L21 phase is observed for Tsub = 600ºC and 400ºC, whereas no reflection is observed for Tsub = RT. Note that, due to the absence of strong reflections from MgO in this experimental setting, we can exclude a possibility that the reflection is multiple scattering of MgO. A preliminary powder neutron diffraction experiment for bulk (cid:1) (cid:1) Mn2VAl at high temperatures has confirmed that 100 reflection in A2 phase at room temperature originates from antiferromagnetic (AFM) ordering [16]. Thus, along with the discussion in the last paragraph, we conclude that 1) Mn2VAl film deposited at 600ºC is L21 phase, 2) one deposited at 400ºC is A2 phase with AFM ordering, and 3) one deposited at RT is A2 phase without AFM ordering. A summary of the neutron experiments is shown in Table 1. FIG. 4 Bragg reflections in the (hk0) planes of A2 and L21 phases. (a)-(c) Observed reflections for Mn2VAl films with Tsub = 600ºC, 400ºC, and RT. (d) Schematic illustration of the observed reflections indicated by blue and red arrows for L21 and AFM A2 phases, respectively. (cid:1) (cid:1) TABLE 1. A summary of the neutron experiments on Mn2VAl films. Deposition temperature Observed reflections Chemical phase and magnetic structure Tsub RT 400ºC 600ºC L21 reflections in Fig. 3 L21 200 or A2 100 (AFM) in Fig. 4 No No Yes No Yes Yes A2 para A2 AFM L21 D. Exchange bias in Mn2VAl/Fe bilayers EB effects in Mn2VAl/Fe bilayer samples were evaluated using A2-type Mn2VAl films with Tsub = RT and 400℃. Both bilayer samples showed no EB shift at RT. M-H curves measured at 10 K are shown in Figure 5. Samples were cooled down to 10 K from 300 K under a magnetic field of 10 kOe before starting measurements. At 10 K, no EB shift appears for Tsub = RT, whereas a shift of 120 Oe appears for Tsub = 400ºC. Although the blocking temperature is low and the shift is too small for device application at this moment, EB has been successfully demonstrated using the newly developed AFM Mn2VAl film. (cid:1) (cid:1) FIG. 5. M-H curves of A2-type Mn2VAl/Fe bilayer samples measured at 10 K. The deposition temperatures, Tsub, for the Mn2VAl layer are (a) RT and (b) 400ºC. IV. CONCLUSIONS The structural and magnetic properties of epitaxially grown Mn2VAl films were investigated. The chemical phases of the films were successfully controlled by changing Tsub. The maximum saturation magnetization of 220 emu/cm3 was achieved in a film with L21 phase, which is larger than a reported value probably due to the improved chemical ordering with better stoichiometry. An A2-type Mn2VAl film deposited at 400ºC showed magnetic superlattice reflection associated with the AFM phase in the neutron diffraction experiments at room temperature, and a Mn2VAl/Fe bilayer sample with an AFM A2-type Mn2VAl layer deposited at 400ºC showed an exchange bias shift of 120 Oe at 10 K. ACKNOWLEDGEMENTS This work was partially supported by Heusler Alloy Replacement for Iridium (HARFIR) under the Strategic International Cooperative Research Program (SICORP) (cid:1) (cid:1) from Japan Science and Technology Agency (JST). A part of this work was done by a cooperative program (No. 16G0407) of the CRDAM-IMR, Tohoku University. Authors would like to thank Dr. R. Y. Umetsu for valuable discussion, and TT and TK would like to thank Mr. I. Narita for his technical support. TT acknowledges a support from the Graduate Program of Spintronics (GP-Spin) at Tohoku University. RK acknowledges a support from the Leading Graduates Schools Program, "Interdepartmental Doctoral Degree Program for Multi-dimensional Materials Science Leaders" by the Ministry of Education, Culture, Sports, Science and Technology. The neutron experiment at the Materials and Life Science Experimental Facility of the J- PARC was performed under the Elements Strategy Proposal (No. 2015E0003). (cid:1) (cid:1) (References) 1 W.H. Meiklejohn and C.P. Bean, Phys. Rev. 102, 1413 (1956). 2 B. Dieny, V.S. Speriosu, S.S.P. Parkin, B.A. Gurney, D.R. Wilhoit, and D. Mauri, Phys. Rev. B 43 1297 (1991). 3 A. Kohn, A. Kova, R. Fan, G. J. McIntyre, R. C. C. Ward, and J. P. Goff, Sci. Rep. 3, 2412 (2013). 4 A. J. Devasahayam, P. J. Sides, and M. H. Kryder, J. Appl. Phys., 83, 7216 (1998). 5 M. Pakala, Y. Huai, G. Anderson, and L. Miloslavsky, J. Appl. Phys., 87, 6653 (2000). 6 L. Fernandez-Outon, K. O'Grady, S. Oh, M. Zhou, and M. Pakala, IEEE Trans. Magn. 44, 2824 (2008). 7 K. Komagaki, K. Yamada, K. Noma, H. Kanai, K. Kobayashi, Y. Uehara, M. Tsunoda, and M. Takahashi. IEEE Trans. Magn. 43, 3535 (2007). 8 J. van Driel, F. R. de Boer, K. M. H. Lenssen, and R. Coehoorn, J. Appl. Phys., 88, 975 (2000). 9 See www.webelements.com for WebElements Periodic Table. 10 K.H.J. Buschow and P.G. van Engen, J. Magn. Magn. Mater, 25, 90 (1981). 11 C. Jiang, M. Venkatesan, and J.M.D. Coey, Solid State Commun., 118, 513 (2001). 12 B. Deka, A. Srinivasan, R.K. Singh, B.S.D.Ch.S. Varaprasad, Y.K. Takahashi, and K. Hono, Journal of Alloys and Compounds, 662, 510 (2016). 13 T. Kubota, K. Kodama, T. Nakamura, Y. Sakuraba, M. Oogane, and K. Takanashi, Y. Ando, Appl. Phys. Lett. 95, 222503 (2009). 14 P. Klaer, E. Arbelo Jorge, M. Jourdan, W. H. Wang, H. Sukegawa, K. Inomata, and H. J. Elmers, Phys. Rev. B, 82, 024418 (2010). (cid:1) (cid:1) 15 M. Meinert, J. M. Schmalhorst, G. Reiss, and E. Arenholz, J. Phys. D. Appl. Phys. 44, 215003 (2011). 16 R. Y. Umetsu et. al., in preparation. 17 S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Nat. Mater. 3, 868 (2004). 18 S. S. P. Parkin, C. Kaiser, A. Panchula, P. M. Rice, B. Hughes, M. Samant, and S.-H. Yang, Nat. Mater. 3, 862 (2004). 19 Y. Sakuraba, M. Ueda, Y. Miura, K. Sato, S. Bosu, K. Saito, M. Shirai, T. J. Konno, and K. Takanashi, Appl. Phys. Lett. 101, 252408 (2012). 20 H.-x. Liu, T. Kawami, K. Moges, T. Uemura, M. Yamamoto, F. Shi, and P. M. Voyles, J. Phys. D: Appl. Phys. 48, 164001 (2015). 21 T. Ohhara, R. Kiyanagi, K. Oikawa, K. Kaneko, T. Kawasaki, I. Tamura, A. Nakao, T. Hanashima, K. Munakata, T. Moyoshi, T. Kuroda, H. Kimura, T. Sakakura, C. H. Lee, M. Takahashi, K. Ohshima, T. Kiyotani, Y. Noda, and M. Arai, J. Appl. Cryst. 49, 120 (2016). (cid:1)
1910.13774	1	1910	2019-10-30T11:11:10	Chemical-Vapor-Deposited Graphene as a Thermally Conducting Coating	[ "physics.app-ph", "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	We performed Scanning Thermal Microscopy measurements on single layers of chemical-vapor-deposited -CVD- graphene supported by different substrates, namely SiO2, Al2O3 and PET using a double-scan technique to remove the contribution to the heat flux through the air and the cantilever. Then, by adopting a simple lumped-elements model, we developed a new method that allows determining, through a multi-step numerical analysis, the equivalent thermal properties of thermally conductive coatings of nanometric thickness. In this specific case we found that our CVD graphene is thermally equivalent, for heat injection perpendicular to the graphene planes, to a coating material of conductivity Keff=2.5+-0.3 mK and thickness teff=3.5+-0.3 nm in perfect contact with the substrate. For the SiO2 substrate, we also measured stacks made of 2 and 4 CVD monolayers and we found that the effective thermal conductivity increases with increasing number of layers and, with a technologically achievable number of layers, is expected to be comparable to that of one order of magnitude-thicker metallic thin films. This study provides a powerful method for characterizing the thermal properties of graphene in view of several thermal management applications.	physics.app-ph	physics	Chemical-Vapor-Deposited Graphene as a Thermally Conducting Coating Mauro Tortello,1, Iwona Pasternak2, Klaudia Zeranska-Chudek2, Wlodek Strupinski2, Renato S. Gonnelli1, Alberto Fina1 1. Dipartimento di Scienza Applicata e Tecnologia, Politecnico di Torino, 10129 Torino, Italy 2. Faculty of Physics, Warsaw University of Technology, Koszykowa 75, 00-662 Warsaw Poland We performed Scanning Thermal Microscopy measurements on single layers of chemical-vapor-deposited (CVD) graphene supported by different substrates, namely SiO2, Al2O3 and PET using a double-scan technique to remove the contribution to the heat flux through the air and the cantilever. Then, by adopting a simple lumped-elements model, we developed a new method that allows determining, through a multi-step numerical analysis, the equivalent thermal properties of thermally conductive coatings of nanometric thickness. In this specific case we found that our CVD graphene is "thermally equivalent", for heat injection perpendicular to the graphene planes, to a coating material of conductivity 𝑘𝑒𝑓𝑓 = 2.5 ± 0.3 𝑊 𝑚𝐾 and thickness 𝑡𝑒𝑓𝑓 = 3.5 ± 0.3 𝑛𝑚 in perfect contact with the substrate. For the SiO2 substrate, we also measured stacks made of 2- and 4- CVD monolayers and we found that the effective thermal conductivity increases with increasing number of layers and, with a technologically achievable number of layers, is expected to be comparable to that of one order of ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ magnitude-thicker metallic thin films. This study provides a powerful method for characterizing the thermal properties of graphene in view of several thermal management applications. KEYWORDS: Graphene, thermal conductivity, Scanning Thermal Microscopy, 2D materials, thermally conductive coating INTRODUCTION It is known that the remarkable electrical [1-3] and thermal [4-7] properties of graphene can change considerably depending on its quality and on the specific system in which graphene is employed. Indeed, the number of layers [8-10], amount of defects [11-15], coupling to the substrate [16,17], production method [18], presence of graphene-substrate adsorbate layer or water adlayers [19,20,21], etc., can give rise to different electrical and thermal properties and/or performances. For example, the exceptionally high thermal conductivity of suspended, mechanically exfoliated graphene decreases by one order of magnitude when it is supported by SiO2, due to the coupling of the flexural ZA vibrational modes to the substrate [22]. Moreover, the thermal conductivity of single layer graphene has also been shown to have a 30% to 50% reduction in an epoxy matrix [23]. Therefore, it is very important to evaluate and investigate the properties of graphene or graphene-related materials (but this consideration holds for all 2D materials) in the specific system in which they have to be employed. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ In the perspective of utilizing graphene in future (possibly flexible) electronics, it is very important to consider the thermal conductivity, heat generation and dissipation of supported (rather than suspended) graphene and its interaction with different substrates, since the performance of electronic devices considerably depends on the temperature [24]. For applications like thermally conductive nanocomposites [25,26], thermal interface materials [27,28,29], thermally conductive coatings for plastic materials [30] and innovative heat spreaders [27,28,31], the interaction between graphene and oxides (like SiO2), metals or polymers can be crucial. Furthermore, the investigation of the thermal conductivity properties of CVD graphene is much more relevant to applications compared to exfoliated graphene, as large-scale CVD processes are currently available and exploited for thin film industrial applications [32]. Scanning Thermal Microscopy (SThM) [33,34] is a powerful technique for investigating the thermal properties at the nanoscale. Despite this technique hardly provides a quantitative determination of the thermal conductivity of the sample [34,35], SThM has an unmatched spatial resolution (a few tens of nanometers or less), which cannot be achieved by other popular methods such as the Raman optothermal technique [10] or by electrical methods [19]. By performing SThM measurements, Pumarol et al. [36] showed that the heat transport in suspended exfoliated graphene is higher than for the supported one and that the thermal conductance per single layer in a 3-layer graphene is about 68% of that of supported single layer graphene. Menges et al. [37] measured single and multilayer graphene supported by SiO2 or crystalline SiC and claimed a sub-10 nm lateral resolution with a thickness sensitivity to the single atomic layer. Furthermore, they observed a decrease of the thermal resistance with increasing number of layers for SiO2-supported, mechanically exfoliated graphene. A 30 nm spatial resolution was reported by Tovee et al. on few-layer graphene by using carbon nanotube tipped thermal probes [38]. K. Yoon et al. [39] quantitatively determined the thermal conductivity of suspended graphene by using the so called null-point SThM that employs a thermocouple as the thermal probe. In this ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ work and in others [40], however, the authors do not usually report thermal maps but only line scans. Tortello et al. reported on the thermal properties of pristine and annealed reduced graphite oxide flakes [35]: a correlation between the reduction of structure defectiveness consequent to annealing and improved thermal properties was demonstrated by SThM measurements on the single flakes. To the best of our knowledge, no SThM studies of graphene grown by chemical vapor deposition (CVD) were previously reported, despite this is currently the best candidate for large-scale production of graphene-based devices, since mechanical exfoliation, that gives the best samples in terms of quality, is certainly not viable in this regard. Here we show SThM results on CVD graphene (1GRL) supported by different substrates i.e. SiO2, Polyethylene terephthalate (PET) and Al2O3. For the SiO2 substrate we also measured samples with 2 (2GRL) and 4 (4GRL) CVD graphene layers stacked one on top of the other (random stacking). EXPERIMENTAL SECTION Scheme 1. a: Temperature vs time diagram of the CVD graphene growth process as described in [41]. b: Sequence of the steps for the marker-frame method used for the transfer of CVD graphene on the different substrates [44]. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ The graphene films were grown by chemical vapor deposition (CVD) on top of 25 m-thick copper substrates, as described in Ciuk et al. [41]. The temperature vs time diagram followed for the growth process is reported in Scheme 1a. Bi or tri-layers of graphene on the original graphene film are usually observed as 1-2 μm hexagons or dendrites scattered on the surface. These layers are presumed to grow underneath the first layer at the same copper active site (impurities) as the first layer [42,43]. These areas can be seen as dark spots in SEM images or as bright spots in optical images. We avoided these regions during SThM measurements, as it will be shown later. The graphene films were then transferred to different substrates by using a special marker-frame method (Scheme 1b) that does not make use of polymers like PMMA or PDMS, thus avoiding leaving polymer residues. [44]. Moreover, this method allows transferring the graphene films on almost any substrate, since there is no need of using dissolving agents, like e.g. acetone, normally employed for removing polymers. Three different substrates were adopted, PET, 285nm silicon dioxide grown on silicon (SiO2/Si) and alumina (Al2O3). The SiO2 substrate was a dry thermal oxide while the Al2O3 one was monocrystalline Epi-ready sapphire. On each of them, we transferred 1 graphene layer (1GRL). In the case of SiO2/Si substrate, we also prepared samples with two (2GRL) and four (4GRL) layers. The different substrates were chosen to span in thermal conductivity by two orders of magnitude (𝑘𝑃𝐸𝑇 = 0.2 𝑊 𝑚𝐾 , 𝑘𝑆𝑖𝑂2 = 1.4 𝑊 𝑚𝐾 and 𝑘𝐴𝑙2𝑂3 = 15 𝑊 𝑚𝐾 ). The samples were characterized by Raman spectroscopy using a Renishaw inVia system and a wavelength of 514 nm. It also worth ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ pointing out here that, unless the samples are prepared in dry conditions (which is not the case here), it has been shown that there is a ubiquitous graphene-substrate adsorbate layer [19,20,21] that will tend to make the interface properties similar among different substrates. For this reason, we will later assume that, to a first approximation, the thermal contact resistance between the graphene and the substrate is the same for all the substrates. Scanning Thermal Microscopy (SThM) measurements were performed on an Innova atomic force microscope (AFM) from Bruker, equipped with a VITA module for the thermal measurements. For the SThM measurements we adopted state-of-the-art resistive probes (Bruker VITA HE-GLA) in which a thin Pd film is deposited near the silicon nitride probe apex. The thin film acts at the same time as the heater and the temperature sensor and is part of a Wheatstone bridge. Before the measurements, the resistance of the probe is first measured (by means of an Agilent 34420A nano-voltmeter) at a low current value, i.e. 100 A, to avoid Joule heating and subsequently at a higher value (1 mA) at which the probe is heated. This is necessary to obtain the value of the resistance because it can slightly change over time (days) of repeated measurements. Then, the measured value is compared to that obtained by using the standard Wheatstone bridge formula that requires, as the input, the bridge voltage provided by the instrument software. This operation is necessary to check that the formula is providing the correct resistance value, since these values in the SThM measurements will be obtained through the mentioned procedure. The heating effect due to the laser is also considered by repeating the procedure first with the laser off and then with the laser turned on. The thermal scans are then performed by applying a current of about 1.3-1.4 mA, since higher values are likely to alter the resistance or even damage the probe. Then, after a thermal map has been acquired, the bridge voltage is converted to a resistance value and the resistance is converted to temperature by using the temperature coefficient of the ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ probes, that we measured to be 8.92 × 10−4𝐾−1, similar to that reported for Palladium [48] but lower than the one measured on the older generation of probes, made of silicon dioxide [49]. The temperature increase due to the laser is normally of about 0.8-1.2 K. The probe is formed by two NiCr "legs" resistors deposited on the cantilever and by the heater part formed by the Pd resistor at the tip apex. Indeed, since the temperature coefficient of Pd is one order of magnitude higher than that of NiCr while their electrical resistances are comparable (around 100  each), we assumed that the temperature coefficient of the resistive part close to the apex is that of the whole probe. This is confirmed by the fact that the total temperature coefficient that we determined differs by less than 5% from that of pure Pd. Thus, we can, to a good extent, consider that most of the temperature variation is occurring at the tip apex that is also hotter than the rest of the probe. Therefore, in the following we will consider that the resistive sensor is localized only at the tip apex. The SThM tips that we employed are state-of-the-art microfabricated probes. We think it is not yet technologically possible to obtain this kind of probes with a higher aspect ratio together with the required fabrication repeatability (especially considering the presence of the Pd resistive film deposited on the tip apex). To the best of our knowledge, a better resolution has been claimed for the silicon probes [37], but the heater is farther from the sample and our AFM has been optimized for the Pd probes that we adopted. Another possibility to enhance the resolution, could be to attach a carbon nanotube to the probe, as it has been done by Tovee et al. [38]. This would be interesting but rather beyond the scope of this paper where we are more interested in a reliable method for determining the thermal properties of 2D materials for heat injection along the cross-plane direction. In the SThM measurements, a lower temperature of the sensor means that a higher heat flux is transferred from the probe to the sample with respect to a region where the temperature is higher. The average temperature in a certain region is obtained by applying a mask and by averaging the temperature of each pixel contained in the mask. The temperature difference between the substrate and the graphene is 𝑇𝑠𝑢𝑏 − 𝑇𝐺𝑅 = ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Δ𝑇. The temperature uncertainty on each mask, 𝛿𝑇 is determined by the standard deviation and the final uncertainty is determined by the propagation of the error on each temperature, i.e. 𝛿Δ𝑇 = √𝛿𝑇𝑠𝑢𝑏 2 + 𝛿𝑇𝐺𝑅 2 . From the instrumental point of view, the minimum resolution in the bridge voltage corresponds to a temperature variation of about 1 mK, which is however not corresponding to the actual achievable precision due to various sources of environmental noise (thermal, electrical etc.). Indeed, the uncertainty on the temperature determination on different areas of the sample will be of the order of some tens of mK. We also point out here that results similar to those obtained with the masking procedure can be obtained by applying a thresholding method in order to single out the flat areas of the sample in the same temperature range. Finally, by knowing the ambient temperature, T0 and the applied power, P (determined by the Joule-heating formula, P=RHI2) the maps of the total thermal resistance of the systems can also be obtained. The SThM measurements are performed in the contact mode and the topography and other typical signals of this mode, like the lateral force, can be recorded while at the same time acquiring the thermal maps. The lateral force was found to be very powerful for clearly distinguishing between the graphene and the substrate regions. RESULTS AND DISCUSSION i. 1, 2 and 4 layers supported by SiO2/Si Figure 1 shows Raman spectra of graphene layers transferred onto SiO2/Si substrates. Raman spectra indicate two prominent and characteristic G and 2D peaks which are the features confirming the presence of graphene. The disorder-related weak D peaks connected with defects are also present. For the spectrum marked as "1GRL", the observed narrow (with the full width at half maximum (FWHM) of 35 cm-1) and ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ symmetric Lorentzian lineshape of the 2D peak is a feature confirming the presence of predominantly single layer graphene [45]. For the "2GRL" and "4GRL" we note a broadening of the 2D band and a slight shift of its position. These observations confirm that the shape and frequency of the 2D band are sensitive to the number of graphene layers. Indeed, in the case of exfoliated graphene (with defined stacking order) they can be used to determine the exact number of layers [46]. However, regarding our experiments where the graphene layers were added one by one, the created multilayer stack is in random alignment configurations [47] and, therefore, it is not possible to determine the number of graphene layers by analyzing the 2D peak. Figure1. Raman spectra of 1GRL, 2GRL and 4GRL on SiO2/Si substrates. Figure 2a shows the topography map of 1GRL supported by a SiO2/Si substrate. The graphene is covering the lower-left half of the image, but it is hardly distinguishable from the substrate also owing to the negligible thickness of graphene as compared with the height of some impurities saturating the scale. The presence of several wrinkles in that region, however, approximately indicates where the monolayer is located. The origin of the wrinkles is twofold. First, graphene was transferred from a copper foil. It is well known that due to the ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 14001600180020002200240026002800Intensity [a.u.]4GRL2DGD2GRL1GRLRaman shift [cm-1]mismatch of the thermal expansion between graphene and copper, the graphene ripples [50]. Second, wrinkles might come from the method of graphene transfer. In the marker-frame method, the graphene almost freely floats on a water surface, and such fluctuations can foster graphene wrinkling. Additionally, the standard procedure of graphene transfer includes annealing at 300-400°C to flatten the ripples. Since we transferred graphene also on PET foil which is not resistant to those temperatures, in the case of our experiments we decided to skip this step and we kept the same conditions for all substrates. The lateral force signal (panel b), on the other hand, clearly and unambiguously shows the presence of the graphene layer, since the friction between the probe and the sample is very different for the graphene or the substrate. Panel c represents the corresponding thermal map. It is possible to see that the temperature of the sensor is lower when the probe is in contact with the graphene layer than when it is on the bare substrate. The temperature on the graphene is determined by the average temperature of the masked unwrinkled region (rectangle in panel c), while the temperature on the substrate is determined by a similar mask placed on the substrate (not shown). The temperature difference between the substrate and the graphene is 𝑇𝑠𝑢𝑏 − 𝑇𝐺𝑅 = Δ𝑇 = 92 ± 44 mK. This temperature difference indicates that a greater heat flux is present when the probe is on the graphene than when it is on the substrate. It is also worth noticing here that the temperature has to be determined on the flat areas of the samples, in order to avoid "topological artifacts" [51]. Indeed, when the probe is, for instance, on the top of a significantly higher and steep region (like the impurities that are shown in red color in the lower- right part of panel c), a lower heat flux is transferred to the sample (via conduction through the air) because the distance from the sample has increased with respect to a flat area and the sensor temperature increases. On the other hand, when the probe is inside a concave structure, air-mediated heat transfer contribution becomes higher, increasing the total heat dissipation and consequently decreasing the sensor temperature. In this regard, the small, higher temperature spot at the center of the mask of panel c was excluded from the average temperature calculation. By looking at the thermal maps, one might also wonder how the thermal ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ conductivity behaves at defects and, especially, at line defects and whether it is possible to resolve its behavior. In this regard, we expect of course a decrease of the thermal conduction properties at defects locations due to increased phonon scattering, but one of the experimental limitations will be the spatial resolution of the tip. The resolution of these probes is around 20-30 nm, thus not enough, in principle, to resolve a line defect, which occurs on a much smaller distance. It might nevertheless be possible that, while scanning over a line defect, a small increase of the temperature is detected. However, this experiment should necessarily be performed on graphene samples deposited on atomically flat substrates, e.g. h-BN. Indeed, for detecting a change in thermal conduction over such a small length scale we should get rid of all possible topological artefacts that might give an apparent temperature variation. Finally, we point out that the scanning direction should also be perpendicular to the line defect because the noise along the scanning direction is lower than between adjacent scan lines. This might help to observe a temperature increase along each scan line in the point where the tip passes over the defects. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Figure 2. a, b and c: Topography, lateral force and SThM maps of 1GRL supported by SiO2/Si substrate, respectively. d, e and f: The same as for a, b and c but for 2GRL. g, h and i: The same as for a, b and c but for 4GRL. Panel d reports the topography map of 2GRL on SiO2/Si. In this case it is easier to identify the graphene sample, mainly because of the presence of some impurities, especially located at its edge, related to residues of chemicals used in the graphene transfer process. Moreover, as for the case of the 1GRL sample, we can also notice here some wrinkles on its surface. A clear contrast is observed in the lateral force map (panel e) also showing that the surface of the sample is in this case less homogeneous and presents a few more irregularities compared to the 1GRL sample. The impurities are also very well highlighted in the thermal map (panel f) due to ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ the above-mentioned topological effects. However, several flat regions are present where the temperature can be reliably determined, as in the area indicated by the rectangular mask. By calculating the average temperature on a similar area on the substrate, we obtained for this sample Δ𝑇 = 111 ± 69 mK, which is slightly higher than that observed for the 1GRL sample. Panel g shows the results of 4GRL on SiO2/Si sample. The sample is characterized by several flat, tile-like areas, surrounded by wrinkles, rather noticeable. This morphology is even more clearly indicated by the lateral force image (panel h). These structures are rather pronounced and look very similar to those reported by Kretinin et al. [52] and might be related to inclusions of organic residues. However, since the SThM probe is injecting the heat and measuring the temperature locally, their contribution to the thermal conduction is confined to the defective regions and their effect may easily be excluded by the proper selection of the analysis areas. Panel k reports the temperature map where we can see, at the same time, a clear temperature contrast between the flat areas and the substrate and the presence, as expected, of high-temperature regions in correspondence of the folds. The temperature contrast obtained in this case is Δ𝑇 = 221 ± 65 mK, clearly higher than for the 1GRL and 2GRL case, indicating that a higher heat flux is dissipated from the tip through the sample. Finally, it is worth noticing that all the thermal maps shown here do not present any lower- temperature area with a size of 1 or 2 m, that could be compatible with the possible presence of bi- or tri- layer regions formed during the growth process. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Figure 3. Summary of the temperature difference 𝑇𝑠𝑢𝑏 − 𝑇𝐺𝑅 = Δ𝑇 between the sensor temperature with the probe on the substrate and on graphene, as a function of the number of graphene layers. The dashed line is only a guide to the eye. Figure 3 reports, for the three cases, a summary of the temperature contrasts obtained by scanning on different areas of the samples. Even though the Δ𝑇 values are affected by a significant experimental error band, a clear trend is visible where the temperature contrast increases with the number of layers. The average values are Δ𝑇 = 64 ± 27 𝑚𝐾, Δ𝑇 = 89 ± 19 𝑚𝐾 and Δ𝑇 = 220 ± 39 𝑚𝐾 for the 1GRL, 2GRL and 4GRL samples, respectively. To analyze the data and discuss the results, we adopt the simplest lumped-elements circuit model for the heat conduction in this system, in a similar way as reported in other works [35,34,53] and as shown in Figure 4. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 0123450100200300Tsubstrate-Tgraphene (mK)Number of Graphene layersFigure 4. a: Sketch of the SThM probe in contact with a supported graphene sample. The arrows represent the different heat conduction channels between the heater at temperature 𝑇𝐻 and the ambient temperature at 𝑇0. b: Equivalent lumped-elements circuit model for the heat conduction paths of the system sketched in a. ii. Lumped-elements model The thermal resistance is defined as 𝑅 = 𝑇𝐻−𝑇0 𝑄 where 𝑇𝐻 is the temperature of the hot region (i.e. the heater), 𝑇0 is the ambient temperature and 𝑄 is the heat flux between them. When the probe is on the graphene, the total thermal conductance can be written as 1 𝑅𝑡𝑜𝑡 = 1 𝑅𝑎/𝑐 + 1 𝑆𝑢𝑏 where 𝑅𝑎/𝑐 𝑅𝑡−𝑠+𝑅𝐺𝑅+𝑅𝐺𝑅−𝑆𝑢𝑏+𝑅𝑠𝑝 describes the heat dissipation from both the heater to the air and from the heater through the cantilever, 𝑅𝑡−𝑠 is the contact resistance between the tip and the sample, 𝑅𝐺𝑅 is the resistance of the graphene sample, 𝑅𝐺𝑅−𝑆𝑢𝑏 is the thermal boundary resistance between the substrate and graphene and 𝑅𝑠𝑝 𝑆𝑢𝑏 represents the ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ spreading resistance through the substrate. The different heat conduction paths are represented in the schematic of the probe shown in Figure 4 a. In the thermal maps reported in Figures 2 c, f and i the only difference is the number of graphene layers (1, 2 and 4, respectively). Therefore, the only quantity that changes from one case to the other is 𝑅𝐺𝑅. Since the temperature contrast between the substrate and the sample increases with increasing number of layers, as reported in Figure 3, 𝑅𝐺𝑅 decreases with increasing number of layers when passing from 1 to 4 layers. This result agrees with what has been reported for SThM measurements on exfoliated graphene [37]. Figure 5. a, b and c: Topography, lateral force and SThM maps of 1GRL supported by PET substrate, respectively. d, e and f: The same as for a, b and c but for an Al2O3 substrate. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Figure 5 a reports the topography image for the 1GRL supported by PET. Again, it is not easy to identify the graphene layer, but the presence of some wrinkles suggests that the right part of the area is covered by the 2D monolayer. Indeed, this is confirmed unambiguously by the lateral force map (panel b) that also in this case shows an evident difference between the graphene and the substrate. Panel c shows the thermal map where a clear temperature contrast between the graphene and the PET region can be observed, the second featuring a higher temperature. The temperature difference we obtained in this case is 𝑇𝑠𝑢𝑏 − 𝑇𝐺𝑅 = Δ𝑇 = 167 ± 64 mK, which is significantly higher than for the monolayer on the SiO2/Si substrate. An enhancement of the temperature contrast when passing from the SiO2/Si to the PET substrate was also observed in the case of SThM measurements of graphite nanoplates with thickness in the 4-15 nm range [35]. Panel d shows the topography of a graphene layer supported by Al2O3. The graphene is located at the right-hand side of the image, as confirmed by the lateral force map of panel e. As for the thermal map, a clear temperature contrast is observed also in this case but with a significant difference: unlike the previous cases, the sensor temperature is now higher when the probe is on the graphene than when it is on the substrate, with Δ𝑇 = −110 ± 32 mK. This is clearly related to the thermal conductivity of the substrate, which for alumina is approximately one order of magnitude higher than for SiO2. The change in sign of the temperature contrast, Δ𝑇 indicates that the heat flux is higher when the probe is on the Al2O3 than on graphene, which is now acting like a sort of thermal barrier or, in other words, thermally resistive coating. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Figure 6. Summary of the temperature difference 𝑇𝑠𝑢𝑏 − 𝑇𝐺𝑅 = Δ𝑇 between the sensor temperature with the probe on the substrate and on one graphene layer, as a function of the thermal conductivity of the substrate. The black line is a log fit of the type 𝑦 = 𝑎 ⋅ ln(𝑥) + 𝑏, where 𝑎 = −79.6 ± 4.6 𝑚𝐾 and 𝑏 = 77.3 ± 7.5 𝑚𝐾. The intercept at 𝑦 = 0 is 𝑥 = 2.6 ± 0.4 𝑊 𝑚𝐾 . The trend of Δ𝑇 as a function of the thermal conductivity of the substrate has been reproducibly observed by performing several measurements on different areas of each sample, as shown in Figure 6. This result indicates that the CVD graphene behaves as an ultrathin coating that improves heat dissipation on substrates whose thermal conductivity is equal or lower than that of SiO2 (𝑘𝑆𝑖𝑂2 = 1.4 𝑊 𝑚𝐾 ) while it behaves as a thin thermal barrier for more thermally conducting substrates. The line reported in Figure 6, is a logarithmic fit of the type 𝑦 = 𝑎 ⋅ ln(𝑥) + 𝑏 which intersects the Δ𝑇 = 0 value at 𝑘𝑠𝑢𝑏 = 2.6 ± 0.4 𝑊 𝑚𝐾 . This is the simplest functional form that fits the data in this range and its physical meaning has to be investigated further. However, we do not expect it to have a wide range of validity, especially at higher conductivity values. With ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 0.1110100-200-1000100200300T<0Al2O3SiO2Tsubstrate-Tgraphene (mK)Substrate thermal conductivity (W m-1K-1)PETT>0increasing values of the substrate thermal conductivity, the thermal spreading resistance of the system will decrease. Indeed, it has been shown [53] that for high values of the sample thermal conductivity the SThM tip is expected to progressively decrease its sensitivity. For example, in the case of a single isotropic sample, it will not be possible to distinguish thermal conductivity values above some tens of W/mK because the thermal resistance of the sample will be negligible with respect to that of the tip-sample contact (the two resistances are in series). iii. Thermal resistance maps and double-scan technique To make a more quantitative analysis, it is convenient to report the thermal maps in terms of the thermal resistance rather than of the temperature. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Figure 7. a: Total thermal resistance, 𝑅𝑡𝑜𝑡 for the SThM probe in contact with 1GRL supported by SiO2/Si substrate. b: Thermal resistance,𝑅𝑎/𝑐 obtained by a backward scan in the lift mode (tip close to the sample but not in contact), thus including only the heat dissipation through the air and the cantilever. c: Map of 𝑅′′, obtained by the maps a and b and by applying to each pixel the formula: 1 𝑅𝑡𝑜𝑡 = 1 𝑅𝑎/𝑐 + 1 𝑅′′, thus including only the tip-sample heat conduction. d, e, f and g, h, i: The same as in a, b, and c but for 4GRL supported by SiO2/Si substrate and 1GRL supported by PET, respectively. Since the resistance of the heater is known, we can calculate the heating power 𝑄 by using the Joule effect formula. From that, we can obtain the total thermal resistance of the system as 𝑅𝑡𝑜𝑡 = 𝑇𝐻−𝑇0 𝑄 . 𝑅𝑡𝑜𝑡 can be expressed by the equation of the lumped elements circuit shown in Figure 4 b. We have seen that the circuit is ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ represented by the parallel of two resistances: 𝑅𝑎/𝑐 (that describes the contribution of heat conduction through the air and the cantilever) and the series 𝑅𝑡−𝑠 + 𝑅𝐺𝑅 + 𝑅𝐺𝑅−𝑆𝑢𝑏 + 𝑅𝑠𝑝 𝑆𝑢𝑏 that we can call, for simplicity, 𝑅′′. 𝑅′′ describes the heat conduction that occurs directly through the tip-sample channel and is present only when the probe is in contact with the sample. Thus, if the tip is very close to the sample but not touching it, the only contribution to the heat conduction will be, as a first approximation, given by 𝑅𝑎/𝑐 only. In the light of this observation, we performed double scans by using the lift-mode technique. In the lift-mode scan, the forward trace is recorded with the tip in contact to the sample while the backward trace is obtained with the probe lifted to a certain height. This procedure is similar to that reported by Kim et al. [40] where, however, only line scans were performed instead of entire thermal maps as it is shown here. Different lift heights were explored, and we found that the optimal one is around 250 nm. Indeed, for lower lift heights the tip starts touching the sample during the backward scan due to the tip-sample electrostatic interaction, thus hindering the possibility of obtaining a clean map of 𝑅𝑎/𝑐. On the other hand, for higher lift heights, 𝑅𝑎/𝑐 is overestimated due to the excessive distance from the sample. From the height of 250 nm going down towards the contact, the tip-sample air transfer will still give a contribution, but it can be seen by performing retract measurements (see Supplementary Information file for more details) that this additional contribution is small compared to the total one. The retract measurements also confirmed that 250 nm is the minimum distance achievable from the experimental point of view to overcome electrostatic attraction of the probe to the sample. Figure 7 a reports the map of the total thermal resistance 𝑅𝑡𝑜𝑡 for the 1GRL sample supported by SiO2/Si. This map has been obtained by a forward scan, i.e. with the tip in contact with the sample. The graphene is visible mostly on the left and right side of the image, while the substrate corresponds to the flat central region. As expected, when scanning on the flat areas of the graphene, the probe features a lower thermal resistance than when it is on the SiO2/Si substrate. Higher resistance values are obtained in correspondence of folds and impurities. Figure 7 b shows the map for 𝑅𝑎/𝑐 obtained from the backward scan in ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ the lift mode. This is the thermal signal that has been obtained when the probe is not in contact with the sample. The signal is obviously more blurred than before, but it is still possible to distinguish the most prominent topological features of the sample. This fact indicates that, as expected, the tip in this configuration is not only dissipating heat through the air and the cantilever, but that there is also an air-mediated heat transfer to the sample. This is exactly the contribution that we want to get rid of, in order to single out only the heat flux through the tip-sample contact. Then, since 1 𝑅𝑡𝑜𝑡 = 1 𝑅𝑎/𝑐 + 1 𝑅′′, it is possible to determine 𝑅′′ , simply by inverting this formula. By applying the above formula using each pixel of the maps of Figure 7a and b, we can calculate the map of 𝑅′′, shown in panel c. It is possible to notice that, since 𝑅𝑡𝑜𝑡 (panel a) and 𝑅𝑎/𝑐 (panel b) have the same order of magnitude, 𝑅′′ turns out to be about two orders of magnitude higher. This means that most of the heat generated at the heater is dissipated through the air and the cantilever. However, this does not hinder the capability of the probe to detect a clear temperature contrast when in contact with the sample. This fact is also confirmed by the much higher spatial resolution (a few tens of nm) that is achieved with the probe in contact than when it is lifted, as it can be seen by comparing panel a and b. On the other hand, it can also be shown that the spatial correlation of 𝑅′′ with the topographic signal is not improved with respect to 𝑅𝑡𝑜𝑡, but it is slightly lower (66.8% vs 68.2% in this case). This is due to the fact that the topological effects on the thermal maps will proportionally contribute more, as expected, to lower the correlation in the case of 𝑅′′ than for 𝑅𝑡𝑜𝑡, since these effects are, by definition, more relevant when the tip is in contact than when it is lifted. The value of 𝑅′′ in correspondence of the masked graphene region is (1.22 ± 0.04) × 107 𝐾 𝑊 while it is (1.28 ± 0.03) × 107 𝐾/𝑊 on the substrate. Panels d, e and f report the maps of 𝑅𝑡𝑜𝑡, 𝑅𝑎/𝑐 and 𝑅′′, respectively, for the 4GRL sample supported by SiO2/Si. Considerations like those of the previous case hold here as well. Now 𝑅′′ is (1.41 ± 0.08) × 107 𝐾 𝑊 when the probe scans in correspondence of the mask and (1.65 ± 0.09) × 107 𝐾 𝑊 when the probe is on the substrate. Panel g shows the 𝑅𝑡𝑜𝑡 map for the 1GRL sample ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ supported by PET. Darker areas with several wrinkles correspond to the graphene that is not continuous and features areas where the probe is in contact with PET (lighter areas). The 𝑅𝑎/𝑐 map is reported in panel h. Edges, wrinkles and other topological irregularities are mostly visible. Panel i shows the 𝑅′′ map where a clear contrast between graphene and PET can be noticed. 𝑅′′ is (1.18 ± 0.10) × 107 𝐾 𝑊 on graphene (masked area) and (1.69 ± 0.18) × 107 𝐾 𝑊 on PET. iv. Analysis of the results for the monolayer supported by different substrates As in the case of the temperature variations, Δ𝑇 (reported in Figure 6), also the thermal resistance decreases when passing from the substrate to the graphene in the case of the samples supported by PET and SiO2 while it is higher on the graphene than on the substrate in the case of the Al2O3 substrate. This fact suggests that a convenient way to look at this type of systems is to regard the graphene deposited on the substrate as an effective material of thermal conductivity 𝑘𝑒𝑓𝑓 (𝑘𝑆𝑖𝑂2 < 𝑘𝑒𝑓𝑓 < 𝑘𝐴𝑙2𝑂3) and thickness 𝑡𝑒𝑓𝑓 (𝑡𝑒𝑓𝑓 > 𝑡𝑔𝑟𝑎𝑝ℎ𝑒𝑛𝑒) in perfect contact with the substrate. The latter condition accounts for the graphene/substrate interface by increasing the thickness with respect to that of the bare graphene. 𝑡𝑒𝑓𝑓 is therefore determined (similarly to what was done by Menges et al. [37]) as 𝑡𝑒𝑓𝑓 = 𝑡𝑔𝑟𝑎𝑝ℎ𝑒𝑛𝑒 + 𝑘𝑒𝑓𝑓 ⋅ 𝑟𝑒𝑓𝑓, where 𝑟𝑒𝑓𝑓 is an effective thermal boundary resistance parameter and 𝑡𝑔𝑟𝑎𝑝ℎ𝑒𝑛𝑒 = 0.34 𝑛𝑚. The quantity 𝑘𝑒𝑓𝑓 ⋅ 𝑟𝑒𝑓𝑓 is also known as the Kapitza length. Since the thickness of the substrates is about 500 m, the system in our case is equivalent to a layer of thermal conductivity 𝑘𝑒𝑓𝑓 and thickness 𝑡𝑒𝑓𝑓 in perfect contact to an infinite half-plane of thermal conductivity 𝑘𝑠𝑢𝑏. The sum of the terms 𝑅𝐺𝑅 + 𝑅𝐺𝑅−𝑆𝑢𝑏 + 𝑅𝑠𝑝 𝑆𝑢𝑏 can thus be described by the formula for the spreading resistance of a "compound half-plane" that, in the isoflux conditions, is [54]: 𝑞 = 𝑅𝑠𝑝 𝜓𝑞 4𝑘𝑠𝑢𝑏𝑎 ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Where 𝑘𝑠𝑢𝑏 is the thermal conductivity of the substrate, 𝑎 the contact radius through which the heat is (1) injected and 𝜓𝑞 = 32 3𝜋2 ( 𝑘𝑠𝑢𝑏 𝑘𝑒𝑓𝑓 2 ) + 8 𝜋 [1 − ( 𝑘𝑠𝑢𝑏 𝑘𝑒𝑓𝑓 2 ∞ ) ] ⋅ ∫ 0 2(𝜁)𝑑𝜁 𝐽1 [1 + 𝑘𝑒𝑓𝑓/𝑘𝑠𝑢𝑏tanh (𝜁𝑡𝑒𝑓𝑓/𝑎)]𝜁2 (2) where 𝐽1 is the Bessel function of the first kind. 𝜓𝑞 is the dimensionless spreading resistance parameter that is defined as 𝜓𝑞 = 4𝑘𝑠𝑢𝑏𝑎𝑅𝑠𝑝 𝑞 and its expression comes from that of the area-averaged temperature rise of the heat source area, 𝑇 since the spreading resistance can be expressed by 𝑅𝑠𝑝 𝑞 = 𝑇 𝑞𝜋𝑎2, where q is the heat flux [54]. The isoflux condition has been chosen mainly for ease of calculation. However, it has been shown that the thermal spreading resistance in the isothermal conditions differs, at maximum, by 8 % [54]. Therefore 𝑹′′ = 𝑹𝒕−𝒔 + 𝑹𝒔𝒑 𝒒 = 𝒓𝒕𝒔 𝒒 𝝅𝒂𝟐 + 𝑹𝒔𝒑 (3) where 𝑟𝑡𝑠 is the interface resistance between the tip and the sample. Since in this model the heat "spreads" down into the sample through the contact area, it consequently accounts for the fact that heat transfer area between the graphene and the substrate is larger than the tip-sample contact radius, while the anisotropy of the graphene is embedded in the 𝑘𝑒𝑓𝑓 and 𝑡𝑒𝑓𝑓 parameters. To determine the characteristic parameters of the effective material, we make a couple of considerations: i) we assume that, in a single measurement, the contact area between the tip and the sample remains constant when passing from the graphene/substrate system to the bare substrate for that specific substrate. For example, the contact area for the tip on the graphene/SiO2 system is the same as for the tip on the SiO2 in the same measurement but it will be different ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ for the case of the PET and Al2O3 substrates. This is reasonable because, as it can be seen from the topographic AFM images, the graphene, being thin and bendable, follows to a very good approximation the topography of the underlying substrate; ii) since the contact between the tip and the sample is formed by several nanocontacts, i.e. it is a multi-asperity contact [34], we assume that 𝑟𝑡𝑠 is mainly determined by the morphology of the contact rather than the intrinsic properties of the two materials forming the contact. Therefore, it is kept constant when changing substrate. This is ascribed to the complex physical nature of the contact. Indeed, as shown in Gomes et al. [34], in the contact region the heat conduction occurs along several different channels: through mechanical contacts, water meniscus and ballistic conduction through the air. The determination of the unique set of the three 𝑘𝑒𝑓𝑓,𝑟𝑒𝑓𝑓 (or, equivalently, 𝑡𝑒𝑓𝑓) and 𝑟𝑡𝑠 values that reproduce the experimental results is a three-step process, that has been implemented by using a Matlab code. Step 1. In principle, given a suitable couple of 𝑘𝑒𝑓𝑓 and 𝑟𝑡𝑠 values, by using the model for 𝑅′′reported in equation (3) and by spanning over a wide range of contact radii 𝑎, we can find for a given substrate the set of 𝑟𝑒𝑓𝑓 values that give, separately, the correct experimental 𝑅𝑔𝑟/𝑠𝑢𝑏 ′′ and 𝑅𝑠𝑢𝑏 ′′ results. Then, we determine the only 𝑟𝑒𝑓𝑓 value that matches the experimental data with the same contact radius 𝑎 for both the tip/graphene/substrate and the tip/substrate contact. Indeed, we have two equations (one for 𝑅𝑔𝑟/𝑠𝑢𝑏 ′′ and one ′′ for 𝑅𝑠𝑢𝑏 ) and two unknown parameters (𝑎 and 𝑟𝑒𝑓𝑓). Step 2. We determine many 𝑟𝑒𝑓𝑓 values and, as a consequence, contact radius values, by spanning over a wide (with respect to all the possible realistic values) range of (𝑘𝑒𝑓𝑓, 𝑟𝑡𝑠) values. For each different substrate, the result is a surface determined by all the sets of three (𝑘𝑒𝑓𝑓, 𝑟𝑒𝑓𝑓,𝑟𝑡𝑠) parameters that match the experimental data for that given substrate. An example of these surfaces for the three substrates used in this work and for a specific set of experimental data is reported in Figure 8. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Step 3. Then, we find the intersection between the three surfaces (one for each substrate) to determine the unique (𝑘𝑒𝑓𝑓, 𝑟𝑒𝑓𝑓, 𝑟𝑡𝑠) set. The black line in Figure 8 represents the intersection between the surfaces related to the SiO2 and Al2O3 substrates, while the blue one is the intersection between those related to the SiO2 and PET ones. The intersection between the two lines is the unique set of the (𝑘𝑒𝑓𝑓, 𝑟𝑒𝑓𝑓, 𝑟𝑡𝑠) values for that specific set of experimental data. Once the 𝑟𝑡𝑠 value is determined, the three contact radii for each substrate are also consequently determined. To summarize, we have six different measurements and six unknown parameters: 𝑘𝑒𝑓𝑓, 𝑟𝑒𝑓𝑓, 𝑟𝑡𝑠, 𝑎𝑆𝑖𝑂2, 𝑎𝑃𝐸𝑇 and 𝑎𝐴𝑙2𝑂3. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Figure 8. (𝑘𝑒𝑓𝑓, 𝑟𝑒𝑓𝑓, 𝑟𝑡𝑠) surfaces representing the solutions, for each substrate and for a particular couple of experimental values of 𝑅𝑔𝑟/𝑠𝑢𝑏 ′′ and 𝑅𝑠𝑢𝑏 ′′ ,of equation (3) by imposing the same contact radius 𝑎. The experimental values used in these calculations are: 𝑅𝑔𝑟/𝑃𝐸𝑇 ′′ = 1.25 × 10−7 𝐾/𝑊, 𝑅𝑃𝐸𝑇 ′′ = 1.67 × 10−7 𝐾/𝑊, ′′ 𝑅𝑔𝑟/𝑆𝑖𝑂2 = 1.22 × 10−7 𝐾/𝑊, 𝑅𝑆𝑖𝑂2 ′′ = 1.28 × 10−7 𝐾/𝑊, 𝑅𝑔𝑟/𝐴𝑙2𝑂3 ′′ = 4.25 × 10−6 𝐾/𝑊and 𝑅𝐴𝑙2𝑂3 ′′ = 4.02 × 10−6 𝐾/𝑊. We performed the above procedure by using different sets of data corresponding to different masked areas on the samples and we obtained 𝑘𝑒𝑓𝑓 = 2.5 ± 0.3 𝑊/𝑚𝐾, 𝑡𝑒𝑓𝑓 = 3.5 ± 0.3 𝑛𝑚 and 𝑟𝑡𝑠 = (2.4 ± 0.6) × 10−8 𝐾𝑚2/𝑊. The corresponding 𝑟𝑒𝑓𝑓 values are 𝑟𝑒𝑓𝑓 = (1.2 ± 0.2) × 10−9𝐾𝑚2/𝑊. The contact radii were found to be in the 30-50 nm range for the SiO2 and Al2O3 substrates, while higher (about 80 nm) for PET. As expected, the 𝑘𝑒𝑓𝑓 values are between the thermal conductivity of SiO2 and Al2O3, but closer to that of SiO2 and the obtained value perfectly coincides, within the uncertainty bar, with that of the intersection between the fit line in Figure 6 and Δ𝑇 = 0, that was found to be 𝑥 = 2.6 ± 0.4 𝑊/𝑚𝐾, indicating that the fitting procedure could be a good method for a quick estimation of 𝑘𝑒𝑓𝑓. Moreover, it is worth pointing out here that this value is related to the heat injection perpendicular to the plane. Therefore, it should not be compared to the in-plane one for the supported graphene which can even be of the order of a few hundreds of 𝑊/𝑚𝐾 [19, 7]. As for the 𝑟𝑒𝑓𝑓 value which determines the effective thickness, 𝑡𝑒𝑓𝑓 of the graphene coating, it has the physical dimensions of a thermal boundary resistance. A comparison between this value and those reported for the thermal boundary resistance between graphene and different substrates [55,56] has some limitations because in our model the graphene and the interface form a single entity (indeed it would be problematic to define the c-axis thermal conductivity for a single graphene layer). Nevertheless, we can notice that the order of magnitude of 𝑟𝑒𝑓𝑓 is in the realistic range for the thermal boundary resistances [57] and that ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ the obtained value is very close to the range reported by ref. [55] for a graphene/SiO2 interface, but lower than others [56,58,59]. Values similar to ours have also been reported for the carbon nanotube (CNT)/SiO2 interface [60] and for the graphene/oil interface [61]. The thermal boundary resistance values for other carbon compounds like diamond [63], metallic single-wall CNTs [64] and graphite [10,65] are close to the upper bound of thermal resistances found for graphene, i.e. of the order of 10-8 Km2/W. It is also worth recalling here that the fact that 𝑟𝑒𝑓𝑓 and, consequently, 𝑡𝑒𝑓𝑓 is assumed to be constant on different substrates is the most severe assumption. However, we believe it is sensible in this case because, as stated in the beginning, the presence of a graphene-substrate adsorbate layer [19,20,21] caused by the wet conditions for the sample preparation will tend to make the interface properties similar among different substrates. Finally, we checked in particular that the contact radius for the Al2O3 case (that was found to be about 40 nm) is larger than the phonon mean free path, because the expression of eq. 1 is based on the diffusive heat conduction. We estimated the phonon mean free path, 𝑙𝑝ℎ from the formula Λ = 1 3 𝐶𝜌𝑣𝑙𝑝ℎ, where Λ is the thermal conductivity, 𝐶 is the specific heat, 𝜌 is the density, 𝑣 the sound velocity. The material properties were taken from the literature [62]. We obtained 𝑙𝑝ℎ ≅ 3.3 𝑛𝑚, much smaller than the obtained tip-sample contact radius. Even though the kinetic expression used here for the calculations might underestimate the mean free path by a factor of 4-5, the diffusive heat conduction conditions would be met anyway. v. Analysis of the results for 2 and 4 layers supported by SiO2. The two-layer and four-layer samples have been obtained by multiple transfer procedures of single CVD layers, i.e. each layer has been subsequently stacked one on top of the other. Therefore, their properties are expected to be quite different from those of the exfoliated bi- and four-layer graphene. In our model of graphene as a thermal coating in perfect contact with the substrate, the addition of one layer can be regarded as equivalent to the addition of one layer of the effective material with thermal conductivity 𝑘𝑒𝑓𝑓. The only ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ difference is that now, besides 𝑟𝑒𝑓𝑓, there is an additional interface parameter that describes also the interaction between different graphene layers and that we name 𝑟𝑚𝑙𝑔−𝑒𝑓𝑓. Therefore, the effective thickness of each additional layer after the first will in principle be different from that of the first one. The total effective thickness can thus be expressed as 𝑡𝑛−𝑒𝑓𝑓 = 𝑛 ⋅ 𝑡 + 𝑘𝑒𝑓𝑓 ⋅ [𝑟𝑒𝑓𝑓 + (𝑛 − 1) ⋅ 𝑟𝑚𝑙𝑔−𝑒𝑓𝑓] where 𝑛 is the number of stacked graphene layers. By using the 𝑘𝑒𝑓𝑓 values found for the monolayer case, we obtained 𝑡2−𝑒𝑓𝑓 = 7.6 ± 3.5 𝑛𝑚 (corresponding to 𝑟𝑚𝑙𝑔−𝑒𝑓𝑓 = (1.6 ± 1.5) × 10−9 𝐾 ⋅ 𝑚2/𝑊) and 𝑡4−𝑒𝑓𝑓 = 26 ± 12 𝑛𝑚 (corresponding to 𝑟𝑚𝑙𝑔−𝑒𝑓𝑓 = (3.0 ± 1.8) × 10−9 𝐾 ⋅ 𝑚2/𝑊). The results are reported in Figure 9. The error bars are rather large because these results have been obtained by averaging over many measurements obtained in different regions of the samples and therefore are affected by local inhomogeneities. By looking at the effective thickness per number of layers, 𝑡𝑛−𝑒𝑓𝑓/𝑛 (see inset to Figure 9), it is possible to notice that the stacking of the second graphene layer only slightly improves the heat conduction because 𝑡𝑛−𝑒𝑓𝑓/𝑛 for two layers (𝑡2−𝑒𝑓𝑓/2 = 3.8 ± 1.7 𝑛𝑚) is very close to that of a single layer (𝑡𝑒𝑓𝑓 = 3.5 ± 0.3 𝑛𝑚). On the other hand, when 4 layers are stacked, a noticeable improvement of the heat conduction can be noticed. In this case, 4 graphene layers are equivalent to about 7.4 effective material layers and 𝑡4−𝑒𝑓𝑓/4 = 6.5 ± 3 𝑛𝑚. It might seem counterintuitive that the heat dissipation improves when the effective thickness of the conductive coating increases, but it is worth recalling here that, since the substrate (SiO2) is less conducting than the coating material, an increase of the effective thickness of the conductive coating will decrease the total spreading resistance of the compound half-plane [54]. Furthermore, let us note that even though the graphene/graphene interface is expected to be more efficient than the graphene/substrate one [10,27], this improvement looks still rather weak in the case of 2 layers, where the interface between the second and first layer is most probably still influenced by the substrate. Then, when the number of layers increases to 4, the improvement is clear. Of course, an exfoliated bi- or four-layer sample is expected to dissipate much more, not only because of the intrinsic higher quality of the individual layers, but also because of the better thermal ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ interface between the different graphene planes due to the non-random stacking and to the absence of adsorbates between the planes. Figure 9. Main panel: Effective thickness of 1GRL, 2GRL and 4GRL supported by SiO2/Si as a function of the number of layers. Inset: The same as in the main panel but now the effective thickness is normalized to the number of layers. The increase of the effective thickness of the coating material in the case of the 2GRL and 4GRL samples is a possible way to model the decrease of the spreading resistance with increasing number of layers. Alternatively, 2GRL and 4GRL could of course be considered as materials with a different effective thermal conductivity, 𝑘2−𝑒𝑓𝑓 and 𝑘4−𝑒𝑓𝑓 (>𝑘𝑒𝑓𝑓), and their relevant effective thickness. For precisely obtaining these values, we should perform SThM measurements on these samples, supported by at least two other different substrates, like PET and Al2O3, but this is beyond the scope of the present work and is the subject of future analyses. At present, we can however safely determine a lower bound for 𝑘2−𝑒𝑓𝑓 and 𝑘4−𝑒𝑓𝑓, by using the experimental data for 2GRL and 4GRL supported by SiO2 and by conservatively supposing that Δ𝑇 for the 2GRL and 4GRL ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 01234505101520253035 tn-efftn-eff (nm)n (number of layers)012345012345678910 tn-eff / ntn-eff / n (nm) n (number of layers)supported by Al2O3 stays unchanged with respect to the 1GRL/Al2O3 sample, i.e. Δ𝑇2𝐺𝑅𝐿/𝐴𝑙2𝑂3 = Δ𝑇4𝐺𝑅𝐿/𝐴𝑙2𝑂3 = Δ𝑇1𝐺𝑅𝐿/𝐴𝑙2𝑂3. By connecting these values, we obtain the dashed blue and olive lines reported in Figure 10, respectively. The intercept is 3.3 𝑊/𝑚𝐾 (dashed blue line in Figure 10) for 2GRL and 5.7 𝑊/𝑚𝐾 for 4GRL (dashed olive line). In other words, by performing the same procedure shown for the 1GRL sample on the 2GRL and 4GRL ones, we would expect to obtain at least 𝑘2−𝑒𝑓𝑓 ≅ 3.3 𝑊/𝑚𝐾 and 𝑘4−𝑒𝑓𝑓 ≅ 5.7 𝑊/𝑚𝐾. The data are reported as symbols in the inset to Figure 10. It is interesting to compare this result with the work of Jang et al. [66], where the thermal properties of graphene encased in SiO2 have been studied for different number of layers. Contrary to what observed for suspended graphene [10], and like the results shown here, an increase of the thermal conductivity has been measured with increasing number of layers. There, the effect was ascribed to the presence of the oxide (on both sides of the sample) that suppresses that thermal conduction over a characteristic length. A quantitative comparison of the obtained thermal conductivity values is not possible because in that case the graphene was exfoliated, and the in-plane conduction was probed while we are here sensitive to an overall effective conductivity. However, a similar effect is very likely to occur here as well. The best fit of the data is obtained with a 2nd order polynomial fit (dashed red line). At about 10 stacked layers the conductivity turns out to be 𝑘𝑒𝑓𝑓 ≅ 20 𝑊/𝑚𝐾. However, since 𝑘𝑒𝑓𝑓 is expected to saturate with increasing number of layers, we also tried to fit our data with the model reported in equation (2) or ref. [66], in order to better estimate the expected trend of the data. In this model, we have three free parameters: the thermal conductivity for thin flakes, 𝑘0, the "bulk" thermal conductivity, 𝑘𝑏𝑢𝑙𝑘, and the characteristic penetration of the detrimental effects of the substrate, 𝛿. First, we impose, of course, 𝑘0 = 2.5 𝑊/𝑚𝐾. Then, since we observed experimentally that the conduction properties for the 4-layer sample are better than the 2- layer one, we conservatively limit the upper bound for 𝛿 to 3 layers. In this case we get that the thermal conductivity at 10 layers is about 15 W/mK and the "bulk value" is 30 W/mK. These values would be higher with a larger 𝛿. For example, if we allowed 𝛿 = 4, we would get a thermal conductivity of 17 W/mK for 10 ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ layers with a bulk value of almost 50 W/mK. Even though we don't have enough experimental information on this characteristic length, and the samples here are different from those of ref. [66], it should be kept in mind that the estimated 𝑘2−𝑒𝑓𝑓 and 𝑘4−𝑒𝑓𝑓 values have been obtained in the most conservative way and represent a lower bound for 𝑘𝑒𝑓𝑓. Figure 10. Main panel: Summary of the temperature difference 𝑇𝑠𝑢𝑏 − 𝑇𝐺𝑅 = Δ𝑇 between the sensor temperature with the probe on the substrate and on one (red), two (blue) and four (olive symbols) graphene layers, as a function of the thermal conductivity of the substrate. The black line is a log fit of the type 𝑦 = 𝑎 ⋅ ln(𝑥) + 𝑏 , where 𝑎 = −79.6 ± 4.6 𝑚𝐾 and 𝑏 = 77.3 ± 7.5 𝑚𝐾. The intercept at 𝑦 = 0 𝑚𝐾 is 𝑥 = 2.6 ± 0.4 𝑊/𝑚 ∙ 𝐾. Blue and olive dashed lines are analogous log fits connecting the average value of the 2GRL and 4GRL samples supported by SiO2, respectively to that of 1GRL supported by Al2O3 (see text for details). Inset: Calculated (full symbol) and estimated lower bound (open symbols) values of 𝑘𝑒𝑓𝑓 as a function of number of stacked graphene layers The red line is a 2nd-order polynomial fit to the data while the blue one is a fit performed by using equation (2) of ref [66]. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 0.1110100-200-100010020030004812510152025 keff (1GRL) keff (lower bound) 2nd order fit Reference 66keff (W/m K)n (number of layers)Equationy = Intercept + B1x^1 + B2x^2PlotBWeightNo WeightingIntercept1.96667 ± --B10.4 ± --B20.13333 ± --Residual Sum of Squares0R-Square(COD)1Adj. R-Square--Modelkt (User)Equationk0+(kb-k0)(1-((2d)/x)tanh(x/(2d)))PlotBkb28.02058 ± 57.40953d3 ± 3.95349Reduced Chi-Sqr0.06577R-Square(COD)0.98814Adj. R-Square0.97629 1GRL 2GRL 4GRLT<0Al2O3SiO2Tsubstrate-Tgraphene (mK)Substrate thermal conductivity (W m-1K-1)PETT>0Thus, with a thickness of a few nanometers, the effective thermal conductivity is expected to reach some tens of 𝑊/𝑚 ∙ 𝐾, which is comparable to that of thin films of metals like Al, Cu and Au, where the thermal conductivity can be 20% of the bulk value when the thickness is of the order of 100 nm [27]. This is the case, for example, of gold thin films deposited on etched Si with a thickness comparable to the electronic mean free path [67,68], while for Al thin films, we would expect 𝑘 ≅ 26 − 48 𝑊/𝑚 ∙ 𝐾 [27]. Finally, it is important to recall that these multilayer systems are highly anisotropic, with an in-plane conductivity that can be orders of magnitude higher than the out-of-plane one. Therefore, they can be very useful for achieving a high in-plane heat dissipation with a very small thickness of coating material. CONCLUSIONS In conclusion, we have reported on the first SThM results on CVD graphene supported by different substrates (SiO2, PET, Al2O3). For the SiO2 substrate, 2- and 4-layer samples were investigated as well. The SThM measurements were performed with a double-scan technique to get rid of the heat dissipation through the air and the cantilever. Then, by using a simple lumped-elements model for the probe/sample system, along with the expressions of the spreading resistance in a compound half plane, we developed a multi-step analysis that allows determining the effective thermal conductivity (and effective thickness) of thermally conductive coatings of nanometric thickness. In the specific study reported here, we have shown that the single CVD graphene layer behaves, for heat injection perpendicular to the graphene planes, as a thermal coating equivalent to an effective material of conductivity 𝑘𝑒𝑓𝑓 = 2.5 ± 0.3 𝑊/𝑚𝐾 and thickness 𝑡𝑒𝑓𝑓 = 3.5 ± 0.3 𝑛𝑚 in perfect contact with the substrate. It is thus conductive in the case of SiO2 and PET substrates (𝑘𝑒𝑓𝑓 > 𝑘𝑠𝑢𝑏) while it is resistive in the case of Al2O3 (𝑘𝑒𝑓𝑓 < 𝑘𝑠𝑢𝑏). We have also shown that the heat conduction properties improve with increasing number of layers on SiO2 and that, with a technologically achievable number of layers, ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ the effective thermal conductivity is expected to be comparable to that of some thin films of metals with a thickness one order of magnitude higher, thus confirming the interest for the application of the industrially viable CVD graphene sheets. This improvement is due to both the fact that with increasing number of layers the detrimental effect of the substrate decreases and that a thicker thermal coating deposited on a resistive substrate will reduce the total thermal spreading resistance. This new method is very helpful for determining the equivalent thermal coating properties of 2D materials and can be used for the design of applications for thermal management and heat dissipation in nanoelectronics devices and thermally conductive coatings. These results also show the importance of carefully determining and investigating the properties of graphene and graphene-related in the specific situations in which they are employed. Corresponding author e-mail: [email protected] Acknowledgements This work has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program, grant agreement 639495 -- INTHERM -- ERC-2014-STG and the European Union's Horizon 2020 Framework Program under Grant Graphene Core2 n°785219 Graphene Flagship. Author contributions M.T. and A.F. designed the research. I.P., K.Z.-C. and W.S. prepared and characterized the samples. M.T. performed the SThM measurements and elaborated the data. M.T. and R.S.G. developed the multi-step analysis method. M.T. mainly wrote the paper. All authors participated to the discussion of the results. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ REFERENCES 1. 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, 666-669 2. Novoselov, K. S.; Geim, A. K.; Morozov, S. V.; Jiang, D.; Katsnelson, M. I.; Grigorieva, I. V.; Dubonos, S. V.; Firsov, A. A. Two-Dimensional Gas of Massless Dirac Fermions In Graphene. Nature 2005, 438, 197-200. 3. Geim, A. K.; Novoselov, K. S. The Rise of Graphene. Nat. Mater. 2007, 6, 183-191. 4. Nika, D. L.; Ghosh, S.; Pokatilov, E. P.; Balandin, A. A. Lattice Thermal Conductivity of Graphene Flakes: Comparison with Bulk Graphite. Appl. Phys. Lett. 2009, 94, 203103. 5. Ghosh, S.; Calizo, I.; Teweldebrhan, D.; Pokatilov, E. P.; Nika, D. L.; Balandin, A. A.; Bao, W.; Miao, F.; Lau, C. N. Extremely High Thermal Conductivity of Graphene: Prospects for Thermal Management Applications in Nanoelectronic Circuits. Appl. Phys. Lett. 2008, 92, 151911. 6. Malekpour H.; Balandin, A. A. Raman‐Based Technique for Measuring Thermal Conductivity of Graphene and Related Materials. J. Raman Spectrosc. 2017, 49, 106-120. 7. Nika, D. L.; Balandin, A. A. Phonons and Thermal Transport in Graphene and Graphene-Based Materials. Rep. Prog. Phys. 2017, 80, 036502. 8. Castro Neto, A. H.; Guinea, F.; Peres, N. M. R.; Novoselov, K. S.; Geim, A. K. The Electronic Properties of Graphene, Rev. Mod. Phys. 2009, 81, 109-162. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 9. Gonnelli, R. S.; Piatti, E.; Sola, A.; Tortello, M.; Dolcini, F.; Galasso, S.; Nair, J. R.; Gerbaldi, C.; Cappelluti, E.; Bruna M.; Ferrari, A. C. Weak Localization in Electric-Double-Layer Gated Few-Layer Graphene, 2D Mater. 2017, 4, 035006. 10. Balandin, A. A.; Thermal Properties of Graphene and Nanostructured Carbon Materials, Nat. Mater. 2011, 10, 569-581. 11. Mortazavi, B.; Ahzi, S. Thermal Conductivity and Tensile Response of Defective Graphene: A Molecular Dynamics Study. Carbon 2013, 63, 460-470. 12. Denis, P. A.; Iribarne, F. Comparative Study of Defect Reactivity in Graphene, J. Phys. Chem. C 2013, 117, 19048-19055. 13. Rutter, G. M.; Crain, J. N.; Guisinger, N. P.; Li, T.; First, P. N.; Stroscio, J. A. Scattering and Interference in Epitaxial Graphene, Science 2007, 317 219-222. 14. Wang, Y.; Huang, Y.; Song, Y.; Zhang, X.; Ma, Y.; Liang, J.; Chen, Y. Room-Temperature Ferromagnetism of Graphene, Nano Lett. 2009, 9 , 220-224. 15. Malekpour, H.; Ramnani, P.; Srinivasan, S.; Balasubramanian, G.; Nika, D. L.; Mulchandani, A.; Lake, R.; Balandin, A. A. Thermal Conductivity of Suspended Graphene with Defects, Nanoscale, 2016, 8, 14608. 16. Ong Z.-Y.; Pop, E. Effect of Substrate Modes on Thermal Transport in Supported Graphene, Phys. Rev. B 2011, 84, 075471. 17. Chen, J.; Zhang G.; Li, B. Substrate Coupling Suppresses Size Dependence of Thermal Conductivity In Supported Graphene, Nanoscale, 2013, 5, 532-536. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 18. Bonaccorso, F.; Lombardo, A.; Hasan, T.; Sun, Z.; Colombo, L.; Ferrari, A. C. Production and Processing of Graphene and 2d Crystals, Materials Today 2012, 15, 564-589. 19. Shearer, C. J.; Slattery, A. D.; Stapleton, A. J.; Shapter, J. G.; Gibson, C. T. Accurate Thickness Measurement of Graphene. Nanotechnology 2016, 27, 125704. 20. Cao, P.; Xu, K.; Varghese, J. O.; Heath, J. R. Atomic Force Microscopy Characterization of Room- Temperature Adlayers of Small Organic Molecules through Graphene Templating. J. Am. Chem. Soc. 2011, 133, 2334 -- 2337. 21. Ochedowski, O.; Bussmann, B. K.; Schleberger, M. Graphene on Mica -- Intercalated Water Trapped for Life Sci. Rep. 2014, 4, 6003. 22. Seol, J. H.; Jo, I.; Moore, A. L.; Lindsay, L.; Aitken, Z. H.; Pettes, M.T.; Li, X.; Yao, Z.; Huang, R.; Broido, D.; Mingo, N.; Ruoff, R. S.; Shi, L. Two-Dimensional Phonon Transport in Supported Graphene, Science 2010, 328, 213-216. 23. Mortazavi, B.; Benzerara, O.; Meyer, H.; Bardon, J.; Ahzi, S. Combined Molecular Dynamics-Finite Element Multiscale Modeling of Thermal Conduction in Graphene Epoxy Nanocomposites, Carbon 2013, 60, 356-365. 24. Majumdar, A. Helping Chips to Keep Their Cool, Nat. Nanotechnol. 2009, 4, 214-215. 25. Chen, H. Y.; Ginzburg, V. V.; Yang, J.; Yang, Y. F.; Liu, W.; Huang, Y.; Du, L. B.; Chen, B. Thermal Conductivity of Polymer-Based Composites: Fundamentals and Applications. Progress in Polymer Science 2016, 59, 41-85. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 26. Su, Y.; Li, J. J.; Weng, G. J. Theory of Thermal Conductivity of Graphene-Polymer Nanocomposites with Interfacial Kapitza Resistance and Graphene-Graphene Contact Resistance. Carbon 2018, 137, 222-233. 27. Yan, Z.; Liu, G.; Khan, J. M.; Balandin, A. A. Graphene Quilts for Thermal Management of High-Power Gan Transistors, Nat. Commun. 2012, 3, 827. 28. Prasher, R. Graphene Spreads the Heat. Science 2010, 328, 186. 29. Prasher, R.; Chiu, C. P. Thermal Interface Materials. In Materials for Advanced Packaging, Second Edition; Lu, D.; Wong, C., Eds; Springer: Cham, 2017; 535-511. 30. Malekpour, H.; Chang, K.-H.; Chen, J.-C.; Lu, C.-Y.; Nika, D. L.; Novoselov, K. S.; Balandin, A. A. Thermal Conductivity of Graphene Laminate, Nano Lett. 2014, 14, 5155 -- 5161. 31. Balandin, A. A. Better Computing through CPU Cooling, IEEE Spectrum, October 2009, https://spectrum.ieee.org/semiconductors/materials/better-computing-through-cpu-cooling 32. Ferrari, A. C.; et al. Science and Technology Roadmap for Graphene, Related Two-Dimensional Crystals, and Hybrid Systems. Nanoscale 2015, 7, 4598 -- 4810. 33. Majumdar, A. Scanning Thermal Microscopy, Annu. Rev. Mater. Sci. 1999, 29 ,505-585. 34. Gomes, S.; Assy, A.; Chapuis, P.-O. Scanning Thermal Microscopy: a Review, Phys. Status Solidi A 2015, 212, 477-494. 35. Tortello, M.; Colonna, S.; Bernal, M.; Gomez, J.; Pavese, M.; Novara, C.; Giorgis, F.; Saracco, G.; Gonnelli, R.S.; Fina, A. Effect of Thermal Annealing on the Heat Transfer Properties of Reduced Graphite Oxide Flakes: a Nanoscale Characterization Via Scanning Thermal Microscopy. Carbon 2016, 109, 390-401. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 36. Pumarol M.E.; Rosamond, M. C.; Tovee, P.; Petty, M. C.; Zeze, D. A.; Falko, V.; Kolosov O. V. Direct Nanoscale Imaging of Ballistic and Diffusive Thermal Transport in Graphene Nanostructures, Nano Letters 2012, 12, 2906 -- 2911. 37. Menges F.; Riel, H.; Stemmer, A.; Dimitrakopoulos, C.; Gotsmann B. Thermal Transport into Graphene through Nanoscopic Contacts, Phys. Rev. Lett. 2013, 111, 205901. 38. Tovee, P. D.; Pumarol, M. E.; Rosamond, M. C.; Jones, R.; Petty, M.C.; Zezeb D.A.; Kolosov, O.V. Nanoscale Resolution Scanning Thermal Microscopy Using Carbon Nanotube Tipped Thermal Probes, Phys. Chem. Chem. Phys. 2014, 16, 1174-1181. 39. Yoon, K.; Hwang, G.; Chung, J.; Kim, H. G.; Kwon, O.; Kihm, K. D.; Lee, J. S. Measuring the Thermal Conductivity of Residue-Free Suspended Graphene Bridge Using Null Point Scanning Thermal Microscopy, Carbon 2014, 76, 77-83. 40. Kim K.; Chung, J.; Hwang, G.; Kwon, O.; Lee J. S. Quantitative Measurement with Scanning Thermal Microscope by Preventing the Distortion Due to the Heat Transfer through the Air. ACS Nano 2011, 5, 8700-8709. 41. Ciuk, T.; Pasternak, I.; Krajewska, A.; Sobieski, J.; Caban, P.; Szmidt, J.; Strupinski, W. The Properties of CVD Graphene Transferred by High-Speed Electrochemical Delamination. J. Phys. Chem. C 2013, 117, 20833 -- 20837. 42. Li, Q.; Chou, H.; Zhong, J-H.; Liu, J-Y.; Dolocan, A.; Zhang, J.; Zhou, Y.; Ruoff, R. S.; Chen, S.; Cai, W. Growth of Adlayer Graphene on Cu Studied by Carbon Isotope Labeling, Nano Lett. 2013, 13, 486 -- 490. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 43. Nie, S.; Wu, W.; Xing, S.; Yu, Q.; Bao, J.; Pei, S-S.; McCarty, K. F. Growth from Below: Bilayer Graphene on Copper by Chemical Vapor Deposition. New J. Phys. 2012, 14, 093028. 44. Pasternak, I.; Krajewska, A.; Grodecki, K.; Jozwik-Biala, I.; Sobczak, K.; Strupinski W. Graphene Films Transfer Using Marker-Frame Method. AIP Advances 2014, 4, 097133. 45. Ferrari, A. C.; Meyer, J. C.; Scardaci, V.; Casiraghi, C.; Lazzeri, M.; Mauri, F.; Piscanec, S.; Jiang, D.; Novoselov, K. S.; Roth S.; Geim, A. K. Raman Spectrum of Graphene and Graphene Layers. Phys. Rev. Lett. 2006, 97, 187401. 46. Gupta, A.; Chen, G.; Joshi, P.; Tadigadapa, S.; Eklund, P. C. Raman Scattering from High-Frequency Phonons in Supported n Graphene Layer Films. Nano Lett. 2006, 6, 2667−2673. 47. Krajewska, A.; Pasternak, I.; Sobon, G.; Sotor, J.; Przewloka, A.; Ciuk, T.; Sobieski, J.; Grzonka, J.; Abramski, K. M.; Strupinski, W. Fabrication and Applications of Multi-Layer Graphene Stack on Transparent Polymer. Appl. Phys. Lett. 2017, 110, 041901. 48. Shivaprasad, S.M.; Angadi, M.A. Temperature Coefficient Of Resistance Of Thin Palladium Films. J. Phys. D: Appl. Phys. 1980, 13, L171-L172. 49. Puyoo, E.; Grauby, S.; Rampnoux, J.-M.; Rouvière, E.; Dilhaire, S. Scanning Thermal Microscopy of Individual Silicon Nanowires, J Appl Phys 2011, 109, 024302. 50. Obraztsov, A. N.; Obraztsova E. A.; Tyurnina, A. V.; Zolotukhin, A. A. Chemical Vapor Deposition of Thin Graphite Films of Nanometer Thickness. Carbon 2007, 45, 2017-2021. 51. Martinek, J.; Klapetek, P.; Campbell, A. C. Methods for Topography Artifacts Compensation in Scanning Thermal Microscopy, Ultramicroscopy 2015, 155, 55-61. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 52. Kretinin, A. V.; Cao, Y.; Tu, J. S.; Yu, G. L.; Jalil, R.; Novoselov, K. S.; Haigh, S. J.; Gholinia, A.; Mishchenko, A.; Lozada, M.; Georgiou, T.; Woods, C. R.; Withers, F.; Blake, P.; Eda, G.; Wirsig, A.; Hucho, C.; Watanabe, K.; Taniguchi, T.; Geim, A. K.; Gorbachev R. V. Electronic Properties of Graphene Encapsulated with Different Two-Dimensional Atomic Crystals. Nano Lett. 2014, 14, 3270 −3276. 53. Tovee P.; Pumarol, M.; Zeze, D.; Kjoller, K.; Kolosov O. Nanoscale Spatial Resolution Probes for Scanning Thermal Microscopy of Solid State Materials, J. Appl. Phys. 2012, 112, 114317. 54. Yovanovich, M. M.; Culham, J. R.; Teertstra, P. Analytical Modeling of Spreading Resistance in Flux Tubes, Half Spaces, and Compound Disks IEEE transactions on components, packaging, and manufacturing technology 1998, 21, 168-176. 55. Chen, Z.; Jang, W.; Bao, W.; Lau, C. N.; Dames, C. Thermal Contact Resistance Between Graphene and Silicon Dioxide. Appl. Phys. Lett. 2009, 95, 161910. 56. Schmidt, A. J.; Chen, X.; Chen, G. Pulse Accumulation, Radial Heat Conduction, and Anisotropic Thermal Conductivity in Pump-Probe Transient Thermoreflectance. Rev. Sci. Instrum. 2008, 79, 114902. 57. Wang, H.; Xu, Y.; Shimono, M.; Tanaka, Y.; Yamazaki, M.; Computation of Interfacial Thermal Resistance by Phonon Diffuse Mismatch Model. Materials Transactions 2007, 48, 2349-2352. 58. Mak, K. F.; Lui, C. H.; Heinz, T. F. Measurement of the Thermal Conductance of the Graphene/SiO2 Interface. Appl. Phys. Lett. 2010, 97, 221904. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 59. Cai, W.; Moore, A. L.; Zhu, Y.; Li, X.; Chen, S.; Shi, L.; Ruoff, R. S. Thermal Transport in Suspended and Supported Monolayer Graphene Grown by Chemical Vapor Deposition. Nano Lett. 2010, 10, 1645 -- 1651. 60. Papavassiliou, D. V.; Bui, K.; Nguyen, H. Thermal Boundary Resistance Effects in Carbon Nanotube Composites. In Advanced Computational Nanomechanics, First Edition; Silvestre, N. (Ed); John Wiley & Sons: Chichester, 2016; 282-294. 61. Konatham, D.; Striolo, A. Thermal Boundary Resistance at the Graphene-Oil Interface. Appl. Phys. Lett. 2009, 95, 163105. 62. Munro, R. G. Evaluated Material Properties for a Sintered -Alumina. J. Am. Ceram. Soc., 1997, 80, 1919 -- 1928. 63. Goodson, K. E.; Käding, O. W.; Rösler, M.; Zachai, R. Experimental Investigation of Thermal Conduction Normal to Diamond-Silicon Boundaries. J. Appl. Phys. 1995, 77, 1385-1392. 64. Pop, E.; Mann, D. A.; Goodson, K. E.; Dai, H. Electrical and Thermal Transport in Metallic Single-Wall Carbon Nanotubes on Insulating Substrates. J. Appl. Phys. 2007, 101, 093710 65. Schmidt, A. J.; Collins, K. C.; Minnich, A. J.; Chen, G. Thermal Conductance and Phonon Transmissivity of Metal-Graphite Interfaces. J. Appl. Phys. 2010, 107, 104907 66. Jang, W.; Chen, Z.; Bao, W.; Lau, C. N.; Dames C. Thickness-Dependent Thermal Conductivity of Encased Graphene and Ultrathin Graphite. Nano Lett. 2010, 10, 3909 -- 3913. 67. Chen, G.; Hui, P. Thermal Conductivities of Evaporated Gold Films on Silicon and Glass. Appl. Phys. Lett. 1999, 74, 2942. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 68. Langer, G.; Hartmann, J.; Reichling, P. Thermal Conductivity of Thin Metal Films Measured by Photothermal Profile Analysis. Rev. Sci. Instrum. 1997, 68, 1510 -- 1513. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Supporting Information: Chemical Vapor Deposition Graphene as a Thermally Conducting Coating Mauro Tortello,1, Iwona Pasternak2, Klaudia Zeranska-Chudek2, Wlodek Strupinski2, Renato S. Gonnelli1, Alberto Fina1 1. Dipartimento di Scienza Applicata e Tecnologia, Politecnico di Torino, 10129 Torino, Italy 2. Faculty of Physics, Warsaw University of Technology, Koszykowa 75, 00-662 Warsaw Poland To whom correspondence should be addressed: [email protected] ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ Retract measurements The variation in the thermal signal due to the tip-sample heat transfer from the contact-mode operating condition (deflection signal = 0.5 V) to a retraction of 250 nm (corresponding to the lift mode), is small compared to the total one. At about 4 m distance from the sample (maximum distance experimentally achievable for the retraction of the probe), the heat transfer to the sample is still far from being negligible. This can be seen by the fact that the thermal signal at -4 m is still varying, even though the magnitude of the slope of (Vs-Vr) vs z is decreasing with increasing tip-sample distance. a 4-layer sample supported by SiO2. Red: deflection signal. 0.5V corresponds to the force set for the SThM maps acquisition. Blue: SThM thermal signal. Figure S1. Retract measurement on Retract measurements also show that 250 nm is the minimum achievable distance from the experimental point of view. Two examples are reported below. When scanning in the lift mode with a 250 nm distance, the tip-sample height is a few percent within the minimum achievable distance below which electrostatic effects keep attracting the tip to the sample. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ -4.0-3.5-3.0-2.5-2.0-1.5-1.0-0.50.00.5-7-6-5-4-3-2-10123 Deflection signalDeflection Signal, VRetrace Z Position, µm0124GRL/SiO2Lift-mode Thermal signalThermal signal, Vs-Vr (V) Contact-modeoperating conditionFigure S2. The same as Figure S1 but magnified. Note the thermal signal corresponding to the deflection setpoint of 0.5 V, at about z=0 nm and that in the lift mode, at about z=-250 nm. Figure S3. Retract measurement on a 1-layer sample supported by Al2O3. Red: deflection signal. 0.5V corresponds to the force set for the SThM maps acquisition. Blue: SThM thermal signal. Note the thermal signal corresponding to the deflection setpoint of 0.5 V, at about z=0 nm and that in the lift mode, at about z=-250 nm. ©2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ -0.5-0.4-0.3-0.2-0.10.0-7-6-5-4-3-2-10123 Deflection signalDeflection Signal, VRetrace Z Position, µm250 nm0.10.20.30.4Snap off4GRL/SiO2Lift-mode Thermal signalThermal signal, Vs-Vr (V) Contact-modeoperating condition-0.35-0.25-0.15-0.050.05-4-3-2-101 Deflection signal250 nmDeflection Signal, VRetrace Z Position, µm0.00.10.20.30.40.5Snap off ThermalsignalThermal signal, Vs-Vr (V)≅ 6% Lift-modeContact-modeoperating condition1GRL/Al2O3
1909.05757	2	1909	2019-10-18T02:59:59	Fundamental Limits of the Dew-Harvesting Technology	[ "physics.app-ph" ]	Dew-harvesting technology radiatively cools a condenser below the dewpoint to achieve condensation of the water vapor from the atmosphere. Due to its passive nature, this technology has attracted a broad interest, in particular in the context of the worldwide drinking-water scarcity. However, the fundamental limit of its performance has not yet been clarified. Moreover, the existing applications have been limited to humid areas. Here, we point out the upper bound of the performance of this technology by carefully considering the spectral directional atmospheric transmittance in a wide range of parameters such as the ambient temperature, the relative humidity, and the convection coefficient. Moreover, we highlight the potential of a condenser consisting of a selective emitter, which is capable of condensing water vapor under significantly more arid conditions as compared with the use of a blackbody emitter. For example, a near-ideal emitter could achieve a dew-harvesting mass flux of 13 gram per square metre per hour even at ambient temperature of 20 degrees Celsius with relative humidity of 40%, whereas the black emitter cannot operate. We provide a numerical design of such a selective emitter, consisting of six layers, optimized for dew-harvesting purposes.	physics.app-ph	physics	Fundamental Limits of the Dew-Harvesting Technology Minghao Dong,† Zheng Zhang,† Yu Shi,‡ Xiaodong Zhao,† Shanhui Fan,‡ and Zhen Chen✻, † †Jiangsu Key Laboratory for Design & Manufacture or Micro/Nano Biomedical Instruments, School of Mechanical Engineering, Southeast University, Nanjing 210096, China ‡Ginzton Laboratory, Department of Electrical Engineering, Stanford University, Stanford, CA94305, USA Supporting information ABSTRACT: Dew-harvesting technology radiatively cools a condenser below the dewpoint to achieve condensation of the water vapor from the atmosphere. Due to its passive nature, this technology has attracted a broad interest, in particular in the context of the worldwide drinking-water scarcity. However, the fundamental limit of its performance has not yet been clarified. Moreover, the existing applications have been limited to humid areas. Here, we point out the upper bound of the performance of this technology by carefully considering the spectral directional atmospheric transmittance in a wide range of parameters such as the ambient temperature (Tambient), the relative humidity (RH), and the convection coefficient (h). Moreover, we highlight the potential of a condenser consisting of a selective emitter, which is capable of condensing water vapor under significantly more arid conditions as compared with the use of a blackbody emitter. For example, a near-ideal emitter could achieve a dew- harvesting mass flux (𝑚 ′′) of 13 gm-2hr-1 even at Tambient = 20 ℃ with RH = 40%, whereas the black emitter cannot operate. We provide a numerical design of such a selective emitter, consisting of six layers, optimized for dew-harvesting purposes. KEYWORDS: dew-harvesting, radiative cooling, selective emitter, drinking water, Sustainable access to fresh water has been recognized as one of the grand challenges for engineering in the 21st century.1 Traditional technologies, such as distillation,2 reverse osmosis,3 and waste water recycling,4 are energy-consuming and expensive. It is therefore difficult to widely apply them especially in developing countries where the crisis to access fresh drinking water is the most serious. Recent passive techniques utilizing solar energy,5-9 on the other hand, are constrained either by the exotic and expensive water absorbing material5-6 or by the bulky solar concentrating components.8-9 Dew-harvesting technology utilizes the ultracold outer space to radiatively cool a surface below the dewpoint and condenses the water vapor from the atmosphere. This passive technology holds great potential for fresh water harvesting due to the fact that a significant amount of water vapor is stored in the atmosphere.10 The theoretical analysis on dew-harvesting has a long history. Following the pioneering work by Lewis on the evaporation of a liquid into a gas,11 Hofmann developed an analogous mathematical framework to analyze the heat and mass transfer in the process of the dew formation.12 Since then, several theoretical works13-19 have been conducted to quantitatively explore the dew- harvesting. However, none of these works takes the spectral atmospheric transmittance into account, and very few of them considers the spectral emissivity of the condenser.14-15 As a result, the fundamental limits of this technique have not been properly clarified, making it hard to evaluate the performance of the experiments13-14, 20-22 and to determine whether or not this technology is applicable under various conditions, in particular in relatively arid areas.5-6 In this paper, building upon the recent developments of the radiative cooling technology,23-30 we develop a theoretical framework to analyze the dew-harvesting mass flux (𝑚 ′′) of a black and a selective emitter in a wide range of parameters including the ambient temperature (Tambient), the relative humidity (RH), and the convection coefficient (h). We point out that the selective emitter surpasses its black counterpart, in particular under high Tambient in relative arid areas, where the needs of the potable water are more demanding but impossible to be met using a black emitter. For example, at Tambient = 20℃ with RH = 40%, a black emitter fails in harvesting water by any means; in contrast, the selective emitter could have a 𝑚 ′′ of 13 gm-2hr-1. To realize the potentials of a selective emitter, we design a simple multilayer stack to demonstrate its dew-harvesting abilities. We also quantify the bifunction of the convection, and propose a setup to adjust h, for the purpose of maximizing 𝑚 ′′.  RESULTS Theoretical Framework. Figure 1(a) shows a typical setup for dew-harvesting in most existing experiments,17-18, 20 in which the condenser is radiatively cooled to be below the dewpoint to condense the water vapor from the adjacent air. The effects of the convection over the top surface of the condenser are twofold: it recharges the condenser with fresh water vapor, which is beneficial for dew- harvesting, but also warms the condenser, which is detrimental. Therefore, there must exist an optimized convection coefficient, h, that maximizes 𝑚 ′′. We will return to this insight shortly in the context of Fig. 5. As a result, we propose a second setup (Fig. 1b), which is equipped with a low-power air pump to actively control the air speed and thus h. Figure 1. Two setups for dew-harvesting, both of which utilize the radiative cooling technology to cool the condenser below the dewpoint temperature and condense the water vapor from the air. (a) Passive setup and the corresponding energy balance (see text): top surface of the condenser is used for both cooling and water collection. While the parasitic heat loss from the bottom surface, characterized by a convection coefficient, hparasitic, weakens the cooling performance and thus reduces water production, the convection from the top surface, characterized by a different convection coefficient, h, has one additional function: supplying fresh air with water vapor. (b) Active setup and the corresponding energy balance (see text): top and bottom surfaces are for cooling and water collection, respectively. In this scenario, fresh air with water vapor is pumped into a heat exchanger underneath the condenser, which also weakens the cooling performance of the condenser. The air conduction/convection from the top surface is a pure parasitic heat loss, which can be significantly reduced by evacuating the enclosure to vacuum. To obtain both the temperature of the condenser, Tcondenser, and the mass flux of dew-harvesting, 𝑚 ′′, we require two constraints. The first one comes from the energy balance (right panels of Fig. 1) of the condenser, in which the latent heat released by the condensation process can be expressed as 𝑄𝑙𝑎𝑡𝑒𝑛𝑡 = 𝑄𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟(𝑇𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟) − 𝑄𝑎𝑡𝑚(𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡, RH) − (ℎ + ℎ𝑝𝑎𝑟𝑎𝑠𝑖𝑡𝑖𝑐)(𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡 − 𝑇𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟) ,(1) where Qcondenser and Qatm represent the thermal radiation emitted from the condenser and absorbed from the atmosphere, respectively.23, 25 The parasitic heat transfer coefficient, hparasitic, accounts for all the parasitic heat transfer between the condenser and the environment except for the convection that carries water vapor to the condenser, which is characterized by h. For theoretical upper bounds, we set ℎ𝑝𝑎𝑟𝑎𝑠𝑖𝑡𝑖𝑐 = 0 in all calculations in the main text, leaving the discussion on the effect of a nonzero ℎ𝑝𝑎𝑟𝑎𝑠𝑖𝑡𝑖𝑐 in Supporting Information Figure S3 and Note 2. Note here we choose the parameter pair (𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡, RH), where RH is the relative humidity, instead of (𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡, PWV), where PWV stands for the precipitable water vapor, to quantify the atmospheric transmittance and thus 𝑄𝑎𝑡𝑚,31 because the relative humidity is the most straightforward indicator for dew-harvesting, as will be evident in the following discussion. For the second constraint, we follow the analyses of Lewis11 and Hofmann.12 Analyzing the heat and mass transfer simultaneously,32 we develop a general expression 𝑄𝑙𝑎𝑡𝑒𝑛𝑡 = (𝐿𝑒)𝑛 ℎ 𝛾 [RH × 𝑃𝐻2𝑂(𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡) − 𝑃𝐻2𝑂(𝑇𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟)], (2) which is valid in various situations characterized by different power, n, of the Lewis number, Le (= 0.87 at 300 K),32 as discussed in detail in METHODS and Supporting Information Note 2. Corresponding to the setup in Fig. 1(a), we set n = -2/3 and thus (𝐿𝑒)𝑛 = 1.1 for the calculations in this work, which introduces an error less than 6% if we consider the other scenarios, e.g. the setup in Fig. 1(b). Here 𝛾 is the psychrometer constant (67 PaK-1 at 20℃),33 and 𝑃𝐻2𝑂 is the saturation vapor pressure, which depends only on the temperature.34 The last term of Eq. (2) assumes that the air close to the condenser surface is saturated in water vapor, which is justified since 𝑇𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟 needs to be lower than the dewpoint temperature, 𝑇𝑑𝑒𝑤−𝑝𝑜𝑖𝑛𝑡, for condensation to start. Solving Eqs. (1) and (2), we obtain Tcondenser and Qlatent. The latter can be converted to the mass flux of dew-harvesting, 𝑚 ′′ = 𝑄𝑙𝑎𝑡𝑒𝑛𝑡 ∆ , (3) in which ∆ is the latent heat per unit mass. Figure 2 shows the RH dependence of Tcondenser (black solid line; left axis) and 𝑚 ′′ (orange solid line; right axis) of a black emitter at 𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡 = 20℃, under a typical rooftop scenario (h = 8 Wm-2K-1). The dewpoint temperature, 𝑇𝑑𝑒𝑤−𝑝𝑜𝑖𝑛𝑡 (black dashed line; left axis), is also calculated35 for reference. At low RH, Tcondenser is higher than 𝑇𝑑𝑒𝑤−𝑝𝑜𝑖𝑛𝑡, and thus no dew forms. As RH increases, 𝑇𝑑𝑒𝑤−𝑝𝑜𝑖𝑛𝑡 increases faster than Tcondenser, resulting in a crossover at RH = 74%, where water vapor starts to condense to liquid. Beyond this crossover, 𝑚 ′′ monotonically increases, and finally reaches 42gm-2hr-1 at RH = 100%. Figure 2. Simultaneously solving the temperature, Tcondenser (black solid line; left axis), and the condensation mass flux, 𝑚 ′′ (orange; right axis), as a function of the relative humidity, RH, for a black emitter at Tambient = 20 oC, under a typical rooftop scenario (h = 8 Wm-2K-1). The dewpoint temperature (black dashed line) is also calculated for reference. Dew does not form until a characteristic relative humidity, RH=74%, at which Tcondenser starts to be lower than Tdew-point. As RH increases, 𝑚 ′′ monotonically increases to 42 gm- 2hr-1 at RH = 100%. Black vs. Selective Emitter. Figure 3 analyzes 𝑚 ′′ for three different emitters: a black emitter (black lines), a near-ideal emitter (blue lines) which emits 100% in the wavelength range of 8 -- 13m and 0% elsewhere,25 and a photonic design in this work (red lines), under two representative scenarios: a typical rooftop setup (h = 8 Wm-2K-1), e.g. in Fig. 1(a), and a setup that reduces the wind speed (h = 2 Wm-2K-1), e.g. in Fig. 1(b). We will return to the photonic design shortly in the context of Fig. 4. The dominant feature of Fig. 3 is that the near-ideal emitter surpasses its black counterpart in any circumstance. In particular, the near-ideal emitter is capable of harvesting dew at much lower RH (Fig. 3a) and much higher Tambient (Fig. 3b). Table 1 presents two examples of this feature. First, in a relatively dry condition (RH = 40% and Tambient = 20℃; first column), the black emitter does not work at all, while the near-ideal one has a 𝑚 ′′ of 13 gm-2hr-1. Second, in a high temperature condition (Tambient = 30℃ and RH = 60%; second column), the black emitter again fails in harvesting water, while the near- ideal one has a 𝑚 ′′ up to 14 gm-2hr-1. We note the comparison between the two groups of scenarios in Fig. 3(a): h = 8 Wm-2K-1 (dashed lines) vs. h = 2 Wm-2K-1 (solid lines). While the former shows less difference among the three emitters, the latter amplifies the advantages of the selective emitters, in particular at low RH. Dashed lines in Fig. 3 represent 𝑚 ′′ values that would have been reached if dew formed. However, under these conditions, Tcondenser is below 0℃, and thus frost forms, which hinders the water-harvesting process, in particular in the setup of Fig. 1(a). Figure 3. Relative humidity (RH) and ambient temperature (𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡) dependence of the mass flux of dew- harvesting (𝑚" ). (a) RH dependence of 𝑚" for three different emitters at Tambient = 20 oC under two representative scenarios: ℎ = 2 Wm−2K−1 and 8 Wm−2K−1. For any emitter under any scenario, 𝑚" increases monotonically as the increase of RH. The near-ideal emitter (blue) surpasses its black counterpart (black) in both scenarios. In particular, the former is capable of harvesting water at significantly low RH. (b) Tambient dependence of 𝑚" . For a fixed RH, the content of the water vapor in the atmosphere increases with the increase of Tambient, which, on one hand, is beneficial to dew-harvesting, but, on the other hand, is detrimental due to the blurred atmospheric transparency, resulting in the reduction of the radiative cooling flux. This competition results in maximal 𝑚" at optimized Tambient. Dashed lines represent 𝑚 ′′ values that would have been reached if dew formed, although in reality frost forms instead, because Tcondenser < 0 oC. The other major feature is the trend that 𝑚 ′′ increases monotonically with the increase of RH in Fig. 3(a), but has maxima at optimized Tambient in Fig. 3(b), regardless of the type of the condenser and the value of the convection coefficient. These trends result from the variation of the water vapor in the atmosphere, whose effects are twofold. On one hand, increasing RH (Tambient) while fixing Tambient (RH), the content of the water vapor in the atmosphere increases, which is beneficial to dew-harvesting. On the other hand, it could also be detrimental since the atmospheric transmission in the transparency window is reduced31 and thus reducing the cooling flux. The former always surpasses the latter at any RH, resulting in a monotonic trend in Fig. 3(a), while these two effects are balanced at specific Tambient, resulting in maxima at optimized Tambient in Fig. 3(b). Table 1. Two representative examples (first and second columns), showing the advantages of the selective emitter over the black one, as well as the fundamental limit (third column) of 𝒎 ′′. While the black emitter is slightly better at ultrahigh RH and h (third column), the selective emitters are significantly better in low RH (first column) and high Tambient (second column). 𝑚 ′′ [gm−2hr−1] Black Near-ideal Design in this work (see Fig. 4) (𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡, RH, ℎ) [℃, %, Wm−2K−1] (20, 40, 2) (30, 60, 2) (12, 100, ∞) 0 13 5 0 14 5 59 56 55 Figure 4(a) shows the photonic design of a selective emitter aiming to maximize the condensation mass flux, 𝑚 ′′. The memetic algorithm36 is used to optimize the spectrum of a multilayer device that approaches the emissivity spectrum of the near-ideal emitter. This design consists of six layers, which are made of three common dielectric materials: MgF2, SiN4, and SiC, with an aluminum (Al) substrate as the back mirror. The emissivity spectrum of this design is shown in Fig. 4(b) (red), with a representative atmospheric transmittance (grey) in a relatively dry climate (Tambient = 10 oC and RH = 30%) for reference.37 We note that there exists broad and large emissivity inside the atmospheric transparency window (8-13 μm), which ensures that the condensation mass flux, 𝑚 ′′, of this photonic design significantly surpasses that of the black emitter, which is evident from the comparison in Figs. 3 and 5, as well as in Table 1 (first two columns). Moreover, this design is also better than our previous designs for radiative cooling purposes,24-25 which are targeted at achieving low temperatures (Fig. S4 of Supporting Information). Note that in Fig. 4(b) we show the emissivity and transmittance only along the zenith angle for clarity, but in the calculations of 𝑚 ′′ we use the angle-dependent properties. Figure 4. A photonic design to maximize 𝑚" . (a) Schematic cross-sectional view of the multilayer stack made of three different materials (MgF2, Si3N4, and SiC) on an aluminum substrate. (b) Calculated spectral emissivity of the photonic design (red) along zero zenith angle, with a typical atmospheric transmittance (grey) under a relative dry condition (Tambient = 10 oC and RH = 30%) for reference. Our designed structure emits significantly in the atmospheric transparency window (8-13 m). Effect of Convection. We now turn to quantifying our intuition on the effect of convection, which motivates us to propose the setup in Fig. 1(b). As an example, Figs. 5(a-b) clearly show maximal 𝑚 ′′ at optimized h, at Tambient = 20 oC with RH = 40% and 80%, regardless of the type of the condenser. These maxima originate from the two competing effects associated with convection: on one hand, convection brings water vapor to the condenser, which helps the dew-harvesting; on the other hand, convection also carries parasitic heat to reduce the cooling flux of the condenser, which hurts the dew- harvesting. As a result, at RH < 100%, 𝑚 ′′ first increases and then decreases with the increase of h. We note that by means of varying h we not only maximize 𝑚 ′′ but also avoid frost formation (indicated by the dashed lines), which is a significant advantage of the setup we proposed in Fig. 1(b). Figure 5. Effect of convection. (a-b) RH < 100%. As h increases from zero, more and more water vapor is carried to the condenser, which increases 𝑚" ; as h keeps increasing, however, the condenser temperature finally surpasses the dewpoint temperature (Tcondenser > Tdew-point), which stops the condensation. (c) RH = 100%. In this case, Tcondenser can never surpass Tdew-point, since now Tdew-point = Tambient. Now the only role of convection is to supply water vapor, and the maximal 𝑚" is defined by the radiative cooling power at Tcondenser = Tambient. While the selective emitters can reach significantly low temperature, the black emitter has the maximal cooling flux at Tcondenser = Tambient, which explains the crossovers. Dashed lines represent 𝑚 ′′ values that would have been reached if dew formed. However, under these conditions, Tcondenser is below 0 oC, and thus frost forms, which hinders the dew-harvesting. At RH = 100%, however, 𝑚 ′′ monotonically increases and finally approaches its fundamental limit at the specific Tambient (here, 20 oC), as shown in Fig. 5(c). We note the logarithmic x-axis to emphasize the transition between the two regimes. The saturation behavior results from two facts. First, Tdew-point strictly equals Tambient at RH = 100%; and second, Tcondenser approaches Tambient at large h. These two facts result in Tcondenser ≈ Tambient = Tdew-point at RH = 100% with large h, which requires almost no further cooling for the condensation to start. At this limit, all the radiative cooling flux is used for condensation, and thus the fundamental limit is approached at the specific Tambient. Another feature of Fig. 5(b) is the crossovers between the black and the selective emitters. Selective emitters have higher 𝑚 ′′ at small h, while black emitter is better at large h. This feature results from the fact that selective emitter reaches significantly lower temperature at relatively small h,25 but black emitter has the maximal cooling flux at Tcondenser = Tambient. This result underlines that while the black emitter can reach the ultimate fundamental limit of dew-harvesting at RH = 100% and ultrahigh h (3rd column of Table 1), the selective emitter is better in most other conditions (Fig. S1 of the Supporting Information).  DISCUSSION AND SUMMARY We develop a general framework to analyze the dew-harvesting technology. In particular, we highlight the effect of important factors, such as Tambient and RH. Distinct from the previous analyses in the literature, we consider the wavelength and angle dependence of atmospheric transmittance. We point out that while the black emitter gives the ultimate upper bound of approximately 60 gm-2hr-1 at 100% of RH, a selective emitter can harvest water at much higher Tambient (e.g. 30 ℃) with much lower RH (e.g. 40%). We design a simple multilayer stack to realize this potential. We also quantify the effect of convection, which carries not only water vapor but also parasitic heat to the emitter. Correspondingly, we propose a setup which can adjust h to maximize 𝑚 ′′ and avoid frost. We end by commenting on the prospect of this technology. Dew-harvesting could be beneficial to both humid and dry areas. The former includes islands and coastal cities which are surrounded by seawater that is not potable, while the latter include deserts which lack of any form of drinking water. Even in the hot and arid desert, there is still some amount of water vapor stored in the atmosphere, e.g. Tambient = 20℃ and RH = 40% at night in the Mojave Desert in California,38 which is ready to be harvested by a carefully designed photonic emitter. This passive fresh water harvesting technology would complement existing technologies, especially in rural and low-income areas where the cost is a big concern. In particularly, it could provide an option for local fresh water harvesting, which is more effective for personal needs.1  METHODS We present the derivation of Eq. (2) in detail. Driven by the difference of the water vapor concentration, the mass transfer flux of the water vapor from the ambient to the condenser surface is32 𝑚 ′′ = 𝑔𝑚,𝐻2𝑂(𝑚 𝐻2𝑂,𝑎𝑚𝑏𝑖𝑒𝑛𝑡 − 𝑚 𝐻2𝑂,𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟), (4) where 𝑔𝑚,𝐻2𝑂 is the mass transfer coefficient of the water vapor, with the same unit of 𝑚 ′′, and 𝑚 𝐻2𝑂,𝑎𝑚𝑏𝑖𝑒𝑛𝑡 and 𝑚 𝐻2𝑂,𝑐𝑜𝑛𝑑𝑒𝑛𝑠𝑒𝑟 are the mass fraction of water vapor in the ambient and near the condenser surface, respectively. From the Ideal Gas Law, we write the mass fraction of water vapor as 𝑚 𝐻2𝑂 = 𝑅𝐻×𝑃𝐻2𝑂𝑀𝐻2𝑂 𝑃∞𝑀𝑎𝑖𝑟 , (5) Where 𝑃∞ is the atmospheric pressure, 𝑀𝐻2𝑂 and Mair are the molecular mass of water and air, respectively, and the saturation vapor pressure, 𝑃𝐻2𝑂, with unit of [10−3bar], has an expression34 of 𝑃𝐻2𝑂 = 𝑇𝑎110𝑐1+𝑏1/T ,where T is the absolute temperature, and the parameters are: 𝑎1 = −4.9283, 𝑏1 = −2937.4 K, and 𝑐1 = 23.5518, which are applicable in the temperature range of -50 ℃ - 100 ℃. The most important step is to bridge the mass transfer coefficient, 𝑔𝑚,𝐻2𝑂, and the convection heat transfer coefficient, h. Applying the strong analogy between the heat and mass transfer,32 we obtain 𝑔𝑚,𝐻2𝑂 = ℎ 𝑐𝑝 (𝐿𝑒)𝑛, (6) where cp is the specific heat capacity of the air at constant pressure, with unit of [Jkg-1K-1], and Le is the Lewis number, which is a measure of the relative thermal and concentration boundary layer thicknesses.32, 39 We summarize the n values in Table 2 for different scenarios, including natural and forced convection at horizontal/inclined plates or in pipes with different shapes of cross section, as discussed in detail in Supporting Information Note 1. Combining Eqs. (3)-(6), we arrive at Eq. (2) of the main text. Table 2. Values of n in Eqs. (2) and (6), corresponding to different scenarios. Detailed derivations see Supporting Information Note 1. Natural convection Forced external convection Forced internal (horizontal or inclined plate) − 3 4⁄ (horizontal or inclined plate) − 2 3⁄ convection (pipe) −1 n  ASSOCIATED CONTENT Supporting Information Figure on the contour plots of the condensation mass flux as a function of the relative humidity (x- axis) and the ambient temperature (y-axis) under two representative scenarios. Figure on the comparison of the condensation mass flux between the two setups in Fig. 1(a) and 1(b). Figure on the effect of the parasitic heat transfer coefficient on the condensation mass flux. Figure on the comparison between the photonic design in this work and the designs in literature. Note on the relation between the heat and mass transfer coefficients. Note on the effect of the parasitic heat transfer coefficient. Note on the comparison between the photonic design in this work and the designs in literature.  AUTHOR INFORMATION Corresponding Author ✻E-mail: [email protected]. Notes The authors declare no competing financial interest. During the preparation of this manuscript, we became aware of a related analysis.40  ACKNOWLEDGEMENTS This work was supported in part by National Natural Science Foundation of China (51776038) and the Innovative and Entrepreneurial Talent Plan (Jiangsu Province, China). S. F. acknowledges the support of the U. S. Department of Energy Office of Basic Energy Sciences (grant No. DE-FG02-07ER46426). References 1. 14 Grand Challenges for Enginerring in the 21st Century. http://www.engineeringchallenges.org/challenges.aspx. 2. Al-Obaidani, S.; Curcio, E.; Macedonio, F.; Profio, G. D.; Al-Hinai, H.; Drioli, E. Potential of membrane distillation in seawater desalination: Thermal efficiency, sensitivity study and cost estimation. Journal of Membrane Science 2008, 323, 85-98. 3. Greenlee, L. F.; Lawler, D. F.; Freeman, B. D.; Marrot, B.; Moulin, P. Reverse osmosis desalination: Water sources, technology, and today's challenges. Water Research 2009, 43, 2317-2348. 4. Gupta, V. K.; Ali, I.; Saleh, T. A.; Nayak, A.; Agarwal, S. Chemical treatment technologies for waste-water recycling -- an overview. RSC Advances 2012, 2. 5. Kim, H.; Yang, S.; Rao, S. R.; Narayanan, S.; Kapustin, E. A.; Furukawa, H.; Umans, A. S.; Yaghi, O. M.; Wang, E. N. Water harvesting from air with metal-organic frameworks powered by natural sunlight. Science 2017, 358, 430. 6. Kim, H.; Rao, S. R.; Kapustin, E. A.; Lin, Z.; Yang, S.; Yaghi, O. M.; Wang, E. N. Adsorption- based atmospheric water harvesting device for arid climates. Nature Communications 2018, 9, 1191. 7. Ni, G.; Li, G.; Boriskina, S. V.; Li, H.; Yang, W.; Zhang, T.; Chen, G. Steam generation under one sun enabled by a floating structure with thermal concentration. Nature Energy 2016, 1. 8. Cooper, T. A.; Zandavi, S. H.; Ni, G. W.; Tsurimaki, Y.; Huang, Y.; Boriskina, S. V.; Chen, G. Contactless steam generation and superheating under one sun illumination. Nature Communications 2018, 9. 9. Zhou, L.; Tan, Y.; Wang, J.; Xu, W.; Yuan, Y.; Cai, W.; Zhu, S.; Zhu, J. 3D self-assembly of aluminium nanoparticles for plasmon-enhanced solar desalination. Nature Photonics 2016, 10, 393-398. 10. Mekonnen, M. M.; Hoekstra, A. Y. Encyclopedia of Climate and Weather, OXford Univ. Press: 1996. 11. Lewis, W. K. The evaporation of a liquid into a gas. Mechanical Engineering 1992, 44, 445- 446. 12. Hofmann, G. Die Thermodynamik der Taubildung. 18, Offenbach am Main, 1955, 1-45. 13. Jorge Ernesto, A.-T.; Jose Jasson, F.-P. Winter Dew Harvest in Mexico City. Atmosphere 2015, 7. 14. Nilsson, T. M. J.; Vargas, W. E.; Niklasson, G. A.; Granqvist, C. G. Condensation of water by radiative cooling. Renewable Energy 1994, 5, 310-317. 15. Nilsson, T. M.; Vargas, W. E.; Niklasson, G. A. Pigmented foils for radiative cooling and condensation irrigation. Proceedings of SPIE - The International Society for Optical Engineering 1994, 2255. 16. Al-Nimr, M. A.; Haddad, O. Water distiller/condenser by radiative cooling of ambient air. Renewable Energy 1998, 13, 323-331. 17. Beysens, D.; Milimouk, I.; Nikolayev, V.; Muselli, M.; Marcillat, J. Using radiative cooling to condense atmospheric vapor: a study to improve water yield. Journal of Hydrology 2003, 276, 1-11. 18. Muselli, M.; Beysens, D.; Marcillat, J.; Milimouk, I.; Nilsson, T.; Louche, A. Dew water collector for potable water in Ajaccio (Corsica Island, France). Atmospheric Research 2002, 64, 297-312. 19. Nikolayev, V. S.; Beysens, D.; Gioda, A.; Milimouka, I.; Morel, J. P. Water recovery from dew. Journal of Hydrology 1996, 182, 19-35. 20. Muselli, M.; Beysens, D.; Mileta, M.; Milimouk, I. Dew and rain water collection in the Dalmatian Coast, Croatia. Atmospheric Research 2009, 92, 455-463. 21. Beysens, D.; Muselli, M.; Milimouk, I.; Ohayon, C.; Berkowicz, S.; Soyeux, E.; Mileta, M.; Ortega, P. Application of passive radiative cooling for dew condensation. Energy 2006, 31, 2303-2315. 22. Beysens, D.; Muselli, M.; Nikolayev, V.; Narhe, R.; Milimouk, I. Measurement and modelling of dew in island, coastal and alpine areas. Atmospheric Research 2005, 73, 1-22. 23. Rephaeli, E.; Raman, A.; Fan, S. Ultrabroadband photonic structures to achieve high- performance daytime radiative cooling. Nano Letters 2013, 13, 1457-1461. 24. Raman, A. P.; Anoma, M. A.; Zhu, L.; Rephaeli, E.; Fan, S. Passive radiative cooling below ambient air temperature under direct sunlight. Nature 2014, 515, 540-4. 25. Chen, Z.; Zhu, L.; Raman, A.; Fan, S. Radiative cooling to deep sub-freezing temperatures through a 24-h day-night cycle. Nat Commun 2016, 7, 13729. 26. Goldstein, E. A.; Raman, A. P.; Fan, S. Sub-ambient non-evaporative fluid cooling with the sky. 2017. 27. Zhai, Y.; Ma, Y.; David, S. N.; Zhao, D.; Lou, R.; Tan, G.; Yang, R.; Yin, X. Scalable- manufactured randomized glass-polymer hybrid metamaterial for daytime radiative cooling. Science 2017, 355, 1062-1066. 28. Mandal, J.; Fu, Y.; Overvig, A. C.; Jia, M.; Sun, K.; Shi, N. N.; Zhou, H.; Xiao, X.; Yu, N.; Yang, Y. Hierarchically porous polymer coatings for highly efficient passive daytime radiative cooling. Science 2018, 362, 315-319. 29. Li, T.; Zhai, Y.; He, S.; Gan, W.; Wei, Z.; Heidarinejad, M.; Dalgo, D.; Mi, R.; Zhao, X.; Song, J.; Dai, J.; Chen, C.; Aili, A.; Vellore, A.; Martini, A.; Yang, R.; Srebric, J.; Yin, X.; Hu, L. A radiative cooling structural material. Science 2019, 364, 760-+. 30. Zhou, M.; Song, H.; Xu, X.; Shahsafi, A.; Xia, Z.; Ma, Z.; Kats, M.; Zhu, J.; Ooi, B. S.; Gan, Q.; Yu, Z. Accelerating vapor condensation with daytime radiative cooling. arXiv:1804.10736 2018. 31. Dong, M.; Zhang, Z.; Shi, Y.; Zhao, X.; Fan, S.; Chen, Z. Nighttime radiative cooing in hot and humid climates. Optics Express 2019, 27, 31587-31598. 32. Lienhard, J. H.; Lienhard, J. H. A Heat Transfer Textbook, 4th ed.; Dover: 2011. 33. Monteith, J.; Unsworth, M. Principles of Environmental Physics, 3rd ed.; ELSEVIER: 2008. 34. Parish, O. O.; Putnam, T. W. Equations for the determination of humidity from dewpoint and psychrometric data. NASN TN 1977, D-8407. 35. B., H. The proceedings of the Third international Symposium on Humidity & Moisture, Teddington, London, England, Thunder Scientific Corporation, Albuquerque, NM, USA: Teddington, London, England, April 1998. 36. Shi, Y.; Li, W.; Raman, A.; Fan, S. Optimization of multi-layer optical films with a memetic algorithm and mixed integer programming. Acs Photonics 2017, 5, acsphotonics.7b01136. 37. Shen, S. S.; Berk, A.; Lewis, P. E.; Anderson, G. P.; Acharya, P. K.; Bernstein, L. S.; Muratov, L.; Lee, J.; Fox, M.; Adler-Golden, S. M.; Chetwynd, J. J. H.; Hoke, M. L.; Lockwood, R. B.; Gardner, J. A.; Cooley, T. W.; Borel, C. C.; Lewis, P. E.; Shettle, E. P., MODTRAN5: 2006 update. for Multispectral, Hyperspectral, and Ultraspectral Imagery XII, 2006. 38. Weather Underground. https://www.wunderground.com/forecast/us/ca/mojave/93501?cm_ven=localwx_10day. 39. Incropera, F. P.; Dewitt, D. P.; BERGMAN, T. L.; LAVINE, A. S. Fundamentals of Heat and Mass Transfer, WiLEY: 2006. 40. Yuval, Z.; Nadav, D.; and John, H. Deep, High-Yield, Passive Dew Collection. arXiv:1909.10937v1 2019. In Algorithms and Technologies Fundamental Limit of the Dew-Harvesting Technology Supporting Information Minghao Dong,1 Zheng Zhang,1 Yu Shi,2 Xiaodong Zhao,1 Shanhui Fan,2 and Zhen Chen1, * 1Jiangsu Key Laboratory for Design & Manufacture or Micro/Nano Biomedical Instruments, School of 2Ginzton Laboratory, Department of Electrical Engineering, Stanford University, Stanford, CA94305, USA Mechanical Engineering, Southeast University, Nanjing 210096, China *Email: [email protected] Figure S1. Contour plots of the condensation mass flux, 𝑚 ′′, as a function of RH (x-axis) and 𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡 (y- axis) for the three condensers (rows) under the two representative scenarios (columns): ℎ = 2 Wm-2K-1 and 8 Wm-2K-1. Figure S2. Comparison of the condensation mass flux, 𝑚 ′′ between the two setups in Fig. 1a and 1b of the main text. The difference is within 6% for all the three condensers. Grey lines represent 𝑚" values that would have been reached if dew formed. However, in reality frost forms instead, because under these conditions, Tcondenser is below 0℃. Figure S3. Effect of the parasitic heat transfer coefficient, hparasitic, on the condensation mass flux, 𝑚 ′′ . Here, the parasitic heat refers to the heat loss through the lower surface of the condenser (Fig. 1a of the main text), or through the upper surface of the condenser (Fig. 1b of the main text). This parasitic heat reduces the net cooling flux for condensation, and thus reducing 𝑚 ′′, regardless of the type of the condenser and the value of the convection coefficient, h. Grey lines represent 𝑚" values that would have been reached if dew formed. However, in reality frost forms instead, because under these conditions, Tcondenser is below 0℃. Figure S4. Comparison with previous designs 1-2 on the condensation mass flux, 𝑚 ′′. Design in this work (red), aiming at maximizing is apparently better than our previous designs (grey and black), which is optimized for radiative cooling only. Supporting Information Note 1: Relation between the heat and mass transfer coefficients: derivation of Eq. (6) of the main text We employ the strong analogy between the heat and mass transfer,3-4 and analyze three scenarios. First, we recall the textbook example of the forced external convection over a horizontal plate, which has a Nusselt number4 𝑁𝑢 = ℎ𝐿 𝑘 = 0.664(𝑅𝑒𝐿)1/2(𝑃𝑟)1/3, (S1) in which h and k are the convection coefficient and the thermal conductivity of the air, L is the characteristic length of the plate, and Re and Pr are the Reynold number and Prandtl number, respectively. Correspondingly, the Nusselt number of the mass transfer is4 𝑁𝑢𝑚 = 𝑔𝑚,𝐻2𝑜𝐿 𝜌𝑎𝑖𝑟𝐷𝐻2𝑂,𝑎𝑖𝑟 = 0.664(𝑅𝑒𝐿)1/2(𝑆𝑐)1/3, (S2) which just simply replaces the Prnadtl number, Pr, with the Schmidt number, Sc. 𝜌𝑎𝑖𝑟 is the density of air, and 𝐷𝐻2𝑂,𝑎𝑖𝑟 is the mass diffusion coefficient of the water vapor in the air. Combining Eqs. (S1-S2), we obtain 𝑁𝑢𝑚 𝑁𝑢 = ( 𝑔𝑚,𝐻2𝑜𝐿 𝜌𝑎𝑖𝑟𝐷𝐻2𝑂,𝑎𝑖𝑟 ℎ𝐿 𝑘 ) = ( 1 3, 𝑆𝑐 𝑃𝑟 ) (S3) in which the first bracket gives 𝑐𝑝 at 𝑔𝑚,𝐻2𝑜 ℎ 𝐿𝑒, and the second bracket gives (𝐿𝑒) 1 3. Therefore, we arrive 𝑔𝑚,𝐻2𝑂 = ℎ 𝑐𝑝 (𝐿𝑒)− 2 3, (S4) which gives the relation between the mass transfer coefficient, 𝑔𝑚,𝐻2𝑂, and the heat transfer coefficient, h, for the forced external convection over a horizontal plate (2nd column of Table 2 of the main text). The relation between 𝑔𝑚,𝐻2𝑂, and h for an inclined plate is exactly the same with Eq. S4, since the only difference in Nu and Num is the replacement of the constant 0.664 with different constants according to the inclined angles.5 Second, we derive a similar relation for the natural convection over a horizontal or an inclined plate. One commonly used correlation is3 𝑁𝑢 = ℎ𝐿 𝑘 1/4 = 0.27𝑅𝑎𝐿𝑐 , (S5) in which Lc is the modified characteristic length of the plate, and the Rayleigh number is 𝑅𝑎𝐿𝑐 = 3 𝑔𝑐𝑜𝑠𝜃𝐿𝑐 ∆𝑇 𝜈𝛼 𝑇 , (S6) where 𝜃 is the angle measured from the vertical direction to the direction along the inclined plate. The corresponding Nusselt number of the mass transfer is 𝑁𝑢𝑚 = 𝑔𝑚,𝐻2𝑜𝐿 𝜌𝑎𝑖𝑟𝐷𝐻2𝑂,𝑎𝑖𝑟 1/4 = 0.27𝑅𝑎𝑚,𝐿𝑐 , (S7) which just simply replaces the heat transfer Rayleigh number, 𝑅𝑎𝐿𝑐, with the mass transfer Rayleigh number,4 𝑅𝑎𝑚,𝐿𝑐 = 3 𝑔𝑐𝑜𝑠𝜃𝐿𝑐 𝜈𝐷𝐻2𝑂,𝑎𝑖𝑟 ∆𝜌𝑎𝑖𝑟 𝜌𝑎𝑖𝑟 . (S8) Combining Eqs. (S5-S8), we obtain 𝑁𝑢𝑚 𝑁𝑢 = ( 𝑔𝑚,𝐻2𝑜𝐿 𝜌𝑎𝑖𝑟𝐷𝐻2𝑂,𝑎𝑖𝑟 ℎ𝐿 𝑘 1 4 ) ) = ( 𝑅𝑎𝑚,𝐿𝑐 𝑅𝑎𝐿𝑐 , (S9) in which the first bracket again gives 𝑐𝑝 ℎ the Ideal Gas Law. Therefore, we arrive at 𝑔𝑚,𝐻2𝑜 𝐿𝑒, and the second bracket gives (𝐿𝑒) 1 4, which utilizes 𝑔𝑚,𝐻2𝑂 = ℎ 𝑐𝑝 (𝐿𝑒)− 3 4, (S10) which gives the relation between 𝑔𝑚,𝐻2𝑂 and h for the natural convection over a horizontal or an inclined plate (1st column of Table 2 of the main text). In the last scenario, we derive the relation for the internal flow through pipes with various shapes of the cross section. In this case, the heat and mass transfer shares with a same constant Nusselt number 𝑁𝑢 = 𝑁𝑢𝑚 = constant, where the constant depends on the shape of the cross section. Therefore, we obtain 𝑁𝑢𝑚 𝑁𝑢 = ( 𝑔𝑚,𝐻2𝑜𝐿 𝜌𝑎𝑖𝑟𝐷𝐻2𝑂,𝑎𝑖𝑟 ℎ𝐿 𝑘 ) = 1, (S11) (S12) in which the first bracket again gives 𝑐𝑝 𝑔𝑚,𝐻2𝑜 ℎ 𝐿𝑒. Thus, we arrive at 𝑔𝑚,𝐻2𝑂 = ℎ 𝑐𝑝 (𝐿𝑒)−1, (S13) which gives the relation between 𝑔𝑚,𝐻2𝑂 and h for the forced internal convection through a pipe with various shapes of the cross section (3rd column of Table 2 of the main text). In Summary, Eqs. S4, S10, and S13 give the relation between 𝑔𝑚,𝐻2𝑂 and h for three different scenarios, which are characterized by different n values, as outlined in Table 2 of the main text. We note that the differences among these scenarios are less than 6%, as shown in Fig. S2 for a comparison between the setups of Fig. 1(a) (2nd column of Table 2) and Fig. 1(b) (3rd column of Table 2) of the main text. Supporting Information Note 2: Effect of the parasitic heat transfer coefficient In Eq. (1) of the main text, we set the parasitic heat transfer coefficient, hparasitic = 0, to reach the fundamental limits. In this note, we analyze the effect of a nonzero hparasitic. As shown in Fig. S3, the mass flux of the dew-harvesting, 𝑚 ′′, decreases as the increase of hparasitic for any condenser with any convection coefficient, h, which is consistent with our insight in the main text. Supporting Information Note 3: Comparison with other photonic designs The photonic design in this work represents the first design aiming at maximizing the mass flux of the dew-harvesting, 𝑚 ′′, in contrast with those designs merely for radiative cooling, e.g. in Refs. 1 and 2. As expected, the 𝑚 ′′ of this design surpasses our previous designs, as shown in Fig. S4. References 1. Chen, Z.; Zhu, L.; Raman, A.; Fan, S. Radiative cooling to deep sub-freezing temperatures through a 24-h day-night cycle. Nat Commun 2016, 7, 13729. 2. Raman, A. P.; Anoma, M. A.; Zhu, L.; Rephaeli, E.; Fan, S. Passive radiative cooling below ambient air temperature under direct sunlight. Nature 2014, 515, 540-4. 3. Incropera, F. P.; Dewitt, D. P.; BERGMAN, T. L.; LAVINE, A. S. Fundamentals of Heat and Mass Transfer, WiLEY: 2006. 4. Lienhard, J. H.; Lienhard, J. H. A Heat Transfer Textbook, 4th ed.; Dover: 2011. 5. Sparrow, E. M.; Tien, K. K. Forced convection heat transfer at an inclined and yawed square plate - Application to solar collectors. Journal of Heat Transfer 1977, 99, 507-512.
1912.06826	1	1912	2019-12-14T11:09:23	The Controlled Large-Area Synthesis of Two Dimensional Metals	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	The rise of nanotechnology has been propelled by low dimensional metals. Albeit the long perceived importance, synthesis of freestanding metallic nanomembranes, or the so-called 2D metals, however has been restricted to simple metals with a very limited in-plane size (< 10{\mu}m). In this work, we developed a low-cost method to synthesize 2D metals through polymer surface buckling enabled exfoliation. The 2D metals so obtained could be as chemically complex as high entropy alloys while possessing in-plane dimensions at the scale of bulk metals (> 1 cm). With our approach, we successfully synthesized a variety of 2D metals, such as 2D high entropy alloy and 2D metallic glass, with controllable geometries and morphologies. Moreover, our approach can be readily extended to non-metals and composites, thereby opening a large window to the fabrication of a wide range of 2D materials of technologic importance which have never been reported before.	physics.app-ph	physics	The Controlled Large-Area Synthesis of Two Dimensional Metals Tianyu Wang1, Quanfeng He1, Jingyang Zhang1, Zhaoyi Ding1, Fucheng Li1, Yong Yang 1,2* 1. Department of Mechanical Engineering, College of Engineering, City University of Hong Kong, Kowloon Tong, Kowloon, Hong Kong SAR, China 2. Department of Materials Science and Engineering, College of Engineering, City University of Hong Kong, Kowloon Tong, Kowloon, Hong Kong SAR, China * Corresponding authors: Y. Y. ([email protected]) Abstract The rise of nanotechnology has been propelled by low dimensional metals. Albeit the long perceived importance, synthesis of freestanding metallic nanomembranes, or the so-called 2D metals, however has been restricted to simple metals with a very limited in-plane size (< 10 m). In this work, we developed a low-cost method to synthesize 2D metals through polymer surface buckling enabled exfoliation. The 2D metals so obtained could be as chemically complex as high entropy alloys while possessing in-plane dimensions at the scale of bulk metals (> 1 cm). With our approach, we successfully synthesized a variety of 2D metals, such as 2D high entropy alloy and 2D metallic glass, with controllable geometries and morphologies. Moreover, our approach can be readily extended to non-metals and composites, thereby opening a large window to the fabrication of a wide range of 2D materials of technologic importance which have never been reported before. 1 Introduction Freestanding two-dimensional (2D) materials, such as graphene and MoS2, have been attracting tremendous interests over the past decade. Because of their unique nanostructures and reduced dimensionality1, these 2D materials usually exhibit exceptional physical and chemical properties2 -- 5 and therefore have a great potential to be used in various advanced applications, such as highly sensitive sensors6, flexible electronics7,8, catalysts9, and macromolecule separation systems10,11. In recent years, the family of 2D materials has been expanding, including not only single-layered graphene12 and MoS2 13 but also various ultrathin freestanding metallic films or the so-called 2D metals14. The atomic bonding in metals is intrinsically three dimensional. Therefore, metals were usually not considered as a typical 2D material. According to the recent first-principles calculations, single layered metals are thermodynamically stable only when their lateral size is less than 2 nm 15,16. This poses a significant limitation on the size of 2D metals. In practice, it is extremely difficult to handle and deploy such tiny 2D metals for any real applications. Alternatively, one can increase the thickness of metals to overcome the issue of thermodynamic stability while keeping their surface-volume ratio high. The metals of such geometry are 2D in the sense that their physical and chemical properties are determined mainly by their surface. As seen in Table 1, the 2D metals reported so far are all based on a single element, having a thickness (t) ranging from 0.8 nm to 50 nm, a lateral dimension (L) from ~20 nm to ~8000 nm and an aspect ratio (L/t) from ~10 to ~103. Despite these geometric and chemical limitations, the obtained 2D metals displayed prominent and unusual properties that are impossible for their bulk counterparts17,18. For instance, 2D Au, Ag and Bi exhibit intense localized surface 2 plasmon resonance (LSPR)19,20, enabling the development of LSPR-based sensors21. In general, 2D metals possess good electrical conductivity, excellent flexibility, and highly active surfaces, making them a very competitive candidate material for advanced flexible electronics22,23 and clean energy production24,25. To synthesize and fabricate 2D metals, people developed numerous top-down methods, such as mechanical compression26 and nanolithography27, as well as bottom-up methods, such as solution-based chemical synthesis28 -- 30. Given the versatility and efficiency in handling pure metals, especially the noble ones30 -- 34, the solution based chemical synthesis method drew great attention and has been used to produce various 2D metals over the past decade. However, the shortcomings of the above methods are also evident. First, they were originally designed only for one or two specific pure metals and are difficult to be adapted to producing chemically and structurally complex alloys, such as 2D metallic glass and 2D high entropy alloys. Second, the in-plane size of the reported 2D metals was limited compared to that of graphene or MoS2, which already reached the wafer scale according to the recent works35,36. In this work, we would like to develop a new method to address the above issues. Our method is versatile and capable of synthesizing 2D metals of extended chemical compositions. While their thickness (t) can be kept as thin as ~10 nm, their lateral dimension (L) can reach as large as ~10 mm with the corresponding aspect ratio (L/t) extended to , which is about three orders of magnitude larger than those of the existing 2D metals 24,37 -- 40. 3 610Table 1. Comparison of the thickness and in-plane dimension of the 2D metals obtained via different methods Thickness (t) Lateral scale (L) Synthesis method Metal Type (nm) (nm) L/t Ref. Pd Pt Ag Cu Organic ligand-confined growth 2D template-confined 0.8 2 0.8 100-200 100-200 41 500 250 33,42 50-500 50-500 5-10 50-100 10 10-40 3000-8000 200-300 2 50 2 500 250 500-1500 10-30 20 ~10 43 28 32 44 45 29 growth Au 2.4 200-500 80-200 34 Hydrothermal method Nanoparticle assembly Mechanical compression Rh Pt Al / Ag Exfoliation method Sb 0.4 10 2 4 Ti / HEA / 1000- 500-600 1500 50-100 5-10 7300 ~3500 1000 250 46 47 26 25 PSBEE (Present work) MG 10-50 104-107 102-106 This work Note HEA represents high entropy alloy; MG represents metallic glass. Results 4 Fabrication of 2D metals Figure 1(a) illustrates the critical steps of the experimental method we developed to synthesize 2D metals. In our experiments, we selected the polyvinyl alcohol (PVA) hydrogel as the substrate material for the deposition of metallic thin films. As shown in Fig. 1(a), we prepared the precursory PVA substrate by 3D printing (see Methods). The as-printed PVA substrate exhibited surface striations along the printing direction (Fig. 1b-c). After that, we pressed the precursory PVA substrate against a single crystal Si (111) wafer with the root-mean-square surface roughness Rrms = 0.5 nm at an elevated temperature of 120℃ to 180℃ for a period of 60 to 180 seconds for surface smoothening. As a result, the Rrms of the hot embossed PVA could be reduced to ~ 1.6 nm (Fig. 1d-e). To facilitate handling, we attached the PVA substrate to a polyimide (PI) membrane during hot embossing. Subsequently, we deposited a thin metallic film with the desired chemical composition onto the PVA substrate via magnetron sputtering (Fig. 1f, see the details in Methods). Once film deposition was completed, we soaked the film-substrate system into pure water. Interestingly, with the swelling of the PVA substrate, the entire metallic film could exfoliate easily from the PVA substrate within minutes, floating in the water as a freestanding 2D material (please see Supplementary Video 1 for demonstration). In sharp contrast to all previous methods 3,21,46,48, we can easily fabricate 2D metals using our current method, which possesses a thickness at the scale of 10 nm and an exceedingly large in-plane dimension at the scale of 10 mm, as seen clearly in Figs. 1h-i and Supplementary Video 1. 5 Figure 1 Synthesis of 2D metals with our method of polymer surface buckling enabled exfoliation (PSBEE). (a) Schematic illustration of the fabrication processes of 2D metals based on PSBEE, which involve the preparation of a Polyvinyl alcohol (PVA) substrate (step 1), thermal surface processing (step 2 -3), film deposition (step 4) and film exfoliation in water (step 5-6). (b) Photo of the as-printed PVA plate. (c) 3D surface topography of the as-printed PVA plate. (d) Photo of the thermally treated PVA substrate. (e) 3D surface topography of the thermally treated PVA substrate. (f) Photo of the Ti film deposited on the PVA substrate. (g) 3D surface topography of the Ti-PVA system. (h-i) Photos of the Ti film peeling off entirely from the PVA substrate. Aside from the large area of the fabricated 2D metals, our method is applicable to a variety of chemical compositions, particularly the multi-component compositions which are not usually accessible by the traditional methods 5,14,41. Figs 2 a-c show the scanning electron microscopy (SEM) images of three kinds of 2D metals fabricated using our method, i.e. 2D 6 pure Titanium (Ti) (Fig. 2a), 2D FeCoNiCrNb high entropy alloy (HEA) with the nominal composition of FeCoNiCrNb0.5 (in atomic percentage) (Fig. 2b), and 2D ZrCuAlNi metallic glass (MG) with the nominal composition of Zr53Cu29Al12Ni6 (Fig. 2c). Regardless of the chemical difference of these 2D metals (see Supplementary Figs 1-3 for the details of their chemical composition), their lateral dimension can easily reach the millimeter scale (Supplementary Fig. 4). To facilitate observation, these 2D metals were suspended across the circular hole of a copper grid with a diameter of ~60 m. Interestingly, we could observe severe folding and some micro-cracking in these 2D metals; however, these damages could not spread over to cause overall breakage, hence implicative of good mechanical toughness that can afford the handling of the 2D metals during subsequent collection, transportation, and fixation. Figs 2 d-f show the high-resolution transmission electron microscopy (HRTEM) images of the 2D metals. Evidently, the 2D Ti is of a nanocrystalline structure (Fig. 2d), while the 2D HEA and 2D MG are of an amorphous structure (Figs. 2e-2f). These results are promising, which demonstrates that we could vary the atomic structure of the 2D metals by altering their chemical composition. To measure the thickness, we transferred the 2D metals from the Petri dish where they were fabricated to a Si substrate. As seen in the atomic force microscope (AFM) images (Figs. 2 g-i), it is clear that the detected thickness is ~21 nm for the 2D Ti (Fig. 2g), ~36 nm for the 2D HEA (Fig. 2h), and ~51nm for the 2D MG (Fig. 2i). In general, these thickness values are comparable to those of other 2D metals reported in the literature 28,32,45, being tabulated in Table 1. Notably, the thickness of the 2D metals fabricated via the current method can be easily reduced or increased by properly adjusting the processing parameters of magnetron sputtering, such as deposition time. 7 Figure 2 Structural and geometric characterization of the as-fabricated 2D metals. (a-c) SEM images of the 2D Ti, 2D FeCoNiCrNb and 2D ZrCuAlNi obtained via PSBEE, which were suspended over the copper grid for observation. (d-f) HRTEM images of the as-fabricated 2D Ti, 2D FeCoNiCrNb and 2D ZrCuAlNi. The insets show the corresponding fast fouier transformation (FFT) patterns of the HRTEM images. (g-i) AFM images of the 2D Ti, 2D FeCoNiCrNb and 2D ZrCuAlNi transferred to the Si substrate. The insets show the line scan across the edge of the 2D metals. Elastic characterization of the as-fabricated 2D metals After we obtained the 2D metals, we characterized their elastic moduli, which are relevant to 8 the mechanics of film exfoliation as discussed in the later text, with nanoindentation modulus mapping (see Supplementary Figs.5a-c, and Methods for the experimental details). Interestingly, we found that the 2D Ti exhibited a slightly higher elastic modulus than thick Ti films (thickness > 1 m) directly deposited on the silicon wafer (see Supplementary Fig. 5 d). Because TiO2 possess a higher modulus than pure Ti according to the literature data49,50, this behavior of stiffening, as observed on the 2D Ti, might be attributed to the formation of surface oxides. To confirm this, we carried out X-ray photoelectron spectroscopy (XPS) experiments on the 2D Ti. As seen in Supplementary Fig. 6, the presence of TiO2 is evident on the surface of the 2D Ti. Notably, when the thickness of the 2D Ti increases from ~20 nm to ~40 nm, its modulus reduces and approaches to that of the thick film, hence implicative of the diminishing effect of the surface oxides (Supplementary Fig. 5d). In contrast, the amorphous 2D FeCoNiCrNb and 2D ZrCuAlNi appeared softer than the corresponding thick films deposited on the silicon wafer. This behavior of softening agrees with the longstanding theory that the glass transition temperature and elastic modulus of amorphous films would reduce if the film thickness was decreased to the nanometer scale51,52. In other words, the effect of surface oxides on the 2D FeCoNiCrNb and 2D ZrCuAlNi, if there is any, is not as significant as that on the 2D Ti. As expected, one can clearly see that the surface of the obtained freestanding metal films plays a important role in their elastic properties. A more in-depth study is certainly needed to understand the mechanical behavior of these 2D metals, which however is already out of the scope of the current work and will be addressed soon. Theoretical modeling and finite element simulation of film exfoliation When water molecules diffuse into a PVA without physical constraints, the PVA swells under 9 osmotic pressures. However, when a hard metal film was deposited onto the surface of the soft PVA, the free swelling would be constrained, which could cause the surface of the PVA to buckle53 -- 55, as illustrated in Figs. 3a-b. Interestingly, we observed that film exfoliation in our experiments was usually preceded by surface buckling after the film-substrate system was immersed in water (Fig. 3c, See Supplementary Video 2). According to Ref56 , the wavelength ( and amplitude () of the buckled surface (see the inset of Fig. 3d for the definition of  and ) being constrained by a hard film can be derived as and , where , Ef and Es denote respectively the reduced elastic modulus of the film and substrate; hf represents the film thickness; and cs denotes the strain of the PVA substrate constrained by the film. To form buckling, cs has to reach the critical strain or above. Figs. 3e-f show the contour plots of the dimensionless variable /hf and /hf as a function of cs and Ef/Es respectively. Evidently, the dimensionless amplitude (/hf) increases with the increasing cs and Ef/Es; by comparison, the dimensionless wavelength /hf mainly depends on Ef/Es and rises with Ef/Es. Note that the white region in Figs. 3e-f corresponds to cs < c or the regime of no surface buckling. Physically, surface buckling induces an out-of-plane stress along the film-substrate interface. In our case, the out-of-plane stress should be tensile near the 'valley' of the buckled surface (see the inset of Fig. 3d) while compressive near the "peak." The maximum tensile out-of-plane stress, which occurs at exactly the bottom of the valley, can be expressed as56 where As seen in Fig. 3g, similar to /hf, the dimensionless variable z/Es increases with cs and Ef/Es. In principle, 10 1/31/32/311ffscshEE1/3/111fcsccsh51/32cscs2/313/4csfEE3422211241ffcszffcsEhAEhAA1/31/31/3ffsAhEEone can envision that a sufficiently high z would initiate film debonding and hence enables film exfoliation. Figure 3 Theoretical modeling and finite element simulation of surface buckling enabled film exfoliation. Illustration of (a) the metallic film deposited on the PVA substrate, (b) water diffusion induced surface instability in the film-substrate system, and (c) complete film exfoliation subsequent to surface buckling. (d) The illustrated buckling geometry described by 11 the wavelength () and amplitude () of the surface oscillation together with the out-of-plane stress z acting at the film-substrate interface. Contour plots of (e) /hf , (f) /hf and (g) z/Es as a function of cs and Ef/Es. (h) Finite element simulations of the constrained swelling induced interfacial crack initiation at the buckled film-substrate interface. (i) Illustration of the geometry of the growing interface crack driven by continuously film stretching. (j) Plot of the critical film stress c versus the tensile strain ns in the film for various buckling geometries. (k) Contour plot of c/Es as the function of Ef/Es and /. To gain quantitative insights into the process of film exfoliation, we carried out extensive finite element (FE) simulations by using a plane strain model with a cohesive-zone interfacial element (see Methods). To mimic the swelling of the PVA substrate constrained by the hard film, we assumed that the thermal expansion coefficient of the substrate was finite while that of the film was negligible. After that, we raised the overall temperature of the substrate-film system, which caused the substrate to expand in a similar fashion as swelling. However, because of the mismatch in the thermal expansion coefficient, the hard film impedes the substrate expansion. As a result, we can compute the constrained strain cs of the substrate based on the setting of the temperature change in our FE simulations (see Methods). Meanwhile, in order to initiate surface buckling, the substrate-film interface took on a sinusoidal profile with a very small amplitude, which mimics surface perturbation in real experiments. As seen in Supplementary Fig.7, the amplitude of the perturbation grows with the substrate expansion, indicative of surface buckling. As shown in Fig. 3h, the surface buckling results in stress localization, leading to an out-of-plane compressive stress around the 'peak' while an out-of- plane tensile stress around the 'valley.' This stress localization at the "valley" continues to 12 grow with cs, which then causes excessive interfacial separation or debonding at a low cs value if the interfacial toughness is small. By comparison, a tougher interface can tolerate a higher cs value before debonding occurs (see Supplementary Fig.8). However, regardless of the interfacial toughness, the general trend remains the same, which agrees with the theoretical modeling. Once an interfacial crack is successfully initiated at the bottom of the 'valley', it can grow further if there is a tensile stress in the metallic film57. Assuming that the edge of the film can be stretched continuously at the strain ns, as depicted in Fig. 3i, we can study the growth of the interfacial crack for different geometries of surface buckling. As shown in Fig. 3j, when ns increases, the film stress xx rises initially but drops at a certain point when xx reaches a critical value c. The sudden drop of the film stress marks the onset of the interfacial crack growth. Therefore, we can take c as the indicator for the easiness or difficulty for the interfacial crack growth. Depending on the geometry of the buckled surface, the film stress either continues to decline or goes up again with further crack growth. In theory, the interfacial crack growth is driven by the tensile stress in the film, which can be developed during film deposition and substrate swelling. According to the dimensional analysis58, we can easily derive that c/Es =f(/, Ef/Es) for a given  value, where f is a continuous function that can be determined by the FE simulations. As shown in Fig. 3k, the c/Es value is determined mainly by the ratio of /when the Ef/Es value is large. The general trend is that the larger is the / ratio the smaller is c. In other words, the rougher is the buckled surface the easier is interfacial cracking. Here, it is worth noting that the c value decreases with the decreasing Ef/Es if / is very small, as manifested by the region on the left bottom corner of Fig. 3k. On the other hand, as seen in 13 Supplementary Fig. 9, the / ratio also scales with Ef/Es and cs in a similar manner as z/Es. Therefore, the Ef/Es ratio and cs determine not only interfacial crack nucleation but also interfacial crack propagation. In principle, one needs to raise cs for the nucleation and propagation of an interfacial crack; however, the effect of Ef/Es is twofolds: it needs to be increased for surface roughening (Supplementary Fig. 9) and the nucleation of the interfacial crack (Fig. 3g) but needs to be decreased for its subsequent propagation (Fig. 3k). This is a rather complicated process which requires one to carefully tune Ef/Es and cs in such a way that film exfoliation in real experiments can be facilitated. Discussion Film stress and metal-hydrogel interfacial toughness Aside from the elastic mismatch, a tensile stress in the film is required to drive an interfacial crack to propagate after its nucleation. Because of the amorphous nature of our hydrogel substrate and the low substrate temperature during film deposition, the common mechanisms for the development of residual stress, such as lattice mismatch and atom diffusion into grain boundaries 59,60, are unlikely to apply. Therefore, the coalescence of atom "islands" during film deposition, as discussed in the literature61 -- 63, offers the possible mechanism for the development of tensile residual stress in our thin metallic films. As seen in Supplementary Video 1, the substrate-film system clearly bent towards the side of the metallic film right after the film deposition, hence indicative of tensile film stress. In addition to the tensile film stress, the relative low toughness of the metal-hydrogel interface also facilitates the process of film exfoliation. According to Ref.64, the interfacial toughness between hydrogel and metal could 14 be very weak, particularly so in a liquid environment, such as water. Physically, it was already shown via XPS in the early works65 -- 67 that the atomic/molecular bonding between hydrogel and metal is mainly through surface oxides and metallic atoms. However, the diffusion of water molecules into the interface can easily undermine such weak bonds with the assistance of flexible C-C bonds68. Therefore, the successful film exfoliation from the PVA substrate immersed in water, as witnessed in our experiments, reflects a rather weak hydrogel-metal interface, which is consistent with the prior results reported in the literature64. Transition of the morphology of 2D metals: From sheets to scrolls As aforementioned, film exfoliation in our experiments depends on the ratio of Ef/Es and the constrained strain cs. Here, we would like to demonstrate that one may change the morphology of the obtained 2D metals by tuning these two design parameters. To this end, we prepared solution based PVA (S-PVA) as an alternative substrate for the deposition of metal films (see Methods). We measured the modulus of the S-PVA and the thermally pressed PVA (T-PVA) (Fig. 1) using dynamic mechanical analyses (DMA) (see Methods). Compared to T-PVA, which has an elastic modulus of ~2 GPa at room temperature, S-PVA has a much lower elastic modulus (~0.1 GPa) because of its higher concentration of water molecules. To understand the swelling behavior of these two different substrates, we soaked them into water and measured their in-plane strain with digital image correlation (DIC) analyses (see Methods). As seen in Supplementary Fig. 10, our DIC results clearly show that, under the same experimental condition, S-PVA undergoes free swelling to a lesser degree than T-PVA, which suggests a lesser degree of constraint or a smaller strain cs in the hydrogel-metal system. In such a case, one could envision that the mechanics of surface buckling would lead to a different film 15 exfoliation behavior if one replaces T-PVA with S-PVA for the synthesis of 2D metals. In general, the S-PVA substrate tends to yield a larger Ef/Es ratio but smaller cs than the T-PVA substrate. Consequently, surface buckling with a large /ratio becomes relatively difficult on the S-PVA, therefore defying interfacial crack nucleation. In such a case, film exfoliation, if there is any, tends to take place from existing cracks or sites of geometric discontinuity, such as the edges of a metallic film. Consequently, the 2D metals tend to roll up into scrolls for the release of the residual stress 23 (see Supplementary Video 3) rather than peel off as a sheet from the hydrogel substrate (see Fig. 4a). Fig. 4b shows the nanoscrolls obtained by applying the S-PVA to the FeCoNiCrNb-hydrogel system. Here, it may be worth mentioning that, compared to the freestanding FeCoNiCrNb nanosheet (Fig. 4c), the FeCoNiCrNb nanoscrolls produced by rolling-up delamination could be more useful in various energy-related applications, such as electrocatalysis, because of their curved morphology. 16 Figure 4 Transition of the morphology of the 2D FeCoNiCrNb from nanosheets to nanoscrolls. (a) Illustration for the transition from sheets to scrolls as controlled by cs and Ef/Es. Note that the color represents the easiness of interfacial crack initiation. (b) SEM images of the FeCoNiCrNb nanoscrolls on Si. (c) SEM images of the FeCoNiCrNb free-standing nanosheets on a copper grid. Control of the geometry of 2D metals Apart from the diversity of atomic structure, chemistry and morphology of the obtained 2D metals, the other advantage of our current method is that we are now able to fabricate 2D metals with a controlled shape in mass production. From the application viewpoint, this is of great 17 importance to biomedical engineering because the shape of 2D metals can affect how these metallic nanosheets enter and interfere with human cells69,70. To synthesize 2D metals with a controlled shape, we can cover the PVA surface (see Fig. 1d) with different masks (see Fig. 5 a-e) during film deposition. As a result, the deposited metallic film on the patterned PVA surface can exfoliate and form 2D metals of the selected shape, as exemplified by Fig. 5 f-h. Here, it should be stressed that the final shape of the 2D metals depends not only on the mask but also on the thickness and physical properties of the material. Our current results suggest that the underlying physics of the shape control for our obtained 2D metals may be related to the mechanics of folding and self-organization, which we will address in the near future. Figure 5 Controlling the geometry of the fabricated 2D metals. (a) Illustration of the 18 deposition process with a mask. (b) Low magnification photo of the deposition mask. Photos of the deposition masks with (c) round, (d) square and (e) hexagon holes. The insets show the high magnification images of the masks. (f-h) SEM images of the round, square and hexagon shaped 2D FeCoNiCrNb fabricated via the masked deposition. Beyond the synthesis of 2D metals Although the current work is focused on 2D metals, however, our method is applicable to other materials owing to the versatility of the state-of-the-art deposition techniques. Very interestingly, the outcome of our research already demonstrated that, in addition to the large- sized 2D metals, one can also obtain large-sized non-metallic nanosheets, such as the free- standing amorphous silicon nanosheets (Supplementary Figs 11a-c) and even large-sized composite nanosheets, such as the freestanding Ag-C composite nanosheet (Supplementary Figs 11b-d). On one hand, synthesis of these freestanding nanosheets, either metallic or non- metallic, greatly enrich the family of 2D materials, which opens up a large window towards the discovery of new 2D materials with technologic significance; on the other hand, we also believe that the chemical, structural, morphologic and geometric diversities brought about by our method, which is not seen in the prior works, also give rise to new opportunities for a variety of applications, which could be related to bioengineering and nanotechnology, such as nano-medicine69 and nano-sensing71. Methods 3D printing of precursory PVA The precursory PVA substrate was prepared using the commercial Ultimaker® 3D printer with 19 PVA filaments. The extruder (AA0.4 type) was heated to about 500 K before a single layer of PVA was printed at the speed of 20 mm/s on a cold glass working plate at room temperature. After printing, the precursory PVA was removed from the working plate for hot pressing. Magnetron sputtering The direct current (DC) magnetron sputtering system was used to deposit metal films on the treated PVA surface. The sputtering targets had a high chemical purity and custom made by the local companies. Before film deposition, we applied a 1min plasma etching to the target to remove oxides or contamination. The deposition process of the three targets (Ti, ZrCuAlNi, and FeCoNiCrNb) all worked at 100W for 4min with Ar (99.9% purity) as the working gas at an overall pressure of 10 mTorr. During the deposition process, the target-substrate distance was kept at around 15cm with the substrate rotating at 10 rpm. Nanoindentation modulus mapping We performed the modulus mapping tests using the TI-950 Hystron nanoindentation platform under the contact mode. The transducer is equipped with a Berkovich diamond tip with a tip radius of ~180nm. The 2D metals (Ti, ZrCuAlNi, and FeCoNiCrNb) were placed on the silicon wafer to facilitate the measurement. During the modulus mapping, the tip was scanned over an area of 10×10 m with the oscillation frequency of 200 Hz. The set-point force and load amplitude applied were 5-7 N and 2-3 N, respectively, depending on the property of the 2D metals. The stiffness, contact force and oscillation amplitude were directly measured from the sample response. Based on the estimated tip radius, we can easily calculate the elastic modulus of the 2D metals. In the meantime, we deposited relatively thick films (>1 m) on the Si wafer and measured their elastic modulus as the benchmark. 20 Finite element (FE) analysis The FE simulation was carried out using the commercial package ANSYS®. The 8-node plane element Plane183 was used to mesh the model. The thermal expansion coefficient of the substrate was set to 2.5×10-3 K-1 while that of the film was set to 1×10-10 K-1, which was negligibly small. Therefore, we could mimic the solvent-induced substrate swelling by assigning uniform temperature increment to the FE model. To simulate interfacial debonding, cohesive zone material (CZM) model was adopted in the mesh of the film-substrate interface. During the simulations, we tuned the interfacial toughness by adjusting max, the maximum normal traction at the interface, and 𝛿 𝑛, the normal separation across the interface where the maximum normal traction is attained. The interfacial toughness could be easily calculated as = e𝑚𝑎𝑥𝛿 𝑛, where e is the Euler's number. Preparation of solution based PVA (S-PVA) The solution based PVA (0.02g/ml) was prepared by dissolving PVA filaments into deionized water, which was subsequently spread evenly onto a polyimide support. After curing for about 1 hour at 360 K in the drying oven, the solution formed a layer of PVA membrane on the polyimide surface. After that, the membrane was transferred into a vacuum chamber for storage before usage. Dynamic mechanical analyses (DMA) We used the METTLER® TOLEDO DMA system to measure the elastic modulus of the T- PVA and the S-PVA. The PVA samples with controlled geometry were fixed on the titanium clamps with the effective length of 15mm and tested under tension mode. The stretching force oscillated around the setpoint of 1.125N with the amplitude of 0.75N at a frequency of 1Hz. 21 The temperature of the chamber increased from room temperature (298K) to 430K at 2K/s during the test. The modulus of the T-PVA and the S-PVA were calculated using the storage modulus and loss modulus obtained from the tests. Digital image correlation To carry out the digital image correlation, we first introduced the tracking speckles onto the PVA surface. This was done by spraying 0.5 m diamond particles onto a flat clean glass plate. After that, the glass plate was placed into a drying cabinet for about 10 minutes to let the solvent evaporate thoroughly. After that, the diamond particles was transferred onto the PVA surface as tracking speckles through mechanical contact. Next, we immersed the PVA into deionized water and recorded the swelling processes using an optical microscope (Leica DM 2700M). The digital image correlation analysis was carried out on the commercial Vic-2D software and the picture at the onset of swelling was selected as the reference. Acknowledgments The research of YY is supported by the Research Grant Council (RGC) through the General Research Fund (GRF) with the grant number CityU11213118 and CityU11209317. Part of the research is supported by City University of Hong Kong with the project number 9610391. The authors declare no conflict of financial interest. Contribution of authors YY supervised the project; TW and YY conceived the idea; TW developed the method and performed the finite element simulations; QH carried out the electron microscopy tests and image correlation analyses; JZ performed the nanoindentation modulus mapping; DZ 22 developed the experimental set-up for 3D printing and carried out the XPS analyses; FL designed the experiments for the synthesis of 2D metals with engineered shape and size. TW and YY wrote the manuscript. All authors all contributed to the analysis and interpretation of the data and have opportunities to comment on the final version of the manuscript. References 1. Naguib, M., Mochalin, V. N., Barsoum, M. W. & Gogotsi, Y. 25th Anniversary Article: MXenes: A New Family of Two-Dimensional Materials. Adv. Mater. 26, 992 -- 1005 (2014). 2. 3. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science 306, 666 -- 9 (2004). Gupta, A., Sakthivel, T. & Seal, S. Recent development in 2D materials beyond graphene. Prog. Mater. Sci. 73, 44 -- 126 (2015). 4. Yang, X. et al. Liquid-mediated dense integration of graphene materials for compact capacitive energy storage. Science 341, 534 -- 7 (2013). 5. Tan, C. et al. Recent Advances in Ultrathin Two-Dimensional Nanomaterials. Chem. Rev. 117, 6225 -- 6331 (2017). 6. Jiang, C., Markutsya, S., Pikus, Y. & Tsukruk, V. V. Freely suspended nanocomposite membranes as highly sensitive sensors. Nat. Mater. 3, 721 -- 728 (2004). 7. Yang, C. et al. Silver Nanowires: From Scalable Synthesis to Recyclable Foldable Electronics. Adv. Mater. 23, 3052 -- 3056 (2011). 8. Ahn, J.-H. et al. Heterogeneous three-dimensional electronics by use of printed semiconductor nanomaterials. Science 314, 1754 -- 7 (2006). 9. Luo, B., Liu, G. & Wang, L. Recent advances in 2D materials for photocatalysis. Nanoscale 8, 6904 -- 6920 (2016). 10. Striemer, C. C., Gaborski, T. R., McGrath, J. L. & Fauchet, P. M. Charge- and size-based separation of macromolecules using ultrathin silicon membranes. Nature 445, 749 -- 753 (2007). 11. Gorgojo, P. et al. Ultrathin Polymer Films with Intrinsic Microporosity: Anomalous Solvent Permeation 12. 13. 14. and High Flux Membranes. Adv. Funct. Mater. 24, 4729 -- 4737 (2014). Geim, A. K. Graphene: status and prospects. Science 324, 1530 -- 4 (2009). Splendiani, A. et al. Emerging Photoluminescence in Monolayer MoS 2. Nano Lett. 10, 1271 -- 1275 (2010). Chen, Y. et al. Two-Dimensional Metal Nanomaterials: Synthesis, Properties, and Applications. Chem. Rev. 118, 6409 -- 6455 (2018). 15. Nevalaita, J. & Koskinen, P. Atlas for the properties of elemental two-dimensional metals. Phys. Rev. B 97, 035411 (2018). 16. Chen, S. & Zeng, X. C. Interaction between Iron and Graphene Nanocavity: Formation of Iron Membranes, Iron Clusters, or Iron Carbides. ACS Appl. Mater. Interfaces 9, 12100 -- 12108 (2017). Blakely, J. M. Introduction to the properties of crystal surfaces. (Pergamon Press, 1973). Lang, N. D. & Kohn, W. Theory of Metal Surfaces: Charge Density and Surface Energy. Phys. Rev. B 1, 17. 18. 4555 -- 4568 (1970). 19. Bastys, V., Pastoriza-Santos, I., Rodríguez-González, B., Vaisnoras, R. & Liz-Marzán, L. M. Formation 23 of Silver Nanoprisms with Surface Plasmons at Communication Wavelengths. Adv. Funct. Mater. 16, 766 -- 773 (2006). 20. Elin M. Larsson, †, Joan Alegret, †, Mikael Käll, † and & Duncan S. Sutherland, ‡. Sensing Characteristics of NIR Localized Surface Plasmon Resonances in Gold Nanorings for Application as Ultrasensitive Biosensors. (2007). doi:10.1021/NL0701612 21. 22. Hussain, N. et al. Ultrathin Bi Nanosheets with Superior Photoluminescence. Small 13, 1701349 (2017). Akinwande, D. et al. A review on mechanics and mechanical properties of 2D materials -- Graphene and beyond. Extrem. Mech. Lett. 13, 42 -- 77 (2017). 23. Chen, Z., Huang, G., Trase, I., Han, X. & Mei, Y. Mechanical Self-Assembly of a Strain-Engineered Flexible Layer: Wrinkling, Rolling, and Twisting. Phys. Rev. Appl. 5, 017001 (2016). 24. Zhao, H., Chen, X., Wang, G., Qiu, Y. & Guo, L. Two-dimensional amorphous nanomaterials: synthesis and applications. 2D Mater. 6, 032002 (2019). 25. Gu, J. et al. Liquid-Phase Exfoliated Metallic Antimony Nanosheets toward High Volumetric Sodium Storage. Adv. Energy Mater. 7, 1700447 (2017). 26. 27. Liu, H. et al. 2D Metals by Repeated Size Reduction. Adv. Mater. 28, 8170 -- 8176 (2016). Yu, Q., Guan, P., Qin, D., Golden, G. & Wallace, P. M. Inverted Size-Dependence of Surface-Enhanced Raman Scattering on Gold Nanohole and Nanodisk Arrays. Nano Lett. 8, 1923 -- 1928 (2008). 28. Mathieu Maillard, †, Suzanne Giorgio, ‡ and & Marie-Paule Pileni, †. Tuning the Size of Silver Nanodisks with Similar Aspect Ratios: Synthesis and Optical Properties. (2003). doi:10.1021/JP022357Q 29. Salzemann, C., Urban, J., Lisiecki, I. & Pileni, M.-P. Characterization and Growth Process of Copper Nanodisks. Adv. Funct. Mater. 15, 1277 -- 1284 (2005). 30. Yujie Xiong, † et al. Poly(vinyl pyrrolidone): A Dual Functional Reductant and Stabilizer for the Facile Synthesis of Noble Metal Nanoplates in Aqueous Solutions. (2006). doi:10.1021/LA061323X 31. Song, Y. et al. Synthesis of Platinum Nanowheels Using a Bicellar Template. J. Am. Chem. Soc. 130, 12602 -- 12603 (2008). 32. Deng, Z., Mansuipur, M. & Muscat, A. J. New Method to Single-Crystal Micrometer-Sized Ultra-Thin Silver Nanosheets: Synthesis and Characterization. J. Phys. Chem. C 113, 867 -- 873 (2009). 33. Garcia, R. M. et al. Templated growth of platinum nanowheels using the inhomogeneous reaction environment of bicelles. Phys. Chem. Chem. Phys. 13, 4846 -- 4852 (2011). 34. 35. 36. Huang, X. et al. Synthesis of hexagonal close-packed gold nanostructures. Nat. Commun. 2, 292 (2011). Lee, Y. et al. Wafer-Scale Synthesis and Transfer of Graphene Films. Nano Lett. 10, 490 -- 493 (2010). Shim, J. et al. Controlled crack propagation for atomic precision handling of wafer-scale two-dimensional materials. Science 362, 665 -- 670 (2018). 37. Sprouster, D., Ruffell, S., … J. B.-J. of M. & 2011, U. Structural characterization of B-doped diamond nanoindentation tips. cambridge.org (2011). 38. Zhang, F. et al. Indentation of a hard film on a soft substrate: Strain gradient hardening effects. Int. J. Plast. 23, 25 -- 43 (2007). 39. Suryanarayana, C. & Inoue, A. Bulk Metallic Glasses, Second Edition. (CRC Press, 2017). doi:10.1201/9781315153483 40. Marton, L. (Ladislaus). Advances in electronics and electron physics. Volume VII. (Academic Press, 1955). 41. Xu, D. et al. Novel surfactant-directed synthesis of ultra-thin palladium nanosheets as efficient electrocatalysts for glycerol oxidation. Chem. Commun. 53, 1642 -- 1645 (2017). 42. Song, Y. et al. Synthesis of Platinum Nanowheels Using a Bicellar Template. J. Am. Chem. Soc. 130, 12602 -- 12603 (2008). 24 43. Yujie Xiong, † et al. Poly(vinyl pyrrolidone): A Dual Functional Reductant and Stabilizer for the Facile Synthesis of Noble Metal Nanoplates in Aqueous Solutions. Langmuir (2006). doi:10.1021/LA061323X 44. Xiong, Y., Washio, I., Chen, J., Sadilek, M. & Xia, Y. Trimeric Clusters of Silver in Aqueous AgNO3 Solutions and Their Role as Nuclei in Forming Triangular Nanoplates of Silver. Angew. Chemie 119, 5005 -- 5009 (2007). 45. Shaik, A. H. & Chakraborty, J. A simple room temperature fast reduction technique for preparation of a 46. 47. copper nanosheet powder. RSC Adv. 6, 14952 -- 14957 (2016). Duan, H. et al. Ultrathin rhodium nanosheets. Nat. Commun. 5, 3093 (2014). Yang, S., Qiu, P. & Yang, G. Graphene induced formation of single crystal Pt nanosheets through 2- dimensional aggregation and sintering of nanoparticles in molten salt medium. Carbon N. Y. 77, 1123 -- 1131 (2014). 48. Zhang, Y. et al. Seedless Growth of Palladium Nanocrystals with Tunable Structures: From Tetrahedra to Nanosheets. Nano Lett. 15, 7519 -- 7525 (2015). 49. Pradhan, S. S., Sahoo, S. & Pradhan, S. K. Influence of annealing temperature on the structural, mechanical and wetting property of TiO2 films deposited by RF magnetron sputtering. Thin Solid Films 518, 6904 -- 6908 (2010). 50. Cao, C. R., Lu, Y. M., Bai, H. Y. & Wang, W. H. High surface mobility and fast surface enhanced crystallization of metallic glass. Appl. Phys. Lett. 107, 141606 (2015). 51. Li, Y. Z. et al. Size effect on dynamics and glass transition in metallic liquids and glasses. J. Chem. Phys. 146, 224502 (2017). 52. Alcoutlabi, M. & McKenna, G. B. Effects of confinement on material behaviour at the nanometre size scale. J. Phys. Condens. Matter 17, R461 -- R524 (2005). 53. Chan, E. P. & Crosby, A. J. Spontaneous formation of stable aligned wrinkling patterns. Soft Matter 2, 324 (2006). 54. 55. Vandeparre, H. et al. Hierarchical wrinkling patterns. Soft Matter 6, 5751 (2010). Yang, S., Khare, K. & Lin, P.-C. Harnessing Surface Wrinkle Patterns in Soft Matter. Adv. Funct. Mater. 20, 2550 -- 2564 (2010). 56. Song, J. et al. Buckling of a stiff thin film on a compliant substrate in large deformation. Int. J. Solids Struct. 45, 3107 -- 3121 (2008). 57. 58. Freund, L. & Suresh, S. Thin film materials: stress, defect formation and surface evolution. (2004). Barenblatt, G. I. Scaling, self-similarity, and intermediate asymptotics. (Cambridge University Press, 1996). 59. Chason, E. A kinetic analysis of residual stress evolution in polycrystalline thin films. Thin Solid Films 526, 1 -- 14 (2012). 60. Doljack, F. A. & Hoffman, R. W. The origins of stress in thin nickel films. Thin Solid Films 12, 71 -- 74 (1972). 61. Hoffman, R. W. Stresses in thin films: The relevance of grain boundaries and impurities. Thin Solid Films 34, 185 -- 190 (1976). 62. 63. Thornton, J. A. & Hoffman, D. W. Stress-related effects in thin films. Thin Solid Films 171, 5 -- 31 (1989). Daniel, R., Martinschitz, K. J., Keckes, J. & Mitterer, C. The origin of stresses in magnetron-sputtered thin films with zone T structures. Acta Mater. 58, 2621 -- 2633 (2010). 64. Awaja, F., Gilbert, M., Kelly, G., Fox, B. & Pigram, P. J. Adhesion of polymers. Prog. Polym. Sci. 34, 948 -- 968 (2009). 65. Burkstrand, J. M. Copper -- polyvinyl alcohol interface: A study with XPS. J. Vac. Sci. Technol. 16, 363 -- 25 365 (1979). 66. Burkstrand, J. M. Metal‐polymer interfaces: Adhesion and x‐ray photoemission studies. J. Appl. Phys. 52, 4795 -- 4800 (1981). 67. Grace, J. M. & Gerenser, L. J. Plasma Treatment of Polymers. J. Dispers. Sci. Technol. 24, 305 -- 341 (2003). 68. Holly, F. J. & Refojo, M. F. Wettability of hydrogels I. Poly(2-hydroxyethyl methacrylate). J. Biomed. Mater. Res. 9, 315 -- 326 (1975). 69. Wang, Z. et al. Biological and environmental interactions of emerging two-dimensional nanomaterials. Chem. Soc. Rev. 45, 1750 -- 1780 (2016). 70. Zou, X., Zhang, L., Wang, Z. & Luo, Y. Mechanisms of the Antimicrobial Activities of Graphene Materials. J. Am. Chem. Soc. 138, 2064 -- 2077 (2016). 71. Chen, M. et al. Core-Shell Pd@Au Nanoplates as Theranostic Agents for In-Vivo Photoacoustic Imaging, CT Imaging, and Photothermal Therapy. Adv. Mater. 26, 8210 -- 8216 (2014). 26 Supplementary Information The Controlled Large-Area Synthesis of Two Dimensional Metals Tianyu Wang1, Quanfeng He1, Jingyang Zhang1, Zhaoyi Ding1, Fucheng Li1, Yong Yang 1,2* 1. Department of Mechanical Engineering, College of Engineering, City University of Hong Kong, Kowloon Tong, Kowloon, Hong Kong SAR, China 2. Department of Materials Science and Engineering, College of Engineering, City University of Hong Kong, Kowloon Tong, Kowloon, Hong Kong SAR, China * Corresponding authors: Y. Y. ([email protected]) Supplementary Figure 1 (a) The composition mapping and (b) corresponding energy spectrum of the 2D Ti. 27 Supplementary Figure 2 (a)-(e) The composition mapping and (b) corresponding energy spectrum of the 2D FeCoNiCrNb. 28 Supplementary Figure 3 (a)-(e) The composition mapping and (b) corresponding energy spectrum of the 2D ZrCuAlNi. 29 Supplementary Figure 4 The optical images of the as-fabricated (a) 2D Ti, (b) 2D and (c) 2D FeCoNiCrNb transferred onto the copper grid in order to facilitate experimental observation. Supplementary Figure 5 The elastic modulus of the 2D metals. The stiffness contour plots of (a) the 2D Ti with a thickness of 21nm, (b) the 2D FeCoNiCrNb with a thickness of 36nm, and (c) the 2D ZrCuNiAl with a thickness of 51nm. (a) the thickness dependency of the elastic modulus measured for different 2D metals, being normalized by the modulus (E0) of the thick film deposited on the Si substrate.Note that E0 = 74.1±4.9 GPa for Ti, 215.5±22.6 GPa for FeCoNiCrNb, and 93.1±5.7 GPa for ZrCuAlNi. The error bars come from multiple experiments. 30 Supplementary Figure 6 The XPS spectrum obtained from the surface of the as- fabricated 2D Ti nanosheet. (a) the signal intensity distribution of the Ti element. (b) the signal intensity distribution of the O element. Supplementary Figure 7 The positive correlation between the relative change of buckling amplitude and the constrained strain of the substrate (cs) obtained from the FE simulations. 31 Supplementary Figure 8 Finite Element simulation of the constrained explansion of the soft substrate with a tougher interface. Supplementary Figure 9 The contour plot of / as a function of the constrained swelling strain cs and the modulus ratio Ef/Es. 32 Supplementary Figure 10 The free swelling of the hydrogel (T-PVA and S-PVA). The optical images of a swelling PVA with surface speckles at different times, including those at (a)10s, (b)18s and (c) 27s. The dashed box indicates the selected area for digital image correlation analyses. The histograms of the change in the highest local principal strain δe1 at different times, including after (d)10s, (e)18s, and (f) 27s. The insets in d-f show the spatial distribution of δe1. (g) The variation of the median value of δe1 with the swelling time. (h) The variation of the mean value of δe1 with the swelling time. Note T-PVA exhibits a higher principal strain δe1 than S-PVA 33 Supplementary Figure 11 The other 2D materials fabricated using PSBEE. (a) The optical image of the 2D amorphous Si on a copper grid. (b) The optical image of the 2D Ag-C composite on a copper grid. The HRTEM images of (c) the 2D amorphous Si and (d) the 2D Ag-C composite. Note that the insets in (c)-(d) are the corresponding FFT image. 34
1807.11748	1	1807	2018-07-31T10:31:56	Visible absorbing TiO2 thin films by physical deposition methods	[ "physics.app-ph" ]	Titanium dioxide is one of the most widely used wide bandgap materials. However, the TiO2 deposited on a substrate is not always transparent leading to a loss in efficiency of the device, especially, the photo response. Herein, we show that atomic layer deposition (ALD) and sputtered TiO2 thin films can be highly absorbing in the visible region. While in ALD, the mechanism is purported to be due to oxygen deficiency, intriguingly, in sputtered films it has been observed that in fact oxygen rich atmosphere leads to visible absorption. We show that the oxygen content during deposition, the resistivity of the film could be controlled and also the photocatalysis response has been evaluated for both the ALD and sputtered films. High resolution TEM and STEM studies show that the origin of visible absorption could be due to the presence of nanoparticles with surface defects inside the amorphous film.	physics.app-ph	physics	Visible absorbing TiO2 thin films by physical deposition methods Litty Varghese, AnuradhaPatra, Biswajit Mishra, DeepaKhushalani, Achanta Venu Gopal Abstract: Titanium dioxide is one of the most widely used wide bandgap materials. However, the TiO2 deposited on a substrate is not always transparent leading to a loss in efficiency of the device, especially, the photo response. Herein, we show that atomic layer deposition (ALD) and sputtered TiO2 thin films can be highly absorbing in the visible region. While in ALD, the mechanism is purported to be due to oxygen deficiency, intriguingly, in sputtered films it has been observed that in fact oxygen rich atmosphere leads to visible absorption. We show that the oxygen content during deposition, the resistivity of the film could be controlled and also the photocatalysis response has been evaluated for both the ALD and sputtered films. High resolution TEM and STEM studies show that the origin of visible absorption could be due to the presence of nanoparticles with surface defects inside the amorphous film. Introduction Titanium dioxide is one of the most commonly used and studied semiconductors in the field of optics, solar cells, integrated optoelectronics and environmental issues due to its high refractive index, excellent optical transmission,large energy band gap, high dielectric constant, very good wear resistance, along with a high chemical resistance against solvents [1-10]. Photocatalytic behavior of TiO2 in UV region has been extensively studied [11-17]. Under UV irradiation, the decomposition of organic compounds has been easily facilitated by TiO2 and it is increasingly being established that the surface of TiO2can be used for self-cleaning and anti-fogging effects [18]. The principle behind photocatalytic reactions involving TiO2 is that though the photoexcited carriers predominantly recombine in a non-radiative manner, few of them diffuse to the surface and take part in a variety of redox reactions such as generation of hydrogen and oxygen (Eq 1), or form different reactive oxygen species (e.g. OH) through simultaneous reduction and oxidation of water or molecular oxygen (Fig. 1). 2H2O (l) + 4h+ = O2 (g) + 4H+ [E0 = 1.23 eV] Eq 1a 4H+ + 4e- = 2H2 (g) [E0 = 0 eV] Eq 1b Fig. 1 Schematic representation of the main processes involved in photocatalysis If any organic pollutant is present, the formed reactive oxygen species react with the pollutants adsorbed on the surface of TiO2 and ideally mineralize them to H2O and CO2. Though this criterion could also be satisfied with a number of other semiconductors [ref] (Reference: A. Ghicov, P. Schmuki, Chem. Commun., 2009, 2791.), the matching of the band edges with the important half reactions to generate reactive oxygen species make TiO2 a special material as a photocatalyst. However, due to energy gap of ~3eV, its efficiencyin absorbing sunlight is minimal as radiation from Sun consists only 5% UV while, predominant, 43% is in the visible and 52% is in infraredregion [19]. To make use of solar energy, viable efforts for reducing the bandgap of TiO2 were started in 1990s. These methods have been mainly based on variation of TiO2 stoichiometry by doping usinga variety of elements [20-21]. As a result of low lying dopant states, improved photo-catalytic activity has been reported [22-23]. Recently, defect engineered black TiO2 nanoparticles [24] and nanowires [25] by hydrogenation have been reported and this has led CBλ<385 nme-h+Eg=3.2 eVVBO2O2.-H+HO2.HO2.O2+H2O2H2O2O2+OH.+ OH-OH-OH.invariably to enhanced visible absorption and photocatalytic activity. Subsequently several other reports have contributed to this area with detailed insight being gained into the manner in which hydrogen incorporation leads to a decrease in the band gap [26]. The development of reduced form of TiO2 by oxygen vacancies as a defect also resulted in visible absorption [27-29] which has improved photo catalytic activity and in some cases splitting of water to form hydrogen has also been observed. It is generally accepted that TiO2 exhibits three crystalline phases. Two are tetragonal phases: rutile and anatase and the third one, brookite, is orthorhombic. Depending on the deposition conditions any of these phases can be realized. Among them rutile is the most thermodynamically stable phase. However anatase phase is known to have better electron mobility. As a result, predominantly this phase has been preferred for photocatalysis. Band structure calculations revealed that rutile and anatase have a direct and indirect band gap, respectively. Although wide range of band-gap energies have been reported for both rutile and anatasephase TiO2 by optical measurements, anatase is known to have higher band-gap energy [31-37]. Techniques used for growing TiO2 films include magnetron sputtering [38, 39], sol-gel method [40], ion-assisted deposition [41,42], pulsed laser deposition [43], molecular-beam epitaxy [44] and atomic layer deposition (ALD) [45-46]. In this paper we present preparation of visible absorbing TiO2 thin films having exclusively rutile phase using two physical vapour deposition techniques:(1) radio frequency (RF)sputtering and (2) atomic layer deposition (ALD). These techniques have been known to be suitable for realizing thin films with controlled structure, composition, desired phase and uniform coating as there is control over the deposition parameters such as temperature, rate and the ambient atmosphere. The thin films have been characterized by x-ray diffraction, white light transmission, resistivity, scanning electron microscope (SEM), Transmission electron microscopy (TEM) and studied their photo-catalytic properties. For ALD, Cambridge Savannah 100 Atomic Layer Deposition System was used for the preparation of uniformly coated thin films. Thin films were deposited on quartz and silicon at the growth temperature of 200οC (Fig.2). Titanium tetrachloride (TiCl4) from Sigma-Aldrich Chemicals Ltd was used as the source for the growth of TiO2 with N2as the carrier gas. The base pressure was 10-5 Torr and operating pressure was 50 mTorr. Pure H2O was used as oxygen precursor. Prevention of gas-phase reactions of the vapors was achieved by evacuation of chamber to a pressure below 10-4Torr between successive pulses. Repetition of requisite number of cycles consisting of alternating H2O and Ti pulses resulted in the formation of TiO2 thin film growth layer by layer with atomic precision. The rate of TiO2 deposition was 0.4Å/cycle. In the second method, sputtered thin films were realized using a confocal sputter system. The base pressure was 4x10-6 mbar and operating pressure was 2x10-2 mbar in Ar plasma. The RF power was 100 W at room temperature with a rotating stage. Ar and O2 gas combination was used. For Ar:O2 ratio of 70:30 sccm, the resulting film was dark. The rate of deposition was 0.67 nm/min. The thickness of thin films was measured using Ambios profilometer and Sentech's Film Thickness probe (FTP) which works in the reflection mode. The SEM studies were done using a ZEISS Ultra 55 plus FESEM. For AFM studies, Park XE70 with Silicon cantilever in non- contact mode was employed and root mean square value (RMS) for roughness was calculated using commercial XEI software. X-ray Diffraction (XRD) measurements were done using X'pert pro XRD with Cu Kɑ ( = 1.54056 Å) source. Data was collected in glancing angle mode with source angle at 1 and analyzer angle varied from 20 to 60 with a step size of 0.025(2ϴ). X-ray Photoelectron Spectroscopy (XPS) was carried out using ESCA from Omicron Nanotechnology. Al Kɑ monochromatic source was used with a hemispherical analyzer. FEI TITAN transmission electron microscope (TEM) was used for transmission electron microscopy studies. The four probe resistivity and photovoltaic studies were done using Signatone 4-probe station, Keithley source-meter and Lock-in amplifier at room temperature. Photoconductive studies were done with the probe station and Lock-in amplifier combination in the presence of white light source (100 W Halogen lamp). Transmission studies were carried out using a Cary 5000 spectrophotometer in the 200 nm to 2500 nm range. Reflection studies were done with Jasco spectrophotometer covering the 200 to 2500 nm wavelength range. Refractive index studies were done using Sentech 800 ellipsometer. The visible image showed that film prepared by ALD was black in color and sputtered ones were dark blue in color. From AFM data, the surface roughness could be obtained to be 1.3 nm for ALD grown film and 0.8 nm for sputtered films (Fig.3a and b). Optical transmission and reflection studies revealed 30% absorption in the visible region for the samples prepared by sputtering in the presence of Ar/O2 compared to the film prepared in the presence of solely Argon (Fig.3a and b). Transmission studies on ALD thin film showed a 50% reduction in transmitted light in the visible region compared to TiO2 deposited in Ar atmosphere (Fig.4a and b). Ellipsometric measurements on thin films showed that refractive index of sputtered thin film was 2.55 and that of ALD film was 2.7 at 500 nm wavelength (Fig.5). Fig.2. AFM image of TiO2 thin film prepared a) by ALD and b) by sputtering Fig.3. Transmission and reflection studies on sputtered TiO2 thin film a) in presence of Argon, b) in presence of Ar/O2 and c) absorption spectrum of sputtered thin film in presence of Ar/O2 3006009001200150018000102030405060708090100 T,R (%)Wavelength (nm) T R40080012001600200024000.00.20.40.60.81.0 Absorbance (a.u.)Wavelength (nm) Fig: 4. a) Transmission and b) reflection studies on ALD grown TiO2 thin film Fig.5. Dispersion measurements on TiO2 thinfilms High resolution Transmission electron microscopy (HRTEM) and Scanning electron microscope (SEM) studies done on these films (Fig.6) showed the presence of highly crystalline defect induced individual TiO2 nanoparticles of around 3 nm in diameter. Clusters of TiO2 nanoparticles with highly resolved lattice features are found in amorphous matrix of TiO2 as shown in Fig.7. Strong peaks obtained from XRD studies showed that highly crystalline TiO2 of rutile phase in both the thin films (Fig.8). The crystallite size calculated from XRD peaks were 29 nm which seems to be in good agreement with the size of clusters shown in TEM. 3004005006007008009000102030405060708090100 T (%)Wavelength (nm)5005506006507007508002.202.252.302.352.402.452.502.55 Refractive indexWavelength (nm) Sputtered ALD20030040050060070080090005101520253035 R (%)Wavelength (nm) Fig.7. Schematic view of TiO2 nanoparticles embedded in amorphous matrix Fig.8. XRD studies on thin films prepared by a) ALD and b) sputtering In order to gain further insight into the nature of the films and the electronic environment around Ti and O centers, XPS measurements were performed. The Ti 2p core level spectrum of TiO2 thin film grown by ALD exhibits symmetric spin orbit split lines at 458.224 eV and 463.999 eV for Ti 2p3/2 and Ti 2p1/2, respectively, with a difference of 5.7 eV between the doublets (Fig. 8). The O 1s region showed a peak at 529.738 eV, characteristic of oxides of transition metals and a 3040506070(b) Intensity (a.u.)2 (deg.)(a)(220)(101)(211)Rutile(110) prominent shoulder at 531.523 eV which is a signature of the presence of disorder in the surface in the form of adsorbed oxygen [47,48]. The peak at 531.523 eV is designated as higher binding energy peak (HBEP) in Fig. 8b. In addition, a broad peak centered at ~400 eV (around N 1s region) is observed and is attributed to the molecularly chemisorbed nitrogen (γ-N2) (Fig.9). However, no trace of atomic nitrogen (β-N, Ti-N bonding) is observed in the XPS spectrum. The O 1s peak in samples annealed in presence of N2 showed a shift of 0.577 eV towards lower binding energy and did exhibit a minor change in intensity. The result indicates that the as-grown ALD samples are oxygen deficient. Fig. 9 (a) XPS Ti 2p spectrum (b) O 1s spectrum of TiO2 film grown by ALD Fig. 10 XPS spectrum of ALD grown TiO2 around N 1s binding energy. 4564584604624644664682 p1/22 p3/2 Counts (a.u.)Binding Energy (eV)TiO2 by ALD As-deposited Annealed528530532534O 1s (HBEP)O 1s Counts (a.u.)Binding Energy (eV)TiO2 by ALD As-deposited Annealed390393396399402405408180200220240260280 Counts (a.u.)Binding Energy (eV) Fig.11 XPS spectra of (a) Ti 2p and (b) O 1s of sputtered TiO2 thin film The XPS spectrum of TiO2 thin film deposited by sputtering in the presence of Argon and oxygen plasma is plotted in Fig.10. The Ti 2p3/2, Ti 2p1/2 and O 1s core levels are located at 457.825 eV, 463.544 eV and at 529.068 eV, respectively. The 2p peaks are symmetric, showed binding energy separation of 5.7 eV and could be fitted with a single Gaussian peak. In addition to the signature of lattice oxygen of TiO2 at 529.068 eV, the O 1s region exhibits a pronounced shoulder at higher binding energy at 531.074 eV. The shoulder peak at higher binding energy has been observed by many research groups [47,48] and is generally attributed to defective oxides or sub-stoichiometric oxide on the surface of TiO2. The sputtered TiO2 films were subjected to heat treatment and the XPS spectrum of the annealed films is also shown in Fig. 10. It is observed that the Ti 2p peaks as well as O 1s peak shift to higher binding energy and there is a considerable enhancement in the recorded signal. Thus, the as-grown sputtered samples are oxygen rich. The Ti-O-Ti bonding strength, however, is stronger in the annealed samples as is evident from the increased signal strength. TiO2 being a wide band gap semiconductor (Eg ~ 3.2 eV) is transparent to visible light. However, surface defects may play a significant role in altering the optical 4564584604624644664682 p1/22 p3/2 Counts (a.u.)Binding Energy (eV)TiO2 by sputtering As-deposited Annealed528530532534O 1s (HBEP)O 1s Counts (a.u.)Binding Energy (eV)TiO2 by sputtering As-deposited Annealedresponse of the films and may lead to visible light absorption [49]. These surface defects, when present in good number introduce mid-gap states within the forbidden region. The formed continuum merges with conduction/valence band edge thus modulating optical behavior. Hence we can conclude from the XPS studies that the presence of excess oxygen results in the visible absorption in sputtered TiO2 thin film which is also supported by a recent report [25]. In addition to structural and composition studies, we also conducted studies on electrical properties of TiO2 thin films using four probe measurements. The measured resistivity was in kΩ-cm for ALD film and in GΩ-cm for sputtered films. For temperature dependent studies, annealing was done on samples at 400°C in atmospheric air for 20 min. The resistivity versus temperature graph (Fig.11a) suggests that the behavior of ALD thin film varies from metal to insulator (MOI) as temperature increases. The resistivity of the sputtered thin film reduces from GΩ-cm to kΩ-cm as the temperature is increased (Fig. 11b). As was inferred from the XPS analysis, the sputtered TiO2 films have large amounts of surface defects (presence of excess Fig.12 Resistivity studies on a) ALD TiO2 thin film and b) sputtered thin film 4080120160200102103104105106107108 (ohm-cm)Temperature (OC)4080120160200100101102103 (ohm-cm)Temperature(C)oxygen). Due to which the resistivity of the film is very high (GΩ-cm). When the film is annealed some of the surface defects may get relaxed and consequently may lead to better electron transport which otherwise was hindered due to the presence of defects. As a result, the resistivity of the sample shows systematic decrease (Fig. 11b) with increase in temperature. The temperature dependent resistivity behavior of ALD grown TiO2 film is different from that of sputtered one. The ALD film shows a decrease in measured resistivity up to 100°C and then a monotonous increase. Although the structure of TiO2 films, irrespective of growth mode (sputtering or ALD), is rutile, the difference in resistivity behavior in the films suggest that there is a considerable difference in surface chemistry. In ALD grown films the chemisorbed nitrogen, owing to low sticking probability, might get desorbed on heat treatment and this may lead to initial decrease in resistivity. However, soon the surface sites are occupied with the chemisorbed oxygen, as it was observed in XPS spectrum of ALD grown TiO2 film (Fig.10b) that the high energy O 1s peak intensity increases significantly on annealing. Thus the surface is densely filled with chemisorbed oxygen which may result in increased resistivity of the film with rise in annealing temperature. As a further to evaluate the effectiveness of these films, photocatalysis experiments were conducted by using directly the films (mounted on the substrates) and suspending them in a solution of a model organic compound (tryptophan). The main impetus for photocatalysis was to evaluate the films (as they have impressive absorption in the visible range) for viability in forming long lived electrons/holes that have strong oxidative/reductive capability. As such, the Fig. 13 Caption films were suspended in a solution of tryptophan and the samples were carefully irradiated with exclusively visible light (long pass filter at 400 nm was used) and the degradation of the model organic compound was monitored by spectrophotometer. Note that the model compound, tryptophan has absorption only in the UV region and as such sensitization artefacts (that lead anomalously to enhanced photocatalysis rates) have been carefully avoided. Shown in Fig 12a is the spectra of tryptophan as a function of time, Fig. 12b shows a comparative plot for the degradation rate of Tryptophan in the presence of a conventional TiO2 catalyst (P25, commercially available) along with the ALD prepared thin film sample. It can be readily observed that the rate of degradation of model compound is faster in the presence of the sputtered TiO2 thin film sample and in fact even the ALD deposited film shows a faster rate than the conventional, oft-cited, standard catalyst P25. This result corroborates the XPS and resistivity data whereby the presence of excess oxygen species in the thin films, leads to longer lifetimes of the excited carriers. These long lived species as a result are able to diffuse readily to the surface and contribute to the formation of large ROS (reactive oxygen species) which are vital for the degradation of the model organic compound. 2002402803203604000.00.20.40.60.81.01.21.4 Absorbance (a.u.)Wavelength (nm) 1 h 12 h 32 h 41 h 49 h0102030405060700.00.10.20.30.40.50.60.7 ln(C0/C)Duration (h) Black TiO2 3 hrs Black TiO2 ALD P25References 1. U.Diebold,"The surface science of TiO2", Surface Science Reports,53,229(2003) 2. F. A. Grant, "Properties of Rutile (Titanium dioxide)", Reviews of Modern Physics, 31, 3 (1956) 3. D. C. Czonzmzvzzf,"Electrical and Optical Properties of Rutile Single Crystals", Phys. Rev. 87, 876 -- 886 (1952) 4. E. Ritter, "Dielectric film materials for optical applications", Phys. Thin Films, 8 (1975) 5. A. Amtout and R. Leonelli ,"Optical properties of rutile near its fundamental band gap", Phys. Rev. B 51, 6842 -- 6851 (1995) 6. M. Gratzel, "Dye sensitized soalrcells",J. Photochem. Photobiol.C 4, 145 (2003) 7. M. Gratzel, "Recent Advances in Sensitized Mesoscopic Solar Cells",Acc. Chem. Res. 42, 1788, (2009) 8. B. O'Regan and M. Grätzel, "A low-cost, high-efficiency solar cell based on dye- sensitized colloidal TiO2 films", Nature353, 737 -- 740(1991) 9. M. R. Hoffmann et al., "Environmental Applications of Semiconductor Photo catalysis",Chem. Rev. 95, 69-96 (1995) 10. M. Pelaez, "A Review on the Visible Light Active Titanium Dioxide Photocatalysts for Environmental Applications", Applied Catalysis B. 125, 331 -- 349(2012) 11. A. FujishimaandK. Honda,"Electrochemical Photolysis of Water at a Semiconductor Electrode", Nature, 238, 37−38 (1972). 12. M. Kitano et al.,"Recent developments in titanium oxide-based photocatalysts",Appl. Catal. A.325,1,(2007). 13. A. L. Linsebigler et al.,"Photocatalysis on TiO2 Surfaces: Principles, Mechanisms, and Selected Results",Chem. Rev. 95, 735-758(1995) 14. A.Fujishima and X. Zhang, "Titanium dioxide photo catalysis:present situation and future approaches", C. R. Chimie. 9,750 -- 760(2006) 15. V. I. Shapovalov,"Nanopowders and Films of Titanium Oxide for Photocatalysis:AReview",Glass Physics and Chemistry36, 2, 121 -- 157(2010) 16. K Hashimoto et al.,"TiO2Photocatalysis: A Historical Overview and Future Prospects", Japanese Journal of Applied Physics, 44, 12, 8269 -- 8285 (2005) 17. MengNiet al., "A review and recent developments in photocatalytic water-splitting using TiO2 for hydrogen production", Renewable and Sustainable Energy Reviews. 11, 401 -- 425(2007) 18. P. Zeman and S. Takabayashi,"Self-cleaning and antifogging effects of TiO2 films prepared by radio frequency magnetron sputtering", J. Vac. Sci. Technol. A 20, 388-393 (2002) 19. R. Levinson et al.," Solar spectral optical properties of pigments -- Part II: survey of common colorant",Sol. Energy Mater. Sol. Cells 89, 319 -- 349 (2005) 20. A. Fuerte., "Visible light-activated nanosized doped-TiO2 photocatalysts", Chem. Commun., 2718 -- 2719 (2001) 21. A.Zaleska,"Doped-TiO2: A Review",Recent Patents on Engineering. 2, 157-164(2008) 22. F.Zuo et al., "Self-doped Ti3+ enhanced photocatalyst for hydrogen production under visible light", Journal of American Chemical Society 132, 11856 (2010) 23. S. Hoang et al., "Enhancing Visible Light Photo-oxidation of Water with TiO2 Nanowire Arrays via Cotreatment with H2 and NH3: Synergistic Effects between Ti3+ and N", Journal of American Chemical Society 134, 3659−3662(2012) 24. X. Chen et al., "Increasing Solar Absorption for Photocatalysis with Black Hydrogenated Titanium Dioxide Nanocrystals",Science 331, 746-750(2011) 25. G. Wang et al., "Hydrogen-Treated TiO2 Nanowire Arrays for Photoelectrochemical Water Splitting", NanoLett. 11, 3026 -- 3033(2011) 26. X. Chen et al., "Properties of Disorder-Engineered Black Titanium Dioxide Nanoparticles through Hydrogenation", Scientific Reports 3, 1510 (2003) 27. Azza Amin El-Hamshary., "Influence of reactive sputtering process parameters on the structure and properties ofTiO2 thin films", Thesis, RWTH Aachen, 146-157(2011) 28. I. Nakamura et al.," Role of oxygen vacancy in the plasma-treated TiO2photocatalyst with visible light activity for NO removal", J. Mol. Catal. A: Chem. 161, 205(2000) 29. I. Justicia et al,"Designed Self-Doped Titanium Oxide Thin Films for Efficient Visible- Light Photocatalysis", Adv. Mater. 14, 1399 (2002). 30. R. Abe et al., "Efficient hydrogen evolution from aqueous mixture of I_ and acetonitrile using a merocyanine dye-sensitized Pt=TiO2 photocatalyst under visible light irradiation",Chem. Phys. Lett., 362, 441(2002) . 31. M Landmannet al.," The electronic structure and optical response of rutile, anatase and brookite TiO2", J. Phys.: Condens. Matter24 195503(2012) 32. Shang-Di Mo and W. Y. Ching," Electronic and optical properties of three phases of titanium dioxide: Rutile, anatase, and brookite", Phys. Rev., B 51 13023(1995) 33. David O. Scanlon et al., "Band alignment of rutile and anatase TiO2", Nat Mat. 3697(2013) 34. Ju-Young Park et al.," "Structure Related Photocatalytic Properties of TiO2", Bull. Korean Chem. Soc.30, 2(2009) 35. U. Diebold," Structure and properties of TiO2 surfaces:a brief review", Appl. Phys. A 76, 1 -- 7 (2002) 36. D Reyes-Coronado et al.," Phase-pure TiO2 nanoparticles: anatase,brookite and rutile", Nanotechnology 19 145605(2008) 37. C. E. Ekuma and D. Bagayoko," Ab-initio Electronic and Structural Properties of Rutile Titanium Dioxide", arXiv:1011.1315(2010) 38. KunioOkimura et al.," Characteristics of rutile TiO2 films prepared by r.f. magnetron sputtering at a low temperature", Thin Solid Films 281-282 427·430 (1996) 39. G.E. Jellison Jr., L.A. Boatner, J.D. Budai, B.-S.Jeong, D.P. Norton, J. Appl. Phys. 93, 9537(2003) 40. Md. Mosaddeq-ur-Rahman, G. Yu, K.M. Krishna, T. Soga, J. Watanabe, T. Jimbo, M. Umeno, Appl. Opt. 37, 691(1998). 41. M.G. Krishna, K.N. Rao, S. Mohan, "Properties of ion assisted deposited titania films ",J. Appl. Phys. 73 , 434(1993). 42. F. Zhang, Z. Zheng, Y. Chen, D. Liu, X. Liu," Study on the effect of ion beam bombardment during deposition on preferred orientation in rutile-type titanium dioxide films ", J. Appl. Phys. 83, 4101(1998) 43. Z. Wang, et al.," Transport properties in chromium-doped Ti2O3 thin films ", J. Appl. Phys. 93,7870 (2003). 44. S.A. Chamberet al., "Epitaxial growth and properties of ferromagnetic co-doped TiO2 anatase", Appl. Phys. Lett. 79, 3467(2001) 45. Rohan P. Chaukulkar and Sumit Agarwal," Atomic layer deposition of titanium dioxide using titanium tetrachloride and titanium tetraisopropoxide as precursors ",Vac. Sci. Technol. A 31, 031509 (2013) 46. Qi Xie et al.," Atomic layer deposition of TiO2 from tetrakis-dimethyl- amidotitaniumorTiisopropoxide precursors and H2O", JOURNAL OF APPLIED PHYSICS 102, 083521 (2007) 47. G.Rocker and W.Gopel,"Titanium overlayers on TiO2 (110)," Surface Science 181, 530 (1987) 48. G. Tyuliev and S. Angelov,"The nature of excess oxygen in Co3O4+ɛ," Appl. Surf. Sci. 32, 381 (1988) 49. X. Chen, L. Liu, P.Y.yu and S.S.Mao,"Increasing solar absorption for photocatalysis with black hydrogenated titanium dioxide nanocrystals," Science 331, 746 (2011)
1811.07828	1	1811	2018-11-03T00:18:00	Photonic crystal-Microring resonators for tunable delay lines	[ "physics.app-ph", "physics.optics" ]	In this paper, a new design of optical delay line based on coupled ring resonators in photonic crystals is proposed and analyzed by means of numerical simulation in CST Microwave Studio. The performance of the proposed photonic crystal ring resonator (PhCRR) based coupled resonant optical waveguide (CROW) is compared to: point-defect cavity based coupled cavity waveguides (CCW) in photonic crystal. microring resonator (MRR) based CROW. The suggested design addresses compactness issues of the MRR CROW while offering normalized delay values better than CCW and comparable to MRR CROW. The PhCRR CROW also enhances design flexibility as the PhCRRs support multiple modes and thereby supporting multi-channel transfer/delay application on a single device.	physics.app-ph	physics	Photonic crystal-Microring resonators for tunable delay lines Sumanyu Chauhan and Roza Letizia Abstract In this paper, a new design of optical delay line based on coupled ring resonators in photonic crystals is proposed and analyzed by means of numerical simulation in CST Microwave Studio. The performance of the proposed photonic crystal ring resonator (PhCRR) based coupled resonant optical waveguide (CROW) is compared to: point-defect cavity based coupled cavity waveguides (CCW) in photonic crystal. microring resonator (MRR) based CROW. The suggested design addresses compactness issues of the MRR CROW while offering normalized delay values better than CCW and comparable to MRR CROW. The PhCRR CROW also enhances design flexibility as the PhCRRs support multiple modes and thereby supporting multi-channel transfer/delay application on a single device. Keywords Photonic crystals · Microring resonators · Coupled cavity · Optical storage · Delay lines ------------------------------------------- R. Letizia Engineering Department, Lancaster University, Lancaster - LA1 4YR, United Kingdom S. Chauhan Engineering Department (Alumni), Lancaster University, Lancaster - LA1 4YR, United Kingdom 1 Introduction Improving optical buffers is one of the most challenging issues in the realization of optical delay lines. Slow- light optical devices have been intensively studied because of their properties for handling the flow of light. Among the variety of features connected to the concept of slow light, the ability of controlling the optical group velocity and/or the group delay of the light can find applications in the realization of variable delay lines and devices for optical storage and buffering [1]. For control of optical pulse in the slow-light regime coupled optical resonators are typically used. Some examples of optical resonators are: a ring, a disc, a toroid, a sphere, an annular Bragg grating, a defect in a photonic crystal lattice and a Fabry-Perot cavity [2]. Over the years (microring resonator) MRR has emerged as a component for integration into photonic circuits because unlike MEMS it does not have any moving part thus a longer life period. To control the group velocity of light, coupled MRRs have been effectively shown as delay lines [3]. It has been reported that as the diameter of the MRR is reduced below 5µm, propagation and bending losses increase exponentially, negatively affecting performance [4]. This brings about the need to find an alternative technology which provides reduction in size and provides opportunity for improving performance in slow-light regime. It is well known that light trapped inside PBG material like photonic crystal (PhC) can be guided through tight bends because whatever modes light can diffract are removed by the PBG. Optical resonators can be created in photonic crystal by creating point and line defects. Point and line defects in PhC can be coupled to gain control over group velocity of light [5] [6]. PhC based delay lines have emerged as an attractive candidate for such applications, as the dimensions of the stored bit is comparable to the wavelength [7]. Other advantages of integrating PhC in microchips include: reduction in back scattering along the circuit path [8] , unlike electronic chips heat generation and dissipation are no issue in 3D optical microchips [9]. Coupled cavity waveguides on a square lattice with point defects have shown velocity reduction up to 0.4c [10] . Apart from point and line defects another type of defect inspired from the microring resonator had been implemented in photonic crystal. Ring resonator defects in PhC have much higher Q-factor than point/line defects [11] . In this paper, we use coupled circular ring resonators [12] in PhC to form a delay line. In this paper, following the introduction, the formulas used for calculating performance metrics of MRR CROW and PhCRR CROW are provided in Section 2. Section 3 provides the simulation results for a MRR CROW and PhCRR CROW obtained in CST Microwave Studio. Finally, conclusions are given. 2 Analysis Microring resonator (MRR) consists of an optical waveguide looped back on itself, when the number of wavelengths in the circular resonator are in whole multiples of the optical path, the energy in the structure resonates. Microring resonators play an important role in silicon photonics and have been heavily researched into. MRRs can be coupled to form CROWs which can be used as optical delay lines. The theory behind working of MRRs and its various configurations such as MRR CROWs has been explained in detail by Bogaerts, et al. [13] . The maximum group velocity at the center of CROW transmission band can be calculated using the formula provided in [14] Mechanisms proposed in the past for optical waveguiding include: total internal reflection, Bragg waveguiding and waveguiding based on coupling of optical resonators or coupled resonator optical waveguide (CROW) [15] [16] . Some of the possible ways realizations of CROW based waveguides are: evanescent-field coupling between the high-Q modes of individual microdisk cavities in MRR CROW and evanescent-field coupling between the individual resonators (point defect cavities) in [17] [18] embedded in a two-dimensional periodic structure such as 2D photonic crystal [19] [20] . The coupling in case of CROWs is due to the evanescent Bloch waves. In [21], Yariv et el., employed a formalism similar to the tight-binding method in solid-state physics, and obtain the relations for the dispersion and the group velocity of the photonic band of the CROW's and explained that they are solely characterized by coupling factor (k). For development of this formalism for CROWs, Yariv et al. assumed that sufficiently large separation is present between the individual resonators so that the resonators are weakly coupled. Consequently, it is expected that the eigenmode of the electromagnetic field in such a coupled-resonator waveguide will remain essentially the same as the high-Q mode in a single resonator. In this paper, we propose a new design for delay lines in PhC. Instead of coupling point defects in PhC, we create a waveguide by coupling ring-resonators in PhC. Ring-resonators in PhC exhibit much higher-Q [11] than point defects. Therefore, the higher-Q of ring-resonators can be utilized in creating delay-lines by coupling them to form a CROW in PhC, similar to the MRR CROW. The coupling in PhCRR CROW, essentially (just like in case of CCWs) occurs due to coupling of evanescent Bloch waves. The ring-resonators in PhCRR CROWs are sufficiently separated such that they are weakly coupled and the EM-field in the PhCRR CROW remains essentially the same as EM-field in a single resonator. Since, the undelaying assumptions in the case of PhCRR CROW and its behavior (observed from CST simulations) are same as that of CCW [21], we anticipate that the same formalism can be applied to PhCRR CROW. Therefore, formulas used for circular PhCRR CROW are given as [10] [21] [22] : 𝑣𝑔 = ±𝑅Ω √𝜅2 − ( − 1) 2 𝜔 Ω (1) Where: 𝑣𝑔 is the group velocity of the EM-mode. 𝑅 is the cavity spacing. 𝜅 is the coupling coefficient. 𝜔 is the operating frequency. Ω is the center frequency. 𝜅 = △𝜔/2 At center frequency (Ω), 𝑣𝑔 becomes: 𝑣𝑔 = 𝜅𝑅 Ω 3 Simulation results 3.1 MRR CROW The MRR in an all-pass filter configuration is shown in figure 1 was simulated in CST Microwave Studio. The design parameters were similar to that of the previous work reported in [23], to validate the results obtained in terms of the resonant peaks and coupling coefficients as given in table 1. Figure 1, shows schematic of the MRR simulated and placement of field probes within the simulated geometry. Input probe 1 and output probe 1 detect the input and output fields respectively, while 1, 2 and 3 detect the resonant fields in the ring resonator. The 2D structure was simulated in the x-y plane, where the perpendicular direction along the z-axis is considered infinite. The frequency range simulated was 180- 220 THz, the first resonant mode within this band for the given parameters has a modal number of m = 25. This means that the resonator can accommodate 24 other resonant modes on the lower side of the frequency range and many other modes on the higher side of the frequency range (depending on the frequency tolerance of the material of MRR). In figure 2, the graph shows the FFT of E-field, as computed in the ring probe 1. It shows resonant frequencies and corresponding modal number within simulated frequency range. Fig. 1 Schematic of MRR in APF configuration with placement of field probes. Fig 2. Resonant frequencies for the MRR within the simulated frequency range. Figure 3 shows four conventional ring-resonators coupled to each other, the placement of field-probes in the ports and in the ring coupled to the bus waveguide. The gap between two adjacent rings is same as the gap between the ring and waveguide. The structure was simulated with periodic boundary conditions along the z-axis and open boundary condition in the x-y plane. The structure was discretized using hexahedron mesh and simulated within a frequency range of 180-220 THz. The maximum duration of the simulation was set to 200 pulse widths (200 times the width of excitation signal). Fig. 3 Schematic of MRR CROW with the placement of field probes. Figure 4 shows the frequencies dropped in the forward port of the MRR based CROW. It was observed that for the odd modal numbers (m=25, 27 and 29), higher amplitude of the field is transmitted, this is due to the unidirectional circulating nature of the resonant modes with odd modal number. 4 for Fig. Transmission characteristics the different modal numbers in the forward port of MRR CROW. Figure 5 (a) shows the transmission characteristics for odd modal number (m=29) in the forward port of MRR based CROW. There are four resonant peaks for the transmission bandwidth. The four peaks are marked as 1, 2, 4 and 5. The four peaks occur as a consequence of four coupled ring-resonators where each ring- resonator adds negative or positive frequency shift around the central frequency of the isolated ring resonator. Figure 5 (b) shows the transmission characteristic for even modal number (m=30) in the forward port of MRR based CROW. There are four resonant peaks for the transmission bandwidth. The four peaks are marked as 1, 2, 4 and 5. The four peaks occur as a consequence of four coupled ring-resonators where each ring-resonator adds a shift in frequency around the central frequency of the transmission bandwidth. The two blue rings for peak 1 represent the two degenerate modes for the first resonant peak. The resonance splitting occurs as a consequence of counter-directional coupling. Same holds true for the resonance- splitting in peak 2, 4 and 5. Fig 5 (a) Transmission band for m=29 Fig. 5 (b) transmission band for m=30 The normalized group velocity of a resonant mode in MRR CROW are shown in table 1. These were determined using equation from [14]. Table 1 Modal number (m), resonant frequency (Fres) and wavelength (λres), effective refractive index (neff), coupling coefficient (k) and normalized group velocity values for the simulated MRR CROW. m 25 26 27 28 29 30 Fres (THz) 184.64 190.56 196.52 202.51 208.47 214.38 λres (nm) 1624.783362 1574.307305 1526.562182 1481.408326 1439.055979 1399.384271 neff 2.585923 2.605811 2.623967 2.640663 2.656781 2.672627 k 0.06 0.0471 0.0389 0.0311 0.025 0.0206 Normalized Group velocity 0.01534 0.01195 0.00980 0.00778 0.00622 0.00509 3.2 PhC ring resonator crow Figure 6(a) shows the schematic of the PhCRR and figure 6(b) shows enlarged view of a portion of the PhCCR with the original rod position, marked by the blue circles, in the square lattice. Figure 6 (a) Figure 6 (b) To validate the simulation results, initially the PhCRR was created with silicon rods placed in air in a square lattice with dimension (r= 0.175a; a= 540; nrod = 3.47) was simulated and the resonances were verified with the results published Calo et al [12] . To improve the Q factor of the PhCRR the fill -- factor was changed to 0.185a and refractive index to nrod = 3.59 (gallium phosphide). The Preliminary analysis of the resonant modes was done by simulating a square lattice with an array of 15x15 dielectric rods placed in air. The cavity was excited by a plane wave using time- domain solver. The FFT of resonances detected is shown in the figure 7. Fig. 7: FFT of resonances detected. Figure 8 (a) shows schematic PhCRR CROW created by coupling circular ring resonator cavities on a PhC lattice (r= 0.185a ; a= 540; nrod = 3.59), for which resonant peaks are shown in figure 7. The cavity spacing (R) or the distance between the centers of the two cavities was taken as 3.240 μm. The PhCRR based CROW in figure 8 (a) was simulated in 2D environment with periodic boundary conditions along the z-axis and open boundary condition in the x-y plane. The structure was discretized using tetrahedral mesh and simulated using a frequency domain solver within a frequency range of 160-250 THz. Fig. 8 (a) PhCRR CROW with cavity spacing as 3.240 μm Figure 8(b) shows the variation of scattering parameters S (2, 1) for circular PhCRR CROW. The graph shows four transmission bands, since a single PhCRR (with lattice parameters as : r= 0.185a ; a= 540; nrod = 3.59) exhibits five resonant modes within 160-250 THz (as shown in figure 8), therefore five transmission bands are expected such that each transmission band in the PhCRR CROW has four phase shifted peaks. However, it can be noticed that there are only four transmission bands such that the first and the fourth transmissions bands have more than four phase shifted peaks each. This indicates that there is some interacting of hybrid modes in the first and fourth transmission bands. In figure 8 (b) , the second and third transmission bands have four phase shifted peaks each which means there is no mixing of modes, therefore these transmission bands will be used to closely analyze the slow-light behavior for the given structure. Fig 8 (b) Variation of S (2, 1) within 160-250 THz in a PhCRR CROW with cavity spacing as 3.240 μm. Figure 8 (c) shows the variation of scattering parameters S (2, 1) for circular PhCRR CROW within frequency range of 190-200 THz. It should be noted that the central frequency of this transmission band (194.73 THZ) is also the resonant frequency of an isolated cavity (hexapole) in the PhCRR but due to change in the structure around the cavity, the center frequency hexapole in PhCRR differs slightly from the resonant frequency of isolated PhCRR. When many such cavities are coupled to form a PhCRR CROW then a frequency shift is added on either side of the central frequency due to each cavity [5]. In this case, four coupled cavities were simulated and as such four peaks can be seen in figure 8 (c). The 3db bandwidth of this transmission band was calculated as 4.92 THz in table 2 (given at the end). 8 Fig (c) Variation of S (2, 1) within 190- 200 THz range in a PhCRR CROW with cavity spacing as 3.240 μm. Transmission band centered at 194.73 THz. Similarly, figure 8 (d) shows the variation of scattering parameters S (2, 1) for circular PhCRR CROW within frequency range of 215-225 THz. It should be noted that the central frequency of this transmission band (220.48 THZ) in PhCRR CROW is also the also very close to the resonant frequency of an octupole in an isolated PhCRR cavity. The 3db transmission bandwidth for this band is 1.60 THZ as shown in table 2. Lower transmission bandwidth (for octupole as compared to hexapole). This property is again reflected in the coupling coefficient and normalized group velocity values shown in table 2. That is, octupole has lower coupling coefficient and lower normalized delay values compared to the hexapole. 8 (d) Fig Variation of S (2,1) within 215-225 THz in a PhCRR CROW with cavity spacing as 3.240 μm. Transmission band centered 220.48 THz. at Figure 8 (e) and 8 (f) show mode profiles Hexapole and Octupole in the PhCRR CROW with cavity spacing of 3.240 μm. Fig 8(e) Mode profile for heaxapole (192.28 THz) in PhCRR CROW with cavity spacing of 3.240 μm. Fig 8(f) Mode profile for octupole (219.71 THz) in PhCRR CROW with cavity spacing of 3.240 μm. To demonstrate to tunability of PhCRR CROW, the cavity spacing (R) was changed. Figure 9 (a) shows schematic of PhCRR CROW created by coupling circular ring resonator cavities on a PhC lattice (with lattice parameters as: r= 0.185a; a= 540; nrod = 3.59), for which resonant peaks are shown in figure 7. However, in this PhCRR CROW, the cavity spacing (R) or the distance between the centers of the two cavities was increased to as 4.320 μm. The PhCRR based CROW in figure 9 (a) was simulated in 2D environment with periodic boundary conditions along the z-axis and open boundary condition in the x-y plane. The structure was discretized using tetrahedral mesh and simulated using a frequency domain solver within a frequency range of 160-250 THz. Fig 9 (a) PhCRR CROW with cavity spacing as 4.320 μm Figure 9 (b) shows the variation of scattering parameters S (2, 1) for circular PhCRR CROW within frequency range of 193-196 THz. It should be noted that the central frequency of this transmission band (194.32 THZ) is also the resonant frequency of an isolated cavity (hexapole) in the PhCRR but due to the change in the structure around the cavity, the resonant frequency of hexapole in isolated PhCRR differs slightly from the center frequency of the transmission band for hexapole in PhCRR CROW. The 3db bandwidth of this transmission band was calculated in table 2 as 0.44 THz; which is much lower than the hexapole in the PhcRR CROW will smaller cavity spacing (R=3.240). This happens because the field confinement improves in a PhCRR CROW with bigger separation between cavities. By contrasting schematics in figure 8 (a) and 9 (a) it can be observed that there are two extra rows of pillars separating the cavities in figure 9 (a).The extra row of pillars increase the confinement of the individual cavities. 9 in Fig. (b) Variation of S (2, 1) within 193-196 THz range a PhCRR CROW cavity with spacing as 4.320 μm. Transmission band centered at 194.32 THz Similarly, figure 9 (c) shows the variation of scattering parameters S (2, 1) for circular PhCRR CROW within frequency range of 216-224 THz. It should be noted that the central frequency of this transmission band (219.85 THZ) is the resonant frequency of an isolated cavity (Octupole) in the PhCRR but due to slight variation in the structure around the cavity the center frequency varies slightly. The 3db bandwidth of this transmission band was calculated in table 2 as 0.48 THz which is again much lower in comparison to the bandwidth of the same mode (octupole) in PhCRR CROW with smaller cavity spacing (R=3.240 μm). Fig 9 (c) Variation of S (2,1) within 216-224 THz in a CROW PhCRR with cavity spacing as 4.320 μm. Transmission band centered at 219.85 THz Figure 9 (d) and 9 (e) show mode profiles of hexapole and octupole in the PhCRR CROW with cavity spacing of 4.320 μm. Fig 9 (d) Mode profile for heaxapole (194.32 THz) in PhCRR CROW with cavity spacing of 4.320 μm Fig 9 (e) Mode profile for heaxapole (219.85 THz) in PhCRR CROW with cavity spacing of 4.320 μm Table 2: The normalized group velocity values for PhCRR at the center of the transmission band (calculated using equation 1). 3 Conclusion The MRR CROW was simulated in CST microwave studio. Resonance splitting was observed for resonant modes with even modal numbers. Taking inspiration from this design of MRR CROW a new design for optical delay line based on coupled ring-resonator cavities in photonic crystals is proposed. A slow-light structure based on four coupled PhCRR cavities was simulated. Multiple resonant modes were observed which demonstrated capability for multiple-channel transfer. Each of the modes had four phase shifted peaks due to delay in each coupled-cavity. To make comparison with MRR CROW delay for heaxapole and octupole was analyzed. For cavity spacing of 3.240 μm: hexapole was centered at 194.73 THz with a normalized group velocity of 0.0266 while octupole was centered at 220.48 THz with a group velocity of 0.0087. When cavity spacing was changed to 4.320 μm: center frequency of hexapole shifted to 194.32 THz and the normalized group velocity changed to 0.0031 while center frequency for octupole shifted to 219.85 THz with a group velocity of 0.0034 . This is comparable to normalized group velocity values 0.0098 and 0.005 for MRR CROW at 196 and 214 THz respectively. Table 1, 2 and 3 give the coupling coefficients and normalized group velocities for MRR, PhCRR & PhC CROW respectively. MRR CORW provides better reduction in group velocity for comparable coupling coefficients in PhC CROW. While PhCRR gives results similar to MRR CROW but on a smaller length scale therefore highlighting it compactness. By the changing the cavity spacing the tunabilty of the performance parameters for hexapole and octupole was demonstrated. Additionally, it offers to selectively design performance in the slow-light regime by tuning design parameters such as cavity-size and fill-factor. It was demonstrated that ring-resonator type delay- lines can be executed in PhCs with added advantages over CCWs [10] like multi-channel transfer. The tunabilty of PhCRR based slow-light structure was demonstrated by altering the cavity spacing. CROWModeLattice constant a, (μm)Cavity spacing R (μm)Peak TypePeak Frequency (ω) THzf1 (THz)f2 (THz)Δf THzCenter frequency (Ω ) THzCoupling coefficient (k)vg(Ω) m/sNormalised vg(Ω)1Hexapole0.5403.2401192.28192.27197.1934.92194.730.012647.98E+060.02661Hexapole0.5403.2402193.91192.27197.1934.92194.730.012647.98E+060.02661Hexapole0.5403.2403195.78192.27197.1934.92194.730.012647.98E+060.02661Hexapole0.5403.2404197.18192.27197.1934.92194.730.012647.98E+060.02661Octupole0.5403.2401219.7219.68221.2821.60220.480.003632.60E+060.00871Octupole0.5403.2402220.03219.68221.2821.60220.480.003632.60E+060.00871Octupole0.5403.2403220.66219.68221.2821.60220.480.003632.60E+060.00871Octupole0.5403.2404221.28219.68221.2821.60220.480.003632.60E+060.00872Hexapole0.5404.3201194.083194.08194.5590.44194.320.001129.42E+050.00312Hexapole0.5404.3202194.224194.08194.5590.44194.320.001129.42E+050.00312Hexapole0.5404.3203194.407194.08194.5590.44194.320.001129.42E+050.00312Hexapole0.5404.3204194.557194.08194.5590.44194.320.001129.42E+050.00312Octupole0.5404.3201219.64219.63220.0660.48219.850.001091.03E+060.00342Octupole0.5404.3202219.76219.63220.0660.48219.850.001091.03E+060.00342Octupole0.5404.3203219.93219.63220.0660.48219.850.001091.03E+060.00342Octupole0.5404.3204220.06219.63220.0660.48219.850.001091.03E+060.0034Table 3: Representative coupling coefficients and group velocities (normalised) for CCWs [15] Lattice constant a, (μm) Cavity spacing R (μm) 5.88E-01 8.70E-01 5.86E-01 2a 3a 4a Coupling coefficient (k) Normalised vg(Ω) 4.66E-02 1.49E-02 4.70E-03 0.221 0.1 0.045 CCW mode TE TE TE References [1] R. Tucker, "Slow-light optical buffers: capabilities and fundamental limitations," Lightwave Technology, Journal of, vol. 23, no. 12, pp. 4046 - 4066, 2006. [2] F. Morichetti, C. Ferrari, A. Canciamilla and A. Melloni, "The first decade of coupled resonator optical waveguides: bringing slow light to applications," Laser & Photonics Reviews, pp. 74 - 96, 11 January 2012. [3] Q. Xu, P. Dong and M. Lipson, "Breaking the delay-bandwidth limit in a photonic structure," Nature Physics, vol. 3, pp. 406-410, 22 April 2007. [4] S. Robinson and R. Nakkeeran, "Photonic crystal ring resonator-based add drop filters: a review," Optical Engineering, pp. 060901-060901, 15 May 2013. [5] A. Reynolds, U. Peschel, F. Lederer, P. Roberts, T. Krauss and P. de Maagt, "Coupled Defects in Photonic Crystals," IEEE Transactions on microwave theory and techniques, vol. 49, no. 10, pp. 1860-1867, 2001. [6] E. Ozbay, M. Bayindir, I. Bulu and E. Cubukcu, "Investigation of localized coupled-cavity modes in two- dimensional photonic bandgap structures," Quantum Electronics, IEEE Journal of, vol. 38, no. 7, pp. 837-843, 2002. [7] R. Tucker, P.-C. Ku and C. Chang-Hasnain, "Slow-light optical buffers: capabilities and fundamental limitations," Lightwave Technology, Journal of, Dec. 2005. [8] J. Joannnopolos, P. Villeneuve and S. Fan, "Photonic Crystals: Putting a new twiston light," Nature, vol. 386, p. 143, 1997. [9] A. Chutinan and S. John, "3-Dimensional Integrated Optics with Localised light in a Photonic bandgap," Optical Express, vol. 14, no. March, pp. 143-149, 2006. [10] M. S. Moreolo, V. Morra and G. Cincotti, "Design of photonic crystal delay lines based on enhanced coupled- cavity waveguides," Journal of Optics A: Pure and Applied Optics, vol. 10, no. 6, p. 064002, 2008. [11] V. Kumar, T. Srinivas and S. A., "Investigation of ring resonators in photonic crystal circuits," Photonics and Nanostructures - Fundamentals and Applications, vol. 2, no. 3, pp. 199 - 206, 2008. [12] G. Calo and V. Petruzzelli, "Wavelength Routers for Optical Networks-on-chip using optimised photonic crystal ring resonators," IEEE Photonics Journal, vol. 5, no. 3, pp. 7901011-7901011, 3013. [13] W. Bogaerts, P. De Heyn, T. Van Vaerenbergh, K. De Vos, S. Kumar Selvaraja, T. Claes, P. Dumon, P. Bienstman, D. Van Thourhout and R. Baets, "Silicon microring resonators," Laser & Photon. Rev, vol. 6, no. 1, pp. 1863- 8899, 2012. [14] J. Poon, J. X. Y. Scheur and A. Yariv, "Designing coupled-resoonator optical delay lines," The Journal of the Optical Society of America, vol. 21, no. 9, pp. 1665-1673, 2004. [15] A. Yariv and P. Yeh, Optical Waves in Crystals, New York: Wiley, 1984. [16] P. Yeh and A. Yariv, "Bragg reflection waveguides," Optical Communication, vol. 19, p. 427 -- 430, 1976. [17] R. Meade, A. Devenyi, J. D. Joannopoulos, O. L. Alerhand, D. A. Smith and K. Kash, "Novel applications of photonic band gap materials: Low‐loss bends and high Q cavities," Journal of Applied Physics, vol. 75, p. 4753, 1994. [18] S. Fan, P. Villeneuve, J. D. Joannopoulos and H. Haus, "Channel drop filters in photonic crystals," Optical Express, vol. 54, no. 7836, 1998. [19] S. John, "Strong localization of photons in certain disordered dielectric superlattices," Physics Review Letters, vol. 58, no. 23, 1987. [20] E. Yablonovitch, "Inhibited Spontaneous Emission in Solid-State Physics and Electronics," Physics Review Letters, vol. 58, no. 20, pp. 2059-2062, 1987. [21] A. Yariv, Y. Xu and R. Lee, "Coupled-resonator optical waveguide: a proposal and analysis," Optical Letters, vol. 24, pp. 711-713, 1999. [22] K. Hosomi and K. T., "A dispersion compensator using coupled defects in a photonic crystal," Quantum Electronics , vol. 38, pp. 825-829, 2002. [23] R. Letizia, N. Abujnah, M. Hameed, S. Obayya and M. Abdelrazzak, "Efficient design design of optical delay lines based on slotted-ring resonators," IET Optoelctronics, vol. 7, no. 3, pp. 63 - 70, June 2013. [24] I. Neokosmidis, T. Kamalakis and T. Sphicopolus, "Optical delay lines based on soliton propogation in photonic crystal coupled resonator optical waveguides," IEEE Quantum Electron , vol. 43, no. 7, pp. 560-567, 2007. [25] X. Fengnian, S. Lidija and Y. Vlasov, "Ultracompact optical buffers on a silicon chip," Nature Photonics , pp. 65- 71, 2007. [26] F. Liu, Q. Li, Z. Zhang, M. Qin and Y. Su, "Optical tunable delay line in silicon microring resonator based on thermal nonlinear effect," Quantum Electronics , vol. 14, no. 3, pp. 706-712, 2008. [27] J. Khurgin, "Optical buffers based on slow-light in electronically induced transparent media and coupled resonator structures: comprative analysis," Optical Society of America, vol. 22, no. 5, pp. 1062-1074, 2005. [28] L. BE., J. Laine and C. ST., "Surface-roughness induced contradirectional coupling in the ring and disk resonators," Optical Letters, vol. 22, pp. 4-6, 1997.
1804.08696	1	1804	2018-03-29T19:30:27	Percolative Mechanism of Aging in Zirconia-Containing Ceramics for Medical Applications	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Recently, several episodes of fracture of zirconia ceramic femoral heads of total hip prostheses have alarmed the medical and scientific community regarding aging problems in zirconia prostheses. Such fractures cause immediate local tissue reactions, which require urgent medical intervention to prevent further complications. As a result, it has been promoted that yttria-stabilized zirconia (Y-TZP) hip prostheses be substituted by alumina and alumina/Y-TZP ceramics. In the present investigation, we have found an upper limit of Y-TZP concentration in alumina/Y-TZP composites (16 vol.%) to avoid future aging problems. This limit coincides with the percolation threshold measured by infrared (IR) reflectance in a series of alumina/Y-TZP composites.	physics.app-ph	physics	Percolative Mechanism of Aging in Zirconia‐ Containing Ceramics for Medical Applications C. Pecharromán1, J.F. Bartolomé1, J. Requena1, J.S. Moya1, S. Deville2, J. Chevalier2, G. Fantozzi2, R. Torrecillas3 1 Instituto de Ciencia de Materiales de Madrid (ICMM), CSIC, Cantoblanco 28049, 2 Department of Research into the Metallurgy and Physical Properties of Materials, Madrid, Spain Associate Research Unit 5510, National Institute of Applied Sciences 69621 Villeurbanne, France 3 Instituto Nacional del Carbón CSIC, C/Francisco Pintado Fe, 26 La Corredoira, 33011 Oviedo, Spain Abstract Recently, several episodes of fracture of zirconia ceramic femoral heads of total hip prostheses have alarmed the medical and scientific community regarding aging problems in zirconia prostheses. Such fractures cause immediate local tissue reactions, which require urgent medical intervention to prevent further complications. As a result, it has been promoted that yttria-stabilized zirconia (Y-TZP) hip prostheses be substituted by alumina and alumina/Y-TZP ceramics. In the present investigation, we have found an upper limit of Y-TZP concentration in alumina/Y-TZP composites (16 vol.-%) to avoid future aging problems. This limit coincides with the percolation threshold measured by infrared (IR) reflectance in a series of alumina/Y-TZP composites. 1 Until now, zirconia hip prostheses were a very popular alternative to alumina implants because of their higher mechanical resistance. Y-TZP indeed exhibits the largest value of crack resistance of all the monolithic ceramics. In particular, the use of zirconia has opened the way towards new implant designs that were not possible with alumina. It has been estimated that about 500 000 patients have been implanted with zirconia femoral head prostheses since 1985, 200 000 during the past four years. Nevertheless, alarming problems related to aging have been reported recently. In particular, resistance to steam sterilization and the hydrothermal stability of yttria containing zirconia in the body have been questioned. Aging occurs by a slow tetragonal-to- monoclinic phase transformation of grains on any surface in contact with water or body fluids. [1] This transformation leads to surface roughening, [2] grain pullout, and microcracking. [3,4] Steam sterilization has been associated with surface roughening of zirconia ceramic heads. [5] As a consequence of this roughening, increase wear of the hip components can cause premature failure and requires early revision. The recognized failure rate of these products was 1 in 10 000, but nowadays, current reports indicate a failure rate as high as 8.8 % in some specific batches. [6] The U.S. Food and Drug Administration (FDA) cautions today against steam sterilization of zirconia implants. [7] More importantly, during the last year, the FDA and the Australian Therapeutic Goods Administration (TGA) announced that firms making orthopaedic implants were recalling series of Y-TZP hip prostheses due to an instance of fracture of zirconia ceramic femoral heads. [6,8] This recall follows the action by the French Agency for the Medical Safety of Health Products (AFSSAPS) and the United Kingdom Medical Devices Agency suspending the sales of Y-TZP ceramic heads. According to the manufacturer of the zirconia ceramic, the failure's origin was an accelerated tetragonal- to-monoclinic phase transformation of zirconia in particular batches. [9] Because of these events, there is a trend today to develop advanced materials such as alumina-zirconia composites, to combine the positive effects of alumina and zirconia while avoiding the negative influence of each. In the recent literature concerning alumina-zirconia composites for biomedical applications, different compositions have been tested, from the zirconia-rich to the alumina-rich. [10-12] One route was already developed by companies (CeramTec BIOLOX delta) with the addition of 3Y-TZP (with an amount of approx. 17 vol.-%) to alumina to reach values as high as KIC=8.5 MPa.m1/2 and sf = 1150 MPa. [13,14] However, so far, little attention has been paid to the stability of these biomedical-grade zirconia-containing materials. In this regard, it is the aim of this work to study the aging behavior of alumina/3Y-TZP composites to predict future failure of the alumina/3Y-TZP prosthesis implanted in patients by using the concept of percolation threshold. 2 It is now established that the tetragonal-to-monoclinic phase transformation in zirconia occurs by nucleation and growth. [2] Transformation appears first on isolated grains on the surface, then propagates to the neighboring grains as a result of various stresses and the accumulation of microcracks. The transformation penetrates step by step inside the bulk of the material, providing there is contact between zirconia grains. In the case of alumina/3Y-TZP composites, this phenomenon should require a continuous path of zirconia grains. The geometrical percolation concept, i.e., the infinite cluster formed at the percolation threshold, could therefore be applied to explain the kinetics of the aging process through the bulk of the materials, and to predict the amount of zirconia that can be added to alumina without causing aging problems. It is well-known that experimental determination of percolation thresholds in heterogeneous composites can only be achieved when there is a large contrast between the physical properties of the different phases: [15±18] i.e., permittivity (e), electrical conductivity (s), thermal conductivity (k), etc. In the case of alumina/3Y-TZP composites, the magnitudes of these variables are roughly similar in most of the spectral regions. In addition, the strong anisotropic character of alumina (the IR zirconia dielectric tensor has been taken to be isotropic, in accordance with the literature [19]) requires that one take into consideration the contrast function between two different tensor magnitudes. As a result, it is extremely difficult to experimentally determine the critical concentration. There is one exception: it has be found that around the infrared region of the longitudinal phonons of alumina (from 900 cm-1 to 1 000 cm-1), the complex dielectric tensor of both alumina and zirconia [19,20] are nearly isotropic and display a significant contrast (Fig. 1), which can be as high as 𝜀 ≤ 100𝜀 . In order to simplify the analysis of the IR-effective properties in the considered region of the spectrum, we will assume an isotropic model for the IR dielectric constant of Al2O3: 𝜀𝜔 =𝜀∞ 2−2 − 2−2 − (eq. 1) This approximation can be done only in this spectral region, owing to the fact that the highest transversal and longitudinal phonons of alumina for both A2u (589 cm-1 and 871 cm-1) and Eu (635 cm-1 and 900 cm-1) modes are very close. Then, the transversal and longitudinal frequencies, wT and wL, respectively, the damping g, and the optical dielectric constant 𝜀" can be approximately taken as the weighted average of both orientations (wT=620 cm-1, wL=890 cm-1, g=15.1 cm-1, and 𝜀"=3.2). Although the permittivity cannot be measured directly with conventional IR spectrometers, the specular reflectance is a function of e and must reveal any large changes in this function. The near-normal specular reflectance is given by: 3 (eq. 2) 𝑅= −1+12 𝜀(𝜔)−1≅0 The combination of Equation 1 for the dielectric constant with Equation 2 means that the reflectance curve has a broad band with values for R very close to 1, which extends from wT to wL and quickly descends for frequencies larger than wL. In fact, R reaches a minimum at the approximate frequency which satisfies: (eq. 3) This frequency is approximately wm=1 000 cm-1 for pure polycrystalline alumina. Therefore, the range of reflectance minimum of alumina/3Y-TZP composites studied extends from the longitudinal modes of alumina (wL=900 cm-1) to wm=1 000 cm-1 (Figs. 1, 2a). The reflectance profile of those composites depends on the shape of the effective dielectric constant below and above the percolation threshold at the frequencies wL and wm, respectively. For 3Y-TZP concentrations below the percolation threshold (f<fc), the effect of the addition of zirconia to the composite on the effective dielectric constant for the whole spectral range is a small increase. This implies that the condition of Equation 3 is satisfied for lower frequencies (Fig. 1). It results in a reflectance profile similar to that of pure alumina with the reflectance minimum shifted to slightly lower frequencies. However, above the percolation threshold, the effective dielectric constant will rise sharply at w = wL, where the dielectric contrast is a maximum, and moderately at w =wm, where the dielectric contrast is less than 4. This occurs in such a way that the effective dielectric constant for the whole spectral region from wL to wm will take values similar to that of 𝜀" of zirconia (𝜀" = 4.2). Such region of small dispersion of e(w) will also induce a nearly flat area to occur in the reflectance spectrum. This behavior is clearly detected experimentally. As it can be seen in Figure 2a, the minimum of the reflectance curve gradually shifts to lower frequencies, for concentrations lower than f = 0.14. Above this point, the shape of the minimum located at 1 000 cm-1 drastically changes. In Figure 2b, this effect is represented as a plot of the reflectance minimum vs f, then it is clear that for f > 0.14 the reflectance increases dramatically. This fact suggests that the percolation threshold of alumina/3Y-TZP composites must be located around 𝑓≈01.6 as predicted using standard models for three-dimensional percolation. [15,16] Figure 3 shows the variation in the fraction of monoclinic phase vs time of steam sterilization for some selected compositions. Monoclinic content was measured by an X-ray diffraction (XRD) technique (Cu-Ka radiation). When values of mass absorption 4 coefficients and densities of the studied alumina/ZrO2 composites are taken into consideration, [21] this technique gives information on penetration depths (defined as an attenuation of 1/e = 0.37 for a beam normal to the surface) ranging from 50 µm to 15 µm, depending on the fraction of zirconia. The penetration depth decreases linearly with an increasing vol.-% of ZrO2. Therefore, when there are more than 10 consecutive layers of grains, we can consider that these data supply an accurate representation of processes happening in three dimensions. For all the materials studied, a first increase of up to 10 % of monoclinic phase was observed after two hours. This first increase was easily related to the presence of some (scarce) 3Y-TZP aggregates in the materials (Fig. 4a). More importantly is that for specimens with zirconia content of less than or equal to 14 vol.-%, after this initial slight increase, no more evolution was observed, even after 100 hours in the autoclave. On the other hand, when f>fc, the monoclinic phase fraction continues to grow without reaching saturation. This clearly means that transformation occurs by a progressive invasion of the material (Fig. 4b). Figure 5 represents the amount of transformation observed using XRD analysis after 40 hours of aging in steam, as compared with the initial 3Y-TZP content. This correlates especially well with the percolation theory. Above the critical amount of 16 vol.-% 3Y- TZP, geometrical percolation allows a continuous path between zirconia particles, so that transformation can proceed. One might question the correlation of these accelerated tests with the reality of the in- vivo situation. The activation energy of zirconia aging [2] is of the order of 100 kJ mol- 1. These accelerated tests in steam give a rough estimation of the monoclinic fraction obtained after a hundred years at 37°C. However, it is also very often stated that during wear, the temperature of the surface of the prosthesis can reach as high as 50°C. [22] This temperature would give a more pessimistic estimation of only several years. In any case, taking into consideration the uncertainties in the real temperature, in the estimation on the activation energy, and the recent events of unexpected accelerated aging of zirconia, no aging at all should be acceptable. In summary, it has been shown that comparing specular IR reflectance measurements with aging experiments, the concept of the percolation threshold is relevant when talking about degradation resulting from aging. As far as biomedical applications are concerned, zirconia-toughened alumina ceramics would be very appropriate materials, provided that the zirconia content is kept below the percolation threshold. This study has established an upper limit of 16 vol.-% 3Y-TZP inside an alumina matrix. Therefore, the authors caution ceramics manufacturers and surgeons against fabrication of zirconia-alumina composites, where the amount of zirconia is above 16 vol.-% (22 wt.-%). 5 Experimental Plates with 80x80x3 mm3 dimensions ranging from 7 vol.-% 3Y-TZP in an alumina matrix to pure 3Y-TZP were obtained by slip casting in a plaster of Paris mold and sintered to >99 % density at 1600C for 2 h in the case of alumina-zirconia composites, and at 1450°C for 2 h in the case of pure 3Y-TZP. They were processed from high- purity biomedical-grade powders (Tosoh TZ3Y and Condea HPA 0.5). Samples were polished with diamond paste in order to reach the surface topography recommended for hip prostheses. The initial monoclinic content was in all cases close to zero. Aging experiments were carried out by leaving samples in a steam autoclave at a temperature of 140°C. The diamond-polished side of each specimen was examined by X-ray diffraction (XRD) before and after aging. XRD data were obtained with a diffractometer using Ni-filtered Cu-Ka radiation. The tetragonal/monoclinic zirconia ratio was determined using the integrated intensity (measuring the area under the diffractometer peaks) of the tetragonal (101) and two monoclinic (111) and (1-11) peaks as described by Garvie et Nicholson [23] and then revised by Toroya et al. [24,25]. For the purpose of comparison, the integrated intensities obtained were individually normalized to the (101) tetragonal integrated intensity for each composition. Diffractometer scans were obtained from 2q = 27-33° at a scan speed of 0.2°min-1 and a step size of 0.02°. Near-normal specular reflectance spectra were taken on diamond- polished surfaces of the different alumina/3Y-TZP composites with a near-normal reflectance attachment (12°) in a FTIR Bruker 66V spectrometer. References [1] M. Yoshimura, T. Noma, K. Kawabata, S. Somiya, J. Mater. Sci. Lett. 1987, 6, 465. [2] J. Chevalier, B. Cal(cid:0)s, J. M. Drouin, J. Am. Ceram. Soc. 1999, 82, 2150. [3] E. Lilley, in Ceramics Transactions (Eds: R. E. Tressler, M. McNallen), Vol. 10, American Ceramic Society, Westerville, OH 1990, p. 387. [4] W. Z. Zhu, X. B. Zhang, Scr. Mater. 1999, 40, 1229. [5] Safety Notice MDA SN 9617, Zirconia Ceramic Heads for Modular Total Hip Femoral Components: Advice to Users on Re-Sterilization, UK Medical Devices Agency, Adverse Incidents Centre. http://www.medical-devices.gov.uk [6] http://www.health.gov.au/tga/docs/html/hazard.htm [7] http://www.fda.gov/cdrh/steamst.html [8] http://www.fda.gov/cdrh/recalls/zirconiahip.html [9] http://www.prozyr.com/PAGES_UK/Biomedical/Committee.htm [10] S. Affatato, M. Testoni, G. L. Cacciari, A. Toni, Biomaterials 1999, 20, 1925. [11] S. Affatato, M. Goldoni, M. Testoni, A. Toni, Biomaterials 2001, 22,717. 6 [12] J. Chevalier, A. De Aza, M. Schehl, R. Torrecillas, G. Fantozzi, Adv. Mater. 2000, 12, 1619. [13] G. Willmann, in Bioceramics in Hip Joint Replacement Proc. 5th Int. CeramTec Symposium (Eds: G. Willmann, K. Zweymüller), Georg Thieme Verlag, New York 2000, p. 127. [14] R. Rack, H.-G. Pfaff, in Bioceramics in Hip Joint Replacement Proc. 5th Int. CeramTec Symposium (Eds: G. Willmann, K. Zweymüller), Georg Thieme Verlag, New York 2000, p. 141. [15] S. Kirkpatrick, Rev. Mod. Phys. 1973, 45, 574. [16] D. Stauffer, Phys. Rep. 1979, 54,3. [17] A. L. Efros, B. I. Shklovskii, Phys. Status Solidi B 1976, 76, 475. [18] C. Pecharroman, J. S. Moya, Adv. Mater. 2000, 12, 294. [19] C. Pecharroman, M Ocana, J. C. Serna, J. Appl. Phys. 1996, 80, 3479. [20] A. S. Barker, Phys. Rev. 1963, 132, 1474. [21] International Tables Cristallography), Vol. C, D. Reidel Company, Dordrecht, The Netherlands 1983. [22] Z. Lu, H. McKellop, in Proc. of the Institution of Mechanical Engineers, Part H, J. Eng. Med. 1997, 211, 101. [23] R. C. Garvie, P. S. Nicholson, J. Am. Ceram. Soc. 1972, 55, 303. [24] H. Toraya, M. Yoshimura, S. Somiya, J. Am. Ceram. Soc. 1984, C119. [25] H. Toraya, M. Yoshimura, S. Somiya, J. Am. Ceram. Soc. 1984, C183. for Crystallography (Ed: International Union of 7 Fig. 1. Absolute value of the dielectric constant for the ordinary (Eu) and extraordinary (A2u) modes of alumina and tetragonal zirconia. Fig. 2. a) Near-normal reflectance spectra of alumina/ 3Y-TZP composites in the region of the minimum of reflectance. b) Minimum of reflectance vs 3Y-TZP volume fraction (f). The solid lines are guides to the eyes. 8 Fig. 3. Fraction of monoclinic phase vs time of steam sterilization for alumina/3Y-TZP composites with different zirconia content. o: 7 vol.-% zirconia, n: 14 vol.-%, m: 22 vol.-%, : ¢40 vol.-%, d: l100 vol.-%. Fig. 4. SEM micrographs of polished surface of alumina/22 vol.-% 3Y-TZP a) before and b) after 115 h steam sterilization. 9 Fig. 5. Fraction of monoclinic phase measured after 40 h steam sterilization as a function of the zirconia volume content in the composite. The dashed area represents the percolation threshold measured by IR study. 10
1710.00047	1	1710	2017-09-15T18:13:00	Ultrathin graphene-based membrane with precise molecular sieving and ultrafast solvent permeation	[ "physics.app-ph", "cond-mat.mtrl-sci", "physics.chem-ph" ]	Graphene oxide (GO) membranes continue to attract intense interest due to their unique molecular sieving properties combined with fast permeation rates. However, the membranes' use has been limited mostly to aqueous solutions because GO membranes appear to be impermeable to organic solvents, a phenomenon not fully understood yet. Here, we report efficient and fast filtration of organic solutions through GO laminates containing smooth two-dimensional (2D) capillaries made from flakes with large sizes of ~ 10-20 micron. Without sacrificing their sieving characteristics, such membranes can be made exceptionally thin, down to ~ 10 nm, which translates into fast permeation of not only water but also organic solvents. We attribute the organic solvent permeation and sieving properties of ultrathin GO laminates to the presence of randomly distributed pinholes that are interconnected by short graphene channels with a width of 1 nm. With increasing the membrane thickness, the organic solvent permeation rates decay exponentially but water continues to permeate fast, in agreement with previous reports. The application potential of our ultrathin laminates for organic-solvent nanofiltration is demonstrated by showing >99.9% rejection of various organic dyes with small molecular weights dissolved in methanol. Our work significantly expands possibilities for the use of GO membranes in purification, filtration and related technologies.	physics.app-ph	physics	Ultrathin graphene-based membrane with precise molecular sieving and ultrafast solvent permeation Q. Yang1,2,†, Y. Su1,2,†, C. Chi1,2,†, C. T. Cherian1,2, K. Huang1,2, V. G. Kravets3, F. C. Wang4, J. C. Zhang5, A. Pratt5, A. N. Grigorenko3, F. Guinea3,6, A. K Geim3, R. R. Nair1,2 1National Graphene Institute, University of Manchester, Manchester, M13 9PL, UK. 2School of Chemical Engineering and Analytical Science, University of Manchester, Manchester, M13 9PL, UK. 3School of Physics and Astronomy, University of Manchester, Manchester M13 9PL, UK. 4Chinese Academy of Sciences Key Laboratory of Mechanical Behavior and Design of Materials, Department of Modern Mechanics, University of Science and Technology of China, Hefei, Anhui 230027, China. 5Department of Physics, University of York, York YO10 5DD, UK. 6Imdea Nanociencia, Faraday 9, 28015 Madrid, Spain. †These authors contributed equally to this work. *[email protected] & [email protected] Graphene oxide (GO) membranes continue to attract intense interest due to their unique molecular sieving properties combined with fast permeation rates1-9. However, the membranes' use has been limited mostly to aqueous solutions because GO membranes appear to be impermeable to organic solvents1, a phenomenon not fully understood yet. Here, we report efficient and fast filtration of organic solutions through GO laminates containing smooth two-dimensional (2D) capillaries made from flakes with large sizes of ~ 10-20 µm. Without sacrificing their sieving characteristics, such membranes can be made exceptionally thin, down to 10 nm, which translates into fast permeation of not only water but also organic solvents. We attribute the organic solvent permeation and sieving properties of ultrathin GO laminates to the presence of randomly distributed pinholes that are interconnected by short graphene channels with a width of 1 nm. With increasing the membrane thickness, the organic solvent permeation rates decay exponentially but water continues to permeate fast, in agreement with previous reports1-4. The application potential of our ultrathin laminates for organic-solvent nanofiltration is demonstrated by showing >99.9% rejection of various organic dyes with small molecular weights dissolved in methanol. Our work significantly expands possibilities for the use of GO membranes in purification, filtration and related technologies. 1 Membrane-based technologies enable efficient and energy-saving separation processes which could play an important role in human life by purifying water or harvesting green energy10,11. Recently, it was shown that molecular separation processes could benefit from development of graphene-based membranes2-4 that show tunability in pore size8,12-15 and ultimate permeance15 defined by their thinness. In particular, GO-based membranes are considered to be extremely promising for molecular separation and filtration applications due to their mechanical robustness and realistic prospects for industrial scale production2-4,7,9. A considerable progress in nanofiltration through GO membranes2-4,16 was achieved mainly for water (due to its ultrafast permeation1-4) while organic-solvent permeation has received limited attention. This disparity is rather surprising as organic solvent nanofiltration (OSN) attracts a tremendous interest due to its prospective applications in chemical and pharmaceutical industries11,17-20. The development of novel inorganic membranes for OSN is particularly vital because of the known instability of many polymer-based membranes in organic solvents. The possible lack of motivation for exploiting graphene-based membranes for OSN could have come from the previous reports on impermeability of organic solvents through sub-micron thick GO membranes that remained highly permeable for water1,2,21. Although some latest studies report the swelling of GO membranes in organic solvents and, accordingly, thick GO membranes22,23, this seems inconsistent with the previous reports1,2,21 and could be explained by the presence of extra defects that produce a molecular pathway. In an another work24 OSN was performed using a solvated reduced GO-polymer composite membrane and only achieved a molecular sieve size of ≈ 3.5 nm due to the larger nanochannels in the membrane than that of pristine GO membranes1,2,5. Molecular rejection for the above membranes involves charge specific separation rather than the physical size cutoff. Membranes with Angstrom size precise sieving along with high organic solvent permeance are of great interests for OSN technology, however, such demostration is still lacking. In this report, we investigate permeability and sieving properties of ultrathin GO membranes with respect to organic solutions using an improved laminar structure and demonstrate the membranes' potential for OSN. indicate permeability of organic molecules even through The preparation of GO membranes used in our work is described in Methods. Figure 1 shows the scanning electron microscope (SEM), atomic force microscope (AFM) images and X-ray diffraction (XRD) of the studied GO membranes. Short duration ultrasonic exfoliation and a stepwise separation (Methods) were used to obtain large GO flakes (lateral size D of 10 – 20 µm) with a relatively narrow size distribution (supplementary Fig. 1). The membranes prepared from these large GO flakes are referred to as highly laminated GO (HLGO) membranes due to their superior laminar structure. They show a narrow XRD peak (full width at half maximum of 0.4 degree) as compared to 1.6 degree for the standard GO membranes prepared from smaller flakes (D  0.1 – 0.6 µm). Below the latter are referred to 2 as the conventional GO (CGO) membrane. The narrow X-ray peak for HLGO laminates suggests the importance of the GO flake size for the alignment process, which can be attributed to stronger interlayer interaction between larger overlapping areas25. The stronger interactions could further assist to eliminate the occasional wrinkles and corrugation found in the CGO membranes2,3, and this could lead to achieving smoother 2D capillaries in HLGO membranes. Figure 1 Ultrathin HLGO membrane. (a) SEM image of an 8 nm thick HLGO membrane on an Anodisc alumina support. Scale bar, 1µm. Inset: SEM image of bare alumina support. Scale bar, 500 nm. (b) X-ray diffraction for HLGO and CGO membranes. Inset (left): AFM image of HLGO membrane transferred from an alumina substrate to a silicon wafer. Scale bar, 500 nm. Inset (right): The height profiles along the dotted rectangle. To probe molecular sieving properties of HLGO membranes, we first performed vacuum filtration of aqueous solutions of several salts and large molecules through HLGO membranes (Methods). Figure 2a shows the molecular sieving properties of an 8 nm thin HLGO membrane. Similar to micron-thick GO membranes5, HLGO membranes also block all ions with hydrated radii larger than 4.5 Å. We emphasize that no molecular sieving was observed in similar experiments but using CGO membranes with thickness of 8-50 nm (Fig. 2a inset). Hence, the ultra-sharp sieving cut-off can be achieved in HLGO membranes that are more than two orders of magnitude thinner than conventional membranes showing same sieving properties5. This drastic improvement can be attributed to the highly laminated nature of our HLGO membranes. We failed to observe a cut-off in sieving only for the membranes thinner than 8 nm, which sets a minimum thickness for HLGO membranes used in this study. Ultrahigh permeance to fluids may occur in ultrathin membranes due to a decreased molecular permeation length6,15. To further evaluate liquid permeance of HLGO membranes, we have performed vacuum filtration and dead-end pressure filtration (supplementary section2) experiments with water and a wide range of organic solvents using only 8 nm thick 3 membranes. All the permeance values were recorded after reaching a steady state condition, typically achieved within 30 minutes. The liquid flux is found to be linearly proportional the differential pressure (ΔP) across an HLGO membrane (Fig. 2b inset). The permeance for various solvents as a function of their inverse viscosity (1/) is shown in Fig. 2b. In contrast to much-thicker GO membranes that exhibit ultrafast water permeation and impermeability for organic solvents1, our HLGO membranes are highly permeable to all tested solvents. The highest permeance is observed for solvents with the lowest viscosity. For example, hexane shows permeance of ~18 Lm-2h-1bar-1, i.e, a permeability of ~144 nm·Lm-2h-1bar-1, despite the fact that its kinetic diameter is almost twice larger than that of water26. On the contrary, 1- butanol with a kinetic diameter similar to that of hexane26 but much higher viscosity exhibits the lowest permeance of 2.5 Lm-2h-1bar-1. The linear dependence of permeance on 1/ (see Fig.2b) clearly indicates that the solvent viscosity dictates its permeability and proves the viscous nature of the solvents' flow through HLGO membranes. Figure 2 Molecular sieving and organic solvent nanofiltration through HLGO membranes. (a) Experiments for salt rejection as a function of ion's hydrated radius (largest ions within the aqueous solutions are plotted). The HLGO membranes are 8 nm thick. The hydrated radii are taken from ref. [5 and 7]. MB- Methylene Blue, RB – Rose Bengal, BB – Brilliant Blue. Inset: MB rejection and water permeance exhibited by the standard GO membrane with different thicknesses (colour coded axes). (b) Permeance of pure organic solvents through an 8 nm HLGO membrane as a function of their inverse viscosity. The used solvents are numbered and named on the right. Inset (top): Methanol permeance as a function of pressure gradient (ΔP). Dotted lines: Best linear fits. (c) Rejection and permeance of several dyes in methanol versus their molecular weight (colour coded axes). The dyes used: Chrysoidine G (CG), Disperse Red (DR), MB, Crystal Violet (CV), BB and RB. Left inset: Photographs of dyes dissolved in methanol before and after filtration through 8 nm HLGO membranes. Right inset: MB rejection and methanol permeance of CGO membrane with different thicknesses (colour coded axes). Note that even though the dye rejection increases and approaches ~ 90% with increasing the CGO membrane thickness their permeance is 4 significantly lower than 8 nm HLGO membranes. All the error bars are standard deviations using at least three different measurements using different samples. Points within the grey bar in Figs.1a and c show the rejection estimated from the detection limit (supplementary Fig. 4 and Methods). High permeance of organic solvents combined with accurate molecular sieving makes ultrathin HLGO membranes attractive for OSN17,18. To evaluate this potential for applications, we have performed filtration experiments with methanol solutions of several dye molecules. The dye molecule rejection rates for an 8 nm thick HLGO membrane are presented in Fig. 2c. While the permeance was reduced by ~10-30% compared to the pure solvent (which is not unusual for nanofiltration16,27), no dye molecule could be detected down to 0.1% (our detection limit) of the feed concentration at the permeate side (Fig. 2b). The observed 100% dye rejection and fast solvent permeation makes our ultrathin HLGO membranes superior to the state-of-the-art polymeric membranes17,19. For example, the highest methanol permeance reported on polymeric membranes17,19 is ~ 1.6 Lm-2h-1bar-1 for 90% RB rejection which is ≈ 5 times lower than the methanol permeance obtained with our HLGO membranes providing ≈ 100% RB rejection. Further comparison of OSN performance of HLGO membrane and the other reported OSN membranes are listed in supplementary section 4. Unambiguously, a high organic solvent permeance along with precise molecular sieving (> 99.9% rejection to the dye molecules) indicates that HLGO membranes could be an outstanding candidate for OSN technology. With the view of practical applications, we have also performed OSN experiments with HLGO deposited on porous polymer (nylon) support (supplementary section 5). The nylon supported HLGO membranes showed nearly the same performance as those on the alumina support. For example, an 8 nm HLGO membrane on nylon showed a >99.9% rejection to MB with ≈ 7 Lm-2h-1bar-1 methanol permeance (supplementary Fig. 6). We have also studied the influences of aging and solvent stability of HLGO membrane, which are key parameters for practical applications, on its membrane performance, and found that HLGO membranes are stable in air for more than a year and also highly stable in different solvents (see supplementary section 6). To elucidate the mechanism of organic-solvent permeation and sieving properties of ultrathin HLGO membranes, we have conducted two sets of additional experiments. First, we have performed XRD for HLGO membranes immersed in different organic solvents, see Fig. 3a. The data clearly indicate that several of the organic solvents, especially polar ones, intercalate between graphene oxide layers and increase the interlayer distance, d. However, non-polar solvents, such as hexane, did not produce any increase in d. At the same time, hexane was the fastest permeating molecule among the solvents used in this study (Fig.2b). This suggests that permeation through ultrathin HLGO membranes is not dominated by molecular transport through interlayer capillaries1. Second, we performed water and organic solvent permeation 5 experiments using HLGO membranes of different thicknesses, h. Fig. 3b shows the exponential decay for methanol and hexane permeance as a function of h. HLGO membranes with h > 70 nm show no detectable solvent permeation, consistent with the impermeability reported for sub-micron thick GO membranes1. Using helium and organic vapours, we also observed a similar, exponential decay with increasing h of our HLGO membranes (supplementary section 7). In contrast, water permeance initially decayed exponentially, too, but for h > 70 nm it followed a much weaker, linear dependence on 1/h (Fig.3b inset). Figure 3 Probing molecular permeation through HLGO membranes. (a) X-ray diffraction for 70 nm thick HLGO membranes immersed in various organic solvents (colour coded). (b) Thickness dependence of permeance for methanol, hexane, and water through HLGO membranes (colour coded). Red and blue dotted lines are the best exponential fits. The black dotted curve is a guide to the eye. Inset: Water permeance as a function of inverse thickness for HLGO membranes with thicknesses ≥100 nm. Dotted line: best linear fit. The slope of linear fit provides the water permeability as ≈ 32 nm·Lm-2h-1bar-1. The solid line in the main figure shows the detection limit for methanol and hexane in our experiment. All the error bars are standard deviations using at least three different measurements using different samples. The exponential decrease of organic-solvent permeance with h is surprising and seemingly contradicts to the viscous flow inferred from the observed 1/ dependence. Indeed, the viscous flow suggests that the permeance should be proportional to the pressure gradient ∆(cid:3017)(cid:3013) , where ∆(cid:1842) is the driving pressure gradient and L is the permeation length (proportional h)28,29. For example, the linear dependence of water permeance on 1/h for the thicker membrane is consistent with the viscous flow. To explain these two functional dependences, we propose two different molecular pathways for permeation through HLGO membranes. The first 6 involves permeation through pin holes (pathway 1) and the second one is through the previously suggested model of a network of graphene capillaries1,5 (pathway 2). Pin holes in GO membranes originate from random stacking of individual GO flakes and can also involve nanometre size holes2 within flakes. At a few nm thicknesses, GO laminates contain many pinholes (supplementary section 8) that pierce through the entire film. Such thin GO films allow relatively easy permeation through pinholes without any atomic-size cutoff observed for thicker laminates. At a certain critical thickness hc, GO films become continuous with all pinholes blocked, as the found onset of atomic-scale sieving indicates. The experiment shows that for HLGO membranes, hc is 8 nm. After this threshold, molecular transport is expected to occur in two steps. Liquids continue to rapidly fill the same pinholes but this is not a limiting process. Molecular transport through the entire film becomes limited by the necessity to reach from one pinhole to another, which involves in- plane diffusion between GO sheets. This bottleneck has to involve interlayer diffusion by a distance of the order of the size of GO sheets, which will provide an atomic-scale sieve size for filtration. Assuming that a probability for a molecule to find a pathway through the thinnest continuous membrane with critical thickness hc is p, we can write the probability of transport through a thicker sheet of thickness h as P = pN where N =h/hc. This can be re- written as P = exp[ln(p)h/hc] and yields the flux Q  exp(-h/a) with a = hc /ln(1/p). By definition p should be of the order of ½ because we define it at the threshold, which means that p  1 for h < hc and p  0 for thicker layers. Therefore, a = hc /ln(1/p)  hc, in agreement with exponential fit in Fig. 3b. This proposed model could also explain why the molecular sieve size of ≈ 4.5 Å is not preserved in thin CGO membranes where the smaller flake size increased the critical thickness and weakened the interlayer alignment. The deviation of water permeance from the exponential decay and it's faster transport at large h can be understood by considering the molecular pathway 2 where the permeation occurs through the graphene capillaries1,5 that develop between GO sheets (typically, an area of 40- 60% remains free from functionalization30,31). The permeation through the pathway 2 is primarily restricted by the hydraulic resistance due to a large L ((cid:3005)(cid:3031)×h)1. However, water permeation through these capillaries experiences three orders of magnitude enhanced flow due to the large slip length1,8,29 and therefore effectively reduces the flow resistance. This suggests that with increasing h, the exponentially growing flow resistance for water in the pathway 1 could be overcome by the lower flow resistance in the pathway 2 due to the large slip length, consistent with the deviation from the exponential decay observed above ~ 50 nm in Fig. 3b. The linear dependence of water permeance on 1/h for the HLGO membranes with a thickness larger than 70 nm (Fig. 3b inset) further proves that the flow through the thicker membranes predominantly occurs through the graphene capillaries29. In contrast, for organic 7 solvents, the experimentally undetectable permeation for h > 70 nm indicate that the permeation through the pathway 2 is negligible and suggests a non-slip flow. This is not surprising because graphitic surfaces are known for their lipophilicity, that is, they interact strongly with hydrocarbons. This is consistent with the recent calculation of larger interfacial friction for ethanol in graphene capillaries compared to water32. The non-slip behavior of organic solvents also explains why certain organic molecules (polar solvents) can uniformly intercalate between GO layers, similar to water, but their permeability remains below our detection limit. To confirm this model further, we have also ruled out the influence of polarity, dynamic diameter, and solubility parameters of organic solvents to their permeance (supplementary section 9). Based on the understanding of organic molecule permeation through GO membranes, we propose a strategy to further improve the permeance through GO membranes without substantially reducing the organic solute rejection, even using relatively thick membranes. To this end, we used partially reduced Mg2+ crosslinked GO membrane with 200 nm thickness, where the randomly distributed Mg2+ ions between GO sheets play a role of spacers that introduces the disorder in the laminar structure and hence increases the permeance (supplementary Fig. 11 and supplementary section 10). These modified membranes show ~ 50% increase in permeance while keeping the dye rejection at 98% (supplementary Fig. 12). We believe that the performance of such ion-modified membranes could be further improved by optimising the cation selection for crosslinking and careful control of the reduction process. In conclusion, we show that HLGO membranes of only several layers in thickness exhibit outstanding sieving properties accompanied by ultrafast solvent permeation. The proposed model based on non-slip permeation of organic solvents and slip-enhanced water permeation offers a long-sought explanation for the ultra-low permeability of sub-micron thick CGO membranes for organic solvents. Taking into account the excellent chemical stability of GO, the reported membrane can be used for organic solvent nanofiltration, with pharmaceutical and petrochemical industries being potential beneficiaries. The proposed strategy to enhance the nanofiltration properties of GO membranes by cation-crosslinking is enticing but further research is needed to optimize the performance. Methods Preparation of GO membranes: Graphite oxide was prepared by the Hummers method and then dispersed in water by sonication1, which resulted in stable GO solutions. It is worth noting that, to avoid other possible influence, we repeatedly washed our GO nanosheets until the pH value of their solutions reaches to 7. GO membranes were prepared by vacuum filtering aqueous GO solutions through Anodisc Alumina or Nylon membrane5 (47 mm diameter Whatman filters with 200 nm pore size). To obtain a uniform membrane, the GO 8 suspension was diluted to less than 0.001 wt% before the vacuum filtration. After filtration, the membrane was allowed to dry under vacuum at room temperature for at least 24 hours before the measurements. The membranes with different thickness were obtained by filtrating different volume of GO suspension through the Alumina or Nylon support. It is noteworthy that the influence of the support membrane on the reported permeation were minimum due the large porosity. The solvent permeance through both the bare Alumina and Nylon support layers were found to be minimum 1000 times larger than the GO membrane on support layers suggesting hydraulic resistance from the support layers were negligible and could be ignored. Two types of GO membranes used in this study are HLGO and CGO membranes. The difference between preparation of HLGO and CGO membrane lies in the ultrasonic exfoliation and centrifugal separation process. For HLGO membranes, the graphite oxide was exfoliated by a 3-minute ultrasonic exfoliation (40 W power) and then subsequently centrifuged twice at 3000 rpm for 10 minutes to separate un-exfoliated thick GO flakes. The supernatant GO solution was further centrifuged at 12000 rpm to separate large and small GO flakes. In this step, the sediment was collected because the small size and hence lighter GO flakes remain in the supernatant and larger GO flakes sediments. This sediment was then collected and re-dispersed in water by mild shaking and then repeated the centrifugation steps at 10000 and 8000 rpm respectively. This repeated centrifugation cycles with sequentially decreasing centrifugation speed enable the separation of medium size GO flakes from the large flakes and allows obtaining uniform large GO flakes required for the preparation of HLGO membranes. For the preparation of CGO membranes, the graphite oxide in water was sonicated for 24 hours and then centrifuged three times at 8000 rpm. The supernatant was then collected and used for the membrane preparation. It is important to note that, the HLGO and CGO membranes are prepared by identical procedures except different exfoliation time. The influence of which on the chemical composition of GO sheets was carefully examined by X-ray photoelectron spectroscopy (XPS) and found no difference in oxygen content between two membranes (supplementary section 1 and supplementary Fig. 2). The flake size distribution of GO used for the preparation of conventional CGO and HLGO membranes were measured by analysing more than 700 flakes with the scanning electron microscopy (SEM) or optical microscopy. Due to long time ultrasonication, all the GO flakes used for the CGO membranes are found to be smaller than 1 µm in nominal size and more than 75% of these flakes are with a size between 0.1-0.4 µm. In comparison, for HLGO membranes, 75% of the flakes used were found to be larger than 10 µm (supplementary Fig. 1). 9 Membrane characterizations: SEM and AFM techniques were used to measure the size of GO flakes and thickness of the membranes. A Veeco Dimension 3100AFM in the tapping mode was used for the AFM measurements. To measure the thickness of the GO membranes, we transferred the membrane from the alumina support to a silicon substrate by floating the alumina supported GO membrane in water and subsequently fishing out the GO membrane onto a silicon substrate. GO membrane transferred silicon substrates were completely dried in vacuum before the AFM measurements. X-ray diffraction measurements in the 2θ range of 5° to 25° (with a step size of 0.02° and recording rate of 0.2 s) were performed using a Bruker D8 diffractometer with Cu Kα radiation (λ = 1.5406 Å). Due to the weak intensity of the X-ray peak from an 8 nm membrane we used 70 nm thick membranes for our experiments. To collect an XRD spectrum from HLGO membranes exposed to different organic solvents, the membranes were first aged in a glovebox filled with dry argon gas for more than 5 days to remove any interlayer water present in the membranes1,5 and then immersed in various solvents for more than 3 days inside a glove box. For the XRD measurements, the samples were collected from the solvents and kept inside an airtight XRD sample holder (Bruker, A100B36/B37) filled with same organic solvent vapour to avoid any influences of the environmental humidity and evaporation of solvent from the membrane on the measurements. Permeation and molecular sieving measurements: For probing the molecular sieving and solvent permeation through various GO membranes we used a vacuum filtration setup, where the membrane is clamped and sealed with a silicone rubber O-ring between the feed and permeate side. For each test, at least three membranes were used to validate the reproducibility. Permeate side was connected to a vacuum pump with a controllable pumping speed and a cold trap. The vacuum on the permeate side creates a pressure gradient (ΔP) which drives the molecular permeation across the membrane. Vacuum degree on the permeate side was controllable from 0.6 to 0.01 bar (VARIO chemistry diaphragm pump, Vacuubrand) and feed pressure was ≈1 bar. For studying the influence of ΔP on the permeance, we have performed filtration experiments with different ΔP created using different pumping speed. The permeance of various solvents was obtained by measuring both the volume and weight of the solvent from the permeate side in a liquid nitrogen cold trap and the liquid leftover in the feed side. The system leakage was examined by replacing the membrane with a 100 µm polyethylene terephthalate plastic sheet, or a 200 µm Cu foil, the leakage was found to be < 0.1 L m-2 h-1 bar-1. The solvent permeance through GO membrane was also measured by a dead-end pressure filtration system at room temperature and found good consistency between two methods (Supplementary section 2). 10 We have noticed that for water due to its high surface tension the HLGO membrane breaks once the water was in contact with the membrane. We, therefore, used a small amount of surfactant (0.6 mg/mL sodium dodecyl benzene sulfonate) to decrease the surface tension of water and thereby avoiding the membrane damage during water permeation experiments. For probing the molecular sieving property of HLGO and CGO membranes we used aqueous solutions of NaCl, MgCl2, K3[Fe(CN)6], pyrenetetrasulfonic acid tetrasodium salt (Na4PTS), MB, RB, and BB. For MB, RB, and BB the feed concentrations were 20 mg/L, and for K3[Fe(CN)6] and Na4PTS, their concentrations were 1000, 250 mg/L, respectively. For NaCl and MgCl2 we used 1M concentration. All the experiments were repeated at least three times. The amount of sodium and magnesium salts permeated were measured by probing the concentration of salt in the permeate side by checking the conductivity of the permeate water. Furthermore, we cross-checked the results of our conductivity analysis by weighing the dry material left after evaporation of water in the permeate. The permeation of other salts and dyes through GO membranes was measured by checking their concentration at the permeate side by UV-vis absorption as detailed below. The salt rejection was calculated as (1-CP/CF), where Cp is the salt concentration at the permeate side and CF is the salt concentration at the feed side. For organic solvent nanofiltration experiments, CG, MB, DR, CV, BB, and RB with a concentration of 200 mg/L were dissolved in methanol. The concentration of the dye at the permeate side was measured by UV-vis absorption as detailed below and the permeance was determined by the same method for the measurement of pure solvent as detailed above. UV-Vis absorption: For obtaining the concentrations of K3[Fe(CN)6], Na4PTS and organic dye molecules in the permeate we used optical absorption spectroscopy. UV-visible-near- infrared grating spectrometer with a xenon lamp source (240-1700 nm) was used for this study. For the HLGO membranes, we could not detect any absorption features of the above salts or dye in the permeate side (supplementary Fig. 4). To cross check this further, we have also measured the concentration of the leftover feed solution after the filtration experiment. The leftover concentrated feed solutions (including the salt or dye absorbed on the membrane) were diluted to the same volume as before the filtration experiment and then the optical absorption features were compared with the pristine original feed solution. We could not find any difference in the absorption spectra, suggesting all the solutes were retained at the feed side. The detection limit in Fig. 2a and c were estimated by measuring a reference solution and gradually decreasing its concentration until the signature peaks completely disappeared. The penultimate concentration is set as the corresponding detection limit. For the case of CGO membranes and Mg2+-crosslinked membranes (Fig. 2a, 2c, and supplementary Fig. 12), the absorbance for the most intense optical absorption peak for various known concentrations 11 of salt and dye molecules were plotted against their concentration and obtained a linear fit. From this linear dependence, we estimated the concentration of salt and dye at the permeate side. Reference 1. Nair, R. R., Wu, H. A., Jayaram, P. N., Grigorieva, I. V., Geim, A. K. Unimpeded permeation of water through helium-leak-tight graphene-based membranes. Science 335, 442- 444 (2012). 2. Sun, P., Wang, K., Zhu, H. Recent developments in graphene-based membranes: Structure, mass-transport mechanism and potential applications. Adv. Mater. 28, 2287-2310 (2016). 3. Liu, G., Jin, W., Xu, N. Graphene-based membranes. Chem. Soc. Rev. 44, 5016-5030 (2015). 4. Fathizadeh, M., Xu, W. L., Zhou, F., Yoon, Y., Yu, M. Graphene oxide: A novel 2- dimensional material in membrane separation for water purification. Adv. Mater. Interf. 1600918 (2017). 5. Joshi, R. K. et al. Precise and ultrafast molecular sieving through graphene oxide membranes. Science 343, 752-754 (2014). 6. Li, H. et al. Ultrathin, molecular-sieving graphene oxide membranes for selective hydrogen separation. Science 342, 95-98 (2013). 7. Akbari, A. et al. Large-area graphene-based nanofiltration membranes by shear alignment of discotic nematic liquid crystals of graphene oxide. Nat. Commun. 7, 10891 (2016). 8. Abraham, J. et al. Tuneable sieving of ions using graphene oxide membranes. Nat. Nanotech. 12, 546–550 (2017). 9. Hong, S. et al. Scalable graphene-based membranes for ionic sieving with ultrahigh charge selectivity. Nano Lett. 17, 728-732 (2017). 10. Mulder, J. Basic principles of membrane technology. (Springer Science & Business Media, 2012). 11. Koros, W. J., Zhang, C. Materials for next-generation molecularly selective synthetic membranes. Nat. Mater. 16, 289-297 (2017). 12. Wang, L. et al. Molecular valves for controlling gas phase transport made from discrete ångström-sized pores in graphene. Nat. Nanotech. 10, 785-790 (2015). 13. Jain, T. et al. Heterogeneous sub-continuum ionic transport in statistically isolated graphene nanopores. Nat. Nanotech. 10, 1053-1057 (2015). 14. O'Hern, S. C. et al. Selective ionic transport through tunable subnanometer pores in single- layer graphene membranes. Nano Lett. 14, 1234-1241 (2014). 15. Celebi, K. et al. Ultimate permeation across atomically thin porous graphene. Science 344, 289-292 (2014). 16. Han, Y., Xu, Z., Gao, C. Ultrathin graphene nanofiltration membrane for water purification. Adv. Funct. Mater. 23, 3693-3700 (2013). 12 17. Marchetti, P., Jimenez Solomon, M. F., Szekely, G., Livingston, A. G. Molecular separation with organic solvent nanofiltration: A critical review. Chem. Rev. 114, 10735-10806 (2014). 18. Vandezande, P., Gevers, L. E., Vankelecom, I. F. Solvent resistant nanofiltration: separating on a molecular level. Chem. Soc. Rev. 37, 365-405 (2008). 19. Jimenez-Solomon, M. F., Song, Q., Jelfs, K. E., Munoz-Ibanez, M., Livingston, A. G. Polymer nanofilms with enhanced microporosity by interfacial polymerization. Nat. Mater. 15, 760-767 (2016). 20. Karan, S., Jiang, Z., Livingston, A. G. Sub–10 nm polyamide nanofilms with ultrafast solvent transport for molecular separation. Science 348, 1347-1351 (2015). 21. Huang, K. et al. A graphene oxide membrane with highly selective molecular separation of aqueous organic solution. Angew. Chem. Int. Ed. 53, 6929-6932 (2014). 22. Huang, L., Li, Y., Zhou, Q., Yuan, W., Shi, G. Graphene oxide membranes with tunable semipermeability in organic solvents. Adv. Mater. 27, 3797-3802 (2015). 23. Aba, N. F. D., Chong, J. Y., Wang, B., Mattevi, C., Li, K. Graphene oxide membranes on ceramic hollow fibers – Microstructural stability and nanofiltration performance. J. Membr. Sci. 484, 87-94 (2015). 24. Huang, L. et al. Reduced graphene oxide membranes for ultrafast organic solvent nanofiltration. Adv. Mater. 28, 8669-8674 (2016). 25. Lin, X. et al. Fabrication of highly-aligned, conductive, and strong graphene papers using ultralarge graphene oxide sheets. ACS Nano 6, 10708-10719 (2012). 26. Wu, H., Gong, Q., Olson, D. H., Li, J. Commensurate adsorption of hydrocarbons and alcohols in microporous metal organic frameworks. Chem. Rev. 112, 836-868 (2012). 27. Karan, S., Samitsu, S., Peng, X., Kurashima, K. & Ichinose, I. Ultrafast viscous permeation of organic solvents through diamond-like carbon nanosheets. Science 335, 444-447 (2012). 28. Secchi, E., et al. Massive radius-dependent flow slippage in carbon nanotubes. Nature 537, 210-213 (2016). 29. Radha, B., et al. Molecular transport through capillaries made with atomic-scale precision. Nature 538, 222-225 (2016). 30. Wilson, N. R. et al. Graphene oxide: structural analysis and application as a highly transparent support for electron microscopy. ACS Nano 3, 2547-2556 (2009). 31. Loh, K. P., Bao, Q., Eda, G. & Chhowalla, M. Graphene oxide as a chemically tunable platform for optical applications. Nature Chem. 2, 1015-1024 (2010). 32. Dai, H., Liu, S., Zhao, M., Xu, Z., Yang, X. Interfacial friction of ethanol–water mixtures in graphene pores. Microfluid Nanofluidics 20, 141 (2016). 13 Supplementary Information Ultrathin graphene-based membrane with precise molecular sieving and ultrafast solvent permeation Q. Yang, Y. Su, C. Chi, C. T. Cherian, K. Huang, V. G. Kravets, F. C. Wang, J. C. Zhang, A. Pratt, A. N. Grigorenko, F. Guinea, A. K Geim, R. R. Nair 1. Graphene Oxide (GO) flakes with different sizes Supplementary Fig.1 GO flake size distribution. (a) SEM image of GO flakes used for the preparation of CGO membranes (Scale bar, 200 nm) and (b) its flake size distribution. (c) Optical image of GO flakes used for the preparation of HLGO membranes (Scale bar, 20 µm) and (d) its flake size distribution. The flake sizes were estimated by taking the square root of the area of each flake measured with the Image J software. The influence of flake size on the chemical composition of GO sheets was carefully examined by X-ray photoelectron spectroscopy (XPS). XPS experiments were performed using a monochromated Al K source (1486.6 eV) in an ultrahigh vacuum system with a base pressure of < 2×10-10 mbar. Survey scans were taken to confirm that only C and O were present in each sample before high-resolution C 1s spectra were obtained (supplementary Fig. 14 2). Using XPS Peak 4.1, each C 1s spectrum was fitted with four components representing the main bonding environments found in graphene oxide: C-C (284.5-284.8 eV), C-O (285.2- 285.4 eV), C=O (286.8-287.2 eV), and C(=O)-(OH) (288.1-289.1 eV)1,2. Fitted peak areas were used to calculate C/O ratios of 3.3 ± 0.3 and 3.6 ± 0.3 for HLGO and CGO membranes, respectively. The corresponding oxygen content of 23 ± 2% for HLGO and 22 ± 2% for CGO clearly indicates that the size of GO flake does not influence their oxygen content and is consistent with the previous reports where the oxygen content is found insensitive to the GO flake size3,4. Supplementary Fig.2 X-ray photoelectron spectroscopy of CGO and HLGO membrane. XPS spectra from (a) CGO and (b) HLGO membrane showing raw data (black line), the fitting envelope (red line), and deconvolved peaks (blue lines) attributed to the chemical environments indicated. With respect to C 1s peak, C-C, C-O, C=O and CO(OH) peaks have an area of 61±3%, 13±2%, 22±2%, and 4±1% respectively for CGO membrane and 58±3%,14±2%, 25±2%, and 3±1% respectively for HLGO membrane. 2. Dead-end pressure filtration In addition to the vacuum filtration, HLGO membranes were also examined by pressure filtration using a home-made dead-end set-up with a pressure up to 2 bar (supplementary Fig. 3 inset). Supplementary Fig. 3a shows the methanol permeance as a function of pressure for 8, 15 and 50 nm thick HLGO membranes. As expected, the permeance was found to increase linearly with the applied pressure. 8 nm membranes were also tested for dye rejection (methylene blue, similar conditions to that of vacuum filtration) and obtained >99.9% rejection. Supplementary Fig. 3b shows the water and methanol permeance as a function membrane thickness. Similar to the case of vacuum filtration, the organic solvents' 15 permeance decreased exponentially with the thickness whereas for water it deviates this behaviour above 50 nm. For the thicker membranes (> 50 nm) we have also used the Sterlitech HP4750 stirred cell for the dead-end filtration experiments (thinner membranes were found easy to get damaged in this cell during the sample mounting) and obtained similar results to that from the home-made pressure cell and the vacuum filtration. We have also studied the influence of membrane thickness on the water permeance through CGO membranes and found that, even though the permeance is larger than HLGO membranes, their thickness dependent permeance follows the same trend as in the HLGO membranes (Supplementary Fig. 3b inset). Note that, despite its high permeance, the dye molecule rejection characteristics of CGO membrane were much inferior to that of HLGO membranes (Fig. 2a and c inset in main text). Supplementary Fig. 3 Permeation measured by dead-end pressure filtration. (a) Pressure dependence of methanol flux for HLGO membrane with a thickness of 8, 15 and 50 nm. Dotted lines are best linear fits. Inset: Schematic of dead-end pressure filtration setup. To avoid solvent leakage, a flat rubber gasket (marked as grey) is placed on top of the GO membrane (marked as brown), which is then clamped between two glass funnel-shaped containers. The upper compartment (marked as blue) is filled with feed solvent/solution and then controllably pressurised with compressed air. The permeate solvent/solution is collected at the bottom compartment (marked as white) and analysed as detailed in the methods session in the main text. (b) Thickness dependence of permeance for methanol and water through HLGO membranes. Inset: Thickness dependence of water permeance through CGO membranes. 16 3. Optical absorption measurements Supplementary Fig.4 Optical detection of permeate concentration. (a) Absorption spectra of the feed and permeate solution of K3[Fe(CN)6] and Na4PTS in water (colour coded). (b) Absorption spectra of the feed and permeate solution of chrysoidine G (CG), disperse red (DR), methylene blue (MB), crystal violet (CV), brilliant blue (BB), and rose bengal (RB) in methanol (colour coded). The absorption spectrum from an empty container was taken as a reference spectrum of all the measurements. 4. Organic solvent nanofiltration (OSN) performance comparison Supplementary Fig.5 Comparison of HLGO membrane performance with other OSN membranes. Permeance as a function of rejection for several dye molecules taken from the literature is plotted together with the data obtained from the HLGO membrane. HLGO membranes provided ≈ 100% rejection for all the tested dye molecules. Dotted line indicates the typical trend found between the rejection and permeance of reported values. All the data points in the green coloured regions are obtained from Ref. [5-10].CG- Chrysoidine G, BB- Brilliant Blue, RB- Rose Bengal. 17 To further demonstrate the superior OSN performance of HLGO membranes, we have compared them with different polymeric membranes. As an example, frequently used dye molecules such as Chrysoidine G (CG), Brilliant Blue (BB), and Rose Bengal (RB) in methanol have been chosen for the comparison. Supplementary Fig. 5 shows a typical trend between the methanol permeance and the molecular rejection values reported for several OSN membranes. Compared with the state-of- the-art polymeric membranes, HLGO membranes shows much higher permeance to solvents with a rejection of > 99.9% to dye molecules including CG whose molecular weight is only 249 g/mol. As an example, the highest reported rejection for BB in methanol is 95%, whereas the HLGO membranes exhibit a permeance of 7.5 Lm-2h-1bar-1 (25 times higher) with ≈ 100% rejection. The high permeance along with ≈100% rejection even for smaller molecules indicates the prospect of HLGO membranes for OSN technology. 5. HLGO membrane on porous nylon support Supplementary Fig.6 Ultrathin HLGO membrane on nylon support. (a) SEM image of an 8 nm HLGO membrane on a nylon support. Scale bar, 1 µm. Inset: SEM image of a bare nylon support. Scale bar, 1 µm. (b) Ambient air XRD spectrum for HLGO membrane on nylon support. The peaks at ~7o and 14o are from the nylon support. (c) Permeance and rejection of MB in methanol through HLGO membranes with different thicknesses on nylon support. The dotted line is the best linear fit. Points within the grey bar show the rejection estimated from the detection limit. In addition to the porous alumina support, which is brittle, we have also tested porous polymer as a support material. It has been reported that due to the roughness and non-uniform macroscopic pore distribution of polymer support, tens of nanometre thin GO membrane (small GO flakes) fails to maintain a good laminar structure11. Here, we show that GO membrane prepared from large GO flakes could form a good laminate, even though the membrane is ultrathin. Supplementary Fig. 6a shows the SEM image of a bare nylon support and an 8 nm HLGO membrane deposited nylon support. X-ray diffraction (XRD) spectrum of a 50 nm HLGO membrane on nylon substrate shows a narrow peak with a full width at half 18 maximum (FWHM) of 0.4 degree (supplementary Fig. 6b), which confirms the highly laminated structure similar to that on the alumina support. To evaluate the organic solvent nanofiltration (OSN), we have tested filtration of methanol solutions of CG and MB through an 8 nm thin HLGO membrane on nylon support. Similar to that of alumina support, HLGO membrane on nylon support also shows a 99.9% rejection to CG and MB with a similar methanol permeance to that of alumina support (supplementary Fig. 6c). Also, the exponential decay of the methanol permeance (supplementary Fig. 6c) with increasing the thickness of HLGO membrane is consistent with that of the alumina supported HLGO membranes (Fig 3b in main text). 6. Stability of HLGO membranes To study the stability of HLGO membrane in air and solvents, we have conducted two sets of experiments. First, for probing air stability, we have compared methanol and water permeance of freshly prepared membranes with aged membranes (two samples aged for 452 days). Both the membranes provided similar permeance indicating no significant degradation of the membrane with aging. This is consistent with ref. [12], where only ~ 6% reduction in oxygen content is reported with GO aging for 100 days. This small change in oxygen content is not expected to affect the membrane performance because the molecular permeation mainly occurs through the pristine graphene capillaries in the GO membrane and moreover, in HLGO membranes, permeation of solvents mainly occurs through the random pin holes and that is not anticipated to change with aging. Secondly, for probing the membrane stability in the solvent environment, we have performed long time filtration experiments. Nylon supported HLGO membranes with thicknesses of 8 nm, and 30 nm were examined under dead end filtration for water and organic solvents. As shown in supplementary Fig. 7, permeance of the solvents and water are stable within the testing period which varies from 5 hours to 4 days, suggesting that the HLGO membrane is intact and capable of long-time filtration process. To further check the solvent stability of membranes we have immersed 50 nm thin HLGO membranes on nylon in water, methanol, and hexane for 7 days. Note that the 50 nm thick samples are chosen because the membrane thinner than that gives poor visual contrast for analysis. As shown in supplementary Fig. 7, all the membranes were found stable in all tested solvents and water. To quantify this further, we have measured methanol permeance before and after immersing the membrane in solvents and could not find any detectable change. To demonstrate the solvent stability of the membrane further, we have performed another vigorous testing. Membranes immersed in solvents for 7 days were further placed in a glass beaker containing the same solvent and bubbled with nitrogen gas to check if the harsh dynamic turbulence would destroy the membrane. Membranes were found intact even in such a harsh environment. This is consistent with our previous report were membranes also found stable under mild sonication13. We do notice that there is a debate on the stability of GO membrane in water 19 and the currently proposed mechanism of stability is cross-linking of GO flakes with metal ions from the support substrate14. However, we believe many other factors could also lead to stability. For example, we found that complete drying of the membrane after its fabrication is critical for obtaining a stable membrane. Supplementary Fig. 7 Stability of GO membranes. (a) Variation of water and solvent (inset) permeance through HLGO membranes as a function of measurement time. (b) Photographs showing 50 nm thin HLGO and CGO membranes on nylon support immersed in water and solvents for 0 day and 7 days. In each experiment, a bare nylon substrate (white colored substrate in the photo) was also immersed as a reference to check visual contrast of thin GO membranes on nylon. Scale bar: 15 mm. 7. Vapour and Helium gas permeation through HLGO membranes Besides liquid permeation, vapour and gas (helium) permeation through HLGO membranes with different thicknesses (h) were measured to further validate the proposed mechanism for molecular transport in GO membranes. The vapour permeation measurements were performed as we reported previously15. Membranes were glued to a Cu foil with an opening 20 of 0.5 cm in diameter. The foil was then clamped between two rubber O-rings sealing a metal container. Permeation was measured by monitoring the weight loss (for ≈ 12 hours) of the container that was filled with water and isopropyl alcohol (IPA) inside a glovebox. Supplementary Fig. 8a shows the weight loss rate for water and IPA through HLGO membranes with different thicknesses. Weight loss rate for IPA was found to decay exponentially with increasing membrane thickness, indicating exponentially decaying permeance, consistent with the mechanism proposed (permeation through pinholes) in the main text. However, for water, we observed a thickness independent weight loss rate, consistent with the previous report15. In this case, unlike liquid permeation reported in the main text, water vapour permeation is limited by the evaporation from the top surface of GO membranes and hence masks the thickness dependence. Supplementary Fig. 8 Vapour and gas permeation through HLGO membranes. (a) Weight loss rate for a container sealed with HLGO membranes with different thicknesses (aperture diameter ≈ 0.5 cm). Weight loss for IPA and water were tested at room-temperature and zero humidity. (b) Thickness dependence of helium permeance through HLGO membrane. Dotted lines are the best fits to the exponential decrease. Inset: Schematics of our experimental setup for helium permeation measurement. For the helium (He) gas permeation experiments, HLGO membranes attached to the Cu foil were placed between two rubber O-rings in a custom made permeation cell and pressurised from one side up to 100 mBar. He gas permeation through the HLGO membrane was monitored on the opposite (vacuum) side by using mass spectrometry (supplementary Fig. 8b inset). We used Hiden quadrupole residual gas analyser for measuring the partial pressure of He gas in the vacuum side. A standard calibrated leak (Open style CalMaster Leak Standard, LACO technologies) is utilised to convert the partial pressure to the leak rate16. 21 Supplementary Fig. 8b shows the He permeance through HLGO membrane as a function of membrane thickness. Similar to the organic solvent and vapour permeation (Fig. 3b and supplementary Fig. 8a), He gas also follows exponential decay indicating the pathway for the gas permeation is dominated by the pinholes. The observed exponential decay of He permeance with increasing thickness is consistent with the earlier study on He and H2 permeance through ultrathin GO membranes17, but the mechanism of exponential dependence was not elucidated. The proposed mechanism in this study (main text) clarifies this ambiguity. 8. Pinholes in ultrathin HLGO membranes Supplementary Fig. 9 Pinholes in GO membrane. (a) Schematic showing continuous interconnected GO plane formed by the random overlap of GO flakes. (b) SEM image from one of our HLGO membrane with a thickness of ≈ 3 nm transferred to ITO coated glass slide showing the presence of pinholes (large pinholes are circled) in the membrane. Scale bar, 20 µm. The membrane was transferred to ITO substrate by floating the alumina supported GO membrane in water and subsequently fishing out the GO membrane onto an ITO substrate. ITO substrate was used to avoid the charging effect during SEM imaging. During the self-assembly of GO membrane, the flakes randomly overlap and provide a continuous interconnected plane that contains a large number of holes (Supplementary Fig. 9). These holes between different flakes are referred as pinholes. Our SEM analysis shows that for ≈ 3 nm membranes the size of these pinholes is of the order of the flake size. With increasing numbers of layers of GO, the newly added layers block these pinholes and form fully covered GO membranes. Our sieving experiments (Fig. 1) confirm that the minimum thickness required for the fully continuous GO membrane is ~ 8 nm. 9. Influence of solvent parameters on their permeance To probe other possible mechanisms (e.g. Solution-diffusion model18) for the faster water transport through thick GO membranes, we have checked the correlation between different solvent parameters such as relative polarity, kinetic diameter and total Hansen solubility 22 parameter19-26 on the permeance through the HLGO membranes. Supplementary Table 1 shows the different parameters for the solvents used in our experiments. To understand the influence of the above parameters on the permeance we have plotted the product of permeance and viscosity as a function of the solvent parameters. As an example, supplementary Fig. 10 shows Hansen solubility parameter vs. product of viscosity and permeance for 8 nm and 70 nm thick HLGO membranes. Despite a small variation (within the grey area in supplementary Fig. 10), it is clear from the figure that the permeance behaviour for 8 nm HLGO membrane is very close to what can be expected for pore flow model27, i.e. product of viscosity and permeance is a constant, independent of the solvents used. On the other hand for 70 nm thick membranes, except water, for all other solvents, the product of permeance and viscosity is nearly a constant. However, water deviates from this trend and permeates faster. A similar behaviour can also be obtained by plotting product of viscosity and permeance as a function other solvent parameters. The unique fast permeation of water through GO membranes is attributed to the enhanced flow of water through graphene capillaries in the GO membranes15. By increasing the GO membrane thickness further, we only obtained water flux and all the organic solvent flux were below our detection limit. The absences of correlation between solvent parameters and permeance of HLGO membranes, especially for the thicker membranes (> 50 nm), further support the validity of pore flow model for mass transport in GO membranes and rules out other mechanisms such as solution-diffusion model18. Supplementary Table 1 Solvent parameters. Viscosity, relative polarity, kinetic diameter, and total Hansen solubility parameter of the solvents used in this study. Solvents Viscosity19,20 Relative polarity21,22 Kinetic diameter23 Total Hansen solubility parameter24,25,26 hexane acetone actonitrile tetrahydrofunan methanol butyl acetate water ethanol iso-proponal butanol nm 0.51 0.47 0.34 0.48 0.38 - 0.265 0.44 0.47 0.5 MPa1/2 14.9 20.1 24.4 19.4 29.7 16.8 47.8 26.6 24.6 23.1 mPa.s 0.3 0.306 0.369 0.456 0.544 0.685 0.89 1.074 2.038 2.544 0.009 0.355 0.46 0.207 0.762 - 1 0.654 0.546 0.586 23 Supplementary Fig. 10 Probing influence of solubility parameter of solvents on permeation. Product of permeance and viscosity of solvents as a function of the total Hansen solubility parameter for 8 nm and 70 nm thick HLGO membrane (colour coded). Variations between the points in the grey colour marked region are minor and within the accuracy of measurements. All the error bars are standard deviations using at least three different measurements using different samples. 10. Mg2+-crosslinked partially reduced GO membrane for OSN Supplementary Fig. 11 Mg2+ crosslinked GO membranes. (a) X-ray diffraction for pristine GO, Mg2+ crosslinked GO (GO-Mg2+) and partially reduced Mg2+ crosslinked GO (rGO-Mg2+) membranes. The thickness of membranes ≈ 200 nm. (b) Schematic showing the structure of the GO-Mg2+ membrane. The dotted line indicates the permeation pathway and blue circles indicate Mg2+ ions. Multivalent cations have previously been used to crosslink the GO sheets by attaching them to the oxidised regions to improve the mechanical strength and to control the ion permeation through the GO membranes14,28,29. Here, we propose the same crosslinking technique to enhance the solvent permeance through the GO membranes because the interlayer cations 24 could act as randomly distributed external spacers to introduce disorder in the laminar structure (supplementary Fig. 11) and hence increase the permeance. We chose Mg2+ for crosslinking due to its large hydrated diameter13, which is comparable to the interlayer spacing in GO membrane. GO crosslinking with Mg2+ was carried out by the drop-by-drop addition of 10 mL of 9.5 g/L MgCl2 into 40 mL GO suspension (0.2 wt. %) under vigorous magnetic stirring followed by at least one day of sonication. After the sonication, the suspensions were stable up to one hour (average flake size ≈ 200 nm) without any stirring, but it starts agglomerating after that. This could be due to the neutralisation of the negative surface charges of GO with the cations. To avoid the agglomeration we stored the suspension under vigorous stirring. Mg2+ crosslinked GO membranes (GO-Mg2+) were then prepared by the vacuum filtration of these suspensions through an Anodisc alumina membrane (200 nm pore size). The incorporation of Mg2+ in the GO membranes was confirmed by XRD analysis, where a broader GO peak was found (supplementary Fig. 11a). An increase of FWHM from 1.6 degree to 2.1 degree indicates a poor interlayer alignment in GO-Mg2+ (supplementary Fig. 11b) compared to pristine GO and suggests the prospect of obtaining higher permeance. The organic solvents permeance and organic solvent nanofiltration (OSN) through GO-Mg2+ membranes (200 nm thick) were measured by vacuum filtration technique as detailed in the main text. Supplementary Fig. 12 shows the pure solvent permeance and dye rejection properties of GO- Mg2+ membranes. Comparing to the performance of the CGO membranes, even though GO- Mg2+ membranes are thicker, they show nearly one order of magnitude higher permeance to methanol but with same dye rejection (84% MB rejection for 35 nm CGO and 200 nm GO- Mg2+ membrane) (supplementary Fig. 12b and Fig. 2c inset). The enhanced permeance through GO-Mg2+ membranes suggests that the addition of Mg2+ increases the disorder in the laminar structure as shown in supplementary Fig. 11b. To further improve the dye rejection performance of the GO-Mg2+ membranes, we partially reduced them in hydroiodic acid vapour for 1 min at room temperature. The partially reduced GO-Mg2+ membranes (rGO-Mg2+ membranes) show a broad XRD peak at ≈ 23.7o (supplementary Fig. 11a), suggesting the collapse of the interlayer channels. However, in comparison to the fully reduced GO membranes30 (peak at ≈ 25o with FWHM of 1.7 degree), where it blocks the permeation of all gases and solvents, the larger FWHM of 3.3 degree confirms larger disorder in the laminar structure which could allow the molecular permeation. Our filtration experiments further support this. After the partial reduction, even though the permeance of all the solvents decreased by a factor of ≈ 3.5 (supplementary Fig. 12a), it is still 30-50% higher than even the permeance for an 8 nm thick HLGO membranes. Besides, the rGO-Mg2+ membranes exhibited 90-99% rejections to the organic dye molecules with molecular weights ranging from 249 g/mol to 1017 g/mol (supplementary Fig. 12b). We 25 explain the relatively lower permeance and high rejection of dye molecules for rGO-Mg2+ membranes compared to GO-Mg2+ membranes by the close packing30 of the interlayer after the reduction, which could make the disordered interlayer channels narrower. Even though further improvement in membrane performance could be achieved with better optimisations in the membrane crosslinking process, our findings show the potential of crosslinked GO membrane for organic solvent nanofiltration applications. Supplementary Fig. 12 Permeation through 200 nm thick Mg2+-crosslinked GO membranes. (a) Permeance of various organic solvents through GO-Mg2+ and rGO-Mg2+ membrane as a function of their inverse viscosity. The used solvents are numbered and named on the top left. Dotted lines are the best linear fit. (b) Rejection of several dyes in methanol versus their molecular weight. The dyes used: CG, MB, CV, and RB. Inset: The corresponding permeance of methanol. Supporting References 1. Bagri, A. et al. Structural evolution during the reduction of chemically derived graphene oxide. Nat. Chem. 2, 581-587 (2010). 2. Ganguly, A., Sharma, S., Papakonstantinou, P. & Hamilton, J. Probing the Thermal In Situ X-ray-Based Deoxygenation of Graphene Oxide Using High-Resolution Spectroscopies. J. Phys. Chem. C 115, 17009-17019 (2011). 3. Cai, C., Sang, N., Shen, Z. & Zhao, X. Facile and size-controllable preparation of graphene oxide nanosheets using high shear method and ultrasonic method. J. Exp. Nanosci., http://dx.doi.org/10.1080/17458080.2017.1303853 (2017). 4. Gonçalves, G. et al. Breakdown into nanoscale of graphene oxide: Confined hot spot atomic reduction and fragmentation. Sci. Rep. 4, 6735 (2014). 26 5. Jimenez-Solomon, M. F., Song, Q., Jelfs, K. E., Munoz-Ibanez, M. & Livingston, A. G. Polymer nanofilms with enhanced microporosity by interfacial polymerization. Nat. Mater. 15, 760-767 (2016). 6. Lim, S. K., Goh, K., Bae, T.-H. & Wang, R. Polymer-Based Membranes for Solvent- Eng. Chem. Chin. Resistant Review. https://doi.org/10.1016/j.cjche.2017.05.009 (2017). Nanofiltration: A J. 7. Roy, S., Ntim, S. A., Mitra, S. & Sirkar, K. K. Facile fabrication of superior nanofiltration membranes from interfacially polymerized CNT-polymer composites. J. Membr. Sci. 375, 81- 87 (2011). 8. Kosaraju, P. B. & Sirkar, K. K. Interfacially polymerized thin film composite membranes on microporous polypropylene supports for solvent-resistant nanofiltration. J. Membr. Sci. 321, 155-161 (2008). 9. Xu, Y. C., Tang, Y. P., Liu, L. F., Guo, Z. H. & Shao, L. Nanocomposite organic solvent nanofiltration membranes by a highly-efficient mussel-inspired co-deposition strategy. J. Membr. Sci. 526, 32-42 (2017). 10. Marchetti, P., Jimenez Solomon, M. F., Szekely, G. & Livingston, A. G. Molecular Separation with Organic Solvent Nanofiltration: A Critical Review. Chem. Rev. 114, 10735- 10806 (2014). 11. Huang, Y. et al. Ultrafiltration membranes with structure-optimized graphene-oxide coatings for antifouling oil/water separation. Adv. Mater. Interf. 2, 1400433 (2015). 12. Kim, S. et al. Room-temperature metastability of multilayer graphene oxide films. Nat. Mater. 11, 544-549 (2012). 13. Joshi, R. K. et al. Precise and ultrafast molecular sieving through graphene oxide membranes. Science 343, 752-754 (2014). 14. Yeh, C.-N., Raidongia, K., Shao, J., Yang, Q.-H., Huang, J. On the origin of the stability of graphene oxide membranes in water. Nat. Chem. 7, 166-170 (2015). 15. Nair, R. R., Wu, H. A., Jayaram, P. N., Grigorieva, I. V., Geim, A. K. Unimpeded permeation of water through helium-leak-tight graphene-based membranes. Science 335, 442-444 (2012). 16. Firpo, G., Angeli, E., Repetto, L., Valbusa, U. Permeability thickness dependence of polydimethylsiloxane (PDMS) membranes. J. Membr. Sci. 481, 1-8 (2015). 17. Li, H. et al. Ultrathin, Molecular-sieving graphene oxide membranes for selective hydrogen separation. Science 342, 95-98 (2013). 18. Wijmans, J. G. & Baker R. W. The solution-diffusion model: a review. J. Membr. Sci. 107, 1- 21 (1995). 19. Buekenhoudt, A. et al. Unravelling the solvent flux behaviour of ceramic nanofiltration and ultrafiltration membranes. J. Membr. Sci. 439, 36-47 (2013). 20. Karan, S., Jiang, Z. & Livingston, A. G. Sub–10 nm polyamide nanofilms with ultrafast solvent transport for molecular separation. Science 348, 1347-1351 (2015). 27 21. Liu, R. et al. Graphene oxide membrane for liquid phase organic molecular separation. Carbon 77, 933-938 (2014). 22. Linstrom, P. J. & Mallard, W. G. The NIST Chemistry WebBook: A Chemical Data Resource on the Internet. J. Chem. Eng. Data 46, 1059-1063 (2001). 23. Wu, H., Gong, Q., Olson, D. H. & Li, J. Commensurate Adsorption of Hydrocarbons and Alcohols in Microporous Metal Organic Frameworks. Chem. Rev. 112, 836-868 (2012). 24. Hansen, C. M. Hansen solubility parameters: a user's handbook. (CRC press, 2007). 25. Konios, D., Stylianakis, M. M., Stratakis, E. & Kymakis, E. Dispersion behaviour of graphene oxide and reduced graphene oxide. J. Colloid Interface Sci. 430, 108-112 (2014). 26. Zeng, W., Du, Y., Xue, Y. & Frisch, H. L. Physical properties of polymers handbook 289-303 (Springer New York, 2007). 27. Mulder, M., Basic principle of membrane technology (Kluwer academic publishers, 1996). 28. Park, S. et al. Graphene oxide papers modified by divalent ions-enhancing mechanical properties via chemical cross-linking. ACS Nano 2, 572-578 (2008). 29. Chen, L. et al. Highly efficient ion rejection by graphene oxide membranes via ion- controlling interlayer spacing. arXiv:1610.06369 (2016). 30. Su, Y. et al. Impermeable barrier films and protective coatings based on reduced graphene oxide. Nat. Commun. 5, 4843 (2014). 28
1811.10093	1	1811	2018-11-25T20:50:15	Experimental parameter uncertainty in PEM fuel cell modeling. Part II: Sensitivity analysis and importance ranking	[ "physics.app-ph" ]	Numerical modeling of proton exchange membrane fuel cells is at the verge of becoming predictive. A crucial requisite for this, though, is that material properties of the membrane-electrode assembly and their functional dependence on the conditions of operation are known with high precision. In this bipartite paper series we determine the most critical transport parameters for which accurate experimental characterization is required in order to enable the simulation of fuel cell operation with sufficient confidence from small to large current densities. In Part II, we employ the two-phase model developed in Part I to carry out extensive forward uncertainty propagation analyses. These include the study of local parameter sensitivity in the vicinity of a baseline parameter set, and a global sensitivity analysis in which a broad range of operating conditions and material properties is covered. A comprehensive ranking list of model parameters is presented, sorted by impact on predicted fuel cell properties such as the current-voltage characteristics and water balance. The top five in this list are, in this order: The membrane hydration isotherm, the electro-osmotic drag coefficient, the membrane thickness, the water diffusivity in the ionomer and its ionic conductivity.	physics.app-ph	physics	Experimental parameter uncertainty in PEM fuel cell modeling Part II: Sensitivity analysis and importance ranking Roman Vetter and Jürgen O. Schumacher Institute of Computational Physics (ICP), Zurich University of Applied Sciences (ZHAW), Wildbachstrasse 21, CH-8401 Winterthur, Switzerland November 27, 2018 8 1 0 2 v o N 5 2 ] h p - p p a . s c i s y h p [ 1 v 3 9 0 0 1 . 1 1 8 1 : v i X r a Numerical modeling of proton exchange membrane fuel cells is at the verge of becoming predictive. A crucial requisite for this, though, is that material properties of the membrane-electrode assembly and their functional dependence on the conditions of operation are known with high precision. In this bipartite paper series we determine the most critical transport parameters for which accurate experimental characterization is required in order to enable the simulation of fuel cell operation with sufficient confidence from small to large current densities. In Part II, we employ the two-phase model developed in Part I to carry out extensive forward uncertainty propagation analyses. These include the study of local parameter sensitivity in the vicinity of a baseline parameter set, and a global sensitivity analysis in which a broad range of operating conditions and material properties is covered. A comprehensive ranking list of model parameters is presented, sorted by impact on predicted fuel cell properties such as the current-voltage characteristics and water balance. The top five in this list are, in this order: The membrane hydration isotherm, the electro-osmotic drag coefficient, the membrane thickness, the water diffusivity in the ionomer and its ionic conductivity. 1 Introduction Modeling the various transport phenomena in proton exchange membrane fuel cells (PEMFCs) with high accuracy can be a challenging task [1], in particular at high current density. In spite of the well-known parametric degeneracy of fuel cell polar- ization curves, it is still common practice to validate PEMFC models with small sets of experimental performance data that cannot accommodate the full complexity of a fuel cell's nonlin- ear response to changes in operating conditions and material properties. It is all the more important that the uncertainties associated with each model parameter are known. This not only assists modelers with appreciating the limitations of their models, but also provides insight for fuel cell designers and engineers who seek to understand which material properties to tackle in order to increase cell performance or to reduce manufacturing costs. Although a large number of numerical PEMFC models have been developed over the past decades, only few publications offer parameter sensitivity analyses [2]. Kimble & White [3] presented an early parameter sensitivity study of a five-layer al- kaline fuel cell model, focusing mainly on the effect of electrode thickness, porosity and interfacial surface area on the limiting current density. Reports of similar efforts for acidic fuel cell models seem to have appeared much more recently only. Guo et al. [4] fitted five mass and charge transport parameters si- multaneously to experimental polarization data, confirming the well-known significance of ionic conductivity and gas transport on PEMFC performance. Grujicic & Chittajallu [5] reported on a flow channel design optimization with a coarsely meshed sen- sitivity analysis for six cathode parameters using a single-phase 2D model. Carnes & Djilali [6] estimated the local sensitivity of a simple 1D PEMFC model with three differential equations to changes in membrane conductivity, exchange current densities and oxygen diffusivities with a least-squares fit to experimental polarization curves. Min et al. [7] carried out a local sensitiv- ity study with eleven varying parameters on a 3D two-phase PEMFC model and found that the cathode kinetic properties, the membrane conductivity and oxygen transport capability of the cathode GDL and are among the most significant ones. They also reemphasized the need to include more information than just the polarization curve in model validation. Mawardi & Pitchumani [8] used latin hypercube sampling to examine the parameter uncertainty of a 1D single-phase PEMFC model. Multivariate uncertainty, in our parlance adopted here, refers to the presence of concurrent uncertainties in more than one input parameter. Their study was, however, limited to uncertainty in the kinetic transfer coefficients and the employed operating conditions rather than material parameterizations. Moreover, the variance of power density was the only criterion for un- certainty in the predicted fuel cell model response. Zhao et al. [9] performed a simple local variation of various parameters of a 1D PEMFC stack model. Laoun et al. [10] were, to the best of our knowledge, the first to report on a global multi- variate sensitivity analysis of a PEMFC model. They drew quasi-random samples over a relatively broad range of oper- ating conditions (gas pressure, temperature, current density) and some geometrical properties of the membrane-electrode assembly (MEA) sandwich (layer thicknesses, GDL porosity, cell area) to quantify the uncertainty of the predicted power density with Sobol indices. The research group found that their 0D model was most sensitive to variations in current density and membrane thickness. Several local parameter sensitivity analyses were also carried out on strongly simplified, lumped or zero-dimensional PEMFC models [11 -- 15] to examine the effect of uncertainty in one up to a few variables such as temperature or GDL porosity on fuel cell polarization. Of these, only the works by Corrêa et al. conclude with a list of model parameters sorted by sensitivity. Recently, Shah [16] proposed to use prin- cipal component analysis to reduce the computational demand associated with uncertainty quantification in PEMFC modeling with Monte Carlo sampling. While the above-cited publications show a moderately in- creasing level of sophistication in PEMFC model uncertainty 1 quantification, they are far from complete in the sense that: • Only a very limited set of varying input parameters are considered, out of the dozens a high-fidelity PEM fuel cell model depends on. • Almost all studies examine only a single model output (typi- cally the power density at a fixed operating point). • Parameter uncertainty is almost exclusively determined only locally, i.e., at a fixed point of operation or model parame- terization, rather than globally, i.e., for a whole spectrum of conditions. • Most reported uncertainty analyses are based on computa- tional models with strongly reduced complexity, and thus with insufficient account for the complex nonlinear multifunc- tional behavior of MEA materials. Therefore, a thorough global parameter uncertainty quantifica- tion with a detailed PEMFC model with spatial resolution is appropriate. In the first part of this bipartite study [17], we demonstrated how considerable epistemic uncertainty associated with measured material properties can induce a large spread in the predicted fuel cell performance. Being an extension of our one-dimensional two-phase MEA model [18], the model devel- oped in Part I offers sufficient efficiency for a large number of numerical evaluations in manageable computation time without neglecting through-plane gradients, as lumped models do. In the present article, we employ our model to carry out the most extensive parameter uncertainty propagation analysis for PEM fuel cells reported in the scientific literature so far. Unlike exist- ing studies, we include a dozen predicted fuel cell state variables in the uncertainty analysis (including performance and water balance characteristics), rather than considering a single point on the polarization curve as the only model response. In Sec. 2, a mathematical formalism is established, suitable for the quan- tification of local error propagation through a strongly nonlinear function such as the present MEA model. To establish some intuition for the model response to input uncertainty, a local sensitivity analysis at typical fuel cell operating conditions is presented in Sec. 3. The analysis is then extended to the global scope by systematic quasi-stochastic variation of all relevant model parameters over a wide range of physically meaningful values in Sec. 4. This allows us to deduce a robust sorted list of input parameters, ranking their absolute and relative signif- icance for predictive state-of-the-art PEMFC modeling. We refer to [17] for all modeling details and adopt the nomenclature introduced therein. 2 Quantification of uncertainty propagation The most appropriate measure for the (local) propagation of uncertainty through a mathematical model or function is its condition number -- a concept that is routinely applied in linear and nonlinear algebra as well as numerical mathematics. Given a model function (cid:126)f : Rn → Rm, the local relative condition number of the i-th output value fi with respect to the j-th input variable xj, at a given set of model parameters (cid:126)x with fi((cid:126)x) (cid:54)= 0, is defined as [19] (cid:12)(cid:12)(cid:12)(cid:12) = (cid:12)(cid:12)(cid:12)(cid:12) xj fi((cid:126)x) (cid:12)(cid:12)(cid:12)(cid:12) ∂fi ∂xj ((cid:126)x) κij((cid:126)x) = ((cid:126)x) (1) (2) (cid:12)(cid:12)(cid:12)(cid:12), if fi is differentiable at (cid:126)x. More generally, for a non-differentiable function, κij((cid:126)x) = lim δ→0 sup ∆xj≤δ fi((cid:126)x + ∆xj(cid:126)ej) − fi((cid:126)x) ∆xj fi((cid:126)x) where (cid:126)ej denotes the j-th standard unit vector and ∆xj a small perturbation or error in the j-th model parameter. The condition number measures the propagation of uncer- tainty through a model function f. A relative error in the 2 ∂ log xj (cid:12)(cid:12)(cid:12)(cid:12) ∂ log fi (cid:12)(cid:12)(cid:12)(cid:12) xj j-th input variable is magnified by the factor κij in the output value fi (assuming no uncertainty in all others). A quantity of interest fi predicted by a model is sensitive to uncertainty in its parameter xj if κij is large, and conversely, insensitive if κij is small. When fi responds (locally) linearly to relative changes in xj, one has κij = 1. In the following, by computing the relative condition number explicitly for a large range of operating conditions and material properties (cid:126)x and for numer- ous fuel cell state characteristics fi, we show in which material parameterizations the largest source of uncertainty resides in macro-homogeneous two-phase fuel cell modeling. In practice, Eq. 2 needs to be approximated for numerical models. The most straightforward way to do this is by fixing δ to a very small constant value and setting ∆xj = δxj, such that (cid:12)(cid:12)(cid:12)(cid:12) fi((cid:126)x + δxj(cid:126)ej) fi((cid:126)x) 1 δ (cid:12)(cid:12)(cid:12)(cid:12) − 1 κij((cid:126)x) ≈ (3) We use δ = 10−3 in all subsequent calculations. First, appropriate model output variables fi need to be se- lected. In order to cover a wide spectrum of potential properties of interest for typical PEMFC operation, we choose twelve key figures of which the former six are performance indicators de- rived from the predicted polarization curve, whereas the latter six characterize the state of heat and water balance within the fuel cell. The selected performance indicators are the limiting current density Imax, the peak power density Pmax, the current densities I0.8, I0.6 and I0.4 at applied cell voltages of 0.8 V, 0.6 V and 0.4 V, respectively, and the cell voltage U1 at 1 A cm−2. All these key properties are evaluated subject to the condition that they are attained under the simulated operating conditions and material parameterizations. Furthermore, at a fixed cell voltage of 0.6 V, we select the peak temperature Tmax across the MEA sandwich, the minimum local water content of the ionomer (including catalyst layers) λmin, the mean water content of the catalyst-coated membrane (CCM) [18] λavg = iλ dx i dx, CCM CCM the ohmic resistance of the membrane (cid:90) (cid:30)(cid:90) (cid:90) Rp = dx σp , PEM (4) (5) the maximum local liquid water saturation smax, and the net water flux across the membrane jλ. Next, a set of model input factors xj are chosen, with respect to which the modeling uncertainty is to be evaluated. They can be classified into two primary categories: the conditions at which the fuel cell is operated, and the MEA properties includ- ing electrochemical, physical and geometrical parameters. In the first category we have the anode and cathode gas feed pressures PA and PC, the relative humidities RHA and RHC in the gas channels, the boundary temperatures TA and TC (assuming that the temperature of the supplied gas equals that of the bipolar plate on either side of the MEA) and the mole fraction of oxygen in the oxidation gas mixture, which equals 21% for operation with ambient air. The parameter set in the second category consists of the following coefficients: The electric conductivity of the catalyst layers (CLs) and gas diffusion layers (GDLs) σe, the protonic conductivity of the ionomer σp, the thermal conductivity k, the effective Fickean diffusivity of water in the ionomer Dλ, the electro-osmotic drag (EOD) coefficient ξ, the microstructure factor of the pores in the CLs and GDLs Mp that reduces the effective gas diffusivities, the liquid water transport coefficient Ds (incorporating the absolute and relative hydraulic permeabilities but also the capillary pressure -- saturation rela- tionship), the anode and cathode exchange current densities jA and jC, the condensation and evaporation rates γc and γe, Table 1: Local model sensitivity to parameter uncertainty for the baseline model parameterization at the reference operating conditions listed in Tab. 2. High sensitivities are highlighted with shaded cells, from white (log10 κij ≤ −1) to dark gray (log10 κij ≥ 0). log10 κij xj fi PA = PC RHA RHC TA = TC αO2 σe σp k Dλ ξ Mp Ds jA jC γc γe ka kd Re RT λv CL i LGDL LCL LPEM 0 Pcl median 0 0 Imax -2.81 -0.23 -1.63 0.07 -1.47 -2.93 -1.95 -1.59 -0.15 0.25 -1.33 -2.09 -2.50 -2.34 -3.30 -2.12 -0.79 -1.66 -2.33 -1.25 -0.24 -0.69 -1.16 -0.73 -0.29 -1.06 -1.40 Pmax -0.90 -0.37 -1.00 -0.24 -0.84 -1.71 -0.70 -1.96 -0.49 -0.06 -0.92 -3.97 -1.87 -1.22 -3.50 -2.51 -1.04 -1.76 -0.80 -1.67 -0.37 -0.78 -0.83 -1.04 -0.45 -0.81 -0.91 I0.8 0.03 -0.93 -0.42 -0.09 -0.31 -1.57 -0.95 -2.70 -1.89 -1.43 -1.31 -1.84 -2.10 -0.25 -4.27 -3.79 -2.99 -2.76 -0.65 -2.67 -0.90 -1.29 -1.07 -0.29 -1.13 -1.16 -1.23 I0.6 -0.71 -0.41 -0.93 -0.43 -0.77 -1.58 -0.69 -2.07 -0.61 -0.17 -0.93 -2.98 -1.87 -1.01 -3.54 -2.68 -1.16 -1.81 -0.67 -1.78 -0.38 -0.82 -0.82 -1.39 -0.53 -0.72 -0.87 I0.4 -1.36 -0.24 -1.13 -0.05 -1.05 -2.05 -0.90 -1.67 -0.31 0.11 -1.01 -2.46 -1.99 -1.60 -3.39 -2.32 -0.86 -1.70 -1.16 -1.43 -0.29 -0.75 -0.90 -0.85 -0.37 -0.97 -1.03 U1 -0.99 -0.76 -1.24 -0.82 -1.07 -1.88 -1.02 -2.42 -0.97 -0.53 -1.26 -3.19 -2.19 -1.29 -3.85 -3.04 -1.52 -2.15 -0.97 -2.14 -0.72 -1.16 -1.14 -1.90 -0.88 -1.03 -1.20 Tmax -2.00 -1.62 -1.56 -0.01 -2.20 -2.99 -2.17 -1.87 -2.21 -1.81 -2.71 -3.03 -3.43 -2.48 -4.44 -3.26 -2.44 -3.38 -2.00 -1.79 -1.78 -2.33 -1.97 -3.17 -2.16 -2.20 -2.20 λmin -0.52 0.01 -1.38 -0.02 -0.72 -1.52 -0.62 -1.40 -0.32 0.12 -1.30 -1.90 -1.88 -0.95 -3.66 -2.57 -0.84 -1.87 -0.60 -1.26 -0.20 -1.10 -1.35 -0.70 -0.82 -0.95 -0.95 λavg -0.74 -0.21 -1.32 -0.29 -0.94 -1.74 -0.80 -1.62 -0.50 -0.05 -1.46 -1.86 -1.90 -1.16 -3.23 -2.46 -1.00 -1.49 -0.81 -1.35 -0.13 -0.87 -1.47 -0.77 -1.20 -1.32 -1.18 Rp -0.35 0.16 -0.85 0.42 -0.54 -1.33 -0.21 -1.42 -0.07 0.35 -1.04 -1.59 -1.43 -0.76 -3.02 -2.25 -0.64 -1.32 -0.42 -1.16 0.14 -0.57 -0.95 -0.33 0.07 -0.83 -0.70 smax -1.44 -0.66 -0.74 -0.86 -1.83 -2.69 -1.76 -1.69 -2.10 -2.36 -2.05 -0.90 jλ -0.82 0.56 -0.49 0.16 -1.38 -2.23 -1.01 -0.90 -0.65 -0.01 -0.64 -2.55 -- -1.70 -1.54 -3.34 -2.35 -0.31 -1.91 -1.44 -0.73 0.12 -1.29 -0.41 -1.34 -0.73 -0.76 -0.86 -2.02 -3.22 -2.03 -1.53 -2.84 -1.85 -1.39 -1.03 -1.89 -1.05 -2.19 -1.87 -1.05 -1.84 med. -0.86 -0.30 -1.07 -0.07 -0.99 -1.81 -0.92 -1.68 -0.55 -0.06 -1.28 -2.28 -1.94 -1.26 -3.44 -2.48 -1.02 -1.84 -0.89 -1.41 -0.33 -0.98 -1.06 -0.95 -0.77 -1.00 -1.01 Table 2: Reference operating conditions (from [17]). Value Symbol Explanation PA PC RHA RHC TA TC αO2 Gas pressure in anode gas channel Gas pressure in cathode gas channel Relative humidity in anode gas channel Relative humidity in cathode gas channel Temperature of anode plate and gas channel Temperature of cathode plate and gas channel Oxygen mole fraction in dry oxidant gas 1.5 bar 1.5 bar 100% 100% 80 ◦C 80 ◦C 21% the vapor absorption and desorption rates at the ionomer -- gas interface ka and kd, the electrical and thermal contact resistivi- ties Re and RT , the hydration number of a vapor-equilibrated membrane λv, the volume fraction of ionomer in the catalyst layer CL and i , and finally, the applied clamping pressure Pcl, which LPEM affects several effective MEA parameters simultaneously. , the uncompressed layer thicknesses LGDL 0 , LCL 0 0 3 Local sensitivity analysis We start off with a local sensitivity analysis to showcase the model behavior using the best material parameterization found in Part I [17], at operating conditions typical for automotive applications. Many of the material parameters are based on different con- stitutive parameterizations in the individual MEA layers, as detailed in Part I. The aim of the present study is to compare the impact of the different kinds of model parameters rather than the impact of individual layers. We therefore vary each parameter xj in all layers simultaneously where applicable. As an example, if the j-th model parameter is the thermal con- ductivity k, we perturb k in each MEA layer by the factor δ for the evaluation of κij using Eq. 3. Furthermore, we set PA = PC and TA = TC to reduce the number of independent operating conditions from seven to five, which simplifies the 3 data presentation. Since the condition number is a logarithmically scaling quan- tity, we report all values in logarithmic form. A fuel cell char- acteristic fi is highly sensitive to uncertainty in parameter xj if log10 κij is positive or nearly so. Tab. 1 lists the simulation results for all input/output pairs at the reference operating conditions given in Tab. 2. Note that the maximum liquid water saturation on the cathode side is insensitive to changes in the anode reaction rate (i.e., the condition number is indis- tinguishable from zero within the numerical error tolerances used here), which is why this value is omitted in Tab. 1. This data provides first insight into the model behavior at reference operating conditions. The peak temperature and the maximum liquid water saturation are the least sensitive of the twelve selected traits quantifying the operative state of the fuel cell. Nevertheless, with the membrane resistance and the net water flux, two quantities from the second category (not derived from the polarization curve) are the most sensitive overall, which accentuates that a comprehensive uncertainty analysis should also include model outputs other than points on the polarization curve. The relative humidity in the anode gas channel and the bound- ary temperature are the most influential operating conditions at reference conditions, which can be explained by the direct impact on the ionic conductivity on the (typically drier) anode side of the membrane they have [17, 18]. In the category of MEA parameters, the water diffusivity, the EOD coefficient and the equilibrium water uptake of the ionomer show the greatest impact on many of the fuel cell properties predicted by the model. How sensitive these findings are with respect to the choice of operating conditions is demonstrated in Fig. 1, where in each subplot one of them is systematically varied. All model parame- ters are fixed but the one operating factor with respect to which the condition number is evaluated. As can be recognized from Fig. 1a, changes in gas pressure mostly affect the polarization Figure 1: Model sensitivity to uncertainty in the operating conditions for the baseline model parameterization. All operating conditions are fixed according to Tab. 2 but the one shown on the respective horizontal axes. Thin solid vertical lines indicate the reference conditions. Local fluctuations are numerical noise. at low current densities (I0.8), whereas the maximum tempera- ture and liquid water saturations are generally very insensitive. An important feature of the relative condition number is that its logarithm is unbounded toward both negative and positive infinity. κij vanishes if xj = 0 or ∂fi/∂xj = 0. Since we varied the gas pressure from 1 to 10 bar, only the second condition can be fulfilled in Fig. 1a. Indeed, the two pressure points at which κij → 0, the current densities I0.4 and Imax are invariants on the polarization curve. For instance, if the gas pressure is varied about the value of 2.6 bar, the resulting polarization curves cross at 0.4 V, because that is where log10 κij = −∞. The reverse scenario occurs with the net water flux through the membrane jλ in Fig. 1b. A high relative humidity is required in the hydrogen gas feed to obtain a flux from anode to cathode. Slightly below RHA = 60% (for fixed RHC = 100%), back diffusion and EOD cancel each other, such that fi = jλ = 0, letting κij diverge to infinity. At lower RHA, the hydration gradient in the membrane is so steep that back diffusion is stronger than EOD at 0.6 V. This reversal of water balance also affects other fuel cell properties, which show a kink at that relative humidity value, such as Tmax (through the latent heat of phase change and sorption) or the ohmic membrane resistance Rp (through the strong hydration-dependence of the ionic conductivity). There is also an invariant point on the polarization curve at 0.8 V slightly above RHA = 90% (Fig. 1b), in the vicinity of which ∂I0.8/∂RHA ≈ 0. As RHA → 0, also the relative condition numbers vanish. The model sensitivity to uncertainty in the RH on the anode side generally increases toward higher humidity. A similar observation is made in Fig. 1c for RHC until it reaches about 50%, above which the model abruptly looses its sensitivity to RH changes in the cathode gas channel. This is because with a more humid air feed, evaporation yields saturated gas in almost the entire CL and GDL, making further humidification of the supplied air irrelevant. Only as RHC → 100%, the sensitivity rises again to significant values, because liquid water starts to accumulate. As shown in Fig. 1d, various points on the polarization curve are invariant points with respect to changes in the boundary temperature at different temperature values, which can be ratio- nalized by appreciating how many material parameterizations in our model are temperature-dependent. Also the net water flux is found to be invariant at about 97 ◦C. As expected, the effect of the boundary temperature on Tmax is nearly linear, as can be recognized by κij ≈ 1 in Fig. 1d. Low temperatures prevent the fuel cell from operating at higher current densities, which is why the condition number for U1 diverges when the temperature is such that Imax = 1 A cm−2, i.e., at about 53 ◦C. The same effect is observed in Fig. 1e for the case of oxygen (cid:46) 11%). Not surprisingly, polarization curve starvation (αO2 features at large current densities (Imax, I0.4, I0.6, Pmax) ex- hibit κij ≈ 1 as oxygen depletes, which means that I responds linearly to changes in αO2 in this oxygen-limited regime. Next, the same analysis is repeated for the 21 major MEA parameters to obtain a quantitative picture of how the model sensitivity responds to deviations from the baseline param- eterization. For those which are parametric expressions or layer-dependent properties rather than constant coefficients, 4 12345678910104103102101100101102(a)xj=PA=PCinvariantpointsGaspressurePA=PC[bar]Relativeconditionnumberij0.00.20.40.60.81.0(b)xj=RHAnetwaterfluxreversalpointinvariantpointAnoderelativehumidityRHA0.00.20.40.60.81.0(c)xj=RHCinvariantpointsCathoderelativehumidityRHC2030405060708090100104103102101100101102(d)xj=TA=TCImax=1Acm2invariantpointsBoundarytemperatureTA=TC[C]Relativeconditionnumberij0.00.20.40.60.81.0(e)xj=↵O2Imax=1Acm2oxygenstarvationOxygenfraction↵O2Polarizationcurvecharacteristicsfi=Imax(limitingcurrentdensity)fi=Pmax(peakpowerdensity)fi=I0.8(currentdensityat0.8V)fi=I0.6(currentdensityat0.6V)fi=I0.4(currentdensityat0.4V)fi=U1(cellvoltageat1Acm2)Othercharacteristics(at0.6V)fi=Tmax(max.temperature)fi=min(min.watercontentinCCM)fi=avg(avg.watercontentinCCM)fi=Rp(ohmicresistanceofmembrane)fi=smax(max.watersaturation)fi=j(netwaterfluxthroughmembrane)Figure 2: Model sensitivity to uncertainty in the parameterization at the reference operating conditions as listed in Tab. 2. All parameterizations are fixed but the one shown on the respective horizontal axes. The graphs represent the same output variables fi as in Fig. 1. Thin solid vertical lines indicate the baseline parameterization. Local fluctuations are numerical noise. 5 0.1110104103102101100101(a)xj=pProtonicconductivityp/refpRelativeconditionnumberij0.1110(b)xj=kThermalconductivityk/kref0.1110(c)xj=DWaterdiffusivityD/Dref0.51.01.52.0(d)xj=⇠EODcoefficient⇠/⇠ref0123104103102101100101(e)xj=MpMicrostructurefactorMp/MrefpRelativeconditionnumberij0.1110(f)xj=DsLiquidwaterdiff.Ds/Drefs0.0010.010.1110(g)xj=jAAnodereactionratejA/jrefA0.010.1110100(h)xj=jCCathodereactionratejC/jrefC0.0010.010.1110104103102101100101(i)xj=cCondensationratec/refcRelativeconditionnumberij0.0010.010.1110(j)xj=eEvaporationratee/refefi=Imaxfi=Tmaxfi=Pmaxfi=minfi=I0.8fi=avgfi=I0.6fi=Rpfi=I0.4fi=smaxfi=U1fi=j0.1110(k)xj=kaAbsorptioncoeff.ka/krefa0.1110(l)xj=kdDesorptioncoeff.kd/krefd0.1110104103102101100101(m)xj=ReElectriccont.resist.Re/RrefeRelativeconditionnumberij0.1110(n)xj=RTThermalcont.resist.RT/RrefT0.51.01.52.0(o)xj=vIonomerhydrationv/refv0.00.20.40.6(p)xj=✏CLiCLionomercontent✏CLi100200300400104103102101100101(q)xj=LGDL0GDLthicknessLGDL0[µm]Relativeconditionnumberij05101520(r)xj=LCL0CLthicknessLCL0[µm]020406080(s)xj=LPEM0PEMthicknessLPEM0[µm]0123(t)xj=PclClampingpressurePcl[MPa]current density drops below 1 A cm−2, resulting in poles for U1. • Tmax and smax are often the most insensitive features, also away from the baseline parameterization. However, it is not possible to identify a single quantity fi which could serve as a robust representative for the overall model sensitivity. Evidently, it is essential to look at more than a single point in the parameter space and also at more than one output vari- able for a robust and generally valid evaluation of parameter uncertainty in a PEMFC model, because one might acciden- tally hit a spot where κij is near zero or extremely large. A global parameter sensitivity analysis is therefore indispensable. Moreover, since zeros and poles will be encountered even in a global analysis, an appropriate robust statistic should be chosen to quantitatively compare the impact of different model parameters on the model prediction. The extrema (minimum and maximum) over all i or j will be of very limited informa- tion value, as they strongly depend on the choice of fi and the screened range of admissible input values xj. We there- fore choose the median of log10 κij (over i or j), which has a breakdown point of 50%, as the decisive statistic to quantify uncertainty. The calculated medians of the local sensitivity analysis are given in Tab. 1. Fig. 3 summarizes the results of the local sensitivity analy- sis. The two input parameter categories are sorted by median over all outputs i. Alongside with the medians, the maximum and mean values over i are also given for comparative pur- poses. At reference automotive conditions with the baseline parameterization, the boundary temperature and the anode feed humidity are the most influential operating conditions. In the category of MEA properties, the ability of the ionomer to accommodate, transport and spatially redistribute dissolved water clearly dominates, followed by the membrane thickness and electrical contact resistivity. This corroborates our previous recommendation that contact resistance should not be neglected in PEMFC modeling [20]. 4 Global sensitivity analysis For the above-mentioned reasons, we seek to estimate the model sensitivity to uncertainty in all n = 26 major parameters in a global sense in the entire n-dimensional domain of physically plausible values. Based on our findings of the (semi-)local uncertainty analysis shown in Figs. 1 and 2, and considering the ranges across which experimental material data are scattered [17], the intervals listed in Tab. 3 were chosen for the global sensitivity analysis. Most intervals are rather generous so as to not miss any critical regions. They also account for the difficulty in determining accurate estimates of the exchange current densities in Pt/C-Nafion catalysts layers [21, 22] and the trend toward thinner MEA layers. In order to obtain a good estimate of global sensitivity with manageable computational effort, this n-dimensional parameter space needs to be sampled with high degree of uniformity -- higher than with Monte Carlo sampling. Mawardi & Pitchumani [8] achieved this with latin hypercube sampling (LHS) in their sensitivity analysis. Here, we use the quasi-random Sobol sequence [23], which offers superior uniformity compared to LHS and the possibility to sequentially add more samples. A Sobol sequence of length N = 2500 is generated and then rescaled according to Tab. 3. With this set of input parameters, the MEA model is evaluated N (n + 1) times to get the relative condition numbers at each sample point with respect to each parameter using Eq. 3. Each evaluation yields m = 12 output values, totaling in N (n + 1)m = 810,000 model responses that are used for the subsequent statistical analysis. Those that are not attainable under the sampled input parameters are excluded from the analysis. For instance, small ionic conductivity of the 6 Figure 3: Local model sensitivity to parameter uncertainty for the baseline model parameterization at the reference operat- ing conditions. Operating conditions (top group) and MEA parameterizations (bottom group) are separately sorted by sig- nificance (i.e., by median of log10 κij taken over all outputs i) in decreasing order. Error bars are standard errors of the statistics over i. we introduce a prefactor C to be varied, while maintaining the functional dependency of the constitutive parameterization. Consider, for instance, the ionic conductivity of the ionomer σp. We set σp = Cσref p , where σref p = Miσ0 max{fw − f0}β (6) is the baseline parameterization from Part I [17]. Thus, the prefactor C = A/Aref is reported in Fig. 2 for A = σp, k, Dλ, ξ, Mp, Ds, jA, jC, γc, γe, ka, kd, Re, RT , λv. We omit the data for the electric conductivity of the CLs and GDLs (σe), which is one of the least interesting, to save space. In short, e )−1 for all i in the we find the expected behavior κij ∼ (σe/σref e ≤ 10 with prefactors as listed in examined range 0.1 ≤ σe/σref Tab. 1, which is generally very low. We restrict the discussion of Fig. 2 to a few relevant points: • Notice that relatively broad parameter ranges are screened -- broader than the scatter of available experimental data for some of them [17]. A few of the more extreme transport coefficients yield noisy sensitivity data, e.g., large values for Dλ (Fig. 2c). • Our baseline parameterizations for the anode exchange cur- rent density and the evaporation/condensation rates are near the upper end of the reported experimental range, which is why their domains analyzed here extend further down than up. • Invariant points (zeros of κij) are scattered all over the pa- rameter domains. • A water flux reversal point is present at high back diffusivi- ties. At low back diffusivity, large EOD coefficients, low gas diffusivity, high electrical contact resistivity, poor membrane hydration, small ionomer content in the CLs, thick GDLs and membranes, and at high clamping pressures, the limiting TA=TCRHAPA=PC↵O2RHC432101⇠vDLPEM0RepLCL0✏CLiPclkaLGDL0jCMpRTkekdjADsecParametersensitivitylog10ijmaximummedianmeanTable 3: Parameter ranges for the global sensitivity analysis. For MEA parameterizations with functional dependencies on other variables, the functional dependency is fixed and only the prefactor C is varied (denoted by C = A/Aref, where A is the effective resulting value of the parameter and Aref the baseline parameterization as detailed in Part I). In these cases, the indicated range of values applies to this prefactor. Parameter Explanation s PA = PC n o i t i d n o c g n i t a r e p o Unit Range of values bar Gas feed pressure -- Relative humidity in anode gas channel -- Relative humidity in cathode gas channel ◦C Temperature of bipolar plates -- Oxygen mole fraction in dry oxidant gas -- Electric conductivity of CLs and GDLs -- Protonic conductivity of the ionomer -- Thermal conductivity -- Diffusivity of water in the ionomer Electro-osmotic drag coefficient -- Microstructure factor of the CLs and GDLs -- -- Liquid water transport coefficient -- Anode exchange current density -- Cathode exchange current density -- Condensation rate Evaporation rate -- -- Vapor absorption rate of the ionomer -- Vapor desorption rate of the ionomer -- Electrical contact resistivity -- Thermal contact resistivity -- Vapor sorption isotherm of the ionomer Volume fraction of ionomer in the CLs -- µm Uncompressed GDL thickness µm Uncompressed CL thickness µm Uncompressed membrane thickness Applied clamping pressure MPa 1 − 4 0.5 − 1 0.5 − 1 50 − 90 0.1 − 0.3 0.1 − 10 0.1 − 10 0.1 − 10 0.1 − 10 0.5 − 2 0.2 − 3 0.1 − 10 0.001 − 10 0.01 − 100 0.0001 − 10 0.0001 − 10 0.1 − 10 0.1 − 10 0.1 − 10 0.1 − 10 0.5 − 2 0.1 − 0.5 100 − 400 1 − 20 10 − 80 0.1 − 3 Scaling in range Details in Part I linear linear linear linear linear logarithmic logarithmic logarithmic logarithmic linear linear logarithmic logarithmic logarithmic logarithmic logarithmic logarithmic logarithmic logarithmic logarithmic linear linear linear linear linear linear Sec. 2.4 Sec. 2.4 Sec. 2.4 Sec. 2.4 Sec. 2.4 Sec. 3.2 Eq. (22) Eqs. (25) -- (32) Eqs. (34), (35), last row of Tab. 3 Eqs. (36), (38) Eq. (48) Eq. (15) Eq. (21) Eq. (21) last row of Tab. 7 last row of Tab. 7 4th row of Tab. 6 4th row of Tab. 6 Eq. (68), Tab. 8 Eq. (68), Tab. 8 Eqs. (44) -- (46) Eq. (23) Eqs. (61), (62) Eqs. (61), (63) Sec. 3.13 Sec. 3.13 e RHA RHC TA = TC αO2 σe/σref σp/σref p k/kref Dλ/Dref λ ξ/ξref Mp/M ref p Ds/Dref s jA/jref A jC/jref C γc/γref γe/γref ka/kref a kd/kref d Re/Rref e RT /Rref T λv/λref v CL i LGDL LCL LPEM 0 Pcl 0 0 c e s n o i t a z i r e t e m a r a p A E M Table 4: Global model sensitivity to parameter uncertainty. High sensitivities are highlighted with shaded cells, from white (log10 κij ≤ −1) to dark gray (log10 κij ≥ 0). Standard errors (calculated with the bootstrap method, not shown) are in the second to fourth significant digit. xj fi PA = PC RHA RHC TA = TC αO2 σe σp k Dλ ξ Mp Ds jA jC γc γe ka kd Re RT λv CL i LGDL LCL LPEM 0 Pcl median 0 0 Imax -0.19 -0.91 -0.98 0.02 -1.86 -2.87 -1.01 -1.75 -0.27 0.21 -0.94 -2.16 -2.04 -2.09 -3.64 -2.83 -0.77 -1.83 -2.39 -1.47 -0.10 -0.80 -0.92 -0.81 -0.29 -0.88 -0.96 Pmax -0.33 -1.07 -0.94 -0.16 -1.20 -2.10 -0.75 -2.14 -0.56 -0.06 -1.07 -2.37 -1.57 -1.25 -3.98 -3.04 -1.17 -1.96 -1.31 -1.91 -0.20 -0.89 -1.03 -1.09 -0.41 -0.88 -1.08 I0.8 -0.58 -1.00 -0.27 -0.40 -0.49 -2.13 -0.86 -3.15 -1.81 -1.62 -1.71 -3.20 -1.73 -0.46 I0.6 -0.29 -1.01 -0.78 -0.22 -1.03 -2.23 -0.80 -2.26 -0.80 -0.25 -1.19 -2.49 -1.73 -1.12 -- -3.83 -3.14 -1.34 -2.10 -1.42 -2.07 -0.20 -0.92 -1.10 -0.93 -0.47 -0.93 -1.06 -3.68 -2.20 -2.54 -1.30 -2.98 -0.51 -1.20 -1.50 -0.42 -0.86 -0.98 -1.40 log10 κij Tmax -1.97 -2.24 -2.46 -0.00 -2.82 -3.84 -2.57 -2.34 -2.65 -2.19 -2.96 -3.89 -3.38 -2.93 -- -4.26 -3.00 -3.91 -2.85 -2.19 -2.00 -2.73 -2.26 -2.71 -2.42 -2.31 -2.68 λmin -0.12 -1.22 -1.06 -0.75 -1.47 -2.51 -0.96 -2.01 -0.91 -0.35 -1.29 -2.60 -1.97 -1.50 -- -3.46 -1.36 -2.43 -1.70 -1.88 -0.19 -1.59 -1.31 -1.19 -1.19 -1.45 -1.40 λavg -0.34 -0.47 -1.12 -0.69 -1.69 -2.71 -1.20 -2.09 -0.87 -0.53 -1.19 -2.09 -2.04 -1.71 -- -2.90 -1.45 -1.66 -1.88 -1.78 -0.13 -1.22 -1.22 -1.11 -1.26 -1.24 -1.25 Rp 0.14 -0.32 -0.67 0.11 -1.02 -2.09 -0.10 -1.81 -0.30 0.23 -0.88 -1.91 -1.30 -1.02 -- -2.73 -0.88 -1.55 -1.31 -1.54 0.14 -1.05 -0.90 -0.55 0.01 -0.95 -0.92 smax -1.06 -1.28 -1.56 -1.16 -2.38 -3.46 -2.05 -2.61 -1.91 -1.50 -1.93 -0.93 -2.88 -2.43 -- -3.16 -2.37 -3.16 -2.74 -2.51 -1.20 -2.13 -1.31 -1.77 -1.80 -1.41 -1.99 jλ 0.50 -0.67 -0.61 0.21 -1.33 -2.40 -0.91 -1.59 -0.50 -0.09 -0.56 -2.31 -1.58 -1.37 -- -2.97 -0.41 -1.99 -1.57 -1.34 0.16 -0.80 -0.48 -0.71 -0.69 -0.90 -0.85 med. -0.31 -1.03 -0.96 -0.19 -1.39 -2.45 -0.93 -2.10 -0.83 -0.30 -1.19 -2.36 -1.83 -1.44 -- -3.06 -1.35 -2.01 -1.63 -1.90 -0.19 -1.12 -1.08 -1.03 -0.76 -0.94 -1.16 I0.4 -0.28 -1.06 -0.98 -0.08 -1.45 -2.61 -0.96 -2.05 -0.50 0.08 -1.09 -2.38 -1.94 -1.70 -3.75 -3.09 -1.04 -2.03 -1.89 -1.81 -0.15 -0.89 -1.04 -0.97 -0.39 -0.92 -1.05 U1 -0.85 -1.41 -0.91 -0.78 -1.04 -1.64 -0.85 -2.11 -1.32 -1.02 -1.33 -2.36 -1.38 -1.20 -4.65 -2.91 -2.25 -1.92 -0.98 -1.93 -0.62 -1.18 -1.06 -1.18 -0.83 -0.79 -1.19 7 current densities only in the midrange of the parameter list, outrun by kinetic properties that have generally received less attention, such as the the vapor absorption coefficient at the ionomer -- gas interface, ka. Just like in the local analysis, the phase change rates γc, γe turn out to be the least influential in the examined range. Although the condition number is the appropriate quantity to measure error propagation through a model, it provides no directly accessible insight into the correlation between input and output in a global sense -- such information can be extracted from scatter plots. A selection of scatter plots from our global sensitivity analysis is shown in Fig. 5 for the three most influ- ential MEA parameterizations (λv, ξ, LPEM ) and three of the most sensitive output quantities (Imax, Pmax, jλ). As can be recognized from Fig. 5a -- f, the limiting current density and the peak power density increase exponentially with the membrane hydration, but they decrease exponentially with the EOD coeffi- cient and also weakly with the membrane thickness. On the net water flux, on the other hand, the effect is more linear. Better ionomer hydration, stronger EOD and thicker membranes tend to let the water flow more from anode to cathode. 0 5 Conclusion In Part II of this paper series, we have carried out the first full- fledged local and global parameter sensitivity analyses of a state- of-the-art two-phase PEMFC model with spatial resolution. Our work showcases the strengths of modeling: A sensitivity study of this scale, with hundreds of thousands of measurements, would be absolutely unfeasible without a numerical model. Unlike previous efforts in this direction, our study rests on a statistical analysis of a dozen model output quantities, including heat and water balance characteristics, and more than two dozen input parameters. This allowed us to compare the modeling uncertainties associated with each major model parameter in a robust quantitative way. Rather than resorting to variance- based sensitivity estimates such as Laoun et al. [10], we have measured the propagation of uncertainty through the model explicitly by introducing the concept of condition numbers to fuel cell modeling. Four out of the five most critical model parameters (other than the environmental conditions at which the fuel cell is op- erated) are constitutive transport properties of the electrolyte membrane. Considering that these are precisely those for which the experimental data available in the open literature are scat- tered most, as shown in Part I, this is an unfortunate circum- stance that calls for better experimental characterization of the ionomer. For PEM fuel cell models to make the final leap to predictiveness, the interplay between the different water trans- port mechanisms and ionic conductivity must be known with high confidence. This article answers the previously unaddressed question how much error to expect from a PEMFC model given a certain error in its constitutive material properties, both locally at given operating conditions with common MEA materials, and globally for the case that the model is employed to predict the fuel cell behavior under unexplored conditions or material substitution. While our local analysis might be most meaningful for a typical automotive application, the results of the global analysis are of more fundamental interest for fuel cell modelers, because a model will unfold its full potential only when being applicable to an entire range of conditions. Acknowledgements Funding: This work was supported by the Swiss National Sci- ence Foundation [project no. 153790, grant no. 407040_153790]; Figure 4: Global model sensitivity to parameter uncertainty. Operating conditions (top group) and MEA parameterizations (bottom group) are separately sorted by significance (i.e., by median of log10 κij taken over all outputs i) in decreasing order. Error bars are standard errors of the statistics over i. membrane and dry gas feeds may in combination prevent the current density from reaching 1 A cm−2, making U1 undefined, in which case only the remaining eleven key figures are used in the statistical analysis for this sample. Tab. 4 presents the medians of log10 κij taken over the Sobol sequence (i.e., over all N samples). Analogously to Tab. 1, condition numbers which are numerically indistinguishable from zero within the used error tolerances are omitted. This is the case for the condensation rate γc, which has generally the lowest impact. Compared to the local sensitivity analysis, the overall picture changes slightly, but some specific observations remain valid. Rp and jλ are the most sensitive features, whereas Tmax and smax are the least sensitive ones. The global sensitivity of polarization curve properties generally increases toward higher current densities (lower cell voltages). Bear in mind that this is found here for the relative condition number, which puts the derivative of fi in relation to the function value fi itself, making this a nontrivial result. A statistical summary of the global analysis is presented in Fig. 4, in which both parameter categories are sorted by median over all outputs i to obtain a quantitative ranking of model parameters by impact on PEMFC modeling. First in the list of operating conditions is the boundary temperature, with the gas pressure close behind. Unlike in the local analysis at reference conditions, the global race between anode and cathode gas channel relative humidities is almost tied. In the category of MEA parameterizations, the hydration isotherm of the membrane ends up first in the list with a condition number (median over all outputs i) of almost unity, closely followed by the EOD coefficient. On ranks three to five are the membrane thickness, water diffusivity and protonic conductivity of the ionomer. The latter is, however, only little more influential than the applied clamping pressure on rank six. Contrarily to what is sometimes commonly presumed, we find the exchange 8 TA=TCPA=PCRHCRHA↵O2432101v⇠LPEM0DpPclLCL0LGDL0✏CLiMpkajCRejARTkdkDseecParametersensitivitylog10ijmaximummedianmeanFigure 5: Scatter plots showing a selection of model responses to variation in the three parameters with highest global sensitivity. Each data point represents a sample from the Sobol sequence in the parameter ranges listed in Tab. 3. The general trend it shown with linear least-square fits in the respective lin/log spaces (solid lines) and expressed as a functional relationship in each subplot label, with all fit values in units of the shown axes. the Swiss Commission for Technology and Innovation [contract no. KTI.2014.0115]; the Swiss Federal Office of Energy; and through the Swiss Competence Center for Energy Research (SCCER Mobility). References [1] A. Z. Weber, R. L. Borup, R. M. Darling, P. K. Das, T. J. Dursch, W. Gu, D. Harvey, A. Kusoglu, S. Litster, M. M. Mench, et al., Journal of The Electrochemical Society 161, F1254 (2014). [2] C. Siegel, Energy 33, 1331 (2008). [3] M. C. Kimble and R. E. White, J. Electrochem. Soc. 139, 478 [4] Q. Guo, V. A. Sethuraman, and R. E. White, J. Electrochem. (1992). Soc. 151, A983 (2004). [5] M. Grujicic and K. Chittajallu, Appl. Surf. Sci. 227, 56 (2004). [6] B. Carnes and N. Djilali, J. Power Sources 144, 83 (2005). [7] C. Min, X. Liu, Y. He, B. Yin, W. Jiang, and W. Tao, J. Power Sources 160, 359 (2006). [8] A. Mawardi and R. Pitchumani, J. Power Sources 160, 232 (2006). [9] D. Zhao, F. Gao, P. Massonnat, M. Dou, and A. Miraoui, IEEE Trans. Energy Convers. 30, 1008 (2015). [10] B. Laoun, M. W. Naceur, A. Khellaf, and A. M. Kannan, Int. J. [11] J. Corrêa, F. Farret, V. Popov, and M. Simões, IEEE Trans. Hydrog. Energy 41, 9521 (2016). Energy Convers. 20, 211 (2005). [12] L. Placca, R. Kouta, J.-F. Blachot, and W. Charon, J. Power Sources 194, 313 (2009). [13] M. Noorkami, J. B. Robinson, Q. Meyer, O. A. Obeisun, E. S. Fraga, T. Reisch, P. R. Shearing, and D. J. Brett, Int. J. Hydrog. Energy 39, 1439 (2014). [14] N. Noguer, D. Candusso, R. Kouta, F. Harel, W. Charon, and G. Coquery, Int. J. Hydrog. Energy 40, 3968 (2015). [15] G. Correa, F. Borello, and M. Santarelli, Int. J. Hydrog. Energy 40, 10354 (2015). [16] A. A. Shah, J. Electrochem. En. Conv. Stor. 14, 011006 (2017). [17] R. Vetter and J. O. Schumacher, J. Power Sources (2018), sub- mitted. 223 (2018). [18] R. Vetter and J. O. Schumacher, Comput. Phys. Commun. 234, [19] L. N. Trefethen and D. Bau III, Numerical Linear Algebra (Society for Industrial and Applied Mathematics, Philadelphia, 1997), ISBN 0-89871-361-7. [20] R. Vetter and J. O. Schumacher, in 14th Symposium on Fuel Cell and Battery Modelling and Experimental Validation (ModVal 14) (Karlsruhe, Germany, 2017), https://doi.org/10.21256/ zhaw-3634. [21] K. C. Neyerlin, G. Wenbin, J. Jorne, and H. A. Gasteiger, J. Electrochem. Soc. 153, A1955 (2006). [22] K. C. Neyerlin, W. Gu, J. Jorne, and H. A. Gasteiger, J. Elec- trochem. Soc. 154, B631 (2007). [23] I. M. Sobol, USSR Comput. Math. Math. Phys. 7, 86 (1967). 9 10−1100101(a)Imax≈100.26(λv/λrefv)−0.61Imax[Acm−2](b)Imax≈10−0.46(ξ/ξref)+0.29(c)Imax≈10−0.005LPEM0−0.0810−210−110010−2(d)Pmax≈100.25(λv/λrefv)−0.98Pmax[Wcm−2](e)Pmax≈10−0.31(ξ/ξref)−0.27(f)Pmax≈10−0.004LPEM0−0.470.51.01.52.0−50510(g)jλ≈0.87(λv/λrefv)−0.40Vaporsorptionisothermλv/λrefvjλ[µmolcm−2s−1]0.51.01.52.0(h)jλ≈1.25(ξ/ξref)−0.87EODcoefficientξ/ξref1020304050607080(i)jλ≈0.007LPEM0+0.36MembranethicknessLPEM0[µm]101
1804.00999	1	1804	2018-04-01T09:05:28	Dynamical homogenization of a transmission grating	[ "physics.app-ph" ]	A periodic assembly of acoustically-rigid blocks (termed 'grating'), situated between two half spaces occupied by fluid-like media, lends itself to a rigorous theoretical analysis of its response to an acoustic homogeneous plane wave. This theory gives rise to two sets of linear equations, the first for the amplitudes of the waves in the space between successive blocks, and the second for the amplitudes of the waves in the two half spaces. The first set is solved numerically to furnish reference solutions. The second set is submitted to low-frequency approximation procedure whereby the pressure fields are found to be those for a configuration in which the grating becomes a homogeneous layer of the same thickness as the height of the blocks in the grating. A simple formula is derived for the constitutive properties of this layer in terms of those of the fluid-like medium in between the blocks. The homogeneous layer model scattering amplitude transfer functions and spectral reflectance, transmittance and absorptance reproduce quite well the corresponding rigorous numerical functions of the grating over a non-negligible range of low frequencies. Due to its simplicity, the homogeneous layer model enables theoretical predictions of many of the key features of the acoustic response of the grating.	physics.app-ph	physics	Dynamical homogenization of a transmission grating Armand Wirgin∗ September 29, 2018 Abstract A periodic assembly of acoustically-rigid blocks (termed 'grating'), situated between two half spaces occupied by fluid-like media, lends itself to a rigorous theoretical analysis of its response to an acoustic homogeneous plane wave. This theory gives rise to two sets of linear equations, the first for the amplitudes of the waves in the space between successive blocks, and the second for the amplitudes of the waves in the two half spaces. The first set is solved numerically to furnish reference solutions. The second set is submitted to low-frequency approximation procedure whereby the pressure fields are found to be those for a configuration in which the grating becomes a homogeneous layer of the same thickness as the height of the blocks in the grating. A simple formula is derived for the constitutive properties of this layer in terms of those of the fluid-like medium in between the blocks. The homogeneous layer model scattering amplitude transfer functions and spectral reflectance, transmittance and absorptance reproduce quite well the corresponding rigorous numerical functions of the grating over a non-negligible range of low frequencies. Due to its simplicity, the homogeneous layer model enables theoretical predictions of many of the key features of the acoustic response of the grating. Keywords: dynamic response, homogenization, gratings. Abbreviated title: Dynamic homogenization of a grating Corresponding author: Armand Wirgin, e-mail: [email protected] 8 1 0 2 r p A 1 ] h p - p p a . s c i s y h p [ 1 v 9 9 9 0 0 . 4 0 8 1 : v i X r a ∗LMA, CNRS, UMR 7031, Aix-Marseille Univ, Centrale Marseille, F-13453 Marseille Cedex 13, France, ([email protected]) 1 Contents 1 Introduction 2 Description of the scattering configuration 3 The boundary-value problem 3.1 Field representations via separation of variables (DD-SOV) . . . . . . . . . . . . . . 3.2 Expressions for each of the four sets of unknowns . . . . . . . . . . . . . . . . . . . . 3.3 System of liner equations for the determination of the sets of coefficients {A[1]+ m } and {A[1]− m } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . m } . . 3.4 Numerical issues concerning the resolution of the system of equations for {A[1]± 3.5 System of liner equations for the determination of the sets of coefficients {A[0]− {A[2]+ n } . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . n } and 3 4 5 6 6 7 8 8 4 Approximate solution for the acoustic response of the transmission grating 9 9 4.1 The iterative scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 The low-order iterates in a low-frequency, large filling factor context 9 . . . . . . . . . 4.3 A conservation principle for the grating . . . . . . . . . . . . . . . . . . . . . . . . . 12 5 Acoustic response of a homogeneous layer situated between two fluid-like half 13 spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 5.1 DD-SOV Field representations 5.2 Solutions for the plane-wave coefficients and displacement fields . . . . . . . . . . . . 14 5.3 Comparison of the layer response to the grating response . . . . . . . . . . . . . . . 15 5.4 Conservation principle for the layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 6 Numerical results 17 6.1 Assumed parameters and general indications of the information contained in the graphs 17 6.2 Response of transmission gratings with wide spaces between blocks . . . . . . . . . . 17 6.3 Response of transmission gratings with medium-width spaces between blocks . . . . 22 6.4 Response of transmission gratings with narrow spaces between blocks . . . . . . . . 26 6.5 Response of deep transmission gratings as a function of incident angle . . . . . . . . 30 6.6 Response of shallow transmission gratings as a function of the incident angle . . . . 32 6.7 Response of transmission gratings as a function of their thickness . . . . . . . . . . . 34 7 Conclusions 39 2 1 Introduction The scheme by which an inhomogeneous (porous, inclusions within a homogeneous host, etc.) medium is reduced (with regard to its response to the solicitation) to a surrogate homogeneous medium is frequently termed 'homogenization'. What is meant by homogenization also includes the manner in which the physical characteristics of the surrogate are related to the structural and physical characteristics of the original, this being often accomplished by theoretical multiple scattering field averaging techniques [50, 51, 42, 29, 45, 49, 2, 10] or multiscale techniques (which actually also involve field averaging) [14, 38, 32]. Usually, the latter two techniques cannot be fully- implemented for other than static, or at least low frequency, solicitations [43, 14, 20]. There exist alternatives to theoretical field-averaging and multiscaling, applicable to a range of (usually-low) frequencies, which can be called: 'computational field averaging' [8] and 'computational parameter retrieval' [58, 40] approaches to dynamic homogenization. Recent research on metamaterials [30, 28, 39, 14, 15, 36, 11, 16, 41, 5, 6, 4] has spurred renewed interest in homogenization techniques [30, 46, 47, 13, 48, 49, 1, 44, 7, 18, 17, 32], the underlying issue being how to design an inhomogeneous medium so that it responds in a given manner (enhanced absorption [18, 20, 22], total transmission [31], reduced broadband transmission [24, 25] or various patterns of scattering [53, 54, 21]) to a wavelike (acoustic [31, 22], elastic [38], optical [53, 54], microwave [23], waterwave [34]) solicitation, the frequencies of which can exceed the quasi-static regime. This design problem is, in fact, an inverse problem that is often solved by encasing a specimen of the medium in a flat layer, treating the latter as if it were homogeneous, and obtaining its constitutive properties (in explicit manner in the NRW technique [37, 52, 3, 46, 9, 47, 19, 26, 27, 55] from the complex amplitudes of the reflected and transmitted waves that constitute the response of the layer (i.e., the data) to a (normally-incident in the NRW technique) plane body wave. To the question of why encase the specimen within a flat layer it can be answered that a simple, closed form solution is available for the associated forward problem of the reflected and transmitted-wave response of homogeneous layer to an incident plane body wave (the same being true as regards a stack of horizontal plane-parallel homogeneous layers employed in the geophysical context). Moreover, this solution is valid for all frequencies, layer thicknesses, and even incident angles and polarizations, such as is required in the homogenization problem since one strives to obtain a homogenized medium whose thickness does not depend on the specimen thickness (otherwise, how qualify it as being a medium?), but, on the contrary, he wants to find out how its properties depend on the characteristics of the solicitation. Moreover, the requirement of normal incidence (proper to the NRW technique) can be relaxed provided one accepts to search for the solution of the inverse problem in an optimization, rather than explicit, manner [40, 55, 57]. Herein we shall carry out the homogenization in a more theoretically-oriented manner by means of a three-step procedure: 1) establish the rigorous system of equations for the amplitudes of the reflected and transmitted waves constituting the far-field response of the grating, 2) find a low- frequency approximation of theses amplitudes from the system of equations, 3) show that the so-obtained low-frequency approximate amplitudes correspond exactly to the amplitudes of the reflected and transmitted waves constituting the response of a homogeneous layer of the same thickness as that of the grating. Moreover, we show, by the same procedure, that the field in the space between blocks of the grating is exactly that of the field in the entire aforementioned homogeneous layer and we establish the relation of the velocity and mass density of the latter to the velocity and mass density of the filler material as well as to the filling fraction of this material. 3 2 Description of the scattering configuration The site in which is situated the grating consists of a lower half space filled with a homogeneous, isotropic non-lossy fluid-like medium wherein the mass density and bulk wave velocity are ρ[0], c[0] (both real) respectively and an upper half space filled with another homogeneous, isotropic fluid-like medium wherein the mass density and bulk wave velocity are ρ[2], c[2] (both real) respectively. The grating is composed of a periodic (along x, with period d) set of identical, perfectly-rigid rectangular blocks whose widths and heights are d − w and h respectively. The spaces in between successive blocks (width w) is filled with a third homogeneous, isotropic generally-lossy fluid-like medium wherein the mass density and bulk wave velocity are ρ[1] (real), c[1] (generally-complex) respectively. All the geometrical features of the grating and site are assumed to not depend on y and the various interfaces (between the given structure and the two half spaces are surfaces of constant z and extend indefinitely along x and y. The solicitation is a homogeneous compressional (i.e., acoustic) plane wave whose wavevector ki lies in the x − z (sagittal) plane)), such that Ai(ω) is the spectral amplitude and θi the incident angle, with ω = 2πf the angular frequency (f the frequency which will be varied). The mass densities and bulk wave velocities are assumed to not depend on f (nor, of course on θi and Ai). The assumed y-independence of the acoustic solicitation, as well as the fact that it was assumed that the blocks of the given structure do not depend on y entails that all the acoustic field functions do not depend on y. Thus, in this 2D scattering problem, the analysis will take place in the sagittal plane in which the vector joining the origin O to a point whose coordinates are x, z is denoted by x (see fig. 1. Figure 1: Sagittal plane view of the configuration. Ω0 is the lower half-space domain, Ω1 = ∪n∈ZΩ1n the domain occupied by the given periodic structure, Ω1n the domain of the filler space of width w between successive blocks, the blocks being of rectangular cross section and height h, with d the period (along x) of the given periodic structure, and Ω2 the upper half space domain. The interface (i.e., the line z = −h2) between Ω0 and Ω1 is designated by Γ0 = ∪n∈ZΓ0n and the interface between Ω1 and Ω2 by Γ1 = ∪n∈ZΓ1n, in which Γm0 is the portion of Γm included between x = −w/2 and x = x/2. 4 3 The boundary-value problem The positive real phase velocities in Ωl; l = 0, 2 are c[l], and the generally-complex phase velocity ′′[1] ≤ 0. All c[j] l = 0, 1, 2 are assumed do ′[1] > 0 and c ′′[1], with c ′[1] + ic in Ω1n n ∈ Z are c[1] = c not depend on ω. The positive real density in Ωl ; l = 0, 2 is ρ[l] > 0 and the real positive real densities in Ω1n n ∈ Z are ρ[1]. The three densities are assumed to not depend on ω. The wavevector ki is of the form ki = (ki x, ki z) = (k[0] sin θi, k[0] cos θi) wherein θi is the angle of incidence (see fig. 1), and k[l] = ω/c[l]. The total pressure wavefield u(x, ω) in Ωl ; l = 0, 1, 2 is designated by u[l](x, ω). The incident wavefield is u[0]+(x, ω) = ui(x, ω) = A[0]+(ω) exp[i(ki xx + ki zz)] , (1) wherein A[0]+(ω) = Aiσ(ω) and σ(ω) is the spectrum of the solicitation. The plane wave nature of the solicitation and the d-periodicity of Γ0 and Γ1 entails the quasi- periodicity of the field, whose expression is the Floquet condition u(x + d, z, ω) = u(x, z, ω) exp(iki xd) ; ∀x ∈ Ω0 + Ω1 + Ω2 . (2) Consequently, as concerns the response in Ω1, it suffices to examine the field in Ω10. The boundary-value problem in the space-frequency domain translates to the following relations (in which the superscripts + and − refer to the upgoing and downgoing waves respectively) satisfied by the total displacement field u[l](x; ω) in Ωl: u[l](x, ω) = u[l]+(x, ω) + u[l]−(x, ω) ; l = 0, 1, 2 , u[l] ,xx(x, ω) + u[l] (ρ[0])−1u[0] ,zz(x, ω) + (k[l])2u[l](x, ω) = 0 ; x ∈ Ωl ; l = 0, 1, 2 . ,z (x, −h, ω) = 0 ; ∀x ∈ [−d/2, w/2] ∪ [w/2, d/2] , (ρ[2])−1u[2] ,z (x, 0, ω) = 0 ; ∀x ∈ [−d/2, w/2] ∪ [w/2, d/2] , (ρ[2])−1u[1] ,x (±w/2, z, ω) = 0 ; ∀z ∈ [0, h2] , u[l](x, ω) − u[1+1](x, ω) = 0 ; x ∈ Γl ; l = 0, 1 , ,z (x, ω) − (ρ[l+1])−1u[1+1] (x, ω) = 0 ; x ∈ Γl ; l = 0, 1 , ,z (ρ[l])−1u[l] (3) (4) (5) (6) (7) (8) (9) wherein u,ζ (u,ζζ) denotes the first (second) partial derivative of u with respect to ζ. Eq. (4) is the space-frequency wave equation for compressional sound waves, (5)-(7) the rigid-surface boundary conditions, (8) the expression of continuity of pressure across the two interfaces Γ0 and Γ0 and (9) the expression of continuity of normal velocity across these same interfaces. Since Ωl ; l = 0, 2 are of half-infinite extent, the field therein must obey the radiation conditions u[l]−(x, ω) ∼ outgoing waves ; x → ∞ ; l = 0, 2 . (10) Various (usually integral equation) rigorous approaches [35, 23, 34] have been employed to solve this boundary value problem. Herein, we outline another rigorous technique, based on the domain decomposition and separation of variables technique previously developed in [53, 54]. 5 3.1 Field representations via separation of variables (DD-SOV) The application of the domain decomposition-separation of variables (DD-SOV) technique, The Floquet condition, and the radiation conditions gives rise, in the two half-spaces, to the field representations: u[0](x, ω) =Xn∈Z(cid:16)A[l]+ 0 (ω) exp[i(kxnx + k[0] znz)] + A[l]− 0 (ω) exp[i(kxnx − k[0] znz)](cid:17) , wherein: u[2](x, ω) =Xn∈Z A[2]+ n (ω) exp[i(kxnx + k[2] znz)] , kxn = ki x + 2nπ d , and, on account of (1), k[l] zn =q(cid:0)k[l](cid:1)2 − (kxn)2 ; ℜk[l] zn ≥ 0 , ℑk[l] zn ≥ 0 ω > 0 , A[0]+ n (ω) = A[0]+(ω) δn0 , with δn0 the Kronecker delta symbol. In the central block, the DD-SOV technique, together with the rigid body boundary condition (7), lead to u[1](x, ω) = in which ∞ Xm=0(cid:16)A[1]+ m (ω)) exp[iK [1] zmz)] + A[1]− m (ω)) exp[−iK [1] zmz)](cid:17) cos[Kxm(x + w/2)] , Kxm = mπ w , K [1] zm =q(cid:0)k[1](cid:1)2 − (Kxm)2 ; ℜK [1] zm ≥ 0 , ℑK [1] zm ≥ 0 ω > 0 . 3.2 Expressions for each of the four sets of unknowns From the remaining boundary and continuity conditions it ensues the four sets of relations: u[0] ,z (x, −h, ω) exp(−ikxj) u[2] ,z (x, 0, ω) exp(−ikxj) dx d dx d −d/2 = (ρ[1])−1Z d/2 = (ρ[1])−1Z d/2 −d/2 u[1] ,z (x, −h, ω) exp(−ikxj) dx d ; ∀j ∈ Z , (19) u[1] ,z (x, 0, ω) exp(−ikxj) dx d ; ∀j ∈ Z , (20) (11) (12) (13) (14) (15) (16) (17) (18) −d/2 (ρ[0])−1Z d/2 (ρ[2])−1Z d/2 Z w/2 −w/2 −d/2 u[0](x, −h, ω) cos[Kxl(x + w/2)] dx w/2 = Z w/2 −w/2 u[1](x, −h, ω) cos[Kxl(x + w/2)] dx w/2 ; l = 0, 1, 2, ... , (21) 6 Z w/2 −w/2 u[2](x, 0, ω) cos[Kxl(x + w/2)] dx w/2 = Z w/2 −w/2 u[1](x, 0, ω) cos[Kxl(x + w/2)] dx w/2 ; l = 0, 1, 2, ... , (22) which should suffice to determine the four sets of unknown coefficients {A[0]−}, {A[1]+}, {A[1]−}, and {A[2]+}. Employing the DD-SOV field representations in these four relations gives rise to: A[0]− j = A[0]+ j j (cid:1)−2 − (cid:0)ε− ρ[0]w 2ρ[1]d ε− j k[0] zj ∞ Xm=0hA[1]+ m e− m − A[1]− m e+ zmE− jm ; ∀j ∈ Z , mi K [1] A[2]+ j = ρ[2]w 2ρ[1]d A[1]+ l l + A[1]− e− l e+ l = zmE− jm ; ∀j ∈ Z , m i K [1] n + A[0]− ε− n nl ; l = 0, 1, 2, .. , ε+ ni E+ A[1]+ l + A[1]− l = nl ; l = 0, 1, 2, .. , (23) (24) (25) (26) 1 k[2] zj ǫl ∞ n ǫl m − A[1]− Xm=0hA[1]+ 2 Xn∈ZhA[0]+ 2 Xn∈Z nm =Z w/2 n E+ A[2]+ znh) , E± −w/2 wherein: m = exp(±iK [1] e± zmh) , ε± n = exp(±ik[0] exp(±ikxnx) cos[Kxm(x + w/2)] dx w/2 (27) and ǫ0 = 1, ǫm>0 = 2. More specifically: E± nm = imsinc([±kxn + Kxm]w/2) + i−msinc([±kxn − Kxm]w/2) (28) in which sinc(ζ) = sin(ζ)/ζ and sinc(0) = 1. 3.3 System of liner equations for the determination of the sets of coefficients {A[1]+ m } and {A[1]− m } We can go a step further by inserting two of the expressions (23)-(26) into the other two and interchange summations to obtain: lmQ1 m + P 12 lmQ2 lmQ1 m + P 22 lmQ2 P∞ m=0(cid:0)P 11 P∞ m=0(cid:0)P 21 l m(cid:1) = R1 m(cid:1) = R2 l ; l = 0, 2, ... , (29) e+ l e− mK [1] zmS[0] lm , P 12 l (cid:1)2 lm =(cid:0)e+ δlm − ǫlwρ[0] 4dρ[1] e+ l e+ mK [1] zmS[0] lm , R1 m = A[0]+ǫle+ l ε− in which P 11 lm = δlm + ǫlwρ[0] 4dρ[1] P 21 lm = δlm − ǫlwρ[2] 4dρ[1] e+ l e− mK [1] zmS[2] lm , P 22 lm = δlm + ǫlwρ[2] 4dρ[1] K [1] zmS[0] lm , R2 m = 0 , S[j] lm = Xn∈Z 0 E+ 0l (30) E+ nlE− nm k[j] zn , (31) and Q1 m = A[1]+ m , Q2 m = A[1]− m . 7 3.4 Numerical issues concerning the resolution of the system of equations for {A[1]± m } We strive to obtain numerically the sets {A[1]± m } from the linear system of equations (29). Once these sets are found, they are introduced into (23)-(24) to obtain the sets {A[0]− n } and {A[2]+ n }. Concerning the resolution of the infinite system of linear equations (29), the procedure is basi- cally to replace it by the finite system of linear equations ; l = 0, 2, ..., M , (32) lm Q1 m + P 12(N ) lm Q2 lm Q1 m + P 22(N ) lm Q2 m=0(cid:16)P 11(N ) PM m=0(cid:16)P 21(N ) PM l m(cid:17) = R1 m(cid:17) = R2 l lm signifies that the series in S[k] in which P jk(N ) lm therein is limited to the terms n = 0, ±1, ..., ±N , N having been chosen to be sufficiently large to obtain numerical convergence of these series, and to increase M so as to generate the sequence of numerical solutions {Qj(0) },....until the values of the first few members of these sets stabilize and the remaining members become very small. This is usually obtained for values of M , that are all the smaller the lower is the frequency. When all the coefficients (we mean those whose values depart significantly from zero) are found, they enable the computation of the far-field and near-field acoustic responses that are of interest herein. The so-obtained numerical solutions for these coefficients and fields, can, for all practical purposes, be considered to be 'exact' since they compare very well with their finite element or integral equation counterparts in [35, 23, 34]. }, {Qj(1) , Qj(1) 0 0 1 3.5 System of liner equations for the determination of the sets of coefficients {A[0]− n } and {A[2]+ n } We can proceed differently by inserting (25)-(26) into (23)-(24) and interchanging summations to obtain: in which X 11 jn = δjn− ρ[0]w ρ[1]4id ε− j k[0] zj Pn∈Z(cid:16)X 11 Pn∈Z(cid:16)X 21 jnY 1 n + X 12 jnY 2 jnY 1 n + X 22 jnY 2 j n(cid:17) = Z 1 n(cid:17) = Z 2 j ; j ∈ Z , (33) ε+ n Σ2 jn , X 12 jn = ρ[0]w ρ[1]4id ε− j k[0] zj Σ1 jn , Z 1 j = A[0]+ j + ρ[0]w ρ[1]4id j (cid:17)2 (cid:16)ε− ε− j k[0] zj Xν∈Z A[0]+ ν ν Σ2 ε− jν (34) X 21 jn = ρ[2]w ρ[1]4id 1 k[2] zj ε+ n Σ1 jn , X 22 jn = δlm − ρ[2]w ρ[1]4id 1 k[2] zj Σ2 jn , Z 2 j = − ρ[2]w ρ[1]4id 1 k[2] zj Xν∈Z A[0]+ ν ν Σ1 ε− jν , (35) Σ1 jn = ǫm K [1] zm Sm ∞ Xm=0 and E− jmE+ nm , Σ2 jn = ǫm K [1] zmCm Sm ∞ Xm=0 E− jmE+ nm , n = A[0]− Y 1 n , Y 2 n = A[2]+ m , Cm = cos(K [1] zmh), Sm = sin(K [1] zmh) . 8 (36) (37) 4 Approximate solution for the acoustic response of the transmis- sion grating We now discuss an iterative approach for the resolution of the second system of equations for {A[0]− n } and {A[2]+ n }. We shall limit ourselves to the first-order iterate. 4.1 The iterative scheme We can write (33) as X 11 jj Y 1 j + X 12 jj Y 2 j = Z 1 jnY 1 n + X 12 jnY 2 X 21 jj Y 1 j + X 22 jj Y 2 j = Z 2 jnY 1 n + X 22 jnY 2 j −P∞ j −P∞ n=−∞,6=j(cid:16)X 11 n=−∞,6=j(cid:16)X 21 n(cid:17) n(cid:17) ; j ∈ Z . (38) It ensues formally: wherein Y 1 j = ∞ W 1 j = Z 1 j − Xn=−∞,6=j(cid:0)X 11 W 1 j X 22 X 11 jj X 22 jj − W 2 jj − X 21 j X 12 jj j X 12 jj , Y 2 j = W 2 j X 11 X 11 jj X 22 jj − W 1 jj − X 21 j X 21 jj j X 12 jj ; j ∈ Z , (39) ∞ jnY 1 n + X 12 jnY 2 j = Z 2 j − jnY 1 n + X 22 jnY 2 . (40) n(cid:1) , W 2 Xn=−∞,6=j(cid:0)X 21 n(cid:1) , The iterative scheme is then; W 1(k−1) j Y 1(k) j = X 22 jj − W 2(k−1) j Dj X 12 jj , Y 2(k) j = W 2(k−1) j X 11 jj − W 1(k−1) j Dj X 21 jj ; j ∈ Z , k = 1, 2, ... , wherein and W 1(k−1) j = Z 1 j − Dj = X 11 jj X 22 jj − X 21 j X 12 jj . ∞ Xn=−∞,6=j(cid:16)X 11 jnY 1(k−1) n + X 12 jnY 2(k−1) n (cid:17) , ∞ W 2(k−1) j = Z 2 j − Xn=−∞,6=j(cid:16)X 21 jnY 1(k−1) n + X 22 jnY 2(k−1) n (cid:17) , In these expressions, we have not yet addressed the question of zeroth-order iterates Y 1(0) j (41) (42) . (43) , Y 2(0) j . 4.2 The low-order iterates in a low-frequency, large filling factor context The preceding formulae show that we must initiate the iteration procedure with a priori as- sumptions concerning Y 1(0) j , Y 2(0) j . We place ourselves in situations characterized by k[0] < kxj ; ∀j > 0 , (44) 9 k[2] < kxj ; ∀j > 0 , (45) which specify that only the j = 0 waves in the bottom and top media are homogeneous. In addi- tion, we assume that under these conditions, all the j 6= 0 (inhomogeneous) waves have vanishing amplitudes, which translates to We also shall assume j6=0 = Y 2(0) Y 1(0) j6=0 = 0 . k[0]w << π ⇒ ki xw << π , which together with (44), (45) and (46) corresponds to an essentially low-frequency context. We make one further assumption w/d ≈ 1 , which corresponds to a large filling fraction (i.e., narrow block) situation. Let us now see what the consequences of these assumptions are. We had E± 0m = im{sinc[(±kx0 + Kxm)w/2] + (−1)msinc[(±kx0 − Kxm)w/2]} = im{sinc[(±ki xw/2 + mπ/2] + (−1)msinc[(±ki xw/2 − mπ/2]} , or, on account of (47), E± 0m ≈ im[1 + (−1)m]sinc(mπ/2) = 2δm0 . Furthermore, (47) and (48) tell us that E± n0 = im{1 + (−1)m}sinc(±nπw/d) ≈ 2sinc(nπ) = 2δn0 , the consequences of which are (via (36)): Σ1 j0 = 4 K [1] z0 S0 δj0 = Σ1 00δj0 , Σ2 j0 = 4 K [1] z0 C0 S0 δj0 = Σ2 00δj0 , It then follows from (46) ad (52) that: W 1(0) j = Z 1 0 δj0 , W 2(0) j = Z 2 0 δj0 , (46) (47) (48) (49) (50) (51) (52) (53) whence Y 1(1) j = Y 2(1) j = Eq. (52) entails W 1(0) j X 22 jj − W 2(0) j X 12 jj =(cid:18)Z 1 Dj W 2(0) j X 11 jj − W 1(0) j X 21 jj Dj D0 =(cid:18) Z 2 0 X 22 00 − Z 2 0 X 12 00 (cid:19) δj0 = Y 1(1) 0 δj0 = N 1(1) 0 D0 δj0 0 X 11 00 − Z 1 0 X 21 00 D0 (cid:19) δj0 = Y 2(1) 0 δj0 = N 2(1) 0 D0 δj0 . (54) X 11 00 = 1− ρ[0]k[1] z0 w ρ[1]k[0] z0 d C0 iS0 , X 12 00 = ρ[0]k[1] z0 w ρ[1]k[0] z0 d ε− 0 iS0 , X 21 00 = ρ[2]k[1] z0 w ρ[1]k[2] z0 d ε+ 0 iS0 , X 22 00 = 1− ρ[2]k[1] z0 w ρ[1]k[2] z0 d C0 iS0 . (55) 10 Consequently: D0 = X 11 00 X 22 00 − X 21 00 = 00 X 12 ρ[0]k[1] z0 w ρ[1]k[0] z0 d 1 − C0 iS0! 1 − ρ[2]k[1] z0 w ρ[1]k[2] z0 d C0 iS0! − ρ[2]k[1] z0 w ρ[1]k[2] z0 d ε+ 0 iS0! ρ[0]k[1] z0 w ρ[1]k[0] z0 d This expression is easily shown to reduce to: D0 =(cid:18) i g0g2S0(cid:19)(cid:2)g1(cid:0)g2 + g0(cid:1)C0 −(cid:0)g2g0 + g1g1(cid:1)iS0(cid:3) , g0 = k[0] z0 ρ[0] , g1 = k[1] z0 w ρ[1]d , g2 = k[2] z0 ρ[2] . in which By the same token, Z 1 0 = A[0]+(cid:0)ε− 0(cid:1)2(cid:20)1 + so that: g1C0 g0iS0(cid:21) , Z 2 0 = −A[0]+ε− 0 (cid:20)1 + g1 g2iS0(cid:21) , ε− 0 iS0! . (56) (57) (58) (59) N 1(1) 0 = Z 1 0 X 22 00 − Z 2 0 X 12 00 = A[0]+(cid:0)ε− N 2(1) 0 = Z 2 0 X 11 00 − Z 1 0 X 21 00 = −A[0]+ε− g0iS0(cid:21)(cid:20)1 − g1C0 0(cid:1)2(cid:20)1 + g2iS0(cid:21)(cid:20)1 − 0 (cid:20) g1 g1C0 g2iS0(cid:21)(cid:20) g1ε− g2iS0(cid:21) + A[0]+ε− g0iS0(cid:21) , 0 (cid:20) g1 g0iS0(cid:21)(cid:20) g1ε+ g2iS0(cid:21) , 0(cid:1)2(cid:20)1 + g0iS0(cid:21) − A[0]+(cid:0)ε− g1C0 g1C0 0 0 (60) which reduce to: N 1(1) 0(cid:1)2 0 = −A[0]+(cid:0)ε− whence: i g0g2 (cid:2)−(cid:0)g1g1 − g2g0(cid:1)iS0 + g1(cid:0)g2 − g0(cid:1)C0(cid:3) , N 2(1) 0 = −A[0]+ε− 0 (cid:20) 2g1 g2iS0(cid:21) , (61) Y 1(1) 0 = A[0]−(1) 0 and we recall (54) = −A[0]+(cid:0)ε− 0(cid:1)2(cid:20) g1(g2 − g0)C0 + (g2g0 − g1g1)iS0 g1(g2 + g0)C0 − (g2g0 + g1g1)iS0(cid:21) , Y 2(1) 0 = A[2]+ 0 = A[0]+ε− 0 (cid:20) g1(g2 + g0)C0 − (g2g0 + g1g1)iS0(cid:21) , 2g1g0 Y 1(1) j6=0 = A[0]−(1) j6=0 = 0 , Y 2(1) j6=0 = A[2]+(1) j6=0 = 0 . (62) (63) We now turn to the first-order iterate of the coefficients of the field in the region between successive blocks. From (25)-(26) we obtain A[1]+(1) l l + A[1]−(1) e− l e+ l = n + A[0]−(1) ε− n ǫl 2 Xn∈ZhA[0]+ n 11 nl ; l = 0, 1, 2, .. , ε+ ni E+ (64) A[1]+(1) l + A[1]−(1) l = ǫl 2 Xn∈Z A[2]+(1) n E+ nl ; l = 0, 1, 2, .. , (65) Using the (62)-(63) gives rise (with the help of (50)) to A[1]+(1) l l + A[1]−(1) e− l e+ l = ǫl 2 hA[0]+ 0 0 + A[0]−(1) ε− 0 A[1]+(1) l + A[1]−(1) l = ǫl 2 A[2]+(1) 0 0 ε+ 0l =hA[0]+ 0i E+ 0l ; l = A[2]+(1) E+ 0 0 + A[0]−(1) ε− 0 ε+ 0i δl0 ; l = 0, 1, 2, .. , δl0 ; l = 0, 1, 2, .. , (66) (67) The introduction of (62)-(63) into (66)-(69) finally yields A[1]−(1) l = A+ 0 ε− 0 (cid:20) g0(g1 − g2) g1(g2 + g0)C0 − (g2g0 + g1g1)iS0(cid:21) δl0 , A[1]+(1) l = A+ 0 ε− 0 (cid:20) g0(g1 + g2) g1(g2 + g0)C0 − (g2g0 + g1g1)iS0(cid:21) δl0 ; l = 0, 1, 2, .. . (68) The higher-order iterates of the various field coefficients can be obtained in similar manner, but we shall content ourselves with the first order results. However, the latter will be compared to the solutions of a homogeneous layer scattering problem to see what are the connections, if any, between the two. A last word is here in order about the fields associated with the first-order approximation of the coefficients. These fields are: u[0](1) = u[0]+ +Xn∈Z A[0]−(1) n exp[i(kxnx − k[0] znz)] = 0 exp[i(kx0x + k[0] A+ z0 z)] + A[0]−(1) 0 exp[i(kx0x − k[0] z0 z)] , (69) u[1](1) = ∞ Xm=0hA[1]+(1) m exp(iK [1] zmz) + A[1]−(1) m exp(−iK [1] zmz)i cos[Kxm(x + w/2)] = A[1]+(1) 0 exp(iK [1] z0 z) + A[1]−(1) 0 exp(−iK [1] z0 z)] , u[2](1) =Xn∈Z A[2]+(1) n exp[i(kxnx + k[2] znz)] = A[2]+(1) 0 exp[i(kx0x + k[2] z0 z)] , (70) (71) 4.3 A conservation principle for the grating Using Green's second identity, it is rather straightforward to demonstrate the following conser- vation principle: ρ + α + τ = 1 , (72) wherein ρ an and τ are what can be termed hemispherical reflected and transmitted fluxes respec- tively, given by: ρ = ℜXn∈Z n k2k[0] kA[0]− zn kA[0]+k2k[0] z0 , τ = ℜXn∈Z 12 n k2k[2] kA[2]− kA[0]+k2k[0] znρ[0] z0 ρ[2] (73) and α is what can be called the absorbed flux given by α = 1 k[0] z0 d ρ[0] ρ[1] d , (74) 2 ℑ(cid:20)(cid:16)k[1](cid:17)2(cid:21)ZΩ10(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) u[1](x, ω) A[0]+(ω)(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) wherein d is the infinitesimal element of area in the central interstitial domain Ω10 of the grating, it having been assumed, as previously, that the media filling both the lower and upper half-spaces are non-lossy. In addition, (74) tells us that the absorbed flux vanishes when the medium filling the interstitial spaces is non-lossy (i.e. when c[1], and therefore k[1], are real. z0 = ki The expression (73) shows (since k[0] z is real) that the only contributions to ρ stem from diffracted-reflected waves for which k[0] zn is real, and the only contributions to τ stem from diffracted- zn is real. Real k[0] transmitted waves for which k[2] zn corresponds to homogeneous plane waves in the lower half space and real k[2] zn to homogeneous plane waves in the upper half space. The angles of emergence of these observable homogeneous waves are θ− n = arcsin(kxn/k[2]) (note that θ− 0 = θi is the angle of specular reflection in the sense of Snell, and θ+ 0 = arcsin(k[0] sin θi/k[2]) the angle of refraction in the sense of Fresnel). Thus, the flux in each half space is composed of a denumerable, finite, set of subfluxes, which fact is expressed by: n = arcsin(kxn/k[0]) and θ+ ρ = Xn∈H− ρn , τ = Xn∈H+ τn . wherein: ρ− n = n kA[0]− k2k[0] zn kA[0]+k2k[0] z0 , τ + n = n kA[2]− k2k[2] kA[0]+k2k[0] znρ[0] z0 ρ[2] . and H− the set of n for which k[0] zn is real, whereas H+ is the set of n for which k[2] zn is real. Actually, (72) can be resumed by the following conservation relation fout = fin , (75) (76) (77) wherein fout = ρ + α + τ is the normalized output flux and fin = 1 the normalized input flux. It is important to underline the fact that the conservation principle is a rigorous consequence of the grating boundary-value problem and therefore does not depend on the low-frequency, large filling fraction approximations made previously. However, this principle can (and will) be employed to test the consistency of both the rigorous and first-iterate solutions of the grating response. 5 Acoustic response of a homogeneous layer situated between two fluid-like half spaces This configuration, depicted in fig. 2, is similar to the previous one, except that the grating is replaced by a homogeneous layer situated between the same two half spaces. Its acoustic response can be treated in the classical manner [12] outlined hereafter. The solicitation, bottom and top half spaces are as previously (i.e., in the problem corresponding to the grating). The surrogate occupies the layer-like domain Ω1 (see fig. 2) in which the properties are designated by the superscript 1. Thus, the boundary-value problem is expressed by (1), (3), (4), 13 Figure 2: Sagittal plane view of the homogeneous layer configuration. Ω0 is the lower half-space domain, Ω1 the surrogate domain in the form of a layer of thickness h. Ω2 is the half-space above the superstructure. The interface (i.e., the line z = −h) between Ω0 and Ω1 is designated by Γ0 and the interface between Ω1 and Ω2 by Γ1. (8), (9), (10) (in which ρ[l] is replaced by R[l]), c[l] by C [l]), and u[l] by U [l]), with the understanding that: C [l] = c[l] , R[l] = ρ[l] ; l = 0, 2 . U [0]+(x, ω) = u[0]+(x, ω) = ui(x, ω) = A[0]+ exp[i(ki xx + ki zz] , with ki x = K [0] sin θi, ki z = K [0] cos θi and K [l] = ω/C [l] ; l = 0, 1, 2. 5.1 DD-SOV Field representations Separation of variables and the radiation condition lead to the field representations: U [l](x, ω) = U [l]+(x, ω) + U [l]−(x, ω) ; ∀x ∈ Ωl , l = 0, 1, 2 . (78) (79) (80) (81) z = k[l] z0 = with U [l]±(x, ω) = A[l]± exphi(kxx ± k[l] z z)i ; ∀x ∈ Ωl , l = 0, 1, 2 . in which A[0]+ is as previously, and the relation to previous wavenumbers is as follows: k[l] q(cid:0)K [l](cid:1)2 −(cid:0)kx(cid:1)2. 5.2 Solutions for the plane-wave coefficients and displacement fields The four interface continuity relations lead to A[0]−e[0] − A[1]+(cid:0)e[1](cid:1)−1 − A[1]−e[1] = −A[0]+(cid:0)e[0](cid:1)−1 −A[0]−e[0] − A[1]+G[10](cid:0)e[1](cid:1)−1 + A[1]−G[10]e[1] = −A[0]+(cid:0)e[0](cid:1)−1 A[1]+ + A[1]− − A[2]+ = A[1]+ − A[1]− − A[2]+G[21] = 0 0 , (82) 14 (with e[j] = exp(ik[j] A[1]−, A[2]+. The solution of (82) is: z h) and G[jk] = G[j]/G[k], G[j] = k[j] z /R[j]) for the four unknowns A[0]−, A[1]+, A[0]− = −A[0]+(cid:0)e[0(cid:1)−2h(cid:0)G[2] + G[1](cid:1)(cid:0)G[1] − G[0](cid:1)(cid:0)e[1](cid:1)−1 +(cid:0)G[2] − G[1](cid:1)(cid:0)G[1] + G[0](cid:1)e[1]i D−1 , wherein These expressions are easily cast into the following forms: A[1]+ = A[0]+(cid:0)e[0(cid:1)−1h2G[0](cid:0)G[1] + G[2](cid:1)i D−1 , A[1]− = A[0]+(cid:0)e[0(cid:1)−1h2G[0](cid:0)G[1] − G[2](cid:1)i D−1 , A[2]+ = A[0]+(cid:0)e[0(cid:1)−1(cid:2)4G[1]G[0](cid:3)D−1 , D =(cid:0)G[2] + G[1](cid:1)(cid:0)G[1] + G[0](cid:1)(cid:0)e[1](cid:1)−1 +(cid:0)G[2] − G[1](cid:1)(cid:0)G[1] − G[0](cid:1)e[1] . A[0]− = −A[0]+(cid:0)e[0(cid:1)−2" G[1](cid:0)(cid:0)G[2] − G[0](cid:1)C +(cid:0)G[2]G[0] − G[1]G[1](cid:1)iS G[1](cid:0)(cid:0)G[2] + G[0](cid:1)C −(cid:0)G[2]G[0] + G[1]G[1](cid:1)iS# , A[1]+ = A[0]+(cid:0)e[0(cid:1)−1" G[1](cid:0)(cid:0)G[2] + G[0](cid:1)C −(cid:0)G[2]G[0] + G[1]G[1](cid:1)iS# , G[1](cid:0)(cid:0)G[2] + G[0](cid:1)C −(cid:0)G[2]G[0] + G[1]G[1](cid:1)iS# , A[1]− = A[0]+(cid:0)e[0(cid:1)−1" G[1](cid:0)(cid:0)G[2] + G[0](cid:1)C −(cid:0)G[2]G[0] + G[1]G[1](cid:1)iS# . A[2]+ = A[0]+(cid:0)e[0(cid:1)−1" G[0](cid:0)G[1] + G[2](cid:1) G[0](cid:0)G[1] − G[2](cid:1) 2G[2]G[0] (83) (84) (85) (86) (87) (88) (89) (90) (91) (92) (93) (94) (95) (96) (97) wherein S = sin(k[1] z h), C = cos(k[1] z h). 5.3 Comparison of the layer response to the grating response We notice that because of the assumption made at the outset that C [l] = c[l] ; l = 0, 2 which implies that (cid:0)e[0](cid:1)±1 = ε± 0 , If, in addition, we assume that z = k[l] k[l] z0 ; l = 0, 2 . ′[1] = c C ′′[1] = c C ′[1] , ′′[1] , then z = k[1] k[1] z0 ⇒ S = S0 , C = C0 . Finally, if we assume (with φ = w/d the so-called filling factor) that R[1] = ρ[1]φ−1 , 15 and recall that we assumed at the outset that then R[l] = ρ[l] ; l = 0, 2, G[l] = gl ; l = 0, 1, 2 , so that the comparison of (88)-(99) with (62), (63) and (69) shows that A[0]−(1) 0 = A[0]− , A[1]±(1) 0 = A[0]± , A[2]±(1) 0 = A[2]± . (98) (99) (100) The conditions (94) and (97) are what one would expect to obtain from a mixture theory [43] approach to homogenization. Also the comparison of (69)-(71) with (80)-(81) shows, on account also of 47), (93) and (96) that U [l] = u[l](1) ; = 0, 1, 2 , (101) which indicates equality of the first-order approximation of the fields in the grating configuration with the corresponding fields in the homogeneous layer configuration when (94) and (97) prevail. This means that the first-order iteration approximation amounts to replacing the transmission grating by a homogeneous layer. Furthermore, the mass density of this 'homogenized' layer is simply the mass density of the generic grating block ρ[1] divided by the filling factor φ, all other parameters of the layer (thickness h, bulk wave velocity C [1] of the layer, bulk wave velocities C [0] and C [2] of the bottom and top half spaces, mass densities ρ[0] and ρ[2] of the bottom and top half spaces, incident angle θi, and amplitude spectrum A[0]+(ω) of the acoustic solicitation, being the same as for the transmission grating configuration. 5.4 Conservation principle for the layer Again using Green's second identity, leads in rather straightforward manner to the following conservation principle: R + A + T = 1 , (102) wherein R and T are the hemisperical= single-wave reflected and transmitted fluxes respectively, given by: R = kA[0]−k2 kA[0]+k2 , T = kA[2]−k2k[2] kA[0]+k2k[0] z R[0] z R[2] and A the absorbed flux given by A = 1 k[0] z d R[0] R[1] ℑ(cid:20)(cid:16)K [1](cid:17)2(cid:21)ZΩ1d(cid:13)(cid:13)(cid:13)(cid:13)(cid:13) U [1](x, ω) A[0]+(ω) (cid:13)(cid:13)(cid:13)(cid:13)(cid:13) (103) , 2 d , (104) with Ω1d the portion of the layer situated between x = −d/2 and x = d/2. Expression (103) shows, unsurprisingly (since k[0] z are real) that the only contri- bution to R stems from the single specularly-reflected homogeneous plane wave(s) and the only contribution to T stems from the single transmitted homogeneous plane wave (there exist no other transmitted) wave(s)). The angles of emergence of these observable homogeneous waves z and k[2] z = ki 16 are θ− = arcsin(kx/k[0]) = θi and θ+ = arcsin(kx/k[2]) (note again that θ− = θi is the angle of specular reflection in the sense of Snell, and θ+ = arcsin(k[0] sin θi/k[2]) the angle of refraction in the sense of Fresnel). Actually, (81) can be resumed by the following conservation relation wherein Fout = R+A+T is the normalized output flux and Fin = 1 the normalized input flux.oth to the rigorous and approximate (i.e., l−order iterate) solutions to the transmission grating problem. Fout = Fin , (105) 6 Numerical results The purpose of the numerical study is essentially to find out how well the surrogate layer model responses compare to the grating responses. 6.1 Assumed parameters and general indications of the information contained in the graphs In the following figures it is assumed, unless written otherwise, that: Ai = A[0]+ = 1, θi = ′′[1] = 0◦, d = 0.004 m, h = 0.00475 m, c[0] = 343 ms−1, ρ[0] = 1.2 Kgm−3, c ′[1] = c ′′[1], R[1] = 0 ms−1 ρ[1] = 2.4 Kgm−3, c[2] = 343 ms−1, ρ[2] = 1.2 Kgm−3, C φ/ρ[1]. Recall that the sites and thicknesses of the grating and layer are identical as are their solicitations. ′[1] = 500 ms−1, c ′[1], C ′′[1] = c In all the figures, the blue curves are relative to the grating whereas the red curves correspond to the surrogate (homogeneous) layer. In the (1,1) (left-hand, uppermost) panel the full curves depict the amplitude transfer functions kA[2]+ k and kA[2]+k whereas the dashed curves depict kA[1]− 0 In the (2,1) (left-hand, lowermost) panel the full curves depict kA[0]− k and kA[1]−k. 0 k and kA[0]−k whereas the 0 dashed curves depict kA[1]+ 0 k and kA[1]+k. In the (1,2) (middle, uppermost) panel the full curves depict the transmitted fluxes (=spectral transmittances) τ (blue) and T (red). In the (2,2) (middle, lowermost) panel the full curves depict the reflected fluxes (=spectral reflectances) ρ (blue) and R (red). In the (1,3) (right-hand, uppermost) panel the full curves depict the input fluxes fin (blue), Fin (red) whereas dashed curves depict the output fluxes fout (blue), Fout(red). In the (2,3) (right- hand, lowermost) panel the full curves depict the absorbed fluxes (=spectral absorptances) α (blue) and A (red). 6.2 Response of transmission gratings with wide spaces between blocks In figs. 3-9 we consider the case w = 0.003 m for various C ′′[1] = c ′′[1]. 17 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 t u o F , t u o f , n F , n i f i 1 1 1 1 1 1 1 a f l A , a f l a 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 3: C ′′[1] = c ′′[1] = 0 5 f (Hz) 10 x 104 2 4 6 f (Hz) 8 10 x 104 . ) . u a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R , o h r 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 1 1 1 1 1 0.8 a f l A , a f l a 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 Figure 4: C ′′[1] = c ′′[1] = −5 18 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 1.0001 t u o F , t u o f , n F , n i f i a f l A , a f l a 1 1 1 1 1 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 5: C ′′[1] = c ′′[1] = −10 ) . u . a ( s e d u t i l p m a p o t ) . . u a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R o h r , 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 1.0001 1 1 0.9999 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 1 0.8 0.6 0.4 0.2 0 a f l A a , f l a 5 f (Hz) 10 x 104 Figure 6: C ′′[1] = c ′′[1] = −20 19 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . . u a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 1.0001 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 7: C ′′[1] = c ′′[1] = −30 ) . u . a ( s e d u t i l p m a p o t . ) . u a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 t u o F , t u o f , 1 i n F n , i f 0.9999 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 a f l A a , f l a 5 f (Hz) 10 x 104 Figure 8: C ′′[1] = c ′′[1] = −40 20 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R o h r , 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 t u o F , t u o f , 1 i n F n , i f 0.9999 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1 0.8 a f l A a , f l a 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 9: C ′′[1] = c ′′[1] = −50 21 These figures show that the agreement between the grating (blue curves) and layer (red curves) responses is very good up till about 30KHz. This is as expected since the first-order iteration grating model= homogeneous layer model derives essentially from a low-frequency approximation. What is less expected is the rather good agreement between these two responses even beyond 30KHz except in the neighborhood f ≈ 80.6KHz of occurrence of a Wood anomaly [33]. We note that flux is perfectly-well conserved for both configurations at all the considered frequencies. Other noticeable features of the grating response, also present in the layer response, are: (i) the total transmission peak near 50KHz when the interstitial material is lossless, (ii) the near- coincidence of frequencies of occurrence of the maxima of transmission and absorption, (iii) the ′′[1]k) and (iv) the fact that more nonlinear increase of absorption with the increase of kC than 35% of the incident flux is absorbed beyond f ≈ 40 KHz, with a peak of 55% at f ≈ 55 KHz, when C ′′[1] = −50 ms−1. ′′[1] = c ′′[1]k = kc 6.3 Response of transmission gratings with medium-width spaces between blocks In figs. 10-16 we consider the case w = 0.002 m for various C ′′[1] = c ′′[1]. 3 2 1 0 4 3 2 1 0 s e d u t i l p m a p o t s e d u t i l p m a m o t t o b 1 0.8 u a T , u a t 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 o h R , o h r t u o F , t u o f , n F , n i f i 1 1 1 1 1 1 1 0.8 a f l A , a f l a 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 Figure 10: C ′′[1] = c ′′[1] = 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 22 u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 t u o F , t u o f , n F , n i f i 1.0001 1.0001 1 1 0.9999 a f l A , a f l a 1 0.8 0.6 0.4 0.2 0 Figure 11: C ′′[1] = c ′′[1] = −5 2.5 2 1.5 1 0.5 s e d u t i l p m a p o t 0 4 3 2 1 0 s e d u t i l p m a m o t t o b 2.5 2 1.5 1 0.5 s e d u t i l p m a p o t 1 0.8 u a T , u a t 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 o h R o h r , 1.0001 1.0001 1 1 0.9999 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 1 0.8 0.6 0.4 0.2 0 a f l A a , f l a 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 0 4 3 2 s e d u t i l p m a m o t t o b 1 0 5 f (Hz) 10 x 104 Figure 12: C ′′[1] = c ′′[1] = −10 23 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 t u o F , t u o f , n F , n i f i 1.0002 1.0001 1 0.9999 a f l A , a f l a 1 0.8 0.6 0.4 0.2 0 Figure 13: C ′′[1] = c ′′[1] = −20 2.5 2 1.5 1 0.5 s e d u t i l p m a p o t 0 4 3 2 1 0 s e d u t i l p m a m o t t o b 2.5 2 1.5 1 0.5 s e d u t i l p m a p o t 1 0.8 u a T , u a t 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 o h R o h r , 1.0001 1.0001 1 1 0.9999 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 1 0.8 0.6 0.4 0.2 0 a f l A a , f l a 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 0 4 3 2 s e d u t i l p m a m o t t o b 1 0 5 f (Hz) 10 x 104 Figure 14: C ′′[1] = c ′′[1] = −30 24 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 t u o F , t u o f , n F , n i f i 1.0001 1.0001 1 1 0.9999 0.9999 a f l A , a f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 15: C ′′[1] = c ′′[1] = −40 2.5 2 1.5 1 0.5 s e d u t i l p m a p o t 0 4 3 2 1 0 s e d u t i l p m a m o t t o b 2.5 2 1.5 1 0.5 s e d u t i l p m a p o t 1 0.8 u a T , u a t 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 o h R o h r , 1.0001 t u o F , t u o f , 1 i n F n , i f 0.9999 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 a f l A a , f l a 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 0 4 3 2 s e d u t i l p m a m o t t o b 1 0 5 f (Hz) 10 x 104 Figure 16: C ′′[1] = c ′′[1] = −50 25 These figures show that the agreement between the grating (blue curves) and layer (red curves) responses is very good up till about 25KHz. This is as expected since the first-order iteration grating model= homogeneous layer model derives essentially from a low-frequency approximation and is should fare less well for smaller w/d. What is less expected is the rather good agreement between these two responses even beyond 25KHz except in the neighborhood f ≈ 80.6KHz of occurrence of the Wood anomaly. We note that flux is nearly perfectly-well conserved for both configurations at all the considered frequencies. Other noticeable features of the grating response, also present in the layer response, are: (i) the total transmission peak near 50KHz when the interstitial material is lossless, (ii) the near- coincidence of frequencies of occurrence of the maxima of transmission and absorption, (iii) the ′′[1]k) and (iv) the fact nonlinear increase of peak abosorption with the increase of kC that more than 10% of the incident flux is absorbed beyond f ≈ 30 KHz, with a peak of 55% at f ≈ 53 KHz, when C ′′[1] = −50 ms−1. ′′[1]k = kc ′′[1] = c 6.4 Response of transmission gratings with narrow spaces between blocks In figs. 17-23 we consider the case w = 0.001 m for various C ′′[1] = c ′′[1]. 6 5 4 3 2 1 0 8 6 4 2 0 s e d u t i l p m a p o t s e d u t i l p m a m o t t o b u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R o h r , 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 t u o F , t u o f , i n F n , i f 2 4 6 f (Hz) 8 10 x 104 1 1 1 1 1 1 1 2 4 6 f (Hz) 8 10 x 104 0.8 a f l A a , f l a 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 Figure 17: C ′′[1] = c ′′[1] = 0 26 6 5 4 3 2 1 0 8 6 4 2 0 6 5 4 3 2 1 0 8 6 4 2 0 s e d u t i l p m a p o t s e d u t i l p m a m o t t o b s e d u t i l p m a p o t s e d u t i l p m a m o t t o b u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R , o h r 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1.0001 1.0001 1 1 0.9999 1 0.8 a f l A , a f l a 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 18: C ′′[1] = c ′′[1] = −5 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0002 1.0001 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 19: C ′′[1] = c ′′[1] = −10 27 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 6 5 4 3 2 1 0 8 6 4 2 0 6 5 4 3 2 1 0 8 6 4 2 0 s e d u t i l p m a p o t s e d u t i l p m a m o t t o b s e d u t i l p m a p o t s e d u t i l p m a m o t t o b u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R , o h r 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1.0002 1.0001 1 0.9999 1 0.8 a f l A , a f l a 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 20: C ′′[1] = c ′′[1] = −20 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 1.0001 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 21: C ′′[1] = c ′′[1] = −30 28 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 6 5 4 3 2 1 0 8 6 4 2 0 6 5 4 3 2 1 0 8 6 4 2 0 s e d u t i l p m a p o t s e d u t i l p m a m o t t o b s e d u t i l p m a p o t s e d u t i l p m a m o t t o b u a T , u a t 1 0.8 0.6 0.4 0.2 0 1 0.8 o h R , o h r 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1.0001 1.0001 1 1 0.9999 1 0.8 a f l A , a f l a 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 22: C ′′[1] = c ′′[1] = −40 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 1.0001 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 2 4 6 f (Hz) 8 10 x 104 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 23: C ′′[1] = c ′′[1] = −50 29 These figures show that the agreement between the grating (blue curves) and layer (red curves) responses is very good up till about 20KHz. This is as expected since the first-order iteration grating model= homogeneous layer model derives essentially from a low-frequency approximation and is should fare less well for smaller w/d. What is less expected is the rather good agreement between these two responses even beyond 20KHz except in the neighborhood f ≈ 80.6KHz of occurrence of the Wood anomaly. We note that flux is near perfectly-well conserved for both configurations at all the considered frequencies. Other noticeable features of the grating response, also present in the layer response, are: (i) the total transmission peak near 50KHz when the interstitial material is lossless, a feature that is unexpected for a grating with such narrow interstices, but in agreement with what has been predicted previously [31] by finite element computations and verified experimentally, (ii) the near- coincidence of frequencies of occurrence of the maxima of transmission and absorption, (iii) the ′′[1]k) nonlinear increase, followed by leveling-off, of absorption with the increase of kC and (iv) the rather unexpected fact (considering the narrowness of the interstices between grating blocks) that more than 5% of the incident flux is absorbed beyond f ≈ 20 KHz, with a peak ′′[1] = −50 ms−1. The large absorption is probably due of 50% at f ≈ 50 KHz, when C to the existence of a very strong acoustic field within the interstices of the grating (and therefore throughout the surrogate layer). ′′[1]k = kc ′′[1] = c 6.5 Response of deep transmission gratings as a function of incident angle In figs. 24-26 we consider the case w = 0.003 m, h = 0.00475 m, C ′′[1] = c ′′[1] = −25 ms−1 for various θi. 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1.0001 1.0001 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 24: θi = 0 ◦ 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 2 4 6 f (Hz) 8 10 x 104 30 2 1.5 1 0.5 ) . u . a ( s e d u t i l p m a p o t 0 3 2 1 0 4 3 2 1 0 5 4 3 2 1 0 ) . u . a ( s e d u t i l p m a m o t t o b ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1 1 0.9999 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 25: θi = 20 ◦ u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 26: θi = 40 ◦ 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 31 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 These figures show that the agreement between the two responses diminishes with increasing incident angle, notably because the frequency at which the lowest-order inhomogeneous waves in the half-spaces become homogeneous (this occurring at the frequencies of the Wood anomalies [33]) diminishes with increasing θi, so that the agreement is good up till 50 KHz for θi = 0◦, up till ≈ 15 KHz for θi = 20◦, and only up till ≈ 10 KHz for θi = 40◦. Nevertheless, there appears to exist agreement as to the secular trends of response (notably as concerns absorbed flux) between the two configurations for the three angles of incidence. A last observation concerns the near-perfect conservation of flux for both the grating and equivalent layer at the three angles of incidence. 6.6 Response of shallow transmission gratings as a function of the incident angle In figs. 27-29 we consider the case w = 0.003 m, h = 0.00175 m, C ′′[1] = c ′′[1] = −25 ms−1 for various θi. 4 3 2 1 0 5 4 3 2 1 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1 1 1 1 1 1 a f l A , a f l a 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 10 8 x 104 2 6 4 f (Hz) 10 8 x 104 Figure 27: θi = 0 ◦ 2 4 6 f (Hz) 10 8 x 104 2 6 4 f (Hz) 10 8 x 104 2 4 6 f (Hz) 10 8 x 104 2 6 4 f (Hz) 10 8 x 104 32 2 1.5 1 0.5 ) . u . a ( s e d u t i l p m a p o t 0 3 2 1 0 ) . u . a ( s e d u t i l p m a m o t t o b ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 t u o F , t u o f , i n F n , i f 2 4 6 f (Hz) 8 10 x 104 1 1 1 1 1 1 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 a f l A a , f l a 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1 1 1 1 1 1 a f l A a , f l a 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 Figure 28: θi = 20 ◦ u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 Figure 29: θi = 40 ◦ 33 These figures show that, contrary to the case of deep blocks/layers, the agreement between the two responses is excellent for all frequencies up to the one at which the Wood anomaly occurs, this frequency diminishing with increasing angle of incidence. Again, there appears to exist agreement as to the secular trends of response (notably as concerns absorbed flux) between the two configurations for the three angles of incidence. A last observation concerns the perfect conservation of flux for both the grating and equivalent layer at the three angles of incidence. 6.7 Response of transmission gratings as a function of their thickness In figs. 30-37 we consider the case θi = 0◦, w = 0.003 m, C ′′[1] = c ′′[1] = −25 ms−1 for various h. 1.6 1.4 1.2 1 0.8 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1 1 1 1 1 1 1 a f l A , a f l a 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 Figure 30: h = 0m 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 34 ) . . u a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 t u o F , t u o f , i n F n , i f 2 4 6 f (Hz) 8 10 x 104 1 1 1 1 1 1 2 4 6 f (Hz) 8 10 x 104 a f l A a , f l a 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 Figure 31: h = 0.00075m 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 t u o F , t u o f , n F , n i f i 1 1 1 1 1 1 a f l A a , f l a 0.8 0.6 0.4 0.2 0 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 Figure 32: h = 0.00175m 2 4 6 f (Hz) 8 10 x 104 2 4 6 f (Hz) 8 10 x 104 35 ) . . u a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 1 1 1 1 1 1 a f l A a , f l a 0.8 0.6 0.4 0.2 0 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 33: h = 0.00275m 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 34: h = 0.00375m 36 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.0001 1.0001 1 1 0.9999 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , i n F n , i f 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 35: h = 0.00475m 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R o h r , 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1.0001 1 0.9999 0.9998 a f l A a , f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 36: h = 0.00575m 37 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 ) . u . a ( s e d u t i l p m a p o t ) . u . a ( s e d u t i l p m a m o t t o b u a T , u a t o h R , o h r 1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0 t u o F , t u o f , n F , n i f i 1.0001 1 0.9999 0.9998 a f l A , a f l a 1 0.8 0.6 0.4 0.2 0 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 Figure 37: h = 0.00675m 5 f (Hz) 10 x 104 5 f (Hz) 10 x 104 38 These figures show that the agreement between the two responses is excellent, for all h, and for all frequencies up to the one at which the Wood anomaly occurs, this frequency being practically constant as a function of h. Of particular interest is the absorbed flux, which is practically the same for the grating and equivalent layer, whose maximum peak gradually shifts to lower frequencies and levels off in height beyond h = 0.00375 m while increasing in height, even with the formation of a second peak, starting at h = 0.00575 m. Thus, the deep gratings (and equivalent layers) turn out to be quite efficient devices for wide-band absorption of acoustic waves. A last observation concerns the near-perfect conservation of flux for both the grating and equivalent layer for all h. 7 Conclusions The principal object of this investigation was to: (a) obtain a simple model of acoustic response of block-like transmission gratings by an approximation procedure applied to a rigorous model of this response (b) evaluate the applicability and precision of approximate responses derived from the simple model by comparing them to the numerical solutions obtained from the rigorous model. It was shown that the approximate model amounts to considering the grating to behave, with respect to an acoustic solicitation, as a homogeneous layer of the same thickness as that of the blocks of the grating, the real and imaginary parts of the bulk wave velocity of this layer being equal to the real and imaginary parts of the bulk wave velocity in the interstitial medium of the grating, and the mass density of the layer being equal to that of the interstitial medium divided by the filling factor (φ = w/d) of the grating, all other constitutive parameters as well as the solicitation, being the same in the layer and grating configurations. It was found, via a series of numerical tests, backed up by conservation of flux checks , that the layer model enables to predict many (but not all of) the principal features of seismic response for a wide range of grating thicknesses provided the φ is close to one (its maximum value) and the frequency of the solicitation is relatively low, these being the conditions invoked in the approxima- tion procedure adopted to derive the (effective) layer response from that of the (rigorous) grating response. It was found that a surprising feature of the layer model is that it enables to predict the first (non-zero frequency) peak of EAT ('Extraordinary Acoustic Transmission' [31], which tranforms to significant absorption in the interstices when the latter are filled with a lossy medium), even for small φ. However, the second peak, which is linked to the Wood anomaly, is not accounted-for in the layer model. In fact, the existence of the Wood anomalies (which occur for frequencies and incident angles at which an inhomogeneous scattered wave becomes homogeneous) is so tightly linked with the d-periodic nature of the scattering configuration (the inhomogeneous waves being absent in the layer modeel) that they cannot make their appearance in the response of a homogeneous layer unless the effective mass density and/or velocity of the surrogate layer are dispersive (this possibility was ruled out a priori herein, but was taken into accout in studies such as [14, 57]). In spite of this shortcoming, the layer model deriving from our low-frequency homogenization scheme, appears to give meaningful predictions of the response of the transmission grating well beyond the static limit and can therefore be qualified as 'dynamical'. Moreover, these predictions can be improved either by the technique outlined in [56, 57] or by taking into account higher-order iterates in the scheme presented herein. Finally, our homogenization scheme may provide a useful 39 alternative to traditional multiscale and field averaging approaches to homogenization of periodic structures as regards their response to dynamic solicitations. References [1] Andryieuski A., Malureanu R. and Lavrinenko A.V., Homogenization of metamaterials: Pa- rameters retrieval methods and intrinsic problems, ICTON 2010 paper Mo.B2.1, (2010). [2] Arist´egui C. and Angel Y.C., Effective mass density and stiffness derived from P-wave multiple scattering, Wave Motion, 44, 153-164 (2007). [3] Barroso J.J., De Paula A.L., Retrieval of permittivity and permeability of homogeneous ma- terials from scattering parameters, J. Electromagn. Waves Appl., 24, 1563-1574 (2010). [4] Brul´e S., Ungureanu B., Achaoui Y., Diatta A., Aznavourian1 R., Antonakakis T., Craster R., Enoch S. and Guenneau S., Metamaterial-like transformed urbanism, Innov. Infrastruct. Solut. 2, 20, DOI 10.1007/s41062-017-0063-x (2017). [5] Brunet T., Merlin A., Mascaro B., Zimny K., Leng J., Poncelet O., Arist´egui C. and Mondain- Monval O., Soft 3D acoustic metamaterial with negative index, Nat Mater. 14(4), 384-388, doi: 10.1038/nmat4164 (2015). [6] Brunet T., Poncelet O., and Arist´egui C., Negative-index metamaterials: is double negativity a real issue for dissipative media?, EPJ Appl. Metamat., 2, 3, DOI: 10.1051/epjam/2015005 (2015). [7] Campione S., Sinclair M.B. and Capolino F., Effective medium representation and complex modes in 3D periodic metamaterials made of cubic resonators with large permittivity at mid- infrared frequencies, Photonics Nanostruct., 11, 423-435 (2013). [8] Chekroun M., Le Marrec L., Lombard B. and Piraux J.,Time-domain numerical simulations of multiple scattering to extract elastic effective wavenumbers, Waves Rand. Complex Media, 22(3), 398-422 (2012). [9] Chen X., Grzegorczyk T.M., Wu B.I., Pacheco J., and Kong J.A., Robust method to retrieve the constitutive effective parameters of metamaterials, Phys. Rev. E 70, 016608 (2004) . [10] Conoir J.M. and Norris A., Effective wavenumbers and reflection coefficients for an elastic medium containing random distributions of cylinders in a fluid-like region, Wave Motion, 46, 522-538 (2009). [11] Dubois J., Homog´en´eisation dynamique de milieux al´eatoires en vue du dimensionnement de m´etamat´eriaux acoustiques, Phd thesis, Univ. Bordeaux 1, Bordeaux (2012). [12] Ewing W.M., Jardetzky W.S. and Press F, Elastic Waves in Layered Media, McGraw-Hill, New York (1957). [13] Fang N. Xi D., Xu J., Ambati M., Srituravanich W., Sun C. and Zhang X., Ultrasonic metamaterials with negative modulus, Nature Mater., 5, 452-456 (2006). 40 [14] Felbacq D. and Bouchitt´e G., Negative refraction in periodic and random photonic crystals, New J. Phys. 7, 159 (2005). [15] Fokin V., Ambati M., Sun C. and Zhang X., Method for retrieving effective properties of locally resonant acoustic metamaterials, Phys. Rev. B 76, 144302 (2007). [16] Forrester D.M. and Pinfield V.J., Shear-mediated contributions to the effective properties of soft acoustic metamaterials including negative index, Nature Scientif. Repts., 5, 18562, DOI: 10.1038/srep18562 (2015). [17] Gao Y., Experimental study and application of homogenization based on metamaterials, Phd thesis, Univ.Pierre et Marie Curie, Paris (2016). [18] Groby J.-P., Lardeau A., Huang W. and Aur´egan Y., The use of slow sound to design simple sound absorbing materials, J. Appl. Phys., 117, 124903 (2015). [19] Groby J.-P., Ogam E., De Ryck L., Sebaa N. and Lauriks W., Analytical method for the ultra- sonic characterization of homogeneous rigid porous materials from transmitted and reflected coefficients, J. Acoust. Soc. Am., 127(2), 764-772 (2010). [20] Groby J.-P., Pommier R. and Aur´egan Y., Use of slow sound to design perfect and broadband passive sound absorbing materials, J. Acoust. Soc. Am., 139(4), 153-164 (2016). [21] Jim´enez N., Cox T.J., Romero-Garcia V. and Groby J.-P., Metadiffusers: Deep-subwavelength sound diffusers, Scientif. Repts., 7, 5389 (2017). [22] Jim´enez N., Romero-Garcia V., Pagneux V. and Groby J.-P., Rainbow-trapping absorbers : Broadband, perfect and asymmetric sound absorption by subwavelength panels with ventilation, ArXiv: 1708.03343v1 [physics.class-ph] (2017). [23] Kobayashi K. and Miura K., Diffraction of a plane wave by a thick strip grating, IEEE. Anten. Prop., 37(4), 459-470 (1989). [24] Lagarrigue C., Groby J.-P. and Tournat V., Sustainable sonic crystal made of resonating bamboo rods, J. Acoust. Soc. Am., 133(1), 145-254 (2013). [25] Lardeau A., Groby J.-P. and Romero Garcia V., Broadband transmission loss using 3D locally resonant sonic crystals, Crystals, 6, 51 (2016). [26] Lepert G., Arist´egui C., Poncelet O., Brunet T. Audoly C. and Parneix P., Recovery of the effective wavenumber and dynamical mass density for materials with inclusions, hal-00811114 (2012). [27] Lepert G., Arist´egui C., Poncelet O., Brunet T. Audoly C. and Parneix P., Determination of the effective mechanical properties of inclusionary materials using bulk elastic waves, J. of Physics, Conf. series, 498, 012007 (2014). [28] Li J. and Chan C.T., Double-negative acoustic metamaterial, Phys. Rev. E, 70, 055602(R) (2004). 41 [29] Linton C.M. and Martin P.A., A new proof of the Lloyd-Berry formula for the effective wavenumber, SIAM J. Appl.Math., 66, 1649-1668 (2006). [30] Liu X.-X., Alu A., Generalized retrieval method for metamaterial constitutive parameters based on a physically-driven homogenization approach, Phys. Rev. B, 87, 235136 (2013). [31] Lu M.-H., Liu X.-K., Feng L., Li J., Huang C.-P. and Chen Y.-F. Extraordinary acoustic transmission through a 1D grating with very narrow apertures, Phys. Rev. Lett., 99, 174301 (2007). [32] Maling B., Colquitt D.J. and Craster R.V. Dynamic homogenisation of Maxwell's equations with applications to photonic crystals and localised waveforms on gratings, Wave Motion, 69, 35-49 (2017). [33] Maradudin A.A. , Simonsen I., Polanco J. and Fitzgerald R.M., Rayleigh and Wood anomalies in the diffraction of light from a perfectly conducting reflection grating, J. Optics, 18(2), 024004 (2016). [34] Mc Lean N.D., Water wave diffraction by segmented permeable breakwaters, Phd thesis, Loughborough Univ. (1999). [35] Miles J.W., The diffraction of a plane wave through a grating, Quart. Appl. Math. 7, 45-64 (1949). [36] Morvan B., Tinel A., Hladky-Hennion A.-C., Vasseur J.O. and Dubus B., Experimental demonstration of the negative refraction of a transverse elastic wave in a two-dimensional solid phononic crystal, Appl. Phys. Lett., 96(10), 101905 (2010). [37] Nicolson A.M. and Ross G., Measurement of the intrinsic properties of materials by time- domain techniques, IEEE Trans. Instrum. Meas., 19, 377-382 (1970). [38] Parnell W.J and Abrahams I.D., Dynamic homogenization in periodic fibre reinforced media. Quasi-static limit for SH waves, Wave Motion, 43, 474-498 (2006). [39] Pendry J.B. , Martin-Moreno L. and Garcia-Vidal F.J., Mimicking surface plasmons with structured surfaces, Science, Aug 6, 305(5685), 847-8 (2004). [40] Qi J., Kettunen H., Wall´en H. and Sihvola A., Different homogenization methods based on scattering parameters of dielectric-composite slabs, Radio Sci., 46, RS0E08, doi:10.1029/2010RS004622 (2011). [41] Raffy S., Mascaro B. , Brunet T., Mondain-Monval O. and Leng J., A soft 3D acoustic metafluid with dual-band negative refractive index, Adv. Mater., DOI: 10.1002/adma.201503524 (2015). [42] Sayers C.M., On the propagation of ultrasound in highly concentrated mixtures and suspen- sions, J. Phys. D, 13, 179-184 (1980). [43] Sihvola A., Mixing models for heterogeneous and granular Media, in Advances in Electromag- netics of Complex Media and Metamaterials, Zouhdi S., Sihvola A. and Arsalane M. (eds.), Kluwer, Amsterdam (2002). 42 [44] Simovski C.R., Tretyakov S.A., On effective electromagnetic parameters of artificial nanos- tructured magnetic materials, Photonics Nanostruct.Fundamen., 8(4), 254-263 (2010). [45] Smith D.R. and Pendry J.B., Homogenization of metamaterials by field averag- ing, J. Opt. Soc. Am. B, 23, 391 (2006). [46] Smith D.R., Schultz S., Markos P., Soukoulis C.M., Determination of effective permittivity and permeability of metamaterials from reflection and transmission coefficients, Phys. Rev. B, 65, 195104 (2002). [47] Smith D.R., Vier D.C., Koschny T., Soukoulis C.M., , Electromagnetic parameter retrieval from inhomogeneous metamaterials, Phys. Rev. E, 71, 036617 (2005). [48] Torrent D., Hakansson A., Cervera F. and Sanchez-Dehesa J., Homogenization of two- dimensional clusters of rigid rods in air, Phys. Rev. Lett., 96, 204302 (2006). [49] Torrent D. and Sanchez-Dehesa J., Effective parameters of clusters of cylinders embedded in a nonviscous fluid or gas, Phys. Rev. B, 74, 224305 (2006). [50] Urick R.J. and Ament W.S., Propagation of sound in composite media, J. Acoust. Soc. Am., 21, 115-119 (1949). [51] Waterman P.C. and Truell R., Multiple scattering of waves, J. Math. Phys., 2(4), 512-537 (1961). [52] Weir W.B., Automatic measurement of complex dielectric constant and permeability at mi- crowave frequencies, Proc. IEEE, 62, 33-36 (1974). [53] Wirgin A. High-efficiency frequency-modulation non-redundant scanning of a light beam in a half space, Opt. Comm., 26(1), 153-157 (1978). [54] Wirgin A., Can the reflectivity of a nonlossy slab of arbitrary thickness loaded with a strip grating be identical to that of a plane interface over a wide range of frequencies?, J. Opt. Soc. Am. A, 5(2), 214-219 (1988). [55] Wirgin A., Retrieval of the frequency-dependent effective permeability and permittivity of the inhomogeneous material in a layer, Prog. Electromag. Res. B, 70, 131-147, (2016). [56] Wirgin A., Apparent attenuation and dispersion arising in seismic body-wave velocity re- trieval, Pure Appl. Geophys. 173, 2473-2488 (2016). [57] Wirgin A. Computational parameter retrieval approach to the dynamic homogenization of a periodic array of rigid rectangular blocks, arXiv:1803.04717v1 [physics.app-ph] (2018). [58] Wirgin A. and Ghariani S., Verification of the Urick-Ament description of the effective dy- namic response of random composite fibrous media, J. Physique IV, C1, C1-17 -- C1-21 (1992). 43
1905.10394	1	1905	2019-05-24T18:07:50	Enhanced solar evaporation using a photo-thermal umbrella: towards zero liquid discharge wastewater management	[ "physics.app-ph" ]	Rising water demands and depleting freshwater resources have brought desalination and wastewater treatment technologies to the forefront. For sustainable water management, there is a global push towards Zero Liquid Discharge (ZLD) with the goal to maximize water recovery for reuse, and to produce solid waste that lowers the environmental impact of wastewater disposal. Evaporation ponds harvest solar energy as heat for ZLD, but require large land areas due to low evaporation rates. Here, we demonstrate a passive and non-contact approach to enhance evaporation by more than 100% using a photo-thermal umbrella. By converting sunlight into only mid-infrared radiation where water is strongly absorbing, efficient utilization of solar energy and heat localization at the water surface through radiative coupling are achieved. The non-contact nature of the device makes it uniquely suited to treat a wide range of wastewater, and the use of commercially available materials enables a potentially low cost and scalable technology for the sustainable disposal of wastewater, with the added benefit of salt recovery.	physics.app-ph	physics	Enhanced solar evaporation using a photo-thermal umbrella: towards zero liquid discharge wastewater management Akanksha K. Menon1$, Iwan Haechler1$, Sumanjeet Kaur1, Sean Lubner1, Ravi S. Prasher1,2* 1Energy Storage & Distributed Resources Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA 2Department of Mechanical Engineering, University of California, Berkeley, CA 94720, USA * e-mail: [email protected] $ Equal contributions Abstract Rising water demands and depleting freshwater resources have brought desalination and wastewater treatment technologies to the forefront. For sustainable water management, there is a global push towards Zero Liquid Discharge (ZLD) with the goal to maximize water recovery for reuse, and to produce solid waste that lowers the environmental impact of wastewater disposal. Evaporation ponds harvest solar energy as heat for ZLD, but require large land areas due to low evaporation rates. Here, we demonstrate a passive and non-contact approach to enhance evaporation by more than 100% using a photo-thermal umbrella. By converting sunlight into only mid-infrared radiation where water is strongly absorbing, efficient utilization of solar energy and heat localization at the water's surface through radiative coupling are achieved. The device's non-contact nature makes it uniquely suited to treat a wide range of wastewater, and the use of commercially available materials enables a potentially low cost and scalable technology for the sustainable disposal of wastewater, with the added benefit of salt recovery. 1 The World Economic Forum recognizes water crises as a major global risk that has arisen from the depletion of natural freshwater resources due to agricultural, industrial and municipal use, while generating vast amounts of wastewater.1 This poses a sustainability challenge that currently threatens four billion people worldwide and is expected to become more severe with population growth and economic development.2,3 Desalination of seawater and inland brackish water has emerged as a solution to meet increasing water demand. However, desalination plants produce concentrated brine as a byproduct, disposal of which is detrimental to land vegetation and the aquatic ecosystem, thereby having a significant environmental impact.4,5 Thus, there has been a push towards maximizing water recovery for reuse from industrial wastewater and desalination brine to achieve Zero Liquid Discharge (ZLD) such that the final waste is a solid. At present, ZLD involves a series of treatment processes that (i) reduce the volume of wastewater using membrane-based systems or thermal brine concentrators, and (ii) reduce the concentrated brine to solid waste using a crystallizer or an evaporation pond.6 The choice for this final step of ZLD depends upon various factors including concentrated brine volume and composition, energy requirement, local climate and land costs.7,8 A crystallizer is a complex mechanical system (forced circulation evaporator) that requires high-grade heat and electricity resulting in a large energy cost and capital cost. Although crystallizers have a small site footprint, their operating costs depend heavily on the composition of wastewater, which can be exorbitant for highly scaling water (e.g. wastewater containing large amounts of silica).9 In contrast, evaporation ponds rely on solar energy to passively evaporate water from any waste stream, resulting in low energy and operating costs. Currently, evaporation ponds are implemented in China, Australia, Europe (Mediterranean region), the Middle East and some areas of the U.S. where they are economically viable owing to inexpensive land and a suitable climate (arid or semi-arid).6,9,10 Although they offer a tremendous advantage of being suitable for different wastewater streams, capital costs are high due to the large land footprint required for natural evaporation. This footprint is inversely proportional to the evaporation rate, which is inherently low due to the passive nature and inefficient use of solar energy in these ponds. To reduce the environmental impact (i.e., smaller 2 areal footprint) and capital costs, evaporation enhancement in brine disposal ponds is essential and different approaches have been implemented in this regard.11,12 Active methods of enhancement include Wind-Aided Intensified Evaporation (WAIV) and droplet spraying, and passive techniques include the use of solar radiation absorbing organic dyes, wetted floating fins and salt tolerant plants.7,8,12 These methods have been shown to enhance evaporation rates by up to 35%, while WAIV has shown a 50% enhancement at high salinities but requires continuous electric pumping making it an active system. Recently, a new passive approach for solar evaporation enhancement has emerged where the emphasis is to avoid wastefully heating a large volume of water and instead perform surface heating by localizing sunlight at the water -- air interface.13,14 Various prototypes using nanomaterial-based absorbers and bio-inspired structures exploiting surface heating have since been reported, with conversion efficiencies of over 90% for these floating structures.15-20 While these are practical for solar stills and steam generation applications, their continuous operation in high salinity wastewater (such as evaporation ponds) causes salt precipitation at the surface in contact with water, resulting in deterioration of the optical and wicking properties over time.16 Thus, there is a need to develop non-contact and passive technologies for enhanced solar evaporation that can reduce the footprint of evaporation ponds and eliminate contamination from fouling and scaling. Here, we demonstrate a novel surface heating approach that enhances evaporation by over a 100% under 1 sun, with the potential to increase evaporation by 160% compared to traditional evaporation ponds through thermal design. The system relies solely on radiative coupling using a photo-thermal conversion device comprising a selective solar absorber and blackbody emitter, as shown in Figure 1. Since this photo- thermal device shields the water in an evaporation pond from direct sunlight, we refer to it as a "solar umbrella." In this paper, we design a lab-scale system to experimentally demonstrate the potential of the solar umbrella for enhancing evaporation from concentrated brines. Using thermal models validated by these lab-scale experiments, we predict the performance of a large-scale evaporation pond for sustainable 3 wastewater disposal. By enhancing the evaporation two-fold, the land required for the same volume of wastewater disposal is halved, which has significant environmental and sustainability impact. Concept: increasing surface temperature with radiative heat localization Efficient utilization of solar energy for evaporation is limited by the transparency of water at visible and near-IR wavelengths owing to its low absorption coefficient of 0.01 m-1.21 As a result, a large fraction (~60%) of incident solar flux results in volumetric or sensible heating of bulk water as shown in Figure 1a. Although the temperature of water increases as sunlight is absorbed, that increase is distributed in the whole volume of liquid (large thermal mass) resulting in a lower temperature at the water surface, as compared to the case where all the radiation is absorbed at the surface. As a result, the transient thermal response of traditional evaporation ponds is slow, which when coupled with the diurnal variation of solar radiation, results in low evaporation rates. At mid-IR and larger wavelengths however, the absorption coefficient of water increases by several orders of magnitude to 104 m-1, and consequently, radiation is absorbed and localized close to the surface (small thermal mass). Given that evaporation is a surface phenomenon, by shifting the solar radiation to mid-IR and larger wavelengths, surface heating can be achieved and evaporation can be enhanced. As an example, a blackbody with a temperature < 150 °C emits ~99.9% of radiation at wavelengths above 2 µm that can be absorbed in a thin layer of water ~100 μm (Supplementary Information). We achieve the shifting of solar radiation to mid-IR and larger wavelengths by utilizing a photo-thermal converter comprising a selective solar absorber and a blackbody emitter shown in Figure 1b. When used as an umbrella over a water surface such as an evaporation pond, the overall efficiency (η) of the system can be expressed as: 𝜂 = ( 𝑚 ℎ𝑓𝑔 𝑞𝑠𝑜𝑙𝑎𝑟 ) = 𝜂1 × 𝜂2 × 𝜂3 = ( 𝛼𝑠𝑞𝑠𝑜𝑙𝑎𝑟−𝜀𝑠𝜎𝑇𝑎𝑏𝑠 4 𝑞𝑠𝑜𝑙𝑎𝑟 ) ( 𝐹𝜀𝑏𝜎𝑇𝑎𝑏𝑠 4 4 ) ( 𝑚 ℎ𝑓𝑔 4 ) 𝛼𝑠𝑞𝑠𝑜𝑙𝑎𝑟−𝜀𝑠𝜎𝑇𝑎𝑏𝑠 𝐹𝜀𝑏𝜎𝑇𝑎𝑏𝑠 (1) where 𝑚 is the water evaporation rate, hfg is the latent heat of vaporization and qsolar is the solar flux. There are three parts to this efficiency: η1 represents the absorber efficiency, η2 incorporates the emitter efficiency 4 as well as the radiative coupling between the solar umbrella and water surface, and η3 represents the fraction of incident radiation that results in evaporation. a c b d Fig 1: Surface heating using a photo-thermal (solar) umbrella. a Schematic of a conventional evaporation pond where incoming sunlight is volumetrically absorbed, causing a bulk water temperature increase that leads to evaporation and conduction losses to the ground. b Rendering of the proposed solar umbrella (spectrally selective absorber and blackbody emitter) that converts incoming sunlight into mid-IR radiation where water is strongly absorbing, thereby increasing the surface temperature and evaporation rate while the bulk remains at a lower temperature. c Schematic of the lab-scale prototype of an evaporation pond comprising water in an acrylic tank separated from the solar umbrella by an air gap and placed under a solar simulator to measure temperatures and mass loss due to evaporation. d Energy balance and modes of heat transfer for the umbrella and water, where the red region represents non- contact heat localization at the surface. Equation (1) can be used to guide the material selection and structure of the solar umbrella. To obtain a high η1, the top surface of the umbrella is coated with a spectrally selective solar absorber; an ideal selective absorber is characterized by a high solar absorbtance, αs, from 0.2 -- 2.5 μm and a near-zero 5 thermal emittance, εs, at wavelengths larger than 2.5 μm.22 This reduces radiation losses to the ambient and achieves efficient conversion of sunlight into heat at an equilibrium temperature of Tabs. To obtain a high η2, the bottom surface of the umbrella is coated with a black emitter characterized by a high IR emittance, εb, which is placed above the water surface with a large view factor, F; σ is the Stefan- Boltzmann constant. Finally, to obtain a high η3, thermal losses from water via conduction, convection and radiation must be minimized. However this efficiency cannot be optimized to a great extent as the surface of the water is exposed to the ambient. To test the effectiveness of the non-contact surface heating approach, we designed a lab-scale prototype that consists of two sub-systems, namely the solar umbrella and a water tank, as shown in Figure 1c. Figure 1d shows the energy balance for the overall system. Lab-scale experimental demonstration In the lab-scale prototype, the solar umbrella is comprised of an aluminum substrate with a selective absorber coating on top (TiNOX, Almeco)23 and a black paint coating (Zynolyte Hi-Temp Paint, Aervoe) on the bottom. The absorber has a solar absorptance of 0.95 and thermal emittance of 0.04 at IR wavelengths that suppresses radiation losses, while the black paint has an emittance of 0.94 at IR wavelengths (Supplementary Information). Since the substrate is thin (0.5 mm) and metallic, temperatures of the absorber and emitter are equal and remain isothermal over the course of the measurements. The tank is made of acrylic and is filled with pure water, with the solar umbrella positioned 4 mm above the water surface. A piece of transparent plastic with a transmittance of 0.91 is used as a convective shield to reduce losses from the top surface of the umbrella as shown in Figure 1c. A solar simulator was used to provide a flux of 1 sun (1000 W m-2) on the top surface of the umbrella. The temperatures of the solar umbrella (Tabs), water surface (Ts, thermocouple placed 2 mm below water surface) and bulk water (Tb, thermocouple placed 40 mm below water surface) were monitored over the course of the experiment. Under illumination, the temperature of the solar umbrella increases rapidly and reaches a steady-state value of Tabs = 70 °C in under three minutes (Supplementary Information). The corresponding increase in temperature of pure water due to radiative heating from the hot emitter is shown in Figure 2a. The water surface temperature 6 steadily rises from 22.5 °C to 40 °C, while the bulk water increases only by 2.5 °C in one hour, which confirms the surface heating effect. To mimic concentrated wastewater in evaporation ponds, the experiment was repeated with a saturated common salt solution (25 wt. % NaCl). Similar trends were observed for the brine, with a 20 °C temperature rise at the surface over one hour (~2 °C higher than with pure water), thereby demonstrating radiative heat localization using the solar umbrella. The presence of salts such as NaCl strongly absorber near-IR radiation which increases the brine temperature compared to pure water. To facilitate a comparison with traditional evaporation ponds where water is illuminated directly by sunlight, the same setup was used without the solar umbrella, and corresponding temperatures are shown in Figure 2b. Due to the small absorption coefficient of water at solar wavelengths, the solar flux is volumetrically absorbed resulting in a 5 °C temperature rise throughout the water in one hour. Given that evaporation varies with surface temperature, the mass change due to evaporation was measured for each case at an ambient temperature, T∞ = 25 °C and a relative humidity of 50%. Figure 2c shows the evaporation rates under one sun after subtracting the evaporation in an otherwise identical but dark environment (0.07 kg m-2 h-1). Without the solar umbrella, pure water absorbs sunlight volumetrically that increases evaporation to 0.3 kg m-2 h-1 compared to the dark case. With the solar umbrella however, the evaporation increases to 0.62 kg m-2 h-1 as a higher surface temperature is achieved; this represents a 2- fold enhancement under the same input solar flux. This evaporation rate and surface temperature of 40 °C under one sun are comparable to other surface heating approaches in literature that use self-supporting floating structures (single layer devices without an insulating foam), and the device demonstrated here has the added advantage of being non-contact.24-26 The saturated NaCl brine showed an evaporation rate of 0.49 kg m-2 h-1, which represents a 21% reduction compared to pure water. This is expected as vapor pressure decreases with salinity but is partially compensated by the higher temperature of brine compared to pure water. Figure 2d compares the evaporation rate over a longer period of time, resulting in precipitation and recovery of salts which confirms that the solar umbrella can be used to concentrate wastewater and achieve zero liquid discharge. The effect of the photo-thermal umbrella on water 7 evaporation was also measured under low optical concentrations of 2, 3 and 5 suns. As expected, higher evaporation rates up to 1.6 kg m-2 h-1 were observed at higher input fluxes (Supplementary Information). We note that these low optical concentrations can be achieved with passive non-tracking concentrators.27 a With solar umbrella (surface heating) b Without solar umbrella (volumetric heating) 32 °C 23 °C c d Fig. 2: Lab-scale experimental results. a Surface and bulk temperature of pure water (solid line) and 25 wt. % NaCl brine (dashed line) over one hour with the solar umbrella showing localized heating at the surface. b Surface and bulk temperatures of pure water without the solar umbrella showing volumetric heating. Inset images are captured using an IR camera after 10 minutes of illumination for visualization of surface vs. volumetric heating with and without the solar umbrella, respectively (IR images are for qualitative observation only; temperature measurements were made with thermocouples as shown in Fig. 1c). c Mass change due to evaporation for pure water under direct illumination (volumetric heating) compared to evaporation of pure water and NaCl brine with the solar umbrella (surface heating). d Evaporation rate over a three-hour period with the solar umbrella for pure water (solid line) and 25 wt. % brine (dashed line) showing a 21% reduction due to salinity. The inset shows precipitated salt crystals from the walls of the water tank owing to evaporation of water from brine to achieve ZLD. All experiments are conducted under one sun (1000 W m-2) at an ambient temperature of 25 °C and relative humidity of 50%. 8 For the case of pure water, these experimentally measured values are used to calculate the sub-system efficiencies from Equation (1): η1 = 91% η2 = 74% and η3 = 63%, resulting in an overall solar-thermal evaporation efficiency of 43% under one sun. This compares favorably with reported single-layer floating evaporation structures20,28 and has the added benefit of being non-contact thus eliminating contamination from the wastewater. The absorber efficiency of 91% can be explained in terms of losses due to reflection and thermal radiation as evidenced from the optical property measurements (Supplementary Information). The emitter efficiency of 74% is due to an emittance and view factor less than unity (εb = 0.94 and F = 0.9) in the lab-scale prototype, as well as convection losses from the umbrella even with a cover. These convective losses can be minimized with an evacuated absorber that is commonly used in domestic solar hot water heaters. Finally, an evaporation efficiency of 63% indicates that there are thermal losses via conduction, convection and radiation in the system that compete with and reduce the energy available for evaporation, as discussed in the next section. Nevertheless, a solar-thermal evaporation efficiency of 43% is achieved in this first demonstration of non-contact surface heating. In comparison, the solar-thermal efficiency of volumetrically heated water under the same solar flux is ~20% based on the experimentally measured evaporation rate (see Figure 2c) and consistent with literature,18 thus confirming the superior performance of surface-based heating for solar evaporation. Thermal model validation To estimate heat losses in the lab-scale prototype and to gain insight into improving performance, a thermal model was developed using COMSOL Multiphysics. Relevant boundary conditions and material properties of the experimental setup were used as inputs (Supplementary Information). The evaporation was modeled as a boundary heat flux: 𝑞𝑒𝑣𝑎𝑝 = ℎ𝑒𝑣𝑎𝑝(𝑇𝑠 − 𝑇∞) = 𝑚 ℎ𝑓𝑔 (2) where qevap is the heat loss due to evaporation characterized by an effective heat transfer coefficient, hevap (which can in general be a function of temperature). Using the experimentally measured 𝑚 and Ts values 9 obtained at different optical concentrations (see Supplementary Information), an average value of hevap = 28 W m-2 K-1 is extracted which matches floating structure prototypes.13 The low value suggests a diffusion-limited process19,29 that is likely to improve when moving from lab scale to deployment scale due to wind. The simulated cross-sectional temperature profile of water in the acrylic tank at steady-state is shown in Figure 3a. Heat localization results in a higher temperature of 42 °C at the surface, while the bulk is close to ambient at 25 °C. Figure 3b shows the transient temperature over an hour, which matches the experimentally measured values, thereby confirming the validity of the thermal model. From this analysis, conduction from the surface into underlying bulk water was calculated to be 21%, convection from the walls of the acrylic tank were 6% and radiation from the acrylic walls were 11%. This leaves 62% of the incident energy on the water surface for evaporation, which is in good agreement with η3 obtained from the experimentally measured mass change due to evaporation. We note here that due to the small size of the lab-scale prototype, side walls of the acrylic tank contribute to a 17% parasitic loss implying that higher evaporation efficiencies are achievable in a larger system. Next, the validated thermal model is used to make performance predictions under different conditions, as shown in Figures 3c and 3d. For instance, on cloudy or winter days, the solar flux can be as low as 300 W m-2 but the umbrella can still reach 45 °C and heat the water surface to 30 °C from an initial temperature of 22 °C to enhance evaporation. Furthermore, by eliminating convective losses from the umbrella surfaces (e.g. evacuated absorber), the steady-state temperature of the solar umbrella can exceed 100 °C under one sun as shown in Figure 3c. This in turn increases the blackbody emission to the water surface, resulting in a temperature rise to over 50 °C as shown in Figure 3d. Given that the vapor pressure of water increases by 70% from 40 to 50 °C, much higher efficiencies and evaporation rates can be obtained using an evacuated umbrella. 10 a c b T (°C) d Fig. 3: Thermal model validation and performance prediction for Solar Umbrella. a Cross-sectional temperature profile of water in an acrylic tank with solar umbrella modeled using COMSOL that shows localized heating to a steady state surface temperature of 42 °C due to a large absorption coefficient (104 m-1) at mid-IR wavelengths. b Water surface and bulk temperatures simulated over one hour (solid lines) and overlaid with experimentally measured temperatures (open symbols) that confirms the validity of the thermal model. c Simulated temperature of the solar umbrella at various solar fluxes indicate that convection losses limit the performance, and d the corresponding simulated water surface temperature predictions show that even at low solar fluxes, the water surface temperature increases and can reach 55 °C if an evacuated system is used. Performance prediction for evaporation ponds Based on the experimentally validated thermal model, the performance of the solar umbrella for real evaporation ponds can be predicted accounting for the diurnal variation of sunlight. The diurnal variation of solar radiation causes a time delay of a few hours between the highest solar intensity and maximum brine temperature in traditional evaporation ponds as sunlight is volumetrically absorbed and used for the 11 sensible heating of water. As a result, the transient response is slow and the evaporation rate is small, thereby requiring large land areas to dispose wastewater. In evaporation ponds having a white salt precipitate layer at the bottom, this is exacerbated due to further losses by reflection and conduction to the ground,30 resulting in even lower solar-thermal evaporation efficiencies. One commonly used approach to address this is the addition of colored organic dyes (e.g., naphthol green and methylene blue) that increase the solar absorption of water during the day. This has been shown to create a ~3 °C increase in brine surface temperature which enhances the evaporation by up to 35%, but at night these dyes lead to lower surface temperatures.30,31 The solar umbrella presented in this work can serve as an alternative to these colored dyes and eliminate the time lag through localizing heat at the evaporation surface. To investigate the effectiveness of this approach, the aforementioned validated COMSOL model was modified to mimic a typical evaporation pond depth7,30 of 0.5 m at an ambient temperature of 20 °C and relative humidity of 50%. The large-scale system surpasses limitations of the lab-scale prototype as the view factor between the umbrella and pond surface is unity (for a real evaporation pond with a large surface area, all the radiation emitted by the solar umbrella reaches the water surface so F = 1 in this case) and parasitic losses (convection and radiation from sides of the acrylic tank) are eliminated. Under 1 sun, the solar umbrella temperature reaches 70 °C, resulting in a thermal emission of ~750 W m-2 to the surface of the pond. The resulting surface temperature (accounting for natural convection, evaporation and radiation) for a complete 24-hour cycle with sunlight during the first eight-hours, is shown in Figure 4a. From these temperatures, the time dependent evaporation rate is predicted using: 𝑚 = ℎ𝑚(𝜌𝑠 − 𝜌∞) where hm is the convective mass transfer coefficient for evaporation, 𝜌𝑠 is the water vapor density at the surface and 𝜌∞is the corresponding density in ambient air. For fixed ambient conditions, 𝜌∞ is known and with an average value of hm = 0.0042 m s-1 (based on the lab-scale experiments and comparable with data on evaporation from a stagnant water surface32), an increase in the surface temperature from 20 to 40 °C in Figure 4a results in a three-fold increase in 𝜌𝑠. This results in a higher evaporation rate as shown in 12 Figure 4a for pure water, where the daily evaporation rate is obtained by integrating the time and temperature-dependent mass flux yielding ~9.8 kg m-2 day-1. To account for the salinity of brines, the evaporation is reduced by 21% to 7.8 kg m-2 day-1. We note that the 21% reduction is based on the experimentally measured evaporation rates for pure water and saturated NaCl in the lab-scale prototype and is not a constant value. However, it has been reported that for a salinity of 25 wt. % and at similar ambient conditions to the one modeled here, this value varies between 18 -- 22%;33 thus a fixed reduction of 21% is valid for this analysis. We also note that in real ponds hm can be higher due to the presence of wind, and other factors such as low humidity in arid regions would also increase evaporation. As such, our model predictions under natural convection and relative humidity of 50% provide a conservative estimate of the performance enhancement. a b Fig. 4: Simulated performance of evaporation ponds for ZLD. a Simulation of water surface temperature and evaporation rate over a diurnal cycle; the shaded region represents sunlight incident on the solar umbrella, resulting in heat localization at the water surface and a corresponding rise in temperature. The evaporation rate is calculated at each temperature, and the brine evaporation rate under these conditions is obtained as a 21% reduction from pure water evaporation. b Daily evaporation enhancement (relative to a traditional volumetrically heated pond under 1 sun) by passive methods including radiation absorbing dyes, solar umbrella and an evacuated solar umbrella at a relative humidity of 50% and ambient temperature of 20 °C. The main advantage is the fast response of the solar umbrella that allows for efficient evaporation during the 8-hour illumination window, which represents over a 100% enhancement compared to volumetrically heated brines under similar ambient conditions over a diurnal cycle.33-36 An evacuated umbrella may be 13 used for further performance enhancement as discussed in Figure 3, and in this case, a daily evaporation of 14.3 kg m-2 day-1 can be obtained (or 11.3 kg m-2 day-1 accounting for brine salinity). This leads to a ~ 160% enhancement in the evaporation rate compared to traditional evaporation ponds, thus surpassing organic dyes and WAIV. In summary, we report a new photo-thermal approach to improve the efficiency of solar evaporation passively without making contact with the water surface. This leverages the inherent absorption properties of water, i.e., a large absorption coefficient at mid-IR wavelengths that causes heat to be localized in a thin layer (<100 μm) at the water's surface. The system comprises a commercially available selective solar absorber, and the emitter is an inexpensive spray on black paint. The solar umbrella with surface heating shows over a 100% enhancement in evaporation compared to volumetric heating that can lower the land requirement, and a 43% solar-thermal efficiency under one sun. The non-contact radiative coupling eliminates risk of fouling and contamination from wastewater streams, making this system uniquely suitable for implementation in brine disposal ponds for achieving zero liquid discharge. Furthermore, by reducing thermal losses (primarily convection from the absorber surface), high temperature vapor or steam generation can be achieved under one sun which has applications in power generation and sterilization. Another potential application of the photo-thermal umbrella is in solar stills where vapor generated can be condensed to produce drinking water for off-grid communities. Methods Materials and characterization: The selective absorber was obtained from Almeco (TiNOXenergy on aluminum) and the black paint was an off-the-shelf spray paint (Zynolyte Hi-Temp Paint, Aervoe) that is stable up to temperatures of 650 °C and was cured at 350 °C for two hours prior to testing. The optical properties of spectrally selective absorber and black paint emitter were measured using an FTIR (Thermo Electron Nicolet 5700), coupled with an integrating sphere accessory (Pike Technologies Mid-IR IntegratIR). As the samples were non-transmitting, the absorptance was calculated as unity minus reflectance. 14 Experimental setup: For lab-scale prototype testing, a solar simulator (Newport, 94081A) with an optical filter for AM 1.5G spectrum is used as the solar input. A power meter and thermopile detector (Newport, 919P-030-18) were used to measure the incoming solar flux at the same location as the water tank. The tank is made out of acrylic and comprises an inner pocket (square cross-section with a side length of 35 mm and a 5 mm depth) that is coated in a reflective foil and filled with water. The inner pocket is surrounded by a 20 mm thick acrylic wall on all sides that minimizes thermal losses from the sides of the tank. Two K-type thermocouples were placed in the tank: one ~2 mm below the water surface to record the surface temperature, Ts, and one 40 mm below the surface to record the bulk water temperature, Tb. The selective absorber-black emitter was mounted on Teflon spacers and placed ~4 mm above the water surface. Another K-type thermocouple was attached to the surface of the emitter to measure its temperature, Tabs, and estimate the blackbody emissive power. A convective shield (plastic with solar transmittance of 0.91) was placed on top of the absorber to reduce convection losses from the surface. The tank was placed on a balance (A&D, GF-4000) that records the mass loss due to evaporation over time. A hygrometer (Onset, HOBO UX100) was kept next to the balance to measure ambient conditions (temperature, T∞ and relative humidity). The K-type thermocouples were connected to a data logger (Pico Technologies, USB TC-08) that records temperature as a function of time. An aperture was placed above the absorber to limit extraneous light during experimental testing. As a first step, evaporation under dark conditions was measured for 30 minutes with the simulator shutter closed. Next, the shutter was opened with the incident flux set to 1000 W m-2, unless specified otherwise. Mass loss and temperature measurements were made for at least two hours, and values reported in the manuscript are averaged over four runs with the dark evaporation subtracted. Acknowledgements This work was supported by the Laboratory Directed Research and Development Program (LDRD) at Lawrence Berkeley National Laboratory under contract # DE-AC02-05CH11231. The authors thank Sishir 15 Mohammed for assistance with thermal modeling. A.K.M acknowledges funding support from the ITRI- Rosenfeld Fellowship. Author Contributions A.K.M. and I.H. contributed equally to this work. The idea of non-contact radiative heating was conceptualized by R.P. and further developed by S.L., S.K. and I.H. A.K.M. and I.H. conducted experiments and developed models. A.K.M., S.K., and R.P wrote the paper. R.P supervised the research. References The Global Risks Report 2018. (World Economic Forum, Geneva, 2018). Grant, S. B. et al. Taking the "Waste" Out of "Wastewater" for Human Water Security and Ecosystem Sustainability. Science 337, 681 (2012). Vörösmarty, C. J. et al. Global threats to human water security and river biodiversity. Nature 467, 555, doi:10.1038/nature09440 1 2 3 4 5 6 7 8 9 10 11 12 13 14 78, Renewable and Energy Reviews https://www.nature.com/articles/nature09440#supplementary-information (2010). Pinto, F. S. & Marques, R. C. Desalination projects economic feasibility: A standardization of cost Sustainable determinants. 904-915, doi:https://doi.org/10.1016/j.rser.2017.05.024 (2017). Gude, V. G. Desalination and sustainability -- An appraisal and current perspective. Water Research 89, 87-106, doi:https://doi.org/10.1016/j.watres.2015.11.012 (2016). Tong, T. & Elimelech, M. The Global Rise of Zero Liquid Discharge for Wastewater Management: Drivers, Technologies, and Future Directions. Environmental Science & Technology 50, 6846-6855, doi:10.1021/acs.est.6b01000 (2016). Morillo, J. et al. Comparative study of brine management technologies for desalination plants. Desalination 336, 32-49, doi:https://doi.org/10.1016/j.desal.2013.12.038 (2014). Giwa, A., Dufour, V., Al Marzooqi, F., Al Kaabi, M. & Hasan, S. W. Brine management methods: Recent 1-23, doi:https://doi.org/10.1016/j.desal.2016.12.008 (2017). Juby, G. et al. Evaluation and Selection of Available Processes for a Zero-Liquid Discharge System. Report No. DWPR No. 149, (U.S. Department of the Interior Bureau of Reclamation, 2008). Mickley, M. Treatment of Concentrate. Report No. DWPR Report No. 155, (U.S. Department of the Interior Bureau of Reclamation, 2008). Ahmed, M., Shayya, W. H., Hoey, D. & Al-Handaly, J. Brine Disposal from Inland Desalination Plants. Water International 27, 194-201, doi:10.1080/02508060208686992 (2002). Hoque, S., Alexander, T. & Gurian, P. L. Innovative Technologies Increase Evaporation Pond Efficiency. IDA Journal of Desalination and Water Reuse 2, 72-78, doi:10.1179/ida.2010.2.1.72 (2010). Ghasemi, H. et al. Solar steam generation by heat localization. Nature Communications 5, 4449, doi:10.1038/ncomms5449 Desalination innovations 407, and current status. https://www.nature.com/articles/ncomms5449#supplementary-information (2014). Tao, P. et al. Solar-driven interfacial evaporation. Nature Energy, doi:10.1038/s41560-018-0260- 7 (2018). 16 & 52, Science Technology Environmental Shi, Y. et al. Solar Evaporator with Controlled Salt Precipitation for Zero Liquid Discharge Desalination. 11822-11830, doi:10.1021/acs.est.8b03300 (2018). Ni, G. et al. A salt-rejecting floating solar still for low-cost desalination. Energy & Environmental Science 11, 1510-1519, doi:10.1039/C8EE00220G (2018). Xu, N. et al. Mushrooms as Efficient Solar Steam-Generation Devices. Advanced Materials 29, 1606762, doi:10.1002/adma.201606762 (2017). Finnerty, C., Zhang, L., Sedlak, D. L., Nelson, K. L. & Mi, B. Synthetic Graphene Oxide Leaf for Solar Desalination with Zero Liquid Discharge. Environmental Science & Technology 51, 11701- 11709, doi:10.1021/acs.est.7b03040 (2017). Ni, G. et al. Steam generation under one sun enabled by a floating structure with thermal concentration. Nature Energy 1, 16126, doi:10.1038/nenergy.2016.126 https://www.nature.com/articles/nenergy2016126#supplementary-information (2016). Bae, K. et al. Flexible thin-film black gold membranes with ultrabroadband plasmonic nanofocusing for efficient solar vapour generation. Nature Communications 6, 10103, doi:10.1038/ncomms10103 https://www.nature.com/articles/ncomms10103#supplementary-information (2015). Segelstein, D. J. The complex refractive index of water, University of Missouri-Kansas City, (1981). Cao, F., McEnaney, K., Chen, G. & Ren, Z. A review of cermet-based spectrally selective solar absorbers. Energy & Environmental Science 7, 1615-1627, doi:10.1039/C3EE43825B (2014). TiNOX energy. Shi, L., Wang, Y., Zhang, L. & Wang, P. Rational design of a bi-layered reduced graphene oxide film on polystyrene foam for solar-driven interfacial water evaporation. Journal of Materials Chemistry A 5, 16212-16219, doi:10.1039/C6TA09810J (2017). Ye, M. et al. Synthesis of Black TiOx Nanoparticles by Mg Reduction of TiO2 Nanocrystals and their Application for Solar Water Evaporation. Advanced Energy Materials 7, 1601811, doi:10.1002/aenm.201601811 (2016). Zhang, L., Tang, B., Wu, J., Li, R. & Wang, P. Hydrophobic Light-to-Heat Conversion Membranes with Self-Healing Ability for Interfacial Solar Heating. Advanced Materials 27, 4889- 4894, doi:10.1002/adma.201502362 (2015). Winston, R. Principles of solar concentrators of a novel design. Solar Energy 16, 89-95, doi:https://doi.org/10.1016/0038-092X(74)90004-8 (1974). Wang, Z. et al. Bio-Inspired Evaporation Through Plasmonic Film of Nanoparticles at the Air -- Water Interface. Small 10, 3234-3239, doi:10.1002/smll.201401071 (2014). Hisatake, K., Tanaka, S. & Aizawa, Y. Evaporation rate of water in a vessel. Journal of Applied Physics 73, 7395-7401, doi:10.1063/1.354031 (1993). Bloch, M. R., Farkas, L. & Spiegler, K. S. Solar Evaporation of Salt Brines. Industrial & Engineering Chemistry 43, 1544-1553, doi:10.1021/ie50499a025 (1951). Gunaji, N. N., Keyes, C. G., New Mexico State, U. & United, S. Disposal of brine by solar evaporation. (U.S. Dept. of the Interior, Office of Saline Water, 1968). Marek, R. & Straub, J. Analysis of the evaporation coefficient and the condensation coefficient of water. 39-53, doi:https://doi.org/10.1016/S0017-9310(00)00086-7 (2001). Harbeck Jr, G. E. The effect of salinity on evaporation. Report No. 272A, (1955). Langbein, W. B. & Harbeck, G. E. Studies of Evaporation. Science 119, 328 (1954). Moore, J. & Runkles, J. R. Evaporation from Brine Solutions under Controlled Laboratory Conditions. Report No. 77, (1968). International and Mass of Heat Transfer Journal 44, 17 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 Turk, L. J. Evaporation of Brine: A Field Study on the Bonneville Salt Flats, Utah. Water Resources Research 6, 1209-1215, doi:10.1029/WR006i004p01209 (1970). 18 Enhanced solar evaporation using a photo-thermal umbrella: towards zero liquid discharge wastewater management Akanksha K. Menon1$, Iwan Haechler1$, Sumanjeet Kaur1, Sean Lubner1, Ravi S. Prasher1,2* Supplementary Information Zero Liquid Discharge Treatment Process Fig. S1: Flow charts showing three different schemes for ZLD using membrane-based and thermal systems. The final step involves the use of either a brine crystallizer or an evaporation pond to extract remaining water and yields a solid waste for disposal. Radiative Heat Localization in Water The absorption of electromagnetic waves (and thus light) in matter is defined by the optical properties of the material through which the wave is travelling. With Beer-Lambert-Bouguer's law, one can determine the intensity of the incident wave at a given depth x: 𝐼(𝑥) = 𝐼0 ∗ 𝑒−𝛼𝑥 where I0 is the incident intensity on the surface of the material, x is the distance from the surface and α is the absorption coefficient. The absorption coefficient is a function of the material properties and also depends on the wavelength of incoming radiation. For wavelengths between 250 -- 800 nm (which comprises 60% of incoming sunlight), the absorption coefficient is very small ~ 0.01 m-1 and water acts as a transparent medium. However, at wavelengths larger than 2 µm (infrared), the absorption coefficient 19 increases by nearly 6 orders of magnitude. Consequently, the penetration depth (intensity reduced by a factor 1/e) decreases by several orders of magnitude and radiation is localized close to the surface. Predominantly IR emission can be achieved using a blackbody with a temperature, T < 150 °C (~99.9% of the thermal radiation lies at wavelengths above 2 µm), where 90-95% of radiation can be absorbed within tens of microns of water. a b Fig. S2: (a) Absorption coefficient of electromagnetic radiation in water with penetration depth of 30 m in visible wavelengths and 20 μm at mid-IR wavelengths. (b) Blackbody emissive power of the sun (T = 5505 °C) that emits in the visible and near-IR, and a blackbody at 100 °C that emits in mid-IR where water is strongly absorbing. 20 Optical Characterization The optical properties of the selective solar absorber and black paint emitter were measured using an FTIR (Thermo Electron Nicolet 5700), coupled with an integrating sphere accessory (Pike Technologies Mid- IR IntegratIR). As the samples were non-transmitting, the absorptance was calculated as unity minus reflectance. a b Fig. S3: (a) Reflectance of the selective absorber (TiNOXenergy, Almeco) and black paint emitter measured using an FTIR. (b) Structure of the solar umbrella comprising a selective solar absorber coated on the top surface of an aluminum substrate and a black paint sprayed on to the bottom surface. Lab-Scale Setup For lab-scale prototype testing, a solar simulator (Newport, 94081A) with an optical filter for AM 1.5G spectrum is used as the solar input. A power meter and thermopile detector (Newport, 919P-030-18) were used to measure the incoming solar flux at the same location as the water tank. The tank is made out of acrylic and comprises an inner pocket (square cross-section with a side length of 35 mm and a 5 mm depth) that is coated in a reflective foil and filled with water (to reflect radiation on the tank walls into water which provides a better representation of a large water body absorbing radiation). The inner pocket is surrounded by a 20 mm thick acrylic wall on all sides that minimizes thermal losses from the sides of the tank. Two K-type thermocouples were placed in the tank: one ~2 mm below the water surface to record the surface temperature, Ts, and one at the bottom of the inner pocket to record the bulk water temperature, Tb. The selective absorber-black emitter was mounted on Teflon spacers and placed ~3-4 mm above the water surface. In this scenario, the air gap between the hot emitter and cooler water surface acts as an insulting layer, thereby eliminating the need for an insulating foam. Another K-type thermocouple was 21 attached to the surface of the emitter to measure its temperature, Tabs, and estimate the blackbody emissive power. A convective shield was placed on top of the absorber to reduce convection losses from the surface. The tank was placed on a balance (A&D, GF-4000) that records the mass loss due to evaporation over time. A hygrometer (Onset, HOBO UX100) was kept next to the balance to measure ambient conditions (temperature, T∞, and relative humidity). The K-type thermocouples were connected to a data logger (Pico Technologies, USB TC-08) that records temperature as a function of time. A 35 mm aperture was placed above the absorber to limit extraneous light during experimental testing. As a first step, evaporation under dark conditions was measured for 30 minutes (solar simulator on with the shutter closed - dark evaporation). Next, the shutter was opened with the incident flux set to 1000 W m-2, unless otherwise noted. Mass loss and temperature measurements were made for two hours under illumination, and values reported in the manuscript are averaged over four runs with the dark evaporation subtracted. Limitations of this prototype include a view factor less than unity (0.9), as well as convection losses from the top surface of the absorber. Fig. S4: Experimental setup for surface heating using a solar umbrella (selective solar absorber and black emitter), where the temperature profile and evaporation rate are monitored using thermocouples and a mass balance, respectively. Holes are drilled into the umbrella to serve as vapor escape pathways. Experimental Data The selective absorber shows a rapid response for conversion of solar energy into heat. Under a solar flux of 1000 W m-2, the absorber reaches a peak operating temperature of 70 ± 2 °C in ~ three minutes owing to the use of a high thermal conductivity substrate (aluminum in this case). Figure S5 also shows the water evaporation rate under dark conditions, under direct illumination corresponding to volumetric heating, and using the solar umbrella corresponding to surface heating. 22 a b Fig. S5: (a) Temperature of the absorber-emitter when exposed to a solar flux showing a fast thermal transient response. (b) Mass change over time due to evaporation under dark conditions, and under one sun illumination with and without the absorber-emitter. Evaporation under Optical Concentration The lab-scale prototype was tested at different optical concentrations of 1, 3 and 5 suns. As expected, at higher solar fluxes, the absorber-emitter temperature increases, which in turn results in a higher emissive power on the water, thereby increasing the surface temperature. As an illustration, at 5000 W m-2, the evaporation rate increases to 1.6 kg m-2 h-1 as the water surface temperature is 60 °C in one hour while the bulk remains at 30 °C. The absorber-emitter can also be used for generating hot vapor as the evaporated water quickly mixes with the air surrounding the hot absorber. A thermocouple was placed 2 mm below the absorber surface to measure the vapor temperature as it escapes from the holes. Significantly higher absorber temperatures can be achieved by using a better convective cover or an evacuated plate that eliminates convection losses from the hot absorber. Non-tracking wide-angle optical concentrators can provide a passive and cost-competitive approach to concentrate sunlight for evaporation. 23 a b 1 3 5 Optical Concentration (# of suns) Fig. S6: (a) Temperatures of the absorber-emitter, water vapor (measured 2 mm below the absorber surface) and the water surface as a function of low optical concentrations. (b) Mass change over time due to evaporation under different solar fluxes. Heat and Mass Transfer Model The overall system comprises two main sub-systems: the solar umbrella (absorber and emitter) and the water tank. The umbrella receives incoming solar flux (Qsolar), and absorbs a majority of it owing to a high solar absorptance of 0.95. As its temperature increases, 5% losses by thermal radiation (Qrad) as well as convection losses at the top surface (Qconv) occur to the ambient. Since the absorber and emitter coatings are thin layers on an aluminum substrate, the structure is isothermal, and the emitter radiates in the IR (Qemit) to the water surface. Performing an energy balance on the umbrella, Qsolar = Qconv + Qrad + Qemit Performing an energy balance on the water surface: Qemit = Qconv,w + Qrad,w + Qcond + Qevap where Qrad,w and Qconv,w are the radiation and convection losses from water respectively (tank walls + water surface), Qcond is the conduction losses from the surface to the underlying bulk water and Qevap is the remaining energy that is available for evaporation. In the lab-scale prototype, due to the large view factor between the umbrella and water surface, radiation emitted by the water surface is largely absorbed by the emitter, which in turn emits back to the water surface. Less than 1% radiation loss occurs between the water surface and ambient, however the water tank walls exchange radiation with the ambient that 24 constitutes thermal losses. Furthermore, since the umbrella surface is hotter than the water surface below, convection losses from the bottom surface of the umbrella and the water are negligible. In this case, heat transfer is by conduction through the air gap which is small compared to the radiation exchange between the two surfaces. However, convection losses from the water tank to the ambient exist and constitute thermal losses. Here: 𝑄𝑟𝑎𝑑,𝑤 = 𝐹𝜀𝜎(𝑇4 − 𝑇∞ 4 ) 𝑄𝑐𝑜𝑛𝑣,𝑤 = ℎ𝑐𝑜𝑛𝑣(𝑇 − 𝑇∞) 𝑄𝑐𝑜𝑛𝑑 = −𝑘 ( 𝐿 𝑇𝑠 − 𝑇𝑏 ) 𝑄𝑒𝑣𝑎𝑝 = 𝑚 ℎ𝑓𝑔 = ℎ𝑒𝑣𝑎𝑝(𝑇𝑠 − 𝑇∞) where F is the view factor, ε is the emittance, T is the temperature of the surface, T∞ is the ambient temperature, hconv is the convective heat transfer coefficient, k is the thermal conductivity of water, Ts is the water surface temperature, Tb is the bulk water temperature and hevap is the evaporation heat transfer coefficient. hevap is calculated as an average using Ts and 𝑚 values at different optical concentrations shown in Fig. S6 to obtain a value of 28 W/m2-K. Fig. S7: Energy balance and heat transfer modes for the solar umbrella and water system. COMSOL Simulation The lab-scale setup is modeled using COMSOL Multiphysics software and the thermal losses are estimated. Water in the acrylic tank is modeled with the following boundary conditions: convection on all side walls = 5 W/m2-K, radiation from all side walls and water surface with emissivity = 0.95, thermal 25 insulation on bottom surface, evaporation coefficient of 28 W/m2-K on water surface and a boundary heat flux of 680 W/m2 on the water surface (this corresponds to the radiation from the emitter at 72 °C under 1 sun accounting for its emissivity and the view factor). The initial water temperature is set at 22.5 °C and the ambient temperature is 25 °C. Spectral radiation in participating media is used to model the high absorption coefficient of water in mid-IR (peak wavelength ~8 μm with absorption coefficient 104 m-1), resulting in heat localization at the surface. The absorber is modeled with a spectrally selective emissivity (absorptance of 0.95 for solar wavelengths and thermal emittance of 0.04 at wavelengths larger than 2.5 μm). The black paint emitter surface is modeled with an emittance of 0.94. For the evaporation pond simulations, the geometry is modified to a surface with 1 m radius and a 0.5 m depth. Owing to the large surface area, convection and radiation losses from the sides of the system are negligible compared to the lab-scale prototype. Furthermore, for a large pond, the view factor between the umbrella and water surface is 1 (i.e., all the radiation leaving the umbrella is intercepted by the water surface) so a boundary heat flux of 750 W/m2 is used at the water surface. Fig. S8: COMSOL simulation of an evaporation pond showing heat localization at the surface when using a photo-thermal umbrella. 26
1905.07303	1	1905	2019-05-17T14:43:12	A Theoretical Perspective on Transient Photovoltage and Charge Extraction Techniques	[ "physics.app-ph" ]	Transient photovoltage (TPV) is a technique frequently used to determine charge carrier lifetimes in thin-film solar cells such as organic, dye sensitized and perovskite solar cells. As this lifetime is often incident light intensity dependent, its relevance to understanding the intrinsic properties of a photoactive material system as a material or device figure of merit has been questioned. To extract complete information on recombination dynamics, the TPV measurements are often performed in conjunction with charge extraction (CE) measurements, employed to determine the photo-generated charge carrier density and thereby the recombination rate constant and its order. In this communication, the underlying theory of TPV and CE is reviewed and expanded. Our theoretical findings are further solidified by numerical simulations and experiments on organic solar cells. We identify regimes of the open-circuit voltage within which accurate lifetimes and carrier densities can be determined with TPV and CE experiments. A wide range of steady-state light intensities is required in performing these experiments in order to identify their 'working dynamic range' from which the recombination kinetics in thin-film solar cells can be determined.	physics.app-ph	physics	A Theoretical Perspective on Transient Photovoltage and Charge Extraction Techniques Oskar J. Sandberg,,† Kristofer Tvingstedt,,§ Paul Meredith,† and Ardalan Armin,† † Sustainable Advanced Materials (Sêr-SAM), Department of Physics, Swansea University, Singleton Park, Swansea SA2 8PP Wales, United Kingdom § Experimental Physics VI, Julius Maximillian University of Würzburg, 97074 Würzburg, Germany Corresponding Authors E-mail: [email protected] E-mail: [email protected] E-mail: [email protected] Abstract Transient photovoltage (TPV) is a technique frequently used to determine charge carrier lifetimes in thin-film solar cells such as organic, dye sensitized and perovskite solar cells. As this lifetime is often incident light intensity dependent, its relevance to understanding the intrinsic properties of a photoactive material system as a material or device figure of merit has been questioned. To extract complete information on recombination dynamics, the TPV measurements are often performed in conjunction with charge extraction (CE) measurements, employed to determine the photo-generated charge carrier density and thereby the recombination rate constant and its order. In this communication, the underlying theory of TPV and CE is reviewed and expanded. Our theoretical findings are further solidified by numerical simulations and experiments on organic solar cells. We identify regimes of the open-circuit voltage within which accurate lifetimes and carrier densities can be determined with TPV and CE experiments. A wide range of steady-state light intensities is required in performing these experiments in order to identify their "working dynamic range" from which the recombination kinetics in thin-film solar cells can be determined. 1. Introduction Thin-film devices based on organic and perovskite semiconductors hold the potential for the use within sustainable energy production1-4 as well as other optoelectronic applications5-9 such as sensors, photodetectors, and light-emitting diodes. In order to further improve and optimize the device performance over both short and long timescales, a better understanding of loss mechanisms such as recombination of the charge carriers is necessary.10-12 For example, various different techniques, both electrical and optical, have been used to investigate the charge carrier recombination in thin-film solar cells. One of the more popular opto-electrical methods used to probe the recombination dynamics in this regard is the small-perturbation transient photovoltage technique (TPV), often measured in combination with charge extraction (CE) or differential charging/transient photocurrent.13-16 In TPV, the device under test is initially held at open-circuit under steady-state illumination.13 To probe the recombination lifetime of the excess charge carriers, i.e. the carriers introduced into the device as a consequence of the steady-state light bias, an additional (weak) short light pulse is applied to the device and the corresponding induced voltage transient, developed over a large load resistance, is measured. The decay of this small photovoltage is then recorded from which the TPV lifetime is extracted. Ideally, this lifetime is associated with the steady-state recombination rate of charge carriers within the bulk. The small light perturbation increases the excess carrier density in the device as 𝑛 = 𝑛oc + 𝛿𝑛, giving rise to a subsequent perturbation in the recombination rate ℛ = ℛ0 + 𝛿ℛ; here, 𝑛oc is the photo-generated steady- state (dc) carrier density and ℛ0 is the associated (time-independent) recombination rate, which is balanced by an equal steady-state photo-generation rate 𝐺0 of charge carriers at open-circuit. The decay of the voltage transient, induced as a result of the small perturbation, is approximately given by13 𝑑Δ𝑉(𝑡) 𝑑𝑡 ∝ 𝑑𝑛(𝑡) 𝑑𝑡 = −δℛ = − δ𝑛 𝜏𝐵 (1) leading to a first-order kinetics of Δ𝑉(𝑡) ∝ exp(− 𝑡 𝜏𝐵⁄ ), where 𝜏𝐵 is the associated pseudo- first-order bulk recombination lifetime. In general, the open-circuit voltage is related to the steady-state light intensity or photo- generation rate via 𝑉oc ∝ (𝑛id𝑘𝑇 𝑞⁄ ) ln(𝐺0), where 𝑛id is the diode ideality factor and 𝑘𝑇 𝑞⁄ is the thermal voltage.17-19 Since the open circuit voltage is logarithmically dependent on the steady-state (dc) photo-generated carrier density, the recombination lifetime from TPV measurements is parametrically dependent on the 𝑉oc, taking the form13,19 𝜏𝐵 = 𝜏0 exp (− 𝑞𝑉𝑜𝑐 𝜈𝑘𝑇 ) (2) where 𝜈 is the slope parameter that ideally depends only on the order of the steady-state recombination dynamics within the bulk, while 𝜏0 is a voltage-independent prefactor. For example, in the case of second-order bulk recombination ℛ = 𝛽𝑛2, the carrier recombination dynamics after the perturbation pulse is given by 𝑑𝑛 𝑑𝑡 = 𝐺0 − 𝛽(𝑛oc + δ𝑛)2 ≈ − δ𝑛 𝜏𝐵 (3) For δ𝑛 ≪ 𝑛oc the bulk recombination lifetime is thus 𝜏𝐵 = 1 2𝛽𝑛oc = 1 2√𝛽𝐺0 (4) Subsequently, with 𝑛id = 1, the open-circuit voltage dependence of Eq. (2) corresponds to 𝜈 = 2 in this case. Here, 𝛽 is the second-order recombination coefficient and 𝑛oc = √𝐺0 𝛽⁄ (since 2 at open-circuit conditions). This can be compared to the first order recombination 𝐺0 = 𝛽𝑛𝑜𝑐 dynamics scenario, occurring for example in heavily doped solar cells under lower illumination conditions where excess carrier density is smaller than the doping level. Then, 𝜈 = ∞ and the lifetime is constant, i.e. independent of the concentration of photo-generated carriers. To obtain the recombination rate and its kinetics from the TPV lifetime at different open-circuit voltages (i.e. different steady-state light intensities), the corresponding steady-state charge carrier density needs to be known as well. However, the exact relation between the steady-state carrier density 𝑛oc (at open-circuit) and the open-circuit voltage is unknown due to the system- specific voltage dependence of the recombination losses. As such, it is never possible to directly convert a 𝑉𝑜𝑐 value to a charge carrier density, but the relationship needs instead to be measured as accurately as possible. To probe the carrier lifetime as a function of the photo- induced carrier density in the active layer, the transient photovoltage technique is therefore typically performed in combination with charge extraction (CE) or differential charging/transient photocurrent measurements. The ability of both of these methods to correctly determine bulk carrier concentrations is heavily debated, both in the organic and perovskite solar cell community, but are both frequently used to assign the carrier density as a function of open-circuit voltage. The extracted carrier density under open-circuit conditions is generally taken to follow an exponential open-circuit voltage dependence according to: 𝑛CE = 𝑛𝐶𝐸,0 exp ( 𝑞𝑉𝑜𝑐 𝑚𝑘𝑇 ) (5) where 𝑚 is the important (and unknown) associated slope parameter, and 𝑛𝐶𝐸,0 is a prefactor.19 Measuring CE in conjunction with TPV allows for the recombination rate 𝑅 ∼ 𝑛CE 𝜏𝑇𝑃𝑉 ⁄ to be mapped as a function of carrier density, providing the sought-after information concerning the dominating recombination mechanism. In accordance with Eq. (1), one expects the lifetime extracted from the TPV experiment to reflect the recombination of photo-induced carriers. However, this interpretation has been questioned by Street who suggested that the extracted TPV lifetimes in organic solar cells is set by the device capacitance and given by20,21 𝜏 = 𝐶 [ 𝜕𝐽dark 𝜕𝑉 −1 ] ≈ 𝑛id𝑘𝑇𝐶 𝑞𝐽dark(𝑉oc) (6) where 𝐽dark is the dark steady-state current density and 𝐶 is the capacitance of the active layer. This "lifetime" can therefore be interpreted as an internal discharging time of the diode itself rather than reflecting the actual recombination dynamics of photo-induced carriers. Similar concerns on lifetime assignments obtained via voltage transient methods were raised for silicon solar cells in 1981 by Mahan and Barnes and analytically explained in more detail by Castener.22,23 Recently, Kiermasch et al. thus proposed the TPV lifetimes to instead be given by the sum of the real recombination lifetime 𝜏𝐵 of photo-induced carriers and a capacitive contribution24 𝜏total = 𝜏𝐵 + 𝑛id𝑘𝑇𝐶 𝑞𝐽0 exp (− 𝑞𝑉oc 𝑛id𝑘𝑇 ) (7) where 𝐽0 is the dark saturation current density of the device. The capacitive contribution, corresponding to the second term on the right-hand side of Eq. (7) and dominating at lower voltages, is equivalent to the internal diode discharging lifetime suggested by Street. In the same work, Kiermasch and co-workers also found the charge density (Eq. (5)), as extracted from CE, to be strongly influenced by capacitive effects, in particular at low voltages, leading to unrealistically large values for the slope parameter 𝑚 at these intensities.24 As a consequence, using the charge carrier density obtained from CE in combination with the lifetimes extracted from TPV to probe the underlying recombination reaction order might often result in recombination orders > 2, which do not necessarily reflect the recombination of steady-state photo-induced carriers within the bulk. These type of higher recombination orders have usually been explained in terms of trap-assisted recombination via localized tail states, both in organic25-27 and perovskite28 solar cells. However, it has also been pointed out that higher recombination orders might be artefacts caused by non-uniform carrier profiles in thin devices.19,29 Given this background, and the recent insights of Kiermasch et al., in this work we further elaborate the limitations of TPV and CE and their physical meaning, from a theoretical view point supported by numerical calculations and demonstrative experiments. Starting from fundamental electrical transient current theory in the time-domain, we derive analytical expressions for the TPV lifetime and extracted charge carrier densities from CE. The validity of the analysis is confirmed with numerical drift-diffusion simulations on organic thin-film solar cell devices. 𝑉(𝑡) = 𝑉ext(𝑡) − 𝑗(𝑡)ሾ𝑅𝑠 + 𝑅𝐿ሿ 𝑅𝑠 𝑉ext(𝑡) 𝑅𝑠ℎ 𝑅𝐿 𝑉𝐿(𝑡) = −𝑗(𝑡)𝑅𝐿 𝑗(𝑡) Figure 1. The general schematic set up for electrical transient measurements, such as TPV and CE, on diode or solar cell devices. The output voltage 𝑉𝐿 (or current 𝑗) is measured across the load resistance 𝑅𝐿 (usually via an oscilloscope). Note that in order to establish open-circuit conditions, the external voltage source is in general not needed, as long as a large enough load resistance is used; the case without the external voltage source corresponds to 𝑉ext = 0. 2. Theory A schematic picture of the electrical set up used for transient measurements such as TPV and CE is shown in Figure 1. In general, for a planar current flow across the (non-ideal) diode or solar cell device, the transient current density is given by30,31 𝑗(𝑡) = 𝐽c(𝑡) + 𝜀𝜀0 𝑑 𝜕𝑉 𝜕𝑡 (8) where 𝐽c(𝑡) ≡ (1 𝑑⁄ ) ∫ 𝐽c(𝑥, 𝑡) 𝑑 0 𝑑𝑥 is the average conduction current density in the active layer and the second term on the right-hand side is the corresponding average displacement current density, with 𝑉(𝑡) being the voltage across the active layer: 𝑉(𝑡) = 𝑉ext(𝑡) − ሾ𝑅L + 𝑅sሿ𝑗(𝑡) (9) where 𝑉ext(𝑡) is the external applied voltage (by a voltage source), 𝑅L is the load resistance (input terminal to the oscilloscope), and 𝑅s is the total series resistance of the electrodes, the external wires, and the internal resistance of the voltage source. In the set up shown in Figure 1, the transient voltage or current is obtained by measuring the voltage drop 𝑉𝐿(𝑡) = −𝑗(𝑡)𝑅L across the load. Note that, under steady-state illumination, open-circuit conditions can either be established by using a large 𝑅L (𝑅L → ∞) and/or by applying an external voltage 𝑉ext equal to the corresponding open-circuit voltage. In conjunction with the Poisson and the charge carrier continuity equations, the average conduction current can generally be expressed as32 𝐽c(𝑡) = 𝑞𝑑 ∂𝑛eff 𝜕𝑡 + 𝐽𝐷(𝑡) (10) as shown in the Supporting Information, with 𝑛eff being an effective charge carrier density in the device, 𝑛eff ≡ 1 2 ሾ𝑝 + 𝑛ሿ + ′ 𝜎el 𝑞𝑑 (11) where 𝑝 and 𝑛 are the spatial averages of the total hole and electron carrier densities across the active layer, respectively, while ′ = 𝜎el 1 𝑑 𝑑 ∫ (𝑥 − 0 𝑑 2 ) 𝜌(𝑥, 𝑡) 𝑑𝑥 (12) where 𝜌(𝑥, 𝑡) is the total space charge density in the active layer. 𝜎el ′ can be interpreted as an additional electrode charge induced by spatially non-uniform space charge distributions inside the active layer. Finally, 𝐽D is the sum of the net recombination-generation current density and the leakage current due to parasitic shunt resistance 𝑅sh (in Ωm2); 𝐽𝐷(𝑡) = 𝑞 ∫ ሾℛ𝐵(𝑥, 𝑡) − 𝐺(𝑥, 𝑡)ሿ 𝑑 0 𝑑𝑥 + 𝐽surf(𝑡) + 𝑉(𝑡) 𝑅sh (13) where 𝐺(𝑥, 𝑡) is the photo-generation rate of free charge carriers, ℛ𝐵(𝑥, 𝑡) is the carrier recombination rate in the bulk, and 𝐽surf(𝑡) = 𝐽n(0, 𝑡) + 𝐽p(𝑑, 𝑡) is the total surface recombination current of electrons at the anode (𝑥 = 0) and holes at the cathode (𝑥 = 𝑑); 𝐽n p⁄ (𝑥, 𝑡) is the local electron/hole current density at the position 𝑥. We note that, in general, 𝐽𝐷 (or 𝐽surf) also includes a contribution from Ohmic conduction currents (e.g. in case of a doped layer). The above set of equations generally describes planar (one-dimensional) electrical current transients in sandwich-type devices. It should be noted that under dc conditions (∂ 𝜕𝑡⁄ = 0), the total current density reduces to 𝐽𝐷, taking its steady-state value. In the dark, this current density can usually be approximated as 𝐽Ddark = 𝐽0 [exp ( 𝑞𝑉 𝑛id𝑘𝑇 ) − 1] + 𝑉 𝑅sh (14) which is the (non-ideal) generalized Shockley diode equation. Here, the applied voltage across the diode under dc conditions is given by 𝑉 = 𝑉ext − 𝐽D𝑅s,tot, with 𝑅𝑠,tot = 𝑅L + 𝑅s being the total series resistance of the external circuit (see Figure 1). Note, however, that in practice, the dark dc current is always measured without a separate load resistor so that 𝑉 = 𝑉ext − 𝐽D𝑅s in Eq. (14). In the following, the general theory [Eqs. (8) to (13)] is applied to TPV and CE relevant conditions. (a) δ𝐺 𝐺(𝑡) 𝐺0 (b) Δ𝑉(𝑡) 𝑉(𝑡) 𝑉𝑜𝑐 𝑡 = 0 𝑡 𝑡 = 0 (c) 𝐺(𝑡), 𝑉(𝑡) 𝑗(𝑡) 𝐺0, 𝑉𝑜𝑐 𝑡 = 0 𝑡 𝑡 = 0 (d) 𝑄𝑒𝑥𝑡𝑟 𝑡 𝑡 Figure 2. Schematic picture of TPV and CE. In TPV, a perturbation light pulse is applied under open-circuit voltage (a), giving rise to an induced voltage transient across the device (b). From the exponential decay of this transient, the TPV lifetime is extracted. In CE, the device initially held under illumination at dc open-circuit conditions is short-circuited with the light simultaneously switched off (c), this give rise to an extraction current transient flowing through the external circuit (d). The extracted charge is obtained by integrating the current transient response. 2.1. Small-Perturbation Transient Photovoltage In TPV, a large load resistance (𝑅L ≫ 𝑅s) is connected in series with the device. This is to maintain open-circuit conditions during the perturbation light pulse; in practice, a load resistance of 1 MΩ is commonly used, although it is preferable to use much larger resistances to avoid shunting over the measurement load at low voltages.33 From the (ideally) exponential decay of the voltage transient, the associated small perturbation TPV lifetime is determined. A schematic picture is illustrated in Figure 2a and b. Prior to the light perturbation, we have 𝑉 = 𝑉ext − 𝑗𝑅L = 𝑉ext + 𝑉𝐿,0 = 𝑉𝑜𝑐, where 𝑉oc is the steady-state (time-independent) open-circuit voltage due to the background light and 𝑉𝐿,0 is the corresponding steady-state voltage drop across the load. In this case, the only function of applying a fixed external voltage 𝑉ext is to set the value for 𝑉𝐿,0, leaving the voltage 𝑉oc across the active layer unchanged. For example, if 𝑉ext = 0 (i.e. no voltage source), then 𝑉𝐿,0 = 𝑉𝑜𝑐; conversely, if 𝑉ext = 𝑉𝑜𝑐, the steady-state voltage across the load is zero (𝑉𝐿,0 = 0). Following the light perturbation, the photo-induced voltage transient over the active layer is given by 𝑉(𝑡) = 𝑉oc + Δ𝑉(𝑡), where Δ𝑉(𝑡) = 𝑉𝐿(𝑡) − 𝑉𝐿,0 is the small additional voltage drop induced over the load. Inserting Eq. (9) and (10) into Eq. (8), and rewriting the current equation in terms of Δ𝑉(𝑡), then reveals 𝐶0 𝜕Δ𝑉 𝜕𝑡 + Δ𝑉(𝑡) 𝑅L = −𝑞𝑑 ∂𝑛eff 𝜕𝑡 − 𝐽𝐷 (15) where 𝐶0 = 𝜀𝜀0 𝑑⁄ is the geometric capacitance per unit area of the active layer. Provided that the perturbation is small, also the change in the current 𝐽D will be small and 𝐽D(𝑉oc + Δ𝑉) ≈ 𝜕𝐽D(𝑉oc) 𝜕𝑉 Δ𝑉, (16) since 𝐽𝐷(𝑉𝑜𝑐) = 0 (open-circuit condition). Making use of the chain rule, Eq. (15) can then be re-expressed as: 𝜕Δ𝑉 𝜕𝑡 = − [ 𝐶0+𝑞𝑑 1 𝑅L + 𝜕𝐽D(𝑉oc) 𝜕𝑉 𝜕𝑛eff(𝑉oc) 𝜕𝑉 ] Δ𝑉(𝑡) ≡ − Δ𝑉(𝑡) 𝜏TPV (17) where 𝜏TPV is the associated pseudo-first-order lifetime of the small-perturbation transient photovoltage decay. The change in the effective charge density can be related to the recombination rate via 𝛿𝑛eff = ሾ𝜕𝑛eff 𝜕ℛ⁄ is the total recombination rate, averaged over the ሿ𝛿ℛ, where ℛ = ℛ𝐵 + 𝐽surf 𝑞𝑑⁄ active layer, incorporating both surface and bulk recombination; ℛ𝐵 = (1 𝑑⁄ ) ∫ ℛ𝐵(𝑥) 𝑑 0 𝑑𝑥 is the average carrier recombination rate inside the bulk. On the other hand, making use of Eq. (13), we see that 𝑞𝑑𝛿ℛ = 𝑞𝑑ሾ𝜕ℛ 𝜕𝑉⁄ ሿ𝛿𝑉. Concomitantly, the TPV ሿ𝛿𝑉 = ሾ𝜕𝐽D 𝜕𝑉⁄ − 1 𝑅𝑠ℎ⁄ lifetime in Eq. (17) is equivalent to 𝜏TPV = (𝜏B,eff + 𝑅0𝐶0) [1 + ( 1 𝑅𝑠ℎ + 1 𝑅L −1 ) 𝑅0] (18) where 𝜏B,eff = ሾ𝜕ℛ 𝜕𝑛eff ⁄ ሿ−1 is an effective carrier recombination lifetime, and 𝑅0 = [ 𝜕𝐽D(𝑉oc) 𝜕𝑉 −1 − 1 𝑅sh ] ≈ 𝑛id𝑘𝑇 𝑞𝐽0 exp (− 𝑞𝑉oc 𝑛id𝑘𝑇 ) (19) is the differential resistance of the diode itself (excluding the shunt resistance), where Eq. (14) was used in the last step. Note that the shunt resistance 𝑅𝑠ℎ plays an identical role to the load resistance 𝑅𝐿 in Eq. (18), consistent with previous experimental findings.24,33 In the limit when the electron and hole densities are homogenous across the active layer, corresponding to sufficiently high steady-state light intensities, we obtain 𝜏B,eff = 𝜕ሾ(𝑝 + 𝑛) 2⁄ ሿ 𝜕ℛ⁄ , becoming identical to 𝜏B in Eq. (1) when 𝑝 = 𝑛. Under these conditions and provided that 𝜏B,eff ≫ 𝑅0𝐶0 (and 𝑅0 ≪ 𝑅𝐿, 𝑅𝑠ℎ), 𝜏TPV thus embodies a real carrier lifetime which is also relevant under steady-state conditions. Note that this lifetime is only a constant in the case of true first order recombination, but the general assumption is that 𝜏B is allowed to be dependent on the carrier concentration and should thus represent recombination dynamics of any order. We also note that for the case 𝑛eff = 𝑝 = 𝑛, Eq. (18) becomes equivalent to the TPV lifetime expression previously proposed by Credgington et al.34,35, and can be simplified as 𝜏TPV = ሾ1 + 𝐶0 (𝜕ሾ𝑞𝑛𝑑ሿ 𝜕𝑉⁄ ⁄ ) ሿ𝜏B when external resistance effects are negligible. At lower light intensities, however, it is expected that the carrier distributions inside the active layer (eventually) becomes very inhomogeneous. Under these conditions (𝑛eff ≠ 𝑛, 𝑝), the effective lifetime 𝜏B,eff = ሾ𝜕ℛ 𝜕𝑛eff ሿ−1 instead reflects how the total effective carrier density ⁄ 𝑛eff changes with the steady-state recombination current. In the limit of small light intensities, the distribution of excess carriers is almost fully dominated by the distribution of back-injected dark carriers. In this case, the TPV lifetime in Eq. (17) can instead be approximated by 𝜏cap = [ where 𝐶 𝜕𝐽D(𝑉oc) + 𝜕𝑉 = [ 1 𝑅0 + 1 𝑅𝑠ℎ + 1 𝑅𝐿 ] −1 𝐶 1 𝑅L ] 𝐶 = 𝐶0 + 𝑞𝑑 𝜕 𝜕𝑉 ሾ𝑛effሿ (20) (21) is given by the 𝑑𝑎𝑟𝑘 capacitance of the active layer, assuming 𝑛eff to be governed by the dark back-injected carriers. In this limit, TPV is limited by a composite RC-time constant of the device, being ruled by the smallest of the load resistance, the shunt resistance, and the differential resistance of the diode itself. We note that the RC decay in the limit of low light intensities is not fully determined by the geometric capacitance 𝐶0, but rather by a total capacitance 𝐶, as given by Eq. (21). This total capacitance is given by the sum of 𝐶0 and a space charge term that monitors the change in the background charge carrier distribution (as induced by the voltage perturbation) during the voltage transient. Only when the change in the space charge distribution is negligible, will 𝐶 be given by the geometric capacitance 𝐶0. However, for the case of a doped active layer, for example, the photo-induced voltage change across the active layer will induce a small change in the (doping-induced) depletion layer thickness 𝑤 as well, and Eq. (20) reduces in this case to the depletion layer capacitance 𝐶 = 𝜀𝜀0 𝑤⁄ , as shown in the Supporting Information. 2.2. Charge Extraction A schematic picture of the CE method is shown in Figure 2c and d. The device is initially held under illumination at open-circuit conditions. The device is then short-circuited (or reverse- biased) with the illumination switched off simultaneously, and the induced extraction current transient is measured. In contrast to TPV, the load resistance in CE needs to be as small as possible to avoid resistive losses; in the following, we assume the total series resistance 𝑅s,tot = 𝑅L + 𝑅s to be negligibly small (i.e. 𝑗𝑅s,tot ≪ 𝑉oc). The extracted charge, obtained by integrating the conduction current and correcting for the geometric capacitive electrode charge, is then given by 𝑄CE ≡ − [∫ 𝑗(𝑡) 𝑡∞ 0 𝑑𝑡 − 𝐶0𝑉oc] = − ∫ 𝐽c(𝑡) 𝑡∞ 0 𝑑𝑡 (22) where 𝐶0𝑉oc corresponds to the charge stored on the electrodes and 𝑡∞ is the time when all excess charge carriers have been collected (and/or recombined), so that 𝐽c(𝑡∞) = 0. Inserting Eq. (10) into Eq. (22) and integrating, we obtain 𝑄CE = 𝑞𝑛CE𝑑 with the associated extracted carrier density 𝑛CE given by 𝑛CE = Δ𝑛eff − ∫ ℛ 𝑡∞ 0 𝑑𝑡 (23) where Δ𝑛eff = 𝑛eff(𝑡 = 0) − 𝑛eff(𝑡 = 𝑡∞) and assuming shunt effects to be negligible. When interpreting charge extraction experiments, the recombination between charge carriers during the charge extraction process of the excess carriers is generally assumed to be negligible (complete charge extraction),36 ℛ = 0. With this assumption, after making use of Eq. (11) and Eq. (12), Eq. (23) can be expressed as 𝑛CE = 1 𝑑 𝑑 ∫ [ 0 𝑥 𝑑 Δ𝑝(𝑥) + (1 − 𝑥 𝑑 ) Δ𝑛(𝑥)] 𝑑𝑥 (24) assuming that Δ𝜌(𝑥) = 𝑞ሾΔ𝑝(𝑥) − Δ𝑛(𝑥)ሿ, where Δ𝑝(𝑥) = 𝑝(𝑥, 𝑡 = 0) − 𝑝(𝑥, 𝑡 = 𝑡∞) and Δ𝑛(𝑥) = 𝑛(𝑥, 𝑡 = 0) − 𝑛(𝑥, 𝑡 = 𝑡∞) are the corresponding excess hole and electron densities, respectively. Concomitantly, even in the ideal case of complete charge extraction, the extracted charge carrier density is, in general, not given by the spatial average of the extracted electron and hole densities within the active layer. Therefore, in order to relate the extracted carrier density 𝑛CE to the steady-state carrier density under open-circuit conditions, the excess carrier distributions are also assumed to be homogenous and uniform throughout the active layer (i.e. Δ𝜎𝑒𝑙 ′ = 0). With this additional (implicit) assumption, Eq. (24) is then finally simplified as 𝑛CE = ሾΔ𝑝+Δ𝑛ሿ 2 = 𝑛oc (25) where Δ𝑝 = Δ𝑛 = 𝑛oc is the excess steady-state carrier density at open-circuit conditions. Note that in case of pure second-order recombination, we then expect 𝑛CE = 𝑛𝑜𝑐 = √𝐺0/𝛽, as before. Figure 3. (a) The simulated current-voltage characteristics of an organic solar cell under consideration at 1 sun. In the inset, the open-circuit voltage is simulated as a function of the steady-state light intensity 𝐺0. In (b), a photo- induced voltage transient, under open-circuit conditions corresponding to 1 sun steady-state light intensity, is simulated. The inset depicts the voltage transient on a log-lin scale, with the related exponential fit indicated by the dashed line. 3. Comparison with Drift-Diffusion Simulations To gain further insights into the physical meaning of the above theoretical analysis, we turn to transient device simulations based on a 1D drift-diffusion model.37-39 The device model numerically solves Eq. (8) and (9), in conjunction with the Poisson equation and the time- dependent carrier continuity equations, assuming 𝐽c(𝑥, 𝑡) = 𝐽𝑝(𝑥, 𝑡) + 𝐽𝑛(𝑥, 𝑡) to be given by the drift-diffusion relations.40,41 In our simulations, an optically thin device with uniform photo- generation profiles (thus neglecting interference effects) is considered. The active layer is assumed undoped and have a thickness of 𝑑 = 100 nm; moreover, balanced charge carrier mobilities of 10−4 cm2/Vs are assumed. The recombination within the bulk is taken to be of a pure second-order type (ℛ = ℛ𝐵 = 𝛽𝑛𝑝), with a second-order recombination coefficient given by 𝛽 = 1.2 × 10−11 cm3/s. The electrical donor-acceptor bandgap of the active layer is set to 1.2 eV, whereas perfectly selective Ohmic contacts are assumed (𝐽𝑛(0) = 𝐽𝑝(𝑑) = 0), thus neglecting surface recombination effects42. The details of the simulations are given in the Supporting Information. Figure 3a depicts the simulated current-voltage characteristics at 1 sun incident light intensity for the solar cell device considered, with pure second-order (bimolecular) recombination in the bulk being the dominant recombination mechanism. Ideally, we expect that 𝑛CE = Δ𝑛eff = √𝐺0 𝛽⁄ and 𝜏𝑇𝑃𝑉 = 𝜏𝐵 = 1 (2√𝛽𝐺0) ⁄ , assuming equal and homogenous carrier distributions (𝑛 = 𝑝) under steady-state open-circuit conditions. In terms of the steady-state open-circuit voltage dependence, these ideal conditions are manifested by slope parameters of 𝑚 = 2 and 𝜈 = 2 for the extracted CE carrier density and TPV lifetime, respectively. These ideal conditions will accordingly also render the steady state parameter of the ideality factor 𝑛𝑖𝑑 equal to unity [see inset in Figure 3a]. In the following, the CE and TPV measurements are simulated and compared to the idealized case. 3.1. TPV - Capacitive Effects vs. Bulk Recombination Figure 3b depicts a simulated TPV transient under 1 sun steady-state light intensity. The lifetime 𝜏TPV is extracted from the exponential decay of the simulated TPV transients (see inset of Figure 3b). The steady-state (dc) open-circuit voltage 𝑉𝑜𝑐 is varied by varying the steady- state carrier photo-generation rate 𝐺0; the simulated 𝑉𝑜𝑐 as a function of 𝐺0 is shown in the inset of Figure 3a. For simplicity the impact of the shunt resistance is assumed negligible, 𝑅𝑠ℎ → ∞. In Figure 4a, the TPV lifetime 𝜏TPV as a function of the steady-state open-circuit voltage 𝑉oc is simulated for a load resistance of 1 GΩ, assuming a device area of 𝑆 = 0.1 cm2 (so that 𝑅L = 104 Ωm2). In accordance with Eq. (18), we expect the TPV lifetime to be determined by the steady-state values of 𝜕𝑛eff 𝜕ℛ⁄ at 𝑉 = 𝑉𝑜𝑐. Indeed, upon comparing the TPV and 𝜕𝐽𝐷 𝜕𝑉⁄ lifetime extracted from the transient simulations (symbols) with Eq. (18), obtained from steady- state simulations (solid line), an excellent agreement is obtained. Furthermore, it can be seen that only at high light intensities (i.e. high 𝑉oc), when the carrier profiles inside the active layer are sufficiently uniform, is the TPV lifetime accurately given by the bulk lifetime Eq. (4), as depicted by the dashed red line with the characteristic slope parameter of 𝜈 = 2. At smaller light intensities (lower 𝑉oc), the TPV lifetime is governed by capacitive effects associated with back-injected dark carriers. In this regime, the TPV lifetime is very well approximated by Eq. (20) (depicted by blue short-dashed line in Figure 4a), assuming 𝐶 to be given by the dark dc capacitance (Figure S1). We note that this capacitance exhibits a voltage dependence under forward bias, increasing from 𝐶 ≈ 1.1𝐶0 at 𝑉 = 0 to 𝐶 ≈ 1.4𝐶0 at 𝑉 = 0.5 V, as can be seen from Figure S1. The increase in 𝐶, relative to 𝐶0, can be understood in terms of back-injected dark carriers at the contacts penetrating deeper into the active layer, effectively decreasing the active layer thickness (in a geometric capacitor approximation). This capacitance is sometimes also referred to as the chemical or diffusion capacitance.43-45 It should be stressed that TPV is ultimately limited at lower voltages by the RC-times of the load resistance and/or the shunt resistance of the cell. The smaller of these two will set the RC- time constant of the system and defines the upper limit for the lifetime that can be extracted by the measurement. This is demonstrated in Figure 4b in case of a 1 MΩ load resistance (assuming 𝑅L ≪ 𝑅𝑠ℎ); as 𝜏TPV approaches the RC-limit set by the load resistance, a saturation of the extracted lifetime 𝜏TPV → 𝑅L𝐶 will occur at low voltages, in accordance with both Eq. (18) and Eq. (20). Based on these findings, it is only the TPV lifetimes at the highest intensities which is governed by recombination between photo-generated carriers inside the active layer, whereas the TPV lifetime at lower intensities is dominated by capacitive effects associated with spatially non- uniform charge carrier distributions within the bulk. Therefore, depending on the value of the shunt/load resistor, the TPV lifetimes at the lowest intensities are ruled only by the associated RC-time constant. Subsequently, a necessary requirement in TPV for the determination of the recombination lifetime is that 𝑅0𝐶0 ≪ 𝜏𝑇𝑃𝑉 ≪ ሾ(1 𝑅𝐿⁄ ) + (1 𝑅𝑠ℎ⁄ )ሿ−1𝐶. Figure 4. The TPV lifetime is shown as a function of the dc open-circuit voltage (steady-state light intensity) for two different load resistances: (a) 1 GΩ and (b) 1 MΩ. The lifetimes, as extracted from the simulated TPV transients (see inset of Figure 3b), are indicated by symbols, whereas Eq. (18) (derived from steady-state simulations) is depicted by the black solid line. The bulk lifetime (𝜏B) [Eq. (4)] and the capacitive RC (𝜏cap) [Eq. (20)] limits are indicated by the red dashed line and dotted blue line, respectively. A device area of 𝑆 = 0.1 cm2 is assumed. 3.2. The Validity of Complete Extraction and Uniform Carrier Distributions in CE We next turn to charge extraction. The CE carrier density, shown in Figure 5, is given by 𝑛CE = 𝑄CE/𝑞𝑑, with 𝑄CE obtained by integrating the simulated extraction current transients and correcting for the geometric capacitance (see Eq. (22)). In Equation (25), it was assumed that (i) the recombination during the extraction process is negligibly small, and (ii) the carrier distributions are homogenous. To check the validity of these two assumptions, we have simulated CE current transients for the device considered in Figure 4a. The corresponding extracted carrier density 𝑛CE is shown in Figure 5a. This is to be compared to the excess carrier density 𝑛oc = √𝐺0/𝛽, as expected from Eq. (25), indicated by the red line (with a slope parameter of 𝑚 = 2). It can be seen, for the case 𝜇 = 10−4 cm2/Vs, that Eq. (25) overestimates the actual 𝑛CE at higher 𝑉oc, but underestimates it at low 𝑉oc. Figure 6 depicts the corresponding excess carrier densities Δ𝑝(𝑥) and Δ𝑛(𝑥), defined as the difference between the local carrier density at dc open-circuit conditions under illumination (𝑡 = 0) and in the dark (𝑡 = ∞), at a high and a low light intensity. At high intensities, shown in Figure 6a, the excess carrier profiles are indeed closely uniform inside the active layer, with the carrier density given by 𝑛oc = √𝐺0/𝛽. This suggests that assumption (ii) may be considered valid at high intensities. To check the validity of assumption (i), we compare the transient simulations (symbols) with Eq. (24) (solid lines), corresponding to the theoretical 𝑛𝐶𝐸 expected in the case of complete charge extraction (as calculated from steady-state simulations) in Figure 5a. It can be seen that excellent agreement is obtained at small and moderate light intensities. Only at high intensities does a deviation occur, suggesting incomplete charge extraction at these intensities. To investigate the reason for the incomplete charge extraction we compare the bulk lifetime Eq. (4) with the charge-carrier transit time, as shown in the inset of Figure 5a. The transit time, given by 𝑡tr ≈ 𝑑2 𝜇𝑉oc , corresponds to the time it takes a carrier to traverse the inter-electrode ⁄ distance. It can be seen that for the simulated second order dynamics, the pseudo-first order bulk lifetime 𝜏B becomes smaller than 𝑡tr as 𝑉oc exceeds 0.8V resulting in a substantial recombination during the extraction process at larger intensities, explaining the incomplete charge extraction at these intensities. On the other hand, a much better agreement is obtained for the case with 𝜇 = 10−2 cm2/Vs, corresponding to a hundred times shorter transit time 𝑡tr (the other parameters are kept the same as before), in Figure 5a. We can thus conclude that the underestimation of the carrier density seen at high intensities is related to the fact that higher-order (>1) recombination during the charge extraction process, starts to compete with extraction at higher intensities (see Eq. (4)). Therefore, a necessary condition to avoid recombination during the extraction pulse is that 𝑡tr ≪ 𝜏B. This limitation can be potentially overcome by applying a strong reverse bias extraction voltage pulse, as shown by Kniepert et al,46 although this will also give rise to a larger amount of displacement charges and thereby a larger noise on the extracted charge evaluation. At low light intensities (small 𝑉oc), the bimolecular recombination during the extraction process is negligible. However, as seen from Figure 6b, the assumption of uniform carrier distributions is no longer valid. Instead, the excess carrier distributions are highly non-uniform, varying exponentially with distance 𝑥 within the active layer, with the electron and hole profiles being mirror-symmetric to each other. This type of behaviour is generally expected for (dark) carriers, originating from the contacts.47-51 Accordingly, at low light intensity, the excess carrier densities of charges injected into the undoped active layer can be approximated by Δ𝑛(𝑥) ≈ 𝑛c exp (− 𝑞ሾ𝑉bi−𝑉ocሿ 𝑘𝑇 [1 − 𝑥 𝑑 ] ) Δ𝑝(𝑥) ≈ 𝑝a exp (− 𝑞ሾ𝑉bi−𝑉ocሿ 𝑘𝑇 𝑥 𝑑 ) (26) (27) for 𝑉oc < 𝑉bi, where 𝑉bi is an effective built-in potential that accounts for the energy-level bending in the vicinity of the injecting contacts,52 and 𝑝a and 𝑛c are the associated effective hole and electron densities in the dark at the anode (𝑥 = 0) and cathode (𝑥 = 𝑑) regions, respectively. Then, by substituting Eq. (26) and (27) into Eq. (24) and integrating, we find 𝑛CE ≈ ( 𝑘𝑇 𝑞ሾ𝑉bi−𝑉ocሿ 2 ) ሾ𝑝a + 𝑛cሿ (28) when (𝑉bi − 𝑉oc) ≫ 𝑘𝑇 𝑞⁄ . In Figure 5b, we have compared Eq. (28) with the simulated CE data and a good agreement can found when assuming 𝑝a = 𝑛c and 𝑉bi in Eq. (28) to be fixed and independent of the light intensity. In general, however, these quantities depend on the applied voltage (the injection level) as well; subsequently, a weak intensity dependence (via 𝑉oc) of 𝑝a, 𝑛c and 𝑉bi is to be expected. We note that these findings are consistent with previous results reported by Deledalle et al.29 Figure 5. In (a), the extracted CE density 𝑛𝐶𝐸 = 𝑄𝐶𝐸 𝑞𝑑⁄ (symbols), obtained from the simulated CE current transients, is shown as a function of open-circuit voltage for the organic solar cell for two different mobilities. The theoretical values in case of complete charge extraction Eq. (24), calculated from steady-state simulations, is indicated by solid lines. The expected analytical approximation Eq. (25) (i.e. 𝑛𝐶𝐸 = 𝑛𝑜𝑐) is indicated by red dashed line. The inset shows the carrier transit time for 𝜇 = 10−4 cm2/Vs, compared to 𝜏𝐵 (Eq. (4)), at different 𝑉𝑜𝑐. In (b), the transient simulations for the case with an undoped (intrinsic) and p-doped active layer are showed for the lower mobility case 𝜇 = 10−4 cm2/Vs. The solid lines depict the exact (numerical) Eq. (24) result, as calculated from steady-state simulations, for the two cases. The corresponding analytical approximations of Eq. (24) for the case with an undoped and a p-doped active layer at low intensities, given by Eq. (28) and (29), respectively, are indicated by dotted lines. Figure 6. The simulated excess electron and hole densities, Δ𝑛(𝑥) and Δ𝑝(𝑥), respectively, as a function of distance 𝑥 inside the active layer. The excess carrier densities are given by the difference between the steady-state carrier density under open-circuit conditions under illumination (𝑡 ≤ 0) and in the dark (𝑡 → ∞); see Figure 2c. In (a), the situation at a high open-circuit voltage (high light-intensity regime, 100 suns) is simulated. The standard analytical approximation Eq. (25), assuming a uniform carrier profile given by 𝑛𝑜𝑐 = √𝐺 𝛽⁄ , is indicated by the dotted line. In (b), the situation at a lower open-circuit voltage (low light-intensity regime, 1/1000 suns) is simulated. For comparison, the exponential approximation Eq. (27) for Δ𝑝(𝑥) has been included (assuming 𝑉𝑏𝑖 = 0.854V and 𝑝𝑎 = 1.25 × 1017 cm-3), as shown by the dashed line. Our findings suggest that at low 𝑉oc, the excess carrier density profiles are dominated by excess carriers, back-injected from the contacts, with 𝑛𝐶𝐸 also being influenced by the associated ′ ≠ 0). To further validate that it indeed is the injected and/or dark capacitive effects (Δ𝜎el background carriers that dominate the response in the capacitive CE regime, we have also included the case with a p-doped active layer in Figure 5b. In this case, the dark background carriers are dominated by doping-induced holes, with a carrier profile given by 𝑝(𝑥) ≈ 𝑁p for 𝑥 < 𝑑 − 𝑤, and 𝑝(𝑥) ≈ 0 else; where the depletion layer thickness is given by 𝑤(𝑉) ≈ √2𝜀𝜀0ሾ𝑉bi − 𝑉ሿ 𝑞𝑁p ⁄ and 𝑁p is the doping concentration.40,44,53 Subsequently, Eq. (24) can be approximated as 𝑛CE ≈ 𝑁p × [ 𝑤(0) 𝑑 − 𝑤(𝑉oc) 𝑑 − 1 2 ([ 𝑤(0) 𝑑 2 ] − [ 𝑤(𝑉oc) ] 𝑑 2 )] (29) which depends on 𝑉oc via 𝑤. Comparing Eq. (29) with the simulated CE data for the device with the doped active layer results in excellent agreement in the capacitive CE regime. These results also agree well with experimental data from Kiermasch et al.24 3.3. Discussion Based on the recombination lifetime obtained from TPV, and the corresponding steady-state carrier density extracted with CE, the recombination rate and the lifetime as a function of carrier density is usually mapped.13 The subsequent recombination rate, as derived from TPV and CE, can be expressed as ℛ𝑇𝑃𝑉/𝐶𝐸 = 1 1+𝜆 × 𝑛𝐶𝐸 𝜏𝑇𝑃𝑉 (30) where 𝜆 = 𝑚 𝜈⁄ and 1 + 𝜆 is the recombination order. In line with the above findings, however, due to the inaccuracy of 𝑛𝐶𝐸 at low light intensities, the TPV/CE recombination rate plotted against this extracted charge density is heavily plagued by capacitive effects. As discussed earlier this is caused by spatially separated carrier distributions and/or shunts. Only at high enough carrier densities, when 𝜏𝑇𝑃𝑉 ≈ 𝜏𝐵 and 𝑛𝐶𝐸 ≈ 𝑛𝑜𝑐, can the recombination rate constant be calculated from the recombination/charge plot. In this high carrier density regime, provided that the recombination during the extraction process in CE can be minimized, the slope of the recombination rate on a log-log plot, directly gives the associated recombination order. We should note that here we considered balanced electron and hole mobilities and a relatively thin junction. Imbalanced mobilities or thicker junctions (which make the transit times of electrons and holes imbalanced) result in more problems in the validity of assumption (i), as one needs to consider the lifetime of the slower carriers and their transit time. On the other hand, in thicker devices, the photo-induced excess charge carrier profiles are generally more uniform throughout the active layer.29 In addition, since capacitive effects can be reduced by increasing the active layer thickness, thicker devices are in general better suited for TPV but also for ensuring the validity of assumption (ii) in CE. This is consistent with the conclusions of previous work by Kirchartz and Deledalle.19,29,54 4. Experimental Demonstration on P3HT:PCBM Solar Cells To demonstrate the relevance of the above theoretical findings, we turn to experimental results obtained from organic bulk heterojunction solar cells, based on P3HT:PCBM. In Figure 7a and Figure 7b the experimentally extracted TPV lifetime and CE charge carrier density, respectively, are shown as a function of the steady-state open-circuit voltage 𝑉𝑜𝑐 of the P3HT:PCBM solar cell device in Reference 24. In Figure 7a, we have also included 𝜏𝑐𝑎𝑝, as estimated from the experimental J-V curve and the low-frequency capacitance of the device in the dark. Since this capacitance was found to show a relatively weak voltage dependence in the voltage range of interest, we assume a fixed value of 𝐶 = 0.55 mF/m2. Further experimental details of fabrication and measurements are outlined in Reference 24. At small 𝑉𝑜𝑐, good agreement is obtained between the estimated 𝜏𝑐𝑎𝑝 and the experimental 𝜏𝑇𝑃𝑉 in this case. Conversely, at high enough 𝑉𝑜𝑐, where we expect the recombination between bulk carriers to instead dominate the TPV response, a slope parameter of 𝜈 = 2 is obtained. A similar situation is true for the CE data in Figure 7b, where 𝑚 = 2 is obtained at high enough 𝑉𝑜𝑐. The corresponding TPV/CE recombination rate, using the obtained slope 𝜆 = 1, is shown in Figure 7c as a function of 𝑛𝐶𝐸. Accordingly, a slope of two is obtained in the relationship 2 , consistent with pure between total recombination rate and carrier density, i.e. ℛ𝑇𝑃𝑉 𝐶𝐸⁄ ∝ 𝑛𝐶𝐸 second-order recombination dominating at these light intensities. Furthermore, under these conditions (slope = 2), the measured data allows one to extract the second-order recombination coefficient via 𝛽 = (2𝑛𝐶𝐸𝜏𝑇𝑃𝑉)−1, in accordance with Eq. (4); from the 𝛽 vs 𝑛𝐶𝐸 plot, depicted in Figure 7d, we find 𝛽 ≈ 6 × 10−12 cm3 s⁄ . The experimental results in this work suggests that the dominating recombination mechanism in P3HT:PCBM is second-order near 1 sun incident light intensities with a constant second- order recombination coefficient, being in line with previous findings by various other methods.46,49,55-58 Furthermore, this also corroborates recent work by Tvingstedt and Deibel, who concluded that trap-assisted recombination via exponentially distributed tail states is not the dominating recombination mechanism in this system.18 Figure 7. Experimental results on P3HT:PCBM bulk heterojunction solar cells. In (a), the experimental TPV lifetime, as indicated by the symbols, is shown. The corresponding 𝜏𝑐𝑎𝑝, as estimated using the experimental J-V curve, is shown by the dashed blue lines. (b) The experimental CE density 𝑛𝐶𝐸 = 𝑄𝐶𝐸 𝑞𝑑⁄ , where the extracted charge 𝑄𝐶𝐸 has been corrected for the capacitance. (c) The experimental CE-TPV recombination rate 𝑛𝐶𝐸 𝜏𝑇𝑃𝑉 ⁄ is shown as a function of 𝑛𝐶𝐸. It can be seen that at high intensities a slope of 2 is obtained, consistent with second- order recombination. (d) The corresponding second-order recombination coefficient 𝛽CE TPV⁄ shown as function of 𝑛𝐶𝐸. 5. Conclusions = 1 (2𝑛𝐶𝐸𝜏𝑇𝑃𝑉) ⁄ is In conclusion, based on the analytical derivations and numerical device simulations on realistic organic solar cells, the relation between the bulk recombination of photo-induced charge carriers and the lifetime, as extracted from the TPV transients, has been clarified. At higher 𝑉𝑜𝑐 the lifetime is given by the sought-after bulk lifetime governed by bulk recombination. At lower light intensities and open-circuit voltages, the TPV lifetimes are instead limited by a composite RC-time constant which is dominated by the smallest of the load resistance of the measurement circuit, the shunt resistance (associated with parasitic leakage currents), and the internal (differential) resistance of the diode itself. For CE, the determination of the photo-generated carrier density 𝑛𝑜𝑐 under open-circuit conditions is more challenging (and requires uniform carrier distributions) for higher order recombination. At higher 𝑉𝑜𝑐, the CE measurement can be susceptible to recombination during the charge extraction process, which might lead to the extracted carrier density underestimating the actual steady-state density 𝑛𝑜𝑐 prior to the pulse. A necessary requirement for the correct determination of 𝑛𝑜𝑐 in this regime is that 𝑡𝑡𝑟 ≪ 𝜏𝐵. At lower 𝑉𝑜𝑐, however, the extracted CE density is dominated by capacitive effects caused by the extraction of spatially separated (back-injected) charge carriers. After accounting for the dark carrier profile (and their displacement current effects), analytical approximations for the capacitive CE regime could be obtained. Finally, the theoretical behaviour is reproduced experimentally on organic solar cells based on P3HT:PCBM. Associated Content Supporting Information: Analytical derivations, details regarding the device model, and additional simulations Acknowledgements The work was supported by the Sêr Cymru Program through the European Regional Development Fund, Welsh European Funding Office and Swansea University strategic initiative in Sustainable Advanced Materials. A.A. is a Sêr Cymru II Rising Star Fellow and P.M. a Sêr Cymru II National Research Chair. K.T. thanks the German Research Foundation (DFG) for funding through project 382633022 (RECOLPER). References (1) Brabec, C.; Scherf, U.; Dyakonov, V. Organic Photovoltaics; Wiley VCH: Weinheim, Germany, 2008. (2) Inganäs, O.; Admassie, S.; 25th Anniversary Article: Organic Photovoltaic Modules and Biopolymer Supercapacitors for Supply of Renewable Electricity: A Perspective from Africa. Adv. Mater. 2014, 26, 830-848. (3) Nelson, J. Polymer:Fullerene Bulk Heterojunction Solar Cells. Mater. Today 2011, 14, 462-470. (4) Meng, L.; Zhang, Y.; Wan, X.; Li, C.; Zhang, X.; Wang, Y.; Ke, X.; Xiao, Z.; Ding, L.; Xia, R., et al. Organic and Solution-Processed Tandem Solar Cells with 17.3% Efficiency. Science 2018, 361, 1094-1098. (5) Meerheim, R.; Lüssem, B.; Leo, K. Efficiency and Stability of p-i-n Type Organic Light Emitting Diodes for Display and Lighting Applications. Proc. IEEE 2009, 97, 1606-1626. (6) Lin, Q.; Armin, A.; Burn, P. L.; Meredith, P. Filterless Narrowband Visible Photodetectors. Nat. Photon. 2015, 9, 687-695. (7) Di, D.; Romanov, A. S.; Yang, L.; Richter, J. M.; Rivett, J. P. H.; Jones, S.; Thomas, T. H.; Abdi Jalebi, M.; Friend, R. H.; Linnolahti, M., et al. High-Performance Light-Emitting Diodes based on Carbene-Metal-Amides. Science 2017, 356, 159-163. (8) Jansen-van Vuuren, R. D.; Armin, A.; Pandey, A. K.; Burn, P. L.; Meredith, P. Organic Photodiodes: The Future of Full Color Detection and Image Sensing. Adv. Mater. 2016, 28, 4766- 4802. (9) Shi, L.; Liang, Q.; Wang, W.; Zhang, Y.; Li, G.; Ji, T.; Hao, Y.; Cui, Y. Research Progress in Organic Photomultiplication Photodetectors. Nanomaterials 2018, 8, 713. (10) Dou, L.; You, J.; Hong, Z.; Xu, Z.; Li, G.; Street, R. A.; Yang, Y. 25th anniversary article: A Decade of Organic/Polymeric Photovoltaic Research. Adv. Mater. 2013, 25, 6642-6671. (11) Deibel, C.; Dyakonov, V. Polymer-Fullerene Bulk Heterojunction Solar Cells. Rep. Prog. Phys. 2010, 73, 096401. (12) Proctor, C. M.; Kuik, M.; Nguyen, T.-Q. Charge Carrier Recombination in Organic Solar Cells. Prog. Polymer Science 2013, 38, 1941-1960. (13) Shuttle, C. G.; O'Regan, B.; Ballantyne, A. M.; Nelson, J.; Bradley, D. D. C.; de Mello, J.; Durrant, J. R. Experimental Determination of the Rate Law for Charge Carrier Decay in a Polythiophene: Fullerene Solar Cell. Appl. Phys. Lett. 2008, 92, 093311 (2008). (14) Foertig, A.; Baumann, A.; Rath, D.; Dyakonov, V.; Deibel, C. Charge Carrier Concentration and Temperature Dependent Recombination in Polymer-Fullerene Solar Cells. Appl. Phys. Lett. 2009, 95, 052104. (15) Maurano, A.; Shuttle, C. C.; Hamilton, R.; Ballantyne, A. M.; Nelson, J.; Zhang, W. M.; Heeney M.; Durrant, J. R. Transient Optoelectronic Analysis of Charge Carrier Losses in a Selenophene/Fullerene Blend Solar Cell. J. Phys. Chem. C 2011, 115, 5947-5957. (16) 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. Adv. Energy Mater. 2015, 5, 1401345. (17) Koster, L. J. A.; Mihailetchi, V. D.; Ramaker, R.; Blom, P. W. Light Intensity Dependence of Open-Circuit Voltage of Polymer:Fullerene Solar Cells. Appl. Phys. Lett. 2005, 86, 123509. (18) Tvingstedt, K.; Deibel, C. Temperature Dependence of Ideality Factors in Organic Solar Cells and the Relation to Radiative Efficiency. Adv. Energy Mater. 2016, 6, 1502230. (19) Kirchartz, T.; Nelson, J. Meaning of Reaction Orders in Polymer:Fullerene Solar Cells. Phys. Rev. B 2012, 86, 165201. (20) Street, R. A. Localized State Distribution and its Effect on Recombination in Organic Solar Cells. Phys. Rev. B 2011, 84, 075208. (21) Street, R. A. Reply to "Comment on 'Interface State Recombination in Organic Solar Cells'". Phys. Rev. B 2010, 82, 207302. (22) Mahan, J. E.; D. L. Barnes, D. L. Depletion Layer Effects in the Open-Circuit-Voltage-Decay Lifetime Measurement. Solid-State Electronics 1981, 24, 989-994. (23) Castaner, L.; Vilamajo, E.; Llaberia, J.; Garrido, J. Investigations of the OCVD Transients in Solar Cells. J. Phys. D: Appl. Phys. 1981, 14, 1867-1876. (24) Kiermasch, D.; Baumann, A.; Fischer, M.; Dyakonov, V.; Tvingstedt, K. Revisiting lifetimes from transient electrical characterization of thin film solar cells; a capacitive concern evaluated for silicon, organic and perovskite devices. Energy Environ. Sci. 2018, 11, 629-640. (25) Kirchartz, T.; Pieters, B. E.; Kirkpatrick, J.; Rau, U.; Nelson, J. Recombination Via Tail States in Polythiophene:Fullerene Solar Cells. Phys. Rev. B 2011, 83, 115209. (26) Blakesley, J. C.; D. Neher, D. Relationship between Energetic Disorder and Open-Circuit Voltage in Bulk Heterojunction Organic Solar Cells. Phys. Rev. B 2011, 84, 075210. (27) Rauh, D.; Deibel, C.; Dyakonov, V. Charge Density Dependent Nongeminate Recombination in Organic Bulk Heterojunction Solar Cells. Adv. Funct. Mater. 2012, 22, 3371. (28) Du, T.; Kim, J.; Ngiam, J.; Xu, S.; Barnes, P. R. F.; Durrant, J. R.; McLachlan, M. A. Elucidating the Origins of Subgap Tail States and Open-Circuit Voltage in Methylammonium Lead Triiodide Perovskite Solar Cells. Adv. Funct. Mater. 2018, 28, 1801808. (29) Deledalle, F.; Tuladhar, P. S.; Nelson, J.; Durrant, J. R.; T. Kirchartz, T. Understanding the Apparent Charge Density Dependence of Mobility and Lifetime in Organic Bulk Heterojunction Solar Cells. J. Phys. Chem. C 2014, 118, 8837-8842. (30) Lampert, M. A.; Mark, P. Current Injection in Solids; Academic Press: New York, 1970. (31) Kao, K. C.; Hwang, W. Electrical Transport in Solids; Pergamon Press: Oxford, 1981. (32) Hawks, S. A.; Finck, B. Y.; Schwartz, B. J. Theory of Current Transients in Planar Semiconductor Devices: Insights and Applications to Organic Solar Cells. Phys. Rev. Applied 2015, 3, 044014. (33) K. Tvingstedt, L. Gil-Escrig, C. Momblona, P. Rieder, D. Kiermasch, M. Sessolo, A. Baumann, H. J. Bolink and V. Dyakonov, Removing Leakage and Surface Recombination in Planar Perovskite Solar Cells. ACS Energy. Lett. 2017, 2, 424-430. (34) Credgington, D.; Jamieson, F. C.; Walker, B.; Nguyen, T.-Q.; Durrant, J. R. Quantification of Geminate and Non-Geminate Recombination Losses within a Solution-Processed Small- Molecule Bulk Heterojunction Solar Cell. Adv. Mater. 2012, 24, 2135-2141. (35) Credgington, D.; Liu, S.-W.; Nelson, J.; Durrant, J. R. In Situ Measurement of Energy Level Shifts and Recombination Rates in Subphthalocyanine/C60 Bilayer Solar Cells. J. Phys. Chem. C 2014, 118, 22858-22864. (36) Wood, S.; Blakesley, J. C.; Castro, F. A. Assessing the Validity of Transient Photovoltage Measurements and Analysis for Organic Solar Cells. Phys. Rev. Applied 2018, 10, 024038. (37) Sandberg, O. J.; Nyman, M.; Österbacka, R. Effect of Contacts in Organic Bulk Heterojunction Solar Cells. Phys. Rev. Applied 2014, 1, 024003. (38) Sandén, S.; Sandberg, O. J.; Xu, Q.; Smått, J.-H.; Juška, G.; Lindén, M.; Österbacka, R. Influence of Equilibrium Charge Reservoir Formation on Photo Generated Charge Transport in TiO2/organic Devices. Org. Electron. 2014, 15, 3506-3513. (39) Neukom, M. T.; Reinke, N. A.; B. Ruhstaller, B. Charge Extraction with Linearly Increasing Voltage: A Numerical Model for Parameter Extraction. Solar Energy 2011, 85, 1250-1256. (40) Sze, S. M. Physics of Semiconductor Devices, 3rd ed.; Wiley & Sons: New York, 1981. (41) Selberherr, S. Analysis and Simulation of Semiconductor Devices; Springer-Verlag: Wien, 1984. (42) Sandberg, O. J.; Sundqvist, A.; Nyman, M.; Österbacka, R. Relating Charge Transport, Contact Properties, and Recombination to Open-Circuit Voltage in Sandwich-Type Thin-Film Solar Cells. Phys. Rev. Applied 2016, 5, 044005. (43) van Mensfoort, S. L. M.; Coehoorn, R. Determination of Injection Barriers in Organic Semiconductor Devices from Capacitance Measurements. Phys. Rev. Lett. 2008, 100, 086802. (44) Kirchartz, T.; Gong, W.; Hawks, S. A.; Agostinelli, T.; MacKenzie, R. C. I.; Yang, Y.; Nelson, J. Sensitivity of the Mott−Schottky Analysis in Organic Solar Cells. J. Phys. Chem. C 2012, 116, 7672-7680. (45) Bisquert, J. Chemical Capacitance of Nanostructured Semiconductors: Its Origin and Significance for Nanocomposite Solar Cells. Phys. Chem. Chem. Phys. 2003, 5, 5360-5364. (46) Kniepert, J.; Lange, I.; van der Kaap, N. J.; Koster, L. J. A.; Neher, D. A Conclusive View on Charge Generation, Recombination, and Extraction in As-Prepared and Annealed P3HT:PCBM Blends: Combined Experimental and Simulation Work, Adv. Energy Mater. 2014, 4, 1301401. (47) Sah, C. T.; Noyce, R. N.; Shockley, W. Carrier Generation and Recombination in P-N Junctions and P-N Junction Characteristics. Proc. IRE 1957, 1228. (48) van Mensfoort, S. L. M.; Coehoorn, R. Effect of Gaussian Disorder on the Voltage Dependence of the Current Density in Sandwich-type Devices based on Organic Semiconductors. Phys. Rev. B 2008, 78, 085207. (49) Deibel, C.; Wagenpfahl, A.; Dyakonov, V. Origin of Reduced Polaron Recombination in Organic Semiconductor Devices. Phys. Rev. B 2009, 80, 075203. (50) Kumar, P.; Jain, S. C.; Kumar, V.; Chand, S.; Tandon, R. P. A Model for the J-V Characteristics of P3HT:PCBM Solar Cells. J. Appl. Phys. 2009, 105, 104507. (51) Dahlström, S.; Sandberg, O. J.; Nyman, M.; Österbacka, R. Determination of Charge-Carrier Mobility and Built-In Potential in Thin-Film Organic M-I-M Diodes from Extraction-Current Transients. Phys. Rev. Applied 2018, 10, 054019. (52) de Bruyn, P.; van Rest, H. P.; Wetzelaer, G. A. H.; de Leeuw, D. M.; Blom, P. W. M. Diffusion- Limited Current in Organic Metal-Insulator-Metal Diodes. Phys. Rev. Lett. 2013, 111, 186801. (53) Sandberg, O. J.; Nyman, M.; Österbacka, R. Direct Determination of Doping Concentration and Built-in Voltage from Extraction Current Transients. Org. Electron. 2014, 15, 3413-3420. (54) Deledalle, F.; Kirchartz, T.; Vezie, M. S.; Campoy-Quiles, M.; Tuladhar, P. S.; Nelson, J.; Durrant, J. R. Understanding the Effect of Unintentional Doping on Transport Optimization and Analysis in Efficient Organic Bulk-Heterojunction Solar Cells. Phys. Rev. X 2015, 5, 011032. (55) Juška, G.; Sliaužys, G.; Genevičius, K.; Arlauskas, K.; Pivrikas, A.; Scharber, M.; Dennler, G.; Sariciftci, N. S.; Österbacka, R. Charge-Carrier Transport and Recombination in Thin Insulating Films Studied via Extraction of Injected Plasma. Phys. Rev. B 2006, 74, 115314. (56) Garcia-Belmonte, G.; Boix, P. P.; Bisquert, J.; Sessolo, M.; Bolink, H. J. Simultaneous Determination of Carrier Lifetime and Electron Density-of-States in P3HT:PCBM Organic Solar Cells under Illumination by Impedance Spectroscopy. Solar Energy Mater. Sol. Cells 2010, 94, 366-375. (57) Baumann, A.; Lorrmann, J.; Rauh, D.; Deibel, C.; Dyakonov, V. A New Approach for Probing the Mobility and Lifetime of Photogenerated Charge Carriers in Organic Solar Cells Under Real Operating Conditions. Adv. Mater. 2012, 24, 4381-4386. (58) Wetzelaer, G. A. H.; Van der Kaap, N. J.; Koster, L. J. A.; Blom, P. W. M. Quantifying Bimolecular Recombination in Organic Solar Cells in Steady State. Adv. Energy Mater. 2013, 3, 1130. TOC (embedded pdf)
1901.10209	1	1901	2019-01-29T10:43:20	Femtosecond laser patterning of graphene electrodes for thin-film transistors	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	The aim of this study is to assess femtosecond laser patterning of graphene in air and in vacuum for the application as source and drain electrodes in thin-film transistors (TFTs). The analysis of the laser-patterned graphene with scanning electron microscopy, atomic force microscopy and Raman spectroscopy showed that processing in vacuum leads to less debris formation and thus re-deposited carbonaceous material on the sample compared to laser processing in air. It was found that the debris reduction due to patterning in vacuum improves the TFT characteristics significantly. Hysteresis disappears, the mobility is enhanced by an order of magnitude and the subthreshold swing is reduced from $S_{sub}~=~2.5~\mbox{V/dec}$ to $S_{sub}~=~1.5~\mbox{V/dec}$.	physics.app-ph	physics	Femtosecond laser patterning of graphene electrodes for thin-film transistors Maren Kasischke Ruhr University Bochum, Faculty of Mechanical Engineering, Applied Laser Technologies, Bochum, Germany Ersoy Suba¸sı Ruhr University Bochum, Faculty of Electrical Engineering and Information Technology, Electronic Materials and Nanoelectronics, Bochum, Germany Claudia Bock Ruhr University Bochum, Faculty of Electrical Engineering and Information Technology, Microsystems Technology, Bochum, Germany Duy-Vu Pham Evonik Resource Efficiency GmbH, Electronic Solutions, Marl Germany Evgeny L. Gurevich Ruhr University Bochum, Faculty of Mechanical Engineering, Applied Laser Technologies, Bochum, Germany Ulrich Kunze Ruhr University Bochum, Faculty of Electrical Engineering and Information Technology, Electronic Materials and Nanoelectronics, Bochum, Germany Andreas Ostendorf Ruhr University Bochum, Faculty of Mechanical Engineering, Applied Laser Technologies, Bochum, Germany Abstract Email address: [email protected] (Maren Kasischke) Preprint submitted to Applied Surface Science January 30, 2019 The aim of this study is to assess femtosecond laser patterning of graphene in air and in vacuum for the application as source and drain electrodes in thin-film transistors (TFTs). The analysis of the laser-patterned graphene with scanning electron microscopy, atomic force microscopy and Raman spec- troscopy showed that processing in vacuum leads to less debris formation and thus re-deposited carbonaceous material on the sample compared to laser processing in air. It was found that the debris reduction due to pat- terning in vacuum improves the TFT characteristics significantly. Hysteresis disappears, the mobility is enhanced by an order of magnitude and the sub- threshold swing is reduced from Ssub = 2.5 V/dec to Ssub = 1.5 V/dec. Keywords: femtosecond laser, graphene, ablation in vacuum, TFTs 1. Introduction Graphene is a monolayer of carbon atoms arranged in a honeycomb lat- tice and presents extraordinary properties such as wavelength-independent absorption in the UV-visible range of 2.3 % [1], high electrical conductiv- ity due to massless electrons [2] and high mechanical strength [3]. This 2D material has a great potential for electronic devices [4]. Much research effort has been put into a high quality and up-scalable production [5, 6] and feasibility of application of graphene [7, 8] in the last years. The han- dling and processing of graphene, however, still represents a great challenge, since several processing steps induce an unwanted modification of the ma- terial properties. For example, the use of photoresists or polymer films on graphene for photolithography and/or transfer processes cause chemical dop- ing of the layer [9 -- 13] or patterning with e-beam lithography hydrogenates the graphene basal plane causing defects [14]. The necessary post-processing cleaning step in Ar/H2 at 250 − 400 ◦C for removing polymer residues can also lead to graphene degradation [15]. In order to use graphene in electronic devices or to study its electrical properties it is usually patterned by lithography. Different methods exist, as for example plasma etching [16, 17], resist-free soft lithography [18], stencil mask lithography [15] and helium ion lithography [19]. One attractive possibility to pattern graphene is with ultrafast lasers [20 -- 26]. Additionally to common benefits of using ultrafast lasers for patterning, like minimized heat affected zone, flexibility and selectivity, this method is especially beneficial for graphene patterning. Indeed, being contactless 2 and requiring no sample preparation, it reduces undesired modification of the graphene. This method can be applied for patterning graphene also on sensitive flexible substrates due to its limited thermal influence. The problem however is that during laser patterning graphene forms folds [27] and debris causing disturbance resulting in unwanted defects and negative impact on its application as transistor electrodes. Femtosecond laser ablation of graphene in vacuum, compared to air, should result in cleaner and more controlled patterning of the layer because rather less material is redeposited on the substrate [28 -- 30]. Moreover, atmospheric oxygen can oxidize the remaining graphene upon laser ablation and reduce performance of graphene- based electronics. This article presents graphene ablation on silicon substrates with a fem- tosecond laser in air and in vacuum and assesses, if the resulting patterns are appropriate for its application as electrodes in metal-oxide thin-film transistors (MOTFTs). The patterned structures are analyzed topographi- cally, with scanning electron microscopy (SEM) and atomic force microscopy (AFM) and additionally by Raman spectroscopy. After measuring the sheet resistance of the patterned graphene electrodes, these are used as source and drain electrodes in MOTFTs. 2. Experimental The graphene samples (Graphenea Inc., Spain) are grown by chemical vapor deposition on copper and subsequently transferred onto silicon wafer with thermally grown silicon oxide layer of 300 nm. For laser ablation ex- periments of graphene the samples are used 'as received'. Laser processing is carried out with a fiber-rod amplified femtosecond laser (Tangerine, max- imal average power P = 20 W, wavelength λ = 1030 nm, repetition rate f = 2 MHz, pulse duration τp ≈ 280 fs, Amplitude Systemes, France), which is guided and focused onto the sample with a galvanometer scanner and a 63 mm f-theta objective. Taking M 2 = 1.05 into account the calculated radius at beam waist is ω0 = 16.8 µm. The graphene electrode itself remains unexposed to the laser radiation during the patterning, but the rest of the graphene layer is removed by the laser with a fluence of F = 79 mJ/cm2. The laser power is controlled by an assembly of a λ\2 plate and a polariz- ing beam splitter. For experiments in vacuum the samples are placed in a chamber where a vacuum of p ≈ 1 × 10−1 mbar is reached. 3 Transistors are prepared in bottom-gate bottom-contact configuration (see Fig. 1 right). After laser-induced patterning of source and drain graphene electrodes Ti/Au bond-pads are e-beam evaporated using a laser-cut shadow mask. The indium based metal-oxide precursor (iXsenic® S, Evonik Indus- tries AG, Germany) is spin-coated and annealed for one hour at 350 ◦C under atmospheric conditions. The resulting indium oxide layer is referred to as MO (metal oxide) in this article. Finally, a surface passivation layer (iXsenic® P, Evonik Industries AG, Germany) was spin-coated, dried, crosslinked in an additional UV-ozone treatment and converted at 350 ◦C. Graphene samples are analyzed with scanning electron microscopy (SEM, Leo Gemini 982, Carl Zeiss AG, Germany) using an acceleration voltage of 1 kV for uncoated graphene and 3 kV for graphene samples with MO and pas- sivation layer coated on top. Additionally, atomic force microscopy (AFM, Nanoscope 5, Bruker Corp., USA) is performed in PeakForce Tapping mode in ScanAsyst mode. Raman measurements (inVia, Renishaw GmbH, Ger- many) are done with an 100 × objective, λ = 532 nm and less than 0.7 mW power in order to avoid damage of the graphene layer. Raman mapping is performed with 1 µm step. Acquired spectra are baseline subtracted and intensity ratios of main peaks are plotted as Raman map. In this study the I(D)/I(G) ratio and I(2D)/I(G) ratio are calculated and plotted for each measured spot. Shading of Raman maps was fit by interpolation. Figure 1: Schematic figure of the graphene-based transistor. Left: graphene electrodes patterned by femtosecond laser on top of a thermally oxidized silicon substrate. Right: bottom-gate bottom-contact metal-oxide thin-film transistor with graphene electrodes. The colored stars mark the positions which are analyzed by SEM, AFM and Raman spectroscopy. 4 Figure 2: Scanning electron microscopy images of graphene patterned in air (first two rows) and graphene patterned in vacuum (last two rows) show a reduction of debris on the sample, when laser patterning graphene in vacuum. Three regions of the transistor are analyzed: the graphene electrode (first column, marked by purple star in Fig. 1), the interface of the electrode to the channel (second column, green star in Fig. 1) and the channel region between source and drain electrodes (third column, yellow star in Fig. 1). The images of these regions are displayed after laser patterning without additional layers (a)-c) and g)-i)) and after TFT preparation with MO and passivation layer (d)-f)-j)-l)). 3. Results 3.1. Monolayer graphene patterning in air and in vacuum Ablation threshold of monolayer graphene (MLG) is determined by the well-known method of Liu [31], the semilog plots of diameter square vs pulse energy can be found in the SI. The thresholds of the layer in air Fth,air = 74 mJ/cm2 and in vacuum Fth,vac = 77 mJ/cm2 are very similar, 5 taking into account the power measurement accuracy of ± 5 %. These values are lower than the ablation threshold of the silicon substrate with thermally oxidized silicon dioxide (Fth,Si = 138 mJ/cm2). For patterning functional structures with an appropriate electrode layout, the ablated areas (see Fig. 1) were scribed with a scanning speed of vsc = 0.2 m/s and distance between lines of ∆y = 11 µm. This means a 95.7 % overlap of pulses within one line and 46.7 % overlap of lines is given, when taking into account the effective radius of ωef f = 11.7 µm calculated with Liu method [31]. The optimal ablation parameters were found by varying laser fluence, laser scanning speed and laser repetition rate. The fluence window below the damage threshold of the silicon substrate was evaluated in steps of 11 mJ/cm2 from 113 mJ/cm2 to 79 mJ/cm2. The scanning speed was varied from 6.1 m/s to 0.2 m/s. Additionally, ablation experiments at different repetition rates in the range of 50 kHz to 200 kHz, in steps of 50 kHz, were performed, while adapting the scanning speed to keep the same overlap of pulses of 95.7 %. The laser ablation parameters, which resulted in the cleanest graphene edge quality, were found to be at a repetition rate of 200 kHz, fluence of 79 mJ/cm2 and a scanning speed of 0.2 m/s. The topography of the patterned graphene electrodes was analyzed with SEM, see Fig. 2. For this purpose, three sections of the transistor are scanned: the patterned graphene electrode (first column), the interface be- tween the electrode and the channel (second column) and the channel be- tween the source and the drain electrodes (third column). These sections are marked by a purple, green and yellow star, respectively, in Fig. 1. The first two rows of Fig. 2 display images of MLG patterned in air and the last two rows contain images of MLG patterned in vacuum. In each case the first row displays the uncoated graphene after laser patterning and the second row displays the complete TFT, which consists of three layers: (1) MLG; (2) MO; (3) passivation layer. When comparing the SEM images it is clear, that the ablation of MLG in vacuum produces less debris than the ablation of MLG in air. Especially, when examining the debris in the channel between source and drain electrodes of the transistor. A reduction of debris formation while patterning MLG in vacuum is con- firmed by AFM images, see Fig. 3. The AFM images were taken of the same three regions of interests as marked in Fig. 1: patterned graphene electrode, the interface between the graphene and the channel, and the channel region. In this case the patterned graphene sample is scanned without additional MO and passivation layers. In the channel region of the sample patterned 6 Figure 3: Atomic force microscopy images of laser-patterned graphene electrodes a)-c) in air show higher occurrence of re-deposited material and formation of debris than the structures laser-patterned d)-f) in vacuum. The arrows mark the border of graphene and Si/SiO2 substrate. in air (Fig. 3 c)) a roughness of RRM S = 6.02 nm with a maximum height of the profile of Rmax = 121 nm was measured. The sample patterned in vacuum also displays redeposited material, but in this case a lower roughness of RRM S = 2.22 nm with a maximum height of the profile of Rmax = 53 nm in the channel region (Fig. 3 f)). Raman mapping of laser patterned graphene shows that the intensity of the D peak (1350 cm−1) is higher compared to the G peak (1600 cm−1) at the edges of graphene electrodes, see Fig. 4 a) and b). This is to be expected, as the D peak emerges due to defects in the carbon network and is more pronounced at the edges due to the disturbed translation symmetry of graphene [32]. Additionally, the D peak may also be attributed to oxidation of graphene edges, which occurs after breaking up the carbon bonds during laser ablation [20, 22, 25]. The intensity of the 2D peak (2700 cm−1) in the Raman spectra of graphene is very sensitive to the number of layers and its intensity is re- ported to be up to 4 times higher than the G peak intensity in MLG [33]. The mapping of the I(2D)/I(G) demonstrates that the peak ratio for the pro- cessed samples is approximately 2, i.e., somewhat lower to the values reported in the literature [33]. The left Raman maps of Fig. 4 a) and b) show that the graphene electrodes display some spots with spectra where I(2D)/I(G) ≤ 1. This signals may be from redeposited carbonaceous material on the 7 a) b) 5 4 3 2 1 r i A m u u c a V c) ) . u . a ( y t i s n e t n I 5 µm 5 µm 5 µm 5 µm D G I(D)/I(G) = 1.00 D+D'' I(2D)/I(G) = 2.14 2D I(D)/I(G) = 0.14 I(2D)/I(G) = 2.14 I(D)/I(G) = 0.04 I(2D)/I(G) = 1.25 0 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000 3200 Raman-Shift (cm-1) Electrode Electrode 2 1.5 1 0.5 0 r i A m u u c a V I(2D) I(G) 5 µm 1 0.5 I(D) I(G) 5 µm Channel 0 1 Channel d) 2 1.5 3 µm 1 I(2D) I(G) 3 µm 0.5 I(D) I(G) e) 0.5 0 0 3 µm 3 µm Figure 4: Optical microscopy image and Raman map of I(2D)/I(G) ratio and I(D)/(G) ratio of the graphene electrodes laser patterned a) in air and b) in vacuum. c) Represen- tative Raman spectra of the graphene electrode patterned in vacuum, normalized to the G peak with an vertical offset for visualization purposes. The corresponding positions in the Raman map are marked by circles with the matching colors in b). Raman map of I(2D)/I(G) ratio and I(D)/(G) ratio in the TFT channel region of samples patterned d) in air and e) in vacuum. graphene electrode itself after laser ablation. Overall, the Raman maps show that the edges of the graphene electrode patterned in vacuum (Fig. 4 b)) are smoother than the graphene electrode patterned in air (Fig.4 a)). Raman spectra taken in the channel region show that the debris and redeposited material observed before with SEM and AFM is attributed to carbonaceous material. These Raman spectra are displayed as a map in Fig. 4 d)-e). The signal features high G and D peaks and a very low 2D peak. Therefore, the debris produced during laser ablation is most probably carbon material that has been redeposited and oxidized during the laser ablation process. 3.2. Thin film transistors with laser patterned graphene electrodes Prior to the preparation of the metal-oxide film the sheet resistance of the graphene electrodes was determined in four-point geometry. The sheet- 8 resistance amounts 583 ± 63 Ω/sq and 542 ± 19 Ω/sq for graphene electrodes patterned in air and in vacuum, respectively. Compared to the values spec- ified by the manufacturer Graphenea (440 ± 40 Ω/sq) the sheet-resistance is not significantly affected by the laser process, whereby laser processing under vacuum has a lower impact on the sheet resistance. This might be a consequence of suppressed oxidation of graphene during laser processing in vacuum. a) b) c) d) Figure 5: Output and transfer characteristics of MOTFTs with graphene source and drain electrodes laser patterned in a)-b) air and c)-d) vacuum. The channel length (width) amounts L = 100 µm (W = 500 µm). The output and transfer characteristics of the TFTs show a great differ- ence in performance with source and drain MLG electrodes laser patterned in air compared to the MLG electrodes patterned in vacuum (Fig. 5). The output characteristics of the transistor patterned in air shows hysteresis and does not saturate. This points to a parasitic conductivity in the channel. The transfer characteristics exhibit a clear shift to negative voltages during measurement. In contrast, the transistors with electrodes patterned in vac- uum exhibit clear current saturation and a pinch-off behavior indicating that 9 the entire thickness of the semiconductor channel layer can be depleted of free electrons (Fig. 5 c) and d)). More negative gate voltages are necessary to achieve full depletion of the active layer compared to TFTs with MLG electrodes patterned in air. A less pronounced shift to negative voltages is measured in the sample with graphene electrodes patterned in vacuum. Both samples have a high negative threshold voltage. The subthreshold swing Ssub which is directly related to the interface trap density NS between the insula- tor and the MO, is extracted from the inverse of the maximum slope of the transfer characteristic. Ssub of the TFTs with patterned graphene electrodes in air and in vacuum at VDS = 5 V is around 2.5 V/dec and 1.5 V/dec, respectively. This corresponds to a maximum interface trap density NS of 3.4×1012 cm−2 and 2.0×1012 cm−2 of the TFTs with graphene electrodes pat- terned in air and in vacuum, respectively. These values are higher compared to MOTFTs based on the same precursor with metal electrodes patterned by conventional lift-off technique (Ssub = 0.37 V/dec) [34]. We attribute this to a rough semiconductor/insulator interface arising from the debris. The roughly estimated field-effect mobility of µF E ≈ 0.1 cm2V−1s−1 and µF E ≈ 2.2 cm2V−1s−1 for TFTs with graphene electrodes patterned in air and vacuum, respectively, strongly underlines the relevance of a clean and smooth semiconductor/insulator interface. Compared to MOTFTs with con- ventional patterning methods and metallic electrodes µF E ≈ 28 cm2V−1s−1 the achieved field-effect mobility is still considerably smaller. Although a direct comparison is virtually impossible due to the large number of factors which must be taken into account, e.g. electrode material, passivation layer, interface treatment. 4. Discussion Laser-patterned graphene electrodes in vacuum, i.e. at reduced ambient pressure, exhibit less debris than the samples patterned in air. The size and the density of redeposited material depend on the ambient conditions due to two following pressure-dependent effects: reduction in the viscous drag force and in the density of the ambient gas. Both these effects reduce the kinetic energy dissipation of the ablated atoms, and hence enable them to fly far away from the sample surface. Measurements by Yoshida et al. demonstrated that the mean diameter of laser-ablated particles decreases with decreasing ambient gas pressure [35]. As demonstrated by Geohegan [36], the cloud of ablated nanoparticles becomes denser and propagates slower at a higher 10 ambient pressure. Thus, at a reduced pressure, the ablated particles are not confined in a vicinity of the sample surface but propagate further in the chamber and can even reach the chamber walls and be deposited there, as it happens in PLD (pulsed laser deposition) [37]. The reduction of debris upon laser patterning of graphene in vacuum has a positive impact on the performance of MOTFTs compared to the tran- sistors, with graphene electrodes processed in air. The high density of the redeposited particles in the channel of the sample laser ablated in air causes great perturbations in the thin MO layer. Since the deposition method of the MO is spin-coating, these particles lead to well known thickness variations and perturbations within the layer or cause the film not to be completely closed (see Fig. 2 and 3). Thicker MO films result in the presence of a high- conductive back-channel layer at a distance beyond the screening length of the metal oxide/dielectric stack that induces humplike subthreshold transfer characteristics and more negative threshold voltages [38, 39]. Additionally, the variations in film thickness might cause an incomplete conversion of the MO precursor, since the conversion parameters are optimized for a homoge- neous layer [34, 40]. These effects lead to a parasitic conductivity causing the measured hysteresis and lacking saturation. Even though the density of particles per area is considerably decreased in the samples patterned in vac- uum, these two effects can also occur to a smaller extent here. The negative threshold voltage and the high subthreshold swing indicate that a further re- duction of debris is essential for high performance TFTs with laser patterned graphene electrodes. Once this is achieved the laser process is favorable com- pared to conventional patterning by UV lithography avoiding doping of the graphene electrodes due to resist residues [13]. 5. Conclusion Femtosecond laser patterning of CVD-grown graphene electrodes on sili- con substrates and their application as source and drain electrodes in MOTFTs were studied. The quality of the patterned electrodes was improved by laser- ablation in vacuum due to debris reduction compared to laser processing in air which is consistent with previous reports on laser processing in vacuum [28 -- 30]. Raman spectroscopy analysis showed that the I(2D)/I(G) ratios for samples processed in air and in vacuum (≈ 2) are comparable, which sug- gests that having air or vacuum as environment during patterning does not considerably influence the graphene layer. A slight increase of I(D)/I(G) ra- 11 tios on the edges of the patterned graphene electrodes with respect to the not laser-processed areas regardless of the ambient conditions indicates an increase of defects in these regions. Graphene laser patterning in vacuum leads to an improvement of MOTFT characteristics and a reduction of the semiconductor/insulator trap density. Most notably, this is the first study to our knowledge to demonstrate MOTFTs using femtosecond laser-patterned graphene electrodes. However, some limi- tations are worth noting. Improving the graphene edge quality and a further reduction of debris, e.g. by enhancing the vacuum quality during laser pro- cessing is desirable. Acknowledgements We would like to thank the financial support of the Federal Ministry of Education and Research of Germany within the m-era.net project "CMOT- Investigation and tuning of graphene electrodes for solution-processable metal oxide thin-film transistors in the area of low-cost electronics" (03XP0014). We gratefully acknowledge the support by the Center for Interface-Dominated High Performance Materials (ZGH) at the Ruhr University Bochum for the access to the AFM. References [1] R. R. Nair, P. Blake, A. N. Grigorenko, K. S. Novoselov, T. J. Booth, T. Stauber, N. M. R. Peres, A. K. Geim, Fine structure constant defines visual transparency of graphene, Science 320 (5881) (2008) 1308. [2] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, A. A. Firsov, Two-dimensional gas of massless dirac fermions in graphene, Nature 438 (7065) (2005) 197 -- 200. [3] C. Lee, X. Wei, J. W. Kysar, J. Hone, Measurement of the elastic prop- erties and intrinsic strength of monolayer graphene, science 321 (5887) (2008) 385 -- 388. [4] K. S. Novoselov, V. I. Fal'ko, L. Colombo, P. R. Gellert, M. G. Schwab, K. Kim, A roadmap for graphene, Nature 490 (7419) (2012) 192 -- 200. 12 [5] X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Vela- makanni, I. Jung, E. Tutuc, et al., Large-area synthesis of high-quality and uniform graphene films on copper foils, science 324 (5932) (2009) 1312 -- 1314. [6] S. Bae, H. Kim, Y. Lee, X. Xu, J.-S. Park, Y. Zheng, J. Balakrishnan, T. Lei, H. R. Kim, Y. I. Song, Y.-J. Kim, K. S. Kim, B. Ozyilmaz, J.-H. Ahn, B. H. Hong, S. Iijima, Roll-to-roll production of 30-inch graphene films for transparent electrodes, Nature nanotechnology 5 (8) (2010) 574 -- 578. [7] A. K. Geim, Graphene: Status and prospects, Science (New York, N.Y.) 324 (5934) (2009) 1530. [8] K. S. Kim, Y. Zhao, H. Jang, S. Y. Lee, J. M. Kim, K. S. Kim, J.-H. Ahn, P. Kim, J.-Y. Choi, B. H. Hong, Large-scale pattern growth of graphene films for stretchable transparent electrodes, nature 457 (7230) (2009) 706. [9] M. Ishigami, J. H. Chen, W. G. Cullen, M. S. Fuhrer, E. D. Williams, Atomic structure of graphene on SiO2, Nano letters 7 (6) (2007) 1643 -- 1648. [10] S. Kumar, N. Peltekis, K. Lee, H.-Y. Kim, G. S. Duesberg, Reliable processing of graphene using metal etchmasks, Nanoscale research letters 6 (1) (2011) 390. [11] K. Kumar, Y.-S. Kim, E.-H. Yang, The influence of thermal annealing to remove polymeric residue on the electronic doping and morphological characteristics of graphene, Carbon 65 (2013) 35 -- 45. [12] J. W. Suk, W. H. Lee, J. Lee, H. Chou, R. D. Piner, Y. Hao, D. Akin- wande, R. S. Ruoff, Enhancement of the electrical properties of graphene grown by chemical vapor deposition via controlling the effects of polymer residue, Nano letters 13 (4) (2013) 1462 -- 1467. [13] C. A. Chavarin, A. A. Sagade, D. Neumaier, G. Bacher, W. Mertin, On the origin of contact resistances in graphene devices fabricated by optical lithography, Applied Physics A 122 (2) (2016) 58. 13 [14] S. Ryu, J. Maultzsch, M. Y. Han, P. Kim, L. E. Brus, Raman spec- troscopy of lithographically patterned graphene nanoribbons, ACS nano 5 (5) (2011) 4123 -- 4130. [15] A. Mahmood, C.-S. Yang, J.-F. Dayen, S. Park, M. V. Kamalakar, D. Metten, S. Berciaud, J.-O. Lee, B. Doudin, Room temperature dry processing of patterned CVD graphene devices, Carbon 86 (2015) 256 -- 263. [16] I. Childres, L. A. Jauregui, J. Tian, Y. P. Chen, Effect of oxygen plasma etching on graphene studied using raman spectroscopy and electronic transport measurements, New Journal of Physics 13 (2) (2011) 025008. [17] H. Al-Mumen, F. Rao, W. Li, L. Dong, Singular sheet etching of graphene with oxygen plasma, Nano-Micro Letters 6 (2) (2014) 116 -- 124. [18] A. George, S. Mathew, R. van Gastel, M. Nijland, K. Gopinadhan, P. Brinks, T. Venkatesan, J. E. ten Elshof, Large area resist-free soft lithographic patterning of graphene, Small 9 (5) (2013) 711 -- 715. [19] D. C. Bell, M. C. Lemme, L. A. Stern, J. R. Williams, C. M. Marcus, Precision cutting and patterning of graphene with helium ions, Nan- otechnology 20 (45) (2009) 455301. [20] G. Kalita, L. Qi, Y. Namba, K. Wakita, M. Umeno, Femtosecond laser induced micropatterning of graphene film, Materials Letters 65 (11) (2011) 1569 -- 1572. [21] R. J. Stohr, R. Kolesov, K. Xia, J. Wrachtrup, All-optical high- resolution nanopatterning and 3D suspending of graphene, ACS nano 5 (6) (2011) 5141 -- 5150. [22] W. Zhang, L. Li, Z. B. Wang, A. A. Pena, D. J. Whitehead, M. L. Zhong, Z. Lin, H. W. Zhu, Ti:sapphire femtosecond laser direct micro-cutting and profiling of graphene, Appl. Phys. A 109 (2) (2012) 291 -- 297. [23] R. Sahin, E. Simsek, S. Akturk, Nanoscale patterning of graphene through femtosecond laser ablation, Applied Physics Letters 104 (5) (2014) 53118. 14 [24] J. van Erps, T. Ciuk, I. Pasternak, A. Krajewska, W. Strupinski, S. van Put, G. van Steenberge, K. Baert, H. Terryn, H. Thienpont, N. Vermeulen, Laser ablation- and plasma etching-based patterning of graphene on silicon-on-insulator waveguides, Optics express 23 (20) (2015) 26639 -- 26650. [25] T. Dong, M. Sparkes, C. Durkan, W. O'Neill, Evaluating femtosecond laser ablation of graphene on SiO2/Si substrate, Journal of Laser Ap- plications 28 (2) (2016) 022202. [26] I. Bobrinetskiy, A. Emelianov, A. Nasibulin, I. Komarov, N. Otero, P. M. Romero, Photophysical and photochemical effects in ultrafast laser pat- terning of CVD graphene, Journal of Physics D: Applied Physics 49 (41) (2016) 41LT01. [27] J.-H. Yoo, J. Bin In, J. Bok Park, H. Jeon, C. P. Grigoropoulos, Graphene folds by femtosecond laser ablation, Applied Physics Letters 100 (23) (2012) 233124. [28] B. N. Chichkov, C. Momma, S. Nolte, F. Von Alvensleben, A. Tunnermann, Femtosecond, picosecond and nanosecond laser abla- tion of solids, Applied Physics A 63 (2) (1996) 109 -- 115. [29] T. Matsumura, A. Kazama, T. Yagi, Generation of debris in the fem- tosecond laser machining of a silicon substrate, Applied Physics A 81 (7) (2005) 1393 -- 1398. [30] P. Feng, N. Zhang, H. Wu, X. Zhu, Effect of ambient air on femtosecond laser ablation of highly oriented pyrolytic graphite, Optics letters 40 (1) (2015) 17 -- 20. [31] J. M. Liu, Simple technique for measurements of pulsed gaussian-beam spot sizes, Optics Letters 7 (5) (1982) 196. [32] A. C. Ferrari, Raman spectroscopy of graphene and graphite: Disorder, electron -- phonon coupling, doping and nonadiabatic effects, Solid State Communications 143 (1-2) (2007) 47 -- 57. [33] A. C. Ferrari, Raman spectrum of graphene and graphene layers, Phys- ical Review Letters 97 (18) (2006) 187401. 15 [34] F. Jaehnike, D. V. Pham, R. Anselmann, C. Bock, U. Kunze, High- quality solution-processed silicon oxide gate dielectric applied on indium oxide based thin-film transistors, ACS applied materials & interfaces 7 (25) (2015) 14011 -- 14017. [35] T. Yoshida, S. Takeyama, Y. Yamada, K. Mutoh, Nanometer-sized sili- con crystallites prepared by excimer laser ablation in constant pressure inert gas, Appl. Phys. Lett. 68 (13) (1996) 1772 -- 1774. [36] D. B. Geohegan, Fast intensified-CCD photography of Y Ba2Cu3O7−x laser ablation in vacuum and ambient oxygen, Appl. Phys. Lett. 60 (22) (1992) 2732 -- 2734. [37] P. Willmott, J. Huber, Pulsed laser vaporization and deposition, Re- views of Modern Physics 72 (1) (2000) 315. [38] C. Y. Koo, K. Song, T. Jun, D. Kim, Y. Jeong, S.-H. Kim, J. Ha, J. Moon, Low temperature solution-processed InZnO thin-film transis- tors, Journal of The Electrochemical Society 157 (4) (2010) J111 -- J115. [39] W. Maeng, J. S. Park, H.-S. Kim, E. S. Kim, K. S. Son, T. S. Kim, M. Ryu, S. Lee, The effect of active-layer thickness and back-channel conductivity on the subthreshold transfer characteristics of Hf -- In -- Zn -- O TFTs, IEEE Electron Device Letters 32 (8) (2011) 1077 -- 1079. [40] C. Weber, M. Oberberg, D. Weber, C. Bock, D. V. Pham, U. Kunze, Im- proved morphology and performance of solution-processed metal-oxide thin-film transistors due to a polymer based interface modifier, Ad- vanced Materials Interfaces 1 (7) (2014) 1400137. 16
1812.09267	1	1812	2018-12-09T09:22:53	Laser-induced Phase-change Perovskite Photodetector Arrays for Optical Information Storage and Imaging	[ "physics.app-ph", "physics.optics" ]	The reversible phase transition between orthorhombic and cubic phases is demonstrated in all-inorganic perovskite CsPbIBr2 grown by a dual-source vapor-deposition method. The phase transition is actuated via heating and moisture exposure. The different crystal structures of two phases result in distinct optoelectronic properties including optical absorption, refractive index, and carrier transport. The perovskite photodetector array for the non-volatile rewritable optical memory application is further investigated. A near-infrared (NIR) laser ({\lambda} = 1064 nm) is used to selectively heat photodetector pixels through the photothermal effect of the interfacial Ag electrodes, resulting in an optically absorbing perovskite phase for the photodetectors. The locations of specific laser-written pixels can be read out by measuring their photocurrents, and the stored optical information can also be erased by moisture exposure. Finally, a proof-of-concept optical imaging application has been demonstrated using perovskite photodetector arrays fabricated on flexible PET substrates. These results show promising applications of vapor-deposited inorganic perovskite for optical memory and image sensors, with unique potential for low-cost manufacturing of large-area and/or flexible devices.	physics.app-ph	physics	Laser-induced Phase-change Perovskite Photodetector Arrays for Optical Information Storage and Imaging Chen Zou1, Jiajiu Zheng1, Cheng Chang1, Arka Majumdar1,2, Lih Y. Lin1* 1Department of Electrical and Computer Engineering, University of Washington, Seattle, WA 98195, USA 2Department of Physics, University of Washington, Seattle, WA 98195, USA Corresponding author. E-mail: [email protected] Abstract The reversible phase transition between orthorhombic and cubic phases is demonstrated in all-inorganic perovskite CsPbIBr2 grown by a dual-source vapor-deposition method. The phase transition is actuated via heating and moisture exposure. The different crystal structures of two phases result in distinct optoelectronic properties including optical absorption, refractive index, and carrier transport. The perovskite photodetector array for the non-volatile rewritable optical memory application is further investigated. A near- infrared (NIR) laser (λ = 1064 nm) is used to selectively heat photodetector pixels through the photothermal effect of the interfacial Ag electrodes, resulting in an optically absorbing perovskite phase for the photodetectors. The locations of specific laser-written pixels can be read out by measuring their photocurrents, and the stored optical information can also be erased by moisture exposure. Finally, a proof-of-concept optical imaging application has been demonstrated using perovskite photodetector arrays fabricated on flexible PET substrates. These results show promising applications of vapor-deposited inorganic perovskite for optical memory and image sensors, with unique potential for low-cost manufacturing of large-area and/or flexible devices. Keywords: all-inorganic perovskite, phase change, laser heating, optical information storage, image sensor Introduction Metal halide perovskite has attracted much research interest in recent years due to its outstanding optoelectronic properties including excellent optical absorption, high carrier mobility and long carrier lifetime.[1-2] Since the first perovskite solar cell was reported,[3] significant research efforts have been devoted to improve the photovoltaic efficiency and understand the optical and electronic properties of the underlying material, viz. metal halide perovskite.[4-5] In the last few years, the power conversion efficiency (PCE) of organic-inorganic perovskite solar cells has rapidly advanced to >22%.[6] The crystal structure of metal halide perovskites has been a topic of interest to explain their unique optoelectronic properties.[7-8] The methylammonium metal halide perovskites (MAPbX3), in which the organic cation (MA+) fills up the voids of corner-sharing PbI6 octahedra framework, were first intensively investigated,[9] and MAPbI3 was reported to undergo phase transition from tetragonal to cubic phase at between 42 and 57 ºC.[10] However, the structural change between the two phases are not significant with only slight titling of octahedral framework.[11] While MAPbI3-based perovskite solar cells have achieved high PCEs, these cells are not stable in ambient conditions.[12] Their poor thermal stability also limits the cell module processing temperature.[13] At high temperatures, methylammonium metal halide perovskites are easily dissociated to organic halides and lead halides, degrading the photovoltaic performance.[14] Replacing the organic cation with cesium (Cs) is considered as one possible way to address the aforementioned issues.[15-16] The solar cells based on all-inorganic cesium lead iodide (CsPbI3) perovskite quantum dots (QDs) have been shown to achieve the highest certified PCE of 13.43% among QD solar cells,[17] and optically-pumped vertical cavity lasers incorporating CsPbBr3 QDs have been shown to achieve ultra-low lasing threshold.[18] Both works reported outstanding stability among perovskite optoelectronic devices. Thus far, most of perovskite thin film fabrications utilize solution-processing methods such as spin-coating and drop-casting. However, CsPbIxBr3-x is much less soluble in solutions, which limits its applications in solution-processed optoelectronic devices.[19] On the other hand, the vapor-deposition method offers many advantages including good film morphology, precisely controlled thickness, ease of patterning, high reproducibility and better adaptability in flexible device fabrications.[20-21] In this work, we reported for the first time a systematic study on the structural phase transitions of vapor-deposited all- inorganic perovskite CsPbIBr2 thin films. The as-deposited inorganic perovskite exhibits the perovskite (PVSK) phase and can be transitioned to the non-perovskite (non-PVSK) phase after appropriate water moisture exposure. This nonvolatile phase transition can be reversed repeatedly through heat/moisture exposure treatment cycles. The two different phases correspond to significant distinctions in the material's optoelectronic properties including photoluminescence (PL), optical absorption, charge transport. To demonstrate potential applications utilizing such characteristics, we also fabricated a photodetector (PD) array by vapor-depositing the CsPbIBr2 film on a substrate pre-patterned with inkjet- printed Ag interdigitated electrodes (IDEs). In addition to heating all PD pixels uniformly on a heater platform, we can use a Nd:YVO4 laser with 1064 nm wavelength to generate heat selectively in specific PD pixels through the photothermal effect in Ag,[22] which makes the transition from the non-PVSK to PVSK phase site-selectable. The PD pixel in the PVSK phase shows high responsivity up to 1.5 A/W with a broad absorption spectrum from 350 to 600 nm. On the contrary, the PD pixel in the non-PVSK phase exhibits a much smaller photoresponsivity and a narrower absorption spectrum. The large difference in the PD performance between these two nonvolatile phases provides the opportunities for potential applications in optical memory and information storage. The laser heating and moisture exposure can be used to "set" and "reset" the selected PD pixels, and the photocurrent is read out to extract the phase information patterned by laser writing. In addition, the PD array in the PVSK phase presents a possibility for flexible image sensors, and imaging of a pre-designed pattern is demonstrated using a flexible perovskite PD array. Results and discussion 1. Characterization of vapor-deposited CsPbIBr2 thin films In this study, we used CsPbIBr2 in further characterizations and device fabrication due to its better stability. The fabrication procedure of vapor-deposited CsPbIBr2 PD array on rigid glass or flexible PET substrates is schematically illustrated in Figure 1a. The substrate was first ultrasonically cleaned and placed onto a printing platform. Conductive silver ink was then inkjet-printed onto the substrate to form an 8×8 interdigitated electrode (IDE) array (see METHODS section). The CsPbIBr2 film was deposited onto the prepatterned substrate via a vapor-deposition process. The CsI, PbBr2, were separately placed into 2 crucibles and simultaneous sublimed. By controlling the evaporation rates of the two sources, the mixed halide inorganic CsPbIBr2 films can be obtained. Energy dispersive (EDS) X-ray analysis was used to estimate the halide stoichiometry of the resulting inorganic perovskite films. Figure S1 (see Supporting Information) shows the element distribution of iodide (I) and bromide (Br), the I/Br molar ratio maintains approximately 1:2 along the feature line, indicating formation of the CsPbIBr2 thin film. The top-view scanning electron microscopy (SEM) image of the CsPbIBr2 film is presented in Figure 1b. The homogenous CsPbIBr2 film shows complete surface coverage without obvious pinholes. The atomic force microscopy (AFM) image of the CsPbIBr2 film in Figure 1c shows the surface roughness of 4.1 nm in a typical area of 5×5 μm2, smaller than that of spin coated film (see Figure S3), further confirming the flat and smooth surface morphology. A unique advantage of the vapor-deposition over solution-cast process is that patterned perovskite thin films can be readily achieved through a shadow mask and more compatible with flexible substrates. A dot array pattern made of vapor-deposited CsPbIBr2 on a flexible PET substrate is clearly observed in Figure 1d. 2. Phase transitions of inorganic perovskites via heating and moisture exposure As shown in Figure 2a, the as-deposited CsPbIBr2 film (PVSK phase) displayed an orange- red color. The synthesized CsPbIBr2 film could maintain its color in an inert or low humidity (<50%) environment when quickly cooled down to room temperature. Lin et.al. have reported the reversible phase transition of solution-cast CsPbIxBr3-x films between the room-temperature non-perovskite phase and high-temperature perovskite phase.[11] A similar phenomenon was observed in the vapor-deposited CsPbIBr2 film. After exposed to an atmosphere with 80% relative humidity (RH) for 30 minutes, the orange-red CsPbIBr2 film gradually turned to be transparent. Reheating the transparent CsPbIBr2 film (non- PVSK phase) around 160 ºC in a N2-filled glovebox triggered the color-change from transparent back to orange-red again. The reversible phase transition of CsPbIBr2 is repeatable and the optical properties in both phases remained consistent after 20 cycles, as discussed below. Figure 2b presents the absorption spectra of the non-PVSK and PVSK phase CsPbIBr2 films, showing the latter have significantly stronger optical absorption. Furthermore, the PVSK phase film shows an absorption edge at around 600 nm, while this is blue-shifted to 430 nm for the non-PVSK phase film. Such a distinction can be used to distinguish CsPbIBr2 films in these two different phases. The reversibility of structural transitions between non-PVSK and PVSK phases was monitored by measuring the absorption spectra after repetitive moisture exposure and heating. Figure 2c shows the absorption of CsPbIBr2 films in the two phases at 500 nm. The data exhibits good consistency over 20 repetitive moisture/heat treatment cycles. The X-ray diffraction (XRD) technique was used to analyze the crystal structural change in more details (Figure 2d). We notice that the PVSK phase CsPbIBr2 film exhibits characteristic peaks located at 14.86º, 21.24º, 30.0º, corresponding to the (100), (110), (200) planes of the CsPbIBr2 cubic structure. This confirms that PbBr2 and CsI layers have reacted to form one cubic-phase perovskite layer through an inter-diffusion process. We also observe a small peak (denoted by ) located at 11.4°, indicative of the remaining PbBr2, which is similar to the finding from Hutter et.al.[9] The XRD pattern of the moisture- exposed CsPbIBr2 film displays main peaks at 10.3º, 13.7º, 18.6º, 23.6º, 27.6º, 28.8º, corresponding to (002), (012), (021), (023), (015), (032) planes of the CsPbIBr2 orthorhombic phase.[23] Furthermore, we compared the XRD patterns of CsPbIBr2 films in the two phases before and after phase-transition cycles. No obvious difference was observed for both phases after 20 cycles (Figure S4). The different crystal structures of CsPbIBr2 films in non-PVSK and PVSK phases also results in noticeable refractive index difference. We measured the complex refractive index using an ellipsometry, Figure 2e shows the real part (n) and imaginary part (k) of the complex refractive index of non-PVSK and PVSK phase CsPbIBr2 films. The k value of the PVSK phase film is almost zero after 600 nm, indicating no optical loss for λ >600 nm. In contrast, the k value of the non-PVSK phase film cuts off at ~430 nm. These results are consistent with the bandgap and absorption spectra in Figure 2b. The large difference of refractive index between two phases in the visible light region may have potential applications in optical modulators, switches, tunable metasurfaces and other photonic devices.[24-25] Figure 2f presents the current-voltage (I-V) characteristics of a hole-only device (ITO/PEDOT:PSS/CsPbIBr2/Au), the CsPbIBr2 layer is either in the non-PVSK or PVSK phase. Both I-V curve (I ~ Vn) shows the ohmic (n = 1), trap filling (n > 2) and child regions (n = 2). The trap filled voltage (VTFL) of the non-PVSK phase film was higher than that of the PVSK phase film, indicating a higher trap density (ntrap).[26-27] This may be attributed to the release of halide atoms and increase of halide vacancies caused by the interaction of water molecules with Pb atoms.[28-29] 3. The patterned non-PVSK to PVSK phase transition through laser direct writing Considering current technological trends toward flexible and wearable devices, a robust processing method such as heating for large-area flexible substrates is becoming more and more important.[30] Laser heating is a promising candidate for such an application. Compared to other more common thermal heating methods, laser heating has the advantage of area-selective rapid heating and cooling, better compatibility with flexible substrates, and processing scalability.[31-33] In this work, we employed a near infrared (1064 nm, Nd:YVO4) laser for laser direct writing (LDW) of the non-PVSK phase CsPbIBr2 film deposited on an interfacial layer (ITO or Ag). This method can precisely control the heating temperature and crystallization process without damage by the laser beam.[34] Figure 3a presents the schematic configuration of the LDW experimental setup. The sample was placed on a XY translation stage and driven laterally, a 1064 nm laser beam was focused onto the sample and laterally scanned the non-PVSK phase CsPbIBr2 film on ITO. Two orange-red stripes were formed as shown in the inset photo of figure 3a, indicating an instantaneous transformation of non-PVSK to PVSK phase triggered by the laser beam. Figure 3b displays the optical (top) and fluorescent (bottom) microscopy images of a LDW pattern on the perovskite film. The laser beam was focused into a ~ 50 μm spot and scanned to form patterns. The PL of the PVSK phase area is much stronger than that of the non- PVSK phase area, which is clearly manifested in these fluorescent images. Figure 3c presents the line-scan PL intensity of the micro-line array pattern in Figure 3b, demonstrating the uniform periodic fluorescence emitted from the PVSK phase micro-line array. Raman spectroscopy has high positional and spectral accuracy, it was used here to investigate the phase mapping of the LDW patterns. In Figure 3d, the top Raman spectrum (blue curve) corresponds to the micro-lines with strong red fluorescence. A prominent peak at 138 cm-1 is observed, which is attributed to the PVSK phase.[35] On the other hand, Raman spectrum of the dark background area (bottom black curve) shows a red-shifted and broader peak at 114 cm-1, which corresponds to the non-PVSK phase.[32] The Raman measurement finding matches well with other reported literature,[36-38] further demonstrating the phase transformation induced by LDW. The COMSOL simulated temperature distribution around a NIR laser spot (1 um diameter) shown in Figure 3e clarifies the underlying mechanism of photothermal heating. The NIR laser beam is not absorbed by the perovskite layer due to its wavelength longer than the absorption edge, and the photothermal effect from the interfacial layer (ITO) contributes to rapid local heating. A relative low laser power (9 mW) could lead a temperature rise of 160 ºC localized to the focused beam position, which is hot enough to locally induce the phase transition. Around the focused laser spot, the temperature decreased to half of the maximum at a distance 1.4 μm from the center (see Figure S6 for thermal simulation in details). The increased surface temperature as a function of laser intensity was measured using an IR thermal camera for both Ag and ITO as the interfacial layer (see Figure S7a). The temperature rise strongly depends on the material of the interfacial layer, and it was observed that the Ag layer induced more photothermal heating and temperature rise than ITO at the same laser intensity. To induce phase transition from the non-PVSK to PVSK phase (160 °C), the laser intensity for Ag and ITO is ~115 and ~235 W/cm2 respectively. 4. Optical memory device for data storage enabled by non-volatile phase transitions The as-fabricated PVSK phase PD array was first exposed to water moisture with relative humidity of 80% until the perovskite film became transparent. As schematically illustrated in Figure 4a, one PD pixel was then heated by the focused laser beam selectively. The induced photothermal heating led to an instantaneous temperature rise at specific pixels, and a color change from transparent to orange-red was observed within a second. The inset photo in Figure 4a is a fluorescent image of one pixel heated by the laser beam. The uniform fluorescence inside the IDE demonstrate the occurrence of phase transformation in that pixel. We noticed that the fluorescence is weaker directly on top of the Ag electrode area, which is attributed to PL quenching effect by Ag. We then measured the current-voltage (I-V) curves of both non-PVSK and PVSK phase PD pixels in dark condition and under white light illumination (tungsten-halogen lamp), the results are presented in Figure 4b. PDs in two phases exhibit similar dark current (Idark) on the level of 10-9 A at 5 V. The non-PVSK phase PD has slightly higher dark current, which may be attributed to the higher conductivity of non-PVSK phase films as shown in Figure 2f. Upon white light illumination, the photoinduced carriers contributed to the current increase of PDs in both phases. However, the light current (Ilight) of PDs in PVSK phase was ~50 times larger than that of the PDs in non-PVSK phase when operated at a bias of 5 V. The large difference of photocurrents between PDs in two phases is attributed to the different crystal structures and optical absorptions. The responsivity (R), a key figure of merit for a PD, is given by[39] 𝑅 = 𝐼𝑝ℎ 𝐸𝑒𝑆 = 𝐼𝑙𝑖𝑔ℎ𝑡 − 𝐼𝑑𝑎𝑟𝑘 𝐸𝑒𝑆 where Ee and S are the illumination intensity and the single PD device area, respectively. As shown in Figure 4c, a large difference of spectral responsivity between PDs in the two phases is observed. The responsivity spectrum shows a cut-off at 430 nm and 600 nm for the non-PVSK and PVSK phase PDs, respectively. Since the PVSK phase PDs have much better photodetection performance in terms of responsivity and photodetection wavelength range compared to the non-PVSK phase PDs, we then focused on systematically investigating the PVSK phase PD performance. The I-V curves of the PVSK phase PDs under 405 nm irradiation with light intensity varying from 0 to 50 mW/cm2 are shown in Figure 4d, the current shows linear relationship with voltage. Response speed is another key parameter for photodetectors. Figure 4e exhibits the time-dependent photocurrent response under periodically modulated UV light illumination. The PDs show stable and reproducible photo-response behaviors under periodic on/off switching cycles. The rise time (defined as the time for the photocurrent to increase from 10% to 90% of the peak value) and the fall time (from 90% to 10% of the peak value) are extracted to be 0.8 and 1 ms (inset of Figure 4f), respectively. We then measured the 3dB frequency bandwidth by recording photocurrent as a function of the modulated light frequency, as presented in Figure 4f. The 3dB bandwidth f3dB is found to be ~400 Hz, which is consistent with theoretical value calculated by f3dB = 0.35 / trise. As discussed above, the crystal structure difference between the two phases leads to dramatically different photo-response under white light irradiation, which can be used to distinguish these two phases. Considering both non-PVSK and PVSK phases are nonvolatile in moderate humidity (RH<50%) environment, the CsPbIBr2 PD array can be potentially applied as a rewritable optical memory device. Figure 5a schematically illustrates the working mechanism of this application. The laser beam directly writes the individual PD pixels to transform the non-PVSK phase to PVSK phase, which represents the 'write' or 'set' process. The moisture exposure triggers the phase transition from the PVSK to non-PVSK phase again, which is the 'erase' or 'reset' process. Mapping of the pixels that have been written (set) can be obtained through reading the photocurrent of each PD pixel under a large-area white light irradiation (the 'read' process). The pixel with a high photocurrent stores '1' and the pixel with a low photocurrent stores '0', then the optical information encoded by the laser beam writing can be restored. Figure S7b (see Supporting Information) shows a PD array photo where the first-row pixels (8 pixels) were heated by NIR laser beam and displayed orange-red color. We also characterized the set/reset cycle endurance and the data retention ability of the PD pixels as optical memories. The photocurrents of the perovskite PD pixels over repetitive laser heating/moisture exposure cycles are presented in Figure 5b. The ratio of the photocurrent from the PD pixel in the PVSK phase to that in the non-PVSK phase was observed to maintain above 50 over 10 cycles. This is consistent with the reversible absorption properties of the perovskite layer shown in Figure 2c and is high enough to distinguish the '1' state from the '0' state. The data retention capability was verified by recording the photocurrent continuously over time. As shown in Figure 5c, the photocurrents of the PD pixels in both phases remain relatively stable. About 20% degradation for the PVSK phase PD was observed after being stored for 16 hours in our lab (RH 45%). The stability can be further improved by reducing the environment humidity during test. These findings show that the perovskite PD array can be used to conduct write- read-erase cycles many times, suggesting the potential of utilizing the non-volatile phase- change properties of vapor-deposited CsPbIBr2 for optical memory and information storage applications. 5. Imaging application of the flexible perovskite PD arrays Through inkjet printing and vapor deposition technologies, high-quality perovskite PD arrays can be fabricated on large-area and flexible substrates without requiring extra-high annealing temperature. These advantages can enable scalable device fabrication and facilitate commercialization of flexible perovskite PD arrays for image sensor applications. Herein, we use a 16×16 PD pixel array to verify the imaging capability as proof-of-concept demonstration. The optical image of the 16×16 PD pixel array is presented in Figure S9b. The as-deposited flexible PD array on PET through the vapor-deposition process was tested without further annealing. In an array of PDs, it is important to examine the uniformity of the photo-response from all 16×16 pixels. A uniform white light source was projected onto the PD array, and the photo-response from all 256 pixels was measured and mapped. Figure 6a presents the photocurrent distribution along selected four lines. The photocurrent fits in a tight range of 0.14 to 0.15 μA, with only 2% fluctuation relative to the average value, demonstrating excellent uniformity over all PD pixels. The imaging capability of the flexible perovskite PD array was verified by reconstructing a projected optical pattern. Figure 6b shows the experimental setup composed of a collimated white light source, a photomask with the desired pattern and the flexible perovskite PD array device. After scan-measuring the light current of each pixel, we obtain the light current distribution of all 256 pixels. Each pixel is given a gray scale number (G) between 0 and 1 based on the measured light current (Imeas) through the following equation.[40] 𝐺 = 𝐼meas − 𝐼dark 𝐼max − 𝐼dark Where Idark is the average dark current and Imax is the maximum light current of all 256 pixels. As shown in Figure 6c, a "UW" pattern was projected onto the PD array and retrieved from the photocurrent mapping. The performance of the flexible perovskite PD array was also characterized at different bending angles and after various numbers of bending cycles. The results show little degradation after bending (Figure S8), supporting the promise of vapor-deposited perovskite films for flexible devices. Since the response speed (trise= 0.8 ms) of our vapor-deposited perovskite PD arrays is faster than the human-eye recognition speed (>40 ms),[20] our demonstrations open up the possibility of using vapor-deposited inorganic perovskite thin film and inkjet printing technology for flexible, large-scale, high-speed image acquisition in a broad spectrum of applications such as wearable electronic eyes and cameras. Conclusions In summary, we have demonstrated uniform inorganic perovskite films with excellent optoelectronic properties grown by a vapor-deposition process. The fabrication method can be easily implemented with various substrates. The reversible phase transition between non-PVSK and PVSK phases was observed and confirmed by XRD. The two phases manifest significant differences in optical absorption, refractive index and other optoelectronic properties. Utilizing the phase-change properties, the fabricated perovskite PD array can be applied as a rewritable optical memory device for information storage. Finally, we demonstrated optical imaging using flexible perovskite PD array devices on PET substrates. These works suggest the versatile utilities of vapor-deposited inorganic perovskite films through the unique reversible phase-changing features and the potential for fabrication of flexible, large-area devices. Experimental Section Materials: CsI (99.9% trace metal purity), PbBr2 (99.99% trace metal purity) and the Ag nanoparticle ink (Silverjet DGP-40LT-15C) for inkjet printing were purchased from Sigma-Aldrich and used without further purification. The evaporation crucibles were purchased from Kurt J. Lesker. Inkjet printing Ag electrodes: PET and glass substrates were first cleaned by ultrasonication with 2-propanol and DI water in sequence. The Ag nanoparticle ink was loaded into the cartridge of an inkjet printer (Diamtrix), the cartridge was ultrasonicated briefly to reduce the aggregation of Ag nanoparticles. The voltage of printing nozzle was set at 23 V, the waveform frequency was set at 5 kHz. The cleaned substrates were well adhered to the printing platform using vacuum, the platform temperature was set at room temperature. The printing nozzle was controlled by the computer with the designed pattern to drop the Ag ink with a proper drop spacing distance (30 μm for PET and 40 μm for glass). The substrates were then heated in oven at 110 °C for 10 minutes to sinter the Ag electrodes. Perovskite vapor deposition: Inorganic perovskite CsPbIxBr3-x was deposited by dual source thermal deposition method. For CsPbIBr2 deposition, CsI and PbBr2 powder are placed into two crucibles separately. The vacuum chamber was pumped down to 10-6 Torr. The evaporation rates of CsI and PbBr2 were 0.53 and 0.5 A/s based on the calculation from their molecular mass and density. Two crucibles were heated simultaneous to appropriate temperature and the evaporation rates were monitored by two quartz crystal microbalances (QCMs) separately, the actual evaporation rates during the deposition process were controlled within 20% away from set values. Two QCMs were calibrated separately before deposition to determine the tooling factors for CsI and PbBr2. Laser direct writing process: Laser heating was performed using our homemade system based on a Nd:YVO4 1064 nm laser source (Spectra Physics). The sample was placed on a XY translation stage, the laser beam was focused through a 10x objective lens at the specific location of the sample. The photothermal effect on ITO or Ag layer induced by laser beam could increase the localized temperature and achieve the phase transition. Reversible phase transition of CsPbIBr2: The non-PVSK phase CsPbIBr2 film was heated on hotplate (160 °C) for 10 minutes or by the 1064 nm laser beam to transform into the PVSK phase CsPbIBr2 film. To convert the PVSK phase to non-PVSK phase, the CsPbIBr2 film was placed in a humidity control box with relative humidity of 80 % for 30 minutes. The reversible phase conversion process was carried out 20 times. Perovskite thin film characterization: The absorption spectra of perovskite films were acquired using a Varian Cary 5000 UV-vis-NIR Spectrophotometer. XRD patterns were obtained using Bruker D8 with a Cu Kα radiation (λ = 1.54184 Å). The surface morphology of perovskite films was measured by SEM (FEI Sirion) and AFM (Bruker Icon). The refractive index was measured using an ellipsometer (J.A. Woollam M2000), the raw data was fitted by the B-spline model on a Cauchy dispersion glass substrate using Complete EASE software. Raman spectra were acquired by a Raman microscope (InVia, Renishaw), 785 nm laser was used as the excitation source to avoid the interruption of fluorescence from sample. The perovskite was deposited on the inkjet-printed Ag layer to enhance the Raman scattering intensity. The fluorescent images were obtained through a fluorescence microscope (EVOS, Thermo Fisher Scientific), the raw images were processed by ImageJ software to adjust the threshold intensity and remove background signal. Perovskite PD array measurement: Current-voltage characteristics of perovskite photodetectors both in dark and under light illumination were measured by a source meter (Keithley 6430). The irradiation light was provided by a 405 nm laser source, the light intensity could be easily adjusted. The time-dependent current response was recorded under periodically on-off light illumination from a 405 nm laser source driven by the square wave from a signal generator. The spectral EQE measurement was performed by using a tungsten-halogen light source to illuminate the device through a monochromator (Acton Research SpectraPro 275), which provides the monochromatic light excitation. The frequency response of the photodetector was measured by modulating the 405 nm laser light at different frequencies using an optical chopper. The generated photocurrent was recorded using a lock-in amplifier (Stanford Research SRS 830). For image sensor application, the perovskite PD array was placed underneath an aluminum mask with pre-designed pattern. The collimated white light from a tungsten-halogen lamp went through the pattern area of the mask and arrives at the PD array platform. The currents of 16×16 PD pixels were recorded and mapped. Author Information Corresponding author *Email: [email protected] ORCID Chen Zou: 0000-0001-9638-6363 Lih Y. Lin: 0000-0001-9748-5478 Notes The authors declare no competing financial interest. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgement This work is supported in part by National Science Foundation (grant ECCS-1807397 and CHE-1836500). Part of the work was conducted at the Molecular Analysis Facility, a National Nanotechnology Coordinated Infrastructure site at the University of Washington which is supported in part by the National Science Foundation (grant NNCI-1542101), the University of Washington, the Molecular Engineering & Sciences Institute, the Clean Energy Institute, and the National Institutes of Health. C. Z. would like to thank Felippe J. Pavinatto from the UW Clean Energy Institute Testbed (WCET) for discussion and help with inkjet printing. 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)) References S. D. Stranks, H. J. Snaith, Nat. Nanotechnol. 2015, 10, 391. E. M. Hutter, R. J. Sutton, S. Chandrashekar, M. Abdi-Jalebi, S. D. Stranks, H. J. B. R. Sutherland, E. H. Sargent, Nat. Photonics 2016, 10, 295. T. M. Brenner, D. A. Egger, L. Kronik, G. Hodes, D. Cahen, Nat. Rev. Mater. 2016, [1] [2] 1. A. Kojima, K. Teshima, Y. Shirai, T. Miyasaka, J. Am. Chem. Soc. 2009, 131, 6050. [3] C. S. Ponseca, T. J. Savenije, M. Abdellah, K. Zheng, A. Yartsev, T. Pascher, T. [4] Harlang, P. Chabera, T. Pullerits, A. Stepanov, J.-P. Wolf, V. Sundström, J. Am. Chem. Soc. 2014, 136, 5189. [5] M. A. Green, A. Ho-Baillie, H. J. Snaith, Nat. Photonics 2014, 8, 506. [6] W. S. Yang, B. W. Park, E. H. Jung, N. J. Jeon, Y. C. Kim, D. U. Lee, S. S. Shin, J. Seo, E. K. Kim, J. H. Noh, S. I. Seok, Science 2017, 356, 1376. [7] [8] M. E. Ziffer, D. S. Ginger, Science 2016, 353, 1365. [9] Snaith, T. J. Sayenije, ACS Energy Lett. 2017, 2, 1901. [10] R. L. Milot, G. E. Eperon, H. J. Snaith, M. B. Johnston, L. M. Herz, Adv. Funct. Mater. 2015, 25, 6218. [11] J. Lin, M. L. Lai, L. T. Dou, C. S. Kley, H. Chen, F. Peng, J. L. Sun, D. L. Lu, S. A. Hawks, C. L. Xie, F. Cui, A. P. Alivisatos, D. T. Limmer, P. D. Yang, Nat. Mater. 2018, 17, 261. [12] Y. Rong, L. Liu, A. Mei, X. Li, H. Han, Adv. Energy Mater. 2015, 5, 1501066. [13] R. J. Sutton, G. E. Eperon, L. Miranda, E. S. Parrott, B. A. Kamino, J. B. Patel, M. T. Horantner, M. B. Johnston, A. A. Haghighirad, D. T. Moore, H. J. Snaith, Adv. Energy Mater. 2016, 6, 1502458. [14] G. E. Eperon, G. M. Paterno, R. J. Sutton, A. Zampetti, A. A. Haghighirad, F. Cacialli, H. J. Snaith, J. Mater. Chem. A 2015, 3, 19688. [15] M. Kulbak, D. Cahen, G. Hodes, J. Phys. Chem. Lett. 2015, 6, 2452. [16] J. Am. Chem. Soc. 2016, 138, 15829. [17] A. Swarnkar, A. R. Marshall, E. M. Sanehira, B. D. Chernomordik, D. T. Moore, J. A. Christians, T. Chakrabarti, J. M. Luther, Science 2016, 354, 92. [18] C. Y. Huang, C. Zou, C. Y. Mao, K. L. Corp, Y. C. Yao, Y. J. Lee, C. W. Schlenker, A. K. Y. Jen, L. Y. Lin, ACS Photonics 2017, 4, 2281. [19] C. Y. Chen, H. Y. Lin, K. M. Chiang, W. L. Tsai, Y. C. Huang, C. S. Tsao, H. W. Lin, Adv. Mater. 2017, 29. [20] G. Tong, H. Li, D. Li, Z. Zhu, E. Xu, G. Li, L. Yu, J. Xu, Y. Jiang, Small 2018, 14, 1702523. [21] Y. Hu, Q. Wang, Y.-L. Shi, M. Li, L. Zhang, Z.-K. Wang, L.-S. Liao, J. Mater. Chem. C 2017, 5, 8144. [22] W. Fang, J. Yang, J. Gong, N. Zheng, Adv. Funct. Mater. 2012, 22, 842. [23] M. Lai, Q. Kong, C. G. Bischak, Y. Yu, L. Dou, S. W. Eaton, N. S. Ginsberg, P. Yang, Nano Research 2017, 10, 1107. [24] J. Zheng, A. Khanolkar, P. Xu, S. Colburn, S. Deshmukh, J. Myers, J. Frantz, E. Pop, J. Hendrickson, J. Doylend, N. Boechler, A. Majumdar, Opt Mater Express 2018, 8, 1551. J. Liang, C. Wang, Y. Wang, Z. Xu, Z. Lu, Y. Ma, H. Zhu, Y. Hu, C. Xiao, X. Yi, J. A. Steele, H. F. Yuan, C. Y. X. Tan, M. Keshavarz, C. Steuwe, M. B. J. Roeffaers, I. Konidakis, T. Maksudov, E. Serpetzoglou, G. Kakavelakis, E. Kymakis, E. [25] C.-H. Liu, J. Zheng, S. Colburn, T. Fryett, Y. Chen, X. Xu, A. Majumdar, Nano Lett. 2018, 18, 6961. [26] Q. Dong, Y. Fang, Y. Shao, P. Mulligan, J. Qiu, L. Cao, J. Huang, Science 2015, 347, 967. [27] M. I. Saidaminov, A. L. Abdelhady, B. Murali, E. Alarousu, V. M. Burlakov, W. Peng, I. Dursun, L. F. Wang, Y. He, G. Maculan, A. Goriely, T. Wu, O. F. Mohammed, O. M. Bakr, Nat. Commun. 2015, 6, 7586. [28] E. Mosconi, J. M. Azpiroz, F. De Angelis, Chem. Mater. 2015, 27, 4885. [29] A. Mattoni, A. Filippetti, C. Caddeo, J. Phys. Condens. Matter 2016, 29, 043001. [30] S. S. Chou, B. S. Swartzentruber, M. T. Janish, K. C. Meyer, L. B. Biedermann, S. Okur, D. B. Burckel, C. B. Carter, B. Kaehr, J. Phys. Chem. Lett. 2016, 7, 3736. [31] S. J. Kim, J. Byun, T. Jeon, H. M. Jin, H. R. Hong, S. O. Kim, ACS Appl. Mater. Interfaces 2018, 10, 2490. [32] J. Hofkens, ACS Nano 2017, 11, 8072. [33] Stratakis, ACS Appl. Energy Mater. 2018, 1, 5101. [34] T. Jeon, H. M. Jin, S. H. Lee, J. M. Lee, H. I. Park, M. K. Kim, K. J. Lee, B. Shin, S. O. Kim, ACS Nano 2016, 10, 7907. [35] S. M. Jain, B. Philippe, E. M. Johansson, B.-w. Park, H. Rensmo, T. Edvinsson, G. Boschloo, J. Mater. Chem. A 2016, 4, 2630. [36] L. Q. Xie, T. Y. Zhang, L. Chen, N. Guo, Y. Wang, G. K. Liu, J. R. Wang, J. Z. Zhou, J. W. Yan, Y. X. Zhao, B. W. Mao, Z. Q. Tian, Phys. Chem. Chem. Phys. 2016, 18, 18112. [37] M. Ledinsky, P. Loper, B. Niesen, J. Holovsky, S. J. Moon, J. H. Yum, S. De Wolf, A. Fejfar, C. Ballif, J. Phys. Chem. Lett. 2015, 6, 401. [38] R. G. Niemann, A. G. Kontos, D. Palles, E. I. Kamitsos, A. Kaltzoglou, F. Brivio, P. Falaras, P. J. Cameron, J. Phys. Chem. C 2016, 120, 2509. [39] C. Zou, Y. Xi, C. Y. Huang, E. G. Keeler, T. Feng, S. Zhu, L. D. Pozzo, L. Y. Lin, Adv. Opt. Mater. 2018, 6, 1800324. [40] L. Gu, M. M. Tavakoli, D. Zhang, Q. Zhang, A. Waleed, Y. Xiao, K. H. Tsui, Y. Lin, L. Liao, J. Wang, Adv. Mater. 2016, 28, 9713. Figure Captions Figure 1. Fabrication and characterization of all-inorganic CsPbIBr2 perovskite thin films. a) The schematic fabrication procedure of all-inorganic perovskite photodetector array via inkjet printing technology and dual- source thermal vapor-deposition method b) The SEM image of the CsPbIBr2 film with a scale bar of 500 nm. c)The AFM image of the CsPbIBr2 film, the RMS roughness is 4.1 nm. d) The optical image of a vapor- deposited perovskite dot array on a flexible PET substrate patterned through a shadow mask. Figure 2. Phase transitions of inorganic perovskite thin films via laser heating and moisture exposure, taking CsPbIBr2 for example. a) Optical images of PVSK (orange-red color) and non-PVSK phase (transparent) CsPbIBr2 thin films. The reversible phase transitions can be achieved by heating/moisture exposure. b) The absorption spectra of PVSK and non-PVSK CsPbIBr2 films. c) The reversible and repeatable switching behavior of the absorption (at 500 nm) of the CsPbIBr2 thin film over 20 phase transition cycles. d) XRD patterns and e) complex refractive index spectra of PVSK and non-PVSK CsPbIBr2 films. f) I-V curves of hole only devices (ITO/PEDOT:PSS/CsPbIBr2/Au) where the CsPbIBr2 layer is either in the non-PVSK or PVSK phase. The I-V curve is fitted by I ~ Vn, the red, green and yellow fitting line represents n = 1, n > 2 and n = 2 respectively. Figure 3. Laser direct writing (LDW) of the perovskite films assisted by photothermal effect. a) The schematic illustration for photothermal heating at the interfacial layer (ITO or Ag). The inset shows the photo of a perovskite thin film with two stripes converted to the PVSK phase through LDW. b) Optical (top) and fluorescent (bottom) microscopy images of the micro-lines pattern by LDW. c) The line-scan PL intensity of the fluorescent image in b). d) The measured Raman spectra at different positions. The two distinct spectra correspond to the non-PVSK phase (black curve) and the PVSK phase (blue curve). e) COMSOL Multiphysics thermal simulation of heat distribution induced by a laser beam (9 mW) with a spot size of 1 um. Figure 4. Detection performance of inorganic perovskite photodetector arrays. a) The schematic illustration of a single perovskite PD pixel heated by a 1064 nm laser. The inset photo is the fluorescent image of the pixel after transitioning to the PVSK phase. b) I-V curves of CsPbIBr2 PD pixels in the non-PVSK phase and PVSK phase, under dark and white light illumination. c) The responsivity spectra of the perovskite PD pixel in two phases. d) I-V curves of the perovskite PD in the PVSK phase under 405 nm light illumination with different intensities. e) The time-dependent photocurrent switching behavior, and f) normalized frequency response of the photodetector in the PVSK phase. The inset shows the magnified photoresponse under 20 Hz light modulation. Figure 5. The optical memory application using the perovskite photodetector arrays enabled by the phase transition between the non-volatile PVSK and non-PVSK phases. a) Schematic illustration of the working principle. b) The photocurrents of perovskite PD pixels in PVSK and non-PVSK phases during 10 cycles of laser heating/moisture exposure test. c) Data retention capability of the non-volatile PVSK and non-PVSK phase perovskite PDs. The photocurrent only decreases to 80% of the original value for the PVSK phase after 1000 minutes under 45% relative humidity. Figure 6. Imaging application of the flexible perovskite PD arrays. a) The line-scan photocurrent profiles measured from four randomly selected lines in a 16x16 array. b) The schematic setup of the imaging experiment. An optical pattern was projected onto the PD array through a prepatterned mask using a collimated white light source. c) a 'UW' pattern reconstructed by the PD array through mapping the photocurrent from each pixel.
1802.02039	1	1802	2017-12-25T13:36:24	Combined Oxy-fuel Magnetohydrodynamic Power Cycle	[ "physics.app-ph" ]	Oxy-fuel carbon capture in power plants is a relatively new concept aiming at reducing carbon dioxide emissions from the plants. This is achieved by burning the fossil fuel using oxygen as oxidizer with no nitrogen, thereby rendering the exhaust gases very rich in carbon dioxide (after condensing water vapor by cooling), which facilitates its capture for environmental or commercial purposes. Despite the worldwide interest in oxy-fuel carbon capture, its progress is at risk given the large energy needed to separate oxygen from air in order to provide the oxidizer, thereby hindering further progress of this concept toward large-scale applications. This paper focuses on alleviating this drawback of oxy-fuel combustion by making it more attractive through combining it with another concept, namely magnetohydrodynamic (MHD) power generators. The end product is a power plant operating on a combined cycle composed of a topping MHD ultra-high-temperature cycle with direct electricity extraction from plasma, followed by a bottoming steam cycle with conventional turbo-generators. Different design aspects and simplified technical analysis for the MHD generator are presented.	physics.app-ph	physics	Conference on energy Challenges in Oman 09-10 December, 2015, Salalah, Sultanate of Oman ECO'2015 Combined Oxy-fuel Magnetohydrodynamic Power Cycle University of Buraimi, college of engineering, Mechanical department Osama A. Marzouk Al Buraimi, Sultanate of Oman [email protected] Abstract Oxy-fuel carbon capture in power plants is a relatively new concept aiming at reducing carbon dioxide emissions from the plants. This is achieved by burning the fossil fuel using oxygen as oxidizer with no nitrogen, thereby rendering the exhaust gases very rich in carbon dioxide (after condensing water vapor by cooling), which facilitates its capture for environmental or commercial purposes. Despite the worldwide interest in oxy-fuel carbon capture, its progress is at risk given the large energy needed to separate oxygen from air in order to provide the oxidizer, thereby hindering further progress of this concept toward large-scale applications. This paper focuses on alleviating this drawback of oxy-fuel combustion by making it more attractive through combining it with another concept, namely magnetohydrodynamic (MHD) power generators. The end product is a power plant operating on a combined cycle composed of a topping MHD ultra-high-temperature cycle with direct electricity extraction from plasma, followed by a bottoming steam cycle with conventional turbo-generators. Different design aspects and simplified technical analysis for the MHD generator are presented. Keywords: MHD; magnetohydrodynamics; oxy-fuel; carbon capture; plasma; power plant Introduction Magnetohydrodynamic power generators (Messerle, 1995; Kayukawa, 2004; Woodside et al., 2012) are chambers where high-speed electrically-conducting hot plasma expands while subject to an applied magnetic field. This leads to the generation of direct-current electricity based on Faraday's law of induction without the conventional turbo-generators. MHD requires ultra-high temperatures (near 3000 K) for attaining reasonably-high levels of electrical conductivity of the plasma. Oxy-fuel combustion (Wall et al., 2009; Zheng and Tan, 2014) is another concept where fossil fuels are burnt using a nitrogen-free oxidizer, leading to ultra-high temperature gases which can be made rich in CO2 after condensing their H2O content through cooling. The CO2-rich flue gas in oxy-fuel combustion facilitates CO2 capture (hence the term 'carbon capture') and then sequestering it in underground formations instead of releasing it into the atmosphere, in an attempt to mitigate the massive emissions of this primary greenhouse gas from anthropogenic activities (Liu et al., 2012). Moreover, the captured CO2 can be sold commercially for later use in enhanced oil recovery by injecting supercritical dense-phase CO2 into the oilfield to reduce the viscosity of the crude oil and enhance its extraction. The ultra-high temperature is thus a common feature between the MHD concept and the oxy-fuel concept, raising the question of combining them if a benefit can be achieved as a consequence. A main drawback of oxy-fuel combustion as a concept to be used in steam power plants is the large energy penalty incurred by separating oxygen from air. This problem is so important that the giant Swedish energy company Vattenfall, which previously announced long-term plans for research investments in oxy-fuel combustion at demonstration scale and commercial scale, has decided in May 2014 to stop all effort on oxy-fuel carbon capture due to "its costs and the energy it requires makes the technology unviable". The company gave priority to other R&D projects "which can contribute more quickly to our business development". This decision aborted an already ongoing project for building a 250 MWe demonstration plant in Jänschwalde (Germany) to be commissioned in 2016, despite the successful commissioning of a smaller oxy-fuel demonstration plant (to produce steam for nearby industry rather than electricity) with 30 MWth capacity at the Schwarze Pumpe power station (Germany) since September 2008 (Anheden et al., 2011). On the other hand, a main drawback of the Marzouk 1 MHD concept as a potential concept for power generation is that it exploits energy only in the very high temperature spectrum, leaving hot gases with huge amounts of unexploited energy. However, having a very high working-fluid temperature in MHD generators enables extremely high levels of energy extraction from this working-fluid (the plasma), which mitigates the energy consumption in air separation units (ASU) for oxy-fuel combustion. In the same time, steam cycle can be used to extract energy from the still-hot MHD exhaust gases, where these gases operate the steam generator to provide the necessary steam for the steam turbine without using a furnace. Thus, the steam cycle handles the extraction of the energy from the working-fluid at the lower temperature spectrum (about 2500 K and below). This releases the MHD limitations and makes it a viable concept for commercial power plants, and in the same time makes oxy-fuel combustion more attractive with the mitigated effect of the air separation cost. What is a Magnetohydrodynamic Generator? Figure 1 illustrates a typical design of an MHD generator. It also shows the Fleming's right-hand rule, governing the direction of the induced current due to the motion of the conducting plasma within the applied magnetic field. The generator consists of 1) Combustor: • The fuel is burnt with O2, plasma is generated. 2) Convergent passage: • Plasma accelerates to sonic speed at the throat. 3) Channel: Plasma continues to accelerate with supersonic speeds. • This is the heart of the MHD generator. • • Electrodes (anode and cathode) are fitted to two opposite walls. • An external magnetic field is applied in the bi-normal direction (normal to both the electrodes and the bulk motion of the plasma). • Energy is extracted from plasma; electric current is collected through the electrodes. 4) Convergent diffuser (not shown): • To slow down the exploited plasma. Figure 1. Ilustration of MHD Generator. While the use of a divergent channel with supersonic plasma is not strictly mandated, it is highly demanded because the power output from the MHD generator is proportional to the square of the plasma velocity. In addition, as energy is extracted from the plasma, it cools down near the rear section of the MHD channel, thereby dropping its electrical conductivity in a sharp manner. This should be compensated by an increase in the plasma velocity, which is again achieved by a supersonic divergent channel where the plasma continues to accelerate downstream the channel despite the loss of enthalpic energy. Marzouk 2 Figure 2. Simple 1-D MHD Channel. Figure 2 further demonstrates the working principle of a simplified one-dimensional MHD channel, having a constant cross-sectional area with a fixed spacing( ) between parallel electrodes. The plasma moves in the axial direction (x-axis) with a velocity( = ), the external magnetic field × =− = ) connected externally to the electrode). The external load is represented by a resistance ( is in the negative y-axis, going from the anode (upper, negative electrode) to the cathode (lower, positive is applied in the z-direction, and the induced electric field electrodes. Plasma Formation for MHD Generator While the hot gases coming from the MHD combustor posses very high temperatures (e.g., 3200 K) that enable partial ionization, the electrical conductivity of these gases (mainly H2O and CO2) will still be negligible because of the relatively-large ionization energy for their molecules. To remedy this, combustion products are augmented with the vapor of an alkali metal (such as cesium: Cs, or potassium: K) at a small mole fraction of about 1%. The alkali metal can be introduced into the combustor in a form of seeded salt compound, such as potassium carbonate (K2CO3). It should be regenerated and re-used for economically-feasible operation. In fact, cesium has the lowest ionization energy among elements, meaning that it is the easiest element to lose an electron and form an ion. On the other hand, helium (He) is the element having the largest ionization energy. Table 1 compares the ionization energies of cesium, potassium, hydrogen, oxygen, and helium (Kramida et al., 2014). Whereas cesium has lower ionization energy than potassium, it is much rarer. element Cesium (Cs) Potassium (K) Hydrogen (H) or Oxygen (O) Helium (He) Ionization energy (eV/atom) Ionization energy (kJ/mol) 3.89 4.34 13.6 24.6 375 418 1310 2370 Table 1. Comparison of the Ionization Energy for Selected Elements. Figure 3 (Swithenbank, 1974) emphasizes the strong dependence of the electrical conductivity of the alkali-seeded plasma on the temperature. The data are for oxy-fired JP-4 combustion products under 1% potassium seed. JP-4 (Jet Propellant 4) is a military aviation fuel having 50-50 kerosene-gasoline composition by volume. For alkali-metal ionization in equilibrium plasma, Saha equation (Saha, 1920) predicts that the number density of released electrons (ne) is strongly affected by the temperature, obeying the following the proportionality expression: (1) is 3 en ∝ 5.1 T exp IE  Tk B where T is the temperature (in kelvins), IE is the ionization energy (in eV, electron volts), Boltzmann constant expressed in eV/K (having the constant value of 8.617×10–5). Marzouk (cid:247) (cid:247) ł (cid:246) (cid:231) (cid:231) Ł (cid:230) - Figure 3. Sensitivity of Electrical Conductivity to Temperature (Oxy-fired JP-4 seeded with 1% potassium). Simplified Electrical Analysis of MHD Channel A simple analysis is presented here to estimate the maximum electrical power output from an MHD channel. We use in Figure 4 the simplified 1-D channel of Figure 2, which greatly facilitates the discussion and permits analytical expressions. Figure 4. Electric Analysis of a Simple 1-D MHD Channel The current flux (current density vector) is denoted by J , which points to the negative y-axis. The external electrical field applied in the channel is V H = E 0 V L h where VL is the low voltage (at anode) and VH is the high voltage (at cathode), both are relative to the ground. The applied magnetic field B (in teslas) is in the z-axis, coming out of the page. There is an internal resistance formed by the plasma inside the channel and is represented by Ri, where = h Ri  A (2) (3) being the electrical conductivity of the plasma, and A is the projected channel area in the x-z with plane, which is also the area of either electrode. Marzouk 4 - (4) (5) (6) (7) The current I in the external circuit is related to the current density across the plasma by magnitude of current density is calculated as ( = 0EBu ) J I = AJ . The The maximum voltage across the electrodes is the open-circuit voltage VO.C., occurring when I = J = 0 (RL = ∞). In this case, manipulating (4) and (2) gives that this voltage is: V CO . . = hBu In the normal operating case, J ≠ 0 (RL is finite), and the current is hBu h +  A hBu + R R i . CO R = (cid:229) = R L = V . I L The power to the load is PL = I VL = I2 RL, where VL is the voltage across the load. In the case of an open circuit; RL = ∞, I = 0, and PL = 0. In the case of a shorted load; RL = 0, I = u B h/Ri = u B σ A, and again PL = 0. Introducing a load factor K = RL/ΣR = RL/(RL + Ri) = E0/(u B), and then expressing PL in terms of that factor, it can be shown that between these two extreme conditions PL has a maximum occurring at the condition of K = ½, or RL = Ri. This maximum value of the load power is where V is the channel volume, A h. PL , max 2 2 V = Bu 4 Considering some representative values for a proposed MHD channel, we have: V = 6 × 2 × 1 = 12 m3, = 15 S/m (siemens/m), u = 1200 m/s, B = 6 T (tesla). This predicts the maximum electric power to the load to be 2333 MWe = 2.3 GWe. This corresponds to a paramount power density (power per volume) of 194 MWe/m3. Although this estimate is rough because it is based on simplified analysis, it is quite conservative because the magnetic field can have larger values with the presence of today's super magnets. Such magnets are already used in the medical field as magnetic resonance imaging (MRI) equipment; and in scientific research as nuclear magnetic resonance (NMR) spectrometers, mass spectrometers, and particle accelerators. They are electromagnets but constructed using coils of a superconducting wire. Thus, they must be cooled to cryogenic temperatures to attain the superconducting phase, where the superconducting wire conducts much larger electric currents than an ordinary wire due to diminished resistance, thereby creating intense magnetic fields. Superconducting magnets can outperform fields all but the strongest ordinary electromagnets in terms of the produced magnetic fields and also can be cheaper to operate because no energy is dissipated as heat in the windings. For the plasma velocity, if the plasma gases have a gas constant of R = 235 J/kg-K (say 33% H2O - 67% CO2, by volume), and have a ratio of specific heats (g ) of 1.15, and an average temperature of T = 2850 K, then the speed of sound within the plasma is = 878 m/s. The above taken value thus corresponds to a modest Mach number of about 1.4. Even if the actual power is 50% of the above estimate, the power density remains attractive. a = TR Layout for Combined Oxy-fuel MHD Cycle Figure 5 shows a schematic layout for a power plant adopting the combined oxy-fuel MHD cycle, described here. The inverter is needed to convert the DC electric output to AC output, suitable for the power grid. The seed regenerator is responsible of recycling the seeded compound of alkali metal back to the combustor. Marzouk 5 - Figure 5. Layout of a Power Plant Utilizing the Combined MHD Oxy-fuel Cycle. Conclusions A combined oxy-fuel magnetohydrodynamic (MHD) power cycle was proposed and discussed as a means to counteract drawbacks of the separate concepts of oxy-fuel and MHD generator by capitalizing on the advantages of each. The cycle combines a topping MHD generator with a bottoming steam plant. Some design aspects were presented and a simplified 1-D model predicts that more than 2 GWe output can be produced from a compact space of only 12 m3. Acknowledgements The author has benefited greatly from different resources made available to him at the National Energy Technology Laboratory or the U.S. Department of Energy. The author is especially grateful to E. David Huckaby and Geo Richards. References Anheden, M., U. Burchhardt, H. Ecke, R. Faber, O. Jidinger, R. Giering, H. Kass, S. Lysk, E. Ramström, and J. Yan (2011), "Overview of Operational Experience and Results from Test Activities in Vattenfall's 30 MWth Oxyfuel Pilot Plant in Schwarze Pumpe," Energy Procedia, 4, 941–950. Kayukawa, N. (2004), "Open-cycle Magnetohydrodynamic Electrical Power Generation: a Review and Future Perspectives,"Progress in Energy and Combustion Science, 30(1), 33–60. Kramida, A., Y. Ralchenko, J. Reader, and NIST ASD Team (2014), NIST Atomic Spectra Database (ver. 5.2), [Online]. Available: http://physics.nist.gov/asd [2015, November 8]. National Institute of Standards and Technology, Gaithersburg, Maryland. Liu, Y., Z.U. Wang and H.-C. Zhou (2012), "Recent Advances in Carbon Dioxide Capture with Metal-Organic Frameworks," Greenhouse Gases: Science and Technology, 2(4), 239–259. Messerle, H.K. (1995), Magnetohydrodynamic Electrical Power Generation, Progress in Energy and Combustion Science, John Wiley & Sons, England. Saha, M.N. (1920), "Ionization in the Solar Chromosphere," Philosophical Magazine, 40, 472–488. Swithenbank, J. (1974), "Magnetohydrodynamics and Electrodynamics of Combustion Systems," in Combustion Technology: Some Modern Developments, Palmer, H.B. and J.M. Beer [eds], Academic Press. Wall, T., Y. Liu, C. Spero, L. Elliott, S. Khare, R. Rathnam, F. Zeenathal, B. Moghtaderi, B. Buhre, C. Sheng, R. Gupta, T. Yamada, K. Makino, and J. Yu (2009), "An Overview on Oxyfuel Coal Combustion: State of the Art Research and Technology Development," Chemical Engineering Research and Design, 87(8), 1003–1016. Woodside, C.R., K.H. Casleton, E.D. Huckaby, T. Ochs, D. Oryshchyn, G. Richards, P.A. Strakey, J. Pepper, I.B. Celik, J. Escobar-Vargas, D.C. Haworth, O.A. Marzouk, and X. Zhao, "Direct Power Extraction with Oxy-Combustion: An Overview of Magnetohydrodynamic Research Activities at the NETL-Regional University Alliance (RUA)," 29thAnnual International Pittsburgh Coal Conference, Pittsburgh, Pennsylvania, October 15–18, 2012. Zheng, L. and Y. Tan (2014), "Overview of Oxy-fuel Combustion Technology for CO2 Capture,"Corner Stone Magazine: the Official Journal of the World Coal Industry, January 7, 2014. Marzouk 6
1902.07341	1	1902	2019-02-19T23:16:33	Generalized Reciprocity Relations in Solar Cells with Voltage-Dependent Carrier Collection: Application to p-i-n Junction Devices	[ "physics.app-ph" ]	Two reciprocity theorems are important for fundamental understanding of the solar cell operation and applications to device evaluation: (1) the carrier-transport reciprocity connecting the dark-carrier injection with the short-circuit photocarrier collection and (2) the optoelectronic reciprocity connecting the electroluminescence with the photovoltaic quantum efficiency at short circuit. These theorems, however, fail in devices with thick depletion regions such as p-i-n junction solar cells. By properly linearizing the carrier-transport equation in such devices, we report that the dark-carrier injection is related to the photocarrier collection efficiency at the operating voltage, not at short circuit as suggested in the original theorem. This leads to the general form of the optoelectronic reciprocity relation connecting the electroluminescence with the voltage-dependent quantum efficiency, providing a correct interpretation of the optoelectronic properties of p-i-n junction devices. We also discuss the validity of the well-known relation between the open-circuit voltage and the external luminescence efficiency. The impact of illumination intensity and device parameters on the validity of the reciprocity theorems is quantitatively investigated.	physics.app-ph	physics	See the publisher version in the Physical Review Applied website (click here) Generalized Reciprocity Relations in Solar Cells with Voltage-Dependent Carrier Collection: Application to p-i-n Junction Devices Kasidit Toprasertpong1,, Amaury Delamarre1,2, Yoshiaki Nakano1,2, Jean-François Guillemoles2,3, and Masakazu Sugiyama1,2 1School of Engineering, the University of Tokyo, Bunkyo-ku, Tokyo 113-8656, Japan 2NextPV LIA CNRS-RCAST, the University of Tokyo, Meguro-ku, Tokyo 153-8904, Japan 3CNRS, UMR IPVF, Photovoltaic Institute of Ile-de-France, Palaiseau 91130, France Corresponding author, e-mail: [email protected] ABSTRACT Two reciprocity theorems are important for both fundamental understanding of the solar cell operation and applications to device evaluation: 1) the carrier-transport reciprocity connecting the dark-carrier injection with the short-circuit photocarrier collection and 2) the optoelectronic reciprocity connecting the electroluminescence with the photovoltaic quantum efficiency at short circuit. These theorems, however, fail in devices with thick depletion regions such as p-i-n junction solar cells. By properly linearizing the carrier transport equation in such devices, we report that the dark-carrier injection is related to the photocarrier collection efficiency at the operating voltage, not at short circuit as suggested in the original theorem. This leads to the general form of the optoelectronic reciprocity relation connecting the electroluminescence with the voltage-dependent quantum efficiency, providing correct interpretation of the optoelectronic properties of p-i-n junction devices. We also discuss the validity of the well-known relation between the open-circuit voltage and the external luminescence efficiency. The impact of illumination intensity and device parameters on the validity of the reciprocity theorems is quantitatively investigated. I. INTRODUCTION The optoelectronic reciprocity relation [1] has been proposed as a theorem which relates the electroluminescence (EL) and the photovoltaic external quantum efficiency (EQE) at short circuit in p-n junction diodes. With the use of the theorem, several new techniques for evaluating the electrical properties of solar cells based on the optical measurement become available, including the indirect measurement of the subcell voltage of multi-junction solar cells [2-4] and the spatial mapping of the local voltage and other electrical properties [5-10]. The derivation of this reciprocity between the EL (carrier injection followed by photon emission) and the EQE (photon absorption followed by carrier collection) is reciprocity of photon based on emission/absorption reciprocity of carrier injection/collection. The photonic reciprocity has been originally proved by the detailed balance principle and the ray optic approach [1], and has been recently discussed with more rigorous physics considering the coupling between the photonic and electronic states [11]. the the and two relations: On the other hand, the carrier-transport reciprocity, or sometimes called the Donolato theorem, has been first discussed in Ref. [12], describing the symmetry between the injection efficiency of minority carriers under applied voltage and the collection efficiency of photogenerated carriers under short-circuit condition. Despite many further attempts to generalize the carrier-transport reciprocity [13-17], all of them have focused on the carrier dynamics in the quasi- neutral region while neglecting the depletion region, where the carrier dynamics is comparatively complicated. This approach is acceptable in most p-n junction solar cells in which the depletion region is thin compared to other active layers. lengths, However, the exclusion of the depletion region is not an appropriate approach for describing p-i-n junction solar cells, in which the intrinsic region (i-region) is inserted in between the n- and p-regions to extend the depletion region. The p-i-n configuration is usually employed for materials with poor carrier diffusion such as amorphous and microcrystalline silicon [18,19], complicated alloys [20,21], and quantum structures [22]. The photocarriers are driven by the internal electric field inside the depletion region in addition to the diffusion process, enhancing the carrier collection efficiency and thus the output photocurrent. Since most photogeneration and recombination take place in the depletion region, the carrier-transport reciprocity becomes invalid in p-i-n junction solar cells, and consequently results in the failure of the optoelectronic reciprocity relation. The invalidity of the theorem in p-i-n junction solar cells has been addressed in the original paper [1], and subsequently confirmed by the numerical simulation [23,24]. When evaluating p-i-n junction solar cells using the optical measurement the optoelectronic reciprocity relation is needed to extract the correct information of the device electrical properties. [25], a careful interpretation of In this paper, we propose the generalized form of the reciprocity relations which is applicable in both p-n and p-i-n junction solar cells and discuss their validity in details. In Sec. II, the original reciprocity relations and the underlying assumptions are briefly discussed. In Sec. III, we show that, even in devices with a thick depletion region, it is possible to linearize the carrier-transport equation by considering well- suited regions, which provides a new aspect of dark-carrier injection and photocarrier collection. In Sec. IV, we derive the generalized form of the optoelectronic reciprocity relation connecting EL and EQE. Readers who are interested in the application on the EL analysis may focus on this section. The requirement of the illumination intensity and the device parameters to keep the reciprocity relations valid is discussed in Sec. V and our findings are summarized in Sec. VI. II. ORIGINAL RECIPROCITY RELATIONS In a p-n junction device, the EL and the photovoltaic EQE at normal emission/incidence, as shown in Figs. 1(a)-(b), are connected by the optoelectronic reciprocity relation [1], ( p n −Φ EL ) , ( ε V ) = (sc) EQE ( ) ε Φ ( ) ε bb ( e qV kT ) 1 , − (1) bb ( )εΦ where ε is the photon energy, is the blackbody radiation flux spectrum, V is the external applied voltage, q is the elementary charge, and kT is the thermal energy. The superscript (sc) is the notation for the short-circuit condition. The optoelectronic reciprocity above has been derived using the carrier-transport the collection efficiency of carriers photogenerated at a given point rg under short-circuit condition -- the situation with only optical perturbation -- with the injection efficiency fi of excess minority carriers δ nminor from the junction rj, i.e. the depletion edge, to the point rg under applied voltage V without illumination -- the situation with only electrical perturbation [12-17], reciprocity, which connects (sc) cf sc) f ( c ( r g ) = = n δ minor n δ mino r ( , r f V g i ( r g ( r ), j , , V n 0 V n 0 ) ) ( ( r g r j ) ) ( p n − ) (2) FIG. 1. Reciprocity relation connecting the electroluminescence (carrier injection/photon emission) and photovoltaic EQE (photon absorption/ photocarrier collection). The dynamics of electrons in the p-region is shown in the figure, whereas the hole dynamics in the n-region is in a similar manner. rg denotes the point of photogeneration, rj the depletion edge, and Sn=p the surface where n = p (see Fig. 2). (a)-(b) Original theorem for p-n junction solar cells considering EQE at short circuit. (c)-(d) Proposed generalized theorem considering EQE at operating voltage. ( gr ) 0 n FIG. 2. Device schematic and region notations of a solar cell with a thick depletion region. where is the minority carrier density at thermal equilibrium. Although it has been mentioned in some literatures [16,17] that the impact of the drift transport by the electric field is included in the above relation, it does not account for the electric field that depends on the applied voltage, i.e. the electric field in the depletion region, as will be discussed in this study. III. GENERALIZED TRANPORT RECIPROCITY The carrier-transport reciprocity in Eq. (2) has been derived by focusing on only the quasi-neutral region, where the density of one type of carriers, the majority carriers, is practically determined by the doping profile and there is only the dynamics of the minority carriers that has to be considered. On the other hand, an essential difference in the depletion region is that it is necessary to consider both the electron density n and the hole density p. Here, we investigate carrier transport dynamics in a device with a thick depletion region shown in Fig. 2, where the depletion region can be either an undoped i-region or a space-charge region of a depleted p- or n-region. To provide a clear picture and quantitative discussion on the theoretical analysis below, which is generalized to devices with arbitrary shapes, a simple p-i-n junction solar cell (Table I) was simulated using the PVcell 1D device simulator [26] as a demonstration (Figs. 3-8). TABLE I. Material parameters for numerical simulation of the p-i-n junction solar cell in Figs. 3-8 Parameters Symbols n- and p-region thickness i-region thickness Bandgap Carrier mobility SRH lifetime Radiative recombination coefficient Auger recombination coefficient Surface recombination velocities Intrinsic carrier density Temperature lp, ln li Eg µn, µp τn, τp B Cn, Cp vs,n, vs,p ni T Values 100 nm 500 nm 1.4 eV 0.1 cm2/Vs 100 ns 10-10 cm3s-1 10-30 cm6s-1 0 107 cm-3 300 K Three types of recombination R are considered: the Shockley-Read-Hall (SRH) recombination, the band-to-band radiative recombination, and the Auger recombination. The SRH recombination rate is given by [27] R SRH = τ p ( 2 np n − i ) τ + n + n n t , ( p + ) p t (3) where ni is the intrinsic carrier density, nt and pt are the carrier densities when the Fermi level aligns at the trap states, and τn and τp are the SRH lifetimes. The n and p subscripts correspond to the parameters for electrons and holes, respectively, and these notations will be used thereafter. In the depletion region, either doped or non-doped, charge carriers are built up in the steady state and can be divided into the region that n>p, i.e. the electron-rich region (Vn>p in Fig. 2), and the region that p>n, i.e. the hole-rich region (Vp>n) [28]. (Therefore, strictly speaking, the word "depletion region" is not the correct word use. We keep using this word as it is the standard word for indicating the region between the two quasi-neutral regions.) As can be seen in Fig. 3(c) showing the carrier distribution of a p-i-n junction solar cell described in Fig. 3(a) and Table I, the both carrier densities vary exponentially in the depth direction of the depletion region. The region where the values of p and n are in the same order of magnitude is very thin, allowing the approximation n >> p and p >> n in the regions Vn>p and Vp>n, respectively. Furthermore, for deep-level defects, nt and pt are small and can be neglected. In this way, RSRH at the position r can be approximated by R SRH     ≈        2 np i − n 2 n i p − n         τ p ; r ∈ V n > p τ n ; r ∈ V p > n , (4) which is confirmed in Fig. 3(d). The radiative recombination is given by FIG. 3. Simulation of (a) band structure, (b) quasi-Fermi levels, (c) carrier distribution, and (d) SRH recombination rate with the open symbols showing the approximation with Eq. (4). A p-i-n junction solar cell with the parameters summarized in Table I at 0.5 V is simulated as a demonstration. R rad = ( B np n − i )2 , (5) where B is the radiative recombination coefficient. Even though the Auger recombination can be large in the quasi- neutral region of some materials, it is generally small and can be omitted in the depletion region where the carrier density is comparatively low. Nevertheless, to generalize the formula which holds in both the quasi-neutral and the depletion regions, we incorporate the Auger recombination process, which can be expressed using the coefficient Cn and Cp [29] R Aug = 2 C n np n i − n ( ) 2 C p np n i − n −  2 C n np n ≈  i 2 C p np n  i − p ( ( ( ∈ p + ) ; r ) ; r V n p > V p n > . ∈ ) (6) The approximation in the last line uses that fact that the carrier distribution varies abruptly in the depletion region, similarly to the approximation in Eq. (4). Electron current density Jn flowing in the semiconductor can be expressed in terms of drift and diffusion currents, nJ µ= n q n E + , qD n ∇ n (7) n nD kT qµ= where µn is the electron mobility, is the electron diffusion coefficient, and E is the electric field. For materials with non-uniform electron affinity, bandgap, or density of states, the effect of the non-uniformity can be included in E as the effective field [30]. In the absence of illumination, we obtain the carrier transport equation from the current continuity and the recombination in Eqs. (4)-(6): qR 0 ( q = −∇ ⋅ µ n ( q = −∇ ⋅ µ n   q p      +    q n      − E n + E n + 2 n i n 2 n i p − n ) ) qD n ∇ + qD n ∇           n 1 τ p 1 τ n + + Bn C n n + Bp C p + p . (8) 2 2        ; r ∈ V n p > ; r ∈ V p n > Here, µ, τ, B, C, and ni can be non-uniform as a function of position r. It is practical to assume that the carrier recombination R has a negligibly small impact on the distribution of the majority carriers, whose density is comparatively large, whereas its impact on the minority-carrier distribution is still important. The validity of this assumption will be discussed in Sec. V B. In this way, as electrons are the majority carriers in the region Vn>p, Eq. (8) can be rewritten by 0 n ) + E n qD n ∇ ( q = −∇ ⋅ µ n 0 ;     q n      Bp C p 1 τ n     2 n i p    + + + − 2 p r ∈ V n > p (9) ; r ∈ Vp > n or 0 n ) n ∇ ( µ ε = −∇ ⋅ Fn 0 ;     q n      Bp C p 1 τ n     2 n i p    + + + − 2 p r ∈ V n p > (10) ; r ∈ V p > n by considering that the current can be expressed by the through gradient nJ the [30]. Similarly, we can write of nµ ε ∇ Fn quasi-Fermi levels εFn = n ε Fn ε Fp    ( ) r ( ) r = = constant for r constant for r ∈ ∈ V n p > V p n > . (12) This implies that the majority-carrier quasi-Fermi levels are flat not only in the quasi-neutral region, but also in the depletion region. The constancy of the quasi-Fermi levels of the majority carriers in the depletion region is confirmed by the simulation in Fig. 3(b). − = Equation (12) tells us that the quasi-Fermi level splitting ∆ ε ε ε at the position rn=p in the interface area F Sn=p (see Fig. 2) between the regions Vn>p and Vp>n, where n , is given by the applied voltage V: p= Fp Fn ( ) r ε ∆ n= p F ≡ = ( ) r − ε ε n= p Fn Fp for r qV n= p ) ( r n= p S ∈ n p = (13) [See Fig. 3(b)]. By reminding the relation between the carrier F kT densities and the quasi-Fermi level splitting , the electron and hole densities at this position are given by 2 i e n np ε∆ = n = ≡ n p n ( r n= p ) = p ( r n= p ) = n i e q V 2 kT . (14) By introducing the variables u ≡ 2 1 np n − i = F k T e ε∆ − 1, (15) we can reexpress the electron photocurrent density at Vp n>∈r as nJ = = = n n − n ∇ n ∇ µ ε Fn ( µ ε ε Fp Fn µ kT u ∇ n 1 u + µ n = n ) 2 kTn i p ∇ u , (16) {Note that Eq. (10) for 0 Fpε∇ Vp n>∈r = at as Vp n>∈r [Eq. (12)]} and rewrite 0 = −∇ ⋅ ( µ n 2 kTn i p ∇ u ) + 2 uqn i p    1 τ n + Bp C p + p 2 .    (17) Under applied voltage V, we define the operator 0 p ∇ ) ( µ ε = ∇ ⋅ Fp p   2   n i q p        n +      − 1 τ p + Bn C n n + 2     ; V ∈ r n p > 0 ; V ∈ r p n > = and = are conditions that make the upper line of Eq. (10) for the hole current continuity. The relations Fpε∇ and the lower line of Eq. (11) satisfied. That is, Fnε∇ 0 0 n,VL ≡ −∇ ⋅ ( µ n 2 kTn i p ∇ + ) 2 qn i p    1 τ n + (11) Bp C p + p 2 .    (18) Using this operator, we can say that in the device at voltage V under dark condition [see Fig. 1(c)], the spatial profile of the parameter r , has to satisfy ( )u r , denoted by D Vu , ( ) n,VL   Vu D , ( ) r  =  0 for r ∈ V . p n > (19) G ( ) r In the next step, we consider the situation where carriers are additionally photogenerated at point rg inside Vp>n with the delta-function-like generation rate of , while the applied voltage V is kept the same [see Fig. 1(d)]. Provided that the carrier generation is low enough that the majority-carrier density (p in Vp>n and n in Vn>p) is unaffected (the validity is discussed in Sec. V A), with the photogeneration at the applied voltage V, denoted by L Vu , has to satisfy , ( , r r g ( gδ= ( )u r r − g r ) ) L n,V u   L , V ( , r r g )  =  ( g q δ r − r g ) for , r r g ∈ V p n > . (20) As suggested in Ref. [17], the Green's identity is useful to analyze the boundary of the carrier transport equation. To investigate the Green's identity for the operator Ln,V, let's consider the integral of ] over the region Vp>n (see the Supplemental Material [31] for more details of the derivation): L n,V L n,V [ u [ u , D V , D V , L V , L V − u u ] L n,V u   , L V  −  u , L V L n,V u   , D V , D V )  d V ( u ∫ V p n > = − ∇ ⋅ ( u , D V µ n 2 kTn i p ∇ u , L V − u , L V µ n 2 kTn i p ∇ u , D V ) V d = − ( u J , D V n,L,V − u , L V J n,D,V ) ⋅ d S ∫ V ∫ p n > p n = S u = − , D V ( r p=n ( J ) ∫ p n = S n,L,V − J n,D,V ) ⋅ S . (21) d Here, Jn,L,V and Jn,D,V are the electron current densities at voltage V with and without illumination, respectively. The last line of Eq. (21) uses the result from Eq. (13) that , Fε∆ thus u, is constant in the surface Sn=p for the given external voltage V, provided that both carrier densities n and p at rn=p are not affected by the illumination (see Sec. V A). The Jn,L,V − Jn,D,V term corresponds to the increment of current density induced by the illumination at the fixed voltage V, or in other words, the photocurrent density. By noting that the electron photocurrent density vector Jn,L,V − Jn,D,V has the opposite direction of the normal vector dS from Vp>n, thus ( , we can say that the term <S d n,D,V n,L,V 0 − J J ) ⋅ − ∫ S p n = ( J n,L,V − J n,D,V ) ⋅ S d = I p ( h r g , V ) (22) the relation ε∆ F D , ( qg e ( gr , V kT ) 1) − = I ( gr , )( V e p h qV kT − 1), (23) where ∆εF,D(rg, V) is the quasi-fermi level splitting at the position rg under dark condition for the applied voltage V. By following the similar derivation in Eqs. (15)-(23) for Vn p>∈gr , it is straightforward to show that the same relation holds as well if carriers are photogenerated in Vn>p and thus the validity of Eq. the photogeneration at any positions r inside the device, including the depletion region. By rearranging Eq. (23) to (23) can be extended to , V ) e = I ph ( r g qg ε∆ , ( F D r g , V kT ) qV kT e − 1 − 1, (24) we obtain an interesting relation between the photocurrent of a device with illumination [Fig. 1(c)] on the left side of the equation and the quasi-Fermi level splitting of the same device without illumination [Fig. 1(d)] on the right side. The left side of Eq. (24) is the ratio of collected photocurrent at voltage V to the photogeneration rate at point rg, which can be defined as the local carrier collection efficiency cf ( r g , V ) ≡ I ( r g ph , V qg ) (25) On the other hand, the right side of Eq. (24) describes how well the quasi-Fermi level splitting ∆εF under carrier injection penetrates from Sn=p, where ∆εF = qV, into the region Vp>n or Vn>p [see Fig. 3(b)]. Since the right side of Eq. (24) is the generalized form of the original injection efficiency in Eq. (2), which has been defined as the penetration of the excess minority carrier from the depletion region, it can be considered as injection efficiency the generalized local carrier e f i ( , r V ) ≡ ε∆ , ( , r F D V kT ) qV T k e − 1 − 1 . (26) constant, considered In the quasi-neutral region, where the majority-carrier density is − 1 holds and fi in Eq. (26) gives fi defined by Eq. n δ= minor (2). In other words, fi given by Eq. (26) is a more general form which can describe the behavior in all regions. n minor 1 − = F D kT , n 0 n 0 ε∆ e From Eqs. (24)-(26), we obtain the generalized carrier- transport reciprocity is the electron photocurrent that is photogenerated at the point Vp n>∈gr and can be subsequently extracted out of the hole- rich region Vp>n. By substituting Eqs. (13), (15), (19), (20), and (22) in the first and the last lines of Eq. (21), we obtain f c ( r g , V ) = f i ( r g , V ) (27) 2 kTn µ n i (sc) p    1 τ n 2 qn i ( sc) p + ≠ L n,V (28) (sc) nL = −∇ ⋅ ( ∇ ) + (sc) Bp + ( C p p (sc) 2 )    for V ≠ 0, which violates the main assumption L n,V in the derivation of the original theorem [12-17]. This inequality limits the use of the original theorem [Eq. (2)] on solar cells with thick depletion regions to only small voltage near zero. (sc) nL = FIG. 4. (a) Band structures and (quasi-)Fermi levels at 0 and 0.5 V. The electron quasi-Fermi level εFn at 0.5 V shown in this scale almost overlaps with the Fermi levels εF at 0 V and is not shown here. (b) Comparison of carrier distribution at 0 V and 0.5 V. connecting the local dark-carrier injection efficiency with the local photocarrier collection efficiency. Although it has been addressed [1,23,24,32] that, in solar cells with thick depletion regions, the majority/minority-carrier concept in the depletion region may cause the non-linearity in the transport equation and violate the transport reciprocity relation, we have shown that it is possible to divide the depletion region into the electron-rich and hole-rich regions and linearize the carrier transport equation as a linear differential equation of u, resulting in the validity of the reciprocity relation. invalidity of the ) ( , Vgr (sc)( cf The significant difference from the original carrier- transport reciprocity is that the photocarrier collection efficiency fc should not be considered at the short-circuit condition [ in Eq. (2)], but should be defined at the ) gr cf operation voltage V [ in Eq. (25)]. This is due to the fact that the governing transport equation in Eqs. (10)-(11) at the applied voltage V ≠ 0 is not the same as that at V = 0. Even though Eq. (12) suggests that the quasi-Fermi levels are almost flat in the corresponding majority-carrier regions regardless of the applied voltage [see Fig. 3(b) and Fig. 4(a)], it is shown in Fig. 4(b) that the electron density n in Vn>p and the hole density p in Vp>n have voltage dependence. Changing the applied voltage V alters the electrostatic potential profile as well as the distance of the bandedges from the quasi-Fermi levels, i.e. for the conduction band EC and E ε− for the valence band EV, which in turn alters the V F p majority-carrier density. This results in the inequality E ε− C Fn The invalidity of the use of the short-circuit carrier collection efficiency can be alternatively explained in terms of the voltage-dependent electric field. At short-circuit condition, the drift-diffusion-based transport equation in Eq. (9) is expressed by 0 = −∇ ⋅ ( q µ n (sc)E n + qD n ∇ + ) n qR , (29) ) gr , n 0( n V = where E(sc) is the electric field profile at V = 0. As the electric field in the depletion region decreases with increasing forward bias voltage V, carriers follow different carrier of the simple linear relation between the dynamics of photogenerated carriers at V = 0 and electrically injected transport equations for V = 0 and V ≠ 0, leading to the failure carriers at V > 0. In addition, the expression for the local dark-carrier injection in the original transport reciprocity relation, the right-hand side of Eq. (2), is not applicable to solar cells containing thick depletion regions since it has been derived based on the assumption minor ( . This assumption is applicable only if the majority-carrier density remains unchanged from the thermal equilibrium, which holds in the highly-doped quasi-neutral regions but not in the depletion region [Fig. 4(b)]. Instead, the local carrier injection efficiency expressed by the penetration of the quasi-Fermi level splitting, Eq. (26), is the general form which can be used in both the quasi-neutral and the depletion regions. Hence, it is more straightforward to interpret the carrier-transport reciprocity as the relation between the photocarrier collection efficiency fc and the uniformity of the dark quasi-Fermi level splitting : solar cells with efficient collection of photocarriers under illumination will have the spatially uniform under the carrier injection mode, and in the opposite way, the inefficient photocarrier collection directly corresponds to the low under carrier injection. ,F Dε∆ ,F Dε∆ e ε∆ ,F Dε∆ F kT gr ) The carrier-transport reciprocity is confirmed by the 1D simulation of the p-i-n device described in Table I. Fig. 5(a) shows the injection efficiency fi to the n-region at the depth zg = 30 nm and the collection efficiency fc of carriers photogenerated in a thin layer at the same depth, simulated using Eq. (25) and Eq. (26), respectively. Fig. 5(b) is similar but is simulated at the depth zg = 500 nm, which is inside the i-region. It can be seen that the voltage-dependent collection efficiency fc, not the collection efficiency at short circuit, agrees well with the value of dark injection, confirming the proposed transport reciprocity relation in Eq. (27). (sc) cf IV. GENERALIZED OPTOELECTRONIC RECIPROCITY When the device at applied voltage V is illuminated with the monochromatic light with photon energy ε, by using the probability is absorbed in the volume element dr at r, EQE can be obtained by integrating the photocurrent induced from each point over the entire volume Vdevice, incident photon ) , dVεr r that the a ( , QE( , E ε V ) = ∫ deviceV a ( r , , ε ) V f ( r , ) V d . r c (30) ( , ) dV , φ εr z On the other hand, in the absence of illumination, the net photon flux r of photon energy ε that is emitted from the volume element dr of the absorber and is able to escape from the device surface without being reabsorbed can be expressed using the generalized Kirchhoff's law [33], which has been further generalized for materials with the field-dependent absorption [11], , φ ε z r ( , V ) r d = a ( r , , ε V ) Φ bb ( , ( , r ε ∆ ε ( ) e F D V kT ) ) 1 − r d . ( 31 ) The total electroluminescence ΦEL from the absorber surfacecan be obtained by integrating the photon flux , φ εr z r contributed from each volume element, ) 1 Φ L( , ε E V ) ) dV ( ) ε ∆ ε F D ( , , ε ( , r V Φ V ( = − a bb kT e ) ( r , ) , ∫ V device ∫ V device EQE( , ε = = a ( r , , ε V ) f c V ) Φ b b V ) Φ bb ( ) ε e q kTV r d ) 1 r ( kT qV e d − ( ) ε ) 1 , (32) − ( , r ( where the generalized transport reciprocity in Eq. (27) is used to obtain the second line and EQE in Eq. (30) to obtain the last line. Equation (32) represents the general form of the optoelectronic reciprocity which links the EL with the photovoltaic EQE [see Figs. 1(c)-(d)]. By following the similar derivation in Ref. [1], the optoelectronic reciprocity relation in Eq. the electroluminescence emitted at the tilt angle θ, with EQE( , )Vε substituted with can be ( , ,Vε θ ) (32) ) (tilt Φ EL extended EQE to (tilt) ( , )Vε θ , FIG. 5. Simulated local dark-carrier injection and photocarrier collection efficiencies of the p-i-n junction solar cell in (a) the n-region at the depth zg = 30 nm and (b) the i-region at zg = 500 nm. Circle symbols show the injection efficiency obtained from the simulated quasi-Fermi level of the device without illumination and Eq. (26). Lines represent the voltage- dependent collection efficiency fc of carriers photogenerated in a thin layer at the corresponding depth zg with the photogeneration rate of 1010 cm-2s-1, corresponds obtained from the simulated photocurrent and Eq. (25). to fc at short circuit. Simplified device schematics indicate the positions of photogeneration. measured at the light incident angle θ. We would like to emphasize, as it might be easily misinterpreted due to the 1) term ( − , that there is no assumption for the formula of EL that the quasi-Fermi level splitting must equal to qV in the entire device. The non-uniformity of the quasi- Fermi level splitting has been included in the EQE( , )Vε term. ,F Dε∆ ,F Dε∆ (sc) cf kV T qe in Equation (32) derived above has the similar expression to the original optoelectronic reciprocity relation in Eq. (1), but the voltage-dependent EQE should be used instead of the short-circuit EQE. This is a consequence of the voltage- dependent carrier collection the carrier-transport reciprocity [Eq. (27)] and the voltage-dependent light absorptance [11]. Previously, it has been addressed in Ref. [34] that in non-linear solar cells, the differential form of the electroluminescence relation ) EQE d = with EQE represented the voltage-dependent EQE. We have shown in this paper that the carrier transport in the depletion region is linear at fixed voltage and the absolute electroluminescence ELΦ itself also follows the optoelectronic reciprocity relation with the voltage-dependent EQE. follows e bb reciprocity ( d qV kT the qV k T Φ Φ EL This finding suggests that when applying the reciprocity relation to analyze the electrical properties of solar cells from optical measurement, for instance extracting the subcell voltage of multi-junction solar cells and mapping the voltage distribution from emitted EL [2-10], the voltage-dependent EQE data are required for correct data analysis. The original reciprocity relation expressed by Eq. (1) may be used only in solar cells with EQE independent of voltage such as p-n junction sufficiently high doping concentration. solar cells with V. VALIDITY DISCUSSION A. Limitation of illumination intensity One main assumption in the derivation above is the low illumination intensity such that the majority-carrier density remains unchanged. It is important to quantitatively estimate such limit as it is directly related to the limitation of the illumination intensity in the EQE measurement. If the electric field in the depletion region is sufficiently high, the transport of photogenerated carriers is dominated by the drift transport. For the photogeneration in Vp>n, the electron photocurrent density Jph,n and the electron density increment ∆n due to illumination are related by [28] J ph,n E nµ= ∆ nq . (33) The similar relation of the hole photocurrent density Jph,p and the hole density increment ∆p for the photogeneration in Vn>p can also be obtained. 2 T qV k n = = n p that given by Remind the electron and hole densities at Sn p=∈n= p r [Eq. (14)] are the minimum values of the majority-carrier density in both the regions Vp>n and Vn>p. The condition for keeping the majority-carrier distribution unaffected by the induced photocarriers ∆n is estimated by n e i J ⊥ ph, n or p ( r n= p ) < q µ n or p n e i qV kT 2 ⊥ E n p = , (34) Φ illu,EQE ( ) ε < n i E n = p e J ⊥ ph, n ) r n= p or ( p n pE⊥ is the electron or hole photocurrent where density normal to the surface Sn=p and = is the electric field normal to Sn=p. It can be interpreted from Eq. (34) that in materials where µn > µp, the theorem validity is more tolerant to photogeneration in Vp<n than photogeneration in J ⊥ induces smaller Vn>p as electron photocurrent density ph,n steady-state charge density ∆n. For planar devices, the output photocurrent density Jph is and the electric ⊥+ given by J ph, field En=p is normal to Sn=p. Then, r n= p r n= p ⊥ ph, ph = J J ) ( ) ( p n J ph < nq E i n p = qV 2 kT e min( µ µ p n , ) (35) n pE = = 22 kV/cm 12 kV/cm respectively. These values well agree with Fig. 6(a) showing For the p-i-n junction solar cell in Table I, Eq. (35) gives the photocurrent limit at V = 0 V [ as shown in Fig. 6(c)] to be 3.5 nA/cm2 and at V = 0.5 V ( ) to be 30 µA/cm2, which are equivalent n pE = = to the carrier fluxes of 2×1010 cm-2s-1 and 2×1014 cm-2s-1, that the photogeneration rate of 1×1014 cm-2s-1 is too high for the validity of the reciprocity relation at 0 V, and 1×1018 cm- 2s-1 is too high for both 0 V and 0.5 V. The carrier distribution under illumination at 0.5 V shown in Fig. 6(b) illustrates that this is due to the remarkable impact of high photogeneration rate on the majority-carrier distribution: the generation rate of 1018 cm-2s-1, but not 1014 cm-2s-1, affects the distribution of the FIG. 6. (a) Failure of the carrier-transport reciprocity at high illumination intensity and high applied voltage for the photogeneration at zg = 500 nm. (b) Electron density n and hole density p at 0.5 V under different photogeneration rates. The distributions of n at the photogeneration rate of 1010 cm-2s-1, and p at 1010 and 1014 cm-2s-1 mostly overlap with the distributions under dark condition. The photogeneration rate as high as 1018 cm-2s-1, roughly equivalent to 5-sun intensity, affects n in the electron-rich region and p in the hole-rich region. (c) Electric field En=p at zn=p. is a sufficient condition to make Eq. (34) simultaneously holds for both the electron and hole photocurrents, ensuring the validity of the transport reciprocity. In the EQE measurement, this is equivalent to the condition for the illuminated photon flux 2 qV k T EQE( m n( i ) , V ε µ µ p n , ) . (36) majority-carrier density at 0.5 V. See also the Supplemental approach cannot give the exact solution of εFn(z), it provides a rough estimation of how the device parameters, such as mobilities, lifetimes, and electric field, affect the uniformity of the majority-carrier quasi-Fermi levels. (0) For non-degenerate carriers, the spatial distribution of n(0) flat n pE = locally follows E kT C quasi-Fermi level. For the electric field uniform near zn=p, it is a fair estimation to say (0) E ε F n C for the ∝ − ∝ − kT l n − n ) ( n (0)( ) z ≈ n n = p e − qE n p = ( z − z n p = ) kT (38) (0) the vicinity of zn=p. By using Eq. p in n can integrate Eq. (37) from zn to z near zn=p and obtain (14) and for large V, we 2( n e i 1) − ≈ 2 n e i qV kT qV kT 2 n i 0) − = (  µ  n  dn dz ′ ε Fn ( z ′ )   z z n = z ∫ q R z dz z ′ ) ( n ′ FIG. 7. (a) Position and width notations for the one-dimensional device discussed in Sec. V B. (b) Quasi-Fermi levels and (c) carrier distribution in the p-i-n device described by Table I at applied voltage of 0.9 V. Material [31] for the behavior of the carrier distributions as well as the quasi-Fermi levels under other device conditions. (0) µ n n d dz (1) ε Fn ( ) z − J n 0 B. Requirement of electric field and mobility Another assumption for the validity of the reciprocity relations is that the quasi-Fermi level of the majority carriers should be flat under carrier injection [Eq. (12)]. Here, we investigate the spatial variation of the electron quasi-Fermi level εFn in the electron-rich region by including the recombination term in the upper line of Eq. (10). To investigate such impact in a quantitative manner, we simplify the discussion to a one-dimensional device with uniform material properties of µn, τn, τp , B, Cn, Cp, and ni, and no surface recombination. The 1D transport equation for electrons in the electron-rich region ( n in Fig. 7, z where zn is the position of the electron contact and zn=p is the position where n = p) gives z = n p ≤ < z ( (0) n (0) p − 2 n i ) dz ′ − 2 n i ) dz ′ = z ∫ z q n (0) (0) n p + + ( (0) (0) (0) qC n n ( qB n τ p z ∫ z n z ∫ z n kT i E τ n p p = ) 2 qBn + i n qC n N L e + n qV kT ( 2 i − n e z z D 2 n n e ≈ (0) p − 2 n i ) dz ′ q E n p = ( z − z n p = ) kT T k q V e q V kT , (39) where Jn0 is the electron current density at the electron contact zn and Ln is the width of the quasi-neutral n-region [see Fig. 7(a)]. The expression for n(0) is approximated by Eq. (38) for the first integral (SRH recombination) as the integrand is large only near zn=p, and approximated by the electron density ND at the undepleted n-region for the third integral (Auger recombination) due to the considerably large n in the quasi- neutral region. In a device without surface recombination, the electron current must vanish at the hole contact zp and Jn0 is given by to the total recombination throughout the device: d dz    µ ε Fn n dn dz  =   ( ) qR z =  q p    − 2 n i n       1 τ p + Bn C n n + 2 . (37)     J n 0 = − = − p ( ) z ∫ q R z dz z z n n p = As Eq. (37) cannot be easily solved analytically, we employ some approximations to estimate a rough tendency of εFn. For the first-order approximation, we first express n and p by n(0) (0) Fnε , and then and p(0) assuming the flat quasi-Fermi levels (1) Fnε . Even though this solve for the first-order solution ≈ − − p z ∫ q n n p = − ( ) R z dz ( ) R z dz q ∫ z n kT i qBn Le E n ) qn C N L C N L e 1 1 τ τ p n z     + qV kT n D = 2 i     qV 2 i + − ( kT e A , 2 p n p p (40) qV kT where NA and Lp are the hole density and the width of the δεFn and δεFp should be sufficiently small compared to kT to keep the relations valid, that is, 1 E n p = µ n     + Bn e i kT E q n p = qV 2 kT ( L 1 E n p = µ p     + Bn e i kT E q n p = qV 2 kT ( L n                     1 1 2 τ τ p n +     p     . (42) + L , i p ) + n C N L e i A p p qV 2 kT     < 1 1 1 2 τ τ p n +     + L , i n ) + n i C N n D L n e qV 2 kT     < 1 Particularly in a symmetric device (all parameters have the same values for the subscripts n and p), the estimated requirement for the electric field when the SRH process is dominant is n pE = > 3 kT qµτ 2 (43) and when the radiative or Auger process is dominant is E = > n p ( 1 2 BL CN L n e i + D n ) qV kT 2 . µ (44) The device described by Table I is dominated by the SRH process and Eq. (43) gives the required electric field of 2 kV/cm, which is consistent with the large deviation of the injection efficiency fi from the collection efficiency fc in Fig. 6(a) when the electric field n pE = [Fig. 6(c)] approaches 2 kV/cm at 0.9 V. This is confirmed by a significant drop of the quasi-Fermi levels of majority carriers at zn=p for V = 0.9 V shown in Fig. 7(b) and Fig. 8. Equations (42)-(44) can also be interpreted as the lower limit of the carrier mobility µ in order to keep the reciprocity relations valid for given device structure and operating voltage. For instance, the same device requires a mobility at least = 160 cm2/Vs to n pE ⊥ validate the relations at 1 V, at which the electric field = is as low as 51 V/cm. This finding is in a good agreement with the simulation results in Ref. [32] suggesting that the reciprocity relations become invalid at high applied voltage, where the electric field becomes weak, and become invalid in the entire voltage range in devices with extremely low carrier mobility. q Eτ ⊥ n p = 3 kT 2 ( ) 2 C. SRH via shallow defects The SRH recombination rate in Eq. (4) assumes the deep- level defects. For shallow-level defects, for instance near the conduction band, nt becomes large and the recombination rate in Eq. (3) becomes Fnδε at zn=p under carrier Fnδε numerically FIG. 8. Electron quasi-Fermi level drop injection mode estimated from Eq. (41) as compared to calculated from the device simulation. quasi-neutral p-region, respectively, and L is the device total thickness. Note that Jn0 is not a function of position z. Then, integrating the deviation of using Eqs. (39)-(40) gives (1) dε Fn dz (1) Fnε ≡ (1) Fn (1) δε ε Fn z n p = ( z n p = ) − ε Fn ( z n ) d dz (1) ε Fn dz ∫ z z n = = n p = ∫ z n 1 n (0) µ n ≈ − kT E n p = µ n qV 2 kT + Bn e i        ( ( 0) µ n n    d dz (1) ε Fn − J n 0    + J n 0 dz    kT E q n p =     1 1 2 τ τ p n +     L + L ′ p p ) + n i C N p L p e A qV 2 kT ,    (41) pL′ is width of the depletion region that has p > n. Due where to the 1 n term, the integrant is particularly large for z near zn=p and the expressions in Eqs. (38) and (39) are used. The deviation Fpδε for the hole quasi-Fermi level in the hole- rich region can be expressed the same way by in interchanging the subscripts n and p. (1) Figure 8 demonstrates Fnδε at zn=p for the device parameters in Table I under carrier injection mode, for that approximated from Eq. (41) and the exact solution solved by the device simulator employed so far. Even though Eq. (41) results in finite approximation error, which is due to the assumptions made during the derivation, the estimation of Fnδε by Eq. (41) provided a good approximation within a factor of 2 and can be used as a convenient tool for further understanding of the behavior of Fnδε . Since the constant quasi-Fermi levels of the majority carriers [Eq. (12)] is the requisite condition for the carrier- transport and the optoelectronic reciprocity relations, both R SRH = ( np n − i )2 nτ p t . (45) V oc = (rad) V oc + kT q ln η ext + kT q ln ( F V oc c ( F V oc i ) ) , (47) This has the same expression as the radiative recombination in Eq. (5) and can similarly be included in the above derivation, provided that the illumination is weak enough that shallow-level defects are not changed to the deep-level type upon illumination. D. Application to various types of solar cells Since the derivation above includes the field-dependent carrier transport in the depletion region, the proposed reciprocity relations are applicable to p-i-n junction solar cells with the i-region having field either uniform or non- uniform, or having field-dependent material parameters such as mobility, lifetime, and density of states. They can also be used to describe the behavior of p-i-n quantum structure solar cells, provided that the carrier mobility in such structures can be defined [35]. Counting for the voltage dependency of EQE also extends the validity of the optoelectronic reciprocity to absorbers with voltage-dependent absorptance [36,37], which is not valid in the original theorem. Moreover in p-n junction solar cells, particularly in low-doped devices, it is well recognized that applying forward bias decreases the depletion width and thus widens the quasi-neutral region. The variation of the depletion width, which has not been taken into account so far, appears as the voltage dependence of carrier collection as well as EQE and can be treated by our modified formulae. E. Open-circuit voltage It is worth noting about the relation of the open-circuit voltage Voc with the external luminescence efficiency ηext, which has been discussed in p-n junction solar cells by assuming the superposition of the short-circuit current and the dark diode current and is expressed by [1,23,38] ( V oc ) p n − = (rad) V oc + kT q ln η ext , (46) (rad) ocV is the ideal open-circuit voltage in the where radiative limit. For devices with voltage-dependent carrier collection such as p-i-n junction solar cells, the non-constant photocurrent at forward bias results in the failure of the simple superposition between the short-circuit and dark currents. Instead, Voc of such devices should be expressed by [32] ( ) ) ( ) cf ( , Vr ( , Vr cF V and iF V are the spatial averages of the where ) local collection efficiency and injection efficiency , respectively. For the illumination intensity and if device parameters that satisfy Eqs. (35) and (42), the carrier- transport thus ( F V = c f , holds. Then we obtain reciprocity, ( F V i and ( , r ( , r V ) V ) f c = i ) ) V oc = (rad) V oc + kT q ln η ext , (48) which is the same expression for p-n junction solar cells [Eq. (46)]. The generalized carrier-transport reciprocity results in the open-circuit voltage Voc unaffected by the characteristics of voltage-dependent carrier collection, which is in a good agreement with the simulation result in Ref. [32] reporting the validity of Eq. (48) in depleted organic solar cells having voltage-dependent photocurrent, and with the experimental result in Ref. [39] showing the satisfactorily high Voc in p-i-n quantum well solar cells in spite of their poor photocarrier collection under forward bias. VI. CONCLUSION the theorems, two reciprocity In this study, we have investigated the validity of the reciprocity relations in solar cells with thick depletion regions. The voltage dependence of the photocarrier collection efficiency and the EQE in such devices results in the failure of the carrier-transport reciprocity relating the dark-carrier injection and short-circuit photocarrier collection, and the optoelectronic reciprocity, which is the consequence of the previous relation, relating the EL with the short-circuit EQE. The analysis of carrier dynamics in the depletion region suggests that the original transport reciprocity theorem should be modified: the photocarrier collection efficiency should be defined at the operating voltage, not at short circuit. As a consequence, we obtain the general form of the optoelectronic reciprocity which connects the EL with the EQE at the operating voltage, extending its application, e.g. the EL diagnosis in solar cells, to solar cells with voltage-dependent EQE. This finding extends the validity of another well-known relation between the open-circuit voltage and the external luminescence efficiency to devices with voltage-dependent photocurrent. The acceptable ranges of the illumination intensity, the depletion-region field, and the carrier mobility to assure the validity of the theorems have been discussed and are analytically expressed in terms of the material parameters. ACKNOWLEDGMENTS A part of this study was supported by the Research and Development of Ultra-high Efficiency and Low-cost III-V Compound Semiconductor Solar Cell Modules project under the New Energy and Industrial Technology Development Organization (NEDO). The authors would like to thank Daniel Suchet for the helpful discussions. REFERENCES [1] U. Rau, Reciprocity relation between photovoltaic quantum efficiency and electroluminescent emission of solar cells, Phys. Rev. B 76, 085303 (2007). [3] [2] T. Kirchartz, U. Rau, M. Hermle, A. W. Bett, A. Helbig, and J. H. Werner, Internal voltages in GaInP/GaInAs/Ge multijunction solar cells determined by electroluminescence measurements, Appl. Phys. Lett. 92, 123502 (2008). S. Roensch, R. Hoheisel, F. Dimroth, and A. W. Bett, Subcell I-V characteristic analysis of GaInP/GaInAs/Ge solar cells using electroluminescence measurements, Appl. Phys. Lett. 98, 251113 (2011). S. Chen, L. Zhu, M. Yoshita, T. Mochizuki, C. Kim, H. Akiyama, M. Imaizumi, and Y. Kanemitsu, Thorough subcells diagnosis in a multi- junction solar cell via absolute electroluminescence-efficiency measurements, Sci. Rep. 5, 7836 (2015). [4] [5] T. Kirchartz, A. Helbig, and U. Rau, Note on the interpretation of electroluminescence images using their spectral information, Sol. Energy Mater Sol. Cells 92, 1621 (2008). [6] M. Seeland, R. Rösch, and H. Hoppe, Quantitative analysis of electroluminescence images from polymer solar cells, J. Appl. Phys. 111, 024505 (2012). [7] H. Nesswetter, W. Dyck, P. Lugli, A. W. Bett, and C. G. Zimmermann, Luminescence based series resistance mapping of III-V multijunction solar cells, J. Appl. Phys. 114, 194510 (2013). [8] A. Delamarre, M. Paire, J. -F. Guillemoles, and L. Lombez, Quantitative luminescence mapping of Cu(In, Ga)Se2 thin-film solar cells, Prog. Photovolt. Res. Appl. 23, 1305 (2015). [9] T. Mochizuki, C. Kim, M. Yoshita, J. Mitchell, Z. Lin, S. Chen, H. Takato, Y. Kanemitsu, and H. Akiyama, Solar-cell radiance standard for absolute electroluminescence measurements and open-circuit voltage mapping of silicon solar modules, J. Appl. Phys. 119, 034501 (2016). [10] A. Delamarre, L. Lombez, K. Watanabe, M. Sugiyama, Y. Nakano, and J. -F Guillemoles, Experimental demonstration of optically determined solar cell current transport efficiency map, IEEE J. Photovolt. 6, 528 (2016). [11] U. Aeberhard and U. Rau, Microscopic perspective on photovoltaic reciprocity in ultrathin solar cells, Phys. Rev. Lett. 118, 247702 (2017). [12] C. Donolato, A reciprocity theorem for charge collection, Appl. Phys. Lett. 46, 270 (1985). [13] C. Donolato, An alternative proof of the generalized reciprocity theorem for charge collection, J. Appl. Phys. 66, 4524 (1989). [14] T. Markvart, Photogenerated carrier collection in solar cells in terms of dark carrier distribution three dimensions, IEEE Trans. Electron Devices 44, 1182 (1997). [15] U. Rau and R. Brendel, The detailed balance principle and the reciprocity theorem between photocarrier collection and dark carrier distribution in solar cells, J. Appl. Phys. 84, 6412 (1998). [16] K. Misiakos and F. A. Lindholm, Generalized reciprocity theorem for semiconductor devices, J. Appl. Phys. 58, 4743 (1985). [17] M. A. Green, Generalized relationship between dark carrier distribution and photocarrier collection in solar cells, J. Appl. Phys. 81, 268-271 (1997). [18] D. E. Carlson and C. R. Wronski, Amorphous silicon solar cell, Appl. Phys. Lett. 28, 671 (1976). [19] B. Rech, T. Repmann, M. N. van den Donker, M. Berginski, T. Kilper, J. Hüpkes, S. Calnan, H. Stiebig, and S.Wieder, Challenges in microcrystalline silicon based solar cell technology, Thin Solid Films 511, 548 (2006). [20] D. J. Friedman, J. F. Geisz, S. R. Kurtz, and J. M. Olson, 1-eV solar cells with GaInNAs active layer, J. Cryst. Growth 195, 409 (1998). [21] K. H. Tan, S. Wicaksono, W. K. Loke, D. Li, S. F. Yoon, E. A. Fitzgerald, S. A. Ringel, and J. S. Harris Jr., Molecular beam epitaxy grown GaNAsSb 1 eV photovoltaic cell, J. Cryst. Growth 335, 66 (2011). [22] K. W. J. Barnham and G. Duggan, A new approach to high-efficiency multi-band-gap solar cells, J. Appl. Phys. 67, 3490 (1990). [23] T. Kirchartz and U. Rau, Detailed balance and reciprocity in solar cells, Phys. Stat. Sol. A 205, 2737 (2008). [24] T. Kirchartz, J. Mattheis, and U. Rau, Detailed balance theory of excitonic and bulk heterojunction solar cells, Phys. Rev. B 78, 235320 (2008). [25] A. Gerber, V. Huhn, T. M. H. Tran, M. Siegloch, Y. Augarten, B. E. Pieters, and U. Rau, Advanced large area characterization of thin-film solar modules by electroluminescence and thermography imaging techniques, Sol. Energy Mater Sol. Cells 135, 35 (2015). [26] PVcell [Software]. soft.com/products/solar/ Available from http://www.str- [27] W. Shockley and W. T. Read, Jr., Statistics of the recombinations of holes and electrons, Phys. Rev. 87, 835 (1958). [28] K. Toprasertpong, T. Inoue, Y. Nakano, and M. Sugiyama, Investigation and modeling of photocurrent collection process in multiple quantum well solar cells, Sol. Energy Mater Sol. Cells 174, 146 (2018). [29] A. Haug, Auger coefficients for highly doped and highly excited semiconductors, Solid State Commun. 28, 291 (1978). [30] A. H. Marshak and K. M. v. Vliet, Electrical current in solids with position-dependent band structure, Solid-State Electron. 21, 417 (1978). [31] See Supplemental Material in http://link.aps.org/supplemental/10.11 03/PhysRevApplied.11.024029 for the derivation of Eq. (21) and the behavior of the carrier distribution and the quasi-Fermi levels under illumination. [32] T. Kirchartz, J. Nelson, and U. Rau, Reciprocity between charge injection and extraction and its influence on the interpretation of electroluminescence spectra in organic solar cells, Phys. Rev. Applied 5, 054003 (2016). [33] P. Würfel, The chemical potential of radiation, J. Phys. C 15, 3967 (1982). [34] J. Wong and M. A. Green, From junction to terminal: Extended reciprocity relations in solar cell operation, Phys. Rev. B 85, 235205 (2012). [35] K. Toprasertpong, S. M. Goodnick, Y. Nakano, and M. Sugiyama, Effective mobility for sequential carrier transport in multiple quantum well structures, Phys. Rev. B 96, 075441 (2017). [36] D. A. B. Miller, D. S. Chemla, T. C. Damen, A. C. Gossard, W. Wiegmann, T. H. Wood, and C. A. Burrus, Electric field dependence of optical absorption near the band gap of quantum-well structures, Phys. Rev. B 32, 1043 (1985). [37] U. Aeberhard, Impact of built-in fields and contact configuration on the characteristics of ultra-thin GaAs solar cells, Appl. Phys. Lett. 109, 033906 (2016). [38] R. T. Ross, Some thermodynamics of photochemical systems, J. Chem. Phys. 46, 4590 (1967). [39] K. Toprasertpong, H. Fujii, T. Thomas, M. Führer, D. Alonso-Álvarez, D. J. Farrell, K. Watanabe, Y. Okada, N. J. Ekins-Daukes, M. Sugiyama, and Y. Nakano, Absorption threshold extended to 1.15 eV using InGaAs/GaAsP quantum wells for over-50%-efficient lattice- matched quad-junction solar cells, Prog. Photovolt. Res. Appl. 24, 533 (2016).
1904.12757	1	1904	2019-04-29T14:56:15	On the Angular Dependence of InP High Electron Mobility Transistors for Cryogenic Low Noise Amplifiers in a Magnetic Field	[ "physics.app-ph", "cond-mat.mes-hall" ]	The InGaAs-InAlAs-InP high electron mobility transistor (InP HEMT) is the preferred active device used in a cryogenic low noise amplifier (LNA) for sensitive detection of microwave signals. We observed that an InP HEMT 0.3-14GHz LNA at 2K, where the in-going transistors were oriented perpendicular to a magnetic field, heavily degraded in gain and average noise temperature already up to 1.5T. Dc measurements for InP HEMTs at 2K revealed a strong reduction in the transistor output current as a function of static magnetic field up to 14T. In contrast, the current reduction was insignificant when the InP HEMT was oriented parallel to the magnetic field. Given the transistor layout with large gate width/gate length ratio, the results suggest a strong geometrical magnetoresistance effect occurring in the InP HEMT. This was confirmed in the angular dependence of the transistor output current with respect to the magnetic field. Key device parameters such as transconductance and on-resistance were significantly affected at small angles and magnetic fields. The strong angular dependence of the InP HEMT output current in a magnetic field has important implications for the alignment of cryogenic LNAs in microwave detection experiments involving magnetic fields.	physics.app-ph	physics	On the Angular Dependence of InP High Electron Mobility Transistors for Cryogenic Low Noise Amplifiers in a Magnetic Field Isabel Harrysson Rodrigues,1, a) David Niepce,2 Arsalan Pourkabirian,3 Giuseppe Moschetti,3, b) Joel Schleeh,3 Thilo Bauch,4 and Jan Grahn1 1)GigaHertz Centre, Department of Microtechnology and Nanoscience - MC2, Chalmers University of Technology, SE-41296 Gothenburg, Sweden 2)Quantum Technology Laboratory, Department of Microtechnology and Nanoscience - MC2, Chalmers University of Technology, SE-41296 Gothenburg, Sweden 3)Low Noise Factory AB, Nellickevagen 22, SE-41663 Gothenburg, Sweden 4)Quantum Device Physics Laboratory, Department of Microtechnology and Nanoscience - MC2, Chalmers University of Technology, SE-41296 Gothenburg, Sweden The InGaAs-InAlAs-InP high electron mobility transistor (InP HEMT) is the preferred active device used in a cryogenic low noise amplifier (LNA) for sensitive detection of microwave signals. We observed that an InP HEMT 0.3-14 GHz LNA at 2 K, where the in-going transistors were oriented per- pendicular to a magnetic field, heavily degraded in gain and average noise temperature already up to 1.5 T. Dc measurements for InP HEMTs at 2 K revealed a strong reduction in the transistor output current as a function of static magnetic field up to 14 T. In contrast, the current reduction was in- significant when the InP HEMT was oriented parallel to the magnetic field. Given the transistor layout with large gate width/gate length ratio, the re- sults suggest a strong geometrical magnetoresistance effect occurring in the InP HEMT. This was confirmed in the angular dependence of the transis- tor output current with respect to the magnetic field. Key device parame- ters such as transconductance and on-resistance were significantly affected at small angles and magnetic fields. The strong angular dependence of the InP HEMT output current in a magnetic field has important implications for the alignment of cryogenic LNAs in microwave detection experiments involving magnetic fields. In many sensitive detection systems, high electron mobility transistor (HEMT) low-noise amplifiers (LNAs) at cryogenic temperatures (1-10 K) are used to read out tiny microwave signals. Some of these systems rely on the presence of a strong magnetic field, e.g. in mass spectrometry1 or detection of dark matter.2,3 A potential application for cryogenic LNAs in a magnetic field is magnetic resonance imaging.4 It has long been known that the sensitivity of the cryogenic LNA is affected by the presence of a magnetic field: when aligned perpendicular to the magnetic field, the noise temperature of a cryogenic AlGaAs- GaAs (GaAs) HEMT LNA was shown to be strongly degraded with increasing magnetic field.5 However, reports on the electrical behavior in a magnetic field for the cryogenic InGaAs-InAlAs-InP (InP) HEMT LNA - the standard component used in today's most sensitive microwave receivers - have so far been absent. Compared to previous work,5 we here report that the gain and noise properties for the cryogenic InP HEMT LNA are much more prone for degradation when exposed to a magnetic field. For the first time, we measured the InP HEMT at 2 K as a function of angular orientation with respect to the magnetic field. It is shown that the InP HEMT output current is limited by a strong geometrical magnetoresistance effect (gMR). The results suggest that even small a)Electronic mail: [email protected] b)Present address: Qamcom Research and Technology AB, Falkenbergsgatan 3, SE-412 85 Gothenburg, Sweden 2 misalignment of the cryogenic InP HEMT LNA in a magnetic environment is detrimental to read-out sensitivity. The sensitivity of the cryogenic InP HEMT LNA in a magnetic field was examined using a 10 T superconducting magnet. The LNA was a monolithic microwave integrated circuit (MMIC) chip consisting of three InP HEMT stages and passive components mounted in a gold-plated aluminum module. Neither the LNA module nor in-going passives exhibited any magnetic-field dependence as verified by additional experiments. The amplifier was a broadband design ranging from 0.3 to 14 GHz with a gain of 42 dB and average noise temperature of 4.2 K (0.06 dB).6 The LNA was mounted at the center of the magnet on the 2 K stage of an Oxford Instruments Triton 200 dilution refrigerator. In this arrangement, the LNA module was oriented perpendicular towards the magnetic field. Taking cable loss and an additional attenuator in account, the input signal was attenuated by 27 dB, which was necessary to thermalize the signal going from 300 K to 2 K as well as avoiding saturation of the amplifier. In Fig. 1, the gain and noise temperature for 3, 5 and 8 GHz are presented for the cryogenic InP HEMT LNA at 2 K when exposed to static magnetic fields up to 1.5 T. For all three measured frequencies, the LNA gain and noise temperature started to be affected around 0.25 T. From 0 to 1.5 T, the gain and average noise temperature degraded from 42 to 24 dB and 4 to 18 K, respectively, at 5 GHz. Beyond this field the gain was so low that we needed post amplification for measurements. The LNA degradation versus magnetic field was similar across the 3-8 GHz band. Compared to the earlier study for a GaAs HEMT LNA,5 the degradation for an InP HEMT LNA in the presence of a magnetic field is much stronger. This is connected to the higher electron mobility and sheet density of the two dimensional electron gas (2DEG) in the InP HEMT compared to the GaAs HEMT used in Ref.5. In order to better understand the cryogenic InP HEMT LNA behavior in a magnetic field, DC measurements were conducted on individual transistors at low temperature. The discrete InP HEMTs were fabricated in the same transistor technology7 used for the cryo- genic InP HEMT LNA measured in Fig. 1. We have fabricated two-finger device layouts with gate width Wg ranging from 10 to 100 µm and gate length Lg from 60 to 250 nm. The DC-characterization was carried out in a Quantum Design Physical Property Measure- ment System (PPMS). The transistor was electrically connected through wire bonding and mounted in a matched LC-network on a sample holder in the vacuum chamber of the cryo- stat and cooled down to 2 K. A static magnetic field ranging up to 14 T was then applied and DC measurements were performed using a Keithley 2604B source meter. The measurement setup in PPMS allows for sample rotation in the B-field. In Fig. 2, the geometry of the DC experiment is shown. A static magnetic field was applied in the z-direction. Fig. 2 presents a top view and side view of the two-finger InP HEMT where an angle of rotation θ is defined with respect to the magnetic field. θ could be varied between 0o and 180o degrees where 0o (180o) and 90o corresponded to measurements of the InP HEMT oriented parallel and perpendicular to the magnetic field, respectively. The cryogenic InP HEMT output current was first measured for θ = 90o. Figure 3 shows the source-drain current Ids versus source-drain voltage Vds for different gate-source voltage Vgs under a magnetic field of 0 T and 10 T. The device was a 2x50 µm InP HEMT with Lg=100 nm. Figure 3 (a) illustrates a typical InP HEMT measured under cryogenic conditions in the absence of a magnetic field. The kink behavior at high current level around Vds=0.4 V originates from electron traps in the interface layers of the heterostructure and is characteristic for these type of devices measured at low temperature.8 Since the LC network in the sample holder did not provide a perfect 50 Ω impedance environment for the transistor, oscillations occurred for lower Ids showing up as small current jumps at Vgs < -0.2 V, which do not influence the interpretations in this study. In Fig. 3 (b), the same measurement was repeated under a magnetic field of 10 T. It stands clear that Ids is extremely suppressed for θ = 90o when exposed to the magnetic field. Comparing Fig. 3 (a) and (b), the maximum Ids (Vds=1 V) was reduced by almost a factor of 100. The device still behaves as a transistor but with a much larger on-resistance and strongly reduced transconductance. Measurements for the InP HEMT shown in Fig. 3 for θ = 0o and θ = 90o in a magnetic field up to 14 T are summarized in Fig. 4. Ids was measured under saturation for two biases 3 FIG. 1. Gain and average noise temperature for the cryogenic InP HEMT MMIC LNA, measured at 2 K, for three different frequencies as a function of applied magnetic field. The LNA was aligned perpendicular to the magnetic field. Vds=0.4 and 0.6 V and normalized with respect to zero field. The strong suppression in output current for θ = 90o is visualized in Fig. 4. As comparison, the effect in Ids from the magnetic field for θ = 0o is almost negligible (a minor reduction beyond 10 T may be due to a slight mis-orientation of the HEMT in the field). No difference with regard to Vds is noted in Fig. 4. We also observe in Fig. 4 that for θ = 90o, the normalized Ids varies as B2. Moreover, it was confirmed that this dependence was the same for a range of various device sizes Lg (60, 100, 250 nm) and Wg (2x10, 2x50, 2x100 µm), see Fig. 5. In our experimental setup, the Hall effect (voltage) is negligible because of the device geometry with Wg >> Lg.9 The output current transport in the InP HEMT is therefore limited by gMR,10 which occurs due to the effect of the Lorentz force on the 2DEG in devices where Wg >> Lg. gMR is expected to be large under the experimental circumstances here, i.e. high channel mobility in the 2DEG and a strong magnetic field. The angular dependence of the cryogenic InP HEMT output current in a magnetic field was investigated by rotating the transistor in the magnetic field. θ was increased from 0o to 180o using a step size of 1o. In Fig. 6, Ids is plotted as a function of θ up to 10 T. It is noted that the Ids is clearly dependent on θ and that this dependence becomes larger with higher magnetic field. Irrespectively of the applied field strength, we observe no change in the output current when (the 2DEG channel of) the transistor has a 0o or 180o rotation. A significant reduction (∼ 20 %) in Ids for rotations as small as 15o occurs at an applied field of 3 T. With increasing magnetic field, the alignment of the InP HEMT becomes even 4 FIG. 2. The two-finger InP HEMT shown in top view (by scanning electron microscopy) and side view (by cross-sectional transmission electron microscopy) perspective. The T-shaped gate is marked in yellow. The static magnetic field was applied in z-direction. The device was rotated from θ = 0o to θ = 180o corresponding to parallel and perpendicular orientation, respectively, of the HEMT output current with respect to the magnetic field. FIG. 3. Ids versus Vds for the InP HEMT at 2 K oriented θ = 90o in a static magnetic field of (a) 0 T and (b) 10 T. Vgs was varied from -0.4 to 0.4 V in steps on 0.1 V. Wg =2x50 µm and Lg=100 nm. Note difference in y-axis scale for (a) and (b). more crucial and the Ids is reduced by a minor tilt in θ of a few degrees. For a transistor layout with Wg >> Lg, the gMR in the channel is expected to vary as 1+µ2B2, where µ is the electron mobility in the channel.9,10 Taking the angular dependence in Fig. 2 into account for a transistor output current in an electrical field subject to a Lorentz force leads to Ids(B, θ) = Vds R0(1 + µ2B2sin2θ) (1) where R0 is a resistance term (including channel and access resistance contributions in the 5 FIG. 4. Normalized maximum Ids versus magnetic field ranging from 0 to 14 T in steps of 1 T with θ = 0o (black) and θ = 90o (red), at Vgs = 0.4 V. Values for Vds = 0.6 V (triangle) and 1.0 V (dot) are plotted. Device size: Wg = 2x50 µm and Lg = 100 nm, measured at 2 K. InP HEMT). Fitting the data to Eq. 1 gives R0 = 11 Ω and µ = 10, 500 cm2/Vs, which is in excellent agreement for the full range of measured θ and B in Fig. 6 reflecting the symmetrical angular dependence in Ids around θ = 90o. The transconductance and on-resistance are two HEMT parameters fundamental for the design of a cryogenic LNA. The maximum transconductance gm,max at 2 K of the cryogenic InP HEMT is plotted in Fig. 7 (a) as a function of the applied magnetic field up to 14 T for various angles θ. The gm,max is around 1.9 S/mm at zero field and strongly decreases as a function of magnetic field for θ = 30o and above. Already at 1 T, the gm,max is reduced with 40 % (60 %) for θ = 30o (90o). Field-effect transistor transconductance versus magnetic field for θ = 90o has been reported in Refs.5 and11 for GaAs and silicon devices at cryogenic and room-temperature conditions, respectively. The dependence on magnetic field is much stronger for the InP HEMTs investigated in this study. In contrast to Ref.5, the gm,max here is directly calculated from measured I-V data for the transistor at cryogenic temperature in a magnetic field. Compared to Fig. 7 (a), Ref.11 shows a much weaker gm(B) dependence which is probably related to the silicon-based transistor subject to study, a device normally not used in cryogenic LNAs. The on-resistance for the InP HEMT (at 2 K) as a function of B for various θ is presented in Fig. 7 (b). For B=0 T, the Ron is around 0.7 Ωmm independent of B for θ = 0o. As expected from Figs. 3 and 4, Ron increases rapidly with B for θ = 90o, two orders of magnitude at 5 T. This dependence is also strong at smaller angles as illustrated for θ = 30o in Fig. 7(b). Ron is found to vary in a similar way as the denominator in Eq. (1): Ron = R0(1 + µ2B2sin2θ). Fig. 7 demonstrates that transistor parameters crucial for the design of a cryogenic LNA are highly sensitive to the alignment of the InP HEMT in a magnetic field at 2 K. This explains the strong degradation of the LNA gain and average noise temperature (measured for θ = 90o) observed in Fig. 1. In conclusion, we have investigated the angular dependence of the output current of cryogenic InP HEMTs in magnetic fields up to 14 T and found that it is greatly attenuated, not only at θ = 90o, but also at small θ. The physical reason is the very strong gMR occurring for Ids in the cryogenic InP HEMT. This was validated by an accurate fitting 6 FIG. 5. Normalized Ids versus magnetic field up to 10 T for InP HEMTs oriented θ = 90o with (a) Lg = 60, 100 and 250 nm and (b) Wg = 2x10, 2x50 and 2x100 µm. A fix Vgs and Vds of 0.4 V were applied in an ambient temperature of 2 K. of experimental Ids data with an equation describing the gMR as a function of B and θ. Furthermore, we have shown the strong influence from θ for the transistor parameters gm and Ron when the cryogenic InP HEMT is exposed to a magnetic field. As a result, the alignment of cryogenic InP HEMT LNAs with respect to a magnetic field is critical in sensitive microwave detection systems: even small deviation from θ = 0o (180o) leads to significantly lower gain and higher average noise temperature. This work was performed in GigaHertz Centre in a joint research project between Chalmers University of Technology, Low Noise Factory AB, Wasa Millimeter Wave AB, Omnisys Instruments AB and RISE Research Institutes of Sweden. We are grateful to Ser- guei Cherednichenko for valuable assistance in the noise measurements and Niklas Wadefalk for the LNA design. 1R. Mathur, R. W. Knepper, and P. B. O'Connor, IEEE Transactions on Applied Superconductivity 18, 1781 -- 1789 (2008). 2C. Hagmann, P. Sikivie, N. S. Sullivan, and D. B. Tanner, Phys. Rev. D 42, 1297 -- 1300 (1990). 3B. M. Brubaker, arXiv e-prints , arXiv:1801.00835 (2018). 4D. H. Johansen, J. D. Sanchez-Heredia, J. R. Petersen, T. K. Johansen, V. Zhurbenko, Ardenkjaer-Larsen, IEEE Transactions on Biomedical Circuits and Systems 12, 202 -- 210 (2018). 5E. Daw and R. F. Bradley, Journal of Applied Physics 82(4), 1925 -- 1929 (1997). 6J. Schleeh, N. Wadefalk, P. Nilsson, J. P. Starski, and J. Grahn, IEEE Transactions on Microwave Theory and Techniques 61, 871 -- 877 (2013). 7J. Schleeh, G. Alestig, J. Halonen, A. Malmros, B. Nilsson, P. A. Nilsson, J. P. Starski, N. Wadefalk, H. Zirath, and J. Grahn, IEEE Electron Device Letters 33, 664 -- 666 (2012). 8H. Rodilla, J. Schleeh, P. Nilsson, and J. Grahn, IEEE Trans. Electron Devices 62(2), 532 -- 537 (2015). 9J. P. Campbell, K. P. Cheung, L. C. Yu, J. S. Suehle, A. Oates, and K. Sheng, IEEE Electron Device Letters 32, 75 -- 77 (2011). and J. H. 10T. Jervis and E. Johnson, Solid-State Electronics 13, 181 -- 189 (1970). 11J. R. Bodart, B. M. Garcia, L. Phelps, N. S. Sullivan, W. G. Moulton, and P. Kuhns, Review of Scientific Instruments 69, 319 -- 320 (1998). 7 FIG. 6. Rotation sweep (θ from 0 to 180o) showing Ids (absolute values) for various externally applied magnetic fields 0, 1, 2, 3, 4, 6, 8 and 10 T at a fix Vgs and Vds of 0.4 V in an ambient temperature of 2 K. The dashed lines (black) are fitting of Eq.(1) to the experimental data points (colored). Device size Wg=2x50 µm and Lg = 100 nm. 8 FIG. 7. (a) gm,max and (b) Ron as a function of applied magnetic field up to 14 T for θ = 0o, 30o, 45o, 60o and 90o. Ambient temperature of 2 K and device size Wg = 2x50 µm, Lg = 100 nm. Vds = 1.0 V, Vgs = 0.4 V.
1910.06685	1	1910	2019-10-15T12:42:05	Atomic-Scale Patterning of Arsenic in Silicon by Scanning Tunneling Microscopy	[ "physics.app-ph" ]	Over the last two decades, prototype devices for future classical and quantum computing technologies have been fabricated, by using scanning tunneling microscopy and hydrogen resist lithography to position phosphorus atoms in silicon with atomic-scale precision. Despite these successes, phosphine remains the only donor precursor molecule to have been demonstrated as compatible with the hydrogen resist lithography technique. The potential benefits of atomic-scale placement of alternative dopant species have, until now, remained unexplored. In this work, we demonstrate successful fabrication of atomic-scale structures of arsenic-in-silicon. Using a scanning tunneling microscope tip, we pattern a monolayer hydrogen mask to selectively place arsenic atoms on the Si(001) surface using arsine as the precursor molecule. We fully elucidate the surface chemistry and reaction pathways of arsine on Si(001), revealing significant differences to phosphine. We explain how these differences result in enhanced surface immobilization and in-plane confinement of arsenic compared to phosphorus, and a dose-rate independent arsenic saturation density of $0.24{\pm}0.04$ ML. We demonstrate the successful encapsulation of arsenic delta-layers using silicon molecular beam epitaxy, and find electrical characteristics that are competitive with equivalent structures fabricated with phosphorus. Arsenic delta-layers are also found to offer improvement in out-of-plane confinement compared to similarly prepared phosphorus layers, while still retaining >80% carrier activation and sheet resistances of $<2 k{\Omega}/{\square}$. These excellent characteristics of arsenic represent opportunities to enhance existing capabilities of atomic-scale fabrication of dopant structures in silicon, and are particularly important for three-dimensional devices, where vertical control of the position of device components is critical.	physics.app-ph	physics	Atomic-Scale Patterning of Arsenic in Silicon by Scanning Tunneling Microscopy Taylor J. Z. Stock1, Oliver Warschkow2, Procopios C. Constantinou1, Juerong Li3, Sarah Fearn1,4, Eleanor Crane1, Emily V. S. Hofmann1,5, Alexander Kölker1, David R. McKenzie2, Steven R. Schofield1,6, Neil J. Curson1,7 1. London Centre for Nanotechnology, University College London, London WC1H 0AH, UK 2. Centre for Quantum Computation and Communication Technology, School of Physics, The University of Sydney, Sydney, NSW 2006, Australia 3. Advanced Technology Institute, University of Surrey, Guildford GU2 7XH, United Kingdom 4. Department of Materials, Imperial College of London, London SW7 2AZ, UK 5. IHP -- Leibniz-Institut für innovative Mikroelektronik, Frankfurt (Oder) 15236, Germany 6. Department of Physics and Astronomy, University College London, London WC1E 6BT, UK 7. Department of Electronic and Electrical Engineering, University College London, London WC1E 7JE, UK. AUTHOR INFORMATION Corresponding Authors Taylor J.Z. Stock [email protected] Neil J. Curson [email protected] 1 ABSTRACT Over the last two decades, prototype devices for future classical and quantum computing technologies have been fabricated, by using scanning tunneling microscopy and hydrogen resist lithography to position phosphorus atoms in silicon with atomic-scale precision. Despite these successes, phosphine remains the only donor precursor molecule to have been demonstrated as compatible with the hydrogen resist lithography technique. The potential benefits of atomic-scale placement of alternative dopant species have, until now, remained unexplored. In this work, we demonstrate the successful fabrication of atomic-scale structures of arsenic-in-silicon. Using a scanning tunneling microscope tip, we pattern a monolayer hydrogen mask to selectively place arsenic atoms on the Si(001) surface using arsine as the precursor molecule. We fully elucidate the surface chemistry and reaction pathways of arsine on Si(001), revealing significant differences to phosphine. We explain how these differences result in enhanced surface immobilization and in- plane confinement of arsenic compared to phosphorus, and a dose-rate independent arsenic saturation density of 0.24±0.04 monolayers. We demonstrate the successful encapsulation of arsenic delta-layers using silicon molecular beam epitaxy, and find electrical characteristics that are competitive with equivalent structures fabricated with phosphorus. Arsenic delta-layers are also found to offer improvement in out-of-plane confinement compared to similarly prepared phosphorus layers, while still retaining >80% carrier activation and sheet resistances of <2 kΩ/□. These excellent characteristics of arsenic represent opportunities to enhance existing capabilities of atomic-scale fabrication of dopant structures in silicon, and are particularly important for three- dimensional devices, where vertical control of the position of device components is critical. TOC GRAPHICS KEYWORDS scanning tunneling microscopy, density functional theory, atomic fabrication, silicon (001), arsenic, arsine, dopant 2 INTRODUCTION Challenges associated with the continued miniaturization of electronic devices toward the atomic- scale limit,1 and proposals for novel atomic-scale device architectures, e.g., to enable quantum information processing,2 have stimulated renewed interest in the investigation of dopant species in silicon. Of particular interest are methods for controlling the placement of dopant atoms in silicon with atomic-scale precision,3,4 and understanding the electronic5,6,7,8,9 and optical10,11,12 properties of single-atom and few-atom clusters of dopants. The atomic-scale control of dopant atoms in silicon using scanning tunneling microscopy (STM) has to date focused on phosphorus donors, which can be introduced into the silicon matrix using the precursor gas, phosphine (PH3). The chemical reaction of phosphine with the silicon (001) surface has been extensively researched and is now understood in exquisite detail.13 This chemical process can be spatially controlled at the atomic-scale via the use of STM hydrogen-desorption lithography 14 to produce atomically-defined patterns of phosphorus donors in silicon.15 This has led to numerous exciting advances in our understanding of silicon-based electronic devices at the atomic-scale,16,17,18,19 and is a promising avenue for the use of individual donor spins as qubits in a solid state quantum computer. 20,21,22 Arsenic donors in silicon are particularly interesting, since they have a lower diffusivity and a higher solid solubility in bulk silicon than phosphorus,23 as well as a higher ionization energy in silicon (53.76 meV, compared to 45.59 meV for phosphorus),24 a larger atomic radius (rAs =115 pm, rP = 100 pm),25 larger atomic spin-orbit interaction (ZAs = 33, ZP = 15), and a higher nuclear spin value (IAs = 3/2, IP = ½) than phosphorus. These differing properties present opportunities for atomic-scale device designs with advanced functionality, such as nuclear spin qudits (where a qudit is a generalized quantum information object with n > 2 quantum states),26,27 and silicon photonic crystal cavity based quantum computation schemes.28 Furthermore, the ability to position multiple dopant species in silicon with atomic-scale precision would allow independent addressing of each donor species, and will enable new principles of device operation, such as implementations of quantum error correction codes,29 and optically driven silicon-based quantum gates.30 Thus, the ability to control the reaction of arsine (AsH3) with Si(001) using STM hydrogen-desorption lithography, in an analogous manner to phosphine, presents enormously exciting opportunities for atomic-scale electronics. Kipp et al.31 have studied the adsorption of arsine on Si(001) using STM and X-ray photoemission spectroscopy and suggest that AsH3, adsorbed at room temperature, decomposes into AsH + 2H and saturates at an As concentration of ~20% of a silicon monolayer (ML). Several other reports on the surface chemistry and growth kinetics of arsenic layers on Si(001) from a gas phase arsine precursor32,33,34,35,36 have been primarily concerned with rapid growth rates and elevated growth temperatures (> 600 °C), which are not compatible with the thermal requirements of STM hydrogen-desorption lithography.37,38 In general however, these reports have established that arsine gas phase dosing of Si(001) is self-limiting at an arsenic coverage of ~0.25-1 ML, dependent on the growth temperature, and therefore on the degree of H2 desorption. Previous density functional theory (DFT) studies have considered likely dissociation configurations of AsH3,39 kinetic activation barriers for the removal of the first H atom,40 and comparative thermodynamic stabilities of dissociated AsH3 and PH3.41 Here, we present a combined atomic-scale STM and DFT investigation of the adsorption and thermal decomposition of arsine on Si(001). We establish the full chemical pathway of arsine 3 decomposition and incorporation of arsenic into the Si(001) surface, and also the spatial control of this reaction via STM hydrogen-desorption lithography. In particular, we demonstrate the following: 1) Individual AsH3 molecules rapidly and fully dissociate on Si(001) at room temperature. 2) Arsenic atoms from dissociated AsH3 molecules incorporate substitutionally into the silicon lattice at temperatures below 350 °C, and at all surface coverages. 3) Arsine adsorption on Si(001), at room temperature, is self-terminating at a saturation coverage of 0.24 ± 0.04 molecules/surface silicon atom, independent of gas exposure rate, providing an arsenic coverage well above the metal-insulator transition, allowing for STM fabrication of metallic interconnects. 4) AsH3 molecules selectively adsorb and dissociate in a STM depassivated region of the Si(001)- H surface, while leaving the surrounding H-resist intact; subsequent annealing incorporates the arsenic atoms into the top layer of the silicon surface, without disturbing the H-resist. 5) Arsenic delta-doped layers can be grown by confining the arsenic incorporated surface below an epitaxial silicon capping layer, grown via temperature-programmed, low temperature epitaxy. 6) Ohmic contacts can be made to the buried, and electrically activated, arsenic-doped layer by aluminum deposition. Fulfillment of these six criteria satisfies the necessary requirements for successful STM fabrication of atomic-scale, electrically contacted, arsenic in silicon electronic devices. While we are aware of unpublished progress towards STM patterning of boron in silicon,42 the present work represents the first development of the capability to atomically pattern any dopant species, other than phosphorus, in silicon, using hydrogen-desorption lithography. Moreover, this work represents a first step in the development of a materials toolbox, where atomic-scale structures can be constructed using multiple dopant elements; an important step closer towards the emerging field of atomic-scale fabrication and manufacturing. RESULTS AND DISCUSSION 1. Dissociation of Isolated AsH3 Molecules To ascertain the atomic-scale behavior of the AsH3/Si(001) adsorption system, we exposed atomically clean Si(001)2×1 to a low dose of arsine (0.015 L at 5×1011 mbar) at room temperature. An example of the surface observed by STM, immediately after dosing, is shown in the filled and empty-state STM images of Figure 1a and 1b. Of the features observed, several are readily identified as native Si(001) surface defects,43 such as dimer vacancies (labeled DV in Figure 1a and 1b ) and C-defects (labeled Cd).44 The new features in these images are attributed to the products of AsH3 dissociation reacting with the surface. These are labelled "type-1" and "type-2". The type-1 features are by far the most common, accounting for more than 75% of all AsH3- associated features detected at these coverages, and these are the principal focus here. The minority type-2 features arise due to steric effects, which we briefly discuss below. 4 Figure 1: Adsorption and dissociation of single AsH3 molecules on Si(001)2×1 a,b) Filled and empty state STM images of low- coverage AsH3 on Si(001). AsH3 features are labelled type-1 and type-2, native surface defects are dimer vacancies (DV), double dimer vacancy (2DV), 2+1 dimer vacancy complex (2+1DV), and C-defect (Cd). V = -1.7 V, +1.6 V, I = 0.10 nA. c) Filled-state STM image of a type-1 feature, reticle highlights alignment with Si dimers. V = -1.7 V, I = 0.10 nA. d) Schematic valence diagram details the inter-row end-bridge As+3H structure of the type-1 feature. Grey shading indicates regions imaging bright in STM. e) Empty-state STM image of a type-1 feature, reticles highlight alignment with Si dimers. V = +1.6 V, I = 0.10 nA. f) Structural schematic of inter-row end-bridge As+3H structure showing the positions of first and second layer silicon atoms (black and grey, respectively), and the single As (dark blue) and three H atoms (light blue) provided by the AsH3 molecule. (g) Five-step reaction path from molecularly-adsorbed AsH3 to the fully dissociated inter-row end-bridge As+3H structure. Calculated adsorption energies are provided in brackets for each sequential structure. Calculated activation energies, EA, for each step are given above the reaction arrows. The STM images are skewed to account for thermal drift and rotated ~45° from the scan direction such that the dimer-rows align with the horizontal. High-resolution, filled- and empty-state images of the type-1 feature are shown in Figure 1c and 1e, respectively. The type-1 feature is characterized by two bright protrusions on two adjacent dimer rows. These protrusions appear in both filled and empty states, but differ in their relative brightness. Across the dimer rows, i.e., along the vertical axis of the images in Figure 1c and 1e, both protrusions are shifted up from the dimer centers, towards an upper dimer end. Divisions of the overlaid white reticles on the STM images indicate the dimer center positions in the horizontal direction. By reference to these reticles, we see that the upper row protrusion is positioned on top of a single dimer end, while the lower-row protrusion is between two dimer ends. The upper protrusion induces strong dimer pinning, which is apparent in the filled state image (Figure 1c) as the characteristic zig-zag appearance along the row. 5 We assign the type-1 STM feature to the structure shown in Figure 1d, which we call the inter- row end-bridge As+3H configuration. Grey highlighting in the schematic indicates sites that appear bright in the corresponding filled-state image. These bright sites correspond to the bare Si- Si dimers that surround the adsorbate, and to the two prominent protrusions of the type-1 feature. The two protrusions in the STM images are assigned to the dangling bond end of a Si-Si-H hemihydride dimer on the upper row, and a dimer end-bridge arsenic atom on the lower row (where the remaining two hydrogen atoms are attached to the lower row, directly across from the end- bridge arsenic atom). These two sites are both characterized by a single unpaired electron, i.e. there are half-filled molecular orbitals at these sites that would nominally image brightly in both filled and empty state. The two dimer ends opposite the bridging arsenic atom appear dark in both the filled and empty state image, which is consistent with the electronically saturated, hydrogen- termination at these sites. Figure 1g shows the proposed reaction pathway giving rise to the inter-row end-bridge As+3H configuration observed as the type-1 feature. This kinetically preferred pathway is based on DFT calculations of adsorption energies of the various intermediate structures involved and the activation energies for each step along the path. Each successive step leads to a structure of lower energy, providing the thermodynamic driving force towards the final formation of the inter-row end-bridge As+3H configuration. The first reaction step, labeled (i) in Figure 1g, is the dissociation of a datively-adsorbed AsH3 molecule into an AsH2+H species by means of an inter- row hydrogen shift. This reaction transfers one hydrogen atom to the adjacent dimer row, producing a hemihydride dimer. The calculated activation barrier for this reaction step is only 0.23 eV, which corresponds to a reaction that is completed on a timescale of nanoseconds at room temperature. We note that the inter-row hydrogen shift is kinetically preferable to the alternative, (and perhaps more intuitive) on-dimer and inter-dimer hydrogen shifts (not shown in Figure 1g), which have calculated activation energies of 0.53 eV and 0.29 eV, respectively. The fact that the inter-row H-shift is preferred, correlates well with the fact that the type-1 STM feature, contains a hemihydride on an adjacent row, and that a majority of adsorbed AsH3 molecules dissociate by this route. The second reaction step, labeled ii) in Figure 1g, is also a hydrogen-shift reaction, which preferentially occurs in the on-dimer direction to produce a structure that we refer to as the dimer- end AsH+2H structure. The calculated activation energy for this reaction is 0.63 eV, corresponding to a reaction timescale of milliseconds at room temperature. The alternative reaction, in which the hydrogen atom shifts in the inter-dimer direction (not shown in Figure 1g), has a slightly larger activation barrier of 0.72 eV, and is thus kinetically disfavored. The dimer-end AsH+2H configuration containing a single-valent AsH fragment is a highly transient intermediate corresponding to an extremely shallow minimum on the potential energy surface. This species near-instantly stabilizes into the end-bridge AsH+2H species, in which the AsH fragment bridges between two dimer-ends, labeled iii) in Figure 1g. The final reaction step, labeled iv) in Figure 1g, transfers a hydrogen atom from the AsH fragment to the silicon dangling bond at the opposite dimer end. This reaction produces the inter-row end- bridge As+3H structure that corresponds to the type-1 STM feature. The reaction barrier for this final hydrogen-shift is calculated to be 0.89 eV; this is the rate-determining activation energy along the full dissociation path. This barrier is still low enough such that full dissociation of an adsorbing AsH3 molecule into inter-row end-bridge As+3H will be completed on a timescale of tens of 6 seconds. Overall, the calculated reaction pathway is thus highly consistent with observation of the type-1 STM feature as the majority species following AsH3 chemisorption on Si(001). Despite the kinetic preference for this dissociation pathway, STM images also reveal small numbers of a few alternative dissociation structures, e.g., the type-2 feature seen in Figure 1a,b. These arrangements, which are kinetically disfavored on the bare Si(001) surface, are thought to result from steric hindrance provided by neighboring AsH3 adsorbates, or native Si(001) defects. A single example of the type-2 feature is shown in Figure 1, in the bottom right of panels a and b, and notably occurs immediately adjacent a type-1 feature. At this low coverage, the type-2 feature comprises roughly 10-15% of the total AsH3 associated STM features, and corresponds to the intra-row end bridge structure, known to be the preferred dissociation configuration for PH3/Si(001).13 It is instructive here to briefly compare the dissociation of AsH3 on silicon to the analogous PH3/Si(001) system. The primary PH3 dissociation process at low coverage is a progressive sequence of three principal species, namely, PH2+H, PH+2H, and P+3H.45 Transitions along this sequence occur on a timescale of minutes, making these transitions directly observable in room- temperature STM experiments (see e.g. Figure. 6 in 45). In contrast, full AsH3 dissociation into the As+3H species is complete before the surface can be imaged by STM after arsine exposure, leaving the type-1 STM feature as the single dominant species at low coverage. Comparison of our DFT calculations for AsH3/Si(001) with those reported for phosphine13 provides some insights into how these differences arise. First, hydrogen-shift reactions in the arsine pathway are more exothermic and have lower activation energies than the corresponding reaction in the phosphine pathway. This results in a much faster rate of dissociation for AsH3. Second, a characteristic of the phosphine pathway is that hydrogen-shift dissociation and PH2--fragment diffusion are in close competition; this becomes manifest in the great diversity of PH3 dissociation species formed,41 and in the observed diffusion of the P atom away from the initial adsorption site.13 This competition no longer exists for arsine due to the much reduced hydrogen-shift activation barrier. In effect, an adsorbing AsH3 molecule rapidly falls apart where it lands, undergoing minimal rearrangement of the fragments. The result is that the arsenic atom does not have opportunity to diffuse away from the initial landing site and a single structure, namely the inter-row end-bridge As+3H structure, is formed as the majority species. 2. Substitutional Incorporation of Arsenic Atoms into the Silicon Lattice Following molecular dissociation of AsH3 on the Si(001) surface, the next step critical to the STM lithography fabrication process, is incorporation of the arsenic atoms into the silicon lattice. Figure 2a and 2b show a clean Si(001) surface exposed to a low dose (0.015 L) of arsine at room temperature (Figure 2a), and then subsequently annealed to promote the incorporation of the arsenic atoms into the surface (Figure 2b). Before annealing, the surface exhibits many As+3H features (type-1), two of which have been highlighted by arrows. High-resolution filled- and empty-state images of the same surface (inset to Figure 2a) provide a close view of a type-1 feature (arrow) and a nearby C-defect. Figure 2b shows the same surface (but a different area) after a one- minute thermal anneal at 500 °C. Arrows in Figure 2b highlight examples that, by their appearance and by analogy to the behavior of PH3 on Si(001),46 we assign to Si-As heterodimers formed during the anneal. High-resolution filled and empty state images of two such As-Si heterodimers (inset of 7 Figure 2b) reveal an appearance that is very similar to that of P-Si heterodimers, confirming that arsenic incorporation into the surface has taken place. At the temperature of 500 °C used in the annealing experiment above, all hydrogen atoms are expected to desorb from the silicon surface, and thus Figure 2b shows a surface completely depleted of hydrogen. At this temperature, the substituted silicon atoms, which are ejected from the surface as the As-Si dimers are formed, rapidly diffuse away from the ejection site and are eventually captured at the existing step edges, or newly formed ad-dimer rows.47 This process of controllably substituting arsenic atoms into the silicon lattice is a requirement for compatibility of arsine and arsenic with STM hydrogen resist lithography, as substitution of isolated dopant atoms is necessary if the dopants are to be activated as electron donors in STM fabricated electronic devices. Figure 2: Substitutional incorporation of isolated arsenic atoms into the silicon lattice and saturation arsine coverage (a) STM image of low coverage (0.02 adsorbates/nm2), room temperature AsH3 on Si(001). Arrows indicate examples of isolated type-1 AsH3 features. V = -1.7 V, I = 0.02 nA The inset shows filled and empty state images of the type-1 feature alongside a C-defect for comparison. (b) Surface from image (a) subsequent to a 500 °C × 1 min anneal. Isolated As-Si heterodimers (HD) are observed in the first layer of the silicon surface, examples are indicated by white arrows. V = -1.6 V, I = 0.10 nA. Two examples of the As-Si HD are shown in filled and empty states in the inset. (c) Kinetic Monte Carlo simulation of total As surface coverage and AsHx species distribution, as a function of arsine partial pressure. (d) STM image of a Si(001) surface following a 1.5 L (5  10-9 mbar × 5 minutes) room temperature, arsine exposure. Inset shows a magnified image of the same surface. V = -2.0 V, I = 0.04 nA. (e) Surface from image (a) subsequent to a 350 °C × 1 min anneal. Inset shows ejected silicon dimer chains indicative of arsenic incorporation. V = -2.0 V, I = 0.05 nA. 8 3. Saturation Arsine Coverage Having resolved the elementary AsH3 single-molecule processes and arsenic surface incorporation in the low-coverage regime, we now move on to the saturation-coverage regime. High-density regions of arsenic dopants are important for STM lithography device fabrication, as these conductive 2D areas serve as device interconnects and contacts. Figure 2d shows a large-scale STM image of the Si(001) surface dosed to saturation by 1.5 L of arsine. The image shows a disordered pattern of protrusions and depressions as expected for a surface densely covered with the products of AsH3 dissociation. Hydrogen atoms bonded to Si(001) are immobile at room temperature, and in the stable configuration of the inter-row end-bridge As+3H structure, the arsenic atom is locked into place by the three hydrogen atoms surrounding it. Activation barrier considerations reveal that arsenic incorporation can only take place when the thermal conditions are such that hydrogen atoms can diffuse away from the arsenic atom. The result of annealing the arsine saturated surface to 350°C for 5 minutes is shown in the STM images in Figure 2e, this anneal allows the adsorbed hydrogen atoms formed in the course of dissociation to diffuse over the surface, creating space for arsenic atoms to incorporate into the surface dimer layer. That incorporation has occurred is evidenced in Figure 2e by the bright linear protrusions running perpendicular to the underlying dimer rows (also see inset to Figure 2b). These bright linear protrusions correspond to short ad-dimer chains, formed from the silicon atoms ejected during arsenic incorporation. These ad-dimer chains are a familiar sight from phosphorus incorporation experiments in the context of PH3 lithographic fabrication.3, 47 Because there is one ejected silicon atom for every incorporated arsenic atom, one can interpret the density of ejected silicon atoms observed in STM images as a direct measure of the density of incorporated arsenic. Using this approach we estimate the incorporated arsenic coverage in samples prepared by a saturation arsine dose of 1.5 L and thermal anneal of 350 °C to be 0.24+/-0.04 ML. We find that this incorporated arsenic coverage is independent of the arsine saturation dose rate, which is interesting since this is not the case for phosphine (discussed further below). Saturation dosing at different rates can be achieved by changing the arsine partial pressure, with simultaneous adjustment of the total exposure time. When the arsine dose rate is varied by decreasing the arsine partial pressure from 5  10-9 mbar to 5  10-10 mbar (along with a dose equalizing change in total exposure time from 5 minutes to 50 minutes), the resulting total arsenic coverage is found to be constant within measurement error. This dose-rate independent saturation arsenic coverage can be explained as follows. (i) At low coverages, AsH3 moieties dissociate rapidly and completely at random locations. (ii) As the density of surface adsorbates increases, the occupied sites begin to sterically hinder the complete dissociation of further incoming AsH3, resulting in the subsequent adsorption of a distribution of partially dissociated AsHx moieties -- a distribution determined by site availability, not dissociation rate. (iii) Eventually all adsorption sites are blocked and coverage becomes saturated, leaving a final surface comprised of all possible AsHx moieties. The constant incorporated arsenic coverage value, measured across a range of arsine dose rates, is also found to be consistent with predictions from simple kinetic Monte Carlo (KMC) simulations. The results of these simulations are plotted in Figure 2c. Here, we use KMC calculations to 9 determine the total arsenic coverage at saturation, for a range of arsine partial pressures, spanning 5 orders of pressure magnitude, in which the distribution of AsHx fragments is allowed to vary. We assume that isolated molecules will dissociate in the type-1 structure, with a 2.5 dimer footprint discussed in section 1, but that as the surface fills in, and silicon site availability becomes reduced, AsH3 dissociation will be halted at different stages leaving behind AsH and AsH2 fragments. Although the ratio of As, AsH, and AsH2 varies as a function of partial pressure, the total number of arsenic atoms present on the surface remains in the 0.24-0.26 ML range, and the arsenic surface concentration only increases by 2-3% as the dose rate is increased across the full range. This dose rate independence of arsenic coverage in the saturated AsH3/Si(001) system is in direct contrast to the PH3/Si(001) system, where phosphorous coverage at saturation is found to be dose- rate dependent at room temperature.48,49,50 When Si(001) is saturated with PH3 there is competition over available bare surface dimers between the adsorbing PH3 molecules and the dissociating PH2+H species. For AsH3, this competition no longer exists, due to a dissociation rate that is three orders of magnitude more rapid than that of PH3. In effect, any adsorbing AsH3 molecule breaks apart on the Si(001) surface to a degree limited only by the availability of bare Si dimer sites surrounding the molecule. The much more rapid dissociation of AsH3 also manifests in a lack of ordering in this system. The saturated surface and magnified inset shown in Figure 2d show very little evidence of long- or short-range ordering. This is once again in direct contrast to the PH3/Si(001) system, where PH2 moieties are found to assemble in small patches of local p(2×2) ordering, with PH moieties aligned along Si dimer rows.48 In the case of PH3, dissociation and removal of successive H atoms from the PHx moieties occurs on the time scale of minutes, and the moieties survive long enough to undergo diffusion from the initial adsorption site to energetically preferred, ordered sites. In the AsH3/Si(001) system, the H atoms are stripped from the molecule much more rapidly, with isolated molecules completely dissociating on the timescale of tens of seconds. The AsHx moieties do not have opportunity to diffuse before fully dissociating, and at saturation coverage they remain immobilized at the initial, random landing site, producing the completely disordered surface shown in Figure 2a. 4. Adsorption and Incorporation of Arsenic through a Hydrogen Resist We now demonstrate that the arsine adsorption, dissociation, saturation, and incorporation on the clean Si(001) surface, as described above, is compatible with hydrogen desorption lithography. For compatibility with the STM fabrication process, AsH3 molecules must selectively adsorb only on depassivated Si(001) regions and not on the passivated Si(001):H surface. Furthermore, to allow for maximum in-plane confinement, and fabrication flexibility, arsenic should also incorporate into the silicon lattice at anneal temperatures below the onset of H2 desorption (380 °C). Fulfilment of both these requirements is demonstrated in Figure 3. Each panel in this figure corresponds to one step in the STM lithography patterned doping process, with the corresponding steps indicated by the inset schematics on each STM image. 10 Figure 3: Compatibility of AsH3 adsorption, dissociation, and incorporation with STM hydrogen resist lithography Sequential STM images of the same 100×100 nm area on a Si(001)-H surface showing that the H-resist remains intact and free from adsorbates throughout each step of the STM lithography process. (a) clean hydrogen terminated Si(001) surface prior to STM lithography or arsine exposure. b) selected area hydrogen desorption of a 50×50 nm square (V = +7.0 V, I = 1 nA, s = 100 nm/s) c) 1.5 L arsine dose results in selective adsorption only within the lithographically defined area. d) 350 C × 1 minute anneal, results in incorporation of the adsorbed arsenic atoms as evidenced by presence of ejected Si. e) shows an enlargement of the upper right-hand corner of the patterned square in figure (c). The disordered saturation arsenic layer is clearly contained within only the patterned region, and there is no adsorption on the resist. f) shows an enlargement of the upper right-hand corner of the patterned square in figure (d). Short Si dimer chains, evidence of arsenic incorporation, are indicated by the arrows. The surrounding hydrogen resist remains essentially intact with only a small density of additional dangling bonds produced by the incorporation anneal. V = -2.0 V, I = 0.10 nA. In the first step, a hydrogen passivated Si(001) surface is prepared, as shown in Figure 3a. Each silicon atom has one or two hydrogen atoms bonded to it, except in a few cases where the hydrogen atom is missing, leaving a dangling bond (DB). The DBs have the appearance of bright dots in the image and the DB density is < 1%. Selective removal of a region of the hydrogen 'resist' layer is achieved by scanning over a 50×50 nm region of the surface using an increased bias/current of +7.0 V and 1.0 nA, at a tip raster speed of 100 nm/s (Figure 3b). The patterned region, in the center of the image, shows the density of bright dots now approaches 100% coverage of the surface, indicating that most of the hydrogen atoms have been desorbed from the surface, leaving behind the bare silicon. 11 Dosing the patterned surface of Figure 3b with 1.5 L of arsine results in AsHx moieties (where x = 0 to 3) selectively adsorbing within the patterned area, producing a saturation arsenic density of 0.24 ML. An image of this surface is shown in Figure 3c. Following arsine dosing, the exposed region of the surface has a different appearance, consisting of a high density of features, whose identity is hard to distinguish due to their tight and disordered packing. An enlargement of the upper right corner of the patterned and dosed area is shown in Figure 3e. It is clear from these images that arsenic containing molecules from the gas phase have attached to the silicon surface where the hydrogen resist has been removed, while the resist protects the surface from AsH3 adsorption elsewhere. In the final step, shown in Figure 3d, annealing the surface to 350°C for 2 minutes results in substitutional incorporation of arsenic and the ejection of silicon onto the surface, exclusively within the patterned area. Once again, an enlargement of the upper right-hand corner of the patterned and dosed area (this time post-anneal) is shown in Figure 3d. Two examples of short dimer chains consisting of ejected silicon atoms are highlighted with arrows. Confirmation of the successful execution of this final step demonstrates that an arsenic dopant layer can be substitutionally incorporated into an arbitrarily defined region of a silicon surface, where the shape of the region is defined by the hydrogen desorption pattern written by an STM tip. 5. Encapsulation of Arsenic Layers: Low Temperature Silicon Epitaxy The final, in situ, ultra-high vacuum step in the STM device fabrication is encapsulation with crystalline silicon. Once the arsenic atoms are incorporated into the silicon surface through a patterned hydrogen resist, the dopant structure must then be buried beneath a few nanometers of epitaxial silicon. This protective capping layer is grown via molecular beam epitaxy, using a solid intrinsic silicon sublimation source (SUSI-40, MBE Komponenten GmbH). The thermal budget of the sample during the epitaxy stage is tightly restricted by two competing requirements: 1) dopant segregation and diffusion must be minimized so that the substitutional arsenic atoms remain in position -- demanding a minimization of sample temperature; 2) crystallographic defect formation must be minimized to ensure that arsenic donor atoms remain electrically activated -- usually achieved by elevated temperature overgrowth.51 In the case of phosphorous, these two requirements can be reasonably satisfied by holding the sample temperature constant at 250 °C, throughout. In the case of arsenic we find (via secondary ion mass spectrometry; SIMS) that this approach is inadequate. We attribute this to the fact that arsenic is considerably more prone to surface segregation.52 As described below, we have determined a variable sample temperature program during overgrowth that suppresses arsenic surface segregation during silicon epitaxy, allowing for the growth of both well confined and electrically activated arsenic δ-layers. To achieve thermal control of arsenic segregation and diffusion during silicon epitaxy, a number of saturation arsenic delta layers were overgrown with 15 nm of silicon at a rate of 1 ML/min, using different sample temperature programs. Arsenic concentration depth profiles were subsequently measured using SIMS, and Figure 4 shows SIMS data for four different temperature programs as described by the temperature vs time insets in each SIMS profile. Figure 4a shows the result of silicon overgrowth at a constant sample temperature of 240 °C. This arsenic distribution is dominated by a high and narrow peak located at a depth of ~1 nm. Beneath this 12 narrow surface peak the arsenic concentration reduces to a much lower, but not an insignificant constant level (> 1  1018 atoms/cc) before terminating in a slight peak at the δ-layer origin depth, 15nm below the surface. Beyond this depth, the arsenic concentration rapidly drops below the instrument's detection limit (1×1018 atoms/cm3). From this we conclude that the arsenic layer is not effectively buried beneath the silicon, but rather, a large percentage of the 0.24 ML of arsenic atoms undergo continual cycles of surface segregation, i.e. they exchange positions with newly arrived silicon adatoms, and assume energetically preferred positions within the top layer.53,54 Figure 4b shows the arsenic profile for overgrowth at a constant but lower temperature of 190 °C. Here we observe a shift in weight from the surface to the subsurface arsenic peak, and find that the arsenic concentration is no longer uniform in the region between peaks. Below the surface peak, it slowly rises to an asymmetric peak at the arsenic origin depth of ~15nm, before rapidly falling off below the detection limit. Segregation is an activated process, and by lowering the constant growth temperature the segregation is found to be partially suppressed. This arsenic segregation suppression can be further enhanced through a greater reduction in substrate temperature during encapsulation. Figure 4c shows the arsenic profile resulting from overgrowth at the lowest achievable sample temperature given our silicon source -- sample geometry. In this case, there is no Joule heating of the sample, only the unavoidable radiative heating from the silicon MBE source. The substrate begins overgrowth at 40 °C increasing to a stable 100 °C after the first 12 minutes of growth (12 ML). From the SIMS profile, we see that arsenic surface segregation is highly suppressed in this sample, resulting in a very well confined (<2 nm FWHM) and symmetric peak (~ 5  1020 atoms/cc) located at the origin depth of 15 nm. This peak, which falls rapidly to below the instrument's detection limit within 3 nm in either direction, represents the best achievable confinement of the arsenic δ-layer grown in our MBE system. However, due to the reduced substrate temperature during overgrowth, and the high concentration of arsenic surface atoms, the critical thickness at which the epitaxial layer becomes amorphous will also be reduced considerably.55 At a temperature of 100 °C, and deposition rate of 1 ML/min, the critical thickness is expected to reduce below 2 nm,51 meaning the upper 13 nm or more of silicon will be amorphous, and unsuitable for device encapsulation. In order to facilitate suppression of arsenic segregation during reduced temperature silicon overgrowth, yet still achieve epitaxy and avoid amorphization, we have developed the strategy (adapted from a similar concept used for phosphorus δ-layers in silicon 56,57,58) of growing a very thin, reduced temperature "locking layer" that is kept below the critical amorphization thickness. With a reduced temperature locking layer in place, the substrate is then heated to a temperature that can support epitaxial growth for the remainder of the overlayer deposition. Figure 4d shows an example of locking layer assisted epitaxial growth applied to an arsenic δ-layer. Here the sample is not actively heated for the first 10 minutes (10 ML for a 1 ML/min growth rate) after which it is then rapidly heated to 250 °C, and held constant for the remaining 230 minutes of overgrowth. Comparing the SIMS profiles of Figure 4c and d, we see that increasing the temperature to 250 °C following growth of a 10 ML locking layer, has no detrimental effect on the confinement. The locking layer strongly suppresses arsenic surface segregation, producing epitaxially capped arsenic δ-layers with very tight confinements; confinements which are limited primarily by the solid-state diffusivity of arsenic in silicon. It is interesting to note here that arsenic has a lower diffusion coefficient than phosphorus in bulk silicon23, which suggests that with the segregation issue 13 overcome, using arsenic in place of phosphorus should allow for fabrication of denser and more tightly confined dopant structures, in every dimension. Figure 4: Arsenic δ-layer distribution and electronic transport controlled through a silicon overgrowth thermal annealing program a-e) SIMS profiles of five different saturation arsenic δ-layer samples overgrown with 15 nm Si, at a deposition rate of 1 ML/min. Sample temperature program during silicon epitaxy was modified for each sample according to the inset temperature vs time profiles. f) δ-layer mobility and confinement as a function of RTA temperature. g) δ-layer carrier density and sheet resistance as a function of RTA temperature. Error bars in the plots in panels f and g account for temperature measurement uncertainty and magnetoresistance curve fitting confidence levels as explained in the supplementary material, section b. 6. Electrical Activation and Contacting of Arsenic δ-layers To establish full compatibility of arsine and arsenic with the STM device fabrication process, the buried arsenic dopant structures must be electrically activated and Ohmic contacts achieved with the arsenic layer beneath the silicon cap. Satisfaction of this requirement has been verified by measurement of during standard magnetotransport characterization of the arsenic δ-layers. As explained in detail in the supplementary material, Hall bars were fabricated from locking layer encapsulated arsenic δ-layers, and Ohmic contacts achieved through deposition of aluminum directly onto the δ-layer edge. linear and symmetric IV curves In addition to fabricating Ohmic contacts, we have also optimized the transport properties of the arsenic δ-layers for improved device performance. The optimizations were achieved by improving the epitaxial silicon quality and minimizing crystallographic defects, resulting in a maximization 14 of carrier density and mobility, and a minimization of sheet resistance. Despite the fact that the locking layer thickness used is below the critical amorphization limit, including this reduced temperature step in the overgrowth is detrimental to crystal quality, and thus to the electrical performance of the δ-layers. To overcome this issue, we include a rapid thermal anneal (RTA) stage in the overgrowth sample temperature program, as has been done for phosphorus δ- layers.56,57,58 Figure 4e shows the SIMS profile of an arsenic δ-layer grown using both locking layer and RTA stages in the substrate temperature program. The temperature vs time profile used here is shown in the inset of Figure 4e. It is similar to the profile shown in the inset of Figure 4d, with the exception of the additional 500 °C × 10 second RTA, occurring between the locking layer growth and constant temperature epitaxy stages. Comparing the SIMS profiles of Figure 4 (d) and (e) we find that introduction of this RTA stage is not detrimental to the enhanced confinement provided by the locking layer in any significant way. The effect of the RTA on transport properties is however significant. Figure 4f and g show the effect of the RTA temperature maximum on δ-layer carrier concentration, sheet resistance, mobility, and confinement. The data for these figures of merit were obtained from Hall measurements on five arsenic δ-layers, with heating profiles identical to that of Figure 4e, except with each using a different RTA peak value, covering the range from 250 °C to 560 °C. For a RTA greater than 500 °C, the electronic layer thickness increases beyond 2 nm, while the improvement in sheet resistance and mobility has levelled out. For this reason, the RTA temperature maximum is considered optimum at ~500 °C. Despite the fact that both the mobility and sheet resistance values plateau at 1.6 ± 0.1 kΩ/□, and 28 ± 1 cm2/(V·s) respectively, the carrier density continues increasing beyond this optimum RTA temperature. At the highest RTA temperature maximum examined, 560 °C, we measure a carrier density of 1.45×1014 cm-2, which corresponds to an arsenic donor atom density of only 0.21 ML. Given our estimate of arsenic saturation coverage at 0.24 ML, this maximum value represents an activation of 88% of the donor atoms. At the optimized RTA temperature of 500 °C, this activation percentage is even lower at 81%. Nevertheless, this carrier concentration is well above the metal insulator transition, and greater than ½ of the typical carrier density of 2.4×1014 cm-2 found in phosphorus δ-layers. The fact that 100% of the donor atoms in the arsenic δ-layers is not surprising, as arsenic is known to suffer from clustering at high concentrations,59,60 resulting in incomplete activation at concentrations > 1020cm-3 (or 2×1013 cm-2) in traditional implantation or diffusion doping of silicon.51 Given the large parameter space involved in optimizing arsenic δ-layer growth for both confinement and electrical transport, there is potentially room for improvement in performance values beyond those presented here. However, comparing the figures of merit achieved for arsenic δ-layers with those typically achieved for phosphorus δ-layers,56,58 we find that they are at most a factor of 2 worse, and that arsine and arsenic are 100% compatible with atomic-scale STM device fabrication, and therefore represent a viable alternative to phosphine and phosphorous. CONCLUSIONS We have demonstrated the full compatibility of arsine as a precursor gas for the atomic-scale positioning of arsenic donors in silicon using STM-based hydrogen-desorption lithography, and, 15 when combined with low temperature silicon epitaxy, the capability to fabricate buried, atomic- scale, dopant structures in silicon. Room temperature dissociation of arsine on Si(001) differs from the dissociation of phosphine on Si(001) in that full dissociation of arsine occurs much more rapidly at room temperature. This has interesting implications, including the potential for tighter control of atomic-scale placement of arsenic compared to phosphorus. We find also that saturation arsine coverage is self-limiting at 24% ML, invariant with dose rate, and that monolayer hydrogen termination acts as an effective resist layer for the adsorption of arsine on Si(001). We show that arsenic can be incorporated in the silicon surface lattice by annealing at 350 °C. We demonstrate successful low temperature silicon epitaxial overgrowth by growing arsenic δ-layers, where segregation was suppressed through the use of an unheated, ultra-thin, locking layer. A rapid thermal anneal stage of 500 C × 10 sec immediately following locking layer growth improves electronic transport properties resulting in 2 nm thick, conductive layers with sheet resistance, and carrier concentration of 1.5 kΩ, and 1.3×1014 cm-3, respectively. It has been a decade and a half since phosphorus dopant placement using STM H-lithography was first demonstrated and the work presented here shows that using arsenic as the n-type dopant in place of phosphorus is not only possible, but has demonstrable advantages over phosphorus. Finally, the parameters used for arsenic and for phosphorus device fabrication are completely compatible with one another, demonstrating the possibility of multispecies dopant devices. EXPERIMENTAL METHODS Sample Preparation and Scanning Tunneling Microscopy: Si(001) samples were diced to 2×9 mm from a 0.3 mm thick, Czochralski grown, bulk phosphorus doped wafer with a resistivity of 1-10 Ω∙cm. Samples were cleaned ultrasonically ex situ in acetone followed by isopropyl alcohol, then thermally outgassed in situ (base pressure < 1×1010 mbar ) for >8 hours at 600 °C, and finally flash annealed multiple times at 1200 °C, using direct current resistive sample heating. Sample temperature was monitored using an infrared pyrometer (IMPAC IGA50-LO plus), with a total estimated measurement uncertainty of +/-30 °C. Silicon epitaxy was performed at a base pressure of 1×1010 mbar, using an all silicon, solid sublimation source (SUSI-40, MBE Komponenten GmbH) operated at a deposition rate of 1.5 Å/min. During silicon overgrowth, sample temperature was indirectly monitoring by measuring sample resistance, while the sample was heated using direct current resistive sample heating. All scanning tunneling microscopy measurements were performed in an Omicron variable temperature series STM at room temperature with a base pressure of <5×1011mbar. Density Functional Theory Calculations: All DFT calculations reported here were conducted using the Gaussian 09 software [48] and methods of energy computation and geometry optimization implemented therein. We calculate adsorption energies for various arsine configurations on the silicon (001) surface using an approach that we refer to as the cluster composite model (CCM).13,61 In this approach, the Si(001) surface is represented using hydrogen- truncated cluster models, such as Si15H16 or Si54H44, and various correction terms are applied to minimize the effects of finite cluster size. We note that the specific approach used here for the AsH3/Si(001) system is the same as in our previous work on the PH3/Si(001) system,13 and the reader is referred to this article for the full technical details. Reaction steps (i) in Fig. 2g was calculated using what is referred to as the 3D2R cluster model in Ref.13 while reaction steps ii), iii), and iv) used the 2D1R cluster model with the hemihydride dimer represented by a separate cluster (see Eq. 5 in Ref. 13). 16 Kinetic Monte Carlo Simulations. KMC methods are explained in detail in the supplementary material. Secondary Ion Mass Spectrometry: Time of flight (ToF) SIMS measurements were made using an IONTOF ToF.SIMS5 system with a 25 keV Bi+ primary ion beam in high current bunch mode (HCBM), and a 500 eV, 35 nA Cs+ sputter beam. Depth profiles were made with a 300 µm2 sputter crater, and the analytical region was gated to the central 50 m2 of the sputter region Hall bar Fabrication and Magnetotransport Measurements: Arsenic delta layer samples were fabricated into 6 terminal Hall bars using standard UV photolithography, with electrical contact made via aluminum thermally evaporated over the edge of the mesa structures. Magneto-transport measurements were made at a temperature of 5 K in a Cryogenics Ltd. cryogen free measurement system using magnetic fields up to 5 Tesla. Details of these measurements, and quantification of the delta layer electrical transport characteristics are explained in detail in the supplementary material. ASSOCIATED CONTENT Supporting Information: Kinetic Monte Carlo calculations and magneto-transport data analysis are each explained in detail in the supporting information document. AUTHOR INFORMATION The authors declare no competing financial interests. ACKNOWLEDGMENT This projected has been supported by the EPSRC project Atomically Deterministic Doping and Readout For Semiconductor Solotronics (grant number EP/M009564/1). P.C.C. and E.V.S.H. were partly supported by the EPSRC Centre for Doctoral Training in Advanced Characterisation of Materials (grant number EP/L015277/1), and also by Paul Scherrer Institute and IHP -- Leibniz- Institut für innovative Mikroelektronik, respectively. O.W. acknowledges the support of the Australian Research Council (ARC) Centre of Excellence for Quantum Computation and Communication Technology (project number CE110001027). Computing support was provided by the Australian National Computational Infrastructure (NCI). All data created during this research is openly available at zenodo.org (DOI: 10.5281/zenodo.3406575). 17 REFERENCES (1) Jehl, X.; Niquet, Y. M.; Sanquer, M. Single Donor Electronics and Quantum Functionalities with Advanced CMOS Technology. J. Phys. Condens. Matter 2016, 28 (10), 103001. https://doi.org/10.1088/0953-8984/28/10/103001. (2) Morton, J. J. L.; McCamey, D. R.; Eriksson, M. A.; Lyon, S. A. Embracing the Quantum Limit in Silicon Computing. Nature. Nature Publishing Group November 16, 2011, pp 345 -- 353. https://doi.org/10.1038/nature10681. (3) Schofield, S. R.; Curson, N. J.; Simmons, M. Y.; Ruess, F. J.; Hallam, T.; Oberbeck, L.; Clark, R. G. Atomically Precise Placement of Single Dopants in Si. Phys. Rev. Lett. 2003, 91 (13), 136104. https://doi.org/10.1103/PhysRevLett.91.136104. (4) Hudak, B. M.; Song, J.; Sims, H.; Troparevsky, M. C.; Humble, T. S.; Pantelides, S. T.; Snijders, P. C.; Lupini, A. R. Directed Atom-by-Atom Assembly of Dopants in Silicon. ACS Nano 2018, 12 (6), 5873 -- 5879. https://doi.org/10.1021/acsnano.8b02001. (5) Sinthiptharakoon, K.; Schofield, S. R.; Studer, P.; Brázdová, V.; Hirjibehedin, C. F.; Bowler, D. R.; Curson, N. J. Investigating Individual Arsenic Dopant Atoms in Silicon Using Low-Temperature Scanning Tunnelling Microscopy. J. Phys. Condens. Matter 2014, 26 (1), 012001. https://doi.org/10.1088/0953-8984/26/1/012001. (6) Salfi, J.; Mol, J. A.; Rahman, R.; Klimeck, G.; Simmons, M. Y.; Hollenberg, L. C. L.; Rogge, S. Spatially Resolving Valley Quantum Interference of a Donor in Silicon. Nat. Mater. 2014, 13 (6), 605 -- 610. https://doi.org/10.1038/nmat3941. (7) Voisin, B.; Salfi, J.; Bocquel, J.; Rahman, R.; Rogge, S. Spatially Resolved Resonant Tunneling on Single Atoms in Silicon. J. Phys. Condens. Matter 2015, 27 (15), 154203. https://doi.org/10.1088/0953-8984/27/15/154203. (8) Salfi, J.; Mol, J. A.; Rahman, R.; Klimeck, G.; Simmons, M. Y.; Hollenberg, L. C. L.; Rogge, S. Quantum Simulation of the Hubbard Model with Dopant Atoms in Silicon. Nat. Commun. 2016, 7 (1), 11342. https://doi.org/10.1038/ncomms11342. (9) Pierre, M.; Wacquez, R.; Jehl, X.; Sanquer, M.; Vinet, M.; Cueto, O. Single-Donor Ionization Energies in a Nanoscale CMOS Channel. Nat. Nanotechnol. 2010, 5 (2), 133 -- 137. https://doi.org/10.1038/nnano.2009.373. (10) Murdin, B. N.; Litvinenko, K.; Li, J.; Bowyer, E.; Pang, M.; Greenland, P. T.; Villis, B.; Aeppli, G.; van der Meer, A. F. G.; Redlich, B.; et al. Nano-Orbitronics in Silicon. In Springer Proceedings in Physics; Springer, Cham, 2015; Vol. 159, pp 92 -- 93. https://doi.org/10.1007/978-3-319-07743-7_30. (11) Li, J.; Le, N. H.; Litvinenko, K. L.; Clowes, S. K.; Engelkamp, H.; Pavlov, S. G.; Hübers, H. W.; Shuman, V. B.; Portsel, L. D.; Lodygin, D. N.; et al. Radii of Rydberg States of Isolated 085423. https://doi.org/10.1103/PhysRevB.98.085423. Silicon Donors. 2018, Phys. Rev. B 98 (8), 18 (12) Moraru, D.; Udhiarto, A.; Anwar, M.; Nowak, R.; Jablonski, R.; Hamid, E.; Tarido, J.; Mizuno, T.; Tabe, M. Atom Devices Based on Single Dopants in Silicon Nanostructures. Nanoscale Res. Lett. 2011, 6 (1), 479. https://doi.org/10.1186/1556-276X-6-479. (13) Warschkow, O.; Curson, N. J.; Schofield, S. R.; Marks, N. A.; Wilson, H. F.; Radny, M. W.; Smith, P. V.; Reusch, T. C. G. G.; McKenzie, D. R.; Simmons, M. Y. Reaction Paths of Phosphine Dissociation on Silicon (001). J. Chem. Phys. 2016, 144 (1), 014705. https://doi.org/10.1063/1.4939124. (14) Lyding, J. W.; Shen, T. C.; Hubacek, J. S.; Tucker, J. R.; Abeln, G. C. Nanoscale Patterning and Oxidation of H-Passivated Si(100)-2×1 Surfaces with an Ultrahigh Vacuum Scanning Tunneling Microscope. Appl. Phys. 2010 -- 2012. https://doi.org/10.1063/1.111722. 1994, (15), Lett. 64 (15) Koch, M.; Keizer, J. G.; Pakkiam, P.; Keith, D.; House, M. G.; Peretz, E.; Simmons, M. Y. Spin Read-out in Atomic Qubits in an All-Epitaxial Three-Dimensional Transistor. Nature Nanotechnology. Nature Publishing Group February 7, 2019, pp 137 -- 140. https://doi.org/10.1038/s41565-018-0338-1. (16) Fuechsle, M.; Miwa, J. A.; Mahapatra, S.; Ryu, H.; Lee, S.; Warschkow, O.; Hollenberg, L. C. L.; Klimeck, G.; Simmons, M. Y. A Single-Atom Transistor. Nat. Nanotechnol. 2012, 7 (4), 242 -- 246. https://doi.org/10.1038/nnano.2012.21. (17) Wyrick, J.; Wang, X.; Kashid, R. V.; Namboodiri, P.; Schmucker, S. W.; Hagmann, J. A.; Liu, K.; Stewart, M. D.; Richter, C. A.; Bryant, G. W.; et al. Atom‐by‐Atom Fabrication of Single and Few Dopant Quantum Devices. Adv. Funct. Mater. 2019, 1903475, 1903475. https://doi.org/10.1002/adfm.201903475. (18) Hile, S. J.; Fricke, L.; House, M. G.; Peretz, E.; Chen, C. Y.; Wang, Y.; Broome, M.; Gorman, S. K.; Keizer, J. G.; Rahman, R.; et al. Addressable Electron Spin Resonance Using Donors and Donor Molecules in Silicon. Sci. Adv. 2018, 4 (7), eaaq1459. https://doi.org/10.1126/sciadv.aaq1459. (19) He, Y.; Gorman, S. K.; Keith, D.; Kranz, L.; Keizer, J. G.; Simmons, M. Y. A Two-Qubit Gate between Phosphorus Donor Electrons in Silicon. Nature 2019, 571 (7765), 371 -- 375. https://doi.org/10.1038/s41586-019-1381-2. (20) Kane, B. E. A Silicon-Based Nuclear Spin Quantum Computer. Nature 1998, 393 (6681), 133 -- 137. https://doi.org/10.1038/30156. (21) Morello, A.; Escott, C. C.; Huebl, H.; Willems Van Beveren, L. H.; Hollenberg, L. C. L.; Jamieson, D. N.; Dzurak, A. S.; Clark, R. G. Architecture for High-Sensitivity Single-Shot Readout and Control of the Electron Spin of Individual Donors in Silicon. Phys. Rev. B - Phys. 081307. Condens. https://doi.org/10.1103/PhysRevB.80.081307. Matter Mater. 2009, (8), 80 (22) Hill, C. D.; Peretz, E.; Hile, S. J.; House, M. G.; Fuechsle, M.; Rogge, S.; Simmons, M. Y.; Hollenberg, L. C. L. Quantum Computing: A Surface Code Quantum Computer in Silicon. 19 Sci. Adv. 2015, 1 (9), 1 -- 12. https://doi.org/10.1126/sciadv.1500707. (23) Solmi, S.; Nobili, D. High Concentration Diffusivity and Clustering of Arsenic and 2484 -- 2490. 1998, Phys. Appl. Phosphorus J. https://doi.org/10.1063/1.367008. Silicon. in 83 (5), (24) Ramdas, A. K.; Rodriguez, S. Spectroscopy of the Solid-State Analogues of the Hydrogen Atom: Donors and Acceptors in Semiconductors. Reports Prog. Phys. 1981, 44 (12), 1297 -- 1387. https://doi.org/10.1088/0034-4885/44/12/002. (25) Slater, J. C. Atomic Radii in Crystals. J. Chem. Phys. 1964, 41 (10), 3199 -- 3204. https://doi.org/10.1063/1.1725697. (26) Salvail, J. Z.; Dluhy, P.; Morse, K. J.; Szech, M.; Saeedi, K.; Huber, J.; Riemann, H.; Abrosimov, N. V.; Becker, P.; Pohl, H. J.; et al. Optically Enabled Magnetic Resonance Study of As 75 and Sb 121 in Si 28. Phys. Rev. B - Condens. Matter Mater. Phys. 2015, 92 (19), 195203. https://doi.org/10.1103/PhysRevB.92.195203. (27) Franke, D. P.; Pflüger, M. P. D.; Itoh, K. M.; Brandt, M. S. Multiple-Quantum Transitions and Charge-Induced Decoherence of Donor Nuclear Spins in Silicon. Phys. Rev. Lett. 2017, 118 (24), 246401. https://doi.org/10.1103/PhysRevLett.118.246401. (28) Abanto, M.; Davidovich, L.; Koiller, B.; De Matos Filho, R. L. Quantum Computation with Doped Silicon Cavities. Phys. Rev. B - Condens. Matter Mater. Phys. 2010, 81 (8), 085325. https://doi.org/10.1103/PhysRevB.81.085325. (29) Waldherr, G.; Wang, Y.; Zaiser, S.; Jamali, M.; Schulte-Herbrüggen, T.; Abe, H.; Ohshima, T.; Isoya, J.; Du, J. F.; Neumann, P.; et al. Quantum Error Correction in a Solid-State Hybrid Spin Register. Nature 2014, 506 (7487), 204 -- 207. https://doi.org/10.1038/nature12919. (30) Stoneham, A. M.; Fisher, A. J.; Greenland, P. T. Optically Driven Silicon-Based Quantum Gates with Potential for High-Temperature Operation. J. Phys. Condens. Matter 2003, 15 (27), 447 -- 451. https://doi.org/10.1088/0953-8984/15/27/102. (31) Kipp, L.; Bringans, R. D.; Biegelsen, D. K.; Swartz, L. E.; Hicks, R. F. Arsine Adsorption on Si(100) 2×1: A Photoemission and Scanning-Tunneling-Microscopy Study. Phys. Rev. B 1994, 50 (8), 5448 -- 5455. https://doi.org/10.1103/PhysRevB.50.5448. (32) Tok, E. S.; Hartell, A. D.; Zhang, J. Kinetics of Si Growth from Hydride Precursors on As- 919 -- 921. Passivated Si(001) Surface. Appl. Phys. Lett. 2001, https://doi.org/10.1063/1.1347396. (7), 78 (33) Soares, J. A. N. T.; Kim, H.; Glass, G.; Desjardins, P.; Greene, J. E. Arsenic-Doped Si(001) Gas-Source Molecular-Beam Epitaxy: Growth Kinetics and Transport Properties. Appl. Phys. Lett. 1999, 74 (9), 1290 -- 1292. https://doi.org/10.1063/1.123527. (34) Cho, J.; Ghosh, M. K.; Choi, C. H. Adsorption Reactions of Trimethylgallium and Arsine on H/Si(100)-2x1 Surface. Bull. Korean Chem. Soc. 2009, 30 (8), 1805 -- 1810. 20 https://doi.org/10.5012/bkcs.2009.30.8.1805. (35) Kim, H.; Glass, G.; Soares, J. A. N. T.; Desjardins, P.; Greene, J. E. Arsenic Incorporation during Si(001):As Gas-Source Molecular-Beam Epitaxy from Si2H6and AsH3: Effects on Film-Growth Kinetics. 7067 -- 7078. https://doi.org/10.1063/1.1324701. 2000, Phys. Appl. (12), 88 J. (36) Kim, H.; Glass, G.; Soares, J. A. N. T.; Foo, Y. L.; Desjardins, P.; Greene, J. E. Temperature-Modulated Si(001):As Gas-Source Molecular Beam Epitaxy: Growth Kinetics and As Incorporation. Appl. Phys. Lett. 2001, 79 (20), 3263 -- 3265. https://doi.org/10.1063/1.1415420. (37) Oberbeck, L.; Curson, N. J.; Hallam, T.; Simmons, M. Y.; Bilger, G.; Clark, R. G. Measurement of Phosphorus Segregation in Silicon at the Atomic Scale Using Scanning Tunneling Microscopy. Appl. Phys. Lett. 2004. https://doi.org/10.1063/1.1784881. (38) Oberbeck, L.; Curson, N. J.; Simmons, M. Y.; Brenner, R.; Hamilton, A. R.; Schofield, S. R.; Clark, R. G. Encapsulation of Phosphorus Dopants in Silicon for the Fabrication of a Quantum Computer. 3197 -- 3199. https://doi.org/10.1063/1.1516859. 2002, Phys. Appl. (17), Lett. 81 (39) Northrup, J. E. Theoretical Studies of Arsine Adsorption on Si(100). Phys. Rev. B 1995, 51 (4), 2218 -- 2222. https://doi.org/10.1103/PhysRevB.51.2218. (40) Miotto, R.; Srivastava, G. P.; Miwa, R. H.; Ferraz, A. C. A Comparative Study of Dissociative Adsorption of NH3, PH3, and AsH3 on Si(001)-(2×1). J. Chem. Phys. 2001, 114 (21), 9549 -- 9556. https://doi.org/10.1063/1.1355766. (41) McDonell, T. L.; Marks, N. A.; Warschkow, O.; Wilson, H. F.; Smith, P. V.; Radny, M. W. Molecular Dissociation of Group-V Hydrides on Si(001). Phys. Rev. B - Condens. Matter Mater. Phys. 2005, 72 (19), 193307. https://doi.org/10.1103/PhysRevB.72.193307. (42) Fuhrer, A.; Köster, S.; Pascher, N. Towards Bipolar Atomic Scale Dopant Devices Defined by STM-Lithography. In Bulletin of the American Physical Society: APS March Meeting; American 60. https://doi.org/http://meetings.aps.org/link/BAPS.2015.MAR.L14.7. San Antonio, 2015; Vol. Physical Society: Texas, (43) Hamers, R. J.; Köhler, U. K. Determination of the Local Electronic Structure of Atomic‐ sized Defects on Si(001) by Tunneling Spectroscopy. J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 1989, 7 (4), 2854 -- 2859. https://doi.org/10.1116/1.576158. (44) Warschkow, O.; Schofield, S.; Marks, N.; Radny, M.; Smith, P.; McKenzie, D. Water on Silicon (001): C Defects and Initial Steps of Surface Oxidation. Phys. Rev. B 2008, 77 (20), 201305. https://doi.org/10.1103/PhysRevB.77.201305. (45) Wilson, H. F.; Warschkow, O.; Marks, N. A.; Schofield, S. R.; Curson, N. J.; Smith, P. V.; Radny, M. W.; McKenzie, D. R.; Simmons, M. Y. Phosphine Dissociation on the Si(001) Surface. 226102. 2004, Phys. (22), Lett. Rev. 93 21 https://doi.org/10.1103/PhysRevLett.93.226102. (46) Curson, N. J.; Schofield, S. R.; Simmons, M. Y.; Oberbeck, L.; O'Brien, J. L.; Clark, R. G. STM Characterization of the Si-P Heterodimer. Phys. Rev. B - Condens. Matter Mater. Phys. 2004, 69 (19), 195303. https://doi.org/10.1103/PhysRevB.69.195303. (47) McKibbin, S. R.; Clarke, W. R.; Fuhrer, A.; Reusch, T. C. G.; Simmons, M. Y. Investigating the Regrowth Surface of Si:P δ -Layers toward Vertically Stacked Three Dimensional Devices. Appl. Phys. Lett. 2009, 95 (23), 233111. https://doi.org/10.1063/1.3269924. (48) Schofield, S. R.; Curson, N. J.; Warschkow, O.; Marks, N. A.; Wilson, H. F.; Simmons, M. Y.; Smith, P. V.; Radny, M. W.; McKenzie, D. R.; Clark, R. G. Phosphine Dissociation and Diffusion on Si(001) Observed at the Atomic Scale. J. Phys. Chem. B 2006, 110 (7), 3173 -- 3179. https://doi.org/10.1021/jp054646v. (49) Wilson, H. F.; Warschkow, O.; Marks, N. A.; Curson, N. J.; Schofield, S. R.; Reusch, T. C. G.; Radny, M. W.; Smith, P. V; McKenzie, D. R.; Simmons, M. Y. Thermal Dissociation and Desorption of P H3 on Si(001): A Reinterpretation of Spectroscopic Data. Phys. Rev. B 656 -- 661. - https://doi.org/10.1103/PhysRevB.74.195310. Condens. Matter Mater. 2006, Phys. (19), 74 (50) Lin, D. S.; Ku, T. S.; Sheu, T. J. Thermal Reactions of Phosphine with Si(100): A Combined Photoemission and Scanning-Tunneling-Microscopy Study. Surf. Sci. 1999, 424 (1), 7 -- 18. https://doi.org/10.1016/S0039-6028(98)00943-1. (51) Liu, X.; Tang, Q.; Kamins, T. I.; Harris, J. S. Heavy Arsenic Doping of Silicon by Molecular 651 -- 656. Growth 2003, 251, Beam https://doi.org/10.1109/MBE.2002.1037754. Epitaxy. Cryst. J. (52) Bashir, R.; Kabir, A. E.; Westrom, P. Phosphorus and Arsenic Profile Control for High Performance Epitaxial Base Bipolar Junction Devices. Appl. Phys. Lett. 1999, 75 (6), 796 -- 798. https://doi.org/10.1063/1.124516. (53) Barnett, S. A.; Greene, J. E. Si Molecular Beam Epitaxy: A Model for Temperature Dependent Incorporation Probabilities and Depth Distributions of Dopants Exhibiting Strong Surface Segregation. Surf. Sci. 1985, 151 (1), 67 -- 90. https://doi.org/10.1016/0039- 6028(85)90455-8. (54) Liu, X.; Tang, Q.; Harris, J. S.; Kamins, T. I. Arsenic Surface Segregation during in Situ Doped Silicon and Si 1-x Ge x Molecular Beam Epitaxy. J. Cryst. Growth 2005, 281 (2 -- 4), 334 -- 343. https://doi.org/10.1016/j.jcrysgro.2005.04.066. (55) Eaglesham, D. J. Semiconductor Molecular-Beam Epitaxy at Low Temperatures. J. Appl. Phys. 1995, 77 (8), 3597 -- 3617. https://doi.org/10.1063/1.358597. (56) Keizer, J. G.; Koelling, S.; Koenraad, P. M.; Simmons, M. Y. Suppressing Segregation in Highly Phosphorus Doped Silicon Monolayers. ACS Nano 2015, 9 (12), 12537 -- 12541. https://doi.org/10.1021/acsnano.5b06299. 22 (57) Wang, X.; Hagmann, J. A.; Namboodiri, P.; Wyrick, J.; Li, K.; Murray, R. E.; Myers, A.; Misenkosen, F.; Stewart, M. D.; Richter, C. A.; et al. Quantifying Atom-Scale Dopant Movement and Electrical Activation in Si:P Monolayers. Nanoscale 2018, 10 (9), 4488 -- 4499. https://doi.org/10.1039/c7nr07777g. (58) Hagmann, J. A.; Wang, X.; Namboodiri, P.; Wyrick, J.; Murray, R.; Stewart, M. D.; Silver, R. M.; Richter, C. A. High Resolution Thickness Measurements of Ultrathin Si:P Monolayers Using Weak Localization. Appl. Phys. Lett. 2018, 112 (4). https://doi.org/10.1063/1.4998712. (59) Berding, M. A.; Sher, A.; Van Schilfgaarde, M.; Rousseau, P. M.; Spicer, W. E. Deactivation in Heavily Arsenic-Doped Silicon. Appl. Phys. Lett. 1998, 72 (12), 1492 -- 1494. https://doi.org/10.1063/1.121036. (60) Duguay, S.; Vurpillot, F.; Philippe, T.; Cadel, E.; Lard, R.; Deconihout, B.; Servanton, G.; Pantel, R. Evidence of Atomic-Scale Arsenic Clustering in Highly Doped Silicon. J. Appl. Phys. 2009, 106 (10), 106102. https://doi.org/10.1063/1.3257178. (61) Warschkow, O.; Gao, I.; Schofield, S. R.; Belcher, D. R.; Radny, M. W.; Saraireh, S. A.; Smith, P. V. Acetone on Silicon (001): Ambiphilic Molecule Meets Ambiphilic Surface. Phys. Chem. Chem. Phys. 2009, 11 (15), 2747 -- 2759. https://doi.org/10.1039/b815542a. 23 Supporting Information: a) Kinetic Monte Carlo Simulations Computer simulations of the evolution of an AsH3-exposed Si(001) surface on time and length scales relevant to the STM experiments were conducted using the kinetic Monte Carlo (KMC) method1 and a discrete lattice representation of the adsorbed molecular species. Key elements of this model are illustrated in Supplementary Figure S1. The lattice used to represent the adsorbed species is shown in Figure S1 (a) with a unit cell (gray background shading) that corresponds to the (2x1) reconstruction of the Si(001) surface, and much larger super cells used in our simulations to represent an extended surface. Square boxes in Figure S1 (a) indicate discrete lattice sites that represent the surface binding sites for an AsH3 molecule or any of its fragments. These sites come in three types, namely, dimer-end, end-bridge, and dimer- bridge, labeled 'a', 'b', and 'c' within the unit cell in Figure S1(a), respectively. Each of the sites exists in one of seven states, which indicates the type of molecular species occupying the site. The states used here are defined in Figure S1(b) and they represent AsH3, AsH2, AsH, As, and H molecular species, as well as a state F indicating a free or unoccupied site. Further defined is a distinct state X, which serves as a blocking state to indicate that a site is unavailable for occupation by another species. This is used in our model to block dimer-end sites when an adjacent dimer- bridge or end-bridge site is occupied, as such a bridging adsorbate would utilize the dangling bond valences at the two dimer ends involved in the bridge. Together, the lattice and the configuration of states at all sites provides a compact representation of an extended Si(001) surface with a coverage of AsH3 molecules in varying degrees of dissociation. An example configuration is shown in Fig. S1(c), describing a low-coverage surface with two AsH3-related species, namely, a molecular AsH3 molecular adsorbate at a dimer-end position near the top of the figure and a fully dissociated end-bridge As+3H species near the bottom. In a KMC simulation, a given initial configuration of states is advanced in time using a set of discrete transformations that occur with a defined rate1. In our case, the initial configuration is the bare surface without any adsorbates (i.e. all sites in the F state) and the set of transformations corresponds to a set of discrete chemical reactions that may occur upon AsH3 exposure, broadly divided into adsorption, dissociation, and rearrangement reactions. These transformations are set out in Figure S1(d) to (o), and they each comprise a pattern of specific states at a subset of lattice sites, a prescription for how these states change when the reaction occurs, and a parameter to define the rate of reaction. Note that an empty box in a transformation pattern indicates that the site can be in any state for the pattern to match. 24 Figure S1: KMC model of the Si(001) surface exposed to AsH3. (a) The lattice representation of binding sites (indicated by squares) on the Si(001) surface with the unit cell shaded gray. (b) The set of seven possible occupation states at each binding site. (c) An example configuration in a 3x3 supercell featuring a molecular AsH3 adsorbate and the end-bridge As+3H structure observed as the type-1 STM feature. (d,e) KMC transformations representing two types of chemisorption processes with the rate of transformation given by the pressure-dependent surface impingement rate. (f-o) KMC transformation representing various hydrogen-shift and rearrangement reactions relevant to AsH3 dissociation on Si(001) with the rate of reaction defined using Arrhenius activation energies. 25 For all other transformations [Figure S1 (f) to (o)] the rates are defined by an activation energy, which is converted into a rate constant using the Arrhenius equation with an assumed attempt frequency of 1013 s-1 and a temperature of 298 K. The activation energies for most of the transformations are derived from the density functional theory calculations of suitable prototype reactions that represent the transformation. Here, these prototype reactions are those calculated for the main AsH3 dissociation pathway (cf. main text Fig. 1g) and some of the competing reaction pathways discussed in the main text. For other transformation which are indicated by an asterisk ('*') in Fig. S1, the activation energies are derived by analogy to a similar reaction within the set. For example, the activation energy of the dimer-bridge AsHAs+H transformation [Fig. S1(n)] assumed to be the same as that calculated by DFT for the end-bridge AsHAs+H transformation [Fig. S1(n)]. One of the key tasks of a KMC simulation is to assemble a list of all possible transformations for a given configuration. This is accomplished by identifying all those transformation patterns that match the given configuration. Using again the example configuration in Fig. S1(c), the only transformation patterns that apply are the two types of adsorption reactions [Fig. S1(d) and (e)] matching many F-state sites in this configuration as well as the dissociation reactions [Fig. S1(f) and (g)] for the molecular AsH3 adsorbate near the top of Fig. S1(c). None of the transformation patterns match to the end-bridge As+3H structure near the bottom of the figure, which reflects the fact that this structure represents an end point in the dissociation pathway of a single AsH3 molecule. From the assembled list of all applicable transformations, a single transformation is selected at random and executed to advance the system to the next configuration, and the time is advanced also. A new list of applicable transformations is then assembled for the new configuration and the process repeats until the desired simulation time has been reached, or the system has reached a stable point where none of the transformations apply. Critically, both the random selection of the transformation and the time advancement are performed such that the all transformations occur at the specified rate when examined using in a sufficiently large ensemble average (see (Gillespie, 1976) for details). The sequence of configurations that result from successive transformations applied in the course of a KMC simulation create a single trajectory in which AsH3 molecules adsorb and dissociate in a stochastic manner. The pressure-dependent species coverages reported in the main text Fig. 2c using large 40x40 super cells and an average over 100 trajectories for each pressure, which sufficiently removes stochastic effects from the data. All KMC simulations were carried out using a software developed by one of the authors (O.W.). 26 b) Hall and Magnetoresistance Measurements of As -layers Eight-terminal Hall bars were fabricated from multiple Si:As delta layer samples and electrically characterized by IV, Hall, and magnetoresistance measurements. The Hall bar geometry is illustrated below in Figure S22, alongside an example of a two-terminal IV measurement (terminals 8-4, sample d) demonstrating Ohmic contacting to the As delta layer. Figure S2: Eight-terminal Hall bar geometry and two-terminal IV measurement on a Si:As delta layer. a) The orientation of the perpendicular magnetic field B⊥, and the measurement terminals for current and voltages Ixx, Vxx, and Vxy are indicated on the Hall bar schematic. b) A two-terminal IV measurement (across terminals 8 and 4) on sample d (RTA = 500 °C) measured at 5 K, is both linear and symmetric indicating that the delta layer contacting is Ohmic. Figure S3 shows Hall and magnetoresistance measurements for all six As delta-layer Hall bar samples along with fits to the data. The approach used in fitting to the weak-localisation data follows the methods of Sullivan et al.,2 and assumes that the signal depends on both the magnitude and direction of the applied magnetic field, , where the change in the conductance, , is for a weak spin-orbit donor. From the Hall and magnetoresistance measurements, Δ is obtained from the measured resistivity : Δ()= 1()− 1(0) . ℏ4−Ψ12+ where Ψ is the digamma function, the dephasing length and the mean free path. For a parallel (∥)= 2ℏln1+∥ , where is the fitting parameter. For a perpendicular magnetic field, the change in conductance is: ()= 2ℏΨ12+ magnetic field, the change in conductance is: ℏ2+ln2 , (1) (2) (3) 27 (4) Once the fits to the experimental weak-localisation data are determined, the fit parameters can be used to obtain an approximate measure of the -layer thickness: =14/ℏ/ , where is the correlation length, assumed to be of the order of the mean donor spacing = 1/√, where is the carrier concentration extracted from the experimental Hall-measurements shown in the insets of Figure S3. Figure S3: Summary of weak-localization data for parallel- (red) and perpendicular- (blue) magnetic fields. (a)-(d) The primary figures show the weak-localization data, whereas the insets show the Hall-measurements, with progressively increasing rapid thermal anneal (RTA) temperatures during the silicon overgrowth. The fits are applied from the equations provided in (Sullivan et al., 2004). No fit could be made to the parallel field data in (c). Table S1 summarises values extracted from the Hall measurements (n, µ, and R0), and the best fit parameters for the weak-localization magnetoresistance plots for each of the arsenic -layers grown using different rapid-thermal-anneal (RTA) temperatures. Using equations (1)-(3), a least- squares method is used to fit to the experimental weak-localisation data and the uncertainties in the fit parameters reflect the confidence intervals dictated by the fit parameters. For samples (a)- (b) and (d)-(e), there is a small feature present in the magnetoresistance near zero-field regime, 28 Sample RTA (°C) (a) 250±15 (b) 384±15 (c) 438±15 (d) 498±15 (e) 562±15 Electron density (n) (10/cm ) 1.15±0.02 1.19±0.07 1.29±0.02 1.32±0.01 1.45±0.10 Sheet Mobility (cm/V∙s) Mean free Resistance () path (L) () (nm) (10 Ω / □ ) 4.18±0.02 13.0±0.2 7.6±0.2 2.49±0.04 20.0±1.3 8.1±0.7 2.41±0.02 20.0±0.3 − 1.69±0.02 28.1±0.2 7.1±0.3 1.54±0.06 28.0±2.0 16.0±0.7 which yields a negative change in conductance. Establishing the source of this behaviour in the Si:As system is beyond the scope of the current work and is the focus of ongoing investigations. For the purposes of the present work, all fits to extract thickness estimates were made by omitting the near zero-field behaviour, which is reflected in the uncertainties of the fit parameters shown in Table S1. For sample (c), no fits could be made to extract thickness estimates to the parallel field as the change in conductance is always negative. Table S1: As delta-layer electrical transport characteristics extracted from the magnetoresistance data shown in Figure S3. The electron density (n) and mobility (μ) are determined from the fits to the inset data in Figure S3. The sheet resistance (R_0), mean- free path (L), dephasing length (Lϕ) and parallel field parameter (γ) are determined from the experimental fits to the primary data shown in Figure S3. The mean thickness (T) is then determined as in (Sullivan et al, 2004). Supporting Information References (1) Gillespie, D.T. A General Method for Numerically Simulating the Stochastic Time Evolution Dephasing length () (nm) 30.9±0.1 58.6±0.6 − 52.2±0.2 55.3±0.8 Fitting Parameter () (10 T-2) 6.5±0.4 27.0±3.5 − 75.0±1.0 67.0±7.3 Mean Thickness () (nm) 0.8±0.1 0.9±0.1 − 1.7±0.1 2.3±0.2 (2) Sullivan, D.F.; Kane, B.E.; Thompson, P.E. Weak localization thickness measurements of Si:P delta-layers, Appl. Phys. Lett. 2004, 85, 6362. 29 of Coupled Chemical Reactions, J. Comput. Phys. 1976, 22, 403.
1904.01103	1	1904	2019-02-22T21:24:53	Charge transport through redox active [H7P8W48O184]33- polyoxometalates self-assembled onto gold surfaces and gold nanodots	[ "physics.app-ph", "cond-mat.mes-hall" ]	Polyoxometalates (POMs) are redox-active molecular oxides, which attract growing interest for their integration into nano-devices, such as high-density data storage non-volatile memories. In this work, we investigated the electrostatic deposition of the negatively charged [H7P8W48O184]33- POM onto positively charged 8-amino-1-octanethiol self-assembled monolayers (SAMs) preformed onto gold substrates or onto an array of gold nanodots. The ring-shaped [H7P8W48O184]33- POM was selected as an example of large POMs with high charge storage capacity. To avoid the formation of POM aggregates onto the substrates, which would introduce variability in the local electrical properties, special attention has to be paid to the preformed SAM seeding layer, which should itself be deprived of aggregates. Where necessary, rinsing steps were found to be crucial to eliminate these aggregates and to provide uniformly covered substrates for subsequent POM deposition and electrical characterizations. This especially holds for commercially available gold/glass substrates while these rinsing steps were not essential in the case of template stripped gold of very low roughness. Charge transport through the related molecular junctions and nanodot molecule junctions (NMJs) has been probed by conducting-AFM. We analyzed the current-voltage curves with different models: electron tunneling though the SAMs (Simmons model), transition voltage spectroscopy (TVS) method or molecular single energy level mediated transport (Landauer equation) and we discussed the energetics of the molecular junctions. We concluded to an energy level alignment of the alkyl spacer and POM lowest occupied molecular orbitals (LUMOs), probably due to dipolar effects.	physics.app-ph	physics	Charge transport through redox active [H7P8W48O184]33- polyoxometalates self-assembled onto gold surfaces and gold nanodots. K. Dalla Francesca,1 S. Lenfant,2 M. Laurans,1 F. Volatron,1 G. Izzet,1 V. Humblot,3 C. Methivier,3 D. Guerin,2 A. Proust,1,* D. Vuillaume2,* 1) Sorbonne Université, CNRS, Institut Parisien de Chimie Moléculaire, IPCM, 4 Place Jussieu, F-75005 Paris, France. 2) Institute for Electronics Microelectronics and Nanotechnology (IEMN), CNRS, Lille Univ., Av. Poincaré, F-59652 Villeneuve d'Ascq, France. 3) Sorbonne Université, CNRS, Laboratoire de Réactivité de Surface, LRS, 4 Place Jussieu, F-75005 Paris, France. * Corresponding authors: [email protected], [email protected] Abstract. Polyoxometalates (POMs) are redox-active molecular oxides, which attract growing interest for their integration into nano-devices, such as high-density data storage non-volatile memories. In this work, we investigated the electrostatic deposition of the negatively charged [H7P8W48O184]33- POM onto positively charged 8-amino-1-octanethiol self-assembled monolayers (SAMs) preformed onto gold substrates or onto an array of gold nanodots. The ring-shaped [H7P8W48O184]33- POM was selected as an example of large POMs with high charge storage capacity. To avoid the formation of POM aggregates onto the substrates, which would introduce variability in the local electrical properties, special attention has to be paid to the preformed SAM seeding layer, which should itself be deprived of aggregates. Where necessary, rinsing steps were found to be crucial to eliminate these aggregates and to provide uniformly covered substrates for subsequent POM deposition and electrical characterizations. This especially holds for commercially available gold/glass substrates while these rinsing steps were not essential in the case of template stripped gold of very low roughness. Charge transport through the related molecular junctions and nanodot molecule junctions (NMJs) has been probed by conducting-AFM. We analyzed the current-voltage curves with different models: electron tunneling though the SAMs (Simmons model), transition voltage spectroscopy (TVS) method or molecular single energy level mediated transport (Landauer equation) and we discussed the energetics of the molecular junctions. We concluded to an energy level alignment of the alkyl spacer and POM lowest occupied molecular orbitals (LUMOs), probably due to dipolar effects. 1. Introduction. Redox active molecules, which can be switched from one redox state to another, have been considered as promising for the development of molecular memory devices.1 2-4 5 6 7 Often viewed as the missing link between conventional molecules and extended oxides, which are ubiquitous in electronics, polyoxometalates (POMs) should thus draw more attention. These molecular oxides obeying to the general formula [XxMpOy]n- (X= P, Si …; M = MoVI, WVI, VV…) display an outstanding chemical versatility and remarkable redox properties.8-13 There are still very few reports on single POM devices.14-15 Charge transport through large POM-based molecular junctions prepared by layer- by-layer deposition or dip-coating on various electrodes has been studied16 and a metal- insulator/POM-semiconductor capacitor cell has been reported.17 As a major milestone, a flash-type memory cell comprising POMs drop-cast at the Si-nanowire channel has been recently described.18 However, in all the above POM-based devices the molecular organization and the POM packing density are not controlled and this was found to confine the performances. Therefore special attention should be paid to the processing of POMs to get uniformly structured thus reproducible layered materials and to reduce device-to-device variability of the ultimate electrical properties which is a main issue in nanoelectronics. Yet POM processing is still a sticking point. Direct deposition of POMs onto electrodes has been probed by microscopy imaging19-22 23-25 and largely exploited for electrocatalysis,26-28 sensing purposes,29 solar-energy conversion30 and in composite materials for molecular batteries or supercapacitors.31 32-34 35-37 Drop-casting18, 38-39 and spin-coating of POM solutions are easy to carry out and have supplied, among others, interfacial layers for solar cells or light-emitting devices.40-41 42 43-45 In those cases no special order in the POM arrangement was sought. As POMs are negatively charged species their immobilization onto electrodes has also commonly relied on their entrapment into positively charged polymers46-48 or on the dip-coating exchange of their counter cations: by positively charged electrolyte through the Layer-by-Layer method to build photo- or electro-chromic materials,49- 50 photo-electrodes,51 interfacial layers for solar cells52-54 or multilayer films for electrocatalysis;55-57 by amphiphilic cations to form Langmuir Blodgett films;58-62 by positively charged groups directly decorating the electrode,63 for application in water splitting,64-68 molecular cluster batteries69 or molecular electronics.17, 70-71 Finally but to a lesser extent, covalent grafting of POMs onto electrodes, directly or in two steps, has been investigated by some of us and others.72-76 This allowed us to describe the electron transfer kinetics from the electrode to a compact POM monolayer, assembled onto carbonaceous materials,76-78 gold 79-80 or silicon wafers81 and electron transport through silicon- POMs-metal junctions,82 as a first step towards integration into nanoelectronic devices. Although a covalent route provides a better control of the POMs/electrode interface, it is highly demanding in terms of synthetic efforts since it requires prior POM functionalization to obtain 2 organic-inorganic POM hybrids with suitable remote anchor. For this reason, it has been restricted to the immobilization of POMs of the Anderson-, Lindqvist- and Keggin-types, since functionalization of larger POMs remains to be rationalized. On the other hand, an electrostatic route is easier to implement and allows drawing into the whole POM library, including larger POMs with higher charge storage capacity, for a rapid benchmarking. In this contribution, we thus investigated the immobilization of [H7P8W48O184]33- by dip-coating onto 8-amino-1-octanethiol self-assembled monolayers (SAMs) preformed onto Au-substrates or onto an array of Au-nanodots and we have characterized the charge transport through the resulting molecular junctions by C-AFM. Our choice of the [H7P8W48O184]33- POM is motivated by its high stability and also by its remarkable redox properties since, in solution and at an appropriate pH, its electrochemistry proceeds through up to three successive, 8-electron reduction waves.83 2. Deposition of [H7P8W48O184]33- onto 8-amino-1-octanethiol SAMs. The immobilization of negatively charged POMs onto electrodes coated with SAMs terminated by positively charged groups is easy to implement but getting a uniformly structured layer not guaranteed. Deposition of H3[PW12O40] onto a Si/SiO2 substrate covered by APTES by dip-coating resulted in the formation of nano-islands of POM aggregates, with a mean diameter of 17 nm and a height of 5-14 nm.17 Hydrosilylation of undecylenic acid on hydrogenated silicon, followed by coupling to N,N-dimethylethylenediamine supplied amino-functionalized SAMs for subsequent deposition of H4[SiW12O40]: mean surface coverage densities deduced from XPS experiments suggested the formation of monolayers but no details were given about the POM organization in the layers.71 The electrochemically-driven deposition of H4[SiW12O40] on glassy-carbon electrode or HOPG modified by a 4-aminophenyl SAM gave close-packed ordered arrays according to STM images.84 These previously published results point out that special attention should be paid to the experimental conditions for the solution processing of POMs. The POM concentration in the dipping solution and the incubation time are obviously key parameters but we have also particularly taken care of the quality of the seeding SAM. These parameters were probed by ellipsometry, PM-RAIRS and AFM for POMs deposited onto Au/Si (ellipsometry, PM-RAIRS) or Au/glass (AFM) respectively. Once the best conditions have been determined on commercially available substrates, POM layers have been prepared onto template-stripped Au surfaces (TSAu)85 and Au nanodots86-87 for electrical characterization. 3 ]33- Au Scheme 1. Polyhedral representation [H7P8W48O184]33-(color code: WO6 octahedra blue, PO4 tetrahedra green) and schematization (not at scale) of its electrostatic deposition onto a 8-amino-1- octanethiol (AOT) SAM. The POM has a wheel-shaped structure with a diameter of ~2.0 nm and a thickness of ~1.0 nm. Preparation of the seeding 8-amino-1-octanethiol SAM. The 8-amino-1-octanethiol (AOT) was chosen with an alkyl chain long enough to ensure a good organization of the SAM on gold but not too long to limit the decrease of the tunneling current through the layer.88-91 The Au substrates were incubated in a 10-4 M solution of AOT hydrochloride in absolute ethanol for 24 h in the dark. Subsequent rinsing steps were found to be crucial to eliminate aggregates and to provide uniformly covered substrates for subsequent POM deposition. According to AFM monitoring (figure 1), our best conditions consisted in a five-step procedure: immersion in ethanol for 5 minutes, followed by ultra-sonication in ethanol for 5 minutes; immersion in 0.01 M Phosphate Buffer Saline (PBS pH = 7.4) for 3h, followed by ultra-sonication in distilled water for 5 minutes. The substrates were finally washed with ethanol and dried under nitrogen. The numerous ca. 2 nm sized aggregates observed on the AFM image (see Experimental Section) before PBS treatment (figure 1-a) disappeared after the 3h immersion in PBS (figure 1-b). 4 Figure 1. AFM images, zoom of the images (dashed line areas) and histograms of heights (from top to bottom) of the AOT SAMs on Au/glass substrates: (a) before and (b) after PBS immersion and (c) with the POM layer. AFM images (1x1 µm2) and histograms of heights of the AOT SAMs on TSAu substrates (d) without PBS rinsing and (e) with the POM layer. The histograms of heights clearly show the improvement of the surface topography for SAMs on the TSAu electrodes, with a less 5 dispersed, single peak, distribution (d and e). The root mean square (rms) roughnesses, calculated from these height histograms, are (b) 1.27 nm, (c) 1.66 nm on the zoom images (d ) 0.6 nm and (e) 0.5 nm. The AOT SAMs were also characterized by PM-RAIRS (see Experimental Section). The spectra (figure 2) displayed the characteristic bands corresponding to the vibrations of the alkyl chains (around 2920 and 2850 cm-1 for the asymmetric and symmetric νC-H stretching modes and a broad band around 1400 cm-1 for the δC-H cissor vibration) together with the bands assigned to the δN-H deformations at about 1650 and 1550 cm-1 for the unprotonated and protonated amino groups, respectively (see Figure 2). 2 1 9 2 6 4 8 2 7 3 4 1 5 5 5 1 1 5 6 1 1 3 1 1 0 3 2 1 2 3 9 8 4 6 21 4 5 1 8 6 2 1 2 3 2 1 4 7 3 1 1 2 4 1 1800 2000 Wavenumber (cm-1) 1600 1400 1200 1000 3000 2800 2000 1800 1600 1400 1200 1000 Wavenumber (cm-1) 6 2 9 2 6 5 8 2 3000 2800 Figure 2. PM-RAIRS spectra of the AOT functionalized Au substrate before (left) and after (right) POM deposition. POM immobilization. The AOT functionalized Au substrates were then incubated in solutions of K28Li5[H7P8W48O184] in LiCl 2M and for various times. The solubility of K28Li5[H7P8W48O184] in water is poor but increased in the presence of LiCl. The deposition of the POM layer was monitored by ellipsometry, PM-RAIRS and AFM to select the optimal conditions, which were found to correspond to an incubation time of around 1 hour for the AOT functionalized gold substrate into a 10-3 M POM solution, followed by washing with distilled water and ethanol and drying under nitrogen. Longer immersion times (until 24 h) led to exactly the same layer features and significantly more concentrated solutions were not possible to obtain because of the low solubility of the K28Li5[H7P8W48O184] in water, even at higher LiCl concentrations (see supporting information). Besides the peaks attributed above to AOT SAM, the PM-RAIRS spectra now displayed the fingerprint of the POM at 1131 cm-1, assigned to the νP-O vibrations (Figure 2).92 The vibration at 932 6 cm-1 could tentatively by assigned to the terminal WO bonds. However it lies at the border of the optical window of the experimental set-up of the PM-RAIRS instrument (KBr and ZnSe optics and MCT detector). The mean thickness of the layer was assessed by ellipsometry (see Experimental Section) with several measurements on each substrate to check for the homogeneity: the thickness was found to increase from 1.2 nm for the AOT SAM to 2 nm after the deposition of the POMs. This 0.8 nm increase was obtained in a very reproducible way, whatever the time of immersion. Moreover, the AFM image of the POM layer showed a homogeneous film (figure 1-c). This proves that the POMs were uniformly deposited. Theoretically, a 2.0 nm increase is expected if the POMs have a vertical orientation and a sub-1 nm increase is expected in the case of a horizontal orientation (see scheme 1), because of the porous aspect of a compact monolayer in this case. We thus assume that the [H7P8W48O184]33- adopt a horizontal configuration during the electrostatic deposition and rapidly saturate the surface, thus stopping further growing of the thickness. Note that if the rinsing of the AOT SAM was not done properly, the POMs clinged to small aggregates and an inhomogeneous layer was obtained (figure S1 in supporting information), which demonstrates how the PBS rinsing step is crucial. The quality of the seeding layer is thus essential to get a uniformly structured POM layer. Once optimized with common Au/Si and Au/glass substrates, the procedure was generalized to the deposition of [H7P8W48O184]33- onto TSAu and Au nanodots to perform electrical measurements. As TSAu was very flat (the roughness is around 0.4 nm and there is no grain boundaries), the thickness of the AOT SAM determined by ellipsometry was quite lower (0.7 nm) and the POM layer was measured slightly thinner (0.6 nm). This decrease of the AOT monolayer thickness observed on the TSAu could be related to the lower roughness of this SAM (0.60 nm, figure 1-d) compared to the one on Au/glass substrates (roughness around 1.27 nm, figure 1-b). We also note that in the case of SAMs on TSAu substrates the PBS rinsing step was not essential, probably because of the higher quality/lower roughness of the bare gold surface (figures 1-d and e). The SAMs on TSAu substrates were characterized by XPS, see Experimental Section (Figure 3). The presence of the POMs was inferred from the W4f doublet at 35.7 and 37.9 eV corresponding to the 4f7/2 and 4f5/2 levels respectively for W(VI) atoms. The P2p peak was also observed at 133.9 eV, in agreement with fully oxidized phosphorous atoms.93 A S2p doublet was observed at 162.1 ± 0.1 and 163.3 ± 0.1 eV, for the S2p3/2 and the S2p1/2 signals respectively, corresponding to S bound to Au (75.5 %). An additional contribution was observed at 163.7 ± 0.1 (S2p3/2) and 164.8 ± 0.1 eV (S2p1/2) attributed to non Au- bonded thiol molecules (24.5%). Finally the protonation of the amino-groups which had driven the electrostatic POM deposition was assessed from the N1s peak, which was decomposed in two almost equal contributions at 400.1 ± 0.1 and 402.1 ± 0.1 eV for the unprotonated (37.7%) and protonated (62.3 %) amino-groups respectively (figure 3). 7 Figure 3. XPS characterization of the [H7P8W48O184]33-layer onto AOT SAM prepared onto TSAu: (a) N1s, (b) S2p, (c) W4f and (d) P2p peaks. 3. Electrical characterizations. Electronic transport properties through molecular junctions are impacted by numerous parameters such as contact geometry, molecule orientation, molecular organization, and number of molecules in large junction, for example. Conductance histograms thus become essential to describe electron transport properties in single molecule junctions94 as well as larger molecular junctions.95 One way to obtain statistics is to repeat measurements on various single molecule junctions or on various points on the same larger molecular junction. In an alternative way, some of us have recently reported on the use of an array of sub-10 nm Au nanodots to record the conductance of up to a million of alkyl-thiol 8 junctions in a single C-AFM image.86-87 Analysis of the collected data is used to determine the electronic structure of the molecular junctions. POMs on Au-TS electrodes. Electronic transport properties of the [H7P8W48O184]33- electrostatically deposited onto the AOT SAMs were investigated at the nanoscale by C-AFM. The SAMs grafted on TSAu substrate were measured under a nitrogen flow using a platinum tip and at a loading force of 6 nN (see Experimental Section). TSAu-S-C8- This conducting +//[H7P8W48O184]33-//Pt nanojunctions (here // denotes a non-covalent interface and -- a covalent NH3 interface). A square grid of 10 × 10 µm is defined with a lateral step of 100 nm. At each point, one I−V curve is acquired (back and forth), leading to the measurements of 100 I−V traces on a given location on the sample. This procedure was repeated at 3 different locations on the sample. The bias was applied on the Au-TS substrate, and the tip was grounded through the input of the current amplifier (figure 4-a). The voltage sweeps (back and forth) were applied from -1.5 to 1.5 V and then from 1.5 to −1.5 V. stationary point contact tip was placed at a form to (a) (b) -8 -9 -10 -11 -12 -13 ) A g o L ( ) ) t n e r r u C ( s b A ( g o L -14 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Bias (V) 50 44 38 31 25 19 13 6.3 0.0 (c) 50 e c n e r r u c c O 40 30 20 10 0 -14 -12 -13 Log (Abs(Current)) at +0.3V (Log A) -11 -10 -9 -8 Figure 4. (a) Scheme of the electronic transport characterization by C-AFM of the [H7P8W48O184]33- electrostatically deposited onto AOT SAM. (b) 2D current histogram of 300 I-V curves measured at several locations on the SAM. (c) Current histogram at + 0.3 V adjusted by a log-normal distribution. The higher bar at log I = -8.4 is due to the saturation of the amperemeter and corresponds to the number of counts of current higher than 4x10-9 A (i.e. the hidden higher tail of the distribution). 9 tip molecular Figure 4-b shows the 2D histogram of the 300 I-V traces measured on the SAM nanojunctions. The I- V curves are quite symmetric with respect to the voltage polarity. The current histograms at a given bias (for example at +0.3 V in figure 4-c) were fitted by a log-normal distribution, leading to a mean value of log I = -10.5 (i.e. a mean current of 3.2 x 10-11 A) with a standard variations log-σ ≃ 2.5. POMs on Gold Nanodots (nanodot molecule junction : NMJ). Molecular junctions of the [H7P8W48O184]33- electrostatically deposited onto AOT SAM were also fabricated on a large array of single crystal Au nanodot electrodes.86-87 An array of gold nanodot (10 nm in diameter) electrodes was fabricated by e-beam lithography and lift-off technique.86-87 Each nanodot is embedded in a highly doped Si substrate (to ensure a back ohmic contact), covered by a SAM of molecules of interest and contacted by the C-AFM tip. Each nanodot/molecules/tip junction contains about 25 POMs (considering a nanodot diameter of 10 nm, a POM diameter of 2 nm, and assuming a close-packing). Figure 5-a shows a typical C-AFM image taken at +0.3 V for 197 Au nanodot-AOT//[H7P8W48O184]33-//C-AFM current histogram corresponding to the current image is shown in figure 5-b for voltage + 0.3 V. This histogram was well fitted by a log-normal distributions giving the average current of log I = - 7.29 (i.e. a mean current of 5.1 x 10-8 A) and a standard deviation log σ ≃ 1.5. It may be possible to consider two peaks in the histogram shown in figure 5-b, as observed in previous works on NMJs.87 However, this feature was not systematically observed for the present samples (see other histograms in the supporting information, figure S3 ). We have mainly observed a tail at low current in the histograms deviating from the log-normal distribution, a feature previously attributed to intermolecular interactions between adjacent molecules in the NMJs.96-97 This confirms a close packing of the [H7P8W48O184]33- molecules in the NMJs. We note that, albeit using the same loading force (6 nN), the currents are higher in the NMJs compared to the SAM on TSAu substrates. This has been rationalized in a previous study combining mechanical and electrical characterizations of the molecule/nanodot structure: it was found that the applied pressure is increased in the NMJs due to the smaller contact area (compared to C- AFM on SAM on large metal surface), leading to higher currents.98 By successive acquisition of the current images at various voltages, a reconstructed I-V curve is obtained for the molecular junction (figure 5-c). The current histograms at the different voltages are given in the supporting information (figure S3). junctions. The 10 (a) (b) e c n e r r u c c O 35 30 25 20 15 10 5 0 -11 -9 -10 -7 Log (Abs(I)) at + 0.3V (Log A) -8 -6 (c) ) A ( ) t n e r r u C M F A C ( g o l -6 -7 -8 -9 -10 -11 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 Bias (V) 70 61 53 44 35 26 18 8.8 0.0 Figure 5. (a) Current image of 8 x 8 µm at + 0.3 V for a loading force fixed at 6 nN (white spot = positive current) of Au nanodots capped with [H7P8W48O184]33-electrostatically deposited onto AOT SAM. (b) Corresponding histograms of the current with 197 junctions fitted with a log-normal distribution : average current of log I = -7.29 (i.e. a mean current of 5.1 x 10-8 A) and a standard deviation log-σ≃1.5. (c) Reconstructed I-V curve from the current image at different voltages: the current value and the incertitude in the curve correspond to the mean current value and standard deviation respectively deduced from the log-normal distribution fitting. The reconstructed I-V curve (black line) is superimposed on the 2D histogram of the current measured at the different voltages on the NMJs. Only a fraction (around 3 %) of the junctions are electrically active (bright spots in figure 5-a) compared to the total number of nanodots on the scanned area (8 x 8 µm, with a nanodot every 100 nm, Experimental Section and images in the supporting information, figure S2). This feature may have several origins (see the supporting information). 11 4. Discussion: molecular junction energetics. The I-V curves were analyzed with three methods to determine the electronic structure of the molecular junctions, i.e. the energy position of the molecular orbital involved in the electron transport process. The first one consists to fit the I-V curves with the Simmons tunnel equation99 (see supporting information), the second one is the TVS (transition voltage spectroscopy) method,100-104 and the third consists to fit the I-Vs with a molecular single energy level model based on the Landauer equation and considering that the electron transfer is dominated by one molecular orbital in the junction ("molecular model", see SI).105-107 In the first case, the fit gives the effective tunnel energy barrier, ΦT, seen by electrons to tunnel from one electrode to the other. In the TVS method, a threshold voltage VT is extracted from the minimum of (LnI/V2) versus 1/V plot. This voltage is related to the energy position of the molecular orbital, ε, relative to the electrode Fermi energy, by ε = 0.86eVT with e the electron charge (see supporting information). This energy level is also determined by fitting the "molecular model" and noted as ε0 in the following. In the present case, since the POMs are strongly accepting molecules, we infer that the molecular orbital involved in the electron transport though the molecular junction is the POM LUMO. Typical examples of these three methods applied on the data shown in figures 4-b and 5-c for the SAM and NMJs respectively are given in figure S4 (supporting information). Figure 6 gives the histograms of the energy levels determined by the three methods for the POMs junctions on TSAu substrates. Table I summarizes the values obtained with these three methods on the SAMs and NMJs. POM SAMs on TSAu Table I. Summary for the values obtained by the three methods on the POM SAMs grafted on Au-TS and NMJs. These values are obtained from the normal distribution of the fitting values presented in Fig. 6 for SAMs and in the supporting information(figure S7) for NMJs. Due to the small numbers of voltages used for the IV measurements, the TVS method is not applicable for NMJs. ΦT (Simmons model) ε (TVS method) ε0 (molecular model) (eV) 0.83 ± 0.16 0.72 ± 0.07 0.93 ± 0.15 NMJs (eV) 0.26 ± 0.33 not applicable 0.24 ± 0.02 12 e c n e r r u c c O (b) 9 8 7 6 5 4 3 2 1 0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 ε (eV) 1.5 2.0 1.0 ε0 (eV) e c n e r r u c c O (a) 10 9 8 7 6 5 4 3 2 1 0 0.0 0.5 1.0 ΦT (eV) 1.5 2.0 (c) 7 6 5 4 3 2 1 0 0.0 e c n e r r u c c O 0.5 Figure 6. Statistical distributions of the energy levels deduced by applying the 3 methods on 21 I-V curves arbitrarily taken from data shown in figure 4: (a) Simmons model, (b) TVS method, (c) molecular model. The error bars in Table I are taken form the FWHM of these distributions. Examples of the I-V curve adjustments with the three models and distributions for the other fitting parameters (electron effective mass and electrical surface contact area for the Simmons model) are presented in the supporting information (figures S4, S5 and S6). These values are now discussed in term of energetics of the molecular junctions. The first observation is that the ε values given by the TVS method or by fitting the molecular level are in good agreement as expected since the relation ε = 0.86eVT is also based on the single molecular energy model (see SI).103, 108 The measured molecular junctions are constituted of 2 parts: the alkyl spacer and the POM (Scheme 1). The alkyl spacer is electrically insulating with a high HOMO-LUMO gap (7-9 eV),109-110 while the POM has a smaller gap (4.7 eV from optical absorption). We can consider a staircase 13 diagram for the energetics of the junction as depicted in figure 7 where the LUMO of the alkyl chains (C8) is supposed higher in energy than the LUMO of the POMs. In such a case, the Simmons model considers only a simple rectangular energy tunnel barrier with an effective tunnel barrier ΦT between the LUMOs of the alkyl chain and the POM, εPOM < ΦT < εC8. When a voltage is applied, the Simmons model considers a linear variation of the electric field between the electrodes (figure 7-c). On the contrary, the TVS method used here (with the relationship ε = 0.86eVT) and the molecular model give the energy of the molecular orbital, considering that the potential drops are mainly located at the molecule (i.e. POM unit)/electrode interface (i.e. through the alkyl spacer and the mechanical contact between the POM and the C-AFM tip) - figure 7-e (for a more detailed discussion of the Simmons model and TVS method, see Refs. 111, 112). We can consider that the measured ε and ε0 (Table I) are reasonable estimate of εPOM. In the WKB (Wentzel-Kramers-Brilloin) approximation, the effective tunnel barrier ΦT for such a staircase tunnel barrier (Figure 7) is given by: !! ΦT(dC8 + dPOM)= εC8dC8 + εPOMdPOM (1) where dC8 and dPOM are the thicknesses of the C8 alkyl chain and POM, respectively, dC8=0.7 nm and dPOM=0.6 nm (as measured by ellipsometry on the TSAu substrates). Since the values for εPOM (i.e. ε or ε0) and ΦT are quite similar (Table I), we conclude that the LUMO of the alkyl spacer, εC8, and the POM, εPOM, are roughly similar (alkyl spacer and POM LUMOs energetically aligned). This feature is consistent with the value εC8 of ca. 1 eV previously determined for the LUMO of C8 (pure alkyl chain, no amine) SAM on Au by TVS104 and IPES.113 We hypothesis that the dipole at the AOT/POM interface (positive charge at the AOT amine end-group, negative charge on the POM side) can induced this alignment by shifting downstairs (upstairs, respectively) the energetics of C8 and POM, respectively. The second observation is that the values for NMJs are smaller than for SAMs on TSAu. Again this is consistent with the higher currents measured on NMJs and with previous report on the mechanical behaviors on NMJs (higher pressure at the same loading force induces a reduction of ε).98 We previously reported a reduction by a factor about 2 of the TVS value (ε) for a C8 monolayer on NMJs98 compared to a C8 SAM on golf surface,104 both measured by C-AFM at the same loading force. The fact that, for NMJs, we still have ΦT ≈ ε0 means that the same conclusion (energy level alignment of the alkyl spacer and POM LUMOs) holds for NMJs. 14 Figure 7. (a) Scheme of the staircase energetic diagram of the Au-AOT//POM//C-AFM tip junction. (b-c) Effective tunnel barrier ΦT determined with the Simmons model at 0 V (b) and at a voltage V (c). (d-e) Determination of the energy position of the POM LUMO,εPOM, by the TVS method and molecular model at 0 V (d) and at a voltage V (e). 5. Conclusion. In this contribution, we have described a reliable procedure for the electrostatic immobilization of POMs onto positively charged SAMs of amino-alkylthiols onto Au. Uniformly structured monolayers of POMs have been directly obtained onto TSAu, while a rinsing methodology at the SAM formation step had to be optimized in the case of usual Au/glass or Au/Si substrates of higher roughness. Controlling the morphology/organization of the POM layers is essential to ensure reproducible properties. This procedure is easy to extend to other POMs, especially those with high electron-storage abilities. We have thus recorded the first electrical data on molecular junctions involving the large [H7P8W48O184]33- POM, analyzed these data by various methods and experimentally determined the energy position of the lowest unoccupied molecular orbital of the POM (with respect of the metal electrode Fermi energy), which governs electron transport through these molecular devices. 15 from Sigma-Alrich and used as solvents were purchased Experimental section. Materials Chemicals and received. K28Li5[H7P8W48O184] was prepared as described in the literature92 and its purity checked by 31P NMR and electrochemistry (figures S8-S10). The Au/Si substrates, purchased from Sigma-Aldrich, consisted of a 100 nm gold layer adhered to the silicon wafer by means of a titanium adhesion layer. The substrates were treated 2 min in pure sulfuric acid and rinsed abundantly with water and isopropanol and dried under N2. Before immersion in the thiol solution, the substrates were exposed to UV-ozone during 20 minutes and rinsed with ethanol. Au/glass substrates, coated successively with a 50 Å thick layer of chromium and a 200 nm thick layer of gold, were purchased from Arrandee (Werther, Germany). To ensure a good crystallinity of the gold top layer, the substrates were annealed in a butane flame, afterwards exposed to UV-ozone during 20 minutes and rinsed with ethanol prior to immersion in the thiol solution. Template-stripped Au substrates. Very flat TSAu surfaces were prepared according to the method reported by the Whiteside group.85 In brief, a 300−500 nm thick Au film is 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 2:1 H2SO4/H2O2 (v/v); Caution: Piranha solution is exothermic and strongly reacts 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 UVirradiated in ozone for 10 min) is glued (UV polymerizable glue) on the 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 AuTS surface is immediately used for the formation of the SAM. Gold nanodots fabrication. The fabrication and detailed characterization of these nanodot arrays have been reported elsewhere.86 For e-beam lithography, an EBPG 5000 Plus from Vistec Lithography was used. The (100) Si substrate (resistivity = 10-3 Ω.cm) was cleaned with UV-ozone and native oxide etched before resist deposition. The e-beam lithography has been optimized by using a 45 nm thick diluted (3:5 with anisole) PMMA (950 K). For the writing, an acceleration voltage of 100 keV was used, which reduces proximity effects around the dots, compared to lower voltages. The beam current to expose the nanodots was 1 nA. The conventional resist development / e-beam Au evaporation (8 nm) / lift-off processes were used. Immediately before evaporation, native oxide is removed with dilute HF solution to allow good electrical contact with the substrate. At the end of the process, these nanodots were covered with a thin layer of SiO2 that was removed by HF at 1% for 1 mn prior to SAM deposition. Spacing between Au nanodots was set to 100 nm (see images in the supporting information, figure S2). 16 Characterization techniques. Ellipsometry Ellipsometry measurements were performed on the Au/Si substrates with a monowavelength ellipsometer SENTECH SE 400 equipped with a He-Ne laser at λ = 632.8 nm. The incident angle was 70°. As the optical indices of the bare gold substrate change from one sample to another, the ns and ks values of the sample with bare Au were systematically measured just before immersion in thiol solution. The ns and ks values were around 0.2 and 3.5 respectively. For the 8- amino-1-octanethiol SAM, typical optical indices of an organic monolayer were used (ns = 1.5, ks = 0). Finally, ns = 1.48 and ks = 0 were used for the layer of POMs.79 At least 6 measurements were performed on a same sample in different zones, to check the homogeneity of the layer. A mean value for the thickness was calculated when the standard deviation was equal or lower than 0.2 nm. X-ray photoelectron spectroscopy XPS analyses were performed on using an Omicron Argus X-ray photoelectron spectrometer. The monochromated AlKα radiation source (hν = 1486.6 eV) had a 300 W electron beam power. The emission of photoelectrons from the sample was analyzed at a takeoff angle of 90° under ultra-high vacuum conditions (≤ 10-10 Torr). Spectra were carried out with a 100 eV pass energy for the survey scan and 20 eV pass energy for the P2p, W4f, C1s, O1s, N1s, S2p regions. Binding energies were calibrated against the Au4f7/2 binding energy at 84.0 eV and element peak intensities were corrected by Scofield factors.114 The peak areas were determined after subtraction of a linear background. The spectra were fitted using Casa XPS v.2.3.15 software (Casa Software Ltd., U.K.) and applying a Gaussian/Lorentzian ratio G/L equal to 70/30. Polarized modulated reflexion absorption infrared spectroscopy (PM-RAIRS) PM-RAIRS measurements were performed on Au/Si or Au/glass substrates using a Nicolet Nexus 5700 FT-IR spectrometer equipped with a nitrogen-cooled HgCdTe wide band detector. Infrared spectra were recorded at 8 cm-1 resolution, by co-addition of 128 scans. A ZnSe grid polarizer and a ZnSe photoelastic modulator were placed prior to the sample in order to modulate the incident beam between p and s polarizations (HINDS Instruments, PM90, modulation frequency = 36 kHz). The sum and difference interferograms were processed and underwent Fourier-transformation to yield the PM- RAIRS signal which is the differential reflectivity (ΔR/R°) = (Rp-Rs)/(Rp+Rs), where Rp is the signal parallel to the incident plane while Rs is the perpendicular contribution. The measurements were done at two different voltages applied to the modulator ZnSe crystal to optimize the sensitivity. Atomic force microscopy. AFM images were recorded on Au/glass substrates using a commercial AFM (NanoScope VIII MultiMode AFM, Bruker Nano Inc., Nano Surfaces Division, Santa Barbara, CA) equipped with a 150 × 150 × 5 µm scanner (J-scanner). The substrates were fixed on a steel sample puck using a small piece of adhesive tape. Images were recorded in peak force tapping mode in air at room temperature (22−24 °C) using oxide-sharpened microfabricated Si3N4 cantilevers (Bruker Nano Inc., Nano Surfaces Division, Santa Barbara, CA). The spring constants of the cantilevers were measured using 17 the thermal noise method, yielding values ranging from 0. 4 to 0.5 N/m. The curvature radius of silicon nitride tips was about 10 nm (manufacturer specifications). The raw data were processed using the imaging processing software NanoScope Analysis, mainly to correct the background slope between the tip and the surfaces. C-AFM measurements. Conducting atomic force microscopy (C-AFM) was performed under a flux of N2 gas (ICON, Bruker), using a solid metal probe in platinum (tip radius of curvature less than 20 nm, force constant 0.3 N/m, reference: RMN-12Pt400B from Bruker). The tip loading force on the surface was fixed at 6 nN thanks to force−distance curves with the controlling software of the ICON. The choice of the loading force of 6 nN is a tradeoff between too low or unstable currents and too high deformation or damage of the organic monolayer.115 Previous works both on SAMs on large electrodes116 and NMJs98 have estimated a deformation below 0.1 nm at 6 nN for C8 alkyl chains (as used here). These values remain smaller than the accuracy of the ellipsometry measurements (see above) used to determine the monolayer thickness, which is further used as a parameter in our model to analyze the current-voltage curves. Images were acquired with a sweep frequency of 0.5 Hz and the voltage was applied on the substrate. Preparation of the seeding 8-amino-1-octanethiol SAM The gold substrate was immersed in a solution of 8-amino-1-octanethiolhydrochloride in absolute ethanol AE (1mg in 50 mL, 10-4 mol.L-1) during 24h, protected from light. Then the surface was rinsed with AE, plunged into an AE bath during 5 min, sonicated 5 min in a new AE bath, rinsed with AE and dried with a N2 flow. The dried substrate was treated by immersion in a PBS solution during 3h (0.01M, pH=7.4), then submitted to several successive rinsing steps: (i) distilled water flow, (ii) sonication 5 min in distilled water, (iii) distilled water then AE flow, before being dried under N2. Immobilization of K28Li5[H7P8W48O184] A 2 mol.L-1 LiCl solution in water was prepared (0.8 g in 10 mL) then 136.1 mg of K28Li5[H7P8W48O184] were added (10-3 mol.L-1). The thiol modified gold substrate was immersed in the clear solution during 1 hour. Finally, the substrate was rinsed with a distilled water flow, AE flow and dried under N2. Decreasing (20 min) or increasing (24 h) the incubation time was shown (see supporting information) to have little effect on the layer thickness according to ellipsometry measurements. The films are at least stable in the course of the electrical and physical characterizations (i.e. days to weeks). All the measurements have been carried out at the solid state and at variance with electrochemical studies carried out in solution, no leaching of the POMs is feared. We also questioned the stability of such electrostatically assembled POM layer in solution. The modified gold substrate (Au/glass) was placed in a 1 M solution of (NBu4)4NPF6 in acetonitrile and sonicated for a few hours, then rinsed with acetonitrile. Despite these quite harsh conditions, ellipsometry did not reveal any significant modification of the thickness, which led us to conclude that 18 POMs were not significantly released from the substrate. This is tentatively ascribed to the high total charge of the POMs. Supporting information. Additional AFM images, current histograms for NMJs, descriptions and equations of the models, examples of fitted I-V curves, histograms of fitted Simmons model parameters (effective mass, electrical contact area), cyclic voltamogram and NMR spectrum of K28Li5[H7P8W48O184].. Conflicts of interest. There are no conflicts to declare. Acknowledgements. K.D.F. thanks the program PER-SU of Sorbonne Universités for funding. We thank the French RENATECH network (IEMN clean-room). The authors would like to acknowledge IMPC (Institut des Matériaux de Paris Centre, FR CNRS 2482) at Sorbonne Université and the C'Nano projects of Région Ile-de-France for Omicron-XPS apparatus funding. References Green, J. E.; Choi, J. W.; Boukai, A.; Bunimovich, Y.; Johnston-Halperin, E.; Pope, M. T., Heteropoly and Isopoly Oxometalates. 1983. 19 Roth, K. M.; Yasseri, A. A.; Liu, Z. M.; Dabke, R. B.; Malinovskii, V.; Schweikart, K. Liu, Z. M.; Yasseri, A. A.; Lindsey, J. S.; Bocian, D. F., Molecular memories that 1. DeIonno, E.; Luo, Y.; Sheriff, B. A.; Xu, K.; Shin, Y. S.; Tseng, H. R.; Stoddart, J. F.; Heath, J. R., A 160-kilobit molecular electronic memory patterned at 10(11) bits per square centimetre. Nature 2007, 445 (7126), 414-417. 2. survive silicon device processing and real-world operation. Science 2003, 302 (5650), 1543- 1545. 3. H.; Yu, L. H.; Tiznado, H.; Zaera, F.; Lindsey, J. S.; Kuhr, W. G.; Bocian, D. F., Measurements of electron-transfer rates of charge-storage molecular monolayers on Si(100). Toward hybrid molecular/semiconductor information storage devices. J. Am. Chem. Soc. 2003, 125 (2), 505-517. 4. Chem. Res. 2011, 44 (8), 638-650. 5. Terminated Silicon Surfaces. Toward the Development of Charge Storage and Communication Devices. Acc. Chem. Res. 2010, 43 (12), 1509-1518. 6. for Multistate Charge Storage Information. Adv. Mater. 2013, 25 (3), 462-468. 7. Acceptor Compound Containing Boron(III) Centers for Small-Molecule-Based High- Performance Ternary Electronic Memory Devices. Angew. Chem. Int. Ed. 2015, 54 (36), 10569-10573. 8. Lindsey, J. S.; Bocian, D. F., Molecules for Charge-Based Information Storage. Acc. Mas-Torrent, M.; Rovira, C.; Veciana, J., Surface-Confined Electroactive Molecules Poon, C. T.; Wu, D.; Lam, W. H.; Yam, V. W. W., A Solution-Processable Donor- Fabre, B., Ferrocene-Terminated Monolayers Covalently Bound to Hydrogen- Sadakane, M.; Steckhan, E., Electrochemical properties of polyoxometalates as Hong, L.; Tanaka, H.; Ogawa, T., Rectification direction inversion in a Balliou, A.; Douvas, A. M.; Normand, P.; Tsikritzis, D.; Kennou, S.; Argitis, P.; Douvas, A. M.; Makarona, E.; Glezos, N.; Argitis, P.; Mielczarski, J. A.; Mielczarski, 9. electrocatalysts. Chem. Rev. 1998, 98 (1), 219-237. 10. Maestre, J. M.; Lopez, X.; Bo, C.; Poblet, J. M.; Casan-Pastor, N., Electronic and magnetic properties of alpha-keggin anions: A DFT study of [XM12O40](n-) (M = W, Mo; X = Al-III, Si-IV,P-V, Fe-III, Co-II, Co-III) and [SiM11VO40](m-) (M = Mo and W). J. Am. Chem. Soc. 2001, 123 (16), 3749-3758. 11. Lopez, X.; Bo, C.; Poblet, J. M., Electronic properties of polyoxometalates: Electron and proton affinity of mixed-addenda Keggin and Wells-Dawson anions. J. Am. Chem. Soc. 2002, 124 (42), 12574-12582. 12. Poblet, J. M.; Lopez, X.; Bo, C., Ab initio and DFT modelling of complex materials: towards the understanding of electronic and magnetic properties of polyoxometalates. Chem. Soc. Rev. 2003, 32 (5), 297-308. 13. Pope, M. T., in Polyoxometalate Molecular Science, Eds J. J. Borras-Almenar, E. Coronado, A. Müller, M. T. Pope, Kluwer Academic Publishers, Dortrecht. 2003, 3-31. 14. phosphododecamolybdic acid/single-walled carbon nanotube junction. J. Mater. Chem. C 2013, 1 (6), 1137-1143. 15. Sherif, S.; Rubio-Bollinger, G.; Pinilla-Cienfuegos, E.; Coronado, E.; Cuevas, J. C.; Agrait, N., Current rectification in a single molecule diode: the role of electrode coupling. Nanotechnology 2015, 26 (29). 16. E., Polyoxometalate-based layered structures for charge transport control in molecular devices. ACS Nano 2008, 2 (4), 733-742. 17. Glezos, N., Tungsten polyoxometalate molecules as active nodes for dynamic carrier exchange in hybrid molecular/semiconductor capacitors. J. Appl. Phys. 2014, 116 (14), 143703. 18. Asenov, A.; Pedersen, R. H.; Gadegaard, N.; Mirza, M. M.; Paul, D. J.; Poblet, J. M.; Cronin, L., Design and fabrication of memory devices based on nanoscale polyoxometalate clusters. Nature 2014, 515 (7528), 545-549. 19. shapes, orientation, and packing of polyoxometalate arrays imaged by scanning tunneling microscopy. Inorg. Chem. 1998, 37 (3), 398-406. 20. From solids and solutions to surfaces. Chem. Rev. 1998, 98 (1), 297-306. 21. polyoxometalate, alpha-SiW11O398-, and electrode surfaces. Langmuir 2003, 19 (21), 8934- 8942. 22. tunneling microscopy (STM) and tunneling spectroscopy (TS) studies of polyoxometalates (POMs) of the Wells-Dawson structural class. Journal of Molecular Catalysis a-Chemical 2007, 262 (1-2), 216-226. 23. U., STM/STS observation of polyoxoanions on HOPG surfaces: the wheel-shaped [Cu20Cl(OH)(24)(H2O)(12)(P8W48O184)](25-) and the ball-shaped [{Sn(CH3)(2)(H2O)}(24){Sn(CH3)(2)}(12)(A-PW9O34)(12)](36-). Inorg. Chem. 2006, 45 (7), 2866-2872. Alam, M. S.; Dremov, V.; Muller, P.; Postnikov, A. V.; Mal, S. S.; Hussain, F.; Kortz, Busche, C.; Vila-Nadal, L.; Yan, J.; Miras, H. N.; Long, D. L.; Georgiev, V. P.; Kaba, M. S.; Song, I. K.; Duncan, D. C.; Hill, C. L.; Barteau, M. A., Molecular Klemperer, W. G.; Wall, C. G., Polyoxoanion chemistry moves toward the future: Kim, J.; Gewirth, A. A., Interactions between the Keggin-type lacunary In, K. S. A.; Kabab, M. S.; Nomiyac, K.; Finked, R. G.; Barteau, M. A., Scanning 20 Keita, B.; Nadjo, L.; Belanger, D.; Wilde, C. P.; Hilaire, M., Electrochemical Quartz- Sloan, J.; Liu, Z.; Suenaga, K.; Wilson, N. R.; Pandey, P. A.; Perkins, L. M.; Rourke, 24. J. P.; Shannon, I. J., Imaging the Structure, Symmetry, and Surface-Inhibited Rotation of Polyoxometalate Ions on Graphene Oxide. Nano Lett. 2010, 10 (11), 4600-4606. 25. Zhong, D. Y.; Sousa, F. L.; Muller, A.; Chi, L. F.; Fuchs, H., A Nanosized Molybdenum Oxide Wheel with a Unique Electronic-Necklace Structure: STM Study with Submolecular Resolution. Angew. Chem. Int. Ed. 2011, 50 (31), 7018-7021. 26. Crystal Microbalance - Evidence for the Adsorption of Heteropoly and Isopoly Anions on Gold Electrodes. J. Electroanal. Chem. 1995, 384 (1-2), 155-169. 27. Kuhn, A.; Anson, F. C., Adsorption of monolayers of P2Mo18O626- and deposition of multiple layers of Os(bpy)(3)(2+)-P2Mo18O626- on electrode surfaces. Langmuir 1996, 12 (22), 5481-5488. 28. reactions: Recent achievements. Journal of Molecular Catalysis a-Chemical 2007, 262 (1-2), 190-215. 29. Ammam, M., Polyoxometalates: formation, structures, principal properties, main deposition methods and application in sensing. Journal of Materials Chemistry A 2013, 1 (21), 6291-6312. Luo, X. Z.; Li, F. Y.; Xu, B. B.; Sun, Z. X.; Xu, L., Enhanced photovoltaic response of 30. the first polyoxometalate-modified zinc oxide photoanode for solar cell application. J. Mater. Chem. 2012, 22 (30), 15050-15055. 31. Kawasaki, N.; Wang, H.; Nakanishi, R.; Hamanaka, S.; Kitaura, R.; Shinohara, H.; Yokoyama, T.; Yoshikawa, H.; Awaga, K., Nanohybridization of Polyoxometalate Clusters and Single-Wall Carbon Nanotubes: Applications in Molecular Cluster Batteries. Angew. Chem. Int. Ed. 2011, 50 (15), 3471-3474. 32. Cronin, L., High-Performance Polyoxometalate-Based Cathode Materials for Rechargeable Lithium-Ion Batteries. Adv. Mater. 2015, 27 (31), 4649-4654. 33. polyoxometalates to carbon nanotubes using a sonication-driven periodic functionalization strategy. Energy Environ. Sci. 2016, 9 (3), 1095-1101. 34. Based Bottom-Up Fabrication of Graphene Quantum Dot/Manganese Vanadate Composites as Lithium Ion Battery Anodes. Chem. Eur. J. 2017, 23 (65), 16637-16643. 35. Rincon, M. E., Electrochemical supercapacitors based on novel hybrid materials made of carbon nanotubes and polyoxometalates. Electrochem. Commun. 2007, 9 (8), 2088-2092. 36. Srinivasan, M., A novel SWCNT-polyoxometalate nanohybrid material as an electrode for electrochemical supercapacitors. Nanoscale 2015, 7 (17), 7934-7941. 37. Asymmetric Supercapacitors Based on Reduced Graphene Oxide with Different Polyoxometalates as Positive and Negative Electrodes. Chemsuschem 2017, 10 (13), 2742- 2750. 38. Musumeci, C.; Luzio, A.; Pradeep, C. P.; Miras, H. N.; Rosnes, M. H.; Song, Y. F.; Long, D. L.; Cronin, L.; Pignataro, B., Programmable Surface Architectures Derived from Hybrid Polyoxometalate-Based Clusters. J. Phys. Chem. C 2011, 115 (11), 4446-4455. 39. the assembly of carboxylic acid-functionalised hybrid polyoxometalates. Crystengcomm 2013, 15 (22), 4422-4430. Chen, H. Y.; Al-Oweini, R.; Friedl, J.; Lee, C. Y.; Li, L. L.; Kortz, U.; Stimming, U.; Chen, J. J.; Symes, M. D.; Fan, S. C.; Zheng, M. S.; Miras, H. N.; Dong, Q. F.; Hu, J.; Ji, Y. C.; Chen, W.; Streb, C.; Song, Y. F., "Wiring" redox-active Ji, Y. C.; Hu, J.; Biskupek, J.; Kaiser, U.; Song, Y. F.; Streb, C., Polyoxometalate- Cuentas-Gallegos, A.; Martinez-Rosales, R.; Baibarac, M.; Gomez-Romero, P.; Dubal, D. P.; Chodankar, N. R.; Vinu, A.; Kim, D. H.; Gomez-Romero, P., Keita, B.; Nadjo, L., Polyoxometalate-based homogeneous catalysis of electrode Hutin, M.; Yvon, C.; Yan, J.; Macdonell, A.; Long, D. L.; Cronin, L., Programming 21 Palilis, L. C.; Vasilopoulou, M.; Douvas, A. M.; Georgiadou, D. G.; Kennou, S.; Alaaeddine, M.; Zhu, Q.; Fichou, D.; Izzet, G.; Rault, J. E.; Barrett, N.; Proust, A.; Palilis, L. C.; Vasilopoulou, M.; Georgiadou, D. G.; Argitis, P., A water soluble Tountas, M.; Topal, Y.; Kus, M.; Ersoz, M.; Fakis, M.; Argitis, P.; Vasilopoulou, M., Liu, S. Q.; Tang, Z. Y., Polyoxometalate-based functional nanostructured films: Yang, H. X.; Song, T.; Liu, L.; Devadoss, A.; Xia, F.; Han, H.; Park, H.; Sigmund, Ohisa, S.; Kagami, S.; Pu, Y. J.; Chiba, T.; Kido, J., A Solution-Processed Heteropoly 40. Stathopoulos, N. A.; Constantoudis, V.; Argitis, P., Solution processable tungsten polyoxometalate as highly effective cathode interlayer for improved efficiency and stability polymer solar cells. Sol. Energy Mater. Sol. Cells 2013, 114, 205-213. 41. Vasilopoulou, M.; Polydorou, E.; Douvas, A. M.; Palilis, L. C.; Kennou, S.; Argitis, P., Annealing-free highly crystalline solution-processed molecular metal oxides for efficient single-junction and tandem polymer solar cells. Energy Environ. Sci. 2015, 8 (8), 2448-2463. 42. Tortech, L., Enhancement of photovoltaic efficiency by insertion of a polyoxometalate layer at the anode of an organic solar cell. Inorganic Chemistry Frontiers 2014, 1 (9), 682-688. 43. inorganic molecular oxide as a novel efficient electron injection layer for hybrid light- emitting diodes (HyLEDs). Org. Electron. 2010, 11 (5), 887-894. 44. Acid Containing MoO3 Units as a Hole-Injection Material for Highly Stable Organic Light- Emitting Devices. ACS Appl. Mater. Interfaces 2016, 8 (32), 20946-20954. 45. Water-Soluble Lacunary Polyoxometalates with Excellent Electron Mobilities and Hole Blocking Capabilities for High Efficiency Fluorescent and Phosphorescent Organic Light Emitting Diodes. Adv. Funct. Mater. 2016, 26 (16), 2655-2665. 46. Current progress and future prospects. Nano Today 2010, 5 (4), 267-281. 47. W.; Kwon, K.; Paik, U., Polyaniline/Polyoxometalate Hybrid Nanofibers as Cathode for Lithium Ion Batteries with Improved Lithium Storage Capacity. J. Phys. Chem. C 2013, 117 (34), 17376-17381. 48. C.; Bonchio, M.; McCormac, T., Surface Immobilization of a Tetra-Ruthenium Substituted Polyoxometalate Water Oxidation Catalyst Through the Employment of Conducting Polypyrrole and the Layer-by-Layer (LBL) Technique. ACS Appl. Mater. Interfaces 2014, 6 (11), 8022-8031. 49. and photochromic dual-mode devices. Langmuir 2006, 22 (5), 1949-1951. 50. Assembly and UV Photoreduction of Graphene-Polyoxometalate Composite Films for Electronics. J. Am. Chem. Soc. 2011, 133 (24), 9423-9429. 51. Walsh, J. J.; Bond, A. M.; Foreter, R. J.; Keyes, T. E., Coord. Chem. Rev. 2016, 217- 234. 52. Yang, Y. B.; Xu, L.; Li, F. Y.; Du, X. G.; Sun, Z. X., Enhanced photovoltaic response by incorporating polyoxometalate into a phthalocyanine-sensitized electrode. J. Mater. Chem. 2010, 20 (48), 10835-10840. 53. based on Dawson type polyoxometalate and porphyrin formed by layer-by-layer self assembly. Chem. Commun. 2013, 49 (5), 496-498. 54. Wang, E. B., Enhanced Visible Photovoltaic Response of TiO2 Thin Film with an All- Inorganic Donor-Acceptor Type Polyoxometalate. ACS Appl. Mater. Interfaces 2015, 7 (24), 13714-13721. Ahmed, I.; Farha, R.; Goldmann, M.; Ruhlmann, L., A molecular photovoltaic system Liu, S. Q.; Mohwald, H.; Volkmer, D.; Kurth, D. G., Polyoxometalate-based electro- Li, H. L.; Pang, S. P.; Wu, S.; Feng, X. L.; Mullen, K.; Bubeck, C., Layer-by-Layer Li, J. S.; Sang, X. J.; Chen, W. L.; Zhang, L. C.; Zhu, Z. M.; Ma, T. Y.; Su, Z. M.; Anwar, N.; Sartorel, A.; Yaqub, M.; Wearen, K.; Laffir, F.; Armstrong, G.; Dickinson, 22 Cheng, L.; Cox, J. A., Nanocomposite multilayer film of a ruthenium Bi, L. H.; Foster, K.; McCormac, T.; Dempsey, E., Preparation of multilayer films Yaqub, M.; Imar, S.; Laffir, F.; Armstrong, G.; McCormac, T., Investigations into the Rosnes, M. H.; Musumeci, C.; Pradeep, C. P.; Mathieson, J. S.; Long, D. L.; Song, Y. de Viguerie, L.; Mouret, A.; Brau, H. P.; Nardello-Rataj, V.; Proust, A.; Bauduin, P., Rosnes, M. H.; Musumeci, C.; Yvon, C.; Macdonell, A.; Pradeep, C. P.; Sartorio, C.; 55. metallodendrimer and a Dawson-type polyoxometalate as a bifunctional electrocatalyst. Chem. Mater. 2002, 14 (1), 6-+. 56. containing a crown heteropolyanion and an osmium functionalised pyrrole monomer. J. Electroanal. Chem. 2007, 605 (1), 24-30. 57. Electrochemical, Surface, and Electrocatalytic Properties of the Surface-Immobilized Polyoxometalate, TBA(3)K[SiW10O36(PhPO)(2)]. ACS Appl. Mater. Interfaces 2015, 7 (2), 1046-1056. 58. Clemente-Leon, M.; Coronado, E.; Gomez-Garcia, C. L.; Mingotaud, C.; Ravaine, S.; Romualdo-Torres, G.; Delhaes, P., Polyoxometalate monolayers in Langmuir-Blodgett films. Chem. Eur. J. 2005, 11 (13), 3979-3987. 59. F.; Pignataro, B.; Cogdell, R.; Cronin, L., Assembly of Modular Asymmetric Organic- Inorganic Polyoxometalate Hybrids into Anisotropic Nanostructures. J. Am. Chem. Soc. 2010, 132 (44), 15490-15492. 60. Musumeci, C.; Rosnes, M. H.; Giannazzo, F.; Symes, M. D.; Cronin, L.; Pignataro, B., Smart High-kappa Nanodielectrics Using Solid Supported Polyoxometalate-Rich Nanostructures. ACS Nano 2011, 5 (12), 9992-9999. 61. Surface pressure induced 2D-crystallization of POM-based surfactants: preparation of nanostructured thin films. Crystengcomm 2012, 14 (24), 8446-8453. 62. Long, D. L.; Pignataro, B.; Cronin, L., Exploring the Interplay Between Ligand Derivatisation and Cation Type in the Assembly of Hybrid Polyoxometalate Mn-Andersons. Small 2013, 9 (13), 2316-2324. 63. nanocarbon materials for energy conversion, energy storage and sensor systems. Energy Environ. Sci. 2015, 8 (3), 776-789. 64. Gonzalez, B. R.; Amenitsch, H.; Da Ros, T.; Casalis, L.; Goldoni, A.; Marcaccio, M.; Scorrano, G.; Scoles, G.; Paolucci, F.; Prato, M.; Bonchio, M., Efficient water oxidation at carbon nanotube-polyoxometalate electrocatalytic interfaces. Nature Chemistry 2010, 2 (10), 826-831. 65. A.; Maccato, C.; Ke, X. X.; Bittencourt, C.; Da Ros, T.; Van Tendeloo, G.; Marcaccio, M.; Paolucci, F.; Prato, M.; Bonchio, M., Knitting the Catalytic Pattern of Artificial Photosynthesis to a Hybrid Graphene Nanotexture. ACS Nano 2013, 7 (1), 811-817. 66. Bonchio, M.; Quintana, M.; Prato, M., Positive graphene by chemical design: tuning supramolecular strategies for functional surfaces. Chem. Commun. 2014, 50 (7), 885-887. 67. Lauinger, S. M.; Sumliner, J. M.; Yin, Q. S.; Xu, Z. H.; Liang, G. J.; Glass, E. N.; Lian, T. Q.; Hill, C. L., High Stability of Immobilized Polyoxometalates on TiO2 Nanoparticles and Nanoporous Films for Robust, Light-Induced Water Oxidation. Chem. Mater. 2015, 27 (17), 5886-5891. 68. Lauinger, S. M.; Piercy, B. D.; Li, W.; Yin, Q. S.; Collins-Wildman, D. L.; Glass, E. N.; Losego, M. D.; Wang, D. W.; Geletii, Y. V.; Hill, C. L., Stabilization of Polyoxometalate Water Oxidation Catalysts on Hematite by Atomic Layer Deposition. ACS Appl. Mater. Interfaces 2017, 9 (40), 35048-35056. Quintana, M.; Lopez, A. M.; Rapino, S.; Toma, F. M.; Iurlo, M.; Carraro, M.; Sartorel, Toma, F. M.; Sartorel, A.; Iurlo, M.; Carraro, M.; Parisse, P.; Maccato, C.; Rapino, S.; Ji, Y. C.; Huang, L. J.; Hu, J.; Streb, C.; Song, Y. F., Polyoxometalate-functionalized Hadad, C.; Ke, X. X.; Carraro, M.; Sartorel, A.; Bittencourt, C.; Van Tendeloo, G.; 23 He, T.; He, J. L.; Lu, M.; Chen, B.; Pang, H.; Reus, W. F.; Nolte, W. M.; Nackashi, D. Rinfray, C.; Izzet, G.; Pinson, J.; Derouich, S. G.; Ganem, J. J.; Combellas, C.; Huder, L.; Rinfray, C.; Rouchon, D.; Benayad, A.; Baraket, M.; Izzet, G.; Lipp- Xie, J. J.; Zhang, Y.; Han, Y. L.; Li, C. L., High-Capacity Molecular Scale Conversion Zhong, Y. L.; Ng, W.; Yang, J. X.; Loh, K. P., Electrostatically Self-Assembled Lu, M.; Nolte, W. A.; He, T.; Corley, D. A.; Tour, J. M., Direct Covalent Grafting of Song, Y. F.; McMillan, N.; Long, D. L.; Kane, S.; Malm, J.; Riehle, M. O.; Pradeep, Fleury, B.; Billon, M.; Duclairoir, F.; Dubois, L.; Fanton, A.; Bidan, G., Electrostatic 69. Anode Enabled by Hybridizing Cluster-Type Framework of High Loading with Amino- Functionalized Graphene. ACS Nano 2016, 10 (5), 5304-5313. 70. Polyoxometalates on Molecular-Dye-Functionalized Diamond. J. Am. Chem. Soc. 2009, 131 (51), 18293-18298. 71. immobilization of polyoxometallates on silicon: X-ray Photoelectron Spectroscopy and electrochemical studies. Thin Solid Films 2011, 519 (11), 3732-3738. 72. Errington, R. J.; Petkar, S. S.; Horrocks, B. R.; Houlton, A.; Lie, L. H.; Patole, S. N., Covalent immobilization of a TiW5 polyoxometalate on derivatized silicon surfaces. Angew. Chem. Int. Ed. 2005, 44 (8), 1254-1257. 73. P.; Franzon, P. D.; Tour, J. M., Controlled modulation of conductance in silicon devices by molecular monolayers. J. Am. Chem. Soc. 2006, 128 (45), 14537-14541. 74. Polyoxometalates onto Si Surfaces. Chem. Mater. 2009, 21 (3), 442-446. 75. C. P.; Gadegaard, N.; Cronin, L., Micropatterned Surfaces with Covalently Grafted Unsymmetrical Polyoxometalate-Hybrid Clusters Lead to Selective Cell Adhesion. J. Am. Chem. Soc. 2009, 131 (4), 1340-+. 76. Kanoufi, F.; Proust, A., Electrografting of Diazonium-Functionalized Polyoxometalates: Synthesis, Immobilisation and Electron-Transfer Characterisation from Glassy Carbon. Chemistry-a European Journal 2013, 19 (41), 13838-13846. 77. Bregolin, F.; Lapertot, G.; Dubois, L.; Proust, A.; Jansen, L.; Duclairoir, F., Evidence for Charge Transfer at the Interface between Hybrid Phosphomolybdate and Epitaxial Graphene. Langmuir 2016, 32 (19), 4774-4783. 78. F.; Proust, A., Electron Transfer to a Phosphomolybdate Monolayer on Glassy Carbon: Ambivalent Effect of Protonation. Inorg. Chem. 2016, 55 (14), 6929-6937. 79. Combellas, C., Control of the Grafting of Hybrid Polyoxometalates on Metal and Carbon Surfaces: Toward Submonolayers. Langmuir 2014, 30 (8), 2287-2296. 80. Yaqub, M.; Walsh, J. J.; Keyes, T. E.; Proust, A.; Rinfray, C.; Izzet, G.; McCormac, T.; Forster, R. J., Electron Transfer to Covalently Immobilized Keggin Polyoxotungstates on Gold. Langmuir 2014, 30 (15), 4509-4516. 81. A., Electron transfer properties of a monolayer of hybrid polyoxometalates on silicon. J. Mater. Chem. C 2015, 3 (24), 6266-6275. 82. Lenfant, S.; Proust, A., Molecular signature of polyoxometalates in electron transport of silicon-based molecular junctions. Nanoscale 2018, 98, 219. 83. electrocatalytic behaviour of the crown heteropolyanion K28Li5H7P8W48O184 center dot 92H(2)O. Electrochem. Commun. 2000, 2 (10), 720-726. 84. preparation and characterization of silicotungstic heteropolyanion monolayer electrostatically linked aminophenyl on carbon electrode surface. Langmuir 1999, 15 (21), 7268-7275. Laurans, M.; Dalla Francesca, K.; Volatron, F.; Izzet, G.; Guerin, D.; Vuillaume, D.; Rinfray, C.; Brasiliense, V.; Izzet, G.; Volatron, F.; Alves, S.; Combellas, C.; Kanoufi, Derouich, S. G.; Rinfray, C.; Izzet, G.; Pinson, J.; Gallet, J. J.; Kanoufi, F.; Proust, A.; Keita, B.; Lu, Y. W.; Nadjo, L.; Contant, R., Salient electrochemical and Liu, S. Q.; Tang, Z. Y.; Shi, Z.; Niu, L.; Wang, E. K.; Dong, S. J., Electrochemical Volatron, F.; Noel, J. M.; Rinfray, C.; Decorse, P.; Combellas, C.; Kanoufi, F.; Proust, 24 Clement, N.; Patriarche, G.; Smaali, K.; Vaurette, F.; Nishiguchi, K.; Troadec, D.; Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.; Whitesides, G. M., Self- Salomon, A.; Cahen, D.; Lindsay, S.; Tomfohr, J.; Engelkes, V.; Frisbie, C., Khattak, G. D.; Salim, M. A.; AlHarthi, A. S.; Thompson, D. J.; Wenger, L. E., Contant, R.; Teze, A., A New Crown Heteropolyanion, K28li5h7p8w48o184.92h2o - Smaali, K.; Clement, N.; Patriarche, G.; Vuillaume, D., Conductance Statistics from a 85. 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 (14), 4336-4349. 86. Fujiwara, A.; Vuillaume, D., Large Array of Sub-10-nm Single-Grain Au Nanodots for use in Nanotechnology. Small 2011, 7 (18), 2607-2613. 87. Large Array of Sub-10 nm Molecular Junctions. ACS Nano 2012, 6 (6), 4639-4647. 88. Bedford, E.; Humblot, V.; Methivier, C.; Pradier, C. M.; Gu, F.; Tielens, F.; Boujday, S., An Experimental and Theoretical Approach to Investigate the Effect of Chain Length on Aminothiol Adsorption and Assembly on Gold. Chem. Eur. J. 2015, 21 (41), 14555-14561. 89. Vericat, C.; Vela, M. E.; Benitez, G.; Carro, P.; Salvarezza, R. C., Self-assembled monolayers of thiols and dithiols on gold: new challenges for a well-known system. Chem. Soc. Rev. 2010, 39 (5), 1805-1834. 90. assembled monolayers of thiolates on metals as a form of nanotechnology. Chem. Rev. 2005, 105 (4), 1103-1169. 91. Comparison of Electronic Transport Measurements on Organic Molecules. Advanced Materials 2003, 15 (22), 1881-1890. 92. Synthesis, Structure, and Properties. Inorg. Chem. 1985, 24 (26), 4610-4614. 93. Structure of molybdenum-phosphate glasses by X-ray photoelectron spectroscopy (XPS). J. Non-Cryst. Solids 1997, 212 (2-3), 180-191. 94. Scale Junctions: Histograms and Beyond. Acc. Chem. Res. 2016, 49 (3), 452-460. 95. Motivation and Challenges. Chem Rev 2017, 117 (5), 4248-4286. 96. Trasobares, J.; Rech, J.; Jonckheere, T.; Martin, T.; Aleveque, O.; Levillain, E.; Diez- Cabanes, V.; Olivier, Y.; Cornil, J.; Nys, J. P.; Sivakumarasamy, R.; Smaali, K.; Leclere, P.; Fujiwara, A.; Theron, D.; Vuillaume, D.; Clement, N., Estimation of π-π Electronic Couplings from Current Measurements. Nano Letters 2017, 17 (5), 3215 -- 3224. 97. Effects and Transport Mechanisms in Conductance Histograms. Nano Letters 2012, 12 (5), 2243-2248. 98. Smaali, K.; Desbief, S.; Foti, G.; Frederiksen, T.; Sanchez-Portal, D.; Arnau, A.; Nys, J. P.; Leclere, P.; Vuillaume, D.; Clement, N., On the mechanical and electronic properties of thiolated gold nanocrystals. Nanoscale 2015, 7 (5), 1809-1819. 99. thin insulating film. J. Appl. Phys. 1963, 34 (9), 2581-2590. 100. Beebe, J. M.; Kim, B.; Gadzuk, J. W.; Frisbie, C. D.; Kushmerick, J. G., Transition from Direct Tunneling to Field Emission in Metal-Molecule-Metal Junctions. Physical Review Letters 2006, 97 (2), 026801. 101. Beebe, J. M.; Kim, B.; Daniel Frisbie, C.; Kushmerick, J. G., Measuring relative barrier heights in molecular electronic junctions with transition voltage spectroscopy. ACS Nano 2008, 2 (5), 827-832. 102. Huisman, E. H.; Guédon, C. M.; van Wees, B. J.; van der Molen, S. J., Interpretation of Transition Voltage Spectroscopy. Nano Letters 2009, 9 (11), 3909-3913. Hybertsen, M. S.; Venkataraman, L., Structure-Property Relationships in Atomic- Vilan, A.; Aswal, D.; Cahen, D., Large-Area, Ensemble Molecular Electronics: Simmons, J. G., Electrical tunnel effect between disimilar electrodes separated by a Reuter, M. G.; Hersam, M. C.; Seideman, T.; Ratner, M. A., Signatures of Cooperative 25 103. Bâldea, I., Ambipolar transition voltage spectroscopy: Analytical results and experimental agreement. Physical Review B 2012, 85 (3), 035442. 104. Ricoeur, G.; Lenfant, S.; Guerin, D.; Vuillaume, D., Molecule/Electrode Interface Energetics in Molecular Junction: A "Transition Voltage Spectroscopy" Study. Journal of Physical Chemistry C 2012, 116 (39), 20722-20730. 105. Cuevas, J. C.; Scheer, E., Molecular Electronics: An introduction to theory and experiment. World Scientific: 2010. 106. Grüter, L.; Cheng, F.; Heikkilä, T. T.; González, M. T.; Diederich, F.; Schönenberger, C.; Calame, M., Resonant tunnelling through a C 60molecular junction in a liquid environment. Nanotechnology 2005, 16 (10), 2143-2148. 107. Zotti, L. A.; Cuevas, J.-C.; Pauly, F.; Huhn, T.; Scheer, E.; Erbe, A., Revealing the Role of Anchoring Groups in the Electrical Conduction Through Single-Molecule Junctions. Small 2010, 6 (14), 1529-1535. 108. Chen, J.; Markussen, T.; Thygesen, K. S., Quantifying transition voltage spectroscopy of molecular junctions. Physical Review B 2010, 82 (12), 121412. 109. Vuillaume, D.; Boulas, C.; Collet, J.; Allan, G.; Delerue, C., Electronic structure of a heterostructure of an alkylsiloxane self-assembled monolayer on silicon. Physical Review B 1998, 58 (24), 16491-16498. 110. Wang, W.; Lee, T.; Reed, M. A., Mechanism of electron conduction in self-assembled alkanethiol monolayer devices. Phys. Rev. B 2003, 68, 035416. 111. Vilan, A., Analyzing Molecular Current-Voltage Characteristics with the Simmons Tunneling Model: Scaling and Linearization. J. Phys. Chem. C 2007, 111 (11), 4431-4444. 112. Vilan, A.; Cahen, D.; Kraisler, E., Rethinking Transition Voltage Spectroscopy within a Generic Taylor Expansion View. ACS Nano 2013, 7 (1), 695-706. 113. Qi, Y.; Yaffe, O.; Tirosh, E.; Vilan, A.; Cahen, D.; Kahn, A., Filled and empty states of alkanethiol monolayer on Au (111): Fermi level asymmetry and implications for electron transport. Chemical Physics Letters 2011, 511 (4-6), 344-347. 114. Scofield, J. H., Hartree-Slater Subshell Photoionization Cross-Sections at 1254 and 1487ev. J. Electron. Spectrosc. Relat. Phenom. 1976, 8 (2), 129-137. 115. Wold, D. J.; Frisbie, C. D., Fabrication and characterization of metal-molecule-metal junctions by conducting probe atomic force microscopy. J. Am. Chem. Soc. 2001, 123, 5549- 5556. 116. Engelkes, V. B.; Daniel Frisbie, C., Simultaneous nanoindentation and electron tunneling through alkanethiol self-assembled monolayers. The Journal of Physical Chemistry B 2006, 110 (20), 10011-10020. 26 Charge transport through redox active [H7P8W48O184]33- polyoxometalate self-assembled onto gold surfaces and gold nanodots. K. Dalla Francesca,1 S. Lenfant,2 M. Laurans,1 F. Volatron,1 G. Izzet,1 V. Humblot,3 C. Methivier,3 D. Guerin,2 A. Proust,1,* D. Vuillaume2,* 1) Sorbonne Université, CNRS, Institut Parisien de Chimie Moléculaire, IPCM, 4 Place Jussieu, F-75005 Paris, France. 2) Institute for Electronics Microelectronics and Nanotechnology (IEMN), CNRS, Lille Univ., Av. Poincaré, F-59652 Villeneuve d'Ascq, France. 3) Sorbonne Université, CNRS, Laboratoire de réactivité de surface, LRS, 4 Place Jussieu, F-75005 Paris, France. * Corresponding authors: [email protected], [email protected] SUPPORTING INFORMATION 5.87 nm 0.88 nm 1. Formation of POM aggregates in the absence of PBS rinsing. Figure S1. 2x2 µm and 1x1 µm AFM images and height profiles of a slightly rinsed AOT modified gold substrate (without PBS treatment) immersed in a 1 mM solution of H3[PW12O40] in water. The POMs tend to cling to the physisorbed thiol molecules to form bigger aggregates (around 2.5 nm according to the line profiles recorded on the aggregates). 100nm 100 X[nm] ] m n [ Z X[nm] ] m n [ Z 300 3 2 1 0 100 150 200 250 150 0.5 3 2 1 0 0 50 2.5 3.5 2.5 1.5 0 50 1.5 0.5 to obtain because of 2. POM incubation The samples were prepared by incubation of the AOT functionalized gold substrate in a 1 mM POM solution for one hour. We have varied the incubation time from 20 mn to 1h and 24h without observing any significant modification of the layer thickness as follows by ellipsometry. Regarding the POM solution concentration, we have started with 1mM which is the concentration often used. As the 0.8 nm increase of the thickness was slightly lower than expected (POM thickness of about 1.0 nm) we did not try to lower the POM solution concentration but rather we tried to increase it. Unfortunately, we were limited by the low solubility of the POM and longer immersion times (until 24 h) led to exactly the same layer features and significantly more concentrated solutions were not possible the K28Li5[H7P8W48O184] in water, even at higher LiCl concentrations. 3. Nanodots Figure S2 shows the scanning electron microscope images of the nanodot array after the fabrication. We clearly observed a large and regular array of 10 nm in diameter nanodots separated by 100 nm. The fact that only a fraction (around 3 %) of the junctions are electrically active (bright spots in figure 5-a in the main text) may have several origins: nanodots removed during the chemical treatments or removed during the C-AFM measurements (if not well embedded in the Si substrate). We can discard the removal during the C-AFM measurements since we have not observed a significant variation of the number of active NMJs during successive scanning of the samples to measure the current at several voltages as shown in the current histograms (see below figure S3). It is likely that the nanodots were not well embedded onto the substrate for the present case and were removed during the grafting process in solution. the low solubility of Figure S2. Scanning electron microscope images of the nanodots arrays at two magnifications. 3. Current histograms at different voltages measured on NMJs. Figure S3. Histograms of the currents measured on NMJs at several voltages (as indicated in the x-axis legend). The black lines are fit with a log-normal distribution. Mean values and standard deviations are given in the table S1. Applied Voltage (V) -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 < log I > (log A) -7.05 -7.41 -7.78 -8.20 -9.59 -8.35 -7.46 -7.29 log σ 0.66 0.74 0.64 0.6 0.72 1.2 0.90 1.50 Table S1. Mean values and standard deviations of the log-normal distributions fitted on the histograms of the measured currents shown in Fig. S2. 5. I-V curves adjustment with the Simmons model. The expression of the tunnel current through a potential barrier is given by Simmons:1 I(V) = S e 4πhd2 ⎡ ⎢ ⎢ ⎣ 2ΦT − eV ( ⎛ )exp − ⎜⎜ ⎝ 4πd m 2ΦT − eV ( h ) ⎞ ⎟⎟ ⎠ − 2ΦT + eV ( ⎛ )exp − ⎜⎜ ⎝ 4πd m 2ΦT + eV ( h ) ⎞ ⎟⎟ ⎠ ⎤ ⎥ ⎥ ⎦ (S1) with e the electron charge, h the Planck's constant, d the thickness of the tunneling barrier, ΦT the energy barrier height, V the voltage applied across the junction, m the effective mass of the electron, I the current and S the electrical contact surface area (smaller than the geometrical contact area, due to defects in the SAM and on the substrate, e.g. roughness, grain boundaries,…).2-3 The electron effective mass m is m = mr m0 with m0 the mass of the electron and mr the reduced mass. Adjustments of the measured I-V curves are systematically done for the positive and negative bias. The parameter d corresponding to the total thickness of the monolayer determined by ellipsometry (1.3 nm, see main text) is fixed and the three other parameters (ΦT, mr and S) are the fitting parameters. Figure S4 gives 3 examples of the fits for SAMs on TSAu surfaces. The histograms for ΦT , mr and S are given in Figs. 6 (main text), S5 and S6, respectively. The average S value corresponds to a C-AFM tip electrical contact with a diameter of about 0.45 nm. Albeit the real value of the C-AFM tip contact area is difficult to estimate (it depends on the loading force, Young modulus of the SAM)4 this value seems reasonable considering that the tip has a radius of 20 nm and the low loading force (6nN). 6. Transition voltage spectroscopy (TVS). The I−V curves are analyzed by the TVS method.5-7 In brief, the I− V data are plotted as Ln(I/V2) vs. 1/V. A minimum in this curve corresponds to a transition from a direct tunneling electron transport through the molecules to a resonant tunneling via a frontier molecular orbital (LUMO or HOMO). The energy position ε of the orbital involved in the transport mechanism with respect to the Fermi energy of the metal electrode is given by : ε= 2 e VT +VT − 2 +10 VT +VT − / 3+VT − VT + 2 (S2) with e the electron charge, VT+ and VT− the voltage of the minima of the TVS plot at positive and negative voltages, respectively.8 Eq. S2 reduces to: ε = 0.86 e VT (S3) when the I-V curves are symmetric with respect of the applied voltages (our case here) and VT+=VT-. Figure S4 shows 3 examples of TVS plots (on the same IV curves fitted with the Simmons equation). The corresponding histograms of ε are given in Fig. 6 (main text). 7. Molecular single energy level model. A simple analytical model to describe electron transport through a molecular junction is the single energy level model.9-13 This model is based on the following assumptions: (i) the transport is phase coherent (tunneling mechanism), (ii) the current is dominated by a single energy level, ε0, of the molecule (HOMO or LUMO) in the voltage range considered to fit the experimental I-V curves, and (iii) the voltage drops exclusively over the contacts which are described by the coupling constants g1 and g2 . 2ε+ eV g1 − g2 g1 + g2 2 g1 + g2 ) ( − eV ⎞ ⎟ ⎟ ⎟ ⎠ ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ (S4) I(V) = N 8e h g1g2 g1 + g2 arctan ⎛ ⎜ ⎜ ⎜ ⎝ ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ 2ε+ eV g1 − g2 g1 + g2 2 g1 + g2 ) ( + eV ⎞ ⎟ ⎟ ⎟ ⎠ − arctan ⎛ ⎜ ⎜ ⎜ ⎝ with h the Planck's constant and N the number of molecules in the junctions. The fitting parameters are ε, g1 and g2. Since the exact number of molecules in the junction is not known (albeit weak, e.g. <25 in the NMJs, see main text) we fixed N=1, but we have observed that the exact value of N does not have a drastic influence on ε, only on the coupling parameters g1 et g2. Albeit the obtained values seem reasonable (in the range 0.1 - few tens of meV), it is known that these fitting parameters are also very sensitive to the I-V curve "quality" (e.g. noise, sudden jumps in the IV,…),13 while the value of ε is not, consequently we have not discussed these values in more detail. Figure S4 shows 3 examples of these fits on the same IVs (SAMs on TSAu) also analyzed with the Simmons equation and the TVS method. The corresponding histogram of ε0 is shown in Fig. 6 (main text). ) A ( I ) A ( I ) A ( I 0.5 1.0 1.5 -1.5 -1.0 -0.5 0.0 V (V) Experimental Data Molecular Model ε0 = 0.94±0.01 eV Simmons Model ΦT = 0.99±2.5 eV Experimental Data Molecular Model ε0 = 0.73±0.01 eV Simmons Model ΦT = 0.98±1.6 eV 6x10-9 4x10-9 2x10-9 0 -2x10-9 -4x10-9 -6x10-9 4x10-9 3x10-9 2x10-9 1x10-9 0 -1x10-9 -2x10-9 -3x10-9 3x10-9 2x10-9 1x10-9 0 -1x10-9 -2x10-9 -3x10-9 Figure S4. Typical examples of the fit (Simmons and molecular models) of 3 different I-V curves measured by C-AFM on the SAM on TSAu substrates (left), and corresponding plot of the TVS method (right). Experimental Data Molecular Model ε0 = 0.85±0.01 eV Simmons Model ΦT = 0.76±0.01 eV 0.0 V (V) 0.0 V (V) -1.5 -1.0 -0.5 -1.5 -1.0 -0.5 1.0 0.5 1.0 1.5 0.5 1.5 Figure S5. Histograms of the effective reduced mass from Simmons model fitted on I-V curves measured for POM SAMs on TSAu electrodes. Black curve: Gaussian fit. Figure S6. Histograms of the electrical contact area from Simmons model fitted on I-V curves measured for POM SAMs on TSAu electrodes. Black curve: Gaussian fit. 8. Application of the Simmons and molecular models on I-V measured on NMJs. Figure S7 shows a typical example of the Simmons model and molecular model fitted on the same I-V curve measured on NMJ. For the Simmons model, the thickness was fixed at 1.3 nm as for SAM on TSAu since it is not easy to measure the POM thickness on the Au nanodots.14 ) A ( t n e r r u C 6x10-8 4x10-8 2x10-8 0 -2x10-8 -4x10-8 -6x10-8 -8x10-8 -1x10-7 Experimental Data Molecular Model ε0 = 0.24±0.02 eV Simmons Model ΦT = 0.26±0.33 eV -0.4 -0.3 -0.2 -0.1 0.0 Voltage (V) 0.1 0.2 0.3 Figure S7. I-V curve (same data as in Fig. 5-c, mean current vs. voltages from histograms in Fig. S2) measured on POM NMJ, and fits with Simmons equation and molecular model. 9. Characterization of [H7P8W48O184]33- ) A µ ( I 4 2 0 -2 -4 -6 -8 -10 -12 -14 -1,0 -0,8 -0,6 -0,4 -0,2 0,0 0,2 0,4 E (V) Figure S8. Cyclic voltammogram of K28Li5[H7P8W48O184] (10-4 M) in CH3COOH/CH3COOLi 0.1M in water at a scan rate 10 mV.s-1. Working electrode = glassy carbon, reference = calomel saturated electrode (SCE), counter-electrode = platinum wire. It is compliant with those published in by Keita et al.15 ) A µ ( I 0,5 0,0 -0,5 -1,0 -1,5 -2,0 -2,5 -3,0 -3,5 -0,8 -0,6 -0,4 -0,2 E (V) 0,0 0,2 Figure S9. Cyclic voltammogram of K28Li5[H7P8W48O184] (10-4 M) in CH3COOH/CH3COOLi 0.1M in water at a scan rate 10 mV.s-1. Working electrode = Au microelectrode, reference = calomel saturated electrode (SCE), counter-electrode = platinum wire. Figure S10. 31P (121 MHz) NMR spectrum of K28Li5[H7P8W48O184] in LiCl 1M / D2O REFERENCES. 1. Simmons, J. G., Electrical tunnel effect between disimilar electrodes separated by a thin insulating film. J. Appl. Phys. 1963, 34 (9), 2581- 2590. Simeone, F. C.; Yoon, H. J.; Thuo, M. M.; Barber, J. R.; Smith, B.; Whitesides, G. M., Defining the Value of Injection Current and Effective Electrical Contact Area for EGaIn-Based Molecular Tunneling Junctions. J. Am. Chem. Soc. 2013, 135 (48), 18131-18144. Sangeeth, C. S. S.; Wan, A.; Nijhuis, C. A., Equivalent Circuits of a Self- Assembled Monolayer-Based Tunnel Junction Determined by Impedance Spectroscopy. J. Am. Chem. Soc. 2014, 136 (31), 11134-11144. Engelkes, V. B.; Daniel Frisbie, C., Simultaneous nanoindentation and electron tunneling through alkanethiol self-assembled monolayers. The Journal of Physical Chemistry B 2006, 110 (20), 10011-10020. Beebe, J. M.; Kim, B.; Gadzuk, J. W.; Daniel Frisbie, C.; Kushmerick, J. G., Transition from Direct Tunneling to Field Emission in Metal-Molecule- Metal Junctions. Phys. Rev. Lett. 2006, 97 (2), 4. Huisman, E. H.; Guédon, C. M.; van Wees, B. J.; van der Molen, S. J., Interpretation of Transition Voltage Spectroscopy. Nano Letters 2009, 9 (11), 3909-3913. Ricoeur, G.; Lenfant, S.; Guerin, D.; Vuillaume, D., Molecule/Electrode Interface Energetics in Molecular Junction: A "Transition Voltage Spectroscopy" Study. Journal of Physical Chemistry C 2012, 116 (39), 20722-20730. Bâldea, I., Ambipolar transition voltage spectroscopy: Analytical results and experimental agreement. Physical Review B 2012, 85 (3), 035442. Datta, S., Electrical resistance: an atomistic view. Nanotechnology 2004, 15 (7), S433. Grüter, L.; Cheng, F.; Heikkilä, T. T.; González, M. T.; Diederich, F.; Schönenberger, C.; Calame, M., Resonant tunnelling through a C 60molecular junction in a liquid environment. Nanotechnology 2005, 16 (10), 2143-2148. Cuevas, J. C.; Scheer, E., Molecular Electronics: An Introduction to Theory and Experiment. World Scientifiic: 2010. Zotti, L. A.; Cuevas, J.-C.; Pauly, F.; Huhn, T.; Scheer, E.; Erbe, A., Revealing the Role of Anchoring Groups in the Electrical Conduction Through Single-Molecule Junctions. Small 2010, 6 (14), 1529-1535. Brunner, J. Gaining Microscopic Insight into Molecular Junctions by Transport Experiments. Basel University, 2013. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. Smaali, K.; Clement, N.; Patriarche, G.; Vuillaume, D., Conductance Statistics from a Large Array of Sub-10 nm Molecular Junctions. ACS Nano 2012, 6 (6), 4639-4647. Keita, B.; Lu, Y. W.; Nadjo, L.; Contant, R., Salient electrochemical and electrocatalytic behaviour of the crown heteropolyanion K28Li5H7P8W48O184·92H2O. Electrochemistry Communications 2000, 2 (10), 720-726. 14. 15.
1806.00083	1	1806	2018-05-23T20:31:28	Topological and Network Analysis of Lithium Ion Battery Components: The Importance of Pore Space Connectivity for Cell Operation	[ "physics.app-ph", "cond-mat.mtrl-sci", "eess.IV" ]	The structure of lithium ion battery components, such as electrodes and separators, are commonly characterised in terms of their porosity and tortuosity. The ratio of these values gives the effective transport of lithium ions in the electrolyte-filled pore spaces, which can be used to determine the ionic resistivity and corresponding voltage losses. Here, we show that these microstructural characteristics are not sufficient. Analysis of tomographic data of commercial separators reveals that different polyolefin separators have similar porosity and through-plane tortuosity, which, in the homogenised picture of lithium ion cell operation, would imply that these different separators exhibit similar performance. However, numerical diffusion simulations indicate that this is not the case. We demonstrate that the extent to which lithium ion concentration gradients are induced or smoothed by the separator structure is linked to pore space connectivity, a parameter that can be determined by topological or network based analysis of separators. These findings enable us to propose how to design separator microstructures that are safer and accommodate fast charge and discharge.	physics.app-ph	physics	Topological and Network Analysis of Lithium Ion Battery Components: The Importance of Pore Space Connectivity for Cell Operation M. F. Lagadec,1 R. Zahn,1 S. Müller1 and V. Wood1 1 Department of Information Technology and Electrical Engineering, ETH Zurich, Zurich CH-8092, Switzerland. The structure of lithium ion battery components, such as electrodes and separators, are commonly characterised in terms of their porosity and tortuosity. The ratio of these values gives the effective transport of lithium ions in the electrolyte-filled pore spaces, which can be used to determine the ionic resistivity and corresponding voltage losses. Here, we show that these microstructural characteristics are not sufficient. Analysis of tomographic data of commercial separators reveals that different polyolefin separa- tors have similar porosity and through-plane tortuosity, which, in the homogenised picture of lithium ion cell operation, would imply that these different separators exhibit similar performance. However, numerical diffusion simulations indicate that this is not the case. We demonstrate that the extent to which lithium ion concentration gradients are induced or smoothed by the separator structure is linked to pore space connectivity, a parameter that can be determined by topological or network based analysis of separators. These findings enable us to propose how to design separator microstructures that are safer and accommodate fast charge and discharge. The structures of components in a lithium ion battery (LIB), such as the electrodes and the separator, influence lithium ion transport1 and therefore play an important role in dictating the cell performance metrics such as (dis)charge-rate dependent capacity and cycle life2. In the homogenised picture of cell operation used in 1D models3–5 that dominate cell modelling today (e.g., Dual- foil6 and COMSOL Multiphysics7), the diffusion coefficient of the cations (D+) and the anions (D–) in the electrolyte- filled pore space is given by their diffusion in a bath of electrolyte scaled by the effective transport coefficient of the microstructure,8 δTP = ε / τTP, where ε is the porosity and τTP is the tortuosity along the through-plane (TP) di- rection between the current collectors. A low effective transport coefficient leads to a low ionic diffusivity and therefore a low ionic conductivity (σ ≈ c(cid:1)(D+ + D–), where c is the concentration of the salt in the electrolyte), which in turn results in large voltage drops (i.e., large overpo- tentials) across the electrolyte-filled pore space9. At fast operation speeds (e.g., at the end of a 5C discharge),10 these overpotentials can account for ~60 % of cell over- potentials, outweighing the contributions of the charge transference resistance at the electrodes. However, this volume-averaged effective transport in the homogenised picture does not account for inhomogenei- ties across the cells. Inhomogeneities lead to incomplete capacity extraction, lithium plating, and hot spots where current preferentially flows.11–14 While inhomogeneities can be determined by running simulations over real 3D microstructures8,11 or statistically assessing many sub- volumes of an imaged microstructure15, these analyses do not tell us about how the structure itself may give rise to or how good it is at compensating for inhomogeneities. In this article, we propose a new approach to character- ise microstructure of lithium ion battery components based on topological and network analysis. We show that this analysis captures how a structure induces or homog- enises ion gradients. While topological analysis of porous media is commonly used in soil physics and geology16,17, it has not previously been applied to the LIB field. Linked to topology is net- work theory, which describes the types of connections that exist in a system that can be characterised by branches and nodes (i.e., points where branches inter- sect). Network analysis has been applied in a wide vari- ety of fields including information and communication (e.g., the world-wide web), energy (e.g., power grids), and biology (e.g., metabolic networks),18 but not to de- scribe the pore space of a battery, which can also be viewed as a network. Here, we show that parameters that can be calculated from topological and network analysis of 3D microstruc- tures, such as pore space connectivity density and per- cent of dead end pores, are important for predicting cell performance and safety. As a case study, we look at lithium ion battery separators. We show that two separators of strikingly different mor- phology have similar TP effective transport coefficients, suggesting that both separators would exhibit similar lith- ium ion transport. However, 3D diffusion simulations highlight that lithium ion transport occurs in different ways in the separator structures. We show that the differences in pore space topology and network properties of the two separators can explain the different transport properties in the separators, particularly the tendency of a structure to allow or prevent lithium ion concentration gradients. These parameters can be used to optimise separator se- lection for a given cell and to guide design of next gener- ation separators. Microporous polyolefin membranes have been used as separators in LIBs for several decades, and have been manufactured with a variety of thicknesses, pore struc- tures, and surface chemistries.19,20 Recently, we have shown that it is possible to obtain quantitative reconstruc- tions of LIB separators using focus-ion-beam scanning electron microscope (FIB-SEM) tomography.15,21 3D mi- crostructure renderings of polyethylene (PE)22 and 1 polypropylene (PP)23 separators obtained using this ap- proach are shown in Figure 1. Figure 1. SEM top view image recorded in secondary electron mode, and 3D microstructure renderings of (a) Targray PE16A and (b) Celgard® PP1615 separators of sub-volumes of 3 µm edge length, imaged as described by Lagadec et al.15 The PE and PP separators exhibit distinct morphologies that stem from the different processes used to manufac- ture them. The PE separator (Figure 1a) microstructure is isotropic,15 while that of PP is anisotropic21 (Figure 1b). However, the respective porosities, ε, TP tortuosities, τTP, and thus the effective transport coefficients8, δTP = ε / τTP, of the PE and PP microstructures are similar (Table I). Topological and network analysis of these structures pro- vides a set of parameters with which to quantify separa- tors. Here, we provide a brief introduction of these pa- rameters for readers unfamiliar with morphological or net- work descriptors and illustrate simple cases in Figure 2. The topological invariant (Euler-Poincaré characteristic, see sections 2-3 in the ESI)24, X, describes an object's shape and structure independent of how it is bent and relates to the object's connectivity,16,25 which is a concept from topology and network theory. The skeleton of the structure (i.e., pink lines in Figure 2) can be used to an- alyse the separator as a network. Χ of a given pore network is N-C†, where N is the number of unique pores, and C is their connectivity, which is defined as the number of cuts needed to obtain a simply connected network (i.e., without redundant connec- tions).26 From network analysis, C is also defined as the number of branches minus the end point branches (i.e., dead-end branches connected only to a single node) mi- nus number of nodes plus 1. To illustrate this connection between the topological in- variant, X, and the connectivity, C, we consider two cases. In Figure 2a, we have N unconnected pores. C is zero and X is positive (X = N). In the case of a single pore network in Figure 2b, N = 1 and C = 2 (the two redundant connections are marked with cyan cuts) such that X is negative (X = 1 - 2 = -1). Alternatively, we see that there are 11 branches, 5 end points, and 5 nodes, also giving C = 2. A more connected network (higher C) implies a more negative X (Figure 2c). From network analysis, we additionally consider the node density, the number of nodes of different order (the order is given by the number of branches connected to the node), the number of end point branches, and the aver- age branch length. Figure 2. Schematics of pore space and pore space skeleton- isation of (a) unconnected single objects with connectivity C = 0, and of (b) single, interconnected network (redrawn from DeHoff et al.27 with branches and nodes) connectivity C = 2. (c) Relationship between Euler-Poincaré characteristic, Χ, and connectivity, C. Table I. Porosity, ε, tortuosity, τ, and effective transport coefficient, δ, for representative volume elements of Targray PE16A 22 (PE) of 2 µm edge length and of Celgard® PP1615 23 (PP) of 3 µm edge length. The tortuosity values are obtained from Fickian diffusion simulations across the pore phase in both in-plane directions (IP1 and IP2) and in the TP direction. Parameter Porosity ε [%] Tortuosity τ [-] Effective transport coefficient δ [%] PE 40.82±1.92 2.99±0.39 2.65±0.31 2.64±0.21 13.9±2.2 15.7±2.2 15.6±1.9 τIP1 τIP2 τTP δIP1 δIP2 δTP 2 PP 40.19±1.03 2.31±0.24 24.89±6.15 2.04±0.19 17.6±2.1 1.7±0.4 19.9±2.0 In the example in Figure 2b, we have one node of order 5 and four nodes of order 3. There are 5 end point branches, but, for our purposes, because we only work with a sub-volume of a separator, we count only those that end within the structure as end-point branches (dashed line). Since the real separator structures are complex, we work with computer-generated, idealised structures as well as with the imaged PE and PP structures. We generate structures (Figure 3) with cylindrical pores in 1, 2, and 3 directions using the algorithm described in section 4 of the ESI. For each type of structure, 3 entities are generated and the reported properties are the aver- age values. The pore (i.e., cylinder) diameter is compa- rable to the geometrical pore size, D50, of the PE separa- tor, and their porosity, ε, is set to be within 40±2 %, which is comparable to the porosity of the PE and PP separa- tors. These cubic datasets have an edge length of 5 µm and an isotropic voxel length of 10 nm. Figure 3. Example 3D microstructure renderings of artificial separator microstructures with randomly distributed, cylindrical pores of diameter 0.13 µm and edge length 3 µm in (a) one, (b) two, and (c) three dimensions. For the three computer generated reference structures as well as the imaged PE and PP, we calculate X using the MATLAB code by Legland et al.25 For the skeleton analysis, the datasets are symmetrically eroded using the 3D thinning algorithm of the Skeletonize 3D plugin in Im- ageJ. The resulting 3D skeletons are evaluated using Im- ageJ's AnalyzeSkeleton (2D/3D) plugin to assess the number of branches, nodes, and end-points, and the or- der of nodes as described in sections 5-6 of the ESI. To determine the proportions of node orders, the inter-tra- becular angle calculation program by Reznikov et al.28 is used. We normalise X and C by dividing them by the an- alysed microstructure volume, V, which gives the corre- sponding densities χ and c. For unconnected, cylindrical pores in one dimension, we obtain zero connectivity density c and a positive value for χ (7.23 µm-3), which corresponds to n, the number of pores per unit volume V, since χ = n - c. For a single, connected pore network, n is given by N/V (N = 1 and V = 125 µm3 yielding n = 0.008 µm-3); therefore, χ and c are almost identical in magnitude but of opposite sign. For interconnected pores in two directions with N = 1, χ becomes negative (χ = -102.71 µm-3, X = -12838.33) and the connectivity density, c, becomes positive (c = 102.71 µm-3, C = 12839.33), indicating that the number density of redundant connections in the pore network has in- creased. The PE microstructure is more connected (Table II; c = 143.16 µm-3, C = 17894.67) than the reference micro- structure with pores in 3 directions (c = 117.37 µm-3, C = 14671.00), which has comparable geometrical pore ra- dius D50. In contrast, the PP separator exhibits a relatively small negative χ (χ = -7.43 µm-3, X = -929.00) and a low connectivity (c = 7.44 µm-3, C = 930.00), which can be understood by noting its straight pores with few redun- dant connections. To further understand these trends in connectivity, we systematically analyse the proportions of node order, node and branch densities, percentage of end point branches, and average branch length (Table III). The ref- erence separator pore networks with pores in two and three dimensions have a similar fraction of nodes of order 3-6 (section 6 in the ESI) and, as designed, zero end point branches within the volume. As larger numbers of perpendicular pore channels are introduced, the node and branch densities increase and the average branch length decreases. For the reference datasets, the pore dead-ends all are at the dataset's boundaries, whereas for the measured datasets, the pore dead-ends also ap- pear within the volume. Table II. Average values and standard deviations of porosity, topological invariant density, χ, and connectivity density, c, for the artificially gener- ated microstructures (1D, 2D and 3D) and the imaged Targray PE16A (PE) and Celgard® PP1615 (PP) separator microstructures of edge lengths 5 µm each. The values for χ and c are calculated via the Minkowski functional, M3. Parameter Porosity [%] Topological invariant density χ [µm-3] Connectivity density c [µm-3] 1D 39.95±0.00 7.23±0.00 0.00±0.00 2D 40.48±0.07 -102.71±0.16 102.71±0.16 3D 41.04±0.04 -117.36±1.15 117.37±1.15 PE 40.53±0.77 -143.15±6.88 143.16±6.88 PP 40.19±0.42 -7.43±0.51 7.44±0.51 3 Consistent with its low connectivity, PP exhibits a lower node density and a larger average branch length than the PE separator. Analysis of the pore orientations (section 7 in the ESI) indicates that PP contains straight pores, while pores in the PE separator are also angled relative to one another. This difference in how pores are con- nected in PE and PP separators is further revealed by the different fractions of node orders. PE and PP exhibit ra- tios of 81:15 and 90:9, respectively for nodes of orders 3 and 4. The larger number of higher order nodes com- bined with the larger connectivity in PE compared to PP separator reflects the high redundancy of connections between nodes and higher spreading power. Thus, on a device level, transport through the separator pore net- work remains unchanged even if some pores are blocked. Finally, the PP separator exhibits a larger per- centage of end point branches within the volume (31.55 %) than the PE separator (9.07 %), see section 8 in the ESI. To understand the impact of these structural differences on battery performance, we perform steady-state Fickian diffusion simulations on the artificial and measured sep- arator structures. We use cubic datasets of 3 µm edge length and iteratively calculate the solution of the Poisson equation on the electrolyte domain of the input geometry along the TP direction. We use Dirichlet boundary condi- tions at the end planes orthogonal to the TP direction and zero flux Neumann boundary conditions on all other boundaries and on the separator surfaces.29,30 Inlet and outlet concentrations of 1.25 and 1.20 M are chosen based on the COMSOL simulation of C-rate dependence of electrolyte salt concentration for Li0separatorLTO cells with Targray PE16A separator shown in the ESI. If only the TP effective transport coefficient δTP calculated across a volume were considered, similar results would be would be obtained for the PE and PP separators as such calculations result in an overall value without local resolution. The effects of local variations in separator mi- crostructure and disturbances (e.g., defects) cannot be resolved. However, Figure 4 shows the concentration profiles and density maps of these simulated, steady-state concentra- tion gradients at different depths in the separator structure, which reveal the influence of separator topol- ogy on the concentration distributions. For 1D pores in the TP direction, all cylindrical pore channels have the same concentration at a given depth, so the concentra- tion profile is a straight line. Upon adding more pores in a second and third direction, the concentration profile at a given depth broadens slightly (~2 mM and ~3 mM, re- spectively). For the PE separator, the concentration pro- file broadens to ~13 mM, indicating a variety of ion con- centrations in different pores at a given depth. The con- centration profile for the PP separator shows a broad dis- tribution of concentrations at each depth. As marked by arrows, there are also regions where the same electrolyte concentration is found over close to 1 µm in length. This comes from dead-end pores, which extend in the TP di- rection but lack a connection with other pore channels.31 Compared to the PE dataset, the PP dataset shows many more such threads, consistent with the network analysis (Table III). Assuming PE separator thicknesses of 12-16 µm, we es- timate a broadening of 45-60 mM. This corresponds to a range of 23-24% of the calculated concentration differ- ences (~190-250 mM) at each depth, as outlined in sec- tion 9 in the ESI. In contrast, for the PP separator, a broadening corresponding to ~50 % of the concentration differences is found. This is consistent with the higher connectivity in the PE separator than the PP separator. Regions of different electrolyte concentrations through- out the separator and impinging on the electrode may contribute to uneven lithium insertion in the electrode ma- terial, resulting in uneven expansion, diffusion induced stress and cracking,13,32,33 as well as local overcharging or deep discharging. This can diminish battery perfor- mance and shorten battery life-time. Therefore, we expect that a highly-connected structure reduces degradation in a battery (Table II). This finding highlights the importance of knowing connec- tivity in a separator structure: the broadening of the local ion concentration distribution across a separator in the TP direction cannot be deduced from the effective transport coefficient δTP. Table III. Pore network properties for the artificially generated microstructures with pores in 2 and 3 directions, as well as for Targray PE16A (PE), and Celgard® PP1615 (PP). The values are averaged for three datasets of edge length 5 µm. Parameter Proportion [%] of nodes of or- der Node density [µm-3] Branch density [µm-3] End point branches [%] Average branch length [nm] 3 4 5 6 2D 74.44±1.07 24.00±0.26 0.68±0.10 0.01±0.01 171.56±1.40 284.35±2.12 0 158.74±1.35 4 3D 74.37±0.50 22.07±0.53 2.82±0.07 0.57±0.03 187.24±1.66 314.46±2.18 0 148.12±0.62 PE 80.97±0.35 15.16±0.14 3.08±0.13 0.60±0.05 282.68±8.64 490.03±14.62 9.07±0.39 129.85±0.42 PP 89.60±0.17 9.28±0.16 1.00±0.03 0.10±0.04 36.50±1.23 69.87±2.06 31.55 ±0.81 189.88±0.92 Figure 4. (a) Concentration profiles and (b) concentration density maps from steady-state Fickian diffusion simulations across the through-plane direction of artificially generated and recorded datasets of 3 µm edge length with a concentration difference of ~50 mM between top (1.25 M) and bottom (1.20 M). In a next step, we assess how efficient the different sep- arator topologies are at smoothing out in-plane ion con- centration gradients that impinge on separator structures and are caused, e.g., by blocked pores. In a commercial lithium ion battery (schematic shown in Figure 5a), active particles in an electrode are typically on the order of 1 to 40 µm in diameter. This means that the electrode pore space structure has typical features approximately one to two orders of magnitude larger than the pore space of the separator. Direct contact between electrodes and sepa- rators can result in different concentrations of lithium ions in regions where the separator's pores are blocked by the electrode particles, and concentrations in regions where the electrode's and separator's pores meet. An example of an interface between a graphite electrode and a sepa- rator is shown in Figure 5b. Alternatively, a defect during separator manufacturing or battery assembly (e.g., ag- glomeration or contamination) as shown in Figure 5c may result in blocked areas and in decreased perfor- mance.34 Pore-blocking defects can create ion-insulated regions, which locally may lead to high Li-ion concentra- tions and over-potentials at the distant separator inter- face. Local defects in separators lead to non-uniform charging and plating around the defect.12 To simulate these types of scenarios and gain an under- standing of how high connectivity in a structure can help compensate for concentration gradients, we assume that a 3 µm circular object (electrode particle or defect) locally prevents electrolyte from impinging on the separator. Figure 5e shows the concentration density maps as in Figure 4, and Figure 5f shows the ion concentration pro- file at 1.5 µm depth for each sub-volume. For structures with zero connectivity, the blocked pores do not contribute to the effective transport and form an ion-insulated region. For structures with intermediate connectivity in the range of >0 to ~100 µm-3 (i.e., the 2D artificial structure or the PP structure, see sections 10-11 in the ESI), the pore network can compensate for the presence of defect structures at a depth of ~3 µm, while for microstructures with high connectivity in the range of 100 to 150 µm-3 (i.e., the 3D artificial structure and PE) are only mildly affected by the presence of defect struc- tures, and the fan-like distortion ends at around 1.5 µm. Due to its many redundant connections and slanted pores, PE is better in equalising in-plane ion gradients than the 3D artificial structures. Conclusions In summary, we quantified the difference in the topologi- cal parameters and node structure of PE and PP separa- tors of comparable porosity, TP tortuosity, and effective transport. High connectivity of the pores, as found in PE separators, enables ion gradients present at the top of the separator to be smoothed out within a fraction of the separator thickness. A structure with multiple straight cylindrical channels, though offering excellent TP tortuosity and ef- fective transport, has zero connectivity density and, due to the likely presence of defects, is more susceptible to Li plating if integrated into a lithium ion battery. In order to understand separator performance and optimise next generator separators for superior performance in cells, connectivity should be considered. 5 Beyond their function in describing homogenisation of ion concentration gradients through separators as described in detail here, topological and network-based analysis can also be used to predict how a structure will respond to mechanical or thermal stress35. By leveraging known trends in how a structure of a given topology shrinks under thermal stress36, deforms in response to compres- sive or tensile stresses37,38, or maintains connectivity de- spite closing of branches or nodes35, it will be possible to predict a separator's response to many of the dynamic processes experienced during cell manufacturing and operation39. Figure 5. (a) Schematic of LIB setup with electrode particles touching the separator. (b) Rendering of separator volume of 3 µm edge length with separator-electrode interface. The electrode particles are of diameters 3-6 µm (i.e., up to three orders of magni- tude larger than separator pores)40 and can thus block a significant number of separator pores. (c) Rendering of separator volume of 3 µm edge length with blocked pores. (d) Concentration distribution from diffusion simulation across separator volume of 3 µm edge length with defect of diameter 3 µm on top (dashed circle) and at half depth (dashed line at 1.5 µm). (e) Concentration density maps for steady-state Fickian diffusion simulations across the through-plane directions of the reference datasets with pores in 1D, 2D, and 3D, and the imaged datasets of PE and PP (3 µm edge lengths). A circular defect structure of 3 µm diameter was placed on top of the structures and a concentration difference of ~50 mM was applied between top and bottom. (f) Ion concentration profiles at 1.5 µm depth (dashed white lines in Figures 5d and e) across sub-volumes of 3 µm edge length. The dashed circle represents the defect structure at 0 µm depth (see Figure 5d). 6 Connectivity density can be calculated via a Minkowski functional and is important when describing homogenisa- tion of ion concentration gradients across microstruc- tures. Similarly, other morphological and topological pa- rameters are helpful when assessing surface interactions and effects31,40,41. Among such parameters are other Min- kowski functionals, which correspond to a microstruc- ture's surface area and curvature (described in detail in sections 2-3 in the ESI). Finally, lithium ion battery separators are just one exam- ple of a component in energy and environmental systems that can benefit from the topological and network analysis presented here42,43. Connectivity can also improve under- standing and design of separators in other electrochemi- cal systems such as fuel cells44,45 or ion-selective mem- branes for desalination46,47, providing insights such as how thick a membrane should be or how transport paths can be designed to prevent mixing of product/reactant streams. Furthermore, beyond separator technology, we propose that all electrochemical systems (catalysis stacks for fuel generation, sensitised solar cells, lithium ion battery anodes and cathodes, etc.) can be viewed as interwoven electronic and ionic networks. An ideal sys- tem will have balanced networks at all length scales that bring together (or carry away) electrons and ions (or re- actants and products) at equal rates while maintaining mechanical stability. Conflicts of interest There are no conflicts to declare. References 1. Martin, M. A. et al. Morphological Influence in Lithium-Ion Battery 3D Electrode Architectures. J. Electrochem. Soc. 162, A991–A1002 (2015). Chen, C.-F., Verma, A. & Mukherjee, P. P. Probing the Role of Electrode Microstructure in the Lithium-Ion Battery Thermal Behavior. J. Electrochem. Soc. 164, E3146–E3158 (2017). Doyle, M. Modeling of Galvanostatic Charge and Discharge of the Lithium/Polymer/Insertion Cell. J. Electrochem. Soc. 140, 1526–1533 (1993). Fuller, Thomas F. Doyle, Marc. Newman, J. Simulation and Optimization of the Dual Lithium Ion Insertion Cell. J. Electrochem. Soc. 141, 1– 10 (1994). Fuller, T. F., Doyle, M. & Newman, J. Relaxation phenomena in lithium-ion-insertion cells. J. Electrochem. Soc. 141, 982–990 (1994). Fortran programs for the simulation of electrochemical systems. (2009). COMSOL Multiphysics. Batteries & Fuel Cells Module. Ebner, M., Chung, D. W., García, R. E. & Wood, V. Tortuosity anisotropy in lithium-ion battery electrodes. Adv. Energy Mater. 4, 1–6 (2014). Newman, J. & Tiedemann, W. Porous-electrode theory with battery applications. AIChE J. 21, 25–41 (1975). 2. 3. 4. 5. 6. 7. 8. 9. Chandrasekaran, R. Quantification of contributions to the cell overpotential during galvanostatic discharge of a lithium-ion cell. J. Power Sources 262, 501–513 (2014). Harris, S. J. & Lu, P. Effects of inhomogeneities - Nanoscale to mesoscale -on the durability of Li- ion batteries. J. Phys. Chem. C 117, 6481–6492 (2013). Cannarella, J. & Arnold, C. B. The Effects of Defects on Localized Plating in Lithium-Ion Batteries. J. Electrochem. Soc. 162, A1365– A1373 (2015). Sun, G. et al. On the fragmentation of active material secondary particles in lithium ion battery cathodes induced by charge cycling. Extrem. Mech. Lett. 9, 449–458 (2016). Leung, K. & Jungjohann, K. L. Spatial Heterogeneities and Onset of Passivation Breakdown at Lithium Anode Interfaces. J. Phys. Chem. C 121, 20188–20196 (2017). Lagadec, M. F., Ebner, M., Zahn, R. & Wood, V. Communication - Technique for Visualization and Quantification of Lithium-Ion Battery Separator Microstructure. J. Electrochem. Soc. 163, A992–A994 (2016). Vogel, H.-J., Weller, U. & Schlüter, S. Quantification of soil structure based on Minkowski functions. Comput. Geosci. 36, 1236– 1245 (2010). Rabot, E., Wiesmeier, M., Schlüter, S. & Vogel, H. J. Soil structure as an indicator of soil functions: A review. Geoderma 314, 122–137 (2018). Strogatz, S. H. Exploring complex networks. Nature 410, 268–276 (2001). Arora, P. & Zhang, Z. Battery separators. Chem. Rev. 104, 4419–4462 (2004). Zhang, S. S. A review on the separators of liquid electrolyte Li-ion batteries. J. Power Sources 164, 351–364 (2007). Finegan, D. P. et al. Characterising the structural properties of polymer separators for lithium-ion batteries in 3D using phase contrast X-ray microscopy. J. Power Sources 333, 184–192 (2016). Lagadec, M. F., Ebner, M. & Wood, V. Microstructure of Targray PE16A Lithium-Ion Battery Separator. (2016). doi:10.5905/ethz- 1007-32 Lagadec, M. F. & Wood, V. Microstructure of Celgard® PP1615 Lithium-Ion Battery Separator. (2018). doi:10.3929/ethz-b-000265085 Scholz, C. et al. Permeability of Porous Materials Determined from the Euler Characteristic. Phys. Rev. Lett. 109, 264504 (1-5) (2012). Legland, D., Kiêu, K. & Devaux, M. F. Computation of Minkowski Measures on 2D and 3D Binary Images. Image Anal Stereol 26, 83–92 (2007). Russ, J. C. The Image Processing Handbook. (CRC Press, Taylor & Francis Group, 2016). 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 7 41. 42. Zahn, R., Lagadec, M. F., Hess, M. & Wood, V. Improving ionic conductivity and lithium-ion transference number in lithium-ion battery separators. ACS Appl. Mater. Interfaces 8, 32637–32642 (2016). Choong, L. T., Yi, P. & Rutledge, G. C. Three- dimensional imaging of electrospun fiber mats using confocal laser scanning microscopy and digital image analysis. J. Mater. Sci. 50, 3014– 3030 (2015). 43. Müllner, T., Unger, K. K. & Tallarek, U. Characterization of microscopic disorder in reconstructed porous materials and assessment of mass transport-relevant structural descriptors. New J. Chem. 40, 3993–4015 (2016). 45. 46. 44. Ostadi, H. et al. 3D reconstruction of a gas diffusion layer and a microporous layer. J. Memb. Sci. 351, 69–74 (2010). Joos, J., Ender, M., Rotscholl, I., Menzler, N. H. & Ivers-Tiffée, E. Quantification of double-layer Ni/YSZ fuel cell anodes from focused ion beam tomography data. J. Power Sources 246, 819– 830 (2014). Calo, V. M., Iliev, O., Lakdawala, Z., Leonard, K. H. L. & Printsypar, G. Pore-scale modeling and simulation of flow, transport, and adsorptive or osmotic effects in membranes: the influence of membrane microstructure. Int. J. Adv. Eng. Sci. Appl. Math. 7, 2–13 (2015). Zhao, X., Li, J. & Liu, C. Improving the separation performance of the forward osmosis membrane based on the etched microstructure of the supporting layer. Desalination 408, 102– 109 (2017). 47. 27. 28. 29. 30. 31. 32. 33. 35. 36. 37. 38. 39. DeHoff, R. T., Aigeltinger, E. H. & Craig, K. R. Experimental determination of the topological properties of three-dimensional microstructures. J. Microsc. 95, 69–91 (1972). Reznikov, N. et al. Inter-trabecular angle: A parameter of trabecular bone architecture in the human proximal femur that reveals underlying topological motifs. Acta Biomater. 44, 65–72 (2016). Kehrwald, D., Shearing, P. R., Brandon, N. P., Sinha, P. K. & Harris, S. J. Local Tortuosity Inhomogeneities in a Lithium Battery Composite Electrode. J. Electrochem. Soc. 158, A1393 (2011). Ebner, M. Designing Better Batteries: Visualization and Quantification of Microstructure and Degradation Mechanisms in Lithium Ion Battery Electrodes. (ETH Zürich, 2014). doi:10.3929/ethz-a-010211757 Zahn, R., Lagadec, M. F. & Wood, V. Transport in Lithium Ion Batteries: Reconciling Impedance and Structural Analysis. ACS Energy Lett. 2, 2452–2453 (2017). Zhao, K., Pharr, M., Cai, S., Vlassak, J. J. & Suo, Z. Large plastic deformation in high-capacity lithium-ion batteries caused by charge and discharge. J. Am. Ceram. Soc. 94, 226–235 (2011). Ji, L., Guo, Z. & Wu, Y. Computational and Experimental Observation of Li-Ion Concentration Distribution and Diffusion-Induced Stress in Porous Battery Electrodes. Energy Technol. 5, 1702–1711 (2017). manufacturing defects on electrochemical performance of lithium-ion batteries: Cognizance of the battery failure sources. J. Power Sources 312, 70–79 (2016). Lagadec, M. F., Zahn, R. & Wood, V. Designing polyolefin separators to minimise the impact of local compressive stresses on lithium ion battery performance. J. Electrochem. Soc. 165, (2018). Love, C. T. Thermomechanical analysis and durability of commercial micro-porous polymer Li-ion battery separators. J. Power Sources 196, 2905–2912 (2011). Cannarella, J. et al. Mechanical Properties of a Battery Separator under Compression and Tension. J. Electrochem. Soc. 161, F3117– F3122 (2014). Zhang, X., Sahraei, E. & Wang, K. Li-ion Battery Separators, Mechanical Integrity and Failure Mechanisms Leading to Soft and Hard Internal Shorts. Sci. Rep. 6, 32578 (2016). Zhang, X., Zhu, J. & Sahraei, E. Degradation of battery separators under charge-discharge cycles. RSC Adv. 7, 56099–56107 (2017). 34. Mohanty, D. et al. Effect of electrode 40. Müller, S. et al. Quantifying Inhomogeneity of Lithium Ion Battery Electrodes and Its Influence on Electrochemical Performance. J. Electrochem. Soc. 165, A339–A344 (2018). 8 Electronic Supporting Information Topological and Network Analysis of Lithium Ion Battery Components: The Im- portance of Pore Space Connectivity for Cell Operation Marie Francine Lagadec, Raphael Zahn, Simon Müller, and Vanessa Wood* Department of Information Technology and Electrical Engineering, ETH Zurich, Zurich CH-8092, Swit- zerland * To whom correspondence should be addressed. E-mail: [email protected] 1. Effective transport properties of the Celgard® PP1615 separator d = 3), can also be expressed as intensive pa- rameters (i.e., normalised functionals), mx(P).4 #$(P)=)$(P∩Ω) V- Figure S1. (a) Scale space analysis of the Celgard® PP1615 dataset3 showing the convergence of the porosity distributions at different sub-volume sizes towards the mean value. (b) Porosity ε, (c) tortuosity τ, and (d) effective transport coefficient δ histograms for TP, IP1, and IP2 directions of the 3 µm edge length sub-volumes. The Celgard® PP1615 separator was analysed as described in our previous work1. This sepa- rator has relatively large pore channels leading to a large representative volume element (RVE) with an edge length of ~3 µm (see Figure S1a). For the Targray PE16A separator dataset2, the RVE edge length is ~2 µm as determined in our previous work1. For the PP RVE edge length of 3 µm, we determine a porosity ε of 40.19±1.03 % (Figure S1b), tortuosities τTP = 2.04±0.19, τIP1 = 2.31±0.24, and τIP2 = 24.89±6.15 (Figure S1c). This is also reflected in the effective transport coefficients δTP = 19.9±2.0 %, δIP1 17.6±2.1 %, and δIP2 = 1.7±0.4 % (Figure S1d); effective transport in the IP2 direction is thus approximately ten times worse than in the TP or IP1 directions. The pore networks of both the Targray PE16A and the Celgard® PP1615 separators consist of a single interconnected pore network. Small ar- eas of non-connected pore space that might be present in these separators do not contribute to ionic transport across the separator and should be omitted for performance evaluations. Our imaging process relies on infilling the con- nected pore structure of the separator with ma- terial of a high imaging contrast1. This infilling process omits the non-connected pore space yielding a single interconnected pore network. 2. Extensive and intensive Minkowski The pore space of a binary structure is defined as P, which has an embedding space Ω (P ⊂ Ω, tionals, Mx(P) (with x ∈ {0, …, d} and d being the with Ω occupying the total dataset volume, VΩ). The pore space's boundary is δP, and its sur- face element for cylindrically-shaped structures is ds = R dz dφ. The extensive Minkowski func- functionals dimensionality of the structure of interest, here: S1 The first Minkowski functional, M0(P), corre- sponds to the pore volume, VPore: M/(P)=V0123(P). The first normalised Minkowski functional, m0(P), corresponds to a structure's porosity, ε. For N non-intersecting cylindrical pores of height h and radius R (see Figure 2a in the main text), M0(P) becomes N(cid:1)π R2(cid:1)h. The second Minkowski functional, M1(P), measures the interfacial area between pores and solid phase. The corresponding normalised functional is the specific surface area, m1(P). M5(P)=6ds :0 third Minkowski For N non-intersecting cylindrical pores of height h and radius R (Figure 2a), the integral yields N(cid:1)2 π R(cid:1)(R + h). The functional, M2(P), measures the mean curvature, H(P), over the interface and is a one-dimensional shape factor for 3D shapes. The corresponding normalised functional is called mean breadth density, m2(P). = 14π6 A 1rCDE+ 1rCGHIds :0 M;(P)= 12π6H(P) ds :0 For N non-intersecting cylindrical pores of height h and radius R (Figures 2a and 3a) the integral yields N/2(cid:1)(h + π(cid:1)R). Thus, for long, small pores, M2(P) becomes N(cid:1)h/2. For complex structures with interconnectivity in (Figures 3b-c), M2(P) several directions becomes more complex, and can be calculated numerically, but no longer analytically. The functional, M3(P), measures the Gaussian curvature, K(P), over the interface (thus, the total curvature), and is proportional to the Euler-Poincaré characteris- tic, Χ(P), a topological invariant. fourth Minkowski MJ(P)=6K(P) ds =6 A =4π∙M :0 1 rCDE∙rCGHIds :0 In 3D, the characteristic Χ is linked to the Betti numbers β0 (number of objects N), β1 (connec- tivity C) and β2 (number of enclosed cavities), = and, for a voxel-based dataset, to NV (number of vertices), NE (number of edges), NF (number of faces), and Nvox (number of voxels, or sol- ids).4,5 M=N/−N5+N;= PQ−PR+PS−PTU$ pressed by the number of pores, N (N ≧ 0), and the connectivity, C (C ≧ 0). M=P−W For percolating networks of pores and solid (i.e., without enclosed cavities), the character- istic Χ of the pore space can simply be ex- The corresponding normalised parameters, χ and c, are the Euler-Poincaré characteristic density and the connectivity density. 3. Minkowski functionals of separator microstructures Minkowski functionals have previously been linked to transport related parameters. The shape factor, in combination with the surface area, provides a first approximation of the diffu- sion coefficient6. Structures with high connec- tivity have large node and branch densities, and higher order nodes are associated with more spreading power.7 Meanwhile, in diffusion sim- ulations, branches that dead-end do not con- tribute to effective transport through the struc- ture.8 The intensive Minkowski functionals (i.e., nor- malised to the analysed volume and designated here with mx) are listed in Table T1. As de- signed, the reference separator microstructures replicate porosities well within the specified po- rosity of 40±2 and 40±5 % of the PE and PP separators. The specific surface area of the ref- erence datasets slightly decreases as more pores are added in a second and third perpen- dicular direction. The specific surface area (m1) of the PE separator (11.72 µm-1) is almost iden- tical to m1 of artificially generated microstruc- tures with pores in three perpendicular direc- tions (11.62 µm-1). This is expected since pore size as well as porosity were chosen to match the parameters of the PE separator. It shows that cylindrical pore segments are a good ap- proximation for the pore shape in the PE sepa- rator. For the PP separator, the value for m1 is lower (5.22 µm-1) due to the larger pores in PP. Table T1. Average values and standard deviations of the intensive Minkowski functionals m0, m1, m2, as well as Euler-Poincaré characteristics and connectivity densities χ and c, respectively, for the artificially generated micro- structures (1D, 2D and 3D) and the imaged Targray PE16A (PE) and Celgard® PP1615 (PP) separator microstruc- tures of edge lengths 5 µm each. The values for χ and c are calculated via the Minkowski functional M3. Parameter Porosity m0 [%] Specific surface area m1 [µm-1] Shape factor density m2 [µm-2] Topological invariant density χ [µm-3] Connectivity density c [µm-3] 1D 39.95±0.00 13.94±0.00 17.40±0.01 7.23±0.00 0.00±0.00 2D 40.48±0.07 12.11±0.02 7.82±0.03 -102.71±0.16 102.71±0.16 3D 41.04±0.04 11.62±0.02 5.44±0.03 -117.36±1.15 117.37±1.15 PE 40.53±0.77 11.72±0.13 6.68±0.35 -143.15±6.88 143.16±6.88 PP 40.19±0.42 5.22±0.14 1.54±0.07 -7.43±0.51 7.44±0.51 S2 tersect. In this case, a new pore generating it- eration is induced (starting from (i)) whereby the shortage in porosity is converted into the new number of circles to be created. This procedure runs until the target porosity of 40±2 % is met. While the artificially generated 1D microstruc- tures consist of individual pores that are not in- terconnected amongst each other, the artifi- cially generated 2D and 3D microstructures are strongly interconnected and form a single con- nected pore network. This interconnectivity is not an implicit result of our algorithm for creat- ing artificial microstructures, however, at 40 % porosity, it is extremely unlikely for cylindrical pores to penetrate 5 µm (= 500 voxels) thick structures without crossing another pore (prob- ability ~ (0.6)500). Therefore, none of the used artificially generated 2D and 3D microstructures contain isolated, non-interconnected pores. Artificially generated 2D and 3D microstruc- tures contain high fractions of third order nodes (see Table III). Nodes of order four (or higher) are only created if the central skeleton lines of two (or more) pores intersect in one single point. For pores of final diameter, such events have a low likelihood, and for most intersecting pores, the central skeleton lines of these pores will not intersect. Thus, several third order nodes are created instead of one single higher order node. with NB being the number of branches, NN being the number of nodes, and NEP being the number of end points.11 The description for connectivity using branches, end-points, and nodes is valid only in 2D; in 3D, the correct description uses edges, faces, and vertices of the single voxels. For our structures with 1D and 2D pore direc- tionality, the values for connectivity using the descriptions for 2D and 3D are the same. For the artificially generated pore structures in 3D and the recorded datasets, the values obtained via the 2D description are off by less than ±1 % compared to the values obtained via the 3D de- scription. For a 2D network, the connectivity C (number of loops) can be calculated via 5. Shape analysis C= NZ−N[−N\0+1 m2 can be interpreted as a measure of shape of the surface. For the all structures (reference and real), the values are positive, indicating that the shape of the pore surface is on average convex.9,10 For the reference datasets with cy- lindrical pores in one direction, the shape factor scales with the number of pores N and with h+π (cid:1)R, where h is the pore length and R is the pore radius. As more pores are introduced in a sec- ond and third perpendicular direction, the shape factor density decreases (from 17.40 µm-2 to 7.82 µm-2 in 2D and 5.44 µm-2 in 3D). The PE separator's shape factor density (6.68 µm-2) is comparable to the ones of the datasets with straight pores in 2 and 3 directions, while the PP separator's shape factor density is lower (1.54 µm-2) because of its higher proportion of concave regions at the pore surface. 4. Algorithm to generate artificial sep- arator geometries The following paragraph describes an algorithm to generate an artificial separator geometry consisting of solid elements and pore space. The pores are cylindrical and their orientation is always parallel to the IP1, IP2, or TP direction. At first, a cuboid with desired dimensions is de- fined. In three different datasets, we create pores in the TP direction (1D pore directional- ity), pores parallel to the TP and IP1 direction (2D pore directionality), and pores parallel to the TP, IP1, and IP2 axis (3D pore directional- ity). The desired porosity of 40 % is divided by the number of pore directions (one, two, or three) in order to get the same porosity in all directions. The pore generating process con- sists of two major steps. (i) On the face orthogonal to each desired pore direction, circles with a predefined radius (130 nm) are generated in an iterative manner. The number of circles is set by the porosity. The lo- cation of the circles is random with the only con- straint that they cannot touch or intersect. (ii) As soon as the necessary quantity of circles is created, they are extended through the entire separator producing the pores. If more than one pore direction is wanted, the total porosity might be smaller than the addition of the directional porosities since pores may in- S3 6. Network analysis log-log-plots In network theory, it is common to assess the node order distributions on a log-log plot. A Poisson distribution indicates a random net- work and power law distribution indicates a scale-free network. Figure S2. Log-log-plots of the node order distribu- tion of the 2D and 3D reference datasets and the im- aged PE and PP datasets. Figure S2 shows that node order distribution in the 2D and 3D dataset follows an exponential distribution, while the distributions of the PE and PP datasets seem to follow a power law. However, to quantify the scaling of the node or- der, the distribution should exhibit a linear rela- tionship on a log-log-plot over at least two or- ders of magnitude in both the x and y axes.12 In standard processing of voxel-based data only nodes of order 3-6 can be reliably identified.13 The x axis of Figure S2, therefore, spans less than one order of magnitude. 7. Pore orientation analysis We determine the pore orientation angle distri- bution using ImageJ's Directionality plugin for the non-processed datasets for the PE (Figure S3a) and PP (Figure S3b) separators. As illus- trated in the left-most images, the orientation angle is calculated for all pores in a plane, and each plane is indexed by the slice number in a specific direction (TP, IP1, or IP2). A vertical cut through any of the pore orientation angle distri- bution density plots (three plots to the right), would yield a histogram that represents the pore orientation distribution for that slice. For a slice (i.e., plane) with pores perpendicular to the slice, the orientation angle is 0°. In the PE separator, for slices along the TP di- rection, the peak of the pore orientation distri- bution varies between 0° and 180°. This may be attributed to the presence of fibres in the IP di- rection located at different separator depths. For slices in the IP1 and IP2 directions, the pore orientation distributions are broadly and asym- metrically centred above 0°, indicating pores slanted at many different angles. For PP, slices along the TP and IP1 are similar with orientation angle histograms centred around 90° due to its straight pore in the TP and IP1 directions. Figure S3. Orientation angle distributions across TP, IP1, and IP2 directions of the greyscale value, non-pro- cessed datasets of (a) Targray PE16A, and (b) Celgard® PP1615 separators. S4 8. End point analysis The skeletonized pore space of the reference and imaged datasets of 5 µm edge length gives the total number of branches in the sub-volume, NB. To obtain the number of network branches, Nn, we prune the skeletonized datasets using ImageJ's AnalyseSkeleton 2D/3D plugin. Sub- tracting the pruned skeleton from the original skeleton gives the number, NEP* = NB – Nn, of the end point branches within the volume and their coordinates. To account for end points that stem from cropping the datasets, we discard end point branches with coordinates within 5 voxels of the sub-volume's surface. We deter- mine the end point density as the number of end point branches per volume (NEP* / V), and the percentage of end point branches as the frac- tion of end point branches and the total number of branches (NEP* / NB). 9. Steady-state diffusion simulations We simulate the C-rate dependence of the elec- trolyte salt concentration gradient across a Li0separatorLTO cell as shown in Figure S4 and as described in our earlier work.14 At 1C, a concentration difference of 0.25 M builds up across the 16 µm thick separator; this corre- sponds to a concentration difference of ~50 mM across a sub-volume of 3 µm edge length and to inlet and outlet concentrations of 1.25 and 1.20 M, respectively. We plot the broadening of the electrolyte salt concentration at each depth in the TP direction for artificially generated and imaged datasets (Figure 4 in the main text). Figure S4. C-rate dependence of electrolyte salt concentration for Li0separatorLTO cells with Tar- gray PE16A separator. S5 Figure S5. Calculated concentration broadening as function of sub-volume edge length of separator vol- ume and corresponding concentration difference. In Figure S5, we plot the concentration broad- ening (i.e., histogram width at half-depth of sub- volume) calculated for sub-volumes of 3, 4 and 5 µm (blue data points) against the concentra- tion difference across a separator sub-volume (lower x-axis, range of values from Figure S4) and the edge length of the separator sub-vol- ume (upper x-axis), and extrapolate the broad- ening of the electrolyte salt concentration to edge lengths of 12-16 µm (orange data points). 10. Adding nanofibers to Celgard® PP1615 separator geometries Due to resolution limitations, the measured PP dataset does not feature the amorphous PP nanofibers spanning the large pore channels (partially visible in the SEM in Figure 1b). To assess how the presence of nanofibers in the large pore channels affects effective transport and topological properties of the separator ge- ometry, cylindrical nanofibers with a diameter of 40 (thin nanofibers) and 60 nm (thick nano- fibers) and distances of 50-70 nm are added to the pore channels of the recorded PP dataset. The diameter and the distances are estimated from FIB-SEM cross-sectional images. At first, the existing separator geometry is loaded and rotated such that the direction along which the fibres have to be created corre- sponds to the IP2 axis. Then, the geometry is up-scaled isotropically in order to decrease voxel size three-dimensional smoothing with a Gaussian kernel. Start points of a given number of fibres are randomly gen- erated in the IP direction. The direction of each fibre is slightly deflected at random such that a direction distribution is created. A fibre ends as soon it (re-)enters the other side of the separa- tor. All fibres are dilated to the desired radius and in a final step, the resulting structure is smoothed again. followed by Table T2. Normalised Minkowski functionals m0, m1, m2, as well as Euler-Poincaré characteristic and connectivity densities, χ and c, respectively, for a PP separator microstructure of edge length 3 µm as imaged, and with thin and broad nanofibers. The values for χ and c are calculated via the Minkowski functional M3. PP1615 with nanofibers of diameter 40 nm 38.56 7.66 -3.31 -94.07 94.11 PP1615 with nanofibers of diameter 60 nm 35.80 7.08 -0.07 -54.89 54.93 Datasets Porosity m0 [%] Specific surface area m1 [µm-1] Shape factor density m2 [µm-2] Topological invariant density χ [µm-3] Connectivity density c [µm-3] PP1615 39.38 5.30 1.82 -6.70 6.74 11. Celgard® PP1615 separator geome- tries without and with nanofibers Figure S6 shows that there is little difference between the calculated density plots for the electrolyte salt concentrations across a sub-vol- ume of the PP1615 separator as imaged and with added nanofibers. Thus, we conclude that – from a geometric per- spective – the effective transport properties are not affected significantly by the presence of the PP nanofibers. The effect of the nanofibers can- not be neglected when modelling the mechani- cal properties of PP separators, as shown by Xu et al.15 for Celgard® 2400 separator. Figure S6. Electrolyte salt concentration across a sub-volume of 3 µm edge length of the Celgard® PP1615 separator dataset with artificially added thin (40 nm) and broad (60 nm) nanofibers. The calculated Minkowski functional densities listed in Table T2 show that adding nanofibers to the PP dataset results in a lower porosity, an increased specific surface area, a negative shape factor, and a more negative topological invariant. The latter corresponds to a more pos- itive connectivity density (up to ~100 µm-3), which is below the calculated connectivity den- sity of PE. Since the surface integral of the mean curvature can be interpreted as the aver- age of the mean curvature, a more positive shape factor indicates the presence of more convex parts. A negative shape factor like in the case of added nanofibers indicates thus more concave regions. For Table T2, we calculate the intensive Minkowski functionals for a single dataset of 3 µm edge length; in Table II in the S6 main text, we calculate the average and stand- ard deviation of the intensive Minkowski func- tionals for three datasets of 5 µm edge length. References 1. Lagadec, M. F., Ebner, M., Zahn, R. & Wood, V. Communication - Technique for Visualization and Quantification of Lithium- Ion Battery Separator Microstructure. J. Electrochem. Soc. 163, A992–A994 (2016). Lagadec, M. F., Ebner, M. & Wood, V. Microstructure of Targray PE16A Lithium- Ion Battery Separator. (2016). doi:10.5905/ethz-1007-32 Lagadec, M. F. & Wood, V. Microstructure of Celgard® PP1615 Lithium-Ion Battery Separator. (2018). doi:10.3929/ethz-b- 000265085 Vogel, H.-J., Weller, U. & Schlüter, S. Quantification of soil structure based on Minkowski functions. Comput. Geosci. 36, 1236–1245 (2010). Legland, D., Kiêu, K. & Devaux, M. F. Computation of Minkowski Measures on 2D and 3D Binary Images. Image Anal Stereol 26, 83–92 (2007). Mitra, P. P., Sen, P. N. & Schwartz, L. M. Short-time behavior of the diffusion coeffcient as a geometrical probe of porous media. Phys. Rev. B 47, 8565–8574 (1993). Strogatz, S. H. Exploring complex networks. Nature 410, 268–276 (2001). Zahn, R., Lagadec, M. F. & Wood, V. Transport in Lithium Ion Batteries: Reconciling Impedance and Structural Analysis. ACS Energy Lett. 2, 2452–2453 (2017). Ohser, J. & Mücklich, F. Statistical Analysis of Microstructures in Materials Science. (Wiley, 2000). San José Martínez, F., Muñoz Ortega, F. J., Caniego Monreal, F. J., Kravchenko, A. N. & Wang, W. Soil aggregate geometry: Measurements and morphology. Geoderma 237–238, 36–48 (2015). Russ, J. C. The Image Processing Handbook. (CRC Press, Taylor & Francis Group, 2016). Stumpf, M. P. H. & Porter, M. A. Critical Truths About Power Laws. Science (80-. ). 335, 665–666 (2012). Reznikov, N. et al. Inter-trabecular angle: A 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. parameter of trabecular bone architecture in the human proximal femur that reveals underlying topological motifs. Acta Biomater. 44, 65–72 (2016). Zahn, R., Lagadec, M. F., Hess, M. & Wood, V. Improving ionic conductivity and lithium-ion transference number in lithium- ion battery separators. ACS Appl. Mater. Interfaces 8, 32637–32642 (2016). Xu, H., Zhu, M., Marcicki, J. & Yang, X. G. Mechanical modeling of battery separator based on microstructure image analysis and stochastic characterization. J. Power Sources 345, 137–145 (2017). 14. 15. S7
1904.10252	1	1904	2019-04-23T11:25:12	Recent Progress and Future Prospects of 2D-based Photodetectors	[ "physics.app-ph" ]	Conventional semiconductors such as silicon and InGaAs based photodetectors have encountered a bottleneck in modern electronics and photonics in terms of spectral coverage, low resolution, non-transparency, non-flexibility and CMOS-incompatibility. New emerging 2D materials such as graphene, TMDs and their hybrid systems thereof, however, can circumvent all these issues benefitting from mechanically flexibility, extraordinary electronic and optical properties, as well as wafer-scale production and integration. Heterojunction-based photodiodes based on 2D materials offer ultrafast and broadband response from visible to far infrared range. Phototransistors based on 2D hybrid systems combined with other material platforms such as quantum dots, perovskites, organic materials, or plasmonic nanostructures yield ultrasensitive and broadband light detection capabilities. Notably the facile integration of 2D-photodetectors on silicon photonics or CMOS platforms paves the way towards high performance, low-cost, broadband sensing and imaging modalities.	physics.app-ph	physics	Recent Progress and Future Prospects of 2D-based Photodetectors Nengjie Huo, Gerasimos Konstantatos* Dr. N. Huo, Prof. G. Konstantatos ICFO -- Institut de Ciencies Fotoniques, The Barcelona Institute of Science and Technology, Castelldefels, 08860, Barcelona, Spain E-mail: [email protected] Prof. G. Konstantatos ICREA -- Institució Catalana de Recerca i Estudis Avançats, Lluis Companys 23, 08010 Barcelona, Spain Keywords: 2D materials, photodetectors, hybrids, quantum dots Abstract: Conventional semiconductors such as silicon and InGaAs based photodetectors have encountered a bottleneck in modern electronics and photonics in terms of spectral coverage, low resolution, non-transparency, non-flexibility and CMOS-incompatibility. New emerging 2D materials such as graphene, TMDs and their hybrid systems thereof, however, can circumvent all these issues benefitting from mechanically flexibility, extraordinary electronic and optical properties, as well as wafer-scale production and integration. Heterojunction-based photodiodes based on 2D materials offer ultrafast and broadband response from visible to far infrared range. Phototransistors based on 2D hybrid systems combined with other material platforms such as quantum dots, perovskites, organic materials, or plasmonic nanostructures yield ultrasensitive and broadband light detection capabilities. Notably the facile integration of 2D-photodetectors on silicon photonics or CMOS platforms paves the way towards high performance, low-cost, broadband sensing and imaging modalities. 1. Introduction 1 Photodetectors that convert light to electrical signals are one of key components in modern multifunctional technologies. So far, the developed photodetector technologies have covered the whole application spectrum that profoundly affects our daily lives: X-rays for biomedical imaging;[1] ultraviolet for lithography and living cell inspection;[2,3] visible light detection for digital camera and video imaging;[4] broad-range infrared detection for night vision, optical communications, atmospheric and quality inspection spectroscopy,[5-8] among others. Photodetectors are characterized with some key figures of merit, including responsivity, quantum efficiency, photo-gain, etc. In Box1, the relevant terminology, units and key figures of merits for photodetectors are summarized. To direct compare the performance among different devices, the detectivity D, in units of Jones (i.e. cm Hz1/2 W-1), is provided which takes into account variations in the detector geometry and noise measurement conditions. Today crystalline silicon photodetectors integrated with complementary metal-oxide-semiconductor (CMOS) technology have reached a high level of maturity and performance in digital cameras and optical sensing systems. Silicon has low dielectric constant which allows the low capacitance and fast operation (~ns), however the indirect bandgap (~1.1 eV) limits its high optical absorption and operation wavelength coverage within visible and near-wavelength infrared (NWIR) region.[9] Currently, infrared photodetectors based on epitaxial-grown InGaAs, InSb, HgCdTe and type-II superlattices have allowed sensing in the infrared spectrum beyond silicon´s reach.[9,10] These technologies have been well developed and commercialized with high sensitivity up to 1013 Jones and broad spectral coverage from NWIR to long-wavelength infrared (LWIR) region.[9] The single photon detectors such as single-photon avalanche photodiode, superconducting single-photon detectors, etc, have also been developed with high efficiency, low timing jitter and well photon-number resolution in NWIR region for specific applications such as quantum communication.[11] Despite the maturity of infrared photodetector technologies, there are still some major roadblocks for large scale deployment. For example, HgCdTe and other exotic semiconductors are usually fabricated with molecular 2 beam epitaxy (MBE) or metal-organic chemical vapour deposition (MOCVD), techniques which sinificantly increase the manufacturing complexity and cost. These detectors also suffer from CMOS-incompatibility due to the large lattice mismatch and require cryogenic temperature of operation to obtain low noise and high sensitivity. Besides, the conventional bulky semiconductors are rigid, brittle and opaque, preventing their specific applications from bendable and flexible optoelectronics. Presently, the diversity of photodetector applications are growing. Some factors in terms of CMOS-compatibility, low-cost manufacturing, transparency and flexibility, which are beyond the limits of conventional semiconductors, are needed for new emerging wearable and portable optoelectronics. Along this direction, two-dimensional (2D) materials, colloidal quantum dots (CQDs) and organic semiconductors have been demonstrated as appealing potential platforms. CQDs based photodetectors have exhibited high photoconductive gain in the range of 103-106 and detectivity up to 1013 Jones from visible to the shortwave-IR range by tuning the size of CQDs.[12,13] Organic photodiodes were recently reported with excellent LDR of 110 dB, -3dB bandwidth of 11.4 MHz and detectivity up to 1014 Jones,[14] showing their potential in high performance photodetectors. In this review, we focus on the development of 2D materials based photodetectors, functionalization with CQDs, organic materials, etc, to improve the performance, as well as the potential challenges to further develop and commercialize these technologies. Since the discovery of graphene in 2004, 2D atomic sheets including graphene, transition metal dichalcogenides (TMDs) and black phosphorus (BP), with intralayer covalent bonding and interlayer van der Waals (vdW) interaction have emerged as a unique and promising material family for photonics and optoelectronics in view of their appealing characteristics.[15-17] Their atomically thin feature enables high transparency and mechanical flexibility, offering the opportunity for bendable, flexible and conformal detectors.[18, 19-22]. Due to their flexible nature, 3 strain engineering can be exploited to modulate the electronic and optical properties. Desai et al. observed an indirect to direct bandgap transition in bilayer WSe2 through applied strain, which led to the improvment in photoluminescence by 35 times.[23] Recently, Ahn et al. also demonstrated a dramatic modulation of the band structure through strain engineering introduced directly during the growth process.[24] Strain engineering can also be exploited in bulk semiconductors to tune their optical and electrical properties. But, the maximum achievable tuning is eventually determined by the elastic limit of the material (∼1% in most bulky semiconductors). The atomically thin thickness of 2D materials leads to a high 'stretchability' up to ∼20% strain magnitudes, providing a larger room for poreprty tuning via strain.[25] When 2D materials, particular TMDs, are thinned to monolayer limit, quantum confinement effects lead to the indirect-to-direct bandgap transitions, which can be used to tune the band structure and absorption wavelength by varying the number of layers.[26] Due to the direct bandgap, the monolayer TMDs have strong light-matter interactions from the strongly bounded excitons.[27,28] High in-plane mobility, of 10,000 cm2 V-1 s-1 reported in graphene and 100-500 cm2 V-1 s-1 in TMDs, at room temperature, facilitates efficient photo-carrier extraction and leads to fast and sensitive detectors.[29-31] Another important advantage of 2D materials is the absence of surface dangling bonds due to the vdW interlayer interactions, enabling seamless integration on any substrate crystalline or amorphous, rigid or flexible.[32-34] Large area growth and easy- processing of 2D materials warrant low-cost and large-scale manufacturability.[35] Finally, the versatility of 2D family materials offer different applications ranging from semi-metal graphene, semiconducting TMDs to insulating h-BN and new monoatomic buckled crystals like black phosphorous (BP)[36] and silicene,[37], which can be synergistically combined to build hybrid platforms for multifunctional optoelectronics.[32,38] Benifiting from these extraordinary properties, 2D material-based photodetectors hold promise to overcome the roadblocks of current conventional photodetector technologies operated at 4 room temperature. On the other hand, 2D materials and their photodetectors also face some challenges. The ultrahigh mobility of graphene makes it suitable for high-speed photodetectors (up to 40 GHz bandwidth),[29] but its single atomic layer and zero bandgap limit its photo- absorption, external quantum efficiency and thus photoresponsivity. Graphene photodetectors are characterized by large dark current that is typically associated to high noise and power consumption. Various approaches have been proposed to introduce a bandgap in graphene, such as building graphene QD-like structures[39] or cutting graphene into nanoribbons.[40] However, these bandgap-opening methods degraded the electronic performance and the speed of graphene photodetectors compared to their pristine counterparts. 2D TMDs with sizable bandgap can overcome the disadvantages of graphene, however, the photoresponse wavelength range is limited from ultraviolet to near-IR, determined by their optical bandgap. Also the response speed of TMDs photodetectors is relatively slow due to significant trapping effects of photocarriers.[31] BP has an intermediate bandgap between graphene and TMDCs, thus covers a broad photoresponse spectrum from near-IR to mid-IR. BP photodetectors were also reported with high responsivity of 657 mA/W and fast speed up to 3 GHz,[41] yet its stability remains a main issue for this class of materials. In the first section of this review, we briefly discuss the classification of photodetectors as photoconductors, photodiodes and phototransistors. Then, we review the state-of-the-art 2D crystal based photodetectors in terms of the underlying physical mechanisms with and without gain. We focus on the strategies developed recently to address the above-mentioned drawbacks and further improve the performance of 2D photodetectors via doping or sensitization with colloidal quantum dots (CQDs), perovskites, metal nanostructures and organic molecules or polymers. In the third section, we demonstrate current applications of 2D photodetectors on flexible substrates or integrated with silicon and CMOS technologies. In the last section, we compare the performance between commercial photodetectors and developed 2D 5 photodetectors, and outline the new challenges needed to be addressed to further boost performance, functionality and commercialization of this new optoelectronic platform. These challenges include improvement in linear dynamic range, contact and mobility engineering, noise suppression issues and advances in wafer-scale manufacturing, among others. Box Figures of merit: Responsivity R, Responsivity is defined as 𝑅 = 𝐼ph 𝑃in , representing the ratio of photocurrent generated in the detector over the incident optical power in units of A W-1, where Iph is photocurrent, Pin incident light power. External quantum efficiency, EQE is the ratio of the number of photogenerated electrons in a detector over the number of incident photons and is given as 𝐸𝑄𝐸 = 𝐼ph 𝑃in ℎ𝜈 𝑒 = 𝑅 ℎ𝜈 𝑒 Internal quantum efficiency, IQE is the ratio of the number of photogenerated electrons in a detector over the number of absorbed photons and is given as 𝐼𝑄𝐸 = 𝑅 hν 𝐴𝑎e , where Aα is the absorbed fraction of light. Linear dynamic range, LDR is used to characterize the light intensity range in which the photodetectors have a constant responsivity in unit of dB and can be expressed as 𝐿𝐷𝑅 = 10 × log10( 𝑃sat 𝑁𝐸𝑃 ) [dB], where NEP is noise equivalent power (defined below), Psat the saturated light intensity at which the photocurrent begins to deviate from linearity. Signal to Noise Ratio, SNR is the ratio of signal power to noise power, which is must be larger than unity so that the signal power can be distinguished from the noise. Photo-Gain, G describes the number of recirculated carriers in the circuit per single incident photon. In a photodiode, the photo-gain is equal to unity, unless carrier multiplication effects are present. In a photoconductor, one type of carrier (e.g. hole) is usually captured in trap states or sensitizing centres 6 with a lifetime of τlife, while the other type of carrier (e.g. electron) is free to traverse the channel with a transit time of τtransit . If carrier lifetime is longer than transit time, the free electrons recirculate many times before recombination with captured holes, leading to a generation of photo- gain, which is defined by 𝐺 = 𝜏life 𝜏transit = 𝜏life 𝐿2 𝜇𝑉DS , where L is length of channel, µ is carrier mobility, VDS is applied bias across channel. Noise equivalent power, NEP is the minimum detectable optical power at which the electrical signal-to-noise ratio (SNR) in the detector is equal to unity, when bandwidth is limited to 1 Hz. The unit is W Hz-1/2. NEP describes the sensitivity of a detector and is determined by both noise current and responsivity, 𝑁𝐸𝑃 = 𝐼𝑛𝑜𝑖𝑠𝑒 𝑅 . Detectivity D, is another common figure of merit to characterize the sensitivity of a detector, which can enable the comparison of detectors with different geometries. The unit is cm Hz1/2 W-1 or Jones. It can be defined by 𝐷∗ = √𝐴𝐵 𝑁𝐸𝑃 = 𝑅√𝐴𝐵 𝐼𝑛𝑜𝑠𝑖𝑒 = 𝑅√𝐴 𝑆𝑛 , where A is active area, B is electrical bandwidth, Sn is noise spectral density. 2. Classification of photodetectors 2.1. Photodiodes Typically, the two most common photodetector types are photodiodes and photoconductors. 2D crystals have been widely explored in both classes of detectors. In a photodiode, a built-in field is formed via a p-n junction or a Schottky junction between metal and semiconductors. The generated electrons and holes by incoming photons move to opposite contact electrodes driven by the built-in potential. The response speed is determined by the transit time of excess charge carriers defined by ttransit = L2/μVbi, where L is channel length, μ is carrier mobility and Vbi is the built-in potential across junction.[42] For graphene based photodiodes, the mobility is over 10,000 cm2 V-1 s-1 leading to ultrafast speed of picoseconds to nanoseconds and large bandwidth 7 on the order of Gigahertz.[29,43-45] On the other hand, the quantum efficiency of photodiodes is limited to unity, unless avalanche or carrier multiplication effects occur when operated near the breakdown regime, offering the possibility of multiple carriers generated per single photon; this comes at the cost of very high bias of 50-100 V that is needed to reach this regime. 2D materials such as MoS2, InSe and BP based avalanche photodiodes (APDs) have been demonstrated with carrier multiplication of 100-1000.[46-48] 2.2. Photoconductors In a photoconductor, the photo-excited electrons and holes are separated by the applied external bias and drift to opposite Ohmic contact electrodes, leading to photocurrent or photo-voltage generation. A gain mechanism, called photoconductive gain, can be present in this case when one type of charge carriers is able to circulate through an external circuit many times before it recombines with its opposite carrier. The gain is then defined as the ratio of lifetime and transit time (see Box 1). Usually, the trap states or sensitizing centres within the band gap of the semiconductors capture and localize one charge carrier type and effectively prolong its carrier lifetime, leading to multiple carriers per single photon and extremely high responsivity compared to a photodiode.[42,49] Due to the fact that both gain and temporal response are determined by the lifetime of trapped carriers, the bandwidth in this class of detectors is usually lower than that of the photodiodes, thus a trade-off between gain and response should be considered when designing detectors for specific applications. 2.3. Phototransistors The strategies to improve the performance of photodetectors are not only maximizing its electrical response to light in terms of gain, but also minimizing the noise in its electrical output. Low noise can be achieved in high quality photodiodes with suppressed generation- recombination noise and negligible shot noise contribution when operated in a photovoltaic mode; the absence of gain though limits the electrical response, giving rise to noise contributions from the read-out electronics. On the other hand, the Ohmic contacts in 8 photoconductors leads to large dark current and thereby large shot and 1/f noise components, that however may compete with the presence of gain to reach high SNR ratio under specific conditions. To address the high noise issue in photoconductors as mentioned above, the concept of phototransistors, a special case of photoconductors with an additional gate terminal which is electrically isolated from the semiconductor channel by a thin dielectric, is demonstrated in 2D materials based photodetectors.[31,49] The applied gate VG can electronically modulate the carrier density by field-effect and switch off the dark current by operating the device in the depletion regime. Thus both photoconductive gain and low noise can be achieved leading to high sensitivity and gain-bandwidth products. As a typical example, monolayer MoS2 with a direct band gap of 1.8 eV possesses large gate-tunable conductivity with on/off ratio exceeding 108 and extremely low off current densities of 25 fA μm-1,[30,50] resulting to reported responsivities of ~1000 A W-1 and experimentally measured detectivity of 1011-1012 Jones.[31,50-52] 3. 2D-based photodetectors In this section, we briefly discuss the fundamental properties of 2D crystals and the progress of 2D based photodetectors. 3.1. Crystal structure and basic properties 2D crystal systems encompass a large number of family members with unique and varying physical properties, including metals, semimetals, semiconductors, insulators. This enables the development of an entirely independent new generation of optoelectronics based exclusively on 2D materials. Depending on the specific application in terms of sensitivity, speed or wavelength spectrum coverage, different 2D materials are selected and designed. The chemical structure of various monolayer 2D materials including semi-metal graphene, semiconducting TMDs, insulating h-BN and monoatomic buckled crystals are shown in Figure 1. Graphene, monolayer carbon atoms bonded together in a hexagonal honeycomb lattice, is a gapless semi- metal material with ultrahigh mobility,[53-55] and as such an appealing material for broadband 9 photodetection from terahertz (THz) to ultraviolet (UV)[56-58] and ultrafast technologies.[43-45] Graphene can also serve as work function tunable electrode due to linear dispersion of the Dirac electrons near the K point.[59-61] 2D TMDs with a chemical formula MX2 (M: Mo, W, etc. and X: S, Se or Te), where each layer is composed of one layer of hexagonally packed metal atoms sandwiched with two layers of chalcogen atoms, have a variable bandgap of 1-2 eV and offer strong light-mater interactions,[62,63] which can be utilized as field effect channel semiconductors and photoactive layers for strong light absorption. 2D h-BN with an ultrawide bandgap of ~6 eV, where boron and nitrogen atoms are covalently bonded in each layer, is an excellent insulator as high-ƙ dielectric to enhance device performance such as mobility and stability.[64,65] Recently, new monoatomic buckled crystals (termed Xenes, Figure 1) like black phosphorene,[36,66,67] silicene,[37,68,69] germanene,[70,71] and bismuthene,[72,73] etc. have been theoretically predicted and experimentally developed, which can offer smaller bandgap of 0.2- 2 eV, high mobility of 100-1000 cm2 V-1 s-1 and the possibility to serve as high performance short- and mid-infrared photodetectors.[74-76] All these 2D crystals share the same advantages of transparency, flexibility, vdW interlayer interactions and CMOS compatibility. The well- developed manufacturing methods such as mechanical/liquid phase exfoliation,[77,78] CVD growth[35,79-81] and inkjet-printing[82,83] enable large-scale fabrication and easy-processing for both fundamental research and practical applications in optoelectronics. 3.2. 2D-based photodetectors without gain In this section, we discuss the 2D materials based photodetectors without gain and the corresponding photo-detection mechanisms. Typically, this class of photodetectors include 2D- photodetectors with photo-thermoelectric and bolometric effects and 2D heterojunction-based photodiodes based on the photovoltaic effect. 3.2.1. Photo-thermoelectric and bolometric detectors 10 The photo-thermoelectric effect (PTE) refers to the carriers generated by the photo-induced temperature gradient △T between different substances upon localized illumination. Thus a photovoltage VPTE can be produced through the Seebeck effect (or thermoelectric effect): VPTE = (S1-S2) △T, where S1 and S2 (in V K-1) are the Seebeck coefficients of the two materials (Figure 2a). The photovoltage can drive the hot carriers to form a net current flowing without external bias. Due to the dominant hot carrier transport and strong e-e interactions in graphene, the PTE effect plays a key role in the photo-response of graphene p-n junctions.[84-89] In some reported metal-graphene-metal photodetectors, the photocurrent is dominated by the photovoltaic effect.[90-92] Graphene photo-thermoelectric detectors exhibit high bandwidth up to 40 GHz, the responsivity however, is limited to 6.1 mA W-1 due to the low optical absorption and the absence of gain.[29,44,45,86-89] Beside graphene, significant PTE effect has also been reported in MoS2 detectors arising from the large mismatch of Seebeck coefficients between MoS2 and metal electrodes.[93-95] The Seebeck coefficient of MoS2 (30-100 mV K-1) is several orders of magnitude larger than that of graphene (4-100 µV K-1).[93-95] Black phosphorene[96,97] and 2D SnS2 [98] have also been reported with large Seebeck coefficients of 60-335 µV K-1 and 34.7 mV K-1, respectively, rendering them promising candidates in thermoelectric detectors. The bolometers, on the other hand, are used to detect the changes in incident photon radiation (dP) by measuring the changes in temperature (dT) of the absorbing element. The bolometric effect is associated with the conductance change (dG) of a semiconductor channel induced by photo-induced heating of an absorber and thus an external bias is needed to monitor the dG which is followed by the dT-induced change of the mobility (Figure 2b).[99,100] The sensitivity of bolometers is ultimately determined by the thermal conductance: Gh = dP/dT. High- sensitivity bolometers, which are made of doped or ion-implanted semiconductors,[101] show a strong temperature-dependence of its conductance at operating temperature upon illumination in the far-infrared and submillimetre (THz) wavelength range. Upon light irradiation, the temperature changes as a function of time with thermal time constant τ = Ch/Gh, where Ch is the 11 heat capacity which determines the response time.[99] The bilayer graphene-based bolometers have exhibited very fast intrinsic speeds (>1 GHz at 10 K) and high sensitivity (33 fW Hz -- 1/2 at 5 K) due to the small heat capacity and weak electron-phonon coupling.[102] Very recently, Wu et al. demonstrated a negative infared photoresponse (up to 1550 nm) in multilayer MoS2 for the first time with incident photon energy lower than bangap of MoS2. The responsivity at 980 nm can reach 2.3 A W-1 with response time of 50 ms. The negative photoresponse was attributed to the bolometric effect with a bolometric coefficient of -33 nA K−1. The increased temperature from light irradiation could lead to the decreased conductivity of MoS2 due to the increased electron-phonon interaction.[103] In bolometers, the temperature coefficient of resistance (TCR in units of % K−1) is another key performance indicator and defined as: TCR(𝑅0) = 1 𝑅0 𝑑𝑅 𝑑𝑇 , representing the percentage change in resistance per Kelvin at the operating resistance R0. Recently, graphene-based pyroelectric bolometers have reached a record TCR up to 900% K−1 at room temperature and temperature resolution down to 15 μK by incorporating a pyroelectric substrate (LiNbO3) and a floating metallic structure (Figure 2c).[104] Goswami et al. demonstrated a TCR up to -2.9% K−1 at room temperature in pulsed laser deposition (PLD) grown MoS2 film on silicon substrate. Their reported MoS2 bolometers were sensitive to mid- IR irradiation (7-8.2 µm) with responsivity of 8.7 V W-1 and response time on the order of 10 s.[105] Superconductors are also known to show a very strong resistance dependence on temperature near the critical temperature Tc, and have been proposed as sensitive superconducting bolometers.[99] Recently, many 2D materials such as MoS2,[106] Mo2C[107] and FeSe[108] are reported with remarkable superconductive properties possessing high critical temperature, thus they may serve as potential new material platforms for 2D superconducting bolometers due to their extremely sensitive resistance change upon temperature. 3.2.2. 2D heterojunction based photodiodes 12 Van der Waals heterojunctions designed by assembling isolated 2D crystals have emerged as a new class of artificial materials with extraordinary optoelectronic functionalities in various applications such as solar cells, photodetectors and light emitting diodes. The 2D heterojunctions are fabricated by either local chemically or electrostatically induced doping or through transfer technologies in both in-plane and out-of-plane directions. These heterojunctions have been typically explored as photodiodes, where ultrafast response is possible followed by the absence of gain. The underlying physical mechanism of photo- detection is the photovoltaic effect based on the separation of photo-generated electron -- hole (e -- h) pairs by the built-in electric field at the junction. Next, we discuss the progress in heterojunction-based photodiodes with both in-plane and out-of-plane configurations. In-plane heterojunctions: The atomically thin profile of 2D crystals enables the efficient modulation of their carrier density, work function and polarity through electrostatic doping[36,109] or chemical doping.[110,111] As a result, a large number of in-plane heterojunction based photodiodes have been fabricated. One class of in-plane photodiodes is realized by the localized electrostatic doping on 2D semiconducting channels including graphene,[53] WSe2 [109] and BP[36] with natural ambipolar transport behaviour. Figure 3a shows a typical device structure of WSe2 based lateral diode, where the split gate electrodes are located under a high- ƙ dielectric such as HfO2, h-BN or SiN.[112-115] The local gate can modulate the carrier type and concentration in the WSe2 channel atop. Electrons or holes can be induced respectively by applying opposite polarity bias on the split gate electrodes, thus forming a lateral P-N junction in a single WSe2 channel. In this case, the device can act as photodiode with evident rectifying behaviour under dark and photovoltaic effect under illumination, leading to the generation of short circuit current ISC and open circuit voltage VOC (Figure 3b). The responsivity of the in- plane WSe2 photodiode has been reported from 0.7-210 mA W-1 with tens of milliseconds response time. The EQE is also limited to 0.1%-0.2% due to the weak photon absorption and finite depletion width.[113,116] In the same device configuration and operation principle, BP[117] 13 and MoSe2 [118] have also been demonstrated as in-plane photodiodes with split gate well- controlled photocurrent generation and transport. Due to the narrow bandgap of BP, the BP photodiodes have an extended spectrum to NIR region (up to 1500 nm) with responsivity of 28 mA W-1, response time constant of 2 ms and EQE of 0.1%.[117] Another class of in-plane photodiodes has been developed via doping or direct CVD-growth. P-type doping techniques for MoS2 have been reported recently including plasma treatment,[119] niobium (Nb) physical doping[120] and gold chloride (AuCl3) chemical doping.[121] In a MoS2 phototransistor, one can use an insulator such as h-NB or HfO2 to cover partially the MoS2 channel protecting it from doping the underneath layer (Figure 3c), thus a lateral PN junction is formed at the interface to perform rectifying and photovoltaic properties with responsivity of 300 mA W-1.[122] The same device concept was also explored in BP[123] where benzyl viologen (BV) was used as electron dopant; the obtained responsivity was 180 mA W-1 at a wavelength of 1.5 µm. Recently, many groups[124-126] have developed the one- or two-step epitaxial growth of lateral TMD-TMD heterostructures, implemented in WSe2-MoS2 and WS2-WSe2 heterojunctions that exhibited atomically sharp interfaces (Figure 3d). Sahoo et al. developed a one-pot synthetic approach, using a single heterogeneous solid source, for the continuous fabrication of lateral multi-junction heterostructures consisting of monolayers of varing TMDs (Figure 3e). They also fabricated the heterojunction based devices (Figure 3f) and observed significant photocurrent across the junctions (Figure 3g).[127] These results enabled continuous growth of multi-junction lateral heterostructures and realization of lateral p-n diodes and photodiodes as well as complex in-plane superlattices. Thanks to their higher quality interface over previous architectures this class of photodide have reached EQE of 9.9% (corresponding to an IQE of 43%) and fast response speed of 100 µs.[126] Out-of-plane heterojunctions: Heterojunctions based on 2D materials can have another form factor in that they may form instead across the van der Waals interfaces or different 2D crystals, also known as out-of-plane heterojunctions. We summarize them herein as TMD-TMD vertical 14 heterojunctions and Gr-TMD-Gr sandwich structures. For the former case, benefiting from the well-developed transfer technologies, the atomically sharp surfaces and the absence of lattice- match constraints, a variety of 2D TMDs are stacked on each other to form high quality vertical heterojunctions.[128-132] As an example, MoS2 and WSe2 possess opposite intrinsic N-type and P-type behaviours, respectively, thus the stacking of those can form vertical PN junctions which have been extensively demonstrated with very promising photovoltaic and electroluminescent properties paving the way towards all-2D photovoltaics, photodetectors and light emitters.[129,133] Other TMDs have also been combined to exhibit diode-like behaviour and efficient photocurrent generation due to the type-II band alignment that can facilitate efficient electron -- hole separation for light detection and harvesting.[134] For example, Hong, et al. reported ultrafast (~50 fs) charge transfer at the MoS2/WS2 interface because of the type-II band alignment, showcasing the large potential of vertical 2D heterojunctions in optoelectronic applications that require very high speed including ultra-fast spectroscopy and optical, on-chip communications.[135] Among the widely reported vertical heterojunction-based photodiodes, the responsivity is on the order of 10 mA/W, with reported EQE up to 80% and response time in the millisecond range.[128-132] The higher quantum efficiency compared to that of in-plane photodiodes is attributed to the larger photoactive area (stacking region), strong interlayer coupling and efficient interlayer charge transfer. In such heterojunstions, some new physical mechanisms emerge resulting in drastic improvments in device performance. Barati et al. demonstrated a monolayer MoSe2 and bilayer WSe2 heterojunction based photocells (Figure 4a) and observed highly efficient multiplication of interlayer e-h pairs which are generated by hot-electron impact excitation at temperature near 300 K. As shown in Figure 4b, the reverse bias photocurrent increases rapidly with increasing temperature indicating that an optically excited electron in the conduction band of WSe2 undergoes efficient multiplication of electrons in MoSe2. Such interlayer impact excitation results in responsivity enhancement exceeding 350% with a multiplication factor over 3.5 at low irradiance.[136] In addition to TMD-TMD 15 junctions, other 2D materials were also combined with TMDs to form photodiodes. Such architectures have been based on the combination of TMDs with graphene (or BP), where broadband detection up to 2.4 µm (or 1.5 µm) range with microsecond response speed and high D* of 1011 Jones were achieved.[137-139] Recently, Long et al. reported a black arsenic phosphorus (b-AsP)-based long-wavelength IR photodetectors with room temperature operation up to 8.2 µm. By combing with multilayer MoS2 (Insert of Figure 4c), the b-AsP- MoS2 heterojunction based devices showed a diode behaviour (Figure 4c), and the photodiode can exhibit a specific detectivity higher than 4.9 × 109 Jones in 3-5 µm range, high responsivity of 115.4 mA W-1 at 4.29 µm and fast speed of < 0.5 ms as well as suppressed flicker noise; these values are well beyond all room temperature Mid-IR photodetectors to date (Figure 4d).[140] Now we turn to the Gr-TMD-Gr sandwich structures for photo-detection applications. This device concept (Figure 4e), where graphene serves as top and bottom transparent electrodes for charge extraction and the TMD acts as the photoactive layer, responsible for photon-absorption and carrier transport, is proposed based on the efficient gate-tunable Fermi level of graphene[141] and strong light-matter interactions of TMDs.[62,142] The first report of this class of detectors was a Gr-WS2-Gr structure which acted as tunnelling transistors under dark.[142] A built-in field across WS2 was created due to the different Fermi level position in graphene induced by electrostatic gating (Figure 4f). Under light illumination, the photo-excited electrons and holes are subsequently separated and drifted towards opposite graphene electrodes. The built-in field and thus photocurrent can be efficiently modulated by varying the gate voltage. The devices reached remarkable performance with responsivity of ~0.1 A W-1 and EQE above 30%, demonstrated also as efficient flexible photovoltaic devices.[142] Then Yu et al. utilized multilayer MoS2, instead of WS2, to demonstrate gate-tunable photocurrent with EQE as high as 55% (IQE of 85%), fast response of 50 µs and responsivity of ~0.2 A W-1 as shown in Figure 4g and 4h.[143] Similarly, M. Massicotte, et al. fabricated Gr-WSe2-Gr heterostructures and 16 reached unprecedented photo-response time as short as 5.5 ps combined with high EQE of 7.3% (IQE of 85%), that can be tuned by applying a bias and by varying the WSe2 thickness, entering the regime of ultrafast photodetectors.[144] All these results indicate that the Gr-TMD-Gr sandwich structures in out-of-plane configuration allow highly efficient photo-detection capabilities owing to the strong light absorption in TMDs, atomically thin carrier transient paths for efficient charge separation and collection as well as practically sizeable junction areas for efficient photon harvesting. In this configuration, large room for improvement remains in particular towards broader spectral coverage employing 2D materials with smaller bandgap such as BP[36] and Bi2Se3.[145] 3.3. 2D-hybrid photodetectors with Gain Although high quantum efficiency and ultrafast response were achieved in 2D heterojunction based photodiodes with both in-plane and out-of-plane configurations, the responsivity of those has been limited to 1 A W-1 determined by the ceiling of unity quantum efficiency that photodiodes offer and the absence of gain. In view of this a new class of photodetectors have been proposed and developed based on 2D materials that offer the possibility of gain and the achievement of very high responsivity. These detectors rely on the photoconductive effect, that yields photoconductive gain due to the difference between the transit time of majority-type of carriers (e.g. electrons) and the lifetime of the minority carriers (e.g. holes), in which electrons recirculate multiple times before recombination with holes. The ratio of the carrier lifetime over the transit time determines the value of gain that can be reached. To demonstrate and even enhance the presence of photo-gain effects in 2D-based photodetectors, several approaches have been reported based on surface doping,[146] sensitizations with QDs,[147] perovskite[148,149] or metal nanostructures.[150] The common aim of those efforts has been to introduce or enhance the photogating effect. In the photogating effect, the minority carrier lifetime is prolonged trough localized trapping in trap states or sensitizing centers, that can be formed either on the 2D semiconductors themselves or via using sensitizer layers based on other material platforms. 17 The presence of such high gain comes at the expense of lower electrical bandwidth typically reported on the order of 1 kHz or lower. This feature determines then the employment of these detectors in applications that require high sensitivity but not very fast operation such as video imaging, sensing and steady-state spectroscopy applications. In this session, we discuss the different strategies developed so far to explore high photoconductive gain utilizing the photogating effect in 2D-hybrid photodetectors. 3.3.1. Photogating effect in graphene sensitized detectors Graphene has demonstrated its potential as photodetecting material in broadband and ultrafast technologies, however the monolayer thickness of graphene absorbs only 2.3 % of incident light.[57,58] The dominant photo-thermoelectric or bolometric effects in photo-response and fast photo-carrier recombination rates of a few picoseconds have limited the responsivity and sensitivity in graphene photodetectors due to the absence of gain mechanism and the limited absorption. To enhance the photon absorption, several approaches have been demonstrated such as the integration of optical microcavities,[151,152] optical waveguides[45] and the field enhancement by plasmons,[153] which have led to improved absorption of more than 60% and responsivity in the 20-130 mA W-1 range. Interested readers can see recent review focussing on this approach.[100] Here we discuss the graphene hybrid phototransistors sensitized with other 2D materials, colloidal quantum dots, perovskites and organic materials, etc, which can further improve the performance through the photogating effect. Graphene-CQDs or perovskite sensitized detectors: The first report of a highly sensitive graphene-based photodetector was based on sensitization with colloidal quantum dots (CQDs) of PbS.[147] CQDs have many unique properties that make them an ideal sensitizing pair for such as strong light absorption, broad absorption range from ultraviolet to SWIR and size-tunable bandgap through quantum confinement effect.[12] Their low-temperature and facile solution processing make them compatible with various substrates and large-scale, low-cost manufacturing processes. The concept of hybrid graphene-CQD phototransistors is shown in 18 Figure 5a. The detector consists of monolayer or bilayer graphene covered with a thin film of colloidal quantum dots. Upon light illumination, the photo-excited holes can transfer to graphene and drift by means of a voltage bias to the drain, while electrons remain trapped in the quantum-dot layer. These trapped carriers lead to a photogating effect, where the presence of these charges changes the conductance of graphene, and shifts the Dirac point to higher back- gate voltage (Figure 5b). Thanks to the high mobility of graphene and strong light absorption of CQDs, the hybrid phototransistors exhibited high EQE of 25%, high gain up to 108 corresponding to responsivity of 107 A W-1 in the shortwave-IR region up to 1.6 µm determined by the size of the CQDs.[147] The high quantum efficiency, fast video-imaging speed (<10ms) and high detectivity of 1013 Jones, taken together with its CMOS compatibility rendered this hybrid systems a promising platform for visible and SWIR photodetection applications. Nikitskiy et al. then further integrated an electrically active colloidal quantum dot photodiode atop a graphene phototransistor (Figure 5c). By applying a bias VTD across the CQDs film, the additional electrical field perpendicular to the graphene-CQDs interface increases the depletion width and charge collection efficiency. As a result, the EQE was improved in excess of 70% (Figure 5d) with a linear dynamic range of 110 dB and 3 dB bandwidth of 1.5 kHz.[154] Alternative sensitizers have then been employed in such architectures. As an example, topological insulator Bi2Te3 nanoplates were decorated on graphene and the responsivity of 35 A W-1 (gain of ~83) with response spectrum from visible to near-infrared (980 nm) and telecommunication band (1550 nm) has been achieved in view of the small bandgap of Bi2Te3.[155] Large bandgap quantum dots like CdS and ZnO were also integrated with graphene to form hybrid phototransistors, showcasing promising applications in UV detectors and image sensors with high sensitivity and gain of 107-109.[156,157] In a recent review,[158] various transition metal oxide (TMO) nanoparticles were also used in conjunction with graphene for different applications. Lee et al. reported graphene-perovskite (CH3NH3PbI3) hybrids phototransistors, which exhibited a responsivity of 180 A W-1 in the UV-Visible range, limited by the absorption 19 range of the perovskite layer.[159] Using synergistically plasmonic effects of metallic nanostructures, the performance in this hybrid system was two-fold improved.[160] Recently, graphene-MAPbI3 perovskite hybrid detectors have shown further progress with responsivity of 107 A W−1 and detectivity up to ~1015 Jones, thanks to the efficient light harvesting from perovskite sensitizers, its very long carrier diffusion length and the efficient charge separation at the interface.[161] Graphene-2D TMDs sensitized detectors: The strong optical absorption (1 × 107 m-1) and a visible-NIR range bandgap of 2D TMDs, make them natural partners of graphene for optically active heterostructures.[142] Roy et al. fabricated the graphene-on-MoS2 binary heterostructures and showed remarkable dual optoelectronic functionality, including highly sensitive photodetection and gate-tunable persistent photoconductivity.[162] The schematic diagram of the device is shown in Figure 5e, where the multilayer MoS2 acting as the photoactive layer lies underneath the graphene and source-drain electrodes located atop render graphene as the carrier transport layer. Under light illumination, the photo-generated electrons in MoS2 transferred into graphene under negative gate bias (Figure 5f) changing the conductance of graphene, while holes were trapped by localized states in the MoS2 acting as a local gate and giving rise to the photogating effect. The ultrahigh mobility of graphene (104 cm2 V-1 s-1) allowed the fast transit time of electrons, whereas holes residing longer time in MoS2 result in a very large photoconductive gain of ~4 × 1010 electrons per single photon and reported responsivity of 5 × 108 A W-1 in the visible at room temperature. At the expense of this gain, the response was very slow and displayed persistent photocurrent which the authors reported for a rewritable optoelectronic switch or memory.[162] Similar device structures have been afterwards reported with CVD grown graphene and MoS2, instead of the exfoliated ones.[163] Very recently, Mehew et al. reported a graphene-WS2 heterostructure-based photodetector encapsulated in an ionic polymer where WS2 underneath graphene acts as the photoactive layer and the ionic polymer serves as the top gate (Figure 5g).[164] Similar to Graphene-MoS2 devices, they also observed 20 significant photo-gain of 4.8 × 106, high responsivity of 106 A W-1 (Figure 5h) and high detectivity D* of 3.8 × 1011 Jones at a bandwidth of 150 Hz. The key aspect of this work is the significant screening of charge impurities due to the highly mobile ions of the top ionic polymer, which leads to the sub-millisecond response times and a -3 dB bandwidth of 1.5 kHz without the need of gate voltage pulsing. Graphene with tunnelling barrier: Besides the sensitization of graphene with 2D TMDs, CQDs and perovskite as photo-sensing layers, the introduction of a thin tunnel barrier in graphene double-layer heterostructures can also induce significant photogating effect.[165] The phototransistors comprise a pair of stacked graphene layers (top gate layer; bottom channel layer) sandwiching a thin tunnel barrier (5nm Ta2O5) (Figure 5i). The photo-excited hot carriers can tunnel efficiently into the nearby graphene layer, thereby minimizing hot carrier recombination. The asymmetric tunnelling barrier favours hot electron tunnelling from the top to the bottom graphene layer (Figure 5j). As a result, positive charges accumulate in the top graphene layer, leading to a photogating effect on the bottom graphene transistor, yielding very high responsivity over an ultra-broad spectral range. The electron accumulation in bottom graphene leads to the shift of Dirac point towards negative gate under light illumination shown in Figure 5k. The responsivity can reach 1000 A W-1 in visible and 1-4 A W-1 from NIR (1.3 µm) to MWIR range (3.2 µm) with bandwidth of 10-1000 Hz, rivalling state-of-the-art mid- infrared detectors and without the need of cryogenic cooling. They noted that the utilization of 2D TMDs (MoS2 or WS2) as tunnelling barrier could further enhance the interlayer hot carrier tunnelling and photogating effect and thus device performance.[165] Zhang et al. also reported high responsivity of 8.6 A W-1 compared to pure graphene phototransistors from the visible (532 nm) up to the mid-infrared (~10 µm) in a single pure graphene photodetector, by introducing electron trapping centres and by creating a bandgap in graphene through band structure engineering.[39] 3.3.2. Photogating effect in 2D TMDs based photodetectors 21 Despite the very high performance reported in hybrid graphene based phototransistors, their power consumption, electronic read-out schemes and noise are all determined by the absence of bandgap in graphene which leads to large dark current flow. To further improve sensitivity, power consumption and read-out integration, 2D TMDs have been instead considered as potential replacement of graphene for transistor channels sensitized either with CQDs or via doping modulation techniques. The use of semiconducting 2D TMDs channels is of particular promise for they enable the operation of the transistor in the depletion mode, offering thus the advantage of low leakage current in dark conditions by applying appropriate gate voltage. 3.3.2.1. Photogating effect in neat 2D TMDs detectors 2D TMDs with a band gap of 1-2 eV have emerged as a promising candidate for next generation of logic transistors, photodetectors and photo-harvesting devices.[62,63] Phototransistors based on a diversity of 2D TMDs including MoS2,[31,166-169] MoSe2,[170] WS2,[171,172] WSe2,[173,174] etc. with monolayer or multilayer structures have been reported with a range of performance reported depending on the quality and the nature of 2D-materials investigated. Among them, MoS2 is one of most studied 2D TMDs because of its outstanding properties in view of its high photon absorption efficiency, high carrier mobility up to 200 cm2 V-1 s-1 and large electrical on/off ratio over 108 as well as natural occurrence of MoS2 single crystals.[30,31] The reported MoS2 phototransistors have shown a very large variation in performance parameters because of the large surface-to-volume ratio, its sensitivity to the surrounding environment and the different substrate treatments. Yin et al. demonstrated the first monolayer MoS2 phototransistors with responsivity of 7.5 mA W-1 and fast response of 50 ms, the low responsivity is likely due to the absence of any gain mechanism.[166] To further enhance photo- response, many strategies have then been explored employing the photogating effect. Lopez-Sanchez et al. designed top HfO2 gated single layer MoS2 phototransistors with improved contact quality and MoS2 mobility, achieving high responsivity of 880 A W-1 but slow response of 0.6-9 s.[31] The high responsivity was attributed to the large photo-gain induced 22 by the trap states in MoS2 or MoS2/SiO2 bottom interface. Then Furchi et al. investigated the origin of photo-response in MoS2, considering both photoconducting and photogating effects.[175] The former contributes to the fast yet low response signal, and the latter is associated with the slow and high-gain response. The long-lived hole traps lying in MoS2 trap states or surface adsorbates (Figure 6a) can cause the photogating effect where holes are captured for a certain time while electrons recirculate many times leading to the high photoconductive gain and thus high responsivity. To achieve fast response speed together with high sensitivity, Kufer et al. proposed a robust passivation scheme by encapsulating MoS2 with HfO2 or Al2O3.[50] The isolation from ambient air improved electronic properties with suppressed hysteresis and fast response speed (Figure 6b) and had led to a gate tunable responsivity of 10-104 A W-1 and an experimentally measured sensitivity of 1011−1012 Jones with decay times of 10 ms. Another efficient strategy for the performance improvement is via gate dielectric engineering. Wang et al. integrated multilayer MoS2 phototransistors with ferroelectric P (VDF-TrFE) gate dielectric which favors ultrahigh local electrostatic field in the channel and passivation of the surface trap states. As a result the performance was significantly improved with responsivity of 2570 A W-1, fast response of ~2 ms and high sensitivity of ~1012 Jones.[176] Interestingly, the spectrum was extended from visible to 1550 nm likely due to the electrostatic field induced defect formation.[176] Lee et al. used a semi-transparent and conducting NiOx as gate electrodes directly on MoS2 channel without an insulator between them (Figure 6c);[177] this device, called metal-semiconductor field effect transistor (MESFET), exhibited extremely high intrinsic-like mobility of 500-1200 cm2 V-1 s-1 attributed to the reduced charge scattering effect (Figure 6d). They noted that the conducting carriers located at the insulator/MoS2 interface in standard MoS2-on-insulator structures, were unavoidably interfered by the interface traps and gate voltage, while the utilization of NiOx gate electrodes allowed scattering-suppressed transport. Due to the high mobility, the responsivity has reached as high as 5000 A W-1 with fast response time of 2 ms (inset of Figure 6d). The high responsivity with thousands A W-1 together with 23 fast speed operation in the above developed phototransistors has been largely due to improved carrier mobility. In addition to TMDs, other types of 2D semiconductors such as group IIIA, IVA, IVB and ternary metal chalcogenides and their photodetectors have also been developed and reviewedextensively elsewhere.[178] 3.3.2.2. Photogating effect in 2D TMDs sensitized detectors Building upon the promising findings of graphene-sensitized phototransistors and in an effort to expand the spectral coverage of 2D TMDs, leveraging at the same time the benefits of those that arise from the presence of a bandgap of around 1-2 eV, in using them as transistor channels, significant efforts were invested in developing high sensitivity infrared photodetectors based on 2D TMDs, sensitized with lower bandgap absorbers. MoS2-PbS CQD Hybrid detectors: In this first report of this class of detectors, a few-layer MoS2 channel was covered by PbS QDs on top (Figure 7a). A built-in field was created at the interface between P-type of PbS and N-type of MoS2, which can facilitate the separation of photo-generated carriers in the PbS QD absorber (Figure 7b).[179] The photoexcited electrons transfer into MoS2 and holes remain trapped in QDs, leading to a large gain of 106, high responsivity of 105-106 A W-1 and D* of 5 × 1011 Jones with extended spectral sensitivity up to 1.5 µm, determined by the absorption of the QDs. While this work had demonstrated the proof of concept and the fact that an efficient charge separating heterojunction can be formed between QDs and 2D semiconductors, the reported detector was suffering by high dark leakage current because of the serious electron doping of the MoS2 channel during the QD deposition process.[179] Kufer et al. then introduced a thin TiO2 buffer layer between MoS2 and PbS QDs[180] that played a twofold role: it first isolated the MoS2 channel from the environment and second, acted as an efficient photogenerated-electron acceptor buffer layer that funnelled the electrons from PbS QDs to the MoS2 channel. Interestingly, the buffer layer preserves the gate modulation of current in MoS2 by suppressing the high density of localized states at the interface. The on-off 24 ratio has been close to that in pristine MoS2, allowing a dark current as low as few tens of pA (Figure 7c). The device also exhibited fast response of ~10 ms, high EQE of 28% and a record D* of 5 × 1012 Jones from visible to NIR spectrum range. The interface engineering presented in this work discloses a new path to control interfaces and doping effects of 2D crystals-based hybrid devices. A similar detector concept based on MoS2 hybrids but with CH3NH3PbI3 perovskite[181,182] or graphene QDs[183] as sensitizers were also reported with improved performance characteristics. MoS2-HgTe CQD hybrid detectors: HgSe and HgTe colloidal quantum dots with even smaller bandgap than PbS, have been considered as a promising low-cost route for mid-IR and far-IR detection due to their tunable bandgap throughout the full infrared spectrum with favourable optical properties.[184-186] The responsivity in their neat photodetectors, however, has been limited to tens of mA/W due to the low carrier mobility, lack of sensitizing centers and the consequent absence of photoconductive gain. Thus, the MoS2-HgTe hybrid phototransistors with TiO2 buffer layer were proposed[187] (Figure 7d). Similar to the MoS2-PbS hybrid case, high gain and responsivity of ~106 A W-1 with fast speed on the order of ms, were achieved benefiting from the synergism of 2D MoS2 and 0D HgTe QDs. By operating in the depletion regime, the noise current was significantly suppressed leading to an experimentally measured D* of ~1012 Jones at a wavelength of 2 µm and room temperature (Figure 7e). The spectrum can be further extended by increasing the size of HgTe QDs. The sensitivity was two orders of magnitude higher than prior reports from HgTe-based photodetectors as well as existing commercially available technologies based in extended-InGaAs, InAs or HgCdTe that also require thermo-electric cooling,[9] demonstrating the great potential of hybrid 2D/QDs detector technology in mid-IR applications with compelling sensitivity. n-MoS2-p-MoS2 sensitized photodetectors: MoS2 doping technologies have enabled the polarity transition to P-type and fabrication of MoS2 based PN photodiodes but with.[119-121] Kang et al. utilized self-assembled monolayer (SAM)-based doping techniques on 2D TMDs 25 (MoS2 and WSe2). Upon such treatments carrier mobility and photodetection performance were improved by more than one order of magnitude over the pristine control devices.[146] Very recently, a novel device architecture composed of all-2D MoS2 was proposed to serve as an out- of-plane charge separating p-n MoS2 homojunction and an in-plane MoS2 phototransistor.[188] In this case the MoS2 homojunction plays the role of the sensitizing layer for the MoS2 transistor channel. Using low concentrated AuCl3 solution, the top layers of multilayer MoS2 (7-11 nm) were P-type doped while bottom layers remained N-type, thus an out-of-plane PN homojunction was formed (Figure 7f). The out-of-plane PN junction can serve as a sensitizing scheme, which can separate the photo-excited carriers efficiently and produce a photo-gain of >105 electrons per photon. The high gain and fast response of ms level taken together with low noise yields record and gate-tunable sensitivity up to 3.5 × 1014 Jones at bandwidth of 1 Hz and 10 Hz (Figure 7g). The same device concept can be in principle applied in other 2D semiconductors, particularly those of low bandgap, such as BP, that hold promise to extend the spectral coverage of the 2D materials realm. 3.3.3. 2D photodetectors sensitized by plasmonic nanostructures An alternative to semiconductor sensitizers for 2D phototransistors has been based on plasmonic metal nanostructures. Very intense and resonant absorption enhancement can be achieved by utilizing the strong local field concentration from plasmons at the resonant wavelength of the nanostructures. Upon decoration with metallic nanoparticles or nanoantenna arrays, the photocurrent in silicon and 2D materials-based photodetectors has been reported to increase at the plasmonic resonance of the sensitizers, enabling narrowband spectrally selective photodetectors.[189-192] For example, in a few-layer MoS2 detector sensitized with Au nanostructures, the recorded photocurrent was improved due to absorption enhancement in the MoS2 from the near field of Au nanoparticles.[193-196] Graphene nanodisks and Au plasmonic nanoantennas have also been reported to enhance the absorption efficiency in graphene from less than 3% to 30%.[197-202] 26 Another important property of plasmonic sensitizers is their ability to generate energetic or "hot" carriers, which can enable photocurrent generation from photons with energy below the bandgap of the 2D semiconducting channel. The incident light can couple into surface plasmons by nanoantennas and the nonradiative decay of the plasmons results in hot electrons that can transfer across the Schottky barrier at the metal -- semiconductor interface and be detected as a photocurrent[203-206] (Figure 8a). Hot-electron-induced photodetection has been reported in 2D graphene[200,201] and MoS2.[150,207,208] By scanning the laser beam along the MoS2 channel (Figure 8b), the spatially resolved scanning photocurrent with different photon energies was recorded as shown in Figure 8c. The strongest photocurrent response was observed at the MoS2- metal junction with extended spectrum coverage up to 1.55 µm due to the generated hot electrons from the metal to MoS2.[207] A bilayer MoS2 integrated with a plasmonic antenna array also exhibited sub bandgap photocurrent with a photo-gain of 105 and responsivity of 5.2 A W- 1 at 1070 nm, the gain in this case has been attributed to charge trapping in MoS2/metal or MoS2/SiO2 interfaces.[150] Based on the same concept a graphene-antenna sandwich photodetectors (Figure 8d) showed 800% enhancement of the photocurrent due to the hot electron transfer and direct plasmon-enhanced excitation of intrinsic graphene electrons.[200] When measuring the local photocurrent in the device, by performing a line scan of the excitation laser between the source and drain electrodes, an antisymmetric photocurrent response is observed (Figure 8e), showing that the plasmonic antenna particularly heptamer structures provide larger field enhancements, multiple hot spots and a greater yield of hot electrons. While intriguing, the concept of having resonant selective photodetectors based on plasmonic sensitizers has, till now, led to much lower performance over the semiconductor-sensitized counterparts due to the challenge of competing efficiently over the ultrafast hot carrier relaxation in metals and the generated heat losses. 3.3.4. 2D-organic hybrid based photodetectors 27 Organic semiconductors possess some very favorable properties particularly relevant in wearable (opto)electronic applications, in view of their optical properties, stretchable/flexible form factor and production-scalable methods.[209-211] An opportunity therefore exists in synergistically combining those with 2D materials for optoelectronic applications. Particular benefits arise from the compatibility of those two material platforms: 2D atomic crystals, for example, provide atomically flat and inert surfaces, that can be ideal for ordered self-assembly of organic molecules.[212] Moreover, high quality organic layers can form on top of 2D materials via thermal evaporation or spin/dip-coating in view of the vdW force interactions and the absence of dangling bond formation at their interface. The vdW interactions can allow the epitaxial growth of organic-based films with larger crystal grain size, thus enhancing the electronic properties of organic semiconductors. Graphene and h-BN have been demonstrated as an ideal template or dielectric substrate for C60 film, dioctylbenzothienobenzothiophene (C8- BTBT) and rubrene transistors with high carrier mobility exceeding 10 cm2/Vs.[213-216] Due to the efficient charge transfer at interface, the organic molecules can significantly tune the Dirac point of graphene, and change the transport behaviour from p-type to ambipolar and finally n- type, offering new functional devices.[217] Ultrafast charge transfer (6.7 ps) and long-lived charge-separated states (5.1 ns) have been observed in pentacene-MoS2 p-n heterojunctions.[218] By employing plasmonic metasurfaces in hybrid MoS2-organic heterojunction, the charge generation within the polymer is enhanced 6-fold and the total active layer absorption band is increased.[219] These features suggest significant promise for 2D-organic hybrid heterostructures in photovoltaics and photodetectors. Recently, a new graphene-organic semiconductor based vertical field effect transistors (VFETs) (Figure 9a) was demonstrated to exhibit high on-off ratio up to 105 (Figure 9b) thanks to the partially-screened field effect and selective carrier injection through graphene.[220,221] Solution-based graphene nanomaterials were modified with organic molecules to open a bandgap due to the charge redistribution between the C-C bonds, which in turn has led to promising photodetector performance 28 characteristics and high responsivity of 2.5 A/W in the mid-infrared spectral region (3-5 μm).[222] Similar to 2D/QDs detectors, the organic molecules or polymers can also be used to sensitize the surface of 2D channels thus producing a significant photoconductive gain through the photo-gating effect. Lee et al. fabricated a hybrid photodetector comprising organic dye molecules (rhodamine 6G) and graphene with broad spectral photo-response and responsivity of 460 A W-1.[223] Liu et al. have epitaxially grown small molecule C8-BTBT on top of graphene using a CVD approach and fabricated C8-BTBT/graphene hybrid phototransistors which can exhibit photo responsivity of 104 A/W, photo-gain larger than 108 and time response of 25 ms.[224] Under the same device concept, graphene-polymer semiconductor (P3HT or PTB7) hybrid phototransistors (Figure 9c) have exhibited responsivity exceeding 104 A W-1 and fast temporal response of ~7.8 ms. In this device, the use of a self-assembled-monolayer (SAM) functionalization to effectively remove surface traps and charged impurities between graphene and SiO2 substrate, further improved responsivity up to 105 A W-1 (Figure 9d).[225,226] Tan et al. demonstrated that the utilization of piezoelectric (PZT) substrate can improve the photocurrent of graphene-P3HT photodetectors by 10 times compared to that based on SiO2 substrate due to the enhanced separation of photogenerated electrons and holes under the electric field of the polarization from the piezoelectric substrate.[227] It is noted that carbon nanotubes were also combined with graphene to form hybrid photodetectors with broadband spectal response (covering 400 -- 1,550 nm), high responsivity of 4100 A W-1 and a fast response time of 100 µs.[228,229] Besides graphene, 2D TMDs such as MoS2 and WS2 have also been combined with organic materials to achieve hybrid photodetectors.[230-232] For instance, Yu et al. reported dye- sensitized MoS2 photodetectors utilizing a single-layer MoS2 treated with rhodamine 6G (R6G) organic dye molecules. The hybrid devices exhibited responsivity of 1.17 A W-1, detectivity of 1.5×107 Jones, and spectral converage up to 980 nm.[233] Overall, 2D-organic semiconductor hybrid systems have attracted a lot of attention and the recent progress in the field has been extensively described elsewhere.[234,235] 29 4. Demonstrated applications of 2D based photodetectors 4.1. Flexible electronics and detectors Flexible electronics and sensors have opened a new realm of functionalities in bendable and portable device technologies and applications such as wearable health monitors, electronic skins, portable touch panels as well as Internet-of-things sensors and transducers seamlessly integrated in an ubiquitous manner. Conventional semiconductors (a-Si) and metal oxide thin- films have been explored and currently used in flexible electronics, yet some of their limitations such as opaqueness, thickness scalability and high-cost manufacturing[236,237] need to be overcome for high performance, low-power consumption, small footprint devices. The nanoribbon or nanomembranes based on single crystalline inorganic materials have emerged as main contenders in high performance bendable and stretchable electronic applications.[238] Organic based semiconductors have also been recently developed particularly in transparent and stretchable optoelectronics for biomedical applications such as electronic skins, health monitoring, medical implants and human-machine interface.[209-211,239-242] Efforts on more robust and higher carrier mobility of organic electronics are still ongoing.[243,244] Along this direction, 2D materials offer an alternative promising route towards these goals.[21,245] Leveraging its extraordinary optical and electronic properties and its large-scale manufacturability, graphene has been widely reported as soft and transparent electrode in modern flexible electronics.[246,247] But the absence of bandgap limits the utilization of graphene in digital electronics. To address this, graphene-organic semiconductor based VFETs were demonstrated to exhibit high-off ratio up to 105 with a bending radius of < 1mm.[220,221] Meanwhile, 2D TMDs such as MoS2 have been proposed and developed in logic circuit components for flexible electronics.[248-250] The MoS2 transistors on flexible substrates have shown ON/OFF ratios over 107 and mobility of ~30 cm2 V-1 s-1, which is on par to that on rigid substrate.[248] 2D heterostructures combined with graphene as electrodes, TMDs as channel 30 materials and hBN as gate dielectric have also been demonstrated as promising architectures for all-2D based flexible electronics and detectors (Figure 10a).[249,251,252] The graphene/MoS2 flexible photodetectors have reached responsivity of 45.5 A W-1 with a gain of ~105 even upon bending curvature of 1.4 cm.[253] Liu et al. fabricated fully transparent and flexible graphene P- N junction-based IR photodetectors through a facile chemical doping technique.[254] Recently, highly stretchable graphene devices by intercalating graphene scrolls in between graphene layers (Figure 10b) were also reported, where the graphene scrolls provide conductive paths to bridge cracks in the graphene sheets, thus maintaining high conductivity under strain. The corresponding all-carbon based flexible transistors exhibited a transmittance of >90% and retained 60% of their original current output at 120% strain, showing superior performance in terms of mobility, on/off ratio and being highly stretchable.[255] BP-based transistors on highly bendable polyimide substrate (Figure 10c) were reported with mobility up to 310 cm2 V-1 s-1 and excellent mechanical durability.[256,257] Hybrid flexible photodetectors based on WSe2, graphene-carbon nanotube and ZnS-MoS2 heterojunctions were also reported with responsivity of 50 A W-1 and fast response (~40 ms).[258-261] 2D materials are expected to play a big role in the next generation of flexible electronics and optoelectronics in view of their atomically thin form factor and their outperforming electronic properties over those of current flexible electronic semiconductors.[21,22,262] 4.2. Silicon and CMOS Integration Conventional silicon semiconductor technology has been the cornerstone of modern electronics and may continue so for the near future. One of the first generation of applications of 2D materials may therefore be in combination with standard silicon technology in providing new functionalities to silicon or replacing some manufacturing processing steps needed to implement junctions on silicon. 2D materials combined with conventional Si has therefore been explored and demonstrated a series of promising electronic and photonic devices. Graphene barristors were realized by combining silicon to form a graphene-silicon Schottky barrier.[263] 31 Large modulation on the barristors current with on/off ratio of 105 and ideality factor of 1.1 was achieved by adjusting the gate voltage to control the barrier height, which overcame the key obstacle in graphene-based electronics. Recently, Wan et al. presented a self-powered, high- performance graphene-enhanced ultraviolet silicon Schottky photodetector with a Al2O3 anti- reflection layer (Figure 10d). At zero-biasing (self-powered) mode, the photodetectors exhibit high photo-responsivity (0.2 A W-1) in the ultraviolet region, fast time response (5 ns) and high specific detectivity (1.6 × 1013 Jones) comparable to that of state-of-the-art Si, GaN, SiC Schottky photodetectors. They also demonstrated high stability of the device exceeding 2 years.[264] MoS2 has also been integrated with silicon to form a heterojunction that has then been exploited in silicon photodiodes, eliminating the need of doping silicon.[265-268] MoS2-Si heterojunctions exhibit high sensitivity of 1013 Jones with fast response times of ~3 μs due to the efficient built-in field at their interface.[266] Lopez-Sanchez et al. reported an avalanche photodiode based on MoS2/Si heterojunctions with multiplication factor exceeding 1000 and respectably low noise floor, enabling low-noise photon counters in 2D APDs.[46] The integration of graphene detectors with a silicon-on-insulator waveguide (Figure 10e) gave rise to the enhancement of graphene absorption and thus photodetection efficiency through coupling the evanescent field from the optical waveguide mode to the graphene absorber, preserving both high speed and broad spectral bandwidth.[269] These detectors had responsivity exceeding 0.1 A W-1 with response rates exceeding 20 GHz and were demonstrated successfully in an instrumentation-limited 12 Gbit/s optical data link. Similarly, Youngblood, et al. also reported a multilayer black phosphorus photodetector integrated on a silicon photonic waveguide. High responsivity of ~0.6 A W-1 with large bandwidth over 3 GHz operating in the near-infrared telecom band was achieved.[41] A silicon waveguide-integrated bilayer MoTe2 was recently demonstrated to act as both a light source and a photodetector on a single chip platform (Figure 10f),[270] paving the way for Si-photonics enabled by 2D materials. Besides silicon, the integration of MoS2 on germanium (Ge) was reported in a band-to-band tunnelling transistor 32 (Figure 10g) with subthreshold swing as low as 3.9 mV/decade and an average value of 31.1 mV per decade which overcome the theoretically predicted limits in conventional bulk transistor configurations.[271] Complementary metal -- oxide semiconductor (CMOS) based integrated circuits are at the heart of the technological revolution in modern society, CMOS integration of any new technology introduced warrants its easier industrial acceptance and market entry leveraging the mass production capabilities and low cost manufacturing processes developed for CMOS electronics. Individual CMOS-compatible graphene photodetectors covering the fibre-optic telecommunication bands[272] and graphene/silicon CMOS hybrid hall integrated circuits (ICs) via low temperature process[273] have been reported. Very recently, graphene-QD hybrid photodetectors have been integrated with CMOS read-out circuits to demonstrate a broadband CMOS-based digital camera capable of capturing both visible and infrared light.[274] The schematic diagram of CMOS integration of graphene with 388 × 288 pixel image sensor read- out circuit is shown in Figure 11a. Graphene was transferred on to a CMOS die and connected with the bottom readout circuity by vertical metal interconnects. Then graphene was patterned to define the pixel shape followed by sensitization of PbS QDs on top (Figure 11b). This hybrid technology has reached high responsivity of 107 A W-1, measured D* of 1012 Jones and fast switching speed of 0.1-1ms that were achieved in the spectral range of 300-1800 nm. The successful integration of 120,000 pixel photodetectors in a single focal plane array has enabled high-resolution imaging in a broadband spectrum under weak light atmospheric conditions or environmental conditions that standard silicon cameras would have not been able to capture. (Figure 11c). Zanjani et al. also demonstrated monolithically integrated CMOS-graphene gas sensors combining the superior gas sensitivity of graphene with the low power consumption and low cost silicon CMOS platform.[275] The graphene -- CMOS integration is pivotal for incorporating 2D materials into the next generation of sensor arrays and CMOS imaging systems. The successful integration of 2D materials on a silicon photonics and microelectronic 33 platform with high performance,[276] offers the possibility of CMOS integrated graphene and related 2D materials (TMDs, BP, etc.) for commercial optoelectronic applications. 5. Future outlook In this concluding section, we first compare the performance between typical 2D and hybrid based photodetectors and commercial silicon, germanium, InGaAs and HgCdTe photodetectors.[9] Figure 12a shows the responsivity versus the response time, demonstrating comparable performance. Graphene and BP photodetectors exhibit ultrafast speed and comparable responsivity. For hybrid and TMDs based photodetectors possessing gain, the responsivity is much higher but the response speed is limited on the order of milliseconds. Figure 12b also shows the detectivity versus the wavelength for different classes of photodetectors. The detectivity of 2D based photodetectors is comparable or even superior than that of conventional ones in the SWIR spectral range at room temperature. Overall, there is still a large room for improvement for 2D based photodetectors particularly aiming at faster response speed and extension of their spectral range. In the following, we discuss in brief the future outlook and some of the challenges needed to be addressed to further boost performance, technology and manufacturing readiness levels in order to promote 2D material photodetector technologies towards commercial applications. 5.1. Improvement in linear dynamic range The linear dynamic range (LDR) of graphene photodetectors are limited to only 7.5 dB due to their intrinsic hot-carrier dynamics, which causes deviation from a linear photoresponse at low incident powers.[29] TMDs and TMD/QDs hybrid based photodetectors also suffer from relatively low LDR due to their dependence of dynamic range on the density of the sensitizing states as well as the parabolic nature of the density of states in the TMDs.[31,180,187] As compared to 2D materials, state-of-the-art conventional (Si, Ge, GaAs, etc.) photodetectors currently exhibit a linear response over a larger range of optical powers. High LDR in 2D-based 34 photodetectors is thus needed for high-resolution sensing and video imaging applications. To address this issue in graphene detectors, Sanctis et al. defined p-p´ junctions by laser-assisted displacement of FeCl3 in FeCl3 intercalated few-layer graphene, these junctions can quench the hot-carrier effects and exhibit photocurrent signals from purely photovoltaic (PV) effect. As a result, they achieved an extraordinary LDR of 44 dB which is 4500 times larger than that of previously reported graphene photodetectors.[277] Their detectors also exhibited high stability under atmospheric conditions and a broad spectral response from ultraviolet to mid-IR wavelength. In 2D-QDs hybrid phototransistors, Nikitskiy et al. employed an electrically active QDs photodiode as the sensitizing element instead of a passive sensitizer atop graphene and achieved a significant enhancement of the LDR up to 110 dB, which was evidenced by a flat responsivity across the wide range of power density.[154] 5.2. Contact and mobility engineering Low contact resistance in 2D material-based electronics and optoelectronics is critical for achieving efficient carrier injection, high responsivity and high frequency operation.[31,278] However, a large Schottky barrier is usually formed between 2D materials and conventional 3D metals due to the lack of chemical bonding on the chemically inert 2D surface, the large energy offset and Fermi level pinning at the interface.[279] Although optimized metal contacts have been studied in 2D materials,[280,281] new contact schemes towards lower contact resistance are needed to further enhance the device performance. Due to the finite density of states and its chemical inertness, graphene has been reported as an ideal electrode on other 2D semiconductors such as MoS2 with a barrier-free contact via tuning the work function of graphene. The low Ohmic contact resistance leads to higher photo-response and measured field effect mobility.[282-284] The one-dimensional edge contact geometry was also developed to make a high-quality electrical contact between graphene and metal, leading to outperforming electronic performance such as the phonon-scattering limited room-temperature mobility up to 140,000 cm2 V-1 s-1.[285] Recently, low-resistance metal -- semiconductor contacts (200-300 Ω 35 µm) were also obtained by interfacing semiconducting 2H-MoS2 layers with 1T-MoS2.[286] The utilization of thin TiO2 layer between MoS2 and metal was also reported to suppress trap state density at the contact interface and thus enhance both extracted mobility and photoresponsivity.[287] Overall, the performance of 2D photodetectors can be drastically improved by these contact engineering schemes due to more efficient photo-carrier extraction and recirculation through the external circuit. Mobility is another important challenge in 2D materials that determines to a large extent their performance as transistors or photodetectors. The utilization of high-ƙ dielectrics as strongly coupled top-gate insulators has improved the room temperature mobility up to the order of 100 cm2 V-1 s-1 due to dampening of Coulomb scattering from the charged traps or impurities.[288,289] This taken together with the low contact resistance through contact engineering, as discussed above, is expected to lead to even higher extracted mobility values for 2D-TMD transistors. In addition, strain effects have been theoretically demonstrated to improve the mobility in MoS2 by one order of magnitude through the suppression of electron-phonon coupling.[290] The mobility engineering in 2D materials is very promising in terms of realization high-speed and high-sensitivity 2D photodetectors. For example, higher mobility leads to faster photocarrier transit time from source to drain, enabling higher photoconductive-gain for a given carrier lifetime. Besides mobility, the subthreshold swing and ON/OFF ratio in 2D based transistors particularly 2D/QDs hybrids phototransistors also play a critical role since they determine the ultimate sensitivity and linearity of corresponding photodetectors when operating around the threshold voltage.[49] 5.3. Challenges in 2D hybrid platforms 2D hybrid systems employing synergistically other materials such as 2D TMDs, CQDs or perovskites have already shown performance on par with detector technology based on silicon and InGaAs for the Vis−SWIR range and have the potential to outperform them in sensitivity and cost. Graphene-PbS CQD hybrid phototransistors have reached specific detectivity of 36 7×1013 Jones with fast speed below 1ms from visible to SWIR range, enabling CMOS compatible, broadband, high-resolution imaging systems.[274] MoS2 based phototransistors have also exhibited detectivities on the order of 1012-1014 Jones at visible and SWIR beyond 2 µm, realized by chemical doping of MoS2 or HgTe CQD sensitization.[187,188] Despite the great progress in this new photodetector platform, significant further improvements are to be expected in terms of optimizing 2D transistor channels, sensitizers and interface towards more sensitive, broadband and faster photodetectors. In 2D hybrid platforms, the 2D channels act as carrier transport layers and therefore must meet the high mobility and gate current modulation features, which contribute to the high gain and low noise, respectively. Graphene has very high mobility but being a semimetal also suffers from large dark current. Strategies to open a bandgap in graphene such via doping or nanostructuring can give rise to lower noise and higher sensitivity. 2D TMDs on the other hand, albeit limited by lower carrier mobility, they benefit from very low dark current when operated in depletion mode, offering the potential for overall improvement in the sensitivity of the detector. Potential pathways to improve their performance lie on enhancing their mobility by low contact resistance or mobility engineering using high-ƙ dielectrics. Alternative or synergistically to that, the use of new emerging semiconducting 2D materials such as BP, with a moderate bandgap of 0.3 eV in bulk and high mobility of 1000 cm2 V-1 s-1, may also be considered.[36] 2D topological insulators such as Bi2Se3 with symmetry protected surface states and insulating or semiconducting internal bulk band structure have been reported with high mobility up to 1750 cm2 V-1 s-1 at room temperature, posing them as new promising material platforms for transistor channels towards broadband, highly sensitive photodetectors.[145,291] As far as the sensitizers is concerned, a high absorption coefficient with large and ideally tunable spectral coverage is a prerequisite. In addition to this and in order to facilitate efficient charge transfer to the transistor channel, thereby high quantum efficiency, the sensitizer layer should also possess favourable electronic properties in terms of carrier diffusion length, carrier 37 mobility, doping and minority carrier lifetime. Colloidal quantum dots in that aspect, can serve as strong absorbers with broadband spectral range from the visible to the infrared region. Particularly the HgSe and HgTe QDs can cover the full spectrum up to 20 µm,[184-186] which offers the possibility of 2D/QDs hybrid systems with sensitivity extending in the MWIR and even LWIR. The type-II band alignment between the CQDs sensitizer layer and the 2D channel can also be engineered by using different capping ligands, which can further extend the depletion width and improve the photocarrier transfer toward the transport channel thereby enhancing the charge collection efficiency.[292] Recently, solution-processed perovskites were also demonstrated with strong light absorption and remarkable diffusion lengths exceeding 10 μm,[293] suggesting them an ideal sensitizer to reach high optical absorption and high quantum efficiency, although limited in the visible and near infrared part of spectrum. In the future, new environmentally friendly CQDs with smaller bandgap, such as Ag2Se QDs with absorption wavelength up to 6.5 µm[294] or other semiconducting 2D materials such as BP can be employed as sensitizers in the 2D hybrid based IR detectors. The interface between 2D channels and sensitizers plays a key role for the performance of the hybrid detectors. The introduction of a thin TiO2 buffer layer has been instrumental in avoiding the electrical doping on the channel from the environment and allowing nearly full modulation of its conductivity upon sensitization.[180] The use of self-assembled monolayers at the interface may also lead to efficient cross-linking between QDs and 2D channels and simultaneously passivate electronic defect states.[295,296] The transformation of the sensitizer layer from an electrically passive one to an active one can really lead to new configurations and even better performance. This concept has been demonstrated in a graphene -- PbS QD photodetector in which the PbS layer, by the use of a top electrode has been transformed into a Schottky photodiode.[154] As a result the charge transfer efficiency from the sensitizer to the graphene was improved leading to an EQE of 80% and the linear dynamic range of the detector increased to 110 dB. Such an approach led to a detector that provided the combined benefits of a 38 photodiode and a phototransistor. Based upon this concept several new photodetector architectures can be implemented as a sensitizing layer, such as p-n or heterojunction diodes leading to even better characteristics in terms of speed, noise etc. This concept can also be applied in 2D TMDs-QDs hybrid systems. 5.4. Noise suppression The suppression of noise in photodetectors is a very demanding and challenging task yet it is required for high sensitivity. Generally, electrical noise is composed of four intrinsic sources: thermal or Johnson noise, shot noise, generation-recombination (G-R) noise and flicker or 1/f noise. Thermal and shot noise components are frequency independent (thus typically called white noise) and originate from the random motion of charge carriers. The spectral density of thermal noise is given by Nyquist's formula 𝑆𝐼(𝑓) = 4𝑘𝐵𝑇/𝑅 , where kB is Boltzmann's constant and T is temperature, R is resistance of detector. Whereas the shot noise is determined by Schottky's theorem 𝑆𝐼(𝑓) = 2𝑞 < 𝐼 >, where < 𝐼 >is the average value of the electrical current. 1/f noise is frequency dependent and caused by electrostatic fluctuations resulting in charge carrier number fluctuation or mobility fluctuation or their combination. The origin of 1/f noise is not always known but has been ascribed, depending on the material platform, to metal- semiconductor interfaces, trap states or effects from semiconductor edges and dislocations.[297] G-R noise bulges are superimposed on the 1/f spectrum at low frequency and associated with trap states of different time constants. In 2D material-based photodetectors, the trap states in the substrate or the gate oxide can capture or emit carriers from/to atomically thin channels leading to large current fluctuations and significant 1/f noise which currently determine the sensitivity of those detectors.[297] Bilayer or few layer graphene has shown lower 1/f noise than single layer due to their stronger electrostatic screening ability,[298] whereas suspended graphene has exhibited very low noise due to the removal of the oxide substrate effects underneath the channel.[299] Defect-free and atomically flat 2D h-BN as a substrate has also been reported to lead to noise suppression in graphene by 39 10-100 times over neat graphene devices.[300] Further improvements can be reached by encapsulating graphene with h-BN on both sides and by making one-dimensional edge contacts[285,301] In semiconducting 2D TMDs, ultralow dark current can be achieved in view of the atomically thin profile of the channel and by operating the transistor in depletion mode, thus shot noise is not expected to be a limiting noise factor in these devices. However, the 1/f noise is still significant due to the trap states, environmentally induced Coulomb scattering centers as well as contact barriers.[302,303] The strategies for suppressing noise in graphene can also be applied in TMD-based devices and this remains to be demonstrated. High quality contact and encapsulation with high-ƙ dielectric were demonstrated to suppress the noise in MoS2.[304,305] Besides, for CVD grown 2D graphene and TMDs, the existing numerous defects and grain boundaries can act as scattering centers that significantly increase the 1/f noise.[306,307] The growth of high quality 2D materials with lower lattice disorder remains a challenge that needs to be addressed also for low noise and high sensitivity detectors. Although 1/f noise is useful to detect environmental chemical species,[308] it ultimately limits the sensitivity of detectors. Readers can see recent reviews focused on noise in 2D materials.[297,309] 5.5. Wafer-scale production To realize widely commercial applications of 2D materials in modern electronics and optoelectronics, wafer-scale growth or processing is instrumental. So far, many groups have developed growth methods towards wafer-scale production of 2D materials. As one of the most studied 2D materials, large area graphene with >95% monolayer uniformity and low defect density has been achieved on 300 mm Si wafer covered with one copper layer using chemical vapour deposition (CVD) method,[310,311] and now it is already commercially available. Polycrystalline graphene with 30-inch size (Figure 13a) can also be produced on flexible substrates such as polyethylene terephthalate (PET) via roll-to-roll processing (Figure13b),[312] enabling massive high-throughput processing of flexible electronics. 2D TMDs, as semiconducting films with high quality, scalable size and controllable thickness, have also been 40 grown using the CVD method.[79-81] Kang, et al. achieved 4-inch wafer-scale films of monolayer MoS2 and WS2 (Figure 13c), grown directly on insulating SiO2 substrates using a newly developed metal -- organic chemical (MOCVD) method (Figure 13d).[35] The resultant monolayer films possess excellent spatial homogeneity over the entire wafer and appreciable mobility of 30 cm2 V-1 s-1 at room temperature. Wafer-scale growth of other 2D semiconducting materials such as WSe2,[313] InSe[314] and GaS[315] has also been demonstrated pointing to the generalization of wafer scale manufacturability of 2D materials. As an important dielectric, few layer h-BN has also been reported based on wafer-scale and wrinkle-free epitaxial growth using high-temperature and low-pressure CVD method.[316] This demonstrated progress on wafer- scale growth of a variety of 2D materials keeps up the promise for the next generation of low cost, easy-processing, flexible and CMOS compatible optoelectronic platforms enabled by 2D materials. Yet, CVD grown 2D materials still exhibit lower quality than small-scale exfoliated counterparts, thus further improvement is still foreseen in terms of suppressed lattice disorder, higher mobility and stronger photoluminence for CVD grown 2D materials. Efficient transfer techniques have been developed[317-319] to stack 2D building blocks towards multifunctional heterostructures, yet issues related to wrinkle formation, contamination and precise control of stacking orientation during the transfer process should be addressed to achieve such functionalities and exploit the full potential of the 2D family, in large scale processes. An alternative approach to large scale manufacturing may rely on printing from solution, a method that may be useful when cost is critical and the requirements for quality are more relaxed. In this aspect, Coleman et al. produced large amount of 2D nanosheets in liquid phase by liquid exfoliation method and dispersed them in common solvents, which can be deposited as individual flakes or formed into films.[320] Carey et al. demonstrated fully inkjet-printed 2D- material active heterostructures with graphene and h-BN inks, and fabricated all inkjet-printed flexible and washable field-effect transistors.[82] McManus et al. recently proposed a new approach to achieve inkjet-printable, water-based, two-dimensional crystal formulations 41 including graphene, TMDs and h-BN (Figure 13e), and fabricated high performance of logic memory arrays (Figure 13f) based on multi-stack films.[83] In summary, we have summarized concisely the recent progress in the field of photodetectors based on 2D materials including graphene, TMDs and BP as well as their heterostructures also hybridly with 0-dimensional or 3-dimensional material platforms. The presence of photoconductive gain or not has led to the classification of two classes of 2D-based photodetectors. One is heterojunction based photodiodes with both in-plane and out-of-plane configurations, which operate as photodiodes exhibiting fast response and high quantum efficiency. The photo-thermoelectric and bolometric detectors can also be applied in 2D materials particularly graphene, enabling the broadband detection from MWIR to submillimetre wavelength range. Another is hybrid based phototransistors, possesing photoconductive gain by combining 2D materials or CQDs, perovskite, metal nanostructures and organic semiconductors. The high performance taken together with the advantages of low-cost and easy-processing, flexibility, integrability with silicon technologies or CMOS compatibility, have lend 2D materials a very promising future in next generation of photodetector applications. Acknowledgements We acknowledge financial support from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (grant agreement No 725165), the Spanish Ministry of Economy and Competitiveness (MINECO) and the "Fondo Europeo de Desarrollo Regional" (FEDER) through grant TEC2017-88655-R. This work was also supported by by European Union H2020 Programme under grant agreement n°696656 Graphene Flagship. We also acknowledge financial support from Fundacio Privada Cellex, the CERCA Programme and the Spanish Ministry of Economy and Competitiveness, through the "Severo Ochoa" Programme for Centres of Excellence in R&D (SEV-2015-0522). 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)) References 42 [1] F. C. Lussani, R. F. C. Vescovi, T. D. Souza, C. A. P. Leite, C. Giles, Rev. Sci. Instrum. 2015, 86, 063705. [2] R. A. Bartels, A. Paul, H. Green, H. C. Kapteyn, M. M. Murnane, S. Backus, I. P. Christov, Y. Liu, D. Attwood, C. Jacobsen, Science 2002, 297, 376. [3] B. J. Zeskind, C. D. Jordan, W. Timp, L. Trapani, G. Waller, V. Horodincu, D. J. Ehrlich, P. Matsudaira, Nat. Methods 2007, 4, 567. [4] B. Fowler, C. Liu, S. Mims, J. Balicki, W. Li, H. Do, J. Appelbaum, P. Vu, Proc. SPIE 2010, 7536, 753607. [5] J.-E. Källhammer, Nature Photon. 2006, 1, 12. [6] S. Kim, Y. T. Lim, E. G. Soltesz, A. M. D. Grand, J. Lee, A. Nakayama, J. A. Parker, T. Mihaljevic, R. G. Laurence, D. M. Dor, L. H. Cohn, M. G. Bawendi, J. V. Frangioni, Nat. Biotechnol. 2004, 22, 93. [7] J. M. Kahn, J. R. Barry, Proc. IEEE 1997, 85, 265. [8] V. Formisano, S. Atreya, T. Encrenaz, N. Ignatiev, M. Giuranna, Science 2004, 306, 1758. [9] E. L. Dereniak, G. D. Boreman, Infrared Detectors and Systems. Wiley, New York, USA 1996. [10] A Rogalski, Infrared Physics & Technology 2002, 43, 187. [11] M. D. Eisaman, J. Fan, A. Migdall, S. V. Polyakov, Rev. Sci. Instrum. 2011, 82, 071101. [12] G. Konstantatos, I. Howard, A. Fischer, S. Hoogland, J. Clifford, E. Klem, L. Levina, E. H. Sargent, Nature 2006, 442, 180. [13] J. P. Clifford, G. Konstantatos, K. W. Johnston, S. Hoogland, L. Levina, E. H. Sargent, Nature Nanotech. 2009, 4, 40. [14] R. Nie, X. Deng, L. Feng, G. Hu, Y. Wang, G. Yu, J. Xu, Small 2017, 13, 1603260. [15] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, A. A. Firsov, Science 2004, 306, 666. [16] Q. H. Wang, K. Kalantar-Zadeh, A. Kis, J. N. Coleman, M. S. Strano, Nature Nanotech. 2012, 7, 699. [17] H. Zhang, ACS Nano 2015, 9, 9451. [18] S. J. Kim, K. Choi, B. Lee, Y. Kim, B. H. Hong, Annu. Rev. Mater. Res. 2015, 45, 63. 43 [19] W. S. Yun, S. W. Han, S. C. Hong, I. G. Kim, J. D. Lee, Phys. Rev. B 2012, 85, 033305. [20] A. Castellanos-Gomez, M. Poot, G. A. Steele, H. S. J. van der Zant, N. Agraït, G. Rubio- Bollinger, Adv. Mater. 2012, 24, 772. [21] D. Akinwande, N. Petrone, J. Hone, Nature Commun. 2014, 5, 5678. [22] L. Gao, Small 2017, 13, 1603994. [23] S. B. Desai, G. Seol, J. S. Kang, H. Fang, C. Battaglia, R. Kapadia, J. W. Ager, J. Guo, A. Javey, Nano Lett. 2014, 14, 4592. [24] G. H. Ahn, M. Amani, H. Rasool, D.-H. Lien, J. P. Mastandrea, J. W. Ager III, M. Dubey, D. C. Chrzan, A. M. Minor, A. Javey, Nature Commun. 2017, 8, 608. [25] J. Martín-Sánchez, R. Trotta, A. Mariscal, R. Serna, G. Piredda, S. Stroj, J. Edlinger, C. Schimpf, J. Aberl, T. Lettner, J. Wildmann, H. Huang, X. Yuan, D. Ziss, J. Stangl, A. Rastelli, Semicond. Sci. Technol. 2018, 33, 013001. [26] K. F. Mak, C. Lee, J. Hone, J. Shan, T. F. Heinz, Phys. Rev. Lett. 2010, 105, 136805. [27] K. F. Mak, K. He, J. Shan, T. F. Heinz, Nature Nanotech. 2012, 7, 494. [28] K. F. Mak, K. He, C. Lee, G. H. Lee, J. Hone, T. F. Heinz, J. Shan, Nature Mater. 2012, 12, 207. [29] T. Mueller, F. Xia, P. Avouris, Nature Phonon. 2010, 4, 297. [30] B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, A. Kis, Nature Nanotech. 2011, 6, 147. [31] O. Lopez-Sanchez, D. Lembke, M. Kayci, A. Radenovic, A. Kis, Nature Nanotech. 2013, 8, 497. [32] A. K. Geim, I. V Grigorieva, Nature 2013, 499, 419. [33] A. Pospischil, M. Humer, M. M. Furchi, D. Bachmann, R. Guider, T. Fromherz, T. Mueller, Nature Phonon. 2013, 7, 892. [34] N. Youngblood, M. Li, Nanophotonics 2016, 6, 1205. [35] K. Kang, S. Xie, L. Huang, Y. Han, P. Y. Huang, K. F. Mak, C.-J. Kim, D. Muller, J. Park, Nature 2015, 520, 656. 44 [36] L. Li, Y. Yu, G. J. Ye, Q. Ge, X. Ou, H. Wu, D. Feng, X. H. Chen, Y. Zhang, Nature Nanotech. 2014, 9, 372. [37] L. Tao, E. Cinquanta, D. Chiappe, C. Grazianetti, M. Fanciulli, M. Dubey, A. Molle, D. Akinwande, Nature Nanotech. 2015, 10, 227. [38] W. Xia, L. Dai, P. Yu, X. Tong, W. Song, G. Zhang, Z. Wang, Nanoscale 2017, 9, 4324. [39] Y. Zhang, T. Liu, B. Meng, X. Li, G. Liang, X. Hu, Q. J. Wang, Nature Commun. 2013, 4, 1811. [40] V. Ryzhii, M. Ryzhii, N. Ryabova, V. Mitin, T. Otsuji, Jpn J. Appl. Phys. 2009, 48, 04C144. [41] N. Youngblood, C. Chen, S. J. Koester, M. Li, Nature Photon. 2015, 9, 247. [42] G. Konstantatos, E. H. Sargent, Nature Nanotech. 2010, 5, 391. [43] F. Bonaccorso, Z. Sun, T. Hasan, A. C. Ferrari, Nature Photon. 2010, 4, 611. [44] F. Xia, T. Mueller, Y.-M. Lin, A. Valdes-Garcia, P. Avouris, Nature Nanotech. 2009, 4, 839. [45] X. Gan, R.-J. Shiue, Y. Gao, I. Meric, T. F. Heinz, K. Shepard, J. Hone, S. Assefa, D. Englund, Nature Phonon. 2013, 7, 883. [46] O. Lopez-Sanchez, D. Dumcenco, E. Charbon, A. Kis, 2014, arXiv: 1411.3232. [47] M. R. M. Atalla, S. J. Koester, IEEE Xplore, DOI: 10.1109/DRC.2017.7999500. [48] S. Lei, F. Wen, L. Ge, S. Najmaei, A. George, Y. Gong, W. Gao, Z. Jin, B. Li, J. Lou, J. Kono, R. Vajtai, P. Ajayan, N. J. Halas, Nano Lett. 2015, 15, 3048. [49] D. Kufer, G. Konstantatos, ACS Photonics 2016, 3, 2197. [50] D. Kufer, G. Konstantatos, Nano Lett. 2015, 15, 7307. [51] M. Buscema, J. O. Island, D. J. Groenendijk, S. I. Blanter, G. A. Steele, H. S. J. van der Zant, A. Castellanos-Gomez, Chem. Soc. Rev. 2015, 44, 3691. [52] A. Pospischil, T. Mueller, Appl. Sci. 2016, 6, 78. [53] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, A. A. Firsov, Science 2004, 306, 666. [54] A. K. Geim, K. S. Novoselov, Nature Mater. 2007, 6, 183. [55] A. K. Geim, Science 2009, 324, 1530. [56] J. M. Dawlaty, S. Shivaraman, J. Strait, P. George, M. Chandrashekhar, F. Rana, M. G. Spencer, D. Veksler, Y. Chen, Appl. Phys. Lett. 2008, 93, 131905. 45 [57] R. R. Nair, P. Blake, A. N. Grigorenko, K. S. Novoselov, T. J. Booth, T. Stauber, N. M. R. Peres, A. K. Geim, Science 2008, 320, 1308. [58] A. B. Kuzmenko, E. van Heumen, F. Carbone, D. van der Marel, Phys. Rev. Lett. 2008, 100, 117401. [59] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, A. A. Firsov, Nature 2005, 438, 197. [60] Z. Q. Li, E. A. Henriksen, Z. Jiang, Z. Hao, M. C. Martin, P. Kim, H. L. Stormer, D. N. Basov, Nature Phys. 2008, 4, 532. [61] K. S. Kim, Y. Zhao, H. Jang, S. Y. Lee, J. M. Kim, K. S. Kim, J.-H. Ahn, P. Kim, J.-Y. Choi, B. H. Hong, Nature 2009, 457, 706. [62] Q. H. Wang, K. Kalantar-Zadeh, A. Kis, J. N. Coleman, M. S. Strano, Nature Nanotech. 2012, 7, 699. [63] S. Z. Butler, S. M. Hollen, L. Cao, Y. Cui, J. A. Gupta, H. R. Gutiérrez, T. F. Heinz, S. Sae Hong, J. Huang, A. F. Ismach, E. Johnston-Halperin, M. Kuno, V. V. Plashnitsa, R. D. Robinson, R. S. Ruoff, S. Salahuddin, J. Shan, L. Shi, M. G. Spencer, M. Terrones, W. Windl, J. E. Goldberger, ACS Nano 2013, 7, 2898. [64] K. Watanabe, T. Taniguchi, H. Kanda, Nature Mater. 2004, 3, 404. [65] I. Meric, C. Dean, A. Young, Proc. IEEE 2013, 101, 1609. [66] Q. Wei, X. Peng, Appl. Phys. Lett. 2014, 104, 251915. [67] H. Liu, A. T. Neal, Z. Zhu, Z. Luo, X. Xu, D. Tománek, P. D. Ye, ACS Nano 2014, 8, 4033. [68] P. Vogt, P. D. Padova, C. Quaresima, J. Avila, E. Frantzeskakis, M. C. Asensio, A. Resta, B. Ealet, G. L. Lay, Phys. Rev. Lett. 2012, 108, 155501. [69] X. Li, J. T. Mullen, Z. Jin, K. M. Borysenko, M. B. Nardelli, K. W. Kim, Phys. Rev. B 2013, 87, 115418. [70] S. Cahangirov, M. Topsakal, E. Akturk, H. Sahin, S. Ciraci, Phys. Rev. Lett. 2009, 102, 236804. [71] Z. Ni, E. Minamitani, Y. Ando, S. Watanabe, Phys. Rev. B 2017, 96, 075427. [72] F. Reis, G. Li, L. Dudy, M. Bauernfeind, S. Glass, W. Hanke, R. Thomale, J. Schäfer, R. Claessen, Science 2017, 357, 287. 46 [73] M. Pumera, Z. So, Adv. Mater. 2017, 29, 1605299. [74] Q. Guo, A. Pospischil, M. Bhuiyan, H. Jiang, H. Tian, D. Farmer, B. Deng, C. Li, S.-J. Han, H. Wang, Q. Xia, T.-P. Ma, T. Mueller, F. Xia, Nano Lett. 2016, 16, 4648. [75] M. Amani, E. Regan, J. Bullock, G. H. Ahn, A. Javey, ACS Nano, 2017, 11, 11724. [76] M. Engel, M. Steiner, P. Avouris, Nano Lett. 2014, 14, 6414. [77] Z. Zeng, Z. Yin, X. Huang, H. Li, Q. He, G. Lu, F. Boey, H. Zhang, Angew. Chemie Int. Ed. 2011, 50, 11093. [78] Z. Zeng, T. Sun, J. Zhu, X. Huang, Z. Yin, G. Lu, Z. Fan, Q. Yan, H. H. Hng, H. Zhang, Angew. Chemie Int. Ed. 2012, 51, 9052. [79] Y.-H. Lee, X.-Q. Zhang, W. Zhang, M.-T. Chang, C.-T. Lin, K.-D. Chang, Y.-C. Yu, J. T.-W. Wang, C.-S. Chang, L.-J. Li, T.-W. Lin, Adv. Mater. 2012, 24, 2320. [80] D. Dumcenco, D. Ovchinnikov, K. Marinov, P. Lazić, M. Gibertini, N. Marzari, O. L. Sanchez, Y.-C. Kung, D. Krasnozhon, M.-W. Chen, S. Bertolazzi, P. Gillet, A. F. Morral, A. Radenovic, A. Kis, ACS Nano 2015, 9, 4611. [81] H. Schmidt, S. Wang, L. Chu, M. Toh, R. Kumar, W. Zhao, A. H. C. Neto, J. Martin, S. Adam, B. Özyilmaz, G. Eda, Nano Lett. 2014, 14, 1909. [82] T. Carey, S. Cacovich, G. Divitini, J. Ren, A. Mansouri, J. M. Kim, C. Wang, C. Ducati, R. Sordan, F. Torrisi, Nature Commun. 2017, 8, 1202. [83] D. McManus, S. Vranic, F. Withers, V. Sanchez-Romaguera, M. Macucci, H. Yang, R. Sorrentino, K. Parvez, S.-K. Son, G. Iannaccone, K. Kostarelos, G. Fiori, C. Casiraghi, Nature Nanotech. 2017, 12, 343. [84] V. N. Kotov, B. Uchoa, V. M. Pereira, F. Guinea, A. H. Castro Neto, Rev. Mod. Phys. 2012, 84, 1067. [85] X. Xu, N. M. Gabor, J. S. Alden, A. M. van der Zande, P. L. McEuen, Nano Lett. 2010, 10, 562. [86] N. M. Gabor, J. C. W. Song, Q. Ma, N. L. Nair, T. Taychatanapat, K. Watanabe, T. Taniguchi, L. S. Levitov, P. Jarillo-Herrero, Science 2011, 334, 648. [87] D. Sun, G. Aivazian, A. M. Jones, J. S. Ross, W. Yao, D. Cobden, X. Xu, Nature Nanotech. 2012, 7, 114. 47 [88] L. Prechtel, L. Song, D. Schuh, P. Ajayan, W. Wegscheider, A. W. Holleitner, Nature Commun. 2012, 3, 646. [89] M. Freitag, T. Low, P. Avouris, Nano Lett. 2013, 13, 1644. [90] J. Park, Y. H. Ahn, C. Ruiz-Vargas, Nano Lett. 2009, 9, 1742. [91] E. J. H. Lee, K. Balasubramanian, R. T. Weitz, M. Burghard, K. Kern, Nature Nanotech. 2008, 3, 486. [92] F. Xia, T. Mueller, R. Golizadeh-Mojarad, M. Freitag, Y.-M. Lin, J. Tsang, V. Perebeinos, P. Avouris, Nano Lett. 2009, 9, 1039. [93] M. Buscema, M. Barkelid, V. Zwiller, H. S. J. van der Zant, G. A. Steele, A. Castellanos-Gomez, Nano Lett. 2013, 13, 358. [94] J. Wu, H. Schmidt, K. K. Amara, X. Xu, G. Eda, B. O zyilmaz, Nano Lett. 2014, 14, 2730. [95] Y. Zhang, H. Li, L. Wang, H. Wang, X. Xie, S.-L. Zhang, R. Liu, Z.-J. Qiu, Sci. Rep. 2015, 5, 7938. [96] E. Flores, J. R. Ares, A. Castellanos-Gomez, M. Barawi, I. J. Ferrer, C. Snchez, Appl. Phys. Lett. 2015, 106, 022102. [97] Y. Saito, T. Iizuka, T. Koretsune, R. Arita, S. Shimizu, Y. Iwasa, Nano Lett. 2016, 16, 4819. [98] M.-J. Lee, J.-H. Ahn, J. H. Sung, H. Heo, S. G. Jeon, W. Lee, J. Y. Song, K.-H. Hong, B. Choi, S.-H. Lee, M.-H. Jo, Nature Commun. 2016, 7, 12011. [99] P. L. Richards, J. Appl. Phys. 1994, 76, 1. [100] F. H. L. Koppens, T. Mueller, P. Avouris, A. C. Ferrari, M. S. Vitiello, M. Polini, Nature Nanotech. 2014, 9, 780. [101] R. A. Soref, Proc. IEEE 1993, 81, 1687. [102] J. Yan, M-H. Kim, J. A. Elle, A. B. Sushkov, G. S. Jenkins, H. M. Milchberg, M. S. Fuhrer, H. D. Drew, Nature Nanotech. 2012, 7, 472. [103] J.-Y. Wu, Y. T. Chun, S. Li, T. Zhang, J. Wang, P. K. Shrestha, D. Chu, Adv. Mater. 2018, 30, 1705880. [104] U. Sassi, R. Parret, S. Nanot, M. Bruna, S. Borini, D. De Fazio, Z. Zhao, E. Lidorikis, F. H. L. Koppens, A. C. Ferrari, A. Colli, Nature Commun. 2017, 8, 14311. 48 [105] A. Goswami, P. Dhandaria, S. Pal, R. McGee, F. Khan, Ž. Antić, R. Gaikwad, K. Prashanthi, T. Thundat, Nano Res. 2017, 10, 3571. [106] D. Costanzo, S. Jo, H. Berger, A. F. Morpurgo, Nature Nanotech. 2016, 11, 339. [107] C. Xu, L.Wang, Z. Liu, L. Chen, J. Guo, N. Kang, X.-L. Ma, H.-M. Cheng, W. Ren, Nature Mater. 2015, 14, 1135. [108] J.-F. Ge, Z.-L. Liu, C. Liu, C.-L. Gao, D. Qian, Q.-K. Xue, Y. Liu, J.-F. Jia, Nature Mater. 2015, 14, 285. [109] S. Das, J. Appenzeller, Appl. Phys. Lett. 2013, 103, 103501. [110] H. Fang, M. Tosun, G. Seol, T. C. Chang, K. Takei, J. Guo, A. Javey, Nano Lett. 2013, 13, 1991. [111] D. Xiang, C. Han, J. Wu, S. Zhong, Y. Liu, J. Lin, X.-A. Zhang, W. Ping Hu, B. Özyilmaz, A. H. C. Neto, A. T. S. Wee, W. Chen, Nature Commun. 2015, 6, 6485. [112] A. Pospischil, M. M. Furchi, T. Mueller, Nature Nanotech. 2014, 9, 257. [113] B. W. H. Baugher, H. O. H. Churchill, Y. Yang, P. Jarillo-Herrero, Nature Nanotech. 2014, 9, 262. [114] J. S. Ross, P. Klement, A. M. Jones, N. J. Ghimire, J. Yan, D. G. Mandrus, T. Taniguchi, K. Watanabe, K. Kitamura, W. Yao, D. H. Cobden, X. Xu, Nature Nanotech. 2014, 9, 268. [115] R. Bratschitsch, Nature Nanotech. 2014, 9, 247. [116] D. J. Groenendijk, M. Buscema, G. A. Steele, S. Michaelis de Vasconcellos, R. Bratschitsch, H. S. J. van der Zant, A. Castellanos-Gomez, Nano Lett. 2014, 14, 5846. [117] M. Buscema, D. J. Groenendijk, G. a Steele, H. S. J. van der Zant, A. Castellanos-Gomez, Nature Commun. 2014, 5, 4651. [118] S. Memaran, N. R. Pradhan, Z. Lu, D. Rhodes, J. Ludwig, Q. Zhou, O. Ogunsolu, P. M. Ajayan, D. Smirnov, A. I. Fernández-Domínguez, F. J. García-Vidal, L. Balicas, Nano Lett. 2015, 15, 7532. [119] F. Xue, L. Chen, J. Chen, J. Liu, L. Wang, M. Chen, Y. Pang, X. Yang, G. Gao, J. Zhai, Z. L. Wang, Adv. Mater. 2016, 28, 3391. 49 [120] M. R. Laskar, D. N. Nath, L. Ma, E. W. Lee, C. H. Lee, T. Kent, Z. Yang, R. Mishra, M. A. Roldan, J.-C. Idrobo, S. T. Pantelides, S. J. Pennycook, R. C. Myers, Y. Wu, S. Rajan, Appl. Phys. Lett. 2014, 104, 092104. [121] X. Liu, D. Qu, J. Ryu, F. Ahmed, Z. Yang, D. Lee, W. J. Yoo, Adv. Mater. 2016, 28, 2345. [122] M. S. Choi, D. Qu, D. Lee, X. Liu, K. Watanabe, T. Taniguchi, W. J. Yoo, ACS Nano 2014, 8, 9332. [123] X. Yu, S. Zhang, H. Zeng, Q. J. Wang, Nano Energy 2016, 25, 34. [124] M.-Y. Li, Y. Shi, C.-C. Cheng, L.-S. Lu, Y.-C. Lin, H.-L. Tang, M.-L. Tsai, C.-W. Chu, K.-H. Wei, J.-H. He, W.-H. Chang, K. Suenaga, L.-J. Li, Science 2015, 349, 524. [125] Y. Gong, J. Lin, X. Wang, G. Shi, S. Lei, Z. Lin, X. Zou, G. Ye, R. Vajtai, B. I. Yakobson, H. Terrones, M. Terrones, B. K.Tay, J. Lou, S. T. Pantelides, Z. Liu, W. Zhou, P. M. Ajayan, Nature Mater. 2014, 13, 1135. [126] X. Duan, C. Wang, J. C. Shaw, R. Cheng, Y. Chen, H. Li, X. Wu, Y. Tang, Q. Zhang, A. Pan, J. Jiang, R. Yu, Y. Huang, X. Duan, Nature Nanotech. 2014, 9, 1024. [127] P. K. Sahoo, S. Memaran, Y. Xin, L. Balicas, H. R. Gutiérrez, Nature 2018, 553, 63. [128] C.-H. Lee, G. Lee, A. M. van der Zande, W. Chen, Y. Li, M. Han, X. Cui, G. Arefe, C. Nuckolls, T. F. Heinz, J. Guo, J. Hone, P. Kim, Nature Nanotech. 2014, 9, 676. [129] M. M. Furchi, A. Pospischil, F. Libisch, J. Burgdörfer, T. Mueller, Nano Lett. 2014, 14, 4785. [130] N. Huo, J. Kang, Z. Wei, S.-S. Li, J. Li, S.-H. Wei, Adv. Funct. Mater. 2014, 24, 7025. [131] N. Huo, J. Yang, L. Huang, Z. Wei, S.-S. Li, S.-H. Wei, J. Li, Small 2015, 11, 5430. [132] A. Pezeshki, S. H. H. Shokouh, T. Nazari, K. Oh, S. Im, Adv. Mater. 2016, 28, 3216. [133] R. Cheng, D. Li, H. Zhou, C. Wang, A. Yin, S. Jiang, Y. Liu, Y. Chen, Y. Huang, X. Duan, Nano Lett. 2014, 14, 5590. [134] J. Kang, S. Tongay, J. Zhou, J. Li, J. Wu, Appl. Phys. Lett. 2013, 102, 012111. [135] X. Hong, J. Kim, S.-F. Shi, Y. Zhang, C. Jin, Y. Sun, S. Tongay, J. Wu, Y. Zhang, F. Wang, Nature Nanotech. 2014, 9, 682. [136] F. Barati, M. Grossnickle, S. Su, R. K. Lake, V. Aji, N. M. Gabor, Nature Nanotech. 2017, 12, 1134. 50 [137] M. Long, E. Liu, P. Wang, A. Gao, H. Xia, W. Luo, B. Wang, J. Zeng, Y. Fu, K. Xu, W. Zhou, Y. Lv, S. Yao, M. Lu, Y. Chen, Z. Ni, Y. You, X. Zhang, S. Qin, Y. Shi, W. Hu, D. Xing, F. Miao, Nano Lett. 2016, 16, 2254. [138] Y. Deng, Z. Luo, N. J. Conrad, H. Liu, Y. Gong, S. Najmaei, P. M. Ajayan, J. Lou, X. Xu, P. D. Ye, ACS Nano 2014, 8, 8292. [139] L. Ye, H. Li, Z. Chen, J. Xu, ACS Photonics. 2016, 3, 692. [140] M. Long, A. Gao, P. Wang, H. Xia, C. Ott, C. Pan, Y. Fu, E. Liu, X. Chen, W. Lu, T. Nilges, J. Xu, X. Wang, W. Hu, F. Miao, Sci. Adv. 2017, 3, e1700589. [141] Y.-J. Yu, Y. Zhao, S. Ryu, L. E. Brus, K. S. Kim, P. Kim, Nano Lett. 2009, 9, 3430. [142] L. Britnell, R. M. Ribeiro, A. Eckmann, R. Jalil, B. D. Belle, A. Mishchenko, Y.-J. Kim, R. V Gorbachev, T. Georgiou, S. V. Morozov, A. N. Grigorenko, A. K. Geim, C. Casiraghi, A. H. C. Neto, K. S. Novoselov, Science 2013, 340, 1311. [143] W. J. Yu, Y. Liu, H. Zhou, A. Yin, Z. Li, Y. Huang, X. Duan, Nature Nanotech. 2013, 8, 952. [144] M. Massicotte, P. Schmidt, F. Vialla, K. G. Schädler, A. Reserbat-Plantey, K. Watanabe, T. Taniguchi, K. J. Tielrooij, F. H. L. Koppens, Nature Nanotech. 2016, 11, 42. [145] L. Braun, G. Mussler, A. Hruban, M. Konczykowski, T. Schumann, M. Wolf, M. Munzenberg, L. Perfetti, T. Kampfrath, Nature Commun. 2016, 7, 13259 [146] D.-H. Kang, M.-S. Kim, J. Shim, J. Jeon, H.-Y. Park, W.-S. Jung, H.-Y. Yu, C.-H. Pang, S. Lee, J.-H. Park, Adv. Funct. Mater. 2015, 25, 4219. [147] G. Konstantatos, M. Badioli, L. Gaudreau, J. Osmond, M. Bernechea, F. P. Garcia de Arquer, F. Gatti, F. H. L. Koppens, Nature Nanotech. 2012, 7, 363. [148] C. Ma, Y. Shi, W. Hu, M. H. Chiu, Z. Liu, A. Bera, F. Li, H. Wang, L. J. Li, T. Wu, Adv. Mater. 2016, 28, 3683. [149] D. Kang, S. R. Pae, J. Shim, G. Yoo, J. Jeon, J. W. Leem, J. S. Yu, S. Lee, B. Shin, J.-H. Park, Adv. Mater. 2016, 28, 7799. [150] W. Wang, A. Klots, D. Prasai, Y. Yang, K. Bolotin, J. Valentine, Nano Lett. 2015, 15, 7440. [151] M. Furchi, A. Urich, A. Pospischil, G. Lilley, K. Unterrainer, H. Detz, P. Klang, A. M. Andrews, W. Schrenk, G. Strasser, T. Mueller, Nano Lett. 2012, 12, 2773. 51 [152] M. Engel, M. Steiner, A. Lombardo, A. C. Ferrari, H. v. Löhneysen, P. Avouris, R. Krupke, Nature Commun. 2012, 3, 906. [153] T. J. Echtermeyer, L. Britnell, P. K. Jasnos, A. Lombardo, R. V. Gorbachev, A. N. Grigorenko, A. K. Geim, A. C. Ferrari, K. S. Novoselov, Nature Commun. 2011, 2, 458. [154] I. Nikitskiy, S. Goossens, D. Kufer, T. Lasanta, G. Navickaite, F. H. L. Koppens, G. Konstantatos, Nature Commun. 2016, 7, 11954. [155] H. Qiao, J. Yuan, Z. Xu, C. Chen, S. Lin, Y. Wang, J. Song, Y. Liu, Q. Khan, H. Y. Hoh, C.-X. Pan, S. Li, Q. Bao, ACS Nano 2015, 9, 1886. [156] A. V. Klekachev, M. Cantoro, M. H. van der Veen, A. L. Stesmans, M. M. Heyns, S. De Gendt, Physica E 2011, 43, 1046. [157] W. Guo, S. Xu, Z. Wu, N. Wang, M. M. T. Loy, S. Du, Small 2013, 9, 3031. [158] A. Jana, E. Scheer, S. Polarz, Beilstein J. Nanotechnol. 2017, 8, 688. [159] Y. Lee, J. Kwon, E. Hwang, C.-H. Ra, W. J. Yoo, J.-H. Ahn, J. H. Park, J. H. Cho, Adv. Mater. 2015, 27, 41. [160] Z. Sun, L. Aigouy, Z. Chen, Nanoscale 2016, 8, 7377. [161] P.-H. Chang, S.-Y. Liu, Y.-B. Lan, Y.-C. Tsai, X.-Q. You, C.-S. Li, K.-Y. Huang, A.-S. Chou, T.-C. Cheng, J.-K. Wang, C.-I Wu, Sci. Rep. 2017, 7, 46281. [162] K. Roy, M. Padmanabhan, S. Goswami, T. P. Sai, G. Ramalingam, S. Raghavan, A. Ghosh, Nature Nanotech. 2013, 8, 826. [163] W. Zhang, C.-P. Chuu, J.-K. Huang, C.-H. Chen, M.-L. Tsai, Y.-H. Chang, C.-T. Liang, Y.-Z. Chen, Y.-L. Chueh, J.-H. He, M.-Y. Chou, L.-J. Li, Sci. Rep. 2014, 4, 3826. [164] J. D. Mehew, S. Unal, E. T. Alonso, G. F. Jones, S. F. Ramadhan, M. F. Craciun, S. Russo, Adv. Mater. 2017, 29, 1700222. [165] C.-H. Liu, Y.-C. Chang, T. B. Norris, Z. Zhong, Nature Nanotech. 2014, 9, 273. [166] Z. Yin, H. H. Li, H. H. Li, L. Jiang, Y. Shi, Y. Sun, G. Lu, Q. Zhang, X. Chen, H. Zhang, ACS Nano 2012, 6, 74. [167] W. Choi, M. Y. Cho, A. Konar, J. H. Lee, G.-B. Cha, S. C. Hong, S. Kim, J. Kim, D. Jena, J. Joo, S. Kim, Adv. Mater. 2012, 24, 5832. 52 [168] W. Zhang, J.-K. Huang, C.-H. Chen, Y.-H. Chang, Y.-J. Cheng, L.-J. Li, Adv. Mater. 2013, 25, 3456. [169] J. Kwon, Y. K. Hong, G. Han, I. Omkaram, W. Choi, S. Kim, Y. Yoon, Adv. Mater. 2015, 27, 2224. [170] Y.-H. Chang, W. Zhang, Y. Zhu, Y. Han, J. Pu, J.-K. Chang, W.-T. Hsu, J.-K. Huang, C.-L. Hsu, M.-H. Chiu, T. Takenobu, H. Li, C.-I. Wu, W.-H. Chang, A. T. S. Wee, L.-J. Li, ACS Nano 2014, 8, 8582. [171] N. Perea-López, A. L. Elías, A. Berkdemir, A. Castro-Beltran, H. R. Gutiérrez, S. Feng, R. Lv, T. Hayashi, F. López-Urías, S. Ghosh, B. Muchharla, S. Talapatra, H. Terrones, M. Terrones, Adv. Funct. Mater. 2013, 23, 5511. [172] N. Huo, S. Yang, Z. Wei, S.-S. Li, J.-B. Xia, J. Li, Sci. Rep. 2014, 4, 5209. [173] J. Chen, B. Liu, Y. Liu, W. Tang, C. T. Nai, L. Li, J. Zheng, L. Gao, Y. Zheng, H. S. Shin, H. Y. Jeong, K. P. Loh, Adv. Mater. 2015, 27, 6722. [174] S.-H. Jo, D.-H. Kang, J. Shim, J. Jeon, M. H. Jeon, G. Yoo, J. Kim, J. Lee, G. Y. Yeom, S. Lee, H.-Y. Yu, C. Choi, J.-H. Park, Adv. Mater. 2016, 28, 4824. [175] M. M. Furchi, D. K. Polyushkin, A. Pospischil, T. Mueller, Nano Lett. 2014, 14, 6165. [176] X. Wang, P. Wang, J. Wang, W. Hu, X. Zhou, N. Guo, H. Huang, S. Sun, H. Shen, T. Lin, M. Tang, L. Liao, A. Jiang, J. Sun, X. Meng, X. Chen, W. Lu, J. Chu, Adv. Mater. 2015, 27, 6575. [177] H. S. Lee, S. S. Baik, K. Lee, S. Min, P. J. Jeon, J. S. Kim, K. Choi, H. J. Choi, J. H. Kim, S. Im, ACS Nano 2015, 9, 8312. [178] C. Xie, C. Mak, X. Tao, F. Yan, Adv. Funct. Mater. 2017, 27, 1603886. [179] D. Kufer, I. Nikitskiy, T. Lasanta, G. Navickaite, F. H. L. Koppens, G. Konstantatos, Adv. Mater. 2015, 27, 176. [180] D. Kufer, T. Lasanta, M. Bernechea, F. H. L. Koppens, G. Konstantatos, ACS Photonics. 2016, 3, 1324. [181] Y. Wang, R. Fullon, M. Acerce, C. E. Petoukhoff, J. Yang, C. Chen, S. Du, S. K. Lai, S. P. Lau, D. Voiry, D. O'Carroll, G. Gupta, A. D. Mohite, S. Zhang, H. Zhou, M. Chhowalla, Adv. Mater. 2017, 29, 1603995. 53 [182] B. Peng, G. Yu, Y. Zhao, Q. Xu, G. Xing, X. Liu, D. Fu, B. Liu, J. R. S. Tan, W. Tang, H. Lu, J. Xie, L. Deng, T. C. Sum, K. P. Loh, ACS Nano 2016, 10, 6383. [183] C. Chen, H. Qiao, S. Lin, C. M. Luk, Y. Liu, Z. Xu, J. Song, Y. Xue, D. Li, J. Yuan, W. Yu, C. Pan, S. P. Lau, Q. Bao, Sci. Rep. 2015, 5, 11830. [184] E. Lhuillier, S. Keuleyan, P. Guyot-Sionnest, Chem. Mater. 2013, 25, 1272. [185] S. E. Keuleyan, P. Guyot-Sionnest, C. Delerue, G. Allan, ACS Nano 2014, 8, 8676. [186] E. Lhuillier, M. Scarafagio, P. Hease, B. Nadal, H. Aubin, X. Z. Xu, N. Lequeux, G. Patriarche, S. Ithurria, B. Dubertret, Nano Lett. 2016, 16, 1282. [187] N. Huo, S. Gupta, G. Konstantatos, Adv. Mater. 2017, 29, 1606576. [188] N. Huo, G. Konstantatos, Nature Commun. 2017, 8, 572. [189] S. H. Lim, W. Mar, P. Matheu, D. Derkacs, E. T. Yu, J. Appl. Phys. 2007, 101, 104309. [190] T. J. Echtermeyer, L. Britnell, P. K. Jasnos, A. Lombardo, R. V. Gorbachev, A. N. Grigorenko, A. K. Geim, A. C. Ferrari, K. S. Novoselov, Nature Commun. 2011, 2, 458. [191] J. Lin, H. Li, H. Zhang, W. Chen, Appl. Phys. Lett. 2013, 102, 203109. [192] A. Sobhani, A. Lauchner, S. Najmaei, C. Ayala-Orozco, F. Wen, J. Lou, N. J. Halas, Appl. Phys. Lett. 2014, 104, 031112. [193] D. M. Schaadt, B. Feng, E. T. Yu, Appl. Phys. Lett. 2005, 86, 063106. [194] S. P. Sundararajan, N. K. Grady, N. Mirin, N. J. Halas, Nano Lett. 2008, 8, 624. [195] S. Najmaei, A. Mlayah, A. Arbouet, C. Girard, J. Léotin, J. Lou, ACS Nano 2014, 8, 12682. [196] J. Miao, W. Hu, Y. Jing, W. Luo, L. Liao, A. Pan, S. Wu, J. Cheng, X. Chen, W. Lu, Small 2015, 11, 2392. [197] Z. Fang, Y. Wang, A. E. Schlather, Z. Liu, P. M. Ajayan, F. J. G. de Abajo, P. Nordlander, X. Zhu, N. J. Halas, Nano Lett. 2014, 14, 299. [198] G. V. Naik, V. M. Shalaev, A. Boltasseva, Adv. Mater. 2013, 25, 3264. [199] T. J. Echtermeyer, L. Britnell, P. K. Jasnos, A. Lombardo, R. V. Gorbachev, A. N. Grigorenko, A. K. Geim, A. C. Ferrari, K. S. Novoselov, Nature Commun. 2011, 2, 458. [200] Z. Y. Fang, Z. Liu, Y. M. Wang, P. M. Ajayan, P. Nordlander, N. J. Halas, Nano Lett. 2012, 12, 3808. 54 [201] Z. Fang, Y. Wang, Z. Liu, A. Schlather, P. M. Ajayan, F. H. L. Koppens, P. Nordlander, N. J. Halas, ACS Nano 2012, 6, 10222. [202] N. K. Emani, T. F. Chung, X. J. Ni, A. V. Kildishev, Y. P. Chen, A. Boltasseva, Nano Lett. 2012, 12, 5202. [203] S. Mubeen, G. Hernandez-Sosa, D. Moses, J. Lee, M. Moskovits, Nano Lett. 2011, 11, 5548. [204] I. Goykhman, B. Desiatov, J. Khurgin, J. Shappir, U. Levy, Nano Lett. 2011, 11, 2219. [205] M. W. Knight, H. Sobhani, P. Nordlander, N. J. Halas, Science 2011, 332, 702. [206] Y. K. Lee, C. H. Jung, J. Park, H. Seo, G. A. Somorjai, J. Y. Park, Nano Lett. 2011, 11, 4251. [207] T. Hong, B. Chamlagain, S. Hu, S. M. Weiss, Z. Zhou, Y. Xu, ACS Nano 2015, 9, 5357. [208] Y. Kang, S. Najmaei, Z. Liu, Y. Bao, Y. Wang, X. Zhu, N. J. Halas, P. Nordlander, P. M. Ajayan, J. Lou, Z. Fang, Adv. Mater. 2014, 26, 6467. [209] J. Xu, S. Wang, G.-J. N. Wang, C. Zhu, S. Luo, L. Jin, X. Gu, S. Chen, V. R. Feig, J. W. F. To, S. Rondeau-Gagné, J. Park, B. C. Schroeder, C. Lu, J. Y. Oh, Y. Wang, Y.-H. Kim, H. Yan, R. Sinclair, D. Zhou, G. Xue, B. Murmann, C. Linder, W. Cai, J. B.-H. Tok, J. W. Chung, Z. Bao, Science 2017, 355, 59. [210] A. Miyamoto, S. Lee, N. F. Cooray, S. Lee, M. Mori, N. Matsuhisa, H. Jin, L. Yoda, T. Yokota, A. Itoh, M. Sekino, H. Kawasaki, T. Ebihara, M. Amagai, T. Someya, Nature Nanotech. 2017, 12, 907. [211] S. Wang, J. Xu, W. Wang, G.-J. N. Wang, R. Rastak, F. Molina-Lopez, J. W. Chung, S. Niu, V. R. Feig, J. Lopez, T. Lei, S.-K. Kwon, Y. Kim, A. M. Foudeh, A. Ehrlich, A. Gasperini, Y. Yun, B. Murmann, J. B.-H. Tok, Z. Bao, Nature 2018, 555, 83. [212] Q. H. Wang, M. C. Hersam, Nature Chem. 2009, 1, 206. [213] T. H. Lee, K. Kim, G. Kim, H. J. Park, D. Scullion, L. Shaw, M.-G. Kim, X. Gu, W.-G. Bae, E. J. G. Santos, Z. Lee, H. S. Shin, Y. Nishi, Z. Bao, Chem. Mater. 2017, 29, 2341. [214] C.-H. Lee, T. Schiros, E. J. G. Santos, B. Kim, K. G. Yager, S. J. Kang, S. Lee, J. Yu, K. Watanabe, T. Taniguchi, J. Hone, E. Kaxiras, C. Nuckolls, P. Kim, Adv. Mater. 2014, 26, 2812. [215] S. J. Kang, G.-H. Lee, Y.-J. Yu, Y. Zhao, B. Kim, K. Watanabe, T. Taniguchi, J. Hone, P. Kim, C. Nuckolls, Adv. Funct. Mater. 2014, 24, 5157. 55 [216] D. He, Y. Zhang, Q. Wu, R. Xu, H. Nan, J. Liu, J. Yao, Z. Wang, S. Yuan, Y. Li, Y. Shi, J. Wang, Z. Ni, L. He, F. Miao, F. Song, H. Xu, K. Watanabe, T. Taniguchi, J.-B. Xu, X. Wang, Nature Commun. 2014, 5, 5162. [217] P. Wei, N. Liu, H. R. Lee, E. Adijanto, L. Ci, B. D. Naab, J. Q. Zhong, J. Park, W. Chen, Y. Cui, Z. Bao, Nano Lett. 2013, 13, 1890. [218] S. B. Homan, V. K. Sangwan, I. Balla, H. Bergeron, E. A. Weiss, M. C. Hersam, Nano Lett. 2017, 17, 164. [219] C. E. Petoukhoff, M. B. M. Krishna, D. Voiry, I. Bozkurt, S. Deckoff-Jones, M. Chhowalla, D. M. O'Carroll, K. M. Dani, ACS Nano 2016, 10, 9899. [220] C.-J. Shih, R. Pfattner, Y.-C. Chiu, N. Liu, T. Lei, D. Kong, Y. Kim, H.-H. Chou, W.-G. Bae, Z. Bao, Nano Lett. 2015, 15, 7587. [221] K. Kim, T. H. Lee, E. J. G. Santos, P. S. Jo, A. Salleo, Y. Nishi, Z. Bao, ACS Nano, 2015, 9, 5922. [222] F. Jabbarzadeh, M. Siahsar, M. Dolatyari, G. Rostami, A. Rostami, IEEE Sens. J. 2015, 15, 2795. [223] Y. Lee, S. H. Yu, J. Jeon, H. Kim, J. Y. Lee, H. Kim, J.-H. Ahn, E. Hwang, J. H. Cho, Carbon 2015, 88, 165. [224] X. Liu, X. Luo, H. Nan, H. Guo, P. Wang , L. Zhang, M. Zhou, Z. Yang, Y. Shi, W. Hu, Z. Ni, T. Qiu, Z. Yu, J.-B. Xu, X. Wang, Adv. Mater. 2016, 28, 5200. [225] E. H. Huisman, A. G. Shulga, P. J. Zomer, N. Tombros, D. Bartesaghi, S. Z. Bisri, M. A. Loi, L. J. A. Koster, B. J. van Wees, ACS Appl. Mater. Interfaces 2015, 7, 11083. [226] P.-H. Chang, Y.-C. Tsai, S.-W. Shen, S.-Y. Liu, K.-Y. Huang, C.-S. Li, H.-P. Chang, C.-I. Wu, ACS Photonics 2017, 4, 2335. [227] W.-C. Tan, W.-H. Shih, Y. F. Chen, Adv. Funct. Mater. 2014, 24, 6818. [228] Y. Liu, F. Wang, X. Wang, X. Wang, E. Flahaut, X. Liu, Y. Li, X. Wang, Y. Xu, Y. Shi, R. Zhang. Nature Commun. 2015, 6, 8589. [229] R. Lu, C. Christianson, B. Weintrub, J. Z. Wu, ACS Appl. Mater. Inter. 2013, 5, 11703. 56 [230] A. J. Molina-Mendoza, L.Vaquero-Garzon, S. Leret, L. de Juan-Fernández, E. M. Pérez, A. Castellanos-Gomez, Chem. Commun. 2016, 52, 14365. [231] D. B. Velusamy, M. A. Haque, M. R. Parida, F. Zhang, T. Wu, O. F. Mohammed, H. N. Alshareef, Adv. Funct. Mater. 2017, 27, 1605554. [232] Z. Jin, D. He, Q. Zhou, P. Mao, L. Ding, J. Wang, ACS Appl. Mater. Interfaces 2016, 8, 33043. [233] S. H. Yu, Y. Lee, S. K. Jang, J. Kang, J. Jeon, C. Lee, J. Y. Lee, H. Kim, E. Hwang, S. Lee, J. H. Cho, ACS Nano 2014, 8, 8285. [234] F. Yang, S. Cheng, X. Zhang, X. Ren, R. Li, H. Dong, W. Hu, Adv. Mater. 2018, 30, 1702415. [235] M. Gobbi, E. Orgiu, P. Samorì, Adv. Mater. 2018, 1706103. [236] H. Klauk, Chem. Soc. Rev. 2010, 39, 2643. [237] K. Nomura, H. Ohta, A. Takagi, T. Kamiya, M. Hirano, H. Hosono, Nature 2004, 432, 488. [238] K. J. Yu, Z. Yan, M. Han, J. A. Rogers, npj Flexible Electronics 2017, 1, 4. [239] Y. Wang, C. Zhu, R. Pfattner, H. Yan, L. Jin, S. Chen, F. Molina-Lopez, F. Lissel, J. Liu, N. I. Rabiah, Z. Chen, J. W. Chung, C. Linder, M. F. Toney, B. Murmann, Z. Bao, Sci. Adv. 2017, 3, e1602076. [240] T. Yokota, P. Zalar, M. Kaltenbrunner, H. Jinno, N. Matsuhisa, H. Kitanosako, Y. Tachibana, W. Yukita, M. Koizumi, T. Someya, Sci. Adv. 2016, 2, e1501856. [241] M. Kaltenbrunner, T. Sekitani, J. Reede, T. Yokota, K. Kuribara, T. Tokuhara, M. Drack, R. Schwodiauer, I. Graz, S. Bauer-Gogonea, S. Bauer, T. Someya, Nature 2013, 499, 458. [242] H. Jinno, K. Fukuda, X. Xu, S. Park, Y. Suzuki, M. Koizumi, T. Yokota, I. Osaka, K. Takimiya, T. Someya, Nat. Energy 2017, 2, 780. [243] A. Chortos, J. Liu, Z. Bao, Nature Mater. 2016, 15, 937. [244] T. Someya, Z. Bao, G. G. Malliaras, Nature 2016, 540, 379. [245] C. Lee, X. Wei, J. W. Kysar, J. Hone, Science 2008, 321, 385. [246] F. Bonaccorso, Z. Sun, T. Hasan, A. C. Ferrari, Nature Photon. 2010, 4, 611. [247] K. Ellmer, Nature Photon. 2012, 6, 809. [248] H.-Y. Chang, S. Yang, J. Lee, L. Tao, W.-S. Hwang, D. Jena, N. Lu, D. Akinwande, ACS Nano 2013, 7, 5446. 57 [249] G.-H. Lee, Y.-J. Yu, X. Cui, N. Petrone, C.-H. Lee, M. S. Choi, D.-Y. Lee, C. Lee, W. J. Yoo, K. Watanabe, T. Taniguchi, C. Nuckolls, P. Kim, J. Hone, ACS Nano 2013, 7, 7931. [250] J. Pu, Y. Yomogida, K.-K. Liu, L.-J. Li, Y. Iwasa, T. Takenobu, Nano Lett. 2012, 12, 4013. [251] S. Das, R. Gulotty, A. V. Sumant, A. Roelofs, Nano Lett. 2014, 14, 2861. [252] H. Xu, J. Wu, Q. Feng, N. Mao, C. Wang, J. Zhang, Small 2014, 10, 2300. [253] D. D. Fazio, I. Goykhman, D. Yoon, M. Bruna, A. Eiden, S. Milana, U. Sassi, M. Barbone, D. Dumcenco, K. Marinov, A. Kis, A. C. Ferrari, ACS Nano 2016, 10, 8252. [254] N. Liu, H. Tian, G. Schwartz, J. B.-H. Tok, T.-L. Ren, Z. Bao, Nano Lett. 2014, 14, 3702. [255] N. Liu, A. Chortos, T. Lei, L. Jin, T. R. Kim, W.-G. Bae, C. Zhu, S. Wang, R. Pfattner, X. Chen, R. Sinclair, Z. Bao, Sci. Adv. 2017, 3, e1700159. [256] W. Zhu, M. N. Yogeesh, S. Yang, S. H. Aldave, J.-S. Kim, S. Sonde, L. Tao, N. Lu, D. Akinwande, Nano Lett. 2015, 15, 1883. [257] W. Zhu, S. Park, M. N. Yogeesh, K. M. McNicholas, S. R. Bank, D. Akinwande, Nano Lett. 2016, 16, 2301. [258] Y. Liu, Y. Liu, S. Qin, Y. Xu, R. Zhang, F. Wang, Nano Res. 2017, 10, 1880. [259] Z. Zheng, T. Zhang, J. Yao, Y. Zhang, J. Xu, G. Yang, Nanotechnology, 2016, 27, 225501. [260] P. T. Gomathi, P. Sahatiya, S. Badhulika, Adv. Funct. Mater. 2017, 27, 1701611. [261] K. Zhang, M. Peng, W. Wu, J. Guo, G. Gao, Y. Liu, J. Kou, R. Wen, Y. Lei, A. Yu, Y. Zhang, J. Zhai, Z. L. Wang, Mater. Horiz. 2017, 4, 274. [262] S. J. Kim, K. Choi, B. Lee, Y. Kim, B. H. Hong, Annu. Rev. Mater. Res. 2015, 45, 63. [263] H. Yang, J. Heo, S. Park, H. J. Song, D. H. Seo, K.-E. Byun, P. Kim, I. Yoo, H.-J. Chung, K. Kim, Science 2012, 336, 1140. [264] X. Wan, Y. Xu, H. Guo, K. Shehzad, A. Ali, Y. Liu, J. Yang, D. Dai, C.-T. Lin, L. Liu, H.-C. Cheng, F. Wang, X. Wang, H. Lu, W. Hu, X. Pi, Y. Dan, J. Luo, T. Hasan, X. Duan, X. Li, J. Xu, D. Yang, T. Ren, B. Yu, npj 2D Materials and Applications 2017, 1, 4. [265] L. Hao, Y. Liu, W. Gao, Z. Han, Q. Xue, H. Zeng, Z. Wu, J. Zhu, W. Zhang, J. Appl. Phys. 2015, 117, 114502. 58 [266] L. Wang, J. Jie, Z. Shao, Q. Zhang, X. Zhang, Y. Wang, Z. Sun, S.-T. Lee, Adv. Funct. Mater. 2015, 25, 2910. [267] Y. Zhang, Y. Yu, L. Mi, H. Wang, Z. Zhu, Q. Wu, Y. Zhang, Y. Jiang, Small 2016, 12, 1062. [268] B. Li, G. Shi, S. Lei, Y. He, W. Gao, Y. Gong, G. Ye, W. Zhou, K. Keyshar, J. Hao, P. Dong, L. Ge, J. Lou, J. Kono, R. Vajtai, P. M. Ajayan, Nano Lett. 2015, 15, 5919. [269] X. Gan, R.-J. Shiue, Y. Gao, I. Meric, T. F. Heinz, K. Shepard, J. Hone, S. Assefa, D. Englund, Nature Photon. 2013, 7, 883. [270] Y.-Q. Bie, G. Grosso, M. Heuck, M. M. Furchi, Y. Cao, J. Zheng, D. Bunandar, E. Navarro- Moratalla, L. Zhou, D. K. Efetov, T. Taniguchi, K. Watanabe, J. Kong, D. Englund, P. Jarillo- Herrero, Nature Nanotech. 2017, 12, 1124. [271] D. Sarkar, X. Xie, W. Liu, W. Cao, J. Kang, Y. Gong, S. Kraemer, P. M. Ajayan, K. Banerjee, Nature 2015, 526, 91. [272] A. Pospischil, M. Humer, M. M. Furchi, D. Bachmann, R. Guider, T. Fromherz, T. Mueller, Nature Photon. 2013, 7, 892. [273] L. Huang, H. Xu, Z. Zhang, C. Chen, J. Jiang, X. Ma, B. Chen, Z. Li, H. Zhong, L.-M. Peng, Sci. Rep. 2014, 4, 5548. [274] S. Goossens, G. Navickaite, C. Monasterio, S. Gupta, J. J. Piqueras, R. Pérez, G. Burwell, I. Nikitskiy, T. Lasanta, T. Galán, E. Puma, A. Centeno, A. Pesquera, A. Zurutuza, G. Konstantatos, F. Koppens, Nature Photon. 2017, 11, 366. [275] S. M. M. Zanjani, M. Holt, M. M. Sadeghi, S. Rahimi, D. Akinwande, npj 2D Mater. Appl. 2017, 36, 1. [276] N. Youngblood, M. Li, Nanophotonics 2016, 6, 1205. [277] A. D. Sanctis, G. F. Jones, D. J. Wehenkel, F. Bezares, F. H. L. Koppens, M. F. Craciun, S. Russo, Sci. Adv. 2017, 3, e1602617. [278] D. Krasnozhon, D. Lembke, C. Nyffeler, Y. Leblebici, A. Kis, Nano Lett. 2014, 14, 5905. [279] A. Allain, J. Kang, K. Banerjee, A. Kis, Nature Mater. 2015, 14, 1195. [280] S. Das, H-Y. Chen, A. V. Penumatcha, J. Appenzeller, Nano Lett. 2013, 13, 100. [281] A. Dankert, L. Langouche, M. V. Kamalakar, S. P. Dash, ACS Nano 2014, 8, 476. 59 [282] H. Tan, Y. Fan, Y. Zhou, Q. Chen, W. Xu, J. H. Warner, ACS Nano 2016, 10, 7866. [283] L. Yu, Y.-H. Lee, X. Ling, E. J. G. Santos, Y. C. Shin, Y. Lin, M. Dubey, E. Kaxiras, J. Kong, H. Wang, T. Palacios, Nano Lett. 2014, 14, 3055. [284] Y. Liu, H. Wu, H.-C. Cheng, S. Yang, E. Zhu, Q. He, M. Ding, D. Li, J. Guo, N. O. Weiss, Y. Huang, X. Duan, Nano Lett. 2015, 15, 3030. [285] L. Wang, I. Meric, P. Y. Huang, Q. Gao, Y. Gao, H. Tran, T. Taniguchi, K. Watanabe, L. M. Campos, D. A. Muller, J. Guo, P. Kim, J. Hone, K. L. Shepard, C. R. Dean, Science 2013, 342, 614. [286] R. Kappera, D. Voiry, S. E. Yalcin, B. Branch, G. Gupta, A. D. Mohite, M. Chhowalla, Nature Mater. 2014, 13, 1128. [287] Y. Pak, W. Park, S. Mitra, A. A. S. Devi, K. Loganathan, Y. Kumaresan, Y. Kim, B. Cho, G.- Y. Jung, M. M. Hussain, I. S. Roqan, Small 2018, 14, 1703176. [288] B. Radisavljevic, A. Kis, Nature Mater. 2013, 12, 815. [289] H. Liu, P. D. Ye, IEEE Electr. Device L. 2012, 33, 546. [290] Y. Ge, W. Wan, W. Feng, D. Xiao, Y. Yao, Phys. Rev. B 2014, 90, 035414. [291] W. Zheng, T. Xie, Y. Zhou, Y. L. Chen, W. Jiang, S. Zhao, J. Wu, Y. Jing, Y. Wu, G. Chen, Y. Guo, J. Yin, S. Huang, H. Q. Xu, Z. Liu, H. Peng, Nature Commun. 2015, 6, 6972. [292] C.-H. M. Chuang, P. R. Brown, V. Bulovic, M. G. Bawendi, Nature Mater. 2014, 13, 796. [293] D. Shi, V. Adinolfi, R. Comin, M. Yuan, E. Alarousu, A. Buin, Y. Chen, S. Hoogland, A. Rothenberger, K. Katsiev, Y. Losovyj, X. Zhang, P. A. Dowben, O. F. Mohammed, E. H. Sargent, O. M. Bakr, Science 2015, 347, 519. [294] A. Sahu, A. Khare, D. D. Deng, D. J. Norris. Chem. Commun. 2012, 48, 5458. [295] A. Tada, Y. Geng, Q. Wei, K. Hashimoto, K. Tajima, Nature Mater. 2011, 10, 450. [296] Z. He, C. Zhong, X. Huang, W.-Y. Wong, H. Wu, L. Chen, S. Su, Y. Cao, Adv. Mater. 2011, 23, 4636. [297] A. A. Balandin, Nature Nanotech. 2013, 8, 549. [298] A. N. Pal, A. Ghosh, Appl. Phys. Lett. 2009, 95, 082105. [299] Y. Zhang, E.E. Mendez, X. Du, ACS Nano 2011, 5, 8124. 60 [300] M. A. Stolyarov, G. Liu, S. L. Rumyantsev, M. Shur, A. A. Balandin, Appl. Phys. Lett. 2015, 107, 023106. [301] C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, K. L. Shepard, J. Hone, Nature Nanotech. 2010, 5, 722. [302] J. Renteria, R. Samnakay, S. L. Rumyantsev, C. Jiang, P. Goli, M. S. Shur, A. A. Balandin, Appl. Phys. Lett. 2014, 104, 153104. [303] V. K. Sangwan, H. N. Arnold, D. Jariwala, T. J. Marks, L. J. Lauhon, M. C. Hersam, Nano Lett. 2013, 13, 4351. [304] T. Paul, S. Ghatak, A. Ghosh, Nanotechnology 2016, 27, 125706. [305] D. Sharma, M. Amani, A. Motayed, P. B. Shah, A. G. Birdwell, S. Najmaei, P. M. Ajayan, J. Lou, M. Dubey, Q. Li, A. V Davydov, Nanotechnology 2014, 25, 155702. [306] S. Najmaei, Z. Liu, W. Zhou, X. Zou, G. Shi, S. Lei, B. I. Yakobson, J.-C. Idrobo, P. M. Ajayan, J. Lou, Nature Mater. 2013, 12, 754. [307] V. Kochat, C. S. Tiwary, T. Biswas, G. Ramalingam, K. Hsieh, K. Chattopadhyay, S. Raghavan, M. Jain, A. Ghosh, Nano Lett. 2015, 16, 562. [308] K. R. Amin, A. Bid, ACS Appl. Mater. Interfaces 2015, 7, 19825. [309] P. Karnatak, T. Paul, S. Islam, A. Ghosh, Advances in Physics: X 2017, 2, 428. [310] X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Velamakanni, I. Jung, E. Tutuc, S. K. Banerjee, L. Colombo, R. S. Ruoff, Science 2009, 324, 1312. [311] S. Rahimi, L. Tao, S. F. Chowdhury, S. Park, A. Jouvray, S. Buttress, N. Rupesinghe, K. Teo, D. Akinwande, ACS Nano 2014, 8, 10471. [312] S. Bae, H. Kim, Y. Lee, X. Xu, J.-S. Park, Y. Zheng, J. Balakrishnan, T. Lei, H. R. Kim, Y. I. Song, Y.-J. Kim, K. S. Kim, B. Özyilmaz, J.-H. Ahn, B. H. Hong, S. Iijima, Nature Nanotech. 2010, 5, 574. [313] S. M. Eichfeld, L. Hossain, Y.-C. Lin, A. F. Piasecki, B. Kupp, A. G. Birdwell, R. A. Burke, N. Lu, X. Peng, J. Li, A. Azcatl, S. McDonnell, R. M. Wallace, M. J. Kim, T. S. Mayer, J. M. Redwing, J. A. Robinson, ACS Nano 2015, 9, 2080. 61 [314] Z. Yang, W. Jie, C.-H. Mak, S. Lin, H. Lin, X. Yang, F. Yan, S. P. Lau, J. Hao, ACS Nano 2017, 11, 4225. [315] B. J. Carey, J. Z. Ou, R. M. Clark, K. J. Berean, A. Zavabeti, A. S. R. Chesman, S. P. Russo, D. W. M. Lau, Z.-Q. Xu, Q. Bao, O. Kavehei, B. C. Gibson, M. D. Dickey, R. B. Kaner, T. Daeneke, K. Kalantar-Zadeh, Nature Commun. 2017, 8, 14482. [316] A-R. Jang, S. Hong, C. Hyun, S. I. Yoon, G. Kim, H. Y. Jeong, T. J. Shin, S. O. Park, K. Wong, S. K. Kwak, N. Park, K. Yu, E. Choi, A. Mishchenko, F. Withers, K. S. Novoselov, H. Lim, H. S. Shin, Nano Lett. 2016, 16, 3360. [317] L. Gao, G.-X. Ni, Y. Liu, B. Liu, A. H. Castro Neto, K. P. Loh, Nature 2014, 505, 190. [318] F. Pizzocchero, L. Gammelgaard, B. S. Jessen, J. M. Caridad, L. Wang, J. Hone, P. Bøggild, T. J. Booth, Nature Commun. 2016, 7, 11894. [319] H. D. Phan, Y. Kim, J. Lee, R. Liu, Y. Choi, J. H. Cho, C. Lee, Adv. Mater. 2017, 29, 1603928. [320] J. N. Coleman, M. Lotya, A. O'Neill, S. D. Bergin, P. J. King, U. Khan, K. Young, A. Gaucher, S. De, R. J. Smith, I. V. Shvets, S. K. Arora, G. Stanton, H. Y. Kim, K. Lee, G. T. Kim, G. S. Duesberg, T. Hallam, J. J. Boland, J. J. Wang, J. F. Donegan, J. C. Grunlan, G. Moriarty, A. Shmeliov, R. J. Nicholls, J. M. Perkins, E. M. Grieveson, K. Theuwissen, D. W. McComb, P. D. Nellist, V. Nicolosi, Science 2011, 331, 568. 62 Figure 1. Crystalline structures of 2D atomic crystals including graphene, TMDs, h-BN and Xenes. Reproduced with permission.[21] Copyright 2014, Nature Publishing Group. Figure 2. Schematic representation of the photo-thermoelectric (a) and bolometric (b) effect in graphene detectors. The red shaded area indicates elevated electron temperature with the temperature gradient (ΔT) and the resistance change across the channel (ΔR); S1 and S2 are the Seebeck coefficient in graphene areas with different doping. (c) Scheme of the graphene pyroelectric bolometer, where the conductance of a single layer graphene channel is modulated by the pyroelectric substrate and by a floating gate. (a,b) Reproduced with permission.[100] Copyright 2014, Nature Publishing Group. (c) Reproduced with permission.[104] Copyright 2017, Nature Publishing Group. 63 Figure 3. (a) Schematic diagram of lateral monolayer WSe2 p-n diode with split-gate electrodes for varying optoelectronic applications. The metal gates underneath are separated by an insulator from WSe2 channel. (b) I -- V characteristics of the WSe2 devices under optical illumination with 1,400 W m−2. Top inset: Schematic of experiment. Lower inset: electrical power Pel versus voltage under incident illumination. (c) Three-dimensional schematic diagrams of lateral MoS2 p-n diodes using chemically doping. (d) Atomic-resolution Z-contrast STEM images of the in-plane interface between WS2 and MoS2 domains. The red dashed lines highlight the atomically sharp interface along the zigzag-edge direction. (e) Optical images of five-junction MoSe2-WSe2 heterostructures. The scale bar is 10 µm. (f) Micrograph of a MoSe2-WSe2 single junction based device. (g) Photocurrent Iph as a function of the illumination power P. The red line is a linear fit, indicating a linear dependence of Iph on P at high bias. (a) Reproduced with permission.[115] Copyright 2014, Nature Publishing Group. (b) Reproduced with permission.[112] Copyright 2014, Nature Publishing Group. (c) Reproduced with permission.[122] Copyright 2014, American Chemical Society. (d) Reproduced with permission.[125] Copyright 2014, Nature Publishing Group. (e-f) Reproduced with permission.[127] Copyright 2018, Nature Publishing Group. 64 Figure 4. (a) Schematics of the atomic-layer heterostructure (top) and the n+-n heterojunction device (bottom). (b) Interlayer photocurrent vs VSD at T = 300, 340 and 350 K. Optical illumination is focused at the heterostructure with a laser at λ = 900 nm, P = 17 µW (see inset schematic). (c) Ids-Vds curves (in logarithmic scale) with and without illumination. The wavelength of the laser is 4.034 μm, and the power density is 1.09 W cm−2. Inset: Optical image of a typical b-AsP/MoS2 heterostructure device. Scale bar, 5 μm. (d) Wavelength dependence of D* at Vds = 0 V. The purple and dark lines are commercial specific detectivity for a thermistor bolometer and PbSe MWIR detectors, respectively, at room temperature. (e) Schematic representation of photoexcited charge carrier dynamics in graphene-TMDs-graphene heterostructures in out-of-plane direction. Following pulsed-laser excitation, electron -- hole pairs are created, separated and transported to the graphene electrodes. (f) Schematic band diagram for Gr/WS2/Gr heterostructure with a built-in electric field to separate the generated e-h pairs. (g) Schematic representation of separation and transfer of photoexcited electrons and holes in Gr-MoS2-Gr heterostructures. (h) External quantum efficiency in this device as a function of light power with different wavelength. (a,b) Reproduced with permission.[136] Copyright 2017, Nature Publishing Group. (c,d) Reproduced with permission.[140] Copyright 2017, American Association for the Advancement of Science. (e) Reproduced with permission.[144] Copyright 2016, Nature Publishing Group. (f) Reproduced with permission.[142] Copyright 2013, American Association for the Advancement of Science. (g,h) Reproduced with permission.[143] Copyright 2013, Nature Publishing Group. 65 Figure 5. (a) Schematic of graphene-CQDs hybrid photodetectors, where graphene acts as carrier transport channel and CQDs act as strong light absorption layers. (b) Resistance as a function of back- gate voltage for the graphene -- quantum dots structure for increasing illumination intensities. Increasing the illumination leads to a photogating effect that shifts the Dirac point to higher back-gate voltage due to the hole doping on graphene channel. Inset: two-dimensional plot of graphene resistance as a function of optical power. (c) Schematic of graphene-PbS QDs hybrid photodetector integrated with a top transparent electrodes. (d) Responsivity and EQE of the phototransistor as function of applied VTD. (e) Schematic of device architecture based on graphene-MoS2 heterostructures. (f) Schematic of charge exchange process for Vg ≪ VT at interface between graphene and MoS2. (g) Schematic diagram of WS2- graphene photodetector with ionic polymer as top gate. (h) Responsivity of the device as a function of light power. (i) Schematic of phototransistors composed of a pair of stacked graphene layers sandwiching a thin tunnel barrier (5nm Ta2O5). (j) Schematic of band diagram and photoexcited hot carrier transport under light illumination. Vertical arrows represent photoexcitation, and lateral arrows represent tunnelling of hot electron (grey) and hole (red). (k) I-V characteristics of the device under 66 different laser powers. Inset: Energy band diagram of the graphene/Ta2O5 /graphene heterostructures. (a,b) Reproduced with permission.[147] Copyright 2012, Nature Publishing Group. (c,d) Reproduced with permission.[154] Copyright 2016, Nature Publishing Group. (e,f) Reproduced with permission.[162] Copyright 2013, Nature Publishing Group. (i-k) Reproduced with permission.[165] Copyright 2014, Nature Publishing Group. Figure 6. (a) Band diagram of a MoS2 based photoconductive detector, taking into account hole trapping in the traps states closed to the valence band. (b)Transfer characteristics and photo-response dynamic of monolayer MoS2 based phototransistors with HfO2 encapsulation. (c) Schematic 3D view of MoS2 based MESFET with 6 μm long channel, 3 μm long NiOx gate. (d) Transfer curves of the MoS2 based MESFET under dark and green light illumination, Inset is temporal response of the devices. (a) Reproduced with permission.[100] Copyright 2014, Nature Publishing Group. (b) Reproduced with permission.[50] Copyright 2015, American Chemical Society. (c,d) Reproduced with permission.[177] Copyright 2015, American Chemical Society. 67 Figure 7. (a) Schematic of MoS2-PbS QDs hybrid phototransistors. (b) Charge transfer at the inerface between MoS2 and QDs under light illumination. (c) On/Off ratio versus dark current density in pristine MoS2, MoS2/PbS and MoS2/TiO2/PbS based phototransistors, the utilization of TiO2 buffer layer at interface preserves the large On/Off ratio and low dark current after QDs sensitization. (d) Schematic diagram of MoS2 and HgTe QDs hybrid based phototransistor with TiO2 buffer layer. (e) Detectivity and responsivity spectral in the depletion mode. (f) Schematic diagram of MoS2 phototransistors integrated with an out-of-plane p-n homojunction, where the built-in field in out-of-plane direction can facilitate the photoexcited carrier separation and lead to the photo-gating effect. (g) Detectivity of the detector as a function of back gate at bandwidth of 1Hz and 10Hz. (c) Reproduced with permission.[180] Copyright 2016, American Chemical Society. (f, g) Reproduced with permission.[188] Copyright 2017, Nature Publishing Group. 68 Figure 8. (a) Schematic illustration of hot electron injection from a metal electrode to MoS2. Eg represents the bandgap of MoS2. (b) Schematic illustration of the MoS2 device and the optical setup scanning across the MoS2 channel. (c) Line profiles of the photocurrent response by scanning the laser beam along the channel with different wavelength. The orange background indicates electrode positions. The appeared photocurrent from NIR laser illumination is owing to the hot electrons from metal to MoS2. (d) Schematic illustration (left panel) and SEM images (right panel) of gold heptamer array sandwiched between two monolayer graphene sheets. The scale bar is 1 µm. (e) Photocurrent measurements show antisymmetric photocurrent responses from the different regions of the device corresponding to specific plasmonic antenna geometries, obtained along the line scan direction in inset. (a-c) Reproduced with permission.[207] Copyright 2015, American Chemical Society. (d,e) Reproduced with permission.[200] Copyright 2012, American Chemical Society. 69 Figure 9. (a) Schematic illustration of graphene-organic semiconductor based vertical field effect transistors (VFETs). (b) Transfer characteristics of Graphene-C60 VFETs. (c) Schematic illustration of a graphene-PTB7 hybrid photodetector. Inset shows the side view of the device. (d) Responsivity of the SiO2 and ODTS devices with respect to light intensities. (a,b) Reproduced with permission.[220] Copyright 2015, American Chemical Society. (c,d) Reproduced with permission.[226] Copyright 2017, American Chemical Society. 70 Figure 10. (a) Photograph (left panel) and schematic diagram (right panel) of all-2D based flexible photodetectors. (b) Schematic of intercalating graphene scrolls in between graphene layers. (c) Schematic of monolayer black phosphorus transistors on flexible substrate. (d) Schematic diagram of graphene-Si based UV photodetectors. (e) Schematic of the graphene photodetectors integrated with silicon waveguide. (f) Cross-sectional schematic of the encapsulated bilayer MoTe2 p -- n junction on top of a silicon PhC waveguide. (g) Schematic diagram illustrating the cross-sectional view of the tunnel field-effect transistors (TFET) composing of bilayer MoS2 as the channel and degenerately doped p- type Ge as the source. (a) Left panel. Reproduced with permission.[251] Copyright 2014, American Chemical Society. Right panel. Reproduced with permission.[249] Copyright 2013, American Chemical Society. (b) Reproduced with permission.[255] Copyright 2017, American Association for the Advancement of Science. (c) Reproduced with permission.[256] Copyright 2015, American Chemical Society. (d) Reproduced with permission.[264] Copyright 2017, Nature Publishing Group. (e) Reproduced with permission.[269] Copyright 2013, Nature Publishing Group. (f) Reproduced with permission.[270] Copyright 2017, Nature Publishing Group. (g) Reproduced with permission.[271] Copyright 2015, Nature Publishing Group. 71 Figure 11. (a) Computer-rendered impression of the CVD graphene transfer process on a single die (real dimensions 15.1 mm height, 14.3 mm width) containing an image sensor read-out circuit that consists of 388 × 288 pixels. (b) Side view explaining the graphene phototransistor and the underlying read-out circuit. The bottom panel represents 3D impression of the monolithic image sensor displaying the top level with graphene carved into S-shaped channels sensitized with a layer of quantum dots, vertical interconnects and underlying CMOS read-out circuitry. (c) Digital camera set-up: the image sensor plus lens module captures the light reflected off objects that are illuminated by an external light source. The right panel is the near-infrared (NIR) and short-wave infrared (SWIR) light photograph of an apple and pear. Reproduced with permission.[274] Copyright 2017, Nature Publishing Group. 72 Figure 12. (a) Comparison of responsivity versus response time between commercial photodetectors and 2D or hybrid based photodetectors. (b) Specific detectivity D* versus response wavelength in conventional, 2D and hybrid photodetectors. Figure 13. (a) A transparent ultra-large area graphene film transferred on a 35-inch PET sheet. (b) Schematic of the roll-based production of graphene films grown on a copper foil. (c) Photographs of monolayer MoS2 and WS2 films grown on 4-inch fused silica substrates, with diagrams of their respective atomic structures. The left halves show the bare fused silica substrate for comparison. (d) Diagram of MOCVD growth setup. Precursors were introduced to the growth setup with individual mass flow controllers (MFC). (e) Optical image of water-based two-dimensional crystal inks including 2D TMDs, h-BN and graphene. (f) Sketch of the fabricated logic memory devices using inkjet-printed method with 2D crystal inks. (a,b) Reproduced with permission.[312] Copyright 2010, Nature Publishing Group. (c,d) Reproduced with permission.[35] Copyright 2015, Nature Publishing Group. (e,f) Reproduced with permission.[83] Copyright 2017, Nature Publishing Group. 73 74
1903.10306	1	1903	2018-12-20T00:39:35	Scaling Law for Criticality Conditions in Heterogeneous Energetic Materials under Shock Loading	[ "physics.app-ph", "physics.comp-ph" ]	Initiation in heterogeneous energetic material (HEM) subjected to shock loading occurs due to the formation of hot spots. The criticality of the hot spots governs the initiation and sensitivity of HEMs. In porous energetic materials, collapse of pores under impact leads to the formation of hot spots. Depending on the size and strength of the hot spots chemical reaction can initiate. The criticality of the hot spots is dependent on the imposed shock load, void morphology and the type of energetic material. This work evaluates the relative importance of material constitutive and reactive properties on the criticality condition of spots. Using a scaling-based approach, the criticality criterion for cylindrical voids as a function of shock pressure, Ps and void diameter, Dvoid is obtained for two different energetic material HMX and TATB. It is shown that the criticality of different energetic materials is significantly dependent on their reactive properties.	physics.app-ph	physics	A. Nassar, N. K. Rai, O. Sen, and H.S. Udaykumar *Department of Mechanical Engineering University of Iowa, Iowa City, IA, USA Shock Loading Abstract. Initiation in heterogenous energetic material (HEM) subjected to shock loading occurs due to the formation of hot spots. The criticality of the hot spots governs the initiation and sensitivity of HEMs. In porous energetic materials, collapse of pores under impact leads to the formation of hot spots. Depending on the size and strength of the hot spots chemical reaction can initiate. The criticality of the hot spots is dependent on the imposed shock load, void morphology and the type of energetic material. This work evaluates the relative importance of material constitutive and reactive properties on the criticality condition of spots. Using a scaling-based approach, the criticality criterion for cylindrical voids as a function of shock pressure, 𝑃𝑃𝑠𝑠 and void diameter, 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is obtained for two different energetic material HMX and TATB. It is shown that the criticality of different energetic materials is significantly dependent on their reactive properties. Scaling Law for Criticality Conditions in Heterogeneous Energetic Materials under Introduction regions known as The meso-scale structure of heterogenous energetic materials (HEMs) is replete with defects such as cracks and voids. These voids collapse when the HEMs are subjected to shock loading, leading to the formation of high temperature localized the hot-spots. Chemical reaction is initiated in HEMs once the formed hot-spot intensity (temperature and size) threshold value. Sensitivity and exceeds a initiation of a given HEM depends on the dynamics of void-collapse and criticality of the hot spots at the meso-scale. Therefore, it is important to understand and establish the criticality criterion of hot spots. Tarver et al.1 proposed critical hot spot curves for HMX and TATB which are based on the hot- spot size and temperature. Tarver et al.1 assumed a cylindrical/spherical hot-spot with a uniform temperature of a certain size in an otherwise unreacted material (HMX or TATB). The criticality criterion of the hot spots for different combinations of size and temperature was established by solving a reaction-diffusion system. If there is insufficient energy in the hotspot (the hotspot is too small or of low intensity, i.e. temperature) diffusion may transport heat away from the hotspot before complete consumption of the surrounding HEM material. Tarver and coworkers1 determined the boundary between the go/no-go regions, corresponding to "critical" hotspots. Large hotspots (millimeter sized ones that may arise in impact scenarios) may take much longer to reach criticality. Smaller hotspots that are micron-sized take much shorter times to reach explosion (order of micro-seconds or sub- microseconds). Note that the criticality of the smaller hotspots requires higher temperature. collapse is sub- and super-critical. This allows for While the Tarver hotspot curve1 indicates whether a hotspot is critical or not, it does not indicate how that hotspot was created in the first place. It is important to know what void size must collapse under what loading condition to form a hotspot of critical size and temperature. This is an issue that is addressed in this work by examining the regions in the parameter space -- defined by shock pressure 𝑃𝑃𝑠𝑠 and void size 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 -- where void prediction of the criticality hypersurface 𝛴𝛴𝑐𝑐𝑐𝑐= 𝛴𝛴 (𝑃𝑃𝑠𝑠,𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣). This criticality criterion has The criticality criterion, 𝛴𝛴𝑐𝑐𝑐𝑐=𝛴𝛴 (𝑃𝑃𝑠𝑠,𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣) is the advantage of being defined in terms of operating conditions of a HEM sensitivity experiment, rather than in terms of hotspot size and temperature. dependent on the material constitutive and reactive properties of the energetic material. Using the criticality criteria for HMX and TATB, the relative importance of constitutive and reaction kinetics on criticality condition is also established in the current work. Methods In framework, the current Eulerian = 0 (2) the governing equations are comprised of a set of hyperbolic conservation laws corresponding to the conservation of mass, momentum and energy: 𝜕𝜕𝜌𝜌𝜕𝜕𝜕𝜕+𝜕𝜕(𝜌𝜌𝑢𝑢𝑖𝑖) 𝜕𝜕𝑥𝑥𝑖𝑖 = 0 (1) 𝜕𝜕(𝜌𝜌𝑢𝑢𝑖𝑖)𝜕𝜕𝜕𝜕 +𝜕𝜕�𝜌𝜌𝑢𝑢𝑖𝑖𝑢𝑢𝑗𝑗−𝜎𝜎𝑖𝑖𝑗𝑗� 𝜕𝜕𝑥𝑥𝑗𝑗 𝜕𝜕(𝜌𝜌𝐸𝐸)𝜕𝜕𝜕𝜕 +𝜕𝜕�𝜌𝜌𝐸𝐸𝑢𝑢𝑗𝑗−𝜎𝜎𝑖𝑖𝑗𝑗𝑢𝑢𝑖𝑖� 𝜕𝜕𝑥𝑥𝑗𝑗 where ρ, and 𝑢𝑢𝑣𝑣, respectively denote the density, and the velocity components, 𝐸𝐸=𝑒𝑒+0.5 𝑢𝑢𝑣𝑣𝑢𝑢𝑣𝑣 is tensor, 𝜎𝜎𝑣𝑣𝑖𝑖, is decomposed into volumetric and 𝜎𝜎𝑣𝑣𝑖𝑖=𝑆𝑆𝑣𝑣𝑖𝑖−𝑝𝑝𝛿𝛿𝑣𝑣𝑖𝑖 (4) the specific total energy, and e is the specific internal energy of the mixture. The Cauchy stress = 0 (3) deviatoric components, i.e., the following evolution equation: The deviatoric stress tensor, 𝑆𝑆𝑣𝑣𝑖𝑖, is evolved using 𝜕𝜕𝜕𝜕 +𝜕𝜕�𝜌𝜌𝑆𝑆𝑖𝑖𝑗𝑗𝑢𝑢𝑘𝑘� 𝜕𝜕�𝜌𝜌𝑆𝑆𝑖𝑖𝑗𝑗� 2𝜌𝜌𝐺𝐺𝐷𝐷𝑣𝑣𝑖𝑖=0 where, 𝐷𝐷𝑣𝑣𝑖𝑖 is the strain rate tensor, and G is the 𝜕𝜕𝑥𝑥𝑘𝑘 +23𝜌𝜌𝐺𝐺𝐷𝐷𝑘𝑘𝑘𝑘𝛿𝛿𝑣𝑣𝑖𝑖− shear modulus of material. (5) Constitutive and Reaction Models for HMX is modeled using The material models for HMX used to perform the current analysis are based on the work of Menikoff et al.2. A Birch-Murnaghan equation of state is used for the dilatational response of HMX. The equation of state properties are provided in Menikoff et al.3. The deviatoric response is obtained by modeling the perfectly plastic behavior of HMX under shock loading. Void collapse under shock loading can lead to the melting of HMX; therefore thermal softening of HMX the Kraut-Kennedy relation with model parameters provided in the work of Menikoff et al.2. Once the temperature exceeds the melting point of HMX the deviatoric strength terms are set to zero. The specific heat of HMX is known to change significantly with temperature. The variation of specific heat is modeled as a function of temperature as suggested by Menikoff et al.3. The chemical decomposition of HMX is modeled using Tarver-Nichols 3 step reaction model1. A detailed description of the Eulerian solver and the implementation of the HMX constitutive models is presented in the previous work4, 5. Constitutive models for TATB Equation of state obtained from the Mie-Grüneisen form of the equation of state: The hydrostatic pressure, 𝑝𝑝, in Eq. (4) is 𝑝𝑝(𝜌𝜌,𝑒𝑒)=𝑝𝑝𝑠𝑠(𝜌𝜌)+𝜌𝜌 𝛤𝛤 [𝑒𝑒− 𝑒𝑒𝑠𝑠(𝜌𝜌)] (6) 3 𝑖𝑖=1 used to find the isentrope pressure6: For TATB, the Davis equation of state form is pressure and specific internal energy, respectively. material dependent parameters and their values are obtained from Davis et al.6. The isentrope specific internal energy for HMX and TATB is defined as: where 𝜌𝜌 is the density, 𝛤𝛤 is the Mie-Grüneisen parameter, and 𝑝𝑝𝑠𝑠(𝜌𝜌) and 𝑒𝑒𝑠𝑠(𝜌𝜌) are the isentrope 𝑝𝑝𝑠𝑠(𝜌𝜌) ⎩⎪⎪⎨⎪⎪⎧𝑝𝑝 [exp(4𝐵𝐵𝐵𝐵)−1] ; 𝜌𝜌< 𝜌𝜌𝑣𝑣 +𝐶𝐶(4𝐵𝐵𝐵𝐵)44! 𝑝𝑝 ⎣⎢⎢⎢⎢⎡�(4𝐵𝐵𝐵𝐵)𝑖𝑖 ⎦⎥⎥⎥⎥⎤ ; 𝜌𝜌≥ 𝜌𝜌𝑣𝑣 (7) 𝑗𝑗! = 𝐵𝐵2 + (1−𝐵𝐵)4 where 𝐵𝐵=1− 𝜌𝜌𝑣𝑣/𝜌𝜌 , 𝜌𝜌𝑣𝑣 is the reference ambient density, 𝑝𝑝= 𝜌𝜌𝑣𝑣𝐴𝐴2/4𝐵𝐵, and A, B, and C are 𝑑𝑑�1𝜌𝜌� 𝑒𝑒𝑠𝑠(𝜌𝜌)=𝑒𝑒0− � 𝑝𝑝𝑠𝑠(𝜌𝜌) 1/𝜌𝜌 1/𝜌𝜌0 where 𝑒𝑒0 is the integration constant and represents 𝑇𝑇(𝑒𝑒,𝜌𝜌)=𝑇𝑇𝑠𝑠(𝜌𝜌)�1+𝛼𝛼𝑠𝑠𝜕𝜕 𝐶𝐶𝑣𝑣0𝑇𝑇𝑠𝑠(𝜌𝜌)[𝑒𝑒−𝑒𝑒𝑠𝑠(𝜌𝜌) +1]� 11+𝛼𝛼𝑠𝑠𝑠𝑠 𝑇𝑇 𝑠𝑠(𝜌𝜌)=�𝑇𝑇𝑣𝑣 �𝜌𝜌𝜌𝜌𝑜𝑜�𝛤𝛤 𝑜𝑜 ; 𝜌𝜌< 𝜌𝜌𝑣𝑣 𝑇𝑇𝑣𝑣𝑒𝑒𝑒𝑒𝑝𝑝(−𝑍𝑍𝐵𝐵)�𝜌𝜌𝜌𝜌𝑜𝑜�𝛤𝛤 𝑜𝑜; 𝜌𝜌≥ 𝜌𝜌𝑣𝑣 𝛤𝛤(𝜌𝜌)=� 𝛤𝛤𝑣𝑣 ; 𝜌𝜌< 𝜌𝜌𝑣𝑣 𝛤𝛤𝑣𝑣+𝑍𝑍 𝐵𝐵 ; 𝜌𝜌≥ 𝜌𝜌𝑣𝑣 The temperature in the HMX is obtained from the calculated internal energy in Eq. (8) using the relationship7: the stored chemical energy of the explosive. The Grüneisen coefficient for the TATB is where, given by6, 7, (8) (9) (10) (11) equals to 0.81686, Z is a constant equal to 0.30936. The specific heat for TATB is calculated as6: where 𝛤𝛤𝑐𝑐𝑣𝑣 the reference Grüneisen coefficient 𝐶𝐶𝑣𝑣(𝜌𝜌,𝑇𝑇)= 𝐶𝐶𝑣𝑣𝑣𝑣 � 𝑇𝑇𝑇𝑇𝑠𝑠(𝜌𝜌)�𝛼𝛼𝑠𝑠𝑠𝑠 where 𝐶𝐶𝑣𝑣𝑣𝑣 is the reference specific heat which equals to 837 J/Kg.K8, 𝛼𝛼𝑠𝑠𝜕𝜕 is a parameter that determines the specific heat change with respect to temperature and its value taken from Davis et al.6 to be 0.7331. Strength model for TATB (12) in is used The TATB material is assumed to exhibit perfectly plastic behavior. The deviatoric part of the perfectly plastic material response. The radial return algorithm for a perfectly plastic material given by Ponthot9 the current framework. The radial return algorithm evolves the stress tensor, 𝑆𝑆 (Eq. (4)) is modeled to capture the stress deviator, 𝑆𝑆 as an elastic response first The values of yield stress, Y and shear modulus, 𝐺𝐺 using a predictor step (Eq. (5)). Then the stress deviator is mapped back to the yield surface using a corrector step that enforces the plastic flow rule. for TATB used in the calculations are obtained from the work of Najjar et al.8. Under high shock compression HEMs can reach melting temperatures and lose their strength. The degradation of TATB strength under melting is modeled in the current framework The melting temperature is taken to be a constant value for the TATB, which is 623 K8. Therefore, as the temperature of the melt temperature the deviatoric component of the stress tensor is set to zero. Reaction modeling of TATB the HMX reaches The thermal decomposition of TATB are modeled using a multistep chemical kinetic model Chemical proposed by Tarver decomposition of TATB takes place in 3 steps involving four different species. The three steps are given as, al.1. et (13) (14) (15) (18) (17) (16) Reaction1: Reaction3: Reaction2: TATB decompositions are given as, The rate equations for all the species in the 𝑇𝑇𝐴𝐴𝑇𝑇𝐵𝐵 (𝐶𝐶6𝐻𝐻6𝑁𝑁6𝑂𝑂6)→𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑑𝑑 2: 𝑓𝑓𝑢𝑢𝑓𝑓𝑆𝑆𝑒𝑒𝑓𝑓𝑓𝑓𝑓𝑓,𝑓𝑓𝑢𝑢𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓,(𝐻𝐻2,𝑓𝑓𝑓𝑓𝑑𝑑 𝐻𝐻2𝑂𝑂) 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑑𝑑 2→𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑑𝑑 3:𝑓𝑓𝑢𝑢𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓, 𝑆𝑆𝑜𝑜ℎ𝑒𝑒𝑓𝑓 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑒𝑒𝑓𝑓𝑜𝑜𝑓𝑓,(𝐻𝐻2𝑂𝑂) 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑑𝑑 3→𝑓𝑓𝑆𝑆𝑓𝑓𝑓𝑓𝑆𝑆 𝑓𝑓𝑓𝑓𝑓𝑓𝑒𝑒𝑓𝑓 (𝑁𝑁2,𝐶𝐶𝑂𝑂,𝐶𝐶𝑂𝑂2,𝐻𝐻2𝑂𝑂) 𝑌𝑌1=−𝑌𝑌1𝑍𝑍1𝑒𝑒𝑒𝑒𝑝𝑝�−𝐸𝐸1𝑅𝑅𝑇𝑇� 𝑌𝑌2=𝑌𝑌1𝑍𝑍1𝑒𝑒𝑒𝑒𝑝𝑝�−𝐸𝐸1𝑅𝑅𝑇𝑇�−𝑌𝑌2𝑍𝑍2𝑒𝑒𝑒𝑒𝑝𝑝�−𝐸𝐸2𝑅𝑅𝑇𝑇� 𝑌𝑌3=𝑌𝑌2𝑍𝑍2𝑒𝑒𝑒𝑒𝑝𝑝�−𝐸𝐸2𝑅𝑅𝑇𝑇�− 𝑌𝑌32𝑍𝑍3𝑒𝑒𝑒𝑒𝑝𝑝�−𝐸𝐸3𝑅𝑅𝑇𝑇� 𝑌𝑌4= 𝑌𝑌32𝑍𝑍3𝑒𝑒𝑒𝑒𝑝𝑝�−𝐸𝐸3𝑅𝑅𝑇𝑇� where, 𝑌𝑌𝑣𝑣 is the mass fraction of the 𝑆𝑆𝜕𝜕ℎ species, 𝑍𝑍𝑣𝑣 is the frequency factor for each reaction, 𝐸𝐸𝑖𝑖 is the activation energy for each reaction, 𝑅𝑅 is the universal gas constant and T is the temperature. The values for each of these constants are obtained from the work of Tarver et al.1. The change in temperature chemical decomposition of TATB is calculated by solving the evolution equation, λ is the thermal conductivity of the reaction 𝜌𝜌𝐶𝐶𝑝𝑝𝑇𝑇=𝑄𝑄𝑅𝑅+𝜆𝜆𝜆𝜆𝑇𝑇 where, 𝜌𝜌 is the density of TATB, 𝐶𝐶𝑝𝑝 is the specific heat of the reaction mixture, 𝑇𝑇 is the temperature, mixture, 𝜆𝜆 is the Laplacian operator and 𝑄𝑄𝑅𝑅 is the 3 𝑄𝑄𝑅𝑅= �𝑄𝑄𝐼𝐼𝑌𝑌𝐼𝐼 𝐼𝐼=1 total heat release rate from all the reactions (Equations (13-15) and (16-19)) and given as, because the of (19) (20) (21) (22) (23) is 𝐼𝐼=1−3 where, the reaction number each of the reactions and its values are obtained weighted mass fraction average of the specific heat and thermal conductivity for the four species, (Equations (16-19)), 𝑄𝑄𝐼𝐼 is the energy release from from the work of Tarver et al.1. The values of 𝜆𝜆 and 𝐶𝐶𝑃𝑃 for the reaction mixture are obtained by 4 𝐶𝐶𝑃𝑃= �𝐶𝐶𝑃𝑃𝑣𝑣𝑌𝑌𝑣𝑣 𝑣𝑣=1 4 𝜆𝜆= �𝜆𝜆𝑣𝑣𝑌𝑌𝑣𝑣 𝑣𝑣=1 where, 𝐶𝐶𝑃𝑃𝑣𝑣 and 𝜆𝜆𝑣𝑣 are the specific heat capacity and thermal conductivity for the four species obtained from the work of Tarver et al.1. Results detailed verification This work is aimed towards understanding and developing criticality criterion for two different energetic materials HMX and TATB. In the previous work4, and validation of the numerical framework focusing on the void collapse in HMX is presented. In this work, the implementation of the constitutive and reaction models for TATB is verified. The results section is organized as follows, the verification of the numerical framework to perform the void collapse simulation of TATB is presented first, following that the approach to obtain the criticality condition for HMX and TATB is discussed and finally the criticality criterion for HMX and TATB is compared. Verification of the Numerical Framework The verification of the numerical framework to perform void collapse analysis in TATB is presented in this section. The results obtained from the void collapse analysis is compared against the simulation results of Najjar et al.8. Najjar et al.8 performed void collapse analysis for a 2 µ𝑓𝑓 shock load of 6 𝐺𝐺𝑃𝑃𝑓𝑓 using ALE3D code10. Fig. 1 diameter single cylindrical void under a sustained time variation of maximum compares the temperature in the primary jet collapse region4 obtained from the current work and Najjar et al. 8 simulation. The predictions from the current analysis is in good agreement with the simulations of Najjar et al.8. the critical hot spot is obtained by repeating these calculations for different hot spot sizes. The threshold curve is in good agreement with the Tarver et al. 1 result for prediction of the critical condition (Fig. 2). This verifies the parts of the implementation that pertain to chemical kinetics modeling and diffusive transport. Fig. 1. Comparison of the time variation of the maximum temperature near the primary jet collapse region with the inert void collapse simulation results of Najjar et al.6. Having verified the numerical framework for the inert void collapse analysis, the next step is to verify the implementation of the reactive model for TATB decomposition. The 3-step chemical kinetic model of Tarver et al.1 for TATB decomposition is used in the current analysis. In the work of Tarver et al.1, the 3-equation reaction kinetics for TATB was used to obtain a criticality curve for TATB initiation. The criticality curve was obtained by performing reaction-diffusion calculations in an otherwise uniform HMX material containing a hot spot. The calculations were performed by instantaneously heating a hot spot of a certain shape and diameter (𝐷𝐷ℎ𝑠𝑠) temperature (𝑇𝑇ℎ𝑠𝑠). The HMX surrounding the hot spot was kept at room temperature (293K). For verification, the current work obtains this critical hot spot curve using reaction kinetics implementation for cylindrical hot spots. To this end, the reaction diffusion calculations to compute the evolution of a cylindrical hot spot are performed by solving Eq. (20). The criterion for (cylindrical, spherical or planar) at a specified the Fig. 2. Comparison of the critical hot spot curve of TATB with the results of Tarver et Criticality Criterion for HMX and TATB The aim of the current work is to understand the relative importance of constitutive and reactive properties on the criticality criterion of two different energetic materials, i.e. HMX and TATB. The physics governing the differences in the criticality of the two material is explained in this section. In the past, Tarver et al.1 obtained critical hot spot curve for HMX and TATB by performing reaction-diffusion analysis as shown in Fig. 2. Tarver et al.1 critical hot spot criterion is expressed their criticality criterion is useful and has been used by modelers11, as a function of hot spot size, 𝐷𝐷ℎ𝑠𝑠 and hot spot temperature, 𝑇𝑇ℎ𝑠𝑠. Although, 12, it involves quantities (𝑇𝑇ℎ𝑠𝑠,𝐷𝐷ℎ𝑠𝑠) that are not such as shock load, 𝑃𝑃𝑠𝑠 and void diameter, 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣. operating parameters in a SDT experiment. It will be more useful to develop a criticality criterion that is a function of controllable input parameters Therefore, in this section a criticality criterion for cylindrical voids is derived as a function of shock strength and void size. i.e. TATB (24) thermal diffusion depending on the ratio: Criticality of a hot spot is governed by two competing mechanisms, reaction growth and thermal diffusion. A hot spot will go critical if the diffusion time scale can be obtained then criticality conditions can be established. reaction time scale 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 is smaller than the time scale 𝜏𝜏𝑣𝑣𝑣𝑣𝑑𝑑𝑑𝑑𝑢𝑢𝑠𝑠𝑣𝑣𝑣𝑣𝑟𝑟, 𝜓𝜓1= 𝜏𝜏𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑠𝑠𝑖𝑖𝑜𝑜𝑟𝑟 𝜏𝜏𝑑𝑑𝑖𝑖𝑑𝑑𝑑𝑑𝑑𝑑𝑠𝑠𝑖𝑖𝑜𝑜𝑟𝑟 Criticality requires 𝜓𝜓1<1. Therefore, if an estimate of reaction time scale, 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 and Estimate of time to ignition, 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 for HMX and The reaction time scale, 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 can be 𝑇𝑇ℎ𝑠𝑠 for a given chemical kinetic model of HMX and TATB. Therefore, for different 𝑇𝑇ℎ𝑠𝑠, 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 models of HMX and TATB1. For a given 𝑇𝑇ℎ𝑠𝑠, time 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 with 𝑇𝑇ℎ𝑠𝑠 for HMX and TATB is shown in provides 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 as a function of 𝑇𝑇ℎ𝑠𝑠 for HMX to ignition is recorded when the solid HMX/TATB is completely decomposed to form the final gaseous species. The variation of time to ignition, and TATB (Eq. (25) and (26)). The time to ignition for a given hot spot temperature is lower for HMX as compared to TATB. Therefore, for the same hot spot temperature TATB will reach ignition at a slower rate as compared to HMX. Fig. 3. An exponential fit to the data points expressed as a function of hot spot temperature, is obtained for the Tarver-Nichols 3-step reactive HMX: TATB: =2×109𝑒𝑒−0.035𝑇𝑇ℎ𝑠𝑠 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 𝐻𝐻𝐻𝐻𝐻𝐻 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 =2.6957×104𝑒𝑒−0.0109𝑇𝑇ℎ𝑠𝑠 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 Time scale of thermal diffusion, 𝜏𝜏𝑣𝑣𝑣𝑣𝑑𝑑𝑑𝑑𝑢𝑢𝑠𝑠𝑣𝑣𝑣𝑣𝑟𝑟: (25) (26) The time scale of thermal diffusion is dependent on the hot spot size (𝐷𝐷ℎ𝑠𝑠) and thermal diffusivity (𝛼𝛼𝐻𝐻𝐻𝐻𝐻𝐻=6.82×10−8𝑓𝑓2 𝑓𝑓⁄ ,𝛼𝛼𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇= 1.01 ×10−7𝑓𝑓2 𝑓𝑓⁄ )1: 𝜏𝜏𝑣𝑣𝑣𝑣𝑑𝑑𝑑𝑑𝑢𝑢𝑠𝑠𝑣𝑣𝑣𝑣𝑟𝑟≈𝐷𝐷ℎ𝑠𝑠2𝛼𝛼 (27) Fig. 3. Comparison of the time to ignition as a function of hot spot temperature for HMX and TATB. (28) Using Eq. (24) for a critical hot spot: criticality condition for Tarver-Nichols model can be obtained as: 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟< 𝜏𝜏𝑣𝑣𝑣𝑣𝑑𝑑𝑑𝑑𝑢𝑢𝑠𝑠𝑣𝑣𝑣𝑣𝑟𝑟 Substituting the expressions for 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟 and 𝜏𝜏𝑣𝑣𝑣𝑣𝑑𝑑𝑑𝑑𝑢𝑢𝑠𝑠𝑣𝑣𝑣𝑣𝑟𝑟 from Equations (25) and (27), the 𝜏𝜏𝑐𝑐𝑟𝑟𝑟𝑟𝑐𝑐𝜕𝜕𝑣𝑣𝑣𝑣𝑟𝑟< 𝐷𝐷ℎ𝑠𝑠2𝛼𝛼 characteristics, i.e. 𝑇𝑇ℎ𝑠𝑠 and 𝐷𝐷ℎ𝑠𝑠. This is similar to realized if the hot spot characteristics, 𝑇𝑇ℎ𝑠𝑠 and 𝐷𝐷ℎ𝑠𝑠 𝑃𝑃𝑠𝑠 and void size, 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣. Then Eq. (29) can yield a the Tarver et al.1 hot spot criterion. However, as mentioned earlier the aim of the current analysis is to obtain criticality criterion in terms of input parameters that can be controlled. This can be Eq. (29) provides an expression for the criticality criterion as a function of hot spot can be expressed in terms of input shock pressure, (29) performed and analyzed. the inert void collapse simulations are criticality expression in terms of experimental control parameters. respect to void diameter is also observed for TATB. This observation simplifies the functional Fig. 4 shows the temperature contours at an intermediate stage of collapse for voids of pressures, i.e. in both plasticity dominated and hydrodynamic collapse modes, is independent of The hot spot temperature, 𝑇𝑇ℎ𝑠𝑠 and size, 𝐷𝐷ℎ𝑠𝑠 can be expressed in functional forms as: 𝑇𝑇ℎ𝑠𝑠= 𝑓𝑓(𝑃𝑃𝑠𝑠,𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣) and 𝐷𝐷ℎ𝑠𝑠=ℎ(𝑃𝑃𝑠𝑠,𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣). To develop the functional form for 𝑇𝑇ℎ𝑠𝑠 and 𝐷𝐷ℎ𝑠𝑠, Ths as a function of Ps and Dvoid for HMX and TATB diameters 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣=0.5 µ𝑓𝑓 and 10 µ𝑓𝑓 under two different shock strengths 𝑃𝑃𝑠𝑠=1.27 𝐺𝐺𝑃𝑃𝑓𝑓 and 6.4 𝐺𝐺𝑃𝑃𝑓𝑓 for HMX. It is clear that 𝑇𝑇ℎ𝑠𝑠, for both void diameter. The independence of 𝑇𝑇ℎ𝑠𝑠 with form of 𝑇𝑇ℎ𝑠𝑠 to: 𝑇𝑇ℎ𝑠𝑠=𝑓𝑓(𝑃𝑃𝑠𝑠) To obtain the functional form of 𝑇𝑇ℎ𝑠𝑠, the inert void collapse simulation for the 10 µ𝑓𝑓 diameter 2 𝐺𝐺𝑃𝑃𝑓𝑓−20 𝐺𝐺𝑃𝑃𝑓𝑓. 𝑇𝑇ℎ𝑠𝑠 is computed by performing 𝑉𝑉ℎ𝑠𝑠∫ 𝑇𝑇ℎ𝑠𝑠=∫ 𝜌𝜌𝑇𝑇𝑣𝑣𝜌𝜌 𝑉𝑉ℎ𝑠𝑠 𝜌𝜌𝑣𝑣𝜌𝜌 where 𝑉𝑉ℎ𝑠𝑠 is the hot spot volume, 𝜌𝜌 is the density and 𝑇𝑇 is the local temperature in the hot spot 𝑇𝑇ℎ𝑠𝑠 with 𝑃𝑃𝑠𝑠. The figure also shows logarithmic region. Note that the hot spot in the current work is defined as the region where the temperature is higher than the bulk temperature in the shocked material. However, the minimum value of the hot spot temperature is taken to be 750 K. Fig. 5 shows the simulation data for the variation of the a volume average of temperature field in the hot spot region at the end of the void collapse: functions fitted to the data. Those functions for HMX and TATB respectively are: void is performed for input pressures ranging from (31) (30) 𝑇𝑇ℎ𝑠𝑠𝐻𝐻𝐻𝐻𝐻𝐻=𝑓𝑓𝐻𝐻𝐻𝐻𝐻𝐻(𝑃𝑃𝑠𝑠) =165.05∗𝑆𝑆𝑓𝑓𝑃𝑃𝑠𝑠+ 718.85 𝑇𝑇ℎ𝑠𝑠𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇=𝑓𝑓𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇(𝑃𝑃𝑠𝑠) =196.08∗𝑆𝑆𝑓𝑓𝑃𝑃𝑠𝑠+683.9 where 𝑇𝑇ℎ𝑠𝑠 is in K and 𝑃𝑃𝑠𝑠 is in 𝐺𝐺𝑃𝑃𝑓𝑓. (32a) (32b) It is interesting to note that the hot spot temperature values for HMX and TATB are comparable across different pressure ranges. Therefore, the constitutive properties of HMX and TATB has no significant influence on the hot spot temperature. the difference in 𝑷𝑷𝒔𝒔=𝟏𝟏.𝟐𝟐𝟐𝟐 𝑮𝑮𝑷𝑷𝑮𝑮 𝑷𝑷𝒔𝒔=𝟏𝟏.𝟐𝟐𝟐𝟐 𝑮𝑮𝑷𝑷𝑮𝑮 𝑫𝑫𝒗𝒗𝒗𝒗𝒗𝒗𝒗𝒗=𝟏𝟏𝟎𝟎 µ𝒎𝒎 𝑫𝑫𝒗𝒗𝒗𝒗𝒗𝒗𝒗𝒗=𝟎𝟎.𝟓𝟓 µ𝒎𝒎 (i) 𝑜𝑜= 0.73 𝑓𝑓𝑓𝑓 (ii) 𝑜𝑜= 14.7 𝑓𝑓𝑓𝑓 𝑷𝑷𝒔𝒔=𝟔𝟔.𝟒𝟒 𝑮𝑮𝑷𝑷𝑮𝑮 𝑷𝑷𝒔𝒔=𝟔𝟔.𝟒𝟒 𝑮𝑮𝑷𝑷𝑮𝑮 𝑫𝑫𝒗𝒗𝒗𝒗𝒗𝒗𝒗𝒗=𝟎𝟎.𝟓𝟓 µ𝒎𝒎 𝑫𝑫𝒗𝒗𝒗𝒗𝒗𝒗𝒗𝒗=𝟎𝟎.𝟓𝟓 µ𝒎𝒎 (iv) 𝑜𝑜= 5.04 𝑓𝑓𝑓𝑓 (iii) 𝑜𝑜= 0.25 𝑓𝑓𝑓𝑓 of 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣=0.5 µ𝑓𝑓 and 10 µ𝑓𝑓 under sustained shock load of 1.27 𝐺𝐺𝑃𝑃𝑓𝑓 and 6.4 𝐺𝐺𝑃𝑃𝑓𝑓 for Dhs as a function of Ps and Dvoid temperature, 𝑇𝑇ℎ𝑠𝑠 is independent of void size; the Fig. 4. Temperature (K) contour plots during the collapse of three different void diameters In the above, we show that the hot spot three shock strengths, defined identical void collapse modes in Fig. 4 are self-similar for different void sizes under loading conditions. To analyze the influence of this self- similar collapse behavior on the hot spot size, the with respect to ⁄ variation of the ratio, 𝐷𝐷ℎ𝑠𝑠 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 is compared for 𝑃𝑃𝑠𝑠=1.27 𝐺𝐺𝑃𝑃𝑓𝑓,2.38 𝐺𝐺𝑃𝑃𝑓𝑓 and 6.4 𝐺𝐺𝑃𝑃𝑓𝑓 for HMX. The value of 𝐷𝐷ℎ𝑠𝑠 = �𝐴𝐴𝑓𝑓𝑒𝑒𝑓𝑓ℎ𝑠𝑠, with the hot spot ⁄ 𝐷𝐷ℎ𝑠𝑠 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 viz. 0.5 µ𝑓𝑓, 10 µ𝑓𝑓 and 30 µ𝑓𝑓. Fig. 6 shows that ⁄ 𝐷𝐷ℎ𝑠𝑠 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 functional form for 𝐷𝐷ℎ𝑠𝑠 can be expressed as 𝐷𝐷ℎ𝑠𝑠𝐷𝐷𝑣𝑣𝑜𝑜𝑖𝑖𝑑𝑑=ℎ(𝑃𝑃𝑠𝑠) (33) varies nonlinearly with respect to shock pressure but linearly with the void diameter. This behavior for TATB. The as mentioned ratio is obtained for the three void diameters is also observed above. The Fig. 5. Variation of hot spot temperature, Ths with incident shock pressure, Ps for a given void diameter of 10 µm. Fig. 6. Variation of the ratio of hot spot diameter and void diameter with respect to different void diameters for different incident shock pressure for HMX. with respect to The data points are fitted to logarithmic functions from the inert void collapse simulations for a To obtain ℎ(𝑃𝑃𝑠𝑠), the values of 𝐷𝐷ℎ𝑠𝑠 for different input shock pressure, 𝑃𝑃𝑠𝑠 are calculated 10 µ𝑓𝑓 diameter for both HMX and TATB. Fig. 7 ⁄ shows the variation of 𝐷𝐷ℎ𝑠𝑠 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 the input shock pressure, 𝑃𝑃𝑠𝑠 for TATB and HMX. here as well to estimate ℎ(𝑃𝑃𝑠𝑠) as: 𝐷𝐷ℎ𝑠𝑠𝐻𝐻𝐻𝐻𝐻𝐻𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣=ℎ𝐻𝐻𝐻𝐻𝐻𝐻(𝑃𝑃𝑠𝑠) = 0.125∗𝑆𝑆𝑓𝑓𝑃𝑃𝑠𝑠−0.035 𝐷𝐷ℎ𝑠𝑠𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣=ℎ𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇(𝑃𝑃𝑠𝑠) = 0.179∗𝑆𝑆𝑓𝑓𝑃𝑃𝑠𝑠−0.044 where 𝑃𝑃𝑠𝑠 is in 𝐺𝐺𝑃𝑃𝑓𝑓. (34b) (34a) Comparison of criticality curve with reactive void collapse simulation data for HMX Fig. 8 shows the criticality envelope (black can be obtained from the data points and the criticality condition for HMX is expressed as: curve) in the (𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣,𝑃𝑃𝑠𝑠) space. A power-law fit 𝑃𝑃𝑠𝑠4.45𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐=2102 (38) where, 𝑃𝑃𝑠𝑠 is in 𝐺𝐺𝑃𝑃𝑓𝑓 and 𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐 is in µ𝑓𝑓. The function of 𝑃𝑃𝑠𝑠 and 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 was obtained also for is: 𝑃𝑃𝑠𝑠4.38𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐=3933 (39) above criticality relationship is obtained from scaling arguments. Using highly resolved reactive meso-scale simulations the criticality criterion as a HMX. A detailed description of the reactive Eulerian framework is provided in the previous work3. The simulation-based criticality condition simulation-based criticality criterion (Eq. (39), the magenta curve in Fig. 8) and the scaling-based criticality criterion (Eq. (38), the black curve in Fig. 8). The sub- critical and super-critical void collapse data points obtained from the reactive simulations are also shown in Fig. 8. compares Fig. 8 the The scaling-based criticality criterion is in simulation-based close agreement with criterion. Also, the scaling-based criterion can distinguish the sub-critical and critical regimes with reasonable accuracy. the The hot spot diameter for TATB is higher than HMX under similar loading condition. Therefore, TATB forms a larger hot spot as compared to HMX, however the temperature of the hot spot is observed to be in the comparable range. and TATB Equations (32) and (34) provide expressions Fig. 7. Variation of the ratio of hot spot equation) are substituted in Equation (29) to yield a hot spot criticality criterion for HMX: diameter, 𝐷𝐷ℎ𝑠𝑠 and void diameter, 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 with incident shock pressure, 𝑃𝑃𝑠𝑠 for a given void of diameter 10 µ𝑓𝑓 for TATB and HMX. Criticality as a function of 𝑃𝑃𝑠𝑠 and 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 for HMX for 𝑇𝑇ℎ𝑠𝑠 and 𝐷𝐷ℎ𝑠𝑠 as functions of 𝑃𝑃𝑠𝑠 and 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣. These expressions for 𝑇𝑇ℎ𝑠𝑠 and 𝐷𝐷ℎ𝑠𝑠 (the (a) part of each 2×109𝑒𝑒−0.035𝑔𝑔(𝑃𝑃𝑠𝑠)< 𝐷𝐷𝑣𝑣𝑜𝑜𝑖𝑖𝑑𝑑2 ℎ2(𝑃𝑃𝑠𝑠) 𝛼𝛼𝐻𝐻𝐻𝐻𝐻𝐻 size 𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐 for HMX as: 𝐻𝐻𝐻𝐻𝐻𝐻 =�2𝛼𝛼𝐻𝐻𝐻𝐻𝐻𝐻𝑟𝑟−0.035𝑔𝑔𝐻𝐻𝐻𝐻𝐻𝐻(𝑃𝑃𝑠𝑠) 𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐 (ℎ𝐻𝐻𝐻𝐻𝐻𝐻)2(𝑃𝑃𝑠𝑠) 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 =�2.6957𝛼𝛼𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑟𝑟−0.019𝑔𝑔𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇(𝑃𝑃𝑠𝑠) 𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐 ×109 (36) ×104 (37) Similarly, for TATB (using Equations. (32b) and (34b)) the criticality expression is obtained as, Rearrangement of Eq. (35) provides a critical void (ℎ𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇)2(𝑃𝑃𝑠𝑠) (35) In the present work, the differences in the criticality prediction of single cylindrical voids for HMX and TATB is analyzed. A numerical framework to perform inert/reactive void collapse simulations of HMX and TATB material is presented. Using a scaling based approach, criticality conditions for both materials are obtained as a function of input shock pressure and void diameter. The scaling based approach is shown to be an effective way to obtain criticality conditions with desirable accuracy. For HMX and TATB, the time to ignition as a function of temperature is obtained. The time to ignition for TATB is observed to be higher as compared to HMX for a given temperature. This is reflected in the criticality predictions of the two materials, where TATB is observed to be less sensitive, although the hot spot size and temperature for both materials are in the comparable range. Therefore, the difference in the constitutive properties of the two material has no significant influence on the criticality. The difference in the criticality of HMX and TATB is attributed to differences in their respective reactive properties. Fig. 8. Criticality curve as a function of incident shock load, 𝑃𝑃𝑠𝑠 and void diameter, 𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 obtained from the scaling arguments and non-dimensional number 𝜓𝜓1. Criticality criterion for TATB Similar to the criticality criterion obtained for HMX (Eq. (38)), using the expressions for time to ignition for the TATB (Eq. (26)), the criticality conditions are obtained. Criticality criterion for HMX and TATB is compared and shown in Fig. 9. (41) HMX: (40) TATB: 𝑃𝑃𝑠𝑠4.45𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐=2102 𝑃𝑃𝑠𝑠2.38𝐷𝐷𝑐𝑐𝑐𝑐𝑣𝑣𝜕𝜕𝑣𝑣𝑐𝑐𝑟𝑟𝑐𝑐=731 𝑃𝑃𝑠𝑠−𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 plane for HMX and TATB. TATB is observed to be insensitive as compared to HMX. In particular, the difference in the sensitivity of TATB and HMX varies greatly for the smaller sized voids. Fig. 9 shows the criticality functions in the Conclusions Fig. 9. Comparison of the criticality conditions for HMX and TATB in the 𝑃𝑃𝑠𝑠−𝐷𝐷𝑣𝑣𝑣𝑣𝑣𝑣𝑣𝑣 space. References 1. C. M. Tarver, S. K. Chidester, andA. L. Nichols, "Critical conditions for impact-and shock-induced hot spots in solid explosives," The Journal of Physical Chemistry 100, 5794 (1996). tri-amino-tri-nitro-benzene 2. R. Menikoff, andT. D. Sewell, "Constituent properties of HMX needed for mesoscale simulations," Combustion theory and modelling 6, 103 (2002). 3. T. D. Sewell, andR. Menikoff, COMPLETE EQUATION OF STATE FOR/3-HMX AND IMPLICATIONS FOR INITIATION (American Institute of Physics, 2003). 4. N. K. Rai, M. J. Schmidt, andH. S. Udaykumar, "High-resolution simulations of cylindrical void collapse in energetic materials: Effect of primary and secondary collapse on initiation thresholds," Physical Review Fluids 2, (2017). 5. N. K. Rai, andH. Udaykumar, "Three- dimensional simulations of void collapse in energetic materials," Physical Review Fluids 3, 033201 (2018). 6. B. Wescott, D. S. Stewart, andW. C. Davis, "Equation of state and reaction rate for condensed- phase explosives," Journal of applied physics 98, 053514 (2005). 7. T. D. Aslam, "The reactants equation of state for the (TATB) based explosive PBX 9502," Journal of Applied Physics 122, 035902 (2017). 8. F. Najjar, W. Howard, andL. Fried, GRAIN‐ SCALE HOT‐SPOT INITIATION FOR SHOCKED TATB (AIP, 2009). 9. J.-P. Ponthot, "Unified stress update algorithms for the numerical simulation of large deformation elasto-plastic and elasto-viscoplastic processes," International Journal of Plasticity 18, 91 (2002). 10. A. Nichols, "ALE3D User's Manual," Lawrence Livermore National Laboratory Tech. Rep. UCRL-MA-152204 (2007). 11. A. Barua, S. Kim, Y. Horie, andM. Zhou, "Ignition criterion for heterogeneous energetic materials based on hotspot size-temperature threshold," Journal of Applied Physics 113, 064906 (2013). 12. S. Kim, C. Miller, Y. Horie, C. Molek, E. Welle, andM. Zhou, "Computational prediction of probabilistic ignition threshold of pressed granular octahydro-1, 3, 5, 7-tetranitro-1, 2, 3, 5-tetrazocine (HMX) under shock loading," Journal of Applied Physics 120, 115902 (2016). SIMULATIONS OF
1811.12845	2	1811	2018-12-03T08:58:33	Quasi isotropic underwater acoustic carpet cloak based on latticed pentamode metafuid	[ "physics.app-ph" ]	In this work, we present a practical design of quasi isotropic underwater acoustic carpet cloak with pentamode microstructure. The quasi conformal transformation is not only used to obtain the required parameters, but also used to deform the retrieved regular pentamode material structure to the desired carpet, during which the effective parameters can be considered as not being affected too much. The nice and broadband cloak effect shows the success of the design. This work will contribute to the experimental demonstration of pentamode acoustic carpet cloak. Moreover, the technique in this work also can be used to design arbitrarily shaped devices with microstructures.	physics.app-ph	physics	Quasi isotropic underwater acoustic carpet cloak based on latticed pentamode metafluid Zhaoyong Sun,1 Xuecong Sun,1, 2 Han Jia,1, 3, ∗ Yafeng Bi,1 and Jun Yang1, 2, 3, † 1Key Laboratory of Noise and Vibration Research, Institute of Acoustics, Chinese Academy of Sciences, 21 North 4th Ring Road, Beijing 100190, China 2School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, 19A Yuquan Road, Beijing, 100049, China 3State Key Laboratory of Acoustics, Institute of Acoustics, Chinese Academy of Sciences, 21 North 4th Ring Road, Beijing, 100190, China (Dated: December 4, 2018) In this work, we present a practical design of quasi isotropic underwater acoustic carpet cloak with pentamode microstructure. The quasi conformal transformation is not only used to obtain the required parameters, but also used to deform the retrieved regular pentamode material structure to the desired carpet, during which the effective parameters can be considered as not being affected too much. The nice and broadband cloak effect shows the success of the design. This work will contribute to the experimental demonstration of pentamode acoustic carpet cloak. Moreover, the technique in this work also can be used to design arbitrarily shaped devices with microstructures. PACS numbers: 46.40.Cd INTRODUCTION Transformation method, which is used to design amaz- ing devices such as cloaks for wave controlling, was ini- tially proposed for electromagnetic waves[1, 2], and then generalized to acoustics soon[3, 4]. It uses coordinate transformations to build connections between geomet- ric space and material space. In several fields, different cloaks, designed with the transformation method to can- cel the scattering of the incident wave to hide objects, have been reported recently[1 -- 3, 5 -- 9]. Another most attracting application is the carpet cloak, which covers the object on a rigid plane and makes the scattering wave perform as being reflected by the plane, resulting in the concealment of the object[10].The idea of carpet cloak was firstly proposed by Li and Pendry[10], and then was fabricated for electromagnetic wave[11 -- 13]. The concept was extended to acoustics immediately[14, 15]. And soon latter, an acoustic carpet composed with perforated plates was designed by a lin- ear transformation, where the materials are anisotropic but homogeneous[16]. The good cloak effect was shown by the experimental results. Since then, more relative researches on carpet cloak for airbone sound have been reported[17 -- 19]. Compared with carpet cloak for airbone sound, the underwater one has more important practical applications, but suffering from difficulty in impedance matching and microstructure design[20, 21]. In our pre- vious works, we experimentally demonstrated the two and three dimensional under water carpets[22, 23]. The carpets were constructed with uniformed mental strips alternating with water layers periodically, which behave like fluid with anisotropic inertia and isotropic bulk mod- ulus. The effectiveness of the prototypes was experi- mentally demonstrated in anechoic tanks. Another ap- proach to under water carpet is using pentamode ma- terial (PM), which can be completely fabricated from rigid solids in practical case[24]. For the ideal PM in three dimensional (3D) case, the six order elastic ten- sor only has one nonzero eigenvalue. The five vanished eigenvalues correspond to five easyways to deform the materials, which leads to the zero shear modulus. This is the reason that the pentamode is so named (penta means five in Greek)[24, 25]. The only eigenmode is a hydrostatic stress. Thus, the ideal PM exhibits simi- lar acoustic properties to fluids that only the compres- sive wave can exist. In two dimensional (2D) case, the name "pentamode" is stilled inherited, even though only two of the three eiganvalues are zeros. The 2D PM can be realized by hexagonal lattice with highly tun- able effective parameters, which can be designed to have anisotropic modulus[26, 27]. The shear modulus of the latticed PM is so small that it can be neglected and regarded as fluid in some frequency bands. PM has attracted great attention in underwater acoustic wave controlling[8, 9, 28 -- 31]. It has been proved that wave equation in PM is also form-invariant under the coor- dinate transformations[25], thus PM can be used to de- sign some valuable under water devices with transfor- mation method[8, 9]. However, it is difficult to design an anisotropic PM carpet (PMC), because of the com- plexity of the transformation method and harsh bound- ary conditions of impedance matching[25, 32]. An pos- sible way for PMC is quasi conformal transformation (QCT), which will bring the quasi isotropic parame- ter distribution. The QCT method can be archived by grid deformation theory[10] or solving inverse Laplace's equations[33, 34]. Although suffering from large size and inhomogeneity, the arbitrary shapes, the minimized anisotropy and broadband character of the QCT based devices still show strong attractions[10, 11, 33, 34]. In this work, we use a 2D version of the PM lattice to design a quasi isotropic latticed PMC. The continu- ous required parameters are obtained with QCT method, which is archived by solving inverse Laplace's equations with Neuman and Dirichelet boundaries[33, 34]. In order to realize the carpet with 2D PM microstructures, the carpet is pre-divided into 300 different cells, where the discrete parameters are obtained by the area average of the continuous one. The 2D PM unit cells, with effective parameters equaling to that of the corresponding discrete ones, are acquired by retrieving the energy bands[8, 31]. Since the retrieved PM unit cells are regular hexagon, it is difficult to build a carpet. Thus, some appropriate defor- mations of the PM cells are necessary. The QCT method is reapplied to deform the retrieved PM cells in order to build the desired carpet. We firstly use the retrieved PM cells to form 300 rectangle shaped bricks which have the same parameters with the discrete ones correspondingly, and then arrange them into an array having the same size with that of the virtual space. The PM array is viewed as a new virtual space with the required param- eters of the carpet, and mapped into a carpet with the same QCT. Due to the conformal preservation, we sup- pose that the deformations will not change the effective parameters of the structure seriously. This hypothesis is further supported by the good cloak effect of simula- tion results. Cosine similarity is used to demonstrate the broadband and good performance of the latticed PMC in different directions. DESIGN METHOD Considering that the virtual space and physical space are described by complex coordinates w = u + iv and z = x+iy correspondingly, the conformal transformation optics builds a relationship between those two spaces in the following way[2, 35 -- 37]: (cid:114) 1 detJ where nz and nw denote the reflective index in physical and virtual spaces, and J is the Jacobian transformation matrix. This result is derived by mapping the Helmholtz equation in physical space into the virtual space with the analytical function w(z)[2, 36]. Since the propaga- tion of the acoustic wave in isotropic media also obeys the Helmholtz equation, the conformal transformation also can be used to design acoustic device. Thus, the relationship in Eq.(1) will still be kept in acoustics with conformal transformations. Conformal transformation provides the possibilities for isotropic carpet cloak designing. Figure 1 shows nz = nw dw dz = nw , (1) shown as follows: Thus, the inverse Jacobian matrix A = J−1, which is (cid:32) ∂u (cid:33) A = ∂x ∂v ∂x ∂u ∂y ∂v ∂y , (5) 2 the transformation of a carpet, where the virtual space ABCDEF presents the free space with reflective index nw and the physical space A(cid:48)B(cid:48)C(cid:48)D(cid:48)E(cid:48)F (cid:48) represents the carpet cloak filled with the transformed media with re- flective index nz. The boundaries A(cid:48)B(cid:48), C(cid:48)D(cid:48), E(cid:48)F (cid:48) and D(cid:48)E(cid:48), F (cid:48)A(cid:48) are all straight lines along horizontal and ver- tical directions, while the arc (cid:91)B(cid:48)C(cid:48) is an arbitrary curve. The background media is water with ρ0 = 1000 Kg/m3, K0 = 2.25 GPa, and nw = n0 = 1. According to Eq.(1), nz is determined by the transformation matrix J. How- ever, finding a suitable mapping z(w) for the designed carpet is a great challange, since analytical function z(w) determines the shape of the physical space. Quasi con- formal transformation (QCT) is an appropriate option for this situation[10, 33, 34, 38] One way to realize QCT is calculating the inverse mapping w(z) by numerically solving the inverse Laplace's equations with predefined boundary conditions[33, 34, 38]. The QCT theory suggests that the Cauchy-Riemann conditions must be satisfied in the mapping w(z): ∂u ∂x ∂u ∂y = ∂v ∂y = − ∂v ∂x which can result to the Laplace equations as: ∇2u = 0, ∇2v = 0. (2a) (2b) (3a) (3b) According to the equivalence of the external boundaries between virtual and physical spaces, the Neuman and Dirichelet boundaries are used: (cid:126)n · (cid:126)∇u(x, y)A(cid:48)B(cid:48)C(cid:48)D(cid:48),E(cid:48)F (cid:48) = 0, (cid:126)n · (cid:126)∇v(x, y)F (cid:48)A(cid:48),D(cid:48)C(cid:48) = 0, v(x, y)A(cid:48)B(cid:48)C(cid:48)D(cid:48),E(cid:48)F (cid:48) = y; (4a) u(x, y)F (cid:48)A(cid:48),D(cid:48)E(cid:48) = x. (4b) can be obtained by solving the Laplace equation Eq.(3) with the boundary Eq.(4) in a partial differential equa- tion solver. Because the calculation is numerical, it is easy to be understood that A11 = ∂u/∂x (A12 = ∂u/∂y) does not exactly equal to A22 = ∂v/∂y (−A21 = −∂v/∂x). This can lead some very soft anisotropy which can be neglected. Thus, the transformation can be regard as quasi conformal. 3 FIG. 1. (Color online) Illustration of conformal transformation acoustics. Conformal mapping ω(z) transforms the virtual space ABCDEF into the physical space A(cid:48)B(cid:48)C(cid:48)D(cid:48)E(cid:48)F (cid:48), while the inverse mapping z(ω) transforms the physical space back to the virtual space. If regarding the density as homogeneous, then the acoustic parameters can be written as: ρ = kρ0, K = ρc2 = kK0 n2 = kK0detJ, (6a) (6b) with k > 0 as an arbitrary constant able to tune the parameters. Considering the identity dw/dz = (cid:112)1/detJ[35], one can find that Eq.(6) is the same with Norris's results[39]. On the other hand, keeping the physical space with an impedance equaling to that of the background media, one can derive that the bulk modulus and density of the physical space have the properties shown as follows: (cid:114) 1 √ detJ detJ ρ = ρ0c0 c = ρ0n = ρ0 K = ρc2 = K0 n = K0 FIG. 2. (Color online) Profiles of (a) the continuous refrac- tive index and (b) the discrete approximation in the acoustic carpet. (7a) (7b) which are the same with Eq.(6) in Ren et al's work[40]. In this paper, we chose the acoustic parameters shown in Eq.(7) to design the latticed PMC. DESIGN OF THE UNIT CELLS In our case, the designed carpet has the scale of E(cid:48)F (cid:48) = 4.8 dm and D(cid:48)E(cid:48) = 2.0785 dm, while the arc (cid:91)B(cid:48)C(cid:48) is a curve described by y = a cos2(πx/b) with a = 0.3 dm, b = 125/36 dm, and x ∈ [−1.44, 1.44] dm. The back- ground media is water with ρ0 = 1000 Kg/m3, K0 = 2.25 GPa, and nw = n0 = 1. The inverse Jacobian matrix A is obtained by solving the Laplace equation Eq.(3) with the boundaries Eq.(4) in a partial differential equation solver. Since the inverse Laplace's equations are numeri- cally solved, the Cauchy-Riemann relationships shown in Eq.(refeq:cr) are not exactly but approximately satisfied. This causes tiny anisotropy. The anisotropic ratio, de- fined as A11/A22 = A12/A21 here, is as small as 1.058, which can be neglected. Then the inverse mapping w(z) and the mapping z(w) can be confirmed and regarded as nearly conformal. The index profile in the acoustic carpet is shown in Fig.2(a), which presents a petal like distribution. It tends to n = 1, the value of water, near the boundary D(cid:48)E(cid:48), E(cid:48)F (cid:48), F (cid:48)A(cid:48), and reaches at the max- imum of n = 1.35 at the peak of curve (cid:91)B(cid:48)C(cid:48) and the minimum of n = 0.75 near the points B(cid:48) and C(cid:48). The carpet is pre-divided into 300 different cells in order to be realized by PM structure. The discrete indexes are shown as Fig.2(b). There are 15 rows and 20 columns. The index in each cell is the area average of the contin- uous one. Thus, it is easy to obtain the required bulk modulus and density of each cell according to Eq.(7). And in this work, only isotropic PM is in considerations. A basic unit cell of PM is shown in Fig.3(a) which is colored in light yellow. The substrate is aluminum with density ρAl = 2700 kg/m3, Young's modulus EAl = 69 GPa and Poisson's ratio ν = 0.33. The regular hexagon cells behave as isotropic metafluid. The geometric pa- rameters of the basic PM unit cell are characterized by side length l, strut thickness t, and the block size (s, h). xyuvABCDA'B'C'D'E'F'FEw=u+ivz=x+iyw(z)z(w)(a) (b)0.70.80.9 11.11.2nA'B'C'D'E'F' A'B'C'D'E'F'4 FIG. 3. (Color online) (a) Microstructure of the 2D PM, with a basic PM unit cell colored in light yellow and a PM brick colored in lemon. (b) The 15×20 sized PM brick array with corresponding expected effective refractive index shown in Fig.2(b). (c) The deformed PM bricks archived by conformally transforming the regular PM bricks shown in (a). (d) The microstructure of the PMC transformed from the array shown in (b). At the quasi-static regime[41], the effective density is the volume average of the mass. Thus the effective density is mainly determined by the block size (s, h). The ratio of thickness and side length, η = t/l, mainly determines the effective bulk modulus of the PM. The effective pa- rameters of the PM can be calculated by Bloch-Floquet analysis[31, 42], and the required geometric parameters can be obtained by retrieving the energy bands[8, 31]. In order to construct the carpet with the latticed PM structures, the rectangle shaped PM brick is composed by one unit cell with four quarter cells adjoining the four oblique sides, shown as the lemon structure in Fig.3(a). And there are four different PM bricks marked by the red dashed lines in Fig.3(a). Every brick, with the side length l = 8 mm, has the same size with that of the grids in the virtual space shown in Fig.1. All the bricks together pill up a 15 × 20 array (Fig.3(b)), where every PM brick has nearly the same effective index with the corresponding cells in Fig.2(b). The red rectangle in Fig.3(b) marks the four PM bricks shown in Fig.3(a). The PM bricks array has the same size with that of the virtual space. Since the bricks are regularly shaped, some deforma- tions should be implemented in order to form the de- signed carpet cloak. We use the QCT technique again to deform the regular PM bricks into a appropriate shape which can form the needed carpet. The PM brick array is regarded as a new virtual space, where the effective pa- rameters are not the same with the back ground media any more. Since the mapping z(w) connects the virtual space and the physical space, z(w) can transform the ge- ometry of the PM array into a shape similar with the physical space. Note that QCT can be regarded as ap- proximate conformal transformation, thus it can keep the angles and shapes locally unchanged during the trans- formation. The basic assumption in the deformations is that the transformation of the bricks can only change the scales and keep the shape almost unchanged. As what have been mentioned above that the effective parameters of the latticed PM are determined by the shape not the size of the structures, the deformations will not change the effective parameters of the bricks seriously. The formed carpet is shown in Fig.3(d). It can be observed that most of the PM unit cells keep a highly approximate regular hexagon lattice. We suppose that the deformed structure will keep the original effective parameters in the long wave approximation, which is also proved by the simulation results. Since that the bottom boundary (cid:91)B(cid:48)C(cid:48) is not specific, the technique can be used to carpet with arbitrarily shaped bottom boundary[33, 34]. SIMULATION RESULTS The performance of the latticed PMC is simulated us- ing finite element solver COMSOL Multiphysics. The simulated results are shown as Fig.4, where a Gassian beam of frequency 15 Hz is incident at 55◦ from the left up side to the bottom. In Fig.4(a), the bottom side is set as acoustic rigid boundary to imitate the rigid ground plane. It is obvi- ous that all the incident waves are reflected by the rigid plane and the reflected waves keep the plane wave prop- erty. Figure 4(b) presents the acoustic field distribution of the ideal PMC with the index shown in Fig.2(a). The field is almost identical to that of the rigid plane shown in Fig.4(a). Figure 4(c) shows the field distribution of lat- (a)(b)(c)(d)slhtQCT z(w) QCT z(w) 5 FIG. 4. (Color online) Acoustic pressure field of (a) rigid plane (b) ideal PMC (c) rigid scatter and (d) latticed PMC. ticed PMC, whose structure is presented in Fig.3(d). In Fig.4(d), the incident beam is scattered into two distinct beams at different angles by the rigid bulge. It shows that the reflected field is very similar to that of the ideal case in Fig.4(b). Thus, it can be concluded that the designed carpet can cloak the scatter well at 15kHz. FIG. 5. (Color online) The cosine similarity of the reflected waves for ideal PMC (red circles), rigid scatter (blue squares) and latticed PMC (green triangles). Moreover, the cosine similarity (CSI), which can be written as follows[21, 43]: CSI = P rP o P rP o = (cid:112)Σ(P r ΣP r i )2(cid:112)Σ(P o i P o i i )2 , (8) is used to evaluate the performance of the latticed PMC In Eq.(8), P r and P o are the at different frequencies. reference and objective pressure in a measured domain. The reference pressure in this work is the reflected pres- sure of the rigid bulge, while the objective one is that in measure. The measure domain is a 10 dm × 3 dm rectangle that is 6 dm above the bottom side, shown as the dashed line S in Fig.4(a). The value of CSI describes the degree of similarity between the objective field and the reference pressure, i.e., a larger SCI means a higher similarity. Figure 5 shows the SCIs of the ideal PMC, rigid scatter and latticed PMC at the frequency ranging from 10 Hz to 20 Hz. It can been seen that the CSIs of the ideal PMC are all 1, which means the ideal PMC has an identical performance with the rigid plane. The CSIs of the latticed PMC coincide with that of the ideal PMC very well, while the CSIs of the rigid scatter are much smaller than that of the ideal and latticed PMC. Thus, the latticed PMC is as effective as the ideal one at a broadband frequency domain. Furthermore, the case of the incoming beam incident in vertical direction is also studied, and the simulation results are shown in Fig.6. It is obvious that in vertical incident case, the latticed PMC exhibits the same good cloak behavior with the oblique incidence. The almost perfect performance of the ideal PMC in- dicates that the QCT can be used to design acoustic de- vice. Also, the good cloak effect of the latticed PMC shows that the deformation discussed in Sec. does not impact the effective parameters seriously. (a) (b) (c) (d) 1-1Normalized PressureS101214161820Frequency (kHz)0.50.60.70.80.91.0Cosine SimilarityIdeal PMCRigid scatterLatticed PMC6 FIG. 6. (Color online) The scattered acoustic pressure of (a) the rigid plane, (b) the ideal PMC, (c) the rigid scatter and (d) the latticed PMC at 15 kHz. (e) The cosine similarity of the scattered pressure for ideal PMC (red circles), rigid scatter (blue squares) and latticed PMC (green triangles) CONCLUSION We have designed a latticed PM carpet via the QCT, which only needs an isotropic reflective index distribu- tion. The Jacobian transformation matrix, which deter- mines the index distribution, is obtained by numerically solving the Laplace's equations in the inverse mapping. The needed density and bulk modulus are calculated by keeping the impedance equal to that of water. The de- signed carpet are divided into 300 cells in order to realize the latticed PMC by the 2D PM unit cells. The corre- sponding 300 PM bricks are obtained and arranged into an array with the same size with that of the virtual space. We use QCT again to map the geometry of the PM brick array, which is regarded as a new virtual space, into a carpet shaped structure, in which process we assume the transformation will not change the effective parameters of the corresponding PM lattices too much. The CSI in a chosen domain is calculated to archive an accurate anal- ysis for the cloak effect, which shows that both the ideal and latticed PMC have a highly similar behavior with the rigid plane. This confirms the validation of the QCT method on carpet designing and unit cell deformations. We hope the technique used in this work can be used to design relative isotropic acoustic devices. ACKNOWLEDGMENT The authors sincerely acknowledge the financial sup- port of the National Natural Science Foundation of China (Grant No. 11874383), the Youth Innovation Promotion Association CAS (Grant No. 2017029) and the IACAS Young Elite Researcher Project (Grant No. QNYC201719). ∗ Correspongding author:[email protected] † Correspongding author:[email protected] [1] J. B. Pendry. Controlling Electromagnetic Fields. Sci- ence, 312(5781):1780 -- 1782, June 2006. [2] U. Leonhardt. Optical Conformal Mapping. Science, 312(5781):1777 -- 1780, June 2006. [3] S. A. Cummer and D. Schurig. One path to acoustic cloaking. New Journal of Physics, 9(3):45 -- 45, March 2007. [4] H. Chen and C. T. Chan. Acoustic cloaking in three di- mensions using acoustic metamaterials. Applied Physics Letters, 91(18):183518, October 2007. [5] S. Zhang, C. Xia, and N. Fang. Broadband Acoustic Cloak for Ultrasound Waves. Physical Review Letters, 106(2), January 2011. [6] H. Chen, U. Leonhardt, and T. Tyc. Conformal cloak for waves. Physical Review A, 83(5), May 2011. [7] F. Gomory, M. Solovyov, J. Souc, C. Navau, J. Prat- Camps, and A. Sanchez. Experimental Realization of a Magnetic Cloak. Science, 335(6075):1466 -- 1468, March 2012. [8] Y. Chen, X. Liu, and G. Hu. Latticed pentamode acous- tic cloak. Scientific Reports, 5:15745, October 2015. [9] Y. Chen, M. Zheng, X. Liu, Y. Bi, Z. Sun, P. Xiang, J. Yang, and G. Hu. Broadband solid cloak for underwater acoustics. Physical Review B, 95(18), May 2017. [10] J. Li and J. B. Pendry. Hiding under the Carpet: A New Strategy for Cloaking. Physical Review Letters, 101(20), November 2008. [11] R. Liu, C. Ji, J. J. Mock, J. Y. Chin, T. J. Cui, and D. R. Smith. Broadband Ground-Plane Cloak. Science, 323(5912):366 -- 369, January 2009. [12] L. H. Gabrielli, J. Cardenas, C. B. Poitras, and M. Lip- son. Silicon nanostructure cloak operating at optical fre- quencies. Nature Photonics, 3(8):461 -- 463, August 2009. [13] J. Valentine, J. Li, T. Zentgraf, G. Bartal, and X. Zhang. An optical cloak made of dielectrics. Nature Materials, 8(7):568 -- 571, July 2009. [14] J. B. Pendry and J. Li. An acoustic metafluid: realizing a broadband acoustic cloak. New Journal of Physics, 10(11):115032, November 2008. 1-1(a)(b)(c)(d)(e)S[15] W. Zhu, C. Ding, and X. Zhao. A numerical method for designing acoustic cloak with homogeneous metamateri- als. Applied Physics Letters, 97(13):131902, September 2010. [16] B. I. Popa, L. Zigoneanu, and S. A. Cummer. Experi- mental Acoustic Ground Cloak in Air. Physical Review Letters, 106(25), June 2011. [17] W. Hu, Y. Fan, P. Ji, and J. Yang. An experimental acoustic cloak for generating virtual images. Journal of Applied Physics, 113(2):024911, January 2013. [18] L. Zigoneanu, B.-I. Popa, and S. A. Cummer. Three- dimensional broadband omnidirectional acoustic ground cloak. Nature Materials, 13(4):352 -- 355, April 2014. [19] H. Esfahlani, S. Karkar, H. Lissek, and J. R. Mosig. Acoustic carpet cloak based on an ultrathin metasurface. Physical Review B, 94, July 2016. [20] B. I. Popa and S. A. Cummer. Homogeneous and com- pact acoustic ground cloaks. Physical Review B, 83(22), June 2011. [21] J. Zhu, T. Chen, Q. Liang, X. Wang, J. Xiong, and P. Jiang. A unidirectional acoustic cloak for multilayered background media with homogeneous metamaterials. Journal of Physics D: Applied Physics, 48(30):305502, August 2015. [22] Y. Bi, H. Jia, W. Lu, P. Ji, and J. Yang. Design and demonstration of an underwater acoustic carpet cloak. Scientific Reports, 7(1), December 2017. [23] Y. Bi, H. Jia, Z. Sun, Y. Yang, H. Zhao, and J. Yang. Experimental demonstration of three-dimensional broad- band underwater acoustic carpet cloak. Applied Physics Letters, 112(22):223502, May 2018. [24] G. W. Milton and A. V. Cherkaev. Which Elasticity Ten- sors are Realizable? Journal of Engineering Materials and Technology, 117(4):483, 1995. [25] A. N. Norris. Acoustic cloaking theory. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 464(2097):2411 -- 2434, September 2008. [26] M. Kadic, T. Bckmann, N. Stenger, M. Thiel, and M. Wegener. On the practicability of pentamode mechanical metamaterials. Applied Physics Letters, 100(19):191901, May 2012. [27] C. N. Layman, C. J. Naify, T. P. Martin, D. C. Calvo, and G. J. Orris. Highly Anisotropic Elements for Acous- tic Pentamode Applications. Physical Review Letters, 111(2), July 2013. [28] A. N. Norris. Acoustic metafluids. The Journal of the Acoustical Society of America, 125(2):839 -- 849, February 2009. [29] Y. Tian, Q. Wei, Y. Cheng, Z. Xu, and X. Liu. Broad- band manipulation of acoustic wavefronts by pentamode metasurface. Applied Physics Letters, 107(22):221906, November 2015. 7 [30] X. Su, A. N. Norris, C. W. Cushing, M. R. Haberman, and P. S. Wilson. Broadband focusing of underwater sound using a transparent pentamode lens. The Journal of the Acoustical Society of America, 141(6):4408 -- 4417, June 2017. [31] Z. Sun, H. Jia, Y. Chen, Z. Wang, and J. Yang. De- sign of an underwater acoustic bend by pentamode metafluid. The Journal of the Acoustical Society of Amer- ica, 143(2):1029 -- 1034, February 2018. [32] Y. Chen, X. Liu, and G. Hu. Design of arbitrary shaped pentamode acoustic cloak based on quasi-symmetric mapping gradient algorithm. The Journal of the Acousti- cal Society of America, 140(5):EL405 -- EL409, November 2016. [33] J. Hu, X. Zhou, and G. Hu. Design method for electro- magnetic cloak with arbitrary shapes based on Laplaces equation. Optics Express, 17(3):1308, February 2009. [34] Z. Chang, X. Zhou, J. Hu, and G. Hu. Design method for quasi-isotropic transformation materials based on in- verse Laplaces equation with sliding boundaries. Optics Express, 18(6):6089, March 2010. [35] J. P. Turpin, A. T. Massoud, Z. H. Jiang, P. L. Werner, and D. H. Werner. Conformal mappings to achieve simple material parameters for transformation optics devices. Optics Express, 18(1):244, January 2010. [36] L. Xu and H. Chen. Conformal transformation optics. Nature Photonics, 9(1):15 -- 23, December 2014. [37] L. Xu, H. Chen, T. Tyc, Y. Xie, and S. A. Cummer. Perfect conformal invisible device with feasible refractive indexes. Physical Review B, 93(4), January 2016. [38] C.-Y. Ren, Z. Xiang, and Z. Cen. Acoustic carpet in- visibility cloak with two open windows using multilay- ered homogeneous isotropic material. Chinese Physics B, 20(11):114301, November 2011. [39] A. N. Norris. Comment on Design of acoustic devices with isotropic material via conformal transformation [Appl. Phys. Lett. 97, 044101 (2010)]. Applied Physics Letters, 100(6):066101, February 2012. [40] C. Ren, Z. Xiang, and Z. Cen. Design of acoustic de- vices with isotropic material via conformal transforma- tion. Applied Physics Letters, 97(4):044101, July 2010. [41] A. N. Norris. Mechanics of elastic networks. In Proc. R. Soc. A, volume 470, page 20140522. The Royal Society, 2014. [42] A.A. Kutsenko. Wave propagation and homogenization in 2d and 3d lattices: A semi-analytical approach. Quar- terly Journal Of Mechanics And Applied Mathematics, 2017. [43] J. Xiong, T. Chen, X. Wang, and J. Zhu. Design and assessment of an acoustic ground cloak with layered structure. International Journal of Modern Physics B, 29(27):1550191, October 2015.
1808.02107	1	1808	2018-06-18T06:23:19	Significantly enhanced photocatalytic degradation of Methylene Blue using rGO-SnO2 nanocomposite under natural sunlight and UV light irradiation	[ "physics.app-ph", "cond-mat.mtrl-sci", "physics.chem-ph" ]	In this paper, we report the synthesis of reduced Graphene Oxide-Tin Oxide nanocomposite and the effectiveness of this composite in decolorizing and degrading Methylene Blue (MB). Tin Oxide was prepared by liquid phase co-precipitation method and reduced Graphene Oxide-Tin Oxide (rGO-SnO2) nanocomposite was prepared by solution mixing method. Tin Oxide nanoparticles (NPs) have been ardently investigated as photocatalyst for water purification and environment decontamination but the photon generated electron and hole pair (EHP) recombination is one of the limiting factors. Reduced Graphene Oxide-Tin Oxide (rGO-SnO2) nanocomposite is very efficient to overcome this limitation for photocatalytic application. The as-synthesized GO, SnO2, GO-SnO2, rGO and rGO-SnO2 nanocomposite were characterized by X-ray Diffraction, Scanning Electron Microscopy, Energy Dispersive X-ray spectroscopy and Fourier Transform Infrared spectroscopy. The XRD data confirms the sharp peak at 2Theta=10.44 degree corresponding to (002) reflection of GO with interlayer d spacing of 8.46 Angstrom indication of successful preparation of GO by oxidation of graphite. Moreover, the diffraction peak shifts from 2Theta=10.44 degree to 2Theta=23.31 degree confirm successful synthesis of rGO as well. SEM image shows the morphology of all the photocatalysts. EDX studies are carried out to investigate the elemental composition and purity of the sample by giving all the elements present in the nanomaterials. The presence of functional groups was identified by FTIR. The rGO-SnO2 (1:10) nanocomposite shows an efficient photodegradation efficiency of ~94% and ~95% under natural sunlight and UV light irradiation respectively for Methylene Blue (MB) within 15 minutes. Furthermore, the degradation kinetics of MB is also studied in this paper as well.	physics.app-ph	physics	MANUSCRIPT Significantly enhanced photocatalytic degradation of Methylene Blue using rGO-SnO2 nanocomposite under natural sunlight and UV light irradiation M. Bakhtiar Azima,e, Muhammad Hasanuzzamanb, Md. Shakhawat Hossain Firozc, Md. Abdul Gafurd, A.S.W Kurnya, Fahmida Gulshana aDepartment of Materials and Metallurgical Engineering, Faculty of Engineering, Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh. aProfessor [retired], Department of Materials and Metallurgical Engineering, Faculty of Engineering, Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh. aProfessor, Department of Materials and Metallurgical Engineering, Faculty of Engineering, Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh. bAssistant Professor, Department of Glass and Ceramics Engineering, Faculty of Engineering, Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh. cAssociate Professor, Department of Chemistry, Faculty of Engineering, Bangladesh University of Engineering and Technology, Dhaka-1000, Bangladesh. dPilot Plant and Process Development Center, Bangladesh Council of Scientific and Industrial Research (PP and PDC, BCSIR), Dhaka-1205, Bangladesh. eSchool of Engineering Science, Faculty of Applied Science, Simon Fraser University, Burnaby, British Columbia, Canada. Dr. Fahmida Gulshan, E-mail: [email protected] M. Bakhtiar Azim, E-mail: [email protected]; [email protected] ABSTRACT Environmental contamination and human exposure to dyes have dramatically increased over the past decades because of their increasing use in such industries as textiles, paper, plastics, tannery and paints. These dyes can cause deterioration in water quality by imparting color to the water and inducing the photosynthetic activity of aquatic organisms by hindering light penetration. Moreover, some of the dyes are considered carcinogenic and mutagenic for human health. Therefore, efficient treatment and removal of dyes from wastewater have attracted considerable attention in recent years. Photocatalysis, due to its mild reaction condition, high degradation, broad applied area and facile manipulation, is a promising method of solving environmental pollution problems. In this paper, we report the synthesis of reduced Graphene Oxide-Tin Oxide (rGO-SnO2) nanocomposite and the effectiveness of this composite in decolorizing and degrading Methylene Blue (MB). Tin Oxide was prepared by liquid phase co-precipitation method and reduced Graphene Oxide-Tin Oxide (rGO-SnO2) nanocomposite was prepared by solution mixing method. Tin Oxide (SnO2) nanoparticles (NPs) have been ardently investigated as photocatalyst for water purification and environment decontamination but the photon generated electron and hole pair (EHP) recombination is one of the limiting factors. Reduced Graphene Oxide-Tin Oxide (rGO-SnO2) nanocomposite is very efficient to overcome this limitation for photocatalytic application. The as-synthesized GO, SnO2, GO-SnO2, rGO and rGO-SnO2 nanocomposite were characterized by X-ray Diffraction (XRD), Scanning Electron Microscopy (SEM), Energy Dispersive X-ray spectroscopy (EDX) and Fourier Transform Infrared spectroscopy (FTIR). The XRD data confirms the sharp peak at 2=10.44 corresponding to (002) reflection of GO with interlayer d- spacing of 8.46 A indication of successful preparation of GO by oxidation of graphite. Moreover, 1 the diffraction peak shifts from 2=10.44 to 2=23.31 confirm successful synthesis of rGO as well. SEM image shows the morphology of GO, SnO2, GO-SnO2 and rGO-SnO2. EDX studies are carried out to investigate the elemental composition and purity of the sample by giving all the elements present in the GO, SnO2, GO-SnO2 and rGO-SnO2. The presence of functional groups was identified by FTIR. The rGO-SnO2 (1:10) nanocomposite shows an efficient photodegradation efficiency of ~94% and ~95% under natural sunlight and UV light irradiation respectively for Methylene Blue (MB) within 15 minutes. Furthermore, the degradation kinetics of MB is also studied in this paper as well. Keywords: reduced Graphene Oxide, Tin Oxide, Kinetics, Photodegradation, Methylene Blue. 1. INTRODUCTION With the rapid development of textile industry in recent years, more and more new types of dyes, such as Methylene Blue, have been produced. According to World Bank report, almost 20% of global industrial water pollution comes from the dyeing and finishing processes of textiles [1]. The discharge of azo dyes, which are stable and carcinogenic, into water bodies are harmful to human health, and cause such illness as cholera, diarrhea, hypertension, precordial pain, dizziness, fever, nausea, vomiting, abdominal pain, bladder irritation, staining of skin etc. [2]. Dyes also affect aquatic life by hindering the photosynthesis process of aquatic plants, eutrophication, and perturbation [3,4]. Numerous techniques, such as activated carbon adsorption (physical method), chlorination (chemical method), and aerobic biodegradation (biochemical method) [4] have been applied to treat textile wastewater. However, further treatments are needed, which create such secondary pollution in the environment, as the breakdown of parent cationic dyes to Benzene, NO2, CO2 and SO2 [6]. Advanced oxidation 2 processes (AOPs) are widely applied to mineralize dyes into CO2 and H2O [7, 8]. AOPs include ozonation, photolysis, and photocatalysis with the aid of oxidants, light, and semiconductors. Photocatalytic degradation is initiated when the photocatalysts absorb photons (UV) to generate electron-hole pairs on the catalyst surface. The positive hole in the valence band (hVB +) will react with water to form hydroxyl radical (•OH), followed by the oxidization of pollutants to CO2 and H2O [9]. Methylene Blue (MB), also known as Basic Blue 9, is a cationic azo dye (Table 1). MB is widely used in textile industries for dye processing, and upto 50% of the dyes consumed in textile industries are azo dyes [9-11]. In the past few years, several catalysts such as TiO2 [6], BiFeO3 [5], ZnS [13] and ZnO [9] have been used to degrade MB and the results are summarized in (Table 2). [Insert Table 1 and Table 2] Carbon based materials combined with metal oxides are known to have improved photocatalytic activity. Graphene, a two-dimensional material having sp2 bonded carbon atoms arranged in a honeycomb lattice with a one-atom-thick, has recently attracted a great deal of scientific attention among the researchers. Owing to its extraordinary advantages, such as large theoretical specific surface area (2630 m2/g), superior electronic and excellent chemical stability, therefore graphene is considered as an outstanding support for photocatalytic application. It was first isolated from 3D graphite by mechanical exfoliation. It has also been reported that the graphene-metal oxide composite possess good photocatalytic activity, compared to the pure metal oxide. 3 Numerous chemical, thermal, microwave and microbial/bacterial methods have been used in the synthesis of rGO [15]. Chemical exfoliation is preferable due to its large-scale production and low cost. Chemical exfoliation involves four steps, oxidation of graphite powder, dispersion of graphite oxide (GTO) to graphene oxide (GO), GTO exfoliation by ultrasonication to produce graphene oxide (GO) and finally, reduction of GO to rGO using a reducing agent [16]. Reduced Graphene oxide (GO) has less oxygen functional groups than graphene oxide (GO). rGO has a surface area of 833 m2/g compared to 736.6 m2/g and 400 m2/g for GO and graphite [30-31]. rGO, with its unique electronic properties, large surface area and high transparency, contributes to facile charge separation and adsorptivity in its structure. As a potential photocatalytic material, rGO-SnO2 has been used in the decolorization of Methylene Blue [20] and Rhodamine B [20]. SnO2 is a capable candidate to photocatalyze and complete oxidative mineralization of MB. Chemical contamination of water streams has become crucial issue of human life. Wastewater treatment plays an important role in reducing the toxic elements in wastewater. Heterogeneous photocatalysis can be applied to remove contaminants existing in wastewater effluent. Heterogeneous photocatalysis includes such reactions as organic synthesis, water splitting, photo-reduction, hydrogen transfer and metal deposition, disinfection, water treatment, removal of gaseous pollutants etc. It has become an increasingly viable technology in environmental decontamination. Photocatalytic oxidation of such organic compounds are derived by semiconductor materials like TiO2, ZnO, CdS and CuO. These are now-a-days widely used in the environment as photocatalysts. They have band-gap of 3.6 eV, when other photocatalysts have higher band-gap. They are used extensively due to their low-cost, non- 4 toxicity, high activity, large chemical stability, very low aqueous solubility and environmental friendly characteristics. Tin-assisted photocatalytic oxidation is an alternative method for purification of air and water streams. Also, in water splitting, the driving force for electrons is provided by energy of light. When SnO2 is exposed to light, photocatalytic reaction is initiated. rGO-SnO2 nanocomposite is very promising to overcome the limitation for photocatalytic application. rGO with its large surface area, unique electronic properties and high transparency, contributes to spatial charge separation and adsorptivity in this hybrid structure. In this investigation, we report a facile method to prepare rGO-SnO2 nanocomposite. It was synthesized via solution mixing method. The photocatalytic performances of the prepared GO, SnO2, GO-SnO2, rGO and rGO-SnO2 nanomaterials were evaluated in the degradation of Methylene Blue (MB) under natural sunlight and UV light irradiation. 2. EXPERIMENTAL SECTION 2.1. Chemicals and Materials Graphite fine powder (~325 mesh) was purchased from Alfa Aesar. Sodium Nitrate, Tin (II) Chloride Di-hydrate (98%), Potassium permanganate (99%), Hydrochloric acid (37%), Hydrogen Peroxide (30%), Silver Nitrate (99.8%), Ammonia solution (25%), Urea and Hydrazine Hydrate (99%) were purchased from Sigma-Aldrich (Steinheim, Germany). Sulfuric acid (98%) was obtained from Merck (Darmstadt, Germany). The chemicals were used without further purifications. Methylene Blue (MB) powder from Merck (Darmstadt, Germany) was used as the 5 model organic dye in this study. Deionized water (DI water) was used throughout the experiments. 2.2. Synthesis of GO Graphene oxide was produced through the modified Hummers' method [35] by oxidizing the graphite powder. In a typical synthesis, 5 gm of graphite powder and 2.5 gm NaNO3 were mixed with 115 ml H2SO4 (conc. 98%). 15 gm of KMnO4 was slowly added and stirred in an ice-bath for 1 h below 20C. The mixture was heated to 35C and was constantly stirred for 2 hours. The beaker was placed on an oil bath, heated to and maintained at a temperature 95C~ 98C, for 15 minutes. 250 ml DI water was added slowly under constant stirring. The mixture was cooled to room temperature. The beaker was then placed on the oil bath for additional 60 minutes at a constant temperature 60C. 150 ml DI water was added under constant stirring. Finally, 50 ml (30%) H2O2 was added in drops and stirred for 2 hours. Washing, filtration and centrifugation (6000 rpm) were done until the removal of Cl- ions by using DI water. Finally, the resulting precipitate was dried at 70C for 24 hours in an oven giving thin sheets which was Graphite Oxide (GTO). Graphite Oxide was made into fine powder form by grinding and then GTO powder was finely dispersed in DI Water. At last, ultrasonication was done for the complete exfoliation of GTO to GO. 2.3. Synthesis of SnO2 SnO2 was produced through the liquid phase co-precipitation method. 2 gm Stannous Chloride Di-hydrate (SnCl2.2H2O) was dissolved in 100 ml DI Water. After complete dissolution, ammonia solution (25%) was added in drops to the above solution under stirring. The resulting gel type 6 precipitate was filtered and dried at 80C for 24 hours to remove water molecules. Finally, tin oxide nanopowders were formed through calcination at 550C for 4-6 hours. 2.4. Synthesis of GO-SnO2 GO-SnO2 was prepared through the solution mixing method. At first, 2.6 gm SnCl2.2H2O was added to HCl (37%) and stirred for 1 hour. Then, 260 mg GO was dispersed in 200 ml DI water by using ultrasonication for 45 minutes. Both solutions were mixed and stirred rigorously for 15 minutes. Again ultrasonication was done for 15 minutes. After that, washing, filtration and centrifugation were done with DI water till a neutral pH was obtained. The sediment was collected and dried at 80C for 24 hours. Finally, thin sheet of GO-SnO2 nanocomposite was collected and grinded into GO-SnO2 nano powder. 2.5. Synthesis of rGO 260 mg GO was dispersed in DI water and then exfoliated by ultrasonication for 1 hour. Subsequently, 5 ml of 99% Hydrazine Hydrate was added and the solution was heated in heating mantle with stirrer at 100C under a water cooled condenser for 24 hours. After heating at 100C, the solution turns from brown to black precipitate. The product was cooled, filtered and washed with DI water for 4 times. Finally, the product was dried overnight at 60C. 2.6. Synthesis of rGO-SnO2 rGO-SnO2 was prepared also through the solution mixing method. At first, 2.6 gm SnCl2.2H2O was added to HCl (37%) and diluted with DI water to give SnCl2-HCl solution. Then, 260 mg GO was dispersed in DI water by using ultrasonication for 30 minutes to form a colloidal suspension. The GO solution was added to a SnCl2-HCl solution and ultrasonication was done for 15 minutes. 2.6 gm Urea was then added to the resulting solution and stirred for 15 minutes. The solution was 7 heated to 95C for 6 hours. The precipitate obtained was separated by centrifugation (6000 rpm) and subsequent washing to remove excess Cl- ions. The resulting product was dried in an oven at 110C for 6 hours. 3. CHARACTERIZATION The X-ray diffraction pattern of GO, SnO2, GO-SnO2, rGO and rGO-SnO2 was recorded by a Bruker, D8 Advance diffractometer (Germany). The sample was scanned from 5 to 80 using Cu K radiation source ( = 1.5406 A) at 40 kV and 30 mA with a scanning speed of 0.01s-1. The surface morphology of GO, SnO2, GO-SnO2 and rGO-SnO2 was observed by FESEM-JEOL (FEG-XL 30S) Field Emission Scanning Electron microscope (FESEM). FTIR spectra of GO, SnO2, and rGO-SnO2 was recorded by Agilent Cary 670 FTIR spectrometer. The photodegradation percentage of MB was determined by using an ultraviolet-visible spectrophotometer (Shimadzu-UV-1601) at max = 664 nm and wavelength region between 400 and 800 nm. DI water was used as a reference material. 4. PHOTOCATALYTIC REACTION Photocatalytic experiments were carried out by photodegrading MB using Shimadzu-UV-1601 UV-Vis spectroscopy. The solution of MB (pH~7) without rGO-SnO2 was left in a dark place for 24 hrs. Then, the dye solution was exposed to natural sunlight irradiation and UV irradiation and there was no decrease in the concentration of dye. In a typical experiment, 7.5 mg of rGO-SnO2 was added into a 50 mL 0.05 mM MB solution. Before illumination, the suspensions were continuously shacked and kept in dark for 60 min to reach an adsorption-desorption equilibrium between the photocatalyst and dye solution. The suspensions were then exposed to natural sunlight and UV irradiation. Samples were taken from irradiation at regular time intervals (0 min, 8 10 min, 15 min, 30 min, 45 min and 60 min for MB) filtered out to remove the rGO-SnO2. Irradiation was carried out in falcon tubes. The same procedure was repeated for both GO, SnO2, rGO and GO-SnO2. The initial pH of solution (pH~7) was adjusted by small amount of 0.1 M NaOH and 0.1 M HCl when required. Photodegradation was also observed for 0.05 mM 50 ml MB solution using only GO, SnO2, GO- SnO2 and rGO. The photodegradation efficiency of MB was determined by using the equation shown below: Photodegradation efficiency (%) = [(C0 - Ct) / C0]  100% = [( A0 - At) / A0]  100% ---------------(1) [According to 'Beer-Lambert Law'] Where, C0 is the initial concentration of MB, Ct is the concentration of MB at time, t and A0 is the initial absorbance of MB, At is the absorbance of at time, t. 5. RESULTS AND DISCUSSION 5.1. Characterization of Photocatalysts XRD The powder X-ray diffraction pattern of GO shows a broadened diffraction peak (Fig. 1) at around 2=10.44, which corresponds to the (002) reflection of stacked GO sheets with interlayer d- spacing of 8.46. rGO shows diffraction peak at around 2=23.31 with interlayer d-spacing of 7.62 A. XRD patterns of SnO2 nanoparticles (Fig. 1) shows the diffraction peaks of (110), (101), 9 (111) and (211) at 2 of 26.8, 33.9, 37.9 and 51.8 respectively which matches well with JCPDS card # 41-1445. The average crystallite size can be determined using "Scherrer Formula": B = (𝒌)/(𝒄𝒐𝒔)----------------------------------------------------(2) Where, B is the average crystallite size (nm), k is the factor of 0.9,  is the wavelength of radiation source used is 1.5406 A,  is the full width of half maximum peak (FWHM),  is the angle at maximum peak. The average crystallite size of SnO2 is ~35 nm (eq. 2). The XRD pattern of rGO- SnO2 nanocomposite (Fig. 1) shows SnO2 peaks but the intensity of the peaks are reduced and broaden when compared with XRD pattern of individual rGO and SnO2. [Insert Fig. 1] SEM and EDX SEM images of rGO-SnO2 structure are shown in ((Fig. 2(e)). The average particle size of SnO2 is ~46 nm using ImageJ analysis ((Fig. 2(b)). SEM images of rGO-SnO2 shows presence of SnO2 nanoparticles in the rGO layer by layer sheets. [Insert Fig. 2] EDX results indicating successful incorporation of oxygen by modified Hummers method in graphite layers and formation of oxygen based functional groups in GO with ratio of C/O=4.58 ((Fig. 3(a)). It also shows successful synthesis of SnO2 nanoparticles ((Fig. 3(b)). It gives clear indication of incorporation of SnO2 nanoparticles in rGO layer by layer sheets ((Fig. 3(d)). [Insert Fig. 3] 10 FTIR The FTIR spectra of GO, SnO2 and rGO-SnO2 was recorded by Agilent Cary 670 FTIR spectrometer. FTIR analysis of GO shows broad absorption spectrum observed at ~3420 cm-1 corresponding O- H stretching vibration indicating existence of absorbed water molecules and structural O-H groups in GO. The broad peak appeared in GO spectrum depicted the presence of O-H and C-H stretching. Besides, a band at 1747 cm-1 might be related to not only the C=O stretching motion of -COOH groups situated at the edges and defects of GO lamellae but also that of ketone or quinone groups. The peak near 1700-1550 cm-1 widens and moves to 1565 cm-1 that reflects the presence of un-oxidized aromatic regions. The FTIR spectrum of Tin Oxide shows broad not sharp absorption band ~3420 cm-1 corresponding O-H stretching vibration due to absorbed water molecules. A band at ~612 cm-1 related to O-Sn-O stretching. Broad and sharp absorption spectrum observed at ~3420 cm-1 corresponding O-H stretching vibration in rGO-SnO2. Peak at 1625cm-1 corresponding O-H bending vibration indicating existence of absorbed water molecules and structural O-H groups in rGO-SnO2 by FTIR. Similarly, a sharp band at ~612 cm-1 related to O- Sn-O stretching in rGO-SnO2 refers to the incorporation of SnO2 particles. (Fig. 4) [Insert Fig. 4] 5.2. Photocatalytic Activity 5.2.1. Conditions and Parameters for Photocatalytic Activity Measurement The concentration and amount of the dye used for the experiment were 0.05 mM and 50 ml respectively. Weight of the samples used for the experiment was 7.5 mg each. For the experiment 11 pH was considered as ~7 and temperature outside was 29 C - 33 C. The most crucial parameters of the experiment were (i) Natural Sunlight Irradiation, (ii) UV light Irradiation, (iii) Irradiation Time Effect and (iv) Catalyst Types. 5.2.2. Photocatalytic Activity Evaluation UV-Vis Spectroscopy was used to measure absorbance of the dye solution at regular time intervals using rGO-SnO2. Controlled experiments were also carried out to confirm that the degradation of MB by UV-Vis for visible range. Experiments were repeated for other photocatalysts also. Photodegradation Efficiency (%) = [( C0 - Ct) / C0]  100%= [( A0 - At) / A0]  100%-------------(3) Here, C0= Concentration of MB at '0' min, Ct= Concentration of MB at 't' min, A0= Absorbance of MB at '0' min, At= Absorbance of MB at 't' min Time-dependent absorption spectra of MB solution shows enhanced photocatalytic activity of rGO-SnO2 compared to SnO2 and rGO under natural sunlight and UV irradiation (Fig. 5 and Fig 7). [Insert Fig. 5 and Fig. 6] [Insert Fig. 7 and Fig. 8] It is clearly seen from the ((Fig. 6(a) and Fig. 8(a)) that rGO-SnO2 having lesser absorbance compared to GO, SnO2, GO-SnO2 and rGO. From ((Fig. 6(b) and Fig. 8(b)) with irradiation time concentration of MB solution reduces significantly by using rGO-SnO2 nanocomposite with 12 respect to others under natural sunlight and UV light irradiation. Under natural sunlight and UV light irradiation rGO-SnO2 nanocomposite showed ~94% and ~95% photodegradation efficiency respectively within 15 minutes compared to other photocatalysts ((Fig. 9(a) and Fig. 9(b)). The most outstanding factor of this experiment is under UV light irradiation the degradation efficiency of GO-SnO2 has increased remarkably from ~61% to 94% after 30 minutes ((Fig. 9(a) and Fig. 9(b)). [Insert Fig. 9] 6. ADSORPTION KINETICS 6.1. Pseudo First Order The MB photo degradation was fitted to pseudo-first order kinetics by referring to the Langmuir- Hinshelwood kinetic model : 𝐥𝐧 ( 𝑪𝟎 𝑪𝒕 ) = 𝒌𝟏𝒕---------------------------------------------------------------------------(4) Where, Ct = Concentration of MB at time, t, C0 = Initial concentration of MB, k1 = Pseudo-first order rate constant. The k1 values of respective concentrations was determined from knowing the values Ct and was listed in Table 3. 6.2. Pseudo Second Order Pseudo second-order equilibrium adsorption equation can be expressed as follows: 𝟏 𝑪𝒕 = 𝟏 𝑪𝟎 + 𝟐𝒌𝟐𝒕-----------------------------------------------------------------------------(5) 13 Where, Ct = Concentration of MB at time t, C0 = Initial concentration of MB, k2 = Pseudo-second order rate constant. The k2 values of respective concentrations was determined from knowing the values Ct and was listed in Table 4. 6.3. Adsorption Kinetics Evaluation The correlation coefficient (R2) values are close to 1 observed in Table 3 than Table 4. Therefore, it gives clear indication that rGO-SnO2 obeys the pseudo-first order kinetic model under natural sunlight irradiation. [Insert Table 3 and Table 4] [Insert Fig. 10] The correlation coefficient (R2) values are close to 1 observed in Table 5 than Table 6. Therefore, it gives clear indication that rGO-SnO2 obeys the pseudo-first order kinetic model under UV light irradiation. [Insert Table 5 and Table 6] [Insert Fig. 11] 7. CONCLUSIONS Results of characterization showed that GO, SnO2, GO-SnO2, rGO and rGO-SnO2 were synthesized successfully. Degradation of 0.05 mM 50 ml Methylene Blue under direct sunlight and UV light 14 irradiation with 7.5 mg rGO-SnO2 nanocomposite as a photocatalyst takes around 15 min for ~94% and ~95% degradation respectively compared to other catalysts. The most significant result of this experiment is under UV light degradation efficiency of GO-SnO2 enhanced remarkably compared to other photocatalysts within 30 minutes (from ~61% to ~94%). Wide band gap of SnO2 is tuned using rGO therefore it shows enhanced photocatalytic performance compared to SnO2 only. Moreover, the degradation kinetics is also studied and rGO-SnO2 maintains the pseudo-first order kinetic model. CONFLICTS OF INTERESTS There are no conflicts to declare. ACKNOWLEDGEMENTS This work was supported by Department of Materials and Metallurgical Engineering (MME, BUET), The Pilot Plant and Process Development Center, Bangladesh Council of Scientific and Industrial Research (PP and PDC, BCSIR), Department of Glass and Ceramics Engineering (GCE, BUET) and Department of Chemistry, BUET. REFERENCES: 1. APHA, 2005. Standard methods for the examination of water and waste water, 21 ed. American Water Works Association and Water Environment Federation. 2. Lee, K.M.; Abdul Hamid, S.B.; Lai, C.W. "Multivariate analysis of photocatalytic-mineralization of Eriochrome Black T dye using ZnO catalyst and UV irradiation," Mater. Sci. Semicond. Process. 2015, 39, 40 -- 48. 15 3. Mehra, M.; Sharma, T.R. "Photocatalytic degradation of two commercial dyes in aqueous phase using photocatalyst TiO2," Adv. Appl. Sci. Res. 2012, 3, 849 -- 853. 4. Vinothkannan, M.; Karthikeyan, C.; Gnana kumar, G.; Kim, A.R.; Yoo, D.J. "One-pot green synthesis of reduced graphene oxide (RGO)/Fe3O4 nanocomposites and its catalytic activity toward methylene blue dye degradation," Spectrochim. Acta A Mol. Biomol. Spectrosc. 2015, 136, 256 -- 264. 5. Soltani, T.; Entezari, M.H. "Photolysis and photocatalysis of methylene blue by ferrite bismuth nanoparticles under sunlight irradiation," J. Mole. Cat. 2013, 377, 197-203. 6. Dariani, R.S.; Esmaeili, A.; Mortezaali, A.; Dehghanpour, S. "Photocatalytic reaction and degradation of methylene blue on TiO2 nano-sized particles," Optik. 2016, 127, 7143-7154. 7. Chan, S.H.S.; Wu, Y.T.; Juan, J.C.; Teh, C.Y. "Recent developments of metal oxide semiconductors as photocatalysts in advanced oxidation processes (AOPs) for treatment of dye waste-water," J. Chem. Technol. Biotechnol. 2011, 86, 1130 -- 1158. 8. Zhang, W.; Naidu, B.S.; Ou, J.Z.; O'Mullane, A.P.; Chrimes, A.F.; Carey, B.J.; Wang, Y.; Tang, S.Y.; Sivan, V.; Mitchell, A.; et al. "Liquid metal/metal oxide frameworks with incorporated Ga2O3 for photocatalysis," ACS Appl. Mater. Interfaces 2015, 7, 1943 -- 1948. 9. Kansal, S.; Kaur, N.; Singh, S. "Photocatalytic degradation of two commercial reactive dyes in aqueous phase using nanophotocatalysts," Nanoscale Res. Lett., 2009, 4, 709 -- 716. 10. Ong,W.J.; Yeong, J.J.; Tan, L.L.; Goh, B.T.; Yong, S.T.; Chai, S.P. "Synergistic effect of graphene as a co-catalyst for enhanced daylight-induced photocatalytic activity of Zn0.5Cd0.5S synthesized 16 via an improved one-pot co-precipitation-hydrothermal strategy," R. Soc. Chem. Adv., 2014, 4, 59676 -- 59685. 11. Sahel, K.; Perol, N.; Chermette, H.; Bordes, C.; Derriche, Z.; Guillard, C. "Photocatalytic decolourization of remazol black 5 (RB5) and procion red MX-5B-Isotherm of adsorption, kinetic of decolourization and mineralization," Appl. Catal. B Environ., 2007, 77, 100 -- 109. 12. Lucas, M.S.; Tavares, P.B.; Peres, J.A.; Faria, J.L.; Rocha, M.; Pereira, C.; Freire, C. "Photocatalytic degradation of reactive black 5 with TiO2-coated magnetic nanoparticles," Catal. Today, 2013, 209, 116 -- 121. 13. Goharshadi, E.K.; Hadadian, M.; Karimi, M.; Azizi-Toupkanloo, H. "Photocatalytic degradation of reactive black 5 azo dye by zinc sulfide quantum dots prepared by a sonochemical method," Mater. Sci. Semicond. Process., 2013, 16, 1109 -- 1116. 14. Laohaprapanon, S.; Matahum, J.; Tayo, L.; You, S.J. "Photodegradation of reactive black 5 in a ZnO/UV slurry membrane reactor," J. Taiwan Inst. Chem. Eng., 2015, 49, 136 -- 141. 15. Bianco, A.; Cheng, H.M.; Enoki, T.; Gogotsi, Y.; Hurt, R.H.; Koratkar, N.; Kyotani, T.; Monthioux, M.; Park, C.R.; Tascon, J.M.D.; et al. "All in the graphene family-A recommended nomenclature for two-dimensional carbon materials," Carbon, 2013, 65, 1 -- 6. 16. Cao, N.; Zhang, Y. "Study of reduced graphene oxide preparation by Hummers' method and related characterization," J. Nanomater., 2015. 17. Singh, V.; Joung, D.; Zhai, L.; Das, S.; Khondaker, S.I.; Seal, S. "Graphene based materials: Past, present and future," Prog. Mater. Sci., 2011, 56, 1178 -- 1271. 17 18. Wang, D.T.; Li, X.; Chen, J.F.; Tao, X. "Enhanced photoelectrocatalytic activity of reduced graphene oxide/TiO2 composite films for dye degradation," Chem. Eng. J. 2012, 198, 547 -- 554. 19. Dong, S.Y.; Cui, Y.R.; Wang, Y.F.; Li, Y.K.; Hu, L.M.; Sun, J.Y.; Sun, J.H. "Designing three- dimensional acicular sheaf shaped BiVO4/reduced graphene oxide composites for efficient sunlight-driven photocatalytic degradation of dye wastewater," Chem. Eng. J., 2014, 249, 102 -- 110. 20. Liu, X.J.; Pan, L.K.; Lv, T.; Sun, Z.; Sun, C.Q. "Visible light photocatalytic degradation of dyes by bismuth oxide-reduced graphene oxide composites prepared via microwaved-assisted method," J. Colloid Interface Sci. 2013, 408, 145 -- 150. 21. Yao, Y.J.; Xu, C.; Miao, S.D.; Sun, H.Q.; Wang, S.B. "One-pot hydrothermal synthesis of Co(OH)2 nanoflakes on graphene sheets and their fast catalytic oxidation of phenol in liquid phase," J. Colloid Interface Sci., 2013, 40, 230 -- 236. 22. Bose, S.; Kuila, T.; Uddin, M.E.; Kim, N.H.; Lau, A.K.T.; Lee, J.H. "In-situ synthesis and characterization of electrically conductive polypyrrole/graphene nanocomposites," Polymer, 2010, 51, 5921 -- 5928. 23. Krishnamoorthy, K.; Kim, G.S.; Kim, S.J. "Graphene nanosheets: Ultrasound assisted synthesis and characterization," Ultrason. Sonochem., 2013, 20, 644 -- 649. 24. Satheesh, K.; Jayavel, R. "Synthesis and electrochemical properties of reduced graphene oxide via chemical reduction using thiourea as a reducing agent," Mater. Lett., 2013, 113, 5 -- 8. 18 25. Zainy, M.; Huang, N.M.; Kumar, S.V.; Lim, H.N.; Chia, C.H.; Harrison, L. "Simple and scalable preparation of reduced graphene oxide-silver nanocomposites via rapid thermal treatment," Mater. Lett., 2012, 89, 180 -- 183. 26. Stankovich, S.; Dikin, D.A.; Piner, R.D.; Kohlhaas, K.A.; Alfred, K.; Jia, Y.Y.;Wu, Y.; Nguyen, S.T.; Ruoff, R.S. "Synthesis of graphene-based nanosheets via chemical reduction of exfoliated graphite oxide," Carbon, 2007, 45, 1558 -- 565. 27. Apollo, S.; Moyo, S.; Mabuoa, G.; Aoyi, O. "Solar photodegradation of methyl orange and phenol using silica supported ZnO catalyst," Int. J. Innov. Manag. Technol., 2014, 5, 203 -- 206. 28. Ong, S.A.; Min, O.M.; Ho, L.N.; Wong, Y.S. "Comparative study on photocatalytic degradation of Mono Azo Dye Acid Orange 7 and Methyl Orange under solar light irradiation," Water Air Soil Pollut., 2012, 223, 5483 -- 5493. 29. Kumar, K.V.; Porkodi, K.; Rocha, F. "Langmuir-Hinshelwood Kinetics-A theoretical study," Catal. Commun., 2008, 9, 82 -- 84. 30. Mohabansi, N.P.; Patil, V.B.; Yenkie, N. "A comparative study on photo degradation of methylene blue dye effluent by advanced oxidation process by using TiO2/ZnO photo catalyst," Rasayan J. Chem., 2011, 04, 814-819. 31. Navajas, P.M.; Asenjo, N.G.; Santamaria, R.; Menendez, R.; Corma, A.; Garcia, H. "Surface Area Measurement of Graphene Oxide in Aqueous Solutions," Langmuir, 2013, 29, 13443 -- 13448. 32. Reza, K.M.; Kurny, A.S.W; Gulshan, F. "Photocatalytic Degradation of Methylene Blue by Magnetite + H2O2 + UV Process," Int. J. Env. Scn. Dev., 2016, 7 (5), 325-329. 19 33. M. Bakhtiar Azim, Intaqer Arafat Tanim, Riad Morshed Rezaul, Rizwan Tareq, Arafat Hossain Rahul, A.S.W Kurny, Fahmida Gulshan, "Degradation of Methylene Blue using Graphene Oxide- Tin Oxide Nanocomposite as Photocatalyst", Proceedings of the 1st International Conference on Engineering Materials and Metallurgical Engineering, Paper ID: C117, 202-209, 22- 24 December, 2016, Bangladesh Council of Scientific and Industrial Research (BCSIR) Dhaka, Bangladesh. 34. M. Bakhtiar Azim, Intaqer Arafat Tanim, Riad Morshed Rezaul, M. Mozammal Hosen, Fahmida Gulshan , A.S.W Kurny, "Graphene-Oxide for Efficient Photocatalytic Degradation of Methylene Blue under Sunlight and Its Kinetic Study", International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering (IC4ME2), Paper ID: 12, 38-43, 26-27 January, 2017. 20 FIGURES Fig. 1: XRD of Graphite, GO, SnO2, GO-SnO2 and rGO-SnO2 21 (a) GO crumbled layer by layer sheets (b) SnO2 particles 22 Oxygen (c) GO crumbled layer by layer sheets SnO2 particles SnO2 particles GO crumbled layer by layer sheets (d) 23 (e) SnO2 particles rGO layer by layer sheets Fig. 2: SEM of (a) GO, (b) SnO2, (c,d) GO-SnO2 and (e) rGO-SnO2 (a) 24 25 (b) (c) (d) Fig. 3: EDX spectra of (a) GO, (b) SnO2, (c) GO-SnO2 and (d) rGO-SnO2 Fig. 4: FTIR curves of GO, SnO2 and rGO-SnO2 26 27 (a) (b) 28 (c) (d) (e) Fig. 5: Time-dependent absorption spectra of MB solution during natural sunlight irradiation in the presence (a) GO, (b) SnO2, (c) GO-SnO2, (d) rGO and (e) rGO-SnO2 29 (a) (b) 30 Fig. 6: (a) Absorbance versus Irradiation time curves for GO, SnO2, GO-SnO2, rGO and rGO-SnO2 and (b) ln(C0/Ct) versus Irradiation time curves illustrating MB photodegradation by GO, SnO2, GO-SnO2, rGO and rGO-SnO2 (a) (b) 31 32 (c) (d) (e) Fig. 7: Time-dependent absorption spectra of MB solution during UV light irradiation in the presence (a) GO, (b) SnO2, (c) GO-SnO2, (d) rGO and (e) rGO-SnO2 (a) 33 (b) Fig. 8: (a) Absorbance versus Irradiation time curves for GO, SnO2, GO-SnO2, rGO and rGO-SnO2 and (b) ln(C0/Ct) versus UV light Irradiation time curves illustrating MB photo degradation by GO, SnO2, GO-SnO2, rGO and rGO-SnO2 (a) 34 (b) Fig. 9: Photodegradation Efficiency (%) under (a) Natural Sunlight and (b) UV light (a) 35 (b) Fig. 10: The correlation coefficient (R2) values for (a) pseudo-first order and (b) pseudo-second order under natural sunlight irradiation. (a) 36 (b) Fig. 11: The correlation coefficient (R2) values for (a) pseudo-first order and (b) pseudo-second order under UV light irradiation. 37 TABLES Table 1. Properties of Methylene Blue (MB). Properties Synonym name Molecular formula Molecular weight Absorbance wavelength(max) Molecular structure Cationic Azo Dye Basic Blue 9 C16H18ClN3S 319.851 g/mol 664 nm Table 2. The photocatalytic degradation of MB using several catalysts. Authors/Year Catalysts Degradation efficiency (%) Conditions References Soltani et al. 2014 BiFeO3 100% MB Time: 80 min; Catalyst loading: 0.5 g/L-1; [5] Irradiation: Natural Sunlight; pH 2.5 Dariani et al. 2016 TiO2 100% MB Time: 2 hr; Catalyst loading: 0.5 g/L-1; [6] Irradiation: UV light; pH 2.5 38 Table 3. Degradation efficiency and pseudo-first order rate constant for photocatalytic degradation of MB by GO, SnO2, GO-SnO2, rGO and rGO-SnO2 under natural sunlight irradiation: Samples Concentration of MB (mM) Degradation Efficiency (%) R2 Rate constant (min-1) H2O/MB/GO H2O/MB/SnO2 H2O/MB/GO-SnO2 H2O/MB/rGO H2O/MB/rGO-SnO2 0.05 0.05 0.05 0.05 0.05 68.68% 62.81% 89.4% 92.66% 93.59% 0.9057 0.9555 0.9122 0.9793 0.8962 0.0374 0.0165 0.0193 0.0870 0.1833 Table 4. Degradation efficiency and pseudo-second order rate constant for photocatalytic degradation of MB by GO, SnO2, GO-SnO2, rGO and rGO-SnO2 under natural sunlight irradiation: Samples Concentration of MB (mM) Degradation Efficiency (%) H2O/MB/GO H2O/MB/SnO2 H2O/MB/GO-SnO2 H2O/MB/rGO H2O/MB/rGO-SnO2 0.05 0.05 0.05 0.05 0.05 68.68% 62.81% 89.4% 92.66% 93.59% 39 R2 0.8229 0.8221 0.8151 0.7712 0.9212 Rate constant (mM-1min-1) 0.3641 0.2816 1.4059 4.2079 9.75 Table 5. Degradation efficiency and pseudo-first order rate constant for photocatalytic degradation of MB by GO, SnO2, GO-SnO2, rGO and rGO-SnO2 under UV light irradiation: Samples Concentration of MB (mM) Degradation Efficiency (%) R2 Rate constant (min-1) H2O/MB/GO H2O/MB/SnO2 H2O/MB/GO-SnO2 H2O/MB/rGO H2O/MB/rGO-SnO2 0.05 0.05 0.05 0.05 0.05 70.01% 63.28% 94.23% 93.6% 94.82% 0.8931 0.9439 0.9957 0.9787 0.9534 0.0201 0.0167 0.0951 0.0916 0.1974 Table 6. Degradation efficiency and pseudo-second order rate constant for photocatalytic degradation of MB by GO, SnO2, GO-SnO2, rGO and rGO-SnO2 under UV light irradiation: R2 0.8826 0.8475 0.8042 0.9547 0.9654 Rate constant (mM-1min-1) 0.3904 0.2872 5.4437 4.8750 12.2033 Samples Concentration of MB (mM) Degradation Efficiency (%) H2O/MB/GO H2O/MB/SnO2 H2O/MB/GO-SnO2 H2O/MB/rGO H2O/MB/rGO-SnO2 0.05 0.05 0.05 0.05 0.05 70.01% 63.28% 94.23% 93.6% 94.82% 40 GRAPHICAL ABSTRACT Fig. (A): Photodegradation of MB using rGO-SnO2 nanocomposite under sunlight irradiation Fig. (B): Photodegradation of MB using rGO-SnO2 nanocomposite under UV light irradiation 41
1810.12129	1	1810	2018-10-29T13:53:52	Towards Sub-micrometer High Aspect Ratio X-ray Gratings by Atomic Layer Deposition of Iridium	[ "physics.app-ph" ]	X-ray grating interferometry is an excellent technique for X-ray phase contrast imaging and X-ray wavefront sensing with applications in materials science, biology and medical diagnosis. Among other requirements, the method depends on the availability of highly X-ray absorbing metallic gratings. Here, we report on the fabrication and characterization of high aspect ratio iridium gratings with a period of one micrometer and a depth of 30 micrometer combining deep reactive ion etching of silicon and atomic layer deposition of iridium. The implementation of such structures can greatly enhance the sensitivity of grating-based X-ray phase contrast imaging and thus, expand further its broad range of applications.	physics.app-ph	physics	Towards Sub-micrometer High Aspect Ratio X-ray Gratings by Atomic Layer Deposition of Iridium Joan Vila-Comamalaa,b,∗, Lucia Romanoa,b,c, Vitaliy Guzenkoa, Matias Kagiasa,b, Marco Stampanonia,b, Konstantins Jefimovsa,b aInstitute for Biomedical Engineering, University and ETH Zurich, 8092 Zurich, bPaul Scherrer Institut, 5232 Villigen PSI, Switzerland cDepartment of Physics and CNR-IMM, University of Catania, 95023 Catania, Italy Switzerland Abstract X-ray grating interferometry is an excellent technique for X-ray phase con- trast imaging and X-ray wavefront sensing with applications in materials sci- ence, biology and medical diagnosis. Among other requirements, the method depends on the availability of highly X-ray absorbing metallic gratings. Here, we report on the fabrication and characterization of high aspect ratio iridium gratings with a period of one micrometer and a depth of 30 µm combining deep reactive ion etching of silicon and atomic layer deposition of iridium. The implementation of such structures can greatly enhance the sensitivity of grating-based X-ray phase contrast imaging and thus, expand further its broad range of applications. Keywords: Grating-based X-ray Interferometry; Deep Reactive Ion Etching of Silicon; Atomic Layer Deposition of Iridium ∗Corresponding author Email address: [email protected] (Joan Vila-Comamala) Preprint submitted to Microelectronic Engineering October 12, 2017 1. Introduction X-ray grating interferometry[1, 2, 3] is a prominent method for X-ray wavefront sensing[4, 5] and X-ray phase contrast imaging with applications in materials science, biology and medical diagnosis[6, 7, 8]. The technique crucially depends on the availability of highly X-ray absorbing metallic grat- ings. For absorbing typical X-ray photon energies of 20 keV and above, thicknesses of at least 25 -- 30 µm of a high atomic number element, such as gold, are required. On the other hand, typical X-ray grating periods range from 15 down to a few micrometers, and thus, high aspect ratio structures are necessary for most grating-based X-ray phase contrast imaging setups. Re- cently, X-ray phase constrat imaging applications using even smaller grating periods have been proposed[7, 9] but currently used fabrication methods can not routinely produce gratings with periods of one micrometer and below. To date, X-ray absorption gratings are commonly fabricated combining opti- cal or X-ray lithography with gold electroplating in silicon templates[10, 11] or polymer resist molds[12]. However, such fabrication techniques are not easily transferred for the production of sub-micrometer high aspect ratio structures. Here, we propose and demonstrate the production of high as- pect ratio (> 50) iridium gratings with a period of one micrometer and a depth of 30 µm combining deep reactive ion etching (RIE) of silicon and atomic layer deposition (ALD) of iridium[13]. Until now, ALD processes have been successfully applied to conformally deposit oxide layers on high aspect (> 180) nanostructures[14, 15] and ALD metal coatings have also been proven very effective to fabricate high resolution diffractive X-ray optics in the sub-100 nm structure size range[16, 17]. In this work, we performed 2 a thorough optimization of the ALD recipe to extend the conformal metal coating to trench depths of tens of micrometers while keeping the one mi- crometer grating periodicity. After scanning electron microscopy inspection, the fabricated gratings were successfully implemented and characterized in a laboratory X-ray phase contrast imaging setup. 2. Materials and Experimental Methods 2.1. Grating-based Phase Contrast X-ray Imaging The interaction of X-rays with an object can be described by introducing a complex refractive index, n = 1 - δ + iβ, in which β and δ respectively account for the absorption and phase shift experimented by X-ray wavefield due to the specimen[18]. For biological soft tissues with small density differ- ences, the phase shift refractive index variation δ can be up to three orders of magnitude larger than its absorption counterpart β, thus allowing X-ray phase contrast imaging to deliver images with much higher contrast in com- parison to those obtained by conventional X-ray absorption imaging. Among different existing X-ray phase contrast imaging techniques, X-ray grating in- terferometry obtains excellent contrast and quantitative information of the sample under investigation[19, 20]. Grating-based phase contrast imaging can be both performed in synchrotron radiation facilities[2] and using inco- herent laboratory X-ray sources[3]. In the latter case, the setup typically requires the use of two highly absorbing X-ray gratings, G0 and G2, and phase-shifting grating, G1, as schematically represented in Fig. 1. The com- bination of G0 and G1 gratings creates an X-ray intensity modulation at a given downstream distance (Talbot distance) from the G1 phase-shifting 3 grating. The G2 grating is necessary to detect the intensity modulation when the pixel size of the X-ray detector is larger than the period of this X-ray intensity modulation. After inserting the sample near the G1 phase-shifting grating, the X-ray intensity modulation is modified. This change allows the detection of the sample differential phase contrast image by step-scanning one of the three gratings in small steps covering a range of one or several grating periods while recording consecutively the transmission X-ray inten- sity images[2]. The exact geometry and grating periods requirements for an X-ray grating interferometer setup can be calculated using the formulas reported elsewhere[21]. Figure 1: Schematics of laboratory grating-based phase contrast X-ray imaging setup composed of two highly absorbing X-ray gratings, G0 and G2, and a phase-shifting X-ray grating, G1 (not to scale). The fabricated X-ray gratings were tested in a laboratory setup using a Hamamatsu L10101 X-ray microsource with a tungsten target and an ex- pected focused source size of 5 -- 10 µm. The X-ray microsource high voltage 4 was set to 40 kV and an electron current of 0.200 mA was employed. The X-ray images were acquired using a scintillator-based sCMOS camera with pixel size of 22 µm (1024x1024 pixels) from Photonic Science Ltd. (UK). The gratings were mounted on motorized stages with nanometer positioning accu- racy for alignment and scanning simplicity. The gratings periods p0, p1 and p2 were chosen to be one micrometer. Thus, two absorbing gratings made of iridium were mounted in our laboratory X-ray phase contrast imaging setup as G0 and G2. The phase-shifting G1 grating was made of silicon and it had a thickness of 25 µm, thus producing the required phase shift of π rad for an X-ray energy of approximately 20 keV. 2.2. Fabrication of the X-ray Gratings Phase contrast imaging X-ray gratings require thin silicon supporting sub- strates to minimize the X-ray absorption when typical photon energies below 30 keV are employed. Nevertheless, the handling of very thin substrates can be challenging because of their fragility and finally, 4-inch silicon double side polished wafers with a thickness of 250 µm were used to fabricate the X-ray gratings reported here. The schematics of the fabrication process is depicted in Fig. 2. Before the lithography step, the silicon wafers were coated with a 100 nm thick layer of chromium by electron beam evaporation. After that, the grating pattern was produced using conventional UV photolithography on positive photoresist layer (MicroChem Corp. S1805) using a Karl Suss Mask Aligner MA6 in vacuum contact mode. After exposure and development of the photoresist, the pattern was further transferred into the chromium hard mask by a RIE Cl2/CO2 based process. Then, the photoresist residuals were removed by immersion in acetone and isopropyl alcohol. Using the patterned 5 chromium layer as hard mask, the high aspect ratio silicon structures were produced by a SF6 / C4F8 based deep RIE, also commonly know as Bosch process, using an inductively coupled plasma (ICP) Plasmalab 100 system from Oxford Instruments Plasma Technology (UK). The high aspect ratio and high resolution structures were achieved after a careful fine tuning of all process parameters such as the gas flows, chamber pressure and radiofre- quency powers. As schematically suggested in Fig. 2(c), the photolithogra- phy and deep silicon RIE processes were adjusted to produce a 2 µm period silicon grating with a duty cycle of 0.25. In this way, the one micrometer iridium grating was only produced after a conformal metal coating at both sides of the silicon trenches. Since silicon is a very low absorbing material in comparison to iridium, the gaps and the silicon lines of the grating can be regarded as equivalent for the X-ray wavefield that will only be sensitive to effective iridium grating period. In the past, such an approach has been suc- cessfully demonstrated for the production of high resolution X-ray diffractive optics[16, 17]. Finally, the last step in grating fabrication was the conformal deposition of iridium by ALD using iridium acetylacetonate, Ir(acac)3, and oxygen gas, O2, as precursors[13], as depicted in Fig. 2(d). In principle, this deposition method can be used to coat conformally any surface structure by repeatedly supplying two complementary reactant vapors in alternating pulses. Because the chemical reactions are forced to happen entirely on the surface and they are self-limited by the amount of precursor reactant that can be adsorb by the surface, the technique is well-suited for a film growth with almost atomic monolayer accuracy[22, 23]. During the investigations presented here, a PicosunTM R-200 Advanced ALD tool with the capability 6 of delivering a plasma-enhanced oxygen precursor was used. This ALD sys- tem has a large reaction chamber suitable for up to 8-inch wafers and it has 6 separate inlets allowing for several precursors to be supplied. The inlets are constantly flushed with a flow of nitrogen, even during the ALD cycle steps, to prevent the growth of materials and the likely clogging of the precursor inlet orifices. As a result, the reactor chamber also requires constant vacuum pumping to prevent an excessive pressure increase, which is kept at around 10 mTorr. Since the precursor species are supplied by mixing the precur- sor vapor with inlet flow of nitrogen, the constant vacuum pumping of the reactor chamber is artificially limiting the precursor exposure of the surface being coated. However, the system can be operated using a PicoflowTM diffu- sion enhancer operation mode which combines the temporary closure of the vaccuum pumping valve during the precursor delivery pulse of every ALD cycle with the addition of a stainless steel lid to reduce the reactor chamber size. Using this operation mode, the effective dosing time is increased and the diffusion of the precursor species is promoted. Concerning the precursor species, the Ir(acac)3, that is a powder at room temperature, is kept at high temperature (195 deg C) and low pressure, so that it sublimates inside its reservoir. The precursor is delivered by shortly opening the container valves and mixing the Ir(acac)3 vapor with nitrogen flow through the line. There- fore, the Ir(acac)3 vapor pressure limits the total amount of precursor that can be delivered for a single ALD pulse to the reactor chamber. The O2 precursor can be both delivered as gas for an exclusively thermal reaction or through a plasma generator for a plasma-enhanced ALD process with a typical radiofrequency power of 2000 W. 7 Figure 2: Fabrication steps of the X-ray gratings by combining deep reactive ion etching of silicon and atomic layer deposition of iridium. After the photolithography step (a), the grating pattern is transferred into the chromium hard mask, (b), used during the silicon deep reactive ion etching, (c). The final atomic layer deposition step, (d), coats the silicon trenches with a conformal layer of iridium. 3. Results and Discussion 3.1. Optimization of X-ray Grating Fabrication The fabrication procedure of the X-ray gratings reported here relies on combining two well-established techniques -- silicon deep RIE and ALD of iridium -- that required an exhaustive tailoring for our particular purpose. On the one hand, the optimization of the X-ray grating fabrication started with the fine-tuning of the deep reactive ion etching by balancing the etching and passivation times to achieve the required depth of 30 µm, vertical sidewalls 8 with a 2 µm periodicity and the duty cycle of 0.25. The details of the deep RIE recipe are summarized in Tab. 1 and its results are shown in the scanning electron micrograph of Fig. 3(a). It can be seen that all the requirements are met and that the targeted 500 nm silicon lines with a periodicity of 2 µm and depth of 30 µm have been achieved. Table 1: Detailed parameters of the silicon deep reactive ion etching recipe. An ICP Plasmalab 100 system from Oxford Instruments Plasma Technology (UK) was used. Etching Step Passivation Step Temperature [deg C] Number of Cycles Presurre [mtorr] SF6 Flow [sccm] C4F8 Flow [sccm] ICP Power RF Power Step Time [s] 0 350 20 100 5 600 30 3.0 5 100 600 20 3.0 On the other hand, the optimization of the ALD recipe was both ap- proached from a theoretical point of view and from the experimental con- siderations directly derived from the operational aspects of the PicosunTM R-200 Advanced ALD tool reported in Sec. 2.2. From a theoretical stand- point, ALD is expected to coat conformally any surface if sufficient precursor exposure and dosing times are supplied. Nevertheless, a complete step cover- age will strongly depend on the particular aspect ratio of the structures under consideration[22, 24]. For example, and from a kinetic theoretical model, it has been shown that a hole with aspect ratio 43 is expected to require a pre- 9 cursor exposure of about 3000 times larger than a flat surface[22]. Therefore, high aspect ratio structures will require maximizing the precursor exposure of the surface by increasing the precursor partial pressure (i.e. concentration) in the reactor chamber or at least, by increasing the dosing time to promote the precursor diffusion and to ensure the precursor saturation of the surface being coated. Taking this into account, the use of silicon gratings with a periodicity of 2 µm and a duty cycle 0.25 is clearly more advantageous than trying to completely fill smaller trenches of 500 nm. After a few experimental trials, it was confirmed that the essential key for a successful ALD conformal coating was to deliver enough Ir(acac)3 precursor and to promote its diffusion into the silicon trenches by increasing the ALD cycle step times and by employing the PicoflowTM operation mode of the sys- tem. The complete parameter details of the ALD recipe that delivered the highest conformal coating to the 30 µm deep silicon trenches are contained in Tab. 2. This optimized iridium ALD recipe started with a deposition of a thin (~ 10 nm) layer of Al2O3 to promote the nucleation of the iridium layer on the silicon surface. After that, the actual iridium deposition was divided into two separate processes: (1) a short deposition using oxygen plasma-enhanced recipe to obtain a uniform nucleation; and (2) a very long deposition using a thermal oxygen process to prioritize the conformality of the iridium coat- ing. It is reported that the use of a plasma precursor is beneficial to obatin a nucleation of smaller grain which will result in a much smoother layer as the deposition progresses. However, it has also been reported that for long depositions the plasma can lead to unconformal deposition as recombination problems can occurs for high aspect ration structures, that is, the precursor 10 ion species may react before reaching the bottom region of the trench[23]. The results of the ALD recipe can be observed in the scanning electron micro- graph of Fig. 3(b) and they demonstrate the conformal coating of the 30 µm deep silicon trenches. Figures 3(c) and (d) compare the iridium thickness obtain at the top and the bottom of the silicon template and Fig. 3(d) shows a larger portion of the fabricated grating. Table 2: Detailed parameters of the iridium atomic layer deposition recipe for conformal coating of high aspect ratio silicon trenches. The recipe is divided in three separate processes: a deposition of a thin layer of Al2O3 to ensure the iridium growth; the nucleation of iridium using O2 plasma-based recipe to obtain smoother iridium layer; the deposition of the required iridium thickness using thermal O2-based recipe to prioritize a conformal coating of the silicon. 1st ALD Pulse 2nd ALD Pulse 1st ALD Pulse 2nd ALD Pulse 1st ALD Pulse 2nd ALD Pulse Al2O3 Deposition Ir Nucleation Ir Deposition Temperature [deg C] Number of Cycles Thickness [nm] Presurre [mTorr] Precursor Line Gas Flow [sccm] Pulse Time [s] Purge Time [s] PicoflowTM Operation Mode Total Cycle Time [s] Total Deposition Time 370 300 10 370 800 30 370 11000 480 12 TMA 150.0 0.5 6.0 No 12 H2O 150.0 0.5 6.0 No 13 < 1 hour 12 Ir(acac)3 150.0 1.0 6.0 No 12 O2 plasma 60 6.0 6.0 No 19 ~ 4 hours 20 Yes, 25.0 s Ir(acac)3 150.0 4.0 35.0 12 O2 gas 100.0 10.0 35.0 Yes, 25.0 s 120 ~ 15 days The current ALD recipe using the PicosunTM R-200 Advanced ALD tool was able to deliver a conformal coating of iridium for 3 grating substrate pieces of 2 × 2 cm2 with a single ALD run. However during the experimental trials, it was found that the iridium deposition was not completely conformal at the central regions of a full 4-inch silicon wafer with a ~ 7 × 7 cm2 grating patterned area. In this latter case, it is suspected that the total amount of 11 Ir(acac)3 precursor supplied to the chamber was not sufficient to saturate the larger area of the fully patterned 4-inch silicon wafer. This result clearly points the importance of operational aspects of the particular ALD system in use. It is expected that with a dedicated ALD system that would be able to deliver the Ir(acac)3 precursor from more than one reservoir at the same time would allow the coating of larger grating areas and it would probably decrease substantially the very long ALD pulse times required by the equipment used for investigations reported here. 3.2. Performance of the X-ray Gratings The X-ray absorbing iridium gratings were characterized at our labora- tory X-ray phase imaging setup. A picture of the actual system is shown in Fig. 4(a). The setup was arranged to employ the 9th and 13th Talbot order configurations with a grating separation distances of 3.7 and 5.2 cm, respec- tively. The fabricated gratings were mounted and aligned using the motor- ized positioning systems. During the tests, a sample consisting of polystyrene microspheres with a diameter of 700 µm was used. The differential phase con- trast images of the spheres at the 9th and 13th Talbot order configurations are respectively shown in Fig. 4(b) and (c). Comparing the two images, the highest sensitivity is achieved by the 13th Talbot order as it is theoretically expected[21]. The main advantage of having smaller periods is the reduction of the grating separation distance to obtain a similar sensitivity. For exam- ple, when using 3 µm grating periods a separation distance between gratings of 16.3 cm would be required for a similar sensitivity to the one achieved by setup reported here. In addition, using shorter grating separation distances is very beneficial in terms of photon flux when using laboratory X-ray mi- 12 crosources. The measured fringe grating visibility, that is an evaluation of the performance of the X-ray phase imaging setup, was found to be 11.5% and 12.1% for the 9th and 13th Talbot orders. The visibility histograms are shown in the inset plots of Fig. 4(b) and (c). Finally, it can be observed that the phase contrast images are clean of defects, which in turn is an indication of the excellent quality of the fabricated one micrometer period gratings. 4. Conclusions In summary, we have demonstrated the fabrication and characterization of high aspect ratio iridium gratings with a period of one micron and a depth of 30 µm for X-ray phase contrast imaging. The combination of deep RIE of silicon and atomic layer deposition of iridium has been proven as an effective fabrication method to produce high quality X-ray absorbing gratings. The technique has the potential of producing even smaller period gratings over larger areas (> 5 × 5 cm2) if used together with new emerging lithography technologies such as Displacement Talbot Lithography[25] that have the ca- pability of producing sub-micrometer grating patterns over large areas (4 and 8-inch wafers). We believe that the proposed method can be used to provide high quality gratings that will greatly enhance the sensitivity of grating- based X-ray phase contrast imaging and broaden its range of applications. In addition, the technique could be easily transferred for the production of other microcomponents and optical devices. 13 Acknowledgements The authors would like to thank Dr. Z. Wang and Dr. C. Arboleda from the Paul Scherrer Institut (Switzerland) for assistance during the phase con- trast X-ray imaging experiments. This work has been partially funded by the ERC-2012-SRG 310005-PhaseX grant and by the Swiss National Science Foundation, SNSF Grant Number 159263. References [1] A. Momose, S. Kawamoto, I. Kowama, Y. Hamaishi, K. Takai, Y. Suzuki, Demonstration of x-ray talbot interferometry, Jpn. J. Appl. Phys. 42 (2003) L866 -- L868. [2] T. Weitkamp, A. Diaz, C. David, F. Pfeiffer, M. Stampanoni, P. Cloetens, E. Ziegler, X-ray phase imaging with a grating interfer- ometer, Opt. Express 13 (2005) 2062 -- 2068. [3] F. Pfeiffer, M. Bech, O. Bunk, P. Kraft, E. F. Eikenberry, C. G. C. Brn- nimann, C. David, Hard-x-ray dark-field imaging using a grating inter- ferometer, Nat. Materials 7 (2008) 134 -- 137. [4] T. Weitkamp, B. Nhammer, A. Diaz, C. David, X-ray wavefront analysis and optics characterization with a grating interferometer, Appl. Phys. Lett. 86 (2005) 054101. [5] S. Rutishauser, L. Samoylova, J. Krzywinski, O. Bunk, J. Grnert, H. Sinn, M. Cammarata, D. M. Fritz, C. David, Exploring the wave- 14 front of hard x-ray free-electron laser radiation, Nat. Commun. 3 (2012) 947. [6] Z. Wang, N. Hauser, G. Singer, M. Trippel, R. Kubik-Huch, C. Schnei- der, M. Stampanoni, Non-invasive classification of microcalcifications with phase-contrast x-ray mammography, Nat. Commun. 5 (2014) 1 -- 9. [7] C. Arboleda, Z. Wang, T. Koehler, G. Martens, U. V. Stevendaal, M. Bartels, P. Villanueva-Perez, E. Roessl, M. Stampanoni, Sensitivity- based optimization for the design of a grating interferometer for clinical x-ray phase contrast mammography, Opt. Express 25 (6) (2017) 6349 -- 6364. [8] L. B. Gromann, F. De Marco, K. Willer, P. B. Noel, K. Scherer, B. Renger, B. Gleich, K. Achterhold, A. A. Fingerle, D. Muenzel, S. Auweter, K. Hellbach, M. Reiser, A. Baehr, M. Dmochewitz, T. J. Schroeter, F. J. Koch, P. Meyer, D. Kunka, J. Mohr, A. Yaroshenko, H.-I. Maack, T. Pralow, H. van der Heijden, R. Proksa, T. Koehler, N. Wieberneit, K. Rindt, E. J. Rummeny, F. Pfeiffer, J. Herzen, In- vivo X-ray Dark-Field Chest Radiography of a Pig, Scientific Reports 7 (2017) 4807. [9] H. Wen, A. A. Gomella, A. Patel, S. K. Lynch, N. Y. Morgan, S. A. Anderson, E. E. Bennett, X. Xiao, C. Liu, D. E. Wolfe, Subnanoradian x- ray phase-contrast imaging using a far-field interferometer of nanometric phase gratings, Nat. Commun. 4 (2013) 2659. [10] C. David, J. Bruder, T. Rohbeck, C. Grunzweig, C. Kottler, A. Diaz, 15 O. Bunk, F. Pfeiffer, Fabrication of diffraction gratings for hard x-ray phase contrast imaging, Microelectron. Eng. 84 (2007) 1172 -- 1177. [11] D. Noda, A. Tokuoka, T. Hattori, Fabrication of high aspect ratio x-ray grating using silicon dry etching method, AIP Conf. Proc. 1466 (2012) 187. [12] J. Mohr, T. Grund, D. Kunka, J. Kenntner, J. Leuthold, J. Meiser, J. Schulz, M. Walter, High aspect ratio gratings for x-ray phase contrast imaging, AIP Conf. Proc. 1466 (2012) 41 -- 50. [13] T. Aaltonen, M. Ritala, V. Sammelselg, M. Leskela, Atomic layer depo- sition of iridium thin films, J. Electrochem. Soc. 151 (8) (2004) G489 -- G492. [14] E. Shkondin, O. Takayama, J. M. Lindhard, P. Voss Larsen, M. Dysse- holm Mar, F. Jensen, A. V. Lavrinenko, Fabrication of high aspect ratio tio2 and al2o3 nanogratings by atomic layer deposition, J. Vac. Sci. Tech- nol. A 34(3) (2016) 031605. [15] R. Zazpe, M. Knaut, H. Sopha, L. Hromadko, M. Albert, J. Prikryl, V. Grtnerov, J. W. Bartha, J. M. Macak, Atomic layer deposition for coating of high aspect ratio tio2 nanotube layers, Langmuir 32(41) (2016) 10551 -- 10558. [16] K. Jefimovs, J. Vila-Comamala, T. Pilvi, J. Raabe, M. Ritala, C. David, Zone-doubling technique to produce ultrahigh-resolution x-ray optics, Phys. Rev. Lett. 99 (2007) 264801. 16 [17] J. Vila-Comamala, S. Gorelick, V. A. Guzenko, E. Farm, M. Ritala, C. David, Dense high aspect ratio hydrogen silsesquioxane nanostruc- tures by 100 kev electron beam lithography, Nanotechnology 21 (2010) 285305. [18] D. Attwood, Soft X-rays and Extreme Ultraviolet Radiation, Cambridge University Press, Cambridge, UK, 2000. [19] F. Pfeiffer, O. Bunk, C. David, M. Bech, G. L. Duc, A. Bravin, P. Cloetens, High-resolution brain tumor visualization using three- dimensional x-ray phase contrast tomography, Phys. Med. Biol. 52 (2007) 6923 -- 6930. [20] M.-C. Zdora, J. Vila-Comamala, G. Schulz, A. Khimchenko, A. Hipp, A. C. Cook, D. Dilg, C. David, C. Grunzweig, C. Rau, P. Thibault, I. Zanette, X-ray phase microtomography with a single grating for high- throughput investigations of biological tissue, Biomed. Opt. Express 8(2) (2017) 1257 -- 1270. [21] T. Donath, M. Chabior, F. Pfeiffer, O. Bunk, E. Reznikova, J. Mohr, E. Hempel, S. Popescu, M. Hoheisel, M. Schuster, B. Joachim, C. David, Inverse geometry for grating-based x-ray phase-contrast imaging, J. Appl. Phys. 106 (2009) 054703. [22] R. G. Gordon, D. Hausmann, E. Kim, J. Shepard, A kinetic model for step coverage by atomic layer depsotion in narrow holes or trenches, Chem. Vap. Depostion 9 (2) (2003) 73 -- 78. 17 [23] H. C. M. Knoops, E. Langereis, M. C. M. van der Sanden, W. M. M. Kessels, Confromality of plasma-assisted ald: Physical processes and modeling, J. Electrochem. Soc. 157 (12) (2010) G241 -- G249. [24] A. Yanguas-Gil, J. W. Elam, Self-limited reaction-diffusion in nanos- tructured substrates: Surface coverage dynamics and analytic approxi- mations to ald saturation times, Chem. Vap. Deposition 18 (2012) 46 -- 52. [25] H. H. Solak, C. Dais, F. Clube, Displacement talbot lithography: a new method for high-resolution patterning of large areas, Opt. Express 19(11) (2011) 10686 -- 10691. 18 Figure 3: (a) Scanning electron micrograph of the silicon grating fabricated by deep re- active ion etching with a periodicity of 2 µm, a depth of 30 µm and a duty cycle of 0.25. (b) One micrometer iridium grating after the atomic layer deposition process. (c) and (d) demonstrate the obtained conformality by comparing the thickness of iridium at the top and bottom of the silicon trenches. (e) (a) Low magnification scanning electron micrograph of the one micrometer iridium X-ray grating. 19 Figure 4: (a) Picture of the laboratory X-ray phase imaging setup with the fabricated one micrometer period iridium gratings mounted as G0 and G2. (b) and (c) display the differential phase contrast images of a sample made of 700 µm silicon oxide spheres obtained at the 9th and 13rd Talbot orde2r0configurations. The inset plots shows the histogram of the visibility obtained, that is a direct evaluation parameter of the grating- based X-ray phase contrast imaging setup.
1705.09388	1	1705	2017-05-23T18:01:05	Reducing Graphene Device Variability with Yttrium Sacrificial Layers	[ "physics.app-ph", "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	Graphene technology has made great strides since the material was isolated more than a decade ago. However, despite improvements in growth quality and numerous 'hero' devices, challenges of uniformity remain, restricting large-scale development of graphene-based technologies. Here we investigate and reduce the variability of graphene transistors by studying the effects of contact metals (with and without Ti layer), resist, and yttrium (Y) sacrificial layers during the fabrication of hundreds of devices. We find that with optical photolithography, residual resist and process contamination is unavoidable, ultimately limiting device performance and yield. However, using Y sacrificial layers to isolate the graphene from processing conditions improves the yield (from 73% to 97%), average device performance (three-fold increase of mobility, 58% lower contact resistance), and the device-to-device variability (standard deviation of Dirac voltage reduced by 20%). In contrast to other sacrificial layer techniques, removal of the Y sacrificial layer with HCl does not harm surrounding materials, simplifying large-scale graphene fabrication.	physics.app-ph	physics	1 Reducing Graphene Device Variability with Yttrium Sacrificial Layers Ning C. Wang1, Enrique A. Carrion2, Maryann C. Tung1, Eric Pop1,3,a) 1Department of Electrical Engineering, Stanford University, Stanford, CA 94305, USA 2Department of Electrical and Computer Engineering & Micro and Nanotechnology Lab, Uni- versity of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 3Department of Materials Science & Engineering, Stanford University, Stanford, CA 94305, USA Graphene technology has made great strides since the material was isolated more than a decade ago. However, despite improvements in growth quality and numerous "hero" devices, challenges of uniformity remain, restricting large-scale development of graphene-based technologies. Here we investigate and reduce the variability of graphene transistors by studying the effects of contact metals (with and without Ti layer), resist, and yttrium (Y) sacrificial layers during the fabrication of hundreds of devices. We find that with optical photolithography, residual resist and process contamination is unavoidable, ultimately limiting device performance and yield. However, using Y sacrificial layers to isolate the graphene from processing conditions improves the yield (from 73% to 97%), average device performance (three-fold increase of mobility, 58% lower contact resistance), and the device-to-device variability (standard deviation of Dirac voltage reduced by 20%). In contrast to other sacrificial layer techniques, removal of the Y sacrificial layer with HCl does not harm surrounding materials, simplifying large-scale graphene fabrication. a) Author to whom correspondence should be addressed. Electronic mail: [email protected] 2 Since the first experimental demonstration of monolayer graphene in 2004,1 academic and indus- trial research labs have extensively explored applications that leverage the unique electrical, mechan- ical, and thermal properties of this material. To date, these efforts have yielded promising results for high-sensitivity chemical sensors,2 transparent and flexible electronics,3 and analog circuits.4 Rapid advances in large-scale production of graphene via chemical vapor deposition (CVD) have also ac- celerated the development and mass-production of graphene towards practical applications.5 While unique physical properties motivate research in graphene electronics, device yield and spa- tial variability will ultimately dictate industrial impact. To this end, many works have focused on im- proving quality and cost of large-scale CVD graphene growth and transfers,6,7 but fewer have at- tempted to quantify and reduce the detrimental effects of subsequent device fabrication, which may lead to inconsistencies in reported results. Optical photolithography-critical for large-scale fabrica- tion-is particularly problematic as photoresist (PR) residues on the graphene surface negatively im- pact device performance.8 Additionally, post-fabrication residue removal methods, such as non-selec- tive physical etches with CO2,9 are difficult to reproduce in a controlled manner and may damage graphene by inducing tears and wrinkles. In contrast to post-fabrication cleaning methods, modifying and optimizing a process flow to protect graphene from its interaction with PR is a more suitable approach. The semiconductor indus- try has employed native SiO2 as a sacrificial protection layer for Si during integrated circuit fabrica- tion. However, since graphene has no equivalent native oxide, thin aluminum (Al) and titanium (Ti) layers have been introduced as sacrificial layers.10,11,12 Subsequent removal is nontrivial, and can af- fect graphene and surrounding materials (i.e. substrates, insulators, PR) during fabrication. For exam- ple, Al removal via AZ422 developer12 lengthens resist development times, potentially leading to overdevelopment and reduced yield of small features. For Ti removal,10 hydrofluoric (HF) acid was 3 utilized. However, HF can etch underneath the graphene and delaminate devices from the substrate, lowering device yield on oxide substrates such as SiO2. In this work, we introduce the use of yttrium (Y) sacrificial layers to protect graphene. Y is an ideal material for this purpose, as it (1) readily forms a sub-stoichiometric oxide that does not de- grade electrical transport13 and (2) is etched in hydrochloric acid (HCl), which is safe for both SiO2 and PR. The latter point greatly simplifies fabrication, resulting in higher device yield and differenti- ates the Y-sacrificial method from existing alternatives. We fabricate and measure hundreds of de- vices, employing transport models and materials analysis to physically quantify the cause of fabrica- tion-induced degradation. We then introduce the use of Y-sacrificial layers to alleviate degradation, reduce variability, increase performance (three-fold intrinsic mobility improvement and 58% de- crease in contact resistance), and improve our device yield up to 97%. Figure 1 outlines the process flow used, including implementation of the Y-sacrificial layer. Complete fabrication details are given in the Supplementary Material, and we outline here materials characterization relevant to Y-sacrificial layer processing. Monolayer graphene is first grown on Cu foils by chemical vapor deposition at 1000 °C, using isopropyl alcohol (IPA) or CH4 carbon source (additional details are given in the Supplementary Material). The graphene is then transferred onto 90 nm SiO2 on p+ Si (< 0.005 Ω⋅cm) substrates with a poly(methyl methacrylate) (PMMA) support scaf- fold14 [Fig. 1(a)], applying a modified RCA cleaning process7 to minimize wrinkles and impurities (i.e. Fe, Cl, Cu). Immediately after transfer and prior to device fabrication, the samples are coated with ~5 nm of Y via electron-beam evaporation in ~10-7 Torr base pressure to isolate and protect the underlying graphene [Fig. 1(b)]. The Y-coated samples are not metallic, indicating that the Y film is oxidized upon exposure to air after deposition. To characterize the effects of process conditions on device quality, we fabricate back-gated gra- phene field-effect transistors (GFETs) using ultraviolet (UV) photolithography, as shown in Fig. 1(c- 4 h). During Y deposition, a glass slide raised by silicon spacers shields half of each sample to yield control regions where the graphene and subsequent devices are fully exposed to process conditions. This approach leads to two types of devices, as shown in Fig. 1(e), "bare" GFETs with potential PR residue and/or processing damage in the channel and under the contacts, and "Y protected" devices. We also repeat GFET fabrication (with and without Y protection) using pure palladium (Pd) con- tacts instead of Ti/Pd contact stacks (where Ti serves as the adhesion layer) and polymethylglutarim- ide/Shipley 3612 (PMGI/SPR3612) instead of lift-off layer 2000/Shipley 3612 (LOL2000/SPR3612) bilayer PR stacks. By eliminating the Ti adhesion layer which can oxidize, the Pd contacted devices should exhibit cleaner, more ideal contact interfaces and lower contact resistance.15 Derived from the same chemistry, PMGI and LOL2000 differ only in the fact that LOL2000 contains additional con- trast-enhancing dye. Thus, use of PMGI may lower the amount of residual PR. These process splits allow us to compare improvements from complete graphene protection with Y-sacrificial layers vs. other common process variations. The Y sacrificial layer is etched in 10:1 DI:HCl after resist exposure and development to expose the underlying graphene when necessary, such as prior to contact metallization [Fig. 1(c)], channel definition [Fig. 1(f)], and final device definition [Fig. 1(h)]. Contact metals for all devices remained intact despite the total of three HCl baths necessary for back-gated GFET fabrication, with no opti- cally visible delamination or lateral removal of metal. AFM measurements show ~1 to 1.5 nm con- tact metal reduction after a single 15 second bath (Supplementary Figure S3), yielding an HCl etch selectivity ~3:1 of Y to the contact metal (here 1.5 nm Ti capped with 40 nm Pd). Processing with sacrificial Y should not affect device yield for contact metals of typical thickness (> 40 nm) since Y layers are thin (< 5 nm) and etch times are short (at most 60 seconds). Auger spectroscopy results shown in Fig. 2(b) and Supplementary Fig. S2 confirm Y removal in desired regions, and the ~1736 eV peak present only over the channel is indicative of a protective, 5 partially oxidized Y layer.16 Raman spectroscopy of graphene after Y removal also indicates that no graphene damage is induced (Supplementary Figure S4). To assess sample cleanliness, we conduct atomic force microscopy (AFM) of graphene in the "open" contact regions [Fig. 2(c-d)] after PR ex- posure, development and Y etching, but prior to metal deposition. Root mean square (RMS) surface roughness decreases by a factor of two (from 1.29 to 0.69 nm) with the use of a Y-sacrificial layer, consistent with a cleaner surface. Figure 3 displays a summary of electrical data for the subset of 44 bare and 56 Y-protected de- vices fabricated using Ti/Pd contacts and LOL2000 resist, and the arrows display the changes be- tween them. Electrical measurements are taken in vacuum (T = 294 K, P ≈ 10-5 Torr) following an in situ vacuum anneal at 200 °C for 1 hour, which also reverses remnant HCl-induced doping (Supple- mentary Figures S5 and S6). ID vs. VBG measurements (at VDS = 50 mV) as in Fig. 3(a) are then fit to our charge transport model17,18 using a least-squares method to extract contact resistance (RC), carrier puddle density (n), and intrinsic mobility (μ). The total device resistance is R = (L/W)RS + 2RC + Rse- ries, where RS = [qμ(n+p)]-1 is the graphene sheet resistance, q is the elementary charge, and (n+p) is the total carrier concentration which depends on VBG (reaching a minimum at n + p = 2n0 when VBG = V0, the charge neutrality point).17,18 We note that n0 = [(n/2)2 + nth 2]1/2, where n* is the carrier puddle density generated by ionized impurities, and nth are the thermally-generated carriers.17,18 Thus, at a given temperature, lower n0 means lower graphene impurity density. RC follows a transfer-length method (TLM) model,17 and both it and the mobility are listed at a carrier density of 5×1012 cm-2 in the remainder of this work. Rseries = 12 Ω is our total parasitic pad, lead, and cable resistance. Figure 3(a) shows much improved transport in Y-protected samples due to cleaner metal-gra- phene contact interfaces. Devices exhibit electron-hole asymmetry due to use of large metal work- function contacts, which increases contact resistance for electrons and leads to lower current at posi- 6 tive VBG – V0.19 Figure 3(b) reveals that the Y-sacrificial process decreases the estimated puddle car- rier density n* from 4.6×1011 to 2.2×1011 cm-2, which is indicative of a cleaner graphene surface. In fact, the puddle carrier density of the Y-protected CVD graphene samples is comparable to that of previous studies on exfoliated graphene on SiO2.17,18 This is an important device metric, as it sets a limit on the minimum GFET current, and therefore on the maximum achievable on/off current ratio. In the ideal case of impurity-free graphene, the minimum carrier density would be given by ther- mally-generated carriers, 2n0 = 2nth ≈ 1.6×1011 cm-2 at 300 K.18 (Thus, an on/off ratio >100 could be achieved in ultra-clean GFETs at room temperature, depending on contact resistance and assuming >1013 cm-2 maximum carrier density achieved via strong gating. To date, the highest GFET on/off ratios we are aware of are ~24 at 300 K and ~35 at 250 K, for graphene samples encapsulated by h- BN.20) In addition, Figs. 3(c-d) show that average RC and μ improve by factors of ~2.5x and ~3x respec- tively (RC decreases from 4.67 to 1.92 kΩ μm while μ increases from ~1200 to ~3100 cm2V-1s-1). While these values are for holes, similar improvements in RC and μ are also observed for electron transport (Supplementary Figure S7). Table I summarizes the electrical properties and yield of 227 measured devices using atmospheric-pressure CVD graphene, which includes those fabricated with pure-Pd contacts and using PMGI lift-off resist as well. We use a conservative definition of yield, i.e. the percentage of devices with a well-defined, single charge neutrality point (CNP) V0 and on/off ra- tio > 3, where on-current is measured at maximum negative gate bias (VBG = -30 V) and off-current is measured at the CNP. This expresses the yield in terms of electrically "well-behaved" devices, which is a better indicator of useful devices than simply the proportion of electrically active devices. When comparing the device splits processed without the Y-protective layer (Table 1, LOL2000 resist and Ti/Pd contacts vs. PMGI resist and Pd contacts), only yield improves as the number of de- vices with single CNP increases for the PMGI process. Since multiple CNPs signify the presence of 7 charged interface traps,21 the improvement in yield is indicative of cleaner graphene interfaces with use of PMGI vs. LOL2000, where the contrast-enhancing dye may leave surface residue. However, despite a more ideal metal-graphene interface with Pd-only contacts, the lack of average device per- formance improvement suggests another mechanism is suppressing device performance. The hypoth- esis is supported by analyzing the Y-sacrificial layer devices, which exhibit both mobility, impurity density, and contact resistance improvement over "bare" devices, regardless of metal and/or resist used (Table 1). This indicates that process-induced contamination ultimately limits the performance of graphene devices, and attempts to reduce residue are insufficient. In other words, graphene must be fully shielded from fabrication conditions in order to obtain optimal device performance. To demonstrate the wide applicability of our approach, we fabricate and measure additional de- vices using graphene from various academic and commercial sources (our atmospheric vs. low-pres- sure CVD growth vs. Graphene Supermarket, see details in Supplement). The extracted average val- ues of mobility, contact resistance and carrier puddle density (Supplementary Table S1) show a clear improvement for all GFETs fabricated with Y-sacrificial layers, underscoring the effectiveness of the technique. To further quantify the variability reduction, we extract the CNP, also known as the Dirac voltage V0, for all measured devices (Supplementary Figure S8). The average V0 is reduced from 1.22 to 0.89 V and the standard deviation decreased from 3.19 to 2.54 V with the Y-sacrificial process. Given that our test devices are using 90 nm SiO2 as the back-gate dielectric, the equivalent V0 for 1 nm thin SiO2 would be 0.01 ± 0.028 V, very close to the true charge neutrality, signifying that the Y- sacrificial layer technique does not induce additional doping. The impact of process-induced contamination on large-scale GFET fabrication is also evident from the measured on/off ratios of the subset of atmospheric-pressure derived graphene devices, as shown in Fig. 4(a). Regardless of PR stack or contact metal, bare devices (without a Y-sacrificial layer) have higher RC due to process-induced contamination, yielding lower on/off ratio at shorter 8 channel lengths as RC begins to dominate. This problem is alleviated with use of Y-sacrificial layers, which enable cleaner contacts and preserve the on/off ratio down to the shortest channel lengths used in this study (L = 3 μm). In Fig. 4(b) we also quantify a simple analog device metric, the maximum ratio of transconductance to current max(gm/ID), extracted from ID vs. VBG sweeps. The Y-sacrificial layers improve average max(gm/ID) by ~60%, reflecting an improvement in channel quality as the max(gm/ID) is obtained at biases near V0, where the channel dominates device transport. Since the ox- ide is the same for both types of samples, the improvement is solely from the Y-sacrificial layer tech- nique, reinforcing the notion that graphene must be protected from PR during processing to maxim- ize device performance. In summary, we introduced a simple and practical method to improve graphene device quality through the use of Y-protective layers during processing. By electrically characterizing hundreds of graphene devices fabricated under various conditions, we identify through physical, quantitative analysis that PR residue restricts device variability and performance, hampering large-scale analysis. Unlike existing methods, the Y-sacrificial technique protects the graphene without affecting sur- rounding materials, leading to increased device performance, reduced variability, and increased yield (here up to 97%). Although demonstrated for GFETs on SiO2, the Y-sacrificial technique is applica- ble to other HCl-compatible two-dimensional materials and substrates, where process contamination could be a concern. See supplementary material for further details and analysis of yttrium sacrificial layer process as well as additional electrical data. This work was supported in part by Systems on Nanoscale Information Fabrics (SONIC), one of six SRC STARnet Centers sponsored by MARCO and DARPA, by the Air Force Office of Scientific Research (AFOSR) grant FA9550-14-1-0251, in part by the National Science Foundation (NSF) EFRI 2-DARE grant 1542883, and by the Stanford SystemX Alliance. We thank Ling Li and Prof. H.-S.P. Wong for providing the CVD graphene samples from Graphene Supermarket. We thank Sam Vaziri for helpful comments during the manuscript preparation process. 9 1 K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V. Grigorieva, and 2 3 A.A. Firsov, Science 306, 666 (2004). P. Yasaei, B. Kumar, R. Hantehzadeh, M. Kayyalha, A. Baskin, N. Repnin, C. Wang, R.F. Klie, Y.P. Chen, P. Král et al., Nature Communications 5, 4911 (2014); A. Salehi-Khojin, D. Estrada, K.Y. Lin, K. Ran, R.T. Haasch, J.M. Zuo, E. Pop, and R.I. Masel, Applied Physics Letters 100, 033111 (2012). S. Lee, K. Lee, C.-H. Liu, G.S. Kulkarni, and Z. Zhong, Nature Communications 3, 1018 (2012); R.- H. Kim, M.-H. Bae, D.G. Kim, H. Cheng, B.H. Kim, D. Estrada, S.W. Hong, Y. Huang, E. Pop, and J.A. Rogers, Nano Letters 11, 3881 (2011). 4 N.C. Wang, S.K. Gonugondla, I. Nahlus, N.R. Shanbhag, and E. Pop, 2016 IEEE Symposium on VLSI Technology, DOI: 10.1109/VLSIT.2016.7573377 (2016); S.-J. Han, A.V. Garcia, S. Oida, K.A. Jenkins, and W. Haensch, Nature Communications 5, 3086 (2014). 5 A.C. Ferrari, F. Bonaccorso, V. Falko, K.S. Novoselov, S. Roche, P. Bøggild, S. Borini, F. Koppens, V. Palermo, N. Pugno et al., Nanoscale 7, 4598 (2014). 6 Y. Hao, M.S. Bharathi, L. Wang, Y. Liu, H. Chen, S. Nie, X. Wang, H. Chou, C. Tan, B. Fallahazad et al., Science 342 (6159), 720 (2013); S. Rahimi, L. Tao, S.F. Chowdhury, S. Park, A. Jouvray, S. Buttress, N. Rupesinghe, K. Teo, and D. Akinwande, ACS Nano 8 (10), 10471 (2014). 7 X. Liang, B.A. Sperling, I. Calizo, G. Cheng, C.A. Hacker, Q. Zhang, Y. Obeng, K. Yan, H. Peng, Q. 8 9 Li et al., ACS Nano 5, 9144 (2011). R. Shi, H. Xu, B. Chen, Z. Zhang, and L.M. Peng, Applied Physics Letters 102, 113102 (2013). S. Gahng, C.H. Ra, Y.J. Cho, J.A. Kim, T. Kim, and W.J. Yoo, Applied Physics Letters 104 (22), 223110 (2014). 10 C.A. Joiner, T. Roy, Z.R. Hesabi, B. Chakrabarti, and E.M. Vogel, Applied Physics Letters 104, 223109 (2014). 11 S. Mzali, A. Montanaro, S. Xavier, B. Servet, J.-P. Mazellier, O. Bezencenet, P. Legagneux, M. Piquemal-Banci, R. Galceran, B. Dlubak et al., Applied Physics Letters 109 (25), 253110 (2016). 12 A. Hsu, H. Wang, K.K. Kim, J. Kong, and T. Palacios, IEEE Electron Device Letters 32, 1008 (2011). 13 L. Wang, X. Chen, Y. Wang, Z. Wu, W. Li, Y. Han, M. Zhang, Y. He, C. Zhu, K.K. Fung et al., 14 Nanoscale 5 (3), 1116 (2013). J.D. Wood, G.P. Doidge, E.A. Carrion, J.C. Koepke, J.A. Kaitz, I. Datye, A. Behnam, J. Hewaparakrama, B. Aruin, Y. Chen et al., Nanotechnology 26 (5), 055302 (2015). 15 E. Guerriero, L. Polloni, M. Bianchi, A. Behnam, E. Carrion, L.G. Rizzi, E. Pop, and R. Sordan, ACS Nano 7 (6), 5588 (2013). J.M. Baker and J.L. McNatt, Journal of Vacuum Science and Technology 9, 792 (1972). 16 17 M.-H. Bae, S. Islam, V.E. Dorgan, and E. Pop, ACS Nano 5, 7936 (2011). 18 V.E. Dorgan, M.-H. Bae, and E. Pop, Applied Physics Letters 97 (8), 082112 (2010). 19 K. Nagashio and A. Toriumi, Japanese Journal of Applied Physics 50, 070108 (2011); B. Huard, N. Stander, J.A. Sulpizio, and D. Goldhaber-Gordon, Physical Review B 78 (12), 121402 (2008). 20 L. Wang, I. Meric, P.Y. Huang, Q. Gao, Y. Gao, H. Tran, T. Taniguchi, K. Watanabe, L.M. Campos, D.A. Muller et al., Science 342 (6158), 614 (2013); A.V. Kretinin, Y. Cao, J.S. Tu, G.L. Yu, R. Jalil, K.S. Novoselov, S.J. Haigh, A. Gholinia, A. Mishchenko, M. Lozada et al., Nano Letters 14 (6), 3270 (2014). 21 B. Antonio Di, G. Filippo, S. Salvatore, R. Francesco, C. Roberta, S. Thomas, and L. Grzegorz, Nanotechnology 22 (27), 275702 (2011). 10 11 FIG. 1. Side-by-side fabrication steps of control ("bare") graphene devices and of those protected with the Y sacrificial layer. (a) CVD-grown graphene is transferred14 onto 90 nm SiO2 on Si after modified RCA clean.7 (b) ~5 nm Y deposition via electron beam evaporation. Half of sample is covered for the control devices. (c) Contact metal photolithography and subsequent Y removal in dilute 10:1 DI:HCl. Some PR residues remain on control devices. (d) Deposition of contact metal by electron beam evap- oration (here Ti/Pd or pure Pd). (e) Devices after metal liftoff. (f) Channel photolithography, Y re- moval, and graphene removal (O2 plasma). (g) Defined devices after PR removal. (h) Final devices (channel lengths L = 3−10 µm and widths W = 3−20 µm) after subsequent channel lithography and final HCl clean. (f) Optical image of single device which underwent Y processing. PR ResidueY ProtectedBareVBG(a)90 nm SiO2p+SiGraphene5 nm YPhotoresistContact Metal(c)10:1 DI:HCl10 µm10:1 DI:HCl+ O2 Plasma(i)VDS(g)(h)(b)Glass Slide(f)(d)(e)PR Residue 12 FIG. 2. (a) Scanning electron micrograph (SEM) of device after channel definition and before final HCl bath. Symbols correspond to acquisition points for Auger spectroscopy shown in (b). The ~1736 eV peak present only over the channel is lower than known peaks associated with elemental Y and is indicative of partially oxidized Y.16 AFM surface comparison of (c) bare (control) and (d) Y-pro- tected graphene devices at contact regions after photolithography and prior to metallization. Using Y- sacrificial layers decreases RMS roughness from 1.29 to 0.69 nm. 172017401760051015x 104Y-ProtectedRRMS= 0.69 nmRRMS= 1.29 nmHeight[nm]10 µmIntensity [arb. unit]Kinetic Energy [eV]BareSiO2Y/Graphene(c)(a)Graphene(d)(b) 13 FIG. 3. Electrical data for 44 bare (pink squares) and 56 Y-protected (green circles) devices, with arrows denoting the change trend. (a) Measured current vs. back-gate voltage (ID vs. VBG) for two devices displaying performance improvement of Y-protected devices. The Dirac voltage distributions of all bare and Y-protected devices are given in Supplementary Fig. S8. (b) Histogram of estimated carrier puddle density n, which is reduced in Y-protected devices to levels comparable to exfoliated graphene on SiO2.17,18 Cumulative distribution function (CDF) of hole (c) contact resistance and (d) mobility. The contact resistance, in particular, becomes smaller and more consistent for Y-protected devices. Distributions of electron contact resistance and mobility are shown in Supplementary Fig. S7. 02.557.51012.51500.20.40.60.812468051015n0 [1011 cm-2]Counts-200200510152025VBG [V]Id [A/m](b)(a)0200040006000800000.20.40.60.81 [cm2/V-s]CDF()(c)(d)L = 3 µmVDS= 50 mVY-ProtectedBareCDFID[µA/µm]RC[kΩ-µm]µ[cm2/V/s]n[1011cm-2]VBG[V]CDF14 TABLE I. Average electrical properties (hole mobility, contact resistance, carrier puddle density) and yield with two types of resists, Ti/Pd vs. pure Pd contacts, for bare and Y-protected devices. We note that n* = 2.2×1011 cm-2 corresponds to puddle surface potential variation ∆ ≈ ℏvF(πn)1/2 ≈ 55 meV, which is comparable to exfoliated graphene devices on SiO2.17,18 LOL2000 Resist Lithography PMGI Resist Lithography Ti/Pd Contacts Pd Contacts µh RC,h n (cm2V-1s-1) (kΩ-µm) (cm-2) µh RC,h n* (cm2V-1s-1) (kΩ-µm) (cm-2) 1100 5.5 6.1×1011 2600 2.2 4.3×1011 Yield 85% (55/64) 93% (42/45) Yield 73% (44/60) 97% (56/58) Bare 1200 4.7 4.6×1011 Y Protect. 3100 1.9 2.2×1011 15 FIG. 4. (a) Measured on/off ratio of GFETs with various channel lengths with data laterally scattered for clarity. Pink squares represent bare devices and green circles represent Y-protected devices. The latter display up to 3× higher on/off ratio, independent of channel length due to lower RC. (b) Y-sacri- ficial layer also increases analog GFET figure of merit, the max(gm/ID) ratio. Low gm/ID values result from thick SiO2 oxide (90 nm) employed as gate dielectric, but the relative change due to process conditions is evident. (a)(b)Supp.-1 Supplementary Material: Reducing Graphene Device Variability with Yttrium Sacrificial Layers Ning C. Wang1, Enrique A. Carrion2, Maryann C. Tung1, Eric Pop1,3,a) 1Department of Electrical Engineering, Stanford University, Stanford, CA 94305, USA 2Department of Electrical and Computer Engineering & Micro and Nanotechnology Lab, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 3Department of Materials Science & Engineering, Stanford University, Stanford, CA 94305, USA Fabrication Details: Monolayer graphene is first grown on Cu foil (99.8%, Alfa Aesar) via chem- ical vapor deposition (CVD) at 1000 °C in a 1" tube furnace. Two types of growths are conducted, at atmospheric pressure and at 100 mTorr conditions. Atmospheric growths use liquid isopropyl alcohol (IPA, reagent grade) in a metallic bubbler as a carbon source. The bubbler is chilled to 4 °C on a cold plate and 50 SCCM Ar serves as a carrier gas. 250 SCCM of Ar and 50 SCCM H2 are mixed downstream, forming the necessary conditions for growth. Low pressure growths use a gaseous feedstock (100 SCCM CH4) along with 50 SCCM H2 and 1000 SCCM Ar. For both growths, the Cu foil is annealed in growth conditions without carbon sources for 1 hr. Figure 1 in the main text summarizes the subsequent process steps. After growth, graphene is transferred onto 90 nm SiO2 on p+ Si (< 0.005 Ω⋅cm) by etching the Cu substrate in FeCl3, using a PMMA scaffold for support and applying a modified RCA cleaning process1 to minimize wrinkles and impurities (i.e. Fe, Cl, Cu). As a control in our study, a glass slide raised by silicon spacers shields half of each sample from Y evaporation, creating samples with no sacrificial layer [Fig. 1(b)]. We then fabricate back-gated graphene field-effect transistors (GFETs) using UV photolithogra- phy to define contact and channel regions, HCl to etch the sacrificial Y layer (where necessary), and e-beam evaporation for metal deposition of contacts. We use two different types of PR bi- layer stacks (0.2/1.2 μm LOL2000/Shipley 3612 and 0.2/1.2 μm PMGI/Shipley 3612) in order to assess impact on devices electrical performance, as described in the main text. After PR exposure/development and Y etching (samples are immersed in a dilute 10:1 deionized (DI) water to HCl solution for 15 s, although the photoresist survives baths up to 180 s without affecting feature definition), we deposit 1.5/40 nm Ti/Pd via e-beam evaporation (~10-7 Torr) as contact metal. A separate set of devices using 40 nm Pd without a Ti interfacial layer was fabricated to study the impact of interfacial Ti adhesion layers on contact resistance. After 1 hour liftoff in Remover PG, we proceed with photolithography for channel definition (length L = 3−10 µm and width W = 3−20 µm). Finally, Y is removed in open regions after resist development to expose the underlying graphene to a subsequent O2 plasma etch (100 SCCM O2, 150 mTorr, 250 W) for graphene removal and channel definition. Finally, all Y is removed prior to device measurements with a prolonged 60 second dilute HCl bath. Supp.-2 Figure S2. 20x optical images of initial graphene on SiO2 sample, after 5 and 10 nm Y deposition, and after baking on a hot plate in air at 110 °C for 1 minute. Although residue spots are present on all samples, the spots are present on the initial graphene and are therefore not a result of the Y deposition. The hot plate step induces irreversible cracks in the 10 nm Y film, which can also affect the underlying graphene, most likely as a result of rapid oxidation and mismatches in coefficient of thermal expansion. However, the thinner ~5 nm Y film is unaffected, and thus we suggest keep- ing Y films < 5 nm to prevent cracking. Additional measures, such as baking in an inert environ- ment, must be taken if a thicker film is necessary to prevent damage to graphene. Figure S3. (a) SEM of device, followed by full Auger mapping of channel region after contact metallization and prior to final Y removal (Phi Instruments, 5 keV, 10 nA). (b) Yttrium mapping is averaged over 20 scans, while (c) oxygen and (d) carbon mappings are integrated over 5. Sig- nificant Y signal is present only in the channel in figure (b), confirming removal of Y in desired regions as well as remaining Y protecting the channel. Graphene on SiO2Post Transfer5 nm Y Deposition10 nm Y DepositionAfter 110 C bakeAfter 110 C bakeSEMYttriumOxygenCarbon(a)(b)(c)(d) Supp.-3 Figure S4. Tapping mode AFM (Digital Instruments) of Ti/Pd contacts (a) before and (b) after exposure to 10:1 DI:HCl for 15 seconds. Minimal Pd metal (~1 nm) is removed with the diluted solution, confirming a high etch selectivity (~ 3:1) to Y over surrounding contact metal. Figure S5. Raman spectroscopy (WiTec 532 nm laser) of graphene under defined contact region (a) before and (b) after Y removal in 10:1 DI:HCl. No noticeable increase in D-peak is observed after Y deposition, nor after HCl etch, confirming that the Y-sacrificial layer processing is safe for graphene. The G-peak is blue-shifted after exposure to HCl, indicative of some doping; however, this doping effect is reversible, as shown in Fig. S5. InitialAfter 15 s 10:1 DI:HCletch(a)(b)120013001400150016001700180000.511.522.53x 1041200130014001500160017001800020004000600080001000012000G Peak:1606 cm-1G Peak:1590 cm-1Y on GraphenePhotoresistPhotoresistGrapheneSiO2SiO2InitialAfter 15s 10:1 DI:HClEtch(a)(b)Counts (AU)Wavenumber (cm-1)Counts (AU)Wavenumber (cm-1) Supp.-4 Figure S6. Electrical measurement of back-gated GFET before and after channel doping in 2:1 HCl for 1 second. Undoped device is shown as grey squares, while doped measurements are shown as pink circles. The doping effect is largely reversed by a 200 °C anneal in vacuum for 2 hours. Figure S6. Electrical measurements (in ~10-5 Torr vacuum) of back-gated GFET (a) with and (b) without 2:1 HCl treatment for 180 seconds. HCl treated devices exhibit (c) lower charge puddle concentration (n) than untreated devices, implying that the HCl treatment cleans the graphene surface, although (d) mobility is largely unaffected. (e) Untreated and (f) HCl treated devices ex- hibit slightly (g) lower charge puddle concentration with further vacuum annealing (2 hours, 200 °C). (h) Annealing improves hole transport similarly for untreated and HCl treated devices. Device channel length (L) is 2 μm and widths (W) range from 5 – 10 μm. -40-200204001020304050R x W [k x m]VBG[V]3-40-200204001020304050R x W [k x m]VBG[V]UntreatedHNO3Doped20 ⁰C20 ⁰C200 ⁰C100 ⁰C-40-200204001020304050R x W [k x m]VBG[V]-40-200204001020304050R x W [k x m]VBG[V]UntreatedHClDoped20 ⁰C~ 5.5x~ 4x05010015020002468T [C]Rmax / RV0050100150200051015T [C]RS [k / sq] HClHNO3(a)(b)(c)Si (p++)SiO2(90 nm)VDVSVGLW-40-200204001020304050R x W [k x m]VBG[V]3-40-200204001020304050R x W [k x m]VBG[V]UntreatedHNO3Doped20 ⁰C20 ⁰C200 ⁰C100 ⁰C-40-200204001020304050R x W [k x m]VBG[V]-40-200204001020304050R x W [k x m]VBG[V]UntreatedHClDoped20 ⁰C~ 5.5x~ 4x05010015020002468T [C]Rmax / RV0050100150200051015T [C]RS [k / sq] HClHNO3(a)(b)(c)Si (p++)SiO2(90 nm)VDVSVGLWVBG-40-2002040010203040R x W [k x m]VG[V]-40-2002040010203040R x W [k x m]VG[V]-40-2002040010203040R x W [k x m]VG[V]-40-2002040010203040R x W [k x m]VG[V]-40-2002040010203040R x W [k x m]VG[V](a)UntreatedHClTreated2 hr, 200 ºCAnneal(b)VBGVBG(c)(e)(g)HClTreatedHClTreatedUntreatedUntreated(d)(f)(h)n[1011cm-2] Supp.-5 Figure S7. Cumulative distribution function (CDF) of electron contact resistance and mobility for devices with Ti/Pd contacts. 44 bare (pink squares) and 56 Y-protected (green circles) samples measured. Similar to hole transport (main text Fig. 3), electron transport properties improve as well with use of Y-sacrificial layers as (a) average RC decreases from 7.4 to 4.3 kΩ-μm and (b) average μ increases from 1792 to 3730 cm2V-1s-1. Figure S8. Histogram of charge neutrality point (CNP) i.e. Dirac voltage V0 for all 378 devices measured in this study. The dashed red line for bare devices and the solid green line for Y protected devices represent fitted Gaussian distributions. The average V0 is reduced from 1.22 to 0.89 V and the standard deviation decreased from 3.19 to 2.54 V with the Y-sacrificial process. These values are measured on 90 nm SiO2 back-gate oxide, thus values for ~1 nm equivalent oxide thickness (EOT) would be reduced by a factor of 90x (see main text). 051015202500.20.40.60.81Rc [k-m]CDF(Rc)0200040006000800000.20.40.60.81 [cm2/V-s]CDF()(a)(b)CDFCDFRC[kΩ-µm]µ[cm2/V/s]-10-5051005101520253035V0 [V]Counts Table S1. Average values of electrical data (electrons, holes, mobility, contact resistance, puddle charge density, CNP voltage) for all 378 GFETs fabricated from various graphene sources. µe Μh RC,e RC,h [cm2V-1s-1] [cm2V-1s-1] [kΩ-µm] [kΩ-µm] n* [cm-2] V0 [V] Supp.-6 Stanford University (Atmospheric Pressure CVD) Bare Y-Sacr. 1800 3700 1200 3100 7.4 4.3 4.7 1.9 4.6×1011 2.2×1011 University of Illinois Urbana-Champaign (Low Pressure ~100 mTorr CVD) Bare Y-Sacr. 2100 3000 Graphene Supermarket Bare Y-Sacr. ---* 3700 1270 2700 ---* 2300 10.2 6.4 17.9 4.3 6.1 3.5 4.3 1.3 *Data not shown due to poor fit. 5.0×1011 2.8×1011 7.3×1011 4.6×1011 -1.4 -0.8 2.5 0.6 13.5 6.1 X. Liang, B.A. Sperling, I. Calizo, G. Cheng, C.A. Hacker, Q. Zhang et al., ACS Nano 5, 9144 (2011). 1
1902.02855	2	1902	2019-11-08T16:39:37	Integrated magnonic half-adder	[ "physics.app-ph", "cond-mat.other", "cs.ET" ]	Spin waves and their quanta magnons open up a promising branch of high-speed and low-power information processing. Several important milestones were achieved recently in the realization of separate magnonic data processing units including logic gates, a magnon transistor and units for non-Boolean computing. Nevertheless, the realization of an integrated magnonic circuit consisting of at least two logic gates and suitable for further integration is still an unresolved challenge. Here we demonstrate such an integrated circuit numerically on the example of a magnonic half-adder. Its key element is a nonlinear directional coupler serving as combined XOR and AND logic gate that utilizes the dependence of the spin wave dispersion on its amplitude. The circuit constitutes of only three planar nano-waveguides and processes all information within the magnon domain. Benchmarking of the proposed device is performed showing the potential for sub-aJ energy consumption per operation.	physics.app-ph	physics	Integrated magnonic half-adder Qi Wang1, Roman Verba2, Thomas Brächer1, Florin Ciubotaru3, Christoph Adelmann3, Sorin D. Cotofana 4, Philipp Pirro1, and Andrii V. Chumak1,5* 1Faculty of Physics and Research Center OPTIMAS, University of Kaiserslautern, Kaiserslautern 67663, Germany 2Institute of Magnetism, Kyiv 03142, Ukraine 3Imec, Leuven, 3001 Belgium 4Department of Quantum and Computer Engineering, Delft University of Technology, Delft, 2600 The 5Faculty of Physics, University of Vienna, Boltzmanngasse 5, A-1090 Vienna, Austria Netherlands Spin waves and their quanta magnons open up a promising branch of high-speed and low-power information processing. Several important milestones were achieved recently in the realization of separate magnonic data processing units including logic gates, a magnon transistor and units for non-Boolean computing. Nevertheless, the realization of an integrated magnonic circuit consisting of at least two logic gates and suitable for further integration is still an unresolved challenge. Here we demonstrate such an integrated circuit numerically on the example of a magnonic half-adder. Its key element is a nonlinear directional coupler serving as a combined XOR and AND logic gate that utilizes the dependence of the spin-wave dispersion on its amplitude. The circuit constitutes of only three planar nano-waveguides and processes all information within the magnon domain. Benchmarking of the proposed device is performed showing the potential for aJ energy consumption. 142 words from 150 A spin wave (SW) is a collective excitation of the magnetic order in magnetic materials that can propagate in both conducting and insulating media. Spin waves and their quanta magnons are considered as a potential data carrier for beyond Moore computing [1-4]. This is due to their ultrashort wavelengths in the micro- to nanometer range [5- 7] in the microwave and THz frequency range [8,9], due to their ultralow losses and the absence of Joule heating [10,11], resulting in long-propagation distances of spin * Corresponding author. Email: [email protected] 1 information [12,13], as well as because of their abundant nonlinear phenomena [14- 16]. These factors make spin wave highly attractive for wave-based computing concepts [17-27]. Separate spin-wave logic gates [17-19], wave-based majority gates [20-23], transistors [15,24,25] and building blocks for unconventional computing [26- 29] have already been proposed in experiments or micromagnetic simulations. Thus, the next step is the realization of integrated magnonic circuits consisting of at least two elements that are suitable for further integration and interconnection with other magnonic components. In general, there are two approaches to realize magnonic circuits. One is the development of magnonic circuits using magnetoelectric cells [30- 32] and modern spintronics structures [4,33-35] that behave as transducers converting information between magnons and electrons. This concept suffers from a large number of required conversions from spin to charge and vice versa, which have been identified as a serious bottleneck, in particular due to the relatively low conversion efficiencies achieved so far [30]. The second approach is based on the development of all-magnon circuits in which one magnonic gate is directly controlled by the magnons from the output of another magnonic gate without any intermediate conversion to electric signal [15]. Here, we demonstrate such an integrated magnonic circuit on the example of a half- adder built from two logic gates by means of micromagnetic simulations. The magnonic half-adder has a strikingly simple design, consisting of two directional couplers [36]: The first one works in a linear regime and acts as a symmetric power splitter for each of the two inputs. The second coupler operates in a nonlinear regime and simultaneously performs XOR and AND logic operations. Magnonic half-adder design. A general schematic layout of a half-adder in electronics is shown in Fig. 1a. It combines an XOR logic gate and an AND logic gate using three-dimensional bridge constructions. The half-adder is an essential primary component of any arithmetic logic system, which is why we chose it as proof of the concept device for an integrated magnonic circuit. It adds two single binary digital Inputs "A" and "B" and has two Outputs, sum ("S") and carry ("C"). The truth table of a half-adder is shown in Fig. 1b. The design of the proposed magnonic half-adder is not a direct reproduction of a complementary metal oxide semiconductor (CMOS) half-adder in the spin-wave 2 domain. It uses advantages proposed by SW physics and is based on the combination of two directional couplers [36-38] operating in different regimes and performing different functionalities. The sketch of the device is shown in Fig. 1c. Directional Coupler 1 in the magnonic half-adder acts as a power splitter for each of the two inputs and, at the same time, replaces the three-dimensional bridge [36] required for sending the signals from Input "A" to the gate AND and from Input "B" to the gate XOR (see Fig. 1 The operational principle of the magnonic half-adder. a Sketch of half-adder in electronics. Building blocks are highlighted by different colors. b Half-adder truth table. c Schematic view of the magnonic half-adder. In this work we consider the following parameters: The widths of the YIG waveguides are w = 100 nm, thicknesses are h = 30 nm, edge to edge distances between different waveguides are d1 = 450 nm, d2 = 210 nm, the angles between the waveguides are j = 20°, the gaps between the coupled waveguides are d1 = 50 nm, d2 = 10 nm, and the lengths of the coupled waveguides are L1 = 370 nm and L2 = 3 µm. The arrows show the magnons flow path from inputs to logic gates. Fig. 1a). The spin-wave flow paths in the magnonic half-adder are shown by the black and red arrows in Fig. 1c: Spin waves from both inputs are split into two identical spin waves of half intensity by the Directional Coupler 1. One pair of the waves is directed to the Directional Coupler 2 via the Waveguide 1 and the other pair is guided into the idle Output "I" via the Waveguide 3. In the present simulation, the Output "I" just features a high damping region at the end (shown in the figure with a dashed rectangle) and it does not contribute to the half-adding function. However, it acts as an XOR logic gate and, with the use of another directional coupler, can perform the same half-adder 3 operation (see Supplementary Materials). Thus, the modified half-adder can be considered as a combination of a half-adder with a fan-out logic gate which doubles each output of the device. The Directional Coupler 2 performs the actual half-adder operation and its operational principle is described in the next section. The material we use for the numerical simulations is Yttrium Iron Garnet (YIG) which features the smallest today's known SW losses together with large velocities of short- wavelength exchange spin waves [6,10,11,13]. One of the greatest advantages of magnon-based data processing systems is their scalability, since the smallest wavelength is limited by the lattice constant of the used magnetic material. In our simulations we have chosen the minimal width of the waveguides to be 100 nm (see Fig. 1 for the sizes of the structure) which can be reliably fabricated using modern patterning techniques [39-43]. Recently, the propagation of coherent spin waves in nanoscale waveguides of the width down to 50 nm was studied analytically, numerically and experimentally in Ref. [41,42]. Moreover, the nanoscale directional coupler, which is the primary building block of the magnonic half-adder, has been fabricated and studied using the micro-focused Brillouin light scattering in linear and nonlinear regimes [43]. Theoretical model of the nonlinear directional coupler. The processing of data, in general, requires the utilization of elements with nonlinear characteristics that are, e.g., provided by a semiconductor transistor in CMOS. Spin waves possess a variety of natural and very pronounced nonlinear phenomena [44] that potentially can be used for computing such as four-magnon scattering, which was, for instance, used in the demonstration of the first magnon transistor [15]. Here, we utilize the phenomenon of the nonlinear dependence of SW frequency on its amplitude [44,45]. The increase in the SW amplitude results in a shift of the SW dispersion and in a phase accumulation due to the change in SW wavelength while conserving its frequency. The utilization of this phenomenon for data processing has the advantage that, unlike in the magnon scattering-based approaches, no magnon energy is lost to idle magnons generated due to the scattering [15]. The basic configuration of the directional coupler consists of dipolarly-coupled straight parallel waveguides and of bent waveguides in order to guide spin waves in and out -- see Fig. 2a. Directional couplers operating in the linear regime were 4 comprehensively studied e.g., in optics and in magnonics [36,37]. A power dependence of the characteristics of SW-based directional couplers was observed in experiments and simulations on mm-scale [16,38] and nanoscale samples [43]. When two parallel magnetic SW waveguides are placed sufficiently close to each other, the dipolar coupling between them results in a splitting of the dispersion curve of the isolated waveguides into symmetric and antisymmetric modes of the coupled waveguides [36,37]. The analytically calculated [36] dispersion relation of the isolated SW waveguide is shown in Fig. 2b by the grey solid line for the case of the lowest SW mode with a quasi-uniform profile across the waveguide width [41]. The split dispersion relations in the linear regime in the coupled waveguides are shown in the figure by the blue lines. To obtain the linear dispersion, small SW amplitudes are excited by a microwave field of hrf = 2 mT. When the excited SW frequencies are above the minimum frequency of the antisymmetric mode (about 2.278 GHz), the symmetric and antisymmetric SW modes with different wavenumber ks and kas will be excited simultaneously in the coupled waveguides. The constructive and destructive interferences of these two propagating SW modes results in a periodic energy exchange between the two waveguides. In doing so, the spin waves excited in one of the waveguides transfer their energy to the other after propagation over a certain distance which is called the coupling length LC. The coupling length LC can be calculated as and depends on various parameters such as the SW wavelength, the applied magnetic field, the geometrical sizes of the waveguides and their magnetizations [16,36-38]. / k p =D= - k as L C / p k s x The output power in the first waveguide normalized by the total power P1out/(P1out+P2out) can be expressed using the characteristic coupling length LC2: P 1out P + 2out P 1out = 2 cos ( L p 2 ( / 2 C L 2 ) ) , (1) where L2 = 3 µm is the length of the coupled waveguide in the Directional Coupler 2. Figure 2c shows the normalized output power in the first waveguide as a function of the SW frequency f in the frequency range from 2.28 GHz to 2.65 GHz. The result of numerical simulations in the linear regime is shown with the blue symbols and the analytic calculation with the blue solid lines (see Methods for details). One can clearly 5 see, that the output power P1out strongly depends on the SW frequency. This is due to the strong dependence of the coupling length LC2 on the SW wavenumber [36,37]. The coupling length, consequently, defines the energy distribution between the output waveguides for a given length of the coupled waveguides. The small mismatch between simulations and theory in the region below 2.3 GHz is mainly caused by the damping, which is not taken into account in the theory, and by the large sensitivity of the coupling coefficient to the dispersion of the antisymmetric mode, which is practically flat in this region. Fig. 2 Modeling and characteristics of the Directional Coupler 2. a Schematic of the directional coupler. b Analytically-calculated dispersion curves for the coupled waveguides for small (blue lines) and large (red lines) excitation fields hrf. The change in the coupling length LC2 is clearly visible that is associated with the decrease in the SW frequency with the increase in its amplitude. c Normalized output power in the first waveguide P1out/(P1out+P2out) as a function of frequencies for different excitation field hrf (symbols - simulations, lines -- analytic theory). A zoom of the region marked with the dashed rectangle is shown in the top panel. d Simulated normalized output power P1out as a function of the excitation field b0 for a fixed frequency of f = 2.282 GHz for Directional Coupler 1 and 2. 6 When the input SW power increases, nonlinear effects start to play an important role. In the range of relatively weak nonlinearity, the main impact is produced by the nonlinear frequency shift of the SW [16,38,43-45]. Since the dipolar coupling between waveguides is much smaller than the internal forces, i.e., the splitting of the collective SW modes is much lesser than the SW frequency, we can neglect a nonlinear correction to the coupling strength. In this case, the power-dependent SW dispersion of the collective modes of coupled waveguides can be calculated as [46]: f ( ) nl , s as ( , k a x k x ) = f (0) , s as ( k x ) + T a k k x x 2 (2) f ) ( k (0) ,s as where are the dispersion relations of symmetric and antisymmetric modes of the coupled waveguides in the linear regime, 𝑎"# is the dimensionless SW amplitude and 𝑇"# the nonlinear frequency shift of SWs in an isolated waveguide. This frequency x shift is mainly due to the decrease in the effective magnetization of a magnetic material with the increase in the magnetization precession angle [44,45]. All these characteristics are described in more detail in the Methods Section. For the further calculations, we used the numerically found relation between the excitation field and the amplitude of the dynamic magnetization with both in units of mT. In in-plane magnetized structures the nonlinear shift is negative [46] and, thus, according to Eq. (2), the SW dispersion shifts down with an increase in SW amplitude -- see red lines in Fig. 2b. Thus, a fixed SW frequency of 2.282 GHz becomes correspondent to different SW wavenumbers what changes the parameter Dkx = p/LC2 from Dkxlin to Dkxnonlin (see red dots and lines in Fig. 2) with an increase in the excitation field from b0 = 2 mT to 4 mT. Consequently, the coupling length LC2 of the directional coupler also changes. Using the Taylor expansion of the frequency dependence of the coupling length, the power dependence of the output of Directional Coupler 2 can be found: P 1out P +¶ 2out P 1out =- cos 2 ae ç è L p 2 2 L lin 2 C L 2 L lin 2 C p 2 L lin 2 C L ¶ C f 2 T a k k x x 2 ö ÷ ø (3) The power-independent term is proportional to the ratio of the directional coupler length to the coupling length in the linear regime L2/LC2lin. The output power P1out 7 periodically changes with a change in the coupling length and is maximal for the cases L2/LC2lin = 0, 2, 4, … (see Fig. 2c). Simultaneously, as it is seen from Eq. (3), the sensitivity to the nonlinear effect increases with the increase in the ratio L2/LC2lin. Therefore, the longer the directional coupler is and the more coupling lengths it spans, the higher is the nonlinear phase accumulation. This is the reason why the Directional Coupler 2 in our half-adder design is long and features a strong coupling provided by the small gap between the waveguides of only 10 nm. It has a length of L2 = 14LC2lin and it is very sensitive to the increase in the SW amplitude passing through it. As a result, a complete energy transfer from Output 1 to Output 2 is observed in the micromagnetic simulations if the SW amplitude is increased by a factor of two (L2 = 13LC2nonlin) -- see black line in Fig. 2d. The normalized output SW power in the first waveguide decreases from 97.3% at b0 = 2 mT to 2.0% at b0 = 4 mT. Due to this nonlinear switching effect, the Directional Coupler 2 performs a combined AND and XOR logic function, as will be described in the following. Fig. 3 Operational principle of the magnonic half-adder. The SW intensity distributions for different inputs combinations: a "A" = "1", "B" = "0", b "A" = "0", "B" = "1" and c "A" = "1", "B" = "1". The truth tables for each inputs combinations are shown on the top of the structure. The normalized output SW intensities in the outputs are shown on the right side. All the outputs "S" are multiplied by a factor of four. 8 At the same time, the first Directional Coupler 1 in the half-adder design should remain in the linear regime and its coupling length should be independent on the SW amplitude. This is achieved by its smaller length of 370 nm as well as by a decreased strength of the coupling via an increased spacing between the waveguides of 50 nm. As a result, Directional Coupler 1 spans only half of the coupling length L1 = 0.5LC1lin. Thus, it requires much higher SW amplitudes to show a power dependence and in our working range of SW amplitudes, its output remains around 50% and is independent on the excitation field -- see green symbols in Fig. 2d. Hence, in contrast to Directional Coupler 2, Directional Coupler 1 works as a stable energy splitter, combiner and bridge for the two input signals. Operational principle of the magnonic half-adder. The operational principle of the half-adder is shown in Fig. 3. Binary data is coded into the SW amplitude, namely, in the ideal case, a spin wave of a given amplitude (e.g., Mz/Ms = 0.057) corresponds to the logic state "1" while zero SW amplitude corresponds to "0". In the following, we normalize all output SW intensity to the input SW intensity. In the more realistic cases considered below, we utilize an approach from CMOS: a normalized SW intensity below 1/3 is considered to be logic "0" and above 2/3, it is logic "1". The operational principle of the half-adder is as follows. Let us first consider the case of logic inputs "A" = "1" and "B" = "0" -- see Fig. 3a. In this case, the SW injected into the Input "A" is split into two equal parts by the Directional Coupler 1. One of them is directly guided to the Directional Coupler 2 by the upper conduit. The SW intensity is chosen in such a way that Directional Coupler 2 remains in the linear regime (LC2lin » 214 nm » 14/L2) and after initial oscillations, the SW is guided into the Output "S" as shown in the figure. Only about 1.9% of the SW energy goes into the Output "C". This corresponds to the logic Outputs "S" = "1" and "C" = "0". If a SW is injected in Input "B" only, this corresponds to the logic inputs A" = "0" and "B" = "1" -- see Fig. 3b. The situation in this case is quite similar to the previous one: The SW intensity is split into two parts by the Directional Coupler 1, one of which is guided to the output "S" via the Directional Coupler 2 and, thus, again "S" = "1", "C" = "0". The situation is different for the input logics states "A" = "1" and "B" = "1" -- see Fig. 3c. It is assumed that the phase of the SW injected into the Input "B" is shifted by p/2 with respect to the one in the Input "A" in order to compensate the -p/2 phase shift 9 caused by the Directional Coupler 1. Since the spin-wave wavelength is fixed as well as the required phase shift, in practice, the phase shifter can be realized by varying, for instance, the length of the magnonic conduits between two half-adders in series, by varying the waveguide width or by creating inhomogeneities in the biasing magnetic field. As a result, the excitation of SWs in both inputs results in their constructive interference and in a twice larger SW amplitude arriving at Directional Coupler 2. As was discussed above, this increase in the SW intensity switches the coupler to the nonlinear regime (LC2nonlin » 230 nm » 13/L2) and the spin wave is guided to the output "C". This corresponds to the logic Outputs "S" = "0" and "C" = "1" (see Fig. 3c) and, thus, the whole truth-table of the half-adder is realized. The combination of two directional couplers preforms the AND and XOR logic functions due to the phenomenon of nonlinear SW frequency shift. Please note that the all-magnon circuit concept [15] requires that the signal from the output of a magnonics gate can be directly guided into the input of the next one. In order to satisfy this condition, the spin-wave intensity at the Outputs "S" still has to be amplified by a factor of four due to the energy splitting in the Directional Coupler 1 and due to parasitic reflections and spin-wave damping in the waveguides. The output signals "S" shown in Fig. 3 are artificially multiplied by 4. The most promising realization of such an amplifier is based on the utilization of parametric pumping [47,48]. In contrast, the SW amplitude in the output "C" is either vanishingly small or is approximately equal to the input SW amplitude as a consequence of constructive interference. Thus, no amplifier is required for the "Carry" output of the half-adder. In general, the idea presented here and the concept of the half-adder is applicable for any magnetic material. Nevertheless, the requirement that the device length Lde is smaller than the spin-wave decay length should be satisfied. Thus, the materials with large values of spin-wave lifetimes and group velocities, as well as directional coupler with a strong coupling between the waveguides are preferable. Discussion In the final part of this paper, we would like to discuss potential power consumption, scalability, and delay time of the presented magnonic half-adder and also analyze the energy consumption of different spin-wave amplifiers. 10 Energy consumption: For the estimation of energy consumption in the magnonic system (neglecting transducers), the minimal energy consumption can be express as (see Supplementary Materials for the details of the derivation): E = 20 p 3 M s g v fS gr T k x (4) where v gr = 2p f ¶ k ¶ is the SW group velocity, S is the cross-section of the waveguide. As one can see, the energy consumption is independent of the characteristics of SW is of the order couplers and SW amplitude. Note that the nonlinear frequency shift Tkx of the SW frequency f ( xkT µ f ), especially in the exchange-dominated region. The conclusion has arrived that the feasible way to reduce the energy consumption is the decrease of waveguide cross-section S. Another alternative is searching for specific points or mechanisms with anomalously high nonlinearity. The use of exchange spin waves will not decrease the energy consumption. Instead, the SW group velocity increases in the exchange region which results in an increase of the energy consumption. It should be noted that the relation of the Eq. (4) is universal and takes place in other realizations of magnonic half-adders which are based on the nonlinear shift. For the other designs, the only change is the pre-factor 20p/3. Scalability and delay time: The width of the device can be estimated by: wde = 2w+ 4 × 5h (5) where w is the width of waveguide and h is the thickness of waveguide. This equation accounts for the minimal distance between all waveguides and neighboring devices at 5h to make dipolar interaction relatively weak. The gaps between different logic gates are taken into account in this width. The length of the device is given by ( (6) Lde = N + 0.5 ) LC + 4 5h sinϕ where j is the angle of the bent waveguide, LC is the coupling length, N = L2 LC is the ratio between the coupled length of Directional Coupler 2 and the coupling length. The minimal N can be estimated from the condition that Directional Coupler 1, working at half the coupling length, does not significantly change its characteristics at power, which is sufficient to switch the Directional Coupler 2. Simple calculations 11 yield that the change of Directional Coupler 1 transmission is given by cos2 π N −1 , while in the linear regime transmission rate is equal to 1/2. ) / 4N( ) ) ( ( . τd = Lde / vgr . The processing delay is This gives the restriction Nmin = 6. The area of the magnonic half-adder is Area = wdeLde Figure 4 presents results for the half-adder with the dimensions which have been mentioned above. The only difference is that the minimal N = 6 is used instead of N = 14 used in the simulations. The total energy consumption and delay time are very close to the estimations. The area of half-adder is slightly larger than the estimation which can be optimized by decreasing the gaps between the two input waveguides and the length of the Directional Coupler 2. Fig. 4 Estimation of energy consumption, scalability and delay time of the magnonic half-adder. The energy consumption (a), scalability (b), and delay time (c) as a function of SW frequency. The black lines are obtained from Eqs. (4)-(6). The red dots are extracted from simulations. The summary of these parameters is shown in Table 1. Parameters for different devices are estimated: The first one is the device that was simulated and discussed above. The width and thickness of the waveguides are equal to 100 nm and 30 nm which can be reliably fabricated according to the recent achievements [41, 42]. The second device is an estimation performed for a device with w = 30 nm, h = 10 nm and minimal gap d = 10 nm. It has to be mentioned that the second device does not constitute a fundamental limit but is merely an estimation based on the current state of the art of fabrication technology. A further improvement in all characteristics is potentially achievable by introducing another coupling mechanism instead of dipolar coupling. 12 (100 nm) Parameters YIG * 5.58 Area (µm2) 150 Delay time (ns) Total energy consumption 24.6 without amplification (aJ) 2.282 SW frequency (GHz) 340 SW wavelength (nm) 25 SW group velocity (m/s) Electric current based parametric pumping [47] Voltage controlled magnetic anisotropy parametric pumping [48,49] Type of amplifier (30 nm) (7 nm) CMOS ‡ 1.024 6×10-2 35.3 - - - YIG † 1.016 18 1.96 2.29 510 137 105 (aJ/operation) 3 (aJ/operation) Energy consumption Table 1. Characteristics of the half-adder investigated in the paper, estimated characteristics of a device miniaturized down to 30 nm, the calculated parameter of the current state of the art of 7 nm CMOS. * The column is extracted from micromagnetic simulation. † The column is estimated using the Eqs. (4) - (6). ‡ The column is calculated using Cadence Genus by Sorin D. Cotofana for 7 nm CMOS technology. According to the table, the area of the simulated 100 nm feature size half-adder is 5.58 µm2 (the spaces between the neighboring logic gates are also included) and is, thus, only a few times larger than a corresponding 7 nm feature size CMOS device. In contrast to a CMOS realization, the magnonic half-adder core part (without amplifier) consists of only three nano-wires made of the material and of only one planar layer. This drastically simplifies its fabrication and decreases its potential costs. The area can be readily decreased down to 1.016 µm2 for the second device. In addition, it should be noted that the largest part of the half-adder is given by the Directional Coupler 2, which could potentially be further decreased via the utilization of other, stronger coupling mechanisms between the SW waveguides like exchange instead of dipolar coupling. To achieve this, the air gap between the coupled waveguides should be filled with another magnetic material. Operational frequency is an important requirement. In the presented half-adder, the delay time is defined by the whole length of the device with respect to the SW group velocity. In our design, the SW propagation time from input to output is about 150 ns. According to Table 1, the calculation time can be reduced to 18 ns in the second device. This value is still much larger than the 60 ps delay time obtained for 7 nm 13 CMOS and assumes that magnon logic would be more suitable for the slow but low- energy applications. Small energy consumption of computing systems is probably the most crucial requirement taking into account the constantly increasing amount of information that has to be processed. In our simulations, we record the total energy of the device as a function of simulation time. The energy injected into the device per nanosecond is equal to 4.1×10-20 J/ns for the input combinations "A" = 1, and "B" = 1. Please note that only the energy propagating along the positive direction is taken into account. For the 300 ns pulse duration the energy consumption is, thus, 12.3 aJ. For all the operations, the total energy consumption is 24.6 aJ. This is similar to current CMOS values (35.3 aJ) which is calculated using Cadence Genus (see Method). It should be especially highlighted that the energy consumption of the miniaturized device is remarkably low which is around 1.96 aJ. At the same time, we have to underline that this energy consumption is related to the energy within the magnonic domain only and it does not take into account the energy consumed during the transfer of signals between the elements and the energy required for the SW amplification. Finally, the operation of the proposed all-magnon half-adder requires the utilization of the amplifiers at the output "S" in order to bring its amplitude to the level of the input signal. The energy consumption of the amplifier should be added to the consumption of the half-adder itself. Currently, from all the variety of spin-wave amplifiers, the following two approaches seem to represent the main interest. The first one is the parallel parametric pumping which converts microwave photons into magnons at half of the microwave frequency [47]. Such amplifier was simulated (see details in the Supplementary Materials) and its energy consumption appeared to be around 105 aJ (see Table 1), which can be further decreased one or two orders of magnitude by optimizing the parameters of the pumping antenna and the half-adder. In general, this energy is much higher than the energy used to process data within the magnonic domain. The second approach relies on the usage of the voltage control magnetic anisotropy parametric pumping that has been reported recently [48]. This approach uses rf electric rather than the magnetic (and associated electric current) field and, therefore, allows for the drastic decrease in the energy consumption down to ~ 3 aJ per device (was estimated for a nm-thick CoFeB film - see the Supplementary Materials 14 for details). The experimental realization of such a device was reported recently [49]. The same approach should also be applicable for YIG if to place an additional layer of piezoelectric on top [50]. In conclusion, we have proposed and numerically tested a magnonic half-adder that constitutes of two logic gates and is potentially suited for further integration with other logic gates relying on the same concept. In this integrated magnonic circuit, the magnons are controlled by magnons themselves without any conversion to the electric domain which ensures low energy consumption. The half-adder consists of two directional couplers, one of which acts as a power splitter and another which acts as a nonlinear switch. The operational principle of the latter is based on the nonlinear shift of spin-wave frequency with an increase in spin-wave amplitude. An analytic theory is developed to describe this phenomenon and is verified numerically as well as the whole functionality of the half-adder. In comparison to CMOS technology, the half-adder consists of three magnetic nano- wires with one amplifier placed on top of the output "S" and substitutes 14 transistors in electronics. The magnonic half-adder, although being based on 100 nm technology, has characteristics comparable to a 7 nm CMOS half-adder. At the same time, the magnonic half-adder has large potential for miniaturization and further improvement. Methods Spin-wave dispersion calculation in the linear region. The details of the calculations have been fully described in our previous paper [36]. The SW dispersion curve can be obtained by solving the Landau-Lifshitz equation of magnetization dynamics in the linear approximation and neglecting the damping term. The SW dispersion relation in the isolated waveguide is given by [36]: ( ( 2 w w l ( ( 2 w w l (7) F yy + k x ( ) 0 ( ) 0 F zz k x ) ) + k 2 x k 2 x ( k f 0 ) x yy zz + + H M H M ) ) 1 =W W= 2 p 1 2 p The dispersion relation for the two coupled waveguides (two modes) is f s,as ( k x 1 ( ) =W± W± 2 p yy where ii W= ( 2 w w l + H M k 2 + x ( ) F d ii k x ) w ) ( d yy x F M k ) ( , i = y, z, zz w ( F d M k zz x ) ) (8) extBw g= H , wM = gµ0Ms, Ms is the saturation magnetization, g is the gyromagnetic ratio, µ0 is the vacuum permeability, 15 0 M )s 2 2 / ( A is the exchange length, A is the exchange constant, is the lµ= distance between the centers of the two waveguides, is the gap between the two waveguides with width w. The coordinate system used is shown in Fig. 1c: The x- axis is directed along the waveguides, the z-axis is out-of-plane. The tensor xkF d w d= + d describes the dynamical magneto-dipolar interaction [51,52]: ò 1 2 p k k x k 2 k k 2 y 2 F k x ( ) d = k 2 x k 2 k k x k 2 ( f kh ) y ( f kh ) 0 ae ç ç ç ç ç ç ç ç è N k = 2 s k w ! N ik d y e k dk y (9) y ( f kh ) ( f kh ) 0 0 0 1 - ( f kh ) ö ÷ ÷ ÷ ÷ ÷ ÷ ÷ ÷ ø (10) where ( f kh ) 1 (1 exp( =- - - kh )) / ( ) kh , k = k 2 x + k 2 y , h is the thickness, is the Fourier ks transform of the SW profile across the width of the waveguide, and is the normalized constant of the mode profile. In this case, the spins are fully unpinned at the edge of the waveguides [41]. The Fourier transform is, then, described by the function and w w=! sinc( s = / 2) . w! w k w y k Nonlinear frequency shift. The nonlinear shift coefficient xkT in the isolated waveguide can be calculated using the framework of [53] and by assuming a uniform mode profile across the waveguide thickness and width. Accounting for the negligible static demagnetization of a waveguide along its length, , the nonlinear shift coefficient becomes equal to [46]: xxF = 0 0 T k x ( w H ae =-+ ç ç è A k x ) B 2 k x 2 2 w 0 where ( 2 w l + 4 ( M k 2 + x F xx 2 k x ( ) 0 ) 3 w H ) ö ÷ ÷ ø 2 p , (11) A k x = w H + ( w M 2 w M 2 =- B k x 2 2 l k 2 x + F k yy x ( ) 0 + F zz k x ( ) 0 ) , (12) ( F yy k x ( ) 0 F zz k x ( ) 0 ) . (13) The relation between the dynamic magnetization component and the canonical SW amplitude is given by: xka M M a s k =- z 2 a - k x 2 ( u k x x ) v k x , (14) 16 with u k x = w 0 x A + k 02 w and v k x =- sign B k é ë x ù û w 0 x A - k 2 w 0 . (15) Micromagnetic simulations. The micromagnetic simulations were performed by the GPU-accelerated MuMax3 [54] code. The simulated structure of the magnonic half- adder is shown in Fig. 1c. The parameters of nanometer thick YIG are obtained from experiment and are as follows [11]: saturation magnetization Ms = 1.4×105 A/m, exchange constant A = 3.5 pJ/m, and Gilbert damping a = 2×10-4. The damping at the ends of the simulated structure and the high damping absorber is set to exponentially increase to 0.5 to prevent SW reflection [55]. The high damping region could be realized in the experiment by putting another magnetic material or metal on top of YIG to enhance the damping or it can just correspond to waves guided into further parts of the magnonic network. No external bias field is applied. The static magnetization orients itself parallel to the waveguides spontaneously due to the strong shape anisotropy in the nanoscale waveguides. The mesh was set to 10×10×30 nm3. To excite propagating spin waves, a sinusoidal magnetic field b = b0sin(2pft) is applied over an area of 100 nm in length, with a varying oscillation amplitude b0 and the microwave frequency f. Mz(x,y,t) of each cell was collected over a period of 300 ns which is long enough to reach the steady state. The fluctuation mz(x,y,t) were calculated for all cells via mz(x,y,t) = Mz(x,y,t) - Mz(x,y,0), where Mz(x,y,0) corresponds to the ground state. The SW spectra of the output signals are calculated by performing a fast Fourier transformation from 250 ns to 300 ns which corresponds to the steady state. We would like to mention that all these simulations were performed for the defect-free waveguides and without taking into account temperature. The influences of the edge roughness, trapezoidal cross-sections of the waveguides, and temperature can be ignored due to their smallness as it has been shown in our previous studies [36,41]. Calculation of energy consumption of 7 CMOS half-adder. We considered a 7nm HA standard cell afferent to the typical processor corner (room temperature, 0.7V power supply) and evaluated its power consumption using Cadence Genus. To this end, we set an inverter standard cell as driver and a capacitance of 2.5fF as output load, and assume for the nets a 50% probability of logic "1" and a toggle rate of 0.02 per ns. 17 Simulation results indicate a total power consumption of 587.994nW out of which the dynamic component (divided into nets power and internal power, which account for 87.7% and 13.3% of the dynamic power, respectively) dominates the less than 1nW leakage component. Acknowledgements: The authors thank Burkard Hillebrands for support and valuable discussions. This research has been supported by ERC Starting Grant 678309 MagnonCircuits and by the DFG through the Collaborative Research Center SFB/TRR-173 "Spin+X" (projects B01) and through the Project DU 1427/2-1. R. V. acknowledges support from the Ministry of Education and Science of Ukraine, Project 0118U004007. Author contributions: Q.W proposed the magnonic half-adder design, performed micromagnetic simulation jointly with T. B. and P. P., and wrote the first version of the manuscript. T. B. and A. V. C. devised and planned the project. R. V. developed the analytical theory and estimated the energy consumption of the half-adder. F. C., C. A. and S. D. C. performed the benchmarking and calculated the parameters of 7 nm CMOS half-adder. P. P and A. V. C led the project. All authors discussed the results and contributed to writing the manuscript. References: 1. Khitun, A. Bao, M. & Wang, K. L. Magnonic logic ciruits. J. Phys. D Appl. Phys. 43, 264005 (2010). 2. Lenk, B. et al. The building blocks of magnonics. Phys. Rep. 507, 107-136 (2011). 3. Nikonov, D. E. & Young, I. A., Overview of beyond-CMOS devices and a uniform methodology for their benchmarking. Proc. IEEE 101, 2498-2533 (2013). 4. Chumak, A. V., et al. Magnon spintronics. Nat. Phys. 11, 453-461 (2015). 5. Wintz, S. et al. Magnetic vortex cores as tunable spin-wave emitters, Nat. Nano., 11, 948-953 (2016). 6. Yu, H., et al. Approaching soft x-ray wavelengths in nanomagnet-based microwave technology. Nat. Commun. 7, 11255 (2016). 18 7. Brächer, T. et al. Detection of short-waved spin waves in individual microscopic spin-wave waveguides using the inverse spin hall effect. Nano Lett. 17, 7234-7241 (2017). 8. Kampfrath, T. et al. Coherent terahertz control of antiferromagnetic spin waves. Nature Photon. 5, 31-34 (2011). 9. Balashov, T. et al. Magnon dispersion in thin magnetic films J. Phys.: Condens. Matter, 26, 394007 (2014). 10. Kajiwara, Y. et al. Transmission of electrical signals by spin-wave interconversion in a magnetic insulator. Nature 464, 262-266 (2010). 11. Dubs, C. et al. Sub-micrometer yttrium iron garnet LPE films with low ferromagnetic resonance losses. J. Phys. D Appl. Phys. 50, 204005 (2017). 12. Pirro, P. et al. Spin-wave excitation and propagation in microstructured waveguides of yttrium iron garnet/Pt bilayers. Appl. Phys. Lett. 104, 012402 (2014). 13. Yu, H. et al. Magnetic thin-film insulator with ultra-low spin wave damping for coherent nanomagnonics. Sci. Rep. 4, 6848 (2014). 14. Demidov, V. E. et al. Nonlinear propagation of spin waves in microscopic magnetic stripes. Phys. Rev. Lett. 102, 177207 (2009). 15. Chumak, A. V., Serga, A. A., & Hillebrands, B. Magnon transistor for all-magnon data processing. Nat. Commun. 5, 4700 (2014). 16. Sadovnikov, A. V. et al. Nonlinear spin wave coupling in adjacent magnonic crystals. Appl. Phys. Lett. 109, 042407 (2016). 17. Schneider, T., Serga, A. A. & Hillebrands, B. Realization of spin-wave logic gate. Appl. Phys. Lett., 92, 022505 (2008). 18. Jamali, M. et al. Spin wave nonreciprocity for logic device applications. Sci. Rep. 3, 3160 (2013). 19. Zografos, O. et al. Exchange-driven magnetic logic. Sci. Rep. 7.12154 (2017). 20. Khitun, A. & Wang, K. L. Non-volatile magnonic logic circuits engineering. J. App. Phys. 110, 034306 (2011). 21. Klingler, S. et al. Design of a spin-wave majority gate employing mode selection. Appl. Phys. Lett. 105, 152410 (2014). 22. Radu, I. P. et al. Spintronic majority gates. in 2015 IEEE International Electron Devices Meeting (IEDM), pp. 32.5.1-32.5.4, 2015. 23. Fischer, T. et al. Experimental prototype of a spin-wave majority gate. Appl. Phys. Lett. 110, 152401 (2017). 19 24. Wu, H. et al. Magnon valve effect between two magnetic insulators. Phys. Rev. Lett. 120, 097205 (2018). 25. Cramer, J. et al. Magnon detection using a ferroic collinear multilayer spin valve. Nat. Comm. 9, 1089 (2018) 26. Torrejon, J. et al. Neuromorphic computing with nanoscale spintronic oscillators. Nat. Nano. 547, 428-431 (2017). 27. Brächer, T. & Pirro, P. An analog magnon adder for all-magnonic neurons. J. Appl. Phys. 124, 152119 (2018). 28. Kozhevnikov, A. et al. Pattern recognition with magnonic holographic memory device. Appl. Phys. Lett. 106, 142409 (2015). 29. Papp, Á. et al. Nanoscale spectrum analyzer based on spin-wave interference. Sci. Rep. 7, 9245 (2017). 30. Cherepov, S. et al. Electric-field-induced spin wave generation using multiferroic magnetoelectric cells. Appl. Phys. Lett. 104, 082403 (2014). 31. Dutta, S. et al. Non-Volatile clocked spin wave interconnect for beyond-CMOS nanomagnet pipelines. Sci. Rep. 5, 09861 (2015). 32. Duflou, R. et al. Micromagnetic simulations of magnetoelastic spin wave excitation in scaled magnetic waveguides. Appl. Phys. Lett. 111, 192411 (2017). 33. Sandweg, C. W. et al. Spin pumping by parametrically excited exchange magnons. Phys. Rev. Lett. 106, 216601 (2011). 34. Collet, M. et al. Generation of coherent spin-wave modes in yttrium iron garnet microdiscs by spin-orbit torque. Nat. Comm. 7, 10377 (2016). 35. Talmelli, G. et al. Spin-wave emission by spin-orbit-torque antennas. Phy. Rev. Appl. 10, 044060 (2018). 36. Wang, Q. et al. Reconfigurable nanoscale spin-wave directional coupler. Sci. Adv. 4, e1701517 (2018). 37. Sadovnikov, A. V. et al. Directional multimode coupler for planar magnonics: Side-coupled magnetic stripes. Appl. Phys. Lett. 107, 202405 (2015). 38. Sadovnikov, A. V. et al. Toward nonlinear magnonics: Intensity-dependent spin- wave switching in insulating side-coupled magnetic stripes. Phys. Rev. B 96, 144428 (2017). 39. Duan, Z. et al. Nanowire spin torque oscillator driven by spin orbit torques. Nat. Comm. 5, 5616 (2014). 20 40. Schneider, M. et al. Bose-Einstein condensation of quasi-particles by rapid cooling. arXiv:1612.07305. 41. Wang, Q. et al. Spin pinning and spin-wave dispersion in nanoscopic ferromagnetic waveguides. Phys. Rev. Lett. 122, 247202 (2019). 42. Heinz, B. et al. Propagation of coherent spin waves in individual nano-sized yttrium iron garnet magnonic conduits. arXiv: 1910.08801 (2019). 43. Wang, Q. et al. Realization of a nanoscale magnonic directional coupler for all- magnon circuits. arXiv: 1905.12353 (2019). 44. Gurevich, A. G. & Melkov, G. A. Magnetization oscillations and waves (CRC- press, London, New York, 1996). 45. Bauer, H. G. et al. Nonlinear spin-wave excitations at low magnetic bias fields. Nat. Comm. 6, 8274 (2015). 46. Verba, R. et al. Excitation of propagating spin waves in ferromagnetic nanowires by microwave voltage-controlled magnetic anisotropy. Sci. Rep. 6, 25018 (2016). 47. Brächer, T., Pirro, P. & Hillebrands, B. Parallel pumping for magnon spintronics: Amplification and manipulation of magnon spin currents on the micron-scale. Phys. Rep. 699, 1-34 (2017). 48. Verba, R. et al. Amplification and stabilization of large-amplitude propagating spin waves by parametric pumping. App. Phys. Lett. 112, 042402 (2018). 49. Chen, Y. et al. Parametric resonance of magnetization excited by electric field. Nano. Lett. 17, 572-577 (2017). 50. Sadovnikov, A. V. et al. Magnon straintronics: reconfigurable spin-wave routing in strain-controlled bilateral magnetic stripes. Phys. Rev. Lett. 120, 257203 (2018). 51. Verba, R. et al. Collective spin-wave excitations in a two-dimensional array of coupled magnetic nanodots. Phys. Rev. B 85, 014427 (2012). 52. Beleggia, M. et al. On the magnetostatic interactions between nanoparticles of arbitrary shape. J. Magn. Magn. Mater. 278, 270-284 (2004). 53. Krivosik, P. & Patton, C. E. Hamiltonian formulation of nonlinear spin-wave dynamics: Theory and applications. Phys. Rev. B 82, 184428 (2010). 54. Vansteenkiste, A. et al. The design and verification of MuMax3. AIP Adv. 4, 107133 (2014). 55. Venkat, G. Fangohr, H. & Prabhakar, A. Absorbing boundary layers for spin wave micromagnetics. J. Magn. Magn. Mater. 450, 34-39 (2018). 21 Supplementary Material Integrated magnonic half-adder Qi Wang1, Roman Verba2, Thomas Brächer1, Florin Ciubotaru3, Christoph Adelmann3, Sorin D. Cotofana 4, Philipp Pirro1, and Andrii V. Chumak1,5,† 1Faculty of Physics and Research Center OPTIMAS, University of Kaiserslautern, Kaiserslautern 67663, Germany 2Institute of Magnetism, Kyiv 03142, Ukraine 3Imec, Leuven, 3001 Belgium 4Department of Quantum and Computer Engineering, Delft University of Technology, Delft, 2600 The 5Faculty of Physics, University of Vienna, Boltzmanngasse 5, A-1090 Vienna, Austria Netherlands 22 In the supplementary material, we first discuss the details of the estimation of the energy consumption for magnonic half-adder in Section S1. The modified device operates as a combination of a half-adder with a fan-out gate is discussed in Section S2. In Section S3, we analyzed three different spin-wave amplifiers and estimated the energy consumption of two of them. S1. Energy consumption for magnonic half-adder For the estimation of energy consumption in the magnonic system (neglecting transducers), we can simply calculate the energy of the SW pulse. The energy of SW pulse is given by E = M V s g 2w c k k (1) where SW amplitude ck is defined so that SW norm Ak = 1 (which is the same as common u-v transformation), and V is the volume of SW pulse. Naturally, V = Svgrt, with S being the cross-section of the waveguide, vgr -- SW group velocity and t -- pulse duration. The SW amplitude is determined from the condition that it should be enough to switch nonlinear directional coupler. From Eq. (3) of the paper, we have the following condition † Corresponding author. Email: [email protected] L L c p 2 L c L ¶ c f ¶ T c k k 2 = p 2 (2) ¶cL f ¶ For simplicity, we neglect the sign of Tk and -- they are not important. Then, using the definition of the coupling length via the coupling frequency , and setting L/Lc=N, N = 0, 2, 4,… (i.e., in the linear regime we work on the N/2 - the transmission maximum), we get N ¶W f W ¶ (3) T c k k W 1 = 2 =D , wD 2t p w The duration of the SW pulse is determined by the maximal frequency width of the pulse , which should be small enough for stable operation of both linear and nonlinear couplers. In the case of the nonlinear coupler, which works near the maximum or minimum of transmission, 2 cos 1 » - L W v gr 1 2 ae ç ç è L v gr ö ÷ ÷ ø DW 2 (4) where 1<<x / w w ) ( DW = ¶W ¶ D , we get means 0.1£x and the last term should be 1<< . Using the convention that DW W < 0.3 N (5) 0.2 2³N DW W < Performing the same calculations for a linear coupler (assuming its length . In fact, the condition for the nonlinear case is more severe, L=Lc/2), we get as . More rigorously, the minimal size of the nonlinear coupler N can be estimated from the condition that the linear coupler, working at half the coupling length, does not significantly change its characteristics at the power which is enough to the switch nonlinear coupler. Simple calculations yield that the change of linear coupler transmission is given by , while the linear regime transmission rate is equal to 1/2. This gives the restriction Nmin = 6 (assuming that linear and nonlinear coupler have the same gap). 1) / (4 )) 2cos ( ( p -N N Combining Eqs. (5, 3, 1), we get the final expression for the minimal energy consumption (Eq. 4 in the paper) E = 20 p 3 M s g v fS gr T k (6) 23 S2 Modified half-adder with fan-out gate In this manuscript, the Output "I" features high damping to avoid the reflection at the end of the waveguide. However, the SW amplitude in this waveguide satisfies the XOR logic gate -- see truth table Table S1. In the further, it could also be used for logical operation. Inputs A 0 1 0 1 B 0 0 1 1 Output I 0 1 1 0 Table S1. Truth of Output "I" Furthermore, the Output "I" could also be used to perform the half-adder operation. In order to do that, an additional coupler was introduced to the Output "I" as shown in Fig. S1(a)-(c). For this, another waveguide is closely placed to the Output "I" and the phase difference between the inputs "A" and "B" has been changed from p/2 to 0. The operation principle is identical to the original half-adder for the inputs combinations "A" = "1", "B" = "0" and "A" = "0", "B" = "1" as shown in Fig. S1 (a)-(b). If the inputs are "A" = "1", "B" = "1", the excitation of spin waves in both inputs results in the transfer of spin waves of the same intensities to both directional couplers. Fig. S1 (a-c) The operational principle and (d) truth table of two half-adders with shared inputs (corresponds to the half-adder with added fan-out logic gate). 24 In this case, the SW intensity reached the nonlinear directional coupler is twice comparing to the single input (but not four times larger like in the original half-adder design). Nevertheless, even the increase in the SW intensity twice between the logic states allowed for the triggering of the nonlinear switching phenomenon (after adapting working frequency and the coupled length of the device) - see Fig. S1(c). This device operates as a combination of half-adder with fan-out gate and each logic output is doubled. S3. Energy consumption of parametric amplification The amplifier is a very important point for the whole all-magnon data processing concept since the amplitude of the output signal should always be equal to the input one. The energy consumption of the amplifier might be crucial. To find the answer which amplifier is the most suitable, we analyzed three different approaches: Spin- Orbit Torque-based approach, electric current-based parametric pumping, and Voltage Control Magnetic Anisotropy (VCMA) electric field-based parametric pumping. 1) The first one is based on the usage of, so-called Spin-Orbit Torque (the combination of spin Hall effect (SHE) with spin transfer torque (STT)) [e.g., described in the review [1]. Unfortunately, this approach does not seem to be suitable since the amplification of the propagating spin waves was not demonstrated so far. Even the result reported in Ref. [2] hardly can be used to amplify the propagation spin waves since the final smallest damping parameter in bilayer YIG/Pt with the current was comparable to the original damping of pure single layer YIG. The spin-orbit torque can only compensate for the damping caused by spin pumping due to the introduction of the Pt layer. To sum up, the output spin-wave intensity is always lower than the input in this system. Moreover, this approach requires the applications of electric currents of high densities and, therefore, hardly can ensure small energy consumption. 2) The second way is a classical parallel parametric pumping process in which the pumping rf magnetic Oersted field is created by sending of rf signal through a metallic strip [3]. This is the most well-studied approach in magnonics. In order to find the energy consumption of parametric amplification, we perform simulations in a single waveguide with a pumping antenna. Figure S2(a) shows the schematic picture 25 of the parametric pumping region where an antenna (width: wa = 500 nm, thickness: ta = 300 nm) is placed on top of the YIG waveguide. Fig. S2 (a)A schematic picture of the parametric amplifier. (b) Output spin-wave intensity as a function of pumping current for different conditions. In the simulations, the pumping frequency (2f) is twice the frequency of the initial spin wave. The spin-wave intensity at the frequency of f is extracted after the pumping region and is plotted as a function of pumping current. The blue dotted line shows the spin-wave intensity when only the pumping current is applied, where a clear threshold is observed. The intensity of the initial spin wave is marked by the dashed line. The black dotted line presents the total energy of the output spin-wave after the amplification when the pumping current and initial spin wave are applied simultaneously. The gap between the dashed line and the black dotted line indicates the amplification due to the parallel pumping. It is worth to mention that the parametric pumping starts to amplify the initial spin wave before the it reaches the threshold of the parametric generation. This means that the parametric pumping will not directly excite parasitic spin waves in the magnonic circuits. The spin-wave intensity drops down in the high pumping current region due to the four magnon scattering. In order to obtain a 4 times amplification of SW intensity, which is needed for the magnonic half-adder, the pumping current should be around 8.8 mA as shown in 26 Fig. S2(b). For each output "S", the length of the pumping antenna is at least La=100 nm (the width of waveguide). Then, the resistance of the pumping antenna can be calculated using R=rLa/(wata)=0.016 Ohm (r = 2.44×10-8 Ohm/m is the resistivity of gold at room temperature). The pumping time is 300 ns. Thus, the energy consumption of amplification is around 105 aJ, which can be further decreased one or two orders of magnitude by optimizing the size of the pumping antenna and the half- adder. In general, this energy consumption is much higher than the energy used to process data within magnonic domain. 3) The most suitable approach appeared to be the VCMA parametric pumping that has been reported in Ref. [4]. In this case, the parametric process is similar to the previous one but instead of the magnetic rf field (which requires large currents), an electric rf field is used. This allows for the drastic decrease in the energy consumption of the amplification down to 3 aJ as was estimated for the example of a nm-thick CoFeB waveguide. The functionality of such kind of parametric pumping was recently reported experimentally in [5]. The detail of estimation is given below. For VCMA parametric pumping, the electric field required is equal to: E = 2 W E d = wee 0 wL tan d + ae ç ç è hM b s th 2b dielectric losses and Ohmic ones, (7) with 𝛽 -- magnetoelectric coefficient and h -- waveguide thickness. Losses consists of where d -- dielectric thickness, 𝜀 -- dielectric permittivity, tan𝛿 -- tangents of dielectric thickness h = 1 nm (for perpendicular anisotropy), 𝛽=100 fJ/Vm, for MgO 𝜀=9.6 and tan𝛿=1078, tunnel resistance per area product 109Ωµm< for d = 2.2 nm and losses, R -- tunnel resistance. Both terms depend on the thickness d and the total losses are minimal at certain optimal d. In the estimation, we use L = 100 nm (the length of pumping area), CoFeB (8) 2 2 E d R ö t ÷ ÷ ø increases in order per each 0.4 nm [6]. Then pumping electric field is E ~ 1-1.5 V/nm, optimal MgO thickness d ~ 2.6 nm, energy consumption ~ 3 aJ. 27 References 56. Demidov, V. E. et al. Magnetization oscillations and waves driven by pure spin currents. Phys. Rep. 673, 1-32 (2017). 57. Evelt, M. et al. High-efficiency control of spin-wave propagation in ultra-thin yttrium iron garnet by the spin-orbit torque. Appl. Phys. Lett. 108, 172406 (2016). 58. Brächer, T. Pirro, P, & Hillebrands, B. Parallel pumping for magnon spintronics: Amplification and manipulation of magnon spin currents on the micron-scale. Phys. Rep. 699, 1-34 (2017). 59. Verba, R. et al. Amplification and stabilization of large-amplitude propagating spin waves by parametric pumping. App. Phys. Lett. 112, 042402 (2018). 60. Chen, Y. et al. Parametric resonance of magnetization excited by electric field. Nano. Lett. 17, 572-577 (2017). 61. D. Mazumdar, Coherent magnetotunneling based on (001) magnesium oxide barrier, PhD thesis, Brown University, (2007). 28
1807.07326	1	1807	2018-07-19T10:06:02	Influence of the Hole Transporting Layer on the Thermal Stability of Inverted Organic Photovoltaics Using Accelerated Heat Lifetime Protocols	[ "physics.app-ph" ]	High power conversion efficiency (PCE) inverted organic photovoltaics (OPVs) usually use thermally evaporated MoO3 as a hole transporting layer (HTL). Despite the high PCE values reported, stability investigations are still limited and the exact degradation mechanisms of inverted OPVs using thermally evaporated MoO3 HTL remain unclear under different environmental stress factors. In this study, we monitor the accelerated lifetime performance of non-encapsulated inverted OPVs using thiophene-based active layer materials and evaporated MoO3 under the ISOS D-2 protocol (heat conditions 65 {\deg}C). The investigation of degradation mechanisms presented focus on optimized P3HT:PC[70]BM-based inverted OPVs. Specifically, we present a systematic study on the thermal stability of inverted P3HT:PC[70]BM OPVs using solution processed PEDOT:PSS and evaporated MoO3 HTL. Using a series of measurements and reverse engineering methods we report that the P3HT:PC[70]BM/MoO3 interface is the main origin of failure of P3HT:PC[70]BM-based inverted OPVs under intense heat conditions.	physics.app-ph	physics	Influence of the Hole Transporting Layer on the Thermal Stability of Inverted Organic Photovoltaics Using Accelerated Heat Lifetime Protocols Felix Hermerschmidt,a Achilleas Savva,a Efthymios Georgiou a , Sachetan M. Tuladhar,b James R. Durrant,c Iain McCulloch,c Donal D. C. Bradley,d Christoph J. Brabec,e Jenny Nelson,b Stelios A. Choulisa,* a Molecular Electronics and Photonics Research Unit, Department of Mechanical Engineering and Materials Science and Engineering, Cyprus University of Technology, 3041 Limassol, Cyprus b Department of Physics, Imperial College London, London SW7 2AZ, UK c Department of Chemistry, Imperial College London, London SW7 2AZ, UK d Departments of Engineering Science and Physics, Division of Mathematical, Physical and Life Sciences, University of Oxford, Oxford OX1 3PD, UK e Institute for Materials in Electronics and Energy Technology, Friedrich-Alexander University Erlangen-Nuremberg, 91054 Erlangen, Germany KEYWORDS. Organic photovoltaics, inverted structure, thermal stability, ISOS-D2 protocol, lifetime, degradation mechanism, hole transporting layer, buffer layer engineering 1 ABSTRACT. High power conversion efficiency (PCE) inverted organic photovoltaics (OPVs) usually use thermally evaporated MoO3 as a hole transporting layer (HTL). Despite the high PCE values reported, stability investigations are still limited and the exact degradation mechanisms of inverted OPVs using thermally evaporated MoO3 HTL remain unclear under different environmental stress factors. In this study, we monitor the accelerated lifetime performance of non-encapsulated inverted OPVs using thiophene-based active layer materials and evaporated MoO3 under the ISOS D-2 protocol (heat conditions 65 °C). The investigation of degradation mechanisms presented focus on optimized P3HT:PC[70]BM-based inverted OPVs. Specifically, we present a systematic study on the thermal stability of inverted P3HT:PC[70]BM OPVs using solution processed PEDOT:PSS and evaporated MoO3 HTL. Using a series of measurements and reverse engineering methods we report that the P3HT:PC[70]BM/MoO3 interface is the main origin of failure of P3HT:PC[70]BM-based inverted OPVs under intense heat conditions. 1. Introduction Organic photovoltaics (OPVs) have attracted great scientific interest during the last decade due to their ease of manufacture with printable techniques and their potential to become flexible, lightweight and low cost energy sources.1-2 High power conversion efficiencies and prolonged lifetimes are essential for OPV commercialization. Exciting power conversion efficiencies (PCEs) of 10% have been demonstrated,3 but long stabilities of OPVs is the next barrier that needs to be overcome. Understanding the degradation mechanisms that influence the stability of different device configurations under various environmental stress factors is still a challenging task. In the inverted structure, the use of a high work function metal such as silver results in enhanced lifetime compared to normal structured OPVs, which are limited in stability mainly due to oxidation of the low work function metals such as calcium or aluminum.4 Poly(3,4-ethylenedioxythiophene) 2 polystyrene sulfonate (PEDOT:PSS) is the most common material used as hole transporting layer (HTL) in both architectures. It is known that due to its hydroscopic nature, ingress of moisture and oxygen from the edges into the device can cause degradation.5-6 Furthermore, heat is one of the environmental factors found to significantly affect the long term stability of OPVs. Heat stability studies at operating temperature performed by Sachs-Quintana et al. on normal structure OPVs show that a thin polymer layer forms at the back (metal) contact, which creates an electron barrier between the active layer and the cathode.7 This is related to the interface between the back contact and the bulk heterojunction and is shown to be the first step in thermal degradation in organic solar cells with normal architecture and especially while using low glass transition temperature conjugated polymers. However, in inverted OPVs this barrier formation is favorable as it helps hole selectivity. This feature has been claimed as an additional advantage of inverted OPVs.7 Thus, the inverted structure is preferable concerning product development targets as it can allow more design flexibility and prolonged lifetime. However, despite its prolonged lifetime, the top electrode (the anode) of inverted OPVs has been identified as one of the most vulnerable parts of the device under various environmental stress factors. Several studies were performed that reveal the crucial influence of the interface between the active layer and HTL in the lifetime performance of inverted OPVs.8-11 Inverted OPVs with PEDOT:PSS treated with different wetting agents present different lifetime behavior in accelerated humidity conditions, thereby showing the crucial influence of processing and HTL/active layer interface formation.12 Furthermore, Norrman et al. have suggested that PEDOT:PSS is the main factor of degradation of inverted OPVs in ambient conditions. They have demonstrated that the phase separation of PEDOT:PSS into PEDOT rich and PSS rich regions and the active layer/PEDOT:PSS interface are the main sources of failure of the long-term lifetime of inverted OPVs.13 3 Therefore, the substitution of the moisture sensitive PEDOT:PSS in inverted OPVs with metal oxide HTLs such as MoO3, V2O5, and WO3 could be beneficial in some cases concerning stability of inverted OPVs. However, the use of solution processed metal oxides as HTL in inverted OPVs is still challenging, since major processing issues have to be addressed. Recent work has shown that solution-processed MoO3 can be used to produce efficient and stable normal structure and inverted OPVs when used as a sole HTL, 14-15 as well as when inserted in addition to PEDOT:PSS.16 Mixing of solution-based MoO3 with PEDOT:PSS has also yielded efficient inverted solar cells.17 However, evaporated metal oxides are still more widely used as HTLs in inverted OPVs, leading to high PCEs with optimum hole selectivity. Evaporated MoO3 is a promising material as an HTL in inverted OPVs in terms of efficiency due to the favorable energy level alignment between its work function measured at 6.7eV,18 and the HOMO of typical active layers in the range of -5.0 eV for polythiophene-based materials.19 There are promising developments regarding the stability of inverted OPVs comparing MoO3 and PEDOT:PSS HTL in air,15 however lifetime studies under humidity comparing MoO3 and PEDOT:PSS HTL show that evaporated MoO3 is also very sensitive to moisture and leads to inverted OPVs with limited lifetime.20 The reason is that oxygen and humidity ingress can alter the conductivity and work function of MoO3.21-22 On the other hand, the hygroscopic PEDOT:PSS in this architecture can act as a getter and protect the active layer from water interactions.23 Furthermore, MoO3 may undergo a change in the oxidation state of the Mo (VI) metal centre under exposure to heat or UV light, leading to changes in its work function and its optical response. Under heat conditions, MoO3 releases oxygen leading to lower Mo oxidation states and a shift in the work function of the oxide.22, 24 In addition, Voroshazi et al. have shown that the reduction of MoO3 can also occur under heat in dark conditions in inverted OPVs with MoO3/Ag/Al top contact due to chemical interactions between MoO3 and Al atoms.25 Another aspect of the degradation mechanism of inverted OPVs with the aforementioned top electrode was pointed out by Rösch et al. through different imaging 4 techniques, whereby they attributed the failure to the migration of silver and diffusion into the MoO3 layer, changing its work function.26 However, a full understanding of the stability of MoO3 HTL in inverted OPVs is still lacking due to the complexity of this material and the varying environmental stress factors that can influence the degradation cause of inverted OPVs with MoO3 HTL. In this study, therefore, we aim to analyze the degradation mechanisms of non-encapsulated inverted OPVs using MoO3 as HTL, focusing on accelerated heat lifetime conditions using the ISOS- D-2 protocol (dark – low RH – 65 °C). As a first step, inverted OPVs using various polymer:fullerene active layers are investigated. Photocurrent mapping measurements provide degradation images at 65 °C at various points in time. Secondly, different electrode configurations are investigated for our reference P3HT:PC[70]BM polymer-fullerene active layer. Using combinations of two interlayers – MoO3 and PEDOT:PSS – at the top electrode of P3HT:PC[70]BM based devices, the device degradation is attributed primarily to the interface between the active layer and MoO3 and secondly to that between MoO3 and Ag. Thirdly, alternative reverse engineering methods are used to enhance our assumption that the MoO3/Ag interface significantly contributes to the degradation of devices under heat. 2. Experimental Pre-patterned glass-ITO substrates (sheet resistance 4 Ω/sq) were purchased from Psiotec Ltd. Zinc acetate dehydrate, 2-methoxyethanol and ethanolamine were purchased from Sigma Aldrich, P3HT from Rieke Metals, PTB7 from 1-Material, PC[70]BM from Solenne BV, PEDOT:PSS PH from H.C. Stark, MoO3 powder from Sigma Aldrich and silver pellets from Kurt J. Lesker. For the fabrication of inverted solar cells, ITO substrates were sonicated in acetone and subsequently in isopropanol for 10 minutes. ZnO electron transporting layer was prepared using a sol-gel process as described in detail elsewhere.27 The ZnO precursor was doctor bladed on top of ITO substrates and annealed for 20 min at 140 °C in ambient conditions. After the annealing step a 5 40 nm ZnO layer is formed as measured with a Veeco Dektak 150 profilometer. The photo-active layers, which are blends of 36 mg/ml in chlorobenzene P3HT:PC[70]BM (1:0.8 by weight), 15 mg/ml in o-DCB DPPTTT:PC[70]BM (1:2 by weight), 25 mg/ml PTB7:PC[70]BM (1:1.5 by weight) were doctor bladed on top of ZnO resulting in a layer thickness of ~180 nm, 100 nm and 90 nm, respectively, as measured with a profilometer. For all PEDOT:PSS-based devices a treatment with two wetting agents (Zonyl and Dynol) has been applied as described in detail previously,28 and will be named as PEDOT:PSS:ZD within the text and figures. The P3HT:PC[70]BM-based inverted OPVs were annealed inside a glovebox at 140°C for 20 minutes prior to thermal evaporation of a silver layer with a thickness of 100 nm, resulting in four solar cells, each with an active area of 9 mm2. For the devices with MoO3 HTL, 10 nm of MoO3 was evaporated prior to silver evaporation. For devices using MoO3 and PEDOT:PSS:ZD as HTL, MoO3 was evaporated onto the active layer, PEDOT:PSS:ZD doctor bladed onto the MoO3 layer, followed by silver evaporation. These devices were annealed for another 2 min at 140 °C after PEDOT:PSS:ZD deposition. For devices with PEDOT:PSS:ZD/MoO3/Ag top electrode the procedure was the same as for PEDOT:PSS:ZD-based devices, except for an extra 10nm MoO3 layer evaporated on top of the PEDOT:PSS:ZD layer. The current density-voltage (J/V) characteristics were measured with a Keithley source measurement unit (SMU 2420). For illumination, a calibrated Newport Solar simulator equipped with a Xe lamp was used, providing an AM1.5G spectrum at 100mW/cm2 as measured by an Oriel 91150V calibration cell equipped with a KG5 filter. Photocurrent measurements were performed under 405 nm laser excitation using a Botest PCT photocurrent system. Thermal aging of the devices has been performed according to ISOS-D-2 protocol,29 namely, inverted OPVs under test were placed in a dark chamber (Binder) at 65 °C with controlled ambient humidity ~45% and aged for several hours. Their J/V characteristics were measured at different stages of degradation. The devices were un-encapsulated and the lifetime study was started 1 day after device fabrication. 6 3. Results and Discussion From our initial heat lifetime experiments comparing inverted OPVs comprised of ITO/ZnO/P3HT:PC[70]BM with a MoO3/Ag vs PEDOT:PSS:ZD/Ag top electrode, we observed a dramatic drop in all photovoltaic parameters after just a few hours of aging of the inverted OPVs using MoO3 as HTL (data not shown). 3.1. Investigating the degradation mechanisms of inverted OPVs under the ISOS-D-2 protocol using different polymer:fullerene blends and top electrode configurations In order to obtain a general picture about the heat degradation origins of inverted OPVs, three different polymer:fullerene blends were tested. The inverted OPVs with the following general configuration: ITO/ZnO/polymer:PC[70]BM/MoO3/Ag or PEDOT:PSS:ZD/Ag were exposed to heat conditions based on the ISOS-D-2 protocol (T = 65 °C, dark, RH = constant ambient ~40%). P3HT:PC[70]BM (1:0.8),27 PTB7:PC[70]BM (1:1.5),30 and DPPTTT:PC[70]BM (1:4)33 were used as the active layers. The polymer:fullerene ratios used were chosen from the best performing devices reported in literature. The average absolute values of the photovoltaic parameters for the inverted OPVs under study were obtained just before starting the heat aging and are summarized in Table 1. Table 1. Average absolute photovoltaic parameter values out of 8 inverted OPVs in each case, obtained before initiating the heat aging. Inverted OPVs Voc (V) Jsc (mA/cm2) FF (%) PCE (%) ITO/ZnO/P3HT:PC[70]BM/MoO3/Ag 0.56 10.42 58.9 3.54 ITO/ZnO/PTB7:PC[70]BM/MoO3/Ag 0.68 12.49 46.8 3.97 ITO/ZnO/DPPTTT:PC[70]BM/MoO3/Ag 0.55 11.22 52.9 3.32 7 ITO/ZnO/P3HT:PC[70]BM/PEDOT:PSS:ZD/Ag 0.58 9.26 58.3 3.16 ITO/ZnO/PTB7:PC[70]BM/PEDOT:PSS:ZD/Ag 0.69 8.26 50.3 2.86 ITO/ZnO/DPPTTT:PC[70]BM/PEDOT:PSS:ZD/Ag 0.54 8.07 54.3 2.53 It should be noted that while some of the device performance values presented in Table 1 may be below some literature values for the device structure, the purpose of this paper is not to produce best- performing "hero" devices. Thus the relative performance values can still be used to obtain a general picture for the degradation behavior. Figure 1 shows the normalized photovoltaic parameters over time of exposure under heat conditions of all the inverted OPVs under study. Figure 1. Degradation trends of the OPV parameters at 65 °C over time for non-encapsulated inverted OPVs with different active layers (red P3HT:PC[70]BM, green DPPTTT:PC[70]BM, blue PTB7:PC[70]BM) using either MoO3 (dashed lines) or PEDOT:PSS (solid lines) as hole transporting 8 layers, plotted as a function of (a) normalized Voc, (b) normalized Jsc, (c) normalized FF and (d) normalized PCE. As shown in Figure 1 all the inverted OPVs with different polymer:fullerene blends using MoO3/Ag top electrodes drop dramatically within the first few hours of exposure under accelerated heat conditions. In contrast, the lifetime performance of the corresponding inverted OPVs using PEDOT:PSS HTL is significantly better to that observed with MoO3 for all the different polymer:fullerene blend combinations studied. It is worth noting that all the polymers used in this study are thiophene-based (P3HT, PTB7 and DPPTTT). For PTB7:PC[70]BM-based inverted OPVs, a fast degradation pattern under 85 °C dark heat has previously been observed and been primarily attributed to morphological changes within the active layer.31 Furthermore, PTB7 has been reported as sensitive to several environmental factors,32- 33 while DPPTTT has shown greater stability under light and oxygen,34 but not under thermal conditions.35 On the other hand, P3HT:PC[70]BM-based inverted OPVs are known to be resistive to several environmental stress factors and impressive lifetime performances in accelerated and outdoor conditions are demonstrated elsewhere.36-37 All the above observations indicate that the incorporation of MoO3 as HTL at the top electrode of inverted OPVs has a dominant influence on the lifetime performance when subjected to the ISOS D- 2 protocol despite the complicated degradation patterns that might also arise due to the variety of active layers. Therefore, our further analyses in the remainder of this study focus on the understanding of the degradation mechanisms of inverted OPVs using our reference and well optimized P3HT:PC[70]BM as the active layer. 9 3.2. Investigating the degradation mechanisms of P3HT:PC[70]BM/MoO3-based inverted OPVs under the ISOS-D-2 protocol using buffer layer engineering. In order to visualize this fast degradation of P3HT:PC[70]BM-based solar cells, photocurrent images that map the device active area and its degradation are shown in Error! Reference source not found.. Inverted OPVs using P3HT:PC[70]BM as active layer and MoO3 as HTL were aged and photocurrent mapping images were extracted at different stages of degradation. The degradation of inverted OPVs with MoO3 as HTL happens very fast, even after 5 hours degradation under heat, and the photocurrent drops dramatically (see Figure 1 and Error! Reference source not found.). Figure 2. Normalized photocurrent mapping images of non-encapsulated inverted OPVs using MoO3 as HTL, showing degradation at 65 °C over time of exposure. The photocurrent mapping measurements indicate that for non-encapsulated inverted P3HT:PC[70]BM solar cells incorporating evaporated MoO3 HTL, the photocurrent generation drops dramatically in just 5 hours of exposure. The photocurrent intensity observed is significantly less within the whole device area. Despite the obvious drop near the edges of the devices which could be 10 attributed to ingress of air and moisture, the center of the device also shows significantly less photocurrent generation compared to the fresh devices. This is in accordance to Figure 1b which shows a fast Jsc degradation of P3HT:PC[70]BM/MoO3 within the first few hours of exposure. To investigate the interfacial interaction of MoO3 with the P3HT:PC[70]BM active layer, we apply buffer layer engineering to isolate the interfaces under study. Inverted OPVs based on P3HT:PC[70]BM with four different configurations of top electrode – MoO3/Ag, PEDOT:PSS:ZD/Ag, PEDOT:PSS:ZD/MoO3/Ag and MoO3/PEDOT:PSS:ZD/Ag – were tested under 65 °C. The photovoltaic parameters of these devices are shown in Table 2. Table 2. Initial photovoltaic parameters of inverted OPVs with different hole transporting layers. Inverted P3HT:PC[70]BM OPVs using different top electrode configurations Voc Jsc FF PCE (V) (mA/cm2) (%) (%) PEDOT:PSS:ZD/Ag 0.58 9.85 54.46 3.13 PEDOT:PSS:ZD/MoO3/Ag 0.57 10.03 53.65 3.08 MoO3/Ag 0.56 11.17 56.41 3.54 MoO3/PEDOT:PSS:ZD/Ag 0.55 8.52 57.20 2.72 We observe that the solar cells using MoO3/Ag top electrode show the highest initial efficiency of 3.54%. This is followed by very similar device efficiencies obtained using PEDOT:PSS:ZD/Ag and PEDOT:PSS:ZD/MoO3/Ag of 3.13% and 3.08%, respectively. Finally, 2.72% PCE is obtained for inverted OPVs with an MoO3/PEDOT:PSS:ZD/Ag top electrode. Interestingly, these values are comparable to those reported by Wang et al. using a solution-based hybrid electrode consisting of MoO3 and PEDOT:PSS,17 indicating an efficient bilayer in our system. 11 Figure 3. Degradation trends of inverted OPVs parameters at 65 °C over time for ITO/ZnO/P3HT:PC[70]BM with different top electrode configurations as a function of (a) normalized Voc, (b) normalized Jsc, (c) normalized FF and (d) normalized PCE. 12 As shown in Figure 3 and in line with the previous observations from Figure 1, inverted OPVs using MoO3/Ag top electrode degrade very fast under heat conditions from the initial efficiency of 3.54% compared to inverted OPVs with PEDOT:PSS:ZD as HTL. As also reported elsewhere,38 all the J/V parameters exponentially drop without allowing precise identification of the failure mechanism. However, by incorporating a PEDOT:PSS:ZD layer between the active layer and MoO3 as well as between MoO3 and Ag, the degradation behavior can be influenced. The inverted OPVs with MoO3/PEDOT:PSS:ZD/Ag as a top contact present an intermediate PCE decay between devices with MoO3 and PEDOT:PSS:ZD as HTL. The majority of the PCE drop on 13 those devices is attributed to FF ( Figure 3). Thus, we assume that the influence on FF in devices with MoO3/Ag top contact is attributed to the interface of MoO3 and active layer. By inserting the layer of PEDOT:PSS:ZD between MoO3 and Ag, the suppressed decay of those devices indicates that direct contact between the MoO3 and Ag influences device degradation. Furthermore, the Voc decline of devices with a direct interface between P3HT:PC[70]BM and MoO3 (organic/metal oxide interface) is more obvious than the devices with P3HT:PC[70]BM and PEDOT:PSS:ZD (organic/organic interface). The latter can again link the failure of the device to the interface of the P3HT:PC[70]BM active layer with MoO3, as has been reported previously.38 14 On the other hand inverted OPVs with PEDOT:PSS:ZD/MoO3 HTL present similar lifetime behavior to inverted OPVs with only PEDOT:PSS:ZD as HTL. In that case no obvious degradation is observed when MoO3 and Ag are in contact. Two possible phenomena could explain this. The first is that carrier selectivity occurs through PEDOT:PSS:ZD and carrier (i.e. hole) collection through MoO3/Ag , suggesting that electrons are responsible for the degradation at the MoO3 / Ag interface and that degradation is prevented when the MoO3 / Ag is protected from electrons. This effect has been seen in work by Zhu et al. on inverted hybrid quantum dot/polymer solar cells,39 in which use of a dual HTL (PEDOT:PSS/MoO3) led to enhanced electron blocking. This could explain the reduced degradation characteristics observed in our OPVs in which PEDOT:PSS interfaces with the active layer. The second might be that silver migration and diffusion in the active layer through defects of the MoO3 layer is prevented by inserting the PEDOT:PSS:ZD layer, as assumed elsewhere.26, 38 This effect of atoms diffusing into the active layer has been investigated in several works: The Wantz group has observed diffusion of silver atoms in MoO3 and organic layers, but, interestingly, they did not observe oxygen migration.40 Additionally, the Österbacka group saw that diffusion of molecules into the active layer material causes doping effects that are detrimental to device performance.41 Our findings for the MoO3/Ag top contact seem to reflect these behaviors. However, since upon insertion of the PEDOT:PSS:ZD layer in our inverted OPVs the detrimental effects are slowed down drastically, we believe this is a strong indication that indeed the silver top contact also influences these degradation effects. Therefore, in our experimental approach, our observations reveal that the interfaces between the active layer and MoO3 as well as between MoO3 and Ag majorly contribute to the degradation of the P3HT:PC[70]BM inverted OPVs under heat aging conditions, and that both electron blocking as well as diffusion of both MoO3 and Ag species may play a role in this. 15 To better understand the above observations on the degradation mechanism, the dark J/V characteristics of the different types of OPVs with different buffer layers at the initial and at different stages of degradation were obtained and are shown in Figure 4. 16 Figure 4. Current density versus voltage characteristics (J/V) in dark over time of exposure under heat conditions for inverted OPVs using different hole transporting layers – (a) PEDOT:PSS:ZD, (b) PEDOT:PSS:ZD/MoO3, (c) MoO3 and (d) MoO3/ PEDOT:PSS:ZD. The inverted OPVs containing MoO3 HTL as produced and measured in the dark (see Figure 4) have low leakage current at reverse bias implying high shunt resistance (Rp). In addition at high forward bias they have high current density and thus low series resistance (Rs) value and therefore their diode-like behavior is better than PEDOT:PSS:ZD-containing devices and accordingly they present higher PCE values, as shown in Table 1. 17 The dark J/V curves of devices with PEDOT:PSS:ZD as HTL show more stable Rs and Rp values, which are slightly improved over time ( Figure 4a). This shows that the organic-organic interface improves over time under heat conditions and that hole transportation is favored at the first stages of degradation. In addition, the performance may be assisted by formation of a hole selective thin P3HT layer on top of the active layer under heating 7. 18 Figure 4b shows the dark J/V for devices with PEDOT:PSS:ZD/MoO3 as HTL. The Rp of these devices is slightly improved over time whereas Rs is slightly increased over time. We assume that the Rp is improved due to changes in the organic/organic interface. However, we attribute the slight reduction in Rs to the interactions of the double interlayers affecting the carrier transportation in the vertical direction. Inverted OPVs using MoO3 as HTL show slightly reduced Rs over time of exposure .On the other hand, a significant leakage current and ideality factor increase over time, implying poor diode 19 properties and MoO3/active layer interactions ( Figure 4c). The increase in Rp of those devices is more intense and more obvious than the other inverted OPVs under study and it occurs even after some hours of degradation. In addition, the Rp and ideality factor issues over time of exposure are also observed when inverted OPVs using MoO3/PEDOT:PSS:ZD as HTL are tested (Figure 4d). The latter reveals the detrimental influence of the P3HT:PC[70]BM/MoO3 interface in carrier selectivity.42 This interfacial interaction over time of exposure promotes high leakage current resulting in poor Rp and ideality factors which are reflected in FF, as also shown in Figure 3. 20 3.3. Investigating the degradation mechanisms of P3HT:PC[70]BM/MoO3-based inverted OPVs under the ISOS-D-2 protocol using reverse engineering. Having seen that the inverted OPVs in which MoO3 directly interfaces with the active layer degrade the fastest, we wanted to investigate whether this interfacial interaction was reversible. We therefore fabricated incomplete device structures with ITO/ZnO/P3HT:PC[70]BM and ITO/ZnO/P3HT:PC[70]BM/MoO3 stacks. These were aged at 65 °C prior to the deposition of the subsequent fresh MoO3/Ag and Ag layers, respectively. Figure 5. (a) Illuminated and (b) dark J/V characteristics of complete devices with MoO3 HTL as produced (black rectangles) and aged for 20 hours (blue circles). The incomplete stacks were aged for 20 hours at 65 °C and then coated with the required fresh electrode. The stacks were ITO/ZnO/P3HT:PC[70]BM, which was coated with fresh MoO3/Ag (red triangles) and ITO/ZnO/P3HT:PC[70]BM/MoO3, which was coated with fresh Ag (green diamonds). 21 Figure illustrates the illuminated and dark J/V curves for devices with MoO3 HTL as produced and after aging for 20 hours at 65 °C. Also shown are the J/V characteristics for devices with degraded stacks after fresh evaporation of the top contact, enhancing the aspect that the MoO3/Ag interface also plays a role in this degradation. After reverse engineering, ,9 illuminated J/Vs show the Voc value is recovered when a new MoO3/Ag is deposited and also when only a new Ag layer is deposited on the ITO/ZnO/P3HT:PC[70]BM/MoO3 aged stack. This seems to suggest that the MoO3/Ag interface is responsible for the Voc decay in devices using MoO3/Ag top contact. 22 In addition, the rectification of devices with an active layer/MoO3 interface changes over time and is getting more symmetrical ( 23 Figure 4c, Figure 4d and Figure b). However, rectification of complete devices based on aged ITO/ZnO/P3HT:PC70BM/MoO3 upon evaporation of a fresh silver electrode, or aged ITO/ZnO/P3HT:PC70BM with fresh MoO3/Ag the ideality factor of devices with 24 ITO/ZnO/P3HT:PC70BM/MoO3 and ITO/ZnO/P3HT:PC70BM aged stacks is recovered (see Figure 5) and is similar in shape to the as-produced devices. This is an indicator that in these devices aging of the MoO3/Ag interface is responsible for the change in ideality factor.38 However, the fresh evaporation of Ag by itself does not lead to a recovery of Jsc. This happens only upon fresh evaporation of both MoO3 and Ag on the aged stack of ITO/ZnO/P3HT:PC[70]BM. This could be another indication that P3HT:PCBM/MoO3 interface is responsible for the degraded Jsc of the device over time of exposure under heat conditions. Overall, with a freshly evaporated MoO3/Ag electrode, we recover 90% of the efficiency of the as-produced devices, which is a clear indication that the device degradation is not affected by ITO/ZnO/P3HT:PCBM interfaces but almost exclusively by the interfaces of the top electrode P3HT:PCBM/MoO3/Ag as shown in Figure . It should be noted that in principle we expected to reach the same efficiency with the as- produced devices. However, we assume that the exposure to moisture in combination with the heat causes faster degradation and interface modification than the uncapped devices, thereby preventing full recovery by reverse engineering the top contact. 4. Summary We monitored the lifetime performance of P3HT:PC[70]BM, PTB7:PC[70]BM and DPPTTT:PC[70]BM inverted OPVs using thermally evaporated MoO3 as HTL under heat conditions 25 according to ISOS-D-2 protocol (T = 65 °C, dark, RH = ambient constant at ~40%). Inverted P3HT:PC[70]BM OPVs using thermally evaporated MoO3 as HTL presented the most dramatic decay of all OPV parameters under 65 °C even after a few hours of exposure. A series of experiments were performed using photocurrent mapping and different device architectures in order to isolate the interfaces and reveal the failure mechanisms of P3HT:PC[70]BM devices under accelerated heat conditions. Upon incorporation of the two compared HTL (evaporated MoO3 and solution-processed PEDOT:PSS:ZD) in different configurations in inverted OPVs with the P3HT:PC[70]BM active layer, we observed that the P3HT:PC[70]BM/MoO3 interface is the most sensitive part of this degradation triggered at 65 °C. This is primarily shown by the reduction in FF and slower degradation of inverted OPVs with MoO3/PEDOT:PSS:ZD double interlayer, which isolates the MoO3 from the Ag. This was in agreement with the dark J/Vs of inverted OPVs, which show the ideality factor of OPVs with the P3HT:PC[70]BM/MoO3 interface is affected, thereby further revealing the detrimental influence of this interface. After alternative reverse engineering method we showed that the Voc and ideality factor of such devices is recovered when fresh Ag is deposited on top of ITO/ZnO/P3HT:PC[70]BM/MoO3 aged stacks. This showed that the Ag electrode influences the stability of inverted OPVs with MoO3 as HTL under heat conditions and was confirmed by the suppressed decay of inverted OPVs using MoO3/PEDOT:PSS:ZD as HTL. To conclude, in the case of inverted P3HT:PC[70]BM OPVs with MoO3/Ag top contact, which were extensively studied in this paper by using a series of measurements and device/reverse engineering methods, the results presented indicate that inverted OPV heat degradation is not affected by the ITO/ZnO/P3HT:PCBM interfaces but almost exclusively by the interfaces of the top electrode P3HT:PCBM/MoO3/Ag. Although the interface between MoO3 and Ag contributes to degradation, the interface between the P3HT:PC[70]BM active layer and MoO3 is the main origin of failure of inverted OPVs under intense heat conditions. In this study of inverted OPVs with 26 thiophene-based active layer materials, the PEDOT:PSS:ZD HTL resulted in more stable inverted OPV performance compared to MoO3 HTL under ISOS-D-2 heat protocol. AUTHOR INFORMATION Corresponding Author *Email: [email protected]. Author Contributions The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. ACKNOWLEDGMENT Funding from the European Research Council (H2020-ERC-2014-GoG project number 647311) is gratefully acknowledged. REFERENCES 1. Hösel, M.; Søndergaard, R. R.; Angmo, D.; Krebs, F. C., Comparison of Fast Roll-to-Roll Flexographic, Inkjet, Flatbed, and Rotary Screen Printing of Metal Back Electrodes for Polymer Solar Cells. Adv. Eng. Mater. 2013, 15 (10), 995-1001. 2. Krebs, F. C., Fabrication and processing of polymer solar cells: a review of printing and coating techniques. Sol. Energy Mater. Sol. Cells 2009, 93 (4), 394-412. 3. Green, M. A.; Emery, K.; Hishikawa, Y.; Warta, W.; Dunlop, E. D., Solar cell efficiency tables (Version 45). Prog. Photovolt. Res. Appl. 2015, 23 (1), 1-9. 4. Lira-Cantu, M.; Tanenbaum, D. M.; Norrman, K.; Voroshazi, E.; Hermenau, M.; Lloyd, M. T.; Teran-Escobar, G.; Galagan, Y.; Zimmermann, B.; Hösel, M.; Dam, H. F.; Jørgensen, M.; 27 Gevorgyan, S.; Lutsen, L.; Vanderzande, D.; Hoppe, H.; Rösch, R.; Würfel, U.; Andriessen, R.; Rivaton, A.; Uzunoğlu, G. Y.; Germack, D.; Andreasen, B.; Madsen, M. V.; Bundgaard, E.; Krebs, F. C. In Combined characterization techniques to understand the stability of a variety of organic photovoltaic devices: the ISOS-3 inter-laboratory collaboration, 2012; pp 847203-847203-6. 5. Adams, J.; Salvador, M.; Lucera, L.; Langner, S.; Spyropoulos, G. D.; Fecher, F. W.; Voigt, M. M.; Dowland, S. A.; Osvet, A.; Egelhaaf, H.-J.; Brabec, C. J., Water Ingress in Encapsulated Inverted Organic Solar Cells: Correlating Infrared Imaging and Photovoltaic Performance. Adv. Energy Mater. 2015, 5 (20), 1501065. 6. Glen, T. S.; Scarratt, N. W.; Yi, H.; Iraqi, A.; Wang, T.; Kingsley, J.; Buckley, A. R.; Lidzey, D. G.; Donald, A. M., Dependence on material choice of degradation of organic solar cells following exposure to humid air. J. Polym. Sci. B 2015, 54 (2), 216-224. 7. Sachs-Quintana, I. T.; Heumüller, T.; Mateker, W. R.; Orozco, D. E.; Cheacharoen, R.; Sweetnam, S.; Brabec, C. J.; McGehee, M. D., Electron Barrier Formation at the Organic-Back Contact Interface is the First Step in Thermal Degradation of Polymer Solar Cells. Adv. Funct. Mater. 2014, 24 (25), 3978-3985. 8. Kawano, K.; Pacios, R.; Poplavskyy, D.; Nelson, J.; Bradley, D. D. C.; Durrant, J. R., Degradation of organic solar cells due to air exposure. Sol. Energy Mater. Sol. Cells 2006, 90 (20), 3520-3530. 9. Drakonakis, V. M.; Savva, A.; Kokonou, M.; Choulis, S. A., Investigating electrodes degradation in organic photovoltaics through reverse engineering under accelerated humidity lifetime conditions. Sol. Energy Mater. Sol. Cells 2014, 130 (C), 544-550. 10. Lloyd, M. T.; Peters, C. H.; Garcia, A.; Kauvar, I. V.; Berry, J. J.; Reese, M. O.; McGehee, M. D.; Ginley, D. S.; Olson, D. C., Influence of the hole-transport layer on the initial behavior and lifetime of inverted organic photovoltaics. Solar Energy Materials and Solar Cells 2011, 95 (5), 1382-1388. 28 11. Lloyd, M. T.; Olson, D. C.; Lu, P.; Fang, E.; Moore, D. L.; White, M. S.; Reese, M. O.; Ginley, D. S.; Hsu, J. W. P., Impact of contact evolution on the shelf life of organic solar cells. Journal of Materials Chemistry 2009, 19 (41), 7638-7642. 12. Savva, A.; Georgiou, E.; Papazoglou, G.; Chrusou, A. Z.; Kapnisis, K.; Choulis, S. A., Photovoltaic analysis of the effects of PEDOT:PSS-additives hole selective contacts on the efficiency and lifetime performance of inverted organic solar cells. Solar Energy Materials and Solar Cells 2015, 132 (0), 507-514. 13. Norrman, K.; Madsen, M. V.; Gevorgyan, S. A.; Krebs, F. C., Degradation Patterns in Water and Oxygen of an Inverted Polymer Solar Cell. J. Am. Chem. Soc. 2010, 132 (47), 16883-16892. 14. Zilberberg, K.; Gharbi, H.; Behrendt, A.; Trost, S.; Riedl, T., Low-Temperature, Solution- Processed MoOx for Efficient and Stable Organic Solar Cells. ACS Appl. Mater. Interfaces 2012, 4 (3), 1164-1168. 15. Qiu, W.; Muller, R.; Voroshazi, E.; Conings, B.; Carleer, R.; Boyen, H.-G.; Turbiez, M.; Froyen, L.; Heremans, P.; Hadipour, A., Nafion-Modified MoOx as Effective Room-Temperature Hole Injection Layer for Stable, High-Performance Inverted Organic Solar Cells. ACS Appl. Mater. Interfaces 2015, 7 (6), 3581-3589. 16. Kim, J.-H.; Park, E.-K.; Kim, J.-H.; Cho, H. J.; Lee, D.-H.; Kim, Y.-S., Improving charge transport of P3HT:PCBM organic solar cell using MoO3 nanoparticles as an interfacial buffer layer. Electronic Materials Letters 2016, 12 (3), 383-387. 17. Wang, Y.; Luo, Q.; Wu, N.; Wang, Q.; Zhu, H.; Chen, L.; Li, Y.-Q.; Luo, L.; Ma, C.-Q., Solution-Processed MoO3:PEDOT:PSS Hybrid Hole Transporting Layer for Inverted Polymer Solar Cells. ACS Appl. Mater. Interfaces 2015, 7 (13), 7170-7179. 18. Hammond, S. R.; Meyer, J.; Widjonarko, N. E.; Ndione, P. F.; Sigdel, A. K.; Garcia, A.; Miedaner, A.; Lloyd, M. T.; Kahn, A.; Ginley, D. S.; Berry, J. J.; Olson, D. C., Low-temperature, 29 solution-processed molybdenum oxide hole-collection layer for organic photovoltaics. J. Mater. Chem. 2012, 22 (7), 3249-3254. 19. Park, S.; Jeong, J.; Hyun, G.; Kim, M.; Lee, H.; Yi, Y., The origin of high PCE in PTB7 based photovoltaics: proper charge neutrality level and free energy of charge separation at PTB7/PC71BM interface. Scientific Reports 2016, 6, 35262. 20. Weerasinghe, H. C.; Watkins, S. E.; Duffy, N.; Jones, D. J.; Scully, A. D., Influence of moisture out-gassing from encapsulant materials on the lifetime of organic solar cells. Solar Energy Materials and Solar Cells 2015, 132, 485-491. 21. Irfan; Ding, H.; Gao, Y.; Small, C.; Kim, D. Y.; Subbiah, J.; So, F., Energy level evolution of air and oxygen exposed molybdenum trioxide films. Applied Physics Letters 2010, 96 (24), 243307. 22. Griffin, J.; Watters, D. C.; Yi, H.; Iraqi, A.; Lidzey, D.; Buckley, A. R., The Influence of MoOx Anode Stoicheometry on the Performance of Bulk Heterojunction Polymer Solar Cells. Advanced Energy Materials 2013, 3 (7), 903-908. 23. Voroshazi, E.; Cardinaletti, I.; Uytterhoeven, G.; Li, S.; Empl, M.; Aernouts, T.; Rand, B. P., Role of electron-and hole-collecting buffer layers on the stability of inverted polymer: Fullerene photovoltaic devices. IEEE Journal of Photovoltaics 2014, 4 (1), 265-270. 24. Zheng, Z.; Yan, X.; Huai-Xin, W.; Guo-Fu, M.; Steffen, D.; Yan-Qing, L.; Jian-Xin, T., Impact of Oxygen Vacancy on Energy-Level Alignment at MoO x /Organic Interfaces. Applied Physics Express 2013, 6 (9), 095701. 25. Voroshazi, E.; Uytterhoeven, G.; Cnops, K.; Conard, T.; Favia, P.; Bender, H.; Muller, R.; Cheyns, D., Root-cause failure analysis of photocurrent loss in polythiophene:Fullerene-based inverted solar cells. ACS Applied Materials and Interfaces 2015, 7 (1), 618-623. 26. Rösch, R.; Tanenbaum, D. M.; Jørgensen, M.; Seeland, M.; Bärenklau, M.; Hermenau, M.; Voroshazi, E.; Lloyd, M. T.; Galagan, Y.; Zimmermann, B.; Würfel, U.; Hösel, M.; Dam, H. F.; Gevorgyan, S. A.; Kudret, S.; Maes, W.; Lutsen, L.; Vanderzande, D.; Andriessen, R.; Teran- 30 Escobar, G.; Lira-Cantu, M.; Rivaton, A.; Uzunoğlu, G. Y.; Germack, D.; Andreasen, B.; Madsen, M. V.; Norrman, K.; Hoppe, H.; Krebs, F. C., Investigation of the degradation mechanisms of a variety of organic photovoltaic devices by combination of imaging techniques-the ISOS-3 inter- laboratory collaboration. Energy Environ. Sci. 2012, 5 (4), 6521. 27. Savva, A.; Choulis, S. A., Cesium-doped zinc oxide as electron selective contact in inverted organic photovoltaics. Appl. Phys. Lett. 2013, 102 (23), 233301. 28. Savva, A.; Neophytou, M.; Koutsides, C.; Kalli, K.; Choulis, S. A., Synergistic effects of buffer layer processing additives for enhanced hole carrier selectivity in inverted Organic Photovoltaics. Org. Electron. 2013, 14 (11), 3123-3130. 29. Reese, M. O.; Gevorgyan, S. A.; Jørgensen, M.; Bundgaard, E.; Kurtz, S. R.; Ginley, D. S.; Olson, D. C.; Lloyd, M. T.; Morvillo, P.; Katz, E. A.; Elschner, A.; Haillant, O.; Currier, T. R.; Shrotriya, V.; Hermenau, M.; Riede, M.; Kirov, K. R.; Trimmel, G.; Rath, T.; Inganäs, O.; Zhang, F.; Andersson, M.; Tvingstedt, K.; Lira-Cantu, M.; Laird, D.; McGuiness, C.; Gowrisanker, S. J.; Pannone, M.; Xiao, M.; hauch, J.; Steim, R.; DeLongchamp, D. M.; Rösch, R.; Hoppe, H.; Espinosa, N.; Urbina, A.; Yaman-Uzunoglu, G.; Bonekamp, J.-B.; van Breemen, A. J. J. M.; Girotto, C.; Voroshazi, E.; Krebs, F. C., Consensus stability testing protocols for organic photovoltaic materials and devices. Sol. Energy Mater. Sol. Cells 2011, 95 (5), 1253-1267. 30. He, Z.; Zhong, C.; Su, S.; Xu, M.; Wu, H.; Cao, Y., Enhanced power-conversion efficiency in polymer solar cells using an inverted device structure. Nat. Photon. 2012, 6 (9), 593-597. 31. Tessarolo, M.; Guerrero, A.; Gedefaw, D.; Bolognesi, M.; Prosa, M.; Xu, X.; Mansour, M.; Wang, E.; Seri, M.; Andersson, M. R.; Muccini, M.; Garcia-Belmonte, G., Predicting thermal stability of organic solar cells through an easy and fast capacitance measurement. Sol. Energy Mater. Sol. Cells 2015, 141 (C), 240-247. 31 32. Razzell-Hollis, J.; Wade, J.; Tsoi, W. C.; Soon, Y.; Durrant, J.; Kim, J.-S., Photochemical stability of high efficiency PTB7:PC70BM solar cell blends. J. Mater. Chem. A 2014, 2, 20189- 20195. 33. Wong, H. C.; Li, Z.; Tan, C. H.; Zhong, H.; Huang, Z.; Bronstein, H.; McCulloch, I.; Cabral, J. T.; Durrant, J. R., Morphological Stability and Performance of Polymer–Fullerene Solar Cells under Thermal Stress: The Impact of Photoinduced PC 60BM Oligomerization. ACS Nano 2014, 8 (2), 1297-1308. 34. Soon, Y. W.; Cho, H.; Low, J.; Bronstein, H.; McCulloch, I.; Durrant, J. R., Correlating triplet yield, singlet oxygen generation and photochemical stability in polymer/fullerene blend films. Chem. Commun. 2013, 49 (13), 1291. 35. Huang, Z.; Fregoso, E. C.; Dimitrov, S.; Tuladhar, P. S.; Soon, Y. W.; Bronstein, H.; Meager, I.; Zhang, W.; McCulloch, I.; Durrant, J. R., Optimisation of diketopyrrolopyrrole:fullerene solar cell performance through control of polymer molecular weight and thermal annealing. J. Mater. Chem. A 2014, 2, 19282-19289. 36. Hauch, J. A.; Schilinsky, P.; Choulis, S. A.; Childers, R.; Biele, M.; Brabec, C. J., Flexible organic P3HT: PCBM bulk-heterojunction modules with more than 1 year outdoor lifetime. Sol. Energy Mater. Sol. Cells 2008, 92 (7), 727-731. 37. Sapkota, S. B.; Spies, A.; Zimmermann, B.; Dürr, I.; Würfel, U., Promising long-term stability of encapsulated ITO-free bulk-heterojunction organic solar cells under different aging conditions. Sol. Energy Mater. Sol. Cells 2014, 130 (C), 144-150. 38. Greenbank, W.; Hirsch, L.; Wantz, G.; Chambon, S., Interfacial thermal degradation in inverted organic solar cells. Appl. Phys. Lett. 2015, 107 (26), 263301. 39. Zhu, L.; Richardson, B. J.; Yu, Q., Inverted hybrid CdSe-polymer solar cells adopting PEDOT:PSS/MoO3 as dual hole transport layers. Physical Chemistry Chemical Physics 2016, 18 (5), 3463-3471. 32 40. Chambon, S.; Derue, L.; Lahaye, M.; Pavageau, B.; Hirsch, L.; Wantz, G., MoO3 Thickness, Thermal Annealing and Solvent Annealing Effects on Inverted and Direct Polymer Photovoltaic Solar Cells. Materials 2012, 5 (12). 41. Nyman, M.; Dahlström, S.; Sandberg, O. J.; Österbacka, R., Unintentional Bulk Doping of Polymer-Fullerene Blends from a Thin Interfacial Layer of MoO3. Advanced Energy Materials 2016, 6 (15), 1600670-n/a. 42. Lee, K. E.; Liu, L.; Kelly, T. L., Effect of Molybdenum Oxide Electronic Structure on Organic Photovoltaic Device Performance: An X-ray Absorption Spectroscopy Study. J. Phys. Chem. C 2014, 118 (48), 27735-27741. 33
1802.07941	1	1802	2018-02-22T08:39:34	Strain effect in highly-doped n-type 3C-SiC-on-glass substrate for mechanical sensors and mobility enhancement	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	This work reports the strain effect on the electrical properties of highly doped n-type single crystalline cubic silicon carbide (3C-SiC) transferred onto a 6-inch glass substrate employing an anodic bonding technique. The experimental data shows high gauge factors of -8.6 in longitudinal direction and 10.5 in transverse direction along the [100] orientation. The piezoresistive effect in the highly doped 3C-SiC film also exhibits an excellent linearity and consistent reproducibility after several bending cycles. The experimental result was in good agreement with the theoretical analysis based on the phenomenon of electron transfer between many valleys in the conduction band of n-type 3C-SiC. Our finding for the large gauge factor in n-type 3C- SiC coupled with the elimination of the current leak to the insulated substrate could pave the way for the development of single crystal SiC-on-glass based MEMS applications.	physics.app-ph	physics	Strain effect in highly-doped n-type 3C-SiC-on-glass substrate for mechanical sensors and mobility enhancement Hoang-Phuong Phan,1, 2, a) Tuan-Khoa Nguyen,1 Toan Dinh,1 Han-Hao Cheng,3 Fengwen Mu,4 Alan Iacopi,1 Leonie Hold,1 Tadatomo Suga,4 Dzung Viet Dao,1, 5 Debbie G. Senesky,2, 6 and Nam-Trung Nguyen1 1)Queensland Micro- and Nanotechnology Centre, Griffith University, QlD, Australia 2)Aeronautics and Astronautics Department, Stanford University, CA, USA 3)Centre for Microscopy and Microanalysis, University of Queensland, QlD, Australia 4)Department of Precision Engineering, The University of Tokyo, Tokyo, Japan 5)School of Engineering, Griffith University, Qld, Australia 6)Department of Electrical Engineering, Stanford University, CA, USA. (Dated: 13 October 2018) This work reports the strain effect on the electrical properties of highly doped n-type single crystalline cubic silicon carbide (3C-SiC) transferred onto a 6-inch glass substrate employing an anodic bonding technique. The experimental data shows high gauge factors of -8.6 in longitudinal direction and 10.5 in transverse direction along the [100] orientation. The piezoresistive effect in the highly doped 3C-SiC film also exhibits an excellent linearity and consistent reproducibility after several bending cycles. The experimental result was in good agreement with the theoretical analysis based on the phenomenon of electron transfer between many valleys in the conduction band of n-type 3C-SiC. Our finding for the large gauge factor in n-type 3C- SiC coupled with the elimination of the current leak to the insulated substrate could pave the way for the development of single crystal SiC-on-glass based MEMS applications. The demand for electronics and sensors which can operate in extreme conditions has propelled the re- search into wide band gap materials such as III-nitride (e.g. GaN, AlN), silicon carbide (SiC) and diamond- like carbon (DLC)1 -- 4. Among these materials, SiC has emerged as an excellent semiconductor for Micro Elec- tro Mechanical Systems (MEMS) applications owing to its large energy gap, varying from 2.3 eV to 3.4 eV, ex- cellent chemical inertness, superior mechanical proper- ties, and outstanding radiation tolerance5,6. Numerous SiC technologies have been transitioned into commercial products, including bipolar transistors, Schottky diodes, field effect transistors, ultraviolet photodetectors and gas sensors7 -- 10. Although great progress has been made to commercial- ize SiC materials and devices, the material cost is still relatively expensive in comparison to the Si counterpart. One promising approach to bring down the cost of SiC materials is the development of epitaxial SiC nanothin films on a Si substrate11,12. This approach can utilize the low cost and worldwide availability of Si substrate, as well as take advantage of Si-orientated MEMS fabri- cation technologies. To date, high quality and large scale SiC wafers grown on Si have been reported13. Appli- cations of SiC-on-Si have also been demonstrated such as mechanical sensors, chemical sensors, bio devices14 -- 16. Nevertheless, when subjected to high temperature, the current leakage from SiC to the Si substrate could ad- versely affect the performance of the SiC sensing element. Therefore, a number of efforts have been made to transfer SiC films from SiC/Si wafers onto electrically insulating a)Electronic mail: [email protected] substrates17. Most of the works reported in the liter- ature are based on SiO2 to SiO2 direct bonding or Si to poly Si diffusion bonding techniques to form an ox- ide layer sandwiched between SiC and Si18 -- 20. However, these methods require an ultra-smooth buffer layer such as SiO2, formed by sputtering/oxidation and polishing or SiO2/poly-Si which typically complicates the process. Furthermore, the Si layer in the bonded wafers generally absorbs visible wavelengths, hindering the merit of high- optical-transmittance in bulk SiC materials. Employing the anodic bonding method, we have recently demon- strated large scale single crystal 3C-SiC films transferred onto a glass substrate21. The bonded wafers possess ro- bust adhesion strength, high visible-wavelength transmit- tance, and excellent electrical insulation to the substrate. In addition, the transferred SiC nanothin films have also been proven to be bio-compatible which enabled the di- rect culture of bio-cells. Furthermore, the large scale of 6-inch in the diameter also makes the single crystal 3C- SiC/glass wafers a promising platform to develop low- cost mass production devices for a wide range of MEMS applications. This work elucidates the strain effect on the electrical conductance of highly doped n-type 3C-SiC nano films bonded onto glass through experimental characteriza- tions and the first principle calculation. The experimen- tal results show a negative gauge factor of -8.6, which is in solid agreement with theoretical analysis based on the many valleys electron transfer phenomenon. The in- sight obtained from this study will be the foundation for the development of SiC/glass mechanical sensing appli- cations as well as opening up opportunities for further investigation into strain engineering in SiC. To conduct this experiment, single-crystalline 3C-SiC thin films were grown on Si substrates using a low pres- 8 1 0 2 b e F 2 2 ] h p - p p a . s c i s y h p [ 1 v 1 4 9 7 0 . 2 0 8 1 : v i X r a 2 FIG. 1. Fabrication of SiC/glass piezoresistors. (a) The concept of anodic bonding process and lithography on the SiC/glass platform; (b) Photograph of a full 6-inch SiC/glass wafer; (c) Photograph of a SiC/glass cantilever used in the bending experiment (In set: Microscope image of SiC piezoresistors aligned in longitudinal and transverse [100] orientations) sure chemical vapor deposition process (LPCVD)22,23. Silane (SiH4) and propene (C3H6) were employed as the precursors in the hot-wall furnace at 1250◦C. Addition- ally, ammonia (NH3) was utilized as the in situ dopant, forming highly-doped n-type 3C-SiC. The thickness of the initial 3C-SiC film was 700 nm, measured using a NANOMETRICS TMNanospec/AFT 210. The film was further smoothed by etching back the SiC layer to a thickness of 600 nm to improve the contact surface be- tween SiC and glass during wafer bonding. Prior to the bonding process, piranha cleaning and oxygen plasma was performed to remove contamination from the sur- face of glass and SiC/Si wafers. The smoothed 6-inch SiC/Si wafers were then subjected to the anodic bonding process, performed using an EVGTM520IS hot embosser under a compressive force between 1.5 and 2.5 kN and bias voltage between -200 to -1000 V at a temperature of 400◦C. Subsequently, the top Si layer with a thickness of 650 µm was completely removed in KOH at 80◦C. Next, a two-mask fabrication process was carried out to fabricate SiC resistors using Inductive Coupled Plasma with an etching rate of 100nm/min. Finally, the SiC on glass samples was diced into cantilevers with dimensions of 0.51 mm×3.5 mm×20 mm for the subsequent bending experiment, as illustrated in Fig 1(a). Figure 1(b) shows a 6-inch 3C-SiC/glass wafer, ex- hibiting excellent transparency. The optical character- ization and the bonding strength of the films can be found elsewhere21. The large scale of the bonded wafers could significantly reduce the material cost as well as allow mass production of SiC/glass based devices by em- ploying a batch fabrication process. Figure 1(c) shows a photograph of SiC/glass cantilever for the bending ex- periment, in which the U-shaped SiC piezoresistors with dimensions of 8 µm×200 µm×600 nm were patterned in the vicinity of the fixed end of the cantilever in order to obtain high tensile strain. In this work, we employed the SiC piezoresistors aligned along the longitudinal and transverse [100] directions to investigate the strain effect on electrical conductance, Fig. 1(c) inset. The carrier concentration of the bonded 3C-SiC films was found to be approximately 1019 cm−3, using a hot probe technique. The current-voltage (I-V) char- acteristic of the SiC piezoresistors was measured using AgilentTMB1500A, indicating that a good Ohmic con- tact was formed between Ni and n-type 3C-SiC. In ad- dition, the similarity in the resistances of longitudinal and transverse SiC piezoresistors also indicates the uni- formity of the carrier concentration in the highly doped film. Furthermore, as the SiC was transferred onto glass, no current leakage was observed from the sensing layer to the substrate, as illustrated in Fig. 2. The bending beam method was applied to induce strain into the SiC piezoresistors, in which a SiC/glass FIG. 2. Current-Voltage (I-V) characteristic of the SiC piezoresistors. The current leakage was found to be in pA order, which is six orders of magnitude smaller than the cur- rent of the SiC resistors. 3 FIG. 3. Characterization of the piezoresistive effect in n-type 3C-SiC on glass using the bending beam method. The change in the current of the longitudinal resistor (a) and (b); and of the transverse resistor (c) and (d) under strain and at a constant applied voltage of 1 V. The spikes were caused by the uncertainty of the experimental setup during loading/unloading forces. cantilever was fixed, while the other free end was de- flected downward by external forces. In addition, to avoid the effect of the boundary condition in the bend- ing test, the clamped area was placed at least 2 mm far from the piezoresistors. The applied strain was estimated based on finite element analysis (FEA) using COMSOL MultiphysicsTM). Accordingly, for the longitudinal re- sistance, nearly 100% of the strain applied to the top surface of the glass cantilever was transmitted to the SiC layer. On the other hand, for the transverse resistor, due to the small ratio of width to length of approximately 1:11, only 48% of applied strain was transmitted to the SiC layer (See supplementary online information). This phenomenon of small strains induced into transverse re- sistors has also been reported in the literature, which is more profound for the case of nanowires with extremely small width to length ratio24. Based on the FEA, when varying the applied force from 0 to 192 mN, the strains induced into longitudinal and transverse SiC piezoresis- tors were estimated to vary from 0 to 310 ppm, and from 0 to 140 ppm, respectively. The response of the n-type 3C-SiC piezoresistors under applied strain was then investigated by monitoring the change in the output current under a constant applied voltage of 1 V, as shown in Fig. 3(a). Accordingly, when increasing the tensile strain, the current passing through the SiC piezoresistors also increased. This characteris- tic in n-type 3C-SiC is different from p-type 3C-SiC as well as metals where the current decreases under tensile strain5,10. The increase in the current in the n-type 3C- SiC also indicates the possibility to improve the mobility of n-type 3C-SiC on glass based devices by employing mechanical tensile strain. The change in the current also exhibits excellent repeatability after multiple bending cy- cles under an applied force of 192mN. Figures 3(c)(d) plot the resistance change of the transverse resistor un- der applied loads, showing the opposite trend to that of the longitudinal resistor (i.e. resistance increased when increasing the tensile strain). Furthermore, it should be pointed out that under the same applied force to the SiC/Si cantilever, the change of the transverse resistance was relatively small in comparison to that of longitudinal resistance. This result is considerable due to the fact that a smaller strain was induced into the transverse resistor as described above. ε Figure 4 shows the resistance change of the n-type 3C- SiC on glass against applied strain to obtain the gauge factor of the material (GF = ∆R/R ). A linear relation- ship between the resistance change and tensile strain was observed in both longitudinal and transverse resistors. This linearity is a favorable property for strain sensing applications. Based on the results from Fig. 4, the lon- gitudinal and transverse gauge factors of n-type 3C-SiC were found to be -8.6 and 10.5, respectively, which are at least 5 times larger than those of metals (See supple- mentary online information). 4 in [100] orientation, the energy of the valley located in [100] axis will shift up while the energies in the other val- leys in [010] and [001] axes will shift down an amount of ∆E = Ξuε. Here, Ξu is an independent constant of the deformation energy. As a result, the change in electron concentration in each valley is: ∆n = Nc × {F1/2( ≈ Nc × ±∆E kBT × F−1/2( Ec ± ∆E − EF ) − F1/2( Ec − EF kBT )} kBT Ec − EF kBT ) (2) where F−1/2 = δF1/2/δE. Consequently, based on the electron transfer phenomenon, the gauge factor of n-type 3C-SiC is given by31 -- 33: FIG. 4. The relationship between resistance change in the n- type 3C-SiC resistances and applied strain. Inset: Schematic sketch of the electron transfer phenomenon. Employing the theory of strained n-type cubic semi- conductors, we qualitatively explain the piezoresistance of n-type 3C-SiC on glass based on the electron transfer phenomenon25,26. That is, under applied strain, the con- duction bands of the n-type 3C-SiC are deformed, lead- ing to the re-population of electrons, following the Boltz- mann distribution. According to the first principle analy- sis of Nakamura et al. on a n-type 3C-SiC nanosheet27,28, under a tensile strain in [100] orientation, the energy val- ley in the [100] (i.e. the longitudinal valley) will shift up, while the energy valleys in [010] and [001] directions (i.e. the transverse valleys) will shift down. These shifts in the 6 energy valleys result in the re-population of free electrons, in which the electrons from the longitudinal direction (i.e. [100] orientation) will fill up the valley in transverse direction (i.e. [010] and [001] orientations). Moreover, since the transverse mobility of electron (µ⊥) is higher than that in the longitudinal direction (µ(cid:107)), more electrons in the [010] and [001] direction leads to a higher effective mobility. This hypothesis is consistent with our results that the tensile strain increases the con- ductivity of the n-type 3C-SiC, indicating the enhance- ment of electron mobility with the assumption that the total number of electrons remain constant under strain for highly doped semiconductors29. Next, we utilize the theoretical model of Kanda devel- oped for Si -- another cubic single crystal30 to numerically estimate the gauge factor in the highly doped 3C-SiC. For strain-free n-type 3C-SiC, the the number of electrons in each equivalent valley is given by: n0 = Nc × F1/2( Ec − EF kBT ) (1) where Nc is the effective density of state in the conduc- tance band; F1/2 is the Fermi-Diract integral; Ec and EF are the energy of each valley in the conduction band and the Fermi level; kB is the Boltzmann constant; and T is the absolute temperature. Under applied strain ε G = 1 + 2ν − Ξu(L − 1) 3kBT (2L + 1) × (2 + ν(cid:48)) × F−1/2F1/2 (3) where L = µ⊥/µ(cid:107) is the ratio of the transverse and longitudinal electron mobilities, and ν and ν(cid:48) are the Poisson ratio of 3C-SiC and the substrate, respectively. The Fermi-Dirac integral was calculated based on Chang- Elizabeth approximation34, while the deformation energy Ξu = 6.3eV was obtained based on the theoretical work of Lambrecht et al. on full potential band calculation35. Additionally, the mobility ratio L = 2.74 was obtained from the experimental work reported in36. Substituting these values into Eq. 3, the gauge factor in degeneracy doped 3C-SiC with carrier concentration of 5×1019 cm−3 to 1020 cm−3 varies from -8 to -14 which is consistent with our experimentally measured value. This result indicates that the electron transfer mechanism can qualitatively and quantitatively explain the piezoresistive effect in the highly doped 3C-SiC. In conclusion, this work reports the piezoresistive effect in highly doped 3C-SiC nanothin film transferred onto a 6-inch glass substrate, employing anodic bonding tech- nique. Experimental data shows a negative gauge factor of approximately -8.6 in the longitudinal [100] direction of the n-type 3C-SiC, which is in solid agreement with nu- merical analysis based on electron transfer effect in cubic materials under stress. The large gauge factors, superior mechanical properties, excellent chemical inertness, and bio-compatibility of the transferred 3C-SiC on glass could open up promising opportunities for MEMS applications in harsh environments, as well as bio-sensing. This work was partially funded by the linkage grants LP150100153 and LP160101553 from the Australian Re- search Council (ARC). This work was performed in part at the Queensland node of the Australian National Fabri- cation Facility, a company established under the National Collaborative Research Infrastructure Strategy to pro- vide nano and micro-fabrication facilities for Australia's researchers. H.P. Phan acknowledges research grants from the Australian Nanotechnology Network Overseas Travel Fellowship and Griffith University Postdoctoral Fellowship (GUPF). 1Senesky, D. G.; Jamshidi, B.; Cheng, K. B.; Pisano, A. P. IEEE Sens. J. 2009, 9, 14721478. 2Eddy, C. R.; Gaskill, D. K. Science 2009, 324, 13981400. 3R. S. Okojie, A. A. Ned, A. D. Kurtz, and W. N. Carr, IEEE Trans. Electron Devices 45(4), 785, 1998. 4Wu, Y., Lin, Y.M., Bol, A.A., Jenkins, K.A., Xia, F., Farmer, D.B., Zhu, Y. and Avouris, P., 2011. Nature, 472(7341), pp.74- 78. 5H.-P. Phan, Piezoresistive effect of p-type single crystalline SiC, Springer, 2017. ISBN: 978-3-319-55543-0. 6T. Dinh, H.-P. Phan, A. Qamar, P. Woodfield, N. T. Nguyen, and D.V. Dao, JMEMS, 26(5), 966-986, 2017. 7Rabkowski, J.; Peftitsis, D.; Nee, H. P. IEEE Ind. Electron. Mag. 2012, 6, 1726. 8Nakamura, D.; Gunjishima, I.; Yamaguchi, S.; Ito, T.; Okamoto, A.; Kondo, H.; Onda, S.; Takatori, K. Nature 2004, 430, 10091012. 9Wright, N. G.; Horsfall, A. B.; Vassilevski, K. Mater. Today, 2008, 11, 1621. 10H.-P. Phan, T. Dinh, T. Kozeki, T.-K. Nguyen, A. Qamar, T. Namazu, N.-T. Nguyen and D. V. Dao, Appl. Phys. Lett., 109, p. 123502, 2016. 11H.-P. Phan, T.-K. Nguyen, T. Dinh, G. Ina, A. R. Kermany, A. Qamar, J. Han, T. Namazu, R. Maeda, D. V. Dao, and N.-T. Nguyen, Appl. Phys. Letts., 110, 141906, 2017. 12Hens, P., Zakharov, A.A., Iakimov, T., Syvajarvi, M. and Yaki- mova, R., 2014. Carbon, 80, pp.823-829. 13Saddow, S. E. Elsevier 2012, 1. 14Yang, T.; Zhang, L.; Hou, X.; Chen, J.; Chou, K. C. Sci. Rep. 2016, 6, 24872. 15Naglieri, V.; Gludovatz, B.; Tomsia, A. P.; Ritchie, R. O. Acta Mater. 2015, 98, 141151. 16J. Bi, G. Wei, L. Wang, F. Gao, J. Zheng, B. Tang, and W. Yang, J. Mater. Chem. C. 1, 4514 (2013). 17Serre, C.; Romano-Rodriguez, A.; Perez-Rodriguez, A.; Morante, J. R.; Fonseca, L.; Acero, M. C.; Kogler, R.; Skorupa, W. Sens. Actuators, A 1999, 74, 169173. 18Suga, T.; Mu, F.; Fujino, M.; Takahashi, Y.; Nakazawa, H.; Iguchi, K. Jpn. J. Appl. Phys. 2015, 54, 030214. 5 19Wu, C.-H.; Zorman, C. A.; Mehregany, M. IEEE Sens. J. 2006, 6, 316. Sensors for High Temperature Applications. 20Vinod, K. N.; Zorman, A. A.; Yasseen, A. A.; Mehregany, M. J. Electron. Mater. 1998, 27, L17L20. Wafer Bonding Techniques. 21Phan, H.P., Cheng, H.H., Dinh, T., Wood, B., Nguyen, T.K., Mu, F., Kamble, H., Vadivelu, R., Walker, G., Hold, L. Iacopi, A., D.V. Dao, T. Suga, and N.-T. Nguyen, 2017, 9(33), pp.27365- 27371. 22Phan, H.P., Qamar, A., Dao, D.V., Dinh, T., Wang, L., Han, J., Tanner, P., Dimitrijev, S. and Nguyen, N.T., 2015. RSC Ad- vances, 5(69), pp.56377-56381. 23D.V. Dao, H.-P. Phan, A. Qamar, T. Dinh, RSC Advances, 6 (26), 21302-21307, 2015. 24Toriyama, T., Tanimoto, Y. and Sugiyama, S. 11(5), pp.605-611, 2002. 25J. C. Doll and B. L. Pruitt, Piezoresistor Design and Applica- tions, 1st ed. New York, NY, USA: Springer, 2013. 26G. L. Bir and G. E. Pikus, Symmetry and Strain-Induced Effects in Semiconductors. Hoboken, NJ, USA: Wiley, Jan. 1974. 27K. Nakamura, T. Toriyama, and S. Sugiyama, in Proc. 27th Sen- sor Symp. Sensors, Micromach., Appl.Syst., Oct. 2010, pp. 16. 28K. Nakamura, T. Toriyama, and S. Sugiyama, Jpn. J. Appl. Phys., vol. 50, no. 6S, p. 06GE05, 2011. 29Y. Sun, S. E. Thompson, and T. Nishida, Strain Effect in Semi- conductors: Theory and Device Applications. 1st ed. New York, NY, USA: Springer, 2009. 30Y. Kanda, Sens. Actuators A, Phys., vol. 28, no. 2, pp. 8391, Jul. 1991. 31Y. Kanda and Y. Kanda, IEEE Trans. Electron Devices, vol. 29, no. 1, pp. 6470, Jan. 1982. 32T. Toriyama, S. Sugiyama, Appl. Phys. Lett., 81, 2797, 2002. 33HP Phan, DV Dao, K Nakamura, S Dimitrijev, NT Nguyen, J. Microelectromech. Syst., 24 (6), 1663-1677, 2015. 34C. Y. Chang and A. Izabelle, J. Appl. Phys. 65, 2162, 1989. 35W. R. L. Lambrecht, B. Segall, M. Methfessel, and M. van Schil- fessel, Phys. Rev. B 44, 3685-1991. 36R. Kaplan, R. J. Wagner, H. J. Kim, and R. F. Davis, Solid State Commun. 55, 67, 1985.
1708.06744	1	1708	2017-07-23T07:52:21	Graphene/SiO2 nanocomposites: The enhancement of photocatalytic and biomedical activity of SiO2 nanoparticles by graphene	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	The exceptional conducting nature of graphene makes it a viable candidate for enhancing the effectiveness of photocatalytic and biomedical nanomaterials. Herein, the immobilization of monodispersed silicon dioxide (SiO2) nanoparticles on multiple graphene layers is demonstrated for intercalation of graphene nanoplatelets (GNPs). Interestingly, the loading of graphene nanoplatelets with SiO2 nanoparticles enhances the photocatalytic efficiency from 46% to 99%. For biomedical applications, it is found that 75% of Gram positive and 50% of Gram negative bacteria have been killed, hence bacterial proliferation is significantly restricted. Further, the cytotoxicity study reveals that the synthesised nanocomposites are non-toxic for both normal (HCEC) and cancerous (MCF-7, HEp-2) cell lines which signifies their potential as carriers for drug delivery. The prepared nanocomposites with controlled amount of carbon in the form of graphene can be employed for photocatalysis based waste water remediation, biomedicine and nano drug delivery	physics.app-ph	physics	PREPRINTS Graphene/SiO2 nanocomposites: the enhancement of photocatalytic and biomedical activity of SiO2 nanoparticles by graphene Aqsa Arshad1,2,a), Javed Iqbal3,a), Qaisar Mansoor4, Ishaq Ahmed5 1Department of Physics, International Islamic University, Islamabad, Pakistan. 2Department of Physics, Durham University, South Road Durham DH1 3LE, United Kingdom. 3 Laboratory of Nanoscience and Technology, Department of Physics, Quaid i Azam University, Islamabad, Pakistan. 4Institute of Biomedical and Genetic Engineering (IBGE), Islamabad, Pakistan. 5Experimental Physics Labs, Professor Abdus Salam Centre of Physics, Islamabad, Pakistan. a) Corresponding author(s): [email protected] ; [email protected] This manuscript is accepted by Journal of Applied Physics (AIP) Volume 121 Issue 24 Page number 244901 Year 2017 doi: 10.10632/1.4979968 Comments & suggestions are welcome [email protected] 1 ABSTRACT: The exceptional conducting nature of graphene makes it a viable candidate for enhancing the effectiveness of photocatalytic and biomedical nanomaterials. Herein, the immobilization of monodispersed silicon dioxide (SiO2) nanoparticles on multiple graphene layers is demonstrated for intercalation of graphene nanoplatelets (GNPs). Interestingly, the loading of graphene nanoplatelets with SiO2 nanoparticles enhances the photocatalytic efficiency from 46% to 99%. For biomedical applications, it is found that 75% of Gram positive and 50% of Gram negative bacteria have been killed, hence bacterial proliferation is significantly restricted. Further, the cytotoxicity study reveals that the synthesised nanocomposites are non-toxic for both normal (HCEC) and cancerous (MCF-7, HEp-2) cell lines which signifies their potential as carriers for drug delivery. The prepared nanocomposites with controlled amount of carbon in the form of graphene can be employed for photocatalysis based waste water remediation, biomedicine and nano drug delivery. KEYWORDS: Toxic nanomaterials; Graphene nanocomposites; Photocatalysis; Waste water treatment; Cytotoxicity; Biomedicine I. INTRODUCTION Health and ecological issues arising due to microorganism and rapid industrialization have posed the excessive risk of diseases and water contamination. These problems have enforced researchers to develop new environment friendly materials to cope with worldwide health hazards and treatment of waste water. Nowadays, finding solutions to clean the water contaminated by dyes is prime focus of research. At the same time, synthesis of bactericidal materials1 is equally significant for cleaning infected water, food packaging, hospital implants, and dentistry instruments.2-5 Furthermore, in the health sector, cancer is the most threatening and incurable disease known till now.6 Therefore, the chemotherapeutic 2 treatments are essential and such treatments require nontoxic nanosized carriers for efficient drug delivery.7 Metal and non-metal oxides have been extensively studied for photocatalysis and biomedical applications.8-12 Among them, SiO2 has the prime importance as it is the promising candidate for future development of catalysts, adsorbents, nanodrug carriers and biomolecular transport agents.13-20 However, some demerits of SiO2 restrict its efficiency in photocatalysis and biomedicine, which are: the wide bandgap (~5eV), aggregation of NPs, quick charge carrier's recombination and low surface area. These factors contribute towards its inert behaviour towards many catalytic processes and it shows only a slight catalytic activity under UV irradiation.21,22 To improve its photocatalytic performance, it is highly desirable to combine carbonaceous materials with SiO2, as various studies indeed describe their efficacy for remediation of contaminated water.23-26 The latest addition to carbonaceous materials is graphene which is a perfectly two dimensional, sp2 hybridized carbon. The ultrafast Dirac Fermions (near the k and k' points) render exceptional transport properties in graphene.27 The enticing features that make graphene ideal for various applications are its high electrical and thermal conductivities, large specific surface area, good chemical stability and outstanding mechanical properties.28-30 These features pave the way for incorporation of graphene in hybrid materials (constituted by its combination with metals, ceramics, polymers, and chalcogenides).31-36 Harnessing the good electronic transport and other physical properties of graphene with those of ceramics can significantly enhance their performance in photocatalysis and biomedical applications.37-39 For instance, the activity of a nanomaterial to remove dye and bacterial contamination is mainly influenced by surface area, surface roughness and functionalization.40 SiO2 NPs decorated graphene nanoplatelets (GNPs) have hybrid properties of both nanomaterials e.g., improved interfacial contact (leading to large surface area), development of conducting pathways, and suppression of charge 3 recombination. Incorporating these features via formation of graphene/SiO2 nanocomposites provides a superior channel for enhancement of the photocatalytic performance as compared to other carbonaceous nanocomposites or SiO2 alone.41 The combined assembly of SiO2 and graphene also makes possible the covalent and non-covalent attachments of drugs on graphene by π–π interactions. This makes it a suitable nano-carrier for delivering drug to target cancer sites.42 This outcome can't be achieved by SiO2 NPs alone. Thus, these merits make graphene/SiO2 nanocomposites, potentially new candidates for averting the ever- growing health and environmental risks associated with contaminated water. Graphene family based nanocomposites with SiO2 have recently been studied in detail for various applications e.g., liquid chromatography, fluorescence enhancement, super hydrophilic coatings.43-47 Despite these few reports, there is still a lot of room to probe various aspects of graphene/SiO2 nanocomposites. For example, health and environmental impacts of graphene/SiO2 nanocomposite need to be thoroughly evaluated before employing it for potential applications. In this work, photocatalytic activity under UV light illumination for a model organic dye methyl orange (MO), antibacterial features of graphene/SiO2 nanocomposite for three model bacteria and the cytotoxicity of graphene/SiO2 nanocomposite for normal (HCEC) and two cancerous cell lines (MCF-7, HEp-2) have been evaluated. For the first time, this study explores the impact of graphene loading on the role of SiO2 in the fast photodegradation of MO and inhibition of the bacterial growth. Moreover, the cytotoxicity evaluations suggest the possible use of graphene/SiO2 nanocomposites as a nano carrier for drug delivery and bio-imaging in cancer treatment procedures. II. EXPERIMENTAL METHODS A. SYNTHESIS OF GRAPHENE/SiO2 NANOCOMPOSITES Tetraethyl ortho silicate (TEOS) (99%, Fluka), liquid ammonia (32%), ethyl alcohol (99%, Merck), GNPs (100%, KNano) and distilled water were used in the fabrication of 4 monodispersed SiO2 particles and graphene/SiO2 nanocomposites. All the chemicals were used as obtained. The synthesis process was initiated by mixing and stirring definite amounts of TEOS, in the double solvent of ethyl alcohol and distilled water in 10:3 (v/v). The pH of the solvent was controlled by NH3. With careful monitoring of pH, different amounts of sonicated GNPs were added to the solution. The reaction was completed in 2 hours. The solution was dried at 373.15 K for 12 hours in an electric oven with post annealing session at 923.15K for 1 hour in a tube furnace. The SiO2 was prepared under similar conditions without addition of GNPs. Three nanocomposites labelled as Sx, Sy, Sz were prepared with different feed ratios of GNPs i.e., x = 40 mg, y = 80 mg, and z = 100 mg. B. PHOTOCATALYTIC DEGRADATION EXPERIMENT A textile dye, MO was selected as a model pollutant to check the photocatalytic performance of graphene/SiO2 nanocomposites. 0.03g of each photocatalyst was added to 15 μM aqueous MO solution. The pH of solution was adjusted to 3 using 1M HNO3 (1M NH3). 100 ml of each solution was subjected to the experiment. After establishing the adsorption- desorption equilibrium between photocatalyst and MO. Thereafter, the zero-time reading was recorded and the solution was exposed to UV light source (Type C, 90 Watt, emission peak 254 nm). 4 ml of each sample was withdrawn at regular intervals from all the solutions. The samples of the experimented solutions were analysed for absorbance immediately after centrifugation. Each experiment was repeated three times to ensure the accuracy of results. C. ANTIBACTERIAL ACTIVITY EXPERIMENT To study the antibacterial performance, the 10 mg/ml of test samples (SiO2, Sx, Sy, Sz) in sterile water were sonicated. The 200 μL of each sonicated solution was added to 5 ml Luria- Bertani Broth (LB) medium. This growth medium was added to 100 μL of the inoculum 5 (bacterial culture in LB). The inoculated media containing the test sample was incubated at 37oC for 24 h and all the bacteria were quantified by OD600nm at different intervals from 2 to 24 h. D. ANTICANCER TEST EXPERIMENT Three model cell lines were selected to check the anticancer and cytotoxicity profile of prepared samples. In this regard two cancer cell lines, i.e., human cervical cancer (HeLa) cells derivative (HEp-2), human breast adenocarcinoma cell line (MCF-7), and a normal cell line i.e., human corneal epithelial cells (HCEC) were subjected to the experiment. All the chemicals used in cell toxicity experiments were purchased from Sigma Aldrich (USA). To test the effect of prepared samples, all the cell lines were grown and maintained within incubator at 37oC, 5% CO2 in PRIM-1640 (Invitrogen-USA) supplemented with 5% antibiotics-antimycotic solution (GPPS) and 10% fetal bovine serum (FBS). In present work, the dosage dependent activity of prepared samples was checked against all cell lines. These cell lines were exposed to different doses i.e., 100 μg/ml to 400 μg/ml of each of the sample. The cells (1 × 105) were seeded in 96-well culture plates and were grown in 5% CO2 and humid atmosphere, for 48 hours at 37oC in incubator both in the presence and absence (control) of test samples. The percent cell viability was obtained using the MTT assay. E. CHARACTERIZATIONS The synthesized nanomaterials were characterized for their physical and chemical properties. The X-ray diffractograms were obtained by PANalytical X'Pert PRO diffractometer using Cu Kα radiation. A Shimadzu (IR Tracer-100) spectrometer was used for FTIR spectra. The morphology of prepared nanomaterials was studied by FE-SEM (MIRA3 TESCAN) and HRTEM (JOEL JEM 4000EX). Room temperature Raman spectra and photoluminescence measurements were taken by Ramboss using excitation laser of 6 wavelength 514 nm and 325 nm wavelength, respectively. A Tristar 3020 Micromeritics (USA) Porosimetry analyser was used to measure surface area of samples using Braunauer- Emmet-Teller (BET) method. To study the photocatalytic properties, the UV-vis spectra were recorded by Perkin-Elmer (Lambda 25 UV). III. RESULTS AND DISCUSSIONS A. STRUCTURAL AND MORPHOLOGICAL INVESTIGATION The crystallinity of prepared nanocomposites is analysed in the range of 20o – 80o as depicted in Figure 1. The presence of diffraction peak at 26o corresponds to C (002) of the graphitic host matrix, and agrees with JCPDS-NO: 75-1621. The X-ray diffractograms show a broad halo in the region 2θ = 20o_30o. This obtuse peak indicates the amorphous nature of SiO2 NPs. The X-ray diffractograms agree with previous study.48 Morphological investigations have been carried out to understand the shape and attachment between the constituent species. The Figure 2(a) illustrates that SiO2 sample shows spherical shape, monodispersed particles with the diameter ranging from 230 nm - 260 nm. Figure 2 (b) and 2(c) reveal the microstructure of graphene/SiO2 nanocomposites. The semi- transparent graphene sheets with anchored SiO2 NPs are clearly visible in Figures 2(b) and (c), thereby confirming the intercalation of GNPs into few layers of graphene sheets. The insertion of SiO2 NPs has thus served to reduce the van der Waal's interaction between the stacked GNPs. It is interesting to note that the particle size of SiO2 has been greatly reduced during nanocomposite formation, which is logical and expected as well. Indeed, the confinement effect of graphene sheets has been observed in previous reports on graphene/metal oxide's nanocomposites.49 Moreover, the large surface area provided by graphene during growth process serves to reduce the excessive aggregation of primary nuclei. An Energy Dispersive X-ray spectroscopy (EDX) coupled with FE-SEM was used to investigate the composition of pristine SiO2 and graphene/SiO2 nanocomposite. The presence 7 of carbon, silicon and oxygen is confirmed by peaks [see insets in left bottom of Figure 2(a) and 2(b)]. To further investigate the nature of SiO2 and nanocomposites, TEM and selected area electron diffraction (SAED) patterns are presented in the Figure 2(d), which demonstrates that the SiO2 NPs are attached to graphene sheets. The SAED pattern supports the observation from X-ray diffractograms. The inset illustrates that the SiO2 NPs attached to graphene sheets show amorphous nature. Their amorphous nature is retained after the formation of nanocomposites. B. GROWTH MECHANISM The growth mechanism is explained schematically in the Figure 3. The graphene sheets offer the active sites and large surface area for the growth of SiO2 NPs. This gives the advantage that a large surface area is available for the nucleation of the primary particles, thereby reducing the particle size of SiO2 NPs as compared to pristine SiO2. The confining effect of graphene sheets also contributes to stop nucleation process after a certain limit thus leading to the reduced particle size as compared to pristine SiO2.48,49 The SiO2 spheres have served to intercalate the graphene sheets as they get drafted on graphene and reduce the van der Waal's interaction between the sheets. C. FTIR AND RAMAN ANALYSIS The FTIR spectroscopic curves are shown in Figure 4. A band at 460 cm-1 can be assigned to the Si-O-Si bending vibrations. The band at 812 cm-1 originates due to Si-O symmetric bending vibrations, where –O vibrations are perpendicular to the Si-Si bond line. However, the band in the range 1045-1107 cm-1 is due to parallel vibrations of oxygen atom in either direction in Si-O-Si linkage. Thus, it manifests asymmetric mode of Si-O-Si bond. The adsorbed water molecules manifest themselves by a band around 1615 cm-1. The chemical 8 bond Si-OH appears as a band around 957cm-1 and 3458 cm-1. The graphene/SiO2 nanocomposites show significant difference with the presence of a broad band in the range 1014-1303 cm-1. The C-Si bond manifests itself in this region around 1260 cm-1. Thus, it can be concluded that SiO2 fabrication modified the surface of graphene. The FTIR results accord well with previous studies.43, 48 Raman spectroscopy is the basic characterization tool to identify graphene and its nanocomposites. The Raman spectra obtained from 1000 cm-1 to 2000 cm-1 are presented in Figure 5. All samples show D band, associated with defects, located around 1350 cm-1. The defect band arises due to termination of sheet at edges and attachment of particles on graphene. The E2g mode arises due to first order scattering and is manifested as G band, located at 1587 cm-1 in case of GNPs. These observations agree with previous reports on graphene's Raman spectra.50,51 The nanocomposites Sx, Sy, Sz have G-band located at 1602 cm-1, 1599 cm-1, and 1602 cm-1. An overall shift towards higher wavenumber of G-band is observed in all the nanocomposites as compared to GNPs. This shift is due to the charge transfer between GNPs and SiO2 NPs, and is an indicator of electrostatic interaction between the two-constituent species. This large shift illustrates the strong attraction between the two constituent phases of nanocomposites. The Raman results confirm the formation of GNPs and SiO2 nanocomposites.52 D. PHOTOLUMINESCENCE AND SURFACE AREA ANALYSIS To monitor the optical changes due to structural defects, a comparison of photoluminescence spectra of SiO2 and graphene/SiO2 nanocomposites is presented in Figure 6. The incomplete Si-O-Si tetrahedral network formation on the surface of graphene may lead to several structural defects. The SiO2 nanoparticles show emission peaks in the visible light region. A weak band near UV region 355.6 nm (3.48 eV) is contributed by silanol groups (-OH related 9 groups). Green emission 512nm (2.43 eV) is observed. A very prominent band 409 nm (3.04eV) is observed related to violet emission.53,54 The comparison of the PL results of graphene/SiO2 nanocomposites and SiO2 indicates that the intensity of all emission peaks quenches significantly. This suppression is attributed to the presence of graphene, which acts as an acceptor of electrons in the nanocomposite.55 Graphene sheets provide an additional path for the conduction electrons of SiO2. The suppression of PL intensity indicates the decrease in carriers' recombination. This quenching behaviour agrees with the previous reports as well suggests the potential photocatalytic use of prepared nanocomposites. The surface area is examined by N2 adsorption-desorption isotherms. The observed values are tabulated in Table 1. The SiO2 particles with relatively larger particle size, possess the smallest BET surface area. With the increment of graphene content, the surface area has been increased significantly. Since, theoretically, graphene possesses a surface area ≈ 2600 m2g-1. The highest BET surface area is observed for Sz i.e., 146.52 m2g-1. Sample Surface area -2 2 (m g ) Pore volume -1 3 (cm g ) Average pore size (Å) SiO 2 S x S y S z 6.3547 0.019405 55.157 16.0839 0.0210532 46.234 30.2417 0.040561 35.904 146.5199 0.393357 33.021 Table 01 E. PHOTOCATALYTIC DEGRADATION OF MO 10 Finally, to evaluate the performance of graphene/SiO2 nanocomposites, MO was employed as water contaminant in photo induced dye-degradation experiments. To study the impact of increased graphene concentration on MO degradation, four experiments were conducted using catalysts SiO2, Sx, Sy, and Sz. Figures 7(a)-(d) show the absorption spectra detailing degradation of MO. The adsorption of catalysts (in the dark) on degrades the dye molecules slightly. The photodegradation efficiencies of MO are presented in Figure 8 using different photocatalysts. In the presence of SiO2, 46% photodegradation is achieved in 160 min. The catalyst with minimal graphene content, i.e., Sx achieves 87.2% efficiency in 160 min. The catalyst Sy has shown 92% photodegradation of MO in 160 min. The photocatalytic efficiency of graphene/SiO2 is maximized at the optimal graphene content in the photocatalyst Sz which shows 99% photodegradation in considerably reduced time. The spectrum becomes flat only in 100 min. The photocatalytic efficiency obtained in present case is much better than previous report on photocatalytic activity of SiO2 NPs with Au/Ag doping.21, 22 The mechanism of photocatalytic activity is explained below. The photons of UV light falling on the SiO2 nanoparticles, excite its valence band electrons to the conduction band and produce e- cb – h+ vb pairs. The number of e- cb – h+ vb pairs increases gradually with time. The graphene attached to SiO2 NPs, being a good acceptor of electrons, provides trapping sites for e- cb. This delays the recombination of e- cb – h+ vb pairs. Meanwhile the e- cb may also interact with the dissolved O2 to produce O- 2 species, which may further produce several reactive oxygen species (ROS) as mentioned in the following equations. The holes in the valence band of SiO2 contribute towards generation of •OH radicals.21,22 These species attack the ring of azo dye, MO, by completely opening its ring structure. This ultimately results into the mineralization of dye.37 The mechanism is schematically illustrated in Figure 9. 11 SiO2 – graphene + hSiO2 (h+ vb) – graphene (e- cb) SiO2 (h+ vb) + (H2O  H+ + OH-) SiO2 + H+ + •OH graphene (e- cb) + O2  graphene + O- 2 O- 2 + 2(H+ + OH-)  H2O• + 2OH- •OH + dye  H2O + CO2 In the case of bare SiO2 particles, the e- cb – h+ vb pairs formed on the surface of SiO2 recombine quickly. Only very few carriers can be trapped on surface states of SiO2 particles, which may further initiate the dye degradation. The amount of ROS generated in the process strictly controls the dye degradation. Fewer ROS can react with adsorbed dye molecules. In this case absence of graphene leads to quick recombination of e- cb – h+ vb pairs. That is why SiO2 particles have shown low photocatalytic activity as compared to graphene/SiO2 nanocomposites. Notably, all the nanocomposites have not shown the similar photoactivity for the degradation of MO. It indicates that the importance of optimum addition ratio of graphene in nanocomposites. The explanation for the graphene content dependence of photocatalytic performance of the nanocomposites is elucidated below. Incorporating graphene in nanocomposites seems to promote electron trapping. This is due to exceptional conductivity of graphene. The retardation in recombination of UV light generated charge carriers in SiO2 by introducing graphene nanoplatelets gets support from photoluminescence spectra (Figure 6). PL findings suggest that the quenching of intensity is due to inhibition of charge carrier's recombination in nanocomposites. As electrons are accepted by graphene lying adjacent to the SiO2 NPs. This inhibition ultimately consequences 12 the efficient photocatalytic performance of nanocomposites as compared to SiO2 alone. The highest quenching is observed in Sz which supports the photocatalytic results. The nanocomposite having highest quenching has shown most efficient photocatalytic activity by 99% degradation of MO in 100 minutes. The surface area of sample is likely to affect the performance of photocatalyst. The high surface area of graphene/SiO2 nanocomposites (Table 1) offers more area (as compared to SiO2) for better adsorption of MO molecules. Therefore, a better contact between the dye and photocatalyst is developed. Hence, the synergistic effect created by conducting graphene and high surface area is helpful towards fast and efficient degradation of dye. The reaction kinetics of experiment give a better insight to the photocatalytic activity. The reactions kinetics for photocatalysis can be described on the bases of Langmuir-Hinshelwood model56 as presented in Figure 10. The UV light induced degradation of MO can be well ascribed by pseudo-first order kinetics. The rate equation that describes the reaction is ln(Ct/Co) = kt, where Ct and Co are the dye concentration at time t and the initial concentration respectively, and k = apparent rate constant. The apparent rate constant determined for different catalysts are 0.003 min-1 (SiO2), 0.014 min-1(Sx), 0.015 min-1 (Sy), and 0.04 min-1(Sz). These results show that the rate constant has been increased significantly by increasing graphene content. The graphene/SiO2 nanocomposite with maximum graphene loading has the highest apparent rate constant (an order of magnitude higher than that of bare SiO2 particles) and therefore it exhibits excellent photocatalytic activity. Hence, it can be concluded that Sz nanocomposite is photocatalytically most active than pristine SiO2 and graphene/SiO2 nanocomposites with low graphene content for the degradation of MO. This observation is ascribed to the synergistic effect of excellent electron acceptor nature of graphene and higher surface area of graphene/SiO2 nanocomposites.10,23 13 The recyclability tests are very important for the practical use of photocatalyst. The recycling performance of Sz was evaluated for consecutive three cycles. It is illustrated by inset in Figure 10 that there is negligible loss in photocatalytic activity of Sz. The Sz nanocomposite shows excellent performance even after three continuous cycles of activity. Therefore, it may be recommended as efficient alternative of traditional photocatalysts. F. ANTIBACTERIAL STUDY To investigate the effect of graphene/SiO2 nanocomposites on Gram negative and Gram positive bacteria, the time kill assay was conducted for three model bacterial strains i.e., Escherichia coli (E. coli), Methicillin resistant Staphylococcus aureus (S. aureus), and Pseudomonas aeruginosa (P. aeruginosa) in the absence and presence all samples. The growth inhibition rate was determined by observing the optical density (600 nm) at different intervals in the total incubation time (24 h). The results obtained from the experiment are presented by the bacterial growth inhibition curves in Figures 11(a)-(c). The control sample in Figure 11(a)-(c) presents the untreated bacterial strains under observation for comparison purposes. The experimental findings suggest the inhibition of bacterial growth to a significant extent. It is observed that bacterial growth inhibition is a strong function of graphene loading in the nanocomposites. The growth of S. aureus has been inhibited up to 26.32%, 30.17%, 50.30%, and 75.40% by SiO2, Sx, Sy, and Sz respectively. For E. coli, the growth inhibition rates are 17.00%, 30.00%, 32.23%, 51.80% for SiO2, Sx, Sy, and Sz respectively. The growth of P. aeruginosa has been inhibited upto 17.75%, 40.52%, 40.77%, and 48.97% by SiO2, Sx, Sy, and Sz respectively. The sample with no graphene contents i.e., SiO2 shows the minimum inhibition of bacterial growth for all the bacterial strains. The increase of graphene loading in nanocomposites decreases the number of viable cells. The maximum growth inhibition rate is achieved in nanocomposite with maximum graphene content. The sample with maximum graphene loading (Sz) is found to possess excellent antibacterial properties for the growth 14 inhibition of S. aureus. Graphene/SiO2 nanocomposites have stopped around 50% growth of Gram negative bacterial strains. These results are much better than the previous reports, where 47% and 49.5 ± 4.8% growth inhibition of E. coli, and 34% growth inhibition of S. aureus was achieved by using rGO and graphene films.57-59 Several research reports are available, documenting the possible mechanisms for growth inhibition of bacterial strains by carbon nano tubes, fullerenes, and graphene family nanostructures.60-65 But the exact mechanism explaining the loss of bacterial integrity is still a researchable topic. One of the suggested mechanism is the destruction of bacterial membrane induced by direct contact between sheet like structure of graphene based materials and bacteria. This mechanism has been proposed previously for GO, rGO, and CNTs.58 In present case, intermingling of bacteria and planar graphene may be thought to induce irreversible destruction of bacterial membrane. The planes and sharp edges of graphene nanosheets produce significant stress on cell membranes. These edges serve as cutters for rupturing the bacterial cell membranes which induce cell death by leakage of cytoplasmic content. Additionally, at the same time the normal respiratory functioning of bacteria is strongly dependent on electronic charge transport between the cell and mitochondrial membranes in respiratory chain reactions. The physical contact between bacteria and graphene/SiO2 nanocomposites may result in Schottky barrier formation, as previously reported for graphene on SiO2 substrate.66 As graphene is an excellent electron acceptor so it can be speculated that cell membranes may lose their electrons, which are eventually transported to graphene. In this manner, a charge imbalance is created in bacterial cells which leads to cell death.66,67 However, the differential toxicity of samples towards Gram positive and Gram negative bacteria can be explained on the bases of their outer membrane's composition. The Gram- negative bacteria have a complex double membrane's structure which is less penetrable as compared to single membrane of Gram positive bacteria. This difference, mainly arising due 15 to chemical composition of membranes of the two classes of bacteria, develops their differential resistance.68,69 Hence, this feature makes S. aureus an easy target for graphene/SiO2 nanocomposites. G. CYTOTOXICITY ANALYSIS It is highly desirable to evaluate the cytotoxic response of a nanoplatform against various cell lines before its recommendation for drug attachment to kill the cancer cells. The dosage dependent toxicity of prepared samples towards different cell lines (HCEC, MCF-7 and HEp- 2) was evaluated. The absorbance at 550 nm was recorded and % cell viability was obtained. The untreated cell lines were considered as control. A comparison between control and treated cell lines show that the number of viable cells doesn't decrease after the exposure to prepared samples. The dosage increment from100 to 400 μg/ml also doesn't create the toxic effects both on cancer cell lines and normal cell lines. No apoptosis or cell death is induced using any of the prepared samples at exposure concentrations as high as 400 μg/ml and exposure time as much long as 48 hours. These findings establish that SiO2 particles and graphene/SiO2 nanocomposites are non-toxic for used cell lines. Their non-toxicity towards healthy cells is an excellent indication for their utilization as a nanoplatform for efficient drug delivery. The prepared graphene/SiO2 nanocomposites can be functionalized with anti-cancer drugs like doxorubicin as previously done for carbon nanoparticles. The biofunctionalization of graphene with SiO2 and doxorubicin is expected to target the cancerous cell lines by making use of combined mechanism of photothermal therapy and chemotherapy.42 Moreover, the slight modification in morphology of prepared SiO2 (that is making them mesoporous) and attachment of doxorubicin can target glioma.70 This nanoplatform can be modified with hypocrellin A suggesting their efficacy in drug delivery for photo dynamic cancer treatment and bio-imaging.71 We conclude that the cytotoxicity analysis presented in this study, opens 16 the doors towards covalent attachment of anti-cancerous drugs on prepared graphene/SiO2 nanocomposites. IV. CONCLUSIONS High surface area graphene/SiO2 nanocomposites are synthesized successfully by a simple chemical route. The composite developed with retention of the exceptional intrinsic properties of graphene leads to achieve the outstanding photocatalytic performance with 99% degradation of MO under UV light illumination. Graphene loaded with SiO2 shows excellent antibacterial activities i.e., 75% growth inhibition of S. aureus and ≈ 50% loss of E. coli, and P. aeruginosa. The prepared nanocomposites show no toxicity towards normal cells (HCEC), human cervical cancer cells (HEp-2), and breast cancer cells (MCF-7) which suggests their possible use as nano drug carriers to target cancerous sites. Our experimental findings greatly recognize graphene/SiO2 nanocomposites for their utilization towards waste water treatment and biomedical applications. ACKNOWLEDGEMENTS This work was funded by the Higher Education Commission of Pakistan (HEC) NRPU (Grant No: 20-4861/R & D/ HEC/14) to Dr. Javed Iqbal and Higher Education Commission of Pakistan (HEC) IRSIP (Grant No: 1-8/HEC/HRD/2016/5995 PIN: IRSIP 32 PSc 04) to Aqsa Arshad. The authors are thankful to Dr. Ian Terry and Dr. Mahavir Sharma, Department of Physics, Durham University, UK, for the useful discussions. The plagiarism check of the article (using Turnitin ID: 801482706) as recommended by HEC Pakistan, shows 2% similarity index which is well below the permissible range. REFERENCES 1 A. Hameed, Z. Shafiq, M. Yaqub, M. Hussain, M. A. Hussain, M. Afzal, M. N. Tahir and M. M. Naseer, New J. Chem. 39, 9351-9357 (2015). 2 L. Zhao, H. Wang, K. Huo, L. Cui, W. Zhang, H. Ni, Y. Zhang, Z. Wu, P. K. Chu, Biomaterials 32, 5706–5716 (2011). 17 3 4 5 6 M. Zhang, K. Zhang, B. D. Gusseme, W. Verstraete, Water Res. 46, 2077–2087 (2012). P. Prombutara, Y. Kulwatthanasal, N. Supaka, I. Sramala, S. Chareonpornwattana, Food Control 24, 184–19 (2012). L. Cheng, M. D. Weir, H. H. K. Xu, J. M. Antonucci, A. M. Kraigsley, N. J. Lin, S. Lin-Gibson, X. Zhou, Dent. Mater. 28, 561–572 (2012). F. Anam, A. Abbas, K. Mun Lo, Z. Rehman, S. Hameed and M. M. Naseer, New J. Chem. 38, 5617-5625 (2014). 7 Kenry, P. K. Chaudhuri, K. P. Loh, C. T. Lim, ACS Nano 10, 3424–3434 (2016). 8 H. R. Jafry, M. V. Liga, Q. Li, A. R. Barron, Environ. Sci. Technol. 45, 1563–1568 (2011). 9 B. Czech, W. Buda, Environ. Res.137, 176–184 (2015). 10 T. Kavitha, A. I. Gopalan, K. P. Lee, S. Y. Park, Carbon 50, 2994–3000 (2012). 11 Y. Li, W. Zhang, J. Niu, Y. Chen, ACS Nano 6, 5164–5173 (2012). 12 G. A. Seisenbaeva, M. P. Moloney, R. Tekoriute, A. H. Dessources, J. M. Nedelec, Y. K. Gun'ko, V. G. Kessler, Langmuir 26, 9809–9817 (2010). 13 A. Popat, S. B. Hartono, F. Stahr, J. Liu, S. Z. Qiao, G. Q. M. Lu, Nanoscale 3, 2801−2818 (2011). 14 X. Yang, Z. Shen, B. Zhang, J. Yang, W. X. Hong, Z. Zhuang, J. Liu, Chemosphere 90, 653−659 (2013). 15 R. D. C. Solatani, A. R. Khataee, M. Safari, S.W. Joo, Int. Biodeterior. Biodegrad. 85, 383−391 (2013). 16 X. Li, J. Zhang, H. Gu, Langmuir 10, 6099−6106 (2011). 17 B. Bharti, J. Meissner, G. H. Findenegg, Langmuir 27, 9823−9833 (2011). 18 Y. Badr, M. G. A. E. Wahed, M. A. Mahmoud, J. Haz. Mat. 15, 245–253 (2008). 18 19 Y. Badr, M. A. Mahmoud, J. Phys. Chem. Solids 68, 413–419 (2007). 20 L. F. D. Oliveira, K. Bouchmella, K. D. A. Gonçalves, J. Bettini, J. Kobarg, M. B. Cardoso, Langmuir 32, 3217–3225 (2016). 21 B. Xu, Y. Ju, Y. Cui, G. Song, Y. Iwase, A. Hosoi, Y. Morita, Langmuir 30, 7789– 7797 (2014). 22 C. Giménez, C. D. Torre, M. Gorbe, E. Aznar, F. Sancenón, J. R. Murguía, R. Martínez-Máñez, M. D. Marcos, P. Amorós, Langmuir 31, 3753–3762 (2015). 23 X. Li, S. Yang, J. Sun, P. He, X. Xu, G. Ding, Carbon 78, 38–48 (2014). 24 G. Jiang, Z. Lin, C. Chen, L. Zhu, Q. Chang, N. Wang, W. Weia, H. Tang, Carbon 49, 2693–2701 (2011). 25 V. Kumara, N. Bahadur, D. Sachdeva, S. Guptaa, G. B. Reddyb, R. Pasricha, Carbon 80, 290-2304 (2014). 26 N. O. Ramoraswi, P. G. Ndungu, Nanoscale Res. Lett. 10, 427-443 (2015). 27 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva1, A. A. Firsov, Science 306, 666-669 (2004). 28 N. A. H. Castro, F. Guinea, N. M. R. Peres, K. S. Novoselov, A. K. Geim, Rev. Mod. Phys. 81, 109–162 (2009). 29 C. N. R. Rao, A. K. Sood, K. S. Subrahmanyam, A. Govindaraj, Angew. Chem. Int. Ed. 48, 7752–7777 (2009). 30 A. K. Geim, K. S. Novoselov, Nat. Mater. 6, 183–191 (2007). 31 R. Muszynski, B. Seger, P. V. Kamat, J. Phys. Chem. C. 112, 5263–5266 (2008). 32 Y. Y. Wen, H. M. Ding, Y. K. Shan, Nanoscale 3, 4411–4417 (2011). 33 H. Seema, K. C. Kemp, V. Chandra, K. S. Kim, Nanotechnol. 23, 355705–355712 (2012). 19 34 S. Stankovich, D. A. Dikin, G. H. B. Dommett, K. M. Kohlhaas, E. J. Zimney, E. A. Stach, R. D. Piner, S. T. Nguyen, R. S. Ruoff, Nature 442, 282–286 (2006). 35 T. S. Sreeprasad, S. M. Maliyekkal, K. P. Lisha, T. Pradeep, J. Hazard. Mat. 186, 921– 31 (2011). 36 C. Xu, X. Wang, J. Zhu, J. Phys. Chem. C 112, 19841–19845 (2008). 37 A. Arshad, J. Iqbal, M. Siddiq, Q. Mansoor, M. Ismail, F. Mehmood, M. Ajmal, and Z. Abid, J. Appl. Phys. 121, 024901 (2017). 38 Y. Gao, X. Pu, D. Zhang, G. Ding, X. Shao, J. Ma, Carbon 50, 4093–4101 (2012). 39 a S. C. Smith, D. F. Rodrigues, Carbon 91, 22-143 (2015). b X. Liang, S. Liu, X. Song, Y. Zhu, S. Jiang, Analyst 137, 5237–5244 (2012). 40 H. M. Hegab, A. E. Mekawy, L. Zou, D. Mulcahya, C. P. Saint, M. G. Markovic, Carbon 105, 362-376 (2016). 41 L. Yang, L. Wang, M. Xing, J. Lei, J. Zhang, Appl. Catal. B: Environ. 180, 106–112 (2016). 42 X. Tu, L. Wang, Y. Cao, Y. Ma, H. Shen, M. Zhang, Z. Zhang, Carbon 97, 35–44 (2016). 43 C. W. Lee, K. C. Roh, K. B. Kim, Nanoscale 5, 9604-9608 (2013). 44 Z. M. Wang, W. Wang, N. Coombs, N. Soheilnia, G. A. Ozin, ACS Nano 4, 7437– 7450 (2010). 45 S. Watcharotone, D. Z. Dikin, S. Stankovich, R. Piner, I. Jung, G. H. B. Dommett, G. Evmenenko, S. E. Wu, S. F. Chen, C. P. Liu, S. T. Nguyen, R. S. Ruoff, Nano Lett. 7, 1888-1892 (2007). 46 D. Yin, B. Liu, L. Zhang, M. Wu, J. Biomed. Nanotechnol. 8, 458-464 (2012). 47 L. Kou, C. Gao, Nanoscale 3, 519–528 (2011). 48 N. D. Singho, M. R. Johan, Int. J. Electrochem. Sci. 7, 5604–5615 (2012). 20 49 B. Zhao, J. Song, P. Liu P, W. Xu, T. Fang, Z. Jiao, H. Zhang, Y. Jaing, J. Mat. Chem. 21, 18792-18798 (2011). 50 T. Tsukamoto, K. Yamazaki, H. Komurasaki, T. Ogino, J. Phys. Chem. C 116, 4732- 4737 (2012). 51 A. C. Ferrari, J. C. Meyer, V. Scardasi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth, A. K. Geim, Phys. Rev. Lett. 97, 187401-187404 (2006). 52 A. K. Rai, J. Gim, L. T. Anh, J. Kim, Partially reduced Co3O4/graphene nanocomposite as an anode material for secondary lithium ion battery, Electrochimica Acta 100, 63-71 (2013). 53 M. Jafarzadeh, I. A. Rahman, C. S. Sipaut, Ceramics Int. 36, 333–338 (2010). 54 I. A. Rahman, P. Vejayakumaran, C. S. Sipaut, J. Ismail, C. K. Chee, Mat. Chem. Phy. 114, 328–332 (2009). 55 X. An, J. C. Yu, Y. Wang, Y. Hu, X. Yu, G. Zhang, J. Mat. Chem. 22, 8525–8531 (2012). 56 K. V. Kumar, K. Porkodi, F. Rocha, Cat. Comm. 9, 82-84 (2008). 57 W. Hu, C. Peng, W. Luo, M. Lv, X. Li, D. Li, Q. Huang, C. Fan, ACS Nano 4, 4317– 4323 (2010). 58 S. Liu, T. H. Zeng, M. Hofmann, E. Burcombe, J. Wei, R. Jiang, J. Kong, Y. Chen, ACS Nano 5, 6971–6980 (2011). 59 L. Dellieu, E. Lawarée, N. Reckinger, C. Didembourg, J. J. Letesson, M. Sarrazin, O. Deparis, J. Y. Matrouleb, J. F. Colomer, Carbon 84, 310–316 (2015). 60 E. Nakamura, H. Isobe, Acc. Chem. Res. 36, 807-815 (2003). 61 B. Sitharaman, L. J. Wilson, J. Biomed. Nanotech. 3, 342-52 (2007). 62 L. Lacerda, A. Bianco, M. Prato, K. Kostarelos, Adv. Drug Deliv. Rev. 58, 1460-1470 (2006). 21 63 Z. Liu, S. Tabakman, K. Welsher, H. Dai, Nano Res. 2, 85-120 (2009). 64 O. Akhavan, E. Ghaderi, ACS Nano 4, 5731–5736 (2010). 65 Y. B. Zhang, S. F. Ali, E. Dervishi, Y. Xu, Z. R. Li, D. Casciano, A. S. Biris, ACS Nano 4, 3181–3186 (2010). 66 J. Li, G. Wang, W. Zhu, M. Zhang, X. Zheng, Z. Di, X. Li, X. Wang, Sci. Rep. 4, 4359 (2014). 67 H. W. Harris, M. Y. El-Naggar, O. Bretschger, M. J. Ward, M. F. Romine, A. Obraztsova, K. H. Nealson, PNAS 107, 326–331 (2010). 68 J. K. Krishnamoorthy, M. Veerapandian, L. H. Zhang, K. Yun, S. J. Kim, J. Phys. Chem. C 116, 17280−17287 (2012). 69 A. M. Jastrzębska, P. Kurtycz, A. R. Olszyna, J. Nanopart. Res. 14, 1320-41 (2012). 70 Y. Wang, K. Wang, J. Zhao, X. Liu, J. Bu, X. Yan, R. Huang, J. Am. Chem. Soc. 135, 4799-4804 (2013). 71 L. Zhou, L. Zhou, X. Ge, J. Zhou, S. Wei, J. Shen, Chem. Comm. 51, 421-424 (2015). Table Legends Table 1. Parameters obtained from N2 adsorption-desorption isotherms. Figure Captions Figure 1. X-ray diffractograms depicting SiO2 and graphene/SiO2 nanocomposites. 22 Figure 2. (a) SEM micrograph of monodispersed SiO2 particles, inset is EDX spectrum of SiO2, (b) and (c) graphene/SiO2 nanocomposite, inset represents EDX spectrum, and (d) TEM image of graphene/SiO2 nanocomposite (Sz), inset shows the SAED pattern of SiO2 NPs attached to graphene sheet. 23 Figure 3. Schematic illustration of intercalation of graphene nanoplatelets by SiO2 NPs. Figure 4. FTIR spectra of SiO2 and graphene/SiO2 nanocomposites. 24 Figure 5. Raman spectra depicting the formation of graphene/SiO2 nanocomposites. Figure 6. PL spectra of SiO2 and graphene/SiO2 nanocomposites. 25 26 Figure 7 (a-d). UV-Vis absorbance spectra of methyl orange in presence of different photocatalysts i.e., SiO2, Sx, Sy, and Sz. Figure 8. Photodegradation of methyl orange by SiO2 and graphene/SiO2 nanocomposites. Figure. 9. Proposed mechanism for UV light induced catalysis of MO using graphene/SiO2 nanocomposites. Figure 10. Pseudo first order rate kinetics for photocatalytic reactions by SiO2 and graphene/SiO2 nanocomposites. Inset: recyclability curves of graphene/SiO2 nanocomposite (Sz) depicting its excellent performance in three consecutive cycles of photocatalysis. 27 Figure 11. Bacterial growth inhibition curves obtained via MTT assay for (a) Methicillin resistant S. aureus, (b) E. coli and (c) P. aeruginosa. Insets: % cell viability of respective bacterial strains. View publication stats View publication stats 28
1903.11447	1	1903	2019-03-18T16:30:46	Zirconia nano-colloids transfer from continuous hydrothermal synthesis to inkjet printing	[ "physics.app-ph" ]	Water dispersion of nanometric yttria stabilized zirconia (YSZ) particles synthesized by continuous hydrothermal synthesis are transferred into nanoinks for thin film deposition. YSZ nanoparticles are synthesized in supercritical conditions resulting in highly dispersed crystals of 10 nm in size. The rheology of the colloid is tailored to achieve inkjet printability (Z) by using additives for regulating viscosity and surface tension. Inks with a wide range of properties are produced. A remarkable effect of nanoparticles on the ink printability is registered even at solid load < 1% vol. In particular, nanoparticles hinder the droplet formation at low values of the printability while suitable jetting is observed at high Z values, i.e. Z ca. 20. For the optimized inks, we achieve high quality printing with lateral and thickness resolution of 70 um and ca. 250 nm respectively, as well as self levelling effect with a reduction of the substrate roughness. Densification is achieved at sintering temperatures below 1200 C.	physics.app-ph	physics	Zirconia nano-colloids transfer from continuous hydrothermal synthesis to inkjet printing M. Rosa 1, P. N. Gooden 2, S. Butterworth 2 , P. Zielke 1, R. Kiebach 1, Y. Xu 1, C. Gadea 1, V. Esposito 1 1DTU Energy, Technical University of Denmark, Risø Campus, Frederiksborgvej 399, 4000, Roskilde, Denmark. 2Promethean Particles Ltd., Unit 1 Genesis Park, Midland Way, Nottingham, NG7 3EF, UK. corresponding author: [email protected] ABSTRACT Water dispersions of nanometric yttria stabilized zirconia (YSZ) particles synthesized by Continuous Hydrothermal Synthesis are transferred into nano-inks for thin film deposition. YSZ nanoparticles are synthesized in supercritical conditions resulting in highly dispersed crystals of 10 nm in size. The rheology of the colloid is tailored to achieve inkjet printability (Z) by using additives for regulating viscosity and surface tension. Inks with a wide range of properties are produced. A remarkable effect of nanoparticles on the ink printability is registered even at solid load < 1 %vol. In particular, nanoparticles hinder the droplet formation at low values of the printability while suitable jetting is observed at high Z values, i.e. Z  20. For the optimized inks, we achieve high quality printing with lateral and thickness resolutions of 70 µm and ca. 250 nm respectively, as well as self-levelling effect with a reduction of the substrate roughness. Densification is achieved at sintering temperatures below 1200 °C. Keywords: Continuous Hydrothermal Synthesis, Nanomaterial, Zirconia, Inkjet, Thin Films 1. Introduction Nanostructured materials show unique properties that can dramatically improve the performances of several technologies in many different fields, from catalysis [1] to optics [2], from optoelectronic [3] to structural applications [4]. From the application point of view, their higher activity compared to bulk materials is due to the combination of size effects [5] and high surface area [6,7]. The first modify the band structure of the material, affecting, e.g., electronic and optical properties [5]. The latter has a direct impact on phenomena taking place at the interfaces, increasing the number of superficial defective sites [7]. As a consequence, nanoparticles typically show a decrease in the sintering temperature [8 -- 13], which can be exploited in their processing. In addition, their nanometric size is ideal for deposition via liquid casting methods such as inkjet printing [14]. Inkjet printing is an additive manufacturing technique that consists in the accurate deposition of liquid droplets onto a substrate. By printing nanoparticles-loaded fluids, it is possible to fabricate customized ceramic patterns with high resolution, reproducibility and automation [15]. Typically, inkjet printing produces droplets with volumes in the picoliter range (10-12 L), which are jetted from hundreds of nozzles with a diameter of a few tens of microns. Depositing picoliter-sized droplets allows controlling accurately the deposition of nanograms of particles, resulting in printing with high lateral resolution. Therefore, the formulation of finely dispersed nano-colloids for inkjet, i.e. nano-inks, paves the way for a further increase in the miniaturization capability of this technique. In particular, the reduced particle size allows decreasing the nozzle diameter and the droplet volume with a further enhancement in the lateral resolution. Moreover, thinner structures can be printed using smaller particles, increasing the thickness resolution. However, using nanomaterials raises remarkable processing difficulties compared to micron-sized structures. One of the reasons for this is their large surface area, which usually leads to a strong increase of the inter-particle interaction, leading, eventually, to aggregation [16]. The preparation of monodispersed nano-colloids from dry powders is particularly challenging and often requires intense deagglomeration treatments and an extensive use of stabilizing agents. The formulation of aqueous inks with suspended nanoparticles has already been successfully demonstrated for Ag colloids [17,18]. In particular, stable inks with a solid loading up to 45 % wt of Ag nanoparticles were prepared with the aid of a polyethilenoxide-polypropilenoxide co- polymer [17]. On the other hand, inkjet printing of several metal oxide nanoparticles (Al2O3, TiO2, WO3) was reported to be more challenging [19 -- 23]. In particular, the ink formulation often required the use of low concentrated inks and a careful ink preparation for avoiding large agglomerates. For example, it was shown that dispersing commercial TiO2 nanopowders involves an ultrasound treatment and a proper selection of the dispersant, despite a low particle content of 1-3 %wt [21,23]. Nonetheless, these efforts do not always lead to singularly dispersed nanoparticles, as a certain degree of agglomeration was observed in some cases [21,23,24] One of the strategies to overcome these problems is synthesizing nanoparticles directly dispersed in a liquid medium. Continuous hydrothermal synthesis (CHS) represents a powerful route for the synthesis of various materials in the form of a liquid dispersion of nanosized particles with a narrow size distribution. An additional feature that increases the effectiveness of this method is operating in supercritical conditions [25 -- 27]. In a typical CHS reactor, a stream of a precursor solution is continuously mixed with a stream of water above its critical point [28]. At the interface between the two flows, the precursor solution reaches rapidly supersaturation resulting in the abundant formation of nucleation sites. A large number of nanoparticles, for a capability in the range of tens of grams per minute, grow from these sites and are transported and collected downstream in an all-wet process. The reaction stoichiometry can be also controlled during synthesis by regulating the streams compositions and flows, providing constant synthesis conditions and uniform products. Moreover, CHS can be scaled-up [29] without handling dry nano-powders at any step of the processing. Due to their highly dispersed particles, low viscosity and good scalability, CHS products have the potential to act as precursors for nano-inks for inkjet printing [14,30] On the other hand, controlling the rheology of a nano-dispersion is a non-trivial task, especially considering the extreme shear variations experienced by a fluid during inkjet printing [31,32]. The fluid dynamics of this process has been studied both from a theoretical [33 -- 35] and experimental point of view [15,36,37]. Most of the experimental studies focus on particle-free fluids [36,38] or solid loaded inks with particles of hundreds of nanometers [39,40]. However, nano-fluids are known to have a different rheology than conventional fluids [41], which might affect inkjet printing [42]. In this work, we investigated the conversion of yttria stabilized zirconia (YSZ) nano-dispersions produced by continuous hydrothermal synthesis into nano-inks for inkjet printing. YSZ is one of the most studied ceramic ionic conductors and its properties are exploited in several applications and with different microstructures; e.g. dense electrolytes or porous backbones for electrodes and catalysts [43 -- 45]. While inkjet printing opens to the possibility of fabricating application-tailored structures and apply novel designs [46,47], the colloidal properties of nano-dispersions need to be tailored after synthesis by CHS. Transferring CHS products to inkjet printing represents a possible strategy for taking advantage of nano-sized YSZ, which allows a lower sintering temperature [48] and a higher printing resolution. 2. Experimental Two CHS reactors were utilized for the synthesis of YSZ nano-colloids in water. The reactor built at DTU Energy and the procedure for synthesis of YSZ had been described elsewhere [25]. Briefly, a precursor stream of 0.184 mol L-1 Zr(NO3)4 and 0.032 mol L-1 Y(NO3)3 with a flow rate of 10 mL min-1 was mixed with a stream of supercritical water (equivalent to 25 mL min-1 at room temperature) at 397 °C and 270 bar. The reactant solution was heated to approx. 390 °C after mixing before cooling down to room temperature. After collection, the products were washed and concentrated by centrifuging the dispersion, and redispersing particles in DI water. The large scale YSZ production at Promethean Particles Ltd. started from yttrium nitrate and zirconium acetate as precursors, which were mixed with water in a feedstock solution. The precursor stream was pumped at 20 mL min-1 in the reactor and mixed with a water flow in supercritical conditions at 375 °C and 241 Bar. The mixing zone is engineered with a high- pressure tube-in-tube counter current mixing system. Samples were depressurized and cooled down before increasing the particle concentration by tangential flow filtration [49]. Different batches were produced with a final solid loading in the range 5.5 ±1 %wt using the two reactors. The solid loading was evaluated by evaporating the solvent from the final dispersion. The as- produced particles were characterized by TEM, using a Jeol JEM 3000F microscope. The particle size distribution (PSD) of a synthesized YSZ dispersion was measured by Dynamic Light Scattering (DLS) with a Malvern Zetasizer Nano on a concentrated sample with a particle concentration of 6.5 %wt. The ζ-potential was measured on a synthesized YSZ dispersion after concentration with a solid loading of 4.5 %wt. The ζ-potential was monitored for 30 minutes at room temperature using a Zeta Probe Analyzer from Colloidal Dynamics. The concentrated dispersions showed little sedimentation after 72 hours on the shelf, and the solid deposit could be dispersed by gently stirring even letting the dispersion settle for more than 4 weeks. Nano-dispersions produced by CHS were the starting materials for the formulation of the inks. The printability parameter Z was used to ensure printability and evaluate the properties of the inks, as reported in other papers [50]. This factor was first proposed by Fromm [33] as a figure of merit for describing the jetting behavior of a fluid. Z is inverse of the Ohnesorge number, Oh, which is independent from the velocity of the fluid and contains several properties of the ink in the form: Z = 1/Oh = (𝜌 ∙ 𝜎 ∙ 𝑎)1/2 / 𝜂 (1), where ρ is the density, σ is the surface tension, η is the viscosity, and a is the characteristic length, which is typically considered as the nozzle diameter [15]. Polyvinylpyrrolidone (PVP) K15 with an average molecular weight of 10K g mol-1 (Applichem) and PVP with average molecular weight of 360K g mol-1 (Sigma-Aldrich) were both used to regulate the viscosity of the ink. 2,4,7,9-Tetramethyl-5-decyne-4,7-diol ethoxylate (TMDE, Sigma-Aldrich) was used as a surfactant for controlling the surface tension. The pH was adjusted using HCl 36 %wt (Alfa Aesar). All the chemicals were used as received. In a typical ink preparation, the CHS dispersions were first sonicated using a sonic probe (Hielscher UP200ST) with a 3 minutes burst at 50% of amplitude. Then, PVP and HCl were dissolved under magnetic stirring and before printing TMDE was added. Viscosity was measured with a rheometer (Anton Paar, MCR 302) in rotational mode and at a constant temperature of 21 °C. Rheological measurements were carried out using a plate-plate system with a diameter of 50 mm and a gap distance of 0.6 mm. Analyzes were performed at increasing shear rates up to 1000 s-1. The η values at the highest shear rate were used for the calculation of the printability according to equation (1). Surface tension was measured with a bubble pressure tensiometer (BP50, KRÜSS). Inks were jetted and printed using a Pixdro LP50 inkjet printer equipped with DMC disposable piezoelectric printheads from Dimatix. These printheads have 16 nozzles with a 21.5 μm diameter and a nominal droplet volume of 10 pL. For each ink, the waveform actuating the piezoelectric elements responsible for the droplet ejection was optimized starting from a standard trapezoidal pulse. This waveform was characterized by a dwell time and fall time of 10 and 5 μs respectively. Optimized printing was achieved by changing the filling time for each ink in a range of 4-7 μs and the maximum voltage between 40-60 V. A jetting frequency of 1000 Hz was used for all the tests. Before printing, the ink was filtered using a syringe filter with a 700 nm mesh. Inks were printed on a pre-sintered YSZ/NiO composite produced by tape casting as described elsewhere [51]. The deposition was carried out placing droplets in squared arrangement at different distances in order to achieve an optimum compromise between substrate coverage and minimum overlap. For this specific substrate, the optimal linear droplet density was 500 dots per inch in the x and y directions. In the case of multilayer printing, ca. 120 s of drying time was applied between the depositions of consecutive layers. YSZ samples were sintered at three different maximum temperatures in air: 800, 1000 and 1200 °C. The following thermal profile was used: 0 -- 600 °C at 0.25 °C min-1; dwell for 4 h; 600 -- Tmax °C at 1 °C min-1; dwell for 6 h; cool to room temperature at 1.67 °C min-1. The SEM characterization of the sintered samples was carried out with a Zeiss MERLIN scanning electron microscope. Top-view images were taken on carbon coated samples while cross sections were observed after embedding samples in epoxy followed by polishing. The porosity of the sintered samples was estimated by image analysis with the software ImageJ [52]. 3. Results and discussion 3.1 Synthesis and inks formulation Water dispersions of YSZ nanoparticles were produced in supercritical conditions by heating and pressurizing the water stream at 375 °C and 241 bar in the Promethean reactor, and 397 °C and 270 bar in the DTU reactor. Figure 1 gives an overview of the typical structure of the CHS reactor, and the particles morphology that were produced in the DTU process. Figure 1: Sketch of the continuous hydrothermal reactor (Promethean Particles LTD.) with a magnified view of the mixing zone (A), TEM image of the YSZ nanoparticles produced at DTU Energy (B), and Fourier transform of a TEM image at low magnification with randomly distributed particles (C). Figure 1(C) reports the Fourier transform used for calculating the (220) and (111) planes distances. Figure 1(D) shows the DLS analysis of a synthesized YSZ dispersion. Particularly, Figure 1(A) shows a sketch of the core components of a continuous hydrothermal reactor operating in supercritical conditions, with a magnified view of the mixing zone. In the configuration showed here, a water stream is first pressurized and heated above its supercritical point and is then continuously mixed with a metal salt stream in a counter-flow fashion [28]. At the interface between the two streams, dehydration and hydrolysis reactions take place leading to the formation of YSZ particles. The morphology for particles synthesized using the DTU reactor was assessed by electron microscopy and is shown in Figure 1(B) and 1(C). The average particle size is 10 nm and each particle consists of a single crystal. Figure 1(C) shows the crystallographic characterization carried out using the fast Fourier transform of an image containing a large number of randomly oriented particles. Two lattice spacings were calculated with values of 2.9(4) and 1.8(1) Å, in good agreement with the crystallographic distances along the [220] and [111] directions of cubic YSZ. TEM characterization especially revealed very similar structures between three different batches that were characterized. Besides the particles in Figure 1, YSZ nanoparticles produced at Promethean by using supercritical water at 375 °C and 241 bar were observed. Despite the use of different CHS conditions, neither morphological nor crystallographic differences were appreciated at the TEM for the material produced in the two reactors. The ink formulation was carried out bearing in mind the 1-10 range of the printability Z reported by Derby [15] and Reis [37], as already done in our previous works [50]. At the same time, a wider range of Z was also explored for assessing the modifications needed to make a CHS product printable. YSZ nano-dispersions were synthesized in water and concentrated to a final solid loading of 5.5 ±1 %wt by water evaporation or tangential flow water filtration. Due to the low particle concentration, viscosity (η) and surface tension (σ) were comparable with the values for pure water, i.e. 1 mPa s and 73 mN m-1 respectively. In particular, the as-produced nano-dispersions showed a surface tension of 71 mN m-1 and a viscosity of 1.2 mPa s at a shear rate of 1000 s-1. These results indicated only a weak impact of the presence of nanoparticles on these properties, leading to a printability value as high as Z = 33. These dispersions showed very little agglomeration and a limited reversible sedimentation after 72h. Figure 1(D) reports the particle size distribution measured by DLS, indicating that the dispersion consisted of 6 -- 20 nm particles with a maximum volume percentage at 10 nm, in good agreement with the TEM characterization. Above 30 nm up to 1 μm, only a small and broad peak with a maximum at 50 nm was visible, indicating substantial absence of aggregates that could influence the printing process. The good dispersion of the particles was explained by measuring the ζ-potential. This analysis resulted in a value of 62 ± 2 mV, in the typical range of dispersions with a strong electrostatic stabilization. From this starting composition, inks with different Z values were prepared to investigate the jetting behavior of YSZ nanoparticle-loaded fluids. Figure 2: Rheological optimization of the YSZ nano-dispersions. Measurement of the surface tension before and after the addition of TMDE (A), viscosities at increasing shear rates and resulting Z values at 1000 s-1 rate (B). The red line marks the surface tension and viscosity recommended by the printhead manufacturer for an optimal jetting. The ink formulation is summarized in Figure 2 and was started with lowering  to the optimal surface tension recommended by the printhead manufacturer, i.e. 30 mN m-1. Figure 2(A) indicates that adding 2 %wt of TMDE allowed reaching an optimal value of the surface tension without increasing the viscosity (black triangles in Figure 2(A) and (B)). In the following steps, the surface tension had been kept constant to 30 mN m-1 while the viscosity was varied. Since the printability dependency with viscosity and surface tension is 1/η and √𝜎 respectively, we chose to control η. Hence, a larger Z interval can be achieved by smaller modification of the viscosity. Several YSZ nano-inks with different viscosities were formulated using PVP as an additive after lowering σ with TMDE. Figure 2(B) reports the viscosities of these inks at increasing shearing rates, highlighting a slight self-thinning behavior of the fluids. This weak viscosity decrease is noteworthy because it suggests a limited agglomeration of particles in the fluid, which is unexpected considering the small particle size and their large surface area [16]. The all-wet synthesis of YSZ nanoparticles likely contributed to the prevention of substantial aggregation. Inks having viscosities measured at 1000 s-1 ranging from 1.2 to 17 mPa s were prepared in this way (Figure 2(B)). The viscosity plots are labelled with the Z value of each ink, which covers a wide range going from 1.6 to 25. It can be noted that the viscosity recommended by the printhead manufacturer translates into a Z value of about 3. All these inks were loaded into the cartridges and subsequently mounted on the printer. 3.2 Jetting behavior and printing Jetting tests were conducted to analyze the jetting behavior of the formulated inks. For each ink, the waveform was optimized for reaching the best droplet shape and the most reproducible droplet generation. This always resulted in a trapezoidal electrical pulse where only the duration of the filling ramp and the maximum voltage were found to have a significant effect on the jetting. Therefore, during the waveform optimization, dwell and fall times were kept constant at 10 and 5 μs, respectively. On the other hand, the duration of the filling ramp and the value for the maximum voltage changed for each ink and were comprised between 4-7 μs and 40-60 V, respectively. In particular, for low Z inks shorter filling ramps and higher voltages were required than for high Z inks. Figure 3: Jetting behavior of the formulated particle-loaded inks with different Z values (A). Jetting behavior of particle-free inks with similar Z values of the particle-loaded inks (B). Both (A) and (B) report a droplet picture, the droplet volume V and speed v obtained at each Z. Figure 3 summarizes the results obtained with several inks, which are displayed here as points defined by different Z values. For particle-loaded inks, Figure 3(A) describes their jetting behavior showing droplet volume, speed and shape obtained with an optimized waveform. To investigate the effect of nanoparticles on the jetting behavior, a similar set of experiments was carried out formulating particle-free inks (Figure 3(B)). Mixtures of water, PVP and TMDE were prepared to explore a printability range similar to the particle-loaded inks, using the same components of their liquid phase. Interestingly, particle-loaded inks with Z comprised between 4.4 and 20 allowed obtaining single, round shaped droplets. On the opposite at Z = 25 the formation of a satellite droplet was unavoidable, while it was impossible to achieve stable jetting with inks having Z values equal to 1.6 and 3.0 (Figure 3(A)). For particle-free inks, Figure 3(B) shows that inks with low Z, i.e. high viscosity, were jettable and produced good quality droplets. An optimal jetting was observed for Z values from 1.6 to 14, while at Z = 25 satellite droplets were formed. For both inks, the droplet volume followed a similar increasing trend with the fluid viscosity. This last observation is in agreement with Mogalicherla et al. [42], but there is no accordance in literature on the relation between Z and drop dimension [53]. In summary, for particle-loaded inks a good printability range was defined by Z values between 4.4 and 20. Fluids with higher viscosities and Z < 4 resulted in no jetting, while for Z > 20 multiple droplet formation was observed. The comparison with particle-free inks indicated that the no-jetting zone found at Z < 4 was not an artifact, but a reproducible behavior induced by the presence of nanoparticles in the fluid. The good printability range for particle-loaded inks is in disagreement with previous experimental observations and theoretical calculations. In particular, modeling and simulation studies indicated that the Z range for good jetting is 1 < Z < 10 [37]. Deposition of zirconia by inkjet printing has been reported also using reactive inks based on sol-gel chemistry [54] and it had been demonstrated that inks with Z between 1 and 2 were printed successfully. Our experiments with particle-free inks confirmed these results but the addition of nanoparticles avoided jetting for Z < 4. This lowest bound for Z is in accordance with the work of Jang et al. who observed a good printability range of 4 < Z < 14 [36]. On the other hand, the top limit of the printability range is higher than the values observed in both theoretical and experimental early studies of inkjet printing [36,37,39] . It is noteworthy that this significant modification of the printability range was caused by a very small nanoparticle concentration. We can thus conclude that the particle nanosize plays a crucial role on the fluid dynamic of the inkjet process. Due to their small size and large surface area, a small quantity of particles with a diameter of 10 nm forms a high amount of solid-liquid interfaces. These interfaces might represent discontinuities into the liquid medium that can affect the transmission of mechanical stresses through the fluid. The addition of larger particles was previously reported to extend the printability range towards higher values. However, a solid loading of 24 %vol leads to an increase of the printability to Z = 16.7 [39], while our results were obtained with a concentration of 0.9 %vol. Our results are in better agreement with the work of Mogalicherla et al., where good jetting was observed at Z as high as 30 using an ink prepared with 1.25 %vol of Al2O3 nanoparticles [42]. Figure 4: separated splats (A), honeycomb pattern (B) and continuous squared layer (C) all obtained using the ink with Z = 20 and by overprinting 10 times with a linear density of 500 dots per inch. After studying the jettability of the nano-inks, those results were exploited for patterning the substrate with YSZ nanoparticles. Firstly, printing experiments were carried out using the ink formulation with Z = 20, in order to understand the optimal conditions for the fabrication of patterns and continuous layers. In particular, several different dots dispositions were tested to achieve the best overlap between the splats. Indeed, when a droplet impacts on the substrate, it generates a splat with a diameter depending on several parameters: droplet volume and speed, ink viscosity, ink surface tension and surface energy of the substrate. The splat diameter corresponds to the smallest feature that can be printed and therefore determines the highest possible lateral resolution for a specific set of conditions. This value was measured by analyzing the splat diameter of separated dots and it resulted to be 70 μm on a sintered YSZ/NiO composite, see Figure 4(A). The droplet volume and splat diameter indicated that in our inkjet process, it was possible to position less than 0.5 ng of nanoparticles in an area of 0.04 mm2. Avoiding accumulation of material and reaching a smooth surface is important for reducing defects during sintering. The optimal arrangement consisted of dots in a squared disposition, as in Figure 4(A), printed with a linear density of 500 dots per inch in both x and y directions. It was possible to draw complex patterns in these conditions as the honeycomb reported in Figure 4(B) and uniform continuous layers as in Figure 4(C). Due to the dependence of the lateral resolution from the surface energy of the substrate, these optimized parameters needed to be adjusted for each different substrate. Using these optimized conditions, square shaped samples (4x4 mm2) were deposited for studying the sintering of YSZ nanoparticles in constrained conditions. To this scope, 10 layers were overprinted onto a pre-sintered substrate. 3.3 Sintering A first set of experiments was carried out to investigate the sintering of YSZ nanoparticles in free conditions at 800, 1000 and 1200 °C in air. Figure 5: microstructure of YSZ nanoparticles after free sintering at 800 °C (A), 1000 °C (B), 1200 °C (C) , and cross-section of the printed film after sintering in constrained conditions at 1000 °C (D). Figure 5 shows the microstructures obtained in these experiments. Figures 5 A-C show the top view of the free sintering samples at 800 °C, 1000 °C and 1200 °C. These images highlight a tendency of the nanoparticles towards necking and grain growth, which is already visible at 800 °C. At this temperature, sintering was already at an advanced stage as substantial necking of the particles was observed leaving a residual porosity of ca. 9 %vol. Grain growth was also measured, as the gain size increased from the starting 10 nm to about 25-30 nm. At 1000 °C, residual porosity was further decreased to ca. 2 %vol and particle started coalescing as a consequence of the grain growth. Full densification was reached at 1200 °C, where we observed the formation of crystallites up to 200 nm with flat grain boundaries. This last feature is a typical evidence of the complete relaxation of the material as a result of an intense mass diffusion. Square shaped layers, as in Figure 4(C), were used for studying the sintering behavior of the material in constrained conditions. Each sample was prepared by printing 10 layers of YSZ nanoparticles on a pre-sintered substrate, so that these films experienced a shrinkage due to the sintering and an opposite stress due to the thermal expansion of the substrate. Figure 5(D) shows the cross section of a sample sintered at 1000 °C. This picture reveals the microstructure of the printed material, which contained extremely fine pores (10 - 40 nm). Similarly to the sample sintered at 1000 °C in free conditions (Figure 5(B)), particles connected into a continuous network with a nearly full densification (estimated residual porosity 2 %vol). The dimension of the superficial porosity was two orders of magnitude higher than the starting particles size. Nonetheless, infiltration through the substrate was not observed and particles formed a continuous layer which covered the whole support. Moreover, the film obtained after sintering at 1000 °C was self-levelling, i.e. for ten printed layers its thickness varied between 250 and 500 nm to smooth the substrate morphology. The major reason for this thickness variation is the pronounced roughness of the substrate, which presented pinning features and large cavities. The smoothing effect is visible in Figure 5(D), which highlights the flat top surface of the inkjet printed material in contrast with the irregular substrate. The SEM analysis showed this behavior for all the printed samples. Nanoparticles arranged in such a way that the material adapted to the irregular surface keeping a constant level over pores and asperities. 4. Conclusions We here report for the first time how to transfer YSZ nano-dispersions produced by continuous hydrothermal synthesis in inks for inkjet printing. The all-wet process for the ink preparation avoided a significant aggregation resulting in continuous networks of particles after sintering. Furthermore, the presence of particles with extremely reduced size affected the jetting behavior of the fluids compared to particle-free inks with similar compositions. In particular, nanoparticles hindered the generation of droplets for inks with Z < 4, and at the same time prevented the formation of satellite droplets for Z as high as 20. The transfer of CHS products into the inkjet process was completed by optimizing the printing process with the formulated inks. This allowed writing patterns with a lateral resolution of about 70 µm. Inkjet printed YSZ layers were sintered on a porous pre-sintered substrate with a very rough surface showing a beneficial self-levelling effect. This allowed fabricating continuous layers with a thickness between 250 nm -- 500 nm depending on the local structure of the surface. Nevertheless, no infiltration of the nanoparticles took place although the porosity of the substrate was orders of magnitude larger than the starting particle size. YSZ nanoparticles showed also a strong tendency toward sintering. In particular, necking and grain growth were already visible at temperatures as low as 800 °C, while full densification was achieved at 1200 °C. A further evidence of the remarkable reactivity of these nanoparticles was given by their sintering in constrained conditions, where the particles were deposited on pre-sintered substrates. Acknowledgements This project has partially received funding from the Fuel Cells and Hydrogen 2 Joint Undertaking under grant agreement No 700266. This Joint Undertaking receives support from the European Union's Horizon 2020 research and innovation program and Hydrogen Europe and N.ERGHY. References [1] G. Li, Z. Tang, Noble metal nanoparticle@metal oxide core/yolk-shell nanostructures as catalysts: recent progress and perspective, Nanoscale. 6 (2014) 3995 -- 4011. doi:10.1039/c3nr06787d. [2] A. Rogov, Y. Mugnier, L. Bonacina, Harmonic nanoparticles: noncentrosymmetric metal oxides for nonlinear optics, J. Opt. 17 (2015) 33001. doi:10.1088/2040-8978/17/3/033001. [3] S.T. Kochuveedu, Y.H. Jang, D.H. Kim, A study on the mechanism for the interaction of light with noble metal-metal oxide semiconductor nanostructures for various photophysical applications, Chem. Soc. Rev. 42 (2013) 8467. doi:10.1039/c3cs60043b. [4] B. Wetzel, P. Rosso, F. Haupert, K. Friedrich, Epoxy nanocomposites - fracture and toughening mechanisms, Eng. Fract. Mech. 73 (2006) 2375 -- 2398. doi:10.1016/j.engfracmech.2006.05.018. [5] N. Satoh, T. Nakashima, K. Kamikura, K. Yamamoto, Quantum size effect in TiO2 nanoparticles prepared by finely controlled metal assembly on dendrimer templates, Nat. Nanotechnol. 3 (2008) 106 -- 111. doi:10.1038/nnano.2008.2. [6] A.T. Bell, G.C. Bond, D.T. Thompson, M. Valden, X. Lai, D.W. Goodman, T. Blasko, J.M.L. Nieto, K. Chen, A.T. Bell, E. Iglesia, T. Koyama, T. Komaya, P.L. Gai, M. Weyland, G. Durscher, N.D. Browning, S.J. Pennycook, F. Besenbacker, P.C. Thune, J. Loss, D. Wonter, P.J. Leustra, J.W. Niemantsverdriet, J. Corker, V. Vidal, A. Theolier, J. Thivolle-Cazat, J.-M. Basset, C. Nozkaki, C.G. Lugmair, A.T. Bell, T.D. Tilley, D. Kolb, K.P. De Jong, J.W. Geus, The impact of nanoscience on heterogeneous catalysis., Science. 299 (2003) 1688 -- 91. doi:10.1126/science.1083671. [7] V. Polshettiwar, R.S. Varma, Green chemistry by nano-catalysis, Green Chem. 12 (2010) 743. doi:10.1039/b921171c. [8] D.Z. De Florio, V. Esposito, E. Traversa, R. Muccillo, F.C. Fonseca, Master sintering curve for Gd-doped CeO2 solid electrolytes, J. Therm. Anal. Calorim. 97 (2009) 143 -- 147. doi:10.1007/s10973-009-0238-6. [9] J.A. Glasscock, V. Esposito, S.P. V Foghmoes, T. Stegk, D. Matuschek, M.W.H. Ley, S. Ramousse, The effect of forming stresses on the sintering of ultra-fine Ce0.9Gd0.1O2-?? powders, J. Eur. Ceram. Soc. 33 (2013) 1289 -- 1296. doi:10.1016/j.jeurceramsoc.2012.12.015. [10] V. Esposito, E. Traversa, Design of electroceramics for solid oxides fuel cell applications: Playing with ceria, J. Am. Ceram. Soc. 91 (2008) 1037 -- 1051. doi:10.1111/j.1551- 2916.2008.02347.x. [11] E. Olevsky, T.T. Molla, H.L. Frandsen, R. Bjørk, V. Esposito, D.W. Ni, A. Ilyina, N. Pryds, Sintering of multilayered porous structures: Part I-constitutive models, J. Am. Ceram. Soc. 96 (2013) 2657 -- 2665. doi:10.1111/jace.12375. [12] D.W. Ni, E. Olevsky, V. Esposito, T.T. Molla, S.P. V Foghmoes, R. Bjørk, H.L. Frandsen, E. Aleksandrova, N. Pryds, Sintering of multilayered porous structures: Part II- experiments and model applications, J. Am. Ceram. Soc. 96 (2013) 2666 -- 2673. doi:10.1111/jace.12374. [13] D.W. Ni, C.G. Schmidt, F. Teocoli, A. Kaiser, K.B. Andersen, S. Ramousse, V. Esposito, Densification and grain growth during sintering of porous Ce0.9Gd0.1O1.95 tape cast layers: A comprehensive study on heuristic methods, J. Eur. Ceram. Soc. 33 (2013) 2529 -- 2537. doi:10.1016/j.jeurceramsoc.2013.03.025. [14] T.H.J. Van Osch, J. Perelaer, A.W.M. De Laat, U.S. Schubert, Inkjet printing of narrow conductive tracks on untreated polymeric substrates, Adv. Mater. 20 (2008) 343 -- 345. doi:10.1002/adma.200701876. [15] B. Derby, Inkjet printing ceramics : From drops to solid, 31 (2011) 2543 -- 2550. doi:10.1016/j.jeurceramsoc.2011.01.016. [16] S. Kango, S. Kalia, A. Celli, J. Njuguna, Y. Habibi, R. Kumar, Surface modification of inorganic nanoparticles for development of organic-inorganic nanocomposites - A review, Prog. Polym. Sci. 38 (2013) 1232 -- 1261. doi:10.1016/j.progpolymsci.2013.02.003. [17] A. Kosmala, Q. Zhang, R. Wright, P. Kirby, Development of high concentrated aqueous silver nanofluid and inkjet printing on ceramic substrates, Mater. Chem. Phys. 132 (2012) 788 -- 795. doi:10.1016/j.matchemphys.2011.12.013. [18] A. Kosmala, R. Wright, Q. Zhang, P. Kirby, Synthesis of silver nano particles and fabrication of aqueous Ag inks for inkjet printing, Mater. Chem. Phys. 129 (2011) 1075 -- 1080. doi:10.1016/j.matchemphys.2011.05.064. [19] M. Arin, P. Lommens, S.C. Hopkins, G. Pollefeyt, J. Van der Eycken, S. Ricart, X. Granados, B. a Glowacki, I. Van Driessche, Deposition of photocatalytically active TiO$_2$ films by inkjet printing of TiO$_2$ nanoparticle suspensions obtained from microwave-assisted hydrothermal synthesis., Nanotechnology. 23 (2012) 165603. doi:10.1088/0957-4484/23/16/165603. [20] S. Lee, T. Boeltken, A.K. Mogalicherla, U. Gerhards, P. Pfeifer, R. Dittmeyer, Inkjet printing of porous nanoparticle-based catalyst layers in microchannel reactors, Appl. Catal. A Gen. 467 (2013) 69 -- 75. doi:10.1016/j.apcata.2013.07.002. [21] R. Cherrington, D.J. Hughes, S. Senthilarasu, V. Goodship, Inkjet-Printed TiO2 Nanoparticles from Aqueous Solutions for Dye-Sensitized Solar Cells (DSSCs), Energy Technol. 3 (2015) 866 -- 870. doi:10.1002/ente.201500096. [22] P.J. Wojcik, L. Santos, L. Pereira, R. Martins, E. Fortunato, Tailoring nanoscale properties of tungsten oxide for inkjet printed electrochromic devices, Nanoscale. 7 (2015) 1696 -- 1708. doi:10.1039/C4NR05765A. [23] I. Fasaki, K. Siamos, M. Arin, P. Lommens, I. Van Driessche, S.C. Hopkins, B.A. Glowacki, I. Arabatzis, Ultrasound assisted preparation of stable water-based nanocrystalline TiO2 suspensions for photocatalytic applications of inkjet-printed films, Appl. Catal. A Gen. 411 -- 412 (2012) 60 -- 69. doi:10.1016/j.apcata.2011.10.020. [24] P.D. Angelo, R. Kronfli, R.R. Farnood, Synthesis and inkjet printing of aqueous ZnS:Mn nanoparticles, J. Lumin. 136 (2013) 100 -- 108. doi:10.1016/j.jlumin.2012.10.043. [25] P. Zielke, Y. Xu, S.B. Simonsen, P. Norby, R. Kiebach, Simulation, design and proof-of- concept of a two-stage continuous hydrothermal flow synthesis reactor for synthesis of functionalized nano-sized inorganic composite materials, J. Supercrit. Fluids. 117 (2016) 1 -- 12. doi:10.1016/j.supflu.2016.06.008. [26] T. Adschiri, K. Kanazawa, K. Arai, Rapid and Continuous Hydrothermal Crystallization of Metal Oxide Particles in Supercritical Water, J. Am. Ceram. Soc. 75 (1992) 1019 -- 1022. doi:10.1111/j.1151-2916.1992.tb04179.x. [27] T. Adschiri, Y. Hakuta, K. Sue, K. Arai, Hydrothermal synthesis of metal oxide nanoparticles at supercritical conditions, J. Nanoparticle Res. 3 (2001) 227 -- 235. doi:10.1023/A:1017541705569. [28] E. Lester, P. Blood, J. Denyer, D. Giddings, B. Azzopardi, M. Poliakoff, Reaction engineering: The supercritical water hydrothermal synthesis of nano-particles, J. Supercrit. Fluids. 37 (2006) 209 -- 214. doi:10.1016/j.supflu.2005.08.011. [29] R.I. Brief, R. Summaries, Sustainable Hydro thermal Manufacturing o f Nano materials Pro ject details, (2016) 1 -- 5. [30] M. Černá, M. Veselý, P. Dzik, C. Guillard, E. Puzenat, M. Lepičová, Fabrication, characterization and photocatalytic activity of TiO2 layers prepared by inkjet printing of stabilized nanocrystalline suspensions, Appl. Catal. B Environ. 138 -- 139 (2013) 84 -- 94. doi:10.1016/j.apcatb.2013.02.035. [31] A. Lee, K. Sudau, K.H. Ahn, S.J. Lee, N. Willenbacher, Optimization of experimental parameters to suppress nozzle clogging in inkjet printing, Ind. Eng. Chem. Res. 51 (2012) 13195 -- 13204. doi:10.1021/ie301403g. [32] X. Wang, W.W. Carr, D.G. Bucknall, J.F. Morris, High-shear-rate capillary viscometer for inkjet inks, Rev. Sci. Instrum. 81 (2010). doi:10.1063/1.3449478. [33] J.E. Fromm, Numerical Calculation of the Fluid Dynamics of Drop-on-Demand Jets, IBM J. Res. Dev. 28 (1984) 322 -- 333. doi:10.1147/rd.283.0322. [34] Q. Xu, O.A. Basaran, Computational analysis of drop-on-demand drop formation, Phys. Fluids. 19 (2007). doi:10.1063/1.2800784. [35] J. Eggers, E. Villermaux, Physics of liquid jets, Reports Prog. Phys. 71 (2008) 36601. doi:10.1088/0034-4885/71/3/036601. [36] D. Jang, D. Kim, J. Moon, Influence of fluid physical properties on ink-jet printability, Langmuir. 25 (2009) 2629 -- 2635. doi:10.1021/la900059m. [37] N. Reis, B. Derby, Ink Jet Deposition of Ceramic Suspensions: Experiments of Droplet Formation, Mater. Res. 625 (2000) 117 -- 122. doi:dx.doi.org/10.1557/PROC-624-65. [38] C. Gadea, D. Marani, V. Esposito, Nucleophilic stabilization of water-based reactive ink for titania-based thin film inkjet printing, J. Phys. Chem. Solids. 101 (2017) 10 -- 17. doi:10.1016/j.jpcs.2016.10.004. [39] E. Özkol, J. Ebert, R. Telle, An experimental analysis of the influence of the ink properties on the drop formation for direct thermal inkjet printing of high solid content aqueous 3Y-TZP suspensions, J. Eur. Ceram. Soc. 30 (2010) 1669 -- 1678. doi:10.1016/j.jeurceramsoc.2010.01.004. [40] M. Bienia, M. Lejeune, M. Chambon, V. Baco-Carles, C. Dossou-Yovo, R. Noguera, F. Rossignol, Inkjet printing of ceramic colloidal suspensions: Filament growth and breakup, Chem. Eng. Sci. 149 (2016) 1 -- 13. doi:10.1016/j.ces.2016.04.015. [41] M. Corcione, Empirical correlating equations for predicting the effective thermal conductivity and dynamic viscosity of nanofluids, Energy Convers. Manag. 52 (2011) 789 -- 793. doi:10.1016/j.enconman.2010.06.072. [42] A.K. Mogalicherla, S. Lee, P. Pfeifer, R. Dittmeyer, Drop-on-demand inkjet printing of alumina nanoparticles in rectangular microchannels, Microfluid. Nanofluidics. 16 (2014) 655 -- 666. doi:10.1007/s10404-013-1260-3. [43] H. Huang, M. Nakamura, P. Su, R. Fasching, Y. Saito, F.B. Prinz, High-Performance Ultrathin Solid Oxide Fuel Cells for Low-Temperature Operation, J. Electrochem. Soc. 154 (2007) B20. doi:10.1149/1.2372592. [44] J.M. Vohs, R.J. Gorte, High-performance SOFC cathodes prepared by infiltration, Adv. Mater. 21 (2009) 943 -- 956. doi:10.1002/adma.200802428. [45] F. Matei, C. Jiménez-Borja, J. Canales-Vázquez, S. Brosda, F. Dorado, J.L. Valverde, D. Ciuparu, Enhanced electropromotion of methane combustion on palladium catalysts deposited on highly porous supports, Appl. Catal. B Environ. 132 -- 133 (2013) 80 -- 89. doi:10.1016/j.apcatb.2012.11.011. [46] N.M. Farandos, L. Kleiminger, T. Li, A. Hankin, G.H. Kelsall, Three-dimensional Inkjet Printed Solid Oxide Electrochemical Reactors. I. Yttria-stabilized Zirconia Electrolyte, Electrochim. Acta. 213 (2016) 324 -- 331. doi:10.1016/j.electacta.2016.07.103. [47] J.C. Ruiz-Morales, A. Tarancón, J. Canales-Vázquez, J. Méndez-Ramos, L. Hernández- Afonso, P. Acosta-Mora, J.R. Marín Rueda, R. Fernández-González, Three dimensional printing of components and functional devices for energy and environmental applications, Energy Environ. Sci. (2017) 846 -- 859. doi:10.1039/C6EE03526D. [48] Q. Zhu, B. Fan, Low temperature sintering of 8YSZ electrolyte film for intermediate temperature solid oxide fuel cells, Solid State Ionics. 176 (2005) 889 -- 894. doi:10.1016/j.ssi.2004.12.010. [49] Pall, Introduction to Tangential Flow Filtration for Laboratory and Process Development Applications, (2014) 1 -- 13. [50] V. Esposito, C. Gadea, J. Hjelm, D. Marani, Q. Hu, K. Agersted, S. Ramousse, S.H. Jensen, Fabrication of thin yttria-stabilized-zirconia dense electrolyte layers by inkjet printing for high performing solid oxide fuel cells, J. Power Sources. 273 (2015) 89 -- 95. doi:10.1016/j.jpowsour.2014.09.085. [51] A. Hauch, M. Mogensen, A. Hagen, Ni/YSZ electrode degradation studied by impedance spectroscopy - Effect of p(H2O), Solid State Ionics. 192 (2011) 547 -- 551. doi:10.1016/j.ssi.2010.01.004. [52] C.A. Schneider, W.S. Rasband, K.W. Eliceiri, NIH Image to ImageJ: 25 years of image analysis, Nat. Methods. 9 (2012) 671 -- 675. doi:10.1038/nmeth.2089. [53] D. Kuscer, G. Stavber, G. Trefalt, M. Kosec, Formulation of an aqueous titania suspension and its patterning with ink-jet printing technology, J. Am. Ceram. Soc. 95 (2012) 487 -- 493. doi:10.1111/j.1551-2916.2011.04876.x. [54] C. Gadea, Q. Hanniet, A. Lesch, D. Marani, S.H. Jensen, V. Esposito, Aqueous metal- organic solutions for YSZ thin film inkjet deposition, J. Mater. Chem. C. (2017). doi:10.1039/C7TC01879G.
1907.03239	1	1907	2019-07-07T07:41:36	Deep learning-based quality filtering of mechanically exfoliated 2D crystals	[ "physics.app-ph", "physics.comp-ph", "physics.data-an" ]	Two-dimensional (2D) crystals are attracting growing interest in various research fields such as engineering, physics, chemistry, pharmacy and biology owing to their low dimensionality and dramatic change of properties compared to the bulk counterparts. Among the various techniques used to manufacture 2D crystals, mechanical exfoliation has been essential to practical applications and fundamental research. However, mechanically exfoliated crystals on substrates contain relatively thick flakes that must be found and removed manually, limiting high-throughput manufacturing of atomic 2D crystals and van der Waals heterostructures. Here we present a deep learning-based method to segment and identify the thickness of atomic layer flakes from optical microscopy images. Through carefully designing a neural network based on U-Net, we found that our neural network based on U-net trained only with the data based on 24 images successfully distinguish monolayer and bilayer MoS2 with a success rate of 70%, which is a practical value in the first screening process for choosing monolayer and bilayer flakes of MoS2 of all flakes on substrates without human eye. The remarkable results highlight the possibility that a large fraction of manual laboratory work can be replaced by AI-based systems, boosting productivity.	physics.app-ph	physics	Deep learning-based quality filtering of mechanically exfoliated 2D crystals Yu Saito1†#, Kento Shin2†, Kei Terayama1,3, Shaan Desai1, Masaru Onga4, Yuji Nakagawa4, Yuki M. Itahashi4, Yoshihiro Iwasa4.5, Makoto Yamada1,6,7, Koji Tsuda1,2,8 1 RIKEN Center for Advanced Intelligence Project (AIP), Tokyo 103-0027, Japan, 2 Department of Computational Biology and Medical Sciences, The University of Tokyo, Kashiwa 277-8561, Japan 3 Graduate School of Medicine, Kyoto University, Kyoto 606-8501, Japan 4 Quantum-Phase Electronics Center (QPEC) and Department of Applied Physics, The University of Tokyo, Tokyo 113-8656, Japan 5 RIKEN Center for Emergent Matter Science (CEMS), Wako 351-0198, Japan 6 Department of Intelligence Science and Technology, Kyoto University, Kyoto 606-8501, Japan 7PRESTO, Japan Science and Technology Agency (JST), 4-1-8, Honcho, Kawaguchi-shi, Saitama 332-0012 Japan 8 National Institute for Materials Science (NIMS), Center for Materials research by Information Integration, Tsukuba,305-0047, Japan *Corresponding author: [email protected] (Y.S.), [email protected] (K. Tsuda.) †These authors equally contributed to this work. # Present Address: California NanoSystems Institute, University of California, Santa Barbara, Santa Barbara, CA 93106 USA 1 ABSTRACT Two-dimensional (2D) crystals are attracting growing interest in various research fields such as engineering, physics, chemistry, pharmacy and biology owing to their low dimensionality and dramatic change of properties compared to the bulk counterparts. Among the various techniques used to manufacture 2D crystals, mechanical exfoliation has been essential to practical applications and fundamental research. However, mechanically exfoliated crystals on substrates contain relatively thick flakes that must be found and removed manually, limiting high-throughput manufacturing of atomic 2D crystals and van der Waals heterostructures. Here we present a deep learning-based method to segment and identify the thickness of atomic layer flakes from optical microscopy images. Through carefully designing a neural network based on U-Net, we found that our neural network based on U-net trained only with the data based on 24 images successfully distinguish monolayer and bilayer MoS2 with a success rate of 70%, which is a practical value in the first screening process for choosing monolayer and bilayer flakes of MoS2 of all flakes on substrates without human eye. The remarkable results highlight the possibility that a large fraction of manual laboratory work can be replaced by AI-based systems, boosting productivity. 2 Two-dimensional (2D) crystals1, such as graphene, transition metal dichalcogenides, and van der Waals heterostructures2 have been intensively studied in a wide range of research fields since they show significant properties that has not been observed in their bulk counter parts. Examples include materials engineering such as electronics and optoelectronics devices3 -- 7, solid state physics including superconductors8 and magnets9,10, chemistry11,12 and biomedical applications13. To manufacture such 2D crystals with atomic layer thickness, especially monolayer or a few-layers samples, mechanical exfoliation, chemical exfoliation, chemical vapor deposition and molecular beam epitaxy have been introduced11. Among them, mechanical exfoliation has been instrumental to 2D materials research because it enables us to obtain highly crystalline and atomic 2D layers as is exemplified by ultraclean and high mobility devices based on exfoliated 2D materials combined with the encapsulation by h-BN. Recently, high-throughput identification of various unexplored 2D materials via machine learning and the development of a machine for mechanically exfoliated 2D atomic crystals to autonomously build van der Waals superlattices14 have been reported. These advancements are suggestive of a new research direction for 2D materials that is aimed to explore efficiently enormous materials' properties in large scale using robotics and machine learning. In such a situation, a rapid and versatile method for layer number identification of mechanically exfoliated atomic- layer crystals on substrates is highly desirable in their fundamental research and practical applications. However, the bottleneck of using mechanical exfoliation is that, through the process, not only do we produce desirable atomic layers (mostly monolayer or bilayer) but also many impractical thicker flakes making it tough to quickly separate the useful layers from the unwanted. To identify the thickness of 2D crystals, several methods such as atomic force microscopy (AFM)15, Raman spectroscopy16,17, and optical microscopy (OM) 18,19 are used. AFM is one of the most versatile methods to measure the thickness of various 2D materials, 3 but it takes a relatively long time to measure one region. In addition, the measured value strongly depends on the offset conditions15 due to, for example, bubbles beneath the sample. OM is nowadays a widely utilized technique to measure the thickness of 2D crystals based on the optical contrast between atomic layers and the substrate20. In fact, we can determine the thickness of 2D atomic layers using the contrast difference between the flakes and substrate obtained from the brightness profile of color or grayscale images18. The methods explained above, however, need manual work and take a relatively long time to identify atomic layer thickness, and are therefore inappropriate for studying various kinds of materials. Recently emerged deep learning, a machine learning technique via deep neural networks, has shown immense potential for regression and classification tasks in a variety of research fields20 -- 25. In particular, deep neural networks have been very successful in image recognition tasks such as distinguishing images of cats and dogs with high accuracy, and also several physical problems in theoretical physics, for example, detection of phase transitions26,27 and searching for exotic particles in high-energy physics28. Given this, deep neural networks provide an alternative pathway to quickly identifying layer thicknesses of 2D crystals from OM images. Here, we introduce a versatile technique to autonomously segment and identify the thickness of 2D crystals via a deep neural network. Using a deep neural network architecture, we reproduced the images of 2D crystals from the augmented data based on 24 OM images of MoS2 used as training data and found that both the cross validation score and accuracy rate for the test images through deep neural networks is surprisingly over 70%, which means that our model can distinguish monolayer and bilayer with a practical success rate for initial screening process. The present results suggest the deep neural network can become another promising way to quickly identify the thickness of various 2D crystals in an autonomous way, suggesting 4 a large fraction of manual laboratory work can be dramatic decreased by replacing AI-based systems. Figure 1 shows an overview of data collection using the OM and the training architecture of the deep neural network. First, we prepared the OM images of MoS2 flakes. Bulk single crystals of MoS2 crystals (SPI Supplies, USA) were employed for the preparation of mechanically exfoliated 2D atomic layers, which were then transferred onto the 300-nm- thick SiO2/ Si substrates in the air. The bright-field OM (BX 51, Olympus) was used to locate and take pictures (optical microscope images) of the MoS2 thin flakes on the Si/SiO2 substrate with one hundred magnification. Each OM image includes different number of flakes with various thicknesses under the different light intensity conditions, we then confirmed the thickness and layer number of each samples using AFM and contrast based on OM images, and prepared 35 OM images, which were randomly divided into the training data set (24 images) and test images (11 images), the latter of which were prepared to compare prediction with a non-expert human as discussed below, by dividing into three regions of monolayer (blue), bilayer (green) and others (black). It is noted that as we mention below, we used 960 images which are augmented from 24 images as a training data set. Next, we constructed one of the deep neural network architectures, U-Net, which is based on the fully convolutional encoder-decoder network29.The encoder extracts the small feature map from the input image by convolution and pooling layers, and decoder expands it to the original image size by convolution and up-sampling layers. Figure 2 shows an overview of our network. The encoder and decoder are composed of 14 and 18 layers respectively. The encoder consists of 4 repeated layers set which consist of 3×3 convolutions, each followed by a rectified linear unit (ReLU) activation and 2×2 max pooling with stride 2 for downsampling. At each downsampling step, the number of the feature map is doubled. The decoder consists of 4 repeated layers set which consist of 2×2 upsampling convolution and 3×3 convolution 5 followed by ReLU. We added 50% dropout layers after each of the first three upsampling convolution layers. At the final layer of the decoder, a 1×1 convolution converts the feature map to the desired number of classes and softmax activation is then applied. To transmit high- resolution information in the input images, the network has skip connections between each convolutional layer of the encoder and the corresponding upsampling layer of the decoder. Each skip connection simply concatenates all channels at the layer of the encoder with one of the decoders. The dimension, width × height × channels, of the input image is 512×512×3 and it changes to 256×256×64, 128×128×128, 64×64×256, 32×32×512, and 16×16×1024 at each downsampling step in the encoder respectively, and changes in reverse order through the upsampling steps in the decoder. In the present study, we used the augmented data based on 24 original OM images and corresponding segmentation maps to train the network and use 11 images to evaluate the performance. Here, we particularly choose the classification of MoS2 monolayer and bilayer because the number of training data set is limited to expand the classifications to thicker multilayer. We employed the data argumentation technique, which is one of learning techniques widely used for deep neural networks to improve learning accuracy and prevent overfitting 20,30 -- 33. By augmentation with randomly cropping, flipping, rotating and changing the color of the original images34, we increased the training data up to 960 data points. Indeed, this kind of random manipulation and augmentation of the original data is useful for averaging the difference of contrast and number of flakes in each original image. For the color changing augmentation, we converted images to grayscale or adjusted the contrast of the image. With this augmentation, we increased the data to 960 points. Finally, we normalized all pixels between 0 and 1 as preprocessing before training. We then used cross entropy as a loss function of the multi-class classification and also used a weighted cross entropy in order to balance the frequency of each class. The normal cross entropy 𝐸 is computed as 6 𝑁 𝐾 𝐸 = ∑ ∑ 𝑦𝑖𝑘 log(𝑥𝑖) . 𝑖=0 𝑘=0 Here, 𝑥𝑖 is the output of the softmax activation of the U-Net at pixel i, 𝑦𝑖 and 𝑦𝑖𝑘 are the one hot vector of the label at pixel i and k-th element (scalar quantity) of 𝑦𝑖, respectively, such that only the element at the position of the true label is 1 and the others are 0, K is the number of labels, and 𝑁 is the number of pixels in the image. Similarly, weighted cross entropy 𝐸𝑤 is computed as 𝑁 𝐾 𝐸𝑤 = ∑ ∑ 𝑤𝑘𝑦𝑖𝑘 log(𝑥𝑖) , 𝑖=0 𝑘=0 where 𝑤𝑘 is the weight of the class k. To compute the weight 𝑤𝑘, we use median frequency balancing35, where the weight assigned to a class is the ratio of the median of class frequencies computed on the entire training set divided by the class frequency. This implies that smaller classes in the training data get heavier weights. In the training, we employed minibatch SGD36,37 and Adam solver38. The learning rate, the batch size and the number of epochs were set to 0.00001, 1 and 100, respectively29. Also, we confirmed that learning rate, the batch size and the number of sizes is optimized by changing each parameter in the rage of 0.00001 ~ 0.001, 1 ~ 100 and 10~500, respectively. Figure 3 shows four examples of the original OM images, the segmented images and generated images using U-Net with weighted loss. The blue, green and black regimes show monolayer, bilayer and other parts including thicker flakes and substrate, respectively. It seems that the training neural network can pick up the information of color contrast (against back ground color of substrate) as well as the edge thickness/sharpness, which is indeed useful information for the real screening process. We calculated receiver operating characteristic (ROC) and precision-recall (precision-true positive rate) curves of two-class classifications for the monolayer/others (blue) and bilayer/others (red) identification and confirmed the learning 7 performance in Fig. 4a and b. The ROC curve shows sharply below the false positive rate of 0.1 and then saturates to 1.0 and precision-recall curve show the high precision value (> 90%) below recalls of 0.76 and 0.94 for monolayer and bilayer, respectively, both of which show the high performance of the U-Net with weighted loss. To perform further evaluations, we investigated and compared the difference of the cross-validation score and accuracy rate for the test images by changing the learning process using the grayscale images and/or contrast adjusted images. To evaluate the performance of the trained U-Net, we used both cross-validation score and an accuracy score on the test images. In the cross-validation, we performed three-fold cross validation, in which a mean of the pixel- wise accuracy in each class was used as an evaluation metric. On the other hand, in the test images, we prepared the 11 images (test images), each of which has problem region to be answered. To check whether the U-Net can predict the test image correctly or not, we predict the class of each pixel of the image by using U-Net and define the class whose pixel is most in the problem region as the predicted class of that region, then compare it with the true class. We summarize the cross-validation scores for each learning process in Table 1. The U-Net with weighted loss using contrast adjusted augmentation shows the highest score of all, 0.789, which is much higher than that of a normal U-Net without any options. The value of 0.789 is surprisingly high, considering that while a deep neural network needs thousand or even more training points, our case used the data based on 24 training data points. This remarkable result allows us to generate a practical number of training sets which can initially be prepared by lab works using OM. This result also suggests that once we prepare training data sets and perform leaning for a 2D material, we can obtain a tool that can quickly identify atomic layer thickness, monolayer/bilayer/other thicker parts, with an accuracy rate of almost 80 %, which is practically value that can be helpful for the initial screening process. More importantly, in practice, it is not necessary to check and distinguish of all monolayer or bilayer candidates on the 8 substrate, but just needed to pick up some amount of monolayer or bilayer with a high accuracy. According to the precision-recall curve in Fig. 3d, at miss rates of 24% and 6.0% (1-true positive rate), which means that the U-net miss the flakes with a target layer number (monolayer or bilayer), it can distinguish monolayer and bilayer with an accuracy of 90%. This value is practically high for the real experiments. These high success rates of the identifications mean that the present technique based on the U-net potentially can apply other 2D materials on various wafers because it is very rigid against external conditions (e.g., the number of flakes surrounding the target flakes and the light intensity of optical microscope.), and can detect sensitively the color contrast of the surface against back ground color and edge thickness from the optical microscope images. Finally, to compare the accuracy rate between U-Net and non-expert humans, we calculated the accuracy rate using 11 randomly selected test images (see Supplementary Materials). We found that the accuracy score of U-Net with weighted loss model is 0.733 in all cases, which is higher than the that of a normal U-Net model. Such a tendency is observed in cross validation score, which indicates that the U-Net with weighted loss is better than the normal U-Net for both segmentation and layer number identification tasks. We then compared the accuracy score with that of twelve researchers who were not familiar with 2D materials (non-expert human). The researchers were given three minutes to learn the training data set (see Supplementary Materials for details) and were then asked to respond with the layer numbers (mono- and bilayer) for each segment of new images, which were the same as the test images used in determining the accuracy score for the deep neural network. The accuracy score for humans was 0.67±0.11, which was comparable value to that of a U-Net with weighted loss. The present results suggest that the deep neural network based on U-Net with weighted loss is a new tool to rapidly and autonomously segment and identify number of layers of 2D crystals with an accuracy rate comparable to non-expert humans, indicating it can be an essential tool to significantly decrease manual work in the laboratory by boosting the first screening process, which has been usually done by human eyes. 9 In conclusion, we introduce a versatile technique to autonomously segment and identify the thickness of 2D crystals via a deep neural network. Constructing an architecture consisting of convolutions, U-Net, we reproduced the images of 2D crystals from the 24 OM images of MoS2 and found that both the cross-validation score and the accuracy rate of generated data through U-Net is over 70 percent, which is comparable with non-expert human level. This means that our neural network can distinguish monolayer, bilayer and other thicker flakes of MoS2 on Si/SiO2 substrates with the practical accuracy in the first screening process for searching desirable before further transport/optical experiments. The present study highlights that deep neural networks have great potential to become a new tool for quickly and autonomously segmenting and identifying atomic layer thickness of various 2D crystals and opening a new way for AI-based quick exploration for manufacturing 2D materials and van der Waals heterostructures in large scale. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. 10 References 1. Novoselov, K. S. et al. Two-dimensional atomic crystals. Proc. Natl. Acad. Sci. U. S. A. 102, 10451 -- 3 (2005). 2. Novoselov, K. S., Mishchenko, A., Carvalho, A. & Castro Neto, A. H. 2D materials and van der Waals heterostructures. Science 353, aac9439 (2016). 3. 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). 4. Xia, F., Wang, H., Xiao, D., Dubey, M. & Ramasubramaniam, A. Two-dimensional material nanophotonics. Nat. Photonics 8, 899 -- 907 (2014). 5. Jariwala, D., Sangwan, V. & Lauhon, L. Emerging device applications for semiconducting two-dimensional transition metal dichalcogenides. ACS Nano 8, 1102 -- 1120 (2014). 6. Fiori, G. et al. Electronics based on two-dimensional materials. Nat. Nanotechnol. 9, 768 -- 779 (2014). 7. Mak, K. F. & Shan, J. Photonics and optoelectronics of 2D semiconductor transition metal dichalcogenides. Nat. Photonics 10, 216 -- 226 (2016). 8. Saito, Y., Nojima, T. & Iwasa, Y. Highly crystalline 2D superconductors. Nat. Rev. Mater. 2, 16094 (2016). 9. Huang, B. et al. Layer-dependent ferromagnetism in a van der Waals crystal down to the monolayer limit. Nature 546, 270 -- 273 (2017). 10. Gong, C. et al. Discovery of intrinsic ferromagnetism in two-dimensional van der Waals crystals. Nature 546, 265 -- 269 (2017). 11. Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal dichalcogenide nanosheets. Nat. Chem. 5, 263 -- 75 (2013). 12. Wang, H., Yuan, H., Sae Hong, S., Li, Y. & Cui, Y. Physical and chemical tuning of two-dimensional transition metal dichalcogenides. Chem. Soc. Rev. 44, 2664 -- 2680 (2015). 13. Chimene, D., Alge, D. L. & Gaharwar, A. K. Two-Dimensional Nanomaterials for Biomedical Applications: Emerging Trends and Future Prospects. Adv. Mater. 27, 7261 -- 7284 (2015). 14. Masubuchi, S. et al. Autonomous robotic searching and assembly of two-dimensional crystals to build van der Waals superlattices. Nat. Commun. 9, 1413 (2018). 11 15. Ni, Z. H. et al. Graphene thickness determination using reflection and contrast spectroscopy. Nano Lett. 7, 2758 -- 2763 (2007). 16. Lee, C. et al. Anomalous lattice vibrations of single-and few-layer MoS2. ACS Nano 4, 2695 -- 700 (2010). 17. Li, H. et al. Optical identification of single- and few-layer MoS2 sheets. Small 8, 682 -- 686 (2012). 18. Li, H. et al. Rapid and reliable thickness identifi cation of two-dimensional nanosheets using optical microscopy. ACS Nano 7, 10344 -- 10353 (2013). 19. Li, Y. et al. Optical identification of layered MoS2via the characteristic matrix method. Nanoscale 8, 1210 -- 1215 (2016). 20. Krizhevsky, A. & Hinton, G. E. ImageNet Classification with Deep Convolutional Neural Networks. Adv. Neural Inf. Process. Syst. 25, (2012). 21. LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436 -- 444 (2015). 22. Silver, D. et al. Mastering the game of Go with deep neural networks and tree search. Nature 529, 484 -- 489 (2016). 23. Fauw, J. De et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat. Med. 24, 1342 -- 1350 (2018). 24. Poplin, R. et al. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nat. Biomed. Eng. 2, 158 -- 164 (2018). 25. Ryan, K., Lengyel, J. & Shatruk, M. Crystal Structure Prediction via Deep Learning. J. Am. Chem. Soc. 140, 10158 -- 10168 (2018). 26. Van Nieuwenburg, E. P. L., Liu, Y. H. & Huber, S. D. Learning phase transitions by confusion. Nat. Phys. 13, 435 -- 439 (2017). 27. Carrasquilla, J. & Melko, R. G. Machine learning phases of matter. Nat. Phys. 13, 431 -- 434 (2017). 28. Baldi, P., Sadowski, P. & Whiteson, D. Searching for exotic particles in high-energy physics with deep learning. Nat. Commun. 5, 4308 (2014). 29. Ronneberger, O., Fischer, P. & Brox, T. U-Net: Convolutional Networks for Biomedical Image Segmentation. MICCAI 2015 (2015). 30. Szegedy, C. et al. Going Deeper with Convolutions. Conf. Proc. CVPR2015 (2015). 31. He, K., Zhang, X., Ren, S. & Sun, J. Deep Residual Learning for Image Recognition. Conf. Proc. CVPR2016 (2016). 32. Kim, E., Huang, K., Jegelka, S. & Olivetti, E. Virtual screening of inorganic materials synthesis parameters with deep learning. npj Comput. Mater. 3, (2017). 12 33. Perol, T., Gharbi, M. & Denolle, M. Convolutional neural network for earthquake detection and location. Sci. Adv. 2016, 2 -- 10 (2018). 34. Dosovitskiy, A., Fischer, P., Springenberg, J. T., Riedmiller, M. & Brox, T. Discriminative unsupervised feature learning with exemplar convolutional neural networks. Adv. Neural Inf. Process. Syst. 27 (NIPS 2014) (2014). doi:10.1109/TPAMI.2015.2496141 35. Eigen, D. & Fergus, R. Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture. Proc. 2015 IEEE Int. Conf. Comput. Vis. 2650 -- 2658 (2015). doi:10.1109/ICCV.2015.304 36. Dekel, O. & Xiao, L. Optimal Distributed Online Prediction Using Mini-Batches. J. ofMachine Learn. Res. 13, 165 -- 202 (2012). 37. Li, M., Zhang, T., Chen, Y. & Smola, A. J. Efficient Mini-batch Training for Stochastic Optimization. KDD '14 Proc. 20th ACM SIGKDD Int. Conf. Knowl. Discov. data Min. 661 -- 670 (2014). 38. Kingma, D. P. & Ba, J. L. ADAM: A Method for stochastic optimization. Conf. Proc. ICLR 2015 (2015). 13 Acknowledgements This work was supported by the 'Materials research by Information Integration' Initiative (MI2I) project and Core Research for Evolutional Science and Technology (CREST) (JSPS KAKENHI Grant Numbers JPMJCR1502 and JPMJCR17J2) from Japan Science and Technology Agency (JST). It was also supported by Grant-in-Aid for Scientific Research on Innovative Areas 'Nano Informatics' (JSPS KAKENHI Grant Number JP25106005) and Grant-in-Aid for Specially Promoted Research (JSPS KAKENHI Grant Number JP25000003) from JSPS. M.O. and Y.M.I. were supported by Advanced Leading Graduate Course for Photon Science (ALPS). Y.N. was supported by Materials Education program for the future leaders in Research, Industry, and Technology (MERIT). M.O. and Y.N. were supported by the Japan Society for the Promotion of Science (JSPS) through a research fellowship for young scientists (Grant-in-Aid for JSPS Research Fellow, JSPS KAKENHI Grant Numbers JP17J09152 and JP17J08941, respectively). M.Y. was supported by JST PRESTO (Precursory Research for Embryonic Science and Technology) program JPMJPR165A Author contributions Y.S., K. Terayama, M.Y. and K. Tsuda conceived the idea, designed and supervised the Y.S., K.S., K. Terayama implemented the proposed method and analyzed the experimental results. Y.S., M.O., Y.N and Y.M.I. exfoliated MoS2 and collected images of atomic layers. All authors discussed the results. Y.S., K.S., K. Terayama and K. Tsuda wrote the manuscript with contributions from all other co-authors. Notes The authors declare that they have no competing interests. 14 Figure Captions Fig. 1. Process of quality-filtering of 2D crystals via deep neural network. First, we mechanically exfoliate MoS2 crystals on a SiO2/Si substrate and take images using optical microscope. After recording data, we train deep neural networks to generate images from optical microscope images using segmented images. Here, our targets are monolayer and bilayer crystals. For training, we prepared segmented data, which is divided into monolayer (blue), bilayer (green) and other parts (black). 15 Fig. 2. Network architectures based on U-Net. The encoder of U-Net extracts the small feature map from the input image by convolution and pooling layers, and decoder expands it to the original image size by convolution and up-sampling layers. Skip connections are added between each layer of the encoder and the corresponding layer of the decoder in order to transmit a high-resolution information to the decoder. In the architecture, the encoder and decoder are composed of 14 and 18 layers respectively. The dimension, width × height × channels, of the input image is 512×512×3 and it changes to 256×256×64, 128×128×128, 64×64×256, 32×32×512, and 16×16×1024 at each downsampling step in the encoder respectively, and changes in reverse order through the upsampling steps in the decoder. 16 Fig. 3 Examples of original optical microscope images (a-d), segmented images (e-h) and generated images based on U-Net with weighted loss (i-j). In the segmented and generated images, blue and green region show monolayer and bilayer region, respectively. 17 Fig.4. Receiver operating characteristic (ROC) curves and precision-recall curves for the monolayer and bilayer identification. a, ROC curve for the monolayer and bilayer identification. Blue and red curve show the ROC for two-class classifications of monolayer/others and bilayer/others, respectively. This ROC curve shows the high performance of the U-Net with weighted loss because the ROC curve rises sharply below the false positive rate of 0.1 and then saturates to 1.0. b, Precision-recall (true positive rate) curves for the monolayer and bilayer identification. Table 1. Cross-validation scores for U-Net and U-Net with weighted loss. U-Net U-Net with weighted loss w/o grayscale & w/o contrast adjusted images w/ grayscale w/ contrast adjusted images w/ grayscale & contrast adjusted images 0.584 0.46 0.678 0.633 0.643 0.56 0.789 0.723 18
1912.04122	1	1912	2019-12-06T03:41:48	Phase-encoded RF signal generation based on an integrated 49GHz micro-comb optical source	[ "physics.app-ph", "physics.optics" ]	We demonstrate photonic RF phase encoding based on an integrated micro-comb source. By assembling single-cycle Gaussian pulse replicas using a transversal filtering structure, phase encoded waveforms can be generated by programming the weights of the wavelength channels. This approach eliminates the need for RF signal generators for RF carrier generation or arbitrary waveform generators for phase encoded signal generation. A large number of wavelengths of up to 60 were provided by the microcomb source, yielding a high pulse compression ratio of 30. Reconfigurable phase encoding rates ranging from 2 to 6 Gb/s were achieved by adjusting the length of each phase code. This work demonstrates the significant potentials of this microcomb-based approach to achieve high-speed RF photonic phase encoding with low cost and footprint.	physics.app-ph	physics	Phase-encoded RF signal generation based on an integrated 49GHz micro-comb optical source Xingyuan Xu, Mengxi Tan, Jiayang Wu, Andreas Boes, Bill Corcoran, Thach G. Nguyen, Sai T. Chu, Brent E. Little, Roberto Morandotti, Arnan Mitchell, and David J. Moss 8 d Abstract -- We demonstrate photonic RF phase encoding based on an integrated micro-comb source. By assembling single-cycle Gaussian pulse replicas using a transversal filtering structure, phase encoded waveforms can be generated by programming the weights of the wavelength channels. This approach eliminates the need for RF signal generators for RF carrier generation or arbitrary waveform generators for phase encoded signal generation. A large number of wavelengths -- up to 60 -- were provided by the microcomb source, yielding a high pulse compression ratio of 30. Reconfigurable phase encoding rates ranging from 2 to 6 Gb/s were achieved by adjusting the length of each phase code. This work demonstrates the significant potentials of this microcomb-based approach to achieve high-speed RF photonic phase encoding with low cost and footprint. Index Terms -- Microwave photonics, micro-ring resonators. I. INTRODUCTION Phase encoded continuous wave (CW) radio frequency (RF) signals are widely employed in low probability of intercept (LPI) radar systems, since they feature low power density and employ random codes to avoid being cracked [1-2]. In such systems, targets are resolved within range when the reflected phase encoded signals are compressed at the receiver, and a large signal bandwidth is required to achieve a high resolution [3]. Photonic techniques have significant potential to overcome the limitations of their electronic counterparts [4] due to their intrinsically broad bandwidth, immunity from electromagnetic interference, and low propagation loss [5-11]. Significant effort has been devoted to realizing photonic-assisted RF phase encoding, including approaches based on polarization modulators [12, 13], Sagnac loops [14, 15], and dual parallel modulators [16, 17]. However, these approaches all face limitations of one form or another, such as the need for complicated modulation schemes to achieve the required π-phase shifts, and the need for both high-frequency RF signal generators and arbitrary waveform generators. The former increases the complexity and instability (such as the bias drift of the modulators) of the system, while the latter greatly increase the cost, size and power consumption. An alternative method of generating RF phase encoded signals based on transversal filtering structures has been proposed [18]. In this method, RF high-order Gaussian pulses are broadcast onto 4 wavelength channels via an intensity modulator, and then progressively delayed through a dispersive medium such that the pulse segments are assembled (concatenated) in the time domain. By controlling the phase of each pulse segment, phase-encoded waveforms can be generated. This approach reduces the need for RF sources and complicated modulation schemes, and offers a high potential phase encoding speed by tailoring the delay step and pulse width. However, it is not without challenges. The use of discrete laser arrays significantly increases the complexity and cost for further integration and limits the number of wavelengths. This in turn leads to a limited sequence length and thus a small pulse compression ratio that would limit the performance of radar systems. Secondly, the employed high-order Gaussian pulses require complicated This work was supported by the Australian Research Council Discovery Projects Program (No. DP150104327). RM acknowledges support by the Natural Sciences and Engineering Research Council of Canada (NSERC) through the Strategic, Discovery and Acceleration Grants Schemes, by the MESI PSR-SIIRI Initiative in Quebec, and by the Canada Research Chair Program. He also acknowledges additional support by the 1000 Talents Sichuan Program in China. Brent E. Little was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences, Grant No. XDB24030000. X. Xu, M. Tan, J. Wu, and D. J. Moss are with the Centre for Micro-Photonics, Swinburne University of Technology, Hawthorn, VIC 3122, Australia. (Corresponding e-mail: [email protected]). A. Boes, T. G. Thach and A. Mitchell are with the School of Engineering, RMIT University, Melbourne, VIC 3001, Australia. B. Corcoran is with the Department of Electrical and Computer System Engineering, Monash University, Clayton, VIC 3168 Australia. S. T. Chu is with the Department of Physics, City University of Hong Kong, Tat Chee Avenue, Hong Kong, China. B. E. Little is with State Key Laboratory of Transient Optics and Photonics, Xi'an Institute of Optics and Precision Mechanics, Chinese Academy of Science, Xi'an, China. R. Morandotti is with INSR-Énergie, Matériaux et Télécommunications, 1650 Boulevard Lionel-Boulet, Varennes, Québec, J3X 1S2, Canada, and adjunct with the Institute of Fundamental and Frontier Sciences, University of Electronic Science and Technology of China, Chengdu 610054, China. 8 generation schemes such as high-order differentiation, which will increase the cost and complexity of the system. Integrated microcombs [19-24], generated through optical parametric oscillation in monolithic microring resonators (MRRs), are promising multi-wavelength sources that offer many advantages including a much higher number of wavelengths and greatly reduced footprint and complexity. Thus, a wide range of RF applications have been demonstrated based on micro-combs, such as transversal filters [25-27], differentiators [28], optical true time delays [29, 30] and channelizers [31, 32], Hilbert transformers [33-35], microwave mixers [36], microwave generation [37], and more. In this paper, we demonstrate phase encoded CW RF signal generation using an integrated soliton crystal micro- comb source. The soliton crystal is a recently discovered coherent and stable solution of the Lugiato-Lefever equation [38, 39], which governs the nonlinear dynamics of the optical parametric oscillator. The soliton crystal stably forms through the background wave generated by a mode crossing, and features over thirty times higher intra-cavity power than traditional dissipative Kerr solitons. Thus, they can remove thermal effects and enable adiabatic pump sweeping for coherent comb generation [39, 40]. These properties, together with the broadband comb spectrum, make soliton crystals promising for demanding integrated RF photonic systems, such as the photonic RF phase-encoder introduced in this work. Here, by multicasting an RF single-cycle pulse (instead of the high-order pulse used in [18]) onto a spectrally flattened micro-comb and progressively delaying the replicas with a dispersive medium, the RF pulse replicas are assembled arbitrarily in time according to the designed binary phase codes. The large number of wavelengths offered by the microcomb enabled 30 bits, representing an enhancement of over 4 times compared to previous work [18], for the phased encoded RF sequence, leading to a pulse compression ratio as high as 29.6. A reconfigurable phase encoding rate ranging from 1.98 to 5.95 Gb/s was also achieved by varying the length of each phase code. These results verify the strong potential of our approach for the realization of photonic RF phase encoding schemes and represent a solid step towards the miniaturization of fully integrated radar transmitters with greatly reduced cost and footprint. Figure 1 illustrates the operation principle of a photonic phase encoder. First, an RF Gaussian pulse f (t) with a duration of Δt is generated. Then a discrete convolution operation between the RF pulse and flattened microcomb spectrum (denoted by a discrete signal g[n] with length N and binary values of 1 or -1) is performed with a delay step of Δt, and can be described as: II. THEORY (2.1) The discrete convolution operation between the RF pulse f (t) and the flattened microcomb spectrum is achieved using the experimental setup in Figure 2. The RF pulse is first multi-cast onto the wavelength channels to yield replicas, which are then delayed with a progressive step that matches the pulse duration Δt. By flipping the sign of the delayed replicas according to designed phase codes g[n], a phase-encoded sequence can be assembled in the time domain. The sign reversal is accomplished by spatially separating the intended positive pulses from the negative ones using the programmable wavelength dependent spatial routing capability of a waveshaper, and then directing the two separate outputs to the positive and negative inputs of a differential photodetector, respectively. The total number of RF pulse replicas N is equal to the number of wavelength channels, which is 60 in our case -- 15 times that of the previous work [18]. Thus, the total time length of the phase-encoded sequence is T=N·Δt. Here, the basic temporal element in the phase-encoded sequence is termed an "RF segment", which is a single-cycle or multi-cycle sine wave assembled from the input RF pulse (i.e., two anti-phase RF pulses assemble a single-cycle sine wave). The centre frequency of the phase-encoded sequence is determined by the frequency of the assembled sine waves, thus is equivalently given by 1/2/Δt. Assuming each RF segment constitutes m pulses, then the length of the phase-encoded sequence would be N/m, together with an equivalent phase encoding speed of 1/Δt/m. We note that the generated phase-encoded sequence has a bandwidth similar to the input RF pulse, which is subject to the Nyquist bandwidth, equivalent to half of the comb spacing (48.9 GHz/2=24.45 GHz). In a radar system, the phase encoded RF waveform is first amplified and then emitted by antennas, followed by reflection from the target, and finally collected and processed at the receiver. The range information of the target is carried on the delay of the received waveforms, which can be readily determined by calculating the autocorrelation function of the reflected phase-encoded RF signal given by 1()[][][]Nifgnfnitgi8 (2.2) In practical systems this is calculated by performing a hardware autocorrelation extracted by an array of range gates. The range gates are composed of matched filters with their tap coefficients determined by the employed phase codes (Fig. 3). The function of the matched filters is to perform single-output autocorrelations for the received signal, and the range gate that has the highest output power denotes the resolved range of the target. III. EXPERIMENTAL RESULTS The experimental setup is shown in Figure 2. A CW laser is amplified and its polarization state adjusted to pump a nonlinear high Q factor (> 1.5 million) MRR, which featured a free spectral range of ~0.4nm, or 48.9 GHz. As the detuning between the pump laser and the MRR was changed, dynamic parametric oscillation states corresponding to distinctive solutions of the Lugiato-Lefever equation were initiated [38]. We thus generated soliton crystal microcombs [39, 40], which were tightly packaged solitons circulating in the MRR as a result of a mode crossing (at ~1552 nm in our case), and which resulted in the distinctive palm-like comb spectrum (Fig. 4). Next, 60 lines of the microcomb were flattened (N=60 in our case), using two stages of WaveShapers (Finisar 4000S) to acquire a high link gain and signal-to-noise ratio. This was achieved by pre-flattening the microcomb lines with the first WaveShaper such that the optical power distribution of the wavelength channels roughly matched with the desired channel weights. The second WaveShaper was employed for accurate comb shaping assisted by a feedback loop as well as to separate the wavelength channels into two parts (port 1 and port 2 of the Waveshaper) according to the polarity of the designed binary phase codes. The feedback loop was constructed by reading the optical spectrum with an optical spectrum analyser and comparing the power of the wavelengths with the designed weights to generate an error signal, which was then fed back into the second WaveShaper (WS2) to calibrate its loss until the error was below 0.2 dB. Here, we used a Gaussian pulse with a duration of Δt = 84 ps, as the RF fragment f [t]. Although we used an arbitrary waveform generator (Keysight, 65 GSa/s) to generate the Gaussian pulse for the sake of simplicity, the arbitrary waveform generator is actually not a necessity and can be replaced with many other readily available approaches that are easier and cheaper [41, 42]. The input RF pulse was imprinted onto the comb lines, generating replicas across all the wavelength channels. The replicas then went through a ~13 km long spool of standard single mode fibre (with a dispersion of 17ps/nm/km) to progressively delay them, leading to a delay step of ~84 ps between the adjacent wavelength channels that matched with the duration of the RF pulse Δt. Finally, the wavelength channels were separated into two parts according to the designed phase codes and sent to a balanced photodetector (Finisar, 40 GHz) to achieve negative and positive replicas for the phase encoding. Figure 5 shows the input Gaussian pulse and the flattened optical comb spectra measured at the output ports of the second Waveshaper (WS2). Each pair of wavelength channels from different ports can assemble a monocycle sine wave, thus with the 60 comb lines, 30 sine cycles can be achieved, with a total time length of T = N·Δt = 60×84ps = 5.04 ns. By applying designed phase codes during the separation of the wavelength channels, the sine cycles could be π- phase shifted at desired times. The phase-encoded results are shown in Fig. 6. The number of Gaussian pulses for each RF segment (denoted by m) was reconfigured from 6 to 2, corresponding to reconfigurable sequence lengths (N/m) ranging from 10 to 30 and phase coding speeds (1/Δt/m) ranging from 1.98 to 5.95 Gb/s. The employed phase codes were denoted both by the shaded areas and the stair waveforms (black solid line). This result shows that our photonic phase coder can offer a reconfigurable sequence length to address the performance tradeoffs between range resolution and system complexity. To acquire a large pulse compression ratio for a high resolution, the sequence length should be maximized, where the number of Gaussian pulses for each RF segment (m) should be set as 2. While to reduce the complexity and cost of the RF system (such as the number of range gates at the receiver), the sequence length could be reduced by either employing fewer wavelength channels or by increasing m. The corresponding optical spectra (Fig. 6 (a, c, e)) were measured at the output of Waveshaper to show the positive and negative phase codes realized by changing the wavelength channels' output ports at the Waveshaper. The encoded RF waveforms (Fig. 6 (b, d, f)) clearly show the flipped phase of the RF segments at the time of negative phase codes, where the number of sine cycles was reconfigured as well, depending on the value of m. This result also shows that our approach is fully reconfigurable for different phase codes and encoding speeds. We note that higher encoding ()()ACFsstd8 speeds can be achieved by reducing the duration of the RF fragment and the delay step Δt. In this work we calculated the autocorrelation (Fig. 7) of the phase-encoded RF waveforms s(t). As the sequence length varied from 10 to 30, the full width at half-maximum (FWHM) of the compressed pulses varied from 0.52 to 0.17 ns, which corresponds to a pulse suppression ratio ranging from 9.7 to 29.6. Meanwhile the peak-to-sidelobe ratio (PSR) also increased with the sequence length from 4.17 to 6.59 dB. These results confirm that the pulse compression ratio of an RF phase-encoded signal is linearly related to its sequence length [43], and that this can be significantly enhanced with our approach by employing a larger number of wavelength channels of the microcomb. Figure 8 shows calculated examples of the estimated outputs of the range gates [43] with different distances. Considering an example with a sequence length N/m = 30, the delay of the matched filters would be 2Δt=168 ps. The tap coefficients for the lth matched filter are c[k-l], where c[k], k=1, 2, …30, is the employed phase codes. The range resolution of the radar, which is the minimum distance between two resolvable targets, is determined by the delay step (2Δt) of the matched filters, which is given by 2Δt ⸱c=5cm, where c=3×108m/s is the speed of RF signals in air. If the distance between the target and radar is R, then the delay would be 2R/c. The range gates would have a maximum output at the lth range gate, l=2R/(c⸱2Δt). We note here that this calculation only shows the basic connections between our phase encoder and radar systems' performances - practical radar systems are subject to more complicated trade- offs involving capability versus performance. We note that although random phase codes were used here as a proof of concept, the design of phase codes can be optimized to achieve a better detection performance of the radar system, which is an NP-hard problem and can be approximated and addressed with the methods in [44]. Further, the routes to further scale up the performance of our system are clear. The sequence length can be further increased by either employing microcombs that have a smaller FSR and thus more wavelength channels in the available optical band, or by introducing spatial-multiplexing techniques to boost the length of parallel channels. Moreover, the increasingly mature nonlinear platforms and recent advances in coherent soliton generation techniques (such as deterministic soliton generation [45] and battery-driven solitons [46]) enable microcombs to have an increasingly wide access and enhanced performance for engineering applications, including the phase-encoded signal generator reported here. IV. CONCLUSION We demonstrate photonic RF phase encoding using an integrated micro-comb source. A single-cycle Gaussian pulse was multicast onto the comb wavelengths to assemble the desired phase-encoded RF waveform. A high pulse compression ratio of 29.6 and phase encoding speed of 5.95 Gb/s were achieved, enabled by the use of 60 wavelengths generated by the microcomb. The sequence length was reconfigured by adjusting the length of each phase code, which led to a reconfigurable encoding speed. These results verify that our approach to high-speed RF phase encoding is competitive in terms of performance, with potentially lower cost and footprint. REFERENCES [1] P. Ghelfi, F. Laghezza, F. Scotti, G. Serafino, A. Capria, S. Pinna, D. Onori, C. Porzi, M. Scaffardi, A. Malacarne, V. Vercesi, E. Lazzeri, F. Berizzi, and A. Bogoni, "A fully photonics-based coherent radar system," Nature, vol. 507, pp. 341, Mar. 2014. [2] M. Skolnik, "Role of radar in microwaves," IEEE Transactions on Microwave Theory and Techniques, vol. 50, no. 3, pp. 625- 632. 2002. [3] P. Ghelfi, F. Scotti, F. Laghezza, and A. Bogoni, "Photonic Generation of Phase-Modulated RF Signals for Pulse Compression Techniques in Coherent Radars," Journal of Lightwave Technology, vol. 30, no. 11, pp. 1638-1644, Jun. 2012. [4] J. Capmany, and D. Novak, "Microwave photonics combines two worlds," Nat. Photonics, vol. 1, no. 6, pp. 319-330, Jun. 2007. [5] J. P. Yao, "Microwave Photonics," J. Lightwave Technol., vol. 27, no. 1-4, pp. 314-335, Jan-Feb. 2009. [6] D. Marpaung, J. Yao, and J. Capmany, "Integrated microwave photonics," Nature Photonics, vol. 13, no. 2, pp. 80-90, Feb. 2019. [7] V. R. Supradeepa, C. M. Long, R. Wu, F. Ferdous, E. Hamidi, D. E. Leaird, and A. M. Weiner, "Comb-based radiofrequency photonic filters with rapid tunability and high selectivity," Nat. Photonics, vol. 6, no. 3, pp. 186-194, Mar. 2012. [8] J. Wu, X. Xu, T. G. Nguyen, S. T. Chu, B. E. Little, R. Morandotti, A. Mitchell, and D. J. Moss, "RF Photonics: An Optical Microcombs' Perspective," IEEE Journal of Selected Topics in Quantum Electronics, vol. 24, no. 4, pp. 1-20. 2018. [9] J. Azaña, "Ultrafast Analog All-Optical Signal Processors Based on Fiber-Grating Devices," IEEE Photonics Journal, vol. 2, no. 3, pp. 359-386. 2010. [10] Z. Jiang, C.-B. Huang, D. E. Leaird, and A. M. Weiner, "Optical arbitrary waveform processing of more than 100 spectral comb lines," Nature Photonics, vol. 1, pp. 463, Aug. 2007. 8 [11] S. T. Cundiff, and A. M. Weiner, "Optical arbitrary waveform generation," Nature Photonics, vol. 4, pp. 760, Oct. 2010. [12] H. Chi, and J. Yao, "Photonic Generation of Phase-Coded Millimeter-Wave Signal Using a Polarization Modulator," IEEE Microwave and Wireless Components Letters, vol. 18, no. 5, pp. 371-373. 2008. [13] Y. Zhang, and S. Pan, "Generation of phase-coded microwave signals using a polarization-modulator-based photonic microwave phase shifter," Optics Letters, vol. 38, no. 5, pp. 766-768, Mar. 2013. [14] Z. Li, W. Li, H. Chi, X. Zhang, and J. Yao, "Photonic Generation of Phase-Coded Microwave Signal With Large Frequency Tunability," IEEE Photonics Technology Letters, vol. 23, no. 11, pp. 712-714. 2011. [15] W. Liu, and J. Yao, "Photonic Generation of Microwave Waveforms Based on a Polarization Modulator in a Sagnac Loop," Journal of Lightwave Technology, vol. 32, no. 20, pp. 3637-3644, Oct. 2014. [16] S. Zhu, Z. Shi, M. Li, N. H. Zhu, and W. Li, "Simultaneous frequency upconversion and phase coding of a radio-frequency signal for photonic radars," Optics Letters, vol. 43, no. 3, pp. 583-586, Feb. 2018. [17] S. Zhu, M. Li, X. Wang, N. H. Zhu, Z. Z. Cao, and W. Li, "Photonic generation of background-free binary phase-coded microwave pulses," Optics Letters, vol. 44, no. 1, pp. 94-97, Jan. 2019. [18] Y. Dai, and J. Yao, "Microwave pulse phase encoding using a photonic microwave delay-line filter," Optics Letters, vol. 32, no. 24, pp. 3486-3488, Dec. 2007. [19] T. J. Kippenberg, A. L. Gaeta, M. Lipson, and M. L. Gorodetsky, "Dissipative Kerr solitons in optical microresonators," Science, vol. 361, no. 6402, pp. 8083. 2018. [20] D. J. Moss, R. Morandotti, A. L. Gaeta, and M. Lipson, "New CMOS-compatible platforms based on silicon nitride and Hydex for nonlinear optics," Nat. Photonics, vol. 7, no. 8, pp. 597-607, Aug. 2013. [21] A. L. Gaeta, M. Lipson, and T. J. Kippenberg, "Photonic-chip-based frequency combs," Nature Photonics, vol. 13, no. 3, pp. 158-169, Mar. 2019. [22] M.-G. Suh, and K. J. Vahala, "Soliton microcomb range measurement," Science, vol. 359, no. 6378, pp. 884-887. 2018. [23] W. Liang, D. Eliyahu, V. S. Ilchenko, A. A. Savchenkov, A. B. Matsko, D. Seidel, and L. Maleki, "High spectral purity Kerr frequency comb radio frequency photonic oscillator," Nature communications, vol. 6, pp. 7957. 2015. [24] D. T. Spencer, T. Drake, T. C. Briles, J. Stone, L. C. Sinclair, C. Fredrick, Q. Li, D. Westly, B. R. Ilic, and A. Bluestone, "An optical-frequency synthesizer using integrated photonics," Nature, vol. 557, no. 7703, pp. 81-85. 2018. [25] X. Xu, J. Wu, T. G. Nguyen, M. Shoeiby, S. T. Chu, B. E. Little, R. Morandotti, A. Mitchell, and D. J. Moss, "Advanced RF and microwave functions based on an integrated optical frequency comb source," Opt. Express, vol. 26, no. 3, pp. 2569-2583, Feb 5. 2018. [26] X. Xu, M. Tan, J. Wu, T. G. Nguyen, S. T. Chu, B. E. Little, R. Morandotti, A. Mitchell, and D. J. Moss, "Advanced Adaptive Photonic RF Filters with 80 Taps Based on an Integrated Optical Micro-Comb Source," Journal of Lightwave Technology, vol. 37, no. 4, pp. 1288-1295, Feb. 2019. [27] X. X. Xue, Y. Xuan, H. J. Kim, J. Wang, D. E. Leaird, M. H. Qi, and A. M. Weiner, "Programmable Single-Bandpass Photonic RF Filter Based on Kerr Comb from a Microring," J. Lightwave Technol., vol. 32, no. 20, pp. 3557-3565, Oct 15. 2014. [28] X. Xu, J. Wu, M. Shoeiby, T. G. Nguyen, S. T. Chu, B. E. Little, R. Morandotti, A. Mitchell, and D. J. Moss, "Reconfigurable broadband microwave photonic intensity differentiator based on an integrated optical frequency comb source," APL Photonics, vol. 2, no. 9, 096104 Sep. 2017. [29] X. Xu, J. Wu, T. G. Nguyen, T. Moein, S. T. Chu, B. E. Little, R. Morandotti, A. Mitchell, and D. J. Moss, "Photonic microwave true time delays for phased array antennas using a 49 GHz FSR integrated optical micro-comb source," Photonics Res, vol. 6, no. 5, pp. B30-B36, May 1. 2018. [30] X. Xue, Y. Xuan, C. Bao, S. Li, X. Zheng, B. Zhou, M. Qi, and A. M. Weiner, "Microcomb-Based True-Time-Delay Network for Microwave Beamforming with Arbitrary Beam Pattern Control," Journal of Lightwave Technology, vol. 36, no. 12, pp. 2312-2321, Jun. 2018. [31] X. Xu, M. Tan, J. Wu, T. G. Nguyen, S. T. Chu, B. E. Little, R. Morandotti, A. Mitchell, and D. J. Moss, "High performance RF filters via bandwidtdh scaling with Kerr micro-combs," APL Photonics, vol. 4, no. 2, pp. 026102. 2019. [32] X. Xu, J. Wu, T. G. Nguyen, S. Chu, B. Little, A. Mitchell, R. Morandotti, and D. J. Moss, "Broadband RF Channelizer based on an Integrated Optical Frequency Kerr Comb Source," Journal of Lightwave Technology, vol. 36, no. 19, pp. 7, 2018. [33] M.Tan, X.Xu, B.Corcoran, J.Wu, A. Boes, T. G. Nguyen, S.T. Chu, B.E. Little, R. Morandotti, A.Mitchell, and D.J. Moss, "Broadband microwave and RF photonic fractional Hilbert transformer based on a 50GHz integrated Kerr micro-comb", Vol. 37 2019. IEEE Journal of Lightwave Technology. DOI:10.1109/JLT.2019.2946606 [34] T.G. Nguyen, M.Shoeiby, S.T. Chu, B.E. Little, R.Morandotti, A.Mitchell, and D. J. Moss, "Integrated frequency comb source based Hilbert transformer for wideband microwave photonic phase analysis", Optics Express Vol. 23, no. 17, 22087 (2015). DOI:10.1364/OE.23.022087 [35] X. Xu, M. Tan, J.Wu, R.Morandotti, A.Mitchell, and D. J. Moss, "Microcomb-based photonic RF signal processing", IEEE Photonics Technology Letters Vol. 31 (2019). DOI: 10.1109/LPT.2019.2940497 [36] X.Xu, J.Wu, Mengxi Tan, T.G. Nguyen, S.T. Chu, B. E. Little, R.Morandotti, A.Mitchell, and David J. Moss, "Broadband microwave frequency conversion based on an integrated optical micro-comb source", IEEE Journal of Lightwave Technology Vol. 37 (2019). DOI: 10.1109/JLT.2019.2930466 [37] A. Pasquazi; M.Peccianti; B. Little; S.T. Chu; D. J. Moss; R. Morandotti, "Stable, dual mode, high repetition rate mode-locked laser based on a microring resonator", Optics Express vol. 20, no. 24, 27355 -- 27362, 2012. DOI: 10.1364/OE.20.027355 8 [38] A. Pasquazi, M. Peccianti, L. Razzari, D. J. Moss, S. Coen, M. Erkintalo, Y. K. Chembo, T. Hansson, S. Wabnitz, P. Del'Haye, X. X. Xue, A. M. Weiner, and R. Morandotti, "Micro-combs: A novel generation of optical sources," Physics Reports, Review Section of Physics Letters, vol. 729, pp. 1-81, Jan 27. 2018. [39] D. C. Cole, E. S. Lamb, P. Del'Haye, S. A. Diddams, and S. B. Papp, "Soliton crystals in Kerr resonators," Nat. Photonics, vol. 11, no. 10, pp. 671-676, Oct. 2017. [40] W. Q. Wang, Z. Z. Lu, W. F. Zhang, S. T. Chu, B. E. Little, L. R. Wang, X. P. Xie, M. L. Liu, Q. H. Yang, L. Wang, J. G. Zhao, G. X. Wang, Q. B. Sun, Y. S. Liu, Y. S. Wang, and W. Zhao, "Robust soliton crystals in a thermally controlled microresonator," Opt. Lett., vol. 43, no. 9, pp. 2002-2005, May 1. 2018. [41] W. Chiming, and N. K. Dutta, "High-repetition-rate optical pulse generation using a rational harmonic mode-locked fiber laser," IEEE Journal of Quantum Electronics, vol. 36, no. 2, pp. 145-150. 2000. [42] H. Jeongwoo, and N. Cam, "A new ultra-wideband, ultra-short monocycle pulse generator with reduced ringing," IEEE Microwave and Wireless Components Letters, vol. 12, no. 6, pp. 206-208. 2002. [43] N. Levanon, and B. Getz, "Comparison between linear FM and phase-coded CW radars," IEE Proceedings - Radar, Sonar and Navigation, vol. 141, no. 4, pp. 230-240. 1994. [44] A. D. Maio, S. D. Nicola, Y. Huang, Z. Luo, and S. Zhang, "Design of Phase Codes for Radar Performance Optimization with a Similarity Constraint," IEEE Transactions on Signal Processing, vol. 57, no. 2, pp. 610-621. 2009. [45] H. Zhou, Y. Geng, W. Cui, S.-W. Huang, Q. Zhou, K. Qiu, and C. Wei Wong, "Soliton bursts and deterministic dissipative Kerr soliton generation in auxiliary-assisted microcavities," Light: Science & Applications, vol. 8, no. 1, pp. 50, 2019. [46] B. Stern, X. Ji, Y. Okawachi, A. L. Gaeta, and M. Lipson, "Battery-operated integrated frequency comb generator," Nature, vol. 562, no. 7727, pp. 401. 2018. 8 Fig. 1. Illustration of the operation of the photonic RF phase encoder. 8 Fig. 2. Schematic diagram of photonic RF phase encoding based on an integrated optical micro-comb source. EDFA: erbium-doped fibre amplifier. PC: polarization controller. MRR: micro-ring resonator. WS: WaveShaper. EOM: Mach-Zehnder modulator. SMF: Single mode fibre. BPD: balanced photodetector. Fig. 3. Structure of the range gates at the radar systems' receiver. 8 Fig. 4. Soliton crystal microcomb. Fig. 5. (a) The input Gaussian pulse. (b) The optical spectra of the flattened microcomb at different ports of the Waveshaper. (c) Assembled RF signal with 30 cosine cycles. 8 Fig. 6. (a, c, e) The optical spectra of the flattened microcomb at different ports of the Waveshaper with different phase codes and sequence length. (b, d, f) The assembled phase-encoded RF waveform. Fig. 7. The calculated autocorrelation of the phase-encoded RF waveforms for (a) m=6, (b) m=4, and (c) m=2. 8 Fig. 8. Calculated outputs of the range gates for (a) R=0.71 m and (b) R=0.76 m.
1807.03660	1	1807	2018-07-02T16:54:34	Open-source automated chemical vapor deposition system for the production of two - dimensional nanomaterials	[ "physics.app-ph", "cond-mat.mtrl-sci", "physics.ins-det" ]	The study of two- dimensional (2D) materials is a rapidly growing area within nanomaterial research. However, the high equipment costs, which include the processing systems necessary for creating these materials, can be a barrier to entry for some researchers interested in studying these novel materials. Such process systems include those used for chemical vapor deposition. This article presents the first open-source design for an automated chemical vapor deposition system that can be built for less than a third of the cost for a similar commercial system. Our design can be easily customized and expanded on, depending upon the needs of the user. With a process chamber built as described, we demonstrate that a variety of 2D nanomaterials and their heterostructures can be grown via chemical vapor deposition. Specifically, our experimental results demonstrate the capability of this open-source design in producing high quality, 2D nanomaterials such as graphene and tungsten disulfide, which are at the forefront of research in emerging semiconductor devices, sensors, and energy storage applications.	physics.app-ph	physics	Manuscript Click here to download Manuscript Open source automated CVD system manuscript.docx Open-source automated chemical vapor deposition system for the production of two- dimensional nanomaterials Lizandra Williams- Godwin1¶, Dale Brown1¶, Richard Livingston2, Tyler Webb1, Lynn Karriem1, Elton Graugnard1, David Estrada1,3* 1Micron School of Materials Science & Engineering, Boise State University, Boise, Idaho, United States of America 2Department of Mechanical and Biomedical Engineering, Boise State University, Boise, Idaho, United States of America 3Center for Advanced Energy Studies, Boise State University, Boise, Idaho, United States of America * Corresponding author Email: [email protected] (DE) ¶These authors contributed equally to this work. Abstract The study of two- dimensional (2D) materials is a rapidly growing area within nanomaterial research. However, the high equipment costs, which include the processing systems necessary for creating these materials, can be a barrier to entry for some researchers interested in studying these novel materials. Such process systems include those used for chemical vapor deposition. This article presents the first open-source design for an automated chemical vapor deposition system that can be built for less than a third of the cost for a similar commercial system. Our design can be easily customized and expanded on, depending upon the needs of the user. With a process chamber built as described, we demonstrate that a variety of 2D nanomaterials and their heterostructures can be grown via chemical vapor deposition. Specifically, our experimental results demonstrate the capability of this open-source design in producing high quality, 2D nanomaterials such as graphene and tungsten disulfide, which are at the forefront of research in emerging semiconductor devices, sensors, and energy storage applications. Introduction The modern research push into two- dimensional (2D) materials was sparked by Novoselov and Geim's isolation of few- and single-layer graphene via the mechanical exfoliation of graphite [1]. Novoselov et al. discovered that these single- and few-atom thick crystals of graphitic carbon, known as graphene, were stable in air, semi-metallic, and had extremely high carrier mobility (about 10,000 cm2 / V-s), though the on/off ratio of graphene transistors was found to be less than 30 at room temperature, limiting its usefulness as an active component in transistors for digital CMOS circuits. Since then, 2D materials such as hexagonal boron nitride (hBN) [2] and tungsten disulfide (WS2) [3] among a host of others [4-10] are garnering much research attention due to their potential use in emerging semiconductor device applications. Moreover, chemical vapor deposition (CVD) is one of the most common methods for the production of these novel 2D materials [3], [9], [11-14]. In the 2D materials and nanotechnology research community, there is a common barrier that potentially slows progress in the field. This barrier is due to the high cost of systems such as the CVD chamber that is needed in order to synthesize high quality films of these advanced materials. CVD is a synthesis process widely used in the production of commercial microelectronics, but the equipment for this process is very costly to purchase and often too expensive for the average researcher. A viable solution to this challenge is open-source technology. Open-source technology, broadly speaking, is the field of technological development in which the design of the final product is published and freely available for others to use and adapt to their own needs. This is in contrast to the standard commercial approach to hardware development where designs and software are often patented or kept as trade secrets, giving the company an advantage in the development and use of their technology [15]. Open-source hardware, such as the RepRap – an open-source 3D printer [16], syringe pumps [17], optics equipment [18] and many others [19- 21], are freely available to the user community and has led to even further accessibility in research. One of the chief benefits of open-source technology for scientific research lies in the significant reduction of the engineering cost and time to conceptualize and develop the hardware. A second benefit of an open-source design is the community of users linked through a common experience to discuss possible methods of use and troubleshooting ideas. A similar argument can be made for the software required to automate these systems. 3 This paper describes a feasible option for those that require a CVD system for their materials research but also have financial limitations that prevent the purchase of a commercial system. We have designed, assembled, and programmed an automated CVD system capable of synthesizing high-quality 2D materials for about $30,000 USD in hardware costs. This paper provides the details of the design and the software making it an open-source system to the research community. Researchers adapting this design will be able to customize the system to their needs and have confidence that the system is capable of producing high-quality 2D materials. Materials and methods Using the design described in this manuscript, a four gas, variable pressure CVD system, as shown in Fig 1, can be built for around $30,000 USD. A three gas, atmospheric pressure CVD system could be built for approximately $16,000 USD worth of parts. The bill of materials is contained in the supporting information (S1 Appendix) and is divided by subsystem such that the builder can easily swap out parts within a subsystem, add or remove entire subsystems, or even adapt subsystems and their respective LabVIEW drivers for use in other systems. The intake side (Fig 1A) includes the following subsystems: (1) gas delivery lines, (2) gas filter lines, (3) mass flow controllers, (4) gas mixing line, (5) intake manifold, and (6) pressure gauges. The deposition chamber (Fig 1B) is comprised of a quartz reaction tube and a Lindberg Blue M tube furnace. The exhaust side (Fig 1C) is divided into the following subsystems: (1) exhaust manifold, (2) overpressure relief, (3) vent valve, (4) exhaust line, (5) butterfly valve, (6) vacuum pump, and (7) snorkel manifold. Fig 1. Photograph of the variable pressure CVD system with 4 gas line option. The figure is annotated to indicate the (A) intake side, (B) deposition chamber, and (C) exhaust side of the system. 4 A brief overview of the main subsystems is given in this section. To assist with the system build, detailed construction notes are included in the supporting information, S2 Appendix. The construction notes are inclusive of all of the details needed for building the system including the order of assembly, high resolution subsystem images clearly labeled with part numbers, cable pinouts, control program setup, and user instructions for the automated program. Gas and pressure monitor system  The gas canisters for the system were placed into a gas cabinet immediately behind the table holding the CVD machine. If a different location is chosen for the gas cabinet, this subsystem may require some modifications. Individual gas filters are placed immediately following the gas delivery lines as shown in Fig 2. The mass flow controllers (MFCs) chosen were the M100 series from MKS Instruments (comparable MKS MFCs available today would be the G-Series). Flow rates were chosen to be 1000 sccm for argon and methane, 500 sccm for hydrogen. A 20 sccm MFC was installed for future expansion. The gases exiting the mass flow controllers are mixed and then routed through a manual diaphragm valve, which is connected to the intake manifold. The diaphragm valve is useful for system seal check and leak isolation as it isolates the MFCs and everything upstream from the remainder of the system. For variable pressure CVD applications, two capacitive manometers were used. One with a 1000 Torr full range, and the other with a 10 Torr full range. These manometers are connected to the intake manifold but can be moved to other KF25 fittings throughout the system to aid in leak detection. Both pressure gauge readings were fed into an integrated butterfly valve controller, which uses the downstream vacuum to match the system pressure to a software- controlled set point. Fig 2. Side view of the intake side of CVD system. The photograph shows the (A) gas lines, (B) gas filter lines, (C) MFCs, and (D) intake manifold. 5 Reaction chamber  The furnace is a Lindberg Blue M Mini-Mite Tube Furnace. This furnace requires a 25.4mm outer diameter tube and can operate at temperatures up to 1100 oC. Control of the furnace is possible with the controller on the front of the furnace, or via a serial connection. The reaction tube is a 609.6 mm long, 25.4 mm outer diameter sintered quartz tube that is coupled to the intake manifold via a KF25 to 25.4 mm quick disconnect and the exhaust manifold via a 25.4 mm quick disconnect to a KF40, as shown in Fig 3. The discoloration seen in the reaction tube downstream of the furnace center (Fig 3B) is mainly due to the copper deposition from graphene growths. Quartz tubes with fire-polished ends should be used to lower the likelihood of tube breakage during handling. The reaction tube can be easily changed by loosening the quick disconnects (Figs. 3A and 3C). Fig 3. CVD system opened to show reaction tube assembly. The quick connects (A) and (C) are identified, as well as the reaction tube (B). Data acquisition and system control  While communication with the furnace and butterfly valve was carried out via serial connections, the MFCs and capacitance gauges required analog inputs and outputs (digital communications options are available for purchase). These analog signals were supplied via a National Instruments CompactDAQ data acquisition system. The final design uses one 16- channel analog input module and two 4-channel analog output modules. Once a sample is placed into the reaction tube, the manual gas line and gas mixing valve are turned on, the furnace is turned on, and the vacuum pump is started, while all system controls are conducted on a computer running LabVIEW. The main elements of the controls programming are called virtual instruments, or VIs. For our system, the VIs include the device drivers, the manual control 6 program, and the automatic control program. All of the VIs used to control our system are included in the supporting files (S1 Folder Programs). Detailed system startup instructions using the VI are given in the construction notes (S1 Appendix). The device drivers are small LabVIEW programs that operate the system's active hardware (MFCs, pressure gauges, furnace, and butterfly valve). These driver programs can be opened and tested individually in order to test the operation of single active components. Or, in the case that the system design is modified, the drivers for the modified components can be reprogrammed independently, making it simple to customize the system design in order to fit individual needs. The manual program (ManualControls.vi) controls the LabVIEW user interface. A screen capture of the graphical user interface for the manual control program is shown in Fig 4. Once the system has been built and the active components have been tested, the manual controls VI can be used to check system seal under vacuum and run simple growths. The left side of the interface is where a user can input the manual growth conditions such as pressure set point, gas flow rate, and furnace temperature. This interface provides feedback on the system conditions in real time on the right side of the interface. The first row of plots indicates the system pressure and conveniently lights up as each pressure threshold is met. The second row provides real time information on the gas flow for each MFC. Again, the buttons below the plots light up once the condition has been met. When the manual control program is in use, both setpoints and actual values for the system pressure, furnace temperature, and gas flow rates are recorded in a dated log file. The interface is set up to be easily controlled and interpreted. To gain even greater accessibility, the open-source programming language Python [22] can be used to interface with the National Instruments data acquisition hardware using the PyDAQmx package [23]. 7 Fig 4. Screen capture of the CVD system manual control graphical user interface. For complex recipes and repeatable growths, the automated control program (Read&RunRecipe.vi) should be used. The automatic control program operates the CVD system through user-generated recipes. A sample recipe 'Quick.csv', is included in the supporting information. The contents of the comma separated value (csv) file is shown in Table 1. Users can use this recipe as a guide to create custom recipes for their desired process. Recipes are organized into stages, where Stage 0 sets the initial setpoints and then immediately ends. The next stages, starting at Stage 1, contain the process steps necessary to grow desired materials. Prior to executing a recipe, the automatic control program will conduct a system check to ensure that the furnace and butterfly valve are communicating with the program, the gases required for the recipe are available, the system is sealed, and the vacuum pump is operating. The system check can be followed by selecting the 'system check' tab in the top left corner of program. If any of these checks fail, the CVD system is re-pressurized to atmospheric pressure with argon and the program is stopped so that the fail can be addressed. Once the system check is successful, the program will begin to run the recipe and the user can monitor the progress by selecting the 'run recipe' tab to monitor the progress. A view of the automated control graphical user interface is shown in Fig 5. The right side of the interface is similar to the manual control program. However, the left side projects a table of the current stage and the corresponding set points for the selected recipe rather than user inputs. The graphical user interface also has a counter that records the time elapsed within the recipe as well as an estimated time of completion for the process. The user can also ensure that a log file is being recorded during the recipe by noting the blinking green box, right above the log file path in the lower left corner of the 8 interface. The graphical user interfaces were designed to make the critical deposition conditions readily available and easily found during a process run. Table 1. Stages, or process steps, within the Quick.csv recipe. Stage Pressure Setpoint (Torr) Temperature Setpoint (°C) Temperature Ramp Rate (°C/min, 30°C/min max) Ar flow (sccm) CH4 flow (sccm) H2 flow (sccm) C2H4 flow (sccm) Dwell time (min) Stage End Condition 0 680 25 0 0 0 0 1 500 200 600 0 0 0 2 300 500 30 0 400 0 0 Start Temp Ramp 3 0 500 4 680 25 500 0 0 0 0 0 200 0 10 Time End Fig 5. Screen capture of the CVD system automated control graphical user interface showing recipe progress. Results Monolayer tungsten disulfide (WS2) and WS2/ graphene heterostructures were made using the CVD system built as described. For the heterostructures, graphene film was grown in the CVD chamber on copper foil (Alfa Aesar, 0.5 mm thick, 99.99%) and transferred to single- side polished sapphire (C plane) substrates (see complete preparation in S2 Appendix- Growth Details). Prior to WS2 growth, the graphene film was annealed at 400 °C for 30 min under Ar/ H2 flow to remove any residual PMMA [24] left over from the transfer process. The annealed graphene sample was then placed face down in a quartz boat and held 7.5 mm above tungsten oxide (WO2.9)(Alfa Aesar) powder, which was placed in the center of the tube furnace. Upstream in a separate quartz boat, sulfur pieces of random sizes (Puratronic 99.9%, Alfa Aesar) were placed approximately 200 mm from the quartz boat containing the graphene sample, which was centered within the uniform hot zone of the furnace. The temperature of the sulfur was held at approximately 200 °C during the growth process, allowing the sulfur to vaporize. The 9 atmospheric pressure CVD growth for WS2 took place at 850 °C, with argon flowing at 70 sccm and hydrogen flowing at 15 sccm for 15 minutes. Similarly, we have also grown WS2 on single- side polished sapphire (C plane) substrates using the same process set up and recipe, omitting the graphene steps. The growth recipe for WS2 for use in the open-source automated CVD system is outlined in the supporting material (S2 Appendix- growth details). Tungsten Disulfide/ Graphene Heterostructures The CVD growth of WS2 on graphene/ sapphire substrates resulted in well-defined equilateral triangles, which is characteristic of single crystal domains [25] as shown in the optical micrograph (Zeiss Axio Imager.M2m) within Fig 6A. Raman spectra were collected (Horiba LabRAM) on graphene and WS2 as shown in Fig 6B and 6C, giving an indication of the structural properties for the sample. Using an excitation source with a wavelength of 532 nm, the sample exhibits a strong peak at 1582 cm-1, which corresponds to the G band, while the peak at 2700 cm-1 is representative of the 2D band, as expected for graphene [26]. Near the edge of the graphene we also see a small peak at 1350 cm-1 due to the disorder- induced D band. The presence of monolayer WS2 is also validated through Raman spectroscopy using a 532 nm excitation wavelength as shown in Fig 6C. The spectra in Fig 6C include first order modes at the Brillouin zone center, 𝐸2𝑔 1 , at 356 cm-1 and A1g at 418 cm-1, typical of monolayer WS2 [27], [28]. Fig 6. WS2/ graphene heterostructures grown in the open-source CVD system. (A) Optical micrograph of 2D heterostructures showing WS2 triangles. (B) Raman spectra of graphene. (C) Raman spectra of WS2. Monolayer WS2 10 Additionally, we demonstrate that controlled growth to a monolayer is possible with this open-source automated CVD system. First, the optical micrograph in Fig 7A confirms growth of WS2 on sapphire in the characteristic triangular domains [25]. The thickness of as- grown WS2 on sapphire was measured by atomic force microscopy (AFM). The height profile in the AFM image (Fig 7B) indicates a height difference of 0.7 nm as evidence of monolayer growth [25], [29]. Fig 7. Synthesis of monolayer WS2 on sapphire grown with the open-source CVD system. (A) Optical micrograph showing WS2 triangle domains. (B) AFM characterization of WS2 on sapphire showing the height profile. (Bruker Nano Dimension FastScan) The AFM color scale represents the sample height with a range of 0- 145 nm. Discussion The benefits of an open-source automated CVD system include: 1) a lower time investment for each growth cycle, 2) increased repeatability through the use of recipes, and 3) the ability to audit past growths since the automated program will record each growth cycle. Additional benefits of this system include the feasibility for a high level of customization, the ability to modify and update the system as needed, and a significant gain of experience in vacuum systems and system automation for those conducting the build. The open-source CVD system is also much cheaper than purchasing a commercial CVD, costing about $30,000 USD for parts as compared with $95,000 USD for a (low end) commercial turn-key system (quote obtained in 2015). Once you factor in the additional savings of time and labor, one can anticipate a savings exceeding half of the cost for a commercial system. The software for the system was also designed to be simple and easy to use, allowing for growths with minimal tuning of process variables. Conclusions 11 In summary, an open-source CVD system has been designed, sourced, built and programmed. This CVD system is the first open-source design that has shown comparable capabilities to commercially manufactured systems. This system is capable of both atmospheric pressure and low-pressure growths, temperatures up to 1100 ℃, and substrates up to ~20 mm in width. Additionally, with the detailed parameter and recipe logging, past growths can quickly be audited for possible issues or repeated, allowing for quick system troubleshooting and high process repeatability. With the system described here, we have grown 2D materials such as graphene, WS2, as well as graphene/ WS2 vertical heterostructures, with exceptional quality. As open-source scientific hardware continues to increase, it is conceivable that the rate of scientific discovery could increase as researchers are freed to spend more time and effort conducting experiments, and less time building or troubleshooting a custom-built system. An open-source CVD system such as this one will help to propel 2D nanomaterials research forward by lowering financial barriers and allowing researchers with smaller budgets to join this emerging area of research. By sharing this design as an open-source, we hope to put the opportunity of research into many more labs, with a fairly reasonable startup cost thus accelerating the development and understanding of 2D materials. Acknowledgements This work was supported through startup funds provided by the Micron School of Materials Science and Engineering at Boise State University and a gift from the Micron Foundation. D.B. acknowledges additional support through fellowships provided by the Idaho Space Grant Consortium and the Nuclear Regulatory Commission. The authors would also like to acknowledge Corey Efaw and Mike Hurley of the Applied Electrochemistry and Corrosion 12 research lab for their assistance with the electrochemical transfer of graphene films. Atomic force micrographs were obtained through the Boise State University Surface Science Laboratory. References 1. Novoselov KS, Geim AK, Morozov SV, Jiang D, Zhang Y, Dubonos SV, et al. Electric field effect in atomically thin carbon films. Science. 2004;306:666–9. doi:10.1126/science.1102896 2. Dean CR, Young AF, Meric I, Lee C, Wang L, Sorgenfrei S, et al. Boron nitride substrates for high-quality graphene electronics. Nature Nanotech. 2010;5:722–6. doi:10.1038/nnano.2010.172 3. He G, Nathawat J, Kwan C-P, Ramamoorthy H, Somphonsane R, Zhao M, et al. Negative differential conductance & hot- carrier avalanching in monolayer WS2 FETs. Sci. Rep. 2017;7:11256. doi:10.1038/s41598-017-11647-6 4. Koski K, Cui Y. The new skinny in two-dimensional nanomaterials. ACS Nano. 2013;7:3739–43. doi:10.1021/nn4022422 5. Cong WT, Tang Z, Zhao XG, Chu JH. Enhanced magnetic anisotropies of single transition-metal adatoms on a defective MoS2 monolayer. Sci. Rep. 2015;5:9361. doi:10.1038/srep09361 6. Vogel EM, Robinson, JA. Two-dimensional layered transition-metal dichalcogenides for versatile properties and applications. MRS Bulletin. 2015;40(7):558-63. doi:10.1557/mrs.2015.120 7. Geim AK, Grigorieva IV. Van der Waals heterostructures. Nature. 2013;499:419–25. doi:10.1038/nature12385 13 8. Wang QH, Kalantar-Zadeh K, Kis A, Coleman JN, Strano, MS. Electronics and optoelectronics of two-dimensional transition metal dichalcogenides. Nature Nanotech. 2012;7:699–712. doi:10.1038/nnano.2012.193 9. Fiori G, Bonaccorso F, Iannaccone G, Palacios T, Neumaier D, Seabaugh A, et al. Electronics based on two-dimensional materials. Nature Nanotech. 2014;9:768-79. doi:10.1038/nnano.2014.207 10. Butler SZ, Hollen SM, Cao L, Cui Y, Gupta JA, Gutierrez HR, et al. Progress, challenges, and opportunities in two-dimensional materials beyond graphene. ACS Nano. 2013;7:2898-2926. doi:10.1021/nn400280c 11. Wang S, Wang X, Warner JH. All chemical vapor deposition growth of MoS2: h-BN vertical van der Waals heterostructures. ACS Nano. 2015;9:5246-54. doi:10.1021/acsnano.5b00655 12. Shi Y, Zhang H, Chang W-H, Shin HS, Li L-J. Synthesis and structure of two- dimensional transition-metal dichalcogenides. MRS Bulletin. 2015;40(7):566-576. doi:10.1557/mrs.2015.121 13. Chen K, Wan X, Xu J. Epitaxial stitching and stacking growth of atomically thin transition-metal dichalcogenides (TMDCs) heterojunctions. Adv. Funct. Mater. 2017;27:1603884. doi:10.1002/adfm.201603884 14. Tan C, Cao X, Wu XJ, He Q, Yang J, Zhang X, et al. Recent Advances in Ultrathin Two- Dimensional Nanomaterials. Chem. Rev. 2017;117(9):6225-6331. doi:10.1021/acs.chemrev.6b00558 15. Deek FP, McHugh JAM. Open source: technology and policy. Cambridge University Press; 2007. 14 16. Reprap.org [Internet]. RepRap - RepRapWiki [cited 2018 May 13]. Available from: http://reprap.org/ 17. Wijnen B, Hunt EJ, Anzalone GC, Pearce JM. Open-source syringe pump library. PLoS ONE. 2014; 9(9):e107216. doi:10.1371/journal.pone.0107216 18. Zhang C, Anzalone NC, Faria RP, Pearce JM. Open-source 3D-printable optics equipment. PLoS One. 2013;8(3):e59840. doi:10.1371/journal.pone.0059840 19. Rosenegger DG, Tran CHT, LeDue J, Zhou N, Gordon GR. A high performance, cost- effective, open-source microscope for scanning two-photon microscopy that is modular and readily adaptable. PLoS One. 2014;9:e110475. doi:10.1371/journal.pone.0110475 20. Dryden MDM, Wheeler AR. DStat: A versatile, open-source potentiostat for electroanalysis and integration. PLoS ONE. 2015;10(10): e0140349. doi:10.1371/journal.pone.0140349 21. Wittbrodt BT, Squires DA, Walbeck J, Campbell E, Campbell WH, Pearce JM. Open- Source Photometric System for Enzymatic Nitrate Quantification. PLoS ONE. 2015: 10(8): e0134989. doi:10.1371/journal.pone.0134989 22. Python.org [Internet]. Python Software Foundation [cited 2018 May 13]. Available from https://www.python.org 23. pythonhosted.org/PyDAQmx/ [Internet] Cladé P. PyDAQmx: a Python interface to the National Instruments DAQmx driver. [cited 2018 May 13]. Available from: http://pythonhosted.org/PyDAQmx/ 24. Salehi‐Khojin A, Estrada D, Lin KY, Bae M-H, Xiong F, Pop E, et al. Polycrystalline graphene ribbons as chemiresistors. Adv. Mater. 2012;24:53-57. doi:10.1002/adma.201102663 15 25. Gutierrez HR, Perea-Lopez N, Elias AL, Berkdemir A, Wang B, Lv, et al. Extraordinary room-temperature photoluminescence in triangular WS2 monolayers. Nano Lett. 2013;13(8):3447-3454. doi:10.1021/nl3026357 26. Malard LM, Pimenta MA, Dresselhaus G, Dresselhaus MS. Raman spectroscopy in graphene. Phys. Rep. 2009;473:51-87. doi:10.1016/j.physrep.2009.02.003 27. Henck H, Aziza ZB, Pierucci D, Laourine F, Reale F, Palczynski P, et al. Electronic band structure of two-dimensional WS2/ graphene van der Waals heterostructures. Phys. Rev. B. 2018;97:155421-6. doi:10.1103/physrevb.97.155421 28. Berkdemir A, Gutierrez HR, Botello-Mendez AR, Perea-Lopez N, Elias AL, Chia C, et al. Identification of individual and few layers of WS2 using Raman spectroscopy. Sci. Rep. 2013;3:1755. doi:10.1038/srep01755 29. Zhang Y, Zhang Y, Ji Q, Ju J, Yuan H, Shi Y, et al. Controlled growth of high-quality monolayer WS2 layers on sapphire and imaging its grain boundary. ACS Nano. 2013;7(10):8963-8971. doi:10.1021/nn403454e Supporting information S1 Appendix. Bill of materials for CVD system divided by subsystem. S2 Appendix. Construction notes. S3 Appendix. Growth details. S1 Folder Programs. Folder contains all of the software needed to automate the CVD system. S2 Folder Support Drawings. Folder contains construction drawings. 16 Click here to download Figure Fig1.tif Click here to download Figure Fig2.tif Click here to download Figure Fig3.tif Figure Click here to download Figure Fig4.tif Figure Click here to download Figure Fig5.tif Figure Click here to download Figure FIG6.tif Figure Click here to download Figure FIG7.tif Supporting Information Click here to access/download Supporting Information S1 Appendix Bill of materials for CVD system.xls Supporting Information Click here to access/download Supporting Information S2 Appendix Construction Notes.pdf Supporting Information Click here to access/download Supporting Information S3 Appendix Growth Details.pdf Supporting Information - Compressed/ZIP File Archive Supporting Information - Compressed/ZIP File Archive Click here to access/download S1 Folder Programs.zip Supporting Information - Compressed/ZIP File Archive Click here to access/download Supporting Information - Compressed/ZIP File Archive S2 Folder Support Drawings.zip
1908.04180	1	1908	2019-08-12T14:44:23	Magnetic resonance force microscopy with a one-dimensional resolution of 0.9 nanometers	[ "physics.app-ph", "cond-mat.mes-hall" ]	Magnetic resonance force microscopy (MRFM) is a scanning probe technique capable of detecting MRI signals from nanoscale sample volumes, providing a paradigm-changing potential for structural biology and medical research. Thus far, however, experiments have not reached suffcient spatial resolution for retrieving meaningful structural information from samples. In this work, we report MRFM imaging scans demonstrating a resolution of 0.9 nm and a localization precision of 0.6 nm in one dimension. Our progress is enabled by an improved spin excitation protocol furnishing us with sharp spatial control on the MRFM imaging slice, combined with overall advances in instrument stability. From a modeling of the slice function, we expect that our arrangement supports spatial resolutions down to 0.3 nm given suffcient signal-to-noise ratio. Our experiment demonstrates the feasibility of sub-nanometer MRI and realizes an important milestone towards the three-dimensional imaging of macromolecular structures.	physics.app-ph	physics	Magnetic resonance force microscopy with a one-dimensional resolution of 0.9 nanometers U. Grob,†,§ M. D. Krass,†,§ M. Héritier,† R. Pachlatko,† J. Rhensius,† J. Košata,† B. A. Moores,†,‡ H. Takahashi,†,¶ A. Eichler,† and C. L. Degen∗,† †Department of Physics, ETH Zurich, Otto Stern Weg 1, 8093 Zurich, Switzerland. ‡Present address: JILA, National Institute of Standards and Technology and the University of Colorado, Boulder, Colorado 80309, USA and Department of Physics, University of Colorado, Boulder, Colorado 80309, USA. ¶Present address: JEOL RESONANCE Inc., Musashino, Akishima, Tokyo 196-8558, Japan. §These authors contributed equally to this work. E-mail: [email protected] Abstract Magnetic resonance force microscopy (MRFM) is a scanning probe technique ca- pable of detecting MRI signals from nanoscale sample volumes, providing a paradigm- changing potential for structural biology and medical research. Thus far, however, experiments have not reached sufficient spatial resolution for retrieving meaningful structural information from samples. In this work, we report MRFM imaging scans demonstrating a resolution of 0.9 nm and a localization precision of 0.6 nm in one di- mension. Our progress is enabled by an improved spin excitation protocol furnishing us with sharp spatial control on the MRFM imaging slice, combined with overall advances in instrument stability. From a modeling of the slice function, we expect that our ar- rangement supports spatial resolutions down to 0.3 nm given sufficient signal-to-noise 1 ratio. Our experiment demonstrates the feasibility of sub-nanometer MRI and real- izes an important milestone towards the three-dimensional imaging of macromolecular structures. Keywords magnetic resonance force microscopy, nanoscale magnetic resonance imaging, scanning probe microscopy, nuclear magnetic resonance Introduction The goal of nanoscale magnetic resonance imaging ("nano MRI") is the three-dimensional visualization of nuclear spin densities inside materials with near-atomic spatial resolution and on length scales of up to a few 100 nm. Such images are expected to provide fundamental insight into the structure and composition of matter, especially in the field of structural molecular biology. For example, nano MRI images could serve as templates for modeling the global arrangement of large protein assemblies. If sub-nanometer resolution can be realized, nano MRI could even allow for a direct imaging of tertiary or secondary protein structure, and ultimately, the atomic arrangement. Important advantages of MRI compared to other structural imaging techniques, such as electron tomography, are its high chemical selectivity and the fact that single copies of molecules can be imaged in a non-destructive manner. One promising candidate for nano MRI is magnetic resonance force microscopy (MRFM).1 -- 5 The method employs an ultrasensitive nanomechanical transducer to detect the interaction between nuclear spins in the sample and a nanoscale magnetic tip by means of a magnetic force. Thanks to major advances in mechanical transduction6 -- 11 and magnetic gradient gen- eration,12,12 -- 16 researchers have in recent years greatly improved the sensitivity of MRFM. Latest imaging experiments reported sensitivities of order 50 − 100 nuclear moments, corre- sponding to voxel sizes between (3.5 nm)3 − (5.5 nm)3 for statistically polarized protons in 2 organic material.17,18 In principle, MRFM even offers the sensitivity required to detect sin- gle proton magnetic moments,15 but it is unclear at present whether this sensitivity can be extended to the context of three-dimensional MRI. Unlike other nano MRI techniques, like nitrogen-vacancy-center NMR,19,20 MRFM is capable of handling larger objects (> 20 nm) by adjusting the sizes of the mechanical sensor and the magnetic tip. In order to provide meaningful information about the macromolecular arrangement of protein complexes, the spatial resolution must be of order ∼ 1 nm. Despite providing suf- Imaging ficient detection sensitivity, recent MRFM scans have fallen short of this goal. experiments on single tobacco mosaic virus particles3 have reported a best-effort resolution of 4 nm in one dimension, limited by a combination of scan step size, available magnetic gradient, thermomechanical force noise, and instrument stability. Other experiments have shown feature sizes of order 5 − 10 nm.17,21 Recently, Rose et al.18 have reported a nomi- nal resolution of ∼ 2 nm for a polystyrene-coated nanowire using a novel Fourier encoding method. In this paper, we demonstrate MRFM scans with a one-dimensional resolution of 0.9 nm± 0.2 nm and a localization precision of 0.6 nm ± 0.1 nm. This progress is enabled through: (i) an improved spin inversion protocol providing a sharply defined imaging slice, (ii) the use of a state-of-the-art magnetic gradient exceeding 106 T/m, (iii) the use of a suitable nanorod sample geometry, and (iv) a high instrument stability with drifts of less than 2 nm over 24h. As a result of our efforts, our experimental resolution is limited solely by the sensor noise (which we have not improved within this work) and not by other factors. Our demonstration is achieved with a top-down fabricated cantilever that is suitable for loading large molecules and amenable to modular sample preparation techniques. With this result, we take a critical step towards three-dimensional imaging of biological samples on the ∼ 1 nm lengthscale. 3 Figure 1: (a) Configuration of the MRFM experiment. (b) Sketch of nanorod apex, 1H-rich adsorbate layer, magnetic tip and stripline in cross section. (c) Scanning electron micrograph of a silicon nanorod and cantilever, similar to the device used in experiments. (d) Calculated magnetic field map B(x, z) near the FeCo nanomagnet.21,22 An external bias field B0 = 5.88 T is applied along z to polarize the nanomagnet and to stabilize the quantization axis of the nuclear spins. Contours indicate the 1H NMR frequency in the given field. The shaded region represents a resonant slice with rf pulse frequency frf = 255 MHz and FM modulation depth fFM = 500 kHz. The field map is calculated from a combination of AFM topography and MRFM calibration scans (see SI). (e) Magnetic gradient Gx ≡ ∂xB calculated from the field map shown in panel d. 4 zcantileverb)a)c)d)e)cantileverlaserbeamstriplinemagnetnanorodxyznanorodxstriplinesubstratemagnetadsorbatelayer2µmExperimental Setup A schematic of our MRFM setup is shown in Fig. 1a. The sample, in our case a silicon nanorod with a thin film of 1H-containing adsorbate molecules (see Fig. 1b,c), is affixed to the end of the cantilever force transducer and positioned above a nanomagnet. The nanomagnet produces a highly localized magnetic field B = B(r) and an associated strong magnetic gradient G = ∇B (see Fig. 1d,e). The gradient generates an attractive or repulsive force on the nuclear spins in the sample. To measure the magnetic force, nuclear spins are periodically inverted by adiabatic radio-frequency (rf) pulses to drive the transducer at its kHz mechanical resonance. The output signal is provided by the oscillation amplitude of the cantilever, which is proportional to the spin force and which we detect by an optical interferometer. As in conventional MRI, image generation takes advantage of a spatially localized ex- citation of nuclear spins inside the sample volume. Radio-frequency pulses act selectively on spins whose Larmor frequencies fL = γnB are within the excitation bandwidth of the rf pulse frequency frf ± fFM. Here, γn is the nuclear gyromagnetic ratio (γn = 42.58 MHz/T for protons), frf is the rf carrier frequency and fFM is the modulation depth of rf pulses. Due to the locally inhomogeneous field, only spins within a thin spatial slice are inverted and generate a driving force on the mechanical sensor (see Fig. 1d). Specifically, the force signal is given by the convolution of the slice with the sample's three-dimensional spin density ρ(r),3,4,23 (cid:90) R3 F 2 spin = µ2 dr(cid:48)ρ(r − r(cid:48))G2 x(r(cid:48))ξ(r(cid:48)) . (1) where µ = hγn/2 is the nuclear magnetic moment, h = 6.63· 10−34 J/Hz is Planck's constant, Gx is the x-component of the magnetic gradient, and ξ(r(cid:48)) contains the shape of the spatial slice. We remind that because we detect statistical rather than thermal spin polarization,24,25 our signal is proportional to the variance F 2 spin of the spin force, rather than the magnitude. 5 The slice function ξ(r(cid:48)) ≡ ξ(fL − frf) ≡ ξ(∆f ) depends on the detuning ∆f = fL − frf between the (location-dependent) Larmor frequency of a nuclear spin, fL = γnB(r(cid:48)), and the rf excitation frequency frf. We control ξ(∆f ) by tuning shape and parameters of the adiabatic rf pulse modulation. In this work, we are particularly interested in the frequency selectivity of pulses, since this determines the sharpness of the slice edges. At the same time, the rf excitation must be robust against variations in pulse amplitude, because hundreds of coherent spin reversals are required to build up a detectable oscillation of the mechanical sensor. In a first part of this study, we have explored several adiabatic rf modulation schemes for their ability to robustly invert spins in a narrow, well-defined frequency bandwidth. We find experimentally and through simulations that hyperbolic secant (HSn) pulses26 -- 28 are well-suited for this task. Other schemes that we explored include Gauss,27 WURSTn,29 and Sin/Cos30 modulation. HSn pulses involve both amplitude (AM) and frequency (FM) modulation of the rf field, B1,mod(T ) = B1 sech(βT n) frf,mod(T ) = frf + fFM (cid:20) c(n, β) sech2(βT (cid:48)n) dT (cid:48)(cid:19) (cid:21) − 1 (cid:18)(cid:90) T −1 (2) (3) ≤ 1 is a normalized where B1 is the in-plane component of the rf field amplitude, T = 4t time running over half a cantilever oscillation period, and c(n, β) is a unit-less factor that normalizes the integral for a symmetrical modulation around frf. The parameters n and β are integers that control the truncation and steepness of the frequency modulation. The HSn profile for n = 2 and β = 8 is shown in Fig. 2a. Tc We calculate slice functions ξ(∆f ) for the HSn and other modulation schemes by simu- lating the spin reversal and computing the Fourier coefficient at the mechanical resonance fc. Fig. 2b shows the expectation value (cid:104) Iz(t)(cid:105) of the single spin operator Iz during the reversal. (cid:104) Iz(t)(cid:105) is calculated using density matrices under a piece-wise constant Hamiltonian (see SI). 6 Figure 2: Signal encoding and generation by hyperbolic secant (HSn) spin reversals. (a) Amplitude B1,mod(t) and frequency frf,mod(t) of adiabatic rf pulses. The HSn modulation is shown in red and conventional trapezoidal modulation is shown in gray for comparison. Parameters are n = 2 and β = 8. Tc = 1/fc represents one cantilever oscillation period (here Tc ∼ 300 µs). (b) Simulated nuclear spin polarization (cid:104) Iz(t)(cid:105) in response to the rf pulses in panel a. (c) Cantilever oscillation amplitude in response to the driving force of panel b. (d) Simulated fidelity F(∆f ) of spin reversals (Eq. 5) as a function of nuclear spin detuning ∆f = fL − frf. The excitation bandwidth is 2fFM = 300 kHz (blue dashed line). Slice profiles are shown for HSn modulation with B1 = 5.3 mT (red solid line), as well as for trapezoidal modulation with B1 = 2.5 mT (gray solid line) and B1 = 2.0 mT (gray dashed line). The arrows indicate the sharp slice edge of 27 kHz. The top scale indicates the spatial slice width in a gradient of G = 2 · 106 T/m. (e) Experimental MRFM signal as a function of the modulation depth fFM for HSn (red) and trapezoidal (gray) modulation. The rf carrier frequency is frf = 254 MHz and the rf amplitude is B1 = 5.3 mT. 7 02004000200400600800TrapezoidHS28spin signal [aN²]mod. depth fFM [kHz]• • •B1,mod...N0fFM-fFM0.501Tc2Tc• • •• • •• • •00frf,mod - frfB1-0.5xc-xcIZcantilevertime1234-150015001fidelitydetuning [kHz]excitation bandwidth-2-1012distance [nm]a)e)d)b)c)The first real Fourier coefficient is a1 = 4 Tc (cid:90) Tc/2 0 2 (cid:104) Iz(t)(cid:105) cos(2πfct) dt . (4) The cantilever motion follows the periodic force generated by (cid:104) Iz(t)(cid:105) as shown in Fig. 2c. Because of order N = 2τ /Tc ∼ 102 − 103 reversals are needed to ring up the mechanical sensor, where τ is the resonator's time constant, spin inversions must be highly efficient with no loss of magnetization over many hundreds of cycles N. We account for the inversion efficiency through the fidelity (cid:104)(cid:104) Iz(Tc/2)(cid:105) − (cid:104) Iz(0)(cid:105)(cid:105)N F = , (5) where (cid:104) Iz(0)(cid:105) and (cid:104) Iz(Tc/2)(cid:105) are the expectation values of Iz at the beginning and the end of the inversion pulse, respectively. For the HSn modulation shown in Fig. 2a, all spins in the slice flip at approximately the same instance in time, therefore a1 ≈ 4/π. The slice function is then given by ξ(∆f ) = a1(∆f )F(∆f ) ≈ 4 π F(∆f ) (6) In Fig. 2d we plot the fidelity F(∆f ) of spin inversions for a sequence of N = 140 HSn pulses as a function of the detuning from the slice center frequency, for a modulation depth of fFM = 150 kHz. This number of pulses N approximately corresponds to the number of reversals within the τ ≈ 20 ms ring-up time of the feedback-damped cantilever. The figure shows that HSn pulses are clearly very effective at inverting spins. Importantly, the frequency slice is sharply defined with only ∼ 30 kHz between complete inversion (F = 1) and no signal (F = 0). By comparison, common trapezoidal pulses with a linear frequency ramp3,17,21 produce an ill-defined slice profile due to the sudden turn-on of the rf amplitude (gray curves). Further advantages of the HSn modulation are its robustness in the presence 8 of B1 variations (see SI) and nuclear spin-spin interactions.28 Results and Discussion Figure 3: Series of 1H NMR spectra as a function of vertical approach distance d = 10 − 150 nm. The schematics show the configuration of nanorod and nanomagnet; the adsorbate layer is color-coded with the gradient Gx (scale of Fig. 1e). (a) Spectra taken with the tip positioned over the edge of the nanomagnet (x = −200 nm, where x is the center-to- center distance between cantilever and nanomagnet). The white star marks the position of highest gradient of ∼ 6 · 106 T/m. (b) Spectra taken with the tip positioned in front of the nanomagnet (x = −400 nm). Modulation depth is fFM = 300 kHz in panel a and fFM = 150 kHz in panel b. The bias field is 5.88 T. We experimentally demonstrate high-resolution 1H NMR imaging using the ∼ 1 nm − thick adsorbate film3,17,31 on a silicon nanorod.32 The nanorod has an edge length of ∼ 500 nm (see Fig. 1b) and is fabricated using standard silicon lithography (see SI). Our choice of test sample has two motivations: first, the nanorod has a well-defined geometry and the natural adsorbate layer provides a strong 1H NMR signal. Second, the nanorod can be batch produced and can undergo water dipping and shock-freezing. This capability is important for preparing biological samples in future experiments. For the present study, no sample is loaded onto the nanorod and the natural adsorbate film provides the only 1H NMR sig- nal. The nanorod is glued to the end of an ultrasensitive silicon cantilever that is in turn mounted in the custom MRFM apparatus (Fig. 1b). The cantilever used in this study has, after loading, a natural frequency of fc ∼ 3.5 kHz, a quality factor of Q ∼ 28, 000, and a 9 250255260265270150nmspin signal [aN²]frf [MHz]100 aN2130nm110nm 90nm 70nm 50nm 40nm 30nm 20nm 10nmb)249254100 aN2frf [MHz]a)zxspring constant of k = 82 µN/m (see SI). The microscope is operated in high vacuum at the bottom of a cryogen-cooled helium cryostat (T ∼ 4.7 K). Further details on the apparatus are given in Ref. 21. In a first set of experiments, we examine the resonant slice profile and optimize the parameters of the HSn protocol. In Fig. 2e, we show the signal magnitude F 2 as a function of modulation depth fFM. We observe a roughly linear scaling of F 2 with fFM, as expected, because the resonant slice volume, and hence the signal, are proportional to fFM. We also find that modulation depths as small as fFM = 75 kHz still lead to a detectable signal. By using the gradient Gx = 2.3 · 106 T/m extracted from the tip model in Fig. 1d and the simulated slice shape for the experimental settings, we find that the total slice width is only spin spin 0.7 nm.1 We have also measured the corresponding signal for a conventional trapezoidal pulse modulation; here, no signal is detectable for fFM < 150 kHz and detection is highly susceptible to B1 miscalibration (see Fig. 2e). We further examined HSn pulses for a range of n, β parameters; these results are provided as SI. Figure 2e confirms that HSn pulse modulation is well-suited for precise, high-resolution imaging. We next record a series of 1H NMR spectra as a function of the vertical approach distance d. These spectra serve to determine the optimum position and rf frequency frf for high- resolution imaging scans. A first series of spectra, shown in Fig. 3a, is taken with the cantilever centered over the magnet's edge. In this position, spectral peaks are broad, because a large number of slices penetrates the sample volume. Peak widths exceeding 15 MHz are found for close approach distances (d < 40 nm), corresponding to tip fields in excess of 450 mT (see SI). In all spectra, the low-frequency ends (fL ≈ 252 MHz) contain signal from spins that experience little tip field, while the high-frequency ends reflect locations over the magnet where the tip field is high. Figure 3b shows a second series of NMR spectra recorded with the cantilever positioned in front of the magnet. In this configuration, the sample surface and slice edge are oriented 1Note that the experimental settings were different from the ones shown in Fig. 2d; see SI for details. 10 nearly tangentially (see schematic in Fig. 3b) and the spectra become narrow. This tangential configuration is therefore well-suited for demonstrating high-resolution imaging scans along the x-direction, because a large portion of the sidewall 1H adsorbate layer can be moved into the imaging slice within a few nm. This results in a large signal change over a short distance. Figure 4: Lateral imaging scans showing 0.9 nm spatial resolution. (a) Two consecutive line scans taken within 31 hours. Averaging time is 4 min per point. (b) Series of x scans for excitation frequencies frf = 252.3− 252.8 MHz. The sharp signal rise around x = 20− 50 nm indicates the position where the nanowire 1H layer enters the slice. The scan height is z = 30 nm. The inset shows the 90 % confidence interval of the fit to Eq. 7 for frf = 252.6 MHz. Fig. 4a shows two lateral scans in the tangential nanorod configuration at fc = 252.6 MHz. The nominal step size for these high-resolution scans is 1.6 nm. We calibrate the tip x position by a separate line scan over the known stripline topography, and correct the nominal scanner position for static cantilever deflections caused by electrostatic tip-magnet interactions (see SI). The sudden onset of signal marks the position where the adsorbate film enters a resonant slice. The two scans are taken 31 hours after each other. The lateral offset between the scans is only 1.8 nm, demonstrating an excellent long-term stability of the experiment. Instrument stability and low drift are critical for recording undistorted images and avoiding artifacts in 11 4050607080901000100200 252.3 MHz 252.4 MHz 252.5 MHz 252.6 MHz 252.7 MHz 252.8 MHzspin signal [aN²]tip position [nm]sx0x050 aN25 nm0204060801001200100200spin signal [aN²]tip position [nm] scan2 31h later1.8nm scan1a)b)three-dimensional MRI. To quantify the image resolution supported by our MRFM configuration we have per- formed x-scans for six resonant slices between frf = 252.3 − 252.8 MHz (Fig. 4b). By fitting the signal onset (see below) and plotting the onset position x0 as a function of slice frequency frf we can directly measure the lateral field gradient at this position, Gx = 0.56 · 106 T/m (see SI). Although this gradient is significantly smaller than the maximum value in Fig. 3a, it still allows for producing nanometer-localized signal features. To analyze the scans, we fit the signal onset by a hyperbolic tangent step function (Fig. 4b, inset), (cid:16) w (x−x0)(cid:17)−1 1 + e− 4 F 2 spin(x) = F 2 spin,max (7) where x0 is the position of the signal onset, w the characteristic width of the signal edge, and F 2 the signal step height. We find a characteristic width for the scans in Fig. 4b of w ∼ 10 nm. This width is not indicative of the spatial resolution, however, because it is determined by the convolution of the sample with the residual curvature of the tangential spin,max imaging slice (see SI). When the signal error is dominated by statistical fluctuations, a suit- able metric for the spatial resolution must be based on the signal-to-noise ratio (SNR) of the measurement. Here, we compare the maximum signal slope to the noise of the measurement (error bars in Fig. 4b, inset) and obtain (cid:12)(cid:12)∂F 2 spin/∂x(cid:12)(cid:12)max F 2 noise σx,slope = . (8) From a slope of (cid:12)(cid:12)∂F 2 spin/∂x(cid:12)(cid:12)max = 14 − 20 aN2/nm and a total noise variance of F 2 noise = 13 − 14 aN2, we find an uncertainty-limited spatial resolution of σx,slope = 0.9 ± 0.2 nm for a set of 11 scans. Another method consists of computing the mean fit uncertainty σx0 = 0.6±0.1 nm of the onset positions x0, from which we gain an estimate of the localization precision for these scans. Overall, our data clearly demonstrate that MRFM is able to 12 perform one-dimensional MRI with sub-nanometer spatial resolution. The representative scans shown in Fig. 4b are taken in a tangential configuration (cf. Fig. 3b) to create a sharp signal feature with little convolution by the slice's point spread function. The magnetic gradient in this configuration is, however, not very large (Gx ∼ 0.56 · 106 T/m), limiting both the sensitivity and the spatial resolution. We can use our tip model (Fig. 1d,e) to estimate the maximum gradients produced in our experimental arrangement. From the numerical model of the tip, we find a maximum G ≈ 6 · 106 T/m at z = 10 nm (white star in Fig. 3a), with a corresponding expected slice width of 0.3 nm. Summarizing, we demonstrate one-dimensional scans with a resolution of 0.9 nm, a pre- cision of 0.6 nm and a minimum measured slice width of 0.7 nm. Our work supplies proof that the complex protocols involved in MRFM are compatible with performing MRI imaging at a sub-nanometer scale. At the core of this improvement is a spin inversion protocol that enables sharply defined imaging slices. We also decreased to below 1 nm all technical sources of blur in our setup, such as stage drift and sample-gradient convolution. Our current experi- ment is sensitive to spin ensembles containing about 103−105 hydrogen atoms, depending on the sample position in the gradient field. Future work will focus on improving the transducer sensitivity, such that three-dimensional images with 1 nm voxel size, corresponding to about 102 hydrogen atoms, become possible. Several routes lead towards this goal, for instance, a reduction of sensor dissipation through surface treatment and spatial design,8 -- 11,33,34 the use of higher magnetic field gradients,12,14,15 or a lower operating temperature.35 Acknowledgement We thank Ute Drechsler for her support with the clean-room fabrication. This work was sup- ported by Swiss National Science Foundation (SNFS) through the National Center of Com- petence in Research in Quantum Science and Technology (NCCR QSIT) and the Sinergia grant (CRSII5_177198/1), and the European Research Council through the ERC Starting 13 Grant "NANOMRI" (Grant agreement 309301). References (1) Sidles, J. A. Applied Physics Letters 1991, 58, 2854. (2) Sidles, J. A.; Garbini, J. L.; Bruland, K. J.; Rugar, D.; Züger, O.; Hoen, S.; Yan- noni, C. S. Reviews of Modern Physics 1995, 67, 249. (3) Degen, C. L.; Poggio, M.; Mamin, H. J.; Rettner, C. T.; Rugar, D. Proceedings of the National Academy of Sciences of the United States of America 2009, 106, 1313. (4) Poggio, M.; Degen, C. L. Nanotechnology 2010, 21, 342001. (5) Nichol, J. M.; Naibert, T. R.; Hemesath, E. R.; Lauhon, L. J.; Budakian, R. Physical Review X 2013, 3, 031016. (6) Mamin, H. J.; Rugar, D. Applied Physics Letters 2001, 79, 3358. (7) Moser, J.; Guttinger, J.; Eichler, A.; Esplandiu, M. J.; Liu, D. E.; Dykman, M. I.; Bachtold, A. Nature Nanotechnology 2013, 8, 493 -- 496. (8) Tao, Y.; Boss, J. M.; Moores, B. A.; Degen, C. L. Nature Communications 2014, 5, 3638. (9) Rossi, N.; Braakman, F. R.; Cadeddu, D.; Vasyukov, D.; Tutuncuoglu, G.; i Mor- ral, A. F.; Poggio, M. Nature Nanotechnology 2017, 12, 150. (10) de Lepinay, L. M.; Pigeau, B.; Besga, B.; Vincent, P.; Poncharal, P.; Arcizet, O. Nature Nanotechnology 2017, 12, 156. (11) Heritier, M.; Eichler, A.; Pan, Y.; Grob, U.; Shorubalko, I.; Krass, M. D.; Tao, Y.; Degen, C. L. Nano Letters 2018, 18, 1814 -- 1818. 14 (12) Mamin, H. J.; Rettner, C. T.; Sherwood, M. H.; Gao, L.; Rugar, D. Applied Physics Letters 2012, 100, 013102. (13) Nichol, J. M.; Hemesath, E. R.; Lauhon, L. J.; Budakian, R. Physical Review B 2012, 85, 054414. (14) Longenecker, J. G.; Mamin, H. J.; Senko, A. W.; Chen, L.; Rettner, C. T.; Rugar, D.; Marohn, J. A. ACS Nano 2012, 6, 9637 -- 9645. (15) Tao, Y.; Eichler, A.; Holzherr, T.; Degen, C. L. Nature Communications 2016, 7, 12714. (16) Wagenaar, J.; den Haan, A.; Donkersloot, R.; Marsman, F.; de Wit, M.; Bossoni, L.; Oosterkamp, T. Physical Review Applied 2017, 7, 024019. (17) Mamin, H. J.; Oosterkamp, T. H.; Poggio, M.; Degen, C. L.; Rettner, C. T.; Rugar, D. Nano Letters 2009, 9, 3020. (18) Rose, W.; Haas, H.; Chen, A. Q.; Jeon, N.; Lauhon, L. J.; Cory, D. G.; Budakian, R. Physical Review X 2018, 8, 011030. (19) Mamin, H. J.; Kim, M.; Sherwood, M. H.; Rettner, C. T.; Ohno, K.; Awschalom, D. D.; Rugar, D. Science 2013, 339, 557 -- 560. (20) Staudacher, T.; Shi, F.; Pezzagna, S.; Meijer, J.; Du, J.; Meriles, C. A.; Reinhard, F.; Wrachtrup, J. Science 2013, 339, 561 -- 563. (21) Moores, B. A.; Eichler, A.; Tao, Y.; Takahashi, H.; Navaretti, P.; Degen, C. L. Applied Physics Letters 2015, 106, 213101. (22) Poggio, M.; Degen, C. L.; Mamin, H. J.; Rugar, D. Physical Review Letters 2007, 99, 017201. 15 (23) Chao, S.; Dougherty, W. M.; Garbini, J. L.; Sidles, J. A. Review of Scientific Instru- ments 2004, 75, 1175. (24) Degen, C. L.; Poggio, M.; Mamin, H. J.; Rugar, D. Physical Review Letters 2007, 99, 250601. (25) Herzog, B. E.; Cadeddu, D.; Xue, F.; Peddibhotla, P.; Poggio, M. Applied Physics Letters 2014, 105, 043112. (26) Silver, M. S.; Joseph, R. I.; Hoult, D. I. Physical Review A 1985, 31, 2753 -- 2755. (27) Tannus, A.; Garwood, M. Journal of Magnetic Resonance. Series A 1996, 120, 133 -- 137. (28) Tomka, I. T.; van Beek, J. D.; Joss, R.; Meier, B. H. Physical Chemistry Chemical Physics 2013, 15, 3438 -- 3441. (29) Kupce, E.; Freeman, R. Journal of Magnetic Resonance. Series A 1995, 115, 273 -- 276. (30) Bendall, M. R.; Pegg, D. T. Journal of Magnetic Resonance 1986, 67, 376 -- 381. (31) Loretz, M.; Pezzagna, S.; Meijer, J.; Degen, C. L. Applied Physics Letters 2014, 104, 033102. (32) Overweg, H. C.; den Haan, A. M. J.; Eerkens, H. J.; Alkemade, P. F. A.; la Rooij, A. L.; Spreeuw, R. J. C.; Bossoni, L.; Oosterkamp, T. H. Applied Physics Letters 2015, 107, 072402. (33) Tsaturyan, Y.; Barg, A.; Polzik, E. S.; Schliesser, A. Nature Nanotechnology 2017, 12 . (34) Ghadimi, A. H.; Fedorov, S. A.; Engelsen, N. J.; Bereyhi, M. J.; Schilling, R.; Wil- son, D. J.; Kippenberg, T. J. Science 2018, 360, 764 -- 768. 16 (35) Usenko, O.; Vinante, A.; Wijts, G.; Oosterkamp, T. H. Applied Physics Letters 2011, 98, 133105. 17
1711.07180	1	1711	2017-11-20T07:47:20	Nano-scale characterization of the formation of silver layers during electroless deposition on polymeric surfaces	[ "physics.app-ph" ]	We report here a quantitative method of Transmission Electron Microscopy (TEM) to measure the shapes, sizes and volumes of nanoparticles which are responsible for their properties. Gold nanoparticles (Au NPs) acting as nucleating agents for the electroless deposition of silver NPs on SU-8 polymers were analyzed in this project. The atomic-number contrast (Z-contrast) imaging technique reveals the height and effective diameter of each Au NP and a volume distribution is obtained. Varying the reducing agents produced Au NPs of different sizes which were found both on the polymer surface and in some cases buried several nanometers below the surface. The morphology of Au NPs is an important factor for systems that use surface-bound nanoparticles as nucleation sites as in electroless metallization. Electrolessly deposited silver layers reduced by hydroquinone on SU-8 polymer are analyzed in this project.	physics.app-ph	physics	ICAMMP 2011 International Conference on Advances in Materials and Materials Processing, Indian Institute of Technology Kharagpur, 9-11 December, 2011 Nano-scale characterization of the formation of silver layers during electroless deposition on polymeric surfaces Aniruddha Dutta1, 2, , Biao Yuan5, Helge Heinrich1, 2, Christopher N. Grabill 3, Stephen M. Kuebler 3, 4, Aniket Bhattacharya1 1Department of Physics; 2Advanced Materials Processing and Analysis Center; 3Department of Chemistry; 4CREOL, The College of Optics and Photonics; and 5Department of Mechanical, Materials and Aerospace Engineering University of Central Florida, Orlando, FL 32816. Abstract [email protected] We report here a quantitative method of Transmission Electron Microscopy (TEM) to measure the shapes, sizes and volumes of nanoparticles which are responsible for their properties. Gold nanoparticles (Au NPs) acting as nucleating agents for the electroless deposition of silver NPs on SU-8 polymers were analyzed in this project. The atomic-number contrast (Z-contrast) imaging technique reveals the height and effective diameter of each Au NP and a volume distribution is obtained. Varying the reducing agents produced Au NPs of different sizes which were found both on the polymer surface and in some cases buried several nanometers below the surface. The morphology of Au NPs is an important factor for systems that use surface-bound nanoparticles as nucleation sites as in electroless metallization. Electrolessly deposited silver layers reduced by hydroquinone on SU-8 polymer are analyzed in this project. Keywords: Polymer, gold, silver, HAADF-STEM. 1. Introduction Determining the sizes and shapes of metal NPs is crucial in understanding the morphology associated with nonmaterial systems. Various reducing agents effect the formation of Au NPs at the surface of SU-8 cross-linked polymers which are increasingly employed for micro and nano- scale fabrication [1, 2]. In this work Au NPs have been formed on the SU-8 polymer surface by in-situ reduction. Quantitative TEM has been used for the evaluation of NP sizes, volumes, structures, and compositions, and to understand their distribution on the surface [3-7]. 2. Experimental Gold and silver NPs deposited on SU-8 polymer samples were provided by the group of Dr. Stephen Kuebler of the Chemistry Department at UCF. Au NPs were formed by binding Au cations to an aminated polymer surface immersed in an aqueous solution of 5.3 × 10-4 M HAuCl4.The ICAMMP 2011 Page 1 solution was then reduced with 0.1M NaBH4. All depositions were done in ambient temperature. Preparation of the Ag NPs was done according to the method introduced by Khalid et al. [8]; the reduction by citrate was accomplished by immersing samples for eight hours in aqueous 1% (w/v) sodium citrate. Reduction with hydroquinone was achieved by immersing samples for one hour in aqueous 0.1 M hydroquinone. Samples for plan-view TEM imaging were prepared by scraping a thin section from the surface of the polymer film. Cross sectional TEM samples were prepared by Focused Ion Beam (FIB) thinning. Each cross-sectional sample was coated with 30 to 50 nm carbon followed by 10 nm Au- Pd coating before the FIB cutting to increase the contrast for TEM micrographs. A Tecnai F30 TEM at 300 kV was used for plan-view and cross-sectional imaging. 3. Results and discussion 3.1. Characterization of Au NPs formed by NaBH4 and hydroquinone reduction. Fig1 (a) shows a plan-view TEM image of Au NPs deposited on the surface of a SU-8 polymer after reduction by NaBH4.The nanoparticles appear evenly spaced with no long-range ordering. Energy-Dispersive X-ray Spectroscopy(EDS) was performed on these small 2 to 5 nm diameter particles confirming the nanoparticles consisted of gold without any Pd. Electron diffraction proved that the particles have a face centered cubic(FCC) structure. Fig 1(b) represents a High- Angle Annular Dark-Field (HAADF) Scanning Transmission Electron Microscopy (STEM) plan- view image confirming the homogeneity of the nanoparticle distribution. Fig1 (a). Bright-field plan-view TEM micrograph of Au NPs generated on the polymer by reduction of gold ions bound at the surface by ethylenediamine (ED) then reduced with NaBH4. (b) HAADF- STEM plan view micrograph of NaBH4-generated Au NPs. The volumes, heights and the cross-sectional areas of individual NPs were obtained from intensity- calibrated HAADF-STEM micrographs. For a statistical analysis 150 NPs were considered. The particle size analysis reveals that the average volume of the Au NPs is 8 nm3 with a Relative Standard Deviation (RSD) of the volume distribution of 56%. The NPs are typically not spherical but rather oblate or prolate spheroids. Fig 2(a) shows the volume distribution of the Au NPs for a Page 2 ICAMMP 2011 plan-view sample. Most of the NPs have volume between 5-7 nm3 with around 23% NPs being very small with volumes below 5 nm3. % q e r F 35 30 25 20 15 10 5 0 0.0 2.5 5.0 7.5 Volume % q e r F 17.5 20.0 22.5 25.0 35 30 25 20 15 10 5 0 Height 0.5 1.0 1.5 2.0 2.5 3.0 Height(nm) 3.5 4.0 4.5 5.0 10.0 12.5 15.0 Volume(nm3) Fig 2(a). Volume distribution of 150 Au NPs. (b) Height distribution of Au NPs from calibrated HAADF- STEM micrographs. Likewise, the measured particle heights from integrated pixel intensities results in an average particle height of 1.8 nm and a RSD of 37%. The mean lateral radius was calculated to be 1.4 nm with a RSD of 13%. From the height distribution in Fig 2(b) we infer that almost half of the particles are less than 2 nm in height. Fig 3(a) Cross-sectional Bright Field (BF) TEM micrographs of Au NPs reduced by NaBH4. (b)Hydroquinone (HQ) reduced Au NPs on the SU-8 polymer surface. Fig 3(a) shows the BF image of Au NPs with 2-5 nm in diameter reduced by NaBH4 on the polymer surface. The Au NPs of 10 to20 nm in diameter in Fig. 3(b) are consistent with Au NPs reduced by HQ as reported in the literature [9]. Being a stronger reducing agent, with NaBH4 much smaller Au NPs are created compared to HQ which has a positive reduction potential compared to NaBH4[10-12]. ICAMMP 2011 Page 3 3.2. Characterization of Ag NPs formed by reduction of hydroquinone from silver citrate. - at surface bound amine Gold NPs are bound to the surface of SU-8 polymer by coordinating AuCl4 sites. Surface bound Au3+ is then reduced with NaBH4, causing Au NPs to form. Ag is deposited onto Au NPs via reduction of Ag+ by hydroquinone for 6 hours in presence of Gum Arabic. Fig 4 shows the chemical reaction of formation of Ag NPs by the above method [1, 13]. 2 O OH + O- Ag+ OH OH Ag2 + 2 O OH OH + O O Hydroquinone Quinone Fig 4. Reaction scheme for electroless silver deposition. Fig 5 shows a plan-view HAADF-STEM micrograph of Ag NPs formed on SU-8 polymer. The Au NPs act as nucleation sites for the silver grains. Comparison of the plan-view images in Fig. 1(b) and Fig. 5 shows that the distance of small Au NPs is smaller than the distance of centers of Ag grains. We can therefore conclude that not all Au NPs are nucleation sites for large Ag grains. The average thickness of the silver layer was found to be 45 nm with a standard deviation of 18 nm. The granular structure of silver grains accounts for the large standard deviation. Gum Arabic, a complex protein (stabilizer), has been used in this case in the solution to control the silver deposition. The average volume for the Ag grains was (65000±9000) nm3 with an average diameter as 45 nm. Fig 5. Plan-view HAADF-STEM of Ag NPs on a SU8 polymer prepared by 6 hours of electroless deposition using [Ag+] = 5.6 mM in the presence of gum Arabic [15]. ICAMMP 2011 Page 4 4. Conclusions The above work suggests that the size of Au NPs generated can be controlled with choice of reducing agents. NPs that are generated on the surface by HQ are much larger than those generated beneath the surface. This happens probably because the rapid reaction of NaBH4 with Au ions on the surface promotes diffusion resulting in growth rather than nucleation. The model also suggests that NPs can be prepared at different depths below the surface for several reducing agents. These NPs are useful for creating material systems with controlled optical enhancement that are not compromised by fluorescence quenching. It follows then that the role of the polymer matrix may be important in determining, or even controlling, the shape, size, and properties of the NPs generated. Controlling the rate of silver growth and shaping its 3D morphology will be the area of further studies. Acknowledgements This work was supported by NSF CAREER grant #0748712 and NSF-CHE grant #0809821.We acknowledge Dr. Aniket Bhattacharya and his group from University of Central Florida (UCF) Physics Department for modeling and simulations. We also thank all the staff members of Advanced Material Processing & Analysis Center (AMPAC) at UCF for providing support and help with handling of the instruments. References [1]. Tal A, Chen YS, Williams HE, Rumpf RC, Kuebler SM. Fabrication and characterization of three dimensional copper metallodielectric photonic crystals. Optics Express 2007; 1(26):18283. [2]. Yang R, Lu BR, Due J, Sheen ZK, CSU ZC, Hub E, Quad XP. Nanoimprint lithography for optic fluidics with phase gratings for environmental monitoring application. Microelectronic Engineering 2010; 87(5-8):824-826. [3]. Sanders JV. Transmission Electron Microscopy of catalyst. J Electron Microscopy Technique 1986; 3:67-93. [4]. Prestridge EB, Via GH, Sinfelt JH. Electron microscopy studies of metal clusters: Ru, Os, Ru- Cu and Ox-Cu. J Catalysis 1977.J Catalysis; 50:115-123. [5]. Menard LD et al. Sub-Nanometer Au Monolayer-Protected Clusters Exhibiting Molecule-like Electronic Behavior: Quantitative High-Angle Annular Dark-Field Scanning Transmission Electron Microscopy and Electrochemical Characterization of Clusters with Precise Atomic Stoichiometry. J. Physical Chemistry B 2006; 110:12874. [6]. Menard LD et al. Metal Core Bonding Motifs of Monodisperse Icosahedral Au13 and Larger Au Monolayer-Protected Clusters as Revealed by X-ray Absorption Spectroscopy and Transmission Electron Microscopy. J. Physical Chemistry B 2006; 110:14564-14573. [7]. Heinrich H et. al. Material Research Society Symposium 2009; 1184:119. ICAMMP 2011 Page 5 [8]. Khalid M, Pala I, Wasio N, Bandyopadhyay K. Functionalized surface as template for in situ generation of two-dimensional metal nanoparticle assembly. Colloids and Surfaces A: Physicochemical and Engineering Aspects 2009; 348(1-3):263-269. [9]. Zhong L, Jiao T, Liu M. Synthesis and assembly of gold nanoparticles in organized molecular films of geminiamphiphiles. Langmuir 2008; 24(20):11677-83. [10]. Shankar R, Shahi V, Sahoo U. Comparative Study of Linear Poly(alkylarylsilane)s as Reducing Agents toward Ag(I) and Pd(II) Ions−Synthesis of Polymer−Metal Nanocomposites with Variable Size Domains of Metal Nanoparticles. Chemistry of Materials 2010; 22(4):1367- 1375. [11]. Troupis A, Triantis T, Hiskia A, Papaconstantinou E. Rate-Redox-Controlled Size-Selective Synthesis of Silver Nanoparticles Using Polyoxometalates. European Journal of Inorganic Chemistry 2008; 36:5579-5586. [12]. Reetz MT, Maase M, Schilling T, Tesche B. Computer Image Processing of Transmission Electron Micrograph Pictures as a Fast and Reliable Tool to Analyze the Size of Nanoparticles. The Journal of Physical Chemistry B 2010; 104(37):8779-8781. [13]. Pérez MA, Moiraghi R, Coronado EA, Macagno VA. Hydroquinone Synthesis of Silver Nanoparticles: A Simple Model Reaction to Understand the Factors That Determine Their Nucleation and Growth. Crystal Growth and Design 2008; 8(4):1377-1383. [14]. Yuan B, Heinrich H, Yao B, Dutta A. Quantitative Measurment of Volumes for Nanoparticles by High Angle Annular Dark-Field Scanning Transmission Electron Microscopy. Microscopy and Microanalysis 2010; 16:1764 -1765. ICAMMP 2011 Page 6
1805.11207	1	1805	2018-05-29T01:10:53	Toward Ka Band Acoustics: Lithium Niobate Asymmetrical Mode Piezoelectric MEMS Resonators	[ "physics.app-ph", "eess.SP" ]	This work presents a new class of micro-electro-mechanical system (MEMS) resonators toward Ka band (26.5-40GHz) for fifth-generation (5G) wireless communication. Resonant frequencies of 21.4 and 29.9 GHz have been achieved using the fifth and seventh order asymmetric (A5 and A7) Lamb-wave modes in a suspended Z-cut lithium niobate (LiNbO3) thin film. The fabricated device has demonstrated an electromechanical coupling (kt2) of 1.5% and 0.94% and extracted mechanical Qs of 406 and 474 for A5 and A7 respectively. The quality factors are the highest reported for piezoelectric MEMS resonators operating at this frequency range. The demonstrated performance has shown the strong potential of LiNbO3 asymmetric mode devices to meet the front-end filtering requirements of 5G.	physics.app-ph	physics	Toward Ka Band Acoustics: Lithium Niobate Asymmetrical Mode Piezoelectric MEMS Resonators Yansong Yang, Ruochen Lu, Tomas Manzaneque, and Songbin Gong Department of Electrical and Computer Engineering University of Illinois at Urbana-Champaign Urbana-Champaign, USA [email protected] Figure 1: Displacement mode shapes of the (a) first-order (A1), (b) second- order (A2), (c) third-order (A3), (d) fourth-order (A4), (e) fifth-order (A5), (f) sixth-order (A6), and (g) seventh-order (A7) asymmetric modes. The ar- rows denote the displacement directions. The stress field for each mode or- der is plotted in black dashed line. FBAR. All these limitations lower the power handling and in- crease the thermal nonlinearity of the resonator [5]. On the other hand, overtone operations, despite providing better power han- dling via a thicker film, would lead to diminished electrome- chanical coupling (kt 2) as it will be discussed in this paper. To overcome these limitations, one alternative is to select an acoustic wave mode in a material that features much higher elec- tromechanical coupling than AlN FBARs so that frequency scal- ing toward Ka Band based on higher orders can still produce reasonable kt 2 to meet 5G bandwidth requirement. It is also pref- erable that the mode has a large acoustic velocity so that scaling would not reduce the feature size to the extent that limits power handling. Recently, several modes (S0, SH0, and A1) have been demonstrated in LiNbO3 with high coupling factors [6-10]. Among these wave modes, A1 Lamb wave mode has the highest phase velocity, and has been demonstrated with very large kt 2 and Qs simultaneously in Z-cut LiNbO3 (kt 2=28% and Q=500 at 5 GHz), and in Y-cut LiNbO3 (kt 2=6.3% and Q=5341 at 1.7 GHz) [9] [10]. Therefore, A1 mode devices are being considered as an alternative resonator technology for sub-6 GHz applica- tions. However, the higher order asymmetrical modes have not yet been explored for scaling the resonant frequency beyond 6 GHz. In this work, we aim to extend asymmetrical Lamb wave modes in Z-cut LiNbO3 to higher orders for beyond-24 GHz ap- plications. We first investigate the excitation of asymmetric modes of various orders. More specifically, the dependence of electromechanical coupling factor on mode order and resonant Abstract- This work presents a new class of micro-electro-me- chanical system (MEMS) resonators toward Ka band (26.5-40 GHz) for fifth-generation (5G) wireless communication. Resonant frequencies of 21.4 and 29.9 GHz have been achieved using the fifth and seventh order asymmetric (A5 and A7) Lamb-wave modes in a suspended Z-cut lithium niobate (LiNbO3) thin film. The fabricated device has demonstrated an electromechanical cou- pling (kt2) of 1.5% and 0.94% and extracted mechanical Qs of 406 and 474 for A5 and A7 respectively. The quality factors are the highest reported for piezoelectric MEMS resonators operating at this frequency range. The demonstrated performance has shown the strong potential of LiNbO3 asymmetric mode devices to meet the front-end filtering requirements of 5G. Keywords-Ka Band; 5G wireless communications; Internet of Things; lithium niobate; asymmetrical modes; MEMS resonators I. INTRODUCTION Recently, due to the market demand for higher data rate for cellular phone applications, below 3 GHz spectrum has become increasingly crowded with little available spectrum for the ex- pansion of fifth-generation (5G) wireless communication sys- tems. As a result, Federal Communications Commission (FCC) has opened frequency bands 27.5-28.35 GHz and 37-40 GHz for licensed use, which overlaps with Ka band, for 5G mobile radio applications [1]. Several European and Asian countries have also planned portions of Ka band for the same purpose [2]. This expansion has sparked new technology development to over- come the remaining bottlenecks in accessing this portion of the spectrum. One key missing piece in beyond-24 GHz front-ends is high-performance and miniature filters that can be arrayed in a smart phone. At fourth-generation (4G) frequency ranges (<3 GHz), filters based on acoustic and MEMS resonators have been widely adopted. However, it remains challenging to scale these devices in frequency to meet 5G requirements. The state-of-the-art acoustic resonators beyond 24 GHz re- sort to either the fundamental FBAR mode in a 100 nm thin AlN film or an overtone FBAR mode in a thicker AlN film [3] [4]. In either case, significant performance compromise has to be made in the process of scaling, which might lead to inadequate perfor- mance for 5G applications. For instance, a suspended 100 nm AlN thin film would produce a high thermal resistance to the surrounding. In addition, matching to system resistance (50 Ω) beyond 24 GHz requires small static capacitance (C0<150 fF), which limit the size (length∙width < 200 μm2) of 100 nm thick Figure 2: Distribution of electrical and stress fields of A1, A2, A3, and A4 modes. frequency is studied with both analytical and finite element methods. Based on these studies, a film thickness of 400 nm is chosen to support the fifth and seventh order asymmetric (A5 and A7) modes to move up and reach Ka band. The fabricated device is measured with kt 2 of 1.5% and 0.94% and extracted mechanical Qs of 406 and 474, at 21.4 GHz and 29.9 GHz for A5 and A7 modes, respectively. II. ASYMMETRIC LAMB WAVE MODES IN LITHIUM NIOBATE THIN FILM Asymmetric modes are a class of Lamb-wave modes charac- terized by their particular anti-symmetry about the median plane of the plate. In other words, they have equal vertical displace- ment components but opposite longitudinal components on dif- ferent sides of the median plane [11]. To visualize the displace- ment mode shapes of asymmetric modes of various orders, Com- sol finite element analysis (FEA) is used to simulate the eigen modes in a 2D LiNbO3 slab with free top and bottom surfaces and periodic boundaries in the lateral direction (Fig. 1). Various order modes from the first (A1) to seventh (A7) are shown with a mode order denoting the number of half-wavelength periodic- ities in the vertical direction. Clearly, overmoding a slab with a fixed thickness would yield a higher resonant frequency, pro- vided the intended higher order mode can somehow be excited in the LiNbO3 slab with transducers. The simplest transducers for such a purpose are interdigital electrodes that are patterned exclusively on top of a transferred LiNbO3 thin film, as they have least fabrication complication. As demonstrated by several prior works on A-modes, top-only in- terdigital transducers (IDT) can effectively excite the A1 modes with lateral electric fields [9] [10]. However, electromechanical coupling to asymmetric modes with mode orders higher than A1 has not been well studied with top-only IDTs. To understand the excitation of higher-order A-modes with IDT, magnitudes of the stress standing waves for different order modes are depicted in Fig. 1 and 2. As seen in Fig. 2, we simplify the E-field introduced by IDTs as laterally polarized with a uniform magnitude across the thickness of the plate. Thus, the integration of stress and the lateral electrical fields, which describes the mutual energy be- tween the electrical and mechanical domains [12], vanishes for even-order modes and leads to zero electromechanical coupling to these modes (according to Berlincourt's definition of kt 2). On the other hand, odd order modes can be excited due to non-zero integral of mutual energy. For the purpose of frequency scaling to Ka band, our overmoding approach will focus on the odd or- der modes with the simple top-only IDTs. Figure 3: (a) Calculated and simulated f×t product and (b) Simulated effective electromechanical coupling factor of A1, A3, A5, and A7 modes vs. the ratios of film thickness to longitudinal wavelength (h/λL). To more precisely predict the resonant frequency and elec- tromechanical coupling of a higher odd order A-mode, we have to resort a more refined model that treats the cross-section of the resonator as a two-dimensional cavity, instead of an infinitely long slab seen in Fig. 1 [11]. The resonant frequency of an odd order mode in a two-dimensional cavity with thickness of t and where m and n are the mode orders in the vertical (z-axis) and ing an odd value larger than 1 can emerge as a spurious mode acoustic phase velocities in the vertical and longitudinal direc- tions. The asymmetric modes of interest in this work have a lon- gitudinal mode order n of 1 with a vertical mode order m that takes a value among 1, 3, 5 and 7. Note that Eq. 1 assumes the lateral boundaries are mechanically free, the same as the top and bottom surfaces. This assumption will be revisited later. Eq. 1 length of l is given by (cid:1858)(cid:2868)(cid:3040)(cid:3041)=(cid:3495)((cid:3040)(cid:3049)(cid:3295)(cid:2870)(cid:3047))(cid:2870)+((cid:3041)(cid:3049)(cid:3261)(cid:2870)(cid:3039))(cid:2870) (1) longitudinal (x-axis) directions, respectively. (cid:1874)(cid:3047) and (cid:1874)(cid:3013) are the also implies that any composite mode of order (cid:1858)(cid:2868)(cid:3040)(cid:3041) with n tak- near the intended mode (cid:1858)(cid:2868)(cid:3040)(cid:2869). Such a phenomenon has been ob- (cid:1858)(cid:2868)(cid:3040)(cid:2869)=(cid:3049)(cid:3261)(cid:2870)(cid:3047)(cid:3495)((cid:2009)(cid:1865))(cid:2870)+(2 (cid:3047)(cid:3090)(cid:3261))(cid:2870) (2) (cid:2009)=(cid:3493)(cid:1855)(cid:2873)(cid:2873)/(cid:1855)(cid:2869)(cid:2869) (3) the simulated and calculated frequency-thickness products ((cid:1858)× (cid:1872)) as a function of the ratio of plate thickness to longitudinal To validate the two-dimensional analysis, Comsol-based FEA is used to calculate the frequency variation of each odd-order asymmetrical mode in a Z-cut LiNbO3 thin film. Fig. 3 (a) shows served for other high coupling resonators with a 2D nature [13] [14]. For modes with n=1, l is equal to half of the longitudinal wavelength (λL/2). Thus, Eq. 1 can be re-written as: where α is the ratio between the velocity of vertical and longi- tudinal directions: wavelength (t/λL). The discrepancies between simulation and calculation are likely caused by several simplifications used in the calculation. Figure 5: SEM image of the fabricated device. Figure 6: Multi-resonance equivalent MBVD circuit model with an added series inductor. tanδ is the loss tangent of LiNbO3, ρ is the resistivity of gold, and μ is the permeability of Au. The relationship, shown in Fig. 3(b), indicates that overmoding into A3, A5, and A7 can still have meaningful keff 2 for 5G appli- cations if it starts with a large keff 2 of 30%, which is attainable for A1 mode with a t/λL<0.1. Based on the above analyses (summarized in Fig. 3), we choose the thickness of the thin film to be 400 nm, and λL to be 6 μm so that the targeted A5 and A7 modes can be scaled toward 2. t of 400 nm and λL of 6 μm Ka band with reasonably large keff are also deemed as a good tradeoff between having an ade- quately low t/ λL and achieving a reasonably large static capaci- tance (C0) for the resonator. Based on Fig. 3 (a), the A7 mode in such a cavity would yield a resonant frequency at 30 GHz. III. DESIGN AND MODELING OF ASYMMETRIC MODE RESONATOR To validate resonant characteristics and factor in the effects of electrodes, the resonator is simulated with 2D Comsol FEA. As shown in the cross-sectional mock-up in Fig. 4 (a), the device consists of a 2-electrode transducer on top of a mechanically sus- pended 400 nm thick Z-cut LiNbO3 thin film. The two elec- trodes, connected to signal and ground respectively, induce lat- eral electric fields in the LiNbO3 thin film, which subsequently excite the resonator into odd-order asymmetric mode vibration. 60 nm gold is used as the electrodes in the simulation. Gold is used for its good conductivity while thickness is kept low to avoid excess mass loading and shifting down the resonances. The design parameters are summarized in Table I. The Comsol simulated response is shown in Fig. 4 (b), including the displace- ment mode shape of each excited asymmetric mode. Different from fundamental Lamb waves (S0, A0), the higher order A- modes are confined between two electrodes, so the longitudinal cavity dimension is approximately the distance between adja- cent electrodes. We already considered this effect in the design process by setting the distance between the electrodes to 3 μm, Figure 4: (a) Mocked-up view of designed asymmetric mode LiNbO3 resona- tor. (b) Simulated response of a Z-cut LiNbO3 A-mode resonator. Displacement mode shapes of (i)A1, (ii)A3, (iii)A5, and (iv)A7 are included. Physical dimensions of the designed asymmetric mode resonator Electrode width, (cid:1849)(cid:3032) 3 μm Device width, W 13 μm Spacing be- tween elec- trodes, g Total length, L 3 µm 100 μm TABLE I 60 nm Electrode thickness, Film thick- ness, t 400 nm (cid:1872)(cid:3032) (cid:1863)(cid:3032)(cid:3033)(cid:3033)(cid:2870) ≈ (cid:3032)(cid:3118)(cid:3084)(cid:3268)(cid:3030)(cid:3254)∙(cid:1863)(cid:3049)_(cid:3040)(cid:2870) 2) [15]: Berlincourt Formula, is given as: on the stress field distribution of nth order mode in the lon- , as part of the In addition to the resonant frequency, each mode is also char- acterized by an effective electromechanical coupling factor (keff where e is the piezoelectric efficient, εS is the permittivity under constant strain, cE is the stiffness under constant electric field. on the stress and electric field distributions of the mth order mode in the ∙(cid:1863)(cid:3039)_(cid:3041)(cid:2870) (4) is a scaling factor capturing the dependence of (cid:1863)(cid:3032)(cid:3033)(cid:3033)(cid:2870) (cid:1863)(cid:3049)_(cid:3040)(cid:2870) vertical direction. Similarly, (cid:1863)(cid:3039)_(cid:3041)(cid:2870) represents the dependence of (cid:1863)(cid:3032)(cid:3033)(cid:3033)(cid:2870) gitudinal direction [11]. The expression of (cid:1863)(cid:3049)_(cid:3040)(cid:2870) (cid:1863)(cid:3049)_(cid:3040)(cid:2870) = ((cid:1516)(cid:3006)(cid:3299)((cid:3053))(cid:3048)(cid:3299)((cid:3053))(cid:3031)(cid:3053))(cid:3118) (cid:1516)(cid:3006)(cid:3299)(cid:3118)((cid:3053))(cid:3031)(cid:3053)∙(cid:1516)(cid:3048)(cid:3299)(cid:3118)((cid:3053))(cid:3031)(cid:3053) (5) integration, (cid:1863)(cid:3049)_(cid:3040)(cid:2870) (cid:1863)(cid:3049)_(cid:3040)(cid:2870) = (cid:2869)(cid:3040)(cid:3118) (6) (cid:1863)(cid:3039)_(cid:3041)(cid:2870) in Eq. 4 is dependent to the ratio between the vertical and (also the dispersive relationship between (cid:1863)(cid:3039)_(cid:3041)(cid:2870) and t/λL), on (cid:1863)(cid:3032)(cid:3033)(cid:3033)(cid:2870) longitudinal dimensions (t/λL), and typically increases with re- 2 of the mth-order spect to m. Therefore, the lower bound for keff (m >1) asymmetric mode is 1/m2 of keff 2 of the A1 mode. By using the simplified field distributions shown in Fig. 2 for can be formulated as a function of mode order To further understand the diminishing effect of overmoding Comsol FEA is used to calculate the keff 2 as a function of t/λL. m: Order (m) f0 1 3 5 7 4.5 GHz 12.9 GHz 21.4 GHz 29.9 GHz Rmi 74 Ω 78 Ω 82 Ω 76 Ω Cmi Lmi 3.84 fF 0.574 fF 0.225 fF 0.15 fF 319 nH 267 nH 247.3 nH 192 nH 12.5 Ω 12.9 Ω 13.2 Ω 13.4 Ω TABLE II Rs((cid:2033)) C0 19.4 fF Cf 13.5 fF Ls 90 pH Key measured values (in bold symbols) and extracted parameters of the multi-resonance MBVD model (cid:1872)(cid:1853)(cid:1866)(cid:2012) Qo 0.07 118 224 287 328 2 kt 24% 3.7 % 1.5 % 0.94% FoM 28.3 8.3 4.3 3.1 Extracted Mechanical Qm 122 277 406 474 IV. MEASUREMENTS AND DISCUSSIONS To validate the analytical and modeling results, the designed Z-cut LiNbO3 A-mode resonator was fabricated on a 400 nm- transferred Z-cut LiNbO3 thin film following the process de- scribed in [9]. The fabricated device is shown in Fig. 5. 60 nm- Au is sputtered and lift-off as the top electrodes. The probing pads, also made of Au, are electroplated to 2.5 μm to reduce loss. To reduce the parasitic effect between the lead lines, a 200 μm pitch between probing pads is used. The fabricated device was characterized over a wide frequency range (up to 40 GHz) with a Keysight N5230A PNA-L network analyzer in dry air and at room temperature. A TRL calibration based on on-wafer stand- ards is used to move the measurement reference planes to the positions as labeled in Fig. 5. Similar to the simulation results in Fig. 4, the measurement results shown in Fig.7(a) exhibit resonant frequencies at 4.5, 12.8, 21.4, and 29.9 GHz, corresponding to the anticipated A1, A3, A5, and A7 modes respectively. As expected, even-order asymmetric modes cannot be observed with this design. Among the first four odd-order asymmetric modes shown in Fig. 7(b-d), a higher order yields a higher measured overall quality factor, Qo (including the effects of all loss mechanisms). The reason be- hind this phenomenon is still under investigation. A multi-resonance modified Butterworth-Van Dyke (MBVD) model, in which each resonance is captured by a mo- tional branch of Rm, Lm, and Cm, is used to interpret the measure- ment results. As shown in Fig. 6, an additional series inductor (Ls) is added to account for the non-negligible inductance of the electrode fingers at high frequencies. An additional resistor (Rs) is added to account for surface resistance of electrodes and bus lines, the value of which is dependent on frequency due to skin effects and calculated based on the equation shown in Fig. 6. To account for the parasitic effects, Cf and Rf are included as the feedthrough capacitance and loss in the substrate, respectively. An on-chip test structure consisting of only the bus lines was included in fabrication to measure the value of Cf. Except for the overall Qo, resonances, and Cf, other parameters in the multi-res- onance MBVD model are extracted from measurements. By ex- cluding the loss in the electrical domain, we can also extract the mechanical Qm at these resonances from the MBVD model. It is worth noting that the A7 mode demonstrates a very high me- chanical Q of 474 at 30 GHz. Qm for other resonances, along with extracted kt 2 of each mode and other key parameters of the MBVD models, are listed in Table II. As seen in Fig. 7, the re- sponse of the MBVD model excellently fits with measured ad- mittance response, hence validating the parameter extraction. The effects of self-inductance and feedthrough capacitance have 2 values. The extracted kt been de-embeded for the extracted kt 2 values match the simulated results shown in Fig. 4, as well as Figure 7: (a) Measured and multi-resonant equivalent MBVD circuit modeled response of an asymmetric mode LiNbO3 resonator with zoomed-in view of (b) A1, (c) A3, (d) A5, and (e) A7 mode resonances. half of the target λL value. This local confinement of displace- ment for A3, A5 and A7 modes is believed to be caused by the acoustic impedance mismatch between the metalized and un- metalized sections, which increases with respect to the mode or- der. Strictly speaking, Eq. 2 can no longer apply to calculating the resonances as such confinement does not have mechanical free boundaries assumed in the derivation of Eq. 2. Instead, com- plex reflection coefficients should be used to capture longitudi- nal boundaries. However, the tight confinement as seen in Fig. 4 leads to the belief that Eq. 2 could be still used for estimates of resonances with good accuracy. Consistent with our theoretical analyses in Section II, FEA results show that only the odd-order modes are excited. The first four odd modes, A1, A3, A5 and A7, have resonant frequencies at 4.5, 12.8, 22, and 30 GHz respectively. The simulated effec- tive electromechanical coupling factors for these modes are 24%, 4.4%, 1.8% and 1%, showing the theoretically predicted diminishing trend. It scales slightly better than 1/m2 due to the aforementioned dispersive relationship between (cid:1863)(cid:3039)_(cid:3041)(cid:2870) and mode order, the simulated results on resonator kt the predictions based on Eq. 4 and in Fig. 3(b). 2 are also on par with the analytical results in Fig. 3(b). The agreement thus confirms the effectiveness of our frequency scaling approach based on overmoding asymmetric modes. The measured results herein have demonstrated the potential of higher-order A-modes for enabling Ka band acoustic resona- tors. A mechanical Q of 474, as demonstrated for A7, has al- ready exceeded the state of the FBAR results [3]. Further Q en- hancement has to reply on more sophisticated understanding the acoustic loss at this frequency range, as well as reducing dissi- pation in the electrical domain. To achieve a large kt 2 and higher filter bandwidth at these frequencies, a thinner LiNbO3 film may be adopted with our approach to scale the A5 mode to Ka band. This, however, will likely comes at the cost of lower power han- dling and linearity. V. CONCLUSION In this work, extending asymmetric modes in a LiNbO3 thin film to higher orders is first theoretically studied for the purpose of scaling asymmetric mode resonances toward Ka band. Sub- sequently, a LiNbO3 MEMS resonator with various resonances, the highest at 29.9 GHz, has been demonstrated for the first time. The developed MEMS resonator exhibits a kt 2 of 0.94%, a meas- ured overall Qo of 328, and an extracted mechanical Q of 474 at 29.9 GHz. Further development and optimization of LiNbO3 asymmetric resonators could lead to miniature Ka-band acoustic front-end filter technology that is currently unavailable for 5G. REFERENCES [1] Federal Communications Commission, "Spectrum frontiers report and order and further notice of proposed rulemaking: FCC16-89," Jul. 2016. [Online]. Available: https://apps.fcc.gov/edocs public/attachmatch/FCC- 16- 89A1 Rcd.pdf. [2] HUAWEI, "5G Spectrum Public Policy Position," 2018. [3] M. Hara, et al., "Super-High-Frequency band filters configured with Air- Gap-Type Thin-Film Bulk Acoustic Resonators," Jpn. J. Appl. Phys., Part 1 49, 07HD13 (2010). [4] S. Ballandras, et al., "High overtone Bulk Acoustic Resonators: application to resonators, filters and sensors," Acoustics Nantes, France, 2012, pp.23-27. [5] R. Lu and S. Gong, "Study of Thermal Nonlinearity in Lithium Niobate Based MEMS Resonators," Solid-State Sensors, Actuators Microsystems Conf. (TRANSDUCERS), 2015 18th Int., Jun. 2015. [6] S. Gong and G. Piazza, "Figure-of-merit enhancement for laterally vibrating lithium niobate MEMS resonators," IEEE Trans. Electron Devices, vol. 60, no. 11, pp. 3888-3894, Nov 2013. [7] R. H. Olsson III, K. Hattar, M. S. Baker, M. Wiwi, J. Nguyen, C. Padilla, S. J. Homeijer, J. R. Wendt, T. A. Friedmann, "Lamb wave micromechanical resonators formed in thin plates of lithium niobate", Hilton head solid-state sensors, actuators and microsystems workshop, 2014, pp 281–284 [8] M. Kadota, T. Ogami, K. Yamanoto, H. Tochishita and Y. Negora, "High- Frequency Lamb Wave Device Composed of MEMS Structure Using LiNbO3 Thin Film and Air Gap," IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 57, no. 11, NOV 2010. [9] Y. Yang, A. Gao, R. Lu, and S. Gong, "5 GHz lithium niobate MEMS resonators with high FoM of 153," 2017 IEEE 30th Int. Conf. on MEMS, Las Vegas, NV, 2017, pp. 942-945. niobate MEMS resonators with FoM of 336 and f∙Q of 9.15×1012," in [10] Y. Yang, R. Lu, T. Manzaneque, and S. Gong, "1.7 GHz Y-cut Lithium Microwave Symposim (IMS), 2018 IEEE MTT-S International, 2018, pp. 1-4. [11] D. Royer and E. Dieulesaint (1996) Elastic waves in Solids I free and guided propagation. Springer, Paris. [12] R. Lu, T. Manzaneque, Y. Yang, A. Kourani, and S. Gong, "Lithium niobate lateral overtone resonators for low power frequency-hopping applications," 2018 IEEE 31th Int. Conf. on MEMS, Belfast, Northern Ireland, UK, 2018, pp. 751-754. [13] C. Cassella, G. Chen, Z. Qian, G. Hummel, and M. Rinaldi, "Aluminum nitride cross-sectional Lamé mode resonators," J. Microelectromech. Syst., vol. 25, no. 2, pp. 275–285, Apr. 2015. [14] Y.-H. Song, R. Lu, and S. Gong, "Analysis and removal of spurious response in SH0 lithium niobate MEMS resonators," IEEE Trans. Electron Devices, vol. 63, no. 5, pp. 2066–2073, May 2016. [15] Hashimoto K (2009) RF bulk acoustic wave filters for communications. Artech House, Norwood, MA.
1811.08621	2	1811	2018-12-03T10:56:58	Super Damping of Mechanical Vibrations	[ "physics.app-ph" ]	We report the phenomenon of coherent super decay, where a linear sum of several damped oscillators can collectively decay much faster than the individual ones in the first stage, followed by stagnating ones after more than 90 percent of the energy has already been dissipated. The parameters of the damped oscillators for CSD are determined by the process of response function decomposition, which is to use several slow decay response functions to approximate the response function of a fast decay reference resonator. Evidence established in experiments and in finite element simulations not only strongly supported the numerical investigations, but also uncovered an unexplored region of the tuned mass damper parameter space where TMDs with total mass less than 0.2 percent of a primary free body can damp its first resonance up to a damping ratio of 4.6 percent. Our findings also shed light onto the intriguing underline connections between complex functions with different singular points.	physics.app-ph	physics	Super Damping of Mechanical Vibrations Ka Yan Au Yeung, Brian Yang, Liang Sun, Kehang Bai, Z. Yang* Department of Physics, The Hong Kong University of Science and Technology, Clearwater Bay, Kowloon, Hong Kong, The People's Republic of China Abstract We report the phenomenon of coherent super decay (CSD), where a linear sum of several damped oscillators can collectively decay much faster than the individual ones in the first stage, followed by stagnating ones after more than 95 % of the energy has already been dissipated. The parameters of the damped oscillators for CSD are determined by the process of response function decomposition, which is to use several slow decay response functions to approximate the response function of a fast decay reference resonator. Evidence established in experiments and in finite element simulations not only strongly supported the numerical investigations, but also uncovered an unexplored region of the tuned mass damper (TMD) parameter space where TMD's with total mass less than 0.2 % of a stainless steel plate can damp its first resonance up to a damping ratio of 4.6 %. Our findings also shed light onto the intriguing underline connections between complex functions with different singular points. Introduction Damped vibrations are common phenomena in many physical systems, including merging black holes [1] and neutron stars [2]. The classical vibrations can be described by a simple expression in time domain and in frequency domain [3]. Mitigating unwanted vibration in machinery, buildings, bridges, airplane, satellites, etc. is still a great challenge in engineering science despite of worldwide intensive efforts over the last century [4, 5]. The distribution of vibration energy is mostly within the first few resonant modes of the structures. Since its invention nearly 100 years ago [6], the toned mass dampers (TMD) or dynamic vibration absorbers (neutralizers) are particularly effective in suppressing vibrations with discrete frequencies, because their maximum effect is within a narrow band width around their individual working frequency, which can be designed and tuned to suit particular application needs. As a result, they have been widely used to specifically target at the first few resonant modes in many structures [7 -- 33]. For comprehensive summary and reviews, see [7 -- 10]. A TMD can be generically described by a damped harmonic oscillator [6]. The first theoretical investigation of the optimum TMD design was carried out in 1928 [11, 12]. Optimizations with generic TMD's soon followed and are still being pursued today [11 -- 29]. Most TMD's are made of cantilevers or mass-spring that each weighs about 100 g or more for working frequency below 100 Hz [30 -- 32]. The optimizations have been limited to the parameter space where the total mass of the TMD's was a few percent of the primary structures, even though in some cases the TMD mass was up to 35 % of the primary structures [27, 28]. Damping the vibration in free bodies, such as satellites and airplanes, are even more challenging. While part of the vibration energy of a 1 supported structure can be dissipated by the supporting bodies, i. e., the vibration energy of a building or bridge can partially be transmitted to the ground, or the vibration of a floor can be partially transmitted to the building, the vibration energy of a free body can only be dissipated internally. As a result, active control is the preferred choice for free bodies [33 -- 42]. In this paper we report solid experimental evidence and theoretical analysis of super damping of the first resonant mode at 100 Hz of a nearly free elastic body (a stainless steel plate about 6.4 kg in mass) brought by effective TMD's with total mass less than 0.2 % of the plate. Under an impulse excitation, the decay of the first mode of the plate was nearly complete within a time scale of 100 ms, or about 10 periods of the oscillation. As the elastic body had negligible internal loss (Q > 3000), almost all the impulse energy was dissipated by the TMD's within 10 periods. This shows that all the previous studies have missed an important parameter space of TMD's with mass about 50 times lighter than the conventional ones. Such phenomenon is explained with the concept of coherent super decay. A design strategy of damper parameters based on response function decomposition is developed to achieve super decay. Basics of Coherent Super Decay Consider a group of five damped oscillators with combined time dependent displacement in the form (1), where , and . As a typical example for illustration, we choose Qn = 20, . The other parameters are {An} = {0.0515, 0.1516, 0.358, 0.229, 0.0792}, { } = { -- 0.09, -- 0.0389, -- 0.0048, 0.0267, 0.0792}. The decay curve of a representative oscillator, which is the n = 3 term in Eq. (1), is shown in Fig. 1(a) as the purple curve. Its decay time constant (DTC) is almost the same as an oscillator with angular frequency 1 Hz and quality factor (Q-factor) Q = 20. As the Q-factors of the other four oscillators are also equal to 20, and their resonant frequencies deviate from the central frequency of 1 Hz by at most 9 %, the DTC of each individual oscillator deviates from each other by at most 10 %. Intuitively, one would expect the DTC of the combined displacement given in Eq. (1) to be about the same as the n = 3 term, i. e., it decays like an oscillator with and Q = 20. After all, if all , the combined decay would almost exactly follow the purple curve. The actual curve obtained by simple numerical computation using Eq. (1) and the parameters given above, however, is the green solid curve (middle) in Fig. 1(a), which decays almost twice as fast as any of the individual oscillators. In fact, the decay curve resembles that of an oscillator with Hz and Q = 10, which is depicted as the red 2 51()cos()IntRnnnxtAet2InnnQ211RnnnQ1nnn010n01(bottom) curve in Fig. 1(a). In other words, when the parameters {An} and { } are properly selected, the combined decay of several oscillators can be much faster than each individual ones. We refer to such phenomenon as the coherent super decay (CSD) of harmonic oscillators. Figure 1, (a) Displacement as a function of time of several damped oscillators obtained by numerical computations. (b) The reference response function (red disks) and the approximation response function (solid green curve) made by the response functions of five individual oscillators (green dashed curve). The CSD is caused by the coherent interference among the individual oscillators. To see in Eq. (1) this more clearly, the time dependent displacement without the damping factors was also calculated and shown as the green dashed curve in Fig. 1(a). It can be seen that the first minimum of the displacement by interference alone occurs at almost the same time when the combined decay (green solid curve) is near completion. At this moment, the phases of the individual oscillators are mutually destructive. Without damping, however, the displacement recovers its strength when the phases become mutually enhancing again as time goes by. With damping, sufficient vibration strength of each oscillator has already been consumed and the resurgence is negligible. We now show an approach to determine the parameters that can realize CSD. According to classical mechanics [3], the time dependent displacement given in Eq. (1) is resulted from the frequency response function given below, (2). This is a typical frequency response function of a vibrating system with 5 degrees of freedom being excited. The time dependent displacement in response to a unit impulse is given by (3). 3 nInte5221()/nnnnnAXiQ()t1()Re()2itxtXedThe integral can be calculated using standard path integral and residue theorem in complex analysis [41]. As the CSD behaves like a reference oscillator with resonant frequency = 1 and Q0 = 10, we plot the imaginary part of the response function of the reference oscillator in the form in Fig. 1(b) as the red curve. The real part is not shown due to space limit. The green points are the imaginary part of the response function in Eq. (2) using the CSD parameters. It is no coincidence that the two curves are nearly the same, because the amplitudes {An} and the resonant frequencies { } were deliberately chosen to approximate, or mimic, the reference oscillator response function with Q0 = 10. The green dashed curves are the response functions of the five mimicking resonators plotted individually. What has been shown here is what we refer to as response function decomposition (RFD), that is, if a response function with large line width can be mimicked by the sum of a number of response functions with much smaller line width, then the time dependent displacement generated by the two response functions will be almost the same, even though each mimicking oscillator decays much slower. That is the underline strategy to identify the amplitude and the central frequency parameters for the response functions with slow individual decay to collectively generate a much faster decay. Figure 2, (a) The computed time dependent displacement of case-B (green curve) together with the reference (red curve). The insert depicts the response function of case-B (green curve) and that of the reference (red curve). (b) The corresponding quantities for case-C. (c) The corresponding quantities for case-D. In the above example, we decompose one reference (mother) resonator of Q0 = 10 into five Q = 20 daughter resonators (Case-A). The same can be done to decompose each daughter resonator into five granddaughters with Q = 40, so as to mimic the mother resonator with 25 granddaughter resonators, each with four times the Q-factor of the mother. The resulting time dependent displacement curve is shown in Fig. 2(a), together with the response function in the insert (Case-B). The corresponding results by 125 great granddaughter response functions with Q = 80 (Case-C) are shown in Fig. 2(b). In Case-C we deliberately kept the mimicking curve relatively rough while keeping the average line shape nearly tracing the original mother resonator. Despite the rough mimicking, the time dependent displacement curve of the 125 great granddaughters is still very close to that of the mother resonator. 4 0220001()/XiQnIf the reference oscillator response function is poorly mimicked (Case-D), as is shown in the insert of Fig. 2(c), the resulting decay will resurge after initial fast decay, as is shown in the main graph, and the decay lasts for quite a long time. For more in-depth examination of the decay behavior, we calculate the quantity that is proportional to the remaining total energy of a vibrating elastic body, given by (4). The curve by the mother (reference) resonator is shown in Fig. 3 as the red curve, which takes the simple form of . We refer to such decay as a uniform decay, in that the DTC remains the same throughout the decay process. The corresponding curve by the five daughters (Case-A) is shown as the green curve in the same figure, together with the one made by the 25 granddaughters (Case-B, blue curve) and the 125 great granddaughter resonators (Case-C) as the purple curve. All of them exhibit similar decay in the first stage as the reference, and staged decays with different DTC's in different stages afterwards. This is the consequence that their mimicking of the uniform decay reference resonator is only approximate. Figure 3, The remaining vibration intensity as a function of time for the cases studied above, together with that of the reference (red line). For Case-A, the decay curve follows closely that of the reference resonator up to t = 38 s (6.05 periods for an angular frequency of 1 Hz), where E = 2.56 %, in the first stage decay. The quick drop levels off at around t = 81 s, after which the system is in the stagnating second decay stage. However, by then only 0.12% of the original vibration energy still remains, so the stagnation will not have any practical implications in vibration damping of a primary structure. The reference resonator will take 69 s to reach such energy reduction. Therefore, for all practical purposes in engineering Case-A can be regarded as decaying as fast as the reference. 5 2()(')'tEtxtdt0.1teThe decay curve of Case-B (blue curve) matches well the reference up to t = 69 s, where E = 0.1 %. The curve for Case-C (purple curve) decays even slightly faster than the reference up to t = 38 s, where E = 2.5 %. This is consistent with its response function being a little bit wider in combined line width than the reference. It levels off at t = 69 s, with E = 1.0 %. Case-D with poorly mimicking response function is also interesting. Its apparent line width is almost twice of that of the reference. So the question is at what time it will level off after the first stage fast decay. The result, shown as the brawn dashed curve in Fig. 3, indicates that it takes only t = 25 s (4 periods) to bring E down to 1.7 %, as compared to 44 s for the reference. At the time of leveling off, only 2 % of the initial energy is still present. Therefore, in practical applications, the decay can already be regarded as complete, and the damper parameters chosen to create such response function are really good for practical optimization purposes. Experimental Verifications The CSD realized by following the RFD strategy can lead to the designs and realization of fast vibration decay of primary structures with dozens times lighter multiple TMD's than the ones reported in the literature [4 -- 29], as will be shown in the experiments below. The primary structure for the damping experiments is a 40 cm × 40 cm × 5 mm stainless steel plate (6.24 kg in mass) with free edges. Its first eigenmode is at 100.24 Hz with a Q-factor over 3000 (damping ratio  = 0.5/Q = 1.7 × 10-4). The plate was hanging by soft threads through the two clear holes near its top edge at about 10 cm apart. For frequency response measurements, the plate at the position 7 cm from the lower right corner along the plate diagonal line was attached to a shaker similar to the one reported earlier [42]. The response displacement was measured by a miniature accelerometer at the corresponding position relative to the left lower corner. Single frequency excitation was used, and the excitation and the response displacement signals were measured by lock-in amplifiers. For free vibration decay measurements, the shaker was removed, and a gentle knock was applied near the same position where the shaker was originally attached, and the time dependent displacement of the plate vibration was recorded by a data collection card controlled. The TMD's used in this study were made of miniature decorated membrane resonators (DMR's) reported earlier [42, 43]. The diameter of each stretched membrane was 25 mm with its boundary fixed on the circular plastic frame about 5 mm in width and 5 mm in height. The thickness of the rubber membrane is 0.2 mm. The frame weighs about 0.5 g. A steel platelet 10 mm in diameter and 1.0 g in mass was attached to the center of the membrane. The total mass of a DMR-type miniature TMD is therefore about 1.5 g, while the oscillating mass is only about 1.0 g. The mass density, Poisson's ratio, Young's modulus, and pre-stress of the membrane were 980 kg/m3, 0.49, 5×105, and 0.4 MPa, respectively. The pre-stress was estimated by matching the theoretical eigen-frequency obtained from simulations to the experimental one near 100 Hz. The method to measure the dynamic effective mass of the DMR's reported earlier [42] was used here. 6 In simple terms, the dynamic mass of the DMR was obtained by the measured force acting on the frame divided by the acceleration of the frame. Vacuum grease was applied on the membrane to reduce its Q-factor from around 100 for a pristine DMR to about 25. The second resonance of the DMR's was above 500 Hz so it will not be considered further in this work. Small amount of putty was added onto the platelet to fine tune the first resonance frequency of each DMR, so that collectively they would spread around the first mode of the primary structure. Due to the very small mass of the miniature TMD's, the first mode of the primary structure only shifted by about 0.3 Hz when 4 DMR's were attached, one to each of the four corners. Figure 4, (a) The real part (red curve) and the imaginary part (green curve) of the experimental dynamic mass of a typical DMR damper. The insert is a photo of a DMR damper. (b) The experimental time trace of the displacement of the primary structure with four dampers attached after single impulse excitation. The insert depicts the experimental response function. (c) The delayed Fourier Transform spectra at different delay time obtained from the time trace in (b). The insert is the time dependent peak amplitude obtained from the spectra. A typical measured dynamic mass of a DMR is shown in Fig. 4(a). The insert shows the photo of a DMR. We chose the convention of negative imaginary mass for energy dissipation and positive imaginary part of the eigen-frequency which then leads to the decay of vibration amplitude with time. At the first resonance of the DMR, the maximum amplitude of the imaginary mass was about 25 g, while the Q-factor obtained from the line width at half height is about 25. It is noted that for a classical TMD, the peak imaginary mass Mi is equal to the Q- factor times the oscillator mass M0, i. e., Mi = QM0, which is well validated here. It is a clear indication that the DMR behaves like a classical TMD as far as its damping effect on the primary structure is concerned. Also, the line shapes of the real and the imaginary parts of the dynamic mass resemble well the Lorentzian form that satisfies the Kramers-Kronig relations. The measured first resonant frequency of each of the four DMR's is listed in the first row in Tab. 1. These frequencies were the results of fine-tuning so that with these DMR's, one at each corner of the primary structure, the resulting frequency response of the structure is shown in the insert of Fig. 4(b). The line shape of the frequency response nearly resembles a single resonance with an apparent Q-factor of 13 as estimated from the line width, even though the total mass of the dampers is less than 0.1 % of the primary structure, which is in stark contrast to the typical ones in the literature which are more than 10 to 100 times heavier [27, 28, 30 -- 32]. 7 Compared to the added mass by multiple TMD's of the order of 5 % or more reported in the literature, our dampers would have been considered as insignificant, let alone causing significant damping. We have also noted that the measured frequency response remained almost unchanged when the damper on the upper left corner was moved to the upper right corner, such that there was no damper on the upper left corner while there were two on the upper right corner. Damper # Group-1 Group-2 Table 1 The measured frequency of the imaginary effective dynamic mass peak of each DMR damper used in the experiments. #2 99.3 98.2 #1 99.3 96.2 105.3 101.4 105.2 100.4 #5 #7 #6 #8 101.6 101.8 102.8 104.2 #3 #4 The question of whether such frequency response with an 'apparent' Q-factor of 13 would indeed produce a fast decay comparable to such Q-factor is clarified by the experimental time dependent displacement curve shown in Fig. 4(b). Fast decay of vibration is indeed observed within the first 70 ms (~ 7 periods) after an impulse excitation. The remaining slow decay is mostly due to the second eigenmode of the primary structure around 125 Hz. To examine the displacement curve more closely, we carried out Fourier transform of the remaining time-dependent curve after certain delayed time from the impulse, i. e, delayed Fourier transform analysis. The selected moments of time were the ones at the consecutive maximum positive displacement. Each spectrum was therefore taken at about 10 ms (~ 1 period) later than the earlier one. The resulting spectra are shown in Fig. 4(c), where a fast amplitude decrease of the peak around the first resonance of 100 Hz of the primary structure can be clearly seen. In the insert is the amplitude of the peak value versus the delay time. A fast decay in the first stage followed by a second stage slow decay is clearly seen. By fitting the first stage decay curve with an exponent function, the Q-factor of the decay was found to be 14.2, which is in good agreement with the one obtained from the response function of the structure (Fig. 4(b)), confirming the relationship of time trace and response function. The second stage decay was also clearly revealed, with its DTC much smaller than that of the first stage. Figure 5, (a) The experimental time trace of the displacement of the primary structure with 8 dampers attached. The insert depicts the experimental response function. (b) The delayed Fourier 8 Transform spectra at different delay time obtained from the time trace in (a). The insert is the time dependent peak amplitude obtained from the spectra. To further increase the damping effect, four more DMR's were added and their resonant frequencies were fine-tuned to obtain the experimental response function as shown in the insert of Fig. 5(a), which has an apparent Q = 11. The resonant frequencies of the eight dampers are listed in the second row in Tab. 1. The delayed Fourier transform spectra of the decay curve are shown in Fig. 5(b), while the peak amplitude decay curve is shown as the insert in the same figure. The obtained Q-factor is 10.8, which again matches well the one from the response function. It also demonstrated the damping efficiency of the DMR's with total mass of 12 g, or less than 0.2 % of the primary structure, that can bring out a damping ratio for the first eigenmode of the primary structure to be as high as 4.6 %. Simulations The finite element simulations for the four-damper case in the above experimental investigation were performed by using COMSOL multi-physics with the following materials parameters for the steel plate, density 7850 kg/m3, Young's modulus 2.0 × 1011 Pa, and Poison ratio 0.33. The first five eigenmodes for the bare steel plate were shown in Fig. 6(a). The red color indicates high vibration amplitude, while deep blue indicates near zero vibration. The vibration patterns resemble approximately the expression of . For damping simulations, instead of calculating the primary structure with the real DMR's, which could take up large amount of computer memory and computational power because of the thin membrane structure of the DMR's, the measured dynamic mass spectra of the DMR's were used instead as the effective input parameters for their damping effects [42]. The dampers were placed at the corners of the plate because it is where the vibration amplitude of the first eigenmode is the largest. 9 cos()cos()mxnyFigure 6, (a) The first five resonant modes obtained from numerical simulations. (b) The frequency response functions of the primary structure with four dampers attached (red curve) obtained directly from simulations, and from mimicking using the five eigenmodes (green curve). (c) The delayed Fourier Transform spectra obtained from the time dependent displacement obtained from simulations. (d) The amplitude of the peak in (c) as a function of delay time (red dashed curve) and the numerical fit by a single exponent function with a small offset (green curve). The numerical simulations were carried out in the following steps. In step-1, the response function of the primary structure in the same excitation-response scheme as the experiments was calculated when four dampers with the parameters shown in Tab. 1 were attached to the primary structure. The resulting response function is shown as the red curve in Fig. 6(b), with an apparent Q = 11. The reason for the discrepancy from the experimental value of Q = 13 is most likely due to the imperfection of the real DMR dampers. In step-2 the eigenmodes of the primary structure with the four attached dampers were calculated. The original eigenmode near 100 Hz now splits into five modes, because four additional virtual degrees of freedom were introduced by the four effective TMD's. The eigen-frequencies are {96.4678 + 1.74668i, 98.2078 + 2.06091i, 100.159 + 1.80042i, 102.455 + 2.4188i, 104.085 + 2.02823i} in Hz, revealing the eigenstates with frequencies spreading across the first resonance of the primary structure and the imaginary parts due to DMR damping. The Q-factor of these resonances ranges from 55 to 43, none being close to the experimental value of 13 under the same situation, or the apparent Q-factor of 11 in simulations. Therefore, the response function obtained in step-1 is a 'mimicked' one by the combination of these resonances. In step-3, the amplitudes of the five eigenmodes were then determined by mimicking the resulting response function following Eq. (2) to that obtained in step-1, assuming that only the eigenmodes around 100 Hz are excited. The resulting response function is shown as the green curve in Fig. 6(b). In step-4, the time dependent displacement was then calculated using Eq. (1) and the outcome from step-2 and step-3. In step-5 the delayed Fourier on the time dependent displacement obtained in step-4 was performed. The resulting spectra are shown in Fig. 6(c). The time interval between two consecutive spectra is about 20 ms, or 2 periods. The evolution of the spectra resembles well the experimental one in Fig. 4(c). Finally, in step-6 the total energy decay curve was calculated by using Eq. (4) and the time dependent displacement found in step-4. The outcome is shown in Fig. 6(d). The decay curve exhibits clearly the characteristic multi-stage fast and slow decays similar to that in Fig. 3. The decay started with a fast drop that lasts for about 100 ms, followed by a slow decay till about 200 ms, then fast decay again. The Q-factor obtained from the first stage fast decay is 12, which is in good agreement with the apparent value of 11 obtained from the spectrum in Fig. 6(b). The decay in the remaining stages is negligible because the vibration amplitude is already minute. Overall, the results from the simulations agree very well with the experimental ones, and are in perfect alignment with the coherent super decay phenomemum. Discussions 10 After nearly 100 years since the invention of TMD's, the parameter space for optimizing the performance of TMD's seem to have already been exhaustively combed. What we report in this work reveals a piece of fertile and yet unexplored land in the parameter space, where TMD's with mass only a fraction of a percent of the primary structure can effectively tame the individual resonant modes of the primary structure. The intuitive method of using apparent line width in the frequency response function to adjust the design parameters of the TMD's for effective damping is simple to use in designs and in experiments. The non-uniform multiple-stage decay processes where the decay time constant changes in different stages, such that at the beginning there could be a fast decay of most vibration energy, followed by a period of almost stagnating decay, could be very beneficial for practical applications, where the real effect of the remaining few percent of the initial vibration energy on the primary structure after a stage-1 rapid decay is almost negligible. Therefore, the effectiveness of the dampers is almost entirely determined by the first stage. Although we have only reported the damping of the first resonance of the primary structure, it is straight forward to damp the second resonance which has maximum vibration at the center of each edge (Fig. 6(a)), away from that of the first resonance. One would expect that four dampers similar to the one in Fig. 2(a) are needed. The total added mass to damp the first two resonances would most likely be below 0.5 %. As in most cases the vibration energy is concentrated in the first few resonances of a structure, such approach could suppress the vibration of nearly free bodies with no more than 1% added mass by TMD's. Figure 7, The path (semicircle) to carry out the path integral in Eq. (5), and the simple poles (red and green dots). Our findings also reveal some interesting underline connections between two complex functions which are mutual approximation of each other. Take the example of the mother resonance being approximated by five daughter resonances . and are a mimicking pair, that is, and , and are expressed in the form of to denote such relation. Furthermore, , where 11 220001()/MXiQ5221()/nDnnnnAXiQ()MX()DXRe()Re()MDXXIm()Im()MDXX()()MDXX()()MDxtxtand Eq. (5A), Eq. (5B). The closed integral path is an infinitely large semicircle shown in Fig. 7, with its straight edge along the real axis of the complex - plane [41]. The expression means that the two functions are nearly equal in the first stage, as exemplified in Fig. 3 and in Fig. 6(d), and the difference between them is negligible when compared to their initial strength. However, the two complex functions cannot be equal in the domain of complex analysis [41]. The complex function has two simple poles with residue '1' in the complex -plane, as represented by the two red dots in Fig. 7. The complex function , on the other hand, has ten simple poles (the green dots in Fig. 7) with proper residues, both taking the values according to the mimicking conditions given in Eq. (1). They are closer to the real axis because they have larger Q-factors and therefore smaller imaginary parts as compared to . If one carries out the path integral along the red loop in Fig. 7, the integral with will be zero while that with will be of non-zero. Likewise, along the green loop the path integral with will be zero while that with is non-zero. In more general terms, consider two complex functions and , then when only has two simple poles as the same residues, and only has ten simple poles and the residues as those in with , as long as the integrals along the big semicircle is negligible. As any response function can be mimicked by almost an infinite number of combinations of mimicking functions with their own simple poles and residues, a complex function with two simple poles would somehow be connected to many other complex functions with the right poles and residues. Such intriguing mathematics is yet to be explored. Acknowledgement This work was supported by AoE/P-02/12 from the Research Grant Council of the Hong Kong SAR government. References Abbott, Benjamin P.; et al. (LIGO Scientific Collaboration and Virgo Collaboration) Observation of Gravitational Waves from a Binary Black Hole Merger. Phys. Rev. Lett. 116 (6): 061102 (2016). 12 [1] 1()Re(())Re()2itMMMxtxtXed1()Re(())Re()2itDDDxtxtXed()()MDxtxt()MX()DX()MX()DX()MX()MX()DX1()F2()F12()()FF1()F()MX2()F()DX[2] Weisberg, J. M.; Nice, D. J.; Taylor, J. H. Timing Measurements of the Relativistic Binary Pulsar PSR B1913+16. Astrophysical Journal. 722 (2): 1030 -- 1034 (2010). [3] Herbert Goldstein, Charles P. Poole, and John Safko, Classical Mechanics, 3rd Edition, Pearson (2011) [4] L. Cremer, M. Heckl, E. E. Ungar, Structure-borne sound, Springer (1988). [5] Modal Testing, Theory, Practice and Application, 2nd Edition, D. J. Ewings, Research Studies Press Ltd., Baldock, Hertfordshire, England (2000). [6] [7] [8] Frahm H. Device for damping vibrations of bodies, US Patent 989958; 1909. D. J. Mead, Passive vibration control, Willey (1998). Housner, G., Bergman, L. A., Caughey, T. K., Chassiakos, A. G. Structural control: past, present, and future. J. Eng. Mech. 123(9), 897 -- 971 (1997) [9] Mohamad S. Qatu, Rani Warsi Sullivan, Wenchao Wang, Recent research advances on the dynamic analysis of composite shells: 2000 -- 2009, Composite Structures 93 14 -- 31 (2010) [10] Spencer, B. F., Nagarajaiah, S. State of the art of structural control J. Struct. Eng. 129(7), 845 -- 856 (2003) [11] Ormondroyd J., Den Hartog J. P. The theory of dynamic vibration absorber. Trans. Am. Soc. Mech. Eng. 50:9 -- 22 (1928). [12] Optimum damper parameters in terms of frequency ratio and damping ratio, Den Hartog J. P. Mechanical vibrations. 3rd ed. New York: McGraw-Hill (1947). [13] M. Strasberg and D. Feit, Vibration damping of large structures induced by attached small resonant structures, J. Acoust. Soc. Am. 99 (1), 335, (1996) [14] J. A. Zapfe, G. A. Lesieutre, Broadband vibration damping using highly distributed tuned mass absorbers, AIAA Journal 35 (4), 753-755 (1997). [15] M. N. Hadi, Y. Arfiadi, Optimum design of absorber for MDOF structures, Journal of Structural Engineering 124 (11) 1272 -- 1280 (1998). [16] K. Nagaya and L. Li, Control of sound noise radiated from a plate using dynamic absorbers under the optimization by neural network, Journal of Sound and Vibration 208, 289 (1997) [17] L. Zuo, S. A. Nayfeh The two-degree-of-freedom tuned-mass damper for suppression of single-mode vibration under random and harmonic excitation, Journal of Vibration and Acoustics 128, 56 -- 65 (2006). [18] M. S. Khun, H. P. Lee, S. P. Lim, Structural intensity in plates with multiple discrete and distributed spring-dashpot systems, Journal of Sound and Vibration 276 (2004) 627 -- 648. [19] M. B. Ozer, Extending den Hartog's vibration absorber technique to multi-degree-of- freedom systems, Journal of Vibration and Acoustics 127 341 -- 350 (2005). 13 [20] R. L. Harne, C.R. Fuller, Modeling of a passive distributed vibration control device using a superposition technique, Journal of Sound and Vibration 331, 1859-1869 (2012). [21] Y. L. Cheung, W. O. Wong, H∞ and H2 optimizations of a dynamic vibration absorber for suppressing vibrations in plates, Journal of Sound and Vibration 320, 29 -- 43 (2009). [22] L. Zuo, S. Nayfeh, Design of passive mechanical systems via decentralized control techniques, in 43rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2002, AIAA 2002 -- 1282. H2 and H∞ optimization [23] D. I. G. Jones, Response and damping of a simple beam with tuned damper, Journal of the Acoustical Society of America 42 (1), 50 -- 53 (1967). [24] H. N. Ozguven, B. Candir, Suppressing the first and second resonances of beams by dynamic vibration absorbers, Journal of Sound and Vibration 111 (3) 377 -- 390 (1986). [25] R. G. JACQUOT, Suppression of random vibration in plates using vibration absorbers Journal of Sound and Vibration 248(4), 596(2001) [26] Paolo Gardonio and Michele Zilletti, L. C. Integrated tuned vibration absorbers: A theoretical study J. Acoust. Soc. Am. 134, (2013) [27] Reduced-order multiple tuned mass damper optimization: A bounded real lemma for descriptor systems approach, M. Jokic, M.Stegic, M.Butkovic, Journal of Sound and Vibration 330, 5259 -- 5268 (2011). 1 square meter 4 mm thick steel plate (3.12 Kg) simply supported at four corners, total TMD mass 1.256 Kg, damping factors ~ 0.05 estimated from the FR figures. [28] J. Michielsen, I. Lopez Arteaga, H. Nijmeijer, LQR-based optimization of multiple tuned resonators for plate sound radiation reduction, Journal of Sound and Vibration 363 166 -- 180 (2016) 0.59 Kg Al plate, up to 35% of added mass of multiple TMD. [29] Qi Qin and Mei-Ping Sheng, Analyses of multi-bandgap property of a locally resonant plate composed of periodic resonant subsystems, International Journal of Modern Physics B32, 1850269 (2018) [30] Yiqing Yang, Wei Dai, Qiang Liu Design and implementation of two-degree-of-freedom tuned mass damper in milling vibration mitigation Journal of Sound and Vibration 335 78 -- 88 (2015), 3% of 8 kg = 72 g [31] L. Zuo, S.A. Nayfeh Minimax optimization of multi-degree-of-freedom tuned-mass dampers Journal of Sound and Vibration 272 (2004) 893 -- 908, two dampers, each is 3.345 kg and 19.8 cm x 8.382 cm x 2.54 cm [32] Jae-Sung Bae, Jai-Hyuk Hwang, Dong-Gi Kwag, Jeanho Park, and Daniel J. Inman, Vibration Suppression of a Large Beam Structure Using Tuned Mass Damper and Eddy Current Damping, Shock and Vibration 2014, Article ID 893914, 26 kg eddy current damper 14 [33] C. R. Fuller, S. J. Elliott, P. A. Nelson, Active control of Vibration, Academic Press (1996). [34] A. Preumont, Vibration Control of Active Structures: An Introduction, Kluwer Academic Publishers (2002). [35] D. Karnopp, M. J. Crosby, R.A. Harwood, Vibration Control using semi-active force generators, ASME Journal of Engineering for Industry 96 (2) 619-626 (1974). [36] C. Maurini, dell'Isola, F. & Del, D. Vescovo, Comparison of piezoelectronic networks acting as distributed vibration absorbers, Mechanical Systems and Signal Processing 18, 1243-1271 (2004). [37] S-M. Kim, S. Wang, M. J. Brennan, Dynamic analysis and optimal design of a passive and an active piezo-electrical dynamic vibration absorber, Journal of Sound and Vibration 330 (4) 603 -- 614 (2011). [38] R. L. Harne, On the linear elastic, isotropic modeling of poroelastic distributed vibration absorbers at low frequencies, Journal of Sound and Vibration 332 3646-3654 (2013). [39] Giovanni Ferrari, Marco Amabili, Active vibration control of a sandwich plate by non- collocated positive position feedback Journal of Sound and Vibration 342 44 -- 56 (2015). [40] Ehsan Omidi, S. Nima Mahmoodi, W. Steve Shepard Jr. Multi positive feedback control method for active vibration suppression in flexible structures Mechatronics 33, 23 -- 33 (2016) . [41] Mary L Boas, Mathematical methods in the physical sciences, 3rd Ed., Wiley (2006) [42] Liang Sun, Ka Yan Au-Yeung, Min Yang, Suet To Tang, Zhiyu Yang, and Ping Sheng. Membrane-type resonator as an effective miniaturized tuned vibration mass damper. AIP Advances, 6(8):085212 (2016). [43] Z. Yang, M. Yang, N. H. Chan, and P. Sheng, Membrane-type acoustic metamaterial with negative dynamic mass, Phys. Rev. Lett. 101, 204301 (2008) 15
1801.08053	1	1801	2018-01-24T16:07:45	Strain Mapping of Two-Dimensional Heterostructures with Sub-Picometer Precision	[ "physics.app-ph", "cond-mat.mes-hall" ]	Next-generation, atomically thin devices require in-plane, one-dimensional heterojunctions to electrically connect different two-dimensional (2D) materials. However, the lattice mismatch between most 2D materials leads to unavoidable strain, dislocations, or ripples, which can strongly affect their mechanical, optical, and electronic properties. We have developed an approach to map 2D heterojunction lattice and strain profiles with sub-picometer precision and to identify dislocations and out-of-plane ripples. We collected diffraction patterns from a focused electron beam for each real-space scan position with a high-speed, high dynamic range, momentum-resolved detector - the electron microscope pixel array detector (EMPAD). The resulting four-dimensional (4D) phase space datasets contain the full spatially resolved lattice information of the sample. By using this technique on tungsten disulfide (WS2) and tungsten diselenide (WSe2) lateral heterostructures, we have mapped lattice distortions with 0.3 pm precision across multi-micron fields of view and simultaneously observed the dislocations and ripples responsible for strain relaxation in 2D laterally-epitaxial structures.	physics.app-ph	physics	Strain Mapping of Two-Dimensional Heterostructures with Sub-Picometer Precision Yimo Han1, Kayla Nguyen3, Michael Cao1, Paul Cueva11, Saien Xie1,2, Mark W. Tate4, Prafull Purohit4, Sol M. Gruner4,5,6,7, Jiwoong Park3, David A. Muller1,6 1. School of Applied and Engineering Physics, Cornell University, Ithaca, NY 14853, USA 2. Department of Chemistry, Institute for Molecular Engineering, and James Franck Institute, University of Chicago, Chicago, IL 60637, USA 3. Chemistry and Chemical Biology Department, Cornell University, Ithaca, NY 14853, USA 4. Laboratory of Atomic and Solid State Physics, Cornell University, Ithaca, NY 14853, USA 5. Physics Department, Cornell University, Ithaca, NY 14853, USA 6. Kavli Institute at Cornell for Nanoscale Science, Ithaca, NY 14853, USA 7. Cornell High Energy Synchrotron Source, Cornell University, Ithaca, NY 14853, USA Correspondence to: [email protected], [email protected] Next-generation, atomically thin devices require in-plane, one-dimensional heterojunctions to electrically connect different two-dimensional (2D) materials. However, the lattice mismatch between most 2D materials leads to unavoidable strain, dislocations, or ripples, which can strongly affect their mechanical, optical, and electronic properties. We have developed an approach to map 2D heterojunction lattice and strain profiles with sub-picometer precision and to identify dislocations and out-of-plane ripples. We collected diffraction patterns from a focused electron beam for each real-space scan position with a high-speed, high dynamic range, momentum-resolved detector – the electron microscope pixel array detector (EMPAD). The resulting four-dimensional (4D) phase space datasets contain the full spatially resolved lattice information of the sample. By using this 1 technique on tungsten disulfide (WS2) and tungsten diselenide (WSe2) lateral heterostructures, we have mapped lattice distortions with 0.3 pm precision across multi-micron fields of view and simultaneously observed the dislocations and ripples responsible for strain relaxation in 2D laterally-epitaxial structures. Strain fields and dislocations play an important role in determining the mechanical and electronic properties of crystalline materials. Conventional transmission electron microscopy (TEM) can identify strain fields and dislocations in bulk materials using the diffraction contrast from two-beam analysis1 or weak-beam dark field imaging2. However, given the rod-like nature of diffraction peaks normal to the film, these methods are very difficult to implement in 2D materials, which are confined to atomic dimensions in the direction of the beam propagation. While geometric phase analysis (GPA)3-6 on atomic-resolution images can provide strain maps for 2D materials, the field-of-view is limited to a few tens of nanometers by the need to resolve every atom. When grains or heterostructures of 2D materials7-10 laterally reach the more typical micron scales, measuring strain and dislocations at atomic resolution using GPA becomes impractical. Conversely, nanobeam diffraction (NBD) combined with scanning TEM (STEM)11-15 effectively decouples the spatial resolution from the strain mapping precision, allowing for high-precision strain measurements across a larger sample area. However, this approach on 2D materials has historically been limited by the speed and dynamic range of the existing detectors, as 2D materials are sensitive to the electron beam and weak scatterers. To overcome these issues, we developed a method to map the strain and identify dislocations in 2D crystals using an electron microscope pixel array detector (EMPAD)16 designed at Cornell. Due to its high-speed, high dynamic range, and high sensitivity, the scanning NBD can be achieved within minutes with no noticeable damage to 2D samples, ultimately providing sub-picometer precision strain mapping over length scales, ranging from angstroms to many micrometers. The EMPAD operates at a range of accelerating voltages from 20 kV to 300 kV (The work presented here was conducted at 80 kV). Its single-electron sensitivity allows for quantitative analysis of diffraction from a single 2 atom17,18, which is highly advantageous for studying 2D materials only one- to three- atoms thick. Moreover, the EMPAD's high dynamic range enables collection of all transmitted electrons at small convergence angles with the primary beam unsaturated and diffracted beams clearly resolved (Fig. 1b), as demonstrated by integrating the center beam (and one diffracted spot) to plot the virtual bright field (and dark field) images, as shown in Fig. 1c and 1d. The samples we examined were laterally stitched epitaxial monolayer transition metal dichalcogenide (TMD) heterojunctions, which are synthesized through metal-organic chemical vapor deposition (MOCVD)19,20. From the EMPAD's 4D dataset of a sample on 20 nm SiO2 TEM grids, we extracted the ADF-STEM signal (Fig. 2a) by integrating the diffraction patterns masked by a virtual ADF detector with an inner angle of 50 mrad. The ADF-STEM image provides little contrast difference between WS2 and WSe2, as the heavy tungsten atoms dominate the contrast. In comparison, the lattice constant map (Fig. 2b) calculated by measuring shifts in the reciprocal lattice vectors clearly distinguishes the two materials, which contain nanometer-sharp interfaces (inset of Fig. 2b) (see SI#1 and #2 for calculation details). From a histogram of lattice constant measurment (Fig. 2c), we extracted the mean values as the statistically averaged lattice constant for WS2 (3.182 ± 0.0005 Å) and WSe2 (3.282 ± 0.001 Å), indicating a 3.1% lattice mismatch (fully relaxed films have 4.5% lattice mismatch21). The histogram (inset of Fig. 2c) from a flat region (gray box in Fig. 2b) indicates this method has a precision higher than 0.3 pm, with local sample distortions placing an upper limit on the spread. The sub-picometer precision of the lattice constant mapping relies on high angular resolution when we measure the centers of the diffraction spots. For 2D materials, the center of mass (CoM) is an efficient approach to achieve a high angular resolution in the diffraction space for strain mapping. The calculation details, optimization of pixel sampling, and error analysis are provided in supplementary information #1. From the CoMs of all diffraction spots, we can also extract the diffraction vectors gi (i=1,2) (Fig. S4), i.e. the reciprocal lattice vectors, which can be used to map the strain and rotation (see details in SI #3). The x-direction uniaxial strain (εxx) map (Fig. 3a) shows 3 clear differences between the WS2 and WSe2 as well as small local variations, indicating the film is largely relaxed, but not completely. This is expected since the width of WS2 is far beyond the critical thickness (a few nanometers for TMD materials), as narrow heterojunctions below the critical thickness remain coherent and exhibit uniaxial strain parallel to the interface (Fig. S5). To see the strain details, we plot the histograms of the x-direction strain map, where the WS2 peak in Fig. 3b fits two Gaussian peaks, corresponding to the unstrained (outer edges) and strained (interfaces) parts of the WS2 lattice in Fig. 3a. From a relatively flat WS2 region (gray box in Fig. 3a), we determined that the precision of our technique is at least 0.18%, as given by the spread of the histogram in the insert of Fig. 3b. The rotation map (Fig. 3c) displays periodic misfit dislocations as dipole fields located at the interface between WS2 and WSe2. The misfit dislocations along the interface contribute to release the lattice strain. However, the observed misfit dislocation spacing (~100 nm) is much larger than the spacing required to fully relax the lattice strain caused by the 3.1% lattice mismatch between WS2 and WSe2 (~10 nm). We note an internal periodic strain field in the outer WS2 region in the rotation map (Fig. 3c). Analogous to bulk epitaxy, which forms periodic ripples within the top layers of the thin films22, the periodic strain fields lower the elastic strain energy in 2D heterojunctions and create intrinsic dislocations inside WS2, as depicted in the magnified rotation map (Fig. 3d). However, the long wavelength of the strain field (in microns) implies that it plays a minor role in releasing strain. A more significant contribution for releasing strain is that 2D materials can also buckle up and form out-of-plane ripples without introducing in-plane lattice distortions. We observed that the WSe2 forms this type of ripples, which is the origin of WSe2's broader peak in Fig. 3b. These out-of-plane ripples contribute to relax the lattice strain dramatically20,23. To accurately identify the out-of-plane ripples and quantitatively map their orientations, we developed a novel approach, using the EMPAD 4D datasets. Due to the cone-shaped diffracting beam from the microscopic corrugation of 2D materials24, the tilt of the 2D materials results in elliptical diffraction spots, as illustrated schematically in Fig. 4a. 4 Here, we mapped the tilted regions (i.e. the ripples) by measuring the relative broadening of the diffraction spots compared to flat regions. We defined the complex ripple measure as 𝑅 = 𝐴 + 𝐵𝑒𝑖2𝜋/3 + 𝐶𝑒𝑖4𝜋/3, where A, B, and C are the characteristic sizes (as measured by second moments) of the corresponding spots in Fig. 4a. As illustrated in the phase plot of the R maps, the ripples form along different orientations, and only appear inside WSe2 (Fig. 4b). In addition, the ripples prefer to form perpendicular to the interface between WS2 and WSe2 on the WSe2 side, thus releasing the lattice strain, as shown in Fig. 4c. The calculation details can be found in supplementary information #4. So far, we mapped the lattice constant, strain, rotation, and ripples in the lateral heterostructure and identified uniaxial strain, dislocations, and ripples. All the lattice information is recorded within a single EMPAD 4D dataset. In electron microscopy, principle component analysis (PCA) has been widely used for multi-dimensional datasets, such as electron energy loss spectra (EELS)25,26. In EMPAD 4D datasets, PCA as a screening tool provides a statistical approach to immediately obtain a hierarchy of information by extracting the linearly uncorrelated patterns that contain the majority of the variance in the data (SI #5). For our 4D data of the WS2-WSe2 multi-junctions, we blocked the center beam for PCA to avoid any dominating features inside the center beam. The principle components of the diffraction patterns (Fig. 5c-h) and their corresponding weighting maps (Fig. 5i-l) capture the highest variance features in the 4D dataset (more maps in Fig. S7). By comparing the maps generated using the former method (Fig. 5a-d), we can relate the second to fifth principle components to the filtered dark field (DF) of the multi-junction (Fig. 5e, i), the rotation of the lattice (Fig. 5f, j), the lattice constant difference (Fig. 5g, k), and the local ripples in WSe2 region (Fig. 5h, l), respectively. PCA offers a swift and facile method to analyze the 4D EMPAD data and quickly highlights the largest changes for our future studies. In conclusion, we developed an approach to map lattice constant, strain, dislocations, and out-of-plane ripples with high precision on all relevant length scales. All lattice information can be extracted from the 4D data which requires only a single fast scan, effectively reducing the electron dose in 2D materials. Moreover, the accuracy of CoM 5 measurements allows us to map the lattice constant and strain with precisions greater than 0.3 pm and 0.18% respectively at WS2-WSe2 lateral heterojunctions. In addition, we observed that the lattice strain is released mainly by misfit dislocations and out-of-plane ripples, both of which play an important role in the mechanical and electrical properties of 2D materials. This method will be crucial for studying the lattice distortions in 2D materials and their atomically thin circuitry by identifying the lattice strain and dislocations and offering essential feedback to material syntheses. 6 Figure 1 EMPAD imaging. a, Schematic of the EMPAD operation, where a full diffraction pattern, including the unsaturated primary beam, is recorded at each scan position. b, Diffraction images taken by EMPAD. The top panel shows the diffraction image of a 5 nm SiNx film, while the bottom panel displays the diffraction pattern of a WSe2 monolayer located on the 5 nm SiNx film. c and d show the virtual bright field and filtered dark field images obtained by integrating the central and the labeled diffracted beam, respectively, as indicated on their top left sections. 7 Figure 2 Lattice constant map. a, Annular dark field (ADF) image extracted from the EMPAD 4D data on a wide WS2-WSe2 lateral heterojunction. The inner detector angle is 50 mrad. b, Lattice constant map of micron-sized triangles. The inset displays a line profile across the interface between WSe2 and WS2. c, Lattice constant histogram from b. The inset is the histogram from a flat region (gray box in b), indicating a precision of at least ~0.3 pm. 8 Figure 3 Strain maps. a, Uniaxial strain map showing most of the strain has been released. b, Strain histogram from a. The WS2 peak fits two Gaussians that correspond to unstrained and strained regions. The inset shows the histogram from a flat region (gray box in a) indicating a precision better than ~ 0.18%. c,d, The rotation map displaying periodic misfit dislocations that contribute to relaxing the lattice strain at the WS2-WSe2 junction. The internal strain in WS2 results in a few dislocations inside WS2, which are indicated by the white circles. 9 z Figure 4 Map of out-of-plane ripples. a, Schematics showing that the tilt causes the ellipsoidal broadening of the diffraction spots due to the corrugation of the 2D materials. The broadening of the spots was measured through the second moment measurement. b, The phase plot of relative broadening shows the oriented ripple map where the WSe2 film forms out-of-plane ripples to release the strain, while WS2 is flat. c, Magnified ripple map showing a nanoscale ripple array along the junction. 10 Figure 5 Extracted data vs PCA on EMPAD 4D data. a Filtered DF, b rotation map, c lattice constant map, and d ripple map. e-h, The PCA scores of the second to fifth principle components of the EMPAD 4D dataset and i-l their corresponding PCA loads. e,i show the crystalline structure, f,j depict the lattice rotation, g,k represent the lattice constant difference, and h,l display the local ripples. 11 Corresponding Author David A. Muller ([email protected]), Yimo Han ([email protected]) Acknowledgements This work was primarily supported by the Cornell Center for Materials Research with funding from the NSF MRSEC program (DMR-1719875). MC was supported by DOD-MURI (Grant No. FA9550-16-1-0031). Detector development at Cornell is supported by DOE award DE- SC0017631 to SMG. We thank Mervin Zhao, Megan Holtz, Zhen Chen, Mengnan Zhao, Gabriela Correa, Lei Wang for helpful discussions. We thank Mariena Ramos and Lena Kourkoutis for help with electron microscopes. Author contributions The project was conceived and designed by Y.H. under the supervision of D.A.M.. The experimental results and data analysis were obtained by Y.H., with help from K.N., M.C., and P.C.. The samples were grown by S.X. under the supervision of J.P.. The EMPAD was developed by M.W.T., P.P., K.N., and M.C. under the supervision of S.M.G. and D.A.M.. 12 REFERENCES 1. Hirsch, P. B.; Whelan, M. J. Phil. Trans. R. Soc. Lond. A 1960, 252, 499–529. 2. Cockayne, D. J. H.; Ray, I. L. F.; Whelan, M. J. Philosophical Magazine 1969, 20:168, 1265-1270. 3. Hytch M.J.; et al. Ultramicroscopy 1998, 74, 131-146. 4. Hytch, M. J.; Putaux, J.; Penisson, J. M. Nature, 2003, 423, 270-273. 5. Han, Y.; et al. Nature Materials 2017, doi:10.1038/nmat5038 6. Han, Y.; et al. Microscopy & Microanalysis 2016, 22(S3), 870-871. 7. van der Zande, A. M.; et al. Nature Materials 2013, 12, 554–561. 8. Li, M.-Y.; et al. Science 2015, 349, 524-528. 9. Zhao, M.; et al. Nat. Nanotech. 2016, 11, 954-959. 10. Zhang, Z.; et al. Science 2017, eaan6814. 11. Uesugi, F.; et al. Ultramicroscopy 2011, 111, 995-998. 12. Muller, K.; et al. Microscopy & Microanalysis 2012, 18, 995-1009. 13. Zuo, J.M.; Spence, J.C.H. Strain measurements and mapping. In: Advanced Transmission Electron Microscopy. 553-580, Springer, New York, 2017. 14. Ozdol, V.B.; et al, Appl. Phys. Lett. 2015, 106, 253107. 15. Pekin, T.C.; et al. Ultramicroscopy 2017, 176, 170-176. 16. Tate, M.W.; et al, Microscopy & Microanalysis 2016, 22, 237-249. 17. Muller, D.A.; et al, Microscopy & Microanalysis 2016, 22(S3), 846-847. 18. Cao, M.; et al, arXiv:1711.07470v2 2017 19. Kang, K.; et al. Nature 2015, 520, 656-660. 20. Xie, S.; et al. arXiv:1708.09539 2017 21. Kang, J.; et al. Appl Phys Lett 2013, 102, 012111 22. Cullis, A. G.; Robbins, D. J.; Barnett, S. J.; Pidduck, A. J. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 1994, 12, 1924–1931. 23. Nandwana, D.; Ertekin, E. Nano. Lett., 2015, 15, 1468-1475 24. Meyer, J.C.; et al. Nature 2007, 446, 60-63. 25. Trebbia, P.; Bonnet, N. Ultramicroscopy 1990, 34, 165–178. 26. Cueva, P.; et al. Microscopy & Microanalysis 2012, 18, 667-675. 13 Supporting Information for Strain Mapping of Two- Dimensional Heterostructures with Sub-Picometer Precision Yimo Han1, Kayla Nguyen3, Michael Cao1, Paul Cueva11, Saien Xie1,2, Mark W. Tate4, Prafull Purohit4, Sol M. Gruner4,5,6,7, Jiwoong Park3, David A. Muller1,6 1. School of Applied and Engineering Physics, Cornell University, Ithaca, NY 14853, USA 2. Department of Chemistry, Institute for Molecular Engineering, and James Franck Institute, University of Chicago, Chicago, IL 60637, USA 3. Chemistry and Chemical Biology Department, Cornell University, Ithaca, NY 14853, USA 4. Laboratory of Atomic and Solid State Physics, Cornell University, Ithaca, NY 14853, USA 5. Physics Department, Cornell University, Ithaca, NY 14853, USA 6. Kavli Institute at Cornell for Nanoscale Science, Ithaca, NY 14853, USA 7. Cornell High Energy Synchrotron Source, Cornell University, Ithaca, NY 14853, USA *Correspondence to: [email protected], [email protected] 14 Contents: 1. Measuring the center of mass (CoM) 2. Mapping lattice constant 3. Mapping the strain 4. Mapping the ripples using second moments 5. Principle component analysis (PCA) 15 1: Measuring the center of mass (CoM) CoM calculation The CoM is calculated from the diffraction pattern, 𝐼(𝑝⃗), using the following equation: ⟨𝑝⃗⟩ = ∫ 𝑝⃗ 𝐼(𝑝⃗)𝑑𝑝⃗ (1) where 𝑝⃗ is the momentum in the diffraction space. Fig. S1 shows the definition of our masks (green circles) that were applied to the diffraction disks and the center disk when we calculated their CoMs. The green dots label the calculated CoMs of all disks. Fig. S1c shows that the diffraction disk span across ~6 pixels in diameter. The mask with a diameter of 12 pixels is aligned to the disk with 1/3-pixel resolution by eye. We attempted aligning the mask as well as possible and, in fact, we achieved that the center of the mask is close to the measured CoM in Fig. S1d. In addition, measuring the centers for the EMPAD 4D data, generally a few gigabytes in size, requires a fast-computational algorithm. The CoM calculation is an O(n) algorithm (see code below), where n can usually be only tens of pixels for each diffraction disk. As a conclusion, CoM is a high-efficiency approach for measuring centers for the EMPAD 4D datasets. 16 Figure S1 CoM measurements. a, A diffraction pattern of WS2 on a 5 nm SiNx window taken by EMPAD. b, The diffraction pattern overlaid by the masks (green circles) and their CoMs (green dots). We manually placed the masks and aligned them to the diffraction spots by eye. c,d, A magnified diffraction disk (c) and its overlay with the mask and the CoM (d). The accuracy of CoM measurements CoM has the advantage of speed and simplicity compared to more elaborate curve fitting procedures. The CoM provides sufficiently accurate centers of diffraction patterns of 2D materials mainly because of the following two reasons: 1) the rod-like nature of diffraction patterns of 2D materials and 2) the high dynamic range of our EMPAD, which counts all transmitted electrons. The errors for CoM measurements come from the unavoidable Poisson noise from the detector. Here, we discuss how Poisson noise affects the CoM. For each pixel in the EMPAD, the Poisson noise is proportional to √𝑁, where N is the number of electrons hitting that pixel. By calculating the error propagation in Equation (1), we achieved that 𝛿𝐶𝑜𝑀 = √⟨𝑝2⟩ 𝑁 ∝ 1 √𝑁 , where 𝛿𝐶𝑜𝑀 is the absolute error of CoM, and ⟨𝑝2⟩ is the 17 second moment defined in Equation (5). This result shows that increasing the electron dose, i.e. using a high beam current, can effectively reduce the error caused by Poisson noise. However, there is a trade-off between high beam current for more accurate CoM and low beam current to avoid electron beam damage in 2D materials. Experimentally, we were working on the regime that the number of electrons in each diffraction spot are on the order of 104 to 105. Moreover, 𝛿𝐶𝑜𝑀 also depends on the number of pixels the disk spans across (the diameter of the disk in the unit of pixels, as shown in Fig. S2a). We simulated 𝛿𝐶𝑜𝑀 for different disk diameters by averaging the errors from 1000 diffraction patterns with Poisson noise, and plotted the results in Fig. S2b. The results show the absolute error (𝛿𝐶𝑜𝑀) is proportional to the disk diameter for a given dose. Designing number of pixels in the detector The parameter determines the angular resolution is the percentage error, which is the ratio between the absolute error (𝛿𝐶𝑜𝑀) and the k vector length – 𝛿𝐶𝑜𝑀/𝑘. We note that k, as well as the disk diameter, changes correspondingly as we change the number of pixels in the detector (or change the camera length to magnify the diffraction patterns), as shown in Fig. S2c. In addition, k is proportional to the disk diameter. Here, we ignored a constant scaling prefactor, assumed k equals to the disk diameter, and plotted the percentage errors in Fig. S2d. The percentage errors are close to a constant with small increment at the smaller disk diameters, especially for the low-dose case. For example, if each diffraction disk only contains 10 electrons (blue curve in Fig. S2d), we would choose 10-15 pixels for a disk diameter for reasonable angular resolution. For doses larger than 1000 electrons per disk (we worked at 104 to 105) the optimized disk diameters will be any one larger than 5 pixels. Above that, the errors stay constant, indicating that we do not benefit from designing more pixels in the detector or magnifying the diffraction patterns. Optimizing convergence angle 18 For diffraction disk at a fixed k and dose, we will have a smaller percentage error if we have smaller convergence angle (as shown in Fig. S2e and f). Although smaller convergence angles (less pixels) are preferred, the lower bound is a 2x2- pixel-sized diffraction disk, which is similar to a differential phase contrast (DPC) detector. In addition, there is also a tradeoff between the angular resolution and spatial resolution: 𝑑0 = 1.22 𝜆 , where d0 is the probe size, 𝜆 is the wavelength of the θ electrons, and 𝜃 is the convergence angle of the electron beam. In our case, we used a 0.5 mrad convergence angle, where the diffraction disks span several of pixels on the EMPAD, resulting in a 5 nm probe size. Different probe sizes can be selected for the length scale appropriate for the sample, with large probes desirable at large length scales to avoid under sampling. 19 Figure S2 Accuracy of CoM measurements. a, Simulated diffraction patterns of different diameters (in pixels) with Poisson noise. b, Absolute errors of CoM (𝛿𝐶𝑜𝑀) proportional to the disk diameters. c, Schematics of diffraction patterns when we increase the number of pixels in the detector (or magnify the diffraction patterns by changing the camera length). d, Percentage error plot (𝛿𝐶𝑜𝑀/𝑘) under the situation described in (c), with arrows indicating the lower bounds of the optimized disk 20 diameters. (We used k = disk diameter.) e, Schematics of diffraction patterns when we change the convergence angle. f, Percentage error plot (𝛿𝐶𝑜𝑀/𝑘) for cases in (e), showing that reducing the convergence angle will reduce the error and improve the angular resolution dramatically. (We used k = 30 pixels.) 2: Mapping lattice constant To calculate the lattice constants from a single diffraction pattern, we averaged the distances between diffracting beams and the center beam, d1 to d6, as shown in Fig. S3a. The averaged lattice constant is: aave = 6a0d0 6 ∑ di i=1 (2) where a0 and d0 are the calibrated ones from a referenced region. We used flat WS2 as the reference. In STEM, EMPAD acquires diffraction patterns at each scan position with 1.86 ms/frame (1 ms exposure time and 0.86 ms readout time) – so a 4D data (x and y in real space and kx and ky in momentum space) at 256×256 scan points can be reached in about two minutes. Using the 4D dataset, we can map the lattice constant throughout the entire sample. Although there are strain and tilt at some regions in the sample, due to the averaging of the six spots in different directions, the strain and tilt effects are negligible. Fig. S3b and c show the schematics of how strain and tilt affect the diffraction pattern. The calculation below describes that the strain and tilt are higher order effects in the lattice constant calculation: a′ave ≅ 3a0d0 d1(1+ε′)+2d2(1−√3 2 νε′) (3) where 𝜈 is the Poisson's ratio (0.25 for WS2) and ε′ is a small uniaxial strain (we used compressive strain here). a′′ave ≅ 3a0d0 𝑑1(1+ 𝜃2 2 )+2𝑑2(1+ ) 𝜃2 8 (4) where 𝜃 is the small tilt angle. 21 Figure S3 Lattice constant map. a, Diffraction pattern of WS2-WSe2 on 20 nm SiNx windows. The gray circles are the masks we used to calculate the CoMs. The reciprocal lattice constants were measured, shown as d1 – d6. b, Schematic showing how strain affects the lattice constant measurements. c, Schematic showing how small tilt affects the lattice constant measurements. The small strain and tilt are higher order effect for lattice constant calculation. 3: Mapping strain To map the strain from the 4D dataset, we calculated the diffraction vectors gi (i=1,2) (i.e. the reciprocal lattice vectors) as shown in Fig. S4a. The reference diffraction vectors gi ref were set by averaging 200 scan positions (or pixels) in real 22 space where half of them are on WS2 and the other half are on the WSe2 region. The choice of these gi ref is for mapping convenience. Afterwards, we derived the transformation matrix T using gi = T gi ref. T can be polar-decomposed into a rotation matrix R and a strain matrix U, from which the uniaxial strain εxx=1-U11 and εyy=1- U22, shear strain εxy=U12, and rotation εrot=asin(R12) can be calculated. Figure S4 Maps of diffraction vectors. a, A single diffraction pattern of WSe2 taken by EMPAD with second order diffraction spots highlighted by the masks. For each spot, we calculated the CoM and achieve the diffraction vectors as labeled by g1 and g2. b,c, g1 and g2 maps over the entire triangle. 23 Figure S5 Strain maps of WS2-WSe2 superlattices. a-d, x-direction uniaxial strain (a), y-direction uniaxial strain (b), shear strain (c), and rotation (d) maps. e,h, Magnified x-direction uniaxial strain and y-direction uniaxial strain maps from the white boxes in (a) and (b). f,g, Histogram and the line profiles of (e). i,j, Histogram and line profiles of (h). 24 Figure S6 Strain maps of thick WS2-WSe2 junctions. a, x-direction uniaxial strain map. b, y-direction uniaxial strain map. c, shear strain map. d, rotation map. 25 Figure S7 PCA of 4D dataset. a,d, The first principle component which represents the SiNx substrate and the averaged diffraction patterns from the lattice. b,e, The sixth principle component representing the ripple with a different orientation from the one in Fig. 5i and j. c,f, The seventh principle component showing higher order derivatives of the diffraction patterns. For principle components more than seventh order, the real space images start to show less feature as the higher order terms mainly contain noise. 26
1910.09354	1	1910	2019-10-21T13:22:04	Quantifying mobile ions in perovskite-based devices with temperature-dependent capacitance measurements: frequency versus time domain	[ "physics.app-ph", "cond-mat.mtrl-sci" ]	Perovskites have proven to be a promising candidate for highly-efficient solar cells, light-emitting diodes, and X-ray detectors, overcoming limitations of inorganic semiconductors. However, they are notoriously unstable. The main reason for this instability is the migration of mobile ions through the device during operation, as they are mixed ionic-electronic conductors. Here we show how measuring the capacitance in both the frequency and the time domain can be used to study ionic dynamics within perovskite-based devices, quantifying activation energy, diffusion coefficient, sign of charge, concentration, and the length of the ionic double layer in the vicinity of the interfaces. Measuring the transient of the capacitance furthermore allows for distinguishing between ionic and electronic effects.	physics.app-ph	physics	Quantifying mobile ions in perovskite-based devices with temperature-dependent capacitance measurements: frequency versus time domain Moritz H. Futscher1, Mahesh K. Gangishetty2, Daniel N. Congreve2 and Bruno Ehrler1 1. AMOLF, Center for Nanophotonics, Science Park 104, 1098 XG Amsterdam, The Netherlands 2. Rowland Institute at Harvard, 100 Edwin H. Land Blvd, Cambridge, Massachusetts, United States Abstract Perovskites have proven to be a promising candidate for highly-efficient solar cells, light- emitting diodes, and X-ray detectors, overcoming limitations of inorganic semiconductors. However, they are notoriously unstable. The main reason for this instability is the migration of mobile ions through the device during operation, as they are mixed ionic-electronic conductors. Here we show how measuring the capacitance in both the frequency and the time domain can be used to study ionic dynamics within perovskite-based devices, quantifying activation energy, diffusion coefficient, sign of charge, concentration, and the length of the ionic double layer in the vicinity of the interfaces. Measuring the transient of the capacitance furthermore allows for distinguishing between ionic and electronic effects. Introduction Measuring the temperature-dependent capacitance as a function of frequency or time is a well-established technique in experimental physics to quantify electronic defect states in semiconductors.1 -- 4 The most famous examples are deep-level transient spectroscopy (DLTS) and thermal admittance spectroscopy (TAS).5,6 These techniques allow for quantifying activation energy, attempt-to-escape frequency, sign, and concentration of electronic defect states. Due to the intriguing properties of perovskites such as high charge-carrier mobilities, long diffusion lengths, strong absorption, and low exciton binding energies, perovskites have 1 been successfully used in many optoelectronic applications including light-emitting diodes, lasers, and X-ray detectors.7 -- 10 Both DLTS and TAS have been used to study perovskite-based devices.11 -- 15 To measure electronic defect states, these techniques rely on the depletion approximation, assuming that the depletion region is free of mobile carriers. In the case of perovskites, however, this assumption is not fulfilled because they are mixed ionic-electronic conductors, leaving slow charged carriers (mobile ions) within the depletion region.16 Consequently, temperature-dependent capacitance measurements may lead to misleading results.17 If used correctly, measuring the capacitance as a function of frequency or time can be used to quantify these mobile ions within the perovskite bulk.18 -- 23 Here we review the difference in studying mobile ions by capacitance measurements in the frequency and in the time domain. Based on the example of a PEABr0.2Cs0.4MA0.6PbBr3 quasi- 2D/3D perovskite layer in the inverted structure consisting of NiOX/perovskite/C60/BCP, we show how both measurements can be used quantify properties of mobile ions such as activation energy, diffusion coefficient, sign of charge, concentration, and the length of the ionic double layer. Furthermore, we show how a distinction between mobile ions and electronic defect states can be made when measuring the capacitance in the time domain, allowing both mobile ions and electronic defect states to be quantified. Capacitance measurement Most optoelectronic devices such as solar cells and light-emitting diodes can be approximated as a parallel-plate device with the active material sandwiched between the two contacts. The capacitance of such a parallel-plate device is given by where 𝜀 is the vacuum permittivity, 𝐴 is the active area of the device, and 𝑤- is the In case of full depletion, 𝑤- corresponds to the thickness of the active layer, i.e the thickness device 𝐶./0. The dielectric permittivity of the material 𝜀=𝜀′+𝑖𝜀′′ is a complex frequency- impedance response by applying a time-varying electric field with amplitude 𝑉* at frequency depletion layer width, plus a possible additional dielectric contribution of the contact layers. dependent tensor. To obtain the complex capacitance of the device we measure the of the perovskite layer, and the capacitance corresponds to the geometric capacitance of the 𝐶=##$%&' ( 1 ) 2 𝜔 and measuring the current response 𝐼* sin (𝜔𝑡+Φ), as indicated in Figure 1. The impedance Z is defined as 𝑍(𝑡)= ?$ @AB (CD) where 𝛷 is the phase delay between the voltage input and the current response, and the modulus 𝑍=𝑉/𝐼* is the ratio between the amplitude of the voltage and the current where 𝐶J and 𝐶′′ are coupled by the Kramers-Kroning relations.24 In semiconductor devices, E$ @AB (CDFG)=𝑍* 𝑒IAG=𝑍J−𝑖𝑍JJ 𝐶∗(𝜔)=(𝑖𝜔𝑍)IO= 𝐶J(𝜔)−𝑖𝐶JJ(𝜔), signal. The complex capacitance is given by ( 2 ) ( 3 ) the measurement of the capacitance thus measures the complex dielectric permittivity of the device. Figure 1. (a) Impedance measurements are based on applying a sinusoidal voltage with a certain frequency and Frequency versus time domain In a dielectric material with a Debye relaxation, i.e. a physical process exhibiting an measuring the current response. (b) Impedance expressed as the modulus Z0 and the phase angle Φ. exponential relaxation with a single time constant τ, the complex dielectric response where 𝜀V and 𝜀R are the static and asymptotic high frequency values of the dielectric constant, respectively.25 The real part (𝜀J) is a measure of the ability to store energy in the dielectric material and the imaginary part (𝜀JJ) is a measure of the dielectric loss (see Figure 2a). In the imaginary part one can see a distinctive peak at a frequency of ω=1/τ, 𝜀(𝜔)=𝜀J(𝜔)−𝑖𝜀JJ(𝜔)=𝜀R+#SI#TOFACU function is ( 4 ) corresponding to the maximum dielectric loss in the material, i.e. the irreversible transfer of energy from the external stimulus to the dielectric material. At lower frequencies, the 3 electrical displacement can follow the applied electric field, whereas at higher frequencies, the material has no time to respond. Alternatively, one can instead measure the transient response of the system following an abrupt change in driving field. Using Laplace transformation to convert Equation 4 from the frequency domain into the time domain then yields 𝜀(𝑡)=𝜀R+(𝜀V−𝜀R)Y1−expY−DU]] ( 5 ) which is illustrated in Figure 2b. Both the frequency and the time domain contain identical information. Figure 2. Debye dielectric material represented in (a) the frequency domain and (b) the time domain after a step function of applied electric field. Both the frequency and the time domain contain identical information. The physical origin of a Debye process leading to the dielectric permittivity shown in Figure 2 can, for example, be the thermal emission of trapped charge carriers or the migration of mobile ions within the perovskite bulk. The magnitude of 𝜀V−𝜀R is then directly related to the number of trapped charge carriers or mobile ions. In the case of electronic defect states within the band gap, the kinetics of charge-carrier capture and emission by these defect states can be described in the context of the Shockley- Read-Hall model.26,27 In this model, the time constant measured is due to thermal emission of trapped charge carriers (see Section 1 in the Supplementary Information (SI) for details). However, the model assumes that the measured time constant results solely from the occupation/emission of electronic defect states. If a significant density of mobile ions is present within the studied material, this analysis may yield misleading results as the migration of mobile ions typically dominates the slow capacitance signal, and it hence cannot be assigned to be solely from electronic defect states.17 4 In the following we show how this dominating signal can be used to study ion migration, both in the frequency domain with impedance spectroscopy, and in the time domain with transient ion drift. Finally, we suggest that the time domain can be used to also measure electronic defect states by suppressing the signal from mobile ions. Impedance spectroscopy Impedance spectroscopy measures the capacitance as a function of frequency to observe dielectric relaxation in a material as shown in Figure 2a. To illustrate the difference between impedance spectroscopy and transient ion drift, we measure a perovskite-based device illustrated in Figure 3a (details of the device are found in Section S2 in the SI and will be published elsewhere). We measure the modulus and the phase angle at short circuit in the dark (Figure 3b) to obtain the complex capacitance from Equation 3. Figure 3c shows the real part and Figure 3d the imaginary part of the complex capacitance. At low frequencies, the phase angle is close to -90°, showing that the device operates in a capacitor-like manner. At high frequencies, the contact resistance is dominating the measured impedance response resulting in a decrease in phase angle. The real part of the capacitance shows a plateau at medium frequencies (Figure 3c). Assuming that the perovskite layer is mostly depleted at short circuit, we obtain a dielectric constant of 7.3 ± 0.1 based on the mean of three measurements, which is in good agreement with the dielectric constant obtained for CsPbBr3.28 We observe a small increase in the real part of the capacitance 𝐶J at low frequencies which is accompanied with a peak in the imaginary part 𝐶JJ, similar to the dielectric relaxation shown in Figure 2a. The peak in the imaginary part shifts to lower frequencies with decreasing temperatures, which corresponds to the presence of a thermally-activated process. In perovskite-based devices such a behavior typically arises from the migration of mobile ions within the perovskite layer.29 We hence assume in the following that the change in capacitance arises from ion migration. Later we validate this assumption with transient ion drift. Due to the different work functions of the contacts, the perovskite layer is subject to an internal electric field in optoelectronic devices. This leads to an accumulation of mobile ions at the contact interfaces.30 Mobile ions that accumulated at the contact interfaces form a diffuse ionic double layer that strongly influence the current injection rates of perovskites- 5 𝜔=2𝜋𝑓=OU=-ab 𝑙-=d##$efg hbi then leads to the migration of mobile ions with a peak in the imaginary part of the complex capacitance with a maximum at the frequency based devices under operation.18 The application of an oscillating voltage with a frequency 𝜔 where 𝑓 is the measured angular frequency and 𝑙 the diffusion length. Often, the ion diffusion length is assumed to correspond to the Debye length 𝑙- given by where 𝑘k is Boltzmann's constant, 𝑇 the temperature, and 𝑁 the doping density.31 The where 𝜈 is the attempt frequency of an ionic jump and 𝑑 the jump distance. Δ𝐺, Δ𝑆, and Δ𝐻 are the change in Gibbs free energy, entropy, and enthalpy during the jump of a mobile ion. The activation enthalpy is often referred to as the activation energy 𝐸%. The attempt 𝐷=o$pbq expY−rsefg]=o$pbq expYrtef] expY−ruefg] diffusion coefficient can be expressed as32,33 ( 8 ) ( 6 ) ( 7 ) frequency is the frequency of an attempt to break or loosen a bond, related to the vibration frequency and often assumed to be in the order of 1012 s-1.34 Assuming that the attempt frequency is temperature independent, Equation 8 can be simplified to where 𝐷* is a temperature-independent pre-factor. Assuming furthermore that the diffusion ( 9 ) of mobile ions is primarily within the Debye layer in the vicinity to the interfaces, Equation 6 can be written as 𝐷=𝐷* expY−}~efg] 𝜏=#$#efg hbi-$expY(cid:128)(cid:129)efg]. ( 10 ) By plotting the measured frequency as function of temperature in an Arrhenius plot, the activation energy for ion migration can be obtained. 6 Figure 3. (a) Schematic representation of the device structure. (b) Modulus (Z) and phase angle (Φ) of the perovskite-based device measured at short circuit in the dark together with the corresponding (c) real and (d) imaginary part of the complex capacitance. The peak in the imaginary part corresponds to the temperature- dependent diffusion of mobile ions within the perovskite. The temperature dependence reveals the activation energy and the pre-factor. Note that also in the presence of electronic defect states, an increase in capacitance at low frequencies is expected,6 which cannot be distinguished from the effect of mobile ions. In addition, the diffusion coefficient can only be obtained if the diffusion length within the perovskite is known. Here, we assumed that the diffusion length is equal to the Debye layer. However, it has also been suggested that the migration of mobile ions extends throughout the perovskite bulk.35,36 Furthermore, it was shown that the interfaces play an important role as charge accumulation at interfaces leads to an additional increase in capacitance.37 -- 40 Capacitance measurements in the time domain offer the complementary information needed to assign the observed feature to the migration of mobile ions. Only when measuring the transient of the capacitance can the diffusion coefficient be obtained without prior knowledge of the diffusion length. Transient capacitance measurements furthermore allow to distinguish between effects caused by electronic defect states and mobile ions. Transient ion drift Transient ion-drift measurements are based on measuring the capacitance transient after applying a voltage bias.41 The voltage is chosen such that it collapses the depletion layer within the perovskite. The depletion layer width 𝑤- is approximated as42 where and 𝑉(cid:132)A the built-in potential. The depletion capacitance is directly related to the 𝑤-=d(cid:131) #$ #h i(𝑉(cid:132)A−𝑉) ( 11 ) depletion width as 7 𝐶pa=𝜀𝜀 %&'= dh #$ # i (cid:131) (?(cid:133)(cid:134)I?). ( 12 ) minority carriers. The diffusion capacitance is given as Applying a bias thus decreases the depletion layer width, increasing the measured capacitance (see Figure 4a). The capacitance of the device is obtained by an equivalent circuit model (see Section S3 in the SI for details). At higher voltages, the diffusion capacitance 𝐶p starts to dominate the measured capacitance, due to the injection of where 𝑁/ is the total equilibrium charge density at a given voltage 𝑉 and 𝑛 is the diode the diffusion capacitance, plotting 𝐶I(cid:131)(𝑉) allows to extract the doping density and the built- efgexpY h?Befg] 𝐶p=hbai(cid:135) ideality factor.43,44 In the case the depletion capacitance can clearly be distinguished from ( 13 ) in potential of the device (see Figure 4b).45,46 This is commonly referred to as the Mott- Schottky analysis. We obtain a built-in potential of 0.95 ± 0.05 V and a doping density of (7 ± 1) 1017 cm-3, based on the mean of three measurements. Figure 4. (a) Capacitance-voltage measurement and (b) Mott-Schottky plot of a perovskite-based device measured at 300 K in the dark at 10 kHz, illustrating different capacitive regimes. Only at a certain voltage regime can the depletion capacitance C(cid:138)(cid:139) clearly be identified. At high voltages, the diffusion capacitance C(cid:138) starts to dominate the measured capacitance. The linear fit reveals the built-in potential and the doping density. The capacitance was calculated assuming a capacitor in series with a resistor (see Section S3 in the SI for details). To measure the transient of the capacitance we apply a bias of 1.25 V, which completely collapses the depletion layer. We note that we are already in the diffusion capacitance regime when applying a 1.25 V bias voltage. However, the initial capacity change caused by the discharge due to the accumulation of minority carriers is much faster than the time resolution of our instrument. The measured capacitance rise and decay are shown in Figure 8 5, measured at 10 kHz where the phase angle is close to -90°, such that the measured capacitance is not affected by the series resistance of the device. Applying a forward bias decreases the depletion width of the device. Mobile ions are now able to diffuse within the perovskite bulk which eventually leads to a uniform ion distribution within the previous depleted region. After removing the bias, the depletion width increases quickly by the movement of electric charge carriers, depleting most of the perovskite bulk according to Equation 11. Mobile ions within the depleted region will drift towards the interfaces following the internal electric field, changing the depletion width as h (i±i(cid:141)(cid:142)(cid:143)(D))(𝑉(cid:132)A−𝑉) ( 14 ) where 𝑁E0B(𝑡) is the density of mobile ions within the depletion region. For ions with the 𝑤-(𝑡)=d (cid:131) #$ # same charge as minority carriers the sign of the capacitance change is positive and for ions with the same charge as majority carriers the sign is negative. The capacitance as a function of time can thus be written as Figure 5. (a) Capacitance rise and (b) decay of the perovskite-based device measured at 10 kHz where the measured capacitance corresponds to the change in depletion capacitance of the device. The capacitance was calculated assuming a capacitor in parallel with a resistor. 𝐶(𝑡)=𝐶(∞)Y1±i(cid:141)(cid:142)(cid:143)(D)i ](cid:145)b≈𝐶(∞)Y1±i(cid:141)(cid:142)(cid:143)(D) (cid:131)i ] where 𝐶(∞) is the junction capacitance at steady state. The approximation is valid as long as the density of the mobile ions is much lower than the doping density. To find 𝑁E0B(𝑡) we ( 15 ) assume that the thermal diffusion is negligible against drift and solve the drift equation following the work of Heiser et al.20,41 The temporal evolution of the mobile ions drifting towards the interface can then be described by (cid:147)D =−(cid:147)(cid:147)(cid:148) 𝑁E0B(𝑡)𝜇𝐸 (cid:147)i(cid:141)(cid:142)(cid:143)(D) 9 ( 16 ) Assuming that the electric field is static and varies linearly within the depletion region as where 𝜇 is the mobility of mobile ions and 𝐸 is the electric field in the depletion region. for 𝑁E0B(𝑡). Equation 15 can then be written as with the time constant τ given by 𝜏=&'(cid:151)(cid:128). Δ𝐶 is the capacitance change due to the drift of mobile ions towards the interface, directly 𝐸(𝑥)=𝐸Y1− (cid:148)&'], 𝐶(𝑡)=𝐶(∞)±Δ𝐶expY−DU] and that the drift of mobile ions is not affecting the electric field, we can solve Equation 16 ( 18 ) ( 17 ) ( 19 ) related to the mobile ion density as Δ𝐶=𝐶(∞)−𝐶(0)=𝐶(∞)i(cid:141)(cid:142)(cid:143)(cid:131)i . 𝐸=h &'i#$# , 𝜏=efg#$# hb-i ( 20 ) ( 21 ) ( 22 ) By expressing the electric field as a function of the doping density as together with the Einstein relation, the time constant can be written as with the diffusion coefficient given by Equation 9. This equation can now be used to extract the diffusion coefficient and activation energy from the measured capacitance transients. Note that this equation is the same as Equation 10, which describes the time constant for ion migration in the frequency domain. Note that we have assumed a linear electric field within the depletion region and that the electric field is unaffected by the drift of mobile ions, which is only true if the density of mobile ions is small compared to the background doping density. In the case of a mobile ion density close to the background doping density, the transient would have a non-exponential behavior making the analysis more complex.20 We have furthermore assumed that the total ion concentration is conserved, i.e. mobile ions are not diffusing into the contact layers. To distinguish mobile ions from electronic defect states we use the difference in rise and decay time of the capacitance change. In the case of mobile ions, the rise time (Figure 5a) due to diffusion of mobile ions is expected to be longer than the decay time due to drift of mobile ions (Figure 5b). In contrast, the capture-rate for electronic defect states is much 10 higher than the emission rate, resulting in a much faster rise time compared to the decay time. Measuring the rise and the decay of the capacitance thus allows to distinguish between capacitance changes due to electronic defect states and mobile ions (see Section S4 in the SI for details). A unique feature of transient ion-drift measurements is that one can differentiate between mobile cations and mobile anions. Here, the negative capacitance change in Figure 5b measures an increase in depletion width due to ions migrating towards the contact interfaces. The Mott-Schottky plot shows a p-type behavior of the perovskite (Figure 4), the negative capacitance change in Figure 5b hence corresponds to the migration of an anion. We thus attribute the measured changes in capacitance due to bromide migration, presumably due to vacancy-mediated migration.47,48 Figure 6. Arrhenius plot of the measured thermal emission rates by (a) impedance spectroscopy and (b) transient ion drift (TID). (c) Activation energies obtained from the two measurement techniques based on the mean out of three measurements each. The Arrhenius plots together with the obtained activation energies for both impedance spectroscopy and transient ion-drift measurements are shown in Figure 6. For impedance spectroscopy and transient ion-drift measurements we obtain very similar activation energies of 0.13 ± 0.01 and 0.14 ± 0.01 eV, respectively. The obtained activation energy is close to theoretical predictions and experimental observations for the migration of bromide in MAPbBr3 and CsPbBr3 (0.09 to 0.25 eV).49 -- 51 From the capacitance transients we obtain a concentration of (5.1 ± 2.5) x 1016 cm-3 and a diffusion coefficient of (3.1 ± 0.4) x 10-11 cm2/s for the mobile anions. We can now use this coefficient together with the frequency of the dielectric loss peak (Figure 3d) to obtain a diffusion length for ion migration of 6.2 ± 0.4 nm from Equation 6. The calculated diffusion length is larger than the Debye length of 3.9 ± 0.3 nm calculated from Equation 7. Hence, the 11 assumption that the diffusion of mobile ions during frequency-dependent capacitance measurements is limited to the Debye layer is not fulfilled. This underestimation of the diffusion length explains the difference between impedance spectroscopy and transient ion-drift measurements in Figure 6a. Note that we assume, for the purpose of the comparison between transient ion drift and impedance measurements, that the ion migration is the same near the interface and through the bulk. Conclusion Capacitance techniques must be applied with caution to mixed ionic-electronic conductors, because the measured capacitance features can be caused by mobile ions rather or electronic defect states. We have shown that transient ion-drift measurements have the virtue that a distinction between electronic defect states and mobile ions can be made. In addition, transient ion drift allows fast and non-destructive quantification of activation energy, diffusion coefficient, sign of charge, and concentration of mobile ions in perovskite- based devices. Using the diffusion coefficient obtained by transient ion drift, the length of the ionic double layer can be determined from impedance spectroscopy measurements. the way to ion migration Since the migration of mobile ions is a key degradation mechanism in perovskite-based devices, reducing ion migration is crucial for the fabrication of stable devices. Capacitance techniques provide a tool to systematically investigate the effects of different passivation in full agents, fabrication methods, and perovskite compositions on perovskite-based devices guiding for commercialisation. Acknowledgements The authors thank Esther Alarcón Lladó for carefully reading and commenting on the manuscript. This work is part of the research program of the Netherlands Organization for Scientific Research (NWO). D.N.C. and M.K.G. acknowledge the support of the Rowland Fellowship at the Rowland Institute at Harvard University. long-lasting devices, critical 12 References 1 G.L. Miller, D. V Lang, and L.C. Kimerling, Annu. Rev. Mater. Sci. 7, 377 (2003). 2 J. Heath and P. Zabierowski, in Adv. Charact. Tech. Thin Film Sol. Cells (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, 2016), pp. 93 -- 119. 3 J. V. Li and G. Ferrari, Capacitance Spectroscopy of Semiconductors (Jenny Stanford Publishing, 2018). 4 E. Von Hauff, J. Phys. Chem. C 123, 11329 (2019). 5 D. V. Lang, J. Appl. Phys. 45, 3023 (1974). 6 T. Walter, R. Herberholz, C. Müller, and H.W. Schock, J. Appl. Phys. 80, 4411 (1996). 7 A. Miyata, A. Mitioglu, P. Plochocka, O. Portugall, J.T.W. Wang, S.D. Stranks, H.J. Snaith, and R.J. Nicholas, Nat. Phys. 11, 582 (2015). 8 D. Shi, V. Adinolfi, R.; Comin, M.J. Yuan, E. Alarousu, A.; Buin, Y. Chen, S. Hoogland, A.; Rothenberger, K. Katsiev, Y.; Losovyj, X. Zhang, P.A.; Dowben, O.F.; Mohammed, E.H.; Sargent, O.M. Bakr, R.; Comin, A.; Buin, Y. Chen, S. Hoogland, A.; Rothenberger, K. Katsiev, Y.; Losovyj, X. Zhang, P.A.; Dowben, O.F.; Mohammed, E.H.; Sargent, and O.M. Bakr, Science 347, 519 (2015). 9 S. Yakunin, M. Sytnyk, D. Kriegner, S. Shrestha, M. Richter, G.J. Matt, H. Azimi, C.J. Brabec, J. Stangl, M. V Kovalenko, and W. Heiss, Nat. Photonics 9, 444 (2015). 10 S.D. Stranks and H.J. Snaith, Nat. Nanotechnol. 10, 391 (2015). 11 Y. Zhao, H. Tan, H. Yuan, Z. Yang, J.Z. Fan, J. Kim, O. Voznyy, X. Gong, L.N. Quan, C.S. Tan, J. Hofkens, D. Yu, Q. Zhao, and E.H. Sargent, Nat. Commun. 9, 1607 (2018). 12 Y. Chen, N. Li, L. Wang, L. Li, Z. Xu, H. Jiao, P. Liu, C. Zhu, H. Zai, M. Sun, W. Zou, S. Zhang, G. Xing, X. Liu, J. Wang, D. Li, B. Huang, Q. Chen, and H. Zhou, Nat. Commun. 10, 1112 (2019). 13 S. Heo, J. Lee, J.-B. Park, K. Kim, D.-J. Yun, Y.S. Kim, J.K. Shin, G. Seo, T.K. Ahn, Y. Lee, M.K. Nazeeruddin, D. Lee, M. Seol, J. Lee, J.-B. Park, K. Kim, D.-J. Yun, Y.S. Kim, J.K. Shin, T.K. Ahn, and M.K. Nazeeruddin, Energy Environ. Sci. 10, 1128 (2017). 14 J.W. Rosenberg, M.J. Legodi, Y. Rakita, D. Cahen, and M. Diale, J. Appl. Phys. 122, 145701 (2017). 15 A.Y. Polyakov, N.B. Smirnov, I. V. Shchemerov, D.S. Saranin, T.S. Le, S.I. Didenko, D. V. Kuznetsov, A. Agresti, S. Pescetelli, F. Matteocci, and A. Di Carlo, Appl. Phys. Lett. 113, 13 263501 (2018). 16 J.M. Frost and A. Walsh, Acc. Chem. Res. 49, 528 (2016). 17 O. Almora, M. García-Batlle, and G. Garcia-Belmonte, J. Phys. Chem. Lett. 10, 3661 (2019). 18 H. Wang, A. Guerrero, A. Bou, A.M. Al-Mayouf, and J. Bisquert, Energy Environ. Sci. 12, 2054 (2019). 19 M.H. Futscher, J.M. Lee, L. McGovern, L.A. Muscarella, T. Wang, M.I. Haider, A. Fakharuddin, L. Schmidt-Mende, and B. Ehrler, Mater. Horizons 6, 1497 (2019). 20 T. Heiser and E. Weber, Phys. Rev. B 58, 3893 (1998). 21 P. Lopez-Varo, J.A. Jiménez-Tejada, M. García-Rosell, S. Ravishankar, G. Garcia-Belmonte, J. Bisquert, and O. Almora, Adv. Energy Mater. 8, 1702772 (2018). 22 A. Pockett, G.E. Eperon, N. Sakai, H.J. Snaith, L.M. Peter, and P.J. Cameron, Phys. Chem. Chem. Phys. 19, 5959 (2017). 23 S. Reichert, J. Flemming, Q. An, Y. Vaynzof, J.-F. Pietschmann, and C. Deibel, (2019). 24 R. de L. Kronig, J. Opt. Soc. Am. 12, 547 (1926). 25 P. Debye, Trans. Faraday Soc. 30, 679 (1934). 26 W. Shockley and W.T. Read, Phys. Rev. 87, 835 (1952). 27 R.N. Hall, Phys. Rev. 87, 387 (1952). 28 Z. Yang, A. Surrente, K. Galkowski, A. Miyata, O. Portugall, R.J. Sutton, A.A. Haghighirad, H.J. Snaith, D.K. Maude, P. Plochocka, and R.J. Nicholas, ACS Energy Lett. 2, 1621 (2017). 29 O. Almora, I. Zarazua, E. Mas-Marza, I. Mora-Sero, J. Bisquert, and G. Garcia-Belmonte, J. Phys. Chem. Lett. 6, 1645 (2015). 30 J. Bisquert, G. Garcia-Belmonte, and A. Guerrero, in Org. Halide Perovskite Photovoltaics from Fundam. to Device Archit. (Springer International Publishing, Cham, 2016), pp. 163 -- 199. 31 E. Hückel, in Ergebnisse Der Exakten Naturwissenschaften (Springer Berlin Heidelberg, Berlin, Heidelberg, 1924), pp. 199 -- 276. 32 G.H. Vineyard, J. Phys. Chem. Solids 3, 121 (1957). 33 D. Meggiolaro, E. Mosconi, and F. De Angelis, ACS Energy Lett. 4, 779 (2019). 34 J. Koettgen, T. Zacherle, S. Grieshammer, and M. Martin, Phys. Chem. Chem. Phys. 19, 9957 (2017). 35 W. Peng, C. Aranda, O.M. Bakr, G. Garcia-Belmonte, J. Bisquert, and A. Guerrero, ACS 14 Energy Lett. 3, 1477 (2018). 36 M. Bag, L.A. Renna, R.Y. Adhikari, S. Karak, F. Liu, P.M. Lahti, T.P. Russell, M.T. Tuominen, and D. Venkataraman, J. Am. Chem. Soc. 137, 13130 (2015). 37 H.-S. Kim, I.-H. Jang, N. Ahn, M. Choi, A. Guerrero, J. Bisquert, and N.-G. Park, J. Phys. Chem. Lett. 6, 4633 (2015). 38 I. Zarazua, G. Han, P.P. Boix, S. Mhaisalkar, F. Fabregat-Santiago, I. Mora-Seró, J. Bisquert, and G. Garcia-Belmonte, J. Phys. Chem. Lett. 7, 5105 (2016). 39 A. Guerrero, G. Garcia-Belmonte, I. Mora-Sero, J. Bisquert, Y.S. Kang, T.J. Jacobsson, J.-P. Correa-Baena, and A. Hagfeldt, J. Phys. Chem. C 120, 8023 (2016). 40 S.A.L. Weber, I.M. Hermes, S.-H. Turren-Cruz, C. Gort, V.W. Bergmann, L. Gilson, A. Hagfeldt, M. Graetzel, W. Tress, and R. Berger, Energy Environ. Sci. 11, 2404 (2018). 41 T. Heiser and A. Mesli, Appl. Phys. A Solids Surfaces 57, 325 (1993). 42 K.K.N. S. M. Sze, Physics of Semiconductor Devices (John Wiley & Sons Inc, 2014). 43 J. Bisquert, Phys. Chem. Chem. Phys. 5, 5360 (2003). 44 S.K. Sharma, D. Pavithra, G. Sivakumar, N. Srinivasamurthy, and B.L. Agrawal, Sol. Energy Mater. Sol. Cells 26, 169 (1992). 45 O. Almora, C. Aranda, E. Mas-Marzá, and G. Garcia-Belmonte, Appl. Phys. Lett. 109, 173903 (2016). 46 M. Fischer, K. Tvingstedt, A. Baumann, and V. Dyakonov, ACS Appl. Energy Mater. acsaem.8b01119 (2018). 47 Y. Luo, P. Khoram, S. Brittman, Z. Zhu, B. Lai, S.P. Ong, E.C. Garnett, and D.P. Fenning, Adv. Mater. 29, 1703451 (2017). 48 A. Senocrate, I. Moudrakovski, G.Y. Kim, T.-Y. Yang, G. Gregori, M. Grätzel, and J. Maier, Angew. Chemie Int. Ed. 56, 7755 (2017). 49 J.M. Azpiroz, E. Mosconi, J. Bisquert, and F. De Angelis, Energy Environ. Sci. 8, 2118 (2015). 50 J. Mizusaki, K. Arai, and K. Fueki, Solid State Ionics 11, 203 (1983). 51 H. Cho, C. Wolf, J.S. Kim, H.J. Yun, J.S. Bae, H. Kim, J.M. Heo, S. Ahn, and T.W. Lee, Adv. Mater. 29, 1700579 (2017). 15 SUPPLEMENTARY INFORMATION FOR Quantifying mobile ions in perovskite-based devices with temperature-dependent capacitance measurements: frequency versus time domain Moritz H. Futscher1, Mahesh K. Gangishetty2, Daniel N. Congreve2 and Bruno Ehrler1 1. AMOLF, Center for Nanophotonics, Science Park 104, 1098 XG Amsterdam, The Netherlands 2. Rowland Institute at Harvard, 100 Edwin H. Land Blvd, Cambridge, Massachusetts, United States Corresponding Author [email protected] S1 S1 Shockley-Read-Hall model In the Shockley-Read-Hall model, the thermal emission rate of trapped electrons is given by 1𝜏=𝑒%=𝜎%𝜈()𝑁+exp/−𝐸2𝑘4𝑇6 e7=A(T) exp/−E>k@T6 where 𝜎% is the capture cross section for capturing electrons, 𝜈() the thermal velocity, 𝑁+ the density of states in the conduction band, 𝐸2 the depth of the trap from the conduction band, 𝑘4 Boltzmann's constant, and 𝑇 the temperature.1,2 Taking the temperature where 𝐴=𝐴B𝑇C is the attempt-to-escape frequency. An analogous expression holds for the dependence of the pre-factors into account, this can be simplified to thermal emission rate of trapped holes. By measuring the time constant as a function of temperature, the depth of the trap and the attempt-to-escape frequency can be obtained. When the time constant is measured from the transient of the capacitance, this technique is called deep-level transient spectroscopy (DLTS) originally developed by Lang et al.,3 and when the time constant is measured from the frequency-dependence of the capacitance, this technique is called thermal admittance spectroscopy (TAS) originally developed by Walter et al.4 S2 Experimental Device fabrication. ITO substrates were cleaned by sonicating sequentially in detergent (Micron - 90), water (twice) and acetone (twice) for 10 minutes each, before soaking them in boiling isopropanol for 20 minutes to remove the leftover organic contamination. The substrates were then treated with O2 plasma for 5 minutes using 0.5 Torr O2 gas at 500 W. On top of these cleaned substrates, 30 µL precursors solution of NiOX was spin coated at 2500 rpm with an acceleration of 2500 rpm/s. The films were then immediately transferred to a hot plate, which was kept at 100 oC and annealed at 300 oC for 3 hours. After cooling the NiOX films, a thin layer of perovskite was made by spin coating the perovskite precursor solution at 1000 rpm for 10 seconds and ramped up to 3000 rpm for 45 seconds. After 20 seconds, 90 μL of chloroform (anti-solvent) was dripped on the spinning perovskite layer. The perovskite precursor was prepared by mixing appropriate ratios of 0.3 M of MABr, CsBr, PbBr3 and PEABr to get PEABr0.2Cs0.4MA0.6PbBr3 in DMF/DMSO. The detailed preparation of S2 precursor solution is reported in our previous work.5,6 After coating the perovskite layer, the films were taken into an evaporation chamber where, 50 nm C60, 5 nm BCP, and 60 nm Ag were sequentially deposited at 10-6 mbar. Electrical measurements. All measurements were performed in the dark in a Janis VPF-100 liquid nitrogen cryostat at pressures below 2 x 10-6 mbar. Samples were loaded into the cryostat inside a nitrogen-filled glovebox to avoid air exposure. Capacitance-voltage characteristics were measured at 10 kHz with an AC frequency of 50 mV from 1.45 to -0.25 V with a step size of 0.01 V. Impedance spectroscopy measurements were performed with an AC voltage of 50 mV at short circuit between 240 and 330 K in steps of 10 K. Transient ion- drift measurements were performed at 10 kHz with an AC voltage of 20 mV between 240 and 330 K in steps of 2 K. The transients were averaged over 20 separate measurements. The temperature accuracy was set to 0.2 K. S3 Equivalent circuit model The complex impedance of a parallel-plate device can be modelled with a resistor in parallel and a resistor in series to a capacitor, as shown in Figure S1. Figure S1. Circuit model of a resistor in parallel (RP) and a resistor in series (RS) with a capacitor (C). If the capacitance is measured at one frequency, however, the complex impedance must be obtained with a model consisting only of two free parameters, i.e. one resistor and one capacitor. Time-dependent and voltage-dependent capacitance measurements can thus be modelled with a resistor either in parallel or in series to the capacitor as shown in Figure S2 and Table S1. At high impedance the series resistance can be neglected, at low impedance the shunt resistance can be neglected. As long as the device does not suffer from resistive losses, however, the capacitance is identical for both models. S3 Figure S2. Circuit model of a capacitor (C) (a) in parallel and (b) in series with a resistor (R). (c) Capacitance as a function of voltage of a perovskite-based device calculated assuming a capacitor in series or in parallel with a resistor measured at 300 K in the dark at 10 kHz. To measure the transient of the capacitance, we neglect the series resistance assuming a resistor in parallel with a capacitor. We thus choose the measuring frequency such that the impedance is not limited by the series resistance of the device, i.e. the phase angle has to be close to -90°. Table S1. Comparison of capacitance calculated with a capacitor either in series or in parallel with a resistor. Impedance Capacitance Series 𝑅+ 1𝑖𝜔𝐶 1𝜔 Im(𝑍) − Parallel /𝑅+ 1𝑖𝜔𝐶6IJ 1𝜔Im/1𝑍6 Due to charge-carrier injection at forward biases, the series resistance can often no longer be neglected. In this case, the capacitance can be calculated as a function of voltage assuming a capacitor in series with a resistor. Figure S2c shows the difference in capacitance as a function of voltage assuming a capacitor in series or in parallel with a resistor. At low biases, the difference between the two models is negligible. At high biases, there is a growing difference in capacitance obtained from the two models. In our case, however, the differences between the two models in the values obtained with the Mott-Schottky analysis are small, i.e. 0.89 and 0.87 V for the built-in potential and 8.0 x 1017 and 8.1 x 1017 cm-3 for the doping density, for a capacitor in parallel and in series with a resistor, respectively. S4 S4 Rise and decay time The characteristic rise and decay times for both mobile ions and electronic defect states are shown in Table S2. For mobile ions, the rise time is assumed to be due to diffusion from the interfaces into the perovskite bulk. The decay time is assumed to be due to drift of mobile ions from the bulk towards the interfaces, i.e. 𝜏=NOPQ, where 𝑤S is the depletion width and 𝜇 is the mobility of mobile ions. The electric field is approximated as 𝐸= UNO. The rise and decay time follow from the Shockley-Read-Hall model. Consequently, electronic defect states and mobile ions show different ratios between rise and decay time, so that a distinction can be made between them. Table S2. Comparison of rise and decay times of electronic defect states and mobile ions at room temperature assuming a defect state with a trap energy between 0.2 and 0.6 eV. 𝐷 is the diffusion coefficient of mobile ions, 𝑞 the elementary charge, and 𝑉Z[ the built-in potential. J\]^_'≈10IJc− 10IJB s J\]^_'expeQfgh2i≈10IJB− 10B s Electronic defect states Mobile ions NOdS ≈10IJ− 10J s SjUkl≈10Ic− 10B s NOdgh2 Rise time Decay time Figure S3 shows measured capacitance transients after a filling voltage of 1.25 V for different periods of time. Applying a voltage pulse for 2 seconds allows mobile ions to be measured, since mobile ions have sufficient time to diffuse from the interfaces into the perovskite bulk (Figure S3a). On the other hand, applying a voltage bias for 0.2 milliseconds only fills electronic defect states present in the material, as ions have not enough time to react (Figure S3c). When applying a bias for 20 milliseconds both ions and electronic defects states are measured (Figure S3b). The Arrhenius plot of the measured thermal emission rates for electronic defect states is shown in Figure S3d. We obtain an activation energy of 0.08 ± 0.01 eV, a concentration of 3 x 1015 cm-3 and an attempt-to-escape frequency of 1.3 ± 0.4 1013 s-1. The obtained concentration of electronic defect states is much lower than the concentration of mobile ions. We therefore assume that the presence of electronic defect states is not influencing the our measurements of mobile ions. S5 Figure S3. Capacitance transient measurements between 240 and 330 K in steps of 3 K measured at 0 V at 10 kHz. Capacitance transients are measured after applying a bias of 1.25 V for (a) 2 seconds, (b) 20 milliseconds, and (c) 0.2 milliseconds. (d) Arrhenius plot of the thermal emission rates from (c). The linear fit reveals the activation energy and the attempt-to-escape frequency of electronic defect states. Reference 1. Shockley, W. & Read, W. T. Statistics of the recombinations of holes and electrons. Phys. Rev. 87, 835 -- 842 (1952). Hall, R. N. Electron-hole recombination in germanium. Phys. Rev. 87, 387 (1952). Lang, D. V. Deep-level transient spectroscopy: A new method to characterize traps in semiconductors. J. Appl. Phys. 45, 3023 -- 3032 (1974). 2. 3. 4. Walter, T., Herberholz, R., Müller, C. & Schock, H. W. Determination of defect distributions from admittance measurements and application to Cu(In,Ga)Se2 based heterojunctions. J. Appl. Phys. 80, 4411 -- 4420 (1996). Gangishetty, M. K., Hou, S., Quan, Q. & Congreve, D. N. Reducing architecture limitations for efficient blue perovskite light-emitting diodes. Adv. Mater. 30, 1706226 (2018). Gangishetty, M. K., Sanders, S. N. & Congreve, D. N. Mn2+ doping enhances the brightness, efficiency, and stability of bulk perovskite light-emitting diodes. ACS Photonics 6, 1111 -- 1117 (2019). S6 5. 6.
1808.06595	1	1808	2018-06-15T05:38:33	Storage and retrieval of electromagnetic waves using electromagnetically induced transparency in a nonlinear metamaterial	[ "physics.app-ph", "physics.optics", "quant-ph" ]	We investigate the storage and retrieval of electromagnetic waves using a nonlinear metamaterial, analogous to the electromagnetically induced transparency (EIT) observed in atomic systems. We experimentally demonstrate the storage of the electromagnetic wave by reducing an auxiliary "control" wave; the stored wave is then released by recovering the control wave. We also confirm that the metamaterial can store and reproduce the phase distribution of the original input wave. These effects confirm a remarkable analogy between the metamaterial and an atomic EIT medium.	physics.app-ph	physics	Storage and retrieval of electromagnetic waves using electromagnetically induced transparency in a nonlinear metamaterial Toshihiro Nakanishi1, a) and Masao Kitano1 Department of Electronic Science and Engineering, Kyoto University, Kyoto 615-8510, Japan (Dated: 21 August 2018) We investigate the storage and retrieval of electromagnetic waves using a nonlinear metamaterial, analogous to the elec- tromagnetically induced transparency (EIT) observed in atomic systems. We experimentally demonstrate the storage of the electromagnetic wave by reducing an auxiliary "control" wave; the stored wave is then released by recovering the control wave. We also confirm that the metamaterial can store and reproduce the phase distribution of the original input wave. These effects confirm a remarkable analogy between the metamaterial and an atomic EIT medium. Electromagnetically induced transparency (EIT) is a non- linear optical phenomenon whereby an opaque medium is made transparent for a probe light by the incidence of an aux- iliary "control" light in an extremely narrow bandwidth1,2. In the frequency regions with high transparency, the group ve- locity of the probe light is dramatically reduced. Character- istic features of EIT, including a narrow-band transparency and slow propagation, have been extensively investigated in quantum systems composed of three-level atoms. Slow-light propagation in an EIT medium can be exploited for "optical memory" applications, or for storing and retrieving light by dynamically modulating the control light, whose intensity is proportional to the group velocity of the probe light3 -- 6. Narrow-band transparency and slow propagation can also be realized in other systems such as an optical waveguide cou- pled with optical cavities7 -- 9 and optomechanical systems10 -- 12, to name but a few13,14. Metamaterials composed of artificially designed structures can also mimic the EIT effects observed in atomic systems. Since the experimental demonstration of the EIT-like effect in the microwave frequency regime15,16, var- ious types of the EIT-like metamaterials have been reported to operate in different frequency ranges17 -- 20. The EIT-like metamaterials can serve various applications, including high- accuracy sensing21,22, nonreciprocal propagation23,24, and nonlinearity enhancement25. For more practical applications, property-tunable metamaterials have been realized using non- linear diodes26, superconductors27,28, air-discharge plasma29, photocarrier excitation in semiconductors30,31, and microelec- tromechanical systems32. The storage of electromagnetic waves in a metamaterial requires the rapid control of the group velocity. The first demonstration of electromagnetic-wave storage was realized using an EIT-like metamaterial with variable capacitors33. The metamaterial had similar functionality to the original EIT medium with three-level atoms, but the transparency or the group velocity was controlled by a bias voltage applied to the variable capacitors, rather than by the electromagnetic waves. In this sense, the effect can be regarded as displaying "static- electric-field-induced transparency." The bias voltage on each constituent, or "meta-atom," should be applied via a bias cir- cuit because a static field cannot propagate in free space. a)E-mail: [email protected] This constitutes a critical obstacle to increasing the number of meta-atoms or extending the frequency range. Here, we focus on a metamaterial showing genuine EIT ef- fect, whereby transparency results from the incidence of an auxiliary electromagnetic wave (the control wave). We previ- ously reported the static control of the EIT effect by using con- trol waves via three-wave mixing in varactor diodes included within the meta-atoms34. In this paper, we demonstrate the storage and retrieval of electromagnetic waves in the true EIT metamaterial by dynamically modulating the control waves in the same way as in the original atomic EIT systems. First, we review a conventional method used to mimic the EIT-related phenomena, namely a narrow-band transparency and slow propagation. EIT-like metamaterials are well de- scribed using a coupled-resonator model14. It is intriguing that Fano resonances35,36 can also be explained by this clas- sical model37,38, because the EIT effect can be regarded as a special case of Fano resonances, where the spectral line- shape becomes symmetric. One of the resonant modes, called a radiative mode, is responsible for directly interacting with propagating electromagnetic waves. The other resonant mode, called a trapped mode, is uncoupled from the external field and is excited only via coupling with the radiative mode. If the resonant frequencies of the radiative and trapped modes are close to the frequency of the incident waves and there is some degree of coupling between them, the waves received through the radiative mode are effectively transferred to the trapped mode. The electromagnetic waves are temporarily stored in the trapped mode and are subsequently re-emitted through the radiative mode. Owing to the high quality fac- tor of the trapped mode, dissipation and scattering are sub- stantially suppressed. In addition, the temporary storage in the trapped mode causes a propagation delay. Consequently, transparency and slow propagation in the metamaterial are re- alized. If the coupling between two resonant modes depends on a controllable parameter, the energy flow between the ra- diative and trapped mode can be dynamically controlled. By removing the coupling during the interaction with the exter- nal waves, the energy in the trapped mode is completely cap- tured in the metamaterial, and thus, storage of electromagnetic waves can be achieved33. Next, we introduce a metamaterial showing genuine EIT, as realized in atomic EIT systems. The metamaterial design, shown in Fig. 1(a), is the same as in our previous work34. The directions of the electric field E and magnetic field B of the 3mm 20mm 15 1mm 24.5mm varactor E 80mm (a) unit cell B metal e v e w a b p r o varactor radiative mode (b) e v r o l w a n t o c e v e w a b p r o e v r o l w a n t o c trapped mode (d) control mode (c) FIG. 1. radiative, (c) control, and (d) trapped modes. (a) Structure of a meta-atom and current flow for the (b) incident waves are also displayed. The dimension of the unit structure is 120 mm × 25 mm. The structure, made of copper, is fabricated on a dielectric material of thickness 0.8 mm and permittivity of 3.3. Two nonlinear capacitors, called varactor diodes, whose capacitances are functions of the voltage across the diode, are inserted with opposite directions on the two lat- eral circuit paths. The metamaterial has three resonant modes: a radiative mode, a trapped mode, and an additional mode called the control mode. The current flow for these modes is illustrated using the dashed lines with arrows in Figs. 1 (b) -- (d). The functions of the radiative and trapped modes are the same as those in the conventional EIT-like metamaterials dis- cussed above, but the resonant frequencies, fr for the radiative mode and ft for the trapped mode, are notably different. The probe wave (with frequency νp) and the control wave (with frequency νc) can directly excite the radiative mode (Fig. 1(b)) and the control mode (Fig. 1(c)), respectively. On the other hand, the trapped mode cannot be excited by an incident wave with one frequency only. The transmission of the control wave can be made relatively strong by adjusting νc to be located in the tail of the control-mode resonance at fc. In the absence of the control wave, the transmission of the probe wave (with νp = fr) is low because of the high radiation loss of the radiative mode. On the other hand, in the presence of an intense control wave, the probe wave becomes mixed with the control wave to produce other components with the frequencies νp ± νc via three-wave mixing in the nonlinear capacitors. If the resonant frequency of the trapped mode ft satisfies the condition ft = νp +νc or ft = νp −νc, the energy received in the radiative mode becomes effectively transferred into the trapped mode owing to the resonance of the generated components. (The proposed metamaterial is designed to sat- isfy ft = νp − νc.) The reverse process also occurs, and the radiative and trapped modes become coupled via three-wave mixing with the control wave. As a result, the metamaterial becomes transparent for the probe wave. 2 Figure 2(a) shows a schematic diagram of the experimen- tal setup for performing the transmission measurements and for demonstrating the storage of electromagnetic waves. The resonant frequencies of the radiative, trapped, and control modes are fr = 1.05 GHz, ft = 0.65 GHz, fc = 0.4 GHz, respectively34. The probe and control waves are combined and sent to an open-type waveguide, in which the meta-atoms are placed. The waveguide, designed as a stripline with a height of 25 mm and a width of 122 mm, supports the prop- agation of a quasi-TEM mode39. A unit comprising a direc- tional coupler and a 5 dB attenuator is inserted at each port of the waveguide. The attenuators serve to suppress undesired multiple reflections. We utilize the directional couplers fol- lowed by frequency filters, which transmit the probe wave, in order to monitor the transmitted and reflected signals for the probe wave. First, we observed transmission spectra for the probe wave in the presence or absence of the control wave for a single- (a) combiner -10dB directional coupler F P H F P L T T A control wave probe wave waveguide to oscilloscope to oscilloscope T T A meta-atoms to oscilloscope LPF HPF -10dB directional coupler (b) without control with control 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 i i n o s s m s n a r t 0.1 0.8 0.9 1 1.1 1.2 1.3 1.4 / GHz FIG. 2. (a) Overview of the experimental setup. Directional couplers pick up 10% of the transmitted or reflected signal. ATT: 5dB attenua- tor. LPF: low-pass filter with an insertion loss of 10 dB at 1.07 GHz. HPF: high-pass filter with an insertion loss of 10 dB at 0.865 GHz. (b) Transmission spectra for the probe wave in the presence or ab- sence of the control wave. layer metamaterial. A network analyzer was introduced to acquire spectra for the transmission between points N1 and N2 in Fig. 2(a), as the probe waves were swept from 0.8 to 1.4 GHz. The transmission spectra were normalized by a sig- nal transmitted through the empty waveguide. The transmis- sion spectrum without the control wave is shown by a dashed line in Fig. 2(b). A broad transmission dip is observed around the resonant frequency of the radiative mode. The solid line in Fig. 2(b) represents the transmission spectrum in the pres- ence of a continuous control wave with a fixed frequency of νc = 0.45 GHz, which is located in the tail of the resonance of the control mode centered at fc = 0.4 GHz and gives a transmission of 88% (not shown in the figure). The amplitude of the control wave is estimated to be 1.2 V at the waveguide input. The observed spectrum clearly exhibits a sharp trans- parency window around 1.035 GHz. In this case, the intense control wave connects the radiative and trapped modes, and consequently induces EIT effect in the metamaterial. Next, we describe procedures for storage and retrieval of electromagnetic waves, which can be realized by dynamical modulation of the control wave. The EIT effect is initially ac- tivated by illuminating the control wave; the probe wave then travels slowly in the metamaterial. During propagation, the control wave is switched off, so that the probe wave is cap- tured in the metamaterial. This happens because the energy in the trapped mode can not escape via the radiative mode retrieved signal -60 -40 -20 0 20 40 60 80 100 120 time / ns retrieved signal (a) V m / e b o r p V m / l o r t n o c (b) V m / e b o r p 3 2 1 0 -1 -2 -3 400 200 0 -200 -400 3 2 1 0 -1 -2 -3 V m / l o r t n o c 400 200 0 -200 -400 -60 -40 -20 0 20 40 60 80 100 120 time / ns FIG. 3. Storage and retrieval of electromagnetic waves, for a storage time of (a) 30 ns or (b) 50 ns. The upper (lower) graph shows the probe (control) signal at port Pf (Pc). 3 in the absence of three-wave mixing. After some time pe- riod τ , called the storage time, the EIT effect is recovered by reintroducing the control wave. Then, the energy stored in the trapped mode is released into the waveguide through nonlinearity-assisted coupling with the radiative mode. This procedures are notably identical to that for an atomic EIT sys- tem. By way of experimental demonstration, we placed a three- layered metamaterial with a separation of 7.5 cm in the waveguide. The amplitude of the control wave, tuned to νc = 0.45 GHz, could be varied using an external controller within several nanoseconds. We prepared a probe pulse of 30 ns duration and carrier frequency 1.035 GHz, which cor- responds to the transparency window in the presence of the control wave. The amplitudes of the probe and control waves at the input port of the waveguide were 260 mV and 1.2 V, respectively. Output signals were acquired using a high-speed oscilloscope connected to ports Pc and Pf in Fig. 2(a). The signal at port Pc was used to monitor the control wave in the absence of the probe wave, and the signal at port Pf (hav- ing passed through two filters to remove the control wave and unnecessary harmonic waves) was used to monitor the probe wave. The timing of the control-wave modulation was ad- justed to allow the capture of the rear part of the probe pulse. Figures 3(a) and (b) show the observed signals for storage times τ = 30 and 50 ns, respectively, which correspond to the periods without the control wave. The upper and lower graphs in each figure represent the time-domain waveforms at ports Pf and Pc, respectively. The modulation of the control wave can be confirmed in each lower graph. In both cases, the probe-wave signals were reduced when the control waves were switched off. This fact means that the probe waves were captured in the trapped mode, which was decoupled from the radiative mode in the absence of the control waves. After rein- troducing the control waves, some signals return, correspond- ing to the probe waves being released from the metamaterial. The decay rate of the retrieved signals is determined by the lifetime of the trapped mode. In the experiment, ohmic loss in the nonlinear capacitors dominates. If the metamaterial stores and retrieves the input pulse while preserving the spatial phase distribution, the retrieved signal is released in the forward direction; the backward wave can be suppressed. In order to confirm this effect, or "co- herent" storage, we compared the retrieved waves propagat- ing in the forward and backward directions for single- and three-layered metamaterials. The backward signal can be ob- served at port Pb in Fig. 2(a). The coupler and the filters intro- duced before port Pb were the same as those for the forward wave, and no calibration was necessary between the forward- propagating wave at port Pf and the backward-propagating wave at port Pb. The experimentally observed retrieved signals for the stor- age time τ = 40 ns are shown in Fig. 4. The upper (lower) graphs show the results for the single-layer (three-layered) metamaterial. The graphs in the left (right) column corre- spond to the forward (backward) waves observed at Pf (Pb). For the single-layer metamaterial, the amplitudes of the re- trieved signals observed in the forward and backward direc- tions are almost the same, as shown in Figs. 4(a) and (b), be- r e y a l l e g n s i forward wave backward wave (a) 2 V m / e b o r p 1 0 -1 (b) 2 V m / e b o r p 1 0 -1 -2 20 30 40 50 60 70 80 90 -2 20 30 40 50 60 70 80 90 time / ns time / ns (c) 2 V m / e b o r p 1 0 -1 s r e y a l e e r h t (d) 2 V m / e b o r p 1 0 -1 -2 20 30 40 50 60 70 80 90 -2 20 30 40 50 60 70 80 90 time / ns time / ns FIG. 4. Signals from a single-layer metamaterial retrieved in the (a) forward or (b) backward directions, and those from three-layered metamaterial retrieved in the (c) forward or (d) backward directions. The storage time is 40 ns. cause a single meta-atom releases the stored energy in both directions with equal intensity. For the three-layered metama- terial, the retrieved signal in the forward direction (Fig. 4(c)) is enhanced, while that in the backward direction (Fig. 4(d)) is suppressed. This result is attributed to the interference in the radiation emitted from the multiple meta-atoms. The fact proves that the metamaterial stores and reproduces the probe wave while preserving its phase distribution. The interference in the experiment is not complete, because of the insufficient number of meta-atoms, incomplete transparency due to ohmic loss, and inhomogeneity in the control wave. In conclusion, we have experimentally demonstrated the storage and retrieval of electromagnetic waves by exploiting EIT effect, implemented with a nonlinear metamaterial de- signed for operation in the microwave region. The storage of the probe wave was achieved by switching the control wave off during the propagation, and the retrieval by recovering the control wave. We also confirmed that the coherence was pre- served in the storage and retrieval processes by comparing the retrieved signals released in the forward and backward direc- tions. These facts indicate that the metamaterial has the same ability as the atomic EIT medium. In principle, it should be possible to store the whole pulse more efficiently by increas- ing the number of meta-atoms and by reducing the undesired dissipation mainly caused by ohmic loss in the nonlinear ca- pacitors. The nonlinear diodes can operate only at low fre- quencies (e.g., microwaves) but the extension to higher fre- quencies (e.g., the optical region) can be made possible by incorporating other nonlinear optical effects into the metama- terial. The present research was supported by Grants-in-Aid for Scientific Research No. 17K05075 and No. 16K13699. 4 1S. Harris, Phys. Today 50, 36 (1997). 2M. Fleischhauer and J. P. Marangos, Rev. Mod. Phys. 77, 633 (2005). 3M. Fleischhauer, A. Imamoglu and M. Lukin, Phys. Rev. Lett. 84, 5094 (2000). 4D. Phillips, A. Fleischhauer, A. Mair, R. Walsworth, and M. Lukin, Phys. Rev. Lett. 86, 783 (2001). 5C. Liu, Z. Dutton, C. H. Behroozi, and L. V. Hau, Nature 409, 490 (2001). 6A. Turukhin, V. Sudarshanam, M. Shahriar, J. Musser, B. Ham, and P. Hemmer, Phys. Rev. Lett. 88, 023602 (2001). 7S. Chu, B. Little, T. Kaneko, W. Pan, and Y. Kokubun, IEEE Photon. Technol. Lett. 11, 1426 (1999). 8Q. Xu, S. Sandhu, M. Povinelli, J. Shakya, S. Fan, and M. Lipson, Phys. Rev. Lett. 96, 123901 (2006). 9K. Totsuka, N. Kobayashi, and M. Tomita, Phys. Rev. Lett. 98, 213904 (2007). 10Q. Lin, J. Rosenberg, D. Chang, R. Camacho, M. Eichenfield, K. J. Vahala, and O. Painter, Nat. Photonics 4, 236 (2010). 11S. Weis, R. Rivi`ere, S. Del´eglise, O. Gavartin, Emanuel a nd Arcizet, A. Schliesser, and T. J. Kippenberg, Science 330, 1520 (2010). 12A. H. Safavi-Naeini, T. P. M. Alegre, J. Chan, M. Eichenfield, M. Winger, Q. Lin, J. T. Hill, D. E. Chang, and O. Painter, Nature 472, 69 (2011). 13I. Yoo, C. K. Han, D.-S. Shin, K. J. B. Lee, J. W. Wu, H. S. Moon, O. B. Wright, and S. H. Lee, Sci. Rep. 4, 4634 (2014). 14C. L. Garrido Alzar, M. A. G. Martinez, and P. Nussenzveig, Am. J. Phys. 70, 37 (2002). 15V. A. Fedotov, M. Rose, S. L. Prosvirnin, N. Papasimakis, and N. I. Zhe- ludev, Phys. Rev. Lett. 99, 147401 (2007). 16N. Papasimakis, V. A. Fedotov, N. I. Zheludev, and S. L. Prosvirnin, Phys. Rev. Lett. 101, 253903 (2008). 17S. Zhang, D. A. Genov, Y. Wang, M. Liu, and X. Zhang, Phys. Rev. Lett. 101, 047401 (2008). 18N. Liu, L. Langguth, T. Weiss, J. Kastel, M. Fleischhauer, T. Pfau, and H. Giessen, Nat. Mater. 8, 758 (2009). 19R. Hokari, Y. Kanamori, and K. Hane, J. Opt. Soc. Am. B 31, 1000 (2014). 20Y. Yang, I. I. Kravchenko, D. P. Briggs, and J. Valentine, Nat. Commun. 5, 5753 (2014). 21B. Lahiri, A. Z. Khokhar, R. M. De La Rue, S. G. McMeekin, and N. P. Johnson, Opt. Express 17, 1107 (2009). 22N. Liu, T. Weiss, M. Mesch, L. Langguth, U. Eigenthaler, M. Hirscher, C. S onnichsen, and H. Giessen, Nano Lett. 10, 1103 (2010). 23S. H. Mousavi, A. B. Khanikaev, J. Allen, M. Allen, and G. Shvets, Phys. Rev. Lett. 112, 117402 (2014). 24D. Floess, M. Hentschel, T. Weiss, H.-U. Habermeier, J. Jiao, S. G. Tikhodeev, and H. Giessen, Phys. Rev. X 7, 021048 (2017). 25Y. Yang, W. Wang, A. Boulesbaa, I. I. Kravchenko, D. P. Briggs, A. Puret- zky, D. Geohegan, and J. Valentine, Nano Lett. 15, 7388 (2015). 26D. Meng, S. Wang, X. Sun, R. Gong, and C. Chen, Appl. Phys. Lett. 104, 261902 (2014). 27C. Kurter, P. Tassin, L. Zhang, T. Koschny, A. Zhuravel, A. Ustinov, S. An- lage, and C. Soukoulis, Phys. Rev. Lett. 107, 043901 (2011). 28O. Limaj, F. Giorgianni, A. Di Gaspare, V. Giliberti, G. de Marzi, P. Roy, M. Ortolani, X. Xi, and D. Cunnane, S. Lupi, ACS Photonics 1, 570 (2014). 29Y. Tamayama, K. Hamada, and K. Yasui, Phys. Rev. B 92, 125124 (2015). 30J. Gu, R. Singh, X. Liu, X. Zhang, Y. Ma, S. Zhang, S. A. Maier, Z. Tian, A. K. Azad, H.-T. Chen, A. J. Taylor, J. Han, and W. Zhang, Nat. Commun. 3, 1151 (2012). 31F. Miyamaru, H. Morita, Y. Nishiyama, T. Nishida, T. Nakanishi, M. Ki- tano, and M. W. Takeda, Sci. Rep. 4, 4346 (2014). 32P. Pitchappa, M. Manjappa, C. P. Ho, R. Singh, N. Singh, and C. Lee, Adv. Opt. Mater. 4, 541 (2016). 33T. Nakanishi, T. Otani, Y. Tamayama, and M. Kitano, Phys. Rev. B 87, 161110 (2013). 34T. Nakanishi and M. Kitano, Phys. Rev. Appl. 4, 024013 (2015). 35U. Fano, Phys. Rev., 124, 1866 (1961). 36A. Miroshnichenko, S. Flach, and Y. Kivshar, Rev. Mod. Phys. 82, 2257 (2010). 37Y. S. Joe, A. M. Satanin, and C. S. Kim, Phys. Scr. 74, 259 (2006). 38S. Satpathy, A. Roy, and A. Mohapatra, Eur. J. Phys. 33, 863 (2012). 39T. Nakanishi, Y. Tamayama, and M. Kitano, Appl. Phys. Lett. 100, 044103 (2012).
1907.02630	1	1907	2019-07-05T01:10:40	Water-based Reconfigurable Frequency Selective Rasorber with Thermally Tunable Absorption Band	[ "physics.app-ph" ]	In this paper, a novel water-based reconfigurable frequency selective rasorber (FSR) at microwave band is proposed, which has a thermally tunable absorption band above the transmission band. The water-based FSR consists of a bandpass type frequency selective surface (FSS) and a 3D printing container. The water substrate is filled into the sealed space constructed by the above two structures. The numerical simulation results show that the FSR can achieve absorption with high absorptivity from 8.3 to 15.2 GHz, and obtain a transmission band of 5.2 to 7.0 GHz. The minimum insertion loss of the transmission band reaches 0.72 dB at 6.14 GHz. In addition, the FSR has the reconfigurable characteristics of absorbing or reflecting electromagnetic waves by filling with water or not. The proposed water-based FSR shows its good transmission/absorption performance under different polarizations and oblique incident angles. Due to the Debye model of water, the absorption band can be adjusted by water temperature, while the passband remains stable. At last, prototype of the FSR based on water has been fabricated, and the experimental results are presented to demonstrate the validity of the proposed structure.	physics.app-ph	physics	Water-based Reconfigurable Frequency Selective Rasorber with Thermally Tunable Absorption Band Xiangxi Yan1, Xiangkun Kong1,2, Qi Wang1, Lei Xing1, Feng Xue1, Yan Xu1, Shunliu Jiang1 1 Key Laboratory of Radar Imaging and Microwave Photonics, Ministry of Education, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, People's Republic of China 2 State Key Laboratory of Millimeter Waves of Southeast University, Nanjing 210096, People's Republic of China E-mail: [email protected] Received xxxxxx Accepted for publication xxxxxx Published xxxxxx Abstract In this paper, a novel water-based reconfigurable frequency selective rasorber (FSR) at microwave band is proposed, which has a thermally tunable absorption band above the transmission band. The water-based FSR consists of a bandpass type frequency selective surface (FSS) and a 3D printing container. The water substrate is filled into the sealed space constructed by the above two structures. The numerical simulation results show that the FSR can achieve absorption with high absorptivity from 8.3 to 15.2 GHz, and obtain a transmission band of 5.2 to 7.0 GHz. The minimum insertion loss of the transmission band reaches 0.72 dB at 6.14 GHz. In addition, the FSR has the reconfigurable characteristics of absorbing or reflecting electromagnetic waves by filling with water or not. The proposed water-based FSR shows its good transmission/absorption performance under different polarizations and oblique incident angles. Due to the Debye model of water, the absorption band can be adjusted by water temperature, while the passband remains stable. At last, prototype of the FSR based on water has been fabricated, and the experimental results are presented to demonstrate the validity of the proposed structure. Keywords: frequency selective rasorber, water, reconfigurable, thermally tunable 1. Introduction Frequency selective surfaces (FSSs) have been found numerous applications due to their attractive frequency filtering characteristics [1]. FSSs can be regarded as a spatial filter which performing a bandpass/bandstop response for electromagnetic wave, and have been widely used as radome, reflectors, and substrates of antennas. The FSSs with bandpass performance are employed in a radome design to suppress mutual interference, enhance radiation performance or reduce the Radar Cross Section (RCS) [2-4]. But such radomes based on FSSs usually reflect the out-of-band xxxx-xxxx/xx/xxxxxx signals and result in lager RCSs in other directions, so they can barely reduce the mono-static RCS. In order to realize the purpose of the true stealth radome system, metamaterial absorbers (MMAs) are usually applied to design new type of radomes which can maintain the transmission at the passband but also absorb the out-of-band waves to reduce the bistatic RCS. Such a low RCS radome based on FSS and MMA is named as frequency selective rasorber (FSR), where the word "rasorber" is a combination of the terms "radome" and "absorber" [5]. In fact, various designs of FSRs have been proposed in recent years. The reported FSRs can be mainly divided into 1 2-D FSR and 3-D FSR according to the structural dimensions. Traditionally, the 2-D FSRs are usually realized by cascading a lossy FSS with a lossless FSS [6-11]. The incident waves of frequencies outside the passband can be absorbed by lossy FSS similar to the operating principle of Salisbury Screens, while retaining the transmission performance at the passband frequencies. The 3-D FSRs are mainly constructed by parallel-plate waveguides realize multiple resonators or propagation modes. Based on this innovative concept, the 3-D FSRs can be designed to achieve high selectivity, low insertion loss or wide absorption band [12- 15]. Of course, the existing FSRs can be sorted into three categories based on the relative locations of the transmission band and the absorption band: that with the transmission band above the absorption band [16-17], that with the transmission band below the absorption band [6],[12- 13],[18], or the transmission band between two absorption bands [7],[9-11]. (PPW) to But those mentioned above FSRs both can be defined as passive FSRs, which the structures, once built, are generally inflexibility, because of their fixed nature after fabrication. Unfortunately, in practical applications, the requirement of tunability of FSRs is often necessary because of the change of the realistic environment, and traditional passive FSRs would be seriously limit its scope of applications. So the active FSRs (AFSRs) is proposed, which can realize the performance of controlling electromagnetic characteristics through actively adjusting external excitation (e.g., dc voltage size, field energy and so on). The reported electrically controlled AFSRs are divided into two categories: the switchable type of AFSRs [19-21] and the tunable type of AFSRs [18], [22]. In general, the AFSRs are commonly loaded with active components, such as PIN diodes and varactor diodes, to achieve the reconfigurable performance. However, many extra feeder lines and bias networks are employed in the structure array, causing design complexity, affecting the reconfigurable performance, high fabrication cost and difficult measured process. In order to simplify the design procedure of the AFSRs and also realize the reconfigurable purpose of rasorber structures, a novel FSR based on water substrate is firstly proposed in this paper. Water substrate is exactly applied to overcome the above disadvantages. According to the particular physics characteristics, water has been found many applications in the design of water antennas [23-24] or filling in metamaterial absorbers as substrate. Traditionally, MMAs usually adopt loading lumped resistive [25-27], using high-impedance surfaces [28- 30] or magnetic materials [31] to realize broadband absorption. Water as a unique liquid material can provide a novel theory for the design of broadband MMAs. For one thing, water has a high value about permittivity and dielectric loss in terms of electrical parameters due to the dispersion characteristic. Water can be used as a multi-layered substrate with the container for supporting multiple resonance modes, so water-based "all-dielectric" MMAs have ultra-broadband absorption operating frequency [32-37]. For another thing, as a liquid, water can be injected into various shapes of a container as a substrate due to fluidity. Compared with traditional MMAs, the absorbers based on water can highly simplify the design procedure, reduce fabrication cost and realize the optical transparency performance. As an example, Qu et al. developed a broadband absorption and optical transparency MMA based on water, polymethyl methacrylate (PMMA) and indium tin oxide (ITO) substrate [37]. The developed MMA achieved broadband microwave absorption with efficiency larger than 90% in 6.4 ~ 30 GHz as well as a high optical transmittance of about 85%. In addition, the thermal tunability performance of such all-dielectric MMAs can be realized by water temperature. So water-based structures have great potential value the practical application. However, due to the information available to the author, all water-based MMA structures typically have a metal or ITO backing plate at the bottom, and no water-based FSR with a transmission band has been designed. in In this study, a new kind of FSR structure based on water substrate is firstly proposed. The water-based FSR composed of a bandpass type FSS and a 3D printing container. The water substrate is filled into the sealed space constructed by the above two structures. The proposed FSR can realize broadband absorption from 8.37 to 15 GHz, and obtain a transmission band covers 5.2 to 7.0 GHz. Furthermore, the transmission and absorption characteristics of the FSR under different polarizations and oblique is analysed, showing the stability of the passband and -10dB absorption band. The thermal tunability and reconfigurability of the proposed rasorber are also investigated. At last, a prototype of the proposed FSR structure is fabricated and experimentally tested. Good agreement with the simulated and the measured results indicate the validity of the proposed design. incident angles 2 Water-based FSR The proposed water-based FSR is a sandwiched structure consisting of a container layer, a water layer, and an FSS layer, as shown in Fig.1. On the bottom layer, the FSS is made up of the cross-slot type unit that is printed on the F4B substrate. The thickness of the F4B substrate is h1. The length and width of the cross-slot type unit are l and w. As shown in Fig.1 (b), the upper layer structure is constructed of resin material by using 3D printing technique, whose shape is the same as that of the FSS layer. A sealed space between the two layers is designed to fill with water substrate. In Fig.1(c), the thickness of the container, the water layer, and the F4B substrate are h3, h2 and h1, respectively. The geometrical 2 parameters are p=20 mm, l=18 mm, w=3 mm, h1=0.5 mm, h2=0.9 mm, h3=3.8 mm. The S-Parameters of the water-based FSR are numerically investigated using the full-wave electromagnetic simulation software CST Microwave Studio. In the simulation, the permittivity of the resin and the F4B are 2.93(1-j0.04) and 2.65(1-j0.001), respectively. The permittivity of water at radio frequency can be described by the Debye formula with the filled water temperature of T as follows [38]: , (1) where (T), (T) and (T) are the optical permittivity, static permittivity and rotational relaxation time, respectively. Meanwhile, (T), (T) and (T) in the Debye formula of water are all directly dependent on water temperature T: , , , (2) (3) (4) where a1=87.9, b1=0.404 K-1, c1=9.59×10-4 K-2, d1=1.33×10-6 K-3, a2=80.7, b2=4.42×10-3 K-1, c2=1.37×10-13 K-2, d2=651 ℃, T0=133 ℃ and T is the water temperature in ℃. (a) (c) (b) Fig 1. Schematic diagram of the water-based FSR. (a) 3D periodic structure; (b) layer by layer of the unit cell; (c) side view of the unit cell. Fig 2. Dielectric permittivity of water for different temperature. 3 Results and performance analysis A Reconfigurable performance Firstly, we simulate the model of water-based FSR at room temperature (25℃). The numerical simulation results of transmission/reflection coefficient under normal incidence are shown in Fig.3 (a). Because of the bandpass response of the FSS layer, the water-based FSR can obtain a transmission window with a minimum insertion loss of 0.72 dB between 5.2 GHz and 7.0 GHz. The absorption band is achieved from 8.3 GHz to 15.2 GHz with the absorptivity higher than 80%, where the S11 and S21 are less than -10 dB. Secondly, the electromagnetic characteristics of the proposed structure without filled water are studied. The simulation results are shown in Fig 3 (b). It can be noticed that the previous absorption band in the high frequency range changes into the reflection band, while the transmission band does not shift significantly. The results demonstrate that the water-based FSR has good reconfigurable performance due to the filled water in the structure has absorption effect on electromagnetic wave. In Fig.3 (b), the gap between the lower and upper effective reflection band is caused by the dielectric resonance of the container substrate. There is a slight offset for the transmission zero of the passband because of the chang of the effective dielectric constant and equivalent permeability caused by filling with water or not. In summary, the designed FSR can realize the reconfigurable performance with or without filled water. the one hand, The real and imaginary parts of the permittivity change significantly at different water temperatures, as shown in Fig.2. On in microwave frequency can be achieved due to the high imaginary part of the permittivity of water. On the other hand, the water-based FSR has the ability of thermal tunability owing to the significant dispersion characteristics of the permittivity of water. the effective absorption 3 0+1(T)(T)(,T)(T)i(T)002301111+TabTcTdT202bTTTae202dTTTce4681012141601020304050607080Permittivity of waterFrequency(GHz) T=0℃ T=25℃ T=50℃ T=75℃ T=100℃Real componentsImaginary components213 (a) (b) (a) (b) Fig 3. Reconfigurable performance of the water-based FSR for different states. (a) with water; (b) without water. Fig 4. Simulated transmission/reflection coefficients of the water-based FSR with different incident angles. (a) TE polarization; (b) TM polarization B Wide-angle performance C Thermally tunable performance Fig 4 shows the transmission/reflection coefficients of the water-based FSR under oblique incidence of electromagnetic waves. As shown in Fig. 4(a), for TE model, when the oblique incident angle increases from 0° to 45°, the intensity of the magnetic field decreases gradually and the electric field remains unchanged, which makes the effect of incident angle on dielectric resonance weaker. The absorptivity is still higher than 80% in the entire absorption band under wide- angle incidences. As shown in Fig.4 (b), for TM model, the electric field intensity decreases significantly with the increase of incident angle, and the dielectric resonant is sensitive to the electric field. As a result, the absorption band decreases remarkably when the incident angle is less than 45°, except for several special frequency points, but the absorption rate maintains high. In the meantime, with the increase of incident angle, the absorption band shifts to higher frequency. Furthermore, the transmission window under both TE and TM polarization waves exhibits the same bandpass the transmission zero of the passband is consistently located at 6.24 GHz. Based on the above analysis, the proposed water- based FSR can display good bipolarization and wide incident angle performance. response. Under all these conditions, As mentioned in Fig.2, the permittivity of water has a certain relationship with its temperature T, so it is necessary to study the thermal tunability of water-based rasorbers. According to the formulas (1) to (4), the numerical simulation results are obtained by changing the water temperature, inwhich only the TE wave under normal incidence is considered. Fig.5 (a) presents the S11 and S21 of the transmission window for the temperature varies from 0~100℃. It can be seen that the transmission window maintains the constant bandwidth of the -3 dB passband and has the function of thermal stability. The transmission window is influenced weakly at different water temperatures, because the permittivity of the substrate of whole structure has little change. In contrast, as shown in Fig.5 (b), the absorption band exhibits good thermal tunability with the increase of the water temperature: in the lower frequency, the absorption bandwidth gradually becomes narrow and the fractional bandwidth obviously decreases, while in the higher frequency, the variation of the absorption bandwidth is very small. Meanwhile, when the temperature increases, the absorptivity falls slowly, but it still keeps over 80%. Therefore, the water-based FSR exhibits the thermal stability of thermal transmission window and the active the 4 Transmission band46810121416-25-20-15-10-50Absorption bandS-Parameters(dB)Frequency(GHz) S11 S21 46810121416-30-25-20-15-10-50Reflection bandTransmission bandS-Parameters(dB)Frequency(GHz) S11 S21 46810121416-30-25-20-15-10-50S-Parameters(dB)Frequency(GHz) =0° ,S11 =0° ,S21 =15°,S11 =15°,S21 =30°,S11 =30°,S21 =45°,S11 =45°,S2146810121416-50-40-30-20-100S-Parameters(dB)Frequency(GHz) =0° ,S11 =0° ,S21 =15°,S11 =15°,S21 =30°,S11 =30°,S21 =45°,S11 =45°,S21 adjustment of the absorption band which varies with the water temperature. produce dielectric resonance in the proposed composite structure. The resonance frequency can be studied by analyzing the Mie resonance theory [39] as: , , (5) (6) where F(θ) is a resonant function and becomes infinite at some values of θ, so that results in the magnetic and electric resonances. Among all the resonances, the first-order Mie resonance mode is the most intense. θ is a function about the resonance frequency and approximately equally to for the first-order Mie resonance; c is the speed of light in vacuum and r is the radius of the dielectric sphere. 𝜀p and 𝑢p are the permittivity and permeability of the dielectric particles. However, it should be noted that in article [40-41], the dielectric materials used to calculate the first Mie resonance are usually cubic or sphere particles, but here, the water in container forms a rectangle, so it is no longer applicable to calculate the resonance frequency by using the parameter r in equation (5). Therefore, we provide a modified formula (7) to illustrate the specific relationship between h2 and r, and further get the functional relationship between h2 and the resonance frequency. The corrected formula can be written as , (7) In order to verify the validity of the mordified formula, the results calculated by Eq. (7) are compared with those obtained by simulation software, as shown in Fig 7. The results show the theoretical calculation results are in good agreement with the simulation results, which proves the correctness of the above formulas. Fig 6 Simulated transmission/reflection coefficients by varying the thickness h2 of the water layer. (a) (b) Fig 5 The transmission/absorption results of the water-based FSR under different temperatures. (a) Transmission window; (b) Absorption band. D Mie resonance theory analysis based on water Moreover, we consider the transmission/absorption band under the condition of optimizing structure parameters. Fig 6 presents the S-Parameters results of the water-based FSR with different thickness of water layer h2. For the absorption band, when the thickness of h2 varies from 0.8 to 1.0mm, the relative bandwidth of the absorption band can be broadened gradually, and the absorptivity keeps higher than 80% at all times. For the transmission band, the bandwidth of -3 dB passband and the minimum insertion loss almost remain unchanged with the thickness of h2 increase to 0.9 mm. When h2 up to 1mm, the minimum insertion loss of the transmission band reach to 1 dB, which is worse than that of the above situation. Thus, in pursuit of the minimum insertion loss of the passband and expand the absorption bandwidth simultaneously, we choose the appropriate thickness h2=0.9mm as the optimal structural parameter of the aquifer. It can be seen that when the value of h2 increases, the resonance frequency moves to lower frequency. The reason is that under the excitation of electromagnetic wave, a mixed dielectric layer consisting of water and resin substrate can 5 2ppfcr22sincos1sincosF251324hr45678-25-20-15-10-50S11S-Parameters(dB)Frequency(GHz) T=0℃ T=25℃ T=50℃ T=75℃ T=100℃S21468101214160.00.20.40.60.81.0Absorption rateFrequency(GHz) T=0℃ T=25℃ T=50℃ T=75℃ T=100℃46810121416-30-25-20-15-10-50S-Parameters(dB)Frequency(GHz) h2=0.8 ,S11 h2=0.8 ,S21 h2=0.9 ,S11 h2=0.9 ,S21 h2=1.0 ,S11 h2=1.0 ,S21 Fig 7 Comparison of the simulated and theoretical resonance frequencies with different thickness of h2. 4 Experimental Verification and Discussion A prototype of the proposed rasorber is fabricated to validate the main functions, as shown in Fig.8. The water- based FSR is composed of 15 × 15 unit cells with the overall size of 300 mm × 300 mm × 5.2 mm. The upper layer plays a role of container constructed by using 3D printing technology. The bottom layer is made up of the bandpass FSS, which is printed on F4BK265 substrate. The epoxy resin adhesive with 0.5mm thick is used to bond the upper and bottom layer to build a sealed space for filling with water substrate. Based on the theory of connected container, two holes of connected hose are arranged to fill and drain water. Fig 8 prototype of the water-based FSR. shows Fig. 9 the measurement environment of transmission /reflection performance of water-based FSR prototype in anechoic chamber. The prototype is set in the middle of the measurement platform. Two standard horn antennas (2-18 GHz), mounted on both ends of the measurement platform for aiming at the center of the prototype, are connected to an Agilent N5245A vector network analyzer (VNA) by using two RF cables. The measured results of the prototype with and without filled water under the normal incidence at room temperature are shown in Fig.10 (a) and (b). Considering the effcet of the exoxy resin adhesive between the two layers, we optimize the previous model and show the optimized simulation results of the water-based FSR with and without filled water in Fig.10. The comparison of the former and optimized simulation results is shown in Fig.3 (a) and Fig.10 (a), it can be seen that the optimized transmission band shifts to lower frequency, while the absorption band stays the same. In Figs.10 (a) and (b), the measured results have good agreement with the optimized ones. Moreover, there is a slight deviation caused by the fabrication tolerance of 3D technology. Fig 9 Measurement setup. (a) (b) Fig 10 Comparison of measured and re-simulated results of the water- based FSR under normal incidence. (a) with water; (b) without water. Next, the wide-angle performance of our designed water- based FSR is further studied, as shown in Fig11. The measured results of water-based FSR under both TE and TM polarized wave at incident angles of 15° and 45° are in good 6 0.750.800.850.900.951.001.051.107891011121314Frequency(GHz)h2(mm) Calculated Simulated 46810121416-25-20-15-10-50S-Parameters(dB)Frequency(GHz) S11_Measured S21_Measured S11_Simulated S21_Simulated46810121416-30-20-100 S11_Measured S21_Measured S11_Simulated S21_SimulatedS-Parameters(dB)Frequency(GHz) agreement with the optimized results. It is noticed that there are two harmonics in the simulation results, which is due to the existence of harmonics of the bandpass FSS at its rejection band. (a) (b) Fig 11 Measured and re-simulated transmission/reflection coefficient of the water-based FSR under oblique incidence. (a) TE polarization; (b) TM polarization. Finally, the transmission/reflection coefficients of water- based FSR at different temperatures are discussed. A thermostat is introduced to create a same measurement environment as the simulation ones. Owing to the volatility of the epoxy resin adhesive and the thermal deformation of the resin material, the measured temperatures are set below 55℃, and the transmission/reflection coefficients at different temperatures are measured, as shown in Fig.12. The measured transmission zero of the passband slightly shifts to a lower frequency when the temperature goes up from 40℃ to 55℃. The discrepancies of the measured and optimized results are generated by the thermal influence of various structural materials and the edge scattering caused by processing size, which are not considered in our model simulation. From the above analysis, the measured results effectively prove the feasibility of the proposed water-based rasober. (a) (b) Fig 12 Measured and re-simulated results of the water-based FSR at different temperature. (a) transimission window; (b) absorption band. 5 Conclusion In conclusion, a water-based reconfigurable FSR is designed to achieve a thermally tunable absorption band while ensuring a high selective transmission band. We adopt FSS layer and 3D printing container to assemble the proposed rasorber. The water layer can be created by filling with water into the sealed space, which is built by the upper container and the bottom FSS layer. The simulated -10 dB absorption band ranges from 8.3 to 15.2 GHz, and the transmission window is generated by the bandpass FSS in the frequency from 5.2 to 7.0 GHz. In addition, compared with traditional AFSRs, the reconfigurability of water-based FSR is obtained by filling with water or not. The wide-angle performance of FSR under both TE and TM polarizations waves is further analyzed and demonstrated. Moreover, the thermal tunability of water-based FSR at different temperatures is discussed. The structure is fabricated and the main functions obtained from theoretical analysis are verified by experiments. The innovative water-based rasorbers have potential application in antenna stealth technology. 7 46810121416-30-20-100S-Parameters(dB)Frequency(GHz)Mea.θ=15°,S11 Mea.θ=15°,S21Mea.θ=45°,S11 Mea.θ=45°,S21Sim.θ=15°,S11 Sim.θ=15°,S21Sim.θ=45°,S11 Sim.θ=45°,S2146810121416-40-30-20-100S-Parameters(dB)Frequency(GHz)Mea.θ=15°,S11 Mea.θ=15°,S21Mea.θ=45°,S11 Mea.θ=45°,S21Sim.θ=15°,S11 Sim.θ=15°,S21Sim.θ=45°,S11 Sim.θ=45°,S2145678-30-20-100S-Parameters(dB)Frequency(GHz) Mea.T=10℃ Mea.T=25℃ Mea.T=40℃ Mea.T=55℃ S11S21468101214160.00.20.40.60.81.0Absorption rateFrequency(GHz) Mea.T=10℃ Mea.T=25℃ Mea.T=40℃ Mea.T=55℃ Sim.T=10℃ Sim.T=25℃ Sim.T=40℃ Sim.T=55℃ Acknowledgements The authors greatly appreciate the help of Professor Qiang Cheng from Southeast University for his meaningful advice and assistance in this paper. This work is supported by 'the Fundamental Research Funds for the Central Universities' (No. kfjj20180401),the Chinese Natural Science Foundation (Grant No.61471368), Aeronautical Science Foundation of China (20161852016), China Postdoctoral Science Foundation (Grant No. 2016M601802), Open Research Program in China's State Key Laboratory ofMillimeter Waves (grant No. K202027) and by Jiangsu Planned Projects for Postdoctoral Research Funds (Grant No. 1601009B). References [22] Ren X, Zhao J, Chen K, Jiang T, Zhu B and Feng Y 2018 IEEE Asia-Pacific Conf. Antennas Propag. 287 -- 8 [23] Xing L, Huang Y, Xu Q, Alja'Afreh S and Liu T 2016 IEEE Antennas Wirel. Propag. Lett. 15 174 -- 7 [24] Xu Q, Xing L, Alrawashdeh R, Shen Y, Al Ja'afreh S and Huang Y 2015 IET Microwaves, Antennas Propag. 9 735 -- 41 [25] Shang Y, Shen Z and Xiao S 2013 IEEE Trans. Antennas Propag. 61 6022 -- 9 [26] Chen J, Shang Y and Liao C 2018 IEEE Antennas Wirel. Propag. Lett. 17 591 -- 5 [27] Cheng Y Z, Wang Y, Nie Y, Gong R Z, Xiong X and Wang X 2012 J. Appl. Phys. 111 044902 [28] Costa F, Monorchio A and Manara G 2010 IEEE Trans. Antennas Propag. 58 1551 -- 8 [29] Kundu D, Baghel S, Mohan A and Chakrabarty A 2019 IEEE Trans. Antennas Propag. 67 3533 -- 8 [30] Pang Y, Cheng H, Zhou Y and Wang J 2012 J. Phys. D. Appl. [1] Munk B A 2000 Frequency Selective Surface: Theory and Phys. 45 215104 Design (New York: Wiley) [31] Chen Q, Bie S, Yuan W, Xu Y, Xu H and Jiang J 2016 J. [2] Omar A A and Shen Z 2018 IEEE Trans. Antennas Propag. Phys. D. Appl. Phys. 49 425102 PP 67 282 -- 90 [3] Liu Z, Liu S, Bian B, Kong X and Zhang H 2019 Int. J. RF Microw. Comput. Eng. 29 1 -- 9 [32] Pang Y, Wang J, Cheng Q, Xia S, Zhou X Y, Xu Z, Cui T J and Qu S 2017 Appl. Phys. Lett. 110 104103 [33] Shen Y, Zhang J, Pang Y, Wang J, Ma H and Qu S 2018 Opt. [4] Mei P, Lin X and Zhao Y 2018 IEEE Trans. Antennas Propag. Express 26 15665 -- 74 66 582 -- 9 [5] Munk B A 2009 Metamaterials: Critique and Alternatives (Hoboken: Wiley) [6] Costa F and Monorchio A 2012 IEEE Trans. Antennas Propag. 60 2740 -- 7 [34] YIN J R A J Y 2018 Opt. Mater. Express 8 2060 -- 71 [35] Zhu W, Rukhlenko I D, Xiao F, He C, Geng J, Liang X, Premaratne M and Jin R 2017 Opt. Express 25 15737 -- 45 [36] Huang X, Yang H, Shen Z, Chen J, Lin H and Yu Z 2017 J. Phys. D. Appl. Phys. 50 385304 [7] Zhang K, Jiang W and Gong S 2017 IEEE Antennas Wirel. [37] Pang Y, Shen Y, Li Y, Wang J, Xu Z and Qu S 2018 J. Appl. Propag. Lett. 16 2586 -- 9 Phys. 123 155106 [8] Chen Q, Sang D, Guo M and Fu Y 2018 IEEE Trans. [38] Andryieuski A, Kuznetsova S M, Zhukovsky S V., Kivshar Y Antennas Propag. 66 4105 -- 14 [9] Huang H and Shen Z 2017 IEEE Antennas Wirel. Propag. Lett. 16 3212 -- 5 [10] Arrays C, Shang Y, Shen Z, Member S and Xiao S 2014 IEEE S and Lavrinenko A V 2015 Sci. Rep. 5 13535 [39] Lewin L 2014 The electrical constants of a material loaded with spherical particles J. Inst. Electr. Eng. 94 65 -- 8 [40] Liu X, Zhao Q, Lan C and Zhou J 2013 Appl. Phys. Lett. 103 Trans. Antennas Propag. 62 5581 -- 9 031910 [11] Xiu X, Che W, Yang W and Han Y 2018 IEEE MTT-S Int. [41] Liu X, Lan C, Bi K, Li B, Zhao Q and Zhou J 2016 Appl. Wirel. Symp. IWS 2018 - Proc. 1 1 -- 3 [12] Li B and Shen Z 2014 IEEE Trans. Antennas Propag. 62 Phys. Lett. 109 062902 6536 -- 41 [13] Yu Y, Shen Z, Deng T and Luo G 2017 IEEE Trans. Antennas Propag. 65 4363 -- 7 [14] Omar A A, Shen Z and Huang H 2017 IEEE Trans. Antennas Propag. 65 6173 -- 8 [15] Zhang Y, Li B, Zhu L, Tang Y, Chang Y and Bo Y 2018 IEEE Antennas Wirel. Propag. Lett. 17 633 -- 6 [16] Chen Q, Sang D, Guo M and Fu Y 2019 IEEE Trans. Antennas Propag. 67 1045 -- 52 [17] Guo M, Sun Z, Sang D, Jia X and Fu Y 2019 IEEE Access. 7 24964 -- 70 [18] Tu H, Liu P, Huang J and Qin Y 2017 Int. Symp. Electromagn. Compat. - EMC Eur. 2017, EMC Eur. 2017 8 -- 11 [19] Han Y, Che W, Xiu X, Yang W and Christopoulos C 2017 IEEE Trans. Antennas Propag. 65 6998 -- 7008 [20] Bakshi S C, Mitra D and Ghosh S 2019 IEEE Antennas Wirel. Propag. Lett. 18 29 -- 33 [21] Zhou Q, Liu P, Yu D, Bian L and Liu C 2018 IET Microwaves, Antennas Propag. Res. 12 1470 -- 6 8

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.