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
1606.07555	1	1606	2016-06-24T03:38:05	Type-II Topological Dirac Semimetals: Theory and Materials Prediction (VAl3 family)	[ "cond-mat.mes-hall" ]	The discoveries of Dirac and Weyl semimetal states in spin-orbit compounds led to the realizations of elementary particle analogs in table-top experiments. In this paper, we propose the concept of a three-dimensional type-II Dirac fermion and identify a new topological semimetal state in the large family of transition-metal icosagenides, MA3 (M=V, Nb, Ta; A=Al, Ga, In). We show that the VAl3 family features a pair of strongly Lorentz-violating type-II Dirac nodes and that each Dirac node consists of four type-II Weyl nodes with chiral charge +/-1 via symmetry breaking. Furthermore, we predict the Landau level spectrum arising from the type-II Dirac fermions in VAl3 that is distinct from that of known Dirac semimetals. We also show a topological phase transition from a type-II Dirac semimetal to a quadratic Weyl semimetal or a topological crystalline insulator via crystalline distortions. The new type-II Dirac fermions, their novel magneto-transport response, the topological tunability and the large number of compounds make VAl3 an exciting platform to explore the wide-ranging topological phenomena associated with Lorentz-violating Dirac fermions in electrical and optical transport, spectroscopic and device-based experiments.	cond-mat.mes-hall	cond-mat	a Type-II Topological Dirac Semimetals: Theory and Materials Prediction (VAl3 family) Tay-Rong Chang∗,1 Su-Yang Xu∗†,2 Daniel S. Sanchez,2 Shin-Ming Huang,3 Guoqing Chang,4, 5 Chuang-Han Hsu,4 Guang Bian,2 Ilya Belopolski,2 Zhi-Ming Yu,6, 7 Xicheng Xu,8, 9 Cheng Xiang,8, 9 Shengyuan A. Yang,7 Titus Neupert,10 Horng-Tay Jeng,1, 11 Hsin Lin†,4, 5 and M. Zahid Hasan†2 1Department of Physics, National Tsing Hua University, Hsinchu 30013, Taiwan 2Laboratory for Topological Quantum Matter and Spectroscopy (B7), Department of Physics, Princeton University, Princeton, New Jersey 08544, USA 3Department of Physics, National Sun Yat-Sen University, Kaohsiung 804, Taiwan 4Centre for Advanced 2D Materials and Graphene Research Centre National University of Singapore, 6 Science Drive 2, Singapore 117546 5Department of Physics, National University of Singapore, 2 Science Drive 3, Singapore 117542 6School of Physics, Beijing Institute of Technology, Beijing 100081, China 7Research Laboratory for Quantum Materials, Singapore University of Technology and Design, Singapore 487372, Singapore 8International Center for Quantum Materials, School of Physics, Peking University, China 9Collaborative Innovation Center of Quantum Matter, Beijing,100871, China 10Department of Physics, University of Zurich, 190, CH-8052, Switzerland, Winterthurerstrass 11Institute of Physics, Academia Sinica, Taipei 11529, Taiwan (Dated: June 27, 2016) ∗ These authors contributed equally to this work. † Corresponding authors (emails): [email protected], [email protected], [email protected] 1 2 Abstract The discoveries of Dirac and Weyl semimetal states in spin-orbit compounds led to the realiza- tions of elementary particle analogs in table-top experiments. In this paper, we propose the concept of a three-dimensional type-II Dirac fermion and identify a new topological semimetal state in the large family of transition-metal icosagenides, MA3 (M=V, Nb, Ta; A=Al, Ga, In). We show that the VAl3 family features a pair of strongly Lorentz-violating type-II Dirac nodes and that each Dirac node consists of four type-II Weyl nodes with chiral charge ±1 via symmetry breaking. Fur- thermore, we predict the Landau level spectrum arising from the type-II Dirac fermions in VAl3 that is distinct from that of known Dirac semimetals. We also show a topological phase transition from a type-II Dirac semimetal to a quadratic Weyl semimetal or a topological crystalline insulator via crystalline distortions. The new type-II Dirac fermions, their novel magneto-transport response, the topological tunability and the large number of compounds make VAl3 an exciting platform to explore the wide-ranging topological phenomena associated with Lorentz-violating Dirac fermions in electrical and optical transport, spectroscopic and device-based experiments. 3 The correspondence between condensed matter and high-energy physics has been a source of inspiration throughout the history of physics. Advancements in topological band theory have uncovered a new and profound relations that has enabled the realization of elemen- tary relativistic fermions in crystals with unique topologically non-trivial properties [1–20]. Specifically, the low-energy quasiparticle excitations of type-I Dirac semimetals [5–10], type-I Weyl semimetals [11–17], and topological superconductors [18–20] are the direct representa- tions of relativistic Dirac, Weyl, and Majorana fermions, respectively. From an application perspective, what makes this realized connection with high-energy physics of importance and interest is the resulting broad range of topologically protected phenomena that can be potentially used for low-power electronics, spintronics, and robust qubits [21–23]. For these reasons, the type-I Dirac semimetal state in Na3Bi [6, 7] and Cd3As2 [8, 9], the type-I Weyl semimetal state in the TaAs family of crystals [15, 17], and the various topological supercon- ductor candidates [18–20] have attracted tremendous interest. Very recently, a new line of thinking that looks for new topological quasiparticles beyond direct analogs in high-energy physics, has gained attention. Such an idea offers inroads into new topological phenomena that are not limited by the stringent constraints in high-energy physics [24–26]. A particu- larly interesting proposal is the prediction of type-II emergent Weyl fermions [27]. Type-I Weyl fermions, which have been realized in the TaAs family of crystals, are the direct analogs of the massless relativistic Weyl fermion from high-energy. They respect Lorentz symmetry and have the typical conical dispersion. In contrast, type-II Weyl fermions are dramatically Lorentz symmetry breaking, which is manifest in a tilted-over cone in energy-momentum space [27]. These Lorentz violating Weyl fermions can give rise to many new properties, such as a direction-dependent chiral anomaly [28], an anti-chiral effect of the chiral Landau level [29], novel quantum oscillations due to momentum space Klein tunneling [30], and a modified anomalous Hall conductivity [31]. The novel type-II Weyl semimetal state has been recently predicted/confirmed in a number of 3D crystals [27, 32–37, 39–41]. Since the type-II behavior only relies on the fact that Lorentz invariance is not a nec- essary symmetry requirement in condensed matter physics, in solid-state crystals, Lorentz symmetry breaking is not limited to the type-II Weyl fermion and, in principle, can emerge in other particles, including the Weyl fermion's most closely related particle, the Dirac fermion. However, to date, three-dimensional (3D) type-II Dirac fermions remain entirely lacking. Here, for the first time, we propose the concept of the 3D type-II Dirac semimetal 4 state and identify it in a large family of transition-metal icosagenides, MA3 (M=V, Nb, Ta; A=Al, Ga, In). Type-II Dirac fermions The VAl3 family of compounds crystalizes in a body-centered tetragonal Bravais lattice with lattice constants a = 3.78 A and c = 8.322 A [42] and the space group I4/mmm (No. 139), as shown in Fig. 1a. In this structure, each Al atom is surrounded by four V atoms in two different local structures: a planer square and a tetrahedron geometry (Fig. 1b). Figure 1c shows the bulk Brillouin zone of the VAl3 crystal. Interestingly, without spin-orbit coupling (SOC), the conduction and valence bands cross to form a triply-degenerate point along the Γ − Z line, which, upon the inclusion of SOC, evolves into the type-II Dirac node. We present a physical picture for the formation of the type-II Dirac fermions in VAl3 (Figs. 1d-f). The important symmetries are time-reversal (T ), space-inversion (I), the C4z rotation, and the Mx mirror reflection. Since the Γ−Z line lies in the Mx mirror plane and coincides with the C4z rotational axis, the electron states along Γ − Z are required to be eigenstates of both operators. In the absence of SOC, the C4z operator has the following four eigenvalues, ±1 and ±i, and we denote the corresponding eigenstates as ψ1, ψ−1, ψi, and ψ−i, respectively. Under the Mx mirror reflection, ψ1, ψ−1 remain unchanged (Mx ψ1 = ψ1, Mx ψ−1 = ψ−1), whereas ψi and ψ−i will transform into each other Mx ψi = ψ−i. Thus, as Mx and C4z do not commute in this subspace, ψi and ψ−i must remain degenerate everywhere along the Γ−Z line. In the case of VAl3, our calculation shows that the observed triply-degenerate point is formed by the crossing between a ψ−1 (singly- degenerate) band and a ψ±i (doubly-degenerate) band (Fig. 1d). This triple degeneracy is protected because the presence of C4z rotational symmetry prohibits hybridization between states of different C4z eigenvalues. Now we explain how the triply-degenerate point evolves into the type-II Dirac node upon the inclusion of SOC. With SOC, the four eigenvalues of C4z become e iπ 4 , e3i π 4 , e5i π 4 , and ei 7π 4 , and we denote the corresponding eigenstates as 4 4 , ψ ei 3π ei 5π 4 , ψ , and ψ ψei π 1e). Along Γ − Z, the doubling of the bands can be tracked by their C4z eigenvalues. ψ−i ⇒ Specifically, we have ψ1 ⇒ ψ . Adding SOC also doubles the number of bands at play (Fig. ψ−1 ⇒ ψ ei 7π , ψ ; ; 4 4 , ψ ei 3π 4 ei 3π 4 ei 5π 4 ψi ⇒ ψei π . Under the Mx mirror reflection, ψei π , ψei π 4 ; ei 7π 4 ψ , ψ 4 4 ei 5π (Mxψei π Therefore, ψei π ei 7π 4 = iψ ei 4 and ψ 7π 4 ei 7π 4 4 and ψ ei 7π 4 transform into each other ), and ψ and ψ 3π 4 5π 4 ei are degenerate everywhere along Γ − Z and the same is true for ei ei ei transform into each other (Mxψ 5π 4 = iψ ). 3π 4 4 and ψ . Taking the above conclusions from symmetry analyses, We can understand ψ ei 3π the evolution from triply-degenerate node to the type-II Dirac node. Specifically, the ψ−1 ei 5π 4 5 (singly-degenerate) band without SOC becomes a ψ (doubly-degenerate) band with SOC. The ψ−±i (doubly-degenerate) band without SOC becomes two doubly-degenerate ei 3π ei 5π , ψ 4 4 bands (ψei π eigenvalues open a gap. On the other hand, crossings between bands with the different C4z ) with SOC. Crossings between bands with the same C4z 4 , ψ ei and ψ , ψ 7π 4 3π 4 5π 4 ei ei eigenvalues remains gapless. This gives rise to the Dirac node of VAl3 (Fig. 1f). We now present the calculated band structure of VAl3 to reveal the Dirac node and its type-II character. Figure 2a shows the calculated bulk band structure along high symmetry directions. We mark the energy gap between the lowest conduction and valence bands by the green shaded areas. The conduction and valence bands cross each other near the Fermi level at discrete k point, revealing the semimetallic ground state. In Fig. 2d, we show the zoomed-in view of the band structure near the crossing point. We see that, along the Γ − Z direction, the conduction and valence bands cross each other, forming the Dirac node near the Z point. Figures 2b,c show the energy dispersion away from the Dirac node along all three k directions. While the two bands have Fermi velocities of opposite signs along kx and ky directions, they have velocities of the same sign along kz. Moreover, the constant energy contour at the energy of the Dirac node consists of an electron pocket and a hole pocket touching at the Dirac node. These observations demonstrate the first type-II Dirac fermion semimetal state. We note that a few recent papers [43–45] have discussed type-II linear band crossings in 2D. Here our focus is 3D. In fact, Weyl fermions are not allowed in 2D so linear band crossings in 2D such as graphene and the surface states of topological insulators are usually called "Dirac" but that mainly refers to the linear dispersions. Only in 3D, the distinction between Dirac and Weyl becomes well-defined, and carries a topological meaning because Weyl fermions have a nonzero chiral charge whereas 3D Dirac fermions do not. Topological invariant and Fermi arc surface states. In order to understand the topological properties of the type-II Dirac semimetal state in VAl3, we calculate the 2D Z2 invariant ν2D and the mirror Chern number nM. A 2D Z2 invariant can be defined on any 2D k plane that is time-reversal invariant and whose band structure has a full energy gap. In VAl3, the 2D planes that satisfy these conditions TABLE I: The product of parity eigenvalue of the occupied states for eight time-reversal symmetry momenta in BZ. 6 Γ 2X Z 4N + + + - include kz = 0 and kz = π. Since the system preserves space-inversion symmetry, ν2D can be computed through a parity analysis. In Table 1, we show the parity total eigenvalue of the occupied states for the eight time-reversal invariant momenta, from which we conclude that both the kz = 0 and kz = π planes have a trivial Z2 number (νkz=0 = νkz=π = 0). A mirror Chern number nM can be defined on any 2D k plane that is invariant under a mirror reflection and whose band structure has a full energy gap. In VAl3, only the kz = 0 plane satisfies these conditions. We computed the nM associated with the Mz mirror plane by the Kubo formula, and our calculations show nM = 2 for the kz = 0 plane. We note that existing Dirac semimetals Na3Bi and Cd3As2 are known to possess a 2D Z2 (ν2D = 1) and a mirror Chern number (nM = 1), respectively [10]. Therefore, the result nM = 2 indicates that the Dirac semimetal state in VAl3 is topologically distinct from that of in Cd3As2 or Na3Bi [10]. We now explore the existence of protected surface states in VAl3 and their connection to the type-II Dirac nodes. In order to do so, we calculate the surface electronic structure of the (100) surface, where the two Dirac nodes are projected onto different k locations in the surface BZ. Figures 3d,e show how the bulk BZ is projected onto the (100) surface. Due to the body-centered structural property, VAl3's (100) surface BZ center ¯Γ corresponds to the projection of both the Σ − Γ line and the Σ1 − Z line. Because the Dirac nodes are near the Z point in the bulk, their surface projections are close to the ¯Γ point. Figure 3a shows the energy dispersion of surface band structure along the ¯Γ − ¯N (kz) direction. The shaded areas represent the projected bulk bands whereas the distinct lines are the surface states. We observe the type-II bulk Dirac fermion cone in the form of a tilted-over cone near the ¯Γ point (kz ≃ 0.15 2π c ). We find surface states that emerge out of the Dirac node, suggesting the existence of Fermi arcs. In Figs. 3b,c, we present the surface constant energy contour at the energy of the bulk Dirac nodes. We see two pairs of Fermi arcs terminated 7 onto a Dirac node. They start from the Dirac node and quickly merge onto the projected electron-like pocket. According to the current theoretical understanding [10], the surface states in a Dirac semimetal are not required to connect the Dirac node because a Dirac node does not carry nonzero chiral charge. However, band structure calculations have now found Fermi arcs connecting Dirac nodes in all known Dirac semimetals including Na3Bi, Cd3As2 [5, 7], and VAl3 (this work). Additionally, we notice that at E ∼ 35 meV above EF , the normal surface states (Fig. 3a) are observed to accidentally cross with the Fermi arcs and form a surface (2D) type-II Dirac cone. Although such crossings are not guaranteed by any topological invariant, they are protected because the surface bands that cross have opposite My eigenvalues. Topological phase transitions in VAl3. To further showcase the novel physics that may be studied in VAl3, we will now turn our attention towards investigating its topological phase transitions. Generically, a Dirac semimetal can be regarded as a critical point of different phases. By starting from the Dirac semimetal phase, it is possible to realize different topological phases by simply tuning various parameters. In Fig. 4e we show a cartoon illustration of the (001) surface Fermi surface of VAl3. Interestingly, the mirror Chern number nM = 2 defined on the kz = 0 plane and the number of Fermi arcs found at each Dirac node are consistent with each other. We first show the topological phase transition from the type-II Dirac semimetal state to a topological crystalline insulator state. We break the C4z rotational symmetry by compressing the lattice along the x direction which makes a 6= b. This opens up a gap at VAl3's type-II Dirac nodes. Because of nM = 2 at kz = 0, the resulting insulating phase is a topological crystalline insulator with two Dirac surface states as shown in Fig. 4d. We now show two consecutive topological phase transitions which transform the type-II Dirac fermions first to quadratic double Weyl fermions with chiral charge ±2 then to linear single Weyl fermions with chiral charge ±1 . As shown in Figs. 4e,f, we apply an Zeeman field along the kz direction. This field breaks time-reversal symmetry but preserves the C4 rotational symmetry. As a result, each type-II Dirac node is found to split into a pair of quadratic Weyl nodes with chiral charge of ±2. Because of the ±2 chiral charge, each quadratic Weyl node is required to have two Fermi arcs. Interestingly, the four Fermi arcs associated with each 8 Dirac node naturally provide the Fermi arcs needed for the pair of quadratic Weyl nodes. Since the C4 rotational symmetry is preserved, the quadratic Weyl nodes are still along the C4 rotational (kz) axis. The k space distribution of the quadratic Weyl nodes (Fig. 4e) breaks time-reversal symmetry. As a result, as shown in Fig. 4e, any (kx, ky) slice whose kz is between the immediate pair of quadratic Weyl nodes has a Chern number of 2. The fact that the BZ carries nonzero chiral number suggests the existence of anomalous Hall current σxy that arises from the quadratic Weyl nodes [46]. Figure 4i shows the calculated band structure along kz in the presence of the Zeeman field, where we see that the type-II Dirac fermion (Fig. 4h) indeed splits into a pair of type-II quadratic Weyl nodes. Each quadratic Weyl fermion disperses quadratically along kx and ky directions (Fig. 4j) but linearly along kz direction (Fig. 4i). In Figs. 4f,g, we further break the C4 rotational symmetry. We find that the C4 breaking splits each double Weyl nodes with chiral charge ±2 into two single Weyl node with chiral charge ±1. Therefore, a net number of four Weyl nodes are generated from a singe type-II Dirac node in VAl3. Depending on whether one compresses the lattice along x or y direction, the splitting of the double Weyl nodes will be along kx or ky direction. Figure 4k-i show the situation where the lattice was compressed along the x direction. We see that each quadric Weyl node split into two single Weyl nodes along ky direction. Landau level spectrum We study the Landau level spectrum of the type-II Dirac fermions in VAl3 (Fig. 5). In the presence of an external magnetic field, electrons in 2D move in cyclotron orbits with quantized energy values, the Landau levels. In a 3D material, the Landau levels further gain dispersions along the direction of the magnetic field. The Landau level spectrum plays a key role in dictating the magneto-transport properties of materials. In Fig. 5a, we show the Landau level spectrum for the type-II Dirac fermions in VAl3. We see that, except the red lines which are the lowest Landau levels, all higher Landau levels are segregated into two groups (the conduction and valence bands) separated by an energy gap. The lowest Landau levels are found to connect the conduction and valence bands across the band gap. The existence of zeroth (0th) Landau levels connecting the band gap is a signature of Dirac/Weyl fermions. In the case of a Weyl fermion, the 0th Landau level is one chiral band (Fig. 5c). In the case of a Dirac fermion, the 0th Landau levels are a pair of counter-propagating chiral 9 bands as a Dirac fermion consists of a pair of Weyl fermions of opposite chirality. This can be easily seen in a type-I Dirac fermion shown in Fig. 5a. For the type-II Dirac fermion in VAl3 (Fig. 5b), the two chiral bands have Fermi velocities of the same sign, which seems to contradict the above picture. This is because the band gap, i.e., the white region in Fig. 5b separating the conduction and valence bands, is heavily tilted due to the type-II character. If one tilts the band gap back to being horizontal, then the two 0th Landau levels are counter-propagating. We find that the type-II character in VAl3 leads to a distinct response in its Landau level spectrum, i.e., the existence of a critical angle of the magnetic field, along which all Landau levels "collapse" into the same energy, giving rise to a large density of states [29, 30]. We start from the condition where the magnetic field is parallel to the tilting direction of the type-II Dirac cone, which is the kz direction in VAl3. In this case, the electrons' cyclotron motions are within the (kx, ky) plane. In the semiclassical picture, the electrons will trace the Fermi contour within this plane. Fig. 5d shows the band dispersion along kkx. Since the dispersion has a typical conical shape, the constant energy contour is a closed loop independent of the chemical potential position (the top panel of Fig. 5g). We now vary the magnetic field direction within the (kx, kz) plane. Figure5e shows the energy dispersion along kθ1 that is perpendicular to the magnetic field Bθ1 (Fig. 5g). We see that the dispersion becomes a tilted cone, while the constant energy contour is a still closed loop (the middle panel of Fig. 5g). As we continue to tilt the magnetic field, there exists a critical angle, at which one of the bands becomes flat. This means that, to the lowest order (k1 term in the k · p theory), the Fermi contour becomes non-closed (the bottom panel of Fig. 5g). Interestingly, our calculations show that, to the lowest order, all Landau levels collapse to the same energy. When considering higher order terms, the Landau levels are still confined to a very narrow energy window at this critical angle, leading to a large density of states (Fig. 5f). Based on the calculated band structure of VAl3, we obtain a critical angle 0.247π (between the magnetic field and kz) for the type-II Dirac fermions in VAl3. Finally, we consider the effects of exchange/Zeeman coupling of the magnetic field. Fig- ures 5h-k show the Landau level spectrum with different magnitudes of the Zeeman field. It can be seen that, the inclusion of an Zeeman coupling moves the two crossings of the two counter-propagating chiral bands closer (Figs. 5h,i). This corresponds to process shown in Figs. 5l-o, where an Zeeman field in VAl3 splits each type-II Dirac node into a pair of 10 quadratic double Weyl nodes with opposite chiral charges ±2 and increasing the Zeeman field will increase the splitting, which moves two of the four double Weyl nodes closer to the ¯Γ point. If one keeps increasing the Zeeman coupling, there exists a critical value of the Zeeman coupling, at which the two crossings meet at kz = 0 and at the same time the band gap closes (Fig. 5j). This corresponds to process shown in Fig. 5n, where the two double Weyl nodes meet and annihilate at the ¯Γ point. If one further increases the Zeeman coupling beyond this critical value, the band gap reopens and two co-propagating chiral bands appear (Fig. 5k). This corresponds to scenario shown in Fig. 5o. Only one pair of double Weyl nodes are left following the annihilation at ¯Γ. Since each double Weyl node has a chiral charge of ±2, it contributes two co-propagating chiral Landau bands (Fig. 5k). The rich and novel Landau level spectra of the type-II Dirac fermions in VAl3 uncovered above suggest distinct and novel magneto-transport phenomena that can be measured in electrical transport, optical transport, and scanning tunneling spectroscopic experiments. [1] Herring, C. Accidental Degeneracy in the Energy Bands of Crystals. Phys. Rev. 52, 365-373 (1937). [2] Weyl, H. Elektron und gravitation. Z. Phys. 56, 330-352 (1929). [3] Turner, A. M. & Vishwanath, A. Beyond Band Insulators: Topology of Semi-metals and Interacting Phases. Preprint at https://arxiv.org/abs/1301.0330 (2013). [4] Hasan, M. Z., Xu, S.-Y., & Neupane, M. Topological Insulators, Topological Dirac Semimet- als, Topological Crystalline Insulators, and Topological Kondo Insulators. in Topological Insulators: Fundamentals and Perspectives edited by F. Ortmann, S. Roche, S. O. Valen- zuela (John Wiley & Sons, 2015). [5] Wang, Z. et al. Dirac semimetal and topological phase transitions in A3Bi (A=Na, K, Rb), Phys. Rev. B 85, 195320 (2012). [6] Liu, Z. K. et al. Discovery of a Three-Dimensional Topological Dirac Semimetal, Na3Bi, Sci- ence 343, 864 (2014). [7] Xu, S.-Y. et al. Observation of Fermi arc surface states in a topological metal, Science 347, 11 294 (2015). [8] Neupane, M. et al. Observation of a three-dimensional topological Dirac semimetal phase in high-mobility Cd3As2 , Nature Commun. 5, 3786 (2014). [9] Borisenko, S. et al. Experimental Realization of a Three-Dimensional Dirac Semimetal, Phys. Rev. Lett. 113, 027603 (2014). [10] Yang, B.-J & Nagaosa, N. Classification of stable three-dimensional Dirac semimetals with nontrivial topology. Nature Commun. 5,4898 (2014). [11] Wan, X., Turner, A. M., Vishwanath, A. & Savrasov, S. Y. Topological semimetal and Fermi- arc surface states in the electronic structure of pyrochlore iridates. Phys. Rev. B 83, 205101 (2011). [12] Burkov, A. A. & Balents, L., Weyl Semimetal in a Topological Insulator Multilayer. Phys. Rev. Lett. 107, 127205 (2011). [13] Huang, S.-M. et al. An inversion breaking Weyl semimetal state in the TaAs material class. Nature Commun. 6, 7373 (2015). [14] Weng, H., et al. Weyl Semimetal Phase in Noncen-trosymmetric Transition-Metal Monophos- phides, Phys. Rev. X 5, 011029 (2015). [15] Xu, S.-Y. et al. Discovery of a Weyl fermion semimetal and topological Fermi arcs. Science 349, 613-617 (2015). [16] Lu, L. et al. Observation of Weyl nodes in a photonic crystal. Science 349, 622-625 (2015). [17] Lv, B. Q. et al. Experimental Discovery of Weyl Semimetal TaAs. Phys. Rev. X 5, 031013 (2015). [18] Mourik, V. et al. Signatures of Majorana Fermions in Hybrid Superconductor-Semiconductor Nanowire Devices. Science 336, 1003-1007 (2012). [19] Xu, J.-P. et al. Experimental Detection of a Majorana Mode in the core of a Magnetic Vortex inside a Topological Insulator-Superconductor Bi2Te3/NbSe2 Heterostructure. Phys. Rev. Lett. 114, 017001 (2015). [20] Matano, K. et al. Spin-rotation symmetry breaking in the superconducting state of CuxBi2Se3. Nature Phys. doi:10.1038/nphys3781 (2016). [21] Parameswaran S. A. et al. A. Probing the chiral anomaly with nonlocal transport in three- dimensional topological semimetals. Phys. Rev. X 4, 031035 (2014). [22] Huang, X. et al. Observation of the chiral anomaly induced negative magneto-resistance in 12 3D Weyl semi-metal TaAs. Phys. Rev. X 5, 031023 (2015). [23] Zhang, C. et al. Observation of the Adler-Bell-Jackiw chiral anomaly in a Weyl semimetal. Nature Commun. 7, 10735 (2016). [24] Grushin, A. G. Consequences of a condensed matter realization of Lorentz violating QED in Weyl semi-metals. Phys. Rev. D 86, 045001 (2012). [25] Bergholtz, E. J., Liu, Z., Trescher, M., Moessner, R. & Udagawa, M. Topology and Interactions in a Frustrated Slab: Tuning from Weyl Semimetals to C > 1 Fractional Chern Insulators. Phys. Rev. Lett. 114, 016806 (2015). [26] Trescher, M., Sbierski, B., Brouwer, P. W. & Bergholtz, E. J. Quantum transport in Dirac materials: Signatures of tilted and anisotropic Dirac and Weyl cones. Phys. Rev. B 91, 115135 (2015) [27] Soluyanov, A. A. et al. Type II Weyl Semimetals. Nature 527, 495-498 (2015). [28] Udagawa, M. & Bergholtz, E.J. Field-Selective Anomaly and Chiral Charge Reversal in Type- II Weyl Materials. Preprint at http://arxiv.org/abs/1604.08457 (2016). [29] Yu, Z., Yao, Y. & Yang, S. A. Unusual Magneto-Response in Type-II Weyl Semimetals. Preprint at http://arxiv.org/abs/1604.04030 (2016). [30] O'Brien, T. E., Diez, M. & Beenakker, C. W. J. Magnetic breakdown and Klein tunneling in a type-II Weyl semimetal. Preprint at http://arxiv.org/abs/1604.01028 (2016). [31] Zyuzin, A. A. & Tiwari, R. P. Intrinsic Anomalous Hall Effect in Type-II Weyl Semimetals. Preprint at http://arxiv.org/abs/1601.00890 (2016). [32] Sun, Y. et al. Prediction of the Weyl semimetal in the orthorhombic MoTe2. Phys. Rev. B 92, 161107 (2015). [33] Chang, T.-R. et al. Prediction of an arc-tunable Weyl Fermion metallic state in MoxW1−xTe2. Nature Commun. 7,10639 (2016). [34] Wang, Z. et al. MoTe2: Weyl and Line Node Topological Metal. Preprint at http://arxiv.org/abs/1511.07440 (2015). [35] Belopolski, I. et al. Unoccupied electronic structure and signatures of topologi- cal Fermi arcs in the type-II Weyl semimetal candidate MoxW1−xTe2. Preprint at http://arxiv.org/abs/1512.09099 (2015). [36] Belopolski, I. et al. Measuring Chern numbers above the Fermi level in the Type II Weyl semimetal MoxW1−xTe2. Preprint at http://arxiv.org/abs/1604.07079 (2016). 13 [37] Bruno, N. et al. Surface states and bulk electronic structure in the candidate type-II Weyl semimetal WTe2. Preprint at http://arxiv.org/abs/1604.02411 (2016). [38] Aut´es, G. et al. Robust Type-II Weyl Semimetal Phase in Transition Metal Diphosphides XP2 (X = Mo, W). Preprint at http://arxiv.org/abs/1603.04624 (2016). [39] Xu, S.-Y. et al. Discovery of Lorentz-violating Weyl fermion semimetal state in LaAlGe ma- terials, Preprint at http://arxiv.org/abs/1603.07318 (2016). [40] Chang, G. et al. A strongly robust Weyl fermion semimetal state in Ta3S2. Preprint at http://arxiv.org/abs/1512.08781 (2015). [41] Koepernik, K. et al. TaIrTe4 a ternary Type-II Weyl semi-metal. Phys. Rev. B 93, 201101(R) (2016) [42] Maas, J. H., Bastin, G. F. & van Loo, F. J. J., Metselaar, R. The texture in diffusion- grown layers of trialuminides MeAl3 (Me = Ti, V, Ta, Nb, Zr, Hf) and VNi3. Zeitschrift fur Metallkunde 74 294-299 (1983). [43] Isobe H. & Nagaosa, N. Coulomb Interaction Effect in Weyl Fermions with Tilted Energy Dispersion in Two Dimensions. Phys. Rev. Lett. 116, 116803 (2016). [44] Muechler, L., Alexandradinata, A., Neupert, T. & Car, R. Tilted (2D) Dirac Fermions. Preprint at http://arxiv.org/abs/1604.01398 (2016). [45] Chiu, C.-K., Chan, Y.-H. Li, X. Nohara, Y. & Schnyder, A. P. Type-II Dirac surface states in topological crystalline insulators. Preprint at http://arxiv.org/abs/1606.03456 (2016). [46] Burkov, A. A. Anomalous Hall Effect in Weyl Metals. Phys. Rev. Lett. 113 187202 (2014). 14 FIG. 1: Symmetry origins for the type-II Dirac fermions in VAl3. a,b, VAl3 crystalizes in a body centered tetragonal Bravais lattice with lattice constants a = 3.78 A and c = 8.322 A and space group I4/mmm (No. 139). The crystals unit cell is shown with the blue and red spheres representing the V and Al atoms, respectively. As can be seen, the crystal lattice has spatial inversion symmetry, C4z symmetry, and mirror reflection planes Mx and Mz. c, The bulk Brillouin zone (BZ) of VAl3 with the relevant high-symmetry points marked with red dots. The (100) surface BZ is also shown with its projected high-symmetry points. d, In the absence of SOC, the triply-degenerate point along the Γ-Z direction is formed by a doubly degenerate conduction band (red band) and singly degenerate valence band (blue-dashed band). Defined by C4z operator, the ±i eigenvalues of the doubly-degenerate band and −1 eigenvalue of the singly degenerate bands is shown. e, In the presence of SOC, every band becomes doubled. The initial doubly degenerate conduction band in the no SOC case gets turned into two doubly degenerate bands conduction bands (blue and red bands).The resulting new eigenvalues for the blue and red doubly degenerate conduction bands are (ei 3π 4 , ei 5π 4 ) and (e iπ 4 , ei 7π 4 ), respectively. FIG. 1: The initial singly degenerate valence band in the no SOC case get turned into a doubly 15 degenerate band (blue band) with eigenvalues (ei 3π 4 , ei 5π 4 ). f, Enclosed by green circles in panel (e) are the touching points between two doubly-degenerate blue bands with the same C4z eigenvalues, which results in these states becoming gapped. However, the two doubly degenerate red and blue bands do not share the same eigenvalues and, therefore, generate crossing points that are not gapped. The new generated crossing points are the type-II Dirac nodes. 16 FIG. 2: Type-II Dirac fermions in the bulk band structure of VAl3. a, The calculated bulk band structure of VAl3 in the presence of spin-orbit coupling. The green shaded region shows the energy gap between the lowest conduction and valence bands. b, c A zoomed-in calculation of the band dispersion along the kz (b) and kx/ky (c) directions in the vicinity of the type-II Dirac node. d, A zoomed-in view of the area highlighted by the gray box in (a). Each of the two crossing bands is marked with their corresponding space group representations Λ6 and Λ7. Because the two bands touching have different space group representations, they are prevented from hybridizing and gapping out. The type-II character of the Dirac node is clearly shown in this plot (i.e. there are both electron and hole bands touching at the energy of the Dirac node). e, Bulk Fermi surface of VAl3 with the red and blue pockets denoting the electron and hole bands, respectively. f, Bulk constant energy contour in (ky, kz) space at kx = 0 and at the energy of the type-II Weyl nodes. An electron pocket (red contour) enclosing the Z point and two hole pockets (blue contours) touch to form a pair of type-II Dirac nodes. 17 FIG. 3: Fermi arc surface states associated with the type-II Dirac nodes. a, Surface and bulk band structure of VAl3 along ¯Γ- ¯N direction on the (100) surface BZ. Two topological surface states are shown to be terminating directly onto the projected type-II Dirac node. The normal surface states are observed to avoid the Dirac node and merge into the bulk band continuum. b, The calculated kz-ky surface and bulk electronic structure at the energy of the bulk Dirac node (∼ 28 meV above EF ). Enclosing ¯Γ is a electron pocket and extending along ¯Γ- ¯N is a hole pocket. The normal surface states are labeled with arrows and the projected type-II Dirac nodes are marked with red and black dots, which reside at the touching point between the electron and hole pockets along the ¯Γ- ¯N . c, Zoomed-in view of the area highlighted by the black box that surrounds the projected type-II Dirac node in (b). Here we observe two pairs of topological surface states emerging from the Dirac node and merging into the bulk band continuum to connect the pair of Dirac nodes along ¯N -¯Γ- ¯N . The fact that the topological surface states are pinned to the Dirac node is an interesting result because they are not constrained by the mirror Chern number to behave in this 18 FIG. 3: fashion and has a net zero Chiral charge. d, The (kx, kz) side of three bulk Brillouin zones is shown to better understand the projection of Dirac nodes (blue dots) and relevant high-symmetry points (red dots) on the (100) surface BZ. e, Similar to panel (d) but for two bulk Brillouin zones that are tilted in such a way that it becomes more clear where the projected Dirac nodes (blue dots) and relevant high-symmetry points (red dots) are positioned on the (100) surface BZ. As shown the Dirac nodes are projected along the ¯Γ − ¯N high symmetry line. 19 FIG. 4: Topological phase transitions in VAl3. a-c, A schematic of the Fermi surface for Na3Bi, (b), undergoing a transition to a topological insulating phase by breaking the C3z rotational symmetry, or a Weyl semimetal phase, (c), by applying an Zeeman field along the rotational symmetry preserving axis. d-e, Topological phase transitions in VAl3. e, an illustration of surface Fermi arcs and type-II Dirac nodes. The double Fermi arcs arise because of the defined nM = 2 mirror Chern number on the kz = 0-mirror plane. 20 FIG. 4: d, Breaking the C4z rotational symmetry drives the type-II Dirac semimetal into a topological crystalline insulator phase, which results in two concentric topological surface states protected by the Mz symmetry, as shown in panel (d). e, Breaking time-reversal symmetry with an applied Zeeman field along the kz-axis splits each Type-II Dirac node into a pair of double Weyl nodes, each with chiral charge ±2. f, By further applying a C4z symmetry breaking perturbation, each double Weyl nodes split into two singleWeyl nodes with ±1 chiral charge. A net number of four Weyl nodes are generated from a singe type-II Dirac node. For panels (c), (f), and (g), the net Chern number for the regions defined by dashed lines is shown. h, The calculated band dispersion of the Type-II Dirac cone in VAl3 along the kz directions. i, The calculated band dispersion in the presence of an Zeeman field. As shown, the type-II Dirac node splits into a pair (red and blue) type-II double Weyl cones with Weyl nodes of equal but opposite chiral charge ±2. j, The band dispersion along the kx/ky direction for the type-II double Weyl cone with −2 chiral charge in panel (i). As expected, the Weyl node with ±2 chiral charge is defined by the touching point between two quadratic bands. k-m, By breaking the C4z symmetry, the double Weyl node in panel (j) splits into two single Weyl nodes with an equal chiral charge of −1. The zoomed-in band dispersions along kx, ky and kz directions for the single Weyl nodes. 21 FIG. 5: Landau level spectrum of the type-II Dirac fermions in VAl3. 22 FIG. 5: a, Landau level spectrum of the type-II Dirac fermions in VAl3. The magnetic field is along the tilting direction of the type-II Dirac fermions, which is the kz direction for VAl3. b,c, Same as panels (a,b) but for type-I Dirac fermions and type-II Weyl fermions. d-f, We vary the magnetic field direction within the (kx, kz) plane. θ defines the angle between the magnetic field and kz. The dispersions in panels (d-f) are along the k space directions (kx, kθ1, kθ2) shown in panel (g). They are within the (kx, kz) plane and at the same time they are perpendicular to the magnetic field directions. g, Schematic illustration of different magnetic field directions and also the k directions (kx, kθ1, kθ2) that are perpendicular to the magnetic fields. The three insets on the top-right corner shows the constant energy contours in the k plane that is perpendicular to the magnetic field. h-k, Landau level spectra of the type-II Dirac fermions in VAl3 under different Zeeman couplings. The lowest Landau levels change from two counter-propagating chiral bands to two co-propagating chiral bands. This corresponds to the annihilation of the two quadratic double Weyl fermions at the ¯Γ point as shown in panels (l-o). 23 Supplementary Information: VAl3 family materials FIG. S1: Band structure of VAl3 family. a, Calculated bulk band structure of VAl3 with the inclusion of SOC. b, Calculated bulk band structure of NbAl3 with the inclusion of SOC. c, Calculated bulk band structure of TaAl3 with the inclusion of SOC. d, Calculated bulk band structure of NbGa3 with the inclusion of SOC. e, Calculated bulk band structure of TaGa3 with the inclusion of SOC. f, A zoomed-in view of a for the area highlighted by the gray box. g, A zoomed-in view of b for the area highlighted by the gray box. h, A zoomed-in view of c for the area highlighted by the gray box. i, A zoomed-in view of d for the area highlighted by the gray box. j, A zoomed-in view of e for the area highlighted by the gray box. 24 FIG. S2: Band structure of k · p model a, Calculated band dispersion of k · p model along kz direction. The fitting parameters α0 = -0.0274, α5 = -0.0179, β0 = 0.3730, β5 = 0.1831, γ0 = 2.0574, γ5 = 0.8662, η = 1.6590 - i0.7754, δ = -(0.8351 + i0.0301), and ξ = -(0.0048 + i0.1961) were used in this work. b, Calculated band dispersion of k · p model along kx and ky directions. c, Calculated band dispersion of k · p model along kz direction under kz exchange field. d, Calculated band dispersion of k · p model along kx and ky directions under kz exchange field. e, Calculated band dispersion of k · p model along kz direction when α0 = γ0 = 0. f, Calculated band dispersion of k · p model along kx and ky directions when α0 = γ0 = 0. 25 k · p model. To better understanding the fundamental properties of Type-II Dirac semimetal, we synthesize a k.p model is as follow. Let us consider 4 × 4 matrix minimal Hamiltonian, which general has the following form, H(k) = aij(k)σiτj =   3X i,j=0 h↑↑(k) h↑↓(k) h↓↑(k) h↓↓(k)   (1) where the σ0 and τ0 are identity matrices. σ1,2,3 and τ1,2,3 are Pauli matrix which indicates spin and orbital degree of freedom, respectively. We assume aij(k) and hσσ′ are real functions. To fit the VAl3, we constrain the form of H(k) by time-reversal symmetry (TRS), spatial inversion symmetry (IS), and C4 rotational symmetry along the z-axis. At first, we impose TRS on H(k). The TRS can be represented by the operator Θ = iσ2κ, where κ denotes complex conjugation and σ2 is the second Pauli matrix acting on the electron spin. Under TRS, H(k) requires ΘH(k)Θ−1 = H(−k), giving rise to h↓↓(k) = h∗ ↑↑(−k) and h↓↑(k) = -h∗ ↑↓(−k). Thus Hamiltonian can be written in the following form. H(k) =   h↑↑(k) h↑↓(k) −h∗ ↓↑(−k) h∗ ↓↓(−k)   (2) Second, we impose IS on H(k). The invariance of the H(k) under IS, that is, P H(k)P −1, where P is IS operator. Since P is independent of the spin rotation, P can be written P = α0τ0 + ·τ , where α1,2,3,4 are complex numbers. In addition, we claim because P 2 = 1 because it should be equivalent to the identity operator up to a global U(1) phase factor, that is, P 2 0 + α2+2α0 α·τ . Therefore, P = ±eiφτ0 or eiπα′·τ , where α′2 = 1. To determine α′ and φ, three conditions constrain are needed, (1) [Θ, P ] = 0, (2) P †P = 1, and (3) (ΘP )2 = ei2φ ·I = α2 = -1. Thus, the general solution of P is P = ±τ0 or cosθτ3 - sinθτ1 with θ ∈ [0, 2π]. In this work, we pick up P = τ0. Finally, let us consider rotational symmetry on H(k). We choose kz as a rotation axis and C4 matrix are diagonal C4 = diag(uA↑, uB↑, uA↓, uB↓) = diag(αp, αq, αr, αs), where αp = ei π 2 ) with p = 0, 1, 2, 3. Thus C4 can be written in 2 (p+ 1 following form. C4 =   eiπ( 1+p+q 4 )+ p−q 4 τ3 0 0 eiπ( 1+r+s 4 )+ r−s 4 τ3   (3) The H(k) under C4 symmetry results in C4H(k+, k−, kz)C −1 2 , kz). where 4 = H(k+ei π 2 , k−e−i π k± = kx ±iky. From this invariance, we can obtain ei π 4 (p−q)τ3h↑↑(k+, k−, kz)e−i π 4 (p−q)τ3 = h↑↑(k+ei π 2 , k−e−i π 2 , kz), ei π 4 (p+q−r−s)ei π 4 (p−q)τ3h↑↓(k+, k−, kz)e−i π 4 (r−s)τ3 = h↑↓(k+ei π 2 , k−e−i π 2 , kz) Since [Θ, C4] =0, C4 can be written C4 = diag(αp, αq, α∗ p, α∗ q). Therefore eq.(4) becomes ei π 4 (p−q)τ3h↑↑(k+, k−, kz)e−i π 4 (p−q)τ3 = h↑↑(k+ei π 2 , k−e−i π 2 , kz), ei π 4 (p−r)ei π 4 (p−q)τ3h↑↓(k+, k−, kz)e−i π 4 (p−q)τ3 = h↑↓(k+ei π 2 , k−e−i π 2 , kz) In terms of τ and hσσ′ can be expressed as h↑↑(k) = f0(k) + f+(k)τ+ + f ∗ +(k)τ− + fz(k)τ3, h↑↓(k) = g0(k) + g+(k)τ+ + g∗ −(k)τ− + gz(k)τ3 26 (4) (5) (6) where τ± = τ1 ± iτ2, and due to hermitian requirement, f0,z are real functions and f+, g0,±,z are complex functions. From eq.(5) and eq.(6), we obtain f0(k+, k−, kz) = f0(k+ei π 2 , k fz(k+, k−, kz) = fz(k+ei π 2 , k ei π 2 (p−q)f+(k+, k−, kz) = f+(k+ei π −i π − , kz), 2 −i π − , kz), 2 −i π − , kz) 2 2 , k and ei π 2 (p−r)(g0 + gz)(k+, k−, kz) = (g0 + gz)(k+ei π 2 , k ei π 2 (q−s)(g0 − gz)(k+, k−, kz) = (g0 − gz)(k+ei π −i π − , kz), 2 2 (q−r)g±(k+, k−, kz) = g±(k+ei π 2 , k 2 , k ei π −i π − , kz), 2 −i π − , kz), 2 (7) (8) i,j=0 aij(k)σiτj becomes H(k) = P5 Combining with TRS and IS, H(k) = P3 i=0eai(k)Γi, where Γ0 = σ0τ0, Γ1 = τ1, Γ2 = σ3τ2, Γ3 = σ1τ2, Γ4 = σ2τ2, and Γ5 = τ3, and eai are real and even functions of k. Hence h↑↑ = ea0 + ea1τ1 + ea2τ2 + ea5τ3 and h↑↓ = (ea3 - iea4)τ2. Imposing C4 constrain, the form of H(k) becomes H(0, 0, kz) ∼ Pi=0eai(0, 0, kz)Γi with Γi along kz axis only include Γ0 and Γ5. When P = τ0, we can obtain eai(0, 0, kz) = eai(0, 0, −kz) for i=0, 5 and eai(0, 0, kz) =0 for i = 1, 2, 3,4. Thus h↑↑ and h↑↓ can be written in terms of f and g functions. h↑↑ = f0(k) + f+(k)τ+ + f ∗ +(k)τ− + fz(k)τ3, h↑↓ = g0(k) + g+(k)τ+ + g−(k)τ− + gz(k)τ3 (9) where f0(k) = ea0(k), f+(k) = (ea1(k) - iea2(k))/2, f ∗ +(k) = (ea1(k) + iea2(k))/2, fz(k) = ea5(k), g+(k) = −i(ea3(k) - iea4(k))/2, g−(k) = i(ea3(k) - iea4(k))/2, and g0(k) = gz(k) = 0. According to eq.(7) and eq.(8), we have 27 2 , k−ei π 2 , kz), 2 , k−ei π 2 , kz), ea0(k+, k−, kz) = ea0(k+ei π ea5(k+, k−, kz) = ea5(k+ei π 2 [ea1(k+ei π 2 [ea3(k+ei π 2 , k−ei π 2 , k−ei π ei π 2 (p−q)f+ = 1 ei π 2 (q−r)g± = ∓ i 2 , kz) − iea2(k+ei π 2 , kz) − iea4(k+ei π 2 , k−ei π 2 , kz)], 2 , k−ei π 2 , kz)] (10) As f+(0, 0, kz) = g±(0, 0, kz) = 0, ei π and αr 6= αq. Hence αp, αq, α∗ p, and α∗ and g functions can be determined. 2 (p−q) and ei π 2 (q−r) cannot be 1. In other words, αp 6= αq q are different rotational eigenvalues. Now the eai, f ea0,5 = α + βk+k− + γk2 f+(k+, k−, kz) = ηk+kz, z , g+(k+, k−, kz) = δk2 + + ξk2 −, g−(k+, k−, kz) = −g+(k+, k−, kz) (11) where α, β, γ and η, δ, ξ are real numbers and complex numbers, respectively. Therefore H(k) can be written in the following form. H(k) =   D1 2η∗k−kz 0 2(δ∗k2 − + ξ∗k2 +) 2ηk+kz −2(δ∗k2 D2 − + ξ∗k2 +) 0 −2(δk2 0 + + ξk2 −) D1 2ηk+kz 2(δk2 + + ξk2 −) 0 2η∗k−kz D2   (12) where D1 = (α0 + α5) + (β0 + β5)k+k− + (γ0 + γ5)k2 (γ0 − γ5)k2 z z, and D2 = (α0 − α5) + (β0 − β5)k+k− + Figure SS2 shows the band structure of k · p model. The fitting parameters α0 = -0.0274, α5 = -0.0179, β0 = 0.3730, β5 = 0.1831, γ0 = 2.0574, γ5 = 0.8662, η = 1.6590 - i0.7754, δ = -(0.8351 + i0.0301), and ξ = -(0.0048 + i0.1961) were used to diagonalize H(k). Our calculation shows a clear Type-II Dirac band dispersion with a four-fold degenerate Dirac point (Fig. SS2(a)). We also check the band structures along kx and ky directions in the vicinity of gapless points and found that the bands disperse linearly along all directions (Fig. 28 SS2(b)), which consistent very well with our DFT results. In the presence of exchange field Mzσ3, two doubly degenerate Dirac bands were split into four singly degenerate bands. The band dispersion along kz direction shows Type-II Weyl band dispersion (Fig. SS2(c)) and depicts double Weyl quadratic band dispersion along kx and ky directions (Fig. SS2(d)). Furthermore, our model not only catch the main features of VAl3 but also providing a platform to study Type-I Dirac and Weyl band dispersion. Setting α0 = γ0 = 0, our model presents a Type-I and dispersion along all three directions (Fig. SS2(e) and (f)).
1903.02455	1	1903	2019-03-06T15:47:48	Effects of geometry on spin-orbit Kramers states in semiconducting nanorings	[ "cond-mat.mes-hall" ]	The holonomic manipulation of spin-orbital degenerate states, encoded in the Kramers doublet of narrow semiconducting channels with spin-orbit interaction, is shown to be intimately intertwined with the geometrical shape of the nanostructures. The presence of doubly degenerate states is not sufficient to guarantee a non-trivial mixing by only changing the Rashba spin-orbit coupling. We demonstrate that in nanoscale quantum rings the combination of arbitrary inhomogeneous curvature and adiabatic variation of the spin-orbit amplitude, e.g. through electric-field gating, can be generally employed to get non-trivial combinations of the degenerate states. Shape symmetries of the nanostructure act to constrain the adiabatic quantum evolution. While for circular rings the geometric phase is not generated along a non-cyclic path in the parameters space, remarkably, for generic mirror-symmetric shape deformed rings the spin-orbit driving can lead to a series of dynamical quantum phase transitions. We explicitly show this occurrence and propose a route to detect such topological transitions by measuring a variation of the electron tunneling amplitude into the semiconducting channel.	cond-mat.mes-hall	cond-mat	Effects of geometry on spin-orbit Kramers states in semiconducting nanorings G. Francica,1 P. Gentile,1 and M. Cuoco1 1CNR-SPIN, c/o Universit`a di Salerno, I-84084 Fisciano (Salerno), Italy The holonomic manipulation of spin-orbital degenerate states, encoded in the Kramers doublet of narrow semiconducting channels with spin-orbit interaction, is shown to be intimately intertwined with the geometrical shape of the nanostructures. The presence of doubly degenerate states is not sufficient to guarantee a non-trivial mixing by only changing the Rashba spin-orbit coupling. We demonstrate that in nanoscale quantum rings the combination of arbitrary inhomogeneous curva- ture and adiabatic variation of the spin-orbit amplitude, e.g. through electric-field gating, can be generally employed to get non-trivial combinations of the degenerate states. Shape symmetries of the nanostructure act to constrain the adiabatic quantum evolution. While for circular rings the geometric phase is not generated along a non-cyclic path in the parameters space, remarkably, for generic mirror-symmetric shape deformed rings the spin-orbit driving can lead to a series of dynam- ical quantum phase transitions. We explicitly show this occurrence and propose a route to detect such topological transitions by measuring a variation of the electron tunneling amplitude into the semiconducting channel. Introduction -- The interest in geometric phases in quantum mechanics has been turned on by the seminal works of Berry and Simon [1, 2]. For an adiabatic and cyclic evolution, it is known that the quantum state ac- quires a phase factor that depends only on the path in the parameters manifold. Indeed, the geometric phase is due to a holonomy for the given fibre bundle [2]. For degener- ate quantum systems the phase factor is non-Abelian, as firstly pointed out by Wilczek and Zee for a cyclic evolu- tion [3]. Geometric phases have been then generalized for non-cyclic paths [4 -- 7] by showing their intimate relation with the geometric structures of the projective Hilbert space [8 -- 10]. One of the major challenges in the context of quan- tum processing points to the achievement of novel mech- anisms and devices which are based on non-Abelian geo- metric phases, as they are robust to local perturbations and can provide means for performing universal quantum information processing [11, 12] with noteworthy perspec- tives towards the realization of geometric quantum com- putation [13 -- 17] . In this framework, low-dimensional semiconductors with inversion asymmetry can be par- ticularly attractive [18 -- 22] because the spin-orbit (SO) interaction [23 -- 25] offers a tantalizing prospect of an all- electrical control over the electron spin in absence of a magnetic field. Furthermore, the electrical manipula- tion preserves time-reversal symmetry which is crucial to guarantee degenerate Kramers pair configurations by which a qubit can be in principle encoded and quantum processed. Along this direction, schemes to perform non-Abelian holonomic operations on the electron spin have been mainly focusing on time-dependent electrostatic con- fining potentials realized through moving quantum dots [26], also including closed loop trajectories [27, 28]. Although experimental realizations of moving quantum dots have been successfully achieved [29], the efficiency of such quantum engineering is sensitively dependent on the concomitant dynamical control of the confining elec- trostatic potential and the strength of the inversion sym- metry breaking via the SO coupling. It would be then highly beneficial to develop alternative paths which can provide expanded functionalities and increase the phase space for quantum manipulation. In this Letter, we aim at demonstrating how spin- orbital quantum states, encoded in the Kramers doublet, can be engineered in nanoscopic semiconducting chan- nels using the combined effects of geometric curvature and Rashba spin-orbit coupling. The potential of this union has already led to augmented paths for the design of topological states [30 -- 34] and electron-transport [35 -- 37]. Such effects mainly arise due to the fundamental role of nanoscale shaping in narrow SO coupled semiconduct- ing channels acting as a source of spatial dependent spin- torque controlling both the electron spin-orientation and its spin-phase through non-trivial spin windings [31, 34]. Here, we show that asymmetrically shaped nanostruc- tures can generally lead to non-trivial mixing of the states forming the Kramers doublet by an adiabatic driving of the spin-orbit amplitude, e.g. through electric-field gat- ing. Moreover, since the geometric curvature of SO cou- pled nanostructures can generate spin-texture with non- trivial topological windings in real space, it is rele- vant to evaluate whether such geometrical resource can lead to quantum driven topological phase transitions. Recently, a subtle interconnection between topological phase changeover and dynamical quantum phase transi- tions (DQPT) [38] has been put into limelight in a variety of quantum systems [38, 39], including low-dimensional topological systems [40]. Remarkably, for a given sym- metry class, a topological invariant, represented by a momentum space winding number of the Pancharatnam geometric phase, can be introduced and employed for characterizing DQPTs [41]. In this paper, for suitable shaped nanostructures with mirror symmetry, we find that the amplitude's variation of the Rashba SO cou- pling can lead to a series of dynamical topological phase transitions. A controlled amplification of the Rashba in- teraction is feasable by electric-field gating [42, 43] as it curvilinear coordinate s. The spatial dependent spin- orbit coupling is expressed via two local Pauli matrices, comoving with the electrons as they propagate along s, expressed by σN (s) = σ · N (s) and σT (s) = σ · T (s), where the σ's are the usual Pauli matrices. T and N are related via the Frenet-Serret type equation ∂s N = κ T with κ being the local curvature. A Rashba SO coupling arises due to the inversion symmetry breaking and can be tuned by an applied electric field transverse to the plane of the nanostructure. An effective model descrip- tion that is able to capture the combination of Rashba spin-orbit coupling and geometrical curvature is given by the Hamiltonian [30, 47 -- 49] 2α 2m (σN (s)∂s + ∂sσN (s)) ∂2 s + i H = − 2 2m 2 (1) where m is the effective mass of the charge carrier, α is the Rashba SO coupling strength and σN is the nor- mal component of the spin with respect to the nanos- tructure geometrical profile. Since we are interested in assessing the role of the geometry to set the character of the Kramers pairs quantum evolution, we consider dif- ferent types of spatial profiles focusing on the ensuing symmetries. The circular quantum nanoring is a highly symmetric case and, thus, it represents an ideal reference with uniform curvature and invariance under continuous rotation around the axis perpendicular to the electron orbital plane. A deviation from the circular shape can bring to two possible paths of shape deformations: a first one preserving few specific point group symmetries of the quantum loop and a second direction corresponding to nanoring with an arbitrary shape. We consider the class C1 of nanoring geometry with point symmetry transfor- mations including the rotation around the z axis (or in- plane inversion), P : (x, y, 0) (cid:55)→ (−x,−y, 0), and a sub- class C2 ⊂ C1 for which there is also a mirror symmetry with respect to the reflection Py : (x, y, 0) (cid:55)→ (x,−y, 0) or equivalently Px. The computational analysis is per- formed by numerically solving the model Hamiltonian (1) and deals with the case of nanoring with circular, ellip- tical or generic asymmetric shape. The elliptical config- uration is a representative and paradigmatic example of planar inversion and mirror symmetric profiles exhibiting an inhomogeneous curvature. In general, for any geomet- rical shape, the model Hamiltonian is symmetric with re- spect to the time-reversal transformation Θ, so that the energies E are two-fold degenerate and the eigenstates arise in Kramers pairs. For inversion symmetric profiles of the nanoring, the Hamiltonian is also invariant under the transformation M = σzP , so that a Kramers pair (E+(cid:105),E−(cid:105)) can be classified such that ΘE+(cid:105) = E−(cid:105) and ME±(cid:105) = ±E±(cid:105), and the Hilbert space is the direct sum H = H+⊕H− with the invariant subspace H± being spanned by {E±(cid:105)}E. It has been shown in Refs. [31, 34] that the elec- tron spin orientation manifests topological features ex- pressed by windings of the spin direction along the spa- tial profile. In particular, for a nonuniform curvature FIG. 1. (a) schematic representation of the spin-orbital Kramers pairs with time-reversal symmetric spin-texture along a curved spin-orbit coupled nanochannel (blue line). Sketches of solid-state platform with applied gate voltage Vg for tuning the strength of the Rashba spin-orbit coupling. A schematic view of the adiabatic evolution of spin-orbital 0(cid:105) and 1(cid:105) states, associated to the spin-textures in (a), is rep- resented on an effective Bloch sphere with the corresponding trajectory (red solid line) for the case of (b) constant curva- ture, (c) mirror symmetric and generically d) shaped nanos- tructure. The dotted line (orange) indicates the geodesic for closing the path for the case of a non-cyclic evolution and provide the geometric phase. is a consolitated practice in a variety of semiconducting nanostructures including small quantum rings. Advance- ments in the design of small nanoring with few electrons [44 -- 46] and different shapes make our proposal accessi- ble in laboratory. To this end, we discuss possible setups for detecting the spin-orbit driven quantum topological transitions. Model system -- We consider a system of electrons prop- agating in a narrow semiconducting channel lying in the xy-plane and forming a spatial profile with a nontrivial spatial curvature. The shape of the narrow nanostruc- ture can be generally specified by introducing two unit vectors T (s) and N (s), which are tangent and normal to the spatial profile at a given position labelled by the b) c) d)VgVgVgVga)the spin orientation (cid:104)σ(cid:105) with respect to the Frenet-Serret frame displays spin textures which correspond to loops on the Frenet-Serret-Bloch sphere, and are characterized by windings around the normal and the out-of-plane direc- tions. With the same spirit, another topological feature of the state can be introduced by considering the rela- tive phase χ = arg ψ (L)ψ(0) = 2πw where ψ indicated the spin configuration of the state, w is an integer and L is the length of the nanoring. These topological aspects are relevant and play an important role in setting the geometric phase produced in the quantum evolution. † Spin-orbit driven quantum geometric phase -- Let us start by providing few general considerations about the phase acquired by a given state for non-cyclic adia- batic time evolution which is achieved by changing suf- ficiently slowly the Rashba SO coupling from the ini- tial value αin into the final one α in a given nanor- ing. If the system is initially prepared in an eigenstate Ea(αin)(cid:105), the adiabatically evolved state Ψa(α)(cid:105) will be a linear combination of the degenerate eigenstates, i.e. Ψa(α)(cid:105) =(cid:80) Ub,a(α)Eb(α)(cid:105) at least of an irrelevant dy- namical phase factor, so that the evolved state acquires a non-Abelian phase factor U (α) which depends on the ba- sis chosen {Ea(α)(cid:105)}a with the eigenstates smooth in α. The holonomic contribution can be taken in account for an open path in the projective Hilbert space by a unitary matrix Ug(α) having gauge-invariant eigenvalues [6, 7], which are e±iγ due to the time reversal symmetry. By considering the initial state E+(αin)(cid:105), the evolu- tion occurs in the invariant subspace H+ and the evolved state can be expressed as Ψ+(α)(cid:105) = eiγ E+(α)(cid:105), where the state E+(α)(cid:105) is defined from E+(α)(cid:105) by fixing the gauge so that arg(cid:104)E+(αin) E+(α)(cid:105) = 0. The phase γ can be so viewed as a Pancharatnam phase [4, 5], and can be expressed as the integral of the Berry-Simon connection evaluated along the curved electronic nanochannel ψ γ = i∂αψ or equivalently in a more compact form γ = (cid:82) A = (cid:82) (cid:104) E+d E+(cid:105), where ψ is the spin component of the state E+(cid:105), which satisfies the differential equation [31] i∂sψ(s) = Q(s)ψ(s), with Q(s) = −ασN (s) − c0, and 0 = 2m2 E + α2. c2 During the evolution, due to the continuous change in the spin direction at the point s, one can identify a local geometric phase γB(s) [4, 5]. This part is uniquely related to the changing in the local spin direction during the evolution, and it is half the solid angle swept by the geodesic closure of the path spanned by the local spin direction during the evolution. Then, the phase γ can be expressed as (cid:90)(cid:90) † (cid:90) L 0 3 where ∆λ = −(cid:82) L 0 ϕ(s)ρds, ϕ = arg(ψ (s, αin)ψ(s, α)) ψ is the spatially con- † † being the total phase, and ρ = ψ stant electron density. Starting from the expression of the phase γ we observe that ∆λ is related to an eventual changing in the wind- ing number ∆w = w(α) − w(αin) through the general equation ∆λ = π∆w − 1 2 (ϕ(s) + ϕ(−s))ρds . (3) Let us consider the geometric phase for various spatial symmetries of the nanoring. The circular quantum ring is a special case of mir- ror symmetric profile because, due to the constant ra- dius, it has also invariance under rotation around the axis perpendicular to the ring plane. Such symmetry aspect is further constraining the possible values of the acquired phase along the adiabatic evolution. Indeed, under a rotation of an angle θ with respect to the z- axis we have ψ (cid:55)→ ei(w− 1 2 )θψ where w is the wind- ing number previously introduced, from which the or- G(α) = thogonality condition cannot be reached, i.e. (cid:104)E+(αin) E+(α)(cid:105) = ψ (s, αin)ψ(s, α)L > 0. Further- more, exploiting the same argumentation it is straightfor- ward to show that there is only another state, indicated +(αin)(cid:105), having a non-zero overlap with respect with E(cid:48) to the state E+(α)(cid:105), so that 2− σz † E+(α)(cid:105) = G(α)E+(αin)(cid:105) + 1 − G2(α)E(cid:48) +(αin)(cid:105) (4) This parameterizes an arc of great circle in the space of the unit vectors S(H), so that the adiabatic evolution moves along a geodesic in the projective Hilbert space, for which the Pancharatnam phase is zero, i.e. γ = 0 [8, 9]. In particular LF S = arccos G(α) is its length calculated with respect to the Fubini-Study metric [10]. phases is zero(cid:82) γB ρds = 0, and the local total phase is Topological phase transition and non-Abelian phase -- Proceeding with our analysis for mirror symmetric nanostructures, we have that Q(s) = Q†(−s) so that ψ(−s) = Kψ(s), with K being the complex conjugate operator. Taking into account such symmetry relations, one can show that the sum of all the local geometric odd ϕ(s) = −ϕ(−s), so that a variation in the phase γ can be uniquely related to a change in the winding w, i.e. γ = ∆χ/2 = π∆w. The value of ∆w is such that the condition arg(cid:104)E+(αin) E+(α)(cid:105) = 0 is satisfied. We have (cid:113) (cid:90) (cid:104)E+(αin) E+(α)(cid:105) = † ψ (s, αin)ψ(s, α)ds (5) (cid:90) L 0 γ = γBρds + ∆λ (2) Hence, we observe that ψ (s, αin)ψ(s, α) = rei∆χ/2eiφ with r ≥ 0, so that under a reflection with respect to the xz plane, rei∆χ/2eiφ (cid:55)→ rei∆χ/2e−iφ. We then obtain † 4 FIG. 3. Color map of the geometric phase γ in unit of π as a function of ∆αL/2π. A geometric non-Abelian phase is generated in the class C1\C2 and a dynamical phase transition does not occur. We have considered a nanoring parameterized by x = b(1 + sin(4t)) cos t, y = a(1 + sin(4t)) sin t with t ∈ [0, 2π]. For = 0 the nanoring is elliptical. The results refer to a/b = 0.4, and the initial configuration is the ground state at αin = 0. Starting from the previous analysis, we observe that the reflection symmetry makes the phase γ quantized so that it prevents the generation of a linear combination of the Kramers doublet. However, it is enough to consider a small modification of the spatial profile that breaks the mirror symmetry by keeping the inversion in order to get a geometric non-Abelian phase factor Ug with γ (cid:54)= nπ and n an integer. To demonstrate such effect, we apply a slight deformation to the elliptical shape of the nanoring by introducing a suitable parameterization of the coordinates (see inset in Fig. 3). As one can observe, by inspection of Fig. 3, the change of the phase γ in proximity of the critical amplitudes αn becomes smooth when the nanoring is deformed away from the elliptical shape (i.e. = 0). Discussion and conclusions -- We show that a non- Abelian quantum manipulation and dynamical quantum phase transitions can be obtained by combining a vari- ation of the Rashba SO coupling and the geometrical curvature in narrow semiconducting nanorings. Apart from the spatial confinement, a non-uniform cur- vature can gap the ground-state energy from the other levels, so that an adiabatic evolution in the ground-state sector can be achieved through a slow variation of the Rashba SO coupling. In particular the gap can be gen- erally expressed as ∆E = gh2/2mL2, where the factor g < 1 depends on the geometric details of the nanos- tructure. We expect that the adiabatic approximation for the time evolution of the ground state is suitable for energy level and ∆E = E(cid:48) − E is the energy separation. Taking into account the expression of the gap ∆E then η = L2 the so-called Compton wavelength. For a length L ∼ 60 πgλcc (cid:12)(cid:12) dH dt E+(cid:105)(cid:12)(cid:12) /∆E2 (cid:28) 1, where E(cid:48) is the first (cid:12)(cid:12) σNE+(cid:105)(cid:12)(cid:12) with λc = h/mc being (cid:12)(cid:12)(cid:10)E(cid:48) η = (cid:12)(cid:12)(cid:10)E(cid:48) dt L dα 2π + + FIG. 2. The contour map of the amplitude G(α) in unit of π as a function of αinL/2π and ∆αL/2π where ∆α = α−αin and L is the length of the nanoring. The orthogonality condition G = 0 is obtained at the zero Fisher αn (red lines), at which a π-jump in the phase γ occurs. The initial state is the ground state for αin, and the wire is an ellipse with the ratio of the semiaxis length set to a/b = 0.4, so that level crossings with other energies do not occur in the interval under consideration. (cid:104)E+(αin) E+(α)(cid:105) = 2ei∆χ/2 (cid:90) L 2 0 r cos(φ)ds (6) and consequently ei∆χ/2 = eiπ∆w = ±1 in order to keep arg(cid:104)E+(αin) E+(α)(cid:105) = 0 during the evolution. (cid:88) Then for shapes with reflection symmetry, the winding number parity ∆w( mod 2) can change at given values αn of the Rashba SO coupling for which the final evolved state becomes orthogonal with respect to the initial one. The eigenvalue eiγ gives a sign factor, whose value can be obtained from the following connection A = πδ(α − αn)dα (7) n By explicitly evaluating the holonomy for an elliptical nanoring, we find that the adiabatic evolution keeps the evolved state in phase with the initial one until the or- thogonality is reached and the phase γ undergoes a π- jump (see Fig. 2). This behavior can be related to the occurrence of a DQPT when the Rashba SO coupling is changed through a value αn. For a generic non-adiabatic evolution the Loschmidt amplitude G(α) = (cid:104)E+(αin)Ψ+(α)(cid:105) describes how the evolved state deviates from the initial one, and can be employed to detect the occurrence of a DQPT when G goes to zero at the so-called Fisher zeros for a given time instant associated to the spin-orbit coupling amplitude αn. In particular, for the analysis reported in Fig. 2 the critical values {αn}, indicated with red lines, are dis- tributed in a non periodic way. We note that the am- plitude G(α) crosses zero linearly when α is nearby the value αn. nm, an effective electron mass m ∼ 0.07me and a geo- metric factor g (cid:38) 0.1, the adiabatic evolution in a range ∆αL ≈ π can be realized with a linear drive of dura- tion ∆t (cid:29) 10−12s. The typical spin decoherence times in semiconducting nanostructures [50] should then allow to observe such quantum evolution in laboratory. An- other relevant outcome of our study is that symmetric nanoring can realize an innovative quantum platform for achieving and characterizing a DQPT. For the system upon examination, the DQPT can be experimentally de- tected by designing a nanoring tunnel coupled to a quan- tum dot. We consider the case in which the Kramers states E±(cid:105) are selected through an external perturba- tion V . The final state Ψ+(α)(cid:105) is the time evolved configuration of the initial ground-state E+(αin)(cid:105) af- ter changing the Rashba SO coupling from αin to α. 5 Then, at the end of the process we perform a sudden change α → αin, so that the final state remains Ψ+(α)(cid:105). Hence, through the coupling with the quantum dot one can probe only the electron tunnelling in a given energy window around E(αin), e.g. the spectral density function is zero above the energy E(αin). Within the first order in the perturbation V only transitions from Ψ+(α)(cid:105) into +(αin)(cid:105) are allowed if the element matrix an eigenstate E(cid:48) +(αin)Ψ+(α)(cid:105) is nonzero. However, at the critical values α = αn, the Loschmidt amplitude is vanishing, i.e. G = 0, so that the transition to the ground state occurs only through high order pro- cesses, thus leading to a sensible decrease of the tunnel current between the nanoring and the dot. +(αin)(cid:12)(cid:12) V Ψ+(α)(cid:105) = VE(cid:48)(cid:104)E(cid:48) (cid:10)E(cid:48) [1] M. V. Berry, Proc. R. Soc. London, Ser. A 392, 45 (1984). [2] B. Simon, Phys. Rev. Lett. 51, 2167 (1983). [3] F. Wilczek and A. Zee, Phys. Rev. Lett. 52, 2111 (1984) [4] J. Samuel and R. Bhandari, Phys. Rev. Lett. 60, 2339 (1988). [25] Y. A. Bychkov and E. I. Rasbha, Pisma Zh. Eksp. Teor. Fiz. 39, 66 (1984) [JETP Lett. 39, 78 (1984)]. [26] V. N. Golovach, M. Borhani, and D. Loss, Phys. Rev. A 81, 022315 (2010). [27] T. Cadez, J.H. Jefferson, and A. Ramsak, Phys. Rev. [5] N. Mukunda and R. Simon, Ann. Phys. N.Y. 228, 205 Lett. 112, 150402 (2014). (1993). [6] A. Mostafazadeh, J. Phys. A 32, 8157 (1999). [7] D. Kult, J. Aberg, E. Sjoqvist, Phys. Rev. A 74, 022106 (2006). [28] A. Kregar, J.H. Jefferson, and A. Ramsak, Phys. Rev. B 93, 075432 (2016). [29] H. Sanada et al., Nat. Phys. 9, 280 (2013). [30] P. Gentile, M. Cuoco, and C. Ortix, Phys. Rev. Lett. [8] J. Anandan and Y. Aharonov, Phys. Rev. Lett. 65, 1697 115, 256801 (2015). (1990). [9] D. Chruscinski, A. Jamiolkowski, Geometric Phases in Classical and Quantum Mechanics Birkhauser Boston (2004). [10] D. N. Page, Phys. Rev. A 36, 3479 (1987). [11] P. Zanardi and M. Rasetti, Phys. Lett. A 264, 94 (1999). [12] E. Sjoqvist, APS Physics 1, 35 (2008); E. Sjoqvist, Int. J. Quantum Chem. 115, 1311 (2015). [13] R. G. Unanyan, B. W. Shore, and K. Bergmann, Phys. Rev. A 59, 2910 (1999). [31] Z. J. Ying, P. Gentile, C. Ortix, and M. Cuoco, Phys. Rev. B 94, 081406(R) (2016). [32] A.A. Reynoso A. A., J.P. Baltans, H. Saarikoski, J. E. V´azquez-Lozano , J. Nitta, and D. Frustaglia, New J. Phys. 19, 063010 (2017). [33] S. Pandey, N. Scopigno, P. Gentile, M. Cuoco, and C. Ortix, Phys. Rev. B 97, 241103 (R) (2018). [34] H. Saarikoski, J. E. V´azquez-Lozano, J. P. Baltan´as, F. Nagasawa, J. Nitta, and D. Frustaglia, Phys. Rev. B 91, 241406(R) (2015). [14] L.-M. Duan, J. I. Cirac, and P. Zoller, Science 292, 1695 [35] F. Nagasawa, D. Frustaglia, H. Saarikoski, K. Richter, (2001). and J. Nitta, Nat. Commun. 4, 2526 (2013). [15] I. Fuentes-Guridi, J. Pachos, S. Bose, V. Vedral, and S. [36] D. Frustaglia and K. Richter, Phys. Rev. B 69, 235310 Choi, Phys. Rev. A 66, 022102 (2002). (2004). [16] A. Recati, T. Calarco, P. Zanardi, J. I. Cirac, and P. [37] D. Bercioux, D. Frustaglia, and M. Governale, Phys. Rev. Zoller, Phys. Rev. A 66, 032309 (2002). B 72, 113310 (2005). [17] P. Solinas, P. Zanardi, N. Zanghi´, and F. Rossi, Phys. Rev. B 67, 121307(R) (2003). [38] M. Heyl, Rep. Prog. Phys. 81, 054001 (2018). [39] J. M. Hickey, S. Genway, and J. P. Garrahan, Phys. Rev. [18] K. C. Nowack, F. H. L. Koppens, Y. V. Nazarov, and L. B 89, 054301 (2014). M. K. Vandersypen, Science 318, 1430 (2007) [19] S. Nadj-Perge, S. M. Frolov, E. P. A.M. Bakkers, and L. [40] S. Vajna and B. Dora, Phys. Rev. B 91, 155127 (2015). [41] J. C. Budich and M. Heyl, Phys. Rev. B 93, 085416 P. Kouwenhoven, Nature (London) 468, 1084 (2010). (2016). [20] S. M. Frolov, S. R. Plissard, S. Nadj-Perge, L. P. Kouwen- hoven, and E. P. A. M. Bakkers, MRS Bulletin 38 809 (2013). [21] D. D. Awschalom, L. C. Bassett, A. S. Dzurak, E. L. Hu, J. R. Petta, Science 339 (6124), 1174 (2013). [22] A. Manchon, H. C. Koo, J. Nitta, S. M. Frolov, and R. A. Duine, Nat. Mater. 14, 871 (2015). [23] G. Dresselhaus, Phys. Rev. 100, 580 (1955). [24] E. I. Rashba, Sov. Phys. Solid State 2, 1109 (1960). [42] J. Nitta, T. Akazaki, H. Takayanagi, and T. Enoki, Phys. Rev. Lett. 78, 1335 (1997). [43] D. Liang and X. P.A. Gao, Nano Lett. 12, 3263 (2012). [44] A. Lorke, R. J. Luyken, A. O. Govorov, J. P. Kotthaus, J. M. Garcia, and P. M. Petroff, Phys. Rev. Lett. 84, 2223 (2000). [45] U. F. Keyser, C. Fuhner, S. Borck, R. J. Haug, M. Bich- ler, G. Abstreiter, and W. Wegscheider, Phys. Rev. Lett. 90, 196601 (2003). [46] A. Fuhrer, S. Luscher, T. Ihn, T. Heinzel1, K. Ensslin, W. Wegscheider, and M. Bichler, Nature (London) 413, 822 (2001). [47] P. Gentile, M. Cuoco, and C. Ortix, SPIN 03, 1340002 (2013). [48] C. Ortix, Phys. Rev. B 91, 245412 (2015). [49] E. Zhang, S. Zhang, and Q. Wang, Phys. Rev. B 75, 085308 (2007). [50] R. Hanson, L. P. Kouwenhoven, J. R. Petta, S. Tarucha, and L. M. K. Vandersypen, Rev. Mod. Phys. 79, 1217 (2007). 6
1203.2025	1	1203	2012-03-09T09:30:57	Conductance quantization in graphene nanoconstrictions with mesoscopically smooth but atomically stepped boundaries	[ "cond-mat.mes-hall" ]	We present the results of million atom electronic quantum transport calculations for graphene nanoconstrictions with edges that are smooth apart from atomic scale steps. We find conductances quantized in integer multiples of 2e2/h and a plateau at ~0.52e2/h as in recent experiments [Tombros et al., Nature Physics 7, 697 (2011)]. We demonstrate that, surprisingly, conductances quantized in integer multiples of 2e2/h occur even for strongly non-adiabatic electron backscattering at the stepped edges that lowers the conductance by one or more conductance quanta below the adiabatic value. We also show that conductance plateaus near 0.52e2/h can occur as a result of electron backscattering at stepped edges even in the absence of electron-electron interactions.	cond-mat.mes-hall	cond-mat	Conductance quantization in graphene nanoconstrictions with mesoscopically smooth but atomically stepped boundaries Department of Physics, Simon Fraser University, Burnaby, British Columbia, Canada V5A 1S6 S. Ihnatsenka and G. Kirczenow We present the results of million atom electronic quantum transport calculations for graphene nanoconstrictions with edges that are smooth apart from atomic scale steps. We find conductances quantized in integer multiples of 2e2/h and a plateau at ∼ 0.5 × 2e2/h as in recent experiments [Tombros et al., Nature Physics 7, 697 (2011)]. We demonstrate that, surprisingly, conductances quantized in integer multiples of 2e2/h occur even for strongly non-adiabatic electron backscattering at the stepped edges that lowers the conductance by one or more conductance quanta below the adiabatic value. We also show that conductance plateaus near 0.5 × 2e2/h can occur as a result of electron backscattering at stepped edges even in the absence of electron-electron interactions. PACS numbers: 72.10.Fk,73.23.Ad,81.05.Uw Electrical conductances quantized in integer multiples of the fundamental quantum 2e2/h are the hallmark of ballistic quantum transport in nanostructures such as semiconductor quantum point contacts,1,2 gold atomic wires,3 and carbon nanotubes.4 These quantized con- ductances are explained theoretically in terms of the Landauer theory of transport.5 However, in the case of graphene nanostructures, quantum transport calcu- lations have shown the conductance quantization to be easily destroyed by disorder6 -- 14 that is ubiquitous in these systems or by abrupt bends in the quantum wire geometry.15 Accordingly, there have been only a few reports16 -- 18 of conductance quantization being observed experimentally in graphene nanostructures: Lin et al.16 and Lian et al.17 demonstrated conductance quantiza- tion experimentally in graphene nanoribbons. However, the conductance steps that they observed were a few or- ders of magnitude smaller than the ballistic conductance quantum 2e2/h. This phenomenon16,17 has been ex- plained theoretically19,20 as arising from strong electron backscattering at the edges of the electronic subbands of the ribbons due to the presence of random defects. More recently, Tombros et al.18 have reported the experimental observation of conductance quantization in integer multi- ples of 2e2/h, as well as a fractional conductance plateau at ∼ 0.6 × 2e2/h, in a graphene nanoconstriction (GNC) at zero magnetic field. To minimize the effects of disor- der on transport in their device Tombros et al.18 stud- ied a short suspended GNC whose width was similar to its length and estimated to be ∼ 300 nm. Their sam- ple was annealed by Joule heating which resulted in the constriction being formed with curved boundaries that were smooth on the mesoscopic length scale of ∼ 100 nm. The atomic-scale structure of the boundaries was not de- termined experimentally, however, the curvature of the constriction's boundaries implies the presence of large numbers of atomic-scale steps (and possibly also other defects) along the boundaries. In this respect the GNC of Tombros et al.18 differs from the well known semi- conductor quantum point contacts (SQPCs)1,2 where the transverse electron confinement is achieved electrostati- cally and thus the constriction boundaries are effectively smooth on the atomic scale as well as on the much larger (submicrometer) length scale of the overall dimensions of the constriction. In the limit of extremely slow spatial variation of the confining potential, electrons are adia- batically transmitted through the SQPC or adiabatically reflected. As was pointed out by Glazman et al.21 such adiabatic transport results in quantized conductances; each electronic subband that is adiabatically transmitted through the narrowest part of the SQPC at the Fermi en- ergy contributes a quantum 2e2/h to the measured con- ductance. If the confining potential of the SQPC varies smoothly but not adiabatically, conductance quantiza- tion may still occur, each electronic subband transmitted through the narrowest part of the SQPC at the Fermi energy again contributing a quantum 2e2/h to the total conductance.22 However, to date there have been no the- oretical studies of conductance quantization in constric- tions with boundaries exhibiting large-scale smoothness but atomic-scale steps, as in the GNC of Tombros et al.18 For this reason a definitive understanding of the conduc- tance quantization in integer multiples of 2e2/h observed by Tombros et al.18 has been lacking. Furthermore it has also been unclear whether the conductance plateau ob- served by Tombros et al.18 at ∼ 0.6×2e2/h was the result of electron-electron interactions (as is widely believed of the 0.7 × 2e2/h plateau in SQPCs23,24) or whether it can be accounted for instead by strong single-electron scat- tering at steps in the constriction's boundaries. In this paper we report the results of quantum trans- port calculations that address these issues. We con- sider a non-interacting electron tight-binding model of graphene constrictions having similar dimensions to the GNC of Tombros et al.18 and having boundaries that are smooth on the length scale of the constriction but with large numbers of steps on the atomic scale. We show that this model exhibits integer and fractional con- ductance plateaus similar to those that were observed experimentally.18 Our results depend qualitatively on both the width of the constriction and its orientation. For the armchair orientation, the calculated integer quan- d a e l W W W=200 nm W=150 nm (b) armchair leads W=250 nm 3 2 1 0 0 (a) ) h / 2 e 2 ( e c n a t c u d n o C 12 10 8 6 4 2 0.005 0.000 0.005 0.010 0.015 Energy (t) 0 dG/dE (arb. units) 1000 FIG. 1: (color online) (a) Conductance vs. Fermi energy for GNCs with constriction widths W =150, 200, 250 nm. The constriction shape follows the cosine function and is smooth apart from steps on the atomic scale. Host and edge orienta- tion of the semi-infinite leads is armchair. Dotted lines show conductances of uniform armchair ribbons of the same widths as the narrowest parts of the GNC's whose conductances are plotted in the same colors. The black solid line shows the con- ductance of a 300 nm wide ribbon, whose width equals that of the leads W lead. (b) Conductance G vs. its energy derivative dG/dE. Dips in dG/dE indicate conductance plateaus. In the dashed rectangle dG/dE is shown for W = 200 nm only. Temperature T = 0. t = 2.7 eV. The subband spacing is an order of magnitude larger than kBT even at 4.2K as in Ref.18. tized conductances of the constrictions have smaller in- teger values than those of uniform armchair graphene ribbons with the same width as the narrowest part of the constriction. This differs qualitatively from the well known behavior of SQPCs where the adiabatic and non- adiabatic quantized conductance values are equal to those of a uniform quantum wire whose width equals that of the narrowest part of the constriction.21,22 We find plateaus with conductance values ∼ 0.5 × 2e2/h as well as the in- teger plateaus. For the zigzag orientation the calculated integer quantized conductances of the constrictions are either the same as or lower than those of uniform rib- bons of the same width as the constriction. We also find integer and fractional quantized conductances for con- strictions whose narrowest parts are neither zigzag nor armchair. We describe GNCs by the standard tight-binding Hamiltonian on a honeycomb lattice, 2 where ǫi is the on-site energy and tij = t = 2.7 eV is the matrix element between nearest-neighbor atoms. This Hamiltonian is known to describe the π band disper- sion of graphene well at low energies.25 Spin and elec- tron interaction effects are outside of the scope of our study. The nanoconstriction and any random edge dis- order and bulk vacancies that are present are introduced by removing carbon atoms from a uniform ribbon and setting appropriate hopping elements tij to zero. It is assumed that atoms at the edges are always attached to two other carbon atoms and all dangling bonds are passi- vated by a neutral chemical ligand, such as hydrogen, so that the bonding between the carbon atoms at the edge and around vacancies is similar to that in bulk graphene. Random bulk and edge disorder (when present) are char- acterized by the probability of the carbon atoms being removed, pb and pe, respectively. pb is normalized rela- tive to the whole sample, while pe is defined relative to an edge only. The long-range potential due to charged im- purities is approximated by a Gaussian form9,11 of range d: ǫi = Pr0 V0exp(− ri − r02/d2), where both the am- plitude V0 and coordinate r0 are generated randomly. In the linear response regime the conductance of the GNC is given by the Landauer formula5 G = 2e2 h X ji Tji. (2) Tji is the transmission coefficient from subband i in the left lead to the subband j in the right lead, at the Fermi energy. Tji is calculated by the recursive Green's function method, see Ref. 26 for details. The average conductance hGi for samples with random disorder was calculated by averaging over an ensemble of samples with different real- izations of the disorder. For the results presented below, d a e l W W (a) ) h / 2 e 2 ( e c n a t c u d n o C 12 10 8 6 4 2 0 zigzag leads W=250 nm W=200 nm W=150 nm (b) 0 0.005 0.010 0.015 Energy (t)(cid:13) 1000 0 dG/dE (arb. units)(cid:13) 2000 H = X i ǫia† i ai − X hi,ji tij (cid:16)a† i aj + h.c.(cid:17) , (1) FIG. 2: (color online) The same as Fig. 1 but for the zigzag configuration of host and edges in the leads. averaging was carried out over ten realization for each disorder type. similar to those chose geometries To investigate the transport properties of GNC's we studied experimentally.18 The shape of constriction was modeled by a cosine function so that its edges were smooth apart from atomic scale steps. The width of narrowest part of the GNC was varied in the range W = 150...250 nm. The GNC was attached at its two ends to semiin- finite leads represented by ideal nanoribbons of width W lead=300 nm. This guarantees that for any W the leads supply more states for propagation than can pass through the narrowest part of the constriction. The region of the constriction itself in our tight-binding quantum transport calculations included up to ∼ 1.5 million carbon atoms. In our modeling of the effects of random disorder we assumed it to be present only in a finite region of width 300 nm and length L = 300 nm; the semi-infinite leads were free from disorder. The calculated conductances of GNCs with different constriction widths W are shown in Figure 1(a) for the armchair orientation of the graphene host and edges of the ideal leads. Note, however, that the edge orienta- tion along most of the constriction itself is neither arm- chair nor zigzag; see the outset in Figure 1. The con- ductance shows faint quantization steps in integer mul- tiples of 2e2/h, similar to those observed experimentally by Tombros et al.18 For better visualization we plot the energy derivative of the conductance dG/dE in Fig. 1(b). Here a dip in dG/dE indicates a plateau in the conduc- tance. The prominent dips in dG/dE in Fig. 1(b) clus- ter around conductance values that are integer multi- ples of 2e2/h, including both odd and even integer multi- ples. The conductance of the GNC decreases as constric- tion becomes narrower, a feature expected theoretically and observed1,2 in conventional semiconductor quantum point contacts: As the constriction width shrinks the number of propagating states for a given Fermi energy de- creases. Interestingly, although the conductance plateaus occur near integer multiples of 2e2/h, in each case the in- teger has a smaller value than that for the ideal infinite ribbon of uniform width whose width equals the width W of the narrowest part of the constriction, for the same electron Fermi energy, calculated with the same tight binding approach. This can be seen by comparing the conductances of the GNC's in Fig. 1(a) with those of the corresponding uniform ideal ribbons27,28 that are shown as the dotted lines of the same color in Fig. 1(a). We also found no correlation between the calculated GNC conductances and the semiconductor/metallic property of uniform ideal armchair ribbons. These findings show that the conductance quantization that we find for the armchair oriented host and leads is not due to adiabatic transmission of individual eigenmodes of the ideal leads through the constriction but that additional scattering along the constriction edges plays an important role. For W = 200 nm, we find an additional conductance step at ∼ 0.5 × 2e2/h; see Fig. 1(b) and the inset of Fig. 3 ) h / 2 e 2 ( e c n a t c u d n o C 8 7 6 5 4 3 2 1 0 0.000 armchair W=200 nm symmetric shifted W 0.005 0.010 0.015 Energy (t) FIG. 3: (color online) Conductance vs. Fermi energy for GNC of width W=200 nm. The red open squares show the con- ductance for a constriction whose top and bottom parts are shifted by 80 nm relative to each other. Black filled squares show for comparison the conductance for the corresponding symmetric constriction of width W = 200 nm, as in 1(a). 1(a). This agrees with the experimental findings in Ref. 18. Whether or not this feature is present in the results of our quantum transport calculations depends on the width W of the constriction; note that experiments have been reported for only a single sample.18 However, as can be seen in Fig. 1 this fractional plateau coexists with integer conductance plateaus at higher electron Fermi energies for the same GNC and occurs for both electron and hole transport (not shown), as in the experimental data.18 Our results for GNCs with the zigzag orientation29,30 of the graphene host and edges of the leads are shown in Fig. 2. Comparison of Fig. 1 and Fig. 2 reveals signifi- cant differences between quantized conductance plateaus in GNCs with the host and leads in the zigzag and arm- chair orientations: For the zigzag orientation the quan- tized conductance plateaus are more pronounced than for the armchair case. Also for the zigzag case the calcu- lated values of the quantized conductances of the GNCs in many (but not all) cases are close to the values of the quantized conductances of ideal uniform zigzag ribbons having the same width as the narrowest part of the GNC and the same electron Fermi energy. By contrast, as we have already mentioned, all of the integer quantized GNC conductances for the armchair case are smaller than those of the corresponding uniform ideal ribbons by inte- ger multiples of 2e2/h. Thus in many cases non-adiabatic electron backscattering is much weaker for GNCs in the zigzag orientation than for those in the armchair orien- tation. This difference may be attributed to the current densities being much lower near zigzag graphene edges than near armchair edges so that the conductances are less affected by edge imperfections for zigzag ribbons.31 The open red squares in Fig. 3 show the calculated conductance of an asymmetric GNC with the armchair orientation of the host and leads. As shown in the inset, the geometry in this case is similar to the W = 200nm armchair-oriented constriction in Fig. 1 except that the (cid:13) leads no defects edge disorder pe=0.2 bulk disorder pb=10-5 long range potential(cid:13) V<0.2t, =1015 m-2, d=10a (a)(cid:13) (b)(cid:13) 12 10 8 6 4 2 0 12 10 8 6 4 2 0 0.000 0.005 0.010 0.015 Energy (t) ) h / 2 e 2 ( e c n a t c u d n o C ) h / 2 e 2 ( e c n a t c u d n o c d e g a r e v A FIG. 4: (color online) (a) Effect of different disorder types on the conductance of armchair-oriented GNCs. (b) Conduc- tance averaged over 10 realization of disorder. Constriction width W =200 nm. Red line with filled circles is for edge dis- order with pe = 0.2. Blue line with filled squares is for bulk vacancy disorder with pb = 10−5. Green line with rhombuses is for long ranged potentials due to charged impurities with effective parameters V ≤ 0.2t, ρ = 5 × 1015 m−2, d = 10a. The black solid line shows the conductance for an ideal, uni- form ribbon 300 nm wide and is given as a reference. upper and lower regions where the carbon atoms have been removed are now offset from each other laterally by 80 nm. Thus the edges of the narrowest part of the con- striction have neither the armchair nor the zigzag orien- tation. We find that the electron backscattering is some- what stronger (the conductance lower) in this case than for the symmetric W = 200nm armchair-oriented con- striction in Fig. 1; the calculated conductance for the latter is replotted as the solid black squares in Fig. 3 for comparison. However the first few quantized conduc- tance plateaus (as well as the plateau at ∼ 0.5 × 2e2/h) are still clearly visible for the asymmetric GNC. The effects of disorder of different types are shown in Fig. 4. As a test system we chose a GNC of width W = 200 nm having the armchair orientation. The effect of disorder on the conductance of the GNC is similar to that for graphene nanoribbons.19 However, the conduc- tance quantization is strongly degraded for every disorder 4 type including bulk vacancies. This may be attributed to the varying width of the GNC along the transport direc- tion that precludes the existence of well-defined subband edges for the whole structure. We find each type of disor- der to suppress the conductance and to result in universal conductance fluctuations.19,32 In conclusion, we have carried out million-atom elec- tronic quantum transport calculations for graphene nanoconstrictions with boundaries that are smooth ex- cept for steps on the atomic scale and have dimen- sions similar to those of the graphene nanoconstrictions that have been found to exhibit conductances quantized in integer multiples of 2e2/h in recent experiments.18 Our results demonstrate quantized conductances simi- lar to those observed experimentally18 in a tight binding model with non-interacting electrons. We find conduc- tances quantized in integer multiples of 2e2/h to occur in graphene nanoconstrictions even in the presence of strong electron backscattering at the stepped constriction edges that depresses the quantized conductance values by one or more 2e2/h conductance quanta below the quantized conductance values of uniform graphene ribbons with the same width and electron Fermi energy as those of the nar- rowest part of the constriction. This integer conductance quantization in the presence of such strong backscattering has no known analog in either adiabatic or non-adiabatic semiconductor quantum point contacts. It may explain why, based on their transport measurements, Tombros et al.18 estimated the width of their GNC to be smaller (200-275 vs. 300 nm) at zero magnetic field than at higher magnetic fields where electron backscattering at the edges of the constriction is reduced.33 We also find that conductance plateaus at ∼ 0.5 × 2e2/h need not be the result of electron-electron interactions in these sys- tems but can result instead from non-adiabatic backscat- tering of electrons at atomically stepped constriction boundaries. However, the plateau observed experimen- tally at ∼ 0.6 × 2e2/h by Tombros et al.18 resembles the plateau that is seen at ∼ 0.7 × 2e2/h in SQPCs and is attributed to electron-electron interactions23,24 in part because in SQPCs the potentials are smooth and there is no analog of the atomic steps present at the edges of GNCs. Therefore further experimental studies are re- quired to clarify whether electron-electron interactions or boundary scattering are primarily responsible for the fractional plateau observed by Tombros et al.18 in the GNC. Our results (see Fig. 1) suggest that systematic ex- perimental studies of GNCs having differing widths may answer this question. Our quantum transport calcula- tions also show random defects to strongly degrade the conductance quantization in graphene nanoconstrictions. This work was supported by NSERC, CIFAR, Com- pute Canada and WestGrid. 1 B. J. van Wees, H. van Houten, C. W. J. Beenakker, J. G. Williamson, L. P. Kouwenhoven, D. van der Marel, and C. T. Foxon, Phys. Rev. Lett. 60, 848 (1988) 5 2 D. A. Wharam, T. J. Thornton, R. Newbury, M. Pepper, H. Ahmed, J. E. F. Frost, D. G. Hasko, D. C. Peacock, D. A. Ritchie, and G. A. C. Jones, J. Phys. C21, L209 (1988). 3 J. I. Pascual, J. M´endez, J. G´omez-Herrero, A. M. Bar´o, N. Garcia, Uzi Landman, W. D. Luedtke, E. N. Bogachek, H.-P. Cheng, Science 267, 1793 (1995); J. I. Pascual, J. M´endez, J. G´omez-Herrero, A. M. Bar´o, N. Garcia, Vu Thien Binh, Phys. Rev. Lett. 71, 1852(1993) 4 S. Frank, P. Poncharal, Z. L. WangandW. A. deHeer, Sci- ence280, 1744(1998) 5 For a recent review see G. Kirczenow in The Oxford Hand- book of Nanoscience and Technology,Volume I: Basic As- pects edited by A. V. Narlikar and Y. Y. Fu, Oxford Uni- versity Press, U.K. (2010). 6 Y.-W. Son, M. L. Cohen, and S. G. Louie, Phys. Rev. Lett. 97, 216803 (2006). 20 S. Ihnatsenka and G. Kirczenow, Phys. Rev. B 83, 245442 (2011). 21 L. I. Glazman, G. B. Lesovik, D. E. Khmel'nitskii, R. I. Shekhter, Pis'ma v ZhETF, 48, 218 (1988); JETP Lett., 48, 238 (1988). 22 E. Castano and G. Kirczenow, Phys. Rev. B 45, 1514(R) (1992). 23 K. J. Thomas, J. T. Nicholls, M. Y. Simmons, M. Pepper, D. R. Mace, and D. A. Ritchie, Phys. Rev. Lett. 77, 135 (1996). 24 See also the special section on the 0.7 feature and inter- actions in one-dimensional systems, J. Phys.: Condens. Matter 20, 160301-164217 (2008). 25 S. Reich, J. Maultzsch, C. Thomsen, and P. Ordej´on, Phys. Rev. B 66, 035412 (2002). 26 Hengyi Xu, T. Heinzel, M. Evaldsson, and I. V. Zo- 7 D. Areshkin, D. Gunlycke, and C. T. White, Nano Lett. zoulenko, Phys. Rev. B 77, 245401 (2008). 7, 204 (2007). 8 D. Gunlycke, D. A. Areshkin and C. T. White, Appl. Phys. Lett. 90, 142104 (2007) 9 M. Yamamoto, Y. Takane, and K. Wakabayashi, Phys. Rev. B 79, 125421 (2009). 10 M. Evaldsson, I. V. Zozoulenko, Hengyi Xu and T. Heinzel, Phys. Rev. B 78, 161407(R) (2008) 11 E. R. Mucciolo, A. H. Castro Neto, and C. H. Lewenkopf, Phys. Rev. B 79, 075407 (2009). 12 A. L´opez-Bezanilla, F. Triozon and S. Roche, Nano Lett. 9, 2537 (2009). 13 A. La Magna, I. Deretzis, G. Forte and R. Pucci, Phys. Rev. B80, 195413 (2009). 14 B. Biel, F. Triozon, X. Blase and S. Roche, Nano Lett. 9, 2725 (2009). 27 For a recent review of conduction in uniform graphene rib- bons see G. Kirczenow and S. Ihnatsenka, Chapter 13 of Graphene Nanoelectronics: Metrology, Synthesis, Proper- ties and Applications edited by H. Raza (Springer, Berlin Heidelberg New York, 2012), in press. 28 For other models of uniform ribbons see Z. Li, H. Qian, J. Wu, B.-L. Gu, and W. Duan, Phys. Rev. Lett. 100, 206802 (2008); J. Tworzyd lo, I. Snyman, A. R. Akhmerov, and C. W. J. Beenakker, Phys. Rev. B76, 035411 (2007); M. Zarea and N. Sandler, New Journal of Physics 11, 095014 (2009). 29 X. Jia, M. Hofmann, V. Meunier, B. G. Sumpter, J. Campos-Delgado, J. M. Romo-Herrera, H. Son, Y.-P. Hsieh, A. Reina, J. Kong, M. Terrones, M. S. Dresselhaus, Science 323, 1701 (2009). 30 M. Engelund, J. A. Furst, A. P. Jauho and M. Brandbyge, 15 A. Iyengar, T. Luo, H. A. Fertig and L. Brey, Phys. Rev. Phys. Rev. Lett. 104, 036807(2010). B 78, 235411 (2008). 31 L. P. Zarbo and B. K. Nikolic, Europhys. Lett. 80, 470001 16 Yu-Ming Lin, V. Perebeinos, Zhihong Chen, and Ph. (2007). Avouris, Phys. Rev. B 78, 161409(R) (2008). 32 P. A. Lee and A. D. Stone, Phys. Rev. Lett. 55, 1622 17 C. Lian, K. Tahy, T. Fang, G. Li, H. G. Xing, and D. Jena, (1985). Appl. Phys. Lett. 96, 103109 (2010). 18 N. Tombros, A. Veligura, J. Junesch, M. H. D. Guimaraes, I. J. Vera-Marun, H. T. Jonkman and B. J. van Wees, Nature Physics 7, 697 (2011). 19 S. Ihnatsenka and G. Kirczenow, Phys. Rev. B 80, 201407(R) (2009). 33 M. Buttiker, Phys. Rev. B 38, 9375 (1988); R. Ribeiro, J.-M. Poumirol, A. Cresti, W. Escoffier, M. Goiran, J.-M. Broto, S. Roche, and B. Raquet, Phys. Rev. Lett. 107, 086601 (2011).
1703.07509	1	1703	2017-03-22T03:53:28	Emergence of Coherent Nanoscale Structures in Carbon Nantotubes	[ "cond-mat.mes-hall" ]	Recently unusual properties of water in single-walled carbon nanotubes (CNT) with diameters ranging from 1.05 nm to 1.52 nm were observed. It was found that water in the CNT remains in an ice-like phase even when the temperature ranges between 105 - 151 C and 87 - 117 C for CNTs with diameters 1.05 nm and 1.06 nm respectively. Apart from the high freezing points, the solid-liquid phase transition temperature was found to be strongly sensitive to the CNT diameter. In this paper we show that water in such CNT's can admit coherent nanoscale structures provided certain conditions are met. The formation of such coherent structures allows for high values of solid-liquid phase transition temperatures that are in qualitative agreement with the empirical data. The model also predicts that the phase transition temperature scales inversely with the square of the effective radius available for the water flow within the CNT. This is consistent with the observed sensitive dependence of the solid-liquid phase transition temperature on the CNT diameter.	cond-mat.mes-hall	cond-mat	a Emergence of Coherent Nanoscale Structures in Carbon Nantotubes Siddhartha Sen ∗1 and Kumar S. Gupta†2 1Centre for Research in Adaptive Nanostructures and Nanodevices Trinity College Dublin , Ireland 2Saha Institute of Nuclear Physics, Theory Division 1/AF Bidhannagar, Kolkata 700064, India Recently unusual properties of water in single-walled carbon nanotubes (CNT) with diameters ranging from 1.05 nm to 1.52 nm were observed. It was found that water in the CNT remains in an ice-like phase even when the temperature ranges between 105 - 151 C and 87 - 117 C for CNTs with diameters 1.05 nm and 1.06 nm respectively. Apart from the high freezing points, the solid- liquid phase transition temperature was found to be strongly sensitive to the CNT diameter. In this paper we show that water in such CNT's can admit coherent nanoscale structures provided certain conditions are met. The formation of such coherent structures allows for high values of solid-liquid phase transition temperatures that are in qualitative agreement with the empirical data. The model also predicts that the phase transition temperature scales inversely with the square of the effective radius available for the water flow within the CNT. This is consistent with the observed sensitive dependence of the solid-liquid phase transition temperature on the CNT diameter. I. INTRODUCTION Recently unusual properties of water in single-walled carbon nanotubes (CNT) with diameters ranging from 1.05 nm to 1.52 nm were observed [1]. It was found that the water in the CNT remains in an ice-like phase even when the temperature ranges between 105 - 151 C and 87 - 117 C for CNTs with diameters 1.05 nm and 1.06 nm respectively. Apart from the high freezing points, the solid-liquid phase transition temperature was found to be very sensitive to the CNT diameter, with the freezing point decreasing by 20% for increase of diameter of .01 nm. The experimental observations reported are in qualitative agreement with the presence of a single layer of the ice like phase [2, 3], although the phase transition points obtained using such a model are significantly different. Neither the sensitive dependence of the transition temperature on the CNT diameter nor its high value can be understood in terms of models considered, as highlighted in [1]. In view of this alternative ideas need to be considered. In this paper we will explore whether these unusual observed features can be understood in terms of an emergent coherent mesoscopic scale phase of water. The idea of emergent mesoscopic scale coherent structure has been previously used to explain puzzling observed features of mesoscopic magnetism [4] and of water microbubbles [5]. It has been shown that stable coherent mesoscopic structures can form due to the interaction of bound electrons of the mesoscopic system with the ever present zero-point electromagnetic (ZPEM) field, provided a number of system specific constraints are met [6]. Here we suggest that such a picture of mesoscopic coherent surface layer formation [6], which has found applications in mesoscopic magnetism [4] and water nanobubbles [5], might shed light on the current problem. II. MODEL The physical reason for coherence structure formation [6] can be briefly summarized as due to a ZPEM field effecting all bound electrons and lowering their ground state energies in a coherent way [7]. This idea is demonstrated by considering a mesoscopic system of volume V with N bound electrons which have two energy levels and can be described by a Hamiltonian H = N Xi=1 hω0a† 0a0 + ω1a† 1a1 + Ω(cid:16)a† 0a1 + a† 1a0(cid:17)ii . (1) The electron-electron interactions are neglected. For the current problem, the two levels appearing in H are that of electronic states of water molecules which are selected to maximize the possibility of stable mesoscopic coherent ∗ [email protected] † [email protected] 2 structure formation in a CNT. We set the ground state energy to zero. The excitation energy is taken to be E = ω and Gω is the mixing element due to the ZPEM field between the two bound states where G = √8πα r c V ω δV 2V ρ01. (2) Here ρ01 represents the mixing of the two energy levels due to the ZPEM field, V is the volume of the mesoscopic surface layer, δV denotes the fluctuation of the mesoscopic volume due to the interaction with ZPEM and α is the electromagnetic fine structure constant. The fluctuations due to the ZPEM field are such that δV V << 1 (see Appendix A for details). The mixing due to the ZPEM field leads to the lowering of the ground state energy from 0 to −G2ω and an increase of the excited state energy by the same amount [6]. A special property of a coherent water layer is that it expels air molecules dissolved in it. The physical reason for this is that molecules of water in a coherent state have a common frequency which leads to an additional time dependent force that is attractive for molecules in resonance but repulsive for those not in resonance [8]. Thus the air molecules not in resonance with the water molecule oscillations, experience a repulsive force, leading to their expulsion (see Appendix B for details). There are sufficient number of these expelled air molecules to form a ring surrounding the water molecules [9]. In addition, the hydrophobic surface of the CNT repels the water molecules. In our model we expect that the volume available for water flow is reduced by the presence of the expelled air molecules. Taking the diameter of air molecule to be 2.92 A◦ we expect a gap of this size between the water layer and the CNT surface [10, 11]. Thus the effective radius re of the water column for a CNT with diameter 1.05 nm would be 2.33 A◦ and the corresponding re for the 1.06 nm diameter CNT would be 2.38 A◦. It is known that water molecules can form clusters [12 -- 16] and there is both theoretical [2, 14] and experimental [14] evidence for water clusters in CNT's. Furthermore, MD simulations of 6-20 water molecules show that most stable clusters are formed by square and pentagonal water structures [17]. Taking the diameter of a water molecule to be 2.75 A◦, the effective CNT volume available for water flow in our model is compatible with the existence of stable water clusters. The mesoscopic coherent structure of water can only exist and be stable as a quasi 2-dimensional layer [6]. In the present case, there are a small number of water molecules in any given cross-section of the CNT, that form the required mesoscopic layer. It was also necessary that the two electronic bound states of the water molecules used in the model are stable. If these conditions are met, a coherent mesoscopic layer of water molecules can form as interactions of the surface layer molecules with the ZPEM field results in the decrease in energy of the original ground state energy and an increase in energy of an excited state that is in resonance with a ZPEM frequency. Such an effect for molecules in a tunable cavity have been observed [7]. However in [6] it was shown that such tuning is not needed for a properly chosen mesoscopic domain volume V . The conditions that the coherent structure be stable at a temperature T are given by [6] kT < ǫ − G2ω ≡ ǫ kT < G2ω (3) (4) where ǫ is the difference between the ionization energy of a bound electron in water and the excited state energy of the two level system. III. RESULTS We are now ready to state the results of [6] for the system of interest. For the coherent layer water volume V = πr2 e λ, νE , E = ω is the energy difference between the two electronic states of the water and ν = 1.5 is the where λ = 2πc refractive index of water, we get G = 5 × 10−12√ν ρ01√N r2 e . (5) The maximum number of electrons N that can contribute to the coherent structure formation is given by N = 3chr2 , 4νEr3 e λ′. Thus, the w and the coherent domain of water has a volume V = πr2 Each water molecule has a volume 4 e w maximum number of water molecules in the coherent volume V is given by 3 πr3 N = V 4 3 πr3 w , (6) where rw denotes the radius of a water molecule. If the energy difference between the two levels of the system is given by ω = E, then the corresponding wavelength in the coherent water volume is given by 3 λ′ = 2πc ν E Thus the maximum number of coherent electrons in the volume V is given by where ν is the refractive index of water in the CNT. N = 3r2 eλ′ 4r3 w = 3r2 4r3 e ch wνE , (7) (8) √Nq 2α The requirement δV 137 . The constraint δV V << 1 places restrictions on N . For the coherent water volume V , δV π ln 1 α2 , where the Compton wavelength λc = 2.42 × 10−10 cm and α = 1 V << 1 then implies that √N << 1010re. Thus N has to be much less than 5.4 × 104 and 5.6 × 104 for the CNT's with diameters 1.05 nm and 1.06 nm respectively. For a typical value of E = ω = 5 eV, the length of the coherent cylindrical water structure is given by λ = 1.65 × 10−5 cm, For a CNT of diameter 1.06 nm, the corresponding maximum number of coherent electrons is given by N = 2.7 × 103, which falls well within the above limit on N . The parameter G depends on the transition matrix element ρ01. Let us estimate its value. ρ01 depends on the two electronic states of the water molecule. The two states are assumed to be effective Rydberg states and carry angular momenta l and l + 1 respectively, since in this simplified model, the transition matrix between the two states is non-zero only if their angular momenta differ by 1. The geometrical picture uses cylindrical coordinates to locate a water molecule and then describes the effective electronic states in terms of standard spherical polar coordinates. Thus we take V = 2λc re r 0 > = l >= Clrl exp(cid:20)− r0(cid:21) Yl,0, 1 > = l + 1 >= Cl+1 rl+1 exp(cid:20)− which gives Cl =s 2l + 1 r0(cid:21) Yl+1,0, Cl+1 =s 2l + 3 4πΓ(2l + 3)(cid:18) 2 4πΓ(2l + 5)(cid:18) 2 r0(cid:19)2l+3 r0(cid:19)2l+5 r ρ01 ≡ ρ(l, l + 1) = r0 l + 1 2 r 2l + 4 2l + 1 (9) (10) (11) If we denote by ǫ = ǫ − G2ω, where E = ω and ǫ is the difference between the dissociation scale and the excited state energy of the two level system electronic system, then we get r0 = √2meǫ Putting these together we get ρ(l, l + 1) = √2meǫ l + 1 2 r 2l + 4 2l + 1 (12) (13) Note that the applicability of our model for the bound electrons of water molecules requires that both the ground and excited states should be stable. This is not the case for free water molecule [18]. However the presence of the CNT boundary and the presence of the nearly free electrons within the CNT can make the relevant electronic states stable so that coherent structures can form and the results used hold. Using these ingredients, we find that the quantity G has the form G = 9.3 × 10−22 l + 1 2 r 2l + 4 2l + 1 1 re √E √ǫ , (14) where re is measured in cm and the energies E = ω and ǫ are measured in ergs. Recall that the mixing of the two electronic states of water due to the ZPEM field increases the energy of the excited state by G2E where E = ω. In order for the system to be stable under dissociation, the condition G2E < ǫ has to be satisfied. This ensures that the T (1.05) max (K) T (1.06) max (K) T T (1.05) max (1.06) max r(1.05) e (A◦) r(1.06) e (A◦) (cid:16) r (1.06) e r1.05 e 4 (cid:17)2 424 390 1.09 2.33 2.38 1.04 Table I: The first three columns refer to the experimental data [1] for the 1.05 nm and 1.06 nm CNT's. We have used the maximum reported temperatures. The last three column show the result of our analysis. The ratios appearing in the third and sixth column give a direct comparison between the data and our theoretical prediction. inrease in the excited state energy due to the mixing caused by the ZPEM field does not lead to the ionization of the bound electron. This condition leads to the constraint ǫ > ǫmin where ǫmin = 9.3 × 10−22 re l + 1 2 r 2l + 4 2l + 1 . (15) For transition between l = 2 and l = 3 electonic states, and for the lowest effective radius re = 2.33 corresponding to the 1.05 nm CNT, we get the condition that ǫmin = 0.05 eV. We are now ready to calculate the temperature of the coherent water layer. Substituting ρ01 and N in G we get G = 9.3 × 10−22 l + 1 2 r 2l + 4 2l + 1 1 re √E √ǫ (16) Recall that for thermal stability, we need kT < G2ω. Thus the maximum limiting temperature upto which the coherent structure could be thermally stable is given by kTmax = G2E, where E = ω. From these conditions we get Tmax = 62.67 × 10−28 (cid:18) l + 1 2 (cid:19)2 2l + 4 2l + 1 1 r2 e ǫ . (17) An immediate consequence of Eqn. transition temperatures are related by (17) is that for two CNT's of effective radii r[1] e and r[2] e , the corresponding Tmax(2) Tmax(1) = r2 e (1) r2 e (2) (18) Thus the model predicts that the phase transition temperature ratio of water in two different CNT's is completely determined by the ratio of their effective radii, which is available for the water flow. It also predicts that the phase transition temperature decreases as the effective radius of the CNT increases. A comparison of the prediction of this temperature ratio with that obtained in recent experiment [1] is given in Table 1. The data [1] further indicates that the ice like phase persists upto 30 C, which is approximately the room temperature. In our framework this corresponds to a CNT diameter of approximately 11.5 nm. The value of Tmax depends on the angular momenta (l, l + 1) of the electronic states between which the transition takes place, ǫ and the effective radius re of the coherent water structure in the CNT. We choose l = 2 and ǫ = 0.06 eV which satisfies the theoretical constraint that ǫ > 0.05 eV. For the effective radius re = 2.33 A◦, corresponding to the 1.05 nm CNT, the model predicts the maximum phase transition temperature Tmax = 433 K or equivalently 160 C. The maximum temperature signifies the upper limit beyond which the thermal instability sets in. Thus the actual temperature is expected to be somewhat less, which is consistent with the experimental data [1]. Once this choice is made, values of the maximum allowed solid-liquid phase transition temperatures for other effective radii can be obtained from Eqn. (18). Also note that as the CNT radius and equivalently the effective radius re increases, the value of G decreases. In our formalism this indicates that the mixing between the two electronic states of water becomes weaker and the coherent domains begin to disappear as the effective radius increases. The model thus predicts that stable coherent water cannot exist when the phase transition temperature is lower than the room temperature. IV. CONCLUSION The simple model proposed can satisfactorily explain the observed sensitive dependence of the phase-transition temperature on the CNT radius. It predicts that for re > 11.5 nm there is no ice like phase since the corresponding transition temperature is 303 K, approximately the room temperature. The model uses a simple geometric picture for air molecules to explain the effective radius size. Since the expulsion of air molecules from coherent water leading to the formation of a ring structure reduces the available volume for the water molecules. The model also predicts that the coherent phases of water appear as domains of length of 100-300 nm separated by a few molecules of air in a state of dynamic equilibrium. The key result is that the ratio of the transition temperatures. The presence of these air molecules should be amenable to detection by spectroscopic means. 5 [1] Agrawal, K.V., Shimizu, S., Drahushuk, L.W., Kilcoyne, D., & Strano, M.S., Observation of extreme phase transition tem- peratures of water confined inside isolated carbon nanotubes, Nature Nanotechnology (2016) doi:10.1038/nnano.2016.254. [2] Takaiwa, D., Hatano,I., Koga, K. & Tanaka, H. Phase diagram of water in carbon nanotubes, Proc. Natl Acad. Sci. USA 105, 39-43, (2008). [3] Kyakuno, H. et al. Confined water inside single-walled carbon nanotubes : global phase diagram and effect of finite length, J. Chem. Phys. 134, 244501 (2011). [4] Coey,M., Ackland,K., Venkatesan, M., & Sen, S. Collective magnetic response of CeO2 nanoparticles, Nat. Phys. 12, 694-699 (2016). [5] Coey,J.M.D., Mobius,M., Gillen, A.J. & Sen,S. Generation and stability of freestanding aqueous microbubbles, Elec- trochem. Comm. 76, 38-41 (2017). [6] Sen,S., Gupta, K.S., & Coey, J.M.D. Mesoscopic structure formation in condensed matter due to vacuum fluctuations, Phys. Rev. B 92, 155115 (2015). [7] Shalabney,A., George,J., Hutchison,J., Pupillo,G., Genet, C., & Ebbesen,T.W. Coherent coupling of molecular resonators with a microcavity mode, Nature Communications 6, 5981 (2015). DOI: 10.1038/ncomms6981, 2015 [8] Frohlich, H. Theory of the Superconducting State. I. The Ground State at the Absolute Zero of Temperature, Phys. Rev. 79, 845 (1950). [9] See www.engineeringtoolbox.com/oxygen-solubility-water-d8 41.html [10] Reiter, G.F., Kolesnikov, A.I., Paddison, S.J., Platzman,P.M., Moravsky, A. P., Adams, M.A., & Mayers, J. Evidence for an anomalous quantum state of protons in nanoconfined water, Phys. Rev. B 85, 045403 (2012). [11] da Silva, L.B. Structural and dynamical properties of water confined in carbon nanotubes, J. Nanostruct. Chem. 4, 104 (2014). [12] Teeter, M. M. Water structure of a hydrophobic protein at atomic resolution: pentagon rings of water molecules in crystals of crambin Proc. Natl. Acad. Sci. U. S. A. 81, 6014-6018 (1984). [13] Fowler, P.W., Quinn, C.M. & Redmond, D. B. Decorated fullerenes and model structures for water clusters, J. Chem. Phys. 95, 7678 (1991). [14] Kolesnikov, A.I., Zanotti,J.-M., Loong, C.-K., Thiyagarajan,P., Moravsky, A.P., Loutfy,R.O., & Burnham, C. J., Anoma- lously Soft Dynamics of Water in a Nanotube: A Revelation of Nanoscale Confinement, Phys. Rev. Lett. 93, 035503 (2004). [15] Chaplin, M. Do we underestimate the importance of water in cell biology? Nat. Rev. Mol. Cell Biol. 7, 861-6 (2006). [16] Johnson, K., Terahertz vibrational properties of water nanoclusters relevant to biology, J. Bio. Phys. 38, 85-95 (2012). [17] Ignatov, I., Mosin, O. Mathematical Models Describing Water Clusters as Interaction Among Water Molecules. Distribu- tions of Energies of Hydrogen Bonds, Journal of Medicine, Physiology and Biophysics, 3, 48 (2014). [18] Shandilya,B.K., Sen,S., Sahoo,T., Talukder,S., Chaudhury,P., & Adhikari, S. J. Chem. Phys. 139, 034310 (2010) and private communication, S.Adhikari. [19] Welton,T.A. Some Observable Effects of the Quantum-Mechanical Fluctuations of the Electromagnetic Field, Phys. Rev. 74, 1157 (1948). Appendix A: General expression for G The transition matrix element between the electronic bound states of water mixed by the zero point electrmagnetic (ZPEM) field is used to define the parameter G as [6] Using E2 8π = ω V and δE = −E δV 2V , gives Gω =< 0ex.δE1 > . G = √8πα r c V ω δV 2V ρ01, where ρ01 ≡< 0x.u1 > [6]. wavelength λ such that ωλ = 2πc, where c is the velocity of light in vacuum. associated with a wavelength λ′ such that λ = νλ′ where ν is the refractive index of water. In vacuum, the fluctutaion of the E-M field with a frequency ω corresponds to a In water, the same frequency is The ZPEM field acts on all bound electrons in the mesoscopic volume V to form a coherent domain of water. The mesoscopic volume is given by V = πr2 e λ′. This gives δV V = 2 δre re , where and λc denotes the Compton wavelength [19] . From these we get (δre)2 = 2N α π λ2 c ln 1 α2 6 G = √8πα r c V ω ( λc r ) r 2α π √ν λc α rln 1 α2 ρ01 √N 1 ln α2 ρ01 √N r2 2√2 π 137 and λc = 2.42 × 10−10, we get = Usung α = 1 G = 5 × 10−12√ν ρ01√N r2 e The important feature of this formula is the dependence on the total number of coherent electrons N . We use simple geometric arguments to estimate the maximum allowed value of N . Appendix B: Expulsion of air molecules from water Dissolved air molecules in water are expelled when water is in a coherent phase. This happens because air molecules not vibrating in resonance with water molecules experience repulsive force as shown in [8]. The expelled air molecules form a ring around the water molecules preventing them in coming into direct contact with the carbon surface. Let us briefly give Frohlich's argument. Consider the following simple model in which a collection of free dipoles interact with an oscillating electromagnetic field. The interaction Hamiltonian H is H = gZ ψ∗(x, t)ψ(x, t)A(x, t) where ψ∗(x, t) represents the dipole creation operator and ψ(x, t) the dipole destruction operator and A(x, t) the oscillating electromagnetic field which we have approximated all fields by scalar fields. The oscillatory character of A(x, t) means we can write it as This gives rise to a positive interaction of the form A(x, t) = 1 √V X[aqe−1ωqt+q.x + hc] Vq ≈ ωq (ǫp − ǫp+q)2 − (ωq)2 where ǫ = ǫp − ǫp+q is energy difference between the two dipoles and ω is the photon energy. Integrating ω over the allowed momenta values of the photon with an appropriate physically determined cut off, Omega, gives the potential energy as a function of the difference in energies of the two dipoles. In the simple model we take the dipoles to be free air molecules in the self generated gas cavity that surrounds a hydrophobic site. The quantum fluctuating momentum scale set by the quantum uncertainty principle is for a cavity size of ∆x ≈ 1 given by ∆p ≈ h ∆x . The corresponding energy is ≈ 10−4 eV. Carrying out the momentum integration gives the dipole- photon potential Vd,γ(ǫ) to be Vd,γ(ǫ) ≈ ( g2 Ω )( ǫ 2Ω ln 1 + ǫ Ω 1 − ǫ Ω − 1) where ǫ = ǫ1− ǫ2 the difference of energies of the two dipoles and Ω is the photon energy corresponding to the thermal momentum cutoff momentum. We see this potential is attractive for ǫ ≤ Ω. We note that this attractive potential, and its associated attractive force, increases as ǫ ⇒ 0 which corresponds to the oscillators being in resonance. This new force, of quantum origin, only appears when the dipoles are oscillating and the electromagnetic field is time dependent. The oscillation of the dipoles in the model come from the fact that they are confined in a bubble cavity of small length and are represented by oscillating waves. This new quantum force brings resonating dipoles close together and repels ones that are not in resonance as stated.
1109.6148	4	1109	2012-07-24T09:35:57	Nonlinear adiabatic response of interacting quantum dots	[ "cond-mat.mes-hall", "cond-mat.str-el" ]	We develop a generic method in Liouville space to describe the dissipative dynamics of coherent interacting quantum dots with adiabatic time dependence beyond linear response. We show how the adiabatic response can be related to effective quantities known from real-time renormalization group methods for stationary quantities. We propose the study of a generalized $RC$-time as a characteristic time scale. We apply the method to the interacting resonant level model and calculate the nonlinear adiabatic charge response to time-dependent gate voltages, tunneling barriers and Coulomb interaction. The generalized $RC$-time is found to acquire a unique form in all cases, in contrast to the capacitance and the charge relaxation resistance. We discuss the observability of the effects in molecular systems and cold atom setups.	cond-mat.mes-hall	cond-mat	epl draft Nonlinear adiabatic response of interacting quantum dots Oleksiy Kashuba(a), Herbert Schoeller and Janine Splettstoesser Institut fur Theorie der Statistischen Physik, RWTH Aachen, 52056 Aachen, Germany, EU and JARA - Fundamentals of Future Information Technology PACS 71.10.-w -- Theories and models of many-electron systems PACS 73.23.-b -- Electronic transport in mesoscopic systems PACS 73.63.Kv -- Quantum dots Abstract -- We develop a generic method in Liouville space to describe the dissipative dynamics of coherent interacting quantum dots with adiabatic time dependence beyond linear response. We show how the adiabatic response can be related to effective quantities known from real-time renormalization group methods for stationary quantities. We propose the study of a delay time as a characteristic time scale. We apply the method to the interacting resonant level model and calculate the nonlinear adiabatic charge response to time-dependent gate voltages, tunneling barriers and Coulomb interaction. The dot charge delay time is found to be given by a unique expression in all cases, in contrast to the capacitance and the charge relaxation resistance. We discuss the observability of the effects in molecular systems and cold-atom setups. Introduction. -- Adiabatic transport through quan- tum dots associated with a slow cyclic time dependence of the system parameters has generated a lot of inter- est in recent years, particularly in connection with quan- tum pumps [1 -- 4] and mesoscopic capacitors [5, 6], see also Ref. [7]. For noninteracting systems the scattering formalism is a powerful tool to describe the adiabatic response [1, 2]. A central challenge in this field is the understanding of the influence of strong interactions as they typically occur in small quantum dots. Although general current formulas have been derived in terms of Green's functions [8 -- 10], their evaluation is quite diffi- cult in the coherent regime at low temperature. Progress has been achieved in the perturbative regime of high tem- perature [11], where it was shown that pure interaction effects can be revealed by the adiabatic response which would be covered by more dominant effects in the steady state [11, 12]. These studies also included the properties of the RC-time in linear response [13]. In contrast, for interacting quantum dots at low temperature, a generic formalism providing the adiabatic time evolution in re- sponse to any parameter in linear or nonlinear response is not yet available. So far, quantum pumping has been studied for special models, like e.g. the 2-channel Kondo model in the strong coupling regime [9], the Kondo model at the exactly solvable Toulouse point [14], and the single- (a)E-mail: [email protected] impurity Anderson model within slave-boson mean-field approximation [15]. In addition, the research on interac- tion effects in mesoscopic capacitors has concentrated on the special case of linear charge response to an external AC gate voltage by using the standard relation to the equilib- rium density-density correlation function. Here, the main object of interest was the charge relaxation resistance R defined by expanding the charge response δQ in the exter- nal frequency ω via δQ = (C + iωRC2)δV , (1) where Q(t) = δQe−iωt denotes the charge, C is the static quantum capacitance and V(t) = δVe−iωt defines the ex- ternal AC gate voltage. For a single transport channel and provided that the Coulomb interaction is weak, a universal relaxation resistance R = h/2e2 was found in the coherent regime at zero temperature [5, 16]. For interacting metal- lic dots and the single-impurity Anderson model the Shiba relation was shown to be a powerful tool to analyze R and its universality [17 -- 19]. Using bosonization, the influence of Luttinger-liquid leads on R has also been studied [20] and a numerical approach has been used away from the Fermi-liquid regime [21]. In this Letter we develop a general approach to deal with the adiabatic dissipative response, where the time scale of the external modulation ∝ ω−1 is much larger than the in- verse of typical relaxation rates Γc, of a coherent quantum dot at low temperature including Coulomb interactions. p-1 O. Kashuba, H. Schoeller and J. Splettstoesser We show that the adiabatic response can be calculated very efficiently by using quantum field theoretical methods in Liouville space developed in Refs. [22, 23] and general- ized here for the case of time-dependent Hamiltonians (for approaches within Keldysh formalism see the review arti- cles [7, 22] and the recent development [24]). We provide a general relationship of the adiabatic response to effec- tive Liouvillians and vertices known from real-time renor- malization group (RTRG) in the stationary limit with in- stantaneous time parametrization, based on powerful tech- niques for the calculation of Laplace-variable derivatives, recently used within the E-flow scheme of RTRG [25]. As a consequence, our formalism is suitable to any model which can be treated by RTRG, which is applicable for many generic models with charge and spin fluctuations [22]. Re- cent applications of RTRG cover the Kondo model for both weak [26 -- 30] and strong [25] coupling, and the in- teracting resonant level model (IRLM) [31]. Most impor- tantly, in contrast to previous research, our formalism al- lows for the adiabatic variation of any parameter in non- linear response, where no relation to equilibrium density- density correlation functions is possible and where certain identities like e.g. the Shiba relation are no longer appli- cable. Therefore, going beyond previous studies of the linear response to an external gate voltage, we also study the response to other parameters, like the tunneling cou- pling or the Coulomb interaction, which experimentally can be either realized intentionally, or indirectly induced by the gate voltage. We even cover the regime of non- linear response, motivated by recent works on mesoscopic capacitors in the nonlinear driving regime [32]. Instead of the linear response formula (1) for the charge variation by an external gate voltage, we decompose the dynamics of any observable A, with a nonvanishing instantaneous contribution and its adiabatic correction in response to any parameter, as A(t) = A(i)(t) + A(a)(t). The central quantity of our interest is the delay time scale τA for the expectation value A, defined by τA = A(a)/ A(i) , (2) which describes the delay of the full solution comparing to the instantaneous one. For A ≡ Q in linear response to a time-dependent gate voltage, it is equivalent to the RC- time. In general τA can be quite different from typical relaxation times, depending on the observable, the type of excitation and its amplitude, and it is of fundamental interest to understand its dependence on interactions. We use our method to consider the IRLM with a single lead, which constitutes a minimal model for the mesoscopic capacitor with one single-particle level, where strong correlations are induced by a local Coulomb inter- action between the dot and the lead. Recently, the IRLM has been extensively used to study nonequilibrium trans- port through interacting quantum dots [31,33 -- 36], includ- ing the dynamics of the time evolution into the stationary state [31]. We calculate the nonlinear adiabatic charge re- sponse and the delay time τQ, including renormalization effects of the tunneling enhanced by correlations. Impor- tantly, we find that the functional form of the charge delay time τQ is robust against the choice of the time-dependent parameter even in nonlinear response, whereas the capac- itance C or the relaxation resistance R get a more com- plex form when the tunneling or the Coulomb interaction are varied. Finally, we analyze further possible experi- mental implementations of the predicted results for the IRLM with time-dependent parameters, namely via the Anderson-Holstein model in molecular electronics or via the spin-boson model in cold-atom setups. Method. -- We start from a general Hamiltonian H(t) = Hres + Pα µα(t) Nα + Hdot(t) + V (t) of an inter- acting quantum dot coupled to noninteracting fermionic reservoirs with time-dependent chemical potentials µα(t) and a flat d.o.s. of width 2D via the coupling V (t). Gen- eralizing the Liouvillian approach of Ref. [22] to the case of time-dependent Hamiltonians, one finds that the dissi- pative dynamics of the reduced density matrix ρ(t) of the dot can be described by the effective Liouvillian equation i∂tρ(t) = Z t t0 L(t, t′)ρ(t′)dt′, (3) where L(t, t′) is the effective dot Liouvillian obtained by integrating out the reservoirs ( = 1). At the initial time t0 the total density matrix factorizes into an arbitrary dot and an equilibrium reservoir part. Since we are only in- terested in the asymptotic dynamics we set t0 = −∞ be- low. Following Ref. [22], the effective Liouvillian can be calculated diagrammatically, where each diagram of order O(V n) consists of a product of vertices G(ti), t1 > ··· > tn, with effective dot propagators Π(ti, ti−1) in between. In addition, the information of the Fermi distribution and the d.o.s. of the reservoirs is contained in time-independent reservoir contractions between the vertices. Using the for- mal definition G(t, t′) = G(t)δ(t − t′ − 0+), we find that each term can be written in terms of a generalized convolu- tion in time space as (G◦ Π◦ ···◦ Π◦ G)(t, t′), where (A ◦ B)(t, t′) ≡ R t t′ dτ A(t, τ )B(τ, t′). Introducing the partial Laplace transform A(t; E) = R t −∞ dt′ei(E+i0)(t−t′)A(t, t′), (A1 ◦ A2 ◦ ··· ◦ An)(t; E) = eDA1(t; E) . . . An(t; E) = ei Pj>k ∂Ej ∂tk A1(t1, E1) . . . An(tn, En)(cid:12)(cid:12)(cid:12)Ej =E,tk=t The special differential operator D = i∂left prescribes the energy derivative to act left to the time derivative. This rule is a natural generalization to Laplace space of analog identities in Fourier space used for gradient expan- sions in the Keldysh formalism [37]. Formally, it allows for the straightforward application of the Liouvillian approach to time-dependent Hamiltonians, with the difference that the exponential differential operator has to be taken be- forehand. In the adiabatic case, the exponential can be E ∂right we get . (4) t p-2 expanded in ∂E∂t ∼ ω the effective Liouvillian, Γc ≪ 1, leading to an expansion of L(t; E) = L(i)(t; E) + L(a)(t; E) + . . . . (5) Here, L(i)(t; E) denotes the instantaneous part, where the time t enters only parametrically via the external parame- ters, and L(a)(t; E) is the first adiabatic correction, which is linear in the time derivatives of the external parameters. Once the effective Liouvillian L(t; E) is known up to the adiabatic correction, one can use it in eq. (3) which reads i∂tρ(t) = (L ◦ ρ)(t; 0) = ei∂L E ∂ρ t L(t; E)ρ(t)(cid:12)(cid:12)(cid:12)E=0 (6) in the mixed (t; E)-representation. Expanding ρ(t) = ρ(i)(t) + ρ(a)(t) + . . . analogously to eq. (5), we find by comparing equal powers in the external frequency , Tr ρ(i) = 1 L(i)ρ(i) = 0 , Tr ρ(a) = 0 , L(i)ρ(a) + L(a)ρ(i) − i(1 − ∂EL(i))∂tρ(i) = 0 . (7) (8) In all arguments of L(i/a) and ∂EL(i), E = 0 has to be taken. From these equations the instantaneous density matrix ρ(i)(t) and the first adiabatic correction ρ(a)(t) can be determined. We emphasize that this approach is even applicable in nonlinear response in the amplitude of the external perturbations, i.e. only the time scale of the ex- ternal modulation needs to be large enough. Furthermore, it allows for an adiabatic modulation of any parameter of the Hamiltonian and is not restricted to a time-dependent gate voltage. The algebra of (7) and (8) can be easily evaluated for quantum dots with two accessible states. If additional conservation laws are present (as, e.g., charge conservation in the IRLM or spin-Sz conservation in the Kondo model), the nonvanishing matrix elements of the Liouvillian can be written as L¯s¯s,ss = −Lss,ss = iΓs = iΓ/2 + isΓ′ , Ls¯s,s¯s = ǫs , where s ≡ ± denotes the two states and ¯s = −s. At E = 0 we get from (7) and (8) that the instantaneous density matrix is diagonal ρ(i) s /Γ(i) = 1/2 + sΓ′(i)/Γ(i) and the adiabatic correction fulfills ρ(a) s = Γ(i) (9) − with + = −ρ(a) ρ(a) + = 1 Γ(i) (cid:26)Γ′(a) − Γ′(i) Γ(i) Γ(a) − (1 + i∂EΓ(i))∂t(cid:18) Γ′(i) Γ(i) (cid:19)(cid:27) (10) Below we use this result to evaluate the adiabatic response for the IRLM. Calculation of L(a). -- We now turn to the central issue of how to relate the adiabatic correction L(a)(t; E) to the instantaneous quantities known from RTRG. Us- ing (4), we can formally write L(a)(t; E) = DL(i)(t; E) = E ∂right i∂left L(i)(t; E). The representation of L(i)(t; E) by its diagrammatic expansion specifies what "left" and t Nonlinear adiabatic response of interacting quantum dots "right" means for the derivatives with respect to E and t. As a first step, we represent the derivative ∂EL(i) by effec- tive vertices and propagators, using a method developed in Ref. [25]. For generic models with two types of vertices, namely single (e.g. tunneling) and double (e.g. Coulomb interaction, exchange, etc.), we decompose it into two con- tributions in leading order and find ∂EL(i)(t; E) = ∂EL(i) Γ (t; E) + ∂EL(i) U (t; E) = ❡ ❡ + 1 2 ❡❡ ❡❡ . (11) 1 The diagrammatic rules are explained in detail in Refs. [22, 25]. The single (double) circles represent full effec- tive single (double) vertices with effective propagators Π(i)(t; E) = E−L(i)(t;E) in between (the Laplace variable is shifted by the frequencies and chemical potentials of all reservoir contractions crossing over the propagator). The left slash indicates ∂E and the grey (green, color online) line represents the reservoir contraction given by the anti- symmetric part f (ω)− 1 2 of the Fermi distribution function. All possible diagrams for ∂EL(i) can be classified by the number of lines over the propagator containing a deriva- tive. In the next step we perform the time derivative i∂t right to the energy derivative. The energy derivative is then shifted by partial integration to the reservoir con- traction (indicated by a (blue) cross) [25]. This yields L(a) Γ (t; E) = ❡ ❡ + ❡ ❡ L(a) U (t; E) = = − ❡ 1 2 ❡❡ ❡ + ❡ 1 2 ❡❡ ❡❡ + ❡ , (12) ❡❡ , (13) t E ∂right where the right slash represents i∂t and the hat indicates the differential operator D = i∂left . The frequency integral in both diagrams of (12) is well-defined in the wide-band limit, so (12) provides an explicit expression for the adiabatic correction containing the tunneling ver- tices in terms of renormalized vertices and propagators. In contrast, the frequency integrals in (13) are logarithmi- cally divergent. We therefore take a second derivative with respect to E, yielding an RG equation for the adiabatic part, L(a) U (t; E), after partial integration. This contains the double vertices ∂E L(a) U (t; E) = 1 2 ❡❡ ❡❡ − 1 2 ❡❡ ❡❡ . (14) Eqs. (12) and (14) are the final results for the evalua- tion of adiabatic corrections of the Liouvillian in leading order, based on the instantaneous values of the renor- malized vertices and Liouvillian, which are obtained from RTRG1. For the adiabatic part of the propagator, appear- ing in the second diagram of (12) and (14) each, we insert 1Provided that the frequency integrals converge, we note that our results can even be applied to a frequency-dependent d.o.s. in the leads. Otherwise, another derivative with respect to E may be required. p-3 O. Kashuba, H. Schoeller and J. Splettstoesser Π(a) = Π(i)L(a)Π(i) +(∂EΠ(i))(i∂tL(i))Π(i). The first term does however not contribute to the adiabatic propagator in leading order. An interesting question is whether derivatives with re- spect to the Laplace and time variable commute in leading order, i.e. whether the adiabatic correction to the effective Liouvillian, eqs. (12) and (14), can be written as Γ/U (t; E) ?= L(a) 1 2 i∂E∂tL(i) Γ/U (t; E) . (15) A similar relation was investigated so far only for noninter- acting systems [38]. To analyze its validity we introduce the complementing differential operator D′ = i∂right ∂left , t where the energy derivative is taken right to the time derivative. Analogously to (12) and (14) one finds in lead- ing order E , ❡ (16) ❡❡ − ❡ + ❡ 1 2 ❡❡ 1 2 ❡❡ Γ (t; E) = − ❡ U (t; E) = D′L(i) ∂E D′L(i) ❡❡ . (17) The inverted hat represents the differential operator D′. Using i∂E∂t = D + D′, we can write D = 1 2 i∂E∂t + 2 (D − D′) and, thus, the correction to eq. (15) for L(a) 1 Γ (∂EL(a) U ) is given by half the difference of (12) and (16) ((14) and (17)). We first address the second diagrams on the r.h.s. of these equations: their differences involve the expression 1 2 (D − D′)Π(i) = Π(i)(cid:18) 1 2 (D − D′)L(i)(cid:19) Π(i) (18) + 1 2 n(∂EΠ(i))(i∂tL(i))Π(i) − Π(i)(i∂tL(i))(∂EΠ(i))o s¯s,s′ ¯s′ = δss′ ǫ(i) for the propagator. Here, the first term on the r.h.s. can be neglected in leading order, whereas the second one is only zero if the Liouvillian and its time and energy deriva- tive commute. For special cases this is indeed possible: it follows trivially for blocks where the Liouvillian is diago- nal, as e.g. the 2 × 2-block L(i) s of eq. (9). For 2-level systems with conservation laws, see eq. (9), it holds also for the block L(i) ss,s′s′ since the zero eigenvalue of the Liouvillian can be omitted in a propagator stand- ing left to a vertex averaged over the Keldysh indices [22]. Therefore, for this block one can replace the Liouvillian by its nonzero eigenvalue −iΓ(i)(t; E) and the second term on the r.h.s. of (18) is again zero. If this is given (or if the term can be neglected in leading order for certain models), we can write the correction to eq. (15) generically as i∂E∂tL(i) Γ (t; E) L(a) Γ (t; E) = 1 2 1 2 ❡ U (t; E) = ∂E (cid:26) 1 + 2 ∂EL(a) 1 ❡ − 2 ❡ U (t; E)(cid:27) i∂E∂tL(i) ❡ , (19) + 1 4 ❡❡ ❡❡ − 1 4 ❡❡ ❡❡ . (20) From this result we observe another condition for the va- lidity of (15), namely that it should not matter whether the right or the left vertex is differentiated with respect to time, i.e. the two vertices should be equivalent. Whether this is the case, depends on the algebra of the model under consideration. For noninteracting systems one can take bare vertices and the reservoir indices of the two vertices have to be the same due to the reservoir contractions con- necting them. In this case the condition is fulfilled if the vertices do not depend on the level index of the dot states, e.g. through differently time-dependent coupling to differ- ent leads [39]. For interacting systems, the validity of (15) is more restrictive. The renormalized vertices can be quite different from the bare ones and the vertices get an ad- ditional dependence on the Laplace variable E which is shifted by the chemical potentials of the reservoir lines crossing over the propagator standing left to that vertex. As a consequence, the two vertices are never equivalent in the presence of a bias voltage and correction terms def- initely occur for time-dependent voltages. As discussed below, for the particular case of the IRLM with one single reservoir, correction terms to eq. (15) do not appear in leading order. Results. -- We use the above developed method to analyze the response of a mesoscopic capacitor at zero temperature, described by the IRLM, where Hres = kak describes a noninteracting reservoir with flat d.o.s. ν of band width 2D, Hdot(t) = ǫ(t)c†c denotes a spinless single-level quantum dot with time-dependent level position ǫ(t), and Pk ǫka† V (t) = r Γ0(t) 2πν Xk ν Xkk′ U (t) (c†ak + h.c.) (c†c − 1/2) a† kak′ (21) + is the dot-reservoir coupling with the bare time-dependent tunneling rate Γ0(t) and the time-dependent dimensionless Coulomb interaction U (t). As shown above, we can evaluate the adiabatic response from eq. (10), where Γ(a) and Γ′(a) can be extracted from eqs. (19) and (20) together with the RTRG results for the instantaneous vertices and the Liouvillian derived in Ref. [31]. For E = 0 and leading order in U , the results of Ref. [31] read Γ = Γ0(cid:18) Γ′ = − Γ π D ǫ − iΓ/2(cid:19)2U arctan , ǫ Γ/2 , ∂EΓ = i ∂EΓ′ = − , U Γ2 ǫ2 + ( Γ i π 2 )2 Γǫ ǫ2 + ( Γ 2 )2 (22) , (23) where we have omitted the index (i) for the instantaneous quantities. Furthermore, the analysis in Ref. [31] shows that the Coulomb vertex is zero in leading order for the Li- ouvillian elements containing Γ and Γ′. Therefore eq. (19) p-4 is sufficient to evaluate Γ(a) and Γ′(a). Inserting the alge- bra for the instantaneous tunneling vertices into eq. (19), one finds that eq. (15) is valid for the calculation of Γ′(a), whereas for Γ(a) a correction term occurs proportional to ∂tU . This yields the total result Γ(a) = − U 2 ∂t Γ2 ǫ2 + ( Γ 2 )2 , Γ′(a) = 1 2π ∂t Γǫ ǫ2 + ( Γ 2 )2 . (24) Since ∂EΓ, Γ(a) ∼ O(U ) we can neglect them in leading order in eq. (10) and, by inserting (22) to (24) into (10), we find after a straightforward analysis for the charge re- sponse given by Q = eρ+, Q(i) = C0Γ∂t ǫ eΓ , Q(a) = −R0C2 0 Γ∂t ǫ eΓ , (25) Γ 2π ǫ2+(Γ/2)2 . 2e2 and C0 = e2 where R0 = h In the special case of linear response and when only ǫ is varied with time, C = C0 is the static capacitance and R = R0 the universal relaxation resistance, in agreement with (1). In contrast, when Γ is varied with intent or via an accidental (but experimentally unavoidable) gate voltage dependence of Γ0 or U , we obtain in linear response eq. (1) with ǫ Γ ∂Γ ∂ǫ(cid:19) , R = R0C0 C , (26) C = C0(cid:18)1 − ∂ǫ ≈ Γ Γ0 ∂Γ0 ∂ǫ + 2Γ ∂U where ∂Γ ǫ−iΓ/2 . As a consequence, C and R are very sensitive to the variation of other param- eters, and logarithmic terms due to renormalization effects occur, if the Coulomb interaction U varies with time. ∂ǫ log D In this general case, where also the renormalized Γ varies with time, we propose to analyze the time scale τQ. From (2) and (25) we get τQ = (cid:12)(cid:12)(cid:12)(cid:12) Q(a) Q(i) (cid:12)(cid:12)(cid:12)(cid:12) = Γ/2 ǫ2 + (Γ/2)2 = R0C0 , (27) which is of O(Γ−1) close to resonance ǫ ∼ Γ and of O(Γ/ǫ2) away from resonance. This result holds for any variation of ǫ, Γ0 and U and is also valid in nonlinear response. In- teraction effects enter only weakly via the renormalized Γ given by (22). Importantly, τQ reveals the static capac- itance C0 for a pure change of the gate voltage in linear response, with the advantage that τQ can be determined in the presence of the variation of any parameter. The experimentally accessible time scale τQ is thus an interesting quantity, which, for the case of the IRLM, is stable for the variation of any parameter in linear or non- linear response. We note that this time scale can vary quite drastically if other observables or other models are studied. E.g., the time scales τQ and τI , when Q is re- placed by the current I = Q, are in general the same only in linear response. For the IRLM, the time scale τI shows similar logarithmic renormalizations in nonlinear response as they occur in C and R for time varying U . Nonlinear adiabatic response of interacting quantum dots Realizations. -- Several experimental realization of the IRLM exist, where the different parameters can be modulated in a controlled way. As we outline here, the applicability of the IRLM is not limited to the description of an interacting quantum dot, but allows the observation of the predicted effects for various physical systems. First, we show that the low-energy physics of the IRLM is equiv- alent to the one of the Anderson-Holstein model, as first predicted in Ref. [40]. This model is widely used in molec- ular electronics [41] and describes a single-level molecular quantum dot, having a vibrational degree of freedom with frequency Ω coupled linearly to the charge of the dot Hdot = ǫM c†c + Ωb†b − λΩ(b + b†)c†c , V = r ΓM 2πν Xk (c†ak + h.c.) . (28) (29) [42, 43]. The parameters ǫM , ΓM , λ and Ω can be related to the effective parameters ǫ, Γ0 and U of the IRLM. Applying a Lang-Firsov transformation [42], the coupling to the vi- brational mode can be incorporated into the tunneling, leading to an effective level position, ǫ = ǫM − λ2Ω, and tunneling rate, Γ0 = ΓM e−λ2 If the vibration frequency Ω is large compared to the other energy scales, the virtual intermediate states between the tunneling se- quences with one or more bosons can be integrated out. This produces terms with n > 2 lead operators in the ef- fective Hamiltonian. At large λ, all cotunneling processes with n > 2 are exponentially suppressed, while the two- particle processes enter as an effective interaction. Hence, by integrating out all vibrational modes the Anderson- Holstein model can be mapped onto the IRLM with ef- fective interaction, U = ΓM 2πλ2Ω , with Ω ∼ D. This is in agreement with Ref. [40], where it was shown numerically that this formula has even a broader range of applicability. A modulation of the tunneling barriers is always accompa- nied by a modulation of the effective interaction U , since it is proportional to the tunneling rate ΓM . In this case, our results predict that logarithmic corrections appear for C and R, whereas the time scale τQ only depends on ǫ and Γ via eqs. (27) and (22). The Holstein coupling in the Hamiltonian allows for the observation of the dot charge via the displacement of the dot ∼ hb + b†i. Finally, our results can be used to extract information on the relaxation behavior of systems described by the spin-boson model, namely, two-level dissipative systems connected to a large ensemble of oscillators H = ǫ 2 σz − ∆ 2 σx +Xq ωqb† qbq + σz 2 Xq gq(bq + b† q). (30) The spin-boson model can be implemented by a Bose con- densate of atoms trapped by a focused laser beam [44]. Such ultracold gases in optical lattices provide experimen- tal realizations for theoretical models with remarkably in- dependent tunability of parameters including the interac- tion strength, in contrast to usual semiconductor quan- tum dot setups. The system's behavior depends crucially p-5 O. Kashuba, H. Schoeller and J. Splettstoesser on the spectral coupling function. For the ohmic case, i.e. when the coupling constant obeys Pq g2 q δ(ω − ωq) = 2αωe−ω/D, the spin-boson model can be mapped onto the IRLM if α ≈ 1/2 (close to the Toulouse limit) [45, 46]. The effective IRLM parameters are U = 1 − √2α and Γ0 = ∆2/D. Changing the coupling of the Bose con- densate to the spin via α one generates a time-dependent effective interaction U . The resulting response hSzi of the spin, identified with (ρ+ − ρ−)/2 in the effective IRLM, allows for the determination of the interesting time scale τSz , given by (27). Especially in the biased case, where τSz ∼ Γ/ǫ2, this time scale is expected to differ signifi- cantly from typical relaxation rates Γ and Γ/2 for the di- agonal and nondiagonal components of the density matrix [31, 47]. Conclusions. -- In this Letter we provide a generic re- lation of the adiabatic response to real-time RG results for the stationary case. The presented scheme allows for the variation of any parameter in linear or nonlinear response and provides criteria when the adiabatic correction to the Liouvillian can be calculated directly via energy and time derivatives of the instantaneous one. We suggest a delay time as an interesting time scale and show for the IRLM that its expression is robust against the choice of time- dependent parameters and their amplitude. We confirm the universality of the AC relaxation resistance, unless a time dependence of tunneling and interaction is present, revealing logarithmic renormalizations due to charge fluc- tuations. We proposed different setups in molecular elec- tronics and cold-atom systems to observe the effects ex- perimentally. ∗ ∗ ∗ We acknowledge valuable discussion with S. Ander- gassen and M. Pletyukhov, and useful comments by M. Buttiker, as well as financial support from the Ministry of Innovation NRW and DFG-FG 723. REFERENCES [1] Buttiker M., Thomas H., and Pretre A., Z. Phys. B, [11] Splettstoesser J. et al., Phys. Rev. B, 74 (2006) 085305. [12] Reckermann F., Splettstoesser J. and Wegewijs M. R., Phys. Rev. Lett., 104 (2010) 226803. [13] Splettstoesser J. et al., Phys. Rev. B, 81 (2010) 165318. [14] Schiller A. and Silva A., Phys. Rev. B, 77 (2008) 045330. [15] Aono T., Phys. Rev. Lett., 93 (2004) 116601. [16] Nigg S. E., L´opez R. and Buttiker, M., Phys. Rev. Lett., 97 (2006) 206804. [17] Rodionov Y. I., Burmistrov I. S. and Ioselevich A. S., Phys. Rev. B, 80 (2009) 035332. [18] Mora C. and Le Hur K., Nat. Phys., 6 (2010) 697. [19] Filippone M., Le Hur K. Phys. Rev. Lett., 107 (2011) 176601. and Mora C., [20] Hamamoto Y. et al., Phys. Rev. B, 81 (2010) 153305. [21] Lee M. et al., Phys. Rev. B, 83 (2011) 201304. [22] Schoeller H., Eur. Phys. J. Special Topics, 168 (2009) 179. [23] Leijnse M. and Wegewijs M. R., Phys. Rev. B, 78 (2008) 235424. [24] D. Kennes and V. Meden, arXiv:1204.2100. [25] Pletyukhov M. and Schoeller H., arXiv:1201.6295. [26] Schoeller H. and Reininghaus F., Phys. Rev. B, 80 (2009) 045117. [27] Pletyukhov M., Schuricht D. and Schoeller H., Phys. Rev. Lett., 104 (2010) 106801. [28] Schuricht D. and Schoeller H., Phys. Rev. B, 80 (2009) 075120. [29] Pletyukhov M. and Schuricht D., Phys. Rev. B, 84 (2011) 041309. [30] Horig C. B. M., Schuricht D. and Andergassen S., Phys. Rev. B, 85 (2012) 054418. [31] Karrasch C. et al., Europhys. Lett., 90 (2010) 30003; Andergassen S. et al., Phys. Rev. B, 83 (2011) 205103. [32] Feve G. et al., Science, 316 (2007) 1169; Moskalets M., Samuelsson P. and Buttiker M., Phys. Rev. Lett., 100 (2008) 08660. [33] Mehta P. and Andrei N., Phys. Rev. Lett., 96 (2006) 216802; cond-mat/0703426 (Erratum). [34] Boulat E. and Saleur H., Phys. Rev. B, 77 (2008) 033409. [35] Boulat E. Saleur H. and Schmitteckert P., Phys. Rev. Lett., 101 (2008) 140601. [36] Karrasch C. et al., Phys. Rev. B, 81 (2010) 125122. [37] Rammer J. and Smith H., Rev. Mod. Phys., 58 (1986) 94 (1994) 133. 323. [2] Brouwer P. W., Phys. Rev. B, 58 (1998) R10135. [3] Pothier H. et al., Europhys. Lett., 17 (1992) 249. [4] Giazzotto F. et al., Nat. Phys., 7 (2011) 857. [5] Buttiker M., Thomas H., and Pretre A., Phys. Lett. A, 180 (1993) 364. [6] Gabelli J. et al., Science, 313 (2006) 499. [7] Andergassen S. et al., Nanotechnology, 21 (2010) 272001. [8] Splettstoesser J. et al., Phys. Rev. Lett., 95 (2005) 246803. [38] Moskalets M. and Buttiker M., Phys. Rev. B, 69 (2004) 205316. [39] Moskalets M. and Buttiker M., Phys. Rev. B, 78 (2008) 035301. [40] Eidelstein E., Goberman D. and Schiller A., in preparation. [41] Boese D. and Schoeller H., Europhys. Lett., 54 (2001) 668; Wegewijs M. R. and Nowack K. C., New J. of Physics, 7 (2005) 239. [42] Lang I. G. and Firsov Yu. A., Sov. Phys. JETP, 16 [9] Sela E. and Oreg Y., Phys. Rev. Lett., 96 (2006) (1963) 1301. 166802. [10] Fioretto D. and Silva A., Phys. Rev. Lett., 100 (2008) 236803. [43] Koch T. et al., J. Phys.: Conf. Ser., 220 (2010) 012014. [44] Recati A. et al., Phys. Rev. Lett., 94 (2005) 040404. [45] Tsvelick A. M. and Wiegmann P. B., Adv. in Physics, p-6 Nonlinear adiabatic response of interacting quantum dots 32 (1983) 453. [46] Leggett A. J. et al., Rev. Mod. Phys., 59 (1987) 1. [47] Egger R., Grabert H. and Weiss U., Phys. Rev. E, 55 (1997) R3809. p-7
1804.02936	1	1804	2018-04-09T12:19:44	Nonreciprocal microwave transmission based on Gebhard-Ruckenstein hopping	[ "cond-mat.mes-hall", "quant-ph" ]	We study nonreciprocal microwave transmission based on the Gebhard-Ruckenstein hopping. We consider a superconducting device that consists of microwave resonators and a coupler. The Gebhard-Ruckenstein hopping between the resonators gives rise to a linear energy dispersion which manifests chiral propagation of microwaves in the device. This device functions as an on-chip circulator with a wide bandwidth when transmission lines are attached.	cond-mat.mes-hall	cond-mat	Nonreciprocal microwave transmission based on Gebhard-Ruckenstein hopping Shumpei Masuda1,∗ Shingo Kono2, Keishi Suzuki2, Yuuki Tokunaga3, Yasunobu Nakamura2,4, and Kazuki Koshino1 1College of Liberal Arts and Sciences, Tokyo Medical and Dental University, Ichikawa, Chiba 272-0827, Japan 2Research Center for Advanced Science and Technology, The University of Tokyo, Meguro-ku, Tokyo 153-8904, Japan 3NTT Secure Platform Laboratories, NTT Corporation, Musashino 180-8585, Japan and 4RIKEN Center for Emergent Matter Science, Wako, Saitama 351-0198, Japan (Dated: November 6, 2018) We study nonreciprocal microwave transmission based on the Gebhard-Ruckenstein hopping. We consider a superconducting device that consists of microwave resonators and a coupler. The Gebhard-Ruckenstein hopping between the resonators gives rise to a linear energy dispersion which manifests chiral propagation of microwaves in the device. This device functions as an on-chip cir- culator with a wide bandwidth when transmission lines are attached. I. INTRODUCTION As erythrocytes transport oxygen from the lungs to the body tissues, microwaves can carry energy and in- formation between electromagnetic components in su- perconducting circuits [1 -- 6], which provide a promising platform for quantum information processing [1 -- 3, 7 -- 15]. Many of the quantum information processing schemes and the superconducting quantum optics experiments re- quire routing of microwaves in a cryostat. Therefore, cryogenic circulators are indispensable tools, and the loss at each circulator is detrimental especially for quan- tum information processings. This motivates the grow- ing body of experimental and theoretical works devoted to lossless on-chip microwave circulators, which possibly replace the commercial ferrite circulators. Various principles for achieving on-chip microwave cir- culators, as well as their practical designs, have been pro- posed [16 -- 23]. Several types of the devices with the nonreciprocal transmission of microwaves have been im- plemented such as, electrically driven nonreciprocity on a silicon chip [24], a circulator based on a Josephson circuit [25], a fiber-integrated quantum optical circula- tor operated by a single atom [26], chiral ground-state currents of interacting photons in three qubits based on a synthetic magnetic field [27], on-chip nonrecipro- cal current based on a combination of frequency conver- sion and delay [28] and the ones based on optomechani- cal circuits [29 -- 31]. The nonreciprocal signal routing in photonic resonator lattice systems has been also studied [32, 33]. In this paper, we investigate the microwave response of a system of the coupled resonators with the Gebhard- Ruckenstein (GR) hopping [34, 35]. The GR hopping gives rise to a linear energy dispersion, which manifests the chiral hopping of cavity photons. This system func- tions as a circulator when transmission lines are attached to some (three or more) of the resonators. This paper is organized as follows. In Sec. II we discuss the property of the system with GR hopping. In Sec. III we intro- duce our circulator based on the GR hopping. In Sec. IV we analyze the microwave response of the system and demonstrate the robustness of the routing efficiency of the circulator. Section V is devoted to conclusion. II. GEBHARD-RUCKENSTEIN HOPPING In this paper we propose a method to route microwaves based on the GR hopping between the resonators. To illustrate the property of GR hopping we introduce a system, which we call GR cluster, consisting of N bosonic sites (resonators) with the GR hopping as depicted in Fig. 1 . FIG. 1. Schematic of a GR cluster for N = 5. The dashed lines represent the coupling between the sites. It is known that the GR hopping gives rise to a lin- ear energy dispersion, which manifests a chiral current in the system. The GR hopping was also used to model the helical current on the edge of a two-dimensional topo- logical insulator [36]. The Hamiltonian of the GR cluster is represented as H (GR) cluster = ηm,nc† mcn, (1) (cid:88) m,n ∗ [email protected] ηm,n = η∗ n,m = with the bosonic annihilation operator cm for site m and the coupling constant between site m and site n given by (cid:40) iπη0(−1)n−m N sin π(n−m) 0 N (n (cid:54)= m) (n = m) , (2) 12543where η0 is a real constant. The single-particle eigen- states and their eigenenergies are represented as N(cid:88) ν(cid:105) = φν(m)m(cid:105), m=1 Eν = η0kν, (3) respectively, with the wave function φν(m) = exp(ikνm), where ν = −(N − 1)/2,··· , (N − 1)/2 for 1√ odd N , and ν = −N/2,··· , N/2 − 1 for even N . Here, N m(cid:105) represents the state in which the particle (photon) is localized at site m. kν is given by kν = 2πν/N for odd N , and kν = 2π(ν + 1/2)/N for even N . Figures 2(a) and 2(b) show ηm,n/(iη0) for N = 9 and 8, respectively. The sign of ηm,n changes alternately with respect to the site number n (except for n = N for even N ). Note that the coupling constants are cyclic for odd N , that is, ηm,n = ηm+j,n+j for integer j, where the indices are understood modulo N . In contrast, this does not hold for even N : ηN,1 = −η1,2 (see Fig. 2(b)). Figures 3(a) and 3(b) plot the eigenenergies as the functions of wave number. Figures 3(c) and 3(d) show the phase of the wave function of the fourth lowest level relative to that of site 1. The phase of the wave func- tion of the eigenstates changes approximately 2πν from site 1 to site N for odd N , while it changes approximately 2π(ν + 1/2) for even N as shown in Figs. 3(c) and 3(d). In Fig. 4, we observe the dynamics of a particle (pho- ton) in the GR clusters. The initial state is set as Ψ(cid:105) = 1(cid:105). Figures 4(a) and 4(b) show the time-evolution of the population at each site for the systems with N = 5 and 6, respectively. The figures clearly show the chiral population transfer in the GR clusters. 2 FIG. 2. Coupling constant in the GR model. ηm,n for m = 1, 2 are shown for N = 9 and 8 in panels (a) and (b), respectively. Note that the rightmost and leftmost values in the panels are identical. The dashed lines are the guide to the eyes. The arrows in panel (a) indicate the cyclicity of the Hamiltonian elements for N = 9, and the arrows in panel (b) indicate the lack of the cyclicity for N = 8. Hamiltonian of the system is given by III. SYSTEM Hcluster = ωma† mam + N(cid:88) m,n((cid:54)=m) gm,n(t)a† man, H = Hcluster + Hdamp, m=1 N(cid:88) (cid:90) N(cid:88) (cid:114) vκm m=1 (cid:104) A. Circulator based on GR hopping In this section, we propose a way of routing microwaves based on the GR hopping between resonators. As de- picted in Fig. 5, our system consists of the N resonators with different resonance frequencies, a coupler of the res- onators and transmission lines coupled to some of the res- onators. We assume that each resonator is coupled to the other resonators with time-dependent coupling strength. (Physical realization is discussed in Appendix D.) The Hdamp = dk † m,kbm,k vkb (cid:105) (a† † m,kam) mbm,k + b . + 2π (4) Here, Hcluster describes the N coupled resonators, and Hdamp describes the interactions between the transmis- sion lines and the resonator modes. am and bm,k are the annihilation operators of the mode of resonator m and the mode of the transmission line m with wave number k, respectively. We refer to the transmission line attached to -1.5-1-0.5 0 0.5 1 1.5 1 2 3 4 5 6 7 8 9(a) 6-1.5-1-0.5 0 0.5 1 1.5 1 2 3 4 5 6 7 8(b) 53 frequency of the mode of resonator m, v is the microwave velocity in the transmission lines, κm is the decay rate of a photon from resonator m into transmission line m. Note that transmission lines are not attached to some of the resonators. For example, in Fig. 5, κ2 = κ5 = 0. The coupling between resonators n and m is modulated periodically in time as gm,n(t) = 2gm,n cos (cid:104) (cid:105) (ωm − ωn)t + θm,n , (5) where gm,n and θm,n are time independent, and gn,m(t) = g∗ m,n(t). The frequency of gm,n in Eq. (5) was set so that resonator m couples to resonator n. In the rotating frame the coupling strength between res- onators m and n becomes independent of time and is given by gm,ne−iθm,n (see Appendix A for details). Thus, the phase of the coupling strength can be tuned in-situ via θm,n. FIG. 3. Eigenenergies of H (GR) cluster as the functions of kν for (a) N = 9 and (b) N = 8. Panels (c) and (d) show the phase of wave function φ−1(l) relative to that of φ−1(1) for N = 9 and 8, respectively. FIG. 5. Schematic of a circulator with N = 5. The input field is applied through one of the transmission lines. The blue circles, the green circle and gray lines represent the resonators, the coupler, and the transmission lines, respectively. The dashed lines represent the coupling between the resonators. Transmission lines are uncoupled to resonators 2 and 5 in this example. Hereafter, we restrict ourselves to the case in which the coupling between the resonators are of the GR type and three of the resonators are coupled to the transmission lines (see Fig. 5). We set gm,n and θm,n in Eq. (5) as πη0(−1)n−m N sin π(n−m) N , gm,n = θm,n = − π 2 , (6) FIG. 4. Time-evolution of the population at each site for the systems with (a) N = 5 and (b) N = 6. pi denotes the population of site i. resonator m as transmission line m. ωm is the resonance so that the coupling strengths between the resonators, gm,ne−iθm,n , become the GR type. As shown in Sec. IV this system has the non-reciprocal transmission property and functions as a circulator. We comment here on the relation between our circu- lator and that studied previously. The N = 3 case in Eq. (4) was studied in Ref. 27 to show that the unit routing efficiency (S12 = S23 = S31 = 1) is achieved 0 0 0-2-1.5-1-0.5 0 1 2 3 4 5 6 7 8 9phase /(cid:1)-2-1.5-1-0.5 0 1 2 3 4 5 6 7 8(a)(b)(c)(d)phase /(cid:1) 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 2 4 6 8 10populationtime 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 2 4 6 8 10populationtime(a)(b)couplerwhen gm,n = g, κm = 2g, θ1,2 = θ2,3 = θ3,1 = π/2. (7) Note that θn,m = −θm,n. Our circulator based on the GR hopping for N = 3 with η0 = 3g sin(π/3)/π and κm = 2g satisfies Eq. (7) and reduces to the same system. IV. RESULTS Wide bandwidth is a desirable property of a circulator. We examine the stability of the routing efficiency against detuning of the incident microwaves, assuming that the strength of the coupling between resonators is limited. We consider the case in which every transmission line is coupled to a resonator with the same strength, κ, and the amplitude of the nearest neighbor hopping of the GR cluster, which is the largest, is fixed to be g unless it is stated that we consider other cases. Hereafter g is used as the unit of detuning. In the following, we first present the results for odd N , and then show the results for even N . A. S-matrix 4 FIG. 6. Circulation properties for the case of N = 3 and κ1,2,3 = κ. (a) Dependence of the forward transmission prob- abilities (S122, S232 and S312) on κ for ∆ω = 0. The inset shows the system configuration. (b) Dependence of the for- ward and backward transmission and reflection probabilities on ∆ω for κ = 2g. We consider the injection of continuous microwave from one of the transmission lines, transmission line p. As shown in Appendix A, the transmission and reflection coefficients of the microwave are given by the S-matrix elements represented as Sp,m = δp,m − √ κpκm[G−1]m,p, (8) where [G−1]m,p is the element of matrix G−1. The ele- ments of matrix G are given by (cid:40) Gm,n = κm/2 − i∆ω (n = m), (n (cid:54)= m), igm,ne−iθm,n (9) C. five-resonator system where ∆ω(≡ ωin − ωp) is the detuning of the incident microwave, and ωin is the frequency of the incident mi- crowave. B. three-resonator system For the reference, we examine the routing efficiency of the three-resonator system. Figure 6(a) shows the depen- dence of the forward transmission probabilities (counter clockwise) on κ for N = 3 without detuning. The for- ward transmission probabilities become unity at κ = 2g. Figure 6(b) shows the dependence of the forward and backward transmission and reflection probabilities on de- tuning ∆ω with κ = 2g. Figure 7(a) shows the dependence of the forward trans- mission probabilities on κ for N = 5 without detun- ing, ∆ω = 0. The transmission lines are coupled to resonators 1, 3 and 4. We numerically confirm that S13 = S34 = S41. These equalities are analytically derived in Appendix B. The forward transmission prob- abilities become unity at κ = 4g. This value of κ is twice larger than the ideal value of κ(= 2g) for N = 3. Fig- ure 7(b) shows the dependence of the forward and back- ward transmission and reflection probabilities on detun- ing ∆ω with κ = 4g. It is seen that the forward trans- mission probabilities for N = 5 are higher than that for N = 3. S33 and S44 (S43) exhibit similar ∆ω-dependence to S11 (S14 and S31), and they are not shown here. The robustness of the routing efficiency against detun- ing depends on which resonators the transmission lines 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8Transmission probability (a) 0 0.2 0.4 0.6 0.8 1-8-6-4-2 0 2 4 6 8(b)Transmission probability 1235 are attached to. In Fig. 8, we make the same plot as Fig. 7 except that the transmission lines are attached to resonators 1,2 and 5. Figure 8(a) shows the dependence of the forward transmission probabilities on κ. The opti- mal value of κ, with which S12, S25, and S51 become unity, is approximately 2.472g. Figure 8(b) shows the de- pendence of the forward and backward transmission and the reflection probabilities on detuning for κ = 2.472g. The transmission probabilities are comparable to that for N = 3 for ∆ω/g < 0.5 as seen in the inset of Fig. 8(b). Therefore, the configuration in Fig. 7(a) is more desirable than the one in Fig. 8(a). S22 and S55 (S52) exhibit a similar ∆ω-dependence to S11 (S15 and S21), and they are not shown here. The results for N = 7 system is shown in Appendix C. Circulation properties for the case of N = 5, FIG. 8. (a) Dependence of the forward κ1,2,5 = κ, and κ3,4 = 0. S252 and S512 ) on κ transmission probabilities ( S122, for ∆ω = 0. The inset shows the system configuration. (b) Dependence of the forward and backward transmission and reflection probabilities on ∆ω for κ = 2.472g. The black line 12 2. The inset is a closeup around ∆ω/g = 0. represents S(3) D. four-resonator system Figure 9(a) shows the dependence of the forward trans- mission probabilities on κ for N = 4 without detuning. The transmission lines are coupled to resonators 1, 2 and 4. Under the condition of κ1 = κ2 = κ4, it is seen that S12, S24, S41 do not reach unity simultaneously. There- fore, this system does not work as a circulator. Now we consider the case with κ1 (cid:54)= κ2 = κ4. We opti- mize κ1 and κ2 so that the product of the forward trans- mission amplitudes defined by S12S24S41 is maximized. Figure 9(b) shows the dependence of the forward and backward transmission and reflection probabilities on the detuning for κ1 = 2.14g and κ2,4 = 4.24g. It is seen that the forward transmission probabilities are almost unity for ∆ω = 0, and the routing efficiency is robust against FIG. 7. Circulation properties for the case of N = 5, κ1,3,4 = κ, and κ2,5 = 0. (a) Dependence of the forward transmission probabilities (S132, S342 and S412) on κ for ∆ω = 0. The inset shows the system configuration. (b) De- pendence of the forward and backward transmission and re- flection probabilities on ∆ω for κ = 4g. The thin black line 12 2. The inset is represents S122 for N = 3, denoted by S(3) a closeup around ∆ω/g = 0. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-8-6-4-2 0 2 4 6 8 0.8 1-1 1Transmission probability (b) 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1 2 3 4 5 6 7 8Transmission probability (a)1345-8-6-4-2 0 2 4 6 8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.75 1-1 1Transmission probability (b) 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8Transmission probability (a)12345the detuning more than the circulator for N = 3. S22 and S44 (S42) show a similar ∆ω-dependence to S11 (S14 and S21), and they are not shown here. 4.328g. As seen in the inset, the routing efficiency is obviously robust against the detuning compared to the systems with other cavity number N . 6 FIG. 9. Circulation properties for the case of N = 4 and κ3 = 0. (a) Dependence of the forward transmission prob- abilities ( S122, S242 and S412 ) on κ for ∆ω = 0 when κ1,2,4 = κ. The inset shows the system configuration. (b) De- pendence of the forward and backward transmission and re- flection probabilities on ∆ω for the case of κ1 = 2.14g and κ2,4 = 4.24g. Circulation properties for the case of N = 6, FIG. 10. (a) Dependence of the forward κ1,3,5 = κ, and κ4,6 = 0. S352 and S512 ) on κ transmission probabilities ( S132, for ∆ω = 0. The inset shows the system configuration. (b) Dependence of the forward and backward transmission and reflection probabilities on ∆ω for κ = 4.328g. The black line represents S(3) 12 2. The inset is a closeup around ∆ω/g = 0. E. six-resonator system Figure 10(a) shows the dependence of the transmis- sion probabilities on κ for N = 6 without detuning. The transmission lines are coupled to resonators 1, 3 and 5. As expected from the rotational symmetry, S13 = S35 = S51. The transmission probabilities become unity when κ (cid:39) 4.328g. Figure 10(b) shows the dependence of the transmis- sion and reflection probabilities on the detuning for κ = F. Systems with large number of resonators To see an asymptotic property we examine large N sys- tems. The transmission lines are attached in a way that the system is geometrically symmetric against 2π/3 rota- tion for concreteness. Figure 11(a) shows the dependence of the transmission probabilities on κ for N = 195 with- out detuning. The transmission probabilities become unity when κ (cid:39) 4g. The dependence of the transmis- sion probabilities on ∆ω asymptotically changes to the ones shown in Fig. 11(b) when N becomes sufficiently 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8Transmission probability (a)1234Transmission probability -1 1 0.75 1(b) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-8-6-4-2 0 2 4 6 8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-8-6-4-2 0 2 4 6 8-1 1 0.75 1(b)Transmission probability 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8(a)Transmission probability 123456large. These asymptotic profiles do not depend much on the parity of N and where the transmission lines are attached as long as they are separated from each other sufficiently. The transmission and the reflection probabil- ities for N = 192 show the ∆ω-dependence qualitatively similar to the ones for N = 195, although they are not exhibited here. Note that the system with N = 6 is more robust against the detuning around ∆ω/g = 0 than these large N systems. The forward transmission probability for large N sys- tem is close to unity for −πg < ∆ω < πg. The asymp- totic value of the bandwidth of 2πg is attributed to the energy band of the GR cluster, which is from −gπ to gπ. The incoming microwave can enter to the GR clus- ter and can propagate as a plane wave if its energy is in that range, otherwise it is reflected. G. Effects of parameter fluctuations Here, we observe the effects of fluctuations of the sys- tem parameters. To observe the effects of fluctuation in κ, we replace κ1 and κ3 with λκ1κ1 and λκ3κ3, respec- tively, for the six-resonator system discussed in Sec. IV E. Figure 12(a) shows the dependence of S132 on λκ1 and S132 is approximately 0.93 when κ1 and κ3 have λκ3. 30% of inhomogeneity. Transmission probabilities S352 and S512 (not shown) are higher than 0.965 in the same range of λκ1 and λκ3. To observe the effects of the fluctuation in g, we make the following replacements: g13/31 → λg13g13/31 and g35/53 → λg35g35/53. Figure 12(b) shows the dependence of S132 on λg13 and λg35. S132 is approximately 0.93 when the coupling strengths have 30% of inhomogeneity. Transmission probability S352 is higher than 0.99 and S512 is higher than 0.93 in the same range of λg13 and λg35 (not shown). Finally, to observe the effects of the fluctuation in the phase of g, we replace g13 and g35 with eiθ1g13, and eiθ2 g35. Figure 12(c) shows the dependence of S132 on θ1 and θ2. Transmission probability S352 is higher than 0.96 and S51 = S13 in the same range of θ1 and θ2, although they are not shown here. It is seen that S13 is sensitive to θ1 compared to θ2. H. Properties of circulator based on GR hoppings Here we discuss the general properties of the circulator based on GR hopping with transmission lines A, B and C. As we observed in Sec. IV C, several equalities hold in the transmission probabilities for odd N . We numerically confirmed that SAB = SBC = SCA, SAC = SCB = SBA and SAA = SBB = SCC for ∆ω = 0 irrespectively of the resonators to which the transmission lines are attached. The equalities for N = 5 and 7 are derived analytically in Appendix B. 7 FIG. 11. Circulation properties for the case of N = 195, κ1,66,131 = κ, and κi = 0 for i (cid:54)= 1, 66, 131. (a) Dependence of the forward transmission probabilities on κ for ∆ω = 0. The inset schematically shows the system configuration for large N . (b) Dependence of the forward and backward transmission and reflection probabilities on ∆ω for κ (cid:39) 4g. The red line is for the forward transmission probability. The green and the purple lines are for the backward transmission and reflection probabilities, respectively. The black solid and dashed lines are for the forward transmission probabilities for N = 6 sys- tem in Fig. 10(b) and N = 3, respectively. The right inset is the closeup around ∆ω/g = 0. On the other hand, these equalities do not generally hold for even N . However similar equalities hold when the transmission lines are attached in a way that the system is geometrically symmetric against 2π/3 rotation (see also Appendix B for the N = 6 case). Importantly, the equalities which is desirable for a circulator, SAB = SBC = SCA, hold for the six-resonator system as seen in Fig. 9(b). I. Drawback and other possible configurations The drawback of our scheme is the number of the re- quired couplings between resonators, N (N − 1)/2, in- Transmission probability (a) 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8(b) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-8-6-4-2 0 2 4 6 8Transmission probability 0.97 1-1 0 1types of long range hopping will be also studied. 8 V. CONCLUSION We have proposed an on-chip microwave circulator based on the Gebhard-Ruckenstein (GR) hopping. The linear energy dispersion of the GR cluster gives rise to a chiral propagation of a microwave, and thus can work as a circulator when transmission lines are attached. Our circulators composed of more than three resonators can have a wider operating bandwidth than that composed of three resonators. Especially, the circulator composed of six resonators with the three fold rotational symmetry has a remarkably wide operating bandwidth. The robust- ness of the routing efficiency against the inhomogeneity in the system parameters has also been examined. ACKNOWLEDGMENTS We acknowledge the support from JST ERATO (Grant No. JPMJER1601). KK is grateful to JSPS for a support from JSPS KAKENHI (Grant No. 16K05497). Appendix A: Derivation of S-matrix We derive the S-matrix of the system, which is de- scribed by the Hamiltonian of Eq. (4) [37,38]. The Heisenberg equation of motion of bm,k is represented as (cid:114) vκm d dt bm,k(t) = −ivkbm,k(t) − i We define the real-space representation(cid:101)bm,r of the trans- 2π am(t). (A1) mission line field as (cid:101)bm,r = 1√ 2π (cid:90) ∞ −∞ dkeikrbm,k. (A2) Tildes are used to distinguish the operator in k-space representation from the one in r-space representation in this section. In this representation, the field interacts with resonator at r = 0, and the r < 0 (r > 0) region corresponds to the incoming (outgoing) field. The input and output field operators of the transmission line field are defined by (cid:101)b(in) m,r(t) =(cid:101)bm,−r(t), (cid:101)b(our) m,r (t) =(cid:101)bm,r(t), (cid:114) κm m,vt(0) − i v (cid:101)b(out) m,+0(t) =(cid:101)b(in) (A3) am(t). (A4) respectively, where r > 0. Using Eqs. (A1) and (A3), the input-output relation is derived as [37, 38] FIG. 12. Effects of the fluctuations in the system param- eters. The system with N = 6 in Fig. 9(c) is investigated. Dependence of S132 on (a) λκ1 and λκ3, (b) λg13 and λg35, (c) θ1 and θ2. The values next to the contour lines indicate the values of S132. creasing with the number N of resonators. Unwanted crosstalk between resonators and unwanted excitations of the coupler should be avoided. Thus, it becomes ex- perimentally more challenging when N increases. We discuss the physical realization of our circulator with a concrete circuit model in Appendix D. In the present article we have mainly studied the sys- tems with small number of resonators N and with only three transmission lines. Only a part of the possible ar- rangements of the transmission lines have been examined, although there are many other choices in the arrange- ment for cases with large N . The optimized arrangement will be studied for larger N in an experimentally feasible range elsewhere, and routing microwaves based on other 0.7 0.8 0.9 1 1.1 1.2 1.3 0.7 0.8 0.9 1 1.1 1.2 1.3 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 0.7 0.8 0.9 1 1.1 1.2 1.3 0.7 0.8 0.9 1 1.1 1.2 1.3 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1-0.1-0.05 0 0.05 0.1-0.1-0.05 0 0.05 0.1 0.84 0.86 0.88 0.9 0.92 0.94 0.96 0.98 1(b)(c)(a)0.9950.9850.9750.9650.9950.9850.9750.9650.990.970.950.930.910.890.87tion of motion of am is represented as m,+0(t). The Heisenberg equa- In Eq. (A4)(cid:101)b(in) d dt am = m,vt(0) =(cid:101)b(in) (cid:17) (cid:16) − iωm − κm vκm(cid:101)b(in) √ −i m,vt. 2 (cid:88) n((cid:54)=m) am − i gm,n(t)an (A5) So far we discussed the operator equations. Here we as- sume that an input microwave is applied from the trans- mission line attached to the p-th resonator. We consider a continuous mode version of a coherent state: (cid:104)(cid:90) ∞ drEin(r)(cid:0)(cid:101)b(in) (cid:1)†(cid:105)0(cid:105), Ψi(cid:105) = N exp 0 (A6) with the overall vacuum state 0(cid:105) and a normalization constant N . Considering that the input wave propagates in the negative r direction, Ein(r) represents the input microwave at the initial moment as given by p,r Ein(r) = Ee−iωinr/v, (A7) where E and ωin are the amplitude and the angular fre- quency of the incident microwave, respectively. We as- sume that at the initial moment the resonators and the transmission lines except for the input one are unexcited, and the input microwave has not arrived at resonator p Ψi(cid:105) is in a coherent state and therefore an eigen- yet. state of the initial field operator(cid:101)bp,r(0). It is confirmed that(cid:101)b(in) m,vtΨi(cid:105) = Ein(vt)δm,pΨi(cid:105) = Ee−iωintδm,pΨi(cid:105) using Eqs. (A3) and (A6). We rewrite Eq. (A5) as (cid:16) (cid:17) vκm(cid:101)b(in) i∆ω − κm 2 √ −i (cid:88) n((cid:54)=m) Am − i m,vtei(ωm+∆ω)t d dt Am = (A8) gm,n(t)ei(ωm−ωn)tAn (A9) to Ψi(cid:105) we obtain d dt (cid:104)Am(cid:105) = (cid:16) i∆ω − κm 2 √ −i vκmEδm,p. (cid:17)(cid:104)Am(cid:105) − i (cid:88) n((cid:54)=m) 9 gm,ne−iθm,n(cid:104)An(cid:105) (A12) To obtain the stationary solution in the rotating frame we put d(cid:104)Am(cid:105)/dt = 0. Then Eq. (A12) is rewritten in the matrix form as where the (m, n) element of matrix G is given by vκpE −→ φp, G−→(cid:104)A(cid:105) = −i √ (cid:40) −→(cid:104)A(cid:105) =(cid:0)(cid:104)A1(cid:105),··· ,(cid:104)AN(cid:105)(cid:1), κm/2 − i∆ω (n = m) (n (cid:54)= m), igm,ne−iθm,n Gm,n = and and the p-th component of −→(cid:104)A(cid:105) is written with the inverse of matrix G as 0. Then −→ φp is 1, while the others are −→(cid:104)A(cid:105) = −i √ vκpEG−1−→ φ p. We multiply Eq. (A4) by ei(ωm+∆ω)t and take the ex- pectation value with respect to Ψi(cid:105) to obtain (cid:104)Am(t)(cid:105), (cid:114) κm (A17) v where (cid:104)(cid:101)B(out) m,+0(t)(cid:105) = (cid:104)(cid:101)B(in) m,vt(cid:105) − i m,+0(t) =(cid:101)b(out) (cid:101)B(out) m,vt =(cid:101)b(in) (cid:101)B(in) the stationary solution of (cid:104)(cid:101)B(out) m,vt(cid:105) − i m,+0(t)ei(ωm+∆ω)t m,vtei(ωm+∆ω)t. (cid:114) κm m,+0(cid:105) is given by (cid:104)Am(cid:105), (A18) Using the stationary solution of (cid:104)Am(cid:105) given by Eq. (A16), v (cid:104)(cid:101)B(out) m,+0(cid:105) = (cid:104)(cid:101)B(in) (cid:104)(cid:101)B(in) m,vt(cid:105) = δm,pE, (cid:104)(cid:101)bm,+0(t)(cid:105) (cid:104)(cid:101)bp,−0(t)(cid:105) . Sp,m = (A13) (A14) (A15) (A16) (A19) (A20) (A21) with Am defined by where Am(t) = ei(ωm+∆ω)tam(t), (A10) where ∆ω is the detuning of the incident microwave with angular frequency ωin, that is, ωin = ωp + ∆ω. Using Eqs. (5) and (A9) and the rotating wave approximation, we obtain d dt Am = (cid:16) (cid:17) vκm(cid:101)b(in) i∆ω − κm 2 √ −i (cid:88) Am − i n((cid:54)=m) m,vtei(ωm+∆ω)t. gm,ne−iθm,n An (A11) as and(cid:101)b(in) m,vt(0) =(cid:101)b(in) With the use of Eqs. (A16), (A18), (A19), (A20), (A21) m,+0(t), the S-matrix element is given because of Eq. (A8). The elements of S-matrix are de- fined by Taking the expectation value of Eq. (A5) with respect Sp,m = δp,m − √ = δp,m − √ κpκm[G−1−→ φp]m, κpκm[G−1]m,p, (A22) where [G−1−→ φp]m denotes the m-th component of vector, G−1−→ φp. This is identical to [G−1]m,p, since the p-th com- −→ φp is 1 and the others are 0. ponent of S−matrix such as 31/53 = −(cid:2)G−1(cid:3) (cid:2)G−1(cid:3) 35/13 = −(cid:2)G−1(cid:3) (cid:2)G−1(cid:3) 11 =(cid:2)G−1(cid:3) (cid:2)G−1(cid:3) 15, 51, 33 =(cid:2)G−1(cid:3) 55. Appendix B: Equalities between S-matrix elements From these equalities, we have S13 = S35 = S51, S15 = S53 = S31, S11 = S33 = S55. 10 (B6) (B7) We analytically show the equalities between the S- matrix elements, which are numerically confirmed in Secs. IV C and IV E. We consider the matrix G in Eq. (A13) for the system with N = 5 depicted in Fig. 5. The matrix G has the form as (B1) (B2) (B3) (B5) Appendix C: Results for seven-resonator system Figure 13(a) shows the dependence of the forward transmission probabilities on κ for N = 7. The transmis- sion lines are coupled to resonators 1, 3 and 6 as shown in the inset. It is numerically confirmed that the forward transmission probabilities, S132, S362 and S612, are identical in spite of the absence of the three fold rota- tional symmetry and become unity at κ (cid:39) 4.45g. We also analytically confirmed the equalities of the forward transmission probabilities in the same manner as Ap- pendix B. Figure 13(b) and 13(c) show the dependence of the forward transmission probabilities on detuning ∆ω for κ = 4.45g. The transmission probabilities for N = 7 are generally higher than those for N = 5 in Fig. 7(b) in a wide range of ∆ω. The profile of the transmission probabilities for N = 7 is close to a rectangular shape compared to the one for N = 3. The backward trans- mission and the reflection probabilities for N = 7 show the ∆ω-dependence qualitatively similar to the ones for N = 5 in Fig. 7(b), although they are not exhibited here. Appendix D: Physical realization In this section we discuss the physical realization of our circulator with a concrete circuit model with N = 5. Figure 14 shows a circuit system of a possible physical realization of the circulator. The system is composed of a Josephson ring, resonators and transmission lines. The Josephson ring works as the coupler depicted in Fig. 5. A magnetic flux, Φ, is penetrating the Josephson ring. We derive an effective photon Hamiltonian for the sys- tem depicted in Fig.14 in a manner analogous to Ref.16. The Hamiltonian of the system is represented by H = Hres + HJring + Hint, (D1) where Hres, HJring and Hint respectively describe the res- onators, the Josephson ring and their interaction. Hres with real constants x, a and b when ∆ω = 0. The matrix elements of G−1 are represented as G =  (cid:2)G−1(cid:3) (cid:2)G−1(cid:3) (cid:2)G−1(cid:3)  , b −b −a x a −a 0 b −b a −b −a x b a b −b −a x a b −b −a 0 a 31/43/14 = 41/34/13 = 11/33/44 = A5 + B5x 3A5x + b2x3 , A5 − B5x 3A5x + b2x3 , A5 + b2x2 3A5x + b2x3 with A5 = a4 − 2a3b − a2b2 + 2ab3 + b4, B5 = −a2b + ab2 + b3. From Eq. (8), we obtain the following equalities between the S-matrix elements: S13 = S34 = S41, S14 = S43 = S31, S11 = S33 = S44. (B4) it is confirmed analytically that the same equalities hold in the system with N = 7. Now we consider the case with N = 6 depicted in Fig. 9(b). Because matrix G has the following form:  G =  , x a b c b a −a 0 a b b c −b −a x a c b −c −b −a 0 a b −b −c −b −a x a −a −b −c −b −a 0 for the case ∆ω = 0, we have the equalities relevant to 11 Circuit diagram of a possible physical realiza- FIG. 14. tion of the system depicted in Fig.5. Z0 is the characteristic impedance of the transmission lines. Cκ is the coupling ca- pacitance between the resonators and the transmission lines. Cc is the coupling capacitance between the resonators and the Josephson ring. CJ and EJ are the junction capacitance and the Josephson energy, respectively. and HJring is given by HJring = 1 2 −→ TC−1−→ Q −→ Q T −→ φ , Φ), Q + V ( (D3) −→ = (Q1, Q2, Q3, Q4, Q5) and with the charge vector the flux vector φ = (φ1, φ2, φ3, φ4, φ5), where Qµ and φµ are the charge and the flux on node µ of the Josephson ring, respectively, and satisfy the commutation relation, [φµ, Qν] = iδµ,ν. C is the capacitance matrix with ma- trix elements Cµ,µ±1 = −CJ and Cµ,µ = 2CJ +Cc, where CJ and Cc are the junction capacitance and the coupling capacitance between a resonator and the Josephson ring, respectively. The inductive energy is represented as −→ φ , Φ) = −EJ V ( cos (φµ+1 − φµ − Φ/5) , 5(cid:88) µ=1 (cid:104) 2π Φ0 with the Josephson energy EJ , where µ is modulo 5, that is, µ = 6 is identical to µ = 1. The interaction Hamilto- nian is given by (cid:88) (cid:0)eT mC−1−→ Q(cid:1)qmϕm, Hint = Cc m and the others are zero, and qm = (cid:112)ωm/2(am + a† mC−1−→ where em is the unit vector of which m-th element is unity m), and ϕm is the amplitude of the mode function at the coupling capacitance Cc. In Eq. (D5), eT Q corre- sponds to the voltage of the superconductive island m, (cid:105) (D4) (D5) FIG. 13. Circulation properties for the case of N = 7, κ1,3,6 = κ, and κ2,4,5,7 = 0. (a) Dependence of the forward transmission probabilities on the coupling to transmission line κ for ∆ω = 0. The inset shows the system configuration. (b) Dependence of the forward transmission probabilities on detuning ∆ω for κ = 4.45g. The thin lines correspond to the system with N = 5 in Fig. 7. (c) Closeup of (b) around ∆ω/g = 0. is given by 5(cid:88) m=1 Hres = ωma† mam, (D2) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-8-6-4-2 0 2 4 6 8 0.75 0.8 0.85 0.9 0.95 1-1.5-1-0.5 0 0.5 1 1.5Transmission probability Transmission probability (b)(c) 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1 2 3 4 5 6 7 8Transmission probability (a)1234567and qmϕm is the voltage at the end of resonator m. Hint can be represented also as Hint = Cc V res with V ring = C−1−→ −→ −→ res = (q1ϕ1, q2ϕ2,··· ). We rewrite V T (cid:88) Eq. (D5) as −→ V ring · −→ Q and ξm,µV (m) rms nµ(am + a† m), Hint = 2eCc (D6) m,µ (cid:112)ωm/2. where nµ = Qµ/(2e), and ξm,µ = [C−1]m,µ, and V (m) ϕm rms = Now we derive an effective photon lattice Hamiltonian in the dispersive regime, where the coupling between a resonator and the Josephson ring is sufficiently smaller than the energy difference between photonic and circuit excitations. We assume that the Josephson ring transfers microwaves via intermediate virtual excitations and re- mains in its ground state during the operation[16]. The effective Hamiltonian is obtained by the canonical trans- formation Hph = P0eiSHe−iSP0 = P0(Hres + HJring)P0 + P0[iS, Hint]P0/2 + O(H 3 int), (D7) where P0 is the projection operator onto the subspace in which the Josephson ring is in its ground state. Here, iS is defined by (cid:88) α,α(cid:48) iS = (cid:104)α(cid:48)Hintα(cid:105) Eα(cid:48) − Eα α(cid:48)(cid:105)(cid:104)α (D8) with εm = ωm + (2eCc)2(cid:88) (cid:104)(cid:16) tmn = 2(eCc)2(cid:88) k>0 1 (cid:104) 1 ωm − Ek Ξm,k2(cid:105) (cid:17) 1 , ωm − Ek + ωn − Ek k>0 Ξ∗ m,kΞn,k 12 (cid:105) , where (cid:88) µ Ξm,k = V (m) rms (D12) ξm,µnµk. (D13) Note that εm and tm,n depend on Φ through nµ,k in Eq. (D9). Therefore, the coupling among resonators can be tuned in-situ via Φ. We consider the tempo- ral modulation of Φ around fixed value Φ0, that is, Φ(t) = Φ0 + ∆Φ(t). Then, tm,n and εm are represented as tm,n(Φ) = tm,n(Φ0) + ∆tm,n(∆Φ), εm(Φ) = εm(Φ0) + ∆εm(∆Φ). (D14) We consider a narrow range of ∆Φ in which ∆tm,n and ∆εm are proportional to ∆Φ. Then, the time dependence of ∆tm,n and ∆εm are represented as (cid:12)(cid:12)(cid:12)Φ0 (cid:12)(cid:12)(cid:12)Φ0 dtm,n dΦ dεm dΦ ∆Φ(t), ∆tm,n(t) = ∆εm(t) = ∆Φ(t). (D15) with the eigenstates α(cid:105),α(cid:48)(cid:105) of Hres + HJring so that the first-order quantities in Hint are eliminated in Hph. The second term in Eq. (D7) is of the second order in Hint and gives rise to the mutual couplings between resonators. m by µ,kN0, 0(cid:105)(cid:104)N0, ka† m, respectively, where N0 is the relevant total charge of the Josephson ring which is conserved, and k(= 1, 2,··· ) runs over the excited states of the Josephson ring, and (cid:80) k nµ,kN0, k(cid:105)(cid:104)N0, 0am and (cid:80) In Eq. (D6) we approximate nµam and nµa† k n∗ Here, (cid:80) nµ,k = (cid:104)N0, knµN0, 0(cid:105). (D9) k nµ,kN0, k(cid:105)(cid:104)N0, 0am annihilates a photon in resonator m and yields an excitation in the Josephson ring. This approximation is based on the assumption that the Josephson ring remains in its ground states, and we neglect the counter rotating terms. Then, the inter- action Hamiltonian in Eq. (D5) is represented as Hint = (2eCc) ξm,µV (m) rms nµ,kamN0, k(cid:105)(cid:104)N0, 0 + h.c.. m,µ,k (D10) Using Eqs. (D7) and (D10), the effective photon lattice Hamiltonian is obtained as Hph = εma† mam + tm,na† man (D11) (cid:88) m,n((cid:54)=m) (cid:88) (cid:88) m (cid:88) The coupling among resonators can be tuned via the time-dependence of ∆Φ(t). We consider ∆Φ(t) repre- sented by ∆Φ(t) = Φl cos(Ωlt + θl), (D16) l where the index l runs over every pair of the resonators, and we set Ωl so that it matches to the difference of the resonance frequencies of resonator pair l. In the RWA the coupling strength between resonators m and n is repre- dΦ Φl=(m,n)e−iθl=(m,n). We assume that the sented as 1 2 influence of the second term of εm(t) is negligible in the RWA because it is rapidly oscillating. dtm,n Several comments are in order. (i) Although the eval- uation of εm and tm,n exceeds the scope of this paper, it was reported that tm,n/(2π) can exceed 20 MHz for N = 3 depending on Φ [16]. The frequency correspond- ing to the excitation energy from the ground state of the Josephson ring can be higher than 10 GHz. Thus, ex- citation of the Josephson ring induced by oscillating Φ with frequency of orders of 100 MHz is negligible. (ii) Ωl is chosen such that the oscillating Φ couples resonators. Each pair of the resonators should have different fre- quency difference for selective coupling. This limits the number of resonators. (iii) The electric potentials of the superconducting islands can be tuned via gate voltages to optimize the resonator coupling strength, although we consider the case that the gate voltages are zero for the simplicity. 13 [1] A. Blais, R.-S. Huang, A. Wallraff, S. M. Girvin and R. 13, 095008 (2011). J. Schoelkopf, Phys. Rev. A 69, 062320 (2004). [19] M. Hafezi and P. Rabi, Opt. Express 20, 7672 -- 7684 [2] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R.-S. Huang, J. Majer, S. Kumar, S. M. Girvin, and R. J. Schoelkopf, Nature 431, 162 (2004). [3] O. Astafiev, K. Inomata, A. O. Niskanen, T. Yamamoto, Y. A. Pashkin, Y. Nakamura and J. S. Tsai, Nature 449, 588 -- 590 (2007). [4] M. Hofheinz, H. Wang, M. Ansmann, R. C. Bialczak, E. Lucero, M. Neeley, A. D. O'Connell, D. Sank, J. Wen- ner, J. M. Martinis and A. N. Cleland, Nature 459, 546 (2009). [5] K. Inomata, Z. R. Lin, K. Koshino, W. D. Oliver, J. S. Tsai, T. Yamamoto and Y. Nakamura, Nat. Commun. 7, 7, 12303 (2016). [6] K. Y. Tan, M. Partanen, R. E. Lake, J. Govenius, S. Ma- suda and M. Mottonen, Nat. Commun. 8, 15189 (2017). [7] Y. Nakamura, Yu. A. Pashkin and J. S. Tsai, Nature 398, 786 -- 788 (1999). [8] O. Astafiev, A. M. Zagoskin, A. A. Abdumalikov, Jr., Y. A. Pashkin, T. Yamamoto, K. Inomata, Y. Nakamura, and J. S. Tsai, Science 327, 840 (2010). [9] E. L. Wolf, Quantum Nanoelectronics: An Introduction to Electronic Nanotechnology and Quantum Computing; (Wiley-VCH, Weinheim, 2009). [10] J. Majer, J. M. Chow, J. M. Gambetta, J. Koch, B. R. Johnson, J. A. Schreier, L. Frunzio, D. I. Schuster, A. A. Houck, A. Wallraff, A. Blais, M. H. Devoret, S. M. Girvin, and R. J. Schoelkopf, Nature 449, 443 (2007). [11] M. A. Sillanpaa, J. I. Park and R. W. Simmonds, Nature (2012). [20] J. Kerckhoff et al. Phys. Rev. Applied. 4, 043002 (2015). [21] A. Metelmann and A. A. Clerk, Phys. Rev. X 5, 021025 (2015). [22] L. Ranzani and J. Aumentado, New J. Phys. 17, 023024 (2015). [23] S. Walter and F. Marquardt, New J. Phys. 18, 113029 (2016). [24] H. Lira, Z. Yu, S. Fan and M. Lipson, Phys. rev. Lett. 109, 033901 (2012). [25] K. M. Sliwa, M. Hatridge, A. Narla, S. Shankar, L. Frun- zio, R. J. Schoelkopf, and M. H. Devoret, Phys. Rev. X 5, 041020 (2015). [26] M. Scheucher, A. Hilico, E. Will, J. Volz and A. Rauschenbeutel, Science 354 1577 (2016). [27] P. Roushan et al., Nat. Phys. 13, 146 (2017). [28] B. J. Chapman et al., Phys. Rev. X 7, 041043 (2017). [29] G. A. Peterson, F. Lecocq, K. Cicak, R. W. Simmonds, J. Aumentado and J. D. Teufel, Phys. Rev. X 7, 031001 (2017). [30] N. R. Bernier, L. D. T´oth, A. Koottandavida, M. A. Ioannou, D. Malz, A. Nunnenkamp, A. K. Feofanov and T. J. Kippenberg, Nat. Commun. 8, 604 (2017). [31] S. Barzanjeh, M. Wulf, M. Peruzzo, M. Kalaee, P. B. Di- eterle, O. Painter and J. M. Fink, Nat. Communications 8, 953 (2017). [32] K. Fang, Z. Yu and S. Fan, Nat. Photon. 6, 782 (2012). [33] A. Metelmann and H. E. Tureci, arXiv:1703.04052v1 449, 438 (2007). (2017). [12] M. H. Devoret and R. J. Schoelkopf, Science 339, 1169 [34] F. Gebhard and A. E. Ruckenstein, Phys. Rev. Lett. 68, (2013). 244 (1992). [13] J. Kelly et al., Nature 519, 66 (2015). [14] N. Ofek, et al., Nature 536, 441 (2016). [15] K. Koshino, K. Inomata, Z. R. Lin, Y. Tokunaga, T. Ya- mamoto and Y. Nakamura, Phys. Rev. Applied 7, 064006 (2017). [35] Y. Kuramoto and Y. Kato, One-Dimensional Quantum Systems; Cambridge, London, 2009. [36] S. Masuda and Y. Kuramoto, Phys. Rev. B 85, 195327 (2012). [37] K. Koshino and Y. Nakamura, New J. Phys. 14, 043005 [16] J. Koch, A. A. Houck, K. L. Hur, and S. M. Girvin, Phys. (2012). Rev. A 82 043811 (2010). [38] K. Koshino, K. Inomata, T. Yamamoto and Y. Naka- [17] A. Kamal, J. Clarke, and M. Devoret, Nat. Phys. 7, 311 mura, New J. Phys. 15, 115010 (2013). (2011). [18] A. Nunnenkamp, J. Koch, and S. M. Girvin, New J. Phys.
1802.01385	1	1802	2018-02-05T13:49:09	Chiral Anomaly, Topological Field Theory, and Novel States of Matter	[ "cond-mat.mes-hall" ]	Starting with a description of the motivation underlying the analysis presented in this paper and a brief survey of the chiral anomaly, I proceed to review some basic elements of the theory of the quantum Hall effect in 2D incompressible electron gases in an external magnetic field, ("Hall insulators"). I discuss the origin and role of anomalous chiral edge currents and of anomaly inflow in 2D insulators with explicitly or spontaneously broken time reversal, i.e., in Hall insulators and "Chern insulators". The topological Chern-Simons action yielding the large-scale response equations for the 2D bulk of such states of matter is displayed. A classification of Hall insulators featuring quasi-particles with abelian braid statistics is sketched. Subsequently, the chiral edge spin currents encountered in some time-reversal invariant 2D topological insulators with spin-orbit interactions and the bulk response equations of such materials are described. A short digression into the theory of 3D topological insulators, including "axionic insulators", follows next. To conclude, some open problems are described and a problem in cosmology related to axionic insulators is mentioned. As far as the quantum Hall effect and the spin currents in time-reversal invariant 2D topological insulators are concerned this review is based on extensive work my collaborators and I carried out in the early 1990's.	cond-mat.mes-hall	cond-mat	Chiral Anomaly, Topological Field Theory, and Novel States of Matter∗ Jurg Frohlich† February 6, 2018 Abstract Starting with a description of the motivation underlying the anal- ysis presented in this paper and a brief survey of the chiral anomaly, I proceed to review some basic elements of the theory of the quantum Hall effect in 2D incompressible electron gases in an external magnetic field, ("Hall insulators"). I discuss the origin and role of anomalous chiral edge currents and of anomaly inflow in 2D insulators with ex- plicitly or spontaneously broken time reversal, i.e., in Hall insulators and "Chern insulators". The topological Chern-Simons action yield- ing the large-scale response equations for the 2D bulk of such states of matter is displayed. A classification of Hall insulators featuring quasi- particles with abelian braid statistics is sketched. Subsequently, the chiral edge spin currents encountered in some time- reversal invariant 2D topological insulators with spin-orbit interactions and the bulk response equations of such materials are described. A short digression into the theory of 3D topological insulators, includ- ing "axionic insulators", follows next. To conclude, some open problems are described and a problem in cos- mology related to axionic insulators is mentioned. As far as the quantum Hall effect and the spin currents in time-reversal invariant 2D topological insulators are concerned this review is based on extensive work my collaborators and I carried out in the early 1990's. ∗Dedicated to the memory of Ludvig Dmitrievich Faddeev – a great scientist who will be remembered. To appear in: "Ludwig Faddeev Memorial Volume: A Life in Mathematical Physics", edited by Molin Ge, Antti Niemi, Kok Khoo Phua, and Leon A Takhtajan (World Scientific, 2018); http://www.worldscientific.com/worldscibooks/10.1142/10811 †Theoretical Physics, ETH Zurich, Wolfgang Pauli Strasse, 8093 Zurich, Switzerland; Email: [email protected] 1 1 General goals of analysis The purpose of this paper is to review some applications of the chiral anomaly and of topological field theory to condensed matter physics, more specifically to the theoretical analysis of novel, topologically protected states of matter. This is an area that has been of much interest to me, ever since the late 1980's, and to which I have made contributions that I feel are in- teresting and important. For this reason, it makes sense to highlight them, one more time, trying to put the emphasis on some of the most important concepts and most elegant ideas. The results sketched in the following are based on the papers [1, 2, 3, 4, 5, 6] (and references given therein) all pub- lished in the early nineties. I am confident that the material exposed here would have interested Ludvig Faddeev. Among many other important contributions he greatly enhanced our mathematical understanding of quantum gauge theories and of the origin and implications of the chiral anomaly. Some of his results on gauge theories and anomalies provide the theoretical underpinning for much of what I will discuss in the following. I start by sketching the general motivation underlying the efforts of my collaborators and myself (see [4, 5, 6, 7, 8]) that have led to the results dis- cussed in the following. Key purposes of our work are: • To classify bulk- and surface states of systems of condensed matter, using concepts and results from gauge theory, current algebra and General Relativity. In particular, effective actions (= generating func- tionals of connected current Green functions yielding expressions for transport coefficients, such as conductivities, via Kubo formulae) are used extensively in our analysis; their form being constrained by gauge- invariance, by the cancellation of gauge anomalies between bulk- and surface terms ("holography"), and by locality, unitarity and "power counting". It turns out that these general principles completely deter- mine the most relevant terms in the effective actions of, for example, insulators. Using effective actions, one discovers new ways of charac- terizing novel states of matter, as indicated next. • To complement the Landau Theory of Phases and Phase Transitions by a "Gauge Theory of Phases/States of Matter", with the purpose of 2 describing novel "topologically protected" states of matter that cannot be characterized by a local order parameter. • To apply our "Gauge Theory of Phases of Matter" to extend or de- velop: – The theory of the Fractional Quantum Hall Effect; (our work extended over the period from 1989 – 2012) – The theory of time-reversal invariant Topological Insulators and of Topological Superconductors; (our first paper appeared in 1993) – A description of higher-dimensional cousins of the Quantum Hall Effect, with applications in cosmology, e.g., towards understand- ing the origin of intergalactic primordial magnetic fields in the Universe, etc.; (1999 – the present) The chiral anomaly While vector currents, J µ, are usually conserved, i.e., ∂µJ µ = 0, axial cur- rents, J µ 5 , defined on space-times of even dimension fail to be conserved, i.e., tend to be anomalous, when coupled to gauge fields – even if they are associated with massless matter fields. This phenomenon is called "chiral anomaly"; (see [9] and references given there). Let (cid:126)E denote the electric field and (cid:126)B the magnetic field (or -induction). In two space-time dimensions, the chiral anomaly is expressed by ∂µJ µ 5 = α 2π E, α := q2 , where q is the electric charge of a massless field that gives rise to the current 5 , and is Planck's constant, and by the (non-vanishing) anomalous equal- J µ time commutator 5 ((cid:126)y, t), J 0((cid:126)x, t)] = iαδ(cid:48)((cid:126)x − (cid:126)y). [J 0 In four space-time dimensions, the anomaly is expressed by ∂µJ µ 5 = and the anomalous commutator [J 0 5 ((cid:126)y, t), J 0((cid:126)x, t)] = i α π (cid:126)E · (cid:126)B, α π (cid:0) (cid:126)B((cid:126)y, t) · ∇(cid:126)y δ(cid:1)((cid:126)x − (cid:126)y). 3 For a massive matter field, there are additional terms contributing to the divergence, ∂µJ µ 5 , of the axial current that are proportional to the mass, m, of that field. It is then obvious that the chiral currents (cid:96) := J µ − J5 J µ µ, and J µ r := J µ + J5 µ are anomalous, too, (the signs of their divergence being opposite to one an- other). In this paper, we will not make use of anomalous commutators. But they play an important role in the study of transport in dissipationless one- dimensional and three-dimensional conductors; see [7, 8]. Faddeev and collaborators [10] contributed important insights to a clear un- derstanding of the mathematics underlying the chiral anomaly and anoma- lous commutators. Contents, Credits and Acknowledgements This paper is devoted to reviewing efforts to understand novel states of mat- ter that I have personally been involved in and that have spanned more than two decades. I think they have yielded a number of important and elegant results. Of course, numerous other people have contriubuted important in- sights towards clarifying issues discussed in this paper, and it is completely impossible to do justice to all their contributions in a short review like this one. Some of the results on the fractional quantum Hall effect described in Section 2 overlap with independent ones due to Xiao-Gang Wen [11].1 Ideas and results concerning 2D time-reversal invariant topological insulators re- lated to some of the ones described in Section 3 of this paper have also been brought forward by Shoucheng Zhang and collaborators and will be referred to later.2 In Section 4, some results on 3D time-reversal invariant topologi- cal insulators, including axionic insulators, are reviewed. These results, too, overlap with ones described by Shoucheng Zhang and collaborators. I am indebted to the following colleagues and friends for collaboration and/or many illuminating discussions on the subject matter covered in this review and for encouragement: 1I will not review work on the integer quantum Hall effect in 2D non-interacting electron gases, and I refrain from quoting papers on this effect, because there are too many of them, and they are well known. 2However, their work only appeared several years after the publication of the one of my collaborators and myself. 4 A. Alekseev, S. Bieri, A. Boyarsky, V. Cheianov, G. M. Graf, B. I. Halperin, T. Kerler, I. Levkivskyi, L. Molenkamp, G. Moore, B. Pedrini, O. Ruchayskiy, C. Schweigert, E. Sukhorukov, J. Walcher, Ph. Werner, P. Wiegmann, and A. Zee. Very special thanks are due to R. Morf, who introduced me to the quantum Hall effect and joined me for countless discussions, and to U. M. Studer and E. Thiran, without whose devoted help and collaboration my efforts to understand aspects of the fractional quantum Hall effect would have failed. 2 Theory of the Fractional Quantum Hall Effect: Anomalous Chiral Edge Currrents, Effective Actions, Classification of Hall Fluids Setup, basic quantities We consider a two-dimensional electron gas (2D EG) forming at the interface between an insulator and a semi-conductor when a gate voltage is applied; see Fig.1. The gas is confined to a domain Ω in the xy-plane. A uniform magnetic field (cid:126)B0 ⊥ Ω is applied to the system. The so-called filling factor ν := n e (cid:126)B0/h c , where n is the particle density of the 2D EG, e is the elementary electric charge and c is the speed of light, is chosen such that the longitudinal resis- tance, RL, of the gas vanishes. It is known, experimentally, that such filling factors exist, and these findings are supported by numerical simulations. But there are essentially no analytical results that enable us to understand why, at certain values of ν, the longitudinal resistance vanishes (or, put differently, why there is a strictly positive mobility gap above the ground- state energy). Here we will not consider this difficult problem, but see, e.g., [12].3 Rather, we assume that, at certain values of the filling factor, RL vanishes and then derive non-trivial, mathematically precise consequences of this assumption. 3I refer the reader to [13] and [14], and references given there, for much useful infor- mation on the quantum Hall effect. 5 Fig.1 – Caption: Schematic representation of the experimental setup and of exper- imental results on the QHE Observations: RL = 0 ↔ (ν, σH ) ∈ plateau; plateau heights ∈ e2 Q; the cleaner the sample, the more numerous are the observed plateaux and the narrower they are; if h Applications: Metrology; novel computer memories; topological quantum comput- ing (?) using quasi-particles with non-abelian braid statistics [15] e2 σH /∈ Z there appear to exist fractional electric charges. h We first study the response of the 2D EG to small perturbations in the external electromagnetic field, (cid:126)E(cid:107)Ω and (cid:126)B ⊥ Ω. We set (cid:126)Btot = (cid:126)B0 + (cid:126)B, B := (cid:126)B, E := (E1, E2). Note that, as long as electron spin is neglected, only the components, E, of the electric field parallel to the sample plane and the component, B, of the magnetic field perpendicular to the sample plane enter the laws of motion of electrons in the domain Ω. The field tensor of the electromagnetic field on the three-dimensional space-time of the 2D EG is then given by  0 F :=  . −E1 −E2 B E1 E2 0 −B 0 1-form, A =(cid:80) Aµdxµ, where x = (x0 ≡ t, x), x = (x1, x2) ∈ R2, and A is When interpreted as a 2-form, F is the exterior derivative, dA, of a the electromagnetic vector potential. 6 Electrodynamics of 2D incompressible electron gases The electric current density in the 2D EG is defined by jµ(x) = (cid:104)J µ(x)(cid:105)A, µ = 0, 1, 2, where J µ(x) is the quantum-mechanical vector current density – an operator- valued distribtuion – and (cid:104)(·)(cid:105)A is the state of the EG in the presence of an external electromagnetic field with vector potential A. We begin by summarizing the basic equations governing the electrodynam- ics in the bulk of a 2D EG, which, for E = 0 and B = 0, is assumed to be incompressible, i.e., RL = 0. (I) Hall's Law (assuming that RL = 0) j(x) = σH (E(x))∗, (1) where σH is the Hall conductivity, and E∗ is the vector arising from counter- clockwise rotation of E in the (xy-) plane of the sample through an angle of 90◦. This equation shows that space reflections in lines, P , and time rever- sal, T , are not symmetries of the gas, i.e., are broken, which is, of course, a consequence of the presence of the external magnetic field (cid:126)B0 ⊥ Ω. (II) Charge conservation ρ(x) + ∇ · j(x) = 0. ∂ ∂t (III) Faraday's induction law 3 + ∇ ∧ E(x) = 0. Btot ∂ ∂t It follows that ∂ρ ∂t (2) = −∇ · j (1) = −σH∇ ∧ E (3) = σH ∂B ∂t . (2) (3) (4) Integrating Eq. (4) in the time variable t, with integration constants chosen as follows j0(x) := ρ(x) + e · n, B(x) =(cid:0) (cid:126)Btot(x) − (cid:126)B0 (cid:1) · (cid:126)e3 , where (cid:126)e3 is the unit vector in the z-direction orthogonal to Ω, we find the 7 (IV) Chern-Simons Gauss law j0(x) = σH B(x) . Combination of laws (1) and (5) yields jµ(x) = 1 2 σH εµνλFνλ(x) (5) (6) (2) = ∂µjµ (3),(6) = εµνλ(∂µσH )Fνλ (cid:54)= 0, 1 2 Next, when combining laws (II), (III) and Eq. (6) we seemingly conclude that 0 (7) wherever σH (cid:54)= const., e.g., at ∂Ω. We seem to arrive at a contradiction. Well, actually, there isn't any contradiction! The point is that the current density jµ appearing in Eq. (6) is the bulk current density, (jµ bulk), which is not the conserved total electric current density, jµ tot: jµ tot = jµ bulk + jµ edge, with ∂µjµ tot = 0, (8) where jµ edge is the "edge current density", a current distribution with sup- port on all lines in Ω across which the value of σH jumps; in particular, supp(jµ edge) contains the boundary/edge, ∂Ω, of the gas. In general, only jµ tot is conserved, but ∂µjµ bulk , ∂µjµ edge (cid:54)= 0, as follows from Eqs. (7) and (8). Anomalous chiral edge currents The edge current density is defined in such a way that supp jµ edge = supp(∇σH ) ⊇ ∂Ω, ⊥ ∇σH . j edge One may view it as an instance of "holography" that, on supp(∇σH ), ∂µjµ edge (8) = −∂µjµ bulksupp(∇σH ) (6) = −σH E(cid:107)supp(∇σH ) , (9) where E(cid:107) is the component of the electric field (cid:126)Esupp(∇σH ) perpendicular to ∇σH . Eq. (9) is the chiral anomaly in 1 + 1 dimensions! 8 edge ≡ jµ The edge current, jµ (cid:96)/r, is an anomalous chiral current in 1 + 1 dimensions: At the edge of the sample, the Lorentz force acting on the electrons must be cancelled by the force confining the electrons to the interior of the domain Ω. Hence Btotv(cid:107) = (∇Vedge), e c where v(cid:107) is the propagation speed of an electron moving along ∂Ω, and Vedge is the potential of the force confining the electrons to the interior of Ω. The chiral nature of jµ egde becomes obvious from Figure 2. Fig.2 – Caption: Skipping orbits of electrons moving along the boundary/edge of a 2D electron gas confined to a disk – electrons near ∂Ω perform a chiral motion Classically, the orbits of electrons moving close to the edge ∂Ω of the sample are so-called "skipping orbits" which describe a chiral motion of electrons, the chirality being determined by the direction of (cid:126)B0. There is an analogous phenomenon in classical physics, namely in the theory of hurricanes: Under the replacements (cid:126)B → (cid:126)ωearth (angular velocity) , Lorentz force → Coriolis force, Vedge → air pressure, the theory of charge transport in a 2D EG close to the edge of the sam- ple is mapped to the theory of air-transport in a hurricane near the sur- face of the Earth. Actually, one may view a hurricane as an example of a "Hall effect" in a rotating gas. It is expected that a quantized Hall effect is encountered in a rapidly rotating layer of a superfluid Bose gas; (see last subsection of Sect. 1). Recall that the chiral anomaly in (1 + 1)D says that (for jµ edge = jµ (cid:96) ) ∂µjµ edge = − 1 h (9)⇒ σH = 1 h q2 α , (10) (cid:32)(cid:88) α (cid:33) q2 α E(cid:107) 9 (cid:88) α where qα = e(cid:101)Qα is the electric charge of an edge current corresponding to a counterclockwise-chiral edge mode α; (similar contributions come from clockwise modes, but with a reversed sign on the right side in Eq. (10)). Equation (10) yields a precise expression for the Hall conductivity σH in terms of charges of chiral currents circulating at the edge of Ω. The integer quantum Hall effect is observed if the charge quantum numbers (cid:101)Qα = ±1, for all α. The observation of a system in a fractional quantum Hall state then implies that there are currents carried by chiral modes with fractional charge circulating at the edge of the system. Apparently, starting from the basic equations of the electrodynamics of a 2D EG in an external electromagnetic field, we are led to rediscover Halperin's edge currents [16]; (see also [17]). h Edge- and bulk effective actions Eq. (10) shows that if σH /∈ e2 Z then there must exist fractionally charged modes (quasi-particles) propagating along supp(∇σH ) and, in particular, along ∂Ω. The quantum-mechanical chiral edge current density J µ edge turns out to generate a U (1)-current algebra that can be described in terms of free massless chiral Bose fields, ϕα. The Green functions of J µ edge can be derived from the anomalous chiral action, Γ∂Ω×R(A(cid:107)) in 1 + 1 dimensions – see Eq. (12), below – where A(cid:107) is the restriction of the vector potential A to the boundary, ∂Ω×R, of the space-time of the 2D EG. The action Γ∂Ω×R(A(cid:107)) is anomalous in the sense that it is not invariant under gauge transformations of A(cid:107). The anomaly of Γ∂Ω×R(A(cid:107)) is a consequence of the fact that the chiral edge current J µ edge is not conserved, i.e., is anomalous. The anomaly of the edge current must be cancelled by the anomaly of the bulk current J µ bulk; see Eq. (9). The generating function of the Green functions of the bulk current density is the bulk effective action, SΩ×R(A), which we will determine presently. This action is not invariant under gauge transformations of A either. Its variation under gauge transformations of A must be cancelled by the gauge variation of an appropriate multiple of the anomalous chiral action Γ∂Ω×R(A(cid:107)). Next, we derive the expression for the bulk effective action. By the definition of effective actions, the bulk current is given by the variational 10 derivative of SΩ×R(A) with respect to A. Thus, bulk(x) = (cid:104)J µ jµ bulk(x)(cid:105)A ≡ δSΩ×R(A) δAµ(x) (6) = 1 2 It follows that σH εµνλFνλ(x), x /∈ ∂Ω × R . (cid:90) SΩ×R(A) = σH 4 Ω×R A ∧ dA (11) (cid:21) (cid:20) (cid:90) 1 2 Note that the action SΩ×R(A) is not invariant under gauge transformations of A not vanishing at the boundary, ∂Ω×R, of the space-time of the sample. The total effective action is obtained by adding to the anomalous Chern- Simons action in Eq. (11) the anomalous chiral action Γ∂Ω×R(A(cid:107)) (see Eq. (12), below), multiplied by a coefficient chosen in such a way that the total action is gauge-invariant. In fact, the coefficient in front of Γ∂Ω×R(A(cid:107)) must be equal to σH ! This observation implies that the bulk definition of the Hall conductivity coincides with its edge definition; a conclusion that first appeared in [2, 4]. of a single chiral matter field on ∂Ω × R coupled to a vector potential a: Finally, we present an expression for the anomalous chiral action Γ∂Ω×R Γ∂Ω×R(a) := (12) where a = a+du+ + a−du− ≡ A(cid:107), and u+, u− are "light-cone coordinates" on the boundary ∂Ω × R of the space-time of the sample. ∂Ω×R a+a− − 2a± ∂2∓ (cid:3) a± d2u, Exercise: Check that the anomaly of the (bulk) Chern-Simons action SΩ×R(A) in (11), which is a boundary term, is cancelled by the one of σH Γ∂Ω×R(A(cid:107)). Whatever we have said about 2D electron gases in a homogeneous ex- ternal magnetic field exhibiting the quantum Hall effect can be extended to so-called Chern insulators [18] (see also [2]), which have the property that reflections in lines ("parity") and time reversal are broken even in the absence of an external magnetic field, (e.g., because of magnetic impurities in the bulk). An experimental realization of Haldane's model [18] has been described in [19]. The low-energy physics of quasi-particles in the bulk of a Chern insulator may resemble the one of two-component relativistic Dirac fermions coupled to the electromagnetic vector potential, with an effective action given by (11), with A = Atot, and σH = e2/2h (= Chern number of a certain vector bundle of Dirac fermion wave functions over the Brillouin zone T2). 11 Classification of "abelian" Hall fluids and -Chern insulators Next, I sketch a general classification [5] of 2D insulators with broken time reversal and parity in topologically protected states (i.e., groundstates of many-body Hamiltonians with a strictly positive mobility gap above the goundstate energy) exhibiting quasi-particles with abelian braid statistics [15]; ("non-abelian states" are discussed in [20]). Explicit results will be limited to Hall fractions in the interval (0, 1]. In the following, Jtot denotes the total electric current density (bulk + edge), which is conserved, i.e., ∂µJ µ tot = 0. We consider a general ansatz: M(cid:88) Jtot = QαJα, α=1 where the densities Jα are separately conserved current densities corre- sponding to different quasi-particle species, and the coefficients Qα ∈ R are "charges". (In the simple case of the integer quantum Hall effect for non-interacting electrons, α labels filled Landau levels of which there are M , and Qα = 1,∀α.) On a three-dimensional, simply connected space-time Λ = Ω × R, a conserved current density J can be derived from a vector potential: If J denotes the 2-form dual to J then conservation of J implies that dJ = 0, and hence J = dB, by Poincar´e's lemma, where the 1-form B is the vector potential of J and is determined by the current density up to the gradient of a scalar function β; i.e., B and B + dβ yield the same current density J. Chern-Simons effective action of conserved currents Henceforth we use units where e2 h = 1. For a 2D insulator (i.e., a state of matter protected by a mobility gap above the groundstate energy), the field theory of the conserved currents (Jα)M α=1 in the limit of very large distance scales and very low frequencies must be topological : If parity and time reversal are broken the "most relevant" contribution to the action of the currents Jα is the Chern-Simons action SΛ(B, A) := {Bα ∧ dBα + A ∧ QαdBα} + boundary terms, (13) (cid:90) M(cid:88) α=1 Λ 12 where A is the electromagnetic vector potential, and the boundary terms must be added to cancel the anomalies of the Chern-Simons terms (1st term on right side) in (13) under the gauge transformations B → B + dβ and A → A + dχ. Carrying out the Gaussian functional integrals over the fields(cid:0)Bα (cid:19) (cid:90) we find that (cid:1)M α=1, (cid:90) M(cid:89) DBα = exp A ∧ dA + σH ΓΛ(A(cid:107)) , exp(iSΛ(B, A)) α=1 where σH 4 i (cid:18) M(cid:88) Λ σH = Q2 α (14) (15) α=1 Physical excitations (states) of a topological Chern-Simons theory are labelled by networks of Wilson lines contained in the half-space Λ− = Ω×R− corresponding to negative values of the time variable and ending in points of the spatial surface Ω × {0}. Scalar products of physical states of the Chern- Simons theory with action given by (13) can be calculated by inserting into the Gaussian functional integral on the left side of Eq. (14) networks of Wilson lines contained in Λ− hooked up to reflected networks of Wilson lines, contained in Λ+ = Ω×R+, at common intersection points in Ω×{0}, so as to yield a network of gauge-invariant Wilson networks/loops. The Hamiltonian of Chern-Simons theory vanishes; bulk excitations are static, reflecting the feature that there is a positive mobility gap above the groundstate energy. Classification of 2D "abelian" topological insulators with broken time reversal – bulk degrees of freedom The operator measuring the electric charge of an excitation of the Chern- Simons theory with action given by (13) contained in a region O of space, Ω × {0}, is given by (cid:90) O (cid:90) M(cid:88) α=1 Qα Bα . ∂O QO = J 0(x)d2x = The electric charge deposited inside the domain O by excitations corre- sponding to a network of Wilson lines can be inferred by inserting, besides 13 the Wilson network (hooked up to its reflection at Ω × {0}), a Wilson loop (cid:90) exp(cid:0)i t M(cid:88) α=1 (cid:1), Qα Bα ∂O t ∈ R, into the functional integral; (see Eq. (14)). If a network of Wilson lines creates an excitation describing n electrons or holes located inside O from the ground state of the insulator then its electric charge, as measured by the operator QO, is given by −n + 2k, k = 1, . . . , n, where k is the number of holes. If n is odd this excitation must have Fermi-Dirac statistics, if n is even it must have Bose-Einstein statistics. This relation between the electric charge of an excitation and its statistics implies that the M -tuples of quantum numbers assigned to lines in Wilson networks creating multi- electron-hole excitations must be sites of an odd-integral lattice, Γ, of rank M , and that the "co-vector" Q := (Q1, . . . , QM ) must belong to the dual lattice Γ∗, (in fact, Q must be a so-called "visible vector" in Γ∗); see [5, 6]. Hence (cid:88) σH = α ∈ Q , Q2 α=1,...,M i.e., the Hall conductivity σH is a rational multiple of e2 h . It takes some more effort to show that the so-called Witt sublattice of Γ is an A-, D- or E- root lattice; see [5, 6]. Classification of 2D "abelian" topological insulators with broken time reversal – edge degrees of freedom The chiral anomaly in the form of Eq. (10) reflects the feature that there are several (namely N ) species of gapless quasi-particles propagating either clockwise or anti-clockwise along the edge ∂Ω of the 2D EG. The dynamics of these quasi-particles is described by N massless chiral scalar Bose fields {ϕα}N α=1, in α=1 is diagonal ). We then find field space such that the velocity matrix(cid:0)vα α=1; (we choose a basis,(cid:0)ϕα(cid:1)N (cid:1)N α=1 with propagation speeds {vα}N that: 1. The chiral electric edge current operator and the Hall conductivity are given by J µ edge = e N(cid:88) α=1 (cid:101)Qα∂µϕα, (cid:101)Q = ((cid:101)Q1, . . . ,(cid:101)QN ), (cid:101)Q · (cid:101)QT , e2 h and σH = 14 respectively. 2. Multi-electron/hole states localized along the edge of the 2D EG are created by vertex operators (cid:32) N(cid:88) (cid:33)  ∈(cid:101)Γ,  q1 ... qN : exp i qαϕα : , q = (16) where the double colons indicate Wick ordering, and (cid:101)Γ is a lattice. α=1 The electric charge carried by the vertex operator in (16) is given by e(cid:101)Q · q . By the same reasoning process as above, the relation between the charge of an edge excitation created by the vertex operator in (16) and its statistics implies that the "quantum numbers" q must correspond to sites in an odd-integral lattice, (cid:101)Γ, of rank N . The Witt sublattice of (cid:101)Γ is an A-, D- or E- root lattice and gives rise to an A-, D- or E- Kac-Moody current algebra at level 1 acting as a quantum symmetry on the edge states of the 2D electron gas. The matrix q∗ :=(cid:0)q∗ k (cid:1) α α,k =1,...,N where, for all k, the vector q∗ k determines a vertex operator creating a one-electron state propagating along the edge of the 2D EG, is a close analogue of the Cabibbo-Kobayashi-Maskawa matrix of the weak sector of the Standard Model. We conclude that: 3. The classifying data for the edge degrees of freedom of a 2D EG ex- hibiting the quantum Hall effect are given by {(cid:101)Γ; (cid:101)Q ∈(cid:101)Γ∗("visible"); q∗; v = (vα)N α=1}. (17) The charged quasi-particles predicted by these data exhibit abelian braid statistics and transform under A-, D- or E- Kac-Moody symme- tries at level 1. For "ideal" Hall insulators, one expects that (cid:101)Γ = Γ and (cid:101)Q = Q; (see lattices Γ and (cid:101)Γ need not be identical. This can happen when the edge previous subsection on bulk degrees of freedom). But, in general, the two 15 of a 2D incompressible EG has a layered structure. However, in any case, the constraint that the edge conductivity e2 Hall conductivity e2 invariance. h (cid:101)Q · (cid:101)QT is equal to the bulk h Q · QT must always be satisfied, as required by gauge A (partial) classification of 2D electron gases with quasi-particles ex- hibiting non-abelian braid statistics and quantized Hall conductivity has been presented in [20]. It is an interesting, partly open problem to classify egde degrees of free- dom localized on interfacial lines at which two different electron gases, with non-vanishing Hall conductivities σ(1) H (cid:54)= 0, join. H (cid:54)= σ(2) Success of classification, comparison with data In [5, 6], we have given a list of "physically plausible" odd-integral lattices Γ and visible vectors Q ∈ Γ∗ with the property that 0 < ( e2 write h )−1σH ≤ 1. We where nH and dH are relatively prime integers, with nH ≤ dH . Plotting dH h )−1σH horizontally (to the right), we have vertically (downwards) and ( e2 obtained the table displayed in Fig. 3, below, comparing data on 2D elec- tron gases exhibiting the quantum Hall effect observed in laboratories with our theoretical predictions. This figure is taken from a plenary lecture I presented at the 1994 International Congress of Mathematicians in Zurich [21].4 ( e2 h )−1σH = nH dH , 4On this occasion, I was kindly introduced to the audience by Ludvig Faddeev. 16 Fig.3 – Caption: Observed Hall fractions σH = nH /dH in the interval 0 < σH ≤ 1 and their experimental status in single-layer 2D electron gases exhibiting the quan- tum Hall effect Explanation of Figure 3: Experimentally well established Hall fractions are indicated by a bullet "•". These are fractions for which the longitudinal resistance, RL, has a minimum very close to 0 and RH shows a plateau. Fractions for which RL has a minimum and RH shows an inflection are indicated by a circle "◦". If there are only weak experimental indications for a Hall insulator state or if the data are controversial the symbol "·" is used. All bullets and circles in this figure, and more (!), are predicted theoretically; see [5, 6]. Hall fractions at which a magnetic field-(B) and/or density-driven transition from one insulator state to another one (at a constant value of σH ) has been ob- served are marked by the letters "B/n − p", "B − p". In the interval [ 1 1 ∪ 3 , 1] = Σ+ Σ− 1 , the agreement between our theory, namely the classification of 2D "abelian" Hall insulator states, as presented above, and experimentally observed Hall frac- tions, including the prediction of transitions at a constant value of σH , is nearly perfect. (Only the fractions 8 17 are somewhat uncertain.) Furthermore, there are (abelian and non-abelian) states at σH = 1 2 in double-layer electron gases and at σH = 1; see [20]. Predictions of plateaux in the intervals Σ+ p , p = 2, 3, ..., are related to those in the interval Σ+ 1 by the so-called "shift map"; see [6, 20]. For a detailed discussion of the shift map and of the Kac-Moody symmetries of Hall insulators alluded to above, with σH < 1 (as indicated in Fig. 3), see [5, 6, 20]. p ∪ Σ− 13 and 10 4 , 1 1 ∪ Σ− 17 Rotating Bose gases, and a duality between pho- tonics and electronics We conclude our discussion of Hall- and Chern insulators with some remarks. Remark 1: A discussion of the relevance of the gravitational anomaly and of some new invariants for the theory of the quantum Hall effect (in a spirit somewhat related to the one of this review) can be found in recent papers by P. Wiegmann and collaborators [22]. Remark 2: A very rapidly rotating 2D layer of a superfluid Bose gas of atoms (with "hard-core repulsion" between atoms) may settle in a state analogous to a Hall insulator. The centrifugal force acting on the bosonic atoms in the gas may be precisely cancelled by a harmonic potential trapping the gas, and the Coriolis force then plays the role of the Lorentz force acting on the electrons in a 2D EG. The energy levels of a bosonic atom precisely correspond to the Landau levels of an electron in a 2D EG. This effect has first been sketched in [4, 6] and discussed in considerably more detail in [23] – among other papers. The basic formalism can be found, e.g., in [24]. "dual" to the vector potential B (:=(cid:80) Remark 3: In two space dimensions, the electromagnetic vector potential A is α QαBα) of the conserved electromagnetic current density J; (see Eqs. (13) and (14)). Under the replacements A (cid:55)→ B, B (cid:55)→ A, conventional 2D insulators with time-reversal symmetry are mapped to 2D su- perconductors, and conversely, as discussed in [6]. Furthermore, electronic Hall insulators are mapped to 2D gapped photonic wave guides exhibiting extended elec- tromagnetic surface waves. Assuming that time reversal is not a symmetry of the system, the leading terms in the action of the electromagnetic field in such a wave guide are given by {E(x) · εE(x) − µ−1B2(x)}d3x A ∧ dA + σH ΓΛ(A(cid:107)), (18) SW G Λ (A) = (cid:90) (cid:90) [25]. The Chern-Simons term, ∝ (cid:82) 1 2e2 σH 4 + Λ Λ where ε is the dielectric tensor of the wave guide and µ is its magnetic permeability. This is the action functional of massive QED in 2 + 1 space-time dimensions; see from coupling the electromagnetic vector potential A to the electric current density J = dB, as in Λ A ∧ dA, on the right side of Eq. (18) arises (cid:90) A ∧ dB, Λ assuming that the action functional for the vector potential B of the current density is given, in the simplest instance, by SΛ(B) = 1 σH B ∧ dB + boundary action + l.r. terms, (cid:90) Λ 18 as appropriate for an insulator with broken time reversal, (see Eq. (13), with N = 1); the "boundary action" (see (12)) is there to restore gauge invariance, i.e., invariance under B (cid:55)→ B+dβ, of SΛ(B), and "l.r. terms" stands for "less (infrared-) relevant terms" in the action functional. We then find the response equations (cid:104)Fµν(x)(cid:105)B = εµνρ = σ−1 δSΛ(B) δBρ(x) H εµνρjρ(x) , (19) with εµνρjρ = (dB)µν. Using arguments similar to those presented in connection with Eqs. (7) and (9), we find that quantized electromagnetic waves can propagate chirally and essentially without dispersion along the edge of such a wave guide. 3 Chiral Spin Currents in Planar Time-Reversal Invariant Topological Insulators So far, we have not paid attention to electron spin, although there are actually 2D EG exhibiting the fractional quantum Hall effect where electron spin must be taken into account; see [6]. But it would take us too far off the main line of this paper to discuss this issue here. In the theory of 2D time-reversal invariant topological insulators, which we will sketch next, electron spin undoubtedly plays a crucial role! In order to prepare the ground for an account of this theory, we briefly review the Pauli Equation for a non-relativistic spinning electron in an arbitrary external electromagnetic field: where m is the (effective) mass of an electron, (cid:126)D · (cid:126)D Ψt, (20) iD0Ψt = − 2 (cid:33) (cid:32) 2m ↑ t ((cid:126)x) ↓ t ((cid:126)x) ψ ψ Ψt((cid:126)x) = ∈ L2(R3) ⊗ C2 is a two-component Pauli spinor describing the state of an electron located at the point (cid:126)x at time t, and the operators D0 and (cid:126)D = (D1, D2, D3) are covariant derivatives given by iD0 = i∂t + eϕ − (cid:124) (cid:123)(cid:122) (cid:125) (cid:126)W0 · (cid:126)σ Zeeman coupling (cid:126)W0 = µc−1 (cid:126)B + , 4 (cid:126)∇ ∧ (cid:126)V , (21) where ϕ is the electrostatic potential, (cid:126)σ = (σ1, σ2, σ3) is the vector of Pauli matrices, µ is the Bohr magneton, (cid:126)B is the magnetic field and (cid:126)V is the velocity field generating an incompressible (i.e., (cid:126)∇ · (cid:126)V = 0) classical motion/flow of the system (relative to the inertial laboratory frame). Furthermore, 19 i Dk = i ∂k + eAk − mVk − (cid:126)Wk · (cid:126)σ , (22) k = 1, 2, 3, where (cid:126)A = (A1, A2, A3) is the electromagnetic vector potential, and (cid:126)Wk is given by (cid:126)Wk · (cid:126)σ := −µ (cid:126)E + (cid:20)(cid:18) (cid:124) (cid:19) (cid:21) ∧ (cid:126)σ (cid:126)V c2 (cid:123)(cid:122) with µ = µ + e 4mc , (23) , (cid:125) k spin-orbit interactions describing spin-orbit interactions and Thomas precession.5 Note that the 2 × 2 matrices (cid:126)W0 · (cid:126)σ and (cid:126)Wk · (cid:126)σ, k = 1, 2, 3, determine an SU (2)-gauge field/connection on the bundle of two-component Pauli spinors over space-time, and that the Pauli equation (20) obeys U (1)em × SU (2)spin - gauge invariance. For a detailed discussion of the Pauli equation and its gauge invariance, as well as plenty of physical consequences thereof, see [4, 6]. Effective action of a 2D time-reversal invariant topo- logical insulator µ · σ3, with W K µ ≡ 0, K = 1, 2, for µ = 0, 1, 2. Next, we consider a 2D gas of interacting electrons confined to a domain Ω in the xy-plane, assuming that (cid:126)B ⊥ Ω and (cid:126)E, (cid:126)V (cid:107)Ω. Then the SU (2)- connection, (cid:126)Wµ, is given by W 3 Thus the connection determining parallel transport of the component ψ↑ of a Pauli spinor Ψ is given by a + w, while parallel transport of ψ↓ is determined by a − w, with a0 = −eϕ, ak = −eAk + mVk, and wµ = W 3 µ . These connections are abelian, i.e., just affect the phases of ψ↑ and ψ↓, respectively. Under time reversal, the components of these connections transform as follows: a0 → a0, ak → −ak, but w0 → −w0, wk → wk. (24) The term in the effective action of a 2D insulator with the smallest scaling dimension, i.e., the term dominating the large-scale physics, is a Chern-Simons term. If there were only the gauge field a, with w ≡ 0, or only the gauge field w, with a ≡ 0, a Chern-Simons term would not be invariant under time reversal, and the dominant term in the effective action would be given by d3x{E · εE − µ−1B2}, (25) (cid:90) S(a) = Ω×R with ε the dielectric tensor and µ the magnetic permeability. This is the effective action of a conventional time-reversal invariant insulator. But, in the presence of 5One should actually also include the "spin connection" of the sample geometry in the definition of the SU (2)-gauge field (cid:126)Wµ; see [4]. 20 electron spin and hence of two gauge fields, a and w, satisfying (24), the dominant term is a time-reversal invariant combination of two Chern-Simons terms: (cid:90) (cid:90) S(a, w) = = σ 4 σ 2 {(a + w) ∧ d(a + w) − (a − w) ∧ d(a − w)} {a ∧ dw + w ∧ da} (26) This action reproduces an action similar to the one in (25) if the connection w is chosen as in (21) and (22). However, the action (26) describes the response of a topological insulator for an arbitrary choice of the gauge field wµ = W 3 µ . The gauge fields a and w transform independently under gauge transformations, and the Chern-Simons action S(a, w) in Eq. (26) is anomalous under gauge transformations of a ± w not vanishing at the boundary, ∂Λ, of a 2D sample with space-time given by Λ = Ω × R. Edge degrees of freedom: Spin currents The anomalous chiral boundary action σ{Γ((a + w)(cid:107)) − Γ((a − w)(cid:107))} , (27) cancels the anomaly of the bulk action S(a, w) in (26). The two terms in the boundary action are the generating functionals of connected Green functions of two counter-propagating chiral electric edge currents, one carried by electrons with "spin-up", the other one carried by electrons with "spin-down", generating two U (1)- current algebras. Thus, whenever the edge degrees of freedom support a current then a net chiral edge spin-current, s3 edge, circulates at the boundary of the sample. Bulk response equations of a 2D time-reversal in- variant topological insulator From the Chern-Simons action S(a, w) in Eq. (26) one easily derives the following response equations in the bulk of a 2D topological insulator with unbroken time- reversal invariance: j(x) = σ(∇B)∗, and sµ 3 (x) = δS(a, w) δwµ(x) = σεµνλFνλ(x), (28) where j is the electric current density in the bulk of the insulator, and sµ 3 is the bulk spin-current density, (more precisely its 3-component in spin space); compare 21 to Eqs. (1) and (6). By arguments analogous to those in Eqs. (7) through (9), we see that Eq. (28) implies that the spin current sµ 3 is anomalous and that its anomaly is cancelled by a chiral egde spin-current. The Chern-Simons effective action in (26), the response equations in (28), chiral spin currents and the role of spin-orbit interactions in connection with the presence of a non-trivial gauge field w have first been described, among many other effects, in [4], in 1993; see also [6]. Quasi-particles in 2D time-reversal invariant topo- logical insulators We should ask what kinds of quasi-particles may induce the effective Chern-Simons actions (a ± w) ∧ d(a ± w) ± σΓ(cid:0)(a ± w)(cid:107)(cid:1) , S±(a ± w) = ± σ 4 see Eq. (26), where "+" stands for "spin-up" and "−" stands for "spin-down". It is well known that a two-component relativistic Dirac fermion with mass M > 0 (M < 0) breaks time-reversal invariance and – when coupled to an abelian gauge field, A – induces a Chern-Simons term (cid:90) (cid:90) 1 4π + (−) A ∧ dA in the effective action; see [25]. Thus, a 2D time-reversal invariant topological insulator with chiral edge spin-currents might exhibit two species of charged quasi- particles in the bulk, with one species (spin-up) related to the other one (spin-down) by time reversal, each species having two degenerate states per wave vector and mimicking, at small wave vectors, a 2-component Dirac fermion. The value of the constant σ is then determined! Experiments, other approaches, chiral spin liquids Materials with properties similar to those predicted by the theory described above have been studied experimentally in the group of L. Molenkamp in Wurzburg [26]. More detailed theoretical aspects have been discussed in [28, 29]. However, because of external electromagnetic fields violating the assumption that (cid:126)B ⊥ Ω and (cid:126)E(cid:107)Ω, and because of magnetic impurities in the material, the property that (cid:126)Wµ = (0, 0, W 3 3 is not conserved; (although (cid:126)s µ is always "covariantly conserved", [4, 6]). There then exist spin-umklapp processes that cause transitions between the two chiralities of edge currents and hide the quantization of the value of σ. For these reasons, one cannot expect that real materials exactly reproduce the theoretical predictions of the theory sketched above. µ ) will usually not hold exactly, and hence the spin current sµ 22 One may, however, expect that there are materials with edge currents, j(cid:96) and jr, of opposite chirality related to each other by time reversal that do not mix. The current j(cid:96) may couple to a chiral abelian gauge field a(cid:96) and the current jr to a gauge field ar. The effective actions of a(cid:96) and ar are given by (±1)× the anomalous action in (12). The two gauge fields may have the form a(cid:96)/r = a ± w, where w is a Berry connection. The transformation of a and w under time reversal is the one given in (24). The theory sketched above provides a description of such materials. This theory has the advantageous feature that it is not based on a single-particle picture. More recently, Z2 - topological invariants based on a single-particle picture and characterizing non-interacting time-reversal invariant topological insulators have been studied by Kane and Mele and numerous further authors; see, e.g., [29] and references given there. In [4, 6], various further 2D systems with somewhat exotic magnetic properties have been described. For example, it has been speculated (see also [30]) that there may be chiral spin liquids with broken time-reversal symmetry not exhibiting any long-range spin order. The response laws for the spin current, (cid:126)s µ, of chiral spin liquids can be derived from a non-abelian Chern-Simons action, (cid:90) k 4π tr(cid:0)w ∧ dw + w ∧ w ∧ w(cid:1) + bd. terms, 2 3 where k is an integer, and w = (cid:80) wµdxµ is an SU (2)-gauge field coupling to Λ electron spin, with wµ = (cid:126)Wµ · (cid:126)σ. 4 3D Topological Insulators, Axions Encouraged by the findings of the last section, we propose to consider insulators in three dimensions confined to a region Λ := Ω × R, Ω ⊂ R3, ∂Λ (cid:54)= ∅ of space-time. We are interested in the general form of the effective action describing the response of such materials to turning on an external electromagnetic field. Until the mid nineties, it was expected that the effective action of a 3D insulator is always given by SΛ(A) = dt d3x{ (cid:126)E · ε (cid:126)E − (cid:126)B · µ−1 (cid:126)B}, (29) (cid:90) 1 2 Λ where A is the electromagnetic vector potential of the electromagnetic field ( (cid:126)E, (cid:126)B), ε is the dielectric tensor of the material and µ is the magnetic permeability tensor. In (3 + 1) dimensions, the action in (29) is dimensionless. In the seventies, particle theorists have taught us that one could add another dimensionless term to the action in (29), the θ-term first introduced in QCD. 23 An effective action with a topological term We replace the action SΛ(A) by SΛ(A) → S(θ) Λ (A) := SΛ(A) + θ IΛ(A), θ ∈ [0, 2π) , (30) where θ is the analogue of the "vacuum angle" in particle physcis, and IΛ is a topological term ("instanton number") given by (cid:90) (cid:90) (cid:90) (4π2)IΛ(A) = 2 dt d3x (cid:126)E · (cid:126)B = FA ∧ FA = Stokes Λ Λ ∂Λ A ∧ dA (31) see [6, 31, 30]. The partition function of a topological insulator in three dimensions whose effective action is given by (30) and (31) is Λ (A) = exp(cid:0)iS(θ) Λ (A)(cid:1). Z (θ) In the thermodynamic limit, Ω (cid:37) R3, Z (θ) invariant under time reversal iff Λ (A) is periodic in θ with period 2π and θ = 0, π . Surface degrees of freedom A conventional insulator corresponds to θ = 0. For θ = π, the partition function Z (θ) Λ (A) has a boundary term given by (cid:18) i (cid:90) exp 4π ∂Λ A ∧ dA (cid:19) , (32) see (31). As mentioned in the last section, this is the leading term in the parti- tion function of (2 + 1)-dimensional charged, "relativistic", two-component Dirac fermions on ∂Λ coupled to an external electromagnetic field A∂Λ. Two species of charged spin-polarized quasi-particles with a conical dispersion law mimicking a two-component relativistic Dirac fermion and propagating along the surface of the system may be encountered in certain 3D topological insulators with two de- generate bands "communicating" with each other and an effective action given by (30), (31), with θ = π. Such systems support surface Hall currents, as is evident from the form of the surface action in (31) and (32); cf. Eq. (11). They have been studied experimentally in [27]; see also [31]. Axionic topological insulators One may wonder whether, following ideas of Peccei and Quinn [32], it might make sense to view the angle θ as the (groundstate) expectation value of a dynamical 24 (cid:90) Λ field, ϕ, and replace the term θ IΛ(A) by IΛ(ϕ, A) := 1 4π2 ϕ FA ∧ FA + S0(ϕ), (33) where S0(ϕ) is invariant under constant shifts ϕ (cid:55)→ ϕ + nπ, n ∈ Z, and ϕ is a pseudo-scalar field called "axion field", (see [8]). The effective action given in (33) then gives rise to Halperin's 3D Hall effect: j = − 1 2π2 dϕ ∧ FA; in particular (cid:126)j = (cid:126)∇ϕ × (cid:126)E 1 2π2 (34) In a crystalline insulator, and for an axion linear in (cid:126)x, (cid:126)∇ϕ = cst. (cid:126)K, where, by invariance under lattice translations, (cid:126)K must belong to the dual lattice. Hence (cid:126)∇ϕ is "quantized".6 It is argued that dynamical axions may emerge in certain topological insulators with anti-ferromagnetic short-range order and two bands of electron states "communicating" with each other. The time derivative of ϕ then has the interpretation of a chemical potential conjugate to the axial charge: The chiral anomaly tells us that (cid:90) (cid:90) dϕ ∧ j5 = −cst. ϕFA ∧ FA + terms ∝ masses (35) The action of Eq. (33) has been proposed in [8] as the effective action describing a higher-dimensional cousin of the Hall effect and has been applied to the problem of unravelling possible origins of intergalactic primordial magnetic fields in the Universe; see also [7]. Later it appeared in [34, 35]. Mott transition and instabilities in axionic topolog- ical insulators It is interesting to analyze physical effects observable in putative states of matter described by the effective action Stot Λ (A) := SΛ(A) + IΛ(ϕ, A), where SΛ(A) is as in (29) and IΛ(ϕ, A) as in (33). Since the action S0(ϕ) on the right side of Eq. (33) has been assumed to be periodic under constant shifts ϕ (cid:55)→ ϕ + nπ, n ∈ Z, one may expect that, at positive (but low) temperatures, the bulk of such a state of matter will exhibit axion domain walls across which the mean value of ϕ changes by an integer multiple of π. Recalling the insight described after (32), one predicts that gapless, charged two-component Dirac fermions propagate along domain walls across which the mean value of ϕ changes by an odd multiple of π. We are thus led to predict that the formation of macroscopic axion domain walls 6I thank Gregory Moore for having explained this point to me; see [33] 25 gives rise to a "Mott transition" from an insulator to a state with non-vanishing bulk conductivity, due to the extended charged surface modes propagating along such domain walls; (see [30]). Axion-electrodynamics exhibits some tantalizing instabilities. One such insta- bility, triggered by the term IΛ(ϕ, A), manifests itself in the growth of long-wave- length magnetic fields in the bulk of the system; see [7, 8]. A related instability has been described in [36]: Above a certain critical field strength, Ec, a constant exter- nal electric field, E(cid:126)n (with (cid:126)n = 1 and E > Ec), applied to an axionic topological insulator is screened, with the excess field, (E − Ec)(cid:126)n, converted into a magnetic field. Expository articles covering theoretical and experimental aspects of topological insulators can be found, e.g., in [37, 38]. Conclusions and outlook • The Physics of 2D (and 3D) insulators is surprisingly rich and has potential to lead to promising technological applications. Interesting mathematical techniques – ranging from abstract algebra over the topology of fibre bun- dles all the way to hard analysis – have non-trivial applications to problems concerning these fascinating states of matter. • 2D electron gases, Bose gases and magnetic materials represent stimulat- ing play grounds for experimentalists and theorists alike – not least because very general principles of quantum theory, such as braid statistics, fractional spin & fractional electric charges, anomalies and their cancellation, current algebra and holography appear to manifest themselves in the arena of 2D condensed matter physics. Unfamiliar species of quasi-particles, in particular two-component Dirac-like fermions and Majorana fermions are encountered in the physics of various exotic two- and three-dimensional systems whose exploration has begun only recently. • It is interesting to consider higher-dimensional cousins of the QHE and of time-reversal invariant topological insulators; see [8, 39]. Some of them might be relevant in cosmology, e.g., in connection with the generation of intergalac- tic primordial magnetic fields in the Universe. Ideas on cosmology somewhat related to those reviewed in this paper are presently actively pursued. Many ideas originally developed in particle physics turn out to have fruitful applications in condensed matter physics. Likewise, ideas originating in condensed matter physics and statistical mechanics promise to have very interesting applica- tions in cosmology. But that's another story that I cannot go into here. – In any event, these fields of physics appear to feature highly stimulating exchanges and fruitful interactions, or call for such. Moreover, different mathematical techniques and styles of research have turned out to be useful and to complement each other in productive ways in particle physics, condensed matter physics and cosmology. 26 New insights gained in pure mathematics sometimes have spectacular applications in physics, and conversely. Yet, disappointingly, interactions across certain disci- plinary boundaries apparently continue to be weaker than they ought to be, and people in different communities are not seldom reluctant to profit from each other's experiences. A determination to change this belongs to the legacy Ludvig Faddeev has left us. References [1] J. Frohlich, P. A. Marchetti, Quantum Field Theory of Anyons, Letters in Math. Phys. 16, 347-358 (1988) [2] J. Frohlich, T. Kerler, Universality in Quantum Hall Systems, Nucl. Phys. B 354, 369-417 (1991) [3] J. Frohlich, A. Zee, Large-Scale Physics of the Quantum Hall Fluids, Nucl. Phys. B 364, 517-540 (1991) [4] J. Frohlich, U. M. Studer, Gauge Invariance and Current Algebra in Non- Relativistic Many-Body Theory, Rev. Mod. Phys. 65, 733-802 (1993) [5] J. Frohlich, E. Thiran, Integral Quadratic Forms, Kac-Moody Algebras, and Fractional Quantum Hall Effect. An ADE – O Classification, J. Stat. Phys. 76, 209-283 (1994) [6] J. Frohlich, U. M. Studer, E. Thiran, Quantum Theory of Large Systems of Non- Relativistic Matter (Course 8, Part I), F. David, P. Ginsparg and J. Zinn-Justin, eds., Les Houches, Session LXII, 1994 Elsevier Science B.V., Amsterdam, 1996 [7] A. Yu. Alekseev, V. V. Cheianov, J. Frohlich, Universality of Transport Prop- erties in Equilibrium, the Goldstone Theorem, and Chiral Anomaly, Phys. Rev. Letters 81, 3503-3506 (1998) [8] J. Frohlich, B. Pedrini, New Applications of the Chiral Anomaly, in: Mathe- matical Physics 2000, A. Fokas, A. Grigoryan, T. Kibble, B. Zegarlinski (eds.), Imperial College Press, London and Singapore, 2000 A. Boyarsky, J. Frohlich, O. Ruchayskiy, Self-consistent evolution of magnetic fields and chiral asymmetry in the early Universe, Phys. Rev. Lett. 108, 031301 (2012) [9] R. Jackiw, D. Gross, S. B. Treiman, Current Algebra and Anomalies, Princeton University Press, Princeton NJ, 1986 [10] J. S. Bell, R. Jackiw, A PCAC Puzzle: π0 → yy in the σ-Model, Nuovo Cimento 60A, 47 (1969) S. L. Adler, Axial-Vector Vertex in Spinor Electrodynamics, Phys. Rev. 177, 2426 (1969) L. D. Faddeev, Operator anomaly for the Gauss law, Physics Letters B 145, 27 81-84 (1984) A. G. Reiman, M. A. Semenov-Tyan-Shanskii, L. D. Faddeev, Quantum anoma- lies and cocycles on gauge groups, Funct. Anal. Appl. 18, 319-326 (1984) L. D. Faddeev, S. L. Shatashvili, Algebraic and Hamiltonian methods in the theory of non-Abelian anomalies, Theor. Math. Physics 60, 770-778 (1984) J. Lott, A Proof of the Axial Anomaly, Commun. Math. Phys. 97, 371-379 (1985) A. Yu. Alekseev, Ya. Madaichik, L. D. Faddeev, S. L. Shatashvili, Derivation of anomalous commutator in the functional integral formalism, Theor. Math. Physics 73, 1149-1151 (1987) [11] X. G. Wen, Chiral Luttinger liquid and the edge excitations in the fractional quantum Hall states, Phys. Rev. B 41, 12838 (1990) B. Blok, X. G. Wen, Effective theories of the fractional quantum Hall effect at generic filling fractions, Phys. Rev. B 42, 8133 (1990) [12] "High Magnetic Fields in Semiconductor Physics III – Quantum Hall Effect, Transport and Optics", Springer Series in Solid-State Sciences vol. 101, G. Landwehr (ed.), Springer Verlag, Berlin, Heidelberg, 1992; (see, in particular, the contribution of R. Morf, and references given therein) [13] "The Quantum Hall Effect", R. E. Prange, S. M. Girvin (eds.), 2nd edition, Graduate Texts in Contemporary Physics, Springer-Verlag, New York, 1990 [14] B. I. Halperin, Theory of the quantized Hall conductance, Helv. Phys. Acta 56, 75-102 (1983) [15] J. Frohlich, Statistics of Fields, the Yang-Baxter Equation, and the Theory of Knots and Links, in: "Non-Perturbative Quantum Field Theory", G. 'tHooft et al. (eds.), Carg`ese 1987, Plenum Press, New York, 1988 [16] B. I. Halperin, Quantized Hall conductance, current-carrying edge states, and the existence of extended states in a two-dimensional disordered potential Phys. Rev. B 25 , 2185-2190 (1982) [17] J. Frohlich, G. M. Graf, J. Walcher, On the Extended Nature of Edge States of Quantum Hall Hamiltonians, Annales Henri Poincar´e 1, 405-442 (2000) [18] F. D. M. Haldane, Model for a Quantum Hall Effect without Landau Levels: Condensed-Matter Realization of the "Parity Anomaly", Phys. Rev. Lett. 61, 2015-2018 (1988) [19] G. Jotzu, M. Messer, R. Desbuquois, M. Lebrat, T. Uehlinger, D. Greif, T. Esslinger, Experimental realization of the topological Haldane model with ul- tracold fermions, Nature 515, 237-240 (2014) [20] J. Frohlich, B. Pedrini, C. Schweigert, J. Walcher, Universality in Quantum Hall Systems: Coset Construction of Incompressible States, J. Stat. Phys. 103, 527-567 (2001) 28 [21] J. Frohlich (RG in M∪Φ), The Fractional Quantum Hall Effect, Chern-Simons Theory, and Integral Lattices, Proceedings of ICM '94, S. D. Chatterji (ed.), Birkhauser Verlag, Basel, Boston, Berlin, 1995 [22] T. Can, M. Laskin, P. Wiegmann, Fractional Quantum Hall Efffect in a Curved Space: Gravitational Anomaly and Electromagnetic Response, Phys. Rev. Lett. 113, 046803 (2014) S. Klevtsov, P. Wiegmann, Geometric Adiabatic Transport in Quantum Hall States, Phys. Rev. Lett. 115, 086801 (2015) [23] N. R. Cooper, N.K. Wilkin, J.M.F. Gunn, Quantum phases of vortices in rotating Bose-Einstein condensates, Phys. Rev. Lett. 87, 120405 (2001) N. R. Cooper, Rapidly Rotating Atomic Gases, Adv. in Physics 57, 539-616 (2008) [24] E.Schmutzer, J. Pleba´nski, Quantum Mechanics in Non-Inertial Frames of Reference, Fortschritte der Physik 25, 37-82 (1977) [25] S. Deser, R. Jackiw, S. Templeton, Topologically massive gauge theories, Ann. Physics (NY) 140, 372-411 (1982) A. N. Redlich, Gauge Noninvariance and Parity Nonconservation of Three- Dimensional Fermions, Phys. Rev. Lett. 52, 18 (1984) [26] M. Konig, H. Buhmann, L. W. Molenkamp, T. Hughes, C. X Liu, X. L. Qi, S. C. Zhang, The Quantum Spin Hall Effect: Theory and Experiment, J. Phys. Soc. Jpn. 77, 031007-031020 (2008) A. Roth, C. Brune, H. Buhmann, L. W. Molenkamp, J. Maciejko, X. L. Qi, S. C. Zhang, Nonlocal Transport in the Quantum Spin Hall State, Science 325, 294-297 (2009) C. Brune, A. Roth, E. G. Novik, H. Buhmann, E. M. Hankiewicz, W. Hanke, J. Sinova, L. W. Molenkamp, Evidence for the ballistic intrinsic spin Hall effect in HgTe nanostructures, Nature physics 6, 448-454 (2010) C. Brune, A. Roth, H. Buhmann, E. M. Hankiewicz, L. W. Molenkamp, J. Maciejko, X. L. Qi, S. C. Zhang, Spin polarization of the quantum spin Hall edge, Nature physics 8, 485-490 (2012) [27] C. Brune, C.-X. Liu, E. G. Novik, E. M. Hankiewicz, H. Buhmann, Y.-L. Chen, X.-L. Qi, Z.-X. Shen, S.-C. Zhang, L. W. Molenkamp, Quantum Hall Effect from the Topological Surface States in Strained Bulk HgTe Phys. Rev. Lett. 106, 126803-126809 (2011) [28] S. Murakami, N. Nagaosa, S. C. Zhang, Spin-Hall Insulator, Phys. Rev. Lett. 93, 156804 (2004) B. A. Bernevig, S. C. Zhang, Quantum Spin Hall Effect Phys. Rev. Lett. 96, 106802-106805 (2006) B. A. Bernevig, T. L. Hughes, S. C. Zhang, Quantum Spin Hall Effect and Topological Phase Transition in HgTE Quantum Wells, Science 314, 1757- 1761 (2006) 29 [29] C. L. Kane, E. J. Mele, Z2 topological order and the quantum spin Hall effect, Phys. Rev. Lett. 95, 146802 (2005) L. Fu, C. L. Kane, Topological insulators with inversion symmetry, Phys. Rev. B 76, 045302 (2007) A. Kitaev, Periodic table for topological insulators and superconductors, Adv. in Theor. Physics: Landau Memorial Conf., AIP Conf. Proc. 1134, 22-30 (2009) M. Z. Hasan, C. L. Kane, Topological Insulators, Rev. Mod. Phys. 82, 3045 (2010) G. M. Graf, M. Porta, Bulk-Edge Correspondence for Two-Dimensional Topo- logical Insulators, Commun. Math. Phys. 324, 851–895 (2013) K. Gawedzki, Bundle gerbes for topological insulators, arXiv:1512.01028 [math- ph] Dec 2015 Hermann Schulz-Baldes, Topological Insulators from the Perspective of Non- commutative Geometry and Index Theory, Jahresbericht der Deutschen Mathematiker-Vereinigung 118, 247-273 (2016) R. Kennedy, M. R. Zirnbauer, Bott Periodicity for Z2 Symmetric Ground States of Gapped Free-Fermion Systems, Commun. Math. Phys. 342, 909-963 (2016) [30] J. Frohlich, Ph. Werner, Gauge theory of topological phases of matter, EPL 101, 47007-p1 - 47007-p6 (2013) [31] Z. Wang, X. G. Qi, S. C. Zhang, Equivalent topological invariants of topolog- ical insulators New Journal of Physics 12, 065007 (2010) [32] R. D. Peccei, H. R. Quinn, CP Conservation in the Presence of Pseudopar- ticles, Phys. Rev. Lett. 38, 1440-1443 (1977) [33] G. Moore, A Comment On Berry Connections, arXiv:1706.01149v2 [hep-th] 6 June 2017 [34] Yu. N. Obukhov, F. W. Hehl, Measuring a piecewise constant axion field in classical electrodynamics, Physics Letters 341, 357-365 (2005) [35] R. Li, J. Wang, X. L Qi, S. C. Zhang, Dynamical axion field in topological magnetic insulators, Nature physics 6, 284-288 (2010) [36] H. Ooguri, M. Oshikawa, Instability in Magnetic Materials with a Dynamical Axion Field, Phys. Rev. Lett. 108, 161803-161806 (2012) [37] J. E. Moore, The birth of topological insulators, Nature 464, 194-198 (2010) [38] "Topological Insulators", Contemporary Concepts of Condensed Matter Sci- ence, E. Burstein, A. H. MacDonald, P. J. Stiles (eds.), vol. 6, edited by M. Franz and L. Molenkamp, Elsevier, Oxford, Amsterdam, Waltham MA and San Diego CA, 2013 [39] H. M. Price, O. Zilberberg, T. Ozawa, I. Carusotto, N. Goldman, Four- Dimensional Quantum Hall Effect with Ultracold Atoms, Phys. Rev. Lett. 115, 195303 (2015) 30
1507.08844	1	1507	2015-07-31T11:57:22	Dephasing effects on coherent exciton-polaritons and the breaking of the strong coupling regime	[ "cond-mat.mes-hall" ]	Using femtosecond pump-probe spectroscopy, we identify excitation induced dephasing as a major mechanism responsible for the breaking of the strong-coupling between excitons and photons in a semiconductor microcavity. The effects of dephasing are observed on the transmitted probe pulse spectrum as a density dependent broadening of the exciton-polariton resonances and the emergence of a third resonance at high excitation density. A striking asymmetry in the energy shift between the upper and the lower polaritons is also evidenced. Using the excitonic Bloch equations, we quantify the respective contributions to the energy shift of many-body effects associated with Fermion exchange and photon assisted exchange processes and the contribution to collisional broadening.	cond-mat.mes-hall	cond-mat	Dephasing effects on coherent exciton-polaritons and the breaking of the strong coupling regime N. Takemura,1, ∗ S. Trebaol,2 M. D. Anderson,1 S. Biswas,1 D. Y. Oberli,1 M. T. Portella-Oberli,1 and B. Deveaud1 1Laboratory of Quantum Optoelectronics, ´Ecole Polytechnique F´ed´erale de Lausanne, CH-1015, Lausanne, Switzerland 2UMR FOTON, CNRS, Universit´e de Rennes 1, Enssat, F22305 Lannion, France. (Dated: June 21, 2021) Preprint Using femtosecond pump-probe spectroscopy, we identify excitation induced dephasing as a major mechanism responsible for the breaking of the strong-coupling between excitons and photons in a semiconductor microcavity. The effects of dephasing are observed on the transmitted probe pulse spectrum as a density dependent broadening of the exciton-polariton resonances and the emergence of a third resonance at high excitation density. A striking asymmetry in the energy shift between the upper and the lower polaritons is also evidenced. Using the excitonic Bloch equations, we quan- tify the respective contributions to the energy shift of many-body effects associated with Fermion exchange and photon assisted exchange processes and the contribution to collisional broadening. PACS numbers: 78.20.Ls, 42.65.-k, 76.50.+g A semiconductor microcavity is a system that confines photons and allows them to strongly interact with quan- tum well excitons [1]. Polaritons are composite parti- cles arising from the coherent superposition of a pho- ton and exciton. The interactions mediated by their excitonic part make semiconductor microcavities a suit- able platform for realizing nonlinear optical devices such as: bistablity memory [2] and polariton switching [3]. Moreover, applications of semiconductor microcavities to quantum information have been also proposed on the ba- sis of their coherent behaviour [4, 5]. When modeling these devices, they are usually described with a nonlin- ear Schrodinger equation, which is formally the same as the Gross-Pitaevskii equations (GPE) used for coherent ground-state of Bose condensed dilute atoms. This di- rectly leads to a wide range of analogies between exciton- polaritons and cold atoms. Actually, in semiconductor microcavity systems, we can investigate a wide range of physics including Bose-Einstein condensation [6, 7] and the collective quantum fluid nature [8 -- 10]. On the other hand, one of the important properties of a semiconduc- tor system is the existence of dephasing [11 -- 13], which induces decoherence. The investigation of the effect of decoherence, induced by excitonic dephasing, on the po- lariton dynamics is important both for understanding the physics of exciton-photon strongly coupled systems and for designing semiconductor microcavity devices such as nonlinear optical devices and polariton-based qubits. In this letter, we show that excitation induced dephas- ing (EID) plays an important role in the dynamics of polaritons in a semiconductor microcavity. The inves- tigation is experimentally carried out by time-resolved femtosecond pump-probe optical spectroscopy. For the theoretical description of our results, we utilize the ex- citonic Bloch equations (EBE) approach, taking into ac- count separately the coherent part of the polariton pop- ulation and an incoherent population of excitons [14, 15]. We study the role of exciton-exciton interactions, photon- assisted exchange scattering, and EID effects on the po- lariton dynamics. The experimental results are very well reproduced by EBE and not by the exciton-photon GPE, which assumes that excitons are in a coherent limit. The experiment is performed in a high quality GaAs- based microcavity [16] at the cryogenic temperature of 4K. A single 8 nm In0.04Ga0.96As quantum well is em- bedded in between two GaAs/AlAs distributed Bragg- reflectors. The Rabi splitting energy is 2Ω=3.45 meV at zero cavity detuning [9]. For the accurate measurement of the transmitted probe beam, we employ a heterodyne pump-probe setup with a degenerated beam configura- tion at k = 0 µm−1 [17], which dramatically increases the signal-to-noise ratio. The pump and probe pulses originate both from a broadband few hundreds femtosec- ond Ti:Sapphire laser. The center of the laser spectrum is set between the lower and upper-polariton peaks. Ad- ditionally, noise coming from laser spectrum envelope is removed with the aid of a numerical low-pass filter. The experimental setup is described in detail in our previous papers [10, 17]. In order to avoid the complex effects of biexcitons [17, 18], the pump and probe beams are co- circularly polarized. We obtain a time delay map through successive measurements of the pump-probe spectrum. The experimental results are presented in Fig. 1 (a) and (c), showing the probe spectra as a function of pump-probe time delay. The cavity detuning is set at c − x = 0.8 meV, where c(x) is cavity mode (exciton) energy. In this figure, the pump pulse arrives before (af- ter) the probe pulse at positive (negative) pump-probe delays. For low pump intensity (Fig. 1 (a)), we observe two polariton branches (lower and upper) at both posi- tive and negative delays and the lower polariton shows a maximum blue-shift at zero delay. The delay dependence of the lower-polariton blue shift is asymmetric with re- spect to zero delay. While the blue shift gradually de- creases at negative delays, it stays at positive delays for long time. No clear energy shift of the upper polariton 5 1 0 2 l u J 1 3 ] l l a h - s e m . t a m - d n o c [ 1 v 4 4 8 8 0 . 7 0 5 1 : v i X r a resonance is seen for all delays. At a high pump intensity (Fig. 1 (c)), a triple peak structure appears at negative delays, while a single peak exists at positive delays. With the aid of numerical simulations based on EBE, we show that such behaviours originate from a long-living incoher- ent population and a short-living coherent polarization of excitons. 2 For the analysis of the experiment, we use EBE [14, 15]. The starting point of EBE is a bosonic exciton-photon Hamiltonian: (cid:90) (cid:20) H = c x dx (cid:90) (cid:90) ψx + ψ† ψ† x +Ω( ψ† ψx + ψ† dx(cid:48)(cid:20) 1 ψc) ψ† + 2 −Vpae(x − x(cid:48))( ψ† ψ†(cid:48) dx x x c c(c − 2∇2 (cid:105) 2mc ) ψc x Vex(x − x(cid:48)) ψx ψ†(cid:48) ψ†(cid:48) ψ(cid:48) x + ψ† ψx x x x x ψ(cid:48) ψ(cid:48) c) ψx (1) (cid:105) . † ψx(c) and ψ x(c) are exciton (photon) field creation and annihilation operators. They obey the Bose commuta- †(cid:48) x(c)] = δ(x − x(cid:48)). This Hamilto- tion relation, [ ψx(c), ψ nian can be obtained from the electron-hole Hamiltonian via a boson mapping method called the Usui transfor- mation [14]. Since the exciton mass is large, the kinetic term of the exciton is neglected. The interactions are as- sumed to be contact interactions: Vex(x−x(cid:48)) = gδ(x−x(cid:48)) and Vpae(x − x(cid:48)) = gpaeδ(x − x(cid:48)) [19, 20]. The exciton- exciton interaction potential Vex is associated with the Coulomb exchange scattering. The term Vpae is a pho- ton assisted exchange scattering [21] and contributes to the reduction of the Rabi coupling, which is the remi- niscence of the fermionic nature of the exciton [14]. In order to obtain a closed set of equations, we truncate the hierarchy by applying the following assumptions such as (cid:104) ψ† ψx(cid:105) (cid:39) (cid:104) ψ†(cid:48) x(cid:105)(cid:104) ψx(cid:105), ψ(cid:48) x(cid:105) = 0. We x(cid:105)(cid:104) ψ(cid:48) (cid:104) ψx ψ(cid:48) ψx(cid:105) and define the exciton population as N (x, t) = (cid:104) ψ† the exciton polarization as P (x, t) = (cid:104) ψx(cid:105). Assuming factorization between the photon and exciton, we define E(x, t) = (cid:104) ψc(cid:105). With the aid of the Heisenberg equation of motion, the EBE then reads [14] i N = −iΓxN − 2i(Ω − 2gpaeN )Im[P E∗] i P = (x + g0N − iγx(N )) P + (Ω − 2gpaeN )E i E = (c − 2 (2) ∇2 − iγc)E + (Ω − gpaeN )P − fext. ψx ψx(cid:105) = 0, and (cid:104) ψ† ψ(cid:48) x(cid:105) (cid:39) (cid:104) ψ† ψ(cid:48) ψx(cid:105)(cid:104) ψx(cid:105), (cid:104) ψ†(cid:48) ψx(cid:105) (cid:39) (cid:104) ψ† ψ(cid:48) x x x x x x x x x 2mc To obtain the above equations, the interaction constant g is phenomenologically divided into a real and imaginary part: g = g0 − ig(cid:48). The real part g0 is associated with an energy renormalization, while the imaginary part g(cid:48) represents the strength of EID, which is also referred to as collisional broadening. The ratio of the constants is estimated as gpae/g0 (cid:39) Ω/6nsEba2 0 [22]. ns is the sat- uration density of excitons. Eb and a0 are respectively FIG. 1. (color online) Measured and simulated probe trans- mission are shown as a function of energy and time delay between pump and probe pulse. The spectra are measured for two different pump intensities: 1.48×1013 (1 mW) (a) and 7.4×1013 (5 mW) (c) photons pulse−1 cm−2. Corresponding simulated spectra are attached below the measured spectra ((b) and (c)). The black dashed lines represent the lower and upper-polariton peak energies without pump pulse. The white dashed lies are the cavity photon and exciton energies. In the simulation, the intensity I0 is defined as I0 = 0.8/g0. the exciton binding energy and Bohr radius. The con- stants, γx(N ) and Γx are respectively polarization de- phasing and population decay rate of excitons. In gen- eral, γx(N ) is written as [11, 23, 24], γx(N ) = Γx/2 + γ∗ x + g(cid:48)N, (3) energy (meV)(c)pump-probe delay (ps)experimentsimulationexperimentsimulation(a)energy (meV)pump-probe delay (ps)1 mW5 mW1 mW(F pu2 = I0)5 mW(F pu2 = 5I0)minMax00.51(b)(d)pump-probe delay (ps)pump-probe delay (ps)energy (eV)energy (eV)εcεxεcεxεcεxεcεxwhere γ∗ x is the pure dephasing term. In the terminol- ogy of two level systems, Γx and γx correspond to the inverse of T1 and T2 times respectively . The EID con- stant g(cid:48) introduces a phenomenological linear increase of the dephasing that depends on the exciton population N , which plays an important role in our experiment. 3 x ψx(cid:105) = (cid:104) ψ† The advantage of EBE, compared to GPE, is that we can apply independent decay rates for the coherent polar- ization and incoherent population and calculate the time evolution of each. Indeed, in the commonly used GPE, a factorization, (cid:104) ψ† x(cid:105)(cid:104) ψx(cid:105) is implicitly assumed [20]. Therefore, the dynamics of the population N (x, t) is uniquely determined by the polarization P (x, t) through N = P2. In the EBE, this condition is satisfied when neither pure dephasing nor EID exist: γ∗ x = 0 meV and g(cid:48) = 0 meV (coherent limit). The EBE are close analogue of the optical Bloch equations (OBE) [13, 25], however differing since OBE are based on a two-level electron-hole system, while EBE are based on a bosonic exciton basis [14]. To reproduce the experiments, Γx and γc are chosen to be 0.01 meV and 0.1 meV respectively. The pure de- phasing is set to γ∗ x = 0.1 meV [26], additionally, we include EID as the primary decoherence mechanism in our simulations. We set the interaction constants as g(cid:48) = 0.4g0 and gpae = 0.3g0. fext is the excitation photon field and is assumed to be a Gaussian pulse: fext = F pu(pr) exp(−(t − t0)2/(τ 2)) exp(−iωpulse t). We set ωpulse at the center of both polariton branches and set τ =0.5 ps. 0 0 For the calculation of the pump-probe dynamics, we use a coupled-mode approximation: N (x, t) = N pu + N preik·x + N pr∗e−ik·x (the population is a real value), P (x, t) = P pu + P preik·x + P ide−ik·x, and E(x, t) = Epu + Epreik·x + Eide−ik·x. For example, P pu, P pr, and P id represent pump, probe, and idler component of the polarization respectively. Considering the conservation of momentum, we obtain 8 coupled equations (See Sup- plementary material). The pump and probe pulses are introduced as Epu and Epr respectively and the trans- mitted probe signal is obtained through Epr. This is the standard method of calculating a transient four-wave mixing signal in optical Bloch equations [27, 28]. Since the wave number of the probe is sufficiently small, we neglect the momentum dispersion of the photon mode. The simulated probe transmission spectra are given in Fig. 1 (b) and (d) for two different pump intensities. There are striking similarities between the experimental and simulated spectra. Firstly, Fig. 1 (b) features a long lasting (∼ /Γx) blue-shift of the lower-polariton in the positive delay, while the blue-shift builds up on a shorter time scale (∼ 2/(γc + γ∗ x)) in the negative delay. On the other hand, the energy shift of the upper-polariton is al- most zero because of the cancellation of the blue and red- shift contributions induced by g0 and gpae respectively. The high-density simulation (Fig. 1 (d)) reproduces both FIG. 2. (color online) Simulated probe transmission as a func- tion of energy and pump-probe time delay without EID or pure dephasing (g(cid:48) = γ∗ x = 0 meV). The other parameters are the same as those used in the simulation of Fig. 1. Fig. 2 (a) and (b) respectively correspond to 1 mW and 5 mW pump intensities. the occurrence of three peaks at negative delays and of a single peak at positive delays. At negative delays, the signal shows a three peak structure, this is because the signal is temporally convoluted due to the finite lifetime of the probe pulse in the sample. Therefore the side peaks arise from the portion of the probe which is transmitted before the arrival of the pump pulse; however, the mid- dle peak builds with the arrival of the pump pulse in the short lifetime of the polarization given by ∼ 2/(γc +γ∗ x). At positive delays, since the probe polarization is always affected by the long lasting pump population, the single middle peak remains for a long time (∼ /Γx). If the pump intensity is further increased, eventually the term γx(N ) becomes comparable to the effective Rabi coupling Ω− 2ggpaeN and the central peak asymptotically reaches the cavity mode c, which is the signature of a strong to weak coupling transition [29]. For the purpose of better understanding the effect of the incoherent exciton population and of EID on the two polariton resonances, we present in Fig. 2 a simulation without EID or pure dephasing (g(cid:48) = γ∗ x = 0 meV). The other parameters are same as for Fig. 1. The probe transmission of Fig. 2 is a simulation of the coherent limit, where the polarization decay (dephasing) rate is a half of the population decay rate (γx(N ) = Γx/2 = 0.005 meV). We find that the polariton branch is broadened to- wards the high energy side because of "dynamical energy- shift". Namely, the mean-field energy-shift of polariton temporarily decreases following the decay of the polari- energy (meV )simulationCoherent limitpump-probe delay (ps)pump-probe delay (ps)simulation1 mW(F pu2 = I0)5 mW(F pu2 = 5I0)minMax00.51(a)(b)εcεxεcεx4 (∼ /γc). On the other hand, with EID, we have to dis- tinguish between the dynamics of the polarization and that of the population. While the exciton polarization P (t) is directly coupled to the cavity photon E(t) (not shown), there is no direct coupling between the exciton population N (t) and the photon E(t). Therefore, while the polarization decays with a lifetime of the same order as that of the cavity photon (∼ /γc), the population de- cays independently and stays for a long time (∼ /Γx), even after the disappearance of the polarization. Intu- itively, the EID process can be understood as follows; microscopically, the exciton-exciton collisions introduce an energy fluctuation, which gives an additional random phase to the time evolution of the exciton field opera- tor ψx(x, 0)ei(x+δ)t/ . Since the phase δ is random, the expectation value P (x, t) = (cid:104) ψx(x, t)(cid:105) shows a decay [31], which is the origin of the imaginary part of the inter- action constant g(cid:48) (EID) [24, 32]. Meanwhile, the the en- ergy fluctuations affect neither the term ψ† x(x, t) ψx(x, t) x(x, t) ψx(x, t)(cid:105) nor its expectation value N (x, t) = (cid:104) ψ† due to a phase cancellation [31]. Finally, we comment that the incoherent exciton population N should be in- terpreted as the "inactive" excitonic reservoir already dis- cussed in the context of non-resonantly excited polariton condensates [33, 34]. Actually, both incoherent exciton population featured in Fig. 3 and inactive excitonic reser- voir have a long lifetime and nevertheless contribute to the energy shift of the polariton resonances [35]. In conclusion, we investigated the coherent dynamics of exciton-polaritons by femtosecond pump-probe spec- troscopy in a high quality semiconductor microcavity. We demonstrated that excitation-induced dephasing, as a manifestation of exciton-exciton interactions, is a ma- jor mechanism for the breaking of the strong coupling regime. We showed that the experimental results could only be simulated with the inclusion of EID in the exci- tonic Bloch equations. The present work is supported by the Swiss National Science Foundation under project N◦135003 and the Eu- ropean Research Council under project Polaritonics con- tract N◦291120. The polatom network is also acknowl- edged. ∗ E-mail: [email protected] [1] C. Weisbuch, M. Nishioka, A. Ishikawa, and Y. Arakawa, Phys. Rev. Lett. 69, 3314 (1992). [2] R. Cerna, Y. L´eger, T. K. Paraıso, M. Wouters, F. Morier-Genoud, M. T. Portella-Oberli, and B. De- veaud, Nature comm. 4 (2013). [3] A. Amo, T. Liew, C. Adrados, R. Houdr´e, E. Giacobino, A. Kavokin, and A. Bramati, Nature Phot. 4, 361 (2010). [4] S. S. Demirchyan, I. Y. Chestnov, A. P. Alodjants, M. M. and A. V. Kavokin, Phys. Rev. Lett. 112, Glazov, 196403 (2014). FIG. 3. (color online) Simulated time evolution of the k = 0 µm−1 polarization P2 and population N as a function of time after an arrival of a single pulse. Simulations with EID and pure dephasing (a,c) and in a coherent limit (g(cid:48) = γ∗ = 0 meV) (b,d) are presented for two different laser pulse intensities 1 mW (a,b) and 5 mW (c,d). The parameters are the same as the simulation of Fig. 1. The dashed lines represent scaled laser pulses. ton density. The time integration of the temporal de- crease of the energy-shift introduces a broadening of the polariton branches. The blue shift of both polariton res- onances decay two times faster at positive delays than they emerge at negative ones. In this limit, the dynamics of the exciton population is uniquely determined by the polarization and we can replace the population N (x, t) with the square of the polarization P (x, t)2. Here the three sets of equations can be reduced to two equations composed of the exciton polarization and electric field, this is the commonly used exciton-photon GPE [20, 30]. Clearly, the GPE cannot reproduce the dynamics of po- laritons in the presence of EID, which is evidenced by the huge differences between Fig. 2 and the experiments (Fig. 1 (a) and (c)). In particular, the high pump in- tensity simulation (Fig. 2 (b)) reproduces neither the three peak structure nor the disappearance of the quan- tum beat pattern for the positive delay. This simulation also implies that the bleaching of the upper and lower polariton resonances in the positive delay (Fig. 1 (c)) is associated with EID. The transition to the weak coupling regime can also be observed with the GPEs, however the very strong dynamical blue-shift effects completely differ from the observed experimental behaviour. In Fig. 3, we present simulated real time evolutions of exciton polarization P (t) and population N (t) at k = 0 µm−1 after the arrival of a single laser pulse. Without EID (coherent limit), the time evolution of the exciton population N (t) is found to coincide with that of P (t)2, which supports the relation N (t) = P (t)2 and the fac- torization (cid:104) ψ† x(cid:105)(cid:104) ψx(cid:105), which is assumed in the exciton-photon GPE. In this case, the lifetime of the sys- tem is mainly determined by the short photon lifetime ψx(cid:105) = (cid:104) ψ† x P2 and N (a.u.)(a)coherent limit(b)(c)(d)P2Nwith EID and pure dephasing5 mW (F pu2 = 5I0)1 mW (F pu2 = I0)time (ps)time (ps)P2 and N (a.u.)5 mW (F pu2 = 5I0)1 mW (F pu2 = I0)[5] L. Dominici, D. Colas, S. Donati, J. P. Restrepo Cuar- tas, M. De Giorgi, D. Ballarini, G. Guirales, J. C. L´opez Carreno, A. Bramati, G. Gigli, E. del Valle, F. P. Laussy, and D. Sanvitto, Phys. Rev. Lett. 113, 226401 (2014). [6] J. Kasprzak, M. Richard, S. Kundermann, A. Baas, P. Jeambrun, J. Keeling, F. Marchetti, M. H. Szyman- ska, R. Andr´e, J. L. Staehli, V. Savona, P. B. Littlewood, B. Deveaud, and L. S. Dang, Nature 443, 409 (2006). [7] H. Deng, G. Weihs, C. Santori, J. Bloch, and Y. Ya- mamoto, Science 298, 199 (2002). [8] A. Amo, J. Lefr`ere, S. Pigeon, C. Adrados, C. Ciuti, I. Carusotto, R. Houdr´e, E. Giacobino, and A. Bramati, Nature Phys. 5, 805 (2009). [9] V. Kohnle, Y. L´eger, M. Wouters, M. Richard, M. T. and B. Deveaud-Pl´edran, Phys. Rev. Portella-Oberli, Lett. 106, 255302 (2011). [10] V. Kohnle, Y. L´eger, M. Wouters, M. Richard, M. T. and B. Deveaud, Phys. Rev. B 86, Portella-Oberli, 064508 (2012). [11] J. Shah, Ultrafast spectroscopy of semiconductors and semiconductor nanostructures, Vol. 115 (Springer Science & Business Media, 1999). [12] M. Kira and S. Koch, Progress in quantum electronics 30, 155 (2006). [13] H. Wang, H. Q. Hou, and B. E. Hammons, Phys. Rev. Lett. 81, 3255 (1998). [14] G. Rochat, C. Ciuti, V. Savona, C. Piermarocchi, A. Quattropani, and P. Schwendimann, Phys. Rev. B 61, 13856 (2000). [15] D. Wang and M.M. Dignam, Phys. Rev. B 79, 165320 (2009). [16] R. Stanley, R. Houdre, U. Oesterle, M. Gailhanou, and M. Ilegems, Appl. Phys. Lett. 65, 1883 (1994). [17] N. Takemura, S. Trebaol, M. Wouters, M. T. Portella- Oberli, and B. Deveaud, Phys. Rev. B 90, 195307 (2014). [18] N. Takemura, S. Trebaol, M. Wouters, M. T. Portella- Oberli, and B. Deveaud, Nature Physics 10, 500 (2014). [19] L. Pitaevskii and S. Stringari, "Bose-einstein condensa- tion (international series of monographs on physics)," (2003). [20] I. Carusotto and C. Ciuti, Reviews of Modern Physics 85, 299 (2013). [21] M. Combescot, M. Dupertuis, and O. Betbeder-Matibet, EPL (Europhysics Letters) 79, 17001 (2007). [22] D. Sarchi and V. Savona, Phys. Rev. B 75, 115326 (2007). [23] T. Baars, M. Bayer, A. Forchel, F. Schafer, and J. P. Reithmaier, Phys. Rev. B 61, R2409 (2000). [24] C. Ciuti, V. Savona, C. Piermarocchi, A. Quattropani, and P. Schwendimann, Phys. Rev. B 58, 7926 (1998). [25] X. Li, T. Zhang, C. N. Borca, and S. T. Cundiff, Physical review letters 96, 057406 (2006). [26] A. Honold, L. Schultheis, J. Kuhl, and C. W. Tu, Phys. Rev. B 40, 6442 (1989). [27] T. Yajima and Y. Taira, Journal of the Physical Society of Japan 47, 1620 (1979). [28] D. B. Turner, P. Wen, D. H. Arias, and K. A. Nelson, Physical Review B 84, 165321 (2011). [29] F. Quochi, G. Bongiovanni, A. Mura, J. L. Staehli, B. De- veaud, R. P. Stanley, U. Oesterle, and R. Houdr´e, Phys. Rev. Lett. 80, 4733 (1998). [30] Strictly speaking, the coupled-mode approximated EBE in coherent limit is slightly different from a coupled-mode GPE. [31] S. Schmitt-Rink, S. Mukamel, K. Leo, J. Shah, and D. S. 5 Chemla, Physical Review A 44, 2124 (1991). [32] C. Cohen-Tannoudji, J. Dupont-Roc, G. Grynberg, and P. Thickstun, Atom-photon interactions: basic processes and applications (Wiley Online Library, 1992). [33] K. G. Lagoudakis, F. Manni, B. Pietka, M. Wouters, T. C. H. Liew, V. Savona, A. V. Kavokin, R. Andr´e, and B. Deveaud-Pl´edran, Phys. Rev. Lett. 106, 115301 (2011). [34] M. De Giorgi, D. Ballarini, P. Cazzato, G. Deligeorgis, S. I. Tsintzos, Z. Hatzopoulos, P. G. Savvidis, G. Gigli, F. P. Laussy, and D. Sanvitto, Phys. Rev. Lett. 112, 113602 (2014). [35] L. Ferrier, E. Wertz, R. Johne, D. D. Solnyshkov, P. Senellart, I. Sagnes, A. Lemaıtre, G. Malpuech, and J. Bloch, Phys. Rev. Lett. 106, 126401 (2011). SUPPLEMENTAL MATERIAL: COUPLED MODE EQUATIONS FOR PUMP-PROBE DYNAMICS We present explicit forms of the coupled mode equa- tions for simulating pump-probe spectra. We restrict in- volving modes into three modes: pump, probe and idler. The polarization P, population N and electric field E are respectively written as N = N pu + N preikr + N pr∗e−ikr, P = P pu + P preikr + P ide−ikr, and E = Epu + Epreikr + Eide−ikr. Substituting these representations into the ex- citonic bloch equations (EBE) in the manuscript and ne- glecting components such as e±i2k and e±i3k, we obtain 8 coupled equations of motions. Firstly, the population N read, i N pu = −iΓxN pu −Ωapu + 2gpae(apuN pu + bidN pr + bprN pr∗) i N pr = −iΓxN pr −Ωbpr + 2gpae(bprN pu + apuN pr + cN pr∗) Now the quantities apu, bpr, bid and c are given by apu = 2iIm(P puEpu∗ + P prEpr∗ + P idEid∗) bpr = P puEid∗ − P id∗Epu + P prEpu∗ − P pu∗Epr bid = P puEpr∗ − P pr∗Epu + P idEpu∗ − P pu∗Eid c = P prEid∗ − P id∗Epr. The equations of motion of exciton polarization is written as i P pu = (x − iγx)P pu + g(N puP pu + N pr∗P pr + N prP id) +ΩEpu − 2gpae(N puEpu + N pr∗Epr + N prEid) i P pr = (x − iγx)P pr + g(N puP pr + N prP pu) +ΩEpr − 2gpae(N puEpr + N prEpu) i P id = (x − iγx)P id + g(N puP id + N pr∗P pu) +ΩEid − 2gpae(N puEid + N pr∗Epu). Finally, the electric fields follow the following equations: i Epu = (c − iγc)Epu + ΩP pu −gpae(N puP pu + N prP id + N pr∗P pr) − f pu ext i Epr = (c − iγc)Epr + ΩP pr −gpae(N puP pr + N prP pu) − f pr ext i Eid = (c − iγc)Eid + ΩP id − gpae(N puP id + N pr∗P pu). 6
1305.4221	1	1305	2013-05-18T03:35:36	Spin force and the generation of sustained spin current in time-dependent Rashba and Dresselhauss systems	[ "cond-mat.mes-hall" ]	The generation of spin current and spin polarization in a 2DEG structure is studied in the presence of Dresselhaus and Rashba spin-orbit couplings (SOC), the strength of the latter being modulated in time by an ac gate voltage. By means of the non-Abelian gauge field approach, we established the relation between the Lorentz spin force and the spin current in the SOC system, and showed that the longitudinal component of the spin force induces a transverse spin current. For a constant (time-invariant) Rashba system, we recover the universal spin Hall conductivity of $\frac e{8\pi}$, derived previously via the Berry phase and semiclassical methods. In the case of a time-dependent SOC system, the spin current is sustained even under strong impurity scattering. We evaluated the ac spin current generated by a time-modulated Rashba SOC in the absence of any dc electric field. The magnitude of the spin current reaches a maximum when the modulation frequency matches the Larmor frequency of the electrons.	cond-mat.mes-hall	cond-mat	Spin force and the generation of sustained spin current in time-dependent Rashba and Dresselhauss systems APS/123-QED Cong Son Ho,1, 2 Mansoor B. A. Jalil,1 and Seng Ghee Tan()1, 2 1Department of Electrical and Computer Engineering, National University of Singapore,4 Engineering Drive 3, Singapore 117576, Singapore. 2Data Storage Institute, Agency for Science, Technology and Research (A*STAR), DSI Building, 5 Engineering Drive 1,Singapore 117608, Singapore. (Dated: May 11, 2018) The generation of spin current and spin polarization in a 2DEG structure is studied in the presence of Dresselhaus and Rashba spin-orbit couplings (SOC), the strength of the latter being modulated in time by an ac gate voltage. By means of the non-Abelian gauge field approach, we established the relation between the Lorentz spin force and the spin current in the SOC system, and showed that the longitudinal component of the spin force induces a transverse spin current. For a constant (time-invariant) Rashba system, we recover the universal spin Hall conductivity of 8π , derived previously via the Berry phase and semiclassical methods. In the case of a time-dependent SOC system, the spin current is sustained even under strong impurity scattering. We evaluated the ac spin current generated by a time-modulated Rashba SOC in the absence of any dc electric field. The magnitude of the spin current reaches a maximum when the modulation frequency matches the Larmor frequency of the electrons. e PACS numbers: 72.25.Dc, 71.70.Ej, 71.10.Ca I. INTRODUCTION can be utilised as spin pumping sources to generate spin current without applying bias. The generation of a sustained spin polarized current constitutes a key requirement in practical spintronics. In two-dimensional electron gas (2DEG) systems with Rashba spin-orbit coupling (SOC), spin precession about the Rashba field and the accompanying Zitterbewegung- like motion would usually result in zero net transverse spin current on average.1 However, a net spin polar- ization can be generated in the transverse direction via the spin Hall effect. Based on the time-gauge field2 or the semi-classical geometric force3 descriptions, it can be shown that a vertical (out-of-plane) spin polarization can be sustained under adiabatic conditions, albeit un- der ballistic limit only. Other methods of generating spin current which have been proposed include the applica- tion of a non-uniform external magnetic field in a SOC medium4 and the use of optical excitation.5,6 Recently, an all-electrical method has been also investigated by utiliz- ing a time-varying gate voltage.7 -- 10 The idea is based on the fact that the Rashba SOC can be tuned by changing the applied gate voltage.11 -- 13 However, it is pointed out that, in the Rashba system, only the in-plane polarized spin currents are generated.7,10 Therefore, it would be in- teresting if the out-of-plane polarized component of the spin current can be generated by applying this method. In this article, we will present a theoretical study of the spin current generation in 2DEG system that combines both time-dependent Rashba and Dresselhauss SOCs. In frame work of non-Abelian gauge field, we will point out that time-modulated gate voltage can indeed induce a spin current polarized in the out-of-plane direction of the 2DEG. This spin current can be maintained even under non-ballistic situation, i.e., in the presence of impurities. Moreover, we will show that such combined SOC systems II. SPIN FORCE AND SPIN HALL CURRENT We first consider a system with Dresselhaus and Rashba SOC, with the Rashba coupling being time- dependent (shown schematically in Fig. 1). The Hamil- tonian of the system is: H = p2 2m + HR(t) + HD + V (r), (1) where m is the effective electron mass, HD = β (pyσy − pxσx) is the Dresselhaus SOC, and HR(t) = α (t) (pxσy − pyσx) is the Rashba SOC, with the Rashba coupling consisting of static and time-varying compo- nents, i.e., α (t) = α0 + α1(t). The last term in Eq. (1) is the potential energy which may include the external applied electric field and the impurity field. To the first order in SOC couplings, Eq. (1) can be rewritten in the form of the Yang-Mills Hamiltonian 2m (p − eA)2, where A is the gauge field given HY M = 1 by A = (Ax,Ay, 0) = m e (−ασy + βσx, ασx − βσy, 0) , (2) and the corresponding curvature is Fµν = ∂µAν − ∂νAµ − ie [Aµ,Aν] . (3) The physical fields extracted from the curvature tensor can be considered as the effective magnetic and electric 2 The velocity operator can be formally derived from the Hamiltonian as v = ∂H/∂p, and the force is F = dv/dt. At some instant t, the average force can be decomposed into two parts as hFi = F0 + Ft in which fn, dvn dt F0 = mZ Xn=± 2 Z γ∇pγ(E+ − E−)(f+ − f−), m Ft = − (9) (10) (11) ∂p where, vn = ∂En(p) is the velocity of electron belong- ing to the nth branch, and the integral is taken over momentum space. The first force term F0 is related to the evolution of the velocity over time due to the ap- plied electric field and the modulating Rashba coupling, while the second component Ft is related to the explicit time-dependence of the eigen-state and will vanish if the Hamiltonian is time-independent. Constant Rashba coupling -- We first consider the static case when α = α0. The second term on the right hand side of Eq. (6) vanishes, hence the spin Hall current (SHC) is just dependent on the total force acting on the electrons. Since the eigen-energy and eigen-state are time-independent, its force component Ft vanishes. For simplicity, we consider an electric field applied along the x-direction, following which the longitudinal force in Eq.(10) is readilyf found to be Fx = 1 ± m(α0 2−β2)2sin θ 2 2+β2+2α0βsin 2θ )3/2! dpx dt p(α0 , (12) where dpx/dt = eEx. Comparing the above force expres- sion with that obtained by the gauge formalism, i.e. Eq. (6), we arrive at the formula of the spin current for the two eigenbranches: jz s,y (p) = ± eEx(cid:0)α0 2 − β2(cid:1) 2sin θ2 2 + β2 + 2α0βsin 2θ )3/2 . 4pm(α0 (13) The total SHC at zero temperature is then found by inte- grating over all occupied states in the momentum space: J z s,y = 1 (2π)2 Z p+ p− pdpZ 2π 0 dθ jz s,y (p) (14) where, p+ and p− are the Fermi momenta cor- responding to the two eigenbranches, and the dif- ference between them is given by: p+ − p− = . With this, the SHC ex- 2m(cid:0)α0 pression simplifies to 2 + β2 + 2α0β sin2θ(cid:1)1/2 2 − β2 0 − β2(cid:16) e α0 α2 J z s,y = 8π(cid:17) Ex. (15) FIG. 1. Schematic diagram of a 2DEG system with Rashba and Dresselhaus SOC. The strength of the Rashba SOC can be modulated in time by applying an ac gate bias. fields, i.e., B = Fxy z = 2m2(α2 − β2) σz z, e E = ∂tA = − m e α(σ × z), (4) (5) where α = ∂α/∂t, z is the unit vector in the z-direction. These Yang-Mills fields exert a spin-dependent force on the electron, with the Lorentz-like magnetic field act- ing in the direction transverse to the electron motion as FY M = e(E + v × B), where v is the velocity. Note that the Lorentz force term is proportional to the spin current polarized along the z-direction, which is given by jz s = (/4){v, σz}. Therefore, one can arrive at the expression for the quantum mechanical force: 4m2(cid:0)α2 − β2(cid:1) (jz s × z) − m α (σ × z) − ∇V . F = (6) 2 From the above, we see the interplay between spin force, spin current and the time modulation of the Rashba SOC. The spin current is dependent on either the spin force or the time-modulated RSOC, and would still be present if either one of the two terms vanishes. One can directly evaluate the spin current from its definition by solving the spin dynamics of the system.14,15 However, this method may be quite involved if the Hamiltonian is non-trivially time-dependent. Alternatively, the spin current may be obtained by evaluating the force by some other means, and substituting its expression into Eq. (6). In the fol- lowing, we evaluate the force by classical consideration, and subsequently apply the force expression to determine the spin current. consider and eigen-energy of the time-dependent SOC Hamiltonian HSOCψn(t)i = En(t)ψn(t)i, given by ± ieiγ(cid:19) , instantaneous eigenstate We the (7) ψ±(cid:11) = E± = 1 √2(cid:18) 1 p2 2m ∓ ∆E, (8) in which tan γ = (αpy + βpx) / (αpx + βpy) , ∆E = p(α2 + β2 + 2αβsin 2θ )1/2 with θ = tan−1py/px . The expectation value of an operator at time t is defined as h Oi = Pn hψn Oψiifn, with fn = f (En) is the Fermi distribution function. The above result is consistent with the well-known uni- versal spin Hall conductivity of (e/8π) obtained previ- ously by Sinova et al.,14 and by Shen,15 the latter by considering linear response (Kubo) transport theory and the Berry phase of the system. Up to this point, we have explicitly shown the relation between the spin Hall cur- rent and the spin force in a time-independent system. In Ref.1, the spin force concept is introduced to describe the spin dynamics in a Rashba SOC system. Specifically, the study focused on the transverse component of the spin force and its effect on the Zitterbewegung dynam- ics of the electrons in the system. However, the con- nection between this spin force and the well-known spin Hall current was not established. In the above analysis, we have shown that it is the longitudinal component of the spin force, whose magnitude is proportional to the driving electric field, which gives rise to the spin Hall current with the universal conductivity of (e/8π). The longitudinal spin force is related to the Lorentz force of the Yang-Mills (non-Abelian) effective fields arising from the Rashba and Dresselhaus SOC effects. Our physical picture of the spin Hall effect is consistent with previ- ous descriptions of the spin Hall effect based on different theoretical models.2,14,15 III. EFFECT OF IMPURITY SCATTERING placed delta potentials: Vim(r) = V PN We now consider the effect on the spin current of impurities, which can be modeled as some randomly i=1 δ (r − Ri). As has been pointed out previously,16 the spin Hall current would be suppressed in an infinitely large disordered sys- tem. In the steady state, the total force is zero, i.e., F = 0. Additionally, the external electric field which is the driving force for the spin current, is also effectively canceled by the effect of impurities.17 Thus from Eq. (6), the spin current vanishes if the Rashba coupling is con- stant in time, as expected. However, in the presence of a time-modulated Rashba SOC strength, the α term in Eq. (6) is non-zero, and this leads to a spin current of J z s,i = 2 4m αSi (α2 − β2) , (16) in which Si = hσii is the spin density, with h. . .i de- noting the expectation values taken over all momentum and spin spaces, and impurity configurations. We see that the transverse spin current in any one direction de- pends on the spin polarization along that direction; in general the latter may be finite and hence the spin cur- rent can be sustained. On the other hand, the spin cur- rent can also be interpreted as the response to the ef- (5). Thus, fective electric field which is given by Eq. we have J z s,i = σijEj, with the "spin conductivity" of σxy = −σyx = 4m2(α2−β2) . We consider a special case in which the Rashba and Dresselhaus coupling strengths are approximately the same, but with the former hav- e2 3 ing an additional (small) time-dependent variation, i.e., the Rashba coupling is given by α(t) = β + α1eiωt with α1 ≪ β. Then to the first order in α1, we can express the spin current in Eq. (16) as: J z s,i(ω) = ω2 8mβ Si(ω). (17) Thus, the spin current is approximately proportional to the modulation frequency, but is independent of the mod- ulating amplitude α1. IV. GATE-MODULATION INDUCED SPIN CURRENTS Normally, when an electron with its spin aligned, say, vertically, is passed through the Rashba (Dresselhauss) 2DEG system, the spin will rotate about the in-plane Rashba field.18,19 This spin precession is associated with Zitterbewegung-like motion of the electron.1,20 Under the influence of an electric field, the precessional dynamics is modified giving rise to a net spin polarization and hence a spin Hall current. Since the spin precession motion originates from the SOC interaction, this suggests that we induce a net spin polarization by modulating the SOC to control the spin precession without the need to apply any electric field. Thus, we will study the possibility to generate a sustained spin current under the influence of time-modulated SOC but in the absence of any dc electric field. First, we separate the Hamiltonian in Eq. (1) into two parts: time-independent Hamiltonian H0 = p2 2m + α0 (pxσy − pyσx) + β (pyσy − pxσx), and time-dependent Hamiltonian H1 (t) = α1(t) (pxσy − pyσx) + eE.r. The evolution of electron spin state can be found by means of the first-order time-dependent perturbation theory, in which H1(t) is treated as the perturbation. We assume that the modulation is started at t = 0; meanwhile the applied electric field is switched on adiabatically from the past as E = lims→0+ Eest.21 If the initial electron spin state is represented by the spinor χ(0) = c+ ψ+(cid:11) + c− ψ−(cid:11), then its subsequent evolution up to time t under the full Hamiltonian is given by χ(t)i = Xi=± ci(t)e− i Eit ψii, (18) in which ψ±(cid:11) and E± are eigen-states and eigen-energies of the unperturbed Hamiltonian. To the first order in both the dynamic Rashba component α1 and the elec- tric field, the time-dependent coefficients ci(t) satisfy the following differential equations: i dci dt = Xj=± hψi H1 (t) ψji cj (0) e i (Ei−Ej )t. (19) The above can be readily solved once the initial condi- tions are specified, and subsequently the spin polarization S(t) = hχ(t)σχ(t)i can be evaluated. m c A y , z s J Ñ e 2 0.03 0.02 0.01 0.00 0.03 0.02 0.01 0.00 m c A y , z s J Ñ e 2 à à à à à à à à à à à à à à à à à à à à à 0.0 0.5 1.0 1.5 2.0 2.5 ΩWF aeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeaeae ììììììììììììììììììììììììììììììììììììììììììììììììììì ae à à à aeaeae à aeaeae aeaeae ae ae ae à ae ae à à ae à à à à à à à -0.01 -0.02 à à à à à à à à à à à à à à à à ae Ω=1012s-1 à Ω=1013s-1 ì Ω=1014s-1 0.0 0.2 0.4 Ω t 0.6 0.8 1.0 FIG. 2. Spin current generated in the system with Rashba coupling α = α0 + α1 sin ωt. The inset shows the dependence of the spin current amplitude on the modulation frequency ω obtained numerically (solid). For comparison, we plot the analytical results for the maximum amplitude given by Eq. (23) (horizontal dashed) and the off-resonant amplitude given by Eq. (24) (bold dashed). The maximum amplitude occurs at ω ≈ ΩF . We first consider the system of electrons, which are ini- tially in an admixture of the two eigen-states. The initial spin polarization vector is ±S(0) = (∓ sin γ,± cos γ, 0), corresponding to the two eigen-branches. To the first or- der in H1, the spin polarization vector at a subsequent time t is easily found to be (20) (21) S(t) = ±[S(0) + S(1) + S(2)], where e Ω (E.∇pγ)z, S(1) = − sin (γ − θ)Z t 2p 0 S(2)(t) = dt′α1(t′)R(t − t′), (22) with R(t) = (− cos γ sin Ωt, sin γ sin Ωt, cos Ωt) being the response functions and Ω = 2∆E/. The two terms S(1) and S(2) are the contributions to the spin polarization due to the applied electric field and time modulation of the Rashba coupling, respectively. To the first order in both E and α1, these two factors can be considered as independent driving forces which induce the spin polar- ization. We focus on S(2), the spin polarization arising from the time-modulation in the Rashba SOC. It is in- teresting to point out that in the absence of Dresselhauss SOC (i.e., β = 0), the phase factor is simply γ = θ, and therefore, from Eq. (22), the spin polarization S(2) be- comes zero. Hence, the time-modulation of the Rashba SOC will only induce a spin polarization if both types of SOCs, i.e., Rashba and Dresselhauss, are present in the system. In addition, we find that S(2) is an odd function of the momentum vector k, so that summing over all oc- cupied states in momentum space would yield a zero net z-spin polarization. However, we will show below that the spin current may still be finite. As discussed in previous papers,14,15 the modification to the electron spin due to an applied electric field can 4 s,i = (/2)(pi/m)S(1) s,i = (/2)(pi/m)S(2) give rise to a spin Hall current (even if the net spin po- larization is zero). This can be readily seen by evaluat- ing the expression J z z , which would recover the result of Eq. (15). We apply the same ex- pression to evaluate the spin current arising from the spin polarization S(2) arising from the time-modulation of the Rashba SOC, i.e., J z z , and average the result over the momentum space to yield the net spin current. Fig. 2 illustrates the ac spin current generated in the system when the Rashba coupling is modulated by a sinusoidal gate voltage, i.e. α = α0 + α1 sin ωt, assum- ing typical parameter values: α = β = 10−11eV.m, α1 = 0.1α0, EF = 0.1eV, m = 0.05me. Note that when the gate voltage is changed, the Fermi energy and ef- fective mass will also be modified as a consequence12. However, for simplicity, we assume a small change in the gate voltage, so that the Fermi energy and effective mass can be treated as constant. We see that the amplitude of the spin current is maximum around the resonant re- 0 + β2 the Larmor frequency of electron near the Fermi surface (see inset of Fig. 2). For the special case of α0 = β, this maximum amplitude can be estimated as gion ω ≈ ΩF , where ΩF = 2kFpα2 J z s,i(ΩF ) ∼ a1 ΩF 4mE2 F 3π2 r Eso EF = a1p2 F 6π , (23) while the off-resonant amplitude is given by J z s,i(ω) ∼ a1ω ω2 − Ω2 F 8mE2 F 3π22r Eso EF . (24) In the above, we have assumed that Eso/EF ≪ 1, where Eso = mα2 being the energy scale associated with the spin-orbit couplings, EF = p2 F /2m the Fermi energy. The analytical results of Eqs. (23) and (24) are in approxi- mate agreement with the numerical results (see inset of Fig. 2). V. CONCLUSION In this paper, we have applied the spin force picture to a general time-dependent SOC system. By considering the Lorentz force arising from the non-Abelian gauge field of the SOC effects we establish the relation between the spin current and spin force in the system. Previously the spin force in a Rashba system was invoked to explain the Zitterbewegung motion of electrons, but no link was made to the spin Hall effect. We applied our gauge field method to show that the longitudinal component of the spin force induces the spin Hall effect. This not only provides a physical picture of the underlying origin of the spin Hall effect, but also recovers the universal spin Hall conductivity in a constant (time invariant) Rashba system. We also showed that the spin current vanishes under strong impurity scattering, a result consistent with previous findings. However, under time modulation of the Rashba SOC a sustained spin current is obtained, the amplitude of which is proportional to the in-plane spin density and the modulation frequency. Finally, we evaluated the ac spin current generated by a Rashba SOC with a sinusoidal time variation in the absence of any dc electric field. For the special case of equal Rashba and Dresselhaus coupling, we derived the analytical expression for the magnitude of the spin current, which approximately agrees with the numerical results. 5 ACKNOWLEDGMENTS We gratefully acknowledge the SERC Grant No. 092 101 0060 (R-398006105) for financial support. 1 S. Q. Shen, Phys. Rev. Lett. 95, 187203 (2005). 2 T. Fujita, M. B. A. Jalil, , and S. G. Tan, New J. Phys. 12, 013016 (2010). 3 Y. Aharonov and A. Stern, Phys. Rev. Lett. 69, 3593 (1992). 4 S. G. Tan, M. B. A. Jalil, X.-J. Liu, and T. Fujita, Phys. Rev. B 78, 245321 (2008). 5 Q. F. Sun, H. Guo, and J. Wang, Phys. Rev. Lett. 90, 258301 (2003). 6 E. Y. Sherman, A. Najmaie, and J. E. Sipe, Appl. Phys. 10 S. Zhang and W. Zhu, J Phys-Condens Mat 25, 075302 (2013). 11 S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990). 12 J. Nitta, T. Akazaki, H. Takayanagi, and T. Enoki, Phys. Rev. Lett. 78, 1335 (1997). 13 D. Grundler, Phys. Rev. Lett. 84, 6074 (2000). 14 J. Sinova, D. Culcer, Q. Niu, N. A. Sinitsyn, T. Jungwirth, and A. H. MacDonald, Phys. Rev. Lett 92, 126603 (2004). 15 S. Q. Shen, Phys. Rev. B 70, 081311 (2004). 16 J. I. Inoue, G. E. W. Bauer, and L. W. Molenkamp, Phys. Lett. 86, 122103 (2005). Rev. B. 70, 041303 (2004). 7 A. G. Mal'shukov, C. S. Tang, C. S. Chu, , and K. A. 17 I. Adagideli and G. E. W. Bauer, Phys. Rev. Lett. 95, Chao, Phys. Rev. B 68, 233307 (2003). 256602 (2005). 8 C. S. Tang, A. G. Mal'shukov, and K. A. Chao, Phys. 18 Y. A. Bychkov and E. I. Rashba, J. Phys. C 17, 6039 Rev. B 71, 195314 (2005). (1984). 9 F. Liang, Y. H. Yang, and J. Wang, Eur. Phys. J. B 69, 19 G. Lommer, F. Malcher, and U. Rossler, Phys. Rev. Lett. 337 (2009). 60, 728 (1988). 20 T. Biswas and T. K. Ghosh, J. Phys.: Condens. Matter 24, 185304 (2012). 21 W. Kohn and J. M. Luttinger, Phys. Rev. 108, 590 (1957).
1301.1943	2	1301	2013-04-29T09:50:57	Finite-gap twists of carbon nanotubes and an emergent hidden supersymmetry	[ "cond-mat.mes-hall", "hep-th", "math-ph", "math-ph" ]	We consider radially twisted nanotubes in the low-energy approximation where the dynamics is governed by a one-dimensional Dirac equation. The mechanical deformation of the nanotubes is reflected by the presence of an effective vector potential. We discuss twisted carbon and boron-nitride nanotubes, where deformations give rise to periodic and nonperiodic finite-gap Hamiltonians. The intimate relation of these systems with the integrable Ablowitz-Kaup-Newell-Segur hierarchy is exploited in the study of their spectral properties as well as in the computation of the (local) density of states. We show that a nonlinear hidden supersymmetry generated by local supercharges arises naturally in the finite-gap configurations of twisted nanotubes with time-reversal symmetry. The properties of the van Hove singularities are encoded in its structure.	cond-mat.mes-hall	cond-mat	Finite-gap twists of carbon nanotubes and an emergent hidden supersymmetry Francisco Correa1, V´ıt Jakubsk´y2 1Centro de Estudios Cient´ıficos (CECs), Avenida Arturo Prat 514, Valdivia, Chile 2Nuclear Physics Institute, Rez near Prague, 25068, Czech Republic We consider radially twisted nanotubes in the low-energy approximation where the dynamics is governed by a one-dimensional Dirac equation. The mechanical deformation of the nanotubes is reflected by the presence of an effective vector potential. We discuss twisted carbon and boron- nitride nanotubes, where deformations give rise to periodic and nonperiodic finite-gap Hamiltonians. The intimate relation of these systems with the integrable Ablowitz-Kaup-Newell-Segur hierarchy is exploited in the study of their spectral properties as well as in the computation of the (local) density of states. We show that a nonlinear hidden supersymmetry generated by local supercharges arises naturally in the finite-gap configurations of twisted nanotubes with time-reversal symmetry. The properties of the van Hove singularities are encoded in its structure. I. INTRODUCTION Since their discovery in 1991 [1], carbon nanotubes have attracted massive attention from both experimental and theoretical physicists. Their remarkable mechanical and electronic properties, extreme mechanical strength [2], and elasticity [3] as well as conductivity [4] make their use in future electronic devices very promising [5]. Carbon nanotubes are also relevant as a low-dimensional test field of fundamental physical theories. For instance, the Klein tunneling, a well known phenomenon in rel- ativistic quantum theory [6], has not been observed for elementary particles so far. However, it was predicted and observed in carbon nanotubes where it stays behind the absence of backscattering on impurities [7]. The single-wall carbon nanotubes are small cylinders with atom-thick shells, that can be created by rolling up graphene nanoribbons [8]. Despite their structural simplicity (as quasi one-dimensional objects), the nan- otubes exhibit a remarkable variety of electronic prop- erties. They can be either metallic or semiconducting, depending on the orientation of the hexagonal lattice in the nanotube. The gap between the valence and conduc- tion band can be regulated by either external fields [9] or by mechanical deformations [10]. This paves the way to strain engineering [11] where graphene-based devices would be fine-tuned by deformations. In the present paper, we will consider a class of ex- actly solvable models of single-wall carbon nanotubes subject to radial twist (axial torsion). The systems will be studied in the low-energy regime where the dynam- ics of charge carriers is well approximated by a (1 + 1)- dimensional Dirac equation [12]. In particular, the twist configurations under investigation will be described by a finite-gap Hamiltonian. Besides the analysis of the spectral properties, the computation of the local density of states (LDOS) of the systems will be addressed. This quantity is measurable by means of scanning tunneling microscopy experiments [13] and is very important for the specification of the electronic properties of nanostructures. When integrated spatially, it provides the density of states (DOS) that re- flects the probability of inserting an electron at given energy into the system. In general, it is a rather complicated task to analyti- cally compute the local density of states. However, as it was suggested in [14] and demonstrated later in [15], the computation can be considerably simplified for a broad class of Dirac and Bogoliubov - de Gennes Hamiltoni- ans that are characterized by a finite number of gaps in the spectrum. This approach is based on the intrinsic properties of the finite-gap systems that arise from the stationary AKNS (Ablowitz-Kaup-Newell-Segur) hierar- chies of integrable systems [16]. In the next section, we will set up the theoretical frame- work in which the twisted nanotubes will be analyzed. It will be explained how the Dirac Hamiltonian emerges in the description of twisted carbon (and boron-nitride) nanotubes. In Sec. III, the relevant points of the con- struction of the AKNS hierarchies are summarized and the closed formula for the LDOS is discussed. Section IV is devoted to the presentation of explicit examples of car- bon and boron-nitride nanotubes where the LDOS and DOS will be computed explicitly. We will show in Sec. V that the singular points of DOS, so-called van Hove singularities, are closely related to the nonlinear super- symmetry that arises naturally for the nanotubes with finite-gap twists and conserved time-reversal symmetry. The last section is left for the discussion and outlook. II. THE MODEL Carbon atoms have four valence electrons; three of them are tightly bound in the interatomic bonds while the fourth one is free and can contribute to the electronic properties of the crystal. The properties of the collec- tive excitations of these electrons in graphene are well described by the tight-binding Hamiltonian [17]. The interactions between the nearest neighbors are only as- sumed, being specified by a constant hopping parameter. The spin degree of freedom of electrons can be neglected; it is irrelevant in the considered interactions. The analysis of the band structure displays the spe- cific feature of graphene: the Fermi surface is formed by six discrete points where valence and conduction bands meet. They are located at the corners of the hexago- nal first Brillouin zone and are called Dirac points. Only two of them are inequivalent1 and correspond to different electronic states. Let us denote them as K and K′ ≡ −K. In the close vicinity of the Dirac points, the energy sur- face acquires conelike shape. It suggests that the dis- persion relation is linear in this region. Indeed, taking k = ±K + δk with δk ∼ 0 and expanding the tight- binding Hamiltonian up to the terms linear in δk, we get the stationary equation for the two-dimensional massless Dirac particle [12]. The Hamiltonian acquires the same form in both valleys of K and K′. In the coordinate representation, we have 2 h(±K)ψ±K = (−iσ2∂x + iσ1∂y)ψ±K = λψ±K . (1) The spinorial degree of freedom in (1), the pseudospin, arises due to the two carbon atoms in the elementary cell; the hexagonal lattice can be thought of as assembled from two triangular lattices. The spin-up or down components of the wave functions are non-vanishing only on one of the two triangular sublattices. The operator h(K) acts on the spinors (ψKA, ψKB)t, while h(K′) on the spinors (ψK′B, ψK′A)t. Here, the first index denotes the valley, the second distinguishes between the sublattices A and B and t denotes transposition. The formula (1) was introduced already in 1984 by Se- menoff [12] and makes the basis for considering the con- densed matter system as a convenient test field for a low- dimensional quantum field theory. Indeed, it makes it possible to observe phenomena in this condensed matter system that are native in high-energy quantum physics, see e.g. [18], [19]. A single-wall carbon nanotube is rolled up from a straight graphene strip. The actual orientation of the hexagonal lattice in the strip is uniquely determined by the chiral (circumference) vector Ch, which is a linear combination of the translation vectors of the lattice [20]. Its length corresponds to the diameter of the nanotube. We can fix the coordinates such that y goes in the circum- ference direction. Then the chiral vector gets the simple form, Ch = (0, Ch). The effect of rolling up the strip is reflected by the periodic boundary condition imposed on the wave func- tions, ψK(x, y + Ch) = ψK(x, y). It leads to the quanti- zation of the momentum in the circumference direction which acquires discrete values ky. In the low-energy ap- proximation, only the value of ky that minimizes the en- ergy is relevant. The system is then governed by a truly 1 The remaining four Dirac points can be reached by translational vectors of the reciprocal lattice, and, hence, do not represent distinct electronic states. 2 We set = e = 1. λ = E vF where E is energy and vF the Fermi velocity of the quasi-particle. 2 one-dimensional Hamiltonian −iσ2∂x + σ1ky. The actual value of this fixed ky depends on the character of the nanotube. Instead of going into more details that can be found, for instance in Ref. [20], let us notice that ky = 0 corresponds to metallic nanotubes as there is no gap in the spectrum. When ky 6= 0, there is a small gap in the spectrum and the nanotube is semiconducting. For purposes of our current analysis, we can suppose that the nanotubes are metallic (i.e. the angular momentum is vanishing, ky = 0) and are infinitely long. The lat- ter approximation is rather reasonable due to the recent experiments where ultralong single-wall nanotubes were created [21]. Up to now, we considered systems where neither exter- nal fields nor any strains were present. By deforming the crystal mechanically, the interatomic distances in the lat- tice are modified. Thus, the hopping parameter ceases to be constant and becomes position dependent. This leads to the appearance of gauge fields in the tight-binding Hamiltonian. It can be approximated in the low-energy limit by Dirac operator with nonvanishing vector poten- tial [10, 22, 23]. We shall consider radial twist of the nanotubes. Let us mention that both single-wall and multi-wall carbon nanotubes with radial twist were used in construction of nanoelectromechanical devices, e.g. single-molecule tor- sional pendulum [24], abacus-type resonators [25] or even rotors [26]. In these systems, the nanotube served as the torsional string that was twisted by deflection of small paddles attached to it; see also [27] for a brief review. The radial twist shifts the atoms in the lattice perpen- dicularly to the axis, preserving the tubular shape of the nanotube. The displacement is reflected by a deforma- tion vector, which measures the difference between actual and equilibrium position of atoms. It can be written in our specific case as d = (0, dy(x))t. We consider the situ- ation where the displacement is smooth and small on the scale of the interatomic distance. Then the interaction does not mix the valleys of K and K′, and the system can be studied in the vicinity of one Dirac point only. The stationary equation for low energy Dirac fermions in the K-valley acquires the following simple form [10], [28], h(K)φ = (−iσ2∂x + Σ(x)σ1)φ = λφ , (2) where the vector potential Σ(x) reflects the twist. It is re- lated to the displacement vector by d = ζ (0,R Σ(x)dx), where ζ is a constant dependent on the crystal charac- teristics.3 In this framework, the constant vector poten- tial Σ(x) = γ > 0 would reflect a linear displacement d = ζ (0, γx). Finally, let us consider the following generalization of 3 ζ = (cid:16)− t a the lattice constant. See [23] for more details. ∂ ln t ∂ ln a(cid:17)−1 where t is the hopping parameter and a is (2), where a mass term is included, h(K)φ = (−iσ2∂x + Σ(x)σ1 + M σ3)φ = λφ . (3) The analogue of (1) with the mass term M was pro- posed by Semenoff for description of the quasiparticles in the boron-nitride crystal in the low-energy approxima- tion [12]. The boron-nitride crystal has the same hexag- onal structure as graphene. However, the atoms in the elementary cell of the crystal cease to be equivalent. It gives rise to the potential term with σ3 that distinguishes between the two triangular sublattices A and B. We will consider (3) as the Hamiltonian of radially twisted carbon (M = 0) or boron-nitride (M 6= 0), de- pending on the value of M , nanotubes. Boron-nitride nanotubes were studied theoretically and observed ex- perimentally, see e.g. [29], [30]. Contrary to the carbon nanotubes, they are always semiconducting. We shall consider the scenario where the term Σ(x)σ1+ M σ3 in (3) belongs to the broad class of the finite-gap potentials. In the next section, we will show how the peculiar properties of finite-gap systems can be utilized for computation of the local density of states. III. FINITE-GAP TWISTS AND THE LDOS VIA AKNS HIERARCHY We review here some properties of the integrable ANKS hierarchies associated with the Dirac Hamiltonian (3). To explain more easily the main features, we use the following unitarily transformed Hamiltonian, σ1π h = exp(cid:16)−i = (cid:18) −i∂x ∆(x) 4 (cid:17) h(K) exp(cid:16)i i∂x (cid:19) , ∆(x)∗ σ1π 4 (cid:17) (4) where ∆ = Σ(x) + iM . This form with diagonal deriva- tive term, the Bogoliubov-de Gennes type Hamiltonian, is used frequently in the analysis of Gross-Neveu and Nambu-Jona-Lasinio models [14], [31] and will make the presentation more coherent with the specialized litera- ture [16]. The vector potential ∆(x) = ∆ in (4) is called finite gap (or algebro-geometric in the mathematical literature) when it solves one of the equations from the stationary AKNS hierarchy of the nonlinear differential equations, namely AKNSN . One of the most intriguing properties of the Hamiltonian (4) with a finite gap potential is man- ifested in its spectrum; it consists of a finite number of bands [16], [32]. The actual number of bands (or gaps) is fixed by the AKNSN equation solved by ∆. The val- ues of band-edge energies of a finite-gap system can be obtained in purely algebraic manner; see [16]. These fea- tures are intimately related with the existence of an in- tegral of motion of the Hamiltonian (4). Nonperiodic finite-gap systems can be obtained as the infinite-period limit of the periodic ones. In this context, the nonperi- odic systems are known as kink or kink antikink models 3 in analogy with the soliton solutions in the Korteweg-de Vries (KdV) hierarchy. Another relevant feature of this class of models is that they can approximate very well any condensed matter systems described, in the low-energy approximation, by the Hamiltonian (2) with a generic periodic potential. The Hamiltonian with a generic periodic potential has an infinite number of spectral gaps, the width of which decreases rapidly with the increasing absolute value of energy. Hence, the spectrum of such system can be fitted well by a finite- gap one. The stationary AKNS hierarchy of nonlinear differen- tial equations can be constructed in terms of a Lax pair which consists of the Hamiltonian h and a matrix differ- ential operator SN +1, defined as SN +1 = i N +1 Xl=0 (cid:18) gN +1−l f ∗ N −l fN −l gN +1−l (cid:19) σ3hl , N ∈ N . (5) The functions fn(x) and gn(x) are defined, recursively, in the following manner, i 2 f ′ n−1 + ∆ gn , fn = − g′ n = i(cid:0)∆∗fn−1 − ∆ f ∗ g0 = 1, f−1 = 0 . n−1(cid:1) , (6) (7) (8) The functions fn and gn depend on ∆(x) and its deriva- tives and also contain integration constants that appear due to the integration of (7); see Ref. [15] for details. The operator (5) satisfies the following commutation relation for any positive integer N , [ SN +1, h] = 2i(cid:18) 0 −f ∗ N +1 fN +1 0 (cid:19) , (9) The stationary AKNS hierarchy of nonlinear differential equations is then defined in terms of the vanishing com- mutator (9), AKNSN = fN +1 = 0 . (10) The Hamiltonian h and the operator SN +1 are called the Lax pair of the stationary AKNS hierarchy. When a function ∆ satisfies the (N + 1)th order differential equa- tion AKNSN , all the next equations of the hierarchy with greater values than N (and with the integration constants fixed appropriately4) are immediately solved. The operators h and SN +1 satisfy the remarkable al- gebraic relation, S2 N +1 = 2N +1 Yn=0 (h − λn) , (11) 4 The AKNSN can be written as a linear combination PN +1 l=0 cl fl where the functions fl are defined like in (8) but with all the integration constants that emerge in (7) fixed to zero. where λn are band-edge energies. The operator valued function on the righthand side is known as the spectral polynomial. The integral SN +1 annihilates all the sin- glet eigenstates φn of h, corresponding to the band-edge energies, (h − λn) φn = 0, SN +1 φn = 0, n = 0, 1, ..., 2N + 1 . (12) The local density of states ρ(x, λ) is defined in terms of the trace of the Green's function, R(x, λ) ≡ G(x, x, λ), (13) Im T r R(x, λ) , lim Imλ→0+ ρ(x, λ) = − 1 π where the trace is computed over matrix degrees of free- dom. The function R(x, λ) is also called diagonal resol- vent or Gorkov resolvent. The spatial integration of LDOS leads to the formula for DOS. In case of periodic quantum systems, the inte- gration can be performed over one period L 5, DOS = 1 L ZL ρ(x, λ) dx . (14) Explicit calculation of the Green's function can be quite difficult. Nevertheless, the definition (13) suggests that the need of its explicit knowledge might be avoided; only the diagonal resolvent is required to find LDOS. This fact was utilized in [14] and further developed in [15]. In- deed, an exact form of the diagonal resolvent was found for a wide class of Hamiltonians (4). The approach was based on the fact that R(x, λ) has to satisfy the Dikii- Eilenberger equation [33], ∂ ∂x R(x; λ) σ3 − i (cid:20)(cid:18) λ ∆∗(x) −λ (cid:19) , R(x; λ) σ3(cid:21) = 0 , −∆(x) (15) where λ belongs to the spectrum of h. Additionally, R(x, λ) has to satisfy the following requirements, R = R†, det R = − 1 4 , (16) where the latter one fixes the normalization of R. For more details on the properties of R and derivation of (16), see e.g. the Appendix in [34]. Making the following ansatz for the diagonal resolvent [15], R(x; λ) = N +1 Xn=0 βn(λ) (cid:18) gn(x) f ∗ n−1(x) fn−1(x) gn(x) (cid:19) , (17) and substituting (17) into (15), the Dikii-Eilenberger equation transforms into the two (mutually conjugated) 4 nonlinear differentials equations of the form of the AKNS hierarchy. The diagonal entries in (15) vanish identically due to the recurrence relations (8). The resulting equa- tion can be written as N +1 Xn=0 βn(λ)fn − λ N +1 Xn=0 βn(λ)fn−1 = 0 , (18) which can be solved by fixing properly the constants βn(λ), see footnote 4 and Ref. It can be shown that the ansatz (17) fulfills the requirements (16). 6 [15]. Making the inverse transformation (4), we can find the Lax operator associated with the finite-gap Hamiltonian (3) as SN +1 = exp(cid:16)i = −i N +1 Xl=0 σ1π 4 (cid:17) SN +1 exp(cid:16)−i σ1π 4 (cid:17) (gN +1−l1 + σ3ImfN −l + σ1RefN −l) σ2hl (19) . The diagonal resolvent for the Hamiltonian h can be ob- tained directly from (3), since the trace of an operator is invariant with respect to similarity transformations. IV. EXACTLY SOLVABLE MODELS OF THE TWISTED NANOTUBES The periodic systems described by (2) can be classified in terms of a quantity which we call average twist. It is defined as Σc = max(Σ) + min(Σ) 2 , (20) and corresponds to the value around which the potential is oscillating. We will present two- and four-gap sys- tems, denoted as Σ(x) = ∆2 and Σ(x) = ∆4, respec- tively, where the average twist is vanishing. Then we will consider two simple cases where it acquires nonzero values. They correspond to the one-, Σ(x) = ∆1, and three-gap Σ(x) = ∆3 systems. The mass term will be identically zero in all these models, M = 0. We will see that the actual value of the average twist is in correlation with the qualitative spectral properties of these models. As the last example, we will consider a nonperiodic system with a constant mass, M 6= 0. It will serve for illustration of a twisted boron-nitride nanotube. 5 For the nonperiodic settings, the spatial integration can be di- vergent. 6 The ansatz (17) is manifestly hermitian. Additionally, it also Indeed, one can check satisfies the second condition in (16). directly that (det R(x, λ))′ = 0 with the use of (18). reads Tr R2(x; λ) = λ2 + 1 2 ∆2 2 p(m2 − λ2)(λ2 + m2(k2 − 1)) 5 , (25) and the associated density of states acquires the following form FIG. 1: Illustration of the two-gap (upper), three-gap (mid- dle) and four-gap (lower) configurations of the twisted carbon nanotubes. In the system without twist, the black line would be straight. A. Configurations with zero average twist λ2 − m2 E(k) K(k) DOS2 = − 1 π lim Imλ→0+ Im p(m2 − λ2)(λ2 + m2(k2 − 1)) (26) where we have used Eqs. (13) and (14). Notice that DOS2 is identically zero when λ belongs to the prohib- ited gaps. The function in the argument is purely real for these values of λ and, thus, the imaginary part is vanishing identically, see Fig. 2. , 1. Two-gap system DOS ∆2 First, let us consider the system governed by (2) with the vector potential ∆2 = mk2 sn mx cn mx dn mx , (21) where m is a real parameter and k ∈ (0, 1). This vector potential is induced by the deformation specified by the following displacement vector (see Fig. 1), d = (0,−ζ ln dn mx) . (22) The crystal kink two-gap potential (21) is given in terms of doubly periodic Jacobi elliptic functions depend- ing on the modular parameter k. It has a real period L = 2K(k), where K(k) is the complete elliptic integral of the first kind. For the definitions and properties of the elliptic functions, we recommend Refs. [35, 36]. The infinite-period limit (k → 1) of (21) is called the single kink vector potential ∆2 = tanh x. Let us notice that the properties of the Dirac electron in graphene in the pres- ence of a single-kink-type vector potential were analyzed in [37]. The potential ∆2 vanishes in the limit when the modular parameter goes to zero. The spectrum of the one-dimensional Dirac Hamilto- nian h(K) has two gaps located symmetrically with re- spect to zero. The band-edge energies are λ0 = −λ3 = −m and λ1 = −λ2 = −m√1 − k2. The corresponding eigenstates ((h(K) − λn)φn = 0, n = 0, 1, 2, 3) are φ0 =(cid:16)−sn mx, φ1 =(cid:18) √1 − k2 dn mx(cid:19)t , φ2 = σ3φ1 . cn mx dn mx(cid:17)t φ3 = σ3φ0 , cn mx, sn mx 1 (23) (24) , 0:8 0:6 0:4 0:2 −3 −2 −1 0 1 2 3 λ FIG. 2: Illustration of DOS (26) of the Hamiltonian (3) with Σ(x) = ∆2 and M = 0 with m = 1.5 and k = 0.7. 2. Four-gap system As the next example, we shall consider the 2K(k)- periodic system described by the Hamiltonian (2) with the vector potential ∆4 = 6mk2 sn mx cn mx dn mx 1 + k2 + δ − 3k2sn2 mx , (27) where δ = √1 − k2 + k4. The crystal kink four-gap po- tential (27) is an isospectral deformation of the crystal kink potential ∆4′ = 2mk2 sn mx cn mx . Both potentials reduce to the single kink ∆4 = ∆4′ = 2m tanh mx when k → 1. The associated displacement vector in this case takes the form dn mx The band-edge energies are nondegenerate, while the en- ergies from the interior of the bands are doubly degener- ated. Using directly the formula (17) for N = 2, we can find the explicit form of the diagonal resolvent. Its trace then d = (0,−ζ ln(1 + k2 + δ − 3k2sn2 mx)) (28) and is illustrated in Fig. 1. The spectrum of h(K) has five bands and eight band-edge states φn, n = 0, ..., 7, which can be defined with help of and an operator D = dx + ∆4 as follows, 7 d Let us note that for the system described by the isospectral potential ∆4′ , the resolvent trace 6 φn = (cid:18)ψn, 1 λnDψn(cid:19)t , (h(K) − λn)φn = 0. (29) Keeping in the mind the spectral symmetry λ ↔ −λ (which is valid for any model (3) with M = 0), it is suffi- cient to find just the first four eigenstates φ0, φ1, φ2, φ3, since the remaining four can be obtained as φn+4 = σ3φn, where n = 0, 1, 2, 3. They are given in terms of the fol- lowing functions ψ0 = m(cid:0)1 + k2 − δ − 3k2sn2 mx(cid:1) , λ0 = −2m√δ , ψ1 = cn x sn x, ψ2 = dn x sn x, ψ3 = cn x dn x, λ1 = −mp2 − k2 + 2δ , λ2 = −mp2k2 − 1 + 2δ , λ3 = −mp2δ − 1 − k2 , where λn are the corresponding eigenvalues. (30) (31) (32) (33) Tr R4′ (x; λ) = η1 + η2∆2 8p(λ0 − λ2)(λ2 − λ2 4′ + η3∆4 4′ 1)(λ2 − λ2 2)(λ2 − λ2 3) (37) can be written just in terms of a polynomial in ∆4′, where η1, η2 and η1 are constant depending on λ. B. Semiconducting carbon nanotubes via nonzero average twist 1. One-gap system The simplest example of a finite-gap system with the nonzero average twist is given by the Hamiltonian (2) with the constant vector potential Σ(x) = γ. The two band-edge energies correspond to λ0 = −λ1 = −γ. The local density of states can be found in the following form The local density of states can be computed using the method described in the preceding section, Tr R4 = . (38) Tr R4(x; λ) = α1 + α2∆2 4 + 3∆4 8p(λ0 − λ2)(λ2 − λ2 4 + ∆′2 1)(λ2 − λ2 4 − 2∆4∆′′ 4 , 2)(λ2 − λ2 3) (34) where the constants are α1 = 8(λ4 − 5m2δλ2 + 4m4δ), α2 = 4(δ2 − 5m2δ2) . (35) (36) The explicit (analytical) form of the density of states is rather cumbersome. In Fig. 3, we present the numerical computed DOS of the current four-gap system. DOS ∆4 0:8 0:6 0:4 0:2 −2 −1 0 1 2 λ FIG. 3: Σ(x) = ∆4 and M = 0 with m = 1 and k = 0.6. Illustration of DOS of the Hamiltonian (3) with 7 This way to express the eigenfunctions is just the essence of usual supersymmetric quantum mechanics applied for finite-gap poten- tials. To avoid the details here, we refer to [38, 39]. λ pγ2 − λ2 The constant potential can be regarded as periodic with the period being equal to any real number L. We can compute the average twist as Σc = γ. The spectrum of the system has two bands separated by a gap of width 2Σc. This suggests that the central gap is twice the average twist. 2. Three-gap system Let us test the suggestion in the case of a more compli- cated system. Its Hamiltonian (2) has the 2K(k)-periodic vector potential ∆3 = cn b dn b sn b + k2 sn b sn (x) sn (x + b) , (39) which is called the crystal kink-antikink, three-gap po- tential, [15]. The real parameter b ∈ (0, K(k)) represents the distance between the kink and the antikink. The vector potential is induced by the displacement d = (0, ζ F (x)), where F (x) is as follows, cn b dn b sn b F (x) = Π(k2 sn2b, am xk)− ln(cid:0)1−k2 sn2b sn2x(cid:1) . The function Π(a; xφ) is the incomplete elliptic integral of the third kind and am x is the Jacobi amplitude. See Fig. 1 for illustration. 1 2 When b = K(k), (39) is reduced to the two-gap vector potential (21). In the infinite period limit, the single kink-antikink solution is recovered [40], limk→1 ∆1(x) = coth b + tanh x − tanh(x + b). The spectrum of (2) with (39) contains three gaps po- sitioned symmetrically with respect to zero. The three band-edge states with negative energies are DOS ∆3 7 φ0 = (−sn (x), sn (x + b))t , φ1 = (−cn (x), cn (x + b))t , φ2 = (−dn (x), dn (x + b))t , 1 , λ0 = − λ1 = − λ2 = − sn b dn b sn b cn b sn b , . (40) The positive energy states are obtained as φn+3 = σ3φn and correspond to the energies λn+3 = −λn, where n = 0, 1, 2. The trace of diagonal resolvent can be computed from (17) for N = 2 in the following form 0:6 0:2 −2 −1 0 1 λ 2 Tr R3(x; λ) = λ 2 (cid:0)α + ∆2 3(cid:1) , (41) FIG. 4: Σ(x) = ∆3 and M = 0 with k = 0.2 and b = 1.5. Illustration of DOS of the Hamiltonian (3) with p(λ2 0 − λ2)(λ2 1 − λ2)(λ2 2 − λ2) where α = 1 + k2 + 2λ2 − 3 sn2 b . The actual integration of the formula above, needed for analytical form of DOS, is rather complicated. We present Fig. 4 of DOS for the three-gap case that was obtained numerically. The average twist associated with the potential (39) can be found as Σc = cn b sn b . (42) Checking the corresponding band-edge energies λ2 and λ3 in (40), we can see that that the gap between the positive and negative energies is exactly of width 2Σc. Comparing the spectra of the presented systems, we can see that the nonvanishing average twist (20) is pro- portional to the magnitude of the central spectral gap in the system. In the two- and four-gap systems, the aver- age twist is vanishing and there is no gap between positive and negative energies. These nanotubes are conducting in the sense that infinitesimal excitation is sufficient to kick the electrons from valence band to conduction band. The systems with the nonvanishing average twist are dif- ferent. They have a gap between positive and negative energies that are equal to 2Σc and, hence, are semicon- ducting. C. Boron-nitride nanotubes As the last example, we shall consider a nonperiodic system with the nontrivial mass term. We take the po- tential term of (3) in the following form, Σ(x) = N sin θ 2 tanh(sin θ 2 x), M = −N cos θ 2 , (43) where N is a positive integer and θ is a real parameter. The potential is classified as (N + 1)-gap as it solves the corresponding equation of the AKNS. It has N + 2 singlet states in the spectrum. Two of them correspond to the energies that form the threshold of the continuum spectrum, the rest is associated with bound states of the system. In the case of N = 1, the eigenstates are then given as ψ0 = (cid:18)tanh(cid:18)sin ψ1 = (cid:18)sech (cid:18)sin θ 2 θ 2 x(cid:19) ,− cot x(cid:19) , 0(cid:19)t , θ 4(cid:19)t , λ0 = −1, (44) λ1 = cos θ 2 , ψ2 = (cid:18)tanh(cid:18)sin θ 2 x(cid:19) , tan θ 4(cid:19)t , λ2 = 1. (45) (46) The trace of the diagonal resolvent can be computed in the following manner, Tr R(x; λ) = λ2 − λ cos θ 2 sin2 θ 2 − 1 √1 − λ2(λ − cos θ 2 sech2 (cid:0)sin θ 2 x(cid:1) 2 ) . (47) It is worth noticing that in the examples of the car- bon nanotubes, the trace of diagonal resolvent (25), (34), (37), and (41) could be written in terms of the finite-gap potential and its derivatives. A similar result is known for the Schrodinger systems with Lam´e potential. The trace of the diagonal resolvent corresponded in that case to a polynomial of the finite-gap potential [41]. This is related to the fact that the square of the Dirac opera- tor with the four-gap potential ∆4′ corresponds to an extended Schrodinger operator with two-gap Lam´e po- tential. In the current case with the nonvanishing mass term, the trace of diagonal resolvent can be written as a func- tion of the amplitude of the corresponding complex po- tential ∆ [see (4)], where ∆2 = Σ(x)2 + M 2. For (43), we can write Tr R(x; λ) = P2N (x, θ) √N 2 − λ2Q2N n=1(λ − λn)2 . Here, P2N (x, θ) = N Xn=0 cn(λ)(cid:18)sin2 θ 2 sech2 (sin x)(cid:19)n θ 2 (48) with cn(λ) being specific constants 8. V. VAN HOVE SINGULARITIES AND THE NONLINEAR SUPERSYMMETRY The densities of states have a set of singular points that are called van Hove singularities. A closer inspec- tion of the corresponding formulas (25), (34), (38), (41), and (47) shows that the number as well as the position of the van Hove singularities coincide precisely with the sin- glet band-edge energies of the finite-gap systems. In this section, we will show that this coincidence is reflected by a nonlinear supersymmetry that underlies the finite-gap configurations of the twisted nanotubes. Quantum systems in presence of a magnetic field cease to be time-reversal invariant. The time-reversal opera- tor changes the sign of momentum while it preserves the coordinate. It changes the sign of the magnetic field. It can be represented by an anti-unitary operator T that satisfies T †T = 1, T iT = −i and T 2 = −1. The latter equality arises due to the half-integer spin of the consid- ered particles. One can check that the anti-unitary operator σ2T (T denotes complex conjugation, T 2 = 1) does not commute with the Hamiltonian (3) due to the symmetry break- ing term Σ(x) (and M ). However, we have to keep in mind that these terms arise from the tight-binding model, which, despite the deformations of the crystal, is time- reversal invariant [17]. The time-reversal symmetry of the system in the low- energy regime emerges when dynamics in both valleys corresponding to K and K′ is taken into account. The total Hamiltonian reads H = (cid:18) h(K) 0 0 h(−K) (cid:19) , (49) where the energy operators h(±K) of the subsystems are given as h(±K) = −iσ2∂x ± Σ(x)σ1 ± M σ3 . (50) 8 For N = 2 the band edge energies are λ0 = −λ4 = −2, λ1 = 2 . The constants in (48) are and λ2 = 2 cos θ −λ3 = −q 7+cos θ 2 then given in the following form c0(λ) = λ4 − 2λ3 cos(cid:18) θ 1 2 2(cid:19) − λ(cid:18)15 cos(cid:18) θ 1 2 c1(λ) = 2(cid:18)−λ2 + λ cos(cid:18) θ λ2(cos(θ) + 7)+ 2(cid:19) + cos(cid:18) 3θ 2 (cid:19)(cid:19) 2(cid:19) + cos(θ) + 1(cid:19) , c2 = 9 4 8 The operator (49) acts on the bispinors Ψ = (ψKA, ψKB, ψK′B, ψK′A), where we use the notation in- troduced in the second section below (1). The Hamiltonian H commutes with the time-reversal operator T which is defined in the following manner9, [H,T ] = 0, T = σ1 ⊗ σ2 T. (51) As the considered system consists of a single fermion, the Kramer's theorem applies; all the energy levels of (49) have to be at least doubly degenerate. In case of a periodic system, the band structure of h(K) is deter- mined by 2N + 2 nondegenerate band-edge energies λn. In the infinite period limit, the operator has N +2 singlet states. As we can see from (50), the operators h(±K) are unitarily equivalent, h(K) = σ2h(−K)σ2. Hence, H has the same band structure as h(K), but the degeneracy is doubled as is required by the Kramer's degeneracy theo- rem. Degeneracy of energy levels is reflected by a set of integrals of motion that are based on the Lax integral SN +1, see (19). In the individual subsystems governed by h(±K), the degeneracy is associated with two diago- nal operators, Q0, and Q3, 0 Q0 = (cid:18) SN +1 0 σ2SN +1σ2 (cid:19) , Q3 = τ30Q0, (52) where τ30 = σ3 ⊗ 1. The intervalley (Kramer's) degener- acy is naturally reflected by the operators Q1 and Q2, Q1 = τ12Q0, Q2 = τ22Q0, (53) where τab = σa ⊗ σb, a, b = 1, 2. All these operators commute with the total Hamiltonian, [Qa,H] = 0 . (54) By construction, these operators close Lie algebra so(3)⊕ u(1), a, b, c = 1, 2, 3. [Qa,Qb] = 2iεabcQc, [Q0,Qa] = 0, (55) The existence of the operators (53) is a direct conse- quence of the time-reversal symmetry of (49). Indeed, (51) implies the unitary equivalence of the valley Hamil- tonians h(±K) and enables the construction of antidiag- onal operators (53). The action of the integrals is quite nontrivial and de- termined by the properties of the Lax operator SN +1. It can be inferred from (12) that all doublet states Ψ2−deg 9 In [19], the real spin of electrons in taken into account. There, the time-reversal operator is defined as (σ1 ⊗ σ1)T σ2, where the last Pauli matrix acts on the spin degree of freedom of the electrons. As we do not consider real spin of electrons in our model, we have to define T as in (51) to keep T 2 = −1. of H, corresponding to the band-edge energies λn, are annihilated by the integrals of motion Qa, a = 0, 1, 2, 3 . (56) QaΨ2−deg = 0, Let us denote by the subscript K and K′(=−K) the states that are nonvanishing in one valley only, i.e. 1 2 (1 ± τ30) Ψ±K = Ψ±K. We can find mutual eigenstates K and Ψ± Ψ± K′ of the Hamiltonian H, the valley-index op- erator τ30, and the integrals Q0 and Q3. They satisfy the following relations, K(K′) = 0, (H − λ)Ψ± and (τ30 − 1)Ψ± K = (τ30 + 1)Ψ± K′ = 0 (57) K, QiΨ± K = ±γλΨ± QiΨ± The eigenvalues γλ can be determined from the spectral polynomial (11) as K′ = ±γλΨ± i = 0, 3 . (58) K′ 2N +1 γλ = pP (λ) = Yn=0 (λ − λn)1/2 . (59) Hence, the operators Q0 and Q3 act on the basis of Ψ± as the multiplication by pP (λ), i.e. as the square root of the operator P (H). As mentioned above, the roots of the spectral polynomial (59) coincide with the van Hove singularities of the analyzed finite-gap systems. The two antidiagonal operators Q1 and Q2 switch the valley in- dex, ±K Q1Ψ± K(K′) = ±γλΨ± K′(K), Q2Ψǫ ±K = ±iǫγλΨǫ ∓K , (60) where ǫ = ±. The action of the operators Qa on the valley index is not indicated by the algebra (55). To reflect better the properties of the system, we can define a superalge- bra graded by the valley index operator τ30. We denote F1(2) ≡ Q1(2) as fermionic operators that change the val- ley index of the wave functions ({F1(2), τ30} = 0) and B1(2) ≡ Q0(3) as bosonic operators that preserve the val- ley index ([B1(2), τ30] = 0). The superalgebra is nonlinear and contains other fermionic operators τ12 and τ22, [H,Ba] = [H,Fa] = 0, {Fa,Fb} = 2δabP (H), (61) (62) [Ba,Fb] = 2i δa2 ǫ3bc τc2 P (H), [Ba, τ22] = −2iδ2a F1, [Ba, τ12] = 2i δ2a F1, (63) (64) {Fa, τb2} = 2δab B1. The fact that we deal with finite-gap systems is mani- fested in the anticommutator of the fermionic operators where the spectral polynomial P (H) emerges naturally. It underlies nonlinearity of the superalgebra and mani- fests the intimate relationship of between the algebraic structure and the van Hove singularities of the considered models. 9 Let us stress that the superalgebra (61)-(64) exists for any finite-gap configuration of the twisted nanotubes de- scribed by H as long as the Hamiltonian possesses the time-reversal symmetry. The choice of the grading operator was not unique. We could use either τ12 or τ22 equally well; both of them ei- ther commute or anticommute with the considered oper- ators. Notice that τ12 corresponds to the unitary compo- nent of the time-reversal (51). Choosing any of them as the new grading operator, qualitatively the same superal- gebra would be obtained. The operators (52)-(53) would be just permuted in the role of bosonic and fermionic generators. Let us notice that in examples presented in the pre- vious section, the single valley Hamiltonians with the vector potentials (21) and (27) commute with the op- erator σ3R where R is the parity10, RxR = −x. Hence, the corresponding Hamiltonian (49) is commuting with τ33R. The latter operator also commutes with τ30 and τ22, whereas it anticommutes with all the operators Qa, a = 0, .., 3. It means that τ33R could be regarded as a grading operator of a distinct, N = 4 superalgebra that would be generated by four fermionic operators (52) and (53). The nonlinear superalgebra of Bogoliubov-de Gennes Hamiltonians generated by nonlocal supercharges was discussed in the literature. We refer to [15] for more details, see also [42]. The formulas for LDOS and DOS computed in the third section with the use of the formula (17) have to be multiplied by four to get the correct form for the cor- responding twisted nanotubes. Indeed, we have to take into account the valley degeneracy that we discussed in this section, as well the double degeneracy of energy lev- els due to (real) spin− 1 2 of the particle that was neglected up this moment. Finally, let us discuss briefly the settings where an ex- ternal magnetic field is present in addition to the twists. The magnetic field breaks the time-reversal symmetry. When the vector potential ∆mg is included into the Hamiltonian, we have h(±K) = −i∂xσ2 ± Σ(x)σ1 ± M σ3 + ∆mgσ1. (65) We can see that as long as mass term M is vanishing and either magnetic field or twists are switched on (i.e. Σ(x)∆mg = 0), all the energy levels have even degeneracy due to the unitary equivalence of h(K) and h(K′). The situation changes when both Σ(x) and ∆mg are nonzero. In that case, we can still have a finite-gap configuration in one valley described by h(K). However, in the second valley the finite-gap potential is violated in general by the changed sign of ∆mg. Curiously enough, we can still get a finite-gap config- uration for each valley by the fine-tuning of the external 10 The three-gap setting with (39) has the nonlocal integral of slightly modified form, see [15]. field. As an example, let us consider the situation when the low-energy dynamics in the K valley is described by Hamiltonian (49) which consists of two copies of the (uni- tarily) equivalent single-valley energy operators. 10 h(K) = −i∂xσ2 + (coth b + tanh x− tanh(x + b))σ1, (66) which is an infinite-period limit of the three-gap system (39). Let us suppose that the vector potential in (66) is induced both by radial twist and by external magnetic field, where Σ(x) = 1 2 (2 coth b + tanh x − tanh(x + b)) and ∆mg = 1 2 (tanh x − tanh(x + b)). Then the subsys- tem in the K′-valley is described by h(K′) = −i∂xσ2 − coth b σ1, (67) which is just the trivial one-gap system. In the current setting, deformation associated with Σ(x) is asymptot- ically uniform but gets changed in the localized region where the (asymptotically vanishing) external magnetic field is nonzero. The spectrum of the corresponding to- tal Hamiltonian H has two singlet discrete energy levels corresponding to the bound states and two doubly de- generate levels ± coth b corresponding to the threshold of the positive and negative continuum. The other en- ergy levels are four-fold degenerate. It is worth noticing that the discussed framework can be understood in the context of (planar) graphene crys- tal in the presence of the external magnetic field and strain, both of which depend on x only. Due to sepa- rability of the stationary equation, the one-dimensional Hamiltonian can be written as h(K) = −iσ2∂x + (ky + Ay(x))σ1, (68) where ky corresponds to the momentum that is paral- lel with the (pseudo-) magnetic barrier. The operator (68) describes a massless Dirac particle that moves with fixed direction in the presence of vector potential Ay, as- sociated with the strain and the external magnetic field. In this context, the setting with the single-valley Hamil- tonians (66) and (67) with the inhomogeneous external magnetic field perpendicular to the surface and given by ∆mg is rather realistic. VI. DISCUSSION AND OUTLOOK The one-dimensional Dirac operator with finite-gap potential appears in a variety of physically interesting models [14], [15], [43], [44]. In the present paper, we il- lustrated how the machinery of the AKNS hierarchy can be used in the analysis of the twisted nanotubes in the low-energy regime, particularly, for the computation of the local density of states. We showed that the finite-gap, time-reversal invariant configurations possess a hidden nonlinear supersymme- try that is associated with the Kramer's degeneracy of energy levels. Physics of these systems, namely the pres- ence of the two valleys at K and K′ and the preserved time-reversal symmetry, is responsible for the form of the The current situation differs from the quantum mod- els with bosonized supersymmetry [45], where a nonlocal integral of motion was identified as the grading operator. Both the Hamiltonian (49) and its integrals of motion (52) and (53), forming so(3) ⊕ u(1) Lie algebra, can be graded by a local operator, e.g. by the valley index oper- ator τ30. This framework represents a nontrivial example of the hidden supersymmetry in the sense that it natu- rally emerges within the unextended, physical Hamilto- nian (49). The explicit results for the presented finite-gap systems can be extended with the use of Darboux transforma- tion [46]. Within this framework, one can construct new finite-gap Hamiltonians h2 from a known one, namely h1. The transformation is given in terms of a matrix differen- tial operator D, which intertwines two one-dimensional Dirac Hamiltonians, Dh1 = h2D. It maps the eigenstates of h1 into those of h2, keeping the operators (almost) isospectral. Moreover, the diagonal resolvent of h2 can be computed directly from the diagonal resolvent of h1 with the use of D, see [28] for details. In the paper, the operator (3) was almost exclusively interpreted as the effective Hamiltonian of the twisted carbon (or boron-nitride) nanotube. As we discussed in the end of the preceding section, the results can be also used in the analysis of the Dirac electrons in graphene in the presence of (pseudo-)magnetic barriers. Such systems with Kronig-Penney or a piece-wise constant (pseudo-) magnetic fields induced by either external field or strains were considered in the literature, see [47] or [48]. In this context, the DOS computed for the twisted nanotubes can be interpreted as the partial density of states in graphene for the ky = 0 channel. It could facilitate the computation of the transition coefficient in the normal direction to the magnetic barrier. The known results [49] on the relation between one-dimensional DOS and the phase of the transmission amplitude could be partic- ularly helpful in this context. Considering Dirac electrons in graphene, it is desirable to extend the analysis for ky 6= 0 as well. Keeping in mind [37] or [50], the infinite-period limit of the finite-gap models could be a feasible starting point in this respect. The analysis of periodic systems could make it possible to observe the phenomena that appear in graphene super- lattices, the new generation of Dirac points in particular [51], [52]. Study of the finite-gap configurations of electrostatic potential represents another possible direction for fu- ture research. Spectral properties of Dirac electron in graphene in presence of both periodic electrostatic and magnetic fields were discussed in various works, see e.g. [53], [54], [55]. In this context, the mapping between the systems with magnetic and electrostatic field [56] could provide an interesting way to extend our results. The finite-gap systems are an approximation of more realistic settings. They can serve as a test field for numer- ical or perturbative methods and can also provide qual- itative insight into the experimental data. Although, to our best knowledge, the experiments with the single-wall carbon nanotubes with the periodically modulated twist have not been prepared yet, the building blocks of such settings seem to be available, see e.g. [24]-[26]. As an experimental implementation of the proposed models, we can imagine a long suspended nanotube an- chored to a substrate at the ends and a periodic array of small paddles attached to it. By deflection of the pad- dles, the twist of the nanotube could be altered. Let us mention that the latter configuration (with one paddle) was employed in [57] for measurement of the torsional properties of the nanotubes. In the presented finite-gap models, the possible inter- action of the nanotube with the anchors and the paddles is not taken into account. Still, we think that the results (e.g. the suggested dependence of the central gap on the average twist) might provide interesting qualitative in- sight into the spectral properties of the settings realized in the experiments. 11 Acknowledgments The authors would like to thank to Gerald Dunne and Dmitry V. Kolesnikov for discussions. F.C. was sup- ported by the Fondecyt Grant No. 11121651 and by the Conicyt grant 79112034 and ACT-91. F.C. wishes to thank the kind hospitality of the Nuclear Physics Institute of the ASCR and the Physics Department of the University of Connecticut. The Centro de Estudios Cient´ıficos (CECs) is funded by the Chilean Government through the Centers of Excellence Base Financing Pro- gram of Conicyt. V.J. was supported by GA CR Grant No. P203/11/P038 of the Czech Republic. [1] S. Iijima, Nature 354, 56 (1991). [2] Q. Zhao, M. Buongiorno Nardelli, and J. Bernholc, Phys. Rev. B 65, 144105 (2002). [17] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). [18] A. Shytov, M. Rudner, Nan Gu, M. Katsnelson, L. Lev- [3] B. I. Yakobson, C. J. Brabec, and J. Bernholc, Phys. Rev. itov, Solid State Commun. 149, 1087 (2009). Lett. 76, 2511 (1996). [19] V. P. Gusynin, S. G. Sharapov, J. P. Carbotte, Int. J. [4] S. J. Tans et al., Nature (London) 386, 474 (1997). [5] P. G. Collins and P. Avouris, Sci. Amer. 283, 62 (2000); P. Avouris, J. Appenzeller, R. Martel, and S. J. Wind, Proc. IEEE 91, 1772 (2003); M. P. Anantram and F. L´eonard, Rep. Prog. Phys. 69, 507 (2006). [6] O. Klein, Z. Phys. 53, 157 (1929). [7] T. Ando, T. Nakanishi, and R. Saito, J. Phys. Soc. Jpn. 67, 2857 (1998); P. L. McEuen, M. Bockrath, D. H. Cobden, Y-G. Yoon, and S. G. Louie, Phys. Rev. Lett. 83, 5098 (1999); T. Ando, H. Suzuura, Physica E 18, 202 (2003); V. Jakubsk´y, L. -M. Nieto and M. S. Plyushchay, Phys. Rev. D 83, 047702 (2011). [8] O. O. Kit, T. Tallinen, L. Mahadevan, J. Timonen, and P. Koskinen, Phys. Rev. B 85, 085428 (2012). [9] see chapter E in [20] and references therein. Mod. Phys. B 21, 4611 (2007). [20] J. -C. Charlier, X. Blase and S. Roche, Rev. Mod. Phys. 79, 677 (2007). [21] Xueshen Wang et al., Nano. Lett. 9, 3137 (2009). [22] D. V. Kolesnikov, V. A. Osipov, Phys. Part. Nucl. 40, 502 (2009). [23] M. A. H. Vozmediano, M. I. Katsnelson and F. Guinea, Phys. Rept. 496, 109 (2010) and references therein. [24] J. C. Meyer, M. Paillet, S. Roth, Science 309, 1539 (2005). [25] H. B. Peng, C. W. Chang, S. Aloni, T. D. Yuzvinsky, and A. Zettl, Phys. Rev. B 76 035405 (2007). [26] A. M. Fennimore, T. D. Yuzvinsky, Wei-Qiang Han, M. S. Fuhrer, J. Cumings and A. Zettl, Nature 424, 409 (2003). [27] E. Joselevich, ChemPhysChem 7, 1405 (2006). [28] V. Jakubsk´y and M. S. Plyushchay, Phys. Rev. D 85, [10] C. L. Kane and E. J. Mele, Phys. Rev. Lett. 78, 1932 045035 (2012). (1997). [11] V. M. Pereira, A. H. Castro Neto, Phys. Rev. Lett. 103, [29] E. J. M. Hamilton et.al., Science 260, 659 (1993). [30] A. Rubio, J. L. Corkill, and M. L. Cohen, Phys. Rev. B 046801 (2009). 49, 5081 (1994). [12] G. W. Semenoff, Phys. Rev. Lett. 53, 2449 (1984). [13] J. Tersoff, D. R. Hamann, Phys. Rev. B 31, 805 (1985); Z. F. Wang, Ruoxi Xiang, Q. W. Shi, Jinlong Yang, Xi- aoping Wang, J. G. Hou, and Jie Chen, Phys. Rev. B 74, 125417 (2006). [31] M. Thies, J. Phys. A 39, 12707 (2006). [32] C. De Concini, R. A. Johnson, Ergodic Theory Dynami- cal Systems 7, 1 (1987). [33] G. Eilenberger, Z. Phys. 214, 195 (1968). [34] I. Kosztin, S. Kos, M. Stone, and A. J. Leggett, Phys. [14] G. Basar, G. V. Dunne, Phys. Rev. Lett. 100, 200404 Rev. B 58, 9365 (1998). (2008); G. Basar, G. V. Dunne, Phys. Rev. D 78, 065022 (2008). [15] F. Correa, G. V. Dunne and M. S. Plyushchay, Annals Phys. 324, 2522 (2009) [16] F. Gesztesy, H. Holden, Soliton Equations and Their Algebro-Geometric Solutions, Cambridge University Press, New York (2003). [35] M. Abramowitz and I. A. Stegun, "Jacobian elliptic func- tions and theta functions", Ch. 16 in Handbook of Math- ematical Functions with Formulas, Graphs, and Mathe- matical Tables, (Dover, New York, Dover, 1972), pp. 567 [36] E. T. Whittaker and G. N. Watson, A course of modern analysis (Cambridge University Press, Cambridge, Eng- land 1980). [37] E. Milpas, M. Torres and G. Murgu´ıa, J. Phys.: Condens. M. Titov, Phys. Rev. B 82, 155417 (2010). Matter 23, 245304 (2011). [48] L. DellAnna and A. De Martino, Phys. Rev. B 79, 045420 [38] D. J. Fernandez, B. Mielnik, O. Rosas-Ortiz and B. F. Samsonov, Phys. Lett. A 294, 168 (2002) [quant-ph/0302204]. (2009). [49] Y. Avishai, Y. B. Band, Phys. Rev. B 32, 2674 (1985). [50] D. S. Miserev, M. V. Entin, J. Experiment. and Theor. [39] F. Correa, V. Jakubsk´y and M. S. Plyushchay, J. Phys. Phys. 115, 694 (2012). A 41, 485303 (2008) [arXiv:0806.1614 [hep-th]]. [51] Ch.-H. Park, Li Yang, Y-W. Son, M. L. Cohen, and 12 [40] R. Jackiw and C. Rebbi, Phys. Rev. D 13, 3398 (1976). [41] E. D. Belokolos et al, Algebro-geometric approach to non- linear integrable equations, (Springer, Berlin, 1994). [42] F. Correa, V. Jakubsk´y, and M. S. Plyushchay, Phys. Rev. Lett. 101, 030403 (2008) [arXiv:0801.1671 [hep-th]]. -M. Nieto L. [43] J. Feinberg, Annals Phys. 309, 166 (2004). [44] A. Arancibia, J. M. Guilarte and M. S. Plyushchay, arXiv:1210.3666 [math-ph]; A. Arancibia and M. S. Plyushchay, Phys. Rev. D 85, 045018 (2012) [45] M. S. Plyushchay, Annals Phys. 245, 339 (1996) [hep-th/9601116]; Int. J. Mod. Phys. A 15, 3679 (2000) [hep-th/9903130]. [46] L.M. Nieto, A.A. Pecheritsin, B.F. Samsonov, Annals Phys. 305, 151 (2003). [47] M. Ramezani Masir, P. Vasilopoulos, F. M. Peeters, New J. Phys. 11 095009 (2009); S. Gattenlohner, W. Belzig, S. G. Louie, Phys. Rev. Lett. 101, 126804 (2008); L. Brey, H. A. Fertig, Phys. Rev. Lett. 103, 046809 (2009). [52] L. DellAnna and A. De Martino, Phys. Rev. B 83, 155449 (2011). [53] D. S. Novikov, Phys. Rev. B 72, 235428 (2005). [54] V. K. Dugaev, M. I. Katsnelson, arXiv:1206.4526. [55] Li-Gang Wang, Shi-Yao Zhu, Phys. Rev. B 81, 205444 (2010); M. Ramezani Masir, P. Vasilopoulos, F. M. Peeters, J. Phys.: Condens. Matter 22, 465302 (2010). [56] Liang Zheng Tan, Ch.-H. Park, and S. G. Louie, Phys. Rev. B 81, 195426 (2010). [57] A. R. Hall, L. An, J. Liu, L. Vicci, M. R. Falvo, R. Su- perfine, and S. Washburn, Phys. Rev. Lett. 96, 256102 (2006).
1904.09202	1	1904	2019-04-19T14:13:23	Absence of inter-layer tunnel coupling of $K$-valley electrons in bilayer MoS$_{2}$	[ "cond-mat.mes-hall" ]	In Bernal stacked bilayer graphene interlayer coupling significantly affects the electronic bandstructure compared to monolayer graphene. Here we present magnetotransport experiments on high-quality $n$-doped bilayer MoS$_{2}$. By measuring the evolution of the Landau levels as a function of electron density and applied magnetic field we are able to investigate the occupation of conduction band states, the interlayer coupling in pristine bilayer MoS$_{2}$, and how these effects are governed by electron-electron interactions. We find that the two layers of the bilayer MoS$_{2}$ behave as two independent electronic systems where a two-fold Landau level's degeneracy is observed for each MoS$_{2}$ layer. At the onset of the population of the bottom MoS$_{2}$ layer we observe a large negative compressibility caused by the exchange interaction. These observations, enabled by the high electronic quality of our samples, demonstrate weak interlayer tunnel coupling but strong interlayer electrostatic coupling in pristine bilayer MoS$_{2}$. The conclusions from the experiments may be relevant also to other transition metal dichalcogenide materials.	cond-mat.mes-hall	cond-mat	Absence of inter-layer tunnel coupling of K-valley electrons in bilayer MoS2 Riccardo Pisoni,1 Tim Davatz,1 Kenji Watanabe,2 Takashi Taniguchi,2 Thomas Ihn,1 and Klaus Ensslin1 1Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland 2National Institute for Material Science, 1-1 Namiki, Tsukuba 305-0044, Japan (Dated: April 22, 2019) In Bernal stacked bilayer graphene interlayer coupling significantly affects the electronic band- structure compared to monolayer graphene. Here we present magnetotransport experiments on high-quality n-doped bilayer MoS2. By measuring the evolution of the Landau levels as a function of electron density and applied magnetic field we are able to investigate the occupation of conduction band states, the interlayer coupling in pristine bilayer MoS2, and how these effects are governed by electron-electron interactions. We find that the two layers of the bilayer MoS2 behave as two inde- pendent electronic systems where a two-fold Landau level's degeneracy is observed for each MoS2 layer. At the onset of the population of the bottom MoS2 layer we observe a large negative com- pressibility caused by the exchange interaction. These observations, enabled by the high electronic quality of our samples, demonstrate weak interlayer tunnel coupling but strong interlayer electro- static coupling in pristine bilayer MoS2. The conclusions from the experiments may be relevant also to other transition metal dichalcogenide materials. Of the multitude of two-dimensional (2D) host materi- als, transition metal dichalcogenides (TMDs) are promis- ing candidates for exploring quantum correlated elec- tronic phases and electron-electron interaction effects due to their intrinsic 2D nature, large spin-orbit interaction and large effective mass carriers. Molybdenum disulfide (MoS2) is one of the most widely studied TMDs and still most of its fundamental quantum electronic prop- erties have thus far been elusive. Contrary to monolayer MoS2, in pristine bilayer MoS2 inversion symmetry is re- stored [1 -- 3]. As a result, the orbital magnetic moment and the valley-contrasting optical dichroism vanish [1, 4]. A potential difference between the two layers breaks the inversion symmetry [5, 6]. The influence of a perpendic- ular electric field on bilayer MoS2 has been extensively probed by optical excitation [5, 6]. Very little is known about the electronic transport properties of bilayer MoS2 when electric and magnetic fields are both applied per- pendicular to the sample plane [7]. Magnetotransport studies of 2D holes have been recently performed in bi- layer WSe2 revealing the presence of two subbands, each localized in the top and bottom layer, and demonstrat- ing an upper bound of the interlayer tunnel coupling of 19 meV [8, 9]. A thorough study of the interlayer cou- pling in the conduction band of bilayer transition metal dichalcogenides is still missing [7, 10]. Here we report a magnetotransport study of electrons in the conduction band of dual-gated bilayer MoS2. All studied bilayer samples exhibit Shubnikov-de Haas (SdH) oscillations with a twofold Landau level degeneracy at T = 1.5 K. At lower temperature the valley degeneracy is lifted and spin-valley coupled Landau levels are resolved. The evolution of the Landau level spectrum as a func- tion of density indicates that electrons occupy states of the K and K(cid:48) valley in each layer. By tuning the Fermi energy in each layer individually we are able to populate lower and upper spin-orbit split bands in both layers. The exchange interaction in a single layer yields a pro- nounced negative compressibility visible in occupation of the states detected via the Landau fan diagram. In addi- tion, we observe an intricate interplay between spin- and valley-polarized Landau levels originating from the two decoupled MoS2 layers. We do not observe any obvious signature in the Landau level spectrum when the electro- static potential difference between the two layers vanishes and the structural inversion symmetry is expected to be restored. Figure 1(a) shows the schematic cross-section of the dual-gated bilayer MoS2 device under study. The MoS2 is encapsulated between two hBN dielectrics with graphite layers as top and bottom gates. We fabricate pre- patterned Au bottom contacts below the bilayer MoS2. Ohmic behavior of these contacts is achieved by applying a sufficiently positive top gate voltage (VTG). The het- erostructure is assembled using a dry pick-up and transfer method [11 -- 13]. We fabricated and measured three bi- layer MoS2 samples, labeled A, B, and C, which show the same behavior. We will mainly discuss samples A and B here. The samples use commercial bulk MoS2 crystal (SPI supplies) mechanically exfoliated on SiO2/Si substrates. Using a combination of optical contrast, photolumines- cence spectroscopy and atomic force microscopy bilayer MoS2 flakes are identified. Figure 1(b) shows the optical micrograph of sample A. The bilayer MoS2 flake is out- lined with a white dashed line. Top and bottom graphite gates are outlined in purple and top and bottom hBN are outlined in blue and cyan, respectively. In the inset of Fig. 1 (b) we sketch the contact geometry where contacts 1 and 4 are used for current injection and extraction and contacts 2 and 3 serve as voltage probes. Figure 1(c) shows the four-terminal resistance, R14,23, as a function of VBG and VTG at T = 1.5 K. Green solid lines denote specific resistance values as a function of VTG 9 1 0 2 r p A 9 1 ] l l a h - s e m . t a m - d n o c [ 1 v 2 0 2 9 0 . 4 0 9 1 : v i X r a 2 from 1.5 K to 4.5 K and a density n = 3.8 × 1012 cm−2. For T = 1.5 K SdH oscillations start at ≈ 3 T. At T = 100 mK the onset of SdH oscillations moves to yet lower magnetic fields yielding a lower bound for the quan- tum mobility of ≈ 5000 cm2/Vs (see Fig. 4). The elec- tron density is calculated from the period of the SdH oscillations in 1/B considering valley degenerate Landau levels at the K and K(cid:48) conduction band minima [13]. At T = 1.5 K and n = 3.8 × 1012 cm−2 we observe the sequence of odd filling factors ν = 21, 23, 25, .... The twofold Landau level's valley degeneracy is lifted at lower temperatures (see below). We determine the electron effective mass m∗ from the temperature dependence of the SdH oscillations by fit- ting ∆R14,23 to ε/ sinh(ε), where ε = 2π2 kBT / ωc and ωc = eB/ m∗ is the cyclotron frequency [14 -- 16]. The inset of Fig. 1 (c) shows the extracted m∗ for the three different samples. We obtain a density-averaged mass of m∗ ≈ 0.62 me which does not show any obvious depen- dence neither on n nor on B [17, 18]. In Fig. 1 (d) we extract the m∗ of the K and K(cid:48) elec- trons localized in the top MoS2 layer, thus effectively calculating the effective mass of a monolayer MoS2 [7]. The effective masses we extract in our bilayer samples are systematically 10−20% lower compared to the ones mea- sured in monolayer MoS2 [13]. In bilayer MoS2 the top MoS2 layer is encapsulated between hBN and the bottom MoS2 layer, which is devoid of electrons. We speculate that the higher dielectric constant ( ≈ 6.4) reported for monolayer MoS2 [19 -- 22] compared to hBN ( ≈ 3.5) causes a weakening of electron-electron interaction effects thus affecting the m∗ value. The interaction strength is characterized by the di- mensionless Wigner-Seitz radius rs = 1/(√π na∗ B), where B = aB(κme/m∗) is the effective Bohr radius and κ the a∗ dielectric constant. For the considered electron density range we estimate rs = 1.9 − 10, placing the system in a regime where interaction effects are important [23 -- 25]. Figure 2(a) shows ∆R14,23 (color scale) as a function of B applied perpendicular to the sample and VBG, at VTG = 9 V and T = 1.5 K. For VBG < 1.3 V (black arrow in Fig. 2 (a)) the Landau level's evolution as a function of VBG resembles the one of monolayer MoS2 [13]. For VBG < −1.5 V only the Landau levels of the lower spin- orbit split K↑ and K(cid:48) ↓ bands are seen. As the electron density increases in this regime, we observe an alternat- ing parity of the filling factor sequence [see filling factor sequences in Fig. 2 (a)]. These results can be explained in an extended single-particle picture where the valley g factor is density dependent, following the interpretation of previous works [7 -- 10, 26]. At VBG = −1.5 V there is a sudden change in the slope of the Landau fan related to the occupation of the higher spin-orbit split K↓ and K(cid:48) ↑ valleys. Where the slope changes, the electron density is n = 3.6× 1012 cm−2. As- suming two-fold valley-degeneracy and using the experi- FIG. 1. (a) Schematic of the device. Bilayer MoS2 is encapsulated between two hBN layers and Au contacts are pre-patterned on the bottom hBN before the MoS2 layer is transferred. Graphite flakes serve as bottom and top gates. (b) Optical micrograph of sample A. The bilayer MoS2 flake is highlighted with a white dashed line. Inset: contact ge- ometry, current is injected to contact 1 and extracted from contact 4, voltage is measured between contacts 2 and 3 (scale bar is 2 µm). (c) Four-terminal resistance R14,23 as a func- tion of VTG and VBG at T = 1.5 K. Equi-resistance lines are highlighted in green. The white dashed line denotes the VBG values at which the bottom MoS2 layer starts to be pop- ulated. Solid and dotted lines represent VTG and VBG val- ues at which the upper spin-orbit split bands are occupied in the top and bottom MoS2 layers, respectively. (d) Temper- ature dependence of the SdH oscillations at VBG = −0.4 V and VTG = 9 V, n ≈ 3.8 × 1012 cm−2. An odd filling factor sequence ν = 21, 23, 25, ..., is observed. Inset: electron effec- tive masses ( m∗) extracted for the three different samples as a function of electron density. Black, blue, and red markers represent samples A, B, and C, respectively. and VBG. At fixed VTG we observe a sudden increase in R14,23 at VBG ≈ 1 V (white dashed line) that we attribute to the population of the bottom MoS2 layer. As a result, for the VBG values on the left side of the white dashed line we probe the electron transport only through the top MoS2 layer. At VBG ≈ 3.4 V (white dotted line) and for VTG and VBG values along the white solid line, we observe additional resistance kinks that we attribute to the occupation of the upper spin-orbit split bands in the bottom and top MoS2 layers, respectively. We investigate magnetotransport phenomena in bi- layer MoS2 using lock-in techniques at 31.4 Hz. Fig- ure 1(d) shows ∆R14,23, the four-terminal linear resis- tance with a smooth background subtracted, as a func- tion of magnetic field B at various temperatures ranging −101234567VBG(V)7891011VTG(V)100140180200400R14,23(Ω)45678B(T)−4004080120∆R14,23(Ω)1.5K4.5K34n(1012cm−2)0.00.40.8m∗(me)(a)(b)VTGVBGGraphiteGraphiteh-BNh-BN123456MoS210 μm(d)ν = 252321(c)3 below this threshold changes sign from positive to nega- tive. Second, an additional set of Landau levels appears [blue dashed lines in Fig. 2 (a)]. At VBG = 3.4 V the slope of these secondary Landau levels changes by about a fac- tor of 2, as indicated with white dashed lines in Fig. 2 (a). To interpret these observations, we extract how the electron densities change as we tune VBG. To determine the electron density of the individual layers and bands from the Landau fan diagram we generate a Fourier transform map of ∆R14,23 vs. 1/B for each VBG value in Fig. 2 (a) (see supplemental information). The Fourier transform of the SdH oscillations shows multiple peaks in the amplitude spectrum as we increase VBG. From these peaks we extract the electron density of the various spin- orbit split bands in bilayer MoS2 using n = (gve/h) × f , where f is the frequency of the Fourier transform peaks and gv = 2 accounts for the valley degeneracy. The results of this procedure are shown in Fig. 2 (b). For VBG < 1.2 V electrons populate only the top MoS2 layer where they occupy the lower (green dashed line) and up- per (orange dashed line) spin-orbit split bands as we in- crease VBG. At VBG = 1.2 V the bottom MoS2 layer starts to be populated (blue dashed line). The secondary Landau fan that appears at VBG = 1.2 V in Fig. 2 (a) orig- inates from the Landau levels of the electrons populating the bottom MoS2 layer. Beyond VBG = 1.2 V we see an increasing density in the bottom layer, whereas the top layer density starts to decrease. This density decrease is direct experimental evidence for the negative compress- ibility of the bottom layer at low densities [28 -- 31]. At VBG > 3.4 V the two valleys of the upper spin-orbit split bands in the bottom layer start to be populated (cyan dashed line). In the following, we will quantify this negative com- pressibility effect in MoS2 based on the data in Fig. 2. To this end, we consider the electrostatic model schemat- ically displayed in the inset of Fig. 3 consisting of three layers of different dielectric constants, in which electric displacement fields exist due to the applied voltages VTG and VBG. The MoS2 bilayer is modeled as two grounded conducting planes of finite density of states with a ge- ometric capacitance CBL and a displacement field DBL between them. It is our goal to express dDBL/dDB, i.e., the change in DBL upon a change in the displacement field DB between back gate and MoS2, at constant top gate voltage in terms of the measured VBG-dependent changes of the layer densities. This quantity allows us to directly compare the strength of the effect with the results obtained by Eisenstein et al [28] in the case of a GaAs double quantum well, and with the numerical results of Tanatar and Ceperley [30]. The model (see supplemental material for details) re- = CBL CB × sults in dDBL dDB (cid:12)(cid:12)(cid:12)(cid:12)VTG FIG. 2. (a) Sample A. Four-terminal resistance ∆R14,23 as a function of VBG and magnetic field at T ≈ 1.5 K and VTG = 9 V. For VBG < 1.3 V electrons are localized in the top MoS2 layer. The slope change at VBG = −1.5 V indicates the occupation of the upper spin-orbit split bands in the top layer. At VBG = 1.3 V the Landau fan of the bottom MoS2 layer appears. At VBG = 3.4 V, the electrons localized in the bottom layer populate the higher energy bands at the K and K(cid:48) valleys. (b) Electron densities in the bilayer MoS2 bands as a function of VBG at T ≈ 1.5 K and VTG = 9 V. Green, orange, blue, and cyan dashed lines correspond to electron densities in the lower (nt,lb, nb,lb), upper (nt,ub, nb,ub) spin- orbit split bands in the top and bottom layer, respectively. Green and blue solid lines represent the total carrier density in the top (nt) and bottom (nb) layer, respectively. Black solid line corresponds to the total electron density (ntot) in the bilayer MoS2. At VBG = 5.4 V (red circle) same carrier density in top and bottom MoS2 layer is achieved. mentally determined electron effective mass, we calculate the Fermi energy to be EF = 14 meV, in good agreement with the intrinsic spin-orbit interaction measured previ- ously for K-valley electrons in monolayer MoS2 [13]. We would like to note that our results justify the assump- tions in [27] that bilayer MoS2 investigated in the right regime behaves as single-layer MoS2 with the caveat that the effective mass is different because of the dielectric en- vironment. The measured Landau level structure for VBG < 1.2 V fully agrees with our previous monolayer MoS2 re- sults [13]. For VBG crossing the voltage 1.2 V from be- low, we observe two important changes in the SdH os- cillations compared to the monolayer system. First, the slope of the SdH oscillation minima vs VBG that existed −3−2−101234567VBG(V)02468101214n(1012cm−2)nt,lbnt,ubnb,lbnb,ubntotntnb−3−2−101234567VBG(V)45678B(T)−70070∆R14,23(Ω)(a)(b)ν = 1820222123251513174 strongly affects the energy momentum dispersion com- pared to monolayer graphene [32]. Previous work probing hole transport in bilayer WSe2 reported an upper bound for the interlayer tunnel coupling of ≈ 19 meV [8]. In our results the interlayer coupling in the conduction band of bilayer MoS2 is not observable. We achieve same electron densities in both layers (red circle in Fig. 2 (b)) for three different samples with no experimental evidence for in- terlayer coupling. Band structure calculations [33] reveal that strong interlayer hybridization in the conduction band of MoS2 occurs predominantly from orbitals which are responsible for the minima at the Q-point, which are not occupied in our samples. Conversely, weak interlayer hybridization is expected from the orbitals forming the K-valleys, which is consistent with our experimental ob- servations. At lower temperatures finer details of the Landau level structure are resolved. In Fig. 4 (a) we show ∆R14,23 as a function of B and VBG at VTG = 13.5 V and T = 100 mK for sample B. For VBG < 2 V (white dashed line) the bottom MoS2 layer is devoid of electrons and we only ob- serve the Shubnikov-de Haas oscillations of the top layer. At T = 100 mK we are able to resolve valley-spin polar- ized Landau levels originating from the lowest conduction band minima in the top layer. The Landau level struc- ture of the spin-valley coupled bands in the bottom layer appears for VBG ≥ 2 V. Figures 4(b-c) show an enlarge- ment of Fig. 4 (a). When only the top layer is populated [see Fig. 4 (b)] we observe a pattern of avoided crossings, a signature of the coupling between the spin-valley polar- ized Landau levels of the lower and upper spin-orbit split bands [13]. Figure 4(c) shows that Landau levels of elec- trons populating the two different layers cross each other, indicating weak coupling between the two electronic sys- tems below our measurement resolution. In conclusion, we observe SdH oscillations at magnetic fields as low as 2 T at T ≈ 100 mK, testifying to the high- mobility of our dual-gated bilayer MoS2 devices. We are able to measure spin-valley polarized LLs originat- ing from the lower and upper spin-orbit split bands of K-valley electrons populating the top and bottom MoS2 layers. Our observations demonstrate that electrons in bilayer MoS2 behave like two independent electronic sys- tems. The exchange interaction at the turn on of the two-dimensional electron gas in the bottom layer leads to the observation of a large negative compressibility. Our work demonstrates fundamental electronic transport properties as well as the importance of interaction effects in pristine bilayer MoS2. These results bear relevance for understanding electronic transport in twisted bilayer TMDs. We thank Guido Burkard, Vladimir Falko, Andor Korm´anyos, Mansour Shayegan and Peter Rickhaus for fruitful discussions. We thank Peter Marki as well as the FIRST staff for their technical support. We acknowledge financial support from ITN Spin-NANO FIG. 3. Sample A. Ratio of the electric displacement fields as a function of VBG at T ≈ 1.5 K and VTG = 9 V. The top green axis represents the electron density in the bottom layer (nB). The blue axis denotes the rS parameter that accounts for intralayer interactions in the bottom layer assuming in-plane dielectric constant ≈ 15.3 [22] and the measured effective mass of 0.6me. Inset: electrostatic model of our dual-gated bilayer MoS2 device. (cid:18) (cid:18) CB CT + e dnt dVBG × CBL CT + e d(nt+nb) dVBG (cid:19) (cid:12)(cid:12)(cid:12)VTG (cid:12)(cid:12)(cid:12)VTG (cid:12)(cid:12)(cid:12)VTG (cid:12)(cid:12)(cid:12)VTG (cid:19) − CTe dnb dVBG + CTe dnb dVBG , (1) where CT and CB are the geometric capacitances per unit area between MoS2 and top- and bottom-gate, respec- tively. The quantities nt and nb are the measured total electron densities in the two layers shown in Fig. 2 (b). In the case of VBG < 1.2 V, where nb and its VBG-derivative are zero, the displacement field ratio in eq. (1) is exactly one. Negative compressibility in the region VBG > 1.2 V will manifest itself by dDBL/dDB < 0. Figure 3 displays the result of applying eq. (1) to the data in Fig. 2 (b). Two curves are shown, in which the quite uncertain value of CBL takes on two plausible ex- treme values (see supplemental information for details). This shows that the result depends very little on the exact value of this parameter. A strong negative compressibil- ity with dDBL/DB ≈ −0.5 is seen at VBG = −2 V where nb = 1 × 1012 cm−2, roughly ten times stronger than the effect observed in Ref. [28]. To compare the value to the numerical results of Ref. [30], we have added an esti- mated scale bar of rs-values to the top axis in Fig. 3. The negative compressibility values measured in our sample agree fairly well with the predictions of the numerical calculations at these rs-values. Resolving individual layer electron densities in Fig. 2 indicates that the two MoS2 layers are weakly coupled. This observation is in contrast to Bernal stacked bi- layer graphene, where the interlayer coupling of ≈ 0.4 eV −3−2−101234567VBG(V)−0.50−0.250.000.250.500.751.00dDBL/dDBCBL=0.01Fm−2CBL=0.65Fm−201234567nB(1012cm−2)234510rSVTG=9.0VdTdBLdBDBLDTDBhBNGraphiteϵϵBLϵVTGVBGFIG. 4. (a) Sample B. Four-terminal resistance ∆R14,23 as a function of VBG and magnetic field at T ≈ 100 mK and VTG = 13.5 V. At VBG = 2.2 V the bottom MoS2 layer is filled with electrons and SdH oscillations appear (white dashed lines). Inset: conduction band minima sketch at the K and K(cid:48) point for top and bottom MoS2 layer. The horizontal black dashed line corresponds to the highest Fermi energy reached in the top layer before the bottom layer is occupied. (b) Avoided crossing patterns between spin-valley coupled LLs originating from the lower and upper spin-orbit split bands in the top MoS2 layer. (c) Crossings between LLs in the lower spin-orbit split bands originating from top and bottom MoS2 layers. Marie Sklodowska-Curie grant agreement no. 676108, the Graphene Flagship and the National Center of Com- petence in Research on Quantum Science and Technol- ogy (NCCR QSIT) funded by the Swiss National Science Foundation. Growth of hexagonal boron nitride crystals was supported by the Elemental Strategy Initiative con- ducted by the MEXT, Japan and JSPS KAKENHI Grant Numbers JP15K21722. 5 [1] D. Xiao, G.-B. Liu, W. Feng, X. Xu, and W. Yao, Phys- ical Review Letters 108, 196802 (2012). [2] K. F. Mak, K. He, J. Shan, and T. F. Heinz, Nature Nanotechnology 7, 494 (2012). [3] T. Cao, G. Wang, W. Han, H. Ye, C. Zhu, J. Shi, Q. Niu, P. Tan, E. Wang, B. Liu, and J. Feng, Nature Commu- nications 3, 887 (2012). [4] W. Yao, D. Xiao, and Q. Niu, Physical Review B 77, 235406 (2008). [5] S. Wu, J. S. Ross, G.-B. Liu, G. Aivazian, A. Jones, Z. Fei, W. Zhu, D. Xiao, W. Yao, D. Cobden, and X. Xu, Nature Physics 9, 149 (2013). [6] J. Lee, K. F. Mak, and J. Shan, Nature Nanotechnology 11, 421 (2016). [7] J. Lin, T. Han, B. A. Piot, Z. Wu, S. Xu, G. Long, L. An, P. K. M. Cheung, P.-P. Zheng, P. Plochocka, D. K. Maude, F. Zhang, and N. Wang, arXiv:1803.08007 [cond- mat] (2018), arXiv: 1803.08007. [8] B. Fallahazad, H. C. Movva, K. Kim, S. Larentis, T. Taniguchi, K. Watanabe, S. K. Banerjee, and E. Tu- tuc, Physical Review Letters 116, 086601 (2016). [9] H. C. Movva, B. Fallahazad, K. Kim, S. Larentis, T. Taniguchi, K. Watanabe, S. K. Banerjee, and E. Tu- tuc, Physical Review Letters 118, 247701 (2017). [10] S. Larentis, H. C. P. Movva, B. Fallahazad, K. Kim, A. Behroozi, T. Taniguchi, K. Watanabe, S. K. Banerjee, and E. Tutuc, Physical Review B 97, 201407 (2018). [11] R. Pisoni, Y. Lee, H. Overweg, M. Eich, P. Simonet, K. Watanabe, T. Taniguchi, R. Gorbachev, T. Ihn, and K. Ensslin, Nano Letters 17, 5008 (2017). [12] R. Pisoni, Z. Lei, P. Back, M. Eich, H. Overweg, Y. Lee, and K. Ensslin, K. Watanabe, T. Taniguchi, T. Ihn, Applied Physics Letters 112, 123101 (2018). [13] R. Pisoni, A. Korm´anyos, M. Brooks, Z. Lei, P. Back, M. Eich, H. Overweg, Y. Lee, P. Rickhaus, K. Watanabe, T. Taniguchi, A. Imamoglu, G. Burkard, T. Ihn, and K. Ensslin, Physical Review Letters 121, 247701 (2018). [14] T. Ando, A. B. Fowler, and F. Stern, Reviews of Modern Physics 54, 437 (1982). [15] A. Isihara and L. Smrcka, Journal of Physics C: Solid State Physics 19, 6777 (1986). [16] V. M. Pudalov, M. E. Gershenson, and H. Kojima, Phys- ical Review B 90, 075147 (2014). [17] Y. Zhang and S. Das Sarma, Physical Review B 72, 075308 (2005). [18] C. Attaccalite, S. Moroni, P. Gori-Giorgi, and G. B. Bachelet, Physical Review Letters 88, 256601 (2002). [19] T. Cheiwchanchamnangij and W. R. L. Lambrecht, Phys- ical Review B 85, 205302 (2012). [20] A. Ramasubramaniam, Physical Review B 86, 115409 (2012). [21] A. Molina-S´anchez and L. Wirtz, Physical Review B 84, 155413 (2011). [22] A. Laturia, M. L. V. d. Put, and W. G. Vandenberghe, npj 2D Materials and Applications 2, 6 (2018). [23] T. Okamoto, K. Hosoya, S. Kawaji, and A. Yagi, Phys- ical Review Letters 82, 3875 (1999). [24] A. A. Shashkin, S. V. Kravchenko, V. T. Dolgopolov, and T. M. Klapwijk, Physical Review Letters 87, 086801 (2001). [25] K. Vakili, Y. P. Shkolnikov, E. Tutuc, E. P. De Poortere, 0123456789VBG(V)345678B(T)−50050∆R14,23(Ω)2.53.03.54.04.55.0VBG(V)6.06.57.07.58.0B(T)−50050∆R14,23(Ω)0.00.51.01.52.0VBG(V)6.06.57.07.58.0B(T)−50050∆R14,23(Ω)(b)(c)(a)KK'KK'Top layerBottom layerK'>K>K' >K>EFand M. Shayegan, Physical Review Letters 92, 226401 (2004). [26] M. V. Gustafsson, M. Yankowitz, C. Forsythe, D. Rhodes, K. Watanabe, T. Taniguchi, J. Hone, X. Zhu, and C. R. Dean, arXiv:1707.08083 [cond-mat] (2017), arXiv: 1707.08083. [27] J. Lin, T. Han, B. A. Piot, Z. Wu, S. Xu, G. Long, L. An, P. Cheung, P.-P. Zheng, P. Plochocka, X. Dai, D. K. Maude, F. Zhang, and N. Wang, Nano Letters 19, 1736 (2019). [28] J. P. Eisenstein, L. N. Pfeiffer, and K. W. West, Physical Review Letters 68, 674 (1992). [29] M. S. Bello, E. I. Levin, and B. I. Shklovskr, , 8 (1981). [30] B. Tanatar and D. M. Ceperley, Physical Review B 39, 5005 (1989). [31] S. V. Kravchenko, D. A. Rinberg, S. G. Semenchinsky, and V. M. Pudalov, Physical Review B 42, 3741 (1990). [32] L. M. Zhang, Z. Q. Li, D. N. Basov, M. M. Fogler, Z. Hao, and M. C. Martin, Physical Review B 78, 235408 (2008). [33] A. Korm´anyos, V. Z´olyomi, V. I. Fal'ko, and G. Burkard, Physical Review B 98, 035408 (2018) + private commu- nication. 6
1211.2939	1	1211	2012-11-13T10:17:43	Landau level spectroscopy of relativistic fermions with low Fermi velocity in Bi2Te3 three-dimensional topological insulator	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	X-band microwave spectroscopy is applied to study the cyclotron resonance in Bi2Te3 exposed to ambient conditions. With its help, intraband transitions between Landau levels of relativistic fermions are observed. The Fermi velocity equals to 3260 m/s, which is much lower than has been reported in the literature for samples cleaved in vacuum. Simultaneous observation of bulk Shubnikov - de Haas oscillations by contactless microwave spectroscopy allows determination of the Fermi level position. Occupation of topological surface states depends not only on bulk Fermi level but also on the surface band bending.	cond-mat.mes-hall	cond-mat	Landau level spectroscopy of relativistic fermions with low Fermi velocity in Bi2Te3 three-dimensional topological insulator A. Wolos Institute of Physics, Polish Academy of Sciences, Al. Lotnikow 32/46, 02-668 Warsaw, Poland. Faculty of Physics, University of Warsaw, ul. Hoza 69, 00-681 Warsaw, Poland. S. Szyszko, A. Drabinska, and M. Kaminska Faculty of Physics, University of Warsaw, ul. Hoza 69, 00-681 Warsaw, Poland. S. G. Strzelecka, A. Hruban, A. Materna, and M. Piersa Institute of Electronic Materials Technology, ul. Wolczynska 133, 01-919 Warsaw, Poland. ABSTRACT: X-band microwave spectroscopy is applied to study the cyclotron resonance in Bi2Te3 exposed to ambient conditions. With its help, intraband transitions between Landau levels of relativistic fermions are observed. The Fermi velocity equals to 3260 m/s, which is much lower than has been reported in the literature for samples cleaved in vacuum. Simultaneous observation of bulk Shubnikov - de Haas oscillations by contactless microwave spectroscopy allows determination of the Fermi level position. Occupation of topological surface states depends not only on bulk Fermi level but also on the surface band bending. Receipt date: PACS: 76.40.+b, 73.20.At, 03.65.Vf. Keywords: topological insulators, cyclotron resonance, relativistic fermions. In the last five years a new class of topological quantum states has emerged, referred to as topological insulators.1,2,3,4,5,6,7,8,9,10 Topological properties of the family of bismuth compound crystals, i.e. Bi2Te3,5 Bi2Se3,8 and Bi2Te2Se,9 have been confirmed in spectacular angle resolved photoemission spectroscopy (ARPES) and scanning tunneling microscopy and spectroscopy11 experiments. Other experiments, including electric transport or optical measurements, are difficult to perform because of high bulk conductance due to native crystal lattice defects.6,8,12 The defects responsible for a high bulk carrier concentration are most probably Bi antisites in Bi2Te3 leading to typically highly metallic p- type, and Se vacancies in Bi2Se3 giving n-type conductivity.6,13,14 In order to suppress the contribution of the conducting bulk states and expose the surface states in the experiment, very thin samples with 1 the thickness of a few quintet layers are typically studied.15,16 An alternative approach is to use surface sensitive experimental techniques or techniques disregarding the metallic bulk. One of them is the already mentioned ARPES, which probes surface regions from a few up to tens of nm. In this communication we describe the results of microwave spectroscopy measurements performed in an X- band (9.5 GHz) resonator. The microwave penetration depth in the applied experimental conditions ranges from 4 μm to 40 μm. The technique can actually probe independently both bulk and surface properties, being partly insensitive to the detrimental influence of the metallic bulk conductivity. Similar approach has been recently applied to study surface states in Bi2Se3 topological insulator, where 2D cyclotron resonance has been revealed thanks to the application of the microwave cavity transmission technique.17 The Bi2Te3 crystal was grown by the vertical Bridgman method, with an excess of tellurium in the melt in order to suppress the formation of Bi antisite defects. Growth details have been described in Ref. 18. Segregation effects, characteristic of the Bridgman method, caused the gradient of stoichiometry along the crystal growth direction. For a given temperature, the equilibrium composition of the liquid phase is typically different than of the solid phase. In the case of Bi2Te3, this causes enrichment of the liquid phase in tellurium during the growth process, and formation of the concentration gradient of defects related to tellurium excess. The crystal, highly p-type at the crystal seed due to large concentration of Bi antisites (p = 1×1020 cm-3 at room temperature), changed its conductivity to n-type at its end (n = 3×1019 cm-3 at room temperature). The crystal was cut with a wire saw and cleaved with the razor edge to obtain samples with surface area not larger than 4 mm by 4 mm, to fit to the resonance cavity. Mirror-like surfaces perpendicular to the Bi2Te3 c-axis were obtained. The thickness of all the samples was typically about 0.5 mm. Three samples from p-type region of the crystal were prepared for the studies, cut at 73 mm (A), 80 mm (B), and 107 mm (C) from the seed. The samples were kept at ambient conditions. The microwave spectroscopy experiment was performed using a Bruker ELEXSYS E580 electron spin resonance (ESR) spectrometer operating in X-band (9.5 GHz), with a TE102 resonance cavity. The temperature was lowered down to 2.5 K using an Oxford continuous-flow cryostat. The magnetic field was swept up to 1.7 T. Due to the use of the magnetic field modulation and the lock-in detection technique, the recorded spectra take the form of the first derivative of the microwave power absorption. Two kinds of signals were observed, the Shubnikov - de Haas (SdH) oscillations and the cyclotron resonance. The SdH oscillations were visible in the high magnetic field range for samples A, B, and C, see Fig. 1 and Table 1. The SdH signal can be detected with the ESR spectrometer due to non-resonant changes in sample conductivity causing variation of the cavity quality factor. The contactless method of the SdH oscillations detection has been earlier applied for a high-mobility two- dimensional electron gas present at the GaN/AlxGa1-xvN interface19 or in InxGa1-xAs quantum wells.20 2 The SdH signal shows characteristic periodicity in 1/B (B is the external magnetic field). The Fourier transform of the signal (Fig. 1 (b)) reveals only one frequency of the oscillations, equal to 14.2 T, 14.0 T, and 7.0 T for samples A, B, and C respectively (for B s). Table 1. summarizes the parameters. The decrease of the SdH frequency, when the sample is taken away from the seed, is consistent with increasing the excess of tellurium in the melt during the growth process what causes systematic suppression of acceptor-like Bi antisite defect formation. The SdH oscillations in Bi2Te3 are well visible for the magnetic field oriented parallel to the Bi2Te3 bisectrix s-axis (in-plane of the cleavage surfaces), and up to about 30 degrees from this orientation towards the Bi2Te3 c-axis. The frequency of the oscillations increases slightly while rotating the sample, the onset moves towards higher magnetic fields, Fig 1(a) and (c). The angular dependence of the oscillations excludes the possibility of their originating from the 2D surface states, the oscillations should vanish for B in-plane of the cleavage surface in that case, and allows assigning them to the bulk holes. The structure of the Fermi surfaces in p-type Bi2Te3 (bulk) has been studied by first principles calculations, revealing a rather complicated picture of irregular elongated surfaces, located in the mirror plane of the Bi2Te3 Brillouin zone, tilted by 21 degrees with respect to the crystal principal axes.21 At least 6 hole pockets exist in the Brillouin zone. A simplified 6-valley model of ellipsoidal Fermi surfaces with non-parabolic dispersion has been applied by Köhler in 1976 to explain the pattern of the SdH oscillations in Bi2Te3.22,23 Significant non-parabolicity effects have been visible for the Fermi energy above 20 meV, while at around 25 meV the slope of mc(EF) has been infinite (mc is the cyclotron mass, EF – the Fermi energy counted from the bulk valence band edge, it increases when increasing hole concentration). The band edge cyclotron masses have been deduced by Köhler as equal to 0.080 m0 for B c, and 0.058 m0 for B s. For the Fermi energy EF = 21.6 meV, the cyclotron masses increase up to 0.102 m0 for B c and 0.075 m0 for B s. In general, the SdH pattern in Bi2Te3 is very complex, consisting of up to three periods at the arbitrary orientation of the magnetic field, with different amplitudes and field and temperature dependencies.22 In our experiment, due to the magnetic field limited to 1.7 T, we can only observe components characterized by the highest mobility and thus by the lowest cyclotron mass. This happens for orientations close to B s. Frequencies of the SdH oscillations, F = 14.2 T and 14.0 T, observed for samples A and B respectively, are very close to the value observed by Köhler when mc equals to 0.075 m0. A lower value of the SdH frequency in sample C, F = 7.0 T, suggests that the corresponding cyclotron mass is of the band edge (0.058 m0 for B s, after Köhler). The SdH period Δ(1/B), the cyclotron mass mc, and Fermi energy EF, are related through the equation:22 Δ(1/B)=e (cid:61) /(mcEF), Eq.1. where e is the electron charge and (cid:61) is a Planck constant over 2π. Using this relation, we can roughly determine the position of the Fermi level in our samples, assuming the cyclotron mass of 0.075 m0 for A and B samples and the band edge cyclotron mass 0.058 m0 for sample C. We obtain EF equal to 3 about 22 meV for samples A and B respectively, and EF = 14 meV for sample C. Table 1 collects the estimated parameters. In sample C, a strong signal shown in Fig.2(a) was detected next to the SdH oscillations. A large amplitude of the signal and a strong dependence of the amplitude on the position in the resonance cavity (optimized for measurements of magnetic dipole transitions) suggests the microwave electric field-driven cyclotron resonance. Cyclotron resonance-related lines are commonly detected in ESR cavities both in bulk materials24 as well as in 2D systems like Si/SiGe quantum wells25 or GaN/AlGaN heterostructures.19 We should mention here that the signal did not show any signs of aging within a time scale of a month. The spectrum in Fig. 2(a). shows clear structures for B c, and becomes flat for B s (in-plane of the cleavage surfaces). Three resonance lines can be distinguished, marked as L1, L2, and L3, at the magnetic field equal to 0.61 T, 1.09 T, and 1.56 T respectively (for B c). Fig. 2(b) shows the position of the resonance lines versus the tilt angle θ, between the magnetic field and the Bi2Te3 c-axis (θ = 0 means B c, θ = 90 corresponds to B s). The position of the resonance fields was determined by fitting three Gaussian components to each recorded spectrum, Fig. 3(b). The position of the three resonance lines shows clear dependence 1/cos(θ) characteristic for the cyclotron resonance of 2D objects, for which the cyclotron frequency depends only on the perpendicular component of the magnetic field. This allows us to assign the spectrum to the cyclotron resonance of the topological surface states. It is worth noting here that we can clearly exclude the possibility of the resonance being due to the bulk cyclotron resonance, which is expected at 27 mT for B c (mc = 0.080 m0 assumed) and 20 mT for B s (mc = 0.058 m0). The 2D quantum well states are either unlikely to be responsible for the observed signal, as they reflect the cyclotron mass of the bulk valence band as well. We can also rule out the bulk plasma-shifted cyclotron resonance due to large difference between the plasma frequency (1012 Hz) and experimentally set microwave frequency (9.5 GHz). The coupling of the surface cyclotron resonance to the surface plasma oscillations is also unlikely as it is dependent on the plasmon wave vector and thus on the sample lateral dimension.26 We did not observe any influence of the sample size on the resonance signal. The linear dispersion of relativistic fermions results in the Landau level (LL) energy spectrum described by the equation: = (cid:61) 2 E sgn(n)v eBn , Eq.2. n F where n = 0, ± 1, ± 2, ± 3,… is a LL index, vF denotes the Fermi velocity, (cid:61) is the Planck constant over 2π, e is electron charge, and B is the magnetic field. In our experiment, the microwave frequency is set constant (f = 9.5 GHz), while the magnetic field is swept in order to tune the transition energy to the energy of microwaves. The condition for the cyclotron resonance is: = − E E h f . n n Eq.3. +1 4 The resonance field depends on both the Fermi velocity and the n-index, while the ratio of two resonance fields depends only on n-indexes. This relation allows us to assign L1, L2, and L3 resonance lines to appropriate transitions. The measured ratio of the resonance fields equals to: L B B 3 : L B B 2 : = 1.43 ± 0.03 and = 1.79 ± 0.03 respectively. The calculated ratio for cyclotron L 2 L 1 resonance transitions between LLs with four lowest indexes is: = 1.41 and B =↔= n n 3 4 : B =↔= n n 2 3 : B = 1.70. The agreement is very promising taking into account the experimental B =↔= =↔= n n n n 3 2 2 1 error in the determination of the resonance field. To sum up, we attribute the L1 line to the cyclotron resonance transition between Landau levels with n = 1 and n = 2, the L2 line to the transition between n = 2 and n = 3, and the L3 line to the transition between n = 3 and n = 4. Once knowing the indexes for particular cyclotron resonance lines we can determine the value of the Fermi velocity which equals to 3260 m/s. A scheme of the energy spectrum of Landau levels in Bi2Te3 is shown in Fig. 3(a). In the description proposed above we have used a notation “ ↔” in order to account for both absorption and relaxation processes occurring simultaneously in the resonance cavity. Based on the observation of the cyclotron resonance with the lowest n-indexes, one can conclude that the Dirac point is located near the Fermi level in sample C. From Shubnikov-de Haas we have already determined that the Fermi level lies 14 meV below the bulk valence band edge. Regarding samples A and B, the Fermi level lies deeper in the bulk valence band (see Tab. 1) which means that it lies also substantially below the Dirac point. This explains the lack of the cyclotron resonance in X-band, because the resonance transitions involving Landau levels with high n-indexes would require magnetic field higher than 40 T, therefore fall out of the range of available magnetic field. Moreover, surface states below the Dirac point seem to be degenerated with bulk valence band states,5 which may prevent observation of the cyclotron resonance at all. ARPES determines position of the Dirac point with respect to the bulk band structure, about 130 meV below the Bi2Te3 valence band edge.5 The location of the Dirac point close to the Fermi level in sample C requires upward band bending by about 116 meV at the surface. It has been schematically shown in Fig. 3 (c). This is different from Ref. 28 where downward band bending has been observed in Bi2Te3, Bi2(Te2.6Se0.4), and Bi2Se3 after short exposition of the sample surface to air. On the other hand in Ref. 8 an upward band bending has been reported for Bi2Se3. It seems that the problem in Bi2Te3 and related materials requires further studies. Nevertheless, it becomes clear that not only bulk doping determines occupation of the topological surface states, but surface doping and resulting band bending is not less important. The most striking conclusion from the cyclotron resonance experiment is the value of the Fermi velocity. ARPES measurements reveal that the surface states of Bi2Te3 consist of a single, non- degenerate Dirac cone at the Γ point with Fermi velocity of the order of 4 × 105 m/s,5 while our cyclotron resonance studies reveal much lower Fermi velocity, equal to 3260 m/s. Still little is known about the nature of topological surface states, thus many explanations of the discrepancies can be 5 considered. We could attribute the difference in the Fermi velocity to imperfectness of the surface morphology. This explanation is however hard to accept, keeping in mind that we deal with topologically protected surface states. On the other hand, similar effects have been recently observed in ultrahigh-mobility Bi2Se3.27 Despite a low bulk concentration of the order of 1016 cm-3, no effects of topological surface could be observed in electric transport measurements, possibly due to low mobility of surface electrons. The authors of Ref. 27 conclude that topological protection does not guarantee high surface mobility, and that controlled surface preparation may be necessary in order to obtain desired transport parameters. An alternative explanation is contamination of the surface exposed to ambient conditions. Recent ARPES works on topological insulators from the Bi2(Se3-xTex) family show that the surface states exposed to N2 or air for only 5 min., although demonstrate the robustness of the topological order, they also show the effects of modification of the electronic structure of the surface states. This includes additional doping of electrons into the surface states, modification of the Dirac cone shape (lowering of the Fermi velocity, enhancement of the warping), and finally formation of 2D quantum well states which occurs due to intercalation of gases between quintuple layers.28 The presented effects of modification of the Dirac cone are not clear enough to explain reduction of the Fermi velocity by two orders of magnitude. However, no data on the long-term exposition to ambient conditions are yet available. To sum up, we observed the cyclotron resonance in Bi2Te3 sample exposed to ambient conditions. The nature of the states responsible for the resonance is unambiguously topological with the linear Dirac dispersion. The corresponding Fermi velocity equals to 3260 m/s, which is considerably lower than derived from ARPES5 for samples cleaved in vacuum. Occupation of topological surface states depends not only on bulk Fermi level but also on surface band bending. The Dirac point is located about 14 meV below the Bi2Te3 valence band edge in the bulk, which requires upward band bending by about 116 meV at the surface. Observation of the cyclotron resonance in a sample exposed to ambient conditions illustrates exceptional properties of the topologically protected surface states. Surface contamination with atmospheric gases will lower Fermi velocity by two orders of magnitude, it will not however cause the extinction of the surface states. Their nature will remain relativistic. ACKNOWLEDGEMENTS This work has been supported by funds for science, grant number: 2011/03/B/ST3/03362, Poland. 6 ) s t i n u . b r a ( e d u t i l p m A l a n g i S 2.5 (a) 2.0 1.5 1.0 A 0.5 0.0 -0.5 -1.0 B C FIGURES: θ = 130ο 120o 110o 100o 90o, B s 80o 70o 60o 90o, B s 90o, B s ) s t i n u . b r a ( r e w o P ) T ( y c n e u q e r F 80 60 40 20 0 17 16 15 14 13 12 (b) A & B C θ = 90o, B s 0 20 40 60 80 100 Frequency (T) (c) B s sample A 60 80 100 Tilt Angle, θ (deg.) 120 0.6 1.2 0.9 Inverse Magnetic Field (T) 1.5 FIG 1. (a) SdH oscillations in p-type Bi2Te3 recorded in the resonance cavity. For sample A, the spectra were taken versus angle θ between the direction of the magnetic field B and Bi2Te3 c-axis. For samples B and C, the spectra at θ = 90o are shown. (b) Fourier transform shows one oscillation frequency equal to 14.2 T, 14.0 T, and 7.0 T for samples A, B, and C, respectively, at θ = 90o. (c) frequency of the SdH oscillations versus tilt angle θ indicates bulk origin. 7 20 15 10 5 0 ) s t i n u . b r a ( e d u t i l p m A l a n g i S L2 (a) θ = 0°, B c L1 Sample C (b) 2.5 2.0 L3 10° 20° 30° 40° 50° 1.5 ) T ( d l e i F e c n a n o s e R 60° 70° 80° 90°, B s 0.0 1.0 0.5 1/cos(θ) L3 L2 L1 0.0 0.5 1.0 Magnetic Field (T) 1.5 B c B s B c 0 45 90 135 180 Tilt Angle, θ (deg) FIG 2. Cyclotron resonance in p-type Bi2Te3, sample C. (a) spectra recorded versus tilt angle θ between B and c. (b) resonance fields of L1, L2, and L3 lines versus the tilt angle θ – symbols. Solid lines show 1/cos(θ) dependence. Both the position of the arrows in (a) and the resonance fields in (b) were determined by fitting three Gaussian components to each recorded spectrum, see Fig 3(b). 8 e c a f r u s DP (a) 0.3 0.2 0.1 0.0 -0.1 -0.2 ) V e m ( y g r e n E -0.3 0.0 15 10 5 0 -5 ) s t i n u . b r a ( e d u t i l p m A l a n g i S -10 0.0 -1 -2 -3 -4 2.0 L3: n = 3 ↔ n = 4 1.4 1.6 1.8 2.0 (c) VB edge EF z L1 L2 L3 = 0.039 meV = h⋅9.5 GHz n = 4 3 2 1 0 0.2 0.4 0.6 0.8 1.2 1.0 Magnetic Field (T) 1.4 1.6 1.8 (b) L2: n = 2 ↔ n = 3 L1: n = 1 ↔ n = 2 Bi2Te3, saple C, T = 2.5 K , f = 9.5 GHz, B c 0.2 0.4 0.6 1.2 1.0 0.8 Magnetic Field (T) FIG 3. (a) diagram of Landau levels in Bi2Te3 exposed to ambient conditions. Arrows indicate cyclotron resonance transitions in X-band assigned to L1, L2, and L3 lines respectively. (b) decomposition of the recorded spectrum (open points) into three Gaussian components (dotted lines). Exponential background was assumed (dashed line). Solid line shows the fit. Note that the signal represents first derivative of the microwave power absorption due to the use of the lock-in detection technique. (c) The inset shows schematic representation of band bending at the surface of Bi2Te3. 9 TABLE: TABLE I. Parameters for Bi2Te3 samples: distance from the crystal seed l, conductivity type, frequency of the SdH oscillations recorded for B s, F. mc Bs is the assumed cyclotron mass (after Köhler, Ref. 22) and EF is the Fermi energy evaluated from Eq. 1 (counted from the bulk valence band edge). Last column says whether the cyclotron resonance in X-band was observed. Sample A B C mc B s (m0) 0.075 0.075 0.058 EF (meV) 22 22 14 CR NO NO YES l (mm) 73 80 107 type p p p F B s (T) 14.2 14.0 7.0 1 C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 146802 (2005). 2 C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005). 3 H. Zhang, Ch.-X. Liu, X.-L. Qi, Xi Dai, Z. Fang, and Sh.-Ch. Zhang, Nature Physics 5, 438 (2009). 4 M. Konig, S. Wiedmann, C. Brune, A. Roth, H. Buhmann, L. W. Molenkamp, X.–L. Qi, and S.-C. Zhang, Science 318, 766 (2007). 5 Y. L. Chen, J. G. Analytis, J. H. Chu, Z. K. Liu, S. K. Mo, X. L. Qi, H. J. Zhang, D. H. Lu, X. Dai, Z. Fang, S. C. Zhang, I. R. Fisher, Z. Hussain, and Z.X. Shen, Science 325, 178 (2009). 6 D. X. Qu, Y. S. Hor, J. Xiong, R. J. Cava, and N. P. Ong, Science 329, 821 (2010). 7 D. Hsieh, Y. Xia, L. Wray, D. Qian, A. Pal, J. H. Dil, J. Osterwalder, F. Meier. G. Bihlmayer, C. L. Kane, Y. S. Hor, R. J. Cava, and M. Z. Hasan, Science 323, 919 (2009). 8 J. G. Analytis, J.-H. Chu, Y. Chen, F. Corredor, R. D. McDonald, Z. X. Shen, and I. R. Fisher, Phys. Rev. B 81, 205407 (2010). 9 T. Arakane, T. Sato, S. Souma, K. Kosaka, K. Nakayama, M. Komatsu, T. Takahashi, Zhi Ren, Kouji Segawa, and Yoichi Ando, Nature Communications 3, 636 (2012). 10 H. Zhang, Ch.-X. Liu, X.-L. Qi, Xi Dai, Z. Fang, and Sh.-Ch. Zhang, Nature Physics 5, 438 (2009). 11 Z. Alpichshev, J. G. Analytis, J.-H. Chu, I. R. Fisher, Y. L. Chen, Z. X. Shen, A. Fang, and A. Kapitulnik, Phys. Rev. Lett. 104, 016401 (2010). 12 Z. Ren, A. A. Taskin, S. Sasaki, K. Segawa, and Y. Ando, Phys. Rev. B 82, 241306 (2010). 13 Y. Xia, D. Qian, D. Hsieh, L. Wray, A. Pal, H. Lin, A. Bansil, D. Grauer, Y. S. Hor, R. J. Cava, and M. Z. Hasan, Nat. Phys. 5, 398 (2009). 14 M.Z. Hasan and C.L. Kane, Rev. Mod. Phys. 82, 3045 (2010). 15 Y. S. Kim, M. Brahlek, N. Bansal, E. Edrey, G. A. Kapilevich, K. Iida, M. Tanimura, Y. Horibe, S.- W. Cheong, and S. Oh, Phys. Rev. B 84, 073109 (2011). 16 H. Peng, K. Lai, D. Kong, S. Meister, Y. Chen, X.-Y. Qi, S.-Ch. Zhang, Z.-X. Shen, and Y. Cui, Nature Materials 9, 225 (2010). 17 O. E. Ayala-Valenzuela, J. G. Analytis, J.-H. Chu, M. M. Altarawneh, I. R. Fisher, and R. D. McDonald, arXiv: 1004.2311v1 (2010). 18 A. Hruban, A. Materna, W. Dalecki, G. Strzelecka, M. Piersa, E. Jurkiewicz-Wegner, R. Diduszko, M. Romaniec, and W. Orłowski, Acta Phys. Pol. A 120, 950 (2010). 10 19 A. Wolos, W. Jantsch, K. Dybko, Z. Wilamowski, C. Skierbiszewski, Phys. Rev. B 76, 045301 (2007). 20 G. Hendorfer, M. Seto, H. Ruckser, W. Jantsch, M. Helm, G. Brunthaler, W. Jost, H. Obloh, K. Kohler, and D. J. As, Phys. Rev. B 48, 2328 (1993). 21 S. J. Youn and A. J. Freeman, Phys. Rev. B 63, 085112 (2001). 22 H. Kohler, Phys. Stat. Sol. (b) 74, 591 (1976). 23 H. Kohler, Phys. Stat. Sol. (b) 75 441 (1976). 24 H. Malissa, Z. Wilamowski, and W. Jantsch, AIP Conf. Proc. 772, pp. 1218-1219, PHYSICS OF SEMICONDUCTORS: 27th International Conference on the Physics of Semiconductors - ICPS-27, 26-30 July 2004, Flagstaff, Arizona (USA). 25 Z. Wilamowski, N. Sandersfeld, W. Jantsch, D. Tobben, and F. Schaffler, Phys. Rev. Lett. 87, 026401 (2001). 26 S. Das Sarma and E. H. Hwang, Phys. Rev. Lett. 102, 206412 (2009). 27 N. P. Butch, K. Kirshenbaum, P. Syers, A. B. Sushkov, G. S. Jenkins, H. D. Drew, and J. Paglione, Phys. Rev. B. 81, 241301R (2010). 28 Ch. Chen, S. He, H. Weng, W. Zhang, L. Zhao, H. Liu, X. Jia, D. Mou, S. Liu, J. He, Y. Peng, Y. Feng, Z. Xie, G. Liu, X. Dong, J. Zhang, X.Wang, Q. Peng, Z. Wang, S. Zhang, F. Yang, Ch. Chen, Z. Xu, Xi, Dai, Z. Fang, and X. J. Zhou, Proc. Nat. Ac. Sci. 109, 3694 (2012). 11
1003.0984	2	1003	2010-05-24T07:52:55	Quantum discrete breathers	[ "cond-mat.mes-hall" ]	We review recent studies about quantum discrete breathers. We describe their basic properties in comparison with their classical counterparts, and the ways they may be addressed theoretically in different quantum lattice models including either bosonic or fermionic excitations. We also review recent experimental work in the field.	cond-mat.mes-hall	cond-mat	Quantum discrete breathers Ricardo A. Pinto1 and Sergej Flach2 1. Department of Electrical Engineering, University of California Riverside, 900 University Ave., Riverside, California 92521, USA 2. Max-Planck-Institut fur Physik komplexer Systeme, Nothnitzer Str. 38, 01187 Dresden, Germany October 30, 2018 1 Introduction In solid-state physics, the phenomenon of localization is usually perceived as arising from extrinsic disorder that breaks the discrete translational invariance of the perfect crystal lattice. Familiar examples include the localized vibra- tional modes around impurities or defects in crystals and Anderson localization of waves in disordered media [1]. The usual perception among solid-state re- searchers is that, in perfect lattices excitations must be extended objects as well, essentially plane wave like. Such firmly entrenched perceptions were severely jolted since the discovery of discrete breathers (DB), also known as intrinsic localized modes (ILM). These states are typical excitations in perfectly peri- odic but strongly nonlinear systems, and are characterized by being spatially localized, at variance to plane wave states [2, 3, 4]. DB-like excitations, being generic objects, have been observed in a large variety of lattice systems that include bond excitations in molecules, lattice vibrations and spin excitations in solids, charge flow in coupled Josephson junc- tions, light propagation in interacting optical waveguides, cantilever vibrations in michromechanical arrays, cold atom dynamics in Bose-Einstein condensates loaded on optical lattices, among others. They have been extensively studied, and a high level of understanding about their properties has been reached. Two decades of intensive research have polished our theoretical understand- ing of DBs in classical nonlinear lattices. Less is known about their quantum counterparts - quantum breathers (QB). This chapter is devoted to a review of the more recent studies in this field. The concept of QBs is closely related with the theme of dynamical tunneling in phase space. 1 1.1 A few facts about classical discrete breathers Let us study the combined effect of nonlinearity and discreteness on the spatial localization of a discrete breather on a basic level. For that we look into the dynamics of a one-dimensional chain of interacting (scalar) oscillators with the Hamiltonian H = Xn (cid:20) 1 2 p2 n + V (xn) + W (xn − xn−1)(cid:21) . (1) The integer n marks the lattice site number of a possibly infinite chain, and xn and pn are the canonically conjugated coordinate and momentum of a degree of freedom associated with site number n. The on-site potential V and the interaction potential W satisfy V ′(0) = W ′(0) = 0, V ′′(0), W ′′(0) ≥ 0. This choice ensures that the classical ground state xn = pn = 0 is a minimum of the energy H. The equations of motion read xn = pn , pn = −V ′(xn) − W ′(xn − xn−1) + W ′(xn+1 − xn) . (2) Let us linearize the equations of motion around the classical ground state. We obtain a set of linear coupled differential equations with solutions being small amplitude plane waves: xn(t) ∼ ei(ωq t−qn) , ω2 q = V ′′(0) + 4W ′′(0) sin2(cid:16) q 2(cid:17) . (3) These waves are characterized by a wave number q and a corresponding fre- quency ωq. All allowed plane wave frequencies fill a part of the real axis which is coined linear spectrum. Due to the underlying lattice the frequency ωq depends periodically on q and its absolute value has always a finite upper bound. The maximum (Debye) frequency of small amplitude waves ωπ = pV ′′(0) + 4W ′′(0). Depending on the choice of the potential V (x), ωq can be either acoustic- or optic-like, V (0) = 0 and V (0) 6= 0, respectively. In the first case the linear spec- trum covers the interval −ωπ ≤ ωq ≤ ωπ which includes ωq=0 = 0. In the latter case an additional (finite) gap opens for ωq below the value ω0 = pV ′′(0). For large amplitude excitations the linearization of the equations of motion is not correct anymore. Similar to the case of a single anharmonic oscillator, the frequency of possible time-periodic excitations will depend on the amplitude of the excitation, and thus may be located outside the linear spectrum. Let us assume that a time-periodic and spatially localized state, i.e. a discrete breather, xn(t + Tb) = xn(t) exists as an exact solution of Eqs.(2) with the period Tb = 2π/Ωb. Due to its time periodicity, we can expand xn(t) into a Fourier series xn(t) = Xk AkneikΩb t . The Fourier coefficients are by assumption also localized in space Ak,n→∞ → 0 . 2 (4) (5) Inserting this ansatz into the equations of motion (2) and linearizing the re- sulting algebraic equations for Fourier coefficients in the spatial breather tails (where the amplitudes are by assumption small) we arrive at the following linear algebraic equations: k2Ω2 bAkn = V ′′(0)Akn + W ′′(0)(2Akn − Ak,n−1 − Ak,n+1) . (6) If kΩb = ωq, the solution to (6) is Ak,n = c1eiqn + c2e−iqn. Any nonzero (whatever small) amplitude Ak,n wil thus oscillate without further spatial decay, contradicting the initial assumption. If however kΩb 6= ωq (7) for any integer k and any q, then the general solution to (6) is given by Ak,n = c1κn + c2κ−n where κ is a real number depending on ωq, Ωb and k. It always admits an (actually exponential) spatial decay by choosing either c1 or c2 to be nonzero. In order to fulfill (7) for at least one real value of Ωb and any integer k, we have to request ωq to be bounded from above. That is precisely the reason why the spatial lattice is needed. In contrast most spatially continuous field equations will have linear spectra which are unbounded. That makes resonances of higher order harmonics of a localized excitation with the linear spectrum unavoidable. The nonresonance condition (7) is thus an (almost) necessary condition for obtaining a time-periodic localized state on a Hamiltonian lattice [2]. The performed analysis can be extended to more general classes of discrete lattices, including e.g. long-range interactions between sites, more degrees of freedom per each site, higher-dimensional lattices etc. But the resulting non- resonance condition (7) keeps its generality, illustrating the key role of discrete- ness and nonlinearity for the existence of discrete breathers. Let us show discrete breather solutions for various lattices. We start with a chain (1) with the functions V (x) = x2 + x3 + 1 4 x4 , W (x) = 0.1x2 . (8) The spectrum ωq is optic-like and shown in Fig.1. Discrete breather solutions can have frequencies Ωb which are located both below and above the linear spec- trum. The time-reversal symmetry of (2) allows to search for DB displacements xn(t = 0) when all velocities xn(t = 0) = 0. These initial displacements are computed with high accuracy (see following sections) and plotted in the insets in Fig.1 [3]. We show solutions to two DB frequencies located above and below ωq - their actual values are marked with the green arrows. To each DB frequency we show two different spatial DB patterns - among an infinite number of other possibilities, as we will see below. The high-frequency DBs (Ωb ≈ 1.66) occur for large-amplitude, high-energy motion with adjacent particles moving out of phase. Low-frequency DBs (Ωb ≈ 1.26) occur for small-amplitude motion with adjacent particles moving in phase. 3 ω 1.8 1.6 1.4 1.2 0 1 0. -1 0.4 0.2 0. q ω q 1 2 3 Figure 1: The frequency versus wavenumber dependence of the linear spectrum for a one-dimensional chain of anharmonic oscillators with potentials (8). Cho- sen DB frequencies are marked with green arrows and lie outside the linear spectrum ωq. Red circles indicate the oscillator displacements for a given DB solution, with all velocities equal to zero. Lines connecting circles are guides for the eye. From [3]. In Fig.2 we show two DB solutions for a Fermi-Pasta-Ulam chain of particles coupled via anharmonic springs V (x) = 0, W (x) = 1 4 x4 (c.f. (1)) which has an acoustic-type spectrum [5]. The DB frequency is in both cases Ωb = 4.5. Again the displacements xn are shown for an initial time when all velocities vanish. In the inset we plot the strain un = xn − xn−1 on a log-normal scale. The DB solutions are exponentially localized in space. Finally we show DB solutions for a two-dimensional square lattice of an- harmonic oscillators with nearest neighbour coupling. The equations of motion read 2 x2 + 1 xi,j = k(xi+1,j + xi−1,j − 2xi,j) + k(xi,j+1 + xi,j−1 − 2xi,j) − xi,j − x3 i,j (9) with oscillator potentials V (x) = 1 In Fig.3 we plot the oscillator displacements with all velocities equal to zero for three different DB frequencies and k = 0.05 [6]. For all cases adjacent oscillators move out of phase. 2 x2 + 1 4 x4. We conclude this section with emphasizing that DB solutions can be typically localized on a few lattice sites, regardless of the lattice dimension. Thus little overall coherence is needed to excite a state nearby - just a few sites have to oscillate coherently, the rest of the lattice does not participate strongly in the excitation. 4 1e+02 1e-02 1e-06 ) 0 = t ( n U 1e-10 -10 -5 0 n 5 10 -2 0 2 n 1e+02 4 6 8 10 1e-02 1e-06 ) 0 = t ( n U 1e-10 -10 -5 0 n 5 10 3 2 1 0 -1 -2 -4 2.5 2 1.5 1 0.5 0 -0.5 ) 0 = t ( n x ) 0 = t ( n x -1 -4 -2 0 2 n 4 6 8 10 Figure 2: Discrete breather solutions for a Fermi-Pasta-Ulam chain (see text). These states are frequently referred to as the Page mode (left) and the Sievers- Takeno mode (right). Adapted from [5]. 0.2 0 -0.2 A 5 5 10 10 5 15 15 0.4 0.2 0 15 10 5 5 B 10 10 15 15 5 0.8 0.6 0.4 0.2 0 15 10 15 10 5 5 C 10 10 15 15 5 Figure 3: Displacements of DBs on a two-dimensional lattice (9) with k = 0.05, all velocities equal to zero. (A) Ωb = 1.188; (B) Ωb = 1.207; (C) Ωb = 1.319. From [6]. 1.2 From classical to quantum A natural question is what remains of discrete breathers if the corresponding quantum problem is considered [7, 8, 9]. The many-body Schrodinger equation is linear and translationally invariant, therefore all eigenstates must obey the Bloch theorem. Thus we cannot expect eigenstates of the Hamiltonian to be spatially localized (on the lattice). What is the correspondence between the quantum eigenvalue problem and the classical dynamical evolution? The concept of tunneling is a possible answer to this puzzle. Naively speak- 5 ing we quantize the family of periodic orbits associated with a discrete breather located somewhere on the lattice. Notice that there are as many such fami- lies as there are lattice sites. The quantization (e.g., Bohr-Sommerfeld) yields some eigenvalues. Since we can perform the same procedure with any family of discrete breather periodic orbits which differ only in their location on the lattice, we obtain N -fold degeneracy for every thus obtained eigenvalue, where N stands for the number of lattice sites. Unless we consider the trivial case of uncoupled lattice sites, these degeneracies will be lifted. Consequently, we will instead obtain bands of states with finite band width. These bands will be called quantum breather bands. The inverse tunneling time of a semiclassical breather from one site to a neighboring one is a measure of the bandwidth. We can then formulate the following expectation: if a classical nonlinear Hamiltonian lattice possesses discrete breathers, its quantum counterpart should show up with nearly degenerate bands of eigenstates, if the classical limit is con- sidered. The number of states in such a band is N , and the eigenfunctions are given by Bloch-like superpositions of the semiclassical eigenfunctions obtained using the mentioned Bohr-Sommerfeld quantization of the classical periodic or- bits. By nearly degenerate we mean that the bandwidth of a quantum breather band is much smaller than the spacing between different breather bands and the average level spacing in the given energy domain, and the classical limit implies large eigenvalues. Another property of a quantum breather state is that such a state shows up with exponential localization in appropriate correlation functions [10]. This approach selects all many-particle bound states, no matter how deep one is in the quantum regime. In this sense quantum breather states belong to the class of many-particle bound states. For large energies and N the density of states becomes large too. What will happen to the expected quantum breather bands then? Will the hybridization with other non-breather states destroy the particle-like nature of the quantum breather, or not? What is the impact of the nonintegrability of most systems allowing for classical breather solutions? Since the quantum case corresponds to a quantization of the classical phase space, we could expect that chaotic tra- jectories lying nearby classical breather solutions might affect the corresponding quantum eigenstates. From a computational point of view we are very much restricted in our abilities to study quantum breathers. Ideally we would like to study quantum properties of a lattice problem in the high energy domain (to make contact with classical states) and for large lattices. This is typically impossible, since solving the quantum problem amounts to diagonalizing the Hamiltonian matrix with rank bN where b is the number of states per site, which should be large to make contact with classical dynamics. Thus typically quantum breather states have been so far obtained numerically for small one-dimensional systems [10, 11, 12]. 6 Figure 4: Spectrum of the quantum DNLS with b = 2 and N = 101. The energy eigenvalues are plotted versus the wavenumber of the eigenstate. Adapted from [13]. 2 Quantum breather models 2.1 The Bose-Hubbard chain Let us discuss quantum breathers within the widely used quantum discrete nonlinear Schrodinger model (also called Bose-Hubbard model) with the Hamil- tonian [13] H = − N Xl=1 (cid:20) 1 2 and the commutation relations a† l a† l alal + C(a† l al+1 + h.c.)(cid:21) ala† m − a† mal = δlm (10) (11) with δlm being the standard Kronecker symbol. This Hamiltonian conserves the total number of particles B = Xl nl , nl = a† l al . For b particles and N sites the number of basis states is (b + N − 1)! b!(N − 1)! . (12) (13) For b = 0 there is just one trivial state of an empty lattice. For b = 1 there are N states which correspond to one-boson excitations. These states are similar to classical extended wave states. For b = 2 the problem is still exactly solvable, because it corresponds to a two-body problem on a lattice. A corresponding numerical solution is sketched in Fig. 4. Note the wide two-particle continuum, and a single band located below. This single band corresponds to quasiparticle 7 states characterized by one single quantum number (related to the wavenumber q). These states are two-particle bound states. The dispersion of this band is given [13] by E = −r1 + 16C2 cos2(cid:16) q 2(cid:17) . (14) Any eigenstate from this two-particle bound state band is characterized by ex- ponential localization of correlations, i.e. when represented in some set of basis states, the amplitude or overlap with a basis state where the two particles are separated by some number of sites is exponentially decreasing with increasing separation distance. Note that a compact bound state is obtained for q = ±π, i.e. for these wave numbers basis states with nonzero separation distance do not contribute to the eigenstate at all [14]. 2.2 The dimer A series of papers was devoted to the properties of the quantum dimer [15, 16, 17, 18]. This system describes the dynamics of bosons fluctuating between two sites. The number of bosons is conserved, and together with the conservation of energy the system appears to be integrable. Of course, one cannot consider spatial localization in such a model. However, a reduced form of the discrete translational symmetry - namely the permutational symmetry of the two sites - can be imposed. Together with the addition of nonlinear terms in the classical equations of motion the dimer allows for classical trajectories which are not invariant under permutation. The phase space can be completely analyzed, all isolated periodic orbits (IPO) can be found. There appears exactly one bifurca- tion on one family of isolated periodic orbits, which leads to the appearance of a separatrix in phase space. The separatrix separates three regions - one invari- ant and two non-invariant under permutations. The subsequent analysis of the quantum dimer demonstrated the existence of pairs of eigenstates with nearly equal eigenenergies [15]. The separatrix and the bifurcation in the classical phase space can be traced in the spectrum of the quantum dimer [17]. The classical Hamiltonian may be written as H = Ψ∗ 1Ψ1 + Ψ∗ 2Ψ2 + 1Ψ1)2 + (Ψ∗ 1 2 (cid:0)(Ψ∗ 2Ψ2)2(cid:1) + C (Ψ∗ 1Ψ2 + Ψ∗ 2Ψ1) . (15) with the equations of motion Ψ1,2 = i∂H/∂Ψ∗ (or number of particles) B = Ψ12 + Ψ22. Let us parameterize the phase space of (15) with Ψ1,2 = A1,2eiφ1,2 , A1,2 ≥ 0. It follows that A1,2 is time independent and φ1 = φ2 + ∆ with ∆ = 0, π and φ1,2 = ω being also time independent. Solving the algebraic equations for the 1,2. The model conserves the norm 8 amplitudes of the IPOs we obtain I : A2 1,2 = II : A2 1,2 = 1 2 1 2 B , ∆ = 0 , ω = 1 + C + B , ∆ = π , ω = 1 − C + 1 2 1 2 B , B , III : A2 1 = 1 2 B(cid:16)1 ±p1 − 4C2/B2(cid:17) , ∆ = 0 , ω = 1 + B . (16) (17) (18) IPO III corresponds to two elliptic solutions which break the permutational symmetry. IPO III exists for B ≥ Bb with Bb = 2C and occurs through a bifurcation from IPO I. The corresponding separatrix manifold is uniquely defined by the energy of IPO I at a given value of B ≥ Bb. This manifold separates three regions in phase space - two with symmetry broken solutions, each one containing one of the IPOs III, and one with symmetry conserving solutions containing the elliptic IPO II. The separatrix manifold itself contains the hyperbolic IPO I. For B ≤ Bb only two IPOs exist - IPO I and II, with both of them being of elliptic character. Remarkably there exist no other IPOs, and the mentioned bifurcation and separatrix manifolds are the only ones present in the classical phase space of (15). To conclude the analysis of the classical part, we list the energy properties of the different phase space parts separated by the separatrix manifold. First it is straightforward to show that the IPOs (16)-(18) correspond to maxima, minima or saddle points of the energy in the allowed energy interval for a given value of B, with no other extrema or saddle points present. It follows E1 = H(IPO I) = B + E2 = H(IPO II) = B + 1 4 1 4 E3 = H(IPO III) = B + B2 + CB , B2 − CB , 1 B2 + C2 . 2 (19) (20) (21) For B < Bb we have E1 > E2 (IPO I - maximum, IPO II - minimum). For B ≥ Bb it follows E3 > E1 > E2 (IPO III - maxima, IPO I - saddle, IPO II - minimum). If B < Bb, then all trajectories are symmetry conserving. If B ≥ Bb, then trajectories with energies E1 < E ≤ E3 are symmetry breaking, and trajectories with E2 ≤ E ≤ E1 are symmetry conserving. The quantum eigenvalue problem amounts to replacing the complex func- tions Ψ, Ψ∗ in (15) by the boson annihilation and creation operators a, a† with the standard commutation relations (to enforce the invariance under the ex- change Ψ ⇔ Ψ∗ the substitution has to be done on rewriting ΨΨ∗ = 1/2(ΨΨ∗ + Ψ∗Ψ)): H = 5 4 + 3 2 (cid:16)a† 1a1 + a† 2a2(cid:17) + 1 2 (cid:16)(a† 1a1)2 + (a† 2a2)2(cid:17) + C(cid:16)a† 1a2 + a† 2a1(cid:17) . (22) Note that = 1 here, and the eigenvalues b of B = a† 2a2 are integers. Since B commutes with H we can diagonalize the Hamiltonian in the basis of 1a1 + a† 9 200000 0.02 ρ 150000 0.00 50000 E 100000 200000 E 50000 0 100 200 300 n~ Figure 5: Eigenvalues versus ordered state number n for symmetric and anti- symmetric states (0 < n < b/2 for both types of states). Parameters: b = 600 and C = 50. Inset: Density of states versus energy. Adapted from [17]. eigenfunctions of B. Each value of b corresponds to a subspace of the dimension (b + 1) in the space of eigenfunctions of B. These eigenfunctions are products of the number states ni of each degree of freedom and can be characterized by a symbol n, mi with n bosographicsns in the site 1 and m bosons in the site 2. For a given value of b it follows m = b − n. So we can actually label each state by just one number n: n, (b − n)i ≡ n). Consequently the eigenvalue problem at fixed b amounts to diagonalizing the matrix 5 0 2 (cid:0)n2 + (b − n)2(cid:1) 2 b + 1 4 + 3 Cpn(b + 1 − n) Cp(n + 1)(b − n) n = m n = m + 1 n = m − 1 else (23) Hnm =   where n, m = 0, 1, 2, ..., b. Notice that the matrix Hnm is a symmetric band matrix. The additional symmetry Hnm = H(b−n),(b−m) is a consequence of the permutational symmetry of H. For C = 0 the matrix Hnm is diagonal, with the property that each eigenvalue is doubly degenerate (except for the state b/2) for even values of b). The classical phase space contains only symmetry broken trajectories, with the exception of IPO II and the separatrix with IPO I (in fact in this limit the separatrix manifold is nothing but a resonant torus containing both IPOs I and II). With the exception of the separatrix manifold, all tori break permutational symmetry and come in two groups separated by the separatrix. Then quantizing each group will lead to pairs of degenerate eigen- values - one from each group. There is a clear correspondence to the spectrum of the diagonal (C = 0) matrix Hnm. The eigenvalues H00 = Hbb correspond to the quantized IPOs III. With increasing n the eigenvalues Hnn = H(b−n),(b−n) correspond to quantized tori further away from the IPO III. Finally the states with n = b/2 for even b or n = (b− 1)/2 for odd b are tori most close to the sep- aratrix. Switching the side diagonals on by increasing C will lead to a splitting of all pairs of eigenvalues. In the case of small values of b these splittings have 10 1050 100 300 200 100 0 0 E ∆ 10−50 E ∆ 10−100 10−150 0 300 150 n~ 20 40 60 80 n~ Figure 6: Eigenvalue splittings versus n for b = 150 and C = 10. Solid line - numerical result, dashed line - perturbation theory. Inset: Same for b = 600 and C = 50. Only numerical results are shown. Adapted from [17]. no correspondence to classical system properties. However, in the limit of large b we enter the semiclassical regime, and due to the integrability of the system, eigenfunctions should correspond to tori in the classical phase space which sat- isfy the Einstein-Brillouin-Keller quantization rules. Increasing C from zero will lead to a splitting ∆En of the eigenvalue doublets of C = 0. In other words, we find pairs of eigenvalues, which are related to each other through the symmetry of their eigenvectors and (for small enough C) through the small value of the splitting. These splittings have been calculated numerically and using perturba- tion theory [15, 17]. In the limit of large b the splittings are exponentially small for energies above the classical separatrix energy (i.e. for classical trajectories which are not invariant under permutation). If the eigenenergies are lowered below the classical separatrix energy, the splittings grow rapidly up to the mean level spacing. In Fig. 5 the results of a diagonalization of a system with 600 particles (b = 600) is shown [17]. The inset shows the density of states versus energy, which nicely confirms the predicted singularity at the energy of the separatrix of the classical counterpart. In order to compute the exponentially small split- tings, we may use e.g. a Mathematica routine which allows to choose arbitrary values for the precision of computations. Here we chose precision 512. In Fig. 6 the numerically computed splittings are compared to perturbation theory re- sults. As expected, the splittings become extremely small above the separatrix. Consequently these states will follow for long times the dynamics of a classical broken symmetry state. 2.3 The trimer The integrability of the dimer does not allow a study of the influence of chaos (i.e. nonintegrability) on the tunneling properties of the mentioned pairs of eigenstates. A natural extension of the dimer to a trimer adds a third degree of 11 freedom without adding a new integral of motion. Consequently the trimer is nonintegrable. A still comparatively simple numerical quantization of the trimer allows to study the behavior of many tunneling states in the large-energy domain of the eigenvalue spectrum [19]. Similarly to the dimer, the quantum trimer Hamiltonian is represented in the form H = + 15 3 8 2 +C(a† 1a1 + a† (a† 1a1 + a† 1a2 + a† 2a2 + a† 1 2a2 + a† 2a1) + δ(a† 3a3) + 1a3 + a† 2 h(a† 3a1 + a† 1a1)2 + (a† 2a3 + a† 2a2)2i 3a2) . (24) Again B = a† 3a3 commutes with the Hamiltonian, thus we can diagonalize (24) in the basis of eigenfunctions of B. For any finite eigenvalue b of B the number of states is finite, namely (b + 1)(b + 2)/2. Thus the infinite dimensional Hilbert space separates into an infinite set of finite dimensional subspaces, each subspace containing only vectors with a given eigenvalue b. These eigenfunctions are products of the number states ni of each degree of freedom and can be characterized by a symbol n, m, li where we have n bosons on site 1, m bosons on site 2, and l bosons on site 3. For a given value b it follows that l = b − m − n. So we can actually label each state by just two numbers (n, m): n, m, (b− n− m)i ≡ n, m). Note that the third site added to the dimer is different from the first two sites. There is no boson-boson interaction on this site. Thus site 3 serves simply as a boson reservoir for the dimer. Dimer bosons may now fluctuate from the dimer to the reservoir. The trimer has the same permutational symmetry as the dimer. The matrix elements of (24) between states from different b subspaces van- ish. Thus for any given b the task amounts to diagonalizing a finite dimensional matrix. The matrix has a tridiagonal block structure, with each diagonal block being a dimer matrix (23). The nonzero off-diagonal blocks contain interaction terms proportional to δ. We consider symmetric Ψis and antisymmetric Ψia states. The structure of the corresponding symmetric and antisymmetric de- compositions of H is similar to H itself. In the following we will present results for b = 40. We will also drop the first two terms of the RHS in (24), because these only lead to a shift of the energy spectrum. Since we evaluate the matrix elements explicitly, we need only a few seconds to obtain all eigenvalues and eigenvectors with the help of standard Fortran routines. In Fig. 7 we plot a part of the energy spectrum as a function of δ for C = 2 [19]. As discussed above, the Hamiltonian decomposes into noninteracting blocks for δ = 0, each block corresponding to a dimer with a boson number between 0 and b. For δ 6= 0 the block-block interaction leads to typical features in the spectrum, like, e.g., avoided crossings. The full quantum energy spectrum extends roughly over 103, leading to an averaged spacing of order 100. Also the upper third of the spec- trum is diluted compared to the lower two thirds. The correspondence to the classical model is obtained with the use of the transformation Ecl = Eqm/b2 + 1 and for parameters C/b and δ/b (the classical value for B is B = 1). The main result of this computation so far is that tunneling pairs of eigen- states of the dimer persist in the nonintegrable regime δ 6= 0. However at certain 12 350 (b) y g r e n E 340 330 320 0 1 δ 2 3 Figure 7: A part of the eigenenergy spectrum of the quantum trimer as a func- tion of δ with b = 40 and C = 2. Lines connect data points for a given state. Solid lines - symmetric eigenstates; thick dashed lines - antisymmetric eigen- states. Adapted from [19]. 100 10−5 E ∆ 10−10 10−15 10−20 0 1 δ 2 3 Figure 8: Level splitting versus δ for a level pair as described in the text. Solid line - numerical result. Dashed line - semiclassical approximation. Filled circles - location of wave function analysis in Fig. 9. Adapted from [20]. pair-dependent values of δ a pair breaks up. From the plot in Fig. 7 we cannot judge how the pair splittings behave. In Fig.8 we plot the pair splitting of the pair which has energy ≈ 342 at δ = 0 [20]. Denote with x, y, z the eigenvalues of the site number operators n1, n2, n3. We may consider the quantum states of the trimer at δ = 0 when z is a good quantum number and then follow the evolution of these states with increasing δ. The state for δ = 0 can be traced back to C = 0 and be thus characterized in addition by x and y. The chosen pair states are then characterized by x = 26(0), y = 0(26) and z = 14 for C = δ = 0. Note that this pair survives approximately 30 avoided crossings before it is finally destroyed at coupling strength δ ≈ 2.67 as seen in Fig. 7. 13 Figure 9: Contour plot of the logarithm of the symmetric eigenstate of the chosen tunneling pair (cf. Fig. 7) for five different values of δ = 0, 0.3, 0.636, 1.0, 1.8 (their location is indicated by filled circles in Fig. 8). (a): three equidistant grid lines are used; (b-e): ten grid lines are used. Minimum value of squared wave function is 10−30, maximum value is about 1. Adapted from [20]. From Fig. 8 we find that the splitting rapidly increases gaining about eight orders of magnitude when δ changes from 0 to slightly above 0.5. Then this rapid but nevertheless smooth rise is interrupted by very sharp spikes when the splitting ∆E rises by several orders of magnitude with δ changing by mere percents and then abruptly changes in the opposite direction sometimes even overshooting its pre-spike value. Such spikes, some larger, some smaller, repeat with increasing δ until the splitting value approaches the mean level spacing of order one. Only then one may say that the pair is destroyed since it can be hardly distinguished among the other trimer levels. Another observation is presented in Fig. 9 [20]. We plot the intensity distri- bution of the logarithm of the squared symmetric wave function of our chosen pair for five different values of δ = 0 , 0.3 , 0.636 , 1.0 , 1.8 (their locations are indicated by filled circles in Fig. 8). We use the eigenstates of B as basis states. They can be represented as x, y, z > where x, y, z are the particle numbers on sites 1, 2, 3, respectively. Due to the commutation of B with H two site occu- pation numbers are enough if the total particle number is fixed. Thus the final encoding of states (for a given value of b) can be chosen as x, z). The abscissa in Fig. 9 is x and the ordinate is z. Thus the intensity plots provide us with information about the order of particle flow in the course the tunneling process. For δ = 0 (Fig. 9(a)) the only possibility for the 26 particles on site 1 is to di- rectly tunnel to site 2. Site 3 is decoupled with its 14 particles not participating in the process. The squared wave function takes the form of a compact rim 14 1 (a) 2 ===> ===> 3 1 ===> = = = > 2 (b) 3 Figure 10: Order of tunneling in the trimer. Filled large circles - sites 1 and 2, filled small circle - site 3. Arrows indicate direction of transfer of particles. Adapted from [20]. in the (x, z) plane which is parallel to the x axis. Nonzero values of the wave function are observed only on the rim. This direct tunneling has been described in chapter 2.2. When switching on some nonzero coupling to the third site, the particle number on the dimer (sites 1,2) is not conserved anymore. The third site serves as a particle reservoir which is able either to collect particles from or supply particles to the dimer. This coupling will allow for nonzero values of the wave function away from the rim. But most importantly, it will change the shape of the rim. We observe that the rim is bended down to smaller z values with increasing δ. That implies that the order of tunneling (when, e.g., going from large to small x values) is as follows: first, some particles tunnel from site 1 to site 2 and simultaneously from site 3 to site 2 (Fig. 10(a)). Afterwards par- ticles flow from site 1 to both sites 2 and 3 (Fig. 10(b)). With increasing δ the structure of the wave function intensity becomes more and more complex, pos- sibly revealing information about the classical phase space flow structure. Thus we observe three intriguing features. First, the tunneling splitting increases by eight orders of magnitude when δ increases from zero to 0.5. This seems to be unexpected, since at those values perturbation theory in δ should be applicable (at least Fig. 7 indicates that this should be true for the levels themselves). The semiclassical explanation of this result was obtained in [20]. The second observation is that the tunneling begins with a flow of particles from the bath (site 3) directly to the empty site which is to be filled (with si- multaneous flow from the filled dimer site to the empty one). At the end of the tunneling process the initially filled dimer site is giving particles back to the bath site. Again this is an unexpected result, since it implies that the particle number on the dimer is increasing during the tunneling, which seems to decrease the tunneling probability, according to the results for an isolated dimer. These first two results are closely connected (see [20] for a detailed explanation). The third result concerns the resonant structure on top of the smooth variation in Fig. 8. The resonant enhancements and suppressions of tunneling are related to avoided crossings. Their presence implies that a fine tuning of the system parameters may strongly suppress or enhance tunneling which may be useful for spectroscopic devices. In Fig. 11 we show the four various possibilities of avoided crossings between a pair and a single level and between two pairs, and 15 s l e v e l y g r e n e s l e v e l y g r e n e 6e−05 4e−05 2e−05 E ∆ 0e+00 0.049 6e−05 E ∆ 4e−05 2e−05 0e+00 0.049 δ δ (a) 6e−05 4e−05 2e−05 E ∆ s l e v e l y g r e n e δ (b) 0.05 δ 0.051 0e+00 0.049 0.05 δ 0.051 (c) 0.01 E ∆ (d) δ s l e v e l y g r e n e 0.05 δ 0.051 0.00 0.01 0.03 0.05 δ 0.07 0.09 Figure 11: Level splitting variation at avoided crossings. Inset: Variation of individual eigenvalues participating in the avoided crossing. Solid lines - sym- metric eigenstates, dashed lines - antisymmetric eigenstates. Adapted from [20]. the schematic outcome for the tunneling splitting [20]. If the interaction to fur- ther more distant states in the spectrum is added, the tunneling splitting can become exactly zero [21] for some specific value of the control parameter. In such a rare situation the tunneling is suppressed for all times. 2.4 Quantum roto-breathers When discussing classical breather solutions we have been touching some aspects of roto-breathers, including their property of being not invariant under time reversal symmetry. In a recent study Dorignac et al have performed [22] an analysis of the corresponding quantum roto-breather properties in a dimer with the Hamiltonian H = 2 Xi=1 (cid:26) p2 i 2 + α(1 − cos xi)(cid:27) + ε(1 − cos(x1 − x2)) . (25) The classical roto-breather solution consists of one pendulum rotating and the other oscillating with a given period Tb. Since the model has two symme- tries - permutation of the indices and time-reversal symmetry - which may be both broken by classical trajectories, the irreducible representations of quantum eigenstates contain four symmetry sectors (with possible combinations of sym- metric or antisymmetric states with respect to the two symmetry operations). Consequently, a quantum roto-breather state is belonging to a quadruplet of weakly split states rather than to a pair as discussed above. The schematic representation of the appearance of such a quadruplet is shown in Fig. 12 [22]. The obtained quadruplet has an additional fine structure as compared to the tunneling pair of the above considered dimer and trimer. The four levels in 16 1 + 2 = 1 2 Rotobreather s> a> s> a> σ >,α > n n 1 σ > α > 1 -uplets 8 -uplets 4 -uplets 4 -uplets 2 singlets Separatrix -uplets 2 singlets σ > 0 Ground states Figure 12: Schematic representation of the sum of two pendula spectra. Straight solid arrows indicate the levels to be added and dashed arrows the symmetric (permutation) operation. The result is indicated in the global spectrum by a curved arrow. The construction of the quantum roto-breather state is explicitly represented. Adapted from [22]. g n i t t i l p S 1 0.01 0.0001 1e-06 1e-08 1e-10 1e-12 0.1 Different splittings Energy and momentum transfer Total momentum reversal ~~ - Ea-Ea ~ ~ Es-Ea ~ ~ - Es-Ea 1 ε 10 Figure 13: Dependence of different splittings of a quadruplet on ε. Only three of them have been displayed, each being associated with a given tunneling process. Adapted from [22]. the quadruplet define three characteristic tunneling processes. Two of them are energy or momentum transfer from one pendulum to the other one, while the third one corresponds to total momentum reversal (which restores time reversal symmetry). The dependence of the corresponding tunneling rates on the cou- pling ε is shown for a specific quadruplet from [22] in Fig. 13. For very weak coupling ε ≪ 1 the fastest tunneling process will be momentum reversal, since tunneling between the pendula is blocked. However as soon as the coupling is increased, the momentum reversal turns into the slowest process, with breather tunneling from one pendulum to the other one being orders of magnitude faster. 17 Note that again resonant features on these splitting curves are observed, which are related to avoided crossings. 2.5 Large lattices with fluctuating numbers of quanta A number of publications are devoted to the properties of quantum breathers in chains and two-dimensional lattices of coupled anharmonic oscillators. For the respective one-dimensional case, the Hamiltonian is given by (cid:20) 1 2 p2 n + V (xn) + W (xn − xn−1)(cid:21) . (26) H = Xn 2 x2 + 1 Here V (x) = 1 4 v4x4 (or similar) and the nearest neighbour coupling W (x) = 1 2 Cx2. The classical version of such models conserves only the energy, but not any equivalent of a norm. Therefore, no matter whether one uses creation and annihilation operators of the harmonic oscillator [23], or similar operators which diagonalize the single anharmonic oscillator problem [24], the resulting Hamiltonian matrix will not commute with the corresponding number operator. Calculations will typically be restricted to 4-6 quanta, and lattice sizes of the order of 30 for d = 1, 13 × 13 for d = 2 [23]. With these parameters one can calculate properties of quantum breather states, which correspond to typically two quanta which are bound (with unavoidable states with different number of quanta, contributing as well). For large enough v4 a complete gap opens between the two-quanta continuum and quantum breather states [10, 23] (Figs.14 and 15). When decreasing the anharmonic constant v4, Proville found, that the gap closes for certain wave numbers, but persists for others, becoming a pseudogap [25, 23] (Fig.14). Involved calculations of the dynamical structure factor (e.g. available by neutron scattering in crystals) have shown, that signatures of quantum breathers are imprinted in these integral characteristics of the underlying lattice dynamics [10, 24], yet the working out of these differences may become a subtle task (see Fig.16 for an example). Finally, Fleurov et al [26] estimated the influence of the tunnel splitting of a dimer, when embedded in an infinite chain. This situation is close to the tunneling of a very localized DB, so that the nonlinearity (interaction between bosons) can be taken into account only on the two sites, which participate in the tunneling, while the nonlinearity can be neglected on all other sites. Using path integral techniques, the computed tunneling splitting has been shown to become smaller as compared to the case of an isolated dimer. This is due to the fact, that a DB in an infinite chain has a core and a localized tail. That tail has to be carried through the tunneling process as well, and in analogy with a single particle tunneling in a double well, the tail increases the effective mass of such a particle. Consequently the exponential tail of a DB in an infinite chain tends to decrease its ability to perform quantum tunneling motion, yet it never leads to a full suppression of tunneling [26]. 18 Figure 14: Eigenspectrum of a chain with 33 sites for parameters (a) C = 0.05, v4 = 0.2, and (b) C = 0.05, v4 = 0.02. The inserts show magnifications of the fundamental branch (left) and overtone region (right). The quantum breather branch is marked by (2), and the two-phonon band by (11). Adapted from [23]. 3 Quantum breather properties 3.1 Evolution of quantum localized states Suppose that we initially excite only one site in the trimer from above. If this initial state has strong enough overlap with tunneling pair eigenstates, its evo- lution in time should show distinct properties as compared to the case when the overlap is vanishing, or when there are simply no tunneling pair states available. Several results have been reported. First, a quantum echo was observed in [19] by calculating the survival probability of the initial state as a function of time. That quantity measures the probability to find the system in the initial state at later times. If the initial state has strong overlap with many eigenstates, it is expected to quickly decohere into these different states. Yet, if a substan- tial overlap with quantum breathers takes place, the survival probability first rapidly decays to zero, but echoes up after regular time intervals (Fig.17, left plot). If one simply measures the dependence of the number of quanta, then a similar situation will show up with a very slow beating of the occupation num- bers in time, if the overlap of the initial state and a tunneling pair is strong 19 Figure 15: Eigenspectrum of a lattice with 13 × 13 sites for parameters (top) C = 0.025, v4 = 0.1, and (bottom) C = 0.025, v4 = 0.025. Left plots - spectra over the whole Brilloin zone. Right plots - profiles of the spectra along the direction [11]. The insets show the magnifications of the phonon branch (left) and the quantum breather energy region (right). Adapted from [23]. Figure 16: A 3D plot of the inelastic structure factor S(q, ω) as a function of the dimensionless energy transfer 0 ≤ ω ≤ 3 and the scalar product of the transfer momentum q and the polarization u. Adapted from [24]. [21, 27] (see Fig.17, right plot). Suppose we have a large lattice, and put initially many quanta on one site. Then any tunneling of this packet as a whole will occur on very long time scales. 20 t P 1.2 0.8 0.4 0.0 1.2 0.8 0.4 0.0 1.2 0.8 0.4 0.0 (a) (b) (c) Y T I S N E T N I L A R T C E P S 0.05 0.04 0.03 0.02 0.01 0.00 Y T I S N E T N I L A R T C E P S 0.06 0.05 0.04 0.03 0.02 0.01 0.00 Y T I S N E T N I L A R T C E P S 0.08 0.06 0.04 0.02 0.00 -400 0 400 ENERGY 800 -400 0 400 ENERGY 800 i 〉 n 〈 -400 0 400 ENERGY 800 (a) 〈n1 〉 〈n2 〉 (b) 〈n1 〉 (c) 〈n2 〉 〈n1 〉 30 20 10 0 30 20 10 0 30 20 10 〈n3 〉 〈n3 〉 〈n3 〉 〈n2 〉 0 1 2 3 4 5 6 time 0 0 200 400 600 800 1000 time Figure 17: Left plot: Survival probability of the initial state Ψ0i = 20 + ν, 0, 20− νi. ν = (a) -6, (b) 0, and (c) 6. Insets: spectral intensity of the initial state Ψ0i. Filled circles - symmetric eigenstates; open circles - antisymmetric eigenstates. Right plot: Time evolution of expectation values of the number of bosons at each site of the trimer for different initial states Ψ0i = 20 + ν, 0, 20− νi. ν = (a) -6, (b) 0, and (c) 6. Adapted from [21]. On time scales much shorter, we may describe the excitation as a classical discrete breather state plus a small perturbation. Treating that perturbation quantum mechanically, one could expect that the time-periodic DB acts as a constant source of quantum radiation for the quantized phonon field. It turns out to be impossible, for very much the same reasons as in the purely classical treatment (see [30]). This result implies, that there is almost no other source of decay for a localized initial state in a quantum lattice, but to slowly tunnel as a whole along the lattice, if nonlinearities allow for the formation of exact classical DB states [31]. Numerical calculations for such a case, but with few quanta, were performed by Proville [32], and, similar to the above trimer discussion, showed that if quantum breather states exist in the system, then localized excitations stay localized for times which are much longer than the typical phonon diffusion times in the absence of anharmonicity. 21 E ∆ ) 1 1 ( , µ f / ) 2 , 1 ( µ f 100 10-3 10-6 10-9 1.0 0.8 0.6 0.4 0.2 0.0 (a) (b) 400 500 600 ENERGY 700 800 900 Figure 18: (a) Energy splitting and (b) correlation function vs. energy of the eigenstates of the dimer (open circles, symmetric eigenstates; solid circles, an- tisymmetric eigenstates). The vertical dashed line marks the energy threshold for appearance of QB states. The thin solid line in (a) is a guide for the eye, whereas in (b) it is the estimation using eq. (30). Here b = 40 and C = 2. 3.2 Splitting and correlations QBs are nearly degenerate eigenstates. For the dimer and the trimer, they come in symmetric-antisymmetric pairs. So one may compute the nearest neighbor energy spacing (tunneling splitting) between pairs of symmetric-antisymmetric eigenstates in order to identify QBs. Since QBs correspond to classical orbits that are characterized by energy localization, they may be identified by defin- ing correlation functions. For large lattices it has been shown that QBs have exponentially localized correlation functions, in full analogy to their classical counterparts. For the dimer and the trimer, the correlation functions may be defined as follows: fµ(1, 2) = hn1 n2iµ, (27) where ni = a† i ai, and h Aiµ = hχµ Aχµi, {χµi} being the set of eigenstates of the system. The ratio 0 ≤ fµ(1, 2)/fµ(1, 1) ≤ 1 measures the site correlation of quanta: it is small when quanta are site-correlated (i.e. when many quanta are located on one site there are almost none on the other one) and close to unity otherwise. fµ(1, 1) = hn2 1iµ, For the dimer case, the relation b = n1 + n2, leads to fµ(1, 2) = bhn1iµ − hn2 1iµ. (28) In Fig. 18-left we show the energy splitting and the correlation function of the eigenstates. We see that beyond a threshold (dashed line), the splitting drops 22 exponentially fast with energy. The corresponding pairs of eigenstates, which are tunneling pairs, are site correlated. Thus they are QBs. Their correlation functions show a fast decrease for energies above the threshold. In these states many quanta are localized on one site of the dimer and the tunneling time of such an excitation from one site to the other (given by the inverse energy splitting between the eigenstates of the pair) is exponentially large. As shown in Ref. [17], this energy threshold is close to the threshold for the existence of DBs in the corresponding classical model. QBs are close to symmetric (S) and antisymmetric (A) eigenstates of the C = 0 case given by n1, n2iS,A = 1 √2 (n1, n2i ± n2, n1i), (29) with n1,2 ≫ n2,1. So we may estimate the dependence of the correlation func- tions on n1 using the eigenstates (29) and b = n1 + n2. The result is: f (1, 2)n1 f (1, 1)n1 = 2n1(b − n1) 1 + (b − n1)2 , n2 (30) where we note that it is equal to unity when n1 = b/2, and vanishes when n1 = 0, b. Using the relation between the eigenenergy ε of the C = 0 case and the number n1 (= 1, 2, . . . , b) εn1 = 5 4 + 3 2 b + 1 2 (cid:2)n2 1 + (b − n1)2(cid:3) , (31) one may obtain the energy dependence of the correlation function (30), which is plotted in Fig. 18-b (thin solid line). We can see that beyond the energy threshold for appearance of QBs, the numerical results are close to the estima- tion (30). 3.3 Entanglement QBs may also be differentiated from other quantum states when measuring the degree of entanglement [33, 34]. For the dimer and the trimer the degree of entanglement in the eigenstates may be measured by minimizing the distance of a given state to the space of product states of the dimer part (expanded by the product basis {n1i⊗n2i}), which depends on the largest eigenvalue of the corresponding reduced density matrix [35, 36, 37, 29]: ∆ = N Xn1,n2 (χn1,n2 − fn1 gn2)2, (32) where for the case of the dimer χn1,n2 = hn1, n2χi, and for the trimer χn1,n2 = hn1, n2, (n3 = b − n1 − n2)χi. The functions fn1 and gn2 are such that ∆ is 23 ∆ 1.0 0.9 0.8 0.7 0.6 0.5 (a) 1 (b) E ∆ 100 10-4 10-8 3 2 4 (a) 0.2 ) 1 n ( ρ 0.1 0.0 0 0.10 (c) ) 1 n ( ρ 0.05 10 20 30 40 400 500 600 ENERGY 700 800 900 0.00 0 10 30 40 20 n1 0.10 (b) (d) 0.05 0.00 0 0.4 0.3 0.2 0.1 0.0 0 10 20 30 40 10 20 n1 30 40 Figure 19: Left panel: (a) Entanglement of the eigenstates and (b) energy split- ting as a function of energy in the dimer (open circles, symmetric eigenstates; solid circles, antisymmetric eigenstates). The vertical dashed line marks the energy threshold for appearance of QB states. Here b = 40, and C = 2. Right panel: The density of the symmetric eigenstates marked by labeled arrows in left panel-a: (a) S-0 (arrow 1), (b) S-7 (arrow 2), (c) S-9 (arrow 3), (d) S-19 (arrow 4). minimum [29]. ∆ measures how far a given eigenstate of the system is from being a product of single-site states, and has values 0 < ∆ < 1. This measure has a direct relation to the distance of a given eigenstate from a possible one obtained after performing a Hartree approximation [35]. For the dimer, since QB states are close to eigenstates of the C = 0 case χiQB ≃ 1 √2 (n, 0i ± 0, ni), (33) with n . b, one expects that the degree of entanglement in QB states is similar to the degree of entanglement in such states. Since only two basis states are involved, it can not be a state of maximum entanglement. For C = 0 the eigenstates of the system are the basis states given by eq. (29), where for n1 = n2 it follows that ∆ = 0, and for n1 6= n2 (which includes the state in eq. 33) ∆ = 0.5. In previous works in a similar quantum dimer model [33, 34], it was shown that at the energy threshold for appearance of QB states the entanglement (in this case measured in a different way) becomes maximum and then decreases with energy. From this, and the above reasoning, we expect that QB states show decreasing entanglement ∆ with energy, tending to 0.5. Results in left panel-a of Fig. 19 agree with this expectation. For C = 0, the entanglement has the values 0 and 0.5 corresponding to the basis states b/2, b/2i and n, b − ni (n 6= b/2) with equal and distinct number of quanta at each site respectively. When C > 0, the eigenstates become linear superpositions of the basis states and the entanglement rises, being larger as long as more basis states are involved in building up an eigenstate. This can be 24 seen in the right panel of Fig. 19, where we plot the density ρ(n1, n2 = b−n1) = hn1, b− n1χi2 ≡ ρ(n1) of four symmetric eigenstates marked by labeled arrows in the left panel: The low-energy eigenstate marked by the arrow 1 consists mainly of one basis state: b/2, b/2i, as seen in the right panel-a, hence the entanglement is relatively small. When going up in energy the entanglement in the eigenstates quickly increases, becoming maximum at the energy threshold, and then decreases. An eigenstate just before the threshold like the one marked by the arrow 2 in the left panel involves many basis states fulfilling n1 + n2 = b = 40 (right panel-b), hence the entanglement is large. However, for a QB state lying in the energy region beyond the threshold, like the one marked by the arrow 3 in the left panel, the number of involved basis states, and thus the entanglement starts to decrease (right panel-c). Finally, in high-energy eigenstates like the one marked by the arrow 4 in the left panel, which has the form shown in eq. (33) (right panel-d), the entanglement is even smaller and gets close to 0.5 as expected. From the above results we see that by measuring entanglement one may gain information not only about the energy threshold for existence of QBs (also visible when measuring the energy splitting and correlation function), but also about how many basis states overlap strongly with the eigenstate under con- sideration. We also computed the von Neumann entropy [38], which is another standard measure of entanglement, and the results were consistent with those discussed above. 4 Quantum edge-localized states Most of the studies about QBs in large lattices (with few bosons) were done considering a system with periodic boundary conditions, and thus, translational invariant. However, ususally real systems have to be modeled with open bound- ary conditions. Hence it is natural to wonder what hapens with QBs when the lattice has finite size, and therefore, no translational invariance. In the classical case, it has been shown that in finite nonlinear lattices, the breaking of the translational symmetry may lead to the formation of so called nonlinear edge states. These are excitations which are localized at the edges of the lattice and they have been studied, in particular, in nonlinear optics experiments employing optical waveguides, being coined with the name of discrete surface solitons [39]. In particular, it is well known that the discrete nonlinear Schrodinger equation (DNLS) has time-periodic solutions localized at the edge of the lattice [40, 41]. It is therefore expected that the large-boson limit of the open-boundary Bose-Hubbard model will show eigenstates in which the bosons are localized at the edge. Numerical studies by Pouthier were done to answer the edge-localization question in a lattice with a few bosons [42], where the mean-field approximation (DNLS equation) cannot a priori be expected to provide the correct intuition. The answer turns out to be subtle -- this phenomenon is not present for the case of two particles, but appears when the particle number is three or more 25 <nj> 2 1 0 ν = 9 (edge state) ν = 10 (edge state) ν = 1 ν = 2 100 <nj> 10-4 10-8 2 4 6 8 10 j 2 4 6 j 8 10 Figure 20: Spatial profile of site occupancies for several eigenstates of the three- boson Bose-Hubbard chain. The index ν counts the eigenstates from the lowest- energy one (ν = 1). Inset shows the same plot in semilog scale, the linear behavior indicating exponential localization in the edge states. This image was taken from Ref. [43] [42]. Further studies focused on the energy spectrum and eigenstates (Fig. 20) were done by Pinto et al [43], where degenerate perturbation theory allowed to explain why edge-localized states exist only if the number of particles is three or more [43]. 5 Quantum breathers with fermions Advances in experimental techniques of manipulation of ultracold atoms in op- tical lattices make it feasible to explore the physics of few-body interactions. Systems with few quantum particles on lattices have new unexpected features as compared to the condensed matter case of many-body interactions, where excitation energies are typically small compared to the Fermi energy. Therefore it is of interest to study binding properties of fermionic pairs with total spin zero, as recently presented in Ref.[44]. We use the extended Hubbard model, which contains two interaction scales - the on site interaction U and the near- est neighbour intersite interaction V . The nonlocal interaction V is added in condensed matter physics to emulate remnants of the Coulomb interaction due to non-perfect screening of electronic charges. For fermionic ultracold atoms or molecules with magnetic or electric dipole-dipole interactions, it can be tuned with respect to the local interaction U by modifying the trap geometry of a con- densate, additional external dc electric fields, combinations with fast rotating external fields, etc (for a review and relevant references see [45]). Consider a one-dimensional lattice with f sites and periodic boundary condi- 26 tions described by the extended Hubbard model with the following Hamiltonian: where H = H0 + HU + HV , a+ j,σ(aj−1,σ + aj+1,σ) , H0 = −Xj,σ HU = −U Xj HV = −V Xj nj,↑nj,↓ , nj,σ = a+ j,σaj,σ , nj nj+1 , nj = nj,↑ + nj,↓ . (34) (35) (36) (37) H0 describes the nearest-neighbor hopping of fermions along the lattice. Here the symbols σ =↑,↓ stand for a fermion with spin up or down. HU describes the onsite interaction between the particles, and HV the intersite interaction of fermions located at adjacent sites. a+ j,σ and aj,σ are the fermionic creation and annihilation operators satisfying the corresponding anticommutation relations: {a+ l,σ′} = {aj,σ, al,σ′} = 0. Note that throughout this work we consider U and V positive, which leads to bound states located below the two-particle continuum. A change of the sign of U, V will simply swap the energies. j,σ, al,σ′} = δj,lδσ,σ′ , {a+ j,σ, a+ 2,↑a+ To observe the fermionic character of the considered states, any two-particle number state is generated from the vacuum Oi by first creating a particle with spin down, and then a particle with spin up: e.g. a+ 1,↓Oi creates a particle with spin down on site 1 and one with spin up on site 2, while a+ 2,↓Oi creates both particles with spin down and up on site 2. Due to periodic boundary conditions the Hamiltonian (34) commutes also with the translation operator T , which shifts all lattice indices by one. It has eigenvalues τ = exp(ik), with Bloch wave number k = 2πν f and ν = 0, 1, 2, ..., f− 1. For the case of having only one fermion (either spin up or spin down) in the lattice (n = 1), a number state has the form ji = a+ j,σOi. The interaction terms HU and HV do not contribute. For a given wave number k, the eigenstate to (34) is therefore given by: 2,↑a+ Ψ1i = 1 √f f Xs=1(cid:16) T τ (cid:17)s−1 1i . The corresponding eigenenergy εk = −2 cos(k). (38) (39) For two particles, the number state method involves Ns = f 2 basis states, which is the number of ways one can distribute two fermions with opposite spins over the f sites including possible double occupancy of a site. Below we 27 8 6 4 2 0 -2 -4 -6 Y G R E N E Antisymmetric bound states Symmetric bound states Symmetric bound states kc/πa kc/πs2 -1 -0.8 -0.6 -0.4 -0.2 0 k/π 0.2 0.4 0.6 0.8 1 Figure 21: (Color online) Energy spectrum of the two fermion states. The eigenvalues are plotted as a function of the wave number k. Here U = 1 , V = 1, f = 101. Symbols are from analytical derivation, lines are the result of numerical diagonalization. The arrows indicate the location of the critical wave numbers (see text). Adapted from [44]. consider only cases of odd f for simplicity. Extension to even values of f is straightforward. The details of the calculations can be found in [44]. In Fig. 21 we show the energy spectrum of the Hamiltonian matrix obtained by numerical diagonalization for the interaction parameters U = 2 and V = 2 and f = 101. At U = 0 and V = 0, the spectrum is given by the two fermion continuum, whose eigenstates are characterized by the two fermions independently moving along the lattice. In this case the eigenenergies are the sum of the two single-particle energies: E0 k1,k2 = −2[cos(k1) + cos(k2)], (40) with k1,2 = πν1,2/(f + 1) , ν1,2 = 1, . . . , f . The Bloch wave number k = k1 + k2 mod 2π. Therefore, if k = ±π, the continuum degenerates into points. The con- tinuum is bounded by the hull curves h±(k) = ±4 cos k 2 . The same two-particle continuum is still observed in Fig. 21 for nonzero interaction. However, in addi- tion to the continuum, we observe one, two or three bound states dropping out of the continuum, which depends on the wave number. For any nonzero U and V , all three bound states drop out of the continuum at k = ±π. One of them stays bounded for all values of k. The two other ones merge with the continuum at some critical value of k upon approaching k = 0 as observed in Fig.21. Note that for k = ±π and U = V , all three bound states are degenerate. Upon increasing U and V , we observe that a second bound state band sep- arates from the continuum for all k (Fig.22). At the same time, when U 6= V , the degeneracy at k = ±π is reduced to two. Finally, for even larger values of U and V , all three bound state bands completely separate from the continuum (Fig.23). 28 8 6 4 2 0 -2 -4 -6 Y G R E N E Antisymmetric bound states Symmetric bound states Symmetric bound states kc/πs2 -8 -1 -0.8 -0.6 -0.4 -0.2 0 k/π 0.2 0.4 0.6 0.8 1 Figure 22: (Color online) Energy spectrum of the two fermion states. The eigenvalues are plotted as a function of the wave number k. Here U = 4 , V = 3 , f = 101. The symbols are from analytical derivation, lines are the results of numerical diagonalization. The arrow indicates the location of the critical wave number (see text). Adapted from [44]. 5.1 Symmetric and antisymmetric state representation In order to obtain analytical estimates on the properties of the observed bound states, we use the fact that the Hamiltonian for a two fermion state is invariant under flipping the spins of both particles. We define symmetric basis states Φj,si = 1 √2 (Φj,+i + Φj,−i) and antisymmetric states Φj,ai = 1 √2 (Φj,+i − Φj,−i) . Note that Φ1i is a symmetric state as well. 5.2 Antisymmetric bound states (41) (42) The antisymmetric states exclude double occupation. Therefore the spectrum is identical with the one of two spinless fermions [13]. Following the derivations in [13] we find that the antisymmetric bound state, if it exists, has an energy Ea 2 (k) = −(V + 4 V cos2( k 2 )) . (43) This result is valid as long as the bound state energy stays outside of the contin- uum. The critical value of k at which validity is lost, is obtained by requesting 29 4 0 -4 -8 Y G R E N E -12 -16 -1 Antisymmetric bound states symmetric bound states Symmetric bound states -0.8 -0.6 -0.4 -0.2 0 k/π 0.2 0.4 0.6 0.8 1 Figure 23: (Color online) Energy spectrum for the two fermion states. The eigenvalues are plotted as a function of the wave number k. Here U = 8 , V = 8 , f = 101. The symbols are from analytical derivation, lines are the results of numerical diagonalization. Adapted from [44]. 2 (k) = h±(k). It follows V = 2 cos( k Ea state merges with the continuum at a critical wave number 2 ). Therefore the antisymmetric bound ka c = 2 arccos( V 2 ) , (44) setting a critical length scale λa Fig.21). c = 2π ka c . For V = 1 it follows ka c /π ≈ 0.667 (see The equation (43) is in excellent agreement with the numerical data in Figs. 21,22,23 (cf. open triangles). We also note, that the antisymmetric bound state is located between the two symmetric bound states, which we discuss next. 5.3 Symmetric bound states A bound state can be searched for by assuming an unnormalized eigenvector of the form c, 1, µ, µ2, µ3, ...i with µ ≡ ρ ≤ 1. We obtain [44] Es 2(k) = −2(ρ + 1 ρ ) cos k/2 . The parameter ρ satisfies a cubic equation aρ3 + bρ2 + cρ + d = 0 (45) (46) with the real coefficients a, b. c and d given by a = 2V cos( k U V , c = 2(U + V )cos( k (cf. open circles and squares). We obtain excellent agreement. 2 ) − 2 ). We plot the results in Figs. 21,22,23 2 ), b = 4cos2( k 2 ), d = −4cos2( k 30 At the Brilloin zone edge k = ±π the cubic equation (46) is reduced to a quadratic one, and can be solved to obtain finally ρ → 0 and 2 (k → ±π) = −V . 2 = Ea Es1 2 (k → ±π) = −U , Es2 In particular we find for k = ±π that Es2 three bound states degenerate at the zone edge. 2 . In addition, if U = V , all If V = 0, the cubic equation (46) is reduced to a quadratic one in the whole range of k and yields [13] Es1 2 (k) = −pU 2 + 16 cos2(k/2) . Next we determine the critical value of k for which the bound state with is joining the continuum. Since at this point ρ = 1, we solve (46) energy Es2 2 with respect to kc and find (47) (48) (49) U V 2(U + 2V )(cid:17) ks2 c = 2 arccos(cid:16) c = 2π ks c . E.g. for U = V = 1 ks2 setting another critical length scale λs in excellent agreement with Fig.21. For U = 4 and V = 3 we find ks2 confirming numerical results in Fig.22. c /π ≈ 0.89, c /π ≈ 0.59 6 Small Josephson junction networks Recent studies of Pinto et al [28, 29] deal with quantum breather excitations in two capacitively coupled Josephson junctions. Such systems are currently under experimental investigation, being candidates for quantum information process- ing, and show remarkably long coherence times up to 100 ns for few quanta excitations. The system does not conserve the number of excited quanta, and can be best compared with the above Bose-Hubbard trimer. Quantum breather signatures are found simultaneously in the spectra (tunneling splittings), corre- lation functions, entanglement, and quanta number fluctuations. We address the excitation of QBs in a system of two coupled Josephson junc- tions in the quantum regime [29]. Josephson junctions are nonlinear devices that show macroscopic quantum behavior, and nowadays they can be manipulated with high precision, in such a way that the energy flow between coupled junc- tions can be resolved in time. Josephson junctions behave like anharmonic oscillators, and by lowering the temperature one can bring them into the quantum regime, where their quan- tization leads to energy levels which are nonequidistant because of the anhar- monicity. These levels can be separately excited by using microwaves pulses, and the energy distribution between the junctions can be measured in time using subsequent pulses [46]. So far these techniques have been used for experiments on quantum information processing with Josephson junctions [47, 48, 49], but we think that arrays of Josephson junctions in the qunatum regime also might 31 Cc bI bI ϕU( ) Ib 1 Cc Ib 2 ∆U CJ CJ (a) π − ϕ ϕ 0 ϕ 0 (b) (c) Figure 24: (a) Sketch of the two capacitively coupled Josephson junctions. (b) Sketch of the washboard potential for a single current-biased Josephson- junction. (c) Circuit diagram for two ideal capacitively coupled Josephson junc- tions. be used as playgrounds for experiments on quantum dynamics of excitations in nonlinear lattices. The system is sketched in Fig.24-a: two JJs are coupled by a capacitance Cc, and they are biased by the same current Ib. The strength of the coupling due to the capacitor is ζ = Cc/(Cc +CJ ). The dynamics of a biased Josephson junction (JJ) is analogous to the dynamics of a particle with a mass proportional to the junction capacitance CJ , moving on a tilted washboard potential U (ϕ) = −Ic Φ0 2π cos ϕ − Ibϕ Φ0 2π , (50) which is sketched in Fig.24-b. Here ϕ is the phase difference between the macro- scopic wave functions in both superconducting electrodes of the junction, Ib is the bias current, Ic is the critical current of the junction, and Φ0 = h/2e the flux quantum. When the energy of the particle is large enough to overcome the barrier ∆U (that depends on the bias current Ib) it escapes and moves down the potential, switching the junction into a resistive state with a nonzero voltage proportional to ϕ across it. Quantization of the system leads to discrete energy levels inside the potential wells, which are nonequidistant because of the anhar- monicity. Note that even if there is not enough energy to classically overcome the barrier, the particle may perform a quantum escape and tunnel outside the well, thus switching the junction into the resistive state. Thus each state inside the well is characterized by a bias and state-dependent lifetime, or its inverse -- the escape rate. The Hamiltonian of the system is H = P 2 1 2m + P 2 2 2m + U (ϕ1) + U (ϕ2) + ζ m P1P2, (51) 32 ) (a) z H G ( h / E ∆ 100 10-1 10-2 10-3 10-4 100 (b) ) 1 , 1 ( µ f / ) 2 , 1 ( µ f 10-1 10-2 10-3 0 1 50 2 3 100 150 E /h (GHz) 200 250 300 Figure 25: (a) Energy splitting and (b) correlation function vs. energy of the eigenstates of the two-junctions system (open circles, symmetric eigenstates; filled circles, antisymmetric eigenstates). The labeled arrows mark the energy corresponding to the peak of the spectral intensity in Fig.26-b, d, and f (see text). The parameters are γ = Ib/Ic = 0.945 and ζ = 0.1 (22 levels per junction). where 2π(cid:19)2 m = CJ (1 + ζ)(cid:18) Φ0 2π(cid:19)2 P1,2 = (Cc + CJ )(cid:18) Φ0 , ( ϕ1,2 − ζ ϕ2,1). (52) (53) Note that the conjugate momenta P1,2 are proportional to the charge at the nodes of the circuit (which are labeled in Fig.24-c). In the quantum case the energy eigenvalues and the eigenstates of the system were computed and analyzed in [29]. In Fig.25 we show the nearest neighbor energy spacing (tunneling splitting) and the correlation function of the eigen- states. For this, and all the rest, we used Ic = 13.3 µA, CJ = 4.3 pF, and ζ = 0.1, which are typical values in experiments. We see that in the central part of the spectrum the energy splitting becomes small in comparison to the average. The corresponding pairs of eigenstates, which are tunneling pairs, are site correlated, and thus QBs. In these states many quanta are localized on one junction and the tunneling time of such an excitation from one junction to the other (given by the inverse energy splitting between the eigenstates of the pair) can be exponentially large and depend sensitively on the number of quanta excited. Note that the tunneling of quanta between the JJs occurs without an obvious potential energy barrier being present (the interaction between the junctions is 33 i > n < 6 5 4 3 2 1 0 20 i > n < 10 i > n < 0 20 10 0 (a) <n1> <n1> + <n2> No breathers <n2> (c) <n1> + <n2> <n2> <n1> Breather (e) <n1> <n2> <n1> + <n2> No breathers (b) (d) (f) 0.4 0.3 0.2 0.1 0.0 0.2 0.1 Y T I S N E T N I L A R T C E P S Y T I S N E T N I L A R T C E P S 0.0 0.06 0.04 0.02 Y T I S N E T N I L A R T C E P S 0 1 2 3 4 5 6 7 8 9 10 time (ns) 0.00 0 100 200 300 E /h (GHz) Figure 26: Time evolution of expectation values of the number of quanta at each junction (left panels) for different initial excitations with corresponding spectral intensities (right panels). Ψ0i = 0, 5i; (c) and (d): Ψ0i = 0, 19i; (e) and (f): Ψ0i = 9, 19i. Open circles, symmetric eigenstates; filled circles, antisymmetric eigenstates. The energies of the peaks in the spectral intensity are marked by labeled arrows in Fig.25-b (see text). The parameters are γ = Ib/Ic = 0.945 and ζ = 0.1 (22 levels per junction). (a) and (b): only through their momenta). This process has been coined dynamical tun- neling [50, 51, 52], to distinguish from the usual tunneling through a potential barrier. In dynamical tunneling, the barrier -- a so-called invariant separatrix manifold -- is only visible in phase space, where it separates two regions of regu- lar classical motion between which the tunneling process takes place. Therefore, when referring to the tunneling between the JJs, we implicitly mean that it is dynamical. The fact that the strongest site correlated eigenstates occur in the central part of the energy spectrum may be easily explained as follows: Let N be the highest excited state in a single junction, with a corresponding maximum energy ∆U (Fig.24). For two junctions the energy of the system with both junctions in the N -th state is 2∆U , which roughly is the width of the full spectrum. Thus states of the form N, 0i and 0, Ni that have energy ∆U are located approximately in the middle. Having the eigenvalues and eigenstates, we compute the time evolution of different initially localized excitations, and the expectation value of the num- 34 20 15 2 n 10 5 0 0 −5 −10 −15 −20 5 10 n 1 15 20 χi = (χ(S) b i + χ(A) b Figure 27: Contour plot of the logarithm of the density of the asymmetric state i)/√2 as a function of the number of quanta at junctions 1 and 2 (see text). The parameters are γ = Ib/Ic = 0.945 and ζ = 0.1 (22 levels per junction). ber of quanta at each junction hnii(t) = hΨ(t)niΨ(t)i. Results are shown in Fig.26a, c, and e. We also compute the spectral intensity I 0 µ = hχµΨ0i2, which measures the strength of overlap of the initial state Ψ0i with the eigenstates. Results are shown in Fig.26-b, d, and f, where we can see a peak in each case, which corresponds to the arrows in Fig.25-b. We can see that the initial state Ψ0i = 0, 5i overlaps with eigenstates with an energy splitting between them being relatively large and hence the tunneling time of the initially localized excitation is short. For the case Ψ0i = 0, 19i QBs are excited: The excita- tion overlaps strongly with tunneling pairs of eigenstates in the central part of the spectrum, which are site correlated and nearly degenerate. The tunneling time of such an excitation is very long, and thus keeps the quanta localized on their initial excitation site for corresponding times. Finally the initial state Ψ0i = 9, 19i overlaps with weakly site correlated eigenstates with large energy splitting. Hence the tunneling time is short again. We computed also the time evolution of the expectation values of the number of quanta for initial conditions which are coherent or incoherent (mixtures) superpositions of product basis states with equal weights. This is relevant for experiments, since it may be hard to excite one junction to a determined state but easier to excite several states of the junction at the same time. We used coherent superpositions (characterized by well defined states Ψ0i), and mixtures (characterized by their corresponding density operators ρ0), of four basis states around the already used initial states: For the state 0, 5i we superposed the basis states 0, 5i, 0, 6i, 0, 7i, and 0, 8i, for 0, 19i the basis states 0, 20i, 0, 19i, 0, 18i, and 0, 17i, and for 9, 19i the basis states 9, 20i, 9, 19i, 9, 18i, and 9, 17i. Both for superposition and mixture of basis states, the results are qualitatively similar to those shown in Fig.26. Therefore we expect that some imprecision in exciting an initial state in the junctions would not affect in a 35 relevant way the results. we may also conclude, that the excitation of QB states does not rely on the phase coherence. Let us estimate how many quanta should be excited in the junctions in order to obtain QBs (tunneling pairs). We compute the density ρ(n1, n2) = hn1, n2χi2 of the asymmetric state χi = (χ(S) i are the eigenstates belonging to a tunneling pair [29]. In Fig. 27 we show a contour plot of the logarithm of the density for the tunneling pair with energy marked by the arrow labeled by number two in Fig. 25-(b). We see that the density has its maximum around n1 = 19 and n2 = 0 which is consistent with the result shown in Fig. 26-c and d where QBs were excited by using this combination of quanta in the junctions. i)/√2, where χ(S,A) b i + χ(A) b b 7 Experimental realizations There is a fast growing amount of experimental and related theoretical work on applying the quantum discrete breather concept to many different branches in physics, like Bose-Einstein condensates, crystals and molecules, surfaces, and others. We will discuss some of these at length, while others will be reviewed more briefly. 7.1 Repulsively bound atom pairs Winkler et al [53] performed experiments with a three-dimensional optical lat- tice with initially each site being either not occupied, or being occupied by two Rb atoms bound in a pair due to attractive interaction. A magnetic field sweep across the Feshbach resonance changes the sign of interaction, turning attraction into repulsion. The dynamics of ultracold atoms loaded into the lowest band of the optical potential is described by the quantum DNLS model, which is equiv- alent to the Bose-Hubbard model (10). Lifetime measurements have shown, that repulsive pairs of Rb atoms have larger lifetimes than pairs of weakly or al- most not interacting atoms (Fig.28). The two-particle bound states discussed in chapter 2.1 - the simplest versions of a quantum discrete breather - are the obvi- ous explanation of the experimental findings. Indeed, neglecting Landau-Zener transitions to higher lying bands in the optical potential, the Bose-Hubbard model is justified. The sign of the interaction does not play any role, since it only changes quantum discrete breathers from being low-lying to being excited states, not affecting their localization properties. The most simple argument of why two quanta (or atoms) placed initially close to each other, do not separate despite they repel each other, is based on the fact, that if they would do so, the (large) interaction energy should be converted into kinetic energy, which is restricted to be less than two times the width of the single particle band. In other words, repulsively bound atom pairs are a straightforward consequence of quantum discrete breather states with two quanta. Another sophisticated experimental investigation aimed at measuring the quasi-momentum distribution of atom pairs in various regimes by mapping it 36 Figure 28: (a) Repulsive interaction between two atoms sharing a lattice site gives rise to an interaction energy U . Breaking up of the pair is suppressed owing to the lattice band structure and energy conservation. This is the simplest version of a quantum discrete breather. (b) The discrete breather can tunnel along the lattice. (c) Long lifetimes of strongly repulsive atoms. The plot shows the remaining fraction of pairs for strong interaction (open diamonds) and for almost vanishing interaction (filled circles). Adapted from [53]. onto a spatial distribution, which was finally measured using standard absorp- tion imaging (Fig.29). Therefore predictions of such states, which were made more than 30 years ago by Ovchinnikov [54], were beautifully confirmed exper- imentally with ultracold repulsive atoms. 7.2 Molecules Intramolecular vibrational energy redistribution (IVR) has been a central is- sue in the field of chemical physics for many decades. In particular, pathways and rates are of importance there, since understanding them allows to describe e.g. the dynamics of various chemical reactions, and dissociation processes [55]. Spectroscopical studies, where single vibrational quanta are excited, allow to measure the frequencies of normal vibrational modes, i.e. to characterize the dynamics of a molecule for small amplitude vibrations. These normal modes consist of coherent combinations of vibrational excitations of several bonds (or rotational groups) in a molecule. However, in order e.g. to dissociate a molecule, a many quanta excitation is needed, and nonlinearities will certainly become im- portant. It was realized then, that strong vibrational excitations of molecules are much better described by so-called local modes, i.e. basically one or few bond 37 Figure 29: (Upper row) Absorption images of the atomic distribution after release from the 3D lattice and a subsequent 15 ms time of flight. The horizontal and vertical black lines enclose the first Brillouin zone. (upper left) Lattice sites are occupied by single atoms; (upper middle) Repulsively bound atom pairs; (upper right) Attractively bound atom pairs; (bottom row) Quasi-momentum distribution for pairs in one direction as a function of the lattice depth after integrating out the other direction. (bottom left) Experiment; (bottom right) Numerical calculations. Adapted from [53]. vibrational excitations. That transition from normal to local modes remained a puzzle for a long time. A practically complete modern theoretical account on these issues can be found in a recent monography by Ovchinnikov, Erikhman and Pronin [56]. On its most abstract level, the transition from normal to local modes is identical with the bifurcation in the dimer model. Thus, local modes are essentially discrete breathers or slight perturbations of them. Note, that the connection between local modes, breathers and periodic orbits has been recently studied by Farantos in the context of large biological molecules [57]. Discrete breathers (ILMs) have been theoretically predicted to exist in ionic crystals [58], ways of optical excitation of DBs (ILMs) have been proposed [59], and their possible presence in hydrocarbon structures has been discussed [60]. Exciting local modes in molecules with discrete symmetries leads to small tunneling splittings of excitation levels [56], and goes back to the work of Child and Lawton [61], see also a recent comprehensive review by Keshavamurthy [62] and references therein. On its most abstract level, this effect is identical with the tunneling splitting in the permutationally symmetric dimer model discussed in chapter 2.2. An early example of experimental evidence of discrete breather excitations in molecules comes from spectroscopical studies of visible red absorption spectra of benzene, naphtalene, and anthracene by Swofford et al [64]. The C-H stretching vibrations have been excited to the sixth quantum level, and red shifts of the lines show, that instead of a delocalized excitation of six bonds to the first level (yielding six quanta), the excitation resides on just one of the six available 38 Figure 30: (a) Absorption spectra of PBLG in chloroform at 293 K (red line, helical conformation) and at 260 K (blue line, random coil). (b) Pump-probe spectra 600 fs after excita tion under the same conditions. Inset: Decay of negative and both positive bands at 293 K. Adapted from [63]. bonds. While it can tunnel (as a quantum discrete breather) to the other bonds, this tunneling time is a new large time scale in the problem, strongly affecting e.g. dissociation rates. A recent study of femtosecond infrared pump-probe spectroscopy of the N- H mode of a stable α-helix (poly-γ-benzyl-L-glutamate (PBLG)) revealed two excited-state absorption bands, which disappear upon unfolding of the helix [63]. PBLG forms extremely stable, long α-helices in both helicogenic solvents and films grown from these solvents. The monomeric unit of PBLG is a non-natural amino-acid with a long side chain that stabilizes the helix. PBLG has served as the standard model helix since the very early days of structural investigations of proteins. Fig.30(a) (red line) shows the absorption spectrum of the helix at 293 K, which is dominated by the strong N-H stretching band at 3290 cm−1. Fig.30(b) (red line) shows the pump-probe response 600 fs after excitation with an ultrashort broadband pulse. One negative (3280 cm−1) and two positive bands (3160 and 3005 cm−1) are observed. If the N-H stretching modes were isolated, a negative band associated with bleach and stimulated emission, and a positive band associated with excited-state absorption, would be expected. This 39 Figure 31: Simulated pump-probe spectrum for 293 K (red line) and 18 K (blue line). Inset: schematic of the energy levels. Adapted from [63]. is indeed observed here for the unfolded molecule. In contrast, the observation of two positive bands for the intact helix rather than just one, is exceptional. Edler et al [63] argue, that these features can not be explained due to inten- sity dependencies, or Fermi resonances. A consistent explanation is reached by assuming that two vibron states are excited, and these vibrons may form two different types of bound states, self trapped either on the same site, or on neighbouring ones. The latter states originate from the acoustic phonons of the helix, which correlate adjacent sites (see also [65]). 7.3 Crystal surfaces Depositing atoms or molecules on crystal surfaces can be controlled experimen- tally, and as a result a planar regular two-dimensional lattice structure of the deposited material can be obtained. Guyot-Sionnest [66] used Hydrogen to be deposited on Si(111) surfaces. The Si-H bonds can be excited using pum-probe techniques with infrared dye lasers. There is substantial interaction between the Si-H bonds on the Si(111) surface. The pump excites one phonon (quantum), while the tunable probe frequency finds a substantial red shift of the two-phonon excitation, and allows to conclude, that two-phonon bound states are observed. Another set of experiments was performed by Jakob [67, 68, 69]. Carbon monoxide (CO) was deposited on a Ru(001) single crystal surface. The C-O stretching modes constitute a two-dimensional array of weakly interacting an- 40 Figure 32: (a) Infrared absorption spectra of the C-O stretching mode at 30 K. The corresponding mode of naturally abundant 13C16O is displayed in an enlarged vertical scale; (b) the overtone band observed at less than twice the frequency of the fundamental mode. Adapted from [67]. Figure 33: Frequency shifts of the vibrational bands with temperature: crosses - overtone band, dots - fundamental of naturally abundant 13C16O, open circles - delocalized fundamental of 12C16O. Adapted from [67]. harmonic oscillators with 4.7 A intermolecular distance. Intermolecular coupling is provided by means of the electric field of the oscillating dipoles. Experimen- tal spectra at 30 K are shown in Fig.32. The one phonon mode frequency is at 2031 cm−1. This has to be compared to the naturally abundant 13C16O frequency at 1941 cm−1. The corresponding blue shift for the adsorbate is thus due to additional stiffness provided by the Ru surface coupling. Excitation of two uncorrelated phonons would yield a two phonon continuum at about 4062 cm−1. The narrow line observed at 3940 cm−1 can be thus attributed to a two-phonon bound state, or a quantum discrete breather excitation. The temperature dependence of the line positions also clearly shows, that the two-phonon bound state line softens much slower than the line of the one-phonon 41 Figure 34: Structure of the PtCl crystal. One PtCl chain is shown on the left. Each Pt atom is coordinated by two ethylenediamine units in a near square planar geometry, while Cl ions connect pt sites along the chain. The packing arrangement of the 1D chains and their ClO− 4 counterions is shown on the right. Adapted from [76]. delocalized state (Fig.33). This is, among other facts, a strong indication that the observed red shift of the overtone line is due to the formation of a localized two-phonon bound state, or a (quantum) discrete breather. 7.3.1 In the bulk of solids Vibrational spectra in the overtone or combination region of molecular crystals have been studied intensively in the 1970s and 1980s. A pioneering theoretical proposal was due to Agranovich, who predicted the existence of two-exciton bound states in various molecular crystal materials [70]. Experimental studies of infrared absorption spectra for CO2 crystals were conducted by Dows et al [71] and gave evidence of two-phonon bound states. Dressler et al studied the slow vibrational relaxation of N2, which also indicates the presence of many- phonon bound states [72]. In a remarkable theoretical paper, Bogani calculated the spectrum of two phonon excitations in molecular crystals [73], to some extent one of the first accurate calculations of quantum discrete breathers. More recently Bini et al reconsidered the theory of three-phonon bound states in crystal CO2 [74]. While there certainly are many other results worth to be mentioned, we recommend reading related chapters in [75],[56]. The pioneering studies of Swanson et al [76] have shown that up to seven phonons can bind and form a localized state. The system of choice was a PtCl based crystal - a halide-bridged mixed-valence transition metal complex, which is a model low-dimensional electronic material where the ground states can be systematically tuned (with chemistry, doping, pressure, and temperature). It is a very strong charge-density wave (CDW) example. The material is a well- 42 Figure 35: Fundamental and overtone spectra of isotopically pure Pt35Cl. Mov- ing upward in each panel, each x axis is offset by the appropriate integral mul- tiple of the 312 cm−1 fundamental frequency. All spectra have been scaled vertically to equal peak intensities. Adapted from [76]. formed crystal with a homogeneous lattice consisting of quasi-1D chains (see Fig.34). The CDW ground state consists of alternating Pt+2 and Pt+4 sites with a corresponding distortion of the chloride ions towards the Pt+4 site. Res- onance Raman spectra were used to probe both ground and photoexcited states. They probe the fundamental Cl-Pt-Cl stretch and the progression of many over- tones. At low temperatures, the fundamental exhibits a fine structure with up to six discrete, well-resolved modes. The analysis of the evolution of the spec- tral structure in the overtones was performed for isotopically pure samples, in order to avoid exciting localized states due to isotopic disorder. The fundamen- tal and overtone spectra for the pure Pt35Cl sample are shown in Fig.35. The data are presented in a stack plot in which each successive trace is offset along the horizontal axis by increasing multiples of the fundamental frequency 312 cm−1. Such plots clearly expose the relation of features in the overtone spec- trum to multiples of the fundamental peak. The lowest energy dominant feature in each trace (marked by vertical lines) demonstrates a strongly increasing an- harmonic redshift. Further, at higher overtones, each of these dominant peaks recurs, offset by the fundamental frequency, in the next trace above. A simple interpretation is that the lowest-energy dominant peaks in the overtone spectra correspond to all quanta of vibrational energy localized in approximately one PtCl2 unit, while the higher energy peaks correspond to having all quanta but one in a localized PtCl2 unit combined with one quanta in the more extended fundamental. The schematic process of the energy transfer is shown in Fig.36 and has been analyzed theoretically in [77]. 43 Figure 36: A simple picture of a resonant Raman scattering event in the localized atomic limit. Large filled circles mark Pt ions, small grey circles mark Cl ions. Open circles mark the positions of electrons. Adapted from [77]. An incoming photon at frequency ν is exciting an electron from a Pt+2 to a Pt+4 site. The Cl ion between them starts oscillating. Finally the electron relaxes back to its original position, and releases a photon with frequency ν′. The energy difference is remaining in a localized vibration. The effect of isotope disorder was treated by Kalosakas et al [78]. The experimentally observed red- shifts were also theoretically described by Fehske et al [79] using Peierls-Hubbard models, and by Wellein et al using the Holstein model [80]. Inelastic X-ray and neutron scattering was used by Manley et al [81, 82, 83] to probe phonon dispersion in α-uranium single crystals. Variation of temperature showed softening, and the abrupt appearance of a new dynamical mode, without a typically observed phase transition. The authors argue that this mode is a discrete breather, and forms due to strong electron-phonon interaction. Russell and Eilbeck reported evidence for moving breathers in the layered crystal muscovite at 300 K [84]. Breathers were created by bombardment of the crystal surface with heavy ions. Ejection of atoms at the opposite (shielded) crystal surface was attributed to breathers, which were able to carry the vibra- tional energy without dispersing over more than 107 unit cells of the crystal. 44 Finally Abrasonis et al [85] reported on anomalous bulk diffusion of inter- stitial Nitrogen in steel microcrystals. N ions were deposited in a micron-thick layer, and are trapped by local structures, with a characteristic binding energy. Ar ion bombardment increases the N mobility at depths far beyond the ion penetration depth. The authors see evidence for coherent transfer of vibrational energy deep into the bulk of the material. 8 Conclusions and outlook Progress in the understanding of classical discrete breathers evolved for two decades, and has significantly improved our understanding of the complexity in the dynamics of anharmonic lattice dynamics. The little input it needs to form discrete breathers - spatial discreteness and nonlinearity - were demonstrated to generate an impressive list of experimental observations in a large variety of different physical systems, with length and time scales ranging over many orders of magnitude. The quantization of the classical equations of motion opened a whole new avenue of research on quantum breathers. For obvious computational reasons numerical studies are much more involved in the quantum domain, re- stricting our potential of detailed understanding to either small lattices, or few participating quanta. Nevertheless, a set of experiments, comparable in size to its classical counterpart, has been reviewed where quantum breathers are detected. Still we anticipate this only to be the beginning of a much more so- phisticated undertaking. We may expect many new qualities to show up in the regime of large lattices and many participating quanta. These limits are concep- tually closing the gap to quantum interacting many body problems, which are at the heart of condensed matter physics, and quantum computing. Yet another highly interesting, complicated and debated problem is the combined influence of discreteness, nonlinearity, and disorder. Linear disordered wave equations allow for Anderson localization. It is a thrilling modern theme to study the constructive and destructive interplay of all three ingredients. The best is yet to come. References [1] P.W. Anderson, Phys. Rev. 109, 1492 (1958). [2] S. Flach and C. R. Willis, Phys. Rep. 295, 181 (1998). [3] D. K. Campbell, S. Flach, Y. S. Kivshar, Physics Today 57 (1) pp. 43-49 (2004). [4] S. Flach and A. V. Gorbach, Phys. Rep. 467, 1 (2008). [5] S. Flach and A. V. Gorbach, Chaos 15, 015112 (2005). [6] M. Eleftheriou and S. Flach, Physica D 202 142 (2005). 45 [7] S. Aubry, Physica D 103, 201 (1997). [8] R. S. MacKay, Physica A 288 174 (2000). [9] V. Fleurov, Chaos 13, 676 (2003). [10] W. Z. Wang, J. Tinka Gammel, A. R. Bishop and M. I. Salkola, Phys. Rev. Lett. 76, 3598 (1996). [11] S. A. Schoefield, R. E. Wyatt and P. G. Wolynes, J. Chem. Phys. 105 940 (1996). [12] P. D. Miller, A. C. Scott, J. Carr and J. C. Eilbeck, Phys. Scr. 44 509 (1991). [13] A. C. Scott, J. C. Eilbeck, and H. Gilhøj, Physica D 78, 194 (1994). [14] J. C. Eilbeck, in Localization and Energy Transfer in Nonlinear Systems, edited by L. Vazquez, R. S. MacKay, and M. P. Zorzano (World Scientific, Singapore, 2003), p. 177. [15] L. Bernstein, J. C. Eilbeck, and A. C. Scott, Nonlinearity 3, 293 (1990). [16] L. J. Bernstein, Physica D 68, 174 (1993). [17] S. Aubry, S. Flach, K. Kladko, and E. Olbrich, Phys. Rev. Lett. 76, 1607 (1996). [18] G. Kalosakas, A. R. Bishop, Phys. Rev. A 65, 043616 (2002). [19] S. Flach and V. Fleurov, J. Phys.: Cond. Matt. 9, 7039 (1997). [20] S. Flach, V. Fleurov, and A. A. Ovchinnikov, Phys. Rev. B 63, 094304 (2001). [21] R. A. Pinto and S. Flach, Phys. Rev. A 73, 022717 (2006). [22] J. Dorignac and S. Flach, Phys. Rev. B 65, 214305 (2002). [23] L. Proville, Phys. Rev. B 71, 104306 (2005). [24] L. Proville, Phys. Rev. B 72, 184301 (2005). [25] L. Proville, Europhys. Lett. 69, 763 (2005). [26] V. Fleurov, R. Schilling and S. Flach, Phys. Rev. E -- bf 58, 339 (1998). [27] G. Kalosakas, A. R. Bishop, V. M. Kenkre, J. Phys. B 36, 3233 (2003). [28] R. A. Pinto and S. Flach, EPL 79, 66002 (2007). [29] R. A. Pinto and S. Flach, Phys. Rev. B 77, 024308 (2008). [30] S. Flach, V. Fleurov, and A. V. Gorbach, Phys. Rev. B 71, 064302 (2005). 46 [31] L. S. Schulman, D. Tolkunov, E. Mihokova, Phys. Rev. Lett. 96, 065501 (2006); Chem. Phys. 322, 55 (2006); L.S. Schulman, D. Tolkunov, E. Mih´okova, Chem. Phys 322, 55 (2006). [32] L. Proville, Physica D 216, 191 (2006). [33] A. P. Tonel, J. Links, and A. Foerster, J. Phys. A: Math. Gen. 38, 1235 (2005). [34] L. B. Fu and J. Liu, Phys. Rev. A 74, 063614 (2006). [35] T.-C. Wei and P. M. Goldbart, Phys. Rev. A 68, 042307 (2003). [36] A. Shimony, Ann. NY. Acad. Sci. 755, 675 (1995). [37] H. Barnum and N. Linden, J. Phys. A.: Math. Gen. 34, 6787 (2001). [38] A. P. Hines, R. H. McKenzie, and G. J. Milburn, Phys. Rev. A 67, 013609 (2003). [39] K. G. Makris, S. Suntsov, D. N. Christodoulides, G. I. Stegeman, and A. Hache, Opt. Lett. 30, 2466 (2005); M. Molina, I. L. Garanovich, A. A. Sukhorukov, and Yu. Kivshar, Opt. Lett. 31, 2332 (2006). [40] M. I. Molina, R. A. Vicencio, and Y. S. Kivshar, Opt. Lett., 31, 1693 (2006). [41] S. Suntsov, et al, Phys. Rev. Lett. 96, 063901 (2006); E. Smirnov et al, Opt. Lett. 31, 2338 (2006), C. R. Rosberg et al, Phys. Rev. Lett. 97, 083901 (2006). [42] V. Pouthier, Phys. Rev. B 76, 224302 (2007). [43] R. A. Pinto, M. Haque and S. Flach, Phys. Rev. A 79, 052118 (2009). [44] J.-P. Nguenang and S. Flach, Phys. Rev. A 80, 015601 (2009). [45] M. A. Baranov, Phys. Rep. 464, 71 (2008). [46] R. McDermott et al, Science 307, 1299 (2005); M. Steffen et al, Science 313, 1423 (2006). [47] Quantum Computing and Quantum Bits in Mesoscopic Systems, edited by A. Leggett, B. Ruggiero, and P. Silvestrini (Kluwer Academic/Plenum Publishers, New York, 2004). [48] Les Houches 2003, Quantum Entanglement and Information Processing, edited by D. Est`eve, J. M. Raimond, and J. Dalibard (Elsevier, Amsterdam, 2004). [49] J. Q. You and F. Nori, Physics Today, November (2005). 47 [50] M. Davis, E. Heller, J. Chem. Phys. 75, 246 (1981). [51] S. Keshavamurthy, Int. Rev. Phys. Chem. 26, 521 (2007). [52] S. Keshavamurthy, J. Chem. Phys. 122, 114109 (2005). [53] K. Winkler, G. Thalhammer, F. Lang, R. Grimm, J. Ecker Denshlag, A. J. Daley, A. Kantian, H. P. Buchler, and P. Zoller, Nature 441, 853 (2006). [54] A. A. Ovchinnikov, Zh. Eksp. Teor. Fiz. / Soviet Phys. JETP 57/30:263, 147 (1969/1970). [55] M. Joyeux, S. Yu. Grebenshchikov, J. Bredenbeck, R. Schinke, S. C. Faran- tos, Adv. Chem. Phys. 130, 267 (2005). [56] A. A. Ovchinnikov, N. S. Erikhman and K. A. Pronin, Vibrational- Rotational Excitations in Nonlinear Molecular Systems, Kluwer Aca- demic/Plenum Publishers (2001). [57] S. C. Farantos, J. Chem. Phys. 126, 175101 (2007). [58] S. A. Kisilev, A. J. Sievers and G. V. Chester, Physica D 123, 393 (1998). [59] T. Rossler and J. B. Page, Physica B 220, 387 (1996); Phys. Rev. Lett. 78, 1287 (1997); Phys. Rev. B 62, 11460 (2000). [60] G. Kopidakis, S. Aubry, Physica B 296, 237 (2001). [61] M. S. Child, R. T. Lawton, Chem. Phys. Lett. 87, 217 (1982). [62] S. Keshavamurthy, Int. Rev. Phys. Chem. 26, 521 (2007). [63] J. Edler, R. Pfister, V. Pouthier, C. Falvo, P. Hamm, Phys. Rev. Lett. 93, 106405 (2004). [64] R. L. Swofford, M. E. Long, A. C. Albrecht, J. Chem. Phys. 65, 179 (1976). [65] P. Hamm, J. Edler, in: T. Dauxois, A. Litvak-Hinenzon, R. MacKay, A. Spanoudaki (Eds.), Energy Localization and Transfer, World Scientific, Singapore, (2004). [66] P. Guyot-Sionnest, Phys. Rev. Lett. 67, 2323 (1991). [67] P. Jakob, Phys. Rev. Lett. 77, 4229 (1996). [68] P. Jakob, Physica D 119, 109 (1998). [69] P. Jakob, Surface Science 428, 309 (1999). [70] V. M. Agranovich, Sov. Phys.-Solid St. 12, 438 (1970). [71] D. A. Dows, V. Schettino, J. Chem. Phys. 58, 5009 (1973). 48 [72] K. Dressler, O. Oehler, D. A. Smith, Phys. Rev. Lett. 34, 1364 (1975). [73] F. Bogani, J. Phys. C.: Solid State Phys. 11, 1283 (1978). [74] R. Bini, P. R. Salvi, V. Schettino, H. J. Jodl, J. Chem. Phys. 98, 164 (1993). [75] V. M. Agranovich, R. M. Hochstrasser, Spectroscopy and excitation dynam- ics of condensed molecular crystals, North-Holland, Amsterdam (1983). [76] B. I. Swanson, J. A. Brozik, S. P. Love, G. F. Strouse, A. P. Shreve, A. R. Bishop, W.-Z. Wang, M. I. Salkola, Phys. Rev. Lett. 82, 3288 (1999). [77] K. Kladko, J. Malek, A. R. Bishop, J. Phys.: Condens. Matter 11, L415 (1999). [78] G. Kalosakas, A. R. Bishop, A. P. Shreve, Phys. Rev. B 66, 894303 (2002). [79] H. Fehske, G. Wellein, H. Buettner, A. R. Bishop, M. I. Salkola, Physica B 281, 673 (2000). [80] G. Wellein, H. Fehske, Phys. Rev. B 58, 6208 (1998). [81] M. E. Manley, M. Yethiraj, H. Sinn, H. M. Volz, A. Alatas, J. C. Lashley, W. L. Hults, G. H. Lander, J. L. Smith, Phys. Rev. Lett. 96, 125501 (2006). [82] M. E. Manley, M. Yethiraj, H. Sinn, H. M. Volz, A. Alatas, J. C. Lashley, W. L. Hults, G. H. Lander, D. J. Thoma, J. L. Smith, J. Alloys and Comp. 444-445, 129 (2007). [83] M. E. Manley, J. W. Lynn, Y. Chen, G. H. Lander, Phys. Rev. B 77, 852301 (2008). [84] F. M. Russell, J. C. Eilbeck, EPL 78, 10004 (2007). [85] G. Abrasonis, W. Moeller, X. X. Ma, Phys. Rev. Lett. 96, 065901 (2006). 49
1202.6357	3	1202	2013-07-24T19:22:36	Josephson-Majorana cycle in topological single-electron hybrid transistors	[ "cond-mat.mes-hall" ]	Charge transport through a small topological superconducting island in contact with a normal and a superconducting electrode occurs through a cycle that involves coherent oscillations of Cooper pairs and tunneling in/out the normal electrode through a Majorana bound state, the Josephson-Majorana cycle. We illustrate this mechanism by studying the current-voltage characteristics of a superconductor-topological superconductor-normal metal single-electron transistor. At low bias and temperature the Josephson-Majorana cycle is the dominant mechanism for transport. We discuss a three-terminal configuration where the non-local character of the Majorana bound states is emergent.	cond-mat.mes-hall	cond-mat	Josephson-Majorana cycle in topological single-electron hybrid transistors Nicolas Didier,1 Marco Gibertini,1 Ali G. Moghaddam,2, 3 Jurgen Konig,2 and Rosario Fazio1 1NEST, Scuola Normale Superiore and Istituto Nanoscienze-CNR, I-56126 Pisa, Italy 2Theoretische Physik, Universitat Duisburg-Essen and CENIDE, 47048 Duisburg, Germany 3Department of Physics, Institute for Advanced Studies in Basic Sciences (IASBS), Zanjan 45137-66731, Iran Charge transport through a small topological superconducting island in contact with a normal and a superconducting electrode occurs through a cycle that involves coherent oscillations of Cooper pairs and tunneling in/out the normal electrode through a Majorana bound state, the Josephson- Majorana cycle. We illustrate this mechanism by studying the current-voltage characteristics of a superconductor -- topological superconductor -- normal metal single-electron transistor. At low bias and temperature the Josephson-Majorana cycle is the dominant mechanism for transport. We discuss a three-terminal configuration where the non-local character of the Majorana bound states is emergent. PACS numbers: 71.10.Pm, 73.23.-b, 74.50.+r. I. INTRODUCTION Majorana bound states are zero-energy states that oc- cur at the boundary or in the vortex core of topologi- cal superconductors.1 -- 3 Besides the genuine interest in understanding their properties, they play a fundamen- tal role in the realization of a topological quantum com- puter.4,5 For these reasons an intense research has started to find physical systems that support Majorana excita- tions. A quantum wire with spin-orbit coupling in close proximity to a superconductor and in the presence of an external magnetic field has been considered among the most promising proposals.6,7 The recent experiments8 -- 11 on this system provide the first evidences of the existence of Majorana bound states in the condensed matter world. The presence of Majorana bound states in topological superconductors has a number of distinct signatures. It is responsible for a fractional Josephson effect,4,6,12,13 it leads to a resonant Andreev reflection14 -- 18 and to anoma- lies in interference experiments,19 -- 21 or it can be detected by measuring the local density of states.22 -- 26 In the presence of quantum fluctuations, affecting the phase coherence of the topological superconductor, new effects appear due to a parity constraint linking the dy- namics of the superconducting phase and of the Majo- rana fermions. A paradigmatic situation to explore these phenomena, in the setup of Refs. 4, 6, and 7, is when the superconducting island in contact with the wire is small enough so that charging effects27 come into play.28 Fu29 first pointed out that the parity constraint leads to non-local effects in electron transport. The role of charging effects on the fractional Josephson effect and on the Coulomb blockade through a topological super- conducting transistor was studied in Refs. 30 and 31, respectively. In this work we introduce and analyze what turns out to be the dominant charge transport mechanism occur- ring at low voltages in hybrid topological single elec- tron transistors: The Josephson-Majorana cycle. It takes place when a topological superconducting island is cou- pled to superconducting and to normal leads (see Fig. 1). FIG. 1. The system consists of a superconducting island con- nected to two normal metals and to a superconductor. A quantum wire with strong spin-orbit interaction is deposited on top of the superconducting island. The whole system is immersed in an external magnetic field. By choosing prop- erly the parameters the proximized wire enters a topological phase with two Majorana bound states localized at the two ends of the wire. Tunneling into the normal leads happens through the Majorana states while the coupling of the central island to the superconducting lead is due to the Josephson effect. Charges can flow through the island due to the combined effect of the coherent oscillations of Cooper pairs in the island and the tunneling between the Majorana state and the normal leads. Although the process bears some simi- larities to the Josephson-quasiparticle mechanism present in Cooper pair transistors32 -- 34 there are also important differences. In the concluding part of the paper we will address this issue and show that some features of the Josephson-Majorana cycle are true consequences of the non-local character of the Majorana bound states. The paper is organized as follows. In the next section we describe the setup, illustrated in Fig. 1, and introduce the underlying Hamiltonian describing its dynamics. In Sec. III we compute the current at low bias through a master-equation approach governing the behavior of the reduced density matrix of the central island. In the Coulomb blockade regime the current is dominated by co- tunneling. These processes will be discussed in Sec. IV. 3 1 0 2 l u J 4 2 ] l l a h - s e m . t a m - d n o c [ 3 v 7 5 3 6 . 2 0 2 1 : v i X r a SΓLLEJϕSSϕΓRRγLγRNon-locality of the zero-energy modes will show up only at this higher order in tunneling. We conclude in Sec. V. II. THE TOPOLOGICAL SINGLE-ELECTRON HYBRID TRANSISTOR The system we consider is illustrated in Fig. 1. It con- sists of a topological superconducting island (a nanowire in close proximity to a superconducting island) tunnel- coupled to normal and superconducting electrodes in a three-terminal configuration. The island hosts two Majo- rana bound states at the ends of the wire associated with the Majorana operators γi (i = L, R), {γi, γj} = δi,j and γi = γ†i . At low energies (much smaller than the supercon- ducting gap), the island couples to the superconducting lead only through Josephson tunneling. The coupling to the normal leads occurs only via the Majorana bound states. Throughout the paper we set the two normal elec- trodes at the same voltage, the only net current will be from the superconductor to the normal electrodes. The three-terminal setup will, however, be crucial to discuss the non-local character of the Josephson-Majorana cycle. The mesoscopic scale of the central island considered here requires to take into account the charging energy of the island. The Hamiltonian of the system reads: H = HM + HCh + HJ + Hleads + HT , (1) where the five terms describe coupling of Majorana states, island charging energy, Josephson coupling to the superconducting electrode, metallic normal electrode, and tunneling to the normal lead, respectively. The coupling between the Majorana fermions is given by HM = iλγLγR where λ is related to the overlap between them, which is exponentially small for distances much larger than the superconducting coherence length. Two Majorana states can form a zero-energy fermionic level described by an annihilation operator d = (γL + iγR)/√2 which can be either occupied or empty. Therefore, not only the number of excess Cooper pairs N but also the occupation of the d-level enters the charging energy of the island, HCh = EC(2N + nd − ng)2 , (2) where nd = d†d counts the occupation of the d-level, ng is the gate charge which can be varied continuously by changing the gate voltage, and EC = e2/(2C) is the charging energy expressed in terms of the total capaci- tance C of the island. For later convenience, we will label the eigenstates of the charging energy with the notation N, nd(cid:105). To describe the Josephson coupling to the su- perconducting electrode we use the effective Hamiltonian HJ = −EJ cos (ϕS − ϕ) , (3) where EJ is the Josephson coupling energy and ϕS and ϕ are the phase of the superconducting electrode and the 2 island condensate respectively. Here, any possible cou- pling to the Majorana state can be safely ignored.35 The normal lead Hamiltonian is given by the noninteracting model Hleads = kc†k,ick,i (4) (cid:88) (cid:88) i=L,R k with creation (annihilation) operator c†k,i (ck,i) corre- sponding to a spinless fermion in the free particle state k with energy k inside the left/right (L/R) normal lead. Finally, the tunneling through the Majorana state takes the form,31 tk,ic†k,i(d + δie−iϕd†) + H.c. , (5) (cid:88) (cid:88) i=L,R k HT = where tk,i is the hopping amplitude to the k-state in the i-th lead and δL/R = ±1. The advantage of this formula- tion introduced by Zazunov et al.31 is the fact that it in- cludes automatically the constraint on the Hilbert space, linking the occupation of the Majorana bound state to the parity of the superconducting condensate. A non-trivial dynamics in the problem arises since ϕ and N are canonically conjugated variables, [N, eiϕ] = eiϕ. The two terms in the tunneling Hamiltonian cor- respond to regular and anomalous tunneling. The first one describes the transfer of an electron from the d-level to the normal lead, and the second one the annihilation of a Cooper pair inside the island accompanied by the creation of two electrons, one in the d-level and one in a normal electrode. III. JOSEPHSON-MAJORANA CYCLE AND CURRENT-VOLTAGE CHARACTERISTICS We start by considering the current-voltage character- istics at second order in the tunneling amplitudes tk,i. At bias voltage and temperature smaller than the supercon- ducting gap, quasiparticle tunneling is suppressed and coherent Josephson (Cooper pair) tunneling is necessary for transport through the transistor. From Eq. (2) it fol- lows that the resonance condition for coherent Cooper pair oscillations between two charge states that differ by one Cooper pair is fulfilled at integer values of the gate charge ng. Around ng ∼ 1, e.g., the lowest energy states are N, nd(cid:105) = 0, 0(cid:105), 0, 1(cid:105) and 1, 0(cid:105). A Josephson- Majorana cycle involving these three states is illustrated in Fig. 2. Starting from 0, 1(cid:105), a regular tunneling process releases an electron into the (left or right) normal lead. Then, the island is recharged with an extra Cooper pair provided by the Josephson coupling. A second (anoma- lous) tunneling which annihilates a Cooper pair in order to create an electron inside the normal electrode and an- other one filling the d-level, completes the cycle. (A cycle with the reverse direction can be obtained by the conju- gates of each tunneling process.) Since the energy of the state 0, 1(cid:105) is lower than the ones connected by Josephson 3 A convenient way to represent and compute the re- duced density matrix, in particular when higher-order tunneling processes are taken into account, is to use a real-time diagrammatic technique that has been devel- oped to describe transport through a metallic single- electron transistor.36,37 In the absence of the Josephson coupling, EJ = 0, the eigenstates ψn(cid:105) are defined by the total island charge 2N + nd(cid:105) ≡ N, nd(cid:105). Formulated in this basis, the diagrammatic rules for calculating the kernels Wnm,n(cid:48)m(cid:48) are the same as given in Ref. 37 but with a different rate function, α+(ω) → (Γ/2π)f (ω) and α−(ω) → (Γ/2π)[1− f (ω)] and extra rules for the overall sign: each crossing of tunneling lines yields a factor −1 and, furthermore, the factor δL/R needs to be included. The transformation of the diagrams into the eigenbasis of HCh + HJ for finite EJ is straightforward. The time evolution of the reduced density matrix ρ(t) can be cast in the form (cid:90) t (cid:88) n(cid:48),m(cid:48) −∞ (cid:88) (cid:104) ρnm + i∆nmρnm = dt(cid:48)Wnm,n(cid:48)m(cid:48)(t, t(cid:48))ρn(cid:48)m(cid:48)(t(cid:48)) . (9) Expanding to second order in the tunneling, this yields for the steady-state limit ( ρ = 0)38, D(i)∗n(cid:48)n D(i) m(cid:48)m(Fmn(cid:48) + F ∗nm(cid:48)) i∆nmρnm = Γ 4 (cid:104) (cid:104) (cid:88) (cid:88) i,l,n(cid:48) i,l,m(cid:48) Γ 4 Γ 4 − − i,n(cid:48),m(cid:48) nn(cid:48)D(i)∗mm(cid:48)( ¯F ∗n(cid:48)m + ¯Fm(cid:48)n) +D(i) D(i) nl D(i)∗n(cid:48)l F ∗lm + D(i)∗ln D(i) ln(cid:48) ¯Fml D(i) m(cid:48)lD(i)∗ml Fln + D(i)∗lm(cid:48) D(i) lm ¯F ∗nl (cid:105) (cid:105) (cid:105) ρn(cid:48)m(cid:48) ρn(cid:48)m ρnm(cid:48) . (10) condition (cid:80) Having fixed a number of relevant charge states, Eq. (10) can be solved, together with the normalization n ρnn = 1 for the reduced density matrix. This solution can be then used to compute the steady- state value of the current as a function of bias and gate voltages. We remark that the current is antisymmet- ric under the transformation (V, ng) → (−V, 2 − ng). To obtain this antisymmetry, we note that the trans- formation results in the substitutions Fnm → ¯F ∗mn and Dnm → D∗mn. From Eqs. (7) and (10) we conclude that ρ remains unchanged while the current changes sign. In the range of the gate charges where only the three charge states 0, 0(cid:105), 0, 1(cid:105) and 1, 0(cid:105) are involved, a simple analytical solution can be obtained for zero temperature when neglecting the imaginary parts of Fnm, which are associated with energy renormalization induced by the coupling to the bath. In this case the relevant eigenstates and eigenvalues are given by ψ0(cid:105) = cos θ eiϕL/2 0(cid:105) + sin θ e−iϕL/2 2(cid:105) ψ1(cid:105) = 1(cid:105) ψ2(cid:105) = − sin θ eiϕL/2 0(cid:105) + cos θ e−iϕL/2 2(cid:105) (11) FIG. 2. Josephson-Majorana cycle for ng ∼ 1. The states differing by one Cooper pair, 0, 0(cid:105) and 1, 0(cid:105), are coupled via the Josephson term. The coupling between 0, 1(cid:105) and 0, 0(cid:105) is given by the regular term in the tunneling Hamiltonian, the connection between 0, 1(cid:105) and 1, 0(cid:105) by the anomalous one. tunneling, there will be a threshold voltage for the onset of current. In the following we will support this simple picture sketched above with more detailed calculations. As already mentioned, the two normal leads are kept at the same voltage. The average of the total (summed over the two normal leads) current operator, I = IL + IR = i=L,R k[c†k,ick,i, H], can be expressed as ie(cid:80) (cid:80) I = −2e Im tk,ic†k,i(d + δie−iϕd†) . (6) (cid:88) (cid:88) i=L,R k To second order in the tunnel couplings the average current can be conveniently expressed in terms of the reduced density matrix ρ of the topological supercon- ducting island which is obtained after tracing out the fermionic degrees of freedom of the normal metals. The steady-state current is (cid:88) (cid:88) (cid:104) i=L,R l,n,m (cid:104)I(cid:105) = e 2 Γi D(i)∗nl D(i) ml(Fln + F ∗lm) (cid:105) −D(i) ln D(i)∗lm ( ¯F ∗nl + ¯Fml) ρnm , (7) where D(i) nm and ρnm are the matrix elements of the op- erators D(i) = d + δie−iϕd† and ρ in the basis defined by the eigenstates ψn(cid:105) of the Hamiltonian HCh + HJ with eigenvalues En. The coefficients Fnm are defined as , (cid:18) 1 (cid:19) Re Ψ + i (8) Fnm = f (∆nm) − i π ∆nm − eV 2πkBT 2 with ¯F = 1 − F , ∆nm = En − Em, f () = 1/[1 + e(−eV )/kB T ] the Fermi function of the normal electrode, and Ψ the digamma function. Finally Γi = 2πτ2N (F ) is the tunneling rate with tk,i ∼ τ assumed constant close to the Fermi energy F , and N (F ) is the density of states in the normal metal (supposed to be equal for both elec- trodes). For simplicity we will assume that also the in- terfaces are identical, therefore ΓL ∼ ΓR = Γ/2. N=1,nd=0iN=0,nd=1iN=0,nd=0ic†kdRegulartunnelingc†kd†e−iϕtunnelingAnomalouseiϕcouplingJosephson4 FIG. 3. Current-voltage characteristics for kBT = EC /100, Γ = kBT /10, EJ = EC /5. In the lowest order in the trans- mission the current is different from zero only in the region ng ∼ 1 where two states coupled by the Josephson matrix el- ement are nearly degenerate. Outside the region of maximum current the threshold voltage V − is visible as a tiny line in the Coulomb blockade region. FIG. 4. Current-voltage characteristics and its dependence on √ the bias charge nV = CV /e and the gate charge ng = CVg/e. Parameters are: kBT = EC /100, Γ = kBT /10, EJ = Γ/ 3. The threshold voltage to observe a current is nV (cid:39) ±0.5. The slight displacement of the resonance condition from ng = 1 is due to the Lamb shift from the normal leads. (cid:21) 1 ± 2 (cid:20) (cid:113)  2/[Γ2 + 4E 2 − E2 J ] 1/[Γ2gc(ng) + 2E 2] 0 and by (cid:113) (cid:113) (E2 − E0)2 + E2 J (E2 − E0)2 + E2 J (12) E0 = 1 E1 = E1 E2 = 1 2 (E0 + E2) − 1 2 2 (E0 + E2) + 1 2 In the previous expression E0 = EC n2 respectively. g, E1 = EC(1 − ng)2 and E2 = EC(2 − ng)2, and the angle θ is defined by tan 2θ = EJ /(E2 − E0). The current, as a function of bias voltage, increases in a stepwise fashion at two thresholds V ± defined as eV ± = EC (1 − ng)2 + (EJ /4EC)2 . (13) For positive voltage V > 0 and 1/2 < ng < 3/2, the steady-state current is V > V + V − < V < V + J (cid:104)I(cid:105) = eΓE2 (cid:112) [4EC(1 − ng)]2 + E2 V < V −(14) J , gc(ng) = (21 + 13 x− where E = x2 − x3)/16, and x = 4EC(1 − ng)/E. We immediately see that the current vanishes when the Josephson cou- (cid:112) pling goes to zero. The width of the resonance peak is δng = EJ /4EC and the peak value for V > V + be- 4E 2 − E2 comes maximal for Γ = J . Going beyond the three-state approximation requires a numerical solution of the master equation. In Fig. 3 we show the cur- rent as a function of the gate charge ng = CVg/e and the bias charge nV = CV /e. The simple analytical ex- pression given above fully captures all the properties of the current-voltage characteristics. The current increases stepwise (smeared by temperature) with the bias voltage and shows resonance peaks at integer values of ng. The effect of energy renormalization terms (Lamb shifts) that were neglected in the analytical formulas pre- sented above, is highlighted in Fig. 4 where the resonance condition is slightly displaced from the value ng = 1. Fi- nally, we comment on the abrupt suppression of the cur- rent for large bias voltage nV > 3/2 that is visible in Fig. 4. Beyond the threshold nV = 3/2, the new charge state 1, 1(cid:105) becomes available from 1, 0(cid:105) by tunneling. The island is then trapped in this new state since the Josephson coupling to the state 0, 1(cid:105) is suppressed be- cause of the large energy difference of 1, 1(cid:105) and 0, 1(cid:105). As a consequence, transport is blocked. As already anticipated in the introduction we now dis- cuss the differences between the Josephson-Majorana cy- cle and the Josephson-quasiparticle cycle32 of Cooper pair transistors. An obvious difference, but important for experimental detection, is the voltage scale at which the Josephson-Majorana cycle takes place. Here there are no excitations above the gap and therefore the thresh- old voltage is set by the charging energy solely. Second, we note that the Josephson-Majorana cycle shows an e- periodicity with respect to ng.39 To give a further characterization of the Majorana states we need to investigate their non-local character. In the sequential-tunneling approximation this goal cannot be achieved. One therefore needs to make one step fur- ther and analyze cotunneling processes where signatures of non-locality first appear. To this aim it is important to exploit the fork configuration of Fig. 1. IV. COTUNNELING PROCESSES It is worth stressing once more that the two metals are kept at the same voltage and transport occurs be- tween the superconducting lead and the normal elec- trodes. Below the threshold voltage V −, second-order transport considered so far is suppressed by Coulomb −0.5(cid:31)I(cid:30)eΓngnV0.00.00.50.5−0.50.00.51.01.0−1.01.52.0−2.0−1.5−1.0−0.50.00.51.01.52.0nV0.99900.99951.00001.00051.0010ng−0.32−0.24−0.16−0.080.000.080.160.240.32blockade. Nevertheless fourth-order (two-particle) pro- cesses sustain a finite current. Let us consider for illus- tration the case ng ∼ 1 where the Coulomb threshold is largest. Here the system remains in its ground state 0, 1(cid:105) and the states 0, 0(cid:105) and 1, 0(cid:105) needed to complete the Josephson-Majorana cycle are only virtually occu- pied. Let us consider for the sake of clarity the simple three-charge states case discussed previously. To evalu- ate the current, we determine the cotunneling rates for transitions in which two electrons are transferred from the superconducting to the normal leads (or vice versa) while leaving the island in state 0, 1(cid:105). As intermediate (virtual) states, we take the two linear combinations of 0, 0(cid:105) and 1, 0(cid:105) that form the eigenstates of HCh + HJ . All processes that differ only in their intermediate but not initial and final states need to be added coherently. This includes pairs of processes in which the order of the tunneling events of the two normal lead electrons is interchanged. Thereby, it is important to distinguish lo- cal from nonlocal cotunneling. For local cotunneling, the amplitudes of those pairs carry opposite sign. For nonlo- cal cotunneling, however, this sign is compensated by the relative sign of δL = −δR in Eq. (5). As a result, local cotunneling is strongly suppressed, and nonlocal cotun- neling dominates. At ng = 1 we get θ = π/4 in Eq. (12), and the two rel- evant eigenstates are the symmetric and antisymmetric combination of Cooper pair states. The non-local contri- bution ILR (Left-Right) to the current is given by (cid:104)I(cid:105)LR = e ΓLΓR π dω −eV E0 0 − ω2 − E2 E 2 E 2 2 − ω2 On the other side the local (on the same lead) contribu- tion to the current is given by (cid:104)I(cid:105)ii = e Γ2 i 2π for i = L, R. dω −eV ω 0 − ω2 − E 2 ω E 2 2 − ω2 The local and non-local contributions to the current behave quite differently. For zero temperature and to lowest order in eV , we find that (cid:104)I(cid:105)LR = e 2ΓLΓR π E2 J C − E2 (E2 J /4)2 eV and, for i = L, R, (cid:104)I(cid:105)ii = e 4Γ2 i 3π E2 CE2 J C − E2 (E2 J /4)4 (eV )3 . (17) (18) These results obviously extent to any integer value of ng. At low voltages the local contribution to the current is strongly suppressed. The results given in Eqs. (15) and (16) are distinct signatures of the nature of Majo- rana bound states. The dominant contribution of the crossed Andreev reflection because of Majorana bound states has been discussed in literature15 in the absence of (cid:90) eV (cid:20) (cid:90) eV (cid:20) (cid:21)2 (cid:21)2 (15) (16) 5 interaction. Here we showed that in the Coulomb block- ade regime this effect manifests in the cotunneling regime and that it appears as a different voltage dependence of the current (linear vs cubic) of the non-local and local contributions respectively. The difference between local and non-local transport does not appear to lowest order since, there, tunneling of different lead electrons occurs incoherently. The non-local character of the Majorana states can be detected by the finite linear cotunneling conductance. Furthermore, using a counting field at the level of the quantum master equation34 to derive the statistics of the current leads to a maximally correlated noise, L(cid:105) = (cid:104)δIRδIL(cid:105) , independently of the length of the wire.15 R(cid:105) = (cid:104)δI 2 (cid:104)δI 2 (19) In the absence of Majorana bound states at the ends of the wire there would be a local contribution to the current due to the Andreev processes between the nor- mal electrodes and the superconducting island. These processes would be resonant at ng = 1 and, therefore, would contribute to the linear part of the current-voltage characteristics.40 In the topological phase, however, the presence of the Majorana bound states changes the sit- uation drastically. In this case, as we already remarked, the ground state is 0, 1(cid:105) and the two-electron processes discussed in Ref. 40 are Coulomb blocked. At low volt- age bias, the cotunneling conductance is thus local in the trivial phase and non-local in the non-trivial topological phase. A measurement of the cross-correlation noise41 would be able to discriminate between the two situations. V. CONCLUSIONS When the superconducting island contacted to the quantum wire is reduced in dimensions, charging effects become important and may affect the dynamics of the system in the Majorana sector. Charging effects in the presence of Majorana bound states can lead to a rich phe- nomenology. In this work we considered a case in which the contemporary presence of tunneling into normal leads and the Josephson coupling to a superconducting elec- trode can provide a new transport mechanism similar to the Josephson-quasiparticle cycle discussed in supercon- ducting transistors. In topological hybrid transistors charge transport is dominated at low voltages by a Josephson-Majorana cy- cle. In this work we analyzed the current-voltage char- acteristics of a topological superconducting island con- nected to a superconducting and two normal leads. When a bias is applied to the leads (both metals are kept at the same voltage) current flows due to this mechanism. It consists in a process in which the Rabi oscillations of Cooper pairs are accompanied by the tunneling of electrons from/to Majorana bound states. To lowest or- der (sequential tunneling) the Josephson-Majorana cycle 6 leads to a stepwise current-voltage characteristics mod- ulated by the gate voltage. The current is maximum at integer values of ng when the two charge states differing by one Cooper pair are almost degenerate. In the sequential-tunneling limit the cycle can reveal the existence of zero-energy states but cannot ascertain their non-local character. This fundamental property of Majorana bound states appears in the next order (co- tunneling) in the transmission. At this order the three- terminal setup of Fig. 1 has two different contributions to the current. There is a local one in which the two elec- trons forming the Cooper pair tunnel in the same normal lead. In addition, there is a non-local contribution where the Cooper pair is split and tunnels through a "virtual" cycle into the two different leads. This process corre- sponds to crossed Andreev tunneling. When the central island is in the topological phase the local contribution is strongly suppressed (the current is proportional to V 3) and only the non-local contribution gives rise to the lin- ear part of the current-voltage characteristics. Acknowledgments -- We acknowledge stimulating dis- cussions with M. Polini and F. Taddei. We are grateful to R. Egger for useful comments. The work was sup- ported by EU through projects IP-SOLID, QNEMS, GE- OMDISS, and NANOCTM and by DFG through Project No. KO 1987/5. 1 X.-L. Qi and S.-C. Zhang, Rev. Mod. Phys. 83, 1057 24 J. D. Sau, C. H. Lin, H.-Y. Hui, and S. Das Sarma, Phys. (2011). Rev. Lett. 108, 067001 (2012). 2 C. W. J. Beenakker, Annu. Rev. Condens. Mat. Phys. 4, 25 A. M. Black-Schaffer and J. Linder, Phys. Rev. B 84, 113 (2013). 3 J. Alicea, Rep. Prog. Phys. 75, 076501 (2012). 4 A. Yu. Kitaev, Physics-Uspekhi 44, 131 (2001). 5 C. Nayak, S. H. Simon, A. Stern, M. Freedman, and S. Das Sarma, Rev. Mod. Phys. 80, 1083 (2008). 6 R. M. Lutchyn, J. D. Sau, and S. Das Sarma, Phys. Rev. Lett. 105, 077001(2010). 7 Y. Oreg, G. Refael, and F. von Oppen, Phys. Rev. Lett. 105, 177002 (2010). 8 V. Mourik, K. Zuo, S. M. Frolov, S. R. Plissard, E. P. A. M. Bakkers and L. P. Kouwenhoven, Science 336, 1003 (2012). 9 A. Das, Y. Ronen, Y. Most, Y. Oreg, M. Heiblum, and H. Shtrikman, Nature Phys. 8, 887 (2012). 180509(R) (2011). 26 M. Gibertini, F. Taddei, M. Polini, and R. Fazio, Phys. Rev. B 85, 144525 (2012). 27 H. Grabert and M. H. Devoret, Single charge tunneling, Proc. NATO ASI 1991 (Plenum, New York, 1992). 28 The fate of Majorana states in the presence of quasi-long range superconducting order was studied in J. D. Sau, B. I. Halperin, K. Flensberg, S. Das Sarma, Phys. Rev. B 84, 144509 (2011) and in L. Fidkowski, R. M. Lutchyn, C. Nayak, and M. P. A. Fisher, Phys. Rev. B 84, 195436 (2011). 29 L. Fu, Phys. Rev. Lett. 104, 056402 (2010). 30 B. van Heck, F. Hassler, A. R. Akhmerov, and C. W. J. Beenakker, Phys. Rev. B 84, 180502(R) (2011). 10 M. T. Deng, C. L. Yu, G. Y. Huang, M. Larsson, P. Caroff, 31 A. Zazunov, A. Levy Yeyati, and R. Egger, Phys. Rev. B and H. Q. Xu, Nano Lett. 12, 6414 (2012). 84, 165440 (2011). 11 L. P. Rokhinson, X. Liu, and J. K. Furdyna, Nature Phys. 32 D. V. Averin and V. Ya. Aleshkin, JETP Lett. 50, 367 8, 795 (2012). (1989). 12 L. Fu and C. L. Kane, Phys. Rev. B 79, 161408(R) (2009). 13 L. Jiang, D. Pekker, J. Alicea, G. Refael, Y. Oreg, and F. von Oppen, Phys. Rev. Lett. 107, 236401 (2011). 33 A. Maassen van den Brink, G. Schon, and L. J. Geerligs, Phys. Rev. Lett. 67, 3030 (1991). 34 M.-S. Choi, F. Plastina, and R. Fazio, Phys. Rev. Lett. 87, 14 C. J. Bolech, and E. Demler, Phys. Rev. Lett. 98, 237002 116601 (2001). (2007). 15 J. Nilsson, A. R. Akhmerov, and C. W. J. Beenakker, Phys. Rev. Lett. 101, 120403 (2008). 35 A. Zazunov and R. Egger, Phys. Rev. B 85, 104514 (2012). 36 H. Schoeller and G. Schon, Phys. Rev. B 50, 18436 (1994). 37 J. Konig, H. Schoeller, and G. Schon, Phys. Rev. Lett. 78, 16 K. T. Law, P.A. Lee, and T. K. Ng, Phys. Rev. Lett. 103, 4482 (1997); Phys. Rev. B 58, 7882 (1998). 237001 (2009). 17 M. Wimmer, A. R. Akhmerov, J. P. Dahlhaus, and C. W. J. Beenakker, New J. Phys. 13, 053016 (2011). 18 K. Flensberg, Phys. Rev. B 82, 180516 (2010). 19 P. Bonderson, A. Kitaev, and K. Shtengel, Phys. Rev. Lett. 96, 016803 (2006). 20 A. Stern and B. I. Halperin, Phys. Rev. Lett. 96, 016802 (2006). 21 G. Strubi, W. Belzig, M.-S. Choi, and C. Bruder, Phys. Rev. Lett. 107, 136403 (2011). 38 Alternatively this result can be derived using a master- equation approach. See for instance K. Blum, Density Ma- trix Theory and Applications, 2nd ed. (Plenum, New York, 1996). 39 If the Majorana level were replaced by a regular spin- degenerate fermionic level, the current would be 2e- periodic due to an even-odd asymmetry of the current- carrying states and the corresponding cycles. 40 F. W. J. Hekking, L. I. Glazman, K. A. Matveev, and R. I. Shekhter, Phys. Rev. Lett. 70, 4138 (1993). 22 A. C. Potter and P. A. Lee, Phys. Rev. B 83, 094525 41 B. Kaviraj, O. Coupiac, H. Courtois, and F. Lefloch, Phys. (2011). 23 T. D. Stanescu, R. M. Lutchyn, and S. Das Sarma, Phys. Rev. B 84, 144522 (2011). Rev. Lett. 107, 077005 (2011).
1707.03868	1	1707	2017-07-12T19:02:44	Effective Interactions in a Graphene Layer Induced by the Proximity to a Ferromagnet	[ "cond-mat.mes-hall" ]	The proximity-induced couplings in graphene due to the vicinity of a ferromagnetic insulator are analyzed. We combine general symmetry principles and simple tight-binding descriptions to consider different orientations of the magnetization. We find that, in addition to a simple exchange field, a number of other terms arise. Some of these terms act as magnetic orbital couplings, and others are proximity-induced spin-orbit interactions. The couplings are of similar order of magnitude, and depend on the orientation of the magnetization. A variety of phases, and anomalous Hall effect regimes, are possible.	cond-mat.mes-hall	cond-mat	Effective Interactions in a Graphene Layer Induced by the Proximity to a Ferromagnet Võ Tin Phong1,∗ Niels R. Walet1,† and Francisco Guinea1,2,3‡ 1School of Physics and Astronomy, University of Manchester, Manchester, M13 9PY, UK 3Donostia International Physics Center, Paseo Manuel de Lardizabal 4, 20018 San Sebastian, Spain 2Imdea Nanoscience, Faraday 9, 28015 Madrid, Spain and (Dated: July 14, 2017) 7 1 0 2 l u J 2 1 ] l l a h - s e m . t a m - d n o c [ 1 v 8 6 8 3 0 . 7 0 7 1 : v i X r a The proximity-induced couplings in graphene due to the vicinity of a ferromagnetic insulator are analyzed. We combine general symmetry principles and simple tight-binding descriptions to consider different orientations of the magnetization. We find that, in addition to a simple exchange field, a number of other terms arise. Some of these terms act as magnetic orbital couplings, and others are proximity-induced spin-orbit interactions. The couplings are of similar order of magnitude, and depend on the orientation of the magnetization. A variety of phases, and anomalous Hall effect regimes, are possible. Introduction. Electrons in graphene placed in proximity to a ferromagnet experience an induced exchange field that modifies spin-transport properties. The exchange field results from virtual electrons hopping between the graphene layer and the ferromagnet. For instance, on an yttrium-iron garnet (YIG) substrate, this leads to magnetoresistance, spin-current- to-charge-current conversion [1], and spin precession from in- duced ferromagnetism [1–3]. Furthermore, in addition to in- duced ferromagnetism, we can also expect induced spin-orbit coupling (SOC) in graphene because atoms in the YIG sub- strate have non-negligible SOC [1]. The YIG ferromagnetism breaks time-reversal symmetry and thus gives rise to orbital couplings in the graphene layer which are not invariant under time inversion. We can expect similar effects on graphene in proximity to an EuO ferromagnet [4–7]. Heavy atoms near a graphene layer can, in general, induce non-trivial spin-orbit couplings [8], and these effects have been experimentally con- firmed [9–12]. With such experimental platforms now avail- able, we can attempt to engineer proximity-induced interac- tions in graphene for spintronics applications [13–15]. In the present work, we analyze the general interactions in- duced in a graphene layer by proximity to a ferromagnetic in- sulator with a significant SOC. We give a classification of the possible terms allowed by lattice symmetries, and present sim- ple models which allow us to determine the relative strengths of the various terms. The resulting electronic structure of graphene depends sensitively on the balance between these couplings, and can exhibit a number of interesting topological features. We discuss the main features of the different possi- ble phases. The model. For simplicity, we consider the interaction be- tween a graphene sheet and the top layer of a magnetic sub- strate. While this approach does not allow us to make quan- titative numerical predictions for the strength of the induced couplings in graphene, this is sufficient for a general analysis of the type of perturbations and their relative strengths. It is worth noting that the couplings between graphene and a sub- strate depend exponentially on the distance between the two systems, so that a quantitative theoretical study is, in any case, extremely challenging. We further assume that the magnetic atoms form a lattice commensurate with the graphene layer, and neglect the effects of disorder. This approximation im- plies that the allowed couplings satisfy translational symme- try. As an average translational symmetry over long scales can be defined for any substrate orientation, this approximation will capture the leading couplings. In the absence of the sym- metries considered here, other terms are possible, although it can be expected that they will be of smaller magnitudes. Fi- nally, we neglect the coupling between magnetic atoms. Such couplings will lead to dispersive bands in the graphene layer. Our goal is to define effective couplings in the graphene layer at the K and K(cid:48) points of the Brillouin zone. Thus, in princi- ple, we need only consider the states at these high symmetry points in the bands of the magnetic layer. The neglect of dis- persion in the magnetic bands does not change the symmetry of the required states, and the hopping between graphene and the magnetic layer does not depend on the hopping within the magnetic layer. Hence, this approximation is sufficient to cap- ture the general properties of the effective couplings in which we are interested. The geometry of the model is sketched in Fig. 1a. The spin-orbit coupling splits the electronic levels of the magnetic layer into multiplets defined by the total angular mo- mentum J = L + S. We expand the couplings induced in the graphene layer in powers of momenta around the K and K(cid:48) points. The leading terms are momentum-independent. The effective Hamiltonian at the K and K(cid:48) points is characterized by three variables which can take two values each: sublattice, valley, and spin [16]. Hence, the effective Hamiltonian can be written as an 8 × 8 matrix. This matrix can be expressed in terms of 2× 2 Pauli matrices in the sublattice, valley, and spin subspaces, σµ, τµ, and sµ, with µ ∈ [0, 1, 2, 3]. The unper- turbed Hamiltonian is then H0 = Dτ0 ⊗ σ0 ⊗ s0, where D is the Dirac energy, which we choose as the zero of energy. Here. we define matrices with µ = 0 as identity matrices, which henceforth will be left implicit. Results. We now consider the effect of a ferromagnetic layer. Since the atoms on this layer are magnetized, we only consider the lowest-energy orbital for a given total angular momentum for each magnetic atom; since it is magnetized, its angular momentum projection is fixed as well. This approx- imation holds when other orbitals lie at energies much fur- ther away from the Dirac energy of graphene, or, alternatively, when the hopping to graphene from these magnetic orbitals is Magnetization Out-of-plane, z T -symmetric τz ⊗ σz ⊗ sz σy ⊗ sx − τz ⊗ σx ⊗ sy In-plane, x In-plane, y σx σy ⊗ sz τz ⊗ sx τz ⊗ σx ⊗ sx τz ⊗ σz ⊗ sy σx σz σy ⊗ sy τz ⊗ sz τz ⊗ σx ⊗ sz τz ⊗ σz ⊗ sx 2 T -antisymmetric sz τz ⊗ σz sx σx ⊗ sx σz ⊗ sz τz τz ⊗ σx τz ⊗ σy ⊗ sz sy σy σx ⊗ sy σz ⊗ sy τz ⊗ σy ⊗ sz Figure 1. (a) Lattice of graphene-ferromagnet bilayer. The A sub- lattice of graphene is shown in red, while the B sublattice is shown in orange. The yellow dots represent the ferromagnetic atoms placed on top of graphene at the center of each hexagon. (b) The phases of the Bloch wavefunction of graphene for each sublattice at the K and K(cid:48) points. (c)-(d) Mirror reflection symmetry of in-plane magneti- zation. The green arrows indicate the projection direction of angular momentum, which behave as pseudovectors under reflection. Table I. Classification of possible proximity-induced interaction terms based on symmetries. The left column lists the direction of the magnetization, and each row lists all the terms which are consistent with the symmetries imposed by that magnetization direction. Terms in the middle column are even under time-inversion, and terms in the right column are odd. We have omitted terms which are related to each other by a 90o-rotation in the spin Hilbert space about the magnetization axis. suppressed. In this limit, we only allow the graphene electrons to hop to the lowest-energy magnetic orbitals on the nearest- neighbor magnetic atoms. This modifies the low-energy band structure of graphene. In principle, these proximity-induced interaction terms can, to lowest order, be written as linear combinations of τµ⊗σν ⊗sυ that respect the underlying sym- metries of the lattice. When the magnetization is out-of-plane, i.e., in the z-direction, the relevant symmetries of the Bloch wavefunctions at the K and K(cid:48) points are 2D inversion, I, and 120◦ rotation, R 2π : I = τx ⊗ σx ⊗ isz, (cid:32) 3 √ (cid:33) (cid:32) R 2π 3 = 1 2 + i 3 2 τz ⊗ σz ⊗ (cid:33) √ 3 2 1 2 + i (1) sz . All the terms which respect these two symmetries are listed in Table I (see the appendix for additional details). Through- out this work, we neglect terms which contain K-K(cid:48) mixing elements, since intervalley scattering is negligible unless the bilayer system forms a superlattice of at least three graphene unit cells. We further classify the remaining terms according to their transformation under time reversal, defined by T = τx ⊗ isyK, (2) where K is the complex conjugation operator. When the magnetization is aligned in the plane of the sub- strate, we break rotational and inversion symmetries. How- ever, when the magnetization points to a graphene carbon atom (e.g. along the x-direction) or points to the midpoint of the line joining two nearest-neighbor carbon atoms (e.g. along the y-direction), we still have mirror reflection symme- try in the plane perpendicular to the magnetization direction, as shown in Fig. 1c and d. In these two special orientations, the number of allowed terms is restricted by reflection sym- metry. For instance, if the magnetization is aligned along the x-direction, then the lattice is invariant under reflection over the y-z plane, and the corresponding Hamiltonian commutes with the mirror operator My = σx ⊗ isx. (3) Similarly, if the magnetization is in the y-direction, then the lattice is invariant under reflection over the x-z plane, and the corresponding Hamiltonian commutes with the mirror opera- tor Mx = τx ⊗ isy. (4) All the terms which are consistent with these reflection sym- metries are listed in Table I. Based only on symmetry considerations, we have arrived at a set of possible proximity-induced interaction terms for three independent magnetization directions. We now construct sim- ple microscopic tight-binding models that give us magnitude estimates of these terms. We consider a magnetic d-orbital with magnetization in a particular direction. The hopping be- tween the graphene orbitals and the magnetic orbitals can be approximated by the Slater-Koster method [17] in terms of 1ββ1ββKA,KB,1ββ1ββ',KA',KB(a)(b)3/2expiπβ(c)(d)zxytwo-center integrals Vpdσ and Vpdπ that are exponentially sup- pressed by increasing distance between the centers of the or- bitals. In the limit of weak coupling between the magnetic ions and the graphene carbon atoms, we can consider the ef- fect of hopping between the two layers as a small perturba- tion on the wavefunction of graphene at the Dirac points. The first-order correction to the graphene wavefunction is found by summing the hopping orbitals over a graphene hexagon (cid:88) τ,σ,s,j δψG(cid:105) = (cid:12)(cid:12)ψj τ,σ,s (cid:11) , tj s M (5) where τ, σ, s, denote the valley, sublattice, and spin indices at the lattice site j around the hexagon. The coefficient tj s is the strength of the hopping from the carbon atom at site j with spin s to the magnetic orbital, M is the energy of the magnetic (cid:11) is the graphene orbital with the appropri- orbital, and(cid:12)(cid:12)ψj ate phase as defined in Fig. 1b. The effective Hamiltonian is found by projecting the full Hamiltonian onto the basis of the perturbed wavefunctions. In the case of magnetization in the out-of-plane direction, τ,σ,s we find the following effective Hamiltonian Heff = λZsz + λQHτz ⊗ σz + λQSHτz ⊗ σz ⊗ sz + λSO (τz ⊗ σx ⊗ sy − σy ⊗ sx) . (6) Here, the sz term is a Zeeman exchange, the τz ⊗ σz term is the Haldane quantum orbital Hall effect [18], the τz ⊗ σz ⊗ sz term is the Kane-Mele quantum spin Hall effect [19], and the (τz ⊗ σx ⊗ sy − σy ⊗ sx) term is the Rashba spin-orbit cou- pling [20]. From this simple tight-binding model, we have found all four possible physically-distinct couplings that are allowed by symmetry. Separately, these terms have been stud- ied and classified topologically. The Haldane term leads to the quantum anomalous Hall effect in the absence of external magnetic fields. The Kane-Mele term, in a system with time- reversal symmetry, gives rise to symmetry-protected topologi- cal edge states. The exchange term and the Rashba term inde- pendently do not give rise to gaps near the Dirac points. How- ever, a combination of these two terms also leads to the quan- tum anomalous Hall effect [21, 22]. Our perturbative micro- scopic model suggests that all of these terms will be present with the same order of magnitude as all of them are due to the same hopping processes from graphene to magnetic orbitals and back. Therefore, they must be considered simultaneously. For instance, for a fully spin-polarized magnetic orbital with j = 5/2 and jz = 5/2, the Rashba coupling term is extremely small, and all the other terms are approximately equal in mag- nitude. This is because the Rashba term describes spin-flip processes that are prohibited by having a spin-polarized or- bital. To obtain the Rashba coupling, we should instead con- sider a magnetic orbital with non-maximal spin projection, such as a state with j = 5/2 and jz = 3/2. In this case, the tight-binding approach predicts that the Rashba term is of the same order of magnitude as the other three terms. These terms in combination give rise to a rich tapestry of topological features [23–27]. For a spin-polarized or- 3 Figure 2. (a) Phase diagram from proximity-induced interactions due to a spin-polarized magnetic orbital in the out-of-plane direction. (b)- (e) Representative band structures around the K point for the half- metallic, semi-metallic, quantum Hall, and quantum spin Hall phases respectively. The coupling constants are given in triplets of the form (λZ, λQH, λQSH) in units of meV. Blue bands belong to up-spins, and red bands belong to down-spins. (cid:90) Cτ,s = 1 2π bital in the z-direction, we can induce a Zeeman splitting, a Haldane orbital-hopping, and a Kane-Mele spin-orbit cou- pling. When the Zeeman term dominates, λZ > λQH and λZ > λQSH, we have a half-metal, meaning that if the Fermi energy is in the bulk band gap for one spin species, it lies in- side the conduction band or the valence band of the other spin species. Consequently, electrons at the Fermi surface are spin- polarized, and can be harnessed for spintronic applications. Different spin species can be selected by adjusting the Fermi level or by changing the sign of λZ. Increasing the strength of the Haldane and the Kane-Mele terms, we enter the insu- lating phase where there is a common bulk gap for both spin species. Assuming the Fermi energy is in the gap, this phase can be classified according to the Chern numbers Cτ,s for the valence bands of spins s at the τ valleys d2k(Ωτ,s)z, (7) where the Berry curvature is given by Ωτ,s = i(cid:104)∇kψτ,s × ∇kψτ,s(cid:105) , zero Chern number, C =(cid:80) ant called the spin Chern number 2Cs = (cid:80) (8) and ψτ,s(cid:105) are valence-band eigenstates of the low-energy Hamiltonian around the τ valley. When an insulator has a non- τ (Cτ,↑ + Cτ,↓), we have a Chern insulator that hosts the quantum Hall effect with Hall conduc- tance C = e2C/h. This occurs when the Haldane term domi- nates the Kane-Mele term, λQSH < λQH. Where the Chern number vanishes, we can define a different topological invari- τ (Cτ,↑ − Cτ,↓). When Cs = ±1, this phase supports the quantum spin Hall effect where there are spin-polarized edge states. However, unlike in a time-reversal-invariant topological insulator, these edge states are not symmetry-protected, and can backscatter [23]. This phase occurs when the Kane-Mele term domi- nates over the Haldane term, λQH < λQSH. The cross-over Spin-Hall InsulatorChernInsulatorHalf-MetalSemi-MetalQHλQSHλZλZλZλZλ)1 ,2.1 ,5.1(E (meV)K)1 ,5.1 ,5.1()1 ,2 ,5.1()3 ,2 ,5.1(E (meV)Ka)(b)(c)(d)(e)(boundary between the insulating phases and the half-metallic phases hosts a semi-metallic phase. Our tight-binding model suggests that the strengths of all three terms are similar when we have a spin-polarized orbital for the magnetic atoms. This means all four phases of the diagram in Fig. 2 are accessible via slight adjustments of the three coupling constants. As dis- cussed below, the value of the gap can be in the order of a few meV, so that extremely clean samples may be required to ob- serve quantized edge conductances. On the other hand, each type of electronic structure gives rise to different distributions of Berry curvature in the Brillouin zone, and will modify in a different way the anomalous Hall effect [28] in the metallic regime. When the magnetization is in-plane, many more coupling terms are allowed, since this configuration breaks both 2D in- version and rotation symmetries. However, at the two spe- cial alignments noted above, the Hamiltonian retains a mir- ror symmetry that restricts the number of possible terms. See the appendix for a detailed analysis of all terms. As an ex- ample, when the magnetization is in the x-direction, we have eleven possible distinct couplings. Seven of these coupling terms commute with sx, and thus conserve spin. These terms do not open gaps in the energy spectrum, but instead describe exchange interactions and pseudo-gauge potentials. As illus- tration, terms containing σx, e.g., σx, σx⊗sx and τz⊗σx, shift the Dirac points along the kx-direction. On the other hand, terms containing sx such as sx, τz ⊗ sx and τz ⊗ σx ⊗ sx lift the spin degeneracy in the x-direction, acting as exchange in- teractions. When the magnetic orbital is not of maximal pro- jection, spin-flip processes give rise to spin-projection non- conserving terms such as σz ⊗ sy, σy ⊗ sz, τz ⊗ σz ⊗ sy, and τz ⊗ σy ⊗ sz. Some of these terms open gaps in the energy spectrum. A similar structure can be found when the mag- netization is aligned along the y-direction. As in the case of out-of-plane magnetization, we expect all of these coupling terms to be of similar order of magnitude. However, in addi- tion to gap-opening and exchange terms, in the in-plane case, we also find pseudo-gauge potentials which can further mod- ify spin-transport properties in graphene [29, 30]. Discussion and conclusions. We have presented an analy- sis of the general properties of the spin-dependent couplings induced in graphene by proximity to a ferromagnetic insu- lator. A more quantitative analysis requires the knowledge of the band structure of the insulator, the alignment of the graphene layer with respect to the insulator, and, more cru- cially, the hopping between graphene and substrate orbitals. The last parameter is still poorly understood. The effective couplings within the graphene layer depend quadratically on this parameter. Theoretical estimates, mostly based on DFT calculations, for related systems give values in the 10−1− 101 meV range [10, 31–36]. Experimental values for these cou- plings also show a significant variety, although they tend to be smaller than the the theoretical predictions, (cid:46) 0.1 meV [2, 3, 6, 10–12, 37–40]. The reason for this discrepancy is likely to be the uncertainty on the graphene-substrate distance, and the degree of commensurability between graphene and the 4 substrate. The existance of levels in the magnetic substrate close to the Dirac energy of graphene may enhance signifi- cantly these couplings. We have analyzed here a few high symmetry orientations of the magnetization of the substrate, which we expect to be representative of the rich variety of regimes possible. Irrespective of the imprecision in the value of the graphene- substrate couplings, our results lead to four main conclusions: i) The coupling of graphene to a magnetic substrate gives rise to a number of effective interactions beyond a simple ex- change term. Some of them break time-reversal symmetry, and depend on the substrate's magnetization, and some others do not break time-reversal symmetry, and describe spin-orbit coupling terms. ii) A large fraction of the interactions mentioned above are of the same order of magnitude, as all of them are propor- tional to the square of the hopping between graphene and the substrate, divided by the energy difference between the states involved. iii) The precise value of the couplings depends, among other factors, on the orientation of the magnetization of the substrate, leading to the possibility of modifying these inter- actions by changing the direction of the magnetization. iv) The interactions can open gaps at the Dirac energy of graphene. A rich phase diagram, which includes topological insulator and anomalous Hall insulator phases, emerges. The above features lead to a rich number of regimes in the anomalous Hall effect typical of a ferromagnetic system. The three main contributions, intrinsic, skew scattering, and side jump [28], are expected to depend on the various electronic structures described here, which also depend on the orienta- tion of the magnetization. These topics will be discussed else- where. We are thankful for helpful discussions with H. Ochoa and J. Fabian. This work was supported by funding from the European Union through the ERC Advanced Grant NOV- GRAPHENE through grant agreement Nr. 290846, and from the European Commission under the Graphene Flagship, con- tract CNECTICT-604391. VTP acknowledges financial sup- port from the Marshall Aid Commemoration Commission. ∗ [email protected] † [email protected] ‡ [email protected] [1] J. B. S. Mendes, O. Alves Santos, L. M. Meireles, R. G. Lac- erda, L. H. Vilela-Leao, F. L. A. Machado, R. L. Rodr´ıguez- Su´arez, A. Azevedo, and S. M. Rezende, Phys. Rev. Lett. 115, 226601 (2015). [2] J. C. Leutenantsmeyer, A. A. Kaverzin, M. Wojtaszek, and B. J. van Wees, 2D Materials 4, 014001 (2017). [3] S. Singh, J. Katoch, T. Zhu, K.-Y. Meng, T. Liu, J. T. Brangham, F. Yang, M. E. Flatt´e, and R. K. Kawakami, Phys. Rev. Lett. 118, 187201 (2017). [4] A. G. Swartz, P. M. Odenthal, Y. Hao, R. S. Ruoff, and R. K. Kawakami, ACS Nano 6, 10063 (2012). [5] Z. Wang, C. Tang, R. Sachs, Y. Barlas, and J. Shi, Phys. Rev. [26] W. Sa-Ke, T. Hong-Yu, Y. Yong-Hong, and W. Jun, Chinese Lett. 114, 016603 (2015). [6] P. Wei, S. Lee, F. Lemaitre, L. Pinel, D. Cutaia, W. Cha, F. Kat- mis, Y. Zhu, D. Heiman, J. Hone, J. S. Moodera, and C.-T. Chen, Nature Mat. 15, 711 (2016). [7] S. Su, Y. Barlas, J. Li, J. Shi, and R. K. Lake, Phys. Rev. B 95, [8] C. Weeks, J. Hu, J. Alicea, M. Franz, and R. Wu, Phys. Rev. X Lett. 103, 146801 (2009). 075418 (2017). 1, 021001 (2011). 5 Physics B 23, 017203 (2014). [27] S. Wang and J. Wang, Physica B: Condensed Matter 458, 22 (2015). [28] N. Nagaosa, J. Sinova, S. Onoda, A. H. MacDonald, and N. P. Ong, Rev. Mod. Phys. 82, 1539 (2010). [29] D. Huertas-Hernando, F. Guinea, and A. Brataas, Phys. Rev. [30] M. Vozmediano, M. Katsnelson, and F. Guinea, Physics Re- ports 496, 109 (2010). [31] M. Gmitra and J. Fabian, Phys. Rev. B 92, 155403 (2015). [32] K. Zollner, M. Gmitra, T. Frank, and J. Fabian, Phys. Rev. B 94, 155441 (2016). [33] A. Hallal, F. Ibrahim, H. X. Yang, S. Roche, and M. Chshiev, 2D Materials 4, 025074 (2017). [34] D. Kochan, S. Irmer, and J. Fabian, Phys. Rev. B 95, 165415 (2017). [35] M. Gmitra and J. Fabian, "Proximity effects in bilayer graphene on monolayer wse2: Field-effect spin-valley locking, spin- orbit valve, and spin transistor," (2017), arXiv:1706.06149, arXiv:1706.06149. [36] H. Yang, G. Chen, A. A. Cotta, A. T. N'Diaye, S. A. Niko- laev, E. A. Soares, W. A. Macedo, A. K. Schmid, A. Fert, and M. Chshiev, "Significant dzyaloshinskii-moriya interaction at graphene-ferromagnet interfaces due to rashba-effect," (2017), arXiv:1704.09023, arXiv:1704.09023. [37] A. Avsar, J. Y. Tan, J. Balakrishnan, G. Koon, Jayeeta Lahiri, A. Carvalho, A. Rodin, T. Taychatanapat, E. O'Farrell, G. Eda, A. H. Castro Neto, and B. Ozyilmaz, Nature Comm. 5, 4875 (2014). [38] L. Zhang, B.-C. Lin, Y.-F. Wu, J. Xu, D. Yu, and Z.-M. Liao, ACS Nano 11, 6277 (2017). [39] A. Avsar, D. Unuchek, J. Liu, O. Lopez Sanchez, K. Watanabe, T. Taniguchi, B. Ozyilmaz, and A. Kis, "Optospintronics in graphene via proximity coupling," (2017), arXiv:1705.10267, arXiv:1705.10267. [40] V. Asshoff, J. L. Sambricio, A. P. Rooney, S. S. lizovskiy, A. Mishchenko, A. M. Rakowski, E. W. Hill, A. K. Geim, S. J. Haigh, V. I. Fal'ko, I. Vera-Marun, and I. V. Grigorieva, 2D Materials 4, 031004 (2017). [9] F. Calleja, H. Ochoa, M. Garnica, S. Barja, J. J. Navarro, A. Black, M. M. Otrokov, E. V. Chulkov, A. Arnau, A. L. V. De Parga, et al., Nature Physics 11, 43 (2015). [10] Z. Wang, D.-K. Ki, H. Chen, H. Berger, A. H. MacDonald, and A. F. Morpurgo, Nature Comm. 6, 8339 (2015). [11] Z. Wang, D.-K. Ki, J. Y. Khoo, D. Mauro, H. Berger, L. S. Levitov, and A. F. Morpurgo, Phys. Rev. X 6, 041020 (2016). [12] I. I. Klimovskikh, M. M. Otrokov, V. Y. Voroshnin, D. Sostina, L. Petaccia, G. Di Santo, S. Thakur, E. V. Chulkov, and A. M. Shikin, ACS Nano 11, 368 (2017). [13] I. Zuti´c, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76, 323 [14] W. Han, R. K. Kawakami, M. Gmitra, and J. Fabian, Nature [15] P. Lazi´c, G. M. Sipahi, R. K. Kawakami, and I. Zuti´c, Phys. (2004). Nano. 9, 794 (2014). Rev. B 90, 085429 (2014). [16] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). [17] J. C. Slater and G. F. Koster, Phys. Rev. 94, 1498 (1954). [18] F. D. M. Haldane, Phys. Rev. Lett. 61, 2015 (1988). [19] C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005). [20] H. Min, J. E. Hill, N. A. Sinitsyn, B. R. Sahu, L. Kleinman, and A. H. MacDonald, Phys. Rev. B 74, 165310 (2006). [21] W.-K. Tse, Z. Qiao, Y. Yao, A. H. MacDonald, and Q. Niu, Phys. Rev. B 83, 155447 (2011). [22] Z. Qiao, W. Ren, H. Chen, L. Bellaiche, Z. Zhang, A. H. Mac- Donald, and Q. Niu, Phys. Rev. Lett. 112, 116404 (2014). [23] Y. Yang, Z. Xu, L. Sheng, B. Wang, D. Y. Xing, and D. N. Sheng, Phys. Rev. Lett. 107, 066602 (2011). [24] M. Ezawa, Phys. Rev. B 87, 155415 (2013). [25] A. Cresti, D. Van Tuan, D. Soriano, A. W. Cummings, and S. Roche, Phys. Rev. Lett. 113, 246603 (2014). Tight-Binding Model 6 (cid:16) ψG(cid:105) = In this section, we give details of the tight-binding model used in the main text. As mentioned there, we consider a bilayer system with a graphene bottom layer and a triangular-lattice ferromagnetic top layer, as shown in Fig. 1 in the main text. Each carbon atom in the graphene layer has a free pz orbital onto which electrons can hop. We use a simple nearest-neighbor hopping model inside the graphene layer, where each carbon atom has three nearest neighbors at a distance a, to which its electron can hop. We shall concentrate on the low-energy properties, which are dominated by the inequivalent K and K(cid:48) points in the Brillouin zone. Here the electronic wavefunction is a combination of the localized orbital wavefunctions on the A and B sublattices with different spins. We can write the graphene wavefunction ψG(cid:105) in this basis as (A1) Here K and K(cid:48) label the valley degree of freedom, A and B label the two sublattices, and s and s(cid:48) are the spin projections. The unperturbed Hamiltonian at these Dirac points is ψK,A,s ψK,A,s(cid:48) ψK,B,s ψK,B,s(cid:48) ψK(cid:48),A,s ψK(cid:48),A,s(cid:48) ψK(cid:48),B,s ψK(cid:48),B,s(cid:48) . H0 = Dτ0 ⊗ σ0 ⊗ s0, (A2) where D is the Dirac energy, which we henceforth choose as zero. The τµ, σµ, and sµ Pauli matrices act on the valley, sublattice, and spin spaces respectively, and µ ∈ [0, 1, 2, 3], where Pauli matrices with a zero index are the 2 × 2 identity matrix. In considering the ferromagnetic layer, we only include one energetically-favorable magnetized orbital per magnetic atom. We model this layer as a flat-band structure, ignoring the interactions inside the substrate that would allow electrons to hop in the magnetic material. Thus, electrons can only move by hopping to graphene atoms. We assume that the graphene electrons hop to nearest-neighbor magnetic orbitals only. In the limit of weak hopping, we can use lowest-order non-vanishing perturbation theory to find the effective Hamiltonian describing the change to the Dirac spectrum. To illustrate the procedure, we consider the case where the magnetization is in the z-direction, perpendicular to the substrate. The simplest case is to assume a fully spin-polarized atomic orbital in the magnet. We consider a d orbital with orbital angular momentum l = 2 and projection lz = 2, and total angular momentum j = 5/2 and projection jz = 5/2. The wavefunction of the magnetic orbital can then be written as (cid:17) (cid:12)(cid:12)(cid:12)ψjz=5/2 M (cid:69) =(cid:12)(cid:12)ψx2−y2,↑(cid:11) + iψxy,↑(cid:105) , where ψx2−y2,↑ and ψxy,↑ are real atomic d orbitals with spatial symmetry indicated by the coordinate subscripts and spin indicated by ↑. In the absence of spin-orbit coupling, the orbital defined in Eq. (A3) is degenerate with an orbital defined similarly but with the spins flipped. In what follows, we assume that there is strong spin-orbit coupling inside the magnetic ion that lifts this degeneracy. With the lattice cell defined in Fig. 1, each unit cell consists of three atoms, two graphene carbon atoms and one magnetic ion. This corresponds to three orbitals per unit cell, two pz orbitals from the carbon atoms, and one d orbital from the magnetic ion. Without loss of generality, we set the energy of the pz orbitals to zero at the Dirac points, and denote the energy of the d orbital as M . We only consider nearest-neighbor couplings. This means that an electron in a d orbital can hop to six possible pz graphene orbitals, while an electron from a graphene orbital can hop to any of the three neighboring graphene orbitals or to any of the three neighboring magnetic orbitals. Therefore, there are nine, (3 × 6)/2 = 9, possible hopping coefficients per unit cell. The hopping between graphene orbitals and magnetic orbitals can be approximated by the Slater-Koster (SK) method [17] in terms of two-center integrals that depend on the distance between centers of the orbitals. The coupling terms also depend on the relative orientation of the orbitals. Let us place the magnetic atom at rM = (0, 0, dM ), where dM is the vertical distance between the graphene plane and the substrate, and the six surrounding graphene atoms are at rj G = a(cos[jπ/3], sin[jπ/3], 0), where j = 0, 1, ..., 5, and a is the distance between adjacent graphene atoms. The relevant SK parameters in this case are where Vpdσ and Vpdπ are the two-center integrals, and lj, mj and nj are direction cosines of rM − rj G defined by (A3) (A4) (A5) √ Ej z,x2−y2 = Ej z,xy = 3 nj(l2 √ 2 3ljmjnjVpdσ − 2ljmjnjVpdπ, j − m2 j )Vpdπ, j − m2 (cid:112) (cid:112) j )Vpdσ − nj(l2 (cid:19) (cid:19) (cid:18) πj (cid:18) πj dM(cid:112)a2 + d2 1 − n2 cos 1 − n2 sin 3 3 . , , M lj = mj = nj = n = 7 Figure A1. Electron wavefunction of the graphene lattice at the K and K(cid:48) Dirac points with different phases around the hexagon. ψK,A(cid:105), ψK,B(cid:105), ψK(cid:48),A(cid:105), and ψK(cid:48),B(cid:105) are the orbital wavefunctions of the A and B sublattices. Since the magnetic orbital is spin-polarized, only spin-up graphene electrons are allowed to hop into this d orbital. Putting in the spin index, the hoppings terms are just (cid:40) 0 Ej z,x2−y2 + iEj tj M,s = z,xy =(cid:0)n − n3(cid:1)(cid:16)√ (cid:17) 3 2 Vpdσ − Vpdπ e 2πi 3 j = tM e 2πi 3 j s =↑ s =↓ . (A6) In essence, we can now augment the graphene tight-binding model to include the d orbitals of the magnetic atoms with on-site energy M and nearest-neighbor hoppings defined by Eq. (A6). M (cid:28) tGM , where tG is the hopping magnitude between carbon orbitals, we can consider the effect of hopping between the two layers as a small perturbation on the wavefunction of graphene at the Dirac points. To first order, the correction to the graphene wavefunction is In the limit of weak hopping between the magnetic atoms and the graphene atoms, t2 (cid:88) δψG(cid:105) = τ∈{K,K(cid:48)},s∈{↑,↓},j tj M,s M,s τ,s(cid:105), φj (A7) σ,τ is the atomic orbital of graphene at site j in the τ ∈ {K, K(cid:48)} point of the Brillouin zone with spin s ∈ {↑,↓} with the where φj appropriate phase shown in Fig. A1. Summing over the atomic orbitals with the correct phases, we find that the phases interfere destructively among the B graphene orbitals and constructively among the A graphene orbitals in the K point. Conversely, the graphene orbitals interfere destructively among the A graphene orbitals and constructively among the B graphene orbitals in the K(cid:48) point. That is, the perturbed wavefunction is 3tM M Neglecting intervalley terms, the effective Hamiltonian is δψG(cid:105) = (φK,A,↑(cid:105) + φK(cid:48),B,↑(cid:105)) . (cid:16) Heff = 9t2 M M † † K,A,↑cK,A,↑ + c K(cid:48),B,↑cK(cid:48),B,↑ c (cid:17) , (A8) (A9) where c† with spin s. Writing Eq. (A9) in the basis of Pauli matrices, we have τ,σ,s and cτ,σ,s are creation and annihilation operators that create or destroy an orbital at the τ valley, σ sublattice, and Heff = 9t2 M 4M (1 + τz ⊗ σz) (1 + sz) = 9t2 M 4M (1 + sz + τz ⊗ σz + τz ⊗ σz ⊗ sz) . (A10) In considering just a d orbital with maximal perpendicular angular momentum, we find that several coupling terms are possible. These terms modify the low-energy spectrum near the Dirac points of graphene. We can consider the effect of coupling to orbitals which are not fully spin-polarized. The relative strengths of the four operators on the right-hand side of Eq. (A10) will no longer be equal. Furthermore, new terms may also be possible. Orbitals KAKφ,AKiπφe,3/2AKiπφe,3/2BKφ,BKiπφe,3/2BKiπφe,3/2K'AKφ,'AKiπφe,'3/2AKiπφe,'3/2BKφ,'BKiπφe,'3/2BKiπφe,'3/2yxzwhich do not have maximal angular momentum projection combine wavefunctions with up and down spins, and can give these new effects. For instance, we consider the same d orbital as before, with j = 5/2 but now take jz = 3/2 with wavefunction (cid:12)(cid:12)(cid:12)ψjz=3/2 M (cid:69) = 1√ 5 (cid:0)(cid:12)(cid:12)ψx2−y2,↓(cid:11) + iψxy,↓(cid:105)(cid:1) + (ψxz,↑(cid:105) + iψyz,↑(cid:105)) . 2√ 5 (A11) 8 As we can see from Eq. (A11), the spin-orbit coupling combines wavefunctions with opposite spins and different orbital angular momenta. In addition to the SK parameters in Eq. (A4), there are two more relevant parameters √ √ Ej Ej z,xz = z,yz = 3ljn2 3mjn2 j )Vpdπ, j )Vpdπ. j Vpdσ + lj(1 − 2n2 j Vpdσ + mj(1 − 2n2 (cid:40) tM,↑e πi 3 j tM,↓e 2πi 3 j s =↑ s =↓ , tj M,s = (cid:32)√ (cid:33) , n − n3√ (cid:112) 5 2√ 5 Vpdσ − Vpdπ 3 2 1 − n2 (cid:16) n2(cid:16)√ 3Vpdσ − 2Vpdπ (cid:17) (cid:17) . + Vpdπ The hopping coefficients are then where the magnitudes are given by tM,↓ = tM,↑ = The perturbed wavefunction is δψG(cid:105) = The effective Hamiltonian is (cid:16) (cid:16) (A12) (A13) (A14) (A15) (A16) (A17) (A18) (cid:17) 3tM,↓ M (φK,A,↓(cid:105) + φK(cid:48),B,↓(cid:105)) + (φK(cid:48),A,↑(cid:105) − φK,B,↑(cid:105)) . 3tM,↑ M † K,A,↓cK,A,↓ + c c † K(cid:48),B,↓cK(cid:48),B,↓ † † K(cid:48),A,↑cK(cid:48),A,↑ + c K,B,↑cK,B,↑ c (cid:17) (cid:16) 9t2 M,↑ M † † K,B,↑cK,A,↓ − c K(cid:48),A,↑cK(cid:48),B,↓ − c + (cid:17) † K,A,↓cK,B,↑ . † K(cid:48),B,↓cK(cid:48),A,↑ + c c Heff = + 9t2 M,↓ M 9tM,↓tM,↑ M (cid:16) In terms of Pauli matrices, the effective Hamiltonian is Heff = λ1 (1 − τz ⊗ σz ⊗ sz) + λ2 (sz − τz ⊗ σz) − λ3 (σy ⊗ sy + τz ⊗ σx ⊗ sx) , where 9 λ1 = (cid:17) (cid:16) (cid:17) t2 M,↓ + t2 M,↑ 4M 9 , λ2 = M,↑ − t2 t2 4M M,↓ , and λ3 = 9tM,↓tM,↑ . 2M The last term in Eq. (A17), upon a 90o-rotation in spin space about the z-axis, is just the Rashba coupling, τz⊗σx⊗sy−σy⊗sx. In this more familiar notation, the Hamiltonian is Heff = λ1 (1 − τz ⊗ σz ⊗ sz) + λ2 (sz − τz ⊗ σz) + λ3 (τz ⊗ σx ⊗ sy − σy ⊗ sx) . (A19) The terms in Eq. (A19) are all the terms that are consistent with the spatial symmetries of the lattice. With this simple microscopic model, we can estimate the relative contributions of the different terms. The procedure above can likewise be used to find the coupling terms when the magnetization is in-plane. In general, we can write the effective Hamiltonian when the magnetization is in the x-direction as Heff = ξ1σx + ξ2sx + ξ3σx ⊗ sx + ξ4τz + ξ5τz ⊗ σx + ξ6τz ⊗ sx + ξ7τz ⊗ σx ⊗ sx + ξ8σz ⊗ sy + ξ9σy ⊗ sz + ξ10τz ⊗ σz ⊗ sy + ξ11τz ⊗ σy ⊗ sz, (A20) where ξi are coupling constants, for i = 1, ..., 11. The first seven terms commute with sx, and therefore conserve spin in the x-direction. When we have spin polarization, only these seven terms are allowed. The remaining four terms mix the spins, and therefore, can only exist if we consider orbitals with non-maximal spin projection. We analyze these terms one-by-one. For the seven spin-conserving terms, we have exchange effects and pseudo-gauge potentials, which may be valley-dependent and spin-dependent. These terms shift the position of the Dirac points at the K and K(cid:48) points and break the degeneracy of the right and left spins. Of the four terms that mix the spins, two terms behave as pseudo-gauge fields, and two terms open gaps in the energy spectrum. This is summarized in Table AI. We now study the case where the magnetization is in the y-direction. In general, the Hamiltonian is Heff = χ1σx + χ2σy + χ3σz + χ4sy + χ5σx ⊗ sy + χ6σy ⊗ sy + χ7σz ⊗ sy + χ8τz ⊗ sz + χ9τz ⊗ σx ⊗ sz + χ10τz ⊗ σy ⊗ sz + χ11τz ⊗ σz ⊗ sx, 9 (A21) where χi are coupling constants, for i = 1, ..., 11. The first seven terms exist when we have spin polarization. The remaining four terms involve spin-flip processes, and can only be non-zero if we include non-maximally-magnetized orbitals. Furthermore, in our tight-binding model, χ3 = χ7 = 0 because these terms correspond to staggered potentials that require sublattice-symmetry breaking. Let us now analyze these terms one-by-one as before. Only terms which contain σz open gaps in the energy spectrum. The other terms correspond to either exchange terms or pseudo-gauge potentials, which can dependent on spin and valley. If we ignore the staggered potentials, then the only remaining term that can open a gap is τz ⊗ σz ⊗ sx. The complete list of terms is summarized in Table AII. Table AI. Effect of coupling terms on the spectrum of graphene when the magnetization is in the x-direction. We denote the unperturbed spectrum (k) = vF k, and the valley index τ = ±1. When the Hamiltonian conserves sx, we label the spectrum by the right and left spins → and ←. Otherwise, we label the states as (1) and (2). Of the eleven terms, only two terms open gaps in the energy spectrum. Term ξ1σx ξ2sx ξ3σx ⊗ sx ξ4τz ξ5τz ⊗ σx ξ6τz ⊗ sx ξ7τz ⊗ σx ⊗ sx ξ8σz ⊗ sy ξ9σy ⊗ sz ξ10τz ⊗ σz ⊗ sy ξ11τz ⊗ σy ⊗ sz (cid:113) Spectrum E±,(cid:28) = ±vF E±,→ = ξ2 ± (k) E±,← = −ξ2 ± (k) E±,→ = ±vF E±,← = ±vF (cid:113) (cid:113) E±,(cid:28) = τ ξ4 ± (k) (cid:113) y + (τ kx + ξ1/vF )2 k2 y + (τ kx − ξ3/vF )2 k2 y + (τ kx + ξ3/vF )2 k2 y + (kx + ξ5/vF )2 k2 y + (kx − ξ7/vF )2 k2 y + (kx + ξ7/vF )2 k2 E±,(cid:28) = ±vF E±,→ = τ ξ6 ± (k) E±,← = −τ ξ6 ± (k) E±,→ = ±vF E±,← = ±vF (cid:113) (cid:113) E±,(1,2) = ±(cid:112)ξ2 (cid:112)(ky + ξ9/vF )2 + k2 (cid:112)(ky − ξ9/vF )2 + k2 E±,(1,2) = ±(cid:112)ξ2 (cid:113) (cid:113) E±,(1) = ±vF E±,(2) = ±vF 10 + (k)2 E±,(1) = ±vF E±,(2) = ±vF x + (τ ky + ξ11/vF )2 k2 x + (τ ky − ξ11/vF )2 k2 8 + (k)2 x x Description Pseudo-gauge potential that shifts the Dirac point along the kx-axis in opposite directions for the two valleys. This does not open a gap. Exchange potential that breaks the degeneracy between the right and left spins. This does not open a gap. Spin-dependent pseudo-gauge potential that shifts the Dirac point along the kx-axis in opposite directions for the two valleys. This does not open a gap. Valley exchange term that breaks the degeneracy between the K and K(cid:48) points. This does not open a gap. Pseudo-gauge potential that shifts the Dirac point along the kx-axis in same direction for the two valleys. This does not open a gap. Valley-dependent exchange potential that breaks the degeneracy be- tween the right and left spins. This does not open a gap. Spin-dependent pseudo-gauge potential that shifts the Dirac point along the kx-axis in the same direction for the two valleys. This does not open a gap. Sublattice-symmetry-breaking term that opens a gap of magnitude 2ξ8. Pseudo-gauge potential that shifts the Dirac point along the ky-axis in the same direction for the two valleys, and also mixes the spins. This does not open a gap. Valley-dependent sublattice-symmetry-breaking term that opens a gap of magnitude 2ξ10. Pseudo-gauge potential that shifts the Dirac point along the ky-axis in opposite directions for the two valleys. This does not open a gap. Table AII. Effect of the induced couplings on the spectrum of graphene when the magnetization is in the y-direction. We denote the unperturbed spectrum (k) = vF k, and the valley index τ = ±1. When the Hamiltonian conserves sy, we label the spectrum by the spin-eigenstates denoted as (cid:37) and (cid:46). Otherwise, we label the spectrum by (1) and (2). Of the eleven terms, three terms open gaps in the energy spectrum. 10 Term χ1σx χ2σy χ3σz χ4sy χ5σx ⊗ sy χ6σy ⊗ sy χ7σz ⊗ sy χ8τz ⊗ sz χ9τz ⊗ σx ⊗ sz χ10τz ⊗ σy ⊗ sz χ11τz ⊗ σz ⊗ sx Spectrum E±,(cid:37)(cid:46) = ±vF (cid:113) (cid:113) y + (τ kx + χ1/vF )2 k2 x + (ky + χ2/vF )2 k2 3 E±,(cid:37)(cid:46) = ±vF E±,(cid:37) = χ4 ± (k) E±,(cid:46) = −χ4 ± (k) E±,(cid:37)(cid:46) = ±(cid:112)(k)2 + χ2 E±,(cid:37) = ±(cid:113) E±,(cid:46) = ±(cid:113) E±,(cid:37) = ±(cid:113) E±,(cid:46) = ±(cid:113) E±,(cid:37)(cid:46) = ±(cid:112)(k)2 + χ2 7 y + (τ kx − χ5/vF )2 k2 y + (τ kx + χ5/vF )2 k2 x + (ky − χ6/vF )2 k2 c + (ky + χ6/vF )2 k2 E±,(1) = τ χ8 ± (k) E±,(2) = −τ χ8 ± (k) E±,(1) = ±vF E±,(2) = ±vF E±,(1) = ±vF E±,(2) = ±vF (cid:112)(kx + χ9/vF )2 + k2 (cid:112)(kx − χ9/vF )2 + k2 (cid:112)(τ ky + χ10/vF )2 + k2 (cid:112)(τ ky − χ10/vF )2 + k2 E±,(cid:37)(cid:46) = ±(cid:112)(k)2 + χ2 y y x x 11 Description Pseudo-gauge potential that shifts the Dirac point along the kx-axis in opposite directions for the two valleys. This does not open a gap. Pseudo-gauge potential that shifts the Dirac point along the ky-axis in the same direction for the two valleys. This does not open a gap. Mass term that breaks sublattice symmetry, and opens a gap of magni- tude 2χ3. Exchange potential that breaks the degeneracy between the in and out spins. This does not open a gap. Spin-dependent pseudo-gauge potential that shifts the Dirac point along the kx-axis in opposite directions for the two valleys. This does not open a gap. Spin-dependent pseudo-gauge potential that shifts the Dirac point along the ky-axis in the same direction for the two valleys. This does not open a gap. Spin-dependent term that breaks sublattice symmetry, and opens a gap of magnitude 2χ7. Valley-dependent exchange-like potential that breaks the degeneracy between the in and out spins. This does not open a gap. Pseudo-gauge potential that shifts the Dirac point along the kx-axis in the same direction for the two valleys. This does not open a gap. Pseudo-gauge potential that shifts the Dirac point along the ky-axis in opposite directions for the two valleys. This does not open a gap. This term opens a gap of magnitude 2χ11.
1711.08647	2	1711	2018-04-11T20:11:48	Structural and electronic transformation in low-angle twisted bilayer graphene	[ "cond-mat.mes-hall", "cond-mat.dis-nn", "cond-mat.mtrl-sci" ]	Experiments on bilayer graphene unveiled a fascinating realization of stacking disorder where triangular domains with well-defined Bernal stacking are delimited by a hexagonal network of strain solitons. Here we show by means of numerical simulations that this is a consequence of a structural transformation of the moir\'{e} pattern inherent of twisted bilayer graphene taking place at twist angles $\theta$ below a crossover angle $\theta^{\star}=1.2^{\circ}$. The transformation is governed by the interplay between the interlayer van der Waals interaction and the in-plane strain field, and is revealed by a change in the functional form of the twist energy density. This transformation unveils an electronic regime characteristic of vanishing twist angles in which the charge density converges, though not uniformly, to that of ideal bilayer graphene with Bernal stacking. On the other hand, the stacking domain boundaries form a distinct charge density pattern that provides the STM signature of the hexagonal solitonic network.	cond-mat.mes-hall	cond-mat	Structural and electronic transformation in low-angle twisted bilayer graphene Fernando Gargiulo and Oleg V. Yazyev∗ Institute of Physics, Ecole Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland (Dated: April 13, 2018) Abstract Experiments on bilayer graphene unveiled a fascinating realization of stacking disorder where triangular domains with well-defined Bernal stacking are delimited by a hexagonal network of strain solitons. Here we show by means of numerical simulations that this is a consequence of a structural transformation of the moiré pattern inherent of twisted bilayer graphene taking place at twist angles θ below a crossover angle θ(cid:63) = 1.2◦. The transformation is governed by the interplay between the interlayer van der Waals interaction and the in-plane strain field, and is revealed by a change in the functional form of the twist energy density. This transformation unveils an electronic regime characteristic of vanishing twist angles in which the charge density converges, though not uniformly, to that of ideal bilayer graphene with Bernal stacking. On the other hand, the stacking domain boundaries form a distinct charge density pattern that provides the STM signature of the hexagonal solitonic network. 8 1 0 2 r p A 1 1 ] l l a h - s e m . t a m - d n o c [ 2 v 7 4 6 8 0 . 1 1 7 1 : v i X r a ∗ [email protected] 1 Bilayer graphene (BLG) shares many of the properties of monolayer graphene while also showing a number of pronounced differences. For instance, its equilibrium structural con- figuration reveals the massive nature of its charge carriers [1], the possibility of inducing a tunable band gap by applying a transverse electric field [2–4] and quantum Hall valley ferromagnetism [5]. These properties are a result of the coupling between the two layers. In order to describe the atomic structure of bilayer graphene the relative position of the two layers has to be defined. In many situations it is sufficient to specify a unique interlayer displacement vector that defines the stacking configuration. As a general property of graphitic structures, the low-energy configuration is represented by the Bernal stacking [6, 7]. However, the stacking configuration is not immune to disorder which can manifest, for example, in boundaries that connect two domains with energetically degenerate yet topologically inequivalent stacking configurations, AB and BA [8–11]. Such stacking domain boundaries are realized by strain solitons, which are segments with a characteristic width where the strain that arises from interfacing two inequivalent stacking domains is confined. Recent studies have shown that strain solitons can be displaced by the action of a scanning tunneling microscope tip, but do not vanish due to their topological nature [8, 12]. From the theoretical point of view, the two-dimensional extension of the Frenkel-Kontorova model predicts the emergence of strain solitons with a typical width of a few nanometers [13] while their density is defined by the twist angle. In other situations the stacking configuration cannot be uniquely defined on the whole surface of the sample since the two layers cannot be superimposed by a rigid in-plane shift. This is the case of twisted bilayer graphene where one layer is rotated relative to another, a system that has been widely reported in samples grown epitaxially or by chemical va- por deposition [14–20]. The two rotated layers form a typical moiré superlattice that has been imaged by means of transmission electron microscopy (TEM) and scanning tunneling microscopy (STM). Importantly, strain solitons in bilayer graphene can form a hexagonal network that delimits triangular domains with inequivalent AB and BA Bernal stacking [8, 21, 22]. Remarkably, as pointed out in Ref. [21], such structures are topologically equiv- alent to twisted BLG. In this work, we have investigated by means of numerical simulations twisted BLG and show that for twist angles θ above a crossover angle θ(cid:63) = 1.2◦ the equilibrium structures do not differ substantially from a rigid twist of the two layers, while for θ below θ(cid:63) a crossover 2 into a different regime takes place. In this regime, the equilibrium configuration consists of a triangular lattice of alternating AB and BA stacking domains separated by shear strain solitons that form a hexagonal network. The electronic structure is profoundly affected by the emergence of this structural phase and exhibits characteristic features determined by the local stacking order. In contrast to the picture valid for low-angle rigidly twisted BLG that predicts low-energy states localized in AA regions [23–25], we find that the charge density in AB and BA domains resembles that of ideal Bernal-stacked bilayer graphene. On the other hand, the stacking domain boundaries form a distinct charge density pattern that provides the scanning tunneling microscopy signature of the hexagonal solitonic network. I. RESULTS A. Structural transformation in low-angle twisted bilayer graphene The underlying physical mechanism responsible for the structural transformation occur- ring in twisted BLG in the limit of small twist angle is the interplay between van der Waals forces, responsible for the interaction between the two graphene layers, and in-plane elastic- ity forces. As shown in Fig. 1(a), along the diagonal of a moiré supercell of twisted BLG the local stacking evolves through the high-symmetry configurations AA, AB, SP, BA, and AA (see Fig. 1(b) for naming conventions). The binding energy is minimal for AB/BA stacking configuration, whereas AA configuration corresponds to the maximum of the potential en- ergy surface, see Fig. 1(b). The local energy maximum in the middle of the path connecting AB and BA corresponds to an additional high-symmetry stacking configuration commonly referred to as SP. This potential energy landscape leads to in-plane forces that displace atoms in order to maximize the area of AB/BA stacking domains. On the other hand, the in-plane atomic rearrangement is hindered by the strain caused by the atomic displacement itself. The ultimate equilibrium structures result from the competition between the minimization of the interlayer energy and the reaction of the strain field [26]. In our simulations, we investigate the equilibrium structure of twisted bilayer graphene by treating atomic interactions using a classical potential. Previous DFT studies of twisted BLG have been performed within LDA or GGA functionals that disregard dynamical charge correlations responsible for the van der Waals (vdW) interaction [14, 23, 25]. As we aim 3 to treat models with up to N = 3 × 105 atoms, ab initio calculations become prohibitive. Currently available implementations of classical potentials for carbon do not reproduce cor- rectly the interlayer energy of layered structures based on sp2-hybridized carbon atoms [27–31]. In order to describe correctly the interlayer interaction, we define a new poten- tial VLCBOP/KC = VSR + VLR formed by a short range contribution, VSR, inherited from the LCBOP potential [28] and a long-range registry-dependent contribution, VLR, described by a reparametrized version of the Kolmogorov-Crespi potential [32]. The parameters of VLR have been fitted in order to reproduce several observables calculated within DFT+vdW (see Section Methods and Supplementary Materials). Notably, our DFT+vdW calculations reproduce well the equilibrium interlayer distance of graphite 3.36 Å and the in-plane bond length 1.42 Å. For bilayer graphene, we have found the atomic bond-length dCC = 1.419 Å and the equilibrium interlayer distances of AB, AA, and SP stacking configurations, respectively, ∆zAB = 3.412 Å, ∆zAA = 3.599 Å, and ∆zSP = 3.439 Å. We have found that the interlayer energies for AA and SP stacking configurations calculated at ∆z = 3.412 Å are, respectively, EAA b = 12.2 meV/atom and b = 1.33 meV/atom relative to AB configuration. ESP Supercells of twisted bilayer graphene have been built according to the rules derived by imposing commensurability conditions [33, 34]. In particular, one class of supercells is defined by an integer w that determines the supercell periodicity vectors t1 = wa1+(w+1)a2 √ 3dCC) being the crystal vectors of the graphene honeycomb lattice, and the corresponding twist angle is defined by cos θ = (3w2 + 3w + 1/2)/(3w2 + 3w + 1) [33]. Vectors t1 and t2 form a 60◦ angle and the moiré pattern has C3 symmetry, see Fig. 1(a). Notably, in the limit w → ∞ the twist angle and the supercell linear size are inversely proportional: θ−1 ∝ t1 = L. We have performed atomic structure relaxation of models with index w up to 160 corresponding to θ = 0.206◦, L = 68.4 nm, and total number of atoms N = 309124. and t2 = −(w + 1)a1 + (2w + 1)a2, with a1 and a2 ((cid:12)(cid:12)a1/2 (cid:12)(cid:12) = Figs. 2(a,b) show a model of rigidly twisted BLG with θ = 0.235◦ and the correspond- ing equilibrium structure resulting from relaxation. Upon relaxation, AB(BA) regions have extended and transformed in approximately triangular domains with (cid:39) 40 nm side, while AA regions reduce to much smaller, nanometer-scale patches in order to minimize the en- ergy penalty payed with respect to Bernal stacking, see Fig. 1(b). The segments sepa- rating AB/BA domains form a hexagonal lattice with vertices corresponding to the AA 4 Figure 1. Interlayer interactions in twisted bilayer graphene. a Ball-and-stick representation of a model of twisted bilayer graphene characterized by twist angle θ = 4.4◦ and moiré periodicity L = 3.2 nm. The moiré supercell is highlighted by dashed lines. Along the black line the stacking order evolves through AA, AB, SP, BA, and AA configurations defined in b. b Interlayer binding energy of bilayer graphene Eb calculated within DFT+vdW as a function of interlayer lateral displacement ∆x (see ball-and-stick schemes where atoms with different colors belong to opposite layers). The interlayer distance is fixed to ∆z = 3.412 Å. The energy reference Eb (∆x = ±0.5 dCC) corresponds to AB/BA stacking configuration. regions. The modulus of the atomic in-plane displacement upon relaxation, ∆d, along a high-symmetry path is shown in Fig. 2(d). Along the line connecting point A or B with point M, carbon atoms increasingly displace to restore AB/BA stacking. Because of the opposite value of ∆d along segments AM and BM, strain concentrates in the vicinity of M. When the energy gained by restoring AB or BA stacking order is compensated by the local strain energy, ∆d reaches a maximum before abruptly vanishing. The distance WD between the two symmetric maxima converges to a constant value W (cid:63) D = 10.5 nm upon increasing the moiré periodicity L. This illustrates the emergence of a shear strain soliton separating two stacking domains. As shown in Fig. 2(f), the modulus of shear strain vector, ∆u, cor- responds to one carbon-carbon bond length, ∆u = 1.42 Å. A similar reasoning is valid for the path BG. Additionally, atoms displace in the out-of-plane direction as shown in Fig. 2(c). Out-of-plane relaxation of twisted BLG has been investigated within DFT/LDA for 2◦ < θ finding out-of-plane corrugation at low twist angles [25]. A work based on a 5 Figure 2. Atomic relaxation of a model of twisted bilayer graphene characterized by θ = 0.235◦ and L = t1 = 59.8 nm (w = 140). a Representation of the initial model with rigidly rotated layers. Pairs of atoms in opposite layers whose lateral positions are closer than 0.2 Å are colored in red, the remaining in blue. The AA stacking regions (within 0.2 Å tolerance) contain only red atoms, whereas atoms in AB stacking regions are alternatively colored in blue and red, and hence such regions appear in purple. Regions with neither AA nor AB stacking are blue. This representation allows distinct stacking domains to be recognized at a glance. b Representation of the relaxed structure with the same color-coding procedure as in a. c Interlayer distance ∆z for the relaxed system along the path AMBG defined in b. d Absolute magnitude of the atomic displacement driven by the in-plane strain along the path AMBG. e Dependence of WD, defined as the distance between the two symmetric maxima with respect to M, on the moiré periodicity L. f Shear soliton separating AB and BA domains arising from structural relaxation, with ∆u representing the shear strain vector. classical potential investigated twisted BLG in the regime 0.46◦ < θ < 2.1◦ [35]. Similarly to what was reported in Ref. [35], we find that ∆z approaches the value ∆zAB in AB/BA stacking domains giving rise to the plateaus seen in Fig. 2(c). Moreover, ∆z adapts to ∆zSP and ∆zAA, respectively, at M and G, consistent with the local stacking configurations [36]. 6 Therefore, this leads to a small corrugation (tilt angle of the normal vectors α < 0.2◦) at the location of shear solitons as well as their junctions [26, 36]. To obtain a deeper insight into the equilibrium configurations of twisted BLG, we have studied the stacking vector field u, defined as the in-plane component of the minimal shift that has to be applied to one layer in order to make it coincide locally with the opposite layer, see Fig. 3(a). Fig. 3(b) shows the presence of triangular domains with almost constant stacking u = 1.42 Å (white regions). By inspecting the local stacking field around A and B (side panels in Fig. 3(b)) one can see the confluence of three orientations of u differing by 120◦. This discontinuity is trivial as the vector u for AB(BA) stacking has three degenerate represen- tations forming 120◦ angles with each other, see Fig. 3(a). However, when following a path connecting one stacking domain with another across a shear soliton, (e.g. from point A to point B), u rotates by 60◦, that is, the stacking configuration changes from AB to BA or vice versa. The variation ∆u (∆u = 1.42 Å) is parallel to the strain soliton and coincides with its shear vector. These stacking domain boundaries are topological defects and ∆u is assigned as their topological invariant. In the following, the denominations "shear soliton" and "stacking domain boundary" will be used interchangeably. S = 9.5 nm, a value close to W (cid:63) D = 10.5 nm. We choose to use W (cid:63) As shown in Fig. 2(c), along the path AMB u has a minimum at M, corresponding to SP stacking (u = 1.23 Å). Upon increasing L, the full width at half maximum of this dip, WS, S to define the saturates to W (cid:63) width of the stacking domain boundaries since its determination does not require reference to the corresponding rigidly twisted structure. The calculated widths of the solitons are in good accordance with the experiments [9, 10, 21]. The vertices of the hexagonal network (G) where six stacking domain boundaries merge are topological point defects with u = 0 and a non-zero winding number, that is, u rotates by 360◦ along a closed path encompassing G [37]. Noteworthy, that in the vicinity of these vertices, due to energetically unfavorable stacking of the latter close to AA, the width of the solitons is smaller than W (cid:63) S . Intuitively, we expect that a transformation involving the creation of strain solitons takes S such that the strain field can be place when the twisted BLG supercell is larger than W (cid:63) efficiently accommodated. The dependence of several observables on θ (or, equivalently, L) reveals further details of the evolution. As shown in Fig. 4(a), for θ = 21.8◦ the distribution of the interlayer distance, ∆z, has a minimal spread ∆zMAX − ∆zMIN (cid:39) 0.01 Å. Upon 7 Figure 3. Representation of the stacking vector field u of a model of twisted BLG. a Illustration of the stacking vector field u for high-symmetry configurations. b Color-coded representation of field u for the equilibrium configuration of a twisted BLG model with θ = 0.235◦, L = 59.8 nm (w = 140). Hue and saturation at each point represent, respectively, the direction and the intensity of the local value of u. Fully saturated colors correspond to AA stacking configuration (u =0 Å), white regions (vanishing saturation) correspond to AB stacking configuration (u =1.42 Å). SP- stacked soliton centers (u =1.23 Å) are half-saturated. Hue varies with a period of 180◦ as shown in the wind-rose diagram. Lateral panels show the stacking vector field in the vicinity of B and A. c Absolute magnitude of u along the path AMBG for equilibrium and rigidly twisted BLG. The full width at half maximum of the dip at M is referred to as a WS and used as definition of the soliton width. The dependence of WS on the periodicity L is shown in d. reduction of θ, ∆zMAX and ∆zMIN increasingly differ and at θ (cid:39) 2◦ saturate to ∆zMAX=∆zAA and ∆zMAX=∆zAB, consistent with the data shown in Fig. 2(c). Out-of-plane relaxation competes with the bending rigidity of graphene, estimated as BM = 1.44 eV [38]. For lower values of θ in-plane atomic displacements become non-negligible. Note that the maximum in-plane displacement of individual atoms, ∆D, is bounded from above by a half bond length 8 Figure 4. Structural evolution of twisted bilayer graphene in the limit of small twist angle. a Largest (smallest) interlayer distance ∆zMAX(∆zMIN) as a function of moiré periodicity L. b Maximum atomic displacement ∆D as a function of twist angle θ. d Density of twist energy γ as a function of moiré periodicity L. Regions with constant and inversely proportional dependence have been fitted, respectively, by red and blue lines, intersecting at L∗ = 11.9 nm (corresponding to crossover angle θ∗ = 1.2◦). √ dCC/2 = 0.71 Å since two inequivalent stacking vectors are connected by ∆u < dCC and the displacement is equally distributed over the atoms in the two layers. The crossover is underpinned by the change in functional dependence of the density of twist energy on the moiré periodicity: γ (L) = (E (L) − EAB) /AS, where E (L) is the total energy for a supercell of periodicity L, and EAB is the energy of a AB bilayer graphene supercell having 3/2, see Fig. 4(c). For small values of L, atomic in-plane the same surface area AS = L2 displacements due to relaxation are negligible and the energy required to introduce a twist arises from those regions whose stacking configuration is not AB/BA, which represent a constant fraction of the supercell surface. Thus, the difference E (L) − EAB is proportional to the surface of the system and the twist energy density equals a constant: γ (L) = γA = 1.2 meV/Å2. This has been confirmed by DFT calculations, see Fig. S1(c) of Supplementary Informations. On the other hand, for large supercells most of whose area is composed of AB and BA stacking domains, only the soliton network contributes to E (L)−EAB. As the width of the solitons asymptotically approaches the constant value W (cid:63) S , the twist energy density is given by γ = 3γSL/AS ∝ 1/L, where γS is the energy per soliton unit length and the factor 3 counts the number of solitons in the moiré supercell. We estimate γS = 42 meV/Å. The crossover length L(cid:63) = γS/γA = 11.9 nm, corresponding to the crossover angle θ(cid:63) = 1.2◦, is defined as the intersection of the constant line and the curve ∝ 1/L fitting the two distinct 9 regimes, as shown in Fig. 4(c). Finally, we can rigorously answer why the transformation takes place at large moiré periodicities. Regardless the values of γS and γA, the constant "rigid" regime is favorable for L < γS/γA, whereas the ∝ 1/L "solitonic" regime is favorable for L > γS/γA. B. Electronic structure of twisted bilayer graphene The low-energy states of twisted BLG with large to intermediate twist angles 3◦ (cid:46) θ (cid:46) 15◦ can be described by a model that introduces the coupling between graphene layers perturba- tively [33, 39, 40]. This model predicts the existence of low-energy massless Dirac fermions with θ-dependent Fermi velocity and a pair of Van Hove singularities slightly asymmetric with respect to the Dirac point. These predictions have been confirmed experimentally [15, 41–43]. For smaller twist angles 1◦ < θ < 3◦, twisted BLG has been predicted to de- velop a flat band responsible for a zero-energy peak in the density of states (DOS) [23, 25]. This peak is due to states localized in AA regions as a result of the super-periodic potential induced by the moiré pattern. However, these results cannot be extrapolated to lower values of θ, as we expect that the structural relaxation suppressing AA-stacked regions strongly affects the electronic structure. We investigate the low-energy electronic properties of the equilibrium structure of twisted bilayer graphene in the limit θ → 0◦ by means of a tight-binding model taking into account 2pz orbitals with hopping parameters depending on the distance between orbital centers as well as the relative orientation of the orbitals. The latter is achieved by means of the Slater-Koster theory [44], see Supplementary Information for details. In Fig. 5(a), the band structure for a model with twist angle θ = 3.8◦ shows two degenerate Dirac cones projected onto the K point of the supercell Brillouin zone, in contrast to the parabolic dispersion of AB graphene [14, 33, 34]. A finite coupling between the states in the two Dirac cones is responsible for low-dispersion bands around the M point, whence the appearance of two low-energy Van Hove singularities in the DOS [33]. We find that the relaxation has negligible effects on the Dirac fermions, except for lifting the degeneracy of the low-energy bands. As the twist angle decreases, the positions of each Van Hove singularity approaches the Dirac energy, eventually merging at θ (cid:39) 2◦. In this regime, the Fermi velocity is zero and the low-energy states are localized in the AA regions. As shown in 10 Figure 5. Electronic structure of equilibrium and rigidly twisted BLG. a,b Energy bands and density of states (DOS) for two models of twisted BLG characterized by twist angles (a) θ = 3.8◦ (L = 3.6 nm, w = 8) and (b) θ = 1.2◦ (L = 12.1 nm, w = 28). The energy bands and the DOS calculated for an equivalent supercell of AB graphene are plotted for comparison. Energies are referenced to the Fermi level. Fig. 5(b), at the crossover angle θ(cid:63) = 1.2◦ the DOS for rigidly twisted BLG shows a triplet of peaks at energies E(cid:39) − 0.06, 0, and 0.07 eV that correspond to the flat low-energy bands observable in the band structure. We have found that the relaxation is responsible for lifting the degeneracy of the central band and shifting the side peaks further away from the Fermi level, however, without introducing qualitative modifications of the electronic structure. Fig. 6(a) shows the density of states plot for a model characterized by θ = 0.235◦, that is, well below the crossover angle θ(cid:63). The DOS of rigidly twisted BLG still exhibits a zero- energy peak with two satellite shoulders. The nature of the low-energy states is revealed by the inspection of local density of states (LDOS) integrated in the energy regions around each of the three peaks, see Fig. 6(b). In all cases the charge density is localized in AA regions and extends on a fraction of surface η (cid:39) 5% for the central peak and η (cid:39) 25% for satellites, that we found to be largely independent of the moiré periodicity L. The picture changes drastically upon in-plane relaxation responsible for the discussed structural transformation. The DOS of relaxed twisted BLG is overall closer to AB bilayer graphene and, in particular, the zero-energy peak due to the localized states in the AA region is strongly suppressed (Fig. 6(a), inset). We note that the localized states are still present and confined to nanometer-size AA regions, as can be seen in the corresponding LDOS maps (compare Figs. 6(b) and 6(c), central map). The suppression can be explained 11 Figure 6. Density of states for equilibrium and rigidly twisted BLG. a Density of states (DOS) as a function of energy E for twisted bilayer graphene with θ = 0.235◦ (L = 59.8 nm). DOS for equilibrium and rigidly twisted bilayer graphene are compared. As a reference, the DOS of AB bilayer graphene (θ = 0◦) is also shown (dashed line). The inset presents the same data in the energy range [−0.5, 0.5] eV. Regions L, C and R individuate, respectively, the energy intervals [−0.25,−0.038] eV, [−0.038, 0.043] eV, and [0.043, 0.26] eV. Energies are referenced to the Fermi level. b,c Local density of states (LDOS) integrated in the energy intervals L, C, and R for (b) rigidly twisted BLG and for (c) the corresponding equilibrium configuration. Solitons are highlighted by dashed lines. d Same as right panel in c restricted to 1 nm2 squares centered around M, B, and G. as follows. In the solitonic regime, for twist angles θ < θ(cid:63), relaxation leads to AA regions of constant area. Therefore, upon decreasing twist angle θ, or equivalently, increasing moiré periodicity L, the weight of these states in the total DOS decreases as L−2. This overall decrease of the zero-energy peak weight is expected to be accompanied by its narrowing as suggested previously [45]. However, our numerical calculations performed on large-scale models do not allow to address adequately this effect. The effect of structural relaxation is even more dramatic in the energy ranges corresponding to the side peaks (Fig. 6(c), maps L and R). Indeed, the charge density is partially depleted in AA regions and AB/BA domains show an overall homogeneous distribution, whereas solitons exhibit a slightly larger charge density. Upon closer investigation of the center of an AB domain (point B in Fig. 6(d)) one can observe an alternation of atoms with high and low charge density. This is typical of 12 graphene layers with AB stacking configuration as demonstrated by STM images of highly ordered pyrolytic graphite [46]. The reason is the nonequivalence of the two sublattices of AB stacked graphene that reflects different out-of-plane matrix elements for atoms in complementary sublattices. Charge densities in the soliton regions and at the vertices of the network do not show local variations on the atomic scale (Fig. 6(d), points M and G). This is consistent with the fact that the stacking configurations SP and AA found in the solitons and in the vertices, respectively, preserve the sublattice equivalence. II. CONCLUSIONS We investigated the equilibrium low-energy structure of twisted bilayer graphene in the limit of vanishing twist angle (down to θ ≈ 0.2◦) by means of simulations based on a classical potential, which is capable of describing the dependence of the interlayer binding energy on the relative position of the two layers. Carbon atoms displace in order to maximize the area of energetically favorable AB/BA stacking domains that assume a triangular shape. The in-plane strain field, thus, appear confined in a hexagonal network of shear solitons of width WS (cid:39) 9.5 nm, that delimit alternating AB and BA stacking domains. This structural transformation is continuous and takes place at twist angles below the crossover value θ(cid:63) = 1.2◦, at which the moiré superlattice period exceeds the soliton width WS. In the limit θ → 0◦, the equilibrium structure of twisted BLG converges to that of ideal AB-stacked BLG (θ = 0◦). However, the convergence is not uniform in the sense that the relative abundance of the AB-stacking regions approaches 1, but the soliton network due to its topological nature vanishes only at θ = 0◦. On the other hand, twisted BLG as such is not stable with respect to AB-stacked BLG (θ = 0◦) and its existence is governed by kinetic bottlenecks. This fact has major consequences on the low-energy electronic states of the moiré super- lattice. Differently from the range 1.2◦ < θ < 2◦ where the DOS of twisted BLG hosts three low-energy peaks due to flat bands of states localized on AA regions, equilibrium structures of twisted BLG with θ < 1.2◦ show a DOS resembling that of AB bilayer graphene with a low-energy charge density distribution that can be directly inferred from the local stack- ing. The charge density is uniform overall in AB/BA-stacked domains, but shows a strong imbalance between the two inequivalent sublattices in each layer. Conversely, the solitons 13 and the network vertices show no breaking of sublattice symmetry. This distinctive pattern enables the identification of the stacking domain boundaries by means of STM experiments. Analogously to the stacking, the relative extent of the regions where the charge distribution differs from that of AB bilayer graphene asymptotically vanishes for θ → 0◦. III. METHODS DFT calculations. We have employed the rVV10 functional that treats exchange- correlation energy within the GGA and includes a non-local van der Waals (vdW) contribu- tion [47, 48] implemented in QUANTUM ESPRESSO [49]. The ion-electron interaction has been described by means of ultra-soft pseudopotentials [50]. Energy cutoff for wavefunc- tions and charge density have been set, respectively, to Ewf = 80 Ry and Erho = 574 Ry and the Brillouin zone has been sampled with a 16 × 16 × 1 Monkhorst-Pack kpoint grid. All computational parameters and technical details are listed and discussed in Supplementary Information. Classical potential simulations. We have employed the long-range carbon bond order potential (LCBOP) replacing the original long-range contribution by a reparametrized ver- sion of Kolmogorov-Crespi registry-dependent potential [32] fitted to match the DFT/rVV10 values of the interlayer binding energy as a function of interlayer distance and relative shift. The fit has been performed with the non-linear minimizer provided by DAKOTA code [51]. Additional details about the fit procedure and the resulting parameters can be found in Supplementary Information. The optimized potential has been implemented in LAMMPS to perform energy minimizations [52, 53]. Electronic structure calculations. We have considered a Slater-Koster [44] tight-binding model taking into account 2pz orbitals for carbon atoms with hopping parameters depending on the distance between orbital centers as well as the relative orientation of the orbitals. This is particularly important in order to describe correctly the interactions in the soliton region where the relative position of carbon atoms in opposite layers changes continuously. Since the equilibrium structures show only weak corrugation, in-plane orbital interactions are predominantly of the ppπ type. For pairs of atoms in opposite layers that are stacked on top of each other, such as those appearing in AA stacking, the orbital interaction is purely of ppσ character. However, when atoms are misaligned such as in SP or AB stacking con- 14 figurations, the interaction is a mixture of ppσ and ppπ types. Tight-binding Hamiltonians have been diagonalized using the massively parallel linear algebra library ELPA [54] that allowed us to treat matrices of order up to N = 236884. Details on Hamiltonian matrix elements and observable calculations are discussed in Supplementary Information. IV. ACKNOWLEDGMENTS We thank Bastien F. Grosso and Gabriel Autès for collaboration, Riccardo Sabatini, Marco Gibertini and Tommaso Grioni for useful discussions. Anton Kozhevnikov and Luca Marsella assisted us with compiling ELPA libraries. This research was supported by the Swiss NSF grant No. PP00P2_133552 and Graphene Flagship. This work took advantage of the computational facilities of the Swiss National Computing Centre (project s675). 15 SUPPLEMENTARY INFORMATION S1. DFT STUDY OF INTERLAYER INTERACTION IN BILAYER GRAPHENE All our DFT calculations are based on the non-local rVV10 functional implemented in QUANTUM ESPRESSO [49]. This functional is composed of the revised-PW86 exchange- correlation functional plus a non-local contribution to account for the van der Waals inter- action [47, 48, 50, 55]. Ion-electron interactions have been treated by means of an ultrasoft pseudopotential generated with the revised-PW86 functional [50]. In fact, non-local con- tributions are not expected to alter the effective potential generated by nuclei and core electrons. As already found for graphite by the authors of the rVV10 functional, relatively high values of wavefunction and charge-density cutoffs, respectively, Ewf and Eρ, are required to describe accurately sp2-carbon systems. We have used Ewf = 80 Ry and Eρ = 574 Ry together with a 16 × 16 × 1 Monkhorst-Pack k-point grid to obtain converged structural observables. Periodic replicas of the system are separated by 24 Å of vacuum to guarantee negligible interaction. The interlayer binding energy Eb has been calculated from the total energy, Etot, the energy of one isolated graphene layer, Emono, and the number of atoms in the unit cell, Nc, as follows Eb = − 1 Nc (Etot − 2Emono) . (S1) As reported in Ref. [48] and confirmed by our calculations, the atomic bond length of graphite is 1.42 Å and the interlayer distance is ∆zgraph = 3.36 Å in accordance with established values [7]. Table S1 reports the interlayer distance and the binding energy for AA and AB bilayer graphene calculated at fixed bond length dCC = 1.42 Å. Table S1. Structural observables calculated within DFT+vdW for graphite and bilayer graphene. ∆z(cid:0)Å(cid:1) Graphite [48] 3.36 Eb (meV/atom) 39 Bilayer graphene - AB Bilayer graphene - AA 3.41 30.2 3.59 25.7 We have found for AB bilayer graphene the interlayer distance ∆zAB = 3.412 Å, about 1.5% larger than for graphite, consistently with previous reports [56]. For AA stacking 16 configuration the equilibrium distance ∆zAA = 3.588 Å. The dependencies of Eb on ∆z and on the interlayer shift, ∆x (see Fig. 1(b) of the main text for definition), are shown in Fig. S1(a,b). S2. DETERMINATION OF THE CLASSICAL CARBON-CARBON POTENTIAL The classical pair potential for carbon atoms that we have employed in structural relax- ations consists in the sum of a short-range contribution VSR and a long-range contribution VLR describing, respectively, covalent bonds and van der Waals interactions between sp2- hybridized carbon atoms. VSR is the short-range term of the LCBOP potential defined in Ref. [28], adopted with no modifications. The long-range term VLR is a reparametrized version of the registry-dependent potential [32]. For a pair of atoms at positions ri and rj, with ni (nj) being the proposed in Ref. normal vector to the sp2 hybridization plane at position ri (rj), VLR is defined as (cid:19)−6 (cid:18) rij ji − (nj · rji)(cid:1)1/2 , z0 , ρji =(cid:0)r2 n = 0, 1, 2, (S2) VLR (rij, ni, nj) = e−λ(rij−z0) (C + f (ρij) + f (ρji)) − A ij − (ni · rij)(cid:1)1/2 , ρij =(cid:0)r2 f (ρ) = e−(ρ/δ)2(cid:88) (cid:16)(cid:0)rx (cid:1)2(cid:17)1/2 (cid:1)2 +(cid:0)ry ij C2n(ρ/δ)2n rij = ri − rj. ij We have made the approximation that normal vectors ni are directed along the z axis. Consequently, ρij = ρji = . This assumption is justified by the inspection of the corrugation of the equilibrium structures. For all relaxed models we have found that the normal vectors form an angle α < 0.2◦ with the z axis. Although the original paper of Kolmogorov and Crespi provides a set of parameters for the potential, we have reparametrized it by fitting Eq. (S2) to three data sets calculated within DFT+vdW. The first two datasets are the binding energy Eb as a function of ∆z for AB- and AA-stacked bilayer graphene (see dot data series in Fig. S1) and the third dataset is Eb as a function of ∆x at fixed interlayer distance ∆z = (cid:0)∆zDFT (cid:1) /2 = 3.50 Å (red symbols in AA + ∆zDFT AB Fig. S1(b)). The fit has been performed employing the non-linear optimizer DAKOTA [51]. Table S2 compares the parameters reported in the original reference and those resulting from our fit. 17 Table S2. Parameters for the Kolmogorov-Crespi potential δ(cid:0)Å(cid:1) λ(cid:16) Å−1(cid:17) A(cid:0)Å(cid:1) C C0 C2 C4 (meV) (meV) (meV) (meV) z0(cid:0)Å(cid:1) x. x 3.030 15.71 12.29 4.933 3.34 0.578 0.578 10.238 This work 7.183 9.806 5.365 4.266 3.516 0.590 3.039 13.17 Fig. S1 shows a remarkable accordance between observables calculated within DFT (sym- bols data series) and KC potential (continuous curves). In particular, the equilibrium dis- tances calculated by means of the reparametrized KC potential are ∆zKC AA = 3.599 Å and AB = 3.416 Å, and the respective binding energies for AA and SP stacking configura- ∆zKC tions calculated at ∆z = 3.412 Å are EAA b = 1.22 meV/atom relative to AB stacking configuration. Finally, the in-plane carbon-carbon bond length is dCC = 1.419 Å. b = 12.2 meV/atom and ESP Figure S1. a Interlayer binding energy Eb as a function of interlayer distance ∆z for AA and AB-stacked bilayer graphene. b Binding energy as a function of interlayer shift ∆x (see Fig. 1(b) of the main text for definition). The energy zero is set at Eb (∆x = ±0.5 dCC) corresponding to AB/BA stacking configuration. Symbols correspond to DFT+vdW values, whereas continuous lines correspond to classical potential results. c Density of twist energy γ as a function of moiré periodicity L calculated with DFT and LCBOP/KC. As a test of the transferability of the classical potential resulting from our fit, in Fig. S1(c) we show a comparison of the density of twist energy calculated within DFT and LCBOP/KC 18 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 10 12 14 16 18 20 22 24γ (meV/Å2)L (Å)LCBOP+KCDFTfor several bilayer graphene models. In particular, we focus on systems with small moiré large twist angles 6◦ < θ < 13◦) in their equilibrium periodicity 10 Å < L < 24 Å (i.e. configuration obtained by atomic relaxation. Treating the interlayer interaction classically introduces a discrepancy that is smaller than 10% in most of the cases. S3. STRUCTURAL RELAXATION The full potential V = VSR + VLR has been implemented in LAMMPS [52, 53]. Twisted BLG structures have been initially relaxed by means of conjugate gradient plus quadratic line search method [57] and fine minimization was obtained using fast inertia relaxation method [58]. The supercell vectors have been kept fixed and the initial interlayer distance has been set to ∆zKC AB = 3.416 Å. At the end of the relaxation the largest force component acting on any atom was below 3 meV/Å. In Fig. S2, the full maps of the interlayer distance and in-plane atomic displacement are shown. With respect to Fig. 3(c,d) of the main text, the full maps allow to appreciate the whole set of symmetries. ∆z is almost constantly equal to ∆zKC SP = 3.439 Å along the soliton lines and to ∆zKC AB = 3.416 Å in AB/BA domains. Figure S2. a Maps of the interlayer distance ∆z and b the absolute magnitude of the atomic displacement ∆d upon relaxation of a rigidly twisted BLG model characterized by the twist angle θ = 0.235◦, corresponding to the moiré periodicity L = 59.8 nm. The upper bound for the atomic displacement is dCC/2 = 0.71 Å, as explained in the main text. 19 S4. ELECTRONIC STRUCTURE CALCULATIONS In this section, we present the details of our electronic structure calculations. The tight- binding model Hamiltonian for bilayer graphene has been taken from Ref. [23] (see also Ref. [24] for thorough electronic structure calculations of rigidly twisted bilayer graphene). In such a model, only pz-orbitals for carbon atoms are considered. The Hamiltonian is defined as H = † i aj, Vija (S3) (cid:88) i(cid:54)=j † where the operators a i and ai, respectively, create and annihilate an electron in the pz- orbital of the atom at position Ri. The matrix elements, Vij, are obtained by combining σ- and π-type Slater-Koster parameters Vppσ and Vppπ in the approximation that the axes of pz-orbitals are parallel, akin to the assumption that the normal vectors of the two graphene layers are parallel. One has Vij = Vppπ sin2 (θ) + Vppσ cos2 (θ) , (S4) where θ is the angle between the orbital axes and the vector Rij = Ri− Rj connects the two orbital centers [44]. For a pair of atoms in the same layer θ = 90◦ and Vij = Vppπ. Conversely, for a pair of atoms placed on top of each other in opposite layers, namely forming a dimer, θ = 0◦ and Vij = Vppσ. Vppπ and Vppσ depend exponentially on the distance between the two orbital centers r = Ri − Rj as Vppπ (r) = V 0 ppπeqπ(1−r/aπ), Vppσ (r) = V 0 ppσeqσ(1−r/aσ). (S5) ppπ = −2.7 eV, V 0 Following Ref. [23] we assume V 0 ppσ = 0.48 eV, aπ = 1.419 Å, qπ = 3.1454. Differently from Ref. [23], in order to be consistent with the interlayer distance for AB stacking configuration calculated in the present work, we have taken aσ = 3.417 Å and qσ = 8.200. The long distance cut-offs for Vppσ (r) and Vppπ (r) have been fixed, respectively, at ¯rσ = 3.5 Å and ¯rπ = 5 Å. We have verified that further increasing these cut-offs does not affect the calculated observables. In all our calculations the charge neutrality point, corresponding to the Fermi energy for undoped systems, is Ef = 0.82 eV. Density of states are calculated as follows DOS (E) = dk δ (E − Enk) , (S6) (cid:90) (cid:88) BZ nk 20 with nk running over all the eigenvalues at position k in reciprocal space. For computational needs the δ-function appearing in eq. S6 has been replaced by a Lorentzian function: δ (E − Enk) → 1 π η (E − Enk)2 + η2 , η → 0+. (S7) The local density of states on the i-th atom at position Ri, integrated in the energy range [E1, E2], has been calculated as follows (cid:90) E2 (cid:90) LDOS (i; E1, E2) = dE = dk E1 BZ δ (E − Enk)(cid:104)ink(cid:105)2 = (cid:90) BZ dk (cid:88) nk (cid:88) (cid:104)ink(cid:105)2 , E1<Enk <E2 (S8) where i(cid:105) represents the pz orbital of the i-th atom. Spin has not been explicitly considered in our calculations. Integration over the Brillouin zone have been performed introducing a discrete grid. The band structures shown in panels (a) and (b) of Fig. 5 of the main text have been calculated with a Monkhorst-Pack grid of 25 × 25 and 5 × 5 k-points, respectively. The density of states of the model shown in Fig. 6 of the main text has been calculated using the high- symmetry points Γ and M of the hexagonal Brillouin zone, taking advantage of the fact that the Hamiltonian represented in reciprocal space H (k = M, Γ) is a real matrix. Matrix diagonalizations have been performed employing the Eigenvalue Solvers for Petaflop Applications library (ELPA) [54]. This allowed us to diagonalize a N × N real matrix with N = 236884 in about 2.5 hours using 1024 CPU cores. [1] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). [2] T. Ohta, A. Bostwick, T. Seyller, K. Horn, and E. Rotenberg, Science 313, 951 (2006). [3] J. B. Oostinga, H. B. Heersche, X. Liu, A. F. Morpurgo, and L. M. K. Vandersypen, Nat. Mater. 7, 151 (2008). [4] 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). [5] K. Lee, B. Fallahazad, J. Xue, D. Dillen, K. Kim, T. Taniguchi, K. Watanabe, and E. Tutuc, Science 345, 58 (2014). 21 [6] J. D. Bernal, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences 106, 749 (1924). [7] G. E. Bacon, Acta Crystallog. 4, 558 (1951). [8] B. Lalmi, J. C. Girard, E. Pallecchi, M. Silly, C. David, S. Latil, F. Sirotti, and A. Ouerghi, Sci. Rep. 4, 4066 (2014). [9] J. Lin, W. Fang, W. Zhou, A. R. Lupini, J. C. Idrobo, J. Kong, S. J. Pennycook, and S. T. Pantelides, Nano Lett. 13, 3262 (2013). [10] B. Butz, C. Dolle, F. Niekiel, K. Weber, D. Waldmann, H. B. Weber, B. Meyer, and E. Spiecker, Nature 505, 533 (2014). [11] L. Gong, R. J. Young, I. A. Kinloch, S. J. Haigh, J. H. Warner, J. A. Hinks, Z. Xu, L. Li, F. Ding, I. Riaz, R. Jalil, and K. S. Novoselov, ACS Nano 7, 7287 (2013). [12] M. Yankowitz, J. I.-J. Wang, A. G. Birdwell, Y.-A. Chen, K. Watanabe, T. Taniguchi, P. Jacquod, P. San-Jose, P. Jarillo-Herrero, and B. J. LeRoy, Nat. Mater. 13, 786 (2014). [13] A. M. Popov, I. V. Lebedeva, A. A. Knizhnik, Y. E. Lozovik, and B. V. Potapkin, Phys. Rev. B 84, 045404 (2011). [14] J. Hass, F. Varchon, J. E. Millán-Otoya, M. Sprinkle, N. Sharma, W. A. de Heer, C. Berger, P. N. First, L. Magaud, and E. H. Conrad, Phys. Rev. Lett. 100, 125504 (2008). [15] G. Li, A. Luican, J. M. B. Lopes dos Santos, A. H. Castro Neto, A. Reina, J. Kong, and E. Y. Andrei, Nat. Phys. 6, 109 (2010). [16] L. Brown, R. Hovden, P. Huang, M. Wojcik, D. A. Muller, and J. Park, Nano Lett. 12, 1609 (2012). [17] C.-C. Lu, Y.-C. Lin, Z. Liu, C.-H. Yeh, K. Suenaga, and P.-W. Chiu, ACS Nano 7, 2587 (2013). [18] K. S. Kim, A. L. Walter, L. Moreschini, T. Seyller, K. Horn, E. Rotenberg, and A. Bostwick, Nat. Mater. 12, 887 (2013). [19] T. E. Beechem, T. Ohta, B. Diaconescu, and J. T. Robinson, ACS Nano 8, 1655 (2014). [20] I. Razado-Colambo, J. Avila, J.-P. Nys, C. Chen, X. Wallart, M.-C. Asensio, and D. Vignaud, Sci. Rep. 6, 27261 (2016). [21] J. S. Alden, A. W. Tsen, P. Y. Huang, R. Hovden, L. Brown, J. Park, D. A. Muller, and P. L. McEuen, Proc. Natl. Acad. Sci. USA 110, 11256 (2013). [22] L. Jiang, Z. Shi, B. Zeng, S. Wang, J.-H. Kang, T. Joshi, C. Jin, L. Ju, J. Kim, T. Lyu, Y.-R. 22 Shen, M. Crommie, H.-J. Gao, and F. Wang, Nat. Mater. 15, 840 (2016). [23] G. Trambly de Laissardière, D. Mayou, and L. Magaud, Nano Lett. 10, 804 (2010). [24] G. Trambly de Laissardière, D. Mayou, and L. Magaud, Phys. Rev. B 86, 125413 (2012). [25] K. Uchida, S. Furuya, J.-I. Iwata, and A. Oshiyama, Phys. Rev. B 90, 155451 (2014). [26] S. Dai, Y. Xiang, and D. J. Srolovitz, Nano Letters 16, 5923 (2016). [27] D. W. Brenner, O. A. Shenderova, J. A. Harrison, S. J. Stuart, B. Ni, and S. B. Sinnott, J. Phys. Condens. Matter 14, 783 (2002). [28] J. H. Los and A. Fasolino, Phys. Rev. B 68, 024107 (2003). [29] L. Spanu, S. Sorella, and G. Galli, Phys. Rev. Lett. 103, 196401 (2009). [30] S. Lebègue, J. Harl, T. Gould, J. G. Ángyán, G. Kresse, and J. F. Dobson, Phys. Rev. Lett. 105, 196401 (2010). [31] M. Reguzzoni, A. Fasolino, E. Molinari, and M. C. Righi, Phys. Rev. B 86, 245434 (2012). [32] A. N. Kolmogorov and V. H. Crespi, Phys. Rev. B 71, 235415 (2005). [33] N. M. R. Lopes dos Santos, J. M. B.and Peres and A. H. Castro Neto, Phys. Rev. Lett. 99, 256802 (2007). [34] S. Shallcross, S. Sharma, and O. A. Pankratov, Phys. Rev. Lett. 101, 056803 (2008). [35] M. M. v. Wijk, A. Schuring, M. I. Katsnelson, and A. Fasolino, 2D Mater. 2, 034010 (2015). [36] S. K. Jain, V. Juričić, and G. T. Barkema, 2D Materials 4, 015018 (2017). [37] N. D. Mermin, Rev. Mod. Phys. 51, 591 (1979). [38] Y. Wei, B. Wang, J. Wu, R. Yang, and M. L. Dunn, Nano Lett. 13, 26 (2013). [39] R. Bistritzer and A. H. MacDonald, Proc. Natl. Acad. Sci. USA 108, 12233 (2011). [40] M. Koshino, Phys. Rev. B 88, 115409 (2013). [41] A. Luican, G. Li, A. Reina, J. Kong, R. R. Nair, K. S. Novoselov, A. K. Geim, and E. Y. Andrei, Phys. Rev. Lett. 106, 126802 (2011). [42] J. Jung, A. M. DaSilva, A. H. MacDonald, and S. Adam, Nat. Commun. 6, 6308 (2015). [43] D. Wong, Y. Wang, J. Jung, S. Pezzini, A. M. DaSilva, H.-Z. Tsai, H. S. Jung, R. Khajeh, Y. Kim, J. Lee, S. Kahn, S. Tollabimazraehno, H. Rasool, K. Watanabe, T. Taniguchi, A. Zettl, S. Adam, A. H. MacDonald, and M. F. Crommie, Phys. Rev. B 92, 155409 (2015). [44] J. C. Slater and G. F. Koster, Phys. Rev. 94, 1498 (1954). [45] P. San-Jose and E. Prada, Phys. Rev. B 88, 121408 (2013). [46] W.-T. Pong and C. Durkan, J. Phys. D: Appl. Phys. 38, R329 (2005). 23 [47] O. A. Vydrov and T. V. Voorhis, J. Chem. Phys 133, 244103 (2010). [48] R. Sabatini, T. Gorni, and S. de Gironcoli, Phys. Rev. B 87, 041108 (2013). [49] P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. De Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, J. Phys. Condens. Matter 21, 395502 (2009). [50] E. D. Murray, K. Lee, and D. C. Langreth, J. Chem. Theory Comput. 5, 2754 (2009). [51] B. Adams, L. Bauman, W. Bohnhoff, K. Dalbey, M. Ebeida, J. Eddy, M. Eldred, P. Hough, K. Hu, J. Jakeman, Swiler, L.P., and D. Vigil, DAKOTA, A Multilevel Parallel Object-Oriented Framework for Design Optimization, Parameter Estimation, Uncertainty Quantification, and Sensitivity Analysis: Version 5.4 User's Manual (Sandia National Laboratories, 2009). [52] S. Plimpton, J. Comput. Phys. 117, 1 (1995). [53] "LAMMPS molecular dynamics simulator." http://lammps.sandia.gov/. [54] A. Marek, V. Blum, R. Johanni, V. Havu, B. Lang, T. Auckenthaler, A. Heinecke, H.-J. Bungartz, and H. Lederer, J. Phys. Condens. Matt. 26, 213201 (2014). [55] J. P. Perdew and W. Yue, Phys. Rev. B 33, 8800 (1986). [56] E. Mostaani, N. Drummond, and V. Fal'ko, Phys. Rev. Lett. 115, 115501 (2015). [57] R. Fletcher, Practical Methods of Optimization (John Wiley & Sons, 2013). [58] E. Bitzek, P. Koskinen, F. Gähler, M. Moseler, and P. Gumbsch, Phys. Rev. Lett. 97, 170201 (2006). 24
1611.05532	3	1611	2018-11-07T17:00:56	A solid state emitter embedded in a microcavity under intense excitation: a variational master equation approach	[ "cond-mat.mes-hall", "quant-ph" ]	In this work, dissipative effects from a phonon bath on the resonance fluorescence of a solid state two level system embedded in a high quality semiconductor microcavity and driven by an intense laser, are investigated. Within the density operator formalism, we derive a variational master equation valid for broader ranges of temperatures, pumping rates, and radiation-matter couplings, than previous studies. From the obtained master equation, fluorescence spectra for various thermal and exciting conditions are numerically calculated, and compared to those computed from weak coupling and polaronic master equations, respectively. Our results evidence the break down of those rougher approaches under increased temperature and strong pumping.	cond-mat.mes-hall	cond-mat	A solid-state emitter embedded in a microcavity under intense excitation: a variational master equation approach Oscar J. G´omez-S´anchez1, 2 and Hanz Y. Ram´ırez1, ∗ 1Grupo de F´ısica Te´orica y Computacional, Escuela de F´ısica, Universidad Pedag´ogica y Tecnol´ogica de Colombia (UPTC), Tunja 150003, Boyac´a, Colombia. 2Department of Electrophysics, National Chiao Tung University (NCTU), Hsinchu 30050, Taiwan (Dated: November 8, 2018) Abstract In this work, dissipative effects from a phonon bath on the resonance fluorescence of a solid-state two level system embedded in a high quality semiconductor microcavity and driven by an intense laser, are investigated. Within the density operator formalism, we derive a variational master equation valid for broader ranges of temperatures, pumping rates, and radiation-matter couplings, than previous studies. From the obtained master equation, fluorescence spectra for various thermal and exciting conditions are numerically calculated, and compared to those computed from weak coupling and polaronic master equations, respectively. Our results evidence the breakdown of those rougher approaches under increased temperature and strong pumping. 8 1 0 2 v o N 7 ] l l a h - s e m . t a m - d n o c [ 3 v 2 3 5 5 0 . 1 1 6 1 : v i X r a 1 I. INTRODUCTION Solid-state emitters embedded in microcavities have become a new paradigm in cav- ity quantum electrodynamics (cavity-QED) [1 -- 3]. Recent developments in fabrication of semiconductor cavities serve a number of research fields, including quantum information processing, photonic circuits and quantum optics [4 -- 8]. Regarding the later, high quality cavities have been crucial for boosting the efficiency of single photon generators [9 -- 11]. For instance, some late experimental studies have focused on the resonant fluorescence of InGaAs quantum dots (QDs) grown inside of microcolumns, which have provided a clear demonstration of induced excitation [12 -- 14]. Thus, in systems with non-resonant laser- cavity coupling, the cavity mode is indirectly excited by the emission of photons from an artificial atom coupled to the acoustic phonon environment (phonon assisted cavity feeding) [15, 16]. The inverse effect of non-resonant coupling, where the quantum emitter is excited by photons emitted from the cavity, has also been observed [17]. D. McCutcheon et al. developed some years ago a variational master equation to describe the dynamics of a cavityless two-level system interacting with a boson environment, which was applied in the study of Rabi's rotations of a quantum dot [18]. They found that the tech- nique of variational master equation captures effects generally considered non-perturbative, such as multiphoton processes and renormalization of the Rabi frequency induced by the phonon bath. By comparing their population dynamics results with path integral numer- ical calculations, the reliability of the variational approach in accounting for those non- perturbative effects in regimes in which the weak and polaronic models was verified. Nevertheless, state of the art experiments use optical resonators embedding the emitter, because of the associated improvement in collection rates and photon purity [10, 19 -- 23]. Thus, our purpose is to investigate the fluorescence spectrum of a solid-state qubit-cavity system under pumping and thermal conditions beyond the scope of previously studied formu- lations like the weak coupling and polaronic approaches. To do that, we derive a variational master equation, which allows for numerical simulations of resonance fluorescence spectra within a wider range of excitation rates, emitter-cavity couplings, and temperatures. Such a master equation might also contribute to the promising research thread on double-dot-cavity systems, regarding phonon dissipation in tunnel-coupled emitters [24 -- 28]. Although this kind of systems have been addressed by means of numerical approaches, which adequately implemented, may render a solution as close as desired to the exact one (e.g. quasi-adiabatic propagator path-integral or real-time path integral techniques) [18, 29 -- 32]; those techniques are highly demanding from the computational point of view and do not yield the physical insight provided by a master equation. This paper is organized as follows: In the next section we present the model Hamiltonian and its modification under an adequate unitary transformation. In section III, the free energy of the system is minimized to determine the variational parameters and in section IV, the corresponding variational master equation is derived. Finally, in section V we obtain and discuss numerical simulations of fluorescence spectra of a semiconductor QD coupled to a cavity mode, and end by drawing overall conclusions in section VI. II. THEORY The system under study is a solid-state two level system (which we will refer to as "quantum dot" although it could either be a vacancy in a 3D crystal, a localized defect in a 2 FIG. 1: a) Schematics of a quantum emitter (pure radiative linewidth γ), embedded in a micropillar cavity (loss rate κ), and driven by a laterally applied CW laser (pumping rate ηx). b) Emitter energy levels (ground g(cid:105) and excited e(cid:105)), and its interactions with the phonon reservoir (λq) and the cavity (g). low dimensional structure, a nanocrystal, or any other suitable artificial atom), embedded in a QED cavity [33]. Carriers confined in the QD interact with a continuum of states in the sample of which it is part, via acoustic phonons. This interaction causes an incoherent pumping of the two level system. Moreover, because the artificial atom mainly interacts with a cavity single mode, the phonon environment produces some decoherence effects in the atom-cavity arrangement. The system is assumed driven by a continuous wave (CW) laser, as shown in the figure 1a), while the corresponding energy levels and interactions are depicted in figure 1b). Working in a rotating frame whose frequency matches that of the exciting laser ωL [34, 35], the considered Hamiltonian reads ( = 1) H = ∆XLσ+σ− + ∆CLa†a + σ+σ−(cid:88) + ηx(σ+ + σ−) + g(σ+a + a†σ−) b† ωq (cid:88) λq(bq + b† q) + bq , (1) q q q where ωq is the frequency of a phonon with momentum q, while bq (b† q) and λq are cor- respondingly the boson annihilation (creation) operator and intensity of the carrier-phonon coupling. The detuning respect to the pumping laser of the two level transition frequency (ωX) and that of the cavity mode (ωC), are respectively ∆CL and ∆XL. The annihilation (creation) operator of photons at the cavity frequency is a (a†), while the QD dipole opera- tors are σ− and σ+. g is the radiation-matter coupling constant, and the pumping rate ηx is the half of the Rabi frequency associated to the driving laser power. Let us consider a generalization of the polaron transformation that displaces the phonon bath oscillators, by an amount that is determined by a set of variational parameters {fq} [18]. Such a variational transformation can be written as H(cid:48) = e S He− S , (2) where 3 𝜂𝜂𝑥𝑔𝜅a)b)TwolevelsystemCavity...S = σ+σ−(cid:88) q νq(b† q − bq) , where νq = fq ωq The transformed Hamiltonian becomes H(cid:48) . S + H(cid:48) I + H(cid:48) B, with S = ∆Rσ+σ− + ∆CLa†a + (cid:104) B(cid:105)ζx , H(cid:48) H(cid:48) I = Bi , ζi H(cid:48) B = b† q bq , ωq (cid:88) (cid:88) i=x,y,z q (cid:80) (β = 1/kBT ) (cid:104) B(cid:105) = exp (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) where the modified detuning ∆R = ∆XL + R, depends on the variational shift R = q fq(fq − 2λq), and the thermal average of the bath displacement operator is given by q ω−1 (cid:34) −1 2 (cid:88) q f 2 q ω2 q (cid:35) coth(βωq/2) . In turn, the system modified operators ζi are explicitly ζx = ηx(σ+ + σ−) + g(σ+a + σ−a†) , ζy = iηx(σ+ − σ−) + ig(σ+a − σ−a†) , ζz = σ+σ− , and the phonon-induced fluctuation operators are defined as Bx = By = Bz = ( B+ + B− − 2(cid:104) B(cid:105)) , 1 2 ( B+ − B−) 1 2i (λq − fq)(b† (cid:88) q + bq) , in terms of the coherent displacement operators q B± = e±(cid:80) † q−bq) . q νq(b In the limit of continuous phonon modes, which is convenient and appropriate as long as the lattice parameter is much smaller than the typical size of the sample embedding the emitter, a spectral density J(ω) must be introduced, so that (cid:104)B(cid:105) and R correspondingly turn into 4 0 = H(cid:48) S + H(cid:48) where H(cid:48) On the one hand, (cid:104) H(cid:48) Au = − 1 β I(cid:105) H(cid:48) B and (cid:104) H(cid:48) I(cid:105) H(cid:48) 0 (cid:16) (cid:110) (cid:111)(cid:17) = Tr{ H(cid:48) Ie−β H(cid:48) 0}. vanishes because in the basis of eigenstates of H(cid:48) 0, all diagonal terms of H(cid:48) S] = 0 and each of those operators act on eigenstates of different subspaces (the dot-cavity and the phonon bath), then Au can be reduced to I are zero. On the other hand, since [ H(cid:48) B, H(cid:48) 0 ln Tr e−β H(cid:48) 0 + (cid:104) H(cid:48) I(cid:105) H(cid:48) 0 , (17) (cid:90) ∞ 0 R = (cid:104)B(cid:105) = exp dω (cid:20) (cid:90) ∞ ω −1 2 0 J(ω) F (ω)(F (ω) − 2), dω J(ω)2F (ω)2 ω2 (cid:21) coth(βω/2) . (15) (16) III. FREE ENERGY MINIMIZATION The variational parameters {fq} must be chosen in such a way that they minimize the free energy associated with the transformed Hamiltonian [36 -- 38]. To do that, we use the Feynman -- Bogoliubov inequality Au ≥ A, according which the free energy of the system (A), is at first order bounded by an upper limit given by (cid:16) (cid:110) (cid:111)(cid:17) Au = AB − 1 β ln e−β H(cid:48) S Tr , (18) with AB the free energy of the phonon bath. Inserting equation (4) into equation (17), the Feynman -- Bogoliubov upper bound reads Au = AB − 1 ×(cid:0)cosh(cid:2) 1 (cid:3) + cosh(cid:2) 1 β ln[2e− β 2βµ1 2 ((2n−1)∆CL+∆R) (cid:3)(cid:1)] , (19) (20) in terms of the phonon mean occupation number at temperature T (n = (cid:104)b†b(cid:105) = (cid:2)eβω − 1(cid:3)−1), and of the quantities 2βµ2 (cid:112)(cid:102)1 + 2(cid:102)2 , µ2 = (cid:112)(cid:102)1 − 2(cid:102)2 , (21) µ1 = that in turn depend on (cid:113) (cid:102)1 = ∆2 (cid:102)2 = R + 2B2(g2n + 2η2 CL + ∆2 (B2g2n − ∆CL∆R)2 + 4B2(B2g2n + ∆2 x) , CL)η2 x . Because the free energy of the phonon bath does not depend on fq, i.e. AB is unchanged by the interaction with the system, and then it is irrelevant in minimizing Au. By imposing ∂Au ∂fq = 0, we obtain 5 1 − λq  sinh(βµ2 /2) fq ≡ λq F (ωq ) = (cid:18) 1 − ∆R+Λ2 µ1 sinh(βµ1/2)+ ∆R−Λ2 cosh(βµ2/2)+cosh(βµ2/2) µ2 sinh(βµ2/2) sinh(βµ1 /2)+ ∆R−Λ2 (cid:32) cosh(βµ1/2)+cosh(βµ2/2) µ2 (ng2+2η2 µ1 x)+Λ1 ∆R+Λ2 (cid:19) µ1 + B2 ωq sinh(βµ1/2)+ (ng2+2η2 µ2 x )−Λ1 sinh(βµ2/2) cosh(βµ2 /2)+cosh(βµ2 /2) (cid:33) coth(cid:0)βωq /2(cid:1) , (22) (cid:102)2 x)−∆CL(g2nδR−2∆CLη2 x) and Λ2 = ∆CL(∆CL∆R−B2g2n) where Λ1 = B2g2n(g2n+4η2 In figure 2 the frequency dependence of the modulating part of the variational parame- ters for different pumping rates, radiation-matters couplings and temperatures is presented. There can be seen how for wave vectors q whose associated frequencies satisfy ηx/ωq (cid:28) 1, the minimization condition yields fq → λq, recovering the polaronic limit [38]. Only for these modes, the bath oscillators can fully follow the atom excitation. Otherwise, the mode frequencies are too slow and the corresponding oscillator shifts are dwindled, so that the carrier-phonon coupling at the corresponding momentum range is inhibited. (cid:102)2 . IV. MASTER EQUATION In this section, a variational master equation for the reduced density operator ρ(t), of the QD-cavity system, is derived within the second order Born-Markov framework [39]. The use of those approximations is justified because even at room temperature, the thermal energy would be much smaller than the typical transition energy of the two level emitter, and the thermalization processes are much faster than the relevant optical dynamics [40]. The validity of convolutionless non-perturbative approaches (regarding the phonon-carrier interaction) for studying strongly coupled dot-cavity systems, has been shown in references [35, 41]. In the case of strong pumping, minimization of the free energy is expected to grab relevant non-Markovian effects. We include the emitter radiative recombination and the cavity losses as Liouvillian decay superoperators, which act on the density matrix of the reduced system [42]. Such operators in the Lindblad form are given by (cid:0)2σ− ρσ+ − σ+σ− ρ − ρσ+σ−(cid:1) + κ(cid:0)2aρa† − a†aρ − ρa†a(cid:1) γ 2 L(ρ) = (23) where γ/2 is the HWHM radiative linewidth and κ is the cavity loss rate for the relevant mode. 6 FIG. 2: a) Variational parameter as function of the phonon frequency, at T = 30 K and g = 26.7 µeV for different pumping rates (Upper panel: from bottom to top, the curves correspond to smaller rates), b) at T = 30 K and ηx = 100 µeV for different coupling constants (Middle panel: from bottom to top, the curves correspond to smaller couplings), and c) at ηx = 500 µeV and g = 26.7 µeV for different temperatures (Lower panel: from bottom to top, the curves correspond to higher temperatures). Thus, inserting the transformed Hamiltonian from equations (4), (5) and (6), the varia- tional master equation takes the form 7 = − i[Hs, ρ(t)] + L(ρ) ∂ ρ ∂t (cid:90) t (cid:90) t 0 − + dτ dτ (cid:88) (cid:88) l=x,y,z m=x,y,z 0 l=x,y,z m=x,y,z Clm[ζm, e−iHsτ ζl eiHsτ ρ(t) C∗ lm[ρ(t)e−iHsτ ζl eiHsτ , ζm] , (24) where Clm(τ ) = (cid:104)Bl(τ )Bm(cid:105) for l, m = x, y, z. Assuming that the phonon bath is in thermal equilibrium [43], the correlation functions become Cyy(τ ) = (cid:104)B(cid:105)2 (cos φ(τ ) − 1) , Cxx(τ ) = (cid:104)B(cid:105)2 sin φ(τ ) , (cid:90) ∞ Czz(τ ) = dωJ(ω)[1 − F (ω)]2 (cid:90) ∞ 0 × (cos ωτ coth(βω/2) − i sin ωτ ) , F (ω)[1 − F (ω)] × (i cos ωτ + sin ωτ coth(βω/2)) , J(ω) dω ω 0 Czy(τ ) = (cid:104)B(cid:105) Cyz(τ ) = −(cid:104)Bz(τ )By(0)(cid:105) , and Cxz(τ ) = Czx(τ ) = Cxy(τ ) = Cyx(τ ) = 0 , (25) which depend on the spectral density and on the variational parameters. The first two correlations also depend on the function (cid:90) ∞ 0 φ(τ ) = dω ∂ ρ(t) ∂t = − i On the other hand, the master equation can be written in the Lindblad form J(ω) ω2 F (ω)2 (cos ωτ coth(βω/2) − i sin ωτ ) . (cid:16)(cid:104) (cid:17) (cid:105) + Dph(ρ) + L(ρ) + Lph(ρ) , H ef S , ρ(t) (26) (27) in terms of the effective Hamiltonian that describes the coherent part of the system evolution S = ∆xLσ+σ− + ∆cLa†a H ef W σ+σ− + (cid:104)B(cid:105)ζx + ∆σ11 + ∆σ+a + ∆a† σ− ph a†σ−σ+a + ∆σ− ph aσ+σ−a† + ∆σ+ ph σ−σ+ ph σ+σ−, 8 (28) and of the dissipative Lindbladian Lph(ρ) and the coherent variational shift Dph(ρ). The former is defined according to Lph(ρ) = + Γσ11 W 2 Γσ− ph 2 ph + L(σ11) + LIntp Γa† σ− L(σ−) + L(a†σ−) + Γσ+ ph 2 L(σ+) , ph 2 Γσ+a ph 2 L(σ+a) (29) where σ11 ≡ σ+σ− and L( D) = 2 D ρ D† − D† D ρ − ρ D† D. The term LIntp ph (ρ) describes the incoherent interpolation processes between the weak coupling approach [44], and the polaronic theory [41]. It explicitly reads LIntp ph (ρ) = Γσ11 σ+ zy 2 + + + Γσ11(σ+a) zy 2 Γσ+ σ11 yz 2 Γ(σ+a)σ11 yz 2 ph (σ11, σ+) + Γσ11 σ− LIntp zy ph (σ11, σ−) LIntp Γσ11(σ−a†) zy Γσ− σ11 2 ph (σ−, σ11) yz LIntp 2 Γ(σ−a†)σ11 yz LIntp ph (σ+, σ11) + LIntp ph (σ+a, σ11) + 2 ph (σ−a†.σ11) . LIntp LIntp ph (σ11, σ+a) + ph (σ11, σ−a†) LIntp (30) ph (A, B) = AB ρ(t) − ρ(t)B†A† − B ρ(t)A + A† ρ(t)B†. with LIntp As for the variational coherent shift (which is also originated from interpolation between the weak coupling and the polaronic models) [40], it is given by ph (σ11, σ+) + ∆σ11 σ− Dph(ρ) = ∆σ11 σ+ zy DIntp ph (σ11, σ+a) + ∆σ11 σ−a† + ∆σ11 σ+a DIntp ph (σ+, σ11) + ∆σ− σ11 + ∆σ+ σ11 yz DIntp yz DIntp ph (σ+a, σ11) + ∆σ−a† σ11 + ∆σ+aσ11 DIntp ph (σ11, σ−) zy DIntp ph (σ11, σ−a†) DIntp ph (σ−, σ11) ph (σ−a†, σ11) , DIntp zy zy yz yz (31) ph (A, B) = AB ρ(t) + ρ(t)B†A† − B ρ(t)A − A† ρ(t)B†. where DIntp By comparing equation (24) with (27), and dropping highly oscillatory terms, we ob- tained the phonon mediated transition probabilities and the variational shifts. The thermal dissipative rates are found to be of three types: Weak coupling-like rates [44, 45] W = 2(cid:60) Γσ11 polaronic-like rates [40] dτ Czz(τ ) , (32) (cid:21) (cid:20)(cid:90) ∞ 0 9 and interpolated rates Γσ11 σ± zy Γσ+a/a† σ− ph = 2g2(cid:60) Γσ+/σ− ph x(cid:60) = 2η2 = ∓2ηx(cid:61) Γσ11(σ+a/σ−a†) zy Γσ± σ11 yz = ∓2ηx(cid:61) 0 0 (cid:20)(cid:90) ∞ dτ(cid:104)B(cid:105)2e±∆cxτ(cid:0)eφ(τ ) − 1(cid:1)(cid:21) dτ(cid:104)B(cid:105)2e∓∆xLτ(cid:0)eφ(τ ) − 1(cid:1)(cid:21) (cid:20)(cid:90) ∞ (cid:20)(cid:90) ∞ (cid:20)(cid:90) ∞ dτ Czy(τ )e±∆CX τ dτ Czy(τ )e∓∆XLτ = ∓2g(cid:61) (cid:21) (cid:21) (cid:21) , , , 0 dτ Cyz(τ ) , 0 (cid:21) (cid:20)(cid:90) ∞ (cid:20)(cid:90) ∞ 0 0 Meanwhile, the energy shift components are identified as Γ(σ+a/σ−a†)σ11 yz = ∓2g(cid:61) dτ Cyz(τ ) . (cid:20)(cid:90) ∞ 0 = g2(cid:61) W = (cid:61) ∆σ11 ∆σ+a/a† σ− ph ∆σ+/σ− ph ∆σ11 σ± zy x(cid:61) = η2 = ±ηx(cid:60) ∆σ11 σ+a/σ−a† zy = ±g(cid:60) ∆σ± σ11 yz = ±ηx(cid:60) 0 dτ Czz(τ ) (cid:21) (cid:20)(cid:90) ∞ dτ(cid:104)B(cid:105)2e±∆cxτ(cid:0)eφ(τ ) − 1(cid:1)(cid:21) dτ(cid:104)B(cid:105)2e∓∆xLτ(cid:0)eφ(τ ) − 1(cid:1)(cid:21) (cid:20)(cid:90) ∞ (cid:20)(cid:90) ∞ (cid:21) (cid:20)(cid:90) ∞ (cid:20)(cid:90) ∞ (cid:20)(cid:90) ∞ dτ Czy(τ )e±∆CX τ dτ Czy(τ )e∓∆XLτ dτ Cyz(τ ) (cid:21) (cid:21) (cid:21) , , , , 0 0 0 0 = ±g(cid:60) dτ Cyz(τ ) . 0 ∆σ+a/σ−a† σ11 yz V. NUMERICAL RESULTS , , (33) (34) (35) (36) (37) (38) (39) (40) (41) (42) (43) (44) (45) As a representative study case, we will focus on the resonance fluorescence of a InAs/GaAs quantum dot coupled to a high quality optical resonator, under resonant continuous wave excitation [46, 47]. It is known that in most III-V semiconductor materials, the main source of dephasing is the carrier-acoustic phonon interaction via deformation potential [48, 49]. Thus, the spectral density Jph(ω) = αω3e−ω2/2ω2 b , is adopted for the simulations. α captures the strength of the exciton-phonon coupling and ωb provides a natural high-frequency cutoff, which is proportional to the inverse of the carrier localization length in the QD [38]. 10 To simulate the fluorescence spectrum, we compute (cid:20)(cid:90) ∞ (cid:20) Sc(ω) ∝ lim t→∞(cid:60) (cid:104)a(t + τ )a†(t)(cid:105) dτ 0 − (cid:104)a(t + τ )(cid:105)(cid:104)a†(t)(cid:105) ei(ωL−ω)τ (cid:21) (cid:21) , (46) where the correlation functions are obtained by the quantum regression formula [50]. To numerically solve the master equation within the different levels of approximation compared here (weak coupling, polaronic and variational), we employ a quantum optics toolbox devel- oped in MATLAB by Tan S. M. [51]. The pumping rate is assumed stable, i.e. ηx is taken independent of time, and the emitter is considered in the base state as initial condition [52]. To make our results comparable with Mollow triplet experiments on semiconductor mi- cropillars by S. M. Ulrich et al. [13], we consider a mode-cavity detuning ωc − ωx = −0.2 meV, and a radioactive decay rate γ = 3 µeV. Those values are also similar to the ones used in experiments by F. Hargart et al. in reference [14] and by H. Kim et al. in reference [53]. As for phonon parameters, typical values for InAs/GaAs QDs are used (cutoff frequency ωb = 0.9 meV y αp = 0.03 ps2) [14, 54, 55]. Figure 3 shows emission spectra from the cavity under various pumping rates and tem- peratures, obtained within the three considered master equation approaches. One can see how the weak coupling model differs greatly from the polar and variational theories as the system temperature increases, because of overestimation of the phonon dis- sipative effects. Concurrently, as long as the pumping rate remains moderate (e.g. ηx = 50 µeV), the polar and variational approaches predict similar behaviors. In this regime, the polaron model has been successfully fitted to resonance fluorescence measurements [56]. However, contrasts between those two later master equations are revealed when the pump- ing rate is strengthened. At median laser power (e.g. ηx = 250 µeV), the variational theory exhibits intermediate results between the weak and the polaronic models, which is particularly observable at the Mollow triplet side peaks. Under high excitation conditions (e.g. ηx = 500 µeV), the polaronic and variational approaches differ significantly in the predicted renormalization of the Rabi frequency and the emission intensity of all the peaks, specially the right one in the triplet, evidencing how in this regime the polaronic approach also misjudges by excess the phonon associated decoherence. Such a breakdown of the polaronic approach for high pumping rates becomes larger as the temperature increases. Surprisingly for strong pumping, as compared with the variational results, predictions from the weak coupling model differ less than those from the polaronic model. In order to check consistency of our results with real-time path integral calculations, we compare the Rabi frequency renormalization in the bottom-right panel of figure 3, to those reported in figure 3b) of reference [57] and figure 5 of reference [58]. There, a renormalization of ∼ 10% is reported for bare Rabi frequencies at the order of 1 meV, in agreement with our simulations from the variational model, while such a renormalization obtained within the polaron approach reaches ∼ 35%, elucidating overvaluation of the thermal effects. It is worth mentioning that in despite of discrepancies regarding its magnitude, all three models account for the phonon assisted cavity feeding phenomenon. 11 FIG. 3: Cavity-emitted fluorescence spectra of a semiconductor QD-cavity system driven via on-resonance exciton pumping (ωL = ωx, with ∆cx = 2 meV) for various values of the exciton pump ηx and phonon-bath temperature T . Black line: spectra obtained from a weak coupling master equation, blue line: from a polaronic master equation, and red: from the variational master equation developed in this work. In all plot panels, the frequency is taken respect to the QD emission and g = 26.7 µeV is used. VI. SUMMARY AND CONCLUSIONS In this work, we derived an optimized master equation for a quantum photon emitter simultaneously coupled to a phonon bath and to an optical resonator, inspired on the po- laronic transformation but with phonon displacements variationally determined by a mode- dependent approach. Thus, a theory flexible enough to encompass the weak and polaronic coupling methods, but applicable on a larger range of experimental conditions, was obtained. We applied the developed theory in the simulation of the resonance fluorescence emission from a single quantum dot embedded into a high quality microcavity, for different tempera- ture and excitation values. Such spectra were also calculated within the weak coupling and conventional polaronic theories, so that pertinent comparison could be carried out among the three considered models. The numerical results showed that in comparison to the more rigorous variational ap- proach, the weak coupling and polaronic theories, correspondingly overestimate the phonon 12 dissipative effects as the temperature and the excitation power increase. In conclusion, the variational master equation obtained here, is expected to provide a valuable tool to simulate and explain experiments on solid-state emitters interacting with phonon reservoirs and QED cavities, carried out under light-matter coupling, pumping rate and temperature values, lying in a much wider range than those spanned by previously available master equation approaches. This is of significance given the increasing excita- tion intensities and emitter-cavity mode couplings achieved in state of art quantum optical experiments. ACKNOWLEDGMENTS The authors acknowledge the Research Division of Universidad Pedag´ogica y Tecnol´ogica de Colombia for financial support. ∗ [email protected] [1] Murch, K. Cavity quantum electrodynamics: Beyond strong. Nat. Phys. 13, 11 (2017). [2] Yoshihara, F., Fuse, T., Ashhab, S., Kakuyanagi, K., Saito, S., & Semba, K. Superconducting qubit-oscillator circuit beyond the ultrastrong-coupling regime. Nat. Phys. 13, 44 (2017). [3] Forn-D´ıaz, P., Garc´ıa-Ripoll, J. J., Peropadre, B., Orgiazzi, J. L., Yurtalan, M. A., Belyansky, R., & Lupascu, A. Ultrastrong coupling of a single artificial atom to an electromagnetic continuum in the nonperturbative regime. Nat. Phys. 13, 39 (2017). [4] Chiorescu, I., Bertet, P., Semba, K., Nakamura, Y., Harmans, C. J. P. M., & Mooij, J. E. Coherent dynamics of a flux qubit coupled to a harmonic oscillator. Nature 431, 159 (2004). [5] Heo, J., Kang, M. S., Hong, C. H., Choi, S. G., & Hong, J. P. Scheme for swapping two unknown states of a photonic qubit and an electron-spin qubit using simultaneous quantum transmission and teleportation via quantum dots inside single-sided optical cavities. Phys. Lett. A 381, 1845 (2017). [6] Coles, R. J., Price, D. M., Royall, B., Clarke, E., Skolnick, M. S., Fox, A. M., & Makhonin, M. N. Path-dependent initialization of a single quantum dot exciton spin in a nanophotonic waveguide, Phys. Rev. B 95, 121401(R) (2017). [7] Stockklauser, A., Scarlino, P., Koski, J. V., Gasparinetti, S., Andersen, C. K., Reichl, C., Wegscheider, W., Ihn, T., Ensslin, K. & Wallraff, A. Strong Coupling Cavity QED with Gate- Defined Double Quantum Dots Enabled by a High Impedance Resonator. Phys. Rev. X 7, 011030 (2017). [8] Dufferwiel, S., Lyons, T. P., Solnyshkov, D. D., Trichet, A. A. P., Withers, F., Schwarz, S., Malpuech, G., Smith, J. M., Novoselov, K. S., Skolnick, M. S., Krizhanovskii, D. N., & Tartakovskii, A. I. Valley-addressable polaritons in atomically thin semiconductors, Nat. Photonics 11, 497 (2017). [9] Song, H. Z., Hadi, M., Zheng, Y., Shen, B., Zhang, L., Ren, Z., Gao, R., & Wang, Z. M. InGaAsP/InP Nanocavity for Single-Photon Source at 1.55-µm Telecommunication Band. Nanoscale Res. Lett. 12, 128 (2017). [10] Ding, X., He, Y., Duan, Z. C., Gregersen, N., Chen, M. C., Unsleber, S., Maier, S., Schneider, C., Kamp, M., Hofling, S., Lu, C. Y., & Pan, J. W. On-demand single photons with high 13 extraction efficiency and near-unity indistinguishability from a resonantly driven quantum dot in a micropillar. Phys. Rev. Lett. 116, 020401 (2016). [11] Unsleber, S., He, Y. M., Gerhardt, S., Maier, S., Lu, C. Y., Pan, J. W., Gregersen, N., Kamp, M., Schneider, C. & Hofling, S. Highly indistinguishable on-demand resonance fluorescence photons from a deterministic quantum dot micropillar device with 74% extraction efficiency. Optics Exp. 24, 8539 (2016). [12] Wang, H., Duan, Z. C., Li, Y. H., Chen, S., Li, J. P., He, Y. M., Chen, M. C., He, Y., Ding, X., Peng, C. Z., Schneider, C., Kamp, M., Hofling, S., Lu, C. Y., & Pan, J. W. Near-transform- limited single photons from an efficient solid-state quantum emitter, Phys. Rev. Lett. 116, 213601 (2016). [13] Ulrich, S. M., Ates, S., Reitzenstein, S., Loffler, A., Forchel, A., & Michler, P. Dephasing of triplet-sideband optical emission of a resonantly driven InAs/GaAs quantum dot inside a microcavity. Phys. Rev. Lett. 106, 247402 (2011). [14] Hargart, F., Roy-Choudhury, K., John, T., Portalupi, S. L., Schneider, C., Hofling, S., Kamp, M., Hughes, S., & Michler, P. Probing different regimes of strong field light -- matter interaction with semiconductor quantum dots and few cavity photons. New J. of Physics 18, 123031 (2016). [15] Ates, S., Ulrich, S. M., Reitzenstein, S., Loffler, A., Forchel, A., & Michler, P. Post-selected indistinguishable photons from the resonance fluorescence of a single quantum dot in a micro- cavity. Phys. Rev. Lett. 103, 167402 (2009). [16] Calic, M., Gallo, P., Felici, M., Atlasov, K. A., Dwir, B., Rudra, A., Biasiol, G., Sorba, L., Tarel, G., Savona, V., & Kapon, E. Phonon-mediated coupling of InGaAs/GaAs quantum-dot excitons to photonic crystal cavities. Phys. Rev. Lett. 106, 227402 (2011). [17] Majumdar, A., Kim, E. D., Gong, Y., Bajcsy, M., & Vuckovi´c, J. Phonon mediated off- resonant quantum dot -- cavity coupling under resonant excitation of the quantum dot. Phys. Rev. B 84, 085309 (2011). [18] McCutcheon, D. P., Dattani, N. S., Gauger, E. M., Lovett, B. W., & Nazir, A. A general approach to quantum dynamics using a variational master equation: Application to phonon- damped Rabi rotations in quantum dots. Phys. Rev. B 84, 081305 (2011). [19] Muller, K. Fischer, K. A. Rundquist, A. Dory, C. Lagoudakis, K. G. Sarmiento, T. Kelaita, Y. A. Borish, V. and Vuckovi´c, J. Ultrafast Polariton-Phonon Dynamics of Strongly Coupled Quantum Dot-Nanocavity Systems. Phys. Rev. X 5, 031006 (2015). [20] Fischer, K. A. Muller, K. Rundquist, A. Sarmiento, T. Piggott, A. Y. Kelaita, Y. Dory, C. Lagoudakis K. G. and Vuckovi´c, J. Self-homodyne measurement of a dynamic Mollow triplet in the solid state. Nat. Photonics 10, 163 (2017). [21] Seyfferle, S., Hargart, F., Jetter, M., Hu, E.,& Michler, P. Signatures of single-photon in- teraction between two quantum dots located in different cavities of a weakly coupled double microdisk structure. Phys. Rev. B 97, 035302 (2018). [22] Wang, H. Li, W. Jiang, X. He, Y. M. Li, Y. H. Ding, X. Chen, M. C. Qin, J. Peng, C. Z. Schneider, C. Kamp, M. Zhang, W. J. Li, H. You, L. X. Wang, Z. Dowling, J. P. Hofling, S. Lu, C. Y. and Pan, J. W. Toward Scalable Boson Sampling with Photon Loss. Phys. Rev. Lett. 120, 230502 (2018). [23] Hanschke, L. Fischer, K. A. Appel, S. Lukin, D. Wierzbowski, J. Sun, S. Trivedi, R. Vuckovi´c, J. Finley, J. J. and Muller, K. Quantum dot single-photon sources with ultra-low multi-photon probability. npj Quantum Information 4, 43 (2018). 14 [24] Liu, Y. Y. Petersson, K. D. Stehlik, J. Taylor, J. M. and Petta, J. R. Photon Emission from a Cavity-Coupled Double Quantum Dot. Phys. Rev. Lett. 113, 036801 (2014). [25] Liu, Y. Y. Stehlik, J. Eichler, C. Gullans, M. J. Taylor, J. M. and Petta, J. R. Semiconductor double quantum dot micromaser. Science 347, 285 (2015). [26] Rojas-Arias, J. S. Rodr´ıguez, B. A. and Vinck-Posada, H. Magnetic control of dipolaritons in quantum dots. J. Phys. Cond. Matt. 28, 505302 (2016). [27] Cotrino-Lemus, J. and Ram´ırez, H. Y. Efficiently tunable photon emission from an optically driven artificial molecule. J. Phys. Conf. Ser. 864, 012081 (2017). [28] Dom´ınguez, M. S. Macias-Pinilla, D. F. and Ram´ırez, H. Y. Dipolariton formation in quantum dot molecules strongly coupled to optical resonators. arXiv:1609.05525v2 (2018). [29] Sim, E. Quantum dynamics for a system coupled to slow baths: On-the-fly filtered propagator method. J. Chem. Phys. 115, 4450 (2001). [30] Vagov, A. Croitoru, M. D. Axt, V. M. Kuhn, T. and Peeters, F. M. Nonmonotonic Field Dependence of Damping and Reappearance of Rabi Oscillations in Quantum Dots. Phys. Rev. Lett. 98, 227403 (2007). [31] Vagov, A. Croitoru, M. D. Glassl, M. Axt, V. M. and Kuhn, T. Real-time path integrals for quantum dots: Quantum dissipative dynamics with superohmic environment coupling. Phys. Rev. B 83, 094303 (2011). [32] Nahri, D. G. Mathkoor, F. H. and Raymond Ooi C. H. Real-time path-integral approach for dissipative quantum dot-cavity quantum electrodynamics: impure dephasing-induced effects. J. Phys. Cond. Matt. 29, 055701 (2017). [33] Kreinberg, S., Chow, W. W., Wolters, J., Schneider, C., Gies, C., Jahnke, F., Hofling, S., Kamp, M., & Reitzenstein, S. Emission from quantum-dot high-β microcavities: transition from spontaneous emission to lasing and the effects of superradiant emitter coupling Running title: Quantum-dot high-β microcavities. Light Sci. Appl. 6, e17030 (2017). [34] Suter, D. The physics of laser-atom interactions (Vol. 19). Cambridge University Press, (1997). [35] Roy, C., & Hughes, S. Polaron master equation theory of the quantum-dot Mollow triplet in a semiconductor cavity-QED system. Phys. Rev. B 85, 115309 (2012). [36] Silbey, R., & Harris, R. A. (1984). Variational calculation of the dynamics of a two level system interacting with a bath. J. Chem. Phys. 80, 2615 (1984). [37] Kuzemsky, A. L. Variational principle of Bogoliubov and generalized mean fields in many- particle interacting systems. Intern. J. Mod. Phys. B, 29, 1530010 (2015). [38] Nazir, A., & McCutcheon, D. P. Modelling exciton -- phonon interactions in optically driven quantum dots. J. Phys. Condens. Mat. 28, 103002 (2016). [39] Schlosshauer, M. A. Decoherence: and the quantum-to-classical transition. Springer Science & Business Media, (2007). [40] Roy, C., & Hughes, S. Influence of Electron -- Acoustic-Phonon Scattering on Intensity Power Broadening in a Coherently Driven Quantum-Dot -- Cavity System. Phys. Rev. X 1, 021009 (2011). [41] Roy, C., & Hughes, S. Phonon-dressed Mollow triplet in the regime of cavity quantum elec- trodynamics: excitation-induced dephasing and nonperturbative cavity feeding effects. Phys. Rev. Lett. 106, 247403 (2011). [42] Carmichael, H. An open systems approach to quantum optics: lectures presented at the Universit´e Libre de Bruxelles, October 28 to November 4, 1991 (Vol. 18). Springer Science & Business Media, (2009). 15 [43] Glauber, R. J. Coherent and incoherent states of the radiation field. Phys. Rev. 131, 2766 (1963). [44] Harsij, Z., Harouni, M. B., Roknizadeh, R., & Naderi, M. H. Influence of electron-phonon interaction on the optical spectrum and quantum statistics in a quantum-dot -- cavity system: Master-equation approach. Phys. Rev. A 86, 063803 (2012). [45] Harouni, M. B., Roknizadeh, R., & Naderi, M. H. Influence of phonons on exciton-photon interaction and photon statistics of a quantum dot. Phys. Rev. B 79, 165304 (2009). [46] Ram´ırez, H. Y. Double dressing and manipulation of the photonic density of states in nanos- tructured qubits. RSC Adv. 3, 24991 (2013). [47] He, Y., He, Y. M., Liu, J., Wei, Y. J., Ram´ırez, H. Y., Atature, M., Schneider, C., Kamp, M., Hofling, S., Lu, C. Y., & Pan, J. W. Dynamically Controlled Resonance Fluorescence Spectra from a Doubly Dressed Single InGaAs Quantum Dot. Phys. Rev. Lett. 114, 097402 (2015). [48] Krummheuer, B., Axt, V. M., & Kuhn, T. Theory of pure dephasing and the resulting ab- sorption line shape in semiconductor quantum dots. Phys. Rev. B, 65, 195313 (2002). [49] Gauger, E. M., Nazir, A., Benjamin, S. C., Stace, T. M. & Lovett B. W. Robust adiabatic approach to optical spin entangling in coupled quantum dots. New J. Phys. 10, 073016 (2008). [50] Carmichael, H. J., & Scully, M. O. Statistical Methods in Quantum Optics 1: Master Equa- tions and Fokker-Planck Equations. Physics Today 53, 78 (2000). [51] Tan, S. M. A computational toolbox for quantum and atomic optics. J. Opt. B 1, 424 (1999). [52] Perea, J. I., Porras, D., & Tejedor, C. Dynamics of the excitations of a quantum dot in a microcavity. Phys. Rev. B 70, 115304 (2004). [53] Kim, H. C. Shen, T. C. Roy-Choudhury, K. Solomon, G. S. and Waks, E. Resonant Interactions between a Mollow Triplet Sideband and a Strongly Coupled Cavity. Phys. Rev. Lett. 113, 027403 (2014). [54] Ramsay, A. J., Gopal, A. V., Gauger, E. M., Nazir, A., Lovett, B. W., Fox, A. M., & Skol- nick, M. S. Damping of exciton Rabi rotations by acoustic phonons in optically excited In- GaAs/GaAs quantum dots. Phys. Rev. Lett. 104, 017402 (2010). [55] Ramsay, A. J., Godden, T. M., Boyle, S. J., Gauger, E. M., Nazir, A., Lovett, B. W., ... & Skolnick, M. S. Phonon-induced Rabi-frequency renormalization of optically driven single InGaAs/GaAs quantum dots. Phys. Rev. Lett. 105, 177402 (2010). [56] Wei, Y. J. He, Y. He, Y. M. Lu, C. Y. Pan, Y. W. Schneider, C. Kamp, M. Hofling, S. McCutcheon, D. P. and Nazir, A. Temperature-Dependent Mollow Triplet Spectra from a Single Quantum Dot: Rabi Frequency Renormalization and Sideband Linewidth Insensitivity. Phys. Rev. Lett. 113, 097401 (2014). [57] Glassl, M. Sorgel, L. Vagov, A. Croitoru, M. D. Kuhn, T. and Axt, V. M. Interaction of a quantum-dot cavity system with acoustic phonons: Stronger light-matter coupling can reduce the visibility of strong coupling effects. Phys. Rev. B 86, 035319 (2012). [58] Vagov, A. Glassl, M. Croitoru, M. D. Axt, V. M. and Kuhn, T. Competition between pure dephasing and photon losses in the dynamics of a dot-cavity system. Phys. Rev. B 90, 075309 (2014). 16
1109.6591	1	1109	2011-09-29T17:02:56	Carbon sp chains in graphene nanoholes	[ "cond-mat.mes-hall" ]	Nowadays sp carbon chains terminated by graphene or graphitic-like carbon are synthesized routinely in several nanotech labs. We propose an ab-initio study of such carbon-only materials, by computing their structure and stability, as well as their electronic, vibrational and magnetic properties. We adopt a fair compromise of microscopic realism with a certain level of idealization in the model configurations, and predict a number of properties susceptible to comparison with experiment.	cond-mat.mes-hall	cond-mat	Carbon sp chains in graphene nanoholes Ivano E. Castelli1,2, Nicola Ferri2, Giovanni Onida2, and Nicola Manini2 1 Center for Atomic-scale Materials Design, Department of Physics, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark 2 ETSF and Dipartimento di Fisica, Universit`a di Milano, Via Celoria 16, 20133 Milano, Italy Abstract. Nowadays sp carbon chains terminated by graphene or graphitic-like carbon are synthesized routinely in several nanotech labs. We propose an ab-initio study of such carbon-only materials, by computing their structure and stability, as well as their electronic, vibrational and magnetic properties. We adopt a fair compromise of microscopic realism with a certain level of idealization in the model configurations, and predict a number of properties susceptible to comparison with experiment. PACS numbers: 81.07.-b, 31.15.A-, 75.70.Ak, 65.80.Ck 1. Introduction The carbon atom, with its three possible hybridization states, originates in nature very different elemental materials. The three possibilities (sp3, sp2, sp) correspond to three different prototypical structures: respectively diamond, graphitic-like structures (such as graphite, graphene, carbon nanotubes, and fullerenes), and linear carbon chains (known in the literature as polyynes [1, 2] or sp carbon chains, spCCs in short). spCCs were discovered in nature around 1968 [3], but their role in the arena of carbon-based nanostructures has been quite marginal: till very recently, spCCs have been considered as exotic allotropic forms, mainly present in extraterrestrial environments. Indeed, in the interstellar clouds formed in the explosions of carbon stars, novae, and supernovae, spCCs have been detected aside with fullerenes [4, 5], amorphous carbon dust, cyanopolyynes, and oligopolyynes. Indeed, in the phase diagram of carbon, the field of existence of spCCs coincides with that of fullerenes [6]. The high reactivity of spCCs [7], and their tendency to undergo cross-linking to form sp2 structures, directed the experimental efforts towards complicate strategies for the stabilization of the spCCs with molecular end-groups or their isolation in inert matrices [1, 8, 9]. Thanks to novel synthetic routes and strategies [10], polyynes of increasing length and different type of termination have been successfully synthesized and characterized [11, 12, 13, 14, 15]. End-capped spCCs (often, but not exclusively, hydrogen-terminated) are being synthesized by chemical / electrochemical / Carbon sp chains in graphene nanoholes 2 photochemical methods [1, 16, 17, 18, 19]. spCCs have also been produced from carbon by dynamic pressure [17, 20]. spCCs or carbynoid material containing up to 300 carbon atoms were synthesized, which opens a promising route toward molecular engineering of sp-carbon structures [16, 17, 21, 22]. Samples of pure carbon films grown by supersonic cluster beam deposition at room temperature have been characterized, and proven to contain a sizeable sp component [23, 24]. These spCCs showed weaker stability relative to the sp2 component upon exposition to low pressure gases at room temperature [25]. spCCs have also been produced from stretched nanotubes [26]. Recently, Jin et al. [27] have realized spCCs by stretching and thinning a graphene nanoribbon from its two free edges, by removing carbon rows until the number of rows becomes one or two. These spCCs show a good stability under the beam of a transmission electron microscope (TEM) for lengths up to a few nanometers. These experimental observations of spCCs formation during the controlled electron irradiation of graphene planes resulted in an rapidly increasing interest in this field [28, 29, 30, 31, 32, 33, 34, 35, 36]. Meanwhile, the existence of intrinsic magnetism in pure carbon has been a matter of debate for quite some time [37]. Possible effects due to magnetic contaminants on the experimental results have been discussed [38]. Subsequently, it has been shown that possible contamination effects are unable to explain quantitatively the measured ferromagnetism, supporting the idea that carbon magnetism has an intrinsic origin [39]. The existence of π-electron magnetism in pure carbon has now been widely accepted (see e.g. [40, 41]). On the theoretical side, it is well known that magnetic instabilities exist at specific graphene edges [42, 43, 44, 45], in defective graphene [46] and nanotubes [47]. The discovery of pure-carbon π-electron magnetism has also lead to speculations about possible applications of carbon-based magnetic materials in molecular electronic devices: for example, spintronic devices built around the phenomenon of spin- polarization localized at the 1-D zig-zag edges of graphene have been proposed [48]. Recently, the accent has been also put on the possibility to modify the magnetization optically [49]. linear carbon chains are nowadays considered promising structures for nano-electronic applications [50, 51]. For example, they can be used as molecular bridges across graphene nanogap devices. Potential applications for the realization of non-volatile memories and two-terminal atomic-scale switches have been demonstrated [51]. Indeed, The remarkable robustness of spCCs terminated on pure carbon, graphene-like fragments, combined with the fast progresses in the synthesis of graphene and graphene derivatives, could open the way towards the realization of actual nanodevices based on sp + sp2 carbon nanostructures. The interest of such systems stems from the possibility to exploit their peculiar semiconducting-magnetic behavior, in order to achieve novel characteristics and functions for the target devices. Clearly, the possibility of designing graphene-based magnetic nanostructures is particularly intriguing. The capability of arranging the spins inside a carbon structure Carbon sp chains in graphene nanoholes 3 in a variety of ways, could open the way for the construction of completely novel devices [52]. Possible future applications for spCCs in interaction with the graphene-type system could be the construction of microchips with ferromagnetic or antiferromagnetic character that can be controlled by nanomanipulation and read out by nanocurrents. In Sect. 2 we introduce the theoretical model, whose electronic structure we address by a standard ab-initio method based on the Density-Functional Theory (DFT). Section 3 presents the investigation of the structural and binding properties of a spCC inside a nanohole (nh) in a graphene sheet representing the sp2 component in a carbon- only sample. Section 4 studies the magnetic properties of the nh edges and of the In Sections 5 and 6 we cover the band structures and selected inserted spCCs. vibrational properties of the studied nanostructures. In Sect. 7 we investigate the dynamical stability of the metastable spCC-nh structures, by means of a tight-binding molecular dynamics model. Section 8 discusses the results of simulations in the light of experimental data. 2. The model In the present paper, we focus on spCCs bound to graphitic structures, represented by a hole in a infinite graphene sheet. This system is representative of a class of sp + sp2 systems, which are at the core of an intense experimental work [23, 24, 25, 27, 28, 53, 54, 55]. To address the structural, vibrational, and electronic properties of spCCs inserted into a nanometer-sized hole defect of a graphene layer, we resort to the DFT. The plane- wave pseudopotential method and the Local Spin-Density Approximation (LSDA) to DFT have provided a simple framework whose accuracy have been demonstrated in a variety of systems [56]. The time-honored LDA is one in many functionals being used for current DFT studies of molecular and solid-state systems: other functionals often improve one or another of the systematic defects of LDA (underestimation of the energy gap, small overbinding and tiny overestimation of the vibrational frequencies), but to date no functional is universally accepted to provide systematically better accuracy than LDA for all properties of arbitrary systems. For a covalent system of s and p electrons as the one studied here, LDA is appropriate, and we expect our results to change by a few percent at most if the calculations were repeated using some other popular functional [57, 58, 59]. We compute the total adiabatic energies by means of the code Quantum Espresso [60], which computes forces by standard Hellmann-Feynman method. Each self- consistent electronic-structure calculation stops when the total energy changes by less than 10−8 Ry. We use ultrasoft pseudopotentials [61, 62], for which a moderate cutoff for the wave function/charge density of 30/240 Ry is sufficient. We terminate atomic relaxation when all residual force components are smaller than 10−4 Ry/a0 (cid:39) 0.04 µN. Plane waves require periodic boundary conditions in all space directions. In our model, we represent a graphene plane in a (x− y)-periodically repeated 1714× 1690 pm2 supercell consisting of 7 × 4 rectangular conventional unit cells containing four carbon Carbon sp chains in graphene nanoholes 4 Figure 1. (Color online) The adopted slab model supercell for graphene, composed by 7 × 4 conventional rectangular cells containing 4-atoms, each equivalent to two √ primitive unit cells of graphene (in gray). Solid: the unit vectors of the conventional 3 (cid:39) 422 pm. The size of the full supercell cell, of length a = 244 pm and b = a (delimited by the solid lines) in the x − y plane is 1714 pm × 1690 pm. atoms each, see Fig. 1. Within this cell, a complete graphene layer is then represented by 112 carbon atoms. We represent the graphite surface in the slab approximation, as one or a few stacked graphene layers: to ensure that the interaction between periodic images of the graphene sheet is negligible we interpose at least 1 nm of vacuum along the z direction. We have selected this cell size as it allows us to create reasonably large nanoholes in the graphene layers, with fairly small interaction between the supercell repeated copies of the nanohole itself, at the price of a manageable computer time. We cover electronic band dispersion the electron bands in the horizontal plane by means of a 5 × 5 × 1 Γ-centered k-point mesh. BZ integration of the metallic band energies are performed using a 2 × 10−4 Ry-wide Gaussian smearing of the fermionic occupations. 2.1. The Nanohole Starting from the perfect graphene foil of Fig. 1, we remove selected atoms in order to form a nh. The size of the nh should be such that inserted spCCs fit and only bind at their ends. If the nh is too small, then spCC atoms would tend to reconstruct the edges of the hole, as illustrated in the examples of Figs. 2 and 3. There, the small spCC-edge distance leads to spontaneous (barrier-free) edge reconstruction with the formation of Carbon sp chains in graphene nanoholes 5 (a) Initial (b) Relaxed Figure 2. (Color online) Initial and relaxed positions of a small nh with a C6 polyyne: the lateral nh size is so small that no barrier prevents two polyyne atoms (brown/clear) to bind to the armchair edges. (a) Initial (b) Relaxed Figure 3. (Color online) Initial and relaxed positions of a small nh with a C6 polyyne: the lateral nh size is so small that no barrier prevents two polyyne atoms (brown/clear) to bind to the zig-zag edges and form pentagonal rings. Carbon sp chains in graphene nanoholes 6 Figure 4. (Color online) A nh of sufficient size to lodge spCCs is obtained by removing 28 atoms from the perfect graphene of Fig. 1. The rectangular hole has size 975 pm × 985 pm, and is delimited by both zig-zag and armchair edges additional squares or pentagons. We must therefore construct a large enough nh, in particular allowing for a distance of at least 300 pm between the spCC and the nearest nh edge in order to prevent recombination reactions. Figure 4 shows the minimal nh with such property. Starting from the perfect graphene of Fig. 1, the nh is obtained by removing 28 atoms forming a rectangular hole of size 975 pm × 985 pm with both zig-zag armchair edges. The size of the nh permits us to investigate the insertion of several spCCs, from C5 to C8 atoms long, with various positions relative to the nh edges. 3. spCCs Binding to a Nanohole In an experiment [23, 24, 25, 53], spCCs are likely to be bound to extended sp2 structures, more reminiscent of graphite than of graphene. Such a configuration could be described for example with one layer of perfect graphene plus a layer with a nh, such as that described above Sect. 2.1. Figure 5 shows this configuration with the insertion of a C6 polyyne. The upper graphene layer and the polyyne are fully relaxed, while the lower layer is kept frozen in ideal graphitic positions. The equilibrium distance between the two layers, 332 pm, is very close to the observed interlayer distance of graphite, 335 pm. The resulting system composed of over 200 atoms is computationally quite expensive: indeed, with a modern parallel computer using 32 Xeon-class processors, it took more than two weeks to obtain a fully relaxed configuration. On the other hand, we verified that the properties of the upper layer do not change significantly without the lower (perfect) sheet, whose effective corrugation is quite small. Indeed, the forces between the two layers are weak long-range forces whose action is mild and Carbon sp chains in graphene nanoholes 7 Figure 5. (Color online) A nano-indented graphite surface represented by a perfect graphene foil (pale gray) plus a superposed sheet with a nh (dark blue) as in Fig. 4 stacked in the standard AB arrangement of graphite. A C6 chain (also dark blue) is connected to the zig-zag edge of the nh. The layer-layer distance equals 332 pm. almost translationally invariant, compared to the intra-layer forces acting in the nh layer. The DFT-LSDA evaluation of such weak dispersive forces is unreliable anyway. These observations suggest that it makes sense to consider the single layer containing the nh and the Cn spCC inserted into it, and leave the fixed substrate layer out. The relaxation of the positions, performed in the same conditions as above, took about one week only in this 1-layer configuration involving 90 C atoms rather than 200 ones. In our calculations, we consider several spCCs, from C5 to C8, placed in different positions inside the nh. We identify such compounds as nh-Cn, with further specification when different local minima are considered. Their relaxed configurations are depicted in Fig. 6. Table 1 summarizes the structural properties of the configurations considered, comparing in particular their bond-length alternation (BLA) [63] and the spCC- graphene bonding energy, defined as the total energy of the empty nh plus that of the isolated spCC minus the total energy of the bonded spCC-nh configuration under consideration. For the selected nh size, Cn chains of different length can fit more or less easily inside the nh. Short chains such as C5 or C6 may fit at the price of a tensile stress, while longer spCCs can be forced inside the nh with a compressive strain, which could be eased by buckling [34, 64, 65]. Strain influences directly the spCC BLA: a tensile strain leads to stretching more the weaker bonds, thus producing an enhanced BLA, typical of polyynic spCC (in nh-C6 arm, whose BLA reaches 12 pm). Likewise, a BLA(cid:39) 10 pm is obtained for nh-C6 zig due to the nh size being approximately 5% larger than the equilibrium length of the spCC. In contrast, a compressive strain leads to a more cumulenic-type structure, e.g. Carbon sp chains in graphene nanoholes 8 (a) nh-C5 (b) nh-C5 1b (c) nh-C6 zig (d) nh-C6 arm (e) nh-C7 straight, top view (f) nh-C8 (g) nh-C7 curved, side view (h) nh-C7 s-curved, side view (i) nh-C7 straight, side view (j) nh-C8 side view 0.0 0.5 1.0 1.5 nm Figure 6. (Color online) The relaxed configurations of the spCC-nanohole structures considered in the present paper and described in Table 1. Carbon sp chains in graphene nanoholes 9 BLA Ebond σp σv σh [pm] [eV] 4.2 (cid:88) (cid:88) (cid:88) Label [Fig.] nh-C5 [Fig. 6(a)] nh-C5 1b [Fig. 6(b)] nh-C6 zig [Fig. 6(c)] nh-C6 arm [Fig. 6(d)] nh-C7 curved [Fig. 6(g)] nh-C7 s-curved [Fig. 6(h)] nh-C7 straight [Figs. 6(e), 6(i)] nh-C8 [Figs. 6(f), 6(j)] wnh-2C6 [Fig. 7] Description A C5 chain stretched across the nh bonding weakly to both zig-zag edges (bond lengths: 170 pm). The C5 chain bonded to one zig-zag edge of the nh; all bond lengths of the spCC are similar to the lengths of cumulenic double bonds ((cid:39) 127 pm). A weakly stretched C6 chain bonded to the zig-zag nh edge. The C6 chain joining opposite armchair nh edges. A C7 connected to the zig-zag edges, and buckling out of the graphene plane (maximum spCC height (cid:39) 120 pm). A C7 connected to the zig-zag edges, buckling in a s shape, with the central atom in the same plane as graphene. A compressed straight C7 joining the zig-zag edges. A C8 curved chain. The maximum height of the spCC equals 297 pm. Two C6 spCCs inserted in a wider nh joining the zig-zag edges. The lateral distance between the spCCs equals 491 pm. 7 1 10 12 3 3 2 6 6.2 (cid:88) (cid:88) 12.9 (cid:88) (cid:88) (cid:88) 8.2 (cid:88) (cid:88) 12.0 (cid:88) (cid:88) 11.9 * (cid:88) * 11.8 (cid:88) (cid:88) (cid:88) 12.5 (cid:88) (cid:88) 11 n.c. (cid:88) (cid:88) (cid:88) Table 1. (Color online) Summary of the individual configurations considered for the spCCs bound to the nh. For the relaxed configurations we report the resulting BLA and total bonding energy Ebond, corresponding to the formation of the (usually two) bonds between the spCC and the nh. The relevant symmetry planes (σp is the reflection across the graphene layer plane, σv is the vertical plane through the spCC, σh is the horizontal plane through the middle of the spCC) are marked for the configurations for which they apply. ∗Individual σp and σh are not symmetries for nh-C7 s-curved, but their product σpσh is. Carbon sp chains in graphene nanoholes 10 Figure 7. (Color online) Relaxed position of two C6 spCCs in the wider nh: the hole width is 1504 pm, large enough to prevent the recombination of the two spCCs. This configuration displays no significant novelty relative to the single polyyne nh-C6 zig configuration. nh-C7 straight has BLA (cid:39) 2 pm. As was observed in a slightly different context [34], very small BLA variations are induced by lateral atomic displacements. The linear size of the hole is approximately 95% of the equilibrium length of the C7 chain: different stable shapes of the spCC in nh-C7 can be stabilized by the compressive strain [64]. We study three equilibrium geometries of the spCC: straight, single-curvature buckling, and s-curved buckling. The compressive strain depresses the BLA, so that for the three of them the BLA ranges from 2 to 3 pm. The nh-C5 is so much stretched that if kept in a central symmetric configuration bonding between the spCC and the two edges of the nh is weak, each highly stretched terminal bond contributing only about (cid:39) −2 eV to lowering the total energy. In such a condition we observe an intermediate BLA (cid:39) 7 pm. This configuration is locally stable, but if we displace the spCC significantly ((cid:39) 50 pm) closer to one nh edge than to the other, and then let it relax, we retrieve an energetically favored configuration (nh-C5 1b) with essentially a single strong bond (total energy lowering: (cid:39) −6 eV) between the spCC and the nh. Here the spCC internal bond lengths are practically equal to those of isolated C5. Due to the small size of the nh, the C8 chains can only fit in a curved geometry: the maximum out-of-plane elevation of the spCC equals 297 pm. The resulting BLA (cid:39) 6 pm is intermediate between cumulenic and polyynic. We also consider a wider nh in which one can insert more than one spCC: in wnh- 2C6, the nanohole contains two C6 spCCs at a distance large enough to keep them separated, see Fig. 7. All properties are essentially equivalent to those of the nh-C6 zig, therefore we will not further investigate this configuration. We evaluate the bonding energy of the configurations described here. Due to Carbon sp chains in graphene nanoholes 11 Figure 8. (Color online) Scheme of the geometric relation between adjacent graphene edges. The angle between edges is defined as the angle between the vectors normal to the edge. The edge atoms belong to the same sublattice (either dark/red or clear/gray) when the zig-zag edges are at a relative angle of 0◦ or 120◦; they instead belong to different sublattice when the relative angle is 60◦ or 180◦. (Adapted from Ref. [67].) its stretching, the nh-C5 has a little value of Ebond (cid:39) 4 eV, while for the nh-C5 1b Ebond = 6.2 eV which can be considered a fair estimate of the spCC-graphene edge binding energy according to DFT-LSDA, and matches previous evaluations [25]. For all other configurations Ebond (cid:39) 12 eV indicative of the formation of two bonds, at the expense of approximately 1 eV which accounts for the elastic deformation energy of the spCC and the connected graphene. Until now the spCC was always connected to zig-zag edges. When a C6 chain binds to the armchair edges (nh-C6 arm), the bonding energy is smaller (Ebond (cid:39) 9 eV), due to the lower reactivity of the armchair edge relative to the zig-zag one [25, 66]. The BLA assumes a highly dimerized value 12 pm associated to a tensile strain, like for the nh-C6 zig isomer. In the following we shall investigate the electronic properties of selected configurations. In particular, we first focus on the interplay of the magnetic behavior of the nh zig zag edges and of the spCC. We will then move on to describe the DFT-LSDA band structure, the vibrational properties, and the high-temperature stability of nh-Cn configurations. 4. Magnetism Zig-zag edges are generally known to be ferrimagnetic [44, 68] due to non-bonding localized edge states. A detailed investigation of the magnetic properties of graphene Carbon sp chains in graphene nanoholes 12 (a) nh-C7 (b) nh-C8 Figure 9. state of nh-C7 and of nh-C8. (Color online) The 0.01 µB/a3 0 magnetic isosurface of the ferromagnetic edge in the context of a nanohole was carried out by Yu et al. [67]. That work focused on zig-zag edges (armchair ones are known to be nonmagnetic [69]), which made it convenient to study diamond- or hexagon-shaped holes with zig-zag edges only. The main conclusion of Ref. [67] regarding consecutive zig-zag edges is that the relative alignment of magnetic moments tends to be ferromagnetic when the edge atoms belong to the same graphene sublattice. This conclusion can be rephrased in terms of the angle between the two consecutive edges, which is defined as the angle between the in-plane outward vectors normal to the edges, as illustrated schematically Fig. 8. Ferromagnetic correlations occur when subsequent zig-zag edges are unrotated (0◦) or rotated by 120◦, as would happen in a triangular hole. In the opposite case, the magnetization is antiferromagnetic, as occurs for zig-zag edges rotated by 60◦ or 180◦ (relevant, e.g. for a diamond or and hexagonal hole). Our rectangular nh involves two armchair edges, which are long enough to isolate rather effectively the magnetic moments localized at the two zig-zag edges. In Appendix A, we study the magnetic properties of the edge of this nh. Following spCC insertion, all structures described in Sect. 3 preserve a nonzero absolute magnetization, associated to unpaired-spin electrons localized at the zig-zag nh edges. A significant magnetization is shared by the Cn spCCs with odd n, while the even-n spCCs are non-magnetic, as illustrated by Fig. 9 for the nh-C7 and nh-C8 structures. The ferromagnetic structures of Fig. 9 are induced by the choice of a uniform starting magnetization used to initialize the electronic self-consistent calculation. To investigate other possible magnetic arrangements, we need to start off the self-consistent calculation with different magnetic arrangements of the individual atoms. To do this, we define two fictitiously different atomic species, both with the same chemical nature of C, but with initial magnetizations of opposite sign (±1 Bohr magneton). We place Carbon sp chains in graphene nanoholes 13 Figure 10. (Color online) To investigate nonferromagnetic structures we consider two fictitious atomic carbon species fixing the spin polarization at the initial stage of the self-consistent calculation. C↑ (dark/red) and C↓ (clear/gray) carry positive and negative initial magnetization, respectively. Arcs mark all symmetry-independent nearest-neighbor Ising-type magnetic couplings Jij, see Eq. (1). these initially magnetically polarized atoms along the zig-zag edges in order to trigger the desired magnetic structure. Figure 10 illustrates one of many possible arrangements of the C↑ (dark/red) and C↓ (clear/gray) atoms to initiate the self-consistent electronic- structure calculation. 4.1. C6-nh We perform several self-consistent calculations for the nh-C6 structure, considering different starting magnetizations, as shown in Fig. 11, and determine the ground magnetic configuration. In agreement with Ref. [67], the ground-state configuration, Fig. 11(a), has atoms of the the same magnetization in the same graphene sublattice (e.g. atoms labeled B, C, H, E in Fig. 10), and magnetization changes sign in passing from one sublattice to the other. The edge atoms bonded to even-n Cn spCCs show little magnetism, mainly induced by the ferromagnetic interaction with neighboring atoms along the same zig-zag edge. This ground-state magnetic configuration is relaxed completely, and the resulting total energy Egs is taken as reference. Keeping fixed this fully relaxed ground atomic configuration, we repeat single self-consistent DFT-LSDA evaluations of the total energy, Etot, integrated magnetization Mtot = (cid:82) Mz((cid:126)r) d3(cid:126)r, and integrated absolute value of magnetization Mabs =(cid:82) Mz((cid:126)r) d3(cid:126)r, which we report next to each structure in Fig. 11. The resulting individual magnetic configurations, violating the opposite-sublattice rule, are low-lying excitations, which we obtain in the DFT-LSDA simulations by changing appropriately the initial magnetizations of selected atoms. Relaxation of one of these Carbon sp chains in graphene nanoholes 14 (a) Ground state: Etot = Egs = −13948.038 eV Mtot = 0.00 Mabs = 9.31 (b) Etot = Egs + 17 meV Mtot = 0.00 Mabs = 8.21 (c) Etot = Egs + 26 meV Mtot = 0.00 Mabs = 8.03 (d) Etot = Egs + 43 meV Mtot = 0.00 Mabs = 8.38 (e) Etot = Egs + 542 meV Mtot = 0.00 Mabs = 8.67 (f) Etot = Egs + 543 meV Mtot = 0.00 Mabs = 9.19 (g) Etot = Egs + 547 meV Mtot = 8.00 Mabs = 8.66 (Color online) Magnetization-density isosurfaces at +0.01 µB/a3 Figure 11. 0 (dark/red) and −0.01 µB/a3 for several different nh-C6 magnetic structures. Frame (a): the DFT-LSDA ground state. Other frames: magnetically excited states. For each frame we report the relevant excitation energy, total magnetization and integrated absolute magnetization in Bohr magnetons µB. 0 (clear/gray), Carbon sp chains in graphene nanoholes 15 Ising Parameter Value [meV] Standard deviation [meV] E0 − Egs J1 J2 J3 281 −259 −4 −17 5 5 10 10 Hspin = − (cid:88) Table 2. (Color online) The interaction parameters of the Ising Model, Eq. (1), fitted on the DFT-LSDA values of the total energy of the magnetic configurations of nh-C6 zig listed in Fig. 11. configurations shows very small displacements, not larger than 6 pm. The excitation energies of such states can be described approximately within a Ising-model scheme. The z component Si of the spin degree of freedom accounting for the magnetization at site i interacts with neighboring spins Sj, with an energy JijSiSj . (1) <i,j> According to the values of the absolute magnetization reported in Fig. 11(a), it is appropriate to assume that each edge atom carries one Bohr magneton, i.e. one unpaired spin 1/2, thus Si = ±1/2. Accordingly, it makes sense to fit Ising-model energies only to configurations with an absolute magnetization significantly close to 8 µB. To avoid parameter proliferation, we neglect interactions between non-nearest-neighbor magnetic atoms. Figure 10 identifies the 3 independent Ising interaction parameters Jk allowed by symmetry: J1 for the interactions between unpaired spins in different sublattices on edges rotated by 60◦ (J1 = JAB = JCD = JEF = JGH); J2 for the interactions within the same zig-zag edge, but "isolated" by the spCC (J2 = JBC = JF G); and J3 for the interactions across the armchair edge, representing opposite sublattices, or edges rotated by 180◦ (J3 = JAH = JDE). We write the energy of a configuration as the sum of the magnetic energy Espin, approximated by the Ising expression (1), plus E0, including covalency and all other interactions establishing the mean value of the total energy, averaged over all possible spin orientations. Specifically, for the nh-C6 zig structure, we have: Espin = − J1 (SASB + SCSD + SESF + SGSH) − J2 (SBSC + SF SG) − J3 (SASH + SDSE) , (2) so that, given the ground configuration of Fig. 11(a), we have Egs = E0 + Espin = E0 + J1 + (J3 − J2)/2. We estimate the Ising-model parameters by means of a linear fit of the DFT energies with expression (2). Table 2 reports the best-fit values of E0 and Jk: all the exchange energies Jk turn out negative, reflecting antiferromagnetic interactions. The most significant value is J1, reflecting the strong antiferromagnetic coupling of adjacent unpaired spins on edge atoms belonging to opposite sublattices. J1 is over Carbon sp chains in graphene nanoholes 16 Figure 12. (Color online) Comparison of the DFT-LSDA energy levels of the magnetic structures of Fig. 11 (left) with the spectrum (right) obtained using the Ising model, based on parameteres fitting the DFT values. one order of magnitude greater than the weakly antiferromagnetic coupling J3 across an armchair edge section. Given the fit standard deviation, the small value of J2 is compatible with null coupling. The obtained small negative value is the result of a strong cancellation between the energy-order reversed configurations of panels 11(c), 11(d), and the regularly ordered states of panels 11(a), 11(b), the latter matching the ordering Ref. [67] as expected. Indeed the Ising-model ground state and first-excited level turn out almost degenerate and actually in reversed order due to the small positive value of J2, as illustrated in Fig. 12. In Fig. 12, we compare the energies of the different configurations of Fig. 11 obtained by DFT calculation with those obtained using the fitted Ising model. A remarkable feature of the DFT excitation spectrum is the tiny splitting of the levels of panels 11(e)-11(g), which is hardly compatible with a simple nearest-neighbor Ising model. Eventually Fig. 12 shows that the simple Ising model fails to describe the fine structure of the magnetic excitation of the edge atoms in the considered geometry. Only the significant J1 energy, fixing the rough structure of the spectrum, is determined with fair accuracy. Of course one could easily modify the model to include e.g. second-neighbor interactions, to fit the detailed level structure, but that would take any predictive power out of the model. 4.2. C7-nh One may attempt a similar analysis for the odd-n spCCs, e.g. nh-C7. As Fig. 9(a) shows, odd-n spCCs are magnetic, thus quite different from the even-n ones. This leads to two consequences for odd spCCs attached to nh: first the spin values at different sites are Carbon sp chains in graphene nanoholes 17 (a) Ground State Etot = Egs = −14102.247 eV Mtot = −0.80 Mabs = 8.97 (b) Etot = Egs + 60 meV Mtot = −1.20 Mabs = 9.63 Figure 13. (Color online) Magnetization isosurfaces at +0.01 µB/a3 −0.01 µB/a3 0 (clear gray) for two magnetic structures of nh-C7 straight. 0 (dark/red) and different, and second the number of spin interactions to be considered is greater. This would leave little significance to a Ising model description. It is possible to at least identify the ground magnetic configuration, like we did for even-n chains. Figure 13 shows two different magnetic configuration of the nh- C7 straight structure. The ground-state configuration is the one of Fig. 13(a), which follows the edge rules of Ref. [67]. The coupling between the spCC and nh edges is antiferromagnetic: this can be seen as a special case of the edge rules if the spCC atoms are seen as graphene atoms belonging to the edge but in the other sublattice relative to the outer zig-zag edge magnetic atoms. This coupling is so strong that it prevails over the weak antiferro long-range J3-type coupling. One can estimate this magnetic coupling energy between the end spCC atom and one of the nearest zig-zag edge atoms to approximately ≈ 100 meV. The intra-spCC interaction is distinctly antiferromagnetic. 5. Electronic Properties The magnetic properties described in the previous section are determined by the electronic structure. Before analyzing the nanohole-spCC system, it is useful to examine the simpler bands of a empty nh. We will track the the bands along the Brillouin- zone path shown in Fig. 14(a). We sample the k-space path with points separated by 1.75 × 10−13 m−1. Figure 14(b) displays the band structure near the Fermi energy for the empty nh superlattice of Fig. 4. One can identify two kinds of bands, with different spatial localization properties of their wave functions: (i) States localized at the hole edge (HE), such as the one depicted in Figs. 15(a). We use magenta crosses to track these Carbon sp chains in graphene nanoholes 18 (a) k-point path (b) nh bands (Color online) (a): The Brillouin-zone (dashed) with the Γ − X − M − Figure 14. Γ − Y − M k-point path adopted for all band-structure calculations of the present paper. (b) Spin-majority Kohn-Sham band structure of the relaxed nh superlattice of Fig. 4, in the ferromagnetic configuration of Fig. 2(f): magenta crosses stand for HE bands localized on the hole-edge; green squares represent delocalized BU states. The plot focuses a 4 eV-wide energy region around the Fermi level (red dashed) for better readability. (a) A HE state (b) A BU state Figure 15. localized HE state; (b): a BU state. Γ-point crossing at −0.81 eV and 1.36 eV, respectively. (Color online) Examples of k = 0 electronic states of the nh. (a): a In Fig. 14(b), these states are located at the Carbon sp chains in graphene nanoholes 19 HE states in the bands-structure plots, such as Fig. 14(b). (ii) States like the one in Fig. 15(b) localized primarily on the bulk graphene atoms, with a weak component on the edge atoms. We label these states as BU, and identify them with green squares in band-structure plots. HE bands are generally flat, with little dispersion. The small but nonzero bandwidth of HE states is due to the residual interaction between the nh and its periodic images. A HE band touches the Fermi energy near X, and is therefore only partly filled, thus becoming the responsible of the edge magnetism discussed in Sect. 4. Liu et al. in their investigation of the band structures of a different graphene nanohole [70], discovered the opening of band gaps for nanoholes with either armchair or zig-zag edges. In contrast, our graphene with a isolated nh exhibits no band gap at the Fermi energy, and retains the (semi)metallic character of graphene. Specifically, a BU metallic band crosses F and shows a modest but distinct dispersion, with sizeable empty hole pockets near the X and Y points. Coming to the nh-Cn systems, Figs. 16, 17, 18, and 19 report details of the computed DFT-LSDA bands for the relaxed structures of Figs. 6 and 7. Also in these band structures we identify HE (magenta crosses) and BU (green squares) bands. In addition, band states significantly localized on the spCC atoms (CB) are identified by blue circles. One such state is depicted in Fig. 20(a). Occasional resonances of localized spCC and nh-edge states lead to hybrid localized states involving both, e.g. the one of Fig. 20(b). We identify such "CHE" states only in nh-C6 arm, and label them by black triangles in Fig. 18(a). Graphene bulk states often hybridize with the spCC molecular orbitals, thus acquiring a significant spCC components, as illustrated for example in Fig. 20(c). Our sample is too small to distinguish clearly between entirely localized states at the spCC/nh edge (whose bands would be perfectly flat in a realistically wide sample) and only partly localized hybrid states. Essentially all structures display a metallic behavior, due to one or several bulk bands crossing the Fermi energy. The different bonded spCCs affect the graphene nh bands quite considerably, by both shifting them and deforming them especially near the Fermi energy. In particular, the positions of several localized states at the nh edge change depending on the spCC state, and moreover spCC-specific localized states occur. For even-n spCCs, the CB states are energetically quite distant from the Fermi level, while odd-n spCCs exhibit a more metallic behavior, with spCC states quite close to the Fermi energy, and significant hybridization with the extended bulk states, consistently with results of Ref. [25, 71]. The comparison of the spin majority and minority bands in Fig. 17 shows that magnetism affects the bulk bands only weakly. Magnetism appears to be associated to an energy shift of a few localized HE and (for odd spCCs) CB states near the Fermi level. The resulting effective exchange energy is (cid:39) 0.3 eV. We perform several calculations of the nh-C7 structure for each of the considered spCC shapes: curved -- Fig. 6(g), s-curved -- Fig. 6(h), and straight -- Fig. 6(i). All these geometries show basically identical band structures, e.g. the one reported in Fig. 18(b). Carbon sp chains in graphene nanoholes 20 (a) nh-C5 (b) nh-C5 1b Figure 16. in Fig. 6(a), and of (b) nh-C5 1b, Fig. 6(b). (Color online) Spin-majority band structures of (a) nh-C5, represented Likewise, no special effect of the spCC curvature is apparent in the bands of nh-C8, Fig. 19(a). Finally, the congestion of the bands near the Fermi level in Fig. 19(b) is a consequence of the larger cell, and greater number of atoms and of electrons of this specific configuration. The general considerations (even-n spCC bands away from the Fermi energy, magnetism related to HE bands near the Fermi energy) apply also in this more intricate configuration. Carbon sp chains in graphene nanoholes 21 (a) nh-C6 zig majority spin (b) nh-C6 zig minority spin Figure 17. (Color online) Band structure for the majority (a) and minority (b) spin components of the ferromagnetic state of nh-C6 zig, depicted in Fig. 6(c). 6. Vibrational spectra We perform phonon calculation for a few stable structures of Sect. 3. We evaluate the phonon frequencies and eigenvectors of using standard density-functional perturbation theory, as implemented in the Quantum Espresso code [60, 72]. For comparison, the theoretical C-C stretching modes of polyynes CnH2 (n = 8 − 12) evaluated with the same method match the experimental frequencies [73] to within 40 cm−1. The size of the system is too large to evaluate the full dynamical matrix: although in principle possible, it would require a huge investment of computer time. We focus specifically on spCC "optical" stretching modes, which are prominent Carbon sp chains in graphene nanoholes 22 (a) nh-C6 arm (b) nh-C7 Figure 18. nanostructure, depicted in Fig. 6(d), and (b) nh-C7, Fig. 6(e). (Color online) Spin-majority band structures of (a) the nh-C6 arm and characteristic in the experimental spectra of sp − sp2 carbon in the spectral region near 2000 cm−1, while all other vibrations (the "acoustic" spCC stretching modes, all bending modes, all graphene vibrations) overlap and lump together in a continuum extending from 0 to 1600 cm−1 [23, 25]. We verified that the spCC stretching modes are influenced very little by faraway ligand atoms [74]. Accordingly, we only compute and diagonalize the part of the dynamical matrix at Γ relative to displacements of the atoms of the spCC, plus its first and second neighbors in the graphene sheet. The error in the vibrational frequency induced by this approximation can be estimated (cid:39) 1 cm−1. By analyzing the displacement pattern of the normal modes of the even-n spCCs, it is straightforward to identify the "α" modes characterized by the strongest Raman Carbon sp chains in graphene nanoholes 23 (a) nh-C8 (b) nh-2C6 Figure 19. (Color online) Spin-majority band structures of nh-C8, see Fig. 6(f), and of nh-2C6, see Fig. 7. Structure name Raman frequencies [cm−1] IR frequencies [cm−1] nh-C5 nh-C6 zig 1323, 1368 1777, 1878 1332 1939 Table 3. Wavenumber of Raman and IR spCC stretching frequencies calculated for the nh-Cn structures. The most intense Raman and IR frequencies are highlighted in bold. Carbon sp chains in graphene nanoholes 24 (a) A CB state (b) A CHE state (c) A resonant bulk state Figure 20. (Color online) Like in Fig. 15, but for nh-C6 zig and nh-C6 arm. (a): a typical CB state localized mainly on the polyyne -- the Γ state at energy −1.16 eV in Fig. 17(a); (b): a CHE state, localized jointly on the polyyne and the nh edge -- the Γ state at energy −0.50 eV in Fig. 18(a); (c): a bulk state showing a significant extension on the polyyne -- the Γ state at energy −0.46 eV in Fig. 17(a). and IR absorption [25, 54, 55, 75]. We take advantage of this pattern recognition to avoid a computationally expensive explicit evaluation of the Raman and IR intensities. Table 3 reports the computed frequencies of the spCC stretching modes, with the most intense Raman and IR mode highlighted. Note that the wavenumbers of the frequencies are significantly lower than the characteristic stretching-frequencies of free spCCs (1950 − 2300 cm−1). The reason is the tensile strain to which the spCCs are subjected by binding to the nh edges. In nh-C5, the length of the chain, including the bonds between the spCC and the nanohole, is 15% longer than the isolated spCC; this elongation leads to frequencies much softer than typical polyyne ones. The chain length in nh-C6 zig is only 5% longer than isolated length, and the frequencies come much closer to the typical frequencies of free spCCs. The results of the present section do not imply that spCCs in a context of nanostructured sp-sp2 carbon material should vibrate at much different frequencies from their molecular counterparts [54]. Quite on the contrary, previous calculations and experiments confirm that fully relaxed spCCs terminated by sp2-type material exhibit very similar frequencies to those of molecular spCCs [25, 54, 55]. The results of the present calculations suggest instead that unrelaxed tensile strain in nanostructured sp - sp2 carbon material is likely to induce significant frequency shifts of the spCC modes. Depending on the method of production of spCC-containing material (e.g. cluster beam formation/deposition [23] vs. atomic wires stretched out from pulled graphene sheets [27, 28]), whenever a sizeable tensile strains remain frozen in the sample, one is to observe a corresponding distribution of the observed vibrational frequencies, quite independent of the frequency shifts associated to different lengths and terminations of the spCCs [55]. Carbon sp chains in graphene nanoholes 25 7. High-temperature stability i.e. All studied configurations are local minima of the adiabatic potential energy, metastable allotropes of carbon. Given sufficiently long time, the spCCs are expected to degrade, for example by recombining with the nh edge and extend energetically favored sp2 graphene. This possibility is however very remote at low temperature, because in this interconversion process the energy barriers to be crossed are quite substantial. Since several detailed types of sp → sp2 processes are possible, a full study of the spCC degradation is beyond the scope of the present paper. We content ourselves with a semi-quantitative estimate of the thermal stability and the degradation mechanisms of spCCs bonded to carbon sp2 nanostructures by running comparably long high- temperature molecular-dynamics (MD) simulations, and monitor the eventuality of spCC decomposition as a function of the simulated temperature. As, due to the size dependency of statistical fluctuations, the longer the spCCs the higher is the chance of chain breaking. We therefore prefer to simulate a larger version of the model of Sect. 3, namely a C10 chain bound to a nh large enough for it to fit loosely. Also the graphene plane is represented by four, rather than three hexagonal rings separating the nh periodic replicas. The resulting nh-C10 structure involves 184 atoms and 736 electrons. Even using Car-Parrinello dynamics, it would be a formidable task to simulate several sufficiently long runs to expect a significant chance to observe spCC dissociation at a temperature comparable to experiment. For the present task therefore we abandon the ab-initio DFT-LSDA method for the treatment of the electronic degrees of freedom, and replace it with a tight-binding (TB) model [76]. We adopt the TB scheme of Xu et al. [77], which has been applied successfully to investigate several low- dimensional carbon systems [78, 79, 80, 81], as it reproduces well the experimental bulk equilibrium distance dgraph = 1.4224 A of graphite, and the bulk structure and elastic properties of sp3 diamond. Since all interatomic interactions vanish at a cutoff distance rc = 2.6 A, which is shorter than the interlayer spacing of graphite, 3.35 A [82], we focus on a single-layer model like we did for the DFT-LDA model above. To validate the TB force field for spCCs, we compute the total energy of a C14 linear chain as a function of the length d7−8 of its central bond (between atoms 7 and 8), which we keep fixed while all other bonds are allowed to relax. Figure 21 compares this total energy, referred to its value at full relaxation, as obtained with DFT-LSDA and with TB. The two models exhibit significant differences, in particular the TB model has a level crossing to a dissociative regime above 1.9 A, while nothing of the sort occurs in the DFT-LDA band-structure calculation, which takes care of level degeneracies by selecting a spin-1 magnetic state. This problem with the TB model is characteristic of unsaturated conditions, while, in close-shell electronic configurations, dissociation is more regular. In general, the TB model is likely to be more "fragile" than a more realistic DFT-LDA. In turn the latter is also expected to be less strongly bounded than real spCCs, due to missing long-range attractive Van-der-Waals polarization correlation effects. We must then conclude that all quantitative stability evaluations based on the Carbon sp chains in graphene nanoholes 26 Figure 21. (Color online) Comparison of the total energy variation for a C14 linear chain as a function of the (constrained) length d7−8 of its central bond (between atoms 7 and 8), as obtained with DFT-LSDA and with TB. TB model are underestimates of the actual stability in experiment. Periodic boundary conditions (matching the zero-temperature lattice parameter of graphene) suppress long wavelength fluctuations, resulting in a stabilization of the thermal fluctuations, which in the thermodynamical limit would make the 1D - 2D structure unstable. We run microcanonical (constant-energy) TB molecular dynamics (TBMD) simulations, since the temperature fluctuations are small enough ((cid:39) 10%) in a sample of this size for temperature to be considered a fairly well defined quantity. The advantage of the microcanonical ensemble is that no thermostat artifacts, and in particular no dissipative term as in Langevin or Nos´e-Hoover thermostats, can affect the atomic dynamics, allowing for a full account of local fluctuations to produce whatever bond breaking they may lead to. A disadvantage of the constant-energy MD is that, if a significant bonding breakdown of a part of the nanostructure occurs, the corresponding potential-energy increase occurs at the expense of the kinetic energy, thus the system may artificially cool down, thus hindering further decomposition. In practice, this problem has little importance for the system size considered. We use a time step of 0.5 fs, small enough to guarantee a rigorous global energy conservation within 0.02 eV, or 0.001%. the model nh-C10 We run extended simulations of structure at different temperatures, starting with a sampling of initial conditions. We generate starting configurations by beginning with the fully relaxed configuration and running three successive equilibration runs (0.1 ps, 0.1 ps and 0.5 ps), with randomized initial velocities, taken from a Gaussian distribution matching the Boltzmann distribution at the target temperature. In the figures we indicate the resulting temperature, with an error bar appropriate to the determination of the average along the whole simulation, (thus not estimating the instantaneous temperature fluctuations, which are much larger). In the determination of this average temperature, we drop the first 100 fs, Carbon sp chains in graphene nanoholes 27 (a) t = 0 ps (b) t = 1.5 ps (c) t = 2.5 ps (d) t = 2.7 ps Figure 22. temperature T = (3991 ± 10) K of nh-C10. (Color online) Successive snapshots of a sample TBMD simulation at a Figure 23. Eq. (3), of the C10 chain in four simulations at different temperature. (Color online) The time dependence of the longest bond length dmax, to minimize the systematic oscillations induced by starting with initial random velocities with little correlation to the forces. Figure 22 displays successive frames of an example simulation illustrating that edge reconstruction processes -- Fig. 22(b) -- often occur before the earliest spCC breakdown event -- Fig. 22(d). To monitor these breakdown events, a clear indicator is the longest C-C bond length relative to the 11 bonds of the C10 chain and of the chain ends to the attached nh edge atoms: dmax = max i=0,1,..10 dCi−Ci+1 . (3) Figure 23 reports the time dependency of dmax following 4 independent simulations carried out at different temperature, with different initial states. Chain breakdown, as happens between frames 22(c) and 22(d), is signaled by the rapid increase of one of the bond lengths beyond 5 A. The spCC breakdown may be followed by recombination of the chain into the nh edge, or even expulsion of a section of the spCC into vacuum. Carbon sp chains in graphene nanoholes 28 Figure 24. (Color online) Circles: the average inverse time before the breakdown of the C10 spCC, as a function of inverse temperature. Dashed line: a Arrhenius fit, Eq. (4). As suggested by Fig. 23, spCC breakdown occurs, on average, earlier and earlier for increasing temperature. By repeating the numerical simulations for different initial conditions but similar temperature we estimate an average decay rate by averaging the inverse times before decay. The average decomposition rate time ¯k is an increasing function of temperature. In simulations done at substantially lower temperature than the ones considered in Fig. 23, one would need to wait too long to observe decomposition, while at much higher temperature decomposition occurs immediately after start. If one can assume that one type of process (bond breaking) dominates over all decomposition channels, the decomposition rate ¯k is expected to be in the Arrhenius form . (4) (cid:18) (cid:19) ¯k = A exp − Ea kBT We estimate the attempt frequency A and the effective energy barrier Ea of the TBMD model by fitting the Arrhenius plot in Fig. 24. The value Ea = 4.2 eV matches the TB breakup of Fig. 21: this effective activation barrier is surely an underestimation of the actual barrier against breakup. The estimated attempt rate A = 1.5 × 1017 s−1, although probably slightly overestimated, reflects the high number of breakup channels available for decomposition of the spCC [83]. If we assume the computed values for Ea and A, we extrapolate thermal decay rates of spCCs in sp2 carbon of the order of ¯k = 7.5 × 10−5 s−1 at 1000 K, and ¯k = 1.5 × 10−54 s−1 at room-temperature (300 K). Overall, the calculations of the present section confirm a substantial stability of spCCs in the solid state and in vacuum. In the lab, whenever sp − sp2 carbon is not kept in vacuum, chemical decay mechanisms are therefore likely to overcome the thermal ones. Carbon sp chains in graphene nanoholes 29 8. Discussion and conclusion The present work collects extensive investigation of the geometry, electronic structure, magnetic properties, and dynamical stability of spCCs attached to sp2 graphitic fragments. When a spCC binds to zig-zag graphene edges, its polyynic character is attenuated to a value intermediate between those typical of cumulenes and polyynes. The attachment of a spCC to the graphene edge is very stable: we predict stabilization energies near 6 eV per bond between each chain end and the sp2 regions. Thermal excitations typically break bonds along the spCC with similar probability to those formed with the graphene edge, which indicates a very solid attachment. Odd spCCs in the nh display a metallic behavior, with at least one band pinned to the Fermi energy, while even spCCs have little overlap with the states at the Fermi level [50, 51, 84, 85]. The partly filled states of odd-n spCCs are associated to nonzero magnetization related to a spin triplet state of the π bonds. Even-n spCCs are instead insulating and non-magnetic, and in this context only the magnetic moments of the graphene edge contribute to the magnetism of the nh-C2m structures, and only the bulk graphene states provide conducting bands. We compute also the vibrational modes, and specifically the optical C≡C stretching modes which emerge as a characteristic signature of spCCs in Raman and IR spectroscopies. Our calculations show that the vibrational frequencies can be quite substantially red-shifted when spCCs are kept under tensile stress. Indeed, the weaker stability and correspondingly faster decomposition rate of strained spCCs is likely to play a significant role in the overall blue shift of the sp-carbon peak in the Raman spectrum of the decaying spCCs in cluster-assembled sp-sp2 film [25]. Acknowledgments We are grateful to L. Ravagnan, P. Milani, E. Cinquanta, and Z. Zanolli for invaluable discussions. The research leading to these results has received funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement No. 211956 (ETSF-i3). We acknowledge generous supercomputing support from CILEA. Appendix A. Magnetism at the edge of the empty nh As magnetism is intrinsic of the zig-zag graphene edge, thus even a empty nanohole exhibits a range of magnetic states similar to those arising in the presence of even-n spCCs. The zig-zag edge atoms involved in magnetism are clearly identified in Fig. A1. Also here, each atomic site A -- L carries a magnetization close to 1 µB. As illustrated in Fig. A1, even though the number of spin-carrying atoms is larger than in the case of Sect. 4.1, only three independent nearest-neighbor interactions need to be considered. Two of them, J1 and J3, represent interaction between two edges belonging to different Carbon sp chains in graphene nanoholes 30 (Color online) Magnetization-density isosurfaces at +0.01 µB/a3 Figure A1. 0 (dark/red) and −0.01 µB/a3 0 (clear gray) for the magnetic ground state of the unpaired- spin electrons localized at the zig-zag edges of the empty nh. Letters label the spin- carrying atomic sites. Arcs mark all symmetry-independent nearest-neighbor Ising- type magnetic couplings Ji, see Eq. (1). Ising Parameter Value [meV] Standard deviation [meV] E0 − Egs J1 J2 J3 300 −248 10 −53 12 12 12 19 Table A1. nh magnetic configurations. (Color online) The individual Ising-model parameters computed for the sublattices (where we expect an anti-ferromagnetic character). In contrast, the coupling J2 accounts for the interaction of spins within the same edge, and is therefore expected to be ferromagnetic, according to Ref. [67]. As done in Sect. 4.1 for the nh-C6 structures, we make a linear fit of all considered magnetic configurations (shown in Fig. A2), to evaluate the values of the Ji and of E0. In the Ising model, the total energy is written as: Etot = E0 + Espin = E0 − J1 (SASB + SDSE + SF SG + SKSL) + − J2 (SBSC + SCSD + SGSH + SHSK) + − J3 (SASL + SESF ) . (A.1) The result of the linear fit of the energies of the magnetic configurations of Fig. A2 is reported in Table A1. Like in nh-C6, J1 is much larger than the other antiferromagnetic Carbon sp chains in graphene nanoholes 31 (a) Ground State Etot = Egs = −13017.539 eV Mtot = 0.00 Mabs = 11.76 (b) Etot = Egs + 52 meV Mtot = 0.00 Mabs = 10.05 (c) Etot = Egs + 72 meV Mtot = −2.00 Mabs = 9.71 (d) Etot = Egs + 86 meV Mtot = 2.00 Mabs = 10.77 (e) Etot = Egs + 513 meV Mtot = 0.00 Mabs = 11.52 (f) Etot = Egs + 565 meV Mtot = 10.00 Mabs = 10.77 Figure A2. (clear/gray) −0.01 µB/a3 (Color online) Magnetic positive (dark/red) +0.01 µB/a3 0 isosurfaces for the empty nanohole structures. 0 and negative coupling J3. Indeed, the J1 interaction is very similar to the one obtained in the calculations with the nh-C6 structure, see Table 2. Moreover, like in the nh-C6 case, given the relevant error bar the weakly ferromagnetic J2 is in fact compatible with a null value, which is somewhat surprising for neighboring atoms along the same zig-zag edge. Figure A3 compares the DFT-LSDA energy levels and those obtained using the fitted Ising model. Here the fit agrees better than in the nh-C6 case, but clearly the role of the J2 coupling is contradictory, which explains its small value. References [1] Polyynes: Synthesis, Properties, and Applications, edited by F. Cataldo (CRC, Taylor&Francis, London, 2005). [2] Since "polyyne" indicates an alternating single-triple-bond carbon chain, we use the more generic "sp carbon chain" to include also undimerized species, such as cumulenes or odd-n Cn chains. Carbon sp chains in graphene nanoholes 32 Figure A3. (Color online) A comparison of the DFT-LSDA magnetic energy levels of the structures of Fig. A2, with those obtained based on the Ising model, Eq. (A.1), fitted on the DFT-LSDA energies for the empty nh. [3] A. El Goresy and G. Donnay, Science 161, 363 (1968). [4] H. Kroto, Carbon 30, 1139 (1992). [5] W. W. Duley, and A. Hu, Astrophys. J. 698, 808 (2009). [6] F. P. Bundy, W. A. Bassett, M. S. Weathers, R. J. Hemley, H. K. Mao, and A. F. Goncharov, Carbon 34, 141 (1996). [7] R. H. Baughman, Science 312, 1009 (2006). [8] H. Matsuda, H. Nakanishi, and M. Kato, J. Polym. Lett. Ed. 22, 107 (1984). [9] Y. P. Kudryavtsev, Progress of Polymer Chemistry, 87, Nauka (Moscow, 1969). [10] F. Cataldo, D. Capitani, Materials Chem. Phys. 59, 225 (1999). [11] W. Mohr, J. Stahl, F. Hampel, and J. A. Gladysz, Chem. Eur. J. 9, 3324 (2003). [12] X. Zhao, Y. Ando, Y. Liu, M. Jinno, and T. Suzuki, Phys. Rev. Lett. 90, 187401 (2003). [13] Y. Liu et al., Phys. Rev. B 68, 125413 (2003). [14] K. Inoue, R. Matsutani, T. Sanada, and K. Kojima, Carbon 48, 4209 (2010). [15] C. A. Rice, V. Rudnev, R. Dietsche, and J. P. Maier, Astron. J. 140, 203 (2010). [16] M. Kijima, T. Toyabe, and H. Shirakawa, Chem. Commun. 19, 2273 (1996). [17] R. B. Heimann, S. E. Evsyukov, and L. Kavan, Carbyne and Carbynoid Structures (Kluwer, Dordrecht, 1999). [18] M. Tsuji, S. Kuboyama, T. Matsuzaki, and T. Tsuji, Carbon 41, 2141 (2003). [19] F. Cataldo, Tetrahedron Lett. 45, 141 (2004). [20] K. Yamada, H. Kunishige, and A. B. Sawaoka, Naturwiss 78, 450 (1991). [21] K. Ohmura, M. Kijima, and H. Shirakawa, Synth. Metals 84, 417 (1997). [22] M. Kijima, Recent Res. Devel. Pure Appl. Chem. 1, 27 (1997). [23] L. Ravagnan, F. Siviero, C. Lenardi, P. Piseri, E. Barborini, and P. Milani, Phys. Rev. Lett. 89, 285506 (2002). [24] L. Ravagnan, P. Piseri, M. Bruzzi, S. Miglio, G. Bongiorno, A. Baserga, C. S. Casari, A. Li Bassi, C. Lenardi, Y. Yamaguchi, T. Wakabayashi, C. E. Bottani, and P. Milani, Phys. Rev. Lett. 98, 216103 (2007). [25] L. Ravagnan, N. Manini, E. Cinquanta, G. Onida, D. Sangalli, C. Motta, M. Devetta, A. Bordoni, P. Piseri, and P. Milani, Phys. Rev. Lett. 102, 245502 (2009). Carbon sp chains in graphene nanoholes 33 [26] H. E. Troiani, M. Miki-Yoshida, G. A. Camacho-Bragado, M. A. L. Marques, A. Rubio, J. A. Ascencio, and M. Jose-Yacaman, Nano Lett. 3, 751 (2003). [27] C. Jin, H. Lan, L. Peng, K. Suenaga, and S. Iijima, Phys. Rev. Lett. 102, 205501 (2009). [28] A. Chuvilin, J. C. Meyer, G. Algara-Siller, and U. Kaiser, New J. Phys. 11, 083019 (2009). [29] I. M. Mikhailovskij, E. V. Sadanov, T. I. Mazilova, V. A. Ksenofontov, and O. A. Velicodnaja, Phys. Rev. B 80, 165404 (2009). [30] M. G. Zeng, L. Shen, Y. Q. Cai, Z. D. Sha, and Y. P. Feng, Appl. Phys. Lett. 96, 042104 (2010). [31] W. A. Chalifoux and R. R. Tykwinski, C. R. Chimie 12, 341 (2009). [32] E. Hobi Jr., R. B. Pontes, A. Fazzio, and A. J. R. da Silva, Phys. Rev. B 81, 201406 (2010). [33] B. Akdim and R. Pachter, ACSNano 5, 1769 (2011). [34] Y. H. Hu, J. Phys. Chem. C 115, 1843 (2011). [35] L. Ravagnan, T. Mazza; G. Bongiorno, M. Devetta, M. Amati, P. Milani, P. Piseri, M. Coreno, C. Lenardi, F. Evangelista and P. Rudolf, Chem. Commun. 47, 2952 (2011). [36] E. Erdogan1, I. Popov, C. G. Rocha, G. Cuniberti, S. Roche, and G. Seifert, Phys. Rev. B 83, 041401 (2011). [37] T. L. Makarova, B. Sundqvist, R. Hohne, P. Esquinazi, Y. Kopelevich, P. Scharff, V. A. Davydov, L. S. Kashevarova, and A. V. Rakhmanina, Nature (London) 413, 716 (2001). [38] P. Esquinazi, A. Setzer, R. Hhne, C. Semmelhack, Y. Kopelevich, D. Spemann, T. Butz, B. Kohlstrunk, and M. Losche, Phys. Rev. B 66, 024429 (2002). [39] J. M. D. Coey, M. Venkatesan, C. B. Fitzgerald, A. P. Douvalis, and I. S. Sanders, Nature 420, 156 (2002). [40] P. Esquinazi, D. Spemann, R. Hohne, A. Setzer, K.-H. Han, and T. Butz, Phys. Rev. Lett. 91, 227201 (2003). [41] H. Ohldag, T. Tyliszczak, R. Hohne, D. Spemann, P. Esquinazi, M. Ungureanu, and T. Butz, Phys. Rev. Lett. 98, 187204 (2007). [42] D. J. Klein and L. Bytautas, J. Phys. Chem A 103, 5196 (1999). [43] Y. W. Son, M. Cohen and S. Louie, Nature 444, 347 (2006). [44] O. V. Yazyev, and M. I. Katsnelson, Phys. Rev. Lett. 100, 047209 (2008). [45] B. Uchoa, V. N. Kotov, N. M. R. Peres, and A. H. Castro Neto, Phys. Rev. Lett. 101, 026805 (2008). [46] L. Pisani, B. Montanari, and N. M. Harrison, New J. Phys. 10, 033002 (2008). [47] Z. Zanolli and J.-C. Charlier, Phys. Rev. B 81, 165406 (2010). [48] Y. W. Son, M. L. Cohen, and S. G. Louie, Phys. Rev. Lett. 97, 216803 (2006); Nature 444, 347 (2006). [49] L. Yang, M. L. Cohen, and S. G. Louie, Phys. Rev. Lett. 101, 186401 (2008). [50] B. Standley, W. Bao, H. Zhang, J. Bruck, C. N. Lau, and M. Bockrath, Nano Lett. 8, 3345 (2008). [51] Y. Li, A. Sinitskii, and J. M. Tour, Nature Mat. 7, 966 (2008). [52] S. Das Sarma, Am. Scientist 89, 516 (2001). [53] C. S. Casari, A. Li Bassi, L. Ravagnan, F. Siviero, C. Lenardi, P. Piseri, G. Bongiorno, C. E. Bottani, and P. Milani, Phys. Rev. B 69, 075422 (2004). [54] F. Cataldo, L. Ravagnan, E. Cinquanta, I. E. Castelli, N. Manini, G. Onida, and P. Milani, J. Phys. Chem. B 114, 14834 (2010). [55] E. Cinquanta, L. Ravagnan, I. E. Castelli, F. Cataldo, N. Manini, G. Onida, and P. Milani, submitted to J. Chem. Phys. [56] W. E. Pickett, Comput. Phys. Rep. 9, 115 (1989). [57] A. D. Becke, J. Chem. Phys. 98, 5648 (1993). [58] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996). [59] X. Xu and W. A. Goddard III, J. Chem. Phys. 121, 4068 (2004). [60] P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Carbon sp chains in graphene nanoholes 34 Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, J. Phys.: Condens. Matter 21, 395502 (2009). [61] D. Vanderbilt, Phys. Rev. B 41, 7892 (1990). [62] F. Favot and A. Dal Corso, Phys. Rev. B 60, 11427 (1999). [63] The BLA measures the degree of dimerization and, excluding the terminal bonds, can be defined as , with di = (cid:126)ri −(cid:126)ri+1, no = (n + 2)/4, (cid:104)(cid:80)no j=1(d2j−1 + dn−(2j−1))/no −(cid:80)ne (cid:105) 1 j=1(d2j + dn−2j)/ne 2 and ne = n/4 (taken as integer part of these fractions). [64] I. E. Castelli and N. Manini, arXiv:1106.0689. [65] S. Cahangirov, M. Topsakal, and S. Ciraci, Phys. Rev. B 82, 195444 (2010). [66] S. Okada, Phys. Rev. B 77, 041408 (2008). [67] D. Yu, E. M. Lupton, M. Liu, W. Liu, and F. Liu, Nano Res. 1, 56 (2008). [68] M. Fujita, K. Wakabayashi, K. Nakada, and K. Kusakabe, J. Phys. Soc. Jpn. 65, 1920 (1996). [69] K. Kusakabe and M. Maruyama, Phys. Rev. B 67, 092406 (2003). [70] W. Liu, Z. F. Wang, Q. W. Shi, J. Yang, and F. Liu, Phys. Rev. B 80, 233405 (2009). [71] Z. Zanolli, G. Onida, and J.-C. Charlier, ACS Nano 4, 5174 (2010). [72] S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, Rev. Mod. Phys. 73, 515 (2001). [73] H. Tabata, M. Fujii, S. Hayashi, T. Doi, and T. Wakabayashi, Carbon 44, 3168 (2006). [74] I. E. Castelli, Structural and Magnetic Properties of sp-Hybridized Carbon, diploma thesis, (University Milan, 2010), http://www.mi.infm.it/manini/theses/castelliMag.pdf. [75] F. Innocenti, A. Milani, and C. Castiglioni, J. Raman Spectrosc. 41, 226 (2010). [76] L. Colombo, Rivista Nuovo Cimento 28, 1 (2005). [77] C. H. Xu, C. Z. Wang, C. T. Chan, and K. M. Ho, J. Phys.: Condens. Matter 4, 6047 (1992). [78] A. Canning, G. Galli, and J. Kim, Phys. Rev. Lett. 78, 4442 (1997). [79] Y. Yamaguchi, L. Colombo, P. Piseri, L. Ravagnan, and P. Milani, Phys. Rev. B 76, 134119 (2007). [80] E. Cadelano, P. L. Palla, S. Giordano, and L. Colombo, Phys. Rev. Lett. 102, 235502 (2009). [81] F. Bonelli, N. Manini, E. Cadelano, and L. Colombo, Eur. Phys. J. B 70, 449 (2009). [82] R. Zacharia, H. Ulbricht, and T. Hertel, Phys. Rev. B 69, 155406 (2004). [83] See Supplementary Material, Document No. ??, for a few movies illustrating several typical decomposition mechanisms. [84] Ph. Avouris, Z. Chen. and V. Perebeinos, Nature Nanotec. 2, 605 (2007). [85] G. P. Zhang, X. W. Fang, Y. X. Yao, C. Z. Wang, Z. J. Ding, and K. M. Ho, J. Phys.: Condens. Matter 23, 025302 (2011).
1601.02291	1	1601	2016-01-11T01:04:30	Coulomb drag in topological insulator films	[ "cond-mat.mes-hall" ]	We study Coulomb drag between the top and bottom surfaces of topological insulator films. We derive a kinetic equation for the thin-film spin density matrix containing the full spin structure of the two-layer system, and analyze the electron-electron interaction in detail in order to recover all terms responsible for Coulomb drag. Focusing on typical topological insulator systems, with film thicknesses d up to 6 nm, we obtain numerical and approximate analytical results for the drag resistivity $\rho_\text{D}$ and find that $\rho_\text{D}$ is proportional to $T^2d^{-4}n^{-3/2}_{\text{a}}n^{-3/2}_{\text{p}}$ at low temperature T and low electron density $n_{\text{a,p}}$, with a denoting the active layer and p the passive layer. In addition, we compare $\rho_{\text{D}}$ with graphene, identifying qualitative and quantitative differences, and we discuss the multi valley case, ultra thin films and electron-hole layers.	cond-mat.mes-hall	cond-mat	Coulomb drag in topological insulator films Hong Liu ICQD, Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, Anhui, China Weizhe Liu, Dimitrie Culcer∗ School of Physics, The University of New South Wales, Sydney 2052, Australia Abstract We study Coulomb drag between the top and bottom surfaces of topological insulator films. We derive a kinetic equation for the thin-film spin density matrix containing the full spin structure of the two-layer system, and analyze the electron-electron interaction in detail in order to recover all terms responsible for Coulomb drag. Focusing on typical topological insulator systems, with film thicknesses d up to 6nm, we obtain numerical and approximate analytical results for the drag resistivity ρD and find that ρD is proportional to T 2d−4n−3/2 n−3/2 at low temperature T and low p electron density na,p, with a denoting the active layer and p the passive layer. In addition, we compare ρD with graphene, identifying qualitative and quantitative differences, and we discuss the multi valley case, ultra thin films and electron-hole layers. a Keywords: Electron-electron interactions, Coulomb drag, Topological insulator 1. Introduction Three-dimensional topological insulators (3DTIs) are a novel class of bulk insulating materials that possess conducting surface states with a chiral spin texture [1 -- 9]. Thanks to their topology, these surface states re- main gapless in the presence of time-reversal invariant perturbations. Following their initial observation [10 -- 19], improvements in TI growth have made them suit- able for fundamental research [20 -- 23]. Although the reliable identification of the surface states in transport, which remains the key to TIs becoming technologically important, has remained elusive, a number of experi- ments have successfully identified surface transport sig- natures in isolated samples. These were initially mostly singled out via quantum oscillations or in gated thin films [18 -- 22]. Recently, four-point transport measure- ments on clean surfaces in an ultrahigh vacuum have reported a surface-dominated conductivity [24]. A cur- rent induced spin polarization also constitutes a signa- ture of surface transport [25, 26] and was reported in re- cent experimental studies [27 -- 29]. Magnetic TIs have also been successfully manufactured [30 -- 32], and the anomalous [33] and quantum anomalous Hall effects [34, 35] have been detected [36 -- 38]. Hybrid struc- Preprint submitted to Physica E tures such as TI/superconductor junctions have been fabricated [39, 40], which are expected to give rise to topological superconductivity and Majorana fermions [41, 42]. Transport experiments and theoretical work have mostly focused on longitudinal [43 -- 48] and Hall trans- port properties [37, 49 -- 51], thermoelectric response [52 -- 54] and weak antilocalization [55 -- 58], all essen- tially single-particle phenomena. The interplay of strong spin-orbit coupling and electron-electron inter- actions in TIs is at present not completely understood [59 -- 66]. An interaction effect that can be tested experimen- tally in transport is Coulomb drag, which is caused by the transfer of momentum between electrons in differ- ent layers due to the interlayer electron-electron scat- tering. Coulomb drag has been used for decades as an experimental probe of interactions [67 -- 69], and has re- cently attracted considerable attention in massless Dirac fermion systems such as graphene [70 -- 83]. Our focus in this paper is on Coulomb drag in TIs with no mag- netic impurities. Unlike graphene, the spin and orbital degrees of freedom are coupled by the strong spin-orbit interaction, TIs have an odd number of valleys on a October 21, 2018 6 1 0 2 n a J 1 1 ] l l a h - s e m . t a m - d n o c [ 1 v 1 9 2 2 0 . 1 0 6 1 : v i X r a single surface, and the relative permittivity is different, while in known band TIs screening is qualitatively and quantitatively different, since it does not involve the in- terplay of the layer and valley degrees of freedom. All these features impact the drag current. We introduce a density matrix method to calculate the Coulomb drag current in topological insulator films, which fully takes into account the spin degree of freedom and interband coherence. The central result of our work is the drag resistivity, which analytically takes the form ρD = − e2 ζ(3) 16π (kBT)2 a n A2r2 sn 3 2 3 2 p d4 , (1) e2AkFa,p τa,p where kB is the Boltzmann constant, A is the TI spin- orbit constant, rs is the Wigner-Seitz radius (effective fine structure constant) which represents the ratio of the electrons' average Coulomb potential and kinetic en- ergies, d is the layer separation and na,p are the elec- tron densities in the active and passive layers, respec- tively. For a single-valley system rs = e2/(2π0rA), with r the relative permittivity. The intralayer resistiv- ity ρa,p = with kFa,p the Fermi wave vectors. 4π2 The outline of this paper is as follows. In Sec. 2 the interlayer electron-electron scattering matrix is given, including the interlayer screened Coulomb interaction. In Sec. 3 we derive the kinetic equation of topological insulators for spin density matrices of top and bottom surfaces with the full scattering term in presence of an arbitrary elastic scattering potential to linear order in the impurity density. In Sec. 4, we calculate the analytical and numerical expressions of drag resistivity. Our find- ings are summarized in Sec. 5, and we also discusses the broader implications of our results and presents a com- parison with graphene. Sec. 6 discusses extensions of our theory to treat the multi-valley case and ultra-thin films, and briefly touches upon exciton condensation. Finally, Sec. 7 contains our conclusions. 2. Electron-electron interaction The system is described by the many-particle density matrix F, which obeys the quantum Liouville equation [84] where H = H1e + Vee with d F dt H1e = Vee = + i [ H, F] = 0, (cid:88) Hαβc† (cid:88) αcβ, αβ 1 2 αβγδ αc† αβγδc† Vee βcγcδ. In a two-layer system the indices α ≡ kskl represent the wave vector, band, and layer indices respectively. The band index sk = ± with + representing the conduc- tion band and − the valence band, while the layer index l = (a, p) with 'a' the active layer and 'p' the passive layer. The two-particle matrix element Vee αβγδ in a basis spanned by a generic set of wave functions {φα(r)} is given by Vee αβγδ = dr dr(cid:48) ∗ ∗ β(r(cid:48))Vee α(r)φ φ r−r(cid:48) φδ(r)φγ(r(cid:48)), (4) (cid:90) (cid:90) where Vee teraction. r−r(cid:48) = e2 4π0rr−r(cid:48) is the unscreened Coulomb in- The one-particle reduced density matrix is the trace ρξη = tr(c† ηcξ F) ≡ (cid:104)c† ηcξ(cid:105) ≡ (cid:104) F(cid:105)1e, which satisfies [59] dρξη dt + i [ H1e, ρ]ξη = i (cid:104)[ Vee, c† ηcξ](cid:105), (5) (6) where the many-electron averages such as (cid:104)[ Vee, c† are factorized as (cid:104)c† βcγcδ(cid:105) =(cid:104)c† αc† βcδ(cid:105)+Gαβγδ. βcγ(cid:105)−(cid:104)c† αcγ(cid:105)(cid:104)c† αcδ(cid:105)(cid:104)c† ηcξ](cid:105) (7) in which we introduce the Gαβγδ as the matrix elements of the two-particle correlation operator G. Gαβγδ give rise to the electron-electron scattering term in the kinetic equation [84]. The first two terms on the right side of the Eq. (7) which represent the Hartree-Fock mean-field part of the electron-electron interactions have been in- vestigated in Ref. [59]. In Ref. [59] it was demonstrated that the electrical current and nonequilibrium spin po- larization undergo a small renormalization due to the mean-field part of electron-electron interactions and are consequently slightly reduced as compared with their non-interacting values. We are not including this weak renormalization here, so the right-hand side of Eq. (6) only gives the electron-electron scattering term Jee(ρt) which has two contributions, representing intralayer and interlayer electron-electron scattering. Moreover, since the intralayer electron-electron scattering does not con- tribute to the drag current, we concentrate on the inter- layer electron-electron scattering, for which the scatter- ing term is denoted by JInter(ρt). We use below the basis of the eigenstate problem and account for only diagonal (2) (3) 2 0 qq1 vqvq1 e−iq·r, dt1eλt1(cid:104) part of the density matrix JInter( fk) = (cid:104)kJInter(ρt)k(cid:105) (cid:88) JInter( fk) =(cid:104)k 1 2L4 S (t, t1)(1 − ρt1)eiq1·r ρt1 (cid:90) ∞ S +(t, t1)(cid:8)(cid:8)eiq·r S (t, t1)e−iq1·r ρt1 S +(t, t1)(cid:9)(cid:9)− S (t, t1)ρt1eiq1·r(1 − ρt1) S +(t, t1)(cid:8)(cid:8)eiq·r S (t, t1)ρt1e−iq1·r S +(t, t1)(cid:9)(cid:9) + S (t, t1)[ρt1 , eiq1·r] S +(t, t1)(cid:9)(cid:9)(cid:105)k(cid:105), S +(t, t1)(cid:8)(cid:8)eiq·r S (t, t1)ρt1e−iq1·r ρt1 (8) where vq = e2 20rq, 1 is the identity matrix, L2 the area of the 2D system, S (t, t1) the time evolution operator and {{ A}} ≡ A − tr A [84]. The momentum transfer q = q1 = k − k1 = k(cid:48) 1 − k(cid:48). Following a series of simplifications, the interlayer Coulomb interaction eventually takes the form v(pa) k−k1. Without screening k−k1 = vqe−qd. To account for screening, we employ v(pa) the standard procedure of solving the Dyson equation for the two-layer system in the random phase approxi- mation (RPA) discussed in Ref. [68]. In this approach, v(pa) k−k1 in Eq. (8) becomes the dynamically screened in- terlayer Coulomb interaction V(q, ω) = vqe−qd (q, ω) . (9) The dielectric function of the coupled layer system is (q, ω) = [1 − vqΠa(q, ω)][1 − vqΠp(q, ω)] −[vqe−qd]2Πa(q, ω)Πp(q, ω), (10) in which the polarization function is obtained by sum- ming the lowest bubble diagram and takes the form (cid:88) kss(cid:48) Πl(q, ω) =− 1 L2 F(l) sk sk(cid:48) ( f (l) k,s−ε(l) ε(l) 0k,s − f (l) 0k(cid:48),s(cid:48)) k(cid:48),s(cid:48) +ω+i0+ , (11) 0k,s ≡ f (l) with f (l) 0 (εks) the equilibrium Fermi distribution = (cid:104)sklsk1l(cid:105)(cid:104)sk1lskl(cid:105) the wavefunc- function, and F(l) tion overlap. In topological insulator with no tunneling sk sk1 = ). 1 2 sk sk1 F(l) In analytical calculations, (1 + ss(cid:48) k + q cos φ k + q the dynamical screened Coulomb interaction V(q, ω) is usually replaced by the static screened Vq at low temperatures, Πl(q, 0) = − kFl 2πA with kFl the Fermi wave vector. This is because the typ- ical frequencies contributing to the integral are only of (12) the order of kBT/. We approximate Coulomb interac- tion with kFd (cid:29) 1 as (cid:18) e2 (cid:19)2 Vq2 = 20r 4k2 TFa q2 sinh2(qd) k2 TFp , (13) where kTFl rs is the Wigner-Seitz radius introduced in Eq. (1). is the Thomas-Fermi wave vector and = rskFl 2 3. Kinetic Equations At this stage one may include explicitly disorder and driving electric field in the one particle Hamiltonian and write H1e = H0 + HE + U, where H0 is the band Hamilto- nian of TIs, HE = e E· r is the electrostatic potential due to the driving electric flied with r is a position operator and U is the disorder potential. According to Sec. 2, the quantum Liouville equation for the reduced density operator ρ satisfies dρ dt + i [ H1e, ρ] + Jee(ρt) = 0 (14) and will be projected onto the time-independent basis k, sk, l(cid:105). Below we do not write the band indices ex- plicitly. Since the current operator is diagonal in wave vector, the quantity of interest in determining the charge current is the part of the density matrix which is diag- onal in wave vector [6], which here we denote by fk. In the absence of interlayer tunneling, the 4 × 4 matrix fk, which describes the two-layer system, is assumed to be approximately diagonal in the layer index and can be reduced to two 2 × 2 density matrices f (a) k . The quantum Liouville equation can then be broken down k and one for f (p) into two separate equations, one for f (a) k , with different driving terms. The electric field, which is only applied to the active layer, is Ea = Ea x. We re- gard Eq. (8) as the driving term for the passive layer, since the only quantity coupling the two layers is the interlayer electron-electron scattering term. The kinetic equations take the form k and f (p) d f (a) k dt + i [H(a) 0k , f (a) k ] + J0( f (a) k ) = − i [HE k , f (a) 0k ], (15a) d f (p) k dt + 0k , f (p) i [H(p) k ] + J0( f (p) k ) = −JInter (15b) k,p , = Aσ·(k× z) ≡ −Akσ· where the band Hamiltonian H(l) 0k θ, with θ the tangential unit vector corresponding to k. The projection of JInter( fk) onto the eigenstates of the 3 passive layer is denoted by JInter JInter k,p,sk (cid:88) k,p , where v(pa) k−k12δk+k(cid:48),k1+k(cid:48) −ε 1 F(p) sk sk1 −ε(a) k(cid:48),sk(cid:48)] (p) k,sk +ε(a) k(cid:48) 1,sk(cid:48) 1 k(cid:48),sk(cid:48) [1− f (a) ] f (a) k(cid:48) 1,sk(cid:48) 1 [1− f (a) k(cid:48),sk(cid:48) ] f (a) k(cid:48) 1,sk(cid:48) 1 ] (cid:9). =− 2π L4 ×F(a) sk(cid:48) sk(cid:48) 1 ×(cid:8) f (p) k,sk k1 k(cid:48) k(cid:48) 1 (p) δ[ε k1,sk1 [1− f (p) k1,sk1 −[1− f (p) k,sk ] f (p) k1,sk1 k ) =(cid:10)(cid:82) ∞ We do not need to consider the interlayer electron- electron collision integral JInter in the active layer, since k,a that does not produce a drag current. The electron- impurity scattering term in the first Born approximation is given by [33] 2 [ U, e−i Ht(cid:48)/[ U, f ]ei Ht(cid:48)/](cid:11) dt(cid:48) J0( f (l) 0 kk, We will write f (l) k (17) where the notation (cid:104)(cid:105) denotes the average over impurity configurations. k a 2 × 2 Her- mitian matrix which can be written in terms of the Pauli spin matrices. Every matrix in this section can be writ- ten in terms of a scalar part, labeled by the subscript n, and two spin-dependent parts σk(cid:107) and σk⊥. The kinetic Eq. (15) can be written as k , with S (l) k 1 + S (l) = n(l) (18a) (18b) (18c) + Pn J( f (l) k ) = D(l) k,n, dn(l) k dt dS (l) k,(cid:107) dt + P(cid:107) J( f (l) k ) = D(l) k,(cid:107), k ) = D(l) k,⊥, dS (l) k,⊥ dt + i [H(l) 0k, S (l) k,⊥] + P⊥ J( f (l) k k k , f (a) = − i[HE where D(l) k is the driving term for layer l. For the ac- tive layer, D(a) 0k ]; for the passive layer, D(p) k,p , which both have three components D(l) = JInter k,n, D(l) k,(cid:107) and D(l) k,⊥. The projection operator P(cid:107) acts on a ma- trix M as tr(Mσk(cid:107)), where tr refers to the matrix (spin) trace. Analogous definitions hold for the operators P⊥ and Pn. The operators P(cid:107), P⊥, and Pn single out the parts of the density matrix which are parallel to H(l) 0k (in matrix language), orthogonal to H(l) 0k, and scalar, respec- tively. 4. Calculation of Coulomb drag To obtain the drag resistivity, the kinetic equation for the passive layer must be solved. We feed the χ++ l 4 equilibrium density matrix for the passive layer f (p) = k n(p) + S (p) k,0 and the full density matrix of the active k,0 = n(a) layer f (a) Ek into Eq. (8), where k,0 k δ f (a) = n(a) + S (a) Ek is a small correction to the distri- Ek Ek bution function caused by the applied electric field in the active layer + S (a) k,0 + δ f (a) (16) eEaτa(k) · k n(a) Ek = 2 eEaτa(k) · k 2 ∂( f (a) 0+ + f (a) 0− ) ∂k 0+ − f (a) 0− ) ∂k σk(cid:107), ∂( f (a) S (a) Ek,(cid:107) = , (19a) (19b) with τa(k) the momentum scattering time [6]. The re- sulting interlayer electron-electron collision integral be- comes the driving term for the passive layer, and we search for the solution of the kinetic equation for the passive layer, which will yield the drag current. It is easy to verify that the electron-electron collision inte- =(cid:2) f (l) gral vanishes if we replace the density matrix of both layers by its equilibrium form f (l) + S (l) = n(l) 0k. Here 0k 0k 0k,+ − f (l) n(l) 0k,− 0k the equilibrium distribution for the charge and spin dy- namics in layer l respectively, where f (l) 0k,± are the Fermi- Dirac functions for the two energy eigenstates. We solve the kinetic Eq. (18b), yielding the density matrix for the passive layer in the steady state (cid:3)/2 and S (l) (cid:3)σz/2 are =(cid:2) f (l) + f (l) 0k,− 0k,+ 0k S (p) k,(cid:107) = (cid:90) (cid:88) k(cid:48)q k1,+ − ε 1,+ − ε(a) δ[ε(a) k(cid:48) 1)vk(cid:48) 1 eπσk(cid:107) 4kBT L4 (p) k,k1 δ[ε ×F++ ×F++ k,k(cid:48) ×Ea ·(cid:2)τa(k(cid:48) 1 dω (p) k,+ V(q, ω)2 sinh2 βω 2 0k,+ − f (p) + ω]( f (p) 0k1,+) k(cid:48),+ − ω]( f (a) 0k(cid:48),+ − f (a) − τa(k(cid:48))vk(cid:48)(cid:3)τp(k), 1,+) 0k(cid:48) (20) where vk = Aak k. The current operator due to the band Hamiltonian is j = eA σ × z. With jp = tr(jS (p) k,(cid:107)) = jp = σDEa, we have σD = e2 16πkBT dω q V(q,ω)2Im[χ++ a (q,ω)]Im[χ++ sinh2 βω 2 p (q,ω)] , (21) where the nonlinear drag susceptibility for the conduc- tion band of one layer is given by (cid:80) (cid:82) (q, ω) = −2π L2 k,k+q( f (l) F++ k,+ − f (l)k+q,+) ×[τl(k)vk − τl(k + q)vk+q] k,+ − ε(l)k+q,+ + ω + i0+ ε(l) (22) . (cid:88) k The drag problem reduces to the calculation of the non- linear susceptibility of the system. This fact reflects the physical mechanism behind the drag phenomenon: the drag current is a result of the rectification by the pas- sive layer of the fluctuating electric field created by the active layer [68]. In special cases when the intralayer electron-electron correlations are absent, the nonlinear susceptibility is reduced to the product of the diffusion constant and the imaginary part of the polarization op- erator [68]. At low temperatures the predominant con- tribution to drag is due to intraband processes near the Fermi surface. When the Fermi level is finite, i.e. above the Dirac point, electrons take more energy to transition from the valence band to the conduction band than to transition within the conduction band or valence band, and with a small excitation energy the channel involving interband transitions becomes inaccessible [73]. The small interlayer momentum transfer and excitation en- ergy, i.e. q < 2kF, ω < Aq is the dominant region of polarization contributing to the drag problem. Writ- ing τl(k) = τ0kl, we have for the imaginary part of the susceptibility (q, ω) = − (cid:16)2kFl − ω A Imχ++ l ×(cid:2)G> q 8π(cid:113) (cid:17) − G> (cid:16)2kFl τ0q2 q 1 − ( ω Aq )2 + ω A q (cid:17)(cid:3), (23) √ with x2 − 1 − arccosh(x). G>(x) = x (24) The susceptibility, which in general is dependent on the electron momentum, reduces to a momentum- independent form at low temperatures, with the scatter- ing time evaluated at the Fermi level τFl = τ0kFl. 5. Discussion To obtain analytical results, we feed the susceptibil- ity Eq.(23) and the approximate Coulomb interaction Eq.(13) into Eq.(21). With (cid:90) ∞ 0 ω2dω sinh2(ω/2kBT) = 4π2 3 (cid:33)3 (cid:32) kBT , . (25) (26) we have the drag resistivity = − e2 ρD ≈ − σD σaσp ζ(3) 16π (kBT)2 A2r2 sn a n 3 2 3 2 p d4 In Fig. 1(a) we present numerical results for the de- pendence of the Coulomb drag resistivity on the elec- tron number density. The drag resistivity displays a 5 Figure 1: Behavior of drag resistivity ρD as a function of electron concentration n, temperature T, layer separation d: (a) is electron den- sity dependence of ρD at T = 5 K; (b) is temperature dependence of = 0.5 nm; (c) is layer separation d dependence of ρD with kFa ρD at T = 5K and kFa = 0.5 nm. Dielectric constant r = 50. Real and dotted (dashed) lines represent the numerical and analytical respectively. = kFp = kFp d=10nmd=20nmd=40nmd=10nmd=20nmd=40nm0.10.20.512510-1510-1210-910-60.0011n(1012cm-2)ρD(ℏ/e2)(a)d=10nmd=20nmd=40nmd=10nmd=20nmd=40nm24681010-1210-1010-810-610-4T(K)ρD(ℏ/e2)(b)NumericalAnalytical1020304050601.×10-95.×10-91.×10-85.×10-81.×10-75.×10-71.×10-6d(nm)ρD(ℏ/e2)(c)1 dependence with α < 1.5 for d = 10, 20, 40 nm. a nα nα p With increasing electron density n the coefficient α ap- proaches 1.5. The fact that the exact numerical results shown in Fig. 1(a) disagree more strongly with the an- alytical results for smaller values of n and d is under- standable, since Eq. (26) applies only in the kFd (cid:29) 1 limit with kF = 4πn. This trend of an increasing quantitative failure of the asymptotic analytical drag formula for small kFd has also been noted in graphene [72, 73, 75] and 2DEG systems [68, 69]. The analytical result becomes more accurate with increasing kFd. √ In Fig. 1(b) we show the Coulomb drag resistivity as a function of temperature T for three different thicknesses d = 10, 20, 40 nm. The overall temperature depen- dence of the drag resistivity increases nearly quadrati- cally and there is no logarithmic correction due to the absence of backscattering in TIs. The T 2 dependence stems from the allowed phase space where electron- electron scattering occurs at low temperature, and is ex- pected for any interaction strength between the top and bottom layers of TIs as Fig. 1(b) shows, provided that the carriers can be described using a Fermi liquid pic- ture. In addition, in TIs the acoustic phonon velocity is smaller than in graphene. These facts make the con- tribution of electron-phonon scattering processes to the resistivity much more important in the surface of 3DTIs than in graphene. For the surfaces of 3DTIs the effect of electron-phonon scattering events becomes important already for T as low as 10K. For this reason we take consider temperatures up to 10K in our numerical cal- culations. In graphene this effect becomes relevant only beyond T (cid:38) 200K. (26) becomes increasingly accurate and approaches the nu- merical results as the layer separation d increases. It is also evident that Eq. = kFp The behavior of the drag resistivity ρD as a function of layer separation d is shown in Fig. 1(c) with T = 5K and kFa = 0.5 nm. The trend of the exact numer- ical results changes more slowly, a fact that is also em- bodied in Fig. 1(a). Interestingly, in drag experiments on graphene [79], the d dependence of the drag resistiv- ity is much slower than the 1/d4 expected in the weakly interacting regime, varying approximately as 1/d2 for d > 4nm, which is comparable to TIs. Fig. 2 shows the ratio between ρD and the intralayer resistivity ρa corresponding to Fig. 1, which illustrates the relative magnitudes of the drag and longitudinal re- sistivities. The ratio is about 10−6 ∼ 10−8, indicating a small drag resistivity and reflecting the weak electron- electron interactions in TIs. In current 3DTIs, the di- electric constant is as large as 100, indicating weak electron-impurity and electron-electron Coulomb scat- 6 Figure 2: Numerical (blue lines) and analytical (green lines) results of ratio between drag resistivity ρD and intralayer resistivity ρa with ρa = ρp: (a) is electron density dependence of ρD ρa at T = 5 K and d = 20nm; (b) is temperature dependence of ρD = kFp = 0.5 nm and d = 20nm; (c) is layer separation d dependence of ρD ρa at T = 5K and kFa = 0.5 nm. Dielectric constant r = 50 and momentum scattering time τa,p = 0.8 ps. ρa with kFa = kFp NumericalAnalytical0.10.20.512510-910-710-510-3n(1012cm-2)ρD/ρa(a)NumericalAnalytical2468101.×10-85.×10-81.×10-75.×10-71.×10-65.×10-61.×10-5T(K)ρD/ρa(b)NumericalAnalytical10203040506010-1010-910-810-710-610-5d(nm)ρD/ρa(c)tering [85]. So the drag resistivity of TIs is much smaller than that of graphene which the relative static dielectric constant is about 4. This places TIs squarely in the weakly-interacting regime, in which RPA-based theories are applicable. For both Fig. 1 and Fig. 2, we have only applied the approximation kFd (cid:29) 1 for TI film thicknesses up to 6nm, with electron densities n ∼ 1012cm−2. This is a somewhat different regime than graphene, in which the condition kFd (cid:28) 1 can usually be satisfied. How- ever, for other systems with small enough kFd and larger rs, the interlayer Coulomb interaction is in general not small, and it remains to be determined whether it is nec- essary to take into account higher-order contributions in the Coulomb interaction. We have found that approx- imations applied to Vq2 can lead to discrepancies be- tween analytical and numerical results. The numerical result is a factor of ∼ 102 lower than the analytic result, an observation which, in the case of graphene, has been confirmed by drag experiments [76]. 6. Extensions of the theory J( fk) = 6.2. Ultra-thin films When the TI film is ultra thin tunnelling is enabled between the top and bottom surfaces. The tunnelling between the surface states on the top and bottom sur- faces may open an energy gap in the energy spectrum [87]. In this case, the massless Dirac Hamiltonian H(l) 0k needs to be augmented by a series of tunnelling terms, and is generally written as: Hk = Aτz ⊗ [σ · (k × z)] + tτx ⊗ 1, (27) where τ matrices represent the layer pseudospin space, with τz = 1 symbolising the up surface and τz = −1 the bottom surface. Here z is the unit vector in the direction of z, and the term t represents the tunnelling matrix ele- ment between two opposite topological surfaces. After the direct diagonalization, the energy spectrum of the TI t2 + A2k2, which has a gap of size 2t. We discuss how this gap affects the Coulomb drag between the two surfaces. thin film is given by k = ±√ We begin with some general results. The scattering term is: (cid:90) 2π 0 t2 nik 4πA2 ×(cid:18) 1 + dθk(cid:48) Ukk(cid:48)2( fk − fk(cid:48)) (cid:19) A2k2 cos γ t2 + A2k2 , + t2 + A2k2 (28) The focus of the paper up to now has been on thin films of band TIs in which both layers are doped with the same type of carrier, that is, either electron-electron or hole-hole layers. Our theory can straightforwardly be extended to treat structures beyond those considered thus far. In this section we present the necessary mod- ifications for treating the cases of multi-valley TIs and ultra-thin films, and discuss briefly the possibility of ex- citon condensation. 6.1. Multi-valley case Certain materials, such as the topological Kondo in- sulator SmB6, have more than one valley. In this case, a valley degeneracy factor gv will need to be introduced. polarisation Π ∝ √ In the following, we will fix the electron density n. The gv, while the susceptibility χ ∝ gv, thus the screening function (q, ω) is proportional to gv [86]. The drag current, and hence the drag conductivity, will remain the same as in the single-valley case. For √ the drag resistivity, based on Eq. (26) we will get the linear gv dependence because σa,p ∝ 1/ gv. Note that there will be an intervalley impurity scattering contribu- tion if short-range disorder is present in the system, but this will simply result in a renormalisation of the scat- tering time tau by an intervalley scattering term. 7 where γ = θk − θk(cid:48) is the angle between the incident and scattered wave vectors. Note that the density ma- trix entering this term is the full density matrix fk of the double-layer system, rather than its projection onto each individual layer. For t = 0 Eq. (28) reduces to the scattering term introduced in Eq.(17), and we recover the well-known factor of 1 − cos γ, which ensures there is no backscattering in TIs. Based on the ratio of the interlayer tunnelling t and Fermi energy εF, two limit- ing cases can be identified: weak tunnelling t (cid:28) εF and strong tunnelling t (cid:29) εF. So the wavefunction overlap Eq. (12) with tunnelling becomes F(l) sk s(cid:48) k(cid:48) = (1 + ss(cid:48) 1 2 t2 + A2k(k + q cos φ) (t2 + A2k2)(t2 + A2k(cid:48)2) ). (29) We recalculate the polarization and susceptibility, then the related G>(x) in Eq. (24) is rewritten as G>(x) = x x2 − x2 0 − (2 − x2 0)arccosh( x x0 ), (30) (cid:113) (cid:113) where x0 = 1 + 4t2 A2q2−2ω2 . For weak tunnelling, t (cid:28) AkF, the carrier wave func- tions are overwhelmingly located in one of the two lay- ers (surfaces), and the notion of Coulomb drag can be (cid:112) retained to a good approximation. The effect of tun- nelling can be taken into account perturbatively. For example, the momentum relaxation time becomes (cid:18) 1 τ → 1 τ 1 + t2 A2k2 F (cid:19) . (31) (cid:20) (cid:18) (cid:19)2(cid:21) . This implies that, as expected, interlayer tunnelling, which brings with it interlayer scattering, slightly de- creases the momentum relaxation time. Screening is qualitatively different [87], but for weak t it is a good approximation to retain the screening function defined in Eq. (10). With the same method of calculating the t = 0 case, we have the analytical result of the drag re- sistivity with tunnelling becomes ρD ρD(t) = t AkF 1 − 2 [1 + (t/AkF)2]2 ≈ ρD (32) We also recalculate the numerical result. Setting t ≈ 0.1AkF which corresponds to an interlayer separation d ≈ 6 nm, the result also shows that ρD(t0) is marginally smaller than the value that would be obtained by ne- glecting tunnelling. When there the interlayer tunnelling is strong, such that t ∼ AkF, the notion of Coulomb drag is not ap- plicable to the thin film system. In this case, the carrier wave functions are spread over the two layers, hence the picture of the Coulomb interaction causing charges in one layer to drag charges in the other is no longer valid. The main effect of electron-electron interactions will be through the Coulomb renormalization of the conductiv- ity [87]. We note, however, that the film needs to be extremely thin for the tunnelling gap to be noticeable [88], therefore we expect realistic samples to lie in the weak tunnelling limit. 6.3. On the possibility of exciton condensation When the active layer is doped with electrons and the passive layer is doped with holes, or vice-versa, exci- ton condensation may occur. This effect is driven by an exchange term in the interlayer Coulomb interaction. In principle, for interlayer exchange to be nonzero tun- nelling also has to be nonzero, but one can think of a situation in which the tunnelling is negligible but the ex- change is not. The interesting problem concerning the way exciton condensation impacts ρxx has been consid- ered in great detail in Ref. [89], where it was shown that the drag resistivity exhibits an upturn at low tem- peratures described by a logarithmic dependence on the temperature. The theory of Ref. [89] has recently been shown to be a good description of experimental observations in 8 graphene [90], where an exciton condensate phase has been identified, with the critical temperature estimated at 10 - 100mK. This estimate was for a sample grown on a GaAs substrate, in which the effective dielectric constant is expected to be r ≈ 6, whereas in current TI films the lowest experimentally reported r ≈ 30 [91, 92], largely due to screening by the unavoidable bulk of the film. These observations suggest that the critical temperature in TI films could be at least an order of magnitude smaller than in graphene, placing it in the range 1 - 10 mK, which would make exciton condensa- tion rather difficult to detect experimentally in currently available samples. 7. Conclusion a p n−3/2 We have carried out a detailed analysis of the intra- layer and inter-layer electron-electron interactions in TIs in order to determine the Coulomb drag resistiv- ity ρD and devise a complete picture of Coulomb drag in these materials. We have found that ρD is propor- tional to T 2d−4n−3/2 at low temperature and elec- tron density. We have compared our results for ρD with graphene, concluding that the drag effect is expected to be weaker in TIs, and that different regimes are acces- sible experimentally in TIs and graphene. The validity of certain analytical approximations for calculating ρD has also been elucidated. The kinetic equation method presented in this work will be extended in a future publi- cation to describe magnetically doped TIs, in which the anomalous Hall effect makes an important contribution to Coulomb drag. Acknowledgments We are grateful to Zhenyu Zhang, Wenguang Zhu, Zhenhua Qiao, Changgan Zeng, Shun-Qing Shen, W. K. Tse for enlightening discussions. This work was sup- ported by the International Center for Quantum Design of Functional Materials HeFei National Laboratory for Physical Sciences at Microscale University of Science and Technology of China. References References [1] C. L. Kane, E. J. Mele, Quantum spin hall effect in graphene, Phys. Rev. Lett. 95 (2005) 226801. [2] M. Z. Hasan, C. L. Kane, Topological insulators, Rev. Mod. Phys. 82 (2010) 3045. [3] X.-L. Qi, S.-C. Zhang, Topological insulators and superconduc- tors, Rev. Mod. Phys. 83 (2011) 1057. [4] J. E. Moore, L. Balents, Topological invariants of time-reversal- invariant band structures, Phys. Rev. B 75 (2007) 121306. [5] Y. Ando, Topological insulator materials, J. Phys. Soc. Jpn. 82 (2013) 102001. [6] D. Culcer, Transport in three-dimensional topological insula- tors: theory and experiment, Physica E 44 (2012) 860. [7] G. Tkachov, E. M. Hankiewicz, Phys. Status Solidi B 250 (2013) 215. [8] Li, Y. Q. , K. H. Wu, J. R. Shi, X. C. Xie, Electron transport properties of three-dimensional topological insula- tors, Frontiers of Physics 7 (2) (2012) 165. doi:10.1007/ s11467-011-0190-3. URL http://dx.doi.org/10.1007/s11467-011-0190-3 [9] J. E. Moore, The birth of topological insulators, Nature 464 (2010) 194. [10] S. Urazhdin, D. Bilc, S. D. Mahanti, S. H. Tessmer, T. Kyratsi, M. G. Kanatzidis, Surface effects in layered semiconductors bi2se3 and bi2te3, Phys. Rev. B 69 (2004) 085313. [11] D. Hsieh, D. Qian, L. Wray, Y. Xia, Y. S. Hor, R. J. Cava, M. Z. Hasan, A topological dirac insulator in a quantum spin hall phase, Nature 452 (2008) 970. [12] D. Hsieh, Y. Xia, L. Wray, D. Qian, A. Pal, J. H. Dil, J. Oster- walder, F. Meier, G. Bihlmayer, C. L. Kane, Y. S. Hor, R. J. Cava, M. Z. Hasan, Observation of unconventional quantum spin textures in topological insulators, Science 323 (2009) 919. [13] Y. Xia, D. Qian, D. Hsieh, L. Wray, A. Pal, H. Lin, A. Bansil, D. Grauer, Y. S. Hor, R. J. Cava, M. Z. Hasan, Observation of a large-gap topological-insulator class with a single dirac cone on the surface, Nat. Phys. 5 (2009) 398. [14] D. Hsieh, Y. Xia, D. Qian, L. Wray, J. H. Dil, F. Meier, J. Osterwalder, L. Patthey, A. V. Fedorov, H. Lin, A. Bansil, D. Grauer, Y. S. Hor, R. J. Cava, M. Z. Hasan, Time-reversal- protected single-dirac-cone topological-insulator states in bi2te3 and sb2te3: Topologically spin-polarized dirac fermions with pi berry's phase, Phys. Rev. Lett. 103 (2009) 146401. [15] A. A. Taskin, Z. Ren, S. Sasaki, K. Segawa, Y. Ando, topo- insulator, Phys. Rev. Lett. 107 (2011) 016801. Observation of dirac holes logical doi:10.1103/PhysRevLett.107.016801. URL PhysRevLett.107.016801 http://link.aps.org/doi/10.1103/ and electrons in a [16] G. Zhang, H. Qin, J. Teng, J. Guo, Q. Guo, X. Dai, Z. Fang, K. Wu, Quintuple-layer epitaxy of thin films of topological in- sulator bi2se3, Appl. Phys. Lett. 95 (2009) 053114. [17] G. Wang, X.-G. Zhu, Y.-Y. Sun, Y.-Y. Li, T. Zhang, J. Wen, X. Chen, K. He, L.-L. Wang, X.-C. Ma, J.-F. Jia, S. B. Zhang, Q.-K. Xue, Topological insulator thin films of bi2te3 with con- trolled electronic structure, Advanced Materials 23 (26) (2011) 2929. doi:10.1002/adma.201100678. URL http://dx.doi.org/10.1002/adma.201100678 [18] Z. Ren, A. A. Taskin, S. Sasaki, K. Segawa, Y. Ando, Phys. Rev. B 82 (2010) 241306. [19] B. Sac´ep´e, J. B. Oostinga, J. Li, A. Ubaldini, N. J. G. Couto, E. Giannini, A. F. Morpurgo, Gate-tuned normal and supercon- ducting transport at the surface of a topological insulator, Nat. Commun. 2 (2011) 575. URL http://dx.doi.org/10.1038/ncomms1586 [20] D. Kim, S. Cho, N. P. Butch, P. Syers, K. Kirshenbaum, S. Adam, J. Paglione, M. S. Fuhrer, Surface conduction of topo- logical dirac electrons in bulk insulating bi2se3, Nat. Phys. 8 (2012) 459. URL http://dx.doi.org/10.1038/nphys2286 [21] D. Kim, P. Syers, N. P. Butch, J. Paglione, M. S. Fuhrer, Co- herent topological transport on the surface of bi2se3, Nat. Com- mun. 4. URL http://dx.doi.org/10.1038/ncomms3040 [22] C. Brune, C. Thienel, M. Stuiber, J. Bottcher, H. Buhmann, E. G. Novik, C.-X. Liu, E. M. Hankiewicz, L. W. Molenkamp, Dirac-screening stabilized surface-state transport in a topolog- ical insulator, Phys. Rev. X 4 (2014) 041045. doi:10.1103/ PhysRevX.4.041045. URL http://link.aps.org/doi/10.1103/PhysRevX.4. 041045 [23] J. Hellerstedt, M. T. Edmonds, J. H. Chen, W. G. Cullen, C. X. Zheng, M. S. Fuhrer, Thickness and growth-condition dependence of in-situ mobility and carrier density of epitaxial thin-film bi2se3, Applied Physics Letters 105 (17) (2014) 173506. doi:http://dx.doi.org/10.1063/1.4900749. URL journal/apl/105/17/10.1063/1.4900749 http://scitation.aip.org/content/aip/ [24] L. Barreto, L. Khnemund, F. Edler, C. Tegenkamp, J. Mi, M. Bremholm, B. B. Iversen, C. Frydendahl, M. Bianchi, P. Hof- mann, Surface-dominated transport on a bulk topological in- sulator, Nano Letters 14 (7) (2014) 3755. doi:10.1021/ nl501489m. URL http://dx.doi.org/10.1021/nl501489m [25] A. A. Burkov, D. G. Hawthorn, Spin and charge transport on the surface of a topological insulator, Phys. Rev. Lett. 105 (2010) 066802. [26] D. Culcer, E. H. Hwang, T. D. Stanescu, S. Das Sarma, Two- dimensional surface charge transport in topological insulators, Phys. Rev. B 82 (2010) 155457. [27] C. H. Li, O. M. J. van 't Erve, J. T. Robinson, Y. Liu, L. Li, B. T. Jonker, Electrical detection of charge-current-induced spin po- larization due to spin-momentum locking in bi2se3, Nat. Nano. 9 (2014) 218. URL http://dx.doi.org/10.1038/nnano.2014.16 [28] J. Tang, L.-T. Chang, X. Kou, K. Murata, E. S. Choi, M. Lang, Y. Fan, Y. Jiang, M. Montazeri, W. Jiang, Y. Wang, L. He, K. L. Wang, Electrical detection of spin-polarized surface states con- duction in (bi0.53sb0.47)2te3 topological insulator, Nano Let- ters 14 (9) (2014) 5423. doi:10.1021/nl5026198. URL http://dx.doi.org/10.1021/nl5026198 [29] C. Kastl, C. Karnetzky, H. Karl, A. W. Holleitner, Ultrafast he- licity control of surface currents in topological insulators with near-unity fidelity, Nat. Commun. 6. URL http://dx.doi.org/10.1038/ncomms7617 [30] Y. S. Hor, P. Roushan, H. Beidenkopf, J. Seo, D. Qu, J. G. Checkelsky, L. A. Wray, Y. Xia, S.-Y. Xu, D. Qian, M. Z. Hasan, N. P. Ong, A. Yazdani, R. J. Cava, The development of fer- romagnetism in the doped topological insulator bi2-xmnxte3, Phys. Rev. B 81 (2010) 195203. [31] J. Zhang, C.-Z. Chang, P. Tang, Z. Zhang, X. Feng, K. Li, L.- l. Wang, X. Chen, C. Liu, W. Duan, K. He, Q.-K. Xue, X. Ma, Y. Wang, Topology-driven magnetic quantum phase transition in topological insulators, Science 339 (6127) (2013) 1582 -- 1586. doi:10.1126/science.1230905. [32] L. J. Collins-McIntyre, S. E. Harrison, P. Schonherr, N.-J. Steinke, C. J. Kinane, T. R. Charlton, D. Alba-Veneroa, A. Pushp, A. J. Kellock, S. S. P. Parkin, J. S. Harris, S. Lan- gridge, G. van der Laan, T. Hesjedal, Magnetic ordering in cr-doped bi 2 se 3 thin films, EPL (Europhysics Letters) 107 (5) (2014) 57009. URL http://stacks.iop.org/0295-5075/107/i=5/a= 57009 [33] D. Culcer, S. Das Sarma, Anomalous hall response of insulators, Phys. Rev. B 83 (2011) 245441. topological doi:10.1103/PhysRevB.83.245441. URL 83.245441 http://link.aps.org/doi/10.1103/PhysRevB. 9 [34] R. Yu, W. Zhang, H. J. Zhang, S. C. Zhang, X. Dai, Z. Fang, Quantized anomalous hall effect in magnetic topological insula- tors, Science 329 (2010) 61. [35] H.-Z. Lu, A. Zhao, S.-Q. Shen, Quantum transport in magnetic topological insulator thin films, Phys. Rev. Lett. 111 (2013) 146802. doi:10.1103/PhysRevLett.111.146802. URL PhysRevLett.111.146802 http://link.aps.org/doi/10.1103/ [36] H. Jiang, Z. Qiao, H. Liu, Q. Niu, Quantum anomalous hall effect with tunable chern number in magnetic topological in- sulator film, Phys. Rev. B 85 (2012) 045445. doi:10.1103/ PhysRevB.85.045445. [37] C.-Z. Chang, J. Zhang, M. Liu, Z. Zhang, X. Feng, K. Li, L.- L. Wang, X. Chen, X. Dai, Z. Fang, X.-L. Qi, S.-C. Zhang, Y. Wang, K. He, X.-C. Ma, Q.-K. Xue, Thin films of magneti- cally doped topological insulator with carrier-independent long- range ferromagnetic order, Advanced Materials 25 (7) (2013) 1065. doi:10.1002/adma.201203493. URL http://dx.doi.org/10.1002/adma.201203493 [38] C.-Z. Chang, J. Zhang, X. Feng, J. Shen, Z. Zhang, M. Guo, K. Li, Y. Ou, P. Wei, L.-L. Wang, Z.-Q. Ji, Y. Feng, S. Ji, X. Chen, J. Jia, X. Dai, Z. Fang, S.-C. Zhang, K. He, Y. Wang, L. Lu, X.-C. Ma, Q.-K. Xue, Experimental observation of the quantum anomalous hall effect in a magnetic topological insula- tor, Science 340 (2013) 167. [39] I. Sochnikov, A. J. Bestwick, J. R. Williams, T. M. Lippman, I. R. Fisher, D. Goldhaber-Gordon, J. R. Kirtley, K. A. Moler, Direct measurement of current-phase relations in superconduc- tor/topological insulator/superconductor junctions, Nano Lett. 13 (7) (2013) 3086. doi:10.1021/nl400997k. URL http://dx.doi.org/10.1021/nl400997k [40] M.-X. Wang, P. Li, J.-P. Xu, Z.-L. Liu, J.-F. Ge, G.-Y. Wang, X. Yang, Z.-A. Xu, S.-H. Ji, C. L. Gao, D. Qian, W. Luo, C. Liu, J.-F. Jia, Interface structure of a topological insula- tor/superconductor heterostructure, New Journal of Physics 16 (12) (2014) 123043. URL http://stacks.iop.org/1367-2630/16/i=12/a= 123043 [41] C. Nayak, S. H. Simon, A. Stern, M. Freedman, S. Das Sarma, Non-abelian anyons and topological quantum computation, Rev. Mod. Phys. 80 (2008) 1083. [42] Z.-Z. Li, F.-C. Zhang, Q.-H. Wang, Majorana modes in a topo- logical insulator/s-wave superconductor heterostructure, Sci. Rep. 4. URL http://dx.doi.org/10.1038/srep06363 and screening [43] S. Adam, E. H. Hwang, S. Das Sarma, Two-dimensional sur- doi: transport face states, Phys. Rev. B 85 (2012) 235413. 10.1103/PhysRevB.85.235413. URL 85.235413 http://link.aps.org/doi/10.1103/PhysRevB. topological insulator in [44] Q. Li, E. Rossi, S. Das Sarma, Two-dimensional elec- three-dimensional insulators, Phys. Rev. B 86 (2012) 235443. transport the of on surface tronic topological doi:10.1103/PhysRevB.86.235443. URL 86.235443 http://link.aps.org/doi/10.1103/PhysRevB. [45] J. M. Shao, H. Li, G. W. Yang, Conductivity oscillation of surface state of three-dimensional topological insulators in- duced by a linearly polarized terahertz field, Journal of Physics: Condensed Matter 25 (42) (2013) 425603. URL http://stacks.iop.org/0953-8984/25/i=42/a= 425603 [46] M. V. Costache, I. Neumann, J. F. Sierra, V. Marinova, M. M. Gospodinov, S. Roche, S. O. Valenzuela, Fingerprints of 10 inelastic transport at the surface of the topological insulator bi2se3: Role of electron-phonon coupling, Phys. Rev. Lett. 112 (2014) 086601. doi:10.1103/PhysRevLett.112.086601. URL PhysRevLett.112.086601 http://link.aps.org/doi/10.1103/ [47] D. A. Kozlov, Z. D. Kvon, E. B. Olshanetsky, N. N. Mikhailov, S. A. Dvoretsky, D. Weiss, Transport prop- erties of a 3d topological insulator based on a strained high-mobility hgte film, Phys. Rev. Lett. 112 (2014) 196801. doi:10.1103/PhysRevLett.112.196801. URL PhysRevLett.112.196801 http://link.aps.org/doi/10.1103/ [48] P. Syers, D. Kim, M. S. Fuhrer, J. Paglione, Tuning bulk and surface conduction in the proposed topological kondo insulator smb6, Phys. Rev. Lett. 114 (2015) 096601. doi:10.1103/PhysRevLett.114.096601. URL PhysRevLett.114.096601 http://link.aps.org/doi/10.1103/ [49] D. J. Kim, S. Thomas, T. Grant, J. Botimer, Z. Fisk, J. Xia, Surface hall effect and nonlocal transport in smb6: Evidence for surface conduction, Sci. Rep. 3. URL http://dx.doi.org/10.1038/srep03150 [50] J. Wang, B. Lian, S.-C. Zhang, Universal scaling of the quan- tum anomalous hall plateau transition, Phys. Rev. B 89 (2014) 085106. doi:10.1103/PhysRevB.89.085106. URL 89.085106 [51] Y. Zhang, http://link.aps.org/doi/10.1103/PhysRevB. insulator of composite fermions, Phys. Rev. Lett. 113 (2014) 016801. doi:10.1103/PhysRevLett.113.016801. URL PhysRevLett.113.016801 J. Shi, Quantum anomalous hall http://link.aps.org/doi/10.1103/ [52] D. Kim, P. Syers, N. P. Butch, J. Paglione, M. S. Fuhrer, Am- bipolar surface state thermoelectric power of topological insu- lator bi2se3, Nano Letters 14 (4) (2014) 1701. doi:10.1021/ nl4032154. URL http://dx.doi.org/10.1021/nl4032154 [53] A. C. Durst, Low-temperature thermal transport at insulator and a d-wave 094519. (2015) the interface of a topological superconductor, doi:10.1103/PhysRevB.91.094519. URL 91.094519 Phys. Rev. B 91 http://link.aps.org/doi/10.1103/PhysRevB. [54] J. Zhang, X. Feng, Y. Xu, M. Guo, Z. Zhang, Y. Ou, Y. Feng, K. Li, H. Zhang, L. Wang, X. Chen, Z. Gan, S.-C. Zhang, K. He, X. Ma, Q.-K. Xue, Y. Wang, Disentangling the magnetoelectric and thermoelectric transport in topologi- cal insulator thin films, Phys. Rev. B 91 (2015) 075431. doi:10.1103/PhysRevB.91.075431. URL 91.075431 http://link.aps.org/doi/10.1103/PhysRevB. [55] H.-T. He, G. Wang, T. Zhang, I.-K. Sou, G. K. L. Wong, J.-N. Wang, H.-Z. Lu, S.-Q. Shen, F.-C. Zhang, Im- purity effect on weak antilocalization in the topologi- cal insulator bi2te3, Phys. Rev. Lett. 106 (2011) 166805. doi:10.1103/PhysRevLett.106.166805. URL PhysRevLett.106.166805 http://link.aps.org/doi/10.1103/ [56] P. Cheng, C. Song, T. Zhang, Y. Zhang, Y. Wang, J.-F. Jia, J. Wang, Y. Wang, B.-F. Zhu, X. Chen, X. Ma, K. He, L. Wang, X. Dai, Z. Fang, X. Xie, X.-L. Qi, C.-X. Liu, S.-C. Zhang, Q.-K. Xue, Landau quantization of massless dirac fermions in topolog- ical insulator, Phys. Rev. Lett. 105 (2010) 076801. [57] G. Tkachov, E. M. Hankiewicz, Weak antilocalization in hgte quantum wells and topological surface states: Massive versus logical doi:10.1103/PhysRevB.84.235411. URL 84.235411 http://link.aps.org/doi/10.1103/PhysRevB. [58] H.-Z. Lu, massless dirac fermions, Phys. Rev. B 84 (2011) 035444. J. Shi, S.-Q. Shen, Competition between topological Phys. Rev. Lett. 107 (2011) 076801. antilocalization in and localization states, weak surface doi:10.1103/PhysRevLett.107.076801. URL PhysRevLett.107.076801 http://link.aps.org/doi/10.1103/ [59] D. Culcer, Linear insulators, response theory of interacting topo- Phys. Rev. B 84 (2011) 235411. [60] Y. Yamaji, M. Imada, Mott physics on helical edges of two- dimensional topological insulators, Phys. Rev. B 83 (2011) 205122. [61] R. Peters, N. Kawakami, T. Pruschke, Spin-selective kondo in- sulator: Cooperation of ferromagnetism and the kondo effect, Phys. Rev. Lett. 108 (2012) 086402. [62] T. Yoshida, S. Fujimoto, N. Kawakami, Correlation effects on a topological insulator at finite temperatures, Phys. Rev. B 85 (2012) 125113. doi:10.1103/PhysRevB.85.125113. URL 85.125113 http://link.aps.org/doi/10.1103/PhysRevB. [63] P. M. Ostrovsky, I. V. Gornyi, A. D. Mirlin, Interaction-induced criticality in z2 topological insulators, Phys. Rev. Lett. 105 (2010) 036803. doi:10.1103/PhysRevLett.105.036803. URL PhysRevLett.105.036803 http://link.aps.org/doi/10.1103/ [64] J. Wang, D. Culcer, Suppression of the kondo resistivity mini- mum in topological insulators, Phys. Rev. B 88 (2013) 125140. doi:10.1103/PhysRevB.88.125140. [65] S. Das Sarma, Q. Li, Many-body effects and possible super- conductivity in the two-dimensional metallic surface states of three-dimensional topological insulators, Phys. Rev. B 88 (2013) 081404. doi:10.1103/PhysRevB.88.081404. URL 88.081404 http://link.aps.org/doi/10.1103/PhysRevB. [66] H.-Z. Lu, S.-Q. Shen, Finite-temperature conductivity and mag- netoconductivity of topological insulators, Phys. Rev. Lett. 112 (2014) 146601. doi:10.1103/PhysRevLett.112.146601. URL PhysRevLett.112.146601 http://link.aps.org/doi/10.1103/ [67] M. Pograbinskii, Mutual drag of carriers in a semiconductor- insulator-semiconductor system, Sov.Phys.Semicond. 11 (1977) 372. [68] A. Kamenev, Y. Oreg, Coulomb drag in normal metals and superconductors: Diagrammatic approach, Phys. Rev. B 52 (1995) 7516. doi:10.1103/PhysRevB.52.7516. URL 52.7516 http://link.aps.org/doi/10.1103/PhysRevB. [69] A. G. Rojo, Electron-drag effects in coupled electron systems, Journal of Physics: Condensed Matter 11 (5) (1999) R31. URL 004 http://stacks.iop.org/0953-8984/11/i=5/a= [70] W.-K. Tse, S. Das Sarma, Coulomb drag and spin drag in the presence of spin-orbit coupling, Phys. Rev. B 75 (2007) 045333. doi:10.1103/PhysRevB.75.045333. URL 75.045333 http://link.aps.org/doi/10.1103/PhysRevB. [71] M. I. Katsnelson, Coulomb drag in graphene single layers separated by a thin spacer, Phys. Rev. B 84 (2011) 041407. doi:10.1103/PhysRevB.84.041407. URL 84.041407 http://link.aps.org/doi/10.1103/PhysRevB. [72] E. H. Hwang, R. Sensarma, S. Das Sarma, Coulomb drag in monolayer and bilayer graphene, Phys. Rev. B 84 (2011) 245441. doi:10.1103/PhysRevB.84.245441. URL 84.245441 http://link.aps.org/doi/10.1103/PhysRevB. [73] W.-K. Tse, B. Y.-K. Hu, S. Das Sarma, Theory of coulomb drag in graphene, Phys. Rev. B 76 (2007) 081401. doi:10.1103/PhysRevB.76.081401. URL 76.081401 http://link.aps.org/doi/10.1103/PhysRevB. [74] B. Amorim, N. M. R. Peres, On coulomb drag in double layer systems, Journal of Physics: Condensed Matter 24 (33) (2012) 335602. URL http://stacks.iop.org/0953-8984/24/i=33/a= 335602 [75] M. Carrega, T. Tudorovskiy, A. Principi, M. I. Katsnelson, M. Polini, Theory of coulomb drag for massless dirac fermions, New Journal of Physics 14 (6) (2012) 063033. URL 063033 http://stacks.iop.org/1367-2630/14/i=6/a= [76] S. Kim, I. Jo, J. Nah, Z. Yao, S. K. Banerjee, E. Tutuc, Coulomb drag of massless fermions in graphene, Phys. Rev. B 83 (2011) 161401. doi:10.1103/PhysRevB.83.161401. URL 83.161401 http://link.aps.org/doi/10.1103/PhysRevB. [77] B. N. Narozhny, M. Titov, I. V. Gornyi, P. M. Ostrovsky, Coulomb drag in graphene: Perturbation theory, Phys. Rev. B 85 (2012) 195421. doi:10.1103/PhysRevB.85.195421. URL 85.195421 http://link.aps.org/doi/10.1103/PhysRevB. [78] S. Kim, E. Tutuc, Coulomb drag and magnetotransport in graphene double layers, Solid State Communications 152 (15) (2012) 1283. doi:http://dx.doi.org/10.1016/j.ssc. 2012.04.032. URL article/pii/S0038109812002384 http://www.sciencedirect.com/science/ [79] R. V. Gorbachev, A. K. Geim, M. I. Katsnelson, K. S. Novoselov, T. Tudorovskiy, I. V. Grigorieva, A. H. MacDon- ald, S. V. Morozov, K. Watanabe, T. Taniguchi, L. A. Pono- marenko, Strong coulomb drag and broken symmetry in double- layer graphene, Nat. Phys. 8 (2012) 896. URL http://dx.doi.org/10.1038/nphys2441 [80] M. Schutt, P. M. Ostrovsky, M. Titov, I. V. Gornyi, B. N. Narozhny, A. D. Mirlin, Coulomb drag in graphene near the dirac point, Phys. Rev. Lett. 110 (2013) 026601. doi:10.1103/PhysRevLett.110.026601. URL PhysRevLett.110.026601 http://link.aps.org/doi/10.1103/ [81] A. Gamucci, D. Spirito, M. Carrega, B. Karmakar, A. Lom- bardo, M. Bruna, L. N. Pfeiffer, K. W. West, A. C. Ferrari, M. Polini, V. Pellegrini, Anomalous low-temperature coulomb drag in graphene-gaas heterostructures, Nat. Commun. 5. URL http://dx.doi.org/10.1038/ncomms6824 [82] M. Titov, R. V. Gorbachev, B. N. Narozhny, T. Tudorovskiy, M. Schutt, P. M. Ostrovsky, I. V. Gornyi, A. D. Mir- I. Katsnelson, K. S. Novoselov, A. K. Geim, lin, M. L. A. Ponomarenko, Giant magnetodrag in graphene at charge neutrality, Phys. Rev. Lett. 111 (2013) 166601. doi:10.1103/PhysRevLett.111.166601. URL PhysRevLett.111.166601 http://link.aps.org/doi/10.1103/ [83] J. C. W. Song, L. S. Levitov, Hall drag and magneto- drag in graphene, Phys. Rev. Lett. 111 (2013) 126601. doi:10.1103/PhysRevLett.111.126601. URL PhysRevLett.111.126601 http://link.aps.org/doi/10.1103/ 11 [84] F. T. Vasko, O. E. Raichev, Quantum Kinetic Theory and Appli- cations, Springer, New York, 2005. [85] P.Zhang, M.W.Wu, Phys. Rev. B 87 (2013) 085319. [86] S. Das Sarma, S. Adam, E. H. Hwang, E. Rossi, Electronic transport in two dimensional graphene, Rev. Mod. Phys. 83 (2011) 407. [87] W. E. Liu, H. Liu, D. Culcer, Phys. Rev. B 89 (2014) 195417. [88] H.-Z. Lu, W.-Y. Shan, W. Yao, Q. Niu, S.-Q. Shen, Massive dirac fermions and spin physics in an ultrathin film of topolog- ical insulator, Phys. Rev. B 81 (2010) 115407. doi:10.1103/ PhysRevB.81.115407. [89] M. P. Mink, H. T. C. Stoof, R. A. Duine, M. Polini, G. Vignale, Probing the topological exciton condensate via coulomb drag, Phys. Rev. Lett. 108 (2012) 186402. doi:10.1103/PhysRevLett.108.186402. URL PhysRevLett.108.186402 http://link.aps.org/doi/10.1103/ [90] A. Gamucci, D. Spirito, M. Carrega, B. Karmakar, A. Lom- bardo, M. Bruna, L. N. Pfeiffer, K. W. West, A. C. Ferrari, M. Polini, V. Pellegrini, Anomalous low-temperature coulomb drag in graphene-gaas heterostructures, Nat. Commun. 5 (2014) 5824. URL http://dx.doi.org/10.1038/ncomms6824 [91] H. Beidenkopf, P. Roushan, J. Seo, L. Gorman, I. Drozdov, Y. S. Hor, R. J. Cava, A. Yazdani, Spatial fluctuations of helical dirac fermions on the surface of topological insulators, Nat. Phys. 7 (2011) 939. [92] D. Kim, Q. L. P. Syers, N. P. Butch, J. Paglione, S. Das Sarma, M. S. Fuhrer, Phys. Rev. Lett. 109 (2012) 166801. 12
1307.4483	4	1307	2019-11-20T08:24:03	Critical exponent for the Anderson transition in the three dimensional orthogonal universality class	[ "cond-mat.mes-hall", "cond-mat.dis-nn" ]	We report a careful finite size scaling study of the metal insulator transition in Anderson's model of localisation. We focus on the estimation of the critical exponent $\nu$ that describes the divergence of the localisation length. We verify the universality of this critical exponent for three different distributions of the random potential: box, normal and Cauchy. Our results for the critical exponent are consistent with the measured values obtained in experiments on the dynamical localisation transition in the quantum kicked rotor realised in a cold atomic gas.	cond-mat.mes-hall	cond-mat	9 Critical exponent for the Anderson transition in the 1 0 2 three dimensional orthogonal universality class Keith Slevin1 and Tomi Ohtsuki2 1 Department of Physics, Graduate School of Science, Osaka University, 1-1 Machikaneyama, Toyonaka, Osaka 560-0043, Japan 2 Department of Physics, Sophia University, Kioi-cho 7-1, Chiyoda-ku, Tokyo 102-8554, Japan E-mail: [email protected] Abstract. We report a careful finite size scaling study of the metal insulator transition in Anderson's model of localisation. We focus on the estimation of the critical exponent ν that describes the divergence of the localisation length. We verify the universality of this critical exponent for three different distributions of the random potential: box, normal and Cauchy. Our results for the critical exponent are consistent with the measured values obtained in experiments on the dynamical localisation transition in the quantum kicked rotor realised in a cold atomic gas. PACS numbers: 71.23.An,72.15.Rn,71.30.+h,73.43.-f,11.10.Hi v o N 0 2 ] l l a h - s e m . t a m - d n o c [ 4 v 3 8 4 4 . 7 0 3 1 : v i X r a Critical exponent for the Anderson transition... 2 1. Introduction The zero temperature dc conductivity σ of a solid can be expressed using the Einstein relation σ = e2ρ (EF) D . (1) Here ρ (EF) is the density of states per unit volume at the Fermi level and D is the diffusion constant. The electrical conductivity of weakly disordered materials can be understood using a semiclassical picture. The motion of electrons on the scale of the lattice constant is quantum mechanical, while the motion on the scale of the (much longer) mean free path is classical. For classical diffusion, the diffusion constant is equal to D = 1 d vFℓ , (2) where d is the dimensionality, vF the Fermi velocity, and ℓ is the mean free path. Provided the diffusion constant is not zero, we expect the material to be a metal unless the Fermi level lies in a band gap. Then, dependent on the size of the gap, we expect the material to be either an insulator or a semiconductor. The occurrence of a band gap may be explainable within a single particle picture, or it may be the result of correlation effects [1]. Is it possible that the diffusion constant becomes zero? For classical diffusion, the answer is no. For strong disorder the mean free path can be short but it is always finite and so is the diffusion constant. Anderson [2] was the first to realise that for quantum diffusion the situation is different. In this case, quantum interference may result in the complete suppression of diffusion even though the mean free path remains finite. This effect is now called Anderson localisation [3]. The suppression of diffusion is a reflection of a change in the nature of the electronic eigenfunctions. The effect is particularly pronounced in lower dimensions. For one and two-dimensional systems the eigenfunctions are, apart from some special cases, always exponentially localised in space. In three dimensions, eigenfunctions are localised only for sufficiently strong disorder. There is thus a metal insulator transition as the strength of the disorder is increased. This transition is called the Anderson transition and it is an example of a zero temperature continuous quantum phase transition. In common with continuous thermal phase transitions, the concept of universality class plays a central role. In the vicinity of the critical point various critical phenomena described by power laws occur. The exponents appearing in these power laws are expected to be universal, i.e. to depend only on the universality class. The universality classes are determined by the dimensionality of the system and the symmetries of the Hamiltonian. Since we are considering disordered systems, the Hamiltonian does not have any translational symmetry. Rather, the important symmetries for the Anderson localisation problem are time reversal symmetry and spin rotation symmetry. In addition, certain discrete symmetries may also play a role [4, 5]. If these discrete symmetries are ignored, we arrive at the three Wigner-Dyson symmetry classes: Critical exponent for the Anderson transition... 3 orthogonal, unitary and symplectic. If discrete symmetries are included, ten symmetry classes need to be considered [6, 7]. In this paper, we report a finite size scaling study of Anderson's model of localisation. This is defined on a three dimensional lattice and the Hamiltonian has both time reversal and spin-rotation symmetries. Thus, our focus is on the three dimensional orthogonal universality class. The critical phenomena of the Anderson transition are described by two independent critical exponents. The first of these is the critical exponent ν that describes the divergence of the correlation length at the transition (3) 1 ξ ∼ x − xcν . Here, x is the parameter that is varied to drive the transition, and xc is its critical value. The second is the dynamic exponent z that describes how frequency is re-normalised near the critical point. For models where electron-electron interactions are neglected, the only relevant energy scale at the transition is the level spacing. From this it follows that the dynamic exponent is equal to the dimensionality z = d. While concerted efforts have been made to calculate the critical exponent ν using an ǫ expansion about the lower critical dimension (d = 2) [8, 9, 10], reliable values have not been obtained in this way. This gap has been filled by extensive numerical simulations. In addition to the critical phenomena mentioned above, scaling of the wavefunction intensity distribution with system size at the transition is described by a multi- fractal spectrum [11]. This multi-fractal spectrum, which is again expected to display universality, has also been the object of careful numerical study [12, 13, 14]. This article is concerned with the Anderson localisation of electrons. However, Anderson localisation is a wave phenomenon and is also observable for classical waves [15]. The periodically driven quantum kicked rotor exhibits an analogue of localisation in momentum space called dynamic localisation [16]. Moreover, if the amplitude of the kick is modulated quasi-periodically in an appropriate way [17], it is possible to observe a dynamical localisation transition which is believed to be in the same universality class as the Anderson transition in the model we study here [18, 19]. 2. Model and Method 2.1. Anderson's model of localisation The Hamiltonian for Anderson's model of localisation is [2] H = Xi Wic† i ci − V Xhiji c† i cj . (4) Here, the sum in the first term is over the sites of a three dimensional simple cubic lattice with lattice constant a. The sum in the second term is over nearest neighbour lattice sites. The creation operator c† i creates an electron in an orbital ii that is localised Critical exponent for the Anderson transition... 4 on-site i. Orbitals on different sites are assumed to be orthogonal. The state vector of the system is then ψi = Xi ψi ii . (5) If the boundary conditions are specified, the eigenstates and eigenenergies may be found by solving the time-independent Schrodinger equation H ψi = E ψi . The constant V sets the energy scale. We take a as the unit of length, so that a = 1 , and V as the unit of energy, so that V = 1 . (6) (7) (8) The on-site potentials Wi are independently and identically distributed random variables with distribution P (Wi) = p (Wi) dWi . We consider three distributions. The first is the box distribution p (Wi) = ( 1/W Wi ≤ W/2 otherwise 0 . (9) (10) The parameter W characterises the strength of the disorder. In what follows, we usually refer to W simply as the disorder. The second is the normal distribution p (Wi) = 1 √2πσ2 exp − W 2 i 2σ2! . (11) To make it easier to compare with the box distribution, we set the variance of the normal distribution equal to that of the box distribution σ2 = W 2 12 . The third is the Cauchy distribution p (Wi) = W i + W 2) π (W 2 . (12) (13) The parameter W again specifies the strength of the disorder. However, since the Cauchy distribution does not have a second moment, its value is not directly comparable with that of the box and normal distributions. Critical exponent for the Anderson transition... 5 2.2. The transfer matrix method Rather than studying the eigenstates of the Anderson model directly, we consider the transmission of electrons with a given energy E through long disordered wires with a uniform square cross-section Ly = Lz = L . (14) As a consequence of Anderson localisation, for wires that are sufficiently long, the amplitude of the transmission decays exponentially with an associated decay length called the quasi-one-dimensional localisation length ξQ1D. For a given distribution, ξQ1D is a function of the energy, disorder and the transverse dimension L ξQ1D ≡ ξQ1D(E, W, L) . We expect the exponential decay of the transmission to be observable when Lx ≫ ξQ1D . (15) (16) We use the transfer matrix method to estimate ξQ1D [20, 21, 22]. We divide the system into slices labelled by their position x in the x-direction. Since the off-diagonal elements of the Hamiltonian are nonzero only between nearest neighbour sites, the time-independent Schrodinger equation reduces to a set of linear equations relating the wave function amplitudes on adjacent slices. These equations can be rewritten in the form of a transfer matrix multiplication " ψx+1 −ψx # = Mx" ψx −ψx−1 # . (17) Here, ψx groups together all the wavefunction amplitudes ψi on the cross section through the wire at position x ... ψi ... ψx =   ,   and Mx is the transfer matrix defined by Mx = " hx H xi − E 1 0 # . −1 Since there are (18) (19) Ly × Lz = L2 ≡ N (20) wavefunction aplitudes on each slice, the size of the transfer matrix is 2N × 2N. The transfer matrix for a given layer x is a function of the energy E and the on-site potentials Wi of lattice sites on the slice. The boundary conditions in the y and z directions also need to be specified. Throughout this work, we impose periodic boundary conditions in these directions. Critical exponent for the Anderson transition... 6 To estimate the quasi-one-dimensional localisation length we consider the Lyapunov exponents of the following random matrix product Lx M = Mx . Yx=1 From this product we define a real symmetric matrix Ω = ln MM T . (21) (22) As a consequence of current conservation the eigenvalues of this matrix occur in pairs of opposite sign. In what follows, we shall suppose that the eigenvalues νi (i = 1,· · · , 2N) are numbered in decreasing order, i.e. so that νi > νj when i < j. For this ordering, current conservation means that νN +i = −νN −i+1 (23) with i = 1,· · · , N. The Lyapunov exponents associated with the random matrix product are defined by taking the following limit γi = lim Lx→∞ νi 2Lx . (24) In numerical simulations, for practical reasons (see below), the Lyapunov exponents are not calculated by diagonalising the matrix Ω. Instead, to estimate the first m largest Lyapunov exponents we start with a 2N × m matrix U with orthogonal columns and consider the QR decomposition of the matrix MU = QR . (25) Here, Q is a 2N × m matrix with orthogonal columns and R is a m× m upper triangular matrix with positive elements on the diagonal. The Lyapunov exponents are related to the diagonal elements of R by γi = lim Lx→∞ 1 Lx ln Ri,i . (26) In practice, we estimate the Lyapunov exponents by truncating the transfer matrix multiplication at a large but finite length Lx γi = 1 Lx ln Ri,i . (27) Here, the tilde denotes that this is an estimate. Note that these estimates do not necessarily obey the exact symmetry of equation (23). In general, this is recovered only in the limit Lx → ∞. Nevertheless, if we estimate all the Lyapunov exponents by setting m = 2N, we find that the sum of the exponents is exactly zero for any Lx 2N γi = 0 . (28) Xi=1 This is proven by taking the determinant of equation (25). It is also helpful to randomise the starting vectors U by performing several transfer matrix multiplications and QR factorisations, and then replacing U with the final Q matrix. For a more complete discussion of these technicalities see [23]. Critical exponent for the Anderson transition... 7 To relate the Lyapunov exponents with the quasi-one dimensional localisation length, we consider the following scattering problem. Suppose that the wire is connected to perfect leads at both left and right. If we consider an outgoing state at, say, the left of the sample, we can use the transfer matrix multiplication to calculate the wavefunction amplitudes in the right lead. This observation is the basis for a practical method for the calculation of the transmission and reflection matrices for this problem [24]. It also makes clear that the decay of the transmission amplitude will be related to the slowest decay rate in the problem, i.e., the smallest positive Lyapunov exponent. It is thus usual to identify the reciprocal of the smallest positive Lyapunov exponent with the quasi-one dimensional localisation length ξQ1D = 1 γN . (29) Let us just note in passing that, while when performing a finite size scaling analysis (as will be described below) it is customary to consider only the smallest positive Lyapunov exponent, similar analyses of the second, third, etc. smallest positive exponent lead to the same results [25]. 2.3. Dealing with round-off error Unfortunately the transfer matrix multiplication is unstable and, as a result, is very sensitive to round-off error. This means that direct calculation of the Lyapunov exponents by diagonalising the matrix Ω, or QR factorisation of the transfer matrix product M, is not possible. The standard way to overcome this difficulty is to perform repeated QR factorisations at intervals throughout the transfer matrix multiplication. If the the number of transfer matrix multiplications between each QR factorisation is sufficiently small, the loss of precision due to round-off error can be avoided. We start the transfer matrix multiplication with Q(0) = U and perform QR factorisations after every q iterations Q(j)R(j) = Mjq · · · M(j−1)q+1Q(j−1) (j = 1,· · · , l) . For simplicity we assume here that Lx is an integer multiple l of q, i.e. that Lx = lq . The estimates of the Lyapunov exponents are then γi = 1 Lx l Xj=1 ln R(j) i,i . (30) (31) (32) (33) Provided q is small enough to control round-off error, the estimates of the Lyapunov exponents are independent of q. In an attempt to determine a reasonable value for q systematically, we checked how large we can make q while still maintaining the sum of the Lyapunov exponents close to Critical exponent for the Anderson transition... 8 zero. This is illustrated in Tables 1 and 2 for the box and Cauchy distributions. For the Cauchy distribution, in particular, the effect of round-off error is clearly visible in the sum unless q is very small. We also show the estimates of the smallest positive Lyapunov exponent. It is noticeable that the estimates of the smallest positive exponent are much less sensitive to round-off error than the sum of all the exponents. Nevertheless, we preferred to be cautious and so chose q to ensure that the sum of exponents is zero to about single precision accuracy. q Σγi 3 6 9 10−13 10−9 10−4 γN 0.10380011865 0.10380011866 0.10380011848 Table 1. An example of the determination of the interval q at which to perform the QR factorisation in order to control round-off error. The data are for the box distribution with Lx = 9000, Ly = Lz = 24, W = 18 and E = 1. The starting vectors U were randomized with 100 transfer matrix multiplications. q Σγi 1 2 4 10−13 10−7 10−3 γN 0.0953267557872 0.0953267557874 0.0953267558688 Table 2. An example showing that it is more difficult to control round-off error for the Cauchy distribution. Here, Lx = 1000, Ly = Lz = 24, W = 4.5 and E = 0. The starting vectors U were randomized with 100 transfer matrix multiplications. 2.4. Determination of the precision of the Lyapunov exponents To perform the finite size scaling analysis, estimates of the Lyapunov exponents alone are insufficient. In addition, accurate estimates of the precision of the estimates of the Lyapunov exponents are required. In most previous work these estimates were obtained by supposing that the terms appearing in the sum (33) are statistically independent. This assumption is not unreasonable if the interval q between QR factorisations is sufficiently large. However, this is not always the case. Particularly for the Cauchy distribution it is necessary to set q to a small value in order to avoid round-off error. In this case, the assumption of statistical independence leads to erroneously small error estimates. This is serious because it means that reliable estimation of the quality of the finite size scaling fit using the goodness of fit probability is not possible. To circumvent this difficulty we define D(k) i = kr Xj=(k−1)r+1 1 p ln R(j) i,i (k = 1,· · · , s) (34) Critical exponent for the Anderson transition... 9 where r is an integer, p = qr, and we have supposed for simplicity that Lx = ps with s an integer. The estimates of the Lyapunov exponents are then given by the mean values of the Di, γi = 1 s s Xk=1 D(k) i ≡ D . (35) If p is sufficiently large, the assumption that the D(k) are statistically independent for different k is a reasonable approximation. The precision of the estimates of the Lyapunov exponents are then given by the usual formulae for the standard error σi of the mean i 1 where σ2 i = D2 ≡ s − 1 (cid:16)D2 − D i (cid:17)2 Xk=1(cid:16)D(k) 1 s s 2(cid:17) . (36) (37) When performing the simulation, we decide in advance the required precision and stop the simulation when the standard error given by equation (36) satisfies this criterion. (We also take care that a sufficient number of transfer matrix multiplications are always performed to ensure that we have sufficient statistics to estimate the standard error reliably using equation (36), i.e, we set a minimum value for s.) In Figure 1, we show some typical results. The simulation has been set to terminate when according to equation (36) the precision is better than 1%. The simulation has been repeated 100 times using independent streams of random numbers. The observed fluctuation is 1.1%, which is within approximately a single standard deviation of the expected value 1%. Moreover, as expected from the form of equation (35), the normal distribution gives a reasonable description of the sample to sample fluctuations. 2.5. Finite size scaling In our simulations we fix the energy E. There is then a critical disorder Wc ≡ Wc(E) that separates the localised and extended phases. We then accumulate data for the smallest positive Lyapunov exponent for a range of disorder around the critical disorder and for as wide a range of system sizes as practicable. The finite size scaling method is then used to extract information about the critical phenomena from this numerical data. This allows us to estimate universal properties such as the critical exponent, the value Γc of Γ (to be explained below) at the critical point, the scaling function etc., as well as non-universal properties such as the critical disorder. The starting point is to assume that the disorder and system sizes dependence of the dimensionless quantity Γ = γN L , are described by a scaling law of the form [26] Γ = F (φ1, φ2) . (38) (39) Critical exponent for the Anderson transition... 10 0.124 0.126 0.128 0.130 0.132 N Figure 1. The distribution of the estimate of the smallest positive Lyapunov exponent obtained in repeated simulations with the same parameters with independent stream of random numbers. The results shown are for the box distribution with Ly = Lz = 10, W = 15.0 and E = 1. QR factorizations were performed every q = 6 transfer matrix multiplications and we set r = 5 for the determination of the precision. Here, F is a universal function, the arguments φ1 and φ2 are scaling variables φi = ui (w) Lαi , and w is the reduced disorder w = W − Wc Wc . (40) (41) The first of the scaling variables is the relevant scaling variable and has associated with it a positive exponent α1 > 0. The critical exponent is given by the inverse of this exponent ν = 1 α1 . (42) The second is an irrelevant scaling variable and has associated with it a negative exponent α2 < 0. The value of this exponent is usually denoted by the letter y y ≡ α2 . The functions ui appearing in the scaling variables are expanded as Taylor series mi (43) (44) ui (w) = bi,jwj . Xj=0 This allows us to take account of possible nonlinearity of the scaling variables in the disorder. For the relevant scaling variable we must have u1 (w = 0) = 0 . (45) Critical exponent for the Anderson transition... The scaling function is expanded as a Taylor series in the scaling variables F = n1 n2 Xj1=0 Xj2=0 aj1,j2φj1 1 φj2 2 . To avoid ambiguity in the definition of the fitting model we set a1,0 = a0,1 = 1 . 11 (46) (47) In principle, there are many irrelevant scaling variables. However, in practice, it is extremely difficult to resolve contributions of different irrelevant variables. Therefore, we have made the approximation that the contribution of a single irrelevant variable dominates and neglected the others. For the purposes of fitting data near the critical point we expect that it is reasonable to truncate the Taylor series at fairly low order. The total number of parameters in the model is NP = 2 + m1 + m2 + (n1 + 1)(n2 + 1). (48) If the irrelevant variable is neglected in the analysis, the total number of parameters in the model becomes NP = 2 + m1 + n1. (49) To find the best fit, we start with reasonable initial estimates for the fitting parameters and use the Levenberg-Marquardt algorithm to minimize the χ2 statistic χ2 = ND Xi=1 (Fi − Γi)2 σ2 i . (50) Here Fi is the value of finite size scaling model evaluated at the parameters used in the ith run of the simulation, Γi is the value of Γ found in that simulation and ND is the total number of simulations performed, i.e. the number of data. The use of the χ2 statistic is rigorously justified provided the deviations between the model and the data are independent normally distributed random errors, and the model is linear in all the parameters. In this case, the minimum value χ2 min of the χ2 statistic is distributed according to the χ2 distribution with NDOF = ND − NP (51) degrees of freedom. The goodness of fit P , which is the probability that a worse value of χ2 min would be obtained by chance, is given by the formula (see Chapter 11 of the book by Bevington and Robinson [27]) P = Z ∞ χ2 min pχ(cid:16)χ2, NDOF(cid:17) dχ2 where pχ is the χ2 distribution with NDOF degrees of freedom exp (−χ2/2) pχ(cid:16)χ2, NDOF(cid:17) = (cid:16)χ2(cid:17)(NDOF−2)/2 2NDOF/2Γ (NDOF/2) This is calculated as described in [28]; we have P = 1 − P (cid:16)NDOF/2, χ2 min/2(cid:17) . (52) (53) (54) Critical exponent for the Anderson transition... where P (a, x) = 1 Γ (a)Z x 0 exp (−t) ta−1dt 12 (55) is the incomplete Gamma function and Γ (a) is the Gamma function. Here, in fact, the model is linear in some but not all of the parameters. We proceed on the basis that the data are sufficient to narrow down the possible values of the fitting parameters to a small enough region such that, at least in principle, the model could be replaced by a linear approximation. This is true for the most important parameters such as the critical exponent, disorder etc., but not for all the parameters. To determine the precision of the fitted parameters we use the Monte Carlo method described in [28]. This involves the generation and fitting of large numbers of pseudo-data sets. We give the errors in the form of 95% confidence intervals.‡ This method also allows us to check the goodness of fit probability by calculating it directly from the histogram of χ2 obtained in the Monte Carlo simulation. We have found no significant difference with the values obtained from equation (52). 3. Results of the finite size scaling analysis For the three distributions of the random potential, we simulated systems with sizes In each case the simulations were L × L × Lx with L = 4, 6, 8, 10, 12, 16, 20, 24. terminated when Eqs. (35) and (36) indicated that the precision of the estimate of the smallest positive Lyapunov exponent had reached 0.1%. This was typically of the order of Lx = 106 ∼ 107 transfer matrix multiplications. The energies, disorder ranges and total numbers of data points are listed in Table 3. For the box and normal distributions of the random potential, QR factorizations were performed every q = 6 transfer matrix multiplications. The precision of the estimate of the smallest positive Lyapunov exponent was checked after every r = 5 factorizations, i.e. after every p = 30 transfer matrix multiplications. For the Cauchy distribution of the random potential, QR factorizations were performed every q = 2 transfer matrix multiplications and the precision was estimated after every r = 10 factorizations, i.e. after every p = 20 transfer matrix multiplications. The data for each distribution of the random potential were then fitted using the finite size scaling model described above. For the box distribution of the random potential, the starting values used in the non-linear least squares fitting were Wc = 16.53, Γc = 1.73, α1 = 0.63, b1,1 = 1.3, α2 = −2.5 and b2,0 = −0.3. For the normal distribution of the random potential, the starting values were Wc = 6.15, Γc = 1.73, α1 = 0.63, b1,1 = 1.1, α2 = −1.5 and b2,0 = −1.0. And for the Cauchy distribution, the starting values were Wc = 4.3, Γc = 1.72, α1 = 0.63, b1,1 = 1.0, α2 = −2.5 and b2,0 = 0.2. All the other parameters were initially set to zero. ‡ In this article, errors expressed using square brackets are 95% confidence intervals, while numbers appearing after the symbol ± are standard errors. Critical exponent for the Anderson transition... 13 For each data set, a series of fits were performed with m1 = 1, 2, m2 = 0, 1, 2, n1 = 1, 2, 3, 4 and n2 = 1. This step was automated using a combination of fortran and Python scripting. Any fit with an unacceptable goodness of fit, i.e. P < 0.05, was discarded. From the remaining fits, a representative fit was chosen for each distribution of the random potential. The orders of the expansions and the values of χ2 min are listed in Table 3. The estimates of the critical disorder, Γc, the critical exponent ν, and the irrelevant exponent y are listed in Table 4. The data and the fit for box, normal and Cauchy distributions of the random potential are displayed in Figures 2, 3 and 4, respectively. Orders of Expansions p(Wi) Disorder L Details of Fit ≥ 4 m1 = 2, m2 = 2, n1 = 3, n2 = 1 ND = 248, χ2 box W ∈ [15, 18] σ ∈ [5.75, 6.55] ≥ 4 m1 = 2, m2 = 1, n1 = 3, n2 = 1 ND = 328, χ2 normal Cauchy W ∈ [4.1, 4.5] ≥ 4 m1 = 2, m2 = 1, n1 = 2, n2 = 1 ND = 328, χ2 ND = 124, χ2 box ND = 164, χ2 normal W ∈ [15, 18] ≥ 12 m1 = 2, n1 = 3 σ ∈ [5.75, 6.55] ≥ 12 m1 = 2, n1 = 3 min = 239 min = 317 min = 318 min = 112 min = 158 Table 3. The range of data, the orders of the expansions, the total number of data, and the value of χ2 min obtained in the finite size scaling analysis. For the box distribution the energy E = 1, and for the normal and Cauchy distributions E = 0. 16.536[.531, .543] 21.293[.287, .304] Γc p(Wi) Wc 1.7339[.7314, .7371] 1.573[.562, .582] −3.3[−3.9,−2.8] box 1.7371[.7351, .7411] 1.566[.549, .576] −3.1[−4.0,−2.1] normal Cauchy 4.2707[.2680, .2731] 1.7318[.7266, .7360] 1.576[.546, .594] −2.0[−2.4,−1.7] box normal 1.7316[.7286, .7345] 1.577[.568, .586] not applicable 1.7364[.7340, .7388] 1.571[.560, .583] not applicable 16.532[.526, .538] 21.291[.284, .298] ν y Table 4. The results of the finite size scaling analyses. Details of the simulations are given in the corresponding row of Table 3. In Figure 5, we plot the contribution of the irrelevant correction for the fit to the data for the box distribution of the random potential, i.e. we plot the modulus of the sum of the terms with j2 = 1 in Eq. (46) expressed as a percentage of the zero order term (the sum of the terms with j2 = 0). It can be seen that the correction is rapidly decaying with system size and that once L > 10 the correction term is smaller than the precision of our data. For the normal distribution of the random potential the correction was also found to decay rapidly and become negligible compared to the precision of our data for L > 10. For the Cauchy distribution of the random potential, however, the decay was much slower and the corrections are still comparable with the precision of our data even for the largest system sizes. Therefore, for the box and normal distributions, it seemed reasonable to fit data for larger system sizes L ≥ 12 without a correction due Critical exponent for the Anderson transition... 14 2.8 2.4 2.0 1.6 1.2 0.8 15.0 15.5 16.0 16.5 W 17.0 17.5 18.0 Figure 2. Fit of the data for the box distribution of the random potential. 2.4 2.2 2.0 G 1.8 1.6 1.4 1.2 20.0 20.5 21.0 21.5 22.0 22.5 W Figure 3. Fit of the data for the normal distribution of the random potential. Critical exponent for the Anderson transition... 15 2.2 2.0 1.8 1.6 1.4 4.1 4.2 4.4 4.5 4.3 W Figure 4. Fit of the data for the Cauchy distribution of the random potential. to an irrelevant scaling variable. The details and results of these fits are also listed in Tables 3 and 4. 4. Discussion We expect that the critical exponent ν and the quantity Γc are universal, i.e. they depend only on the universality class and not on other details of the model. The results we have presented in Table 4 are clearly consistent with this expectation and confirm the results of our previous analysis [26] based on data for smaller system sizes. In addition, while the irrelevant exponent is not very precisely determined, it seems reasonable to conclude that the irrelevant correction for the Cauchy distribution of the random potential is different from that for the box and normal distributions. There seem to be two possible alternative explanations for this. The first possibility, which we think unlikely, is that there are two fixed points. The scaling at the critical point of the box and normal distributions is controlled by one, and at the critical point of the Cauchy distribution by the other fixed point. Coincidentally, both these fixed points have the same (or at least very close) values of ν and Γc. The second possibility, which we consider more likely, is that there is only one fixed point but that the critical points for the box and normal distributions are positioned in the relevant space in such a way relative to the critical surface as to miss the direction associated with the smallest irrelevant index. This would be the case, for example, if the irrelevant correction of smallest index were associated in some way with distributions which do not have second Critical exponent for the Anderson transition... 16 % ( ) m r e t n o i t c e r r o c 1 0.1 0.01 1E-3 L 15 16 17 18 W Figure 5. The relative value of the irrelevant correction to scaling in the fit of the data for the box distribution of the random potential. moments. In Table 5 we give the weighted average of our estimates for the critical exponent for the box, normal and Cauchy distributions of the random potential, i.e. a weighted average of the values in the first three rows of Table 4. Our result is sightly below the value ν = 1.590[1.579, 1.602] obtained from scaling analysis of the multi-fractal spectrum [13, 14]. There is also a clear difference with the value ν = 1.5 obtained from the formula proposed by Garcia-Garcia [29], which demonstrates that his semi-classical theory is not exact. In Table 5 we also compare our result for the three dimensional orthogonal universality class with published results for the three dimensional unitary and symplectic universality classes. The breaking of time reversal symmetry changes the exponent by roughly ten percent and the breaking of spin-rotation symmetry changes the exponent by a slightly larger but similar amount. The estimates of the exponents for the three dimensional symplectic and unitary classes differ only by a few percent, which is similar to the precision of the estimates themselves. It remains a challenge to reliably distinguish exponents for these latter universality classes in a numerical simulation. Our results for the critical exponent cannot be compared with the results of experiments on metal-insulator transitions observed in disordered electronic systems because interactions between electrons are neglected in Anderson's model of localisation and how these affect the critical behaviour is not yet known. Nevertheless, we can compare our result with the value ν = 1.63 ± .05 found in measurements of the Critical exponent for the Anderson transition... 17 dynamical localisation transition observed in a realisation of the quasi-periodically quantum kicked rotor in a cold atomic gas [19]. Our numerical results are consistent with this experimental measurement. ν Universality Class 1.571[.563, .579] 3D orthogonal (this paper) 1.43[.39, .47] 3D unitary [30] 1.375[.359, .391] 3D symplectic [31] Table 5. List of critical exponents for the three dimensional orthogonal, unitary and symplectic university classes. Before concluding, we comment on the Anderson transition in two dimensions. It is commonly believed that states are always localized in two dimensions [32]. In fact, this is true only for the orthogonal symmetry class. The other nine symmetry classes exhibit an Anderson transition in two dimensions. This includes both the unitary class (Class A) and the symplectic class (Class AII). As described below, in both cases, the finite size scaling method has played a vital role. The transition between quantum Hall plateaux in the integer quantum Hall effect (QHE) that occurs in the unitary class in high perpendicular magnetic fields is an Anderson transition. The critical exponent ν has been well studied both experimentally [33] and numerically. For the Chalker-Coddington model [34, 35], the exponent is estimated to be ν ≈ 2.6 [36, 37, 38, 39, 40] (and see Table 6). The universality of this value is supported by a study of the quantum Hall transition using a periodically driven Hamiltonian model [41]. These results, however, disagree with the experimentally measured value ν ≈ 2.38 [33]. The origin of this discrepancy has not yet been determined but there is a strong suspicion that it originates in the neglect of electron-electron interactions in the numerical simulations. Systems with symplectic symmetry are realized in the presence of strong spin-orbit interaction. Such systems have been attracting renewed interest because the insulating phase is now known to be classified into ordinary and topological insulators [42]. The critical exponent ν for the transition between the metal and the ordinary insulator transition is estimated to be ν ≈ 2.75 [43, 44] (and see Table 6). Most estimates of the exponent for the metal to topological insulator transition [45, 46, 47] are consistent with the conjecture [48] that the exponent for both transitions is the same. The exception is the quite different value ν ≈ 1.6 found [49] in a numerical analysis of the metal- topological insulator transition in the Kane-Mele model [50]. This discrepancy has not yet been explained. Acknowledgments The authors acknowledge F. Evers, A. Furusaki, H. Obuse, and L. Schweitzer for fruitful discussions. This work was supported by Grants-in-Aid for Scientific Research (C) Critical exponent for the Anderson transition... 18 ν Universality Class 2.593[.587, .598] 2.746[.737, .755] 2D unitary (QHE) [36] 2D symplectic [44] Table 6. The critical exponents for the plateau transition in the integer quantum Hall effect and the 2D symplectic university class. (Grants No. 23540376) and Grants-in-Aid 24000013. Appendix A. The Wigner-Dyson classification needs to be extended to take into account discrete symmetries, in particular, the chiral and particle-hole symmetries, that occur in certain disordered systems[4, 5, 6, 7]. The classification is based on Lie algebra, and in addition to the Wigner-Dyson classes, there are 3 chiral and 4 Bogoliubov de Gennes classes. Here we summarize how the Wigner-Dyson classes discussed in this paper are classified according to Lie algebra. Consider an N × N Hermitian matrix H and set X = iH. X is anti-Hermitian, is an element of the Lie algebra u(N), and exp(X) is an element of the Lie group U(N). In the absence of any additional symmetries nothing further can be said, in general, about the Hamiltonian. This is the unitary class in the Wigner-Dyson classification. Any Hermitian matrix may be decomposed as H = H1 + iH2, where H1 is a real symmetric matrix while H2 a real antisymmetric matrix. The matrices H2 are the elements of a Lie algebra that is a subalgebra of u(N) with the corresponding Lie group SO(N), which is a subgroup of U(N). The tangent space to the symmetric space U(N)/O(N) is the space of real symmetric matrices (up to a factor i). The orthogonal class consists of real symmetric matrices, i.e. it spans U(N)/O(N). For the symplectic class (class AII in Table A1) we must consider spin. When spin is included in the description, the number of degrees of freedom is doubled. The Hamiltonian is a 2N × 2N Hermitian matrix, which may be decomposed into 2 × 2 blocks cij containing matrix elements between up and down spin states. Each block may be expressed in the form ij + ib0 ij)τ0 + (a1 ij)τ1 + (a2 ij)τ2 + (a3 cij = (a0 ij + ib1 ij + ib2 (A.1) ij + ib3 ij)τ3 , with τ0 = 12 the 2-dimensional identity matrix, τk = iσk (k = 1, 2, 3) with σk the Pauli matrices, and ak ij (k = 0, 1, 2, 3) real numbers. Since H is Hermitian, ai,j and bi,j must satisfy ij, bk a0 ij = a0 ji , ak ij = −ak ji , bk ij = bk ji (k = 1, 2, 3) .(A.2) ji (k = 1, 2, 3) , b0 ij = −b0 By use of (A.1), a general Hamiltonian is decomposed into H = H1 + H2 where H1 is a matrix with cij of the form cij = a0 ijτ0 + a1 ijτ1 + a2 ijτ2 + a3 ijτ3 , (A.3) Critical exponent for the Anderson transition... 19 and H2 is expressed by the bk X = iH2. The matrices X satisfy ij (k = 0, 1, 2, 3), i.e. the remainder. We then define JX + X TJ = 0 , Jij = δijτ2 , (A.4) and are the elements of a Lie algebra that is a subalgebra of u(2N) with corresponding Lie group Sp(2N). The tangent space to the symmetric space U(2N)/Sp(2N) spans the space of matrices H1 (up to a factor i). The Hamiltonians of systems belonging to the symplectic class are quaternion real, and of precisely this form. WD [H, T ] Orthogonal Symplectic Unitary 0 0 6= 0 [H, S] Symmetric space Symbol 0 ∀S 6= 0 -- U(2N)/Sp(2N) AI AII A U(N)/O(N) U(N) Table A1. Universality classes, corresponding Lie group (symmetric spaces) and Cartan symbol. The first column is the Wigner-Dyson classification. In the second and the third columns, the commutation relations of H with the time reversal operator T and the spin rotation operator S are shown. The 4th column indicates the corresponding symmetric spaces, and the 5th column the Cartan symbols for the symmetry classes of the Hamiltonian. Random Hamiltonians can be mapped to non-linear sigma models [51, 8]. Reflecting the symmetries of the Hamiltonian, the non-linear sigma models are associated with different symmetric spaces. More details can be found in review articles such as [11, 35, 6]. References [1] Dobrosavljevic V, Trivedi N and Valles J M 2012 Conductor-insulator quantum phase transitions (Oxford: Oxford University Press) [2] Anderson P W 1958 Phys. Rev. 109 1492 [3] Abrahams E (ed) 2010 50 years of Anderson localization (World Scientific) [4] Gade R and Wegner F 1991 Nuclear Physics B 360 213 [5] Gade R 1993 Nuclear Physics B 398 499 [6] Zirnbauer M R 1996 J. Math. Phys 37 4986 [7] Altland A and Zirnbauer M R 1997 Phys. Rev. B 55 1142 [8] Hikami S 1981 Phys. Rev. B 24 2671 [9] Bernreuther W and Wegner F J 1986 Phys. Rev. Lett. 57 1383 [10] Wegner F J 1989 Nuclear Physics B 316 663 [11] Evers F and Mirlin A D 2008 Rev. Mod. Phys. 80 1355 [12] Rodriguez A, Vasquez L J and Romer R A 2009 Phys. Rev. Lett. 102 106406 [13] Rodriguez A, Vasquez L J, Slevin K and Romer R A 2010 Phys. Rev. Lett. 105 046403 [14] Rodriguez A, Vasquez L J, Slevin K and Romer R A 2011 Phys. Rev. B 84 134209 [15] Sheng P 1995 Introduction to wave scattering, localization, and mesoscopic phenomena (San Diego: Academic Press) [16] Haake F 2010 Quantum signatures of chaos (Berlin ; New York: Springer) [17] Casati G, Guarneri I and Shepelyansky D L 1989 Phys. Rev. Lett. 62 345 Critical exponent for the Anderson transition... 20 [18] Chab´e J, Lemari´e G, Gr´emaud B, Delande D, Szriftgiser P and Garreau J C 2008 Phys. Rev. Lett. 101 255702 [19] Lopez M, Cl´ement J F, Szriftgiser P, Garreau J C and Delande D 2012 Phys. Rev. Lett. 108 095701 [20] MacKinnon A and Kramer B 1981 Phys. Rev. Lett. 47 1546 [21] Pichard J L and Sarma G 1981 J. Phys. C14 L127 [22] MacKinnon A and Kramer B 1983 Z. Phys. B53 1 [23] Slevin K, Asada Y and Deych L I 2004 Phys. Rev. B 70 054201 [24] Pendry J B, MacKinnon A and Roberts P J 1992 Proceedings of the Royal Society of London. Series A: Mathematical and Physical Sciences 437 67 [25] Slevin K and Ohtsuki T 2001 Phys. Rev. B 63 045108 [26] Slevin K and Ohtsuki T 1999 Phys. Rev. Lett. 82 382 [27] Bevington P R and Robinson D K 2003 Data reduction and error analysis for the physical sciences (Boston: McGraw-Hill) [28] Press W H, Flannery B P, Teukolsky S A and Vetterling W T 1992 Numerical recipes in FORTRAN : the art of scientific computing (Cambridge: Cambridge University Press) [29] Garcia-Garcia A M 2008 Phys. Rev. Lett. 100 076404 [30] Slevin K and Ohtsuki T 1997 Phys. Rev. Lett. 78 4083 [31] Asada Y, Slevin K and Ohtsuk T 2005 J. Phys. Soc. Jpn. Supplement 74 238 [32] Abrahams E, Anderson P W, Licciardello D C and Ramakrishnan T V 1979 Phys. Rev. Lett. 42 673 [33] Li W, Vicente C L, Xia J S, Pan W, Tsui D C, Pfeiffer L N and West K W 2009 Phys. Rev. Lett. 102 216801 [34] Chalker J and Coddington P 1988 J. Phys. C21 2665 [35] Kramer B, Ohtsuki T and Kettemann S 2005 Phys. Rep. 417 211 [36] Slevin K and Ohtsuki T 2009 Phys. Rev. B 80 041304 [37] Obuse H, Subramaniam A R, Furusaki A, Gruzberg I A and Ludwig A W W 2010 Phys. Rev. B 82 035309 [38] Amado M, Malyshev A V, Sedrakyan A and Dom´ınguez-Adame F 2011 Phys. Rev. Lett. 107 066402 [39] Fulga I C, Hassler F, Akhmerov A R and Beenakker C W J 2011 Phys. Rev. B 84 245447 [40] Obuse H, Gruzberg I A and Evers F 2012 Phys. Rev. Lett. 109 206804 [41] Dahlhaus J P, Edge J M, Tworzydlo J and Beenakker C W J 2011 Phys. Rev. B 84 115133 [42] Bernevig B A and Zhang S C 2006 Phys. Rev. Lett. 96 106802 [43] Asada Y, Slevin K and Ohtsuki T 2002 Phys. Rev. Lett. 89 256601 [44] Asada Y, Slevin K and Ohtsuki T 2004 Phys. Rev. B 70 035115 [45] Obuse H, Furusaki A, Ryu S and Mudry C 2007 Phys. Rev. B 76 075301 [46] Kobayashi K, Ohtsuki T and Slevin K 2012 Int. J. Mod. Phys. Conf. Series 11 114 [47] Yamakage A, Nomura K, Imura K I and Kuramoto Y 2013 Phys. Rev. B 87 205141 [48] Fu L and Kane C L 2012 Phys. Rev. Lett. 109 246605 [49] Onoda M, Avishai Y and Nagaosa N 2007 Phys. Rev. Lett. 98 076802 [50] Kane C L and Mele E J 2005 Phys. Rev. Lett. 95 146802 [51] Effetov K B, Larkin A I and Khmel'nitsukii D E 1980 Soviet Phys. JETP 52 568 Appendix B. Errata Caption of Figure 1 In the original version of the article, the system size was given incorrectly as Ly = Lz = 12. The data shown are for Ly = Lz = 10. The caption has been corrected. Critical exponent for the Anderson transition... 21 Table 3 The table has been corrected to make clear that, for the normally distributed random potential, the disorder range given refers to the range of the standard deviation σ not the parameter W (see Eq. 12). Table 4 For the normal distribution, the estimate of the critical value σc of the standard deviation σ was given rather than the critical value Wc of the parameter W . The table has been corrected and now shows the estimate of Wc. Figure 3 The abscissa was labeled as the parameter W but the data were incorrectly plotted versus the standard deviation σ. The data are now correctly plotted versus W .
1012.3459	2	1012	2011-06-29T03:58:34	Anomalous Hall response of topological insulators	[ "cond-mat.mes-hall" ]	The anomalous Hall effect due to the surface conduction band of 3D topological insulators with an out-of-plane magnetization is \textit{always} dominated by an intrinsic topological term of the order of the conductivity quantum. We determine the contributions due to the band structure, skew scattering, side jump and magnetic impurities on the same footing, demonstrating that the topological term, renormalized due to disorder, overwhelms all other terms, providing an unmistakable signature of $Z_2$ topological order. Uncharacteristically, skew scattering contributes in the Born approximation as well as in the third order in the scattering potential, while in addition to the side-jump scattering term we identify a novel intrinsic side-jump term of a similar magnitude. These, however, never overwhelm the topological contribution.	cond-mat.mes-hall	cond-mat	Anomalous Hall response of topological insulators ICQD, Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China, Hefei 230026, Anhui, China Dimitrie Culcer S. Das Sarma Condensed Matter Theory Center, Department of Physics, University of Maryland, College Park, Maryland 20742-4111, USA (Dated: October 10, 2018) The anomalous Hall effect due to the surface conduction band of 3D topological insulators with an out-of-plane magnetization is always dominated by an intrinsic topological term of the order of the conductivity quantum. We determine the contributions due to the band structure, skew scattering, side jump and magnetic impurities on the same footing, demonstrating that the topological term, renormalized due to disorder, overwhelms all other terms, providing an unmistakable signature of Z2 topological order. Uncharacteristically, skew scattering contributes in the Born approximation as well as in the third order in the scattering potential, while in addition to the side-jump scattering term we identify a novel intrinsic side-jump term of a similar magnitude. These, however, never overwhelm the topological contribution. I. INTRODUCTION Topological insulators (TIs) have been an intense focus of research and witnessed impressive experimental break- throughs in the past years1 -- 4. Several materials were predicted to exhibit TI behavior5, and experiment has observed TI behavior in Bi Sb6, Bi2Se3,7 -- 9 Bi2Te3,7,10,11 11, with Heusler alloys also recently predicted and Sb2Te3 to belong to this class12. This extraordinary novel state of matter is remarkable for being insulating in the bulk while conducting along the surface, as well as for the absence of backscattering, which leads to expected high mobilities, Klein tunneling, and the suppression of An- derson localization. The existence of chiral surface states is well estab- lished experimentally through angle-resolved photoe- mission (ARPES) and scanning-tunneling microscopy (STM) studies, yet in transport their contribution is overwhelmed by the bulk, which, due to unintentional defect-induced doping, is a low-density metal. A key aim of current physics research is the identification of a definitive signature of surface transport. One avenue is doping with magnetic impurities, whose moments align beyond a certain doping concentration and make TIs magnetic, opening a gap13. Recent theories of mag- netic TIs include the Kerr and Faraday effects in a thin TI film coupled to a ferromagnet14 and in a strong ex- ternal magnetic field15, spin transfer16, and anomalous magnetoresistance17, and edge contributions18, while ex- periment has observed magnetization dynamics19 and at- tempted to control surface magnetism20. The interplay of magnetism, strong spin-orbit coupling, disorder scat- tering and driving electric fields leads to the anomalous Hall effect (AHE) in magnetic TIs. The AHE has been a mainstay of condensed matter physics, studied exten- sively over the years21 -- 31, and the scattering mechanisms involved, skew scattering22 and side jump23,24,32, are well known to be non-trivial. Furthermore, a fascinating in- stance of this effect has been predicted to occur in 2D magnetic TIs, where edge states are believed to give rise to a quantized AHE when the chemical potential lies in the magnetization-induced gap between the surface valence and conduction bands.13 (See Ref. 33 for a de- scription of the transition between the ordinary and TI regimes in magnetotransport.) This work is devoted to the experimentally relevant problem of the AHE on the surface of a three-dimensional magnetic TI, focusing on the contribution from the sur- face conduction band. This already challenging prob- lem is complicated by the necessity of a quantum me- chanical starting point and of a matrix formulation de- scribing strong band structure spin-orbit coupling, spin- dependent scattering including relativistic corrections, Klein tunneling and Zitterbewegung on the same foot- ing. The fascinating and counterintuitive conclusion of our study is that a topological band structure contri- bution, known to exist in magnetic spin-orbit coupled systems34,35, yields the dominant term, overwhelming all scattering contributions. The result, which is distinct from the topological magnetoelectric effect2,3, is of the order of the conductivity quantum, independent of the magnetization, and unambiguously reflects Z2 topologi- cal order in transport. The outline of this paper is as follows. In Section II the Hamiltonian of the system is introduced. In Section III the kinetic equation for the density matrix is derived, containing all spin-dependent scattering terms in the first and second Born approximations. This includes the usual scalar scattering term in the first Born approximation, the skew scattering and side-jump scattering terms in the first Born approximation, and the usual skew scat- tering term in the second Born approximation (of third order in the scattering potential). In Section IV the vari- ous nonequilibrium corrections to the density matrix are determined, and their contributions to the AHE are cal- culated. Comparison is made with existing experiments as well as predictions for future experiments. Section V contains a summary and conclusions. II. HAMILTONIAN We study a topological insulator doped with mag- netic impurities, which give a net magnetic moment and spin-dependent scattering. The latter are described by Hmag(r) = σ · PI V(r − RI ) sI , where the sum runs over the positions RI of the magnetic ions with sI , and the vector of Pauli spin matrices σ represents the spins of the conduction electrons. We approximate V(r − RI ) = J δ(r − RI ), with J the exchange con- stant between the localized moments and the itinerant carriers, and V the sample volume. The magnetic ions are assumed spin-polarized so that sI = s z. Fourier transforming to the crystal momentum representation {ki}, with k the wave vector, the k-diagonal term, H k=k′ mag = nmag J sσz ≡ M σz, gives the magnetization, with nmag the density of magnetic ions. The k-off- diagonal term leads to spin-dependent scattering H k6=k′ mag = Js V σz XI ei(k−k′)·RI . (1) We consider zero temperature, so that the splitting due to M is resolved, and the Fermi energy εF ≫ M lies in the surface conduction band. We require εF τ / ≫ 1 for our theory to be applicable, with τ the momentum relaxation time. The effective band Hamiltonian is H0k = −Ak σ · θ + σ · M ≡ 2 σ · Ωk, (2) k + b2 where θ is the tangential unit vector in 2D k-space, and the constant A = 4.1eVA for Bi2Se3 5. Scalar terms are negligible at relevant transport densities ≈ 1013cm−25,36. The eigenenergies are ε± = ±√A2k2 + M 2. We will frequently use ak = 2Ak/Ωk and bk = 2M/Ωk, so that a2 k = 1, and aF and bF at k = kF , with bF ≪ 1 at usual transport densities. Interaction with a static, uniform electric field E is contained in HE,kk′ = E,kk′ + H sj H sc E,kk′ , the scalar part arising from the ordi- nary position operator H sc E,kk′ = (eE · r)kk′ 11, with 11 the identity matrix in spin space, and the side-jump part H sj E,kk′ = eλ σ · (k × E) δkk′ arising from the spin-orbit modification to the position operator, r → r + λσ × k,29 with λ a material-specific constant. Elastic scattering off charged impurities, static defects and magnetic ions (but not phonons or electrons) is contained in the scat- tering potential Ukk′ = ¯Ukk′ PJ ei(k−k′)·RJ , with RJ the random locations of the impurities, which incorpo- rates the magnetic scattering term H k6=k′ (1). The potential of a single magnetic impurity ¯Ukk′ = (1 − i λ σ · k × k′)Ukk′ − (Js/V ) σz, with Ukk′ the ma- trix element between plane waves, and λk2 F ≪ 1, while J = 0 for non-magnetic impurities. The full Hamiltonian mag of Eq. 2 Hk = H0k + HEkk′ + Ukk′ + H k=k′ mag . We do not include weak localization, which in the regime εF τ / ≫ 1 pro- duce only a small correction to the conductivity, and is known to be unimportant in the AHE.26 It is well known that in magnetic TIs the Berry cur- vatures of the surface valence and conduction bands are different, so that when the chemical potential is in the band gap the conductivity is e2/2h.35,37 This quantity is regarded as a constant offset in our work, which is con- cerned rather with the response of the conduction elec- trons, that is, the surface conduction band. Nontrivial processes such as spin-flip scattering and the Kondo effect can be accounted for in our theory, provided that one works in the grand canonical ensemble in order to account fully for the spin-flip factors. Several reasons compel us to avoid doing so. Given that the magnetiza- tion must be large enough to be resolved, the local mag- netic moments are assumed to be all lined up. Moreover, although it is reasonable to assume that the Kondo effect may affect the diagonal term in the conductivity signif- icantly, there is no evidence that it should influence the AHE. We will find in this work that all corrections to the AHE are zeroth order or higher in the scattering poten- tial, in other words next-to-leading order. Considering the very few studies of the Kondo effect in topological insulators,38 we have no clear indication of how high the Kondo temperature is expected to be, and the nontrivial effect of spin-momentum locking on Kondo singlet for- mation has not been determined. These factors will be considered in a forthcoming paper. III. KINETIC EQUATION The Liouville equation for the density operator ρ is projected onto the basis {ki} as in Ref. 36. The density matrix ρkk′ = fk δkk′ + gkk′ , where gkk′ is off-diagonal in k, and all quantities are matrices in spin space. To first order in E, the diagonal part fk satisfies36 dfk dt + i [H0k, fk] + J(fk) = Dk. (3) [H E The scattering term J(fk) is given below, the driv- ing term Dk = − i k , f0k], and f0k is the equilib- rium density matrix, which is diagonal in k. We write fk = nk11 + Sk, where Sk is a 2 × 2 Hermitian matrix. Every matrix in this problem can be written in terms of a scalar part, labeled by the subscript n, and σ. Rather than choosing Cartesian coordinates to express the latter, it is more natural to identify three orthogonal directions in reciprocal space, which we denote by Ωk, k, and zef f , and project σ onto them. The three orthogonal direc- tions in reciprocal space that we use for reference are Ωk = −ak θ + bk z kef f = k zef f = ak z + bk θ. 3 The projector Pk acts on a matrix M as tr (Mσk,k), with tr the matrix trace, while P⊥ singles out the part orthogonal to H0k. The total scattering term J(fk) = J Born(fk) + J 3rd(fk), with (4) We project σ as σk,k = σ · Ωk, σk,k = σ · k, and σk,zef f = σ · zef f . Note that σk,k commutes with H0k, while σk,k and σk,zef f do not. Therefore we shall often use the abbreviation ⊥ to refer to vectors in the plane spanned by k and zef f . We project Sk onto the three directions in k-space, writing Sk = Sk,k + Sk,k + Sk,zef f , and defining Sk,k = (1/2) sk,k σk,k, Sk,k = (1/2) sk,k σk,k and Sk,zef f = (1/2) sk,zef f σk,zef f . Equation (18a) can be written as dnk dt dSk,k + Pn J(fk) = Dk,n, J(fk) = Dk,k, [Hk, Sk⊥] + P⊥ J(fk) = Dk,⊥. + Pk dt (5a) (5b) (5c) dSk,⊥ dt + i J Born(fk) = hZ ∞ 0 dt′ 2 [ U , e− i Ht′ [ U , f ] e i Ht′ ]ikk, (6) Considering the negligible size of bk at k = kF , the con- dition yielding the suppression of backscattering is effec- tively unmodified from the non-magnetic case. In the absence of scalar terms in the Hamiltonian the scalar and spin-dependent parts of the density matrix, nk and Sk, are decoupled in the first Born approximation. In the second Born approximation we obtain the additional scattering term J 3rd(fk) = − i 3hZ ∞ 0 dt′Z ∞ 0 dt′′[ U , e− i H t′ [ U , e− i Ht′′ [ U , f ] e i Ht′′ ] e i Ht′ ]ikk, (7) sj ss (fk) + J Born while hi represents averaging over impurity configura- tions. The former can be further broken down into J Born(fk) = J0(fk) + J Born (fk). The first term, J0(fk), in which λ = E = 0 in the time evolution operator, represents elastic, spin-independent, pure mo- mentum scattering. In J Born (fk) we allow λ to be finite but E = 0. In J Born (fk) both λ and E are finite, thus J Born (fk) acts on the equilibrium density matrix f0k. sj Beyond the Born approximation we retain the leading term J 3rd(fk) ≡ J 3rd ss (fk), with λ finite but E = 0, and which is customarily responsible for skew scattering29. The contribution due to magnetic impurities is also ∝ σz ss sj and is contained in J Born ss (fk). The formalism we adopt has a quantum mechanical foundation and accounts for the three mechanisms known to be important in the AHE. The interplay of band struc- ture spin-orbit interaction, adiabatic change in k in the external electric field, and out-of-plane magnetization leads to a sideways displacement of carriers even in the absence of scattering (intrinsic)21. The spin dependence of the impurity potentials via exchange (J) and extrin- sic spin-orbit coupling causes asymmetric scattering of up and down spins (skew scattering)22, and a sideways displacement during scattering (side jump)23. A. Spin-independent scattering term in the Born approximation The projections of J0(fk) that we will need in this work, derived in an analogous manner to Ref. 36, are Pk J0(Skk) = (2π)2 Ukk′2 (skk − sk′k) (1 + b2 k + a2 k cos γ) σkk δ(ε′ + − ε+) πni πni 2 Z d2k′ 2 Z d2k′ 2 Z d2k′ πni P⊥ J0(Skk) = − (2π)2 Ukk′2(skk − sk′k) [( Ωk′ · k) σk,k + ( Ωk′ · zef f ) σk,zef f ] δ(ε′ + − ε+) (8) Pk J0(Sk⊥) = − (2π)2 Ukk′2[(sk,ef f + s′ k,ef f )( Ωk′ · k) + (sz,ef f − s′ z,ef f ) ( Ωk′ · zef f )] σkk δ(ε′ + − ε+), where ni is the impurity density, which emerges once one averages over impurity configurations. B. Born approximation skew scattering and magnetic impurity scattering 4 The effect of this term on the scalar part of the density matrix gives a scalar contribution, which does not yield an electrical current. Next, J Born ss acting on the spin-dependent part of the density matrix gives J Born ss (Sk) = − iniλk2 2 Z d2k′ (2π)2 Ukk′2 sin γ(cid:2)σze−i Hk′ t′ (Sk − Sk′ )eiHkt′ − e−i Hk′ t′ (σzSk − Sk′ σz)eiHk t′ (cid:3) + h.c. (9) In this subsection we omit factors of in the time evolution operators, which are restored in the final results. The two terms in the square brackets are divided as J Born (Sk) = J Born ss,2 (Sk). The first yields ss,1 (Sk) + J Born ss J Born ss,1 (Sk) = πniλk2 2 σz Z d2k′ (2π)2 Ukk′2 sin γ (Sk − Sk′ ) · ( Ωk × Ωk′ ) δ(ε′ + − ε+). (10) This scattering term becomes a source term in the kinetic equation. One will only obtain a contribution linear in τ if one feeds Skk into the RHS, since Skk ∝ τ . But at the same time Skk ∝ Ωk, and its contribution vanishes. The surviving scattering term is J Born (Sk). Its parallel projection is Pk J Born ss (Sk) = πniλk2bk 2 πniλk2ak ss,2 (Sk), and we refer to this term simply as J Born σk,kZ d2k′ σk,kZ d2k′ (2π)2 Ukk′2 sin γ {(sk,ef f + sk′,ef f )(1 + cos γ) ss 2 − − sin γ [ak(skk − sk′k) − bk(skz,ef f − sk′z,ef f )]} δ(ε′ + − ε+). It does not contribute to the Hall current. Its perpendicular projections are (2π)2 Ukk′2 sin γ ak [sin γ(skz,ef f − sk′z,ef f ) − bk (1 − cos γ) (sk − s′ k)] δ(ε′ + − ε+) (11) Pk J Born ss (Sk) = Pz,ef f J Born ss (Sk) = + πniλk2ak 2 πniλk2a2 k 2 πniλk2ak 2 σk,kZ d2k′ σk,zef f Z d2k′ σk,zef f Z d2k′ (2π)2 Ukk′2 sin γ [ak sin γ (sk + s′ k) − (1 − cos γ) (skk − sk′k)] δ(ε′ + − ε+) (2π)2 Ukk′2 sin γ [2 sin γ (sz,ef f − sz,ef f ′ ) − bk(1 − cos γ) (sk − s′ k)] δ(ε′ + − ε+) (2π)2 Ukk′2 sin2 γ bk (skk − sk′k) δ(ε′ + − ε+). (12) Next, in exactly the same manner as above, taking into account the magnetic term in ¯Ukk′ ∝ J, we obtain the scattering term due to magnetic impurities to linear order in bF (i.e. linear in J), referred to as J Born mag (fk). We find that J Born mag (Sk) is a scalar and therefore does not contribute to transport, while J Born mag (nk) = − where ¯U = (1/2π)R dγ Ukk′ (γ) and nmag ≤ ni. nmagJsakk ¯U 22A eEτ cos θ 2 ∂f0+ ∂k σ · zef f , (13) C. Skew scattering beyond the Born approximation The skew scattering term of third order in the scattering potential is J 3rd ss (fk) = − i 3 hZ ∞ 0 dt′ e−ηt′ Z ∞ 0 dt′′ e−ηt′′ [ U , e−i H0t′ [ U , e−i H0t′′ [ U , f ] ei H0t′′ ] ei H0t′ ]kki. (14) In each term the λ-dependent part can appear in three places, so we have a total of 12 terms to evaluate. In each term in []kk we retain the imaginary, spin-dependent part. We find that only J 3rd ss (nk) contributes to skew scattering, so we only retain this term to simplify the algebra below. We let k1, k2 → k2 since this term will eventually be evaluated at the Fermi energy. Skipping a large amount of laborious, yet straightforward, detail, the parallel projection of the third order skew scattering term is 5 Pk J 3rd ss = λk2bk 4 U3 (nk2 − nk1 ) (σ · Ωk) [ (9 + 3b2 k − a2 k)(sin γkk1 + sin γk1 k2 + sin γk2 k) 3a2 k 2 − (sin 2γkk1 + sin 2γk1k2 + sin 2γk2k)] Dk1kDk2k. (15) This term vanishes when we integrate over k1 and k2, thus there is no skew scattering contribution to the AHE that is ∝ τ . The projections k k can be written as Pk J 3rd,A ss = + Pk J 3rd,B ss = 2akbkλk2 3 2akbkλk2 3 2akbkλk2 3 + 2akbkλk2 3 U3 (nk2 − nk) (σ · k)(cid:2) − sin2 γkk1 + sin γk1k2 sin γkk1 − sin γkk1 sin γkk2(cid:3) U3 (nk1 − nk) (σ · k)(cid:2) − sin2 γkk2 − sin γk1k2 sin γkk2 − sin γkk1 sin γkk2(cid:3) U3 (nk2 − nk1) (σ · k)(cid:0) − sin2 γkk1 + sin γk1k2 sin γkk1 − sin γkk1 sin γkk2(cid:1) U3 (nk2 − nk1) (σ · k)(cid:0) sin2 γkk2 + sin γk1 k2 sin γkk2 + sin γkk1 sin γkk2(cid:1). (16) Pk Jss vanishes for isotropic scattering and does not contribute to the Hall current. The projections k zef f can be written as Pz,ef f J 3rd,A ss = λk2ak 3 U3 (nk2 − nk) (σ · zef f )(cid:8) sin γkk1 [(1 + b2 k)(1 + cos γkk1 ) + a2 k cos γkk2 + a2 k cos γk1 k2] + sin γk1k2 [(1 + b2 + sin γk2k[(1 + b2 k)(1 − cos γkk1 ) + a2 k)(1 − cos γkk1 ) − a2 k cos γkk2 − a2 k cos γkk2 + a2 k cos γk1 k2] k cos γk1 k2](cid:9) + (k1 ↔ k2) Pz,ef f J 3rd,B ss = λk2ak 3 U3 (nk2 − nk1) (σ · zef f )(cid:8) sin γkk1 [(1 + b2 k)(1 + cos γkk1 ) + a2 k cos γkk2 + a2 k cos γk1 k2] + sin γk1k2 [(1 + b2 + sin γk2k[(1 + b2 k)(1 − cos γkk1 ) + a2 k)(1 − cos γkk1 ) − a2 k cos γkk2 − a2 k cos γkk2 + a2 k cos γk1 k2] k cos γk1 k2](cid:9) + (k1 ↔ k2). (17) We will leave the results in this form until Section IV, where they will be evaluated explicitly. In the Born approximation, we solve Eq. (18b) to first order in U D. Side jump One contribution due to side jump arises from the in- trinsic side-jump driving term, considered in Section IV. The other two contributions come from corrections to the scattering term. We retain only the parallel projec- tion of the side-jump scattering terms, the others being of higher order in /(εF τ ). The easiest way is to write the kinetic equation in terms of the k−diagonal and off- diagonal components of the density matrix i [ U , g]kk,(18a) [H E i i + k , fk] − dfk dt [H0k, fk] = − [ H, g]kk′ = − dgkk′ dt + i i [ U , f + g]kk′ . (18b) gkk′ = − i Z ∞ 0 dt′ e−i Ht′ h U , f (t − t′)i ei Ht′(cid:12)(cid:12)(cid:12)(cid:12)kk′ . (19) To include side jump, we take into account an additional correction to g arising from HE. Still in the Born ap- proximation we write g = g0 + gE dg0,kk′ dt + i dgE,kk′ dt + i i [ U , f ]kk′ [ H0, g0]kk′ = − i g0,kk′ = − (cid:26)Z ∞ 0 dt′ e−ηt′ e−i H0t′ [ U , f (t − t′)] ei H0t′(cid:27)kk′ i [ HE, g0]kk′ [ H0, gE]kk′ = − i gE,kk′ = − (cid:26)Z ∞ 0 dt′ e−ηt′ e−i H0t′ [ HE, g0(t − t′)] ei H0t′(cid:27)kk′ . 6 (20) Factors of have been left out of the time evolution operators to shorten the equations. We adopt the Markovian approximation f (t − t′) ≈ f (t) ≡ f , and, remembering that HE is first order in the electric field, we replace f → f0, the equilibrium density operator gE,kk′ (t) ≈ − 1 2 (cid:26)Z ∞ 0 dt′′ e−ηt′′ e−i H0t′′ [ HE, e−i H0t′ [ U , f0] ei H0t′ ] ei H0t′′(cid:27)kk′ . (21) Both HE and U have scalar and spin-orbit parts. We first evaluate the easier contribution from H sj part of the scattering potential, which we call gsj E,k1k2 . This contribution can be expressed as E,kk′ and the scalar gsj E,k1k2 = − i Z ∞ 0 dt′ e−ηt′ e−iH0k1 t′ (H sj Ek1g0k1k2 − g0k1k2 H sj Ek2) eiH0k2 t′ where g0k1k2 is taken to zeroth order in λ. This gives rise to the scattering term Xk1 where only the scalar part of ¯Ukk1 contributes, since gsj Jsj(f0k) = We require in addition the scattering term i ¯Ukk1gsj E,k1k + h.c, E,k1k is already first order in λ. Jsc(f0k) = i Xk1 ¯Ukk1 gsc E,k1k + h.c. . (22) (23) (24) The term Jsc(f0k) has two contributions. One contribution arises from the spin-independent part of ¯Ukk′ and g0 to first order in λ. The other term comes from the spin-dependent part of ¯Ukk′ and g0 to zeroth order in λ. Therefore we only need to evaluate one other term to first order in λ, gsc E,k1k3 = − i Z ∞ 0 dt′ e−ηt′ e−iH0k1 t′ (H sc Ek1 k2g0k2k3 − g0k1k2 H sc Ek2k3 ) eiH0k3 t′ . (25) We will leave the result in this form until Section IV when it will be evaluated explicitly. IV. ANOMALOUS HALL EFFECT correctly is crucial, since if terms ∝ n−1 they will be dominant at high mobilities. i exist in the AHE We solve Eq. (5) perturbatively in the small parameter /(εF τ ) (which is proportional to the impurity density ni) as described in Ref. 36, as well as in λ. In transport this expansion starts at order [/(εF τ )]−1 (i.e. n−1 ), reflecting the competition between the driving electric field and impurity scattering resulting in a shift of the Fermi surface. The leading terms in [/(εF τ )]−1 appear 36, namely in the components of the density in nk and Sk,k matrix which commute with H0k. Identifying these terms i We first solve Eqs. (5) to zeroth order in λ, setting J(fk) → J0(fk). We refer to this part of the solution as the bare part, which is gives the intrinsic contribu- tion to the AHE (i.e. λ = 0.) Once the bare compo- nents of Sk are determined, we project repeatedly be- tween Sk,k and Sk,⊥ in Eq. (5) until we have found all terms to [/(εF τ )]−1 and [/(εF τ )]0. These projections give disorder renormalizations, explained in a separate subsection below. Subsequently, the scattering terms The transport scattering time for screened Coulomb im- purities is defined as 7 1 τ = ε+ A niWk 2A (ζ0 − ζ1) Wk = Z 2e4 rk2 0ε2 F 4ε2 (29) ζ = 1 2 (1 + b2 k + a2 k cos γ) 1 2 + kT F 2kF (cid:1)2 . (cid:0) sin γ ζ contains a factor of 1/2 as it is defined in analogy with the B = 0 case,36 while ζ0,1 refer to its Fourier compo- nents, and kT F 4 , with the Wigner-Seitz radius rs = 2kF = rs e2/(2πǫ0ǫrA). This way Ukk′2 = Wk/(cid:0) sin γ The (bare) perpendicular part of the spin density ma- 4 (cid:1)2 2 + rs . trix is Sbare Ek,zef f = Sbare Ek,k = 1 2 1 2 eE · θ kΩ eE · k kΩ σ · zef f σ · k ak f0+ (30) akbk f0+. This completes the evaluation of the bare correction to the spin density matrix, nbare Ek , which will yield the intrinsic contribution to the Hall current and its disorder renormalization. Ek +Sbare B. Disorder renormalizations The correction to the bare density matrix is renormal- ized by scattering between Sk,⊥ and Sk,k. In addition, a number of terms appearing below also need to be renor- malized in this way. We briefly describe the technical procedure in this subsection. Projections of the form −P⊥ J(Sk,k) involve no compli- cations, since Eq. (5c) is solved immediately by applying the time evolution operator.36 For projections of the form J(Sk,⊥), it is helpful to observe that the Fourier ex- −Pk pansions of sk and sz,ef f throughout this work will only have Fourier component 1 (i.e. e±iθ), thus generally, us- ing Eqs., J Born (fk) + J 3rd(fk) act on the bare solution to gen- ss erate new driving terms to first order in λ, for which an analogous solution is found to the density matrix. These give the skew scattering contributions, and the contribu- tion due to magnetic impurities. Finally the side-jump driving terms are found, in which fk → f0k. Given the lengthy expressions, below we present only the terms that contribute to the AHE. The appearance of terms to order zero in /(εF τ ) is standard in transport and important in spin-dependent transport. These terms depend on the angular structure of the scattering potential, thus it is not enough to study only the band structure contribution, but also its disorder renormalization. Disorder renormalizations are crucial in spin-related transport, where they can cancel band structure contributions39. Though additional terms of order [/(εF τ )]0 may exist, they arise from scattering terms ∝ n2 i , describing quantum interference and related effects relevant at higher impurity densities. A. Bare driving term To first order in E, we write fk = f0k + fEk, with cor- responding notation for nk and Sk. The Fermi-Dirac function for the positive energy branch is denoted by f0+ = f0(ε+), with f0− = 0 for electron doping. The equilibrium density matrix is f0k = 1 2 (f0+ + f0−)11 + 1 2 (f0+ − f0−)σk,k. The driving term is Dk = − i [H E k , f0k] = (eE/) · ∂f0k ∂k − ieλ [σ · (k × E), f0k]. (26) Its projections are Dk = Dk,n + Dk,k + Dk,k + Dk,zef f . To zeroth order in λ Dk,n = Dk,k = Dk,k = eE · k 2 (cid:18) ∂f0+ ∂k (cid:19) eE · k (cid:18) ∂f0+ ∂k (cid:19) eE · θ k ak (f0+) σk,k σk,k 1 2 1 2 Dk,zef f = − 1 2 σk,zef f eE · k k akbk (f0+). (27) J0(SEk,k) = −Pk akbksk τc+ σk,k J0(SEk,zef f ) = −Pk akbkszef f τµ σk,k. (31) The effective scattering times τµ, τc+ are given below. The projected terms −P⊥ J0(Sk,k) and −Pk J(Sk,⊥) then act as new driving terms for Sk,⊥ and Sk,k respectively, in effect renormalizing the bare results by a multiplicative constant which depends on the angular characteristics of the scattering potential. All terms in this problem have to be renormalized in this way. Yet all terms in the density matrix consist of Fourier component 1 (that Henceforth E k x. We obtain eEτ · k nbare Ek = 2 Sbare Ek,k = 1 2 σ · Ωk ∂f0+ ∂k eEτ · k (28) ∂f0+ ∂k . is, they are of the form e±iγ, or alternatively sin γ and cos γ), thus the renormalizations all follow the pattern of Eqs. (31). Only the bare term Sbare Ek needs to be renormalized due to scattering from Skk to Sk⊥. The corrections to the perpendicular part of the spin density matrix due to scattering from Sbare Ekk into Sk,k and Sk,zef f are Svtx Ek,k = eExbkτ cos θ τc−Ω ∂f0+ ∂k σ · k Svtx Ek,zef f = − eExτ sin θ τsΩ ∂f0+ ∂k σ · zef f , (32) The effective scattering times are 1 τs 1 τc± 1 τµ = = = c(γ) = s(γ) = µ(γ) = 2A ε+ niWks1 A ε+ niWk (c0 ± c1) A 2A ε+ niWk (µ0 − µ1) A 2A cos γ 1 4 (cid:1)2 2 (cid:0) sin γ 2 + rs 1 sin γ 4 (cid:1)2 2 (cid:0) sin γ 2 + rs 1 − cos γ 1 4 (cid:1)2 . 2 (cid:0) sin γ 2 + rs The renormalizations of all the other terms appearing be- low due to scattering from Skk to Sk⊥ are of higher orders in λ or /(εF τ ) and need not be considered. Renormal- izations due to scattering from Sk⊥ to Skk can be worked out immediately from Eq. (8). We emphasize that, aside from the transport relaxation time τ , the effective scattering times appearing in this work are defined for convenience and they do not nec- essarily represent specific scattering processes, but are auxiliary quantities in achieving a final result. C. Extrinsic contributions to the density matrix 8 full spin-dependent Ukk′ in both, generating new source terms ∝ λ in the kinetic equation. The terms linear in λ in J Born(fEk) yield Sss,Born Ek,k = −λk2akbk eEx cos θ τ Ω τ Born ss δ(k − kF ) σ · k. (34) The effective scattering time τ Born ss is defined as 1 τ Born ss = niWkε+ 2A2 η(γ) = 1 2 (η0 − η1) 4 (cid:1)2 . sin2 γ 2 + rs (cid:0) sin γ (35) Magnetic impurity scattering is also part of J Born(fEk). The scattering term J Born (13) yields a driving term for SEk⊥, which gives a correction ∝ niτ and therefore independent of the impurity density mag (nk) evaluated in Eq. (33) Smag Ek⊥ = ak τ eE cos θ τmag 2Ωk δ(k − kF ) σ · k. (36) ss (f bare ss (fEk). In the term J 3rd Here 1/τmag = nmagJsk ¯U/(22A). We come to the skew scattering term beyond the Born approximation, J 3rd Ek ) only J 3rd ss (nbare Ek ) is real and finite, and we find that the pro- jection Pk J 3rd does not contribute to the AHE. We ss concentrate on the sum of the terms Pz,ef f J 3rd,A + Pz,ef f J 3rd,B ss . This scattering term will act on the electric-field induced correction to the scalar part of the density matrix, nEk, which has the angular struc- ture nEk = n(k) cos θ. In light of this detail, we can boil the term Pz,ef f J 3rd ss = Pz,ef f J 3rd to ss ss Pz,ef f J 3rd ss = λk2akn(k) 1 τ 3rd ss = niWk 2A ε+ A sin θ τ 3rd ss πrsφs z 4 (σ · zef f ) , (37) The next step comprises two parts. Firstly, we feed into J Born(fk) and J 3rd(fk), with the Ek + Sbare nbare Ek where the following auxiliary quantities have been de- fined φs z = 1 π Z dθZ dθ1 2π Z dθ2 2π F s z sin θ (2 cos θ2 − cos θ − cos θ1) 4 (cid:1)(cid:0) sin γk2 k 2 + rs 4 (cid:1)(cid:0) sin γk1 k2 2 + rs (cid:0) sin γkk1 2 + rs 4 (cid:1) F s z = sin γkk1 [(1 + cos γkk1 ) + cos γkk2 + cos γk1k2 ] + sin γk1 k2[(1 − cos γkk1) + cos γkk2 − cos γk1k2 ] + sin γk2k[(1 − cos γkk1 ) − cos γkk2 + cos γk1k2 ]. (38) The contribution of this scattering source term to the spin density matrix is Sss,3rd k,k = λk2akn(k) Ω sin θ τ 3rd ss (σ · k). (39) The term Sss,3rd found in Eq. (39) above does not con- tribute to the Hall current, but its disorder renormaliza- tion yields, k,k Sss,3rd Ek,k = −λk2a2 kbk eE sin θ 2Ωk τ 2 τ 3rd ss τc+ δ(k − kF ) σk,k. (40) This term contributes to the Hall current. It is the only skew scattering contribution to the anomalous Hall effect, and its value is determined explicitly below. 9 (f0k) con- Finally, the side-jump scattering term J Born E,kk′ +H sj sj tains both electric field interaction terms H sc It is ∝ [/(εF τ )]−1, in the time evolution operators. thus only its parallel projection yields a term of order [/(εF τ )]0. The side-jump scattering term consists of E (f0k)+ Jsc(f0k), defined in Eqs. (23) and (24) respec- J sj tively. Substituting for nEk we obtain for PkJ sj E (f0k), after a lot of algebra, E,kk′ PkJ sj E (f0k) = − πniebkλE ∂f0+ ∂ε+ σk,k Z d2k′ (2π)2Ukk′2 (1 + Ωk · Ωk′ ) (ky − k′ y) δ(ε′ + − ε+). The term Jsc(f0k) has two contributions: one from the spin-independent part of ¯Ukk′ and g0 to first order in λ πebλE Pk J 1 sc(f0k) = − (2π)2 Ukk′2 (ky − k′ and another from the spin-dependent part of ¯Ukk′ and g0 to zeroth order in λ (2π)2 Ukk′2 sin γkk′ (1 + Ωk′ · Ωk)(cid:20)E · σk,kZ d2k′ J 2 sc(f0k) = ∂f0+ ∂ε+ πnieλbk2 σk,k Pk 2 ∂f0k ∂εk Z d2k′ y) δ(εk′ − εk), ∂ ∂k (cid:0) Ωk · Ωk′(cid:1)(cid:21) δ(ε′ + − ε+). (41) (42) (43) by and J 2 sc(f0k) = Pk J 1 parts sc(f0k) + Integrating J sj E (f0k) yields an overall factor of Pk 2 in the parallel projection of the side-jump scattering term. The overall result for the side-jump scattering contribution is adding Pk SJsj Ek,k = − 4ebkλExε+ A2 sin θ δ(k − kF ) σk,k. (44) A further contribution due to side jump arises from the intrinsic side-jump driving term, i − [H sj E , f0k] = − eakEλky (f0+ − f0−) σ · k. (45) This term only contributes to the perpendicular driving term and yields a correction to SEk⊥, found straightfor- wardly evaluated by applying the time evolution opera- tor. We note in passing that this term is also responsible for the survival of side-jump spin-Hall effect in systems with band structure spin-orbit coupling40. In the AHE in TI it yields Ssj,int Ek,zef f = − eakExλk sin θ Ω f0+ σk,zef f . (46) This term is projected back onto Skk as described in Sec. IV B, giving an additional renormalization. Both Eqs. (44) and (46) contribute to the AHE. The current operator j has contributions from the σ × z, as well as jE = band Hamiltonian, j0 = eA 2e2λ τ τ 1 + + + e2 π 8 τc− τ τµ 8τ τµ τ Born ss σext yx = bF (λk2 τ 2 2τ 3rd τs(cid:19) + 2h (cid:20)1 − F )(cid:18)9 − + bF τ(cid:18) 1 σint yx = − e2 2h σ × E, and jU = 2ieλ σ × (k − k′)Ukk′ . These latter two cancel, as they represent the force acting on the system. The central result of our work is the AHE conductivity, which, ignoring terms of order b2 F , can be divided into two terms (int ≡ intrinsic, ext ≡ extrinsic) 2τmag(cid:21) ss τc+(cid:19). (47) All τ s are defined in the supplement, and 1/τmag ∝ bF , rendering this term negligible. Both Sbare and Sbare k,k contribute − e2 2h ) can also be ex- pressed in terms of the Berry curvature34, and is thus a topological quantity. It is similar to the result of Ref. 35 in the vicinity of a ferromagnetic layer, and can be identified with a monopole. The corrections are dis- order renormalizations due to scattering between Sk,⊥ and Sk,k. yx the first two terms in brackets rep- resent the combined contribution of side jump scatter- ing and intrinsic side jump plus their disorder renormal- izations, and the last two terms represent the contribu- tions due to skew scattering in the Born approximation and beyond it respectively. An important unprecedented finding of our study is that neither extrinsic spin-orbit coupling nor magnetic impurity scattering give a driv- ing term in the kinetic equation parallel to H0k (that is, yx . The total (− e2 4h to σint In σext k,zef f i (i.e. ∝ τ only) in the AHE response of TI. k Ωk), thus the AHE response does not contain a term of order [/(εF τ )]−1 due to extrinsic spin-orbit scatter- ing or magnetic impurity scattering. Such a term would have overwhelmed the intrinsic topological term in the ballistic limit. Since the intrinsic topological term is also of order [/(εF τ )]0, we conclude that there is no term ∝ n−1 In TI the spin and charge degrees of freedom are inher- ently coupled and the AHE current can also be viewed a steady-state in-plane spin polarization in a direction parallel to the electric field. The (bare) topological term has two equal contributions, which are part of the cor- rection to the density matrix orthogonal to the effective Zeeman field, therefore they represent an electric-field in- duced displacement of the spin in a direction transverse to its original direction. They are also obtained in the Heisenberg equation of motion if one takes into account the fact that k is changing adiabatically. Physically, the x-component of the effective Zeeman field Ωk is chang- ing adiabatically, and the out-of-plane spin component undergoes a small rotation about this new effective field. Consequently, each spin acquires a steady state compo- nent parallel to E, which in turn causes k to acquire a small component in the direction perpendicular to E. Elastic, pure momentum scattering [contained in J0(fk)] reduces this spin polarization because, in scattering from one point on the Fermi surface to another, the spin has to line up with a different effective field Ωk. The extra spin component of each electron is ∝ M , however the final result is independent of M , as the integrand con- tains a monopole located at the origin in k-space34,35. As M → 0 the effect disappears, since the correction to the orthogonal part of the density matrix vanishes. Finally, though the form of this term would hint that it is observ- able for infinitesimally small M , it was tacitly assumed that M exceeds the disorder and thermal broadening. Experimentally, for charged impurity scattering the figures depend on the Wigner-Seitz radius rs, represent- ing the ratio of the Coulomb interaction energy and the kinetic energy, with ǫr the relative permittivity. For Bi2Se3 with rs = 0.1436, the dominant term by far is 10 yx ≈ −0.53 (e2/2h) ≈ −e2/4h. We can consider also σint the (artificial) limit rs → 0, which implies εr → ∞, arti- ficially turning off the Coulomb interaction. As rs → 0, the prefactor of −e2/2h tends to 0.61, while at rs = 4, the limit of RPA in this case, it is ≈ 0.12. Interestingly, for short-range scattering, the AHE current changes sign, yx ≈ 0.18 (e2/2h). We note that λ is not known with σint for TI, yet that does not affect the central result of this work, since the intrinsic term is dominant. We have not taken into account the term in J 3rd linear in E, compa- ss rable in magnitude to the side jump. We recall that the result presented above represents the contribution from the conduction band, and in prin- ciple an extra e2/2h needs to be added to the total result to obtain the signal expected in experiment. Since the conduction band is expected to contribute ≈ −e2/4h, this does not make a difference in absolute terms. Com- parison with experiment must await further advances in materials growth, yet the most recent developments offer renewed grounds for hope.41 V. SUMMARY We have determined the AHE conductivity due to the surface conduction band of 3D TI, including all intrin- sic and extrinsic contributions. We have identified a skew scattering term in the Born approximation and one of third order in the scattering potential, as well as an intrinsic side-jump term and a side jump scat- tering term. The intrinsic topological term, renormal- ized by disorder, is dominant and of the order of e2/4h, though non-universal, and easily observable experimen- tally. This finding provides an unmistakable signature of surface transport in TI. DC acknowledges the partial support of the Chinese Academy of Sciences. We are grateful to N. A. Sinitsyn, Z. Fang, J. R. Shi, W. M. Liu, X. Dai, C. G. Zeng, Y. Q Li and A.-P. Li for enlightening discussions. 1 C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005). 2 M. Z. Hasan and C. L. Kane, arXiv:1002.3895 (unpub- lished). 3 X.-L. Qi and S.-C. Zhang, arXiv:1008.2026 (2010). 4 X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Phys. Rev. B 78, 195424 (2008). 5 H. Zhang, C.-X. Liu, X.-L. Qi, X. Dai, Z. Fang, and S.-C. Zhang, Nat. Phys. 5, 438 (2009). 6 D. Hsieh, D. Qian, L. Wray, Y. Xia, Y. S. Hor, R. J. Cava, and M. Z. Hasan, Nature 452, 970 (2008). 7 S. Urazhdin, D. Bilc, S. D. Mahanti, S. H. Tessmer, T. Kyratsi, and M. G. Kanatzidis, Phys. Rev. B 69, 085313 (2004). 8 Y. Xia, D. Qian, D. Hsieh, L. Wray, A. Pal, H. Lin, A. Ban- sil, D. Grauer, Y. S. Hor, R. J. Cava, et al., Nat. Phys. 5, 398 (2009). 9 J. Chen, H. J. Qin, F. Yang, J. Liu, T. Guan, F. M. Qu, G. H. Zhang, J. R. Shi, X. C. Xie, C. L. Yang, et al., Phys. Rev. Lett. 105, 176602 (2010). 10 Y. L. Chen, J. G. Analytis, J. H. Chu, Z. K. Liu, S. K. Mo, X. L. Qi, H. J. Zhang, D. H. Lu, X. Dai, Z. Fang, et al., Science 325, 178 (2009). 11 D. Hsieh, Y. Xia, D. Qian, L. Wray, J. H. Dil, F. Meier, J. Osterwalder, L. Patthey, A. V. Fedorov, H. Lin, et al., Phys. Rev. Lett. 103, 146401 (2009). 12 D. Xiao, Y. Yao, W. Feng, J. Wen, W. Zhu, X.-Q. Chen, G. M. Stocks, and Z. Zhang, Phys. Rev. Lett. 105, 096404 (2010). 13 R. Yu, W. Zhang, H. J. Zhang, S. C. Zhang, X. Dai, and 11 Z. Fang, Science 329, 61 (2010). 14 W.-K. Tse and A. H. MacDonald, Phys. Rev. Lett. 105, 057401 (2010). Dugaev, and J. Sinova, Phys. Rev. B 75, 045315 (2007). 29 N. A. Sinitsyn, J. Phys. Cond. Mat. 20, 023201 (2008). 30 S. Onoda, N. Sugimoto, and N. Nagaosa, Phys. Rev. B 77, 15 W.-K. Tse and A. H. MacDonald, Phys. Rev. B 82, 161104 165103 (2008). (2010). 31 S. A. Yang, H. Pan, Y. Yao, and Q. Niu, Phys. Rev. B 83, 16 K. Nomura and N. Nagaosa, Phys. Rev. B 82, 161401 125122 (2011). (2010). 17 T. Yokoyama, Y. Tanaka, and N. Nagaosa, Phys. Rev. B 32 E. I. Rashba, Fiz. Tech. Semicond. 42, 923 (2008). 33 G. Tkachov and E. M. Hankiewicz, Phys. Rev. B 83, 81, 121401 (2010). 155412 (2001). 18 R.-L. Chu, J. Shi, and S.-Q. Shen, arXiv:1007.0497 (un- 34 D. Culcer, A. H. MacDonald, and Q. Niu, Phys. Rev. B published). 19 L. A. Wray, Y. Xia, S.-Y. Xu, Y. Hor, R. Cava, A. Bansil, H. Lin, and M. Hasan, arXiv:1009.6216 (unpublished). 20 J.-J. Zhu, D.-X. Yao, S.-C. Zhang, and K. Chang, 1010.4134 (unpublished). 21 J. M. Luttinger, Phys. Rev. 112, 739 (1958). 22 J. Smit, Physica 24, 39 (1958). 23 L. Berger, Phys. Rev. B 2, 4559 (1970). 24 P. Nozi`eres and C. Lewiner, J. Phys. (Paris) 34, 901 (1973). 25 A. Crepieux and P. Bruno, Phys. Rev. B 64, 014416 (2001). 26 V. K. Dugaev, A. Crepieux, and P. Bruno, Phys. Rev. B 64, 104411 (2001). 27 A. A. Kovalev, Y. Tserkovnyak, K. Vyborny, and J. Sinova, Phys. Rev. B 79, 195129 (2009). 28 N. Sinitsyn, A. H. MacDonald, T. Jungwirth, V. K. 68, 045327 (2003). 35 J. Zang and N. Nagaosa, Phys. Rev. B 81, 245125 (2010). 36 D. Culcer, E. H. Hwang, T. D. Stanescu, and S. Das Sarma, Phys. Rev. B 82, 155457 (2010). 37 K. Nomura and N. Nagaosa, Phys. Rev. Lett. 106, 166802 (2011). 38 M.-T. Tran and K.-S. Kim, Phys. Rev. B 82, 155142 (2010). 39 J. Inoue, G. E. W. Bauer, and L. W. Molenkamp, Phys. Rev. B 70, 041303 (2004). 40 W.-K. Tse and S. Das Sarma, Phys. Rev. B 74, 245309 (2006). 41 D. Kim, S. Cho, N. P. Butch, P. Syers, K. Kirshenbaum, J. Paglione, and M. S. Fuhrer, arXiv:1105.1410 (unpub- lished).
1905.12876	1	1905	2019-05-30T06:47:36	Electron heating and mechanical properties of graphene	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	The heating of electrons in graphene by laser irradiation, and its effects on the lattice structure, are studied. Values for the temperature of the electron system in realistic situations are obtained. For sufficiently high electron temperatures, the occupancy of the states in the $\sigma$ band of graphene is modified. The strength of the carbon-carbon bonds changes, leading to the emergence of strains, and to buckling in suspended samples. While most applications of `strain engineering' in two dimensional materials focus on the effects of strains on electronic properties, the effect studied here leads to alterations of the structure induced by light. This novel optomechanical coupling can induce deflections in the order of $\sim 50$ nm in micron size samples.	cond-mat.mes-hall	cond-mat	Electron heating and mechanical properties of graphene Key Laboratory of Artificial Micro- and Nano-structures of Ministry of Education and School of Physics and Technology, Wuhan University, Wuhan 430072, China Jose Angel Silva-Guill´en∗ Fundaci´on IMDEA Nanociencia, C/Faraday 9, Campus Cantoblanco, 28049 Madrid, Spain Department of Physics and Astronomy, University of Manchester, Oxford Road, Manchester M13 9PL, UK and Donostia International Physics Center (DIPC) -- UPV/EHU, E-20018 San Sebasti´an, Spain Francisco Guinea The heating of electrons in graphene by laser irradiation, and its effects on the lattice structure, are studied. Values for the temperature of the electron system in realistic situations are obtained. For sufficiently high electron temperatures, the occupancy of the states in the σ band of graphene is modified. The strength of the carbon-carbon bonds changes, leading to the emergence of strains, and to buckling in suspended samples. While most applications of "strain engineering" in two dimensional materials focus on the effects of strains on electronic properties, the effect studied here leads to alterations of the structure induced by light. This novel optomechanical coupling can induce deflections in the order of ∼ 50 nm in micron size samples. PACS numbers: 31.15.A,81.05.ue,44.05.+e,68.60.Bs Introduction -- Graphene, and other two dimen- sional materials, show a unique coupling between the electronic and mechanical properties[1, 2]. As a result, electronic transport and optical transitions depend on the shape of the sample, and on strains which may be present. The term "strain engineering" is commonly used[3] to de- scribe techniques which use modifications of the strains in the system to induce desirable electronic properties. The inverse effect, the modification of structural proper- ties by changes in the electronic structure is hampered by the high mechanical stiffness of these materials. The low optical absorption of a graphene single layer is an additional difficulty, if the changes in the electronic dis- tribution are induced by light. Proposals for the creation of significant forces by optical means rely on non trivial combinations of graphene and dielectric layers[4, 5]. Un- usual effects of light on macroscopic, graphene like sys- tems has also been reported[6]. It is well documented that intense laser pulses lead to the excitation of high energy electron-hole pairs, and, ultimately, to an electron plasma in thermal equilib- rium at a temperature much higher than the lattice temperature[7 -- 12]. The cooling of a hot electron plasma is mediated by optical and acoustic phonons[13, 14]. The decoupling between electronic and lattice degrees of freedom has been studied extensively. The difference between electronic and sound velocities suppresses the phase space available for electron-phonon scattering[15, 16], and reduces the transfer of energy from electrons to phonons. This obstruction is partially relieved by co- herent processes involving phonons and impurities[15], named supercollisions, which have been observed in dif- ferent experiments[17 -- 20]. In order to analyze the influence of a high temperature electron gas on the structural properties of graphene we FIG. 1: Schematic drawing of the device used to heat up electrons in a graphene lattice. The substrate is depicted in gray and has a hole. A region of suspended graphene is illuminated by a laser beam (depicted in green) of radius R. consider the setup sketched in Fig. [1]. A region in a suspended graphene layer of radius R is illuminated by a laser. The graphene layer absorbs energy from the laser beam, at a rate defined by the laser power, W . We calcu- late the electron temperature once a steady state regime has been reached. The electron temperature modifies the occupancy of the graphene bands, which, in turn, changes the forces between atoms and induce strains and defor- mations in the lattice, irrespective of the lattice temper- ature, which is assumed to be much lower than the elec- tron temperature. The effect of the electron temperature on the lattice constant is obtained from a self consistent band structure calculation, using a Boltzmann distribu- tion for the occupancy of the electronic states. This cal- culation includes the non negligible coupling between the σ bands and the lattice parameter[21, 22]. For comple- tion, we include a brief discussion of the radiation pres- sure due to the momentum transfer from laser photons to the graphene layer in the Supplementary Information[23]. Electron and lattice temperature -- We assume that the laser energy is absorbed by the graphene layer 9 1 0 2 y a M 0 3 ] l l a h - s e m . t a m - d n o c [ 1 v 6 7 8 2 1 . 5 0 9 1 : v i X r a Rvia the creation of electron-hole pairs of energies com- parable to ν, where ν is the frequency of the laser. We study a suspended layer, and we do not need to take into account the degrees of freedom of a substrate. A steady state is reached through the combined effect of electron-electron interactions, the scattering between electrons and optical phonons, and heat diffusion, which transfers energy away from the region illuminated by the laser. The electron-electron interactions, which ther- malize the electron plasma, include plasmon emission processes[24]. The steady state is characterized by a tem- perature T (W, R) which describes both the electron gas and the optical phonons[25]. The temperature depends on the laser power, W , and the radius of the laser spot, R. The rate at which the energy flows away from the il- luminated area depends on the electron thermal conduc- tivity, κel, which is given by[14] 18ζ(3)k3 BT 2 π22v2 F a vF (cid:96)el, κel = e−h (1) where ζ(x) is the Riemann zeta function, a ≈ 3.5 A is the thickness of the graphene layer, kB is the Boltzmann constant, vF the Fermi velocity, and (cid:96)el is the electronic mean free path. At high temperatures, kBTe−h (cid:29) µ, where µ is the chemical potential, the value of (cid:96)el is given by (cid:96)el ≈ e2 kBTe−h ≈ vF kBTe−h . (2) Hence, cκk2 BTe−h a κel ≈ 18ζ(3) π2 k2 BTe−h a , = (3) where cκ ≈ 18ζ(3)/π2 is a dimensionless constant. This result, as well as the existence of a universal electrical conductivity, σ ∼ e2/, are a consequence of the fact that neutral graphene is a critical system. If we take values for the thermal conductivity for the lattice from previous works, that is κl ≈ 5000 W m·K [26, 27], Eq. 1 implies that κl (cid:29) κe−h, even for electron-hole tem- peratures Te−h (cid:38) 103K. Therefore, in the following, we can assume that the lattice dissipates heat rapidly, and remains in equilibrium with the external environment. The rate of heat flow from the electrons to the acous- tic modes can be divided into two contributions, one where the total momentum is conserved[13], and the other from supercollisions, which is mediated by elastic scattering[15]. The heat flow rates per unit area for these processes are ∂Q ∂t ∂Q ∂t ≈ D2(kBTdis)2(kBTe−h)3 ≈ D2(kBTe−h)5 (cid:12)(cid:12)(cid:12)(cid:12)ac1 (cid:12)(cid:12)(cid:12)(cid:12)ac1 5ρv4 5ρv6 F c2 (4) F , , 2 FIG. 2: Power dissipated by optical phonons over an area A = 1µm2 as function of electron temperature. where D is the deformation potential, ρ is the mass den- sity, and c is the sound velocity. We describe supercol- lisions in terms of an effective temperature related to elastic scattering, Tdis. We assume that elastic scatter- ing leads to a mobility independent of the carrier density. Then, (kBTdis)2 ≈ F · (vF /(cid:96)elastic is the ∼ 100 nm, el electronic elastic mean free path. For (cid:96)elastic Tdis ∼ 102 K. ), where (cid:96)elastic el el The rate of heat flow to the optical modes is (cid:12)(cid:12)(cid:12)(cid:12)op ∂Q ∂t ≈ αop (kBTe−h)3 (vF )4 F , (5) (cid:18) ωop (cid:19) kBTe−h where ωop is an average optical phonon frequency, and the parameter αop and the function F are described in the Supplementary Information (see also Ref. [8, 28]). The power dissipated to optical modes in an area A = 1µm2 as function of electron temperature is plotted in Fig. [2]. By comparing Eqs. (4) and (5), we conclude that the energy transfer from high temperature electrons to opti- cal phonons is much larger than the energy transfer to acoustic phonons, for physically accessible temperatures, Te−h (cid:46) 104 K. Therefore, in the following, we can con- sider only the role of the optical phonons. The fraction of the laser power, W , absorbed by the electron-hole pairs in the graphene layer is παW , where α ≈ 1/137 is the fine structure constant, and πα is the optical absorption of a graphene layer[29]. In order to ob- tain the steady state temperature of the electron plasma, we take into account the energy dissipated away from the laser spot, which depends on the electron heat conduc- tivity, and the transfer of energy from electrons to optical 020040060080010001200T(K)0.000.020.040.06W(mWatt)3 FIG. 4: Equilibrium lattice constant (blue) and gap between the σ bands (red) vs. electronic temperature. between plasma temperature and laser power when the laser is focused on a region of radius R = 1µm. The op- tical phonon absorption dominates for W (cid:38) 0.2 mWatt. For this laser power, the electron plasma reaches a tem- perature T ≈ 1300 K. For a power W ∼ 5 mWatt, in the regime dominated by optical phonons, the electron temperature is T ≈ 2250 K. Outside the region illuminated by the laser, electronic thermal conduction will bring the electron-hole plasma to equilibrium with the external environment. From Eq. (6) we can define a length scale eq = πR2 cκ(kBT )2 (cid:96)2 . (8) For T ∼ 2200 K and W ∼ 5 mWatt, we obtain (cid:96)eq ∼ 600 nm. αW Effect of the electronic temperature on the graphene lattice -- As we have just seen, it is pos- sible to change the temperature of electrons in graphene without modifying the actual temperature of its lattice. Now, we center our attention on the possible conse- quences that the change of the electronic temperature has on the lattice of graphene. For that, we carried out first-principles calculations. These were performed us- ing a numerical atomic orbitals approach to DFT,[30, 31] which was developed for efficient calculations in large sys- tems and implemented in the Siesta code.[32, 33] We have used the generalized gradient approximation (GGA) and, in particular, the functional of Perdew, Burke and Ernzerhof.[34] Only the valence electrons are considered in the calculation, with the core being replaced by norm- conserving scalar relativistic pseudopotentials [35] factor- ized in the Kleinman-Bylander form.[36] The non-linear core-valence exchange-correlation scheme [37] was used for all elements. We have used a split-valence triple-ζ basis set including polarization functions.[38] The energy cutoff of the real space integration mesh was set to 1000 Ry. To build the charge density (and, from this, ob- FIG. 3: Temperature of the electron-hole plasma versus laser power for a laser spot of radius R = 1µm. Red: Dissipation by electronic heat diffusion. Blue: Dissipation by optical phonon absorption. Black: total dissipation. For a laser power of W = 5 mWatt, the calculated temperature is T = 2250 K. phonons. We obtain k2 B ∂r[T (r)∂rT (r)] + cκ = αop[kBT (r)]3 (cid:26) αW F (vF )4 πR2 r ≤ R 0 R < r (cid:18)ωop (cid:19) kBT = , (6) where cκ was defined in Eq. (3). As mentioned previ- ously, we assume that the laser has power W , and it irradiates uniformly a circular spot of radius R. Quali- tatively, the two terms in Eq. (6) allow us to define two cooling regimes: - For low values of R, or large values of W , the dissipa- tion is dominated by electronic thermal conduction into the non irradiated region, r > R, described by the first term in Eq. (6). - If W is sufficiently low, or R is large enough, dissipa- tion is mostly the local transfer of heat to optical modes, given by the second term in Eq. (6). The values, W∗ and R∗, which define the crossover between these regimes takes place approximately, are αW∗R4 ∗ ∼ (vF )8 3 , α2 op (7) where we have replaced F[(ωop)/(kBT )] by its con- stant value in the limit (ωop)/(kBT ) → ∞ (see Supple- mentary Information for more details). A more precise determination can be obtained from computing the electron temperature as function of W and R considering only one relaxation mechanism. The crossover between the two regimes takes place when the two temperatures are similar. Fig. [3] shows the relation 05001000150020002500T(K)0246W(mWatt)TotalElectronicheatdiffusionOpticalphononabsorption0.00.51.01.52.0T(eV)2.502.522.542.56Eq.latt.const.(A)10.810.911.011.111.2∆σc−σv(eV)4 FIG. 5: Profile of a disk shaped suspended graphene flake with clamped edges of radius R = 5000 nm with a thermally generated stress of = 0.01%. FIG. 6: As in Fig.[3] for a graphene bilayer. The calculated temperature is T ≈ 4570 K. tain the DFT total energy and atomic forces), the Bril- louin zone (BZ) was sampled with the Monkhorst-Pack scheme[39] using grids of (60×60×1) k-points. To simu- late the effect of increasing the electronic temperature of graphene, we changed the electronic temperature of the Fermi-Dirac (FD) distribution of the electrons. It is im- portant to note that, once a finite temperature has been chosen, the relevant energy is not the Kohn-Sham (KS) energy, but the Free energy since the atomic forces are derivatives of this.[40 -- 42] The change of the lattice con- stant with electronic temperature is shown in Fig. Fig. 4, where we can see that with increasing temperature, the lattice constant becomes larger. This result can be understood looking at the effect that the electronic temperature has on the σ bands which are the responsible for the bonds in graphene and, therefore, its lattice constant. Looking at Fig. 4, we can see that changing the electronic temperature slightly changes the population of these bands. As a result, the σ bonds will be become weaker, and the lattice will expand. Electronic temperature and strains -- We have just seen that the temperature of the electron-hole plasma can modify the interatomic forces and the lo- cal lattice constant. Hence, strains are induced in the graphene layer when shining a laser beam to a graphene layer. The results in the preceding paragraph suggest that a lattice expansion of order = ∆(cid:96)/(cid:96) ∼ 0.01% is pos- sible when the temperature of the electron plasma is T ∼ 1000− 2000 K. In a suspended system with clamped edges (see Fig. 1), such an expansion will make the sheet to buckle. A simple calculation, using the techniques de- veloped for graphene bubbles in Ref. [43], and for a cir- cular region of radius R = 5µm gives the profile shown in Fig.[5]. Note that the average strain is ≈ h2 max/R2. Generalization to multilayer graphene. -- A graphene bilayer. The rate of heat transfer in a graphene bilayer from the electron-hole plasma to optical modes is calculated in the Supplementary Information. The main change is a suppression in the heat transfer, due to in- terference effects in the electron-phonon matrix element, partially compensated by an increase in the electronic density of states. Note that, for a given power, the ab- sorption is α ≈ 2π/137, twice the absorption of a mono- layer. Results are shown in Fig. [6]. The temperature of the electron-hole plasma, for a given laser power, is significantly increased in bilayer graphene. Graphene stacks with more than two layers. In a sys- tem with N layers, the absorbed energy per unit time is ∂Q/∂t = N αW , distributed over the N layers, so that In or- the energy absorbed per layer does not change. der to estimate the electronic thermal conductivity, we make use of the fact that the low energy band structure of multilayered graphene can be described as a combina- tion of quadratic and linear bands touching at the Dirac point[44]. The resulting electronic thermal conductivity is determined by the quadratic bands. Its temperature BT 2τN (T )]/(2a), where we assume that, at high temperatures, the scattering time, τN (T ), is determined by electron-electron interactions. These interactions couple with similar strength an elec- tron in a given layer and electron-hole pairs in any layer. Combining this result with the criticality provided by the band touching, we obtain τN (T ) ∼ /(N kBT ), so that κe−h(N ) ∼ (k2 BT )/(Na). Finally, we obtain that the electron temperature scales as T (N ) ∝ √ dependence is κe−h(N ) ≈ [k3 N . Conclusions -- We have estimated the electron temperature in a graphene layer under laser irradiation. The temperature is determined by a balance between the power input from the laser, the transfer of energy to op- tical phonons, and the conduction of heat away from the 010002000300040005000r(nm)010203040h(nm)0100020003000400050006000T(K)02468W(mWatt)TotalElectronicheatdiffusionOpticalphononabsorptionirradiated region. Temperatures in the order of 1000- 2000 K can be reached for a laser power W ≈ 5 mWatt in regions of radius R ≈ 1µm. Similar, or higher temper- atures can be expected in multilayer stacks. The weak coupling between electrons and acoustic phonons, and the large heat conductivity of these phonons imply that the lattice temperature changes only slightly. The electron temperature leads to changes in the lat- tice constant of graphene, even if the lattice temper- ature does not vary. We find that strains of order ≈ 0.01 − 0.02% are likely. These strains can induce a significant buckling in a suspended sample. Our anal- ysis suggests that light can be used to modify the struc- tural properties of graphene and other two dimensional materials. Acknowledgments -- This work was supported by funding from the European Commission under the Graphene Flagship, contract CNECTICT-604391. Nu- merical calculations presented in this paper have been performed on a supercomputing system in the Supercom- puting Center of Wuhan University. We thank Thomas Frederiksen for fruitful discussions. ∗ Electronic address: [email protected] [1] B. Amorim, A. Cortijo, F. de Juan, A. G. Grushin, F. Guinea, A. Guti´errez-Rubio, H. Ochoa, V. Parente, R. Rold´an, P. San-Jose, et al., Physics Reports 617, 1 (2016). [2] R. Rold´an, L. Chirolli, E. Prada, J. A. Silva-Guill´en, P. San-Jose, and F. Guinea, Chemical Society Reviews 46, 4387 (2017). [3] V. M. Pereira and A. H. Castro Neto, Phys. Rev. Lett. 103, 046801 (2009). [4] S. H. Mousavi, P. T. Rakich, and Z. Wang, ACS photon- 5 T. Taniguchi, et al., Nature Physics 12, 455 (2016). [13] R. Bistritzer and A. H. MacDonald, Phys. Rev. Lett. 102, 206410 (2009). [14] W.-K. Tse and S. Das Sarma, Phys. Rev. B 79, 235406 (2009). [15] J. C. W. Song, M. Y. Reizer, and L. S. Levitov, Phys. Rev. Lett. 109, 106602 (2012). [16] J. C. W. Song and L. S. Levitov, J. Phys.: Condens. Matter 27, 164201 (2015). [17] N. M. Gabor, J. C. W. Song, Q. Ma, N. L. Nair, T. Tay- chatanapat, K. Watanabe, T. Taniguchi, L. S. Levitov, and P. Jarillo-Herrero, Science 334, 648 (2011). [18] K. J. Tielrooij, J. C. W. Song, S. A. Jensen, A. Centeno, A. Pesquera, A. Zurutuza Elorza, M. Bonn, L. S. Levitov, and F. H. L. Koppens, Nature Phys. 9, 248 (2013). [19] J. C. W. Song, K. J. Tielrooij, F. H. L. Koppens, and L. S. Levitov, Phys. Rev. B 87, 155429 (2013). [20] K. J. Tielrooij, L. Piatkowski, M. Massicotte, A. Woess- ner, Q. Ma, Y. Lee, K. S. Myhro, C. N. Lau, P. Jarillo- Herrero, N. F. van Hulst, et al., Nature Nanotech. 10, 437 (2015). [21] F. Mazzola, J. W. Wells, R. Yakimova, S. Ulstrup, J. A. Miwa, R. Balog, M. Bianchi, M. Leandersson, J. Adell, P. Hofmann, et al., Phys. Rev. Lett. 111, 216806 (2013). [22] F. Mazzola, T. Frederiksen, T. Balasubramanian, P. Hof- mann, B. Hellsing, and J. W. Wells, Phys. Rev. B 95, 075430 (2017). [23] See Supplementary Information. [24] J. M. Hamm, A. F. Page, J. Bravo-Abad, F. J. Garcia- Vidal, and O. Hess, Phys. Rev. B 93, 041408 (2016). [25] Note that, as we are considering a suspended system, we do not include substrate modes[45]. [26] A. A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Tewelde- brhan, F. Miao, and C. N. Lau, Nano Lett. 8, 902 (2008). [27] A. A. Baladin, Nature Materials 10, 569 (2011). [28] D. Yadav, M. Trushin, and F. Pauly, Phys. Rev. B 99, 155410 (2019). [29] R. R. Nair, P. Blake, A. N. Grigorenko, K. S. Novoselov, T. J. Booth, T. Stauber, N. M. R. Peres, and A. K. Geim, Science 320, 1308 (2008). [30] P. Hohenberg and W. Kohn, Physical Review 136, B864 ics 1, 1107 (2014). (1964). [5] M. M. Salary, S. Inampudi, K. Zhang, E. B. Tadmor, and [31] W. Kohn and L. J. Sham, Physical Review 140, A1133 H. Mosallaei, Phys. Rev. B 94, 235403 (2016). (1965). [6] T. Zhang, H. Chang, Y. Wu, P. Xiao, N. Yi, Y. Lu, Y. Ma, Y. Huang, K. Zhao, X.-Q. Yan, et al., Nature Photonics 9, 471 (2015). [7] T. Winzer, A. Knorr, and E. Malic, Nano Letters 10, 4839 (2010), 1008.1904. [8] C. H. Lui, K. F. Mak, J. Shan, and T. F. Heinz, Phys. Rev. Lett. 105, 127404 (2010). [9] J. C. Johannsen, S. Ulstrup, F. Cilento, A. Crepaldi, M. Zacchigna, C. Cacho, I. C. E. Turcu, E. Springate, F. Fromm, C. Raidel, et al., Phys. Rev. Lett. 111, 027403 (2013). [10] I. Gierz, J. C. Petersen, M. Mitrano, C. Cacho, E. Turcu, E. Springate, A. Stohr, A. Kohler, U. Starke, and A. Cav- alleri, Nature Materials 12, 1119 (2013). [11] J. C. Johannsen, S. Ulstrup, A. Crepaldi, F. Cilento, M. Zacchigna, J. A. Miwa, C. Cacho, v. Chapman, E. Springate, F. Fromm, et al., Nano Lett. 15, 326 (2015). [12] Q. Ma, T. I. Andersen, N. L. Nair, N. M. Gabor, M. Mas- sicotte, C. H. Lui, A. F. Young, W. Fang, K. Watanabe, [32] J. M. Soler, E. Artacho, J. D. Gale, A. Garc´ıa, J. Jun- quera, P. Ordej´on, and D. S´anchez-Portal, Journal of Physics: Condensed Matter 14, 2745 (2002). [33] E. Artacho, E. Anglada, O. Di´eguez, J. D. Gale, A. Garc´ıa, J. Junquera, R. M. Martin, P. Ordej´on, J. M. Pruneda, D. S´anchez-Portal, et al., Journal of Physics: Condensed Matter 20, 064208 (2008). [34] J. P. Perdew, K. Burke, and M. Ernzerhof, Physical Re- view Letters 77, 3865 (1996). [35] N. Troullier and J. L. Martins, Physical Review B 43, 1993 (1991). [36] L. Kleinman and D. M. Bylander, Physical Review Let- ters 48, 1425 (1982). [37] S. G. Louie, S. Froyen, and M. L. Cohen, Physical Review B 26, 1738 (1982). [38] E. Artacho, D. S´anchez-Portal, P. Ordej´on, A. Garc´ıa, and J. M. Soler, Physica Status Solidi (b) 215, 809 (1999). [39] H. J. Monkhorst and J. D. Pack, Physical Review B 13, 5188 (1976). [40] R. M. Wentzcovitch, J. L. Martins, and P. B. Allen, Phys. and I. V. Grigorieva, Nature Comm. 7, 12587 (2016). Rev. B 45, 11372 (1992). [44] F. Guinea, A. H. Castro Neto, and N. M. R. Peres, Phys. [41] S. de Gironcoli, Phys. Rev. B 51, 6773 (1995). [42] G. Kresse and J. Furthmller, Computational Materials Science 6, 15 (1996). [43] E. Khestanova, F. Guinea, L. Fumagalli, A. K. Geim, Rev. B 73, 245426 (2006). [45] T. Low, V. Perebeinos, R. Kim, M. Freitag, and P. Avouris, Phys. Rev. B 86, 045413 (2012). 6
1501.07599	1	1501	2015-01-28T21:59:59	Spin Seebeck devices using local on-chip heating	[ "cond-mat.mes-hall" ]	A micro-patterned spin Seebeck device is fabricated using an on-chip heater. Current is driven through a Au heater layer electrically isolated from a bilayer consisting of Fe$_3$O$_4$ (insulating ferrimagnet) and a spin detector layer. It is shown that through this method it is possible to measure the longitudinal spin Seebeck effect (SSE) for small area magnetic devices, equivalent to traditional macroscopic SSE experiments. Using a lock-in detection technique it is possible to more sensitively characterize both the SSE and the anomalous Nernst effect (ANE), as well as the inverse spin Hall effect in various spin detector materials. By using the spin detector layer as a thermometer, we can obtain a value for the temperature gradient across the device. These results are well matched to values obtained through electromagnetic/thermal modeling of the device structure and with large area spin Seebeck measurements.	cond-mat.mes-hall	cond-mat	a Spin Seebeck devices using local on-chip heating Stephen M. Wu,1, a) Frank Y. Fradin,1 Jason Hoffman,1 Axel Hoffmann,1 and Anand Bhattacharya1 Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439, USA (Dated: 9 September 2018) A micro-patterned spin Seebeck device is fabricated using an on-chip heater. Current is driven through a Au heater layer electrically isolated from a bilayer consisting of Fe3O4 (insulating ferrimagnet) and a spin detector layer. It is shown that through this method it is possible to measure the longitudinal spin Seebeck effect (SSE) for small area magnetic devices, equivalent to traditional macroscopic SSE experiments. Using a lock-in detection technique it is possible to more sensitively characterize both the SSE and the anomalous Nernst effect (ANE), as well as the inverse spin Hall effect in various spin detector materials. By using the spin detector layer as a thermometer, we can obtain a value for the temperature gradient across the device. These results are well matched to values obtained through electromagnetic/thermal modeling of the device structure and with large area spin Seebeck measurements. The spin Seebeck effect (SSE) has been widely stud- ied due to the implications it has on the generation of pure spin currents1,2. In the SSE, applying a thermal gradient across a magnetic insulator generates a pure spin current that flows into an adjacent material with- out any charge current3. Experiments involving the SSE have taken many forms, but the canonical method has been to use large area ferromagnetic insulators on the order of several millimeters combined with Peltier ele- ments to generate a thermal gradient. There are several disadvantages to this technique including being sensitive to material non-uniformity, being confined to using large samples, and being limited by complicated experimental setups. Here we introduce a micro-patterned spin Seebeck de- vice to solve many of these issues. We measure a mi- croscale SSE using confined local on-chip heating, which is electrically separated from both the spin current source and spin current detector. Using this method it is possi- ble to sensitively examine local magnetization, local spin current, and local spin to charge conversion in a sim- ple and scalable device structure. Since the devices are small, it is possible to easily integrate them into con- ventional cryostat systems. While there are alternative small-area spin Seebeck techniques such as laser heating4 or current induced heating using the spin detector layer5, each has its disadvantages. Laser heating involves addi- tional experimental setup, making it harder to integrate into existing device measurement setups, and current in- duced heating using the spin detector layer generates sev- eral undesired conventional charge transport effects that must be separated to extract the SSE. In this work, our devices consist of Fe3O4 (60 nm)/X/MgO (100 nm)/Au (100 nm) on MgAl2O4 sub- strates, where X is either Ti (15 nm), Ti (1.5 nm)/Pt (5 nm) or absent (Fig. 1a). On-chip heating is provided by the Au layer, which is electrically isolated from the rest of the device by a layer of MgO. Fe3O4 serves as the spin a)Electronic mail: [email protected] current source, while X serves as a spin detector layer that responds to spin current through the inverse spin Hall effect (ISHE). Fe3O4 is grown on MgAl2O4 (100) substrates using ozone assisted oxide molecular beam epitaxy (MBE), de- scribed elsewhere6. The film is then patterned into 10 µm x 800 µm strips through standard photolithography and liquid nitrogen cooled Ar+ ion milling. Liquid ni- trogen cooling is used to limit defects introduced by the ion milling process. Using a lift-off process with standard photolithography and electron beam evaporation, a layer 7489:# !'# 6# ;# 45/6+# !"# $%&# '()*+#,-#'(# ./0&1# $%!23&1# 9# FIG. 1. A schematic of a typical spin Seebeck device is pre- sented in (a). A model of the actual device structure used in experiments and simulations is presented in (b). The pillars represent wirebonds made to the device, the wirebonds on the second set of pads are hidden for viewing clarity. of Ti/Pt or Pt is then deposited onto the Fe3O4 strip. Fi- nally, using the same lift-off process the MgO/Au heater layer is patterned onto the device, with separate contacts to the side. The finished device is shown in a 3D model in Fig. 1b. The response in the spin current detector layer is ~EISHE ∝ ~JS × σ, where ~JS is the injected spin cur- rent, and σ is the unit vector of the spin. Because the spin current injected into the adjacent spin detector layer is directly proportional to the thermal gradient and the magnetization, the total response is ~ESSE ∝ ∇T × ~M . We are not limited to using DC techniques to detect the ISHE signal from the spin detector layer. By using AC lock-in detection, the sensitivity of the measurement can be greatly increased. Since the power generated through Joule heating P = I 2 heater R ∝ ∇T , the voltage measured will also be ∝ I 2 heater. If an AC signal is used to excite the heater Iheater ∝ sin(ωt), then the measured voltage V ∝ 1 2 (1 − cos(2ωt)). By lock-in detecting the 90◦ out- of-phase component at the 2ω frequency it is possible to detect the spin Seebeck effect with much higher sensitiv- ity because parasitic effects occur at the ω frequency. By using this technique and ignoring the constant term, it is possible to detect signals as small as 2 nV depending on the integration period of the lock-in amplifier. The sample is mounted onto a standard circuit board using silver paint, and contact to the device is made us- ing wirebonds before it is loaded into a Quantum De- sign Physical Properties Measurement System (PPMS) cryostat for temperature and magnetic field control. Ex- periments performed on three different device stacks are presented in Fig. 2, each with a different spin current de- tector layer (Ti, Ti/Pt, or Fe3O4 alone). In each exper- iment a constant 5 Vpp signal at 99 Hz is applied across the Au heater and a 60 ohm load resistor. Since the re- sistance of the heater layer changes with temperature, the power applied using this measurement also changes from 17.2 mWrms at 300 K to 21.9 mWrms at 15 K. At each temperature the voltage across the device is mea- sured with respect to an in-plane magnetic field applied transverse to ~ESSE. A typical curve is presented in the inset of Fig. 2. The magnitude of the voltage response, as defined as the voltage difference at zero applied mag- netic field for the two different magnetization states of Fe3O4,is shown against temperature in Fig. 2. Fe3O4 is only insulating below a well known metal- insulator transition at 120 K known as the Verwey transi- tion. At temperatures above the Verwey transition there is a contribution from both the anomalous Nernst effect from conducting Fe3O4, and the ISHE due to the SSE in Ti or Ti/Pt. At these temperatures it is not possible to separate the two effects7. However, below the Verwey transition ,there is a recovery in both the Ti/Pt and the Ti devices due to the SSE, while the signal in the Fe3O4 control device goes to zero. The 1.5 nm Ti spacer layer in between the Fe3O4 and the Pt in the Ti/Pt device serves to remove any effects from proximity magnetism8 -- 10. The large difference in magnitudes between the Fe3O4 control Fe3O4/Ti/Pt Fe3O4/Ti Fe3O4 -2 0 Magnetic Field (kOe) 2 0.4 0.2 0.0 V ( V -0.2 -0.4 1.5 1.2 0.9 0.6 0.3 0.0 ) V ( V 2 25 20 15 10 5 0 0 50 150 100 Temperature (K) 200 250 300 FIG. 2. The magnitude of the spin Seebeck/anomalous Nernst response with respect to temperature. Devices with spin de- tector layers of Ti (15 nm), Ti(1.5 nm)/Pt (5 nm), and a control device with Fe3O4 are presented. The inset curve is an example of a spin Seebeck signal with respect to magnetic field for a Ti/Pt device at 15 K. The linear contribution to this voltage response is likely due to the ordinary Nernst effect from the spin detector materials. device and devices with spin detector layers are due to the large difference in resistivities between the different films. The low resistance spin detector layers essentially act as a shunt that reduces the total voltage response7,11. At the same time, the difference in the magnitudes between the Ti and the Ti/Pt devices is explained by the individual spin Hall angles of the two materials. The spin Hall an- gle in Ti is small and negative, as predicted theoretically through tight binding calculations12 and experimentally observed13, while Pt has been the long standing canoni- cal standard for ISHE experiments due to its large ISHE response14. Our measurements reproduce these results well, showing that our method is equivalent to large area macroscopic measurements. To look further in detail into the heat flow in our de- vices, thermal simulations were performed on the device model presented in Fig. 1b, using the Computer Simu- lation Technology (CST) Studio Suite software modeling package. Values for material resistivity were taken from measured values, while values for thermal conductivity and heat capacity were taken from standard sources in literature15,16. Since the device sits on a large substrate that is mounted using silver paint onto a standard PPMS rotator circuit board solidly heat sinked to the PPMS, the back side of the substrate is assumed to be held at the bath temperature. The same condition is assumed of the four wirebond contacts to the device, since each wirebond is also connected directly to the circuit board. The DC heater current chosen for this simulation was equivalent to the maximum peak current in our AC measurements, 24.5 mA at 300 K and 30.1 mA at 15 K. The resulting heater power loss density is presented in Fig. 3a for a Ti/Pt device at 300 K. This shows that all the heater power is constrained to the narrowest and most resistive a b 3 Using thermal simulation it is possible to model the out-of-plane thermal gradient across each thin film in the center of our device (Fig. 4). The thermal gradient across just the Fe3O4 layer is 0.298 K/µm at 300 K and 0.014 K/µm at 15 K. This difference is consistent with the rel- ative magnitudes of the thermal conductivity of Fe3O4 at 300 and 15 K. The inset of Fig. 4 shows that the much larger temperature drop is across the substrate due to its size relative to the thin film (550 µm vs. 0.06 µm). It also shows that the temperature drop across the sub- strate is non-uniform due to the heating being localized to the device and not the entire top surface of the chip. Locally at the surface, on the scale of our thin films, the temperature distribution is highly linear. FIG. 3. Electromagnetic and thermal simulation results for a Ti/Pt spin Seebeck device. (a) shows the calculated power loss density due to Joule heating in the heater layer, while (b) shows the calculated local temperature within the device that results from heating. part of the device where the Fe3O4/Ti/Pt device stack is. The resulting temperature within the device is presented in Fig 3b. The ∆T from the back of the substrate to the top of the heater is 2.226 K at 300 K, and 0.565 K at 15 K. Given this ∆T, the voltage response measured in our devices match well with large area SSE experiments on both thin film and bulk ferromagnets11,17. Since the large area experiments show that there is a linear voltage response to thermal gradient up to ∆T= 20 K, by opti- mizing the power transferred to our heater our signal to noise ratio could be increased even further by increasing ∆T in our system. The averaged ∆T along the device length, shows the uniformity of the thermal gradient is within 3% according to simulations. Although there is a large ∆T across the entire sub- strate and device, the temperature drop across just the Fe3O4 layer is not as large and depends its individual thermal conductivity. Characterizing ∆T across thin film ferromagnets has been an ongoing challenge in SSE ex- periments both on the macroscale and the microscale since there is no standard method to probe the tem- perature at both sides of the thin film7. By applying a constant ∆T across the substrate and the thin film fer- romagnet there is no guarantee that the thermal gradient across the thin film will be constant with respect to tem- perature since the thermal conductivities of the substrate and the thin film will change relative to each other. By using an on-chip heater, we effectively send a constant heat current through the device instead of setting up a constant ∆T. This is analogous to performing a current biased vs. voltage biased electrical measurement. Using this method it may be easier to eliminate substrate based effects since the thermal gradient across the thin film fer- romagnet will only depend on the properties of the thin film and not its relative value compared to the substrate. 302.23 302.22 302.21 302.20 ) K t ( e r u a r e p m e T 302.19 -0.1 Pt MgO Au MgAl2O4 Fe3O4 Ti 302 301 300 -500 -250 0 0.0 0.2 Distance from substrate surface ( m) 0.1 FIG. 4. Temperature in the out-of-plane direction at the cen- ter of a Ti/Pt spin Seebeck device as calculated from thermal modeling. The inset shows the same temperature distribution at an expanded range through the entire substrate. Using the metallic spin detector layer in our device as a thermometer, it is possible to compare the results of thermal modeling with measured values for ∆T. By characterizing the resistance of the Ti (15 nm) strip in our Fe3O4/Ti spin Seebeck device we have a one-to-one relation between temperature and resistance. By running an AC signal through our on-chip heater and measuring the change in resistance in the Ti strip it is possible to directly measure the change in temperature due to the heater. The resistance of the Ti strip and its derivative with respect to temperature is measured and shown in Fig. 5a-b. Next, the same 99 Hz 5 Vpp signal is sent to the heater through a 60 ohm load like in the experiment presented in Fig. 2. The peak to peak amplitude of the change in resistance of the Ti strip is presented in Fig 5c. Finally, the change in temperature is derived from ∆T = ∆R dR/dT (Fig. 5d). Measurements of the thermal time constant of our system revealed two time scales for changes in the resistance of Ti strip at all temperatures. One of which occurs faster than our data acquisition sys- tem can resolve (5000 samples/s), and one of which oc- curs on the scale of seconds. The longer time constant is likely due to heating the entire experimental probe within ) k ( R ) K / ( T d R d / ) ( R ) K ( T 5.5 5.0 4.5 4.0 3.5 9 6 3 0 4 2 0 1.2 0.6 0.0 a b c d 0 50 100 150 200 250 300 350 Temperature (K) The measured resistance vs. FIG. 5. temperature of Fe3O4/Ti strip is presented in (a), along with its derivative with respect to temperature in (b). (c) shows the change in resistance of the Fe3O4/Ti strip upon heating with a constant 5 Vpp applied across a 60 ohm load resistor and the heater layer. Using the results from (b) and (c), the temperature difference across the device is calculated with respect to tem- perature and presented in (d). the cryostat along with the sample. At 99 Hz the only contribution we measure is the short time constant ∆T across just the device. The values at both low and high temperatures show values close to the predicted results of thermal modeling. The remaining differences are likely due to the differences between the values of thermal con- ductivity taken from literature and our samples. Since our measurements are heat current biased, the results in Fig. 5d scale inversely to measurements of thermal conductivity on MgAl2O4, which seems to be the largest determining factor of ∆T in the device. There is a local increase in ∆T near the Verwey transition in Fig. 5d that cannot be directly explained by thermal conductiv- ity changes in MgAl2O4 or Fe3O4 single crystals15 alone, but does resemble the discontinuity in Fe3O4 heat capac- ity at this temperature16. We have introduced a method to measure the spin See- beck effect using a micropatterned device with an on-chip heater. By using a small scale device it is possible to sen- sitively measure local magnetization, local spin current, and local spin to charge conversion on the microscale and potentially the nanoscale. By using the spin current de- 4 tection layer as a thermometer it is possible to extract a equivalent thermal gradient to the constant ∆T measure- ments in conventional spin Seebeck experiments. These measurements match well with thermal simulations of our device structure. This type of device structure allows for easier access to lower temperature, higher magnetic field, and smaller magnetic materials all of which have been challenging to explore using other methods, which require more custom experimental setup. Through the exploration in these regimes it may be possible to pro- vide further insight into the basic mechanisms behind the SSE, as well as potentially finding other interesting thermal spin transport phenomenon. ACKNOWLEDGMENTS All authors acknowledges support of the U.S. Depart- ment of Energy (DOE), Office of Science, Basic Energy Sciences (BES), Materials Sciences and Engineering Di- vision. The use of facilities at the Center for Nanoscale Materials, was supported by the U.S. DOE, BES under contract No. DE-AC02-06CH11357. 1K. Uchida, S. Takahashi, K. Harii, J. Ieda, W. Koshibae, K. Ando, S. Maekawa, and E. Saitoh, Nature 455, 778 (2008). 2K. Uchida, H. Adachi, T. Ota, H. Nakayama, S. Maekawa, and E. Saitoh, Appl. Phys. Lett. 97, 172505 (2010). 3K. Uchida, J. Xiao, H. Adachi, J. Ohe, S. Takahashi, J. Ieda, T. Ota, Y. Kajiwara, H. Umezawa, H. Kawai, et al., Nature Mater. 9, 894 (2010). 4M. Weiler, M. Althammer, F. D. Czeschka, H. Huebl, M. S. Wag- ner, M. Opel, I.-M. Imort, G. Reiss, A. Thomas, R. Gross, and S. T. B. Goennenwein, Phys. Rev. Lett. 108, 106602 (2012). 5M. Schreier, N. Roschewsky, E. Dobler, S. Meyer, H. Huebl, R. Gross, and S. T. Goennenwein, Appl. Phys. Lett. 103, 242404 (2013). 6M. Liu, J. Hoffman, J. Wang, J. Zhang, B. Nelson-Cheeseman, and A. Bhattacharya, Sci. Rep. 3 (2013). 7S. M. Wu, J. Hoffman, J. E. Pearson, and A. Bhattacharya, Appl. Phys. Lett. 105, 092409 (2014). 8S.-Y. Huang, X. Fan, D. Qu, Y. P. Chen, W. G. Wang, J. Wu, T. Y. Chen, J. Q. Xiao, and C. L. Chien, Phys. Rev. Lett. 109, 107204 (2012). 9T. Kikkawa, K. Uchida, Y. Shiomi, Z. Qiu, D. Hou, D. Tian, H. Nakayama, X.-F. Jin, and E. Saitoh, Phys. Rev. Lett. 110, 067207 (2013). 10S. Geprags, S. Meyer, S. Altmannshofer, M. Opel, F. Wilhelm, A. Rogalev, R. Gross, and S. T. B. Goennenwein, Appl. Phys. Lett. 101, 262407 (2012). 11R. Ramos, T. Kikkawa, K. Uchida, H. Adachi, I. Lucas, M. H. Aguirre, P. Algarabel, L. Morell´on, S. Maekawa, E. Saitoh, et al., Appl. Phys. Lett. 102, 072413 (2013). 12T. Tanaka, H. Kontani, M. Naito, T. Naito, D. S. Hirashima, K. Yamada, and J. Inoue, Phys. Rev. B 77, 165117 (2008). 13K. Uchida, M. Ishida, T. Kikkawa, A. Kirihara, T. Murakami, and E. Saitoh, J. Phys.: Condens. Matter 26, 343202 (2014). 14A. Hoffmann, IEEE Trans. Magn. 49, 5172 (2013). 15G. A. Slack, Phys. Rev. 126, 427 (1962). 16J. P. Shepherd, J. W. Koenitzer, R. Arag´on, C. J. Sandberg, and J. M. Honig, Phys. Rev. B 31, 1107 (1985). 17K. Uchida, T. Ota, H. Adachi, J. Xiao, T. Nonaka, Y. Kajiwara, G. Bauer, S. Maekawa, and E. Saitoh, J. Appl. Phys. 111, 103903 (2012).
1905.02391	1	1905	2019-05-07T07:31:33	Layer degree of freedom for excitons in transition metal dichalcogenides	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	Layered transition metal dichalcogenides (TMDCs) host a variety of strongly bound exciton complexes that control the optical properties in these materials. Apart from spin and valley, layer index provides an additional degree of freedom in a few-layer thick film. Here we show that in a few-layer TMDC film, the wavefunctions of the conduction and valence band edge states contributing to the K (K') valley are spatially confined in the alternate layers - giving rise to direct (quasi-)intra-layer bright exciton and lower-energy inter-layer dark excitons. Depending on the spin and valley configuration, the bright exciton state is further found to be a coherent superposition of two layer-induced states, one (E-type) distributed in the even layers and the other (O-type) in the odd layers. The intra-layer nature of the bright exciton manifests as a relatively weak dependence of the exciton binding energy on the thickness of the few-layer film, and the binding energy is maintained up to 50 meV in the bulk limit - which is an order of magnitude higher than conventional semiconductors. Fast stokes energy transfer from the intra-layer bright state to the inter-layer dark states provides a clear signature in the layer-dependent broadening of the photoluminescence peak, and plays a key role in the suppression of the photoluminescence intensity observed in TMDCs with thickness beyond monolayer.	cond-mat.mes-hall	cond-mat	Layer degree of freedom for excitons in transition metal dichalcogenides Sarthak Das, Garima Gupta, and Kausik Majumdar* Department of Electrical Communication Engineering, Indian Institute of Science, Bangalore 560012, India These authors contributed equally Corresponding author, email: [email protected] ABSTRACT: Layered transition metal dichalcogenides (TMDCs) host a variety of strongly bound exciton complexes that control the optical properties in these materials. Apart from spin and valley, layer index provides an additional degree of freedom in a few-layer thick film. Here we show that in a few-layer TMDC film, the wavefunctions of the conduction and valence band edge states contributing to the 𝐾 (𝐾′) valley are spatially confined in the alternate layers - giving rise to direct (quasi-)intra-layer bright exciton and lower-energy inter-layer dark excitons. Depending on the spin and valley configuration, the bright exciton state is further found to be a coherent superposition of two layer-induced states, one (E-type) distributed in the even layers and the other (O-type) in the odd layers. The intra-layer nature of the bright exciton manifests as a relatively weak dependence of the exciton binding energy on the thickness of the few-layer film, and the binding energy is maintained up to 50 meV in the bulk limit -- which is an order of magnitude higher than conventional semiconductors. Fast stokes energy transfer from the intra-layer bright state to the inter-layer dark states provides a clear signature in the layer-dependent broadening of the photoluminescence peak, and plays a key role in the suppression of the photoluminescence intensity observed in TMDCs with thickness beyond monolayer. 1 1. Introduction: The binding energy of exciton is a strong function of quantum confinement of the electrons and holes. A two-dimensional exciton is thus expected to exhibit stronger binding energy than its three- dimensional counterpart1. This, coupled with heavy carrier effective mass2,3, and small dielectric constant3 -- 5, results in a large binding energy of excitons in monolayer Transition Metal Dichalcogenide (TMDC) materials4 -- 9. This has led to recent efforts in exploring the physics of various exciton complexes including excitons10, biexcitons11, trions and their dark states10,12,13, using monolayer TMDC as a testbed. The inversion symmetry of the crystal is broken in the monolayer limit, and more generally, in TMDCs with odd number of layers, giving rise to rich spin and valley physics7,14,15. While exciton complexes have been extensively studied in monolayer TMDCs, the effort in few-layer thick films remains limited16 -- 23. This is primarily due to the transition from direct bandgap in monolayer to indirect bandgap in few-layer suggesting fast relaxation of valley carriers from the 𝐾 (𝐾′) points. Also, inversion symmetry is either explicitly restored (in even number of layers) or smears out (in odd number of layers) in multi-layer films, suppressing valley controllability. On the other hand, few-layer films allow the provision to use layer as an additional degree of freedom. In the 2-H structure of TMDCs, the consecutive layers are rotated by 180° with respect to each other24. Consequently, the electrons at the 𝐾 point in a bilayer system are not allowed to spill over the other layer due to the symmetry of the constituent 𝑑𝑧2 orbital contributing to the conduction band. On the other hand, for the holes, there exists a finite inter-layer coupling. However, there is also a large spin splitting in the valence band, the magnitude of which is larger than the inter-layer coupling term. This results in a confinement of the holes to a single layer as well. The spilling of the hole wavefunction to the consecutive layers is particularly weak in W 2 based TMDCs25,26 compared with Mo based TMDCs owing to larger spin-orbital interaction. Such suppression of inter-layer hopping for both electrons and holes in bilayer TMDCs gives rise to layer pseudospin24. However, this argument of single layer confinement is strictly true only at the 𝐾 (𝐾′) points of the Brillouin zone, particularly for few-layer thick film with number of layers more than two. On the contrary, the momentum space distribution of exciton, as predicted from Bethe-Salpeter (BS) equation27, spreads well beyond the 𝐾 (𝐾′) points, and the wavefunctions spill over to the other layers due to band mixing. In this work, taking the finite momentum space distribution of excitons into account, we generalize the concept of layer degree of freedom for an arbitrary number of layer thickness of WSe2 in the context of the direct exciton to reveal three important properties. First, for a given spin and valley, the layer degree of freedom introduces an additional selection rule for optical brightness. This results from intra- and inter-layer spatial distribution of excitons arising due to electron and hole wavefunctions being distributed either in the odd or in the even layers. Second, the non-radiative scattering from the bright intra-layer to the dark inter-layer states has a clear signature in the layer dependent luminescence linewidth, and plays a key role in luminescence suppression in few-layer TMDC. Third, owing to a pseudo-confinement arising from the quasi-intra-layer nature of the bright exciton, its binding energy is a relatively weak function of thickness of the film, and remains significantly large (~50 meV) even in the bulk limit28 -- 30 compared to conventional semiconductors31,32. 2. Exciton states in few-layer TMDC and their radiative decay: To understand the excitonic structure in a few-layer TMDC, we model the exciton using a combination of 𝒌. 𝒑 Hamiltonian and Bethe-Salpeter theory27,33. Each layer of WSe2 belongs to the 𝐶3ℎ point group at the high symmetry 𝐾 and 𝐾′ points of the Brillouin zone, and the W atoms 3 has a trigonal prismatic coordination with the Se atoms. Close to the band edges around the 𝐾 and 𝐾′ points in the Brillouin zone (Figure 1a), the bands are contributed primarily from the W 𝑑 orbitals. The symmetry driven basis states for the conduction band and the valence band edges for the 𝑙𝑡ℎ layer can be written as26 𝑐⟩ = 5𝑑𝑧2 𝑙 ⟩ and 𝑣⟩ = 1 √2 𝑙 (5𝑑𝑥2−𝑦2 ⟩ + 𝑖𝜏𝑧5𝑑𝑥𝑦 𝑙 ⟩), respectively. Here 𝜏𝑧 = ±1 are the 𝐾 and 𝐾′ valley indices. For AB stacked TMDC, the adjacent layers are rotated by 180∘ around the 𝑐 axis. The Hamiltonian for an 𝑛-layer WSe2 film is obtained by expanding the monolayer 𝒌. 𝒑 Hamiltonian upon incorporating the interlayer coupling of the VBs with the immediate neighbour layers24. In Supplemental Material S134, we show the generalized multi-layer Hamiltonian used in this work. In the same Supplemental Material34, we also show the Hamiltonians for the 2L, 3L and 4L systems explicitly. An exciton state Ψ𝑠(𝐐)⟩ in an exciton band 𝑠 at a centre of mass momentum 𝐐 = 𝒌𝒆 + 𝒌𝒉 is a coherent superposition of hole (with crystal momentum 𝒌𝒉) and electron (with crystal momentum 𝒌𝒆) states from band-pairs (𝑣, 𝑐) in an 𝑛-layer system in the reciprocal space and can be written as Ψ𝑠(𝐐)⟩ = ∑ 𝑣,𝑐,𝐤 (𝑠) (𝐤) 𝜆v,c,𝐐 𝑣, 𝐤⟩𝑐, 𝐤 + 𝐐⟩ … (1) (𝑠) (𝐤) and the exciton eigen energies 𝐸𝑒𝑥 𝜆v,c,𝐐 (𝑠)(𝐐) are obtained from the solution of the BS equation27: 𝑐𝑐′ ⟨𝑣, 𝑐, 𝒌, 𝑸𝐻𝑣′, 𝑐′, 𝒌′, 𝑸⟩ = 𝛿𝑣𝑣′𝛿𝑐𝑐′𝛿𝒌𝒌′(𝜀(𝒌+𝑸)𝑐 − 𝜀𝒌𝑣) − (𝜉 − 𝜚)𝑣𝑣′ (𝒌, 𝒌′, 𝑸) … (2) Here 𝜀 is the quasiparticle energy eigenvalue obtained by diagonalizing the quasiparticle Hamiltonian in Supplemental Material S134. 𝜚 is the exchange term and we neglect this term since in this work we are primarily interested in the exciton band structure for the direct exciton at 𝐾 (𝐾′), with 𝑸 ≈ 𝟎. The direct term 𝜉 is evaluated using Keldysh form of coulomb interaction 4 potential27: 𝑉𝑞 = 2𝜋𝑒2 𝜅𝑞(1+𝑟0𝑞) . The effective dielectric constant 𝜅 and the characteristic screening length 𝑟0 are used as fitting parameters which we vary with the number of layers in the TMDC film. In the rest of the paper, we only consider the spin allowed, bright transitions, and ignore the selection rule governed dark excitons. Also, we shall limit our discussions to A series excitons only, keeping in mind there exists higher energy exciton series (for example, B series and above). Figure 1a schematically illustrates the one-particle bandstructure for bilayer WSe2. The low- energy bands (1s and 2s) of the 𝐴 series exciton in the 𝑸 space are shown in Figure 1b. In bilayer, for each of 𝐴1𝑠 and 𝐴2𝑠 excitons, there are two layer-induced exciton bands [for example, 𝐴1𝑠 (1) and 𝐴1𝑠 (2) for 1s states], which are closely spaced in energy. Figure 1c shows the light cone, within which the energy and momentum conservation laws are obeyed during an exciton recombination to emit a photon. Thus, any spin allowed bright exciton state with 𝑄 < 𝑄0 can emit light by radiative recombination. Owing to the small momentum of the photon compared to the in-plane momentum of the exciton, the light cone constitutes of a small part (< 0.1%) of the Brillouin zone. To compare the strength of the photoluminescence from the different exciton states, we evaluate the radiative decay rate using the following relation35,36: Γ𝑅(𝐐) = 𝜂𝑜 2 𝜒𝑒𝑥(𝐐)2 ∫ 𝑑𝑞𝑧 ħ𝑒2 2𝑚𝑜 ∞ 0 1 √𝑄2 + 𝑞𝑧 2 × (1 + 2 𝑞𝑧 𝑄2 + 𝑞𝑧 2) × Γ(𝐐) 𝜋⁄ [𝐸𝑒𝑥(𝐐) − ħ𝑐√𝑄2 + 𝑞𝑧 2] + Γ(𝐐)2 … (3) Here Γ(𝐐) is the total radiative and non-radiative broadening: Γ(𝐐) = Γ𝑅(𝐐) + Γ𝑁𝑅. We assume Γ𝑁𝑅 to be 𝐐-independent for simplicity. See Supplemental Material S234 for detailed calculation 5 of 𝜒𝑒𝑥(𝐐). Note that equation (3) is a self-consistent equation and provides the fundamental radiative broadening of the exciton states when Γ𝑁𝑅 = 0. Figure 1d shows the calculated intrinsic radiative decay rate (for Γ𝑁𝑅 = 0) for both 𝐴1𝑠 (1) (in black) and 𝐴1𝑠 (2) (in red) as a function of Q for bilayer WSe2. The implications of the large difference between the two rates which will be discussed later. 3. Layer distribution of exciton states -- layer-induced bright and dark states: Fig. 2a schematically shows the conduction and the valence bands of 2L WSe2 for a given spin (𝑠𝑧) and valley (𝜏𝑧) index. The top panel describes the doubly degenerate case 𝜏𝑧. 𝑠𝑧 = 1, which includes (𝜏𝑧 = 1; 𝑠𝑧 = 1) and (𝜏𝑧 = −1; 𝑠𝑧 = −1), while the bottom panel describes the other doubly degenerate case 𝜏𝑧. 𝑠𝑧 = −1, including (𝜏𝑧 = 1; 𝑠𝑧 = −1) and (𝜏𝑧 = −1; 𝑠𝑧 = 1). At 𝒌 = 𝑲, owing to the weak inter-layer coupling, states from both conduction and valence bands are confined to either layer L1 or L2, respectively. However, this is not strictly true for 𝒌 ≠ 𝑲. As an example, Fig. 2b shows the layer distribution of the band edge electron and hole states in 2L WSe2 for 𝒌 = 𝑲 + ∆𝒌, with ∆𝒌 = 2𝜋 𝑎 (0.0033, 0.0033) where 𝑎 = 3.28 Å. The momentum and transition resolved probability distribution (𝜆v,c,𝐐=𝟎 (𝑠) (𝐤)2) of the lowest lying [𝐴1𝑠 (1) and 𝐴1𝑠 (2)] and the next higher energy [𝐴2𝑠 (1) and 𝐴2𝑠 (2)] direct excitons, as obtained from the BS equation, is shown in Figure 2c-d. Note that, each exciton predominantly consists of a single transition between a specific (v, c) band pair, with negligible contribution from the other transitions. The real space layer distribution of the dominant transition for each exciton state is also schematically shown in Figure 2c-d. The unique layer distribution of the electron and hole basis states governs the formation of intra-layer and inter-layer exciton. For example, for 𝜏𝑧. 𝑠𝑧 = 1, the lower energy exciton [𝐴1𝑠 (1)] almost entirely arises from C1V2 transition, and hence is an 6 inter-layer exciton as inferred from the top panel of Figure 2b. On the other hand, the higher energy exciton [𝐴1𝑠 (2)] results primarily from C2V2 transition and hence forms an intra-layer exciton confined in the bottom layer. On the other hand, for 𝜏𝑧. 𝑠𝑧 = −1, the intra-layer exciton is confined in the top layer, as explained in the bottom panel of Figure 2b. Also note that with an increase in quantum number (from 1s to 2s), the more confined areal distribution of the exciton in the k-space suggests a larger spread in real space distribution. The intra- and inter-layer spatial distributions of the different exciton states are expected to strongly affect their radiative decay. As mentioned earlier, since all these excitons are spin allowed bright states, any exciton with 𝐐 lying within the light cone (in Figure 1c) can, in principle, recombine radiatively emitting photons in a spontaneous fashion. However, in Figure 1d, we observe that 𝐴1𝑠 (1) exciton is an order of magnitude weaker compared to the 𝐴1𝑠 (2) state in terms of light emission due to its inter-layer nature. The primary contributing orbitals (W 5𝑑) for the excitons exhibit a spatial extent along the out-of-plane (𝑧) direction that is much smaller than the interlayer separation (see Figure 2b), suppressing the matrix element for the decay rate in the case of inter-layer exciton. Henceforth, we call these inter-layer states as layer-induced dark excitons. We can thus conclude that the light emission from the 1s state predominantly happens due to the radiative recombination of the intra-layer 𝐴1𝑠 (2) exciton. The analysis can be readily extended to the tri-layer (3L) system, and the results are summarized in Figure 3, where three different layer-induced 1s excitons (from the A-series) are formed, namely 𝐴1𝑠 (1), 𝐴1𝑠 (2) and 𝐴1𝑠 (3). Figure 3a schematically shows the electronic bandstructure around 𝐾 and 𝐾′ points. Similar to bilayer, the components of the eigenstates are significant only in the alternate layers, that is, they are either confined to the even layers or to the odd layers, as illustrated 7 in Figure 3b. The transition and momentum resolved probability distribution of the resulting 𝐴1𝑠 excitons are explicitly shown in Fig. 3c-e. The probability distributions indicate the dominance of one transition out of 9 possible transitions for an exciton. Similar to the bilayer case, the resulting excitons also follow inter-layer pattern for low energy [𝐴1𝑠 (1) and 𝐴1𝑠 (2)] excitons, while intra- layer pattern for the higher energy [𝐴1𝑠 (3)] state. Interestingly, for 𝜏𝑧. 𝑠𝑧 = 1, the 𝐴1𝑠 (3) exciton is confined in the middle layer (L2), as shown in the top panel of Figure 3e. However, for the other spin-valley configuration (𝜏𝑧. 𝑠𝑧 = −1), the 𝐴1𝑠 (3) exciton is confined to the L1 and L3 (bottom panel of Figure 3e). Thus, it maintains its intra-layer structure, but gets distributed in the odd numbered layers. We term the latter case as quasi-intra-layer exciton. The calculated decay rates of the different exciton states for tri-layer WSe2 are shown in Supplemental Material S334. Both types of 𝐴1𝑠 (3) excitons exhibit more than an order of magnitude higher decay rate compared to the rest, and responsible for photoluminescence. In Figure 4, we schematically depict the real space layer-resolved distribution of only the bright excitons (𝐴1𝑠 (𝑛)), for bilayer (2L) to six-layer (6L) thick WSe2 films. We can generalize that for an 𝑛-layer thick TMDC, there are 2 doubly-degenerate bright (quasi-)intra-layer excitons. The rest 2𝑛-2 exciton states are inter-layer and hence layer-induced-dark in nature, which are otherwise bright from a conventional selection rule (spin and azimuthal quantum number selection) perspective. Between the two doubly-degenerate bright excitons, one exciton is distributed in the even layers and the other in the odd layers, and we call them as E-type (with layer index 𝑙𝑧 = +1) and O-type (𝑙𝑧 = −1) exciton, respectively. In the case of 1L and 2L systems, the bright excitons are confined to a single layer. For 3L system, the E-type exciton is confined to a single (middle) layer, while the O-type one is quasi-intra-layer in nature, being distributed between the top and the bottom layers. For 4L and thicker samples, we only have quasi-intra-layer doubly-degenerate E- 8 type and O-type bright excitons. A careful observation reveals that the spin, valley and layer indices of a bright exciton are coupled by the simple rule 𝑙𝑧𝑠𝑧𝜏𝑧 = +1, which dictates the possible quantum states allowed in a few-layer TMDC system. 4. Experimental evidence and implications: We next explore indirect experimental evidences and subsequent implications of the above- mentioned layer distribution of the exciton states. In order to do so, we employ temperature dependent photoluminescence measurement from WSe2 films of varying layer thickness. A. Experiment: We mechanically exfoliate WSe2 flakes on a clean Si substrate covered with 285 nm thick SiO2. The thickness of the flake is identified by a combination of Raman and AFM. Photoluminescence (PL) measurement is carried out by varying the sample temperature from 3.3 K to room temperature. The pressure of the sample chamber is kept below 10−4 torr at all measurement temperatures. The PL is collected through a 50X objective with a numerical aperture of 0.5 in confocal mode. The optical power density on the sample is kept below 100 µW to avoid any laser induced heating effect. Figure 5 summarizes the temperature and thickness dependence of the acquired photoluminescence spectra from WSe2 samples using a 532 nm laser excitation. In Figure 5a, both the neutral (𝐴1𝑠) and charged (𝐴1𝑠 𝑇 ) 𝐴-series exciton peaks are distinctly visible in the temperature range up to ~90 K for monolayer sample. The red shift of the peak positions with an increase in temperature is due to a corresponding decrease in the quasiparticle bandgap. The weak, but distinct 2s (𝐴2𝑠) and 3s (𝐴3𝑠) peaks of the A exciton are observed in the zoomed-in Figure 5b around 1.87 9 eV and 1.93 eV at T = 3.3 K, which smear out as the sample temperature is increased. To confirm that the higher energy peaks originate from the higher order free exciton bright states, we perform polarization resolved photoluminescence measurement at T = 3.3 K. The sample is excited with a 𝜎+ circularly polarized light from a 633 nm laser, and the emitted light is passed through a 𝜎+ or 𝜎− analyzer. The results for the 1L flake are summarized in Figure 5c. We observe that the 𝐴1𝑠 exciton peak and the 𝐴1𝑠 𝑇 trion peak show a degree of circular polarization (𝜌) of ~8.5% and ~10.2% respectively, where 𝜌 = 𝐼𝜎+ − 𝐼𝜎− 𝐼𝜎+ + 𝐼𝜎− . In the inset of Figure 5c, we show a magnified portion of the next higher order peak, which shows a strong polarization contrast of ~26%, confirming its 𝐴2𝑠 assignment. The enhancement of 𝜌 from 1s to 2s is because the 633 nm laser excites the 2s excitons in a near-resonant manner, suppressing the depolarization due to intervalley scattering. In Figure 5d, we show the acquired PL spectra of WSe2 flakes with varying thickness, namely 1L, 2L, 3L, and 6L, all taken at T = 3.3 K. In the left panel, apart from the neutral and charged exciton peaks, we also observe several peaks at energy lower than trion emission energy. The origin of these lower energy peaks has been previously attributed to defect bound localized excitons10 -- 12 and multi-particle excitonic states10 -- 13. On the other hand, the higher energy peaks, as shown in a magnified energy range in the right panel in Figure 5d, are only distinctly visible for 1L, 2L and 3L cases. B. Weak dependence of exciton binding energy on thickness: The positions of the 𝐴1𝑠, 𝐴2𝑠 and 𝐴3𝑠 peaks remain almost unaltered (within ~5 meV error bar due to the variation in the individual spectrum obtained from these samples) irrespective of the thickness of the sample. Such layer independence of the 𝐴1𝑠 exciton peak position has been widely reported previously37 -- 40. 10 The (quasi-) intra-layer nature of the bright exciton irrespective of the number of layers in the film forces a spatial pseudo-confinement of the exciton to individual layers. This allows the excitons to retain their two-dimensional character even in multi-layer samples. This effect manifests itself as a weak dependence of the bright exciton binding energy on the number of layers of the film. The bright excitons being accessible by photoluminescence experiment, allows us to readily verify this hypothesis. The exciton emission energies, calculated from Equation 2 for different layers, are plotted as a function of the quantum numbers in Figure 6a as the open symbols, which are in good agreement with the experimental data, shown by the solid symbols. The insets show zoomed-in views of the data from individual layers. The corresponding continuum levels obtained from the BS equation for different layer numbers are also shown in the same figure by solid horizontal lines. The corresponding binding energy of the different exciton states is then calculated by taking the difference between the continuum level and the emission energy (obtained from both photoluminescence spectra as well as BS equation) and plotted as a function of the thickness of the WSe2 flake in Figure 6b. The agreement between BS theory and experiment is quite remarkable. The observation of the weak dependence of the exciton binding energy on WSe2 film thickness is in stark contrast with a conventional semiconductor when the out of plane quantum confinement is relaxed. Also, the binding energy of the exciton for bulk TMDC is about 50 meV, which was measured long back28 -- 30. This is about an order of magnitude higher than typical exciton binding energies of III-V semiconductor samples31,32. The retention of the large binding energy in the bulk limit is another implication of such quasi-intra-layer configuration of the bright excitons in TMDCs, which maintains a quasi-two-dimensional nature due to layer confinement even in thick samples. A summary of the layer dependence on the energy and the binding energy of different excitonic states is provided in Supplemental Material S434. 11 C. Layer dependent exciton linewidth broadening: Using a Voigt fit to the exciton peaks for samples with varying layer thickness, we deconvolute the homogeneous (Lorentzian) and the inhomogeneous (Gaussian, shown in Supplemental Materials S534) components of the exciton photoluminescence linewidth. The extracted homogeneous linewidth is found to increase monotonically as a function of number of layers (𝑛) in Figure 7a (green symbols). The total homogeneous linewidth is a result of both radiative and non-radiative dephasing processes. Using the self-consistency of Equation 3, we deconvolute the corresponding non-radiative part [2Γ𝑁𝑅(𝑛)] of the homogeneous broadening as a function of 𝑛 from the photoluminescence homogeneous linewidth36. 2Γ𝑁𝑅(𝑛) is found to increase linearly with 𝑛. As the excitation density was maintained low (< 109 cm−2) during measurements, the exciton- exciton scattering induced dephasing41 is small, and the exciton-phonon scattering is the dominating non-radiative dephasing process in a monolayer sample in our experiment. For 𝑛 ≥ 2, (𝑛) apart from the exciton-phonon scattering within the bright 𝐴1𝑠 band, scattering to the indirect valleys (𝜞 and 𝜦) and to the lower energy inter-layer dark states are the additional non-radiative (𝑛) dephasing mechanisms. We assume that the phonon scattering within the 𝐴1𝑠 band is independent of layer number, and therefore, is equal to the monolayer non-radiative linewidth (Γ0). The layer dependence of the non-radiative linewidth can then be given by Γ𝑁𝑅(𝑛) = Γ0 + Γ𝐼 + ∑ 𝑛−1 𝑖=1 Γ (𝑛) 𝐴1𝑠 (𝑖) →𝐴1𝑠 ; 𝑛 > 1 … (4) Here, Γ𝐼 quantifies the lumped effect of dephasing due to exciton-phonon scattering to the indirect valleys. Due to large inter-valley momentum mismatch, Γ𝐼 is expected to be small compared to intra-valley scattering rates. In the last term, Γ (𝑛) 𝐴1𝑠 (𝑖) is the scattering of the bright (quasi-)intra- →𝐴1𝑠 (𝑛) layer 𝐴1𝑠 (𝑖) exciton to the 𝑖𝑡ℎ dark inter-layer 𝐴1𝑠 exciton, maintaining both their spin and valley 12 (𝑖) indices (that is, conserving total angular momentum). Note that 𝐴1𝑠 ranges from 𝐴1𝑠 (1) → 𝐴1𝑠 (𝑛−1) , and this results in a proportionately increasing number of scattering channels as the number of layers increases (see Figure 7b for 2L and 3L cases). For a first order estimate, we assume the same scattering rate [denoted by Γ (𝐵) 𝐴1𝑠 (𝑛) (𝐷)] from bright 𝐴1𝑠 →𝐴1𝑠 to any of the 𝑖𝑡ℎ lower energy inter- layer dark state. We can then rewrite Equation 4 as Γ𝑁𝑅(𝑛) = Γ0 + Γ𝐼 + (n − 1)Γ (𝐵) 𝐴1𝑠 (𝐷) ; 𝑛 > 1 →𝐴1𝑠 … (5) This explains the linear increment in the non-radiative exciton linewidth as the number of layers increases. D. Photoluminescence suppression beyond monolayer: Γ (𝐵) 𝐴1𝑠 (𝐷) is extracted from the slope of the linear fit from Figure 7a, and is found to be ~2.25 →𝐴1𝑠 meV, which translates to a scattering rate of 3.4 × 1012 s-1 per channel. This is on the order of radiative decay rate of the bright exciton (see Figure 1d) as well as the carrier transfer to indirect valleys42. Equation (5) suggests that with an increase in the number of layers, the total non- radiative decay rate due to intra-layer to inter-layer stokes energy transfer increases proportionately. Since the inter-layer states do not contribute to the luminescence, such non- radiative scattering competes with the exciton radiative decay process. This suggests that apart from carrier transfer to the indirect valleys, the fast scattering to the inter-layer dark states also plays a key role in suppressing luminescence in few-layer TMDCs. Conclusions: In summary, the symmetry driven even- and odd-layer distribution of the band edge states close to the zone corner forces intra- (or quasi-intra-) and inter-layer distribution of excitons in few-layer 13 TMDCs. The intra-layer exciton states exhibit more than an order of magnitude higher radiative decay rate compared to the inter-layer states, and hence only these excitons contribute to the luminescence. These bright intra-layer excitons can further be classified into E-type and O-type excitons (denoted as layer index), depending on their spatial layer distribution over either even or odd numbered layers, respectively. This layer index (𝑙𝑧) is coupled to the spin (𝑠𝑧) and valley (𝜏𝑧) indices by the rule 𝑙𝑧𝑠𝑧𝜏𝑧 = +1. Such unique layer distribution has direct implication in maintaining large exciton binding energy in TMDCs up to the bulk limit. Further, the layer index (E or O) can be treated as an additional degree of freedom of the exciton quantum state in a few- layer system, and can be used for quantum information manipulation. ACKNOWLEDGMENTS K. M. acknowledges the support a grant from Indian Space Research Organization (ISRO), grants under Ramanujan Fellowship, Early Career Award, and Nano Mission from the Department of Science and Technology (DST), Government of India, and support from MHRD, MeitY and DST Nano Mission through NNetRA. Competing interests The authors declare no competing financial or non-financial interests. 14 References: (1) X. L. Yang, S. H. Guo, F. T. Chan, K. W. W. and W. Y. C. Analytic Solution of a Two- Dimensional Hydrogen Atom. I. Nonrelativistic Theory. Phys. Rev. A 1991, 43 (3), 1186 -- 1196. (2) Majumdar, K.; Hobbs, C.; Kirsch, P. D. Benchmarking Transition Metal Dichalcogenide MOSFET in the Ultimate Physical Scaling Limit. IEEE Electron Device Lett. 2014, 35 (3), 402 -- 404. https://doi.org/10.1109/LED.2014.2300013. (3) Ramasubramaniam, A. Large Excitonic Effects in Monolayers of Molybdenum and Tungsten Dichalcogenides. Phys. Rev. B - Condens. Matter Mater. Phys. 2012, 86 (11), 1 -- 6. https://doi.org/10.1103/PhysRevB.86.115409. (4) He, K.; Kumar, N.; Zhao, L.; Wang, Z.; Mak, K. F.; Zhao, H.; Shan, J. Tightly Bound Excitons in Monolayer WSe2. Phys. Rev. Lett. 2014, 113 (2), 1 -- 5. https://doi.org/10.1103/PhysRevLett.113.026803. (5) Komsa, H. P.; Krasheninnikov, A. V. Effects of Confinement and Environment on the Electronic Structure and Exciton Binding Energy of MoS2 from First Principles. Phys. Rev. B - Condens. Matter Mater. Phys. 2012, 86 (24), 1 -- 6. https://doi.org/10.1103/PhysRevB.86.241201. (6) Chernikov, A.; Berkelbach, T. C.; Hill, H. M.; Rigosi, A.; Li, Y.; Aslan, O. B.; Reichman, D. R.; Hybertsen, M. S.; Heinz, T. F. Exciton Binding Energy and Nonhydrogenic Rydberg Series in Monolayer WS2. Phys. Rev. Lett. 2014, 113 (7), 1 -- 5. https://doi.org/10.1103/PhysRevLett.113.076802. 15 (7) Xu, X.; Yao, W.; Xiao, D.; Heinz, T. F. Spin and Pseudospins in Layered Transition Metal Dichalcogenides. Nat. Phys. 2014, 10 (5), 343 -- 350. https://doi.org/10.1038/nphys2942. (8) Shi, H.; Pan, H.; Zhang, Y. W.; Yakobson, B. I. Quasiparticle Band Structures and Optical Properties of Strained Monolayer MoS2 and WS2. Phys. Rev. B - Condens. Matter Mater. Phys. 2013, 87 (15), 1 -- 8. https://doi.org/10.1103/PhysRevB.87.155304. (9) Gupta, G.; Kallatt, S.; Majumdar, K. Direct Observation of Giant Binding Energy Modulation of Exciton Complexes in Monolayer MoSe2. Phys. Rev. B 2017, 96 (8), 081403. https://doi.org/10.1103/PhysRevB.96.081403. (10) Jones, A. M.; Yu, H.; Ghimire, N. J.; Wu, S.; Aivazian, G.; Ross, J. S.; Zhao, B.; Yan, J.; Mandrus, D. G.; Xiao, D.; et al. Optical Generation of Excitonic Valley Coherence in Monolayer WSe2. Nat. Nanotechnol. 2013, 8 (9), 634 -- 638. https://doi.org/10.1038/nnano.2013.151. (11) You, Y.; Zhang, X.-X.; Berkelbach, T. C.; Hybertsen, M. S.; Reichman, D. R.; Heinz, T. F. Observation of Biexcitons in Monolayer WSe2. Nat. Phys. 2015, 11 (6), 477 -- 481. https://doi.org/10.1038/nphys3324. (12) Zhang, X.-X.; Cao, T.; Lu, Z.; Lin, Y.-C.; Zhang, F.; Wang, Y.; Li, Z.; Hone, J. C.; Robinson, J. A.; Smirnov, D.; et al. Magnetic Brightening and Control of Dark Excitons in Monolayer WSe2. Nat. Nanotechnol. 2017, No. June, 1 -- 7. https://doi.org/10.1038/nnano.2017.105. (13) Wang, G.; Bouet, L.; Lagarde, D.; Vidal, M.; Balocchi, A.; Amand, T.; Marie, X.; Urbaszek, B. Valley Dynamics Probed through Charged and Neutral Exciton Emission in Monolayer WSe2. Phys. Rev. B - Condens. Matter Mater. Phys. 2014, 90 (7), 1 -- 6. 16 https://doi.org/10.1103/PhysRevB.90.075413. (14) Yao, W.; Xiao, D.; Niu, Q. Valley-Dependent Optoelectronics from Inversion Symmetry Breaking. Phys. Rev. B - Condens. Matter Mater. Phys. 2008, 77 (23), 1 -- 7. https://doi.org/10.1103/PhysRevB.77.235406. (15) Srivastava, A.; Sidler, M.; Allain, A. V.; Lembke, D. S.; Kis, A.; Imamoglu, A. Valley Zeeman Effect in Elementary Optical Excitations of a Monolayer WSe2. 2014, 11 (February), 141 -- 147. https://doi.org/10.1038/nphys3203. (16) Dai, X.; Li, W.; Wang, T.; Wang, X.; Zhai, C. Bandstructure Modulation of Two- Dimensional WSe2 by Electric Field. J. Appl. Phys. 2015, 117 (8). https://doi.org/10.1063/1.4907315. (17) Li, D.; Cheng, R.; Zhou, H.; Wang, C.; Yin, A.; Chen, Y.; Weiss, N. O.; Huang, Y.; Duan, X. Electric-Field-Induced Strong Enhancement of Electroluminescence in Multilayer Molybdenum Disulfide. Nat. Commun. 2015, 6 (May), 7509. https://doi.org/10.1038/ncomms8509. (18) Palummo, M.; Bernardi, M.; Grossman, J. C. Exciton Radiative Lifetimes in Two- Dimensional Transition Metal Dichalcogenides. Nano Lett. 2015, 15 (5), 2794 -- 2800. https://doi.org/10.1021/nl503799t. (19) Zhang, Y.; Chang, T.-R.; Zhou, B.; Cui, Y.-T.; Yan, H.; Liu, Z.; Schmitt, F.; Lee, J.; Moore, R.; Chen, Y.; et al. Direct Observation of the Transition from Indirect to Direct Bandgap in Atomically Thin Epitaxial MoSe2. 2014, 9 (February), 111 -- 115. https://doi.org/10.1038/nnano.2013.277. 17 (20) Brumme, T.; Calandra, M.; Mauri, F. First-Principles Theory of Field-Effect Doping in Transition-Metal Dichalcogenides: Structural Properties, Electronic Structure, Hall Coefficient, and Electrical Conductivity. Phys. Rev. B - Condens. Matter Mater. Phys. 2015, 91 (15), 1 -- 19. https://doi.org/10.1103/PhysRevB.91.155436. (21) Li, X.; Zhang, F.; Niu, Q. Unconventional Quantum Hall Effect and Tunable Spin Hall Effect in Dirac Materials: Application to an Isolated MoS2 Trilayer. Phys. Rev. Lett. 2013, 110 (6), 1 -- 5. https://doi.org/10.1103/PhysRevLett.110.066803. (22) Roldán, R.; López-Sancho, M. P.; Guinea, F.; Cappelluti, E.; Silva-Guillén, J. A.; Ordejón, P. Momentum Dependence of Spin -- orbit Interaction Effects in Single-Layer and Multi- Layer Transition Metal Dichalcogenides. 2D Mater. 2014, 1 (3), 034003. https://doi.org/10.1088/2053-1583/1/3/034003. (23) Roldán, R.; Silva-Guillén, J. A.; López-Sancho, M. P.; Guinea, F.; Cappelluti, E.; Ordejón, P. Electronic Properties of Single-Layer and Multilayer Transition Metal Dichalcogenides MX2 (M = Mo, W and X = S, Se). Ann. Phys. 2014, 526 (9 -- 10), 347 -- 357. https://doi.org/10.1002/andp.201400128. (24) Gong, Z.; Liu, G.-B.; Yu, H.; Xiao, D.; Cui, X.; Xu, X.; Yao, W. Magnetoelectric Effects and Valley-Controlled Spin Quantum Gates in Transition Metal Dichalcogenide Bilayers. Nat. Commun. 2013, 4 (May), 1 -- 6. https://doi.org/10.1038/ncomms3053. (25) Latzke, D. W.; Zhang, W.; Suslu, A.; Chang, T. R.; Lin, H.; Jeng, H. T.; Tongay, S.; Wu, J.; Bansil, A.; Lanzara, A. Electronic Structure, Spin-Orbit Coupling, and Interlayer Interaction in Bulk MoS2 and WS2. Phys. Rev. B - Condens. Matter Mater. Phys. 2015, 91 (23), 1 -- 6. https://doi.org/10.1103/PhysRevB.91.235202. 18 (26) Xiao, D.; Liu, G. Bin; Feng, W.; Xu, X.; Yao, W. Coupled Spin and Valley Physics in Monolayers of MoS 2 and Other Group-VI Dichalcogenides. Phys. Rev. Lett. 2012, 108 (19), 1 -- 5. https://doi.org/10.1103/PhysRevLett.108.196802. (27) Wu, F.; Qu, F.; Macdonald, A. H. Exciton Band Structure of Monolayer MoS2. Phys. Rev. B - Condens. Matter Mater. Phys. 2015, 91 (7), 1 -- 8. https://doi.org/10.1103/PhysRevB.91.075310. (28) Beal, A. R.; Knights, J. C.; Liang, W. Y. Transmission Spectra of Some Transition Metal Dichalcogenides : 11. Group VIA : Trigonal Prismatic Coordination. J. Phys. J. Phys. C Solid State Phys 1972, 5. (29) Beal, A. R.; Liang, W. Y. Excitons in 2H-WSe2 and 3R-WS2. J. Phys. C. Solid State Phys. 1976, 9, 2459 -- 2466. (30) Wilson, J. A.; Yoffe, A. D. The Transition Metal Dichalcogenides Discussion and Interpretation of the Observed Optical, Electrical and Structural Properties. Adv. Phys. 1969, 18 (73), 193 -- 335. https://doi.org/10.1080/00018736900101307. (31) Greene, R. L.; Bajaj, K. K.; Phelps, D. E. Energy Levels of Wannier Excitons in GaAs-Ga1- XAlxAs Quantum-Well Structures. Phys. Rev. B 1984, 29 (4), 1807 -- 1812. https://doi.org/10.1103/PhysRevB.29.1807. (32) Miller, R. C.; Kleinman, D. A.; Tsang, W. T.; Gossard, A. C. Observation of the Excited Level of Excitons in GaAs Quantum Wells. Phys. Rev. B 1981, 24 (2), 1134 -- 1136. https://doi.org/10.1103/PhysRevB.24.1134. (33) Qiu, D. Y.; Cao, T.; Louie, S. G. Nonanalyticity, Valley Quantum Phases, and Lightlike 19 Exciton Dispersion in Monolayer Transition Metal Dichalcogenides: Theory and First- Principles Calculations. Phys. Rev. Lett. 2015, 115 (17), 1 -- 5. https://doi.org/10.1103/PhysRevLett.115.176801. (34) Das, S.; Gupta, G.; Majumdar, K. Supplemental Material for : Layer Degree of Freedom for Excitons in Transition Metal Dichalcogenides. (35) Wang, H.; Bang, J.; Sun, Y.; Liang, L.; West, D.; Meunier, V.; Zhang, S. The Role of Collective Motion in the Ultrafast Charge Transfer in van Der Waals Heterostructures. Nat. Commun. 2016, 7, 11504. (36) Gupta, G.; Majumdar, K. Fundamental Exciton Linewidth Broadening in Monolayer Transition Metal Dichalcogenides. 2019, 085412 (September 2018), 1 -- 9. https://doi.org/10.1103/PhysRevB.99.085412. (37) Maciej R. Molas, Karol Nogajewski, Artur O. Slobodeniuk, Johannes Binder, Miroslav Bartos, and M. P. Optical Response of Monolayer, Few-Layer and Bulk Tungsten Disulfide. Nanoscale 2017, 7, 10421 -- 10429. https://doi.org/10.1039/b000000x/. (38) Arora, A.; Koperski, M.; Nogajewski, K.; Marcus, J.; Faugeras, C.; Potemski, M. Excitonic Resonances in Thin Films of WSe2: From Monolayer to Bulk Material. Nanoscale 2015, 7 (23), 10421 -- 10429. https://doi.org/10.1039/c5nr01536g. (39) Zeng, H.; Liu, G. Bin; Dai, J.; Yan, Y.; Zhu, B.; He, R.; Xie, L.; Xu, S.; Chen, X.; Yao, W.; et al. Optical Signature of Symmetry Variations and Spin-Valley Coupling in Atomically Thin Tungsten Dichalcogenides. Sci. Rep. 2013, 3, 2 -- 6. https://doi.org/10.1038/srep01608. (40) Zhao, W.; Ghorannevis, Z.; Chu, L.; Toh, M.; Kloc, C.; Tan, P.-H.; Eda, G. Evolution of 20 Electronic Structure in Atomically Thin Sheets of WS2 and WSe2. ACS Nano 2012, 7 (1), 791 -- 797. (41) Moody, G.; Kavir Dass, C.; Hao, K.; Chen, C.-H.; Li, L.-J.; Singh, A.; Tran, K.; Clark, G.; Xu, X.; Berghäuser, G.; et al. Intrinsic Homogeneous Linewidth and Broadening Mechanisms of Excitons in Monolayer Transition Metal Dichalcogenides. Nat. Commun. 2015, 6 (May), 8315. https://doi.org/10.1038/ncomms9315. (42) Bertoni, R.; Nicholson, C. W.; Waldecker, L.; Hübener, H.; Monney, C.; De Giovannini, U.; Puppin, M.; Hoesch, M.; Springate, E.; Chapman, R. T.; et al. Generation and Evolution of Spin-, Valley-, and Layer-Polarized Excited Carriers in Inversion-Symmetric WSe2. Phys. Rev. Lett. 2016, 117 (27), 1 -- 5. https://doi.org/10.1103/PhysRevLett.117.277201. 21 (1) and 𝐴2𝑠 Figure 1. Electronic and excitonic band structure in bilayer TMDCs. (a) Band diagram showing the conduction band (CB) and valence band (VB) around the corners of the hexagonal Brillouin zone in bilayer TMDCs. Each band is spin degenerate. (b) Exciton band dispersion of 1𝑠 and 2𝑠 excitons with its centre-of-mass momentum (𝑸). Contrary to a monolayer system, there are two possible 1𝑠 states arising from layer degree of freedom. (c) Light cone for excitons, where 𝑄0 is the edge of the conventional light cone and the boundary of the light cone is given by the (1) light line ℏ𝑐𝑄. The excitons occupying the lower energy 𝐴1𝑠 bands (dashed line) are dark (2) in nature because of their inter-layer character, whereas, the higher energy 𝐴1𝑠 excitons (solid line) are bright due to their intra-layer character. (d) Radiative decay rate variation of the two 1𝑠 exciton bands in (b) with 𝑄 (in units of 10−4 2𝜋 ) on varying the non-radiative linewidth of 𝑎 (2) the exciton band. In the absence of non-radiative scattering (Γ𝑁𝑅 = 0), the decay rate of the 𝐴1𝑠 (1) exciton (red) is roughly two orders of magnitude larger than the 𝐴1𝑠 exciton (black), showing that (1) the lower energy 𝐴1𝑠 (2) As Γ𝑁𝑅 increases, the decay rate outside the light cone boundary 𝑄0 for the bright 𝐴1𝑠 increases due to enhanced participation of the broadened exciton states above the light line. (2) exciton is radiatively inefficient compared to the higher energy 𝐴1𝑠 (2) and 𝐴2𝑠 state. exciton 22 Figure 2. Exciton formation in 2L WSe2. (a) Schematic of layer induced bands at the zone corner for different spin (𝑠𝑧) and valley (𝜏𝑧) configurations. Vi and Ci correspond to the ith valence and conduction band, respectively. The spin and valley configuration for the top and bottom rows has been followed in the subsequent figures also. (b) Real space distribution of different bands at 𝒌 = 𝑲 + ∆𝒌, with ∆𝒌 = (0.0033, 0.0033) with the same spin and valley configuration indicated in 2𝜋 𝑎 in figure (a). The conduction and valence bands are shown in the left and right panels respectively. The physical locations of the layers are shown in the middle. (c-d) 𝒌-space distribution of (c) the 𝐴1𝑠 exciton and (d) the 𝐴2𝑠 exciton, for all the possible individual transitions for a bilayer system. The two different 𝐴1𝑠 states have been indicated according to their dark (in grey boxes) and bright (in yellow boxes) nature. The corresponding real space layer-resolved distribution for each exciton configuration (green and copper spheres indicate the hole and the electron, respectively) is illustrated schematically above the top panel and below the bottom panels. 𝐴2𝑠 excitons are more separated in real space while more confined in the 𝑘-space. 23 Figure 3. Exciton formation in 3L WSe2. (a) Schematic of layer induced bands at the zone corner for different spin (𝑠𝑧) and valley (𝜏𝑧) configurations. Vi and Ci correspond to the ith valence and conduction band, respectively. (b) Real space distribution of different bands at 𝒌 = 𝑲 + ∆𝒌, with ∆𝒌 = (0.0033, 0.0033) with the same spin valley configuration indicated in in figure (a). The 2𝜋 𝑎 conduction and valence bands are shown in the left and right panels, respectively. The physical locations of the layers are shown in the middle. (c-e) 𝑘-space distribution of the 𝐴1𝑠 exciton for all the possible individual transitions for a tri-layer system. The three different 𝐴1𝑠 states have been indicated according to their dark (in grey boxes) in (c-d) and bright (in yellow boxes) nature in (e). The corresponding real space layer-resolved distribution for each exciton configuration (green and copper spheres indicate the hole and the electron, respectively) is illustrated schematically above the top panel and below the bottom panels. Depending on layer distribution, low energy in (e). forms the intra-layer excitons 24 Figure 4. Layer distribution of bright excitons. Schematic for the real space layer-resolved distribution of different intra-layer bright exciton states for bilayer (2L) to six-layers (6L) for two doubly degenerate configurations. Depending on their layer distribution, these are classified as O- type (distributed in the odd numbered layers, top panel, layer index 𝑙𝑧 = −1) and E-type (distributed in the even numbered layers, bottom panel, layer index 𝑙𝑧 = +1) excitons. 25 Figure 5. Exciton states in WSe2 probed through photoluminescence. (a) PL intensity variation of 1L WSe2 at sample temperature ranging from 3.3 K to 295 K with 532 nm laser excitation. The 𝑇 ) peaks are indicated. (b) The magnified view of the corresponding exciton (𝐴1𝑠) and trion (𝐴1𝑠 higher order exciton (𝐴2𝑠 and 𝐴3𝑠) peaks with increasing temperature, as indicated by arrows. (c) Circular polarization resolved PL spectra for monolayer WSe2 with a 633 nm laser excitation at 3.3 K with a polarization contrast (𝜌) of ~8.5% for 𝐴1𝑠. Inset. The degree of polarization of 𝐴2𝑠 peak is around 26%. (d) PL spectra of 1L, 2L, 3L, and 6L WSe2, with 532 nm laser excitation at 3.3 K in the left panel. PL spectra showing higher order exciton peaks (𝐴2𝑠 and 𝐴3𝑠) for the same samples in (d) in the right panel. The higher order peaks (indicated by arrows) are discernible only for 1L, 2L and 3L samples. 26 1.41.61.82.0 1.81.92.0 1.61.71.81.9 1.801.861.92 ~ . %~ %𝜎+/𝜎+𝜎−𝜎+/𝜎- = . Energy (eV)PL Intensity (counts)Energy (eV)𝐴2𝑠Energy (eV)1.51.61.71.8 1L𝐴1𝑠3L6L2L×50×250×401.841.922.002.08 𝐴3𝑠𝐴2𝑠PL Intensity (counts)1L3L6L2L𝐴2𝑠𝐴3𝑠(a)(b)(c)(d)𝐴1𝑠𝐴1𝑠𝑇 Figure 6. Thickness dependent binding energy of excitons. (a) PL emission energy plotted as a function of quantum number. The solid and open symbols represent the emission energies as obtained from photoluminescence experiment ("Experimental") and the Bethe-Salpeter (BS) equation ("Theory"), respectively. The "Theory" values correspond to the bright (quasi-)intra- (𝑛) exciton for the n-layer thick film. The continuum, as obtained from BS equation for layer 𝐴1𝑠 different layers, is also shown as solid horizontal lines. The binding energy for a given state (1s, 2s, 3s) and sample thickness is extracted by subtracting the emission energy of that state from the corresponding continuum level, as indicated by the dashed red arrows for 1L case. The zoomed- in view of the energy states corresponding to the individual quantum numbers are also shown in the insets. (b) The extracted binding energies from (a), plotted as a function of number of layers. Solid symbols represent experimental values, and open symbols with dashed lines indicate BS equation predicted (Theory) values. 27 1234560.20.30.45 1231.82.02.2 11.741.76 21.861.89 31.891.921.95 ExperimentalTheoryQuasi-particle Bandgap L2L3L6LEnergy (eV)Quantum number Binding Energy (eV)Number of LayersExperimentalTheory (b)(a)Figure 7. Exciton scattering to inter-layer dark states. (a) The experimental Lorentzian linewidth (green symbols) and the corresponding extracted non-radiative broadening 2Γ𝑁𝑅 (orange symbols), as a function of number of layers (𝑛). Red line is the fitted function demonstrating linear relationship between non-radiative broadening and 𝑛. The fitted expression is shown in the inset. (b) Real space layer resolved distribution of the 𝐴1𝑠 exciton for each spin-valley {𝑠𝑧, 𝜏𝑧} configuration for a monolayer (top panel), bilayer (middle panel) and a tri-layer (bottom panel) system. 28 Supplemental Material Layer degree of freedom for excitons in transition metal dichalcogenides Sarthak Das, Garima Gupta, and Kausik Majumdar Department of Electrical Communication Engineering, Indian Institute of Science, Bangalore 560012, India These authors contributed equally Corresponding author, email: [email protected] 29 S1. k.p quasiparticle Hamiltonian for few-layer TMDCs The k.p quasi-particle Hamiltonian 𝐻𝑙 used in this work for an arbitrary number of layers 𝑙-1 𝑙 𝑙+1 Below, the specific Hamiltonian for bilayer (2L), tri-layer (3L) and four-layer (4L) systems are given. 30 000000()00000()0000000()00000()0000000()00000()00000izxyizxyzzizxylizxyzzizxyizxyzzatkikatkikstatkikHtatkikstatkiktatkiks⊥⊥−+−+=−−+−2()00()000()0()izxyizxyzzLizxyizxyzzatkikatkikstHatkiktatkiks⊥⊥+−−=−+3()0000()00000()000()00000()000()izxyizxyzzizxyLizxyzzizxyizxyzzatkikatkikstatkikHtatkikstatkiktatkiks⊥⊥⊥⊥+−−−=++−− 𝑙-1 𝑙 𝑙+1 Here, is the monolayer bandgap, is the lattice constant, is the nearest-neighbour intra-layer hopping, is the spin-valley coupling for holes in monolayer, is the interlayer hopping for is the valley degree of freedom (+1 for 𝐾 and −1 for 𝐾′), and sz is spin degree of freedom holes, (±1). S2. Decay rate calculation for excitons in an 𝒏-layer system The electronic band structure of WSe2 in the basis of 5𝑑𝑧2 denoting the layer index) is obtained by expanding the monolayer 𝑘. 𝑝 Hamiltonian upon incorporating the interlayer coupling of the VBs with the immediate neighbouring layers. ⟩ + 𝑖𝜏5𝑑𝑥𝑦 (5𝑑𝑥2−𝑦2 𝑙 ⟩) (𝑙 𝑙 ⟩ and 1 √2 𝑙 Δ is the quasi-particle energy gap at 𝐾, 𝐾′, 𝑎 is the lattice constant and 𝑡 is the effective hopping integral. 𝑐, 𝐤⟩, 𝑣, 𝐤⟩ denotes the conduction band (𝑐) and valence band (𝑣) state at k at energy 𝜀𝑐(𝐤) and 𝜀𝑣(𝐤). An exciton state Χ(𝐐)⟩ at a centre of mass momentum 𝐐 is a coherent superposition of electrons and holes from band-pairs (𝑣, 𝑐) in an 𝑛-layer system in the reciprocal space as Χ(𝐐)⟩ = ∑ Υ(𝐐) 𝑣, 𝐤⟩𝑐, 𝐤 + 𝐐⟩ 𝑣,𝑐,𝐤 31 4()000000()0000000()00000()0000000()00000()0000000()00000()izxyizxyzzizxyizxyzzLizxyizxyzzizxyizxyzzatkikatkikstatkiktatkikstHatkiktatkikstatkiktatkiks⊥⊥⊥⊥⊥⊥+−−−+=+−−−+az000000()00000()0000000()00000()0000000()00000()00000izizzzizizzzizizzzatkxikyatkxikystatkxikytatkxikystatkxikytatkxikys⊥⊥−+−+−−+−it⊥tThe solution of the Bethe-Salpeter equation for Χ(𝐐)⟩ is the exciton eigen energy 𝐸𝑠(𝐐) and wavefunction Ψ𝑠(𝐐) = ∑ (𝑠) (𝐤) 𝜆v,c,𝐐 𝑣,𝑐,𝐤 𝑣, 𝐤⟩𝑐, 𝐤 + 𝐐⟩, for an exciton in band 𝑠. In the dipole approximation, the momentum matrix element given by 𝐏𝑣,𝑐,𝐐(𝐤) = ⟨𝑣, 𝐤𝐩𝑐, 𝐤 + 𝐐⟩ = 𝑚 ħ (𝜀𝑐(𝐤 + 𝐐) − 𝜀𝑣(𝐤)) ⟨𝑣, 𝐤 ∂ ∂𝐤 𝑐, 𝐤 + 𝐐⟩ Following this, the quantity 𝜒𝑒𝑥(𝐐) including contributions from the band-pairs (𝑣, 𝑐) and radiative decay rate Γ𝑅(𝐐) is obtained by 𝜒𝑒𝑥(𝐐) = ∑ (∫ 𝑣,𝑐 𝑑2𝐤 (2𝜋)2 𝐏𝑣,𝑐,𝐐(𝐤). 𝑥 𝜆v,c,𝐐 (𝑠) (𝐤)) Γ𝑅(𝐐) = 𝜂𝑜 2 𝜒𝑒𝑥(𝐐)2 ∫ 𝑑𝑞𝑧 ħ𝑒2 2𝑚𝑜 ∞ 0 1 √𝑄2 + 𝑞𝑧 2 × (1 + 2 𝑞𝑧 𝑄2 + 𝑞𝑧 2) × Γ(𝐐) 𝜋⁄ [𝐸𝑒𝑥(𝐐) − ħ𝑐√𝑄2 + 𝑞𝑧 2] + Γ(𝐐)2 32 S3. Radiative decay rate for excitons in tri-layer (3L) WSe2 Figure S3. Decay rate calculation for three-layer (3L) excitons with two doubly-degenerate spin valley configurations, taking Γ𝑛𝑟 = 0 meV. While the 𝐴1𝑠 (3) shows the highest decay rate for both the cases, the other two excitons possess low decay rate. 33 0.04.04.55.010-5010-4010-3010-2010-101001010 0.04.04.55.010-5010-4010-3010-2010-101001010Decay rate (𝑠−1) ( ),Γ𝑛𝑟=0 𝑚𝑒𝑉 ( ), Γ𝑛𝑟=0 𝑚𝑒𝑉 ,( ), Γ𝑛𝑟=0 𝑚𝑒𝑉 10−4×2𝜋 𝑎S4. Summary of layer dependent excitonic states Table for experimental excitation energies of excitons: No of layers Energy of A1s Energy of A2s Energy of A3s 1 2 3 6 1.746±0.006 (1.743) 1.876±0.008 (1.894) 1.940±0.006 (1.935) 1.743±0.005 (1.746) 1.875±0.015 (1.887) 1.934±0.007 (1.922) 1.746±0.007 (1.75) 1.881±0.007 (1.881) (1.91) 1.741±0.007 (1.745) - - * all the values are in eV. Values are in parenthesis are calculated values for the bright exciton. Table for experimentally extracted binding energies of excitons: No of layers BE of A1s BE of A2s BE of A3s 1 2 3 6 0.447±0.006 (0.45) 0.317±0.008 (0.299) 0.252±0.006 (0.258) 0.44±0.005 (0.437) 0.308±0.015 (0.296) 0.249±0.007 (0.261) 0.427±0.007 (0.423) 0.292±0.007 (0.293) (0.264) 0.422±0.007 (0.418) - - All the values are in eV. Values are in parenthesis are calculated values for the bright exciton. 34 Summary of calculated energy eigenvalues for excitonic states in bilayer and tri-layer WSe2 : 𝜏𝑧. 𝑠𝑧 = +1 System A1s A2s A3s 1.741 (1) 𝐴2𝑠 1.884 (1) 𝐴3𝑠 1.920 1.746 (2) 𝐴2𝑠 1.887 (2) 𝐴3𝑠 1.922 𝜏𝑧. 𝑠𝑧 = −1 A2s A3s 1.741 (1) 𝐴2𝑠 1.884 (1) 𝐴3𝑠 1.920 1.746 (2) 𝐴2𝑠 1.887 (2) 𝐴3𝑠 1.922 A1s (1) 𝐴1𝑠 (2) 𝐴1𝑠 2L WSe2 3L WSe2 (1) 𝐴1𝑠 (2) 𝐴1𝑠 (1) 𝐴1𝑠 (2) 𝐴1𝑠 1.746 1.746 (3) 𝐴1𝑠 1.75 (1) 𝐴2𝑠 (2) 𝐴2𝑠 (3) 𝐴2𝑠 1.877 1.877 1.881 (1) 𝐴3𝑠 (2) 𝐴3𝑠 (3) 𝐴3𝑠 1.908 (1) 𝐴1𝑠 1.746 1.908 (2) 𝐴1𝑠 1.75 1.909 (3) 𝐴1𝑠 1.75 (1) 𝐴2𝑠 (2) 𝐴2𝑠 (3) 𝐴2𝑠 1.877 1.881 1.881 (1) 𝐴3𝑠 (2) 𝐴3𝑠 (3) 𝐴3𝑠 1.908 1.909 1.909 All the values are in eV. Energy values marked in bold front are used for experimental analysis. 35 S5. Layer dependent inhomogeneous broadening of exciton emission linewidth Figure S5. The inhomogeneous (Gaussian) component of the 𝐴1𝑠 exciton linewidth as a function of number of layers of the WSe2 film, as obtained from the Voigt fitting. The inhomogeneous broadening reduces from 1L to 3L, likely due to improved isolation from SiO2 substrate. However, at 6L, the inhomogeneous broadening is found to be quite large. One possible reason could be due to the layered induced spatial spread of the exciton wavefunction in the different layers, effectively increasing the total inhomogeneity of the sample, with a contribution from each layer. 36 123456681012141618Gaussian Broadening (meV)Number of layers
1705.06895	2	1705	2018-08-17T17:57:50	Topological dynamics and excitations in lasers and condensates with saturable gain or loss	[ "cond-mat.mes-hall", "physics.optics" ]	We classify symmetry-protected and symmetry-breaking dynamical solutions for nonlinear saturable bosonic systems that display a non-hermitian charge-conjugation symmetry, as realized in a series of recent groundbreaking experiments with lasers and exciton polaritons. In particular, we show that these systems support stable symmetry-protected modes that mirror the concept of zero-modes in topological quantum systems, as well as symmetry-protected power-oscillations with no counterpart in the linear case. In analogy to topological phases in linear systems, the number and nature of symmetry-protected solutions can change. The spectral degeneracies signalling phase transitions in linear counterparts extend to bifurcations in the nonlinear context. As bifurcations relate to qualitative changes in the linear stability against changes of the initial conditions, the symmetry-protected solutions and phase transitions can also be characterized by topological excitations, which set them apart from symmetry-breaking solutions. The stipulated symmetry appears naturally when one introduces nonlinear gain or loss into spectrally symmetric bosonic systems, as we illustrate for one-dimensional topological laser arrays with saturable gain and two-dimensional flat-band polariton condensates with density-dependent loss.	cond-mat.mes-hall	cond-mat	a Topological dynamics and excitations in lasers and condensates with saturable gain or loss SIMON MALZARD,1 EMILIANO CANCELLIERI,1,2 AND HENNING SCHOMERUS1,* 1Department of Physics, Lancaster University, Lancaster, LA1 4YB, UK 2Department of Physics and Astronomy, University of Sheffield, Sheffield, S3 7RH, UK *[email protected] Abstract: We classify symmetry-protected and symmetry-breaking dynamical solutions for nonlinear saturable bosonic systems that display a non-hermitian charge-conjugation symmetry, as realized in a series of recent groundbreaking experiments with lasers and exciton polaritons. In particular, we show that these systems support stable symmetry-protected modes that mirror the concept of zero-modes in topological quantum systems, as well as symmetry-protected power-oscillations with no counterpart in the linear case. In analogy to topological phases in linear systems, the number and nature of symmetry-protected solutions can change. The spectral degeneracies signalling phase transitions in linear counterparts extend to bifurcations in the nonlinear context. As bifurcations relate to qualitative changes in the linear stability against changes of the initial conditions, the symmetry-protected solutions and phase transitions can also be characterized by topological excitations, which set them apart from symmetry-breaking solutions. The stipulated symmetry appears naturally when one introduces nonlinear gain or loss into spectrally symmetric bosonic systems, as we illustrate for one-dimensional topological laser arrays with saturable gain and two-dimensional flat-band polariton condensates with density-dependent loss. OCIS codes: (140.3430) Laser theory; (140.3290) Laser arrays; (190.3100) Instabilities and chaos; (270.2500) Fluctuations, relaxations, and noise. References and links 1. M. Z. Hasan and C. L. Kane, "Colloquium: Topological insulators," Rev. Mod. Phys. 82, 3045 -- 3067 (2010). 2. X.-L. Qi and S.-C. Zhang, "Topological insulators and superconductors," Rev. Mod. Phys. 83, 1057 -- 1110 (2011). 3. C. W. J. Beenakker, "Random-matrix theory of Majorana fermions and topological superconductors," Rev. Mod. Phys. 87, 1037 -- 1066 (2015). 5. 4. S. Ryu, A. P. Schnyder, A. Furusaki, and A. W. W. Ludwig, "Topological insulators and superconductors: tenfold way and dimensional hierarchy," New J. Phys. 12, 065010 (2010). J. C. Y. Teo and C. L. Kane, "Topological defects and gapless modes in insulators and superconductors," Phys. Rev. B 82, 115120 (2010). 6. P. Heinzner, A. Huckleberry, and M. R. Zirnbauer, "Symmetry classes of disordered fermions," Commun. Math. Phys. 257, 725 -- 771 (2005). 7. L. Fidkowski and A. Kitaev, "Topological phases of fermions in one dimension," Phys. Rev. B 83, 075103 (2011). J. Alicea, "New directions in the pursuit of Majorana fermions in solid state systems," Rep. Prog. Phys. 75, 076501 8. (2012). 9. M. Leijnse and K. Flensberg, "Introduction to topological superconductivity and Majorana fermions," Semicond. Sci. Technol. 27, 124003 (2012). 113 -- 136 (2013). 10. C. W. J. Beenakker, "Search for Majorana fermions in superconductors," Annu. Rev. Condens. Matter Phys. 4, 11. A. Yu. Kitaev, "Unpaired Majorana fermions in quantum wires," Phys. Usp. 44 (suppl.), 131 -- 136 (2001). 12. L. Fu and C. L. Kane, "Josephson current and noise at a superconductor/quantum-spin-hall-insulator/superconductor junction," Phys. Rev. B 79, 161408 (2009). 13. C. W. J. Beenakker, D. I. Pikulin, T. Hyart, H. Schomerus, and J. P. Dahlhaus, "Fermion-parity anomaly of the critical supercurrent in the quantum spin-hall effect," Phys. Rev. Lett. 110, 017003 (2013). 14. R. J. Potton, "Reciprocity in optics," Rep. Prog. Phys. 67, 717 -- 754 (2004). 15. L. Lu, J. D. Joannopoulos, and M. Soljačić, "Topological photonics," Nat. Photon. 8, 821 -- 829 (2014). 16. K. Y. Bliokh, D. Smirnova, and F. Nori, "Quantum spin Hall effect of light," Science 348, 1448 -- 1451 (2015). 17. V. Peano, M. Houde, C. Brendel, F. Marquardt, and A. A. Clerk, "Topological phase transitions and chiral inelastic transport induced by the squeezing of light," Nat. Commun. 7, 10779 (2016). 18. V. Peano, C. Brendel, M. Schmidt, and F. Marquardt, "Topological phases of sound and light," Phys. Rev. X 5, 031011 19. V. Peano, M. Houde, F. Marquardt, and A. A. Clerk, "Topological quantum fluctuations and traveling wave amplifiers," 20. R. Süsstrunk and S. D. Huber, "Observation of phononic helical edge states in a mechanical topological insulator," 21. Z. Yang, F. Gao, X. Shi, X. Lin, Z. Gao, Y. Chong, and B. Zhang, "Topological acoustics," Phys. Rev. Lett. 114, 22. C. L. Kane and T. C. Lubensky, "Topological boundary modes in isostatic lattices," Nat. Phys. 10, 39 -- 45 (2014). 23. N. Goldman, J. C. Budich, and P. Zoller, "Topological quantum matter with ultracold gases in optical lattices," Nat. (2015). Phys. Rev. X 6, 041026 (2016). Science 349, 47 -- 50 (2015). 114301 (2015). Phys. 12, 639 -- 645 (2016). 24. V. G. Sala, D. D. Solnyshkov, I. Carusotto, T. Jacqmin, A. Lemaître, H. Terças, A. Nalitov, M. Abbarchi, E. Galopin, I. Sagnes, J. Bloch, G. Malpuech, and A. Amo, "Spin-orbit coupling for photons and polaritons in microstructures," Phys. Rev. X 5, 011034 (2015). 25. T. Karzig, C.-E. Bardyn, N. H. Lindner, and G. Refael, "Topological polaritons," Phys. Rev. X 5, 031001 (2015). 26. A. V. Nalitov, D. D. Solnyshkov, and G. Malpuech, "Polariton Z topological insulator," Phys. Rev. Lett. 114, 116401 27. L. P. Pitaevskii and S. Stringari, Bose-Einstein Condensation (Oxford University, Oxford, 2003). 28. O. Morsch and M. Oberthaler, "Dynamics of Bose-Einstein condensates in optical lattices," Rev. Mod. Phys. 78, 29. C. Ciuti and I. Carusotto, "Quantum fluid effects and parametric instabilities in microcavities," Phys. Status Solidi B 30. Y. Kawaguchi and M. Ueda, "Spinor Bose-Einstein condensates," Phys. Rep. 520, 253 -- 381 (2012). 31. R. Barnett, "Edge-state instabilities of bosons in a topological band," Phys. Rev. A 88, 063631 (2013). 32. R. Shindou, R. Matsumoto, S. Murakami, and J.-I. Ohe, "Topological chiral magnonic edge mode in a magnonic 33. G. Engelhardt and T. Brandes, "Topological Bogoliubov excitations in inversion-symmetric systems of interacting (2015). 179 -- 215 (2006). 242, 2224 -- 2245 (2005). crystal," Phys. Rev. B 87, 174427 (2013). bosons," Phys. Rev. A 91, 053621 (2015). Phys. Rev. B 93, 020502 (2016). lattices," New J. Phys. 17, 115014 (2015). Rev. Lett. 115, 245302 (2015). Rev. Lett. 117, 045302 (2016). 065703 (2009). Phys. Rev. Lett. 110, 013903 (2013). 34. C.-E. Bardyn, T. Karzig, G. Refael, and T. C. H. Liew, "Chiral Bogoliubov excitations in nonlinear bosonic systems," 35. S. Furukawa and M. Ueda, "Excitation band topology and edge matter waves in Bose-Einstein condensates in optical 36. B. Galilo, D. K. K. Lee, and R. Barnett, "Selective population of edge states in a 2D topological band system," Phys. 37. G. Engelhardt, M. Benito, G. Platero, and T. Brandes, "Topological instabilities in ac-driven bosonic systems," Phys. 38. M. S. Rudner and L. S. Levitov, "Topological transition in a non-hermitian quantum walk," Phys. Rev. Lett. 102, 39. H. Schomerus and N. Y. Halpern, "Parity anomaly and Landau-level lasing in strained photonic honeycomb lattices," 40. H. Schomerus, "Topologically protected midgap states in complex photonic lattices," Opt. Lett. 38, 1912 -- 1914 41. C. Poli, M. Bellec, U. Kuhl, F. Mortessagne, and H. Schomerus, "Selective enhancement of topologically induced interface states in a dielectric resonator chain," Nat. Commun. 6, 6710 (2015). 42. D. Leykam, S. Flach, and Y. D. Chong, "Flat bands in lattices with non-Hermitian coupling," Phys. Rev. B 96, 064305 43. D. I. Pikulin and Yu. V. Nazarov, "Topological properties of superconducting junctions," JETP Lett. 94, 693 -- 697 44. D. I. Pikulin and Y. V. Nazarov, "Two types of topological transitions in finite Majorana wires," Phys. Rev. B 87, (2013). (2017). (2012). 235421 (2013). 45. S. Malzard, C. Poli, and H. Schomerus, "Topologically protected defect states in open photonic systems with non-hermitian charge-conjugation and parity-time symmetry," Phys. Rev. Lett. 115, 200402 (2015). 46. P. San-Jose, J. Cayao, E. Prada, and R. Aguado, "Majorana bound states from exceptional points in non-topological superconductors," Sci. Rep. 6, 21427 (2016). 47. J. Avila, F. Peñaranda, E. Prada, P. San-Jose, and R. Aguado, "Non-hermitian topology: a unifying framework for the Andreev versus Majorana states controversy," arXiv:1807.04677 (2018). 48. P. St-Jean, V. Goblot, E. Galopin, A. Lemaître, T. Ozawa, L. Le Gratiet, I. Sagnes, J. Bloch, and A. Amo, "Lasing in topological edge states of a 1D lattice," Nat. Photon. 11, 651-656 (2017). 49. C. E. Whittaker, E. Cancellieri, P. M. Walker, D. R. Gulevich, H. Schomerus, D. Vaitiekus, B. Royall, D. M. Whittaker, E. Clarke, I. V. Iorsh, I. A. Shelykh, M. S. Skolnick, and D. N. Krizhanovskii, "Exciton-polaritons in a two-dimensional Lieb lattice with spin-orbit coupling," Phys. Rev. Lett. 120, 097401 (2018). 50. H. Zhao, P. Miao, M. H. Teimourpour, S. Malzard, R. El-Ganainy, H. Schomerus, and L. Feng, "Topological hybrid silicon microlasers," Nat. Commun. 9, 981 (2018). 51. M. Parto, S. Wittek, H. Hodaei, G. Harari, M. A. Bandres, J. Ren, M. C. Rechtsman, M. Segev, D. N. Christodoulides, and M. Khajavikhan, "Edge-mode lasing in 1D topological active arrays," Phys. Rev. Lett. 120, 113901 (2018). topological edge state lasers," arXiv:1804.01587 (2018). 52. R. Yao, H. Li, B. Zheng, S. An, J. Ding, C.-S. Lee, H. Zhang, and W. Guo, "Electrically tunable and reconfigurable 53. These considerations can also be extended to include an inhomogeneity g(t) = −Xg∗(t). 54. This exploits the U(1) gauge freedom. 55. At the juncture of cases (a) and (b), the nonlinearities can also stabilize symmetry-breaking modes with vanishing frequency, which then form a degenerate pair. 56. This again exploits the U(1) gauge freedom to fix the overall phase factor exp(iα) of the wavefunction, up to an overall sign. If a self-symmetric state is periodic this can therefore be realized in two variants Ψ(T) = Ψ(0) and Ψ(T) = −Ψ(0). 57. B. Sutherland, "Localization of electronic wave functions due to local topology," Phys. Rev. B 34, 5208 -- 5211 (1986). 58. W. P. Su, J. R. Schrieffer, and A. J. Heeger, "Solitons in polyacetylene," Phys. Rev. Lett. 42, 1698 -- 1701 (1979). 59. S. Ryu and Y. Hatsugai, "Topological origin of zero-energy edge states in particle-hole symmetric systems," Phys. 60. E. H. Lieb, "Two theorems on the Hubbard model," Phys. Rev. Lett. 62, 1201 -- 1204 (1989). 61. H. Aoki, M. Ando, and H. Matsumura, "Hofstadter butterflies for flat bands," Phys. Rev. B 54, R17296 -- R17299 Rev. Lett. 89, 077002 (2002). (1996). 62. N. Malkova, I. Hromada, X. Wang, G. Bryant, and Z. Chen, "Observation of optical Shockley-like surface states in photonic superlattices," Opt. Lett. 34, 1633 -- 1635 (2009). 63. R. Shen, L. B. Shao, B. Wang, and D. Y. Xing, "Single Dirac cone with a flat band touching on line-centered-square optical lattices," Phys. Rev. B 81, 041410 (2010). 64. V. Apaja, M. Hyrkäs, and M. Manninen, "Flat bands, Dirac cones, and atom dynamics in an optical lattice," Phys. Rev. A 82, 041402 (2010). Rev. A 83, 063601 (2011). 65. N. Goldman, D. F. Urban, and D. Bercioux, "Topological phases for fermionic cold atoms on the Lieb lattice," Phys. 66. M. Atala, M. Aidelsburger, J. T. Barreiro, D. Abanin, T. Kitagawa, E. Demler, and I. Bloch, "Direct measurement of the Zak phase in topological Bloch bands," Nat. Phys. 9, 795 -- 800 (2013). 67. R. A. Vicencio, C. Cantillano, L. Morales-Inostroza, B. Real, C. Mejía-Cortés, S. Weimann, A. Szameit, and M. I. Molina, "Observation of localized states in Lieb photonic lattices," Phys. Rev. Lett. 114, 245503 (2015). 68. S. Mukherjee, A. Spracklen, D. Choudhury, N. Goldman, P. Öhberg, E. Andersson, and R. R. Thomson, "Observation of a localized flat-band state in a photonic Lieb lattice," Phys. Rev. Lett. 114, 245504 (2015). 69. J. M. Zeuner, M. C. Rechtsman, Y. Plotnik, Y. Lumer, S. Nolte, M. S. Rudner, M. Segev, and A. Szameit, "Observation of a topological transition in the bulk of a non-hermitian system," Phys. Rev. Lett. 115, 040402 (2015). 70. F. Baboux, L. Ge, T. Jacqmin, M. Biondi, E. Galopin, A. Lemaître, L. Le Gratiet, I. Sagnes, S. Schmidt, H. E. Türeci, A. Amo, and J. Bloch, "Bosonic condensation and disorder-induced localization in a flat band," Phys. Rev. Lett. 116, 066402 (2016). 71. H.-I. Lu, M. Schemmer, L. M. Aycock, D. Genkina, S. Sugawa, and I. B. Spielman, "Geometrical pumping with a Bose-Einstein condensate," Phys. Rev. Lett. 116, 200402 (2016). 72. C. Poli, H. Schomerus, M. Bellec, U. Kuhl, and F. Mortessagne, "Partial chiral symmetry-breaking as a route to spectrally isolated topological defect states in two-dimensional artificial materials," 2D Materials 4, 025008 (2017). 73. M. Tlidi, P. Mandel, and M. Haelterman, "Spatiotemporal patterns and localized structures in nonlinear optics," Phys. 74. G. Wen, D. Xu, and X. Han, "On creation of Hopf bifurcations in discrete-time nonlinear systems," Chaos 12, Rev. E 56, 6524 -- 6530 (1997). 350 -- 355 (2002). 75. T. Kitagawa, M. A. Broome, A. Fedrizzi, M. S. Rudner, E. Berg, I. Kassal, A. Aspuru-Guzik, E. Demler, and A. G. White, "Observation of topologically protected bound states in photonic quantum walks," Nat. Comm. 3, 882 (2012). 76. T. Harayama, S. Sunada, and K. S. Ikeda, "Theory of two-dimensional microcavity lasers," Phys. Rev. A 72, 013803 (2005). Lett. 100, 250401 (2008). 77. J. Keeling and N. G. Berloff, "Spontaneous rotating vortex lattices in a pumped decaying condensate," Phys. Rev. 78. S. Malzard, and H. Schomerus, "Nonlinear mode competition and symmetry-protected power oscillations in topological lasers," New J. Phys. 20, 063044 (2018). 79. S. Longhi, Y. Kominis, and V. Kovanis, "Presence of temporal dynamical instabilities in topological insulator lasers," Europhys. Lett. 122, 14004 (2018). Introduction 1. A wide range of topological quantum effects manifest themselves in symmetries of an excitation spectrum. This relation has been explored extensively for fermionic systems, for which the single-particle Hamiltonian can obey a chiral symmetry or a charge-conjugation symmetry [1 -- 3]. Allowing also for time-reversal invariance one can identify ten universality classes featuring a variety of topological quantum numbers. These quantum numbers determine the formation of zero modes and unidirectional transport channels at edges, surfaces, and interfaces of various dimensions [4,5]. In the fermionic case these symmetries are fundamental as they are owed to the structure of Fock or Nambu space [6], and therefore also extend to the interacting case [7]. Prime examples are superconducting systems, for which a charge-conjugation symmetry dictates that excitations ψ exp(−iωt) at a positive frequency ω are paired with excitations Xψ∗ exp(iωt) at the corresponding negative frequency −ω, where X is a suitable unitary transformation. The systems can then also feature robust Majorana zero modes ψ = Xψ∗ at ω = 0 [8 -- 10], an effect which translates to the fermion parity anomaly in the full many-body theory [11 -- 13]. Some of these symmetries also play a natural role beyond the fermionic context. For instance, time-reversal symmetry is intimately related to optical reciprocity [14], a classical concept whose modification leads to topological effects in photonic structures [15] and analogous optical [16 -- 19], acoustic [20,21] and mechanical [22] systems. Topological effects can also be engineered into bosonic quantum systems, such as cold atomic gases [23] and exciton polaritons [24 -- 26]. For weakly interacting bosons the excitations are again characterized by a Bogoliubov theory [27 -- 30], and recent work has explored a variety of mechanisms to engineer topological features into this description [31 -- 37]. The charge-conjugation symmetry can also be induced into the single-particle theory of bosonic systems, where one again can admit linear gain or loss [38 -- 42]. The spectrum of the effective Hamiltonian then displays symmetric pairs of complex frequencies (Ωn, −Ω∗ n), which can be interpreted as resonances. Topological features persist because this spectral constraint can enforce a number of purely imaginary self-symmetric resonances Ωn = −Ω∗ n, which provide an analogue to broadened fermionic zero modes [43 -- 47]. Five recent experiments aimed at realizing topological zero modes in polaritonic condensates [48,49] and lasers [50 -- 52]. As these are inherently nonlinear systems, the question arises whether the notion of zero modes and topological protection persists. Here, we provide a unifying perspective on these systems based on a notion of charge-conjugation symmetry that applies directly to the time-dependent nonlinear wave equation. Our general strategy is as follows: Topological states in linear systems are protected by symmetry, but their number can change discretely in phase transitions, which are generally linked to degeneracies (such as when a band gap closes). This notion is underpinned by the continuity of the spectrum under smooth parameter changes (deformations of the system), a feature at the heart of linear spectral analysis. Analogously, we show that nonlinear systems can display dynamical solutions that are protected by symmetry. Their number can change in dynamical degeneracies, which correspond to bifurcations. This notion is then underpinned by the structural stability of dynamical solutions, which is at the heart of nonlinear dynamics. This nonlinear notion of topological states reduces to the conventional spectral notions in the linear limit. New effects also appear: We identify a new class of topological solutions that does not have a linear counterpart, and also exploit the link of bifurcations to a qualitative change of the linear stability against changes of initial conditions, which again does not feature in the linear context. In detail, as we will see, nonlinearities in the gain or loss lead to complex-wave dynamics where the time-dependent solutions appear in pairs Ψ(t) and X Ψ∗(t). In contrast to a general time-reversal symmetry, which induces a partner solution X Ψ∗(−t), the solutions Ψ(t) and X Ψ∗(t) both exhibit the same arrow of time. This time-preserving symmetry therefore allows for self-symmetric states -- including stationary states Ψ(t) = Ψ0 = X Ψ∗ 0, which we interpret as zero modes. As a notable feature without analogue in the linear case, the symmetry also protects a twisted variant of time-dependent power-oscillating states with Ψ(t + T/2) = X Ψ∗(t), for which periodicity Ψ(t + T) = Ψ(t) is enforced by symmetry -- suggesting that the nonlinear setting admits for a richer notion of protected states than the linear case. That these states enjoy topological protection can be further ascertained by identifying topological excitations in the stability spectrum, which we find to govern the spectral phase transitions between the different types of symmetry-protected solutions -- corresponding to different topological phases of the dynamical system. We illustrate our findings for two model systems representing one-dimensional topological laser arrays with saturable gain and two-dimensional flat-band polariton condensates with density-dependent loss. 2. Results 2.1. Model and classification of states Consider a bosonic system whose classical limit is described by a complex-wave equation Ψ(t) = HΨ(t) + F[Ψ(t), Ψ∗(t)]Ψ(t). i d dt (1) The operator H provides an effective single-particle description, while the functional F describes the nonlinear effects which may encompass gain and loss. We assume these effects to be local in time and often will drop the time argument. In keeping with the quantum origin of the wave function, the nonlinear effects should preserve the global U(1) gauge freedom, so that any solution can be multiplied by a fixed phase factor exp(iα). This can be guaranteed by assuming FΨΨ = FΨ∗ Ψ∗, where subscripts denote functional derivatives. In addition, we here stipulate that H displays a charge-conjugation symmetry, XHX = −H∗ where X is a unitary operator with X2 = 1. To extend this notion to the nonlinear case, we similarly demand that XF[Ψ, Ψ∗]X = −(F[X Ψ∗, X Ψ])∗, with the same operator X. We will describe practical ways to realize such systems soon below but first discuss the consequences of the following general observation. For any solution Ψ(t) of Eq. (1), we obtain a partner solution (2) Given our assumptions this correspondence even applies when H and F contain an explicit time dependence [53]; however, we focus on the autonomous case. Let us first reflect on the possible classes of solutions admitted by Eq. (1). Because of the underlying charge-conjugation symmetry of H, the linear system possesses pairs of solutions Ψ(t) = Ψn exp(−iΩnt) and Ψ(t) = X Ψ∗ nt), where the latter expression corresponds to a frequency Ωn = −Ω∗ n. This includes the possibility of topologically protected zero modes with purely imaginary frequency Ωn = Ωn = i Im Ωn and time dependence Ψ(t) = Ψ(t) = Ψn exp(Im Ωnt) [39 -- 41, 43 -- 47]. If for any of these states Im Ωn is positive the linear system is unstable, but the nonlinear effects can stabilize the system. This results in stationary states Ψ(t) = Ψn exp(−iΩnt) with real Ωn, which are self-consistent solutions of the condition [54] n exp(iΩ∗ Ψ(t) = X Ψ∗(t). ΩnΨn = HΨn + F[Ψn, Ψ∗ n]Ψn. (3) The solutions either and chaotic solutions. These will either (a) occur in pairs Ψn, Ψn where Ωn and Ωn = −Ωn now are both real, or (b) are time-independent zero modes Ψ0 = Ψ0 whose frequency Ω0 = 0 now vanishes [55]. Alternatively, the system may tend to time-dependent solutions, including periodic, aperiodic (c) still occur in pairs Ψ(t), Ψ(t) that bear no further relation besides Eq. (2), or may be (d) The time-dependent solution may be self-symmetric, Ψ(t) = Ψ(t); given this condition at some point of time, it will remain true for all times [56]. (e) Two partner solutions may be related by a finite time-offset, Ψ(t) = Ψ(t + T/2). It then follows that the solutions must be periodic, Ψ(t + T) = Ψ(t), which amounts to the twisted states mentioned in the introduction. constrained in two possible ways: Fig. 1. Illustration of the five types of modes for a topological laser array based on a Su-Schrieffer-Heeger chain with background loss γA = γB = 0.3 and various amounts of saturable gain [see Eqs. (4), (5) and (6)]. (a) Stationary symmetry-breaking mode (gA = 0.4, gB = 0.7). (b) Stationary self-symmetric zero mode (gA = 0.8, gB = 0.0). (c) Oscillating symmetry-breaking mode (gA = 0.7, gB = 0.4). (d) Oscillating self-symmetric mode (gA = 0.7, gB = 0.4 with symmetry-preserving initial conditions). (e) Twisted oscillating mode (gA = 0.8, gB = 0.2). The sketches at the very top symbolize the traces of the solutions in low-dimensional cross-sections of the dynamical phase space, where in (a-d) the symmetry is represented as a reflection and in (e) as a rotation. In the second row, the circles represent the resonators, where the area denotes the intensity (A and B sublattice in red and blue; in the time-dependent case, we show two circles corresponding to the largest and smallest intensity over a cycle.) The third row shows the time traces of the intensities IA = A2 (red), IB = B2 (blue), and Itot = IA + IB (black). The bottom panels in (a,b) show the stability spectra of the stationary states, while in (c-e) they show the correlation functions C = (cid:104)Ψ(0)Ψ(t)(cid:105) (orange), C = (cid:104)Ψ(0) Ψ(t)(cid:105) (brown) of the oscillating states. The nature of a state can be assessed, e.g., by comparing the correlation functions C = (cid:104)Ψ(0)Ψ(t)(cid:105), C = (cid:104)Ψ(0) Ψ(t)(cid:105), which coincide or alternate for self-symmetric or twisted solutions. Based on the possible invariance of time-parameterized orbits under discrete involutions, this categorization is minimal and complete. Note that the solutions of class (a), (d) and (e) all describe orbits Ψ(t) that are invariant under the symmetry operation (2). In the self-symmetric cases (a) and (d), this invariance is local at every point along the orbit, while for the twisted states (e) the invariance occurs under a translation by T/2. The latter leaves the orbit invariant as the solution is periodic. All these solutions are furthermore symmetry-protected, in that their number can only change due to structural reconfigurations of the dynamical solutions. That such changes are indeed possible at all is necessary in order to have a setting with meaningful topological features, and not just symmetry-imposed constraints. In linear systems, the phase transitions at which the number and nature of symmetry-protected solutions changes involves degeneracies; as we will explore further in Section 2.3, for the nonlinear setting this naturally extends to bifurcations. First, however, we provide concrete model systems with the time-preserving symmetry of Eq. (1) and examples of the different classes of solutions. 2.2. Realization in lasers and condensates Figures 1 and 2 illustrate the different types of solutions for two model systems, which are both formed of bipartite lattices (NA sites A and NB sites B) that give rise to a pseudospin degree of -0.4 0-2.5 2.5Im ωRe ω-0.4 0-2 2ω−ω+Im ωRe ω 0 20 0 50Itot,A,Bt 0 20 0 50Itot,A,Bt 0 20 0 50Itot,A,Bt 0 20 0 50Itot,A,Bt 0 20 0 50Itot,A,Bt 0 20 0 50C,C~t 0 20 0 50C,C~t 0 20 0 50C,C~t(a)(b)(c)(d)(e)Fig. 2. Same as Fig. 1 but for a polaritonic flat-band condensate based on a Lieb lattice with linear gain and density-dependent loss [see Eq. (7)]. To demonstrate the generality of our findings we include 50% relative disorder in all parameters, including for the couplings around their mean value tkl = 1, and the losses with average strength γA = γB = 0.3 (see Fig. 3 for details of the configuration). (a) Stationary symmetry-breaking mode (average gain gA = 0.15, gB = 0.3). (b) Stationary self-symmetric zero mode (average background loss gA = −0.2 and gain gB = 0.5). (c) Slowly oscillating symmetry-breaking mode (gA = 0.35, gB = 0.2). (d) Oscillating self-symmetric mode (gA = 0.5, gB = 0.3 with symmetry-preserving initial conditions). (e) Twisted oscillating mode (gA = 0.1, gB = 0.4). freedom. The dynamics is governed by the equations Ak = Bk = <l> tklBl + fA(Ak2)Ak, tlk Al + fB(Bk2)Bk, (4) (5) i i d dt d dt <l> where we grouped the amplitudes on both sublattices into vectors A, B. The real nearest-neighbour couplings tkl form a linear Hamiltonian H that supports at least ν = NA − NB zero modes, irrespective of whether the system is homogenous or not [57]. In the examples this describes a finite segment of a Su-Schrieffer-Heeger (SSH) chain with alternating couplings tk,k+1 = 1, 0.7 (Fig. 1), where a topological midgap state arises from a central coupling defect [40,58,59], or a finite segment of a disordered Lieb lattice (Fig. 2), which exhibits multiple zero modes associated with a flat band [60, 61]. Both of these models can be realized on a large variety of platforms, including bosonic systems [41, 48 -- 52, 62 -- 72]. The detailed geometric configurations in both models are given in Fig. 3. For the SSH chain, we consider nonlinearities that represent saturable gain, fs(Ψk2) = i gs 1 + Ψk2 − iγs (saturable gain), (6) as often adopted in laser models [76], while for the Lieb lattice we consider density-dependent loss, fs(Ψk2) = i(gs − γsΨk2) (density-dependent loss) (7) as considered, e.g., in studies of polaritonic condensates [77]. For any solution with amplitudes A(t) and B(t), Eqs. (4) and (5), exhibits a partner solution with amplitudes A∗(t) and −B∗(t), so that X = 1A ⊕ (−1B) acts as a Pauli-z matrix in pseudospin -0.6 0-5 5Im ωRe ω-1 0-4 4ω-ω+Im ωRe ω 0 20 0 20Itot,A,Bt 0 20 0 20Itot,A,Bt 0 20 0 200Itot,A,Bt 0 20 0 20Itot,A,Bt 0 20 0 20Itot,A,Bt 0 20 0 200C,C~t 0 20 0 20C,C~t 0 20 0 20C,C~t(a)(b)(c)(d)(e)Fig. 3. Detailed geometry of the two illustrative models investigated in this work. (a) The model based on the SSH chain consists of a linear arrangement of 21 sites (11 A sites and 10 B sites) with alternating couplings 1, 0.7. The centre contains a defect with two consecutive couplings 0.7. This separates two configurations, denoted α and β, which can be characterized by topological features of their band structure. In the hermitian model the coupling defect induces one zero mode, which is spatially localized and exhibits an anomalous response to loss and gain of different strength on the two sublattices. For this model, we introduce nonlinearities in the form of saturable gain. (b) The model based on the Lieb lattice also consists of 21 sites, but these are arranged in two dimensions so that 12 are A sites and 9 are B sites. In the hermitian limit, there are now at least three zero modes, even in the case of disorder in the couplings. In this model we study density-dependent losses, and include disorder in the couplings tkl as well as in the gain and loss parameters gA, B and γA, B. This disorder is generated from independent random numbers rn with a box distribution over [0.75, 1.25], so that pn = prn for any model parameter pn with average p. space. Admitting different amounts of gain gA, gB and loss γA, γB on the two sublattices allows us to study the competition between topological and nontopological states, which leads to the different examples shown in the figures (for a detailed mapping of the phase space in the SSH model see [78]). Note that the form of X in these models implies that self-symmetric states display a rigid phase-shift of i between the two sublattices. Given that Ψ(t) = Ψ(t), the amplitudes A(t) are real while the amplitudes B(t) are imaginary, which then persists for all times. On the other hand, symmetry-breaking stationary states with a finite frequency must have equal weight on both sublattices [40]. The different types of states can therefore also be discriminated by their distinct spatial features. 2.3. Phase transitions and topological features of the excitation spectrum The presented classification of solutions is exhaustive in terms of symmetry-protection. To complete the analogy to topological notions in linear systems, it is essential to explore how the number of symmetry-protected solutions can change. In linear systems, this is generally connected to degeneracies, relying on the structural stability of the spectrum against parametric changes. In nonlinear systems, this is paralleled by the notion of structural stability of dynamical solutions (again against parameter changes), with qualitative changes mandated by bifurcations. In this context, we can then exploit the general link of bifurcations to a change of the linear stability (against initial conditions), which allows an emerging spectral interpretation in terms of the stability of linear perturbations. To conclude this paper, we therefore further illuminate the topological aspects of the symmetry-protected solutions by spectral signatures. Here, we 1.00.7AB(a) SSH chaini[gA/(1+Ψk2)−γA]i[gB/(1+Ψk2)−γB]i[gArgA−γA rγAΨk2] i[gBrgB−γBrγBΨk2]AB(b) Lieb latticeαβ0.97900.87650.86881.03441.16990.9527 1.09301.02410.87280.8259 1.20930.87220.93971.05230.81900.79941.21301.09010.82240.84551.02830.78400.78641.1362rgArγArgBrγB1.1389 0.77930.8688 1.17851.1621 0.93641.2329 1.08991.2363 0.87810.9767 0.92381.0545 0.75471.1378 0.92921.0708 1.22451.1110 0.85890.7675 0.90970.8992 1.20890.7660 1.00460.7825 0.98991.0649 1.22781.1869 0.75000.7544 0.87351.1233 1.10611.1564 0.91230.7879 0.88851.0782 1.0977in particular exploit the pinning of excitations to symmetry protected positions, in analogy to Majorana zero modes in fermionic systems with charge-conjugation symmetry [1,2] and analogous zero modes in periodically driven systems [5, 75]. These connections can be established by utilizing the link between dynamical stability of nonlinear systems and their excitation spectrum, which is addressed by Bogoliubov theory. We first present the results of this analysis, and describe the technical details in the following subsection. The analysis amounts to identifying the eigenmodes ψ = (u, v)T of linear perturbations ue−iωt + v∗eiωt, and here results in a spectrum of 2(NA + NB) excitation frequencies ωm. For a stable system all of these excitations must obey Im ωm ≤ 0. For non-self-symmetric pairs of stationary modes Ψn, Ψn, the time-preserving symmetry of Eq. (1) implies that their excitation spectra are identical, but does not impose any further constraints on these spectra. For self- symmetric stationary zero modes Ψ0 = Ψ0 with Ω0 = 0, on the other hand, we can classify the perturbations into symmetry-preserving modes v+ = u+ and symmetry-breaking modes v− = −u−. For our models of gain or loss, these perturbations fulfill the reduced eigenvalue equations ω+u+ = (H + f + 2 f (cid:48)Ψ02)u+, ω−u− = (H + f )u−, (8) (9) where f is diagonal with entries fA(Ak2) or fB(Bk2) (depending on the sublattice), and the prime denotes the derivative with respect to the argument. The excitation spectrum is thus composed of two parts, eigenvalues ω+ that display an enhanced nonlinearity and eigenvalues ω− that coincide with those of the nonlinear wave operator H + f , with Ψ = Ψ0 fixed to the zero mode. Panel (b) in Figs. 1 and 2 shows that the eigenvalues ω− determine the stability of the stationary zero modes while the eigenvalues ω+ lie deep in the complex plane. Therefore, if we restrict the perturbations to preserve the symmetry, the stability of such modes is greatly enhanced. Note that both reduced spectra in these examples have an odd number of excitations. Thus, each reduced spectrum has an odd number of eigenvalues pinned to the imaginary axis. Setting u− = Ψ0 we recover that one of the eigenvalues ω−,0 = Ω0 = 0 vanishes, in accordance with the U(1) symmetry of the wave equation. A similar analysis can be carried out for time-dependent solutions Ψ(t), whose stability is then characterized by a corresponding quasienergy spectrum obtained from a propagator U(t) with eigenvalues λm = exp(−iωmt). This again includes the U(1) Goldstone mode with λ0 = 1, but also an additional Goldstone mode λT = 1 that describes the time-translation freedom of any solution Ψ(t). For general pairs of states Ψ(t), Ψ(t), the time-preserving symmetry again dictates that their quasienergy excitation spectra are identical. For self-symmetric states Ψ(t) = Ψ(t), we can once more separate perturbations that preserve or break the symmetry. This leads to time-dependent variants of Eqs. (8) and (9), u+(t) = [H + f (t) + 2 f (cid:48)(t)Ψ(t)2]u+(t), u−(t) = [H + f (t)]u−(t), (10) i d dt d dt i (11) where f and f (cid:48) inherit their time dependence from Ψ(t). The stability spectrum thus again contains a component ω− inherited from the nonlinear wave operator H + f , which includes the vanishing eigenvalue ω−,0 = 0 (with u−,0 = Ψ(t)) dictated by the U(1) gauge freedom. In addition, there is now also an eigenvalue ω+,0 ≡ ωT = 0 arising from the time-translation freedom of any solution Ψ(t), which is associated with u+,0 = dΨ(t)/dt. For the final case of a twisted periodic state with Ψ(t + T/2) = Ψ(t), the evolution operator of the perturbations factorizes as U(T) = U(cid:48)2 with a half-step propagator U(cid:48) = XΣxU(T/2), where X = 12⊗ X while Σx = σx⊗1 interchanges u and v. The U(1) gauge and time-translation freedoms Fig. 4. (a) Phase transition from a stationary zero mode to a twisted oscillating state of period T as signalled by the linear stability excitation spectrum, here shown for the SSH laser array with gA = 0.693, gB = 0.1. At the transition the excitations match up via the relation λ = exp(−iωT). This leads to a three-fold degeneracy of marginally stable excitations with λ = 1. (b) Away from the transition (gA = 0.8, gB = 0.1), the excitations rearrange to describe the stabilization of the oscillation amplitude of the emerging twisted mode (λ f ), leaving a two-fold degeneracy of marginal excitations λ0 = λt = 1 corresponding to U(1) and time-translation invariance. In the half-step operators, these two excitations are separated at λ(cid:48) = 1 and hence structurally stabilized, which provides a signature of twisted states in terms of topological excitations. 0 = −1, λ(cid:48) t 0,TT/2 = 0 or π. Inspecting can now in principle be satisfied in two ways, corresponding to ω(cid:48) the Goldstone mode ψ0(t) = (Ψ(t), −Ψ∗(t))T we invariably find that the twisted solutions realize 0T/2 = π, while the time-translation freedom remains associated with ω(cid:48) TT/2 = 0. This ω(cid:48) separation of eigenvalues to symmetry-protected positions is again a robust spectral feature that can only change in further phase transitions. The key observation in this respect is that these phase transitions naturally link the appearance of twisted modes to the instability of zero modes, which occurs when a pair of excitations ω−,(cid:63) = − ω−,(cid:63) = 2π/T crosses the real axis (see Fig. 4), and hence corresponds to a Hopf bifurcation [73, 74] into a time-dependent state that here is still protected by symmetry. At the transition, the excitations ωn map onto the stability spectrum λn = exp(−iωnT) of the emerging twisted state, where ω−,(cid:63) and ω−,(cid:63) both map onto eigenvalues λ−,(cid:63) = λ−,(cid:63) = 1. These are degenerate with the U(1) Goldstone mode λ0 = 1. Away from the transition, these three excitations rearrange into a decaying excitation λ f < 1 describing the stabilization of the oscillation amplitude of the twisted mode, and the Goldstone modes λ0 = λT = 1 from U(1) gauge and time-translation invariance. These phase transitions therefore provide the mechanism by which the proposed class of nonlinear systems acquires robust dynamical features that do not have a counterpart in the linear case. 2.4. Detailed derivation of the symmetry-constrained excitation spectrum We here provide the technical details on the derivation of the stability excitation spectrum for the different solutions identified in the previous subsection. For stationary states, their stability can be analyzed by setting Ψ(t) = [Ψn + u exp(−iωt) + v (12) and expanding in the small quantities u and v, which we group into a vector ψ = (u, v)T . This leads to the Bogoliubov equation ∗ exp(iωt)] exp(−iΩnt) (13) where Σz = σz ⊗ 1 is a Pauli matrix in the space of u and v while the Bogoliubov Hamiltonian = (FΨ)Ψ∗. In our models with saturable is given by Huu = −H∗ = H + (FΨ)Ψ, Huv = −H∗ ωψ = (H[Ψn] − ΣzΩn)ψ vv vu -0.2 0-2 2zero-mode sideω−,★ω~−,★ω0λ=e-iωT(a)Im ωRe ωtwisted-mode sideλ−,★λ~−,★λ0awayλTλfλ0(b)half-stepλ′Tλ′fλ′0λ=(λ′)2gain or density-dependent loss [Eq. (6) and (7)], this becomes f (cid:48)Ψ2 (cid:18) H + f + f (cid:48)Ψ2 H[Ψ] = −[ f (cid:48)Ψ2]∗ −[H + f + f (cid:48)Ψ2]∗ (cid:19) , (14) where the terms containing f have to be read as diagonal matrix with entries fA(Ak2) or fB(Bk2) (depending on the sublattice), and the prime denotes the derivative with respect to the argument. As any Bogoliubov equation, Eq. (13) exhibits the charge-conjugation symmetry (H[Ψn] − ΣzΩn)∗ = −Σx(H[Ψn] − ΣzΩn)Σx, with the corresponding Pauli matrix Σx = σx ⊗ 1. This implies that the excitation spectrum is symmetric under the transformation ωm → ωm = −ω∗ m. Furthermore, due to the U(1) gauge freedom, Eq. (13) always admits the Goldstone mode ψ0 = (Ψ, −Ψ∗)T with eigenvalue pinned at ω0 = 0. As the dimension of H is even, an additional odd number of eigenvalues must lie on the imaginary axis. We need to explore the interplay of these spectral features with the analogous symmetries in Ω. The time-preserving symmetry of the nonlinear wave equation (1) induces the relation X(H[Ψn] − ΣzΩn)∗X = −(H[ Ψn] + ΣzΩn) (15) where X = 12 ⊗ X. For non-self-symmetric pairs of stationary modes Ψn, Ψn, this implies that their excitation spectra are identical. For self-symmetric stationary zero modes Ψ0 = Ψ0 with Ω0 = 0, on the other hand, Eq. (15) turns into an additional charge-conjugation symmetry which imposes a constraint on the excitation spectrum. In this case, we can classify the perturbations into joint eigenstates of X with eigenvalue σ = ±1. These take the form vσ = σXuσ, and fulfill the reduced eigenvalue equations ω±u± = [H + F + FΨΨ0 ± FΨ∗ Ψ0X] u±. (16) For our models with saturable gain or density-dependent loss, this simplifies to Eqs. (8) and (9). As described there, the excitation spectrum is thus composed of two parts, eigenvalues ω+ that display an enhanced nonlinearity and eigenvalues ω− that coincide with those of the nonlinear wave operator H + f , with Ψ = Ψ0 fixed to the zero mode. According to Eq. (12), the perturbations [e−iω−tu− − eiω−t Xu∗−] corresponding to ω− break the time-preserving symmetry of the zero mode. Setting u− = Ψ0 we recover that one of the eigenvalues ω− = Ω0 = 0 vanishes, in accordance with the U(1) symmetry of the wave equation. For time-dependent solutions Ψ(t), their stability against perturbations u(t) + v∗(t) is governed by the time-dependent Bogoliubov equation d i dt ψ(t) = H[Ψ(t)]ψ(t), (17) whose solutions ψ(t) = U(t)ψ(0) define the propagator U(t). The eigenvalues λm = exp(−iωmt) of U(t) can be represented by complex quasienergies that are constrained in similar ways as the excitations in the stationary case. The conventional charge-conjugation symmetry H∗(t) = −ΣxH Σx implies U∗ = ΣxUΣx, so that each ωm comes with a partner ωm = −ω∗ m (thus m) or is purely imaginary (whereupon λm is real). This includes the U(1) Goldstone mode λm = λ∗ ψ0(t) = (Ψ(t), −Ψ∗(t))T , as well as an additional Goldstone mode ψT(t) = (dΨ/dt, dΨ∗/dt)T which describes the time-translation freedom t0 of any solution Ψ(t + t0). For general pairs of states Ψ(t), Ψ(t), the propagators are related as U(t) = X U∗(t)X = XΣx U(t)ΣxX, which again implies that their quasienergy excitation spectra are identical. This also applies to the case of periodic solutions with Ψ(t) = Ψ(t + T), where the propagator U(T) represents a Bogoliubov-Floquet operator. Such solutions are unstable if λm > 1, hence Im ωm > 0, apart from the two Goldstone modes that are fixed at λ0 = λT = 1. For the more general case of periodic solutions with Ψ(t) = exp(iϕ)Ψ(t + T), this applies when the Floquet operator is redefined as diag(e−iϕ, eiϕ)U(T). For self-symmetric states Ψ(t) = Ψ(t), we have at all times U(t) = XU∗(t)X = XΣxU(t)ΣxX. This allows us to separate perturbations that preserve or break the symmetry, and leads to the time- dependent variants (10) and (11) of Eqs. (8) and (9). For periodic states Ψ(t) = exp(iϕ)Ψ(t + T), the self-symmetry constraints the phase to ϕ = 0, π, hence τ = exp(iϕ) = ±1, which needs to be respected in the proper definition of the Bogoliubov-Floquet operator F = τU(T) to result in two Goldstone modes pinned at λ0 = λT = 1. For the final case of a twisted periodic state with Ψ(t + T/2) = Ψ(t), we have automatically Ψ(T) = Ψ(0) without any additional phase [furthermore, the case Ψ(t + T/2) = − Ψ(t) is not separate since we can then redefine Ψ(t) → iΨ(t)]. The Floquet operator factorizes as U(T) = XU∗(T/2)XU(T/2) = XΣxU(T/2)ΣxXU(T/2). (18) In this case, the stability spectrum λm = (λ(cid:48) m)2 thus follows from the eigenvalues λ(cid:48) m of the half-step propagator U(cid:48) = XΣxU(T/2). The U(1) gauge freedom can now in principle be 0 = ±1. However, inspecting the Goldstone mode satisfied in two ways, corresponding to λ(cid:48) ψ0(t) = (Ψ(t), −Ψ∗(t))T we invariably find ψ0(T/2) = −XΣxψ0(0) so that the twisted solutions 0 = −1. The Goldstone mode ψT(t) = (dΨ/dt, dΨ∗/dt)T of indeed always realize the case λ(cid:48) the time translation, on the other hand, fulfills ψT(T/2) = XΣxψT(0), and hence always lies at λ(cid:48) = 1. Note that this spectral separation persists if we were to redefine the half-step propagator as −XΣxU(T/2) (as we are entitled to do; both conventions make sense), upon which the two symmetry-protected excitations swap their positions. T 3. Conclusion In summary, we showed that spectral symmetries underpinning topological quantum systems can be extended to nonlinear complex-wave equations, where they lead to robust constraints of the dynamics. Conceptually, this allowed us to identify a generalization of zero modes from the linear to the nonlinear setting, as well as symmetry-protected power-oscillating modes that do not have a counterpart in the linear case. These states furthermore support symmetry-protected excitations, which play a crucial role in their dynamical features. Practically, nonlinearities are of fundamental importance to stabilize systems with loss and gain in quasistationary operating regimes. In particular, the symmetry-protected modes can be induced into weakly interacting bosonic systems with saturable gain or density-dependent loss, such as recently realized in polaritonic condensates [24, 49] and topological lasers [50 -- 52]. The results can also be reinterpreted in classical wave settings, as encountered, e.g., in conventional optics and acoustics. These findings show that the notion of symmetry-protected topological states persists in nonlinear bosonic systems, and that the nonlinearities lead to an even richer phenomenology than in the original single-particle context. On the practical side, it is worthwhile to explore further extensions to include the dynamics of the medium [79], which is particularly important when one considers time-dependent solutions. More generally, it would be highly worthwhile to explore nonlinear extensions for other representatives of linear topological universality classes, for which experimental realizations and a general classification are currently unknown. Funding UK Engineering and Physical Sciences Research Council EP/P010180/1). The research data is accessible at doi 10.17635/lancaster/researchdata/235 (EPSRC) (EP/N031776/1, Disclosures The authors declare that there are no conflicts of interest related to this article.
1501.05042	1	1501	2015-01-21T03:02:37	Spatially resolved resonant tunneling on single atoms in silicon	[ "cond-mat.mes-hall" ]	The ability to control single dopants in solid-state devices has opened the way towards reliable quantum computation schemes. In this perspective it is essential to understand the impact of interfaces and electric fields, inherent to address coherent electronic manipulation, on the dopants atomic scale properties. This requires both fine energetic and spatial resolution of the energy spectrum and wave-function, respectively. Here we present an experiment fulfilling both conditions: we perform transport on single donors in silicon close to a vacuum interface using a scanning tunneling microscope (STM) in the single electron tunneling regime. The spatial degrees of freedom of the STM tip provide a versatility allowing a unique understanding of electrostatics. We obtain the absolute energy scale from the thermal broadening of the resonant peaks, allowing to deduce the charging energies of the donors. Finally we use a rate equations model to derive the current in presence of an excited state, highlighting the benefits of the highly tunable vacuum tunnel rates which should be exploited in further experiments. This work provides a general framework to investigate dopant-based systems at the atomic scale.	cond-mat.mes-hall	cond-mat	Spatially resolved resonant tunneling on single atoms in silicon B. Voisin1, J. Salfi1, J. Bocquel1, R. Rahman2 and S. Rogge1 1 Centre for Quantum Computation and Communication Technology, School of Physics, The University of New South Wales, Sydney, NSW 2052, Australia 2 Purdue University, West Lafayette, IN 47906, USA E-mail: [email protected] Abstract. The ability to control single dopants in solid-state devices has opened the way towards reliable quantum computation schemes. In this perspective it is essential to understand the impact of interfaces and electric fields, inherent to address coherent electronic manipulation, on the dopants atomic scale properties. This requires both fine energetic and spatial resolution of the energy spectrum and wave-function, respectively. Here we present an experiment fulfilling both conditions: we perform transport on single donors in silicon close to a vacuum interface using a scanning tunneling microscope (STM) in the single electron tunneling regime. The spatial degrees of freedom of the STM tip provide a versatility allowing a unique understanding of electrostatics. We obtain the absolute energy scale from the thermal broadening of the resonant peaks, allowing to deduce the charging energies of the donors. Finally we use a rate equations model to derive the current in presence of an excited state, highlighting the benefits of the highly tunable vacuum tunnel rates which should be exploited in further experiments. This work provides a general framework to investigate dopant-based systems at the atomic scale. 5 1 0 2 n a J 1 2 ] l l a h - s e m . t a m - d n o c [ 1 v 2 4 0 5 0 . 1 0 5 1 : v i X r a 2 1. Introduction Solid-state devices have now reached the single-atom level due to the constant improve- ment of nanofabrication techniques over the last few decades. It is now possible to probe, couple and manipulate single atomic orbitals using a wide range of systems and techniques [1]. Nevertheless the presence of control electrodes and interfaces, essen- tial to manipulate such quantum orbitals or spin degrees of freedom, can induce strain or quantum confinement which modify the properties of such objects (spectrum, spin states, couplings, lifetimes). Additionally these objects can be very sensitive to their atomic position in the crystal host [2, 3, 4, 5]. Silicon represents a prime candidate for future quantum computation schemes due to the extraordinarily long coherence times of donors [6, 7, 8], together with intensive work devoted to their atomically precise po- sitioning in the silicon host [9]. However the situation is here more complex because of the peculiar presence of the six degenerate conduction band minima ("valleys") [10, 11]. Valleys govern the donor spectrum and properties as they mix in the deep confinement potential of such atoms. The valley mixing, and thus the resulting spectrum, is also per- turbed by the local environment [12, 13]. For this reason, understanding the complexity around donors, interfaces, electric fields and valleys has become a crucial challenge in the view of donor-based quantum computation [14]. Field-effect transistors issued from the microelectronics industry have been natural tools to investigate single donors in silicon [15, 16, 17, 18, 19, 20, 21, 22]. The well- known Moore's law has accurately predicted the dramatic reduction of their dimensions over the last decades and the sub 10 nm range has now been reached [23]. This allows to electrically address single donors in the channel. At low temperature, in the resonant tunneling regime, it becomes possible to perform a transport spectroscopy of the donor states [18, 24]. However the extreme compactness of such systems from the large scale integration scheme, affording strong couplings to the different electrodes, leads to a very complex 3D electrostatic problem preventing a precise spatial analysis, which is necessary to evaluate the atomic-scale fabrication methodology for donors. On the other hand, a technique like Scanning Tunneling Microscopy (STM) and Spectroscopy (STS) readily offers atomic resolution. STM experiments have been conducted on subsurface single impurities in III-V semiconductors. In particular, the peculiar geometry of cross- sectional STM and the absence of surface states, has allowed to probe and manipulate shallow donors and acceptors [25, 26, 27, 28, 29] as well as deep magnetic ions [3, 30]. In such experiments, the relative position of the dopants and their depth is determined with atomic precision. However the absolute energy scale has usually remained unknown mainly because of tip-induced band bending [31, 32], even if efforts have been made to convert the bias voltage scale into an absolute energy scale [33]. With similar purpose, we show here how we can combine advantages of both transport spectroscopy and STM techniques. We achieve the resonant tunneling regime in a STM/STS setup allowing for an accurate determination of the different energies along with an atomic spatial 3 resolution [34, 35, 36], a general framework which can be applied to various systems. We perform transport through single donors located in the depletion layer of a highly n-type doped silicon substrate and determine the energy scale from the lineshape of the resonant tunneling onset at 4.2 K. Moreover, varying the tunneling barrier allows to define the electric field at the donor site. The outer tunnel rate can be tuned over more than two orders of magnitude, not possible in transport experiments. A direct application is the determination of the charging energy of such donors close to a vacuum interface. Finally we develop a rate equations model to derive the current in the presence of an excited state, taking into account relaxation [37, 38]. We demonstrate that the ability to tune the different tunnel rates, by varying tip height and sample engineering [35], and taking into account their intrinsic asymmetries, is essential to further investigate the spectrum of single dopant-based systems. 2. Sample preparation We start with a commercial n-type doped silicon (100) wafer with a resistivity of 0.004- 0.001 Ω.cm, corresponding to an As concentration of about 3 · 1019 cm−3. After cleaving and cleaning, the sample is placed in ultra-high vacuum (pressure below 5 · 10−11 mbar) and degassed at 600◦C for 10 h, to remove the last contaminants. The sample is then flashed three times at 1080◦C for 10 s. After the final flash, the sample is slowly cooled down to room temperature ensuring a good crystallisation and a 2 × 1 surface reconstruction. The flashing step has two notable effects: first it removes the native oxide and secondly it creates a depletion layer at the surface of the silicon [39]. In these conditions, the depletion layer (low As concentration) is about 10 nm-thick as shown in Fig. 1(a). Finally to passivate the surface states, the reconstructed Si surface is hydrogen terminated employing a thermal cracker at a nominal H2 pressure of 5 · 10−7 mbar for 10 mn at 330◦C [40]. After preparation the sample is transfered into a low temperature STM for electrical measurements. All the experimental data in the article were taken at 4.2 K with a base pressure below 2 · 10−11 mbar. 3. Formation and electrostatics of the barriers The depletion layer created during the flashing step induces a tunnel barrier, with a rate Γin, between the substrate and any of the remaining isolated As atoms located in this depletion layer. The vacuum barrier between a dopant and the STM tip has a tunneling rate Γout. Eventually, as depicted in Fig. 1(a), such a dopant is confined between two tunnel barriers, similarly to a dopant centred in the channel of a MOSFET. At low temperature the thermal energy kBT drops below the charging energy EC (the energy necessary to bind an extra electron to the donor, usually a few tens of meV [41]), as well as below the single level spacing (the energy difference between the ground and the first excited state, given the by the so-called valley-orbit splitting on the order of a few meV [13, 24]). In these conditions transport occurs in the single electron resonant tunneling 4 Figure 1. (a) Layout of the sample. Resonant tunneling is performed on an isolated As donor located in the depletion layer of a doped silicon wafer. Electrons tunnel from the n-type substrate to the tip. (b) Band diagram in the resonant tunneling conditions. The voltage bias (U) pulls the donor ground state in resonance, between the Fermi levels of the sample's impurity band and of the tip. The donor is located between two tunnel barriers with rates Γin and Γout. regime [42]. We can give a general expression for the tunnel current as a function of the sample bias U following [43], assuming Γ = Γin + Γout (cid:28) kBT : I(U ) = [I0(z) + ∆Γ(Uthres − U )] cosh−2 dU(cid:48) (1) (cid:90) U 0 (cid:32)αe(U(cid:48) − Uthres) (cid:33) 2kBT Uthres is called the threshold voltage and refers to the onset of the resonance. The parameter ∆Γ indicates that the triangular (because of the presence of an electric field) vacuum barrier is lowered when the bias is varied towards more negative values, and so the electric field increased. Therefore the tunnel rate is increased and we assume for simplicity a linear dependence of the current with the bias, i.e. constant ∆Γ, as the first order correction to the barrier lowering. The parameter α expresses the potential variation at the donor site as a function of the bias: it is the key to convert the sample bias into an absolute energy scale on the donor, independently of the initial tip-induced band bending at zero bias. Finally, I0(z) represents the amplitude of the current step and varies as a function of the tip height as developed below (note that I0 also depends on the plane coordinates x and y). Under these assumptions, a characteristic feature of the resonant tunneling regime is the exponential increase of the current at the onset of the resonant peak when the ground state of the donor is pulled in resonance with, and further goes below, the Fermi level defined by the conduction impurity band of the substrate (see Fig. 1(b)). We have developed in [36] a procedure to identify single donors using a two-pass scan, summarised in Fig. 2(a,b). First, in (a), a filled-states topography is performed (U =-1.25 V, I=-100 pA). Dopants appear in this image as bright protusions caused by the overlap between the evanescent tail of the donor's wavefunction in the vacuum and depletion region ~10 nmsubstrateisolatedAs atomvacuumbarrierSTMtipeUU<0HiSAsyxSTMtipzdepletion region~10 nmsubstrate(a)(b)impurity bandEFoutinoutinvalence band5 Figure 2. Donor 1. (a,b) Two-pass scan over a single donor. First a filled-states topography is performed, shown on (a) (set-point: U =-1.25 V, I=-100 pA) where the donor is barely visible due to the direct tunneling from the valence band to the tip. Secondly a current measurement (the topography, recorded on (a), is now played with a z-offset of 100 pm towards the surface. U =-1.0 V) is performed, shown on (b). The voltage is set in the band gap such that only resonant tunneling through the donor is measured, allowing to directly relate the tunnel current to the donor's wave-function. (c) I(U ) curves taken above the donor (at the red cross on (b)) at 9 different tip heights (28 pm-step increase). The exponential increase of current as the tip gets closer to the surface proves that Γin (cid:29) Γout. (d) The fit of the amplitude of the current as a function of z (left axis) allows to extract the vacuum decay length. The variation in the resonance's threshold (right axis) is related to the electric field at the donor site. −1.2−1.1−1−0.910−210−1100101102U(V)I(pA)(c)I0(z) α e-κzUthres(z)010020010−1100101102I0(pA)01002000.980.9850.99Uthres (V)∆z (pm)Δz(d)κ = 1.0 1010 m-1Ez = 5.3 MV.m-1U (V)1 nm (cid:239)(cid:24)(cid:19)(cid:3)(cid:83)(cid:80)(cid:19)(cid:24)(cid:19)(cid:3)(cid:83)(cid:80)(cid:19) (cid:239)(cid:24)(cid:19)(cid:3)(cid:83)(cid:80)(cid:19)(cid:24)(cid:19)(cid:3)(cid:83)(cid:80)1 nm (cid:19)1(a)(b)I (pA)x (110)y (110)-conduction from the valence bandthe tip's wavefunction. [44]. Its signal is weak compared to the direct tunneling from the valence band to the tip. Immediately after, a second pass records the current while the topography of the first one is played (minus a constant offset to enhance the tunnel current). The voltage of this second pass is set below the onset of the valence band where only resonant tunneling through donors contributes to the tunnel current (b). Fig. 2(c) shows nine I(U ) curves taken over this donor for tip heights varying with 28 pm-step increase, fitted using eq. (1). The z-displacement was carefully calibrated using the well-known distance between atomic planes. Here, several observations are noteworthy. First we observe a exponential increase of the current at the onset of the resonance, demonstrating thereby the validity of the assumption Γ (cid:28) kBT . Secondly we observe an exponential increase of the tunnel current I0(z) when the tip gets closer to the silicon surface (see Fig. 2(d), left axis). This evolution corresponds to the situation where Γout (cid:28) Γin, as expected because of the presence of the vacuum. Moreover this situation also corresponds to a classical STM experiment allowing to directly convert the tunnel current measured in Fig. 2(b) into the electronic probability distribution at the silicon surface, that is imaging in real space the donor's wave-function. This has been applied in [36] to directly image the valley interferences occurring in the vicinity of such a donor. We can extract the inverse of the vacuum decay length from this exponential behaviour, plotted in Fig. 2(d), left axis, giving κ=1.0·1010 m−1, which reflects the expected vacuum barrier height of about 4.0 eV for shallow donor states. Finally the evolution of Uthres as a function of z, plotted in Fig. 2(d), right axis, can be related to the electric field felt by the donor, following the relation derived in [36]: 6 (2) Ez = 1 Si dUthres dz We obtain a linear dependence of Uthres with z corresponding to an electric field of 5.3 MV.m−1 on the donor. An ionization at low electric field, i.e. close to flat-band, is expected as the energy difference between the conduction band and the Fermi level of a doped substrate equals the ionization energy of a single donor. The flat-band voltage UF B can vary from donor to donor due to differences in the tip and sample work func- tions but usually found around -0.8 V. Due to variations in the alpha parameter, from the donor's electrostatic environement, as well as in the donor's ionization energy (a few meV [36]), we usually found single donor's ionization occuring between -0.7 V and -1 V. The electrostatics around the donor can be accessed by understanding the depen- dence of the tunnel current on the vacuum barrier, and to the ability to move the tip in the three spatial directions. The absolute energy scale is deduced from the level arm parameter α obtained in resonant tunneling regime. We obtained an average α of 0.08, with no significant variation as a function of the tip height. We have assumed a simple 1D electrostatic model, since the STM ensures to measure a donors isolated from any defect or cluster. We also obtained for the parameter ∆Γ to exponentially vary between 3.4 MHz.V−1 and 2.6 · 102 MHz.V−1 as the tip is brought closer to the surface. Finally we can give a rough estimate for Γin, since it should be smaller than the 7 thermal broadening and larger than the maximum tunnel out rate (about 1 nA). This gives 6 GHz < Γin < 360 GHz, below the range where inelastic cotunneling or Kondo processes occur and would drastically modify the physics of our system. The STM current comes from the overlap between the tip orbital wave-function and the evanescent tail of the donor's wave-function in the vacuum. Thus one could expect an exponential decrease of the tunnel current when the donor gets deeper in the silicon. However this exponential dependence should only occur when the donor's depth exceeds several nanometers. For shallower donors, the quantum confinement of the vacuum interface is more important than that of the dielectric-screened Coulomb potential. In this regime the surface repels the wave-function towards the substrate, and the donor orbital probability density at the surface remains almost constant. Thus we are able to measure donors located up to 2 to 3 Bohr radii (the Bohr radius equals 2.2 nm for As in Si) below the silicon surface. The donor's depth can be independently obtained by fitting the shift of the conduction band's threshold due to the Coulomb potential of the positively charged donor at positive bias, or by comparing the current images to tight-binding simulations, with the wave-functions exhibiting a succession of type A/type B (see ref [36]) and different lateral extents when the donor's depth varies by atomic planes. Here we have ignored for simplicity the various contributions of tip orbitals [45] and the local density of states of the substrate [46, 19]. 4. Transport characteristics Having established a single electron transport regime for single-dopant measurements in scanning tunneling spectroscopy in Fig. 2, several aspects of the observed tunneling spectrum can be understood. In the following section we focus on the charging energy required to add another electron to the system, and on the impact of the asymmetric tunnel rates on the tunneling through excited states. 4.1. Charging states We now increase the bias until the point where two electrons can be bound and tunnel through the donor. According to the diagram of Fig. 1(b), two new transport channels can open when the bias U is increased (with U < 0): first electrons can directly tunnel from the valence band to the tip and secondly a potential well is created at the vac- uum interface as the conduction band is pulled below the Fermi level of the substrate. Therefore this well can be filled with an electron resulting in a so-called tip-induced state. If the valence band tunneling presents only a moderate interest, the tip-induced state is a very rich system as it can hybridize with the donor states. A detailed study goes beyond the scope of this article and we will here only give qualitative arguments to highlight this topic. 8 In Fig. 3(a), a map of the differential conductance (dI/dU , obtained numerically) vs. the bias is plotted for different points along a line crossing transversly a donor (donor 2, see white dotted line in the inset of Fig. 3(b)). The resonant tunneling peak that shifts from -1.0 V to -1.1 V (indicated by the upper black line) is found for all tip positions. Together with the observation that the tip field is attractive at this bias U < UF B, these resonances can be readily ascribed to a tip-induced state. An I(U ) curve, taken at x=0 along the vertical brown dotted line (i.e. right above the donor), is plotted in Fig. 3(b): we observe two current steps corresponding to the transition D+/D0 (1-electron process, lower black line) and D0/D− (2-electron process, upper black line). We have fitted this curve by summing two independent one-electron I(U) given by eq. 1, with thus two different thresholds, amplitudes and α parameters, getting αD0 = 0.06 and αD− = 0.07, but using a common ∆Γ for the two steps. We have computed the charging energy as the energy difference between the two peaks, assuming for simplicity that α is a linear function of the energy: we obtain EC=17 meV. This is a lower value as compared to the bulk (53 meV) but in agreement with reference [36] and with the charging energies observed in MOSFET structures [18, 21, 22], where interfaces, leads (tip, substrate) and electric fields [41] have a crucial role. We can finally note that the tip-induced state, highlighted by the upper black dotted line in Fig. 3(a), exists even away from the donor, and goes towards lower energy (or lower bias) above the donor as a result of the hybridization with the donor. The existence of this tip-induced quantum state, its coupling to the donor and finally the absence of nearby extra features such as vacancies or dangling bonds (which would be easily identified in the topography) undoubtedly rule out the possibility of any defect, dopants dimer or cluster, to account for this two-electron state. 4.2. Master equation - Probing excited states The spectrum of a single donor nearby an interface in silicon is nontrivial due to the 6- fold degenerate Si conduction band minimum, called "valleys". The symmetry-breaking perturbation of a planar interface influences the lifetime of spin and valley-orbit excited states, the energy of the latter, and the hyperfine coupling with the nucleus, to name a few effects. Here we use simple rate equations, in the single electron regime (i.e. maximum 1 electron on the donor) to understand how excited states can be probed in STM experiments [37, 38]: dρg dt dρe dt = ρg = (1 − ρg − ρe)Γg = ρe = (1 − ρg − ρe)Γe in − ρgΓg in − ρeΓe out + W ρe out − W ρe (3a) (3b) The term ρg (resp. ρe) represents the average occupation number of the ground (resp. excited) state. Note the asymmetric roles played by Γin and Γout: Γin adds an electron from the reservoir on the donor with respect to the total occupation number 9 Figure 3. Donor 2. (a) Spectroscopy I(U ) taken across a donor. We identify two current steps, highlighted by the two black dotted lines corresponding to the D+/D0 (lower one) and D0/D− (upper one) transitions. The onset of the valence band is found at large negative bias. (b) Cut taken along the vertical brown dotted line in (a). The fit gives an α parameter for each transition and allows to deduce a charging energy of 17 meV. Inset: current image of donor 2 (U =-0.9 V). The white dotted line indicates the spatial direction along which the spectroscopy shown in (a) has been performed. ρg + ρe, ensuring a maximum of one electron in the system, fulfilling the single electron regime condition. Γout empties each state at a rate proportional to its occupation number. W represents a phenomenological relaxation rate from the excited state to the ground state. We can easily derive the stationnary regime solutions ρstat , ρstat ρe = 0 and then compute the current I g+e flowing through the donor: g e by solving ρg = (4a) (4b) (4c) 1 + W Γe out (1 + Γe in/Γg ) + Γe in Γg in in) ( W +Γg Γe out out ) ρstat g = (1 + W Γe out )(1 + Γg out Γg in 1 ρstat e = 1 + W Γg g Γg I g+e = eρstat out in + (1 + Γg Γg out + eρstat e Γe out out )( W +Γe Γe in out ) These expressions are very general and can be applied to any system. Here we x (nm)U (V) −505−1.2−1−0.8−0.6−12−10−8−1.2−1−0.8−0.60200400600800U(V)I(pA)log(dI/dU)D0D+D-1 nm2 nmU=-0.9 V(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)      (cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)      1-electron process2-electron process(a)(b)10 in, Γe in (cid:29) W, Γg,e give some insights on the different situations which can occur in our STM setup. To simplify these expressions we assume Γg out , limit which indeed corresponds to our experimental situation. As previously mentioned, the vacuum tunnel barrier for electrons tunneling to the tip is much slower than the one formed by the depleted region between the reservoir and the donor, and therefore equation (4c) becomes independent of Γin. We finally derive our quantity of interest which is the variation in the tunnel current when the excited state enter the bias window. For this purpose we define a function ∆I which reads as follows: ∆I(u, v) = I g+e I g = out and v = Γe 2v(1 + u) 1 + v(1 + 2u) out/Γg (5) with u = W/Γe out two dimensionless parameters and I g the cur- rent when only the ground state is in the bias window (thus Ig = eΓg out). As mentioned above, this formalism is general: the excited state could be for instance a valley or an orbital state, with usually W (cid:29) Γout, or a spin excited state, with possibly W (cid:28) Γout. As shown in Fig. 2(c), we are able to tune the outer tunnel rate over three orders of magnitude, from around 60 MHz (corresponding to a tunnel current of 100 fA, noise floor) to around 6 GHz (1 nA, higher currents dehydrogenate the surface), such that the tunnel-out rate could cross the relaxation rate in this range. Secondly we show in Fig. 4(a) and (b) two examples of outer tunnel rates, normalised to the maximum value of the overlap betwen the tip and both of these wave-functions, for the ground and the first valley excited states of a single donor at two different depths: 1.75 a0 for (a) and 3.75 a0 for (b), a0 = 0.543 nm is the Si lattice constant. These tunnel rates are spatially resolved along a line crossing the donor wave-function at the surface (i.e. along x). They have been obtained by computing the density probability of each state at the surface, with the wave-functions being calculated by tight-binding simulations [36, 47, 48]. We observe in (a) that the ground state, plotted in black, is much brighter than the first ex- cited state, plotted in red, which is not the case for the second depth (b). Other depths (not shown) demonstrate even stronger asymmetries without a conclusive trend. Again this shows the sensitivity of the donor properties at the atomic scale due to the presence of valleys, whose coherent populations are modified by the presence of the surface, and are associated with rapid modulation of donor's electronic probability density. Current investigations, concerning single donors excited states, two-electron states or two-donor correlations (exchange for instance), shall give further insights in the role played by silicon valleys on donors' properties. To summarize this work on rate equations and the possibility it offers in our STM experiments, we have plotted in Fig. 4(c) the function ∆I as a function of u = W/Γe out varying from 0.01 to 100, for various v = Γe out also varying in the same range. We can notably expect a decrease of current (i.e. ∆I(u, v) < 1) if u (cid:28) 1 (slow relaxation) and v (cid:28) 1 (slow excited outer rate). This can be understood as a blocked situation where an electron gets trapped in the excited state with both low probabilities to relax out/Γg 11 Figure 4. (a), (b) Γout (in arbitrary units) computed for the ground state (GS, in black) and excited state (ES, in red) states of a donor located at 1.75 a0 (a) and 3.75 a0 (b) below the surface. This shows the possible differences in the brightnesses ratio Γe out. (c) ∆I calculated for various (u, v) both varying from 0.01 to 100. A decrease of current is observed when the excited state enters the bias window if u, v (cid:28) 1. out/Γg and to tunnel to the tip. More conventionally, ∆I(u, v) > 1 is obtained for u (cid:28) 1 and v (cid:29) 1 (fast excited out rate). Indeed, low energy single-hole excited states were found in recent STM experiments with acceptors in silicon [34]. For donors, the excited state energies can well exceed the thermal energy at 4.2 K. From the preceding discussion we can expect an interplay between tunnel-out rates, relaxation rates, and energies, in determining the influence of excited states on the tunneling current. The spatial dependence of the tunnel rates should be combined with sample engineering [35] to further tackle excited states in future STM experiments. 5. Conclusion In conclusion, we have described a scheme for quantitative single electron transport experiments performed with spatial resolution at low temperatures, using cryogenic scanning tunneling spectroscopy. We have demonstrated the combination of atomic spatial resolution, required to investigate the properties of single donors close to an interface, and of the determination of the absolute enegy scale, essential to obtain addition [36] and excited state energies [34]. The method makes use of a level arm parameter extracted from the analysis of the single-electron transport line-shape, which is given by the temperature of a reservoir in the substrate. Some new insights into the −50500.51d (nm)Γout (a.u.)−50500.51d (nm)Γout (a.u.)10−210−1100101102012W/ΓouteΔI(u,v) v0.011100(a)(b)(c)GSESGSES1.75 a03.75 a012 electrostatics, based on tip-height dependence measurements of single-dopant transport spectrum, have also been presented. Furthermore we have developed a simple rate equations model showing the strong dependence of the tunnel current as a function of the tunable tunnel rates and relaxation rates. This combination of the STM and of this general formalism can be readily applied to dopants in other systems, and therefore allows to investigate the properties of single and multiple dopants in solid-state devices. 6. Acknowledgments This work is supported by the European Commission Future and Emerging Technologies Proactive Project MULTI (317707) and the ARC Centre of Excellence for Quantum Computation and Communication Technology (CE110001027), and in part by the US Army Research Office (W911NF-08-1-0527). This work is part of the research program of the Foundation for Fundamental Research on Matter (FOM), which is part of the Netherlands Organization for Scientific Research (NWO). S.R. acknowledges a Future Fellowship (FT100100589). The use of nanoHUB.org computational resources operated by the Network for Computational Nanotechnology funded by the US National Science Foundation under grant EEC-0228390 is gratefully acknowledged. References [1] Paul M. Koenraad and Michael E. Flatte. Single dopants in semiconductors. Nat Mater, 10(2):91 -- 100, February 2011. [2] B. E. Kane. A silicon-based nuclear spin quantum computer. Nature, 393(6681):133 -- 137, May 1998. [3] A. M. Yakunin, A. Yu. Silov, P. M. Koenraad, J.-M. Tang, M. E. Flatte, J.-L. Primus, W. Van Roy, J. De Boeck, A. M. Monakhov, K. S. Romanov, I. E. Panaiotti, and N. S. Averkiev. Warping a single mn acceptor wavefunction by straining the gaas host. Nat Mater, 6(7):512 -- 515, July 2007. [4] Y. L. Hao, A. P. Djotyan, A. A. Avetisyan, and F. M. Peeters. Shallow donor states near a semiconductor-insulator-metal interface. Phys. Rev. B, 80(3):035329 -- , July 2009. [5] Mamadou Diarra, Christophe Delerue, Yann-Michel Niquet, and Guy Allan. Screening and polaronic effects induced by a metallic gate and a surrounding oxide on donor and acceptor impurities in silicon nanowires. J. Appl. Phys., 103(7):073703 -- 5, April 2008. [6] Jarryd J. Pla, Kuan Y. Tan, Juan P. Dehollain, Wee H. Lim, John J. L. Morton, David N. Jamieson, Andrew S. Dzurak, and Andrea Morello. A single-atom electron spin qubit in silicon. Nature, 489(7417):541 -- 545, September 2012. [7] Jarryd J. Pla, Kuan Y. Tan, Juan P. Dehollain, Wee H. Lim, John J. L. Morton, Floris A. Zwanenburg, David N. Jamieson, Andrew S. Dzurak, and Andrea Morello. High-fidelity readout and control of a nuclear spin qubit in silicon. Nature, 496(7445):334 -- 338, April 2013. [8] John J. L. Morton, Alexei M. Tyryshkin, Richard M. Brown, Shyam Shankar, Brendon W. Lovett, Arzhang Ardavan, Thomas Schenkel, Eugene E. Haller, Joel W. Ager, and S. A. Lyon. Solid- state quantum memory using the 31p nuclear spin. Nature, 455(7216):1085 -- 1088, October 2008. [9] Martin Fuechsle, Jill A. Miwa, Suddhasatta Mahapatra, Hoon Ryu, Sunhee Lee, Oliver Warschkow, Lloyd C. L. Hollenberg, Gerhard Klimeck, and Michelle Y. Simmons. A single-atom transistor. Nat Nano, 7(4):242 -- 246, April 2012. 13 [10] W. Kohn and J. M. Luttinger. Theory of donor states in silicon. Phys. Rev., 98(4):915 -- 922, May 1955. [11] A K Ramdas and S Rodriguez. Spectroscopy of the solid-state analogues of the hydrogen atom: donors and acceptors in semiconductors. Reports on Progress in Physics, 44(12):1297 -- , 1981. [12] Belita Koiller, Xuedong Hu, and S. Das Sarma. Exchange in silicon-based quantum computer architecture. Phys. Rev. Lett., 88:027903, Dec 2001. [13] R. Rahman, J. Verduijn, N. Kharche, G. P. Lansbergen, G. Klimeck, L. C. L. Hollenberg, and S. Rogge. Engineered valley-orbit splittings in quantum-confined nanostructures in silicon. Phys. Rev. B, 83(19):195323 -- , May 2011. [14] Floris A. Zwanenburg, Andrew S. Dzurak, Andrea Morello, Michelle Y. Simmons, Lloyd C. L. Hollenberg, Gerhard Klimeck, Sven Rogge, Susan N. Coppersmith, and Mark A. Eriksson. Silicon quantum electronics. Rev. Mod. Phys., 85(3):961 -- 1019, July 2013. [15] M. Hofheinz, X. Jehl, M. Sanquer, G. Molas, M. Vinet, and S. Deleonibus. Individual charge traps in silicon nanowires. The European Physical Journal B - Condensed Matter and Complex Systems, 54(3):299 -- 307, 2006. [16] H. Sellier, G. P. Lansbergen, J. Caro, S. Rogge, N. Collaert, I. Ferain, M. Jurczak, and S. Biesemans. Transport spectroscopy of a single dopant in a gated silicon nanowire. Phys. Rev. Lett., 97(20):206805 -- , November 2006. [17] M. A. H. Khalafalla, Y. Ono, K. Nishiguchi, and A. Fujiwara. Identification of single and coupled acceptors in silicon nano-field-effect transistors. Applied Physics Letters, 91(26):263513, 2007. [18] G. P. Lansbergen, R. Rahman, C. J. Wellard, I. Woo, J. Caro, N. Collaert, S. Biesemans, G. Klimeck, L. C. L. Hollenberg, and S. Rogge. Gate-induced quantum-confinement transition of a single dopant atom in a silicon finfet. Nat Phys, 4(8):656 -- 661, August 2008. [19] PierreM., WacquezR., JehlX., SanquerM., VinetM., and CuetoO. Single-donor ionization energies in a nanoscale cmos channel. Nat Nano, 5(2):133 -- 137, February 2010. [20] V. N. Golovach, X. Jehl, M. Houzet, M. Pierre, B. Roche, M. Sanquer, and L. I. Glazman. Single- dopant resonance in a single-electron transistor. Phys. Rev. B, 83:075401, Feb 2011. [21] J. Verduijn, G. C. Tettamanzi, and S. Rogge. Wave function control over a single donor atom. Nano Lett., 13(4):1476 -- 1480, March 2013. [22] B. Voisin, M. Cobian, X. Jehl, M. Vinet, Y.-M. Niquet, C. Delerue, S. de Franceschi, and M. Sanquer. Control of the ionization state of three single donor atoms in silicon. Phys. Rev. B, 89:161404, Apr 2014. [23] Isabelle Ferain, Cynthia A. Colinge, and Jean-Pierre Colinge. Multigate transistors as the future of classical metal-oxide-semiconductor field-effect transistors. Nature, 479(7373):310 -- 316, November 2011. [24] B. Roche, E. Dupont-Ferrier, B. Voisin, M. Cobian, X. Jehl, R. Wacquez, M. Vinet, Y.-M. Niquet, and M. Sanquer. Detection of a large valley-orbit splitting in silicon with two-donor spectroscopy. Phys. Rev. Lett., 108:206812, May 2012. [25] R. M. Feenstra. Tunneling spectroscopy of the (110) surface of direct-gap iii-v semiconductors. Phys. Rev. B, 50:4561 -- 4570, Aug 1994. [26] Z. F. Zheng, M. B. Salmeron, and E. R. Weber. Empty state and filled state image of znga acceptor in gaas studied by scanning tunneling microscopy. Applied Physics Letters, 64(14):1836 -- 1838, 1994. [27] R. de Kort, M. C. M. M. van der Wielen, A. J. A. van Roij, W. Kets, and H. van Kempen. Zn- and cd-induced features at the gaas(110) and inp(110) surfaces studied by low-temperature scanning tunneling microscopy. Phys. Rev. B, 63:125336, Mar 2001. [28] G. Mahieu, B. Grandidier, D. Deresmes, J. P. Nys, D. Sti´evenard, and Ph. Ebert. Direct evidence for shallow acceptor states with nonspherical symmetry in gaas. Phys. Rev. Lett., 94:026407, Jan 2005. [29] A. P. Wijnheijmer, J. K. Garleff, K. Teichmann, M. Wenderoth, S. Loth, and P. M. Koenraad. Single si dopants in gaas studied by scanning tunneling microscopy and spectroscopy. Phys. 14 Rev. B, 84:125310, Sep 2011. [30] J. Bocquel, V. R. Kortan, C. S¸ahin, R. P. Campion, B. L. Gallagher, M. E. Flatt´e, and P. M. Koenraad. Core-state manipulation of single fe impurities in gaas with a scanning tunneling microscope. Phys. Rev. B, 87:075421, Feb 2013. [31] S. Loth, M. Wenderoth, L. Winking, R. G. Ulbrich, S. Malzer, and G. H. Dohler. Probing semiconductor gap states with resonant tunneling. Phys. Rev. Lett., 96:066403, Feb 2006. [32] Gerhard Munnich, Andrea Donarini, Martin Wenderoth, and Jascha Repp. Fixing the energy scale in scanning tunneling microscopy on semiconductor surfaces. Phys. Rev. Lett., 111:216802, Nov 2013. [33] R. M. Feenstra. Electrostatic potential for a hyperbolic probe tip near a semiconductor. Journal of Vacuum Science & Technology B, 21(5):2080 -- 2088, 2003. [34] J. A. Mol, J. Salfi, J. A. Miwa, M. Y. Simmons, and S. Rogge. Interplay between quantum confinement and dielectric mismatch for ultrashallow dopants. Phys. Rev. B, 87:245417, Jun 2013. [35] J. A. Miwa, J. A. Mol, J. Salfi, S. Rogge, and M. Y. Simmons. Transport through a single donor in p-type silicon. Applied Physics Letters, 103(4):043106, 2013. [36] J. Salfi, J. A. Mol, R. Rahman, G. Klimeck, M. Y. Simmons, L. C. L. Hollenberg, and S. Rogge. Spatially resolving valley quantum interference of a donor in silicon. Nat Mater, 13(6):605 -- 610, June 2014. [37] T. H. Stoof and Yu. V. Nazarov. Time-dependent resonant tunneling via two discrete states. Phys. Rev. B, 53(3):1050 -- 1053, January 1996. [38] Edgar Bonet, Mandar M. Deshmukh, and D. C. Ralph. Solving rate equations for electron tunneling via discrete quantum states. Phys. Rev. B, 65(4):045317 -- , January 2002. [39] Jason L. Pitters, Paul G. Piva, and Robert A. Wolkow. Dopant depletion in the near surface region of thermally prepared silicon (100) in uhv. Journal of Vacuum Science & Technology B, 30(2):1806 -- , 2012. [40] John J. Boland. Structure of the h-saturated si(100) surface. Phys. Rev. Lett., 65:3325 -- 3328, Dec 1990. [41] R. Rahman, G. P. Lansbergen, J. Verduijn, G. C. Tettamanzi, S. H. Park, N. Collaert, S. Biesemans, G. Klimeck, L. C. L. Hollenberg, and S. Rogge. Electric field reduced charging energies and two-electron bound excited states of single donors in silicon. Phys. Rev. B, 84(11):115428 -- , September 2011. [42] C. W. J. Beenakker. Theory of coulomb-blockade oscillations in the conductance of a quantum dot. Phys. Rev. B, 44(4):1646 -- 1656, July 1991. [43] E. B. Foxman, P. L. McEuen, U. Meirav, Ned S. Wingreen, Yigal Meir, Paul A. Belk, N. R. Belk, M. A. Kastner, and S. J. Wind. Effects of quantum levels on transport through a coulomb island. Phys. Rev. B, 47:10020 -- 10023, Apr 1993. [44] Kitiphat Sinthiptharakoon, Steven R Schofield, Philipp Studer, Veronika Brzdov, Cyrus F Hirjibehedin, David R Bowler, and Neil J Curson. Investigating individual arsenic dopant atoms in silicon using low-temperature scanning tunnelling microscopy. Journal of Physics: Condensed Matter, 26(1):012001, 2014. [45] C. Julian Chen. Tunneling matrix elements in three-dimensional space: The derivative rule and the sum rule. Phys. Rev. B, 42:8841 -- 8857, Nov 1990. [46] M. Mottonen, K. Y. Tan, K. W. Chan, F. A. Zwanenburg, W. H. Lim, C. C. Escott, J.-M. Pirkkalainen, A. Morello, C. Yang, J. A. van Donkelaar, A. D. C. Alves, D. N. Jamieson, L. C. L. Hollenberg, and A. S. Dzurak. Probe and control of the reservoir density of states in single-electron devices. Phys. Rev. B, 81(16):161304 -- , April 2010. [47] G. Klimeck, S.S. Ahmed, Hansang Bae, N. Kharche, S. Clark, B. Haley, Sunhee Lee, M. Naumov, Hoon Ryu, F. Saied, M. Prada, M. Korkusinski, T.B. Boykin, and R. Rahman. Atomistic simulation of realistically sized nanodevices using nemo 3-d - part i: Models and benchmarks. Electron Devices, IEEE Transactions on, 54(9):2079 -- 2089, 2007. [48] Rajib Rahman, Cameron J. Wellard, Forrest R. Bradbury, Marta Prada, Jared H. Cole, Gerhard Klimeck, and Lloyd C. L. Hollenberg. High precision quantum control of single donor spins in silicon. Phys. Rev. Lett., 99:036403, Jul 2007. 15
1602.04044	1	1602	2016-02-12T13:16:21	Field-induced dissociation of excitons in two-dimensional MoS$_{2}$/hBN heterostructures	[ "cond-mat.mes-hall" ]	Atomically thin semi-conductors are characterized by strongly bound excitons which govern the optical properties of the materials below and near the band edge. Efficient conversion of photons into electrical current requires, as a first step, the dissociation of the exciton into free electrons and holes. Here we calculate the dissociation rates of excitons in monolayer MoS$_2$ as a function of an applied in-plane electric field. The dissociation rates are obtained as the inverse lifetime of the resonant states of a two-dimensional Hydrogenic Hamiltonian which describes the exciton within the Mott-Wannier model. The resonances are computed using complex scaling, and the effective masses and screened electron-hole interaction defining the Hydrogenic Hamiltonian are computed from first-principles. For field strengths above 0.1 V/nm the dissociation lifetime is shorter than 1 picosecond, which is shorter than the lifetime of other, competing, decay mechanisms. Interestingly, encapsulation of the \moly layer in just two layers of hBN, enhances the dissociation rate by around one order of magnitude due to the increased screening showing that dielectric engineering is an effective way to control exciton lifetimes in two-dimensional materials.	cond-mat.mes-hall	cond-mat	Field-induced dissociation of excitons in two-dimensional MoS2/hBN heterostructures Center for Nanostructured Graphene (CNG), Department of Physics, Technical University of Denmark and Sten Haastrup, Simone Latini, and Kristian S. Thygesen Center for Atomic-Scale Materials Design (CAMD), Department of Physics, Technical University of Denmark Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee 37235, United States Kirill Bolotin Atomically thin semi-conductors are characterized by strongly bound excitons which govern the optical properties of the materials below and near the band edge. Efficient conversion of photons into electrical current requires, as a first step, the dissociation of the exciton into free electrons and holes. Here we calculate the dissociation rates of excitons in monolayer MoS2 as a function of an applied in-plane electric field. The dissociation rates are obtained as the inverse lifetime of the resonant states of a two-dimensional Hydrogenic Hamiltonian which describes the exciton within the Mott-Wannier model. The resonances are computed using complex scaling, and the effective masses and screened electron-hole interaction defining the Hydrogenic Hamiltonian are computed from first-principles. For field strengths above 0.1 V/nm the dissociation lifetime is shorter than 1 picosecond, which is shorter than the lifetime of other, competing, decay mechanisms. Interestingly, encapsulation of the MoS2 layer in just two layers of hBN, enhances the dissociation rate by around one order of magnitude due to the increased screening showing that dielectric engineering is an effective way to control exciton lifetimes in two-dimensional materials. Two-dimensional (2D) semiconductors, such as single- and few layer transition metal dichalcogenides, are presently being intensively researched due to their ex- traordinary electronic and optical properties which in- clude strong light-matter interactions, spin-valley cou- pling, and easily tuneable electronic states[1–9]. One of the hallmarks of the 2D semiconductors is the pres- ence of strongly bound excitons with binding energies reaching up to 30% of the band gap. These large bind- ing energies are mainly a result of the reduced dielectric screening in two dimensions[10–14]. While such strongly bound excitons are highly interesting from a fundamental point of view (for example in the context of Bose-Einstein condensates[15]) they are problematic for many of the envisioned applications of 2D materials such as photode- tectors and solar cells which rely on efficient conversion of photons into electrical currents. This is because the strong attraction between the electron and the hole makes it difficult to dissociate the excitons into free carriers. Photocurrent measurements on suspended MoS2 sam- ples have found that the photocurrent produced by below-band gap photons is strongly dependent on the applied voltage indicating that the electric field plays an important role in the generation of free carriers[16]. One way of increasing the photoresponse, could be to embed the active 2D material into a van der Waals heterostructure[17–19], and thereby enhance the screen- ing of the electron-hole interaction without altering the band structure of the material. While it has been demon- strated that this strategy can indeed be used to control the exciton binding energy, the influence of environmen- tal screening on the exciton dissociation has not been previously studied. In general, rigorous calculations of exciton binding en- ergies require a many-body approach such as the Bethe- Salpeter Equation (BSE) which directly solves for the (real) poles of the interacting response function which correspond to the neutral excitation energies of the system[20, 21]. Such calculations are computationally demanding and typically only used to study excitations from the ground state, i.e. not in the presence of exter- nal fields. We mention, however, that the BSE has been used to study field induced exciton dissociation in car- bon nanotubes by fitting the BSE absorption spectrum to the Fano line shape[22]. In this work we take a different approach using that, under certain simplifying circum- stances, the calculation of the many-body excitonic state can be reformulated as an effective hydrogenic Hamil- tonian whose eigenvalues and eigenstates represent the exciton binding energies and the envelope wave function describing the relative electron-hole motion, respectively. This is the so-called Mott-Wannier model which has been instrumental in the description of excitons in inorganic bulk semiconductors. A 2D version of the Mott-Wannier model has recently been shown to yield exciton binding energies in good agreement with BSE calculations and experiments for both freestanding[10, 11, 13, 14, 23] and supported[10, 23, 24] transition metal dichalcogenide lay- ers. The dissociation rate of the exciton is then obtained by complex scaling which is a formally exact technique to compute resonance energies and life times. By employing a recently developed quantum-classical method for cal- 6 1 0 2 b e F 2 1 ] l l a h - s e m . t a m - d n o c [ 1 v 4 4 0 4 0 . 2 0 6 1 : v i X r a culating the dielectric function of general van der Waals heterostructures, we predict the effect of embedding the MoS2 in hBN on the screened electron-hole interaction and exciton dissociation rate. When an in-plane, constant electric field is applied to an exciton, it will eventually decay into a free electron and hole. This effect belongs to a class first studied by Keldysh[25] and Franz[26], who looked at how the opti- cal properties of semiconductors change in the presence of a static electric field. The application of a constant electric field changes the exciton from a bound state to a resonance with a finite lifetime equal to the inverse dis- sociation rate. The literature on resonances in quantum physics is vast, and we shall not go into the topic here but simply mention a few few important facts. First, it should be understood that even the definition of a resonance is non-trivial. The reason for this is clear from Howland's Razor which states that no satisfactory definition of a res- onance can depend only on the structure of a single op- erator on an abstract Hilbert space[27]. To illustrate the content of the statement consider the Stark effect in hy- drogen: Let H() = −∆− 1/r + x. It can be shown that H() is unitarily equivalent to H((cid:48)) for all nonzero , (cid:48). Since we expect the properties of the resonances, and in particular their lifetime, to depend on field strength, , this example shows that the resonance cannot be viewed as a property of the operator H() alone. Instead the no- tion of resonance is only meaningful when the real space geometry of the given system and relevant boundary con- ditions on the wave functions are considered. There are generally two approaches used to compute resonances. The so-called indirect methods identify resonances as the poles of the scattering amplitude analytically extended to the complex energy plane[28], while the direct meth- ods obtain the resonance states directly as eigenstates of a complex scaled non-hermitian Hamiltonian[29, 30]. In this work we shall use the latter approach. 2 with α being the polarizability of the material. An ana- lytical expression can then be obtained for the screened electron-hole interaction[12]: W (r) = 1 4α [Y0(x) − H0(x)]x=r/2πα , (3) where Y0 is a Bessel function of the second kind, and H0 is a Struve function. For later use we note that both of these functions are meromorphic, and readily admit complex arguments. The expression (3) for the screened interaction relies on a first order expansion of 2D(q) around q = 0, and thus the results obtained from the Mott-Wannier model could be questioned. However, the validity of this approximation has been demonstrated for a number of freestanding 2D semiconductors[11, 13, 14] and recently also for MoS2 embedded in a few layers of hBN[10]. As a rule of thumb, the linear screening approx- imation Eq. (2) remains valid for intra-layer excitons in van der Waals heterostructures as long as the in-plane exciton radius is large compared to the thickness of the heterostructure[10]. For thicker slabs, the linear approx- imation breaks down and the fully q-dependent 2D(q) must be used to obtain W (r). For details on how we calculate the dielectric functions of 2D layers and het- erostructures we refer to Ref. 24. In fig. 1, we show the dielectric functions, 2D(q), of an MoS2 layer in the three different configurations shown in Fig. 2(a)-(c). The linear approximations to 2D(q) are shown by full lines. We also show the dielectric function of isolated MoS2 for different values of ω. For values of ω which are below the exciton binding energy, the dielectric function is very similar to the static one, justifying our use of the latter. As expected, embedding of the MoS2 layer in hBN leads to an increase in the screening which reduces the binding energy of the exciton, see table I. To describe excitons in a 2D semiconductor we use a Mott-Wannier model of the form + W (r) F (r(cid:107)) = EbF (r), (1) (cid:20) −∇2 2D 2µex (cid:21) ex = m−1 where µex is the exciton effective mass, µ−1 e + m−1 h , W is the screened electron-hole interaction, and Eb denotes the exciton binding energy. Based on density functional theory (DFT) band structure calculations us- ing the local density approximation (LDA) we obtain an exciton effective mass for MoS2 of 0.27me. The screened electron-hole interaction is obtained as the inverse Fourier transform of [2D(q)q]−1, where 2D(q) is the static dielec- tric function of the 2D material and 1/q is the 2D Fourier transform of 1/r. For small q we can approximate epsilon as a linear function of q[11–14], so that 2D(q) = 1 + 2παq, (2) Figure 1. (a) The effective dielectric function for isolated MoS2 as a function of q for different values of ω. (b) The effective dielectric function (dashed line) for the three differ- ent MoS2-hBN heterostructure configurations. The linear ap- proximation to the dielectric function is shown by solid lines. 1357(q,ω)(a)MoS2ω=0.00eVω=0.20eVω=0.40eVω=0.61eVω=0.81eVω=1.02eV0.000.050.100.150.200.25q[A−1]1357(q)(b)MoS2MoS2-hBNhBN-MoS2-hBN3 Figure 2. (a)-(c): The three different structures considered in this work: isolated MoS2 , MoS2 on a single layer of hBN, and MoS2 sandwiched between two hBN layers. (d): Illustration of the Mott-Wannier model for monolayer MoS2 in the absence (left) and presence (right) of an in-plane constant electric field. The exciton potential is shown in blue, the exciton wave function is sketched in green, and the energy is shown in red. When an electric field is applied, the energy of the exciton shifts down and the sharp energy peak is broadened due to the coupling to the continuum of states. Table I. Calculated values for the polarisability (α) and ex- citon binding energy (Eb) for single-layer MoS2 in the three configurations shown in Fig. 2 (a)-(c). Material α (a.u.) Eb (eV) MoS2 11.1 13.0 hBN@MoS2@hBN 16.1 MoS2@hBN 0.62 0.55 0.47 Once an in-plane constant electric field is applied to the system, the bound states of the Mott-Wannier Hamilto- nian become metastable. The situation is illustrated in Figure 2(d). Within the so-called direct methods, a reso- nance is defined as an eigenstate of the Hamiltonian un- der the boundary condition that only outgoing waves ex- ist outside the scattering region. Such an eigenstate must necessarily have a complex eigenvalue, E = ε0− iγ, and a wave function that adopts the asymptotic form e±iKx for x → ±∞ (focusing on the one-dimensional case for sim- plicity) where K = k − iκ with k > 0 (an outgoing wave) and κ > 0. The latter condition implies that the wave function increases exponentially away from the scattering region. The decay rate of the resonance state, evaluated as the rate of decay of the probability for finding the par- ticle in any finite region of space, is given by γ = kκ. It can be shown that the resonance eigenvalue, E, is a pole of the analytically continued scattering matrix[31]. To compute the resonance, one could in principle solve the Schrodinger equation with the appropriate boundary conditions. In practice, however, it is more convenient to perform a "complex scaling" of the Hamiltonian, whereby the coordinate r → eiθr and ∇ → e−iθ∇, and then solve for the eigenstates of the resulting (non-hermitian) oper- ator, Hθ, with the more standard zero boundary condi- tions. For θ > tan−1(γ/k), the complex scaled resonance wave function (that is the wavefunction analytically con- tinued to the complex plane and then evaluated on the line reiθ) is an eigenstate of Hθ of eigenvalue E, but now decaying exponentially as r → ±∞. The resonances thus appear as isolated complex eigenvalues of Hθ with energy independent of θ and square integrable wave function[32]. The complex scaled wave functions of the bound states re- main exponentially decaying eigenstates of Hθ with real eigenvalues[29]. The unbound continuum states have a different behavior: If the potentials involved are local- ized, the asymptotic form of these states as r → ∞ is eikr, with k, r ∈ R. They are thus finite at infinity, but nonnormalizable. If this is to remain true after the com- plex scaling is performed, the transformation r → reiθ must be accompanied by the transformation k → ke−iθ. As the energy of a plane wave is proportional to k2, the complex scaling operation results in the energy of the continuum states rotating into the complex plane at an angle of 2θ. This is also true for the Coulomb potential, despite its long-ranged nature[32]. On figure 3 we show an example of the spectrum of the complex-scaled exciton Hamiltonian for isolated MoS2 and zero field, for different values of θ. The two classes of states, bound and unbound, can clearly be dis- tinguished; for zero field there are no resonances. In order to apply the complex scaling procedure to a single-particle Hamiltonian, a number of requirements must be fulfilled by the potential, V . First, in the original (a)(b)(c)−200−1000100200r(A)−0.8−0.6−0.4−0.20.0W(r)(eV)(d)ContinuumE=ε0−200−1000100200r(A)ContinuumE=εR+iγ~E4 Figure 3. The different behaviour of bound and continuum states under complex scaling, for the potential corresponding to isolated MoS2 . The black dashed lines start at -0.15 eV and have been rotated into the complex plane by −2θ for each of the complex scaling angles. The continuum starts at -0.15 eV because of the finite size of the simulation box. Figure 4. The dissociation rate of an exciton in the MoS2 layer as a function of in-plane field strength for the three heterostructures. The intrinsic decay rate spans between the defect-assisted fast decay of the excitons of 2-5 ps (upper limit) and the much slower radiative recombination of the excitons (lower limit). derivation of the complex scaling technique, the potential V (r) should be dilatation analytic[30]. This is the case for the Coulomb potential, and for many other potentials which have bound states, such as Yukawa potentials[33]. A constant electric field is not dilatation analytic, but it has been proven that the technique works nonetheless[34– 36]. Secondly, most potentials of interest for physical sys- tems are known only on the real axis. In order to perform the complex scaling operation, it must be possible to find the analytic continuation of the potential in the complex plane. This is the case for the potential considered here. We mention that a full first-principles implementation of the complex scaling was recently reported and applied to the problem of Stark ionization of simple atoms[37]. The 2D eigenvalue problem for the complex scaled Hamiltonian is solved on a real space grid using radial coordinates. In order to converge the exciton energies, a large simulation cell is needed - significantly larger than the exciton radius, which is around 10 A for all of the systems considered. As the screened potential has a logarithmic singularity at the origin while being vir- tually flat at the edge of the simlation cell, a nonlinear grid is used which allows us to perform simulations in a disk of radius 250 A. The Laplacian is represented by a finite-difference stencil. In order to avoid diagonalization of the full Hamiltonian, we use the iterative eigensolver ARPACK. For each of the systems shown in Fig. 2 (a)- (c) we compute the screened interaction between charges located in the MoS2 layer, and then compute the reso- nance eigenvalue of the complex scaled 2D Mott-Wannier Hamiltonian. Figure 4 shows the MoS2 exciton dissociation rate as a function of in-plane field strength for three different het- erostructures. As expected, larger fields lead to shorter lifetimes, and the rate is seen to depend roughly exponen- tially on 1/E for the considered field strenghts. Further- more, the dissociation rate can be tuned to a high degree by changing the environment of the MoS2 . When MoS2 is placed on a single layer of boron nitride, the extra screen- ing greatly increases the dissociation rate, and when the MoS2 is sandwiched between two layers of BN, the rate is even larger. This is as expected, since larger screening results in more weakly bound excitons, which should dis- sociate more readily. Adding more hBN layers on either side is expected to enhance the screening and hence the dissociation rates even further. However, as the linear screening model breaks down in this regime[10], this has not been pursued here. In a real device, the field-induced dissociation of ex- citons described here is in competition with other decay mechanisms, such as direct radiative recombination[38]. defect-assisted recombination[39] and exciton-exciton annihilation[40]. The relative importance of these effects is highly dependent on the temperature of the MoS2 , the concentration of defects and the exciton density. At very low temperatures, the direct radiative decay of zero momentum excitons dominates, with a characteris- tic lifetime of ∼200 fs[38, 41, 42]. At room temperature, most of the excitons have non-vanishing momenta, and the radiative recombination lifetime is ∼ 1 ns[38, 39]. For these systems, defect-assisted recombination there- fore becomes an important mechanism, with a character- istic lifetime of 2-5 ps[39, 43, 44]. Exciton-exciton an- nihilations become important only when the density of excitons in a sample is large; equivalently when the av- erage distance between excitons is small. At a density of 1 × 1012 cm−2, the effective lifetime from annihilation is −1.0−0.50.00.51.01.5<(E)[eV](εR)−0.6−0.5−0.4−0.3−0.2−0.10.00.1=(E)[eV](γ)Boundandcontinuumstatesθ=0.05θ=0.1θ=0.20.40.20.10.05Fieldstrength[V/nm]10111012101310141015Dissociationrate/[s−1]IntrinsicdecayrateDissociationratesMoS2MoS2-BNBN-MoS2-BN11 ps[40]. Our calculations indicate that for field strengths larger than 0.1 V/nm, the dissociation lifetime is shorter than 1 picosecond in all the systems considered. This is shorter than the smallest characteristic lifetimes of the alterna- tive decay channels at room temperature (indicated by the gray shaded region in Fig. 4) and thus field-induced dissociation should dominate. We note that a potential gradient of 0.1 V/nm is not unlikely to exist in the metal- MoS2 contact region where charge transfer and interface dipole formation driven by Fermi level mismatch can lead to significant variations in the potential and band energies even in the absence of an applied bias voltage. Recently, a chemical treatment has successfully been used to elimi- nate the influence of defects on exciton decay[45], leaving radiative decay as the main decay channel and leading to exciton lifetimes of 10 ns. In this case, field-induced dissociation should be the main photocurrent generation mechanism. In summary we have used complex scaling to com- pute the lifetime of excitons in two-dimensional MoS2 and MoS2/hBN structures under an applied static elec- tric field. The exciton was simulated using a 2D Mott- Wannier model which has previously been found to yield a reliable description of the lowest lying excitonic states in transition metal dichalcogenides. We found that for field strengths above 0.1 V/nm, the exciton dissociation rate is larger than the intrinsic exciton decay rate in MoS2 at room temperature. Moreover, encapsulation in a few layers of hBN increases the dissociation rate by an order of magnitude for fixed field strength due to the increased screening provided by the electrons in the hBN. The Center for Nanostructured Graphene (CNG) is sponsored by the Danish National Research Foundation, Project DNRF58. 5 [9] K. F. Mak, K. He, C. Lee, G. H. Lee, J. Hone, T. F. Heinz, and J. Shan, Nature Materials 12, 207 (2013). [10] S. Latini, T. Olsen, and K. S. Thygesen, Physical Re- view B, Physical Review B Condensed Matter and Mate- rials Physics, Physical Review B, Phys Rev B 92, 245123 (2015). [11] T. C. Berkelbach, M. S. Hybertsen, and D. R. Reichman, Phys. Rev. B 88, 045318 (2013). [12] P. Cudazzo, C. Attaccalite, I. V. Tokatly, and A. Rubio, Phys. Rev. Lett. 104, 226804 (2010). [13] P. Cudazzo, I. V. Tokatly, and A. Rubio, Phys. Rev. B 84, 085406 (2011). [14] O. Pulci, P. Gori, M. Marsili, V. Garbuio, R. Del Sole, and F. Bechstedt, EPL (Europhysics Letters) 98, 37004 (2012). [15] M. M. Fogler, L. V. Butov, and K. S. Novoselov, Nature Communications 5, 4555 (2014). [16] A. R. Klots, A. K. M. Newaz, B. Wang, D. Prasai, H. Krzyzanowska, J. Lin, D. Caudel, N. J. Ghimire, J. Yan, B. L. Ivanov, K. A. Velizhanin, A. Burger, D. G. Mandrus, N. H. Tolk, S. T. Pantelides, and K. I. Bolotin, Scientific Reports 4 (2014), 10.1038/srep06608. [17] H. Terrones, F. L´opez-Ur´ıas, and M. Terrones, Scientific reports 3 (2013). [18] L. Britnell, R. Ribeiro, A. Eckmann, R. Jalil, B. Belle, A. Mishchenko, Y.-J. Kim, R. Gorbachev, T. Georgiou, S. Morozov, et al., Science 340, 1311 (2013). [19] A. Geim and I. Grigorieva, Nature 499, 419 (2013). [20] G. Strinati, Phys. Rev. B 29, 5718 (1984). [21] G. Onida, L. Reining, and A. Rubio, Rev. Mod. Phys. 74, 601 (2002). [22] V. Perebeinos and P. Avouris, Nano Letters 7, 609 (2007). [23] T. Olsen, S. Latini, F. Rasmussen, and K. S. Thygesen, Physical Review Letters, Phys. Rev. Lett, Phys Rev Lett, Phys Rev L, Prl 116, 5 (2016). [24] K. Andersen, S. Latini, and K. S. Thygesen, Nano letters 15, 4616 (2015). [25] L. V. Keldysh, J. Exptl. Theoret. Phys. 33, 994–1003 (1957). [26] W. Franz, Z. Naturforschung 13a, 484–489 (1958). [27] B. Simon, International Journal of Quantum Chemistry 14, 529 (1978). [28] J. Taylor, Scattering Theory : The Quantum Theory on Nonrelativistic Collisions (John Wiley, 1972) p. 477 s. [29] Balslev and Combes, Communications in Mathematical [1] H.-P. Komsa and A. V. Krasheninnikov, Phys. Rev. B Physics 22, 280 (1971). 86, 241201 (2012). [2] Q. H. Wang, K. Kalantar-Zadeh, A. Kis, J. N. Coleman, and M. S. Strano, Nature nanotechnology 7, 699 (2012). [3] K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, Phys. Rev. Lett. 105, 136805 (2010). [4] M. M. Ugeda, A. J. Bradley, S.-F. Shi, H. Felipe, Y. Zhang, D. Y. Qiu, W. Ruan, S.-K. Mo, Z. Hussain, Z.-X. Shen, et al., Nature materials (2014). [5] K. He, N. Kumar, L. Zhao, Z. Wang, K. F. Mak, H. Zhao, [30] J. Aguilar and J. M. Combes, Communications in Math- ematical Physics (1971). [31] N. Hatano, K. Sasada, H. Nakamura, and T. Petrosky, Progress of Theoretical Physics 119, 187 (2008). [32] W. P. Reinhardt, Annual Review of Physical Chemistry 33, 223 (1982). [33] B. Simon, Annals of Mathematics 97, 247 (1973). [34] I. W. Herbst and B. Simon, Physical Review Letters 41, 1759 (1978). and J. Shan, Phys. Rev. Lett. 113, 026803 (2014). [35] W. P. Reinhardt, International Journal of Quantum [6] D. Jariwala, V. K. Sangwan, L. J. Lauhon, T. J. Marks, Chemistry 10, 359 (2009). and M. C. Hersam, ACS nano 8, 1102 (2014). [36] I. W. Herbst, Communications in Mathematical Physics [7] M. Bernardi, M. Palummo, and J. C. Grossman, Nano 64, 279 (1979). letters 13, 3664 (2013). [8] J. S. Ross, P. Klement, A. M. Jones, N. J. Ghimire, J. Yan, D. Mandrus, T. Taniguchi, K. Watanabe, K. Ki- tamura, W. Yao, et al., Nature nanotechnology 9, 268 (2014). [37] A. H. Larsen, U. De Giovannini, D. L. Whitenack, A. Wasserman, and A. Rubio, Journal of Physical Chem- istry Letters 4, 202734 (2013). [38] M. Palummo, M. Bernardi, and J. C. Grossman, NanoLetters (2015). [39] H. Shi, R. Yan, S. Bertolazzi, J. Brivio, B. Gao, and L. Huang, 23273148, A. Kis, D. Jena, H. G. Xing, ACS Nano http://dx.doi.org/10.1021/nn303973r. pMID: 7, 1072 (2013), [40] D. Sun, Y. Rao, G. A. Reider, G. Chen, Y. You, L. Br´ezin, A. R. Harutyunyan, and T. F. Heinz, Nano Letters 14, 5625 (2014), pMID: 25171389, http://dx.doi.org/10.1021/nl5021975. [41] H. Wang, C. Zhang, W. Chan, C. Manolatou, S. Tiwari, and F. Rana, Physical Review B 93, 045407 (2016). [42] C. Poellmann, P. Steinleitner, U. Leierseder, P. Nagler, G. Plechinger, M. Porer, R. Bratschitsch, C. Schueller, 6 T. Korn, and R. Huber, Nature Materials, Nat. Mater, Nat Mater 14, 889 (2015). [43] T. Korn, S. Heydrich, M. Hirmer, J. Schmutzler, and C. Schuller, Applied Physics Letters 99, 102109 (2011). [44] D. Lagarde, L. Bouet, X. Marie, C. R. Zhu, B. L. Liu, T. Amand, P. H. Tan, and B. Urbaszek, Physical Review Letters, Phys Rev Lett, Phys Rev L, Prl, Phys. Rev. Lett 112, 047401 (2014). [45] M. Amani, D.-H. Lien, D. Kiriya, J. Xiao, A. Azcatl, J. Noh, S. R. Madhvapathy, R. Addou, K. C. Santosh, M. Dubey, K. Cho, R. M. Wallace, S.-C. Lee, J.-H. He, J. W. Ager, X. Zhang, E. Yablonovitch, and A. Javey, Science 350, 1065 (2015).
1310.4012	1	1310	2013-10-15T11:29:15	Control of Radiation Damage in MoS2 by Graphene Encapsulation	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	Recent dramatic progress in studying various two-dimensional (2D) atomic crystals and their heterostructures calls for better and more detailed understanding of their crystallography, reconstruction, stacking order, etc. For this, direct imaging and identification of each and every atom is essential. Transmission Electron Microscopy (TEM) and Scanning Transmission Electron Microscopy (STEM) are ideal, and perhaps the only tools for such studies. However, the electron beam can in some cases induce dramatic structure changes and radiation damage becomes an obstacle in obtaining the desired information in imaging and chemical analysis in the (S)TEM. This is the case of 2D materials such as molybdenum disulfide MoS2, but also of many biological specimens, molecules and proteins. Thus, minimizing damage to the specimen is essential for optimum microscopic analysis. In this letter we demonstrate, on the example of MoS2, that encapsulation of such crystals between two layers of graphene allows for a dramatic improvement in stability of the studied 2D crystal, and permits careful control over the defect nature and formation in it. We present STEM data collected from single layer MoS2 samples prepared for observation in the microscope through three distinct procedures. The fabricated single layer MoS2 samples were either left bare (pristine), placed atop a single-layer of graphene or finally encapsulated between single graphene layers. Their behaviour under the electron beam is carefully compared and we show that the MoS2 sample 'sandwiched' between the graphene layers has the highest durability and lowest defect formation rate compared to the other two samples, for very similar experimental conditions.	cond-mat.mes-hall	cond-mat	Control of Radiation Damage in MoS2 by Graphene Encapsulation Recep Zan1, 2,, Quentin M. Ramasse3,, Rashid Jalil1, Thanasis Georgiou1, Ursel Bangert2 and Konstantin S. Novoselov1,* 1School of Physics and Astronomy, University of Manchester, Manchester, M13 9PL, U.K. 2School of Materials, University of Manchester, Manchester, M13 9PL, U.K. 3SuperSTEM Laboratory, STFC Daresbury Campus, Daresbury WA4 4AD, U.K. *To whom correspondence should be addressed: [email protected], [email protected] and [email protected] Abstract Recent dramatic progress in studying various two-dimensional (2D) atomic crystals and their heterostructures calls for better and more detailed understanding of their crystallography, reconstruction, stacking order, etc. For this, direct imaging and identification of each and every atom is essential. Transmission Electron Microscopy (TEM) and Scanning Transmission Electron Microscopy (STEM) are ideal, and perhaps the only tools for such studies. However, the electron beam can in some cases induce dramatic structure changes and radiation damage becomes an obstacle in obtaining the desired information in imaging and chemical analysis in the (S)TEM. This is the case of 2D materials such as molybdenum disulfide MoS2, but also of many biological specimens, molecules and proteins. Thus, minimizing damage to the specimen is essential for optimum microscopic analysis. In this letter we demonstrate, on the example of MoS2, that encapsulation of such crystals between two layers of graphene allows for a dramatic improvement in stability of the studied 2D crystal, and permits careful control over the defect nature and formation in it. We present STEM data collected from single layer MoS2 samples prepared for observation in the 1 microscope through three distinct procedures. The fabricated single layer MoS2 samples were either left bare (pristine), placed atop a single-layer of graphene or finally encapsulated between single graphene layers. Their behaviour under the electron beam is carefully compared and we show that the MoS2 sample 'sandwiched' between the graphene layers has the highest durability and lowest defect formation rate compared to the other two samples, for very similar experimental conditions. Keywords MoS2, graphene, heterostructure, encapsulation, TEM, STEM Similar to graphene, other 2D nanomaterials such as transition metal dichalcogenides (TMD) have attracted considerable attention due to their unique physical, chemical, and structural properties as well as their great potential for applications.1-3 MoS2 is one of the better known and most studied of this new class of layered TMD materials and is already widely used in industrial processes as a nano-catalyst and dry lubricant.4, 5 Owing to the unique properties of single layer MoS2, and in particular the fact that it possesses a direct band gap,6- 8 much effort has been put into exploring its potential use in combination with graphene in electronic and optoelectronic devices with a view to build logic transistors, field emitters and detectors.9 Single layer MoS2 consists of two distinct sub-lattices: Mo (metal) atoms are trigonal- prismatically bonded to pairs of S atoms and arranged hexagonally in plane. The resulting S- Mo-S slabs are then stacked to various degrees, 2H stacking being the most common. Similar to graphite and h-BN, chemical bonds in-plane (within the slab) are covalent (strong) while weak van der Waals bonds are established between stacked layers to form a bulk (3D) crystal. As for graphene, this enables single MoS2-layer mechanical exfoliation,1 but there are several other fabrication methods such as liquid exfoliation 10 and chemical vapour deposition.11, 12 In conductive materials such as graphene, 'knock-on' (the displacement of atoms from their original positions in the lattice) is the dominant radiation damage mechanism during observation in an electron microscope, whereas ionization damage (radiolysis) prevails in semiconductors and insulators.13 If knock-on is the dominant damage mechanism, reducing the primary beam energy below the knock-on threshold will prevent it.14, 15 If radiolysis is the main damage process, there is no sharp energy threshold below which no damage occurs although cooling the specimen can help to reduce it. Radiation damage in single layer MoS2 2 has so far not been investigated to a great extent despite the fact that being able to control the defect formation is crucial: the presence of a controlled number of defects in 2D materials can lead to physical properties entirely different from their pristine form.16 Furthermore, if optimal beam-damage prevention is achieved, the time to acquire an image can be increased until a sufficient signal-to-noise (S/N) ratio is obtained.13 Ab initio studies of the displacement threshold of S atoms in pristine MoS2 layer report a knock-on energy of ~6.5 eV.17, 18 Interestingly, this corresponds approximately to the maximum knock-on energy transfer from 80 keV electrons to 32S atoms while 60 keV electrons can transfer up to 4.3 eV to 32S (and 7.4 eV at 100 keV). In TEM observations of single layer MoS2 both knock-on and ionization damage mechanisms are thus likely at play although at the low primary beam energies often used to image 2D materials the ionization is probably dominant.19, 20 Atom by atom investigation is nonetheless difficult without any protective measure. A commonly used method to preserve the crystalline form of a specimen over extended electron exposure times, and hence to avoid damage be-it caused by beam knock-on or through radiolysis, is to encapsulate the sample within a protective (conductive) layer.21 This is typically achieved for TEM observation by evaporating thin layers of amorphous carbon. This technique has obvious drawbacks, chief of which is the reduced contrast from the actual specimen because of its carbon coating. For 2D materials, one to a few atoms thick, such an approach would be almost impossible. Instead, we demonstrate here the use of graphene to encapsulate the MoS2 crystal. Graphene, which has also been suggested as a support for nano- and bio-particle imaging,22, 23 and used as membrane for liquid cells for in-situ TEM studies,24-26 generates extremely low background signals compared to amorphous carbon films, resulting in a dramatic increase in the S/N ratio in (S)TEM micrographs. Such an approach not only protects the MoS2 from environmental effects such as chemical etching, but also, thanks to graphene’s excellent electrical and thermal conductivity, increases the stability of the MoS2 layer under the electron beam through minimisation of charging effects and vibrations. Furthermore, its crystalline structure can be easily subtracted from the (S)TEM micrographs by Fourier filtering, thus enabling the visualisation of the particles as if surrounded by vacuum.27, 28 Results and Discussion 3 Our initial investigations, carried out on the bare (pristine) single layer MoS2 sample, revealed that even with a 60 keV beam and near-UHV conditions, defect formation to a free-standing sample was unavoidable. Ab-initio modelling predicts this beam energy is much lower than the calculated displacement threshold for S atoms in single layer MoS2,17, 18 and severe knock- on damage is thus unlikely in our experimental conditions. We therefore suggest that ionization damage, which has severe effects on semiconductors and insulators in comparison to knock-on damage, is responsible for the defect creation. Figure 1 illustrates how imaging at high magnification immediately initiated damage: the image in fig. 1b was acquired approximately 30 s after that in fig. 1a, demonstrating the ease with which damage is introduced. The dwell time per pixel was 39 μs and the pixel size 0.098 Ǻ corresponding to an electron dose of 2.5x106 e/Ǻ2. Following the vacancy formation in fig. 1b, a movie was recorded by drastically reducing the scanning time per frame (5.1s dwell time per pixel, 256x256 pixels only) to observe dynamically any further damage progression (see supplementary information movie_1). From the movie, it can be seen that as soon as the MoS2 sheet is perforated by losing first a single S atom and then the other Mo-bonded S atom,17 this hole enlarges easily under the scanning probe. Not only are atoms at the edge of the hole less coordinated, thus facilitating ionization damage, it is also likely that the knock- on threshold of edge atoms is significantly lowered compared to the bulk allowing this mechanism to also come into play. Due to the higher displacement threshold of Mo compared to S, and to the propensity of Mo atoms to form metallic clusters, Mo atoms appear to aggregate on the edges of the damaged area while S atoms are simply sputtered away (this suggestion is supported by EELS measurements, not shown here). These heavy, bright aggregates are clearly seen in fig. 1d and g.18 The first and last frames were extracted from movie_1 and are shown in figs 1c and d, respectively. The images, taken 65 s apart, show how the initial vacancy expands in a short time to a 2nm hole even under a relatively low (in materials science terms) electron dose-rate (1x105 e/Ǻ2/frame). High dose exposures are often required in order to obtain good enough S/N ratios in chemical (spectral) analyses. It is therefore instructive to scrutinise the HAADF images before and after an attempt at acquiring an electron energy loss spectrum image (SI) as they are shown in fig. 1e and 1g, respectively. The map was collected on the blue-framed rectangular area in the fig. 1e (survey image), and the HAADF image acquired simultaneously with the SI is presented in fig. 1f. The Gatan Enfina spectrometer was configured to acquire one spectrum every 50ms, corresponding to a total dose for the map of 2.6x1010 e/Ǻ2. Note that even though modern energy loss spectrometers allow for much shorter pixel dwell times (and therefore lower dose-rates), 4 resulting maps often suffer from poor signal-to-noise (S/N) ratio and for 2D materials in particular 50ms/spectrum was found to be the minimum for interpretable signal. Even though no damage is visible in the region of the SI on the survey image, taken immediately before the start of the EELS acquisition, the 'after' image shows that a large hole has been created. The HAADF image taken simultaneously with the EELS map reveals that the hole opened up very early into the SI acquisition and the damage became very severe, making this dataset unexploitable for chemical analysis (the resulting chemical maps are shown in Supplementary material). (figure 1 here) By contrast, graphene is known to be remarkably stable under the electron beam in very similar experimental conditions. Thanks to the combination of low primary beam energy (60 keV) and clean vacuum at the sample (below 5x10-9 Torr), clean patches of pristine single-layer graphene can thus be observed repeatedly at very high electron doses and without any observable damage formation, even for extremely long periods of time.14, 29, 30 A possible way to minimize the damage in MoS2 (or in other beam sensitive materials), whilst perhaps also increasing the stability of the flake against vibrations under the beam, is therefore to use graphene as support.23 To verify the presence of both single layer graphene and MoS2 in the samples fabricated, a diffraction pattern was obtained on this double layer stack (graphene/MoS2); it is presented in fig. 2a and reveals two clear sets of diffraction spots. The MoS2 diffraction spots are more intense than those of graphene, and they form the very inner circle (red dashed on fig. 2a) of the diffraction pattern due to the larger lattice constant of MoS2 (d100= 2.7 Ǻ) compared to graphene. The HAADF images presented in Figs. 2b and 2c were taken in this region with a dose of 5.1x106 e/Å2, slightly higher than the dose applied to the bare MoS2 sample (and in otherwise identical conditions): remarkably, no damage is observed. MoS2 placed on graphene thus appears more stable under the scanning electron beam and provides longer working time without defect formation compared to bare MoS2. Damage occurred only after longer electron beam exposure during the acquisition of a movie (see movie_2 in supplementary information, acquired in similar conditions to those employed for movie_1): figures 2d and 2e are the first and last frames of movie_2, taken 243s apart. The movie was taken with ~3.8x105 e/Å2 dose per frame (again slightly higher than the one used in movie_1), but at identical frame rate. The hole formation took longer than in the pristine sample even at higher applied dose (the first sign of damage was seen at frame 56, 5 corresponding to a total accumulated dose of 2.3x108 e/Å2), but once the damage started in the MoS2, the expansion of the defective region was again quite rapid. Note the presence in fig. 2e of bright Mo atoms in the middle of the 'hole' confirming that the graphene support is still undamaged underneath the MoS2 sheet (a copy of this micrograph is provided in supplementary material, with the contrast and brightness levels adjusted to reveal the graphene lattice more clearly). As with the bare MoS2 experiment, we attempted to collect an EEL SI of the undamaged region highlighted by the blue rectangular box in the survey image in fig. 2f. The SI parameters were almost identical to the bare MoS2 case (with slightly higher pixel time, pixel density and slightly wider SI spatial extent): the dose rate during the acquisition was therefore 5.5x108 e/Å2.s, for a total accumulated dose of 9.2x1010 e/Å2. The 'after' image taken after collecting the spectra is presented in fig. 2h and shows again clear damage to the MoS2 sheet, although perhaps not quite as severe as in the bare membrane case. The simultaneously-acquired HAADF image is shown in fig. 2g, revealing that the sheet was punctured again early through the SI acquisition, although at a higher total dose (after a larger number of pixels). While a few unit cells were mapped before the onset of damage, the resulting chemical maps (see supplementary material) would again not be exploitable for quantitative analysis. A single layer graphene support therefore enables longer imaging times for MoS2, but it is not capable of preventing damage at the higher electron doses employed for chemical mapping. The same set of experiments was carried out on the same sample flipped over in the holder, so the beam would hit the graphene first and then the MoS2, leading to identical results and conclusions. (figure 2 here) As a last step, single layer MoS2 was encapsulated between single layer graphene sheets. The prepared sample stack was initially checked via electron diffraction to make sure that MoS2 and graphene are present together as a stack (graphene/MoS2/graphene) and all as single layers. As can be seen in the diffraction pattern in figure 3a, the stack consists of single layer MoS2 (red dashed circle) 31 and two rotated graphene layers (blue dashed circle),32 whose diffraction spots can be distinguished as they are separated by almost 23° and far less intense than the MoS2 spots as mentioned in fig. 2. This time, we did not observe any damage to the MoS2 structure during imaging even for the acquisition of images with longer exposure times and higher pixel densities (therefore at higher dose and dose rates) than in the previous two cases. Figure 3b and 3c are the first and the last frames from movie_3 (see supplementary 6 information): they were taken 68 s apart after a total accumulated dose of 4.5x108 e/Å2 and clearly no hole or defect is observed, supporting the claim that MoS2 is undamaged under these conditions. However, an image taken immediately after recording the movie (exposed to a dose of ~5.2x106 e/Å2) showed some interesting contrast features as seen in fig. 3d. The area encircled by a blue dashed line in fig. 3d appears slightly darker than the surrounding area: we attribute this variation to a creation of a hole in one of the protective graphene sheets. Due to the double transfer process to fabricate this sample, the graphene membranes were covered with a slightly larger concentration of contaminants than the previous two samples, Si atoms and SiO2 clusters in particular. These contaminants are known to participate as catalysts in a localized graphene etching mechanism in the experimental conditions employed here, resulting in this case in the formation of a hole in the graphene layer.29,33 The encapsulated MoS2 appears mostly unaffected, though. Moreover, the contribution of the graphene to the atomic contrast in the HAADF images is minimal, as could be expected from simple considerations based on the large difference in atomic weights of the elements involved (2xZS=32, ZMo=42 compared to ZC=6) and the approximate Z2 dependence of HAADF images. Some minor loss of sharpness and contrast is noticeable nonetheless. This effect can be easily quantified by drawing an intensity line profile across a Mo-2S 'dumbbell' and comparing the signal-to-background ratios for the Mo and 2S columns, defined as the ratio between the column peak intensity and the intensity in the middle of the hexagons (i.e. in a hole), having subtracted any dark current offset. Using figure 1a as a representative example of the contrast obtained in bare MoS2 we determined signal-to-background ratios of 2.9 and 1.8 for Mo and 2S, respectively. When the MoS2 sheet is encapsulated with graphene, such as on figure 3d, these signal-to-background values drop to 2.5 and 1.5 for Mo and 2S, respectively, in otherwise almost identical imaging conditions. This corresponds to a loss of contrast of approximately 15% between the bare and encapsulated cases. It is interesting to note that the graphene lattice is not clearly visible in encapsulated images (or indeed where only one layer of graphene is used as support). This is even clearer in images of an area where the MoS2 sheet terminates, leaving only the graphene support visible (for more details, see Supplementary Material, figure S3). Both these experimental observation are borne out by multislice images simulations of single-layer MoS2 and of a model of the encapsulated structure: see Supplementary Material. To make sure that graphene encapsulation of MoS2 supports our basic idea of defect-free chemical mapping, an EEL SI was acquired in the blue rectangular box in the survey image in fig. 3e. As shown in fig. 3g, which was taken after the mapping, no damage was created in the mapped region, which is clearly apparent from the 7 very sharp simultaneously acquired HAADF image of fig. 3f: both Mo and S sub-lattices are resolved throughout the SI acquisition. The apparent stability of the MoS2 sheet even allowed to increase both the total dose and dose rate for the EELS acquisition by choosing more demanding parameters: 0.08 s dwell time per spectrum and higher pixel density. This resulted in a total electron dose of 1.7x1011 e/Ǻ2, which is substantially higher than the one used for the SIs in figs. 1 and 2. Thus, the encapsulation of the single layer MoS2 enabled us to investigate the sample without defect formation and, being able to employ high electron doses, we managed to obtain images and SIs with adequately high S/N ratios, to obtain accurate chemical information at the atomic scale (see supplementary information for the resulting Mo and S chemical maps). The underlying mechanisms responsible for such effective protection against beam damage are difficult to determine with certainty. The remarkable conduction properties of graphene (both thermal and electric) are certainly expected to contribute to a very effective dissipation of accumulated charge or heat under the beam. Graphene is therefore the ideal conductive 'coating' to help mitigate ionization, as suggested for instance by Egerton et al.21 A full encapsulation will result in further advantages: the impermeability of graphene provides very effective protection against environmental effects such as chemical etching under the beam (due for instance to residual gases in the microscope column - unlikely in our case thanks to the near UHV conditions at the sample).34 Should S or Mo atoms be displaced by knock-on (despite the low probability of such an event given the low beam energy), or ionised, the top and bottom carbon layers will also provide added stability to the structure and possibly prevent the displaced or ionised atoms from complete ejection. Finally, we also note that the close proximity between the graphene and MoS2 layers may lead to the formation of interlayer bonds (and consequently a modification of the electronic structure of the encapsulated material). Although this suggestion is only speculative, such bond formation would be expected to favour electron transport between the graphene and MoS2 layers and thus contribute to mitigating ionization. (figure 3 here) Conclusions In summary, employing a 60 keV electron beam and near-UHV conditions (<5x10-9 torr at the sample) to reduce knock-on damage and to minimise ionization damage did not prevent any occurrence of severe damage to bare, free-standing MoS2. Radiation damage was somewhat 8 mitigated in images obtained after placing the MoS2 layer on a single layer of graphene. However, the damage reduction was not sufficient at the high electron doses typically required for quantitative chemical mapping. Our results demonstrate that damage of single layer MoS2 (whether arising from knock-on or ionization effects in the electron microscope), can be prevented altogether by encapsulation between graphene layers. The three-layer stack (graphene/MoS2/graphene) allows the application of high electron doses for high resolution, defect free imaging and, importantly, for chemical analysis of MoS2. We envisage this technique could also be employed for detailed studies of other beam sensitive materials, e.g., molecules and nano- and bio- particles. Methods Sample Preparation Both MoS2 and graphene single layer flakes were prepared by mechanical exfoliation. For the first sample, following exfoliation, the single layer MoS2 was directly transferred to a standard QuantifoilTM TEM grid via polymer based wet-transfer technique as samples need to be freely suspended for transmission electron microscopy measurements.35 The second sample was fabricated by transferring by mask aligning techniques a single MoS2 layer onto a single layer of graphene, which had been positioned (exfoliated) on Si/SiO2 wafer. The two-layer stack was then transferred to a TEM grid in the same way as the single layer MoS2. The third sample required a fabrication process consisting of two transfers: firstly, MoS2 was transferred onto single layer graphene and secondly a further single layer graphene sheet was placed on top of the MoS2 layer. This was followed by the removal of the Si/SiO2 wafer substrate, thus releasing completely the three-layer stack for wet transfer onto a TEM grid. Each transfer during the sample fabrication was followed by a dip into acetone to remove protective polymer layers (PMMA). Once transferred to the TEM grid, the samples were dipped one final time in acetone and dried in a critical point dryer to avoid the surface tension damage to the flakes. Characterization Microscopy measurements were performed at the SuperSTEM Laboratory, on a Nion UltraSTEM100TM aberration-corrected dedicated scanning transmission electron microscope. The design of the column allows for clean high vacuum conditions at the sample (<5x10-9 torr), reducing the probability of damage through chemical etching and preventing build-up of contamination which hinders high resolution observations. The Nion UltraSTEM has a cold field emission gun with a native energy spread of 0.35 eV and was operated at 60 keV primary beam energy. The beam was set up to a convergence semi-angle of 30 mrad with an estimated beam current of ~100 pA. Note that in a cold field emission instrument the probe current drops slightly with time until the tip is cleaned ('flashed'): all electron doses estimated here assume a freshly flashed tip and a current of 100 pA (the tip was systematically flashed shortly before all the acquisition of the data presented). Under these operating conditions, the estimated probe size is ~1.1 Å, providing the perfect tool for atom-by-atom chemical analysis;14 these conditions are particularly adequate for MoS2 as the distance between Mo 9 and S atoms is 1.8 Å when the single layer slab is viewed along an (001) direction. These experimental conditions (scanning probe, low primary beam energy, high vacuum conditions) are significantly different from those in most other studies of MoS2, which are typically performed with stationary and slightly higher energetic beams under poorer vacuum conditions. High Angle Annular Dark Field (HAADF) imaging was employed to produce atomically resolved images whose intensity is approximately proportional to the square of the average atomic number Z of the material under investigation. This chemically-sensitive ‘Z- contrast’ mode is ideally suited to directly identify the nature of individual atoms.14 HAADF imaging is complemented by further chemical fingerprinting through Electron Energy Loss Spectroscopy (EELS). Conflict of Interest The authors declare no competing financial interest. Supporting Information Available This additional material consists of Movies (.avi) showing the dynamic behaviour of the samples under the electron beam. Further images, electron energy loss chemical maps and multislice simulations are also provided in a separate .pdf document. This material is available free of charge via the Internet at http://pubs.acs.org. Acknowledgements This work was supported by the EPSRC (UK). The SuperSTEM Laboratory is the EPSRC National Facility for Aberration-Corrected STEM. References 1. Novoselov, K. S.; Jiang, D.; Schedin, F.; Booth, T. J.; Khotkevich, V. V.; Morozov, S. V.; Geim, A. K., Two-Dimensional Atomic Crystals. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 10451-10453. 2. Novoselov, K. S.; Falko, V. I.; Colombo, L.; Gellert, P. R.; Schwab, M. G.; Kim, K., A Roadmap for Graphene. Nature 2012, 490, 192-200. 3. Kisielowski, C.; Ramasse, Q. M.; Hansen, L. P.; Brorson, M.; Carlsson, A.; Molenbroek, A. M.; Topsøe, H.; Helveg, S., Imaging MoS2 Nanocatalysts with Single-Atom Sensitivity. Angew. Chem., Int. Ed. Engl. 2010, 49, 2708-2710. 4. Jaramillo, T. F.; Jorgensen, K. P.; Bonde, J.; Nielsen, J. H.; Horch, S.; Chorkendorff, I., Identification of Active Edge Sites for Electrochemical H2 Evolution from MoS2 Nanocatalysts. Science 2007, 317, 100-102. 5. Hansen, L. P.; Ramasse, Q. M.; Kisielowski, C.; Brorson, M.; Johnson, E.; Topsøe, H.; Helveg, S., Atomic-Scale Edge Structures on Industrial-Style MoS2 Nanocatalysts. Angew. Chem., Int. Ed. Engl. 2011, 50, 10153-10156. 6. Radisavljevic, B.; Radenovic, A.; Brivio, J.; Giacometti, V.; Kis, A., Single-Layer MoS2 transistors. Nat. Nanotechnol. 2011, 6, 147-150. 7. Mak, K. F.; Lee, C.; Hone, J.; Shan, J.; Heinz, T. F., Atomically Thin MoS2: A New Direct-Gap Semiconductor. Phys. Rev. Lett. 2010, 105, 136805. 10 8. Britnell, L.; Ribeiro, R. M.; Eckmann, A.; Jalil, R.; Belle, B. D.; Mishchenko, A.; Kim, Y. J.; Gorbachev, R. V.; Georgiou, T.; Morozov, S. V.; et al. Strong Light-Matter Interactions in Heterostructures of Atomically Thin Films. Science 2013, 340, 1311-1314. 9. Wang, Q. H.; Kalantar-Zadeh, K.; Kis, A.; Coleman, J. N.; Strano, M. S., Electronics and Optoelectronics of Two-Dimensional Transition Metal Dichalcogenides. Nat. Nanotechnol. 2012, 7, 699-712. 10. Nicolosi, V.; Chhowalla, M.; Kanatzidis, M. G.; Strano, M. S.; Coleman, J. N., Liquid Exfoliation of Layered Materials. Science 2013, 340, 1226419 11. Shi, Y.; Zhou, W.; Lu, A.-Y.; Fang, W.; Lee, Y.-H.; Hsu, A. L.; Kim, S. M.; Kim, K. K.; Yang, H. Y.; Li, L.-J.; et al. van der Waals Epitaxy of MoS2 Layers Using Graphene As Growth Templates. Nano Lett. 2012, 12, 2784-2791. 12. Liu, K.-K.; Zhang, W.; Lee, Y.-H.; Lin, Y.-C.; Chang, M.-T.; Su, C.-Y.; Chang, C.-S.; Li, H.; Shi, Y.; Zhang, H.; et al. Growth of Large-Area and Highly Crystalline MoS2 Thin Layers on Insulating Substrates. Nano Lett. 2012, 12, 1538-1544. Egerton, R. F., Control of Radiation Damage in the TEM. Ultramicroscopy 2013, 127, 13. 100-108. 14. Krivanek, O. L.; Dellby, N.; Murfitt, M. F.; Chisholm, M. F.; Pennycook, T. J.; Suenaga, K.; Nicolosi, V., Gentle STEM: ADF Imaging and EELS at Low Primary Energies. Ultramicroscopy 2010, 110, 935-945. 15. Meyer, J. C.; Eder, F.; Kurasch, S.; Skakalova, V.; Kotakoski, J.; Park, H. J.; Roth, S.; Chuvilin, A.; Eyhusen, S.; Benner, G.; et al. Accurate Measurement of Electron Beam Induced Displacement Cross Sections for Single-Layer Graphene. Phys. Rev. Lett. 2012, 108, 196102. 16. Mathew, S.; Gopinadhan, K.; Chan, T. K.; Yu, X. J.; Zhan, D.; Cao, L.; Rusydi, A.; Breese, M. B. H.; Dhar, S.; Shen, Z. X.; et al. Magnetism in MoS2 Induced by Proton Irradiation. Appl. Phys. Lett. 2012, 101, 102103-5. 17. Komsa, H.-P.; Kotakoski, J.; Kurasch, S.; Lehtinen, O.; Kaiser, U.; Krasheninnikov, A. V., Two-Dimensional Transition Metal Dichalcogenides under Electron Irradiation: Defect Production and Doping. Phys. Rev. Lett. 2012, 109, 035503. 18. Liu, X.; Xu, T.; Wu, X.; Zhang, Z.; Yu, J.; Qiu, H.; Hong, J.-H.; Jin, C.-H.; Li, J.-X.; Wang, X.-R.; et al. Top-Down Fabrication of Sub-Nanometre Semiconducting Nanoribbons Derived from Molybdenum Disulfide Sheets. Nat. Commun. 2013, 4, 1776. 19. Krivanek, O. L.; Chisholm, M. F.; Nicolosi, V.; Pennycook, T. J.; Corbin, G. J.; Dellby, N.; Murfitt, M. F.; Own, C. S.; Szilagyi, Z. S.; Oxley, M. P.; et al. Atom-by-atom Structural and Chemical Analysis by Annular Dark-Field Electron Microscopy. Nature 2010, 464, 571-574. 20. Zhou, W.; Zou, X.; Najmaei, S.; Liu, Z.; Shi, Y.; Kong, J.; Lou, J.; Ajayan, P. M.; Yakobson, B. I.; Idrobo, J.-C., Intrinsic Structural Defects in Monolayer Molybdenum Disulfide. Nano Lett. 2013, 13, 2615-2622. 21. Egerton, R. F.; Wang, F.; Crozier, P. A., Beam-Induced Damage to Thin Specimens in an Intense Electron Probe. Microsc. Microanal. 2006, 12, 65-71. 22. Nair, R. R.; Blake, P.; Blake, J. R.; Zan, R.; Anissimova, S.; Bangert, U.; Golovanov, A. P.; Morozov, S. V.; Geim, A. K.; Novoselov, K. S.; et al. Graphene as a Transparent Conductive Support for Studying Biological Molecules by Transmission Electron Microscopy. Appl. Phys. Lett. 2010, 97, 153102-3. 23. Pantelic, R. S.; Meyer, J. C.; Kaiser, U.; Stahlberg, H., The Application of Graphene as a Sample Support in Transmission Electron Microscopy. Solid State Commun. 2012, 152, 1375-1382. 11 24. Yuk, J. M.; Kim, K.; Aleman, B.; Regan, W.; Ryu, J. H.; Park, J.; Ercius, P.; Lee, H. M.; Alivisatos, A. P.; Crommie, M. F.; et al. Graphene Veils and Sandwiches. Nano Lett. 2011, 11, 3290-3294. 25. Yuk, J. M.; Park, J.; Ercius, P.; Kim, K.; Hellebusch, D. J.; Crommie, M. F.; Lee, J. Y.; Zettl, A.; Alivisatos, A. P., High-Resolution EM of Colloidal Nanocrystal Growth Using Graphene Liquid Cells. Science 2012, 336, 61-64. 26. Chen, Q.; Smith, J. M.; Park, J.; Kim, K.; Ho, D.; Rasool, H. I.; Zettl, A.; Alivisatos, A. P., 3D Motion of DNA-Au Nanoconjugates in Graphene Liquid Cell Electron Microscopy. Nano Lett. 2013, 13, 4556-4561. 27. Meyer, J. C.; Girit, C. O.; Crommie, M. F.; Zettl, A., Imaging and Dynamics of Light Atoms and Molecules on Graphene. Nature 2008, 454, 319-322. 28. Lee, Z.; Jeon, K.-J.; Dato, A.; Erni, R.; Richardson, T. J.; Frenklach, M.; Radmilovic, V., Direct Imaging of Soft-Hard Interfaces Enabled by Graphene. Nano Lett. 2009, 9, 3365- 3369. 29. Ramasse, Q. M.; Zan, R.; Bangert, U.; Boukhvalov, D. W.; Son, Y.-W.; Novoselov, K. S., Direct Experimental Evidence of Metal-Mediated Etching of Suspended Graphene. ACS Nano 2012, 6, 4063-4071. 30. Ramasse, Q. M.; Seabourne, C. R.; Kepaptsoglou, D.-M.; Zan, R.; Bangert, U.; Scott, A. J., Probing the Bonding and Electronic Structure of Single Atom Dopants in Graphene with Electron Energy Loss Spectroscopy. Nano Lett. 2013, 13, 4989-4995. 31. Brivio, J.; Alexander, D. T. L.; Kis, A., Ripples and Layers in Ultrathin MoS2 Membranes. Nano Lett. 2011, 11, 5148-5153. 32. Zan, R.; Bangert, U.; Ramasse, Q.; Novoselov, K. S., Imaging of Bernal Stacked and Misoriented Graphene and Boron Nitride: Experiment and Simulation. J. of Microsc. 2011, 244, 152-158. 33. Zan, R.; Bangert, U.; Ramasse, Q.; Novoselov, K. S., Interaction of Metals with Suspended Graphene Observed by Transmission Electron Microscopy. J. Phys. Chem. Lett. 2012, 3, 953-958. Berry, V., Impermeability of Graphene and its Applications. Carbon 2013, 62, 1-10. 34. 35. Suk, J. W.; Kitt, A.; Magnuson, C. W.; Hao, Y.; Ahmed, S.; An, J.; Swan, A. K.; Goldberg, B. B.; Ruoff, R. S., Transfer of CVD-Grown Monolayer Graphene onto Arbitrary Substrates. ACS Nano 2011, 5, 6916-6924. 12 Figure Captions Figure 1. Atomic resolution HAADF images (raw data) of pristine single layer MoS2. a) before and b) after consecutive scans, showing vacancy formation; c) first and d) last frame of movie_1 showing expansion of the damaged region; e) before, f) during g) after acquisition of an EEL spectrum image (SI) (f) is the HAADF image of the SI area (blue frame in (e)), taken simultaneously with the SI). The electron dose was 2.6x1010 e/Ǻ2. Figure 2. Atomic resolution Z-contrast images (raw data) of single layer MoS2 on graphene. a) diffraction pattern showing spots from both MoS2 (red dashed circle) and graphene (blue dashed circle); image b) before and c) after consecutive scans, showing no vacancy formation; d) the first and e) the last frame of movie_2 showing hole formation; image f) before, g) during and h) after acquisition of the EEL SI, showing damage in MoS2 at high electron dose (9.2x1010 e/Å2). Figure 3. Atomic resolution Z-contrast images (raw data) of single layer MoS2 encapsulated by graphene layers. a) diffraction pattern showing spots from both MoS2 (red dashed circle) and two graphene layers rotated by 23° with respect to each other (blue dashed circle); b) first and c) last frame of movie_3 showing no vacancy formation; d) after the movie showing damage in the graphene layer; e) before, f) during and g) after the EEL SI acquisition, showing no damage in the MoS2 even at the high electron dose of 1.7x1011 e/Å2. 13 Figure 1. 14 Figure 2. 15 Figure 3. 16 TOC Encapsulating a sheet of MoS2 within two single layers of graphene radiation damage helps control sustained when observing this material in a scanning transmission electron microscope. 17
1109.1958	2	1109	2012-12-14T10:06:48	Mode conversion by symmetry breaking of propagating spin waves	[ "cond-mat.mes-hall" ]	We study spin-wave transport in a microstructured Ni81Fe19 waveguide exhibiting broken translational symmetry. We observe the conversion of a beam profile composed of symmetric spin-wave width modes with odd numbers of antinodes n=1,3,... into a mixed set of symmetric and asymmetric modes. Due to the spatial homogeneity of the exciting field along the used microstrip antenna, quantized spin-wave modes with an even number n of antinodes across the stripe's width cannot be directly excited. We show that a break in translational symmetry may result in a partial conversion of even spin-wave waveguide modes	cond-mat.mes-hall	cond-mat	Mode conversion by symmetry breaking of propagating spin waves APS/123-QED P. Clausen,1 K. Vogt,1,2 H. Schultheiss,1,3 S. Schafer,1,4 B. Obry,1 G. Wolf,1 P. Pirro,1 B. Leven,1 and B. Hillebrands1 1Fachbereich Physik and Forschungszentrum OPTIMAS, Technische Universitat Kaiserslautern, D-67663 Kaiserslautern, Germany 2Graduate School of Excellence Material Science in Mainz, Staudinger Weg 9, D-55128 Mainz, Germany 3Material Science Division, Argonne National Laboratory, USA 4MINT Center, University of Alabama, Tuscaloosa, AL 35487, USA (Dated: July 22, 2011) Abstract We study spin-wave transport in a microstructured Ni81Fe19 waveguide exhibiting broken translational symmetry. We observe the conversion of a beam profile composed of symmetric spin-wave width modes with odd numbers of antinodes n = 1, 3, . . . into a mixed set of symmetric and asymmetric modes. Due to the spatial homogeneity of the exciting field along the used microstrip antenna, quantized spin-wave modes with an even number n of antinodes across the stripe's width cannot be directly excited. We show that a break in translational symmetry may result in a partial conversion of even spin-wave waveguide modes. 2 1 0 2 c e D 4 1 ] l l a h - s e m . t a m - d n o c [ 2 v 8 5 9 1 . 9 0 1 1 : v i X r a 1 f_Fig1.eps FIG. 1: (Color online) Concept of the investigated lateral spin-wave waveguide. The distribution of the spin-wave intensity measured by Brillouin light scattering (BLS) microscopy is color coded where black (white) indicates minimum (maximum) intensity, respectively. The measured intensity has been normalized to compensate for damping. For the advancement of the emergent field of magnonics and magnon-spintronics [1 -- 5], the understanding and characterization of spin waves in magnetic microstructures are essential. Many mechanisms for the excitation of spin waves on the micrometer length scale either by uniform microwave field [6], microstrip antennas [7, 8], spin-transfer torque [9, 10], or oscillating do- main walls [11] have been explored recently. However, the transport of spin waves in waveguide structures was mainly focused on one-dimensional geometries [12, 13]. The implementation of spin waves for the transport and processing of information even in devices as simple as a signal splitter will require changes of the spin-wave propagation direction. Modifications in the shape of a magnetic waveguide often result in inhomogeneous demagnetizing fields and non-uniform magnetization distributions. However, the spin-wave dispersion as well as the mode profiles adia- batically adjust to the local variations in the internal field [14]. This makes the overall propagation and quantization pattern complex and difficult to predict. A way out of this problem are specific design rules for the layout of spin-wave waveguide structures. In this Letter, we demonstrate that such design rules exist and specific mode properties can be utilized. We demonstrate this using the example of spin-wave transport in a waveguide with broken translational symmetry. The waveguide is shifted transverse to the spin-wave propagation direction resulting in a parallel offset of the waveguide axis behind the skew section (see Fig. 1). Despite being only a rather small deviation from a straight stripe, we show, that this skew has a profound impact on the spin-wave modes propagating in the waveguide. 2 The schematic layout of the investigated microstructure is shown in Fig. 1. A spin-wave wave- guide with dimensions of w = 2.5 µm × l = 100 µm is fabricated from a 60 nm thick Ni81Fe19 film using electron-beam lithography and conventional lift-off techniques. Spin waves are excited by the magnetic Oersted field of a 2 µm wide and 500 nm thick Cu microstrip antenna driven with frequencies ωrf/2π up to 18 GHz. In a distance of 4 µm from the antenna, a 3 µm long skew results in a 1 µm parallel offset of the waveguide. We apply a static magnetic field of Bext = 50.7 mT along the short axis of the waveguide (y-direction) to ensure the excitation of propagating spin waves travelling perpendicular to the magnetization direction with high group velocities. This allows for the investigation of propagation phenomena over large distances by means of Brillouin light scattering (BLS) microscopy [7, 15] as can be seen in the spin-wave intensity pattern overlaid in Fig. 1. A waveguide with a finite transversal width causes a discretization of the spin-wave wave vector in y-direction leading to a well-defined set of dispersion relations as shown in Fig. 2(a). These dispersion relations display the frequency dependence of the spin-wave modes as a function of the wave vector component kx in x-direction along the waveguide. The wave vector in y-direction, perpendicular to the waveguide, is quantized according to kn,y = nπ/weff, where the mode number n equals the number of antinodes in y-direction across the width of the stripe. In this geometry, the demagnetizing fields lead to a reduction of the effective quantization width weff smaller than the geometrical width w of the stripe. The calculation of the dispersion relations is made following the approach in [7] for an internal magnetic field of Bint = 39.5 mT and standard material parameters for Ni81Fe19 as summarized in [16]. The internal magnetic field is extracted from a micromagnetic simulation (oommf code, see [17]) and set to an averaged value over the assumed quantization width. The smaller value compared to the externally applied field is due to the demagnetizing fields originating from the magnetic charges at the boundaries of the waveguide. In the further discussion, we assume a sinusoidal mode profile in y-direction, as schematically displayed in Fig. 2(b) for the three lowest width modes. This leads to an overall spatial spin-wave amplitude given by ψn(x, y, t) = Ane−i(kxx−ωrf t) · sin(cid:16)nπ w y(cid:17), (1) where An is the maximum amplitude of the respective partial mode. For a given excitation fre- quency (7 GHz in Fig. 2(a)) several modes with different n can be excited simultaneously. These modes interfere with each other and form a stationary intensity distribution [18, 19]. The resulting 3 f_Fig2.eps FIG. 2: (Color online) (a) Frequencies of different lateral spin-wave modes as a function of the wave vector component along the waveguide (x-direction). The dashed lines depict the wave vectors for a fixed excitation frequency. (b) Schematic spin-wave mode profiles in y-direction. time-averaged interference pattern is the coherent superposition I(x, y) = (cid:12)(cid:12)(cid:12) X n 2 ψn(x, y)(cid:12)(cid:12)(cid:12) (2) of the amplitudes given in equation (1) and reflects the symmetry of the interfering modes. Only in the case when all spin-wave modes have an odd mode number n, the resulting intensity distribution is symmetric with respect to the center of the waveguide. Since in our experiment the exciting field of the antenna is homogeneous across the wave- guide's width, only spin-wave modes with a net component of the dynamic magnetization, i.e., odd modes with n = 1, 3, 5, . . ., can be directly excited. Therefore, only a symmetric spin-wave interference pattern is found in the first 4 µm behind the antenna as can be seen in the measured BLS intensity in Fig. 1. Inside the skew section, the spin-wave intensity distribution changes significantly. The incident spin-wave beam is not simply reflected at the edge of the skew section but rather attenuated. Furthermore, it excites a different mode which is strongly localized at the edge of the waveguide 4 in a spin-wave well [20]. Calculations show, that the inhomogeneity of the internal magnetic field Bint, which is modified by the demagnetizing field of the stripe, traps the spin waves in a region extending only several hundred nanometers from the edge of the stripe which is comparable to the lateral dimensions of the edge mode in Fig. 1. Behind the skew section, in 7 µm distance from the antenna, the interference pattern has an asymmetric component. This asymmetric spin-wave intensity distribution observed behind the skew can be described in good approximation as a superposition of the lowest even and odd mode with n = 1 and n = 2. This conversion from a purely symmetric system to a mixed set of sym- metric and asymmetric spin-wave modes is caused by an asymmetric distribution of the dynamic magnetization, which originates from the originally excited modes ending up with their intensity maximum displaced from the center of the stripe by the skew. As a result, the asymmetric mode with n = 2 is excited. To quantitatively analyze this effect we consider the periodicity of the spin-wave intensity. For each stationary spin-wave intensity distribution, being either symmetric or asymmetric, its period- icity dper in x-direction is determined by the difference ∆kx(ωrf) of the wave vector components kx of the interfering spin waves and is a function of the excitation frequency ωrf. In case the two lowest spin-wave modes are excited, this leads to a periodicity dper given by: dper(ωrf) = 2π (cid:0)kx,1(ωrf) − kx,2(ωrf)(cid:1) . (3) In the experiment, the values of kx,1 and kx,2 are determined by the spin-wave dispersion rela- tions shown in Fig. 2(a). Consequently, the periodicity dper can be controlled by either tuning the excitation frequency ωrf or the externally applied magnetic field Bext. To support the stated interpretation of the asymmetric interference pattern being composed of the spin-wave modes with n = 1 and n = 2, we measured and calculated the intensity distributions of these two interfering modes for excitation frequencies ωrf/2π ranging from 6.6 to 8.0 GHz. For calculating the interference pattern we use equations (1) and (2) with n = 1, 2 and the wave vectors determined from the corresponding dispersion relations as in Fig. 2(a) for each excitation frequency. The amplitudes An are free parameters in the calculation and were set to match the measured intensity distributions. However, their specific values will not influence the periodicity of the interference pattern. For three exemplary excitation frequencies, the intensity distributions measured behind the skew can be seen in Fig. 3(a) both for experiment (left sub-panels) and calculations (right sub- 5 panels). The resulting change in the periodicity dper of the snake-like mode pattern is evident in the experiment as well as in the calculations, both showing an excellent agreement. Figure 3(b) summarizes the results obtained from BLS microscopy measurements (filled sym- bols) in direct comparison with the theoretical expectations (line), which are given by equation (3). For increasing excitation frequencies, the split in the dispersion relations becomes more pro- nounced leading to an increase of ∆kx(ωrf). Therefore, the periodicity dper decreases in agreement with equation (3) as can be seen in Fig. 3(a) and (b). The agreement between measured data and calculations further supports our assumption of observing interference between the modes n = 1 and n = 2 behind the skew. In conclusion, we were able to manipulate the spin-wave mode patterns of propagating spin waves in a microscopic waveguide from being a superposition of symmetric modes to a mixed system of symmetric/asymmetric spin-wave modes. This controlled mode conversion is achieved by a shift of the waveguide transverse to the propagation direction of the spin waves creating a variation in the magnetization distribution within the skew section and leading to a strong concen- tration of the spin-wave amplitude on one side of the waveguide. Ultimately, this inhomogeneity causes the excitation of asymmetric spin-wave modes behind the skew section. This is not only an eminent chance to create spin-wave modes in microstructures which cannot be excited by antennas alone due to their homogeneous field. The study of spin-wave propagation and mode conversion in this structure may also lead to further insight into the interaction between spin waves and their excitation via the local dynamic magnetization. In addition, it is a first step to better understand and utilize spin-wave transport in truly two-dimensional microstructures, an essential step for the implementation of magnonics and magnon-spintronics into future devices. The authors thank Dr. P. A. Beck for deposition of the magnetic thin film and the Nano+Bio Center of the Technische Universitat Kaiserslautern for assistance in sample preparation. Financial support by the Carl-Zeiss-Stiftung and the Graduiertenkolleg 792 is gratefully acknowledged. [1] A. Khitun, M. Bao, and K. L. Wang, J. Phys. D 43, 264005 (2010). [2] V. V. Kruglyak, S. O. Demokritov, and D. Grundler, J. Phys. D 43, 264001 (2010). [3] A. A. Serga, A. V. Chumak, and B. Hillebrands, J. Phys. D 43, 264002 (2010). [4] V. V. Kruglyak and R. J. Hicken, J. Magn. Magn. Mater. 306, 191 (2006). 6 f_Fig3.eps FIG. 3: (Color online) (a) Comparison of the measured (left sub-panels) and calculated (right sub-panels) spin-wave interference patterns for different frequencies. (b) Black diamonds show the periodicity dper extracted from the interference patterns as illustrated in (a) as a function of the spin-wave frequency. The curve depicts calculated values. 7 [5] T. Schneider, A. A. Serga, B. Leven, B. Hillebrands, R. L. Stamps, and M. P. Kostylev, Appl. Phys. Lett. 92, 022505 (2008). [6] Y. Au, T. Davison, E. Ahmad, P. S. Keatley, R. J. Hicken, and V. V. Kruglyak, Appl. Phys. Lett. 98, 122506 (2011). [7] K. Vogt, H. Schultheiss, S. J. Hermsdoerfer , P. Pirro, A. A. Serga, and B. Hillebrands, Appl. Phys. Lett. 95, 182508 (2009). [8] V. E. Demidov, M. P. Kostylev, K. Rott, P. Krzysteczko, G. Reiss, and S. O. Demokritov, Appl. Phys. Lett. 95, 112509 (2009). [9] J. Slonczewski, J. Magn. Magn. Mater. 159, L1 (1996). [10] V. E. Demidov, S. Urazhdin, V. V. Tiberkevich, A. Slavin, and S. O. Demokritov, Phys. Rev. B 83 060406 (2011). [11] S. Hermsdoerfer, H. Schultheiss, C. Rausch, S. Schafer, B. Leven, S. Kim, and B. Hillebrands, Appl. Phys. Lett. 94, 223510 (2009). [12] M. Bauer, C. Mathieu, S. O. Demokritov, and B. Hillebrands, J. Appl. Phys. 81, 3971 (1997). [13] V. E. Demidov, J. Jersch, K. Rott, P. Krzysteczko, G. Reiss, and S. O. Demokritov, Phys. Rev. Lett. 102 177207 (2009). [14] C. Bayer, J. P. Park, H. Wang, M. Yan, C. E. Campbell, and P. A. Crowell, Phys. Rev. B 69, 134401 (2004). [15] V. E. Demidov, S. O. Demokritov, B. Hillebrands, M. Laufenberg, and P. P. Freitas, Appl. Phys. Lett. 85, 2866 (2004). [16] The material parameters used for calculating the spin-wave dispersions are: Saturation magnetization Ms = 800 kA/m Gyromagnetic ratio: γ /2π = 28 GHz/T Exchange stiffness constant: A = 1.6 · 10−11 J/m [17] M. J. Donahue and D. G. Porter, Interagency Report NISTIR 6376, National Institute of Standards and Technology, Gaithersburg, MD (1999). [18] O. Buttner, M. Bauer, C. Mathieu, S. O. Demokritov, B. Hillebrands, P. Kolodin, M. Kostylev, S. Sure, H. Dotsch, V. Grimalsky, Y. Rapoport, and A. N. Slavin, IEEE Trans. Magn. 34, 1381 (1998). [19] V. E. Demidov, S. O. Demokritov, K. Rott, P. Krzysteczko, and G. Reiss, Phys. Rev. B 77, 064406 (2008). [20] J. Jorzick, S. O. Demokritov, B. Hillebrands, M. Bailleul, C. Fermon, K. Y. Guslienko, A. N. Slavin, D. V. Berkov, and N. L. Gorn, Phys. Rev. Lett. 88, 047204 (2002). 8
1202.1690	1	1202	2012-02-08T13:30:44	Superconducting proximity effect through graphene from zero field to the Quantum Hall regime	[ "cond-mat.mes-hall" ]	We investigate the superconducting proximity effect through graphene in the long diffusive junction limit, at low and high magnetic field. The interface quality and sample phase coherence lead to a zero resistance state at low temperature, zero magnetic field, and high doping. We find a striking suppression of the critical current near graphene\rq{}s charge neutrality point, which we attribute to specular reflexion of Andreev pairs at the interface of charge puddles. This type of reflexion, specific to the Dirac band structure, had up to now remained elusive. At high magnetic field the use of superconducting electrodes with high critical field enables the investigation of the proximity effect in the Quantum Hall regime. Although the supercurrent is not directly detectable in our two wire configuration, interference effects are visible which may be attributed to the injection of Cooper pairs into edge states.	cond-mat.mes-hall	cond-mat	July 2, 2018 Superconducting proximity effect through graphene from zero field to the Quantum Hall regime. Katsuyoshi Komatsu, Chuan Li, S. Autier-Laurent, H. Bouchiat and S. Gu´eron Laboratoire de Physique des Solides, Univ. Paris-Sud, CNRS, UMR 8502, F-91405 Orsay Cedex, France. We investigate the superconducting proximity effect through graphene in the long diffusive junc- tion limit, at low and high magnetic field. The interface quality and sample phase coherence lead to a zero resistance state at low temperature, zero magnetic field, and high doping. We find a striking suppression of the critical current near graphene's charge neutrality point, which we attribute to specular reflexion of Andreev pairs at the interface of charge puddles. This type of reflexion, specific to the Dirac band structure, had up to now remained elusive. At high magnetic field the use of superconducting electrodes with high critical field enables the investigation of the proximity effect in the Quantum Hall regime. Although the supercurrent is not directly detectable in our two wire configuration, interference effects are visible which may be attributed to the injection of Cooper pairs into edge states. I. INTRODUCTION The celebrated electronic band structure of graphene leads to many interesting features. Among them is the possibility to tune its carrier density from electron to hole, with the consequence that the Integer Quantum Hall effect is observed over a wide range of magnetic fields. Another consequence is the fact that transport can proceed via carriers of either the conduction or the valence band, depending on the doping, and may even proceed via a conversion of one type of carrier into the other, across regions of different doping1 the so called Klein tunneling effect. It was suggested2 that a superconductor/graphene in- terface should also reveal the fact that the valence and conduction band touch at the so called Dirac point. In- FIG. 1: Sketch of the retro- and specular Andreev reflex- ion at a G/S interface. Left: Retro-reflexion occurs in usual conductors and in doped graphene, where the Fermi energy much exceeds the superconducting electrode's energy gap ∆, EF (cid:29) ∆. Right: The specular Andreev reflexion occurs in graphene at doping small enough that EF (cid:28) ∆. FIG. 2: Scanning electron micrograph of the graphene sample connected to Nb electrodes. The distance between electrodes is L = 1.2 µm and the graphene width is W = 12 µm. deed, transport across a Superconductor/Normal metal (S/N) interface at subgap energy implies extracting two electrons from the superconductor and injecting them into the N, which produces a correlated Andreev pair in the normal metal. In a usual normal metal, which is highly doped in the sense that the Fermi level lies well within the conduction band, both electrons are in- jected in the conduction band of the N. The two injected members of the Andreev pair then follow the same, al- beit time-reversed, diffusive path in the normal conduc- tor, so that coherent propagation can occur over several micrometers (the phase coherence length at low tem- perature). This coherent propagation leads to super- currents that flow through such normal conductors sev- eral microns long connected to two superconductors. In contrast, at a superconductor/graphene (S/G) interface, if the superconductor's Fermi level is aligned with the graphene Dirac point, the two electrons of a Cooper pair S S G G Nb Nb must split into an electron in the conduction band and the other in the valence band. The two members of the injected pair in the graphene now have the same velocity (rather than opposite) parallel to the S/G interface (see Fig. 1) and thus do not follow the same diffusive path. The observation of this special type of pair injection, also called "specular Andreev reflexion", has so far remained elusive. This is because the doping inhomogeneities in the graphene samples, of several millielectronvolts10, are much larger than the superconductor's energy gap. Thus only the usual injection of counter-propagating electron pairs (also called Andreev retroreflexion) sets in. In this article we show that diffusive transport of An- dreev pairs through quantum coherent graphene reveals an analog of specular Andreev reflexion at an S/G inter- face, in the form of specular reflexions of Andreev pairs at the interface between a doped charge puddle and a zero density region. These processes result in the destruction of counter-propagation upon specular reflexion, and lead to a large phase accumulation withing each Andreev pair. Since all pairs contribute to the supercurrent with their phase, the resulting supercurrent is suppressed. We ar- gue that this specular reflexion explains the suppression of the critical current that we observe near the Charge Neutrality Point (CNP) in our quantum coherent, long and diffusive SGS junctions. In the second part of the article, we explore the possi- bility of injecting Cooper pairs in graphene in the Quan- tum Hall regime. In contrast to the low field proxim- ity effect, the supercurrent is no longer carried by many diffusing pairs, but must be carried exclusively by the chiral edge states. Thus the two injected electrons must propagate on opposite edges of the graphene sheet. We present a long SGS junction which sustains a tunable su- percurrent at low magnetic field. The superconducting electrodes, made of a high critical field superconductor, remain superconducting at fields such that the graphene exhibits integer quantum Hall plateaus, indicating that transport proceeds via edge states. We present non lin- ear transport features which hint to the existence of in- terference, controlled by gate voltage or magnetic field, between the electrons propagating along different edges of the graphene. II. LONG JUNCTION SAMPLES IN LOW MAGNETIC FIELD A. Sample fabrication Several SGS junctions were fabricated, in which the length of graphene between S electrodes was greater than one micron, more than twice as long as previously reported3–7. Such lengths place a great constraint on the sample in order for a full proximity effect to develop: the phase coherence length must be longer than the sample length, and the interface quality must be excellent since a low transparency decreases the critical current through 2 the junction15. The critical current itself, in the case of a perfect interface, scales as the inverse length cubed (see discussion further down). In addition, the temperature must be low since the critical current is roughly exponen- tially suppressed by temperature with a coefficient pro- portional to the diffusion time across the sample, which scales as the square of the sample length16. Thus it is not surprising that not all samples we fab- ricated showed a full superconducting proximity effect at low temperature. Out of 12 samples with supercon- ducting electrodes (of different superconducting materi- als and contact layers), 3 exhibited a full proximity ef- fect when cooled to low temperature. All samples were mechanically exfoliated with the tape method and de- posited on a doped silicon substrate previously cleaned in an oxygen plasma. Standard electron-beam lithography was performed and the contacts were sputtered onto the samples after an hour long annealing step in vacuum at 100◦C. The contacts consist of a thin Pd layer, 4 to 8 nm thick, over which the superconducting layer, either Nb or ReW8, is deposited without breaking vacuum, with a thin Pd cover layer. We report in this article results on a SGS junction consisting of a W = 12 µm-wide graphene sheet with a L = 1.2 µm separation between Nb elec- trodes (Fig. 2). A second junction, with ReW electrodes, is W = 2.6 µm-wide with a length of L = 0.7 µm between electrodes. The samples are tested at room temperature and then thermally anchored to the mixing chamber of a dilution refrigerator, and measured via low pass filtered lines. B. Critical current in zero magnetic field Both SGS junctions display a gate tunable supercur- rent at low temperature, as shown in Fig. 3 and Fig. 4. As is clear in the figures, the critical current is strongly suppressed near the charge neutrality point, and we ar- gue that this suppression is due to the specular reflexion at the charge puddle interfaces. To quantify this suppres- sion, we compare the measured critical current (assumed here to be equal to the switching current, the current at which the junction resistance switches from zero to a fi- nite value) to the theoretically expected critical current (see Fig. 5 for the definition of the critical current). In the theory of the proximity effect in the diffusive, long junction limit, the critical current has a maximum zero temperature value given by the Thouless energy ET h divided by the normal resistance state RN , multiplied by a numerical factor α which depends on the junction length L: Ic = αET h/eRN , where ET h = ¯hD/L2, with D = vF le/2 the diffusion constant in two dimensions, vF the Fermi velocity and le the mean free path. The tunability of graphene is an asset to probe this relation. As shown in Fig. 5, one can compare the mea- sured switching current to the Thouless energy divided by the normal state resistance as the gate voltage is var- ied. It is clear from the figure that there is not a constant 3 factor between ET h/eRN and Ic but that Ic is strongly suppressed at small gate voltage, as the charge neutrality point is approached. This suppression has not been reported in the other graphene based SNS junctions, which are more than two or three times shorter than the devices reported in this article. To interpret the data of Fig. 5, we first discuss the maximum critical current and its temperature depen- dence (Fig. 7), which we explain by a non ideal interface. We then address the gate-voltage induced suppression of the critical current. The maximum critical current to be expected depends on how long the junction is, compared to the superconducting coherence length in the graphene layer, defined as ξs = (cid:112)¯hD/∆, with ∆ the electrode's superconducting gap, and D the diffusion constant in graphene. We find that L/ξs = 5 for the ReW sample and 7 for the Nb one, which places these junctions in the long (but not infinitely long) junction limit. As com- puted in16, this gives an expected coefficient α between Ic and ET h/eRN of 9 and 8 at zero temperature for the Nb and ReW samples respectively, close to the 10.8 value of the infinitely long junction. These theoretical values are more than twenty times larger than the maximum measured α coefficient of 0.5 for Nb and 0.3 for ReW. This reduced critical current is a feature noted in prac- tically all experiments on S/graphene/S junction, and is attributed to partial transmission at the S/graphene interface. The temperature dependence of the critical current (Fig. 7) confirms the partial transmission of the interface, since the critical current decay with temper- ature is faster than expected for a perfect interface, as described in15. Fig. 7 shows the variations of the differ- ential resistance curves with temperature, as well as the comparison of the critical current suppression with theo- retical prediction considering an opaque interface. From the comparison one can extract a rather large interface resistance, roughly five times larger than the resistance of the graphene sheet itself. We now argue that the critical current suppression near the CNP cannot be attributed to finite tempera- ture. The effect of temperature is twofold: First, the thermal fluctuations induced by kBT must be smaller than the Josephson coupling EJ = ¯h 2e Ic, which gives a minimal supercurrent of 44 nA/K. Thus the minimal crit- ical current at the experimental temperatures are 9 nA at 200mK and 2 nA at 50 mK, and do not depend on gate voltage. Second, temperature decreases the switch- ing current in a manner that is predicted by the Usadel equations, that has been numerically solved exactly16, and that can be approximated by an exponential decay as Ic(T ) ≈ Ic(0)e−T /10ET h for a perfect interface and Ic(T ) ≈ Ic(0)e−T /3ET h for an opaque interface. Since the overall variation with gate voltage of the Thouless energies of both samples (deduced from the measured re- sistance R via the diffusion constant D = with n the carrier density) is less than a factor 50% (between L W 1 ne2R FIG. 3: Proximity effect in graphene connected to Nb elec- trodes at 200 mK. Upper left panel: dV /dI vs Idc for different galte voltages, and, bottom left panel, its 2 dimensional color plot. The suppression of critical current in a gate voltage region of ±10V around the charge neutrality point is notice- able. Upper right panel: I(V) curves for different gate volt- ages, showing how the proximity effect varies between a full proximity effect with zero resistance at high doping, and quasi normal behavior with a linear IV around the charge neutral- ity point. Lower right panel: Zero bias differential resistance as a function of gate voltage in the normal state, from which the RN is determined. A small magnetic field was applied to destroy the constructive interference leading to the supercur- rent. FIG. 4: Proximity effect in graphene connected to ReW elec- trodes at 55 mK. Top panel: dV /dI vs Idc. Bottom left panel: 2 dimensional colour plot emphasizing the suppression of the supercurrent around the charge neutrality point. Right panel: Resistance as a function of gate voltage in a small magnetic field which suppresses the constructive interference leading to supercurrent. 4 ular reflection of an Andreev pair at the charge pud- dle contours, as sketched in Fig. 6. Indeed, around the CNP, electron-doped regions coexist with hole doped ones, forming a network of so called puddles10. Where the doping varies from n to p doping there is necessar- ily a boundary with exactly zero doping, to within kBT , termed a 0 region. Thus a time-reversed Andreev pair formed by the usual Andreev retroreflexion at the super- conductor/graphene interface has, near the CNP, a large probability of encountering a n/0 or p/0 boundary. At such boundaries such junctions, a specular-like reflexion must occur when two counter propagating electrons dif- fusing in the n-doped region are converted into two elec- trons belonging to two different bands in the 0 region. The change in relative velocity destroys the counter prop- agation of the pair. As the two electrons diffuse across the rest of the graphene, they undergo uncorrelated scatter- ing events and their relative phase difference increases. Since the total supercurrent is the sum of all contri- butions from the propagating Andreev pairs, construc- tive interference is destroyed when counter-propagation is lost, and thus the supercurrent is suppressed (see fig. 6). Interestingly, the effect of these puddles is immense in the superconducting state (and presumably all the more so that the superconducting coherence length, the "size" of the pair, is small with respect to the puddle size), whereas it is much weaker in the normal state where thanks to Klein tunneling, the puddles do not suppress single quasiparticle propagation so much. In summary, whereas the specular Andreev reflexion in ballistic S/G/S junctions can yield a supercurrent11, we have shown that in diffusive S/G/S junctions a specular- like reflexion of Andreev pairs at p/0 or n/0 junctions leads to accumulation of phase difference within the An- dreev pair. The critical current is then suppressed, in a manner which depends on the number of such n/0 (or p/0) junctions within the sample. This translates into a critical current suppressed most near the charge neu- trality point. The supercurrent suppression by charge puddles is thus expected to be largest in samples that are long (large ratio of sample length L to puddle size, typically larger than 50 nm10) and connected to super- conductors with large gaps, corresponding to smaller su- perconducting coherence lengths (ξs = (cid:112)¯hD/∆ is typ- ically 125 nm in graphene for Nb (∆ = 1.6 meV) or 170 nm for ReW (∆ = 1.2 meV , as compared to 350 nm for Al (∆ = 0.2 meV), given the diffusion constant D = 4.10−2m2/s in these graphene samples. FIG. 5: Comparison of switching current with Thouless en- ergy. Upper left panel: Two ways of defining the switching current: Ic, the largest current for which the differential resis- tance dV /dI is zero, and I∗ c , the inflection point of the jump in dV /dI towards large resistance. Upper right panel: Varia- tions of the Thouless energy with gate voltage, deduced from the sample resistance in the normal state, for both samples. The resistance of the Nb sample was measured at 1K. The resistance of the ReW sample was measured at 55 mK at a current bias above the critical current of the proximity effect. ∗ Bottom panels: Comparison of Ic and I c with ETh/eRN for the sample with Nb electrodes at 200 mK, and with ReW electrodes at 55 mK. FIG. 6: Sketch of how the specular reflexion of an Andreev pair at an n/0 junction can lead to loss of counterpropagation and thus large phase accumulation within an Andreev pair. The red region is electron doped, the blue one hole doped, and the green region in between has zero doping. 20 and 30 µeV for the Nb sample, and between 40 and 60 µeV for the ReW sample, see Fig. 5), it cannot ex- plain the gate-voltage induced suppression by a factor 10. Thus it is clear that the remarkable suppression of the su- percurrent near the Charge Neutrality Point cannot be explained by temperature-induced effects. We attribute this suppression close to the CNP to spec- C. Junction under radiofrequency irradiation As also reported by others, the junctions display Shapiro steps, i.e. replica of the zero resistance state, which appear at finite dc voltage, when submitted to ra- diofrequency irradiation (via an antenna placed near the sample). This is shown in the top panel of Fig. 8 for the sample with ReW electrodes, at high doping, which * Ic* Vg(V) Vg(V) 5 FIG. 7: Temperature dependence of the proximity effect through the ReW sample. Left panel, differential resistance curves at temperatures ranging from 100 mK to 800 mK. Right panel, comparison of the extracted critical currents with the theoretically expected decay with temperature, for differ- ent ratios r of the contact resistance to the graphene sheet resistance. Both the overall suppression of the critical cur- rent with respect to the Thouless energy at low temperature, and the decay with temperature, are accounted for assuming a ratio r of roughly 7. displays a full proximity effect with a critical current of 130 nA. What is more original is the observation of se- quential non linearities in the IV curves of the junctions at gate voltages such that a full proximity effect with a zero resistance state does not develop, demonstrating that non linearities in the IV curve are sufficient to induce phase locking and replica of non linear features (bottom panel). D. Suppression of supercurrent by small magnetic field Fig. 9 displays the differential resistance as a func- tion of current for different magnetic fields, and shows that the supercurrent is suppressed in an oscillatory man- ner, as expected for wide proximity junctions13,14. How- ever the supercurrent is not recovered periodically, but rather the resistance oscillates away from zero in a peri- odic manner. We attribute the absence of full supercur- rent recovery to the asymmetric (trapezoid-like) shape of the graphene samples, and to probable irregularities in the transmission between electrodes and graphene, which lead to inhomogeneous supercurrent densities26. The fact that the oscillation period is smaller than one flux quan- tum Φ0 through the sample is attributed to the focusing effect of the field by the superconducting electrodes. Al- though the interference patterns look similar for both samples, one can notice an asymmetry in the field de- pendence of the sample with ReW electrodes, which we attribute to trapped flux in these high Hc2 but low Hc1 FIG. 8: Effect of radiofrequency irradiation on the junction with ReW electrodes. Top panel: Junction under irradiation of 2.4 GHz, at a gate voltage of Vg = −25V . Bottom panel: junction under irradiation of 2.4 GHz, at a gate voltage of Vg = −7V for which no full proximity effect (supercurrent) is observed, but only a lower low bias differential resistance. Nonetheless Shapiro like features develop under irradiation. The insets display the rf power dependence of the proximity effect. The arrows point to the dc voltage plateaus, distant by ∆V = 5.3 µV , close to the expected interval ∆V = ¯hω/2e = 4.9 µV . electrodes. III. PROXIMITY EFFECT IN THE INTEGER QUANTUM HALL REGIME The observation of supercurrent through graphene con- tacted to the high Hc superconductor ReW (see previous part) suggests the exciting possibility of observing a su- percurrent running through a conductor in the Quantum Hall regime. Such a supercurrent would have to be car- ried by edge states, so that the time reversed electrons in- jected from the superconductor would be injected into the edge states at the opposite edges of the sample. Only few authors have considered this scenario theoretically17,18. They have shown that in principle such a proximity ef- 6 FIG. 9: Low field dependence of dV /dI(I) for the sample with Nb (left) and ReW (right) electrodes, at T=200 mK for Nb and 55 mK for ReW, and at high doping. Bottom panels: Line traces of dV /dI at zero current bias as a function of magnetic flux through the graphene. We attribute the small period of the flux dependence to strong focusing of the magnetic field by the large superconducting electrodes. fect is possible in the integer quantum Hall regime, with a maximal critical current given by the ballistic limit of evd/L, where L is the perimeter of the sample and vd the drift velocity. In the following we show that we achieve the quantum Hall regime in graphene with supercon- ducting electrodes, and present elements which suggest the existence of coherent interference within the sample, modulated by magnetic field or gate voltage, hinting to a tunable proximity effect through graphene in the quan- tum Hall regime. A. Integer Quantum Hall regime Fig. 10 displays the zero current differential resistance of the SGS junction as a function of gate voltage, for fields between 0 and 7.5 T, at low temperature (70 mK). The quantum Hall effect is visible, in the form of plateaus, at fields above 5 T. Indeed, it has been shown that the quantum Hall regime is detectable in a two wire mea- surement, in the form of regions in which the conduc- tance is quantized at the Hall conductance value28. The exact shape of the conductance versus filling factor curve (i.e., whether peaks or dips separate the plateau regions) depends on the sample aspect ratio since the two wire resistance is a weighed combination of the sample's ρxx 28. Fig. 11 shows that the filling factors corre- and ρxy sponding to the plateaus are those expected for graphene (ν = nh/(eB) = ±2,±6, ...), but that the values of the conductance plateaus are larger than those expected for graphene. We attribute this discrepancy to scattering, which broadens the Landau levels, and to sample inhomo- geneities typical of wide graphene sheets, which change the plateaus conductance values, as has been observed by FIG. 10: Two wire differential resistance as a function of gate voltage for the sample with ReW (Hc > 7.5 T ), at magnetic fields from 0 to 7.5 T, every Tesla between 0 and 5 T, and every 0.5 Tesla above 5 T. Temperature is 70 mK. The inset displays how the Hall plateau at 7.5 T and Vg = −14 V flattens out as temperature is increased. FIG. 11: Quantum Hall effect of graphene sample with ReW electrodes, plotted as a function of filling factor ν = en/Bh. Inset: Two wire differential conductance as a function of fill- ing factor in the Quantum Hall regime. Main panel: zoom of the ν = −6 region, which displays oscillations in conductance of up to 10% at the edge of the plateau. others28. The factor of almost 3 in conductance enhance- ment could also be interpreted as due to three effective samples in parallel. We note that this lack of correct quantification is found in wide samples and not in square samples (see e.g. the quantum Hall regime in a different sample with Nb electrodes and a square shape, shown in the Appendix in Fig. 16). 3000300025002500200020001500150010001000500500dV/dI (W)-20-1001020Vg (V)B. Proximity effect in the Quantum Hall regime The curves of Fig. 10 show no obvious trace of su- percurrent at high field in the form of a zero resistance state, even though the electrodes are superconducting up to more than 7.5 T (we found a critical current of 3.5 µA at 7.5 T and low temperature, measured through slightly wider ReW leads, see appendix). This is in fact to be expected, given the two wire configuration which mixes the ρxx and ρxy components. It is actually interesting to ask what should be the signature of a supercurrent in the quantum Hall regime, especially measured in a two wire configuration. Some hints of the superconducting proximity effect can be found, however. We show below that we find sig- natures of the proximity effect both in the incoherent regime, where the S/graphene/S junction can be viewed as two uncorrelated S/graphene junctions in series, and in the coherent regime, where signatures of the coherent propagation of pairs through the graphene via quantum Hall edge states are visible. The incoherent proximity effect is visible in the shape of the plateaus themselves. As shown in the inset of Fig. 10 for the ν = −6 plateau, and also reported in AlGaAs/GaAs heterostructures connected to high Hc2 NbN electrodes19, the plateaus are far less flat at low temperature than at high temperature. The resistance at the transition between two Hall plateaus exhibits a non monotonous variation with filling factor, with a de- crease of resistance of up to ten percent. This amplitude variation of the resistance was interpreted in19 as the ef- fect of a change in conductance at an NS interface with respect to an NN interface as the edge channel trans- mission coefficient changes with filling factor. Analytical and numerical computations of the NS conductance in the specific case of the quantum Hall regime were con- sidered in21,22. They predict that the NS conductance is not twice the NN conductance, in contrast to what one might naively expect for two electrons being transmit- ted via perfectly conducting edge channels at the Quan- tum Hall plateau. This is because the two electrons of a pair must travel along different edges, much as in the normal case. However interference effects at the NS in- terfaces lead to a predicted oscillatory behavior around the Quantized Hall conductance in21. When disorder at the interface is included,22 find that the two-wire conduc- tance is at most the Quantized Hall value , in contrast to our experimental results and those of19. Signatures of a coherent proximity effect (i.e. a co- herent propagation of pairs and a supercurrent) in the quantum Hall regime are visible when one exploits the non linearity of the reproducible fluctuations in the con- ductance (or resistance) as a function of magnetic field or gate voltage. These fluctuations, which stem from quantum interference between different diffusive trajec- tories, are known to be amplified in the case of supercon- ducting contacts23. But in some instances in this sample we find that the interference leads to a decrease of dif- 7 ferential resistance around zero current, in contrast to the peaked differential resistance at zero current that is commonly observed in disordered samples at low tem- perature (due to electron-electron interactions or to the effect of the electromagnetic environment24,25). Since in our two wire geometry the quantized Hall resistivity adds to the zero longitudinal resistance of a supercurrent, we do not expect a zero two-wire resistance. But the signa- ture of the supercurrent should be the differential resis- tance dip at zero bias. In addition, it was predicted in17 that the supercurrent intensity should be modulated by the Fermi energy or the magnetic field, in an Aharonov Bohm- like way. And, interestingly, we do observe al- ternating constructive and destructive interference, as a function of changing gate voltage or magnetic field, are demonstrated in Figs. 12, 13, and 14. Similar features have been reported in 2D electron gases made in het- erostructures in20 with varying magnetic fields, but not gate voltages, and in samples in which no supercurrent was demonstrated at low field, in contrast to what we have achieved (see section II). In fact, we find that the dips in the differential resistance have an amplitude of up to 50 Ω, and a current range of about 100 nA (see Figs. 12, 13, and 14), comparable to the critical current measured in zero field. Fig. 15 illustrates how the low bias curvature of the differential resistance Vs current curves alternates in sign as the magnetic field is swept: the third derivative of the voltage Vs current curve is negative if the differential re- sistance is dipped at zero bias (induced proximity effect), but positive if the differential resistance is peaked (be- cause of destructive interference, disorder, interactions). The oscillations, reminiscent of mesoscopic fluctuations, are reproducible and can be characterized by a correla- tion field Bc ≈ 100 G which varies with magnetic field and ac current excitation. The fact that we find signatures of supercurrent (dips in the zero bias differential resistance) at low gate volt- ages (from -7 to +4V, see Fig. 12), for which the su- percurrent was much reduced in zero field (see Fig. 4), points to the radically different effect of charge puddles at low and high magnetic field: we argued above that in zero field charge puddles tend to destroy the supercurrent because specular reflexion at the boundary between two oppositely charged puddles separates the two members of the Andreev pair, leading to large phase accumulation within each Andreev pair. The averaging of the many dif- fusing Andreev pairs leads to destructive interference. In contrast, in the quantum Hall regime, conduction pro- ceeds in a ballistic-like fashion via a small number of channels (the edge states). If an edge state encounters a puddle boundary, it has been shown27–29 that edge trans- port can proceed via an "ambipolar snake state", made up of cyclotronic propagation with opposite rotations in the p and n regions. Such edge states may contain a phase that depends on the specific disorder and puddle configuration at each edge, but the total accumulated dephasing over the round trip between the two super- 8 FIG. 12: Differential resistance versus dc current at selected gate voltages between -7 and 4 V at 7.5 T at 70 mK in the sample with ReW electrodes. The thicker curves are those which display a negative differential resistance at zero cur- rent, indicative of a superconducting proximity effect in the quantum Hall regime. Curves are not offset vertically. FIG. 14: Differential resistance as a function of dc current for selected curves at slightly different magnetic fields, around 4.2 T, at 55 mK and Vg = 0, for the sample with ReW electrodes. The zero current differential resistance alternates between a peak and a dip, signaling the alternating nature of interference between transmitted Andreev pairs. The curves are offset vertically by 100 Ω for clarity. IV. CONCLUSION In conclusion, we have shown that a proximity effect can be induced in a graphene junction up to 1.2 microm- eters long. We find a stong suppression of the super- current near the charge neutrality point, and attribute it to the specular Andreev reflexion specific to mono- layer graphene, at the boundaries between p and n pud- dles. This effect is all the stronger that the supercon- ducting coherence length is short and that the junction is long, since Andreev pairs cannot avoid these junction regions. In the Quantum Hall regime, a two wire mea- surement cannot reveal directly a supercurrent carried by edge states. But we argue that the dip in differential resistance at zero current is a signature of a supercur- rent, flowing through the graphene via edge states which interfere constructively. This interference is modulated by gate voltage and magnetic field, as expected theoret- ically. The question that needs to be addressed in the future is how to demonstrate that a supercurrent is cir- culating in the structure in the quantum Hall regime. Since a two wire transport measurement necessarily dis- plays non zero resistance, one must find a different exper- imental configuration. In addition, it will be necessary to devise a method of distinguishing the dissipation-less supercurrent from the dissipation-less edge state trans- port. The detection of an orbital magnetic moment with a signature of pairs (via its field periodicity)30 may be a route towards this fascinating goal. FIG. 13: Differential resistance versus dc current at 7.5 T, for gate voltages regularly distributed between -13 V and -15 V, in the sample with ReW electrodes. The alternation from dipped to peaked differential resistance at zero bias, with a 10% variation, confirms that the oscillation in the ν = −6 Hall plateau (seen in Fig. 10) is due to the proximity effect. Curves have been offset vertically for clarity. conducting electrodes should not average to zero, at low filling when only few edge channels propagate. Therefore the tuning of interference and thus of the proximity ef- fect in the quantum Hall regime is expected over a larger gate voltage range than in the low field diffusive trans- port case. Acknowledgements 9 We acknowledge H. Raffy for suggesting ReW as a su- perconducting material and the use of her cryostat, and B. Altshuler, J. Cayssol, B. Dassonneville, R. Deblock, V. Falko, M. Ferrier, J-N. Fuchs, M. O. Goerbig, A. Ka- sumov, and L. K. Lim for discussions. Appendix A. Quantum Hall effect in two samples with Nb electrodes and different aspect ratios We show in Fig. 16 how the quantization expected for a two wire measurement of monolayer graphene in the Quantum Hall effect is better verified in a square sample than a wide geometry. B. Superconductivity of the ReW electrodes Although we could not test the critical field of the elec- trode portion lying directly in contact with the graphene, we measured the critical current as a function of mag- netic field of slightly wider ReW wires, and found that the critical current was larger than 3 µA at 55 mK, see Fig. 17. FIG. 15: Third derivative as a function of magnetic field at Vg = 0 and T=55 mK, around 4.3 and 6.25 T. Increasing and decreasing field sweep directions are shown to demonstrate the reproducibility of the curves. A negative third derivative corresponds to a dipped differential resistance near zero bias, indicative of a superconducting proximity effect in the quan- tum Hall regime. Such a dipped differential resistance curve, alternating with peaked ones, is shown in Fig. 14. 1 M.I. Katsnelson, K.S. Novoselov and A.K. Geim, Nature Physics 2, 620 (2006); V. V. Cheianov and V. I. Falko, Phys. Rev. B 74, 041403 (2006). 2 C.W. J. Beenakker, Phys. Rev. Lett. 97, 067007 (2006). 3 H. B. Heersche, P. Jarillo-Herrero, J. B. Oostinga, L. M. K. Vandersypen, and A. F. Morpurgo, Nature 446, 56 (2007). 4 X. Du, I. Skachko, and E. Y. Andrei, Phys. Rev. B 77, 184507 (2008). 5 C. Ojeda-Aristizabal, M. Ferrier, S. Gu´eron, and H. Bouch- iat, Phys. Rev. B 79, 165436 (2009). 6 I. V. Borzenets, U. C. Coskun, S. J. Jones, and G. Finkel- stein, Phys. Rev. Lett. 107, 137005 (2011). 7 G.H. Lee, D. Jeong, Jae-Hyun Choi, Yong-Joo Doh, and Hu-Jong Lee, Phys. Rev. Lett. 107, 146605 (2011); Dongchan Jeong, Jae-Hyun Choi, Gil-Ho Lee, Sanghyun Jo, Yong-Joo Doh, and Hu-Jong Lee, Phys. Rev. B 83, 094503 (2011). 8 H. Raffy and R. B. Laibowitz, Phys. Rev. B 30, 5541 (1984). 9 K. D. Usadel, Phys. Rev. Lett. 25, 507 (1970). 10 J. Martin, N. Akerman, G. Ulbricht, T. Lohmann, J. H. Smet, K. von Klitzing, A. Yacoby, Nature Physics 4, 144- 148 (2008). 11 Titov, M., and C. W. J. Beenakker, 2006, Phys. Rev. B 74,041401(R). 12 J. R. Williams, D. A. Abanin, L. DiCarlo, L. S. Levitov, and C. M. Marcus, Phys. Rev. B 80, 045408 (2009) ; J. R. Williams and C. M. Marcus, Phys. Rev. Lett. 107, 046602 (2011). 13 3 J. C. Hammer, J. C. Cuevas, F. S. Bergeret, and W. Belzig, Phys. Rev. B 76, 6 (2007). 14 Chiodi et al., arXiv:1201.3509v1. 15 J. C. Hammer, J. C. Cuevas, F. S. Bergeret, W. Belzig, Phys. Rev. B 76, 064514 (2007); J. C. Cuevas and F. S. Bergeret, Phys. Rev. Lett. 99, 217002 (2007). 16 P. Dubos, H. Courtois, B. Pannetier, F. K. Wilhelm, A. D. Zaikin, and G. Schon, Phys. Rev. B 63, 064502 (2001); P. Dubos, H. Courtois, O. Buisson, and B. Pannetier, Phys. Rev. Lett. 87, 206801 (2001). 17 Ma and Zyuzin, Europhys. Lett. 21, 941 (1993); A. Yu. Zyuzin, Phys. Rev. B 50 (1994). 18 Stone and Lin, Phys. Rev. B 83, 224501 (2011). 19 Hideaki Takayanagi, Tatsushi Akazaki, Minoru Kawamura, Yuichi Harada, Junsaku Nitta, Physica E 12, 922 (2002); 20 Takayanagi and T. Akazaki, Physica 249, 462 (1998); T.D. Moore and D.A. Williams, Phys. Rev. B 59, 7308 (1999). 21 Y. Takagaki, Phys. Rev. B 57, 4009 (1998). 22 H. Hoppe, U. Zulicke, and G. Schon, Phys. Rev. Lett. 84, 1804 (2000). 23 C.W.J. Beenakker, Reviews of Modern Physics 69, 731 (1997). 24 B. L. Altshuler and A. G. Aronov, Electron-electron inter- actions in disordered systems, A.L. Efros and M. Pollak eds., p. 1 (Elsevier 1985) . 10 25 F. Pierre, H. Pothier, P. Joyez, N.O. Birge, D. Esteve and M. Devoret, Phys. Rev. Lett. 86, 1590 (2001). 26 A. Barone and G. Paterno, Physics and Application of Physics and Applications of the Josephson Effect, Wiley (1982), chap. 4. 27 C. W. J. Beenakker, Rev. Mod. Phys. 80, 1337 (2008). 28 J. R. Williams, L. DiCarlo, and C. M. Marcus, Science 317, 638 (2007). 29 P. Carmier, C. Lewenkopf, and D. Ullmo, Phys. Rev. B, 81, 241406(R) (2010). 30 U. Sivan and Y. Imry, Phys. Rev. Lett. 61,1001 (1988); J. P. Eisenstein it et al., Phys. Rev. Lett. 55, 875 (1985); S. A. J. Wiegers et al., Phys. Rev. Lett. 79, 3238 (1997). FIG. 16: Quantum Hall effect in two samples with Nb elec- trodes: the wide sample of Fig.2 which displays a supercur- rent at low field, and a square sample, that displayed no full proximity effect, but whose Hall quantization is closer to that expected for graphene. The data shows the two wire differ- ential resistance as a function of filling factor, measured at 5T. FIG. 17: Inset: Critical current as a function of magnetic field of two ReW leads that are slightly wider than the sample elec- trodes, at low temperature. Main panel, two wire differential resistance of one of the wire at different magnetic fields. 383430262218141062G(e2/h)-16-12-8-404812nSquare sampleWide sampleNb
1810.05646	2	1810	2019-02-04T19:21:32	Wiedemann-Franz law and Fermi liquids	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci", "cond-mat.quant-gas" ]	We consider in depth the applicability of the Wiedemann-Franz (WF) law, namely that the electronic thermal conductivity ($\kappa$) is proportional to the product of the absolute temperature ($T$) and the electrical conductivity ($\sigma$) in a metal with the constant of proportionality, the so-called Lorenz number $L_0$, being a materials-independent universal constant in all systems obeying the Fermi liquid (FL) paradigm. It has been often stated that the validity (invalidity) of the WF law is the hallmark of an FL (non-Fermi-liquid (NFL)). We consider, both in two (2D) and three (3D) dimensions, a system of conduction electrons at a finite temperature $T$ coupled to a bath of acoustic phonons and quenched impurities, ignoring effects of electron-electron interactions. We find that the WF law is violated arbitrarily strongly with the effective Lorenz number vanishing at low temperatures as long as phonon scattering is stronger than impurity scattering. This happens both in 2D and in 3D for $T<T_{BG}$, where $T_{BG}$ is the Bloch-Gr\"uneisen temperature of the system. In the absence of phonon scattering (or equivalently, when impurity scattering is much stronger than the phonon scattering), however, the WF law is restored at low temperatures even if the impurity scattering is mostly small angle forward scattering. Thus, strictly at $T=0$ the WF law is always valid in a FL in the presence of infinitesimal impurity scattering. For strong phonon scattering, the WF law is restored for $T> T_{BG}$ (or the Debye temperature $T_D$, whichever is lower) as in usual metals. At very high temperatures, thermal smearing of the Fermi surface causes the effective Lorenz number to go below $L_0$ manifesting a quantitative deviation from the WF law. Our work establishes definitively that the uncritical association of an NFL behavior with the failure of the WF law is incorrect.	cond-mat.mes-hall	cond-mat	Wiedemann-Franz law and Fermi liquids Ali Lavasani, Daniel Bulmash, and Sankar Das Sarma Department of Physics, Condensed Matter Theory Center, University of Maryland, College Park, Maryland 20742, USA and Joint Quantum Institute, University of Maryland, College Park, Maryland 20742, USA We consider in depth the applicability of the Wiedemann-Franz (WF) law, namely that the electronic thermal conductivity (κ) is proportional to the product of the absolute temperature (T ) and the electrical conductivity (σ) in a metal with the constant of proportionality, the so-called Lorenz number L0, being a materials-independent universal constant in all systems obeying the Fermi liquid (FL) paradigm. It has been often stated that the validity (invalidity) of the WF law is the hallmark of an FL (non-Fermi-liquid (NFL)). We consider, both in two (2D) and three (3D) dimensions, a system of conduction electrons at a finite temperature T coupled to a bath of acoustic phonons and quenched impurities, ignoring effects of electron-electron interactions. We find that the WF law is violated arbitrarily strongly with the effective Lorenz number vanishing at low temperatures as long as phonon scattering is stronger than impurity scattering. This happens both in 2D and in 3D for T < TBG, where TBG is the Bloch-Gruneisen temperature of the system. In the absence of phonon scattering (or equivalently, when impurity scattering is much stronger than the phonon scattering), however, the WF law is restored at low temperatures even if the impurity scattering is mostly small angle forward scattering. Thus, strictly at T = 0 the WF law is always valid in a FL in the presence of infinitesimal impurity scattering. For strong phonon scattering, the WF law is restored for T > TBG (or the Debye temperature TD, whichever is lower) as in usual metals. At very high temperatures, thermal smearing of the Fermi surface causes the effective Lorenz number to go below L0 manifesting a quantitative deviation from the WF law. Our work establishes definitively that the uncritical association of an NFL behavior with the failure of the WF law is incorrect. (cid:18) kB (cid:19)2 e L0 = π2 3 I. INTRODUCTION In 1853, Franz and Wiedemann1 made the experimental discovery that the ratio of the thermal (κ) to the electrical con- ductivity (σ) in several metals is approximately the same at the same temperature. Almost thirty years after this discov- ery, Lorenz established2 that this ratio of κ/σ is in fact propor- tional to the absolute temperature T , and therefore κ/(σT ) is a universal constant in all metals: κ σT = L(T ) = L0, (1) where L0, dependent only on the fundamental constants kB and electron charge e, is universally called the Lorenz number, given by: = 2.45 × 10−8 W/K2. (2) We will call L0 the ideal Lorenz number, and L(T ), the effective Lorenz number in case that this ratio deviates from the ideal L0 value. The finding that κ/(σT ) = L0 is univer- sally called the Wiedemann-Franz (WF) law. In usual 3D met- als, the room temperature value of L0 is remarkably universal (with L ∼ L0 within 5%), making the WF law one of the most applicable defining characteristics of metallic (and hence, FL) properties3. If L(T ) deviates from L0 in a substantive man- ner, it is referred to as the failure of the WF law, which is then often attributed to the breakdown of the underlying quasiparti- cle picture and a failure of the Fermi liquid description for the relevant physics. The current work is on a theoretical study of L(T ) in 2D and 3D metals (Fermi liquids) where the elec- tron liquid is coupled to static random impurities (disorder) and acoustic phonons (phonon bath). We show that L(T ), (cid:18) kB (cid:19)2 e L(T ) = 3 2 depending on the temperature and the details of the parame- ters describing the coupled electron-impurity-phonon system, could manifest arbitrary values of L(T ) with the effective Lorenz number strongly suppressed from the ideal value L0 entirely within the Fermi liquid theory without invoking ei- ther a breakdown of the quasiparticle picture or the Fermi liq- uid paradigm. We establish beyond any doubt that the mere inapplicability of the WF law to a metal does not necessarily imply an underlying NFL behavior, and the widespread use of the validity (invalidity) of the WF law as a smoking gun for an underlying FL (NFL) behavior is unwarranted. Drude4,5 provided a simple classical kinetic theory for the WF law, leading to the following formula: = 1.24 × 10−8 W/K2. (3) It is interesting that Drude's purely classical derivation of the WF law provides a result which is fortuitously within a factor of 2 of the ideal Lorenz number later derived by Sommerfeld6 using the appropriate quantum theory of solids. This arises from the fact that both derivations use elastic im- purity scattering (e.g. quenched impurities or defects) as the driving kinetic mechanism for the electrical (σ) or thermal (κ) transport by the same carriers, consequently leading to the canceling out of the materials parameters (e.g. effective mass, carrier density) in the ratio κ/σ of the system. This has led to the belief that any lack of such a cancellation, with L(T ) manifesting strong temperature dependence, must automati- cally imply a hypothetical NFL situation where the particles carrying charge current and the particles carrying heat current are distinct, leading to the failure of the WF law. While an intrinsic separation of charge and energy transport occurring through different channels would most likely lead to a failure of the WF law (since the two transport mechanisms are then 9 1 0 2 b e F 4 ] l l a h - s e m . t a m - d n o c [ 2 v 6 4 6 5 0 . 0 1 8 1 : v i X r a no longer kinetically connected), such a failure does not have to necessarily imply the breakdown of the FL theory. By con- trast, it has actually been known7 since the early days of the theory of metals that the same particles (namely, electrons or FL quasiparticles) could in fact carry charge and heat current through very different kinetic rates if the operational scatter- ing mechanism is inelastic. Since phonons provide such an in- elastic scattering mechanism, in general, a coupled electron- phonon system should always violate the WF law. The fact that ordinary metals obey the WF law at room temperatures in spite of their transport properties (both electrical and thermal) being dominated by phonons is simply a manifestation of the fact that the room temperature is a very high (very low) char- acteristic temperature for phonons (electrons) since the typ- ical phonon Debye temperature TD (Fermi temperature TF ) for metals is TD ∼ 100 (TF ∼ 50, 000)K. In any system with high Debye temperature (much higher than room temper- ature), the WF law should be violated even at room tempera- tures (since, then, the phonon scattering will be inelastic even at room temperatures), and in all metals, the WF law is indeed strongly violated at low temperatures T (cid:28) TD unless impu- rity scattering starts dominating over phonon scattering. In a very clean metal L(T ) will be vanishingly small at T (cid:28) TD as long as impurity scattering is negligible. Thus, the validity or the failure of the WF law has little to do with NFL behavior, and is connected with the elastic or in- elastic nature of the resistive scattering mechanism dominat- ing transport in the system in the temperature regime where L(T ) is being measured. In particular, in an FL coupled to impurities and phonons there are two important energy scales (assuming the Fermi temperature TF to be very high as it al- ways is in the usual 3D metals): Ti and Tp. For T < Ti, impu- rity scattering dominates, and the WF law is strictly valid by virtue of scattering by quenched impurities being elastic. For Ti < T < Tp, inelastic phonon scattering dominates, and the WF law is strongly violated. For T > Tp, phonon scattering is quasielastic as one enters the so-called equipartition regime, and the WF law is restored again. Note that for strongly disor- dered systems, where Ti > Tp, WF law is obeyed at all tem- peratures. These are the main theoretical results we present for both 2D and 3D metals in our work. We note that Tp may or may not be directly connected to the Debye temperature TD except that Tp < TD. In particular, Tp may be of the or- der of the characteristic temperature scale TBG, where TBG is the so-called Bloch-Gruneisen temperature of the system as- sociated with the energy of a phonon with wave vector equal to 2kF , i.e., kBTBG = 2cskF , where cs is the phonon ve- In situations where TBG > TD (as in normal met- locity. als), Tp ∼ TD/3. In our work, we consider the situation TBG < TD in contrast to regular 3D normal metals where TD < TBG. The reason for our choice of TBG < TD is that our interest is in relatively low-density metals (e.g. cuprates) where this condition is likely to be met by virtue of kF being small. There are a few caveats one must keep in mind in this con- text. First, when optical phonons, with a fixed energy EO, are present in the system (we consider only acoustic phonons whose energy goes as csq for phonon wave number q) then 2 the definition of Tp simply becomes Tp ∼ EO. Thus in the presence of strong optical phonon scattering, the violation of the WF law may persist to rather high temperatures since EO could be large. We do not consider optical phonons since they are typically absent in metals as a strong scattering mecha- nism. Second, if the Fermi temperature is low (e.g. low- density systems) so that T ∼ TF (an impossibility in usual metals since TF > 10, 000K), then the WF law is weakly violated even for T > Tp because thermal Fermi surface smearing makes the system behave classically. Third, all ef- fects of electron-electron interaction are ignored in this work; electron-electron interaction effects on the WF law (without any phonon effects) have recently been discussed in Ref. 8. Thus, our work includes disorder and phonons whereas Ref. 8 includes disorder and electron-electron scattering effects. Fi- nally, the physics should be similar if phonons are replaced by some other bosonic scattering mechanisms leading to the resistivity, e.g. magnons, paramagnons or spin fluctuations. This type of scattering, if present, should also produce the violation of the WF law at low enough temperature in clean enough systems as long as the temperature for measuring L(T ) is below the characteristic temperature scale for inelas- tic scattering by these bosonic excitations to be operational. We mention that some of the results we present are qualita- tively known. But no existing work covers the whole subject of the WF law in the context of Fermi liquids as we do in the current work, covering both 2D and 3D systems and impu- rity and phonon scattering. (All our 2D results are completely new.) We believe that it is important to have all of these results for the temperature-dependent WF law in one comprehensive paper since there seems to be much confusion in this topic. In particular, a large fraction of the community seems to believe that the failure of the WF law (i.e. L(T ) < L0) is sufficient to conclude that the underlying material is a NFL with no well- defined quasiparticles. This is simply untrue. The violation of the WF law may or may not be a necessary condition for the NFL behavior9, but an observation of such a violation most certainly is not sufficient to conclude that the relevant system is a NFL. We refrain from reviewing the rather large literature connecting the violation of the WF law as an automatic sig- nature for NFL behavior since our focus is entirely on a well- defined FL theory (with impurities and acoustic phonons) for the validity or not of the WF law. There are many publica- tions discussing the violation of the WF law in the context of putative NFL behavior, and we cite a few here as representa- tive examples10 -- 13, simply to emphasize the importance of the subject matter of our work where the violation of the WF law is studied entirely in the context of FL physics. The rest of this article is organized as follows: In Sec. II we provide the main theory and the associated numerical re- sults for the calculated effective Lorenz number as a function of temperature for both 2D and 3D systems; Sec. III provides extensive discussions and a conclusion putting our results in the appropriate context of the violation or not of the WF law with reference to the applicability or not of the Fermi liquid paradigm. Five relevant appendices (A-E) provide the details of the theory complementing the presented results in the main text. we relegated the theoretical details to a series of appen- dices so that our main message (sections II and III and the figures in the main text) can be read and understood without referring at all to the theoretical details. II. THEORY AND RESULTS We use the standard Boltzmann kinetic theory with appro- priate approximations to calculate the temperature-dependent effective Lorenz number in a FL metal in the presence of static disorder (arising from random quenched impurities) and acoustic phonons (treated within the Debye model) in the continuum long wavelength jellium model. We start with a brief review of the Boltzmann equation and the mathematical framework we used to study transport coefficients while high- lighting the approximations which were employed to this end. We only present the final results in this section and leave most of the detailed and step by step calculations to the appendices (A-E) which should be consulted for the technical details. A. Boltzmann equation Let f (r, k, t) denote the distribution function of electron wave packets at position r with wave vector k at time t. Evo- lution of f (r, k, t) is governed by the Boltzmann equation14: ∂f ∂t + r · ∂f ∂r + k · ∂f ∂k = Ik{f}, (4) where r and k are given by semiclassical equations of motion15: 1 ∂ε(k) ∂k − k × Ω(k) r = v = e k = − E(r, t) − e c r × B(r, t), (5) where ε(k) is the band dispersion, Ω(k) is the Berry curva- ture and E (B) is the external electric (magnetic) field. For the parabolic band dispersion which we assume in this paper, Ω(k) = 0. We also set B to zero since we are only interested in the zero field longitudinal conductivity. Ik{f} in Eq.(4) is the collision integral given by (2π)3 [S(k, k(cid:48))fk(1 − fk(cid:48)) Ik{f} = − (cid:90) d3k(cid:48) + S(k(cid:48), k)fk(cid:48)(1 − fk)], (6) where S(k, k(cid:48)) is the differential scattering rate from state k to state k(cid:48) and can be computed using Fermi's golden rule for various scattering mechanisms. We write f as f (r, k, t) = f0(r, k, t) + δf (r, k, t), where f0 is the distribution function of fermions in local equi- librium, given by the Dirac distribution (cid:20) (cid:18) ε(k) − µ(r, t) kBT (r, t) (cid:19) (cid:21)−1 f0(r, k, t) = exp + 1 , Once the relaxation times appearing in the ansatz (8) have been calculated, transport coefficients can be obtained using 3 (cid:20) (cid:21)(cid:18) (cid:19) and δf is the deviation from that. If we plug in this form into the Boltzmann equation and only keep terms of first order in external fields and temperature gradient, we arrive at the linearized Boltzmann equation ∂δfk ∂f0 ∂ε eE + ε(k) − µ ∇T T − + v · = Ik{f0 + δf}, ∂t (7) where E = E +∇µ/e is the electrochemical force and δfk = δf (r, k, t). We have assumed that temperature and electric field are both slowly varying in space. Since our interest is in linear response, we will work with the linearized Boltzmann equation in the rest of this article. We are interested in the steady state solution and hence the first term in Eq.(7) can be dropped. In the linear response regime, considering the symmetries of the problem, the fol- lowing ansatz can be used to solve the linearized Boltzmann equation7: δfk =τσ(ε, T ) e v · E + τκ(ε, T ) (cid:19) (cid:18) ∂f0 v · ∇T T ∂ε ε(k) − µ (cid:18) ∂f0 (cid:19) ∂ε , (8) where τσ and τκ are generally unknown functions which are generically distinct (hence allowing for the generic possibil- ity of a failure of the WF law). We call τσ and τκ electrical and thermal relaxation times respectively. As we will men- tion shortly, whenever the scattering mechanism is elastic, these two relaxation times become equal, leading necessarily to the WF law. By contrast, for inelastic scattering, thermal and electrical relaxation times could be completely different, thus leading to a total failure of the WF law independent of the validity or not of the FL paradigm. The key for the va- lidity (or not) of the WF law is the elastic or inelastic nature of carrier scattering, and not the FL or NFL nature of the un- derlying electron system. Whenever transport is dominated by inelastic scattering (e.g. low temperatures T (cid:28) TD for acoustic phonons), the corresponding scattering mechanism may strongly violate the WF law. Some formal details of the Boltzmann transport theory are provided in Appendix A for completeness. B. Transport coefficients For a given external electric field and temperature gradient in the linear response regime, electrical current Je and thermal current Jq would be(cid:18)Je (cid:19) (cid:18)LEE LET (cid:19)(cid:18) E (cid:19) = Jq LT E LT T ∇T . (9) the following expressions: ∂f0 ∂ε ∂f0 ∂ε − (cid:18) (cid:18) (cid:18) (cid:18) − dε dε −∞ (cid:90) +∞ (cid:90) +∞ (cid:90) +∞ (cid:90) +∞ −∞ −∞ −∞ dε ∂f0 ∂ε − dε − ∂f0 ∂ε x(ε)τσ(ε), D(ε) v2 (cid:19) (cid:19) (cid:19) (ε − µ) D(ε) v2 (cid:19) (ε − µ) D(ε) v2 LEE = e2 LET = e T LT E = −e LT T = −1 T x(ε)τκ(ε), x(ε)τσ(ε), (ε − µ)2 D(ε) v2 x(ε)τκ(ε), (10) where D(ε) is the density of states at energy ε. (cid:17) (cid:16) LT T − LT E LET Electrical conductivity σ is simply the LEE coefficient. The thermal conductivity κ, is defined such that Jq = −κ∇T when Je = 0, and with a little bit of algebra turns out to be κ = − In the following sections, we study transport properties of FL in the presence of two different scattering mechanisms; electron-impurity scattering and electron-phonon scattering. First, we consider each scattering mechanism separately and then we study their combined effect. LEE . C. Impurity scattering The constant of proportionality depends on the parameters of the system and can be found in Appendix B. Having computed the relaxation time, one can obtain the electrical and thermal conductivities using Eq.(10). The calculated results are shown using dimensionless units in Figs.1(a) and (b). As is clear from the plots, the only temperature scale that appears in this model is the Fermi temperature defining the zero point energy of the noninteracting electrons. For T (cid:28) TF , the electrical conductivity does not have any temperature dependence and the thermal conductivity is linear in T : σ(T ) ∝ T 0, κ(T ) ∝ T. The WF law is obeyed in this regime (see Fig. 1c): L L0 = 1 (T (cid:28) TF ). (15) (16) On the other hand, for T (cid:29) TF , we get the following temper- ature scalings which differ based on the spatial dimension: (cid:40) (cid:40) 4 2D and 3D (we note that there could be temperature depen- dence if somehow the impurity potential u0 itself has temper- ature dependence, a possibility we ignore in the current work): (cid:40) τ (ε) ∝ ε0 ε−1/2 2D 3D (14) We consider the model of randomly distributed impurities with short range potential, σ(T ) ∝ T 0 T −1/2 2D 3D , κ(T ) ∝ T T 1/2 2D 3D. (17) Vimp(r) = u0 δ(r), (11) where u0 is some constant characterizing the scattering strength. The impurities are assumed to be fixed and hence the scattering would be elastic. The differential scattering rate can be calculated by using Fermi's golden rule and averaging over impurity locations: S(k, k(cid:48)) = 2π nimp u2 0 δ(ε(k) − ε(k(cid:48))), (12) where nimp corresponds to the number density of impurities. Due to the elasticity of scattering, the ansatz in Eq.(8) makes it simple to find an exact solution of the linearized Boltzmann equation for any arbitrary external electric field and tempera- ture gradient. By plugging the ansatz in Eq.(8) into Eq.(7), we can find an explicit closed form for the relaxation times. As is shown in Appendix B, due to the elasticity of scattering, thermal and electrical relaxation times are equal and can be determined from the following integral, τ (ε)−1 = (2π)d (1 − cos(θ))S(k, k(cid:48)), (13) where θ is the so-called scattering angle between k and k(cid:48). The integral can be carried out and we get a temperature inde- pendent relaxation time with different energy dependence in (cid:90) ddk(cid:48) System parameters still cancel out in the κ/σ ratio in this regime, which is related to the fact that the energy and charge currents both relax with the same rate. However, smearing of the Fermi surface at T (cid:29) TF causes the Lorenz ratio L/L0 to deviate from 1, approaching 6 π2 < 1 as is shown in Fig.1c. The full expressions for σ(T ) and κ(T ) as well as details of the calculation can be found in Appendix B. Thus, a modified WF law still applies for elastic impurity scattering at very high temperatures with a suppressed effec- tive Lorenz number L < L0. This is of possible experimental relevance in low density metallic systems where the T > TF regime may be attainable. Obviously, this high-T result is of no relevance to normal metals where TF ∼ 50, 000K. D. Phonon scattering In this section we consider the effect of electron-phonon scattering on transport coefficients. Other than the electron- phonon interaction, we do not incorporate any other lattice effect into our model and work in the continuum approxima- tion. In particular, we assume a parabolic energy dispersion for the electrons and ignore Umklapp scattering. We also ne- glect phonon drag. Finally, since our interest is strictly in the behavior of the Fermi liquid itself, we calculate only the elec- tronic contribution to the thermal conductivity. 5 FIG. 1. (a) electrical conductivity versus temperature for free electron model in presence of random impurity. (b) thermal conductivity for the same system. (c) the Lorenz ratio L/L0 for the same system. In all figures, solid line and dashed line correspond to 2D and 3D systems respectively. Note that electrical and thermal conductance has been plotted relative to their value at T = 0.1 TF to make them dimensionless. The electron-phonon interaction is given by the so-called deformation potential model: (cid:34)(cid:115) D2 (cid:88) (cid:35) 1 √V q 2ρωq k,k(cid:48),q (aq + a† He-ph = −q) ck†ck(cid:48) δk−k(cid:48)−q,0 (18) where a† and c† are phonon and electron creation operators respectively. Here D is the deformation potential strength, ρ is the ion density, and ωq corresponds to the energy of a phonon with momentum q, which, for acoustic phonons, is given as: ωq = csq (19) with cs the speed of sound. The corresponding scattering rate, obtained from Fermi's golden rule, is: (cid:19) (cid:18) 2π + (Nq + 1) δ(k − k(cid:48) − ωq)], D2q2 2ρωq [Nq δ(k − k(cid:48) + ωq) S(k, k(cid:48)) = where q = k − k(cid:48), and Nq is the phonon distribution function. Since all calcula- tions are carried out to the leading order in external fields, Nq can be replaced by the equilibrium Bose-Einstein distribution function, Nq = 1 eβωq − 1 . (22) In contrast to the elastic case, the linearized Boltzmann equation cannot be solved exactly here because inelastic electron-phonon scattering couples the distribution function at one energy to the distribution function at another energy. To arrive at a closed form for the relaxation times, we use the relaxation time approximation, discussed in detail in Ap- pendix C; this uncontrolled approximation assumes that the relaxation time changes slowly enough as a function of energy that, within the collision integral, τ (ε) ≈ τ (ε(cid:48)). Importantly, We find that the relaxation times for electrical transport and thermal transport are generally different. Note that we are not getting into the discussion of whether a relaxation time approximation is valid here or not as we are simply and uncritically assuming it to apply. (See Appendix C and E for more details.) One can of course solve the linearized Boltzmann integral equation directly numerically without as- suming the relaxation time approximation (which may indeed be necessary if one is interested in a quantitative compari- son with experimental results), but such a completely numer- ical calculation would destroy the whole purpose of our work since we are then unable to make general comments about the validity or not of the WF law. Assuming the existence of a relaxation time (albeit possibly different ones for charge and heat transport) enables us to make considerable analytical progress without losing generality (but sacrificing quantitative accuracy). To get the relaxation time for charge transport, we use the ansatz in Eq.(8) with ∇T set to zero. Assuming T (cid:28) TF and using the relaxation time approximation, we get: 1 − f (ε(cid:48)) 1 − f (ε) (1 − cos(θ))S(k, k(cid:48)). (23) On the other hand, to obtain the thermal relaxation time τκ, we set electrochemical force E to zero and solve the linearized Boltzmann equation in the presence of a non-zero temperature gradient ∇T . By comparing to a more direct (but more com- plicated) calculation, we show in Appendix C that the uni- versal features of the thermal relaxation time can be approxi- mately captured by the same expression as in Eq.(23), but by simply dropping the (1 − cos(θ)) "forward-scattering" sup- pression factor in the integral16: τκ(ε)−1 = S(k, k(cid:48)). (24) Intuitively, one can understand this difference between τσ and τκ by noting that different types of scatterings are responsible for relaxing charge and heat currents. Note that forward scat- tering events can not change the charge current significantly; (cid:90) ddk(cid:48) (2π)d 1 − f (ε(cid:48)) 1 − f (ε) (20) (21) (cid:90) ddk(cid:48) (2π)d τσ(ε)−1 = 10−2100102T/TF10−1100σ(T)/σ(0.1TF)σ2D3D10−2100102T/TF10−1101103κ(T)/κ(0.1TF)κ2D3D10−2100102T/TF0.60.70.80.91.0L/L0L/L02D3D(a)(b)(c)T0T−1/2TTT1/2L/L0=6π2L/L0=16 −1 versus temperature for thermal and electrical transport. (b) Electrical conductivity σ versus temperature. (c) FIG. 2. (a) Scattering rate τ Thermal conductivity κ versus temperature. (d) Lorenz ratio L/L0 versus temperature. In all graphs, solid and dashed lines correspond to 2D and 3D systems respectively. Relaxation rates and transport coefficients are plotted relative to their values at T = TBG to make them dimensionless. The Fermi energy is chosen such that TF = 103TBG a fact that explains the (1 − cos(θ)) factor in Eq.(23). On the other hand, the thermal current which is caused by the imbal- ance in the populations of electrons and holes in the vicinity of the Fermi surface (see Eq.(8)), can be relaxed effectively by forward scattering which justifies the absence of (1 − cos(θ)) in Eq. (24)16,17. Thus, in the presence of inelastic scattering processes, it is sensible to include (exclude) the forward scat- tering suppression factor for charge (heat) current within the relaxation time approximation. We remark that when the scat- tering is elastic, e.g. the impurity scattering, backscattering is the only relaxation mechanism for the thermal current (as well as the charge current) and hence the (1 − cos(θ)) factor can- not be dropped in that case. Detailed calculation of relaxation times is left to the Appendix C whereas in Appendix E we provide a detailed discussion of the relaxation time approxi- mation (RTA) in this context. In particular, we discuss how using different relaxation times could result in violations of the Onsager relations. With electron-phonon scattering present, both thermal and electrical relaxation times become functions of temperature. Fig. 2(a) shows their temperature dependence over a range of temperatures which covers multiple orders of magnitude. There are two different regimes, with the crossover occurring roughly at kBTBG = 2cskF . Note that because T is much less than TF , only electrons in the vicinity of the Fermi surface participate in charge and energy transport. This in turn means that momentum transfer in a scattering event is bounded by ∼ 2kF . Therefore, kBTBG represents an upper bound on the energy of phonons contributing to the current relaxation. We assume that TBG < TD, where TD is the Debye temperature; since TD only enters the problem as another upper bound on phonon energy, TD is not an important scale in the problem. We note that textbooks usually do not emphasize the impor- tance of TBG in the context of metallic transport limited by phonon scattering14 since for normal metals, kF is typically very large (since normal metals have very high carrier den- sity) leading to TBG > TD, and hence the phonon energy cut off for normal metals is invariably TD and not TBG. Since the results for TD being the cut off are already available in the lit- erature, we focus on the relatively low density situation where TBG < TD, leading to TBG being the appropriate phonon cut off. For high density regular metals, TBG in our results should simply be replaced by TD; basically, the phonon cut off is ei- ther TBG or TD depending on whichever is smaller for the particular material. For T (cid:29) TBG, thermal and electrical relaxation rates be- 10−2100102104106T/TBG10−1010−610−2102106τ(εF,T)−1/τ(εF,TBG)−1τ−1σ,2Dτ−1κ,2Dτ−1σ,3Dτ−1κ,3D10−2100102104106T/TBG10−41011061011σ(T)/σ(TBG)2D3D10−2100102104106T/TBG10−1101103105κ(T)/κ(TBG)2D3D10−2100102104106T/TBG0.000.250.500.751.00L/L02D3DT4T2TFT3T5TT−1T0TTFT−1/2T−2TFT−4T−1T−1/2TFT−5T−3/2(a)(b)(c)(d)L/L0=6π2L/L0=152π2L/L0=1L/L0∼T2(cid:40) (cid:40) come equal and scale linearly with temperature, independent of the spacial dimension (this is the so-called phonon equipar- tition regime where the acoustic phonon scattering is essen- tially quasi-elastic in metals): τσ(T )−1 = τ−1 (25) But for T (cid:28) TBG, charge current relaxes at a smaller rate than the thermal current(see Fig. 2a): κ (T ) ∝ T. τ−1 σ (T ) ∝ T 4 T 5 2D 3D , τ−1 κ (T ) ∝ T 2 T 3 2D 3D. (26) This can be understood intuitively as follows. Note that charge and thermal current carried by an electron can be roughly writ- ten as its charge and energy respectively times its velocity: je ∼ e v, jq ∼ ε v (27) where ε denotes the energy relative to the chemical poten- tial. Now, the only way a scattering event can relax the charge current is by changing the electron's velocity vector since its charge is strictly conserved. This is the reason that backscattering is the most effective way to relax the charge current. On the other hand, the thermal current can be re- laxed by either changing the electron's velocity or just chang- ing its energy when inelastic scattering processes are opera- tional. When TBG (cid:28) T , the scattering becomes quasi-elastic since phonon's energy is bounded by TBG so a single scat- tering event can only change an electron's energy by a small fraction of its average value ε ∼ T . As a result both thermal and charge current relaxations are dominated by backscatter- ing events for T (cid:29) TBG and hence the two relaxation times become equal. On the other hand for T (cid:28) TBG, backscatter- ing is exponentially suppressed due to the Bose distribution function whereas thermal current can now be relaxed effec- tively by changing the electrons' energy. As a result, charge current relaxes much more slowly than the thermal current. As can be seen from Eq.(26), for both 2D and 3D we have: τ−1 σ κ ∝ T 2 τ−1 (28) and this ratio goes to zero as as T goes to 0. As one would expect, due to different relaxation times, the WF law will no longer be obeyed in this regime. In fact, if the electron-phonon scattering is the only resistive mechanism (i.e. very clean met- als with no impurities), then this WF law violation is arbitrar- ily strong since L(T ) vanishes as T approaches zero, making L(T ) (cid:28) L0 even for a simple FL! Thus, all one needs is a very clean FL to see an arbitrarily strong violation of the WF law at low temperatures. Note that to violate the WF law, suppressing backward scat- tering just by itself is insufficient; it is crucial to allow for inelasticity. To see this clearly, one may consider the sim- ple case of random impurities, but with a scattering potential which suppresses scattering events with momentum transfers larger than some constant value q0. This situation actually arises in, for example, delta-doped two-dimensional electron 7 gases where the mobile carriers and impurities live in differ- ent layers; q0 is then given by 2π/z0 where z0 is the layer separation18. Although backscattering is suppressed in these systems for q0 < kF , WF law is still obeyed due to the elas- tic nature of the scatterings. This problem is studied in de- tail in Appendix B 2, clearly establishing that the absence of backscattering by itself, without any inelasticity, does not lead to any violation of the WF law. By plugging in the thermal and charge relaxation times into Eq. (10), electrical and thermal conductivity, σ(T ) and κ(T ) can be obtained (See Appendix C). The result is plotted in Figs. 2b and c. Although we expect that the relaxation time approximation to become less valid as one approaches T ∼ TF , we have used the same expression for the relaxation time throughout all temperature scales, even for T (cid:29) TF . Hence, our results at T (cid:29) TF should not be interpreted quantitatively but are rather only intended to demonstrate qualitatively that at high temperatures, the Lorenz number would deviate from unity due to the smearing of the Fermi surface. In the equipartition regime where TBG (cid:28) T (cid:28) TF , ther- mal conductivity is independent of temperature whereas elec- trical conductivity decreases as 1/T (the well known linear growth of resistivity in metals due to phonons19), both inde- pendent of dimension: σ(T ) ∝ T −1, κ(T ) ∝ T 0. (29) The WF law is obeyed in this regime due to the quasi-elastic nature of scatterings, L L0 = 1 (TBG (cid:28) T (cid:28) TF ). (30) On the other hand for T (cid:28) TBG, temperature scalings of σ and κ become dimension dependent: σ(T ) ∝ T −4 T −5 2D 3D , κ(T ) ∝ T −1 T −2 2D 3D, (31) recovering the well known T 5 scaling for electrical resistivity (known as the Bloch-Gruneisen formula20,21) and T 2 scaling for thermal resistivity (1/κ) in 3D metals22. As one also ex- pects from Eq.(28), WF law is parametrically violated in this regime with the Lorenz ratio vanishing as L L0 ∝ T 2 (T (cid:28) TBG), (32) for small temperatures. The calculated Lorenz ratio through- out all three temperature regimes is plotted in Fig.2d for both 2D and 3D systems. We remark that even though the violation of the WF law in this system can be traced back to different relaxation times (see Eq.(28)), the WF law in FL could still be violated at low temperatures when both energy and charge transport are char- acterized by a single relaxation time. In Appendix D, we pro- vide results under the assumption of a single relaxation time controlling both charge and heat currents (which is not valid generally for phonon scattering). This violation is, however, not arbitrarily strong as L(T )/L0 eventually becomes a con- stant (< 1) for T (cid:28) TBG. (cid:40) (cid:40) E. Impurity & phonon scattering Finally we consider the case where both scattering mecha- nisms (impurity scattering and electron phonon scattering) are present. With our current approximations, the scattering rates add, which leads to 1 τtotal(ε) = 1 τimp(ε) + 1 τel-ph(ε) (33) 8 imp and τ−1 el-ph correspond to the scattering rate from Where τ−1 impurities and phonons respectively. Using Eq.(10) it is straightforward to compute transport coefficients and hence the Lorenz ratio. The Lorenz ratio for 2- and 3- dimensional systems for three different impurity coupling strength is plotted in Fig.3. As can be seen in the figure, at low enough temperatures, the WF law is always obeyed due to the fact that eventually impurity scattering dominates transport because phonons will no longer be thermally excited at sufficiently low tempera- tures (but the impurity scattering is present even at T = 0). However, as one increases the temperature, phonon scattering become stronger and, at some intermediate temperature scale Ti, eventually overcomes impurity scattering as the dominant scattering mechanism. At Ti, the Lorenz ratio starts to devi- ate from unity. Clearly, Ti is not universal and depends on the specific parameters of the sample. WF law is violated for Ti (cid:28) T (cid:28) TBG, but is recovered again for TBG (cid:28) T (cid:28) TF where the system is in the equipartition regime. Thus, the vi- olation or not of the WF law in a pure Fermi liquid depends entirely on the details of the electron-phonon and electron- impurity scattering. As long as the impurity scattering is weak (i.e. in a relatively clean metal), the WF law will be violated strongly for Ti (cid:28) T (cid:28) TBG, where Ti is determined by the strength of the impurity scattering in the system. For a hypo- thetical absolutely clean Fermi liquid Ti = 0 and the WF law is violated infinitely strongly at low temperatures (T (cid:28) TBG) as L(T ) vanishes. At "high" temperatures (T (cid:29) TBG), how- ever, the WF law is strictly obeyed since phonon scattering becomes quasi-elastic in this equipartition regime (where the electrical resistivity of the metal should manifest the well- known linear-in-T metallic behavior due to electron-phonon scattering) even if impurity scattering is weak (i.e. even for a very clean metal) Thus NFL is by no means necessary for vio- lating the WF law; the violation is automatic in a standard FL at low temperatures provided the system is clean. Note that in relatively dirty impure metals, we may have Ti > TBG, leading to the WF law being obeyed at all temperatures. III. DISCUSSION AND CONCLUSION We have revisited the old topic of the Wiedemann-Franz law in 2D and 3D electron liquids interacting with quenched impurities and acoustic phonons, providing detailed results for the temperature dependent effective Lorenz number (defined as the ratio of κ/σ T ) from T = 0 to T = TF . We neglect ef- fects of electron-electron interaction, and use the Boltzmann FIG. 3. (a) Lorenz ratio versus temperature for 2D electron gas in presence of impurity and phonons for three different impurity densities. (b) The same results for 3D system. For both systems TF = 103 TBG. transport theory for obtaining the results. Our main quali- tative finding is that the WF law is strongly violated at low temperatures (T < TBG) in clean Fermi liquids coupled to phonons. While most of our presented theoretical results are new, the main conclusion is neither surprising nor unknown, but seems to have been forgotten or overlooked in the cur- rently active research on non-Fermi-liquid physics where one often associates the failure of the WF law as synonymous with the breakdown of the Fermi liquid paradigm. Of course, in a narrow technical sense a coupled electron-phonon system is not a precise Fermi liquid23,24 because the interacting sys- tem has additional structures associated with phonon coupling with no analogs in the corresponding noninteracting Fermi gas, so perhaps the statement that the failure of the WF law may imply an NFL behavior is strictly speaking applicable to our system. But the WF law is in fact restored in the coupled electron-phonon system, as our results clearly show (and as is well-known), at higher temperatures (T > TBG), and indeed normal metals all obey the WF law rather accurately at room temperatures in spite of being a coupled electron-phonon sys- 10−2100102104106T/TBG0.00.20.40.60.81.0L/L02Du0=0.1u0=1u0=1010−2100102104106T/TBG0.00.20.40.60.81.0L/L03Du0=0.1u0=1u0=10TFTF(b)(a)tem. In any case, electron-phonon coupling is generic in all electronic materials, and branding such a common system to be an NFL simply because it strongly violates the WF law at low temperatures is not a meaningful advance. We show that in the presence of both impurity and phonon scattering, both 2D and 3D metals have four distinct tem- perature regimes, in principle, with respect to the WF law behavior: At very low temperatures, where impurity scat- tering dominates over phonon scattering (with the electrical resistivity not manifesting any temperature dependence) the WF law is obeyed; at low to intermediate temperatures (but T < TBG), where phonon scattering is stronger than impu- rity scattering (e.g. in clean systems) and the phonon-induced electrical resistivity shows the strong Bloch-Gruneisen tem- perature dependence, the WF law is strongly violated due to the inelastic nature of phonon scattering; at intermediate to high temperatures, where phonons are in the equipartition regime with phonon scattering being quasielastic in nature with the electrical resistivity reflecting a linear-in-T resistivity (as normal metals always do at room temperatures), the WF law is obeyed; and finally, at very high temperatures, where T approaches TF , the system becomes nondegenerate and the WF law is violated weakly with the effective Lorenz num- ber being somewhat smaller than the ideal Lorenz number. The existence of these four distinct regimes is generic both in 2D and 3D, but it is quite possible that a real material may not manifest all of these distinct regimes, depending on the parameter values controlling the various scattering strengths. For example, a normal 3D metal with TF ∼ 50, 000K obvi- ously never manifests the nondegeneracy effect of L(T ) < L0 at high temperatures, but 2D and 3D doped semiconductors, with TF ∼ 100K or less, should have a room temperature Lorenz number typically smaller than the ideal Lorenz num- ber by virtue of the Fermi surface nondegeneracy effect. If the impurity scattering is strong (i.e. relatively dirty systems), then it is possible that the WF law is obeyed at all temperatures with the impurity scattering dominating at low to intermedi- ate temperatures (up to TBG or above) and then quasi-elastic phonon scattering taking over at intermediate to high temper- atures (T > TBG). This appears to be the situation in most normal metals where any violation of the WF law is uncom- mon at any temperature and requires very clean samples. In fact, this accidental universal applicability of the WF law in normal 3D metals, by virtue of the overlapping elastic phonon and impurity scattering at intermediate temperatures, is what may have led to the misleading characterization of the validity or not of the WF law as implying the validity or not of the FL theory. In fact, an arbitrarily clean FL metal would violate the WF law at arbitrarily low temperature with L(T )/L0 ∼ T 2 for T (cid:28) TBG, directly reflecting the inelastic nature of low temperature phonon scattering (and the absence of elastic im- purity scattering by virtue of purity). Our results clearly bring this physics out both for 2D and 3D metals. Our work shows that it is, in principle, possible to use the validity or not of the WF law in order to check the appli- cability of the FL paradigm through careful measurements with some caveats (and some assumptions about the appli- cable materials parameter values for the system under con- 9 sideration). For example, the quasielastic acoustic phonon scattering for T > TBG invariably produces a temperature dependent electrical resistivity going as linear in T . In a FL, however, this linear-in-T resistivity regime should manifestly obey the WF law as our work shows, provided that T (cid:28) TF constraint is also satisfied. So, if a metallic system clearly manifesting a linear-in-T electrical resistivity over a temper- ature regime also violates the WF law at the same time, this would be a strong indicator of a possible NFL behavior. Sim- ilarly, impurity scattering typically leads to T -independent electrical resistivity (again assuming TF (cid:29) T ), and there- fore, an observed violation of the WF law concomitant with a T -independent resistivity (or a linear-in-T resistivity) would be an indicator of a possible NFL behavior. It may be worth- while to mention in this context that the cuprate high-Tc su- perconductors often exhibit a linear-in-T resistivity in the nor- mal phase (although the origin of this linear-in-T resistivity is not agreed upon and is considered by most to be caused by a non-phononic mechanism in contrast to a similar linear-in-T resistivity in normal metals at room temperatures). The WF law seems to be well-obeyed experimentally by the cuprate systems in the normal phase, indicating that a dominant part of its normal state transport is likely to be of a FL nature, but our lack of understanding of the underlying transport mech- anism makes a definitive conclusion difficult. It is, however, interesting to note that the cuprates are often referred to as "strange" or "bad" metals, but the fact that such strange met- als seem to obey the WF law is itself rather strange. One pos- sibility is that the main transport scattering mechanism in the cuprate normal phase arises from spin fluctuations associated with the nearby antiferromagnetic Mott phase, which could provide a simple explanation for the validity of the WF law (as well as the linear-in-T resistivity) assuming that the cor- responding Bloch-Gruneisen temperature for the bosonic spin fluctuations is low so that the scattering is primarily quasielas- tic in nature. Of course, it is also possible that the linear-in-T resistivity does indeed arise from phonon scattering with a low TBG, in which case the WF law emerges naturally. Obviously, much more work is necessary before a definitive conclusion is possible, and our comments here should be construed only as speculative ideas. There have been experimental studies of the violation of the WF law in the context of the breakdown of the quasipar- ticle picture and the FL description. In most of these studies the WF failure seems to occur near a quantum critical point (e.g. magnetic criticality10, metal-insulator transition11, Dirac point25) where the quasiparticle picture may indeed be ques- tionable, but it is also possible that this failure is an inherent effect of electron-electron interactions (neglected in our work) within the Fermi liquid description. A complete theory of the Wiedemann-Franz behavior leading to a quantitative calcu- lation of L(T ) including electron-impurity, electron-phonon, and electron-electron interactions for arbitrary system param- eters is a challenging task which has not been undertaken yet even for a model Fermi liquid, let alone for systems with com- plicated quantum phase transitions. Recent work has con- sidered the status of the Wiedemann-Franz law in the pres- ence of electron-electron and electron-impurity interactions (but without any phonons) in continuum Fermi liquids us- ing the hydrodynamic approximation within the Boltzmann theory26. The key finding is that the ideal WF law is indeed violated at some intermediate temperature range, but the WF law is recovered at low enough temperatures with L(T ) going as L/L0 = Γ γ+Γ, where Γ (γ) is respectively the electron- impurity (electron-electron) interaction strength. Thus, the electron-electron interaction effects vanish in the clean limit (Γ = 0) as it must in the absence of Umklapp and Baber scattering since the electron-electron interaction is manifestly momentum conserving by itself. This is of course very differ- ent from the effect of electron-phonon scattering, where the Lorenz number vanishes at low temperatures in the absence of electron-impurity scattering with L/L0 ∼ T 2 for T (cid:28) TBG in a perfectly clean metal. Using the fact that in a Fermi liq- uid, γ ∼ (T /TF )2 for T (cid:28) TF , we conclude, the violation of the WF law due to the electron-electron interaction is a higher- order effect, going as L/L0 ∼ (1 − O(T 2)) in a dirty system for T (cid:28) TF , whereas the corresponding electron-phonon in- teraction induced violation of the WF law is a leading-order effect in a clean system, going as L/L0 ∼ T 2 for T (cid:28) TBG. This difference arises because the electron-phonon interaction breaks momentum conservation and leads to resistive scat- tering even without any disorder whereas electron-electron interaction necessitates the presence of disorder (within the hydrodynamic theory) for breaking the momentum conserva- tion. (Inclusion of Umklapp scattering in a lattice changes the picture somewhat, but not qualitatively, and is not considered here.) Thus, in principle it should be possible using detailed low-temperature (T (cid:28) TBG) measurements of L(T ) in sam- ples with controlled disorder to distinguish between effects of electron-phonon and electron-electron interactions, but it is likely to be an extremely challenging task. Before concluding, we note that inelastic scattering pro- cesses considered in the current work always suppress the ef- fective Lorenz number L(T ) below the ideal Lorenz number L0, i.e. L(T ) < L0 in all our results, a point also empha- sized in Ref. 9. This implies that inelastic scattering generi- cally enhances the thermal resistivity compared with the elec- trical resistivity, arising simply from the fact that the elec- trical resistivity is dominated by large-angle backscattering ("2kF -scattering") across the Fermi surface relaxing the max- imum possible momentum whereas the thermal resistivity is affected equally by large-angle and small-angle inelastic scat- tering processes. At low temperatures, when kBT is much smaller than the typical phonon energy, large-angle scattering is strongly suppressed compared with the small-angle scatter- ing, thus enhancing thermal resistivity relative to the electrical resistivity, thus suppressing L(T ) well below L0. This sup- pression of L(T ) well below L0 thus is generic in the pres- ence of strong inelastic scattering independent of the FL or NFL nature of the underlying system. It is, however, pos- sible for L(T ) to exceed L0 in special situations. Experi- mentally, this can happen (and often does) when lattice ther- mal conductivity cannot be separated out from the electronic contribution. Since the lattice (i.e. phonons themselves) can carry heat rather efficiently, but does not carry any charge, any lattice contribution would enhance the thermal conductivity, 10 making the apparent L(T ) exceed L0. Ensuring that the mea- sured thermal conductivity is all electronic without any lattice contribution whatsoever is a difficult experimental challenge. Thus, if phonons themselves are conducting heat, the WF law can be violated with the apparent L(T ) > L0. In a simi- lar vein, it is possible for the electrons to lose energy directly to the lattice via electron-phonon interaction through the hot electron energy relaxation process. Such a direct energy loss from the electrons to the phonons is not a transport or con- duction phenomenon, but experimentally this may appear as an enhanced thermal conductivity with L(T ) > L0 and an apparent violation of the WF law. This process could in fact enhance L(T ) arbitrarily above L0 unless one is careful. In the presence of bipolar diffusion (i.e. when both electrons and holes are present in the system in equal numbers), again the thermal conductivity would surpass the WF constraint mak- ing L(T ) > L0. In fact, if the electrons and holes are strongly interacting, the L/L0 ratio could be very large as found re- cently in graphene25. There could be other processes, not con- sidered by us, which could also enhance L/L0 above unity in violation of the WF law. Our work has focused entirely on the issue of electron-phonon scattering leading to a para- metric violation of the WF law at low temperatures in clean systems, where the inelasticity of the scattering process sup- presses electrical conductivity much more strongly than the thermal conductivity making L/L0 ∼ T 2 at low temperatures in the absence of impurity scattering. Finally, we mention several other complications which are likely to cause problems in the study of the WF law in real materials. In particular, as mentioned above, phonons them- selves carry heat (but not electricity) and hence all measure- ment of L(T ) must necessarily ensure that any lattice ther- mal conductivity contributions are either absent or carefully subtracted out. This is not an easy task in general. Second, phonon drag, whence the carriers carry the lattice phonons with them, could be important complicating the extraction of an electronic thermal conductivity. Similarly, the electrons may not be in equilibrium with the lattice (the so-called "hot electron effect" mentioned above where the electrons and the phonons are at different temperatures), and in such a situation, the direct energy loss of the electrons to the lattice (through phonon emission for example) may manifest itself as a heat loss from the electrons, but this energy loss (the so-called hot electron energy relaxation) is completely distinct from the heat diffusion process associated with the electronic thermal conductivity. It is not always easy to separate hot electron en- ergy loss from electrnic thermal conduction, which may again produce erroneous experimental values of L(T ). Thus, there could be many reasons, some fundamental (e.g. inelastic scat- tering, nondegeneracy) and some practical (e.g. hot electron energy loss, lattice thermal conductivity), contributing to a breakdown of the WF law, and therefore, it is unwise to au- tomatically accept a breakdown of the WF law (i.e. finding that L(T ) differs from L0) as an indicator of an underlying NFL description. One must carefully consider all the carrier scattering processes contributing to κ and σ in quantitative depth to see if a FL description with quantitative corrections arising from the details of the scattering processes themselves are leading to the deviation of L(T ) from L0. This is the key message of our work. actions. Note added: A recent work by Hwang and DasSarma27, shows that the linear-in-Tresistivity, often as-sociated with the failure of Fermi liquid paradigm, is alsoconsistent with electron-phonon interactions just as we findthat the break- down of the WF law may arise from electron-phonon inter- This work is supported by Laboratory for Physical Sci- ences. A.L. was supported by JQI-PFC-UMD. The authors thank Maissam Barkeshli for several discussions before and during the course of this work. ACKNOWLEDGMENT 11 1 R. Franz and G. Wiedemann, Ann. Phys. 165, 497 (1853). 2 L. Lorenz, Ann. Phys. 249, 422 (1881). 3 G. Kumar, G. Prasad, and R. Pohl, J. Mater. Sci. 28, 4261 (1993). 4 P. Drude, Ann. Phys. 1, 566 (1900). 5 P. Drude, Ann. Phys.(Berlin) 7, 687 (1902). 6 A. Sommerfeld, Zeitschrift fur Physik 47, 1 (1928). 7 A. H. Wilson, The theory of metals (Cambridge Univ. Press, 8 A. Lucas and S. D. Sarma, Phys. Rev. B 97, 245128 (2018). 9 R. Mahajan, M. Barkeshli, and S. A. Hartnoll, Phys. Rev. B 88, 1954). 125107 (2013). 316, 1320 (2007). 10 M. A. Tanatar, J. Paglione, C. Petrovic, and L. Taillefer, Science Physical Kinetics (Pergamon Press, 1981). 20 F. Bloch, Zeitschrift fur Physik 59, 208 (1930). 21 E. Grneisen, Ann. Phys. 408, 530 (1933). 22 W. Kroll, Zeitschrift fur Physik 81, 425 (1933). 23 S. Engelsberg and J. Schrieffer, Phys. Rev. 131, 993 (1963). 24 R. E. Prange and L. P. Kadanoff, Phys. Rev. 134, A566 (1964). 25 J. Crossno, J. K. Shi, K. Wang, X. Liu, A. Harzheim, A. Lucas, S. Sachdev, P. Kim, T. Taniguchi, K. Watanabe, et al., Science 351, 1058 (2016). 26 A. Lucas and S. D. Sarma, Phys. Rev. B 97, 115449 (2018). 27 E. H. Hwang and S. Das Sarma, Phys. Rev. B 99, 085105 (2019). 28 E. Lifshitz and L. P. Pitaevskii, Course of Theoretical Physics: 29 T. Ando, A. B. Fowler, and F. Stern, Rev. Mod. Phys. 54, 437 (1982). 30 L. Pourovskii, J. Mravlje, A. Georges, S. Simak, and I. Abrikosov, New Journal of Physics 19, 073022 (2017). 31 C. Herring, Physical Review Letters 19, 167 (1967). 32 L. Onsager, Phys. Rev. 37, 405 (1931). Appendix A: Boltzmann Equation Formalism In this appendix, we show the explicit form of the linearized Boltzmann equation, including the collision integral. In all of the appendices we have set = kB = 1. In order to use the linearized Boltzmann equation Eq. (7), the collision integral Eq. (6) must also be linearized. Using the detailed balance relation S(k, k(cid:48))f0(k)(1 − f0(k(cid:48))) = S(k(cid:48), k)(1 − f0(k))f0(k(cid:48)) (A1) the linearized collision integral is 11 S. Lee, K. Hippalgaonkar, F. Yang, J. Hong, C. Ko, J. Suh, K. Liu, K. Wang, J. J. Urban, X. Zhang, et al., Science 355, 371 (2017). 12 J. Dong, Y. Tokiwa, S. Budko, P. Canfield, and P. Gegenwart, Phys. Rev. Lett. 110, 176402 (2013). 13 N. Wakeham, A. F. Bangura, X. Xu, J.-F. Mercure, M. Greenblatt, and N. E. Hussey, Nat. Commun. 2, 396 (2011). 14 N. W. Ashcroft and N. D. Mermin, Solid state physics (holt, rine- hart and winston, new york, 1976). 15 D. Xiao, M.-C. Chang, and Q. Niu, Rev. Mod. Phys. 82, 1959 16 J. M. Ziman, Principles of the Theory of Solids (Cambridge uni- 17 G. D. Mahan, Many-particle physics (Springer Science & Busi- 18 S. D. Sarma and F. Stern, Phys. Rev. B 32, 8442 (1985). 19 N. F. Mott and H. Jones, The theory of the properties of metals and alloys (Courier Corporation, 1958). (2010). versity press, 1979). ness Media, 2013). (cid:90) ddk(cid:48) (2π)d S(k, k(cid:48)) (cid:20) Ik = − (cid:21) δf (k) 1 − f0(k(cid:48)) 1 − f0(k) − δf (k(cid:48)) f0(k) f0(k(cid:48)) (A2) For simplicity, we assume throught this paper an isotropic quadratic band of effective mass m. Substituting in the ansatz Eq. (8), the linearization allows the integral equation to be broken into separate equations for the thermal and electrical lifetimes: (cid:90) ddk(cid:48) (cid:90) ddk(cid:48) (2π)d S(k, k(cid:48)) (2π)d S(k, k(cid:48)) (cid:20) (cid:20) τσ(ε) − τκ(ε) − 1 − f0(ε(cid:48)) 1 − f0(ε) 1 − f0(ε(cid:48)) 1 − f0(ε) 1 = 1 = k(cid:48) cos α(cid:48) k cos α ε(cid:48) − µ ε − µ (cid:21) τσ(ε(cid:48)) k(cid:48) cos α(cid:48) k cos α τκ(ε(cid:48)) (cid:21) 12 (A3) (A4) (cid:40) where α (resp. α(cid:48)) is the angle between k (resp. k(cid:48)) and the applied field. With a bit of algebra, it is straightforward to show that cos α(cid:48) cos α = cos θ + tan α sin θ cos φ d = 3 cos θ − tan α sin θ d = 2 (A5) where θ is the angle between k and k(cid:48) and φ is the polar angle for k(cid:48) about k. In all cases we consider in this paper, S de- pends only on k, k(cid:48), and cos θ, where θ is the angle between k and k(cid:48). Therefore, in d = 3, the integral over φ of the cos φ term will be zero. In d = 2, since S depends on θ only as cos θ, by orthogonality the sin θ term will integrate to zero. The upshot is that we may replace cos α(cid:48)/ cos α with cos θ in both d = 2 and d = 3. Once these equations have been solved for τ, the transport coefficients may be obtained using Eq. (10). Appendix B: Impurity Scattering In this appendix we calculate the electrical and thermal con- ductivities in Boltzmann theory for impurity scattering, in- cluding a model where the impurity scattering is purely elastic but also primarily forward. We show that the Wiedemann- Franz law still holds when T (cid:28) TF . Impurity scattering is elastic, so S(k, k(cid:48)) ∝ δ(ε − ε(cid:48)). The lifetime equations Eq. (A3) and (A4) simplify dramat- ically and actually become the same equation, which is easy to solve: 1 1 = τσ(ε) (2π)d S(k, k(cid:48)) (1 − cos θ) We now consider specific impurity scattering models. = − τκ(ε) (B1) (cid:90) ddk(cid:48) 1. Conventional Impurity Scattering The textbook impurity model is isotropic and short-range with the scattering rate Eq. (12). The Boltzmann equation Eq. (B1) amounts to a simple integral in this case and we find (cid:40) 1 τ (ε) = nimpmu2 0 nimpu2 π 0√2m3ε d = 2 d = 3 (B2) Substitution into Eq. (10) leads to the conductivities (per spin) (cid:16) 1 + eµ/T(cid:17) (cid:16) 3 log(cid:0)1 + eµ/T(cid:1) − log (cid:20) 1 + eµ/T(cid:17) (cid:16) µ2(cid:16) (cid:17) log π2 + − 4 1 + eµ/T(cid:17) (cid:18) + 2T µ σ = e2AT log S = µ T LT T = A 2Li2((1 + eµ/T )−1) log(cid:0)1 + eµ/T(cid:1) − log2(1 + eµ/T ) + Li2((1 + eµ/T )−1) − (cid:19) 2π2 3 − 6T 2Li3(−eµ/T ) (cid:21) (B3) where the thermopower S = LET /σ = −LT E/(T σ) one must remember that the chemical potential µ is a function of T and (cid:40) A−1 = 2πnimpmu2 0 3πnimpmu2 0 d = 2 d = 3 (B4) Note that nimp and u0 have different units in d = 2 and d = 3, and that the functional dependence µ(T ) is different in d = 2 and d = 3. In the regime T (cid:28) TF , µ ≈ TF and we expand the poly- logarithm Lis(−ex) at large values of x using the series rep- resentation ∞(cid:88) Lis(−ex) = (−1)k(1−21−2k)(2π)2k B2k k=0 Γ(s + 1 − 2k) (B5) where the B2k are the Bernoulli numbers. This series is essen- tially the Sommerfeld expansion. The resulting conductivities (2k)! xs−2k are (cid:40) (cid:40) σ(T (cid:28) TF ) = κ(T (cid:28) TF ) = e2 4πnimpu2 e2TF 0m2 3πmnimpu2 0 πT 12nimpu2 0m2 πTF 9mnimpu2 0 d = 2 d = 3 d = 2 d = 3 (B6) (B7) and can be neglected). At T (cid:29) TF , the temperature dependence of the equilib- rium chemical potential µ must be accounted for. A textbook calculation yields (cid:40)µ = T log(cid:0)eEF /T − 1(cid:1) ≈ T log(cid:0) EF (cid:1)3/2(cid:17) (cid:0) EF (cid:16) 4 T (cid:1) d = 2 d = 3 µ ≈ T log 3√π T and the Wiedemann-Franz law is obeyed (S is of order T /TF Plugging this in and expanding Eq. (B3) leads to σ ≈ e2A S ≈ F d = 2 4E3/2 3√π T −1/2 (cid:40)EF (cid:40)2 − log(cid:0) EF (cid:1) (cid:1)3/2(cid:17) (cid:16) 4 (cid:0) EF (cid:40)T EF (cid:0)6 − 4 log(cid:0) EF (cid:1) + log2(cid:0) EF (cid:1)(cid:1) (cid:1)3/2(cid:17) (cid:16) 4 (cid:16) (cid:0) EF 2 − log T 3√π d = 3 d = 3 d = 2 T T T 4 3√π E3/2 F √T 3√π T 6 − 4 log LT T ≈ A + log2(cid:16) 4 3√π (cid:0) EF T (cid:1)3/2(cid:17)(cid:17) d = 2 d = 3 13 (B8) (B9) (B10) (B11) The Lorenz number can then be computed straightforwardly for T (cid:29) TF ; in both d = 2 and d = 3, 6 π2 L0 L = 2 = (B12) 2. Forward Scattering We now demonstrate that even when the dominant scatter- ing mechanism is elastic forward scattering (i.e. not isotropic as in Appendix B as above), the Wiedemann-Franz law is still obeyed at low temperature. Thus, pure elastic scatter- ing always leads to the WF law independent of the isotropic or strongly anisotropic nature of the scattering. This result is a special case of what is known on very general grounds from the Sommerfeld expansion28, but we still find these cal- culations enlightening; we can show explicitly that even when there is a parameter which we can tune to be in the forward scattering limit, WF is unaffected. We will use the scattering rate (per unit of momentum space) (cid:90) pressed. Taking kF z0 (cid:29) 1 corresponds to the extreme for- ward scattering limit18. Substitution into Eq. (B1) yields 1 τ (ε) = U 2 0 nimp (2π)d ddq e−2qz0 (q + qs)2 δ (ε(k) − ε(k + q)) (1−cos θ) (B14) where θ is the angle between k and k(cid:48). We have used k = k(cid:48) for elastic collisions on a circular Fermi surface to rewrite things in terms of θ. It is most convenient to use some geometry to find that 1 − cos θ = −(q/k) cos β where β is the angle between k and q. Likewise, ε(k) − ε(k + q) = − 2kq cos β + q2 2m (B15) In two dimensions, substituting and changing variables to u = cos β we obtain 1 τ (ε) U 2 0 nimp 2π2 = − (cid:90) 1 (cid:90) ∞ (cid:90) 2kz0 dq 0 −1 dx 0 (cid:16) (cid:17) mq k2 e−2qz0 (q + qs)2 δ u + q 2k u √1 − u2 x2 du (cid:112) 1 − (x/2kz0)2 (B16) e−2x (x + qsz0)2 (B17) S(k, k(cid:48)) = U 2 0 nimp e−2qz0 (q + qs)2 δ(ε(k) − ε(k(cid:48))) (B13) = − U 2 0 mnimp 4π2z0k3 where q = k − k(cid:48). Physically, this is the scattering rate obtained from Fermi's Golden Rule for charged impurities placed a distance z0 from a 2D electron gas29, with nimp the impurity concentration, qs a screening wavevector, and U0 a prefactor characterizing the strength of scattering with dimen- sions of energy times length. The precise form is unimpor- tant - what matters is that the scattering is elastic and that scattering wavevectors larger than 1/z0 are exponentially sup- where we have made the change of variables x = z0q. Since we are interested in computing the conductivities at T (cid:28) TF , we may take k ∼ kF . In the forward scattering limit kF z0 (cid:29) 1, a straightforward series expansion about x = 2kz0 shows that the contribution to the integral of the region with x ∼ 2kz0 (cid:29) 1 is exponen- tially suppressed in kF z0. Therefore, the integral is dominated by the regime x (cid:28) 2kz0. In said regime, the square root fac- tor is, to leading order, 1, so, it is safe to neglect the square 14 (cid:90) ∞ 0 root and to extend the upper limit of integration to +∞: 1 e−2x dxx2 (x + qsz0)2 ≡ A(qsz0)ε3/2 1 τ (ε) ≈ U 2 0 mnimp 4π2z0k3 (B18) The precise form of A is unimportant for the WF law since it is independent of ε and T . Using Eq. (10), the transport coefficients can be computed explicitly in terms of polylogarithms: T 5/2Li5/2(−eTF /T ) σ = − κ = − 15Ae2 16√π 15A 64√πT T 5/2(cid:16) (cid:17) F Li5/2(−eTF /T ) − 28TF T Li7/2(−eTF /T ) + 63T 2Li9/2(−eTF /T ) 4T 2 (B19) (B20) As we will show in the next subsection, T σS2 ∼ (T /TF )2LT T , so we have taken κ ≈ LT T . Using the expansion Eq. (B5), where we have left implicit the fact that B is a complicated function of qsz0. We can now expand σ = κ = T 5/2 F Ae2 2π π2 3 A 2π T 5/2 F T (B21) (B22) where we used the k = 0 term for σ and the k = 1 term for κ (the k = 0 term for κ is zero, as expected). The Wiedemann- Franz law is obeyed. 3. Forward Scattering: Corrections to WF We now want to estimate the leading corrections to the WF law at T (cid:28) TF in the elastic forward scattering model used in App. B 2. These will be of order (T /TF )2, arising from doing the next order of the Sommerfeld expansion. Said (T /TF )2 term will have an order-1 coefficient, but we would also like to obtain the corrections to that coefficient to leading order in 1/kF z0. To do so, we need to start by calculating the leading-order corrections to τ as a function of 1/kF z0. As discussed previ- ously, power-law corrections appear only at small x/2kz0 and arise from the lowest-order correction when the square root is expanded. The error is approximately (cid:90) (cid:18) x (cid:19)2 U 2 0 mnimp 4π2z0k3 ≈ − dxx2 1 2 0 τ (x + qsz0)2 (B23) where 1 (cid:28) (cid:28) 2kz0 is some cutoff where the expansion of the square root is valid. For the same reasons as before we may take the upper limit to infinity and we obtain 2kz0 e−2x (cid:18) 1 (cid:19) δ (cid:18) 1 (cid:19) δ τ U 2 0 mnimp 8π2z0k3 ≈ − 1 Bε5/2z2 0 ≡ (cid:90) ∞ 1 (2kz0)2 0 dxx4 e−2x (x + qsz0)2 (B24) 1 τ = 1/Aε3/2 + 1/Bε5/2z2 0 ≈ Aε3/2 − where the expansion is controlled by 1/εz2 A2ε1/2 Bz2 0 (B25) 0 ∼ 1/k2 0. F z2 This expression can be plugged straightforwardly into Eqs. (10), and we wish to take the next highest order in the Som- merfeld expansion Eq. (B5). We define α3/2 = A and α1/2 = A2/Bz2 0. After expand- ing the polylogarithms to the appropriate order, we find e2 2π σ ≈ n=1/2,3/2 LET ≈ e 2π π2 3 LT T ≈ T 1 2π π2 3 (cid:33) n(n + 1) (B26) (B27) (cid:32) (cid:18) T (cid:19)2 TF 1 + π2 6 (cid:88) (cid:18) T TF αnT n+1 F (cid:19) (cid:88) (cid:88) n=1/2,3/2 (cid:34) αnT n+1 F × (cid:18) T (cid:19)2 TF n=1/2,3/2 7π2 60 1 − αnT n+1 F n2(n + 1)(n + 5) (cid:35) (B28) (B29) (B30) After a considerable amount of algebra and Taylor expansion, we find (cid:18) T (cid:19)2 π2 (cid:18) 1339 TF 24 L L0 ≈ 1 − (cid:19) + 743 5 A Bz2 0TF From the definitions, A Bz2 0TF = 2(kF z0)2 I2(qsz0) I4(qsz0) 8 1 with (cid:90) ∞ 0 In(qsz0) = xn e−2x (x + qsz0)2 (B31) We note that the correction to the WF law arising in Eq. (B29) from the forward scattering physics is of O(T /TF )2, which is the same order where electron-electron scattering also shows up as a correction8 of the WF law, thus considerably compli- cating interpretation of experiments. Appendix C: Electron-Phonon Transport Calculations In this appendix, we discuss our Boltzmann theory electron-phonon calculations in detail. Throughout we as- sume a quadratic band of effective mass m and the scattering rate (C1) obtained by Fermi's Golden Rule for electrons of momentum k scattering off of acoustic phonons. Here q is the momentum transfer, equal to k(cid:48) − k in the first term and equal to k − k(cid:48) in the second term. Also, Nq is the Bose distribution, D is the deformation potential, ωq = csq, cs is the speed of sound in the material, ωD is the Debye frequency, and Θ is the Heav- iside step function. We assume throughout that the system is sufficiently clean so that electron-impurity scattering can be neglected at the temperatures in question. We also neglect ef- fects such as phonon drag. (Note also that for the results in our main text we assume TBG < TD throughout so that the effective phonon frequency cut off is TBG for our analysis.) S(k, k(cid:48)) = πD2q2 ρωq [Nqδ( − (cid:48) + ωq) + (Nq + 1)δ( − (cid:48) − ωq)] Θ(ωD−ωq) = 1 1. Relaxation Time Approximations 15 In principle, the integral equation Eq. (A3) can be solved. As we have seen from Appendix B, this is straightforward when the scattering is purely elastic. However, electron- phonon scattering is inelastic, so the Boltzmann equation remains a complicated integral equation for τ. To make progress, we need to perform an uncontrolled approxima- tion on Eqs. In particular, we will replace τσ,κ(ε(cid:48)) → τσ,κ(ε). Although the terminology is used in am- biguous or inconsistent ways in the literature, this is our form of the "relaxation time approximation." (A3) and (A4). With this approximation the Boltzmann equation becomes (cid:90) ddk(cid:48) (cid:90) ddk(cid:48) (2π)d S(k, k(cid:48)) (2π)d S(k, k(cid:48)) 1 τσ(ε) = τκ(ε) (cid:21) (cid:20) (cid:20) 1 − f0(ε(cid:48)) 1 − f0(ε) 1 − f0(ε(cid:48)) 1 − f0(ε) 1 − 1 − cos θ k(cid:48) k ε(cid:48) − µ ε − µ k(cid:48) k (cid:21) cos θ (C2) Plugging in the form of S(k, k(cid:48)) and using the expressions − (cid:48) = − cos θ = ∓ k(cid:48) k 1 − 2m q2 ± 2kq cos β q k cos β (C3) (C4) (cid:90) ddq (2π)d 1 τσ(ε) = πD2m ρcsk (cid:20) 1 − f0(ε + csq) (cid:16) q 1 − f0(ε) (Nq + 1) 1 − f0(ε − csq) + 1 − f0(ε) where the sign corresponds with k(cid:48) = k ± q (depending on whether a phonon is being absorbed or emitted), we find (cid:16) (cid:19) (cid:18) (cid:17) Nq cos β δ − q k (cid:17) (cid:18) cos β − −q + 2mcs 2k + (cid:19)(cid:21) cos β k δ cos β − q + 2mcs 2k Θ(qD − q) (C5) where qD is the Debye wavevector ωD/cs. Similar substitu- tions can be made for the thermal lifetime. 2. 3D Calculations We discuss these approximations further in Appendix E We start with the electrical conductivity. The angular in- tegrals in Eq. (C5) are done mostly straightforwardly, with one important caveat. Since cos β is only integrated over the range (−1, 1), the delta functions only lead to nonzero con- tributions for certain values of q; this restriction is where the Bloch-Gruneisen temperature plays a key role. We find (cid:34)(cid:90) q(1) max q(1) min 1 τ (ε) = D2m 4πρcsk (cid:18) mcsq k2 + − q2 2k2 (cid:19) (cid:90) q(2) max + 0 dqq2 1 − f0(ε + csq) 1 − f0(ε) Nq dqq2 1 − f0(ε − csq) 1 − f0(ε) (Nq + 1) (cid:18) mcsq k2 + (cid:19)(cid:35) q2 2k2 (C6) Defining η = (ε − µ)/T and z = csq/T , 16 Under the assumption TD (cid:29) TBG (and noting that in most systems TBG (cid:28) TF ), q(1) min = −2k + 2mcs q(1) max = 2k + 2mcs q(2) max = 2k − 2mcs (C7) (C8) (C9) (cid:34)(cid:90) z(1) max z(1) min 1 τσ(ε) = D2mT 3 4πρc4 sk dz 1 + eη 1 + eη+z z2 1 − e−z (cid:18) − mT z k2 + T 2z2 2k2c2 s (cid:19) (cid:90) z(2) max + 0 dz 1 + eη 1 + eη−z z2 ez − 1 (cid:18) mT z k2 + T 2z2 2k2c2 s (cid:19)(cid:35) (C10) with the definitions of zmax,min following from those of qmax,min. A very similar computation for the thermal transport life- time yields (cid:34)(cid:90) z(1) (cid:90) z(2) z(1) min max (cid:18) (cid:18) z2 1 − e−z z2 ez − 1 (cid:18) (cid:18) 1 τκ(ε) = D2mT 3 4πρc4 sk dz 1 + eη 1 + eη+z η + z η 1 − 1 + mT z k2 − T 2z2 2k2c2 s max + 0 dz 1 + eη 1 + eη−z η − z η 1 − mT z k2 − 1 − T 2z2 2k2c2 s (cid:19)(cid:19) (cid:19)(cid:19)(cid:35) + To make progress, we now need to look at asymptotic regimes. the lowest nontrivial order in z: (cid:90) zmax 0 dz z3T 2 k2c2 s (cid:18) 1 + mc2 s T 1 τσ(ε) ≈ D2mT 3 4πρc4 sk D2mT k πρc2 s ≈ (C11) (cid:19) tanh(η/2) (C12) (C13) a. Equipartition Regime in 3D This regime is the traditional T -linear resistivity regime: TBG (cid:28) T (cid:28) TF . When T (cid:28) TF , only ε ∼ TF is important so we may estimate k ∼ kF . Then to leading order z(1) max ≈ z(2) max ≈ 2kcs/T ∼ TBG/T (cid:28) 1, and z(1) min = 0. Since z ∈ (0, zmax) and zmax (cid:28) 1, we may expand Eq. (C10) to where we neglected the term of order mc2 s/T ∼ BG/T TF (cid:28) 1. T 2 The conductivity is computed straightforwardly in the lowest-order Sommerfeld expansion to obey a Drude formula σ = ne2τ (εF ) m = 8e2ρk2 F c2 s 6πD2m2T (C14) This is the familiar result that the electron-phonon scattering induced resistivity goes as T at high temperatures where the phonons are in the equipartition regime. Strictly speaking, this linear-in-T regime applies for T (cid:38) TBG/5 (or TD/5) depending on whether TBG < TD or not. In calculating the thermal lifetime, we can similarly equate the limits of the two integrals in Eq. (C11) and expand. We find (cid:90) zmax 0 dz z3T 2 k2c2 s (cid:18) 1 + 1 τκ(ε) ≈ D2mT 3 4πρc4 sk mc2 s T tanh(η/2) − 2c2 sm T η + c2 sk2 T 2η tanh(η/2) (cid:19) 17 (C15) In calculating the thermal conductivity, τκ is integrated against (ε − µ)2∂f0/∂ε, which is peaked at ε − µ ∼ T and equal to zero at ε = µ. Therefore, when calculating τκ, we can safely estimate ε− µ ∼ T , that is, η ∼ 1, when estimating which terms are important (as long as τκ does not diverge at ε → µ). For η ∼ 1 we can neglect all of the η-dependent terms, BG/T 2 (cid:28) 1. We find which are of order T 2 τκ ≈ τσ, in agreement with our results in the main text. The Sommerfeld expansion immediately leads to the Wiedemann- Franz law. Thus, a linear-in-T resistivity arising from phonon scattering is automatically associated with the validity of the BG/T TF (cid:28) 1 or T 2 WF law. b. Bloch-Gruneisen Regime in 3D This regime is T (cid:28) TBG (cid:28) TF . As before, z(1) max ≈ z(2) max ≈ 2kcs/T , but in this regime both of these limits are large. Since the integrand in Eq. (A3) is suppressed exponen- tially at large z, it is a good approximation to take zmax → ∞. After taking z → −z in the second term of Eq. (A3) we obtain (cid:90) ∞ (cid:20) 2mc2 −∞ dz s T 1 τσ(ε) = = D2mT 3 4πρc4 sk 3D2mT 5 4πρc6 sk3 z2 1 + eη 1 + eη+z (cid:0)Li4(−e−η) − Li4(−eη)(cid:1) + 4(cid:0)2ζ(5) − Li5(−e−η) − Li5(−eη)(cid:1)(cid:21) mT z k2 + 1 − e−z T 2z2 2k2c2 s − (C16) (C17) (cid:18) (cid:19) Again, the electrical conductivity is found at leading order in the Sommerfeld expansion σ = ne2τ (εF ) m = sk6 2ρc6 F 9πm2D2 1 T 5 = ρT 6 BG 288m2D2 1 T 5 (C18) which leads to the expected ρ ∼ T 5 behavior (often called the Bloch-Gruneisen behavior). The same approximations can be used in calculating the thermal lifetime 1 τκ(ε) = D2mT 3 4πρc4 sk (cid:90) ∞ −∞ dz 1 + eη 1 + eη+z z2 1 − e−z (cid:20) (cid:18) 1 − 1 + z η (cid:19)(cid:18) 1 + mT z k2 − T 2z2 2k2c2 s (cid:19)(cid:21) (C19) In the z (cid:46) 1 regime where the integrand is not exponen- tially suppressed, we can use mT /k2 ∼ T /TF (cid:28) 1 and BG (cid:28) 1 to simplify the integral dramati- T 2/k2c2 (cid:19) cally for η ∼ 1: s ∼ T 2/T 2 (cid:18) dz 1 + eη 1 + eη+z (cid:90) ∞ (cid:0)Li4(−e−η) − Li4(−eη)(cid:1) 1 − e−z −∞ z2 1 τκ(ε) ≈ D2mT 3 4πρc4 sk = 6D2mT 3 4πρc4 skη − z η (C20) (C21) In the Sommerfeld expansion, the leading-order contribu- tion is zero as expected. The next-leading-order contribution yields κ ≈ LT T ≈ 8π2ρc4 sT 2 F 54ζ(3)D2 1 T 2 (C22) The numerical prefactor should, of course, not be taken very seriously, but we obtain the 1/T 2 behavior as expected. The Wiedemann-Franz law is violated as L ∼ T 2 (C23) The scalings σ ∼ 1/T 5, κ ∼ 1/T 2, and L ∼ T 2 are in agreement with the calculations in the main text. Thus, in the Bloch-Gruneisen regime the WF law is violated strongly as long as impurity scattering contribution to resistivity is much smaller than the phonon scattering contribution -- in other words, any observation of a Bloch-Gruneisen transport behavior must automatically be associated with a strong vio- 18 lation of the WF law. 3. 2D Calculations The angular integral in Eq. (C5) is slightly more tedious in 2D. Changing variables to u = cos β introduces a factor of 2 and a Jacobian. In the same variables z = csq/T and η = (ε − µ)/T as for 3D, we obtain 1 τσ() = D2mT 2 πρc3 sk 1 τκ() = D2mT 2 πρc3 sk max max z(1) min (cid:90) z(1) (cid:90) z(2) (cid:90) z(1) (cid:90) z(2) z(1) min + 0 max max + 0 dz (cid:114) (cid:114) (cid:114) (cid:114) z z (cid:16) T (cid:16) T (cid:16) T (cid:16) T z z s T z − 2mc2 (cid:17)2(cid:16) (cid:17)2 (cid:17)2(cid:16) (cid:17)2 −z − 2mc2 (cid:17)2(cid:16) (cid:17)2 (cid:17)2(cid:16) (cid:17)2 z − 2mc2 T T s s −z − 2mc2 T s dz (1 − e−z) 1 − 2kcs (ez − 1) dz dz 1 − 2kcs (1 − e−z) 1 − 2kcs (ez − 1) 1 − 2kcs (C24) (cid:19)(cid:19) (cid:19)(cid:19) (C25) 1 + eη 1 + eη+z mT z k2 + T 2z2 2k2c2 s 1 + eη 1 + eη−z k2 + T 2z2 2k2c2 s (cid:19) (cid:19) − (cid:18) (cid:18) mT z (cid:18) (cid:18) 1 − (cid:18) (cid:18) 1 + eη 1 + eη+z η + z η 1 + mT z k2 − T 2z2 2k2c2 s 1 + eη 1 + eη−z η − z η 1 − mT z k2 − 1 − T 2z2 2k2c2 s with the limits of the integrals defined in the same way as in 3D. We must take limits carefully to proceed. the square root makes the integrals a bit more complicated. We expand only the exponentials in z to obtain a. Equipartition Regime in 2D As in 3D, this regime is TBG (cid:28) T (cid:28) TF , which has max ∼ z(2) z(1) min = 0. However, max ∼ TBG/T (cid:28) 1 and z(1) (cid:90) z(1) 0 1 τσ() ≈ D2mT 4 2πρc5 sk3 (cid:88) ± = D2mT 4 2πρc5 sk3 2D2mT c2 sρ ≈ (cid:18) (cid:17)2 z − (cid:19) 2mc2 s T (cid:18) z + (cid:17)2 −z − 2mc2 T s max dz (cid:114) (cid:90) z(2) (cid:90) 2kcs/T + 0 max 1 − dz dz ∓2mc2 s/T s T z 2kcs (cid:16) T (cid:114) (cid:16) (cid:114) (cid:17)2(cid:16) z − 2mc2 (cid:17)2(cid:16) (cid:16) T (cid:17) 1 − (cid:17)2 (cid:16) zT z ± 2mc2 1 − 2kcs 2kcs z z T s (cid:19) 2mc2 s T 19 (C26) (C27) (C28) s/T . Again we where we changed variables z → z ± 2mc2 can use the Sommerfeld expansion to lowest order to obtain a Drude-type formula σ = ne2τ (εF ) m = sk2 e2ρc2 F 8πD2m2 1 T (C29) with the electrical resistivity linear in temperature. Doing a similar expansion for τκ we obtain 1 τκ() ≈ D2mT 2 πρc3 sk max (cid:90) z(1) (cid:90) z(2) z(1) min + 0 dz (cid:114) (cid:114) 1 − 2kcs (cid:16) T (cid:16) T s T 1 (cid:17)2 (cid:17)2(cid:16) z − 2mc2 (cid:17)2(cid:16) −z − 2mc2 1 T s 1 − 2kcs max dz (cid:18) (cid:17)2 (cid:18) (cid:18) 1 + (cid:18) η + z η 1 − mT z k2 − T 2z2 2k2c2 s η − z η 1 − mT z k2 − 1 − T 2z2 2k2c2 s (cid:19)(cid:19) (cid:19)(cid:19) ≈ 2D2mT c2 sρ = 1 τσ(ε) (C30) where we have neglected terms of order T 2 BG/(T TF ) (cid:28) 1 (and again taken η ∼ 1 when estimating the size of terms). Since the thermal and electrical lifetimes are equal and energy-independent, in agreement with the results in the main text, the Sommerfeld expansion immediately tells us that the Wiedemann-Franz law is obeyed. b. Bloch-Gruneisen Regime in 2D As in 3D, this regime is T (cid:28) TBG (cid:28) TF . We again have z(1) min = 0 and z(1) max ∼ TBG/T , but now zmax (cid:29) 1. If, as in 3D, we wish to take zmax → ∞, we must deal max ≈ z(2) carefully with the square root factor. When z ∼ zmax (cid:29) 1, the whole integrand is exponentially suppressed, although there is a divergent prefactor scaling as (z − zmax)−2. It can be checked in a straightforward Taylor expansion that contribution of the large-z regime is finite and exponentially small in zmax. At z (cid:28) zmax, the term under the square root is of order 1 − (z/(TBG/T ) ± TBG/TF )2. The correction to 1 is small whenever z (cid:28) zmax since zmax ≤ TBG/T . Therefore, over the entire region of integration, the square root may simply be set to 1. Note that this argument holds for both the electrical and thermal relaxation times. It is then safe to take zmax → ∞. With a substitution z → −z in the second integral, Eq. (C24) becomes 1 τσ() = = D2mT 4 2πρc5 sk3 D2mT 4 2πρc5 sk3 z2 dz 1 + eη 1 + eη+z z − 1 − e−z T −∞ 15 − 6Li4(−eη) − 6Li4(−e−η) − 4mc2 s T (cid:18) (cid:19) 2mc2 s 20 (C31) (C32) (cid:0)Li3(−eη) − Li3(−e−η(cid:1)(cid:21) (cid:90) ∞ (cid:20) 2π4 The electrical conductivity is again found in the lowest-order term of the Sommerfeld expansion σ = ne2τ (εF ) m = sk5 2e2ρc5 F π4D2m2T 4 (C33) 1 τκ(ε) ≈ D2mT 2 πρc3 sk (cid:90) ∞ z −∞ 1 − e−z 1 + eη 1 + eη+z The thermal lifetime, under the same approximations, is leading to ρ ∼ T 4 as expected. (cid:18) (cid:18) η + z 1 − η 1 + mT z k2 − T 2z2 2k2c2 s (cid:19)(cid:19) (C34) In the regime z (cid:46) 1 and with η ∼ 1, the leading-order term in the parentheses is −z/η. Since the integrand is exponen- tially suppressed at z (cid:38) 1, (cid:19) (cid:18) (cid:90) ∞ (cid:0)Li3(−e−η) − Li3(−eη)(cid:1) 1 + eη 1 + eη+z 1 − e−z −∞ z z η − 1 τκ(ε) ≈ = D2mT 2 πρc3 sk 2D2mT 2 πρc3 skη The Sommerfeld expansion yields κ ≈ LT T ≈ sT 3/2 c3 F ρ √2D2√m 1 T (C35) (C36) (C37) which leads to a violation of the Wiedemann-Franz law L ∼ T 2 (C38) Again, the scalings σ ∼ 1/T 4, κ ∼ 1/T , and L ∼ T 2 all agree with the results in the main text. Thus, both in 2D and 3D FL systems, the WF law is obeyed (violated) in the linear-in-T high-temperature (Bloch-Gruneisen low tempera- ture) regime as long as impurity scattering remains weak. Appendix D: WF law violation with single relaxation time It was shown in the main text that thermal relaxation time can differ significantly from charge relaxation time due to dif- ferent mechanisms underlying each relaxation process which results in WF law violation. In this appendix we will show that even with a single relaxation time, WF law could still be violated at very low temperatures. Derivation of WF law at low temperatures in systems which are described by a single relaxation time, relies on the Som- merfeld expansion of the listed integrals in Eq.(10)14. Gener- ally, the Sommerfeld expansion can be used to evaluate low temperature limits of any integral which involves Fermi dis- tribution function: (cid:90) +∞ H(ε) e(ε−µ)/T + 1 −∞ (cid:90) µ (cid:18) T (cid:19)4 µ −∞ + O dε = H(ε)dε + π2 6 T 2 H(cid:48)(µ) (D1) (cid:16) T (cid:17) µ For the expansion to be controlled by , one needs to make sure that derivatives of H(ε) do not involve powers of T . However, non-trivial energy dependences in H(ε) could 1 result in such factors. For example, terms like e(ε−µ)/T in H(ε) could potentially makes keeping first few terms in Som- merfeld expansion incorrect. As we will show bellow, this could be the case whenever relaxation time involves exponen- tial factors related to statistical distributions. Consider the system studied in section II D and Appendix C, where only electron-phonon scattering is present. To arrive at the integral expressions in Eq. (C2) using the RTA, we replaced τσ,κ(ε(cid:48)) in Eqs. (A3) and (A4) by τσ,κ(ε). However, if we use a different type of RTA and replace (ε(cid:48)−µ)τκ(ε(cid:48)) → (ε − µ)τκ(ε) and τσ(ε(cid:48)) → τσ(ε), we will find that the two relaxation times become equal and both can be evaluated from the expression for τ−1 σ (ε) in Eq. (C2). Using this single relaxation time, transport coefficients and hence the Lorenz ratio can be evaluated easily using Eq. (10). Figure 4 shows the result for both 2D and 3D systems. The plot in Fig.4a is exactly the same as the plots for τσ in Fig.2a as one would expect. However this time we have only a single relaxation time characterizing both transports. Fig.4b is also identical to Fig.2b. The thermal conductivity for T (cid:29) TBG remains the same, but for T (cid:28) TBG we get a different temperature scaling such that the Lorenz ratio L = κ/(σT ) becomes independent of temperature. However, as 21 FIG. 4. Transport properties of a clean, coupled electron-phonon system using a single-relaxation-time version of the relaxation time ap- −1 at the Fermi energy versus temperature. (b) Electrical conductivity σ versus temperature. (c) Thermal proximation. (a) Scattering rate τ conductivity κ versus temperature. (d) Lorenz ratio L/L0 versus temperature. In all graphs, solid and dashed lines correspond to 2D and 3D systems respectively. Relaxation rates and transport coefficients are plotted relative to their values at T = TBG to make them dimensionless. The Fermi energy is chosen such that TF = 103TBG one can see from Fig.4d, L/L0 no longer saturates to unity but rather approaches a value which is almost half of what WF law predicts. This number can be expressed in terms of definite integrals over polylogarithm functions and turns out to be: (cid:40) section, the main point is that the WF law could still be vi- olated while both thermal and electrical transports are de- scribed by a single relaxation time. The validity of the RTA in general is discussed in Appendix E. L L0 ≈ 0.47 0.43 2D 3D , (T (cid:28) TBG) (D2) Appendix E: Discussion of the Approximations Generally the exact value depends on the specific form of in- teractions but is independent of system parameters. Note that the T (cid:28) TBG regime is exactly where the exponential factors in phonons' distribution function become important, which in turn makes the Sommerfeld expansion inapplicable. It is worth noting that modification of the Lorenz number due to the energy dependence of relaxation times has already been discussed in the context of electron-electron scattering30,31. Regardless of the validity of the RTA which is used in this The relaxation time approximation (RTA), as implemented in Eqs. (C2) is uncontrolled. One could easily imagine repeat- ing the calculations with a slightly different ansatz (for exam- ple, absorbing the factor of (ε−µ)/T into τκ in Eq. (8)); doing so can in fact lead to qualitatively different results. As such, we should give some justification for our choices. Note that in general the Boltzmann equation, being an integral equation, can be numerically solved iteratively, but such an iterative nu- merical solution has no mathematical transparency or physical 10−2100102104106T/TBG10−1010−610−2102106τ(εF,T)−1/τ(εF,TBG)−1τ−12Dτ−13D10−2100102104106T/TBG10−41011061011σ(T)/σ(TBG)2D3D10−2100102104106T/TBG100103106109κ(T)/κ(TBG)2D3D10−2100102104106T/TBG0.40.60.81.0L/L02D3D(a)(b)(c)(d)T5T4TT−5T−4T−1T−1/2T−3/2T−4T−3T0T1/2T−1/2L/L0=6π2L/L0=152π2L/L0=1L/L0≈0.47L/L0≈0.43TFTFTFTFunderstanding, forcing us to resort to the RTA which provides qualitatively correct, but numerically inaccurate, results. Our approximation does have the drawback that the On- First, we note that, as discussed in Appendix B, the RTA in the form we have used is exact when the scattering is elastic and isotropic. Therefore, in the high-temperature equiparti- tion regime T (cid:29) TBG where the scattering is quasi-elastic, the form of the RTA we have used is physically justified. Furthermore, in the low-temperature regime T (cid:28) Ti (cid:28) TBG, a controlled perturbative calculation is available7. In this regime, Matthiessen's rule is approximately valid, so the phonon contribution to the transport coefficients can be dis- entangled from the impurity contribution. The phonon con- tribution found in the perturbative approach is in qualitative agreement with our RTA results, both from the main text and Appendix C. Our choice of RTA is a good one in the sense that choosing other forms of the RTA will often lead to qualitative disagreement with the perturbative calculation in the regime where it is valid; see Appendix D for a one-relaxation-time example. Our choice of RTA therefore yields qualitative agreement with controlled results in the high- and low-temperature regimes. We therefore expect that our RTA results should give qualitatively correct results when interpolating between these two regimes, in particular in the regime of interest Ti (cid:28) T (cid:28) TBG (except if there are regimes where other en- ergy scales become important). As we are not concerned with quantitative predictions, this is sufficient for our purposes: to show that parametrically large violations of the Wiedemann- Franz law can occur in ordinary metals in a regime set by an energy scale TBG which may differ dramatically from TD. For accurate numerical results for the purpose of comparison with specific experimental results, one must resort to a full nu- merical solution of the Boltzmann integral equation which is well beyond the scope of the current work. 22 sager relation LT E = −T LET is violated. This is a very generic feature of any two-relaxation-time version of the RTA. The physical reason is that the Boltzmann equation, in its to- tal derivative form df /dt = 0, is exactly the statement of conservation of particle number. As such, an uncontrolled ap- proximate solution to the Boltzmann equation will generically lead to an uncontrolled non-conservation of particle number. Particle number conservation is assumed when proving this Onsager relation32, so there is no reason to expect that the Onsager relation will continue to hold for the approximate so- lution which violates this assumption. It so happens that a single-relaxation-time approximation does preserve the On- sager relation, but it will not typically lead to qualitatively correct results in the perturbative regime (see App. D). Although our approximation violates the Onsager relations, this does not lead to qualitative changes in our conclusions as long as the qualitative behavior of σ and LT T are cor- rect. In particular, since the thermopower at low temperatures can only provide a negative correction to the approximation κ ≈ −LT T , the parametric suppression of the Lorenz number below L0 that we have found can only be made more severe when the thermopower is accurately accounted for. Thus, the technical violation of the Onsager relation is an unimportant nuisance in our theory which we understand completely. It arises simply from the fact that RTA by itself cannot provide an exact solution of the Boltzmann integral equation except in certain special situations. In the main text, the numerical results are obtained from a schematic calculation for the thermal lifetime Eq. (24). This was done both for simplicity and for numerical stability. All of the results in the main text are in qualitative agreement with the RTA results in App. C in all asymptotic regimes, so the schematic numerical calculations are sufficient for our pur- poses.
1905.06027	1	1905	2019-05-15T08:30:32	Electrical control of lifetime-limited quantum emitters using 2D materials	[ "cond-mat.mes-hall", "physics.optics" ]	Solid state quantum emitters are a mainstay of quantum nanophotonics as integrated single photon sources (SPS) and optical nanoprobes. Integrating such emitters with active nanophotonic elements is desirable in order to attain efficient control of their optical properties but typically degrades the photostability of the emitter itself. Here, we demonstrate a tuneable hybrid device that integrates lifetime-limited single emitters (linewidth : 40 MHz) and 2D materials at sub-wavelength separation without degradation of the emission properties. Our device s nanoscale dimensions enable ultra-broadband tuning (tuning range larger than 400 GHz) and fast modulation (frequency : 100 MHz) of the emission energy, which renders it an integrated, ultra-compact tuneable SPS. Conversely, this offers a novel approach to optical sensing of 2D material properties using a single emitter as a nanoprobe.	cond-mat.mes-hall	cond-mat	Electrical control of lifetime-limited quantum emitters using 2D materials K. G. Schädler 1, C. Ciancico 1, S. Pazzagli 2 , 3, P. Lombardi 2 A. Bachtold 1, C. Toninelli 2 , 4, A. Reserbat-Plantey 1 and F.H.L. Koppens 1 , 5 1ICFO-Institut de Ciencies Fotoniques, The Barcelona Institute of Science and Technology, 08860 Castelldefels, 2LENS and CNR-INO, Via Nello Carrara 1, 50019 Sesto Fiorentino, Italy Spain 3Dipartimento di Fisica ed Astronomia, Università di Firenze, Via Sansone 1, 50019 Sesto Fiorentino, Italy 4QSTAR, Largo Fermi 2, 50125 Firenze, Italy 5ICREA -- Institució Catalana de Recerça i Estudis Avancats, 08010 Barcelona, Spain. Solid state quantum emitters are a mainstay of quantum nanophotonics as integrated single photon sources (SPS) and optical nanoprobes. Integrating such emitters with active nanophotonic elements is desirable in order to attain efficient control of their optical properties but typically degrades the photostability of the emitter itself. Here, we demonstrate a tuneable hybrid device that integrates lifetime-limited single emitters (linewidth ~ 40 MHz) and 2D materials at sub-wavelength separation without degradation of the emission properties. Our device's nanoscale dimensions enable ultra-broadband tuning (tuning range > 400 GHz) and fast modulation (frequency ~ 100 MHz) of the emission energy, which renders it an integrated, ultra-compact tuneable SPS. Conversely, this offers a novel approach to optical sensing of 2D material properties using a single emitter as a nanoprobe. Hybrid nanophotonic systems blend the strengths of distinct photonic elements to strongly enhance light-matter interactions1 in integrated photonic circuits. In these systems, narrow-linewidth quantum light emitters play a key role as single photon sources (SPS) which interact with their nanoscale environment2,3. Controlling these interactions provides versatile SPS tuning4 required for coupling quantum resources5 -- 7. Integrating nanoscale light emitters with two-dimensional (2D) materials is motivated by the rich physics of near-field interactions8 and new hybrid light-matter states9,10. This approach unites integrated solid-state SPS such as nitrogen vacancy centres11, quantum dots12 and single molecules13 with the diverse optoelectronic properties of 2D materials that facilitate emitting14, controlling15 -- 17 and detecting18 light at the nanoscale. In such hybrid devices, quantum emitters can be integrated at sub-wavelength separation to the 2D interface to achieve efficient near-field coupling8, which modifies the emitter's radiative decay rate19 -- 21 or transition energy22,23. Recent experimental studies integrated 2D materials with ensembles of broadband emitters to demonstrate electrical24 -- 26 and electromechanical27 tuning of the decay rate by controlling non-radiative energy transfer (nRET) or the energy flow to confined electromagnetic modes such as 2D polaritons26,28. Therefore, hybrids of 2D materials and SPS have the potential for in situ control of the conversion and channelling of single photons at the nanoscale. So far, these studies have been limited to ensembles and broad linewidth emitters. Integrating bright and narrow quantum emitters in such systems paves 1 the way towards a tuneable quantum light-matter interface, which is an essential ingredient for integrated quantum networks. Here, we demonstrate hybrid integration of 2D materials (semi-metallic graphene or semi-conducting MoS2) with single, lifetime-limited quantum emitters in nanocrystals to provide active emission control. Using the 2D materials as transparent electrodes, we show broadband Stark tuning of the emission energy over 40,000 times the emitter linewidth and fast modulation of the emitter's optical resonance on the time scale of its radiative lifetime. Such tuning can mitigate inhomogeneous broadening in solid-state environments to enable resonant and synchronised interaction between distinct quantum systems5 and allows for controlled coupling of narrowband quantum emitters to broadband nanophotonic circuitry29. Our approach is particularly suited for making integrated devices: although the 2D material is just tens of nanometres from the quantum emitter, we observe only weak emission linewidth broadening and spectral diffusion. In contrast, emitters close to bulk transparent electrodes30 such as ultra-thin metal films suffer strong emission quenching31, which hampers nanoscale integration. Furthermore, we find that the deposition process required to integrate a transparent conducting oxide inhibits molecular fluorescence, which highlights the potential of 2D materials for integration with ultra-sensitive quantum emitters. At the same time, we show that a single quantum emitter can be used as a transducer of the 2D materials' electronic properties. We choose single dibenzoterrylene (DBT) molecules as bright, photostable single photon sources32,33 emitting at 785 nm (1.58 eV) with lifetime-limited linewidth34 (∼ 40 MHz) at 3 K even when hosted in a sub-micron environment35. Experimentally, we perform scanning laser spectroscopy to address individual DBT molecules at sub-wavelength separation to a 2D metallic or semiconducting interface. The device comprises DBT molecules embedded in anthracene nanocrystals (see Supplementary Note 1), interspersed in a thin polyvinyl alcohol (PVA) film on a Si++/SiO2 substrate (Fig. 1a). MoS2 or graphene flakes are placed on top of the polymer film by a dry transfer technique and electrically contacted by gold electrodes. This transfer process is performed at room temperature and without other nanofabrication steps which potentially degrade emission36. Spatial maps of DBT ensemble fluorescence (see methods) at 3 K show bright, localised emission beneath both 2D materials (Fig. 1b). 2 Figure 1: Single molecules integrated with 2D materials. a Hybrid device schematic; Atomically thin layers of graphene and MoS2 cover fluorescent molecules embedded in a PVA film (hPVA = 300 nm) on SiO2 (hSiO2 = 285 nm). Single molecules are resonantly excited (inset) and their red-shifted fluorescence detected with a single photon counting module (SPCM). Electric fields are controlled by applying DC (𝑉𝑔) and AC (𝛿𝑉𝑔) potentials to the Si++ backgate. b Top to bottom: optical micrograph, AFM topography and DBT emission map for a MoS2 (left column) and bilayer graphene (right column) device. White dashed lines outline the flakes. Scale bars are 20 𝜇m. c Top panel: fluorescence excitation spectrum of an ensemble of single molecules in an uncovered nanocrystal at 3 K. Bottom panel: detail of two emission peaks with Lorentzian line shape (solid line), the narrower peak shows a FWHM of 43 ± 7 MHz. d Anti-bunching measurement for resonant excitation of a single peak as shown in c. The solid line is a fit to the data using a second-order correlation function. A simplified DBT energy level scheme is shown in the inset of Fig. 1a. The transition of interest is the zero-phonon line (ZPL) between the ground vibrational levels (v=0) of the electronic ground and excited states. Upon resonant excitation of this transition, the molecule relaxes either to the electronic ground state (30-40% of emission37) or to a higher vibrational state (v≠0), thereby emitting a red- shifted photon. Scanning laser spectroscopy (see methods) on a pristine nanocrystal (Fig. 1c) at 3 K reveals a series of sharp peaks from a DBT ensemble. This peak dispersion arises from local variations of strain and charge and enables spectrally addressing a single molecule38. The narrowest peaks in such ensembles display a typical linewidth of Γ = 2𝜋 × (43 ± 7) MHz. To confirm that we can address single molecules in a nanocrystal, we measure statistics of photon emission in Hanbury-Brown and Twiss (HBT) configuration (Fig. 1d) and fit the normalised data with the second-order intensity autocorrelation function39 g(2)(𝜏) = 1 − 𝐶 𝑒−𝜏/𝜏𝐷𝐵𝑇, where 𝐶 is the contrast. At zero delay time, we find g(2)(0) = 0.04 ± 0.02 < 0.5, which is signature of a single photon source. We extract an excited state lifetime39 of 𝜏𝐷𝐵𝑇 = 4.7 ± 0.5 ns, implying a lifetime-limited −1 = 2𝜋 × (34 ± 5) MHz, comparable to values observed in bulk anthracene linewidth Γ0 = 𝜏𝐷𝐵𝑇 crystals37. Therefore, single molecules in our device can exhibit lifetime-limited linewidth within the measurement error. In a solid state environment, this linewidth can be broadened by dephasing and near-field interactions, which can hamper the performance of emitters in nanostructured devices36,40. 3 In our device, a 2D material is placed in the near-field of a single emitter (separation 𝑑 ≪ 𝜆). In this near-field interactions non- as radiative energy transfer (nRET) from the molecule to the 2D material41 and Casimir-Polder level shifts8 are (CP) energy expected Both scale divergently interactions with separation as 𝑑−4, in contrast for bulk interfaces22. The nRET process leads to a linewidth broadening while the CP energy level shift is a quantum effect related to the modification vacuum fluctuations by the 2D interface. At fixed emitter-2D material separations, we can quantify the nRET contribution due to the presence of the 2D material by its impact emission linewidth. The CP contributions are difficult to quantify because for one specific emitter, the emission energy with and without the 2D material cannot be measured independently. Figure 2: Emission broadening and time stability in the presence of 2D materials. a Linewidth distribution for uncovered DBT (green), DBT covered by monolayer MoS2 (red) and bilayer graphene (grey). Solid lines are fits to the data as described in the main text. b Time trace of single DBT molecule emission (3 s / line) for uncovered DBT (bottom) and DBT covered by monolayer MoS2 (middle) and bilayer graphene (top). to 𝑑−3 to occur. on the regime, such of By measuring the emission linewidths of over 500 molecules (Fig. 2a) in three different configurations (uncovered, covered by graphene/MoS2), we quantify the effect of proximity to 2D materials on emission linewidth. We quantify the most probable linewidth Γ in each case by fitting the linewidth histogram with a Smirnov distribution42 which describes single molecule linewidth distributions perturbed by long-range coupling to two-level fluctuators, e.g. in polymers42. For uncovered nanocrystals, we obtain Γ𝑢𝑛𝑐𝑜𝑣. = 2𝜋 × 62 MHz, which is comparable to Γ0. This implies that the anthracene nanocrystal is a highly stable and crystalline environment35 and most molecules within it do not experience significant spectral diffusion due to defects or proximity to the surface35. While emitters covered by MoS2 experience extremely weak spectral broadening (Γ𝑀𝑜𝑆2 = 2𝜋 × 65 MHz), they are significantly broadened when covered by BLG (Γ𝐵𝐿𝐺 = 2𝜋 × 108 MHz) and their linewidth distributed over a larger range. We attribute this broadening to higher nRET efficiency to gapless graphene as electronic transitions can be optically excited over a large range of energies and in particular at ℏ𝜔𝑍𝑃𝐿. In contrast, MoS2 has bandgap and excitonic resonances at energies > ℏ𝜔𝑍𝑃𝐿, resulting in weak nRET which preserves the narrow linewidth. These results show that neither the presence nor the integration process of 2D materials have a strong detrimental impact on the emission properties. In contrast, we find that the deposition of a commonly used transparent electrode material such as ITO results in the complete disappearance of DBT emission (see Supplementary Note 2) due 4 to the deposition process. This highlights the weakly invasive nature of 2D material integration, which makes it particularly suitable for fragile quantum emitters. To confirm that emission stability in time is preserved after 2D material integration, we measure DBT spectra over time for the three configurations (uncovered, covered by graphene/MoS2) as shown in Fig. 2b. In each case, we observe stable emission intensity and fluctuations of ZPL frequency and linewidth below 42 MHz (see Supplementary Note 3). Slightly higher fluctuations of 𝜔𝑍𝑃𝐿 - possibly induced by charge fluctuations - are resolved in the case of MoS2. Overall, these results show that 2D semiconductors are particularly suitable for integration with sensitive quantum emitters at nanoscale proximity, introducing negligible perturbation in the photostability. Figure 3: Stark tuning of a single molecule with a 2D electrode. a Spectral map showing the Stark shift of an ensemble of single molecules under BLG with back-gate voltage 𝑉𝑔. b Gate-induced line shift of a single molecule under BLG (grey circles) and monolayer MoS2 (red circles). Solid lines are parabolic fits (see main text).c Back-gate dependence of MoS2 photoluminescence (PL) spectra at 532 nm (2.33 eV) excitation. d PL intensity of exciton (1.96 eV, blue), negatively charged trion (1.93 eV, green) and defect band (1.7 eV, red) with 𝑉𝑔. e Emission linewidth for a single molecule under BLG (grey) and monolayer MoS2 (red). We now turn to electrical manipulation of single emitters. To achieve Stark43 tuning of DBT emission energy, we apply an electric potential 𝑉𝑔 over the capacitor formed by the 2D electrode and the Si++ back-gate, separated by a PVA/SiO2 layer (see Fig. 1a). We reach comparatively large44,45 field strengths above 2 MV.cm-1 before dielectric breakdown takes place. DBT ensemble spectra below BLG at different 𝑉𝑔 (Fig. 3a) show a large, dominantly quadratic shift of the emitters in the whole ensemble on the order of hundreds of GHz (~ 104 Γ0/2𝜋), comparable to the inhomogeneous broadening of the entire ensemble35. While this tuning range is comparable to state-of -- the-art devices employing semiconductor quantum dots45 and diamond defects46, we highlight that our device displays emission linewidths within a smaller footprint, which is attractive for nanophotonic integration. We also verify that DBT emission energy remains stable under large applied electric fields by repeating stability measurements as shown in Fig. 2b up to large values of 𝑉𝑔 (see Supplementary Note 3). We find small, 5 gate-dependent emission energy drifts (< 3 MHz/s), likely due to small leakage currents in the device, which can be reduced by using a smaller 2D electrode area or by active feedback29. For a centrosymmetric molecule such as DBT, we expect a quadratic Stark shift with linear contributions arising from dipole moments induced by distortion of the molecule's insertion site44. 2 The Stark shift is modelled as Δ𝜔𝑍𝑃𝐿 = 𝑎𝐸⃗ + 𝑏𝐸⃗ , where 𝑎 and 𝑏 are the linear and quadratic Stark coefficients, and functions of the dipole moment and polarizability change, respectively, between ground and excited state. Here, 𝐸⃗ = 𝐸⃗ 𝑒𝑥𝑡 − 𝐸⃗ 0 is the net electric field experienced by the molecule upon application of an external field 𝐸⃗ 𝑒𝑥𝑡 = 𝑉𝑔/ℎ𝑡𝑜𝑡, (ℎ𝑡𝑜𝑡 = ℎ𝑃𝑉𝐴 + ℎ𝑆𝑖𝑂2 ~ 600 nm) with a correction factor 𝐸⃗ 0 that accounts for trapped charges and work function differences47. Fitting detuning data for different emitters, we extract typical Stark coefficients of 𝑎 ~ 500 MHz/(kV cm-1) and 𝑏 ~ -0.5 MHz/(kV cm-1)2 (see Supplementary Note 4), comparable to reported results34. The observed quadratic stark shift shows that 𝐸𝑒𝑥𝑡 is linear in 𝑉𝑔, as is expected for a metallic electrode such as BLG (Fig. 3b). In contrast, emitters under MoS2 deviate from the parabolic detuning below 𝑉𝑔,0 ~ -30 V (Fig. 3b), where the detuning flattens and Δ𝜔𝑍𝑃𝐿 is almost independent of 𝑉𝑔. We attribute this deviation to a gate-induced change of resistivity48: as the charge carrier density in a semiconductor changes more abruptly with 𝑉𝑔 , we expect a sub-linear dependence of 𝐸𝑒𝑥𝑡(𝑉𝑔) as excess charge carriers are depleted in MoS2. Electrostatic doping of the MoS2 electrode is independently confirmed by gate- dependent photoluminescence (Fig. 3c, d). The ratio of charged (𝑋−) to neutral (𝑋0) exciton PL intensities increases with 𝑉𝑔 due to n-doping of the MoS2 electrode. We also observe a broad emission 49 - which overlaps peak at lower energy for negative 𝑉𝑔 - associated to emission from defects in MoS2 with ℏ𝜔𝑍𝑃𝐿. With the appearance of this defect band, we observe a linewidth broadening 𝛤𝐷𝐵𝑇 below 𝑉𝑔,0 for some emitters, which is absent in the case of a metallic BLG electrode (Fig. 3e). This suggests that defects could act as n-RET acceptors, leading to emission linewidth broadening of emitters close to MoS2. Our system combines large Stark tuning with lifetime-limited emitters that potentially enables adiabatic control up to a frequency set by their natural linewidth Γ0. We investigate the range of this dynamical control by applying oscillating potentials 𝑉𝑡𝑜𝑡 = 𝑉𝑔 + 𝛿𝑉𝑔(𝑓𝐴𝐶) to monolayer graphene and MoS2 electrodes. For fast modulation frequencies compared to the spectral acquisition time (𝑓𝐴𝐶 ≫ −1 ∼ 0.1 Hz), DBT emission peaks show a splitting 𝜎(𝑓𝐴𝐶) (Fig. 4a) proportional to the modulation 𝑡𝑚𝑒𝑎𝑠 amplitude 𝛿𝑉𝑔 and the local Stark slope ∂𝜔𝑍𝑃𝐿 : 𝜕𝑉𝑔 𝜎 = ∂𝜔𝑍𝑃𝐿 𝜕𝑉𝑔 𝛿𝑉𝑔 𝑇 (1) Here, 𝑇(𝑓𝐴𝐶) is the transmission of the 𝑅𝐶-low-pass filter governed by the effective device resistance 𝑅 which includes contact resistance as well as the sheet resistance of the 2D material and backgate capacitance 𝐶 (see Supplementary Note 5). At fixed 𝑉𝑔 and 𝑓𝐴𝐶, the DBT emission spectra reflect the oscillation turning points of the modulating waveform (Fig. 4a). Under square modulation with a fixed amplitude 𝛿𝑉𝑔, the splitting vanishes with increasing 𝑓𝐴𝐶 for both MoS2 and graphene electrodes (Fig. 4b), albeit at a lower frequency for MoS2. From this measurement, we extract 𝑇(𝑓𝐴𝐶) (Eq. 1) and confirm low-pass behaviour with a characteristic cut-off frequency 𝑓−3𝑑𝐵 (Fig. 4c). Interestingly, 𝑇(𝑓𝐴𝐶) shows a strong gate voltage 𝑉𝑔 dependence for MoS2 electrode, which is not observed for graphene. This dependence is quantified by 𝑓−3𝑑𝐵, which increases with 𝑉𝑔 and saturates at ∼ 5 kHz ≪ Γ0 for 𝑉𝑔 > 0 V (Fig. 4d). We attribute this behaviour to a change of sheet resistivity and contact 6 resistance due to gate-induced electrostatic n-doping48 of the MoS2, consistent with PL measurements (Fig. 3c,d). As a result, 𝑓−3𝑑𝐵 ∝ (𝑅𝐶)−1 increases as we sweep 𝑉𝑔 from negative to positive values. Figure 4: Dynamical emission control of quantum emitter. a Single molecule emission as a function of AC amplitude 𝛿𝑉𝑔 for (left to right) sinusoidal, square and pulsed modulation (𝑓𝐴𝐶 = 1 kHz), and for pink noise (100 kHz bandwidth). b Single DBT emission spectra vs. 𝑓𝐴𝐶 (𝑉𝑔 + 𝛿𝑉𝑔 = 20 + 1 V) under MoS2 (upper panel) and monolayer graphene (lower panel). c Transmission function 𝑇(𝑓𝐴𝐶) for a single molecule at different 𝑉𝑔 under MoS2 (filled circles) and graphene (open circles). d Back-gate dependence of MoS2 cut-off frequency 𝑓−3𝑑𝐵. e Transmission 𝑇(10 𝑘𝐻𝑧) as a function of 𝑉𝑔 for MoS2 (red) and BLG (grey). f Histogram of time-resolved single molecule emission intensity modulation at 𝑓𝐴𝐶 = 100 MHz. The transition energy 𝜔𝑍𝑃𝐿 is modulated electrically while the excitation energy 𝜔𝑒𝑥𝑐 is kept constant, leading to a periodic emission strength in time. To capture the gating efficiency of our 2D electrodes, we extract transmission 𝑇 at fixed 𝑓𝐴𝐶 = 10 kHz varying 𝑉𝑔 (Fig. 4e, and SI). The strong reduction of 𝑇 for 𝑉𝑔 < 0 𝑉 is consistent with a large gate- induced change of resistance in the semiconducting MoS2 device, while graphene's metallicity maintains 𝑇 ∼ 1 over the full 𝑉𝑔 range. Therefore, graphene enables Stark tuning of emitters over a large energy range and at high frequency at the expense of weak linewidth broadening. To extend this modulation bandwidth, we reduce the gate capacitance 𝐶 by using a thicker PVA dielectric (ℎ𝑃𝑉𝐴 = 800 nm). Then, we modulate ℏ𝜔𝑍𝑃𝐿 around a fixed excitation energy by applying 𝛿𝑉𝑔 at 𝑓𝐴𝐶 = 100 MHz ∼ Γ0/2𝜋 to a graphene electrode. Using a time-correlated single photon counter synchronised to the modulation 𝛿𝑉𝑔, we observe a periodic oscillation of emission intensity (Fig. 4f) with a modulation depth of ~ 50 %. This oscillation is a signature of the excitation laser periodically exploring a fraction of the emitter's absorption line. These measurements show that the dynamical modulation bandwidth of our devices - determined by the 2D electrode material and device geometry -- approaches Γ0. Conversely, single emitters act as local nanoprobes of the 2D material's electronic properties. 7 In conclusion, we demonstrate a novel hybrid device where 2D materials are integrated with lifetime- limited single photon emitters to achieve broadband and fast emission energy tuning. Our results highlight the potential of atomically thin electrodes for integration with sensitive quantum emitters in a nanoscale device without perturbing narrow emission linewidth, in contrast to commonly employed bulk transparent electrodes. At high frequencies, Stark modulation reveals low-pass transmission behaviour related to the 2D materials' sheet resistivity. Using a graphene electrode, we show emission energy modulation at frequencies approaching the emitters' linewidth. Our device thus provides resonant tuning and high-frequency modulation of SPS on chip, required for obtaining indistinguishable and synchronised single photons. Finally, our device is a platform for studying novel forms of light-matter interaction with plasmon-polaritons in graphene9 and exciton-polaritons in TMDs10 at the single excitation level. Conversely, strong coupling to such excitations could allow the observation of normally forbidden higher-order transitions of the emitter50. Acknowledgments: We acknowledge Rinu Abraham Maniyara for advice regarding transparent electrodes. F.H.L.K. acknowledges financial support from the Government of Catalonia trough the SGR grant (2014-SGR- 1535), and from the Spanish Ministry of Economy and Competitiveness, through the "Severo Ochoa" Programme for Centres of Excellence in R&D (SEV-2015-0522), support by Fundacio Cellex Barcelona, CERCA Programme / Generalitat de Catalunya and the Mineco grants Ramón y Cajal (RYC-2012-12281) and Plan Nacional (FIS2013-47161-P and FIS2014-59639-JIN). Furthermore, the research leading to these results has received funding from the European Union Seventh Framework Programme under grant agreement Orquid. Methods: Device fabrication. We make a suspension of anthracene nanocrystals hosting DBT molecules in PVA by reprecipitation. The suspension is spin-cast onto a p-doped Si wafer coated with 285 nm thermally grown SiO2. Electrodes are deposited onto the coated chip by thermal evaporation of 100 nm Au through a shadow mask. Nanocrystals close to the electrode displaying DBT fluorescence are localised at room temperature using off-resonant wide-field illumination. Bulk MoS2 and graphite is exfoliated mechanically using commercial polydimethylsiloxane (PDMS) sheets. Mono- and bilayer flakes of both materials are identified optically by absorption contrast (see Supplementary Note 6) and transferred onto anthracene nanocrystals in PVA by dry stamp transfer. Electrical device actuation and optical readout. Measurements are performed under vacuum in a cryostat at 3 K. Our hybrid devices are actuated electrically using a low-noise voltage source and an arbitrary waveform generator to provide DC and AC voltages. Using a custom-built confocal microscope to locally illuminate the device, we excite single molecules with a tuneable 785 nm laser with circular polarisation at 5 nW, well below saturation power ∼ 20 nW (see Supplementary Note 7). Red-shifted single photon emission is detected with a single photon counting module (SPCM) combined with spectral ZPL filtering by a long-pass filter. Emission maps are made using pseudo- broadband excitation of DBT ensemble fluorescence by fast modulation (200 Hz) of the laser detuning compared to the SPCM integration time (∼10 ms). MoS2 photoluminescence is excited using a 532 nm laser and detected with a spectrometer. 8 Author contributions K.G.S., A.R.-P., C.T. and F.H.L.K. conceived the experiment. K. G. S. and C.C. made the samples with the support of S.P., P.L. and C.T. Measurements and analysis were performed by K.G.S., C.C. and A.R.-P. K.G.S, A.R.-P. and F.H.L.K. wrote the manuscript with critical comments from all authors. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. Benson, O. Assembly of hybrid photonic architectures from nanophotonic constituents. Nature 480, 193 -- 199 (2011). Iwasaki, T. et al. Direct Nanoscale Sensing of the Internal Electric Field in Operating Semiconductor Devices Using Single Electron Spins. ACS Nano 11, 1238 -- 1245 (2017). Thompson, J. D. et al. Coupling a Single Trapped Atom to a Nanoscale Optical Cavity. Science (80). 340, 1202 -- 1205 (2013). Petruzzella, M. et al. Fully tuneable, Purcell-enhanced solid-state quantum emitters. Appl. Phys. Lett. 107, (2015). Siyushev, P., Stein, G., Wrachtrup, J. & Gerhardt, I. Molecular photons interfaced with alkali atoms. Nature 509, 66 -- 70 (2014). Lettow, R. et al. Quantum interference of tunably indistinguishable photons from remote organic molecules. Phys. Rev. Lett. 104, 26 -- 29 (2010). O'Brien, J. L., Furusawa, A. & Vučković, J. Photonic quantum technologies. Nat. Photonics 3, 687 -- 695 (2010). Henkel, C. & Sandoghdar, V. Single-molecule spectroscopy near structured dielectrics. Opt. Commun. 158, 250 -- 262 (1998). Koppens, F. H. L., Chang, D. E. & de Abajo, F. J. Graphene Plasmonics: A Platform for Strong Light-Matter Interactions. Nano Lett. 11, 3370 -- 3377 (2011). Karanikolas, V. D., Marocico, C. A., Eastham, P. R. & Bradley, A. L. Near-field relaxation of a quantum emitter to two-dimensional semiconductors: Surface dissipation and exciton polaritons. Phys. Rev. B 94, 195418 (2016). Tamarat, P. et al. Stark shift control of single optical centers in diamond. Phys. Rev. Lett. 97, 083002 (2006). Senellart, P., Solomon, G. & White, A. High-performance semiconductor quantum-dot single- photon sources. Nat. Nanotechnol. 12, 1026 -- 1039 (2017). Orrit, M., Bernard, J., Zumbusch, A. & Personov, R. I. I. Stark effect on single molecules in a polymer matrix. Chem. Phys. Lett. 196, 595 -- 600 (1992). Sundaram, R. S. et al. Electroluminescence in Single Layer MoS 2. Nano Lett. 13, 1416 -- 1421 (2013). Liu, M. et al. A graphene-based broadband optical modulator. Nature 474, 64 -- 67 (2011). Chakraborty, C. et al. Quantum-Confined Stark Effect of Individual Defects in a van der Waals Heterostructure. Nano Lett. 17, 2253 -- 2258 (2017). 17. Roch, J. G. et al. Quantum-Confined Stark Effect in a MoS2 Monolayer van der Waals 9 Heterostructure. Nano Lett. 18, 1070 -- 1074 (2018). 18. 19. 20. 21. Konstantatos, G. et al. Hybrid graphene -- quantum dot phototransistors with ultrahigh gain. Nat. Nanotechnol. 7, 363 -- 368 (2012). Gaudreau, L. et al. Universal Distance-Scaling of Nonradiative Energy Transfer to Graphene. Nano Lett. 13, 2030 -- 2035 (2013). Tisler, J. et al. Single Defect Center Scanning Near-Field Optical Microscopy on Graphene. Nano Lett. 13, 3152 -- 3156 (2013). Raja, A. et al. Energy Transfer from Quantum Dots to Graphene and MoS 2 : The Role of Absorption and Screening in Two-Dimensional Materials. Nano Lett. 16, 2328 -- 2333 (2016). 22. Muschik, C. A. et al. Harnessing Vacuum Forces for Quantum Sensing of Graphene Motion. Phys. Rev. Lett. 112, 223601 (2014). 23. 24. 25. 26. 27. 28. 29. Chang, C.-H., Rivera, N., Joannopoulos, J. D., Soljačić, M. & Kaminer, I. Constructing "Designer Atoms" via Resonant Graphene-Induced Lamb Shifts. ACS Photonics 4, 3098 -- 3105 (2017). Lee, J. et al. Switching Individual Quantum Dot Emission through Electrically Controlling Resonant Energy Transfer to Graphene. Nano Lett. 14, 7115 -- 7119 (2014). Prasai, D. et al. Electrical Control of near-Field Energy Transfer between Quantum Dots and Two-Dimensional Semiconductors. Nano Lett. 15, 4374 -- 4380 (2015). Tielrooij, K. J. et al. Electrical control of optical emitter relaxation pathways enabled by graphene. Nat. Phys. 11, 281 -- 287 (2015). Reserbat-Plantey, A. et al. Electromechanical control of nitrogen-vacancy defect emission using graphene NEMS. Nat. Commun. 7, 10218 (2016). Low, T. et al. Polaritons in layered two-dimensional materials. Nat. Mater. 16, 182 -- 194 (2016). Schmidgall, E. R. et al. Frequency Control of Single Quantum Emitters in Integrated Photonic Circuits. Nano Lett. 18, 1175 -- 1179 (2018). 30. Maniyara, R. A., Mkhitaryan, V. K., Chen, T. L., Ghosh, D. S. & Pruneri, V. An antireflection transparent conductor with ultralow optical loss (<2 %) and electrical resistance (<6 Ω sq−1). Nat. Commun. 7, 13771 (2016). Amos, R. M. & Barnes, W. L. Modification of the spontaneous emission rate of Eu3+ ions close to a thin metal mirror. Phys. Rev. B 55, 7249 -- 7254 (1997). Trebbia, J.-B., Tamarat, P. & Lounis, B. Indistinguishable near-infrared single photons from an individual organic molecule. Phys. Rev. A 82, 063803 (2010). Toninelli, C. et al. Near-infrared single-photons from aligned molecules in ultrathin crystalline films at room temperature. Opt. Express 18, 6577 (2010). Nicolet, A. A. L. et al. Single Dibenzoterrylene Molecules in an Anthracene Crystal: Main Insertion Sites. ChemPhysChem 8, 1929 -- 1936 (2007). Pazzagli, S. et al. Self-Assembled Nanocrystals of Polycyclic Aromatic Hydrocarbons Show Photostable Single-Photon Emission. ACS Nano 12, 4295 -- 4303 (2018). Liu, J. et al. Single Self-Assembled InAs/GaAs Quantum Dots in Photonic Nanostructures: The Role of Nanofabrication. Phys. Rev. Appl. 9, 64019 (2018). 10 31. 32. 33. 34. 35. 36. 37. Trebbia, J.-B., Ruf, H., Tamarat, P. & Lounis, B. Efficient generation of near infra-red single photons from the zero-phonon line of a single molecule. Opt. Express 17, 23986 (2009). 38. Moerner, W. E., Orrit, M., Wild, U. P. & Basché, T. Single-Molecule Optical Detection, Imaging and Spectroscopy. (Wiley, 2008). 39. Michler, P. et al. Quantum correlation among photons from a single quantum dot at room temperature. Nature 406, 968 -- 70 (2000). 40. 41. 42. 43. 44. 45. 46. 47. 48. Aharonovich, I., Englund, D. & Toth, M. Solid-state single-photon emitters. Nat. Photonics 10, 631 -- 641 (2016). Gómez-Santos, G. & Stauber, T. Fluorescence quenching in graphene: A fundamental ruler and evidence for transverse plasmons. Phys. Rev. B 84, 165438 (2011). Gmeiner, B., Maser, A., Utikal, T., Götzinger, S. & Sandoghdar, V. Spectroscopy and microscopy of single molecules in nanoscopic channels: spectral behavior vs. confinement depth. Phys. Chem. Chem. Phys. 18, 19588 -- 19594 (2016). Basché, T., Moerner, W. E., Orrit, M. & Talon, H. Photon antibunching in the fluorescence of a single dye molecule trapped in a solid. Phys. Rev. Lett. 69, 1516 -- 1519 (1992). Brunel, C., Tamarat, P., Lounis, B., Woehl, J. C. & Orrit, M. Stark effect on single molecules of dibenzanthanthrene in a naphthalene crystal and in a n-hexadecane Shpol'skii matrix. J. Phys. Chem. A 103, 2429 -- 2434 (1999). Bennett, A. J. et al. Giant Stark effect in the emission of single semiconductor quantum dots. Appl. Phys. Lett. 97, 1 -- 4 (2010). Acosta, V. M. et al. Dynamic stabilization of the optical resonances of single nitrogen-vacancy centers in diamond. Phys. Rev. Lett. 108, 6 -- 11 (2012). Fontana, M. et al. Electron-hole transport and photovoltaic effect in gated MoS2 Schottky junctions. Sci. Rep. 3, 1634 (2013). Baugher, B. W. H., Churchill, H. O. H., Yang, Y. & Jarillo-Herrero, P. Intrinsic Electronic Transport Properties of High-Quality Monolayer and Bilayer MoS2. Nano Lett. 13, 4212 -- 4216 (2013). 49. Mak, K. F. et al. Tightly bound trions in monolayer MoS2. Nat. Mater. 12, 207 -- 211 (2012). 50. Rivera, N., Kaminer, I., Zhen, B., Joannopoulos, J. D. & Soljačić, M. Shrinking light to allow forbidden transitions on the atomic scale. Science (80). 353, 263 -- 269 (2016). 11
1306.4855	1	1306	2013-06-20T12:48:59	Quantum Transport With Two Interacting Conduction Channels	[ "cond-mat.mes-hall" ]	The transport properties of a conduction junction model characterized by two mutually coupled channels that strongly differ in their couplings to the leads are investigated. Models of this type describe molecular redox junctions (where a level that is weakly coupled to the leads controls the molecular charge, while a strongly coupled one dominates the molecular conduction), and electron counting devices in which the current in a point contact is sensitive to the charging state of a nearby quantum dot. Here we consider the case where transport in the strongly coupled channel has to be described quantum mechanically (covering the full range between sequential tunneling and co-tunneling), while conduction through the weakly coupled channel is a sequential process that could by itself be described by a simple master equation. We compare the result of a full quantum calculation based on the pseudoparticle non-equilibrium Green function method to that obtained from an approximate mixed quantum-classical calculation, where correlations between the channels are taken into account through either the averaged rates or the averaged energy. We find, for the steady state current, that the approximation based on the averaged rates works well in most of the voltage regime, with marked deviations from the full quantum results only at the threshold for charging the weekly coupled level. These deviations are important for accurate description of the negative differential conduction behavior that often characterizes redox molecular junctions in the neighborhood of this threshold.	cond-mat.mes-hall	cond-mat	Quantum Transport With Two Interacting Conduction Channels Alexander J. White,1 Agostino Migliore,2, a) Michael Galperin,1 and Abraham Nitzan2 1)Department of Chemistry & Biochemistry, University of California San Diego, La Jolla, CA 92093, USA 2)School of Chemistry, Tel Aviv University, Tel Aviv, 69978, Israel (Dated: 11 January 2021) The transport properties of a conduction junction model characterized by two mutually coupled channels that strongly differ in their couplings to the leads are investigated. Models of this type describe molecular redox junctions (where a level that is weakly coupled to the leads controls the molecular charge, while a strongly coupled one dominates the molecular conduction), and electron counting devices in which the current in a point contact is sensitive to the charging state of a nearby quantum dot. Here we consider the case where transport in the strongly coupled channel has to be described quantum mechanically (covering the full range between sequential tunneling and co-tunneling), while conduction through the weakly coupled channel is a sequential process that could by itself be described by a simple master equation. We compare the result of a full quantum calculation based on the pseudoparticle non-equilibrium Green function method to that obtained from an approximate mixed quantum-classical calculation, where correlations between the channels are taken into account through either the averaged rates or the averaged energy. We find, for the steady state current, that the approximation based on the averaged rates works well in most of the voltage regime, with marked deviations from the full quantum results only at the threshold for charging the weekly coupled level. These deviations are important for accurate description of the negative differential conduction behavior that often characterizes redox molecular junctions in the neighborhood of this threshold. PACS numbers: 73.63.-b,82.20.Xr,85.65.+h,82.20.Wt Keywords: electron tunneling, redox molecular junctions, quantum correlation, rate equations, pseudoparticle nonequilibrium Green functions I. INTRODUCTION Transport in mesoscopic and nanoscopic junctions is usually a multichannel phenomenon. Model studies of transport in junctions that comprise two, often interact- ing, conduction channels have been carried out in order to describe the essential features of different physical phe- nomena. Prominent examples are studies of interference effects in quantum conduction, analysis of single electron counting, where a highly transmitting junction (a point contact) is used to monitor the electronic state of a poorly transmitting one, and redox molecular junctions, where (transient) electron localization in one channel, stabilized by environmental polarization, determines the transition between redox states that are observed by the conduc- tion properties of another channel. These three classes of phenomena are described by different flavors of the two-channel model. Interference is usually discussed as a single electron problem and interaction with the envi- ronment is minimized (often disregarded in model stud- ies) so as to maintain phase coherent transport. Single electron counting with a point-contact detector is by def- inition a many electron problem, however environmental interactions are again minimized (and again often dis- regarded in theoretical analysis) by lowering the exper- imental temperature in order to obtain detectable sig- a)Present address: Department of Chemistry, Duke University, Durham, NC 27708, USA nals. Conduction in redox junctions is usually observed in room temperature polar environments and is charac- terized by large solvent reorganization that accompanies the electron localization at the redox site. In recent work1 -- 3,7 we have studied the conduction properties of junctions of the latter type. We first ana- lyzed, for a model involving a single conduction channel, the consequence of large solvent reorganization in the limit where the coupling between the molecular bridge and the metal leads is large relative to the frequency of the phonon mode used to model the solvent dynamical response.1 -- 3 It was shown (using a mean field descrip- tion essentially equivalent to the Born Oppenheimer ap- proximation) that solvent induced stabilization of differ- ent charging states of the molecule can result in multi- stable operation of the junction, offering a possible ratio- nalization of observations of negative differential resis- tance (NDR) and hysteretic response in molecular redox junctions. Such multistability was indeed observed re- cently in numerical simulations that avoid the mean field approximation.8,9 Many redox junctions, however, oper- ate in the opposite limit of relatively small molecule-lead coupling, where a single conduction channel model can- not show multistable transport behavior. Two of us have recently advanced a two channel model that can account for such observations.7 In the absence of electron-phonon interaction (solvent polarization) this model is given by 3 1 0 2 n u J 0 2 ] l l a h - s e m . t a m - d n o c [ 1 v 5 5 8 4 . 6 0 3 1 : v i X r a 2 amounts to molecular redox states that affect the trans- mission, therefore the observed current, through the fast channel. Bistability and hysteretic response on experi- mentally relevant timescales are endowed into the model in a trivial way23 and, as was shown in Ref. 7 (see also Refs. 24 -- 26), NDR also appears naturally under suitable conditions. Obviously, this behavior is generic and results from the timescale separation between the W and S channels to- gether with the requirement that the observed current is dominated by the S channel. In Refs. 7 and 27, we have described the expected phenomenology of such junction model in the limit where transport through both chan- nels is described by simple kinetic equations with Marcus electron transfer rates. While, as indicated above, it is natural to model the slow dynamics (observed timescales ∼ 10−6 s) in this way, it is also of interest to consider fast channel transport on timescales where transport co- herence is maintained. For example, one could envision a redox junction that switches between two conduction modes, which shows interference pattern associated with the structure of the fast channel. As a prelude for such considerations, we have studied in Ref. 27 also a model in which the weakly coupled channel W is described by Mar- cus kinetics, however conduction through the strongly coupled channel S is described as a coherent conduc- tion process by means of the Landauer formula, assum- ing that the timescale of transport through this channel is fast enough to make it possible to ignore any inter- action with the polar environment. As in any mixed quantum-classical dynamics, such description is not con- sistently derived from a system Hamiltonian, and ad-hoc assumptions about the way the quantum and classical subsystems interact with each other must be invoked, as described in Section II. In this paper we present a full quantum calculation of the current-voltage response of the two channel model described above, and use it to assess the approximate so- lution obtained using Eqs. (2)-(6) with models A and B (see Section II). The quantum calculation is done with the pseudoparticle non-equilibrium Green function (PP- NEGF) technique,28 -- 31 named the slave boson technique when applied to a 3-states system (Anderson problem at infinite U ),32 -- 35 which was recently used by two of us to study effects of electron-phonon and exciton-plasmon interactions in molecular junctions.36,37 We note that all the methods used in the paper have their own limita- tions. In particular, PP-NEGF is perturbative in the system-bath coupling. However, it accounts exactly for the intra-system interactions, and it is the role of these interactions (quantum correlations due to system chan- nels interactions) which is missed by the mixed quantum classical approaches and is the focus of the present study. In Section II we present our model, briefly review the master equation description and introduce two approxi- mate descriptions of mixed classical-quantum dynamics. The PP-NEGF technique and other details of the fully quantum calculation are described in Section III. Sec- FIG. 1. The two channel model discussed in the paper. Each channel comprises one level coupled to the left and right electrodes. W and S denote weakly and strongly coupled levels, respectively. the Hamiltonian (see Fig. 1) m=S,W (cid:88) (cid:88) (cid:88) k∈L,R k∈L,R (cid:16) (cid:16) H = + + d† m εm dm + U nS nW + (cid:17) (cid:17) † VkW c k dW + H.c. † VkS c k dS + H.c. (cid:88) k∈L,R † kck εkc (1) m † where d† m (c k) creates electron in level m (state k of the contact), and nm = d† dm, m = S, W . In this model, the two channels are coupled only capacitively (no inter- channel electron transfer). U represents the standard Coulomb interaction between them. Two coupled chan- nel models such as (1) also characterize single electron counting devices,10 -- 15 where the current in a point con- tact (that can be represented by channel S) measures the charging state of a quantum dot used as a bridge in a nearby junction (channel W ). The noise properties of such junctions have been studied extensively.16 -- 21 In this model, supplemented by electron phonon cou- pling that represents the response of a polar environ- ment to the electronic occupations in levels W and S, the molecular redox site dominates the properties of one channel (addressed below as "weakly coupled" or "slow" and denoted by W ), characterized by strong transient localization stabilized by large reorganization of the po- lar environment and weak coupling to the metal leads. Transport through this channel, that is, charging and discharging of the molecular redox site, was described by sequential kinetic processes. A second channel (ad- dressed below as "strongly coupled" or "fast" and de- noted by S) is more strongly coupled to the leads and is responsible for most or all of the observed current.22 Switching between charging states of the slow channel  tion IV presents our results and discusses the validity of the approximate calculations. Section V concludes. II. MIXED QUANTUM CLASSICAL APPROXIMATIONS To account for the current-voltage behavior of a junction characterized by the Hamiltonian (1), several workers16 -- 21 have used a master equation level of descrip- tion, whereupon, for a given voltage, the dynamics of populating and de-populating the levels S and W is de- scribed by classical rate equations involving only their populations, with occupation and de-occupation rates given by standard expressions (see Eq. (4) below). Here, in order to focus on redox junction physics, the coupling of channel W to the contacts is assumed to be much smaller than that of channel S, so that in the absence of correlations channel W can be assumed to be classi- cal and treated within such rate equations approach. At the same time channel S will be treated as quantum, as discussed in the previous section. In Ref. 27, we have assumed that on the timescale of interest the junction can be in two states: 1 and 0, where the weakly coupled channel, that is the molecular redox site - is occupied or vacant, respectively. The probability P1 = 1 − P0 that the junction is in state 1 satisfies the kinetic equation dP1 dt = (1 − P1) k0→1 − P1k1→0 (2) where the rates k0→1 and k1→0 are electron transfer rates between a molecule and an electrode, here the rates to occupy and vacate the redox molecular site, respectively. These rates are sums over contributions from the two electrodes ki→j = k(L) i→j + k(R) i→j; i, j = 0, 1 (3) and depend on the position of the redox molecular orbital energy εr relative to the Fermi energy (electronic chemi- cal potential) of the corresponding electrode. In Ref. 27 we have used Marcus heterogeneous electron transfer the- ory to calculate these rates, thus taking explicitly into ac- count solvent reorganization modeled as electron-phonon coupling in the high temperature and strong coupling limit. For the purpose of the present work it is enough to use the simpler, phonon-less, model k(K) 0→1 (εr) =ΓK k(K) 1→0 (εr) =ΓK r fK (εr) r [1 − fK (εr)] (4) where εr is the energy of the "redox level" (see below), −1 (K = L, R) is the fK(E) = [exp ((E − µK)/T ) + 1] Fermi-Dirac function of the electrode K, µK is the cor- responding electronic chemical potential and T is the temperature (in energy units). ΓK r , K = L, R are the widths of the redox molecular level due to its electron transfer coupling to the electrodes.38 In terms of the 3 W = 2π(cid:80) Hamiltonian, Eq. (1) above, these widths are given by k∈KVW k2δ (E − εk). We have assumed that ΓK in the relevant energy regions these widths do not depend on energy. From Eqs. (2) and (3), the steady state population of the redox site is P1 = 1 − P0 = k0→1/ (k0→1 + k1→0), and the current through the weakly coupled channel is IW = k(L) 0→1P0. This current is however negligible relative to the contribution from the strongly coupled channel. In each of the states 0 and 1, the current IS as well as the average bridge population (cid:104)nS(cid:105) in this channel, are assumed to be given by the standard Landauer theory for a channel comprising one single electron orbital of energy εS bridging the leads, dis- regarding the effect of electron-phonon interaction,39,40 1→0P1−k(R) 0→1P0−k(L) 1→0P1 = k(R) (cid:90) +∞ (cid:90) +∞ −∞ −∞ IS (V ; εS) = e (cid:104)nS (V ; εS)(cid:105) = dε 2π dε 2π S [fL(E) − fR(E)] ΓL S ΓR (ε − εS)2 + (ΓS/2)2 ΓL S fL(ε) + ΓR S fR(ε) (ε − εS)2 + (ΓS/2)2 (5) (6) S i.e. S S S +ΓR S = ε(0) S + U , ΓK(1) in state 0, and ε(1) S and where εS and ΓK where ΓS = ΓL S take the values ε(0) S , ΓK(0) = ΓK(0) in state 1. U is essentially a Coulomb energy term that measures the effect of electron occupation in chan- at the redox site, on the energy of the nel W , bridging orbital in channel S. ΓL S , εS, and U are model parameters. The average population and current in channel S are given by (cid:104)nS(cid:105) = P0(cid:104)nS(cid:105)(0) + P1(cid:104)nS(cid:105)(1); (cid:104)IS(cid:105) = P0I (0) ((cid:104)nS(cid:105)(0)) and I (1) ((cid:104)nS(cid:105)(1)) are the values of IS, Eq. (5) ((cid:104)nS(cid:105), Eq. (6)) in system states 0 (redox level empty), and 1 (redox level populated). Finally, the total current at a given voltage is I = IS + IW ≈ IS. S , where I (0) S + P1I (1) S , ΓR S S It should be noted that the rates defined by Eq. (4) are not completely specified, because the "redox energy level" εr is not known: it is equal to εW only if the capacitive interaction between the S and W channels is disregarded. To take this interaction into account, two models were examined in Ref. 27: Model A. The rates are written as weighted averages over the populations 0 and 1 of channel S with respective weights 1 − (cid:104)nS(cid:105) and (cid:104)nS(cid:105): (cid:16) (cid:16) 1 − (cid:104)nS(cid:105)(0)(cid:17) 1 − (cid:104)nS(cid:105)(1)(cid:17) k0→1 = k1→0 = 0→1 + (cid:104)nS(cid:105)(0)k(S0) k(S0) 1→0 0→1 + (cid:104)nS(cid:105)(1)k(S1) k(S1) 1→0 (7) 0→1, k(S1) where k(S0) 0→1 are the rates to occupy and vacate, respectively, the redox site when the fast channel is not occupied, while k(S1) 1→0 are the corresponding rates when this channel is occupied. The dependence of these rates on the occupation of the fast channel is derived from the dependence of εr in Eq. (4) on the occupation of level S: εr = εW when this level is not occupied, and 0→1, k(S1) εr = εW + U when it is. That is, k(K,S0) 0→1 =ΓK k(K,S0) 1→0 =ΓK k(K,S1) 0→1 =ΓK k(K,S1) 1→0 =ΓK r fK(εW ) r [1 − fK(εW )] r fK(εW + U ) r [1 − fK(εW + U )] (8) 4 (cid:17)(cid:104)nS(cid:105)(1) + ε(0) Here K = L, R. Model B. The rates are given by Eq. (4), with εr cal- culated as the difference between the energies of two molecular states, one with the redox level populated, and E1 = S (cid:104)nS(cid:105)(0): the other with the redox level empty, E0 = ε(0) S (cid:104)nS(cid:105)(0) (9) (cid:17) − ε(0) S (cid:104)nS(cid:105)(1) + ε(0) ε(1) 2 ε(0) S + U 2 = ε(1) S (cid:104)nS(cid:105)(1) + ε(0) 2 (cid:16) (cid:16) εr = These two models are associated with different physi- cal pictures that reflect different assumptions about rela- tive characteristic timescales. Model A assumes that the switching rates between states 0 and 1 follow the instan- taneous population in channel S, while model B assumes that these switching rates are sensitive only to the aver- age population (cid:104)nS(cid:105). Model B results from a standard Hartree approximation that would be valid if the elec- tronic dynamics in channel W is slow relative to that of channel S (see Appendix). From the discussion above it may appear at first glance to be the case, since transmis- sion through channel W is small, implying that the rates k0→1 and k1→0 are small. However, the electronic pro- cess that determines the timescale on which these rates change is not determined by the magnitude of these rates but by the response of the electrodes to changes in εr following changes in the bridge level population of the strongly coupled channel. This characteristic time (or times), τB, which is bounded below by the inverse elec- trode bandwidth, may depend also on temperature and the energy dependence of the spectral density, and can be shorter than the timescale of order of Γ−1 S on which population in channel S is changing (note that τB is van- ishingly short in the wide band limit). In this case model A would provide a better approximation. For compari- son, we also present below results for model C, in which the effect of the interaction between the two channels on the electron transfer kinetics in channel W is disregarded so that kK 0→1 =ΓK kK 1→0 =ΓK W fK(εW ) W [1 − fK(εW )] (10) while the current through channel S continues to be sen- sitive to the difference between states 0 and 1, as before. FIG. 2. (Color online) Current (panels a-c) and popula- tions of the channels (panels d-f). Results for the models A (panels a and d), B (panels b and e), and C (panels c and f) are shown for the channels S (dash-dotted line, red) and W (dotted line, blue), and compared to the PP-NEGF results for the same channels (solid, red and dashed, blue lines, re- spectively). Note, the PP-NEGF data is the same in panels a-c and d-f. See text for parameters. sical rate equations description of channel W . The exis- tence of capacitive coupling between the channels makes such mixed quantum-classical description potentially in- valid, since it misses quantum correlations between the two channels. To estimate the performance of these ap- proximations we shall compare them to a fully quantum calculation based on the pseudoparticle nonequilibrium Green function technique.36 In the PP-NEGF approach, a set of molecular many- body states, {N(cid:105)}, defines the set of pseudoparticles to be considered, so that one pseudoparticle represents each state. In particular, for the model (1) the molecular sub- space of the problem is represented by four many-body † states: N(cid:105) = nW , nS(cid:105), where nW,S = 0, 1. Let p N (pN ) be the creation (annihilation) operator for the state N(cid:105). These operators are assumed to satisfy the usual fermion or boson commutation relations depending on the type of the state. In our case the pseudoparticles associated with the states 1, 0(cid:105) and 0, 1(cid:105) are of Fermi type, while those corresponding to states 0, 0(cid:105) and 1, 1(cid:105) follow Bose statistics. The PP-NEGF is defined on the Keldysh con- tour as † GN1,N2(τ1, τ2) ≡ −i(cid:104)Tc pN1 (τ1) p N2 (τ2)(cid:105) (11) In the extended Hilbert space it satisfies the usual Dyson equation, thereby providing a standard machinery for their evaluation. Reduction to the physically relevant subspace of the total pseudoparticle Hilbert space is achieved by imposing the constraint (cid:88) † p N pN = 1 (12) III. THE PSEUDOPARTICLE GREEN FUNCTION METHOD N Models A and B above represent attempts to partly ac- count for the coupling between channels within the clas- on the Dyson equation projections. The resulting system of equations for the Green function projections has to be solved self-consistently (see e.g. Ref. 36 for details). 0.10.20124812I(a.u.)0.10.20120.10.20120.511.5Vsd(eV)00.20.4n0.511.5Vsd(eV)0.511.5Vsd(eV)(a)(b)(c)(d)(e)(f)5 (Color online) Same as Fig. 2 except ΓL W = 1.9 meV FIG. 4. and ΓR W = 0.1 meV. and has been suggested before4 -- 7 as a model for negative differential resistance in molecular junctions. Finally, in Fig. 5, the parameters are the same is in Fig. 2 except that T = 0 K. The voltage was changed by moving the Fermi level of the left electrode, keeping the right elec- trode static. The insets in the I/V plots show a closeup look at the contribution from channel W . The following observations are notable: (a) In comparison with the full quantum calculation, Model A performs considerably better than model B and, not surprisingly, than model C. The failure of model B is notable in view of the common practice to use the timescale separation as an argument for applying mean field theory in such calculations; however, as argued above, it follows from the use of the wide band limit for the electrodes in the calculations. (b) While model A seems to be quite successful in much of the voltage regime, it fails, as expected, near and around V = 0.3 V, the (bare) threshold to populate the W level. It is at this point of maximal fluctuations in the W popu- lation that electronic correlation is most pronounced, as this population is strongly correlated with that in S. (c) The deviation of the kinetic approximation from the full quantum result is considerably larger for the current and population of channel W (the redox site) than for channel S. This reflects the fact that the rates of charg- ing and discharging the redox site are sensitive to its correlation with the population on the strongly coupled level, while the dynamics of the latter responds most of the time just to the static population in W . Of course, these large deviations in the current carried by channel W have only an insignificant effect on the overall observed current. To see these important quantum correlation ef- fects one would need to monitor directly the electronic population of the redox site, which is possible in princi- ple using spectroscopy probes. (d) As a model for negative differential resistance (Fig. 4), model A performs qualitatively well, however the full cal- culation sets the NDR threshold considerably higher than that predicated by the approximate calculation. (e) As expected, the differences between the full quan- FIG. 3. (Color online) Same as Fig. 2 except U = 500 meV Finally, connections to Green functions of the standard NEGF formulation can be obtained by using relations between the electron operators in the molecular subspace of Eq. (1) and those of the pseudoparticles (cid:88) d† m = (cid:104)N1 d† † mN2(cid:105)p N1 pN2 (13) N1,N2 Thus the current through the junction can be obtained ei- ther by the usual NEGF expression,40 or within its pseu- doparticle analog.36 Results of calculations based on this procedure and on the kinetic schemes described in Section I are presented and discussed next. IV. RESULTS AND DISCUSSION In Figures 2 and 3 we compare results from the fully quantum calculation based on the PP-NEGF technique with those based on the kinetic approximations defined by models A-C of Section I. Panels (a), (b) and (c) in Fig. 2 show the current through channels S (red) and W (blue) as function of voltage, while the correspond- ing panels (d), (e) and (f) show, with the same color and line-forms codes, the electronic populations in these channels. The full and dashed lines in these plots corre- spond to the PP-NEGF calculations for channels S and W , respectively, and are identical in the panels (a-c) and in panels (d-f). The dash-dotted and dotted lines show results based on models A (panels (a) and (d)), B (panels (b) and (e)) and C (panels (c) and (f)). The parameters used in these calculations are EF = 0, T = 300 K, ΓL W = ΓR W = 1 meV, ΓL S = 100 meV, εS = 150 meV, εW = 300 meV, and U = 10 eV. For this choice of U states S and W cannot be populated simultaneously. The corresponding panels of Figs. 3 and 4 show similar results for the same choice of parameters, except that in Fig. 3 U is taken 500 meV while in Fig. 4 ΓL W = 1.9 meV and ΓR W = 0.1 meV (so ΓW = ΓL W = 2 meV as before). The latter choice designates level W as a blocking level - current goes down considerably when the voltage bias exceeds the threshold (300 meV) needed to populate it), S = ΓR W + ΓR 0.10.2012481216I(a.u.)0.10.20120.10.20120.511.5Vsd(eV)00.20.4n0.511.5Vsd(eV)0.511.5Vsd(eV)(a)(b)(c)(d)(e)(f)0.050124812I(a.u.)0.050120.050120.511.5Vsd(eV)00.40.8n0.511.5Vsd(eV)0.511.5Vsd(eV)(a)(b)(c)(d)(e)(f)6 pean Union's Seventh Framework Program (FP7/2007- 2013; ERC grant agreement no 226628). MG grate- fully aknowledges support by the Department of Energy (Early Career Award, DE-SC0006422) and the US-Israel Binational Science Foundation (grant no. 2008282). We thank Kristen Kaasbjerg for useful discussions. MG and AN thank the KITPC Beijing for hospitality and support during the time when this work was completed. Appendix A: Timescale considerations leading to the models A and B When it is reasonable to speak about rate of a channel, the formal expression for the W channel rate is?? t(cid:48) ds εr(s)V (t) C(t − t(cid:48)) V (t(cid:48)) (A1) (cid:90) t −∞ dt(cid:48) ei(cid:82) t where r is the position of the redox level, V (t) is the coupling between the channel W and the bath, and C(t− t(cid:48)) is the bath correlation time. At least two timescales have to be taken into account: one related to the dynamics of the redox level, εr(t), the other representing characteristic timescale of the bath. Note, that in general the bath is characterized by several timescales (e.g. the bandwidth of the metal, tempera- ture, and variation of spectral density). In our case the characteristic timescale for the dynamics of the level in the W channel is given by the rate of population change in the S channel. The latter is proportional to Γ−1 (Coulomb interaction is instantaneous). Let assume that the characteristic time of the bath is τB. The two ex- tremes are τB (cid:28) Γ−1 S . The former case corresponds to slow motion of the level relative to the bath dynamics, so that expression (A1) yields a set of rates (2 in our case) for different positions of the redox level. This corresponds to the model A of the paper. S and τB (cid:29) Γ−1 S The other extreme, τB (cid:29) Γ−1 S , corresponds to quick motion of the redox level position, which requires aver- aging of the exponential factor in (A1). This leads to appearance of a single rate, calculated at the average po- sition of the level, which is model B. 1M. Galperin, M. A. Ratner, and A. Nitzan, Nano Lett. 5, 125-130 (2005). 2M. Galperin, A. Nitzan, and M. A. Ratner, J. Phys.: Condens. Matter 20, 374107 (2008). 3M. Galperin, A. Nitzan, and M. A. Ratner, arXiv:0909.0915 (2009). 4M. H. Hettler, H. Schoeller, and W. Wenzel, Europhys. Lett. 57, 571 (2002). 5B. Muralidharan and S. Datta, Phys. Rev. B 76, 035432 (2007). 6R. Hartle and M. Thoss, Phys. Rev. B 83, 115414 (2011). 7A. Migliore and A. Nitzan, ACS Nano 5, 6669 (2011). 8K. F. Albrecht, H. Wang, L. Muhlbacher, M. Thoss, and A. Kom- nik, Phys. Rev. B 86, 081412 (2012). 9It should be emphasized that the theory addresses only locally stable states that will not persist beyond some finite lifetime, and do not imply multistability in the thermodynamic sense. 10M. Field, C. G. Smith, M. Pepper, D. A. Ritchie, J. E. F. Frost, G. A. C. Jones, and D. G. Hasko, Phys. Rev. Lett. 70, 1311 (1993). FIG. 5. (Color online) Same as Fig. 2 except T = 0 K. tum calculation and the results of model A become more pronounced at T = 0 K. While the results of model A display sharp threshold behavior, the full calculation is much less sensitive to temperature for the present choice of parameters because the width of the transition region is dominated by ΓS that is substantially greater than the thermal energy. V. CONCLUSION We have examined the electronic transport behavior of a generic junction model that comprises a bridge charac- terized by two interacting transport channels whose cou- plings to the leads are vastly different from each other. This is a model for a molecular redox junction and also for a point contact detector interacting with a weakly coupled nanodot bridge. We have compared approximate kinetic schemes for the dynamics of this junction to a full quantum calculation based on the pseudoparticle NEGF methodology. We found that a kinetic model in which the electron transfer rates in the weakly coupled chan- nel (redox site) respond instantaneously to occupation changes in the strongly coupled channel works relatively well in comparison with a mean field calculation. Still, this model fails quantitatively when the molecular level comes close to the electrochemical potential of the lead, reflecting the significance of electronic correlations in this voltage range. This paper has focused on the steady state current. Correlations between the two channels are expected to become considerably more pronounced in the noise prop- erties of such junctions and, most probably, would not be amenable to analysis using the kinetic approximation of model A. We defer this interesting issue to future work. ACKNOWLEDGMENTS The research of AN is supported by the Israel Science Foundation, the Israel-US Binational Science Founda- tion and the European Research Council under the Euro- 0.10.20124812I(a.u.)0.10.20120.10.20120.511.5Vsd(eV)00.20.4n0.511.5Vsd(eV)0.511.5Vsd(eV)(a)(b)(c)(d)(e)(f)7 11J. M. Elzerman, R. Hanson, L. H. W. v. Beveren, B. Witkamp, L. M. K. Vandersypen, and L. P. Kouwenhoven, Nature 430, 431 (2004). 12L. M. K. Vandersypen, J. M. Elzerman, R. N. Schouten, L. H. W. v. Beveren, R. Hanson, and L. P. Kouwenhoven, App. Phys. Lett. 85, 4394 (2004). 13R. Schleser, E. Ruh, T. Ihn, K. Ensslin, D. C. Driscoll, and A. 26K. Kaasbjerg and K. Flensberg, Phys. Rev. B 84, 115457 (2011). 27A. Migliore and A. Nitzan, to be published (2013). 28P. Coleman, Phys. Rev. B 29, 3035-3044 (1984). 29N. E. Bickers, Rev. Mod. Phys. 59, 845-939 (1987). 30M. Eckstein and P. Werner, Phys. Rev. B 82, 115115 (2010). 31J. H. Oh, D. Ahn, V. Bubanja, Phys. Rev. B 83, 205302 (2011). 32M. H. Hettler, J. Kroha and S. Hershfield, Phys. Rev. B 58, C. Gossard, Appl. Phys. Lett. 85, 2005 (2004). 5649-5664 (1998). 14T. Fujisawa, T. Hayashi, R. Tomita, and Y. Hirayama, Science 33T. Schauerte, J. Kroha, and P. Wolfle, Phys. Rev. B 62, 4394- 312, 1634 (2006). 15S. Gustavsson, R. Leturcq, B. Simovic, R. Schleser, T. Ihn, P. Studerus, K. Ensslin, D. C. Driscoll, and A. C. Gossard, Phys. Rev. Lett. 96, 076605 (2006). 16S. A. Gurvitz and Y. S. Prager, Phys. Rev. B 53, 15932 (1996). 17S. A. Gurvitz, Phys. Rev. B 56, 15215 (1997). 18G. Bulnes Cuetara, M. Esposito, and P. Gaspard, Phys. Rev. B 84, 165114 (2011). 19A. Carmi and Y. Oreg, Phys. Rev. B 85, 045325 (2012). 20G. Kiesslich, P. Samuelsson, A. Wacker, and E. Scholl, Phys. Rev. B 73, 033312 (2006). 21G. Kiesslich, E. Scholl, T. Brandes, F. Hohls, and R. J. Haug, Phys. Rev. Lett. 99, 206602 (2007). 22The "fast" channel carries all the current if the "slow" channel 4402 (2000). 34N. S. Wingreen and Y. Meir, Phys. Rev. B 49, 11040-11052 (1994). 35N. Sivan and N. S. Wingreen, Phys. Rev. B 54, 11622-1629 (1996). 36A. J. White and M. Galperin, Phys. Chem. Chem. Phys. 14, 13809-13819 (2012). 37A. J. White, B. D. Fainberg, and M. Galperin, J. Phys. Chem. Lett. 3, 2738-2743 (2012). 38Note that while in quantum mechanics damping rates and level widths are synonymous, it is exactly the energetic consequence of the finite lifetime, that is, the level broadening, which is dis- regarded in the kinetic approximation. 39S. Datta, Electric transport in Mesoscopic Systems (Cambridge is coupled only to one of the leads. University Press, Cambridge, 1995). 23A. Migliore, P. Schiff, and A. Nitzan, Phys. Chem. Chem. Phys. 40H. Haug and A.-P. Jauho, Quantum Kinetics in Transport and 14, 13746 (2012). 24B. Muralidharan and S. Datta, Phys. Rev. B 76, 035432 (2007). 25M. Leijnse, W. Sun, M. B. Nielsen, P. Hedegard, and K. Flens- berg, J. Chem. Phys. 134, 104107 (2011). Optics of Semiconductors (Springer, 2008). 41M. Esposito and M. Galperin, Phys. Rev. B 79, 205303 (2009).
1209.2037	2	1209	2012-12-13T09:35:14	Decoherence effects on weak value measurements in double quantum dots	[ "cond-mat.mes-hall", "quant-ph" ]	We study the effect of decoherence on a weak value measurement in a paradigm system consisting of a double quantum dot continuously measured by a quantum point contact. Fluctuations of the parameters controlling the dot state induce decoherence. We find that, for measurements longer than the decoherence time, weak values are always reduced within the range of the eigenvalues of the measured observable. For measurements at shorter time scales, the measured weak value strongly depends on the interplay between the decoherence dynamics of the system and the detector backaction. In particular, depending on the postselected state and the strength of the decoherence, a more frequent classical readout of the detector might lead to an enhancement of weak values.	cond-mat.mes-hall	cond-mat	Decoherence effects on weak value measurements in double quantum dots Dahlem Center for Complex Quantum Systems and Fachbereich Physik, Freie Universitat Berlin, 14195 Berlin, Germany Mark Thomas and Alessandro Romito (Dated: June 22, 2021) We study the effect of decoherence on a weak value measurement in a paradigm system consisting of a double quantum dot continuously measured by a quantum point contact. Fluctuations of the parameters controlling the dot state induce decoherence. We find that, for measurements longer than the decoherence time, weak values are always reduced within the range of the eigenvalues of the measured observable. For measurements at shorter time scales, the measured weak value strongly depends on the interplay between the decoherence dynamics of the system and the detector backaction. In particular, depending on the postselected state and the strength of the decoherence, a more frequent classical readout of the detector might lead to an enhancement of weak values. PACS numbers: 73.23.-b, 03.65.Ta, 03.65.Yz I. INTRODUCTION In quantum mechanics the measurement process is most simply described as a probabilistic event through the projection postulate.1 While it satisfactory describes several simple experimental configurations, some mea- surement protocols, including conditional quantum mea- surements, can lead to results that cannot be interpreted in terms of classical probabilities, due to the quantum correlations between measurements. A striking evidence of that is provided by the so-called weak values (WVs) obtained from the measurement scheme originally devel- oped by Aharonov, Albert, and Vaidman.2 The WV mea- surement protocol consists of (i) initializing the system in a certain state Ψ(cid:105) (preselection), (ii) weakly measur- ing an observable A of the system by coupling it to a detector, and (iii) retaining the detector output only if the system is eventually measured to be in a chosen final state Φ(cid:105) (postselection). The average signal monitored by the detector will then be proportional to the real part of the so-called WV Φ (cid:104) A(cid:105)weak The most surprising property of WVs is that they can be complex or negative2,3 whereas a strong conventional measurement would lead to positively definite values. Af- ter the original debate on the meaning and significance of weak values,4 -- 6 they have proven to be a successful con- cept in addressing fundamental problems and paradoxes of quantum mechanics,7,8 in accessing elusive quantities (e.g., the definition of the time a particle spends under a potential barrier in a tunneling process,9 the direct mea- surement of the wavefunction10), in defining measure- ments in counterintuitive situations (e.g., the simultane- ous measurement of two non-commuting observables7,11), as well as in generalizing the definition of measurement.12 By now a number of experiments in quantum optics has reported the experimental observation of WVs and its ap- plication to quantum paradoxes.13 -- 16 Recently, a series of interesting works has explored the potential of WVs mea- surement protocols for precision measurements. Weak- value-based measurement techniques have been success- fully employed in quantum optics experiments to access tiny effects13 and detect ultrasmall (subnanometric) dis- Ψ = (cid:104) Φ AΨ(cid:105)/(cid:104) ΦΨ(cid:105). placements.15,16 Parallel research has introduced the idea of weak values also in the context of solid state-state sys- tems.17 -- 19 Here, further works have shown that weak val- ues are related to the violation of classical inequalities in current correlation measurements,20 and a WV measure- ment technique for ultrasensitive charge detection has also been proposed.21 Due to the fact that WVs stem from quantum- mechanical correlations between two measurements they are expected to be particularly sensitive to decoherence. The effect of decoherence is important for WV ultrasensi- tive measurements where decoherence could suppress the amplification effect and become crucial in possible solid- state implementations, where it is known that decoher- ence plays a significant role in most systems. This is in fact the case for all the actual proposed implementations of WVs in solid state systems.17 -- 19,21,22 So far, the effect of decoherence on WVs has recently been addressed at a formal level showing how WVs are defined in a general open quantum system,3 while a quantitative evaluation of the effects of decoherence in a specific system exists only for WVs of spin qubits in a simple limit17,23 and for correlated spin measurements of (unpolarized) electronic currents.24 Therefore, a general characterization of the effects of decoherence within a weak value measurement in an open quantum system is, in addition to its theoret- ical significance, a relevant step in the direction of weak value implementation in condensed matter systems. In this work we precisely address this question. We approach the problem by considering the effect of deco- herence in a paradigm system, namely a quantum point contact (QPC) sensing the charge in a nearby double quantum dot.25,26 The model captures all the essential features of a continuous quantum measurement, corre- sponding to the typical measurement schemes of quan- tum states in nanoscale solid-state systems (which is the case for all the above-mentioned proposals), and allows us to fully describe the interplay between the detector backaction and the decoherence process. The key features of our analysis and the main results are as follows. We describe the double quantum dot as a two-level system, L(cid:105), R(cid:105), corresponding to the electron being in the left or right dot, respectively. In the sys- tem dynamics, we introduce fluctuations of the system's parameter, for example the gate voltages, that suppress the quantum mechanical oscillations between these two states, at the decoherence time, 1/γ. The QPC detector, while distinguishing between the two system states, af- fects the system dynamics on the scale of the backaction time, τD. As long as the measurement duration, τ , is shorter than τD, the measurement is weak and can lead to WVs upon proper postselection. The postselection, obtained, for instance by a second detector, is effectively described as a projective measurement on a specific state, ψf(cid:105). The quantity of interest in such a WV protocol is the detector signal conditional to a positive postselection, with a particular attention to the apparence of peculiar weak values, that is, WVs which lie beyond the range of eigenvalues of the measured observable. By taking ad- vantage of the Bayesian formalism, which allows us to consider the correlations for single shot measurements, we obtain a general expression for the WV in terms of the system state only [cf. Eqs. (16) and (17)]. We identify two different regimes depending on whether the detector readouts (i.e., relaxation processes) are slow or fast compared to the duration of the mea- surement. In both regimes, measurements longer than the decoherence time lead to the WVs bounded by the eigenvalue spectrum. In the former case, dubbed co- herent detection, we show that the weak value is exclu- sively determined by the system dynamics undergoing decoherence -- cf. Eqs. (26) and Figs. 3 and 4. In the latter case, named continuous readout, instead, even at time scales τ (cid:28) τD, the WVs are affected by the inter- play of the detector and the decoherence dynamics [cf. Eqs. (30( and (31) and Figs. 5 and 6]. In the coher- ent detection regime, the WV is sensitive to the average quantum-coherent correlation between measurement and postselection [Eq. (27)], vanishing for long-time measure- ments. The frequent projection in the continuous regime freezes the postselection, leading to a finite WV for long measurements. This difference reflects at shorter time scales, where a continuous detection can enhance the cor- responding WV obtained for a coherent detection. In particular, depending on the various parameters, for ex- ample orientation of the postselection, it could give rise to WVs beyond the range of eigenvalues, whereas a co- herent detection would not (cf. Fig. 6). The paper is primarily separated into four parts. In Sec. II we present the model and its description in terms of the Bayesian formalism. Hereafter, Sec. III presents a general expression for the weak value in terms of the density matrix of the combined qubit-detector system. In Sec. IV we discuss the system dynamics of the two regimes of "coherent detection" (Sec. IV A) and continu- ous readout (Sec. IV B). In the former regime, the detec- tor and system dynamics are decoupled from each other in the weak measurement regime; in the latter regime, we show that the detector induced decoherence and the 2 intrinsic system decoherence act together in affecting the measurement outcome. Section V summarizes our re- sults. II. THE MODEL The system under study is illustrated in Fig. 1. It consists of a double quantum dot with an adjacent QPC, which serves as a detector continuously measuring the charge state in the double dot. We consider the case where, due to charging effects, the double dot can host only one electron in two orbital lev- els, L(cid:105) and R(cid:105) corresponding to the lowest orbital level for the left and right dot, respectively. In this case the double dot can be thought as a charge qubit. Through- out the paper we refer to the "dot plus QPC" as the combined qubit-detector system which is then described by the Hamiltonian H(t) = H0(t) + Hint + H, (1) where H0(t) denotes the Hamiltonian of the double dot, Hint characterizes the system-detector interaction and H describes the Hamiltonian of the detector. Specifically H0(t) = (t) σz + ∆(t) σx H + Hint = ∆ H + + ∆ H− ⊗ σz (cid:17) (2) (3) 1 2 (cid:16) (cid:88) l where ∆ H + = (cid:88) (cid:88) r (cid:16) Era† rar + † l al + Ela (cid:88) (cid:16) r,l Ωlra (cid:17) † l ar + h.c. (4) (5) (cid:17) ∆ H− =2 δΩlra † l ar + h.c. lr Here σz = R(cid:105)(cid:104) R − L(cid:105)(cid:104) L, σx = R(cid:105)(cid:104) L + L(cid:105)(cid:104) R and R(cid:105) , L(cid:105) are, without loss of generality, the eigen- states of the operator measured by the QPC.27 The on- site energy difference, 2(t), and the tunneling between such states, ∆(t), are controlled by the externally applied voltage biases, VL,VR, Vh. In Eqs. (3) and (5), the oper- ators in the QPC space are indicated by a , and we do so henceforth. The creation and annihilation operator † in the two leads of the QPC are denoted by a i and ai respectively, where i = l, r refer to the left and the right reservoir, respectively. Ei characterizes the energy of the reservoir states, which are maintained at the correspond- ing Fermi energies µL = µR + eVd > µR, and Ωl,r ± δΩl,r denotes the tunneling amplitude between the reservoirs when R(cid:105) or L(cid:105) is occupied. This model has been extensively studied in the liter- ature.25,28 -- 30 The current through the detector directly measures the "position" of the electron in the dot. Un- der the assumptions of uniform tunneling matrix ele- ments, δΩl,r ≡ δΩ, Ωl,r ≡ Ω, and density of states 3 FIG. 1. (a) Scheme of a double quantum dot with a nearby QPC measuring its charge states, labeled as L(cid:105), R(cid:105). (b), (c) Scheme of the detection mechanism. The conductance of the detector is sensitive to the qubit's occupancy: The tunneling amplitude in the QPC is Ω+δΩ when the right dot is occupied (b) and Ω − δΩ when the left dot is occupied (c). FIG. 2. (a) Qubit plus detector interaction with a classical pointer reading out the number mk of electrons having pen- etrated to the detector's right reservoir at certain discrete times t = tk. Illustration of the continuous readout (b) and coherent (c) measurement regimes, respectively. in the QPC's reservoirs, ρL, ρR, the average current for an electron being in the left or right dot, reads (cid:104)I(cid:105)± = eD/2 · (1 ± δΩ/Ω)2, where D ≡ T Vd/(2π) = 2πρlρrΩ2Vd. Throughout the work we set = 1. The QPC's shot noise power SI = eI(1−T ) sets the time scale τD ∼ SI /(I+− I−)2 ≈ 1/D(Ω/δΩ)2(1− T ) needed to dis- tinguish the detector's signal form the background noise. The weak measurement regime is then identified by mea- surements of duration τ (cid:28) τD, which can be controlled in principle by tuning the system-detector coupling, the duration of the measurement, or the voltage bias across the QPC. The QPC is effectively a detector at finite Vd where it leads to a finite signal (cid:104)I(cid:105) ∝ Vd. In particular, we assume Vd (cid:29) kBT (the temperature T is the small- est energy scale throughout the paper) and Vd (cid:29) ρΩ2 as clarified below. In this regime we neglect the extra deco- herence effect due to the detector equilibrium backaction, for instance orthogonality catastrophe dephasing.31,32 An important aspect of the model is that the Hamil- tonian in Eq. (1) describes the qubit and the detector as a closed quantum system. However, the QPC is contin- uously converting the information about the state of the system in a classical -- macroscopic -- information out- put which is the current. In other words, while Eq. (1) will evolve the detector to a coherent superposition of states with different charges in the reservoirs, the classi- cal knowledge of the current would correspond to a well- defined number of electrons in each reservoir. A solution of this problem, as pointed out in Ref. 30, consists of introducing a macroscopic pointer which interacts with the detector. The pointer provides in fact an effective description of the various relaxation processes confining the QPC electrons in one of the two QPC reservoirs. The pointer can be modeled to interact instantaneously at certain times, t1, t2 . . . tk . . . tN with t0 = 0 and tN = τ , and reads out the change of the number of electrons in the right reservoir, µR, as schematically depicted in Fig. 2. At any time t = tk the pointer reads the number of elec- trons, mk, transmitted to the collector within the time interval ∆tk = tk+1−tk, and collapses the qubit-detector system onto a corresponding state depending on the mea- sured value mk. Note that the introduction of the pointer results in a new time scale ∆tk in the problem which is a free parameter in our model. We discuss in the Results section the different regimes corresponding to the relation of this time scale to other time scales in the problem. The model allows us to include decoherence by consid- ering fluctuations, which naturally arise in the system, of the external parameters, namely the voltage biases VL,VR, Vh. In turn they lead to fluctuations of the dots' parameters, (t) = 0 + ξ(t) and ∆(t) = ∆0 + η(t) around their average values 0 and ∆0. Since the ex- plicit relation between Vh, VL, VR and 0, ∆0 depends on microscopic details, we effectively assume here a gen- eral linear relation and the effect of decoherence is there- fore described by replacing the dot's Hamiltonian with H0(t) → H0 + Hξ(t), with Here ω = (cid:112)2 0 + ∆2 0 is the oscillation frequency of the system and m defines the corresponding eigenstates. To this regard, the index i labels the different independent decoherence sources, ideally corresponding to the inde- pendent voltage sources. For each of them ki indicates the direction of the fluctuations with ki = 1 and ξi(t) is assumed to be a Gaussian white noise, that is, (cid:104) ξi(s)(cid:105)ξ = 0, (cid:104) ξi(s1) ξj(s2)(cid:105)ξ = γi δ(s1 − s2) δij (8) where γi describes the strength of the correlation func- tion. For the sake of simplicity we present in the follow- H0 = 0 σz + ∆0 σx = ωm · σ, Hξ(t) = ξi(t) ki · σ. (cid:88) i (6) (7) L>R>VRVhVL(a)V∆(t)Ω+δΩμLμRεL(t)ε(t)R(b)εL(t)εR(t)∆(t)μLμRΩ-δΩ(c)dεL(t)Δ(t)μLΩ + δΩpointer"qubit plus detector" R(t)mkread-out:classical informationcollapseεR(t)μRtm1read-outm2read-outm3read-outread-out+ post-selectiontm1read-out+ post-selectiont1t2t3tNt0t0tN = t1 (a)(b)(c)ing our general results for the case of a single decoherence source. The results in the case of several noise sources are a straightforward extension and are discussed at the end. Finally, we can include in the model the description of postselection, as required by the WV measurement pro- tocol. As pointed out in Ref. 17, a second QPC strongly measuring the charge on the dot at any time after the weak measurement can effectively realize a postselection in any given qubit state, Ψf(cid:105). Without loss of gener- ality we therefore consider the situation where the post- selection takes place immediately after the weak mea- surement. Within our model the postselection into the state Ψf(cid:105) is described by the action of the correspond- ing projection operator Πf acting at the end of the weak measurement. III. GENERAL EXPRESSION FOR THE WEAK VALUE IN THE PRESENCE OF DECOHERENCE The WV protocol we are interested in consists of preparing the double dot in a given initial state Ψ(t0)(cid:105) at time t0, making the system interact with the detec- tor for a time τ , and finally strongly measuring it in the postselected state in Ψf(cid:105). The quantity of interest is the WV of the electron's occupancy in the double dot, that is, the value of σz(τ ) conditional to a positive post- selected outcome, which we denote by Πf(cid:104)σz(τ )(cid:105)Ψ(t0). In fact, such a quantity is inferred from the postselected output of the detector, that is, the average current in the QPC conditional to the postselection Πf(cid:104)I(τ )(cid:105)Ψ(t0) = e/τ Πf(cid:104)n(τ )(cid:105)Ψ(t0) through Πf(cid:104)σz(τ )(cid:105)Ψ(t0) = Πf(cid:104)I(τ )(cid:105)Ψ(t0) − eD/2h I+ − I− . (9) (cid:88) The average conditional (postselected) value of the num- ber of electrons, Πf(cid:104)n(τ )(cid:105)Ψ(t0), having passed through the QPC during the measurement time τ is (cid:104)n(τ )(cid:105)Ψ(t0) = Πf m P (m Πf ) . (10) m In Eq. (10) m indicates the total number of electrons having reached the collector and P (m Πf ) is the condi- tional probability that m electrons have been transmit- ted through the QPC given that the qubit is finally found to be in a state represented by the projection operator Πf . Note that Eq. (10) is valid for any strength of the measurement. We keep our analysis valid for a general measurement strength until specified differently. The conditional probabilities in Eq. (10), can be di- rectly expressed in terms of the total density matrix R(t) of the qubit-detector system. Following the for- malism of Gurvitz25 and Korotkov,30 a pure state of the qubit-detector system is described by a wave function Ψ(t)(cid:105) = ( Ψ↑(t)(cid:105), Ψ↓(t)(cid:105)), where σ =↑,↓ labels the 4 eigenstates of σz and Ψσ(t)(cid:105) is a many-body state of the QPC, b(0) σ(t) + (cid:88) l,l(cid:48)≤µl;r,r(cid:48)>µr Ψσ(t)(cid:105) = + (cid:88) l≤µl;r>µr b(1) σ lr (t) a† ral † r(cid:48)alal(cid:48) + ··· ll(cid:48)rr(cid:48)(t) a† b(2) σ ra  0(cid:105) , (11) and b(i)(t)2 describes the probability of finding the en- tire qubit-detector system in the corresponding state de- scribed by the creation and annihilation operators with l, r labeling the single particle states in the left, right reser- voirs, respectively. The corresponding qubit-detector density matrix R(t) has components  .  R(0,0) σσ(cid:48) (t) R(1,0) σσ(cid:48) (t) R(1,1) ... R(0,1) ... σσ(cid:48) (t) . . . . . . σσ(cid:48) (t) Rσσ(cid:48)(t) = R(m,n) σσ(cid:48) (t) . . . (cid:110)R(m)(t) (cid:111) by are (t) b (t) by is the labeled Each of (n) σ(cid:48) l1...lnr1...rn each entry R(m,n) (12) a matrix whose infinitely many and entries σσ(cid:48) l, r, l(cid:48), r(cid:48), l(cid:48)(cid:48), r(cid:48)(cid:48) . . . l(m), r(m) Here dimensions given states l, r, l(cid:48), r(cid:48), l(cid:48)(cid:48), r(cid:48)(cid:48) . . . l(n), r(n). b(m) σ l1...lmr1...rm complex conjugate) in R(m,n) (t) characterizes the co- herences between all the possible states with m and n electrons detected in the collector of the QPC at time t. In particular, the trace of each diagonal matrix R(m,m) (t) identifies the probability that exactly m electrons have passed through the detector until time t, namely indicates (the the the (t) σσ(cid:48) σσ(cid:48) · (cid:110) (cid:110) (cid:111)(cid:111) P (m) = trsys = trsys trdet R(m,m)(t) , σσ(cid:48) quantity R(m) (13) σσ(cid:48) (t) = introduced the }. Also the reduced density matrix of where we trdet{R(m,m) the dot, ρ(t) = trdet {R(t)}, can be written as P (m) · ρ(m)(t) , where ρ(m)(t) = R(m)(t)/(tr(cid:8)R(m)(t)(cid:9)) describes the R(m) = (cid:88) (cid:88) ρ(t) = (14) m m state of the qubit where m electrons have reached the collector. Besides the inherent quantum-mechanical fluctuations, the stochastic parameter ξ(t) assumes different values at each replica of the experiment according to its proba- bility distribution. In order to properly take into ac- count the average over fluctuations, we can first rewrite the conditional probability Eq. (10) using Bayes' the- m P (Πf ). The WV of each run of the experiment is now weighted with orem as P (mΠf ) = P (m)P (Πfm)/(cid:80) where ∂t = ∂/∂t. Here, ∆ H + + ∆ · σx + 1 2 M = N = σ · k . (cid:17) ⊗ σz , ∆ H− + 2 0 (cid:16) 1 2 5 (19) Introducing an interaction picture with respect to M , which transforms the arbitrary operator A as AI (t) = e−iM t · A · eiM t, transfers Eq. (18) into ∂tRI (t) = −i ξ(t) [ NI (t), RI (t) ]. Iteratively solving this equation, one obtains after taking the average with respect to ξ(t) 0 (cid:90) t (cid:90) s1 ds1(cid:104)ξ(s1)(cid:105)ξ [NI (s1),(cid:104)RI (0)(cid:105)ξ] (cid:90) s1 ds2(cid:104)ξ(s1)ξ(s2)(cid:105)ξ 0 (cid:90) t (cid:104)RI (t)(cid:105)ξ = (cid:104)R(0)(cid:105)ξ − i (cid:90) t + (−i)2 0 ds1 + (−i)3 (cid:90) s2 · [NI (s1), [NI (s2),(cid:104)RI (0)(cid:105)ξ]] ds1 ds2 ds3(cid:104)ξ(s1)ξ(s2)ξ(s3)(cid:105)ξ 0 0 · [NI (s1), [NI (s2), [NI (s3),(cid:104)RI (0)(cid:105)ξ]]] 0 + . . . (20) Due to the δ-like time correlations in Eq. (8), the average can be performed exactly order by order. The so obtained integral equation for R(t) is more conveniently written in a differential form as ∂t (cid:104)R(t)(cid:105)ξ = − γ 2 (cid:104) N, (cid:104)R(t)(cid:105)ξ (cid:105)(cid:105) − i M,(cid:104)R(t)(cid:105)ξ (21) The density matrix is expressed as a linear combination of Pauli-matrices of the dot's space (cid:104) (cid:104) (cid:105) N, 3(cid:88) j=0 the probability of a successful postselection in the corre- sponding run of the experiment, which also depends on the specific noise realization. This means that the aver- age over the fluctuations ξ(t) in the weak value is properly taken into account by separately averaging over ξ both the conditional average value of m and the postselection probability.23 This leads to (cid:104)n(τ )(cid:105)Ψ(t0) = Πf m m P (Πf (mξ)) P (mξ)(cid:105)ξ m P (Πf (mξ)) P (mξ)(cid:105)ξ . (15) (cid:104)(cid:80) (cid:104)(cid:80) Identifying the emerging probabilities in terms of the ma- trix R(m) similar to Eq. (13) yields the conditional cur- rent m m tr(cid:8)Πf · (cid:104)R(m)(τ )(cid:105)ξ (cid:9) (cid:80) m tr(cid:8)Πf · (cid:104)R(m)(τ )(cid:105)ξ (cid:9) . (cid:80) (cid:104)I(τ )(cid:105)Ψ(t0) = Πf e τ As already pointed out the weak measurement regime is obtained when (δΩ/Ω)2τ (cid:28) 1/D(1 − D/Vd). The WV can be identified from the coefficients in the expansion (cid:104)I(τ )(cid:105)Ψ(t0) = (cid:104)I(τ )(cid:105)+ Πf Πf (cid:104)I(τ )(cid:105)weak Ψ(t0)· δΩ Ω +O (17) We discuss the validity of this expansion later. From the definition of the weak measurement regime, we expect to be sensitive to coherence effects for time scales τ (cid:28) 1/D(1 − D/Vd)(Ω/δΩ)2. The WV is com- pletely determined by the knowledge of the probabilities P (m) and the conditional reduced dot's density matrix (cid:104)ρ(m)(τ )(cid:105)ξ. Further analysis now focuses on the evalua- tion of (cid:104)R(m)(τ )(cid:105)ξ. (16) (cid:18) δΩ2 (cid:19) . Ω2 IV. SYSTEM-DETECTOR DYNAMICS IN PRESENCE OF DECOHERENCE (cid:104)R(t)(cid:105)ξ = 1 2 vj(t) · σj. (22) A. Coherent detection First we consider the case where a single readout by the external pointer takes place at the end of the measure- ment process, that is, at t = τ , immediately followed by the postselection. In this case the coherent evolution of the system and detector is not disturbed by the pointer, and we thus name it coherent detection. The WV is fully determined by the knowledge of the averaged matrices (cid:104)R(m)(τ )(cid:105)ξ. We can derive a differen- tial equation for (cid:104)R(m)(τ )(cid:105)ξ starting from the von Neu- mann equation for the qubit-detector system, i∂tR(t) = [H(t), R(t)]. After inserting the Hamiltonian in Eq. (1) with the specific choice of a single fluctuation source, that is, (t) = 0 + kz · ξ(t), ∆(t) = ∆0 + kx · ξ(t), one obtains for the qubit-detector evolution ∂tR(t) = −i [ M, R(t) ] − i ξ(t) [ N, R(t) ] (18) where each vj(t) is a matrix in the QPC Hilbert space with tr{v0(t)} ≡ 1. Substituting Eq. (22) into the aver- aged von Neumann equation (21) yields a set of differen- tial equations for vj(t) which fully describes the averaged evolution of the qubit-detector system: (cid:104) (cid:104) (cid:18) (cid:104) (cid:18) (cid:18) (cid:104) (cid:105) − i (cid:110) (cid:105) − 1 2 2 ∆ H +, vx ∆ H−, vy ∆ H−, vz(t) ∆ H +, v0(t) (cid:111) x − 1)vx + kxky vy + kxkz vz (k2 (cid:110) y − 1)vy + kykz vz 1 2 (cid:105) z − 1)vz kxkz vx + kykz vy + (k2 kxky vx + (k2 ∆ H +, vy (cid:104) (cid:105) (cid:104) + (cid:105) − i ∆ H−, v0 ∆ H +, vz . (cid:105) (cid:19) (cid:19) (cid:19) 2 − 2∆vz − i 2 v0 = − i 2 vx = 2γ − i 2 vy = 2γ vz = 2γ − i 2 − 20vy + 20vy (cid:111) ∆ H−, vy + 2∆vy (23) Equations. (23) describe a set of infinitely many coupled differential equations. They are a generalization to a den- sity matrix of the results obtained in Ref. 25 for the sim- ple case of a pure state. Note that such a generalization is essential in our case to properly treat fluctuations of ξ. One may further note that the fluctuations treated here differ from Ref. 30, where the system's evolution for a given stochastic measurement output I(t) is stud- ied. Considering the matrix elements of R(t) between two sectors (m, n) with m and n electrons having passed through the QPC, we trace out the QPC degrees of free- dom to obtain a differential equation for (cid:16) (cid:17) v(m)(t) · σ , (24) j j (t)}. The details of the cal- where v(m) culation are presented in appendix A. Here we report the resulting differential equation, which can be recast as a differential equation for v(m) ∂t v(n)(t) = (G0 + Gk + G1) · v(n)(t) + G2 · v(n−1)(t) , (25) (t), that is, j with (cid:104)R(m)(t)(cid:105)ξ = 1 2 (t) = trdet{v(m,m)  ,  , 0 0 0 0 0 −2 0 0 − 2 ∆0 2 0 2 ∆0 0 0 0 0 0 x − 1 kx ky k2 kx kz y − 1 ky kz kx ky k2 z − 1 ky kz k2 kx kz  ,  . 0 1 + δΩ2 Ω2 0 0 0 0 1 + δΩ2 Ω2 0 2 δΩ Ω 0 0 1 + δΩ2 Ω2 0 1 − δΩ2 Ω2 0 0 0 0 1 − δΩ2 Ω2 0 2 δΩ Ω 0 0 1 + δΩ2 Ω2 Gk = 2 γ G0 = 0 0 0 0 0 0  0  0  1 + δΩ2  1 + δΩ2 0 0 2 δΩ Ω 0 0 2 δΩ Ω Ω2 Ω2 G1 = − D 2 G2 = D 2 The above equation is obtained to lowest order in ρΩ2 (cid:28) Vd, under the assumptions that: (i) the detector's tran- sition amplitude only weakly depends on the energies, that is, Ωlr ≡ Ω; (ii) the densities of states in the QPC's collectors are constant, ρl(Elk ) ≡ ρl and ρr(Erk ) ≡ ρr; (iii) at t = 0 the energy levels of the detector are filled up to the Fermi-level so that m = 0, that is, v(n)(0) = (1, vx, vy, vz) · δn,0. As evident in Eq. (25), for D → 0 or δΩ → 0, respectively, the system evolves undisturbed, while for γ → 0 our result reduces to that in Ref. 25. 6 According to Eqs. (16) and (17) we can obtain the expression for the WV by solving the differential equa- tions (25) perturbatively in the regime δΩ (cid:28) Ω. The details of the derivation are given in Appendix B. We highlight here that the perturbative solution is a valid approximation for τ (cid:28) 1/D(Ω/δΩ)2 exactly correspond- ing to the weak measurement regime. This finally yields the expression for the WV (cid:82) τ (cid:82) τ (cid:104)I(τ )(cid:105)weak Ψ(t0) = e D Πf 1 τ 0 ds vz(s) + 1 n · v(τ ) τ 0 ds nz(s) . (26) Here τ is the duration time of the weak measurement and we effectively introduced a time dependence in the postselection operator Πf (τ ) = 1/2 (n(τ )·σ), n = n(τ ) is then the postselected state at time t = τ instantaneously after the measurement. In the notation in Eq. (26), v(s) is defined by v(s) = exp [(G0 + Gk) s]·v(0), while n(s) = exp [(−G0 + Gk) s] · n(0). The result in Eq. (26) is already captured by a minimal model where the coupling to the detector is described by a von Neumann Hamiltonian1 Hint(t) = λg(t) p σz. It lin- early couples the measured observable σz to a detector's variable p, λ indicates the coupling constant and the time dependency of the interaction is included in the function g(t). The WV of σz is inferred from the conditional value of the conjugate variable q, (cid:104)q(τ )(cid:105)weak Ψ(t0) = λ Re Πf (cid:104)Πf (τ ) σz(s)(cid:105) (cid:104)Πf (τ )(cid:105) ds (27) (cid:90) τ 0 obtained to leading order in the coupling. The effect of decoherence is included in the correlation function (cid:104)Πf (τ ) σz(s)(cid:105) resulting in Eq. (26). Equation (27) eluci- dates the role of coherent system evolution between the measurement at time s and the postselection at time τ . B. Continuous readout Opposed to a coherent detection, the regime of con- tinuous readout is characterized by the detector's state being frequently read out by the pointer. This limit is described by a sequence of readouts at times t = tk, k = 1 . . . N , where the time interval between readouts ∆tk := tk+1 − tk ≡ ∆t is the shortest time scale in the problem, that is, ∆t (cid:28) min{1/ω, 1/γ, 1/D}. The condi- tional number of transmitted electrons Eq. (16) can now be expressed as the sum over all permutations describing quantum jumps at all possible times (cid:80)(cid:80) (cid:80)(cid:80) j mj =m j mj =m m tr(cid:8)Πf · (cid:104)R(m1,...,mN )(τ )(cid:105)ξ (cid:9) (cid:9) . tr(cid:8)Πf · (cid:104)R(m1,...,mN )(τ )(cid:105)ξ (cid:104)n(τ )(cid:105)Ψ(t0) = Πf (28) where j = 1 . . . N and R(m1,...,mN )(τ ) characterizes the qubit's density matrix weighted with the probability that exactly mk electrons have passed within each time inter- val ∆tk. Each readout corresponds to a collapse of the qubit-detector system in the sector of mk electrons hav- ing passed with the corresponding probability P (mk; tk). In the regime ∆t (cid:28) min{1/ω, 1/γ, 1/D} at most one electron penetrates through the QPC between two sub- sequent readouts.33 The probabilities that either exactly one (quantum jump30,34) or zero electrons accumulate in the collector within a readout period time are com- puted in appendix C. They are given by P (0; ∆t) = {A · v(nk)(tk)}0 and P (1; ∆t) = {B · v(nk)(tk)}0, respec- tively, where the index {. . .}0 denotes the zeroth compo- nent and A = 1(4) + (G0 + Gk + G1) ∆t + O(cid:0)∆t2(cid:1) B = G2∆t + O(cid:0)∆t2(cid:1) . indicates higher order (29) O(∆t2) for ∆t (cid:28) In the limit of N → ∞, ∆t → 0 min{1/ω, 1/γ, 1/D}. while keeping N · ∆t = τ constant, the conditional num- ber of transferred electrons in Eq. (28) can be analytically evaluated (cf. appendix C), yielding the exact result n · G2 · e(G0+Gk+G1+G2) τ · v(0) terms Πf n · e(G0+Gk+G1+G2) τ · v(0) The WV can be easily extracted by Πf(cid:104)I(τ )(cid:105)weak limδΩ/Ω→0(Πf(cid:104)I(τ )(cid:105)Ψ(t0)), where (cid:104)n(τ )(cid:105)Ψ(t0) = τ (30) Ψ(t0) = . (cid:104)I(τ )(cid:105)weak Ψ(t0) = eΩ δΩ · Πf(cid:104)n(τ )(cid:105)Ψ(t0) − D 2 +O Πf τ Ω2 (31) The simultaneous presence of G1 +G2 and Gk in Eq. (30) defines a new time scale 1/γ (cid:39) 1/D(Ω/δΩ) which de- scribes the extra source of decoherence emerging from the detector itself. Note also that Eqs. (30) and (31) are valid at any time. The derivation indeed relies on the perturbative solution in Appendix B, which is valid in the limit ∆t → 0; hence, the exact composition of subse- quent evolution between different readouts holds at any time. (cid:18) 1 (cid:19) (cid:18) δΩ2 (cid:19) V. RESULTS Equations. (26), (30) and (31) represent the main re- sults of our paper. They express the WVs in the two limiting cases of coherent detection and continuous read- out in terms of the system dynamics. Generally, they give rise to four different time scales, which characterize (i) the system's dynamics, 1/ω, (ii) the decoherence, 1/γ, (iii) the detector dynamics, 1/D, and (iv) the backaction, 1/D · (Ω/δΩ)2. We realistically assume 1/D to be the shortest time scale in our problem and focus on τ (cid:29) 1/D. The effects of the detector at this time scale dominated by the Zeno effect,26 though inherent in Eq. (25), do not play a significant role at the larger time scales of interest where 7 (a) (b) (c) (d) FIG. 3. Schematic evolution of v(s) and n(s) on the Bloch- sphere for (a) m (cid:54)= k and (b) m = k. (c) WV as a function of τ for different γ; (d) WV for different γ at τ1 = 0.10 ns and τ2 = 1.31 ns, marked in panel (c). In all plots the parameters are chosen as 0 = 20 µeV, ∆0 = 3 µeV, k ≈ −0.89 ex + 0.45 ez, v(0) = (0.5,−0.33, 0.80) and n = (−0.20, 0.75, 0.63). decoherence takes place. Consistently with our pertur- bative analysis, we further consider τ (cid:28) 1/D · (Ω/δΩ)2. The WV can be visualized via the motion of v(t) and n(t) on the Bloch-sphere for v = (vx, vy, vz) and n = (nx, ny, nz), as depicted in Fig. 3. v(t) ≤ 1 charac- terizes the coherence of the qubit's state, which is initially prepared in a pure quantum state with v(t = 0) = 1. We start with analyzing the case of coherent detection, that is, the dynamical evolution obtained in Eq. (26). In this regime the effect of decoherence essentially reduces to the dynamics of the qubit alone in the presence of decoherence. Accordingly, the WV presents different be- haviors in different regimes defined by the durations of the measurement, τ , compared to the remaining time scales 1/ω, 1/γ. For long measurement durations τ (cid:29) 1/γ, the qubit's state generally relaxes towards a fully statistical mixture, v(t) = 0, as shown in Fig. 3(c), except for the special case (cid:126)k (cid:107) (cid:126)m, which is discussed later. Consequently, at time scales where decoherence effects come to play we Ψ(t0) ≤ eD and, hence, the peculiar obtain max Πf (cid:104)I(τ )(cid:105)weak characteristics of WVs are washed out. For measurements shorter than the decoherence time, the results depend on the system's evolution time scale. For short measurements, τ (cid:28) 1/ω, which correspond to negligible dynamics and fluctuations, the vectors v(s) and n(s) in Eq. (26) are constant so that a measure- ment of the averaged detector's response trivially re- flects the WV of the observable σz independent from the measurement duration time τ , Πf Ψ(t0) = Re((cid:104) f σz Ψ(cid:105) /(cid:104) f Ψ(cid:105) ). (cid:104)σz(τ → 0)(cid:105)weak exeyezv(s)mn(s)v (s)zn (s)zkexeyezveffm = kn(s)neffv (s)zn (s)zv(s)Τ1Τ24ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakΓ(cid:61)1.6ΜeVΓ(cid:61)1.1ΜeVΓ(cid:61)0.6ΜeVΓ(cid:61)0.1ΜeV0.4ns(cid:45)10.8ns(cid:45)11.2ns(cid:45)1Γ(cid:209)1.20.80.4(cid:80)f(cid:60)ΣzΤj(cid:62)(cid:89)t0weakΤ2(cid:61)1.31nsΤ1(cid:61)0.10ns8 (a) (b) (a) (b) (c) FIG. 4. WV as a function of τ for (a) different k (θ = 0 corresponds to k (cid:107) m) and (b) different n (φ = 0 corresponds to n (cid:107) m). (c) Same as in (a) in the presence of multiple decoherence sources, where γ = 1.0 µeV, k = 1√ ez, 2 k1 = 1√ µeV, k1 = m, k2 = m⊥, γ1 = (k · m) µeV, γ2 = (k · m⊥) µeV; all the other parameters are as in Fig. 3. µeV γ2 = 1√ ex, k2 = 1√ ez, γ1 = 1√ ex + 1√ 2 2 2 2 2 The system's dynamics for intermediate durations of the measurement, τ (cid:29) 1/ω, however, are appreciable. Here, both vectors n(s) and v(s) precess about the eigen- vector m of G0 with a frequency of 2 ω. Note that, due to its backward-in-time evolution, n(s) precesses in the opposite direction as compared to v(s). Due to the oscil- latory dependence of the denominator in Eq. (26), pecu- liar WVs may occur for properly fine-tuned measurement duration times. WVs much larger than 1 are realized, for instance, for orthogonal states when 1 + n · v(τ ) (cid:29) 1, which leads to τ = n · π sin2 α 1 2 ω , where α denotes the altitude angle between v(0) and m. The effect of the strength of the fluctuations on short and intermediate measurements is shown in Figs. 3(c) and 3(d), resulting in a trivial decay towards the steady- state value. ω ± ∆τ , with ∆τ (cid:28) 1 As already anticipated, an important role is played by the "direction" of the noise term. Its effect is illustrated in Fig. 4(a). As long as k becomes more and more par- allel to the m, the relaxation time toward the steady state becomes longer. In the limiting case, it relaxes to a partially coherent state. The direction of the post- selection orientation plays a similar role, as shown in Fig. 4(b). We also note that, in the presence of multiple decoherence sources, even for a given effective direction and strength of the fluctuating term, the decay of the WV still strongly depends on the relative directions between different sources [cf. Fig. 4(c)]. The case of a continuous readout is characterized by a rather different behavior. While in both, the coherent and continuous readout regimes, a significant strong de- coherence destroys the peculiarities of WVs [cf. Fig. 5(a)], FIG. 5. WV for continuous detection for (a) different γ with D = 0.270 eV and (b) different D with γ = 0.1 µeV, other parameters as in Fig. 3; in all plots δΩ/Ω = 0.001. in the latter the detector itself introduces decoherence on the time scale of its backaction. This is visualized in Fig. 5(b). It also shows that D effects the results at the time scales τ (cid:28) 1/D(Ω/δΩ)2 relevant for WVs, in contrast to the coherent case. The difference between the two cases is highlighted in Fig. 6, comparing the two regimes for different decoherence strength and postselec- tion. The WV in the continuous case can be enhanced as compared to the coherent measurement, leading to pecu- liar WVs, where a coherent measurement would not [cf. Figs. 6(c) and (d)]. This effect depends on the chosen postselection [cf. Fig. 6(a) vs. Fig. 6(c); Fig. 6(b) vs. Fig. 6(d)] and is suppressed by decoherence [cf. Fig. 6(c) vs Fig. 6(d)]. We can understand these results by analyz- ing the asymptotic behavior after decoherence has taken place. Equation (31) gives a WV (cid:104)σz(cid:105)weak Ψ(t0) = Πf tr{(1 + n · σ) σz(s) ρ} tr{(1 + n · σ) ρ} = nz (32) for a fully incoherent state ρ ∝ 1. This is the WV of an incoherent state.17 The results from Eq. (26) for the coherent case lead instead to Πf(cid:104)σz(cid:105)weak Ψ(t0) = 0 for ρ ∝ 1. This difference arises because of the coherent evolution in the correlation between measurement and postselection [cf. Eq. (27)] which does not take place in the continu- ous readout due to the frequent "pointer" readout. The different steady states are shown in Fig. 6. Though the steady states correspond to times beyond the weak mea- surement regimes, their difference reflects at shorter time scales as well. There it leads to enhanced WVs exceed- ing the strong boundary in one case and not in the other (cf. Fig. 6(c) and 6(d)). This explains the sensitivity of the effect to postselection (that counts the steady state of the continuous case) and to decoherence (that suppresses faster peculiar WVs within the standard range in both cases). The perturbative solution of the system's dynamics un- derlying the result of the coherent detection allows us to discuss, to some extent, the validity of the WV ex- pression in Eq.( 26). As a first check we can require that the second-order contribution is irrelevant compared to the first-order contribution discussed so far, that is, Πf (cid:104)I(τ )(cid:105)weak(2) Ψ(t0) ·δΩ2/Ω2 (cid:28) Πf (cid:104)I(τ )(cid:105)weak Ψ(t0)·δΩ/Ω, lead- 4ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakΘ(cid:61)Π2Θ(cid:61)Π4Θ(cid:61)Π8Θ(cid:61)04ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakΦ(cid:61)3Π4Φ(cid:61)Π2Φ(cid:61)Π4Φ(cid:61)04ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakΓ1,k1;Γ2,k2Γ1,k1;Γ2,k2Γ,k4ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakΓ(cid:61)1.0ΜeVΓ(cid:61)0.5ΜeVΓ(cid:61)0.0ΜeV4ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakD(cid:61)270meVD(cid:61)27.0meVD(cid:61)2.70meV9 VI. CONCLUSIONS In this work we have addressed the effects of decoher- ence on weak value measurements involving postselec- tion. We have considered the paradigm model of a charge measurement in a double quantum dot by a nearby QPC where we have included fluctuations of the parameters due to external noise sources. After deriving a general ex- pression for the postselected signal (current) in the QPC in terms of the reduced density matrix of the qubit, we have evaluated it explicitly in two different regimes de- termined by the detector's readout, namely continuous vs "single-time" detector's readout. We have characterized the WV's dependence on the various parameters of the system. In particular, we have shown that statistical fluctuations of the qubit's param- eters generally reduce the WV into the classical range for measurements longer than the decoherence time. On shorter time scales we have determined a boundary for the region of validity of the WV result. Remarkably, there the continuous readout can lead to an enhance- ment of peculiar weak values as compared to the case of coherent detection. (a) (b) (c) (d) FIG. 6. Weak value for continuous detection and coherent readout for different γ and pre- and postselection states n, v; γ = 1.0 µeV in (a), (c) and γ = 0.1 µeV in (b), (d); v(0) = (0.5,−0.33, 0.80), n = (−0.20, 0.75, 0.63) in (a,b) and v(0) = (−0.5, 0.33,−0.80), n = (0.50, 0.20, 0.84) in (c), (d). In all plots D = 0.270 eV, δΩ/Ω = 0.001 and the other parameters as in Fig. 3. (cid:82) τ (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) (cid:28) 1 . ACKNOWLEDGMENTS We would like to thank O. Ziberberg and Y. Gefen for very useful discussions. We acknowledge the financial support of ISF and the Minerva Foundation. ing to the condition (cid:12)(cid:12)(cid:12)(cid:12)(cid:12) (cid:82) τ η(τ ) = δΩ Ω · Dτ 2 · vx(τ ) nx + vy(τ ) ny 1 τ 0 ds vz(s) + 1 τ 0 ds nz(s) (33) This imposes a restriction on the validity of the WV's result also within the regime of weak measurement, as shown in Fig. 7. Indeed, for specific qubit's parameters 0 and ∆0 and particular boundary conditions vz(0) and nz, the numerator in Eq. (33) vanishes at finite duration times for τ (cid:28) 1/γ so that the perturbation is valid at discrete times which depend on the chosen parameters (cf. Fig. 7). For τ (cid:29) 1/γ, on the contrary, η ∝ τ 2 e−c τ where c > 0 and the perturbation is asymptotically valid unless k = m = ex. (a) (b) FIG. 7. Validity perturbation. (a) Left-hand side of Eq. (33) and (b) the corresponding WV for γ = 1.0 µeV, D = 0.270 eV, δΩ/Ω = 0.001; all other parameters are as in Fig. 3. ∞(cid:90) Appendix A DERIVATION OF AVERAGED RATE EQUATIONS In this appendix we derive the differential equa- tion (25) out of Eqs. (23) of the main text. In the fol- lowing, the derivation is presented exemplarily only for x = tr{ v(m,m) , v(m) v(m) ) are treated completely analogously. It is useful to perfom a "Laplace" transform for the whole matrices, (t)} since the other terms (v(m) x y z v(m,n) j (E) = lim δ→0 v(m,n) j (t) exp [i (E + i δ) t] dt , 0 (A.1) in order to include the initial conditions of the differential equations that m = 0 electrons have penetrated through the collector at t = 0 and, hence, the qubit-detector sys- tem is in a pure state. Here, δ > 0 ensures the conver- gence of the integral. A high-energy cutoff is introduced concerning the inverse transform, so that the upper limit of the integral in the inverse transform is Λ → ∞ We can write the differential equations for the Laplace transformed components, (t). In this regard, the matrix products in Eq. (23) can be easily calculated by evaluating each (m, n) block as introduced in Eq. (12). v(m,m) j 4ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakcontinuouscoherent4ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakcontinuouscoherent4ns8nsΤ1.02.0(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakcontinuouscoherent4ns8nsΤ1.02.0(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakcontinuouscoherent4ns8ns12nsΤ12ΗΤ1054ns8ns12nsΤ1.20.80.4(cid:80)f(cid:60)ΣzΤ(cid:62)(cid:89)t0weakDue to the fact that the definition of ∆ H + includes terms † proportional to a† l al, diagonal blocks of ∆ H + r ar and a are nonzero, whereas the diagonal blocks ∆ H− (m,m) vanish. Moreover, since ∆ H± consists of combinations † a l ar + h.c. which raise or lower, respectively, the number of electrons in the detector by exactly one electron, only the off-diagonal blocks ∆ H± (m,m+1) and ∆ H± (m+1,m) neighboring the diagonal blocks are nonzero. Conse- quently, a blockwise evaluation of the matrix products can be written as 10 vj · ∆ H−(cid:17)(m,n) (cid:16) vj · ∆ H +(cid:17)(m,n) (cid:16) j = v(m,n−1) = v(m,n−1) j · ∆ H− (n−1,n) + v(m,n+1) · ∆ H + (n−1,n) + v(m,n) j j · ∆ H− (n+1,n) , · ∆ H + (n,n) + v(m,n+1) j · ∆ H + (n+1,n) , (A.2) (A.3) · ∆ H− (k,n) still describes a product of matrices which are of and analogous for (∆ H +/− · vj)(m,n). Note that v(m,k) infinite dimension. Thus, applying the "Laplace" transform to Eqs. (23) for j = x by considering exemplarily k = ez leads to j tr (i E − δ) v(m,m) x (E) + v(m,m) x (0) (cid:111)(cid:105) + ∆ H + (m,m) · v(m,m) (E) (E) + ∆ H + (m,m+1) · v(m+1,m) (E)  (cid:125)  (cid:125) x =ve(E) (cid:123)(cid:122) (cid:123)(cid:122) − 2 0 v(m,m) =vb(E) (E) (cid:125) x x tr (t) (E) =va(E) (cid:110) − i 2 = £−1 = −£−1(cid:104) (cid:111) (cid:110) v(m,m) −2 γ v(m,m) tr (cid:124) (cid:123)(cid:122) (cid:125) (cid:124) ∆ H + (m,m−1) · v(m−1,m) (cid:124) (cid:123)(cid:122) v(m,m−1) (cid:123)(cid:122) (cid:124) ∆ H− (m,m−1) · v(m−1,m) (cid:123)(cid:122) (cid:124) (cid:123)(cid:122) (cid:124) + v(m,m−1) − 1 2 =vf (E) =vc(E) =vi(E) i 2 + x x y y =vk(E) y (cid:123)(cid:122) (cid:125) (cid:125) (cid:125) (cid:124) (cid:124) (cid:124) x =vd(E) (cid:123)(cid:122) (cid:123)(cid:122) =vj (E) (cid:123)(cid:122) (cid:123)(cid:122) =vl(E) (cid:125) (cid:125) (cid:124) (cid:124) (cid:125)  (cid:125) (E) · ∆ H + (m−1,m) + v(m,m) x (E) · ∆ H + (m,m) + v(m,m+1) x (E) · ∆ H + (m+1,m) =vg(E) =vh(E) + ∆ H− (m,m+1) · v(m+1,m) (E) (E) y (E) · ∆ H− (m−1,m) (E) · ∆ H− (m+1,m) (cid:125) (cid:124) + v(m,m+1) y   , (A.4) where £−1 denotes the inverse "Laplace" transform. We analyze the various terms separately. We start with observing that £−1[va(E)] = −2 γ tr{v(m,m) (t)}. Moreover, due to the cyclic invariance of the trace £−1[vd(E)] + £−1[vg(E)] = 0. Employing the explicit expression of ∆ H±, reduces the evaluation of the remaining terms to the calculation of ∆ H + (m,m+1) ·v(m+1,n) (−1)m2 Ωlm+1rm+1 ·(cid:16) (t)} and £−1[vb(E)] = −2 0 tr{v(m,m) (cid:88) v(m+1,m) j (cid:16) (cid:17) (cid:17) = x y j (l1r1...lmrmlm+1rm+1);(l(cid:48) (l1r1...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) lm+1rm+1 nr(cid:48) n) 1r(cid:48) 1...l(cid:48) (A.5) and(cid:16) (cid:17) ∆ H + (m,m−1) ·v(m−1,n) j (l1r1...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) (cid:17) (cid:17) (cid:17) j = − 2 Ωl1r1 ·(cid:16) + 2 Ωl2r2 ·(cid:16) − 2 Ωl3r3 ·(cid:16) + (−1)m 2 Ωlmrm ·(cid:16) v(m−1,m) v(m−1,m) v(m−1,m) . . . j j (l2r2...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) (l1r1l3r3...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) (l1r1l2r2l4r4...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) 11 (A.6) (cid:16) for products concerning the off-diagonal boxes and ∆ H + (m,m) ·v(m,n) j (l1r1...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) (cid:17) (cid:17) = 2 (Er1 + . . . + Erm − El1 − . . . − Elm) ·(cid:16) v(m−1,m) j (cid:17) v(m,n) j (l1r1...lm−1rm−1);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) (l1r1...lmrm);(l(cid:48) 1r(cid:48) 1...l(cid:48) nr(cid:48) n) (A.7) j lkrk 1r(cid:48) (cid:16) nr(cid:48) v(m,n) j To lowest order in ρ Ω2 (cid:28) Vd, we find that for products with non-zero matrices on the diagonal (m, m). Here, the indices (l1r1 . . . lmrm); (l(cid:48) 1r(cid:48) 1 . . . l(cid:48) n) precisely determine the scalar entries of the matrices. The analogous multiplications involving ∆ H− are evaluated by replacing Ω → δΩ. In principle, these products are not restricted to the usage of vj(t) and, thus, are valid for a generic matrix of the dimension of v(m±1,m) . It is sufficient to consider products where ∆ H± is on the left, since in the end we need to evaluate the traces and all products can always be arranged such that ∆ H± is on the left by relying on the cyclic invariance of the traces. (cid:17) (cid:101)E + Elk − Erk + i (δ + 2 γ) (l1r1...lkrk...lmrm);(l(cid:48) (l1r1...lmrm);(l(cid:48) Ω2 (cid:88) ≈ 0, Ω2 (cid:88) (E) with the abbreviation (cid:101)E = E − El1 − . . . − Elk−1 − Elk+1 − if the sum runs over indices occurring in v(m,m) → (cid:82) ρl(Elk ) ρr(Erk ) dElk dErk . Additionally, the hopping amplitudes are assumed to depend weakly on the (cid:80) . . . − Elm + Er1 + . . . + Erk−1 + Erk+1 + . . . + Erm. This result holds within the approximation where the energy levels of the detector's reservoirs are almost continuous so that the sum can be replaced by an integral, that is, energy levels, hence, being constant, that is, Ωlr ≡ Ω, and also the density matrices of the emitter and collector, respectively, are approximated to be constant, with ρl(Elk ) = ρl and ρr(Erk ) = ρr. In order to understand that the integral vanishes, it is helpful to realize that the all entries of v(m,n) essentially characterize higher-order retarded Green's functions which describe the averaged evolution of the density matrix. These Green's functions have poles in the lower half of the complex plane proportional to [(cid:101)E + Elk − Erk + i δ]−1 which can be shown by iteratively solving k...l(cid:48) + i (δ + 2 γ) (cid:17) (cid:101)E + El(cid:48) − Er(cid:48) v(m,n) j 1r(cid:48) 1...l(cid:48) ≈ 0 (A.8) 1...l(cid:48) nr(cid:48) n) (cid:16) nr(cid:48) n) the Laplace transformed differential equations (23) for each v(m,n) integrand decreases ∝ 1/E2. On the contrary, if the sum does not include a summation over indices of v(m,n) can be calculated within the same approximation as being (E). Thus, a contour integral yields zero since the (E), it j l(cid:48) kr(cid:48) kr(cid:48) lkrk j j j k k k Ω2 (cid:88) lkrk (cid:101)E + Elk − Erk + i (δ + 2 γ) 1 ≈ −i π ρl ρr Ω2 (cid:16) Vd + (cid:101)E (cid:17) ≈ − i 2 D (A.9) to lowest order in ρ Ω2 (cid:28) Vd with D = 2 π ρl ρr Ω2 Vd.25 Concerning this, the integral is separated into a singular part and the principal-value part. While the principal part redefines the energy levels, the singular parts lead to the D · vc(E) ≈ i 4 − tr ve(E) ≈ − i 4 − tr (cid:111) (cid:110) (cid:110) (cid:18) (cid:110) 2 tr v(m−1,m−1) x (E) + 2 i δΩ Ω tr v(m−1,m−1) y (E) v(m−1,m) (0) · x (i E − δ − 2 γ) 1(m,m) + ∆ H + (m,m) i 2 (cid:111) − 2 i (cid:110) δΩ Ω tr v(m,m) y (E) 2 tr v(m,m) x (E) (cid:111)(cid:19) i 2 (cid:110) (cid:111) ∆ H + (m,m+1) · (i E − δ − 2 γ) 1(m+1,m+1) + ∆ H + (m+1,m+1) D · vi(E) ≈ i 4 − tr − 2 δΩ Ω v(m−1,m) y v(m−1,m−1) (E) + 2 i v(m−1,m−1) x (0) · (i E − δ − 2 γ) 1(m,m) + ∆ H + (m,m) δΩ2 Ω2 tr i 2 (cid:110) (cid:111)(cid:19) (cid:111) 2 δΩ Ω tr v(m,m) y (E) + 2 i δΩ2 Ω2 tr v(m,m) x (E) D · vj(E) ≈ i 4 − tr ∆ H− (m,m+1) · (i E − δ − 2 γ) 1(m+1,m+1) − i 2 ∆ H + (m+1,m+1) (cid:111)(cid:19) (cid:19)−1 · ∆ H + (m,m−1) (cid:41) , (cid:19)−1 · v(m+1,m) (cid:41) (cid:111)(cid:19) (cid:19)−1 · ∆ H− (m,m−1) (E) x 12 (A.10) (cid:41) (0) , (A.11) (A.12) (cid:19)−1 · v(m+1,m) y (cid:41) (0) , (A.13) (cid:18) y (cid:18) (cid:18) (cid:18) (cid:110) tr (cid:110) (cid:18) (cid:40) D · (cid:40) (cid:18) (cid:40) (cid:18) (cid:40) (cid:18) (cid:18) presented result by relying on the Sokhatsky-Weierstrass theorem. Thus, one obtains which couples (m,m) blocks to (m − 1,m − 1) and (m + 1,m + 1) blocks, respectively. Note that it is trivial to take the inverses since the corresponding matrices are diagonal. Similar calculations eventuate in analogous expression for all the other terms by replacing m with m + 1 or m − 1, respectively, that is, vf (E) = ve(E)(m → m − 1), vh(E) = vc(E)(m → m + 1), vk(E) = vj(E)(m → m − 1) and vl(E) = vi(E)(m → m + 1). Inserting the achieved results for va(E) to vl(E) into Eq. (A.4) and iteratively solving in m yields the desired differential equation for v(m) j with j = x after performing the re-Laplace transform, that is, v(m,m) j (t) := tr (cid:110) (cid:111) (t) (cid:18) (cid:18) (cid:19)(cid:19) (cid:18) (cid:19) v(m) x (t) = −2 γ − D 2 1 + δΩ2 Ω2 v(m) x (t) + D 2 1 − δΩ2 Ω2 v(m−1) x (t) − 2 0 v(m) y (t) . (A.14) Likewise calculations for each v(m) (t) finally end up in the rate equations j (cid:19) (cid:18) v(m) 0 v(m) y 1 + (cid:18) (t) = − D 2 −2 γ − D 2 (t) = δΩ2 Ω2 (cid:18) v(m) z (t) = − D 2 1 + δΩ2 Ω2 1 + (cid:19) D 2 (t) + (cid:19)(cid:19) v(m) y v(m) 0 δΩ2 Ω2 v(m) z (t) + D 2 1 + (t) + (cid:18) 1 + (cid:19) (cid:18) (cid:19) δΩ2 Ω2 D 2 δΩ2 Ω2 (cid:19) 1 − δΩ2 Ω2 v(m−1) z v(m−1) (t) − D 0 δΩ Ω v(m) z (t) + D v(m−1) z (t) δΩ Ω v(m−1) y (t) + 2 0 v(m) x (t) − 2 ∆ v(m) z (t) (A.15) (A.16) (t) − D δΩ Ω v(m) 0 (t) + D v(m−1) 0 δΩ Ω (t) + 2 ∆ v(m) y (t) (A.17) for the special case of considering fluctuations in the z direction. The general case is treated analogously. Appendix B PERTURBATIVE SOLUTION OF THE RATE EQUATIONS In this appendix we present the perturbative solution for δΩ (cid:28) Ω of the rate equations (25) which describe the averaged evolution of the density matrix with the initial conditions that m = 0 electrons have penetrated to the right reservoir at t = 0 so that the initial state of the system is described by v(m) (t = 0), j = 0, x, y, z. (t = 0) = δm,0 · v(0) j j Hereto, we introduce the Fourier transform (cid:101)vj(q, t) = (cid:88) m v(m) j (t) e− imq , (B.1) with respect to m. Hence, (cid:104)R(m)(t)(cid:105)ξ becomes (cid:104)(cid:101)R(q, t)(cid:105)ξ = 1/2 · ((cid:101)v(q, t) · σ) which eventuates in an expression of the conditional current in terms of(cid:101)vj(q, t) by comparing to Eq. (16), that is, 13 (cid:104)I(τ )(cid:105)Ψ(t0) = − i e τ Πf ∂q ln tr If assuming that(cid:101)v(q, t) is an analytic function of q and t, it can perturbatively be expanded in order to describe the evolution of the system's density matrix within a weak measurement regime, (cid:27)(cid:21)(cid:12)(cid:12)(cid:12)(cid:12)q=0 . (cid:20) (cid:26) Πf · ((cid:101)v(q, t) · σ) (cid:18) δΩ (cid:19)n , Ω (cid:101)v(q, t) = ∞(cid:88) n=0 un(q, t) where n denotes the order of perturbation. Substituting this perturbation into Eq. (B.2) the zeroth-order contribution reads (cid:104)I(τ )(cid:105) = − i e τ ∂q ln tr (cid:26) (cid:20) (cid:26) Πf (cid:104)I(τ )(cid:105)weak Ψ(t0) = − i e τ − tr Πf · (u0(q = 0, t) · σ) · tr tr Πf · u1(q = 0, t) · σ) · tr (cid:27) (cid:20) (cid:26) Πf · (u0(q, t) · σ) (cid:27)(cid:21)(cid:12)(cid:12)(cid:12)(cid:12)q=0 (cid:27) Πf · ( ∂qu1(q, t)q=0 · σ) (cid:20) (cid:27)(cid:21) · (cid:26) Πf · ( ∂qu0(q, t)q=0 · σ) (cid:27) (cid:26) (cid:26) and the first-order contribution, that is, the averaged WV for the current, is identified as Πf · (u0(q = 0, t) · σ) tr (cid:27)(cid:21)−2 . (B.5) Thus, the WV is completely expressed in terms of the averaged density matrix. Further analysis now focuses on the evaluation of un(q, t) for n = 0, 1, aiming at finding an illustrative expression for the WV. Inserting the power series of Eq. (B.3) into the Fourier transformed rate equations (25) leads to a set of differential equation for each un(q, t). The resulting equations for the lowest orders are (B.2) (B.3) (B.4) (B.6) (B.7) (B.8) (B.9) (cid:20) (cid:20) (cid:20) ∂ ∂ t ∂ ∂ t ∂ ∂ t u0(q, t) = u1(q, t) = u2(q, t) = G0 + Gk − D 2 G0 + Gk − D 2 G0 + Gk − D 2 (1 − ei q) · 1(4) (1 − ei q) · 1(4) · u0(q, t), · u1(q, t) − D (1 − ei q) · G01 · u0(q, t) (1 − ei q) · 1(4) · u2(q, t) − D 2 · Gq · u0(q, t) (cid:19) (cid:21) (1 − ei q) · 1(4) · u1(q, t) t − D (1 − ei q) · G01 · exp (cid:21) (cid:21) (cid:21) G0 + Gk − D 2 (cid:20)(cid:18)  and Gq := Here, we have introduced  0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0  , G10 :=  0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 G01 :=  1 − ei q 0 0 0  . 0 1 + ei q 0 0 0 0 1 + ei q 0 0 0 0 1 − ei q The initial conditions are equivalent to u0(q, t = 0) = (1, v(0) Higher orders in un(q, t) are not relevant for the expression for the WV. Pertinent for the expression of the WV are solutions for the special cases un(q = 0, t) and ∂qun(q, t)q=0. z (t = 0)). Thus, u0(q, t = 0) does not depend on q initially and un(q, t = 0) ≡ 0 for n ≥ 1. Furthermore, ∂qun(q, t = 0)q=0 ≡ 0 for all n ∈ N and tr{(cid:104)ρ(t)(cid:105)ξ} ≡ 1 implies (u0(q = 0, t))0 ≡ 1 and (un(q = 0, t))0 ≡ 0 at any time t. Additionally, ∂tun(q = 0, t) ≡ 0 at any t to keep tr{(cid:104)ρ(t)(cid:105)ξ} unchanged. The perturbative solution is obtained iteratively with v(t) = exp [(G0 + Gk) t] · v(0) where v(0) = x (t = 0), v(0) y (t = 0), v(0) (v(0) x (0), v(0) y (0), v(0) z (0)). It reads (cid:19) (cid:18) 0 0 (cid:18) 0 0 (cid:19) , (cid:18) (cid:19) , 0 D t (vx(t) ex + vy(t) ey) v(t) (cid:19) (cid:18) 1 (cid:19) (cid:18) 1 z ·(cid:16)  e T (cid:16) D 2 v(t) t (cid:110) u0(q = 0, t) = , u1(q = 0, t) = , u2(q = 0, t) = ∂qu0(q, t)q=0 = i ∂qu2(q, t)q=0 = , [G0 + Gk] [G0 + Gk] −1 · exp [(G0 + Gk) t] − [G0 + Gk] −1 · exp [(G0 + Gk) t] − [G0 + Gk] (cid:111) (cid:110)  . −1(cid:17) · v(t = 0) −1(cid:17) · ez (cid:111) Πf · (u0(q = 0, t) · σ) tr = 1 + v(τ ) · n, tr Πf · (u1(q = 0, t) · σ) = 0 ∂qu1(q, t)q=0 = i D Hereafter we conclude that and (cid:110) (cid:111) Πf · ( ∂qu1(q, t)q=0 · σ) − i tr = −i t(cid:90) 0 ∂ ∂ s tr (cid:110) (cid:111) Πf · ( ∂qu1(q, s)q=0 · σ) ds = D τ(cid:90) 0 vz(s) + nz(s) ds . (B.14) Inserting Eqs. (B.14) and (B.13) into the expression for the conditional value in Eq. (B.2) then yields the expression for the WV, that is, Eq. (26), which is presented in the main text. Similar, the second-order contribution to the WV is evaluated by noting that (cid:110) Πf · ( ∂qu0(q, t)q=0 · σ) (cid:111) 1 + v(τ ) · n (cid:18) D τ 2 = (cid:18) (cid:111) Πf · (u2(q = 0, t) · σ) = D τ 2 vx(τ ) nx + vy(τ ) ny − i tr (cid:110) (cid:19) (cid:19) and tr 14 (B.10) (B.11) (B.12) (B.13) (B.15) (B.16) (B.17) which eventuates in the expression Πf (cid:104)I(τ )(cid:105)weak(2) Ψ(t0) = − e D2 τ 2 · vx(τ ) nx + vy(τ ) ny 1 + n · v(τ ) . Note that the second-order contribution in Eq. (B.17) has the same characteristics, that is, the same denominator, as the first-order term in Eq. (26), which implies analogous conditions for divergencies or peculiar WVs. Appendix C ANALYSIS OF THE CONTINUOS DETECTION Here, we derive Eq. (30) of the main text for the case of a continuous detection. We start by assuming ∆t (cid:28) min{1/ω(Ω/δΩ), 1/γ(Ω/δΩ), 1/D(Ω/δΩ)}. The evolution between two subsequent readouts is still exactly given by Eq. (25) with the modified initial condition that precisely nk electrons have been read out at t = tk so that v(n)(tk) = δn,nk · (1, v(nk) (tk)). In order to solve these modified differential equations it is useful to introduce a vector w(t) = (v(0)(t), v(1)(t), . . . , v(n)(t), . . .) where n labels the number of transferred electrons so that Eq. (25) reads (tk), v(nk) (tk), v(nk) x y z w(t) = (M1 + M2) · w(t) d dt (C.1) = (G0 + Gk − D/2) δm,n + D 1 with M (m,n) 2 1(4)) δm,n+1. This differential equation is solved trivially in the limit ∆t (cid:28) 1/ω(Ω/δΩ), 1/γ(Ω/δΩ), 1/D(Ω/δΩ) we are interested in, by noting that M1 is a block-diagonalized matrix in Jordan form, with the solution 2 1(4) δm,n+1 and M (m,n) = (G1 + D 2 2 1(4)) δm,n + (G2 − D (cid:20) (cid:20) (cid:21) (cid:18) (cid:21) (cid:18) (cid:19) v(nk)(t) = exp (G0 + Gk − D 2 1)(t − tk) · 1(4) + (G1 + D 2 1(4))∆t · v(nk)(tk) , v(nk+m)(t) = fm · exp (G0 + Gk − D 2 1)(t − tk) · (G2 − D 2 1(4))∆t + (t − tk)(1(4) + (G1 + D 2 D 2 1(4))∆t) (cid:19) (C.2) · v(nk)(tk) , 2 (t − tk)(cid:1)m−1 (cid:0) D ensured by tr{ρ(system)(t)} =(cid:80)∞ where fm = 1 m! is, ∆t (cid:28) 1/D, which corresponds to a continuous readout, the solution in Eq. (C.2) becomes with m ≥ 1. Within our approximation, in Eq. (C.2) probability conservation is (t) ≡ 1. Setting the reading time scales as the smallest in the problem, that n=0 v(n) 0 v(nk)(t) = exp [(G0 + Gk)(t − tk)] ·(cid:0)1(4) + G1(t − tk)(cid:1) · v(nk)(tk) v(nk+1)(t) = exp [(G0 + Gk)(t − tk)] · G2(t − tk) · v(nk)(tk) v(nk+m)(t) = 0 , m ≥ 2 . This is the limit where at most one electron penetrates through the QPC between two subsequent readouts. The probability that exactly one electron will have accumulated in the collector within a readout period time is given by P (0; ∆t) = tr{A · v(nk)(tk)}0, while zero electrons penetrate with a probability P (1; ∆t) = tr{B · v(nk)(tk)}0, where the matrices A and B are given by Eqs. (29) and {. . .}0 denotes the zeroth component. With these definitions the conditional number of transmitted electrons can be expressed as (cid:32) N(cid:80) m (cid:80) (cid:32) N(cid:80) (cid:80) m=0 perm n · n · (cid:2)AN−m · Bm(cid:3) (cid:33) (cid:33) · v perm [AN−m · Bm]perm · v (cid:104)n(τ )(cid:105)Ψ(t0) = Πf 15 (C.3) (C.4) where v is evaluated at t = 0. N describes the total number of readouts and(cid:80) Defining the auxiliary function f (m, ∆t, N ) =(cid:80) perm indicates the sum over all possible orders of A and B in the string of products AN−mBm. In the limit of N → ∞, ∆t → 0 while keeping N · ∆t = τ constant, the sum over all permutations in Eq. (C.4) can be analytically evaluated. perm, the numerator of Eq. (C.4) is evaluated as perm perm m=0 (cid:2)AN−m · Bm(cid:3) (cid:33) follows n · (cid:32) N(cid:88) m=0 = i n · ∂q = i n · ∂q = i n · ∂q m · f (m, ∆t, N ) · v m=0 · v (cid:35) f (m, ∆t, N ) e−imq (cid:34) N(cid:88) (cid:34) N(cid:88) (cid:88) (cid:104)(cid:0)1 + (G0 + Gk + G1 + G2 e−iq)∆t(cid:1)N(cid:105) · v (cid:2)AN−m · Bm(cid:3) (cid:2)AN−m · (B e−iq)m(cid:3) (cid:35) · v perm perm m=0 (cid:80) where the last step is valid since(cid:80)N perm = (A + B)N and due to the definitions in Eqs. (29). In the limit N → ∞, ∆t → 0, N · ∆t = τ , this readily yields τ n · G2 · exp((G0 + Gk + G1 + G2)τ ) · v(0). The denominator is treated analogously, which finally leads to Eq. (30). perm m=0 1 J. von Neumann, in Mathematische Grundlagen der Quan- tentheorie (Springer-Verlag Berlin Heidelberg New York, 1996) zweite auflage ed., ISBN 3-540-59207-5 2 Y. Aharonov, D. Z. Albert, and L. Vaidman, Phys. Rev. Lett. 60, 1351 (1988) 3 Y. Shikano and A. Hosoya, Journal of Physics A: Mathe- matical and Theoretical 43, 025304 (2010) 4 A. Peres, Phys. Rev. Lett. 62, 2326 (1989) 5 Y. Aharonov and L. Vaidman, Phys. Rev. Lett. 62, 2327 7 Y. Aharonov, A. Botero, S. Popescu, B. Reznika, and J. Tollaksen, Phys. Lett. A 301, 130 (2002) 8 J. S. Lundeen and A. M. Steinberg, Phys. Rev. Lett. 102, 020404 (2009) 9 A. M. Steinberg, Phys. Rev. Lett. 74, 2405 (1995) 10 J. S. Lundeen, B. Sutherland, A. Patel, C. Stewart, and C. Bamber, Nature 474, 188 (2011) 11 H. Wei and Y. V. Nazarov, Phys. Rev. B 78, 045308 (2008) 12 J. Dressel and A. N. Jordan, Phys. Rev. A 85, 022123 (1989) (2012) 6 A. J. Leggett, Phys. Rev. Lett. 62, 2325 (1989) 13 O. Hosten and P. Kwiat, Science 319, 787 (2008) 16 14 K. Yokota, T. Yamamoto, M. Koashi, and N. Imoto, New J. Phys. 11, 033011 (2009) 15 C. Simon and E. S. Polzik, Phys. Rev. A 83, 040101 (2011) 16 P. B. Dixon, D. J. Starling, A. N. Jordan, and J. C. Howell, Phys. Rev. Lett. 102, 173601 (2009) 17 A. Romito, Y. Gefen, and Y. M. Blanter, Phys. Rev. Lett. 100, 056801 (2008) 18 N. S. Williams and A. N. Jordan, Phys. Rev. Lett. 100, 026804 (2008) 19 V. Shpitalnik, Y. Gefen, and A. Romito, Phys. Rev. Lett. 101, 226802 (2008) 20 A. Bednorz and W. Belzig, Phys. Rev. Lett. 105, 106803 (2010) 21 O. Zilberberg, A. Romito, and Y. Gefen, Phys. Rev. Lett. 106, 080405 (2011) 22 A. Morello, J. J. Pla, F. A. Zwanenburg, K. W. Chan, K. Y. Tan, H. Huebl, M. Mottonen, C. D. Nugroho, C. Yang, J. A. van Donkelaar, A. D. C. Alves, D. N. Jamieson, C. C. Escott, L. C. L. Hollenberg, R. G. Clark, and A. S. Dzurak, Nature 467, 687 (2010), ISSN 0028-0836 23 A. Romito and Y. Gefen, Physica E 42, 343 (2010) 24 A. Di Lorenzo and Y. V. Nazarov, Phys. Rev. Lett. 93, 046601 (2004) 25 S. A. Gurvitz, Phys. Rev. B 56, 15215 (1997) 26 S. A. Gurvitz and D. Mozyrsky, Phys. Rev. B 77, 075325 (2008) 27 In the general case where the eigenstates of the operator measured by the QPC do not coincide with L >, R >, H0 written in such an eigenstates basis takes the same form with renormalized coefficients , ∆. 28 S. A. Gurvitz, arXiv:quant-ph/9808058v2(1998) 29 A. N. Korotkov, Phys. Rev. B 60, 5737 (1999) 30 A. N. Korotkov, Phys. Rev. B 63, 115403 (2001) 31 I. L. Aleiner, N. S. Wingreen, and Y. Meir, Phys. Rev. Lett. 79, 3740 (1997) 32 B. Rosenow and Y. Gefen, Phys. Rev. Lett. 108, 256805 (2012) 33 While we are neglecting the terms with more than one electron having been transferred, the probability is still exactly conserved at any time since tr{ρ(system)(t)} = 34 A. N. Korotkov and D. V. Averin, Phys. Rev. B 64, 165310 (cid:80)∞ (2001) n=0 v(n) 0 (t) ≡ 1.
1212.2570	1	1212	2012-12-11T18:13:02	Determining molecule diffusion coefficients on surfaces from a locally fixed probe: On the analysis of signal fluctuations	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci", "cond-mat.stat-mech" ]	Methods of determining surface diffusion coefficients of molecules from signal fluctuations of a locally fixed probe are revisited and refined. Particular emphasis is put on the influence of the molecule's extent. In addition to the formerly introduced autocorrelation method and residence time method, we develop a further method based on the distribution of intervals between successive peaks in the signal. The theoretical findings are applied to STM measurements of copper phthalocyanine (CuPc) molecules on the Ag(100) surface. We discuss advantages and disadvantages of each method and suggest a combination to obtain accurate results for diffusion coefficients.	cond-mat.mes-hall	cond-mat	Determining molecule diffusion coefficients on surfaces from a locally fixed probe: On the analysis of signal fluctuations Susanne Hahne,1 Julian Ikonomov,2 Moritz Sokolowski,2 and Philipp Maass1, ∗ 1Fachbereich Physik, Universitat Osnabruck, Barbarastrasse 7, 49076 Osnabruck, Germany 2Institut fur Physikalische und Theoretische Chemie, Universitat Bonn, Wegelerstrasse 12, 53115 Bonn, Germany (Dated: December 7, 2012) Methods of determining surface diffusion coefficients of molecules from signal fluctuations of a locally fixed probe are revisited and refined. Particular emphasis is put on the influence of the molecule's extent. In addition to the formerly introduced autocorrelation method and residence time method, we develop a further method based on the distribution of intervals between successive peaks in the signal. The theoretical findings are applied to STM measurements of copper phthalocyanine (CuPc) molecules on the Ag(100) surface. We discuss advantages and disadvantages of each method and suggest a combination to obtain accurate results for diffusion coefficients. PACS numbers: 68.43.Jk,68.35.Fx,82.37.Gk I. INTRODUCTION be used. To describe the growth kinetics of adsorbates on sur- faces, knowledge of diffusion coefficients of atoms and molecules on the surface is of vital importance. Different experimental techniques can be utilized to measure cor- responding diffusion coefficients. Two of the first meth- ods were Field Emission Microscopy1,2 and Field Ioniza- tion Microscopy.3 Another technique is the laser induced thermal desorption,4,5 where first an area of adparticles is cleared by desorption and then the successive repop- ulation measured by desorption. For scanning tunneling microscopy (STM) several modes of operation have been described: Taking scans of the surface with the adsorbed particles at different time instants to identify individual particle displacements,6 recording current signals in time with a locally fixed STM tip,7 and coupling the tip to a single adsorbate and tracking its motions.8 Different pro- cedures to extract diffusion coefficients from these and other techniques have been reviewed by Barth in 2000.9 Analysis of island densities with predictions from rate equations for submonolayer growth kinetics is a further powerful method,10,11 which allows one also to deter- mine binding energies between adatoms or of adatoms to small clusters. An extension of this method to multi- component systems is given in Refs. 12 and 13. In recent years the self-assembly of organic molecules on surfaces has attracted much attention in connection with the idea of developing electronic devices on a molec- ular level.14 Molecules in organic surface growth often have high mobilities and sizes large compared to the lat- tice constant of the substrate. In this situation not all techniques are equally well suited. A convenient and powerful method is the recording of temporal current fluctuations with a fixed scanning tunneling microscope tip15 or, in principle, of signal fluctuations from any locally fixed microscopic or other probe.2 For example, for molecules on insulating substrates, the frequency or height fluctuations in an atomic force microscopy could In order to extract diffusion coefficients from signal fluctuations of a probe, a theoretical basis for the anal- ysis of the measurement is necessary. One approach was developed for the auto-correlation function (ACF) of an STM signal16,17 and applied, on the basis of the corre- sponding power spectrum, to oxygen atom diffusion on Si(111),15 and more recently to hydrogen atom diffusion on Si(111).18 Another approach was followed in Ref. 19 by analyzing the distribution of peak widths in the signal, which correspond to residence times of single molecules under an STM tip. We will refer to it as the residence time distribution (RTD) method in the following. For de- termining absolute values of diffusion coefficients, a gen- eral problem is that the theoretical modeling involves a commonly not precisely known "detection function" with some finite range. However, when the extension of the diffusing molecules is larger than the detection range, this problem becomes essentially irrelevant and the size of the molecule enters as the relevant length scale. In this work we first revisit both the ACF and the RTD method. For the ACF method we study, as a new aspect, modifications implied by the finite size of molecules and for the RTD method we give a detailed account of the functional form of the RTD in different time regimes. In the RTD, time intervals are consid- ered where the molecule is under (or very close to) the tip. Corresponding time intervals are also included in the calculation of the ACF. It can therefore not be pre- cluded that the interaction of the molecule with the tip is influencing the results. A possible influence can be evaluated systematically by changing the bias voltage,19 or the distance between tip and substrate. In order to ensure that the influence is small independent of adjust- ments, we develop a further method, which is based on the analysis of the distribution of peak-to-peak distances in the signal. This method is referred to as the interpeak time distribution (ITD) method in the following. All three methods are applied to diffusion measure- ments of copper phthalocyanine (CuPc) on the Ag(100) surface previously reported in Ref. 19. We discuss ad- vantages and disadvantages of the three procedures men- tioned above and a possible combination of them, which enables us to improve the quality of results for the diffu- sion coefficient. Accordingly, the detection function can be defined as "in- dicator function" of the molecule G(r) = (cid:26) 1 if r ∈ M 0 else , (3) 2 II. AUTO-CORRELATION FUNCTION (ACF) We consider N molecules diffusing on a flat substrate area A and a probing tip of an STM placed at a fixed position above the surface. The lateral tip position is chosen as the origin of an x-y-coordinate system on the surface. Each diffusing molecule α with its center at position rα(t) = (xα(t), yα(t)) gives a contribution iα(t) = G(rα(t)) to the total tunneling current I(t) = α=1 iα(t), where G(r) is the "detection function". We regard the molecules as non-interacting, which should be a valid assumption at low concentrations c = N/A, or more precisely on time scales smaller than l2/D, where l ∼ c−1/2 is the mean intermolecular distance and c the number density of molecules. The autocorrelation function C(t) = hδI(0)δI(t)i of the current fluctuations δI(t) = I(t) − hIi can then be written as PN C(t) = Nhi(0)i(t)i r0Z d2 = cZ d2 r1G(r1)f (r1, tr0)G(r0) , (1) where f (r1, tr0) is the two-dimensional diffusion propa- gator f (r1, tr0) = 1 4πDt exp(cid:18)− (r1 − r0)2 4Dt (cid:19) . (2) To provide a quantitative prediction for C(t), knowl- edge of the detection function G(r) is necessary. For the STM this requires a detailed treatment of the tunnel- ing problem, which is a long-standing problem since the first pioneering work by Tersoff and Hamann.20 Based on a simple approach, Sumetskii and Kornyshev16 sug- gested a Gaussian detection function for small particles (adatoms) as a result of the tunneling barrier. For ex- tended molecules, the lateral structure of the electronic charge density needs to be taken into account. In fact, without referring to first-principle calculations for the specific system under consideration, it is difficult to ob- tain accurate expressions for the tunneling current. While such more detailed treatments are interesting in themselves, they are not really required for deter- mining the diffusion coefficient D. A more practical approach is to transform I(t) into a rectangular signal Irec(t) = Θ(I(t) − Ic), where Θ(.) is the Heaviside jump function (Θ(x) = 1 for x > 0 and zero else), and Ic is a threshold current to exclude background noise. If we con- sider situations, where at most one molecule can be in the detection region under the tip, each single molecule con- tribution iα(t) to Irec(t) also transforms to a rectangular signal irec α (t) ∈ {0, 1}. α (t) ∈ {0, 1}, and Irec(t) = Pα irec where M is the set of center positions of the molecule which give rise to a tunneling current I > Ic. To get a de- scription that is independent of details of the molecule's shape, we assign a circle to the set M. The radius R of this circle can, for example, be determined by setting the covered area of the molecule equal to πR2, by the gyration radius of the set M, or by some other reason- able requirement. The function G(r) from Eq. (3) is then approximated by Θ(R − r). The mutual exclusion under the tip implies that the molecules are no longer non-interacting as assumed when deriving Eq. (1). This leads to nonzero cross-correlations α (t)irec hirec β (0)i, α 6= β. While these cross-correlations could be treated approximately, we concentrate here on times smaller than the mean residence time ∼ τR := R2/D of a molecule under the tip. In this regime we can set hirec β (0)i ≃ 0 for α 6= β. Accordingly, the self-part Nhirec α (0)i appearing in Eq. (1) gives the correlation function Crec(t) = hIrec(t)Irec(0)i (i.e. with- out subtracting a term hIreci2). Evaluation of Eq. (1) then yields α (t)irec α (t)irec Crec(t) = 4πcZ R 0 dr0r0Z R 0 dr1r1 exp(cid:16)− r2 4Dt 1 0 +r2 4Dt (cid:17) × (4) I0(cid:18) 2r0r1 4Dt (cid:19) , where I0 is the modified Bessel function of zeroth order. The knowledge of Crec(t) for times t < τR, as pre- dicted by Eq. (4) for single-particle diffusion, is suffi- cient to determine D. For t > τR, collective proper- ties and the associated cross-correlations had to be taken into account. In a rough treatment one could factorize hirec α (t)irec β (0)i, which amounts to add (cπR2)2 to the expression in Eq. (4) for t ≫ τR. β (0)i ≃ hirec α (t)ihirec Application to CuPc on Ag(100) A relaxed structure of the CuPc molecule in vacuum after energy minimization, as obtained from a DFT cal- culation, is displayed in Fig. 1, together with. A circle of radius R = 7.6 A is assigned to it. The relaxed struc- ture can be expected to be essentially unmodified when the CuPc molecule is adsorbed on the Ag(100) surface. STM studies combined with DFT calculations showed that transition metal phtalocyanines generally lie flat on the Ag(100) surface.21 This is in agreement with STM images of CuPc islands on Ag(100).22 After deposition of CuPc molecules on Ag(100) up to a coverage of 10-15%, islands form and a quasi-stationary state is reached, where the rates of detachment from and 106 103 100 s e c n e r r u c c o f o r e b m u n 3 T = 149K I c I c 106 103 100 −0.2 0 0.2 0 1 2 I [nA] 3 FIG. 3. Histogram of measured current values for T = 192 K after shifting the maximum to I = 0. The solid line marks a Gaussian fit with zero mean and standard deviation de- termined from the histogram for negative currents. The in- set shows an enlargement to demonstrate the determination of the threshold value Ic used for separating the diffusion- induced signal from the noise (see text). length 40 s, is depicted in Fig. 3. In this figure, the max- imum of the histogram was shifted to I = 0.23 Negative I values are associated with noise and their distribution can be fitted to a half-sided Gaussian. Extending this Gaussian distribution to positive I values yields the solid line in Fig. 3. For small positive I values this curve fits very well the data, implying that these values can also be attributed to noise. The diffusing CuPc molecules cause deviations from the Gaussian distribution at larger I val- ues. To separate the diffusion-induced fluctuations from the noise, we define a threshold value Ic > 0, where the Gaussian histogram of the number of occurrences drops below one, cf. Fig. 3. This threshold is marked by the dashed line in Fig. 2(a) and the corresponding Irec(t) is shown in Fig. 2(b). Using this procedure we determined the ACF of Irec(t) for seven different temperatures inves- tigated in the experiments.19 Normalized ACFs Crec(t)/Crec(0) for two temperatures are shown in Fig. 4(a) (symbols) and fits to them with Eq. (4) and under the constraint t < τR are marked as solid lines. The resulting diffusion coefficients are dis- played in the Arrhenius plot in Fig. 4(b) (symbols), where a least square fit of D = D0 exp(−Ea/kBT ) to these data (solid line) yields an activation energy Ea = 30 meV and a pre-exponential factor D0 = 10−9.4 cm2/s. III. RESIDENCE TIME DISTRIBUTION (RTD) FIG. 1. Relaxed structure of the CuPc molecule. Notice that the terminating bonds are connected to hydrogen atoms. The assigned detection area marked in gray is a circle with radius R = 7.6 A. It corresponds to the smallest circle that entails all nuclei. attachment to islands balance each other. In this quasi- stationary state the diffusing CuPc molecules have an effective small number density c. Then the tunneling cur- rent was measured for a stationary STM tip and constant bias voltage. The height of the tip was preset indirectly by the choice of setpoint current Isetpoint and bias volt- age Ubias. Further details of the experiment are given in Ref. 19. The diffusing molecules cause fluctuations in the tun- neling current I(t) as shown in Fig. 2 (a) for a time in- terval of 2 ms at T = 192 K. A histogram of the tun- nelling current, determined for the whole time series of (a) ] A n [ I (b) c e r I 2 1 0 0.0 1 0 0.0 I c 0.4 0.8 1.2 1.6 t [ms] τ' τ 0.4 0.8 1.2 1.6 t [ms] FIG. 2. (a) Fluctuating STM tunneling current I(t) observed during CuPc diffusion on Ag(100) at T = 192 K (Ubias=1.7 V, Isetpoint=20 pA). The dashed line marks the threshold value Ic, above which the distribution of the current values has no longer a Gaussian shape, see Fig. 3. (b) The rectangular signal Irec(t) = Θ(I(t) − Ic) associated with I(t). The widths of the rectangles give the residence times τ and the distances between the rectangles give the interpeak times τ ′. The residence times τ are the time intervals between entrance and exit of a molecule into the detection region M of the probe, see Fig. 2. By sampling these τ values, the RTD Ψ(τ ) is obtained. For a theoretical description of the RTD we need to tackle the problem of the diffusion of a molecule center in a circle with radius R and absorbing boundaries, see Fig. 5a. Since the molecule center cannot be placed at (a) 10−5 10−4 4 T = 166 K τ R 100 10−1 100 10−1 c e r ) 0 ( C / ) t ( c e r C c e r ) 0 ( C / ) t ( c e r C T = 192 K τ R FIG. 5. Illustration of the geometry used for the calculation of (a) the RTD and (b) the ITD. The dashed lines indicate the uniformly distributed initial probabilities on the rings dis- placed by ǫR (a) and (ǫ′R) (b) from the absorbing boundary (solid lines). MC stands for the diffusing molecule center. 10−5 10−4 t [s] T [K] (b) 222 192 183 174 166 152 140 ] s / 2 m c [ D 10−10 10−11 55 60 65 70 T)−1 [eV] (k B 75 80 FIG. 4. (a) Normalized ACFs Crec(t)/Crec(0) as a function of time for two temperatures (symbols) and best fits with Eq. (4) for t < τR (solid lines). (b) Arrhenius plot of the resulting diffusion coefficients (symbols). A least square fit (solid lines) according to the Arrhenius law yield Ea = 30 meV and D0 = 10−9.4 cm2/s. the absorbing boundary, its initial center position is con- sidered to be at a distance ǫR from the boundary. This means that the initial probability density of the center position is a delta-function concentrated on the ring with radius (1 − ǫ)R. Physically the length ǫR may be associ- ated with an elementary step size of the molecule on the substrate. Due to the symmetric initial condition, the probabil- ity density for the molecule center position r at time t depends on r = r only and is given by (see Appendix) p(r, t) = ∞ Xn=1 J0(αnr/R) πR2 J0(αn(1 − ǫ)) J 2 1 (αn) e−α2 nt/τR , (5) where Jk(.) are the Bessel functions of order k, and the αn are the (positive) zeros of J0, J0(αn) = 0 with 0 < α1 < α2 < . . . With p(r, t) from Eq. (5), the calculation of the RTD amounts to calculating the first passage time distribu- tion with respect to the absorbing boundary. This can be carried out by applying standard techniques from ran- dom walk theory, as described, for example, in Ref. 24. The probability that the molecule center has not left the circle until time t is φ(t) = 2πR R 0 dr rp(r, t). The probability that the center leaves the circle in the time interval [τ, τ + ∆τ ] is φ(τ ) − φ(τ + ∆τ ), implying that the probability density ψ(τ ) for the residence times τ is ψ(τ ) = −2π∂τ R R 0 dr rp(r, τ ). This yields e−α2 nτ /τR . (6) ψ(τ ) = 2 τR ∞ αn Xn=1 J0(αn(1 − ǫ)) J1(αn) We note that the characteristic time (ǫR)2/D sets a lower limit above which Eq. (5) becomes applicable, because one would have to refine the continuum treatment for smaller τ . For the experiments analyzed here, this is not of relevance because such small τ are not resolved. For τ > ∼ (ǫR)2/D the functional form of the solution changes with respect to the characteristic time τR the molecule needs to explore the circle area. For τ ≪ τR, we find ψ(τ ) ∝ 1 τR (cid:18) τ τR(cid:19)−3/2 . (7) This power law can be understood by considering the (negative) time derivative of the probability that the molecule is next to the absorbing boundary at time t [i.e. the efflux rate, which equals ψ(τ )]. For τ ≪ τR this probability is given by the ratio of the explored bound- ary section (∝ √τ ) to the explored area inside the circle (∝ τ ), yielding Ψ(τ ) ∼ −∂τ√τ /τ ∼ τ −3/2. For τ ≫ τR, we find ψ(τ ) ∼ 2 τR α1J0(α1(1 − ǫ)) J1(α1) e−α2 1τ /τR . (8) This result can be interpreted by noting that for τ ≫ τR, where the occupation probability of the molecule is spread over the circle, the efflux rate is essentially constant and given by the inverse of the time τR for a molecule to reach the boundary. A Poisson process with this constant rate yields Ψ(τ ) ∼ τ −1 R exp(−const. τ /τR) Application to measurements For the diffusion of CuPc on Ag(100) the τ values were extracted from the rectangular current signal as de- scribed in Fig. 2. Distributions Ψ(τ ) of these τ values are shown in Fig. 6(a) for two representative tempera- tures. By fitting the exponential tail with Eq. (8), we obtain the diffusion coefficient D. In a self-consistency check we have assured that the tail regime used for the fitting fulfills the requirement τ ≫ τR = R2/D. In addi- tion the value ǫ can be calculated from the prefactor of the fit. We find ǫ ≃ 0.3, which means that ǫR is about the lattice constant a = 2.89 A of the Ag(100) substrate. Inserting D and ǫ, Eq. (6) yields the full distribution ψ(τ ), which is marked by the solid lines in Fig. 6(a). The very good agreement of the theoretical prediction with the measured data demonstrates the reliability of the approach. The diffusion coefficient D is shown in Fig. 6(b) for the two temperatures in (a) together with the five further investigated temperatures in an Arrhenius plot. From the slope of the fitted line, we find an activation energy Ea = 30 meV and a pre-exponential factor D0 = 10−9.4 cm2/s. These values agree with the ones of the ACF analysis. We note, however, that these values deviate from those reported in Ref. 19 (Ea = 81 meV and D0 = 10−8.4 cm2/s). The difference is due to the fact that error bars were taken into account in the analysis performed in Ref. 19. One point in the Arrhenius plot, which was de- termined from the data set measured for T = 222 K, had a particularly small error and was largely influencing the slope of the fit line because of its exposed position with respect to the other points. When excluding this par- ticular point from the fitting, values Ea = 38 meV and D0 = 10−9.7 cm2/s are obtained, in fair agreement with the present analysis. We have refrained from including error bars here because the fits of the RTD for different temperatures yielded comparable errors and small differ- ences between them seem to be insignificant with respect to other possible sources of errors, as, for example, mi- nor temperature fluctuations or the influence of spatial inhomogeneities in diffusion profiles, that are associated with the fact that islands act as sinks for the molecule diffusion. IV. INTER-PEAK TIME DISTRIBUTION (ITD) The inter-peak times τ ′ are the time intervals between the end of a peak and the beginning of the next peak in 5 T = 166 K τ R 192K 10−5 τ R τ [s] T [K] 10−4 105 104 103 105 104 103 102 (a) ) τ ( Ψ ) τ ( Ψ (b) ] s / 2 m c [ D 222 192 183 174 166 152 140 10−10 10−11 55 60 70 65 T)−1 [eV] (k B 75 80 FIG. 6. (a) Residence time distribution for CuPc/Ag(100) at two temperatures (symbols). The solid line marks the re- sult predicted by Eq. (5) after determining the parameters D and ǫ from the exponential tail region τ ≫ τR (cf. Eq. (8) and discussion in text). (b) Arrhenius plot of the extracted diffusion coefficients. A least square fit (solid line) with the Arrhenius law yields an activation energy Ea = 30 meV and a pre-exponential factor D0 = 10−9.4 cm2/s. the rectangular signal, see Fig. 2. The statistics of them is, for small τ ′, dominated by entrance and exit of the same molecule into the detection region M. For calcu- lating the contribution of these return processes to the ITD we analyze the diffusion of a molecule center with initial distance ǫ′R from a circular absorbing boundary with radius R, see Fig. 5. For the probability density of the molecule center to be at position r at time t one obtains (see Appendix) p(r, t) = Z ∞ 0 dα 2πR2 α W0(αr/R, α) (9) W0(α(1 + ǫ′), α) 0 (α) + Y 2 J 2 0 (α) × e−α2t/τR , 6 105 103 101 10−1 105 103 101 (a) ) ' τ ( Ψ ) ' τ ( Ψ (b) T = 166 K τ R T = 192 K 100 10−1 101 100 0.1 0.3 0.5 0.7 τ R 10−5 10−4 10−3 0.02 0.07 0.12 10−1 100 10−2 τ' [s] T [K] 222 192 183 174 166 152 140 ] s / 2 m c [ D (12) 10−10 10−11 55 60 70 65 T)−1 [eV] (k B 75 80 where W0(x, y) = J0(x)Y0(y) − J0(y)Y0(x) and Y0 is the zeroth order Bessel function of second kind. The ITD can be derived analogously to the treat- ment of the RTD by taking the time derivative of the integral of p(r, t) over the outer area with respect to the circle. In the present case it is more convenient to take the flow through the absorbing boundary, ψ(τ ′) = H ds · [−D∇p(r, t)]r=R, which, when making use of the 0 (x)] = 2/πx, yields Wronskian [J ′ 0(x)Y0(x) − J0(x)Y ′ ψ(τ ′) = 2 πτR Z ∞ 0 dα α W0(α(1 + ǫ′), α) 0 (α) + Y 2 J 2 0 (α) e−α2τ ′/τR . (10) For τ ′ ≪ τR, the asymptotic behavior for τ ′ → 0 is τR(cid:19)−3/2 4τ ′ (cid:19)(cid:18) τ ′ exp(cid:18)− ψ(τ ′) ∼ ǫ′2τR ǫ′ . 2τRpπ(1 + ǫ′) (11) Accordingly, ψ(τ ′) rapidly rises for small τ ′ and, after go- max = (ǫ′R)2/6D, ap- ing through a maximum ψmax at τ ′ proaches a power law with exponent (-3/2). This power law has an analogous physical origin as the power law in the RTD, see the discussion in Sec. III after Eq. (7). Here τR is the typical time where the molecule center in Fig. 5 realizes the finite extent of the detection area, or, in other words, where the molecules realizes its size. For τ ′ ≫ τR, Eq. (10) can be approximated by ψ(τ ′) ≃ 2 ln(1 + ǫ′) τ ′ ln2(τ ′/τR) . The asymptotics ∼ (τ ln2 τ )−1 follows from the fact that for large τ ′ the detection area becomes very small with respect to the area explored by the molecule. Accord- ingly, Ψ(τ ′) scales as the probability of first return to the origin of a two-dimensional random walk. The large τ behavior predicted by Eq. (12) is, however, of limited use, because another molecule can enter the detection area before the molecule, which has left this area at last, returns to it. The memory to a molecule that leaves the detection area is lost on time scales of order l2/D ∼ 1/cD. On these time scales different molecules can be regarded as entering the detection area with a constant rate. This rate should scale with the inverse mean time D/l2 for a molecule outside the detection area to enter it. Hence in the limit of large τ ′, an exponential distribution is expected, ψ(τ ′) ∼ cD exp (−κπDcτ ′) , (13) where κ is a constant of order unity. Application to measurements The inter-peak intervals were sampled from the cur- rent signal as explained in Fig. 2 and the ITDs for two temperatures are shown in Fig. 7(a). Note that for very FIG. 7. (a) ITD for CuPc/Ag(100) for two temperatures. The inset shows the exponential decay at large times, where the straight solid lines (dashed lines in the main log-log plots) are fits with Eq. (13), yielding D (see text). Using these D values, the solid lines in the main plots mark the short time behavior after fitting with Eq. (11). (b) Arrhenius plot of the extracted diffusion coefficients for all seven investigated temperatures. A least square fit (solid line) with the Arrhenius law yields an activation energy Ea = 31 meV and a pre-exponential factor D0 = 10−9.6 cm2/s. small times the data indicate a plateau, which is not con- tained in Eq. (11). A refinement of the continuum de- scription would be necessary to capture this behavior, analogous to the very small times in the RTD. For times τ > ∼ τR the power law predicted by Eq. (11) is reflected by the straight line behavior in the double-logarithmic plot, before eventually the exponential decay according to Eq. (13) takes over. To obtain D from the ITD, we first concentrate on the long-time behavior. Setting κ = 1, the rate cD of the exponential decay is provided by the slopes of the lines shown in the insets of Fig. 7(a). The concentration is determined from the probability cπR2 = Pi τi/T for a TABLE I. Strengths and weaknesses of the three methods for determining D. Characteristics ACF RTD ITD Signal processing ⊕ Convenient by FFT ⊖ Peak widths affected by Ic ⊕ Negligible influence of Ic Tip influence Assumptions ⊖ Possible ⊖ Non-interacting particles value ⊖ Possible ⊕ None on interpeak distances ⊕ Less likely ⊕ None in short-time regime ⊖ Non-interacting particles in long-time regime 7 2 molecule to be in the detection area, where T = Pi(τi + τ ′ i ) is the total measurement time. For the seven data sets taken at different temperatures, we find c values in the range 0.6 − 5 × 10−6 A , corresponding to coverages Θ = 0.01 − 0.08% of the diffusing CuPc molecules. The resulting D values are shown in the Arrhenius plot in Fig. 7(b) and yield an activation energy of Ea = 31 meV and a pre-exponential factor of D0 = 10−9.6 cm2/s. Note that the assumed value κ = 1 affects only the prefactor D0 but not the activation energy Ea. With D obtained from the long-time behavior, we can fit the remaining part of the ITD with Eq. (11). Corre- sponding curves are shown as solid lines in the double- logarithmic plots of Fig. 7(a). They give a good agree- ment with the experimental data. The fits yield ǫ′ ≃ 0.5, which is consistent with ǫ ≃ 0.3 obtained in the analysis of the RTD, see Sec. III. Accordingly, we find again that ǫ′R is of the order of the lattice constant a, as one should expect. In principle, the part of the ITD dominated by the single molecule diffusion can also be used to determine D. For this we have to notice that the necessary refinements of the continuum theory to describe the behavior of the ITD left to the maximum 3√6D D a2 . (14) ψmax = pπ(1 + ǫ′)e3/2(ǫ′R)2 ≈ are not expected to yield larger values of the ITD. In fact, when considering jump dynamics of the molecules with a rate D/a2 for short times, the ITD should behave as ∼ (D/a2) exp(−const. Dτ ′/a2), i.e. the largest value of the ITD should be of order D/a2. Due to matching with the continuum treatment, we can identify ψmax with the maxima seen in Fig 7(a). This then is a convenient way to determine D/(ǫ′R)2, and knowing this value, to extract D by fitting the part right to the maximum predicted by Eq. (11). Application of this alternative method indeed yields values for D and ǫ′ in good agreement with those discussed above. V. COMPARISON OF THE METHODS For applications it is important to clarify how the three methods are best combined to obtain most accurate re- sults for D. To this end we need to evaluate the strengths and weaknesses of each method. Let us first note that all methods work with a single input parameter, which is the molecule radius R. All other variables arise from the analyses described in Secs. II-IV. Because τR = R2/D is actually determined, uncertainties with respect to R slightly affect the diffusion coefficient. ∼ τR is governed by single Table I summarizes the advantages and disadvantages of the three methods. The ACF can be readily calculated by a fast Fourier transformation without caring about peak identification in the signal. A disadvantage is that only the short-time regime t < particle diffusion, while an accurate theoretical descrip- tion of the crossover to the long-time regime, governed by collective particle diffusion, requires a careful consid- eration of the mutual exclusion of the molecules (and possibly other interaction effects). The decay of the cor- relation function within this regime can be small, which then affects the accuracy of the D values resulting from the fitting. Another drawback is that the determination of the ACF includes time intervals where molecules are under the tip and possible interactions with the tip can thus have an influence on the diffusion properties. Both the RTD and ITD method require some more effort in preparation of the data, which is connected to the determination of the threshold current Ic and iden- tification of the peaks. Once Ic is set, both the peak widths and interpeak distances can be extracted simul- taneously. Note that any method of determining Ic is associated with some uncertainty. For the ITD this is no problem in practice, because the interpeak intervals are comparatively large. We have checked that, when taking the interpeak distances as time intervals between peak maxima, the results do not change significantly. For the RTD the uncertainty of Ic is a more severe problem. Because molecules diffuse slowly into the detection area, the peaks in the original tunneling current signal have rather flat flanks. As a result the peak widths change more sensitively with Ic than the interpeak intervals. An advantage of the RTD is that D can be determined solely by analyzing the exponential tail for large residence times. One should note, however, that it may be diffi- cult to obtain a good statistics in the tail regime, when the molecules are highly mobile or small. In this case the peaks are narrow and it could be difficult to resolve them accurately. In the RTD method the interaction with the tip can influence the residence times and in this case one would not determine the free diffusion of the molecules on the substrate. By systematically changing the bias voltage, a possible influence can be reduced to a minimum.19 A strength of the RTD is that it is related to a single-particle problem. The ITD method has the advantage that tip-molecule interactions can be expected to have, if at all, a marginal influence on the interpeak times (irrespective of tip- substrate distances or bias voltages). For small interpeak times, the ITD is essentially also related to a single- particle problem. For large interpeak times, an approx- imate value of D can be obtained based on an estimate for processes, where one molecule in the detection area under the tip is followed by another molecule. The values obtained in this way turn out to be close to those result- ing from the other methods. Compared to the residence times, the interpeak times are quite large and are thus less prone to the experimental time resolution and the threshold value Ic. VI. CONCLUSIONS In order to exploit the strength of each method one can combine the different methods, if the corresponding data in the relevant time regime have sufficient statis- tics. For a first classification of the measurement it is helpful to choose the ACF method, because it yields fast results without analyzing the time series in detail. In a subsequent step the peak widths and interpeak inter- vals should be extracted with the procedure described in Sec. II. We then recommend to use the RTD method for determining D whenever the exponential tail is suffi- ciently pronounced and the tip influence on the diffusion can be neglected. Otherwise the ITD method should be preferred. In any case, one should always apply both the RTD and the ITD method to perform consistency checks and to obtain most reliable results. Our analysis for CuPc on Ag(100) shows that, despite diffusion coef- ficients extracted from the different methods may differ for one or another sample, the activation energies deter- mined from the Arrhenius plots should have comparable values, see Fig. 8. An interesting question is whether the methods dis- cussed here can be taken over to other fields. Obviously, the ACF method is well known in fluorescence correla- tion spectroscopy (FCS),25 where intensity fluctuations reflect concentration fluctuations, typically in some fi- nite detection volume. Similarly, analysis of density fluc- tuations has been applied as one variant to determine diffusion coefficients in Field Emission Microscopy.2 In the situation discussed in Sec. II only one particle is in the detection area and accordingly information on the tagged particle diffusion (tracer diffusion) is obtained. In the hypothetical limit of very small particle concentra- tions in FCS, where the mean interparticle distance be- comes larger than the linear size of the detection volume, one should essentially recover the behavior discussed in 8 T [K] 222 192 183 174 166 152 140 10−10 ] s / 2 m c [ D 10−11 50 55 60 70 65 T)−1 [eV] (k B 75 80 85 FIG. 8. Comparison of the diffusion coefficients obtained from the three methods. Squares, dashed line: ACF method (Ea = 30 meV, D0 = 10−9.4 cm2/s); Circles, dotted line: RTD method (Ea = 30 meV, D0 = 10−9.4 cm2/s); Dia- monds, dashed-dotted line: ITD method (Ea = 31 meV, D0 = 10−9.6 cm2/s). Sec. II (with R then playing the role of the detection length). In common applications of this method one is, however, not able to extract the tagged particle infor- mation and this makes a difference to the ACF method discussed in Sec. II. This also prevents the use of the RTD and ITD in the analysis of typical fluorescence sig- nals. These methods can yet become useful in light of the ongoing development of sophisticated techniques to measure single-molecule properties. In summary, we provided formulas for different means of analyzing fluctuations in the signal of a locally fixed probe on a surface, where the fluctuations are caused by diffusing particles. By exploiting the fact that the molecule extension enters as a scale bar into the equa- tions, a quantitative determination of the diffusion con- stant becomes possible. So far, we have considered cir- cular shapes for planar molecules. The situation will ev- idently be more involved when rectangular or even more complicated shapes are considered. In addition rotational diffusion and anisotropies induced by the substrate then can modify the results. As mentioned above, the inter- action of the STM tip with the diffusing molecule was neglected so far. The interaction is influenced by the ex- perimental setup (tip shape, tip-substrate distance etc.) and it is specific for the tip and molecule material. One objective could be to minimize the tip influence exper- imentally. On the other hand, it could be interesting to study how the interaction potential affects the diffu- sion of the molecule. This potential has so far not been included in the analysis, but it should be possible and bears an interesting route to obtain information on the tip-molecule interaction. ACKNOWLEDGMENTS J.I. and M.S. gratefully acknowledge fincancial support by the SFB 624 of the Deutsche Forschungsgemeinschaft. Appendix: Diffusion propagators for RTD and ITD the expansion coefficients Cn follow from the initial con- dition Eq. (A.1): 9 The results given in Eqs. (5) and (9) for p(r, t) have been derived earlier in the literature, in particular Eq. (9) by using the Heaviside method.26 We give a straightfor- ward derivation here, using separation of variables and eigenfunction expansions. Let us first consider the equation for diffusion of a par- ticle in a circular stripe a < r < c (a here not equal to the lattice constant used in the main text) with absorbing boundaries and initial distribution p0(r) = 1 2πr δ(r − b) , a < b < c . (A.1) Due to the radial symmetry p(r, t) = ρ(r, t)/2π, where ρ(r, t) satisfies the radial diffusion equation ∂ρ(r, t) ∂t = D(cid:18) ∂2 ∂r2 + 1 r ∂ ∂r(cid:19) ρ(r, t) (A.2) with ρ(a, t) = ρ(c, t) = 0 and ρ(r, 0) = δ(r − b)/r. Making the product ansatz ρ(r, t) = ρ(r, 0)g(t), Eq. (A.2) separates in the variables r and t. One obtains g(t) = e−λ2Dt, where the allowed values for λ2 are the eigenvalues of the radial Laplace operator in the circular stripe, (cid:18) ∂2 ∂r2 + 1 r ∂ ∂r(cid:19) Ψn(r) = −λ2 nΨn(r) , Ψn(a) = Ψn(c) = 0 . (A.3) Because the Laplacian is negative definite, λ2 n > 0. The eigenfunctions Ψn are given by linear combinations of the zeroth order Bessel functions J0(.) and Y0(.) of first and second kind, Ψn(r) = AnJ0(λnr) + BnY0(λnr) (where restriction to λn > 0 gives linear indepen- dent eigenfunctions). The Dirichlet boundary condi- tions yield AnJ0(λna) = −BnY0(λna) [or AnJ0(λnc) = −BnY0(λnc)], and J0(λna)Y0(λnc) − J0(λnc)Y0(λna) = 0 (A.4) as determining equation for the λn, n = 1, 2, . . . (0 < λ1 < λ2 < . . .). Introducing the cross product W (x, y) = J0(x)Y0(y) − J0(y)Y0(x) , (A.5) the solution becomes ρ(r, t) = ∞ Xn=1 CnW0(λnr, λna)e−λ2 nDt , (A.6) where W0(λna, λnc) = 0. Utilizing the orthogonality of the eigenfunctions, Z c a dr r W0(λmr, λma)W0(λnr, λna) = δmnZ c a dr r W 2 0 (λnr, λna) , (A.7) Cn = R c a dr r W0(λnr, λn, a)ρ(r, 0) R c a dr r W 2 W0(λnb, λna) R c a dr r W 2 The result for p(r, t) thus is = 0 (λnr, λna) 0 (λnr, λna) . (A.8) p(r, t) = ∞ Xn=1 W0(λnb, λna)W0(λnr, λna) 0 (λnr, λna) a dr r W 2 2πR c e−λ2 nDt . (A.9) The diffusion propagator in Eq. (9) relevant for the ITD corresponds to the limit c → ∞, where the spectrum of eigenvalues determined by Eq. (A.4) becomes continuous. Analogous to the change of a Fourier series to a Fourier integral, we are led to consider the Weber transform27 of a function f = f (r) F (λ) = 1 a2 Z ∞ a dr r W0(λr, λa)f (r) , (A.10) with back-transformation f (r) = a2Z ∞ 0 dλ λ W0(λr, λa) J 2 0 (λa) + Y 2 0 (λa) F (λ) . (A.11) Accordingly, Eq. (A.6) becomes ρ(r, t) = Z ∞ 0 dλ C(λ)W0(λr, λa) e−λ2Dt , (A.12) where C(λ) = = This yields λ 0 (λa) Z ∞ a 0 (λa) + Y 2 J 2 λW0(λb, λa) 0 (λa) + Y 2 J 2 0 (λa) drrW0(λr, λa)ρ(r, 0) . (A.13) p(r, t) = Z ∞ 0 dλ 2π λ W0(λr, λa)W0(λb, λa) J 2 0 (λa) + Y 2 0 (λa) e−λ2Dt . (A.14) Equation (9) follows by setting a = R, b = (1 + ǫ)R, and λ = α/R. For the diffusion propagator in Eq. (5) relevant for the RTD only one boundary condition ρ(c, t) = 0 has to be taken into account. In this case the Bessel func- tions of second kind cease to apply, because their log- arithmic singularity at the origin eliminates them from the space of functions, where the radial Laplace oper- ator is Hermitean. Notice that the logarithmic singu- larity would be no problem with respect of the integra- bility of p(r, t). The eigenfunctions thus are given by Ψn(r) = AnJ0(λnr), where the λn are determined by J0(λnc) = 0. Equation (A.6) becomes ρ(r, t) = ∞ Xn=1 AnJ0(λnr) e−λ2 nDt , (A.15) and the An are again determined by the initial condition, corresponding to an expansion of ρ(r, 0) into a Fourier- Bessel series, This yields p(r, t) = 1 πc2 ∞ Xn=1 J0(λnr)J0(λnb) J 2 1 (λnc) e−λ2 nDt . (A.17) 10 An = R c R c 0 dr r J0(λnr)ρ(r, 0) 0 dr r J 2 0 (λnr) = 2J0(λnb) c2J 2 1 (λnc) . (A.16) Equation (5) follows by setting c = R, b = (1 − ǫ)R and λn = αn/R. ∗ [email protected] 1 E. W. Mulller, Zeits. f. Physik 108, 668 (1938). 2 R. Gomer, Rep. Prog. Phys. 53, 917 (1990). 3 E. W. Muller and T. T. Tsong, Field Ion Microscopy 537 (1995). 16 M. Sumetskii and A. A. Kornyshev, Phys. Rev. B 48, 17493 (1993). 17 M. Sumetskii, A. A. Kornyshev, and U. Stimming, Surf. (American Elsevier Publishing Company, 1969). Sci. 307-309, 23 (1994). 4 R. Viswanathan, D. Burgess, P. Stair, and E. Weitz, J. 18 M. C. Tringides and M. Hupalo, J. Phys. Condens. Matter Vac. Sci Technol. 20, 605 (1982). 5 S. George, A. Desantolo, and R. Hall, Surf. Sci. 159, L425 (1985). 6 E. Ganz, S. K. Theiss, I.-S. Hwang, and J. Golovchenko, Phys. Rev. Lett. 68, 1567 (1992). 7 R. Gomer, Appl. Phys. A 39, 1 (1986). 8 B. S. Swartzentruber, Phys. Rev. Lett. 76, 459 (1996). 9 J. V. Barth, Surf. Sci. Rep. 40, 75 (2000). 10 H. Brune, Surf. Sci. Rep. 31, 125 (1998). 11 J. W. Evans, P. A. Thiel, and M. C. Bartelt, Surf. Sci. Rep. 61, 1 (2006). 12 M. Einax, S. Ziehm, W. Dieterich, and P. Maass, Phys. Rev. Lett. 99, 016106 (2007). 13 M. Einax, W. Dieterich, and P. Maass, J. Appl. Phys. 105, 054312 (2009). 14 A. Kuhnle, Curr. Opin. Colloid Interface Sci. 14, 157 (2009). 15 M. L. Lozano and M. C. Tringides, Europhys. Lett. 30, 22, 264002 (2010). 19 J. Ikonomov, P. Bach, R. Merkel, Phys. Rev. B 81, 161412(R) (2010). 20 J. D. Phys. Rev. Lett. 50, 1998 (1983). Tersoff and and M. Sokolowski, R. Hamann, 21 A. Mugarza, R. Robles, C. Krull, R. Koryt´ar, N. Lorente, and P. Gambardella, Phys. Rev. B 85, 155437 (2012). 22 P. Bach, diploma thesis, University of Bonn (2009). 23 The offset corresponding to the shift is associated with the preset setpoint current that is mainly related to tunneling to/from the bare Ag surface. 24 S. Redner, A guide to first-passage processes (Cambride University Press, 2001). 25 E. P. Petrov and P. Schwille, State of the Art and Novel Trends in Fluorescence Correlation Spectroscopy (Springer, 2008) pp. 145 -- 197. 26 S. Goldstein, Proc. London Math. Soc. 34, 51 (1932). 27 B. Davies, Integral Transforms and Their Applications (Springer, 1978).
1209.5502	1	1209	2012-09-25T06:13:52	A subKelvin scanning probe microscope for the electronic spectroscopy of an individual nano-device	[ "cond-mat.mes-hall", "cond-mat.supr-con", "physics.ins-det" ]	We present a combined scanning force and tunneling microscope working in a dilution refrigerator that is optimized for the study of individual electronic nano-devices. This apparatus is equipped with commercial piezo-electric positioners enabling the displacement of a sample below the probe over several hundred microns at very low temperature, without excessive heating. Atomic force microscopy based on a tuning fork resonator probe is used for cryogenic precise alignment of the tip with an individual device. We demonstrate the local tunneling spectroscopy of a hybrid Josephson junction as a function of its current bias.	cond-mat.mes-hall	cond-mat	A subKelvin scanning probe microscope for the electronic spectroscopy of an individual nano-device T. Quaglio, F. Dahlem, S. Martin, A. G´erardin, C. B. Winkelmann, and H. Courtois Institut N´eel, CNRS, Grenoble INP and Universit´e Joseph Fourier, 25 rue des Martyrs, BP 166, 38042 Grenoble, France. (Dated: May 31, 2018) We present a combined scanning force and tunneling microscope working in a dilution refrigerator that is optimized for the study of individual electronic nano-devices. This apparatus is equipped with commercial piezo-electric positioners enabling the displacement of a sample below the probe over several hundred microns at very low temperature, without excessive heating. Atomic force microscopy based on a tuning fork resonator probe is used for cryogenic precise alignment of the tip with an individual device. We demonstrate the local tunneling spectroscopy of a hybrid Josephson junction as a function of its current bias. PACS numbers: I. INTRODUCTION Scanning Tunneling Spectroscopy (STS) has been ex- tensively used in the context of condensed matter physics for the local study of fundamental electronic properties including superconductivity, magnetism, Schockley sur- face states. Operation at very low temperature [1 -- 4] im- proves the spectroscopic energy resolution, with values down to 10 µeV being demonstrated [5]. Recently, new cryogenic scanning tunneling microscopes enabling tun- neling studies under out-of-equilibrium conditions have been developed [6, 7]. The next step is to build instru- ments for the scanning tunneling spectroscopy of an indi- vidual nano-scale device. As electron tunneling is inop- erative on an insulator, locating at the submicron scale a device sitting on an insulating substrate requires the use of Atomic Force Microscopy (AFM). AFM and STS can be combined on a single probe by using a quartz resonator, which remains stiff at low frequency, function- nalized with a metallic tip allowing to collect a tunnel current [6, 8]. As the initial tip approach is usually re- alized at a random distance from the sample, one needs to realize large lateral displacements at low temperature. Due to the difficulty of combining these different require- ments, the tunneling spectroscopy of an individual nano- device, that can be current-biased, has to the best of our knowledge not yet been demonstrated. In this paper, we demonstrate the local spectroscopy of an individual current-biased superconducting nano- device in the subKelvin temperature range. In this scope, we have developed a home-made combined AFM/STM based on a quartz tuning fork probe and equipped with commercial piezo-electric positioners. We describe the very low temperature operation, calibration and ther- mal behavior of these positioners. An optimized pro- cess for in-situ single nano-object localization is pre- sented. As a proof of concept, we discuss a local spec- troscopy experiment on a Superconductor-Normal metal- Superconductor (SNS) Josephson junction under current bias. (a) i ii iii iv v vi 1cm (b) (c) VT 2mm IT I Tunnel current amplifier Differential amplifier V FIG. 1: a) View of the cryogenic microscope head sus- pended on a 10 cm-high copper frame: (i) piezoelectric scan- ner tubes, (ii) AFM/STM probe, (iii) sample holder, (iv) micro-coax filters, (v) Attocube piezo-electric positioners and (vi) mechanical isolation springs. b) Zoom-in picture on the AFM/STM probe showing the tungsten tip glued by silver epoxy on a quartz tuning fork, here without its shielding box. c) Schematic of the circuit connecting the SNS sample, en- abling to simultaneously perform I(V ) transport experiments and local tunnel spectroscopies IT (VT ). II. THE CRYOGENIC MICROSCOPE Fig. 1a shows the overview of the microscope head that is suspended on a copper frame through six springs. It includes an arrangement of piezo-electric tubes at the top and a set of piezo-electric positioners at the bottom. The microscope head top part (number i in Fig. 1a) holds the tip. It is made of two concentric EBL2 piezo- electric tubes connected in series but in opposite direc- tions. The outer tube is used to adjust the tip-sample dis- tance and, in the AFM mode, to mechanically excite the tuning fork. The inner tube controls the lateral scanning of the probe over the sample surface. The tube lengths are 1.25 and 2.5 cm, diameters are 1.25 and 0.62 cm, wall thicknesses are 1 and 0.5 mm, respectively. For a 220 V applied voltage on the inner tube and at low temperature, the maximum expansion in Z direction is about 1.2 µm and the X-Y scanning area is about 6.2 × 6.2 µm2. We took special care when gluing the piezoelectric tubes to their metallic supporting parts, since even a small elec- tric leakage to the ground may, under high voltage bias, create an important Joule heat, which could heat signif- icantly the cryogenic system. A set of three piezo-electric positioners from Attocube [9] (number v in Fig. 1a) is used to move the sample under the probe for coarse positioning in the three space directions. From bottom to top, one finds the Z (vertical) direction positioner, the Y and the X (horizontal) direc- tion ones. The operation of this system will be described in detail in section III. The sample holder (number iii in Fig. 1a) is made from a thermal clad plate that is screwed to the upper piezo-electric positioner. The sam- ple is fixed on the sample holder with Apiezon N vacuum grease, which offers a good thermal link and leaves the possibility to easily remove the sample. Sample on-chip electrical contacts are connected to millimeter-sized cop- per pads defined on the sample holder through micro- bonded 30 µm-diameter Al wires. During low temperature operation, the whole micro- scope is enclosed in a hermetic copper chamber that is mounted on the cold plate of an inverted dilution refriger- ator. This refrigerator called Sionludi [10] is compact and thus stiff. It presents a large available volume of about 1 dm3 at very low temperature that is most suitable for the operation of a scanning probe microscope. Its cool- ing power is about 30 µW at 100 mK [1]. The cryostat sits on on a set of four damped legs that filter mechanical vibrations from the ground floor with a cut-off at about 2 Hz. Special care is taken for proper thermalization of the whole system. A 10−1 mbar of Helium exchange gas in- serted in the microscope chamber before cool-down im- proves thermalization of the whole system down to about 4.2 K. In order to ensure a proper thermalization and a good electrical ground, every stage of the positioners tower is connected, through a copper wire, to the base of the microscope frame. In this way, the temperature on the sample stage reaches a base temperature below 100 mK. The copper chamber also protects the measure- ment set-up against electromagnetic radiation. In order to ensure proper electronic thermalization, every elec- trical line is filtered at the microscope chamber input through lossy micro-coaxial cables [11] that are thermally anchored to the refrigerator cold plate. III. THE AFM-STM PROBE The local probe (Fig. 1b and number ii in Fig. 1a) is a quartz tuning fork of 32768 Hz bare resonance frequency with an etched tungsten tip glued on it. The tuning fork 2 oscillation is electrically detected. Compared to an opti- cal detection, this approach is well adapted to very low temperatures: no optical window is necessary and dissi- pation is less than 1 nW [13, 14]. The original metallic casing of the tuning fork is usually kept in order to shield the tip from electromagnetic noise and to simplify its glu- ing to the probe holder. In this case, only a little hole is drilled into it to access the fork electrode on which the tip will be glued. A sharp and stable tip is crucial to perform AFM on small structures with a large relief. The tip is produced by a 2 minutes electrochemical etch of a 150 µm-diameter tungsten wire in a 4 mol/l KOH solution [15]. The typ- ical geometry of the tip apex is conical with an apex curvature of about 20 nm. The tip wire is then glued with silver epoxy [16] on one electrode of the tuning fork. This silver epoxy guarantees both an ohmic contact and a very stiff gluing. A relatively large mass of silver epoxy is used, so that a resonance frequency of about 30 kHz and a quality factor Q of about 7000 is obtained in air at room temperature. This routinely gives a quality factor of about 105 at low temperature under cryogenic vacuum. Obtaining larger Q values is possible but this was found to offer no useful increase of sensitivity while bringing oscillation instabilities during imaging. As a tip, we have also used commercial AFM cantilever-tips coated with 25 nm of platinum-iridium. By using a very thin layer of silver epoxy and a micro manipulator, the cantilever can easily be attached by capillarity to the end of the tun- ing fork and broken at the desired position, thus leaving a pyramidal and metal-coated tip glued on the tuning fork prong. The latter method can bring a more stable tip geometry at the expense of the risk of loosing tun- nel contact if the tip metallization gets damaged. This probe is finally glued again with silver epoxy on a cop- per plate, which is screwed to the piezo-electric scanner tube. This configuration offers proper mechanical cou- pling to the excited piezo-electric tube, which avoids any unwanted resonance frequency drift. The control of the tuning fork oscillation is made through a Phase-Locked Loop included in a SPM elec- tronics [17] that is also used to control the whole system, including coarse approach and positioning, imaging and spectroscopy measurements. IV. COARSE POSITIONING OPERATION Our piezo-electric positioners [9] operation is based on a slip-stick motion. A sawtooth voltage with an ampli- tude up to 70 V is applied with a repetition rate of 1 ms. The decay time of the sawtooth fast slope is about 1 µs. Due to the large capacitance of the piezo-electric actua- tor, a small line resistance is required in order to maintain a small RC time and thus an efficient transfer of the sig- nal waveform to the piezo-electric actuator. Between the room temperature stage and the cold plate, we use a 1- meter cable of 12 constantan wires of 100 Ω resistance. 3 FIG. 2: Operation of the coarse positioning system at a bath temperature of 90 mK. a) AFM topographical image of a tracking symbol. b) Superposed images after a move in - Y and + Y directions, with 8 pulses of 45 V amplitude. c) Superposed images after a move - X and + X movements, with 3 pulses of 45 V amplitude. d) Temperature evolution of the refrigerator cold plate due to a X, Y or Z piezo-electric positioners displacement of 10 steps at a 45 V bias amplitude. FIG. 3: Scanning electron microscopy image showing the pattern of different symbols used to localize by AFM the sam- ple at low temperature. Added on top of this image, the AFM pictures and associated numbers represent the different steps of a typical localization procedure performed at 90 mK. On the left: zoom on one of the low resolution image recorded during this procedure. At low temperature, the piezo-electric actuator capaci- tance is only 22 or 100 nF for the X, Y or Z positioners respectively, so that 2 or 4 wires in parallel are enough to maintain an efficient operation. In order to reduce the number of required wires, the reference electrode of every motor is put to ground at the microscope copper frame. At room temperature, the larger capacitance of the piezo-electric actuators leads us to shunt the main wiring described above with a copper wires cable, which is used for the first rough adjustment of the tip above the device in air. Before any transport or tunneling spec- troscopy measurement, every positioner is connected to the electrical ground in order to avoid electrical noise. Sample displacements have been calibrated using a substrate pattern made by electron beam lithography. Fig. 2a illustrates a calibration sequence performed below 100 mK on one single symbol having a well-defined size. For X-translation (Fig. 2b) and Y-translation (Fig. 2c), only few pulses of 45 V are needed to move the sample on a 100 nm scale. Back and forth movements are in aver- age reproducible at this scale. For the same translation, less pulses are needed for the X-stage that is at the top of the piezo-electric tables set. The Z-movement calibra- tion (not shown here) was done via the distance variation of the tunnel contact position adjusted by the displace- ment of the calibrated piezo-electric scanner tube. Little difference between the top and down z-movements was found. In general, the main obstacle to displacements is given by the positioner electrical wires, which have to be positioned carefully around the microscope head. Fig. 2d shows the thermal response of the dilution re- frigerator mixing chamber after a series of ten steps of each positioner with a 45 V bias amplitude. Starting from temperature of 135 mK, the temperature increases of only 10 to 20 mK, whereas the temperature of the sam- ple stage jumps to about 500 mK (not shown here). A few minutes are needed for the system to recover is base temperature after the sample move. Furthermore, it is possible to quickly move the sample over 10 micrometers by several hundred steps (of 45 V amplitude), while keep- ing the refrigerator stable at a temperature below about 1 K. These experimental observations are compatible with basic thermal considerations. The positioners' nominal friction force of 5 N, a step size of about 30 nm and a displacement rate of 1 kHz determine a mechanical power of about 150 µW during movement. At a 45 V bias, the charging energy CU 2/2 stored in the 22 nF capacitance of the X or Y positioner is 560 µJ. Under voltage pulses at a frequency f , the heat generation is 4π CU 2f tan δ, where the loss angle δ is about 1 degree at low temper- ature [18]. This gives a value of about 780 µW . The total load of about a milliWatt is compatible with the estimated cooling power at 500 mK. Some extra heat dissipation can also originate from Ohmic losses in the electrical wires used to bias the piezoelectric positioners. V. SINGLE DEVICE LOCALIZATION METHOD A prerequisite for performing a local spectroscopy of an individual nano-device is to align the scanning probe with the device. For this purpose, the probe is first roughly aligned with the device at room temperature, using a binocular microscope. This is easily feasible due to the open design of the microscope head and the tip reflection visibility on the silicon substrate. While cooling down the system, the tip drifts by approximately 50 µm due to the different thermal coefficients of the microscope com- ponents. We have therefore developed a reliable method for a precise tip alignment at low temperature, in practice at a temperature of about 1 K when thermal contractions are complete and the He pressure inside the chamber is low enough to avoid any sparkling. The method has been optimized for a minimum AFM imaging time, which is essential for saving the tip apex integrity for tunneling experiments. In this scope, the sample chip was previously covered with different kinds of symbols of micrometer size over a square area of 150 ×150 µm2 centered on the nano-device of interest. The pattern design was defined such that one full symbol is always present within the low temperature scanning range. The patterned region is divided into 10 different zones, the 50 × 50 µm2 central zone being it- self divided into 10 different smaller zones. Moreover, the patterns array is symmetric with respect to the two current leads. Each zone features a different symbol ge- ometry that can be easily recognized in low-resolution AFM images. The individual pattern geometry is comb- shaped with a variable number of teeth. Typically, af- ter 10 lines of scan with an inter-distance of 500 nm or 100 nm, the symbol recognition is achieved. Using the piezo-electric tables calibration, the tip is aligned above the nano-device by dichotomy after a maximum of ten AFM images. Fig. 3 shows a localization process case achieved in six steps, meaning that in this case only 60 scan lines in total were needed to find the nano-device. Once the device is located, the tip is moved over the metallic device, the tuning fork is let static so that elec- tron tunneling spectroscopy can be performed. VI. LOCAL SPECTROSCOPY ON A CURRENT-BIASED PROXIMITY JOSEPHSON JUNCTION We have used the instrument and the method de- scribed above for the study of Superconductor-Normal metal-Superconductor (SNS) Josephson junctions. Sam- ples were realized by shadow evaporation technique under ultra high vacuum on a Si wafer covered with 300 nm- thick SiO2. A Josephson junction is made of a 1.2 µm- long, 0.3 µm-wide and 35 nm thick copper island bridg- ing two 35 nm-thick aluminum leads. The two Al/Cu overlapping regions have a length about 0.2 µm each. Low-power ultrasound and oxygen plasma cleaning were performed at the end of the lift-off process in order to en- sure a clean surface. These precautions are helpful since traces of resist can be gathered by the tip during AFM measurements and affect the quality of both AFM images and tunnel spectroscopies. The AFM image in Fig. 4a shows the geometry and structure of the SNS junction. 4 FIG. 4: Typical measurements obtained with our cryogenic AFM/STM on a hybrid Josephson junction at 90 mK. a) AFM image of the Al/Cu/Al junction. The cross indicates where spectroscopies in out-of-equilibrium conditions were re- alized, see Fig. 5b. b) Color plot of local spectroscopic data dIT /dVT (VT ) performed every 100 nm along the dotted arrow with a tunnel resistance of 10 MΩ. Fig. 5a shows a typical current-voltage characteristics I(V ) obtained at 90 mK on one of our samples. As ex- pected, the data show a superconducting branch with a critical current of about 600 nA. At higher bias, the voltage reaches an Ohmic behavior with a slope given by the normal-state resistance of the junction of 4.5 Ω. We observe an hysteretic behavior that is characteristic of heating effects in Josephson junctions with a large criti- cal current [19]. At a larger bias range, see Fig. 5b, we observe the transition of the Al lead from the supercon- ducting state to the resistive state. Again, this transition shows thermal hysteresis. Local tunneling spectroscopies of the equilibrium lo- cal density of states (LDOS) were performed at various locations on the sample. The tunnel resistance between the tip and the device surface was fixed to 10 MΩ and the tunnel current IT was measured when sweeping the voltage between tip and sample VT . The data IT (VT ) was afterwards numerically differentiated as the tunnel differential conductance dIT /dVT (VT ) gives a measure of the local electronic density of states, smeared by tem- perature. Fig. 4b 2D-color plot shows a set of differen- tial conductance spectroscopies performed every 100 nm along the junction length. We can clearly see how the superconducting gap in every Aluminum lead (deep blue region) decreases in width when approaching the Al/Cu overlap region and finally vanishes in the copper metal part. The spatial evolution of the measured spectra along (a) 0 0 (c) 4 4 I (m A) I ( A)μ 8 8 ) V μ ( V (b) 4 2 0 -2 -4 -1 0 I ( A)μ 1 ) V m V ( 6 6 4 4 2 2 0 0 ) V m ( V 1.5 T V d / T I d N R 1 0.5 I= 0 Aμ 6 Aμ 7 Aμ -0.4 -0.2 0 0.2 0.4 0.6 V (mV) T 0 -0.6 FIG. 5: Typical measurements obtained with our cryogenic AFM/STM on a hybrid Josephson junction at 90 mK. a) and b) Current-voltage characteristics I(V ) of the junction measured as described in Fig. 1c and over two different bias current ranges. Some hysteresis is observed. c) Local spec- troscopic data dIT /dVT (VT ) measured on one aluminum lead, see Fig. 4, while current-biasing the sample close to the local critical current of the lead. the X-axis was shown to be reproducible with a fidelity of at worst 50 nm even when the tunnel contact was es- tablished through the oxide covering an aluminum lead. Tunneling spectroscopies measured on the supercon- ducting Al leads are shown in Fig. 5c. The spectrum measured at zero bias current can be fitted assuming the Bardeen, Cooper and Schrieffer (BCS) density of states expression with a superconducting gap of 240 µeV and an electronic temperature of 270 mK for the tip. This in- creased temperature compared to the measured sample temperature could be understood as due to some residual electronic noise in the set-up. In the normal metal bridge of a SNS junction, the- ory predicts a density of states with a mini-gap given by 3.1 times the Thouless energy ET h = ¯hD/L2 [3, 20], where D is the electronic diffusion constant and L is the junction length. Assuming a typical value D = 70 cm2 for Cu, taking L = 1 µm, we obtain an expected mini- gap value of 14 µeV , which is not observed. However, the thermal broadening due to the effective tip temper- 5 ature 2kBT=46 µeV (at T = 270 mK) also determines the energetic resolution of our measurement and signif- icantly smears smaller structures. The copper oxide at the surface may also destroy very locally the quantum correlations present inside the normal metal bridge. Finally, we have investigated the effect of a current bias on the local spectroscopies on the normal metal bridge as well as on the superconducting electrodes region. Fig. 5c shows a series of local spectroscopies realized on one of the Al leads close to the junction, see Fig. 4. No sizeable effect is detected until a current of about 6 µA, where the heating from the normal island becomes enough to affect locally superconductivity in the closely Al lead. At a bias current of 6 µA, the superconducting gap decreases from 240 µeV at equilibrium down to about 200 µeV . For a current of 7 µA, the aluminum leads turn fully normal. The retrapping current of the superconducting wire measured in transport is actually 6.5 µA, see Fig. 5a . This last experiment demonstrates the possibility to perform the local electronic spectroscopy of a quantum device under actual operation. VII. CONCLUSION We have set up and characterized a home-made scan- ning probe microscope working in a dilution refrigerator that combines scanning force microscopy and tunneling spectroscopy, by using a tuning fork resonator function- nalized with a metallic tip. This system has the speci- ficity to be equipped with commercial piezo-electric po- sitioners whose operation is demonstrated to be compat- ible with subKelvin temperatures. We have developed a protocol based on a specially patterned substrate to position the tip over a single object at very low tempera- ture. Thanks to this experimental setup, we succeeded in realizing the scanning tunneling spectroscopy of a nano- device biased with an electrical current. VIII. ACKNOWLEDGMENT The authors thank S. Rajauria, L. M. A. Pascal, T. Crozes, A. Fay, R. Stomp, A. Schmalz and S. Falk for discussions. Samples have been fabricated at Nanofab fa- cility at Institut N´eel. We acknowledge the support from Grenoble Nanoscience Foundation and MicroKelvin, the EU FRP7 low temperature infrastructure grant 228464. [1] N. Moussy, H. Courtois and B. Pannetier, Europhys. Lett. 55, 861 (2001); Rev. Sci. Instrum. 72, 128 (2001). [2] H. Suderow, I. Guillam`on, and S. Vieira, Rev. Sci. In- strum. 82, 033711 (2011). Est`eve, Phys. Rev. Lett. 100, 197002 (2008). [4] Y. J. Song, A. F. Otte, V. Shvarts, Z. Zhao, Y. Kuk, S. R. Blankenship, A. Band, F. M. Hess, and J. A. Stroscio, Rev. Sci. Instrum. 81, 121101 (2010). [3] H. le Sueur, P. Joyez, H. Pothier, C. Urbina, and D. [5] H. Le Sueur and P. Joyez, Rev. Sci. Instrum. 77, 123701 6 (2006). (1995). [6] J. Senzier, P. S. Luo, and H. Courtois, Appl. Phys. Lett. 90, 043114 (2007). [7] A. Maldonado, I. Guillam`on, H. Suderow and S. Vieira, Rev. Sci. Instrum. 82, 073710 (2011). [8] R. H. M. Smit, R. Grande, B. Lasanta, J. J. Riquelme, G. Rubio-Bollinger et N. Agrait, Rev. Sci. Instrum. 78, 113705 (2007). [9] ANPxyz100 from Attocube, Munchen. [10] This cryostat has been designed at Institut N´eel-CNRS [14] J. Rychen, T. Ihn, P. Studerus, A. Herrmann, K. En- sslin, H.-J. Hug, P. J. A. Van Schendel and H. J. and Guntherodt, Appl. Surf. Sci., 157, 290 (2000). [15] M. Kulawik, M. Nowicki, G. Thielsch, L. Cramer, H.-P. Rust, H.-J. Freund, T. P. Pearl and P. S. Weiss, Rev. of Sci. Instrum. 74, 1027 (2003). [16] Epo-Tek H21D [17] Nanonis - SPECS, Zurich. [18] C. Bodefeld, Technical note: cooling of attocube systems' by A. Benoit and M. Caussignac. positioners (2009). [11] D. C. Glattli, P. Jacques, A. Kumar, P. Pari, and L. [19] H. Courtois, M. Meschke, J. T. Peltonen and J. P. Pekola, Saminadayar, J. Appl. Phys. 81, 7350 (1997). Phys. Rev. Lett. 101, 067002 (2008). [12] R. D. Grober, J. Acimovic, J. Schuck, D. Hessman, P. J. Kindlemann, J. Hespanha, A. S. Morse, K. Karrai, I. Tiemann, and S. Manus, Rev. Sci. Instrum. 71, 2776 (2000). [13] K. Karrai and R. D. Grober, Appl. Phys. Lett. 66, 1842 [20] F. Zhou, P. Charlat, B. Spivak, and B. Pannetier, J. of Low Temp. Phys. 110, 841 (1998). [21] C. Meyer, O. Sqalli, H. Lorenz, and K. Karrai, Rev. Sci. Instrum. 76, 063706 (2005).
1901.04679	1	1901	2019-01-15T07:03:23	Moir\'e quantum well states in tiny angle two dimensional semi-conductors	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	The valence band edge in tiny angle twist bilayers of MoS$_2$ and phosphorene is shown to consist of highly localized energy levels created by a `moir\'e quantum well', i.e. trapped by the interlayer moir\'e potential. These approximately uniformly spaced energy levels exhibit a richly modulated charge density, becoming ultra-localized at the valence band maximum. The number and spacing of such levels is controllable by the twist angle and interlayer interaction strength, suggesting the possibility of `moir\'e engineering' ordered arrays of quantum dots in 2d twist semi-conductors.	cond-mat.mes-hall	cond-mat	Moir´e quantum well states in tiny angle two dimensional semi-conductors M. Fleischmann1, R. Gupta1, S. Sharma2, and S. Shallcross1∗ 1 Lehrstuhl fur Theoretische Festkorperphysik, Staudstr. 7-B2, 91058 Erlangen, Germany, and 2 Max-Born-Institute for non-linear optics and Short Pulse Spectroscopy, Max-Born Strasse 2A, 12489 Berlin, Germany. (Dated: January 16, 2019) The valence band edge in tiny angle twist bilayers of MoS2 and phosphorene is shown to consist of highly localized energy levels created by a "moir´e quantum well", i.e. trapped by the interlayer moir´e potential. These approximately uniformly spaced energy levels exhibit a richly modulated charge density, becoming ultra-localized at the valence band maximum. The number and spacing of such levels is controllable by the twist angle and interlayer interaction strength, suggesting the possibility of "moir´e engineering" ordered arrays of quantum dots in 2d twist semi-conductors. Introduction: "Moir´e materials", created by the small angle twist of a bilayer material, are an emerging fam- ily of two dimensional systems with a remarkable physics of localization. The prototypical moir´e material, twist bilayer graphene1 -- 7, exhibits at the single-particle level strong localization on the moir´e and, at certain "magic angles", low energy bands of vanishing velocity4,7. This, in turn, leads to a rich physics of correlation includ- ing Mott states and signatures of high temperature su- perconductivity, recently observed in experiment8,9 and currently the subject of intense theoretical investigation. State of the art fabrication techniques now allow for the creation of angle controllable twist bilayers of any 2d material, either semi-metallic or semiconducting. As scattering potentials in semiconductors generally create states in or at the gap edge, moir´e materials generated from 2d semi-conductors -- in which single layer states scatter off a "moir´e potential" -- might be expected to display a rich physics of localized moir´e induced impu- rity states. Recent theoretical work at small twist angles 11 (θ > 2.5◦), that in phosphorene10 (θ > 1.8◦) and MoS2 find localization and flat bands, suggest that this is in- deed the case. A continuum approach allows access to the tiny angle limit of these two semi-conductors (0.3◦ < θ ≤ 1◦) and, just as in graphene, this regime is found to present an enrichment of the physics at larger angles. The nearly dispersionless bands found for θ > 1.8◦ and θ > 2.5◦, in phosphorene and MoS2 respectively, are seen to develop into an almost uniformly spaced series of energy levels, well described by the concept of an emergent "moir´e quantum well". This physics of moir´e confinement gener- ates intensely localized states, with a rich density modu- lation whose nodal structure is related the level index n. As the energy spacing and number of levels is sensitive to the twist angle and interlayer coupling strength, such systems offer the tantalizing possibility of engineering in- trinsic moir´e quantum wells in 2d semi-conductors. Model : The divergence of the moir´e length in the limit of θ → 0 renders atomistic methods computationally prohibitive for the tiny angle regime, and makes natu- ral the use of continuum methodologies. The continuum approach was thus adopted at an early stage in the in- vestigation of twist graphene4,5,7, and has been employed to study several van der Waals heterostructures12 as well as partial dislocation networks in bilayer graphene13 -- 15. A crucial criteria that any continuum method must sat- isfy is that it reproduces, as a controlled approximation, the results of the underlying tight-binding method it is based on. Here we adopt an approach based on an oper- ator equivalence between the tight binding method and a continuum Hamiltonian H(r, p): (cid:68) (cid:69) (cid:68) k1α HT B Ψ(m) Ψ(n) k2β = k1α H(r, p) φ(m) φ(n) k2β (cid:69) . (1) (cid:69) (cid:12)(cid:12)(cid:12)Ψ(n) k1α k1α (cid:69) (cid:12)(cid:12)(cid:12)φ(n) On the left hand side of this equation are Bloch states (with layer in- from the constituent single layers dex n, crystal momentum k, and a combined atomic in- dex α), and on the right hand side their natural contin- uum counterpart: generalized pseudospinor plane waves . These are given by a plane wave phase function eik.r augmented by a pseudospin vector, which in the case of graphene would be simply (1, 0) or (0, 1) but for the more general materials we consider here are unit vec- tors in a larger space comprising all the atomic degrees of freedom, i.e. spin, angular momentum, and sub-lattice indices. As shown in Ref. 16 a H(r, p) exists that satisfies Eq. (1) exactly. The interlayer coupling has the general form [S(r, p)]αβ = 1 VU C Mjαβe−ig(m) i .rtαβ(K0 + Gj + p) (2) where p is the momentum operator, and the sum is over all reciprocal lattice vectors Gj of the unrotated layer. The three factors in the sum encode the high symmetry bilayer, the twist momentum boosts, and the electronic coupling through, respectively, the "M matrices" Mjαβ = eiGj .(να−νβ ) (3) = Gi − RGi (R the rota- the moir´e momentum g(m) tion operator), and tαβ(q) the Fourier transform of the interlayer hopping amplitude between orbitals of atomic indices α and β. Here the atomic indices represent a com- pound index of angular momentum, sub-lattice, and spin i (cid:88) j 9 1 0 2 n a J 5 1 ] l l a h - s e m . t a m - d n o c [ 1 v 9 7 6 4 0 . 1 0 9 1 : v i X r a 2 FIG. 1: Band structure in extended zone scheme for MoS2 (top row) and phosphorene (bottom row). The weight function is given by the projection of the twist bilayer wavefunction onto the complete set of single layer states at k; see Eqs. (5) and (6). For the small twist angles shown here the interlayer interaction causes both band broadening, see panels 2,3,5,6, and, near the Γ point, the creation of a series of localized levels, see panels 1 and 4. ω(k, ) = ρkjδ( − Ekj) (cid:69) ki Ψkj ψ(n) (cid:88) (cid:68) j (cid:88) (cid:12)(cid:12)(cid:12)ψ(n) (cid:69) ni ki ρkj = (5) (6) labels. This approach represents a generalization of that pioneered for graphene4,7, and requires as input only the Slater-Koster tight-binding amplitude functions of a high symmetry bilayer, for which accurate parameterizations exist for both MoS2 17 and phosphorene18. Calculation method : The continuum Hamiltonian ex- where pressed in layer space has the form (cid:18) H (1) H = S(r, p) S(r, p)† H (2) (cid:19) is a weight function in which Ψkj(cid:105) is an eigenstate of a single layer eigenstate from the twist bilayer and (4) with H (n) the single layer Hamiltonians, treated exactly at the tight-binding level, and S(r, p) the interlayer cou- pling operator given by Eq. (2). To solve the electronic structure problem we use a basis set of combined single layer eigenstates obtained from H (n). In this basis the intra-layer blocks are diagonal, consisting just the single layer eigenvalues, with elements of the inter-layer blocks non-zero only if the difference of the single layer crystal momenta of the basis functions falls on the moir´e mo- mentum lattice {g(m) }. For the tiny angle calculations we undertake 0.3◦ < θ < 3◦ we find we require up to 5000 single layer states. i In the small twist angle regime it is useful to present the band structure in the extended zone scheme by pro- jection of the twist eigenfunction onto single layer eigen- states, i.e., to plot layer n with (single layer) band index i. With no inter- layer interaction ρkj = 1 and ω(k, ) simply represents the superposition of the two single layer band structures. In contrast, in the physical bilayer the interlayer interac- tion scatters single layer states into other momenta re- sulting in ρkj < 1 and an ω(k, ) describing the formation of mini-bands and mini-gaps in the extended zone. Extended zone band structures: In Fig. 1 we show band structures in the extended zone scheme for MoS2 and phosphorene for θ = 1◦. At these twist angles the moir´e momentum is orders of magnitude smaller than the single layer reciprocal lattice vectors, and the resultant multiple scattering of single layer eigenstates generates an effec- tive broadening of the band structure. In MoS2 this is substantial close to the Γ point, but negligible at the va- lence band K-point, where the twist wavefunctions con- sist of almost pure single layer states. This reflects both 3 FIG. 2: Moir´e quantum well states in a θ = 1◦ MoS2 twist bilayer. Panels 1 and 2 show the density of states in a wide energy range and close to the valence band edge respectively. The labels in the latter panel correspond to the integrated charge density shown in panels (3-11); panel 12 exhibits the density for a localized state close to the edge at ∼ 2 eV. FIG. 3: Moir´e quantum well states in a θ = 1◦ twist bilayer of phosphorene. In panel (1) is exhibited the density of states of the valence band, with the inset displaying the localized energy levels at the gap edge. Panels (2) and (3-6) show, respectively, the electron density corresponding the the energy window within the valence band, and the edge states indicated in the inset figure. the much weaker coupling at the K point as compared to the Γ point, since t(Γ) < t(K), but also the absence of states to scatter into near the K valence band edge due to the larger effective mass; note that broadening is seen at K for higher energies (where the effective mass is much reduced), although it is still weaker than that seen close to the Γ-point. Similarly, in phosphorene the effective mass anisotropy leads to stronger twist induced broadening in the Γ-X direction of the Brillouin zone as compared to the Γ-Y direction. Most interesting, however, is the low energy sector near the Γ-point, see panels 3 and 6 of Fig. 1. A series of uni- formly spaced and almost dispersionless gap edge states can be seen, which have non-zero band velocity only as their projection weight falls to zero. Moir´e quantum well states: A natural explanation for such dispersionless states is that they represent the effec- tively isolated energy levels of a quantum well created by the moir´e lattice. If this is so, one would expect (i) level wavefunctions to be isolated on a specific region of the moir´e and (ii) a structure to the wavefunctions connected to a quantum number n labelling the energy levels, just as occurs in any truly isolated quantum well. As we show in Figs. 2 and 3, that display the density of states and integrated electron densities of 1◦ twist bilayers for MoS2 and phosphorene respectively, this is indeed the case. In panels 6-11 of Fig. 2 the integrated density of the gap edge states in MoS2 is shown corresponding to the edge states labelled in panel 2, revealing both complete isolation from other regions of the moir´e, as well a level density that is circularly symmetric with n/2 periods in the radial density. (The highest energy level has index n = 1.) As can be seen, the level density become in- creasing localized towards the valence band maximum. Correspondingly, as the levels join the continuum scat- tering occurs between quantum wells resulting in a C6 modulation of the level density by the moir´e lattice and the vanishing of regions of zero charge, panel 5. Into the continuum the density becomes ever more homogeneous, see panels 3 and 4, although remaining weakly modulated by the moir´e. The quantum well states are localized on the open AA regions of the moir´e, with the angular mo- mentum character dominated by pz and d3z2−r2, just as at the Γ point in single layer MoS2. A similarly structure of the wavefunctions of the en- ergy levels is observed for phosphorene, with the number of maxima in the level density exactly equal to the level index n (with n = 1 again the highest energy state), see panels 3-6 of Fig. 3. The electron density is, however, no longer circularly symmetric as in MoS2, but exhibits a rectangular structure inherited from the rectangular moir´e of this material, and is localized on the AA' stacked region. We thus conclude that the picture of a gap edge governed by the physics of a moir´e quantum well provides a good description for both MoS2 and phosphorene. In- terestingly, in contrast to MoS2 the electron density in phosphorene is strongly modulation by the moir´e at all energies. This can be seen in panel 2 of Fig. 3 in which 4 the integrated density in the window indicated in the density of states plot is shown. It is worth contrasting the localized and highly struc- tured density of the gap edge states shown in Figs. 2 and 3, with the well known moir´e density modulation found in graphene. While semi-classical calculations do lend some credence to the idea of a moir´e quantum well in graphene19, the absence of a gap renders the concept problematic, and indeed the electron density, which is strongly localized on the AA stacked regions, displays neither a structural modulation connected to a level in- dex, or the profound isolation from other regions of the moir´e seen in Figs. 2 and 3. FIG. 4: Moir´e engineering quantum well states in MoS2. In the left hand panel is shown the level spacing of moir´e quan- tum well states as a function of twist angle. The right hand panel shows the density of states at the valence band edge for the θ = 0.5◦ twist bilayer at two interlayer interaction strengths. Moir´e quantum well engineering: The level spectrum of a moir´e quantum well must, evidently, be controllable through the twist angle. Most obviously, the length scale of the moir´e well wavefunctions will scale by the moir´e length. However, the level spacing ∆E can also be con- trolled through the twist angle, as shown in Fig. 4, with ∆E falling from ∼ 10 meV to ∼ 2 meV at 0.3◦ twist angle. Interestingly, the number of levels is quite sensi- tive to the strength of the interlayer interaction; shown in panel (2) of Fig. 4 are the level states for a 0.5◦ twist bilayer with the interlayer interaction scaled by a factor λ = 0.6 (21 energy levels) as compared to λ = 1.0 (10 energy levels). Discussion: We have shown that the Γ-point valence band edge in frozen moir´e lattices of tiny angle MoS2 and phosphorene twist bilayers is governed by the physics of a "moir´e quantum well", and features essentially iso- lated states with wavefunction structure determined by the level index n. The n = 1 state has a spatial ex- tension of the order of a nanometer, comparable to the excitonic radius in single layer MoS2 and thus ordered arrays of quantum wells, created by a tiny twist of a 2d semi-conducting bilayer, form a tantalizing platform for the "moir´e quantum well" engineering of localization and excitonic effects. Spin orbit interaction in MoS2 is likely generate interesting physics in the context of moir´s quantum well states. While the impact of lattice relax- ation in the tiny angle regime will change details of the physics we present here, experience from graphene (where the physics of localization survives relaxation) and first 11 and principles calculations at larger angles in MoS2 phosphorene10 (in which the creation of flat bands oc- curs in both ideal and relaxed structures) suggests that the fundamental physics of a moir´e quantum well we de- scribe will survive relaxation. 5 Acknowledgement This work was carried out in the framework of SFB 953 of the Deutsche Forschungsgemeinschaft (DFG). ∗ Electronic address: [email protected] 1 S. Shallcross, S. Sharma, and O. A. Pankratov, Phys. Rev. Lett. 101, 056803 (2008), URL http://link.aps.org/ doi/10.1103/PhysRevLett.101.056803. 2 S. Shallcross, S. Sharma, E. Kandelaki, and O. A. Pankra- tov, Phys. Rev. B 81, 165105 (2010), URL http://link. aps.org/doi/10.1103/PhysRevB.81.165105. 3 E. J. Mele, Phys. Rev. B 84, 235439 (2011), URL http: //link.aps.org/doi/10.1103/PhysRevB.84.235439. 4 R. Bistritzer and A. H. MacDonald, Proceedings of the National Academy of Sciences 108, 12233 (2011), URL http://www.pnas.org/content/108/30/12233.abstract. 5 J. M. B. Lopes dos Santos, N. M. R. Peres, and A. H. Castro Neto, Phys. Rev. B 86, 155449 (2012), URL http: //link.aps.org/doi/10.1103/PhysRevB.86.155449. 6 S. Shallcross, S. Sharma, and O. Pankratov, Phys. Rev. B 87, 245403 (2013), URL http://link.aps.org/doi/10. 1103/PhysRevB.87.245403. 7 D. Weckbecker, S. Shallcross, M. Fleischmann, N. Ray, S. Sharma, and O. Pankratov, Phys. Rev. B 93, 035452 (2016), URL http://link.aps.org/doi/10.1103/ PhysRevB.93.035452. 8 Y. Cao, V. Fatemi, A. Demir, S. Fang, S. L. Tomarken, J. Y. Luo, J. D. Sanchez-Yamagishi, K. Watanabe, T. Taniguchi, E. Kaxiras, et al., Nature 556, 80 EP (2018), URL https://doi.org/10.1038/nature26154. 9 Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi, E. Kaxiras, and P. Jarillo-Herrero, Nature 556, 43 EP (2018), article, URL https://doi.org/10.1038/ nature26160. 10 P. Kang, W.-T. Zhang, V. Michaud-Rioux, X.-H. Kong, C. Hu, G.-H. Yu, and H. Guo, Phys. Rev. B 96, 195406 (2017), URL https://link.aps.org/doi/10. 1103/PhysRevB.96.195406. 11 M. H. Naik and M. Jain, Phys. Rev. Lett. 121, 266401 (2018), URL https://link.aps.org/doi/10. 1103/PhysRevLett.121.266401. 12 A. M. Alsharari, M. M. Asmar, and S. E. Ulloa, Phys. Rev. B 98, 195129 (2018), URL https://link.aps.org/ doi/10.1103/PhysRevB.98.195129. 13 F. Kisslinger, C. Ott, C. Heide, E. Kampert, B. Butz, E. Spiecker, S. Shallcross, and H. B. Weber, Nat Phys 11, 650 (2015). 14 S. Shallcross, S. Sharma, and H. B. Weber, Nature Com- munications 8, 342 (2017), ISSN 2041-1723. 15 D. Weckbecker, R. Gupta, F. Rost, S. Sharma, and S. Shallcross, arXiv e-prints arXiv:1812.03343 (2018), 1812.03343. 16 N. Ray, F. Rost, D. Weckbecker, M. Vogl, S. Sharma, Shallcross, O. Pankratov, and S. R. Gupta, arXiv:1607.00920 (2016). 17 E. Cappelluti, R. Rold´an, J. A. Silva-Guill´en, P. Ordej´on, and F. Guinea, Phys. Rev. B 88, 075409 (2013), URL https://link.aps.org/doi/10.1103/PhysRevB.88. 075409. 18 A. N. Rudenko, S. Yuan, and M. I. Katsnelson, Phys. Rev. B 92, 085419 (2015), URL https://link.aps.org/doi/ 10.1103/PhysRevB.92.085419. 19 M. Vogl, O. Pankratov, and S. Shallcross, Phys. Rev. B 96, 035442 (2017), URL https://link.aps.org/doi/10. 1103/PhysRevB.96.035442.
1905.03977	1	1905	2019-05-10T07:26:55	Emergent elastic waves in skyrmion crystals with finite frequencies at long wavelength limit	[ "cond-mat.mes-hall" ]	A fundamental fact in solids is that the frequencies of elastic waves vanish as the wave number approaches zero\cite{6}. Here we theoretically show that this fact is overturned when studying the lattice vibration of skyrmion crystals (SkX), i.e., periodic alignment of topologically nontrivial spin solitons called magnetic skyrmions. As emergent crystals, SkX possess collective excitations called "emergent phonons", which describe dynamics of SkX caused by lattice vibration (resembling acoustical branches of ordinary phonons) and in-lattice vibration (resembling optical branches of ordinary phonons). We find that lattice vibration and in-lattice vibration of the emergent phonons in SkX are coupled even at long wavelength limit, such that multiple types of "emergent elastic waves" (modes causing lattice vibration of SkX) with finite frequencies exist. This phenomenon, which originates from the Berry phase form of kinetic energy, is generally true for emergent crystalline states of spins. Our results show that the dynamics of magnetic emergent crystals are intrinsically different from that of ordinary crystals.	cond-mat.mes-hall	cond-mat	a Emergent elastic waves in skyrmion crystals with finite frequencies at long wavelength limit Yangfan Hu∗ Sino-French Institute of Nuclear Engineering and Technology, Sun Yat-sen University, 519082, Zhuhai, China A fundamental fact in solids is that the frequencies of elastic waves vanish as the wave number approaches zero[1]. Here we theoretically show that this fact is overturned when studying the lattice vibration of skyrmion crystals[2 -- 4] (SkX), i.e., periodic alignment of topologically nontrivial spin solitons called magnetic skyrmions[5, 6]. As emergent crystals, SkX possess collective excitations called "emergent phonons", which describe dynamics of SkX caused by lattice vibration (resembling acoustical branches of ordinary phonons) and in-lattice vibration (resembling optical branches of ordinary phonons). We find that lattice vibration and in-lattice vibration of the emergent phonons in SkX are coupled even at long wavelength limit, such that multiple types of "emergent elastic waves" (modes causing lattice vibration of SkX) with finite frequencies exist. This phenomenon, which originates from the Berry phase form of kinetic energy, is generally true for emergent crystalline states of spins. Our results show that the dynamics of magnetic emergent crystals are intrinsically different from that of ordinary crystals. Magnetic skyrmions[2, 5, 6] are particle-like spin textures stabilized by the noncollinear Dzyaloshinsky- Moriya interaction (DMI) in chiral magnets. Skyrmions appear as basic units in an emergent world such that they usually condensate into a crystalline state[2, 3, 7] called SkX. Similar to ordinary crystals, the presence of SkX induce a novel type of collective excitations accord- ing to previous studies[8 -- 13], which we call it "emer- gent phonons". The significance of this novel type of collective excitations is twofold: on one hand, similar to phonons in ordinary crystals[14], they determine the dynamical behaviors of the system, as well as some of the most fundamental macroscopic properties such as the elasticity[15] and heat capacity. On the other hand, they are intrinsically collective behaviors of spins. This elec- tromagnetic nature permits direct interaction with static or dynamic electromagnetic fields, inducing exotic prop- erties such as high electric current mobility[8, 16], elastic- ity under electric[17] and magnetic fields[18], which are promising to possible microwave- and magnonics-related application[19]. Phonons in ordinary crystals are divided into two cat- egories known as the acoustical branches and the opti- cal branches[1]. At long wavelength limit, the acoustical branches, corresponding to elastic waves, describe vibra- tions of lattice as a whole and have vanishing frequen- cies, while the optical branches describe relative vibra- tions of atoms inside a lattice and have finite frequencies. Such a categorization is subtly used in previous analyt- ical studies of emergent phonons in SkX. Zang et. al[8] consider the acoustical branches of emergent phonons in SkX at long wavelength limit, and obtained a ω ∝ k2 form of dispersion relation. In their study, SkX are re- garded as monatomic crystals whose deformation at long wavelength is described merely by a continuous emergent displacement field, representing vibration of the lattices ∗ Corresponding [email protected] as a whole. On the other hand, SkX can be described by a Fourier series[2, 21], and the vibration of the Fourier magnitudes, seemingly not inducing any lattice deforma- tion, can also cause a change of the field patterns of SkX. Schwarze[9] et. al study the vibration of these Fourier magnitudes, recovering the three distinct modes termed clockwise (CW), counterclockwise (CCW) and breathing in the few GHz range observed in Cu2OSeO3[10], M nSi and F e0.8Co0.2Si[9]. Nevertheless, the displacement of the lattice as a whole is not considered in their formu- lation, for which their analysis seems to be focusing on the optical branches of emergent phonons in SkX. A key issue unsettled in these studies is that at long wavelength limit, do the acoustical branches and the optical branches of the emergent phonons in SkX couple, or do they be- have like independent vibrational modes such as those of ordinary crystals. By systematically studying the two types of vibrations together in SkX, we find that they are strongly coupled even at long wavelength limit, such that multiple modes exist which cause vibration of the lattice as a whole, while the so-called "Goldstone mode" cause simultaneously a rigid translation of the lattice and a dis- tortion of the field pattern of skyrmion inside the lattice. We study Bloch-type SkX in B20 chiral magnets, where the following Landau-Ginzburg functional describes the rescaled free energy density of the system[2] φ (m) = + 2m · (∇ × m) − 2b · m (1) 3 ∂ri(cid:19)2 Xi=1(cid:18) ∂m + T m2 + m4, where m is the rescaled magnetization, b is the rescaled magnetic field and T is the rescaled temperature. The rescaling process is described in the Methods section. In , so that eq. (1) is isotropic in the x − y plane. The effect of anisotropy is discussed in a subsequent work of ours[20]. Deformable SkX with long range order permits the following Fourier this work we assume that b =(cid:2)0 0 b(cid:3)T expansion of m[15, 21] mqe l eiql(εe ij , ωe)·r, (2) m =Xl 1 ue by εe ij = 1 2(cid:3)T 2 (ue i,j + ue 2 (ue 1,2 − ue 1 and qe l = l1qe l denotes the Fourier magnitudes, qe 2 where l1 and l2 are integers and qe displacement field ue =(cid:2)ue 1 + where mqe l2qe 2 are the ba- sic reciprocal vectors of SkX, which are deformable under external disturbance. The deformation of qe l is described by the emergent elastic strains εe ij and the emergent rota- tional angle ωe[15], which are defined from the emergent j,i) and ωe = 1 2,1). Similar to atomic lattice, rigid translation of SkX does not induce a change of free en- ergy, thus it is εe ij and ωe instead of ue which matter in the expression of m. The expression of ql in terms of εe ij and ωe depends on the crystalline structure of SkX, and is introduced in the Methods section for hexagonal SkX. At appropriate condition of b and T , minimization of the free energy based on eq. (1) yields a metastable SkX[6, 21], while the thermodynamically stable state cor- responds to the conical phase. Nevertheless, concerning the exotic robustness of metastable SkX observed in a wide area of the temperature-magnetic field phase dia- gram of helimagnets[4, 22], the analysis of this metastable SkX phase is not only of theoretical interest but also with practical significance. zation of a metastable SkX by (mql)st,(cid:0)εe ij(cid:1)st From eq. (2), we can describe the equilibrium magneti- and (ωe)st. Consider a small vibration around this metastable state, which induces simultaneously a vibration of ue de- noted by (ue)v(r, t) and a vibration of mql denoted by (mql)v(r, t). In this case, eq. (2) becomes (3) l )st = qe , (ωe)sti. It is convenient to m =Xl h(cid:0)mqe + (mql)vi ei(ql)st·[r−(ue)v], l(cid:1)st l h(cid:0)εe ij(cid:1)st where (qe write components of all Fourier magnitudes mql in a sin- gle vector mq, for which the two vectors (ue)v and (mq)v include all the variables to be solved. In previous theo- retical studies of the emergent elastic waves[8, 11], (ue)v is considered solely, while in previous theoretical stud- ies of the internal modes of SkX[12], (mq)v is considered solely. To establish an analytical framework which de- termines the collective spin excitations in SkX, we have to derive the coupled Euler-Lagrange equations for (ue)v and (mq)v, which can be obtained from the least ac- tion principle by using eq. (3) (See the Methods section for details). Consider the plane-wave form of solution k·r), a general- (ue)v = ue0ei(ωt− ized eigenvalue problem of the frequency ω can be ob- tained at long wavelength limit k·r), (mq)v = mq0ei(ωt− (Rω − K)(cid:20) ue0 mq0(cid:21) = 0, (4) where the expression of R and K are derived in the Methods section. Solution of eq. (4) determines the dis- persion relation of the coupled vibration of (ue)v and 2 (mq)v, as well as the eigenvectors for different modes. We denote the frequencies of different modes by ωi(k) (ωi(k) = ωi(k)/η, where η is a material dependent factor such that ωi(k) is material independent), ordered in such a way that ω1(0) < ω2(0) < ω3(0) < · · · . We study the coupled vibration of (ue)v and (mq)v at the long wavelength limit. The real-space magneti- zation distribution undergoing the first 6 modes calcu- lated at T = 0.5, b = 0.3, k1 = 10−5, k2 = 0 is shown in FIG. 1, resembling those of a previous study[23], ex- cept that (ue)v is neglected in their work and consid- ered here. The experimentally confirmed[10, 12] counter clockwise (CCW) mode (ω3), clockwise (CW) mode (ω6) and breathing mode (ω5) all shows up, and the varia- tion of their frequencies with the applied magnetic field plotted in FIG. 2 agree with corresponding experimental results[23] (see Supplementary Videos 1-6 for the motion of these modes). Unexpectedly, we find that even at the Γ point, there is still strong coupling between (ue)v and (mq)v, such that 3 out of the first 6 modes cause vibration of the lattices as a whole, while 2 of them possess nonzero frequency. When we consider more modes by incorporating a higher order Fourier expansion of m in eq. (3), we find that 7 out of the first 20 modes have similar properties, which implies that a finite proportion of all modes present cou- pled vibration of (ue)v and (mq)v at the Γ point. Mean- while, we find that the "Goldstone mode" (ω1), i.e., the emergent elastic wave with vanishing frequency at the Γ point, has nonzero components of mq0 in its eigenvector of vibration. It means that the Goldstone mode (ω1) no- longer describes pure rigid translational vibration of SkX, but is accompanied by deformation of the field pattern of skyrmion inside the lattice (as illustrated by Fig. 1(a1- a4)). This Goldstone mode solved by considering cou- pling between (ue)v and (mq)v still possesses a ω ∝ k2 form of dispersion relation at long wavelength limit, yet with a smaller coefficient than the one obtained by solv- ing the vibration of (ue)v alone (See the phonon spec- trum plotted in FIG. 3 for comparison). Actually, this accompanied particle shape deformation of the Goldstone mode has long been confirmed when studying the dynam- ics of isolated skyrmions[24 -- 26] and SkX[27]. In isolated skyrmions the particle shape deformation of skyrmion in the Goldstone mode is found to induce a large inertia of the skymion and significantly affects its dynamics. Our results also explain why the experimentally obtained in- ertia of isolated skyrmion[25] significantly exceeds that of the related theoretical prediction[24]. Actually the vibrational modes studied in the two works are differ- ent: the experimentally excited modes[25] in the GHz range corresponds to the CCW and CW modes of iso- lated skyrmion[28], while the theoretically studied one is the Goldstone mode[24]. The magnitude of internal de- formation in the Goldstone mode is much smaller than that of the CCW and CW modes, which dominates the effective mass of the skyrmion particle during motion and thus explains the discrepancy. Moreover, by analyzing its eigenvector of vibration, we find that this Goldstone mode no longer possesses two vibration direction of the lattices (corresponding to the longitudinal and transverse elastic waves of ordinary crystals), but has a unique mode of vibration where the longitudinal motion and trans- verse motion of lattices are coupled. In this case, the lat- tices of SkX undergo clockwise elliptical rotation around their equilibrium position (Figure 1(a1-a4) and Supple- mentary Video 1). The particle shape deformation and rotational motion of the Goldstone mode presented in Supplementary Video 1 are observed in simulation of the current-driven motion of SkX[27]. The physical origin of this exotic finite frequency emergent elastic waves in SkX lies in the berry phase form of the kinetic energy. To be more specific, it is caused by coupling terms such as d i ) , (i = 1, 2) appear in the kinetic energy density, where mc i , (i = 1, 2) denote components of the constant vector in the Fourier representation of rescaled magnetization. As a result, the motion of ue i , (i = 1, 2) and mc i , (i = 1, 2) are coupled in the mass matrix R in eq. (4). Although in the stiffness matrix K, the coupling i , (i = 1, 2) and (mq)v vanishes at the Γ point, between ue mc i , (i = 1, 2) is coupled with half of the components in (mq)v, which together render a strong coupling between (ue)v and (mq)v. The above analysis shows that this feature should be common all emergent crystalline states composed of periodic spin textures. dt (ue i mc Nonzero ue0 i , (i = 1, 2) components in the eigenvec- tor of a vibrational mode means that when the mode is excited, global displacement of the lattice as a whole can be achieved. An immediate consequence of this exotic fi- nite frequency emergent elastic waves is that AC external fields with different frequencies may be used to transport SkX by exciting different emergent elastic waves. Since these modes corresponds to different internal deforma- tion pattern of SkX, we should expect different dynamic behaviors when the SkX is moved by different modes. Consider AC magnetic field as an example, an important question is which of all the calculated modes can be ex- cited. This depends on whether the considered mode has a nonzero component of mc i , (i = 1, 2, 3). E.g., CCW and CW modes are excited by an in-plane ac-magnetic field[10, 12] since they have nontrivial mc0 in their eigenvector of vibration. The breathing mode is excited by an out-of-plane ac-magnetic field since they have nontrivial mc0 In 3 Table 1, we list the value of components of ue0 and mc0 in the unit eigenvector for the first 20 modes calculated at T = 0.5, b = 0.3 near the Γ point. The variation of these components for the first 6 modes with the applied magnetic field is plotted in FIG. 2(b, c), which shows that the eigenvector of modes can be tuned by changing the bias magnetic field. in its eigenvector of vibration. 1 and mc0 2 Our study shows that the emergent elastic wave exci- tations of emergent crystals in chiral magnets are fun- damentally different from those of ordinary crystals, for which the incorporation of emergent lattice deformation and displacements when studying their collective spin ex- 3 citations is generally indispensable. Further research in this direction not only helps us in understanding the dis- crepancies between emergent crystalline states and or- dinary crystals in terms of their dynamics, but may also stimulate novel applications. We construct a reliable and analytical framework to study the collective spin excita- tions in magnetic emergent crystals, which can easily be extended to cases where they are deformed[20]. I. METHODS F ree energy density f unctional of B20 chiral magnets and its rescaling We use the following free energy density functional to study magnetic skyrmions in cubic helimagnets φ (M) = 3 Xi=1 ∂xi(cid:19)2 A(cid:18) ∂M + DM · (∇ × M) (5) − B · M + α(T − T0)M2 + βM4, where M denotes the magnetization, B denotes the magnetic field, and T denotes the temperature. The (5) denote respectively the terms on the rhs. of eq. exchange energy density with a coefficient A, the Dzyaloshinskii-Moriya interaction (DMI) with a coeffi- cient D, the Zeeman energy density, and the second and fourth order Landau expansion terms. Eq. (5) can be simplified by rescaling the spatial variables as r = x , LD D , K = D2 m = M 4A , M0 φ (m), , b = B B0 β , B0 = 2KM0, which yield φ (M) = K 2 , and LD = 2A , T = α(T −T0) K β while φ (m) is given in eq. (1). The benefit of eq. (1) compared with eq. (5) is that it provides a free energy density functional that is independent of material parameters, so that results obtained from analyzing eq. (1) has general significance to Bloch-type SkX in any B20 helimagnets. M0 =q K F ourier representation of SkX and f ree energy minimization To determine a metastable SkX state at given T and b, one has to substitute the analytical expression of m for the SkX phase into eq. (1), and minimize the free energy of the system with respect to all independent variables. In practice, we use the following Fourier representation of SkX instead of eq. (2)[15] m =mc + n ni Xi=1 Xj=1 mqij ei[I−Fe(r)]T qij·r, (6) where mc denotes a constant vector, F e ij = ue i,j, and qij denote the undeformed wave vectors orga- nized according to the following rules: q1j < q2j < q3j < ......, and qi1 = qi2 = ...... = qini . When truncated at a specific value of n, the nth order Fourier ij (r) = εe ij + ωe εεεea = [εe 11, εe 22, εe 12, ωe]T , (7) − representation given in eq. (6) saves all the significant Fourier terms up to the nth order, which is hard to achieve if one uses eq. (2). ij1 + icim ij3 + icim ij2, cij3 = cre It is convenient to expand mqij as mqij = cij1Pij1 + cij2Pij2 + cij3Pij3, where cij1 = cre ij1, cij2 = ij2 + icim cre ij3 are complex variables to be determined, and Pij1 = 1 [−iqijy, iqijx, siq]T , √2siq Pij2 = 1 siq [qijx, qijy, 0]T , Pij3 = 1 [iqijy , −iqijx, siq]T √2siq with qij = [qijx, qijy]T , qij = siq. For 2-D hexago- nal SkX, we can assume without loss of generality that q11 = [0, 1]T , q12 = [− 2 ]T , and q = 1. Com- ij , ωe) = paring eq. (6) and eq. [I−(Fe(r))T ]qij , which gives for the basic reciprocal vec- tors qe 11) + 1 2 (εe 12 − ωe), In this case, all the independent variables describing the rescaled magnetization of the SkX phase can be gath- ered in two vectors, which are given by 11 = [−εe 12 + ωe, 1 − εe √3 12 + ωe) − 1 2 (εe √3 2 , − 1 (2), we have qe 22]T , qe 2 (1 − εe 12 = [− 22)]T . √3 2 (1 − εe ij (εe and mq = [mc 1, mc cre 123, cre 2, mc 131, cre 3, cre 132, cre 111, cre 133, cim 112, cre 111, cim 113, cre 112, cim 121, cre 113 · · · ]T , 122, (8) where the length of mq depends on the order of Fourier representation used. For 1st, 2nd, and 3rd order Fourier representation, the length of mq is 21, 39 and 57, respectively. At given T and b, minimization of the free energy based on eq. (6) determines the equilibrium value of the two vectors εεεea and mq for a metastable SkX phase, which are denoted by (εεεea)st and (mq)st. the emergent of equation f or the F ourier magnitudes Euler − Lagrange displacements and SkX Here we briefly introduce the procedure to deduce the basic equations describing the coupled wave motion of (ue)v(r, t) and (mql)v(r, t)[29]. The action of any cubic helimagnet undergoing small vibration of magnetization reads S = SK −Z Φdt (9) where Φ =R φdV denotes the free energy of the system, and SK = R EKdt derives from the Berry phase action of a spin[30], whose variational form reads δSK = M γ Z Z (n × n)δndV dt. (10) Here n denotes the unit vector of magnetization, M the averaged modulus of magnetization and γ the gyromag- netic ratio. It is convenient to replace n in eq. (10) by the rescaled magnetization m, which gives δSK = M γm3 Z Z (m × m)δmdV dt. (11) 4 where the value of m = m depends on the rescaled magnetic field b and the rescaled temperature T . Φ = RV φdV denotes the free energy of the system, where φ (M) = K 2 φ (m) is deduced above and φ (m) is given in β eq. (1). The rescaled magnetization for deformable SkX can be expanded as follow instead of eq. (3) ∞ Xi=1 ni Xj=1 m =(mc)st + (mc)v + [(mqij )st + (mqij )v] × ei[I−(Fe(r))st]T qij·[r−(ue)v ], where (F e ij (r))st = (εe ij )st + (ωe ij )st. The Euler-Lagrange equations of (ue)v and (mq)v read (12) − d dt(cid:20) ∂L ∂( ue)v(cid:21) + ∂L ∂(ue)v −Xi d dri " ∂L ,i(cid:1)v# = 0, ∂(cid:0)ue ,i(cid:1)v# = 0, dri " ∂L ∂(cid:0)mq d (13) d dt(cid:20) ∂L ∂( mq)v(cid:21) + ∂L ∂(mq)v −Xi ∂ri (14) where L = EK − Φ is the Lagrangian of the system, and ,i = ∂ue ue . To actually use eqs. (13, 14) for small vi- bration of (ue)v and (mq)v, we first expand the averaged rescaled free energy density ¯φ = 1 of (ue)v and (mq)v and their derivatives and retain the lowest order terms. Substituting (ue)v = ue0ei(k·r−ωt), (mq)v = mq0ei(k·r−ωt) into eqs. (13, 14), eq. (4) can be derived at long wavelength limit, with V R φ (m)dV in terms Req R =(cid:20) Re (Req∗)T Rq(cid:21) , K =(cid:20) Ke Keq (Keq∗)T Kq(cid:21) , (15) (16) where Req∗ and Keq∗ denote complex conjugate of Req and Keq. , In eqs. (15, 16), Re , Req δmq ∂ mq Rq ∂ ∂ue K e ij ∂ ∂mq kpks ij = −i 1 Vh ∂ = Pp,s i (cid:17)ist j,ps (cid:16) d i,p(cid:17) −Pp i(cid:17) +Pp j (cid:16) δEK kpksh j,ps (cid:16) d j (cid:16) ∂ ¯φ i,p(cid:17)(cid:17) +Pp,s "Pp,s ij = " ∂ (cid:16) d drp (cid:16) ∂φ kpks ∂ ∂mq ∂φ ∂ue K q ikp ∂mq ∂mq ∂mq drp , = drp (cid:16) ∂ ¯φ ∂ue ikp ij = −i 1 ij = −i 1 ∂ ∂mq δue ∂ ue ∂ mq δue , K eq ij i (cid:17)ist j (cid:16) δEK i (cid:17)ist j (cid:16) δEK i,p(cid:17)#st Vh ∂ Vh ∂ i,p(cid:17)(cid:17)ist j,p (cid:16) d i(cid:17) −Pp i,p(cid:17)(cid:17)#st ∂ ∂mq j,p ∂φ ∂ue ikp ∂mq drp ∂mq j,p (cid:16) ∂φ drp (cid:16) ∂φ , j,ps (cid:16) d ∂ ∂mq . Here a subscript "st" means that the term is calculated at the equilibrium state ue = (ue)st and mq = (mq)st. One should notice that the stiffness matrix K is completely determined by the emergent elasticity of the SkX under magnetic field[15, 29]. The dispersion relation for different modes ωi = ωi(k) can be obtained by solving eq. (4). By defining the rescaled frequency ωi = 1 η ωi, , one obtains a rescaled dispersion relation ωi = ωi(k) that is material-independent. where η = (cid:12)(cid:12)(cid:12) γm3D4 16M A2 β(cid:12)(cid:12)(cid:12) Relation between the emergent displacement and the F ourier magnitudes of SkX at long wavelength limit At long wavelength limit, eq. (3) can be expanded as m =Xl (cid:2)(mql)st + (mql)v − i(mql)st (ql)st · (ue)v(cid:3) × ei(ql)st·r, which shows that by setting (mql)v = i(mql)st (ql)st · (ue)v (17) (18) for all ql considered, the vibration of the whole system vanished. This means that at long wavelength limit the vibration of ue is equivalent to a specific type of vibra- tion of the Fourier magnitudes identified by eq. (18). In fact, when solving the eigenvalue problem of eq. (4), two orthogonal modes can be specified from eq. (18), which 5 correspond to two vanishing eigenvalues of the two ma- trix R and K. As a result, the phonon spectrum obtained by solving eq. (4) at long wavelength limit is equivalent to that obtained by solving the following equation (Rqω − Kq) mq0 = 0. (19) The difference is that by solving eq. (19), it is hard to distinguish between deformation of the lattice as a whole and deformation of the field pattern inside the matrix. To do this at long wavelength limit, one has to specify the two orthogonal modes from eq. (18), which are the modes that does not cause any vibration of the system (referred to as the V-modes). The effective vibrational modes solved from eq. (4) satisfy an orthogonal relation with the V-modes, which can be used to calculate the value of ue0 in the eigenvector of a specific mode if one solves eq. (19) instead of eq. (4). ACKNOWLEDGMENTS The work was supported by the NSFC (National Nat- ural Science Foundation of China) through the funds 11772360, 11472313, 11572355 and Pearl River Nova Pro- gram of Guangzhou (Grant No. 201806010134). [1] Born, M. & Huang, K. Dynamical theory of crystal lat- tices. (Oxford university press, 1998). [2] Muhlbauer, S. et al. Skyrmion Lattice in a Chiral Mag- net. Science 323 915919 (2009). [3] Yu, X. Z. et al. Real-space observation of a two- dimensional skyrmion crystal. Nature 465 901904 (2010). [4] Karube, K. et al. Robust metastable skyrmions and their triangular-square lattice structural transition in a high-temperature chiral magnet. Nature Materials 15 12371242 (2016). [5] Bogdanov, A. & Hubert, A. Thermodynamically stable magnetic vortex states in magnetic crystals. Journal of Magnetism and Magnetic Materials 138 255269 (1994). [6] Bogdanov, A. N., Pfleiderer, C. & Roler, U. K. Sponta- neous skyrmion ground states in magnetic metals. Nature 442 797 (2006). [7] Seki, S., Yu, X. Z., Ishiwata, S. & Tokura, Y. Observa- tion of Skyrmions in a Multiferroic Material. Science 336 198201 (2012). [8] Zang, J., Mostovoy, M., Han, J. H. & Nagaosa, N. Dy- namics of Skyrmion Crystals in Metallic Thin Films. Physical Review Letters 107 136804 (2011). [9] Schwarze, T. et al. Universal helimagnon and skyrmion excitations in metallic, semiconducting and insulating chiral magnets. Nature Materials 14 478483 (2015). [10] Onose, Y., Okamura, Y., Seki, S., Ishiwata, S. & Tokura, Y. Observation of Magnetic Excitations of Skyrmion Crystal in a Helimagnetic Insulator Cu2OSeO3. Physi- cal Review Letters 109 037603 (2012). [11] Petrova, O. & Tchernyshyov, O. Spin waves in a skyrmion crystal. Physical Review B 84 214433 (2011). [12] Schwarze, T. et al. Universal helimagnon and skyrmion excitations in metallic, semiconducting and insulating chiral magnets. Nature Materials 14 478483 (2015). [13] Mochizuki, M. Spin-Wave Modes and Their Intense Ex- citation Effects in Skyrmion Crystals. Physical Review Letters 108 017601 (2012). [14] Anderson, P. W. Basic notions of condensed matter physics; 2nd ed. (Addison-Wesley, 1984). [15] Hu, Y. & Wan, X. Thermodynamics and elasticity of emergent crystals. arXiv: 1905.02165 (2019). [16] Jonietz, F. et al. Spin Transfer Torques in MnSi at Ul- tralow Current Densities. Science 330 16481651 (2010). [17] White, J. S. et al. Electric-Field-Induced Skyrmion Dis- tortion and Giant Lattice Rotation in the Magnetoelec- tric Insulator Cu2OSeO3. Physical Review Letters 113 107203 (2014). [18] Wang, C. et al. Enhanced Stability of the Magnetic Skyrmion Lattice Phase under a Tilted Magnetic Field in a Two-Dimensional Chiral Magnet. Nano Letters 17 29212927 (2017). [19] Garst, M., Waizner, J. & Grundler, D. Collective spin ex- citations of helices and magnetic skyrmions: review and perspectives of magnonics in non-centrosymmetric mag- nets. Journal of Physics D-Applied Physics 50 293002 (2017). [20] Hu, Y. Long-wavelength emergent phonons in dis- torted skyrmion crystals induced by intrinsic exchange anisotropy and a tilted magnetic field. ArXiv (2019). [21] Hu, Y. Wave nature and metastability of emergent crys- tals in chiral magnets. Communications Physics 1 82 (2018). [22] Oike, H. et al. Interplay between topological and ther- modynamic stability in a metastable magnetic skyrmion lattice. Nature Physics 12 6266 (2016). [23] Waizner, J. Spin wave excitations in magnetic helices and skyrmion lattices. (Thesis Universitat zu Koln, 2017). [24] Makhfudz, I., Krueger, B. & Tchernyshyov, O. Inertia and Chiral Edge Modes of a Skyrmion Magnetic Bubble. Physical Review Letters 109 217201 (2012). [25] Buettner, F. et al. Dynamics and inertia of skyrmionic spin structures. Nature Physics 11 225228 (2015). [26] Litzius, K. et al. Skyrmion Hall effect revealed by di- rect time-resolved X-ray microscopy. Nature Physics 13 170175 (2017). [27] Iwasaki, J., Mochizuki, M. & Nagaosa, N. Universal current-velocity relation of skyrmion motion in chiral magnets. Nature Communications 4 ncomms2442 (2013). [28] Lin, S.-Z., Batista, C. D. & Saxena, A. Internal modes of a skyrmion in the ferromagnetic state of chiral magnets. Physical Review B 89 024415 (2014). [29] Hu, Y. Lagrangian formulation for emergent elastic waves in magnetic emergent crystals. ArXiv (2019). [30] Altland, A. & Simons, B. D. Condensed Matter Field Theory. (Cambridge University Press, 2008). TABLE I. Specific component of the eigenvectors of the first six modes calculated at T = 0.5, b = 0.3, k1 = 10−5, k2 = 0. 6 ue0 1 2 mc0 1 (mc0 3 2 ) mc0 0 0 0 0 0.558 0 0 0 0 0 0 0.138 0 0 0.281 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.088 0.088 0.011 0.011 ue0 Modes ω1(Goldstone) 0.105 0.178 ω2 ω3(CCW) ω4 ω5(Breathing) ω6(CW) ω7 ω8 ω9 ω10 ω11 ω12 ω13 ω14 ω15 ω16 ω17 ω18 ω19 ω20 0.018 0.018 0.001 0.001 0.014 0.014 0.008 0.008 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.532 0.167 0 0 0 0.158 0 0 0.214 0.063 0 0 0.132 0 0 0 t =0 (a1) t =T/4 (a2) (a2) t =T/2 (a3) (a3) t =3T/4 (a4) (a4) 7 (b1) (b2) (b3) (b4) (c1) (c2) (c3) (c4) (d1) (d2) (d3) (d4) (e1) (e2) (e3) (e4) (f1) (f2) (f3) (f4) 1 Goldstone 2 3 CCW 4 5 Breathing 6 CW FIG. 1. Field configurations of SkX undergoing the first 6 modes of emergent phonons calculated at T = 0.5, b = 0.3 near the Γ point (k1 = 10−5, k2 = 0), different plots at four time points t during a period T are shown. In all figures, the vectors illustrate the distribution of the in-plane magnetization components with length proportional to their magnitude, while the colored density plot illustrates the distribution of the out-of-plane magnetization component. The black solid line plots the displaced Wigner-Seitz cell of SkX due to wave motion, while the white solid line plots the static position of the Wigner-Seitz cell. The first mode ω1 corresponds to the Goldstone mode; the third mode ω3 corresponds to the counter clockwise (CCW) mode; the fifth mode ω5 corresponds to the breathing mode; and the sixth mode ω6 corresponds to the clockwise (CW) mode. 6 5 4 3 2 1 (a) CW Breathing ci m CCW 0.7 0.6 0.5 0.4 0.3 0.2 0.1 (b) mc 3( , Breathing) mc 1=mc 2 ( , CCW) ei u mc 1=mc 2 ( , CW) 0.25 0.20 0.15 0.10 0.05 8 (c) ue 2 ( ,Goldstone) ue 1 1=ue ue 2 ( , CCW) 0 0.0 0.1 0.2 b Goldstone 0.3 0.4 0.0 0.0 0.1 0.3 0.4 0.2 b 0.00 0.0 1=ue ue 2 ( 0.1 , CW) 0.2 b 0.3 0.4 FIG. 2. Variation of (a) the frequency, (b) the value of mc i , (i = 1, 2) in the unit eigenvector of the first 8 modes of emergent phonons with the applied magnetic field calculated at T = 0.5, k1 = 10−5, k2 = 0. i , (i = 1, 2, 3) in the unit eigenvector, and (c) the value of ue 9 6 3 0 K M FIG. 3. Spectrum for the first 10 modes of emergent phonons in the SkX phase of bulk helimagnets within the first Bril- louin zone. The red curve plots the dispersion relation of the Goldstone mode solved by considering the vibration of ue alone.
1201.1903	2	1201	2012-09-14T04:05:44	Spinless basis for spin-singlet FQH states	[ "cond-mat.mes-hall", "hep-th" ]	We investigate an alternative description of the SU(M)-singlet FQH state by using the spinless basis. The SU(M)-singlet Halperin state is obtained via the q-deformation of the Laughlin state and its root of unity limit, by applying the Yangian Gelfand-Zetlin basis for the spin Calogero-Sutherland model. The squeezing rule for the SU(M) state is also investigated in terms of the spinless basis.	cond-mat.mes-hall	cond-mat	RIKEN-MP-39 Spinless basis for spin-singlet FQH states Taro Kimura∗ Department of Basic Science, University of Tokyo, Tokyo 153-8902, Japan and Mathematical Physics Laboratory, RIKEN Nishina Center, Saitama 351-0198, Japan Abstract We investigate an alternative description of the SU(M )-singlet FQH state by using the spinless basis. The SU(M )-singlet Halperin state is obtained via the q-deformation of the Laughlin state and its root of unity limit, by applying the Yangian Gelfand-Zetlin basis for the spin Calogero-Sutherland model. The squeezing rule for the SU(M ) state is also investigated in terms of the spinless basis. ∗E-mail address: [email protected] 2 1 0 2 p e S 4 1 ] l l a h - s e m . t a m - d n o c [ 2 v 3 0 9 1 . 1 0 2 1 : v i X r a Contents 1 Introduction 2 Spin-singlet FQH states 2.1 SU(2) theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 SU(M ) theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 From Laughlin to Halperin 3.1 The q-bosonic field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 The root of unity limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Spinful FQH states with admissible condition 4.1 Admissible condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 2 3 4 5 5 7 8 4.2 Squeezing rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 5 Relation to conformal field theory 6 Conclusion 1 Introduction 11 14 Recently it has been shown the Jack polynomial provides a universal description of the fractional quantum Hall states (FQH states) [1, 2, 3], satisfying the admissible condition [4] which characterizes the generalized statistics of the FQH states. This description is understood from the viewpont of the conformal field theory (CFT): the Jack polynomial for the generic (k, r)-admissible condition is regarded as the conformal block of the CFT with the extended chiral algebra WAk−1(k + 1, k + r) [5, 6, 7]. It is also applied to the FQH states in the presence of the internal (spin) degrees of freedom [8, 9, 10, 11] by considering the non-symmetric Jack polynomial [12, 13, 14, 15]. The Jack polynomial is the eigenfunction of the Laplace-Belrtami operator, which is regarded as the Hamiltonian of the Calogero-Sutherland model, up to the gauge transforam- tion. If we consider the spin degrees of freedom, we have to deal with the spin Laplace- Beltrami operator. In this case there are degeneracies in its spectrum due to the spin degeneracy. Thus it is difficult to apply it naively to numerical applications [16, 17]. On the other hand, it is shown that such a degeneracy is resolved by utilizing the Yangian symmetry of the spin Calogero-Sutherland model [18, 19]. Indeed the orthogonal basis for the spin Calogero-Sutherland model is constructed by applying such a basis. The Yangian Gelfand-Zetlin basis for the spin Calogero-Sutherland model is realized as the root of unity limit of the q-deformed theory [19]. This means we have an alternative method to describe 1 the SU(M )-singlet FQH state, i.e. the root of unity limit of the Macdonald polynomial, which is called the Uglov polynomial [20], instead of the non-symmetric Jack polynomial. In this paper we investigate a novel description of the FQH states with the spin degrees of freedom by utilizing the Yangian Gelfand-Zetlin basis, which is well investigated in [19] for the spin Calogero-Sutherlnad model. First the SU(M )-singlet condition is discussed by considering the generic property of the Lie algebra. We obtain the Fock condition as well as the standard SU(2)-singlet Halperin state. Starting from the Laughlin state, we construct the spin-singlet Halperin state through the q-deformation of the U(1) primary field. This procedure is based on the Yangian Gelfand-Zetlin basis for the spin Laplace- Beltrami operator, which is well studied from the viewpoint of the spin Calogero-Sutherland model. We then show the relation between the SU(M )-singlet state and a certain spinless state in terms of the occupation number representation. The squeezing rule for the SU(M ) state is consistently translated into the spinless basis. We finally comment on the underlying CFT for the SU(M )-singlet state, and its relation to the four dimensional gauge theory on the orbifold and the root of unity limit of the q-deformed CFT. 2 Spin-singlet FQH states Before discussing the spinful FQHE states, let us first introduce the most fundamental ex- ample of the FQH state, which is called the Laughlin state [21], N(cid:89) ΦL({zi}) = (zi − zj)r. For convenience we now omit the Gaussian factor, exp(cid:2)−(cid:80) i<j i zi2/(4(cid:96)2)(cid:3), with the magnetic (2.1) length being (cid:96)2 = /(eB). The filling fraction of this state is given by ν = 1/r, and the power r has to be odd integer due to the anti-symmetricity of the wavefunction for a fermionic system. Thus this wavefunction gives rise to the odd denominator series of the FQH states. Actually this Laughlin state does not include internal degrees of freedom. However we will see the spin-singlet FQH states are built from this fundamental wavefunction. 2.1 SU(2) theory A natural generalization of this Laughlin state (2.1) is the Halperin state [8], which accom- panies SU(2) spin degrees of freedom, N↑(cid:89) N↓(cid:89) (wi − wj)r(cid:89) (zi − wj)s. (2.2) ΦH({zi, wi}) = (zi − zj)r i<j i<j i,j Here zi and wi stand for posisions of up- and down-spin particles, respectively. Since we usually investigate the FQH system in the presence of strong magnetic fields, it is natural to consider fully spin-polarized states. This means the spin degrees of freedom are frozen in 2 such a case. However, when there are pseudo-spin degrees, e.g. valley degeneracy, multi-layer systems and so on, the spinful FQH state would provide a good description. We consider the spin operators for this wavefunction to discuss the spin-singlet state.When the spin-raising operator S+,i acts on this wavefunction (2.2), one of the down-spin particles is changed into the up-spin state, i.e. wi → zN↑+1. We often omit the index labeling the particle. If the state is fermionic, it has to be anti-symmetrized with all the up-spin particles, S+ΦH({zi, wi}) = ΦH({zi, wi}) − ΦH(zj ↔ wi). (2.3) Thus the spin-raising operator yields j=1 e(wi, zj) (for fermions), (2.4) N↑(cid:88) S+ = 1 − j=1 where the operator e(wi, zj) exchanges wi and zj. On the other hand, for a bosonic state, the wavefunction has to be symmetrized with up-spin particles. This means the sign factor arising in the spin-raising operator is modified as N↑(cid:88) N↑(cid:88) j=1 N↓(cid:88) S− = 1 ∓ S+ = 1 + e(wi, zj) (for bosons). (2.5) Similarly an up-spin particle is changed into the down-spin state under the spin-lowring operation, zi → wN↓+1. Therefore the spin-lowering operator is written as e(zi, wj). (2.6) j=1 The sign factor takes minus for fermionic and plus for bosonic systems. The other spin operator is simply given by Sz = (N↑ − N↓)/2. Therefore the spin-singlet condition, which is given by S+ΦH({zi, wi}) = 0, S−ΦH({zi, wi}) = 0, SzΦH({zi, wi}) = 0, (2.7) yields r = s + 1 and N↑ = N↓ for the Halperin state (2.2). This is just regarded as the Fock condition [22]. This also implies that r is odd/even for fermions/bosons. 2.2 SU(M ) theory We extend the singlet-condition (2.7) to arbitrary M -component systems. A simple gener- alization of the Halperin state (2.2) is given by H ({z(u) ΦM i }u=1,···,M ) = i − z(u) (z(u) j )r M(cid:89) N (u)(cid:89) M(cid:89) (cid:89) i − z(v) (z(u) j )s. (2.8) u=1 i<j u<v i,j 3 This is an M -state wavefunction: z(u) i stands for a position of an i-th u-state particle. To assign a similar manipulation to this wavefunction, we then need the raising and lowering operators for SU(M ) group. According to the general theory of the Lie algebra, the generators can be split into Hi, Ei and Fi. Here Ei and Fi are just regarded as the raising and lowering operators, while Cartan subalgebra consists of Hi. They are simply represented by introducing creation and annihilation operators for the u-th state, a(u)† and a(u), Hu = a(u)†a(u) − a(u+1)†a(u+1) (u = 1,··· , M − 1), E(u,v) = a(u)†a(v) F(u,v) = a(u)†a(v) (u > v), (u < v). (2.9) (2.10) (2.11) This means E(u,v) and F(u,v) convert v-th state into u-th state. For u > v and u < v, we call it raising and lowering, respectively. Remark the numbers of these generators are M (M − 1)/2 for E and F , M − 1 for Hi. Thus the total dimension of SU(M ) becomes M 2 − 1. For † † ↑a↑− a ↓a↓)/2. example, SU(2) algebras are represented as S+ = a The total number of the generators is consistent with dim SU(2) = 3. † † ↑a↓, S− = a ↓a↑ and Sz = (a We then explicitly show the singlet condition for SU(M ) theory. The corresponding condition to (2.7) is given by HuΦM H = 0, E(u,v)ΦM H = 0, F(u,v)ΦM H = 0. (2.12) The first one is simply satisfied when N (1) = N (2) = ··· = N (M ). As the case of SU(2), the other operators are represented as 1 ∓ N (u)(cid:88) i=1 (cid:40) e(z(u) i , z(v) j ) = E(u,v) F(u,v) (u > v) (u < v) . (2.13) Again the sign factor takes minus for fermionic and plus for bosonic systems. Thus we have an essentially the same condition for SU(M ) theory as the SU(2) theory. The singlet condition is just given by r = s + 1 and N (1) = N (2) = ··· = N (M ). Remark the filling fraction of the Halperin state (2.8) satisfying this SU(M )-singlet state is given by ν = M M (r − 1) + 1 . (2.14) 3 From Laughlin to Halperin We show the spin-singlet FQH state can be obtained from a certain spinless state. Here we concentrate on the Abelian FQH state, i.e. the Laughlin state. Its validity for the generic cases is discussed in section 4. The method we discuss here is based on the Yangian Gelfand-Zetlin basis for the spin Calogero-Sutherland model [19], which is also applied to the gauge theory partition function 4 for the orbifold theory [23, 24]. We often use the non-symmetric Jack polynomial to describe the spinful model. However there are degeneracies in its spectrum due to the internal spin degrees of freedom. By resolving the spin degeneracy with the Yangian symmetry, we obtain the alternative basis for the spinful model, which can be written in terms of symmetric polynomials. We apply this method to the FQH states to obtain an alternative way of describing the spinful FQH states. 3.1 The q-bosonic field We introduce the q-deformed FQH state by considering the q-bosonic field. The q-boson is quite similar to the standard free boson field, satisfying the slightly modified commutation relations [25, 26], [an, am] = n 1 − qn 1 − tn δn+m,0, [an, Q] = 1 r δn,0, where we parametrize t = qr. The corresponding bosonic field is given by ϕ(z) = Q − a0 log z − i (cid:88) n(cid:54)=0 1 − tn 1 − qn anz−n. 1 n (3.1) (3.2) As the case of the standard CFT, the operator product expansion (OPE) of this q-boson plays an essential role in considering the conformal block, which is interpreted as the FQH wavefunction. The singular part of the OPE is given by (cid:20) (w/z; q)∞ where we use the Pochhammer symbol, (x; q)n =(cid:81)n−1 N(cid:89) ϕ(z)ϕ(w) ∼ − log (tw/z; q)∞ (cid:104)V (z1)··· V (zN )(cid:105) = (cid:21) (zi/zj; q)∞ (tzi/zj; q)∞ i<j m=0(1− xqn). As a result, the conformal block of the q-primary field V (z) = : exp(iϕ(z)) : is calculated by utilizing this OPE relation, zr , (3.3) zr j . (3.4) Remark this is almost the same as the q-Vandermonde determinant, which is just the weight function of the Macdonald polynomial [27], with the extra contribution of the zero mode. Actually, as well known, the Laughlin state is interpreted as the conformal block of the U(1) primary fields. Thus, we now regard this conformal block (3.4) as the q-Laughlin state. 3.2 The root of unity limit We can easily show the q-deformed state (3.4) goes back to the standard Laughlin state (2.1) by taking the limit, t = qr and then q → 1. Note that the Macdonald polynomial is reduced to the Jack polynomial in this limit. On the other hand, to implement the Yangian Gelfand-Zetlin basis, we take another limit of the q-state, i.e. the root of unity limit, q −→ ωM q, t −→ ωM t = ωM qr, q −→ 1, (3.5) 5 Figure 1: The branch of Z = zM for M = 4. Left and right show Z and z planes, respectively. where ωM = exp(2πi/M ) is the M -th primitive root of unity. This limit is singular because there are inifinite products appearing in the q-state, thus its convergence radius is q = 1. Therefore we have to deal with it appropriately [23, 24]. Using the formula for the root of unity limit (3.5) [23, 24], (z; q)∞ (tz; q)∞ −→ (1 − zM )(r−1)/M (1 − z), the OPE of the q-state (3.2) in the limit (3.5) turns out to be ϕ(z)ϕ(w) ∼ − log(zM − wM )(r−1)/M (z − w). To discuss the spin-singlet wavefunction, we now introduce another set of coordinates, zM = Z, wM = W. (3.6) (3.7) (3.8) This means the original one is described as z = ωnz M W 1/M , with nz, nw = 0, 1,··· , M − 1. Here nz and nw stand for the branches of the M -th roots of Z and W , respectively, as shown in Fig. 1. We rewrite the OPE (3.7) with these variables, M Z1/M , w = ωnw ϕ(Z)ϕ(W ) ∼ − log(Z − W )(r−1)/M (ωnz M Z1/M − ωnw M W 1/M ). (3.9) We then show the second part of this OPE singularity depends on the choice of the M -th log(ωnz root branch. In the case of nz = nw, its singular behavior is given by M W 1/M ) ∼ log(Z − W ). (3.10) This is because when we write the right hand side as (Z−W ) = (Z1/M −W 1/M )(Z(M−1)/M + Z(M−2)/M W 1/M + ··· + W (M−1)/M ), the latter part becomes regular at Z ∼ W . Thus the OPE (3.9) should be rewritten as M Z1/M − ωnw ϕ(Z)ϕ(W ) ∼ − log(Z − W )(r−1)/M +1. (3.11) On the other hand, when nz (cid:54)= nw, log(ωnz other words, we cannot obtain the form of log(Z − W ) from log(ωnz M Z1/M − ωnw M W 1/M ) is regular at Z ∼ W . M Z1/M − ωnw In M W 1/M ), as 6 discussed in the case of nz = nw, without modifying its singular behavior at Z ∼ W due to the identity ω(M−1)n at x = y, n (cid:54)≡ 0 (mod M ). xM−2y + ··· + yM−1 = 0 xM−1 + ω(M−2)n (3.12) M M Therefore the OPE (3.9) behaves as ϕ(Z)ϕ(W ) ∼ − log(Z − W )(r−1)/M . (3.13) As a result, the singularity of the OPE (3.9) depends on the choice of the branch for the M -th root of the variable. Thus we parametrize positions of particles as zi = ωu )1/M , satisfying zM . Then the FQH state, which is represented as the corresponding M (z(u) I i = z(u) I conformal block of the primary field, is given by M(cid:89) (cid:89) I − z(v) (z(u) J )(r−1)/M . (3.14) M(cid:89) N (u)(cid:89) Φ({z(u) i }) = I − z(u) (z(u) J )(r−1)/M +1 u=1 I<J u<v I,J This multi-component FQH state is just the SU(M )-singlet Halperin state (2.8). Here we obtain this spin-singlet state from only one kind of the q-boson field by taking the root of unity limit. This is a specific result coming from its SU(M ) symmetry. We have to introduce the corresponding M types of bosons to construct a generic SU(M ) state, i.e. a non-singlet state. Although the singlet-state (3.14) includes the index labeling the particle state, u = 1,··· , M , its role can be freely changed due to the SU(M )-singlet property. The filling fraction of this state is obtained by substituting r → (r − 1)/M + 1 into the formula (2.14), ν = M M (( r−1 M + 1) − 1) + 1 = M r . (3.15) This is almost the same as the fraction of the Laughlin state ν = 1/r, up to the factor corresponding to the number of internal states. This factor is interpreted as a result of changing the variables as zM . i = z(u) I 4 Spinful FQH states with admissible condition We extend the relation between the spinless and spin-singlet FQH wavefunctions to more generalized FQH states. In this section we describe them in terms of the occupation number represenatation with emphasis on the admissible condition and its connection to the Jack polynomial, which is useful to study the generic FQH states. The scheme discussed in this section can describe not only the Abelian FQH state shown in section 3, but also the non- Abelian state. Thus it is expected that the spinless description of the spin-singlet state is possibe even for the non-Abelian state. 7 4.1 Admissible condition We introduce the occupation number representation of the FQH states.1 We represent a partition λ = (λ1, λ2,··· , λN ) with length (cid:96)λ ≤ N as a (bosonic) occupation number configuration n(λ) = {nm(λ), m = 0, 1, 2,···}. This means each of the lowest Landau level (LLL) orbitals, where nm(λ) is the multiplicity of m in λ. We can implement the dominance order for partitions λ > µ, with the squeezing rule which connects configurations n(λ) → n(µ). The ground state of the FQH system on the sphere can be uniquely represented with the root partition, which is the most dominant partition with fixing the particle number. The Jack polynomial is a symmetric polynomial labeled by a partition λ, which can be expanded with non-interacting states (monomial polynomials) obtained by squeezing procedure, This is just an eigenstate of the Laplace-Beltrami operator, λ ({zi}) = mλ({zi}) + J α (cid:18) (cid:88) N(cid:88) (cid:19)2 zi ∂ ∂zi + 1 α i=1 i<j HLB = (cid:88) µ<λ cλµ(α)mµ({zi}). (cid:18) zi + zj zi − zj zi ∂ ∂zi − zj ∂ ∂zj (cid:19) . (4.1) (4.2) It has been shown that when the parameter α is negative as α = −(k + 1)/(r − 1), the Jack polynomial obeys the admissible condition [4], λi − λi+k ≥ r. (4.3) The root configuration for the (k, r)-admissible state, whose Lz component of angular mo- mentum is zero, is given by Φ(k,r)(cid:105) = k0r−1k0r−1k0r−1 ···(cid:105) and the corresponding magnetic flux for the spherical system yields Nφ = r k N − r ≡ 1 ν N − r. (4.4) Therefore the filling fraction reads ν = k/r. The Laughlin state (2.1) satisfies the (k = 1, r) admissible condidtion, ΦL(cid:105) = 10r−110r−1 ···(cid:105). Note that the non-Abelian FQH states, e.g. Moore-Read [28] and Read-Rezayi states [29], can be also described in this way: the RR state associated with Zk parafermion obeys the (k, 2)-admissible condition [1, 2, 3]. We then consider the mapping from the spinless to the spin-singlet states, discussed in section 3, in terms of these representations. The corresponding partition to the root configuration for the (k, r, M = 1) state k0r−1k0r−1 ···(cid:105) is given by λ = (λ1, λ2,··· , λN ) (cid:124) (cid:123)(cid:122) k = (r (N/k − 1) ,··· , r (N/k − 1) ,··· , 2r,··· , 2r (cid:125) (cid:124) (cid:123)(cid:122) k (cid:125) , r,··· , r (cid:124) (cid:123)(cid:122) (cid:125) , 0,··· , 0 (cid:124) (cid:123)(cid:122) (cid:125) ) k k (4.5) 1See, for example, [1, 2, 3]. 8 where N is the particle number and we can see λ1 = Nφ defined in (4.4). In order to obtain the spin-singlet state as discussed in section 3, we then introduce the following partition from this root configuration (4.5), which corresponds to the modified variables (3.8), λ = ,··· , (cid:21) (cid:18)(cid:20) λ1 (cid:124) M (cid:21)(cid:19) (cid:20) λN M , (cid:20) λN−1 (cid:21) (cid:123)(cid:122) M k = ([r (N/k − 1) /M ],··· , [r (N/k − 1) /M ] ,··· , [r/M ],··· , [r/M ] , [0],··· , [0] ). (cid:125) (cid:124) (cid:123)(cid:122) k (cid:125) (cid:124) (cid:123)(cid:122) k (cid:125) Here [x] denotes the floor function, providing the largest integer not greater than x. We can read the following relation between the parameters of the spinless and the singlet state from (4.6) the expression shown in (3.14) r − 1 M r = + 1 ⇐⇒ r = M (r − 1) + 1. Thus each component is written as p = p(r − 1) + r M p M , p ∈ Z. Therefore we have(cid:104) r M (cid:105) −(cid:104) r M p (cid:105) (p − 1) (cid:40) = r − 1 for p (cid:54)≡ 0 (mod M ) for p ≡ 0 (mod M ) r . (4.7) (4.8) (4.9) In the occupation number basis this configuration is represented as X0r−1X0r−1 ···(cid:105) with X = k(0)0r−2k(1)0r−2 ··· k(M−1). (4.10) Superscripts stand for λi modulo M , and correspond to internal degrees of freedom of the SU(M )-singlet state. This is just the root configuration for the sipn-singlet FQH state [14, 15]. Note that the state u = 0 is equivalent to u = M . The SU(M )-singlet (k, r)-state is obtained from the spin Laplace-Beltrami operator [15] (cid:19)2 (cid:18) N(cid:88) i=1 (cid:88) i<j (cid:18) (cid:19) (cid:88) i(cid:54)=j HsLB = zi ∂ ∂zi + 1 α zi + zj zi − zj zi ∂ ∂zi − zj ∂ ∂zj − 1 α (1 − Kij) zizj (zi − zj)2 , (4.11) where Kij stands for the exchange operator. In this case the condition (4.3) is modified as If λi − λi+k = r − 1, the spin part satisfies λi − λi+k ≥ r − 1. σi > σi+k. (4.12) (4.13) Note that the definition we use in this paper is slightly different from the notation used in [15]. Thus we can see (4.10) is just the root configuration for the SU(M )-singlet (k, r)-state. 9 Conversely, starting with an M -component state labeled by (λ, σ), where the latter stands for the internal degree with 0 ≤ σ ≤ M −1, the corresponding spinless state can be recovered as λi = M λi + σi. (4.14) Essentially this relation has been already shown in [19, 30, 23, 24], and corresponds to the method discussed in section 3. This translation procedure is already explicitly shown in [19] for the positive parameter case α > 0. 4.2 Squeezing rule We study the squeezing rule for the spin-singlet state in terms of the spinless basis. From the root configuration we squeeze a-th and b-th components in i-th and j-th blocks of X given in (4.10), and exchange their states, as shown in Fig. 2. First we consider the case i < j. This operation is represented in terms of (λ, σ) as ((i − 1)(M (r − 1) + 1) + a(r − 1), a) −→ ((i − 1)(M (r − 1) + 1) + a(r − 1) + 1, b) , ((j − 1)(M (r − 1) + 1) + b(r − 1), b) −→ ((j − 1)(M (r − 1) + 1) + b(r − 1) − 1, a) . (4.15) The corresponding operation in terms of the spinless basis λ is just given by squeezing M − a + b boxes, M [(i − 1)(M (r − 1) + 1) + a(r − 1)] + a −→ M [(i − 1)(M (r − 1) + 1) + a(r − 1) + 1] + b, M [(j − 1)(M (r − 1) + 1) + b(r − 1)] + b −→ M [(j − 1)(M (r − 1) + 1) + b(r − 1) − 1] + a. (4.16) We can see M − a + b > 0, since they satisfy 0 ≤ a, b ≤ M − 1. Next let us study the case i = j. This operation is given by ((i − 1)(M (r − 1) + 1) + a(r − 1), a) −→ ((i − 1)(M (r − 1) + 1) + a(r − 1), b) , ((j − 1)(M (r − 1) + 1) + b(r − 1), b) −→ ((j − 1)(M (r − 1) + 1) + b(r − 1), a) , (4.17) and the corresponding one yields M [(i − 1)(M (r − 1) + 1) + a(r − 1)] + a −→ M [(i − 1)(M (r − 1) + 1) + a(r − 1)] + b, M [(j − 1)(M (r − 1) + 1) + b(r − 1)] + b −→ M [(j − 1)(M (r − 1) + 1) + b(r − 1)] + a. (4.18) In this case the number of squeezed boxes is b − a. Therefore we can perform this operation only when b > a. Fig. 2 shows the squeezing rule for the SU(M ) states and the corresponding spinless description. 10 Figure 2: The squeezing rule for the singlet admissible states in terms of (a) the original spinful configuration and (b) the spinless description. The parameters are related as r = M (r − 1) + 1. As an example we now consider the (k, r, M ) = (1, 2, 2) state, which is well investigated in [15]. The corresponding spinless state yields the (k, r, M ) = (1, 3, 1) state. The root configuration for N = 4 is given by ↑,↓, 0,↑,↓(cid:105). (4.19) This is also represented by the following partitions, (λ, σ) = (4(↑), 3(↓), 1(↑), 0(↓)) ⇐⇒ λ = (9, 6, 3, 0), (4.20) where we assign ↑≡ 1, ↓≡ 0 (mod 2). The descendants of the root configuration (4.19) are given by ↑,↓, 0,↓,↑(cid:105) = (4(↑), 3(↓), 1(↓), 0(↑)) ⇐⇒ (9, 6, 2, 1) ↓,↑, 0,↑,↓(cid:105) = (4(↓), 3(↑), 1(↑), 0(↓)) ⇐⇒ (8, 7, 3, 0) ↑,↓, 0,↑,↓(cid:105) = (4(↓), 3(↑), 1(↓), 0(↑)) ⇐⇒ (8, 7, 2, 1) ↑,↑, 0,↓,↓(cid:105) = (4(↓), 3(↓), 1(↑), 0(↑)) ⇐⇒ (8, 6, 3, 1) (4.21) The last one is equivalent to ↓,↓, 0,↑,↑(cid:105) due to the spin-singlet condition, which is also represented as ↓,↓, 0,↑,↑(cid:105) = (4(↑), 3(↑), 1(↓), 0(↓)) ⇐⇒ (9, 7, 2, 0). (4.22) Note that (9, 7, 2, 0) cannot be obtained from the root configuration (9, 6, 3, 0) by squeezing it. Therefore, to be consistent with the symmetry, it turns out to be the zero-weight state, as discussed in [14, 15]. 5 Relation to conformal field theory One of the most important properties of the FQH state is its exotic statistics. Such a crucial property can be well described in terms of the two dimensional CFT: the non-Abelian 11 statistics is interpreted as a consequence of the fusion rule of the corresponding CFT. In this paper we have discussed a novel description of the spin-singlet state by utilizing the q- deformed state. Thus, in this section, we comment on CFT relevant to the spin-singlet FQH state with emphasis on its relation to the q-deformed CFT. The typical underlying CFT for the FQH state is the Zk-parafermion model. It describes the k-th Read-Rezayi state [29], which includes the Moore-Read state [28] at k = 2. The Zk-parafermion model can be obtained from the coset construction, SU(2)k U(1) . (5.1) Furthermore, it has been shown the generic (k, r)-admissible state is coming from the ex- tended chiral algebra, WAk−1(k + 1, k + r) [5, 6, 7]. A CFT model for the FQH state has been also investigated for the spin-singlet states. The SU(M )-singlet state corresponds to the generalized parafermion model, SU(M + 1)k U(1)M . (5.2) This is a natural extention of the k-th Read-Rezayi state [10, 11]. The interesting property of the parafermion models is the level-rank duality.2 For exam- ple, for the k-th Read-Rezayi state, we have SU(k)1 × SU(k)1 SU(2)k U(1) (5.3) This means the Zk-parafermion is also realized as the lowest one of the W-minimal series, SU(k)2 = . SU(k)l × SU(k)1 SU(k)l+1 , (5.4) which gives rise to the central charge, c = k − 1 − k(k2 − 1) (p − q)2 (5.5) Here we identify k + l = p/(q − p), k + l + 1 = q/(q − p). This duality is also applied to the spin-singlet theory, pq . SU(M + 1)k U(1)M = (SU(k)1)M +1 SU(k)M +1 . (5.6) Recently the corresponding CFT series, which reproduces the model in the right hand side of (5.6) at the lowest level, has been proposed [15], SU(k)l × (SU(k)1)M SU(k)M +l . (5.7) Substituting k + l = M p/(q − p), k + l + M = M q/(q − p), the central charge is given by c = M (k − 1) − k(k2 − 1) M (p − q)2 pq . (5.8) 2See a review article, e.g. [31]. 12 This novel CFT series is not well investigated yet, but let us comment on its connection to the four dimensional gauge theory. Recent progress on the supersymmetric gauge theory reveals the remarkable relation between the four dimensional gauge theory and the two dimensional CFT [32]: the instanton partition function of the gauge theory is directly interpreted as the conformal block of the two dimensional CFT.3 The standard SU(N ) Yang-Mills theory corresponds to the AN−1 Toda CFT. Its central charge is given by c = N − 1 + N (N 2 − 1)Q2, (5.9) where Q is related to the regularization parameter of the four dimensional theory. This CFT is essentially the same as the model shown in (5.4): they are equivalent under the identification Q2 = −(p − q)2/(pq). A similar connection is also suggested for the generalized model (5.7). In this case the gauge theory on the type AM−1 ALE space, which is given by resolving the singularity of the orbifold C2/ZM , gives rise to the corresponding CFT [34]. Its central charge is given by c = M (N − 1) + N (N 2 − 1) Q2. M When we consider a gauge group G, the central charge is written in a generic form, c = M rG + dGhG M Q2, where rG, dG and hG are rank, dimension and the dual Coxeter number of the group G, respectively. This generalized CFT would be realized as the following coset model, Gl × (G1)M GM +l . (5.12) (5.10) (5.11) Let us then comment on the q-deformed CFT. It has been shown that, by taking the limit, t = qr and then q → 1, the central charge of the corresponding CFT is given by [25, 26] . When we start with the q-deformed W CFT, it becomes r c = 1 − 6 (1 − r)2 c = N − 1 − N (N 2 − 1) (1 − r)2 r (5.13) (5.14) . Indeed they are equivalent to the SU(N ) minimal models via r = p/q. Thus, when we apply the root of unity limit, q → ωM q, t → ωM qr and q → 1, it is natural to obtain the CFT, which describes the SU(M )-singlet FQH state as a result of the Yangian Gelfand-Zetlin basis. Its central charge is expected to be given by c = M (N − 1) − N (N 2 − 1) M (1 − r)2 r . (5.15) This model corresponds to the standard Macdonald polynomial, which is associated with the type A root system. Thus we now expect that the central charge of the model, which is coming from the q-deformed theory related to other root systems, e.g. the BC type theory [35, 36], can be written in a form similar to (5.11). 3An attempt to connect the gauge theory partition function with the FQH state is found in [33]. 13 6 Conclusion We have investigated the SU(M )-singlet FQH states with the spinless basis. We have shown the raising and lowering operators for SU(M ) states with emphasis on its similar- ity to the standard SU(2) spinful states. The SU(M )-singlet condition can be written in a quite similar form to the SU(2)-singlet condition for the Halperin state. We have obtained the SU(M )-singlet Halperin state from the corresponding Laughlin state by considering the q-deformation and its root of unity limit. This is just the prescription to impliment the Yangian Gelfand-Zetlin basis in terms of a certain spinless state, which is proposed in [19]. As well known, the FQH trial wavefunction is regarded as the correlation function of the primary fields in the corresponding CFT. Thus, to discuss such a correlation function, we have studied the q-boson fields and its OPE in the root of unity limit. We have also shown the relation between the SU(M )-singlet and the corresponding spinless states for much gen- eralized FQH states, which obey the (k, r)-admissible condition. We have discussed them in terms of the occupation number representation of the FQH states, and thus the squeezing rule is naturally assigned for them. We then have commented on the underlying CFT for the FQH states discussed in this paper. There would be the interesting structure in the CFT for the multi-component states, and the root of unity limit of the q-deformed CFT. Acknowledgments The author would like to thank Y. Hidaka and T. Nishioka for useful comments. The author is supported by Grant-in-Aid for JSPS Fellows (No. 23-593). References [1] B. A. Bernevig and F. D. M. Haldane, "Model Fractional Quantum Hall States and Jack Polynomials," Phys. Rev. Lett. 100 (2008) 246802, arXiv:0707.3637 [cond-mat.mes-hall]. [2] B. A. Bernevig and F. D. M. Haldane, "Generalized clustering conditions of Jack polynomials at negative Jack parameter α," Phys. Rev. B77 (2008) 184502, arXiv:0711.3062 [cond-mat.mes-hall]. [3] B. A. Bernevig and F. D. M. Haldane, "Properties of Non-Abelian Fractional Quantum Hall States at Filling ν = k/r," Phys. Rev. Lett. 101 (2008) 246806, arXiv:0803.2882 [cond-mat.mes-hall]. [4] B. Feigin, M. Jimbo, T. Miwa, and E. Mukhin, "A differential ideal of symmetric polynomials spanned by Jack polynomials at β = −(r − 1)/(k + 1)," Int. Math. Res. Notices 2002 (2002) 1223 -- 1237, arXiv:math/0112127 [math.QA]. 14 [5] B. A. Bernevig, V. Gurarie, and S. H. Simon, "Central charge and quasihole scaling dimensions from model wavefunctions: toward relating Jack wavefunctions to W-algebras," J. Phys. A42 (2009) 245206, arXiv:0903.0635 [cond-mat.mes-hall]. [6] B. Estienne and R. Santachiara, "Relating Jack wavefunctions to WAk−1 theories," J. Phys. A42 (2009) 445209, arXiv:0906.1969 [hep-th]. [7] B. Estienne, B. A. Bernevig, and R. Santachiara, "Electron-quasihole duality and second-order differential equation for Read-Rezayi and Jack wave functions," Phys. Rev. B82 (2010) 205307, arXiv:1005.3475 [cond-mat.str-el]. [8] B. I. Halperin, "Theory of the quantized Hall conductance," Helv. Phys. Acta 56 (1983) 75 -- 102. [9] F. D. M. Haldane and E. H. Rezayi, "Spin-singlet wave function for the half-integral quantum Hall effect," Phys. Rev. Lett. 60 (1988) 956 -- 959. [10] E. Ardonne and K. Schoutens, "New Class of Non-Abelian Spin-Singlet Quantum Hall States," Phys. Rev. Lett. 82 (1999) 5096 -- 5099, arXiv:cond-mat/9811352 [cond-mat.mes-hall]. [11] E. Ardonne, N. Read, E. Rezayi, and K. Schoutens, "Non-abelian spin-singlet quantum Hall states: wave functions and quasihole state counting," Nucl. Phys. B607 (2001) 549 -- 576, arXiv:cond-mat/0104250. [12] M. Kasatani, "Subrepresentations in the polynomial representation of the double affine Hecke algebra of type GLn at tk+1qr−1 = 1," Int. Math. Res. Not. 2005 no. 28, (2005) 1717 -- 1742, arXiv:math/0501272 [math.QA]. [13] M. Kasatani and V. Pasquier, "On Polynomials Interpolating Between the Stationary State of a O(n) Model and a Q.H.E. Ground State," Comm. Math. Phys. 276 (2007) 397 -- 435, arXiv:cond-mat/0608160 [cond-mat.stat-mech]. [14] E. Ardonne and N. Regnault, "The structure of spinful quantum Hall states: a squeezing perspective," Phy. Rev. B84 (2011) 205134, arXiv:1107.2232 [cond-mat.str-el]. [15] B. Estienne and B. A. Bernevig, "Spin-Singlet Quantum Hall States and Jack Polynomials with a Prescribed Symmetry," Nucl. Phys. B857 (2012) 185 -- 206, arXiv:1107.2534 [cond-mat.str-el]. [16] B. A. Bernevig and N. Regnault, "Anatomy of Abelian and Non-Abelian Fractional Quantum Hall States," Phys. Rev. Lett. 103 (2009) 206801, arXiv:0902.4320 [cond-mat.mes-hall]. 15 [17] R. Thomale, B. Estienne, N. Regnault, and B. A. Bernevig, "Decomposition of fractional quantum Hall model states: Product rule symmetries and approximations," Phys. Rev. B84 (2011) 045127, arXiv:1010.4837 [cond-mat.str-el]. [18] K. Takemura and D. Uglov, "The orthogonal eigenbasis and norms of eigenvectors in the spin Calogero-Sutherland model," J. Phys. A30 (1997) 3685 -- 3718, arXiv:solv-int/9611006 [nlin.SI]. [19] D. Uglov, "Yangian Gelfand-Zetlin bases, glN -Jack polynomials and computation of dynamical correlation functions in the spin Calogero-Sutherland model," Commun. Math. Phys. 193 (1998) 663 -- 696, arXiv:hep-th/9702020 [hep-th]. [20] Y. Kuramoto and Y. Kato, Dynamics of One-Dimensional Quantum Systems: Inverse-Square Interaction Models. Cambridge University Press, 2009. [21] R. B. Laughlin, "Anomalous quantum Hall effect: An incompressible quantum fluid with fractionally charged excitations," Phys. Rev. Lett. 50 (1983) 1395 -- 1398. [22] E. Lieb and D. Mattis, "Theory of Ferromagnetism and the Ordering of Electronic Energy Levels," Phys. Rev. 125 (1962) 164 -- 172. [23] T. Kimura, "Matrix model from N = 2 orbifold partition function," JHEP 09 (2011) 015, arXiv:1105.6091 [hep-th]. [24] T. Kimura, "β-ensembles for toric orbifold partition function," Prog. Theor. Phys. 127 (2012) 271 -- 285, arXiv:1109.0004 [hep-th]. [25] J. Shiraishi, H. Kubo, H. Awata, and S. Odake, "A Quantum deformation of the Virasoro algebra and the Macdonald symmetric functions," Lett. Math. Phys. 38 (1996) 33 -- 51, arXiv:q-alg/9507034 [math.QA]. [26] H. Awata, H. Kubo, S. Odake, and J. Shiraishi, "Quantum WN algebras and Macdonald polynomials," Commun. Math. Phys. 179 (1996) 401 -- 416, arXiv:q-alg/9508011 [math.QA]. [27] I. G. Macdonald, Symmetric Functions and Hall Polynomials. Oxford University Press, 2nd ed., 1997. [28] G. W. Moore and N. Read, "Nonabelions in the fractional quantum Hall effect," Nucl. Phys. B360 (1991) 362 -- 396. [29] N. Read and E. Rezayi, "Beyond paired quantum Hall states: Parafermions and incompressible states in the first excited Landau level," Phys. Rev. B59 (1999) 8084 -- 8092, arXiv:cond-mat/9809384 [cond-mat.mes-hall]. 16 [30] R. Dijkgraaf and P. Su(cid:32)lkowski, "Instantons on ALE spaces and orbifold partitions," JHEP 03 (2008) 013, arXiv:0712.1427 [hep-th]. [31] P. Bouwknegt and K. Schoutens, "W symmetry in conformal field theory," Phys. Rept. 223 (1993) 183 -- 276, arXiv:hep-th/9210010. [32] L. F. Alday, D. Gaiotto, and Y. Tachikawa, "Liouville Correlation Functions from Four-dimensional Gauge Theories," Lett. Math. Phys. 91 (2010) 167 -- 197, arXiv:0906.3219 [hep-th]. [33] R. Santachiara and A. Tanzini, "Moore-Read Fractional Quantum Hall wavefunctions and SU(2) quiver gauge theories," Phys. Rev. D82 (2010) 126006, arXiv:1002.5017 [hep-th]. [34] T. Nishioka and Y. Tachikawa, "Para-Liouville/Toda central charges from M5-branes," Phy. Rev. D84 (2011) 046009, arXiv:1106.1172 [hep-th]. [35] T. Koornwinder, "Askey-Wilson polynomials for root system of type BC," Contemp. Math. 138 (1992) 189 -- 204. [36] M. Kasatani, "Zeros of symmetric Laurent polynomials of type (BC)n and Koornwinder-Macdonald polynomials specialized at tk+1qr−1 = 1," Comp. Math. 141 (2005) 1589 -- 1601, arXiv:math/0312327 [math.QA]. 17
1010.6047	1	1010	2010-10-28T18:49:04	Electronic states and optical properties of PbSe nanorods and nanowires	[ "cond-mat.mes-hall" ]	A theory of the electronic structure and excitonic absorption spectra of PbS and PbSe nanowires and nanorods in the framework of a four-band effective mass model is presented. Calculations conducted for PbSe show that dielectric contrast dramatically strengthens the exciton binding in narrow nanowires and nanorods. However, the self-interaction energies of the electron and hole nearly cancel the Coulomb binding, and as a result the optical absorption spectra are practically unaffected by the strong dielectric contrast between PbSe and the surrounding medium. Measurements of the size-dependent absorption spectra of colloidal PbSe nanorods are also presented. Using room-temperature energy-band parameters extracted from the optical spectra of spherical PbSe nanocrystals, the theory provides good quantitative agreement with the measured spectra.	cond-mat.mes-hall	cond-mat	Electronic states and optical properties of PbSe nanorods and nanowires A. C. Bartnik,1 Al. L. Efros,2 W.-K. Koh,3 C. B. Murray,3, 4 and F. W. Wise1 1Applied Physics Department, Cornell University, Ithaca, NY 14853 2Naval Research Laboratory, Washington, DC 20375 3Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 19104 4Department of Materials Science and Engineering, University of Pennsylvania, Philadelphia, Pennsylvania 19104 Abstract A theory of the electronic structure and excitonic absorption spectra of PbS and PbSe nanowires and nanorods in the framework of a four-band effective mass model is presented. Calculations con- ducted for PbSe show that dielectric contrast dramatically strengthens the exciton binding in narrow nanowires and nanorods. However, the self-interaction energies of the electron and hole nearly cancel the Coulomb binding, and as a result the optical absorption spectra are practically unaffected by the strong dielectric contrast between PbSe and the surrounding medium. Mea- surements of the size-dependent absorption spectra of colloidal PbSe nanorods are also presented. Using room-temperature energy-band parameters extracted from the optical spectra of spherical PbSe nanocrystals, the theory provides good quantitative agreement with the measured spectra. 0 1 0 2 t c O 8 2 ] l l a h - s e m . t a m - d n o c [ 1 v 7 4 0 6 . 0 1 0 1 : v i X r a 1 I. INTRODUCTION Solution-based chemical synthesis of semiconductor nanostructures has allowed tremen- dous flexibility in crystal morphology. After much work on zero-dimensional (0D) nanocrys- tals (NCs), attention is shifting to one-dimensional (1D) nanorods (NRs) and nanowires (NWs)1–4 and the variation of material properties in the transition from 0D to 1D. The electronic structure of these crystals is the foundation for understanding their properties. Previously, the electronic structure of 1D nanocrystals has been modeled using a variety of methods, including effective-mass theories based on k · p Hamiltonians5–9, pseudopotential techniques10–12, tight binding models13–17, and density functional theory9,17–21. The relax- ation of confinement in going from 0D to 1D goes hand–in–hand with an increase in the importance of Coulomb effects mediated through the nanocrystal's dielectric environment22. Lead–salt (PbS, PbSe, PbTe) nanocrystals offer unique advantages to study the interplay of these two effects. Their large exciton Bohr radii places them at the limit of strong confinement, while their large dielectric constants coupled with their mirror–like electron and hole spectra substantially reduce the Coulomb interaction in spherical quantum dots23,24. However, in a 1D structure the Coulomb interaction can act primarily through the host medium, so it will not be screened as effectively as in 0D7. Thus, the lead salts provide a unique system to study the transition from strong confinement to strong Coulomb binding as the length of the nanocrystal changes. Within k· p theory, the general treatment of the optical properties of NWs and NRs sur- rounded by media with small dielectric constant was developed in Refs. [5–7]. A type of adi- abatic approximation naturally separates the calculation into pieces. In recognition of strong confinement perpendicular to the NR or NW axis, one first calculates the 1D subband ener- gies and wavefunctions, while neglecting the Coulomb interaction. Next, using these wave functions of transverse electron and hole motion, one can calculate the longitudinal motion of the exciton, including corrections from image forces in the surrounding medium. To do that, the three-dimensional Coulomb potential is averaged to a one-dimensional Coulomb interaction between the electron and hole along the NW or NR axis. Using this potential, the spectra of 1D excitons and their transition oscillator strengths are found. Finally, in NRs one should find the spectrum of the exciton center of mass motion, in order to include this additional effect of confinement. The main aspects of this framework were performed 2 for lead–salt nanowires recently by Rupasov25, although approximations to the simplified band structure used in that paper preclude the description of real experimental results. In this paper we present calculations of the 1D subband energy spectra of lead–salt nanowires with arbitrary axis orientation, taking into account the multi-valley structure and accurate electron and hole energy-level dispersions in these semiconductors. For PbSe NWs with axis along the h100i direction, we calculate the spectra of 1D excitons including self- interaction corrections. Surprisingly, the calculations show that although the binding energy of excitons in the smallest NWs reaches 350 meV, the optical transition energies are not affected by the small dielectric constant of the surrounding medium and are almost identical to the transitions between non-interacting electron and hole subbands. The cancelation of the exciton binding energy and the self-interaction corrections to the electron and hole levels is a consequence of the almost mirror symmetry of the conduction and valence bands of PbSe. The theoretical results agree well with the measured absorption spectra of h100i PbSe NRs. The paper is organized as follows. In Section II we will describe the Hamiltonian governing the 1D nanowire system, with solutions in Section III. In Section IV we present the effects of dielectric confinement and Coulomb forces on the 1D exciton, with 1D wavefunction solutions in Section V. Experimental data and comparison with theory are presented in Section VI, followed by discussion and conclusion. II. FOUR BAND EFFECTIVE MASS MODEL PbS, PbSe, and PbTe are direct-gap semiconductors, with extrema of the conduction and valence bands at the L points in the Brillouin zone. The energy bands near the L point can be well-described within the four–band k · p model26,27. This model takes into account the direct interaction between the nearest conduction and valence bands, as well as the contributions of the remote bands to the electron and hole effective masses. Following Ref. [24], we use the multiband effective mass approximation and expand the full wave functions inside the nanorod as Φ(r) = Xµ=±1/2 Ψc µ(r)L−6,µi + Xµ=±1/2 µ(r)L+ Ψv 6,µi , (1) 3 where L−6,µi and L+ 6,µi are the Bloch functions of the conduction band and valence band edge, respectively, at the L–point. The upper sign " ± " in the notation reflects the invariance of these functions with respect to the operation of spatial inversion. The smooth functions Ψc ±1/2( r ) and Ψv ±1/2( r ) are the components of the conduction band and valence band spinor envelopes, respectively: Ψc 1/2 Ψc Ψv Ψc =    , The bi–spinor envelope function Ψ =  Ψv  is the solution of the Schrodinger equation  H( p)Ψ = EΨ , where p = k = −i ∇ is the momentum operator, and the Hamiltonian H( p) of Ref. [24] can be written in compact form as Ψv =    . −1/2 −1/2 Ψv Ψc (2) 1/2 H(p) =  (cid:18) Eg 2 Pl m0 + p2 z 2m−l pz σz + + Pt m0 p2 ⊥ Pl m0 2m−t (cid:19) U2 (p⊥σ⊥) −(cid:18)Eg 2 pz σz + Pt m0 + p2 z 2m+ l + (p⊥σ⊥) p2 ⊥ 2m+ t (cid:19) U2   . (3) = p2 x + p2 free electron mass, p2 ⊥ (3) U2 is the 2 × 2 unit matrix, σ = {σx, σy, σz} are the Pauli matrices that In Eq. act on the spinor components of the wave functions ( µ = ±1/2 ), Eg is the bulk energy gap, E is the electron or hole energy measured from the middle of the gap, m0 is the y , (p⊥σ⊥) = pxσx + py σy , Pt and Pl are the transverse and longitudinal momentum matrix elements taken between the conduction and valence band edge Bloch functions24, and m±t and m±l are the remote-band contribution to the transverse and longitudinal band edge effective masses, respectively. For electrons and holes, 0Eg]−1 and l,t = [1/m+ 0Eg]−1 , respectively. In each valley, the z axis in Eq. (3) is directed mh toward the L–point of the Brillouin zone, e.g. along the h111i direction of the cubic lattice. As a result, for each of the four valleys, the z axis will point in different directions. these band edge effective masses can be expressed as me l,t = [1/m−l,t + 2P 2 l,t + 2P 2 l,t/m2 l,t/m2 Although the Hamiltonian of Eq. (3) has cylindrical symmetry with respect to, e.g., the h111i crystallographic direction, this direction may not coincide with the NR growth direction. For a description of NR electronic and optical properties it is convenient to use coordinates connected with the latter direction instead, even though the cylindrical symmetry of the Hamiltonian is generally broken. In PbS and PbSe, the small anisotropy of conduction and valence bands allows us to treat deviations from cylindrical symmetry 4 perturbatively. The Hamiltonian (3) can be written H = H0 + Han , where the cylindrically- symmetric part H0 is H0(p) =  (cid:18) Eg 2 Pz m0 + p2 z 2m−z pz σz + + P⊥ m0 Pz m0 p2 ⊥ 2m− ⊥(cid:19) U (p⊥ σ⊥) −(cid:18)Eg 2 The modified band parameters are pz σz + P⊥ m0 + p2 z 2m+ z + (p⊥σ⊥) p2 ⊥ 2m+ ⊥(cid:19) U Pt 2 (1 + cos2 θ) + Pl 2 (1 + cos2 θ) + 1 2m±t P⊥ = 1 m± ⊥ = sin2 θ 1 2m±l sin2 θ Pz = Pt sin2 θ + Pl cos2 θ 1 m±l sin2 θ + 1 m±z 1 m±t = . (4)   (5) (6) cos2 θ where θ is the angle between the growth axis and the h111i direction. The anisotropic part of the Hamiltonian is given in Appendix A. Note that Eq. (4) has a form identical to Eq. (3), but the z axis is now directed along the growth axis. For arbitrary orientation of the growth direction, there will be four different angles θ for each of the four valleys, and therefore four different sets of modified band parameters defined in Eq. (5). As a result, each valley will have unique electronic structure. The energy spectra associated with the different valleys become degenerate when the growth direction leads to identical values of θ for them. The highest degree of degeneracy is reached when the growth direction is along the h100i crystal axis. In this case all four valleys have the same θ ; cos2 θ = 1/3 , which results in P⊥ = Pz and m⊥ = mz in Eq. (4). All of the spectra are degenerate. The anisotropic part Han of the full Hamiltonian can be considered as a perturbation if Pl − Pt ≪ Pl + Pt and 1/m±l − 1/m±t ≪ 1/m±l + 1/m±t . The first-order corrections to the solutions of H0 caused by Han vanish in the 2-fold Kramers-degenerate subspace at each energy level. As a result, only second-order perturbation theory gives corrections to the energy levels. We will neglect these corrections from this point on, although an example higher-order calculation appears in Appendix A. III. ENERGY SPECTRA OF ELECTRONS AND HOLES IN PBSE NANOWIRES The first step in our modeling process is to find the energy spectra of 1D subbands of infinitely-long cylindrical nanowires, temporarily ignoring the Coulomb interaction. The 5 cylindrical symmetry of the Hamiltonian of Eq. (4) allows the solutions to take the form Ψn(kz) =   Rn 1 (ρ) exp(i(n − 1/2)φ) iRn 2 (ρ) exp(i(n + 1/2)φ) Rn 3 (ρ) exp(i(n − 1/2)φ) iRn 4 (ρ) exp(i(n + 1/2)φ)   exp(ikzz) , (7) where φ is the azimuthal angle, n = ±1/2,±3/2,±5/2, ... is the total angular momentum projection on the nanowire axes defined by the operator Jz = −i∂/∂φ + Sz , kz is the momentum along the nanowire z axis, and ρ = px2 + y2 is the radial coordinate in the plane perpendicular to the NW axis. The chosen phase of each component of the function Ψn(kz) allows the radial functions Rn (4) yields the system of differential equations that defines these functions: ∆n−1/2(cid:19)Rn ∆n+1/2(cid:19)Rn i (ρ) to be pure real. Substitution of Eq. (7) into Eq. 4 (ρ) = 0 , 4 (ρ) = 0 , P⊥ m0 1 (ρ) + 2 (ρ) + 2 2m− ⊥ 2 2m− ⊥ (cid:18)α− + (cid:18)α− + kzPz m0 Rn 1 (ρ) − − P⊥ D+ n−1/2Rn m0 − P⊥ m0 1 (ρ) + D−n+1/2Rn kzPz m0 Rn 3 (ρ) + kzPz m0 Rn P⊥ D+ n−1/2Rn m0 2 (ρ) +(cid:18)α+ + 2 (ρ) +(cid:18)α+ + D−n+1/2Rn kzPz m0 Rn ∆n−1/2(cid:19)Rn ∆n+1/2(cid:19)Rn 3 (ρ) − 2 2m− ⊥ 2 2m− ⊥ 3 (ρ) = 0 , 4 (ρ) = 0 , (8) (9) where α± = Eg/2 ± E + 2k2 z/(2m±z ) . The differential operators D±m = ∓ ∂ ∂ρ + m ρ are the raising and lowering operators D±mJm(kρ) = kJm±1(kρ) for the Bessel functions Jm(kρ) with integer index, and the operator ∆m = D−m+1 D+ m = −(1/ρ)(∂/∂ρ)ρ(∂/∂ρ) + m2/ρ2 . It is easy to check using the raising and lowering properties of the D±m operators that the radial eigenfunctions of Eqs. (8) should take the form   Rn 1 (ρ) Rn 2 (ρ) Rn 3 (ρ) Rn 4 (ρ)   = C1Jn−1/2(kρρ) C2Jn+1/2(kρρ) C3Jn−1/2(kρρ) C4Jn+1/2(kρρ)   .   6 (10) Substitution of this into Eqs. (8) yields a 4x4 system of linear equations for the coefficients C1,2,3,4 . Setting the determinant of this system to zero produces the relation between the quasi-momentum kρ and the energy of electrons or holes E : where 2k2 ρ = −α(E) ±pα(E)2 + β(E) , 2 (cid:19) − m− Eg Eg + ⊥(cid:18)E − 2 (cid:19)(cid:18)E − 2k2 z 2m−z − 2k2 Eg z 2m−z − 2k2 z 2m+ z Eg + ⊥(cid:18)E + 2k2 z 2m+ z ⊥(cid:18)E + α(E) = m+ β(E) = 4m+ ⊥ m− (11) 2k2 2P 2 ⊥ m2 m+ ⊥ ⊥ m− m+ m2 P 2 ⊥ ⊥ 2 (cid:19) + m− 2 (cid:19) − 4 ρ = α(E) +pα(E)2 + β(E) . ρ results (12) z . z From Eq. (11) it is clear that k2 ρ can be positive or negative. The negative value of k2 in an imaginary kρ = iλρ , with λρ defined by Eq. (11) as 2λ2 The complex arguments in Eq. (10) are then simplified by replacing the Bessel functions Jm(iλρρ) with the modified Bessel functions Im(λρρ) using the relationship Jm(iλρρ) = imIm(λρρ) . For each value of k2 ρ , there are two independent solutions of the 4x4 linear system for the coefficients C1,2,3,4 . These two solutions can be chosen such that either C3 = 0 or C4 = 0 , which allows the remaining coefficients Ci to be found. Taking into account the positive and negative value of k2 ρ , there are four independent solutions for each energy E . The energy spectrum is determined by the boundary conditions at the NW surface. The boundary conditions are defined on all four components of the wave function, which inside of the NW can be always written as a linear combination of the four degenerate solutions discussed above  Rn 1 (ρ, kz)  Rn 2 (ρ, kz) Rn 3 (ρ, kz) Rn 4 (ρ, kz)   = a  kρP⊥Jn−1/2(kρρ)  −kzPzJn+1/2(kρρ) ΓkJn+1/2(kρρ) 0   + b kzPzJn−1/2(kρρ) kρP⊥Jn+1/2(kρρ) ΓkJn−1/2(kρρ) 0 +   +c + d  λρP⊥In−1/2(λρρ)  −kzPzIn+1/2(λρρ) ΓλIn+1/2(λρρ) 0  kzPzIn−1/2(λρρ)  −λρP⊥In+1/2(λρρ) ΓλIn−1/2(λρρ) 0 , (13)       7 where Γk = z m− ⊥ ⊥ − λ2 and a , b , c , and d are determined by the boundary conditions. 2m− ⊥ 2m− ⊥ z m− Γλ = + k2 m0 m−z m−z (k2 (k2 m0 (cid:18)E − (cid:18)E − m0 Eg 2 (cid:19) − 2 (cid:19) − Eg m0 ρm−z ) , ρm−z ) , (14) For NWs with an impenetrable surface, the standard boundary conditions require each component of the wave function defined in Eq. (13) to vanish, leading to Rn i (R, kz) = 0 , where i = 1, 2, 3, 4 and R is the NW radius. These four equations define the 4x4 system for the a, b, c, d coefficients. Requiring the determinant of this system to be zero yields the following implicit equation for the 1D energy bands for angular momentum n , and as a function of the parameter kz : t (k2 ρΓ2 λ − λ2 ρΓ2 k) I n − +J n I n − J n + = 0 , (15) kρλρ(cid:2)(I n +)2(J n − )2 − (I n − )2(J n where we use the notation J n ± k2 z P 2 z (Γk − Γλ)2 + P 2 t ΓkΓλ +)2(cid:3) + = Jn±1/2(kρR) and I n ± P 2 = In±1/2(λρR) . After determining the energy from Eq. (15), the wavefunctions can be constructed from Eq. (13), with only the normalization undetermined. We will use the following notation for normalized eigenfunctions: Ψn,k e and Ψn,k h for the electron and hole levels given by Eq. (15), correspondingly, where k = 1, 2, 3... is the index of the 1D subband with angular momentum n , and R R 0 Ψn,k e 2ρedρe2π =R R 0 Ψn,k h 2ρhdρh2π = 1 . Using Eq. (15) we calculated the energy levels for a 4-nm PbSe NW with various growth directions. The energy band parameters of PbSe which we used in this calculation will be described in a later section. The effective energy gap of the NW, which is the energy distance between the top of the highest 1D sub-band of the valence band and the bottom of the lowest 1D sub-band of the conduction band, impacts many material properties. Figure 1 shows the effective energy gap for all four valleys as a function of the growth direction of the nanowire. Because the plot is calculated along high–symmetry directions in the Brillouin zone, the degeneracy of the four valleys is never completely lifted. Without any intervalley coupling, each of these energy gaps would have separate optical absorption and emission peaks associated with it. Figures 2a and 2b show the dispersion of the several lowest 1D subbands of the conduction and valence bands in NWs grown along the h111i and h100i directions, respectively. NWs grown along h111i have one valley oriented parallel to the growth direction and the other 8 1.0 0.9 0.8 0.7 ] V e [ p a G y g r e n E 1 2 1 2 2 1 1 2 <111> <100> Rod Growth Direction <110> <111> FIG. 1. Energy gaps of a 4-nm diameter PbSe NW at each of the four valleys as a function of the growth direction of the NW (red lines). The numbers indicate the valley degeneracy of the energy gaps. Dashed grey lines are the same energy gaps after accounting for the self-Coulomb interaction, described later in the text. three valleys oriented at the equal angles θ = 71o from it. For the h100i NW, all four valleys are at the same angle θ = 55o from the growth direction. It is clear that both the band-edge energies and the effective masses of the 1D subbands depend strongly on the growth direction. IV. DIELECTRIC CONFINEMENT The optical properties of all semiconductor nanostructures are controlled by the strength of the Coulomb interaction between the electron–hole pair participating in the emission and absorption of photons28. Compared to the screened Coulomb interaction in a bulk crystal, the interaction is usually enhanced because the electric field of the electron and hole localized inside the nanostructure penetrates into the surrounding medium, which commonly has a dielectric constant smaller than that of the semiconductor. In addition, any charge in the vicinity of this interface polarizes it. In the case of a flat interface, for example, this polarization can be described easily using an image charge that interacts with the primary charge29. In the case of small external dielectric constant the interaction is repulsive. This repulsive potential in nanostructures of any shape leads to an additional confinement of carriers, which is referred to as dielectric confinement. 9 1 0 -1 ] V e [ y g r e n E (a) = 70.5o (x3) = 0.0o (x1) 1 0 -1 ] V e [ y g r e n E -2 -1 0 kz [nm-1] 1 2 -2 -1 1 e 5/2 1 e 3/2 2 e 1/2 (b) 1 e 1/2 1 e 3/2 3/2 1 h 1/2 1 e 1/2 1 h 1/2 1 h = 55o (x4) 1 h 5/2 1 h 3/2 2 h 1/2 0 kz [nm-1] 1 2 FIG. 2. 1D band structure of a 4-nm PbSe NW for the cases of the axis along the directions (a) h111i and (b) h100i . The bands are labeled by the angle θ between the considered valley and the rod growth axis and also by their multiple valley degeneracy, up to a maximum of (x4). In (b), the individual subbands are labeled using notation adopted from molecular physics: kX e,h n for the kth electron or hole level of certain symmetry with total z angular momentum n , where X = Σ , Π , ∆ ,. . . , is used for m = 0 , 1, 2,. . . , respectively, where m is the angular momentum projection of the conduction (valence) band component of the wavefunction of the electron, ' e ', (hole, ' h ') state. In (a), the order of the levels is the same, and the labeling is suppressed for clarity. To model these effects in NRs and NWs, the analytic potential for two charges in an infinite dielectric cylinder U(re, rh) 30 is used. It was shown previously7 that this approx- imation works well as long as the rod length is larger than the size of the exciton. The potential naturally divides into four terms31: the unscreened direct interaction of the two charges Ud , the modification of this interaction due to the image effects of the solvent Us , and the two self-interactions of each charge with its own image Ue and Uh : U(re, rh) = −e2/(κsre − rh) − eVs(re, rh) + eVs(re, re)/2 + eVs(rh, rh)/2 where the function Vs has the form Ue(re) ≡ Ud(re − rh) + Us(re, rh) + 2π2κsZ ∞ cos(u(ze − zh)) cos(m(φe − φh))(2 − δm0) × (κs − κm)Im(uρe)Im(uρh)Km(Ru) (Km−1(Ru) + Km+1(Ru)) ∞Xm=0 du + Uh(rh) e 0 Vs(re,rh) = × κsKm(Ru) (Im−1(Ru) + Im+1(Ru)) + κmIm(Ru) (Km−1(Ru) + Km+1(Ru)) and where κs and κm are the optical dielectric constants of the bulk semiconductor and the surrounding medium, respectively. Im and Km are the modified Bessel functions of the 10 (16) (17) first and second kind. For PbSe we will use κs = 23 , and for the medium, if not explicitly stated otherwise, κm = 2 throughout this work. The self-interaction terms Ue(re) and Uh(rh) always contribute to the energy of each electron and hole subband calculated in Section III. In narrow NWs and NRs, where the self– interaction energy is smaller than the confined energies, this contribution can be calculated perturbatively for electron and hole levels, respectively: En,k self,e =Z ρedρedφeΨn,k e 2Ue(re) , En′,k′ self,h =Z ρhdρhdφhΨn′,k′ h 2Uh(rh) . (18) The self-interaction terms En,k self,h increase the energy of all electron and hole 1D subbands and consequently the effective energy gap in nanowires. The perturbed electron self,e and En′,k′ and hole subbands with n = n′ = 1/2 and k = k′ = 1 are shown in Fig. 1. In addition, in narrow NWs and NRs one can used an adiabatic approximation of the Coulomb interaction32,33, which replaces the three-dimensional potential of electrons and holes of Eq. (16) by a one-dimensional Coulomb potential that describes their interaction along the NW/NR axis. The adiabatic potential is obtained by averaging the potential over wave functions Ψn,k e and Ψn′,k′ h of the corresponding electron and hole subband. Averaging the first two terms of Eq. (16) results in the 1D adiabatic potential V n′k′ n,k (ze − zh) =Z ρedρedφeZ ρhdρhdφhΨn,k e 2Ψn′,k′ h 2(Ud(re − rh) + Us(re, rh)) , (19) which describes the interaction of electrons and holes occupying different subbands. This adiabatic potential is a function of the electron and hole separation, ze− zh , only. One can show that at large distances ze − zh ≫ R it takes the form of a one-dimensional Coulomb potential with the dielectric constant of the surrounding medium, V n′k′ n,k ∼ −e2/(κmze−zh) . The adiabatic potential for the ground electron and hole subbands with n = n′ = 1/2 and k = k′ = 1 is shown in Fig.3. V. 1D EXCITONS IN PBSE NANOWIRES AND NANORODS The attractive 1D potential described by Eq. (19) creates a series of one-dimensional exciton states for each pair of electron and hole subbands (n, k) and (n′, k′) . The effective masses of electrons and holes along the NW axis mn,k e and mn′,k′ e at the bottom and the top of each subband, correspondingly, is determined by Eq. (15). This allows us to write a 11 0.0 ] V e [ y g r e n E -0.1 -0.2 V1/2,1 1/2,1 Ueff -50 -25 0 (ze - zh)/R 25 50 FIG. 3. Points show the effective binding potential, V 1/2,1 1/2,1 , between an electron and a hole occupying the ground one dimensional subband n = n′ = 1/2 and k = k′ = 1 as a function of their separation, calculated for a 4-nm radius PbSe NW. The solid line shows the approximation of this dependence by the Elliott & Loudon effective potential described by Eq. (21) one-dimensional Schrodinger equation for these 1D excitons: 2 2µn′k′ n,k ∂2 ∂z2 Ψ1D − − 2 2M n′k′ n,k ∂2 ∂Z 2 Ψ1D + U n′k′ n,k (z)Ψ1D = εn′k′ n,k Ψ1D , (20) where we introduce the electron-hole separation, z = ze − zh and the exciton center-of-mass coordinate Z = (mn,k ) is the ) . µn′k′ /(mn,k zh)/(mn,k n,k = mn,k e mn′,k′ e + mn′,k′ h e + mn′,k′ h h e ze + mn′,k′ n,k = mn,k h reduced mass and M n′k′ e +mn′,k′ tantly, the exciton binding energy εn′k′ n,k h is the total effective mass of the 1D exciton. Impor- in this equation is calculated relative to the distance between the bottom of the (n, k) conduction subband and the top of the (n′, k′) valence subband, assuming the self–interaction energy terms En,k self,h are already taken into account. The solution of Eq. (20) can be separated into Ψ1D(z, Z) = ψ1D(z)Ψcm(Z) . self,e and En′,k′ The wave function ψ1D(z) describes relative electron-hole motion and gives the spectrum of 1D excitons. The second component, Ψcm(Z) , describes the exciton center of mass motion, and in the case of an infinite NW Ψcm(Z) ∼ exp(iKZ) , where K is the exciton momen- tum along the NW axis. This replaces the second term in Eq. (20) by the exciton kinetic energy, 2K 2/2M n′k′ n,k . Equation (20) allows us to numerically calculate the energy spectrum of 1D excitons created from any pair of electron and hole subbands. In this paper, we will be interested primarily in the spectrum that arises from the lowest electron and hole subbands 1Σe 1/2 and 12 1Σh 1/2 , and we will use the approach suggested by Elliott & Loudon33 to describe the spec- trum of one-dimensional excitons in a strong magnetic field. They suggest approximation of the one-dimensional adiabatic potential by an effective one-dimensional potential, which has well-known Schrodinger equation solutions, Ueff(z) = − e2 κm(z + ρeff) − Aρeff e2 κm(z + ρeff)2 , (21) where ρeff and A are the two fitting parameters. The medium dielectric constant κm is used in Eq. (21) so that the correct asymptotic form of the potential is maintained. For a 4-nm PbSe NW immersed in a medium with κm = 2 , the numerically-calculated effective potential is described very well by the potential Ueff with ρeff = 5.49R and A = 2.73 , as seen in Fig. 3. The slight dependence of these fit parameters on NW size is shown in Fig. 4a and the much stronger dependence on κm is shown in Fig. 4b. (a) 5.55 5.50 5.45 5.40 R / f f e r e t e m a r a p t i F 5.35 0 eff /R A 1 2 3 4 5 Nanowire radius [nm] 2.75 2.74 2.73 2.72 2.71 2.70 F i t p a r a m e t e r A R / f f e r e t e m a r a p t i F 8 6 4 2 0 0 (b) eff / R 5 20 Medium dielectric constant m 10 15 F i t p a r a m e t e r A 3 2 1 0 25 FIG. 4. Fitting parameters used in the effective potential described by Eq. (21) in PbSe NWs of various radius and medium dielectric constant. The parameter is plotted vs. (a) nanowire diameter with κm = 2 (b) medium dielectric constant with R = 2 nm. The energy spectrum and eigenfunctions of Eq.(20) with effective attractive potential Ueff(z) can be obtained analytically. The eigenfunctions of each 1D exciton level, ψα(z) , can be written as32,33 ψα(z > 0) = a1Wα,− 1 ψα(z < 0) = ±ψα(z) 2√1−4Aαρ(z + ρ) + a2Mα,− 1 2√1−4Aαρ(z + ρ) (22) (23) where Wα,β(x) and Mα,β(x) are the Whittaker functions, z = 2z/(a0α) , ρ = 2ρeff/(a0α) , a0 = 2κm/(µ1/2,1 1/2,1e2) is the effective Bohr radius of a 1D exciton, and a1 and a2 are 13 arbitrary coefficients. The sign of Eq. (23) is " + " for an even eigenfunction and "− " for an odd one. The coefficients a1 , a2 , and parameter α in Eq. (22) as well as the exciton binding energy: εα = − are determined by the boundary conditions. 2 2µ1/2,1 1/2,1a2 0α2 (24) There are two boundary conditions to impose on the solution in Eq. (22): one at z = ze − zh = ±L and one at z = 0 . We first consider infinite nanowires; the effects of finite length will be treated in the following section. In this case, the first boundary condition is 2√1−4Aαρ(z+ ρ) diverges as z → ∞ . The second satisfied by letting a2 = 0 , because Mα,− 1 boundary condition, requiring ψα(z) to be either an even or odd function of z , determines α and the energy spectrum of the exciton. It was shown in Refs. [32 and 33] that for excited doubly-degenerate exciton states, α takes almost-exactly integer values α = 1, 2, 3, ... and that α → 0 for ground states with decreasing exciton transverse radius. Following Refs. [32 and 33] we use ε0 for the ground exciton binding energy. 0.4 0.2 0.0 -0.2 -0.4 ] V e [ t f i h S y g r e n E (a) E1/2,1 self 0 Total 1 2 3 Radius [nm] 4 5 0.4 0.2 0.0 -0.2 -0.4 ] V e [ t f i h S y g r e n E 0 (b) E1/2,1 self 0 Total 10 5 20 Medium dielectric constant m 15 25 FIG. 5. Coulomb energies calculated for (a) κm = 2 with varying R and (b) R = 2 nm with varying κm . Lines are the sum of the electron E1/2,1 self,h self interaction energies self,e and hole E1/2,1 (red circles); the electron-hole binding energy ε0 (blue triangles); and their total (black squares). Figure 5 shows the calculated binding energy of the ground exciton state ε0 and the self,e and E1/2,1 Coulomb self–interaction energies E1/2,1 1D subbands 1Σe,h self,h of electrons and holes from the ground 1/2 . The binding energy decreases dramatically with NW radius or external dielectric constant. The exciton binding energy in the narrowest NW surrounded with κm ∼ 2 − 3 reaches values > 300 meV. 14 Surprisingly, however, the binding energy is almost exactly compensated by the electron and hole self–interaction terms, which leads to practical cancelation of most effects connected with the small dielectric constant of the surrounding medium. Because of this cancelation, the optical transitions between 1D subbands will be determined primarily by the energies calculated in Section III. This result has important practical consequences. For example, the linear optical spectra of PbSe NWs will not be sensitive to the dielectric constant of the surrounding medium. This cancelation is well–known in spherical semiconductor NCs. The exact cancelation of these three terms was shown for parabolic valence and conduction bands in Ref. [34]. This is because in a parabolic-band approximation the wave function of electrons and holes are identical and depend only on the NC radius. As a result the electron and hole charge distributions exactly compensate each other at each point in the NC. If there is no local charge in the NC, there is no electric field outside of the NC, and the external medium does not affect the optical properties. This cancelation is nearly exact even when the electron and hole masses are different35. The cancelation of the Coulomb energies in the ground exciton of PbSE NWs can be attributed to a similar charge compensation. The mirror symmetry of the conduction and valence bands in PbSe makes the wave functions of the electron and hole transverse motion nearly identical. The similar values of effective masses along the NW axes also makes the electron and hole contributions to the 1D exciton wave function identical. It is interesting to note here that because of the large binding energy, the electron and hole in the exciton are remarkably tightly bound, with average separation only slightly larger than the NW radius. Fig. 6 shows the average separation, calculated as ph(z − ¯z)2i , as a function of radius, with inset showing the wavefunction ψ1D for the case of R = 2 nm. One can see that the average electron-hole separation in the exciton is an order of magnitude smaller than the 46 nm Bohr radius in bulk PbSe. Further calculations show that this unusual increase in the strength of the binding is due entirely to the 1D shape of the NR, and is only weakly affected by the dielectric contrast. For the weakest dielectric contrast when κm = κs = 23 , the average separation increases slightly to ≈ 4 nm, still much closer to the 4-nm diameter than to the Bohr radius. 15 2 D 1 ] m n [ n o i t a r a p e s . s . m . r 5 4 3 2 1 0 10 5 -10 0 ze - zh [nm] 0 1 3 2 Radius [nm] 4 FIG. 6. Dependence on PbSe NW radius of the average (r.m.s.) separation of the electron and hole in the 1D exciton. Inset shows the square of the ground exciton wavefunction ψ1D2 for a NW with 2 nm radius. A. Finite length effects For a nanorod, which has finite length, the relative and center-of-mass (CM) motions of the electron and hole can never be completely separated. If the NR is much longer than the radius of the 1D exciton, one can still approximately separate variables to create effective boundary conditions for the exciton CM motion. No other boundary condition (BC) is needed for the exciton separation coordinate, because the assumption is that the tightly–bound wavefunction is already zero well before any additional confinement is felt. On the other hand, the CM motion can be considered as the motion of a free particle confined in a 1D box of length L . If the box is much larger than the exciton radius one can apply the standard boundary conditions on Ψcm to obtain the well-known spectrum Ecm(l) = 2π2l2/(2M 1/2,1 1/2,1 L2) , where l is the level number. Even though this CM boundary condition makes intuitive sense, it is difficult to justify, because the true BCs are for the electron and hole individually. To test our assumption, we calculated the CM energies numerically by solving the two-particle Schrodinger equation with the correct impenetrable boundary conditions on the electron and hole individually. Details of the calculation are in Appendix C. The numerically calculated wavefunctions and energies were best matched to those obtained for a free particle with an effective mass of the exciton which is confined in the 1D box of length Lcm = L − R . The existence 16 of such a simple expression is probably connected with the approximately-equal effective masses of the electrons and holes and their small separation in PbSe NRs. The first few numerically-calculated energy levels are shown in Fig. 7, along with the analytic energies Ecm = 2l2π2/(2M 1/2,1 cm) for various confinement lengths Lcm . This modified CM length 1/2,1 L2 works well for all rod sizes studied, as long as the NR aspect ratio is & 2 . ] V e [ y g r e n E 1.00 0.95 0.90 0.85 0.80 0.75 1 2 Numerical Lcm = L Lcm = L-R Lcm = L-2R 5 4 3 Level number 6 7 8 FIG. 7. Numerically-calculated energies for the lowest few exciton states in a 4 x 20 nm PbSe NR (black circles). The lines are the energies from the analytic model using two different confinement lengths for the center of mass. B. Oscillator strength of the interband optical transitions The decrease of the electron-hole separation within a 1D exciton leads to a dramatic increase of the optical transition strength. It was shown by Elliott & Loudon33 that the oscillator strength of practically the entire spectrum of 1D excitons becomes concentrated in the ground exciton state. The expression for the transition strength in PbSe NRs can be obtained by combining the results derived for PbSe NCs24 and CdSe NRs7. The total oscillator strength Ototal can be written as a product Ototal = O⊥Ok , where the tranverse oscillator strength is24 O⊥ = 2P 2 l 9m0ω(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) Z R 0 ρdρZ 2π 0 dφ"Ψ1/2,1 h #†" 0 σz σz 0#"Ψ1/2,1 e 2 #(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) 17 (25) with ω the total energy of the optical transition. We have neglected the second term from Ref. [24], as it is negligible except for very small NRs, where the envelope function approximation likely breaks down anyway. The oscillator strength of the 1D exciton7 is dZΨcm(Z)(cid:12)(cid:12)(cid:12)(cid:12) where we normalize the 1D exciton wave function such that R L −L dzR L Ok = ψ1D(z = 0)2(cid:12)(cid:12)(cid:12)(cid:12) Z L 1 . 2 0 (26) 0 dZψ1D(z)Ψcm(Z)2 = The transverse oscillator strength provides the selection rule that there is no change in the z-component of the angular momentum, ∆n = 0 , while the longitudinal component focuses the oscillator strength into the ground exciton state. This is because optical transitions are only allowed to the even states of the exciton CM motion with l = 1, 3, 5... , and the oscillator strength decreases as 1/l2 . Even the second allowed transition will be 9 times weaker than the lowest transition. This has practical implications for the optical absorption spectra. Even though the density of allowed transitions increases dramatically with energy in NRs, most of the oscillator strength is concentrated in the lowest-energy transition for each pair of NR subbands. Thus, isolated peaks should still be observable in experimental spectra. VI. EXPERIMENT A. Synthesis and characterization of colloidal PbSe nanorods Although the synthesis of lead salt nanowires was reported several years ago36,37, the fabrication of high quality lead-salt nanorods with small diameter has proved challenging. PbSe NRs were synthesized with noble metals as seeds38, but the resulting NRs did not have good optical spectra. Some high-quality NRs have been reported, but the syntheses were too challenging for us to reproduce39–41. A simple synthesis for high-optical-quality PbSe NRs was recently demonstrated42, and the properties of these NRs will be compared to the theoretical results. Following Ref. [42], the NR synthesis was carried out using standard Schlenk-line tech- niques under dry nitrogen. Tris(diethylamino)phosphine (TDP, Aldrich, 97%), oleic acid (OA, Aldrich, 90%), 1-octadecene (ODE, Aldrich, 90%), squalane (Aldrich, 99%), amor- 18 phous selenium shots (Se, Aldrich, 99.999%), and lead(II) oxide (PbO, Aldrich, 99.9%) were used as purchased without further purification. Anhydrous ethanol, chloroform, acetone, hexane, and tetrachloroethylene (TCE) were purchased from various sources. To prepare 1.0 M stock solutions of TDPSe, 7.86 g of Se was dissolved in 100 mL of TDP. Typically, 0.22 g of PbO was dissolved in 5 mL of squalane in the presence of 1 mL OA. (Squalane can be replaced by ODE.) After drying under nitrogen at 150 C for 30 min, the solution was heated to 170 C and 3 mL of a 1 M TDPSe solution in TDP was injected under vigorous stirring. Once the reaction finished, the reaction mixture was cooled to room temperature using a water bath. The crude solution was mixed with hexane and precipitated by ethanol. The precipitated NRs were isolated by centrifugation (at 5000 rpm for 3 min) and redispersed in chloroform or other organic solvents. Size-selective precipitation can be carried out to obtain better monodispersity of NRs samples using chloroform/acetone or other solvent/nonsolvent pairs. The size of the synthesized NRs was determined from transmission electron microscopy. In-plane powder X-ray diffraction shows that the NRs grow along the h100i direction42. Absorption was measured on a Shimadzu UV-3101PC spectrophotometer at room temper- ature. Emission spectra were recorded at room temperature with an infrared fluorimeter equipped with a 200-mm focal length monochromator, a single mode fiber coupled laser source (S1FC635PM, 635 nm, Thorlabs, Inc) as the excitation source, and an InGaAs photodiode (New Focus Femtowatt model 2153). Fluorescence lifetime was measured us- ing an InP/InGaAs PMT (Hamamatsu H10330A-75) with 120-fs excitation pulses from a Ti:sapphire regenerative amplifier (Spectra-Physics Hurricane) with 1 kHz repetition rate. NRs were dissolved in tetrachloroethylene (TCE) for all measurements to avoid spurious ab- sorbance in the near-IR. Quantum yield measurements were performed using an integrating sphere, with the method described in Ref. [43]. B. Absorption Spectra First, we will highlight the qualitative differences between the absorption spectra of NRs and spherical NCs. Figure 8 shows the absorption spectrum of 3.3 nm diameter x 12 nm length PbSe NRs along with that of 4.4 nm diameter spherical NCs, chosen to have a nearly identical first absorption peak. The spectrum of the NRs has fewer obvious features than 19 the NC spectrum. The first peak in the NR spectrum has a broad high energy side, even though its narrower low energy side is nearly identical to that of the NCs (inset of Fig. 8). Both of these observations indicate the presence of more densely-spaced transitions in the NR spectrum, which have the effect of smoothing out the peaks. Interestingly, the second NC peak appears where there is a dip in the NR spectrum. 0.75 0.50 0.25 e c n a b r o s b A (a) 1100 1300 1500 NR NC 0.00 500 750 1000 1250 Wavelength [nm] 1500 0.0 1200 1.0 (b) ] s t i n u . b r a [ n o i s s i m E 0.5 0 2 4 6 8 10 Time [ s] NR NC 1500 1800 Wavelength [nm] 2100 FIG. 8. (a) Absorption spectra of PbSe NRs (black line, vertically offset for clarity) and spherical PbSe NCs (red line) are compared. The inset shows detail of the first peak. (b) Emission spectra and fluorescence decays measured at the emission peak (inset) of the same two samples. The broadening of the NR absorption peak seen in Fig. 8 is connected with the dispersion of NR diameter and length. Our best PbSe NR samples have around 5% size distribution in radius, but a much larger 20% in length. This large length polydispersity will blur out many of the NR transitions in an ensemble, except for those that are roughly independent of length- specifically, the lowest energy exciton for each pair of NW subbands. Fortunately, this is also the transition predicted to have the largest oscillator strength. As we have shown above, the energies of the optical transitions of the ground exciton states practically coincide with the energies between non-interacting electron and hole subbands, even though their respective wave functions differ greatly. This greatly simplifies the interpretation of the absorption spectra of NRs. We performed second–derivative analysis on the absorption spectra to determine the transition energies accurately. To avoid the problems inherent in this method44, only the peaks in the second-derivative spectra that correspond to obviously-visible peaks in the measured spectra were used. NRs produced by our first syntheses showed instability in 20 solution and would slightly aggregate during the absorption measurement. This adds a moderate scattering background, so only the absorption peak location is recorded for these samples. NRs synthesized more recently are more stable, and at least four peaks can be discerned, with an additional peak in the three samples with narrowest size distribution. Fig. 9a has an example measured spectrum of a 3.9 nm diameter x 17 nm length PbSe NR that shows all five peaks, and the locations of all measurable peaks from all samples are shown in Fig. 9b. The measured peaks are plotted vs. D−3/2 following the similar graph in Ref. [45]. This power of the diameter is chosen to make the trend linear over the measured range, allowing rough extrapolation to bulk as D−3/2 → 0 . In this manner, the peaks originating from the L-point and Σ -point are easily distinguished. 1.5 1.0 0.5 0.0 e c n a b r o s b A P5 (a) 500 700 900 P4 2 0 300 P3 P2 500 1000 Wavelength [nm] P1 1500 (b) 3.5 3.0 2.5 2.0 1.5 1.0 0.5 ] V e [ y g r e n E 0.0 0.00 2 1 1 1 1 1 0.20 0.05 Diameter-1.5 [nm-1.5] 0.10 0.15 FIG. 9. (a) Example absorption spectra of a 3.9 x 17 nm PbSe NR. Inset shows the same data, but on a scale where the 5th peak is visible. (b) Peaks in 2nd derivative spectrum as a function of NR diameter (symbols), calculated allowed transitions (grey lines), simple parabolic effective mass calculation around the Σ -point (dashed grey line), and linear fits (colored dashed lines.) Quantitative theoretical description of the size-dependent absorption spectra of PbSe NRs shown in Fig. 9 requires a set of 6 room temperature energy band parameters for this semiconductor: m±t , m±l t,l . The parameters extracted from low temperature cyclotron resonance and interband magnetooptical experiments in bulk PbSe46 describe quite , and P 2 well the average two-dimensional effective mass of electrons and holes at the bottom of the conduction band and the top of the valence band, respectively. The fitting procedure that gives this set is not sensitive, however, to the separation of 1/ml,t and the 2P 2 l,t/m2 0Eg terms, and describes well only the sum of these terms, because the all measurements are 21 conducted a the narrow energy range comparable with the PbSe energy gap. This procedure is also not very sensitive to the anisotropy of the carrier energy spectra, because a magnetic field averages out the 2D motion of electrons and holes. On the other hand, in order to predict nanocrystal energy levels quantitatively, both the separation of components of the effective masses and the band anisotropy are crucial. Finally, the energy band parameters are expected to be temperature dependent. Thus, we conclude that parameters inferred from cyclotron resonance and magneto-optical measurements might not describe the energy spectra of NRs and NCs measured at room temperature. In principle, spatial confinement of carriers in nanostructures provides a more-sensitive way to determine the energy band parameters, due to the large modification of the energy spectra of confined carriers. With this motivation, we used the previously-measured ab- sorption spectra of PbSe NCs in Refs. [45, 47–51] and extracted room-temperature band parameters using a global fitting procedure. Importantly, this new set of parameters not only quantitatively describes the low-energy transitions of PbSe NCs, but may also help resolve the long-standing controversy over the symmetry of the second peak in the NC ab- sorption spectra (see Appendix D). These band parameters (Table VI B) are used in all graphs presented in this work. Name Ref. [46] Best Fit Aniso. ratio Ref. [46] Best Fit m+ t /m0 m−t /m0 2P 2 0.29 0.27 0.59 0.79 t /m0 3.6 (eV) 4.25 (eV) t m+ l /m+ m−l /m−t t /P 2 P 2 l 1.28 3.53 1.82 1.6 1.6 3.0 TABLE I. Energy band parameters that provide the best fits to the room temperature data from PbSe NCs. The left columns show the transverse band components, while the right columns show the ratio of transverse to longitudinal components. The theoretical size dependence of the optical transitions in PbSe NRs is calculated within our 4 band model and shown in Fig. 9b by solid lines. The lowest two transitions agree well with the theory. The third predicted transition is not observed, possibly owing to its proximity to other strong transitions in our NR samples. The third and fourth peaks are strong transitions that do not appear to be associated with the L-point. Their energies extrapolate back to the Σ -point energy. The third peak is fit well by the same parabolic 22 band model used to model spherical PbSe NCs, and thus we assign this transition to the lowest-energy excitonic state at the Σ point. This line was calculated for both spheres and rods with me Σ = 0.45m0 and Eg(Σ) = 1.65 eV. Without more-detailed knowledge of the band structure there, we cannot predict the excited states with any accuracy. Thus, the Σ = mh identity of the fourth transition cannot be determined, but as the energies approach the same 1.65-eV bulk value, it is reasonable to tentatively attribute it to a higher-energy exciton from the Σ point. Finally, the fifth peak was perhaps the strongest in the absorption spectra, but showed no size dependence. We tentatively ascribe this to a metal-complex transition on the surface of the nanocrystal based on its proximity to absorption peaks of Pb(II) complexes52. The identities of these transitions are summarized in Table VI B. Label Assigned Transitions P1 P2 P3 P4 P5 1Σh 3/2 → 1Πe 1/2 → 1Σe 1/2 3/2 and 1Πh 1Πh 1/2 → 1Πe 1/2 Σ–point ground state Σ–point excited state (?) Surface metal complex mode TABLE II. Transitions observed in the absorption spectra of PbSe NRs. The fluoresence spectra and decays (Fig. 8b) are nearly identical for NCs and NRs, with a slightly larger Stokes shift in the NRs along with a slightly broader peak. The ensemble quantum yield of the nanorods is around 15%, around half that of the NCs. This might indicate that the radiate lifetime of the rods is longer than the that of the NCs, but it is also possible that the QY reflects an ensemble with 15% emitting and 85% non-emitting rods. Two effects would be expected to modify the radiative lifetime in nanorods. First, be- cause the radiative lifetime is inversely proportional to the oscillator strength, the increased electron–hole correlation in NRs should decrease the lifetime compared to NCs. Second, the effect of screening is reduced in NRs, which is believed to be the cause of the long lifetime in PbSe NCs49. Approximating the NR as a dielectric prolate spheroid, the screening will substantially decrease along the rod axis, while slightly increasing along the other two axes, with an overall effect of a reduction in screening of the lifetime. Compared to a spherical NC of the same diameter, the larger oscillator strength and the reduced screening should 23 each produce about a factor of 3 reduction in lifetime in NRs with typical aspect ratios. To- gether this amounts to almost an order of magnitude reduction, and should be measurable even considering other sample–related uncertainties. However, the measured lifetime (Fig. 8b) is nearly identical in NCs and NRs. This discrepancy is not understood. It might be explained by a dark ground exciton state that controls the photoluminescence decay in PbSe NRs and NCs, with the same activation mechanism in both structures. To be thorough, the nonradiative rate must be determined, and completing this along with exploring this phenomenon is a topic of future work. VII. DISCUSSION AND CONCLUSIONS Our model of the electronic structure of lead–salt NRs is based on the 4 band k · p [27], using the standard boundary condition of a vanishing Hamiltonian suggested in Ref. envelope wave function at the NR surface. All calculations are conducted within a cylindrical approximation. To use this model for description of various properties of NRs or NWs, one needs to know a set of the 6 temperature-dependent band parameters that describe a specific bulk lead–salt semiconductor. For the PbSe NRs studied in this paper, we extracted the set of room-temperature parameters from analysis of the size-dependence of previously- measured room temperature absorption spectra of spherical PbSe NCs. The most significant conclusion of this work is that the fundamental excitations in PbSe NRs are one-dimensional excitons under each pair of optically coupled electron–hole sub- bands. The binding energy of the ground exciton state, which accumulates the most oscilla- tor strength, increases with decreasing NR thickness and reaches 400 meV in the narrowest rods. Surprisingly, the large binding energy of the exciton is almost exactly compensated by the self–interaction of electrons and holes with their own images, which makes the energies of the optical transitions nearly independent of the solvent dielectric constant. Although the finite length of NRs affects the spacing between excited exciton states, it has a negligible effect on the energy of the exciton ground states. With the set of PbSe band parameters extracted from spherical NC absorption spectra (Table VI B), the model presented here describes the absorption spectra of PbSe NRs, and potentially resolves some troublesome aspects of k · p theory of spherical PbSe NCs. The energy of the optical transitions to the exciton ground states calculated within a cylindrical 24 approximation match the two lowest-energy transitions observed experimentally. Although the effect of anisotropy in important for description of the absorption in spherical PbSe NCs, it is diminished in NRs (see Appendix A & D), and the energy of the first two transitions is unaffected by it. The absorption spectra of PbSe NRs have another remarkable feature. The size depen- dence of the third and fourth absorption peaks is strong evidence that they originate from the Σ point of the Brillouin zone. Similar states connected with the Σ point were ob- served previously in the absorption spectra45 and in the hot carrier dynamics53 of spherical PbSe NCs. These observations provide clear experimental evidence that even in the smallest nanostructures, wave functions from distinct critical points ( L and Σ , in this particular case) are not mixed if both their corresponding conduction band minima and valence band maxima are well-separated energetically. This provides strong justification for the applica- bility of our multiband effective mass approximation in such small nanostructures. A large energetic separation of L and Σ band edges is supported theoretically by recent ab ini- tio calculations54, which for PbSe predict larger than 500 meV energy separation for these extrema, in both the valence and conduction bands, in contradiction with similar earlier calculations, which placed the separation in the valence band closer to 150 meV55. The predicted strong increase in electron–hole Coulomb interaction in PbSe NWs should have major implications for other properties. This enhancement should increase the rate of the nonradiative Auger recombination as well as the rate of the inverse process, impact ionization. A high rate of impact ionization or efficient multiple exciton generation, combined with good conductivity that might be expected in PbSe NWs, suggests that these structures may be promising for photovoltaic applications. To summarize, we have developed a theory that describes both the energy spectra of individual electrons and holes and the absorption spectra of lead–salt NWs and NRs. Cal- culations show that even though spatial and dielectric confinement dramatically increase the exciton binding energy, the absorption spectra of PbSe NWs and NRs are practically unaffected, which should lead to insensitivity of these spectra to the surrounding media. The size dependence of lowest absorption peaks measured in PbSe NRs is very well described by the developed theory. It should be straightforward to apply this model to PbS and PbTe NRs. 25 Appendix A: Effect of anisotropy on the nanowire energy spectra The cylindrically symmetric Hamiltonian in Eq. (4) can be derived from the full Hamil- tonian in Eq. (3) by transformation to the new coordinate system connected with NW direction. The full Hamiltonian is defined with respect to a crystallographic direction of the Brillouin zone, where the z–axis is pointed towards one of the L–points, and we will call this coordinate system the primed system, {x′, y′, z′} . We need to express Eq. (3) in the new coordinate system where the z–axis is directed along the rod axis, called the unprimed system, {x, y, z} . To do this, we use a coordinate rotation, and define the x–axis such that the rotation occurs in the x–z plane. In the rotation, vector quantities, such as p or σ are transformed using the rotation matrix, p′ = R(θ)p , with R defined as R(θ) =  cos θ 0 − sin θ 0 0 1 sin θ 0 cos θ   . (A1) This transformation expresses the squared momenta in Eq. (3) as: p′2 x = cos2 θ p2 z = sin2 θ p2 p′2 x − sin 2θ px pz + sin2 θ p2 x + sin 2θ px pz + cos2 θ p2 z . z (A2) (A3) and the diagonal and off-diagonal elements of the matrix of Hamiltonian in Eq. (3) in new coordinate system as: 1 mt (p′2 x + p′2 y ) + 1 ml p′2 1 mt p2 y + + mt x + sin2 θ z =(cid:18) cos2 θ +(cid:18)sin2 θ + sin 2θ(cid:18) 1 ml − ml (cid:19) p2 ml (cid:19) p2 mt(cid:19) px pz cos2 θ z + + mt 1 +(Pt sin2 θ + Pl cos2 θ)σz pz + + 1 2 sin 2θ(Pl − Pt)(σz px + σx pz) (A4) (A5) Ptσ′x p′x + Ptσ′y p′y + Plσ′z p′z = (Pt cos2 θ + Pl sin2 θ)σx px + Ptσy py + Notice that neither elements are cylindrically symmetric in the new coordinates. To en- force this symmetry, we rewrite these expressions in a form that separates a cylindrically symmetrical part, formally: a Ox + b Oy = (1/2)(a + b)( Ox + Oy) + (1/2)(a − b)( Ox − Oy) . 26 The first term, which has cylindrical symmetry, is used in the zero-th order Hamiltonian, and the second term creates the asymmetric perturbation. This procedure produces the Hamiltonian in Eq. (4), along with the perturbation matrix Han =   1 2 m− m− 2 sin2 θ(p2 U(cid:16) 1 x − p2 t (cid:17)× l − 1 ×(cid:0) 1 y) + sin 2θ px pz(cid:1) 2m (Pl − Pt)(cid:8) sin2 θ(σx px − σy py)+ + sin 2θ(σz px + σx pz)(cid:9) 1 1 2m(Pl − Pt)(cid:8) sin2 θ(σx px − σy py)+ + sin 2θ(σz px + σx pz)(cid:9) t (cid:17)× − 1 U(cid:16) 1 x − p2 l − 1 y) + sin 2θ px pz(cid:1) m+ m+ 2 2 sin2 θ(p2 ×(cid:0) 1   (A6) We study the effect of anisotropy described by Eq. (A6) on the energy spectrum of electrons and holes. Figure 10 compares the energy of the lowest electron levels in a 4 nm PbSe NW calculated within the cylindrical approximation and with complete numerical inclusion of the anisotropy. The anisotropy was taken into account by diagonalizing the matrix elements of Han in the space of the highest 20 valence and lowest 20 conduction states (that is, including the highest ten and lowest ten doubly degenerate electron and hole levels.) One can see in Fig. 10 that the anisotropy in PbSe splits the nearly degenerate energy levels, whose radial or angular quantum momentum numbers differ by one in radial or angular quantum momentum numbers, while necessarily leaving the Kramer's degeneracy unbroken. The splitting should broaden the energy levels without an overall shift in the level position. Appendix B: Calculations of the one dimensional Coulomb potential Calculation of the one dimensional Coulomb potential in Eq. (19) and self interaction energy in Eq. (18) can be greatly simplified by initial averaging over angular variables. For the U1 term of Eq. (19) the angular integration results hU1i(z) =Z R 0 dρeρeZ R 0 dρhρhΨe2Ψh2V1(ρe, ρh, z) , where V1(ρe, ρh, z) = −4π e2 κs√ρeρh Q−1/2(cid:18)z2 + ρ2 2ρeρh e + ρ2 h (cid:19) and Qn is the Legendre function of the second kind. The two remaining radial integrals are (B1) (B2) evaluated numerically. 27 1.2 1.0 0.8 0.6 0.4 0.2 ] V e [ y g r e n E (a) ] V e [ y g r e n E 1.2 1.0 0.8 0.6 0.4 0.2 (b) = 55o, approximate = 55o, full = 70o, approximate = 70o, full = 0o FIG. 10. Effect of the energy spectrum anisotrpy on the energy of the 1D subband bottom in a 4 nm PbSe NW grown along the (a) h111i and (b) h100i crystal axes. The "approximate" calculations are conducted within the cylindrical approximation, which gives Eq. (15) for the energy levels. The "full" calculations are performed as described in the text. The energy levels are labeled by the angle between the L-point and the rod growth axis. Note that the θ = 0 energy levels do not require perturbation, as Han = 0 for that angle. For the second term in Eq. (19), U2 , the angular integrals vanish unless m = 0 leaving e2 du 0 only this term from the sum. This results in the following expression for hU2i(z) : hU2i(z) = −8π ×(cid:18)Z R κsZ ∞ (cid:18)Z R dρe ρeΨe2I0(uρe)(cid:19) } To calculate the integrals ie and ih in Eq. (B3), we approximate the squared wavefunctions n=1 An(1− ρ2n e ) , with N ≈ 8 . Even with so few terms, the maximum relative error is typically < 10−7 . This allows us to solve these two integrals (κs − κm)K0(Ru)K1(Ru) cos(uz) κsI1(Ru)K0(Ru) + κmI0(Ru)K1(Ru) × dρh ρhΨh2I0(uρh)(cid:19) } {z as a short sum of the form Ψe2 =PN {z (B3) ih(u) ie(u) . 0 0 analytically: ie(u) = = e )I0(uρe) N N Xn=1 Xn=1 0 dρe ρe(1 − ρ2n R2+2n AnZ R An(cid:18)RI1(u) u − where pFq is the generalized hypergeometric function. The remaining integral over u in 1F2 (1 + n; 1, 2 + n; R2u2/4) 2 + 2n (cid:19) , (B4) Eq. (B3) is performed numerically. 28 Lastly, the two self interaction terms in Eq. (18), Ue and Uh , after angular integrations are reduced to hUe,hi = 2e2 κs ∞Xm=0Z ∞ 0 du(cid:18)Z R 0 dρe,h ρe,hΨe,h2I 2 m(uρe,h)(cid:19) × × (κs − κm)Km(Ru)(Km−1(Ru) + Km+1(Ru))(2 − δm0) κsKm(Ru)(Im−1(Ru) + Im+1(Ru)) + κmIm(Ru)(Km−1(Ru) + Km+1(Ru)) . (B5) The two dimensional integrals in Eq. (B5) was taken numerically. It is summed over only the first ≈ 20 values of m , as the sum converges rapidly. Appendix C: Numerical calculation of the exciton binding in PbSe nanorods Our analytic model makes the assumption that the 1D exciton is only weakly confined along the NR axis. In this case the finite length of the NR affects only the exciton center of mass motion. To verify this assumption, the 1D Hamiltonian was numerically diagonalized, while treating both binding and confinement exactly. As an orthogonal basis for this diago- nalization we used a sufficiently large set of electron and hole plane waves that satisfied the single particle boundary conditions. The 1D exciton wavefunction in this basis set can be written as: Ψ1D = Ne Xne=1 Nh Xnh=1 Ane,nh 2 L sin(cid:16) neπze L (cid:17) sin(cid:16) nhπzh L (cid:17) (C1) where Ane,nh are the numerical coefficients. The kinetic energy is diagonal in this basis, and matrix elements of Eq. (21) can be eval- uated analytically. Calculation time was dominated by evaluation of these matrix elements and scaled as O( NeNh ). For Ne = Nh ≈ 30 , calculations were sufficiently converged for the lowest few dozen states, and required roughly one minute of computation time on a desktop computer. Fig. 11 shows the square of 1D wavefunctions, Ψ1D2 , calculated both the numerically and analytically as a function of ze and zh . For the lowest two exciton states Ψ1D2 shows good agreement between the numerical model and the analytical calculation. This is because the electron and hole are strongly localized around each other and do not feel the effects of confinement at the edges of box. As a result, the wavefunction orients along the coordinates associated with Coulomb binding, z and zcom , roughly along the graph diagonals. On the other hand, by the 17th excited state, also shown in Fig. 11, the numerical and analytical 29 calculations disagree greatly. This is because the higher kinetic energy of this state causes the wavefunction to reach the edges of the box and feel confined. And, as a result, it begins to orient along the box coordinates, ze and zh , associated with confinement. In general, our analytic model shows good agreement for the lowest ≈ 10 states for each pair of nanowire bands. l a c i r e m u N h z n o i t i s o p e l o h c i t y l a n A h z n o i t i s o p e l o h L L/2 0 L L/2 0 0 State #1 State #2 State #17 L/2 L 0 L/2 L 0 L/2 L electron position ze electron position ze electron position ze FIG. 11. Comparison of the numerically and analytically calculated 1D exciton wavefunctions Ψ1D2 . Each subplot has axes ze and zh ranging along the length of the nanorod from 0 to L . The two lowest energy states and the 17th state are shown. Appendix D: Choice of the room temperature band parameters The absence of reliable room temperature energy band parameters for bulk PbSe has lead to several problems in the quantitative description of spherical PbSe NC electronic prop- erties within effective mass theory, and as a result, to some controversy on their electronic structure45,55–58. As has been noted56,59, effective mass theory significantly overestimates the energy gap in PbSe NCs (though not in PbS.) In addition, the nature of the 2nd optical tran- sition is still a source of debate60–63, whether it is of symmetry type S-P or P-P. Considering 30 the body of experimental evidence, the explanation put forward by Franceschetti62 seems to offer the simplest explanation of this controversy, that the electron and hole P states are split into P⊥ and Pk states by the anisotropy of the bands, and the second transition is of type Pk - Pk . These two problematic aspects of experimental spectra of PbSe NCs for effective mass theory– overestimation of the bandgap and the symmetry of the 2nd transition– as well as the observation of several optical transitions in a wide range of energies can be used for extraction of a real set of the energy band parameters. Although the extraction of the set of energy band parameters from room temperature absorption spectra is possible, it is likely that many sets of parameters will equally well fit the first few optical transitions. In order to increase the accuracy of the fit, we want to somehow incorporate the energy band parameters in low temperature experiments in bulk PbSe. So, the total band edge effective masses for electrons and holes at T = 4 K are held constant at the values from experiment46. In addition, to limit the degrees of freedom in equal. That is, m+ the fit, the anisotropy of the far-band contributions to both the electron and hole are held t = m−l /m−t , even though their individual values will differ. With these constraints, a fit is performed using the body of literature data45,47–51 for the first l /m+ transition, and the data from Koole45 for the second and third transitions. The final set of room temperature parameters are shown in Table VI B together with the set of low temperature parameters reported for bulk PbSe in Ref. [46]. The transition energies calculated using these parameters are shown in Fig. 12. The anisotropic effective mass calculations were performed using the method outlined in Ref. [64] and the results compared to the energies measured in Ref. [45], ignoring those points criticized in Ref. [44] as possibly being 2nd derivative artifacts. ACKNOWLEDGMENTS A. C. Bartnik thanks Jun Yang for assistance with numerical optimizations and for en- lightening discussions. This work was supported by the Cornell Center for Nanoscale Sys- tems (CNS) through National Science Foundation Grant EEC-0646547, and in part by the Cornell Center for Materials Research (CCMR) with funding from the Materials Research Science and Engineering Center program of the National Science Foundation (cooperative agreement DMR 0520404). Al. L. Efros thanks the U.S. Office of Naval Research (ONR) and 31 ] V e [ y g r e n E 0.8 0.7 0.6 0.5 0.4 (a) 1S+ 1/2 1P+ 1/2,1P+ 3/2 (b) 1D- 1/2 1S+ 1/2 1P+ 3/2= P 1P+ 1/2= P 3 2 1 ] V e [ y g r e n E 1F1/2 1P1/2 1D3/2 1D1/2 1S1/2 1P3/2 1P1/2 1S1/2 1S- 1/2 1S- 1/2 Isotropic Anisotropic 0 0.00 0.03 0.09 0.06 0.12 Diameter-1.5 [nm-1.5] 0.15 0.18 FIG. 12. Calculations of the lowest electron levels in spherical PbSe NCs. (a) Splitting of the P state induced by the fully anisotropic Hamiltonian in a 4 nm radius NC. Anisotropic states are labeled by writing the state in the basis of isotropic states, and labeling it by the isotropic state with largest coefficient. (b) The size dependence of the transition energies in spherical PbSe NCs. Experimental data45 are shown by symbols. The solid lines show the size dependence of optically allowed transitions calculated in a fully anisotropic effective mass model. The optically allowed transitions occur between the states of the same symmetry but opposite parity, and we label them by a symmetry type, which is common for both states. Open points indicate transitions originating from the L-point in the Brillouin zone, while half-open points are suggested to be from the Σ point as in Ref. [45]. The dashed line shows the size dependence of lowest confined level connected with the Σ point of the Brillouin zone, calculated in a parabolic effective mass approximation as explained in the text. the U.S. Department of Energy (DOE). W.-k. Koh and C. B. Murray acknowledge financial support from the NSF through DMS-0935165. This research was partially supported by the Nano/Bio Interface Center through the National Science Foundation (NSEC DMR08-32802) with a seed grant that initiated the investigation of the synthesis of the nanorods. 1 M. Kuno, Phys. Chem. Chem. Phys. 10, 620 (2008). 2 Y.-w. Jun, J.-s. Choi, and J. Cheon, Angew. Chem. Int. Edit. 45, 3414 (2006). 3 C. Burda, X. Chen, R. Narayanan, and M. A. El-Sayed, Chem. Rev. 105, 1025 (2005). 4 E. C. Scher, L. Manna, and A. P. Alivisatos, Philos. T. Roy. Soc. A 361, 241 (2003). 32 5 E. A. Muljarov, E. A. Zhukov, V. S. Dneprovskii, and Y. Masumoto, Phys. Rev. B 62, 7420 (2000). 6 E. A. Mulyarov and S. G. Tikhodeev, J. Exp. Theor. Phys. 84, 151 (1997). 7 A. Shabaev and Al. L. Efros, Nano Lett. 4, 1821 (2004). 8 U. Bockelmann and G. Bastard, Phys. Rev. B 45, 1688 (1992). 9 J. Sun, W. E. Buhro, L.-W. Wang, and J. Schrier, Nano Lett. 8, 2913 (2008). 10 M. Califano and A. Zunger, Phys. Rev. B 70, 165317 (2004). 11 C.-Y. Yeh, S. B. Zhang, and A. Zunger, Phys. Rev. B 50, 14405 (1994). 12 A. Franceschetti, L. W. Wang, G. Bester, and A. Zunger, Nano Lett. 6, 1069 (2006). 13 M. P. Persson and H. Q. Xu, Appl. Phys. Lett. 81, 1309 (2002). 14 M. P. Persson and H. Q. Xu, Nano Lett. 4, 2409 (2004). 15 M. P. Persson and H. Q. Xu, Phys. Rev. B 70, 161310 (2004). 16 Y. M. Niquet, A. Lherbier, N. H. Quang, M. V. Fernandez-Serra, X. Blase, and C. Delerue, Phys. Rev. B 73, 165319 (2006). 17 J.-C. Charlier, X. Blase, and S. Roche, Rev. Mod. Phys. 79, 677 (2007). 18 M. V. Fernandez-Serra, C. Adessi, and X. Blase, Phys. Rev. Letters 96, 166805 (2006). 19 U. Landman, R. N. Barnett, A. G. Scherbakov, and P. Avouris, Phys. Rev. Letters 85, 1958 (2000). 20 B. F. Habenicht, C. F. Craig, and O. V. Prezhdo, Phys. Rev. Letters 96, 187401 (2006). 21 M. Bruno, M. Palummo, A. Marini, R. Del Sole, and S. Ossicini, Phys. Rev. Letters 98, 036807 (2007). 22 L. Brus, Nano Lett. 10, 363 (2010). 23 F. W. Wise, Accounts Chem. Res. 33, 773 (2000). 24 I. Kang and F. W. Wise, J. Opt. Soc. Am. B 14, 1632 (1997). 25 V. I. Rupasov, Phys. Rev. B 80, 115306 (2009). 26 J. O. Dimmock and G. B. Wright, Phys. Rev. 135, A821 (1964). 27 D. L. Mitchell and R. F. Wallis, Phys. Rev. 151, 581 (1966). 28 Al. L. Efros and A. L. Efros, Sov. Phys. Semicond. 16, 772 (1982). 29 J. D. Jackson, Classical Electrodynamics Third Edition, 3rd ed. (Wiley, 1998) ISBN 047130932X. 30 W. R. Smythe, Static and Dynamic Electricity (McGraw-Hill, New York, 1950). 31 L. E. Brus, J. Chem. Phys. 80, 4403 (1984). 33 32 R. Loudon, Am. J. Phys. 27, 649 (1959). 33 R. J. Elliott and R. Loudon, J. Phys. Chem. Solids 15, 196 (1960). 34 S. Schmitt-Rink, D. A. B. Miller, and D. S. Chemla, Phys. Rev. B 35, 8113 (1987). 35 Al. L. Efros and A. V. Rodina, Solid State Commun. 72, 645 (1989). 36 E. Lifshitz, M. Bashouti, V. Kloper, A. Kigel, M. S. Eisen, and S. Berger, Nano Letters 3, 857 (2003). 37 K.-S. Cho, D. V. Talapin, W. Gaschler, and C. B. Murray, J. Am. Chem. Soc. 127, 7140 (2005). 38 K.-T. Yong, Y. Sahoo, K. R. Choudhury, M. T. Swihart, J. R. Minter, and P. N. Prasad, Nano Lett. 6, 709 (2006). 39 J. M. Luther, H. Zheng, B. Sadtler, and A. P. Alivisatos, J. Am. Chem. Soc. 131, 16851 (2009). 40 S. Acharya, U. K. Gautam, T. Sasaki, Y. Bando, Y. Golan, and K. Ariga, J. Am. Chem. Soc. 130, 4594 (2008). 41 H. W. Jamie and C. Huaqiang, Nanotechnology 19, 305605 (2008). 42 W.-k. Koh, A. C. Bartnik, F. W. Wise, and C. B. Murray, J. Am. Chem. Soc. 132, 3909 (2010). 43 J. C. de Mello, H. F. Wittmann, and R. H. Friend, Adv. Mater. 9, 230 (1997). 44 I. Moreels and Z. Hens, Small 4, 1866 (2008). 45 R. Koole, G. Allan, C. Delerue, A. Meijerink, D. Vanmaekelbergh, and A. Houtepen, Small 4, 127 (2008). 46 H. Pascher, G. Bauer, and R. Grisar, Phys. Rev. B 38, 3383 (1988). 47 C. B. Murray, S. Sun, W. Gaschler, H. Doyle, T. A. Betley, and C. R. Kagan, IBM J. Res. Dev. 45, 47 (2001). 48 I. Moreels, K. Lambert, D. De Muynck, F. Vanhaecke, D. Poelman, J. C. Martins, G. Allan, and Z. Hens, Chem. Mater. 19, 6101 (2007). 49 B. L. Wehrenberg, C. Wang, and P. Guyot-Sionnest, J. Phys. Chem. B 106, 10634 (2002). 50 J. Steckel, S. Coe-Sullivan, V. Bulovic, and M. Bawendi, Adv. Mater. 15, 1862 (2003). 51 W. W. Yu, J. C. Falkner, B. S. Shih, and V. L. Colvin, Chem. Mater. 16, 3318 (2004). 52 K. Polk, M. Nikl, E. Mihkov, and K. Nitsch, J. Lumin. 94-95, 397 (2001), ISSN 0022-2313. 53 B. Cho, W. K. Peters, R. J. Hill, T. L. Courtney, and D. M. Jonas, Nano Lett. 10, 2498 (2010). 54 A. Svane, N. E. Christensen, M. Cardona, A. N. Chantis, M. van Schilfgaarde, and T. Kotani, Phys. Rev. B 81, 245120 (2010). 55 J. M. An, A. Franceschetti, S. V. Dudiy, and A. Zunger, Nano Lett. 6, 2728 (2006). 34 56 G. Allan and C. Delerue, Mater. Sci. Eng. C 25, 687 (2005). 57 S. V. Kilina, C. F. Craig, D. S. Kilin, and O. V. Prezhdo, J. Phys. Chem. C 111, 4871 (2007). 58 A. Franceschetti, Phys. Rev. B 78, 075418 (2008). 59 A. Lipovskii, E. Kolobkova, V. Petrikov, I. Kang, A. Olkhovets, T. Krauss, M. Thomas, J. Silcox, F. Wise, Q. Shen, and S. Kycia, Applied Physics Letters 71, 3406 (1997). 60 H. Du, C. Chen, R. Krishnan, T. D. Krauss, J. M. Harbold, F. W. Wise, M. G. Thomas, and J. Silcox, Nano Lett. 2, 1321 (2002). 61 J. J. Peterson, L. Huang, C. Delerue, G. Allan, and T. D. Krauss, Nano Lett. 7, 3827 (2007). 62 A. Franceschetti, J. W. Luo, J. M. An, and A. Zunger, Phys. Rev. B 79, 241311 (2009). 63 M. T. Trinh, A. J. Houtepen, J. M. Schins, J. Piris, and L. D. A. Siebbeles, Nano Lett. 8, 2112 (2008). 64 G. E. Tudury, M. V. Marquezini, L. G. Ferreira, L. C. Barbosa, and C. L. Cesar, Phys. Rev. B 62, 7357 (2000). 35 0.8 0.6 0.4 0.2 h t g n e r t S r o t a l l i c s O 0.0 500 Calculated Broadened 700 900 1100 Wavelength [nm] b 0.8 0.6 0.4 0.2 h t g n e r t S r o t a l l i c s O 0.0 500 Calculated Broadened 700 900 1100 Wavelength [nm] 0.00 -0.05 -0.10 -0.15 -0.20 -0.25 ] V e [ y g r e n E Vavg Ueff -50 -25 0 (ze - zh)/R 25
1902.01405	1	1902	2019-02-04T19:00:01	Gauge phonon dominated resistivity in twisted bilayer graphene near magic angle	[ "cond-mat.mes-hall" ]	Recent experiments on twisted bilayer graphene (tBG) close to magic angle show that a small relative rotation in a van der Waals heterostructure greatly alters its electronic properties. We consider various scattering mechanisms and show that the carrier transport in tBG is dominated by a combination of charged impurities and acoustic gauge phonons. Charged impurities still dominate at low temperature and densities because of the inability of Dirac fermions to screen long-range Coulomb potentials at charge neutrality; however, the gauge phonons dominate for most of the experimental regime because although they couple to current, they do not induce charge and are therefore unscreened by the large density of states close to magic angle. We show that the resistivity has a strong monotonically decreasing carrier density dependence at low temperature due to charged impurity scattering, and weak density dependence at high temperature due to gauge phonons. Away from charge neutrality, the resistivity increases with temperature, while it does the opposite close to the Dirac point. A non-monotonic temperature dependence observed only at low temperature and carrier density is a signature of our theory that can be tested in experimentally available samples.	cond-mat.mes-hall	cond-mat	Gauge phonon dominated resistivity in twisted bilayer graphene near magic angle Indra Yudhistira,1, 2, ∗ Nilotpal Chakraborty,3, ∗ Girish Sharma,1, 2 Derek Y. H. Ho,1, 3 Evan Laksono,1 Oleg P. Sushkov,4 Giovanni Vignale,1, 3, 5 and Shaffique Adam1, 2, 3, † 1Centre for Advanced 2D Materials, National University of Singapore, 6 Science Drive 2, 117546, Singapore 2Department of Physics, National University of Singapore, 2 Science Drive 3, 117551, Singapore 3Yale-NUS College, 16 College Avenue West, 138527, Singapore 4School of Physics, The University of New South Wales, Sydney 2052, Australia 5Department of Physics and Astronomy, University of Missouri, Columbia, Missouri 65211, USA (Dated: February 6, 2019) Recent experiments on twisted bilayer graphene (tBG) close to magic angle show that a small relative rotation in a van der Waals heterostructure greatly alters its electronic properties. We consider various scattering mechanisms and show that the carrier transport in tBG is dominated by a combination of charged impurities and acoustic gauge phonons. Charged impurities still dominate at low temperature and densities because of the inability of Dirac fermions to screen long-range Coulomb potentials at charge neutrality; however, the gauge phonons dominate for most of the experimental regime because although they couple to current, they do not induce charge and are therefore unscreened by the large density of states close to magic angle. We show that the resistivity has a strong monotonically decreasing carrier density dependence at low temperature due to charged impurity scattering, and weak density dependence at high temperature due to gauge phonons. Away from charge neutrality, the resistivity increases with temperature, while it does the opposite close to the Dirac point. A non-monotonic temperature dependence observed only at low temperature and carrier density is a signature of our theory that can be tested in experimentally available samples. The remarkable observations of superconductivity and insulating behaviour near magic angle in twisted bilayer graphene (tBG) [1 -- 3] have underlined the importance of twist angle as an additional control knob in van der Waals heterostructures. The band structure of tBG can be significantly altered with just small variations in the twist angle [4]. Special cases emerge near the so called "magic angles" when the lowest energy bands become almost flat [5], providing a platform for exotic physics arising from strong correlations. These recent exper- iments have inspired a large body of theoretical work which aim to understand both the origin of this strongly correlated phase [6 -- 17] and phonon-driven superconduc- tivity in tBG [18 -- 20]. More recent experimental work has focused on electron transport [21, 22]. A striking char- acteristic of transport in tBG close to magic angle is its extremely high and T -linear resistivity at high tempera- tures [1, 3]. Understanding the physics of these transport features could provide the insights necessary to under- stand the observed strongly correlated phases. Carrier transport in monolayer and bilayer graphene has been studied extensively over the past decade, both in theory and in experiment [23]. Traditionally, charged impurities dominate the electronic carrier transport at low temperature for both graphene monolayers and bi- layers, and these also induce fluctuations in the carrier density close to the Dirac point. Acoustic phonons be- come relevant at intermediate temperatures (T (cid:38) 100 K) ∗ These two authors contributed equally to this work † [email protected] and give rise to a characteristic linear in T resistivity. At still higher temperatures (T (cid:38) 250 K) optical phonons take over as the dominant scattering mechanisms. While the linear-in-T resistivity in tBG has been previously at- tributed to phonons [24], the magnitude of the deforma- tion potential extracted from experiment and the relevant temperature scales do not match what we know from our extensive studies on graphene [23]. In this Letter, we present a complete theory for trans- port in tBG at low densities near magic angle tBG (θM ), identifying all the relevant scattering mechanisms responsible at various temperatures probed experimen- tally. We find that for tBG close to magic angle, there is a crossover from charged-impurity limited transport to phonon-limited transport. Usually, phonon-dominated resistivity is modeled by the deformation potential. How- ever, we show that the deformation potential contribu- tion becomes irrelevant due to screening, and instead the phonon contributions arise from a gauge-field term [25]. The dominance of these gauge phonons arise due to the immunity of these particular phonons to the enhanced screening from the flat bands in tBG. The enhanced screening reduces the importance of all other scattering mechanisms including that of charged impurities and deformation potential phonons. Our the- ory shows that away from charge neutrality, the resistiv- ity increases linearly with temperature with weak depen- dence on both carrier density and impurity density (and therefore shows little sample-to-sample variations). How- ever, close to the Dirac point at sufficiently low tempera- ture, a non-monotonic in temperature and strong density dependence reveals the role of charged impurities. As shown in Fig. 1, acoustic phonons in graphene can 2 Table I. Relevant scattering mechanisms and the dependence of resistivity on Fermi velocity ρ(v∗). Only the βA gauge phonons grow in importance as v∗ → 0 when θ → θM . Mechanism Unscreened Screening (q)−2 Net effect Charged imp. Phonons (DA) Phonons (βA) Viscous friction Puddles v∗−2 v∗−2 v∗−2 v∗−2 v∗−2 No screening v∗2 v∗2 v∗2 v∗2 constant constant 1/v∗2 constant constant compensated by the usual v∗−2 dependence of resistiv- ity on carrier velocity. The inability of Dirac fermions to screen the gauge phonons dramatically increases their importance for the transport close to magic angle. In particular, the crossover temperature for which gauge phonons dominate over charged impurities drops from Tcross ∼ 500 K for monolayer graphene to Tcross ∼ 5 K for tBG. It is remarkable that phonons which were tra- ditionally neglected all the way until room temperature, now become important at such low temperatures. The electron-phonon (e-ph) interaction for monolayer graphene within a single-valley Dirac model contains both scalar and vector potential components [25] and is given by Ve−ph = ΦI + σ · A, where φ = DA (uxx + uyy) −2uxy (cid:18) uxx − uyy A = βA (cid:19) (1) (2) where DA and βA are bare (unscreened) coupling con- stants. The electron-phonon vertex therefore has a two-dimensional matrix structure, and its diagonal (off- diagonal) part comes from the scalar (vector) potential. While the diagonal component constitutes a scalar de- formation potential, the off-diagonal component corre- sponds to distortions which do not induce any variation in the unit cell area, but rather induce bond length mod- ulations which are often represented by a synthetic gauge field [26]. The random phase approximation (RPA) screening of the e-ph vertex by Coulomb interactions can be calculated from the polarizability bubble shown in Fig. 1(b). After including higher-order bubbles, the screening of a general vertex g is given by the Dyson equation g → g + IVqΠg + IVqΠCVqΠg + ··· , where I is the identity matrix, Vq is the Fourier transform of the Coulomb interaction, ΠC is the polarizability bub- ble, while Πg is the polarizability bubble with the vertex g at the right end as shown in Fig. 1(b). The RPA series is summed giving the screened vertex as βAσx RPA−→ βAσx + βAVqΠg 1 − VqΠC 12. (3) The key result is that the off-diagonal component is un- affected by RPA screening i.e. the gauge phonons remain Figure 1. Scalar and gauge deformations of twisted bilayer graphene (a) and how they are screened within the random phase approximation (b). The uniform stretching of the moir´e Brillouin zone (top sketch) couples to both charge and cur- rent and is strongly screened by the electrons in the flat moir´e bands at low energy. The asymmetric and shear modes (bot- tom two sketches) are area preserving and therefore act like a gauge potential that couple to current but not charge. As a result, these remain unscreened even as the Fermi velocity vanishes. Wavy, solid and springy lines refer to Coulomb in- teraction Vq, electrons and phonons, respectively. ΠC is the RPA polarization bubble. be classified as either deformation potential modes DA that change the area of the Brillouin zone, or gauge field modes βA that do not. The DA contribution comes from the longitudinal acoustic phonons and causes charge separation that is susceptible to electronic screen- ing. However, we show below that the gauge field term, which comes from both the longitudinal and transverse phonons, is unaffected by screening because although they couple to the current, they do not couple to charge. These modes remain unscreened even as the Fermi ve- locity vanishes. Charged impurity scattering that has dominated the transport properties of 2D materials are screened by the enhanced density of states. As explained below, this weaker electron-impurity interaction is com- pensated by a lower velocity (see Table I), so in the end, the qualitative behavior of electron-impurity resistivity is the same as in monolayer graphene. Due to the van- ishing density of states of Dirac fermions as n → 0 and T → 0, charged impurity scattering still dominates the transport properties for sufficiently low carrier density and temperature. To qualitatively illustrate the role of screening in tBG, we consider in Table I the Thomas-Fermi screening model where the polarizability is given by the density of states ND. Within a Dirac model (justified below), as θ → θM , the linear bands become flat, v∗ → 0 and the static dielectric function diverges as (q) ∼ v∗−1. For most mechanisms, screening gives a v∗2 contribution, that is (a) (b) (c) 3 Figure 2. (Color online) (a) Electronic structure of twisted bilayer graphene. A low energy effective Dirac Hamiltonian is valid for energies up to the blue line (marked Dirac) which for θ = 1.3◦ is ∼ 25 percent of the van Hove singularity energy (marked VHS). (b) Gauge phonons dominate the transport at temperatures higher than Tcross, while charged impurities dominate at lower temperatures. This crossover occurs within the window probed in recent experiments. (c) Similarly, phonons dominate at high density, while charged impurities dominate at low carrier density. unscreened by the large density of states (these phonon modes could also be germane to the observed supercon- ductivity). We neglect the scalar component both be- cause it is screened by ΠC (similar to the DA phonons), and because Πg vanishes at charged neutrality. To test the validity of the Dirac model, in Fig. 2 we plot the lowest energy bands for the continuum model [4] at θ = 1.3◦. The linear regime breaks down due to the emergence of Fermi pockets close to the Γ point. To stay within the regime of validity of the Dirac model, we limit our considerations to carrier densities (cid:46) 8 × 1010 cm−2 (although we expect our results to hold qualita- tively for even higher densities). In Fig. 2 we show that the crossover from charged impurity limited scattering to gauge phonon limited scattering occurs well within the regime where the Dirac Hamiltonian is valid. The Boltzmann transport theory for charged impurity scattering is now well-established, and we refer the reader to Ref. [23] for details. The scattering time is given by 1 τe−imp(ε) niε 2πv∗2 = (cid:90) π 0 (cid:12)(cid:12)(cid:12)(cid:12) Vei(q) (q) (cid:12)(cid:12)(cid:12)(cid:12)2 dθ (1 − cos2 θ), (4) where ni is the impurity concentration and Vei(q) is the Coulomb impurity matrix element. The v∗ depen- dent prefactor outside the integral comes from the en- hanced ND, while the dielectric function (q) is also en- hanced by v∗−2. As a result 1/τe−imp scales roughly as v∗ and is suppressed near magic angle. Throughout this work we use the RPA dielectric function (q, T ) = 1 − Vq(q)ΠRPA(q, T ), where Vq(q) = 2πe2/(κq) is the Coulomb potential, κ is the background dielectric con- stant and ΠRPA(q, T ) is static RPA polarizability [27]. The resistivity (ρe−imp) is obtained from τe−imp by the usual energy average 1 ρe−imp = e2 dε ND(ε) v∗2 2 τe−imp(ε) −∂nF (ε − µ) ∂ε . (5) (cid:90) By contrast, the e-ph interaction in TBG is drasti- cally different from that of monolayer graphene due to the emergence of hybrid folded phonon branches [28]. These hybrid phonons depend very sensitively on twist angle. For the temperature range we consider, the lowest acoustic phonon branch (which is the same as monolayer graphene) dominates. The gauge phonon contribution to the resistivity is [25, 29] (cid:18) TBG (cid:19) , (6) ρe−ph βA 16 β2 AkF = e2µsvA (v(cid:63))2 F the Bloch-Gruneisen T is is kF the where TBG is (kBTBG = 2vAkF ), µs graphene, is F (x) =(cid:82) 1 temperature the mass density of the Fermi wave-vector and vA Here 0 dy[xy4(cid:112)1 − y2exy]/ (exy − 1)2. The effective effective acoustic phonon velocity. coupling constant βA is proportional to strain, and therefore βA ≈ βA(v∗/vF )/[2 tan(θ/2)] and βA = 3.6eV is the value obtained for monolayer graphene from DFPT and tight-binding calculations [18, 25]. The DA deformation potential phonons have a similar form to Eq. 6. However, since they are heavily screened, this contribution becomes irrelevant in the temperature and density regimes we considered. Figure 3 shows our results for the transport proper- ties of tBG including both charged impurities and gauge phonons. Gauge phonons dominate for T > Tcross and as θ → θM . The dependence of the crossover temper- ature and carrier density as a function of twist angle is shown in Fig. 2 and we observe that the phonon- dominated regime becomes more prominent as one ap- proaches the magic angle. For example, we find that the crossover temperature Tcross from impurity-dominated to phonon-dominated resistivity is 5K at n = 2× 1010 cm−2 and twist angle θ = 1.2o. We identify three distinct transport regimes: (i) For T > Tcross, the resistivity is dominated by gauge phonons and is linear in temper- ature as one would expect from acoustic phonons; (ii) -30-20-10010203001234510010110210310-1100101102012345100102(a) (b) 4 Figure 3. (Color online) (a) Total resistivity as a function of twist angle (above magic angle), showing a crossover from phonon to impurity-dominated carrier transport. The black solid, red dashed and blue dashed lines are total, phonon and impurity limited resistivity, respectively. Gauge phonons are not screened by the large density of states associated with the flat bands, and dominate close to magic angle. (b) Total resistivity as a function of temperature at fixed density and twist angle. The crossover temperature in monolayer graphene is ∼ 500K, while near magic angle, the crossover temperature reduces to 5 K due to the strong screening of charged impurities by the flat bands, and the immunity of gauge phonons towards screening. The non-monotonicity arises from charged impurity scattering component crossing over from the degenerate (kBT < εF ) to non-degenerate regime kBT > εF . We use an effective dielectric constant of κeff = 3.5 appropriate for TBLG on h-BN. (a) (b) Figure 4. (Color online) Full effective medium theory for carrier transport in twisted bilayer graphene as a function of tempera- ture (a) and density (b) considering both charged impurity and gauge phonon scattering. At high temperature or high density, the gauge phonons dominate and there is a linear in T resistivity with negligible density dependence. At low temperature and close to charge neutrality, there is a non-monotonic temperature dependence arising from charged impurity scattering. There is an inversion in temperature dependence as we move away from charge neutrality to higher densities because of a crossover from charged impurity to gauge phonon scattering. Inset: Charge density fluctuations vs twist angle. While charged impurities are strongly screened by the large density of states, the electron-hole puddles are weakly affected by twist angle. For intermediate temperatures k−1 B εF < T < Tcross, the resistivity decreases with increasing temperature. This is understood as the non-degenerate Fermi liquid scat- tering off charged impurities, where both the effects of screening and energy averaging reduce the resistivity as ρe−imp ∝ (εF /kBT )2; (iii) For T < k−1 B εF , Tcross, we have the degenerate Fermi liquid scattering theory of charged impurities where energy averaging has little ef- fect on the resistivity and screening increases resistivity as δρe−imp ∝ (kBT /εF )2. This non-monotonic bump in the total resistivity ρ(T ) for T (cid:46) Tcross, and linear-in-T for T > Tcross are the main predictions of this work. For a theory with homogeneous carrier density, the charged impurity limited resistivity diverges as n → 0. 01234500.0050.010.0150.020.0250510152000.010.020.030510152000.010.020.030.0401234500.0050.010.0150.020.02502401However, we know that these same charged impurities that dominate transport give rise to carrier density inho- mogeneities that cure this divergence [30]. In Fig. 4 we show that the magnitude of the density inhomogeneity is weakly affected by the enhanced screening close to magic angle. We use the well-established effective medium the- ory [31] to average over these density inhomogeneities close to the Dirac point. Our results for the resistiv- ity as a function of temperature and carrier density for θ = 1.2◦ are shown in Fig. 4. There is a strong carrier density dependence at low temperature (dominated by charged impurities) and a weak density dependence at higher temperature (dominated by gauge phonons). Our theory also predicts a temperature and density regime where resistivity decreases with increasing temperature causing an inversion in the temperature dependence at low and high density e.g. in Fig. 4b, close to charge neu- trality, resistivity decreases with increasing temperature, while at 4×1010cm−2 resistivity increases with increasing temperature. (The curve for T = 0.5K is strongly influ- enced by the carrier density inhomogeniety and doesn't follow this general trend). Finally, in our analysis we considered other possible relevant mechanisms for transport in tBG and found them to be negligible. For example, Umklapp scatter- ing [32] is irrelevant for the densities we consider because only a small area of the moir´e Brillouin zone is occu- pied and the hydrodynamic electron-hole scattering [33] contribution to resistivity remains weaker than impurity scattering for T < 100 K. We conclude that for the experimentally relevant temperatures and densities the resistivity is determined only by the interplay between charged impurities and gauge phonons. Acknowledgement: We acknowledge the Singapore Ministry of Education AcRF Tier 2 grant MOE2017- T2-2-140, the National University of Singapore Young Investigator Award (R-607-000-094-133) and use of the dedicated research computing resources at CA2DM. [1] Y. Cao, V. Fatemi, S. Fang, K. Watanabe, T. Taniguchi, and P. Jarillo-Herrero, Nature 556, 43 E. Kaxiras, (2018). [2] Y. Cao, V. Fatemi, A. Demir, S. Fang, S. L. Tomarken, J. Y. Luo, J. D. Sanchez-Yamagishi, K. Watanabe, T. Taniguchi, E. Kaxiras, et al., Nature 556, 80 (2018). [3] M. Yankowitz, S. Chen, H. Polshyn, Y. Zhang, K. Watan- abe, T. Taniguchi, D. Graf, A. F. Young, and C. R. Dean, Science , eaav1910 (2019). [4] E. Laksono, J. N. Leaw, A. Reaves, M. Singh, X. Wang, S. Adam, and X. Gu, Solid State Communications 282, 38 (2018). [5] R. Bistritzer and A. H. MacDonald, Proceedings of the National Academy of Sciences 108, 12233 (2011). 5 [6] C. Xu and L. Balents, Phys. Rev. Lett. 121, 087001 (2018). [7] B. Roy and V. Juricic, arXiv preprint arXiv:1803.11190 (2018). [8] H. C. Po, L. Zou, A. Vishwanath, and T. Senthil, Phys. Rev. X 8, 031089 (2018). [9] M. Koshino, N. F. Q. Yuan, T. Koretsune, M. Ochi, K. Kuroki, and L. Fu, Phys. Rev. X 8, 031087 (2018). [10] J. Kang and O. Vafek, Phys. Rev. X 8, 031088 (2018). [11] B. Padhi, C. Setty, and P. W. Phillips, Nano letters 18, 6175 (2018). [12] H. Guo, X. Zhu, S. Feng, and R. T. Scalettar, Phys. Rev. B 97, 235453 (2018). [13] C.-C. Liu, L.-D. Zhang, W.-Q. Chen, and F. Yang, Phys. Rev. Lett. 121, 217001 (2018). [14] H. Isobe, N. F. Q. Yuan, and L. Fu, Phys. Rev. X 8, 041041 (2018). [15] Y.-Z. You and A. Vishwanath, arXiv preprint arXiv:1805.06867 (2018). [16] J. Gonz´alez and T. Stauber, Phys. Rev. Lett. 122, 026801 (2019). [17] M. Xie and A. H. MacDonald, arXiv preprint arXiv:1812.04213 (2018). [18] B. Lian, Z. Wang, and B. A. Bernevig, (2018), arXiv:1807.04382. [19] F. Wu, A. H. MacDonald, and I. Martin, Phys. Rev. Lett. 121, 257001 (2018). [20] Y. W. Choi and H. J. Choi, Phys. Rev. B 98, 241412 (2018). [21] Y. Cao, D. Chowdhury, D. Rodan-Legrain, O. Rubies- Bigord`a, K. Watanabe, T. Taniguchi, T. Senthil, and P. Jarillo-Herrero, arXiv preprint arXiv:1901.03710 (2019). [22] A. L. Sharpe, E. J. Fox, A. W. Barnard, J. Finney, K. Watanabe, T. Taniguchi, M. Kastner, and D. Goldhaber-Gordon, arXiv preprint arXiv:1901.03520 (2019). [23] S. Das Sarma, S. Adam, E. Hwang, and E. Rossi, Re- views of Modern Physics 83, 407 (2011). [24] F. Wu, E. Hwang, and S. Das Sarma, arXiv preprint arXiv:1811.04920 (2018). [25] T. Sohier, M. Calandra, C.-H. Park, N. Bonini, and F. Mauri, Phys. Rev. B 90, 125414 N. Marzari, (2014). [26] F. von Oppen, F. Guinea, and E. Mariani, Phys. Rev. B 80, 075420 (2009). [27] E. H. Hwang and S. Das Sarma, Phys. Rev. B 75, 205418 (2007). [28] A. I. Cocemasov, D. L. Nika, and A. A. Balandin, Phys. Rev. B 88, 035428 (2013). [29] E. H. Hwang and S. Das Sarma, Phys. Rev. B 77, 115449 (2008). [30] S. Adam, E. H. Hwang, V. M. Galitski, and S. Das Sarma, Proceedings of the National Academy of Sciences 104, 18392 (2007). [31] E. Rossi, S. Adam, and S. Das Sarma, Phys. Rev. B 79, 245423 (2009). [32] J. Wallbank, R. K. Kumar, M. Holwill, Z. Wang, G. Auton, J. Birkbeck, A. Mishchenko, L. Ponomarenko, K. Watanabe, T. Taniguchi, et al., Nature Physics 15, 32 (2019). [33] D. Y. H. Ho, I. Yudhistira, N. Chakraborty, and S. Adam, Phys. Rev. B 97, 121404 (2018).
1906.10581	1	1906	2019-06-25T15:00:26	Tuning the electronic structure and magnetic coupling in armchair B$_2$S nanoribbons using strain and staggered sublattice potential	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	Monolayer B$_2$S has been recently unveiled as a desirable honeycomb monolayer with an anisotropic Dirac cone. We investigate the Ruderman-Kittel-Kasuya-Yoshida (RKKY) interaction, between two magnetic impurity moments in armchair-terminated B$_2$S nanoribbons in the presence of strain and staggered sublattice potential. By using an accurate tight-binding model, we firstly study the electronic properties of all infinite-length armchair B$_2$S nanoribbons (ABSNRs), with different edges, in the presence of both strain and staggered potential. The ABSNRs show different electronic and magnetic behaviors due to different edge morphologies. The band gap energy of ABSNRs depends strongly upon the applied staggered potential and thus one can engineer the electronic properties of the ABSNRs via tuning the external staggered potential. A complete and fully reversible semiconductor (or insulator) to metal transition has been observed via tuning the external staggered potential, which can be easily realized experimentally. A prominent feature is the presence of a quasiflat edge mode, isolated from the bulk modes in the ABSNRs belong to the family $M=6p$, with $M$ the width of the ABSNR and $p$ an integer number. The position of the quasi-flatbands(QFBs) in the energy diagram of ABSNRs can be shifted by applying the in-plane strains. At a critical staggered potential, for nanoribbons of arbitrary width, the QFB changes to a perfect flatband. The RKKY interaction has an oscillating behaviour in terms of the applied staggered potentials and width of the ribbon, such that for two magnetic adatoms randomly distributed on the surface of an ABSNR the staggered potential can reverse the RKKY from antiferromagnetism to ferromagnetism and vice versa. Our findings pave the way for applications in spintronics and pseudospin electronics devices based on ABSNRs.	cond-mat.mes-hall	cond-mat	Tuning the electronic structure and magnetic coupling in armchair B2S nanoribbons using strain and staggered sublattice potential Moslem Zare1, ∗ 1Department of Physics, Yasouj University, Yasouj, Iran 75914-353, Iran (Dated: June 26, 2019) Magnetically-doped two dimensional honeycomb lattices are potential candidates for application in future spintronic devices. Monolayer B2S has been recently unveiled as a desirable honeycomb monolayer with an anisotropic Dirac cone. Here, we investigate the carrier-mediated exchange coupling, known as Ruderman-Kittel-Kasuya-Yoshida (RKKY) interaction, between two magnetic impurity moments in armchair-terminated B2S nanoribbons in the presence of strain and stag- gered sublattice potential. By using an accurate tight-binding model for the band structure of B2S nanoribbons near the main energy gap, we firstly study the electronic properties of all infinite- length armchair B2S nanoribbons (ABSNRs), with different edges, in the presence of both strain and staggered potential. It is found that the ABSNRs show different electronic and magnetic behaviors due to different edge morphologies. The band gap energy of ABSNRs depends strongly upon the applied staggered potential ∆ and thus one can engineer the electronic properties of the ABSNRs via tuning the external staggered potential. A complete and fully reversible semiconductor (or insulator) to metal transition has been observed via tuning the external staggered potential, which can be easily realized experimentally. A prominent feature is the presence of a quasiflat edge mode, isolated from the bulk modes in the ABSNRs belong to the family M = 6p, with M the width of the ABSNR and p an integer number. As a key feature, the position of the quasi-flatbands in the energy diagram of ABSNRs can be shifted by applying the in-plane strains εx and εy. At a critical staggered potential (∆c ∼ 0.5 eV), for nanoribbons of arbitrary width, the quasi-flatband changes to a perfect flatband. The RKKY interaction has an oscillating behaviour in terms of the applied staggered potentials, such that for two magnetic adatoms randomly distributed on the surface of an ABSNR the staggered potential can reverse the RKKY from antiferromagnetism to ferromagnetism and vice versa. The RKKY in terms of the width of the ribbon has also an oscillatory behavior. It is shown that the magnetic interactions between adsorbed magnetic impurities in ABSNRs can be manipulated by careful engineering of external staggered potential. Our findings pave the way for applications in spintronics and pseudospin electronics devices based on ABSNRs. I. INTRODUCTION The intriguing prospect of the potential nanoelectronic and optoelectronic applications, which may take advan- tage of the novel two dimensional (2D) materials with exciting electronic properties, has inspired researchers to explore possibilities of such materials with outstanding characteristics. As a typical two-dimensional material, pristine graphene is one of the most attractive materi- als due to its outstanding potential applications in many fields [1], but unfortunately, the lack of a finite band gap in graphene is a major obstacle for using graphene in nanoelectronic and optoelectronic devices. However, a big challenge for graphene science is how to open a sub- stantial band gap for graphene without significantly de- grading its excellent outstanding advantages in graphene based nanoelectronic devices [2, 3]. In the aspect of na- noelectronic and optoelectronic 2D research, the major issue is the availability of 2D materials with a wide band gap window in their band structure. In this regard, B2S monolayer, an atomically thin layer of boron and sul- fur atoms arranged in a honeycomb pattern with perfect ∗ [email protected] planar structure, appears in the research field again, by using global structure search method and first principles calculation combined with tight-binding model [4, 5]. It is reported that this new 2D anisotropic Dirac cone mate- rial has a Fermi-velocity up to 106m/s in the same order of magnitude as that of graphene. It is thermally and dy- namically stable at room temperature and is a potential candidate for future nanoelectronic applications [4, 5]. B2S monolayer is the first pristine honeycomb lattice with a tilted anisotropic Dirac cone structure, stabilized by sulfur atoms in free standing condition. Since boron atom has one electron less than carbon, all the reported 2D boron-based Dirac cone materials have much more complicated geometries in comparison with the pristine honeycomb structure of graphene [6 -- 10]. In the last decades, magnetic atoms embedded in a non-magnetic host material have been intensively stud- ied in solid state physics due to their functionalities for application in spintronic devices and magnetic recording media [11, 12]. Dilute magnetic semiconductors, as po- tential materials for spintronics and optoelectronics, have been studied since early 90-es. These investigations re- sulted in establishing a unified picture of the nature of indirect exchange interaction between magnetic adatoms, known as the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction [13 -- 15], mediated by a background of con- 9 1 0 2 n u J 5 2 ] l l a h - s e m . t a m - d n o c [ 1 v 1 8 5 0 1 . 6 0 9 1 : v i X r a duction electrons of the host material. In diluted met- als and semiconductors it is often the dominating mag- netic interaction and has played a key part in the de- velopment of magnetic phases, e.g., spiral [16, 17], spin- glass [11, 18, 19], ferromagnetic (FM) [20 -- 25] and anti- ferromagnetic (AFM) [26, 27]. This long-range spin-spin interaction leads to spin relevant effects in giant magne- toresistance devices [28, 29], spin filters [30], drives ferro- magnetism in heavy rare-earth elements [31] as well as in diluted magnetic semiconductors [12]. It was shown that the RKKY interaction consists of three terms, namely Heisenberg, Ising, and Dzyaloshinskii -- Moriya (DM) on the surface of zigzag silicene nanoribbons as well as the three dimensional topological insulators [11, 16, 32], and the competition between them leads to rich spin con- figurations. An additional term, namely spin-frustrated has discovered in a three-dimensional Weyl semimetal (WSM) that along with the Dzyaloshinskii-Moriya term lies in the plane perpendicular to the line connecting two Weyl points [33]. In a spin polarized system [34] and a material with multi-band structure [35], these oscillations become more complicated than a monotonic oscillation with sin(2kFR) behavior, where kF is the wave vector of the electrons (holes) at the Fermi level and R is the distance of two magnetic impurities. In addition, it is important to note that the magnitude of the RKKY interaction can be severely affected by the density of states (DOS) at the Fermi level [17, 36]. Owing to the bipartite nature of the honeycomb sublattice, the RKKY coupling in graphene is highly sensitive to the direction of the distance vec- tor between impurities [36, 37]. In materials with spin- orbit interaction of Rashba type [38], the exchange in- teraction depends on the direction of the magnetic mo- ments and, as a result, the RKKY interaction becomes anisotropic [17]. We have recently addressed the problem of isolated magnetic adatoms placed on silicene [39] and phospho- rene [40] sheets as well as on zigzag silicene nanorib- bons [16] and bilayer phosphorene nanoribbons [41]. In a detailed study, we found that the RKKY interaction in silicene can be written in an anisotropic Heisenberg form for the intrinsic case where the spin coupling could realize various spin models, e.g., the XXZ-like spin model [39]. In another work, it has concluded that the RKKY in- teraction in the bulk phosphorene monolayer is highly anisotropic and the magnetic ground-state of two mag- netic adatoms can be tuned by changing the spatial con- figuration of impurities as well as the chemical potential varying [40]. Importantly, the occurrence of these mag- netic phases not only depends on the magnetic impurity concentration, but also on the concentration of free car- riers in the host material [42]. This effective interaction can also be viewed as an indi- rect coupling mediated by pure spin current in quantum spin Hall systems, due to the helicity [43]. This interac- tion oscillates as a function of the distance between two magnetic adatoms (with wavelength π/kF ), due to the 2 sharpness of the Fermi surface. Besides these practical magnetic phases, the RKKY interaction can provide in- formation about the intrinsic properties of the material, since this coupling is proportional to the spin suscepti- bility of the host system. In the last decades, dilute magnetic semiconductors have emerged as a research hotspot due to their func- tionalities for application in spintronic devices and mag- netic recording media. Inducing magnetism in otherwise nonmagnetic 2D materials has been a subject of intense research due to the unique physical characteristics origi- nating from 2D confinement of electrons, foe concurrent applications in electronic and optoelectronic devices [44 -- 46]. Controllable magnetic properties of nanoribbon-based spintronic devices allow the development of the next gen- eration of magnetic and spintronic devices to be realized, and thus much attention has been focused on determin- ing the magnetic properties of 2D honeycomb nanorib- bons [16, 41, 47, 48]. Motivated by the future potential of the honeycomb nanoribbons decorated by magnetic impurities, in this work we have addressed the problem of indirect exchange coupling between localized magnetic moments mediated by the conduction electrons of B2S nanoribbons with armchair-terminated edges. Within the tight-binding model we exploit the Green's function formalism (GF), to reveal how the RKKY interaction between two mag- netic impurities, placed on a B2S nanoribbon, is affected by mechanical strains in the presence of a sublattice stag- gering potential. It is found that armchair B2S nanoribbons (ABSNRs) show different electronic and magnetic behaviors due to different edge morphologies. The band gap energy of ABSNRs depends strongly upon the applied staggered potential ∆, and thus one can engineer the electronic properties of the ABSNRs with desirable characteristics via tuning the external staggered potential. A complete and fully reversible semiconductor (or in- sulator) to metal transition has been observed via tun- ing the external staggered potential, which can be easily realized experimentally. Interestingly, for the ABSNRs belong to the family M = 6p, with M the width of the ABSNR and p an integer number, one can see that a band gap, in which a quasi-flatband completely detached from the bulk bands, is always observed. As a key fea- ture, the position of the quasi-flatbands in the energy diagram of ABSNRs can be shifted by applying the in- plane strains εx and εy. At a critical staggered potential (∆c ∼ 0.5 eV), for ABSNRs with any width, the quasi- flatband changes to a perfect flatband. It is shown that the RKKY interaction has an os- cillating behaviour in terms of the applied staggered potentials, such that for two magnetic adatoms ran- domly distributed on the surface of an ABSNR the stag- gered potential can reverse the RKKY from antiferro- magnetism to ferromagnetism and vice versa. Mean- while, the RKKY interaction has an oscillatory behavior 3 II. THEORY AND MODEL A. The RKKY interaction To study indirect magnetic interaction between two local moments in armchair-terminated B2S nanoribbons, we consider the indirect exchange coupling between mag- netic impurities to be of the RKKY form, mediated by the conduction electrons in armchair B2S nanoribbon system. The contact magnetic interaction between the spin of itinerant electrons and two magnetic impurities located at positions r and r(cid:48) with magnetic moments S1 and S2 is given by V = −λ (S1 · s(r) + S2 · s(r(cid:48))), (1) where s(r), s(r(cid:48)) are the conduction electron spin densi- ties at positions r and r(cid:48) and λ is the contact potential between the impurity spins and the itinerant carriers. Using a second-order perturbation [13 -- 15], the effec- tive magnetic interaction between local moments induced by the free carrier spin polarization, the RKKY interac- tion, which arises from quantum effects, reads as E(r, r(cid:48)) = J(r, r(cid:48))S1 · S2, (2) The RKKY interaction J(r, r(cid:48)) is explained using the static carrier susceptibility, the response of the charge density n(r) to a perturbing potential V (r(cid:48)), i.e., χ(r, r(cid:48)) ≡ δn(r)/δV (r(cid:48)) , which is given by J(r, r(cid:48)) = λ22 4 χ(r, r(cid:48)). (3) The static spin susceptibility can be written in terms of the integral over the unperturbed Green's function G0 as χ(r, r(cid:48)) = − 2 π dε Im[G0(r, r(cid:48), ε)G0(r(cid:48), r, ε)], (4) (cid:90) εF −∞ where εF is the Fermi energy. The expression for the susceptibility may be obtained by using the spectral rep- resentation of the Green's function G0(r, r(cid:48), ε) = ψn,s(r)ψ∗ n,s(r(cid:48)) ε + iη − εn,s , (5) (cid:88) n,s where ψn,s is the sublattice component of the unper- turbed eigenfunction with the corresponding energy εn,s. For a crystalline structure, n, s denotes the band index and spin. Substituting Eq. (4), after integration over energy, we will get the result for the RKKY interaction. The analytical background of this approach has been already described in details in previ- ous works [16, 41] and is not rediscussed here. Finally, from those analytical calculations the RKKY interaction can be expressed in the following desired result (5) into Eq. Figure 1. (Color online) Schematic illustration of the opti- mized geometry structure of armchair B2S nanoribbons (AB- SNRs). The top panels : armchair B2S nanoribbons of type 1 (a), Type 2 (b), and Type 3 (c). An example of an AB- SNR of type 1 with width M = 7 and length N = 12, in which the edges (or the 1D periodicity direction) lie along the x direction, is shown in the bottom panel (d). The red- dashed rectangle represents the super unit cell. The number of atoms in the vertical zigzag lines across the ABSNRs width, M , is used to indicate the width of a B2S nanoribbon and accordingly we use the number of vertical zigzag lines (N ) to measure its length (in the units of a) The black balls represent the boron atoms (B) and the white ones correspond to the sulfur atoms (S). The nearest-neighbor hopping parameters, used in the tight-binding Eq. 7, are denoted by t1 and t2. For simplicity, each atom is labeled with a set (n, m), where n, m represent the x and y coordinates of the lattice sites. in terms of the width of the ribbon. This paper is organized as follows: In Sec. II, we intro- duce the system under consideration, i.e., an armchair- terminated B2S nanoribbon under the influence of strain and staggered potential applied to it. A tight-binding model Hamiltonian for monolayer B2S is presented and then the band spectrum of ABSNRs with different edge configurations, under a staggered potential, have been in- vestigated then, we introduce the theoretical framework which will be used in calculating the RKKY interaction, from the real space Green's function. After that, we discuss our numerical results for the proposed magnetic doped ABSNRs in the presence of both strain and stag- gered sublattice potential. Finally, our conclusions are summarized in Sec. III. (a) Type 1 (p=1, M=7)(c) Type 3 (p=1, M=6)(b) Type 2 (p=1, M=7)Bm=1m=2m=3m=5m=4m=7m=6n=1n=2n=3n=4n=12xSUnit celly(d)a(cid:88) [ χ(r, r(cid:48))= 2 n,,s n(cid:48),s(cid:48) ×ψn,s(r)ψ∗ f (εn,s) − f (εn(cid:48),s(cid:48)) εn,s − εn(cid:48),s(cid:48) n,s(r(cid:48))ψn(cid:48),s(cid:48)(r(cid:48))ψ∗ n(cid:48)s(cid:48)(r)]. (6) where, f (ε), is the Fermi function. This is a well-known formula in the linear response theory that is the main equation in this work. B. Armchair B2S nanoribbons In this section, we reintroduce B2S monolayer to the layered-material family as an anisotropic material for op- toelectronic and spintronic applications. Since indirect exchange interaction between magnetic moments is sig- nificantly affected by the electronic structure of the host material, tailoring electronic properties of this nanostruc- ture is crucial. To do so, the electronic structures of armchair-terminated B2S nanoribbons are studied using tight-binding model. As very recently reported [4], the most energetically stable structure of B2S monolayer predicted by using global structure search method and first principles cal- culation combined with tight-binding model, is shown in Fig.1 from which we can see that the planar 2D structure consists of honeycomb lattices, similar to graphene. This honeycomb structure is a global minimum in the space of all possible 2D arrangements of B2S in which each hexag- onal ring is distorted with the bond angles ranging from 114 A to 123 A, because B and S atoms have different covalent radii and electronegativities [4, 5]. From the Figs.1 (a-c), it has been demonstrated that different types of ribbons are specified by their edge ge- ometry and width. As seen, the armchair B2S nanorib- bons can be divided into three families, i.e., M = 6p + 1 (type 1, with N boron atoms on the edges of the AB- SNR), M = 6p + 1 (type 2, with N/2 boron atoms on the edges of the ABSNR), M = 6p (type 3, with N boron atoms on the one edge and N/2 boron atoms on the an- other edge of the ABSNR), with p as an integer number. The bottom panel shows an ABSNR of type 1 in which the armchair edge is at the x direction. For an infinite- sized ABSNR this system shows translational symmetry along x axis. The red-dashed rectangle represents the super unit cell. The usage of such geometry division had two aims, first to evaluate the behaviour of the infinite length ABSNRs, and second to investigate some impor- tant finite size effects which will be discussed further. The geometrical structure of the pristine armchair edge B2S nanoribbon, lying in the xy plane, is depicted in Fig.1 (d). As shown in this figure, each hexagon consists of four B atoms and two S atoms, with an orthogonal primitive cell with a space group of P BAM and a point group D2h. As shown in Fig.1, the bonding length be- tween two adjacent B atoms (B-B bonds) was calculated 4 to be 1.62 A, from the relaxed structure whereas the dis- tances between B and S atoms (B-S bonds), are all of the same length 1.82 A [4]. The black balls correspond to the boron atoms (B) and the white ones correspond to the sulfur atom (S). For simplicity, each atom is la- beled with a set (n, m), where n, m represent the x and y coordinates of the lattice sites. The number of atoms in the vertical zigzag lines across the ABSNRs width, M , is used to indicate the width of a BSNR and accordingly we use the number of vertical zigzag lines (N ) to measure its length. From the analysis of symmetry and orbital characters of the wave functions in a B2S monolayer it is clear that a tight-binding (TB) model involving just the tilted pz orbitals should be able to describe the band structure of this 2D layered material near the Fermi level [4]. We be- gin with describing this nearest-neighbor effective tight- binding Hamiltonian, given by H = −(cid:88) (cid:104)i,j(cid:105) † ti,jc i cj + (cid:88) i † Ui,Bc i,Bci,B + (cid:88) i † Ui,Sc i,Sci,S, (7) with nearest neighbor hopping energies t1 = 0.8 eV and t2 = 1.7 eV [4] (see Fig. 1(d), the bottom panel) where c(†) is the annihilation (creation) operator of the electron at the i-th lattice site and (cid:80)(cid:104)i,j(cid:105) sums over all nearest i neighbor pairs. We consider a general situation where a staggered sublattice potential is applied throughout the sheet, ∆/2 for sublattices B, and −∆/2 for sublattices S. However, it has been found that the onsite energies for B and S atoms are VS = 5.4 eV and VS = 6.4 eV, respec- tively and thus, the corresponding parameters Ui,S(B), are as follows: Ui,B = VB + ∆/2 for sublattices B, and Ui,S = VS − ∆/2 for sublattices S, respectively. Having an accurate tight-binding model, as presented in the equation above (Eq.7), we can numerically cal- culate the momentum space dispersion of a monolayer armchair nanoribbon of B2S. To do so, we make a one- dimensional (1D) Fourier transform (owing to the trans- lational invariant along the ribbon direction, x), in accor- † kHkψk, dance with Bloch's theorem obtained from(cid:80) k ψ with respect to the x direction: Hk = H00 + H01e−ikxa + H † 01eikxa (8) in which a is the unit-cell length along the x-axis. More- over, H00 and H01 describe coupling within the principal unit cell (intra-unit cell) and between the adjacent prin- cipal unit cells (inter-unit cell), respectively based on the real space tight binding model given by Eq.7. And the real space Hamiltonian can now be written in the desired tridiagonal form:  0 0 0 HAA HAB H † AB HBB HBA 0 H 0 ... † BA HAA HAB † AB HBB 0 H  , ··· . . . HABSNR = where HAA(BB) and HAB(BA) are intra-unit cell and inter-unit cell (M × N )× (M × N ) matrices, respectively. Furthermore, to understand the effects of the im- purity position on the RKKY properties of ABSNRs, we have studied the local density of state (LDOS) of the ABSNRs. Corresponding site-resolved LDOS for site i−th, at a given position r and energy E, is ob- tained from the imaginary part of the Green's func- tion as ρ(r, E) = −G0(r, r, E)/π, calculated using the unperturbed Green's function matrix as G0(r, r, E) = (E − H + iη)−1, where η is a positive infinitesimal num- ber. C. Influence of strain and staggered sublattice potentials on the electronic properties of the ABSNRs The controlled introduction of strain into semiconduc- tors, a key strategy for manipulating the magnetic cou- pling in 2D nanostructures, has a perfect platform for its implementation in the atomically thin materials in both scientific and engineering applications [48]. Motivated by the search for spintronic materials, a huge number of works have been performed to examine the effectiveness of mechanical strain in modulating the magnetic proper- ties of 2D layered materials [48 -- 55]. To gain insight on how B2S nanoribbons can be fruitful in the realization of high-performing magnetic devices, fundamental stud- ies on the strain-induced variation of the electronic and magnetic properties of this new material are essential. In this subsection, the effect of both strain and staggered sublattice potential on the band structure and magnetic exchange interaction is analyzed and discussed. We first consider an armchair B2S nanoribbon lattice in the xy plane, in the presence of uniaxial strains x and y while a staggered sublattice potential is applied throughout the ABSNR. Let the x-axis be in the direction of the armchair edge of B2S nanoribbon and the y-axis in that of the lateral zigzag edge, as shown in Fig.1. Within the context of con- tinuum mechanics and in the linear deformation regime, application of a uniaxial strain will cause the following change of the bond length r, in terms of strain compo- nents x and y(cid:18) x(cid:48) y(cid:48) (cid:19) = (cid:18) 1 + x (cid:19)(cid:18) x (cid:19) γ γ 1 + y y , (9) where r = xi + yj and r(cid:48) = x(cid:48)i + y(cid:48)j denote the positions of an atom before and after deformation, respectively. 5 In the linear deformation regime, an expansion of the norm of r to first order in strains x and y can be written as r(cid:48) (cid:39) (1 + αxx + αyy)r, where αx = (x/r)2 and αy = (y/r)2 are the strain-related geometrical coefficients in the ABSNRs. According to the Harrison's formula the transfer integral (t) between s and p orbitals scales with the bond length (r) as t ∝ 1 r2 [56 -- 58]. By invoking the Harrison's relationship, we get the following geometrical strain effect on the hopping parameter, t = t0 1 − 2 r αxx − 2 r αyy . (10) (cid:18) (cid:19) One of the fascinating properties of the new families of 2D layered materials is their possibility to use a staggered potential to manipulate their electronic properties. Mo- tivated by this important problem, we examine the effect of staggered sublattice potential on the electronic struc- ture, by breaking the discrete sublattice symmetry of this honeycomb structure. Here, we investigate the band dis- persion of the ABSNRs of infinite length L (N → ∞) under the influence of the staggered potential. We first present the calculated electronic band struc- tures of ABSNRs superlattices. The energy band struc- tures of infinite length ABSNRs with width of M = 7, for different geometry types are plotted in Fig.2. The panels (a),(d) are for ABSNR of type 1, (b),(e) are for type 2, and (c),(f) are for type 3. Top panels are for zero staggered potential (∆ = 0) and the bottom ones are for nonzero staggered potentials (∆ = 3 eV). Interest- ingly, for the ABSNRs of type 2, one can see that a band gap in which a near-midgap band (red curve) completely detached from the bulk bands, is always observed and disappears by introducing staggered sublattice potential term. Indeed, this near-midgap band is shifted and goes away from the flatness mode by applying the staggered potential. As is known, these near-midgap energy bands are due to the edge states whose wave functions are con- fined near the ABSNR edges [59 -- 61]. What the Fig. 3 shows is the same as Fig. 2 but for AB- SNRs with p = 5. As shown, a large electronic band gap is appeared in the band structure of ABSNRs by apply- ing the staggered potential. As can be seen, the resulting band structures are completely different at various values of the strength of the staggered potential. The quasi flat- band in the ABSNR of type 2, in the absence of staggered potential (panel (b)), is shown with a red color. D. Quasiflat band tunability in the ABSNRs Designing the lattice structures which produce the flat band at Fermi energy has attracted much attention re- cently because of its potential applications in nanoelec- tronics, and magnetoectronics. The presence of flat bands at Fermi energy gives rise to the large density 6 Figure 2. (Color online) The energy band structure for several types of the ABSNRs with infinite lengths with p = 1 : (a),(d) type 1, (b),(e) type 2, and (c),(f) type 3. Top panels are for zero staggered potential (∆ = 0) and the bottom ones are for nonzero staggered potentials (∆ = 3 eV). The quasi-flat band is seen just in the ABSNR of type 2 (red curve). Figure 3. (Color online) The energy band structure for several types of the ABSNRs with infinite lengths with p = 5 : (a),(d) type 1, (b),(e) type 2, and (c),(f) type 3. Top panels are for zero staggered potential (∆ = 0) and the bottom ones are for nonzero staggered potentials (∆ = 3 eV). The quasi-flat band is seen in the ABSNR of type 2 with a red curve. of states and is responsible for the flat band ferromag- netism [62 -- 65]. There are primarily three ways toward creating flat bands in nanoribbons [62 -- 65]. The modi- fication of zigzag edge by attaching Klein's bonds gives rises to the partial flat band in Ggraphene nanoribbons. One of simple ways to obtain the flat bands is given by the nonequality between the sublattice sites in bipartite lattices. In such lattices, N-degenerated flat bands ap- pear at the Fermi energy. with N = NA − NB, where NA and NB are the number of A and B -sublattice sites, respectively [63, 66 -- 68]. As suggested by Soleimanikah- noj et al., the quasiflat band (midgap-band) modulation provides a platform for pseudospin electronics [59 -- 61], it is interesting to study the band gap and quasiflat band Figure 4. (Color online) Shift of the quasi-flat band for dif- ferent values of the strains εx and εy, for an ABSNR of type 2, with M = 31. modulation in the ABSNRs. The aim of this subsection is to elucidate the effect of the both strain and staggered potential on the spectral properties of the quasiflat edge modes in the ribbon ge- ometry of type 2, specially the formation and tunability of the quasiflat bands in the semiconducting gap in the APNRs. As shown in Fig.4, as a key feature, the position of the quasiflat bands in the energy diagram of APNRs can be shifted by applying the in-plane strains εx and εy. Particularly, the quasiflat band energy move up under strains εx, while shift down with strains εy. To achieve a superior performance of the ABSNRs in optoelectronic devices based on B2S nanoribbons, a feasi- ble band gap modification is crucial for nanoribbons. To reveal the staggered potential dependence of the band gap of ABSNRs, we have calculated the band gaps for all three types of ABSNR superlattices as a function of the applied staggered potential ∆ for different values of strain εx (see Fig.5). It is visible that the band gap energy of ABSNRs depends strongly upon the applied staggered potential ∆ and thus one can engineer the elec- tronic properties of the ABSNRs via tuning the external staggered potential. From the Fig.5 a complete and fully reversible semiconductor (or insulator) to metal transi- tion has been observed via tuning the external staggered potential, which can be easily realized experimentally. It should be emphasized that a negative energy gap cor- responds to a metallic state and a positive energy gap corresponds to a semiconductor or insulator electronic state, depending on the energy gap values. As we have demonstrated, depending on the ap- plied staggered potential in various strain configurations, one may have ABSNRs with favor electronic structure, namely, semiconductor, insulator or the metallic state. We elaborated more on the potential tunability below and show how to get highly improved flat bands with 0.00.51.068101214160.00.51.08101214161820220.00.51.08101214161820220.00.51.08101214161820220.00.51.068101214160.00.51.06810121416(c)(d)(e)(f)Energy [eV]kxa [p]kxa [p]kxa [p]Energy [eV](a)(b)0.00.51.01.52.068101214160.00.51.01.52.081012141618200.00.51.01.52.081012141618200.00.51.01.52.081012141618200.00.51.01.52.068101214160.00.51.01.52.06810121416(c)(d)(e)(f)Energy [eV]kxa [p]kxa [p]kxa [p]Energy [eV](a)(b)0.00.51.011.6511.6611.6711.6811.6911.7011.71kxa [p]Energy [eV] ey=0.05 ey=0.1 ey=0.15 ey=0.2 ex=0 ex=0.05 ex=0.1 ex=0.15 ex=0.27 Figure 5. (Color online) Variation of energy band gap of ABSNRs with p = 5 as a function of sublattice staggered potential for different values of strain εx. perfect flatness. Figure 6 presents the flat band bandwidth versus the applied staggered potential for ABSNRs with different widths. It is clear that, the response of the midgap bands to the applied staggered potential depends on the width of the ribbon. It is worthwhile to note that the band- width is generally defined as the energy difference be- tween the upper and lower band edges. It is important to note that at a critical staggered potential (∆c = 0.5 eV), for ABSNRs with any width, the quasi-flatband changes to a perfect flatband. Moreover, at a fixed staggered potential the midgap bandwidth (MBW) decreases with increasing the width of the ABSNR. Interestingly, the graph of bandwidth for ABSNRs with any width is sym- metric with respect to the critical staggered potential (at the interval [0-1] eV), as shown in the inset. For staggered potentials greater than the critical staggered potential (∆ > ∆c), the bandwidth of the quasi flat bands monotonically increases with increasing the stag- gered potential. For staggered potentials smaller than the critical staggered potential (∆ < ∆c), the trend in reverse ı.e., the bandwidth of the bands decreases with increasing the sublattice staggered potentials. As shown, the critical staggered potential (the potential at which semiconductor (or insulator)-to-metal transition occurs) is different for different types of the ABSNRs and changes with strain. E. Numerical results for the RKKY interaction in zigzag B2S nanoribbons In this section, we present in the following our main nu- merical results for the numerical evaluation of the RKKY exchange (Eq.6) in the armchair B2S nanoribbons, based on the tight-binding model (Eq.7). For simplicity, all ob- tained data for the RKKY interaction are multiplied by 103. Figure 7 shows the effective exchange interaction for doped ABSNRs (EF = 2 eV ) with N = 300 and p = 2 as Figure 6. (Color online) The effect of the staggered sublattice potential ∆ on the flatband bandwidth (FBBW) for ABSNRs with different widths. a function of distance between the impurities for possible impurity configurations for different types of ABSNRs. The details of the panels are as follows: (a) Both the impurities are located at the same edge, the first impurity at the edge site with coordinate (5, 1) and the second one at lattice sites (n, 1). (b) Both the impurities are situated in the interior region of the ABSNR, the first impurity at lattice site (5, 7) and the second impurity at lattice points (n, 7), (c) One impurity is at the edge site (5, 1) and the other one is moved interior of the ABSNR along the line n = 7 at the lattice sites (n, 7), and (d) The impurities are located at the opposite edge sites (interedge magnetic coupling) with coordinates (150, 1) and (150, 13) for the types 1 and 2 and (150, 1) and (150, 12) for the ABSNR of types 3. It is worth pointing out that for small distance between the impurities, the impurities interact very strongly with each other, but then rapidly decay with R until its flat- tens out to a constant value. As a result, a beating pat- 012345-0.8-0.6-0.4-0.20.00.20123450.00.20.40.60.80.9012345-1.50.01.53.04.55.3(b)Eg [eV]D[eV]D[eV] ex=0 ex=0.05 ex=0.1 ex=0.15 ex=0.2(c) ex=0 ex=0.05 ex=0.1 ex=0.15 ex=0.2D[eV] ex=0 ex=0.05 ex=0.1 ex=0.15 ex=0.2(a)0123450.00.20.40.60.81.00.00.51.00.000.020.04D[eV]Bandwidth [eV] M=7 M=13 M=19 M=25 M=31D[eV]BW [eV]tern of oscillations of the RKKY interaction occurs for all types of ABSNR, for the doped systems. It is clear that the edge structure of the ABSNRs has a strong effect on the RKKY coupling. The edge-geometry contribu- tions to the RKKY interaction were found to be more important for the geometry with both impurity spins are situated in the interior of the ABSNR (panel (b)), be- cause in this configuration the RKKY interaction is very strong for edge-geometry of type 1 in comparison to the other two geometries. For the case when both spins are inside the ABSNRs (the panel (b)) the result is quite dif- ferent, because in this situation the RKKY interaction is at least one order of magnitude greater than the other configurations. The staggered potential dependence of the RKKY in- teraction is shown in Fig.8, where different distance con- figurations between two impurities are considered. (a) Both the impurities are located on the same edge at the edge sites with coordinates (145, 1) and (155, 1) (b) Both the impurities are located in the interior of the ABSNR, on lattice points with coordinates (145, 7) and (155, 7), (c) One of the impurities is located on the edge site (145, 1) and the second one is on the lattice site (155, 7), and (d) The impurities are located on the opposite edge sites with coordinates (150, 1) and (150, 13). It is shown that the RKKY interaction has an oscil- lating behaviour in terms of the applied staggered po- tentials, such that for two magnetic adatoms randomly distributed on the surface of an ABSNR the staggered po- tential can reverse the RKKY from antiferromagnetism to ferromagnetism and vice versa. Importantly, the vari- ous edge geometries of ABSNRs show tunability in mag- netic RKKY coupling on the application of external stag- gered potentials, strain. This proves to be an alternative approach to tuning the impurity interactions in ABSNRs. We further investigate the effect of the nanoribbon's width and edge geometry on the RKKY exchange cou- pling in ABSNRs (see Fig. 1). Figure 9 shows the depen- dence of the RKKY coupling on the ribbon width for all three types of ABSNRs. The RKKY interaction has an oscillatory behavior in terms of the width of the ribbon. One can find from these figures that, with an increase in width, the exchange couplings drop at first and then their oscillating amplitudes decay with increasing the width of the ABSNRs finally approach converged value (almost zero). Such an oscillatory behavior versus the ribbon's width for graphene has been reported previously [69]. We observe that for the case that both of two impurities are situated within the interior of the nanoribbon (panel b) the magnetic coupling of ABSNRs with a finite width is always ferromagnetism that is very robust against the impurity movement. As is known, understanding the sublattice-dependent of local density of states (LDOS) is essential to assess the configuration-dependent exchange interaction. To do so, it is necessary to obtain the diagonal components of the unperturbed Green's function matrix (G(r, r, E)), for a lattice site at position r and energy E. Fig. 10 illustrates 8 the LDOS for an ABSNR with N = 300, p = 2, for both edge and bulk sites: (a) an edge lattice site with coordi- nate (150, 1) and (b) a bulk lattice site with coordinate (150, 7). Clearly for ABSNRs of type 2 and type 3, there is a high LDOS peak in the edge sublattice, for energies around E ∼ 4.7 eV (panel (a)). Here also two peaks at different energies around E ∼ 2.8 and E ∼ 8 eV appear for ABSNR of type 1. On the contrary, for a bulk site there is a high LDOS peak for energies around E ∼ 2.7 eV for ABSNRs of type 1 and type 3 (panel (b)). III. SUMMARY . To summarize, in this work, we numerically investigate the RKKY exchange coupling between two magnetic im- purities located on an armchair B2S nanoribbon, a new anisotropic Dirac cone material, as a strained graphene. In the first part of our study, employing a tight- binding approach, we investigate the electronic proper- ties of armchair-terminated B2S nanoribbons in the pres- ence of both strain and staggered sublattice potential. It is found that armchair B2S nanoribbons (ABSNRs) show different electronic and magnetic behaviors due to differ- ent edge morphologies. The band gap energy of ABSNRs depends strongly upon the applied staggered potential ∆ and thus one can engineer the electronic properties of the ABSNRs via tuning the external staggered potential. A complete and fully reversible semiconductor (or insula- tor) to metal transition has been observed via tuning the external staggered potential, which can be easily realized experimentally. Interestingly, for the ABSNRs belong to the family M = 6p, with M the width of the ABSNR and p an integer number, one can see that a band gap, in which a quasi-flatband completely detached from the bulk bands is always observed. As a key feature, the posi- tion of the quasi-flatbands in the energy diagram of AB- SNRs can be shifted by applying the in-plane strains εx and εy. At a critical staggered potential (∆c ∼ 0.5 eV), for ABSNRs with any width, the quasi-flatband changes to a perfect flatband. Then, within the tight-binding model we exploit the Green's function formalism, to reveal how the RKKY interaction between the impurities placed on a ABSNR is affected by mechanical strain, in the presence of a staggering potential. In particular, the effects of ribbon width,strain and staggered sublattice potential on the be- havior of RKKY interaction have been investigated. For impurities at fixed values distance, the increase of applied staggered potential leads to higher values of exchange coupling. It is shown that the RKKY interaction has an oscillating behaviour in terms of the applied staggered potentials, such that for two magnetic adatoms randomly distributed on the surface of an ABSNR the staggered po- tential can reverse the RKKY from antiferromagnetism to ferromagnetism and vice versa. The RKKY interac- 9 Figure 7. (Color online) The variation of χ versus distance between two impurities, for doped ABSNRs with N = 300, p = 2, for different types of ABSNRs with EF = 2 eV. The details of the panels are as follows: (a) Both the impurities are located at the same edge, the first impurity at the edge site with coordinate (5, 1) and the second one at lattice sites (n, 1). (b) Both the impurities are situated in the interior region of the ABSNR, the first impurity at lattice site (5, 7) and the second impurity at lattice points (n, 7), (c) One impurity is at the edge site (5, 1) and the other one is moved interior of the ABSNR along the line n = 7 at the lattice sites (n, 7), and (d) The impurities are located at the opposite edge sites (interedge magnetic coupling) with coordinates (150, 1), (150, 13) for the ABSNRs of types 1 and 2 and (150, 1), (150, 12) for the ABSNR of types 3. tion has an oscillatory behavior in terms of the width of the ribbon. It is shown that the magnetic interactions be- tween adsorbed magnetic impurities in ABSNRs can be manipulated by careful engineering of external staggered potential. Therefore, the ABSNRs would be expected to be a very promising candidate for spintronics and pseu- dospin electronics devices based on ABSNRs. REFERENCES [1] S. V. M. D. J. M. I. K. I. V. G. S. V. D. . A. A. F. K. S. Novoselov, A. K. Geim, Nature 438, 197 -- 200 (2005). [2] L. Z. S. L. H. D. X. Li, X. Wang, Science 319, 1229 -- 1232 [5] Z. L. P. Li and J. Yang, J. Phys. Chem. Lett. 9, 4852 -- 4856 (2018). [6] N. G. Szwacki, Nanoscale Research Letters 3, 49 (2007). [7] Y. Z. L. P. D. Ciuparu, R. F. Klie, J. Phys. Chem. B (2008). 108, 13 (2004). [3] M. L. C. Y.-W. Son and S. G. Louie, Phys. Rev. Lett. , [8] Z. S. X. D. S. D. J. C. N. X. H. G. F. Liu, C. Shen, J. 97, 216803 (2006). Mater. Chem. 20 (6267 ), 2197 (2010). [4] J. L. C. Z. Y. Zhao, X. Li and Q. Wang, J. Phys. Chem. [9] Y. G. X. C. Z. W. An, S. Bulusu, J. Chem. Phys. 124, Lett. 9, 1815 (2018). 154310 (2006). 501001502002505295-0.02-0.010.000.010.02501001502002505295-4-20246501001502002505295-0.050-0.0250.0000.0250.050501001502002505295-0.3-0.2-0.10.00.10.20.3(a) Type 1 Type 2 Type 3 Type 1 Type 2 Type 3(b) Type 1 Type 2 Type 3R/ax(c) Type 1 Type 2 Type 3cc(d)R/ax10 Figure 8. (Color online) The variation of χ with applied staggered potential for different types of the ABSNRs with p = 2, N = 300, for a fixed spatial distance between impurities. (a) Both the impurities are located on the same edge at the edge sites with coordinates (145, 1) and (155, 1) (b) Both the impurities are located in the interior of the ABSNR, on lattice points with coordinates (145, 7) and (155, 7), (c) One of the impurities is located on the edge site (145, 1) and the second one is on the lattice site (155, 7), and (d) The impurities are located on the opposite edge sites (interedge magnetic coupling) with coordinates (150, 1) and (150, 13). [10] H.-k. M. R. J. H. M. I. Eremets, V. V. Struzhkin, Science [21] L. Brey, H. Fertig, and S. D. Sarma, Phys. Rev. Lett. 293 (5528), 272 (2001). 99, 116802 (2007). [11] D. A. Abanin and D. A. Pesin, Phys. Rev. Lett. 106, [22] J. D.J. Priour, E. H. Hwang, and S. D. Sarma, Phys. 136802 (2011). Rev. Lett. 92, 117201 (2004). [12] F. M. J. C. . D. F. T. Dietl, H. Ohno, Science 287, 1019 [23] F. Matsukura, H. Ohno, A. Shen, and Y. Sugawara, (2000). [13] M. A. Ruderman and C. Kittel, Phys. Rev. 96, 99 (1954). [14] T. Kasuya, Prog. Theor. Phys. 16, 45 (1956). [15] K. Yosida, Phys. Rev. 106, 893 (1957). [16] M. Zare, F. Parhizgar, and R. Asgari, Phys. Rev. B 94, 045443 (2016). Phys. Rev. B 57, 2037(R) (1998). [24] K. T. Ko, K. Kim, S. B. Kim, H. D. Kim, J. Y. Kim, B. I. Min, J. H. Park, F. H. Chang, H. J. Lin, A. Tanaka, and S. W. Cheong, Phys. Rev. Lett. 107, 247201 (2011). [25] H. Ohno, Science 281, 951 (1998). [26] E. Minamitani, W. A. Dino, H. Nakanishi, and H. Kasai, [17] M. Shiranzaei, H. Cheraghchi, and F. Parhizgar, Phys. Phys. Rev. B 82, 153203 (2010). Rev. B 96, 024413 (2017). [27] C.-H. Hsu, P. Stano, J. Klinovaja, and D. Loss, Phys. [18] P. J. T. Eggenkamp, H. J. M. Swagten, T. Story, V. I. Litvinov, C. H. W. Swuste, and W. J. M. de Jonge, Phys. Rev. B 51, 15250 (1995). [19] F. sui Liu, W. A. Roshen, and J. Ruvalds, Phys. Rev. B 36, 492 (1987). Rev. B 92, 235435 (2015). [28] A. F. F. N. V. D. F. P. P. E. G. C. A. F. M. N. Baibich, J. M. Broto and J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988). [29] F. S. G. Binasch, P. Grunberg and W. Zinn, Phys. Rev. [20] M. A. H. Vozmediano, M. P. Lopez-Sancho, T. Stauber, B 39, 4828(R) (1989). and F. Guinea, Phys. Rev. B 72, 155121 (2005). [30] H. Z. F. Ye, G. H. Ding and Z. B. Su, EPL (Europhysics 012345-0.3-0.2-0.10.00.10.20.3012345-0.0150.0000.0150.030012345-0.04-0.020.000.020.040.06-0.05012345-0.6-0.4-0.20.00.2-0.7c Type 1 Type 2 Type 3(a) Type 1 Type 2 Type 3(b) Type 1 Type 2 Type 3(c)cD[eV] Type 1 Type 2 Type 3(d)D[eV]11 Letters) 90, 47001 (2010). [31] A. E. W. H. M. L. J. P. J. B. S. A. S. Z. S. . W. M. T. I. D. Hughes, M. Dane, Nature 446, 650 (2007). [32] S.-C. Z. J.-J. Zhu, D.-X. Yao and K. Chang, Phys. Rev. Lett. 106, 097201 (2011). [33] M. V. Hosseini and M. Askari, Phys. Rev. B 92, 224435 (2015). [34] F. Parhizgar, R. Asgari, S. Abedinpour, and M. Zareyan, Phys. Rev. B 12, 125402 (2013). [35] F. Parhizgar, H. Rostami, and R. Asgari, Phys. Rev. B 87, 125401 (2013). [36] F. Parhizgar, M. Sherafati, R. Asgari, and S. Satpathy, Phys. Rev. B 16, 165429 (2013). [37] M. Sherafati and S. Satpathy, Phys. Rev. B 83, 165425 (2011). [38] M. Shiranzaei, F. Parhizgar, J. Fransson, and H. Cher- aghchi, Phys. Rev. B 95, 235429 (2017). [39] M. Zare, Phys. Rev. B (2019). [40] R. A. M. Zare, F. Parhizgar, J. Magn. Magn. Mater. 456, 307 (2018). [41] M. Zare and E. Sadeghi, Phys. Rev. B 98, 205401 (2018). [42] T. S. V. I. L. C. H. W. S. P. J. T. Eggenkamp, H. J. M. Swagten and W. J. M. de Jonge, Phys. Rev. B 51, 15250 (1995). [43] X. C. X. J. Gao, W. Chen and F.-C. Zhang, Phys. Rev. B 80, 241302(R) (2009). [44] I. Zuti´c, J. Fabian, and S. D. Sarma, Rev. Mod. Phys. 73, 323 (2004). [45] R. Babar and M. Kabir, J. Phys. Chem. C 120, 27 (2016). [46] M. G. J. F. W. Han, R. K. Kawakami, Nat. Nano. 9, 794 -- 807 (2014). [47] J. Klinovaja and D. Loss, Phys. Rev. B 87, 045422 (2013). [48] S.-H. Z. Z. S. M. Y. H. Duan, S. Li and R.-Q.Wang, New J. Phys. 19, 103010 (2017). [49] V. M. Pereira and A. H. C. Neto, Phys. Rev. Lett. 103, 046801 (2009). [50] P. M. F. Liu and J. Li, Phys. Rev. B 76, 064120 (2007). [51] A. H. C. N. V. M. Pereira and N. M. R. Peres, Phys. Rev. B 80, 045401 (2009). [52] J. M. D. S. R. Power, P. D. Gorman and M. S. Ferreira, Phys. Rev. B 86, 195423 (2012). [53] F. Peng and W. Hongbin, Phys. B (Amsterdam) 407, 3434 (2012). [54] M. I. K. F. Guinea and A. K. Geim, Nat. Phys. 6, 30 (2010). [55] V. N. K. A. Sharma and A. H. C. Neto, Phys. Rev. B 87, 155431 (2013). [56] W. A. Harrison, (1999). [57] B. S. W. H. Tang, J. W.Jiang and Z. B. Su, Solid State Commun. 149, 82 (2009). [58] J. W. Jiang and H. S. P. (2015). , Phys. Rev. B 91, 235118 [59] J. H. A. D. Z. C. H. Zhang, Y. Li, Nano Lett. 16, 6124 (2016). [60] A. N. A. Carvalho, A. Rodin, Europhys. Lett. (EPL) 108 (4), 47005 (2014). [61] J. D. X. W. H. Guo, N. Lu and X. C. Zeng, J. Phys. Chem. C 118 (25), 14051 (2014). [62] G. D. M. S. D. K. Nakada, M. Fujita, Phys. Rev. B 54, 17954 (1996). [63] E. Lieb, Phys. Rev. Lett. 62, 1201 (1989). [64] K. Kusakabe and M. Maruyama, Phys. Rev. B 67, 092406 (2003). Figure 9. (Color online) The variation of χ as a function of the system width for ABSNRs with N = 150, in the absence of both strain and staggered potential. (a) Both the impuri- ties are located on a same edge, (b) Both the impurities are located on the interior of the ABSNR, (c) One of the impuri- ties is located on an edge site and the second one is on a bulk lattice site and (d) The impurities are located on the opposite edge sites. Figure 10. (Color online) Calculated local density of states for an ABSNR with N = 300, p = 2, for both edge and bulk sites: (a) an edge lattice site with coordinate (150, 1), (b) a bulk lattice site with coordinate (150, 7). 2468101-5052468101-0.50.00.51.02468101-2-101246810101020 Type 1 Type 2 Type 3(c)cc(a) Type 1 Type 2 Type 3 Type 1 Type 2 Type 3Width (p)Width (p)(d) Type 1 Type 2 Type 3(b)036912150123036912150.00.51.01.512345678910 Type 1 Type 2 Type 3(b)LDOS [eV-1]Energy [eV]03691215012312345678910 Type 1 Type 2 Type 3LDOS [eV-1](a)LDOS [eV-1][65] K. N. M. Fujita, K. Wakabayashi and K. Kusakabe, J. Phys. Soc. Jpn. 65, 1920 (1996). [67] M. Ezawa, Physica E 40, 1421 (2008). [68] T. K. H. Tamura, K. Shiraishi and H. Takayanagi, Phys. [66] J. Fern´andez-Rossier and J. J. Palacios, Phys. Rev. Lett. Rev. B 65, 085324. (2002). 99, 177204 (2007). [69] K. E. N. Gorjizadeh, A. A. Farajian and Y. Kawazoe, Phys. Rev. B 78, 155427 (2008). 12
1703.10336	1	1703	2017-03-30T07:29:06	Chiral topological excitons in the monolayer transition metal dichalcogenides	[ "cond-mat.mes-hall" ]	We theoretically investigate the chiral topological excitons emerging in the monolayer transition metal dichalcogenides, where a bulk energy gap of valley excitons is opened up by a position dependent external magnetic field. We find two emerging chiral topological nontrivial excitons states, which exactly connects to the bulk topological properties, i.e., Chern number =2. The dependences of the spectrum of the chiral topological excitons on the width of the magnetic field domain wall as well as the magnetic filed strength are numerically revealed. The chiral topological valley excitons are not only important to the excitonic transport due to prevention of the backscattering, but also give rise to the quantum coherent control in the optoelectronic applications.	cond-mat.mes-hall	cond-mat	a Chiral topological excitons in the monolayer transition metal dichalcogenides Z. R. Gong1,, W. Z. Luo1, Z. F. Jiang1, and H. C. Fu1 1College of Physics and Energy, Shenzhen University, Shenzhen, 518060, P. R. China [email protected] ABSTRACT We theoretically investigate the chiral topological excitons emerging in the monolayer transition metal dichalcogenides, where a bulk energy gap of valley excitons is opened up by a position dependent external magnetic field. We find two emerging chiral topological nontrivial excitons states, which exactly connects to the bulk topological properties, i.e., Chern number =2. The dependences of the spectrum of the chiral topological excitons on the width of the magnetic field domain wall as well as the magnetic filed strength are numerically revealed. The chiral topological valley excitons are not only important to the excitonic transport due to prevention of the backscattering, but also give rise to the quantum coherent control in the optoelectronic applications. Introduction The topological states on the boundary or surface is one of the most fascinating phenomena in solid state physics. The electronic topological states have been extensively investigated both theoretically1 -- 5 and experimentally6 -- 10 based on the re- alization of the spin-orbit interaction in topological insulator. The topology protected conducting edge states have distinct properties from bulk states, and play important role in the electronic transport and facilitate the implementation of the topo- logical states based electronic devices. The photonic topological states are also found in various systems such as microwave range photonic crystals,11, 12 the arrays of coupled optical resonators and waveguides13 -- 19 and metamaterials.20 -- 22 It is mo- tivated by the same idea of the electronic counterpart that the chiral edge states are prevent from backscattering and thus insensitive to disorder. Recently this idea has been generalized to the system consisting composite particles such as excitons and polarons.23 -- 25 The topological states on the boundary of the 2D materials have attracted a lot of interests due to its 1D nature of the boundary.26, 27 It is the band inversion at the Dirac points that leads to the band structure of the topological states in 2D materials, such as graphene.28 -- 33 As a new member of the 2D materials family, the monolayer TMDs, realized in laboratories recently, shed light on the valleytronics, exciton physics and photoelectronic applications. The monolayer TMDs are direct bandgap semiconductors, where the conduction and valence band edges locate at the doubly degenerate corners of the Brillouin zone, as known as the Dirac points.34 -- 36 In fact there are two obstacles to generate electronic topological states in monolayer TMDs through the same mechanism in graphene: one is that as the direct gap semiconductor the TMDs possesses huge band gap (∼ 1.6− 2eV); and another is that both the valence and conduction bands consist of the transition metal d orbitals on the same site, which prevents the staggered sublattice potential induced band inversion in graphene.32 Those properties basically forbid the electronic band inversion in pristine TMDs. Nevertheless, we can study the realization of the composite particles such as valley excitons instead of the electrons in the TMDs. For the bright excitons, there are two valley pseudospin configurations where the electron and hole both locate at either the K or −K valley.34 -- 36 The valley excitons follow optical selection rule, which means that the valley exciton locating at K(−K) valley only couples to σ+(σ−) circularly polarized light field.37 -- 43 Additionally, the Coulomb interaction between the electron and hole is exceptionally strong because of the 2D quantum confinement. The TMDs possesses valley degree of freedom, the valley-related optical selection rule and the strong Coulomb interaction, offering a new 2D system to explore the exciton physics. In addition, the two kinds of the valley excitons have opposite responses to the external out-of-plane magnetic field,44 which implies a possible band inversion for valley excitons. In this sense, the TMDs provide a unprecedent platform to investigate the topological states of valley excitons. In this paper, we shall theoretically investigate the chiral topological excitons emerging in the monolayer TMDs, where a bulk energy gap of valley excitons is opened up by a position dependent external magnetic field. The band dispersion of the bulk valley exciton realizes the massive Dirac cone, which exploits the strong valley-orbital coupling induced by Coulomb exchange interaction and the valley Zeeman splitting in the external magnetic field. For this unique Dirac cone of valley excitons, there are chiral topological exciton states emerging in the gap, whose number is determined by the bulk topological properties, i.e., Chern number =2. Since the time reversal symmetry is broken by the magnetic field, there are only two different chiral topological exciton modes. We numerically reveal the dependences of the spectrum of the chiral topological excitons on the width of the magnetic field domain wall as well as the magnetic filed strength. The chiral topological valley excitons are not only important to the excitonic transport due to prevention of the backscattering, but also have potential application to the quantum coherent control in the optoelectronics. Results The bulk topological properties The precise state of the valley exciton at K point is the superposition of the electron-hole pairs with all possible wave vector of the relative motion q and the definite wave vector of the center-of-mass motion k +ik = exp(−ik·r)∑ q φ (q) e† q+ k 2 ,↑ h† −q+ k 2 ,⇓vaci , (1) with the profile of relative motion φ (q), the electron (hole) creation operators e†(h†) and the electron (hole) spin up ↑ (spin down⇓). The quantum state −ik of the another valley exciton at −K point is the time reversal of the above one. In the low excitation limit of the valley excitons, the Hilbert space is spanned by the pseudospins described by the spin up(down) state +ik (−ik). We start from the Hamiltonian of the valley excitons in the out-of-plane magnetic field H (k) = ∑ (Hk + HE + HB) , k (2) where Hk = ¯h2k2 −→σ = (σx,σy), and HB = −gBµBB (r)σz is the position dependent valley Zeeman splitting. Here, M is total effective mass 2M I is the kinetic energy of the valley excitons, HE = −→h (k)·−→σ +(cid:12)(cid:12)(cid:12) of the electron and hole consisting valley excitons, k = (kx, ky) = (k cosθ, k sinθ) and r = (x, y) are the wave vector and the position coordinate of the valley excitons' center-of-mass motion respectively, µB is the Bohr magneton, I and σα (α = x, y, z) are respectively the identity matrix and Pauli matrices, and B (r) is the position dependent magnetic field. I is the valley-orbit coupling term with −→h (k)(cid:12)(cid:12)(cid:12) The valley-orbit coupling actually introduces the inter-valley transition, where the effective field −→h (k) = −J (k) (cos 2θ, sin 2θ) originates from the long range part of the Coulomb exchange interaction,46 where J(k) = Jk/K is the valley orbit coupling strength scaling linearly of the wave numbers k, and K = 4π/3a is the wave vector from the K to Γ point of the Brillouin B∆2 only depends on the parameters of the monolayer TMDs, the lattice constant a, the hooping zone. The constant J ≈ 2π Ka2t2 constant t, the effective dielectric constant ε, the Bohr radius of the exciton aB and the bang gap ∆. εa2 The magnetic response of the valley excitons, caused by the out-of-plane magnetic field, gives rise to the last term HB. Since the valley exciton is roughly regarded as a bounded electron-hole pair, both the electron and hole magnetic moments contribute to the exciton's Zeeman splitting in the magnetic field, leading to the effective g-factor of the valley excitons gB (k) = ∑ q φ (q)2 [ge (q, k)− gh (q, k)] . (3) The minus sign before the hole's g factor originates from the electron-hole duality in the semiconductor. Both the transition metal d orbital and valley magnetic moments contribute to the Lande g-factor of electron and hole. In the recent experiment based on WSe2, the typical measured gB is about 1.8.44 In this sense, the typical valley Zeeman splitting is about several meV when the applied magnetic field is up to 10T . Since the momentum scale of the profile φ (q) is about the reciprocal of the of Bohr radius of the valley exciton, which is much smaller than the momentum scale of the Lande g-factor of electron and hole. In this sense, we obtain the approximate effective Lande g-factor for the valley excitons gB ≈ gB (k = 0), which becomes independent of the wave vector of the center-of-mass motion. The σz term in the HB implies that the valley Zeeman energies are exactly opposite at K and −K points, which results from the valley-index-dependent magnetic moments of the valley excitons. The setup is schematically shown in Fig. 1. It is noticed that the inhomogeneity of the magnetic field U(rα)(α = e, h) shifts the momentum of the electron and hole and eventually cause the overlap of their wavefunctions, which gives rise to an additional inter-valley coupling of single states. It is straightforwardly calculated as V α inter = (cid:10)αK+p(cid:12)(cid:12)U (rα)(cid:12)(cid:12)α−K+q(cid:11) = Z drαu∗K+p (rα) exp (−i(K + p)· rα)U (rα)× u−K+q (rα) exp (i(−K + q)· rα) 1 drαU (rα) exp (−2iK· rα) , V ≈ (4) 2/9 where αki = exp (ik· rα)uk (rα)i are the Bloch wave functions of the electron and hole. In the last step we apply the condition p, q ≪ K. For a slowly varying magnetic field domain wall with a typical length of domain wall more than hundreds of lattice constant, the inter-valley couplings of the single states are about 10−5EB, where EB is the Zeeman energy. Obviously, they are sufficiently small. Additionally, since the inter-valley coupling of the excitons basically are the summation of the inter-valley coupling of single states, both of them are neglected in the following discussion. In the bulk, the Hamiltonian can be written as a matrix form45 −J(k)e−2iθ Hbulk = ∑ ∆ k (cid:18) ¯h2k2 2M + J (k)(cid:19) I +(cid:20) −J(k)e2iθ −∆ (cid:21) (5) on the basis {+ik ,−ik}, where ∆ = gBµBB is the valley Zeeman energy. The valley Zeeman energy approximately takes the fixed value because we only consider the bulk topological properties at the region where the applied magnetic field is homogeneous. The valley exciton dispersion splits into two branches with energies E± (k) = ¯h2k2 2M + J (k)±q∆2 + J(k)2 and corresponding eigen-wavefunctions u+ (k)i = cos u− (k)i = sin α 2 α 2 +ik − sin α e−2iθ+ik + cos 2 e2iθ−ik , α 2 −ik , (6) (7) (8) where tanα = J(k) ∆ . So it realizes a massive Dirac cone. The valley Zeeman energy plays the role of the mass in the Dirac-like equation and opens up a gap between two bands of valley excitons (see Fig.2(a)). As composite particles, the valley excitons still share the same Brillouin zone of the electron and hole in the monolayer TMDs. In contrast of the Dirac cones of the electron which locates at the corner of the Brillouin zone, the unique Dirac cone of the valley exciton locates at the center of the Brillouin zone. In order to describe the bulk topological property of the valley excitons, the Berry connection A (k) and Berry curvature Ω (k), defined respectively as A (k) ≡ hu− (k) i∇ku− (k)i Ω (k) ≡ ∇k × A (k) (9) (10) are introduced as the gauge potential and the gauge field of the lower valley exciton band. The Berry curvature is regarded as a magnetic field in the valley exciton center-of-mass momentum space, the integral of which over the k-space area gives rise to the Berry phase of the valley exciton if it adiabatically go around the area boundary. The Chern invariant is defined as the flux of the Berry curvature threading the entire Brillouin zone C = 1 2πZBZ dkΩ (k) . For the valley excitons described by equation (2), one find the Berry curvature centered at the Dirac cone Ω (k) = J2∆K (J2k2 + ∆2K2) 3 2 (11) (12) and thus the Chern invariant C = sign (∆) . The nonzero Chern number implies the existence of topology states of valley excitons. When the position dependent magnetic field is applied, it leads to the band inversion of the valley excitons, and the number of the topological charge equals to the difference of the bulk topological charges on the both side of the domain wall,32 which reads ν = C (left)-C (right)=2. Therefore, it is imaginable that two topological states will emerge at the vicinity of the magnetic field domain wall. Since the magnetic field breaks the time reversal symmetry and the Dirac cone is uniquely centered at zero momentum point, such topological states become chiral ones without time reversal symmetry. For the sake of simplicity, we assume the position-dependent magnetic field varies only along x-direction, namely B (r) ≡ B (x). When the magnetic field varies slowly along x-direction, the wavefunction of the topological exciton can be written as a two-component vector Φ (r) = (cid:20) Φ+ (x) Φ− (x) (cid:21) eipyy (13) 3/9 where Φ± (x) are the wavefunction profile, and py is the y-component momentum. The wavefunction profile Ψ (x) = {Φ+ (x) , Φ− (x)}T satisfies the following equation H (−i∇) Ψ (x) = EΨ (x) , (14) where the momentum in the original Hamiltonian is substituted by the operator in the real space as k → −i∇. According to the symmetry analysis of the above equation, there are two solutions Φ1 − (−x) corresponding to two topological excitons. However, the chirality index of the valley exciton Dirac cone equals to 2 in contrast with the well known Dirac cone, and the off-diagonal element ∝ ke−2iθ does not possess a simple operator form in the real space. So it is appropriate to solve the equations in the momentum space numerically. The details is presented in the Method section. − (−x) and Φ2 + (x) = −Φ2 + (x) = Φ1 Numerical results. In order to confirm the existence of two chiral topological excitons, we numerically evaluate the energy spectrum and the corresponding wavefunctions from equation (14). Here the constant J is chosen as 1eV, and the lattice constant for MoS2 is 3.49 A. In Fig.2 (a-c), the magnetic field domain wall is assumed as B (x) = B0 tanh(x/l). The typical spectrum of the topological excitons are demonstrated in Fig.2 (a). Obviously there are two different topological excitons, which are consistent with the Chern number. When qy tends to positive (negative) infinity, the dispersion of both topological excitons is convergent to the edge of the conduction (valence) band. The dependence of the spectrum on the magnetic field strengthes, widths and types of the magnetic domain wall is depicted in Fig.2 (b-d). In Fig.2.(b) and (c), the solid and dashed lines corresponds to the first and the second solution of the equation of wavefunction profile. The red, blue and black lines in Fig.2 (b) correspond to different magnetic valley Zeeman energy EB ≡ gBµBBmax = 1meV, 10meV, 50meV and l = 500a with a the lattice constant. Fig. 2(c) corresponds to different width of the magnetic domain wall l = 100a, 500a, 1000a and EB = 10meV, respectively. Obviously, the smaller the width of magnetic domain wall is, or the stronger the magnetic field strength is, the larger the energy difference between two topological excitons becomes. It actually results from the stronger quantum confinement of the magnetic domain wall, which suggests to adjust the spectrum of the topological excitons through both the magnetic field strength and the width of the magnetic domain wall. We compare the different types of the magnetic domain wall and find their spectrum almost coincide, which results from the similar behavior of those functions in the vicinity of x = 0 as tanh (x/l) ∼ sinh(x/l) ∼ x/l. It implies that the topological excitons basically locates at the vicinity of the magnetic domain wall. We also present the corresponding wavefunctions in Fig. 3. Since the four components of the two solutions are related to each other as Φ1 − (x;−qy), we only present the contour plot of wavefunction profile Φ1 + (x) versus x and qy in Fig. 3. Parameters in Fig. 3(a) and (b) are the same as ones in Fig. 2(b) and (c). Obviously under stronger quantum confinement with smaller width of the magnetic field domain wall or stronger magnetic field strength, the topological valley excitons become more local in the vicinity of the domain wall. + (−x;−qy) = −Φ2 − (−x; qy) and Φ2 + (x; qy) = Φ1 + (x) and Φ2 Discussion + (x)(cid:12)(cid:12) =(cid:12)(cid:12)R dxΦi where e is the polarization of pumping light field and p is the electric dipole moment. Since the relative and center-of-mass The optoelectronic properties of the topological excitons depend on their optical dipole defined as Di ≡ e·(cid:10)Φi(r)p0(cid:11), i = 1, 2, motions are independent for the topological exciton, the optical dipole can be factorized as Di = Ai(cid:0)D+ + (−1)i+1D−(cid:1) , where − (x)(cid:12)(cid:12) are the integrals of the wavefunction profile and D± = e·h±p 0i are the optical dipole for Ai =(cid:12)(cid:12)R dxΦi the valley excitons ±ik=0 . It indicates that the first and second topological excitons exactly inherit the optical selection rules from the linear combinations of the valley excitons +i±−i, which means that the two topological excitons can be initialized by utilizing the linear-polarized pumping light fields along x- and y-direction, respectively. Additionally, the optical dipole are adjustable by tuning the magnetic field because Ai depends on the widths of the magnetic field domain wall and magnetic field strength. These two topological excitons possess controllable gap(∼meV), optical initialization and robust transportation protected by topology, and thus may shed light on the quantum coherent optoelectronic devices based on these topological excitons in TMDs. The typical time to generate the valley excitons is much shorter than the the lifetime of the valley excitons, which is about tens of ps for monolayer TMDs. Additionally, the optical dipole of the topological chiral excitons can be adjusted by the applied magnetic field, which suggests a longer lifetime for the topological excitons. Therefore, the lifetime of the excitons would not affect the generation of the topological excitons. The longer lifetime allows more subtle control, which may facilitates the application of the optoelectronics based on the topological excitons. Methods The spectrum and the corresponding wave-functions of the chiral topological excitons can be obtained by solving equation (14). Actually, the quadratic term together with the linear dispersion from the massless Dirac equation gives rise to energy minimum 4/9 K = JM K2 ¯h2 ∼ 0.08 corresponding to the minimum energy − MJ2 around k 2K2 ¯h2 ∼ −0.04eV, which is much larger than the magnetic field induced Zeeman splitting. Therefore with the position dependent magnetic field B(r) and considering the emergent edge state around Dirac point, we drop the quadratic terms and obtain the Hamiltonian of topological excitons as Hex = ∑ k (cid:20) J (k) + ∆ −J(k)e−2iθ J (k)− ∆ (cid:21) . −J(k)e2iθ (15) Additionally, we define parameters KB (x) = KgEµBB (x) /J and KE = EK/J. The Schrodinger equation for the wave-function profile Φ± (x) becomes ((k− KB (x)− KE ) Φ+ (x)− ke−2iθΦ− (x) = 0, −ke−2iθΦ+ (x) + (k + KB (x)− KE ) Φ− (x) = 0. Although the momentum operator can be written as keiθ = kx + iky = −i∂x +∂y in the real space, the operators k and ke−2iθ do not possess a simple operator form. However, they are classical numbers in the momentum space. Thus it is convenient to solve above eigen-equations in the momentum space. By applying the Fourier transformation, the equation (16) becomes (16) −R d pxKB (kx − px) Ψ+ (px) + (k− KE ) Ψ+ (kx)− (kx+iqy)2 √k2 − x +q2 y Ψ− (kx) = 0, Ψ+ (kx) +R d pxKB (kx − px) Ψ− (px) + (k− KE) Ψ− (kx) = 0, (kx−iqy)2 √k2 x +q2 y   (17) The convolution terms means all the equations for different kx are coupled to each other. If we discretize the momentum kx, all the equations are linear and thus the eigenvalues KE and the corresponding wavefunctions can be obtained numerically. References 1. Hasan, M. Z. & Kane, C. L. Colloquium: Topological insulators. Rev. Mod. Phys. 82, 3045 (2010). 2. Qi, X.-L. & Zhang, S.-C. The quantum spin Hall effect and topological insulators. Phys. Today63, 33 (2010). 3. Moore, J. E. The birth of topological insulators. Nature464,194 (2010). 4. Qi, X.-L. & Zhang S.-C. Topological insulators and superconductors. Rev. Mod. Phys. 83, 1057 (2011). 5. Fu, L. & Kane, C. L. Topological insulators with inversion symmetry. Phys. Rev. B76, 045302 (2007). 6. Bernevig, B. A., Hughes, T. L. & Zhang S. C., Quantum Spin Hall Effect and Topological Phase Transition in HgTe Quantum Wells. Science 314, 1757 (2006). 7. Konig, M. et al., Quantum Spin Hall Insulator State in HgTe Quantum Wells. Science 318, 766 (2007). 8. Hsieh, D. et al., A topological Dirac insulator in a quantum spin Hall phase. Nature(London) 452, 970 (2008). 9. Xia, Y. et al., Observation of a large-gap topological-insulator class with a single Dirac cone on the surface. Nat. Phys. 5, 398 (2009). 10. Zhang, H. et al., Topological insulators in Bi2Se3, Bi2Te3 and Sb2Te3 with a single Dirac cone on the surface. Nat. Phys. 5, 438 (2009). 11. Wang, Z., Chong, Y., Joannopoulos, J. D. & Soljaci´c, M. Observation of unidirectional backscattering-immune topologi- cal electromagnetic states. Nature 461, 772 (2009). 12. Poo,Y. et al., Experimental Realization of Self-Guiding Unidirectional Electromagnetic Edge States. Phys. Rev. Lett. 106, 093903 (2011). 13. Cho, J., Angelakis, D. G. & Bose, S. Fractional Quantum Hall State in Coupled Cavities. Phys. Rev. Lett.101, 246809 (2008). 14. Koch, J., Houck, A. A., Hur, K. L. & Girvin, S. M. Time-reversal-symmetry breaking in circuit-QED-based photon lattices. Phys. Rev. A 82, 043811 (2010). 15. Umucalilar, R. O. & Carusotto, I. Artificial gauge field for photons in coupled cavity arrays. Phys. Rev. A 84, 043804 (2011). 16. Hafezi, M., Demler, E. A., Lukin, M. D. & Taylor, J. M. Robust optical delay lines with topological protection. Nat. Phys. 7, 907 (2011). 5/9 17. Hafezi, M. et al., Imaging topological edge states in silicon photonics. Nat. Photon.7, 1001 (2013). 18. Liang, G. Q. & Chong, Y. D. Optical Resonator Analog of a Two-Dimensional Topological Insulator. Phys. Rev. Lett. 110, 203904 (2013). 19. Ningyuan, J. et al., Time- and Site-Resolved Dynamics in a Topological Circuit. Phys. Rev. X 5, 021031 (2015). 20. Yannopapas, V. Topological photonic bands in two-dimensional networks of metamaterial elements. New J. Phys.14, 113017 (2012). 21. Khanikaev, A. B. et al., Photonic topological insulators. Nat. Mater.12, 233 (2013). 22. Chen, W.-J. et al., Experimental realization of photonic topological insulator in a uniaxial metacrystal waveguide. Nat. Commum.5, 5782 (2014). 23. Karzig, T., Bardyn, C.-E., Lindner, N. H. & Refael, G. Topological Polaritons. Phys. Rev. X 5, 031001 (2015). 24. Bardyn, C.-E., Karzig, T., Refael, G. & Liew, T. C. H. Topological polaritons and excitons in garden-variety systems. Phys. Rev. B 91, 161413(R) (2015). 25. Nalitov, A. V., Solnyshkov, D. D. & Malpuech, G. Polariton Z Topological Insulator Phys. Rev. Lett. 114, 116401 (2015). 26. Ryu, S. & Hatsugai, Y. Topological Origin of Zero-Energy Edge States in Particle-Hole Symmetric Systems. Phys. Rev. Lett. 89, 077002 (2002). 27. Li, H., Peng, H.-L. & Liu, Z.-F. Two-Dimensional Nanostructures of Topological Insulators and Their Devices. Acta Physico-Chemica Sinca 28, 2423 (2012). 28. Brey, L. & Fertig, H. A. Electronic states of graphene nanoribbons studied with the Dirac equation. Phys. Rev. B73, 235411 (2006). 29. Peres, N. M. R., Castro Neto, A. H. & Guinea, F. Conductance quantization in mesoscopic graphene. Phys. Rev. B 73, 195411 (2006); Electronic properties of disordered two-dimensional carbon. Phys. Rev. B 73, 125411 (2006). 30. Castro Neto, A. H. et al., The electronic properties of graphene. Rev. Mod. Phys. 81, 109 (2009). 31. Castro, E. V. et al., Localized States at Zigzag Edges of Bilayer Graphene. Phys. Rev. Lett. 100, 026802 (2008). 32. Yao, W., Yang, S. A. & Niu, Q. Edge States in Graphene: From Gapped Flat-Band to Gapless Chiral Modes. Phys. Rev. Lett. 102, 096801 (2009). 33. Qiao, Z., Tse, W.-K., Jiang, H., Yao, Y. & Niu, Q. Two-Dimensional Topological Insulator State and Topological Phase Transition in Bilayer Graphene. Phys. Rev. Lett. 107, 256801(2011). 34. Xiao, D. et al., Coupled Spin and Valley Physics in Monolayers of MoS2 and Other Group-VI Dichalcogenides. Phys. Rev. Lett. 108, 196802 (2012). 35. Xu, X., Yao, W., Xiao, D. & Heinz, T. F. Spin and pseudospins in layered transition metal dichalcogenides. Nat. Phys. 10, 343 (2014). 36. Liu, G.-B. et al., Electronic structures and theoretical modelling of two-dimensional group-VIB transition metal dichalco- genides. Chem. Soc. Rev.44, 2643 (2015). 37. Mak, K. F., He, K., Shan, J. & Heinz, T. F. Control of valley polarization in monolayer MoS2 by optical helicity Nat. Nanotechnol.7, 494-498 (2012). 38. Zeng, H. et al., Valley polarization in MoS2 monolayers by optical pumping. Nat. Nanotechnol. 7, 490 (2012). 39. Cao, T. et al., J. Valley-selective circular dichroism of monolayer molybdenum disulphide. Nat. Commum. 3, 887 (2012). 40. Jones, A. M. et al., Optical generation of excitonic valley coherence in monolayer WSe2. Nat. Nanotechnol. 8, 634-638 (2013). 41. Wang, G. et al., Valley dynamics probed through charged and neutral exciton emission in monolayer WSe2. Phys. Rev. B 90, 075413 (2014). 42. Kadantsev, E. S. & Hawrylakb, P., Electronic structure of a single MoS2 monolayer. Solid State Comm. 152, 909 (2012). 43. Qiu, D. Y., Cao, T. & Louie, S. G., Nonanalyticity, Valley Quantum Phases, and Lightlike Exciton Dispersion in Mono- layer Transition Metal Dichalcogenides: Theory and First-Principles Calculations. Phys. Rev. Lett. 115, 176801 (2015). 44. Aivazian, G. et al., Magnetic control of valley pseudospin in monolayer WSe2. Nat. Phys. 11,148 (2015). 45. Yu H., Cui X., Xu X., & Yao W., Valley excitons in two-dimensional semiconductors. National Science Review 2, 57 (2015). 6/9 46. Yu, H. et al., Dirac cones and Dirac saddle points of bright excitons in monolayer transition metal dichalcogenides. Nat. Commum. 5, 3876 (2014). Acknowledgements (not compulsory) This work is supported by NSFC Grants No. 11504241 and the Natural Science Foundation of SZU Grants No. 201551. Author contributions statement Z.R.G. proposed the project and H. C. F. supervised the project. W.Z.L. carried out the study. Z. F. J. provide the numerical calculations. All authors analysed the results and co-wrote the paper. Competing financial interests The authors declare no competing financial interests. Figure 1. Carton of generating the topological exciton in the vicinity of the magnetic field domain wall. The orange wave line represents the pumping light field. The wavefuntion of its relative motion is represented by the orange area, in which the electron and the hole are respectively represented by the purple sphere with "-" symbol and the green sphere with "+" symbol. The colored region of the monolayer TMDs represent the position dependent magnetic field. 7/9 (a) (c) 15 10 5 0 (b) 15 10 5 0 -3 -2 -1 0 1 2 3 (d) 15 10 5 0 3.2 3.1 -1 0 1 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 Figure 2. (a) Typical spectrum of the topological excitons, which are denoted by the red and blue solid lines. Here, the light blue, light red and purple region respectively denotes the upper band, the lower band and the light cone of the pumping light field. (b) The spectrum of the topological excitons versus qy for different magnetic field strength. (c) The spectrum of the topological valley excitons versus qy for different widths of the magnetic domain wall. (d) The spectrum of the topological excitons versus qy for different types of the magnetic domain wall. Insert: magnified view of the spectrum. See text for the details. 8/9 (a) (b) 10 5 0 -5 -10 5 0 -5 -10 -3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 5 2.5 0 -2.5 -5 2.5 0 -2.5 -5 -3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 -2 -1 0 1 2 3 0.14 0.07 0 -0.07 0.17 0.08 -0.02 -0.11 Figure 3. Contour plot of the wavefunction profile Φ1 magnetic field domain wall and (b) different magnetic field strengths. Obviously under stronger quantum confinement with smaller widths of the magnetic field domain wall or stronger magnetic field strengths, the topological valley excitons becomes more local in the vicinity of the domain wall. + (x) versus x and qy for (a) different widths of the + (x) and Φ2 9/9
1607.02321	1	1607	2016-07-08T11:32:48	Magnetic properties of Cobalt films described by second order perturbed Heisenberg Hamiltonian	[ "cond-mat.mes-hall" ]	Second order perturbed Heisenberg Hamiltonian was employed to investigate the magnetic properties of hexagonal Cobalt films. Initially the number of nearest neighbors and the constants arisen from the partial summation of the dipole interactions of the structure of cobalt were calculated using some special algorithms. Minimization of the energy difference between the easy and hard direction of a memory device is very important. When the energy difference between the easy and hard directions is significantly small, the magnetic moments in a memory device can be quickly rotated between easy and hard directions under the influence of a small magnetic field. The thickness of a cobalt film corresponding to this minimum energy difference calculated using this theoretical model agrees with some experimental data of cobalt based magnetic memory devices.	cond-mat.mes-hall	cond-mat	Magnetic properties of Cobalt films described by second order perturbed Heisenberg Hamiltonian P. Samarasekara and Amila D. Ariyaratne Department of Physics, University of Peradeniya, Peradeniya, Sri Lanka Abstract Second order perturbed Heisenberg Hamiltonian was employed to investigate the magnetic properties of hexagonal Cobalt films. Initially the number of nearest neighbors and the constants arisen from the partial summation of the dipole interactions of the structure of cobalt were calculated using some special algorithms. Minimization of the energy difference between the easy and hard direction of a memory device is very important. When the energy difference between the easy and hard directions is significantly small, the magnetic moments in a memory device can be quickly rotated between easy and hard directions under the influence of a small magnetic field. The thickness of a cobalt film corresponding to this minimum energy difference calculated using this theoretical model agrees with some experimental data of cobalt based magnetic memory devices. I. INTRODUCTION Thin films of cobalt based materials find potential applications in memory discs and storage devices [1]. But any theoretical investigation related to the ferromagnetic cobalt thin films have not been performed by any researcher. The magnetic properties of ferromagnetic thin and thick films with simple cubic (sc), body centered cubic (bcc) and face centered cubic (fcc) have been explained using oriented, second order perturbed and third order perturbed Heisenberg Hamiltonian by us previously [2-4, 8]. The easy and hard directions have been determined in each case. According to those studies, easy and hard direction of ferromagnetic thin films depend on magnetic exchange interaction, dipole interaction, second and fourth order anisotropy, demagnetization factor, magnetic field and stress induced anisotropy. In all above cases, c-axis of the lattice was assumed to be perpendicular to the substrate. In addition, the ferromagnetic 1 properties of Fe and Ni have been explained using a similar model by some other researchers [5, 6]. Due to the complexness of the hexagonal closed packed (hcp) structure, the determination of constants arisen from the partial summations of dipole interactions were complicated compared to determination of those of sc, fcc and bcc lattices. So an algorithm has been employed to evaluate these constants. This same strategy has been applied to determine these constants of Nickel ferrite by us previously [7, 11, 12, 13, 14]. According to experimental data, the stress induced anisotropy of ferrite films is considerable [9, 10, 15]. II. THE MODEL The Heisenberg Hamiltonian of any ferromagnetic film can be basically expressed as following. r r SS nm ω J . r r SS nm . + H=- ∑ 2 nm , 2 ∑ ( nm ≠ r SN d n 3 r mn / µ 0 rr . rS m mn (3 − r r Sr mn . n ) ) − ∑ m D λ m )2( S ( m z 2 ) − ∑ m D λ m )4( S ( m z 4 ) )( 5 r mn ∑ − [ r H − ( nm , r S m )]. ∑− m SinK s θ 2 m (1) For a thick ferromagnetic film, the solution of above equation can be given as, E −=θ )( J 2 [ NZ 0 + (2 N − )1 Z 1 ] +Φ+ N { 0 (2 N Φ− )1 1 ωω 3 }( + 8 − N (cos )2( 2 θ D m + cos 4 θ D m )4( + H in sin θ + H out cos θ − cos θ )2 + K s θ )2sin 8 N d µ 0 [ − ω 3 4 ( − +Φ+Φ 2 ) 1 0 )2( D m + 2 D m )4( cos 2 θ 2 (] N − sin)2 2 θ 2 2 C 22 − 1 C 11 [ − ω 3 4 ( +Φ+Φ ) 1 0 )2( D m + 2 D m )4( cos 2 2 θ ] sin 2 θ 2 (2) In above equation; N, J, Z − Φ , nm , nm − ω, θ, Dm (2), Dm (4), Hin, Hout, Nd, Ks are total number of layers, spin exchange interaction, number of nearest spin neighbors, constants arisen from partial summation of dipole interaction, strength of long range dipole interaction, azimuthal angles of spins, second order anisotropy, 2 fourth order anisotropy, in plane applied field, out of plane applied field, demagnetization factor and the stress induced anisotropy factor, respectively. Here C11 and C22 are given by, C11= JZ 1 +Φ− 31( 1 ω 4 cos θ )2 − (sin2 2 θ− cos 2 mDθ ) )2( + 4 cos 2 θ (cos 2 θ − sin3 2 )4( θ ) D m + H in sin θ + H out cos θ − 2 N d µ 0 + 4 K s 2sin θ C22= 2 JZ 1 +Φ− 31( 1 ω 2 cos θ )2 − (sin2 2 θ− cos 2 mDθ ) )2( + 4 cos 2 θ (cos 2 θ − sin3 2 )4( θ ) D m + H in sin θ + H out cos θ − 2 N d µ 0 + 4 K s 2sin θ III. RESULTS AND DISCUSSION The diagram of conventional unit cell of cobalt with lattice parameters (a and c) is given in figure 1. The c/a ratio for Cobalt is 1.62. Fig. 1: Conventional unit cell of cobalt 3 Find the translational vectors a, b, c for the structure Select a lattice point in 1st layer with position vector na0a + nb0b + nc0c Select a lattice point in 2nd layer with position vector naa + nbb + ncc Calculate the distance "d" between the two lattice points. d= (na-na0)a + (nb-nb0)b + (nc-nc0)c Add 1 to Z Yes Is d <= 1 ? No Generate another lattice point on 2nd layer. I.e. values for na and nb Fig. 2: Algorithm to calculate Z 4 Find the translational vectors a, b, c Find the position vector of a lattice point on the 2nd layer ( n0aa + n0bb + n0cc ) n1 =-5000, n2 =-5000 r = (n0aa + n0bb + n0cc) + n1a + n2b Let R =r r = r R Find W_temp matrix W_temp = 1 3 R  2 31 − r a  3  rr ba  3 rr  ca − − 3 − rr ab 2 31 − r b 3 rr cb − − 3 rr ac 3 − rr bc 2 31 r c −      W = W + W_temp Increment n2 by 1 Increment n1 by 1 Is n2 =5000 ? Yes Set n2 = -5000 No No Is n1 =5000 ? Yes End Fig. 3: Algorithm to calculate Φ 5 Then the algorithm given in figure 2 was implemented to evaluate the number of nearest neighbors in a cobalt film with spin layers parallel to the substrate. The 2nd algorithm given in figure 3 was applied to calculate the constants arisen from the partial summations of dipole interactions. Finally following values were found for hcp lattice. The number of nearest neighbors in one lattice plane=Z0=6 Number of nearest neighbors between two adjacent lattice planes=Z1=3 Constants arisen due to the partial summation of dipole interactions in one layer=Φ0=11.0324 Constants arisen due to the partial summation of dipole interactions between two adjacent layers=Φ1=0.4210 Since the experimental values of D )2( , D m )4( , m KHH , in out , , J and ω have not been s measured for cobalt thin films by any researcher yet, the simulations were carried out for a reasonable set of D m )2( , D m )4( , below. KHH , in out , , J and ω values as given s J = D ( ) 2 m = H in H out = = ωωµωωωω N d 0 K s = = 10 and D ( ) 4 m ω = 5 The graph of θ)(E ω versus angle is plotted in Figure 4. As indicated, the easy direction is found at an angle of about 40° and the hard direction occurs at an angle of about 140°. For a material with a simple structure, the angle between the easy and hard directions is 900. But the angle between easy and hard directions is 1000 in this case due to the complexness of the structure of cobalt. The Figure 5 shows the variation of the energy difference between the easy direction and the hard direction against the number of layers for a Cobalt thin film. It could be observed from this graph that the energy difference is a minimum for a film of 50 layers. Therefore, a hard disk drive would require a less amount of energy to store data if the magnetic film is synthesized with the number of layers being in the above region. This theoretical result agrees with that of modern hard 6 disks. The optimum experimental results for Co based magnetic memory devices have been obtained for a thin film with the same number of layers by some other researchers [1]. Fig. 4: The graph of θ)(E ω versus θ for a Cobalt thin film with 5000 layers Fig. 5: The variation of the energy difference between the easy and the hard directions against the number of layers N 7 3-D plot of θ)(E ω versus angle and number of layers is given in figure 6. The angle and number of layers corresponding to easy and hard directions can be determined using this plot. The difference between the maximum and minimum energies is really small around N=50 according to this graph too. So the energy required to rotate from easy to hard direction (crystal anisotropy) is significantly small for films with 50 layers. 19 x 10 4 2 0 -2 ω / ) θ ( E -4 100 50 angle θ(radians) 20 0 0 40 N 100 80 60 Fig. 6: 3-D plot of θ)(E ω IV. CONCLUSION versus angle (θ) and number of layers (N) Values of number of nearest neighbors and the constants arisen from the partial summations of dipole interactions calculated using the algorithms given in figures 2 and 3 are Z0=6, Z1=3, Φ0=11.0324 and =Φ1=0.4210 for cobalt thin films. This simulation was carried out for a selected set of values of energy parameters in order to study the variation of total magnetic energy with angle (θ) and the 8 number of layers (N). According to the energy curves, the energy difference between the easy and hard directions can be minimized at N=50. This number of layers (N=50) is approximately equal to the thickness of cobalt films synthesized for magnetic memory applications by some other researchers [1]. This implies that the magnetic moments of a cobalt based memory device can be easily rotated between easy and hard directions, when the number of layers is 50. REFERENCES 1. U-Hwang Lee et al., 2006. Templated synthesis of nanostructured cobalt thin Film for potential terabit magnetic recording. NANO 1(1), 41-45. 2. P. Samarasekara, 2006. Second order perturbation of Heisenberg Hamiltonian for non-oriented ultra-thin ferromagnetic films. Electronic Journal of Theoretical Physics 3(11), 71-83. 3. P. Samarasekara and William A. Mendoza, 2010. Effect of third order perturbation on Heisenberg Hamiltonian for non-oriented ultra-thin ferromagnetic films. Electronic Journal of Theoretical Physics 7(24), 197-210. 4. P. Samarasekara and S.N.P. De Silva, 2007. Heisenberg Hamiltonian solution of thick ferromagnetic films with second order perturbation. Chinese Journal of Physics 45(2-I), 142- 150. 5. K.D. Usadel and A. Hucht, 2002. Anisotropy of ultrathin ferromagnetic films and the spin reorientation transition. Physical Review B 66-024419, 1-6. 6. A. Hucht and K.D. Usadel, 1999. Theory of the spin reorientation transition in ultra-thin ferromagnetic films. Journal of Magnetism and Magnetic Materials 203, 88-91. 7. P. Samarasekara, 2007. Classical Heisenberg Hamiltonian solution of oriented spinel ferrimagnetic thin films. Electronic Journal of Theoretical Physics 4(15), 187-200. 8. P. Samarasekara, 2006. A solution of the Heisenberg Hamiltonian for oriented thick ferromagnetic films. Chinese Journal of Physics 44(5), 377-386. 9. P. Samarasekara, 2002. Easy Axis Oriented Lithium Mixed Ferrite Films Deposited by the PLD Method. Chinese Journal of Physics 40(6), 631-636. 9 10. P. Samarasekara, 2003. A pulsed rf sputtering method for obtaining higher deposition rates. Chinese Journal of Physics 41(1), 70-74. 11. P. Samarasekara, 2010. Determination of energy of thick spinel ferrite films using Heisenberg Hamiltonian with second order perturbation. Georgian electronic scientific journals: Physics 1(3), 46-49. 12. P. Samarasekara, 2011. Investigation of Third Order Perturbed Heisenberg Hamiltonian of Thick Spinel Ferrite Films. Inventi Rapid: Algorithm Journal 2(1), 1-3. 13. P. Samarasekara and William A. Mendoza, 2011. Third order perturbed Heisenberg Hamiltonian of spinel ferrite ultra-thin films. Georgian electronic scientific journals: Physics 1(5), 15-18. 14. P. Samarasekara, M.K. Abeyratne and S. Dehipawalage, 2009. Heisenberg Hamiltonian with Second Order Perturbation for Spinel Ferrite Thin Films. Electronic Journal of Theoretical Physics 6(20), 345-356. 15. P. Samarasekara and Udara Saparamadu, 2013. Easy axis orientation of Barium hexa-ferrite films as explained by spin reorientation. Georgian electronic scientific journals: Physics 1(9), 10-15. 10
1603.03485	1	1603	2016-03-10T23:15:47	Nanometer-scale Exchange Interactions Between Spin Centers in Diamond	[ "cond-mat.mes-hall" ]	Exchange interactions between isolated pairs of spin centers in diamond have been calculated, based on an accurate atomistic electronic structure for diamond and any impurity atoms, for spin-center separations up to 2~nm. The exchange interactions exceed dipolar interactions for spin center separations less than 3~nm. NV$^-$ spin centers, which are extended defects, interact very differently depending on the relative orientations of the symmetry axis of the spin center and the radius vector connecting the pair. Exchange interactions between transition-metal dopants behave similarly to those of NV$^-$ centers. The Mn\---Mn exchange interaction decays with a much longer length scale than the Cr\---Cr and Ni\---Ni exchange interactions, exceeding dipolar interactions for Mn\---Mn separations less than 5~nm. Calculations of these highly anisotropic and spin-center-dependent interactions provide the potential for design of the spin-spin interactions for novel nanomagnetic structures.	cond-mat.mes-hall	cond-mat	a Nanometer-scale Exchange Interactions Between Spin Centers in Diamond Optical Science and Technology Center and Department of Physics and Astronomy, University of Iowa, Iowa City, IA 52242 V. R. Kortan, C. S¸ahin, and M. E. Flatt´e Exchange interactions between isolated pairs of spin centers in diamond have been calculated, based on an accurate atomistic electronic structure for diamond and any impurity atoms, for spin- center separations up to 2 nm. The exchange interactions exceed dipolar interactions for spin center separations less than 3 nm. NV− spin centers, which are extended defects, interact very differently depending on the relative orientations of the symmetry axis of the spin center and the radius vector connecting the pair. Exchange interactions between transition-metal dopants behave similarly to those of NV− centers. The Mn -- Mn exchange interaction decays with a much longer length scale than the Cr -- Cr and Ni -- Ni exchange interactions, exceeding dipolar interactions for Mn -- Mn separations less than 5 nm. Calculations of these highly anisotropic and spin-center-dependent interactions provide the potential for design of the spin-spin interactions for novel nanomagnetic structures. A single spin, such as from a defect or dopant, can control the properties of a nanomagnetic system[1], sug- gesting pathways to constructing novel magnetic mate- rials or magnetic behavior through designed assembly e.g. of spins in metals, insulators, and semiconductors[2 -- 8]. Spin centers in wide-gap semiconductors such as diamond exhibit exceptionally long room-temperature spin coherence times[9], permitting coherent interactions among such spin centers over length scales of many nanometers, and the corresponding shaping of spin dy- namics in the spin assemblies. As the interactions oc- cur through weak, long-range, largely isotropic dipolar interactions[10, 11] the interaction effects on spin dynam- ics are slow (less than 1 µeV). Continued improvement of control in spin-center positioning, such as through ion implantation[4, 12, 13], will lead to assemblies with short- range coupling, where exchange interactions may domi- nate over dipolar interactions, producing anisotropic[3] interactions that are orders of magnitude greater than dipolar interactions. The current focus on NV− centers in diamond, due especially to the convenience of its lev- els and optical selection rules for spin initialization and readout[14], may also shift to other spin centers that are easier to address and manipulate electrically, especially transition-metal dopants that possess partially-filled d levels[15, 16]. Here we construct a highly-accurate theoretical de- scription of the spin center in bulk diamond, and a very efficient theoretical methodology to evaluate the exchange-coupling between spins in diamond, including both NV− centers and transition-metal spin centers. We include the weak spin-orbit interaction in bulk diamond and the strong spin-orbit interaction of a transition-metal dopant, as well as the dependence of an NV− spin cen- ter's interaction on the N-V axis direction. We find that exchange interactions dominate over dipolar interactions for spin-center separations smaller than 3 nm, except for the more delocalized Mn spins, which are exchange- dominated for separations less than 5 nm. The theo- retical techniques that have been previously applied to diamond find calculations of spin-spin interactions very challenging, either (as with density functional theory[15 -- 17]) due to the very large supercell sizes required for such calculations, or (as with symmetry-based group-theory analyses[18]) due to the inability to constrain the prob- lem to a very small number of experimentally-determined quantities. Our approach is a rigorously tested spds∗ de- scription of the bulk electronic structure[19] and a set of effective impurity potentials, including for d states, that replicate the energies of the spin-center states found in density functional theory calculations or experimen- tal measurements. Once those are known the elec- tronic properties of the pair are efficiently evaluated us- ing a Green's function-based Koster-Slater method[20] as described in Ref. [21], and here extended to the spds∗ system required to accurately describe bulk dia- mond and the d levels of transition-metal dopants. This approach[21], by exactly solving for the electron propaga- tor in the regions between defects, permits calculations of the exchange interaction of a defect pair to proceed with a rapid speed that is independent of the defect sep- aration. The Hamiltonian for a point defect (impurity atom or vacancy) has the form H = H0 + V , where H0 is the spds∗ Hamiltonian of Ref. 19 and V = † (cid:96)ms(R0)c(cid:96)ms(R0) (cid:88) 4(cid:88) (cid:96),m,s + j=1 +(2/3) U os (cid:96)msc (cid:88) (cid:88) (cid:96) † (cid:96)msc (cid:96)ms(Rj)c(cid:96)ms(Rj) U nn (1) † (cid:96)ms(R0)c(cid:96)m+1s−1(R0) + H.c.]. ∆(cid:96)[c (cid:96),m,s Here U(cid:96)ms is the energy difference for the orbital with spin s, angular momentum (cid:96) and azimuthal quantum number m, either at the point defect site (U os) or at the nearest neighbors (U nn), and ∆(cid:96) is the point de- fect's spin-orbit interaction for the (cid:96) angular-momentum † states. c (cid:96)ms(R) (c(cid:96)s(R)) is the creation (annihilation) operator for a spin-s electron in the (cid:96), m orbital at site R. The point defect is located at R0, and the four nearest- TABLE I. On-site potentials (eV) for transition-metal impu- rities in diamond, including the nonmagnetic and magnetic potentials for d electrons of t2 and e symmetry, and the spin- orbit interaction strength ∆ for p and d electrons, in eV. nonmagnetic t2 e magnetic t2 e spin-orbit p d Cr Mn Fe Co Ni -18.89 -21.45 -0.26 -1.85 0.09 0.02 -19.30 -22.50 -0.14 -1.00 -0.03 -0.08 -19.20 -23.15 0 0 -0.15 -0.12 -20.57 -24.64 -0.26 -0.21 -0.10 -0.19 -21.67 -27.03 -0.43 -0.38 -0.08 -0.33 TABLE II. On-site and nearest-neighbor p-orbital potentials, magnetic and nonmagnetic, for nitrogen and a vacancy in diamond. on-site nearest-neighbor nonmagnetic magnetic nonmagnetic magnetic N V -5.33 50 2.93 0 0 -0.26 0 -2.97 neighbor sites are labeled by R1-R4. The spin-orbit po- tential has been calculated from atomic energies[22 -- 24] and using the Land´e interval rule. Spin-orbit interac- tions are positive for angular-momentum shells less than half full, and negative otherwise. For transition-metal dopants, to position the d states of correct tetrahedral symmetry (t2 or e) at the correct locations within the di- amond band gap, U os magnetic and nonmagnetic poten- tials are determined for the t2 and e states, and reported in Table I. U nn = 0 for transition-metal dopants. For the NV− spin center, defect potentials are only required on the p orbitals, however the shift in the atomic posi- tions requires nonzero defect potentials on the nearest neighbors as well. These values are reported in Table II. We calculate the retarded Green's function for the bulk Hamiltonian H0, G0(k, ω) = [ω−H0(k)+iδ]−1, and from this the real-space Green's function G0(Ri, Rj, ω), where G0 is a matrix with rows and columns labeled by (cid:96), m, and s. The properties of the defects, either point defects or pairs, are determined from solving the Dyson equation in real space, G(ω) = [I − G0(ω)V ] −1 G0(ω) . (2) Due to the limited number of positions in real space where the potential is non-zero, Eq. (2) can be solved rapidly once the G0(Ri, Rj, ω) have been tabulated. Figure 1 compares the on site and nearest neighbor spin resolved local density of states (LDOS) for the two 2 FIG. 1. Spin resolved local density of states (LDOS) on the impurity site and nearest neighbor carbon site for Cr and Ni spin centers. The continuum states in the conduction and valance bands are plotted on the scale of the left axis. The probabilities of finding the electron on the impurity for mid- gap impurity states are plotted on the scale of the left axis. The nearest-neighbor contributions are in red, whereas the on-site contributions are in black. transition-metal spin-1 dopants, Ni and Cr. Within the diamond band gap, the Cr spin center forms one doubly- degenerate spin-up and one doubly-degenerate spin-down e level as well as one triply degenerate spin-up and one triply-degenerate spin-down t2 level. The ground state for Cr has two electrons in the spin-up e state and the rest empty. The Ni dopant levels are arranged differently, with the t2 levels in the gap and the e levels below the edge of the valence band, showing as a broad resonance. The t2 levels for Ni show a visible splitting in Fig. 1 due to the large spin orbit coupling for Ni. The ground state for the Ni spin center has two electrons in the spin-up t2 states. As found in Ref. 15 and 16 with density functional theory calculations, the Cr ground state possesses more spectral weight on the site of the dopant than the Ni ground state, with a ratio of ∼ 2:1. The construction of the NV− center requires tracking different mid-gap levels. The NV− center exhibits four levels in the gap, the lower two have a1 symmetry and the upper two are spin- split, orbitally-degenerate ex and ey levels, all of which originate from p orbitals (t2 character)[17]. The ground state for the NV− center fills electrons up through the spin-up ex and ey states. These trends are reflected in the real space proba- bility density of the highest occupied molecular orbital (HOMO) of each of the spin centers in Fig. 2. The ground state spins for each dopant in diamond are Fe: spin 0, Mn and Co: spin 1/2, and NV−, Cr and Ni: spin 1. All of the transition-metal dopant HOMOs show the same overall spatial symmetry regardless of spin, which is expected because the propagation of electron waves in the host material most determines the probability den- sity symmetry[21]. The Fe, Mn, and Cr dopants all have e-like HOMOs whereas the NV−, Co and Ni spin cen- ters have t2-like HOMOs, and therefore among the point defects Fe, Mn and Cr all have larger wave function prob- 10101Integrated Spectral WeightCrNinearestneighbor-0.5 0 0.5 1 1.5-4-2 0 2 4 6 8 10LDOS (1/eV)Energy (eV)3 FIG. 2. Real space probability density for (a) Fe, (b) NV−, (c) Mn, (d) Co, (e) Cr, (f) Ni dopants with any background contribution from the homogeneous diamond crystal removed. The slices are taken in the (110) plane and three atomic layers above the dopant. The logarithmic color scale for all plots is the same, and is in units of the inverse volume of an atomic site. ability near the dopant location and appear less extended than the Co and Ni wave functions. Once the properties of an individual spin center have been determined the exchange interaction between two can be calculated by comparing the energies of filled mid- gap states for parallel and antiparallel alignment of the spin centers[3, 21]. The exchange interaction found be- tween pairs of transition metal spin centers is shown in Fig. 3. For pairs spaced along the [1¯10] direction the Mn- -Mn pair has the largest and slowest-decaying exchange, followed by Cr -- Cr pairs and then Ni -- Ni pairs. The exchange interaction between Cr and Ni appears often smaller than either the Cr -- Cr or Ni -- Ni exchange, which is likely due to the smaller hybridizations of the energy levels of Cr and Ni (relative to homodopant pairs) due to their different energies. Along the [001] direction the Ni- -Ni pair does not decrease logarithmically for the closest pair spacings. The exchange interaction along the [1¯11] interaction is the largest for the Ni -- Ni pair and excluding the Ni -- Ni pair it is the direction for which the exchange interaction between other transition metal pairs is the least. At pair spacings greater than ∼ 2 nm the energy broadening of the calculation (10 µeV) limits the ability to resolve the exchange splittings, and for several pairs of spin centers the exchange interaction is obscured at shorter distances by this broadening. At the first nearest neighbor spacing in the [001] direction and the first and FIG. 3. Magnitude of the exchange interaction for several parings of transition metal spin centers along [001], [1¯10] and [1¯11] denoted by triangles, squares and circles respectively. The four sets of spin center pairs are Mn-Mn (light blue), Ni-Ni (pink), Cr-Cr (gold) and Cr-Ni (purple). second nearest neighbor spacing in the [1¯10] direction the energy broadening in the calculation is on the order of 1 meV and thus the error for these points is larger than the others. The exchange interaction is strongly anisotropic and can vary greatly depending on the direction of in- teraction, the energy of the spin center states as well as the symmetry of the HOMO (which produces the great- est hybridization and splitting), ie e or t2. For all these calculations the strength of the exchange interaction ex- ceeds the dipolar interaction (also shown on Fig. 3) by orders of magnitude. Only for spin center separations in excess of 3 nm would the dipolar interaction become comparable to the exchange interaction. The NV− center exhibits an additional form of ex- change interaction anisotropy, corresponding to the de- pendence of the exchange interaction on the relative orientation of the NV− center atoms themselves. The vacancy and the nitrogen can either be oriented near- parallel to [1¯11] or near-perpendicular to [1¯11]. This in- troduces four orientations for a pair of NV− centers, (1) both near-parallel to [1¯11], (2) both near-perpendicular to [1¯11] and (3) and (4) corresponding to types with one of the pair near-parallel and the other near-perpendicular to [1¯11], pictured in Fig. 4. The choice of near-parallel or near-perpendicular orientation of the NV− center has a large effect on the exchange interaction. Due to the geometry of NV− center pairs along the [1¯10] direction, for some pairs the first nearest neighbor and in one case the second nearest neighbor exchange interactions are not presented due to overlapping impurity potentials. As ex- pected from the symmetry of the different pairs, in some directions there are pairs which have similar exchanges. For example along the [001] direction the exchange in- teractions between the near-parallel near-perpendicular (blue) and near-perpendicular near-parallel (green) over- lay each other in the plot as do the values for the near- FeNV-MnCoCrNi1.07 nm[001][110]1x10-53.2x10-40.010.3210Mn MnCr CrNi NiCr Ni[111][110][001]Exchange Along 0.001 0.01 0.1 10 100 1000 0 5 10 15 20 J (meV)Pair Separation (Å) 1dipoledipole4 exchange interactions, in excess of the dipolar interac- tions between spin centers, even at 2-3 nm separations. Transition metal dopants in diamond offer distinct prop- erties compared to NV− spin centers due to the inclu- sion of d-orbitals and the resulting spin-orbit interaction that permits high-speed electrical control of spin[25] and spin-sensitive optical selection rules. Additionally, based on the exchange between a Ni and Cr dopant pair, one could envision a quantum register where information is transferred to the spin of a Ni spin center and then that information is stored in the less accessible Cr spin. We acknowledge support from an AFOSR MURI. [1] P. M. Koenraad and M. E. Flatt´e, Nature Materials 10, 91 (2011). [2] A. J. Heinrich, J. A. Gupta, C. P. Lutz, and D. M. Eigler, Science 306, 466 (2004). [3] D. Kitchen, A. Richardella, J.-M. Tang, M. E. Flatt´e, and A. Yazdani, Nature 442, 436 (2006). [4] D. M. Toyli, C. D. Weis, G. D. Fuchs, T. Schenkel, and D. D. Awschalom, Nano Letters 10, 3168 (2010). [5] A. A. Khajetoorians, B. Chilian, J. Wiebe, S. Schuwalow, F. Lechermann, and R. Wiesendanger, Nature 467, 1084 (2010). [6] D. H. Lee and J. A. Gupta, Science 330, 1807 (2010). [7] A. A. Khajetoorians, J. Wiebe, B. Chilian, and R. Wiesendanger, Science 332, 1062 (2011). [8] A. Spinelli, B. Bryant, F. Delgado, J. Fern´andez-Rossier, and A. F. Otte, Nature Materials 13, 782 (2014). [9] G. Balasubramanian, P. Neumann, D. Twitchen, M. Markham, R. Kolesov, N. Mizuochi, J. Isoya, J. Achard, J. Beck, J. Tissler, V. Jacques, P. R. Hem- mer, F. Jelezko, and J. Wrachtrup, Nature Materials 8, 383 (2009). [10] K. Ohno, F. Joseph Heremans, L. C. Bassett, B. a. My- ers, D. M. Toyli, A. C. Bleszynski Jayich, C. J. Palm- strom, and D. D. Awschalom, Applied Physics Letters 101, 082413 (2012). [11] B. A. Myers, A. Das, M. C. Dartiailh, K. Ohno, D. D. Awschalom, and A. C. Bleszynski Jayich, Phys. Rev. Lett. 113, 027602 (2014). [12] J. R. Rabeau, Y. L. Chin, S. Prawer, F. Jelezko, T. Gaebel, and J. Wrachtrup, Applied Physics Letters 86, 131926 (2005). [13] I. Aharonovich, S. Castelletto, B. C. Johnson, J. C. Mc- Callum, D. A. Simpson, A. D. Greentree, and S. Prawer, Phys. Rev. B 81, 121201 (2010). [14] F. Jelezko and J. Wrachtrup, Phys Status Solidi A 203, 3207 (2006). [15] T. Chanier, C. E. Pryor, and M. E. Flatt´e, Europhys. Lett. 99, 67006 (2012). [16] T. Chanier, C. Pryor, and M. E. Flatt´e, Phys. Rev. B 86, 085203 (2012). [17] J. R. Weber, W. F. Koehl, J. B. Varley, A. Janotti, B. B. Buckley, C. G. Van de Walle, and D. D. Awschalom, Proc. Nat. Acad. Sci. 107, 8513 (2010). [18] M. W. Doherty, F. Dolde, H. Fedder, F. Jelezko, J. Wrachtrup, N. B. Manson, and L. C. L. Hollenberg, Phys. Rev. B 85, 205203 (2012). FIG. 4. Anisotropy of the exchange interaction for pairs of NV− centers along [001], [1¯10] and [1¯11] denoted by triangles, squares and circles respectively. The inserts are the four real space probability densities representing the different orienta- tions of the NV− centers with respect to [1¯11] direction. They are plotted in the (110) plane containing the centers, for two NV− centers separated by 6.17 A with the same logarithmic color scale used in Fig. 2. perpendicular near-perpendicular (red) and near-parallel near-parallel pairs (black). At the largest spacings the near-parallel near-perpendicular (blue) and near-parallel near-parallel (black) pairs have the largest exchange in- teractions along [1¯11] in direct contrast with the tran- sition metal pairs where the interactions along [11¯1] are in general the smallest. Once again, beyond these pair spacings the exchange interaction is hidden by the 10 µeV broadening included in the homogenous Green's function calculations. The exchange interactions between pairs of transition metal pairs of spin centers and pairs of NV− centers are comparable in magnitude. For all the species and orienta- tions of pairs at the calculated separations the exchange interactions exceed the dipole-dipole interaction between two electrons regardless of dipole orientation. Taking a linear fit to the logarithmic decrease of the exchange in- teraction along the [1¯10] direction, the exchange interac- tion between two Mn equals the dipolar interaction at 47 A; this crossover occurs at roughly 22 A and 25 A for the other transition metal pairs and different orientations of NV− pairs respectively. We have constructed a detailed and accurate theoreti- cal description of NV− and transition-metal point defect spin centers in diamond. The exchange interactions for pairs of transition metal spin centers are on the order, and in some cases, larger than the exchange interaction for pairs of NV− centers. The spin 1 transition metal dopants, Cr -- Cr and Ni -- Ni, show experimentally relevant 1x10-50.0010.010.1110100 0 5 10 15 20 J (meV)Pair Separation (Å)1x10-4Exchange AlongdipoledipoleNV-NV-NV-[111][110][001]NV-[19] J.-M. Jancu, R. Scholz, F. Beltram, and F. Bassani, Phys. Rev. B 57, 6493 (1998). [20] G. F. Koster and J. C. Slater, Phys. Rev. 95, 1167 (1954). [21] J.-M. Tang and M. E. Flatt´e, Phys. Rev. Lett. 92, 047201 (2004). [22] C. E. Moore, Atomic Energy Levels. As Derived From the Analyses of Optical Spectra, Vol. I (National Bureau of Standards, 1949). [23] C. E. Moore, Atomic Energy Levels. As Derived From the Analyses of Optical Spectra, Vol. II (National Bureau of Standards, 1952). [24] C. E. Moore, Atomic Energy Levels. As Derived From the Analyses of Optical Spectra, Vol. III (National Bureau of Standards, 1958). [25] J.-M. Tang, J. Levy, and M. E. Flatt´e, Phys. Rev. Lett. 97, 106803 (2006). 5
1710.06224	1	1710	2017-10-17T11:46:12	Dirac-Surface-State-Dominated Spin to Charge Current Conversion in the Topological Insulator $(Bi_{0.22}Sb_{0.78})_2Te_3$ Films at Room Temperature	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci", "cond-mat.str-el" ]	We report the spin to charge current conversation in an intrinsic topological insulator (TI) $(Bi_{0.22}Sb_{0.78})_2Te_3$ film at room temperature. The spin currents are generated in a thin layer of permalloy (Py) by two different processes, spin pumping (SPE) and spin Seebeck effects (SSE). In the first we use microwave-driven ferromagnetic resonance of the Py film to generate a SPE spin current that is injected into the TI $(Bi_{0.22}Sb_{0.78})_2Te_3$ layer in direct contact with Py. In the second we use the SSE in the longitudinal configuration in Py without contamination by the Nernst effect made possible with a thin NiO layer between the Py and $(Bi_{0.22}Sb_{0.78})_2Te_3$ layers. The spin-to-charge current conversion is attributed to the inverse Edelstein effect (IEE) made possible by the spin-momentum locking in the electron Fermi contours due to the Rashba field. The measurements by the two techniques yield very similar values for the IEE parameter, which are larger than the reported values in the previous studies on topological insulators.	cond-mat.mes-hall	cond-mat	Dirac-Surface-State-Dominated Spin to Charge Current Conversion in the Topological Insulator (Bi0.22Sb0.78)2Te3 Films at Room Temperature J. B. S. Mendes1, O. Alves-Santos2, J. Holanda2, R. P. Loreto1, C. I. L. de Araujo1, Cui-Zu Chang3,4, J. S. Moodera3,5, A. Azevedo2, and S. M. Rezende2 1Departamento de Física, Universidade Federal de Viçosa, 36570-900, Viçosa, MG, Brazil 2Departamento de Física, Universidade Federal de Pernambuco, 50670-901, Recife, PE, Brazil 3Francis Bitter Magnet Lab, Massachusetts Institute of Technology, Cambridge, MA 02139, USA 4Department of Physics, The Pennsylvania State University, University Park, PA16802, USA 5Department of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA We report the spin to charge current conversation in an intrinsic topological insulator (TI) (Bi0.22Sb0.78)2Te3 film at room temperature. The spin currents are generated in a thin layer of permalloy (Py) by two different processes, spin pumping (SPE) and spin Seebeck effects (SSE). In the first we use microwave-driven ferromagnetic resonance of the Py film to generate a SPE spin current that is injected into the TI (Bi0.22Sb0.78)2Te3 layer in direct contact with Py. In the second we use the SSE in the longitudinal configuration in Py without contamination by the Nernst effect made possible with a thin NiO layer between the Py and (Bi0.22Sb0.78)2Te3 layers. The spin-to-charge current conversion is attributed to the inverse Edelstein effect (IEE) made possible by the spin-momentum locking in the electron Fermi contours due to the Rashba field. The measurements by the two techniques yield very similar values for the IEE parameter, which are larger than the reported values in the previous studies on topological insulators. PACS: 72.25.Dc, 72.25.Mk, 75.70.Cn, 75.76.+j, 03.65.Vf Corresponding author: [email protected] 1 Topological Insulators (TIs) constitute a novel state of matter, which have been the subject of intensive investigations in condensed matter physics in the last decade. They are a new class of quantum materials that present insulating bulk, but metallic dissipationless surface states topologically protected by time reversal symmetry, opening several possibilities for practical applications in many scientific arenas including spintronics, quantum computation, magnetic monopoles and highly correlated electron systems [1-4]. More recently, it has been shown that TI-particles behave as optically induced oscillators in an optical tweezers [5]. The surface states are characterized by a single gapless Dirac cone and exhibit the remarkable spin-momentum locking: charge carrier move in such a way that their momenta are always perpendicular to their spin [4,6]. In addition, topological insulators have strong spin-orbit coupling (SOC) and as well have large spin-torque which are essential for efficient spin-charge conversion [7-9]. In turn, the conversion of charge currents into spin currents, and vice versa, are key phenomena for encoding and decoding information carried by electron spins in the active field of spintronics. Until recently, the only known mechanisms for conversion in both directions were the spin Hall effect (SHE) and its Onsager reciprocal, the inverse spin Hall effect (ISHE), that rely on electron scattering processes with spin-orbit interaction in 3D materials [10-14]. Studies of the spin-to-charge conversion by the ISHE have been conducted in metallic films with heavy elements, such as paramagnetic Pt, Pd, and Ta [15-27], ferromagnetic Py [28], antiferromagnetic materials such as IrMn and PtMn [29,30], and semiconductors [31-37]. Recent developments in thin-film growth techniques have made possible the fabrication of samples with atomically flat surfaces and interfaces that have led to the observation of new phenomena induced by SOC in 2D systems [38-40]. Among them are the Edelstein effect, predicted some time ago [41], and its Onsager reciprocal, the inverse Edelstein effect, that enable new means to convert charge currents into spin currents, and vice versa. The direct Edelstein effect and the inverse Edelstein effect (IEE) are made possible by the Rashba effect that arises from SOC and broken inversion symmetry at material surfaces and interfaces [42-45]. The Rashba field produces spin-momentum locking in the electron Fermi contours that enables the conversion between spin and charge currents. The conversion of spin currents produced by ferromagnetic resonance (FMR) spin pumping into charge currents due to the IEE has been observed at Bi/Ag interfaces [46], single-layer graphene [47] and in a few TIs [8,48-50]. In this Rapid Communication, we report the observation of spin-to-charge current conversion by means of the IEE in the topological insulator (Bi0.22Sb0.78)2Te3 at room 2 temperature. The spin currents are generated by two different arrangements, microwave-driven spin pumping and the spin Seebeck effect. The experiments were carried out with two sample structures: i) consisting of the TI grown on a 0.5 mm thick sapphire (0001) substrate and with a Ni81Fe19 (permalloy-Py) top layer, ii) a trilayer in which a NiO layer is grown between the TI and the Py layers. In both, we have used a commercial 0.5 mm thick (0001) sapphire substrate onto which the TI is grown as follows. After high temperature annealing (~800 ºC) of the sapphire substrate, a six-quintuple layer (QL)-thick (BixSb1-x)2Te3 film is grown on top at a temperature ~230 oC in a custom-built ultrahigh vacuum molecular beam epitaxy (MBE) system and capped by a 3-nm-thick epitaxial Te layer. X-ray diffraction patterns confirm the high crystalline, single phase quality of the films, with growth along the c axis (see in the Supplemental Material more details about the conditions of growth and the crystallographic structure of the (Bi,Sb)Te films [51]). We have chosen the Bi concentration x=0.22 to locate the Fermi level close to the Dirac point [52-54]. The Py layer is deposited by DC magnetron sputtering, either directly on the TI film or separated by an insulating NiO layer, grown by RF sputtering at 160 ºC. The Py and NiO films were deposited in a 3 mTorr argon atmosphere in the sputter-up configuration, with the substrate at a distance of 9 cm from the target, and with a deposition rate fixed in 1 Å/s and 0.3 Å/s, respectively. Therefore, the Py and NiO layers were gently deposited over the TI to minimize any detrimental effect on the surface chemistry. Finally, two silver electrodes were attached to the ends of the TI layer for measuring the induced voltages. For the ferromagnetic resonance and spin pumping experiments the sample was mounted on the tip of a PVC rod and inserted through a hole drilled in the center of the back wall of a rectangular microwave cavity operating in the TE102 mode, at a frequency of 9.4 GHz with a Q factor of 2000. The sample is slightly inserted into the cavity in the plane of the back wall, in a position of maximum rf magnetic field and minimum rf electric field to avoid the generation of galvanic effects driven by the electric field. With this arrangement the static magnetic field H and the microwave field are in the film plane and kept perpendicular to each other as the sample is rotated for the measurements of the angular dependence of the FMR spectra and the dc voltage induced by the magnetization precession. Field scan spectra of the derivative of the microwave absorption dP/dH are obtained by modulating the field at 1.2 kHz and using lock-in detection. All FMR and voltage measurements were taken at room temperature. Figure 1(a) shows a schematic illustration of the 6QL (Bi0.22Sb0.78)2Te3/Py (12 nm) bilayer sample used in the SPE experiments, that has length 3 mm and width 1.5 mm. The Py films have 3 rfhin-plane magnetization and thus the magnetic proximity effect is expected to shift the Dirac cone sideways along the momentum direction and does not open an exchange gap (i.e. in our heterostructures, the Dirac cone of the TI film will be preserved). Figure 1(b) shows the FMR absorption spectrum of the Py layer in contact with the TI film measured with microwave power of 24 mW. The FMR line has the shape of a Lorentzian derivative with peak-to-peak linewidth of 38.1 Oe, corresponding to a half-width at half-maximum (HWHM) linewidth of Oe. As shown in the inset of Fig. 1(b), an identical Py layer deposited on a Si substrate has linewidth Oe, showing that the contact of the TI layer produces an additional damping due to the spin pumping process [55,56], similar that observed in Pt/Py bilayers [17,19]. Figure 1(c) shows the field (H) scan dc voltage measured directly with a nanovoltmeter connected by copper wires to the electrodes, for a microwave power of 24 mW, for three angles of the in- plane field. For the voltage lineshape is the superposition of symmetric and antisymmetric components, changes sign with inversion of the field, and vanishes for the field along the sample strip . Figure 1. (color online) (a) Sketch of the bilayer sample 6QL (Bi0.22Sb0.78)2Te3/Py(12 nm) and coordinate system, where  is the in-plane angle. (b) FMR absorption derivative versus magnetic field H measured at 9.4 GHz and microwave power of 24 mW. Inset shows the FMR spectrum for a single Py (12 nm) layer on a Si substrate. (c) Voltage measured between the electrodes for three angles of the in-plane field, with the same microwave frequency and power as in (b). Inset shows the I-V curve of the (Bi0.22Sb0.78)2Te3/Py structure demonstrating the formation of Ohmic contacts between the electrodes. (d) Angular dependence of the symmetric (peak) component of the voltage line. The experimental data are represented by the solid circle symbols and the theoretical fit by the solid curve. 4 0.33H0.28PyHº0º90 The field dependence voltage measured between the electrodes can be described by the sum of two components, , where is the (symmetric) Lorentzian function and is the (antisymmetric) Lorentzian derivative centered about the FMR resonance field . The voltage lineshape measured as function of the field angle can be fit with the expression , (1) where and denote the amplitudes of the symmetric and antisymmetric components of the classical contributions, such as the galvanic effect, or spin rectification, generated in the Py layer [19], and is the peak value of the symmetric contribution to the voltage of quantum origin which will be discussed later. Figure 1 (d) shows the measured angle dependence of the symmetric component, which is the one of interest here, and a solid curve representing the fit obtained with Eq. (1). Figure 2. (color online) (a) Voltage measured between the electrodes at the field angle for several microwave power levels as indicated. (b) and (c) Variation with power of the symmetric and antisymmetric components of the voltage obtained by the fitting of Lorentzian a Lorentzian derivative functions to the lineshapes in (a). (d) Power dependence of the measured symmetric peak component of the voltage at . 5 )(HV)()()(RasymRsymHHDVHHLVHV)(RHHL)(RHHDRHsin2sin)]()([cos)(),(RasymCLRsymCLRpeakQHHDVHHLVHHLVHVsymCLVasymCLVpeakQVº0º0(a)(b)(c)(d)0.40.81.21.6-2-1012345 = 0ºf = 9.4GHz (Bi1-xSbx)2Te3/Py(12nm) 24 mW 19 mW 12 mW 8 mW 5 mW 3 mW 2 mWV (V)H (kOe)-0.4-0.20.00.20.40123 24 mW 19 mW 12 mW 8 mW 5 mW 3 mw 2 mWV (V)H-HR(kOe) (Bi1-xSbx)2Te3/Py(12nm)-0.4-0.20.00.20.4-202 24 mW 19 mW 12 mW 8 mW 5 mW 3 mW 2 mW V (V)H-HR(kOe) (Bi1-xSbx)2Te3/Py(12nm)0510152025012 = 0ºf = 9.4GHzVsym (V)Power (mW)Figure 2(a) shows the voltage lineshapes measured at several power levels and Figs. 2(b) and 2(c) show the corresponding symmetric and antisymmetric components of the lineshapes, obtained by fitting the sum of a Lorentzian function and a Lorentzian derivative to the data. Figure 2 (d) shows that the symmetric component at , which is , exhibits a linear dependence with the microwave power. From Figs. 1(d) and 2(b) we have the value for the voltage of quantum origin at , µV, for a microwave power of 24 mW, part of which is due to the magnonic charge pumping (MCP) that is produced in a single Py layer [57]. In order to separate the contributions, we have measured the voltage induced in a single layer of Py(12 nm) on Si substrate in the same conditions of the data in Figs. (1) and (2). For a microwave power of 24 mW the symmetric component at has a peak value of 0.4 µV. The most important source for the symmetric component of the voltage, and the one of interest here, is the conversion of the spin current produced by spin pumping into charge current in the TI layer. As is well known [12-19], in a ferromagnetic (FM) layer under ferromagnetic resonance, the precessing magnetization generates a spin current density (in units of charge/time area) at the FM/TI interface given by , (2) where e is the electron charge, is the real part of the effective spin mixing conductance at the interface that takes into account the spin-pumped and back-flow spin currents [12-19,55,56], and are, respectively, the frequency and amplitude of the driving microwave magnetic field, and p is the precession ellipticity factor , where is the effective magnetization that appears in the expression for the FMR frequency . The spin current produced by the FMR spin pumping flows through the FM/TI interface into the TI layer. We follow Refs. [8, 44-50] and interpret the spin-to-charge conversion in the TI layer as arising from the inverse Edelstein effect (IEE), that has its origin in the spin-momentum locking in the Fermi contours due to the Rashba SOC interaction. The 3D spin current in Eq. (2) flows into the TI layer and is converted by the IEE into a lateral charge current with a 2D density , where is a coefficient characterizing the IEE, with dimension of length and proportional to the Rashba coefficient, and hence to the magnitude of the SOC [8,44-46]. The measured voltage is related to this current density by , 6 º0peakQVº03.2peakQV0)(82RrfeffSHHLHhgpeJeffgrfh2)42/()4()/(4effReffRMHMHpeffM42/1)]4([effRRMHHfSIEECJej)/2(IEECSIEEjwRVwhere is the shunt resistance, w the width of the (Bi,Sb)Te/Py bilayer in the length of the Py layer and 𝑗𝐶 has units of A/m. In order to obtain the IEE length from the experimental data, we need initially to calculate the SPE spin current. The real part of the spin mixing conductance of the (Bi,Sb)Te/Py interface that enters in Eq. (2) can be inferred from the broadening of the FMR linewidth due to the spin pumping process using [12-14,17,55,56]. With kG, nm, GHz, we find that the additional linewidth of 5 Oe measured in Py due to the contact with the TI layer corresponds to m-2, a value similar to the one for Py/Pt interfaces [12-19]. The amplitude of the microwave field in Eq. (1), in Oersted, is related to the incident power , in watt, by , calculated for a microwave cavity made with a shorted standard X-band rectangular waveguide, operating in the TE102 mode with Q factor of 2000, at a frequency of 9.4 GHz. Using these values, we obtain for mW, , the spin current density at the interface produced by the FMR spin pumping A/m2. The charge current density due to the conversion from the spin current by the IEE, given by , corresponding to the measured voltage of µV, considering that the shunt resistance is approximately the one of the Py layer, Ω, and mm, is A/m. Thus, the IEE length obtained from the spin pumping measurements is nm. The spin pumping origin of the spin current pumped into the TI was confirmed by using a broadband microwave microstrip setup and measuring the voltage between the electrodes on the TI strip with scanning H for several frequencies, keeping constant the incident power at the mW. With this power level, the rf magnetic field produced in the sample placed on a copper microstrip 0.5 mm wide with characteristic impedance Ω, is very similar to the one in the microwave cavity with Q=2000 with power 24 mW previously described. The field dependencies of the voltages measured at several frequencies are shown in the lower panel of Fig. 3 with Lorentzian fits. One clearly see the broadening of the voltage lines with increasing frequency, which is a characteristic feature of the spin pumping damping [55,56]. The upper panel in Fig. 3 shows the variation of the measured field value for the peak voltage, that is the FMR field , with the driving frequency. In order to compare the voltages measured at the various 7 SR)()/4(0PyPyeffHHtMg1140M12Pyt4.92/19100.1effgiP2/1)(776.1irfPh24iPRHH5103.2SJ)/(SpeakIEECRwVj9.1peakIEEV71SR5.1w5107.1CjSCIEEJj/075.0IEE28iP500ZRHfrequencies we need to consider that the peak value of the pumped spin current varies inversely with the FMR linewidth squared, , as in Eq. (2). Thus we introduce the normalized peak voltage, defined by , where V and are, respectively, the peak voltage and the linewidth measured at a frequency f, and is the linewidth at 5 GHz. The linear increase of the normalized peak voltage shown in the inset of the upper panel of Fig. 3 provides another evidence of the spin pumping origin of voltage induced in the TI layer. Figure 3. (color online) Lower panel: the voltage measured between the electrodes on the TI strip at the field angle for several microwave driving frequencies as indicated, and power 28 mW. Upper panel: Driving frequency versus the field value for the peak voltage. Inset shows the variation with frequency of the peak voltage normalized by the FMR linewidth squared referred to the value at 5 GHz. In order to confirm the spin-to-charge current conversion in (Bi1-xSbx)2Te3 by the IEE, we have used another process to generate spin currents, namely, the spin Seebeck effect (SSE). We used a sample arrangement illustrated Fig. 4(a), in which a 5 nm thick NiO layer provides electrical isolation between the TI and Py films. The Py layer has width 1.0 mm, smaller than the NiO and TI layers, to avoid possible contacts at the edges. A commercial Peltier module, of width 4 mm, is used to heat or cool the side of the Py layer while the substrate is maintained in thermal contact with a copper block at room temperature. The temperature difference ΔT across the sample is calibrated as a function of the current in the Peltier module by means of a differential thermocouple. The temperature gradient perpendicular to the Py layer has two effects: One is to generate a voltage along the layer by means of the classical anomalous Nernst effect (ANE) [58- 8 2H25)/(HHVVH5Hº00.250.500.751.001.251.500.00.51.01.5 5 GHzSapphire/(Bi,Sb)Te/Py(12 nm)V (V)Magnetic field (kOe)6810 11 GHz9 GHz7 GHzFrequency (GHz)4681012012345V (V)Frequency (GHz)61]; the other is to generate a spin current across the Py layer by the longitudinal spin Seebeck effect [61-66]. Since NiO is a room temperature antiferromagnet, it blocks the flow of charge current but transports spin currents [66-69], thus allowing the measurements of the voltage generated in the TI layer separated from the voltage induced in the Py layer by the ANE. Figure 4(b) shows the voltages measured between the two electrodes in the (Bi1-xSbx)2Te3 layer that are produced by the electric current resulting from the IEE spin-to-charge conversion of the spin current generated by the spin Seebeck effect in the Py layer that is injected into the Py/NiO interface and transported by the magnons in the NiO layer. The magnetic field dependencies of the SSE-IEE voltages in the TI layer for several values of the temperature difference across the sample structure are shown in Fig. 4 (b). Note that corresponds to the Peltier module warmer than the substrate. The data have the shape of the hysteresis curve of Py with very small coercitivity in the field scale of the measurements. The change in the voltage sign with the field reversal is due to the sign change of the spin polarization. Figure 4 (c) shows the measured variation of the voltage plateau with the temperature difference for applied fields of kOe. The linear dependence of Figure 4. (color online) (a) Schematic illustration of the sample 6QL (Bi0.22Sb0.78)2Te3/NiO/Py(12 nm) structure used to measure the voltages generated by the ANE and SSE. (b) Variation with magnetic field of the SSE-IEE voltage measured in the TI layer, created by the combined SSE in the Py layer and IEE in the TI, for several values of ΔT as indicated. Positive corresponds to the Peltier module warmer than the substrate. (c) Variation with temperature difference of the SSE-IEE voltage measured with H=0.4 kOe, in two field directions. 9 T0TT4.0HSSEVT(a)-606-606-606-606-400-2000200400-606 T = 12 K T = 10 K VSSE (V) T = 6 K T = 0 K H (Oe) T = -6 K-404812-6-4-20246H=-0.4 kOeVSSE (V)T (K)H=+0.4 kOe(b)(c)on results from the fact that the spin current generated by the LSSE in Py is proportional to the temperature gradient across the Py layer [66]. In order to calculate the IEE parameter obtained from the SSE-IEE measurements we use data for the SSE in Si/Py/NiO/Pt in Ref. [66] and rescale the numbers accordingly. The value of the charge current measured in the Pt layer in the sample Si/Py(20 nm)/NiO(5 nm)/Pt (6 nm) for K is 24.3 nA, corresponding to a current density in the Pt layer of width 2.5 mm and thickness 6 nm of A/m2 [66]. Considering for the spin Hall angle of Pt the value [12-14], this gives for the spin current density reaching the Pt after attenuation in the NiO layer the value A/m2 for K. However, we must consider that the temperature gradients in the Py layer in the Si/Py/NiO/Pt and in the Al2O3/TI/NiO/Py samples are different. The temperature gradient across the Py layer in the Si/Py/NiO/Pt sample is given by , where and are the thermal conductivities of Py and Si, and is the sample thickness. For K, mm, W/(K m), W/(K m), K/cm. On the other hand, in the Al2O3/TI/NiO/Py sample, the gradient corresponding to a temperature difference is , which gives for K, mm, W/(K m), K/cm. Thus, the same K would generate in the TI layer a spin current density A/m2. From the data in Fig. 4 (c), for K the voltage in the TI layer is V, which, for a resistance of Ω of the TI in the length of the Peltier module in a width of 1 mm corresponds to a 2D current density of A/m. In turn, the resistance between the Py film and the TI layer is R ≥ 1.0×106 Ω. Therefore, the resistance measurements give us assurance that the NiO layer (5nm) is a good electrical insulator. Thus, the IEE parameter obtained from the SSE measurements is nm, which is nearly the same obtained from the SPE measurements. In summary, we have demonstrated the conversion of a spin current into a charge current in the topological insulator (Bi0.22Sb0.78)2Te3 at room temperature, that is attributed to the inverse Edelstein effect (IEE) made possible by the spin-momentum locking in the electron Fermi contours due to the Rashba field. The spin currents were generated in a thin layer of permalloy by two different processes, spin pumping (SPE) and spin Seebeck effects (SSE). In the former we have used microwave-driven ferromagnetic resonance of the Py film to generate a spin current that is injected into the TI film in direct contact with Py. In the latter we have used the SSE in the 10 T12T3106.1CJ05.0SH4102.3SJ12T)/()/(SPySiPytTKKTPyKSiKSt12T4.0St148SiK4.46PyK957PyTT)/()/(32SPyOAlPytTKKT12T5.0St9.4132OAlK217PyT12T3102.7SJ12T9.5SSEV41008.1R71046.5CjSCIEEJj/076.0IEElongitudinal configuration in Py with no contamination by the Nernst effect made possible with the use of a thin NiO layer between the Py and TI layers. The results of the two measurements yield nearly identical values for the IEE coefficient, 0.076 nm, which is about twice the value measured for the topological insulator Bi2Se3 (6 QL) and about four times larger than the value for (Bi,Sb)2Te3 (6 QL) [50]. ACKNOWLEDGEMENTS This research was supported in Brazil by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Financiadora de Estudos e Projetos (FINEP), Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG), and Fundação de Amparo à Ciência e Tecnologia do Estado de Pernambuco (FACEPE). CZC and JSM acknowledge the support from NSF Grants No. DMR-1207469, DMR-1700137, No. DMR- 0819762 (MIT MRSEC), ONR Grant No. N00014-16-1-2657, and the STC Center for Integrated Quantum Materials under NSF Grant No. DMR-1231319. CZC thanks the support from the startup provided by Penn State. References [1] C. Nayak et al., Rev. Mod. Phys. 80, 1083 (2008). [2] X-L. Qi, R. Li, J. Zang, S-C. Zhang, Science 323, 1184 (2009); W.H. Campos, W.A. Moura- Melo, and J.M. Fonseca, Phys. Lett. A, 381, 413-562 (2017). [3] L. Fu, and C. L. Kane, Phys. Rev. Lett. 100, 096407 (2008). [4] M. Z. Hasan, and C. L. Kane, Rev. Mod. Phys. 82, 3045 (2010). [5] W. H. Campos, J. M. Fonseca, V. E. de Carvalho, J. B. S. Mendes, M. S. Rocha, W. A. Moura-Melo, arXiv:cond-mat/1703.04556 (2017). [6] X. L. Qi, and S. C. Zhang, Rev. Mod. Phys. 83, 1057 (2011). [7] C. H. Li, O. M. J. van't Erve, J. T. Robinson, Y. Liu, L. Li, and B. T. Jonker, Nat. Nanotechnol. 9, 218 (2014). [8] J.-C. Rojas-Sanchez, S. Oyarzún, Y. Fu, A. Marty, C. Vergnaud, S. Gambarelli, L. Vila, M. Jamet, Y. Ohtsubo, A. Taleb-Ibrahimi, P. Le Fèvre, F. Bertran, N. Reyren, J.-M. George, and A. Fert, Phys. Rev. Lett. 116, 096602 (2016); A. R. Mellnik et al., Nature 511, 449-451 (2014). [9] Y. Fan et al., Nat. Nanotechnol. 11, 352 (2016). [10] M. I. D'yakonov and V. I. Perel, JETP Lett. 13,467 (1971). 11 [11] J. E. Hirsch, Phys. Rev. Lett. 83, 1834 (1999). [12] A. Hoffmann, IEEE Trans. Mag. 49, No. 10, 5172 (2013). [13] S. Maekawa, H. A. Adachi, K. Uchida, J. Ieda, and E. Saitoh, J. Phys. Soc. Japan 82, 102002 (2013). [14] J. Sinova, S. O. Valenzuela, J. Wunderlich, C. H. Back, and T. Jungwirth, Rev. Mod. Phys. 87, 1231 (2015). [15] A. Azevedo, L. H. Vilela-Leão, R. L. Rodríguez-Suárez, A. B. Oliveira, and S. M. Rezende, J. Appl. Phys. 97, 10C715 (2005). [16] E. Saitoh, M. Ueda, H. Miyajima, and G. Tatara, Appl. Phys. Lett. 88, 182509 (2006). [17] O. Mosendz, J. E. Pearson, F. Y. Fradin, G. E. W. Bauer, S. D. Bader, and A. Hoffmann, Phys. Rev. Lett. 104, 046601 (2010). [18] C. W. Sandweg, Y. Kajiwara, K. Ando, E. Saitoh, and B. Hillebrands, Appl. Phys. Lett. 97, 252504 (2010). [19] A. Azevedo, L. H. Vilela-Leão, R. L. Rodríguez-Suárez, A. F. Lacerda Santos, and S. M. Rezende, Phys. Rev. B 83, 144402 (2011). [20] L. H. Vilela-Leão, C. Salvador, A. Azevedo, and S. M. Rezende, Appl. Phys. Lett. 99, 102505 (2011). [21] L. Liu, C.-F. Pai, Y. Li, H. W. Tseng, D. C. Ralph, and R. A. Buhrman, Science 55, 336 (2012). [22] C. Hahn, G. de Loubens, O. Klein, M. Viret, V. V. Naletov, and J. Ben Youssef, Phys. Rev. B 87, 174417 (2013). [23] M. Weiler et. al., Phys. Rev. Lett. 111, 176601 (2013). [24] V. Castel, N. Vlietstra, B. J. van Wees, and J. Ben Youssef, Phys. Rev. B 90, 214434 (2014). [25] D. Wei, M. Obstbaum, M. Ribow, C. H. Back, and G. Woltersdorf, Nat. Comm. 5, 3768 (2014). [26] L. Bai, M. Harder, Y. P. Chen, X. Fan, J. Q. Xiao, and C.-M. Hu, Phys. Rev. Lett. 114, 227201 (2015). [27] C. Du, H. Wang, P. Chris Hammel, and F. Yang, J. Appl. Phys. 117, 172603 (2015). [28] P. Hyde, L. Bai, D. M. J. Kumar, B. W. Southern, C.-M. Hu, S. Y. Huang, B. F. Miao, and C. L. Chien, Phys. Rev. B 89, 180404(R) (2014). [29] J. B. S. Mendes, R. O. Cunha, O. Alves Santos, P. R. T. Ribeiro, F. L. A. Machado, R. L. Rodríguez-Suárez, A. Azevedo, and S. M. Rezende, Phys. Rev. B 89, 140406(R) (2014). 12 [30] W. Zhang, M. B. Jungfleisch, W. Jiang, J. E. Pearson, A. Hoffmann, F. Freimuth, and Y. Mokrousov, Phys. Rev. Lett. 113, 196602 (2014). [31] K. Ando, S. Takahashi, J. Ieda, H. Kurebayashi, T. Trypiniotis, C. H. W. Barnes, S. Maekawa, and E. Saitoh, Nat. Mat. 10, 655 (2011). [32] K. Ando and E. Saitoh, Nat. Comm. 3, 629 (2012). [33] E. Shikoh, K. Ando, K. Kubo, E. Saitoh, T. Shinjo, and M. Shiraishi, Phys. Rev. Lett. 110, 127201 (2013). [34] J.-C. Rojas-Sánchez et. al., Phys. Rev. B 88, 064403 (2013). [35] S. Dushenko, M. Koike, Y. Ando, T. Shinjo, M. Myronov, and M. Shiraishi, Phys. Rev. Lett. 114, 196602 (2015). [36] F. Bottegoni et. al., Phys. Rev. Lett. 118, 167402 (2017). [37] J. B. S. Mendes, S. L. A. Mello, O. Alves Santos, R. O. Cunha, R. L. Rodríguez-Suárez, A. Azevedo, and S. M. Rezende, Phys. Rev. B 95, 214405 (2017). [38] A. Soumyanarayanan, N. Reyren, A. Fert, and C. Panagopoulos, Nature 539, 509 (2016). [39] Y. Ando and M. Shiraishi, J. Phys. Soc. Japan 86, 011001 (2017). [40] F. Hellman et al., Rev. Mod. Phys. 89, 025006 (2017). [41] V. M. Edelstein, Solid State Commun. 73, 233 (1990). [42] E. I. Rashba, Sov. Phys. Solid State 2, 1109 (1960). [43] A. Manchon, H. C. Koo, J. Nitta, S. M. Frolov, and R. A. Duine, Nat. Mater. 14, 871 (2015). [44] K. Kondou, R. Yoshimi, A. Tsukazaki, Y. Fukuma, J. Matsuno, K. S. Takahashi, M. Kawasaki, Y. Tokura, and Y. Otani, Nat. Phys. 12, 1027 (2016). [45] S. Zhang and A. Fert, Phys. Rev. B 94, 184423 (2016). [46] J. C. Rojas Sánchez, L. Vila, G. Desfonds, S. Gambarelli, J. P. Attané, J. M. De Teresa, C. Magén, and A. Fert, Nat. Commun. 4, 2944 (2013). [47] J. B. S. Mendes, O. Alves-Santos, L. M. Meireles, R. G. Lacerda, L. H. Vilela-Leão, F. L. A. Machado, R. L. Rodríguez-Suárez, A. Azevedo, and S. M. Rezende, Phys. Rev. Lett. 115, 226601 (2015). [48] Y. Shiomi, K. Nomura, Y. Kajiwara, K. Eto, M. Novak, K. Segawa, Y. Ando, and E. Saitoh, Phys. Rev. Lett. 113, 196601 (2014). [49] A. A. Baker, A. I. Figueroa, L. J. Collins-McIntyre, G. van der Laan, and T. Hesjedal, Sc. Rep. 5, 7907 (2015). [50] H. Wang et al., Phys. Rev. Lett. 117, 076601 (2016). 13 [51] See the Supplemental Information at [URL will be inserted by publisher] for more details about the conditions of growth, transport properties and the crystallographic structure of the (Bi,Sb)Te films. [52] J. Zhang et al., Nat. Commun. 2, 574 (2011). [53] Z. Jiang, C-Z. Chang, M. R. Masir, C. Tang, Y. Xu, J. S. Moodera, A. H. MacDonald, and J. Shi, Nat. Commun. 7, 11458 (2016). [54] Z. Jiang, C-Z. Chang, C. Tang, J-G. Zheng, J. S. Moodera, and J. Shi, AIP Advances 6, 055809 (2016). [55] Y. Tserkovnyak, A. Brataas, G. E. W. Bauer, Phys. Rev. B 66, 224403 (2002). [56] Y. Tserkovnyak, A. Brataas, G. E. W. Bauer, and B. I. Halperin, Rev. Mod. Phys. 77, 1375 (2005). [57] A. Azevedo, R. O. Cunha, F. Estrada, O. Alves Santos, J. B. S. Mendes, L. H. Vilela-Leão, R. L. Rodríguez-Suárez, and S. M. Rezende, Phys. Rev. B 92, 024402 (2015). [58] S. Y. Huang, W. G. Wang, S. F. Lee, J. Kwo, and C. L. Chien, Phys. Rev. Lett. 107, 216604 (2011). [59] A. Slachter, F. L. Bakker, and B. J. van Wess, Phys. Rev. B 84, 020412 (2011). [60] A. D. Avery, M. R. Pufall, and B. L. Zink, Phys. Rev. Lett. 109, 196602 (2012). [61] G. E. W. Bauer, E. Saitoh, and B. J. van Wees, Nature Mat. 11, 391 (2012). [62] K. Uchida, S. Takahashi, K. Harii, J. Ieda, W. Koshibae, K. Ando, S. Maekawa, and E. Saitoh, Nature 455, 778 (2008). [63] H. Adachi, K. Uchida, E. Saitoh, and S. Maekawa, Rep. Prog. Phys. 76, 036501 (2013). [64] S. R. Boona, R. C. Myers, and J. P. Heremans, Energy Environ. Sci. 7, 885 (2014). [65] K. Uchida, H. Adachi, T. Kikkawa, A. Kirihara, M. Ishida, S. Yorozu, S. Maekawa, and E. Saitoh, Proc. IEEE 104, 1946 (2016). [66] J. Holanda, O. Alves-Santos, R. O. Cunha, J. B. S. Mendes, R. L. Rodríguez-Suárez, A. Azevedo, and S. M. Rezende, Phys. Rev. B 95, 214421 (2017). [67] H. Wang, C. Du, P. Chris Hammel, and F. Yang, Phys. Rev. Lett. 113, 097202 (2014). [68] C. Hahn, G. de Loubens, V. V. Naletov, J. Ben Youssef, O. Klein, and M. Viret, Europhys. Lett. 108, 57005 (2014). [69] S. M. Rezende, R. L. Rodríguez-Suárez, and A. Azevedo, Phys. Rev. B 93, 054412 (2016). 14 Supplementary information: Dirac-Surface-State-Dominated Spin to Charge Current Conversion in the Topological Insulator (Bi0.22Sb0.78)2Te3 Films at Room Temperature J. B. S. Mendes1*, O. Alves-Santos2, J. Holanda2, R. P. Loreto1, C. I. L. de Araujo1, Cui-Zu. Chang3,4, J. S. Moodera3,5, A. Azevedo2, and S. M. Rezende2 1Departamento de Física, Universidade Federal de Viçosa, 36570-900, Viçosa, MG, Brazil 2Departamento de Física, Universidade Federal de Pernambuco, 50670-901, Recife, PE, Brazil 3Francis Bitter Magnet Lab, MIT, Cambridge, Massachusetts 02139, USA 4Department of Physics, The Pennsylvania State University, University Park, PA16802, USA 5Department of Physics, MIT, Cambridge, Massachusetts 02139, USA I. The growth of topological insulator (Bi1-xSbx)2Te3 thin films The topological insulator (TI) thin film growth was performed using a custom-built ultrahigh vacuum molecular beam epitaxy (MBE) system. The insulating heat-treated sapphire (0001) was used as substrate and has been outgassed at 800°C for one hour before the deposition. The Bi, Sb and Te were evaporated from Knudsen effusion cells using high purity Bi(99.999%), Sb(99.9999%) and Te (99.9999%). During the growth, the substrate was maintained at 230 °C. To reduce Te vacancies, the growth is kept in Te-rich condition (the flux ratio of Te per (Bi +Sb) was set to approximately ∼10). The Bi, Sb and Te concentrations in the films were determined by their ratio, obtained in situ during growth using properly calibrated quartz crystal monitors. The growth rate for the TI films was ~0.2 quintuple layers (QLs) per minute. Following the growth, the TI films were annealed at 230℃ for 30 minutes to improve the crystal quality before being cooled down to room temperature. The sharp and streaky 1×1 reflective high energy electron diffraction (RHEED) patterns (Figs. S1a and S1b) indicate the high single crystalline quality of 6QL (Bi0.22Sb0.78)2Te3 film grown on sapphire (0001) substrate. To avoid possible contamination, a 3-nm thick epitaxial Te capping layer was deposited at room temperature on top of the TI films before taking the samples out of the MBE chamber for the device fabrications. 15 Fig. S1. (color online) RHEED patterns of 6 QL (Bi0.22Sb0.78)2Te3 grown on sapphire (0001) with the electron beam parallel to along (a) and (b). II. X-ray diffraction of the TI films The crystallographic structure of the (Bi0.22Sb0.78)2Te3 film was assessed by X-ray diffraction (XRD) measurements. The XRD spectrum at high resolution detailing the position of the peaks of the TI film is shown in Fig. S2. The diffraction patterns were obtained at angles between 5° and 70° (2θ). The XRD pattern indicates that the reflections are only from (00l) family of planes of (Bi1-xSbx)2Te3, with the sapphire (006) peak or with the (001) peak of the Te capping layer, suggesting that no impurity phase is present. The inset in Fig. S1 shows XRD rocking measured at 2θ=17.24°, corresponding to the (Bi0.22Sb0.78)2Te3 (006) peak. A Gaussian fit to the spectrum gives a full width at half maximum (FWHM) of only 0.07°, demonstrating that a perfectly aligned domain epitaxial film has grown. The XRD patterns were recorded using the Bruker D8 Discover Diffractometer equipped with the Cu Kα radiation (λ=1.5418Å). 16 [1110][2110] Figure S2. (color online) X-ray diffraction result of a typical 6 QL (Bi0.22Sb0.78)2Te3 grown on sapphire (Al2O3). The inset shows the XRD rocking curve, corresponding to the (Bi,Sb)Te (006) peak that gives a FWHM of 0.07°. The results of XRD measurements indicate that the present TI film is epitaxially grown on the sapphire substrate. III. Transport properties of TI films Based on previous work we have chosen the Bi concentration x=0.22 to locate the Fermi level close to the Dirac point and so the resistivity in the bulk part should show insulating behavior. On this point, it is important to mention that the results involving the variation of Bi concentration to locate the Fermi level close to the Dirac point, have been explored in previous publications [52, 53 and 54]. In these previous publications, transport measurements (including resistivity, ordinary Hall measurements and anomalous Hall conductance) were also extensively explored. Figure S3 displays the temperature dependence of the longitudinal resistance, Rxx, for the TI-samples used in this manuscript. The resistance data indicate insulator-like behaviour for the samples. In the measure, Rxx increases with decreasing temperature in almost the entire temperature range, reflecting the depletion of bulk carriers. 17 10203040506070101102103104105106107 (6QL) (Bi0.22Sb0.78)2Te3 /SapphireTe (0 0 1)(0 0 9) (0 0 3)Al2O3(0 0 6)(0 0 6)(0 0 21)(0 0 18)(0 0 15)Intensity(counts/sec)2 (deg)8.58.68.78.8 (Bi,Sb)Te (0 0 6)FWHM 0.07°Intensity(counts/sec)(deg) Figure S3. (color online) Temperature dependence of longitudinal resistance Rxx 6 QL (Bi0.22Sb0.78)2Te3 sample. 18 0501001502002503005678910(Bi,Sb)2Te3 RXX(k)T(K)
1704.04352	2	1704	2018-05-08T08:52:09	High mobility dry-transferred CVD bilayer graphene	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	We report on the fabrication and characterization of high-quality chemical vapor-deposited (CVD) bilayer graphene (BLG). In particular, we demonstrate that CVD-grown BLG can mechanically be detached from the copper foil by an hexagonal boron nitride (hBN) crystal after oxidation of the copper-to-BLG interface. Confocal Raman spectroscopy reveals an AB-stacking order of the BLG crystals and a high structural quality. From transport measurements on fully encapsulated hBN/BLG/hBN Hall bar devices we extract charge carrier mobilities up to 180,000 cm$^2$/(Vs) at 2 K and up to 40,000 cm$^2$/(Vs) at 300 K, outperforming state-of-the-art CVD bilayer graphene devices. Moreover, we show an on-off ration of more than 10,000 and a band gap opening with values of up to 15 meV for a displacement field of 0.2 V/nm in such CVD grown BLG.	cond-mat.mes-hall	cond-mat	High mobility dry-transferred CVD bilayer graphene Michael Schmitz,1 Stephan Engels,1, 2 Luca Banszerus,1 Kenji Watanabe,3 Takashi Taniguchi,3 Christoph Stampfer,1, 2 and Bernd Beschoten1 1)2nd Institute of Physics and JARA-FIT, RWTH Aachen University, 52074 Aachen, Germany 2)Peter Grunberg Institute (PGI-9), Forschungszentrum Julich, 52425 Julich, Germany 3)National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan (Dated: 9 May 2018) We report on the fabrication and characterization of high-quality chemical vapor- deposited (CVD) bilayer graphene (BLG). In particular, we demonstrate that CVD- grown BLG can be detached mechanically from the copper foil by an hexago- nal boron nitride (hBN) crystal after oxidation of the copper-to-BLG interface. Confocal Raman spectroscopy reveals an AB-stacking order of the BLG crystals and a high structural quality. From transport measurements on fully encapsu- lated hBN/BLG/hBN Hall bar devices we extract charge carrier mobilities up to 180,000 cm2/(Vs) at 2 K and up to 40,000 cm2/(Vs) at 300 K, outperforming state-of-the-art CVD bilayer graphene devices. Moreover, we show an on-off ratio of more than 10,000 and a band gap opening with values of up to 15 meV for a displacement field of 0.2 V/nm in such CVD grown BLG. Bilayer graphene (BLG) is a unique two-dimensional material hosting massive chiral quasi-particles,1 which give rise to a number of interesting phenomena. This includes, for example, an unconventional quantum Hall effect,2 a tunable Lifshitz transition3 or topo- logically protected valley transport.4,5 Most importantly, BLG offers high carrier mobility and the possibility of opening a tunable band gap by breaking the intrinsic inversion sym- metry using a perpendicular electric field,6–8 (the so-called displacement field) making this material particularly interesting for electronic applications. While most of the experimental work has been employed by using mechanical exfoliation of BLG from natural graphite, there are only a few reports on implementing synthetically grown BLG into high performance devices.9,10 Chemical vapor deposition (CVD) is among the most promising growth methods when aiming for large-scale applications. The rapid advancement in this field has recently lead to the growth of continuous single-layer graphene (SLG) sheets as well as individual flakes of arbitrary lateral sizes ranging up to centime- ter scale.11–14 For years, the electronic properties of the CVD-grown SLG and BLG could hardly compete with high-quality mechanically exfoliated samples.15,16 The main reasons for the striking difference in device performance are material degradations and contamina- tions caused by wet transfer techniques which typically have been used for separating CVD graphene from the catalyzing metal surface.17 The direct exposition of graphene to wet chemicals such as etchants and solvents introduce defects, ripples and inhomogene doping in graphene and thereby results in minor electronic performance when incorporated into transport devices.17 Recently, we introduced a dry transfer technique where van-der-Waals forces are used to directly pick-up CVD-grown SLG from the metal surface by hexagonal boron-nitride (hBN) stamps.18,19 Using this method the SLG exhibit a high structural qual- ity and ultra-high carrier mobilities up to 3 × 106cm2/(Vs). This dry transfer method has not yet been demonstrated for CVD-grown BLG. In this Letter, we present the fabrication of high-quality CVD-grown BLG fully encapsu- lated in hBN with charge carrier mobilities matching the largest values reported so far for exfoliated BLG at room temperature.20 Several hBN/BLG/hBN heterostructures are fab- ricated by a dry transfer method where the BLG crystals are mechanically detached from the underlying Cu-foil. Confocal Raman spectroscopy on encapsulated structures proves the AB-stacking order of the CVD-grown BLG and demonstrates its high structural quality in terms of low strain inhomogeneities. Low temperature transport measurements of a top- 2 FIG. 1. (a) Optical image of CVD-grown SLG crystals on a Cu foil with a BLG area in their centers. The scale bar represents 100 µm. (b) Schematic of mechanical delamination process. (c) Optical image of contacted hBN/BLG/hBN Hall bar before (left panel) and after (right panel) the top gate deposition (scale bar is 4 µm). gated Hall bar device reveal ballistic transport thanks to its high charge carrier mobility and allow exploring the formation of the band gap when applying a perpendicular displacement field. We use low pressure CVD growth of graphene on the inside of copper foil enclosures following the same recipe as reported by us for SLG.12,18 Using this recipe, up to 10 % of the graphene flakes exhibit aligned bilayer parts in their centers grown underneath the SLG crystals (see optical image in Fig. 1(a)) with lateral sizes varying between 20 and 100 µm. After growth, the copper surface is left under ambient conditions for at least five days which results in an oxidation of the copper-to-graphene interface, weakening the respective adhesion force. We note that the oxidation process for BLG crystals is signifi- cantly slower than for SLG crystals. For detaching of the graphene crystals we fabricate a transfer stamp by covering a glass slide with scotch tape and a layer of poly(vinyl-alcohol) (PVA) and poly(methyl-methacrylate) (PMMA). Subsequently, we exfoliate hBN crystals onto this stack. The PVA/PMMA is cut around a desired flake, detached from the glass slide and rested on a poly(dimethylsiloxane) (PDMS) block. Using a mask aligner, the hBN/PMMA/PVA/PDMS stamp is brought into contact with the CVD-grown BLG and heated for one minute at 125 ◦C. A schematic of the pick-up process is illustrated in Fig. 1(b). After mechanical pick up of the BLG from the copper foil the hBN/BLG stack is placed on another hBN flake exfoliated on Si/SiO2 which was cleaned beforehand by an oxygen plasma (base pressure of 0.1 mbar, 600 W for 300 s) and the PDMS block is thereafter removed. The resulting hBN/BLG/hBN heterostructure on SiO2 is cleaned by in DI water acetone and isopropanol. The inset in Fig. 2(a) shows an optical false color image of a typ- ical hBN/graphene/hBN heterostructure where the bilayer part is highlighted by a black dashed line. We first characterize the hBN/BLG/hBN heterostructures by confocal Raman spec- troscopy using a laser excitation wavelength of 532 nm. Fig. 2(a) shows a typical Raman spectrum detected on one of our structures recorded at the cross as displayed in the optical image of the inset. The Raman peak originating from the hBN is at ωhBN = 1366 cm−1.21 Moreover, the graphene G-peak is located at ωG = 1583 cm−1 and exhibits a full-width- at-half-maximum (FWHM) of ΓG = 13 cm−1 indicating an low overall doping.22,23 The hBN stampLGBSLGSLGBLGABBLGCu(a)(b)(c)TG3 graphene 2D-peak at around 2700 cm−1 exhibits the well-known BLG sub-peak structure.24 Fig. 2(b) shows a close-up of the 2D-peak where four Lorentzians are needed to fit the spec- trum which proves the AB-stacking of our CVD-BLG.25 Furthermore, as a result of the low FWHM of the individual sub-peaks, we observe distinct minima between the left peaks (see arrows in Fig. 2(b)), which suggests low nanometer-scale strain variations in our CVD BLG.26–29. To gain better statistical information on the strain variations over the BLG crystal, we record a high resolution Raman map in the area marked by the red box in Fig. 2(a). For each spectrum we fit both the peak position and the peak width of all four sub peaks of the 2D-peak. The result is summarized in a scatter plot in Fig. 2(c) where we plot respective Γ2D values for each sub-peak as function of their peak positions. The obtained distributions are in good agreement with data reported for high-quality exfoliated BLG28. Together with the spatial homogeneity seen in the Raman map of the 2D line width (inset of Fig. 2(a)) this underlines the high structural quality of the sample. For magneto-transport experiments, we pattern the hBN/BLG/hBN heterostructure into a Hall bar using standard electron beam lithography, reactive ion etching and metal evapo- ration. The latter is used to fabricate one-dimensional side contacts (see Ref. 18 for details). The left panel of Fig. 1(c) shows a contacted Hall bar prior to the deposition of the top gate. For fabricating the top gate, another hBN flake is transferred onto the device and a Cr/Au top gate is deposited over the entire structure. A finished device is shown in the lower panel of Fig. 1(c). We note that the top gate exceeds the outer edges of the Hall bar transport channel to suppress inhomogeneous gate tuning along the edges of the BLG. To characterize the tuning behavior of the Hall bar structure by means of low temperature transport measurements we record the two-terminal resistance R2T as a function of both top gate voltage (VTG) and back gate voltage (VBG) where the latter is applied across a 285 nm thick SiO2 layer to the underlying highly doped Si wafer. The resulting two-dimensional map is shown in Fig. 3(a). A line of elevated resistance (white dotted line) can be tuned by both VTG and VBG. Along this line the vertical displacement field D is tuned at n = 0 cm−2. Every perpendicular path will change n at D = const. In a simple plate capacitor model, the displacement field equals to D = e 20 [αBG(VBG − VBG,0) − αTG(VTG − VTG,0)] (1) with the charge carrier density being n = αBG(VBG − VBG,0) + αTG(VTG − VTG,0), (2) where αBG = 6.65×1010 cm−2V−1 and αTG = 31.37×1010 cm−2V−1 are the respective gate lever arms which are extracted from independent measurements of the Hall effect and the corresponding offset voltages VBG,0 = 7.500 V and VTG,0 = -0.2212 V result from the finite doping of the sample. Using the lever arm we estimate the total thickness of the top gate dielectrics to be ≈ 68 nm. The device shows an on-off ratio of more than 10,000 which is comparable to that achieved for exfoliated BLG.30 We now focus on the transport properties at zero displacement field D = 0 V/nm. The inset of Fig. 3(a) shows the four-terminal BLG resistance R measured between contacts A and B (see Fig. 1(d)) as function of n at the two temperatures of 2 (blue) and 300 K (purple), where VBG and VTG are simultaneously varied along the red dashed line in Fig. 3(a). The maximum resistance is reached at the charge neutrality point (CNP).31 Fig. 3(b) shows the corresponding conductivity σ = l/(wR) for a larger set of temperatures between 2 and 300 K, where l = 2 µm is the length between the probing contacts A and B and w = 1 µm is the width of the Hall bar. From the n-dependence of σ we extract the charge carrier mobilities by performing fits to the Drude formula σ = enµ. At 2 K, we find µ = 180, 000 cm2/(Vs) for electrons and µ = 80, 000 cm2/(Vs) for holes. The respective values at 300 K are 40, 000 cm2/(Vs) and 30, 000 cm2/(Vs). Fits are included as black dashed lines in Fig. 3(b), which are vertically offset for clarity32. On the electron side (n > 0) there is a plateau evolving near np = 1012cm−2 for temperatures below 200 K 4 FIG. 2. (a) Raman spectrum of a hBN/CVD-BLG/hBN heterostructure. The inset shows an optical image of the structure (scale bar: 20 µm). The black line outlines the BLG crystal. The BLG can easily be identified in the Raman map (right panel of inset) which shows the total 2D line width of both BLG and SLG. (b) Close-up of the 2D-peak in (a) with a fit curve (black dotted line) using four Lorentzians. (c) Γ2D of all 2D line sub-peaks vs. respective 2D-sub-peak positions ω2D extracted from a Raman map recorded over the full BLG area in (a) with a color code equal to (b). which is fully developed at 2 K. We attribute this feature to a transition from diffusive to ballistic transport. To support this notion, we consider the electron mean free path lm = (h/2e)µ(n/π)1/2 with h being the Planck constant, which has to exceed the distance between the voltage probes l = 2 µm in order to allow for ballistic transport. Together with np this provides an estimate for the electron mobility of µ = 175, 000 cm2/(Vs) for the device to become ballistic which is in good agreement to the extracted low temperature values. The broadening of the plateau at elevated temperature marks the crossover from ballistic to diffusive transport due to temperature-induced electron-phonon scattering. A detailed evaluation of the temperature dependent electron and hole carrier mobilities is shown in Fig. 3(c) where we restrict our fitting range to the diffusive transport regime at 060-1Γ (cm)2DbottomhBNtop hBNBLGSLG(a)Intensity (a.u.)12001400160018002200240026002800hBNG2D-1Raman Shift (cm)280027002600(c)-1Raman Shift (cm)2D273025-1Γ (cm)2D26402670270030152035-1ω (cm)2DIntensity (a.u.)(b)5 FIG. 3. (a) Two terminal resistance R2T of the CVD BLG device in Fig. 1(d) vs. VTG and VBG (T=2 K). The dashed lines represent axes with only displacement field D (white) and charge carrier density n (black) variations. The inset shows the four terminal resistance R (scale bar: 1 kΩ) vs. n (scale bar: 1012 cm−2) for D = 0 V/nm at 2 K (blue) and at 300 K (purple). (b) σ vs. n for T = 2, 40, 70, 100, 130, 160, 190, 240 and 300 K. Black dashed lines indicate linear fits to σ for 2 and 300 K. The plateau on the electron side marks the onset of ballistic transport. (c) Electron (blue) and hole (purple) mobilities µ vs. T . (d) σxy vs. VBG taken in a perpendicular magnetic field of B = 1 T for D = 0 (blue) and D = 0.1 V/nm (purple). (e) dσxy/dVBG vs. VBG for different B. Landau levels with filling factors ν = -20,..,20 are highlighted by dashed lines. nVBG (V)VTG (V) (a)−30030−6060510σ (102 e2/h)(b)010020030000.511.5T (K)µ (105 cm2/(Vs))(c)−12−8−44812−16−201620ν = VBG (V)B (T) (e)−20−10010201357dσxy/dVBG (e2/(hV))−0.4−0.200.20.4electronsholes110210310410R2T (Ω)−202n (1012 cm−2)T = 2 KT = 2 K300 KDRn(d)−202−12−8−404812VBG (V)σxy (e2/h)B = 1 TD = 0 V/nmD ≈ 0.1 V/nm6 smaller carrier densities for T < 200 K, where lm < l. Remarkably, the room temperature mobilities are about two times larger than previously reported values for CVD BLG9 and match those values obtained in state-of-the-art heterostructures fabricated from exfoliated BLG.20,33,34 We now explore the quantum Hall effect (QHE) in our hBN/BLG/hBN Hall bar device. Fig. 3(d) shows the transverse conductance σxy as function of VBG measured in a perpen- dicular magnetic field of B = 1 T for two displacement fields. The measurements show a clear Landau level formation with steps of 4 e2/h. These steps together with the observa- tion that an additional Landau level emerging at 0 e2/h with non-zero displacement field directly proves the AB stacking order in our CVD-BLG.2 To further visualize the mag- netic field dependence of the QHE, we plot in Fig. 3(e) the derivative of σxy with respect to VBG as function of VBG and B. Landau level formation is seen above B = 1 T with filling factors ν being multiples of four. Furthermore, there is a partial degeneracy lifting at larger B-fields. This observation is in good agreement with measurements performed on high quality exfoliated BLG2,33 and underlines the remarkable electronic quality of our CVD-grown BLG. In the following, we focus on the displacement field-induced formation of the energy band gap. Fig. 4(a) shows the maximum two-terminal resistance Rmax measured at the CNP along the white dashed line in Fig. 3(a) as a function of the displacement field D. Rmax exhibits a global minimum at D = 0 V/nm and continuously increases for both positive and negative D-fields as expected for band gap formation. At low temperature (T = 2 K), Rmax shows maximum values at around D = 0.1 V/nm and decreases again at larger displacement fields. Such a maximum is not observed at 40 K. The low temperature behavior contradicts to ideal AB-stacked BLG where Rmax is expected to increase monotonically with increasing D-values, but has also been observed in devices with exfoliated BLG.7,34–37 In order to analyse the low temperature transport behavior, we plot in Fig. 4(b) Rmax as a function of 1/T on a semi-log scale. As expected for thermally activated transport across an energy gap, Rmax varies linearly at higher temperatures (low 1/T values). The energy gap Eg can be extracted from38 (cid:18) Eg (cid:19) 2kBT Rmax(T ) ∝ exp , (3) where kB is the Boltzmann constant. The corresponding fits for selected D fields (color code identical to the data in Fig. 4(b)) show good agreement for low 1/T values. In Fig. 4(c) we plot the extracted Eg as function of D. The values of Eg exhibit a distinct minimum at D = 0 V/nm and show a monotonic and symmetric increase towards positive and negative D fields. The gray lines in Fig. 4(c) represent theoretical predictions for BLG35,36 highlighting the good agreement with the expected behavior for the band gap opening in AB-stacked BLG. We note that the observed dependence of Eg on D clearly indicates the opening of a band gap of up to 15 meV while the deviation from the thermally activated behavior at low temperatures (see Fig. 4(b)) points to the presence of additional transport channels. These channels may result from the presence of stacking folds which have been previously reported in BLG39–41. They can lead to the formation of AB/BA stacking domain walls which may serve as ballistic transport channels42 at sub-gap energies. Another possible explanation could be the formation of conducting edge channels.34 The conductance in our device at the CNP is on the order of few e2/h which matches expectations for ballistic transport channels shunting the device. As our device becomes increasingly conductive with increasing temperature, the contribution of one-dimensional channels strongly weakens due to their metallic behavior.42 Further investigations are needed to explore these transport channels in more detail. In summary, we applied a dry-transfer pickup technique to delaminate CVD-grown BLG from the catalytic copper substrate using exfoliated hBN crystals. Raman spectroscopy and quantum Hall effect measurements prove an AB stacking order and a high electronic quality of our BLG crystals. Transport measurements on hBN/CVD-BLG/hBN Hall bar structures reveal very high charge carrier mobilities of up to 40,000 cm2/Vs at 300 K which is about two 7 FIG. 4. (a) Two terminal maximum resistance Rmax at the CNP (n = 0) as function of displacement field D for T = 2 K and T = 40 K. (b) Rmax as function of the inverse temperature T −1 for various D values according to the colors in (a). The solid lines are fits to a temperature activated transport at higher temperatures. (c) Extracted activation energies (energy gaps) Eg vs. D which are compared to theoretical predictions (gray lines).35,36 times larger than previously reported values for CVD based BLG devices. We demonstrate the opening of a band gap of up to 15 meV when applying vertical displacement field of 0.2 V/nm. While we have identified sub-gap transport channels at the lowest temperatures which might be attributed to AB/BA stacking folds we demonstrate a device performance which well compares to state-of-the-art exfoliated BLG-based devices. Work supported by the EU project Graphene Flagship (contract no. 696656) and the DFG (BE 2441/9-1). Growth of hexagonal boron nitride crystals was supported by the Elemental Strategy Initiative conducted by the MEXT, Japan and JSPS KAKENHI Grant Numbers JP26248061, JP15K21722 and JP25106006. 1E. McCann and M. Koshino, Rep. Prog. Phys. 76, 056503 (2013). 2K. S. Novoselov, E. McCann, S. V. Morozov, V. I. Fal'ko, M. I. Katsnelson, U. Zeitler, D. Jiang, F. Schedin, and A. K. Geim, Nat. Phys. 2, 177 (2006). 3A. Varlet, D. Bischoff, P. Simonet, K. Watanabe, T. Taniguchi, T. Ihn, K. Ensslin, M. Mucha-Kruczynski, and V. I. Fal'ko, Phys. Rev. Lett. 113, 116602 (2014). −0.200.2051015D (V/nm)Eg (meV)(c)−0.200.2051015202530D (V/nm)Rmax (kΩ)(a)2K40K00.050.11·1041/T (1/K)Rmax (Ω)(b)Eg2·1042·1035·1038 4Y. Shimazaki, M. Yamamoto, I. V. Borzenets, K. Watanabe, T. Taniguchi, and S. Tarucha, Nat. Phys. 11, 1032 (2015). 5M. Sui, G. Chen, L. Ma, W.-Y. Shan, D. Tian, K. Watanabe, T. Taniguchi, X. Jin, W. Yao, D. Xiao, and Y. Zhang, Nat. Phys. 11, 1027 (2015). 6E. McCann and V. I. Fal'ko, Phys. Rev. Lett. 96, 086805 (2006). 7H. Min, B. Sahu, S. K. Banerjee, and A. H. MacDonald, Phys. Rev. B 75, 155115 (2007). 8J. B. Oostinga, H. B. Heersche, X. Liu, A. F. Morpurgo, and L. M. K. Vandersypen, Nat. Mater. 7, 151 (2007). 9Y. Hao, L. Wang, Y. Liu, H. Chen, X. Wang, C. Tan, S. Nie, J. W. Suk, T. Jiang, T. Liang, J. Xiao, W. Ye, C. R. Dean, B. I. Yakobson, K. F. McCarty, P. Kim, J. Hone, L. Colombo, and R. S. Ruoff, Nat. Nanotech. 11, 426 (2016). 10L. Liu, H. Zhou, R. Cheng, W. J. Yu, Y. Liu, Y. Chen, J. Shaw, X. Zhong, Y. Huang, and X. Duan, ACS Nano 6, 8241 (2012). 11Y. Hao, M. S. Bharathi, L. Wang, Y. Liu, H. Chen, S. Nie, X. Wang, H. Chou, C. Tan, B. Fallahazad, H. Ramanarayan, C. W. Magnuson, E. Tutuc, B. I. Yakobson, K. F. McCarty, Y.-W. Zhang, P. Kim, J. Hone, L. Colombo, and R. S. Ruoff, Science 342, 720 (2013). 12X. Li, C. W. Magnuson, A. Venugopal, R. M. Tromp, J. B. Hannon, E. M. Vogel, L. Colombo, and R. S. Ruoff, J. Am. Chem. Soc. 133, 2816 (2011). 13S. Chen, H. Ji, H. Chou, Q. Li, H. Li, J. W. Suk, R. Piner, L. Liao, W. Cai, and R. S. Ruoff, Adv. Mater. 25, 2062 (2013). 14H. Zhou, W. J. Yu, L. Liu, R. Cheng, Y. Chen, X. Huang, Y. Liu, Y. Wang, Y. Huang, and X. Duan, Nat. Commun. 4, (2013). 15N. Petrone, C. R. Dean, I. Meric, A. M. van der Zande, P. Y. Huang, L. Wang, D. Muller, K. L. Shepard, and J. Hone, Nano Lett. 12, 2751 (2012). 16L. 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, and C. R. Dean, Science 342, 614 (2013). 17J. W. Suk, A. Kitt, C. W. Magnuson, Y. Hao, S. Ahmed, J. An, A. K. Swan, B. B. Goldberg, and R. S. Ruoff, ACS Nano 5, 6916 (2011). 18L. Banszerus, M. Schmitz, S. Engels, J. Dauber, M. Oellers, F. Haupt, K. Watanabe, T. Taniguchi, B. Beschoten, and C. Stampfer, Sci. Adv. 1, e1500222 (2015). 19L. Banszerus, M. Schmitz, S. Engels, M. Goldsche, K. Watanabe, T. Taniguchi, B. Beschoten, and C. Stampfer, Nano Lett. 16, 1387 (2016). 20C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, S. K. L., and J. Hone, Nat. Nanotech. 5, 722 (2010). 21R. V. Gorbachev, I. Riaz, R. R. Nair, R. Jalil, L. Britnell, B. D. Belle, E. W. Hill, K. S. Novoselov, K. Watanabe, T. Taniguchi, A. K. Geim, and P. Blake, Small 7, 465 (2011). 22J. Yan, Y. Zhang, P. Kim, and A. Pinczuk, Phys. Rev. Lett. 98, 166802 (2007). 23S. Pisana, M. Lazzeri, C. Casiraghi, K. S. Novoselov, A. K. Geim, A. C. Ferrari, and F. Mauri, Nat. Mater. 6, 198 (2007). 24D. Graf, F. Molitor, K. Ensslin, C. Stampfer, A. Jungen, C. Hierold, and L. Wirtz, Nano Lett. 7, 238 (2007). 25F. Herziger, M. Calandra, P. Gava, P. May, M. Lazzeri, F. Mauri, and J. Maultzsch, Phys. Rev. Lett. 113, 187401 (2014). 26C. Neumann, S. Reichardt, P. Venezuela, M. Drogeler, L. Banszerus, M. Schmitz, K. Watanabe, T. Taniguchi, F. Mauri, B. Beschoten, S. V. Rotkin, and C. Stampfer, Nat. Commun. 6, (2015). 27N. J. Couto, D. Costanzo, S. Engels, D.-K. Ki, K. Watanabe, T. Taniguchi, C. Stampfer, F. Guinea, and A. F. Morpurgo, Phys. Rev. X 4, 041019 (2014). 28S. Engels, B. Terr´es, F. Klein, S. Reichardt, M. Goldsche, S. Kuhlen, K. Watanabe, T. Taniguchi, and C. Stampfer, Phys. Status Solidi B 251, 2545 (2014). 29L. Banszerus, H. Janssen, M. Otto, A. Epping, T. Taniguchi, K. Watanabe, B. Beschoten, D. Neumaier, and C. Stampfer, 2D Materials 4, 025030 (2017). 30J. Yan, M.-H. Kim, J. Elle, A. Sushkov, G. Jenskins, H. Milchberg, M. Fuhrer, and H. Drew, Nat Nano 7, 472 (2012). 31A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). 32While we used the same fitting range at all temperatures for the hole side (n < 0), there is ballistic transport evolving at lower temperatures above n = 1012cm−2 on the electron side. We therefore have to restrict the Drude fit, which only applies for diffusive transport, to lower carrier densities below 160 K where the σ vs. n dependence also becomes linear towards lower n values (black dotted line for n > 0 at T = 2 K). On the hole side, the σ vs. n slope also becomes steeper towards the CNP at lower temperatures. The extracted hole mobilities are thus conservative estimates at lower temperatures. For example, if we use the same fitting region as for the electron transport at 2 K we extract hole mobilites of 100,000 cm2/(Vs). 33S. Engels, B. Terr´es, A. Epping, T. Khodkov, K. Watanabe, T. Taniguchi, B. Beschoten, and C. Stampfer, Phys. Rev. Lett. 113, 126801 (2014). 34M. J. Zhu, A. V. Kretinin, M. D. Thompson, D. A. Bandurin, S. Hu, G. L. Yu, J. Birkbeck, A. Mishchenko, I. J. Vera-Marun, K. Watanabe, T. Taniguchi, M. Polini, J. R. Prance, K. S. Novoselov, A. K. Geim, and 9 M. Ben Shalom, Nat. Commun. 8, 14552 (2017). 35L. M. Zhang, Z. Q. Li, D. N. Basov, M. M. Fogler, Z. Hao, and M. C. Martin, Phys. Rev. B 78, 235408 (2008). 36Y. 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). 37E. McCann, Phys. Rev. B 74, 161403 (2006). 38T. Taychatanapat and P. Jarillo-Herrero, Phys. Rev. Lett. 105, 166601 (2010). 39J. S. Alden, A. W. Tsen, P. Y. Huang, R. Hovden, L. Brown, J. Park, D. A. Muller, and P. L. McEuen, Proc. Natl. Acad. Sci. U. S. A. 110, 11256 (2013). 40P. San-Jose, R. V. Gorbachev, A. K. Geim, K. S. Novoselov, and F. Guinea, Nano Lett. 14, 2052 (2014). 41L.-J. Yin, H. Jiang, J.-B. Qiao, and L. He, Nat. Commun. 7, 11760 (2016). 42L. Ju, Z. Shi, N. Nair, Y. Lv, C. Jin, J. Velasco Jr, C. Ojeda-Aristizabal, H. A. Bechtel, M. C. Martin, A. Zettl, J. Analytis, and F. Wang, Nature 520, 650 (2015).
1005.3879	1	1005	2010-05-21T03:26:51	Anomalous structure in the single particle spectrum of the fractional quantum Hall effect	[ "cond-mat.mes-hall" ]	The two-dimensional electron system (2DES) is a unique laboratory for the physics of interacting particles. Application of a large magnetic field produces massively degenerate quantum levels known as Landau levels. Within a Landau level the kinetic energy of the electrons is suppressed, and electron-electron interactions set the only energy scale. Coulomb interactions break the degeneracy of the Landau levels and can cause the electrons to order into complex ground states. In the high energy single particle spectrum of this system, we observe salient and unexpected structure that extends across a wide range of Landau level filling fractions. The structure appears only when the 2DES is cooled to very low temperature, indicating that it arises from delicate ground state correlations. We characterize this structure by its evolution with changing electron density and applied magnetic field. We present two possible models for understanding these observations. Some of the energies of the features agree qualitatively with what might be expected for composite Fermions, which have proven effective for interpreting other experiments in this regime. At the same time, a simple model with electrons localized on ordered lattice sites also generates structure similar to those observed in the experiment. Neither of these models alone is sufficient to explain the observations across the entire range of densities measured. The discovery of this unexpected prominent structure in the single particle spectrum of an otherwise thoroughly studied system suggests that there exist core features of the 2DES that have yet to be understood.	cond-mat.mes-hall	cond-mat	Anomalous structure in the single particle spectrum of the fractional quantum Hall effect O. E. Dial∗, R. C. Ashoori∗, L. N. Pfeiffer†, K. W. West† October 13, 2018 Abstract The two-dimensional electron system (2DES) is a unique laboratory for the physics of interacting particles. Application of a large magnetic field produces massively degenerate quantum levels known as Landau levels. Within a Landau level the kinetic energy of the electrons is suppressed, and electron- electron interactions set the only energy scale 1. Coulomb interactions break the degeneracy of the Landau levels and can cause the electrons to order into complex ground states. In the high energy single particle spectrum of this system, we observe salient and unexpected structure that extends across a wide range of Landau level filling fractions. The structure appears only when the 2DES is cooled to very low temperature, indicating that it arises from delicate ground state correlations. We characterize this structure by its evolution with changing electron density and applied magnetic field. We present two possible models for understanding these observations. Some of the energies of the features agree qualitatively with what might be expected for composite Fermions, which have proven effective for interpreting other experiments in this regime. At the same time, a simple model with electrons localized on ordered lattice sites also generates structure similar to those observed in the experiment. Neither of these models alone is sufficient to explain the observations across the entire range of densities measured. The discovery of this unexpected prominent structure in the single particle spectrum of an otherwise thoroughly studied system suggests that there exist core features of the 2DES that have yet to be understood. Our measurements are performed using time domain capacitance spectroscopy (TDCS) 2;3, which allows measurement of the single particle density of states (SPDOS) with accurately calibrated energy and density scales. The technique uses a repeated series of electronic pulses to measure I/V tunnelling characteristics, ∗Massachusetts Institute of Technology, Cambridge, MA 02139, USA †Alcatel-Lucent Bell Laboratories, Murray Hill, New Jersey 07974, USA (present address: Princeton, Princeton New Jersey, 08544, USA. 1 Figure 1: High field TDCS spectra show "sash" features: bright and dark diagonal lines across the spectrum. The horizontal axis in each spectrum is the electron density in the quantum well, expressed as a filling factor ν with ν = 1 corresponding to completely filling the lowest spin-polarized Landau level. The vertical axis is energy measured from Ef , with E > 0 corresponding to injecting electrons into empty states in the quantum well and E < 0 corresponding to ejecting electrons from filled states. Bright regions correspond to high SPDOS. In the spectrum taken at a temperature of 100 mK with a 4 Tesla perpendicular magnetic field shown in a, features associated with the ν = 1 sash are highlighted with a blue arrow, while another sash about ν = 3 is indicated with a dashed blue arrow. In the 13.5 Tesla data in b, taken at 80 mK, the ν = 1 sash is similarly indicated with blue arrows, while the ν = 1/2 sash about ν = 1/2 is indicated with green. Sharp downward steps in the spectrum corresponding to chemical potential jumps at filling fractions corresponding to fractional quantum Hall plateaus are indicated with yellow arrows. The contrast for positive and negative energies has been adjusted separately in this spectrum. c-f, selected regions of the spectrum are blown up and their contrast enhanced to ease identification of the features, indicated with arrows that match the colors in b. with delays between the pulses to allow the sample to re-equilibrate. Two experimental enhancements make the results described in this letter possible. First, compared to our previous work, we use a sample with a wider, 230A quantum well to confine the electrons in two-dimensions. This reduces the scattering from fluctuations in the well thickness 4. Secondly, we have enhanced our experimental technique to allow us to measure spectra at higher magnetic fields (see supplement). The spacing between Landau level orbit centers is smaller at higher fields, increasing the energy scale of the Coulomb interaction and separating features more clearly in the spectrum. As described in our previous work 3, as we vary the filling factor from ν = 0 to 1 the inter-electron spacing decreases. This causes an increased Coulomb interaction that opens an exchange gap between spin- up and spin-down states at the Fermi surface. Generically, we then expect to see one peak above the Fermi energy Ef (E = 0 in our plots), and one peak below Ef in the SPDOS, and those peaks should move away from each-other as we raise the density. Similarly, we expect to see exchange gaps collapse as we raise the density in the 1 < ν < 2 range, so the two peaks will approach Ef again as we raise the density to ν = 2. 2 0.6 (cid:30)⁄(cid:28) 0.7 (cid:27)⁄(cid:26) d 1357 f -2 -3 -4 -5 -6 -7 0.1 0.2 3579 c e -1 -2 -3 -4 -5 b c 13.5 T b 10 7.5 5 2.5 0 d e -2.5 Energy (meV) 4 T 0.5 1 2 1.5 2.5 Filling Factor ν 3 3.5 4 -5 -7.5 -10 0 (cid:25)⁄(cid:26) (cid:27)⁄(cid:28) (cid:27)⁄(cid:26) (cid:30)⁄(cid:28) Filling Factor ν 1 1(cid:25)⁄(cid:26) 1(cid:27)⁄(cid:26) 0.8 0.9 1.2 1.4 1.6 a 5 EF= 0 -5 Energy (meV) -10 -15 0 At higher magnetic fields, several new features become apparent that are not explained within this simple exchange-based framework. At a perpendicular magnetic field B of 4 Tesla, an additional tunneling feature (the edge of this feature closest to Ef is delineated with blue arrows in Figure 1a) appears at roughly 4 meV near zero density, with an energy that decreases towards E = 0 as it approaches ν = 1. This trend with density is surprising; this is a region where we expect to see the exchange gap opening, and states correspondingly moving away from the Fermi energy. A second feature that appears to be a continuation of this line extends towards increasingly negative energies between ν = 1 and 2, once again moving counter to the behaviour of the exchange gap. We will refer to these features as the "ν = 1 sash". Near E = 0, we see a dark band of suppressed tunnelling caused by the magnetic field induced Coulomb gap 2;5 -- 8. On increasing B to 13.5 Tesla (Figure 1b-f ), we find this ν = 1 sash (blue arrows) becomes more distinct and shifts to higher energies. In addition, we see "jumps" in our data at ν = 1/3, 2/5, 3/5, and 2/3 (yellow arrows). These result from chemical potential variations at the fractional quantum Hall (FQH) plateaus, as observed elsewhere 9;10. A second "ν = 1/2" sash also becomes visible at smaller energies, extrapolating to zero energy at ν = 1/2 (green arrows). An additional sash also appears below the ν = 1 sash, in the vicinity of 1 < ν < 2 (purple arrow). These sashes are destroyed at 13.5 Tesla when we raise the sample temperature to 4 K, while the exchange splitting and integer quantum Hall features survive (see supplemental figure 1). Despite their high energies in the spectrum (up to 8 meV), a relatively small thermal energy of 0.4 meV suffices to eliminate these sashes. This demonstrates a remarkable property of the single particle spectrum: the high energy spectral features depend on fragile properties of the 2DES that only develop at low temperatures. All of these sashes have the opposite density dependence than that expected from the opening and closing of the exchange gap, show a linear dependence in energy on filling fraction which becomes steeper as B is raised, and appear to cross the Fermi energy close to either ν = 1 or ν = 1/2. These similarities suggest that we may be able to explain all these additional features using a single underlying origin. The asymmetry of the sashes about the Fermi energy rules out inelastic tunnelling as a possible origin. Their appearance in quantum wells in several different heterostructures excludes defect-induced resonances in the tunnelling barrier. One hypothesis, suggested by the location of the ν = 1 sash, would be the identification of the quasi- particles as "skyrmions," a coupled spin-charge excitation known to be important near ν = 1 11. Although skyrmions exist as part of the ground state only near ν = 1 12 -- 14, they may exist as high-energy excitations away from ν = 1 where the sash is visible. However, the energy and existence of skyrmions is sensitive to the total magnetic field, not only the perpendicular magnetic field 13. Measurements with tilted magnetic fields 3 Figure 2: A √B energy dependence indicates the sashes originate with electron-electron inter- actions within the lowest Landau levels. Linecuts at several magnetic fields, all at 100 mK, and at a variety of filling factors are plotted in a, showing a number of peaks owing to the unexpected sashes. In b, the energy (horizontal) axis has been scaled by 1/√B, collapsing the peaks associated with the ν = 1/2 and ν = 1 sashes. Residual mismatch in the peak locations is predominantly due to small mismatches in the filling fraction selected at each magnetic fields. Best-fit scaling factors averaged across all densities for each magnetic field are shown in c, with a square-root dependence on B shown for reference. Error bars indicate the sample standard deviation for the best-fit scaling factor as a function of density. For fitting details and alternate functional dependences see supplement. do not reveal any change in the energy of the sash (See suppl. figure 2). Furthermore, the presence of a second sash (purple arrow in Figure 1b) in the fan of features at ν > 1 is inconsistent with this interpretation. In the absence of a model for the lineshape of these new features, it is difficult to extract their exact energies to study their dependence on magnetic field. However, just as we can identify the linear dependence of the "sash" energies with density by examining their evolution with filling fraction at fixed magnetic field, we can also make precise comparisons between spectra at different B but similar ν (Figure 2a). We find an energy scaling factor that depends only on B can be used to collapse all of the peaks in each spectrum onto the 13.5 Tesla spectrum (Figure 2b). This scaling factor grows as √B (Figure 2c). The only fundamental energy scale in this system with this magnetic field dependence is the Coulomb energy scale, proportional to the inverse inter-electron separation. This √B scaling indicates that it may be possible to interpret the new "sash" features in terms of composite Fermions, weakly interacting quasiparticles whose properties are determined entirely by electron-electron interactions. One interpretation of the paired nature of our sashes above and below the Fermi energy is that they are part of a single, continuous excitation that projects through the Fermi energy, with the region near zero energy obscured by the Coulomb gap. The ν = 1/2 sash then extrapolates through Ef at ν = 1/2 (green arrows in Figure 1b,c,e) Composite Fermions with two attached flux quanta (2CFs) experience an effective magnetic field B∗ = B(1 − 2ν). If we imagine sweeping the density in the 2DES while injecting a probe 4 0 0 2 4 6 8 10 12 14 B (Tesla) E ∝ B½ c 1 0.8 0.6 0.4 0.2 Best-(cid:28)t scale factor ν≈1.46 ν≈0.76 ν≈0.33 b TDOS (arb. units) ν≈0.76 ν≈1.46 4 Tesla 8 Tesla 10 Tesla 13.5 Tesla ν≈0.33 a TDOS (arb. units) -8 -6 -4 -2 0 2 4 6 8 Energy (meV) -8 -6 -4 -2 0 2 4 6 8 Energy x (B/13.5 Tesla)(cid:29)½ (meV) 2CF the 2CF will experience zero magnetic field at ν = 1/2, with an effective magnetic field that increases as one moves away from ν = 1/2 (red line in Figure 3a). A Landau fan of 2CFs would then follow the same trend in energy that we see in our ν = 1/2 sash (Figure 3c). We measure our spectra relative to the Fermi energy, which is the same for composite fermions and electrons 15; shifting the Landau fan to place Ef at E = 0 yields the fan in Figure 3d. Note the whole electron may not be able to directly break up into composite Fermions; if the CF interpretation is correct, the features in our spectrum are clouds of incoherent excitations whose lower energy edge is the true CF energy (see supplement). For ν < 1/3, the 2CF filling factor ν∗ is less than 1 (Figure 3b); this means that there are no quasi-Landau levels below the Fermi energy, and we only expect to see states at E > 0. However, for ν > 2/3, ν∗ is between one and two; the model predicts a single quasi-Landau level below the Fermi energy. Examination of the fan in Figure 3e,f clearly shows this expected asymmetry. We find that we can observe this sash over a wide variety of filling fractions, including those corresponding to FQH plateaus caused by other quasiparticles. This is consistent with prior experiments 16 and theoretical work 17. We do not observe 2CF Landau levels that would correspond to n∗ ≥ 2 in these spectra. In the regions where we observe 2CF Landau levels, these would be high energy excitations that should not be well described by composite Fermion theory and may not even exist. Unsurprisingly, we do not observe 2CF states near ν = 1/2 where they would be expected to fall inside of the Coulomb gap. In this interpretation, the slopes of the lines in our Landau fans dE/dν relate directly to the quasiparticle effective masses; for each 2CF splitting, E = ±¯heB(1− 2ν)/m∗, so dE/dν = ±¯he2B/m∗. Using this relation we estimate a mass of (0.35 ± 0.06)m0, where m0 is the free electron mass, for the 2CF at 13.5 Tesla magnetic field. Theoretical estimates of the width of the gap at 13.5 Tesla give typical values of m∗/m0 of 0.23 18 or 0.31 15. Finite well width effects will increase these masses slightly 19. The aforementioned √B dependence of the sash features yields a √B dependence of the cyclotron mass, as is predicted and as has been observed in thermally activated transport measurements 20. The observed Landau gap for the ν = 1/2 sash is roughly a factor of four larger than that reported from thermally activated conductivity, and the mass is correspondingly smaller 20. However, similar discrepancies in measurements of the exchange gap suggest that thermal activation measurements may underestimate the widths of gaps at the Fermi energy 3. The ν = 1 sash features do not afford a simple explanation in terms of composite Fermions; extending the model to other composite quasiparticles with p vortices attached allows us to explain any "sashes" which intersect the Fermi energy at a filling factors ν = 1/p where p is even 21. For a fully spin polarized excitation, p = 1 would correspond to an excitation which has the wrong symmetry under exchange to be an electron wavefunction. However, the ν = 1 sash features form a "fan" of states with evenly spaced energies (Figure 4a), suggesting it may be possible to explain them in terms of injecting electrons into states that 5 Figure 3: From the viewpoint of a composite quasiparticle, sweeping the density in a quantum well also sweeps the effective magnetic field. Panel a shows the effective field experienced by a 2CF as the electron density is swept through the first Landau level. As the effective field and particle density are swept, the filling factor of 2CF composite fermions ν∗ = ν/(1 − 2ν) changes (b). In c, a Landau fan, simplified by assuming a constant CF effective mass, E = ¯hωc(n + 1/2) for the 2CF is shown in absolute energy, with the Fermi energy shown in red: Ef = ¯hωc(nf + 1/2), with nf the greatest integer less than ν∗ (nf = bν∗c). The Fermi energy rises as the 2CF density increases in the quantum well. In d, the fan is shifted to place Ef at E = 0, showing the fan as it is expected to appear in TDCS spectra. The 13.5 Tesla TDCS spectrum without (e) and with (f ) the fan superimposed allows identification of the two ν = 1/2 sash features with a Landau level of the 2CF fan closest to the Fermi surface. Lines suggested by the fan but not observed have been included to show alignment with chemical potential jumps associated with known FQH states, as well as to demonstrate the origin of the asymmetry when high energy CF Landau levels (lightest lines) are not observed. 6 -10 0 (cid:30)⁄(cid:28) (cid:27)⁄(cid:28)½ Filling Factor ν 1 0 (cid:30)⁄(cid:28) (cid:27)⁄(cid:28)½ Filling Factor ν -10 1 10 0 (cid:30)⁄(cid:28) (cid:27)⁄(cid:28) Filling Factor ν 1 0 (cid:30)⁄(cid:28) (cid:27)⁄(cid:28) Filling Factor ν 1 Energy (relative to EF) 5 0 -5 (cid:30)⁄(cid:28) ½ (cid:27)⁄(cid:28) Filling Factor ν 1 10 2CF Landau Fan e,f -10 -7.5 -5 -2.5 0 Energy 2.5 5 7.5 b 4 3 ν* 2 1 0 0 1 d N=2 N=1 N=0 2CF (cid:30)⁄(cid:28) ½ (cid:27)⁄(cid:28) Filling Factor ν Ef 2CF Landau Fan 1 a 0.5 0 B/B -0.5 -1 0 c 10 5 0 -5 Energy (Absolute) Figure 4: The ν = 1 sash features can be emphasized by taking an additional derivative of the data, easing comparison to simulated spectra from a model of electrons localized on a lattice. Panel a shows the wide range of fractions over which the ν = 1 sash persists. Noise introduced in the image by taking the second derivative has been diminished by smoothing with a σ = 160µeV Gaussian. The onset energies of the "clouds" of excitations are evenly spaced in energy, allowing a simple fan to describe all of the excitations. In b, a sample spectrum from the lattice model is shown. The right axis is in units of e2/(4πlb), roughly 16 meV at 13.5 Tesla. The "sashes" that move downward in energy as the density is raised originate with different types of sites that can be populated or depopulated. The origins of selected prominent features are indicated with cartoons of the relevant lattice situations in c. Each cartoon is accompanied by a replica of the spectrum for 0 < ν < 1 with the relevant portion highlighted. The hexagons represent lattice cells of a hexagonal lattice, with single particle states in the center of each hexagon. Tunneling an electron of either spin (yellow) onto a site adjacent to an occupied site (i) may give rise to our ν = 1/2 sash, the lowest energy sash at low densities. A similar sash occurs at slightly higher densities (ii) corresponding to tunneling an electron onto a site adjacent to two occupied sites. Tunneling a minority spin (green) electron onto a site already occupied by a majority spin (red) electron (iii) appears to give rise to our ν = 1 sash, the highest energy sash at low densities. For reference, a breakdown of the spectrum by spin is provided in d, with the minority spins only (i), the majority spins only (ii), and the majority and minority spin states color coded. Note that the minority carriers in the lowest Landau level are always completely spin polarized in this model if the on-site repulsion is larger than the peak-to-peak disorder amplitude. 7 -0.5 0.5 1 0 Filling Factor ν -0.5 0.5 0 1 Filling Factor ν ii iii -0.5 0.5 1 0 Filling Factor ν di 1 0 1 0 1 0.5 0 Energy (e2/4πεlb) 0.5 0 Energy (e2/4πεlb) 0.5 0 Energy (e2/4πεlb) (cid:28)⁄(cid:29) Filling Factor ν 1 1(cid:31)⁄(cid:29) 1(cid:28)⁄(cid:29) 0 0.25 0.5 1 0.75 1.25 Filling Factor ν 1.5 1.75 2 0.5 0 Energy (e2/4πεlb) -0.5 -1 b filename silpot3_setback=5.000000_ll=0_a0=0.500000_disorder=0.01_40x20_b.pgm colormap Colormap 0greyscale.ppm 1 hmin 16384 -7.003838e-03 counts (rotate cw;flip-1-1) hmax 26950 1.480015e+02 plane_a 0 plane_b 0 gamma 0.55 gammacenter 0 imop "(rotate cw;flip-1-1)" box_int (0,0)-(799,99) box (0,1.12496)-(2.00001,-1.12558) xscale 1 yscale 0.1995 c 10a 7.5 5 2.5 0 -2.5 Energy (meV) -5 -7.5 -10 13.5 Tesla (cid:31)⁄(cid:29) 0 experience a density dependent effective magnetic field. The theory of composite Fermions makes an ansatz to identify weakly interacting quasiparticles that give rise to the fractional quantum Hall effect. It is expected to accurately capture the low energy physics near ν = 1/2. Given that our observations are at quite large energies and far from ν = 1/2, it is remarkable that this identification appears consistent with our results. Despite this, it is worth noting that inelastic light scattering measurements of for 1/5 < ν < 1/3 measure collective excitations of composite Fermions at energies which, when adjusted for well width and magnetic field, are comparable to our 2CF energies 16. With these concerns in mind, we note the sashes persists down to very low densities where it seems likely a semiclassical description is possible. One very simple model is that described by Fogler et al 22; consider a hexagonal lattice of sites, each of which has an occupation number for each electron spin n↑↓i equal to zero or one (each site can have zero, one, or two electrons) and uncorrelated Gaussian distributed disorder potential. The inter-site separation is selected such that the lattice is completely filled at ν = 1. The electrons interact through a repulsive effective potential that includes screening from the tunnel electrode, while electrons of the same spin experience an additional very short range attractive exchange potential. For each density, the total energy is minimized by rearranging the electrons and the excited state spectrum is calculated (see supplement for details). In the resulting spectra (figure Figure 4b), we see, as expected, the development of an exchange splitting as the total density approaches ν = 1 23 as well as a Coloumb gap at zero energy 24. The calculated exchange splitting is much larger than the observed, possibly due to the aforementioned finite well width effects. Several additional bands of states emerge that are reminiscent of the sashes in our spectra. These bands are insensitive to the exact heterostructure: screening by the tunneling electrode is unimportant. Near zero density, the lattice is populated by a small number of spin up (majority spin) electrons located at minima in the disorder potential. The highest energy "sash" here corresponds to putting a spin down (minority spin) electron onto a site that has already been occupied by a spin up electron; it is displaced upwards in energy by roughly e2/(4πlb) by the on-site Coulomb repulsion. The next highest energry "sash" corresponds to putting either spin electron onto a site that neighbors an occupied site. These bands have a similar structure to the experimentally observed ν = 1 and ν = 1/2 sashes respectively. The energies of these features decrease as the density is raised because the Fermi energy, the energy required to add an electron to the system after allowing the system to relax, is increasing; as the system is filled, new electrons are forced to be placed closer to their nearest neighbors. Eventually, the type of site associated with each sash will become the lowest energy type of site left to fill, causing the sash to merge with the Coulomb gap. For example, the sash associated with diagram (i) in Figure 4c merges with the Fermi energy at ν = 1/4. The sash originating with diagram (iii) is more complicated. At low densities, this sash corresponds to injecting a 8 minority spin electron into already occupied lattice sites. However, near ν = 1, the simulated sash's physical origin has changed, now corresponding to adding an up or down spin electron on top of a site that has not been occupied. This suggests that at low densities the ν = 1 sash is an electrostatically unfavorable excitation involving a spin-singlet composed of the injected electron and an electron already present in the 2DES. Additional features in the simulation not seen in our measurements correspond to clusters of more than two electrons and are probably an artifact of the lattice onto which we have forced the electrons. We expect this lattice model to be a fair description near ν = 0 and ν = 1 where the 2DEG can be described as a Wigner lattice of disorder pinned electrons (making our initial choice of localized single particle states accurate). The main physics the model then brings to light is the emergence of several discrete high energy bands due to quantization of the electron separations at short distances. However, even in these regimes we note that all of the sashes in the calculation have the same slope, while several distinct slopes are present in the experimental data. The lattice model should become increasingly inaccurate at the intermediate densities where the artificial localization of the electronic states will fail. Here, the composite Fermion model may be more appropriate. Both models predict similar features at the same energies and agree roughly with our experimental spectra but also predict features that do not appear in our spectrum. It is unclear if these two views each have a distinct realm of validity or if they are equivalent in some range of energies and densities in the sense of approximately describing the same physics. Whatever the precise explanation for the sashes, their existence and prominence in a fundamental spec- trum of a system otherwise so thoroughly studied comes a stunning surprise. A detailed and accurate explanation of this spectrum will provide key insights into our microscopic understanding of the ground state of a 2D electronic system. References [1] Jain, J. K. Composite fermions (Cambridge University Press, 2007). [2] Chan, H. B., Glicofridis, P. I., Ashoori, R. C. & Melloch, M. R. Universal linear density of states for tunneling into the two-dimensional electron gas in a magnetic field. Phys. Rev. Lett. 79, 2867 -- 2870 (1997). [3] Dial, O. E., Ashoori, R. C., Pfeiffer, L. N. & West, K. W. High-resolution spectroscopy of two- dimensional electron systems. Nature 448, 176 -- 179 (2007). [4] Luhman, D. R., Tsui, D. C., Pfeiffer, L. N. & West, K. W. Electronic transport studies of a systematic series of GaAs/AlGaAs quantum wells. Appl. Phys. Lett. 91, 072104 (2007). [5] Ashoori, R. C., Lebens, J. A., Bigelow, N. P. & Silsbee, R. H. Equilibrium tunneling from the 2- dimensional electron-gas in GaAs - evidence for a magnetic-field-induced energy-gap. Phys. Rev. Lett. 64, 681 -- 684 (1990). [6] Eisenstein, J. P., Pfeiffer, L. N. & West, K. W. Coulomb barrier to tunneling between parallel two- dimensional electron systems. Phys. Rev. Lett. 69, 3804 -- 3807 (1992). 9 [7] Ashoori, R. C., Lebens, J. A., Bigelow, N. P. & Silsbee, R. H. Energy gaps of the two-dimensional electron gas explored with equilibrium tunneling spectroscopy. Phys. Rev. B 48, 4616 -- 4628 (1993). [8] Yang, S.-R. E. & MacDonald, A. H. Coulomb gaps in a strong magnetic field. Phys. Rev. Lett. 70, 4110 -- 4113 (1993). [9] Eisenstein, J. P., Pfeiffer, L. N. & West, K. W. Compressibility of the two-dimensional electron gas: Measurements of the zero-field exchange energy and fractional quantum Hall gap. Phys. Rev. B 50, 1760 -- 1778 (1994). [10] Khrapai, V. S. et al. Direct measurements of fractional quantum Hall effect gaps. Phys. Rev. Lett. 99, 086802 (2007). [11] Sondhi, S. L., Karlhede, A., Kivelson, S. A. & Rezayi, E. H. Skyrmions and the crossover from the integer to fractional quantum Hall effect at small Zeeman energies. Phys. Rev. B 47, 16419 -- 16426 (1993). [12] Barrett, S. E., Dabbagh, G., Pfeiffer, L. N., West, K. W. & Tycko, R. Optically pumped NMR evidence for finite-size skyrmions in GaAs quantum wells near Landau level filling ν = 1. Phys. Rev. Lett. 74, 5112 -- 5115 (1995). [13] Schmeller, A., Eisenstein, J. P., Pfeiffer, L. N. & West, K. W. Evidence for skyrmions and single spin flips in the integer quantized Hall effect. Phys. Rev. Lett. 75, 4290 -- 4293 (1995). [14] Gallais, Y., Yan, J., Pinczuk, A., Pfeiffer, L. N. & West, K. W. Soft spin wave near nu = 1: Evidence for a magnetic instability in skyrmion systems. Phys. Rev. Lett. 100, 086806 (2008). [15] Halperin, B. I., Lee, P. A. & Read, N. Theory of the half-filled Landau level. Phys. Rev. B 47, 7312 -- 7343 (1993). [16] Hirjibehedin, C. F., Pinczuk, A., Dennis, B. S., Pfeiffer, L. N. & West, K. W. Crossover and coexistence of quasiparticle excitations in the fractional quantum Hall regime at ν ≤ 1/3. Phys. Rev. Lett. 91, 186802 (2003). [17] Peterson, M. R. & Jain, J. K. Flavor altering excitations of composite fermions. Phys. Rev. Lett. 93, 046402 (2004). [18] Morf, R. & d'Ambrumenil, N. Stability and effective masses of composite fermions in the first and second Landau level. Phys. Rev. Lett. 74, 5116 -- 5119 (1995). [19] Park, K., Meskini, N. & Jain, J. K. Activation gaps for the fractional quantum Hall effect: realistic treatment of transverse thickness. Journal of Physics: Condensed Matter 11, 7283 -- 7299 (1999). [20] Leadley, D. R., van der Burgt, M., Nicholas, R. J., Foxon, C. T. & Harris, J. J. Pulsed-magnetic-field measurements of the composite-fermion effective mass. Phys. Rev. B 53, 2057 -- 2063 (1996). [21] Jain, J. K. Composite-fermion approach for the fractional quantum Hall effect. Phys. Rev. Lett. 63, 199 -- 202 (1989). [22] Fogler, M. M., Koulakov, A. A. & Shklovskii, B. I. Ground state of a two-dimensional electron liquid in a weak magnetic field. Phys. Rev. B 54, 1853 -- 1871 (1996). [23] Ando, T. & Uemura, Y. Theory of oscillatory g factor in an MOS inversion layer under strong magnetic fields. J. Phys. Soc. Jpn. 37, 1044 -- 1052 (1974). [24] Pikus, F. G. & Efros, A. L. Coulomb gap in a two-dimensional electron gas with a close metallic electrode. Phys. Rev. B 51, 16871 -- 16877 (1995). 10 Supplementary Information is linked to the online version of the paper at www.nature.com/nature. Acknowledgments We are grateful to Alan MacDonald and Yigal Meir, who independently suggested the identification of sash features with nearby localized states that formed the basis for the lattice model. We thank Patrick Lee, Leonid Levitov and Bert Halperin for helpful discussions regarding the interpretation of our results. This work was sponsored by the Office of Science, Basic Energy Sciences, of the US Department of Energy. Author Contributions O.E.D. built the apparatus and performed measurements and analysis. R.C.A. supervised the work and performed analysis. O.E.D. and R.C.A. prepared the manuscript. L.N.P. and K.W.W. performed the crystal growth. Author Information Reprints and permissions information is available at www.nature.com/reprints. Correspondence and requests for materials should be addressed to O.E.D. ([email protected]) or R.C.A. ([email protected]). 11 Supplementary Material 1 Discharge Pulses Previous measurements were limited to low magnetic fields be- cause of the magnetic field induced coulomb gap at the Fermi energy3,5−8; vanishing tunnel current near equilibrium caused the re-equilibration time to grow rapidly with increasing magnetic field, making the measurement impractical. We have extended the technique to include a "discharge pulse", a brief pulse of the opposite polarity of the pulse used to measure the tunnel current, whose width is tuned to remove exactly the amount of charge that tunneled in during the application of the measure- ment pulse. This rapidly returns the device to near-equilibrium. The tuning is performed by monitoring the tunneling current for several tens of microseconds (this time is increased as the magnetic field is increased) after the discharge pulse to insure that the sample is re- equilibrated. This time period also allows the 2DES to re-thermalize. We confirm that this re-thermalization is complete by varying the delay between pulse repetitions and confirming that we measure the same I/V curve. 2 Energy Scaling Supplementary Figure 1 Alternate functional forms for the fit to the magnetic field dependence of the scale factor are less satisfactory than pB. Error bars indicate the standard deviation of the scale factor across the range of available densities at each magnetic field. The fitting functions are pB (solid black line, χ2 = 2.4), power-law B α (dotted black line, α = .52, χ2 = 3.1), linear with an offset (blue line, χ2 = 6.9), and linear without an offset (green line, χ2 = 20). The energy scaling factor for each magnetic field shown in Figure 2 of the main text was determined by maximizing the correlation between d 2I /dV 2 spectra taken at the same density but at two different magnetic fields. First, for each filling in the higher field spectrum, the closest filling factor in the lower field spectrum is selected. Then, the spectra are numerically differentiated a second time, giving d 2I /dV 2; this minimizes any overall background to the spectrum. High frequency noise resulting from this double- differentiation is then removed by convolving each spectrum with a σ = 0.38 meV Gaussian. The lower field spectrum ρ1(E) is then stretched along the energy axis by a scale factor and resampled to the same energy resolution as the higher field spectrum ρ2(E) using linear interpolation. The value of that maximizes the correlation χ() =R ρ1(E)ρ2(E)dE between the two spectra is then searched for and located. At this point, the best estimate for is noisy due to discretization of the data in energy. To remove this noise, χ() is evaluated at 25 values of over a range stretching from 10% below to 10% above this absolute best correlation, and the resulting curve of χ versus is fit using a quadratic. The value of at the peak of this quadratic curve is then selected as the best scale factor for aligning the two spectra at this density. This process is repeated for every density from 0.1 < ν < 0.9 and 1.1 < ν < 1.9 in the spectrum. The resulting map of versus ν is roughly constant; maximum deviations from the mean value are of the order 10%. This flatness indicates that a single value of is sufficient at all densities to explain the magnetic field dependence of our data. This value is found by averaging the value of across this range of ν, discarding any outliers that fall more than 3 standard deviations away from the mean. This mean is used in figure 2c. The standard deviation of across this range of densities is taken as width of the error bars, essentially the range of found to be a reasonable fit. As such they are not representative of normally distributed statistical errors. However, the reduced sum-squared error (χ2) still a rough guide as to the quality of different fits. Supplementary Figure 1 shows alternate fits to the B-field dependence. 3 Supplemental Data Supplementary figure 1 demonstrates that raising the sample tem- perature to 4.1 Kelvin eliminates the "sash" features while leaving Supplementary Figure 3 Tilting the sample while maintaining the same magnetic field perpendicular to the quantum well increases the total mag- netic field applied. This increases the Zeeman energy (which varies with total field), while keeping the exchange energy (which varies with perpendicular field) fixed. Excitations which originate with the interplay between the Zeeman and exchange energies, including skyrmions, are sensitive to this variation of the total magnetic field. By comparing the spectrum in b to a, we see that increasing the total magnetic field by 41% does not affect the ν = 1 sash (blue arrows). This data was taken in a device with a 175 Å quantum well at 80 mK. the exchange splitting intact. The broadening of the Fermi function in the tunnel electrode is not sufficient to explain this change. Sup- plementary Figure 2 demonstrates the insensitivity of the ν = 1 sash to presence of magnetic field in the plane. Supplementary Figure 3 displays the effect of disorder on both the ν = 1/2 sash and the ν = 1 sash. The ν = 1/2 sash is destroyed by disorder in a narrower quantum well whereas the ν = 1 sash survives this disorder. 4 6 8 10 12 14 Magnetic Field B 0 0 2 1.2 1 0.8 0.6 0.4 0.2 Best-(cid:31)t Scale Factor 45º BTot=5.66 Tesla b 5 2.5 0 Energy (meV) -2.5 0º BTot= 4 Tesla 350 400 DC Bias (mV) -5 B⊥=4 Tesla 300 350 DC Bias (mV) 400 a 5 2.5 0 Energy (meV) -2.5 -5 Supplementary Figure 2 The effect of temperature on the ν = 1 and ν = 1/2 sashes is shown by comparing a spectrum acquired in a 175 Å quantum well at 8 Tesla and 80 mK (a) to one measured at 4.1 K (c). The effect of Fermi function broadening in the electrode and quantum well can be demonstrated by convolving with a 4.1 K Fermi function (b). This does not destroy the sash features, demonstrating that some change in the many-body state of the quantum well is responsible. Similar data at 13.5 Tesla in a 230 Å well also shows the ν = 1/2 sash features completely destroyed at 4.1 Kelvin, while some very slight trace of the ν = 1 sash survives (d-f). Low T Broadened to High T 80 mK, Broadened to 4K High Temperature Data c 4 Kelvin filename 8t_4k.pgm colormap Colormap brygbw.ppm hmin 18371 -1.514865e+06 Z (neg;rotate cw;flip-1-1) hmax 44523 1.069118e+07 plane_a 0 plane_b 0 gamma 1 gammacenter imop "(neg;rotate cw;flip-1-1)" box_int (0,0)-(94,122) box (-222.409,6.5148)-(-801.955,-6.5148) xscale 90.9091 yscale 267 0 5 filename data.pgm colormap Colormap brygbw.ppm hmin 16889 -3.118571e+06 Z (neg;rotate cw;flip-1-1;lowpass-0-6.5-3) hmax 48184 1.036467e+07 plane_a 0 plane_b 0 gamma 1 gammacenter imop "(neg;rotate cw;flip-1-1;lowpass-0-6.5-3)" box_int (0,24)-(92,268) box (-204.545,6.675)-(-727.273,-6.3546) xscale 90.9091 yscale 267 0 -400 DC Bias (mV) -600 80 mK, Broadened to 4K filename data.cal.pgm colormap Colormap brygbw.ppm hmin 27812 -6.076256e+06 Z (neg;rotate cw;flip-1-0;lowpass-0-4.5-3) hmax 42129 2.828750e+07 plane_a 6.41917 plane_b -0.387866 gamma 1 gammacenter imop "(neg;rotate cw;flip-1-0;lowpass-0-4.5-3)" box_int (0,27)-(300,291) box (0,-10.494)-(1.50055,10.494) xscale 1 yscale 318 2.5 0 -2.5 -5 f -10 -7.5 -5 -2.5 0 2.5 5 7.5 10 -400 -600 DC Bias (mV) -800 4 Kelvin filename data.cal.pgm colormap Colormap brygbw.ppm hmin 20595 -7.598771e+05 Z (neg;rotate cw;flip-1-0) hmax 44908 2.314957e+06 plane_a 24.8226 plane_b -3.1971 gamma 1 gammacenter imop "(neg;rotate cw;flip-1-0)" box_int (-1,34)-(299,365) box (-0.00197233,-10.4994)-(1.49953,10.4994) xscale 1 yscale 318 0 b 5 2.5 0 -2.5 -5 e -10 -7.5 -5 -2.5 0 2.5 5 7.5 10 Low Temperature Data a 80 mK filename 8t.pgm colormap Colormap brygbw.ppm hmin 16840 -3.257729e+06 Z (neg;rotate cw;flip-1-1) hmax 48025 1.172139e+07 plane_a 0 plane_b 0 gamma 1 gammacenter imop "(neg;rotate cw;flip-1-1)" box_int (0,27)-(92,271) box (-204.545,6.5148)-(-727.273,-6.5148) xscale 90.9091 yscale 267 0 5 -400 DC Bias (mV) -600 80 mK filename data.cal.pgm colormap Colormap brygbw.ppm hmin 31335 -3.986827e+06 Z (neg;rotate cw;flip-1-0) hmax 39015 3.314118e+07 plane_a 6.41917 plane_b -0.387866 gamma 1 gammacenter imop "(neg;rotate cw;flip-1-0)" box_int (0,27)-(300,291) box (0,-10.494)-(1.50055,10.494) xscale 1 yscale 318 2.5 0 -2.5 -5 d -10 -7.5 -5 -2.5 0 2.5 5 7.5 10 Energy (meV) 8 Tesla Energy (meV) 13.5 Tesla 0 (cid:31)⁄(cid:29) (cid:28)⁄(cid:29) 1 Filling Factor ν 1 (cid:31)⁄(cid:29) 0 (cid:31)⁄(cid:29) (cid:28)⁄(cid:29) 1 Filling Factor ν 1 (cid:31)⁄(cid:29) 0 (cid:31)⁄(cid:29) (cid:28)⁄(cid:29) 1 Filling Factor ν 1 (cid:31)⁄(cid:29) Supplementary Figure 4 The effect of disorder on the ν = 1 and ν = 1/2 sashes is shown by comparing a spectrum acquired in a 175 Å quantum well (a) and a wider 230 Å quantum well (b), both with an applied field of 4 Tesla at 80 mK. Although the ν = 1 sash is clearly visible in both spectra (blue arrows), the added disorder from the narrower quantum well almost completely destroys the ν = 1/2 sash (green arrows). 4 Orthogonality Issues in the Compos- ite Fermion Model Given that whole electrons are tunnelling into the system, not com- posite particles, why do we observe associated structures in our spectra? Jain has suggested that when the 2DES is in an incom- pressible fractional state, it may be possible to construct several composite fermions from a whole electron, resulting in a peak in the SPDOS 25;26. Our peaks sharpen as the 2DES enters fractional Hall plateaus, which we may understand as due to an increase in quasiparticle lifetimes as the system becomes gapped. However, we do not observe any qualitative change in the spectra, nor do we detect a new, sharp resonance. The spectral weight of the peak suggested by Jain may be too small to observe in our otherwise crowded spectra. Instead, we expect tunnelling directly into a CF Landau level to be strongly suppressed. This suppression reflects the same physics as that responsible for the Coulomb gap at zero energy. In the case of the Coulomb gap, although there is a high thermodynamic density of states at the Fermi energy 27 in a partially filled Landau level, the tun- nelling density of states at low energies is reduced by inter-electron repulsion between the tunnelling electron and the electrons already in the quantum well. Similarly, we expect a Coulomb pseudo-gap for tunnelling an electron into a highly correlated CF Landau level. The true energy of the CF level occurs at an energy closer to E f than the "clouds" of incoherent excitations we observe as the bright part of the "sash"; this is reflected in the placement of the arrows and fans at the edge of the bright part of the sash in the spectra. However, regardless of the exact relationship between the energy of the quasiparticle and the energy of this "cloud", the behaviour of the "cloud" can be used to track the evolution of the quasiparticle as the 2DES density and magnetic field are varied. This measurement is insensitive to quasiparticle charge; when tunnelling an electron into an FQH state with fractional charge 1/q, conservation of charge de- mands that we create q excitations, each with energy eB∗ qm∗ , requiring a total energy of eB∗/m∗. Supplementary Figure 5 Sample effective potentials calculated for the lattice model show the relatively short range of the exchange attraction. The repulsion between electrons of unlike spins is given by the green curve, while that for like spins by the red curve. The blue line shows the strength of the exchange coupling (the difference between the red and green curves.) Note that the potential for like spins is meaningless at very short distances due to Pauli exclusion. The lattice constant of the hexagonal lattice is shown with the black dotted line. 5 Lattice Model We use a lattice model based roughly on the one described by Fogler et. al22. Each electron is assumed to be localized in a coherent state on a hexagonal lattice. The lattice constant is selected such that one magnetic flux quantum penetrates each unit cell. Periodic boundary conditions are used with a hexagonal lattice, as shown in Supplementary Figure 6d. The resulting basis set is almost or- thogonal and has the same degeneracy as the true single particle states. Wavefunction overlap is neglected. Electrons of the same spin interact through an effective potential as given by equation 60 of reference 4. However, to address concerns that the spectrum may be modified by the nearby tunneling electrode, the bare Coulomb interaction v(r ) is replaced by¡e2/4π¢‡r−1 − (r 2 + s2)−1/2· where s is the distance between the 2D system and the nearby 3D electrode. Electrons of different spins interact through the same potential, but with the exchange term uE X set to zero (Supplementary Figure 5). On each iteration of the software, a single electron is added to the lattice at the most energetically favorable position, and then the electron occupations are selected to minimize the total energy as described in reference 4. The system is then modified by adding a single electron or hole on one lattice site, and the energy change is calculated and stored. This process is repeated for each possible lattice site, carrier type, and spin. The chemical potential is taken as the mean of the largest hole addition energy and the smallest electron addition energy. The energy difference between the chemical potential and the addition energy is then histogrammed. This histogram is used as an approxi- mation to the SPDOS at this density. The entire process is then repeated until the lattice is filled, gen- erating a complete simulated TDCS spectrum. The process of filling the lattice is demonstrated in the supplemental movie. Sample results are shown in Supplementary Figure 6. It is worth noting that in the simulations, the nearby electrode does not play a large role once it is more than a magnetic length away. This condition is met in all of the measurements described in this letter. The main role of the screening from this electrode is to diminish distant electron- electron interactions, but the physics of the spectrum is dominated by nearest-neighbor interactions (Supplementary Figure 6a,c). The calculated spectra show many "sashes" not present in the experimental spectrum. Although the additional sashes appear to be -7.5 300 400 DC Bias (mV) 500 600 -7.5 300 400 DC Bias (mV) 500 b 7.5 5 2.5 0 Energy (meV) -2.5 -5 7.5 a 5 2.5 0 Energy (meV) -2.5 -5 0.5 1 Distance, units of lb 1.5 2 1 0.8 0.6 0.4 0.2 Energy, units of e2/(4πεlb) 0 0 Supplementary Figure 6 Sample spectra generated by the lattice model exhibit features similar to the experimentally observed sashes. a shows a series of simulations with differing setbacks (distances from the center of the quantum well to the 3D electrode) and fixed disorder of 3% of the Coulomb energy scale e2/(4πlb. Note that setbacks of more than lb have very little impact on the spectrum. The effects of disorder are explored in b, where a series of spectra with a fixed setback of 2 lb but varying disorder, measured as a percentage of the Coulomb energy scale, are shown. The actual setback varies with magnetic field and sample design; the setback appropriate for the device used in these spectra is shown in c. Prominent sash features are visible from about 4 to 13.5 Tesla, corresponding to setbacks of 1 to 3 lb. All spectra are calculated with a 25x25 unit cell. Sample annealed electron lattices are illustrated using a smaller 20x20 unit cell at several filling factors in d, with a single unit cell shown in bold. Note the uniformity in the number of occupied neighbor sites. These lattices are for a disorder of 3% and a setback of 2 lb. See also supplemental movie A. artifacts of the lattice in our semi-classical model, it is interesting to note that while the addition of disorder diminishes all of the features in the spectrum, these additional sashes appear to diminish more as disorder is increased (see, for example the 6% plot in Supplementary Figure 6b). Note that the simulation does not included any lifetime effects. It is not possible to estimate the disorder in our devices from these simulations as it is likely that lifetime broadening influences line-widths in our spectrum and which features are visible in our spectrum. 25. Jain, J. K. & Peterson, M. R. Reconstructing the electron in a fractionalized quantum fluid. Phys. Rev. Lett. 94, 186808 (2005). 26. Vignale, G. Integral charge quasiparticles in a fractional quantum Hall liquid. Phys. Rev. B 73, 073306 (2006). 27. Smith, T. P., Goldberg, B. B., Stiles, P. J. & Heiblum, M. Direct measurement of the density of states of a two-dimensional electron gas. Phys. Rev. B 32, 2696 -- 2699 (1985). a 1 Setback 0.4 lb Setback 0.8 lb Setback 1.2 lb Setback 2 lb Setback ∞ lb 0.5 0 -0.5 Energy (e2/4πεlb) 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 0.5 1 1.5 2 Filling Factor ν Filling Factor ν Filling Factor ν Filling Factor ν Filling Factor ν Disorder 0.2% Disorder 2% Disorder 3% Disorder 6% Disorder 12% -1 0 1 0.5 0 -0.5 b Energy (e2/4πεlb) 0.5 1 1.5 Filling Factor ν 2 0 0.5 1 1.5 2 0 0.5 1 1.5 Filling Factor ν Filling Factor ν ν=(cid:28)⁄(cid:26) ν=¼ 2 0 ν=½ 0.5 1 1.5 2 Filling Factor ν 0.5 1 1.5 Filling Factor ν 2 0 d 2 4 8 10 12 14 6 B (Tesla) -1 0 4 3 2 1 0 0 Setback, units of lb c
1911.04451	1	1911	2019-11-11T18:53:04	Dynamics and efficient conversion of excitons to trions in non-uniformly strained monolayer WS$_2$	[ "cond-mat.mes-hall", "cond-mat.soft" ]	We investigate the transport of excitons and trions in monolayer semiconductor WS$_2$ subjected to controlled non-uniform mechanical strain. We actively control and tune the strain profiles with an AFM-based setup in which the monolayer is indented by an AFM tip. Optical spectroscopy is used to reveal the dynamics of the excited carriers. The non-uniform strain configuration locally changes the valence and conduction bands of WS$_2$, giving rise to effective forces attracting excitons and trions towards the point of maximum strain underneath the AFM tip. We observe large changes in the photoluminescence spectra of WS$_2$ under strain, which we interpret using a drift-diffusion model. We show that the transport of neutral excitons, a process that was previously thought to be efficient in non-uniformly strained 2D semiconductors and termed as "funneling", is negligible at room temperature in contrast to previous observations. Conversely, we discover that redistribution of free carriers under non-uniform strain profiles leads to highly efficient conversion of excitons to trions. Conversion efficiency reaches $\simeq 100\%$ even without electrical gating. Our results explain inconsistencies in previous experiments and pave the way towards new types of optoelectronic devices.	cond-mat.mes-hall	cond-mat	a Dynamics and efficient conversion of excitons to trions in non-uniformly strained monolayer WS2 Moshe G. Harats, Jan N. Kirchhof, Mengxiong Qiao, Kyrylo Greben, and Kirill I. Bolotin Department of Physics, Freie Universitat Berlin, 14195 Berlin, Germanya) We investigate the transport of excitons and trions in monolayer semiconductor WS2 sub- jected to controlled non-uniform mechanical strain. We actively control and tune the strain profiles with an AFM-based setup in which the monolayer is indented by an AFM tip. Op- tical spectroscopy is used to reveal the dynamics of the excited carriers. The non-uniform strain configuration locally changes the valence and conduction bands of WS2, giving rise to effective forces attracting excitons and trions towards the point of maximum strain un- derneath the AFM tip. We observe large changes in the photoluminescence spectra of WS2 under strain, which we interpret using a drift-diffusion model. We show that the transport of neutral excitons, a process that was previously thought to be efficient in non-uniformly strained 2D semiconductors and termed as funneling, is negligible at room temperature in contrast to previous observations. Conversely, we discover that redistribution of free carriers under non-uniform strain profiles leads to highly efficient conversion of excitons to trions. Conversion efficiency reaches (cid:39) 100% even without electrical gating. Our results explain in- consistencies in previous experiments and pave the way towards new types of optoelectronic devices. I. INTRODUCTION Two-dimensional materials from the class of tran- sition metal dichalcogenides (TMDCs) are actively considered for applications in photonics, electronics, and optoelectronics. TMDCs feature a direct band- gap in the monolayer limit1,2, exhibit an unusual spin/valley locking3 -- 5,and can host tunable single photon emitters6 -- 10. Prototype TMDC-based electronic devices including transistors11, p-n junctions12, and solar photo- conversion devices13,14 have already been demonstrated. Moreover, high Young's modulus of TMDCs (∼ 170GP a 15) invites applications of these materials in flex- in WS2 ible electronics15 -- 17. Physical properties of TMDCs change under mechan- ical strain18. In the simple case of constant uniaxial strain, the band-gap energy is reduced by 50 meV /%19,20 and the phonon-assisted coupling is altered20,21. The band-gap reduction is twice higher, 100 meV /%, for uniform biaxial strain22. In conventional semiconductor materials, the advent of "strain-engineering", controlled strain-induced modification of the band-gap, allowed di- recting flows of excitons in quantum wells23 and led to the performance improvement of strained Silicon MOSFET transistors24,25. In the same vein, strain-engineering of TMDCs has recently been analyzed. The theoretical pro- posal of Ref. 18 considered a spatially-varying strain profile induced in a suspended TMDC membrane by a sharp tip of an atomic force microscope (AFM). In such a configuration, a force proportional to the band-gap en- ergy gradient acts on a photoexcited electron/hole pair (an exciton), transporting it to the center of the strain "funnel" at the location of the tip (Fig. 1(b)). Two fea- tures make this setup attractive for efficient solar photo- conversion. First, spatial variation of the local bandgap broadens the absorption spectrum of the TMDC. Second, excitons transported to the location of the AFM tip can be efficiently extracted and converted to electrical cur- rent. Experimental signatures consistent with the funnel effect have been previously observed in wrinkled few-layer 26 and monolayer MoS2 nanobubbles27. Neverthe- MoS2 less, these previous experiments did not allow for induc- tion of predictable strain profiles, dynamic control and tunability of strain magnitude, or quantitative analysis of the funneling efficiency. Because of that, the mech- anisms governing transport and dynamics of excitons in non-uniform strain profile has not been fully investigated. Here, we experimentally realized the setup originally proposed by Ref. 18. Highly non-uniform and in-situ tunable strain profiles are induced in suspended mono- layer WS2 by a tip of an all-electrical AFM both in am- bient and under vacuum conditions, while optical fin- gerprints of funneling are collected by a high-resolution optical system. We observe large changes in the PL spec- tra as a function of strain. By comparing our results to a simple drift-diffusion model, we decouple the contri- butions of two effects: funneling of excitons and trions (charged excitons) and funneling of free charge carriers. Contrary to prior expectation, we find that funneling of excitons is a very inefficient process with less than 4% of photoexcited excitons reaching the funnel center even at highest achievable strain. In contrast, funneled free carriers are found to dominate optical spectra by bind- ing to neutral excitons to form trions with a conversion efficiency up to ∼ 100%. Taken together, our results ex- plain inconsistencies in prior experiments and open a new pathway towards ultra-efficient TMDC-based photocon- version and optoelectronic devices. II. RESULTS To study a controllably strained TMDC, we require a suspended sample which can be approached from one side by an AFM tip while allowing optical access for excitation and detection from the other side. To pro- duce suspended samples, we perforated holes with dif- ferent diameters ranging from 0.5 to 2 µm in a 50 nm- thick Silicon Nitride (SiNx) using a Focused Ion Beam (FIB) milling. Monolayer WS2 was mechanically exfo- liated onto a PDMS film and transferred on top of the holes using a dry-transfer technique28 (Fig. 1(c)). Full details on the fabrication of the samples is given in Ref. 29. To fulfill the main experimental challenge of this work, the induction of well-defined and controlled non-uniform strain profiles in a suspended monolayer of WS2, we de- veloped a novel AFM-based apparatus (Fig. 1(a)). A suspended monolayer is indented by an AFM tip from be- low, while being optically excited from above. Critically, the AFM cantilever is based on a piezo-resistive technol- ogy allowing all-electric motion actuation and deflection readout30. All-electrical operation of the cantilever al- lows for optical excitation/detection of WS2 without the disturbing effect of laser sources normally employed in conventional AFMs. The described AFM setup is ca- pable of topographic imaging both in tapping (see Fig. 1(d)) and contact mode, recording force-distance curve for nanoidentation experiments12,31,32, and, most impor- tantly, applying a constant force on the suspended mem- brane whilst optical measurements are performed. Last but not least, the whole AFM system, being all-electrical and compact, can be incorporated into an optical cryo- stat, allowing additional measurements of the system un- der vacuum. The optical part of the setup consists of a high- resolution objective (NA=0.75) and a periscope mounted on a scanning stage capable of nanometer resolution po- sitioning. The sample is excited by a CW laser source at 532 nm (spot size FWHM 600 nm, power 30 µW unless specified otherwise) and the photoluminescence (PL) is directed to a spectrometer (Andor). Overall, our unique AFM/spectroscopy setup combines full AFM and full spectroscopic characterization capabilities. This combi- nation is critical for the investigation of exciton transport in non-uniform strain profiles both in ambient and vac- uum conditions. We start by spatially locating a pristine WS2 mono- layer suspended over a hole. This is done using tap- ping mode AFM imaging to avoid puncturing delicate monolayers6. After locating the center of the suspended portion of the WS2 monolayer, we perform a nanoidenta- tion experiment at that location16,31,32. From this data, we extract the Young's modulus (Y ) and pre-tension (σ0) of the membrane, the parameters necessary for the accu- rate determination of the strain profile29,33. Once pre- liminary characterization is complete, the sample is con- trollably indented by locking the PID loop at the desired 2 Fig. 1 (a) Schematics of the measurement apparatus. The WS2 flake suspended on top of a hole in the SiN/Si membrane is indented from the bottom by an AFM cantilever while being optically in- terrogated from the top. (b) The schematic strain profile (top red curve) and band structure (bottom blue curves) of WS2 vs. dis- tance from the membrane center under non-uniform straining in our experiment. An effective force drives photoexcited excitons towards the point of minimum bandgap in the middle of the mem- brane where electrons are funneled and holes are inverse funnelled. (c) Optical image of a WS2 monolayer (dashed black line) on top of perforated SiN/Si substrate. (d) A tapping mode scan of WS2 shown in (c) recorded using our all-electrical AFM setup. The inset shows the height profile along a line crossing a suspended flake. force value, with the AFM tip still positioned at the cen- ter of the membrane (determined from maximal red-shift of the PL with respect to zero strain), and PL spectra vs. strain are acquired. Figure 2(a) shows the evolution of the PL spectra of sample A as it is progressively indented. Non-uniform strain profiles are parameterized by a single value, the maximum strain εmax that is reached at the middle of the membrane29. The sample is indented up to its breaking point, typically about εmax ∼ 2.5% (limited by the rup- ture of the flake caused by the sharp AFM tip). At zero strain, the PL spectrum is described by a non-symmetric Gaussian peak20,21. As strain is increased, this peak broadens and finally evolves at high strain into a two- peak structure: the red-shifted 'red' peak and the blue- shifted 'blue' peak. Similar strain-dependent two-peak structure is seen in every measured sample, for example in sample B (Fig. 2(b)). Amplitudes of the two peaks are sample-dependent. In sample B, for instance, the 'red' peak has much higher spectral weight compared to sample A. Previous work26 suggests a tempting interpretation of the two-peak structure. The 'red' peak could stem from emission of the excitons funneled to the point of the high- est strain, whereas the 'blue' peak - to excitons that did not reach the funnel center, thus emitting throughout the 3 sample as the laser excitation spot exceeds the character- istics funneling length. A very high density of excitons is expected at the membrane's center in this interpretation. This should lead, in turn, to a rapid non-radiative Auger recombination of excitons which is known to be effective in TMDCs34. To test the role of Auger recombination, we recorded PL spectra for sample B at a relatively low power of 8nW . At this power, only a few excitons are present in the entire sample at any given time and the role of Auger recombination should be negligible34. Fig- ure 2(b) shows that only the 'red' peak remains at low power while the 'blue' peak vanishes. In principle, such behavior is consistent with reduced Auger recombination and implies more efficient funneling at low power. Finally, we tested for the contribution of charge exci- tons (trions) emission in our sample, as the binding en- ergy of charged excitons in WS2 is suspiciously close to the energy separation between the 'red' and 'blue' peaks. While PL spectra of samples A and B do not exhibit any trion contribution at zero strain, we later demonstrate that such contribution can arise when strain is increased. Therefore, we n-doped sample C by measuring it in vac- uum. The desorption of water and Nitrogen from the sample surface increases the density of free electrons35, which, in turn, bind to neutral excitons to form neg- atively charged trions. Indeed, well-understood peaks corresponding to neutral (at 2.01 eV ) and charged exci- tons (at 1.965 eV ) are seen in sample C at zero strain (Fig. 2(c)). Interestingly, as strain is increased, only the 'red' peak grows, while the 'blue' peak vanishes. Overall, the experimental data of Fig. 2 poses the following questions. What is the physical origin of the 'red' and 'blue' peaks? How efficient is the funneling of neutral excitons in our sample? What is the role of charged excitons and why does their contribution appear to be strain-dependent? Fig. 2 (a) PL spectra of sample A at various strain levels - normalized and shifted for clarity. The data recorded during loading (solid lines) and unloading (dashed lines) cycles exhibit no hysteresis indicating the absence of mechanical slipping in our experiments. Maximal strain εmax (reached underneath the AFM tip) is shown next to each curve; The curve with εmax = 0% corresponds to an unstrained device. Two-peak Gaussian fits are shown along with the data. (b) Power- and strain- dependent PL spectra of sample B. The blue and purple curves are PL recorded at the excitation power of 30µW for unstrained and strained (εmax = 1.5%) device. The red curve corresponds to the same strain, but with PL spectrum recorded at 8nW . The data is normalized for better visibility. (c) Strain-dependent PL spectra for sample C that was measured in vacuum. Well-resolved neutral and charged exciton peaks evident at zero strain indicate high doping level in that device. III. DISCUSSION yields34: To answer these questions, we analyze the drift- diffusion equations governing exciton transport in our system. For non-uniform density of excitons n(r), the steady-state continuity condition for excitonic diffusion (cid:126)JD = D∇n(r) and drift (cid:126)Jµ = µn(r)∇u(r) currents ∇(D∇n(r)) + ∇(µ n(r)∇u(r))− Here D is the diffusion coefficient, µ = D − n2(r)RA + S(r) = 0 (1) kB T is the mo- n(r) τ 4 √ time due to drift and diffusion respectively can be eval- uated within a simple Drude approximation. We find Dτ (cid:39) 180nm is that the diffusion length ldif f usion = much larger compared to the drift length averaged over the excitation spot, < ldrif t >=< ∇u(r) > µτ (cid:39) 5nm. Dominating contribution of the diffusion leads, in turn, to inefficient funneling. One could claim that we observe a rather low funneling efficiency due to charging in the system18. This is not the case here as our analysis us- ing the drift-diffusion equation is blind to any charging effects and is also valid to a type I funnel that does not exhibit charging effects. We can analyze the relative contributions of drift and diffusion in another, more quantitative way. It is easy to show that, on average, exciton current flows towards the funnel center (drift dominates over diffusion) if the following condition is met29: kBT < − S(r)∇u(r) ∇S(r) (2) Fig. 3 Comparison between the experimentally measured PL spec- tra in all measured samples (red curves) and the predictions of two models considered in the text: the model of Eq. 1 which only in- cludes funneling of neutral excitons (thin dashed blue line), and the full model of Eq. 3 that adds the effects of carrier funnel- ing and neutral-to-charged exciton conversion (thick dashed black line). The shaded blue area is the unstrained PL shown for refer- ence. The spectra from the following devices are shown: (a) Sample A for the highest strain. (b)-(c) Sample B for high (b) and low (c) laser excitation intensities and the highest strain. (d) Sample C for the highest strain. bility, RA -- the Auger recombination rate, τ -- the exciton 2πσ2 e−r2/2σ2 lifetime, and S(r) = I0 -- the exciton genera- √ tion rate in a Gaussian illumination profile with intensity I0 and σ = F W HM/2 2 ln 2. Unless stated otherwise, we use material constants D = 0.3 cm2/s, µ = 12 cm2 eV ·s , RA = 0.14 cm2/s, and τ = 1.1 ns experimentally deter- mined in Ref. 34. The change of the band-gap due to the strain is assumed to be u(r) = Eg − 0.05· ε(r), where ε(r) is the trace of the strain tensor19,22. A detailed analysis and simulations of Eq. 1 along with various calculations of u(r) are shown in29. Figure 3 (blue dashed curves) shows the PL spectra obtained from the numerical solutions of Eq. 129. These solutions clearly do not match the experimental data (red curves). Indeed, while the 'red' peak of Fig. 2 could be interpreted as corresponding to a very efficient funneling process, the numerical solution of Eq. 1 exhibit funneling efficiency (defined as the fraction of all photoexcited ex- citons reaching the location of the AFM tip) that never exceeds 4%29. It is instructive to develop intuitive understanding for the observed low funneling efficiency. While the drift term in the Eq. 1 'pushes' the exciton towards the fun- nel center with the force proportional to ∇u(r), the dif- fusion term randomizes that motion (see Fig. 1(b)). The average distances travelled by an exciton during its life- Both the left-hand side and the right-hand side of this formula are plotted in Figure 4(b) as red dotted line and as black solid line respectively. We see that the condition above is only fulfilled for the small portion of the mem- brane (r < 250nm), and that diffusion term dominates the rest of the membrane leading to inefficient funneling. We therefore posit that funneling cannot be as efficient at room temperature as predicted in Ref. 18. We note that Eq. 2 also suggests that higher funneling efficiency may be possible at cryogenic temperatures. If funneling is so inefficient, what other physical mecha- nism is responsible for the data of Fig. 2? The hint comes from the data of sample C suggesting that the 'red' peak at high strain evolves from the trion peak at zero strain. To include the contribution of trions into our model, we again use the same Eq. 1, but with n(r) = nex(r)+ntr(r) where nex(ntr) is the exciton (negatively charged trion) density, respectively. Although there are differences in the physical constants D, µ, RA, τ between excitons and trions, we have found that the solution of n(r) does not change significantly in a broad range of possible values29, thus we used the same values for both species. It is easy to see that the density of trions near the de- vice center is expected to be strongly strain-dependent. Indeed, the relative densities of neutral and charged ex- citons depend on the density of background electrons (doping level) nb(r) in our device. While at zero strain background carriers are uniformly distributed through- out the device, the applied nonuniform strain lowers the top of the conduction band uc(r). Quantitatively, assuming that the density of background electrons is described by the Boltzmann distribution, we obtain nB(r) = N0e∆uc (r)/kB T 29. Here N0, which is strain de- pendent, represents the number of free carriers in the whole area of the strained membrane for any given time, and ∆uc(r) is change of the energy of the top of the con- duction band from the zero strain value29. The expres- sion above makes it clear that the electrons are effectively (cid:82) e∆uc (r)/kB T rdr 'funneled' towards the point of the highest strain at the center of the membrane. As a consequence, photoexcited neutral excitons present near the membrane center bind to free electrons forming trions. To quantitatively deter- mine the intensity of trion emission, Eq. 1 is solved for n(r) as before, and nex, ntr are determined from nB(r) using the law of mass action29,36,37. 5 Once the carriers densities nex, ntr are determined, the entire PL spectrum is calculated by the following expres- sion: (cid:90) ∞ (cid:104)P L(cid:105) = [P Lex(uex(r))nex(r)+ 0 P Ltr(utr(r))ntr(r)]rdr (3) Here P Lex(P Ltr) are the spectral lines of the excitons (trions), respectively and are taken from the Gaussian fits of the spectrum at zero strain. Figure 3 shows a comparison between the model (black dashed line), us- ing N0 as a single fit parameter, and the experimen- tal results (red line) for relevant strain and excitation intensities for all samples. The model is in much bet- ter agreement with the experimental data, especially in comparison with the model that does not include trion effects (thin blue dashed line). We therefore conclude that strain-related free carrier funneling, followed by con- version of neutral to negatively charged excitons, is the dominant process in our samples. To illustrate the mechanism responsible for the appear- ance of the strain-dependent trion contribution, we show in Fig. 4(c) the calculated spatial dependencies of neu- tral and charge exciton densities in Sample A (calcula- tions for other samples are shown in Ref. 29). We see that the trion density ntr(r) steadily increases and be- comes much larger than nex(r) towards the center of the funnel, consistent with funneling of free carriers to that region. At the same time, in the region of the sample that predominantly contributes to the observed signal due to the Jacobian rdr (defined as the shaded area in Fig. 4(a)), nex(r) > ntr(r). This explains comparable magnitudes of the trion and exciton peaks in Sample A (Fig. 3(a). In contrast, in samples B (low excitation) and C the doping level is higher. In that situation, we find29 that nex(r) (cid:28) ntr(r) in the relevant area of the de- vice, meaning that photoexcited neutral excitons are con- verted into trions with conversion efficiency approaching 100%. Our findings suggest several important implications. First, strain-dependent exciton-to-trion conversion pro- duces experimental signatures that may appear similar to that of neutral excitons funneling, but much stronger in amplitude. This suggests that previous reports of ex- citon funneling might have strongly overestimated its ef- ficiency. Second, while we experimentally realize, for the first time, the controlled funneling geometry of the the- oretical proposal18, the dominant process in such a de- vice is found to be diffusion rather than drift, at least Fig. 4 (a) The source term S(r) · r in Eq. 1 corresponding to the illumination spot in our experiments. The shaded region, de- fined as the region where the term falls to less than 1/2 from the maximal value, represents the sample area producing the dominant contribution to the measured PL. (b) The ratio between the drift and the diffusion terms in the Eq. 1 (solid black curve). Equation 2 indicates that when this ratio is higher than kBT = 25 meV (dashed red line), drift dominates over diffusion. Inset: A sketch of the forces acting on a neutral exciton in our straining conditions. (c) Normalized densities of neutral nex(r) (dotted blue curve), and charged excitons ntr(r) (solid red curve) calculated using the for- malism described in the text. Large ntr(r) near the center of the membrane reflects effective funneling of free electrons towards the point of the highest strain, followed by their binding into trions. Inset: A sketch of the carrier funneling and trion conversion pro- cesses. at room temperature. This means that the photoconver- sion mechanism proposed by Ref. 18 may not be feasi- ble. Finally, and perhaps most importantly, the strain- dependent exciton-to-trion conversion may constitute an- other, more efficient photoconversion mechanism com- pared to that of Ref. 18. We speculate that the energet- ics of energy-harvesting of weakly-bound trions may be advantageous to that of strongly-bound neutral excitons. To summarize, in this work we have presented a novel experimental setup that allows full dynamical control of strain amplitude and profile in optically-interrogated TMDC monolayers. We revealed that even in TMDCs strained to the point of breakage, the funneling of the excited carriers is not nearly as efficient as previously thought. On the other hand, we discovered that in the presence of non-uniform strain, another process, neutral- to-charged exciton conversion becomes dominant. It is noteworthy that this former process, while being phys- ically very different than the latter, can produce simi- lar experimental signatures leading to possible misinter- pretations. Finally, we note that in the future it will be especially interesting to study the role of funneling at cryogenic temperatures, where the role of diffusion is minimized. a)Electronic mail: [email protected] 1K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz, "Atomi- cally Thin MoS 2 : A New Direct-Gap Semiconductor," Physical Review Letters 105, 136805 (2010). 2W. Zhao, Z. Ghorannevis, L. Chu, M. Toh, C. Kloc, P.-H. Tan, and G. Eda, "Evolution of Electronic Structure in Atomically Thin Sheets of WS 2 and WSe 2," ACS Nano 7, 791 -- 797 (2013). 3X. Xu, W. Yao, D. Xiao, and T. F. Heinz, "Spin and pseudospins in layered transition metal dichalcogenides," Nature Physics 10, 343 -- 350 (2014). 4D. Xiao, G.-B. Liu, W. Feng, X. Xu, and W. Yao, "Coupled Spin and Valley Physics in Monolayers of MoS 2 and Other Group-VI Dichalcogenides," Physical Review Letters 108, 196802 (2012). 5K. F. Mak, K. L. McGill, J. Park, and P. L. McEuen, "Valleytron- ics. The valley Hall effect in MoS transistors." Science (New York, N.Y.) 344, 1489 -- 92 (2014). 6P. Tonndorf, R. Schmidt, R. Schneider, J. Kern, M. Buscema, G. A. Steele, A. Castellanos-Gomez, H. S. J. van der Zant, S. Michaelis de Vasconcellos, and R. Bratschitsch, "Single-photon emission from localized excitons in an atomically thin semicon- ductor," Optica 2, 347 (2015). 7C. Chakraborty, L. Kinnischtzke, K. M. Goodfellow, R. Beams, and A. N. Vamivakas, "Voltage-controlled quantum light from an atomically thin semiconductor," Nature Nanotechnology 10, 507 -- 511 (2015). 8M. Koperski, K. Nogajewski, A. Arora, V. Cherkez, P. Mallet, J.-Y. Veuillen, J. Marcus, P. Kossacki, and M. Potemski, "Sin- gle photon emitters in exfoliated WSe2 structures," Nature Nan- otechnology 10, 503 -- 506 (2015). 9A. Srivastava, M. Sidler, A. V. Allain, D. S. Lembke, A. Kis, and A. Imamoglu, "Optically active quantum dots in monolayer WSe2," Nature Nanotechnology 10, 491 -- 496 (2015). 10Y.-M. He, G. Clark, J. R. Schaibley, Y. He, M.-C. Chen, Y.-J. Wei, X. Ding, Q. Zhang, W. Yao, X. Xu, C.-Y. Lu, and J.-W. Pan, "Single quantum emitters in monolayer semiconductors," Nature Nanotechnology 10, 497 -- 502 (2015). 11T. Roy, M. Tosun, J. S. Kang, A. B. Sachid, S. B. Desai, M. Hettick, C. C. Hu, and A. Javey, "Field-Effect Transistors Built from All Two-Dimensional Material Components," ACS Nano 8, 6259 -- 6264 (2014). 12C.-H. Lee, G.-H. 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, and P. Kim, "Atomically thin pn junctions with van der Waals heterointerfaces," Nature Nanotechnology 9, 676 -- 681 (2014). 13D. Jariwala, A. R. Davoyan, J. Wong, and H. A. Atwater, "Van der Waals Materials for Atomically-Thin Photovoltaics: Promise and Outlook," ACS Photonics 4, 2962 -- 2970 (2017). 14T. Akama, W. Okita, R. Nagai, C. Li, T. Kaneko, and T. Kato, "Schottky solar cell using few-layered transition metal dichalco- genides toward large-scale fabrication of semitransparent and flexible power generator," Scientific Reports 7, 11967 (2017). 15K. Liu, Q. Yan, M. Chen, W. Fan, Y. Sun, J. Suh, D. Fu, S. Lee, J. Zhou, S. Tongay, J. Ji, J. B. Neaton, and J. Wu, "Elastic Prop- erties of Chemical-Vapor-Deposited Monolayer MoS ¡sub¿2¡/sub¿ , WS ¡sub¿2¡/sub¿ , and Their Bilayer Heterostructures," Nano Letters 14, 5097 -- 5103 (2014). 16C. Lee, X. Wei, J. W. Kysar, J. Hone, A. Zettl, F. Guinea, A. H. C. Neto, and M. F. Crommie, "Measurement of the elastic properties and intrinsic strength of monolayer graphene." Science (New York, N.Y.) 321, 385 -- 8 (2008). 17R. Rold´an, A. Castellanos-Gomez, E. Cappelluti, and F. Guinea, "Strain engineering in semiconducting two-dimensional crystals," Journal of Physics: Condensed Matter 27, 313201 (2015). 18J. Feng, X. Qian, C.-W. Huang, and J. Li, "Strain-engineered artificial atom as a broad-spectrum solar energy funnel," Nature Photonics 6, 866 -- 872 (2012). 19H. J. Conley, B. Wang, J. I. Ziegler, R. F. Haglund, S. T. Pan- telides, and K. I. Bolotin, "Bandgap Engineering of Strained Monolayer and Bilayer MoS 2," Nano Letters 13, 3626 -- 3630 (2013). 6 20I. Niehues, R. Schmidt, M. Druppel, P. Marauhn, D. Chris- tiansen, M. Selig, G. Berghauser, D. Wigger, R. Schneider, L. Braasch, R. Koch, A. Castellanos-Gomez, T. Kuhn, A. Knorr, E. Malic, M. Rohlfing, S. Michaelis de Vasconcellos, and R. Brats- chitsch, "Strain Control of ExcitonPhonon Coupling in Atomi- cally Thin Semiconductors," Nano Letters 18, 1751 -- 1757 (2018). 21D. Christiansen, M. Selig, G. Berghauser, R. Schmidt, I. Niehues, R. Schneider, A. Arora, S. M. de Vasconcellos, R. Bratschitsch, E. Malic, and A. Knorr, "Phonon Sidebands in Monolayer Tran- sition Metal Dichalcogenides," Physical Review Letters 119, 187402 (2017). 22D. Lloyd, X. Liu, J. W. Christopher, L. Cantley, A. Wadehra, B. L. Kim, B. B. Goldberg, A. K. Swan, and J. S. Bunch, "Band Gap Engineering with Ultralarge Biaxial Strains in Suspended Monolayer MoS 2," Nano Letters 16, 5836 -- 5841 (2016). 23S. Lazi´c, A. Violante, K. Cohen, R. Hey, R. Rapaport, and P. V. Santos, "Scalable interconnections for remote indirect exciton systems based on acoustic transport," Physical Review B 89, 085313 (2014). 24H. M. Manasevit, I. S. Gergis, and A. B. Jones, "Electron mo- bility enhancement in epitaxial multilayer SiSi ¡sub¿ 1 ¡i¿x¡/i¿ ¡/sub¿ Ge ¡sub¿ ¡i¿x¡/i¿ ¡/sub¿ alloy films on (100) Si," Applied Physics Letters 41, 464 -- 466 (1982). 25R. People, J. C. Bean, D. V. Lang, A. M. Sergent, H. L. Stormer, K. W. Wecht, R. T. Lynch, and K. Baldwin, "Modulation doping in Ge ¡sub¿ ¡i¿x¡/i¿ ¡/sub¿ Si ¡sub¿ 1 ¡i¿x¡/i¿ ¡/sub¿ /Si strained layer heterostructures," Applied Physics Letters 45, 1231 -- 1233 (1984). 26A. Castellanos-Gomez, R. Rold´an, E. Cappelluti, M. Buscema, F. Guinea, H. S. J. van der Zant, and G. A. Steele, "Local Strain Engineering in Atomically Thin MoS 2," Nano Letters 13, 5361 -- 5366 (2013). 27A. V. Tyurnina, D. A. Bandurin, E. Khestanova, V. G. Kravets, M. Koperski, F. Guinea, A. N. Grigorenko, A. K. Geim, and I. V. Grigorieva, "Strained Bubbles in van der Waals Heterostructures as Local Emitters of Photoluminescence with Adjustable Wave- length," ACS Photonics 6, 516 -- 524 (2019). 28A. Castellanos-Gomez, M. Buscema, R. Molenaar, V. Singh, L. Janssen, H. S. J. van der Zant, and G. A. Steele, "Determin- istic transfer of two-dimensional materials by all-dry viscoelastic stamping," 2D Materials 1, 011002 (2014). 29"See Supplementary Information available online," See Supple- mentary Information available online. 30M. Dukic, J. D. Adams, and G. E. Fantner, "Piezoresistive AFM cantilevers surpassing standard optical beam deflection in low noise topography imaging," Scientific Reports 5, 16393 (2015). 31A. Castellanos-Gomez, M. Poot, G. A. Steele, H. S. J. van der Zant, N. Agraıt, and G. Rubio-Bollinger, "Elastic Properties of Freely Suspended MoS2 Nanosheets," Advanced Materials 24, 772 -- 775 (2012). 32R. Zhang, V. Koutsos, and R. Cheung, "Elastic properties of sus- pended multilayer WSe2," Applied Physics Letters 108, 042104 (2016). 33D. Vella and B. Davidovitch, "Indentation metrology of clamped, ultra-thin elastic sheets," Soft Matter 13, 2264 -- 2278 (2017). 34M. Kulig, J. Zipfel, P. Nagler, S. Blanter, C. Schuller, T. Korn, N. Paradiso, M. M. Glazov, and A. Chernikov, "Exciton Diffusion and Halo Effects in Monolayer Semiconductors," Physical Review Letters 120, 207401 (2018). 35D. Ovchinnikov, A. Allain, Y.-S. Huang, D. Dumcenco, and A. Kis, "Electrical Transport Properties of Single-Layer WS ¡sub¿2¡/sub¿," ACS Nano 8, 8174 -- 8181 (2014). 36J. Siviniant, D. Scalbert, A. V. Kavokin, D. Coquillat, and J.-P. Lascaray, "Chemical equilibrium between excitons, elec- trons, and negatively charged excitons in semiconductor quan- tum wells," Physical Review B 59, 1602 -- 1604 (1999). 37J. S. Ross, S. Wu, H. Yu, N. J. Ghimire, A. M. Jones, G. Aivazian, J. Yan, D. G. Mandrus, D. Xiao, W. Yao, and X. Xu, "Electrical control of neutral and charged excitons in a monolayer semicon- ductor," Nature Communications 4, 1474 (2013).
1710.02001	3	1710	2017-11-16T19:59:17	On the bulk boundary correspondence and the existence of Majorana bound states on the edges of 2D topological superconductors	[ "cond-mat.mes-hall" ]	The bulk-boundary correspondence establishes a connection between the bulk topological index of an insulator or superconductor, and the number of topologically protected edge bands or states. For topological superconductors in two dimensions the first Chern number is related to the number of protected bands within the bulk energy gap, and is therefore assumed to give the number of Majorana band states in the system. Here we show that this is not necessarily the case. As an example we consider a hexagonal-lattice topological superconductor based on a model of graphene with Rashba spin orbit coupling, proximity induced s-wave superconductivity, and a Zeeman magnetic field. We explore the full Chern number phase diagram of this model, extending what is already known about its parity. We then demonstrate that despite the high Chern numbers that can be seen in some phases these do not strictly always contain Majorana bound states.	cond-mat.mes-hall	cond-mat	a On the bulk boundary correspondence and the existence of Majorana bound states on the edges of 2D topological superconductors Nicholas Sedlmayr,1, 2, ∗ Vardan Kaladzhyan,3, 4 Cl´ement Dutreix,5, 6 and Cristina Bena3 1Department of Physics and Medical Engineering, Rzesz´ow University of Technology, al. Powsta´nc´ow Warszawy 6, 35-959 Rzesz´ow, Poland 2Department of Physics and Astronomy, Michigan State University, East Lansing, Michigan 48824, USA 3Institute de Physique Th´eorique, CEA/Saclay, Orme des Merisiers, 91190 Gif-sur-Yvette Cedex, France 4Laboratoire de Physique des Solides, CNRS, Univ. Paris-Sud, Universit´e Paris-Saclay, 91405 Orsay Cedex, France 5Univ. Bordeaux, LOMA, UMR 5798, Talence, France and CNRS, LOMA, UMR 5798, Talence, F-33400, France 6Univ. Lyon, Ens de Lyon, Univ Claude Bernard, CNRS, Laboratoire de Physique, F-69342 Lyon, France (Dated: November 20, 2017) The bulk-boundary correspondence establishes a connection between the bulk topological index of an insulator or superconductor, and the number of topologically protected edge bands or states. For topological superconductors in two dimensions the first Chern number is related to the number of protected bands within the bulk energy gap, and is therefore assumed to give the number of Majorana band states in the system. Here we show that this is not necessarily the case. As an example we consider a hexagonal-lattice topological superconductor based on a model of graphene with Rashba spin orbit coupling, proximity induced s-wave superconductivity, and a Zeeman magnetic field. We explore the full Chern number phase diagram of this model, extending what is already known about its parity. We then demonstrate that despite the high Chern numbers that can be seen in some phases these do not strictly always contain Majorana bound states. I. INTRODUCTION The search for Majorana bound states (MBS) in con- densed matter systems1 has already produced a large volume of theoretical work1–7 and promising, though not conclusive, experiments.8–15 A one dimensional (1D) topologically non-trivial superconductor will have Majo- rana bound states present at its ends. This is the result of the well known bulk-boundary correspondence,16 which relates the topological invariant to the number of topo- logically protected edge states. Two-dimensional topo- logical superconductors, including for example p-wave or s-wave pairing, have also received a lot of attention,3,17–30 and including the model we consider here of a system with spin-orbit coupling, in the presence of superconduct- ing proximity and a Zeeman field, as well as other related models.31–40. In two dimensions (2D) the bulk-boundary correspondence relates the Chern number, ν, to the num- ber of protected bands connecting the bulk states above and below the gap41–45 arising in a ribbon structure. These bands correspond to protected edge states and it is often assumed that their zero-energy crossing cor- responds to the formation of a MBS.28,29,38 Indeed the usual expression of the bulk boundary correspondence is in terms of zero energy modes, rather than protected bands.41 However as we will demonstrate this is not nec- essarily the case, and a topologically protected band does not necessarily contain a MBS. As an example we focus on a single-layer-hexagonal topological superconductor. This allows us to easily con- sider two very different types of boundary, both zig-zag and armchair edges. We calculate explicitly the Chern number and we construct a full phase diagram based on the value of the Chern number, and not only on its parity as has been done previously40 We identify the band struc- tures corresponding to each value of the Chern number and we confirm that it is equal to the number of protected edge bands. We show that the Chern number can be changed by gap closings at many points in the Brillouin zone (BZ), however only gap closings at specific points in the BZ can lead to the formation of protected MBS. The Chern number itself varies from -5 to 5. The number of MBS on an edge can vary however from 0 to 3, rather than 0 to 5, and depends on the type of nanoribbon one is considering. One recent work reported MBS near the Dirac points of this model,38 in a phase with ν = 4. Such states were not found in other studies.27,33,34,36,37,40 Here we will clarify that although there are protected bands in this phase, strictly speaking there are no MBS present in the lattice model in this phase. To understand why this is the case we will introduce a more careful definition of MBS. Thus we show that one must additionally consider if the states found near zero energy scale exponentially to zero in the thermody- namic limit. Also we test the low-energy states in terms of their Majorana-like properties; to this end we use the Majorana polarization,46–48 a direct local check of the Majorana nature of an eigenstate. Also we provide sym- metry arguments, as well as study the specific manner in which the gap closing may occur in order to give rise to a change in the number of MBS. Finally we consider the effects of disorder on the formation of MBS. This article is organized as follows. In Sec. II we in- troduce the model of a two-dimensional hexagonal lattice with induced superconducting proximity, spin-orbit cou- pling and a Zeeman magnetic field perpendicular to the plane. In Sec. III we calculate numerically the Chern number for this model. In Sec. IV we calculate its par- ity analytically. In Sec. V we consider the bulk-boundary correspondence and how the Chern number relates to the formation of topologically protected bands and of MBS. We conclude in Sec. VI. II. MODEL 2 (cid:88) (cid:88) σ <i,j> <i,j> σ,σ(cid:48) (cid:88) (cid:104) (cid:105) The model under consideration is a hexagonal-lattice topological superconductor that can be realized in graphene with Rashba spin orbit coupling α, proxim- ity induced s-wave superconductivity ∆, and a Zeeman magnetic potential B. We define t as the strength of the nearest-neighbor hopping, and µ is the chemical poten- tial. Then the Hamiltonian can be written as † ((cid:126)δij × (cid:126)σσσ(cid:48))zc H = −t iσcjσ(cid:48) † iσcjσ − iα (cid:88) (1) c + † iσ [Bσz σσ − µ] ciσ + ∆ c † i↑c † i↓ + ci↓ci↑ c . i,σ i where c(†) iσ denotes the annihilation (creation) operator of an electron of spin σ at site i, and (cid:126)δij are the nearest neighbor vectors. This model has been carefully stud- ied both in the presence of a superconducting proximity effect36–40, as well as without it31,32. We will define the lattice spacing a = 1 and = 1 throughout. The lattice has a length of L(cid:107) unit cells along the nanoribbon di- rection, which always has periodic boundary conditions; and a width of Lw unit cells, which can have periodic or open boundary conditions in the numerical simula- tions. We are interested in nanoribbons with open edges aligned along both the armchair and zigzag directions of the hexagonal lattice. III. NUMERICAL CALCULATION OF THE CHERN NUMBER √ In Eq. (7) in Ref. 38 the authors give their results for the Chern number for this model, and show that it can reach values of up to 4, in one region of phase space. They note that, the gap closings at the Dirac points (cid:126)k = (±4π/3 3, 0) in the BZ need to be taken into account, √ in addition to those at the time reversal invariant (TRI) √ Γ points, Γ0 = (0, 0), Γ1 = (0, 2π/3), Γ2 = (π/ 3, π/3), 3,−π/3), which were already known .36,37 and Γ3 = (π/ Here we will show explicitly what the topological phase diagram looks like. In fact there are other points in the BZ where a gap closing changes the Chern number. Following Ref. 49 the Chern number, or equivalently the Thouless-Kohmoto-Nightingale-den Nijs (TKNN) FIG. 1. (Color online) Numerically determined topological phase diagrams for Eq. (1) using Eq. (2), where ν is the Chern number. The parameters are: (a) α = ∆ = 0.5t; (b) 3α = ∆ = 1.5t; (c) α = 0.5t and ∆ = 0.4t; and (d) α = ∆ = 0.1t. Solid black lines show the phase boundaries caused by the gap closings at the TRI momenta and the blue dashed line corresponds to the gap closes at the Dirac point. These are not however the only phase boundaries. (cid:90) ν = i 8π2 invariant50, can be calculated numerically using d2k dω Tr(cid:2)G2(∂ky H)G(∂kx H) −G2(∂kxH)G(∂ky H)(cid:3) , (2) where G = (H − iω)−1 and (cid:126)k = (kx, ky) is the momen- tum. This can be implemented numerically, and some examples are given in Fig. 1. The Chern number can be as large as -5, which is unusually high for such a model. The phase diagrams in Fig. 1 show changes in Chern numbers away from the analytically calculated bulk gap closing lines for the Γ and Dirac points. To be certain that this is not a numerical error we track the bulk gap across some of these transitions, see Fig. 2. For all the changes in the Chern number we can see that the gap does close at some point in the BZ as required. One can also explicitly check that the bulk-boundary correspon- dence holds, demonstrating that these regions are not caused by numerical errors. However, as we shall see in what follows, these high Chern numbers do not necessar- ily lead to large numbers of protected MBS. In Fig. 1 it can be observed that the gap closings at the Γ points alter the Chern number from odd to even or vice versa. In the following section we will demonstrate why this occurs. We note that this model can still be a topological in- sulator when ∆ = 0, though this is by no means guar- anteed. However there is no general connection between the protected edge modes we see here and those in the topological insulator. In some cases the edge modes per- -5-4-3-2-101234ν0123401234B[t]μ[t](a)-5-4-3-2-101234ν0123401234B[t]μ[t](b)-5-4-3-2-101234ν0123401234B[t]μ[t](c)-5-4-3-2-101234ν0123401234B[t]μ[t](d)In momentum space the total Hamiltonian (1) is (cid:88) (cid:126)k H = 1 2 † Ψ (cid:126)k H((cid:126)k)Ψ(cid:126)k 3 (3) † Ψ (cid:126)k (cid:19) (cid:18) H((cid:126)k) field with fermion = † Here a (cid:126)kσ create electrons of spin σ and with momentum the † † † † (cid:126)k↑, b (cid:126)k↓, a−(cid:126)k↑, b−(cid:126)k↑, a−(cid:126)k↓, b−(cid:126)k↓). (cid:126)k↓, b (cid:126)k↑, a (a † and b (cid:126)kσ (cid:126)k on the two sublattices. The Hamiltonian matrix is H((cid:126)k) = ∆((cid:126)k) −∆∗((cid:126)k) −H∗(−(cid:126)k) , (4) where each entry is itself a 4 × 4 matrix. The pairing matrix satisfies ∆∗(−(cid:126)k) = ∆((cid:126)k) and the Hamiltonian matrix satisfies particle hole symmetry: C†H((cid:126)k)C = −H∗(−(cid:126)k) , (5) where C = σ0 ⊗ λ0 ⊗ τ x. We will use σ, λ, and τ Pauli matrices for the spin, sublattice, and particle-hole sectors and σ0 = λ0 = τ 0I2, the 2 × 2 identity matrix. One consequence of the particle hole symmetry is that all the non-zero energy states are paired with a state of opposite energy. The topological invariant can be related to a parity-like operator, P = σx ⊗ λz ⊗ τ z, of the negative energy bands at the time reversal invariant momenta.51 All eigenstates at the points, Ψn(Γi), have a definite parity πn(Γi) = ±1.40 Note that a sign change in (cid:89) (cid:90) δi = πn(Γi) , (6) En<0 where En is the eigenenergy of the eigenstates Ψn(Γi), implies a gap closing at zero energy. The Chern number can be defined as the integral of the Berry curvature over the Brillouin zone for the negative energy bands ν = 1 2π d2k ∇(cid:126)k × A−((cid:126)k) (7) Where the Berry connection is (cid:88) (cid:104)Ψn((cid:126)k)∇(cid:126)kΨn((cid:126)k)(cid:105) = A−((cid:126)k) = i n ∇(cid:126)k ln Det M ((cid:126)k) , i 2 (8) with Mmn = (cid:104)Ψm((cid:126)k)PCΨn((cid:126)k)(cid:105) and Ψn((cid:126)k)(cid:105) an eigen- state of the Hamiltonian. Thus one finds (cid:89) (−1)ν = πn(Γ0)πn(Γ1)πn(Γ2)πn(Γ3) (9) En<0 for the topological invariant. This results in40: FIG. 2. (Color online) Numerically determined topological phase diagrams for Eq. (1) using Eq. (2), where ν is the Chern number (red squares), compared to the gap closing (blue circles) where G is the gap. (a) shows a transition from ν = −2 → 4 as a function of B (accompanied by a gap closing at the Dirac point) with the gap calculated for a system of length and width L = Lw = L(cid:107) = 500. The parameters are ∆ = α = 0.5t and µ = 0. This is a cut through the phase diagram in Fig. 1(a). (b) shows transitions through ν = 1 → −5 → −2 as a function of α (accompanied by gap closings at points in the BZ) with the gap calculated for L = 500 as for (a). The parameters are ∆ = 0.6t B = 1.5t and µ = 1.3t. sist and remain topologically protected, but this is cer- tainly not a generic feature. In general as ∆ → 0 the gap can close and reopen, changing the topology, and some gapped phases here simply become ungapped metallic phases in this limit. IV. PARITY OF THE CHERN NUMBER √ 3, π/3), and Γ3 = (π/ A detailed calculation of the parity of the Chern num- ber for Eq. (1) has been performed and presented in detail in Refs. 40 and 37, and an example is available also in the appendix of Ref. 27. We show that a rela- tion between the Chern number and the parity of the √ bands at the TRI momenta, Γ0 = (0, 0), Γ1 = (0, 2π/3), 3,−π/3), can be Γ2 = (π/ √ proven for Eq. (1). Here we have set the lattice spacing a = 1. The K = (4π/3 3, 0) Dirac points do not affect the parity of the Chern num- ber, and neither do any of the other gap closing points. This is due to the fact that they always paired, with the gaps at (cid:126)k and −(cid:126)k necessarily closing at the same time, thus the parity of the Chern number is only altered by the special Γ points. However, as we have seen, the Chern number itself can be changed by gap closings at various points in the BZ. √ 3, 0) and K(cid:48) = (−4π/3 ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■(a)0.70.80.91.-2024-2024B[t]100ϵG/tν●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■(b)0.20.40.60.8-4-20246-5-20246α[t]100ϵG/tν(−1)ν = sgn(cid:2)(cid:0)B4 + 92t4 + 18t2[∆2 − µ2] + [∆2 + µ2]2 − 2B2(9t2 + ∆2 + µ2)(cid:1) ×(cid:0)B4 + t4 + 16α4 + 8α2[∆2 − µ2] + ∆4 + 2∆2µ2 + µ4 + 2t2[4α2 + ∆2 − µ2] − 2B2[t2 − 4α2 + ∆2 + µ2](cid:1)(cid:3) . 4 (10) FIG. 3. (Color online) Topological phase diagrams for Eq. (1). The parameters are (a) 3α = ∆ = 1.5t, and (b) α = ∆ = 0.1t. Compare with Fig. 1(b,d) respectively. The red regions are satisfy (−1)ν = −1 and the white regions satisfy (−1)ν = 1, the solid black lines show the phase boundaries and ν is the Chern number. Two exemplary phase diagrams as a function of magnetic field and chemical potential are shown in Fig. 3, which are consistent with the results in Fig. 1. V. NANORIBBON BANDSTRUCTURES AND MAJORANA BOUND STATES A. Correspondence between the Chern number and the band structure Before we consider the formation of MBS, we will demonstrate in what way the bulk-boundary correspon- dence manifests itself in this system. The nanoribbons we consider are periodic in one direction and finite with open boundary conditions in the perpendicular direction. We define k(cid:107) ∈ [−π, π) as the momentum parallel to the edges. We will consider several examples of ribbons, with both zigzag and armchair edges. Figs. 4(a,b) correspond to a phase with ν = 4. In both cases we observe four pairs of edge bands crossing the bulk gap. The energy of these bands has a monotonic dependence on k(cid:107), with the right-moving states being located on one edge, and the left-moving ones on the other. Figs. 4(c,d) correspond to a phase with ν = −5, and we observe five pairs of protected bands. Fig. 5(a,b) correspond to a phase with ν = −2. Both nanoribbons exhibit two pairs of protected bands crossing zero en- ergy, thus expected to support MBS, however for the armchair nanoribbon, see Fig. 5(b), one can see that a pair of bands exhibits additional unprotected zero energy crossings close to k(cid:107) = ±0.5π. Such crossings are unpro- tected in the sense that a continuous deformation of the band can remove these zero energy crossings, and the cor- FIG. 4. (Color online) (a,b) The bandstructures of zigzag (a) and armchair (b) nanoribbons in a regime with ν = 4. The parameters are α = ∆ = 0.5t, µ = 0.1t, and B = 1.4t. (c,d) The bandstructures corresponding to ν = −5 for zigzag (c) and armchair (d) nanoribbons. The parameters are α = ∆ = 0.5t, µ = 1.5t, and B = 1.3t. In (d) three pairs of edge bands are crossing at k(cid:107) = 0 and there are three MBS per edge in this case. The topologically protected bands localized on one edge are depicted by a dashed red line, while those localized on the other edge are represented by a dotted green line. The K and K' points are marked in the figures for reference. FIG. 5. (Color online) (a,b) The bandstructure of zigzag (a) and armchair (b) nanoribbons in a regime with ν = −2. The parameters are α = ∆ = µ = 0.5t, and B = 1.5t. In (b) extra unprotected crossings can be seen at k(cid:107) (cid:54)= 0 which can be removed by continuously deforming the bands, see Fig. 7(a). responding zero energy states are not topologically pro- tected MBS, i.e. a perturbation can gap out these states. Indeed, this can be seen for example in Fig. 7 where we study the effects of disorder on the band structure. B. Zero-energy states and their identification as MBS based on scaling arguments Given the arguments in the previous section, we note that each topologically protected edge band has one pro- tected zero-energy crossing. We would thus naturally ex- pect that each edge band gives rise to a MBS, and thus (a)KK'-1.0-0.50.00.51.0-0.3-0.2-0.10.00.10.20.3k/πϵ[t](b)K,K'-1.0-0.50.00.51.0-0.3-0.2-0.10.00.10.20.3k/πϵ[t](c)KK'-1.0-0.50.00.51.0-0.100.1k/πϵ[t](d)K,K'-1.0-0.50.00.51.0-0.100.1k/πϵ[t](a)KK'-1.0-0.50.00.51.0-0.3-0.2-0.10.00.10.20.3k/πϵ[t](b)K,K'-1.0-0.50.00.51.0-0.3-0.2-0.10.00.10.20.3k/πϵ[t]that the Chern number gives the number of topologically protected MBS. However we will argue in what follows that this is not always the case. We first note that the band structure is the result of a Fourier transform of Eq. (1) along the direction parallel to the edge of the ribbon, along which we have imposed periodic boundary conditions. This yields a set of 1D Hamiltonians, H1D(k(cid:107)), labelled by the quantum num- bers k(cid:107) = 2πn/L(cid:107), with L(cid:107) being the length of the ribbon in the direction parallel to edges, and n = 0, 1, . . . L(cid:107). In the thermodynamic limit k(cid:107) becomes a continuous vari- able. In order that a MBS forms, one edge band needs to contain a state with exactly zero energy. While this is of course automatic in the thermodynamic limit, in the finite-size system, when k(cid:107) is only taking discrete values, this can only happen at special points in the bandstruc- ture, here for example at k(cid:107) = 0, π, since the energy of a state corresponding to an arbitrary k(cid:107) = 0, π is never exactly zero, but is of the order of 1/L(cid:107). We propose that the deciding difference between real MBS and non-MBS states lies in how their energy scales to zero in the ther- modynamic limit, with the energy of the real MBS de- creasing exponentially, while the energy of the non-MBS decreasing inverse proportionally to the system size. To exemplify this we note that the energy of the states at 0 and π only depends on the width of the ribbon, be- ing due to the exponential overlap of the two MBS on the edges, and is given by52 0,π ∼ e−Lw/ L, with L be- ing the localization length of the MBS and Lw is the width of the ribbon. Hence, limLw=L(cid:107)=L→∞ 0,π/λ = limL→∞ L2 e−L/ L /4π2 = 0, with λ = 4π2/L2 is the mean level spacing. However, for the bands crossing zero away from k(cid:107) = 0, π, the lowest energy states typi- cally have a dominant contribution ∼ 2π/L(cid:107), and thus obey limLw=L(cid:107)=L→∞ /λ ∼ L. Thus these states never appear as exact-zero energy states. An example for a ν = −5 phase with one MBS and 4 additional zero en- ergy crossings is shown in Fig 6, the same parameters as for Fig. 4(c) are used. The crossing at k(cid:107) = 0, depicted in blue, shows clear exponential scaling to zero relative to the mean level spacing consistent with /λ ∼ L2 e−L/ L. The alternative crossings, depicted in red and black, do not scale to zero, but rather diverge as a power law con- sistent with /λ ∼ L. C. Gap closing arguments Moreover, we can find additional arguments for the al- lowed values of momentum for which one can form MBS. Thus, considering k(cid:107) as a parameter, if there are MBS present for a particular value of k(cid:107), one of two things must occur at this particular value of k(cid:107). Either the band gap must close or the symmetry of the model must change at these points. In the absence of the first possi- bility, when the 2D model is fully gapped for a given set of parameters, MBS can only exist at the high symmetry 5 FIG. 6. (Color online) The energy scaling of the lowest energy state in each band normalized by the mean level spacing, /λ. Here we consider a zigzag nanoribbon (left hand panel) and an armchair nanoribbon (right hand panel) in a ν = −5 phase, same as in Figs. 4(c,d). We take L = Lw = L(cid:107) to be the length and width of system. The low energy states at k(cid:107) = 0 (denoted in blue), show an exponential scaling to zero as /λ ∼ L2 e−L/ L (dashed blue line), the other low energy states at k(cid:107) (cid:54)= 0, π, (denoted in red and black) do not scale to zero but show a weaker positive divergence, ln /λ ∼ L (dashed red and black lines). points, which for the current model are k(cid:107) = 0, π. How- ever situations in which the bulk gap closes as a function of k(cid:107) and for which MBS can exist at arbitrary values of k(cid:107) may occur. One example is presented in Ref. 27 for a slightly different model, allowing for example for the formation of flat band of MBS in between these special k(cid:107) points at which the gap closes. The variation of the Chern number with the param- eters in our model (B and ∆) was described in Fig. 1. However, we note that the change in the Chern number, while indicating that the number of edge bands is chang- ing, is not always equivalent to a change in the num- ber of the real MBS states. We argue that a change in the Chern number corresponding to a gap closing at any point which does not correspond to the (1D) TRI mo- menta cannot change the number of MBS. This includes the Dirac points for the case of a zigzag nanoribbon. In the bandstructures in Fig. 4(c,d) for which we observe 1 MBS on the zigzag edge and 3 on the armchair edge, the 2 additional MBS in the armchair case originate in the gap closing at the Dirac points. Such gap closing at the Dirac point do correspond to a change in the number of MBS for an armchair ribbon, as in this configuration this corresponds to k(cid:107) = 0. Nevertheless, since the existence of such states depends on the direction of the ribbon, the extra MBS are not fully stable, but rather an example of weak topology. The Γ points in the 2D BZ are TRI momenta for both types of nanoribbon, and therefore a closing of the gap at these points yields a change in the number of MBS. D. Majorana polarization arguments To test whether the edge states are MBS we can use, along with the energy of the states, the Majorana po- larization vector C((cid:126)r) = (cid:104)ΨC rΨ(cid:105).39,46–48,53 As an MBS state is an eigenstate of the particle hole operator C, a ●●●●●●●●●●■■■■■■■■■■■◆◆◆◆◆◆◆◆◆◆◆00.00020.00040.0006-30-20-100L-2ln[ϵ/λ]●●●●●●●●●●●■■■■■■■■■■■◆◆◆◆◆◆◆◆◆◆◆▲▲▲▲▲▲▲▲▲▲▲00.00020.00040.0006-10-505L-2ln[ϵ/λ]Majorana-like state localized inside a spatial region R must satisfy C = 1 where C is the normalized magnitude of the integral of the Majorana polarization vector over the spatial region R: C = . (11) (cid:126)r∈R(cid:104)ΨC rΨ(cid:105)(cid:12)(cid:12) (cid:12)(cid:12)(cid:80) (cid:80) (cid:126)r∈R(cid:104)ΨrΨ(cid:105) (cid:88) (cid:88) Here r is the projection onto site (cid:126)r, and the local Majo- rana polarization vector C((cid:126)r) is simply the expectation value of the local particle-hole transformation: C((cid:126)r) = (cid:104)ΨC rΨ(cid:105) = −2 σu(cid:126)rσv(cid:126)rσ . (12) σ Here we have written the real space wavefunction in Nambu space as Ψ(cid:126)r = (u(cid:126)r↑, u(cid:126)r↓, u(cid:126)r↓, v(cid:126)r↑). We note that in momentum space this mixes eigenstates of different momenta as the conjugation in the particle-hole trans- formation obeys Kψ((cid:126)k) = ψ†(−(cid:126)k) where ψ((cid:126)k) is a wave- function in momentum space. In our case where we are interested in nanoribbons we have wavefunctions in a mixture of representations with both spatial and momentum dependence. We then find that C((cid:126)r) = −2 σux,k(cid:107),σvx,−k(cid:107),σ , (13) σ wavefunctions = for (ux,k(cid:107),↑, ux,k(cid:107),↓, ux,k(cid:107),↓, vx,k(cid:107),↑) with x the position and k(cid:107) the momentum. Ψx,k(cid:107) given by This is a direct test of whether the states in question are MBS and we will apply this test to our candidate MBS states. Thus, for the examples in Fig. 4, see the Tables I and II where we list the energies and the Ma- jorana polarizations for the lowest positive energy states corresponding to various bands. We note that the states at k(cid:107) = 0 have the lowest energies and a C = 1, in- dicative of being MBS, and consistent also with previous arguments. The other states have smaller C's, however this does not automatically imply that they are not MBS since they usually come in pairs, and as the spectrum is de- generate we could also consider combinations of states that can give rise to a maximal C.51,54 One can rule out their MBS character via disorder tests, such as is the case for the non-protected crossings which occur in Fig. 5(b). However this is not so straightforward for other states. Among the possible combinations we can take we can for example consider ψ(k(cid:107)) and ψ(−k(cid:107)), however, as these states are localized on different edges in this chiral sys- tem it does not increase C (which is indicative of the electron-hole overlap)46–48. Alternatively one can com- bine states on the same edge, which belong to different bands, for example the left most and right most green bands in Fig. 4(a). Indeed in this case one obtain a larger Majorana polarization, reaching values of C ≈ 0.9 for this example (with L(cid:107) = 400 and Lw = 160). How- ever for a finite-size system these states do not have the ν 4 4 −5 −5 −5 −5 Edge µ ZZ AC ZZ ZZ AC AC 0.1t 0.1t 1.3t 1.3t 1.3t 1.3t 5 B 110 k(cid:107) − 4π 1.4t 1.4t − 29π 1.5t 0 1.5t − 11π 1.5t 0 1.5t − 29π 20 200 6 6.81 · 10−4t 1.61 · 10−3t O(10−13)t 3.84 · 10−4t O(10−8)t 1.19 · 10−3t C 0.0725 0.404 1 0.613 1 0.218 TABLE I. The energies and Majorana polarization C for the lowest positive energy states corresponding to the different band crossings in Fig. 4. We take the length of the system to be L(cid:107) = 400 unit cells and its width Lw = 160. ZZ refers to a zigzag nanoribbon and AC to an armchair nanoribbon. Note that the two additional crossings for k(cid:107) = 0 in the last armchair case also have an exponentially small energy and C = 1. ν −2 −2 −2 Edge µ ZZ AC AC 0.5t 0.5t 0.5t k(cid:107) B 1.5t − 41π 1.5t 0 1.5t − 11π 100 60 1.03 · 10−4t O(10−7)t 1.68 · 10−4t C 0.392 1 0.491 TABLE II. The energies and Majorana polarization C for the lowest positive energy states corresponding to the different band crossings in Fig. 5. The energies for the additional band crossings at k(cid:107) = 0 for the armchair case are exponentially small, and the corresponding states have C = 1. same energy (here the energy difference between them is 0.00481t), neither are exact combinations of k(cid:107) and −k(cid:107), so they are not exact particle-hole eigenstates. Nonethe- less in the continuum limit their superposition will be- come an eigenstate of the particle-hole operator, and their energies become degenerate and equal to zero, so in this limit we cannot distinguish between the real MBS and the non-MBS, and we need to refer back to our scal- ing arguments to differentiate them. E. Disorder To further test the nature of the bands crossing be- tween the bulk states we introduce some disorder. We consider an onsite electronic potential which fluctu- ates randomly along the direction perpendicular to the nanoribbon edge and taking value between −s → s, thus ensuring that k(cid:107) remains a good quantum number. In Fig. 7 we present two exemplary cases, with the same parameters as Fig. 4(b) and Fig. 5(b) for s = 0.1t. Typ- ically in clean systems the bands of left and right mov- ing electrons cross zero energy and each other simultane- ously. In the disordered case this is no longer the case, and indeed for this particular form of disorder, left- and 7 edge bands predicted by the bulk-boundary theorem for a 2D topological superconductor, and the number of MBS present along nanoribbon edges. We show that for a lattice model, contrary to expectations, a topologically protected band crossing zero energy does not necessar- ily contain a state which has full MBS properties. We illustrate this point with several examples for a topologi- cal superconductor on a hexagonal lattice. The existence of an MBS inside a given edge band is quite subtle and we provided various arguments that allow one to identify a real MBS from a non-MBS zero-energy crossing. We showed for example that for this model the real MBS ap- pear at high-symmetry points in the band structure, in this case the TRI momenta, while zero-energy crossings occurring at arbitrary momenta can be characterized by scaling arguments to be non-MBS. Thus, the edge states forming close to Dirac points, and reported to be MBS in Ref. 38 are not actually real MBS. An open question is whether the results presented here also apply to systems which do not allow for a labelling by a transverse ribbon momentum, such as systems con- taining vortices, or presenting inhomogeneity either in the bulk or at the edges. While this work was being finalized we became aware of a related work56 studying the relationship between Chern numbers and MBS in p-wave superconductors. ACKNOWLEDGMENTS FIG. 7. (Color online) (a,b) The bandstructures of two arm- chair nanoribbons: (a) in a regime with ν = −2 and (b) in a regime with ν = 4. The parameters are the same as for Fig. 5(b) and Fig. 4(b) respectively with a disorder strength of s = 0.3t (a) and s = 0.1t (b). The bulk bands are those for the clean systems, provided here for reference. right-movers become mixed, see Fig. 7 and the band are shifted up and down in energy. However, we cannot ar- gue that this corresponds to an obvious destruction of the MBS, the low-energy crossings preserve the same charac- ter as in the non-disordered case and the same arguments as above can be applied to justify that the low-energy states are not real MBS. Nevertheless we can find some zero-energy states that are destroyed by disorder, for example in Fig. 7(a) we can see that the non-protected crossings described in Fig. 5(b) are actually gapped when introducing disor- der, confirming the fact that if a band can be continually deformed to eliminate a zero-energy crossing, such cross- ing is not topologically protected, and thus does not give rise to stable MBS.55 VI. DISCUSSION AND CONCLUSIONS We have analyzed the relation between the Chern num- ber or equivalently the number of topologically protected Support for this research at Michigan State University (N.S.) was provided by the Institute for Mathematical and Theoretical Physics with funding from the office of the Vice President for Research and Graduate Studies. ∗ [email protected] 1 X.-L. Qi and S.-C. Zhang, Rev. Mod. Phys. 83, 1057 (2011). 2 A. Y. Kitaev, Physics-Uspekhi 44, 131 (2001). 3 L. Fu and C. L. Kane, Phys. Rev. Lett. 100, 096407 (2008). 4 L. Fu and C. L. Kane, Phys. Rev. Lett. 102, 216403 (2009). 5 M. Sato, Y. Takahashi, and S. Fujimoto, Phys. Rev. Lett. 103, 020401 (2009). 6 R. M. Lutchyn, J. D. Sau, and S. Das Sarma, Phys. Rev. Lett. 105, 077001 (2010). 7 Y. Oreg, G. Refael, and F. von Oppen, Phys. Rev. Lett. 105, 177002 (2010). 8 V. Mourik, K. Zuo, S. M. Frolov, S. R. Plissard, E. P. and L. P. Kouwenhoven, Science 336, A. M. Bakkers, 1003 (2012). 9 M. T. Deng, C. L. Yu, G. Y. Huang, M. Larsson, P. Caroff, and H. Q. Xu, Nano Letters 12, 6414 (2012). 10 A. Das, Y. Ronen, Y. Most, Y. Oreg, M. Heiblum, and 11 E. J. H. Lee, X. Jiang, M. Houzet, R. Aguado, C. M. Lieber, and S. De Franceschi, Nature Nanotechnology 9, 79 (2014). 12 S. Nadj-Perge, I. K. Drozdov, J. Li, H. Chen, S. Jeon, J. Seo, A. H. MacDonald, B. A. Bernevig, and A. Yazdani, Science 346, 602 (2014). 13 M.-X. Wang, C. Liu, J.-P. Xu, F. Yang, L. Miao, M.-Y. Yao, C. L. Gao, C. Shen, X. Ma, X. Chen, Z.-A. Xu, Y. Liu, S.-C. Zhang, D. Qian, J.-F. Jia, and Q.-K. Xue, Science 336, 52 (2012). 14 J.-P. Xu, M.-X. Wang, Z. L. Liu, J.-F. Ge, X. Yang, C. Liu, Z. A. Xu, D. Guan, C. L. Gao, D. Qian, Y. Liu, Q.-H. Wang, F.-C. Zhang, Q.-K. Xue, and J.-F. Jia, Phys. Rev. Lett. 114, 017001 (2015). 15 R. Pawlak, M. Kisiel, J. Klinovaja, T. Meier, S. Kawai, T. Glatzel, D. Loss, and E. Meyer, npj Quantum Inf. 2, 16035 (2016). 16 S. Ryu and Y. Hatsugai, Phys. Rev. Lett. 89, 077002 H. Shtrikman, Nat Phys 8, 887 (2012). (2002). (a)-1.0-0.50.00.51.0-0.3-0.2-0.10.00.10.20.3k/πϵ[t](b)-1.0-0.50.00.51.0-0.3-0.2-0.10.00.10.20.3k/πϵ[t]8 17 A. C. Potter and P. A. Lee, Phys. Rev. Lett. 105, 227003 http://dx.doi.org/10.1063/1.4820463. (2010). 36 C. Dutreix, M. Guigou, D. Chevallier, and C. Bena, Eur. 18 T. Mizushima and M. Sato, New Journal of Physics 15, Phys. J. B 87, 296 (2014). 075010 (2013). 19 D. Wang, Z. Huang, and C. Wu, Phys. Rev. B 89, 174510 (2014). 20 K. Poyhonen, A. Weststrom, J. Rontynen, and T. Ojanen, Phys. Rev. B 89, 115109 (2014). 21 I. Seroussi, E. Berg, and Y. Oreg, Phys. Rev. B 89, 104523 (2014). 37 C. Dutreix, Impurity and boundary modes in the honey- comb lattice, Ph.D. thesis, Universit´e Paris Sud-Paris XI (2014). 38 L. Wang and M. W. Wu, Phys. Rev. B 93, 054502 (2016). 39 V. Kaladzhyan and C. Bena, SciPost Phys. 3, 002 (2017). 40 C. Dutreix, Phys. Rev. B 96, 045416 (2017). 41 J. C. Y. Teo and C. L. Kane, Phys. Rev. B 82, 115120 22 R. Wakatsuki, M. Ezawa, and N. Nagaosa, Phys. Rev. B (2010). 89, 174514 (2014). 42 M. Z. Hasan and C. L. Kane, Rev. Mod. Phys. 82, 3045 23 S. Deng, G. Ortiz, A. Poudel, and L. Viola, Phys. Rev. B (2010). 89, 140507 (2014). 43 C.-K. Chiu, J. C. Y. Teo, A. P. Schnyder, and S. Ryu, 24 P. San-Jose, E. Prada, and R. Aguado, Phys. Rev. Lett. Rev. Mod. Phys. 88, 035005 (2016). 112, 137001 (2014). 44 C. Chan and X.-J. Liu, Phys. Rev. Lett. 118, 207002 25 M. Thakurathi, K. Sengupta, and D. Sen, Phys. Rev. B (2017). 89, 235434 (2014). 26 N. Mohanta and A. Taraphder, EPL (Europhysics Letters) 108, 60001 (2014). 27 N. Sedlmayr, J. M. Aguiar-Hualde, and C. Bena, Phys. 45 C. Chan, L. Zhang, T. F. J. Poon, Y.-P. He, Y.-Q. Wang, and X.-J. Liu, Phys. Rev. Lett. 119, 047001 (2017). 46 N. Sedlmayr and C. Bena, Phys. Rev. B 92, 115115 (2015). 47 D. Sticlet, C. Bena, and P. Simon, Phys. Rev. Lett. 108, Rev. B 91, 115415 (2015). 096802 (2012). 28 J. Rontynen and T. Ojanen, Phys. Rev. B 93, 094521 48 N. Sedlmayr, J. M. Aguiar-Hualde, and C. Bena, Phys. (2016). Rev. B 93, 155425 (2016). 29 V. Kaladzhyan, J. Rontynen, P. Simon, and T. Ojanen, 49 P. Ghosh, J. D. Sau, S. Tewari, and S. Das Sarma, Phys. Phys. Rev. B 94, 060505 (2016). Rev. B 82, 184525 (2010). 30 G. Yang, P. Stano, J. Klinovaja, and D. Loss, Phys. Rev. 50 D. J. Thouless, M. Kohmoto, M. P. Nightingale, and B 93, 224505 (2016). 31 C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 146802 (2005). 32 Z. Qiao, S. A. Yang, W. Feng, W.-K. Tse, J. Ding, Y. Yao, J. Wang, and Q. Niu, Phys. Rev. B 82, 161414 (2010). M. den Nijs, Phys. Rev. Lett. 49, 405 (1982). 51 M. Sato, Phys. Rev. B 79, 214526 (2009). 52 K. Wakabayashi, K. and T. Enoki, Science and Technology of Advanced Materials 11, 054504 (2010). ichi Sasaki, T. Nakanishi, 33 C. Chamon, C.-Y. Hou, C. Mudry, S. Ryu, and L. Santos, 53 V. Kaladzhyan, J. Despres, I. Mandal, and C. Bena, Physica Scripta 2012, 014013 (2012). arxiv:1611.09367 (2016). 34 A. M. Black-Schaffer, Phys. Rev. Lett. 109, 197001 (2012). 35 F. Liu, Y. Liu, J. Hu, D. L. Smith, and P. P. Ruden, Journal of Applied Physics 114, 093708 (2013), 54 M. Sato, Phys. Rev. B 73, 214502 (2006). 55 0. Dmytruk and J. Klinovaja, arxiv:1710.01671v1 (2017). 56 X.-P. Liu, Y. Zhou, Y.-F. Wang, and C.-D. Gong, New Journal of Physics 19, 093018 (2017).
1503.03756	1	1503	2015-03-12T14:57:14	Thermal radiation from optically driven Kerr ($\chi^{(3)}$) photonic cavities	[ "cond-mat.mes-hall", "physics.optics" ]	We study thermal radiation from nonlinear ($\chi^{(3)}$) photonic cavities coupled to external channels and subject to incident monochromatic light. Our work extends related work on nonlinear mechanical oscillators [Phys. Rev. Lett. 97, 110602 (2006)] to the problem of thermal radiation, demonstrating that bistability can enhance thermal radiation by orders of magnitude and result in strong lineshape alternations, including "super-narrow spectral peaks" occurring at the onset of kinetic phase transitions. We show that when the cavities are designed so as to have perfect linear absorptivity (rate matching), such thermally activated transitions can be exploited to dramatically tune the output power and radiative properties of the cavity, leading to a kind of Kerr-mediated thermo-optic effect. Finally, we demonstrate that in certain parameter regimes, the output radiation exhibits Stokes and anti-Stokes side peaks whose relative magnitudes can be altered by tuning the internal temperature of the cavity relative to its surroundings, a consequence of strong correlations and interference between the emitted and reflected radiation.	cond-mat.mes-hall	cond-mat	AIP/123-QED Thermal radiation from optically driven Kerr (χ(3)) photonic cavities Chinmay Khandekar,1 Zin Lin,2 and Alejandro W. Rodriguez1 1)Department of Electrical Engineering, Princeton University, Princeton, NJ 08540 2)School of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02139 We study thermal radiation from nonlinear (χ(3)) photonic cavities coupled to external channels and subject to incident monochromatic light. Our work extends related work on nonlinear mechanical oscillators [Phys. Rev. Lett. 97, 110602 (2006)] to the problem of thermal radiation, demonstrating that bistability can enhance thermal radiation by orders of magnitude and result in strong lineshape alternations, including "super-narrow spectral peaks" ocurring at the onset of kinetic phase transitions. We show that when the cavi- ties are designed so as to have perfect linear absorptivity (rate matching), such thermally activated transitions can be exploited to dramatically tune the output power and radiative properties of the cavity, leading to a kind of Kerr-mediated thermo-optic effect. Finally, we demonstrate that in certain parameter regimes, the output radiation exhibits Stokes and anti-Stokes side peaks whose relative magnitudes can be altered by tuning the internal tem- perature of the cavity relative to its surroundings, a consequence of strong correlations and interference between the emitted and reflected radiation. 5 1 0 2 r a M 2 1 ] l l a h - s e m . t a m - d n o c [ 1 v 6 5 7 3 0 . 3 0 5 1 : v i X r a 1 Driven nonlinear oscillators, including optical,1 optomechanical,2 and MEMS3,4 resonators, have been studied for decades and exploited for many applications, from mass detection5 to sens- ing6 and tunable filtering.7 When driven to a non-equilibrium state, these systems can exhibit a wide range of unusual thermal phenomena,8 leading for instance to cooling and amplification of thermal fluctuations in optomechanical systems,2 generation of squeezed states of light in Kerr media,9 and stochastic resonances.10 Previous studies of Duffing oscillators have also identified novel effects arising from the nonlinear interaction of coherent pumps with thermal noise,11 -- 13 leading to phase transitions and lineshape alterations that were recently observed in a handful of systems, e.g. mechanical oscillators14,15 and Josephson junctions.16 In this letter, we study thermal radiation effects in optically driven χ(3) photonic cavities cou- pled to external channels. We demonstrate that in certain parameter regimes, bistability17,18 in photonic resonators leads to thermally activated transitions that amplify thermal fluctuations by or- ders of magnitude and cause dramatic changes in the cavity spectrum, analogous to noise-induced switching in mechanical oscillators.15 We find that when the photonic cavity is critically coupled to the radiation channel (enforced by designing the cavity to have equal dissipation and radiation rates),19 the coherent part of the output power varies dramatically with temperature, leading to a kind of Kerr-mediated thermo-optic effect. A simple perturbative analysis also shows that outside of the bistability region, the interaction of the coherent drive with thermal noise leads to amplified, Raman-type Stokes and anti-Stokes side peaks in the radiation spectrum, the relative amplitudes of which depend on a sensitive interference between the externally incident and reflected thermal radiation. Related phenomena have been long studied13 and more recently observed14,15 in the context of driven nonlinear mechanical oscillators as well as resonators based on rf-driven Joseph- son junctions16, and microscopic theories have also been used to describe related optical effects in the quantum regime.20 Our work is an extension of these studies to the particular problem of thermal radiation from photonic resonators. As we show below, additional considerations arising in the case of radiation from cavities but absent in mechanical oscillators or bulk media, such as the strong coupling of the cavity to an external channel, dramatically impact the outgoing radia- tion. The ability to tune the radiation properties of resonators via temperature and optical signals offers potentially new avenues for applications in nano-scale heat regulation,21 detection,22 recti- fication,23,24 photovoltaics,25 or incoherent sources.26 We propose a practical, photonic structure where these effects can arise near room temperature and at mW powers. The system under consideration belongs to the class of nonlinear photonic resonators depicted 2 FIG. 1. Schematic of a wavelength-scale silicon ring resonator of radius R = 4.4µm, height h = 220nm, and width w = 350nm, coupled to a silicon waveguide (channel), both on a silica substrate. Also shown is the Ey mode profile of a resonance designed to have azithmutal number m = 25, wavelength λ = 1.5µm, radiative lifetimes & 106, and relatively large nonlinear coupling coefficient α = 0.032χ(3)ω0/(8ǫ0λ3). The loss γd and waveguide -- coupling γe rates are much larger than the corresponding radiation rate. in Fig. 1, involving a cavity coupled to an external channel (e.g. a waveguide). The description of thermal radiation in this system can be carried out via the coupled-mode theory framework27, which we recently employed to study thermal radiation in a related system28 but now extend to consider the addition of a coherent pump. The equations describing the cavity mode a are given by:28 da dt = [i(ω0 − αa2) − γ]a +p2γdξd +p2γes+, s− = −s+ +p2γea, (1) (2) where a2 is the energy of the cavity mode and s±2 are the incident (+) and outgoing (−) power from and into the external channel, respectively. The latter arises due to dissipative noise inside the cavity ξd as well as externally incident light s+ consisting of both thermal radiation ξe and a monochromatic pump sp exp(iωpt). The dynamics of the cavity field are described by its res- onance frequency ω0 and decay rate γ = γe + γd, which includes linear absorption γd as well as decay into the external channel γe. The real and imaginary parts of the nonlinear coefficient α = 3 fields,29 and lead to self-phase modulation (SPM) and two-photon absorption (TPA), respectively. 4ω0R ε0χ(3) ~E4/(R ε ~E2)2 depend on a complicated overlap integral of the linear cavity We mainly focus on the effects of SPM (real α > 0) since we find that TPA leads to thermal broad- ening of the kind explored in Ref. 28. Both internal and external thermal sources are represented by stochastic, delta-correlated white-noise sources ξe and ξd satisfying (assuming γ ≪ ω0), hξ∗(t)ξ(t′)i = Θ(ω0, T )δ(t − t′), (3) 3 where h. . .i denotes a thermodynamic or ensemble average, and Θ(ω, T ) = ω/(eω/kBT − 1) is the mean energy of a Planck oscillator23 at local temperature T ; the temperatures of the internal and external baths are denoted as Td and Te, respectively. Above, we assumed γ/kB ≪ Te, Td al- lowing us to ignore the frequency-dispersion and temporal correlations (colored noise) associated with Θ. (Note that in the limit ω0/kBT → 0 one obtains the classical result Θ → kBT .) Thermal amplification and power tunability. -- We show that bistability can amplify thermal fluctuations and lead to enhanced, temperature -- tunable emission from the cavity. We begin by reviewing a number of key features of the system in the absence of thermal noise, whose contri- butions are considered perturbatively due to the generally weak nature of thermal noise, sp2 ≫ γkBT . The steady-state cavity field a0 due to the pump is given by the well-known cubic equa- tion:13,30 "(cid:18)∆ + γ (cid:19)2 αa02 + 1# αa02 γ = 2ζ, (4) ωp−ω0 γ where ζ ≡ αsp2γe/γ3 is the effective nonlinear coupling associated with the pump and ∆ ≡ is the dimensionless detuning. Equation (4) describes a number of extensively studied non- linear effects,30,31 including bistability arising in the regime ∆ < −√3 and ζ (1) < ζ < ζ (2), as illustrated by the hysterisis plot on the inset of Fig. 2(a) which shows the dimensionless cavity energy x = αa2/γ as a function of ζ. An effect that seems little explored but that plays an important role on the thermal properties of this system is perfect absorption, which occurs when a photonic cavity is driven on resonance and its dissipation and radiation rates are equal, also known as rate matching.19 In the presence of nonlinearities, the cavity frequency and hence the absorbed power depend on ζ. For instance, in the non-bistable regime, the output power varies slowly with ζ, as illustrated by the green curve in Fig. 2(a) for ∆ = −1, increasing and then decreasing as ζ → ∆/2, at which point the cavity and pump frequencies are in resonance, i.e. αa02/γ = −∆. Bistability can lead to a more pronounced dependence on ζ: the two stable steady states experience different frequency shifts and hence loss rates, and ultimately which state is excited in the steady state depends on the specific initial (or excitation) conditions.32 Figure 2(a) shows the steady-state output power s−2 as ζ is adiabatically increased (solid lines) from zero and above the critical point ζ (2), for multiple ∆. The dashed blue line shows the power as ζ is adiabatically decreased below ζ (2) for the particular case ∆ = −2.5, demonstrating that only the upper branch experiences perfect absorption, occuring at ζ = ∆/2 4 3 2 1 γ / 2 a α ζ=1.25 (1) ζc (B) (A) ζc (2) 1.2 1.4 ζ 1.6 (a) 1 0.8 2 + s / 2 -- s 0.6 0.4 0.2 Δ=-2.5 γe >> γd (b) 1 Δ=-4 γe=γd (A) 0.8 (A) Δ=-2.5 γe=γd Δ=-2.5, γe >> γd Coherent ouput Thermal output 1 2 + s / 2 -- s 0.6 0.2 (A) ζT=0 (B) ζTQ2=20 (C) ζTQ2=50 (B) -2.5 Frequency,δω/γ -2.48 T B k / 2 -- s δ 8 4 0 (C) (B) -4 Frequency,δω/γ -2 0 (C) Δ=-2.5, ζ=1.3 Δ=-2.5, ζ=1.4 Δ=-4 Δ=-1.8 Δ=-1 0.6 0.4 0.2 Δ=-1.8 γe=γd U l ζ=1.25 ζ=1.6 1 0 a i t n e t o P -1 (A) (B) Δ=-1 γe=γd (B) ζ=1.25 100 Nonlinearity, ζ 0 10-1 -1 1 αa2/γ 3 101 0 0 0.5 1 1.5 2 Temperature, ζTQ2 (x 10-2) 2.5 3 FIG. 2. (a) Output power s−2 normalized by the input power s+2 of the pumped system described in Fig. 1, in the absence of thermal noise and as a function of ζ, for different values of detuning ∆ = ωp−ω0 . The top left inset shows a hysterisis plot of the energy αa02/γ as a function of ζ, the solutions of (4), for the particular choice of ∆ = −2.5 while the bottom inset shows the corresponding potential energy U as a function of the cavity energy for two different ζ = ∆/2 and ζ (2). (b) The same normalized output power s−2/s+2 as a function of temperature ζT Q2, where ζT = αΘ(ω0, T )γe/γ2 and Q = ω0/γ, for different values of ∆ and ζ . ζ (2). The insets illustrate the change in the coherent (left) and thermal radiation (right) spectra. Both internal and external baths have equal temperatures Td = Te = T . γ and marked by the white circle. The corresponding change in the output power as the system transitions from the lower (A) to the higher (B) energy state at ζ (2) is given approximately by: s+2(cid:18)1 − γe + γd(cid:19) γe − γd (∆ + x1)2 − (∆ + x2)2 [1 + (∆ + x1)2][1 + (∆ + x2)2] (5) 3 (2∆ +√∆2 − 3) and x2 = −2(∆ + x1) are the cavity energies associated with the where x1 = − 1 lower and higher energy state, respectively. Given (5), one can show that the difference in output power is largest under the rate matching condition γe = γd and at ∆ ≈ −7/3, decreasing with smaller or larger detuning. The presence of noise complicates this picture due to finite -- temperature fluctuations which cause the system to undergo transitions between the two states, where the rates of forward/backward transitions are a complicated function of the potential energy U and temperature of the system.13,32 (For convenience and without loss of generality, we take both thermal baths to have the same tem- the cavity energy dx perature T .) In particular, U is obtained by integrating the steady-state equation associated with dt = [(∆ + x)2 + 1]x − 2ζ = 0 with respect to x = αa2/γ. Examples of U are shown on the lower inset of Fig. 2(a) for two values of ζ. Thermally activated hopping 5 leads to significant enhancement of amplitude fluctuations, which manifest as large changes in the radiation spectrum of the cavity. This is illustrated by the top inset of Fig. 2(b), which shows the thermal spectrum of the power δs−(ω)2 for the particular choice of ∆ = −2.5 and ζ = 1.3 and for multiple values of ζT Q2, where for convenience (below) we have introduced the dimensionless effective thermal coupling ζT ≡ αΘ(ω0, T )γe/γ2 and cavity-lifetime Q = ω0/γ.19 Such enhance- ments were predicted to occur and recently observed in nonlinear mechanical oscillators,13,14 where the authors showed that at special ζ, the system undergoes a so-called kinetic phase tran- sition associated with equal rates of forward/backward hopping and exhibits a "supernarrow" and highly amplified spectral peak. Interestingly, we find that in the case of optical resonators, thermal amplification can be accompanied by a significant decrease in the coherent output power despite the fact that sp2 ≫ γkBT , a consequence of perfect absorption. In particular, operating under rate matching and near ζ (2) allows for temperature to initiate transitions (A) ⇋ (B), leading to significant changes in s−2 with respect to T . Essentially, as ζ → ζ (2), the potential barrier separating the lower x1 from the higher x2 energy states begins to dissapear, resulting in increased rate of forward transitions and hence larger absorption. These features are illustrated in Fig. 2(b) which shows the total output power as a function of ζT Q2 for different combinations of ζ and ∆. (We found numerically that for a given ζ and ∆, changing either ζT or Q2 while leaving ζT Q2 unchanged leaves s−2/s+2 unaltered.) Noticeably, while the change in the output power is gradual in the non-bistable regime (∆ > −√3), there is a significantly stronger dependence in the bistable regime -- the slope becomes increasingly sharper as ζ → ζ (2) and ζT → 0 since it becomes increasingly easier for lower T fluctuations to induce hopping unto the higher-energy state. At sufficiently large ζT , s−2 is found to increase with increasing ζT as the cavity field no longer probes the hysterisis regime. While the maximum change in s−2 can be estimated from the steady-state analysis in the absence of noise (with the largest change occuring for ∆ ≈ −7/3), its dependence on ζT is a complicated function of ζ and ∆. For instance, for ∆ = −2.5 and ζ = 1.4 [blue line in Fig. 2(b)], the sharp decrease in output power occurs at ζT Q2 ≈ 10 and yields a slope δ(ζT Q2) ≈ 0.05. δ(s−2) 1 s+2 Side peaks. -- We now show that the radiation spectrum also exhibits other interesting features, including the emergence of Raman-type Stokes and anti-Stokes side peaks previously observed in driven mechanical oscillators.13,14 Interestingly, we find that in our photonic resonator, the pres- ence of the external channel dramatically alters the relative amplitudes of the side peaks, e.g. leading to a symmetric spectrum when the two baths have equal temperatures. We begin by ex- 6 (a) 3 T B k / 2 -- s δ , n o i t a d a r l i a m r e h T 2.5 2 1.5 1 0.5 0 Eq. (7) Te << Td ζ = 2 ζ = 4 ζ = 0.2 Eq. (7) (b) 4 3 2 1 Te >> Td γe = γd -8 -6 -4 -2 Frequency, δω / γ 0 2 0 −10 −7.5 −5 −2.5 0 Frequency, δω / γ Te >> Td γe >> γd Te = Td γe = γd Te << Td γe = γd 2.5 5 FIG. 3. Thermal radiation δs−2 for the system described in Fig. 1, normalized by the maximum of the internal and external bath temperatures T = max{Td, Te}, as a function of the dimensionless frequency δω/γ where δω = (ω − ω0), with fixed ∆ = −2.5 and under various operating conditions. The radiation spectrum is shown (a) in the limit Td ≫ Te of negligible externally incident radiation, for different values of ζ and (b) for fixed ζ = 4 but different Td, Te and linear decay rates γd, γe. ploiting a simple perturbation theory in which the thermal fluctuations of the cavity-field δa and radiation δs− = −ξe + √2γeδa are treated perturbatively, leading to analytical expressions for the corresponding thermal energy and radiation spectra. Assuming sp2 ≫ γkB{Td, Te}, we obtain: kBf+(2γdTd + 2γeTe) hδa2i = hδs−2i = kBTe + D 4γekB D [γd(f+Td − f−Te) + γe(f+ − f−)Te] (6) (7) where f+, f− and D are given by f± = (ω + ω0 − 2ωp − 2αa02)2 + γ2 ± α2a04 D = [γ2 + (ω0 − ωp − 2αa02)2 − (ω − ωp)2 − α2a04]2 + 4γ2(ω − ωp)2, and where a02 denotes the steady-state cavity energy in the absence of fluctuations, the solution of (4). Here, for simplicity we have assumed the classical limit Θ(ω0, T ) → kBT . In the absence of the external bath Te = 0, the above equations are similar to those obtained in the case of me- chanical oscillators.13 The situation changes with the channel due to the unavoidable intereference and induced correlations of the emitted and reflected radiation, described in (7) by the f− terms. 7 Figure 3 illustrates the radiation spectrum of the cavity under different operating conditions, showing excellent agreement between the numerically (noisy) and analytically (dashed lines) com- puted spectra. We note that all of the results shown in Fig. 3 correspond to cavities operating out- side of the bistable regime: although it is possible to obtain a complete and analytical description of the spectrum based on (6) and (7), such an analysis is difficult in the bistable regime due to hopping between states, requiring a complicated description of the transition rates and stationary distributions of the system.13 For instance, in the bistable regime, one observes the appearance of a temperature-dependent supernarrow spectral peak whose amplitude decreases with increas- ing ζ. In the absence of bistability, a similar but weaker amplification occurs as ζ → ∆/2 or equivalently, as the cavity frequency becomes resonant with the pump. Regardless of regime, at suffiiently large ζ (once the cavity resonance has crossed ωp), modulation of the thermal noise by the pump causes the spectrum to transition from being singly to doubly resonant due to the emer- gence of an additional anti-Stokes peak.13 In the limit as ζ → ∞, both peaks move farther apart and their amplitudes assymptote to a system-dependent constant. Figure 3(b) explores the dependence of the peak amplitudes on various cavity parameters, in- cluding Td = Te, Td ≫ Te, and Td ≪ Te, corresponding to a resonator that is either at thermal equilibrium, heated, or cooled with respect to its surroundings, respectively. When noise entering the system through the external bath is negligible Td ≫ Te, similar to the previously explored situation involving mechanical oscillators,13 one finds that the Stokes peak is always much larger than the anti-Stokes peak (green line). Essentially, for α > 0 the cavity nonlinearity favors down- conversion, as captured by the asymmetric f+ terms above. The peak radiation associated with the Stokes peak can be readily obtained from (6) in the non-bistable regime ∆ ≤ −√3, and is given by maxδs−2 = 4γeγd γ2 (1 + 2∆2)kBTd, reaching 7kBTd precisely at the onset of bistability and when γe = γd. At larger ζ > ∆/2, the amplitude of both peaks decreases with increasing ζ where, as ζ → ∞ (not shown), the amplitude of the Stokes peak → kBTd while the anti-Stokes peak dissapears. The situation changes dramatically when the noise entering the system through the external bath cannot be ignored, i.e. Te & Td. In particular, as observed from (6), although the cavity spectrum favors Stokes to anti-Stokes conversion regardless of the relative tempera- tures or decay rates, we find that the spectrum of outgoing radiation can be dramatically different depending on the regime of operation. When Te ≫ Td where noise is dominated by external ra- diation, we find that the anti-Stokes peak dominates (red line) except when γe ≫ γd, at which point the spectrum exhibits a symmetric lineshape (orange line). Such a reversal in relative am- 8 plitudes is captured by the f− terms above, which include correlations and interference between the emitted and reflected radiation. The maximum radiation in the non-bistable regime in this case maxδs−2 = kBTe(1 + 8γ2 γ2 ), reaching 25kBTe at the onset of bistability and when γe ≫ γd (rather than under rate matching). Interestingly, we find that when the two baths lie at the same temperature Te = Td, both peaks have equal amplitudes regardless of γe/γd, though the maximum amplitude in this regime also occurs in the limit γe ≫ γd. This unexpected symmetriza- tion of the spectrum arising due to interference effects seems to be a unique property of thermal γ2 ∆2 − 4γeγd e radiation in this system. Although previous work on nonlinear fluctuations in the quantum regime observed similar peaks in the spectrum, a symmetric spectrum was found to arise only at zero temperature (a singular point of the theory33) due to quantum tunneling.20 Although a number of the abovementioned effects have been observed in mechanical oscil- lators, they remain unobserved in the context of thermal radiation where they could potentially be exploited in numerous applications.21,25,26 As demonstrated above, the interplay between the internal and externally incident radiation and the coherent pump leads to new effects in thermal radiators, including dramatic changes in both the coherent and thermal output spectrum with tem- perature, along with temperature -- tunable Stokes and anti-Stokes side peaks. Finally, we conclude δ(s−2) by proposing a realistic, silicon ring resonator design, depicted schematically in Fig. 1, where one could potentially observe these effects near room temperature and with operating Q ∼ 105 and input power sp2 ∼ 1mW, leading to αsp2Q2 ∼ ∆. For these parameters we find that δ(T ) ∼ 0.04K−1 at T ≈ 300K. Although this is almost two orders of magnitude smaller in comparison with thermo-optic effects in silicon, which lead to tunable powers ∼ K−1 for the same structure, at lower temperatures T . 100K where the thermo-optic coefficient is much smaller34, our fluctuation-induced effects offer significantly better temperature tunability. Other cavity de- 1 s+2 signs such as the nanobeam cavity described in Ref. 35 yield much larger α and allow smaller Q to be employed, leading to even larger tunability compared to that obtained via thermo-optic effects. We are grateful to Mark Dykman for very helpful comments and suggestions. This work was supported in part by the National Science Foundation under Grant No. DMR-145483. REFERENCES 1Kerry J. Vahala. Optical microcavities. Nature, 424:839 -- 846, 2003. 2Tobias J. Kippenberg and Kerry J. Vahala. Cavity opto-mechanics. Opt. Express, 15(25):17172 -- 9 17205, 2007. 3R. Lifshitz and M.C. Cross. Nonlinear dynamics of nanomechanical and micromechanical res- onators. Reviews of nonlinear dynamics and complexity, 1:1 -- 50, 2008. 4Romain Quidant, Jan Gieseler, and Lukas. Novotny. Thermal nonlinearities in a nanomechanical oscillator. Nature Physics, 9:806 -- 810, 2013. 5J. Chaste, A. Eichler, J. Moser, G. Cellabos, R. Rurali, and A. Bachtold. A nanomechanical mass sensor with yoctogram resolution. Nature Nanotechnology, 7:301 -- 304, 2012. 6A.N. Clelan and M.L. Roukes. Noise processes in nanomechanical resonators. J. Appl. Phys., 92(5):2758 -- 2769, 2002. 7R. Almog, S. Zaitsev, O. Shtempluck, and E. Buks. High intermodulation gain in a microme- chanical duffing resonator. Appl. Phys. Lett., 88:213509, 2006. 8Mark Dykman, editor. Fluctuating Nonlinear Oscillators: From Nanomechanics to Quantum Superconducting Circuits, chapter 13. Oxford Univrsity Press, 2012. 9Ling-An Wu, Min Xiao, and H. J. Kimble. Squeezed states of light from an optical parametric oscillator. JOSA-B, 4:1465 -- 1476, 1987. 10Luca Gammaltoni, Peter Hangi, Peter Jung, and Fabio Marchesonl. Stochastic resonance. Rev.Mod.Phys., 70:223, 1998. 11M.I. Dykman. Theory of nonlinear nonequilibrium oscillators interacting with a medium. Zh.Eksp.Theor.Fiz, 68:2082 -- 2094, 1975. 12M.I. Dykman and P.V.E. McClintok. Power spectra of noise-driven nonlinear systems and stochastic resonance. Physica D, 58:10 -- 30, 1992. 13M.I. Dykman, D.G. Luchinsky, and R. Manella. Supernarrow spectral peaks and high- frequency stochastic resonane in systems with coexisting periodic attractors. Physical Review E, 49(2):1198 -- 1215, 1994. 14C. Stambaugh and H.B. Chan. Supernarrow spectral peaks near a kinetic phase transition in a driven nonlinear micromechanical oscillator. Phys. Rev. Lett., 97:110602, 2006. 15C. Stambaugh and H.B Chan. Noise-activated switching in a driven nonlinear micromechanical oscillator. Phys. Rev. B, 73:172302, 2006. 16Stephan Andre, Lingzhen Guo, Vittorio Peano, Michael Mathaler, and Gerd Schon. Emission spectrum of the driven nonlinear oscillator. Phys. Rev. A, 85:053825, 2012. 17Masaya Notomi, Akihiko Shinya, Satoshi Mitsugi, Goh Kira, Eiichi Kuramochi, and Takasumi Tanabe. Optical bistable switching action of Si high-q photonic-crystal nanocavities. Opt. Ex- 10 press, 13(7):2678 -- 2687, 2005. 18A. R. Cowan and J. F. Young. Optical bistability involving photonic crystal microcavities and fano line shapes. Phys. Rev. E, 68:046606, 2003. 19John D. Joannopoulos, Steven G. Johnson, Joshua N. Winn, and Robert D. Meade. Photonic Crystals: Molding the Flow of Light. Princeton University Press, second edition, February 2008. 20P.D. Drummond and D.F. Walls. Quantum theory of optical bistability I. nonlinear polarisability model. Journal of Physics A: Math. and Gen., 13:725, 1980. 21Lei Wang and Baowen Li. Thermal memory: A stroage of phononic information. Phys. Rev. Lett., 101:267203, 2008. 22Kengo Nozaki, Shinji Matsuo, Koji Takeda, Tomonari Sato, Eiichi Kuramochi, and Masaya No- tomi. Ingaas nano-photodetectors based on photonic crystal waveguide including ultracompact burried heterostructure. Optics Express, 21:19022, 2013. 23Clayton R. Otey, Wah Tung Lau, and Shanhui Fan. Thermal rectification through vacuum. Phys. Rev. Lett., 104(15):154301, 2010. 24Nick A. Roberts and D.G. Walker. A review of thermal rectification observations and models in solid materials. Journal of Thermal Sciences, 50:648 -- 662, 2011. 25Andrej Lenert, David M. Bierman, Youngsuk Nam, Walker R. Chan, Ivan Celanovic, Marin Soljacic, and Evelyn N. Wang. A nanophotonic solar thermophotovoltaic device. Nature Nan- otechnology, 9:126 -- 130, 2014. 26S. Noda, M. Fujita, and T. Asano. Spontaneous-emission control by photonic crystals and nanocavities. Nature Photonics, 1:449 -- 458, 2007. 27H. A. Haus. Waves and Fields in Optoelectronics. Prentice-Hall, Englewood Cliffs, NJ, 1984. Ch. 7. 28Chinmay Khandekar, Adi Pick, Steven G. Johnson, and Alejandro W. Rodriguez. Radiative heat transfer in nonlinear kerr media. Phys. Rev. B, 91:115406, 2015. 29Alejandro Rodriguez, Marin Soljaci´c, J. D. Joannopulos, and Steven G. Johnson. χ(2) and χ(3) harmonic generation at a critical power in inhomogeneous doubly resonant cavities. Opt. Ex- press, 15(12):7303 -- 7318, 2007. 30Marin Soljacic, Mihai Ibanescu, Steven J. Johnson, Yoel Fink, and J.D. Joannopoulos. Optimal bistable switching in nonlinear photonic crystals. Phys.Rev.E., 66:055601, 2002. 31Steven G. Johnson, Attila Mekis, Shanhui Fan, and J. D. Joannopoulos. Molding the flow of 11 light. Computing Sci. Eng., 3(6):38 -- 47, 2001. 32Steven H. Strogatz. Nonlinear Dynamics and Chaos. Westview Press, Boulder, CO, 1994. 33M.I. Dykman and M.A. Krivoglaz. Theory of nonlinear oscillators interacting with a medium. Soviet Physics Reviews, 5, 1984. 34J. Komma, C. Schwarz, G. Hoffman, D. Heinert, and R. Nawrodt. Thermo-optic coefficient of silicon at 1550nm and cryogenic temperatures. Appl. Phys. Lett., 101:041905, 2012. 35Z. Lin, T. Alcorn, M. Loncar, S.G. Johnson, and A.W. Rodriguez. High-efficiency degenerate four wave-mixing in triply resonant nanobeam cavities. Phys. Rev. A, 89:053839, 2014. 12
1004.3396	4	1004	2011-11-22T22:16:03	Electronic Properties of Graphene in a Strong Magnetic Field	[ "cond-mat.mes-hall", "cond-mat.str-el" ]	We review the basic aspects of electrons in graphene (two-dimensional graphite) exposed to a strong perpendicular magnetic field. One of its most salient features is the relativistic quantum Hall effect the observation of which has been the experimental breakthrough in identifying pseudo-relativistic massless charge carriers as the low-energy excitations in graphene. The effect may be understood in terms of Landau quantization for massless Dirac fermions, which is also the theoretical basis for the understanding of more involved phenomena due to electronic interactions. We present the role of electron-electron interactions both in the weak-coupling limit, where the electron-hole excitations are determined by collective modes, and in the strong-coupling regime of partially filled relativistic Landau levels. In the latter limit, exotic ferromagnetic phases and incompressible quantum liquids are expected to be at the origin of recently observed (fractional) quantum Hall states. Furthermore, we discuss briefly the electron-phonon coupling in a strong magnetic field. Although the present review has a dominating theoretical character, a close connection with available experimental observation is intended.	cond-mat.mes-hall	cond-mat	Electronic Properties of Graphene in a Strong Magnetic Field M. O. Goerbig 1Laboratoire de Physique des Solides, Univ. Paris-Sud, CNRS UMR 8502, F-91405 Orsay, France (Dated: May 16, 2014) We review the basic aspects of electrons in graphene (two-dimensional graphite) exposed to a strong perpendicular magnetic field. One of its most salient features is the relativistic quan- tum Hall effect the observation of which has been the experimental breakthrough in identifying pseudo-relativistic massless charge carriers as the low-energy excitations in graphene. The effect may be understood in terms of Landau quantisation for massless Dirac fermions, which is also the theoretical basis for the understanding of more involved phenomena due to electronic interactions. We present the role of electron-electron interactions both in the weak-coupling limit, where the electron-hole excitations are determined by collective modes, and in the strong-coupling regime of partially filled relativistic Landau levels. In the latter limit, exotic ferromagnetic phases and incompressible quantum liquids are expected to be at the origin of recently observed (fractional) quantum Hall states. Furthermore, we discuss briefly the electron-phonon coupling in a strong magnetic field. Although the present review has a dominating theoretical character, a close con- nection with available experimental observation is intended. PACS numbers: 81.05.ue, 73.43.Lp, 73.22.Pr Contents I. Introduction to Graphene A. The Carbon Atom and its Hybridizations B. Crystal Structure of Graphene C. Electronic Band Structure of Graphene 1. Tight-binding model for electrons on the honeycomb lattice 2. Continuum limit D. Deformed Graphene 1. Dirac point motion 2. Tilted Dirac cones II. Dirac Equation in a Magnetic Field and the Relativistic Quantum Hall Effect A. Massless 2D Fermions in a Strong Magnetic Field 1. Quantum-mechanical treatment 2. Relativistic Landau levels B. Limits of the Dirac Equation in the Description of Graphene Landau Levels C. Landau Level Spectrum in the Presence of an Inplane Electric Field D. Landau Levels in Deformed Graphene 1. The generalized Weyl Hamiltonian in a magnetic field 2. Tilted Dirac cones in a crossed magnetic and electric field III. Electronic Interactions in Graphene – Integer Quantum Hall Regime A. Decomposition of the Coulomb interaction in the Two-Spinor Basis 1. SU(2) valley symmetry 2. SU(4) spin-valley symmetric Hamiltonian B. Particle-Hole Excitation Spectrum 1. Graphene particle-hole excitation spectrum at B = 0 2. Polarizability for B 6= 0 3. Electron-electron interactions in the random-phase approximation: upper-hybrid mode and linear magnetoplasmons 4. Dielectric function and static screening 1 2 3 4 4 7 10 11 12 13 13 14 14 17 19 19 19 20 20 22 23 24 24 25 26 28 29 IV. Magneto-Phonon Resonance in Graphene A. Electron-Phonon Coupling 31 31 1. Coupling Hamiltonian 32 2. Hamiltonian in terms of magneto-exciton operators 32 33 33 34 1. Non-resonant coupling and Kohn anomaly 2. Resonant coupling B. Phonon Renormalization and Raman Spectroscopy V. Electronic Correlations in Partially Filled Landau Levels A. Electrons in a Single Relativistic Landau Level 1. SU(4)-symmetric model 2. Symmetry-breaking long-range terms 3. Qualitative expectations for correlated electron phases 4. External spin-valley symmetry breaking terms 5. Hierarchy of relevant energy scales B. SU(4) Quantum Hall Ferromagnetism in Graphene 1. Ferromagnetic ground state and Goldstone modes 2. Skyrmions and entanglement 3. Comparison with magnetic catalysis 4. The quantum Hall effect at ν = ±1 and ν = 0 1. Generalized Halperin wave functions 2. The use of generalized Halperin wave functions in C. Fractional Quantum Hall Effect in Graphene graphene 3. Experiments on the graphene FQHE VI. Conclusions and Outlook Acknowledgments A. Matrix Elements of the Density Operators References 35 35 36 37 37 39 40 41 41 42 45 46 48 48 49 51 51 52 52 53 I. INTRODUCTION TO GRAPHENE The experimental and theoretical study of graphene, two-dimensional (2D) graphite, has become a major is- sue of modern condensed matter research. A milestone was the experimental evidence of an unusual quantum Hall effect reported in September 2005 by two different groups, the Manchester group led by Andre Geim and a Columbia-Princeton collaboration led by Philip Kim and Horst Stormer (Novoselov et al., 2005a; Zhang et al., 2005). The reasons for this enormous scientific interest are manyfold, but one may highlight some major motiva- tions. First, one may underline its possible technological potential. One of the first publications on graphene in 2004 by the Geim group reported indeed an electric field effect in graphene, i.e. the possibility to control the car- rier density in the graphene sheet by simple application of a gate voltage (Novoselov et al., 2004). This effect is a fundamental element for the design of electronic de- vices. In a contemporary publication Berger et al. re- ported on the fabrication and the electrical contacting of monolayer graphene samples on epitaxially grown SiC crystals (Berger et al., 2004). Today’s silicon-based elec- tronics reaches its limits in miniaturization, which is on the order of 50 nm for an electric channel, whereas it has been shown that a narrow graphene strip with a width of only a few nanometers may be used as a transistor (Ponomarenko et al., 2008), i.e. as the basic electronics component. Apart from these promising technological applications, two major motivations for fundamental research may be emphasized. Graphene is the first truely 2D crystal ever observed in nature and possess remarkable mechanical properties. Furthermore, electrons in graphene show rel- ativistic behavior, and the system is therefore an ideal candidate for the test of quantum-field theoretical models that have been developed in high-energy physics. Most promenently, electrons in graphene may be viewed as massless charged fermions living in 2D space, particles one usually does not encounter in our three-dimensional world. Indeed, all massless elementary particles hap- pen to be electrically neutral, such as photons or neutri- nos.1 Graphene is therefore an exciting bridge between condensed-matter and high-energy physics, and the re- search on its electronic properties unites scientists with various thematic backgrounds. Several excellent reviews witness the enormous re- search achievements in graphene. In a first step those by Geim and Novoselov (Geim and Novoselov, 2007) and by de Heer (de Heer et al., 2007) aimed at a rather global experimental review of exfoliated and epitaxial graphene, respectively. Furthermore, the review by Castro Neto (Castro Neto et al., 2009) was concerned with general theoretical issues of electrons in graphene. Apart from the review by Abergel et al. (Abergel et al., 2010), more recent reviews concentrate on the subfields of graphene 1 The neutrino example observed vors (Fukuda, Y. et al. (Super-Kamiokande Collaboration), 1998). The fla- indeed a tiny non-zero mass is only partially correct. (νµ ↔ ντ ) oscillation between different neutrino requires 2 ground state excited state ( ~ 4 eV) 2p x 2p y 2p z ~ 4 eV 2p x 2p y 2p z 2s 1s y g r e n E 2s 1s Figure 1 Electronic configurations for carbon in the ground state (left) and in the excited state (right). research, which have themselves grown to a consider- able size and that require reviews on their own. As an example one may cite the review by Peres (Peres, 2010), which is concerned with transport properties of graphene, or that by Kotov and co-workers on interac- tion effects (Kotov et al., 2010). The present theoretical review deals with electronic properties of graphene in a strong magnetic field, and its scope is delimited to mono- layer graphene. The vast amount of knowledge on bilayer graphene certainly merits a review on its own. A. The Carbon Atom and its Hybridizations In order to understand the crystallographic structure of graphene and carbon-based materials in general, it is useful to review the basic chemical bonding properties of carbon atoms. The carbon atom possesses 6 electrons, which, in the atomic ground state, are in the configu- ration 1s22s22p2, i.e. 2 electrons fill the inner shell 1s, which is close to the nucleus and which is irrelevant for chemical reactions, whereas 4 electrons occupy the outer shell of 2s and 2p orbitals. Because the 2p orbitals (2px, 2py, and 2pz) are roughly 4 eV higher in energy than the 2s orbital, it is energetically favorable to put 2 electrons in the 2s orbital and only 2 of them in the 2p orbitals (Fig 1). It turns out, however, that in the presence of other atoms, such as e.g. H, O, or other C atoms, it is favorable to excite one electron from the 2s to the third 2p orbital, in order to form covalent bonds with the other atoms. In the excited state, we therefore have four equivalent quantum-mechanical states, 2si, 2pxi, 2pyi, and 2pzi. A quantum-mechanical superposition of the state 2si with n 2pji states is called spn hybridization. The sp1 hybridization plays, e.g., an important role in the context of organic chemistry (such as the formation of acetylene) and the sp3 hybridization gives rise to the formation of diamonds, a particular 3D form of carbon. Here, how- ever, we are interested in the planar sp2 hybridization, which is the basic ingredient for the graphitic allotropes. As shown in Fig. 2, the three sp2-hybridized orbitals are oriented in the xy-plane and have mutual 120◦ angles. (a) (b) o 120 (c) C C H H H H C C H C C H (d) Figure 2 (a) Schematic view of the sp2 hybridization. The orbitals form angles of 120o. (b) Benzene molecule (C6H6). The 6 carbon atoms are situated at the corners of a hexagon and form covalent bonds with the H atoms. (c) The quantum-mechanical ground state of the benzene ring is a superposition of the two configurations which differ by the position of the π bonds. (d) Graphene may be viewed as a tiling of benzene hexagons, where the H atoms are replaced by C atoms of neighboring hexagons and where the π electrons are delocalized over the whole structure. The remaining unhybridized 2pz orbital is perpendicular to the plane. A prominent chemical example for such hybridization is the benzene molecule the chemical structure of which has been analyzed by August Kekul´e in 1865 (Kekul´e, 1865, 1866). The molecule consists of a hexagon with carbon atoms at the corners linked by σ bonds [Fig. 2 (b)]. Each carbon atom has, furthermore, a covalent bond with one of the hydrogen atoms which stick out from the hexagon in a star-like manner. In addition to the six σ bonds, the remaining 2pz orbitals form three π bonds, and the resulting double bonds alternate with single σ bonds around the hexagon. Because a double bond is stronger than a single σ bond, one may ex- pect that the hexagon is not perfect. A double bond (C=C) yields indeed a carbon-carbon distance of 0.135 nm, whereas it is 0.147 nm for a single σ bond (C–C). However, the measured carbon-carbon distance in ben- zene is 0.142 nm for all bonds, which is roughly the av- erage length of a single and a double bond. This equiv- alence of all bonds in benzene was explained by Linus Pauling in 1931 within a quantum-mechanical treatment of the benzene ring (Pauling, 1960). The ground state is indeed a quantum-mechanical superposition of the two possible configurations for the double bonds, as shown schematically in Fig. 2 (c). These chemical considerations indicate the way to- wards carbon-based condensed matter physics – any graphitic compound has indeed a sheet of graphene as its basic constituent. Such a graphene sheet may be viewed simply as a tiling of benzene hexagons, where the hydro- gen are replaced by carbon atoms to form a neighboring carbon hexagon [Fig. 2 (d)]. However, graphene has re- mained the basic constituent of graphitic systems during a long time only on the theoretical level. From an exper- imental point of view, graphene is the youngest allotrope (a) y a 2 a 1 x δ 2 δ3 δ 1 a=0.142 nm 3 (b) K M’’ K’ Γ M * M’ a 2 * K a1 M K’ M’ K K’ M’’ : A sublattice : B sublattice 1 and a∗ Figure 3 (a) Honeycomb lattice. The vectors δ1, δ2, and δ3 connect nn carbon atoms, separated by a distance a = 0.142 nm. The vectors a1 and a2 are basis vectors of the triangular Bravais lattice. (b) Reciprocal lattice of the triangular lattice. Its primitive lattice vectors are a∗ 2. The shaded region represents the first Brillouin zone (BZ), with its center Γ and the two inequivalent corners K (black squares) and K ′ (white squares). The thick part of the border of the first BZ represents those points which are counted in its definition such that no points are doubly counted. The first BZ, defined in a strict manner, is, thus, the shaded region plus the thick part of the border. For completeness, we have also shown the three inequivalent cristallographic points M , M ′, and M ′′ (white triangles). and accessible to electronic-transport measurements only since 2004. For a detailed discussion of the different fabrication techniques, the most popular of which are the exfoliation technique (Novoselov et al., 2005b) and thermal graphi- tization of epitaxially-grown SiC crystals (Berger et al., 2004), we refer the reader to existing experimental re- views (Geim and Novoselov, 2007; de Heer et al., 2007). Notice that, more recently, large-scale graphene has been fabricated by chemical vapor deposition (Reina et al., 2009) that seems a promising technique not only for fun- damental research but also for technological applications. B. Crystal Structure of Graphene As already mentioned in the last section, the carbon atoms in graphene condense in a honeycomb lattice due to their sp2 hybridization. The honeycomb lattice is not a Bravais lattice because two neighboring sites are in- equivalent from a crystallographic point of view.2 Fig. 3 (a) illustrates indeed that a site on the A sublattice has nearest neighbors (nn) in the directions north-east, north-west, and south, whereas a site on the B sublattice has nns in the directions north, south-west, and south- east. Both A and B sublattices, however, are triangular Bravais lattices, and one may view the honeycomb lat- tice as a triangular Bravais lattice with a two-atom basis (A and B). The distance between nn carbon atoms is a = 0.142 nm, which is the average of the single (C–C) and double (C=C) covalent σ bonds, as in the case of 2 This needs to be clearly distinguished from a chemical point of view according to which they may be equivalent as in the case of graphene where both types of sites consist of carbon atoms. benzene. C. Electronic Band Structure of Graphene 4 The three vectors which connect a site on the A sub- lattice with a nn on the B sublattice are given by a 2 (cid:16)√3ex + ey(cid:17) , δ2 = a 2 (cid:16)− √3ex + ey(cid:17) , δ3 = −aey, δ1 = (1) and the triangular Bravais lattice is spanned by the basis vectors a1 = √3aex a2 = and (2) √3a 2 (cid:16)ex + √3ey(cid:17) . The modulus of the basis vectors yields the lattice spac- ing, a = √3a = 0.24 nm, and the area of the unit cell is Auc = √3a2/2 = 0.051 nm2. The density of carbon atoms is, therefore, nC = 2/Auc = 39 nm−2 = 3.9 × 1015 cm−2. Because there is one π electron per carbon atom that is not involved in a covalent σ bond, there are as many valence electrons as carbon atoms, and their den- sity is, thus, nπ = nC = 3.9× 1015 cm−2. As discussed in detail below, this density is not equal to the carrier den- sity in graphene, which one measures in electric transport measurements. As we have discussed in the previous section, three electrons per carbon atom in graphene are involved in the formation of strong covalent σ bonds, and one electron per atom yields the π bonds. The π electrons happen to be those responsible for the electronic properties at low energies, whereas the σ electrons form energy bands far away from the Fermi energy (Saito et al., 1998). This section of the introduction is, thus, devoted to a brief discussion of the energy bands of π electrons within the tight-binding approximation, which was originally calcu- lated for the honeycomb lattice by P. R. Wallace in 1947 (Wallace, 1947). 1. Tight-binding model for electrons on the honeycomb lattice In the case of two atoms per unit cell, we may write down a trial wave function ψk(r) = akψ(A) k (r) + bkψ(B) k (r), (5) The reciprocal lattice, which is defined with respect to the triangular Bravais lattice, is depicted in Fig. 3 (b). It spanned by the vectors where ak and bk are complex functions of the quasi- momentum k. Both ψ(A) k (r) are Bloch func- tions with k (r) and ψ(B) a∗1 = 2π √3a(cid:18)ex − ey√3(cid:19) and a∗2 = 4π 3a ey. (3) Physically, all sites of the reciprocal lattice represent equivalent wave vectors. The first Brillouin zone [BZ, shaded region and thick part of the border of the hexagon in Fig. 3 (b)] is defined as the set of inequivalent points in reciprocal space, i.e. of points which may not be con- nected to one another by a reciprocal lattice vector. The long wavelength excitations are situated in the vicinity of the Γ point, in the center of the first BZ. Furthermore, one distinguishes the six corners of the first BZ, which consist of the inequivalent points K and K′ represented by the vectors ± K = ± 4π 3√3a ex. (4) The four remaining corners [shown in gray in Fig. 3 (b)] may indeed be connected to one of these points via a translation by a reciprocal lattice vector. These cristal- lographic points play an essential role in the electronic properties of graphene because their low-energy excita- tions are centered around the two points K and K′, as is discussed in detail in the following section. We empha- sise, because of some confusion in the literature on this point, that the inequivalence of the two BZ corners, K and K′, has nothing to do with the presence of two sub- lattices, A and B, in the honeycomb lattice. The form of the BZ is an intrinsic property of the Bravais lattice, in- dependent of the possible presence of more than one atom in the unit cell. For completeness, we have also shown, in Fig. 3 (b), the three crystallographically inequivalent M points in the middle of the BZ edges. ψ(j) k (r) =XRl eik·Rlφ(j)(r + δj − Rl), (6) where j = A/B labels the atoms on the two sublattices A and B, and δj is the vector which connects the sites of the underlying Bravais lattice with the site of the j atom within the unit cell. The φ(j)(r + δj − Rl) are atomic orbital wave functions for electrons that are in the vicinity of the j atom situated at the position Rl − δj at the (Bravais) lattice site Rl. Typically one chooses the sites of one of the sublattices, e.g. the A sublattice, to coincide with the sites of the Bravais lattice. Notice furthermore that there is some arbitrariness in the choice of the phase in Eq. (6) – instead of choosing exp(ik· Rl), one may also have chosen exp[ik·(Rl−δj)], for the atomic wave functions. The choice, however, does not affect the physical properties of the system because it simply leads to a redefinition of the weights ak and bk which aquire a different relative phase (Bena and Montambaux, 2009). With the help of these wave functions, we may now search the solutions of the Schrodinger equation Hψk = ǫkψk. Multiplication of the Schrodinger equation by ψ∗k from the left yields the equation ψ∗kHψk = ǫkψ∗kψk, which may be rewritten in matrix form with the help of Eq. (5) (a∗k, b∗k)Hk(cid:18) ak bk (cid:19) = ǫk (a∗k, b∗k)Sk(cid:18) ak bk (cid:19) . Here, the Hamiltonian matrix is defined as Hk ≡ ψ(A)∗k Hψ(A) ψ(B)∗k Hψ(A) k k ψ(A)∗k Hψ(B) ψ(B)∗k Hψ(B) k ! = H†k, k (7) (8) and the overlap matrix Sk ≡ ψ(A)∗k ψ(A) ψ(B)∗k ψ(A) k k ψ(A)∗k ψ(B) ψ(B)∗k ψ(B) k ! = S†k k (9) accounts for the non-orthogonality of the trial wave func- tions. The eigenvalues ǫk of the Schrodinger equation yield the electronic bands, and they may be obtained from the secular equation det(cid:2)Hk − ǫλ kSk(cid:3) = 0, (10) which needs to be satisfied for a non-zero solution of the wave functions, i.e. for ak 6= 0 and bk 6= 0. The label λ denotes the energy bands, and it is clear that there are as many energy bands as solutions of the secular equation (10), i.e. two bands for the case of two atoms per unit cell. a. Formal solution. Before turning to the specific case of graphene and its energy bands, we solve formally the sec- ular equation for an arbitrary lattice with several atoms per unit cell. The Hamiltonian matrix (8) may be writ- ten, with the help of Eq. (6), as 5 B 1 a 1 a 3 a 2 B 2 A δ3 B 3 Figure 4 Tight-binding model for the honeycomb lattice. b. Solution for graphene with nearest-neighbor and next- nearest-neighour hopping. After these formal considera- tions, we now study the particular case of the tight- binding model on the honeycomb lattice, which yields, to great accuracy, the π energy bands of graphene. Be- cause all atomic orbitals are pz orbitals of carbon atoms, we may omit the onsite energy ǫ0, as discussed in the last paragraph. We choose the Bravais lattice vectors to be those of the A sublattice, i.e. δA = 0, and the equivalent site on the B sublattice is obtained by the displacement δB = δAB = δ3 (see Fig. 4). The nn hopping amplitude is given by the expression t ≡Z d2r φA∗(r)∆V φB(r + δ3), (14) k = N(cid:16)ǫ(j)sij Hij k + tij k(cid:17) (11) and we also take into account next-nearest neighbor (nnn) hopping which connects neighboring sites on the same sublattice where (δij ≡ δj − δi), k ≡XRl sij eik·RlZ d2r φ(i)∗(r)φ(j)(r + δij − Rl) = Sij k N (12) and we have defined the hopping matrix tij k ≡XRl eik·RlZ d2r φ(i)∗(r)∆V φ(j)(r+δij−Rl) . (13) Here, N is the number of unit cells, and we have sep- arated the Hamiltonian H into an atomic orbital part H a = −(2/2m)∆ + V (r − Rl + δj), which satisfies the eigenvalue equation H aφ(j)(r+δj−Rl) = ǫ(j)φ(j)(r+δj− Rl) and a “perturbative part” ∆V which takes into ac- count the potential term that arises from all other atoms different from that in the atomic orbital Hamiltonian. The last line in Eq. (11) has been obtained from the fact that the atomic wave functions φ(i)(r) are eigenstates of the atomic Hamiltonian H a with the atomic energy ǫ(i) for an orbital of type i. This atomic energy plays the role of an onsite energy. The secular equation now reads det[tij k ] = 0. Notice that, if the the atoms on the different sublattices are all of the same electronic configuration, one has ǫ(i) = ǫ0 for all i, and one may omit this on-site energy, which yields only a constant and physically irrelevant shift of the energy bands. k − ǫ(j))sij k − (ǫλ tnnn ≡Z d2r φA∗(r)∆V φA(r + a1) (15) Notice that one may have chosen any other vector δj or a2, respectively, in the calculation of the hopping ampli- tudes. Because of the normalization of the atomic wave functions, we have R d2rφ(j)∗(r)φ(j)(r) = 1, and we con- sider furthermore the overlap correction between orbitals on nn sites, s ≡Z d2r φA∗(r)φB(r + δ3). (16) We neglect overlap corrections between all other orbitals which are not nn, as well as hopping amplitudes for larger distances than nnn. If we now consider an arbitrary site A on the A sub- lattice (Fig. 4), we may see that the off-diagonal terms of the hopping matrix (13) consist of three terms corre- sponding to the nn B1, B2, and B3, all of which have the same hopping amplitude t. However, only the site B3 is described by the same lattice vector (shifted by δ3) as the site A and thus yields a zero phase to the hopping matrix. The sites B1 and B2 correspond to lattice vectors shifted by a2 and a3 ≡ a2 − a1, respectively. Therefore, they contribute a phase factor exp(ik · a2) and exp(ik · a3), respectively. The off-diagonal elements of the hopping k )∗, as matrix may then be written as3 tAB k )∗, well as those of the overlap matrix sAB (sAA k = 1, due to the above-mentioned normaliza- tion of the atomic wave functions), where we have defined the sum of the nn phase factors k = sγ∗k = (sBA k = tγ∗k = (tBA k = sBB γk ≡ 1 + eik·a2 + eik·a3. (17) The nnn hopping amplitudes yield the diagonal elements of the hopping matrix, tAA k = tBB k = 2tnnn 3 Xi=1 cos(k · ai) = tnnn(cid:0)γk2 − 3(cid:1) , (18) and one obtains, thus, the secular equation 6 y g r e n E π∗ π K’ K K K’ K’ K ky k x det(cid:20) tAA k − ǫk (t − sǫk)γk (t − sǫk)γ∗k k − ǫk (cid:21) = 0 tAA with the two solutions (λ = ±) ǫλ k = tAA k + λtγk 1 + λsγk . Figure 5 Energy dispersion as a function of the wave-vector com- ponents kx and ky, obtained within the tight-binding approxima- tion, for tnnn/t = 0.1. One distinguishes the valence (π) band from the conduction (π∗) band. The Fermi level is situated at the points where the π band touches the π∗ band. The energy is measured in units of t and the wave vector in units of 1/a. (19) (20) This expression may be expanded under the reasonable assumptions s ≪ 1 and tnnn ≪ t, which we further jus- tify at the end of the paragraph, k + λtγk − stγk2 = t′nnnγk2 + λtγk cos(k · ai)# k ≃ tAA ǫλ = t′nnn"3 + 2 +λtvuut3 + 2 3 Xi=1 Xi=1 3 cos(k · ai), (21) where we have defined the effective nnn hopping ampli- tude t′nnn ≡ tnnn − st, and we have omitted the unim- portant constant −3tnnn in the second step. Therefore, the overlap corrections simply yield a renormalization of the nnn hopping amplitudes. The hopping ampli- tudes may be determined by fitting the energy dispersion (21) obtained within the tight-binding approximation to those calculated numerically in more sophisticated band- structure calculations (Partoens and Peeters, 2006) or to spectroscopic measurements (Mucha-Kruczy´nski et al., 2008). These yield a value of t ≃ −3 eV for the nn hopping amplitude and t′nnn ≃ 0.1t, which justifies the above-mentioned expansion for t′nnn/t ≪ 1. Notice that this fitting procedure does not allow for a distinction be- tween the “true” nnn hopping amplitude tnnn and the contribution from the overlap correction −st. We, there- fore, omit this distinction in the following discussion and drop the prime at the effective nnn hopping amplitude, but one should keep in mind that it is an effective pa- rameter with a contribution from nn overlap corrections. 3 The hopping matrix element tAB k corresponds to a hopping from the B to the A sublattice. c. Energy dispersion of π electrons in graphene. The en- ergy dispersion (21) is plotted in Fig. 5 for tnnn/t = 0.1. It consists of two bands, labeled by the index λ = ±, each of which contains the same number of states. Be- cause each carbon atom contributes one π electron and each electron may occupy either a spin-up or a spin- down state, the lower band with λ = − (the π or valence band) is completely filled and that with λ = + (the π∗ or conduction band) completely empty. The Fermi level is, therefore, situated at the points, called Dirac points, where the π band touches the π∗ band. Notice that only if tnnn = 0 the energy dispersion (21) is electron-hole symmetric, i.e. ǫλ k . This means that nnn hop- ping and nn overlap corrections break the electron-hole symmetry. The Dirac points are situated at the points kD where the energy dispersion (21) is zero, k = −ǫ−λ ǫλ kD = 0. (22) Eq. (22) is satisfied when γkD = 0, i.e. when ReγkD = 1 + cos"√3a + cos"√3a (−kD y )# x + √3kD (kD y )# = 0 x + √3kD 2 2 and, equally, ImγkD = sin"√3a + sin"√3a 2 2 y )# x + √3kD (kD x + √3kD (−kD y )# = 0. (23) (24) The last equation may be satisfied by the choice kD and Eq. (23), thus, when y = 0, 1 + 2 cos √3a 2 kD x ! = 0 ⇒ kD x = ± 4π 3√3a . (25) Comparison with Eq. (4) shows that there are, thus, two inequivalent Dirac points D and D′, which are situated at the points K and K′, respectively, 7 k = λtγkΨλ the eigenvalue equation Hk(tnnn = 0)Ψλ k, which does not take into account the nnn hopping cor- rection. Indeed, these eigenstates are also those of the Hamiltonian with tnnn 6= 0 because the nnn term is pro- portional to the one-matrix 1. The solution of the eigen- value equation (29) yields aλ k = λ γ∗k γk k = λe−iϕk bλ bλ k (30) kD = ±K = ± 4π 3√3a ex. (26) and, thus, the eigenstates Although situated at the same position in the first BZ, it is useful to make a clear conceptual distinction between the Dirac points D and D′, which are defined as the contact points between the two bands π and π∗, and the crystallographic points K and K′, which are defined as the corners of the first BZ. There are, indeed, situations where the Dirac points move away from the points K and K′, as we will discuss in Sec. I.D. Notice that the band Hamiltonian (8) respects time- reversal symmetry, Hk = H∗ −k, which implies ǫ−k = ǫk for the dispersion relation. Therefore, if kD is a solution of ǫk = 0, so is −kD, and Dirac points thus necessarily occur in pairs. In graphene, there is one pair of Dirac points, and the zero-energy states are, therefore, doubly degenerate. One speaks of a twofold valley degeneracy, which survives when we consider low-energy electronic excitations that are restricted to the vicinity of the Dirac points, as is discussed in Sec. I.C.2. d. Effective tight-binding Hamiltonian. Before considering the low-energy excitations and the continuum limit, it is useful to define an effective tight-binding Hamiltonian, Hk ≡ tnnnγk2 1 + t(cid:18) 0 γ∗k Here, 1 represents the 2 × 2 one-matrix γk 0 (cid:19) . 0 1(cid:19) . 1 =(cid:18) 1 0 (27) (28) This Hamiltonian effectively omits the problem of non- orthogonality of the wave functions by a simple renor- malization of the nnn hopping amplitude, as alluded to above. It is therefore simpler to treat than the origi- nal one (8) the eigenvalue equation of which involves the overlap matrix Sk, while it yields the same dispersion re- lation (21). The eigenstates of the effective Hamiltonian (27) are the spinors Ψλ k = 1 √2(cid:18) 1 λeiϕk (cid:19) , (31) where ϕk = arctan(Imγk/Reγk). As one may have expected, the spinor represents an equal probability to find an electron in the state Ψλ k on the A as on the B sublattice because both sublattices are built from carbon atoms with the same onsite energy ǫ(i). 2. Continuum limit In order to describe the low-energy excitations, i.e. electronic excitations with an energy that is much smaller than the band width ∼ t, one may restrict the exci- tations to quantum states in the vicinity of the Dirac points and expand the energy dispersion around ±K. The wave vector is, thus, decomposed as k = ±K + q, where q ≪ K ∼ 1/a. The small parameter, which gov- erns the expansion of the energy dispersion, is therefore qa ≪ 1. It is evident from the form of the energy dispersion (21) and the effective Hamiltonian that the basic entity to be expanded is the sum of the phase factors γk. As we have already mentioned, there is some arbitrariness in the definition of γk, as a consequence of the arbitrary choice of the relative phase between the two sublattice components – indeed, a change γk → γk exp(ifk) in Eq. (17) for a real and non-singular function fk does not af- fect the dispersion relation (21), which only depends on the modulus of the phase-factor sum. For the series ex- pansion, it turns out to be more convenient not to use the expression (17), but one with fk = k · δ3, which renders the expression more symmetric (Bena and Montambaux, 2009), Ψλ k (cid:19) , k =(cid:18) aλ k bλ (29) eik·δ3γk = eik·δ1 + eik·δ2 + eik·δ3 (32) the components of which are the probability amplitudes of the Bloch wave function (5) on the two different sublat- tices A and B. They may be determined by considering In the series expansion, we need to distinguish further- more the sum at the K point from that at the K′ point, 3 e±iK·δj eiq·δj (q · δ1)2(cid:21) (q · δ2)2(cid:21) 1 2 γ±q ≡ eik·δ3γk=±K+q = 1 2 Xj=1 ≃ e±i2π/3(cid:20)1 + iq · δ1 − +e∓i2π/3(cid:20)1 + iq · δ2 − +(cid:20)1 + iq · δ3 − = γ±(0) q + γ±(1) q + γ±(2) 1 2 q (q · δ3)2(cid:21) 8 the framework of the discussion of the zero-energy states at the BZ corners. From Eq. (38) it is apparent that the continuum limit qa ≪ 1 coincides with the limit ǫ ≪ t, as described above, because ǫq = 3taq/2 ≪ t then. It is convenient to swap the spinor components at the K′ point (for ξ = −), Ψk,ξ=+ =(cid:18) ψA k,+ (cid:19) , Ψk,ξ=− =(cid:18) ψB k,− (cid:19) , (39) k,− ψB ψA k,+ (33) i.e. to invert the role of the two sublattices. In this case, the effective low-energy Hamiltonian may be represented as By definition of the Dirac points and their position at the BZ corners K and K′, we have γ±(0) = γ±K = 0. We limit the expansion to second order in qa. q a. First order in qa. The first-order term is given by Heff,ξ q = ξvF (qxσx + qyσy) = vF τ z ⊗ q · σ, (40) i.e. as two copies of the 2D Dirac Hamiltonian HD = vF p · σ (with the momentum p = q), where we have introduced the four-spinor representation a = i 3a 2 γ±(1) q (qx ± iqy), −iqya = ∓ 2h(√3qx + qy)e±i2π/3 − (√3qx − qy)e∓i2π/3i which is obtained with the help of sin(±2π/3) = ±√3/2 and cos(±2π/3) = −1/2. This yields the effective low- energy Hamiltonian q = ξvF (qxσx + ξqyσy), Heff,ξ (34) (35) where we have defined the Fermi velocity4 vF ≡ − 3ta 2 = 3ta 2 and used the Pauli matrices 1 0(cid:19) σx =(cid:18) 0 1 and i 0 (cid:19) . σy =(cid:18) 0 −i (36) (37) Furthermore, we have introduced the valley pseudospin ξ = ±, where ξ = + denotes the K point at +K and ξ = − the K′ point at −K modulo a reciprocal lattice vector. The low-energy Hamiltonian (35) does not take into account nnn-hopping corrections, which are propor- tional to γk2 and, thus, occur only in the second-order expansion of the energy dispersion [at order O(qa)2]. The energy dispersion (21) therefore reads ǫλ q,ξ=± = λvFq, (38) independent of the valley pseudospin ξ. We have already alluded to this twofold valley degeneracy in Sec. I.C.1, in q,+ ψA ψB ψB ψA Ψq =    in the last line via the 4 × 4 matrices 0 −σ (cid:19) , τ z ⊗ σ =(cid:18) σ 0 q,+ q,− q,− (41) (42) and σ ≡ (σx, σy). In this four-spinor representation, the first two components represent the lattice components at the K point and the last two components those at the K′ point. We emphasise that one must clearly distin- guish both types of pseudospin: (a) the sublattice pseu- dospin is represented by the Pauli matrices σj , where “spin up” corresponds to the component on one sublat- tice and “spin down” to that on the other one. A rotation within the SU(2) sublattice-pseudospin space yields the band indices λ = ±, and the band index is, thus, in- timitely related to the sublattice pseudospin. (b) The valley pseudospin, which is described by a second set of Pauli matrices τ j , the z-component of which appears in the Hamiltonian (40), is due to the twofold valley degen- eracy and is only indirectly related to the presence of two sublattices. The eigenstates of the Hamiltonian (40) are the four- spinors Ψξ=+ q,λ = 1 λeiϕq 0 0 1 √2    , Ψξ=−q,λ = 1 √2  0 0 1 −λeiϕq where we have, now, ,   (43) 4 The minus sign in the definition is added to render the Fermi velocity positive because the hopping parameter t ≃ −3 eV hap- pens to be negative, as mentioned in the last section. ϕq = arctan(cid:18) qy qx(cid:19) . (44) y g r e n e 0 conduction band (λ = +) valence band (λ = −) η = + η = − η = − (ξ = +) K η = + (ξ = −) K’ momentum Figure 6 Relation between band index λ, valley pseudospin ξ, and chirality η in graphene. In high-energy physics, one defines the he- b. Chirality. licity of a particle as the projection of its spin onto the direction of propagation (Weinberg, 1995), ηq = q · σ q , (45) which is a Hermitian and unitary operator with the eigen- values η = ±, ηqη = ±i = ±η = ±i. Notice that σ describes, in this case, the true physical spin of the particle. In the absence of a mass term, the helicity op- erator commutes with the Dirac Hamiltonian, and the helicity is, therefore, a good quantum number, e.g. in the description of neutrinos, which have approximately zero mass. One finds indeed, in nature, that all neutri- nos are “left-handed” (η = −), i.e. their spin is antipar- allel to their momentum, whereas all anti-neutrinos are “right-handed” (η = +). For massive Dirac particles, the helicity operator (45) no longer commutes with the Hamiltonian. One may, however, decompose a quantum state Ψi describing a massive Dirac particle into its chiral components, with the help of the projectors 1 + ηq 2 1 − ηq 2 Ψi and ΨRi = ΨLi = Ψi. (46) In the case of massless Dirac particles, with a well-defined helicity Ψi = η = ±i, one simply finds Ψ+ +i = +i +i = 0, 1 − ηq Ψ+ Ri = Li = 1 + ηq 2 2 (47) and 1 + ηq 1 − ηq 2 Ψ−Li = Ψ−Ri = −i = −i, −i = 0, (48) such that one may then identify helicity and chirality. Because we are concerned with massless particles in the context of graphene, we make this identification in the remainder of this review and use the term chirality. 2 For the case of graphene, one may use the same def- inition (45), but the Pauli matrices define now the sub- lattice pseudospin instead of the true spin. The operator 9 ηq clearly commutes with the massless 2D Dirac Hamil- tonian (40), and one may even express the latter as Heff,ξ q = ξvFqηq , (49) which takes into account the two-fold valley degeneracy, in terms of the valley pseudospin ξ = ±. The band index λ, which describes the valence and the conduction band, is therefore entirely determined by the chirality and the valley pseudospin, and one finds λ = ξη , (50) which is depicted in Fig. 6. We notice finally that the chirality is a preserved quan- tum number in elastic scattering processes induced by impurity potentials Vimp = V (r)1 that vary smoothly on the lattice scale. In this case, inter-valley scattering is suppressed, and the chirality thus conserved, as a conse- quence of Eq. (50). This effect gives rise to the absence of backscattering in graphene (Shon and Ando, 1998) and is at the origin of Klein tunneling according to which a massless Dirac particle is fully transmitted, under nor- mal incidence, through a high electrostatic barrier with- out being reflected (Katsnelson et al., 2006). This rather counter-intuitive result was first considered as a paradox and led to the formulation of a charged vacuum in the potential barrier (Klein, 1929), which may be indentified in the framework of band theory with a Fermi level in the valence band. c. Higher orders in qa. Although most of the fundamen- tal properties of graphene are captured within the effec- tive model obtained at first order in the expansion of the energy dispersion, it is useful to take into account second- order terms. These corrections include nnn hopping cor- rections and off-diagonal second-order contributions from the expansion of γk. The latter yield the so-called trigo- nal warping, which consists of an anisotropy in the energy dispersion around the Dirac points. The diagonal second-order term, which stems from the nnn hopping, is readily obtained from Eq. (34), nnn = tnnnγξ Hξ q2 1 ≃ tnnnγξ(1) q 2 1 = independent of the valley index ξ. 9a2 4 tnnnq2 1, (51) q The off-diagonal second-order terms are tγξ(2) = −vF a(qx − iξqy)2/4. Notice that there is a natural energy hierarchy between the diagonal and off-diagonal second-order terms when compared to the leading lin- ear term; whereas the off-diagonal terms are on the or- der O(qa) as compared to the energy scale vFq, the diagonal term is on the order O((tnnn/t)qa) and thus roughly an order of magnitude smaller. We therefore take into account also the off-diagonal third order term tγξ(3) q = −ξvF a2(qx+iξqy)q2/8, which also needs to be considered when calculating the high-energy corrections of the energy levels in a magnetic field (see Sec. II.B). Up to third order, the off-diagonal terms therefore read (a) tγξ q = ξvF h(qx + iξqy) − ξ 8 q2(qx + iξqy)(cid:21) , − a2 a 4 (qx − iξqy)2 K’ -2 -3 -1 (52) (b) Γ 1 k x K 2 3 -0.4 -0.2 k y 3 2 1 -1 -2 -3 10 2eV 1.5eV 1eV K’ 0.4 0.2 xq 0.4 q y 0.2 -0.2 -0.4 where one may omit the valley-dependent sign before the y-components of the wave vector by sweeping the sublat- tice components in the spinors when changing the valley. In order to appreciate the influence of the second-order off-diagonal terms on the energy bands, we need to cal- culate the modulus of γξ q, γξ q ≃ 3a 2 q(cid:20)1 − ξqa 4 cos(3ϕq)(cid:21) , (53) 4 where we have used the parametrization qx = q cos ϕq and qy = q sin ϕq, and where we have restricted the expansion to second order. Finally, the energy dispersion (21) expanded to second order in qa reads tnnnq2 + λvFq(cid:20)1 − ξqa ǫλ q,ξ = cos(3ϕq)(cid:21) . 9a2 4 q,ξ = −ǫλ As mentioned in Sec. (54) I.C.1, it is apparent from Eq. (54) that the nnn correction breaks the electron-hole symmetry ǫ−λ q,ξ. This is, however, a rather small correction, of order qatnnn/t, to the first-order effective Hamiltonian (40). The second-order expansion of the phase factor sum γq yields a more relevant cor- rection – the third term in Eq. (54), that is of order qa ≫ qatnnn/t – to the linear theory. It depends ex- plicitly on the valley pseudospin ξ and renders the energy dispersion anisotropic in q around the K and K′ point. The tripling of the period, due to the term cos(3ϕq), is a consequence of the symmetry of the underlying lattice and is precisely the origin of trigonal warping. The trigonal warping of the dispersion relation is visu- alized in Fig. 7, where we have plotted the contours of constant (positive) energy in Fourier space. The closed energy contours around the K and K′ points at low en- ergy are separated by the high-energy contours around the Γ point by the dashed lines in Fig. 7 (a) at en- ergy t + tnnn the crossing points of which correspond to the M points. As mentioned above, the dispersion relation has saddle points at these points at the border of the first BZ, which yield van Hove singularities in the density of states. In Fig. 7 (b), we compare constant- energy contours of the full dispersion relation to those obtained from Eq. (54) calculated within a second-order expansion. The contours are indistinguishable for an en- ergy of ǫ = t/3 ≃ 1 eV, and the continuum limit yields rather accurate results up to energies as large as 2 eV. Notice that, in today’s exfoliated graphene samples on SiO2 substrates, one may probe, by field-effect doping of the graphene sheet, energies which are on the order of 100 meV. Above these energies the capacitor breaks Figure 7 Contours of constant (positive) energy in reciprocal space. (a) Contours obtained from the full dispersion relation (21). The dashed line corresponds to the energy t + tnnn, which sepa- rates closed orbits around the K and K ′ points (black lines, with energy ǫ < t + tnnn ) from those around the Γ point (gray line, with energy ǫ > t + tnnn). (b) Comparison of the contours at energy ǫ = 1 eV, 1.5 eV, and 2 eV around the K ′ point. The black lines correspond to the energies calculated from the full dispersion re- lation (21) and the gray ones to those calculated to second order within the continuum limit (54). a 2 a t nnn t’nnn 1 t t a a t’nnn t’nnn a’ t’ axis of deformation t’nnn t nnn Figure 8 Quinoid-type deformation of the honeycomb lattice – the bonds parallel to the deformation axis (double arrow) are mod- ified. The shaded region indicates the unit cell of the oblique lattice, spanned by the lattice vectors a1 and a2. Dashed and dashed-dotted lines indicate next-nearest neighbors, with charac- teristic hopping integrals tnnn and t′ nnn, respectively, which are different due to the lattice deformation. down, and Fig. 7 (a) indicates that the continuum limit (54) yields extremely accurate results at these energies. We finally mention that, when higher-order terms in qa are taken into account, the chirality operator (45) no longer commutes with the Hamiltonian. Chirality is therefore only a good quantum number in the vicinity of the Dirac points. D. Deformed Graphene In the previous section, we have considered a per- fect honeycomb lattice which is invariant under a 2π/3 rotation. As a consequence, all hopping parameters along the nn bonds δj were equal. An interesting situation arises when the graphene sheet is deformed, such that rotation symmetry is broken. In order to illustrate the consequences, we may apply a uniax- 11 whereas in quinoid-type graphene b′ = b(1 + ε/2), which gives t′nnn = tnnn(1 − 2ε + bε/2d). (58) The electronic properties of quinoid-type graphene may then be described in terms of an effective Hamil- tonian of the type (27), Hk = tnnnhk1 + t(cid:18) 0 γ∗k γk 0 (cid:19) , with (Goerbig et al., 2008) hk = 2 cos√3kxa + 2 √3kxa + cos"− 2 t′nnn tnnn (cos"√3kxa + ǫ(cid:19)#) , + kya(cid:18) 3 2 2 + kya(cid:18) 3 2 (59) + ǫ(cid:19)# (60) and the off-diagonal elements γk = 2eikya(3/2+ǫ) cos √3 2 kxa! + (1 − 2ǫ). (61) The resulting energy dispersion (62) ǫλ k = tnnnhk + λtγk 9 for an unphysically large defor- is plotted in Fig. mation, ǫ = 0.4, for illustration reasons. Notice that the reversible deformations are limited by a value of ǫ ∼ 0.1...0.2 beyond which the graphene sheet cracks (Lee et al., 2008). One notices, in Fig. 9, two effects of the deformation: i) the Dirac points no longer coin- cide with the corners of the first BZ, the form of which is naturally also modified by the deformation; and ii) the cones in the vicinity of the Dirac points are tilted, i.e. the nnn hopping term (60) breaks the electron-hole symme- try already at linear order in qa. These two points are discussed in more detail in the following two subsections. 1. Dirac point motion In order to evaluate quantitatively the position of the Dirac points, which are defined as the contact points be- tween the valence (λ = −) and the conduction (λ = +) bands, one needs to solve the equation γkD = 0, in anal- ogy with the case of undeformed graphene discussed in Sec. I.C.1. One then finds kD y = 0 and kD x a = ξ 2 √3 arccos(cid:18)− t′ 2t(cid:19) , (63) where the valley index ξ = ± denotes again the two in- equivalent Dirac points D and D′, respectively. As al- ready mentioned, the Dirac points D and D′ coincide, for undistorted graphene, with the crystallographic points K and K′, respectively, at the corners of the first BZ. The Figure 9 Band dispersion of the quinoid-type deformed the hon- eycomb lattice, for a lattice distortion of δa/a = −0.4, with t = 3 eV, tnnn/t = 0.1, ∂t/∂a = −5 eV/A, and ∂tnnn/∂a = −0.7 eV/A. The inset shows a zoom on one of the Dirac points, D′. ial strain in the y-direction,5 a → a′ = a + δa, in which case one obtains a quinoid-type deformation 8). The hopping t′ along δ3 is then differ- (Fig. ent from that t along δ1 and δ2 (Dietl et al., 2008; Farjam and Rafii-Tabar, 2008; Hasegawa et al., 2006; Wunsch et al., 2008; Zhu et al., 2007), 2009; Goerbig et al., t → t′ = t + ∂t ∂a δa. (55) Furthermore, also four of six nnn hopping integrals are affected by the strain (see Fig. 8), tnnn → t′nnn = tnnn + ∂tnnn ∂a δa. (56) If one considers a moderate deformation ǫ ≡ δa/a ≪ 1, the effect on the hopping amplitudes may be estimated with the help of Harrison’s law (Harrison, 1981), accord- ing to which t = C2/ma2, where C is a numerical pref- actor of order 1. One therefore finds a value and = − 2t a ∼ −4.3 eV/A ∂t t′ = t(1 − 2ǫ) (57) ∂a which coincides well with the value ∂t/∂a ≃ 5 eV/A, which may be found in the literature (Dillon et al., 1977; Saito et al., 1998). The estimation of the modified nnn hopping integral t′nnn is slightly more involved. One may use a law tnnn(b, a) ≈ t(a) exp[−(b − a)/d(a)] familiar in the context of the extended Huckel model (Salem, 1966), where b is the nnn distance, and d ≈ a/3.5 ≈ 0.4 A is a caracteristic distance related to the overlap of atomic In undeformed graphene one has b = a√3, orbitals. 5 In our simplified model, we only consider one bond length changed by the strain. The more general case has been con- sidered by Peirera et al. (Pereira et al., 2009). However, the main effects are fully visible in the simplified model. (a) (b) (c) (d) Figure 10 Topological semi-metal insulator transition in the model (64) driven by the gap parameter ∆. (a) Two well-separated Dirac cones for ∆ ≪ 0, as for graphene. (b) When lowering the modulus of the (negative) gap parameter, the Dirac points move towards a single point. (c) The two Dirac points merge into a single point at the transition (∆ = 0). The band dispersion remains linear in the qy-direction while it becomes parabolic in the qx-direction. (d) Beyond the transition (∆ > 0), the (parabolic) bands are separated by a band gap ∆ (insulating phase). From Montambaux et al., 2009a. distortion makes both pairs of points move in the same direction due to the negative value of ∂t/∂a. However, unless the parameters are fine-tuned, this motion is dif- ferent, and the two pairs of points no longer coincide. One further notices that Eq. (63) has (two) solutions only for t′ ≤ 2t. Indeed, the two Dirac points merge at the characteristic point M′′ at the border of the first BZ (see Fig. 3). The point t′ = 2t is special insofar as it characterizes a topological phase transition between a semi-metallic phase (for t′ < 2t) with a pair of Dirac cones and a band insulator (for t′ > 2t) (Dietl et al., 2008; Esaki et al., 2009; Montambaux et al., 2009a,b; Pereira et al., 2009; Wunsch et al., 2008). In the vicin- ity of the transition, one may expand the Hamiltonian (59) around the merging point M′′ (Montambaux et al., 2009a,b), and one finds6 q = HM 0 ∆ + 2q2 x 2m∗ + icqy ∆ + 2q2 x 2m∗ − icqy 0 ! , (64) in terms of the mass m∗ = 22/3ta2 and the velocity c = 3ta/ (Montambaux et al., 2009b). The gap param- eter ∆ = t′ − 2t changes its sign at the transition – it is negative in the semi-metallic and positive in the insulat- ing phase, where it describes a true gap (Fig. 10). The Hamiltonian (64) has quite a particular form in it is linear in the the vicinity of the merging points: 6 We do not consider the diagonal part of the Hamiltonian, here, i.e. we choose tnnn = 0, because it does not affect the position of the Dirac points. 12 qy-direction, as one would expect for Dirac points, but it is quadratic in the qx-direction (Dietl et al., 2008). This is a general feature of merging points, which may only occur at the Γ point or else at half a reciprocal lattice vector G/2, i.e. in the center of a BZ border line (such as the M points) (Montambaux et al., 2009a). Indeed, one may show that in the case of a time-reversal symmetric Hamiltonian, the Fermi velocity in the x-direction then vanishes such that one must take into account the quadratic order in qx in the energy band. Notice that such hybrid semi-Dirac points, with a linear-parabolic dispersion relation, are unaccessible in graphene be- cause unphysically large strains would be required (Lee et al., 2008; Pereira et al., 2009). However, such points may exist in other physical systems such as cold atoms in optical lattices (Hou et al., 2009; Lee et al., 2008; Zhao and Paramekanti, 2009; Wunsch et al., the quasi-2D organic ma- 2006; Zhu et al., 2007), terial α−(BEDT-TTF)2I3 2006; Kobayashi et al., 2007) or VO2/TiO2 heterostructures (Banerjee et al., 2009). (Katayama et al., 2. Tilted Dirac cones Another aspect of quinoid-type deformed graphene and a consequence of the fact that the Dirac points no longer coincide with the BZ corners K and K′ of high crystallo- graphic symmetry is the tilt of the Dirac cones. This may be appreciated when expanding the Hamiltonian (59) to linear order around the Dirac points ξkD, in- stead of an expansion around the point M′′ as in the last subsection. In contrast to the undeformed case (51), the diagonal components hk now yield a linear contribu- tion (Goerbig et al., 2008), tnnnhξkD+q1 ≃ ξw0· q 1, in terms of the tilt velocity 2√3 w0x = (tnnna sin 2θ + t′nnna sin θ) and w0y = 0, (65) where we have defined θ ≡ arccos(−t′/2t). The linear model is therefore described by the Hamiltonian,7 Hξ q = ξ(w0 · q1 + wxqxσx + wyqyσy), (66) with the renormalized anisotropic velocities √3ta wx = sin θ and wy = 3 2 t′a (cid:18)1 + 2 3 ǫ(cid:19) . 7 This model may be viewed as the minimal form of the generalized Weyl Hamiltonian (with σ0 ≡ 1) HW = X µ=0,...,3 vµ · q σµ, which is the most general 2 × 2 matrix Hamiltonian that yields a linear dispersion relation. Diagonalizing the Hamiltonian (66) yields the disper- sion relation ǫξ λ(q) = w0 · q + λqw2 xq2 x + w2 yq2 y, (67) λ(q) = ǫξ and one notices that the first term (w0·q) breaks indeed the symmetry ǫξ λ(−q) in each valley, i.e. it tilts the Dirac cones in the direction opposite to w0, as well as the electron-hole symmetry ǫλ(q) = −ǫ−λ(q) at the same wave vector.8 Indeed, the linearity in q of the generalized Weyl Hamiltonian (66) satisfies only the symmetry Hξ q = −Hξ −q inside each valley. Furthermore, one notices that the chiral symme- try is preserved even in the presence of the tilt term if one redefines the chirality operator (45) as ηq = y, which naturally commutes with the Hamiltonian (66). The eigenstates of the chirality operator are still given by (wxqxσx + wyqyσy)/qw2 x + w2 xq2 yq2 ψη = 1 √2(cid:18) 1 ηe−iϕq (cid:19) , (68) with tan ϕk ≡ wyqy/wxqx, and one notices that these states are also the natural eigenstates of the Hamiltonian (66). One finally notices that not all values of the tilt pa- rameter w0 are indeed physical. In order to be able to associate λ = + to a positive and λ = − to a negative energy state, one must fulfill the condition w0 < 1, in terms of the tilt parameter w0 ≡s(cid:18) w0x wx (cid:19)2 wy (cid:19)2 +(cid:18) w0y . (69) (70) In the particular case of the deformation in the y-axis, which is discussed here and in which case w0y = 0 [see Eq. (65)], the general form of the tilt parameter reduces to w0 = w0x/wx. Unless this condition is fulfilled, the iso-energetic lines are no longer ellipses but hyperbolas. In quinoid-type deformed graphene, the tilt parameter may be evaluated as (Goerbig et al., 2008) t′nnn t + sin 2θ sin θ t (cid:19) ≃ w0 = 2(cid:18) tnnn 2 t2 (tt′nnn−t′tnnn) ≃ 0.6ǫ, (71) where we have used Eqs. (57) and (58). Even at mod- erate deformations (ǫ < 0.1), the tilt of the Dirac cones is on the order of 5%, and one may therefore hope to 8 In the absence of the tilt term w0 · q1, this is a consequence of the symmetry σzHσz = −H, which is satisfied both by the effective Hamiltonian (27) for tnnn = 0 and the linearised version (40) in each valley for undeformed graphene. 13 observe the effect, e.g. in angle-resolved photoemis- sion spectroscopy (ARPES) measurements (Damascelli, 2004) that have been successfully applied to epitaxial graphene (Bostwick et al., 2007) and graphitic samples (Zhou et al., 2006). Notice that the Dirac cones are naturally tilted in α−(BEDT-TTF)2I3 (Katayama et al., 2006; Kobayashi et al., 2007), where the Dirac points oc- cur at positions of low crystallographic symmetry within the first BZ. II. DIRAC EQUATION IN A MAGNETIC FIELD AND THE RELATIVISTIC QUANTUM HALL EFFECT As already mentioned in the introduction, a key ex- periment in graphene research was the discovery of a particular quantum Hall effect (Novoselov et al., 2005a; Zhang et al., 2005), which unveiled the relativistic nature of low-energy electrons in graphene. For a deeper under- standing of this effect and as a basis for the following parts, we discuss here relativistic massless 2D fermions in a strong quantizing magnetic field (Sec. II.A). The limits of the Dirac equation in the treatment of the high- II.B, field properties of graphene are discussed in Sec. and we terminate this section with a discussion of the relativistic Landau level spectrum in the presence of an in-plane electric field (Sec. II.C) and that of deformed graphene (Sec. II.D). A. Massless 2D Fermions in a Strong Magnetic Field In order to describe free electrons in a magnetic field, one needs to replace the canonical momentum p by the gauge-invariant kinetic momentum (Jackson, 1999) p → Π = p + eA(r), (72) where A(r) is the vector potential that generates the magnetic field B = ∇ × A(r). The kinetic momentum is proportional to the electron velocity v, which must natu- rally be gauge-invariant because it is a physical quantity. In the case of electrons on a lattice, the substitution (72), which is then called Peierls substitution, remains correct as long as the lattice spacing a is much smaller than the magnetic length lB =r eB , (73) which is the fundamental length scale in the presence of a magnetic field. Because a = 0.24 nm and lB ≃ 26 nm/pB[T], this condition is fulfilled in graphene for the magnetic fields, which may be achieved in today’s high-field laboratories (∼ 45 T in the continuous regime and ∼ 80 T in the pulsed regime). With the help of the (Peierls) substitution (72), one may thus immediately write down the Hamiltonian for charged particles in a magnetic field if one knows the Hamiltonian in the absence of the field, H(p) → H(Π) = H(p + eA) = H B(p, r). (74) Notice that because of the spatial dependence of the vector potential, the resulting Hamiltonian is no longer translation invariant, and the (canonical) momentum p = q is no longer a conserved quantity. For the Dirac Hamiltonian (40), which we have derived in the preced- ing section to lowest order in qa, the Peierls substitution yields Hξ B = ξvF (qxσx + qyσy) → Heff,ξ B = ξvF (Πxσx + Πyσy). (75) We further notice that, because electrons do not only possess a charge but also a spin, each energy level re- sulting from the diagonalization of the Hamiltonian (75) is split into two spin branches separated by the Zeeman effect ∆Z = gµBB, where g is the g-factor of the host material [g ∼ 2 for graphene (Zhang et al., 2006)] and µB = e/2m0 is the Bohr magneton, in terms of the bare electron mass m0. In the remainder of this sec- tion, we concentrate on the orbital degrees of freedom which yield the characteristic level structure of electrons in a magnetic field and therefore neglect the spin degree of freedom, i.e. we consider spinless fermions. Effects related to the internal degrees of freedom are discussed in a separate section (Sec. V) in the framework of the quantum-Hall ferromagnet. 1. Quantum-mechanical treatment One may easily treat the Hamiltonian (75) quantum- mechanically with the help of the standard canonical quantization (Cohen-Tannoudji et al., 1973), according to which the components of the position r = (x, y) and the associated canonical momentum p = (px, py) sat- isfy the commutation relations [x, px] = [y, py] = i and [x, y] = [px, py] = [x, py] = [y, px] = 0. As a conse- quence of these relations, the components of the kinetic momentum no longer commute, and, with the help of the commutator relation (Cohen-Tannoudji et al., 1973) [O1, f (O2)] = df dO2 [O1,O2] (76) between two arbitrary operators, the commutator of which is an operator that commutes itself with both O1 and O2, one finds [Πx, Πy] = −ie(cid:18) ∂Ay ∂x − ∂Ax ∂y (cid:19) = −i 2 l2 B , (77) in terms of the magnetic length (73). For the quantum-mechanical solution of the Hamilto- nian (75), it is convenient to use the pair of conjugate op- erators Πx and Πy to introduce ladder operators in the same manner as in the quantum-mechanical treatment 14 of the one-dimensional harmonic oscillator. These lad- der operators play the role of a complex gauge-invariant momentum (or velocity), and they read a = lB√2 (Πx − iΠy) and a† = lB√2 (Πx + iΠy) , (78) where we have chosen the appropriate normalization such as to obtain the usual commutation relation [a, a†] = 1. (79) It turns out to be helpful for practical calculations to invert the expression for the ladder operators (78), Πx = √2lB (cid:0)a† + a(cid:1) and Πy = i√2lB (cid:0)a† − a(cid:1) . (80) 2. Relativistic Landau levels In terms of the ladder operators (78), the Hamiltonian (75) becomes B = ξ√2 H ξ vF lB (cid:18) 0 a a† 0 (cid:19) . (81) One remarks the occurence of a characteristic frequency ω′ = √2vF /lB, which plays the role of the cyclotron frequency in the relativistic case. Notice, however, that this frequency may not be written in the form eB/mb because the band mass is strictly zero in graphene, such that the frequency would diverge.9 The eigenvalues and the eigenstates of the Hamilto- nian (81) are readily obtained by solving the eigenvalue equation H ξ Bψn = ǫnψn, in terms of the 2-spinors, ψn =(cid:18) un vn (cid:19) . (82) We thus need to solve the system of equations ξω′a vn = ǫn un and ξω′a† un = ǫn vn, (83) which yields the equation ω′(cid:17)2 a†a vn =(cid:16) ǫn vn (84) for the second spinor component. One may therefore identify, up to a numerical factor, the second spinor com- ponent vn with the eigenstate ni of the usual number operator a†a, with a†ani = nni in terms of the inte- ger n ≥ 0. Furthermore, one observes that the square of the energy is proportional to this quantum number, 9 Sometimes, a density-dependent cyclotron mass mC is formally introduced via the equality ω′ ≡ eB/mC . (a) 4 y g r e n 2 e 0 -2 -4 (b) B +,n=4 +,n=3 +,n=2 +,n=1 1 2 magnetic field 3 4 B n=0 5 −,n=1 −,n=2 −,n=3 −,n=4 η R r Figure 11 (a) Relativistic Landau levels as a function of the mag- netic field. (b) Semi-classical picture of cyclotron motion described by the cyclotron coordinate η, where the charged particle turns around the guiding center R. The gray region depicts the uncer- tainty on the guiding center, as indicated by Eq. (98). n = (ω′)2n. This equation has two solutions, a posi- ǫ2 tive and a negative one, and one needs to introduce an- other quantum number λ = ±, which labels the states of positive and negative energy, respectively. This quantum number plays the same role as the band index (λ = + for the conduction and λ = − for the valence band) in the zero-B-field case discussed in the preceding section. One thus obtains the spectrum (McClure, 1956) ǫλ,n = λ √2n vF lB (85) of relativistic Landau levels (LLs) that disperse as λ√Bn as a function of the magnetic field [see Fig. 11(a)]. Notice that, as in the B = 0 case, the level spectrum is two-fold valley-degenerate. Once we know the second spinor component, the first component is obtained from Eq. (83), which reads un ∝ a vn ∼ ani ∼ n − 1i because of the usual equations a†ni = √n + 1n + 1i and ani = √nn − 1i (86) for the ladder operators, where the last equation is valid for n > 0. One then needs to distinguish the zero-energy LL (n = 0) from all other levels. Indeed, for n = 0, the first component is zero because an = 0i = 0 In this case one obtains the spinor ψn=0 =(cid:18) 0 n = 0i(cid:19) . (87) (88) In all other cases (n 6= 0), one has positive and nega- tive energy solutions, which differ among each other by a relative sign in one of the components. A convenient representation of the associated spinors is given by ψξ λ,n6=0 = 1 ξλni (cid:19) . √2(cid:18) n − 1i (89) 15 The particular form of the n = 0 spinor (88) associated with zero-energy states merits a more detailed comment. One notices that only the second spinor component is non-zero. Remember that this component corresponds to the B sublattice in the K-valley (ξ = +) and to the A sublattice in the K′-valley (ξ = −) – the valley pseu- dospin therefore coincides with the sublattice pseudospin, and the two sublattices are decoupled at zero energy. No- tice that this is also the case in the absence of a magnetic field, where the relation (50) between the chirality, the band index and the valley pseudospin is only valid at non- zero values of the wave vector, i.e. not exactly at zero energy. Indeed, the chirality can no longer be defined as the projection of the sublattice pseudospin on the direc- tion of propagation q/q, which is singular at q = 0. At zero energy, it is therefore useful to identify the chiral- ity with the valley pseudospin. Notice, however, that this particularity concerns, in the absence of a magnetic field, only a non-extensive number of states (only two) because of the vanishing density of states at zero energy, whereas the zero-energy LL n = 0 is macroscopically degenerate, as discussed in the following paragraphs. a. LL degeneracy. A particular feature of LLs, both rel- ativistic and non-relativistic ones, consists of their large degeneracy, which equals the number of flux quanta NB = A× B/(h/e) threading the 2D surface A occupied by the electron gas. From the classical point of view, this degeneracy is related to the existence of a constant of motion, namely the position of the guiding center, i.e. the center of the classical cyclotron motion. Indeed, due to translation invariance in a uniform magnetic field, the energy of an electron does not depend on the position of this guiding center. Translated to quantum mechanics, this means that the operator corresponding to this guid- ing center R = (X, Y ) commutes with the Hamiltonian H(p + eA). In order to understand how the LL degeneracy is re- lated to the guiding-center operator, we formally decom- pose the position operator r = R + η (90) into its guiding center R and the cyclotron variable η = (ηx, ηy), as depicted in Fig. 11(b). Whereas the guiding center is a constant of motion, as mentioned above, the cyclotron variable describes the dynamics of the electron in a magnetic field and is, classically, the time-dependent component of the position. Indeed, the cyclotron variable is perpendicular to the electron’s ve- locity and thus related to the kinetic momentum Π by ηx = Πy eB and ηy = − Πx eB , (91) which, as a consequence of the commutation relations (77), satisfy [ηx, ηy] = [Πx, Πy] (eB)2 = −il2 B, (92) whereas they commute naturally with the guiding-center components X and Y . Equation (92) thus induces the commutation relation [X, Y ] = −[ηx, ηy] = il2 B, (93) in order to satisfy [x, y] = 0. These commutation relations indicate that the compo- nents of the guiding-center operator form a pair of conju- gate variables, and one may introduce, in the same man- ner as for the kinetic momentum operator Π, the ladder operators b = 1 √2lB (X + iY ) and b† = 1 √2lB (X − iY ), (94) which again satisfy the usual commutation relations [b, b†] = 1 and which naturally commute with the Hamil- tonian. One may then introduce a number operator b†b associated with these ladder operators, the eigenstates of which satisfy the eigenvalue equation b†bmi = mmi. (95) One thus obtains a second quantum number, an integer m ≥ 0, which is necessary to describe the full quantum states in addition to the LL quantum number n, and the completed quantum states (88) and (89) then read ψξ n=0,m = ψξ n=0 ⊗ mi =(cid:18) 0 n = 0, mi(cid:19) and ψξ λn,m = ψξ λn ⊗ mi = 1 ξλn, mi (cid:19) , √2(cid:18) n − 1, mi (96) (97) respectively. One may furthermore use the commutation relation (93) for counting the number of states, i.e. the degener- acy, in each LL. Indeed, this relation indicates that one may not measure both components of the guiding cen- ter simultaneously, which is therefore smeared out over a surface ∆X∆Y = 2πl2 B, (98) as it is depicted in Fig. 11(b). The result (98) for the surface occupied by a quantum state may be calculated rather simply if one chooses a particular gauge, such as the Landau or the symmetric gauge for the vector poten- tial, but its general derivation is rather involved (Imry, 1997). This minimal surface plays the same role as the surface (action) h in phase space and therefore allows us to count the number of possible quantum states of a given (macroscopic) surface A, = A 2πl2 B = nB × A, NB = A ∆X∆Y (99) where we have introduced the flux density nB = 1 2πl2 B = B h/e , 16 (100) which is nothing other than the magnetic field measured in units of the flux quantum h/e, as already mentioned above. The ratio between the electronic density nel and this flux density then defines the filling factor ν = nel nB = hnel eB , (101) which characterizes the filling of the different LLs. effect. The relativistic quantum Hall integer b. The quantum Hall effect (IQHE) in 2D electron systems (v. Klitzing et al., 1980) is a manifestation of the LL quantization and the macroscopic degeneracy (100) of each level, as well as of semi-classical electron localiza- tion due to the sample impurities.10 In a nutshell, this energy quantization yields a quantization of the Hall re- sistance RH = h e2N , (102) where N = [ν] is the integer part of the filling factor (101), while the longitudinal resistance vanishes.11 The resistance quantization reflects the presence of an incom- pressible quantum liquid with gapped single-particle and density excitations. In the case of the IQHE, at integer filling factors, the gap is simply given by the energy dif- ference between adjacent LLs which must be overcome by an electron that one adds to the system. Notice that if one takes into account the electron spin and a vanish- ing Zeeman effect, the condition for the occurence of the IQHE is satisfied when both spin branches of the last LL n are completely filled, and one thus obtains the Hall- resistance quantization at the filling factors νIQHE = 2n, (103) i.e. for even integers. Odd integers may principally be observed at higher magnetic fields when the Zeeman ef- fect becomes prominent, and the energy gap is then no longer given by the inter-LL spacing but by the Zeeman gap. This picture is naturally simplistic and needs to be modified if one takes into account electronic interactions 10 Strictly speaking, the IQHE requires only the breaking of trans- lation invariance, which in a diffusive sample is due to impuri- ties. In a ballistic sample, translation invariance is broken via the sample edges (Buttiker, 1992). 11 A simultaneous measurement of the Hall and the longitudinal re- sistance requires a particular geometry with at least four electric contacts [for a recent review on the quantum Hall effect, see Ref. (Goerbig, 2009)]. – their consequences, such as the fractional quantum Hall effect or ferromagnetic states are discussed, in the con- text of graphene, in Sec. V. The phenomenology of the relativistic quantum Hall effect (RQHE) in graphene is quite similar to that of the IQHE. Notice, however, that one is confronted not only with the two-fold spin degeneracy of electrons in graphene (in the absence of a strong Zeeman effect), but also with the two-fold valley degeneracy due to the presence of the K and K′ points in the first BZ, which govern the low-energy electronic properties. The filling factor therefore changes by steps of 4 between adjacent plateaus in the Hall resistance. Furthermore, the fill- ing factor (101) is defined in terms of the carrier density which vanishes at the Dirac point. This particle-hole symmetric situation naturally corresponds to a half-filled zero-energy LL n = 0, whereas all levels with λ = − are completely filled and all λ = + levels are unoccupied. In the absence of a Zeeman effect and electronic inter- actions, there is thus no quantum Hall effect at ν = 0, and the condition of a completely filled (or empty) n = 0 LL is found for ν = 2 (ν = −2). As a consequence, the signature of the RQHE is a Hall-resistance quantization at the filling factors (Gusynin and Sharapov, 2005, 2006; Peres et al., 2006) 17 Figure 12 Transmission spectroscopy on epitaxial multilayer graphene (Plochocka et al., 2008). The inset shows a representative transmission spectrum. The main figure represents the positions of the absorption lines as a function of the square-root of the mag- netic field. The dashed lines correspond to transitions calculated at linear order, in agreement with the Dirac equation, whereas one notices downward deviations in the high-energy limit. the conduction and the valence band, and the transitions have the energies vF lB hp2(n + 1) − ξ√2ni , (105) νRQHE = 2(2n + 1), (104) ∆n,λ = which needs to be contrasted to the series (103) of the IQHE in non-relativistic 2D electron systems. The se- ries (104) has indeed been observed in 2005 within the quantum Hall measurements (Novoselov et al., 2005a; Zhang et al., 2005), which thus revealed the relativistic character of electrons in exfoliated graphene. More re- cently, the RQHE has been observed also in epitaxial graphene with moderate mobilities (Jobst et al., 2010; Shen et al., 2009; Wu et al., 2009). c. Experimental observation of relativistic Landau levels. The √Bn dispersion of relativistic LLs has been observed experimentally in transmission spectroscopy, where one shines monochromatic light on the sample and measures the intensity of the transmitted light. Such experiments have been performed both on epitaxial (Sadowski et al., 2006) and exfoliated graphene (Jiang et al., 2007a). When the monochromatic light is in resonance with a dipole-allowed transition from the (partially) filled (λ, n) to the (partially) unoccupied LL (λ′, n ± 1), it is ab- sorbed due to an electronic excitation between the two levels. Notice that, in a non-relativistic 2D electron gas, the only allowed dipolar transition is that from the last occupied LL n to the first unoccupied one n + 1. The transition energy is ωC, independently of n, and one therefore observes a single absorption line (cyclotron res- onance) that is robust to electron-electron interactions, as a consequence of Kohn’s theorem (Kohn, 1961). In graphene, however, there are many more allowed transitions due to the presence of two electronic bands, where λ = + denotes an intraband and λ = − an interband transition (Abergel and Fal’ko, 2007; Iyengar et al., 2007; Sadowski et al., 2006). One there- fore obtains families of resonances the energy of which disperses as ∆n,λ ∝ √B, as it has been observed in the experiments [see Fig. 12, where we show the results from Ref. (Plochocka et al., 2008)]. Notice that the dashed lines in Fig. 12 are fits with a single fitting parameter (the Fermi velocity vF ), which matches well all experi- mental points for different values of n in the low-energy regime. Moreover, the relativistic LLs have later been directly observed in scanning-tunneling spectroscopy in graphene on a graphite substrate12 (Li et al., 2009a) as well as on epitaxial graphene (Song et al., 2010). B. Limits of the Dirac Equation in the Description of Graphene Landau Levels Transmission spectroscopy is an ideal tool for the study of the high-energy part of the LL spectrum when consid- ering the transitions (λ = −, n) → (λ = +, n ± 1), for n ≫ 1. As discussed in Sec. I.C.2, one expects devi- ations [of order O(q2a2)] from the linear dispersion in this limit. These deviations renormalize the energy of the LLs and thus the transition energies. 12 With the help of the same technique, relativistic LLs had before been identified even in graphite (Li and Andrei, 2007). In order to quantify the effect (Plochocka et al., 2008), we may use the Peierls substitution (72) and the expres- sions (80) in the terms (51) and (52) corresponding to the higher-order diagonal and off-diagonal band terms, respectively. This yields the Hamiltonian H ξ B =(cid:18) h′ h∗ξ hξ h′ (cid:19) , where the diagonal elements read h′ = ω′ 3tnnna √2tlB a†a, and the off-diagonal ones are hξ = ξω′(cid:18)a† − ξ aw1 2√2lB a2 − a2w2 2 4l2 B a†2a(cid:19) . (106) (107) (108) Naturally, to lowest order in a/lB, one obtains again the Hamiltonian (81). The dimensionless parameters w1 and w2 are artificially added to the expressions and play the role of fitting parameters in the comparison with experi- mental measurements, as will be discussed below. They measure the deviation from the tight-binding-model ex- pectation, w1 = w2 = 1. Notice furthermore that, since we are interested in the n ≫ 1 limit, we do not care about corrections related to the ordering of the ladder operators, such that we identify a†2a2 ≃ a2a†2 ≃ (a†a)2 in the following parts. In the calculation of the LL spectrum, one may pro- ceed in the same manner as in Sec. II.A.2 – the eigen- value equation (84) for the second spinor component now becomes hξh†ξ vn ≃ (ǫn − h′)2 vn, (109) which is asymptotically correct in the large-n limit, where we may neglect the commutator [hξ, h′] on the right-hand side of the equation.13 The combination Hξ ≡ hξh†ξ is now interpreted as some fake Hamiltonian which needs to be diagonalized in order to obtain the modified LLs. Notice that n remains a good quantum number if one considers h′ on the right-hand side of the eigenvalue equa- tion. The left-hand side consists of a term H0 ≃ (ω′)2"a†a − 4w2 2 − w2 8 1 lB(cid:19)2 (cid:18) a (cid:0)a†a(cid:1)2# (110) that contains powers of a†a and thus respects the quan- tum number n, but in addition it contains the trigonal- warping term Ht.w. = −ξ w1(ω′)2a 2√2lB (a†3 + a3), (111) 18 which does not commute with a†a and which needs to be treated apart. If we neglect this trigonal-warping term for a moment, the LL energies are obtained from the quadratic equation (ω′)2(cid:20)n − 4w2 2 − w2 8 1 a2 l2 B n2(cid:21) ≃(cid:18)ǫn − ω′ 3tnnna √2tlB n(cid:19)2 . (112) In order to account for the trigonal-warping term in the eigenvalue equation (109), we may use a perturbative treatment, which is justified because of the small param- eter a/lB. There is no contribution at first order since hna(†)3ni = 0 due to the orthogonality of the eigenstates hnn′i = δn,n′. At second order, one obtains δn = − (ω′)2 lB(cid:19)2 8 (cid:18) a × 3n [1 + O(1/n)] , (113) which needs to be added to the right-hand side in Eq. (112). Interestingly, trigonal warping thus yields the same correction to the energies of the relativistic LLs as the third-order term in the expansion of the band disper- sion, although trigonal warping occurs at second order in the absence of a magnetic field, as we have discussed in Sec. I.C.2. This effect is due to the anisotropy of the band correction; in the presence of a magnetic field, the cos(3ϕq) term in Eq. (54) is averaged over the angle ϕq, and therefore only contributes at second order in the per- turbation theory described above. This eventually yields a correction of order (a/lB)2 to the LL energy, as does the third-order term in the correction of the band dispersion. One finally obtains, in the large-n limit, where these corrections become relevant, the energies of the relativis- tic LLs (Plochocka et al., 2008) ǫλn = vF lB 3tnnn t a lB n (114) 3w2 +λ vF lB √2n(1 − lB(cid:19)2 8 (cid:18) a n [1 + O(1/n)]) , independent of the valley index ξ, where O(1/n) stands for corrections of order 1/n. Notice that the fitting pa- rameters w1 and w2 cannot be determined independently from a fit to the spectroscopic measurement, but only the combination w2 ≡ (w2 2)/3. Equation (114) gener- alizes a calculation for the relativistic LLs when only nnn hopping is taken into account (Peres et al., 2006). 1 + 2w2 In Fig. (12), we show experimental results obtained from high-field transmission spectroscopy on multi-layer epitaxial graphene (Plochocka et al., 2008). Qualita- tively, one notices a downward renormalization of the transition energies ∆n = ǫλ=+,n − ǫλ=−,n (115) 13 The commutator would yield relative corrections that are on the order of 1/n and a/lB as compared to the energy scale (tnnn/t)(a/lB )n that dominates h′. in the interband regime for large values of n, in agree- ment with Eq. 114. Notice that because transmission spectroscopy is sensitive to energy-level differences, the nnn correction in Eq. (114) yields only a correction on the order of (tnnn/t)(a/lB)/n <∼ 1%/n at B ∼ 25 T as compared to the energy scale t(a/lB)n of the transition, whereas the other term yields a correction on the order of (a/lB)2×n ∼ 0.5%×n. The latter corrections thus be- come more relevant in the large-n limit than the nnn cor- rection. Indeed, the experiment (Plochocka et al., 2008) was not capable of probing the electron-hole symmetry breaking associated with the nnn term, whereas a quan- titative study of the high-energy transitions revealed a good semi-quantitative agreement with the calculated LL spectrum (114). However, it has been shown that the simple-minded tight-binding approach (with w = 1) un- derestimates the higher-order band corrections and that the best fit to Eq. (114) is obtained for a value of w = 2.8. The origin of this discrepancy is yet unexplained, and it may be interesting to perform high-field transmission spectroscopy measurement also on single-layer exfoliated graphene in order to understand whether the stronger downward renormalization of the LLs is due to interlayer couplings in the epitaxial multi-layer sample. C. Landau Level Spectrum in the Presence of an Inplane Electric Field A remarkable consequence of the relativistic character of electrons in graphene and the Lorentz invariance of the Dirac equation is their behavior in crossed magnetic and electric fields, where the magnetic field remains per- pendicular to the graphene sheet and the electric field is applied in the plane. Remember that in a non-relativistic 2D electron systems, the electric field E = Eey (in the y-direction) simply lifts the LL degeneracy and adds a term (E/B)k to the LL energies, where k is the wave vector in the x-direction. At a fixed wave vector k, the LL spacing is unaffected by the in-plane field. The situation is different for relativistic electrons in graphene, as a consequence of the Lorentz invariance of the Dirac equation. One may choose a reference frame in which the electric field vanishes as long as the drift velocity vD = E/B is smaller than the Fermi velocity, which plays the role of an upper bound for the physically significant velocities in the same manner as the speed of light in relativity (magnetic regime).14 In addition to the lifted LL degeneracy, the LL spacing is reduced (Lukose et al., 2007; Peres and Castro, 2007), as may be seen from a Lorentz boost into the reference frame which moves at the drift velocity and in which the electric field vanishes. In this reference frame, the magnetic field is reduced by the factor p1 − (E/vF B)2, (116) 14 In the opposite case, vD > vF , one may choose a reference system in which the magnetic field vanishes (electric regime) (Jackson, 1999). 19 such that the LLs (85), which scale as √B′ = √B[1 − (E/vF B)2]1/4 with the magnetic field, read ǫ′λ,n = λ vF lB [1 − (E/vF B)2]1/4√2n, (117) where the primes indicate the physical quantities in the moving frame of reference. When measuring the energy in the original (lab) frame of reference, the above energy spectrum also needs to be transformed into this frame of reference, which amounts to being multiplied by an- other factor (116), such that the spectrum of relativistic LLs in the presence of an in-plane electric field becomes (Lukose et al., 2007) ǫλ,n;k = λ vF lB [1 − (E/vF B)2]3/4√2n + E B k. (118) The quantum-mechanical derivation of this result will be discussed in more detail in Sec. II.D.1 in the context of the generalized Weyl Hamiltonian in a magnetic field. D. Landau Levels in Deformed Graphene As we have discussed in Sec. I.D, a uniaxial strain deforms the graphene sheet and modifies the electronic structure. The induced anisotropy of the Fermi velocity wx 6= wy is essentially washed out by the magnetic field, which yields an effective averaging over the Fermi surface, vF → v′F = √wxwy. More spectacular are the two fur- ther consequences of the deformation; (a) the tilt of the Dirac cones accounted for in the generalized Weyl Hamil- tonian (66) and (b) the topological phase transition due to the Dirac point motion. The implication for the LL spectrum are briefly reviewed in the following sections. 1. The generalized Weyl Hamiltonian in a magnetic field With the help of the Peierls substitution (72) and the expression of the kinetic momentum in terms of ladder operators (80), the generalized Weyl Hamiltonian (66) may be cast into the form H ξ B = ξ p2wxwy lB where (cid:18) w0 2 (aeiϕ + H.c.) a† w0 a 2 (aeiϕ + H.c.)(cid:19) . (119) w0eiϕ ≡ w0x wx + i w0y wy , (120) in terms of the effective tilt parameter (70) and the angle ϕ between the x-axis and the direction of the effective tilt (w0x/wx, w0y/wy), renormalized by the Fermi velocities wx and wy in the x- and y-direction, respectively. The Hamiltonian (119) may be solved quantum- mechanically in a straight-forward, but lengthy manner (Morinari et al., 2009; Peres and Castro, 2007). Instead, one may also obtain the result in a simpler semi-classical treatment (Goerbig et al., 2008), with the help of the Onsager relation (Lifshitz and Kosevich, 1956; Onsager, 1952) according to which the surface S(ǫ) enclosed by a trajectory of constant energy ǫ in reciprocal space is quantized as S(ǫ)l2 B = (2π)2Z ǫ 0 dǫ′ ρ(ǫ′) = 2π(n + γ), (121) where n is an integer denoting the energy level which co- incides with the Landau level in the full quantum treat- ment. The additional contribution γ is related to a Berry phase acquired by an electron during its cyclotron orbit. Usually, one has γ = 1/2 except if there is an extra Berry phase of π, which in our case yields γ = 0, as for graphene with no tilt (Mikitik and Sharlai, 1999). If one considers a density of states which scales as ρ(ǫ) ∝ ǫα, the energy levels thus scale as ǫn ∼ [B(n + γ)]1/(1+α), (122) in the large-n limit. 20 2. Tilted Dirac cones in a crossed magnetic and electric field One notices that the form (126) of LLs for tilted Dirac cones is the same as that of the LL spectrum (118) if one interprets the drift velocity vD = E/vF B as an ef- fective electric-field induced tilt. The magnetic regime E/B < vF corresponds then to the regime of closed or- bits ( w0 < 1) and the open hyperbolic orbits may be identified with the electric regime E/B > vF . Mathe- matically, the generalized Weyl Hamiltonian with an in- plane electric field may still be cast into the form (119) H ξ B → H ξ E/B = H ξ′B + E B k1, (127) where H ξ′B is the same as that of Eq. (119) if one replaces the tilt parameter w0 exp(iϕ) by (Goerbig et al., 2009) wξ(E)eiϕξ(E) ≡ wξx wx + i wξy wy . Here, the renormalized tilt velocity is given by wξ = (wξx, wξy) ≡ w0 − ξ E × B B2 , (128) (129) Because the density of states vanishes linearly at the Dirac point, as in the case of no tilt, i.e. α = 1, the scaling argument (122) yields the energy levels, and the angle ϕξ is the angle between this velocity and the x-axis. The resulting energy spectrum is given by ǫλ,n ≃ λ√2 v∗F lB √n, (123) as for unconstrained graphene, apart from a renormaliza- tion of the Fermi velocity. The latter is readily obtained from the calculation of the total number of states below the energy ǫ within the positive energy cone, 1 N+(ǫ) = (2π)22wxwy Zǫ+(q)≤ǫ ǫ2 2 (124) where we have defined qx/y ≡ wx/yqx/y, and the renor- malized Fermi velocity is 2π2v∗2 F dqxdqy = 1 , v∗2 F =hwxwy(1 − w2 0)3/2i , (125) in terms of the effective tilt parameter (70). This yields the result ǫλ,n = λ √wxwy lB 0)3/4√2n (1 − w2 (126) for the LL spectrum in the presence of a tilt, which coincides with the one obtained from the full quan- tum treatment (Morinari et al., 2009; Peres and Castro, 2007). One finally notices that the LL spacing becomes zero for w0 = 1, which corresponds to the condition (69) of maximal tilt for the Dirac cones, as discussed in Sec. I.D – indeed for values of w0 larger than 1, the isoener- getic lines are no longer closed elliptic orbits but open hyperbolas, for which the energy is not quantised. √wxwy lB k . ǫξ λ,n;k(E) = λ (cid:2)1 − wξ(E)2(cid:3)3/4 √2n + E B (130) Naturally one obtains the result (118) for undeformed graphene in an in-plane electric field, for wx = wy = vF and w0 = 0, as well as the LL spectrum (126) for the generalized Weyl Hamiltonian with tilted Dirac cones for zero in-plane field (E = 0). However, the most interest- ing situation arises when both the tilt and an in-plane field are present, in which case one observes a lifting of the valley degeneracy that is maximal when the electric field is applied perpendicular to the tilt velocity, E ⊥ w0 (Goerbig et al., 2009). Notice that, in order to obtain an effect on the or- der of ∼ 1%, extremely large electric fields would be required (on the order of 106 V/m) for a 10% deforma- tion of the lattice (Goerbig et al., 2009). It seems there- fore difficult to observe the effect in graphene, e.g. in high-field transmission spectroscopy or transport mea- surments, whereas the effect may be more visible in α−(BEDT-TTF)2I3, where the Dirac cones are naturally tilted (Katayama et al., 2006; Kobayashi et al., 2007) and where lower electric fields would be required for a comparable effect due to a roughly ten times smaller ef- fective Fermi velocity. III. ELECTRONIC INTERACTIONS IN GRAPHENE – INTEGER QUANTUM HALL REGIME In the preceding sections, we have discussed the elec- tronic properties of graphene within a one-particle model, i.e. we have neglected the Coulomb interaction between the electrons. In many materials, the one-particle picture yields the correct qualitative description of the electronic properties and is modified only quantitatively if one in- cludes the electron-electron interactions within perturba- tion theory (Giuliani and Vignale, 2005; Mahan, 1993). Notice, however, that there exists a class of materials – strongly correlated electron systems – the electronic properties of which may not be described correctly, not even on the qualitative level, within a one-particle pic- ture. In order to quantify the role of the electronic inter- actions, i.e. the correlations, in graphene, one needs to compare the characteristic Coulomb energy Eint = e2/εℓ at the average inter-electronic distance (ε is the dielectric constant describing the environment the 2D electron gas is embedded in) to the kinetic one Ekin(kF ) at the same length scale, given in terms of the Fermi wave vector kF , ℓ ∼ k−1 F , rs = Eint Ekin . (131) If this dimensionless interaction parameter becomes very large, rs ≫ 1, the electrons are strongly correlated. In non-relativistic 2D metals with a parabolic band dis- persion, Ekin ∼ 2k2 F /mb, the dimensionless parameter reads rs = mbe2 2ε ℓ ∼ 1 a∗0kF , (132) in terms of the effective Bohr radius a∗0 = a0εm0/mb, where a0 = 0.5 A is the Bohr radius in vacuum and mb/m0 the ratio between the band and the bare electron mass. The relevance of electronic correlations therefore increases in the dilute limit when ℓ ≫ a∗0. Notice that the parameter rs, which is also called Wigner-Seitz radius, plays the role of a length measured in units of the effective Bohr radius a∗0. The same argument applied to graphene yields a com- pletely different result. Whereas the scaling of the Coulomb energy remains the same, that of the kinetic energy is changed due to the linearity of the band dis- persion. As a consequence the dimensionless interaction parameter in graphene reads 21 expression (133) is the same as that of the fine structure constant α = e2/εc = 1/137 in quantum electrodynam- ics (Weinberg, 1995) if one replaces the Fermi velocity by the velocity of light, which is roughly 300 times larger. One therefore calls αG alternatively the graphene fine structure constant. Long-range versus short-range interactions. Another im- portant aspect of interacting electrons is the range of the interaction potential. Whereas the underlying Coulomb potential e2/εr is long-range, short-range interaction models such as the Hubbard model are often – and suc- cessfully – used in the description of correlated met- als. The use of such a short-range interaction may be justified by the screening properties of interacting elec- trons, which are correctly captured in a Thomas-Fermi approach (Giuliani and Vignale, 2005; Mahan, 1993) ac- cording to which the Coulomb interaction potential is screened above a characteristic screening length λT F ∼ 1/kT F .16 In 2D, the Thomas-Fermi wave vector kT F ≃ rskF (134) is given in terms of the dimensionless interaction param- eter (131) and the Fermi wave vector kF .17 One notices that, for metals with a parabolic dis- persion relation, the Thomas-Fermi wave vector is sim- ply given in terms of the inverse effective Bohr radius, kT F ∼ 1/a∗0, independent of the electronic density. Un- less the band mass is very small as compared to the bare electron mass or the dielectric constant of the host ma- terial very large, the Coulomb interaction is therefore screened on the atomic scale. A description of such sys- tems in the framework of short-range interaction mod- els, such as the Hubbard model, then becomes better justified than in systems with a small band mass or a prominent dielectric constant (such as in 2D electron systems in GaAs heterostructures). Typical examples, where short-range interaction model yields valuable phys- ical insight, are heavy-fermion compounds [for a review see Ref. (Coleman, 2003)]. The situation is again drastically different in graphene where the Thomas-Fermi wave vector (134) becomes kG T F ≃ αGkF ≃ 2.2 ε kF ∼ √nel, (135) αG = Eint Ekin = e2 εvF ≃ 2.2 ε , (133) i.e. it vanishes at the Dirac points where the carrier density goes to zero, and the screening length then di- verges.18 Notice that even for doped graphene, where independent of the carrier density.15 The correlations are therefore in an intermediate regime but may be de- creased if the graphene sheet is embedded in an environ- ment with a large dielectric constant. Notice that the 15 In contrast to an electron system with a parabolic band disper- sion, this parameter can no longer be interpreted as a dimension- less radius, and we therefore use the notation αG rather than rs. 16 Notice that the Thomas-Fermi approach is restricted to static screening effects, whereas dynamic screening require a more com- plex treatment, e.g. in the framework of the random-phase ap- proximation. 17 In three space dimensions, the relation reads k2 18 Due to this divergence of the screening length, one principally needs to describe screening beyond the level of linear-response theory (Katsnelson, 2006). T F ≃ rsk2 F . one may typically induce carrier densities on the order of 1012 cm−2, the screening length is λT F >∼ 10 nm, i.e. much larger than the lattice scale. One thus comes to the conclusion that the rele- vant electronic interactions in graphene are long-range Coulomb interactions that may not be captured, in con- trast to other materials with a parabolic band dispersion, within models such as the Hubbard model (Herbut, 2006, 2007a). We therefore investigate, in this section, the fate of the long-range Coulomb interaction in a strong mag- netic field. In Sec. III.A, we decompose the Coulomb in- teraction Hamiltonian in the two-spinor basis of the low- energy electronic wave functions in graphene and com- ment on its symmetry with respect to the valley pseu- dospin. The role of these interactions in the particle-hole III.B, where we excitation spectrum is studied in Sec. discuss the resulting collective excitations in the IQHE regime, which allows for a perturbative treatment. The strong-correlation regime of partially filled LLs (regime of the fractional quantum Hall effect) is presented sepa- rately in Sec. V. A. Decomposition of the Coulomb interaction in the Two-Spinor Basis Generally, the Coulomb interaction for 2D electrons may be accounted for by the Hamiltonian Hint = 1 2Xq v(q)ρ(−q)ρ(q), (136) of the ρ(q) in terms components the Fourier = R d2r exp(−iq · r)ψ†(r)ψ(r) of electronic den- sity ψ†(r)ψ(r) and the 2D Fourier transformed 1/r Coulomb potential, v(q) = 2πe2/εq. If one takes into account the electronic spin σ =↑,↓, the Coulomb interaction respects the associated SU(2) symmetry, and the Fourier components are then simply the sum of the densities ρσ(q) in both spin orientations, 22 ρ(q) = ρ↑(q) + ρ↓(q). For notational convenience, we neglect the spin index in the following discussion keeping in mind that the spin SU(2) symmetry is respected. The density operators may be decomposed in the basis of the spinor wave functions (96) and (97) for relativistic electrons in graphene, ρ(q) = Xλn,m;ξ λ′n′,m′;ξ′ ψ†λn,m;ξe−iq·rψλ′n′,m′;ξ′ c†λn,m;ξcλ′n′,m′;ξ′ , (137) where c(†) λn,m;ξ are fermion operators in second quantiza- tion that annihilate (create) an electron in the quantum state ψλn,m;ξ=+ = (cid:18) 1∗nn − 1, mi and ψλn,m;ξ=− = (cid:18) −λ2∗nn, mi λ2∗nn, mi (cid:19) eiK·r 1∗nn − 1, mi(cid:19) e−iK·r. (138) In order to avoid confusion in the case of inter-valley coupling, we use now a representation in which the first spinor component represents the amplitude on the A sublattice and the second on the B sublattice for both valleys. Contrary to the expressions (96) and (97), the state (138) is valid for both n = 0 and n 6= 0 by us- ing the short-hand notation 1∗n ≡ p(1 − δn,0)/2 and 2∗n ≡ p(1 + δn,0)/2. Furthermore, we have explicitly taken into account the rapidly oscillating part exp(iξK·r) due to the two different valleys, whereas the expressions (96) and (97) are only concerned with the slowly varying envelope function. Explicitly, the Fourier components of the density operator (137) then read ρ(q) = Xλn,λ′n′ ξ,ξ′ F ξ,ξ′ λn,λ′n′(q)¯ρξ,ξ′ λn,λ′n′ (q), (139) in terms of the reduced density operators ¯ρξ,ξ′ λn,λ′n′ (q) = Xm,m′Dm(cid:12)(cid:12)(cid:12) e−i[q+(ξ−ξ′)K]·R(cid:12)(cid:12)(cid:12) m′E c†λn,m;ξcλ′n′,m′;ξ′ , (140) which may also be interpreted as magneto-exciton operators associated with a particular inter-LL transition (see Sec. IV.A), and the form factors Fλn,λ′n′ (q) ≡ F ξ,ξ for intra-valley and F +,−λn,λ′n′ (q) = λ1∗n′ 2∗nDn(cid:12)(cid:12)(cid:12) λn,λ′n′ (q) = 1∗n1∗n′(cid:10)n − 1(cid:12)(cid:12)e−iq·η(cid:12)(cid:12) n′ − 1(cid:11) + λλ′2∗n2∗n′(cid:10)n(cid:12)(cid:12)e−iq·η(cid:12)(cid:12) n′(cid:11) e−i(q+2K)·η(cid:12)(cid:12)(cid:12) n′E =hF−,+ n′ − 1E − λ′1∗n2∗n′Dn − 1(cid:12)(cid:12)(cid:12) e−i(q+2K)·η(cid:12)(cid:12)(cid:12) λ′n′,λn(−q)i∗ (141) (142) for inter-valley processes. Here, we have used the decomposition r = R + η of the position operator into its guiding center and cyclotron coordinate (see Sec. II.A.1) and the fact that f1(η)f2(R)n, mi = f1(η)ni ⊗ f2(R)mi, for two arbitrary functions f1 and f2. The full expressions for the matrix elements in Eqs. (140), (141), and (142) may be found in Appendix A. In terms of the reduced density operators (140), the interaction Hamiltonian (136) reads 23 Hint = 1 2Xq Xλ1n1...λ4n4 ξ1...ξ4 vξ1...ξ4 λ1n1...λ4n4 (q)¯ρξ1,ξ3 λ1n1,λ3n3 (−q)¯ρξ2,ξ4 λ2n2,λ4n4 (q), (143) (a) ν , ξ 3 1 ν , −ξ 4 2 (b) ν , −ξ 3 ν , −ξ 4 ν , ξ 1 1 ν , ξ 2 2 ν , ξ 1 (c) ν , −ξ 3 ν , ξ 4 (d) ν , ξ 3 1 ν , ξ 1 ν , −ξ 2 ν , ξ 1 1 ν , ξ 2 ν , ξ 4 2 ν , ξ 2 2 Figure 13 Diagrammatic representation of the interaction ver- tex (we use the short-hand notation νi = (λini, mi) for the quantum numbers); (a) vertex associated with terms of the form vξ,ξ,ξ′ ,−ξ′ λ1n1...λ4n4 vξ,ξ,−ξ,−ξ (q) λ1n1...λ4n4 and (d) vertex respecting the SU(2) valley-pseudospin symmetry vξ,−ξ,ξ′ ,−ξ′ λ1n1...λ4n4 (q) or vξ,−ξ,ξ′ ,ξ′ (q), (c) vertex of backscattering type, vξ,−ξ,−ξ,ξ (q), (b) vertex of Umklapp type, λ1n1...λ4n4 λ1n1...λ4n4 (q). where the interaction vertex is defined as vξ1...ξ4 λ1n1...λ4n4 (q) = 2πe2 εq F ξ1,ξ3 λ1n1,λ3n3 (−q)F ξ2,ξ4 λ2n2,λ4n4 (q). (144) 1. SU(2) valley symmetry One notices that, in contrast to the SU(2) symmetry associated with the physical spin, the Hamiltonian (143) does not respect a similar valley-pseudospin symmetry due to possible inter-valley couplings. An SU(2) valley- pseudospin symmetry would be respected for the case ξ1 = ξ3 and ξ2 = ξ4, i.e. if the interaction vertex (144) vξ1...ξ4 λ1n1...λ4n4 (q) ∝ δξ1,ξ3δξ2,ξ4. (145) One may show, however, the SU(2) valley- pseudospin symmetry is approximately respected when considering the different classes of interaction vertices de- picted in Fig. 13. that • Consider the diagram in Fig. 13(a), which rep- (q) or resents a vertex of the type vξ,ξ,ξ′,−ξ′ vξ,−ξ,ξ′,ξ′ In this case, the particle on the λ1n1...λ4n4 left remains in the same valley whereas that on the λ1n1...λ4n4 (q). right changes the valley. Such a process would re- quire a momentum transfer of ±K, i.e. of the wave vector connecting the two valleys, and therefore does not respect momentum conservation, in the absence of a magnetic field. Naturally, momentum is not a good quantum number here due to to the magnetic field, but momentum conservation man- ifests itself by an exponential suppression of such processes. In order to appreciate this point, we need to consider the Gaussian in the form factors (141) and (142), F ξ,ξ′ λn,λ′n′ (q) ∝ e−q+(ξ−ξ′)K2l2 B /4, (146) as discussed in Appendix A [see Eq. (A2)]. One therefore sees that the interaction vertex contains a Gaussian term vξ,ξ,ξ′,−ξ′ λ1n1...λ4n4 (q) ∝ e−(q2+q±K2)l2 B /4 ∼ e−(q′2+K2/4)l2 ∼ e−K2l2 B /8 ∼ e−#l2 B /2 B /a2 (147) , where # represents an unimportant numerical fac- tor and where we have shifted the momentum q′ = q ± K/2 in the second step. The processes associated with the diagram in Fig. 13(a) are thus B/a2 ≃ 104/B[T] and exponentially suppressed in l2 may safely be neglected in the range of physically accessible magnetic fields. • The same fate is reserved for the diagram in Fig. 13(b), which represents a process of Umklapp type. In this case, the vertex reads vξ,ξ,−ξ,−ξ λ1n1...λ4n4 (q) ∝ e−(q+K2+q−K2)l2 B /2 ∼ e−#l2 which is again exponentially small in l2 ∼ e−K2l2 B /4 B/a2. (148) , B /a2 • The situation is different for backscattering-type di- agrams [Fig. 13(c)], in which case the interaction vertex is vξ,−ξ,−ξ,ξ λ1n1...λ4n4 (q) ∝ e−(q±K2+q±K2)l2 B /4. (149) One may then redefine the wave vector q′ = q± K, which is eventually an integration variable in the interaction Hamiltonian (143), and the interaction vertex becomes vξ,−ξ,−ξ,ξ λ1n1...λ4n4 (q′) ∝ e−q′ 2l2 B /2 (150) 2πe2 εq′ ∓ K ∼ 2πe2 εK e−q′2l2 B /2. compared to the leading energy scale e2/εlB. As an order of magnitude, with K ∼ 1/a, one then notices that the backscattering interaction vertex is suppressed by a factor of a/lB ∼ 0.005×pB[T] as • The leading interaction vertex is therefore the SU(2) valley-pseudospin symmetric one depicted in Fig. 13(d), for which the rapidly oscillating contri- bution at K vanishes, as may be seen directly from the form factors (142). 24 The above argument, which generalizes symmetry con- siderations for the interactions in a single relativistic LL (Alicea and Fisher, 2006; Doretto and Morais Smith, 2007; Goerbig et al., 2006; Herbut, 2007b), shows that although the valley SU(2) symmetry is not an exact sym- metry, such as the SU(2) symmetry associated with the physical spin, it is approximately respected by the long- range Coulomb interaction. Valley-symmetry breaking terms are due to lattice effects beyond the continuum limit and therefore suppressed by the small factor a/lB, which quantifies precisely corrections due to effects on the lattice scale. If one takes into account the additional spin degree of freedom, the resulting four-fold spin-valley de- generacy may then be described within the larger SU(4) symmetry group, which turns out to be relevant in the description of strong-correlation effects in partially filled LLs (Sec. V). 2. SU(4) spin-valley symmetric Hamiltonian The SU(4)-symmetric part of the interaction Hamiltonian (143) finally reads H sym int = 1 2Xq Xλ1n1...λ4n4 vsym λ1n1...λ4n4 (q)¯ρλ1n1,λ3n3(−q)¯ρλ2n1,λ4n4 (q), where the symmetric interaction vertex is vsym λ1n1...λ4n4 (q) = 2πe2 εq Fλ1n1,λ3n3 (−q)Fλ2n2,λ4n4(q), in terms of the reduced density operators ¯ρλn,λ′n′ (q) ≡ Xξ=± ρξ,ξ λn,λ′n′ (q) = Xξ=± Xσ=↑,↓ Xm,m′(cid:10)m(cid:12)(cid:12)e−iq·R(cid:12)(cid:12) m′(cid:11) c†λn,m;ξ,σcλ′n′,m′;ξ,σ, (151) (152) (153) where we have explicitly taken into account the spin in- dex σ =↑,↓ in the last line. may also be rewritten in terms of the LL form factors We finally notice that the graphene form factors (141) In the remainder of this section, we neglect the symmetry-breaking part of the Hamiltonian and consider the Coulomb interaction to respect the SU(2) valley sym- metry. (154) Fn,n′(q) =(cid:10)n(cid:12)(cid:12)e−iq·η(cid:12)(cid:12) n′(cid:11) , which arise in a similar decomposition of the Coulomb interaction in Landau states in the non-relativistic 2D electron gas, as Fλn,λ′n′ (q) = 1∗n1∗n′ Fn−1,n′−1(q) + λλ′2∗n2∗n′Fn,n′ (q). (155) To summarize the differences and the similarities between the interaction Hamiltonians in graphene and the non- relativistic 2D electron system, one first realizes that its structure is the same if one replaces the LL form factor (154) by the graphene form factors (141) and if one takes into account the larger (approximate) internal symmetry SU(4), due to the spin-valley degeneracy, instead of the spin SU(2) symmetry. B. Particle-Hole Excitation Spectrum The considerations of the previous subsection allow us to discuss the role of the Coulomb interaction within a perturbative approach in the IQHE regime for ν = ±2(2n + 1), where the (non-interacting) ground state is non-degenerate and separated by the cyclotron gap √2(vF /lB)(√n + 1−√n) from its excited states. Quite generally, the inter-LL transitions evolve into coher- ent collective excitations, as a consequence of these Coulomb interactions. Prominent examples in the non-relativistic 2D electron gas are the upper-hybrid mode (sometimes also called magneto-plasmon), which is the magnetic-field counterpart of the usual 2D plas- (a) (b) CB Ia Ib E F IIa −q F IIb q F q VB Figure 14 Zero-field particle-hole exctiation spectrum for doped graphene. (a) Possible intraband (I) and interband (II) single-pair excitations in doped graphene. The exctiations close to the Fermi energy may have a wave-vector transfer comprised between q = 0 (Ia) and q = 2qF (Ib), in terms of the Fermi wave vector qF . (b) Spectral function Im Π0(q, ω) in the wave-vector/energy plane. The regions corresponding to intra- and interband excitations are denoted by (I) and (II), respectively. mon (Giuliani and Vignale, 2005), and magneto-excitons (Kallin and Halperin, 1984). In the present subsection, we discuss how these modes manifest themselves in graphene in comparison with the non-relativistic 2D elec- tron gas. 1. Graphene particle-hole excitation spectrum at B = 0 Before discussing the particle-hole excitation spectrum (PHES) for graphene in the IQHE regime, we briefly re- view the one for B = 0 as well as its associated collec- tive modes (Ando, 2006a; Hwang and Das Sarma, 2007; Shung, 1986; Wunsch et al., 2006). Quite generally, the PHES is determined by the spectral function S(q, ω) = − 1 π Im Π(q, ω), (156) which may be viewed as the spectral weight of the al- lowed particle-hole excitations, in terms of the polariz- ability Π(q, ω), which plays the role of a density-density response function (Giuliani and Vignale, 2005; Mahan, 1993). The particle-hole excitations for non-interacting elec- trons in doped graphene are depicted in Fig. 14.19 In contrast to the PHES of electrons in a single parabolic band (the non-relativistic 2D electron gas), there are two different types of excitations: intraband excitations [la- beled by I in Fig. 14(a)], where both the electron and the hole reside in the conduction band (CB), and interband excitations [labeled by II in Fig. 14(a)], where an electron is promoted from the valence band (VB) to the CB. In un- doped graphene, there exist naturally only interband ex- citations (II). If the electron and the hole have an energy 19 We consider here only the case of a Fermi energy ǫF in the con- duction band, for simplicity. 25 close to the Fermi energy, the allowed excitations imply a wave-vector transfer that lies in between q = 0 (Ia) and q = 2qF (Ib). At non-zero values ǫ of the transfered en- ergy, one needs to search for available quantum states at larger wave vectors, and the particle-hole pair wave vec- tor is then restricted to ǫ/vF < q < 2qF + ǫ/vF , as a consequence of the linear dispersion relation in graphene. This gives rise to the region I, which describes the intra- band particle-hole continuum, and its linear boundaries in the PHES described by the spectral function in Fig. 14(b). In addition to intraband excitations, one notices that interband excitations become possible above a threshold energy of ǫF , where an electron at the top of the VB (at q = 0) may be promoted to an empty state slightly above the Fermi energy. The associated wave-vector transfer is naturally q = qF . The point (qF , ǫF ) marks the bottom of the region II in Fig. 14(b), which determines the region of allowed interband excitations (interband particle-hole continuum). Direct interband excitations with zero wave- vector transfer are possible above an energy of 2ǫF . Indeed, (50) indicates that the chirality, i.e. Another aspect of the PHES in Fig. 14 is the strong concentration of spectral weight around the central diag- onal ω = vFq. This concentration is a particularity of graphene due to the electrons’ chirality (Polini et al., 2008). if one considers a 2qF backscattering process in the vicinity of the Fermi energy in the CB, Eq. the projec- tion of the sublattice-pseudospin on the direction of the wave vector, is preserved. The inversion of the direc- tion of propagation in the 2qF process would therefore require an inversion of the A and B sublattices that is not supported by most of the scattering or interaction processes. This effect is reflected by a strong suppres- sion of the spectral weight when approaching the right boundary of the region I in the PHES associated with processes of the type Ib in Fig. 14(a). Similarly, the conservation of the electrons’ chirality (50) favors 2qF processes in the interband region (II) and the suppres- sion of direct q = 0 interband excitations of the type IIa in Fig. 14(a). Notice that, although the direction of the wave vector is inverted in a 2qF process, this indicates still the absence of backscattering because the group ve- locity v = ∇qǫλ q/ = λvF q/q remains unchanged – the change in the sign due to the inversion of the wave vector is indeed canceled by the one associated with the change of the band index. In order a. Formal calculation of the spectral function. it is apparent from to obtain the spectral function, Eq. (156) that one needs to calculate the polarizabil- ity Π(q, ω) of the 2D system, which may be found with the help of the Green’s functions G(q, ω), Π(q, ω) = −iTrZ dω′ 2π Xq′ G(q′, ω′)G(q + q′, ω + ω′), (157) (a) (b) q+q’,ω+ω’ q’, ω’ II I Figure 15 (a) Particle-hole bubble diagram (polarizability), in terms of Green’s functions G(q, ω) (lines). (b) Spectral func- tion Im ΠRP A(q, ω) for doped graphene in the wave-vector/energy plane. The electron-electron interactions are taken into account within the RPA. We have chosen αG = 1 here. where Tr means the trace since the Greens functions are 2 × 2 matrices as a consequence of the matrix char- acter of the kinetic Hamiltonian. Diagrammatically, the polarizability may be represented by the so-called 15(a), and one finds bubble diagram shown in Fig. for non-interacting electrons in graphene (Ando, 2006a; Hwang and Das Sarma, 2007; Polini et al., 2008; Shung, 1986; Wunsch et al., 2006) Π0(q, ω) = g A Xq′,λ,λ′ q′(cid:1) − n(cid:16)ǫλ′ n(cid:0)ǫλ q′ − ǫλ′ ǫλ q′+q(cid:17) q′+q + ω + iδ Cλλ′ (q′, q′+q), (158) where ǫλ q = λvFq − ǫF is the energy of the quantum state ψλ(q) measured from the Fermi energy ǫF , g = 4 takes into account the four-fold spin-valley degeneracy, and n(ǫλ q) is the Fermi-Dirac distribution function that reduces to a Heavyside step function n(ǫλ q) = Θ(−ǫλ q) at zero temperature. Equation (158) is nothing other than the Lindhard function (Giuliani and Vignale, 2005; Mahan, 1993), apart from the factor Cλλ′ (q′, q′ + q) ≡ 1 + λλ′ cos θq′,q′+q 2 , (159) in terms of the angle θq′,q′+q between q′ and q′ + q, which takes into account the particular chirality prop- erties of graphene – as already mentioned above, this chirality factor vanishes for backscattering processes, i.e. for intraband (λ = λ′) 2qF processes with θq′,q′+q = π as well as for interband (λ = −λ′) q = 0 processes with θq′,q′+q = 0 or 2π. Notice finally that the quantity δ in Eq. (158) is an infinitesimal energy in the case of pure graphene and may be used (for finite values) as a phenomenological measure of the impurity broadening δ ≃ /τ , in terms of a life time τ of the excitations. 26 level of the random-phase approximation (RPA), which amounts to calculating a geometric series of bubble diagrams and which has shown to yield reliable re- sults for doped graphene (Hwang and Das Sarma, 2007; Sabio et al., 2008; Wunsch et al., 2006). The RPA has also been applied to undoped graphene (Gonz´alez et al., 1994, 1999), but its validity has been questioned (Gangadharaiah et al., 2008; Kotov et al., 2007) because of the vanishing density of states, which would require to take into account diagrams beyond the RPA (Katsnelson, 2006). The RPA polarizability then becomes ΠRP A(q, ω) = Π0(q, ω) εRP A(q, ω) , (160) in terms of the polarizability (158) for non-interacting electrons and the dielectric function εRP A(q, ω) = 1 − 2πe2 εq Π0(q, ω). (161) The spectral function associated with the RPA polar- izability (160), which is shown in Fig. 15(b), reveals the characteristic coherent 2D plasmon mode, which corresponds to the solution of the implicit equation εRP A(q, ωpl) = 0 and the dispersion relation of which reads ωpl(q) ≃r 2e2ǫF 2ε q (162) in the small-q limit (Shung, 1986; Wunsch et al., 2006). Interestingly, this equation is valid also for non- relativistic electrons in conventional 2D electron sys- tems (Stern, 1967) if one takes into account the dif- ference in the density dependence of the Fermi energy (ǫF = πnel/mb for non-relativistic 2D electrons and ǫF = vF√πnel in graphene) as well as that in the Fermi velocity (vF = p2ǫF /mb for non-relativistic electrons, as compared to a constant vF in graphene). Notice that the dispersion relation is restricted to small values of q (as compared to the Fermi wave vector kF ), whereas the numerical solution presented in Fig. 15 indicates that the asymptotic dependence of the plasmon mode is in- deed given by the central diagonal ωuh(q) >∼ vF q (Shung, 1986; Wunsch et al., 2006). Therefore, contrary to the plasmon in 2D metals with a parabolic dispersion rela- tion, the plasmon in graphene does not enter region I, but only the interband particle-hole continuum (region II). In this region, the Landau damping is less efficient, and the coherence of the mode thus survives to a certain extent without decaying into incoherent particle-hole ex- citations. b. Polarizability in the random-phase approximation. The diagrammatic approach is particularly adapted for tak- ing into account the electronic interactions on the In the case of a strong magnetic field applied perpen- dicular to the graphene sheet, one needs to take into ac- 2. Polarizability for B 6= 0 count the quantization of the kinetic energy into rela- tivistic LLs described in Sec. II.A.1, as well as the spino- rial eigenfunctions ψξ nλ,m in the calculation of the po- larizability. One finds a similar expression for the zero- temperature polarizability of non-interacting electrons as in Eq. (158), 27 Π0 B(q, ω) = g Xλn,λ′n′ Θ(ǫF − λω′√n) − Θ(ǫF − λ′ω′√n′) λω′√n − λ′ω′√n′ + ω + iδ Fλn,λ′n′ (q)2, (163) in terms of the graphene form factors (141) and the char- acteristic frequency ω′ = √2vF /lB introduced in Sec. II.A.1. One notices that the first part is nothing other than a Lindhard function (Giuliani and Vignale, 2005; Mahan, 1993) for relativistic LLs filled up to the Fermi energy ǫF = (vF /lB)√2NF , which is chosen to be situ- ated between a completely filled (NF ) and a completely empty (NF +1) LL in the CB (IQHE regime). The second factor is the modulus square of the graphene form factors which plays the role of the chirality factor Cλ,λ′ (q′, q′+q) in the absence of a magnetic field (Rold´an et al., 2009, 2010; Shizuya, 2007).20 As for the zero-field case, one may distinguish two con- tributions to the polarizability, one that may be viewed as a vacuum polarizability Πvac(q, ω) and that stems from interband excitations when the Fermi level is at the Dirac point, and a second one that comes from intraband exci- tations in the case of doped graphene. Because undoped graphene with zero carrier density does not correspond to an IQHE situation – as we have already discussed in II.A, the zero-energy LL n = 0 is only half-filled Sec. then –, we define, here, the vacuum polarizability with respect to the completely filled zero-energy level. In order to describe more explicitly the different con- tributions to the polarizability, we define the auxiliary quantities (Rold´an et al., 2009) Πλn,λ′n′(q, ω) = Fλn,λn′ (q)2 λvF√n − λ′vF√n′ + ω + iδ + (ω+ → −ω−) (164) where ω+ → ω− indicates the replacement ω + iδ → −ω − iδ and n−1 Πλn(q, ω) = Xλ′ Xn′=0 Xn′=n+1 +Xλ′ Nc Πλn,λ′n′ (q, ω) (165) Πλn,λ′n′ (q, ω) + Πλn,−λn(q, ω) 20 A similar expression for the polarizability has also been obtained in Refs. (Berman et al., 2008; Tahir and Sabeeh, 2008) though with approximate form factors. (a) II (b) I Figure 16 Particle-hole excitation spectrum for graphene in a per- pendicular magnetic field. We have chosen NF = 3 in the CB and a LL broadening of δ = 0.05vF /lB (a) and δ = 0.2vF /lB (b). The ultaviolet cutoff is chosen such that Nc = 70. which verify Πλn(q, ω) = −Π−λn(q, ω). The vacuum po- larization may then be defined as Πvac(q, ω) = − Nc Xn=1 Π+n(q, ω) (166) where Nc is a cutoff that delimits the validity of the continuum approximation. Notice that, already in the absence of a magnetic field, the validity of the contin- uum approximation is delimited by a maximal energy ∼ t. One may then introduce an upper level cutoff with the help of ǫNc = (vF /lB)√2Nc ∼ t, that leads to Nc ∼ 104/B[T ], which is a rather high value even for strong magnetic fields. However, due to the fact that the separation between LLs in graphene decreases with n, it is always possible to obtain reliable semi-quantitative results from smaller values of Nc. The spectral function S(q, ω) = −Im Π0 B(q, ω)/π is shown in Fig. 16 for NF = 3 and for two different values of the phenomenologically introduced LL broadening δ. One notices first that the spectral weight is restricted to the two regions I and II corresponding to the intraband and interband particle-hole continuum, respectively, in the zero-field limit. This is not astonishing because the electron-hole pair wave vector remains a good quantum number also in the presence of a magnetic field and be- cause the overlap between the electron and hole wave functions is largest in these regions; if one considers the pair with its overall charge neutrality, its motion is unaf- fected by the magnetic field. Indeed, the pair momentum may be viewed as the sum of the pseudomomenta asso- ciated with the guiding-center variable for the electron, R× ez/l2 B, respectively. Each of the pseudomomenta is naturally a constant of motion because so is the guiding center, as we have discussed in Sec. II.A.1. One therefore obtains the relation B, and the hole −R′ × ez/l2 q = ∆R × ez/l2 B or ∆R = ql2 B, (167) where ∆R = R − R′ is the distance between the guiding center of the electron and that of the hole. The bound- aries of the PHES in Fig. 16 may then be obtained from the decomposition (90), which yields η′ − η ≤ ∆R ≤ η′ + η, with the help of the average values η ≡ hηi = lB√2n + 1 and η′ = lB√2n′ + 1, √2n′ + 1−√2n + 1 ≤ qlB ≤ √2n′ + 1+√2n + 1. (168) Because the energy scales also with √n, one obtains the linear boundaries of the particle-hole continua as in the zero-field case mentioned above. In contrast to these similarities with the zero-field PHES, one notices a structure in the spectral weight that is due to the strong magnetic field. As a con- sequence of the relativistic LL quantization, the spec- tral weight corresponds to inter-LL transitions at ener- gies ω = √2(vF /lB)(√n − λ√n′), where n > NF and n′ ≤ NF (for λ = +) or n′ > 0 (for λ = −). For larger values of NF , or quite generally when increasing the en- ergy, the level density increases due to the √n scaling of the LLs and the transitions. The LL structure is there- fore only visible in the lower part of PHES, in the clean limit δ = 0.05vF /lB [Fig. 16(a)], whereas the inter-LL transitions are blurred at larger energies or even for the lower transitions in the case of less clean samples [Fig. 16(b), for δ = 0.2vF /lB].21 In addition to the (blurred) LL structure in the PHES, one notices another structure of the spectral weight, which is organised in lines parallel to the central diago- nal ω = vFq. This weight is again decreased when ap- proaching the right boundary of the intraband continuum (region I) and the left one of the interband continuum (region II), due to the above-mentioned chirality proper- ties of electrons in graphene. The emergence of diagonal lines is a consequence of the graphene form factors (141) the modulus square of which intervenes in the polariza- tion function. Indeed, these form factors Fλ(n+m),λ′n(q) are (associated) Laguerre polynomials with n + 1 zeros (Gradshteyn and Ryzhik, 2000) due to the overlap be- tween the wave function of the hole in the level λ′n and that of the electron in the LL λ(n + m) (Rold´an et al., 2010). These zeros in the inter-LL transitions are organ- ised in lines that disperse parallel to the central diagonal and thus give rise to zones of vanishing spectral weight. 21 The value δ = 0.2vF /lB is a reasonable estimate for today’s exfoliated graphene samples on an SiO2 substrate (Ando, 2007a). 28 (a) II (b) II I I Figure 17 Particle-hole excitation spectrum for graphene in a per- pendicular magnetic field. The Coulomb interaction is taken into account within the RPA. We have chosen NF = 3 in the CB and a LL broadening of δ = 0.05vF /lB (a) and δ = 0.2vF /lB (b). The ultaviolett cutoff is chosen such that Nc = 70. Interestingly, it is this structure of diagonal lines that survives in more disordered samples in which the hori- zontal lines associated with inter-LL transitions start to overlap, i.e. once the LL spacing is smaller than the level broadening δ.22 3. Electron-electron interactions in the random-phase approximation: upper-hybrid mode and linear magnetoplasmons The PHES of non-interacting electrons in graphene gives already insight into the collective modes which one may expect once electron-electron interactions are taken into account. Indeed the regions of large spec- tral weight evolve into coherent collective excitations as a consequence of these interactions. Because the re- gions of large spectral weight are organised in lines par- allel to the central diagonal ω = vFq, as mentioned above, one may expect that the dominant collective ex- citations are roughly linearly dispersing modes instead of the more conventional weakly dispersing magneto- excitons, which emerge from the inter-LL transitions (Kallin and Halperin, 1984). It has though been argued that such magneto-excitons may play a role at low en- ergies in clean samples with low doping (Iyengar et al., 2007) and that they may renormalize the cyclotron en- ergy at zero wave vector (Bychkov and Martinez, 2008). As in the zero-field case, we take into account the Coulomb interaction within the RPA [see Eq. (160)]. The resulting spectral function is shown in Fig. 17 for the same choice of parameters as in the non-interacting case (Fig. 16). Furthermore, we have chosen a dimensionless interaction parameter αG = 1, here, which corresponds 22 This behavior is in stark contrast to that of non-relativistic 2D electrons, where the LL spacing is constant and given by the cyclotron energy eB/mb. If this quantity is larger than the level broadening δ, there is no qualitative difference between low and high energies, and the horizontal lines associated with the inter-LL excitations (multiples of the cyclotron energy) remain well separated. (a) (b) 29 to a dielectric constant of ε ≃ 2. One notices the prominent mode that evolves in the originally forbidden region for particle-hole excitations. This mode, which is called upper-hybrid mode, is the magnetic-field descendent of the 2D plasmon mode (162) and acquires a density-dependent cyclotron gap ωC = F /ǫF = eBvF /vF√πnel. Its dispersion relation in eBv2 the small-q limit is then given by ωuh(q) ≃ qω2 pl(q) + ω2 C ≃ s 2e2vF√πnel 2ε (169) , Static polarization function Π0(q, ω = 0) for non- Figure 18 interacting electrons in graphene without (a) and with (b) a mag- netic field. To compare both cases, we have chosen a Fermi wave vector qF = √2NF + 1/lB = √7/lB that corresponds to NF = 3. The full black line represents the total polarizability, whereas the red dotted and the blue dashed lines show the intraband and the interband contributions, respectively. From (Rold´an et al., 2010). q +(cid:18) eBv2 vF√πnel(cid:19)2 F as may be shown easily within a hydrodynamic ap- proach that has been successfully applied to the upper- hybrid mode in non-relativistic 2D electron systems (Chiu and Quinn, 1974). It is apparent from Fig. 17 that this mode remains coherent also within the region II, which corresponds to the B = 0 interband particle- hole continuum. In addition to the upper-hybrid mode, one notices lin- early dispersing coherent modes in the regions I and II that emerge from the lines of large spectral weight in the non-interacting PHES, as expected from the quali- tative discussion above. We may call these modes lin- ear magneto-plasmons (Rold´an et al., 2009, 2010) in or- der to distinguish them clearly from the upper-hybrid mode (169) and the weakly dispersing magneto-excitons at low doping (Iyengar et al., 2007). These modes are more prominent in the interband than in the intraband region although they are also visible there. They are gen- uine to graphene with its characteristic √Bn LLs that inevitably overlap in energy, above a critical LL n, if level broadening is taken into account, and they may not be captured in the usual magneto-exciton approximation where the collective modes are adiabatically connected to the inter-LL excitations (Bychkov and Martinez, 2008; Iyengar et al., 2007; Kallin and Halperin, 1984). 4. Dielectric function and static screening One notices that up to 2qF , the zero-field static po- larizability [Fig. 18(a)] remains constant. Indeed, the interband contribution (blue dashed line) increases lin- early with the wave vector (Ando, 2006a; Gonz´alez et al., 1999) − Πvac(q) = q 4vF (172) and thus cancels the linear decrease of the intraband con- tribution (red dotted line), − Πdop(q <∼ 2qF ) ≃ ρ(ǫF )(cid:18)1 − q 2qF(cid:19) , where ρ(ǫF ) = ǫF 2π2v2 F (173) (174) is the density of states per unit area at the Fermi energy. At wave vectors larger than 2qF , the intraband contri- bution dies out, and the polarizability is dominated by interband excitations. With the help of these two contributions, we may rewrite the static dielectric function (170) as εRP A(q) = ε∞(cid:18)1 + παG 2ε∞ Πdop(q) Πvac(q)(cid:19) , where we have defined (175) (176) We terminate this section with a brief discussion of the dielectric function (161) in the static limit ε∞ ≡ εRP A(q → ∞) = 1 + π 2 αG, εRP A(q) ≡ εRP A(q, ω = 0) = 1 − Π0(q, ω = 0), 2πe2 εq (170) by comparing the B 6= 0 to the zero-field case, as shown in Fig. 18. As mentioned above, one may distinguish two separate contributions to the static polarizability, the vacuum polarizability Πvac(q) ≡ Πvac(q, ω = 0) due to interband excitations and the intraband contribution Πdop(q) ≡ Πdop(q, ω = 0), which is only present in doped graphene, Π0(q, ω = 0) = Πvac(q) + Πdop(q). (171) i.e. the value that the static dielectric function ap- proaches at large wave vectors. Notice that this is pre- cisely the RPA result for the intrinsic dielectric constant for undoped graphene (Gonz´alez et al., 1999). The above form of the static dielectric function may be cast into a Thomas-Fermi form, εT F (q) ≃ ε∞(cid:18)1 + α∗G qF q (cid:19) , (177) in terms of the effective coupling constant α∗G = αG ε∞ = αG 1 + παG/2 , (178) α ∗ G 2/π Out[22]= 1.0 0.8 0.6 0.4 0.2 α G 2 4 6 8 10 Figure 19 Effective coupling constant α∗ G as a function of the bare coupling αG. The dashed line indicates the asymptotic value 2/π obtained for large values of the bare coupling (rs ≫ 1). which is plotted in Fig. 19. One notices that interband excitations yield a contri- bution to the dielectric constant that originally takes into account the dielectric environment in which the graphene sheet is embedded, ε → ε∗ = εε∞. This is a direct conse- quence of the linear behavior of the vacuum polarization (172) as a function of the wave vector and thus specific to graphene. Furthermore, one may also define an effec- tive Thomas-Fermi wave vector q∗T F = qT F /ε∞ = α∗GqF , which describes the screening length in the presence of the vacuum polarization. As a consequence of the sat- uration of the effective coupling constant (178) at large values of αG, the effective Thomas-Fermi vector is thus always on the order of the Fermi wave vector unless αG ≪ 1, where α∗G ∼ αG. The relevant effective param- eters are summarized in the table below for freestanding graphene and graphene on mainly used substrates.23 graphene ε αG ε∞ α∗ G in vacuum 1 2.2 4.5 0.5 2.5 0.9 2.4 0.38 on SiO2 on h-BN 2.3 1 2.4 0.39 on SiC 5.5 0.4 1.6 0.25 Table I Dielectric constant ε, ε∞, bare coupling αG, and effective coupling α∗ G for graphene in vacuum and popular substrates. Finally, the screened Coulomb interaction potential may be approximated as vscr(q) ≃ 2πe2 εε∞(1 + α∗GqF /q)q . (179) One notices from this expression that processes at wave vectors q ≪ qF , where the interband excitations play a 30 Figure 20 Static dielectric function for graphene in the IQHE regime for NF = 1, 2 and 3 (blue, red and green curves, in increas- ing order). From (Rold´an et al., 2010). negligible role [see Fig. 18(a)], are still governed by the bare coupling constant αG ∼ vscr(q ≪ qF )/vF q. How- ever, processes at or above the Fermi wave vector, such as those that are relevant in the electronic transport, are characterized by the effective coupling constant α∗G ∼ vscr(q >∼ qF )/vF q, which saturates at a value of 2/π as mentioned above. If we consider the value (133), αG ≃ 2.2, for the bare coupling constant of graphene in vac- uum, the effective coupling is roughly four times smaller α∗G ≃ 0.5, such that the electrons in doped graphene ap- proach the weak-coupling limit. The situation is differ- ent in undoped graphene, where recent renormalization- group (Herbut et al., 2009a,b; Jurici´c et al., 2009) and lattice gauge theoretical calculations (Drut and Lahde, 2009a,b) indicate a flow towards strong coupling at mod- erate values of αG. In Fig. 18(b), we have plotted the static polarizability for graphene in the IQHE regime. Qualitatively, the re- sult agrees with the zero-field behavior, with a (roughly) linearly increasing vacuum polarizability and a decreas- ing intraband contribution, apart from some superim- posed oscillations due to the overlap functions that are reflected by the form factors (141). An important differ- ence is manifest in the small wave-vector limit of the po- larizability. In contrast to the zero-field case, where the polarizability saturates at a non-zero density of states, the system is gapped in the IQHE regime with a result- ing vanishing density of states at the Fermi energy. This gives rise to a q2 behavior of the polarizability at small wave vectors. Furthermore, the static dielectric function, which is shown in Fig. 20 (Rold´an et al., 2010; Shizuya, 2007), does no longer diverge in this limit, contrary to the zero-field Thomas-Fermi result (177). Indeed, the small-q behavior may be approximated as εRP A(q) − 1 ∝ αGN 3/2 F qlB, (180) 23 The dielectric constant ε is then the average of the dielectric constant of the substrate material and that of the vacuum. which is the same as for non-relativistic 2D elec- trons (Aleiner and Glazman, 1995).24 The maximum of the static dielectric function is obtained at qlB ∼ 1/qF lB ∼ 1/√2NF + 1, at the value εmax ≃ εRP A(q ∼ 1/lB√2NF + 1) ∼ αGNF . It therefore scales as εmax ∝ NF in contrast to a √NF scaling in non-relativistic 2D systems (Aleiner and Glazman, 1995). At large wave vectors, the static dielectric function saturates at the same value ε∞ as in zero magnetic field. IV. MAGNETO-PHONON RESONANCE IN GRAPHENE In the previous section, we have discussed the role of electron-electron interactions in the IQHE regime, where a perturbative (diagrammatic) approach may be applied. Similarly, one may treat the electron-phonon interaction in a perturbative manner in this regime. This is the subject of the present section, before discussing again electron-electron interactions in the strong coupling limit of partially filled LLs (Sec. V). As a consequence of the honeycomb-lattice structure of graphene, with two inequivalent sublattices, there are four in-plane phonons, two acoustic and two optical ones. Each phonon type occurs in a longitudinal (longitudinal acoustic, LA, and longitudinal optical, LO) and a trans- verse (transverse acoustic, TA, and transverse optical, TO) mode. In addition to these four phonons, one finds two out-of-plane phonons, one acoustic (ZA) and one op- tical (ZO) [for a review of phonons in graphene, see Refs. (Saito et al., 1998) and (Wirtz and Rubio, 2004)]. Here, we concentrate on the in-plane optical phonons LO and TO, which couple to the electronic degrees of freedom. More specifically, we discuss these phonons at the Γ point (E2g modes) in the center of the first BZ, which are at- tributed to the G-band at ωph ≃ 0.2 eV in the Raman spectra [see e.g. Refs. (Ferrari et al., 2006; Graf et al., 2007; Gupta et al., 2006; Pisana et al., 2007; Yan et al., 2007)]. One of the most prominent effects of electron-phonon coupling in metals and semiconductors is the so-called Kohn anomaly (Kohn, 1959), which consists of a sin- gularity in the phonon dispersion due to a singular- ity in the electronic density-density response function. The analog of the Kohn anomaly in graphene yields a logarithmic divergence of the phonon frequency when the bare phonon frequency coincides with twice the Fermi energy (Ando, 2006b; Castro Neto and Guinea, 2007; Lazzeri and Mauri, 2006). We investigate how this renormalization manifests itself in graphene in a strong magnetic field (Ando, 2007b; Goerbig et al., 2007). In Sec. IV.A, we consider the specific form of the electron- phonon coupling and discuss its consequences for the 24 Notice, however, that the expression becomes exact only in the large-NF limit and that in non-relativistic 2D electron systems, the coupling constant rs also depends on NF , rs ∼ N −1/2 F . (a) −dt/da TO phonon LO phonon q u(r) (b) q electron on A k+q g ~ −dt/da (c) 31 K’ Γ (q~0) K phonon k electron on B optical phonons electron−phonon coupling Brillouin zone Figure 21 Electron-phonon coupling in graphene. (a) Optical phonons in graphene, with a wave vector q in the vicinity of the Γ point at the center of the first BZ [see part (c)]. The amplitude of the LO phonon is in the direction of propagation (black arrows), that of the TO phonon perpendicular (red arrows). The optical phonons modify the bond lengths of the honeycomb lattice. (b) As a consequence of the modified bond lengths, the electronic hopping is varied, and the electron-phonon coupling is off-diagonal in the sublattice index. renormalization of the optical phonons at the Γ point in Sec. IV.B. More specifically, we consider both non-resonant (Sec. IV.B.1) and resonant coupling (Sec. IV.B.2), the latter being specific to graphene in a mag- netic field when the phonon is in resonance with an al- lowed inter-LL transition (magneto-phonon resonance) (Goerbig et al., 2007). A. Electron-Phonon Coupling The LO and TO phonons in graphene are schematically represented in Fig. 21(a). As already mentioned above, we concentrate on phonons at small wave vectors, in the vicinity of the Γ point. The origin of the electron-phonon coupling may easily be understood from the variation of the bond length caused by the phonon, which affects the electronic hopping amplitude between nn carbon atoms. As we have discussed in Sec. I.D, the effect may be quan- tified with the help of Harrison’s law (Harrison, 1981), which yields ∂t/∂a ≃ −4.3 eV/A [see Eq. (57)]. The order of magnitude for the bare electron-phonon energy is then obtained by multiplying this variation with the typical length scale p/M ωph, which characterizes the amplitude of a lattice vibration of frequency ωph within the harmonic approximation. The intervening mass M is that of the carbon atom. Indeed, a tight-binding calcu- lation (Ando, 2006b; Ishikawa and Ando, 2006) corrob- orates this argument, apart from a numerical prefactor 3/2, and yields a bare electron-phonon coupling g = − 3 2 ∂t ∂as M ωph ≃ 0.26 eV. (181) This value agrees well with ab-initio calculations (Piscanec et al., 2004), although it is though slightly lower than the value determined experimentally, which from g ≃ 0.3 eV (Faugeras et al., 2009; ranges Pisana et al., 2007) to g ≃ 0.36 eV (Yan et al., 2007). Furthermore, one notices that, because the electron- phonon coupling is mediated by a bond variation between sites that belong to two different sublattices, the coupling constant is off-diagonal in the sublattice basis. This is diagrammatically depicted in Fig. 21(b). where σ ∧ u(r) = [σ × u(r)]z = σxuy(r) − σyux(r) is the 2D cross product between the Pauli matrices and the displacement field. 1. Coupling Hamiltonian 2. Hamiltonian in terms of magneto-exciton operators 32 The above considerations help us to understand the different terms in the Hamiltonian, H = Hel + Hph + Hcoupl which serves as the basis for the analysis of the electron- phonon coupling. The Hamiltonian for 2D electrons in a magnetic field, Hel = Xξ Z d2r ψ†ξ(r)Heff,ξ B ψξ(r) ǫλ,nc†λn,m;ξcλn,m;ξ (182) = Xλn,m;ξ may be written, in second quantization, in terms of the one-particle Hamiltonian (75) and the fermionic fields ψξ(r) = Xλn,m ψλn,m;ξ(r)cλn,m;ξ, where ψλn,m;ξ(r) is the wave function in position space associated with the spinor (138). The lattice vibration is characterized by the relative displacement u(r) between the two sublattices, which may be decomposed in terms of the bosonic operators bµ,q and b†µ,q, u(r) =Xµ,qs 2NucM ωµ(q)(cid:16)bµ,q + b†µ,−q(cid:17) eµ,qe−iq·r, (183) where eµ,q denotes the two possible linear polariza- tions (µ = LO,TO) at the wave vector q and Nuc = A/(3√3a2/2) is the number of unit cells in the system. The phonon Hamiltonian then reads, on the level of the harmonic approximation, Hph =Xµ,q ωµ(q)b†µ,qbµ,q, (184) in terms of the phonon dispersion ωµ(q). Notice that, at the Γ point, the frequencies of the LO and TO phonons coincide, and one has ωph ≡ ωµ(q = 0). It is then con- venient to describe the phonon modes in terms of circu- larly polarized modes, u(cid:9)(r) = [ux(r) + iuy(r)]/√2 and u(cid:8)(r) = u∗(cid:9)(r). Finally, taking into account the above considerations on the electron-phonon coupling, the coupling Hamilto- nian reads (Ando, 2006b; Castro Neto and Guinea, 2007; Ishikawa and Ando, 2006) Hcoupl = gr 2M ωph Xξ Z d2r ψ†ξ(r) [σ ∧ u(r)] ψξ(r), (185) As a consequence of the off-diagonal character of the electron-phonon coupling (185), one notices that the in- tervening matrix elements are proportional to δn,n±1, and one thus obtains the selection rules λn → λ′(n ± 1), (186) in analogy with the magneto-optical selection rules dis- cussed in Sec. II.A.1. Furthermore, if we fix the energy of the dipole transition (105) to be25 ∆n ≡ ∆n,λ=− = √2(vF /lB)(√n + 1 + √n), there are two possible tran- sitions, which may be distinguished by the circular po- larization of the phonon they are coupled to. Indeed, the form of the electron-phonon coupling (185) indicates that the (cid:9)-polarized phonon is coupled to the transi- tion −(n + 1) → +n, whereas the (cid:8)-polarized phonon couples to the −n → +(n + 1) interband transition (Goerbig et al., 2007). It is then convenient to introduce magneto-exciton op- erators, associated with the above-mentioned inter-LL transitions φ†(cid:9)(n, ξ) = φ†(cid:8)(n, ξ) = ip1 + δn,0 ip1 + δn,0 n Xm N (cid:9) n Xm N (cid:8) c†+n,m;ξc−(n+1),m;ξ, c†+(n+1),m;ξc−n,m;ξ, (187) where the index A =(cid:9), (cid:8) characterizes the angular mo- mentum of the excitation and where the normalization factors and n = q2(1 + δn,0)NB(cid:2)¯ν−(n+1) − ¯ν+n(cid:3) N (cid:9) n = q2(1 + δn,0)NB(cid:2)¯ν−n − ¯ν+(n+1)(cid:3) N (cid:8) are used to ensure the bosonic commutation rela- tions of the exciton operators, [φA(n, ξ), φ† A′ (n′, ξ′)] = δA,A′ δξ,ξ′δn,n′ , including the two-fold spin degeneracy. These commutation relations are obtained within the mean-field approximation with hc†λn,m;ξcλ′n′,m′;ξ′i = δξ,ξ′δλ,λ′ δn,n′ δm,m′(δλ,− + δλ,+ ¯νλn), where 0 ≤ ¯νλn ≤ 1 is the partial filling factor of the n-th LL, normalized to 1. One notices that the magneto-exciton operators are, apart from a normalization constant, nothing other 25 We consider mainly interband transitions here, which may have a chance to be in resonance with the phonon of energy ωph ∼ 0.2 eV. respectively, at zero wave vector. than the reduced density operators (140), φ†(cid:9)(n, ξ) ∝ +n,−(n+1)(q = 0) and φ†(cid:8)(n, ξ) ∝ ¯ρξ,ξ ¯ρξ,ξ +(n+1),−n(q = 0), Notice fur- thermore that, because of the relative sign ξ between the kinetic part (182) and the electron-phonon cou- pling Hamiltonian (185), the optical phonons couple to the valley–anti-symmetric magneto-exciton combina- tion φA,as(n) = [φA(n, ξ = +) − φA(n, ξ = −)]/√2. This needs to be contrasted to the magneto-optical coupling (Abergel and Fal’ko, 2007; Iyengar et al., 2007; Sadowski et al., 2006), where the photon couples to the valley-symmetric mode φA,s(n) = [φA(n, ξ = +) + φA(n, ξ = −)]/√2. The magneto-exciton operators (187) allow one to rewrite the electron-phonon Hamiltonian at the Γ point (q = 0) in a bosonic form as (Goerbig et al., 2007) H = Xτ =s,asXA,n +XA,n gA(n)hb† ∆nφ† A,τ (n)φA,τ (n) +XA AφA;as(n) + bAφ† ωphb† A bA A;as(n)i , (188) in terms of the effective coupling constants g(cid:9)(n) = gq(1 + δn,0)γp¯ν−(n+1) − ¯ν+n, and g(cid:8)(n) = gq(1 + δn,0)γp¯ν−n − ¯ν+(n+1), with the constant γ ≡ 3√3a2/2πl2 B. One therefore re- marks that, although the bare coupling constant g is rather large [see Eq. (181)], the effective coupling is re- duced by a factor of a/lB, gA(n) ∼ g a lB ∼ 1 . . . 2 meVpB[T]. (190) B. Phonon Renormalization and Raman Spectroscopy The Hamiltonian (188) shows that a phonon may be destroyed by exciting a magneto-exciton, and the associ- ated Dyson equation for the dressed phonon propagator D(ω) reads DA(ω) = D0(ω) + D0(ω)χA(ω)DA(ω), (191) in terms of the bare bosonic phonon propagator D0(ω) = 1 2ω ω2 − ω2 ph (192) and χA(ω) = Nc Xn=NF +1 2∆ng2 A(n) 2ω2 − ∆2 n + 2 ∆NF g2 A(n) 2ω2 − ∆2 NF . (193) The form of the last expression is transparent; the magneto-exciton is a boson, and its propagator is there- fore of the same form as that of the bare phonon. It 33 is equivalent to a particle-hole propagation associated with a polarization bubble [see Fig. 15(a)], but the ex- pression (193) also takes into account the square of the effective coupling constant which is due to the double occurence of the electron-phonon coupling – first when the phonon is converted into a magneto-exciton and the second time when the magneto-exciton is destroyed by creating a phonon. The last term in Eq. (193) takes into account the (only) possible intra-band magneto- exciton from the last filled LL NF to NF + 1 with energy omitted in the Hamiltonian (188) because it is irrelevant for resonant coupling. The parameter Nc is the same ∆NF = √2(vF /lB)(√NF + 1 − √NF ), which we have high-energy cutoff, defined by ǫNc = (vF /lB)√2Nc ∼ t, The renormalized phonon frequencies ωA may be ob- tained from the Dyson equation (191), by searching the poles of the dressed phonon propagator as in Sec. III.B of the preceding chapter. DA(ω)−1 = 0 = DA(ωA)−1 − χA(ωA), (194) and one finds (Ando, 2007b; Goerbig et al., 2007) ω2 A−ω2 ph = 4ωph " Nc Xn=NF +1 (189) ∆ng2 2 ω2 A(n) A − ∆2 n + ∆NF g2 2 ω2 NF # . A(NF ) A − ∆2 (195) 1. Non-resonant coupling and Kohn anomaly Before discussing resonant coupling, i.e. when the phonon frequency is in resonance with a possible inter- LL excitation, in a strong magnetic field, we comment on the relation between Eq. (195) and the (non-resonant) renormalization of the phonon frequency in zero mag- netic field. The zero-field limit may indeed be obtained from Eq. (195) if one replaces the sum Pn by an in- tegral R dn, i.e. if the spacing between the LLs van- ishes, ∆NF → 0. Linearizing Eq. (195), if one replaces ωA → ωph in the denominators, and using the approxi- mation √n + √n + 1 ≈ 2√n yields ln(cid:18) ωph + 2√2NF vF /lB ωph − 2√2NF vF /lB(cid:19)(cid:21) , ω ≃ ω0+λ(cid:20)p2NF where λ = (2/√3π)(g/t)2 ≃ 3.3 × 10−3 is the dimen- sionless electron-phonon coupling constant introduced in Refs. (Ando, 2006b, 2007b), and vF lB − ωph 4 (196) ω0 ≃ ωph + 2 Z Nc 0 dn ∆ng2 2ω2 A(n) ph − ∆2 n (197) is the physical phonon frequency at zero doping. In- deed, the frequency ωph is not relevant in a physi- cal measurement in graphene even if it occurs in the Hamiltonian, but one measures ω0 at zero doping and B = 0. Equation (196) coincides precisely with the zero-field result (Ando, 2006b; Castro Neto and Guinea, (a) 2 2 1.75 ω h p 1.5 f 1.25 o s t i n=2 n=1 n n u 1 1 0.75 i y g r e n E 0.25 0.5 0 5 10 10 15 20 20 (a) n=0 B0 25 30 30 40 35 Magnetic Field [T] (b) Figure 22 (a) From Ref. (Goerbig et al., 2007). Anticrossing of coupled phonon-magneto-exciton modes as a function of the mag- netic field. (b) From Ref. (Faugeras et al., 2009). Colour plot of the Raman spectra as a function of the magnetic field. The contin- uous white lines indicate the magnetic field for which the phonon is in resonance with an inter-LL excitation of energy ∆n. Top: data of the full B-field range. Bottom: zoom on the range from 0 to 10 T. 2007; Lazzeri and Mauri, 2006) if one identifies the chem- ical potential with the energy of the last filled LL, µ = √2NF vF /lB (Goerbig et al., 2007). 2. Resonant coupling Apart from the non-resonant coupling discussed in the preceding section, the high-field electron-phonon cou- pling reveals a linear effect when the phonon is in res- onance with a particular magneto-exciton, ωph ≃ ∆n. In this case, the sum on the right-hand side in Eq. (195) is dominated by a single term and may be ap- proximated by 2(ωph/)g2 A(n)/(ωA − ∆n). This re- sults in a fine structure of mixed phonon–magneto- exciton modes, φA,as(n) cos θ + bA sin θ with frequency A and φA,as(n) sin θ−bA cos θ with frequency ω− ω+ A [where cot 2θ = (∆n−ω0)/2gA]. The frequencies of these mixed boson modes read (Ando, 2007b; Goerbig et al., 2007) 34 1 2(cid:18) ∆n + ω0(cid:19) ∓s 1 ω± A(n) = A(n), (198) and the resulting phonon–magneto-exciton anticrossing is depicted in Fig. 22(a). 4(cid:18) ∆n − ω0(cid:19)2 + g2 The above-mentioned anticrossing of the coupled phonon–magneto-exciton modes has been observed in recent Raman experiments on epitaxial graphene. Re- member that Raman spectroscopy is sensitive to the phonon component of the mixed modes (Faugeras et al., 2009). The results are shown in Fig. 22(b) and corrob- orate the theoretically expected behavior (Ando, 2007b; Goerbig et al., 2007). Indeed, one may obtain the oscil- lating behavior from a numerical solution of Eq. (195) if one expresses the equation in terms of ω0 instead of ωph and if one takes into account a finite broadening of the levels. If the phonon is out of resonance with an inter-LL transition, its frequency is essentially field-independent and coincides with the energy of the E2g line at 1586 cm−1 ≃ 0.2 eV. When it approaches the resonance (by increasing the magnetic field), its energy is shifted up- wards as a consequence of the anticrossing but rapidly dies out in intensity once the magneto-exciton component becomes dominant in the ω+ A mode. Upon further in- crease of the magnetic field, the ω− A mode becomes more phonon-like and therefore visible in the Raman spectra. The fine structure of the high-field resonant electron- phonon coupling may furthermore be investigated by sweeping the chemical potential when the magnetic field is held fixed at resonance. The effect is most pronounced for the resonance ωph ≃ ∆n=0, which is expected at B ≃ 30 T [see Fig. 22(a)]. In this case, the mode consists of an equal-weight superposition of the phonon and the magneto-exciton (cos θ = sin θ = 1/√2), and the E2g band would appear as two lines, at the energies ω± = ω0 ± gA, for the case of undoped graphene.26 With the above estimation (181) for the bare electron- phonon coupling constant, one obtains for the line split- ting 2gA ∼ 16 meV (∼ 130 cm−1), which largely exceeds the G-band width observed in Refs. (Ferrari et al., 2006; Graf et al., 2007; Gupta et al., 2006; Pisana et al., 2007; Yan et al., 2007). It is apparent from the expressions (190) for the ef- fective coupling constants g(cid:9) and g(cid:8) that the splitting may be controled by the LL filling factor. Exactly at zero doping, both coupling constants coincide, g(cid:9) = g(cid:8), but upon electron doping the transition −1 → 0 associ- ated with the (cid:9)-polarization becomes weaker due to the reduced number of final states in n = 0, whereas the 0 → +1 transition (with polarization (cid:8)) is strengthened. 26 Notice, however, that only an oscillation of the phonon mode, and not a splitting, has been observed in the experiment by Faugeras et al. (Faugeras et al., 2009). As a consequence, the associated coupling constants are increased and decreased, respectively, until the coupling constant g(cid:9) vanishes at ν = 2. The above-mentioned filling-factor dependence has a direct impact on the Raman lines (Goerbig et al., 2007). Whereas at ν = 0, one expects two lines separated by the energy 2g(cid:9) = 2g(cid:8), the degeneracy in the circular po- larization is lifted between 0 < ν < 2.27 One therefore expects to observe four lines instead of two, where the in- ner ones are associated with the polarization (cid:9), whereas the outer ones with increased splitting correspond to the opposite polarization (cid:8). The separation between the in- ner lines vanishes then at ν = 2, where the splitting of the outer lines is maximal and where one expects to observe three lines. V. ELECTRONIC CORRELATIONS IN PARTIALLY FILLED LANDAU LEVELS This last section is devoted to the physics of in- teracting electrons in the strong-correlation limit of a partially filled LL. The motivation stems from non- relativistic quantum Hall systems in GaAs heterostruc- tures, where these correlations lead to the formation of incompressible quantum-liquid phases, which display the fractional quantum Hall effect (FQHE) (Tsui et al., 1982), as well as of exotic electron-solid phases, such as the high-field Wigner crystal (Andrei et al., 1988; Williams et al., 1991) or the theoretically pre- dicted bubble and stripe phases (Fogler et al., 1996; Koulakov et al., 1996; Moessner and Chalker, 1996). The latter are likely to be at the origin of highly anisotropic transport properties at half-filled higher LLs (Du et al., 1999; Lilly et al., 1999), particular elec- tron transport under microwave irradiation (Lewis et al., 2005, 2004, 2002), and an intriguing reentrance of the IQHE in n = 1 and n = 2 (Cooper et al., 1999; Eisenstein et al., 2002). It is therefore natural to ask whether such strongly- correlated phases exist also in graphene and if so what the differences are with respect to non-relativistic 2D electrons. Moreover, the fact that the electrons reside at the surface opens up the possibility to probe these phases by spectroscopic means, such as scanning tunnel- ing spectroscopy, which has already been applied success- fully in the analysis of the electron density distribution of exfoliated (Martin et al., 2008) and epitaxial graphene (Mallet et al., 2007), as well as graphene on graphite sub- strates (Li et al., 2009b). After a short discussion of the Coulomb interaction in graphene as compared to non-relativistic 2D electrons, we introduce the basic model of interacting electrons 27 We present the argument for a Fermi energy in the CB, i.e. ν > 0, but the situation is generic, and the argument also applies in the VB if one interchanges the polarizations. 35 in a partially filled relativistic LL (Sec. V.A). This model yields a qualitative understanding of the above- mentioned correlated electronic phases in the context of graphene, as compared to non-relativistic electrons. In Sec. V.B, we apply this model to the quantum Hall ferro- magnetism with an internal SU(4) symmetry that is the relevant symmetry in graphene LLs and discuss its rela- tion with the experimentally observed degeneracy lifting of the zero-energy LL n = 0 (Zhang et al., 2006). We ter- minate this section with a review of the specific FQHE in graphene (Sec. V.C), which has recently been observed in the two-terminal (Bolotin et al., 2009; Du et al., 2009) as well as in the four-terminal geometry (Dean et al., 2011; Ghahari et al., 2011). A. Electrons in a Single Relativistic Landau Level Quite generally, the origin of strongly-correlated elec- tron phases is a quenched kinetic energy, where the par- tially filled LL is separated by the cyclotron gap from the neighboring ones such that inter-LL excitations con- stitute high-energy degrees of freedom. The Coulomb in- teraction, which may though be small with respect to the cyclotron gap, remains then as the only relevant energy scale which dominates the low-energy degrees of free- dom if we can neglect disorder effects. This leads to the seemingly counter-intuitive finding of strongly-correlated phases in weakly-correlated matter. In order to quantify the degree of separation between the energy scales, one may use a similar argument as the one that led us to the definition of the dimension- less interaction parameter (131), introduced at the be- ginning of Sec. III. Indeed, one needs to compare the Coulomb interaction energy Eint = e2/εRC at the char- acteristic length scale RC = lB√2n + 1 to the LL spacing ωC = eB/mb, where we concentrate on non-relativistic electrons first, rB s = e2 εvF (n, B) , with vF (n, B) ≡ RC ωC . If one identifies the Fermi wave vector kF ≃ √2n/lB, one obtains the same expression as for the zero-field coupling constant (132), (199) rB s = rs = mbe2 2ε k−1 F ∼ 1 a∗0kF . (200) This means that the degree of LL mixing is still governed by rs, and the inter-LL excitations are well separated from the low-energy intra-LL degrees of freedom unless rs becomes very large. Notice, however, that rs ∼ 1 in most 2D electron systems. In the case of partially filled relativistic LLs in graphene, one is tempted to apply the same argument – if the Coulomb interaction e2/εRC is sufficiently small as compared to the LL spacing ∆n, the relevant degrees of freedom are those which couple quantum states in the (a) (b) LL separation X describe the density fluctuations inside the LL λn that interest us here. Notice that we have dropped the index λn in the definition of the projected density operators; they satisfy the quantum-mechanical commutation rela- tions (Girvin et al., 1986) X x x x x 36 Figure 23 (a) Completely filled topmost LL. Due to the Pauli principle, the only possible excitations are inter-LL transitions. (b) Partially filled LL. For sufficiently small values of rs (or αG), the inter-LL excitations constitute high-energy degrees of freedom that may be omitted at low energies, where the relevant degrees couple states within the same LL. same LL, whereas inter-LL excitations may be considered as frozen out (see Fig. 23). Although this seems a rea- sonable assumption for the lowest LLs, one is confronted with the apparent problem that the LL spacing rapidly decreases once the Fermi level resides in higher LLs, ∆n = √2 vF lB (√n + 1 − √n) ≃ vF lB√2n . (201) Notice, however, that this decrease is balanced by the 1/√2n scaling of the characteristic Coulomb interaction, such that even in higher LLs the separation between low- and high-energy degrees of freedom is governed by the dimensionless coupling constant [¯ρ(q), ¯ρ(q′)] = 2i sin(cid:18) q′ ∧ q l2 2 B (cid:19) ¯ρ(q + q′), (204) where q′ ∧ q = (q′ × q)z = q′xqy − q′yqx is the 2D vec- tor product between q′ and q, and these commutation relations are independent of the LL index λn. The in- formation about the LL is indeed waved to the effective interaction potential vn(q) = 2πe2 εq [Fn(q)]2 , (205) in terms of the LL form factors [see Eq. (141) and their explicit form (A2), discussed in Appendix A] Fn(q) = B 1 2(cid:20)(1 − δn,0)Ln−1(cid:18) q2l2 2 (cid:19) +(1 + δn,0)Ln(cid:18) q2l2 2 (cid:19)(cid:21) e−q2l2 B B /4, (206) αB G = e2/εlB√2n vF /lB√2n = e2 εvF = αG, (202) independent of the band index λ (Goerbig et al., 2006; Nomura and MacDonald, 2006). The Hamiltonian re- sulting from Eq. (151) which coincides with the scale-invariant zero-field cou- pling constant (133). From the interaction point of view, the restriction of the electron dynamics to a single par- tially filled LL in the large-n limit is therefore as justified as for the lowest relativistic LLs. Naturally, this state- ment only holds true in the absence of disorder that in- duces stronger LL mixing for n ≫ 1 than in n = 0 or ±1. Hn = 1 2Xq vn(q)¯ρ(−q)¯ρ(q) (207) therefore defines, together with the commutation rela- tion (204) the model of strongly-correlated electrons re- stricted to a single relativistic LL. The model respects the SU(4) spin-valley symmetry, and naturally, there is no kinetic energy scale because all processes involve states within the same LL. 1. SU(4)-symmetric model Formally, the above-mentioned separation into high- and low-energy degrees of freedom may be realized with the help of the reduced density operators (140). For the moment, we only consider the case where ξ = ξ′, i.e. we concentrate on the valley-symmetric model, in which case the reduced (intra-valley) density operators (153) fall into two distinct classes: for n 6= n′ or λ 6= λ′, the operators ¯ρλn,λ′n′(q) describe density fluctuations corre- sponding to inter-LL transitions of an energy equal to or larger than the LL separation ∆n, whereas the projected density operators ¯ρ(q) ≡ ¯ρλn,λn(q) (203) = Xξ=± Xσ=↑,↓ Xm,m′(cid:10)m(cid:12)(cid:12)e−iq·R(cid:12)(cid:12) m′(cid:11) c†λn,m;ξ,σcλn,m′;ξ,σ a. Algebraic properties. The SU(4) spin-valley symmetry is formally described with the help of the spin and valley- pseudospin operators ¯Sµ(q) = (Sµ ⊗ 1) ⊗ ¯ρ(q) and ¯I µ(q) = (1 ⊗ I µ) ⊗ ¯ρ(q), (208) respectively, that are tensor products between the pro- jected density operators (203) and the operators Sµ and I µ, which are (up to a factor 1/2) Pauli matri- ces and that describe the spin and valley degrees of freedom, respectively. The operators (Sµ ⊗ 1) and (1 ⊗ I µ) are the generators of the SU(2)×SU(2) sub- group of SU(4). However, once combined in a ten- sor product with the projected density operators ¯ρ(q), the SU(2)×SU(2)-extended magnetic translation group is no longer closed due to the non-commutativity of the Fourier components of the projected density opera- tors. The commutators [ ¯Sµ(q), ¯I ν (q′)] yield the remain- ing generators of the SU(4)-extended magnetic trans- lation group (Dou¸cot et al., 2008; Ezawa and Hasebe, 2002; Ezawa et al., 2003). Physically, the operators introduced in Eq. (208) play the role of projected spin and valley-pseudospin densi- ties, where the LL projection is induced by the projected charge-density operator ¯ρ(q). Their non-commutativity with the projected charge density, [ ¯Sµ(q), ¯ρ(q′)] 6= 0 and [ ¯I µ(q), ¯ρ(q′)] 6= 0, which are due to the commutation re- lation (204), is at the origin of the (pseudo-)spin-charge entanglement in quantum Hall systems: as we discuss in more detail in Sec. V.B.2, this entanglement yields (pseudo-)spin-texture states that carry an electric in ad- dition to their topological charge. b. Validity of the model. With the help of the Hamilto- nian (207), we may render more transparent the model assumption of electrons restricted to a single relativistic LL. We need to show that the energy scale that governs the model (207) and its resulting phases is indeed given by e2/εRC and not e2/εlB. As an upper bound for the energy scale, one may use the energy of a completely filled LL described by hc†λn,m;ξ,σcλn,m′;ξ,σi = δm,m′, the mean- field energy hHni/N of which is simply the exchange en- ergy,28 En X = − 1 2Xq vn(q) = − e2 2εZ ∞ 0 dq [Fn(q)]2 . (209) In order to estimate the integral in the large-n limit, one may use the scaling form (Abramowitz and Stegun, 1970; Gradshteyn and Ryzhik, 2000) of the Laguerre polyno- mials Ln(cid:18) q2l2 2 (cid:19) e−q2l2 B B /4 ≃ J0(qlB√2n + 1), (210) in terms of the Bessel function J0(x), such that one obtains by a simple change of the integration variable 0 dq[Fn(q)]2 ≃ (lB√2n)−1R ∞ R ∞ where c is a numerical factor of order one. The exchange energy of a completely filled LL n therefore scales with n ≫ 1 as 0 dx[J0(x)]2 = c/lB√2n, En X ≃ −c e2 εlB√2n ≃ −c e2 εRC , (211) in agreement with the model assumption of a separation between high- and low-energy degrees of freedom and the definition (202) of the coupling constant αB G. 28 The direct energy is compensated by the positively charged back- ground (“jellium model”) (Mahan, 1993) 37 2. Symmetry-breaking long-range terms When decomposing the Coulomb interaction in the two-spinor basis (Sec. III.A), we have seen that the SU(4)-symmetric model yields the leading energy scale, whereas the only relevant symmetry-breaking term is as- sociated with backscattering processes at an energy scale roughly a/lB times smaller than the leading one. When restricted to a single relativistic LL λn, these backscat- tering terms yield a contribution H sb n = 1 2 Xξ=±Xq n (q)¯ρξ,−ξ(−q)¯ρ−ξ,ξ(q), vsb (212) in terms of the effective backscattering potential vsb n (q) = = 2 2πe2 2πe2 εq εq (cid:12)(cid:12)(cid:12)F +,−λn,λn(q)(cid:12)(cid:12)(cid:12) (1 − δ0,n) n−1(cid:18)q − K2l2 ×(cid:20)L1 (qy − Ky)2l2 (cid:19)(cid:21)2 2n 2 B B (213) e−q−K2l2 B /2, where we have made use of Eq. (142) and of the explicit expressions for the intervening matrix elements (A4). The effect of this symmetry-breaking term will be dis- cussed in more detail in Sec. V.B in the context of the SU(4) quantum Hall ferromagnetism. The term (212) is only relevant in relativistic LLs n 6= 0 as a consequence of the factor (1 − δn,0) in the expression (213) for the backscattering potential (Goerbig et al., 2006). This is a consequence of the chiral symmetry of the zero-energy LL (Arikawa et al., 2008) where the sublattice index is con- funded with the valley pseudospin, as may be seen from the expression (88) for the associated wave functions. Notice, however, that there may occur other symmetry- breaking terms in n = 0 as a consequence of short-range interactions on the lattice scale (Alicea and Fisher, 2006; Doretto and Morais Smith, 2007; Herbut, 2007b). 3. Qualitative expectations for correlated electron phases The model of interacting electrons in a single rela- tivistic LL has the same structure as the one for non- relativistic LLs – in both cases, one has an interaction Hamiltonian that is quadratic in the projected density operators (203) which satisfy the commutation relations (204). This is a noticeable result in the sense that, whereas non-relativistic 2D electron systems are gov- erned by Galileian invariance, the electrons in graphene are embedded in a Lorentz-invariant “space-time”. How- ever, once restricted to a single LL, the electrons forget about their original spatial symmetry properties and are governed by the magnetic translation algebra, which is at the origin of the commutation relations (204). As a con- sequence and in contrast to the IQHE, the differences between strongly-correlated electrons in graphene and (a) l a i t 12 n e t o p 10 10 8 6 5 4 2 n o i t c a r e n t i e v i t c e f f e (b) nm 0.8 V l t i t s a 0.6 n e o p 0.4 o d u e s p n=0 (relativistic and non−relativistic) n=1 (relativistic) n=1 (non−relativistic) n=5 (relativistic) n=5 (non−relativistic) 2 2 4 4 6 6 r/l B 8 8 10 10 n=0 n=1 (rel.) n=1 (non−rel.) 0.8 0.6 0.4 1 0 2 1 3 2 m 4 3 5 4 6 5 Figure 24 (a) From Ref. (Goerbig et al., 2006); comparaison between the relativistic (black curves) and non-relativistic (grey curves) potentials for the LLs n = 0, 1, and 5 in real space. The dashed line shows the potential in n = 0, which is the same in both the relativistic and the non-relativistic case. (b) Pseudopo- tentials for n = 0 (black circles), n = 1 relativistic (blue) and n = 1 non-relativistic (green). The lines are a guide to the eye. The open circles represent pseudopotentials with even relative pair angular momentum that are irrelevant in the case of completely spin-valley polarized electronic states. The energies are given in units of e2/ǫlB. non-relativistic 2D electrons do not stem from their re- spective space-time properties, as one would expect from a mean-field Chern-Simons approach (Khveshchenko, 2007; Peres et al., 2006). The differences between graphene and non-relativistic 2D electrons are rather to be seeked in the larger inter- nal symmetry – instead of an SU(2) spin symmetry, one has an SU(4) spin-valley symmetry if one neglects the small symmetry-breaking term (212) in the interactions. Another difference arises from the different effective in- teraction potential (205), instead of vnon−rel n (q) = 2πe2 εq (cid:20)Ln(cid:18) q2l2 2 (cid:19)(cid:21)2 B e−q2l2 B /2 (214) for the usual 2D electron gas. As one may see from the graphene form factors (206), the effective interaction po- tential in graphene for n 6= 0 is the average of the non- relativistic ones in the adjacent LLs n and n− 1, whereas for n = 0 there is no difference between the relativistic and the non-relativistic case (see Fig. 24), as a conse- quence of the above-mentioned chiral properties. One notices, furthermore, that the difference between the relativistic and non-relativistic effective interaction potentials become less prominent in the large-n limit [see n = 5 in Fig. 24(a)]. This may be understood from the approximate expression (210) of the form factors, which yields Fn(q) ≃ [J0(qlB√2n + 1) + J0(qlB√2n − 1)]/2 ≃ 38 J(qlB√2n) + O(1/n). This result agrees indeed to lead- ing order in 1/n with the scaling expression of the form factors (154) for usual non-relativistic 2D electrons. The strongest difference in the interaction potentials is thus found for n = 1, which in graphene is quite reminis- cent to that in n = 0, apart from a reduced repulsion at very short distances, whereas for non-relativistic 2D elec- trons it has an additional structure [see Fig. 24(a)]. The behavior of the effective interaction potential may also be analyzed with the help of Haldane’s pseudopotentials (Haldane, 1983) V n ℓ = 1 2π Xq vn(q)Lℓ(q2l2 B)e−q2l2 B /2, (215) which represent the interaction between pairs of elec- trons in a magnetic field, in a relative angular momentum state with quantum number ℓ. This quantum number is related to the average distance lB√2ℓ + 1 between the two particles constituting the pair and is a good quan- tum number for any two-particle interaction potential v(ri − rj). The pseudopotentials for graphene are shown in Fig. 24(b) for n = 0 and 1. Haldane’s pseudopotentials are an extremely helpful quantity in the understanding of the possible FQHE states which one may expect in 2D electron systems. One notices first that as a consequence of the anti-symmetry of a two-particle wave function under fermion exchange, the relative angular-momentum quantum number ℓ must be an odd integer, i.e. only the pseudopotentials with odd values of ℓ play a physical role in the description of two interacting electrons of the same type (spin or valley). Even-ℓ pseudopotentials become relevant if the SU(4) spin-valley pseudospin is not completely polarized, in the treatment of two electrons with different internal quantum number σ or ξ. One then notices that the n = 1 pseudopotentials, apart from a the difference in Vℓ=0, are much more reminiscent of those in n = 0 than of those for non-relativistic 2D electrons in the same LL n = 1 [see Fig. 24(b)]. If one considers polarized electrons, one therefore expects essentially the same strongly-correlated electronic phases in graphene for n = 1 as for n = 0 (Goerbig et al., 2006), as also corroborated by numerical studies for FQHE states (Apalkov and Chakraborty, Papi´c et al., 2006; 2009; 2006) (Poplavskyy et al., 2009; and electron-solid phases Zhang and Joglekar, 2007, 2008). Because the pseu- dopotentials (215) are systematically larger in n = 1 than in n = 0 (apart from the short-range component for ℓ = 0), the gaps of the FQHE states in n = 1 are larger than the corresponding ones in n = 0, as one may also see from numerical calculations (Apalkov and Chakraborty, 2006; Toke et al., 2006). Goerbig and Regnault, Toke and Jain, 2007; 2007; Toke et al., As much as we have emphasized the similarity between the n = 0 and n = 1 LLs in graphene, we need to stress the difference between the n = 1 LL in graphene as compared to n = 1 in non-relativistic 2D electron sys- tems. Remember that the physical phase diagram in the non-relativistic n = 1 LL is extremely rich; an intrigu- ing even-denominator FQHE has been observed at ν = 5/2 (Willett et al., 1987) and probably possesses non- Abelian quasiparticle excitations (Greiter et al., 1991; Moore and Read, 1991). Furthermore, a particular com- petition between FQHE states and electron-solid phases, which is characteristic of the non-relativistic n = 1 LL (Goerbig et al., 2003, 2004), is at the origin of the reen- trance phenomena observed in transport measurements (Eisenstein et al., 2002; Lewis et al., 2005). These phe- nomena are absent in the n = 0 LL, and the similarity between the n = 0 and the relativistic n = 1 LL thus leads to the expectation that FQHE states corresponding to the 5/2 state in non-relativistic quantum Hall systems and the above-mentioned reentrance phenomena are ab- sent in the n = 1 LL in graphene. This expectation has recently been corroborated in exact-diagonalization studies on the non-Abelian 5/2 state (Wojs et al., 2010). 4. External spin-valley symmetry breaking terms Before we consider the different phases due to electron- electron interactions, we start with an analysis of the different external effects,29 which are capable of lifting the four-fold spin-valley degeneracy. familiar external The probably most symmetry- breaking term is the Zeeman effect, which lifts the two- fold spin degeneracy while maintaining the SU(2) symme- try associated with the valley pseudospin. The size of the Zeeman splitting is given by the energy ∆Z ∼ 1.2B[T] K, for a g-factor that has been experimentally determined as g ∼ 2 (Zhang et al., 2006). If we adopt a compact eight- spinor notation to take into account the four different spin-valley components, in addition to the two sublattice components, the Zeeman term has the form ∆spin Z ∼ Ψ†(cid:0)1valley ⊗ 1AB ⊗ τ z K,↑ − ψA†K,↓ ψA spin(cid:1) Ψ K,↓ + ψA†K′,↑ ψA ∼ ψA†K,↑ + (A ↔ B), K′,↑ − ψA†K′,↓ ψA (216) ψA K′,↓ where the tensor product consists of the valley pseu- dospin (represented by the Pauli matrices τ µ valley and 1valley), the sublattice pseudospin (τ µ AB and 1AB), and the true spin (τ µ spin and 1spin). For a better understand- ing, we have given the explicit expression in terms of spinor components in the second line of Eq. (216). A possible valley-degeneracy lifting, that one could de- scribe with the help of a “valley Zeeman effect” similarly to Eq. (216), ∆valley Z ∼ Ψ†(cid:0)τ z valley ⊗ 1AB ⊗ 1spin(cid:1) Ψ, (217) 29 By external effects we mean those that are not caused by the mutual Coulomb repulsion between the electrons. 39 is more involved because there is no physical field that couples directly to the valley pseudospin, as suggested by the otherwise intuitive form (217). There have however been proposals that such an effect may be achieved with the help of strain-induced disordered gauge fields that mimic large-scale ripples (Meyer et al., 2007) and that yield an easy-plane anisotropy in n = 0 (Abanin et al., 2007a), similarly to the backscattering term (213) in higher LLs. Quite generally, a valley-degeneracy lifting may be achieved indirectly in the zero-energy LL n = 0 via fields that couple to the sublattice index. This is due to the fact that the components ψA K′,σ vanish as a consequence of the chiral properties, which identify the sublattice and the valley pseudospins in n = 0, as we have discussed in Sec. II.A.2. K,σ and ψB In order to illustrate this indirect lifting of the valley degeneracy, we consider the term (Haldane, 1988) MH = M Ψ† (1valley ⊗ τ z AB ⊗ 1spin) Ψ, (218) which breaks the lattice inversion symmetry and opens a mass gap at the Dirac point in the absence of a magnetic field. In the presence of a B-field, the LL spectrum (85) is modified by the term (218) and reads ǫλ,n;ξ = λsM 2 + 2 2v2 F l2 B n, (219) for n 6= 0, independent of the valley index ξ, whereas the fate of the n = 0 LL depends explicitly on ξ, ǫn=0;ξ = ξM, (220) such that the valley degeneracy is effectively lifted. No- tice, however, that due to the vanishing components ψA K,σ and ψB K′,σ, the mass term (218) is now indistinguishable (in n = 0) from the above-mentioned valley Zeeman term (217), Mn=0 H ∼ Ψ† (1valley ⊗ τ z AB ⊗ 1spin) Ψ valley ⊗ 1AB ⊗ 1spin(cid:1) Ψ, ∼ Ψ†(cid:0)τ z (221) whereas this is not the case for the LLs n 6= 0, where the valley degeneracy is only lifted by an explicit val- ley Zeeman effect. A mass term of the form (218) typically arises in the presence of a frozen out-of-plane phonon that yields a crumbling of the graphene sheet (Fuchs and Lederer, 2007). More recent studies have concentrated on a sponta- neous deformation of the graphene sheet due to frozen in-plane phonons that yield a Kekul´e-type distortion (Hou et al., 2010; Nomura et al., 2009). This distor- tion, which is associated with a characteristic wave vec- tor 2K and which therefore couples the two valleys, di- rectly breaks the valley degeneracy, via a term MK = Mx + My, with ∆x,y (222) Mx,y = 2 Ψ†(cid:16)τ x,y valley ⊗ 1AB ⊗ 1spin(cid:17) Ψ. x + ∆2 with the characteristic energy scale ∆kek =q∆2 Such a term yields the same energy spectrum (219) and (220) as the mass term (218) if one replaces M by ∆kek/2, y ≃ 2B[T] K, (Ajiki and Ando, 1995; Hou et al., 2010). No- tice that this energy scale is slightly larger than, but roughly on the same order as, the Zeeman energy scale. Finally, we mention another class of terms that break the spin-valley degeneracy and that have received recent interest in the framework of research on topological in- sulators [for recent reviews see (Hasan and Kane, 2010; Qi and Zhang, 2011)]. In an original work, Haldane ar- gued that a time-reversal-symmetry breaking term with an inhomogeneous flux distribution inside each hexagon opens a gap in a honeycomb lattice with zero magnetic field (Haldane, 1988). Most saliently, he showed that one may thus achieve a quantum Hall effect without an external magnetic field, a system that is now often re- ferred to as the “quantum anomalous Hall insulator” (Hasan and Kane, 2010; Qi and Zhang, 2011). A simi- lar situation arises when spin-orbit interactions are taken into account, which are of the form HSO = ∆SO 2 Ψ†(cid:0)τ z (223) valley ⊗ τ z AB ⊗ τ z spin(cid:1) Ψ and which provide again the same LL spectrum (219) and (220) if one replaces M by ∆SO/2 (Kane and Mele, 2005). In spite of the conceptually appealing prospect of the quantum spin Hall effect, which is revealed by this model because the spin orientation is locked to a particular valley index via the term (223), the associ- ated energy scale ∆SO ∼ 10 mK turns out to be van- ishingly small in graphene, whereas an extrinsic Rashba- type spin-orbit coupling in graphene can be on the order of 1 K (Min et al., 2006). 40 5. Hierarchy of relevant energy scales These energy scales associated with external fields need to be compared to the characteristic (bare) interaction energy e2/ǫlB ≃ 625(pB[T]/ε) K, which is, for experi- mentally accessible magnetic fields, much larger than ∆Z or ∆kek. As we have discussed in Sec. III.B.4, inter-band LL excitations screen the bare Coulomb interaction and yield a contribution to the dielectric constant. In the ab- sence of a quantizing magnetic field, we have seen that this dielectric constant is given by [see Eq. (176)] ε∞ = 1 + π 2 αG = 1 + π 2 e2 εvF , (224) where we remember that ε is the extrinsic dielectric con- stant of the surrounding medium. As one may notice from Fig. 18, the vacuum contribution Πvac(q) (thick dashed lines) is only marginally modified by the magnetic field, such that one may use e2/εε∞lB as an approxima- tion for the interaction-energy scale for graphene taking into account inter-band screening. The relevant energy scales are summarized in the table below for different values of the magnetic field, in com- parison with the interaction energy scales, taking into ac- count the effective dielectric constants for several widely used substrates from Tab. I. energy value for arbitrary B for B = 6 T for B = 25 T ∆Z ∆kek 1.2B[T] K 2B[T] K 7 K 12 K 30 K 50 K e2/εlB (bare) 625(pB[T]/ε) K (1550/ε) K (3125/ε) K e2/εε∞lB (vacuum) e2/εε∞lB (on SiO2) e2/εε∞lB (on h-BN) e2/εε∞lB (on SiC) 139pB[T] K 104pB[T] K 109pB[T] K 71pB[T] K ∆sb < (e2/εlB)(a/lB) < 1B[T] K 344 K 258 K 270 K 176 K < 6 K 694 K 521 K 543 K 355 K < 25 K Table II Energy scales at different magnetic fields. The first two lines show the energy scales associated with the major external symmetry- breaking fields (Zeeman and Kekul´e-type lattice distortion, ∆Z and ∆kek, respectively), which scale linearly in B. Below are shown the √B), the bare one with an unspecified dielectric constant and the ones for different substrates taking into interaction-energy scales (∝ account inter-band screening via the term ε∞ [Eq. (224)]. The last line yields the interaction-energy scale associated with the intrinsic symmetry breaking due to inter-valley coupling, discussed in Sec. V.A.2. In view of the above discussion, one may conclude that the SU(4)-symmetric part of the Coulomb interaction yields the leading energy scale in the problem of elec- trons in partially filled lower LLs, whereas external terms, such as the Zeeman effect or spontaneous lattice distor- tions, play a subordinate role. The remainder of this section is therefore concerned with a detailed discussion of strongly-correlated electron phases that are formed to minimize the Coulomb interaction. B. SU(4) Quantum Hall Ferromagnetism in Graphene A prominent example of the above-mentioned strongly- correlated phases is the generalized quantum Hall ferro- magnet. It arises in systems with a discrete internal de- gree of freedom described by an SU(N ) symmetry, such that each single-particle quantum state ψn,m occurs in N copies. Prominent examples are the non-relativistic quantum Hall systems when taking into account the elec- tronic spin σ =↑,↓ (N = 2) or bilayer quantum Hall sys- tems that consist of two parallel 2D electron gases, where the layer index may be viewed as a “spin” 1/2 [N = 2, or N = 4 if one also takes into account the physical spin (Ezawa and Hasebe, 2002; Ezawa et al., 2003)].30 In this sense, graphene may be viewed as an SU(4) quantum Hall system as a consequence of its four-fold spin-valley degeneracy. 1. Ferromagnetic ground state and Goldstone modes Quite generally, quantum Hall ferromagnetism arises when the filling factor, defined from the bottom of the LL,31 is an integer that is not a multiple of N (Arovas et al., 1999). From the point of view of the ki- netic Hamiltonian, one is thus confronted with a macro- scopic ground-state degeneracy. Even if one has an inte- ger filling factor, the situation is thus much more remi- niscent of the FQHE, i.e. the relevant energy scale is the Coulomb interaction, and the system may be described in the framework of the model (207) of interacting electrons in a single (relativistic) LL. For the moment, we consider that there are no symmetry-breaking terms, such as the backscattering term (212) or Zeeman-type terms that are discussed below in Sec. V.A.4. Qualitatively, one may understand the formation of a ferromagnetic ground state as a consequence of the repul- sive Coulomb interaction. In order to minimize this inter- action, the electrons prefer to form a state described by a maximally anti-symmetric orbital wave function that must then be accompanied by a fully symmetric SU(N ) spin wave function to satisfy an overall fermionic (anti- symmetric) wave function. In a usual metal with a fi- nite band dispersion, this ferromagnetic ordering (e.g. all 41 electrons in the spin-↑ states) is accompanied by a cost in kinetic energy – indeed, the Fermi energy for spin- ↑ electrons is increased whereas that of spin-↓ electrons is lowered. The competition between the gain in inter- action and the cost in kinetic energy defines the degree of polarization, i.e. how ferromagnetic the electrons ef- fectively are. In the quantum Hall effect, however, we are confronted with a highly degenerate LL that may be viewed as an infinitely flat band, such that the kinetic- energy cost for complete spin polarization is zero. As an example of an SU(N ) quantum Hall ferromag- net, one may consider the state FMi = k Yi=1 NB−1 Ym=0 c†m,ivaci, (225) which consists of k < N arbitrarily chosen completely filled subbranches [i ∈ {(K,↑), (K′,↑), (K,↓), (K′,↓)}, for the SU(N = 4) symmetry in graphene LLs], where we have omitted the LL index λn at the fermion operators to simplify the notation. The arbirariness in the choice of the SU(N ) spin subbranches may be viewed as a spon- taneous symmetry breaking that accompanies the ferro- magnetism. Indeed, the state (225) is no longer invariant under an SU(N ) rotation, but only under a rotation de- scribed by the subgroup SU(k)×SU(N − k), where the first factor indicates a symmetry transformation in the fully occupied subbranches i = 1, ..., k and the second fac- tor one in the empty subbranches i = k +1, ...,N . There- fore the quantum Hall ferromagnet (225) is associated with an order parameter with a spontaneous symmetry breaking described by the coset space SU(N )/SU(k) × SU(N − k) × U(1) ∼ U(N )/U(k) × U(N − k), where the additional U(1) is due to the phase difference between the occupied and the unoccupied subbranches (Arovas et al., 1999; Yang et al., 2006). The coset space, with its N 2 − k2 − (N − k)2 = 2k(N − k) complex generators, defines also the Gold- stone modes, which are nothing other than the k(N − k) spin-wave excitations of the ferromagnetic ground state (225).32 The number of the spin-wave modes may also have been obtained from a simple inspection into the LL- subbranch spectrum. Indeed, a spin wave can be de- scribed with the help of the components of the projected density operators (203), ¯ρij (q) = Xm,m′(cid:10)m(cid:12)(cid:12)e−iq·R(cid:12)(cid:12) m′(cid:11) c†m,icm′,j, (226) which represent coherent superpositions at wave vector q of excitations from the occupied subbranch j to the empty subbranch i. One has then k possibilities for the choice of the initial subbranch j and N − k for the final 30 For a review on non-relativistic multi-component systems see Ref. (Ezawa, 2000; Moon et al., 1995). 31 Remember that the filling factor in graphene is defined with re- spect to the center of the n = 0 LL. There is thus a shift of 2 in the filling factor as compared to the non-relativistic case. 32 The complex generators come in by pairs of conjugate operators, and each pair corresponds to one mode. Notice one, and one obtains therefore k(N − k) different spin- wave excitations, in agreement with the above group- theoretical analysis. that all the same dispersion, which may be calculated within a mean-field approximation (Alicea and Fisher, 2006; Doretto and Morais Smith, 2007; Kallin and Halperin, 1984; Yang et al., 2006), excitations have spin-wave Eq = hFM¯ρij (−q)Hn ¯ρij (q) − HnFMi = 2Xk vn(k) sin2(cid:18) q ∧ k l2 (cid:19) , which saturates at large values of q = q, vn(k), Eq→∞ = 2En 2 B X =Xk (227) (228) i.e. at twice the value of the exchange energy (209). This result is not astonishing insofar as the large-q limit cor- responds, as we have discussed in Sec. III.B.2 [see Eq. (167)], to an electron-hole pair where the electron is situ- ated far away from the hole. The energy (228) is therefore nothing other than the cost in exchange energy to create a spin-flip excitation, i.e. an electron with reversed spin and a hole in the ferromagnetic ground state. Because of the large distance between the electron and the hole in such an excitation and the resulting decoupled dynamics, one may be tempted to view this energy as the activa- tion gap of the quantum-Hall state at ν = k, but we see below in Sec. V.B.2 that there exist elementary charged excitations (skyrmions) that have, in some LLs, a lower energy than these spin-flip excitations. In the opposite limit of small wave vectors (qlB ≪ 1), one may not understand the excitation in terms of decou- pled holes and electrons, and the excitation can there- fore not contribute to the charge transport. A Taylor expansion of the sine in the spin-wave dispersion (227) yields the usual q2 dispersion of the spin-wave Goldstone modes, Eq→0 = in terms of the spin stiffness ρn s 2 q2l2 B, ρn s = 1 4π Xk vn(k)k2l2 B. (229) (230) One notices that the above results for the excitation en- ergies do not depend on the size of the internal symmetry group, but they can be derived within the SU(2) model of the quantum Hall ferromagnetism (Moon et al., 1995; Sondhi et al., 1993) – the enhanced internal symmetry of graphene (or of a general N -component system) affects only the degeneracies of the different modes. 2. Skyrmions and entanglement In addition to the above-mentioned spin-wave modes, the SU(N ) ferromagnetic ground state (225) is charac- (a) (b) 42 y y z z x x Figure 25 Excitations of the SU(2) ferromagnetic state. (a) Spin waves. Such an excitation can be continuously deformed into the ferromagnetic ground state [spin represented on the Bloch sphere (right) by fat arrow] – the grey curve can be shrinked into a sin- gle point (b) Skyrmion with non-zero topological charge [from Ref. (Girvin, 1999)]. The excitation consists of a a reversed spin at the origin z = 0, and the ferromagnetic state is recovered at large distances z/ζ → ∞. Contrary to spin-wave excitations, the spin explores the whole surface of the Bloch sphere and cannot be trans- formed by a continuous deformation into the majority spin (fat arrow). terized by a particular elementary excitation that con- sists of a topological spin texture, the so-called skyrmion (Sondhi et al., 1993). Similarly to a spin-wave in the limit qlB ≪ 1, the variation of the spin texture in a skyrmion excitation is small on the scale of the magnetic length lB, such that its energy is determined by the spin stiffness (230) in the small-q limit. Indeed, one may show that its energy is given by (Ezawa, 2000; Moon et al., 1995; Sondhi et al., 1993) Esk = 4πρn sQtop, (231) in terms of the topological charge Qtop, which may be viewed as the number of times the Bloch sphere is covered by the spin texture [see Fig. 25(b)] and which we discuss in more detail below. Skyrmions are the relevant elemen- tary excitations of the quantum Hall ferromagnetism if the energy (231) is lower than that of an added electron (or hole) with reversed spin, which is nothing other than the exchange energy (209), i.e. X . Whereas this condition is fulfilled in non-relativistic LLs only in the lowest one n = 0, skyrmions are the lowest-energy elementary excitations in the graphene LLs n = 0, 1 and 2 (Toke et al., 2006; Yang et al., 2006), as a consequence of the difference in the form factors. if Esk < En As in the case of the spin waves discussed above, the skyrmion energy is independent of the size of the internal symmetry group, and we will first illustrate the skyrmion texture in an effective SU(2) model, where the texture is formed only from states within the last occupied (k) and the first unoccupied (k + 1) LL subbranch. The skyrmion may then be described with the help of the wave function, in terms of the complex coordinate z = (x − iy)/lB, Sk,k+1i = 1 pζ2 + z2 (cid:0)z(cid:12)(cid:12)↑k (z)(cid:11) + ζ(cid:12)(cid:12)↓k+1 (z)(cid:11)(cid:1)(232) where ↑k (z)i corresponds to states in the subbranch k and ↓k+1 (z)i to those in k + 1, at the position z. One notices that at the origin z = 0 the “spin” associated with these two components is ↓ because the first component of Eq. (232) vanishes, whereas the spins are ↑ at z/ζ → ∞ (see Fig. 25), where the ferromagnetic ground state is then recovered. The parameter ζ plays the role of the skyrmion size, measured in units of lB – indeed, for z = ζ, both components are of the same weight and the spin is therefore oriented in the xy-plane. The skyrmion excitation (232) can also be illustrated on the so-called Bloch sphere on the surface of which the (normalized) spin moves (see Fig. 25). The angles (θ for the azimuthal and φ for the polar angle) of the spin orientation on the Bloch sphere correspond to the SU(2) parametrization ψi = cos(θ/2) ↑i + sin(θ/2) exp(iφ) ↓i, and the spin orientation at the circle z = ζ in the com- plex plane describes the equator of the Bloch sphere. The topology of the skyrmion excitation becomes apparent by the number of full circles the spin draws when going around the origin of the xy-plane on the circle z = ζ. More precisely, the topological charge Qtop is not defined in terms of such closed paths, but it is the number of full coverings of the Bloch sphere in a skyrmion excitation [Qtop = 1 in the example (232)]. Notice that a spin-wave excitation has a topological charge Qtop = 0 and corre- sponds to an excursion of the spin on the Bloch sphere that is not fully covered and that can then be reduced continuously to a single point describing the ferromag- netic ground state [Fig. 25(a)]. i.e. The above considerations may be generalized to sys- to SU(N ) tems with larger internal symmetries, quantum Hall ferromagnets. The state (232) is invariant under the SU(N ) subgroup SU(k − 1) × SU(N − k − 1), where the first factor describes a rotation of the occupied subbranches that do not take part in the skyrmion exci- tation and the second factor is associated with a symme- try transformation of the corresponding unoccupied sub- branches k + 2, ...,N . A similar group-theoretical analy- sis as the one presented in Sec. V.B.1 yields the number of residual symmetry transformations (Yang et al., 2006) 2k(N − k) + 2(N − 1), where the first term describes the Goldstone modes of the ferromagnetic ground state, and the second one corresponds to the N − 1 internal modes of the skyrmion excitation. In addition to the topological charge, skyrmions in quantum Hall systems carry an electric charge that co- incides, for ν = k with the topological charge. Indeed, the skyrmion state (232) describes an electron that is ex- pelled from the origin z = 0 in the ↑-component, and its net electric charge is therefore that of a hole. This means that skyrmions are excited when sweeping the filling fac- tor away from ν = k, and the net topological charge is 43 given by Qtot = ν−kNB. The number of internal modes is then Qtot(N − 1), in addition to one mode per charge that corresponds to a simple translation z → z + a of the excitation (Dou¸cot et al., 2008). As a consequence of the Coulomb repulsion, it is energetically favorable to form a state in which Qtot skyrmions of charge 1 are ho- mogeneously distributed over the 2D plane than a single defect with charge Qtot (Moon et al., 1995). A natural (semi-classical) candidate for the ground state of Qtot skyrmions is then a skyrmion crystal (Brey et al., 1995) that has recently been revisited in the framework of the SU(4) symmetry in graphene (Cot´e et al., 2007, 2008). In this case, the Qtot(N − 1) internal modes, which are dis- persionless zero-energy modes in the absence of electronic interactions or Zeeman-type symmetry-breaking terms, are expected to yield N − 1 Bloch bands of Goldstone- type, in addition to the Qtot translation modes that form a magnetic-field phonon mode of the skyrmion crystal with a characteristic ω ∝ q3/2 dispersion (Fukuyama, 1975). a. Skyrmions and activation gaps in graphene. Quite gen- erally, the activation gap in quantum Hall states is the energy required to create a quasi-particle–quasi-hole pair, in which the two partners are sufficiently well separated to contribute independently to the charge transport. In the framework of the quantum Hall ferromagnet, the activation gap may be viewed as the energy to create a skyrmion of topological charge Q = 1 and an anti- skyrmion of charge Q = −1 that are well-separated from each other such that one may neglect their residual in- teraction. The energy of such a skyrmion–anti-skyrmion pair is then given, in the absence of symmetry-breaking terms, by twice the energy in Eq. (231), ∆sym a = 8πρn s . (233) For graphene, the energies of the theoretical activation gaps for n = 0 and n = 1 are shown in the table below. n = 0 n = 1 1 activation gap arbitrary value of B 2 p π 400(pB[T]/εε∞) K 32 p π 175(pB[T]/εε∞) K e2 7 2 2 e2 εε∞lB εε∞lB Table III Theoretical estimates for the activation gaps in the n = 0 and 1 graphene LLs due to well-separated skyrmion–anti- skyrmion pairs. For further illustration, we consider the scenario in which the Zeeman effect is the only SU(4)-symmetry breaking term.33 Due to the Zeeman effect, spin-↓ elec- 33 The energetic argument remains valid in the case where the dom- inant term is a valley-Zeeman effect if one interchanges the role of the spin and the valley pseudospin. trons are energetically favored. If only one spin-valley branch of a particular LL is filled (k = 1), the spin mag- netization of the spin-valley ferromagnet is preferentially oriented in this direction whereas the valley polarization may point in any direction. The activation gap would then be dominated by valley (anti-)skyrmions with no reversed physical spin such that one would not expect any dependence of the gap on the in-plane component of the magnetic field, in agreement with the experimental findings (Jiang et al., 2007b). The situation is different when both valley branches of the spin-↓ branch are occupied; an excitation of the SU(4) ferromagnet with a full spin polarization would then nec- essarily comprise reversed spins, and the corresponding Zeeman energy must be taken into account in the energy of the (spin) skyrmion–anti-skyrmion pair (233), ∆Z a = 8πρn s + 2Nrs∆Z , (234) where Nrs ∼ ζ2 is the number of reversed spins in a sin- gle (anti-)skyrmion. Notice that this number depends on the competition between the Zeeman effect itself, which tries to reduce the skyrmion size ζ, and the cost in ex- change energy due to the strong variation in small tex- tures (Moon et al., 1995; Sondhi et al., 1993).34 The en- ergy of a skyrmion–anti-skyrmion pair in the spin channel (with two completely filled valley sublevels) is therefore larger than that (233) of a pair in the valley channel when only one valley subbranch of the LL is completely filled. Notice that this energy increase may even be sig- nificant for large skyrmions because of the larger num- ber of reversed spins. As a thumbrule, the stability of a quantum Hall state is proportional to the activation gap, which has in the present case been identified with the skyrmion–anti-skyrmion energy and which is dominated by the Coulomb interaction energy. Additional external symmetry-breaking terms, such as those discussed in Sec. V.A.4, may enhance this stability although they provide only a small correction to the activation energy. In an experimen- b. Spin-valley entanglement in graphene. tal measurement, one typically has not direct access to the full SU(4) spin that describes the internal degrees of freedom in graphene LLs, but only to the SU(2) part associated with the physical spin, e.g. in a magnetiza- tion measurement. It is therefore useful to parametrise the SU(4) spin in a manner such as to keep track of the two SU(2) copies associated with the physical spin and with the valley pseudospin, respectively. This may be achieved with the so-called Schmidt decomposition of the 34 This energy cost may be evaluated from a gradient expansion of the energy in the magnetization fields. At leading order, one ob- tains, however, a non-linear sigma model that is scale-invariant, such that the energy cost must be calculated at higher orders (Moon et al., 1995). 44 (a) z θ S (b) y φ S x θ I z φ I y x spin pseudospin (c) z α=0 α=0 y β x entanglement Figure 26 From Ref. (Dou¸cot et al., 2008); Bloch spheres for entangled spin-pseudospin systems. Bloch sphere for the spin (a), pseudospin (b), and a third type of spin representing the entanglement (c). In the case of spin-pseudospin entanglement ( cos α 6= 1), the (pseudo)spin-magnetizations explore the interior of their spheres, respectively (black arrows). four-spinor Ψ(z)i = cos α 2 ψSiψIi + sin α 2 eiβχSiχIi , (235) where α and β are functions of the complex position z, and the local two-component spinors ψSi, χSi, ψIi, and χIi are constructed according to ψi = cos θ χi = − sin θ 2 eiφ! ! . 2 e−iφ cos θ 2 2 sin θ and The angles θ and φ define the usual unit vector (236) n(θ, φ) = (sin θ cos φ, sin θ sin φ, cos θ) , (237) which explores the surface of the Bloch sphere depicted in Fig. 25. Notice that one has two Bloch spheres, one for the unit vector n(θS, φS) associated with the spin angles θS and φS and a second one for n(θI , φI ) for the valley- pseudospin angles θI and φI (see Fig. 26). In addition, one may associate a third Bloch sphere with the angles α and β that describe the degree of factorizability of the wave functions and thus the degree of entanglement be- tween the spin and the valley pseudospin (Dou¸cot et al., 2008). With the help of the Schmidt decomposition (235), one obtains immediately the reduced density matrices for the spin and the valley-pseudospin sectors ρS = TrI (ΨihΨ) = cos2 α ρI = TrS (ΨihΨ) = cos2 α 2 ψSihψS + sin2 α 2 ψIihψI + sin2 α 2 χSihχS , 2 χIihχI , (238) respectively, and the local spin and valley-pseudospin densities are simply ma S = Tr(ρSSa) = cos αhψSSaψSi = cos α na(θS, φS) (239) and mµ I = Tr(ρI I µ) = cos αhψII µψIi = cos α nµ(θI , φI ), (240) where Sa and I µ represent the components of the spin and valley-pseudospin operators, respectively [see Eq. (208)]. One notices from these expressions that, for the case α 6= 0 or π (i.e. cos2 α < 1), the local (pseudo)spin densities are no longer normalized, but they are of length mS/I2 = cos2 α. Thus, in a semiclassical picture, the (pseudo)spin dynamics is no longer restricted to the sur- face of the Bloch sphere, but explores the entire volume enclosed by the sphere (Fig. 26) (Dou¸cot et al., 2008). This result indicates that one may be confronted, in the case of full entanglement (e.g. α = π/2), with an SU(4) quantum Hall ferromagnet the (spin) magnetization of which completely vanishes, as one would naively expect for an unpolarized state. 3. Comparison with magnetic catalysis An alternative scenario proposed for the degeneracy lifting in n = 0 is that of the magnetic catalysis (Ezawa, 2007; Gorbar et al., 2008; Gusynin et al., 2006; Herbut, 2007b, 2008), which was discussed even before the dis- covery of graphene (Gorbar et al., 2002; Khveshchenko, 2001). According to this scheme, the Coulomb interac- tion spontaneously generates a mass term for the (origi- nally massless) 2D electrons once the magnetic field in- creases the density of states at zero energy by the for- mation of the highly degenerate n = 0 LL. As a con- sequence of this mass generation, the particles condense in a state of coherent particle-hole pairs (excitonic con- densation). The effect is at first sight reminiscent of the excitonic condensation at ν = 1 in non-relativistic bi- layer quantum Hall systems (Ezawa and Iwazaki, 1993; Fertig, 1989; Wen and Zee, 1992a). Its superfluid behav- ior gives rise to a zero-bias anomaly in the tunneling con- ductance between the two layers (Spielman et al., 2000) as well as to a simultaneous suppression of the longitudi- nal and the Hall resistance in a counterflow experiment (Kellogg et al., 2004; Tutuc et al., 2004). The bilayer ex- citonic condensate may be described as an easy-plane quantum Hall ferromagnet (Moon et al., 1995), where the spin mimics the layer index. The origin of this easy- plane anisotropy stems from the difference in the inter- actions between electrons in the same layer as compared to the weaker one for electron pairs in different layers. This comparison with non-relativistic 2D electrons in bilayer systems indicates that there may exist a close re- lation between the quantum Hall ferromagnetism and the scenario of the magnetic catalysis also in graphene in a strong magnetic field. Notice, however, that the excitonic 45 state in graphene is not in the same universality class as that of the quantum-Hall bilayer – in the latter case the symmetry of the (interaction) Hamiltonian is U(1) as a consequence of the easy-plane anisotropy, and the sym- metry breaking is associated with a superfluid mode that disperses linearly with the wave vector, ω ∝ q. In con- trast to this system, the interaction Hamiltonian (207) has the full SU(4) symmetry, and even for a sufficiently strong Zeeman effect, the symmetry is quite large with SU(2)↑ × SU(2)↓, i.e. each spin projection ↑ and ↓ is gov- erned by the residual SU(2) valley symmetry and has the characteristic ω ∝ q2 pseudospin-wave modes. The connection between the two scenarios becomes transparent within a mean-field treatment of the Coulomb interaction Hamiltonian. The quantum Hall ferromagnetic states discussed in the previous subsec- tions may be described equivalently with the help of the mean-field order parameters DΨ†(cid:16)τ ν valley ⊗ 1AB ⊗ τ µ spin(cid:17) ΨE , (241) where Ψ denotes the same eight-spinor as in Sec. V.A.4. Remember that a pure spin quantum Hall ferromagnet is obtained for τ ν valley = 1valley, whereas a pure valley- pseudospin ferromagnet is described by an order parame- ter (241) with τ µ spin = 1spin. The remaining order param- eters describe states with a certain degree of spin-valley entanglement, as discussed above. Notice, however, that the choice of order parameters is not restricted to those in Eq. (241). Indeed, one may also opt for a mean-field calculation of the interaction Hamiltonian with the order parameters (Gorbar et al., 2008; Gusynin et al., 2006) (242) (243) and valley ⊗ τ z Ms =(cid:10)Ψ†(cid:0)τ z Mt =(cid:10)Ψ† (1valley ⊗ τ z AB ⊗ 1spin(cid:1) Ψ(cid:11) AB ⊗ 1spin) Ψ(cid:11) , which describe mass gaps. Indeed, we already encoun- tered a term of the form (243) in Sec. V.A.4 and showed that it lifts the valley degeneracy of the n = 0 LL. Whereas such a term arises naturally in the context of an out-of-plane distortion of the graphene lattice, here, it is generated dynamically via the repulsive electron- electron interaction. The difference between the two mass terms Ms and Mt stems from the residual sym- metry of the SU(2)σ groups. The term (242), which may be viewed as a singlet mass term explicitly breaks this symmetry, whereas the term (243) has been coined triplet mass (Gorbar et al., 2008; Gusynin et al., 2006). In Sec. V.A.4, we have argued that mass terms of the above form only lift the valley degeneracy in the zero- energy LL n = 0, whereas they simply renormalize the LL energy for n 6= 0. Furthermore we have seen that as a consequence of the vanishing spinor components ψA and ψB K′,σ, the mass term Mt is indistinguishable, in K,σ n = 0, from a valley-pseudospin ferromagnetic state, Mn=0 t = (cid:10)Ψ† (1valley ⊗ τ z ∼(cid:10)Ψ†(cid:0)τ z AB ⊗ 1spin) Ψ(cid:11) valley ⊗ 1AB ⊗ 1spin(cid:1) Ψ(cid:11) , (244) whereas the singlet mass term simply renormalizes the overall chemical potential, Mn=0 s valley ⊗ τ z = (cid:10)Ψ†(cid:0)τ z ∼(cid:10)Ψ† (1valley ⊗ 1AB ⊗ 1spin) Ψ(cid:11) . (245) AB ⊗ 1spin(cid:1) Ψ(cid:11) These arguments lead to the conclusion that the mag- netic catalysis in n = 0, i.e. the spontaneous genera- tion of a mass gap due to electron-electron interactions, may be fully described in the framework of the SU(4) quantum Hall ferromagnetism. Furthermore, the recent observation of a fully lifted spin-valley degeneracy in the n = 1 graphene LL (Dean et al., 2011) is naturally under- stood in the framework of quantum Hall ferromagnetism, whereas the mass terms (242) and (243), obtained from magnetic catalysis, would not provide a fully lifted spin- valley degeneracy. 4. The quantum Hall effect at ν = ±1 and ν = 0 Before discussing the experimental results on the quan- tum Hall effect, a clarification on the filling factor is re- quired. In the preceding parts of this section, which were concerned with general aspects of the quantum Hall fer- romagnet in LLs with internal degrees of freedom, the filling factor ν = k has been defined with respect to the bottom of the partially filled LL. However, in graphene, this is at odds with the natural definition of the filling factor (101) in terms of the electronic density measured from the charge neutrality point in undoped graphene – a zero filling factor therefore corresponds to two completely filled spin-valley subbranches (k = 2) of the n = 0 LL. In the remainder of this section, we therefore make a clear distinction between the two filling factors, and ν denotes the filling of the n = 0 LL measured from the bottom of the level, whereas the natural filling factor (101) is from now on denoted by νG. Explicitly, the relation between the two filling factors reads ν = νG + 2. (246) Early transport measurements in exfoliated graphene on a SiO2 have revealed broken spin-valley-symmetry states at at νG = 0,±1 and ±4 (Jiang et al., 2007b; Zhang et al., 2006), where the latter corresponds to the LLs ±1. More recent experiments on exfoliated graphene on a h-BN substrate have furthermore revealed quantum Hall states at νG = ±3 (Dean et al., 2011), thus com- pleting the full resolution of the spin-valley quartet, not only in n = 0, but also in ±1. The observed states may generally be understood in the framework of the quantum Hall ferromagnetism, but the understanding of the situation at νG = 0 requires 46 an additional consideration of the subleading external symmetry-breaking terms discussed in Sec. V.A.4. The two-stage picture, which we adopt here based on the above discussions, may be summarized as follows. (a) The quantum Hall ferromagnetic states are formed to minimize the leading energy given by the Coulomb in- teraction. However, because of the (approximate) SU(4) symmetry of the interaction, the orientation of the quan- tum Hall ferromagnets is not fixed – a polarization in the spin channel is as probable as one in the valley chan- nel, and this yields the high degeneracy of the Goldstone modes described in Sec. V.B.1. (b) Therefore, in spite of the small energy scale of the external fields, the latter are relevant for the orientation of the ferromagnets and for the degeneracy lifting of the Goldstone modes. a. The quantum Hall effect at νG = ±1. For νG = −1, only one spin-valley branch is completely filled by elec- trons.35 The Zeeman effect would give a small ener- getic advantage to spin-↓ electrons, such that the two spin Goldstone modes associated with collective excita- tions to the spin-↑ acquire a q = 0 gap, given by ∆Z . In contrast to the spin excitations, the Goldstone mode, which couples the two valleys in the spin-↓ branch of n = 0, remains gapless, and the ground state may thus be viewed as a valley-pseudospin ferromagnet in the spin- ↓ branch. The activation gap would be given by Eq. (233) for pseudospin skyrmion–anti-skyrmion pairs, and its associated scaling e2/εlB ∝ √B has indeed been ob- served experimentally (Jiang et al., 2007b). The resid- ual valley SU(2) symmetry may be broken by the lat- tice distortions, which we have discussed in Sec. V.A.4. Whereas an out-of-plane lattice distortion would yield a gapped valley-pseudospin-wave mode, a Kekul´e-type in- plane distortion orients the pseudospin ferromagnet in the XY -plane, associated with a gapless U(1) superfluid mode (Nomura et al., 2009). Notice that the lattice dis- tortion, characterized by the energy scale ∆kek is not in competition, at νG = ±1, with the Zeeman effect, such that the resulting ferromagnetic state is the same for ∆Z > ∆kek as for ∆Z < ∆kek. In the remainder of this section we restrict the discussion of the valley- pseudospin degeneracy lifting to in-plane distortions that seem to be energetically more relevant than out-of-plane distortions, but the overall picture remains unchanged if the latter are more relevant. b. The quantum Hall effect at νG = 0. The situation is more subtle at νG = 0, where it is not possible to fully po- larize both the spin and the valley pseudospin and where the Zeeman effect is in competition with a lattice distor- 35 For νG = +1, the same arguments apply in terms of holes due to particle-hole symmetry. (a) y g r e n E ∆ kek ∆ Z K, K’, K, K’, (b) y g r e n E y ∆ Z ∆ kek K, K, K’, K’, y Figure 27 Possible scenarios for the lifted spin-valley degeneracy at νG = 0. (a) ∆Z > ∆kek in the bulk. When approaching the edge, the energy difference between the two valleys increases drastically, and two levels (K ′, ↑) and (K, ↓) cross the Fermi energy at the edge depicted by the dashed line (Quantum Hall state). (b) ∆kek > ∆Z in the bulk. The K subbranches are already located above the Fermi energy, and those of K ′ below, such that the energy difference is simply increased when approaching the edge with no states crossing the Fermi energy (Insulator). tion that orients the valley pseudospin. For ∆Z > ∆kek, it is favorable to fill both valley sublevels of the spin- ↓ branch and the resulting state is a spin ferromagnet with gapped spin-wave excitations. For ∆Z < ∆kek, a pseudospin-ferromagnetic ground state is favored with both spin sublevels completely filled. The two different situations are depicted in Fig. 27. Most saliently, the two phases reveal drastically different transport properties as one may see from their behavior at the sample edges. The electronic behavior at the edges may be described within a model of electron confinement, in which the sample edge is described via a mass confinement term M (y)τ z AB in the Hamiltonian, which has the symmetry of the term (218) or else, in n = 0, that of a valley- Zeeman term (217), as argued in Sec. V.A.4. The pa- rameter M (y) is zero in the bulk and increases drasti- cally at the edge at a certain value of the coordinate y.36 Although the model is a simplification to treat the graphene edges in the continuum description of the Dirac equation, a more sophisticated treatment that takes into account the geometry of the edges yields, apart from a fine structure of the levels at the edge, qualitatively simi- lar results (Brey and Fertig, 2006). The mass term M (y) modifies the valley coupling due to the lattice distortion kek + M (y)2, which and yields a y-dependent term p∆2 therefore equally diverges at the sample edge.37 These preliminary considerations on the gap behav- ior at the edges allow us to appreciate the difference in the expected electronic transport between a spin ferro- magnet and a valley-pseudospin ferromagnet at νG = 0. Indeed, for ∆Z > ∆kek, one obtains a quantum Hall 36 For the present argument, we consider translation invariance in the x-direction. 37 In the case of an out-of-plane distortion, the term M (y) simply (217)], but the adds up to the energy scale ∆valley physical picture remains unaltered. [see Eq. Z 47 state at νG = 0 that is characterized by a bulk gap as- sociated with two counter-propagating edge states [Fig. 27(a)]. In the bulk, where M (y) = 0, both valley sub- levels of the spin-↓ branch are filled (spin ferromagnet). When approaching the edge, however, the energy term p∆2 kek + M (y)2 is enhanced by the rapidly increasing contibution from M (y), and the (K,↑) level eventually crosses the (K′,↓) one at the edge at the Fermi energy (Abanin et al., 2007b). This situation corresponds to a quantum Hall state with a bulk insulator and (two counter-propagating) conducting channels. In contrast to usual quantum Hall states, the edge states are not chi- ral, but the chiralitiy, i.e. the transport direction, of each channel is linked to its spin orientation.38 The quantum Hall state therefore remains stable unless magnetic impu- rities couple the two chiralities (Shimshoni et al., 2009). One notices furthermore a change in the spin polariza- tion at the edge; whereas the spin polarization in the bulk is complete, the system becomes spin unpolarized at the edge. If one takes into account the exchange inter- action, the change in the polarization takes place over a certain distance, and the conducting properties may be described in terms of spin-carrying one-dimensional edge excitations (Shimshoni et al., 2009). and an increase of p∆2 In the opposite limit with ∆Z < ∆kek in the bulk [Fig. 27(b)], the system at νG = 0 is already valley-polarized, kek + M (y)2 when approaching the edge does not induce a level crossing at the Fermi energy. Thus, there are no zero-energy states at the edge, and the system would be insulating both in the bulk and at the edge. From an experimental point of view, it is not fully set- tled which of the two phases describes the state at νG = 0. Whereas early experiments in exfoliated graphene on SiO2 samples were discussed in the framework of a dom- inant Zeeman effect (Abanin et al., 2007b; Jiang et al., 2007b; Zhang et al., 2006), more recent experiments at very large magnetic fields (Checkelsky et al., 2008, 2009) and on suspended graphene samples with increasing mo- bility (Du et al., 2009) favor the insulating scenario of Fig. 27(b) with a dominant valley degeneracy lifting. Especially the high-field measurements hint at an easy- plane or XY (valley-pseudospin) ferromagnetic ground state because the transition between the metallic and the insulating state is reminiscent of a Kosterlitz-Thouless phase transition (Kosterlitz and Thouless, 1973) if one replaces the temperature by the magnetic field as the parameter driving the transition (Checkelsky et al., 2008, 2009; Hou et al., 2010; Nomura et al., 2009). However, it has also been argued that this effect may be understood within the above scenario of a Zeeman-dominated quan- tum Hall ferromagnet in the bulk, in the framework of a Luttinger-liquid description of the domain wall seperat- 38 These helical edges are the signature of a quantum spin Hall effect (Hasan and Kane, 2010; Qi and Zhang, 2011). ing the polarized from the unpolarized region at the edge (Shimshoni et al., 2009). One notices that both the Zeeman effect and the Kekul´e-type lattice distortion are very close in energy II) such that one may speculate that other (see Tab. effects, as e.g. impurities, strongly influence the forma- tion and the orientation of the quantum Hall ferromag- net. Further experimental and theoretical studies there- fore seem to be necessary to clearly identify the leading symmetry-breaking mechanisms, which need not be uni- versal, in the zero-energy LL at νG = 0 and ±1. We finally mention scanning-tunneling spectroscopic results for the level splitting at νG = 0 that were per- formed on graphene on a graphite substrate (Li et al., 2009a). Although a gap has been observed as one may expect in the framework of the above scenario, it satu- rates as a function of the magnetic field. This is in dis- agreement with both the √B-behavior of an interaction- dominated gap as well as with the linear dependence of the Zeeman or lattice-distortion effects. A probable ori- gin of this gap is the commensurability of the graphene lattice with the graphite substrate that may break the in- version symmetry between the two sublattices by a term of the type (218). The coupling to the substrate being es- sentially electrostatic, one would then expect no or only a weak magnetic-field dependence of the splitting, as ob- served in the experiment (Li et al., 2009a). C. Fractional Quantum Hall Effect in Graphene The most salient aspect of strongly-correlated 2D elec- trons in partially filled LLs is certainly the FQHE, which is due to the formation of incompressible liquid phases at certain magical values of the filling factor. As we have already argued in Sec. V.A.3 on the basis of the pseu- dopotentials, the FQHE is expected to be present in the graphene LLs n = 0 and n = 1, and the main differ- ence with respect to non-relativistic 2D electron systems should arise from the internal SU(4) symmetry [for a re- cent theoretical review see Ref. (Papi´c et al., 2009)]. On the experimental level, recent progress in the fabri- cation of high-mobility samples, e.g. via current anneal- ing (Bolotin et al., 2008; Du et al., 2008), has allowed for the observation of several FQHE states in graphene. The first observations of a state at νG = ±1/3 were reported in 2009 on current-annealed suspended graphene samples in the two-terminal configuration (Bolotin et al., 2009; Du et al., 2009).39 More recently, in 2010, the FQHE has also been observed in the four-terminal geometry, which allows for the simultaneous measurement of the longi- tudinal and the Hall resistance, in suspended graphene (Ghahari et al., 2011) and on graphene on a h-BN sub- 39 There are also some weak indications for FQHE states at other filling factors than νG = ±1/3 in these samples. 48 strate (Dean et al., 2011). Before commenting in more detail on these first exper- imental results (indeed, this part of graphene research has just started), we introduce the theoretical four- component or SU(4) picture of the FQHE in graphene, in terms of generalized Halperin wave functions. These wave functions, which may be viewed as multi-component generalizations of Laughlin’s wave function, provide the natural framework for the description of the phenomenon in view of the model of electrons restricted to a single rel- ativistic LL (Sec. V.A) 1. Generalized Halperin wave functions The theoretical study of the FQHE is intimitely related to trial N -particle wave functions. In 1983, Laughlin pro- posed a one-component wave function (Laughlin, 1983), φL m ({zk}) = N Yk<l (zk − zl)m e− PN k zk2/2, (247) which allows for an understanding of incompressible FQHE states at the filling factors ν = 1/m that are de- termined by the exponent m for the particle pairs k, l in Eq. (247). The variable zk = (xk−iyk)/lB is the complex position of the k-th particle, and the form of the Laughlin wave function (247) is dictated by the analyticity condi- tion for wave functions in the lowest LL.40 Furthermore, the exponent m must be an odd integer as a consequence of the fermionic statistics imposed on the electronic wave function. Even if Eq. (247) describes only a trial wave function, one can show that it is the exact ground state for a class of model interactions that yield, with the help of Eq. (215), the pseudopotentials (Haldane, 1983) and Vℓ > 0 for ℓ < m Vℓ = 0 for ℓ ≥ m. (248) Although the Coulomb interaction does not fulfill such strong conditions, the pseudopotentials decrease as 1/√m for large values of m. Because the incompressible ground state is protected by a gap that is on the order of V1, one may view the pseudopotentials Vℓ≥m as an irrelevant perturbation that does not change the nature of the ground state. Indeed, exact-diagonalization calcu- lations have shown that, for the most prominent FQHE at ν = 1/3, the overlap between the true ground state and the Laughlin state (247) is extremely large (> 99%) (Fano et al., 1986; Haldane and Rezayi, 1985). Soon after Laughlin’s original proposal, Halperin gen- eralized the wave function (247) to the SU(2) case of 40 The lowest-LL condition of analytic wave functions may seem a very strong restriction when discussing FQHE states in higher LLs. However, the model (207) indicates that all LLs can be treated as the lowest one, n = 0, if the interaction potential is accordingly modified. We adopt this point of view here. electrons with spin, in the absence of a Zeeman effect (Halperin, 1983) – one has then two classes of particles, N↑ spin-↑ and N↓ spin-↓ particles, which are described by the complex positions z(↑) , respectively. In k↑ the (theoretical) absence of interactions between elec- trons with different spin orientation, the most natural ground-state candidate would then be a simple product of two Laughlin wave functions (247), and z(↓) k↓ φL m↑(cid:16)nz(↑) k↑ o(cid:17) φL m↓(cid:16)nz(↓) k↓ o(cid:17) , (249) one for each spin component with the exponents m↑ and m↓, respectively, that need not necessarily be identical. Inter-component correlations may be taken into account by an additional factor N↑ Yk↑ N↓ Yk↓ (cid:16)z(↑) k↓ (cid:17)n k↑ − z(↓) , (250) where the exponent n can now also be an even integer be- cause the fermionic anti-symmetry condition is concerned only with electrons in the same spin state. Halperin’s idea is easily generalized to the case of more than two components, and the corresponding trial wave function for an SU(N ) quantum Hall system with N com- ponents reads (Goerbig and Regnault, 2007) ψSU(N ) m1,...,mN ;nij = φL m1,...,mN φinter nij , (251) in terms of the product Nj φL m1,...,mN = N Yj=1 Ykj <lj(cid:16)z(j) kj − z(j) lj (cid:17)mj e− PNj=1 P Nj kj =1 z(j) kj 2/4 of N Laughlin wave functions and the term kj (cid:17)nij Ykj (cid:16)z(i) ki − z(j) φinter nij = Yki Yi<j N Nj Ni (252) , (253) which describes inter-component correlations. As in the case of Halperin’s two-component wave function (Halperin, 1983), the exponents mj must be odd inte- gers for fermionic particles whereas the exponents nij may also be even integers. These exponents define a symmetric N × N matrix M = nij, where the diago- nal elements are nii ≡ mi. This exponent matrix encodes the statistical properties of the quasi-particle excitations, such as their (fractional) charge and their statistical an- gle (Wen and Zee, 1992a,b). Moreover, the exponent matrix M determines the com- ponent densities ρj – or, equivalently, the component fill- ing factors νj = ρj/nB, ν1 ... νN     = M−1  1 ... 1 ,   (254) 49 where ν = ν1 + . . . + νN is the total filling factor mea- sured from the bottom of the lowest LL. Notice that Eq. (254) is only well-defined if the exponent matrix M is in- vertible. In this case, all component filling factors νj are completely determined, whereas otherwise some of the component fillings remain unfixed, e.g. ν1 and ν2 for the sake of illustration, although the sum of them (ν1 + ν2) is fixed. This is nothing other than a consequence of the underlying ferromagnetic properties of the FQHE state that, similarly to the states at ν = k discussed in Sec. V.B, are described by subgroups of SU(N ). Finally, we notice that not all SU(N ) wave functions (251) describe incompressible quantum liquids with a ho- mogeneous charge density for all components. A general- ization of Laughlin’s plasma picture, according to which the modulus square of the trial wave function corre- sponds to the Boltzmann weight of a classical 2D plasma (Laughlin, 1983), shows that all eigenvalues of the ex- ponent matrix M must be positive (or zero for states with ferromagnetic order). Otherwise, some of the differ- ent components phase-separate in the 2D plane because the inter-component repulsion between them exceeds the intra-component repulsion (de Gail et al., 2008). 2. The use of generalized Halperin wave functions in graphene These general considerations allow us to define the framework for a basic description of the FQHE in graphene where the SU(4) spin-valley symmetry imposes N = 4. Four-component Halperin wave functions are therefore expected to play an equally central role in the description of the graphene FQHE as Laughlin’s in a one- component or Halperin’s in two-component systems. In the remainder of this section, we attribute the four spin- valley components as 1 = (↑, K), 2 = (↑, K′), 3 = (↓, K), and 4 = (↓, K′). In a first a. Fractional SU(4) quantum Hall ferromagnet. step, we consider a four-component Halperin wave func- tion in which all components are equal (odd) integers, mj = nij = m, regardless of whether they describe intra- or inter-component correlations. One obtains then a completely anti-symmetric orbital wave function that is accompanied by a fully symmetric SU(4) spin-valley wave function. As we have argued in Sec. V.B.1, this situation repre- sents precisely a perfect SU(4) quantum Hall ferromagnet – indeed, for m = 1, the generalized Halperin wave func- tion (251) is nothing other than the orbital wave function of the state at ν = 1, i.e. when one of the subbranches is completely filled. The SU(4) symmetry is then sponta- neously broken, and the group-theoretical analysis pre- sented in Sec. V.B.1 yields 3 degenerate Goldstone modes that are generalized spin waves. The situation is exactly the same for any other odd exponent m, but the orbital wave function (251) is then a Laughlin wave function (247) in terms of the particle positions zk regardless of their internal index j = 1, ..., 4. The ferromagnetic properties of these wave functions may be described by the same equations as the spin-wave and skyrmion modes derived in Sec. V.B if one takes into ac- count a renormalization of the spin stiffness, as it has been discussed extensively in the literature for SU(2) quantum Hall ferromagnets (Ezawa, 2000; Moon et al., 1995; Sondhi et al., 1993). States described by such a wave function are ground-state candidates for the filling factors ν = 1/m, which correspond to the graphene fill- ing factors [see Eq. (246)] νG = −2 + 1/m or hole states at νG = 2 − 1/m. There are now two different manners to break the in- ternal SU(4) symmetry explicitly. The simplest one is the same as for the quantum Hall ferromagnetism at νG = 0 or ±1, in terms of external symmetry-breaking fields such as those discussed in Sec. V.A.4. However, one may also change some of the exponents in the gen- eralized Halperin wave function (251), in which case one also changes the filling factor. One may for instance con- sider the [m; m− 1, m] wave function with mj = m for all j, n13 = n24 = m − 1, and n12 = n14 = n23 = n34 = m, which correspond to a filling factor41 ν = 2 2m − 1 or νG = −2 + 2 2m − 1 . (255) Indeed, the difference in the inter-component exponents explicitly breaks the spin-valley symmetry – electrons in different valleys are more weakly correlated (with an ex- ponent m − 1) than electrons in the same valley (expo- nent m), regardless of their spin orientation. As a con- sequence, the filling factors in each of the two valleys, νK = ν1 + ν3 and νK′ = ν2 + ν4, respectively, are fixed, νK = νK′ = 1/(2m − 1), and one may view the wave function as a state with ferromagnetic spin ordering, but that is valley-pseudospin unpolarized. Alternatively, the [m; m−1, m] wave function may be interpreted as a tensor product of an SU(2) Halperin (m, m, m − 1) pseudospin- singlet wavefunction (Halperin, 1983) and a completely symmetric (ferromagnetic) two-spinor that describes the physical spin. The relevance of the [m; m − 1, m] wave function, with m = 3 (ν = 2/5) has been corroborated in recent exact-diagonalization studes, both in the graphene LLs n = 0 and n = 1 (Papi´c et al., 2009; Toke and Jain, 2007). The SU(4) spin-valley symmetry is fully broken, e.g., in the case of the [m; m − 1, m − 1] wave function with all mj = m and off-diagonal nij = m − 1. This wave function, which describes a state at ν = 4 4m − 3 or νG = −2 + 4 4m − 3 , (256) 41 We only discuss electronic states here, but the arguments are equally valid for the particle-hole symmetric states at νG = 2 − 2/(2m − 1). 50 may be viewed as an SU(4) singlet where the filling fac- tors of all spin-valley components are 1/(4m− 3). Exact- diagonalization calculations for N = 4 and 8 particles have shown that the [m; m− 1, m− 1] wave function with m = 3 (at ν = 4/9) describes to great accuracy the ground state for a Coulomb interaction (205), with over- laps ON =8 = 0.992 in n = 0 and ON =8 = 0.944 and in the n = 1 graphene LL (Papi´c et al., 2009). b. A route to understanding the graphene FQHE at νG = ±1/3. The discussion of the above-mentioned states was based on the understanding acquired from quantum Hall systems in semiconductor heterostructures, where the fill- ing factor is defined with respect to the bottom of the n = 0 LL. First experimental observations, however, in- dicated a prominent FQHE at νG = ±1/3, which cor- responds to two completely filled spin-valley sublevels of the graphene n = 0 LL, and a third one that is 1/3 filled, ν = 2 + 1/3. Such a state would naturally arise in a system where the SU(4) symmetry is strongly broken, e.g. by a strong Zeeman effect. However, as argued in Sec. V.A.4, these external fields are weak as compared to the leading interaction energy scale, and it is therefore natural to ask how such a state may arise from the inter- action point of view in the framework of four-component Halperin wave functions. A Halperin wave function that describes the above- mentioned situation is (Papi´c et al., 2010) ψ2+1/3 = Yξ=K,K′Yi<j(cid:16)z↑,ξ × Yξ=K,K′Yi<j(cid:16)z↓,ξ j (cid:17)3 Yi,j (cid:16)z↑,K i − z↑,ξ j (cid:17) , i − z↓,ξ i − z↑,K′ j (cid:17)3 (257) or any permutation of the spin-valley components. One notices that this wave function implicitly breaks the SU(4) spin-valley symmetry and, moreover, is not an eigenstate of the full SU(4) pseudospin, such that it can- not describe the ground state in the total absence of an external symmetry-breaking field. However, exact- diagonalization calculations have shown that even a tiny external Zeeman field is capable to stabelize the state (257), which becomes the ground state for ∆1 Z ≃ 0.01e2/εlB (Papi´c et al., 2010). Furthermore, the state in addition to the valley-pseudospin- (257) possesses, wave Goldstone mode in the spin-↑ branch, low-lying spin-flip excitations for moderately small Zeeman fields, even if the charge (activation) gap is the same as for the usual 1/3 Laughlin state. This particular interplay be- tween the leading Coulomb energy and subordinate ex- ternal spin-valley symmetry breaking terms, illustrated at the νG = 1/3 example, shows the complexity of the graphene FQHE, and further surprises may be expected in future experiments. 3. Experiments on the graphene FQHE We terminate this section on the graphene FQHE with a short discussion of experimental observations in the light of the above-mentioned theoretical four-component picture. In the first observations a. Two-terminal measurements. of the FQHE, the two-terminal configuration was used, where the voltage (and thus the resistance) is measured between the same two contacts used to drive the elec- tric current through the sample (Bolotin et al., 2009; Du et al., 2009). In this two-terminal configuration, it is not possible to measure simultaneously the Hall and the longitudinal resistance. It is nevertheless possible to extract the Hall and the longitudinal conductivities from the two-terminal resistance with the help of a conformal mapping, as a consequence of the 2D nature of the quan- tum transport in these systems (Abanin and Levitov, 2008; Williams et al., 2009). This technique has been ap- plied to obtain insight into the longitudinal conductivity the expected activated behavior of which yields a rough estimate of the activation gap at νG = 1/3 (∆1/3 ∼ 4.4 K at B = 12 T) (Abanin et al., 2010), which is an order of magnitude smaller than the theoretically expected value (Apalkov and Chakraborty, 2006; Toke et al., 2006).42 b. Four-terminal measurements. The activation gap of the 1/3 FQHE state has also been measured in suspended graphene in the four-terminal configuration, in which the longitudinal resistance can be measured directly and independently from the Hall resistance (Ghahari et al., 2011). In this case, the activation gap has been estimated to be ∆1/3 ∼ 26...50 K at B = 14 T, a value that agrees reasonably well with the theoretically expected value (Apalkov and Chakraborty, 2006; Toke et al., 2006) if one considers the energy scale e2/εε∞lB, which takes into account the RPA dielectric constant ε∞ for graphene in vacuum (see Sec. III.B.4). Finally, we would mention very recent high-field trans- port measurements in the four-terminal configuration on graphene on a h-BN substrate (Dean et al., 2011). These experiments allowed for a clear identification of several states of the 1/3 family, at νG = ±1/3,±2/3, and ±4/3 corresponding to the zero-energy LL n = 0, as well as at νG = ±7/3,±8/3,±10/3, and ±11/3 which reside in 42 Notice that the theoretical estimates have been obtained within a simplified two-component model, with a completely frozen spin degree of freedom. In spite of this simplification, the above-mentioned exact-diagonalization calculations with an im- plemented SU(4) symmetry have shown that the charge gap, which is responsible for the activated behavior, coincides indeed with that obtained in the two-component model (Papi´c et al., 2010). 51 the n = 1 LL. The estimation of the activation gap at νG = 4/3 agrees again reasonably well with the theo- retical expectation for the 1/3 state. The perhaps most salient (and unexpected) feature of the transport mea- surement is the absence (or extreme weakness) of the νG = ±5/3 representative of the 1/3 family, which would correspond to the Laughlin state (ν = 1/3 ↔ νG = −5/3 and the corresponding hole state) with a full SU(4) spin- valley ferromagnetic order, as argued in Sec. V.C.2. Whereas the absence of this state remains to be un- derstood, these findings corroborate the theoretical four- component picture of the graphene FQHE. Indeed, it clearly shows that the SU(4) symmetry of the n = 0 LL is essential because the only correspondence between the FQHE states is particle-hole symmetry that maps νG ↔ −νG. If the SU(4) symmetry were broken, e.g. by a sufficiently strong Zeeman effect, the only symme- try would be the valley-SU(2) symmetry in each spin branch of the n = 0 LL, in which case there exist the further mappings −2 + ν ↔ −ν in the spin-↓ branch and ν ↔ 2 − ν in the spin-↑ branch. However, the (observed) ±1/3 state would than be mapped on the (unobserved or extremely weak) ±5/3 state, and the strong difference in the visibility between these two states is therefore dif- ficult to understand. This is also the case if the SU(4) symmetry is fully broken by strong external spin and valley Zeeman fields, such that all spin-valley sublevels are completely resolved, and ±5/3 would be mapped on ±4/3, in the same manner as ±1/3 on ±2/3. VI. CONCLUSIONS AND OUTLOOK We have reviewed the quantum-mechanical properties of relativistic 2D electrons in monolayer graphene ex- posed to a strong magnetic field. The main parts of this review are concerned with the role of electronic interac- tions in graphene LLs. Whereas we have argued that these interactions may be treated perturbatively in the regime of the relativistic (integer) quantum Hall effect (RQHE), they constitute the relevant energy scale in par- tially filled graphene LLs due to the quenching of the ki- netic energy. This is reminiscent of partially filled LLs in non-relativistic 2D electron systems, and the most promi- nent consequence of this quenched kinetic energy and the macroscopic LL degeneracy is certainly the FQHE. The graphene FQHE is expected to be reminiscent of that of non-relativistic 2D electrons but it is governed by a larger internal degeneracy described to great accuracy by the SU(4) group. The experimental study of the FQHE in graphene is still in its infancy, and novel surprises may be expected. Only recently have been reported mea- surements in the four-terminal geometry which allow for an analysis of prominent characteristics of FQHE states, such as the activation gaps. In view of the generally accepted universality of the quantum Hall effect, it will certainly be interesting to make a systematic compari- son with the activation gaps of related FQHE states in conventional 2D electron gases with a parabolic band. Appendix A: Matrix Elements of the Density Operators 52 In the perturbative regime of the RQHE, the theoret- ical study of electron-electron interactions indicates the presence of fascinating novel collective modes, such as linear magneto-plasmons, that are particular to graphene and do not have a counterpart in non-relativistic 2D elec- tron systems in a perpendicular magnetic field. Also the upper-hybrid mode, which is the magnetic-field counter- part of the usual 2D plasmon, is expected to behave in a particular manner in graphene as a consequence of the linear disperison relation and the vanishing band mass. Whereas these studies are at present only theoretical, these collective modes may find an experimental verifica- tion in inelastic light-scattering measurements. Similarly to the role of electron-electron interactions in the RQHE regime, the electron-phonon coupling yields exciting resonance phenomena in graphene in a strong magnetic field. The electron-phonon interaction in graphene LLs has been discussed in the framework of a perturbative approach. Indications for the magneto- phonon resonance, e.g., have recently been found in Ra- man spectroscopy of epitaxial graphene. The present review has been limited to monolayer graphene, and it is definitely a reasonable research pro- gram to ask how the effects described here manifest them- selves in bilayer graphene. For example, the particular collective excitations described in Sec. III have been at- tributed to the lack of equidistant LL spacing and the presence of two bands. Whereas bilayer graphene also consists of two (particle-hole-symmetric) bands in the low-energy regime, the approximate parabolicity there yields almost equidistant LLs. The presence of additional high-energy bands (in the 300 meV range) certainly also affects the plasmonic modes. Acknowledgments I would like to express my deep gratitude to nu- merous collaborators without whom the realisation of this review would not have been possible. Above all, I must acknowledge the very fruitful long-term col- laborations with Jean-Noel Fuchs, on electron-electron interactions in the IQHE regime and electron-phonon coupling, and with Nicolas Regnault on the FQHE in graphene. I would furthermore thank my collabora- tors Claire Berger, Raphael de Gail, Benoıt Dou¸cot, Volodya Fal’ko, Cl´ement Faugeras, Kostya Kechedzhi, Pascal Lederer, Roderich Moessner, Gilles Montambaux, Cristiane Morais Smith, Zlatko Papi´c, Fr´ed´eric Pi´echon, Paulina Plochocka, Marek Potemski, Rafael Rold´an, and Guangquan Wang. Many thanks also to my colleagues H´el`ene Bouchiat, Antonio Castro Neto, Jean-Noel Fuchs, Christian Glattli, Paco Guinea, Anuradha Jagannathan, Philip Kim, and Bernhard Wunsch for fruitful discussions and a careful reading of this review. The matrix elements that intervene in the expres- sion for the density operators (139) are of the form hn, m exp(−iq · r)n′, m′i and may be calculated with the help of the decomposition of the cyclotron variable η and the guiding centre R into the ladder operators a and b, respectively [see Eqs. (78), (91) and (94)]. We further- more define the complex wave vectors q ≡ (qx + iqy)lB and ¯q = (qx − iqy)lB,43 One finds hn, me−iq·rn′, m′i = hme−iq·Rm′i ⊗ hne−iq·ηn′i = hme− i√2 ⊗hne− i√2 (qb†+¯qb)m′i (¯qa†+qa)n′i. (A1) The two matrix elements may be simplified with the help of the Baker-Hausdorff formula exp(A) exp(B) = exp(A + B) exp([A, B]/2), for the case [A, [A, B]] = [B, [A, B]] = 0 (Cohen-Tannoudji et al., 1973). The sec- ond matrix element thus becomes, for n ≥ n′ hne−iq·ηn′i = hne− i√2 (¯qa†+qa)n′i ¯qa† e− i√2 ¯qa† qan′i jihje− i√2 (A2) qan′i hne− i√2 n! (cid:18)−i¯q = e−q2/4hne− i√2 = e−q2/4Xj = e−q2/4r n′! × = e−q2/4r n′! Xj=0 n′ √2(cid:19)n−n′ n! (n − j)!(n′ − j)!j!(cid:18)−q2 n! (cid:18)−i¯q √2(cid:19)n−n′ 2 (cid:19)n′−j n′ (cid:18)q2 2 (cid:19) , Ln−n′ where we have used hne− i√2 ¯qa†ji =  0 q n! j! 1 (n−j)!(cid:16)− i√2 ¯q(cid:17)n−j for j > n for j ≤ n in the third line and the definition of the associated La- guerre polynomials (Gradshteyn and Ryzhik, 2000), Ln−n′ n′ (x) = n′ Xm=0 n! (n′ − m)!(n − n′ + m)! (−x)m m! . 43 We use this notation solely in the present appendix. Throughout the main text, q denotes the modulus of the wave vector q, q = q. hme−iq·Rm′i = hme− i√2 In the same manner, one obtains for m ≥ m′ (qb†+¯qb)m′i √2(cid:19)m−m′ = e−q2/4r m′! m! (cid:18)−iq (cid:18)q2 2 (cid:19) . ×Lm−m′ m′ With the help of the definition (A3) n! (cid:18)−iq √2(cid:19)n−n′ Gn,n′(q) ≡r n′! one may rewrite the expressions without the conditions n ≥ n′ and m ≥ m′, n′ (cid:18)q2 2 (cid:19) , Ln−n′ +Θ(m′ − m − 1)Gm′,m(−¯q)] e−q2/4. References hne−iq·ηn′i = [Θ(n − n′)Gn,n′ (¯q) +Θ(n′ − n − 1)Gn′,n(−q)] e−q2/4 and hme−iq·Rm′i = [Θ(m − m′)Gm,m′ (q) (A4) (A5) 53 Bena, C., and G. Montambaux, 2009, New J. Phys. 11, 095003. Berger, C., Z. Song, T. Li, A. Y. Ogbazghi, R. Feng, Z. Dai, A. N. Marchenkov, E. H. Conrad, P. N. First, and W. A. de Heer, 2004, J. Phys. Chem. 108, 19912. Berman, O., G. Gumbs, and Y. E. Lezovik, 2008, Phys. Rev. B 78, 085401. Bolotin, K. I., F. Ghahari, M. D. Shulman, H. L. Stormer, and P. Kim, 2009, Nature 462, 196. Bolotin, K. I., K. J. Sikes, Z. Jiang, M. Klima, G. Fudenberg, J. Hone, P. Kim, and H. L. Stormer, 2008, Solid State Commun. 146, 351. Bostwick, A., T. Ohta, T. Seyller, K. Horn, and E. Rotenberg, 2007, Nat. Phys. 3, 36. Brey, L., and H. Fertig, 2006, Phys. Rev. B 73, 195408. Brey, L., H. A. Fertig, R. Cot´e, and A. H. MacDonald, 1995, Phys. Rev. Lett. 75, 2562. Buttiker, M., 1992, The Quantum Hall Effect in Open Con- ductors, in M. Reed (Ed.) Nanostructured Systems (Semi- conductors and Semimetals, 35, 191) (Academic Press, Boston). Bychkov, Y. A., and G. Martinez, 2008, Phys. Rev. B 77, 125417. Castro Neto, A. H., and F. Guinea, 2007, Phys. Rev. B 75, 045404. Castro Neto, A. H., F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, 2009, Rev. Mod. Phys. 81, 109. Checkelsky, J. G., L. Li, and N. P. Ong, 2008, Phys. Rev. Lett. 100, 206801. Abanin, D. A., P. A. Lee, and L. S. Levitov, 2007a, Phys. Checkelsky, J. G., L. Li, and N. P. Ong, 2009, Phys. Rev. B Rev. Lett. 98, 156801. 79, 115434. Abanin, D. A., and L. S. Levitov, 2008, Phys. Rev. B 78, 035416. Abanin, D. A., K. S. Novoselov, U. Zeitler, P. A. Lee, A. K. Geim, and L. S. Levitov, 2007b, Phys. Rev. Lett. 98, 196806. Abanin, D. A., I. Skachko, X. Du, E. Y. Andrei, and L. S. Levitov, 2010, Phys. Rev. B 81, 115410. Chiu, K. W., and J. J. Quinn, 1974, Phys. Rev. B 9, 4724. Cohen-Tannoudji, C., B. Diu, and F. Laloe, 1973, Quantum Mechanics (Hermann, Paris). Coleman, P., 2003, An.. Henri Poincar´e 4, S559. Cooper, K. B., M. P. Lilly, J. P. Eisenstein, L. N. Pfeiffer, and K. W. West, 1999, Phys. Rev. B 60, R11285. Cot´e, R., D. B. Boisvert, J. Bourassa, M. Boissonneault, and Abergel, D. S. L., V. Apalkov, J. Berashevich, K. Ziegler, and H. A. Fertig, 2007, Phys. Rev. B 76, 125320. T. Chakraborty, 2010, Advances in Physics 59, 261. Cot´e, R., J.-F. Jobidon, and H. A. Fertig, 2008, Phys. Rev. B Abergel, D. S. L., and V. I. Fal’ko, 2007, Phys. Rev. B 75, 78, 085309. 155430. Abramowitz, M., and I. Stegun, 1970, Handbook of Mathe- matical Functions (9th Ed.) (Dover Publications, Dover). Ajiki, H., and T. Ando, 1995, J. Phys. Soc. Jpn. 64, 260. Aleiner, I. L., and L. I. Glazman, 1995, Phys. Rev. B 52, 11296. Alicea, J., and M. P. A. Fisher, 2006, Phys. Rev. B 74, 075422. Ando, T., 2006a, J. Phys. Soc. Jpn. 75, 074716. Ando, T., 2006b, J. Phys. Soc. Jpn 75, 124701. Ando, T., 2007a, J. Phys. Soc. Jpn. 76, 024712. Ando, T., 2007b, J. Phys. Soc. Jpn. 76, 024712. Andrei, E. Y., G. Deville, D. C. Glattli, F. I. B. Williams, E. Paris, and B. Etienne, 1988, Phys. Rev. Lett. 60, 2765. Apalkov, V. M., and T. Chakraborty, 2006, Phys. Rev. Lett. Damascelli, A., 2004, Phys. Scripta T109, 61. de Gail, R., N. Regnault, and M. O. Goerbig, 2008, Phys. Rev. B 77, 165310. Dean, C., A. Young, P. Cadden-Zimansky, L. Wang, H. Ren, K. Watanabe, T. Taniguchi, P. Kim, J. Hone, and K. Shep- ard, 2011, Nature Phys. 7, 693. Dietl, P., F. Pi´echon, and G. Montambaux, 2008, Phys. Rev. Lett 98, 236405. Dillon, R. O., I. L. Spain, and J. W. McClure, 1977, J. Phys. Chem. Solids 38, 635. Doretto, R. L., and C. Morais Smith, 2007, Phys. Rev. B 76, 195431. Dou¸cot, B., M. O. Goerbig, P. Lederer, and R. Moessner, 2008, Phys. Rev. B 78, 195327. Drut, J. E., and T. A. Lahde, 2009a, Phys. Rev. Lett. 102, 97, 126801. 026802. Arikawa, M., Y. Hatsugai, and H. Aoki, 2008, Phys. Rev. B 78, 205401. Arovas, D. P., A. Karlhede, and D. Lilliehook, 1999, Phys. Rev. B 59, 13147. Banerjee, S., R. R. P. Singh, V. Pardo, and W. E. Pickett, Drut, J. E., and T. A. Lahde, 2009b, Phys. Rev. B 79, 165425. Du, R. R., D. C. Tsui, H. L. Stormer, L. N. Pfeiffer, K. W. Baldwin, and K. W. West, 1999, Solid State Comm. 109, 389. Du, X., I. Skachko, A. Barker, and E. Y. Andrei, 2008, Nat. 2009, Phys. Rev. Lett. 103, 016402. Nanotech. 3, 491. 54 Du, X., I. Skachko, F. Duerr, A. Luican, and E. Y. Andrei, 241405. 2009, Nature 462, 192. Gonz´alez, J., F. Guinea, and M. A. H. Vozmediano, 1994, Eisenstein, J. P., K. B. Cooper, L. N. Pfeiffer, and K. W. Nucl. Phys. B 424, 595. West, 2002, Phys. Rev. Lett. 88, 076801. Gonz´alez, J., F. Guinea, and M. A. H. Vozmediano, 1999, Esaki, K., M. Sato, M. Kohmoto, and B. I. Halperin, 2009, Phys. Rev. B 59, R2474. Phys. Rev. B 80, 125405. Ezawa, M., 2007, J. Phys. Soc. Jpn. 76, 094701. Ezawa, Z. F., 2000, Quantum Hall Effects – Field Theoretical Approach and Related Topics (World Scientific, Singapore). Ezawa, Z. F., and K. Hasebe, 2002, Phys. Rev. B 65, 075311. Ezawa, Z. F., and A. Iwazaki, 1993, Phys. Rev. B 47, 7295. Ezawa, Z. F., G. Tsitsishvili, and K. Hasebe, 2003, Phys. Rev. B 67, 125314. Gorbar, E. V., V. P. Gusynin, V. A. Miransky, and I. A. Shovkovy, 2002, Phys. Rev. B 66, 045108. Gorbar, E. V., V. P. Gusynin, V. A. Miransky, and I. A. Shovkovy, 2008, Phys. Rev. B 78, 085437. Gradshteyn, I. S., and I. M. Ryzhik, 2000, Table of Inte- grals, Series and Products (6th Ed.) (Academic Press, San Diego). Graf, D., F. Molitor, K. Ensslin, C. Stampfer, A. Jungen, Fano, G., F. Ortolani, and E. Colombo, 1986, Phys. Rev. B C. Hierold, and L. Wirtz, 2007, Nano Lett. 7, 238. 34, 2670. Greiter, M., X.-G. Wen, and F. Wilczek, 1991, Phys. Rev. Farjam, M., and H. Rafii-Tabar, 2009, Phys. Rev. B 80, Lett. 66, 3205. 167401. Faugeras, C., M. Amado, P. Kossacki, M. Orlita, M. Sprinkle, C. Berger, W. A. de Heer, and M. Potemski, 2009, Phys. Rev. Lett. 103, 186803. Ferrari, A. C., J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth, and A. K. Geim, 2006, Phys. Rev. Lett. 97, 187401. Fertig, H. A., 1989, Phys. Rev. B 40, 1087. Fogler, M. M., A. A. Koulakov, and B. I. Shklovskii, 1996, Phys. Rev. B 54, 1853. Gupta, A., G. Chen, P. Joshi, S. Tadigadapa, and P. Eklund, 2006, Nano Lett. 6, 2667. Gusynin, V. P., V. A. Miransky, S. G. Sharapov, and I. A. Shovkovy, 2006, Phys. Rev. B 74, 195429. Gusynin, V. P., and S. G. Sharapov, 2005, Phys. Rev. Lett. 95, 146801. Gusynin, V. P., and S. G. Sharapov, 2006, Phys. Rev. B 73, 245411. Haldane, F. D. M., 1983, Phys. Rev. Lett. 51, 605. Haldane, F. D. M., 1988, Phys. Rev. Lett. 61, 2015. Haldane, F. D. M., and E. H. Rezayi, 1985, Phys. Rev. Lett. Fuchs, J.-N., and P. Lederer, 2007, Phys. Rev. Lett. 98, 54, 237. 016803. Fukuda, Y. et al. (Super-Kamiokande Collaboration), 1998, Phys. Rev. Lett. 81, 1562. Fukuyama, H., 1975, Solid State Commun. 17, 1323. Gangadharaiah, S., A. M. Farid, and E. G. Mishchenko, 2008, Halperin, B. I., 1983, Helv. Phys. Acta 56, 75. Harrison, W. A., 1981, Phys. Rev. B 24, 5835. Hasan, M. Z., and C. L. Kane, 2010, Rev. Mod. Phys. 82, 3045. Hasegawa, Y., R. Konno, H. Nakano, and M. Kohmoto, 2006, Phys. Rev. Lett. 100, 166802. Phys. Rev. B 74, 033413. Geim, A. K., and K. S. Novoselov, 2007, Nat. Materials 6, 183. Ghahari, F., Y. Zhao, P. Cadden-Zimansky, K. Bolotin, and P. Kim, 2011, Phys. Rev. Lett. 106, 046801. Girvin, S. M., 1999, The Quantum Hall Effect: Novel Excita- tions and Broken Symmetries, in A. Comptet, T. Jolicoeur, S. Ouvry and F. David (Eds.) Topological Aspects of Low- Dimensional Systems – ´Ecole d’ ´Ete de Physique Th´eorique LXIX (Springer, Berlin). Girvin, S. M., A. H. MacDonald, and P. M. Platzman, 1986, Phys. Rev. B 33, 2481. de Heer, W. A., C. Berger, X. Wu, P. N. First, E. H. Conrad, X. B. Li, T. B. Li, M. Sprinkle, J. Hass, M. Sadowski, M. Potemski, and G. Martinez, 2007, Solid State Comm. 143, 92. Herbut, I. F., 2006, Phys. Rev. Lett. 97, 146401. Herbut, I. F., 2007a, Phys. Rev. B 76, 085432. Herbut, I. F., 2007b, Phys. Rev. B 75, 165411. Herbut, I. F., 2008, Phys. Rev. B 78, 205433. Herbut, I. F., V. Jurici´c, and B. Roy, 2009a, Phys. Rev. B 79, 085116. Herbut, I. F., V. Jurici´c, and O. Vafek, 2009b, Phys. Rev. B Giuliani, G. F., and G. Vignale, 2005, Quantum Theory of 80, 075432. Electron Liquids (Cambridge UP, Cambridge). Goerbig, M. O., 2009, Quantum Hall Effects (Lecture notes of the Les Houches Summer School 2009 (Singapore Session), arXiv:0909.1998). Goerbig, M. O., B. Dou¸cot, and R. Moessner, 2006, Phys. Rev. B 74, 161407. Hou, C.-Y., C. Chamon, and C. Mudry, 2010, Phys. Rev. B 81, 075427. Hou, J.-M., W.-X. Yang, and X.-J. Liu, 2009, Phys. Rev. A 79, 043621. Hwang, E. H., and S. Das Sarma, 2007, Phys. Rev. B 75, 205418. Goerbig, M. O., J.-N. Fuchs, K. Kechedzhi, and V. I. Fal’ko, Imry, Y., 1997, Introduction to Mesoscopic Physics (Oxford 2007, Phys. Rev. Lett. 99, 087402. UP). Goerbig, M. O., J.-N. Fuchs, G. Montambaux, and Ishikawa, K., and T. Ando, 2006, J. Phys. Soc. Jpn. 75, F. Pi´echon, 2008, Phys. Rev. B 78, 045415. 084713. Goerbig, M. O., J.-N. Fuchs, G. Montambaux, and Iyengar, A., J. Wang, H. A. Fertig, and L. Brey, 2007, Phys. F. Pi´echon, 2009, EPL 85, 57005. Rev. B 75, 125430. Goerbig, M. O., P. Lederer, and C. Morais Smith, 2003, Phys. Jackson, J. D., 1999, Classical Electrondynamics (Wiley, 3rd Rev. B 68, 241302. Goerbig, M. O., P. Lederer, and C. Morais Smith, 2004, Phys. Rev. B 69, 115327. Goerbig, M. O., and N. Regnault, 2007, Phys. Rev. B 75, ed., New York). Jiang, Z., E. A. Henriksen, Y.-J. W. L. C. Tung, M. E. Schwartz, M. Y. Han, P. Kim, and H. L. Stormer, 2007a, Phys. Rev. Lett. 98, 197403. Jiang, Z., Y. Zhang, H. L. Stormer, and P. Kim, 2007b, Phys. 55 Rev. Lett. 99, 106802. Jobst, J., D. Waldmann, F. Speck, R. Hirner, D. K. Maude, T. Seyller, and H. B. Weber, 2010, Phys. Rev. B 81, 195434. Jurici´c, V., I. F. Herbut, and G. W. Semenoff, 2009, Phys. Rev. B 80, 081405(R). Kallin, C., and B. I. Halperin, 1984, Phys. Rev. B 30, 5655. Kane, C., and E. J. Mele, 2005, Phys. Rev. Lett. 95, 226801. Katayama, S., A. Kobayashi, and Y. Suzumura, 2006, J. Phys. Soc. Jpn. 75, 054705. Katsnelson, M., K. Novoselov, and A. Geim, 2006, Nat. Phys. 2, 620. Martin, J., N. Akerman, G. Ulbricht, T. Lohmann, J. H. Smet, K. von Klitzing, and A. Yacobi, 2008, Nat. Phys. 4, 144. McClure, J. W., 1956, Phys. Rev. 104, 666. Meyer, J. C., A. K. Geim, M. I. Katsnelson, K. S. Novoselov, T. J. Booth, and S. Roth, 2007, Nature 446, 60. Mikitik, G. P., and Y. V. Sharlai, 1999, Phys. Rev. Lett. 82, 2147. Min, H., J. E. Hill, N. A. Sinitsyn, B. R. Sahu, L. Kleinman, and A. H. MacDonald, 2006, Phys. Rev. B 74, 165310. Moessner, R., and J. T. Chalker, 1996, Phys. Rev. B 54, 5006. Montambaux, G., F. Pi´echon, J.-N. Fuchs, and M. O. Goer- Katsnelson, M. I., 2006, Phys. Rev. B 74, 201401. Kekul´e, A., 1865, Bulletin de la Societe Chimique de Paris 3, big, 2009a, Phys. Rev. B 80, 153412. Montambaux, G., F. Pi´echon, J.-N. Fuchs, and M. O. Goer- (2) 98. Kekul´e, A., 1866, Annalen der Chemie und Pharmazie 137, (2) 129. Kellogg, M., J. P. Eisenstein, and L. N. P. an K. W. West, 2004, Phys. Rev. Lett. 036801. Khveshchenko, D. V., 2001, Phys. Rev. Lett. 87, 206401. Khveshchenko, D. V., 2007, Phys. Rev. B 75, 153405. Klein, O., 1929, Z. Phys. 53, 157. v. Klitzing, K., G. Dorda, and M. Pepper, 1980, Phys. Rev. Lett. 45, 494. big, 2009b, Europhys. J. B 72, 509. Moon, K., H. Mori, K. Yang, S. M. Girvin, A. H. MacDonald, I. Zheng, D. Yoshioka, and S.-C. Zhang, 1995, Phys. Rev. B 51, 5143. Moore, G., and N. Read, 1991, Nucl. Phys. B 360, 362. Morinari, T., T. Himura, and T. Tohyama, 2009, J. Phys. Soc. Jpn. 78, 023704. Mucha-Kruczy´nski, M., O. Tsyplyatyev, A. Grishin, E. Mc- Cann, V. I. Fal’ko, A. Bostwick, and E. Rotenberg, 2008, Phys. Rev. B 77, 195403. Kobayashi, A., S. Katayama, Y.Suzumura, and H. Fukuyama, Nomura, K., and A. H. MacDonald, 2006, Phys. Rev. Lett. 2007, J. Phys. Soc. Jpn. 76, 034711. 96, 256602. Kohn, W., 1959, Phys. Rev. Lett. 2, 393. Kohn, W., 1961, Phys. Rev. 123, 1242. Kosterlitz, J. M., and D. J. Thouless, 1973, J. Phys. C 6, 1181. Kotov, V. N., B. Uchoa, and A. H. Castro Neto, 2007, Phys. Rev. B 78, 035119. Kotov, V. N., B. Uchoa, V. M. Peirera, A. H. Castro Neto, and F. Guinea, 2010, preprint , arXiv:1012.3484. Nomura, K., S. Ryu, and D.-H. Lee, 2009, Phys. Rev. Lett. 103, 216801. Novoselov, K. S., A. K. Geim, S. V. Morosov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, 2005a, Nature 438, 197. Novoselov, K. S., A. K. Geim, S. V. Morosov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, 2004, Science 306, 666. Koulakov, A. A., M. M. Fogler, and B. I. Shklovskii, 1996, Novoselov, K. S., D. Jiang, T. Booth, V. V. Khotkevich, S. M. Phys. Rev. Lett. 76, 499. Laughlin, R. B., 1983, Phys. Rev. Lett. 50, 1395. Lazzeri, M., and F. Mauri, 2006, Phys. Rev. Lett. 97, 266407. Lee, C., X. Wei, J. K. Kysar, and J. Hone, 2008, Science 321, 385. Morozov, and A. K. Geim, 2005b, PNAS 102, 10451. Onsager, L., 1952, Philos. Mag. 43, 1006. Papi´c, Z., M. O. Goerbig, and N. Regnault, 2009, Solid State Comm. 149, 1056. Papi´c, Z., M. O. Goerbig, and N. Regnault, 2010, Phys. Rev. Lee, K. L., B. Gr´emaud, R. Han, B.-G. Englert, and Lett. 105, 176802. C. Miniatura, 2009, Phys. Rev. A 80, 043411. Lewis, R. M., Y. Chen, L. W. Engel, D. C. Tsui, L. N. Pfeiffer, Partoens, B., and F. Peeters, 2006, Phys. Rev. B 74, 075407. Pauling, L., 1960, The Nature of Chemical Bonds (Cornell and K. W. West, 2005, Phys. Rev. B 71, 081301(R). UP). Lewis, R. M., Y. Chen, L. W. Engel, D. C. Tsui, P. D. Ye, L. N. Pfeiffer, and K. W. West, 2004, Phys. Rev. Lett. 93, 176808. Pereira, V. M., A. H. Castro Neto, and N. M. R. Peres, 2009, Phys. Rev. B 80, 045401. Peres, N. M., F. Guinea, and A. H. Castro Neto, 2006, Phys. Lewis, R. M., P. D. Ye, L. W. Engel, D. C. Tsui, L. N. Pfeiffer, Rev. B 73, 125411. and K. W. West, 2002, Phys. Rev. Lett. 89, 136804. Li, G., and E. Andrei, 2007, Nat. Phys. 3, 623. Li, G., A. Luican, and E. Y. Andrei, 2009a, Phys. Rev. Lett. 102, 176804. Li, G., A. Luican, and E. Y. Andrei, 2009b, Phys. Rev. Lett. 102, 176804. Peres, N. M. R., 2010, Rev. Mod. Phys. 82, 2673. Peres, N. M. R., and E. V. Castro, 2007, J. Phys.:Condensed Matter 19, 406231. Pisana, S., M. Lazzeri, C. Casiraghi, K. S. Novoselov, A. K. Geim, A. C. Ferrari, and F. Mauri, 2007, Nature Mat. 6, 198. Lifshitz, I. M., and A. M. Kosevich, 1956, Sov. Phys. JETP Piscanec, S., M. Lazzeri, F. Mauri, A. C. Ferrari, and 2, 636. Lilly, M. P., K. B. Cooper, J. P. Eisenstein, L. N. Pfeiffer, and K. W. West, 1999, Phys. Rev. Lett. 82, 394. Lukose, V., R. Shankar, and G. Baskaran, 2007, Phys. Rev. Lett. 98, 116802. J. Robertson, 2004, Phys. Rev. Lett. 93, 185503. Plochocka, P., C. Faugeras, M. Orlita, M. L. Sadowski, G. Martinez, M. Potemski, M. O. Goerbig, J.-N. Fuchs, C. Berger, and W. A. de Heer, 2008, Phys. Rev. Lett 100, 087401. Mahan, G. D., 1993, Many-Particle Physics (Plenum Press, Polini, M., R. Asgari, G. Borghi, Y. Barlas, T. Pereg-Barnea, 2nd Ed., New York). Mallet, P., F. Varchon, C. Naud, L. Magaud, C. Berger, and J.-Y. Veuillen, 2007, Phys. Rev. B 76, 041403(R). and A. H. MacDonald, 2008, Phys. Rev. B 77, 081411. Ponomarenko, L. A., F. Schedin, M. I. Katsnelson, R. Yang, E. W. Hill, K. S. Novoselov, and A. K. Geim, 2008, Science 56 320, 356. Weinberg, S., 1995, The Quantum Theory of Fields (Cam- Poplavskyy, O., M. O. Goerbig, and C. Morais Smith, 2009, bridge UP). Phys. Rev. B 80, 195414. Qi, X.-L., and S.-C. Zhang, 2011, Rev. Mod. Phys. 83, 1057. Reina, A., X. Jia, J. Ho, D. Nezich, H. Son, V. Bulovic, M. S. Dresselhaus, and J. Kong, 2009, Nano Lett. 9, 30. Rold´an, R., J.-N. Fuchs, and M. O. Goerbig, 2009, Phys. Rev. B 80, 085408. Rold´an, R., M. O. Goerbig, and J.-N. Fuchs, 2010, Semicond. Sci. Technol. 25, 034005. Sabio, J., J. Nilsson, and A. H. Castro Neto, 2008, Phys. Rev. B 78, 075410. Wen, X.-G., and A. Zee, 1992a, Phys. Rev. Lett 69, 1811. Wen, X.-G., and A. Zee, 1992b, Phys. Rev. B 46, 2290. Willett, R. L., J. P. Eisenstein, H. L. Stormer, D. C. Tsui, A. C. Gossard, and J. H. English, 1987, Phys. Rev. Lett. 59, 1776. Williams, F. I. B., P. A. Wright, R. G. Clark, E. Y. Andrei, G. Deville, D. C. Glattli, O. Probst, B. Etienne, C. Dorin, C. T. Foxon, and J. J. Harris, 1991, Phys. Rev. Lett. 66, 3285. Williams, J. R., D. A. Abanin, L. DiCarlo, L. S. Levitov, and Sadowski, M. L., G. Martinez, M. Potemski, C. Berger, and C. M. Marcus, 2009, Phys. Rev. B 80, 045408. W. A. de Heer, 2006, Phys. Rev. Lett. 97, 266405. Saito, R., G. Dresselhaus, and M. S. Dresselhaus, 1998, Physical Properties of Carbon Nanotubes (Imperial College Press, London). Salem, L., 1966, Molecular Orbital Theory of Conjugated Sys- tems (Benjamin, New York). Shen, T., J. J. Gu, M. Xu, Y. Q. Wu, M. L. Bolen, M. A. Capano, L. W. Engel, and P. D. Ye, 2009, Appl. Phys. Lett. 95, 172105. Shimshoni, E., H. A. Fertig, and G. Vanketeswara Pai, 2009, Phys. Rev. Lett. 102, 206408. Shizuya, K., 2007, Phys. Rev. B 75, 245417. Shon, N. H., and T. Ando, 1998, J. Phys. Soc. Jpn. 67, 2421. Shung, K. W. K., 1986, Phys. Rev. B 34, 979. Sondhi, S. L., A. Karlhede, S. A. Kivelson, and E. H. Rezayi, 1993, Phys. Rev. B 47, 16419. Song, Y. J., A. F. Otte, Y. Kuk, Y. Hu, D. B. Torrance, P. N. First, W. A. de Heer, H. Min, S. Adam, M. D. Stiles, and A. H. MacDonald, 2010, Nature 467, 185. Spielman, I. B., J. P. Eisenstein, L. N. Pfeiffer, and K. W. West, 2000, Phys. Rev. Lett. 84, 5808. Stern, F., 1967, Phys. Rev. Lett. 18, 546. Tahir, M., and K. Sabeeh, 2008, J. Phys.:Condens. Matter 20, 425202. Toke, C., and J. K. Jain, 2007, Phys. Rev. B 75, 245440. Toke, C., P. E. Lammert, V. H. Crespi, and J. K. Jain, 2006, Phys. Rev. B 74, 235417. Tsui, D. C., H. Stormer, and A. C. Gossard, 1982, Phys. Rev. Lett. 48, 1559. Wirtz, L., and A. Rubio, 2004, Solid Stat. Comm. 131, 141. Wojs, A., G. Moller, and N. R. Cooper, 2010, preprint , arXiv:1007.3006. Wu, X., Y. Hu, M. Ruan, N. K. Madiomanana, J. Hankinson, M. Sprinkle, C. Berger, and W. A. de Heer, 2009, Appl. Phys. Lett. 95, 223108. Wunsch, B., F. Sols, and F. Guinea, 2008, New Journal of Physics 10, 103027. Wunsch, B., T. Stauber, F. Sols, and F. Guinea, 2006, New Journal of Physics 8, 318. Yan, J., Y. Zhang, P. Kim, and A. Pinczuk, 2007, Phys. Rev. Lett. 98, 166802. Yang, K., S. Das Sarma, and A. H. MacDonald, 2006, Phys. Rev. B 74, 075423. Zhang, C.-H., and Y. N. Joglekar, 2007, Phys. Rev. B 75, 245414. Zhang, C.-H., and Y. N. Joglekar, 2008, Phys. Rev. B 77, 205426. Zhang, Y., Z. Jiang, J. P. Small, M. S. Purewal, Y.-W. Tan, M. Fazlollahi, J. D. Chudow, J. A. Jaszczak, H. L. Stormer, and P. Kim, 2006, Phys. Rev. Lett. 96, 136806. Zhang, Y., Y.-W. Tan, H. L. Stormer, and P. Kim, 2005, Nature 438, 201. Zhao, E., and A. Paramekanti, 2006, Phys. Rev. Lett. 97, 230404. Zhou, S. Y., G.-H. Gweon, J. Graf, A. V. Fedorov, C. D. Spataru, R. D. Diehl, K. Kopelevich, D.-H. Lee, S. G. Louie, and A. Lanzara, 2006, Nat. Phys. 2, 595. Zhu, S.-L., B. Wang, and L.-M. Duan, 2007, Phys. Rev. Lett. Tutuc, E., M. Shayegan, and D. A. Huse, 2004, Phys. Rev. 98, 260402. Lett 93, 036802. Wallace, P. R., 1947, Phys. Rev. 71, 622.
1201.0466	1	1201	2012-01-02T13:49:38	Spin transistor action from Onsager reciprocity and SU(2) gauge theory	[ "cond-mat.mes-hall" ]	We construct a local gauge transformation to show how, in confined systems, a generic, weak nonhomogeneous SU(2) spin-orbit Hamiltonian reduces to two U(1) Hamiltonians for spinless fermions at opposite magnetic fields, to leading order in the spin-orbit strength. Using an Onsager relation, we further show how the resulting spin conductance vanishes in a two-terminal setup, and how it is turned on by either weakly breaking time-reversal symmetry or opening additional transport terminals. We numerically check our theory for mesoscopic cavities as well as Aharonov-Bohm rings.	cond-mat.mes-hall	cond-mat	Spin transistor action from Onsager reciprocity and SU(2) gauge theory I. Adagideli,1 V. Lutsker,2 M. Scheid,2 Ph. Jacquod,3, 4 and K. Richter2 1Faculty of Engineering and Natural Sciences, Sabanci University, Orhanli-Tuzla, Istanbul, Turkey 2Institut fur Theoretische Physik, Universitat Regensburg, D-93040 Regensburg, Germany 3Physics Department and College of Optical Sciences, University of Arizona, Tucson, AZ 85721, USA 4Theoretical Physics Department, University of Geneva, 1211 Geneva, Switzerland (Dated: November 18, 2018) We construct a local gauge transformation to show how, in confined systems, a generic, weak non- homogeneous SU (2) spin-orbit Hamiltonian reduces to two U (1) Hamiltonians for spinless fermions at opposite magnetic fields, to leading order in the spin-orbit strength. Using an Onsager relation, we further show how the resulting spin conductance vanishes in a two-terminal setup, and how it is turned on by either weakly breaking time-reversal symmetry or opening additional transport terminals. We numerically check our theory for mesoscopic cavities as well as Aharonov-Bohm rings. PACS numbers: 72.25.Dc, 73.23.-b, 85.75.-d Transistor action is often based on symmetries. To switch on and off a field effect transistor, an exter- nal gate turns a three-dimensional insulator into a two- dimensional metal and back. Compared to the off-state, the on-state has thus reduced dimensionality and sym- metry. The relevance of symmetries in transistor ac- tion is even more pronounced in some recently proposed spin-based transistors, whose action follows directly from the breaking of spin rotational symmetry, by tuning spin-orbit interaction (SOI) around a special symmetry point [1], where the SOI field reduces to two identical U (1) fields with opposite coupling constants [2]. In this manuscript, we propose a new class of spin tran- sistors whose action is based on an Onsager reciprocity relation. We show that in confined quantum coher- ent systems with spatially inhomogeneous SOI (Rashba, Dresselhaus or impurity SOI, or a combination of the three), an appropriate SU (2) gauge transformation al- lows to express the spin conductance G↑ ij between two terminals labeled i and j through the charge magne- toconductance Gij (B) as G↑ ij = Gij (B) − Gij (−B). This holds to leading order in the ratio L/ℓso ≪ 1 of the system size L and the spin-orbit (precession) length ℓso. The pseudo magnetic field B arises from the gauge- transformed SOI. Current conservation together with the Onsager relation Gij (B) = Gji(−B) [3, 4] then forces G↑ ij = 0 to leading order for a two-terminal setup. This is the off state of our transistor. The on state is ob- tained by either opening additional terminals or breaking time-reversal symmetry with a true magnetic field B0, in which case G↑ ij = Gij (B0 + B) − Gij (B0 − B) 6= 0, even in a two-terminal setup. Our Onsager spin transis- tor can thus be controlled either electrically or magneti- cally. In both instances, this turns on a spin conductance G↑ so with an on/off ratio ∝ (ℓso/L)2 ≫ 1. The mechanism works in diffusive as well as ballistic sys- tems, and is more pronounced in regular systems with few channels. ij ∝ ℓ−1 ij − G↓ ij − G↓ ij − G↓ ij − G↓ ij − G↓ to show that, in confined systems with L/ℓso ≪ 1, a homogeneous k-linear SOI has a much weaker effect on charge transport than the naive expectation ∝ ℓ−1 so [5]. Brouwer and collaborators later argued that terms ∝ ℓ−1 so in the charge conductance survive the gauge transfor- mation for SOI with spatially varying strength [6]. The relevance of the pseudo magnetic field for a specific meso- scopic system with inhomogenous SOI was noticed in Ref. [7]. It is however not clear how much of the gauge arguments of Refs. [5, 6, 8] carry over to spin transport in generic systems [9]. Below we show that gauge trans- formations result in different symmetries for charge and for spin transport [10]. Our starting point is a two-dimensional Hamiltonian for electrons with SOI, which we write as ( ≡ 1) H = − 1 2m DµDµ + V (x) . (1) Here, V (x) is a spin-diagonal potential and the covariant derivative Dµ = ∂µ − (ikso/2)σaAa µ contains the SOI via µ, with the Pauli matrix σa. the SU (2) gauge field σaAa From here on, Latin indices are spin indices, while Greek letters denote spatial indices. The SOI constant kso de- termines the spin-orbit length as ℓso = πkso−1. We consider a gauge transformation O → O′ = U OU −1 with U = exp(iσaΛa/2) ≃ 1 + iσaΛa/2, and search for a Λa that reduces the leading order, kso-linear part of the SOI to a spin-diagonal U (1) × σz structure. We use the well- known decomposition (ǫµν is the totally antisymmetric tensor of order two) for each spin component Aa µ = −(∂µχa + ǫµν∂ν φa) , (2) with φa given by ∇2φa = ǫµν∂µAa In particular, φa ν. is necessarily nonzero for spatially varying SOI. It is straightforward to see that the choice Λa = ksoχa gauges away the gradient part of the SU (2) vector potential to linear order in kso, Aleiner and Falko constructed a gauge transformation Aa µ → (A′)a µ = −ǫµν∂νφa + O(kso) . (3) Note that O(kso) corrections in Aa so) correc- tions in the Hamiltonian. If the SOI strength is spatially constant, φa = 0 and one recovers the result of Ref. [5] that all O(kso)-terms are gauged away. µ lead to O(k2 We next want to extract the leading order, linear in ksoχa ≪ 1 dependence of transport properties such as conductances, and thus use O′ = U OU −1 = O + ikso[σaχa, O]/2 . (4) In particular we have Dµ → D′ µ with σa → σa′ = σa + ksoǫabcχb(x)σc . (5) To calculate spin conductances we need to gauge trans- form the operator for spin current through a cross-section Cj in terminal j, I a dα {nα · j(α), σa}. We obtain j = RCj ( I ′)a j = ZCj = J a dα (cid:2){nα · j′(α), σa + kso ǫabcχb(α)σc}(cid:3) j + ksoδ J a j (6) where J a j is the "naive" spin current of the transformed Hamiltonian, not accounting for the rotation (5) of the spin axes. We further need the Heisenberg picture oper- ators I a j e−iHt which transform as j (t) = eiHt I a ( I ′)a j (t) = J a j (t) + ksoδ J a j0(t) + O(k2 so) . (7) Here the subscript 0 means that the time-evolution is through the kso = 0 Hamiltonian. ijµj/e. i = Pj Ga ij )a, with the conductance matrix (G′ Linear response relates chemical potentials in exter- nal reservoirs and currents in the leads via the spin- conductance matrix as I a It is some- how tedious, though straightforward to show that, to lin- ear order in kso, the gauge transformation gives Ga ij → ij )a evaluated (G′ in the same way as Ga ij but with the spin current op- erators J a j of the transformed Hamiltonian in Eq. (7). Thus, to leading order in kso, infinitesimal nonabelian gauge transformations preserve the form of the spin con- ductance. Note that global gauge transformations (i.e. global spin rotations), whether infinitesimal or finite, are easy to introduce via the corresponding rotation matrix R as Ga ij )b. All global or local spin gauge transformations leave the potential V (x) invariant. ij = Rab(G′ We are now equipped to use the gauge transformation to explore the spin conductance. We first focus on the exactly solvable case of a Rashba SOI [11] with a spatially varying strength α(x) = kso α (x·f ), with a dimensionless function α, whose gradient always points in the direction µ = −2α(x·f )ǫaµ, φa(x) = of the unit vector f . One has Aa ϕ(x)f a, and Eq. (3) gives (A′)a µ = −ǫµν∂νϕ(x)f a + O(kso) , D′ µ = ∂µ + i 2 ksoǫµν∂ν ϕ(x)σ ·f . (8a) (8b) After the global spin rotation σ ·f →σz, Eq. (1) becomes 2 H = (cid:18) h(a) 0 h(−a) (cid:19) + O(k2 [∇ + iksoa(x)]2 + V (x) . so) , (9a) (9b) h(a) = − 0 1 2m ij = (cid:2)Gij(B) − Gij (−B)(cid:3)f a + O(k2 Thus to linear order in kso, the Hamiltonian is mapped onto a block spin Hamiltonian where the opposite spins feel opposite, purely orbital pseudo magnetic fields B = (∇ × a)z generated by the U (1) vector potential aµ = 1 2 ǫµν∂ν ϕ(x). We obtain B(x) = ksof · ∇¯α. Transforming back to the original gauge, the spin conductance is ob- tained as Ga so). In this simple example, the spin conductance is thus the differ- ence of two charge conductances Gij at opposite pseudo magnetic fields. For generally varying SOI, one cannot choose a spin quantization axis as before. Thus we need to define one pseudo-magnetic field per spin polarization, i.e. we define Ba = ∂xAa x as the magnitude of a pseudo magnetic field (pointing always in z-direction) that arises solely from the a component of φa. To linear order in ksoL, the superposition principle gives the spin conductance along axis a as solely due to the component of φa, Ga ij = Gij (Ba) − Gij (−Ba). The same argument gives the leading-order spin conductance in the presence of an externally applied (i.e. true) magnetic field B0 as y − ∂yAa Ga ij (B0) = Gij(B0 + Ba) − Gij (B0 − Ba) + O(k2 so). (10) This is our main result. It expresses the spin conductance of the original dot with SOI in terms of charge conduc- tances of the dot without SOI, but with effective mag- netic fields B0 ± Ba arising from the true applied field, B0, and the pseudo field, Ba, generated by the gauge transformation and the SOI. The key observation is then that the reciprocity rela- tion Gij (B) = Gji(−B) [4], together with gauge invari- ance, Pj Gij(B) = 0, imply that the spin conductance (10) vanishes to order O(kso) in two-terminal geometries in the absence of external magnetic field, since only then Gij (−Ba) = Gji(Ba) = Gij(Ba). On the contrary, Ga ij is linear in kso, i.e. much larger, when an external magnetic field is applied or when one (or more) additional termi- nals are open. Thus, multi-terminal spin conductances linearly depend on ksoL, whereas two-terminal local con- ductances are quadratic or higher order in ksoL. These restrictions imply that any coherent conductor with spa- tially varying SOI can be operated as a spin transistor, whose action is controlled by either opening an extra ter- minal or applying an external magnetic field. This is the fundamental mechanism on which the Onsager spin tran- sistor we propose is based. We numerically confirm these results by computing [12] the charge and spin conductances for two- and three- terminal mesoscopic cavities and rings (sketched in the inset of Figs. 1 -- 3). We first assume a Rashba SOI . . 0,1 0,05 0 -0,05 -0,1 2 2,5 3 3,5 4 Comparison of the spin conductance Gy RL = Figure 1: G↑ RL − G↓ RL with the difference in the magnetoconductance, Eq. (10), for transport (from left to right lead) through the three-terminal ballistic quantum dot (see inset) with linear size L, leads of width W and spatially varying SOI α(x) = kso ¯α(x) = kso(y/L) (i.e. B = ∂yα = kso/L) with ksoL ≈ 0.3. RL = G↑ RL − G↓ with constant gradient over the whole conductor, α(x) = (y/L)kso, and check the prediction (10) that the spin conductance can be expressed in terms of the charge conductance of the transformed system without SOI but In Fig. 1, the spin conduc- with a magnetic field B. tance Gy RL (from now on the y-axis is the spin quantization axis) in the absence of magnetic field is compared to the difference of the charge conductance, GRL(B)−GRL(−B) in the absence of SOI, but with mag- netic field B = ∂yα. Both quantities exhibit precisely the same mesoscopic conductance fluctuations as a function of Fermi momentum, as predicted by Eq. (10). We found that this level of agreement holds up to ksoL ≈ 1, be- yond which terms quadratic and higher order in kso are no longer subdominant. For weak magnetic fields (with an associated cyclotron radius larger than L), Gij(B) − Gij (−B) is predomi- nantly given by quantum coherent contributions only. They give rise, on top of the mesoscopic fluctuations dis- played in Fig. 1, to a shift δG in the (energy) averaged conductance, known as weak localization correction. In the presence of a magnetic field, δG exhibits a damp- ing that is Lorentzian-like, δG(B) = δG(0)/(1 + ξB2), for chaotic ballistic cavities [13] with δG(0) ∼ (1/4)e2/h and ξ proportional to the dwell time in the cavity. Ac- cording to the prediction (10) for the two-terminal case, the presence of an external magnetic field B0 leads to a finite spin conductance Gy(B0) = G(B0+B)−G(B0−B), with B = ∂yα. Then its energy average is hGy(B0)i = δG(0) 1 + ξ(B0 + B)2 − δG(0) 1 + ξ(B0 − B)2 . (11) This line of reasoning is confirmed in Fig. 2(a) where numerically calculated spin conductances (symbols) for the chaotic cavity with linearly varying SOI are compared 3 Figure 2: Spin conductances for two-terminal geometries as a function of an applied magnetic field B0. (a) Average spin conductance of a chaotic cavity (inset) for four different strengths of a linearly varying SOI (same as inset Fig. 1) with ksoL ≈ 0.16, 0.33, 0.67 and 1.0 from bottom to top curve. Symbols with statistical errorbars mark numerical results for the average spin conductance, full lines depict the theoretical prediction (11). The grey dashed line shows predicted spin conductance maxima (from Eq. (11)) for varying gauge field. (b) corresponding on-to-off ratios hGy(B0)i / hGy(0)i. (c) spin resolved conductances G↑(↓)(Φ) for an AB ring (inset panel (d)) as a function of flux Φ = πR2B0, showing a shift ±B due to the gauge field B = ∇α arising from SOI α = (y/L)kso (d) resulting spin conductance Gy(Φ) of a with ksoL = 1. single AB ring. Inset panel (a): Sinai-type billard: linear size L, stopper disk with radius Ri = L/10, leads of width W = L/15 hosting 4 transverse channels. Inset panel (d): AB ring: radius R = L/2, width W = L/15 with 4 open channels. consider Alternatively, we to the prediction (11) (full lines). Figure 2(b) shows the corresponding on-off ratios hGy(B0)i/hGy(0)i. few-channel regular Aharonov-Bohm (AB) rings where kso-linear spin cur- rents can be turned on by a magnetic flux [7]. These systems exhibit large almost periodic AB conductance oscillations instead of the weaker, randomly-looking conductance fluctuations. In Fig. 2(c) we present nu- merically computed spin resolved conductances G↑(↓)(Φ) as a function of flux Φ = πR2B0 (in units of the flux quantum Φ0 = h/e) for an AB ring (inset panel (d)) in presence of the same linearly varying SOI as for the cav- ity. As expected, the conductance traces for the spin-up and -down channels are shifted against each other by ±B = ±∂yα. This shift gives rise to a finite B0-periodic spin conductance Gy = G↑−G↓ as displayed in Fig. 2(d). At B0 = 0, first order spin conductance is forbidden by the Onsager relation. Gy vanishes further for fields corresponding to Φ0, Φ0/2 and Φ0/4, where maxima and minima of the usual charge magnetoconductance occur. Maxima of Gy appear at points where the shifted spin resolved G↑(↓) have their minima. This holds for regular, or quasi-regular electronic dynamics which requires clean AB rings with few-channels. Of particular interest in the 4 probe our theory. Inhomogeneous SOI could, e.g., be realized through a top gate covering only part of the sys- tem. Additionally, a measurement protocol for spin cur- rents based on symmetries of charge transport through quantum point contacts [16] could be implemented. (ii) Inhomogeneous SOI is also a prerequisite for vari- ous specific proposals for spin splitting [17, 18] and ana- logues of the Stern-Gerlach effect [19]. Our theory pro- vides a rather general, common footing to interpret them. For instance, the Stern-Gerlach based spin separation, usually explained in terms of a Zeeman coupling in a non-uniform (in-plane) magnetic field (associated with Rashba SOI), finds its explanation in the opposite bend- ing of electron paths owing to the Lorentz force associ- ated with our gauge field ±B. (iii) Another gauge transformation, dual to ours, al- lows to transform a nonuniform Zeeman term into two de- coupled components with an additional gauge field [20]. (iv) While the spin conductance fluctuations are sim- ilar in a (phase coherent) diffusive system, its classical magnetoconductance has a linear in magnetic field con- tribution originating from the classical Hall effect. Thus in a diffusive system with inhomogeneous SOI, we expect a spin conductance with a nonzero average value propor- tional to the classical Hall conductance. This spin con- ductance can be estimated [21] as hGai ∼ (e2/h)(ksoℓ) where ℓ is the mean free path. We stress that hGai is based on a classical effect in that it is robust against ef- fects such as dephasing and temperature broadening. We thank M. Duckheim for carefully reading our manuscript, and D. Loss, J. Nitta and M. Wimmer for helpful conversations. This work was supported by TUBITAK under grant 110T841 and the funds of the Inonu chair (IA), by NSF under grant DMR- Erdal 0706319 and the Swiss Center for Excellence MANEP (PJ), and by DFG within SFB 689 (MS,KR). [1] J. Schliemann, J.C. Egues, and D. Loss, Phys. Rev. Lett. 90, 146801 (2003). [2] M. Duckheim et al., Phys. Rev. B 81, 085303 (2010). [3] L. Onsager, Phys. Rev. 38, 2265 (1931). [4] M. Buttiker, Phys. Rev. Lett. 57, 1761 (1986). [5] I.L. Aleiner and V.I. Fal'ko, Phys. Rev. Lett. 87, 256801 (2001). [6] P. W. Brouwer, J. N. H. J. Cremers, and B. I. Halperin, Phys. Rev. B 65, 081302 (2002). [7] Y. Tserkovnyak and A. Brataas, Phys. Rev. B 76, 155326 (2007). [8] For a comprehensive recent account see: I.V. Tokatly and E. Ya. Sherman, Ann. Phys. 325, 1104 (2010). [9] To give but one example, a nonuniversal behavior of spin currents has been pointed out for systems with universal charge current characteristics in: I. Adagideli et al., Phys. Rev. Lett. 105, 246807 (2010). [10] Our gauge transformation gives G(B) + G(−B) for the Figure 3: Onsager symmetry-based transistor action re- sulting from the difference in spin conductance of two- and three-terminal mesoscopic rings (insets panel (b)). (a) On -- off ratio and (b) separate spin conductances Gy 3T and Gy 2T for AB ring in three- and two-terminal mode as a func- tion of a spatially nonuniform SOI, α = (y/L)kso. (c) double-log plot of same data as in (b) (top and third sym- bol sequence) and of corresponding Gy 2T (second and fourth sequence) for a more generic nonuniform SOI α = kso cos2(2πx/L1) cos2(2πy/L2). 3T and Gy AB case are: (i) the magnitude of the spin conductance, which exceeds its value in chaotic systems by one to two orders of magnitude (compare the vertical axes scales in Fig. 1 and 2(d)), and (ii) the control one has over the spin conductance: Applying an integer or half-integer flux quantum gives the off state of our transistor, while the on state is recovered at B0 = ±[(Φ0/4)/(πR2) − B]. The on/off spin current ratio can be made arbitrarily large, as it exactly vanishes in the off state. As said above, kso-linear spin conductances can also be turned on by adding an additional terminal. As shown in Fig. 3(a,b) we find a difference of at least three or- ders of magnitude in spin conductance, Gy 3T, for two- and three-terminal rings. In panel (c) a double log representation of the data from (b) reveal the cubic vs. linear ksoL dependence of Gy 2T (top symbol sequence in (c)) and Gy 3T (third sequence from top) in line with our predictions. 2T vs. Gy So far we have considered linearly varying SOI. How- ever, our theory holds generally and works well also for more generic spatial dependence of the SOI. We confirm this by calculating Gy(ksoL) for a ring with SOI α(r) = kso cos2(2πx/L1) cos2(2πy/L2) with L/L1 = 15, L/L2 = 6 giving rise to SOI bumps on scales of the ring width. As demonstrated in Fig. 3(c) we recover again the linear vs. cubic scaling with ksoL for the two- and three-terminal setting (second and fourth symbol sequence from top), in full accordance with our theory. We conclude with a few remarks: (i) Mesoscopic rings based on InAs [14] or p-doped GaAs samples which are known to exhibit large and tun- able SOI [15] are excellent candidates to experimentally charge conductance, which is a linear function of ℓ−1 agreement with Ref. [6], regardless of the geometry. [11] E.I. Rashba, Sov. Phys. Solid State 2, 1109 (1960). [12] The transport calculations are performed using a recur- sive Green's function technique, see: M. Wimmer and K. Richter, J. Comp. Phys. 228, 8548 (2009). so , in [16] P. Stano and Ph. Jacquod, Phys. Rev. Lett. 106, 206602 (2011). [17] M. Khodus, A. Shekhter, and A.M. Finkel'stein, Phys. Rev. Lett. 92, 086602 (2004). [18] Q.-F. Sun and X.C. Xie, Phys. Rev. B 71, 155321 (2005). [19] J.-I. Ohe, M. Yamamoto, T. Ohtsuki, and J. Nitta, Phys. [13] H.U. Baranger, R.A. Jalabert, and A.D. Stone, Phys. Rev. B 72, 041308 (2005). Rev. Lett. 70, 3876 (1993). [20] V. Korenman, J.L. Murray, and R.E. Prange, [14] T. Bergsten, T. Kobayashi, Y. Sekine, and J. Nitta, Phys. Rev. Lett. 97, 196803 (2006). [15] B. Grbi´c et al., Phys. Rev. Lett. 99, 176803 (2007). Phys. Rev. B 16, 4032 (1977). [21] I. Adagideli et al., unpublished. 5
1512.02143	2	1512	2015-12-09T00:17:49	Phonon-assisted resonant tunneling of electrons in graphene-boron nitride transistors	[ "cond-mat.mes-hall" ]	We observe a series of sharp resonant features in the differential conductance of graphene-hexagonal boron nitride-graphene tunnel transistors over a wide range of bias voltages between $\sim$10 and 200 mV. We attribute them to electron tunneling assisted by the emission of phonons of well-defined energy. The bias voltages at which they occur are insensitive to the applied gate voltage and hence independent of the carrier densities in the graphene electrodes, so plasmonic effects can be ruled out. The phonon energies corresponding to the resonances are compared with the lattice dispersion curves of graphene-boron nitride heterostructures and are close to peaks in the single phonon density of states.	cond-mat.mes-hall	cond-mat	Phonon-assisted resonant tunneling of electrons in graphene-boron nitride transistors E.E. Vdovin1,5,6, A. Mishchenko2, M.T. Greenaway1, M.J. Zhu2, D. Ghazaryan2, A. Misra3, Y. Cao4, S. V. Morozov5,6, O. Makarovsky1, A. Patan`e1, G.J. Slotman7, M.I. Katsnelson7, A.K. Geim2,4, K.S. Novoselov2,3, L. Eaves1 1School of Physics and Astronomy, University of Nottingham NG7 2RD UK 2School of Physics and Astronomy, University of Manchester, Manchester, M13 9PL, UK 3National Graphene Institute, University of Manchester, Manchester, M13 9PL, UK 4Centre for Mesoscience and Nanotechnology, University of Manchester, M13 9PL, UK 5Institute of Microelectronics Technology and High Purity Materials, RAS, Chernogolovka 142432, Russia 6National University of Science and Technology "MISiS", 119049, Leninsky pr. 4, Moscow, Russia 7Radboud University, Institute for Molecules and Materials, Heyendaalseweg 135, 6525 AJ Nijmegen, The Netherlands (Dated: October 8, 2018) We observe a series of sharp resonant features in the differential conductance of graphene- hexagonal boron nitride-graphene tunnel transistors over a wide range of bias voltages between 10 and 200 mV. We attribute them to electron tunneling assisted by the emission of phonons of well-defined energy. The bias voltages at which they occur are insensitive to the applied gate voltage and hence independent of the carrier densities in the graphene electrodes, so plasmonic effects can be ruled out. The phonon energies corresponding to the resonances are compared with the lattice dispersion curves of graphene-boron nitride heterostructures and are close to peaks in the single phonon density of states. The discovery of the remarkable electronic properties of graphene [1, 2] has been followed by an upsurge of interest in other layered materials such as hexagonal boron nitride (hBN), the transition metal dichalcogenides and the III-VI family of layered semiconductors. These anisotropic layered materials have strong in-plane bonds of covalent character, whereas the inter-layer bonding arises from weaker van der Waals-like forces, so that crys- talline flakes, one or a few atomic layers thick, can be exfoliated mechanically from bulk crystals. These devel- opments have led to the realisation of a new class of ma- terials, van der Waals (vdW) heterostructures, in which nanoflakes of different materials are stacked together in an ordered way [3 -- 5]. These heterostructures possess unique properties which can be exploited for novel device applications. An example is a tunnel transistor in which a barrier is sandwiched between two graphene layers and mounted on the oxidised surface layer of a doped Si sub- strate [6, 7]. The tunnel current flowing between the two graphene layers can be controlled by applying a gate volt- age to the doped Si layer and arises predominantly from resonant processes in which the energy, in-plane momen- tum and chirality of the tunneling electron are conserved [8 -- 10]. Previous investigations of electron tunneling in a wide variety of metal-insulator diodes [11] and conventional semiconductor heterostructures [12] have demonstrated that electrons can tunnel inelastically, with the emission of one or multiple phonons. Atomically-resolved scanning tunneling spectroscopy measurements on mechanically- cleaved graphene flakes with a tuneable back gate have revealed the presence of phonon-assisted tunneling [13 -- 16]. The multi-component nature of our vdW het- erostructure gives rise to a more complex set of lattice dis- persion curves than for graphene [17 -- 21] and to phonon- assisted tunneling, as shown recently for a graphite-hBN- graphene transistor [22]. An impetus to the study of electron-phonon interactions is the recent discovery of su- perconductivity in graphene-based vdW heterostructures [23 -- 26]. Here we investigate tunnel transistors in which a ∼1 nm layer of hBN is sandwiched between monolayer graphene electrodes. We observe a series of sharp res- onant peaks in the electrical conductance over a wide range of bias voltage, gate voltage and temperature. This spectrum can be understood in terms of inelastic transi- tions whereby electrons tunneling through the hBN bar- rier emit phonons of different and well-defined energies between ∼ 12 and 200 meV, covering the range of lat- tice phonon energies in these heterostructures. The res- onances correspond closely to van Hove-like peaks in the single phonon density of states of the heterostructure, with the strongest peaks arising from the emission of low and high energy optical mode phonons. A schematic energy band diagram of our device and circuit is shown in the inset of Fig. 1(a). The bot- tom graphene layer is mounted on an atomically-flat hBN layer, placed above the silicon oxide substrate, and the active region of the device is capped with an hBN pro- tective top layer. The tunnel current, I, was measured as a function of the bias voltage, Vb, applied between the two graphene electrodes and the gate voltage, Vg, applied across the bottom graphene electrode and the doped Si gate electrode. Fig. 1(a) shows plots of differential tunnel conduc- tance, G(Vb) = dI/dVb, measured at a temperature of T = 4 K. The form of the G(Vb) curves is strongly depen- dent on Vg. Close to Vb = 0, G ≈ 0 at all gate voltages. With increasing Vb, the conductance increases in a series of well-defined steps, indicated by vertical arrows. We at- 2 FIG. 1. Differential conductance of Device 1 at T = 4 K. (a) G(Vb) for Vg = −30 V (red) to 30 V (blue) and inter- vals, ∆Vg = 2 V. Inset: schematic band diagram of Device 1 with bias, Vb and gate, Vg, voltages applied to the monolayer graphene electrodes which are separated by an hBN barrier of thickness d = 0.9 nm. µb,t are the chemical potentials of the bottom (b) and top (t) electrodes and F is the electric field across the barrier. A phonon assisted tunnel process is shown schematically. (b) Color map of dI/dVb for a range of Vg and Vb. tribute each step to inelastic phonon-assisted tunneling in which an electron emits a phonon and tunnels from close to the Fermi energy in one electrode to an empty state near the Fermi energy in the other electrode, with the emission of a phonon of well-defined energy, ωp. Fig. 1(b) shows a color map of G(Vb, Vg) in which some of these step-like features are discernible as a series of faint vertical striped modulations in the color map. Also visi- ble is a dark blue, X-shaped region in which G is small. This corresponds to the passage of the chemical potential through the Dirac point of the two monolayer graphene electrodes as Vb and Vg are varied; here the conductance is suppressed due to the small density of electronic states into which electrons can tunnel. Using an electrostatic model [7], which includes a small amount of doping in the bottom electrode (p-type, 2.5 × 1011 cm−2) as a fitting parameter, we determine the condition for the intersec- FIG. 2. (a) Measured plots αdG/dVb for Device 1 and (b) calculated grey scale map of αdG/dVb for T = 4 K, best fit to data in Fig. 2(a) using the model described in the text and in Table 1. Yellow dashed curves show when the chemical potential in a graphene layer intersects with the Dirac point in that layer. tion of the chemical potential with a Dirac point in each of the two graphene electrodes. The calculated loci of these intersections are shown by the yellow dashed line in Fig. 1(b); they correspond closely with the measured X-shaped low conductance region. The vertical stripes in the G(Vb, Vg) map are faint be- cause the step-like modulation in G(Vb) is only a small fraction of the total conductance. Most of the monotonic increase of G(Vb) with Vb can be partly eliminated by taking the second derivative, dG/dVb, which reveals the weak but sharp phonon-assisted resonances more clearly. Fig. 2(a) shows a grey scale contour map of αdG/dVb, where α = Vb/Vb = ±1. Here the phonon-assisted tun- neling features appear as easily discernible bright vertical stripes, indicated by arrows, at well-defined values of Vb, at which G(Vb) rapidly increases. These values are inde- pendent of gate voltage but their amplitudes at low Vb are significantly stronger at large positive and negative values of Vg. We can exclude the possibility that the fea- tures are plasmon-related as the sheet density, n, in both graphene electrodes is strongly dependent on Vg: n varies G (μS) 0123400.10.20.3-0.1-0.2-0.3-30-20-10010203054321Vb (V)Vg (V)(a)(b)G (μS) μbμteFdeVbeVgdVgVbIħωp-30-20-10010203020100-10-20α dG/dVb (μA V -2)Vg (V)00.10.20.3-0.1-0.2-0.3-30-20-10010203030200-10-20Vb (V)Vg (V)(a)(b)100α dG/dVb (arb. units)from ∼ 1012 cm−2 (holes) through zero to ∼ 1012 cm−2 (electrons) between Vg = −30 V and 30 V. Even though the plasma frequency of carriers in graphene varies rela- tively weakly with n (∼ n1/4), [27] plasmon-related reso- nances would have a significant gate voltage dependence which is not observed. Since the bias voltage values, Vb, of the weak reso- nant features are independent of gate voltage and are the same in both negative and positive bias, we can display them more clearly by averaging over all sixty of the mea- sured αdG/dVb plots between -30 V < Vg < 30 V. This procedure reduces significantly the level of background noise. The result of this averaging procedure for Device 1 is shown in Fig. 3(a). It reveals the phonon-assisted resonances as a series of well defined peaks. The corre- sponding plot for another device, Device 2, is also shown. The overall forms of the two curves are qualitatively sim- ilar, with the exception of some notable differences e.g. the position of the strong peaks at high Vb > 0.12 V. To understand the physical origin of the peaks in Fig. 3(a), we compare them to the one-phonon densities of states of monolayer graphene (red curve) and a graphene- hBN bilayer (green); the lower three curves show the par- tial density of states associated with the predominant motion of the carbon, boron and nitrogen atoms in the bilayer; the solid and dashed curves show contributions by in- and out of-plane phonons respectively. The full phonon dispersion curves of the graphene-hBN bilayer [17] are shown in Fig. 3(b). The phonon density of states and the phonon disper- sion curves were determined by using the "phonopy" package [28] with the force constants obtained by the finite displacement method [29, 30], using the Vienna ab initio simulation package (VASP) [31, 32]. For the phonon density of states a tetrahedron smearing was ap- plied for higher accuracy. A detailed description of the computational methods can be found in ref. [17]. At high bias, the two peaks labeled (i) and (ii) are close to the energies of the large densities of states associated with the weakly-dispersed, high energy optic phonons of monolayer graphene (Device 1) and a bilayer of graphene and hBN (Device 2). Note that peak (iii) at 130 meV in Device 2 is absent in Device 1. This energy corresponds closely to the flat region of the dispersion curve of the graphene-hBN bilayer near the K-point of the Brillouin zone, which arises predominantly from vibrations of the nitrogen atom. This difference, and the variation of the position of peaks (i) and (ii), between the two devices may arise from small differences in the relative lattice orientation of the graphene and hBN layers in the device. Both devices exhibit peaks around 110, 84 and 53 mV, labeled (iv), (v) and (vi), corresponding to prominent features in the calculated density of states plots and the flat regions of the dispersion curves. An interesting fea- ture of the data for both devices is the strong peak (viii) at low bias, around 12 mV, which we attribute to the weakly dispersed low-energy phonons close to the Γ-point of graphene-hBN. This low-energy "out-of-plane" mode 3 FIG. 3. (a) The three top curves (blue): dG/dVb averaged over all gate voltages for our model (dashed) and measured data for Devices 1 and 2 (both solid). Lower curves: total den- sity of phonon states for monolayer graphene (red, monolayer Gr) and a graphene-hBN heterostructure (green, Gr-hBN (to- tal)). Lower three curves: the partial density of states associ- ated with the carbon (red, C), boron (blue, B) and nitrogen (black, N) atoms of a graphene-hBN heterostructure. Solid and dashed curves show contributions by in- and out of-plane phonons respectively. (b) phonon dispersion of a graphene- hBN heterostructure [17]. Vertical dotted curves are guides to the eye highlighting the alignments. was intensively studied in inelastic X-ray spectroscopy measurement of bulk graphite and hBN [18]. Note that the resonant peak (vii) observed at 32 mV can be asso- ciated with a peak in the phonon density of states of the graphene-hBN bilayer which arises predominantly from the motion of the nitrogen atoms and corresponds to the flat region of the lowest-energy acoustic mode at ∼ 30 C B N monolayer Gr Device 1 Device 2 Vb (V)Gr-hBN (total)ΓMKΓDoS (arb. units)Phonon energy (eV)00.050.10.15BNC(a)(b)dG / dVb (μA V-2)0.050.10.150.200255000.5Model(ii)(i)(ii)(i)(iii)(iv)(iv)(v)(v)(vi)(vi)(viii)(viii)(vii)(vii)4 louin zone. As shown in Fig. 4, the resonant peaks broaden and their amplitudes decrease with increasing temperature, disappearing completely at temperatures T ≥ 150 K. This is consistent with the thermal broadening of the electron distribution functions around the Fermi ener- gies of the two graphene electrodes. Pauli blocking of electron tunneling for eVb < ωp between the graphene layers is diminished as more states become available with increasing thermal smearing around the Fermi energies. We fit the data in Fig. 2(a), using a model in which an inelastic tunneling transition becomes allowed only when the difference between the chemical potentials, µb in the bottom (b) and top (t) graphene lay- and µt, ers respectively, exceeds ωi p, which corresponds to the bias voltage of a particular phonon-assisted resonance peak, i, in the conductance. At low temperatures (4 K), eVb = µb − µt − eF d greatly exceeds the smearing, ∼ 2kBT , of the Fermi seas of the two graphene electrodes. The emission of a phonon of energy, ωi p, becomes possi- ble when eVb exceeds ωi p, thus opening up an inelastic scattering channel and giving rise to a step-like rise in the current and a resonant peak in dG/dVb when eVb = ωi p. In our model the current is given by b vs. Vb in De- FIG. 4. Temperature dependence of d2I/dV 2 vice 2 measured from T = 2.3 K to T = 180 K (blue to red) for Vg = 40 V. Inset: peak amplitude vs. T , curve colors match peaks marked by correspondingly colored arrows in the main plot. meV in the vicinity of the M- and K-points of the Bril- (cid:90) (cid:90) dEb T (i) (cid:88) i I = dEtDb(Eb)Dt(Et)(cid:8)Γ(Eb − Et − ωi p)fb(Eb)[1 − ft(Et)] − Γ(Et − Eb − ωi p)ft(Et)[1 − fb(Eb)](cid:9) (1) where Eb,t is the electron energy in the b and t layer, Db,t(E) is the density of electronic states in the b and t graphene layers (which are shifted energetically by eFbd), Γ(E) = exp(−E2/2γ2) with energy broadening γ = 5 meV and fb,t is the Fermi function in the bottom and top electrodes. T (i) is a relative weighting factor that de- pends on the electron-phonon coupling and phonon den- sity of states for each inelastic transition. We show the values of T (i) used in our model in Table I which pro- vides a qualitative indication of the relative strengths of the phonon-assisted processes. of Vb, we obtain the grey-scale plot in Figure 2(b), which simulates accurately with the measured data in Figure 2(a). In particular, the relative intensities of the vertical stripes are in good agreement with the measured data. At high positive and negative Vg, the asymmetry in the measured intensities of the resonances for positive and negative Vb is replicated by the model. This confirms that the peaks arise from phonon-assisted tunneling of carriers from filled states near the chemical potential in one electrode into the empty states just above the chem- ical potential in the other. TABLE I. Phonon energies, ωi p, and weighting factors, T (i), used in Eq. (1) to obtain the fit to the experimental data shown in Fig. 2(b) i 1 ωi p (meV) 12 2 32 3 53 4 5 6 7 8 84 110 143 174 201 T (i) 1.0 0.58 0.30 0.26 0.24 0.23 3.53 1.81 Using this model, and including phonon emission pro- cesses at threshold energies corresponding to the values In conclusion, our measurements reveal a rich spec- trum of inelastic phonon-assisted electron tunneling processes in monolayer graphene - hBN - monolayer graphene tunnel transistors. They allow us to probe electron-phonon interactions in this type of device and identify spectroscopically the energies and nature of the emitted phonons. Our results suggest that slight misori- entations of the component crystalline lattices of these vdW heterostructures may result in differences in the energies and intensities of the measured phonon-assisted tunnel transitions. −0.2−0.15−0.1−0.0500.050.10.150.200.20.40.60.811.21.41.61.8Vb(V) dG / dVb (μA V-2)050100150200012345peak: 1 2 3 4peak amplitude (a.u.)T (K)ACKNOWLEDGMENTS This work was supported by the EU FP7 Graphene Flagship Project 604391, ERC Synergy Grant, Hetero2D, EPSRC (Towards Engineering Grand Challenges and Fel- lowship programs), the Royal Society, US Army Research Office, US Navy Research Office and US Airforce Re- 5 search Office. M.T.G. acknowledges The Leverhulme Trust for support of an Early Career Fellowship. S.V.M. and E.E.V. were supported by NUST MISiS (grant K1- 2015-046) and RFBR (15-02-01221 and 14-02-00792). G.J.S. and M.I.K. acknowledges financial support from the ERC Advanced Grant No. 338957 FEMTO/NANO. We are grateful to Gilles de Wijs and Annalisa Fasolino for useful discussions. [1] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, Science 306, 666-669 (2004). [2] K.S. Novoselov, D. Jiang, F. Schedin, T.J. Booth, V. V Khotkevich, S. V Morozov, and A.K. Geim, Proc. Natl. Acad. Sci. USA 102, 10451 (2005). [3] C.R. Dean et al., Nat. Nanotechnol. 5, 722-726 (2010) [4] A.K. Geim and I.V. Grigorieva, Nature 499, 419-425, (2013). [5] A.C. Ferrari, et al., Nanoscale 7, 4598-4810 (2015) [6] Feenstra, R. M, Jena, D. and Gu, G. J. Appl. Phys. 111, 043711 (2012). [7] L. Britnell, et al., Science 335, 947 (2012). [8] Britnell, L. et al. Nat. Commun. 4, 1794, (2013). [9] A. Mishchenko, et al. Nat. Nanotechnol. 9, 808813 (2014) [10] M.T. Greenaway, et al., Nature Phys. 11, 1057 (2015) [11] A.G. Chynoweth, R.A. Logan, and D.E. Thomas, Phys. Phys. Rev. Lett. 98 095503 (2007) [19] J. Maultzsch, S. Reich, C. Thomsen, H. Requardt, and P. Ordej´on, Phys. Rev. Lett. 92 075501 (2004) [20] L. Wirtz and A. Rubio, Sol. St. Comm. 131 141-152 (2004) [21] A.C. Ferrari, et al., Phys. Rev. Lett. 97 187401 (2006) [22] S. Jung, M. Park, J. Park, T.-Y. Jeong, H.-J. Kim, K. Watanabe, T. Taniguchi, D.H. Ha, C. Hwang and Y.-S. Kim Sci. Rep. 5, 16642 (2015) [23] J. Chapman, Y. Su, C.A. Howard, D. Kundys, A. Grig- orenko, F. Guinea, A.K. Geim, I.V. Grigorieva and R.R. Nair, arXiv:1508.06931v1 (2015) [24] M. Xue, G. Chen, H. Yang, Y. Zhu, D. Wang, J. He, and T. Cao, J. Am. Chem. Soc. 134, 6536-6539 (2012). [25] M. Petrovi´c, et al., Nature Comm. 4, 2772 (2013). [26] D.M. Guzman, H.M. Alyahyaei, and R.A. Jishi, 2D Ma- terials 1, 021005 (2014). Rev. 125, 877 (1962). [27] E. H. Hwang and S. Das Sarma Phys. Rev. B 75, 205418 [12] L. Eaves, P.S.S. Guimaranes, B.R. Snell, D.C. Taylor, (2007) and K.E. Singer, Phys. Rev. Lett. 55, 262 (1985) [28] A. Togo, I Tanaka, Scr. Mater. 108, 1-5 (2015). [13] V.W. Brar, et al., Appl. Phys. Lett. 91 122102 (2007) [14] Y. Zhang, V.W. Brar, F. Wang, C. Girit, Y. Yayon, M. Panlasigui, A. Zettl and M.F. Crommie Nature Phys. 4, 627-630 (2008). http://phonopy.sourceforge.net/ [29] G. Kresse, J. Furthmller, and J. Hafner, Europhys. Lett. 32 (9), 729 (1995). [30] K. Parlinski, Z-.Q. Li, and Y. Kawazoe, Phys. Rev. Lett. [15] T. O. Wehling, I. Grigorenko, A.I. Lichtenstein, and A. 78 (21), 4063 (1997). V. Balatsky, Phys. Rev. Lett. 101, 216803 (2008) [31] G. Kresse and J. Furthmuller, Comp. Mater. Sci. 6 (1), [16] F.D. Natterer, et al., Phys. Rev. Lett. 114, 245502 (2015) [17] G.J. Slotman, G.A. de Wijs, A. Fasolino, and M.I. Kat- 1550 (1996). [32] G. Kresse and J. Furthmuller, Phys. Rev. B 54 (16), snelson, Ann. Phys. (Berlin) 526, 381-386 (2014). 11169 (1996). [18] J. Serrano, A. Bosak, R. Arenal, M. Krisch, K. Watan- abe, T. Taniguchi, H. Kanda, A. Rubio, and L. Wirtz,
1711.04509	1	1711	2017-11-13T10:41:09	Conditional electron confinement in graphene via smooth magnetic fields	[ "cond-mat.mes-hall" ]	In this article we discuss confinement of electrons in graphene via smooth magnetic fields which are finite everywhere on the plane. We shall consider two types of magnetic fields leading to systems which are conditionally exactly solvable and quasi exactly solvable. The bound state energies and wave functions in both cases have been found exactly.	cond-mat.mes-hall	cond-mat	a Conditional electron confinement in graphene via smooth magnetic fields Dai-Nam Le,1, 2, a) Van-Hoang Le,3, b) and Pinaki Roy4, c) 1)Atomic Molecular and Optical Physics Research Group, Ton Duc Thang University, Ho Chi Minh City, Vietnam 2)Faculty of Applied Sciences, Ton Duc Thang University, Ho Chi Minh City, Vietnam 3)Department of Physics, Ho Chi Minh City University of Pedagogy, 280 An Duong Vuong St., Dist. 5, Ho Chi Minh City, Vietnam 4)Physics and Applied Mathematics Unit, Indian Statistical Institute, Kolkata-700108, India (Dated: 1 October 2018) Abstract In this article we discuss confinement of electrons in graphene via smooth magnetic fields which are finite everywhere on the plane. We shall consider two types of magnetic fields leading to systems which are conditionally exactly solvable and quasi exactly solvable. The bound state energies and wavefunctions in both cases have been found exactly. I. INTRODUCTION In recent years graphene which is a sheet of carbon atom in honeycomb lattice1 -- 3 has drawn widespread attention because of its possible applications in various devices. The dynamics of charge carriers or electrons in graphene is described by the (2 + 1) dimensional massless Dirac equation, except that the a)Electronic mail: [email protected] b)Electronic mail: [email protected] c)Electronic mail: [email protected] 2 electrons move with the much smaller Fermi velocity vF = 106 m/s instead of the velocity of light c. For graphene to have practical applications one of the most important problem is controlling or confining the electrons. Attempts have been made to confine electrons e.g, by using position dependent mass4, modulating Fermi velocity5,6, electrostatic fields or magnetic fields. However, confinement using electrostatic fields is usually difficult although zero energy states7 -- 11and sometimes some states of non zero energy12 can be found us- ing different field configurations. On the other hand magnetic confinement of electrons has been studied by many authors. For example, square well mag- netic barrier13,14, radial magnetic field15, decaying gaussian magnetic field16, hyperbolic magnetic fields17, inhomogeneous magnetic fields18 -- 22, one dimen- sional magnetic fields leading to solvable systems23, etc. have been used to create bound states in graphene. In particular, of the different types of mag- netic fields mentioned above, there are some smooth inhomogeneous magnetic fields19 -- 21 for which the pseudo spinor components satisfy equations with quasi exactly solvable effective potentials24. In this context, it may be noted that in- homogeneous magnetic field profiles can be produced in many ways e.g, using ferromagnetic materials25, non planar substrate26, integrating superconduct- ing elements27 etc. In the present paper, our objective is to search for smooth everywhere finite magnetic fields which produce conditionally exactly solvable effective potentials28,29 i.e, potentials which admit exact solutions when pa- rameter(s) of the model assume particular values. More precisely, it will be shown that the electrons remain confined for certain values of the magnetic quantum number while for other values of hte magnetic quantum number they enter the deconfining phase. We shall also explore the possibility of obtaining quasi exactly solvable systems when some of the constraints on the param- eters are relaxed. The organization of the paper is as follows: in section II we shall present the formalism; in section III we shall obtain several magnetic fields which leads to conditionally exactly solvable systems; in section IV we shall examine under what conditions the magnetic fields produce quasi exactly solvable systems and finally section V is devoted to a conclusion. II. FORMALISM The dynamics of quasi particles in graphene is governed by the Hamiltonian 3 H = vF ~σ · ~π= vF ~σ ·(cid:16)~p + ~A(cid:17) = vF   0 Π− Π+ 0   , where vF is the Fermi velocity, σ = (σx, σy) are Pauli matrices, and Π± = πx ± iπy = (px + Ax) ± i(py + Ay). We now choose the vector potentials to be of the form Ax = yf (r), Ay = −xf (r) (1) (2) (3) where the specific form of the function f (r) will be chosen later. With the above choice of the vector potentials, the magnetic field is given by The eigenvalue equation Bz = −2f (r) − rf ′(r). Hψ = Eψ, where ψ = (ψ1, ψ2)T is a two component pseudospinor, can be written as Π−ψ2= ǫψ1, Π+ψ1= ǫψ2, (4) (5) (6) where ǫ = E/vF . Now eliminating ψ1 in favor of ψ2 (and vice-versa), the equations for the components can be written as Π− Π+ψ1= ǫ2ψ1, Π+ Π−ψ2= ǫ2ψ2. (7) (8) Since the magnetic field is a radial one, the pseudospinor components can be taken as ψ1 = eimθr−1/2φ1(r), ψ2 = ei(m+1)θr−1/2φ2(r), m = 0,±1,±2,· · · , (9) where m is the magnetic quantum number. Then eigenvalue equations for the components can be written as 4 d2 dr2 + d2 dr2 + (cid:20)− (cid:20)− 4 m2 − 1 r2 + r2f 2 − 2(m + 1)f − rf ′(cid:21) φ1 = ǫ2φ1, (m + 1)2 − 1 + r2f 2 − 2mf + rf ′(cid:21) φ2 = ǫ2φ2. r2 4 (10) (11) Before closing this section, we note that the intertwining relations (5) and (6) can also be written in terms of polar coordinates and are given by (cid:18) ∂ ∂r − (cid:18) ∂ + ∂r m + 1 2 r + rf(cid:19) φ1 = iǫφ2, r − rf(cid:19) φ2 = iǫφ1. m + 1 2 (12) The set of intertwining relations (12) is particularly important since knowing solution of one of the two equations (10) or (11), the other can be obtained through the above relations. III. CONDITIONALLY EXACTLY SOLVABLE MAGNETIC FIELDS Here we shall consider several conditionally exactly solvable magnetic field profiles i.e, magnetic fields for which all or some bound state solutions can be found only when the parameters of the model assume some specific values. To this end we choose the function f (r) to be of the form f (r) = λ 2 + 2gi 1 + gir2 , λ > 0, g1, g2, ..., gN > 0. N Xi=1 Then the resulting magnetic field is given by Bz(r) = −λ − 4gi (1 + gir2)2 . N Xi=1 (13) (14) From equation (14) it can be observed that the magnetic field is everywhere finite with a maximum value of −λ and a minimum of −λ−4PN i=0 gi. We shall now consider different values of N and examine if the corresponding magnetic field can support bound states when the parameters assume some particular values. A. N = 1 In this case the magnetic field becomes Bz(r) = −λ − 4g1 (1 + g1r2)2 , and the profile of this field can be seen in Fig 1. 5 (15) FIG. 1. Magnetic field profile for N = 1, λ = 1. Then, from (10) and (11) the equations for the components φ1 and φ2 can be obtained as m2 − 1 4 r2 + λ2r2 4 − 2Z − 4g1 1 + g1r2 − (m + 1)2 − 1 4 r2 + λ2r2 4 − d2 (cid:20)− dr2 + "− d2 dr2 + where Z = 2mg1 + λ. 8g1 (1 + g1r2)2(cid:21) φ1 =(cid:2)ǫ2 + λ(m − 1)(cid:3) φ1, 1 + g1r2# φ2 =(cid:2)ǫ2 + λ(m − 2)(cid:3) φ2, (17) (16) 2Z Conditional exact solutions: Let us now consider equation (17) for the lower component. This equation can be interpreted as the radial Schrodinger equation for a particle moving in a two dimensional nonpolynomial oscillator potential. Next, we choose the parameter g1 in such a way that the nonpoly- nomial part vanishes i.e, 1 Z = 0 ⇒ g1 = −λ/2m. (18) Now recalling that g1 and λ are always positive, the admissible values of m are m < 0. With g1 as given above, equation (17) becomes the radial Schrodinger 1 Note that if λ < 0, solutions can be obtained in a similar way for the sector m > 0 equation for the two-dimension isotropic harmonic oscillator : 6 d2 dr2 + (cid:20)− (M − 1)2 − 1 4 r2 + λ2r2 4 (cid:21) φ2 =(cid:2)ǫ2 − λ(M + 2)(cid:3) φ2, M = −m. (19) It may be pointed out the effective potential becomes that of the radial har- monic oscillator only when g1 assumes the particular value given by (18). The eigenvalues and the corresponding wave functions of (19) are standard and are given by : En,M = ±vFp2λ(n + M + 1), n = 0, 1, 2,· · · , M = 1, 2,· · · φ2(r) ∼ rM −1e−λr2/4LM −1 n (cid:0)λr2/2(cid:1) , (20) (21) n (x) is the associated Laguerre polynomial. Then, the lower compo- where LM nent of the pseudospinor wave function ψ2 is ψ2(r, θ) ∼ rM −1e−λr2/4LM −1 n (cid:0)λr2/2(cid:1) ei(1−M )θ. (22) From the intertwining relation (12), the upper component can be determined through the lower component (22) and we obtain the pseudospinor wave func- tion: ψ(r, θ) ∼ rM −1e−λr2/4e−iM θ × n,M λr(cid:20)LM iǫ−1 n (λr2/2) + eiθLM −1 n ×  2 λr2 + 2M LM −1 (λr2/2) n  (λr2/2)(cid:21)  . (23) It may be noted that as the magnetic quantum number decreases and be- comes M < 1, the electrons enter the deconfining phase and are no longer confined. In Figs 2 and 3 we have presented plots of the effective potentials in Eqs.(16) and (17) for Z = 0 and probability density for different values of the parameters. 7 l a i t n e t o P e v i t c e f f E 40 30 20 10 0 1 2 3 4 5 r FIG. 2. Plots of effective potentials in Eq.(16) (dotted curve) and Eq.(17) (solid curve) for λ = 2, M = 4 and Z = 0. y t i s n e D y t i l i b a b o r P 0 n,M=0,1 n,M=1,3 n,M=2,2 n,M=3,4 10 r FIG. 3. Plots of the probability density for some values of n and M . B. N = 2 Here we shall consider a more general magnetic field and put N = 2 in (13) and obtain: f (r) = λ 2 + 2g1 1 + g1r2 + 2g2 1 + g2r2 , (24) and the corresponding magnetic field is given by Bz(r) = −λ − 4g1 (1 + g1r2)2 − 4g2 (1 + g2r2)2 . 8 (25) Fig 4. shows the profile of the magnetic field for different values of the pa- rameters. FIG. 4. Magnetic field profile for N = 2, λ = 1. From (10) and (11), the equations for φ1 and φ2 can be obtained as + λ2r2 4 − 2Z1 − 4g1 1 + g1r2 − 8g1 (1 + g1r2)2 − 2Z2 − 4g2 1 + g2r2 − 8g2 (1 + g2r2)2(cid:21) φ1 (26) (m + 1)2 − 1/4 r2 + λ2r2 4 − 2Z1 1 + g1r2 − 2Z2 1 + g2r2 ,# φ2 =(cid:2)ǫ2 + λ(m − 4)(cid:3) φ2, r2 d2 dr2 + m2 − 1/4 (cid:20)− = [ǫ2 + λ(m − 3)] φ1, "− d2 dr2 + (27) where Z1 = 2mg1 + λ + 4g1g2/(g1 − g2) and Z2 = 2mg2 + λ + 4g2g1/(g2 − g1). Conditional exact solutions: As in the previous example, we consider equation (27) for the lower component. In this case, the nonpolynomial in- teraction is a more general one and consists of two terms representing the nonlinearities. However, the potential reduces to the two dimensional har- monic oscillator potential if Z1 = 0 = Z2 i.e, 2mg1 + λ + 4g1g2/(g1 − g2) = 0, 2mg2 + λ − 4g1g2/(g1 − g2) = 0. (28) The solution of the above set of coupled equation is given by g1 = g2 = λ 2(M + 1 + √M + 1) 2(M + 1 − √M + 1) λ > 0, > 0, 9 (29) where M = 1, 2, 3,· · · and is related to the the magnetic quantum number by m = −M. With the above choice of g1,2, equation (27) immediately becomes radial Schrodinger equation of an isotropic harmonic oscillator : d2 dr2 + (cid:20)− (M − 1)2 − 1/4 r2 + λ2r2 4 (cid:21) φ2 =(cid:2)ǫ2 − λ(M + 4)(cid:3) φ2. (30) The energy and corresponding eigenfunctions are given by En,M = ±vFp2λ(n + M + 2), n = 0, 1, 2· · · , M = 1, 2, 3,· · · φ2(r) ∼ rM −1e−λr2/4LM −1 n (cid:0)λr2/2(cid:1) . (31) (32) Then using the intertwining relation (12), the pseudospinor can be obtained as ψn,M (r, θ) ∼ rM −1e−λr2/4e−iM θ × n,M r(cid:20)λLM iǫ−1 ×  n (λr2/2) +(cid:18) 2g1 eiθLM −1 n 1 + g1r2 + 2g2 1 + g2r2(cid:19)LM −1 n (λr2/2) (33) .  (λr2/2)(cid:21)  As before the electrons remain confined for M ≥ 1. In Figs 5 and 6 we have plotted the effective potentials in Eq.(26) and Eq.(27) for Z1 = 0 = Z2 and probability density for several values of the parameters. 10 l a i t n e t o P e v i t c e f f E 40 30 20 10 0 1 2 3 4 5 r FIG. 5. Plots of effective potentials in Eq.(26) (dotted curve) and Eq.(27) (solid curve) for λ = 2, M = 4 and Z1 = 0 = Z2. y t i s n e D y t i l i b a b o r P 0 n,M=0,1 n,M=1,3 n,M=2,2 n,M=3,4 10 r FIG. 6. Plots of the probability density for some values of n and M . Before we conclude this section, let us examine degeneracy of the eigen- values. In both the cases considered above it is seen that the ground state E0,1 is non degenerate while all other states are degenerate. For example, E1,1 = E0,2, E1,2 = E2,1 = E0,3, E1,3 = E3,1 = E2,2 = E0,4 and so on. Thus degeneracy of the level En,M is (n + M). 11 IV. QUASI EXACT SOLUTIONS Here we shall explore whether or not the magnetic fields considered in the previous section may produce effective potentials which are quasi exactly solvable, that is, potentials for which only some solutions can be found ana- lytically. It will be seen that one may indeed find some exact solutions even when some of the constraints are relaxed. Before considering the quasi exact solutions for N = 1 and N = 2, we note that from here now the notation m + 1 2 − if m ≥ 0, m −(m + 1) if m < 0, is used to separate the cases of m ≥ 0 and of m < 0. M =(cid:12)(cid:12)(cid:12)(cid:12) 1 2(cid:12)(cid:12)(cid:12)(cid:12) =  A. N = 1. Let us first consider the case N = 1. In this case the pseudospinor com- ponents satisfy equations (16) and (17) and we obtained conditionally exact solutions under the condition Z = 0. The question which comes up imme- diately is the following: Can we still find bound states when Z 6= 0? Before answering this question, we would like to note that the parameter g1 should always be positive since for g1 < 0, the magnetic field becomes singular and the potentials V1(r) = V2(r) = λ2r2 4 − λ2r2 4 − 2Z − 4g1 1 + g1r2 − 2Z 1 + g1r2 8g1 (1 + g1r2)2 , (34) also become singular and they may not share the same spectrum. In order to obtain bound states, let us first consider the zero energy ones. For Z 6= 0 and g1 > 0 the zero energy solutions can be easily obtained from equation (12) for λ < 0 and they are given by E = 0, ψ0,M (r, θ) ∼ e−iM θ  0 rM (1 + g1r2)eλr2/4  , M = 0, 1, 2,· · · . (35) 12 On the other hand for λ > 0, the zero energy solutions are given by E = 0, ψ0,M (r, θ) ∼ eiM θ  rM e−λr2/4/(1 + g1r2) 0   , M = 0, 1, 2,· · · . (36) Note that in both the cases the solutions (1) are infinitely degenerate with respect to the quantum number M, (2) exist only for some values of the magnetic quantum number. To determine non zero energy solutions, we choose V2(r) and consider a wave function of the form N φ2 ∼ rm+1+1/2(1 + g1r2)e−λr2/4 cnr2n. Xn=0 Now substituting (37) in (17) we find after some calculations that30 EN ,M = ±vFpλ(2N + m + 1 + (m + 1)) − (λ + λ)(m − 2), φ2 ∼ rm+1+1/2(1 + g2r2)e−λr2/4 (−1)n(m + 1)! 2nn!(n + m + 1)! Dn(Z, E)r2n, N Xn=0 where Dn(Z, E) = a1 b1 c1 a2 0 b2 0 ... 0 · · · . . . . . . . . . . . . 0 c2 a3 ... . . . bn−1 0 · · · 0 cn−1 an (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) , (cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12) (37) (38) (39) (40) ak = 2g2k(k + m + 1) + Z + λ(N + 2 − k), bk = 2g2k(k + m + 1), ck = λ(N + 1 − k). Clearly for the energy levels (38) to be admissible ones Dn(Z, E) = 0 for some n > 1. However, we have not found any such solutions consistent with the constraint g1 > 0. B. N = 2 In this case, the potentials in Eqs. (26) and (27) are given by 8g2 8g1 λ2r2 4 − 2Z1 − 4g1 1 + g1r2 − (1 + g1r2)2 − 2Z2 − 4g2 1 + g2r2 − V1(r) = (1 + g2r2)2 , (41) 1 + g1r2 − A plot of these potentials are shown in Fig 7. V2(r) = λ2r2 4 − 2Z1 13 (42) 2Z2 1 + g2r2 . l a i t n e t o P e v i t c e f f E 25 20 15 10 5 0 1 2 3 4 5 r FIG. 7. Plots of effective potentials in Eq.(41) (dotted curve) and Eq.(42) (solid curve) for λ = 1, m = −3 and g1 = 0.3484878472, g2 = 0.6593973328. The zero energy states in this case can be found as in the previous example and for λ < 0 are given by E = 0, ψ0,M (r, θ) ∼ e−iM θ rM (1 + g1r2)(1 + g2r2)eλr2/4  0  , M = 0, 1, 2,· · · , (43) while for λ > 0 are given by E = 0, ψ0,M (r, θ) ∼ eiM θ  rM e−λr2/4/(1 + g1r2)(1 + g2r2) 0   , M = 0, 1, 2,· · · (44) In order to find whether any of the potentials above can support quasi exact solutions, we first choose V2(r) and consider Z1 = 0. In this case, we obtain g2 = g1(2mg1 + λ) (2m − 4)g1 + λ . (45) Then equation (27) becomes (m + 1)2 − 1/4 + "− d2 dr2 + r2 λ2r2 4 − 2Z2 1 + g2r2# φ2 =(cid:2)ǫ2 + λ(m − 4)(cid:3) φ2, (46) where Z2 is given by Z2 = 4mg1(cid:18)1 + 2g1 2(m − 2)g1 + λ(cid:19) + 2λ. 14 (47) It may be noted that equations (17) and (46) look quite similar although they can not be identified with one another since Z 6= Z2. The procedure for obtaining non zero energy states is similar to the previous case and the results are EN ,M = ±vFpλ(2N + m + 1 − (m + 1) + 8) + (λ − λ)(m − 4), φ2 ∼ rm+1+1/2(1 + g2r2)e−λr2/4 (−1)n(m + 1)! 2nn!(n + m + 1)! Dnr2n, N Xn=0 (48) (49) where the expression for Dn reads exactly as in (40) except that one has to make the change Z → Z2. The bound state solutions can be obtained from the condition Dn(Z2, E) = 0 (50) subject to the condition g1, g2 > 0. In general, this condition is a polynomial equation for g2 with degree of (N + 1). Some specific solutions can be obtained as follows: for λ > 0 and m = −M − 1 < 0 (M = 0, 1, 2, . . . ), one can find exact solutions for all non- negative integers N . For example, for N = 0, the admissible values of g1 and g2 are g1 = 3λ 2M + 2 , g2 = 3λ 2M + 5 . The energy and the pseudospinor wave function now are E = ±2vF√2λ, φ2 ∼ rM e−λr2/4(cid:0)3λr2 + 2M + 5(cid:1) e−iM θ, iE−1r(cid:18)3λ2r2 + (2M + 11)λr + (3λr2 + 2M + 5) and ψ(r, θ) ∼ rM e−λr2/4e−iM θ×  18 3λr2 + 2M + 2(cid:19) e−iθ .   15 In general for N > 0, it is difficult to solve equation (50) analytically. So we have solved it numerically and a sample of the results for λ = 1 are shown in Table I. TABLE I. Allowed values of g1, g2 and exact energy values E for m = −1,−2,−3,−4,−5 and N = 1, 2, 3. m = −M − 1 g1 g2 N = 1 E nodes −1 −2 −3 −4 −5 −1 −2 −3 −4 −5 −1 −2 −3 −4 −5 0.4743416490E + 00 0.1790569415E + 01 ±0.3162277660E + 01 0.3535533906E + 00 0.8535533906E + 00 ±0.3464101615E + 01 0.2834733548E + 00 0.5538126093E + 00 ±0.3741657387E + 01 0.2371708245E + 00 0.4081138830E + 00 ±0.4000000000E + 01 0.2041241452E + 00 0.3224744871E + 00 ±0.4242640687E + 01 N = 2 0.5715576511E + 00 0.2077086572E + 01 ±0.3464101615E + 01 0.4058817282E + 00 0.9555757090E + 00 ±0.3741657387E + 01 0.3162345421E + 00 0.6068766798E + 00 ±0.4000000000E + 01 0.2596159465E + 00 0.4408038959E + 00 ±0.4242640687E + 01 0.2204606934E + 00 0.3446849045E + 00 ±0.4472135955E + 01 N = 3 0.6674965423E + 00 0.2361139093E + 01 ±0.3741657387E + 01 0.4573940766E + 00 0.1056561074E + 01 ±0.4000000000E + 01 0.3484878472E + 00 0.6593973328E + 00 ±0.4242640687E + 01 0.2817316368E + 00 0.4731709640E + 00 ±0.4472135955E + 01 0.2365739000E + 00 0.3666868582E + 00 ±0.4690415760E + 01 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 V. CONCLUSION In this paper we have proposed electron confinement in graphene using smooth magnetic fields which are finite everywhere. Interestingly, when the parameters of the model are subjected to certain constraints the magnetic fields lead to systems which are conditionally exactly solvable. It has also been shown that when these constraints are relaxed it is still possible to determine part of the spectrum analytically, especially the zero energy states. Depending on the orientation of the magnetic field, these states can be found for some values of the magnetic quantum number. For non zero energy states in some cases the algebraic part of the calculations become quite cumbersome and we 16 have obtained the eigenvalues numerically in such cases. It may mentioned that we have examined two values of N in (14) but it is possible to consider higher values of N. We believe it would be interesting to investigate whether solutions in closed form can be found for a general value of N and if so, what the constraints on the parameters may be. Finally we would like to mention that although we have considered massless electrons it is possible to carry out the entire analysis when a mass term of the form ∆σz is present in the Hamiltonian (1). In such a case the model can be used to study other Dirac materials like silicene31. ACKNOWLEDGMENTS One of the authors (PR) wishes to thank AMOG, Ton Duc Thang Uni- versity, Ho Chi Minh City for supporting a visit during which this work was carried out. This research is funded by the Vietnam National Foundation for Science and Technology Development (NAFOSTED) under Grant No. 103.01-2014.44. References 1K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, and A. A. Firsov, Nature (London). 438, 197 (2005). 2Y. Zheng, J. W. Tan, H. L. Stormer, and P. Kim, Nature (London). 438, 201 (2005). 3A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, Rev. of Mod. Phys. 81, 109 (2009). 4N. M. R. Peres, A. H. Castro Neto, and F. Guinea, Phys. Rev. B73, (2006). 5P. Ghosh and P. Roy, Eur. Phys. J. Plus, (2017) 132: 32. 6C. A. Downing and M. E. Portnoi, J. Phys:Condens Matter. 29, (2017) 315301. 7C. A. Downing, D. A. Stone, and M. E. Portnoi, Phys. Rev. B84, 155437 (2011). 8R. R. Hartmann, N. J. Robinson, and M. E. Portnoi, Phys. Rev. B81, (2010) 245431. 9C. L. Ho and P. Roy, Euro. Phys. Lett. 108, (2014) 20004; ibid. 112, (2015) 47004. 10P. Ghosh and P. Roy, Phys. Lett. A380, (2016) 567. 17 11C. A. Downing, A. R. Pearce, R. J. Churchill, and M. E. Portnoi, Phys. Rev. B92, (2015) 165401. 12R. R. Hartmann and M. E. Portnoi, Phys. Rev. A89, (2014) 012101. 13A. De Martino, L. Dell'Anna, and R. Egger, Phys. Rev. Lett. 98, (2007) 066802. 14L. Dell'Anna and A. De Martino, Phys. Rev. B79, (2009) 045420. 15G. Giavaras, P. A. Maksym, and M. Roy, J.Phys:Condens Matter. 21, (2009) 102201. 16T. K. Ghosh, J. Phys:Condens Matter. 21, (2008) 045505. 17E. Milpas, M. Torres, and G. Murguia, J. Phys:Condens Matter. 23, (2011 245304. 18M. Ramezani Masir, P. Vasilopoulos, and F. M. Peeters, J. Phys:Condens Matter. 23, (2011) 315301. 19P. Roy, T. K. Ghosh, and K. Bhattacharya, J. Phys:Condens Matter. 24, (2012) 055301. 20C. A. Downing and M. E. Portnoi, Phys. Rev. B94, (2016) 165407. 21C. A. Downing and M. E. Portnoi, Phys. Rev. B94, (2016) 045430. 22M. Eshghi, H. Mehraban and I. Ahmadi Azar, Physica. E94, (2017) 106. 23S. Kuru, J. Negro, and L. M. Nieto, J. Phys:Condens Matter. 21, (2009) 455305. 24A. V. Turbiner, Comm. Math. Phys. 118, (1988) 467. 25A. Matulis, F. M. Peeters and P. Vasilopoulos, Phys. Rev. Lett. 72, (1994) 1518. 26M. L. Leadbeater et al., Phys. Rev. B 52, (1995) R8629. 27S.J. Bending, K. von Kiitzing and K. Ploog, Phys. Rev. Lett. 85, (1990) 1060. 28A. de Souza Dutra, Phys. Rev. A47, (1993) R2435. 29G. Junker and P. Roy, Phys. Lett. A232, (1997) 155; Ann. Phys. 270, (1998) 155. 30Dai-Nam Le, Ngoc-Tram D. Hoang, and Van-Hoang Le, J. Math. Phys. 58, (2017) 042102. 31M. Tahir and U. Schingenschlogl, Scientific Reports. 3, (2013) 1075.
1710.07259	1	1710	2017-10-19T17:32:03	AC transport in Correlated Quantum Dots: From Kondo to Coulomb blockade regime	[ "cond-mat.mes-hall" ]	We explore the finite bias DC differential conductance of a correlated quantum dot under the influence of an AC field, from the low-temperature Kondo to the finite temperature Coulomb blockade regime. Real-time simulations are performed using a time-dependent generalization of the steady-state density functional theory (i-DFT) [Nano Lett. {\bf 15}, 8020 (2015)]. The numerical simplicity of i-DFT allows for unprecedented long time evolutions. Accurate values of average current and density are obtained by integrating over several periods of the AC field. We find that (i) the zero-temperature Kondo plateau is suppressed, (ii) the photon-assisted conductance peaks are shifted due to correlations and (iii) the Coulomb blockade is lifted with a concomitant smoothening of the sharp diamond edges.	cond-mat.mes-hall	cond-mat	AC transport in Correlated Quantum Dots: From Kondo to Coulomb blockade regime G. Stefanucci1, 2 and S. Kurth3, 4 1Dipartimento di Fisica, Universit`a di Roma Tor Vergata, Via della Ricerca Scientifica 1, 00133 Rome, Italy; European Theoretical Spectroscopy Facility (ETSF) 2INFN, Laboratori Nazionali di Frascati, Via E. Fermi 40, 00044 Frascati, Italy 3Nano-Bio Spectroscopy Group and European Theoretical Spectroscopy Facility (ETSF), Dpto. de F´ısica de Materiales, Universidad del Pa´ıs Vasco UPV/EHU, Av. Tolosa 72, E-20018 San Sebasti´an, Spain 4IKERBASQUE, Basque Foundation for Science, Maria Diaz de Haro 3, E-48013 Bilbao, Spain We explore the finite bias DC differential conductance of a correlated quantum dot under the in- fluence of an AC field, from the low-temperature Kondo to the finite temperature Coulomb blockade regime. Real-time simulations are performed using a time-dependent generalization of the steady- state density functional theory (i-DFT) [Nano Lett. 15, 8020 (2015)]. The numerical simplicity of i-DFT allows for unprecedented long time evolutions. Accurate values of average current and density are obtained by integrating over several periods of the AC field. We find that (i) the zero- temperature Kondo plateau is suppressed, (ii) the photon-assisted conductance peaks are shifted due to correlations and (iii) the Coulomb blockade is lifted with a concomitant smoothening of the sharp diamond edges. Over the last decades there have been tremendous advances in manufacturing nanoscale devices such as nanotubes, artificial atoms (quantum dots), or single molecules attached to metallic leads [1 -- 4]. The theo- retical description of their transport properties often re- mains a challenge since it involves the physics of strong correlations out of thermal equilibrium. Even the (min- imal) single impurity Anderson model [5] (SIAM), after almost sixty years is still a matter of intense research and debate. The development of clever analytical and numerical techniques has provided us with a fairly good understanding of its equilibrium and steady state prop- erties. Unfortunately, the same cannot be said for time- dependent responses. In particular, very little is known about the behavior of the average current when an AC field is superimposed to a DC voltage. What is the fate of the zero-temperature Kondo plateau in the conductance? How does the charging energy affect the photon-assisted conductance peaks? Is the finite-temperature Coulomb blockade lifted? In this Letter we provide an answer to these and related fundamental questions. One way to access AC (linear as well as nonlinear) transport properties consists in performing numerical simulations of the time-evolution of the system. Sev- eral time-propagation algorithms have been put for- ward. Time-dependent (TD) density matrix renor- malization group [6 -- 10], TD numerical renormalization group [11], functional renormalization group [12 -- 14], real-time Monte Carlo [15 -- 19], hierarchical equations of motion approach [20], iterative real-time path inte- gral [21] and real-time effective action [22] have all been used to investigate the transient response to a sudden quench of bias or gate voltages, and a reasonable agree- ment between them has been reached. These methods, however, are limited to short propagation times and AC responses are difficult to address since convergence often requires averaging the TD current over several periods of the driving field. An alternative to time-propagation is the Floquet Green's function approach. SIAM responses to an oscillating gate have been reported in Refs. [23, 24] using either a second-order self-energy (reasonable only at the particle-hole symmetric point) or an analytically interpolated equilibrium self-energy (questionable for AC voltages). To the best of our knowledge, no attempts to calculate the SIAM conductance under AC voltages have been made so far. Lately, strongly correlated systems have been studied also with density functional theory (DFT) both in its static [25 -- 29] and time-dependent version [30 -- 35]. In a recent work we proposed a steady-state density func- tional theory [36] (i-DFT) to calculate the steady density on and the steady current through a quantum junction sandwiched between metallic leads. In Ref. [37] we ap- plied i-DFT to the SIAM and found excellent agreement with numerically exact methods in a wide range of bias and gate voltages, from weak to strong charging energies U and from low to high temperatures T . We here extend i-DFT to the time domain, thereby laying down a computationally feasible scheme to shed light on the AC transport properties of the SIAM. Our main findings are (i) AC voltage suppression of the T = 0 Kondo plateau (ii) interaction-induced shift of the photon-assisted conductance peaks and (iii) lifting of the finite-temperature Coulomb blockade with concomitant smoothening of the diamond shape. Time-Dependent i-DFT - i-DFT establishes a one-to- one correspondence between the pair density on and cur- rent through the impurity, (n, I), and the pair gate and bias voltages (v, V ). Accordingly, there exists a unique pair (vs, Vs) that in the noninteracting SIAM produces the same density and current as in the interacting one. The pair (vs, Vs) consists of the Kohn-Sham (KS) gate vs = v + vH + vxc and KS voltage Vs = V + Vxc where the Hartree potential vH = U n whereas the exchange- correlation (xc) potentials vxc and Vxc are universal func- tionals (i.e., independent of the external fields v and V ) of both density and current. Knowledge of these function- als allows for calculating (n, I) by solving self-consistently the equations n = 2 Xα=L,R Z dω 2π f (ω − Vs,α)Aα(ω) (1) I = 2Z dω 2π(cid:2)f (ω − Vs,L) − f (ω − Vs,R)(cid:3)T (ω) (2) where Vs,L = −Vs,R = Vs/2 is the bias applied to the left (L) and right (R) leads and f (ω) = 1/(eβ(ω−µ) + 1) is the Fermi function at inverse temperature β = 1/T and chemical potential µ. In Eq. (1) the partial spectral function is given by Aα(ω) ≡ G(ω)Γα(ω)G†(ω) with the (retarded) KS Green's function G(ω) = 1/(ω − v − vHxc − ΣL(ω) − ΣR(ω)) and broadening Γα(ω) = −2Im [Σα(ω)], Σα being the embedding self-energy of lead α. The cur- rent in Eq. (2) is expressed in terms of the KS transmis- sion coefficient T (ω) = AL(ω)ΓR(ω). The self-consistent solution of Eqs. (1) and (2) is extremely efficient [37, 38]. It is therefore natural to extend i-DFT to the time domain and explore dynamical (as opposed to steady- state) transport properties. Setting the external bias V (t) = θ(t)[V + VAC sin(Ωt)] we can provide a full char- acterization of the AC conductance. In the same spirit of the adiabatic approximations in standard TD-DFT [39, 40] we calculate the TD density n(t) and current I(t) by propagating the KS SIAM with TD gate potential vs(t) = v(t) + vHxc[n(t), I(t)] and volt- age Vs(t) = V (t) + Vxc[n(t), I(t)]. The propagation is performed by solving the coupled TD KS equation i d dt ψL,k ψC,k ψR,k     =   0 hLL hLC hCL hCC hCR hRC hRR 0     ψL,k ψC,k ψR,k   (3) for every KS state ψk. Equation (3) is written in a block form where the blocks L and R refer to the left and right leads whereas the block C refers to the impurity site, hence hCC = vs(t). We model the leads as semi-infinite tight-binding chains with nearest neighbor hopping tlead and onsite energy Vs,α(t). Only the boundary site of the semi-infinite chains is connected to the impurity and the corresponding hopping amplitude is tlink. We take the leads at half-filling (µ = 0), choose both tlead and tlink much larger than any other energy scale and set the ratio t2 link/tlead = γ/2 to stay in the wide band limit (WBL). Equation (3) is solved using a generalization of the al- gorithm of Ref. 41 to finite temperatures. The time prop- agation is performed with a predictor-corrector scheme at each time step for vHxc and Vxc. The TD occupation and current to plug into the xc potentials are obtained from the KS wavefunctions according to n(t) = 2Xk f (ǫk)ψC,k(t)2, (4) I(t) = IL(t) − IR(t) 2 . The current at the α interface Iα(t) = 4Xk f (ǫk)Im(cid:2)tlinkψ∗ α,k(t)ψC,k(t)(cid:3) 2 (5) (6) is expressed in terms of the wavefunction ψα,k(t) at boundary site of lead α. Notice that current conserva- tion implies IL(t) + IR(t) + n(t) = 0 at all times. For symmetric voltages VL(t) = −VR(t) = V (t)/2 and at the particle-hole symmetric point (ph -- SP) v = −U/2 we have IL(t) = −IR(t). In Ref. [37] we proposed an accurate parametrization of vHxc and Vxc. This parametrization has been shown to agree well with results from the functional and numerical renormalization group in a wide range of temperatures and interaction strengths. We have used the xc potentials of Ref. [37] to solve Eq. (3). A somewhat related current- dependent approximation for vHxc, valid for only one lead and at temperatures higher than the Kondo temperature, has recently been suggested in Ref. 42. Results - We consider the SIAM initially (times t ≤ 0) in thermal equilibrium and then driven out of equilibrium by a symmetric bias VL(t) = −VR(t) = V (t)/2. The bias is the sum of a DC and an AC contribution, i.e., V (t) = V + VAC sin(Ωt) for t > 0. (7) τ = 2π/Ω, i.e., IDC = R tℓ+N τ After a sufficiently long propagation time tℓ the transient features die out and all physical observables become pe- riodic functions of time. We then calculate the DC com- ponent of the current by averaging I(t) over N periods I(t)/(N τ ). Depending on the parameters, a good convergence may require N of the order of 10 or larger. Finally we calculate the finite-bias DC conductance G = dIDC/dV , highlighting its most rel- evant features as VAC and Ω are varied. We consider over two physically distinct regimes: the Kondo regime at zero temperature and the Coulomb blockade (CB) regime at temperatures larger than the Kondo temperature. tℓ First we checked that for U = V = 0 our TD algorithm agrees with the exact (in the WBL approximation) non- interacting formula [43] GU=0 = ∞ Xk=−∞ J 2 k (cid:18)− VAC Ω (cid:19) γ 2 4 (v − kΩ)2 + γ 2 4 (8) where Jk(x) is the Bessel function of order k. Without interactions photon-assisted transport (PAT), i.e., charge transfer accompanied by the emission or absorption of photons, is the only scattering mechanism. Turning on the interaction the scenario changes dra- matically, especially for gates v in the range (−U, 0) where correlation effects are enhanced. In Fig. 1 we show results for the zero-bias DC conductance as function of 1 T/U = 0.0 T/U = 0.125 0 G G / 0,5 0,2 0,1 0 G G / 0 G G / 0,4 0,2 VAC/U = 0.0 VAC/U = 0.5 VAC/U = 0.5 ad VAC/U = 1.0 VAC/U = 1.0 ad VAC/U = 1.5 VAC/U = 1.5 ad 0 -2 -1 v/U 0 1 -1 v/U 0 0 1 0 -2 Ω/U = π/8 Ω/U = π/8 NI Ω/U = π/4 Ω/U = π/4 NI Ω/U = π/2 Ω/U = π/2 NI 3 0,4 0,2 0 G G / VAC/U = 0.5 -1 v/U 0 VAC/U = 1.0 1 -1 0 v/U 0 1 FIG. 1. Zero-bias DC conductance (solid lines) as well as its adiabatic counterpart of Eq. (9) (dashed) versus v for different AC amplitudes. The SIAM parameters are U/γ = 4, Ω/U = π/4 and temperature T = 0 (left panel) and T /U = 0.125 (right panel). G0 = 1 π is the quantum of conductance. v for different VAC. At temperature T = 0 (left panel) and already for small VAC the Kondo plateau is dras- tically suppressed. The DC conductance resembles the one at VAC = 0 in the CB regime with two peaks sepa- rated by roughly the charging energy U and a minimum at the ph -- SP. The suppression of Kondo correlations has already been observed in Refs. [44, 45] using a perturba- tive approach as well as in Refs. [23, 24] where, rather than an AC bias, the gate potential is perturbed harmon- ically. As VAC increases, however, the minimum is first converted into a plateau and eventually into a maximum with the concomitant washing out of the side peaks. The decrease and subsequent increase of the Kondo peak at the ph -- SP predicted by our calculations is in agreement with experimental findings reported in Ref. [46]. In the CB regime (T /U = 0.125, right panel) a small AC bias barely changes the values of G at VAC = 0. However, for VAC ≈ U the DC conductance develops a plateau and a maximum at the ph -- SP for even larger amplitudes. In both regimes, see left and right panels, we also observe increasingly higher side peaks at v ≈ Ω and v ≈ −U − Ω with increasing VAC, a clear signature of PAT. We will discuss the precise position of these peaks below. To highlight nonadiabatic (memory) effects we define the "adiabatic" DC conductance according to Gad = 1 τ Z τ 0 dt GDC(V (t)), (9) where GDC(V ) = dI/dV VAC=0 is the finite-bias DC con- ductance at vanishing AC amplitude. The results for Gad are shown in both panels of Fig. 1 (dashed lines). For small VAC and around the ph -- SP, Gad is remarkably close to G both in the Kondo and the CB regime while away from ph -- SP the two quantities only share qualita- tive features at best. For larger VAC, adiabatic and non- FIG. 2. Zero-bias DC conductance versus v for different AC frequencies and for amplitudes VAC/U = 0.5 (left panel) and VAC/U = 1.0 (right panel). Other parameters as in Fig. 1. 0,8 0,12 0,6 0,1 0 G G / 0 G G / 0,4 0,08 0,2 2,8 v/γ 3,2 U/γ = 0.0 U/γ = 2.0 U/γ = 2.0 NI KS U/γ = 4.0 U/γ = 4.0 NI KS γ / c x H v 0 3 2 1 0 -8 2 n 1 0 -8 -4 v/γ 0 4 -6 -4 -2 v/γ 0 2 4 FIG. 3. Upper panel: Zero-bias DC conductances (solid) ver- sus v for Ω/γ = π, VAC/γ = 4.0 and different values of U . Dashed lines are the noninteracting result of Eq. (8) with v replaced by the time-averaged KS potential vs. Inset: zoom in around v = Ω showing shifts in the peak position of the PAT peak. Lower panel: time-averaged Hxc gate and density (inset). adiabatic DC conductances increasingly differ, pointing to the importance of memory effects. In Fig. 2 we show the dependence of the zero-bias G at T = 0 on the AC frequency for two different AC ampli- tudes. At the small amplitude VAC/U = 0.5 (left panel) the behavior of G is qualitatively similar for different frequencies (suppression of the Kondo plateau and aris- ing of PAT peaks). Interestingly, for the larger ampli- tude VAC/U = 1.0 (right panel), G develops a plateau around the ph -- SP for all frequencies, with a height that decreases monotonically with Ω. For this larger am- plitude the PAT peaks emerge more distinctly and we clearly appreciate the nonlinear (in VAC) effect of ab- sorption/emission of two photons (for Ω/U = π/8). The dashed curves starting at v = 0 in both panels 0 0.25 0.5 0.75 1 4 a) b) c) d) U V / 1 0 -1 -2 -1 0 v/U -2 -1 0 v/U -2 -1 0 v/U -2 -1 0 v/U 1 FIG. 4. DC conductance in the (v, V ) plane at zero temperature for a) VAC/U = 0.0, b) VAC/U = 0.5, c) VAC/U = 1.0 and d) VAC/U = 1.5. Other parameters as in Fig. 1. of Fig. 2 are the values of GU=0, see Eq. (8) (which is an even functions of v), exhibiting PAT peaks at v = kΩ with k integer [43]. For v > 0, the interacting and noninteracing results agree reasonably well while for v < 0 correlations shift the PAT peaks by approximately −U . It is worth noting that in the interacting case the PAT peak at v ≈ Ω is slightly shifted to lower gates. To shed some light on this shift we calculated the zero-bias G for various charging energies U . The results are shown in the upper panel of Fig. 3 where the inset is a magnification of the PAT peak at v ≈ Ω. For these gate values correlations are weak as confirmed by the small value n ≈ 0.15 of the time-averaged density, see inset in the lower panel. In the weakly correlated regime the DC conductance is well approximated by the noninteracting formula of Eq. (8). However, from the i-DFT perspective, the v in Eq. (8) is not just the bare gate but instead the time-averaged KS gate vs = v + vHxc, i.e., the sum of the bare gate and the time-averaged Hxc potential vHxc shown in the lower panel of Fig. 3. Consequently, the PAT peak in G occurs at v = Ω−vHxc, i.e., it is shifted to lower energies. This is precisely the shift we observe in our calculations. In fact, if we evaluate Eq. (8) making the replacement v → vs we get a remarkable good agreement with the interacting G (including position and height of the PAT peaks) for v outside the interval (−U, 0), see dashed curves in upper panel of Fig. 3. Obviously, this good agreement breaks down for gates in the interval (−U, 0) since the system can no longer be considered weakly correlated. So far we presented results for the zero-bias DC con- ductance. However, the proposed method is not limited to this very special case. In Fig. 4 we display the finite- bias G as function of v and V for different amplitudes of the AC field. The calculations are done at temperature T = 0. To appreciate the effects of the AC field the well- known pure DC result, i.e., VAC = 0, is shown in panel a). We correctly reproduce the CB diamond as well as the Kondo resonance at V = 0. A moderate amplitude of the AC field, panel b), leads to a suppression of the Kondo resonance (see also Fig. 1). Moreover, starting from the finite bias corners at v/U = −1/2 and V /U = ±1, the CB is lifted inside the diamond. As VAC increases fur- ther, panels c) and d), the CB is lifted everywhere in the diamond, leaving for the largest amplitude, panel d), an area of increased conductance around the ph -- SP at V = 0. We also find that the side peaks due to PAT seen in Figs. 1 and 2 branch at finite bias into straight lines with negative and positive slopes, reminiscent of the typical G lines at VAC = 0. Not surprisingly, these lines become more pronounced as VAC increases. In summary, we have investigated electron transport through a correlated quantum dot subjected simultane- ously to DC and AC biases by explicit and numerically efficient time propagation in a DFT framework. For the corresponding Hxc gate and xc bias potentials we have suggested a time-local approximation based on recently proposed accurate i-DFT functionals for the steady state. We find that in the Kondo regime already for small AC amplitude the Kondo plateau in the zero-bias DC con- ductance is strongly suppressed while in the CB regime the changes are less pronounced. The observed shift of the photon-assisted transport peaks due to electronic in- teractions can readily be explained within DFT. At finite DC bias, Coulomb blockade is lifted with increasing AC amplitude and the CB diamond is deformed. G.S. acknowledges funding by MIUR FIRB Grant No. RBFR12SW0J, EC funding through the RISE Co-ExAN (GA644076), and funding through the INFN-Nemesys project. S.K. acknowledges funding by a grant of the "Ministerio de Economia y Competividad (MINECO)" (FIS2016-79464-P) and by the "Grupos Consolidados UPV/EHU del Gobierno Vasco" (IT578-13). [1] G. Cuniberti, G. Fagas, and K. Richter, Introducing Molecular Electronics (Springer, Heidelberg, 2005). [2] Y. V. Nazarov and Y. M. Blanter, Quantum Trans- port: Introduction to Nanoscience (Cambridge Univer- sity Press, Cambridge, 2009). [3] J. C. Cuevas and E. Scheer, Molecular Electronics: An Introduction to Theory and Experiment (World Scientific, London, 2010). [4] I. Baldea, Molecular Electronics: An Experimental and Theoretical Approach (Pan Stanford, 2015). [5] P. W. Anderson, Phys. Rev. 124, 41 (1961). [6] K. A. Al-Hassanieh, A. E. Feiguin, J. A. Riera, C. Busser, and E. Dagotto, Phys. Rev. B 73, 195304 (2006). [7] L. G. G. V. Dias da Silva, F. Heidrich-Meisner, A. E. Feiguin, C. A. Busser, G. B. Martins, E. V. Anda, and E. Dagotto, Phys. Rev. B 78, 195317 (2008). [8] F. Heidrich-Meisner, A. E. Feiguin, and E. Dagotto, Phys. Rev. B 79, 235336 (2009). [9] M. Nuss, M. Ganahl, H. G. Evertz, E. Arrigoni, and W. von der Linden, Phys. Rev. B 88, 045132 (2013). [10] S. Kirino, T. Fujii, J. Zhao, and K. Ueda, J. Phys. Soc. Jpn. 77, 084704 (2008). [11] F. B. Anders and A. Schiller, Phys. Rev. Lett. 95, 196801 (2005). [12] S. G. Jakobs, M. Pletyukhov, Phys. Rev. B 81, 195109 (2010). [13] J. Eckel, F. Heidrich-Meisner, M. Thorwart, M. Pletyukhov, New J. Phys. 12, 043042 (2010). and H. Schoeller, S. G. Jakobs, and R. Egger, [14] C. Karrasch, V. Meden, and K. Schonhammer, Phys. Rev. B 82, 125114 (2010). and Muhlbacher [15] L. E. Rabani, Phys. Rev. Lett. 100, 176403 (2008). [16] A. E. Antipov, Q. Dong, and E. Gull, Phys. Rev. Lett. 116, 036801 (2016). [17] P. Werner, T. Oka, M. Eckstein, and A. J. Millis, Phys. Rev. B 81, 035108 (2010). [18] P. Werner, T. Oka, and A. J. Millis, Phys. Rev. B 79, 035320 (2009). [19] A. Droghetti and I. Rungger, Phys. Rev. B 95, 085131 (2017). [20] Y. Cheng, W. Hou, Y. Wang, Z. Li, J. Wei, and Y. Yan, New J. Phys. 17, 033009 (2015). [21] S. Weiss, J. Eckel, M. Thorwart, and R. Egger, Phys. Rev. B 77, 195316 (2008). [22] S. Bock, A. Liluashvili, and T. Gasenzer, Phys. Rev. B 94, 045108 (2016). 5 [23] R. L´opez, R. Aguado, G. Platero, and C. Tejedor, Phys. Rev. B 64, 075319 (2001). [24] R. L´opez, R. Aguado, G. Platero, and C. Tejedor, Phys. Rev. Lett. 81, 4688 (1998). [25] K. Schonhammer, O. Gunnarsson, and R. M. Noack, Phys. Rev. B 52, 2504 (1995). [26] N. A. Lima, M. F. Silva, L. N. Oliveira, and K. Capelle, [27] F. Phys. Rev. Lett. 90, 146402 (2003). P. Phys. Rev. Lett. 109, 246402 (2012). Malet and Gori-Giorgi, [28] J. Lorenzana, Z.-J. Ying, and V. Brosco, Phys. Rev. B 86, 075131 (2012). [29] A. Mirtschink, M. Seidl, and P. Gori-Giorgi, Phys. Rev. Lett. 111, 126402 (2013). [30] C. Verdozzi, Phys. Rev. Lett. 101, 166401 (2008). [31] S. Kurth, G. Stefanucci, E. Khosravi, C. Verdozzi, and E. K. U. Gross, Phys. Rev. Lett. 104, 236801 (2010). [32] A.-M. Uimonen, E. Khosravi, A. Stan, G. Stefanucci, and E. K. U. Gross, S. Kurth, R. van Leeuwen, Phys. Rev. B 84, 115103 (2011). [33] A. Kartsev, D. Karlsson, A. Privitera, and C. Verdozzi, Sci. Rep. 3, 2570 (2013). [34] M. J. P. Hodgson, Chapman, P. Lillystone, Phys. Rev. B 88, 241102 (2013). J. D. Ramsden, J. B. J. and R. W. Godby, [35] M. J. P. Hodgson, J. D. Ramsden, and R. W. Godby, Phys. Rev. B 93, 155146 (2016). [36] G. Stefanucci and S. Kurth, Nano Lett. 15, 8020 (2015). [37] S. Stefanucci, Kurth and G. Phys. Rev. B 94, 241103(R) (2016). [38] S. Kurth and G. Stefanucci, J. Phys.: Condens. Matter 29, 413002 (2017). [39] C. Ullrich, Time-Dependent Density-Functional Theory (Oxford University Press, Oxford, 2012). [40] N. T. Maitra, J. Chem. Phys. 144, 220901 (2016). [41] S. Kurth, G. Stefanucci, C.-O. Almbladh, A. Rubio, and E. K. U. Gross, Phys. Rev. B 72, 035308 (2005). [42] N. Dittmann, J. Splettstoesser, arXiv:cond-mat/1706.04547 (2017). [43] A.-P. Jauho, N. S. Wingreen, and N. Helbig, and Y. Meir, Phys. Rev. B 50, 5528 (1994). and Goldin [44] Y. Phys. Rev. Lett. 81, 5394 (1998). [45] Y. Goldin and Phys. Rev. B 61, 16750 (2000). Y. Y. Avishai, Avishai, [46] J. M. Elzerman, S. De Franceschi, D. Goldhaber- Gordon, W. G. van der Wiel, and L. P. Kouwenhoven, J. Low Temp. Phys. 118, 375 (2000).
1209.2977	2	1209	2012-10-09T09:53:16	The Leggett-Garg inequality in electron interferometers	[ "cond-mat.mes-hall", "quant-ph" ]	We consider the violation of the Leggett-Garg inequality in electronic Mach-Zehnder inteferometers. This set-up has two distinct advantages over earlier quantum-transport proposals: firstly, the required correlation functions can be obtained without time-resolved measurements. Secondly, the geometry of an interferometer allows one to construct the correlation functions from ideal negative measurements, which addresses the non-invasiveness requirement of the Leggett-Garg inequality. We discuss two concrete realisations of these ideas: the first in quantum Hall edge-channels, the second in a double quantum dot interferometer.	cond-mat.mes-hall	cond-mat	The Leggett-Garg inequality in electron interferometers Clive Emary,1 Neill Lambert,2 and Franco Nori2, 3 1Institut fur Theoretische Physik, Technische Universitat Berlin, D-10623 Berlin, Germany 2Advanced Science Institute, The Institute of Physical and Chemical Research (RIKEN), Saitama 351-0198, Japan 3Physics Department, University of Michigan, Ann Arbor, Michigan, 48109, USA (Dated: August 29, 2018) We consider the violation of the Leggett-Garg inequality in electronic Mach-Zehnder inteferome- ters. This set-up has two distinct advantages over earlier quantum-transport proposals: firstly, the required correlation functions can be obtained without time-resolved measurements. Secondly, the geometry of an interferometer allows one to construct the correlation functions from ideal negative measurements, which addresses the non-invasiveness requirement of the Leggett-Garg inequality. We discuss two concrete realisations of these ideas: the first in quantum Hall edge-channels, the second in a double quantum dot interferometer. PACS numbers: 03.65.Ud, 73.23.-b, 03.65.Ta, 42.50.Lc Bell inequalities set bounds on the nature of the cor- relations between spatially-separated entities within local hidden variable theories1,2. In contrast, Leggett-Garg in- equalities (LGIs) set bounds on the temporal correlations of a single system3,4, and are derived under the assump- tions of macroscopic realism (MR) and non-invasive mea- surability (NIM)5 . Bell and Leggett-Garg inequalities are related in that their assumptions both imply the existence of a classi- cal probability distribution that determines experimental outcomes. The probability amplitudes of quantum me- chanics allow for violation of these inequalities: with Bell, the violation is due to entanglement between the two sys- tems; with Leggett-Garg, the violation occurs due to the superposition of system states and their collapse under measurement. The simplest LGI, henceforth referred to as the LGI, reads K ≡ C21 + C32 − C31 ≤ 1, (1) where Cαβ = hQ(tα)Q(tβ)i is the correlation function of the dichotomous variable Q = ±1 at times tα and tβ. Since the first experimental violation6 of this inequality with weak measurements of a superconducting qubit, the Leggett-Garg inequality has been experimentally probed in systems as diverse as photons7 -- 9, defects in diamonds centers10, nuclear magnetic resonance11, and phosphorus impurities in silicon12. Whilst the subjects of these stud- ies may not be macroscopic, the LGI performs a useful role for microscopic systems as an indicator that the de- vice is operating beyond classical probability laws. More- over, if one accepts that the alternative to classical proba- bilities is quantum mechanics, the LGI provides a decisive indicator of the "quantumness" on a system13. In this paper, we are interested in the violation of the LGI in quantum transport, and in particular, in electron- interferometers. Although there has been much work on Bell inequalities in electron transport, e.g. Refs 14 -- 22, the LGI has only relatively recently been consid- ered in this setting23,24. Specifically, the charge flow- ing through a confined nanostructure, e.g. double quan- tum dot (DQD), has been shown to violate an inequal- ity similar to Eq. (1) out of equillibrium23. Further- more, the moment-generating function of charge trans- ferred through a device has also been shown to be subject to a set of LG-style inequalities, which are violated for various quantum dot models. The violation of LGIs in excitonic transport has also attracted recent interest25,26. There are several difficulties which make the investi- gation of the LGI in electronic transport challenging in practice. Ostensibly, the measurement of Eq. (1) requires time-resolved measurements where the time between suc- cessive measurements is smaller than the decoherence time of the system. For the double quantum dot of Ref. 23, for example, this decoherence time is of the order of 1ns27, which makes the necessary time-resolved mea- surements very challenging (but, in principle, possible28). Furthermore, for the violation of Eq. (1) to be a mean- ingful indicator of non-classical behaviour, it must be ensured that the measurements are non-invasive. This "clumsiness loophole"29 that allows violations of Eq. (1) to be associated with invasiveness of measurement, along with possible circumventions, have been the subject of much discussion3,26,30. The transport set-ups we consider here are based on the electronic Mach-Zehnder Interferometer (MZI), and can overcome both of these problems. The basic idea is that an electron travelling through a MZI can take one of two paths, and this path index defines the variable Q = ±1. Unidirectional passage of the electron through the system allows us to map the time indices of Eq. (1) onto positions within the interferometer. As we show below, this removes the need for time-resolved measurements. We consider two realisations of the MZI in which measurements of Q are performed in two different ways. In the first, the MZI is formed from quantum Hall edge channels, a set-up which has been realised experimentally31 -- 38 and also attracted a large degree of theoretical attention39 -- 51 . By interrupting the edge channels at various points and diverting electron flow to current meters, we show that K can be obtained from measurements of mean currents alone. Furthermore, due to the spatial separation of the Q = +1 and Q = −1 channels, our detectors interact with only one of the two Q-states at any given time. Thus, our scheme provides a natural way to implement ideal negative measurements, as advanced by Leggett and Garg as a way to satisfy the NIM criterion3. The second set-up we consider is a MZI with a quan- tum dot (QD) in each arm. This geometry is similar to several experiments52 -- 56 that have investigated transport through Aharanov-Bohm rings with QDs in the arms. The difference here being that the dots are fed by two tunnel-coupled leads57, rather than just one. The QDs are monitored by quantum point contacts, whose trans- mission is sensitive to the charge state of the QD58 -- 63. In contrast with the first set-up, electrons are not diverted out of the MZI at any point, and the influence of the de- tectors occurs as a pure dephasing effect. The three cor- relation function in Eq. (1) are obtained through a com- bination of mean currents, both through the MZI and the quantum point contacts, and zero-frequency noise mea- surements, which cross-correlate current fluctuations in the MZI and quantum point contacts. As in the previ- ous scheme, we construct an ideal negative measurement scheme with this set-up. Both of these techniques exploit a combination of su- perpositions of paths through an interferometer com- bined with a gathering of "which-way" information to violate the LGI. The first set-up is a particularly simple realisation of the LGI, and is by no means restricted to transport, but could be used e.g. with photons, atoms or molecules. The paper proceeds as follows. In Sec. I we describe the basics of testing the LGI in a MZI. Sec. II describes how this may be translated into experiments with quantum Hall effect edge-channels. Finally, Sec. III considers the alternative DQD-QPC geometry studied here. I. MACH-ZEHNDER INTERFEROMETER We begin by describing an abstract version of our MZI scheme to outline the basic ideas. The MZI is a two- channel interferometer with two beam-splitters that di- vide the MZI into three zones which we label: 1, the input ports; 2, the arms of the interferometer; and 3, the output ports (see Fig. 1). We inject one electron at a time into the MZI and the path taken by the electron will be the degree-of-freedom under test with Q = +1 when the electron is located in the upper channel of the MZI, and Q = −1 the lower. Since the electron passes sequentially through the three zones, we can map a mea- surement of Q at time tα to a "which-way" measurement at any point in the region α of the interferometer. In particular, Q1 and Q3 are measured at the input and output ports, and Q2 is measured by placing detectors in the arms of the interferometer at 2±, where 2 refers to the zone, and ± the upper or lower channel. In this section we assume that we have ideal single-electron detectors 2 (a) 1+ (cid:84) A 1− (b) 3+ 3− 2+ 2− (cid:84) B (c) FIG. 1. (Color online) The Mach-Zehnder interferome- ter with three different detector configurations for the non- invasive measurement of the LGI of Eq. (1). Electrons are injected into the 1+ port. (a) Complete MZI configuration with detectors only at the final outputs 3±. With this set-up we can measure the probabilities P D 3±(1) and construct C31. (b) An additional detector is inserted into the MZI + arm. With this configuration we can measure probabilities P D 2+(0) and P D 3±;2+(1, ·). (c) A detector in the '−' arm allows us to obtain P D 3±;2−(1, ·). Combining the results of (b) and (c) allows us to construct correlation functions C21 and C32. 2−(0) and P D that "click" on detecting an electron, which is then re- moved from the system (i.e., the detectors act essentially as electronic analogues of photodetectors). More realis- tic measurements in terms of currents are discussed in section II. A. Ideal negative measurements A detector placed in one of the arms interacts strongly with electrons in that path (they are completely removed from the MZI) and has no effect on electrons in the other. With a detector placed at Q = +1, say, then the ab- sence of a detector response (combined with MR and ideal detectors) allows us to infer the state of the sys- tem (Q = −1) without any disruption. This is exactly the form of detector required to perform an ideal negative measurement as envisioned in Ref. 3. To make the measurement scheme as simple as possi- ble, let us inject electrons into the 1+ port, such that the initial state is known64. We do not need to mea- sure in zone 1 and there is no question about the NIM of Q1. The correlation function C21 and C31 boil down to measuring hQ2i and hQ3i respectively. α±(n) as the probability that the detec- tor placed at position α± either detects (n = 1) or fails to detect (n = 0) the electron. Since no further measure- ments are made past point 3, it is irrelevant whether we measure non-invasively or not at point 3. Placing detec- tors at 3±, we measure the probabilities P D 3±(1), and the Let us define P D C31 correlation function can simplify be expressed as B. Leggett-Garg Inequality 3 C31 = P D 3+(1) − P D 3−(1). (2) The set-up for this measurement is shown in Fig. 1a. Since, in measuring C21, no further measurements are made after region 2, it is also not necessary to measure C21 non-invasively. We can measure hQ2i (and thus C21) by running the experiment once with a detector in chan- nel 2+, and once in 2− (Fig. 1b and c) and writing C21 = P D 2+(1) − P D 2−(1). (3) It is perhaps instructive to discuss how to obtain this quantity using the ideal negative measurement technique and measure C21 in terms of the probabilities of absence of detector clicks. With the detector at 2+, we can equate the probability that no electron is detected, P D 2+(0), with the probability that the electron travels the path 2−. Swapping the detector to the other arm, we measure P D 2−(0) and infer the probability that the electron takes path 2+. Whence, we obtain the non-invasively mea- sured C21 = P D 2−(0) − P D 2+(0). (4) Since P D same result. 2±(0) = 1 − P D 2±(1), Eq. (3) and Eq. (4) give the We now consider C32, where it is essential that we mea- sure Q2 non-invasively, since a subsequent measurement is performed. On the face of it, measuring C32 requires a correlation measurement between two detectors. This, however, is not the case, as we now show. Let us begin by placing one detector at 2+ and another one at 3+ (Fig. 1b). We can then obtain the four prob- abilities, P D 3+,2+(n, n′), that the detectors at 3+ and 2+ give the results n, n′ = 0, 1 respectively. Of these, the one we are interested in is P D 3+,2+(1, 0), since this allows us to infer (non-invasively) the probability that the electron took path 2− to detector 3+. Moreover, we do not actu- ally need to actively detect at 2+, since, if the electron reaches the 3+ detector, it is clear that it has not en- tered channel 2+ (because the detector there would have removed the electron from the system)65. With all four probabilities, P D 3q,2q′ (1,·), obtained in this non-invasive way, we can construct C32 = − Xq,q′=± qq′P D 3q,2q′ (1,·), (5) where we have replaced the measurement value at posi- tion 2 with a dot to indicate that we do not actually have to measure there (the value is guaranteed to be zero). In this way we obtain all the required correlation functions, measured non-invasively where necessary. Al- though we have concentrated on the simplest case here, the above non-invasive techniques are extensible to the case where the input state is unknown and all Cαβ must be measured in a non-invasive way, or to more compli- cated LGIs11. The action of the MZI can be specified by two beam- splitter scattering matrices sX ; X = A, B. With aαq the annihilation operator for an electron in channel αq, the beamsplitter input-output relations read a2− (cid:19) = sA(cid:18) a1+ (cid:18) a2+ a1− (cid:19) ; (cid:18) a3+ a3− (cid:19) = sB(cid:18) a2+ a2− (cid:19) .(6) Parameterizing the scattering matrices as sX = (cid:18) cos( 1 2 θX ) 2 θX )e−i 1 2 φX − sin( 1 2 φX sin( 1 2 θX )ei 1 cos( 1 2 θX ) (cid:19) , (7) we obtain the correlation functions C21 = cos θA; C31 = cos θA cos θB − sin θA sin θB cos φ; C32 = cos θB, (8) (9) (10) such that the LG correlator reads K(θA, θB, φ) = cos θA + cos θB − cos θA cos θB + sin θA sin θB cos φ, (11) with φ = 1 2 (φA − φB) being the phase difference accu- mulated between the two paths. This is a familiar ex- pression. If we identify θA = Ωτ1 and θB = Ωτ2, then Eq. (11) is exactly that obtained for a qubit evolving un- der the Hamiltonian H = 1 2 Ωσx measured in the σz basis at times t1, t2 = t1 + τ1, and t3 = t2 + τ2. The properties of Eq. (11) are discussed in Sec. II II. QUANTUM HALL EDGE-CHANNELS Quantum Hall edge channels have been shown to al- low a direct translation of the MZI into electronic trans- port experiments31 -- 38 and Fig. 2 shows a sketch of the quantum Hall geometry needed to realise our proposal. Each channel in the MZI is realised with a single edge- channel and the electronic beam-splitters are realised by quantum point contacts (QPCs). Backscattering is sup- pressed between edge-channels such that transport is uni- directional. This set-up is the same as the MZIs of ex- periment except for the addition of extra contacts to the arms of the interferometer. These contacts are connected to the edge-channels via adjustable quantum point con- tacts, such that the detectors can be coupled into and out of the MZI as required. This method of coupling probes to the MZI arms has been realised in Ref. 38. Port 1+ is raised to a voltage +V and electrons are in- jected into this channel. The output ports (detectors) are all grounded. When the correlation function C31 is being measured, the detectors at 2± are not required and are isolated from the MZI by closing their QPCs. (Fig. 2 shows detector 2− closed off in this way). To measure the remaining correlation functions, the detectors at 2± are, one then the other, connected into the MZI by opening 2+ V 1+ sA 2- 2+ 3- 2- 3+ 3+ sB 3- FIG. 2. (Color online) Quantum Hall edge-channel real- isation of the MZI set-up for measurement of Legget-Garg inequality. The MZI set-up is similar to that of Ref. 31 but with two extra detectors (2±). These additional detectors can be isolated from the circuit by closing off the QPCs between them and the edge channel. The configuration shown has the detector at the 2+ position active such that transmission to beamsplitter B via channel 2+ is blocked, and the detector at 2− is pinched off. This detector combination corresponds to that of Fig. 1b. up their respective QPCs. In Fig. 2, the detector at 2+ is connected into the circuit and fully prevents electrons in channel 2+ from reaching the outputs 3±. A. Current measurements Let hIαqi be the mean stationary current flowing into output αq, given that when α = 3, the detectors at po- sitions 2± are closed off. Further, let hI3q;2q′i be the current flowing at output 3q when the output at 2q′ is open. Since, in the linear regime, the current operator αqaαq, with G0 = e2/h for each output is Iαq = G0V a† the conduction quantum66, these mean currents are pro- portional to the probability that an electron travels in the corresponding channel. The correlation functions re- quired for the LGI can then be constructed, as with the CHSH inequality15,16,67, as ; Cα1 = hIα+i − hIα−i hIα+i + hIα−i C32 = −P qq′hI3q;2q′i PhI3q;2q′i . (12) (13) Division by the sum of detector currents removes pro- portionality factors and, if all detector are identical, also removes detector inefficiencies. Writing the scattering matrices as (a) φ= 0 φ = π/4 φ = π/2 (b) ∆ = 0 ∆ = π/2 ∆ = π ∆ = 3π/2 ∆ = 2π 4 1.4 K 1.2 0.5 1 ϕ/π 1.5 0 0.5 1 φ/π 1.5 1 2 1 0 K -1 -2 -3 0 FIG. 3. (Color online) (a) LG correlator K(θA, θB, φ) of Eq. (11) as a function of the beamsplitter angle θ = θA = θB for three values of the phase φ = 0, π/4, π/2. The shaded blue region indicated violation of the LG inequality (K > 1). The maximum violation, Kmax = 3 2 , occurs for φ = 0 and, e.g., θ = π/3. (b) The influence of dephasing. Shown is the LG correlator K of Eq. (16) maximized over θA/B as a function of the phase φ. Results shown for values of the dephasing parameter ∆/π = 0, 1 2 , 2. Violations of the LG are only observed for cos φ > 0. 2 , 1, 3 we obtain the correlation functions C21 = rA2 − tA2; BtA2; C31 = rBrA + t′ C32 = rA2(cid:8)rB2 − tB2(cid:9) − tA2(cid:8)t′ BtA2 − tBrA + r′ B2 − r′ B2(cid:9) .(15) With scattering matrices as in the previous section, the LG parameter K obtained from current measurements is the same as Eq. (11). This quantity is plotted in Fig. 3a. A maximum violation of Kmax = 3 2 is obtained for pa- rameters θA = θB = π/3 and φ = 0. The violation the LGI in this set-up arises because the measurements at 2± remove electrons from the in- terferometer arms, preventing interference between the two paths. The presence of this interference in C31 com- bined with its absence in C32 leads to the violation. B. Dephasing We can account for the effects of dephasing by allowing the phase φ to fluctuate. We replace φ → φ + δφ in Eq. (11) and integrate δφ over a flat distribution in the range −∆/2 < δφ < ∆/2. The resulting LG parameter with dephasing reads K deph = cos θA + cos θB − cos θA cos θB +f (∆) sin θA sin θB cos φ, (16) sX = (cid:18) rX t′ X tX r′ X (cid:19) , (14) with f (∆) = 2∆−1 sin(∆/2) the function containing the dephasing effects68. If all angles are freely variable then the maximum of this function is K deph max (∆) = 1 + f (∆)(1 + f (∆)) 1 + f (∆) , (17) obtained for cos θA = cos θB = [1 + f (∆)]−1. Expanding 32 ∆2. In the op- for small ∆, we find K deph posite limit, where the dephasing is total, ∆ → 2π, we have f (∆) → 0 and the maximised Leggett-Garg corre- lator reverts to the classical value, lim∆→2π K deph max = 1 as required. max (∆) = 3 2 − 1 One interesting feature occurs if we assume that the phase φ is fixed (e.g., we are not able to vary the magnetic field) and maximise over θA/B (see Fig. 3b). Providing that cos φ > 0, the maximum value is K deph max(θA/B)(φ, ∆) = f (∆) cos φ + 1 1 + f (∆) cos φ ,(18) found by setting cos θA = cos θB = [1 + f (∆) cos φ]−1. If, however, cos φ ≤ 0, the maximum value is just the classical value, K deph max(θA/B) = 1, found by setting cos θA = cos θB = 1. This reversion to the classical value occurs when the scalar product between the axis of the rotation of beamsplitter B and that of beamsplitter A becomes negative. C. Multi-channel case The above scheme is easily modified to include multi- ple channels. We take the same geometry as before but assume that each lead supports M channels. The M channels of the upper lead are all associated with qubit state Q = +1; the M channels in the lower lead, with state Q = −1. The scattering matrices of Eq. (14) are thus generalised to 2M×2M matrices with M×M blocks, rX , tX , r′ X . Assuming a large source-drain volt- age, such that all channels are equally populated, the correlation functions read: X , and t′ C21 = 1 C32 = 1 C31 = 1 + 1 A (R′ B − T ′ B − T ′ B)o ; B)o M Tr {RA − TA} ; M TrnR† M TrnR† M TrnrAt† A (RB − TB) + T † A (RB − TB) − T † †rB − r′ A(cid:16)r† Bt′ with RA = r† AtA, etc. The second term in the expression for C31 arises from interference between the paths. In the single channel case, these results reduce to those of Eq. (15). A (R′ †tB(cid:17) B(cid:17)o , B − t† Br′ A(cid:16)t′ +tAr† ArA, TA = t† B B (19) An important observation can be made about the multi-channel case by considering that the scattering ma- trices preserve the channel-index, i. e. we essentially have M independent interferometers. In this case, the LG pa- rameter reads K = 1 m=1 K (m), where K (m) is the M PM 5 LG parameter for channel m. If we could tune by hand all the parameters of the scattering matrices, then the maximum violation of Kmax = 3 2 can be reached. How- ever, in an experiment, there will typically only be a few controllable parameters and this could make viola- tions hard to observe. Let us assume that we can adjust the parameters such that one of the K (m) is maximised, m = 1, say. Whether we see a violation or not very much depends on what happens with the parameters of the other channels. If these parameters are all roughly sim- ilar to those of channel 1, then violations should still be observed. Generically, however, this will not be the case, and the K (m) for the other channels will take unrelated values in the range from −3 to 3 2 . The negative values are particulary troublesome as they will tend to over- whelm any positive contribution to the violation from other channels. This lack of controllability means that multi-channel geometries are best avoided if violations of the LGI are sought. III. DOUBLE QUANTUM DOT INTERFEROMETER The above MZI scheme functions by having the de- tectors remove electrons from the interferometer arms. In this section we study a second MZI realisation which leaves the electrons within the system and the effects of measurement are only felt through dephasing. This sec- ond set-up is shown in Fig. 4. As in the foregoing, the basic structure is of two (single-channel) leads that are joined at two beamsplitters. Beamsplitters between non- edge-channel leads can be realised by tunnel junctions, as in the recent experiments by Yamamoto et al.57. In each arm of the interferometer there is a QD and alongside each QD is a QPC charge detector. When connected to a voltage supply, the current flowing through the QPCs serves as read-out of the occupation of their respective QDs. Note that although similar detectors we used in e.g. Refs. 69 and 70, the way in which they are used here is different. A. Model We first consider the system without detectors. Our MZI model is related to that of, e.g., Refs. 71 and 72, but with different leads. Far from the junctions, we describe the four leads as non-interacting Fermi reser- voirs with Hamiltonian Hres = P ωkαqc† kαqckαq with k the wavenumber of the electron, and where α = 1, 3 and q = ± specify the lead (we set = 1 and ignore spin). We assume that there is but a single orbital of relevance in each dot and that the DQD system is in the strong Coulomb blockade regime, such that it is restricted to 'empty', 0i; or with one excess elec- just three states: tron in either the upper or lower dots, +i = d† +0i and −i = d† −0i, respectively. Assuming the dot levels are D+ + - 1+ 1- s A s B 3+ 3- FIG. 4. (Color online) Sketch of a Mach-Zehnder interferom- eter with a quantum dot in each arm. The charge state of each QD can be monitored by the currents flowing through QPCs next to the dots. Here only the QPC monitoring the dot + is active, such that the correlation function to be measured is as Fig. 1b. detuned by an energy ǫ from one another, the dot Hamil- tonian reads HS = 1 qdq. In the following, we set ǫ = 0 for simplicity. 2 ǫPq qd† We assume that the effect of the beam-splitters is to modify the amplitudes with which the leads couple to the QDs. So, for example, an electron in lead 1+ tunnels into a superposition of upper and lower dot states, with the details of the superposition being determined by the scattering matrix sA. The tunnel Hamiltonian connect- ing lead and dots therefore reads A · d + C† k3 · sB · d + H.c.(cid:17) , (20) HT = Xk (cid:16)C† k1 · s† kα+, Tα−c† kα = (Tα+c† where sA,B are scattering matrices, assumed to be energy independent, d = (d+, d−) is a vector of dot operators, and C† kα−) are vectors of lead op- erators with tunnel matrix elements Tαs, also assumed to be energy-independent. The corresponding sequential tunnel rates are Γαq = 2πTαq2αq, where αq is the density-of-states of reservoir αq, also assumed constant. In the infinite-bias limit, the system can be described by a quantum master equation of Lindblad-form73 -- 75. Let us introduce the super-operator notations J [d]ρ = dρd† and A[d]ρ = − 1 2 (cid:8)d†d, ρ(cid:9)76 -- 78, and introduce the operators d3q = pΓ3qeq · sB · d, (21) d1q = pΓ1q eq · sA · d; with unit vectors e+ = (1, 0) and e− = (0, 1). With introduction of counting fields χαq to facilitate the cal- culation of current statistics (see e.g. Refs. 74, 79), the χ-resolved master equation for the DQD system reads ρ(χ) = −i [HS, ρ] +Xαq (cid:16)eiχαqJ [ dαq] − A[ dαq](cid:17) ρ. (22) 6 Ref. 59. In including the detectors in our theory, we fol- low Gurvitz58,61. When dot q is unoccupied, the Hamil- tonian for QPC Dq reads HDq = Xks ksqa† ωD ksqaksq + Ωq Xk a† kLqakRq + H.c, (23) where ωD ksq is the energy of an electron in state k on side s = L, R of the the QPC, and Ωq is the coupling amplitude between the two sides, (assumed energy inde- pendent). When dot q is occupied, we assume that this Hamiltonian is modified such that the coupling constants shift to different values, Ωq → Ω′ q. In the limit of large bias across the QPC, the detector at location q gives rise to an extra Liouvillian WDq(χDq) = eiχDqJ [ dDq] − A[ dDq], (24) q = 2πΩ′ which adds to the DQD Liouvillian. Here, dDq = qqihq + √γq (1 − qihq) with γq the rate of elec- pγ′ tron transfer through the QPC q when its dot is empty, and γ′ q the rate when the dot is occupied. The count- ing field χDq here allows us to calculate the statistics of the detector currents. Microscopically, the rates are q2ρLqρRqVDq, γq = 2πΩq2ρLqρRqVDq and γ′ with ρsq the density of states of the QPC reservoir sq and VDq the applied voltage. Detectors may be decou- pled or coupled from the MZI-QD system by adjusting the QPC voltages such that the differences between the amplitudes Ωq and Ω′ q is either zero (decoupled) or finite (coupled). Here, we only couple at most one detector to the system at a given time. Furthermore, we assume bal- anced detectors such that with the D+ detector coupled we have γ+ = γ, γ′ − = 0, and when − = γ′, the D− detector is coupled we have γ− = γ, γ′ and γ+ = γ′ + = γ′, and γ− = γ′ + = 0, B. Current, correlation functions and probabilities Our approach to measuring the LGI with this set-up is similar to that with the quantum Hall edge-channels with the main exception being how C32 is obtained. We inject electrons into the '+'-channel of lead 1 and close the '1−' channel: Γ1+ → ΓL and Γ1− → 0. For simplicity, we set the output rates equal: Γ3+ = Γ3− = ΓR. To obtain C31, we switch off the QPC detectors and measure the output currents at 3±. Arranging the el- ements of the density matrix into a vector in the basis (00, ++,−−, +−,−+), the stationary state of the DQD system reads 1 ρstat = 2Γ (cid:16) 2ΓR, ΓL(1 + cos θA), ΓL(1 − cos θA), −eiφA/2ΓL sin θA,−e−iφA/2ΓL sin θA(cid:17) ,(25) The QDs are monitored by QPCs in a set-up similar to the single dot in an interferometer in the experiment of with total width Γ = ΓL + ΓR. Here, we have assumed the same scattering matrices as in Eq. (7). The total current flowing is hIitot = hIi1+ = ΓLΓR/Γ, which is divided between the output ports as where q = −q and the total probability at output 3q′ is obtained from the currents 7 hIi3± = hIitot 2 (1 ± cos θA cos θB ∓ cos φ sin θA sin θB) . Constructing C31 as in Eq. (12) we obtain C31 = cos θA cos θB − cos φ sin θA sin θB, (26) which agrees with that of Eq. (9). Next we can obtain C21 by turning on the QPC de- tectors one at a time. As shown in Ref. 58, the mean current flowing through the QPC can be used to extract the mean current flowing through the corresponding dot. With a detector coupled to dot q, the current through the detector is hIDqi = hIitot 2ΓR (cid:20)γ(cid:18)1 + 2 ΓR ΓL(cid:19) + γ′ + q(γ′ − γ) cos θA(cid:21) . The current flowing through the QPC when the DQD is empty is hI 0 Dqi = γ, such that the difference is h∆IDqi = hIDqi − hI 0 Dqi = γ′ q − γq 2ΓR hIitot (1 + q cos θA) , (27) which is proportional to the probability that an electron takes the path 2q. Assuming balanced detectors, we ob- tain C21 = h∆ID+i − h∆ID−i h∆ID+i + h∆ID−i = cos θA, (28) as in Eq. (12). Whereas these two correlation functions can be deter- mined with just mean-current measurements, to deter- mine C32 we need to consider current cross-correlations. Let us first imagine that we can measure the current through dot q. Then, in the limit ΓL → 0, such that there is only ever at most one electron in the interferom- eter at a given time, the zero-frequency noise correlator S3q′2q ≡ 1 2 R dth{I3q′ , I2q}ic, (29) where h. . .ic denotes the cumulant average, is propor- tional to the joint probability, P3q′2q, that the electron travels through dot q and ends up at output 3q′. This result follows in the same way as in Ref. 15; the differ- ence here being that we correlate the position of a single electron in subsequent regions, as opposed to the corre- lation of two spatially-separate electrons. Measuring all four such correlators, we obtain the probabilities P3q′2q = . (30) S3q′2q Pr′r S3r′2r From these directly-obtained probabilities, we construct the ideal-negative-measurement ones as P INM 3q′2q = P3q′ − P3q′2q, (31) P3q = hI3qi PrhI3ri . (32) These relations follow from charge conservation and the unidirectional nature of the transport. The QPC detectors couple not the current flowing though the dot, but rather to their occupations. In terms of the zero-frequency correlation function between cur- rent fluctuations in the QPC and those in one of the 3± ports, S3q′Dq ≡ 1 2 Z dth{I3q′ , IDq}ic, (33) the required probabilities read P3q′2q = h∆IDqiS3q′Dq Pr′rh∆IDriS3r′Dr . (34) This can be understood as follows. Whereas S3q′2q corre- lates two delta-function peaks corresponding to the pas- sage of the electron through the regions 2 and 3, S3q′Dq correlates a delta-function in region 3 with a signal of fi- nite duration in region 2, which corresponds to the finite time for which the dot is occupied. This mean occupa- tion time is proportional to the inverse of mean current through the dot, which can be obtained (up to a pro- portionality constant) from the mean detector current h∆IDqi. limit ΓL → 0, the third correlation function reads Calculating these probabilities, we find that in the C32 = cos θB, (35) in accordance with Eq. (10). Since, in the ΓL → 0 limit, all three correlation function are identical with their ideal counterparts, the LGI for this set-up is identical to that of Eq. (11). In the way that we have described the QPC de- tectors here, it does not make any difference whether we calculate K using the ideal negative measurement prob- abilities or the direct ones since, in our theoretical de- scription, the QPC detectors act as ideal detectors and only influence the system through their dephasing effect. Experimentally, the ideal negative measurement protocol should be used, and actually, the comparison between the case with ideal negative measurement and that without would give an interesting method for studying to what extent the QPC measurements are non-invasive. Let us just add that, whilst the above results were derived in the symmetric case with ΓR+/ΓR− = 1 and with balanced detector rates, if these ratios are unequal but known, then the difference can be accounted for by weighting the terms in the correlation functions accordingly. C. Dephasing A simple way to include the effects of dephasing in this model is to "leave the detectors switched on" when calculating C31. With empty and occupied rates γdephase and γ′ dephase, the measured K function has the form of Eq. (16), with the function f (∆) replaced by f =  1 + (cid:16)(γ′  dephase)1/2 − (γdephase))1/2(cid:17)2 2ΓR −1   . (36) 8 relies on the weak-tunnel limit17 -- 20, so it is here that our measurements are only isomorphic with those required by the LGI in the ΓL → 0 limit. Away from this limit, there exists the possibility that our measurements mis- takenly correlate subsequent electrons, rather than the same electron with itself. f = 1 is the ideal no-dephasing case, and f = 0 gives the classical limit. To obtain strong violations of the LGI, therefore requires that the difference in rates γ′ dephase and γdephase is small compared with the tunnel rate ΓR. D. Detection errors Just as the direct relation between the Bell inequal- ities and noise measurements of, e.g., Refs. 15 and 16 The LGI quantity can be calculated using the currents and zero-frequency noise, as described above, away from the ΓL → 0 limit to assess the error. Assuming for sim- plicity that the detector is faster than the system dynam- ics γ′ − γ ≫ ΓL/R (although the general case can easily be investigated too), we obtain for the LG correlator K ′ = 1 (ΓL − ΓR)ΓR (cid:8)(cid:2)−(ΓL + ΓR)2 + ΓL(ΓL + 3ΓR) cos2 θA(cid:3) cos θB +(ΓL − ΓR)ΓR(cid:2)2 cos θA sin2(θB/2) + cos φ sin θA sin θB(cid:3)(cid:9) . (37) This expression is again maximised with cos φ = 1, but, unlike the ΓL → 0 case, the maximizing angles θA and θB are not equal. If we assume that ΓL/ΓR ≪ 1, we can expand to leading order (γ′ − γ → ∞) to obtain ΓR(cid:17)2(cid:19) , cos θB sin2 θA + O(cid:18)(cid:16) ΓL K ′ = K + 3 ΓL ΓR where K is the ΓL → 0 value. We can also calculate the corresponding quantity in the classical limit (this we do by calculating C31 in limit (γ′ dephase − γdephase) → ∞). In this case, we obtain Ccl 31 = cos θA cos θB, and the expansion of B′ = C21 + C32 − Ccl 31 for small ΓL gives B′ = B + 3 ΓL ΓR cos θB sin2 θA + O(cid:18)(cid:16) ΓL ΓR(cid:17)2(cid:19) , (38) where B = cos θA + cos θB − cos θA cos θB is the classi- cal value in the ideal case which, when maximised gives Bmax = 1, the bound of Eq. (1). Maximising B′ over the angles, we obtain a value bigger that unity. To low- est order then, classical and quantum LG correlators are affected in the same way. Fig. 5 shows the maximum values of both quantum and classical correlators. Thus, assuming that we know the ratio of ΓL/ΓR from current and noise measurements, the effects of a finite tunneling rate ΓL can be included in assessment of whether LGI is violated or not. The conservative ap- proach is say that the quantity (cid:0)K ′ systematic error in the measurement, and assuming that this error works against us, we can only conclude that we violate the LGI when the measured value of K exceeds 2(cid:1) represents a max − 3 unity by an amount equal to this error. Alternatively, one can say that since one knows how the classical bound be- haves at finite Γ/ΓR, we can simply use B′ max of Eq. (38) as a bound. However, providing that we are in the cor- rect operating limit of ΓL/ΓR ≪ 1, these modifications will be very small, such that whether they are taken into account or not will only effect the question of violation in marginal cases. IV. CONCLUSIONS We have considered the violation of the LGI in MZ inteferometer geometries. The key to the violation is a combination of the interference at the second beamsplit- ter and the inhibition of this interference by the measure- ment process. In the two proposals we have considered this inhibition occurs in two different ways. In the first realisation, we physically interrupt transmission through one of the arms of the MZI, obviously preventing inter- ference. On the other hand, in the DQD proposal, the detectors act in a more traditional way and introduce dephasing between the paths. In this MZ geometry both the state of the electron and measurement time are mapped onto real-space co- ordinates -- the qubit states Q = ± are physically sep- arate paths, and the regions within the interferometer correspond to different time instances. This mapping has several advantages for seeking a violation of the LGI in transport. The mapping of the time-coordinate means that we do not need to make time-resolved correlation 2 x a m K ' 1.5 quantum, K'max classical bound, B'max 1 0 0.05 Γ 0.1 L/Γ R 0.15 0.2 FIG. 5. (Color online) The maximum value K ′ max (blue solid line) and the corresponding classical value B ′ max (black dashed line) away from the ΓL → 0 limit. Both are higher than their ideal ΓL → 0 values. A symmetric system was assumed and the fast detector limit (γ ′ − γ)/ΓR → ∞ taken. measurements. All the measurements required here are either mean stationary currents or zero-frequency noise correlators. Furthermore, the spatial separation of the qubit degrees of freedom facilitates the realisation of ideal negative measurement, since it is relatively easy to cou- ple to just one of the qubit states when they are spatially distinct. In this respect, increasing the separation of the detector arms should decrease the plausibility of claims that detection in one arm is, from a macro-realist point- of-view, influencing the other. The general principles described here can easily be ex- tended to further systems. Within transport, for exam- ple, our second scheme could also be realised with an 9 edge-channel MZI plus QPC detector channel without the quantum dots47 -- 49,51. An alternative setting for the realisation of our first scheme might be the flying qubit experiment of Ref. 57, which is essentially a MZI away from the quantum Hall regime. Two challenges are ob- vious with this realisation. Firstly, the leads reported in the experiment have multiple channels, which potentially gives rise to the problems discussed in section II C. The second problem is that of backscattering at the beam- splitters and detectors, which has (justifiably) been ne- glected here but probably can not be eliminated in set- ups such as that of Ref. 57. Applications away from electronic transport are also possible. The application of the first scheme in optics is obvious but the notion of the qubit state is predicated on the source being a single-photon source. So whilst a classical wave might also exceed the right-hand-side of Eq. (1), this would not constitute a violation of the LGI, as it represents an application of the concepts outside their proper realm of definition (i.e., non-dichotomic ob- servables). Going further, the same principles could be used to test the LGI with electrons in free space, neu- trons, atoms and molecules, all of which have had in- terference experiments in the MZI geometry conducted on them80,81. Of these, molecules offer the most excit- ing prospect, as there the nature of the coherence being tested could potentially be macroscopic, in line with the original goals of Ref. 3. ACKNOWLEDGMENTS We are grateful to S. Huelga, Y. Ota, P. Roche, P. Samuelsson, and M. Yamamoto for useful discussions. This work was supported by the DFG through SFB 910. FN acknowledges partial support from the ARO, JSPS- RFBR contract No. 09-02-92114, MEXT Kakenhi on Quantum Cybernetics, and the JSPS-FIRST Program. 1 J. S. Bell, Physics 1, 195 (1964). 2 J. S. Bell, Speakable and Unspeakable in Quantum Mechan- ics, 2nd ed. (Cambridge University Press, 2004). 3 A. J. Leggett and A. Garg, Phys. Rev. Lett. 54, 857 (1985). 4 A. J. Leggett, Journal of Physics: Condensed Matter 14, R415 (2002). 5 From Ref. 3, these assumptions are as follows. Macroscopic realism: A macroscopic system with two or more macro- scopically distinct states available to it will, at all times, be in one or the other of these states. Noninvasive measur- ability: It is possible, in principle, to determine the state of the system with arbitrary small perturbation on its subse- quent dynamics. More information can be found in Refs. 3 and 4. 6 A. Palacios-Laloy, F. Mallet, F. Nguyen, P. Bertet, D. Vion, D. Esteve, and A. N. Korotkov, Nat Phys 6, 442 (2010). 7 M. E. Goggin, M. P. Almeida, M. Barbieri, B. P. Lanyon, J. L. O'Brien, A. G. White, and G. J. Pryde, Proc. Natl. Acad. Sci. USA 108, 1256 (2011). 8 J.-S. Xu, C.-F. Li, X.-B. Zou, and G.-C. Guo, Sci. Rep. 1, 101 (2011). 9 J. Dressel, C. J. Broadbent, J. C. Howell, and A. N. Jor- dan, Phys. Rev. Lett. 106, 040402 (2011). 10 G. Waldherr, P. Neumann, S. F. Huelga, F. Jelezko, and J. Wrachtrup, Phys. Rev. Lett. 107, 090401 (2011). 11 V. Athalye, S. S. Roy, and T. S. Mahesh, Phys. Rev. Lett. 107, 130402 (2011). 12 G. C. Knee, S. Simmons, E. M. Gauger, J. J. Morton, H. Riemann, N. V. Abrosimov, P. Becker, H.-J. Pohl, K. M. Itoh, M. L. Thewalt, G. A. D. Briggs, and S. C. Benjamin, Nat Commun 3, 606 (2012). 13 A. Miranowicz, M. Bartkowiak, X. Wang, Y.-x. Liu, and F. Nori, Phys. Rev. A 82, 013824 (2010). 14 N. M. Chtchelkatchev, G. Blatter, G. B. Lesovik, and T. Martin, Phys. Rev. B 66, 161320 (2002). 10 15 P. Samuelsson, E. V. Sukhorukov, and M. Buttiker, Phys. 42 H. Forster, S. Pilgram, and M. Buttiker, Phys. Rev. B 72, Rev. Lett. 91, 157002 (2003). 075301 (2005). 16 C. W. J. Beenakker, C. Emary, M. Kindermann, and J. L. 43 V. S.-W. Chung, P. Samuelsson, and M. Buttiker, Phys. van Velsen, Phys. Rev. Lett. 91, 147901 (2003). Rev. B 72, 125320 (2005). 17 P. Samuelsson, E. V. Sukhorukov, and M. Buttiker, Phys. 44 E. V. Sukhorukov and V. V. Cheianov, Phys. Rev. Lett. Rev. Lett. 92, 026805 (2004). 99, 156801 (2007). 18 A. V. Lebedev, G. B. Lesovik, and G. Blatter, Phys. Rev. 45 H. Forster, P. Samuelsson, and M. Buttiker, New Journal B 71, 045306 (2005). 19 C. Beenakker, in Quantum Computers, Algorithms and Chaos, Proc. Int. School Phys. E. Fermi, Vol. 162, edited by G. Casati (IOS Press, Amsterdam, 2006). 20 P. Samuelsson, I. Neder, and M. Buttiker, Phys. Rev. Lett. 102, 106804 (2009). 21 C. Emary, Phys. Rev. B 80, 161309 (2009). 22 A. Bednorz and W. Belzig, Phys. Rev. B 83, 125304 (2011). 23 N. Lambert, C. Emary, Y.-N. Chen, and F. Nori, Phys. Rev. Lett. 105, 176801 (2010). 24 C. Emary, Phys. Rev. B 86, 085418 (2012). 25 Y.-N. Sun, Y. Zou, R.-C. Ge, J.-S. Tang, C.-F. Li, and G.-C. Guo, arXiv:1110.5537v2 [quant-ph] (2011). 26 M. M. Wilde, J. M. McCracken, and A. Mizel, Proc. R. Soc. A 466, 1347 (2010). 27 T. Hayashi, T. Fujisawa, H. D. Cheong, Y. H. Jeong, and Y. Hirayama, Phys. Rev. Lett. 91, 226804 (2003). 28 K. D. Petersson, J. R. Petta, H. Lu, and A. C. Gossard, Phys. Rev. Lett. 105, 246804 (2010). 29 M. Wilde and A. Mizel, Foundations of Physics 42, 256 (2012). 30 L. E. Ballentine, Phys. Rev. Lett. 59, 1493 (1987); A. J. Leggett and A. Garg, Phys. Rev. Lett. 59, 1621 (1987); A. Peres, Phys. Rev. Lett. 61, 2019 (1988); A. J. Leggett and A. Garg, Phys. Rev. Lett. 63, 2159 (1989); C. D. Tesche, Phys. Rev. Lett. 64, 2358 (1990); A. Elby and S. Foster, Physics Letters A 166, 17 (1992); J. P. Paz and G. Mahler, Phys. Rev. Lett. 71, 3235 (1993); F. Be- natti, G. Ghirardi, and R. Geassi, Il Nuovo Cimento B (1971-1996) 110, 593 (1995); T. Calarco and R. Onofrio, Physics Letters A 198, 279 (1995); R. Onofrio and T. Calarco, Physics Letters A 208, 40 (1995). 31 Y. Ji, Y. Chung, D. Sprinzak, M. Heiblum, D. Mahalu, of Physics 9, 117 (2007). 46 H. Forster, P. Samuelsson, S. Pilgram, and M. Buttiker, Phys. Rev. B 75, 035340 (2007). 47 I. Neder, M. Heiblum, D. Mahalu, and V. Umansky, Phys. Rev. Lett. 98, 036803 (2007). 48 I. Neder, F. Marquardt, M. Heiblum, D. Mahalu, and V. Umansky, Nat Phys 3, 534 (2007). 49 I. Neder and F. Marquardt, New Journal of Physics 9, 112 (2007). 50 S.-C. Youn, H.-W. Lee, and H.-S. Sim, Phys. Rev. Lett. 100, 196807 (2008). 51 D.-I. Chang, G. L. Khym, K. Kang, Y. Chung, H.-J. Lee, M. Seo, M. Heiblum, D. Mahalu, and V. Umansky, Nat Phys 4, 205 (2008). 52 A. W. Holleitner, C. R. Decker, H. Qin, K. Eberl, and R. H. Blick, Phys. Rev. Lett. 87, 256802 (2001). 53 M. Sigrist, A. Fuhrer, T. Ihn, K. Ensslin, S. E. Ulloa, W. Wegscheider, and M. Bichler, Phys. Rev. Lett. 93, 066802 (2004). 54 T. Ihn, M. Sigrist, K. Ensslin, W. Wegscheider, M. Reinwald, New Journal of Physics 9, 111 (2007). and 55 A. Muhle, W. Wegscheider, and R. J. Haug, Applied Physics Letters 92, 013126 (2008). 56 T. Hatano, T. Kubo, Y. Tokura, S. Amaha, S. Teraoka, and S. Tarucha, Phys. Rev. Lett. 106, 076801 (2011). 57 M. Yamamoto, S. Takada, C. Bauerle, K. Watanabe, A. D. Wieck, and S. Tarucha, Nat Nano 7, 247 (2012). 58 S. A. Gurvitz, Phys. Rev. B 56, 15215 (1997). 59 E. Buks, R. Schuster, M. Heiblum, D. Mahalu, and V. Umansky, Nature 391, 871 (1998). 60 A. N. Korotkov, Phys. Rev. B 60, 5737 (1999). 61 S. Gurvitz, Physics Letters A 311, 292 (2003). 62 D. V. Averin and E. V. Sukhorukov, Phys. Rev. Lett. 95, and H. Shtrikman, Nature 422, 415 (2003). 126803 (2005). 32 I. Neder, M. Heiblum, Y. Levinson, D. Mahalu, and 63 S. Ashhab, J. Q. You, and F. Nori, New Journal of Physics V. Umansky, Phys. Rev. Lett. 96, 016804 (2006). 33 L. V. Litvin, H.-P. Tranitz, W. Wegscheider, 11, 083017 (2009). and 64 It is known that the initial state of the system does not C. Strunk, Phys. Rev. B 75, 033315 (2007). 34 P. Roulleau, F. Portier, D. C. Glattli, P. Roche, A. Ca- vanna, G. Faini, U. Gennser, and D. Mailly, Phys. Rev. B 76, 161309 (2007). 35 I. Neder and E. Ginossar, Phys. Rev. Lett. 100, 196806 (2008). 36 P. Roulleau, F. Portier, P. Roche, A. Cavanna, G. Faini, U. Gennser, and D. Mailly, Phys. Rev. Lett. 100, 126802 (2008). 37 P. Roulleau, F. Portier, P. Roche, A. Cavanna, G. Faini, U. Gennser, and D. Mailly, Phys. Rev. Lett. 101, 186803 (2008). 38 P. Roulleau, F. Portier, P. Roche, A. Cavanna, G. Faini, U. Gennser, and D. Mailly, Phys. Rev. Lett. 102, 236802 (2009). 39 G. Seelig and M. Buttiker, Phys. Rev. B 64, 245313 (2001). 40 F. Marquardt and C. Bruder, Phys. Rev. Lett. 92, 056805 (2004). affect the degree of violation of the LGI82. 65 We do not actually need it to be a detector at 2±, anything that removes the electron from the system would suffice. However, seeing as we need detectors for measuring C21, it seems sensible to use these. 66 Y. M. Blanter and M. Buttiker, Physics Reports 336, 1 (2000). 67 J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, Phys. Rev. Lett. 23, 880 (1969). 68 This function f (∆) is the amplitude of AB oscillations (normalised to the maximum current possible) that would be observed by sweeping the magnetic flux through the device with θA = θB = π/2. 69 R. Ruskov, A. N. Korotkov, and A. Mizel, Phys. Rev. Lett. 96, 200404 (2006). 70 A. N. Jordan, A. N. Korotkov, and M. Buttiker, Phys. Rev. Lett. 97, 026805 (2006). 71 F. Marquardt and C. Bruder, Phys. Rev. B 68, 195305 41 F. Marquardt and C. Bruder, Phys. Rev. B 70, 125305 (2003). (2004). 72 D. Urban and J. Konig, Phys. Rev. B 79, 165319 (2009). 73 S. A. Gurvitz, Phys. Rev. B 57, 6602 (1998). 74 D. A. Bagrets and Y. V. Nazarov, Phys. Rev. B 67, 085316 (2003). 75 T. Brandes, Physics Reports 408, 315 (2005). 76 C. W. Gardiner, Phys. Rev. Lett. 70, 2269 (1993). 77 H. J. Carmichael, Phys. Rev. Lett. 70, 2273 (1993). 78 H. M. Wiseman, Phys. Rev. A 49, 2133 (1994). 79 L. S. Levitov, H. Lee, and G. B. Lesovik, Journal of Mathe- matical Physics 37, 4845 (1996); L. S. Levitov, in Quantum Noise in Mesoscopic Physics, Nato Science series, Vol. 97, edited by Y. V. Nazarov (Kluwer, 2002) p. 373. 80 J. Schmiedmayer, M. S. Chapman, C. R. Ekstrom, T. D. Hammond, D. A. Kokorowski, A. Lenef, R. A. Rubenstein, E. T. Smith, and D. E. Pritchard, in Atom Interferometry, edited by P. R. Berman (Academic Press, San Diego, 1997) pp. 1 -- 83. 81 A. D. Cronin, J. Schmiedmayer, and D. E. Pritchard, Rev. Mod. Phys. 81, 1051 (2009). 82 G. C. Knee, E. M. Gauger, G. A. D. Briggs, and S. C. Benjamin, New Journal of Physics 14, 058001 (2012). 11
1604.08601	3	1604	2017-05-09T18:15:30	The chiral anomaly factory: Creating Weyl fermions with a magnetic field	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci", "cond-mat.str-el" ]	Weyl fermions can be created in materials with both time reversal and inversion symmetry by applying a magnetic field, as evidenced by recent measurements of anomalous negative magnetoresistance. Here, we do a thorough analysis of the Weyl points in these materials: by enforcing crystal symmetries, we classify the location and monopole charges of Weyl points created by fields aligned with high-symmetry axes. The analysis applies generally to materials with band inversion in the $T_d$, $D_{4h}$ and $D_{6h}$ point groups. For the $T_d$ point group, we find that Weyl nodes persist for all directions of the magnetic field. Further, we compute the anomalous magnetoresistance of field-created Weyl fermions in the semiclassical regime. We find that the magnetoresistance can scale non-quadratically with magnetic field, in contrast to materials with intrinsic Weyl nodes. Our results are relevant to future experiments in the semi-classical regime.	cond-mat.mes-hall	cond-mat	The chiral anomaly factory: Creating Weyl fermions with a magnetic field Jennifer Cano,1, ∗ Barry Bradlyn,1, ∗ Zhijun Wang,2 Max Hirschberger,2 N. P. Ong,2 and B. A. Bernevig2 1Princeton Center for Theoretical Science, Princeton University, Princeton, NJ 08544 2Department of Physics, Princeton University, Princeton, NJ 08544 (Dated: October 15, 2018) Weyl fermions can be created in materials with both time reversal and inversion symmetry by applying a magnetic field, as evidenced by recent measurements of anomalous negative magnetore- sistance. Here, we do a thorough analysis of the Weyl points in these materials: by enforcing crystal symmetries, we classify the location and monopole charges of Weyl points created by fields aligned with high-symmetry axes. The analysis applies generally to materials with band inversion in the Td, D4h and D6h point groups. For the Td point group, we find that Weyl nodes persist for all directions of the magnetic field. Further, we compute the anomalous magnetoresistance of field-created Weyl fermions in the semiclassical regime. We find that the magnetoresistance can scale non-quadratically with magnetic field, in contrast to materials with intrinsic Weyl nodes. Our results are relevant to future experiments in the semi-classical regime. INTRODUCTION Following their insulating counterparts[1], topological semi-metals have attracted much recent theoretical and experimental interest. Weyl and Dirac semimetals have recently been theoretically predicted[2 -- 6] and experimen- tally observed[7 -- 9]. Both display topologically protected Fermi-arc surface states, as well as large negative magne- toresistance due to the "chiral anomaly"[10 -- 15]. While Weyl fermions in these semimetals are robust to small perturbations due to their topological character, Dirac points require a combination of crystal symmetry, time reversal, and inversion symmetry[16]. This suggests that Weyl fermions can be engineered by breaking inversion or time reversal symmetry in materials with four-band crossings. While breaking inversion symmetry can be accomplished by adding strain[17], it is more straight- forward to break time reversal symmetry by turning on a magnetic field. This route to creating Weyl fermions has already been carried out in GdPtBi[18, 19], NdPtBi[19], Na3Bi[13] and Cd3As2[20 -- 22]. We predict that the same route can be used to observe Weyl fermions in the exper- imentally relevant materials HgTe[23, 24] and InSb[25]. Here, we consider turning on a magnetic field in ma- terials with four-band crossings. We consider two types of four-band crossings: symmetry-enforced band cross- ings at the Γ point and Dirac points near the Γ point on a high-symmetry line. In both cases, the magnetic field breaks the four-band crossing into an even number of Weyl nodes. We demonstrate the emergence of Weyl nodes explicitly in GdPtBi, HgTe and InSb, which host a symmetry-enforced four-band crossing near the Fermi level, and Cd3As2 and Na3Bi, which host Dirac points near the Fermi level, using k·p Hamiltonians generated by ab initio calculations. We then show that this is a general result when the magnetic field is along a high-symmetry line. However, the emergent Weyls do not always reside on the axis parallel to the magnetic field: instead, a com- plex map of Weyl points (and nodal lines[26 -- 29]) emerges for different directions of the magnetic field. Since Weyl points do not require symmetry protection, they persist when the magnetic field is moved away from these axes. Surprisingly, we find that in GdPtBi and HgTe, Weyl points exist for all directions of the magnetic field. In the Supplement, we give numerical evidence for this state- ment and then prove it explicitly for all materials with Td symmetry and j = 3/2 orbitals at the Fermi level. We emphasize that while we use the language of apply- ing an external magnetic field, our results apply equally well to magnetically ordered materials that host a four- band crossing above the N´eel temperature and can thus be tuned to display Weyl points below this temperature. This is relevant for antiferromagnetic Heuslers[19]. Finally we consider the semiclassical negative mag- netoconductance that is a consequence of the chiral anomaly. This has previously been considered for Weyl points that exist independent of the magnetic field[30, 31]. Here, we show that when the Weyl points are cre- ated by a magnetic field, the magnetoconductance takes a different scaling form. In particular, it can scale as high as B5/2, where the exact scaling depends on the lin- earized Hamiltonian near the Weyl point. We apply this model to GdPtBi, in which recent experiments[18] have observed the chiral anomaly. EMERGENT WEYL NODES FROM A SYMMETRY-ENFORCED FOUR-BAND CROSSING Here we focus on GdPtBi, which is in the Td point group. However, because our analysis depends only on symmetry and band inversion, it applies to all materials with the same symmetry and relevant orbitals near the Γ point, e.g., HgTe and InSb. In GdPtBi, the low-energy spectrum near the Γ point is described by the four p-orbitals with jz = ±3/2,±1/2. Thus, the symmetry operators comprise a four-dimensional representation of the Td symmetry group. The k · p Hamiltonian takes the form[32], (cid:16) )k2(cid:17) HΓ(kx, ky, kz) = + C(k2 x − k2 y)Γ1 + C√ A0 + (A + C√ z − k2 (2k2 3 3 I4×4 + x − k2 y)Γ2 + + E(kxkyΓ3 + kxkzΓ4 + kykzΓ5) + + D(kxU1 + kyU2 + 2kzU3), (1) √ 3Γ15 − Γ25, U2 = −√ where the Γ1,...,5 are Clifford algebra matrices, Γij = 3Γ14 − [Γi, Γj]/(2i) and U1 = Γ24, U3 = Γ23. The parameters A0, A, C, D, E are ob- tained by an ab initio fit and given in the Supplementary Material. We note here only that A0 ≈ 0, meaning that at the Fermi level the spectrum is four-fold degenerate at the Γ point[18]; two bands disperse upwards and two dis- perse downwards. The D (cid:54)= 0 parameter breaks inversion symmetry. 2 (cid:126)B Emergent Weyl points [001], [010], [100] 6 + 8k1 Weyl points [110], [1¯10], [101], 0, 1, or 2 line nodes, [10¯1], [011], [01¯1] 4 + 8k2 Weyl points [111], [11¯1], [1¯11], [¯111], [¯1¯11], [¯1¯11], [1¯1¯1] 4 + 6k3 Weyl points TABLE I. Weyl points in GdPtBi when a magnetic field is ap- plied along one of the high-symmetry axes in the first column. The integers ki indicate Weyl points that appear at generic points Brillouin zone; for GdPtBi, a numerical analysis of our k · p model yields k1 = k2 = 0, k3 = 1, but these numbers are material-dependent. We now consider what happens in the presence of a magnetic field by adding an effective Zeeman coupling to Eq (1), HZ = (cid:126)B · (cid:126)J, (2) where (cid:126)J is a vector of the spin-3/2 matrices. Using this model, oppositely-dispersing bands have either a pro- tected or avoided crossing, regardless of their g-factor. Notice that band inversion is crucial here: if all bands dispersed in the same direction, then the presence or ab- sence of a crossing would depend crucially on the precise values g-factors of the bands. However, as long as the system exhibits band inversion, and all bands have the same sign of the g-factor, then the physics described be- low is universal. In the following analysis, we will ignore the effects of cyclotron motion to lowest order, focusing purely on the Zeeman splitting. This is a good approxi- mation for large g−factor materials such as GdPtBi. When the magnetic field is along an axis of rotation, band crossings along this axis between bands with dif- ferent eigenvalues under the rotation are protected; thus, Weyl points are guaranteed to exist. Because the original four-band quadratic crossing was at the Fermi energy at zero field, these Weyl points will also lie near the Fermi level. Additionally, since the nontrivial Chern number of a Weyl point cannot disappear under small deforma- tions of the Hamiltonian, these Weyls will continue to exist even as the magnetic field is moved away from the rotation axis. This protection is crucial to experimental observation. However, as the magnetic field is moved far away from a high-symmetry axis, it is possible for two Weyl points of opposite chirality to meet and annihilate. Surprisingly, we have verified numerically, using the k · p Hamiltonian (1), that this is not the case for every pair of nodes: Weyl points exist in GdPtBi for all directions of the magnetic field. Weyl points can also be protected between high- symmetry planes with different Chern numbers; we show[32]that in GdPtBi, when the magnetic field is ap- plied along the z direction, two such Weyl points exist in the kx = 0 plane and another pair in the ky = 0 plane. These points persist -- albeit moving in momentum-space -- when the magnetic field is moved off this axis until they reach each other and annihilate. The same analysis ap- plies to a magnetic field in the x and y directions. When the magnetic field is in the x + y direction, we find four Weyl nodes which, for D → 0, are confined to the (k, k, kz) plane; these nodes persist as D is con- tinuously varied from zero to its experimentally relevant value in either GdPtBi or HgTe. Additionally, at least one line node appears in the (k,−k, kz) plane.[32]For the ab initio parameters, two line nodes appear; however, for other choices of parameters, one of these lines moves towards the edge of the Brillouin zone and disappears. Last, when the magnetic field is in the x + y + z direc- tion, there are four Weyl points along the kx = ky = kz axis. Additional Weyl points at generic points in the Bril- louin zone must occur in multiples of six or eight, de- pending on the symmetry that remains when a particu- lar magnetic field is present. A summary of Weyl points that emerge in GdPtBi upon applying a magnetic field along a high-symmetry axis are shown in Table I. The symmetry-protected Weyl points we describe in this section also persist for arbitrary magnetic fields. In particular, at the special point E = 2C, D = 0 in Eq (1), the Hamiltonian HΓ + HZ is exactly solvable and has Weyl points along the momentum axis parallel to the magnetic field. We show in the Supplement that as E is moved away from this fine-tuned value, the Weyl points move in space, but do not annihilate; we confirm this claim with a numerical analysis. Because the Weyl points are topologically protected, they will also persist for small values of D. EMERGENT WEYL NODES FROM DIRAC SEMICLASSICAL MAGNETOTRANSPORT POINTS NEAR THE Γ POINT 3 In Cd3As2 and Na3Bi, a different, but similar, scenario develops, where again band inversion plays a crucial role. There are two pairs of relevant orbitals at the Γ point near the Fermi level: the s-orbitals with jz = ±1/2 and the p-orbitals with jz = ±3/2; each pair transforms as a two-dimensional representation of Dnh, where n = 4 in Cd3As2 and n = 6 in Na3Bi. The two representa- tions have different energies at the Γ point, but, because the bands are inverted, they can cross elsewhere in mo- mentum space. In the materials of interest, the crossings occurs at the Fermi level and are protected by Cnz sym- metry. Furthermore, the crossings occur close enough to the Γ point to be described by the effective k · p model, (cid:16) (cid:17) (cid:17) C0 + C1k2 z + C2k2(cid:107) M0 + M1k2 HDirac(kx, ky, kz) = (cid:17) τz ⊗(cid:16)(cid:16) on the kz axis at kz = ±(cid:112)−M0/M1 (M0 and M1 have y. Eq (3) describes two Dirac points σz + A(σxkx + σyky) where k2(cid:107) = k2 x + k2 I4×4 + (3) z + M2k2(cid:107) , opposite signs in the ab initio fit; the fitting parameters for the quadratic terms and the symmetry-allowed third order terms are included in the Supplementary Mate- rial.) Because the four relevant bands come in pairs with distinct angular momentum character, we allow for two different g-factors gs and gp in the Zeeman coupling[32]. In the presence of a magnetic field, each Dirac point can split into up to four Weyl points. As in the previous section, putative crossings will exist regardless of the g- factors of the different orbitals because the bands dis- perse in opposite directions. Furthermore, Weyl points that emerge when the magnetic field is along a partic- ular high-symmetry axis must persist when the field is slightly off-axis because they can only annihilate in pairs of opposite chirality. We now summarize our results[32]. When the mag- netic field is along the z direction all band crossings are protected by Cnz symmetry: depending on the value of the magnetic field, this implies between four and eight Weyl points. Band crossings between the jz = ±3/2 and ∓1/2 bands are double Weyl points -- these are not robust to small changes in the magnetic field; instead, they can split into two single Weyl points. Line nodes, protected by M001, can also emerge in the kz = 0 plane for large enough magnetic field. When the magnetic field is along the x direction, C2x symmetry can protect between two and four total Weyl points and M100 symmetry protects line nodes in the kx = 0 plane. The same is true for the other symmetry- related directions. The presence of Weyl points near the Fermi surface in a material - whether intrinsic or created by an external field - leads to an experimentally measurable negative magnetoresistance.[13 -- 15, 18] The origin of this effect is due to the non-trivial Berry curvature surrounding each Weyl node, and is a manifestation of the so-called "chi- ral anomaly." Previous theoretical analysis of this effect has been carried out for materials with intrinsic, field independent Weyl nodes, in both the semiclassical and ultra-quantum (only the lowest Landau level occupied) limits.[12, 30, 31, 33] For these intrinsic Weyls, chiral ki- netic theory implies that, in the semiclassical limit and at low temperature, there is an anomalous positive mag- netoconductance of the form (cid:88) i σµν a = τ v3 i 8π2µ2 i BµBν, where τ is the inter-nodal scattering rate, vi is the (ge- ometric) mean velocity of node i, and µi is the chem- ical potential measured from node i. We wish to gen- eralize this result to the case where the Weyl nodes are created by an external magnetic field. Relegating the details of the derivation to the Supplementary Material, we find that the magnetoconductance acquires additional field dependence due to the field dependence of the Weyl velocities, which enters both explicitly and through the now-strongly-field-dependent scattering time. We find for low fields (cid:88) σµν a = τ i(B) det Ai 8π2µ2 i i BµBν(1 + O(B)). (4) where Ai, along with the vector ui, which, in the materi- als we consider, enters the O(B) corrections, parameter- ize the linearized two-band Hamiltonian near the Weyl point at ki:[34] H = ui j(kj − ki j)I + (kj − ki j)Ai jkσk (5) τ i(B) is the rate for scattering out of node i. For short- range impurities we find τ i(B) = τ0 2π2 det Ai µ2 i (1 + O(B)) (6) We now make three observations. First, because det Ai depends on magnetic field, we expect that the magnetoconductance for field-created Weyls scales dif- ferently than that for intrinsic Weyl semimetals; in par- ticular, we provide a simple model in the Supplemen- tary Material where detAi ∼ B1/2. Second, because det Ai is not a rotationally invariant function of B, we expect that the magnetoconductance will also fail to be rotationally invariant. Finally, as B increases, we expect the O(B) corrections to Eq. (4), given explicitly in the Supplementary material, to become significant. This can cause the directionality of the magnetoconductance to acquire additional field-dependence. Similarly, thermoelectric transport in Weyl materials is also influenced by the chiral anomaly. In particular, the thermoelectric conductivity αµν, relating the current response to a temperature gradient, is experimentally rel- evant. Using Onsager reciprocity[35, 36], we can compute this by looking instead at the energy current response to an electric field. Using the same semiclassical treatement as above and neglecting interactions we recover the Mott formula αµν = π2T 2 3 ∂σµν ∂µ , (7) valid for both the anomalous and non-anomalous parts of the thermoelectric conductivity. In particular, we ex- pect that the magnetic field dependence of the anomalous thermoelectric conductivity should simply follow that of the ordinary conductivity. Differences between these two effects serves to measure the significance of electron- electron interactions, which explicitly modify the heat- current[37]. Lastly, we remark on the effects of higher-order Weyl crossings on magnetotransport. In particular, we focus on double-Weyl points, since -- as mentioned above -- these are present in Na3Bi and Cd3As2. Using the fact that the Berry curvature transforms as a tensor under reparametrizations of the Brillouin zone, we can easily repeat the semiclassical analysis above for double (or even n-fold) Weyl points. We find that the forms of all transport coefficients remain the same. The only change is that the response coefficients are proportional to the square of the Chern number (i.e. 4 in the case of a dou- ble Weyl), and that the form of the density of states changes. In particular, the density of states for a double Weyl point is linear in the chemical potential. VALIDITY OF SEMICLASSICAL TRANSPORT We now consider whether Weyl points can be well sep- arated when the system is in the semiclassical regime. This introduces two competing criteria. First, the Weyl points must be well-resolved: the Fermi level must be close enough to the nodal point that the Fermi surface consists of disconnected pockets encircling each node. Quantitatively, this translates to the constraint, kF ∼ µ v (cid:28) k0 (8) where v = (det A)1/3 is the mean velocity of the Weyl point at position k0, kF is the Fermi wavevector mea- sured as the deviation from k0, and the chemical poten- tial µ is the deviation in energy from the Weyl point. Second, we demand that the number ν of filled Landau 4 levels is large. Recall that for a single Weyl point, √ µ ∼ 2Bν. Hence, we demand, √ 2B µ (cid:29) (9) (10) We now consider when the two constraints (8) and (10) are simultaneously satisfiable. For Weyl points that orig- inate from a symmetry-enforced band touching, such as B, and hence we need si- those in GdPtBi, v ∼ k0 ∼ √ multaneously that B (cid:29) γ1µ and B (cid:28) γ2µ2 (11) where γ1 and γ2 are material-dependent parameters. Whether or not there exists a regime that satisfies Eq (11) depends on µ, which is nearly fixed (for B not too large) for a bulk 3d material. For GdPtBi, we find from experiment[18] that the Weyl points become well resolved for B ∼ 6T ; however quantum oscillations reveal that the Landau level index ν ∼ 5 at this value of the field. However, the preceding analysis ignores the magnetiza- tion of GdPtBi. Near the N´eel temperature of about 9K, the spins will have a large magnetic susceptibility, in which case a smaller field will have the same effect. Additionally, this will be compounded by the quenching of the orbital magnetism in the crystal, leading to an enhancement of the Zeeman energy relative to the cy- clotron energy. In this case, it is quite possible that the experimental regime satisfies Eq (11)[19]. larger than the magnetic field, then k0 ∼ √ If the scale of inversion breaking, D in Eq (1), is much B and v ∼ D. Then Eq (11) is replaced by, √ γ1µ (cid:28) D B and √ B (cid:28) γ2µ, (12) which is satisfied for small enough fields when D exceeds other scales. We now consider Weyl points that emerge from split- In this case, ting a Dirac point with a magnetic field. for the two Weyl points to be well-resolved, the spacing between the Weyl points, which scales like B, must be greater than kF ∼ µ/v. To leading order, v depends on the initial dispersion (i.e., determined by Eq (3)) and only has sub-leading B dependence. This again leads to Eq (11). The recent experiment on Na3Bi[13] reports the nodes to be well separated when B = 12T , but the onset of the lowest Landau level to be near 6 − 8T . Hence, the semiclassical regime will likely not quantitatively de- scribe this experiment. DISCUSSION A magnetic field can create Weyl points from four- band crossings by breaking time reversal symmetry. This is a powerful technique for creating Weyl points whose position in energy-momentum space is tunable. Here, we have studied two canonical and experimentally rele- vant examples: a symmetry-enforced four-band crossing at the Γ point and a Dirac node near the Γ point. We have shown that a complex map of Weyl points (and line nodes) emerges, depending on the direction and mag- nitude of the magnetic field. It would be interesting to experimentally track the movement of these points by ob- serving how the surface Fermi arcs move as the magnetic field is changed, e.g., in STM experiments. Furthermore, for the particular cases of GdPtBi and HgTe, our numer- ical analysis indicates that Weyl points exist near the Fermi level for all directions of the magnetic field: this should prompt future experiments that probe the chiral anomaly with fields away from the high-symmetry axes. We computed the anomalous longitudinal conductance in the semiclassical regime for Weyl points created by a magnetic field. The conductance scales with a higher power of the magnetic field than the conductance for in- trinsic Weyl points. Naively, this is consistent with ex- perimental data: for example, the low-field data in Ref 18 shows that σxx scales like a higher power of B than B2; we plot this data in the Supplement. However, this agree- ment should be taken with a grain of salt, because, as mentioned in the previous section, the experiment is not fully in the semi-classical regime. Our theory will be bet- ter tested in future experiments that are in this regime, where we expect the scaling of longitudinal conductance to go beyond B2 for magnetic-field created Weyl points. Our analysis readily generalizes to other point groups. This would be a useful course of study to identify future candidates for magnetic field created Weyl points. In addition, an analysis of the quantum regime could be used to describe existing experiments. We leave these questions for future works. ACKNOWLEDGEMENTS The authors thank Alexey Soluyanov and Claudia Felser for helpful discussions. BAB acknowledges the hospitality and support of the Donostia International Physics Center and the ´Ecole Normale Sup´erieure and Laboratoire de Physique Th´eorique et Hautes Ener- gies and the support of the Department of Energy de- sc0016239, NSF EAGER Award DMR -- 1643312, Si- mons Investigator Award, ONR-N00014-14-1-0330, ARO MURI W911NF12-1-0461, NSF-MRSEC DMR-1420541, the Packard Foundation, the Keck Foundation, and the Schmidt Fund for Innovative Research. 5 [1] M. Z. Hasan and C. L. Kane, Rev. Mod. Phys. 82, 3045 (2010). [2] Z. Wang, Y. Sun, X.-Q. Chen, C. Franchini, G. Xu, H. Weng, X. Dai, and Z. Fang, Phys. Rev. B 85, 195320 (2012). [3] Z. Wang, H. Weng, Q. Wu, X. Dai, and Z. Fang, Phys. Rev. B 88, 125427 (2013). [4] Z. Liu, J. Jiang, B. Zhou, Z. Wang, Y. Zhang, H. Weng, D. Prabhakaran, S. Mo, H. Peng, P. Dudin, et al., Nat. Mater. 13, 677 (2014). [5] Z. Liu, B. Zhou, Y. Zhang, Z. Wang, H. Weng, D. Prab- hakaran, S.-K. Mo, Z. Shen, Z. Fang, X. Dai, et al., Sci- ence 343, 864 (2014). [6] H. Weng, C. Fang, Z. Fang, B. A. Bernevig, and X. Dai, Physical Review X 5, 011029 (2015). [7] M. N. Ali, Q. Gibson, S. Jeon, B. B. Zhou, A. Yazdani, and R. J. Cava, Inorganic Chemistry 53, 4062 (2014). [8] L. Yang, Z. Liu, Y. Sun, H. Peng, H. Yang, T. Zhang, B. Zhou, Y. Zhang, Y. Guo, M. Rahn, et al., Nat. Phys. 11, 728 (2015). [9] B. Lv, N. Xu, H. Weng, J. Ma, P. Richard, X. Huang, L. Zhao, G. Chen, C. Matt, F. Bisti, et al., Nat. Phys. 11, 724 (2015). [10] X. Wan, A. M. Turner, A. Vishwanath, Savrasov, Phys. Rev. B 83, 205101 (2011). and S. Y. [11] A. A. Burkov and L. Balents, Phys. Rev. Lett. 107, 127205 (2011). [12] A. A. Burkov, Phys. Rev. B 91, 245157 (2015). [13] J. Xiong, S. K. Kushwaha, T. Liang, J. W. Krizan, M. Hirschberger, W. Wang, R. J. Cava, and N. P. Ong, Science 350, 413 (2015). [14] X. Huang, L. Zhao, Y. Long, P. Wang, D. Chen, Z. Yang, and H. Liang, M. Xue, H. Weng, Z. Fang, X. Dai, G. Chen, Phys. Rev. X 5, 031023 (2015). [15] C. Zhang, S.-Y. Xu, I. Belopolski, Z. Yuan, Z. Lin, B. Tong, N. Alidoust, C.-C. Lee, S.-M. Huang, T.-R. Chang, et al., Nature Communications 7 (2016). [16] B.-J. Yang and N. Nagaosa, Nat. Comm. 5, 4898 (2014). [17] J. Ruan, S.-K. Jian, H. Yao, H. Zhang, S.-C. Zhang, and D. Xing, (2015), arXiv:1511.08284 [cond-mat]. [18] M. Hirschberger, S. Kushwaha, Z. Wang, Q. Gib- son, C. A. Belvin, B. A. Bernevig, R. J. Cava, and N. P. Ong, Nature Materials (2016), 10.1038/nmat4684, arXiv:1602.07219. [19] C. Shekhar, A. K. Nayak, S. Singh, N. Kumar, S.-C. Wu, Y. Zhang, A. C. Komarek, E. Kampert, Y. Skourski, J. Wosnitza, W. Schnelle, A. McCollam, U. Zeitler, J. Kubler, S. S. P. Parkin, B. Yan, and C. Felser, (2016), arXiv:1604.01641. [20] T. Liang, Q. Gibson, M. N. Ali, M. Liu, R. Cava, and N. Ong, Nat. Mater. 14, 280 (2015). [21] C.-Z. Li, L.-X. Wang, H. Liu, J. Wang, Z.-M. Liao, and D.-P. Yu, Nat. Comm. 6, 10137 (2015). [22] H. Li, H. He, H.-Z. Lu, H. Zhang, H. Liu, R. Ma, Z. Fan, S.-Q. Shen, and J. Wang, Nat. Comm. 7, 10301 (2016). [23] B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Science 314, 1757 (2006). [24] M. Konig, S. Wiedmann, C. Brune, A. Roth, H. Buh- mann, L. W. Molenkamp, X.-L. Qi, and S.-C. Zhang, Science 318, 766 (2007). 6 [25] V. Mourik, K. Zuo, S. M. Frolov, S. Plissard, E. Bakkers, and L. Kouwenhoven, Science 336, 1003 (2012). [26] A. A. Burkov, M. D. Hook, and L. Balents, Phys. Rev. B 84, 235126 (2011). [27] C.-K. Chiu and A. P. Schnyder, Phys. Rev. B 90, 205136 (2014). [28] Y.-H. Chan, C.-K. Chiu, M. Chou, and A. P. Schnyder, arXiv preprint arXiv:1510.02759 (2015). [29] C. Fang, Y. Chen, H.-Y. Kee, and L. Fu, Phys. Rev. B 92, 081201 (2015). [30] D. T. Son and B. Z. Spivak, Phys. Rev. B 88, 104412 (2013). [31] B. Z. Spivak and A. V. Andreev, (2015), arXiv:1510.01817. [32] See Supplementary material. [33] H. B. Nielsen and M. Ninomiya, Phys. Lett. B 130, 389 (1983). [34] A. A. Soluyanov, D. Gresch, Z. Wang, Q. Wu, M. Troyer, X. Dai, and B. A. Bernevig, Nature , 495 (2015). [35] L. Onsager, Phys. Rev. 37, 405 (1931). [36] L. Onsager, Phys. Rev. 38, 2265 (1931). [37] M. Jonson and G. D. Mahan, Phys. Rev. B 21, 4223 (1980). [38] C. Fang, M. J. Gilbert, X. Dai, and B. A. Bernevig, Phys. Rev. Lett. 108, 266802 (2012). [39] M. A. Stephanov and Y. Yin, Phys. Rev. Lett. 109, 162001 (2012). [40] D. Xiao, M. Chang, and Q. Niu, Rev. Mod. Phys. 82, 1959 (2010). [41] D. Xiao, J. Shi, and Q. Niu, Phys. Rev. Lett. 95, 137204 (2005). [42] D. T. Son and N. Yamamoto, Phys. Rev. D 87, 085016 (2013) Supplementary Material for The chiral anomaly factory: Creating Weyl fermions with a magnetic field 7 0 0 1 0 0 −1 0 0 1 0 0 0  0 0 0 −1  i 0 0 0 0 −i 0 0 0 0 i 0 0 0 0 −i C2y = C2z = 0 −(−1)3/4 0 0 0 0 √−1 4 0   0 0 0 (−1)3/4 0 −(−1)3/4 √−1 4 0 0 (−1)3/4 0 0 0 0 0 0 3 √ 3 i −i −√ √ −i √ −i −i 3 −1 −i −√ 1 √ 3 3 3 −1 −i √ √ 1 3 i 3  √−1 − 4 0 0 0   0 0 √−1 0 − 4 C4zI = M110 = √ 2 4 e−iπ/4 C3,111 = (S6) (S7)  (S8)  (S9)  (S10) (S11) (S12) (S13) (S14) 4D IRREPS AT Γ: GdPtBi, HgTe, InSb In this Appendix, we perform the detailed symmetry analysis of materials with a quadratic four-band crossing at the Γ point in a Zeeman field. We will first derive the most general low-energy k · p Hamiltonian near the Γ point. We then deduce the location and multiplicities of all sets of Weyl points that emerge when the magnetic field is aligned with high-symmetry crystallographic di- rections. In all cases we consider, the low-energy bands come from the p orbitals with total angular momentum J = 3/2. We have chosen the ordered basis J, mj(cid:105) = 2(cid:105). The J = 3/2 angular mo- 3 2 , 3 mentum operators are: √ 2 ,− 1 2 ,− 3 2(cid:105), 3 2(cid:105), 3 2(cid:105), 3 2 , 1  (cid:0)i  0 0 0 √  3(cid:1) (S1) (S2) (S3) 3 2 0 0 √ √ Jx =  0  0 − 1 (cid:0)i  Jz = 0 i 0 2 0 0 √ 2 3 i 3(cid:1) 3 2 0 1 0 0 1 0 √ 3 2 0 0 √ 3 2 0 0 −i 0 − 1 √ i 3 2 2 3 2 0 0 0 1 0 2 0 0 − 1 0 0 0 0 0 0 − 3 2 2 Symmetries Jy = In this basis, we can represent the elements of the sym- metry group Td along with time reversal by the matrices (K represents complex conjugation) T = 0 0 1 0 0 −1 0 0 1 0 0 0  0 0 0 −1  0 0 0 i 0 0 i 0 0 i 0 0 i 0 0 0  K  C2x = Notice the unconventional C4zI symmetry, where I is in- version; inversion by itself is not a symmetry. The Hamil- tonian H(kx, ky, kz) then must satisfy: T H(kx, ky, kz)T −1 = H(−kx,−ky,−kz) C2xH(kx, ky, kz)C−1 2x = H(kx,−ky,−kz) C2yH(kx, ky, kz)C−1 2y = H(−kx, ky,−kz) (S4) C2zH(kx, ky, kz)C−1 2z = H(−kx,−ky, kz) C4zIH(kx, ky, kz)(C4zI)−1 = H(ky,−kx,−kz) (S15) (S5) M110H(kx, ky, kz)M−1 110 = H(−ky,−kx, kz) C3,111H(kx, ky, kz)C−1 3,111 = H(kz, kx, ky) (S16) (S17) Hamiltonian The most general Hamiltonian consistent with Eqs. (S11)-(S17) is given -- to order k2 -- by, H(kx, ky, kz) =(cid:0)A0 + Ak2(cid:1) I4×4 + + C(k2 x − k2 y)Γ1 + C√ 3 (2k2 z − k2 x − k2 y)Γ2 + + E(kxkyΓ3 + kxkzΓ4 + kykzΓ5) + + D(kxU1 + kyU2 + 2kzU3) (S18) From here on, we ignore the overall energy shift by setting A0 = 0. Inversion symmetry is preserved when D = 0. The matrices Γ1...5, U1,2,3 are given by: Γ1 = 1√ 3 x − J 2 (J 2 y ) = Γ2 = 1 3 (2J 2 z − J 2 x − J 2 y ) = 8 and U1,2,3 can be written down in terms of the Clifford algebra matrices and their commutations (the full su(4) algebra): Γab = 1 2i [Γa, Γb] (S21) as: √ 3Γ15 − Γ25; U1 = U2 = −√ 3Γ14 − Γ24; U3 = Γ23 (S22) The ab initio fit yields C A D E GdPtBi −1.46 8.20 0.08 −5.4 2.85 11.20 0.04 −12.0 HgTe 8.85 14.20 0.01 −22.0 InSb (S23)  (S19) (S20) where A, C, and E are in units of eVA2 and D is in units of eVA. Along the kz axis, this structure is in- verted, with the Jz = ±3/2 heavy bands dispersing up- wards (electron-like) and the Jz = ±1/2 bands dispersing downwards (hole-like). Weyl Nodes When time-reversal symmetry is broken, a complicated phase diagram of Weyl points appears. The number of these Weyl points depends on the direction of the magnetic field, on the g-factors of the problem, and on whether the magnetic field couples differently to heavy and light holes. We can use the k · p model (S18), with the parameters (S23), to find the Weyl points in most of the interesting cases, and we expect it to be reliable. Since simulating g-factors is notoriously hard, we use an effective model for the Zeeman coupling where we add to the Hamiltonian in Eq. (S18) the Zeeman term: HZ = (cid:126)B · (cid:126)J (S24) where Jx, Jy, Jz are the spin-3/2 matrices of Eqs. (S1, S2, S3). We have assumed that the g-factor is positive, without loss of generality. We now analyze the spectrum as we align the magnetic field with high-symmetry crys- tallographic axes. Field in [001] direction At (cid:126)k = 0, the presence of the magnetic field splits the 3/2 multiplet into four bands with energies 3/2B, 1/2B,−1/2B,−3/2B; the first and fourth bands disperse upwards, while the second and third disperse downwards. We first consider the spectrum when (cid:126)k is in the [001] direction. There are two Weyl nodes on the 001 high 0 0 0 −1 0 0 0 0 −1 0 0 1 0 0  0 0 0 1 1 0 0 0 0 1 0 0 0 0 0 −i 0 i 0 0 0 0 i 0  0 0 1 0  1 0  0 0 −i 0  0 1 0  0 −i 0 0   i 0 0 0 0 i 0 0 0 0 −i 0 √ 0 0 1 0 0 0 0 0 0 −1 0 0 −1 0 3 0 1 0 √ −i 0 i 0 3 3 3    Γ3 = 1√ 3 {Jx, Jy} = Γ4 = 1√ 3 {Jx, Jz} = Γ5 = 1√ {Jy, Jz} = U1 = U2 = 3  0  0 −i √ 0 i U3 = 0 −√ 1 √ 3 1 0 −√ 0 3 0 1 √ i 0 0 −i √ 0 3 −i i 3 0  0 0 −1 0 0 0 0 1 −1 0 0 0 0 1 0 0  Notice the Γ1...5 are Clifford algebra matrices: {Γa, Γb} = 2δab at (cid:126)k = (0, 0,±(cid:113)√ symmetry axis between the −3/2 and −1/2 bands. Since the band of energy −3/2B at the Γ point disperses up- wards, while the bands of energy 1/2B,−1/2B disperse downwards, they will invariably intersect. The crossing between the −1/2 and −3/2 bands is protected, as the bands have different eigenvalue under C2z -- their eigen- values are ±i, as given by Eq (S7). There are two Weyls 3B 4C ). They are related by C4zI, which means that they have opposite chirality due to the inver- sion operation (inversion changes the Weyl charge, TR leaves it invariant, rotations leave it invariant and mir- rors -- the product of rotation times inversion -- change it.) They are shown in Fig. S2a. We move to analyze the physics between the −3/2 and 1/2 bands (the −3/2 band disperses upwards, the 1/2 band downwards and they start inverted). Their cross- ings in the [001] direction are avoided. However, we now give a topological proof why they must give rise to Weyl points elsewhere in momentum space. Consider the plane kz = 0. It is a gapped plane with respect to the bands −3/2, 1/2. At the Γ point, the C4I eigenvalues of these bands are exp(i3π/4), exp(i7π/4) respectively. The −3/2 rests below the 1/2 band at the Γ point. Away from the Γ point, if we believe the k·p model cor- rectly describes the physics, the bands at very large mo- mentum (or (π, π) on the lattice, which is infinity in the continuum) are ordered in energy in the normal 1/2,−3/2 way and so their C4 eigenvalues are exactly the opposite of those at the Γ point. Now take the plane kz = π(= ∞). In this plane the bands have C4zI eigenvalues exp(i7π/4), exp(i3π/4), or- dered in energy. Hence, going from the kz = 0 to kz = π plane, the C4zI eigenvalue of the lower band has changed from exp(i3π/4) to exp(i7π/4). Ref [38] relates the eigen- values of the C4 operator of the occupied bands to the quantity iC where C is the Chern number of the bands: iCkz =0−Ckz =∞ = exp(i3π/4) exp(i7π/4) = −1 these points is at (S25) Hence, Ckz=0 − Ckz=∞ = 2. Since both those planes are gapped with respect to the −3/2 and 1/2 bands, we see that two (mod four) Weyl nodes have to reside between kz = 0 and kz = ∞ planes. Suppose one of some (cid:126)k = (kx0, ky0, kz0). By C2z symmetry, its partner is at (−kx0,−ky0, kz0). However, unless kx0 = 0 or ky0 = 0, the product C2xT , which remains a symmetry in the presence of a magnetic field (with a magnetic field in the [001] direction, C2x flips the field, but time-reversal flips it back), implies two additional Weyl points between kz = 0 and kz = ∞ planes at (∓kx0,±ky0, kz0). Thus, the two Weyl points required by the Chern number argu- ment in Eq (S25) must lie in either the kx = 0 or ky = 0 plane. Figure S2b shows the pair of Weyls in the kx = 0 plane. 9 Suppose they lie in the kx = 0 plane, at (0, ky0, kz0). Then, by C4zI symmetry, there must be two more Weyl points at (±ky0, 0,−kz0) (which could have been de- rived by applying the same Chern number argument in Eq (S25) to the kz = 0 and kz = −∞ planes). In summary, we have proven that (at least) six Weyl points must exist -- two on a high symmetry line and two on high-symmetry planes. In general, additional Weyl points might exist at generic points in the Brillioun zone. The combined presence of C4zI and C2xT implies that these Weyl points come in sets of eight. Thus, the total number of Weyl points that can exist is 6 + 8k, where k ∈ Z. This is shown in Table I. Due to the C3 symmetry of the crystal, an identical argument goes through for (cid:126)B aligned with the 100 and 010 directions as well. Field in [110] direction With (cid:126)B in the [110] direction, it is convenient to work in the basis of eigenstates of Jx + Jy ≡ J110. (S26) At the Γ point, eigenstates of the Hamiltonian coincide with eigenstates of J110. As before, the bands with J110 eigenvalue ±3/2 disperse upwards, and the bands with eigenvalue ±1/2 disperse downwards. With the magnetic field in this direction, the only re- maining symmetries of the Hamiltonian are the mirror symmetry M110, and the combined symmetry C2zT . The points (cid:126)k = (k,−k, kz) are fixed by the action of M110, while the points (cid:126)k = (kx, ky, 0) are fixed by C2zT ; these two subspaces are the high-symmetry planes we will an- alyze. (The intersection line (cid:126)k = (k,−k, 0) of these two planes is invariant under M110C2zT . However, as an anti-unitary symmetry that squares to +1, it does not permit band crossings on a line without fine-tuning, for the following reason: a generic two-band Hamiltonian on the line takes the form H(k) = d(k) · σ. An anti- unitary symmetry that squares to +1 can be represented by K, the complex conjugation operator. The require- ment [H(k), K] = 0 implies dy(k) = 0. Thus, H(k) is described by two functions, dx,z, which are a function of one variable, k. Without fine-tuning, it is not generically true that there exists a k0 where dx(k0) = dz(k0) = 0.) We will start by analyzing the plane fixed by M110. Along this plane the energy eigenstates are M110 eigen- states. Going away from the Γ point, the bands with J110 eigenvalues (3/2, 1/2,−1/2,−3/2) have, respectively, the mirror eigenvalues (i,−i, i,−i). Based on the direction of the band dispersion, we then expect the J110 = −1/2 band and the J110 = −3/2 bands to cross in the high symmetry plane due to their differing mirror eigenvalues. It is not hard to show that such crossings produce line nodes. In the vicinity of the crossing, we can represent the mirror symmetry, which squares to −1, as M110 = iσy. The effective Hamiltonian on the high symmetry plane takes the form Hef f (k,−k, kz) = di(k,−k, kz)σi. (S27) Imposing the symmetry M110Hef f (k,−k, kz)M−1 Hef f (k,−k, kz) requires 110 = Hef f (k,−k, kz) = dy(k,−k, kz)σy. (S28) Since there is one parameter dy and two momenta (k, kz), any point where dy(k,−k, kz) = 0 is generically part of a line node where dy(k, kz) = 0. Using the ab initio parameters (S23), we are guaran- teed such a point. Furthermore, when D = 0, so that in- version symmetry is present, we have two such line nodes at inversion-symmetric points. We expect this to hold true for small values of D as well, and for the physically relevant situation D = 0.08, we do, in fact, observe two line nodes, which are shown in Fig. S3a. However, as inversion symmetry breaking is increased to unphysical values, the bands tilt so that they only cross on one side of the origin and only one line node exists. Let us now examine the high symmetry (kx, ky, 0) plane. This plane is fixed by C2zT , which is antiuni- tary and squares to +1. Assuming there exists a crossing on this plane, we can write an effective two-band Hamil- tonian, and represent C2zT by K. An analysis similar to the previous section shows that after enforcing this symmetry, the Hamiltonian consists of two parameters, which are a function of two momenta. Hence, on this plane, there can generically be isolated band crossings in the (kx, ky, 0) plane. The M110 symmetry requires that a putative Weyl point at (kx, ky, 0) have a partner Weyl at (−ky,−kx, 0). If the line nodes in the (k,−k, kz) plane persist to kz = 0, they will intersect the (kx, ky, 0) plane, twice each. By tuning the parameters in the Hamiltonian (for instance the degree of inversion symmetry breaking), we can change the number of line nodes. Hence, the line nodes can create pairs of band degeneracies in the plane (kx, ky, 0), with the number of pairs determined by the ab initio parameters. Using the particular ab initio values (S23), we observe four band crossings in the (kx, ky, 0) plane, and all four lie on the line nodes. Similar arguments go through when considering the 3/2 and 1/2 band, which also have differing mirror eigen- values. We find from the k · p Hamiltonian that there is also a pair of line nodes between these bands, however, using the ab initio parameters, this pair sits much farther from the Γ point in the Brillouin zone; whether or not this pair is an artifact of the k · p expansion would have to be confirmed by further ab initio calculations. Finally, we find four additional Weyl points near the (k, k, kz) plane. We prove their existence in the inversion- symmetric case D = 0, where they lie exactly on this 10 plane; the nontrivial Chern number associated with the Weyl points guarantees that they survive small inversion breaking. When D = 0, there is an additional unitary symmetry, C2,110, and antiunitary symmetry, C2,1¯10T , which satisfy, 2,110 = −1; (C1¯10T )2 = 1;{C2,110, C2,1¯10T} = 0 (S29) C 2 Both of these symmetries leave invariant the line (cid:126)k = (k, k, 0); the former also leaves invariant the plane (cid:126)k = (k, k, kz). We start by considering the special value of parame- ters, E = 2C in Eq (S18). On the line kx = ky, kz = 0, the Hamiltonian reduces to: (cid:18) (cid:19) H( k110 2 , k110 2 , 0) = A k2 110 2 I + Ck2 110 2 − Γ2√ 3 + Γ3 √ (S30) where we define k110(1¯10) = (kx + (−)ky) /2. The Hamil- tonian (S30) commutes with Jx + Jy. Thus, when the Zeeman term (S24) is added, the four bands can be la- belled not only by their C2,110 eigenvalue, but by their eigenvalue under (Jx + Jy)/ 2. The bands with eigen- values ±1/2 both cross the band with eigenvalue −3/2. We now analyze each of these crossings to show which are protected as we move away from the E = 2C point -- if the band crossings are Weyl points, they will persist for small perturbations, but we want to know whether they persist for a large change in parameter space, where E and C are given by the parameters (S23). We first consider the crossings between the −1/2 and −3/2 bands. An effective Hamiltonian for these two bands is given by H = di(k110, k1¯10, kz)σi, for some functions di. In this effective space, we can choose C2,110 to be represented by iσy and C1¯10T by σzK, where K is the complex conjugation operator; this choice is con- sistent Eq (S29) and consistent with the fact that these bands have distinct eigenvalues ±i under C2,110. Enforc- ing C2,110 yields 2 ,− 1 − 3 eff 2 dx,z(k110, k1¯10, kz) = −dx,z(k110,−k1¯10,−kz) dy(k110, k1¯10, kz) = dy(k110,−k1¯10,−kz), (S31) while enforcing C1¯10T yields, dx(k110, k1¯10, kz) = −dx(k110,−k1¯10, kz) dy(k110, k1¯10, kz) = dy(k110,−k1¯10, kz) dz(k110, k1¯10, kz) = dz(k110,−k1¯10, kz) (S32) To linear order in k, these sets of constraints yield dx ∝ k1,¯10, dz ∝ kz, dy ∝ k110 + d0, for some constant d0 that determines the location of the band crossing. Hence, the band crossings between these bands are two single Weyl points. When E is tuned away from 2C, the Weyl points will remain on the line (k, k, 0) (Weyl points off this line come in sets of four or eight). Thus, to determine in (cid:115) what parameter regime the Weyl points persist, we need only diagonalize H(k, k, 0) + (Jx + Jy)B/ 2 and check for crossings between the −3/2 and −1/2 bands. This Hamiltonian has eigenvalues, √ ± η B 2 B2 + η B√ 3 k2 (2C + 3E) + 4 3 C 2k4 + E2k4 (S33) where η = ±1 and we have omitted the shift 2Ak2 in all band energies. The −3/2 and −1/2 bands intersect only when E(2C + E) > 0. Thus, for C > 0, the Weyl points exist only when E > 0 or E < −2C (when C < 0, they exist when E < 0 or E > 2C); at the transition points, the Weyls annihilate at k → ∞. Referring to the ab initio values in (S23), we see that in the regime of interest for this particular material, the Weyl points between these two bands do not exist. We have confirmed this analysis of the k · p model numerically. We next consider the crossing between the 1/2 and −3/2 bands. Instead of applying a symmetry argument √ similar to that in the previous paragraph, we directly 2 about the desired expand H(kx, ky, kz) + B(Jx + Jy)/ project onto the 1/2 and −3/2 bands to compute the 2×2 effective Hamiltonian, H band crossing at k110 = 31/4(cid:112)B/C, k1¯10 = kz = 0 and (cid:1) = c0I + ciσi, where − 3 2 , 1 eff √ √ 2 (cid:0)k2 + 1¯10 + 2k2 z (A c0 = 3 − 3C)B √ 2C cx = − 31/4 BC√ √ 2 31/4 BC√ √ 2 BC 31/4 cz = − cy = δ110 + δ110 − C√ 2 C√ 3 δ110 + BC 31/4 C√ 1¯10 − 2Ck1¯10kz k2 2 1¯10 − 2Ck1¯10kz k2 1¯10 − 4k2 (cid:0)k2 (cid:1) δ110 − C√ 3 where δ110 ≡ k110 − 31/4(cid:112)B/C. Defining a rotated set (S34) z of Pauli matrices σi = Oijσj, where  1/ √ 1/(2 3/(2 O = √ √ √ √ 2) −√ 2 1/ 2 2) −1/(2 √ √ 2) −√ 3/(2 2) 1/2 0 3/2  , (S35) 2 , 1 − 3 eff 2 we find H = ci σi, where √ 2Ck1¯10kz 1¯10 − 2k2 z) cx = −2 cy = C(k2 cz = δ110 (S36) √ √ where δ110 ≡ −2(3−1/4) 3 z)/ In this is a shift and rescaling of the δ110 axis. describes a double Weyl form, it is evident that H point. When E is tuned away from its special value of 2C, this double Weyl, and its partner at k110 = BCδ110 + C(k2 1¯10 + 2k2 − 3 eff 2 , 1 2 −31/4(cid:112)B/C, k1¯10 = kz = 0, split into four single Weyl 11 points, which are restricted to lie in the (k110, kz) plane (because Weyl points off this plane come in sets of eight). The single Weyl points can pairwise annihilate only on the kz axis. Thus, if the kz axis is gapped for all value of E at fixed C, then these four single Weyl points persist through the large parameter change from E = 2C to E and C described by (S23). The Hamiltonian along the kz axis, including the Zeeman √ 2, which term, is given by H(0, 0, kz) + (Jx + Jy)B/ 1 + 2√ 3 x4 has eigenvalues B 3 (cid:17) (cid:112)C/Bkz. It is straightforward to check that these eigen- (cid:16)−1/2 + x2/ 3 ±(cid:113) 3 ±(cid:113) , where x = 1/2 + x2/ 1 − 2√ x2 + 4 x2 + 4 and B (cid:16) (cid:17) 3 x4 √ √ 3 values are distinct for all real values of x. Thus, we have shown that the two double Weyl points that describe the crossings between the −3/2 and +1/2 bands at the spe- cial parameter value E = 2C split into four single Weyl points, which survive as E and C are varied to the values in (S23). Restoring D = 0.08 (cid:54)= 0, we argue that, while these nodes may move off the symmetry plane, they do not an- nihilate. This is because their opposite chirality partners are generically far away in the Brillouin zone. Lastly, due to the symmetries of the crystal, a com- pletely analogous analysis holds for (cid:126)B aligned with any of the [101], [011], [¯110], [¯101], or [0¯10] directions. Field in [111] direction Define J111 = 1√ 3 (Jx + Jy + Jz). Then C3,111J111C−1 √ 3,111 = J111. 3)/2, respectively. √ 3)/2, (1 + i (S37) Thus, when (cid:126)B[111], eigenstates of J111 are simultane- ously eigenstates of C3,111. In particular, the states with J111 eigenvalues 3/2,−3/2, 1/2,−1/2 have C3,111 eigen- values −1,−1, (1 − i At (cid:126)k = 0, the bands corresponding to ±3/2 curve downwards, while the bands corresponding to ±1/2 curve upwards (notice that along the x, y, z directions, along the (cid:126)k (cid:126)B axis, the opposite is true, that is, the ±3/2 bands curve upwards.) Hence, there are possible cross- ings between the 3/2 band and the ±1/2 bands (while the −3/2 band has no potential crossings.) Along the high-symmetry axis kx = ky = kz, these bands all have distinct eigenvalues of C3,111; hence, these crossings are protected, and we see four Weyl points. Weyl points are exactly at (cid:126)k = ±(cid:112)B/E(1, 1, 1) and ±(cid:112)B/(2E)(1, 1, 1), where (cid:126)B = B(1, 1, 1). When D (cid:54)= 0, At the inversion symmetric point (D = 0), these we can also find an analytical expression for the Weyl points along this axis, but it is not elucidating. These four Weyls are shown in Fig. S3b. Notice that if there is an additional Weyl node at some point (kx0, ky0, kz0), then the combination of C3,111 and C4zI dictates that there are five more symmetry-related Weyls. Hence, the total number of Weyl nodes is 4 + 6k; numerical diagonalization of the k · p Hamiltonian in this case indicates that k = 1 for the ab initio param- eters (S23). Proof that Weyl points exist for all directions of B We now prove that for any generic material parame- ters, Weyl points exist in all directions of the magnetic field when inversion symmetry is present; the topological nature of the Weyl points ensures that they will persist for small inversion symmetry-breaking. For arbitrary B, consider the Hamiltonian H(k) + HZ ≡ H(k) + J · B, where H(k) is defined in Eq (S18); then set D = 0 to preserve inversion symmetry and, without loss of gener- ality, set A = A0 = 0, which fixes the zero of energy. The Hamiltonian is now a function of the parameters E and C, as well as of k and B. Now consider the fine- tuned point in parameter space, E = 2C: with this choice of parameters, [H(k), k · J] = 0, and, in partic- ular, [H(kB), B · J] = 0, rendering the Hamiltonian ex- actly solvable along the axis k (cid:107) B. The eigenvalues of H(kB) + J · B are given by 1 3Ck2B2 ± 3B(cid:1) and 3Ck2B2 ± 9B(cid:1). For finite B, Weyl points exist at (cid:0)−4 (cid:0)4 1 6 √ 4C and k = ±31/4/ √ k = ±31/4/ 2C, as shown below. The band crossings are protected because all four bands have different eigenvalues of B · J. As E is varied away √ √ 6 FIG. S1. Schematic depiction of the four Weyl points that exist along the direction k (cid:107) B when E = 2C. from 2C, the four Weyl points move out of the k (cid:107) B plane, to generic positions ±k1,2, consistent with inver- sion symmetry. Notice that the Weyl points at ±k1 can- not annihilate each other (nor can the points at ±k2) be- cause to do so while enforcing inversion symmetry would require two bands to become degenerate at the origin; however, for finite B, the Hamiltonian has four distinct eigenvalues at the origin (it is equal to J · B, which has eigenvalues ± 1 2 B). Thus, the Weyl points can only 2 B, ± 3 12 annihilate if there exist values of E and C such that k1 = ±k2. If k1 is near ±k2, the effective Hamiltonian of the two nearby Weyl points requires only three bands, i.e., it is a 3 × 3 Hermitian matrix, which is described by eight functions, di(k), i = 1, . . . , 8, which are determined by E and C. The band crossing occurs when di(k) = 0 for all i. Generically, a system of eight equations involving five variables (E, C, and kx,y,z) has no solutions. If solutions exist, then they exist as isolated points, (E∗, C∗, k∗). We now argue that these Weyl points persist for physi- cal values of the material parameters. Given a particular material, described by generic values E1 (cid:54)= 2C1, consider tuning E and C from (E, C) = (2C0, C0), for arbitrary C0 (cid:54)= 0, to (E1, C1). We can always choose a path in parameter space that avoids the isolated points (E∗, C∗) which permit three-band crossings. This leaves no op- portunity for the Weyl points that exist at (2C0, C0) to annihilate. Hence, for arbitrary B and generic values of E and C, at least four Weyl points exist. Numerical Observations We confirmed the above analysis for GdPtBi, HgTe and InSb by numerically diagonalizing the k · p Hamil- tonian, using the parameters in Eq (S23). Surprisingly, this procedure showed that there exist Weyl nodes for the magnetic field in any direction; the field-induced Weyls are surprisingly universal. Furthermore, we find that while all of the Weyl nodes on the high symmetry planes mentioned in the previous sections were of Type I, some of the Weyl fermions that emerge in the numerics are the Type II Weyls introduced in Ref 34. We now list some mention some specific results for GdPtBi. Focus- ing on the middle two bands and for magnetic fields of unit magnitude in the [11x] direction, we find that there are at least four Weyl nodes for any x ∈ [0, 1], and that for x = 0.4121, two of these nodes transition from Type I to Type II. These critical Weyls are located at k = (−0.0461,−0.302, 0.244) and k = (0.302, 0.0461,−0.244) (The symmetry in location is due to M110C2zT , which preserves any magnetic field of the form (B, B, Bz)). The energy of these crossings is E = −0.225. The two Type I Weyls are located at k = (0.318, 0.312, 0.330) and k = (−0.312,−0.318,−0.330), with energies E = −0.03. Cd3As2 k · p Hamiltonian In Cd3As2, there are eight bands of interest: two s- orbitals with J, Jz(cid:105) = 1/2,±1/2(cid:105) and six p-orbitals with J, Jz(cid:105) = 3/2,±3/2(cid:105),3/2,±1/2(cid:105),1/2,±1/2(cid:105). In the presence of spin-orbit coupling, the states nearest to the 13 (a) (b) with (cid:126)v = ±(0, 0, 4.85) and A = FIG. S2. (a) shows two Weyl nodes along the line (0, 0, kz) with (cid:126)B = (0, 0, 20). In the figure we have increased D by a factor of 40 in order to resolve the avoided crossings. The Weyl nodes are type I, and are described by an effective two-band Hamiltonian  −9.39 0 0 0 −9.23 0 0 0 ±42.98 . (b) shows the non-trivial Weyl nodes in the kx = 0 plane, for . This is a Type I Weyl. The other is  0 1 0 3.23 0 2.94 19.55 0 57.57 the same values of the field and material parameters (D = 5) as in the previous figure. The linearized Hamiltonian for the rightmost Weyl has (for D = 0.08) (cid:126)v = (0,−0.24, 6.12) and A = related to it by C2z. (a) (b) (a) shows the spectrum with (cid:126)B = (20, 20, 0), plotted along the high symmetry line (cid:126)k = (kx,−kx, 0). The two FIG. S3. crossings extend to line nodes for kz (cid:54)= 0. This can be seen from the linearized effective Hamiltonian, which near the left-most  (b) Shows the spectrum with (cid:126)B = (20, 20, 20), plotted along node has (cid:126)v = (−17.84, 17.84, 0) and A = (cid:126)k = (k, k, k). There are four Weyl nodes protected by C3,111 symmetry. At B = 20(1, 1, 1), we have computed the velocities and A matrix to verify that the crossings are linear and that the Weyls are Type I. For the nodes at k ≈ .707 and k ≈ 1.00 on the plot, we find the velocities v = −6.938(1, 1, 1) and v = −9.812(1, 1, 1), respectively, with corresponding matrices: 0 0 43.56 17.65 0  0 43.56 17.65  −32.02 58.48 −23.09 −34.63 56.97 −23.09 1.510 −23.09 66.65 0   24.15 −39.93 −16.33 40.88 −16.33 22.51 −46.66 −0.95 −16.33  and A = A = The C3,111 symmetry dictates that the velocities are the same and that the entries in the last column of the A matrix are the same. We have checked that (AAT )ij − vivj has three positive eigenvalues in both cases. Fermi level come from the heavy-hole p-states 3/2,±3/2(cid:105) and the conduction s-states 1/2,±1/2(cid:105)[3]. Since the bands are inverted (the p-orbitals disperse downwards and the s-orbitals upwards), they will intersect at some point in the Brillouin zone, but not at the Γ point. These facts comprise the major difference between the previous analysis of GdPtBi and the current analysis of Cd3As2: in the former case, the low-energy bands comprise a four- dimensional irreducible representation (irrep) and, con- sequently, form a four-band multiplet at the Γ point; in contrast, the low-energy bands in C3As2 come from two two-dimensional irreps, which are not generically degen- erate at the Γ point, but must cross somewhere in the Brillouin zone. In both cases, band inversion is crucial for Weyl points to emerge in the presence of a magnetic field. We work in the basis jz = 3/2, 1/2,−1/2,−3/2. We take the D4h symmetry group for Cd3As2[7], which has generators  C4z = −(−1)1/4 0 −(−1)3/4 0 0 0 0 (−1)1/4 0 0 0 (−1)3/4 0 0 0 14 It is evident that the doubly-degenerate bands cross each other when M(k) = A(k) = B(k) = 0. For generic parameters, this occurs when kz = ±(cid:112)−M0/M1, k(cid:107) = 0. Thus, it is crucial that M0 and M1 have opposite signs, in agreement with the fit in Eq (S45). Material parameters Ab initio calculations give the order k2 parameters in (S42): M0 = .0205, M1 = −18.77, M2 = −13.5, A = 0.889 C0 = −.0145, C1 = 10.59, C2 = 11.5 (S45) Energy is in units of eV when momentum is in units A−1.  (S38) (S39) (S40) C2x = M001 = 0 0 i 0 0 −i 0 0 −i 0 0 0 0 i 0  0 0   0 0 0 −1  K −i 0 0 0 0 −i 0 0 0 i 0 0 0 0 i 0 0 0 −1 0 0 0 1 0 1 0 0 0 T = in addition to time reversal, which takes the form Magnetic field The Zeeman term takes the form: HZ = −(gsJs + gpJp) · B where, in our basis, 0 0 (cid:0) 1 0  , (Jp)z = (cid:1) 0 0 0 0 2 σi 0 0 0 (Js)i = (S41) (S46)  (S47)  3 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 − 3 2 These symmetries imply C2z, C2y, C2,110, C2,1¯10, M100, M010, M110, M1¯10, inversion and the product of inversion and C4z. The k · p Hamiltonian to order k3 derived from the H0(k) = 0(k) + symmetries above takes the form: M(k) A(k) A∗(k) −M(k) B∗(k)  A(k) ≡ Ak−(cid:0)1 + Azk2 (cid:1) + k+ + + B2k2−(cid:1) kz B(k) ≡(cid:0)B1k2 M(k) ≡ M0 + M1k2 z + M2k2(cid:107) where 0 z 0(k) ≡ C0 + C1k2 z + C2k2(cid:107) and k± = kx ± iky and k2(cid:107) = k2  B(k) 0 0 B(k) −M(k) −A(k) B∗(k) −A∗(k) M(k) 0 (S42) (cid:0)A1k2 + + A2k2−(cid:1) (S43) x + k2 y = k+k−. The energies of this Hamiltonian can be solved for ex- actly and are found to be doubly degenerate everywhere: E(k) = 0(k) ±(cid:112)M(k)2 + A(k)2 + B(k)2 (S44) Since (Jp)x,y mix with bands further away in energy that are not included in the k · p, we determine their effec- tive forms perturbatively below when necessary. Notice that in (S46), we have included two g-factors because the bands come from different representations. We now consider the band crossings when the magnetic field is along one of the high-symmetry axes. Field in the z direction The kz-axis is invariant under C4z. Since H0 + HZ is diagonal along this axis when B = Bz z, Eq (S38) shows that each of the four bands has a different eigenvalue of C4z. Hence, band crossings along this axis are protected; we show below that there are between four and eight band crossings along this axis, depending on the magnitude of Bz and gp,s. The kz = 0 plane is also a high-symmetry plane, invari- ant under M001 = Diag[−i,−i, i, i]. For small Bz, this plane is gapped. As Bz is increased, line nodes emerge between different bands depending on the relative signs and magnitudes of gs and gp. We now compute the effective two-band Hamiltonians that describe the band crossings along the kz axis. We first consider the crossing between the jz = 3/2 and jz = z0 = ((gs − 3gp)Bz/4 − 1/2 bands, which occurs at k2 M0)/M1. The linearized Hamiltonian is: 3 2 1 2 H eff = 2M1kz0δzσz + a(kxσx + kyσy), (S48) z0) and δz ≡ kz − kz0. This band where a = A(1 + Azk2 crossing describes a single Weyl point. The band crossing between jz = −3/2 and jz = −1/2 is similar: the Hamil- tonian has an overall negative sign and kz0 is the same after the replacement Bz → −Bz. For any magnitude of B, at least one of these pairs of single Weyl points will exist; for small enough B, both pairs exist. The band crossing between the jz = 3/2 and −1/2 z0 = (−Bz(gs + 3gp)/4 − M0) /M1. (cid:33) bands occurs at k2 The effective Hamiltonian is: 2(cid:0)B2k2− + B1k2 (cid:1) kz0 2M1kz0δz (cid:32) 2 ,− 1 2(cid:0)B2k2 + + B1k2−(cid:1) kz0 −2M1kz0δz H = eff + 3 2 (S49) which describes a double Weyl point. The crossing be- tween the jz = −3/2 and 1/2 bands is similar: the diag- onal entries of the Hamiltonian have the opposite sign, while the off-diagonal entries are the same, and kz0 is found by taking Bz → −Bz. For any magnitude of B, at least one of these pairs of double Weyl points will exist; for small enough B, both pairs exist. Thus, when B = Bz z, at least one pair of single and one pair of double Weyl points emerge on the kz axis. For small enough B, two pairs of single and two pairs of double Weyl points emerge. Field in the x direction The kx axis is invariant under C2x symmetry. This can protect crossings between any two bands with opposite C2x eigenvalues, for a maximum of four total Weyl points. The kx = 0 plane is invariant under the mirror symme- try, M100. Under M100, one of the s-orbitals has eigenval- ues +i and the other −i; the same holds for the p-orbitals. This protects up to two line nodes, between the s-orbital with jx = ±1/2 and the p-orbital with jx = ∓3/2. For any value of Bx, it is guaranteed that at least one of these will exist; if B is small enough, both will appear. The other two potential crossings are avoided. The ky = 0 plane is also a high-symmetry plane, invari- ant under the product C2yT , but this plane is generically gapped. We now now verify these claims using the k · p model: only two of the four p-orbitals with J = 3/2 are included in the low-energy k · p model (S42); in particular, the Jz = ±3/2 orbitals are included while the Jz = ±1/2 orbitals are not. However, the Zeeman term will mix all four orbitals. Assume the Jz = ±1/2 orbitals are separated by some energy, ∆, from the Jz = ±3/2 bands 15 in the k·p model (S42). Then when B = Bx x, the Zeeman term can be added perturbatively to (S42), which yields the following Hamiltonian, H(k)= 0(k)+ A∗(k) B∗(k) −Bxδ2 M(k)+Bxδ A(k) B(k) −M(k) − 1 2 gsBx − 1 2 gsBx −M(k) B∗(k) −A∗(k) M(k)+Bxδ (S50) where δ ∝ Bx/∆. This is the lowest order term that is consistent with C2x symmetry, which remains a symme- try when B = Bx x. −Bxδ2 B(k) −A(k)   Since the kx axis is gapped when Bx = 0, Weyl points do not emerge along this axis until Bx is large enough. When this value is reached, two or four Weyl points emerge. By computing effective two-band Hamiltonians, we have verified that these are all single Weyl points. , Along the kz axis, the eigenvalues are given by, E+± = M(k) + Bxδ ± Bxδ2 E−± = −M(k) ± 1 2 gsBx (S51) There are four possible crossings, either E+± = E−± or E+± = E−∓. In the first case, the band crossings occur at kz0 satis- fying 2M(kz0) = Bx( gs 2 − δ − δ2). To leading order, the effective Hamiltonian between the bands E++ and E−+ is given by, Heff = −2(cid:0)δzkz0M1 + (k2 (cid:0)1 + Azk2 (cid:1) σx + 4kz0(B2 − B1)kxkyσy(S52) x + k2 y)M2 (cid:1) σz + − Akx z0 where δz ≡ kz − kz0. This is part of the line node in the kx = 0 plane: setting kx = 0, Heff → −2(δzkz0M1 + k2 yM2)σz, for which there is a line of degenerate eigenval- ues. The effective Hamiltonian between E+− and E−− is the complex conjugate of this one. The other band crossings occur at kz1 satisfying 2 − δ + δ2). The effective Hamilto- 2M(kz1) = Bx(− gs nian at the crossing between E++ and E−− is, to leading order, Heff = −2(cid:0)δzkz1M1 + (k2 (cid:1) σz (cid:0)1 + Azk2 y)M2 z1 (S53) (cid:1) σy − 2(B1 + B2)(k2 x − k2 x + k2 y)kz1σx − Aky This describes a point crossing, however, the Chern num- ber is zero. Thus, it is not a Weyl point. The same analysis holds when B is in the [010], [110], or [1¯10] directions, which also have C2 symmetry. Band crossings along the kz axis for arbitrary B On the kz axis, the k · p Hamiltonian to the form: H(kz) = (cid:48)(kz) + M(cid:48)(kz)diag[1,−1,−1, 1], where (cid:48)(kz) and M(cid:48)(kz) are symmetry constrains in addition to time reversal, which takes the form 0 0 0 −1 0 0 −1 0 0 0 1 0 1 0 0 0  K T = 16 (S57) These symmetries imply C3z; C2 rotations about and mirror planes perpendicular to: y, z, a, b, c, d; and inver- sion. We have defined the unit vectors corresponding to π/6 rotations of x: √ 3y √ x + (cid:16) (cid:16)−x + (cid:16)√ (cid:16)− 3x + y √ (cid:17) (cid:17) 3y (cid:17) (cid:17) a = b = c = d = 1 2 1 2 1 2 1 2 even functions of kz that are equal to 0(kz) and M(kz) to order k2 z. Thus, the Hamiltonian with the Zeeman term, HZ in Eq (S46), splits into two 2 × 2 blocks, (cid:48)(kz) +M(cid:48)(kz)− gpJp · B and (cid:48)(kz)−M(cid:48)(kz)− gsJs · B. 4 B. From the pre- The eigenvalues of gsJs · B are ± gs vious subsection, we know the eigenvalues of gpJp · B do not take such a simple form, but take the form E0(Bx,y)± Ep(Bx,y,z). Since HZ has no zero eigenvalues for finite B, there will be band crossings along the kz axis 4 B, where the ± signs are when 2M(cid:48)(kz) = E0 ± Ep ∓ gs uncorrelated. As long as E0 < M0 (recall from the pre- vious subsection that E0 ∼ 1/∆, where ∆ is the splitting between the p orbitals with Jz = 1/2 and Jz = 3/2), this equation will always have solutions. However, as we saw in the previous subsection, these band crossings are not always Weyl points. Na3Bi k · p Hamiltonian The analysis proceeds similar to Cd3As2: in Na3Bi there are eight bands of interest, two s-orbitals with (J, Jz) = (1/2,±1/2) and six p-orbitals with (J, Jz) = (3/2,±3/2), (3/2,±1/2), (1/2,±1/2). Following Ref [2], in the presence of spin-orbit coupling, the states near- est to the Fermi level come from the heavy-hole p- states j, jz(cid:105) = 3/2,±3/2(cid:105) and the conduction s-states 1/2,±1/2(cid:105). Thus, the system is represented by two two- dimensional irreps. Since the bands are inverted (the p- orbitals disperse downwards and the s-orbitals upwards), they will intersect at some point in the Brillouin zone away from the Γ point. Below, we work in the basis or- dered by jz = 3/2, 1/2,−1/2,−3/2. We take the D6h symmetry group for NaBi, which has generators C6z =  0 −i 0 −(−1)5/6 0 0 0 0 0 0 0 0 (−1)5/6 0 i 0    0 0  0 C2x = 0 i 0 0 −i 0 0 −i 0 0 0 0 i 0 M100 = 0 −i 0 0 −i 0 0 0 0 0 −i 0 −i 0 0 3x + y (S58) The k·p Hamiltonian to order k3 derived from the sym- metries above has the same structure as that of Cd3As in Eq (S42), but with the matrix elements, z + M2k2(cid:107) 1 + A1k2 z + A2k2(cid:107) (cid:17) (cid:16) M(k) ≡ M0 + M1k2 A(k) ≡ Ak− B(k) ≡ − 1 Bk2−kz 2 0(k) ≡ C0 + C1k2 z + C2k2(cid:107), x + k2 y = k+k−. (S59) where k± = kx ± iky and k2(cid:107) = k2 actly and are doubly degenerate everywhere: The energies of this Hamiltonian can be solved for ex- E(k) = 0(k) ±(cid:112)M(k)2 + A(k)2 + B(k)2 ±(cid:112)−M0/M1, k(cid:107) = 0. Thus, it is crucial that M0 and M1 The doubly-degenerate bands cross each other when M(k) = A(k) = B(k) = 0. This occurs when kz = (S60) have opposite signs, as agrees with the fit in Eq (S61). (S54) Material parameters Ref [2] provides the fit to the coefficients of the quadratic terms: (S55) (S56) M0 = 0.08686 M1 = −10.6424 M2 = −10.361 C0 = −0.06382 C1 = 8.7536 C2 = −8.4008 A = 2.4598 (S61) where the energy is in units of eV when the momentum is in units A−1. Magnetic field The Zeeman term takes the form: HZ = −(gsJs + gpJp) · B (S62) where, if we exclude mixing with bands outside the k · p model, the Js and Jp matrices are given by Eq (S47). As in the previous section, the p-orbitals with J = 3/2, Jz = ±1/2 are not included in the k · p. We will include the mixing between these orbitals and the Jz = ±3/2 orbitals perturbatively below when needed. Field in the z direction When the magnetic field is in the z direction, the kz- axis remains a high-symmetry line and C6z preserves points on this line. From Eq (S54), it is evident that all bands have different eigenvalues under C6z; hence cross- ings between all bands are protected. As in Cd3As2, this can yield up to eight Weyl points (and a minimum of four.) The crossing between jz = ±1/2 and jz = ∓3/2 occurs z0 = (±Bz(gs + 3gp)/4 − M0) /M1; at (0, 0, kz0), where k2 the crossing point can be easily found, since H0(0, 0, kz), (Js)z and (Jp)z are all diagonal. The effective two-band Hamiltonian describing the band crossing is given by, 2 H 2 ,∓ 3 ± 1 eff − B(k2 = −2(M1kz0δz + (k2 x − k2 y)kz0σx ± 2Bkxkykz0σy, which describes a double Weyl point at kz0. x + k2 y)M2)σz (S63) The other crossings are single Weyl points, de- = z0 = 2 ,± 3 scribed by the two-band Hamiltonian: H −2δzkz0M1σz + A(±kxσx + kyσy), where here k2 (±Bz(gs − 3gp)/4 − M0) /M1. ± 1 eff 2 The kz = 0 plane is a high-symmetry plane protected by M001; since mirror symmetries have two distinct eigen- values, this protects two of the four possible crossings. These protected crossings yield line nodes. 17 (α + akx + bk2 y + ck2 Shifting kx by (bk2 Heff into the canonical form of a single Weyl point. z)σz + (dky + ekz)σx + (f ky + gkz)σy. y + ck2 z)/a and rotating σx and σy puts We now consider the kx = 0 plane, which is protected by M100 mirror symmetry. This symmetry can protect half of the possible band crossings; the protected cross- ings form two line nodes in the plane. In addition, along the kz line, there is a non-topological crossing; the two- band Hamiltonian near the crossing is identical in form to Eq (S53), which describes a band crossing in Cd3As2 in the same field configuration. The same analysis applies to a magnetic field in the y, a, b, c and d directions, defined in Eq (S58), because they are related by C6z symmetry. We now discuss the pertubative effects of the Zeeman term: the leading order term, of order B2/∆, contributes to the diagonal entries of the Hamiltonian, while the next order term, B3/∆2, mixes the p orbitals; ∆ is the energy gap to other p-orbitals outside the k · p. This is identical to the case in Cd3As2. Thus, when B is along the x-axis, HZ ∼ B2/∆I + B3/∆2σx, where the matrices act on the two p orbitals. These are the lowest order terms consis- tent with C2x symmetry. Similarly, when B is along the y-axis, HZ ∼ B2/∆I + B3/∆2σy. By including HZ, we have verified the claims above: when B is along a high- symmetry direction, there are a maximum of four Weyl points, which are all single Weyl points, and line nodes exist in the plane perpendicular to B. Band crossings along the kz axis for arbitrary B The logic at the end of the previous section regard- ing Cd3As2 applies equally well to Na3Bi: in particular, symmetry restricts the Hamiltonian to be diagonal along the kz axis and thus exactly solvable. As long as the p-orbitals with Jz = 1/2 are well-separated in energy from the p orbitals with Jz = 3/2, there will always be band crossings along the kz axis. However, these band crossings are not always Weyl points. Field in the x direction MAGNETOTRANSPORT When the magnetic field is in the x direction, the anal- ysis is similar: the kx-axis remains a high-symmetry line and C2x preserves points on this line. However, since C2x only has two distinct eigenvalues, it can only pre- serve half of the possible band crossings. Thus, there can be a maximum of four Weyl points. These must all be single Weyl points: we can understand this by writing an effective two-band Hamiltonian, Heff = di(kx, ky, kz)σi, which describes the band structure near the Weyl point. In this space, C2x can be represented by iσz (because it squares to −1 and mixes bands with different C2x eigen- values). Enforcing C2x symmetry, to lowest order, Heff = Semiclassical equations of motion In this section, we will derive the equations of motion governing semiclassical dynamics near a Weyl node. We work in units = c = e = 1 throughout. We assume that we have a Fermi surface composed of N disjoint pockets surrounding Weyl points, which may be magnetic-field induced or intrinsic. The conditions for this assumption to hold in our models of field-induced Weyls are given in Eqs. (8) -- (12) in the main text. We assume that scatter- ing between the different pockets is weak, so that to first approximation we can describe the low-energy behavior 18 of the system with the Hamiltonian (cid:77) H(k) = Hi(k) (S64) where i = 1, . . . , N indexes the different Weyl nodes, and i Hi(k) = uµ i (k − k0i)µ + (k − k0i)µ(Ai)µ ν σν. the Weyl Hamiltonian is diagonalized at every step. Be- cause of the path integral over momentum, this intro- duces the Berry connection into the classical action. Af- ter imposing the condition that classical particle trajecto- ries have positive energy, we find that the classical action is (cid:90) (S65) Scl = dt(k + A) · x − Φ − + − a · k (S74) Here µ = 1, 2, 3 indexes the spatial direction, k0i is the location of the i-th Weyl node in the Brillouin zone, and σν are the usual Pauli matrices. We assume that the Ai are either positive definite or negative definite, so that the model describes point nodes (as opposed to line nodes). Let us introduce new coordinates ν = (k − k0i)µ(Ai)µ ki ν (S66) in terms of which the linearized Hamiltonians Hi may be written where the Berry connection a satisfies ∇ × a = Ω (S75) By varying the classical action, we find for the equations of motion x = v + k × Ω k = E + x × B (S76) (S77) Hi(k) = uµ i ki µ + ki µσµ (S67) where E and B are the electric and magnetic field, and we have defined where we have introduced uµ i = uν i (A−1 i We assume that ui < 1 )µ ν (S68) (S69) so that all our Weyl points are Type I. The spectrum of each Hi is given by ± i = ui · ki ±(cid:12)(cid:12)(cid:12)ki(cid:12)(cid:12)(cid:12) . (S70) vµ = ∂+ ∂kµ = uµ + Aµ ν kν k We can solve the equations of motion to find v + E × Ω + (Ω · v)B E + v × B + (E · B)Ω 1 + B · Ω 1 + B · Ω x = k = Kinetic Equation (S78) (S79) where the plus sign corresponds to particle energies, and the minus sign corresponds to holes/antiparticles. The Berry curvature around each Weyl node is given by Ωi µ = det(Ai) kµ − k0i 2ki3 µ , (S71) and hence the monopole charge of each node is given by ci = sgn(det Ai) (S72) Using the equations of motion (S79), and assuming that scattering is sufficiently weak that every node can be considered independently, we can write a Boltzmann equation for the distribution function of electrons near each node. To do so, we must first identify a (time- evolution) invariant volume element with which to de- fine a conserved particle density[40 -- 42]. Let dV0 = dxdk/(2π)3 be the "standard" phase space volume el- ement. By computing the Jacobian determinant for the change of variables x(t) → x(t + δt), k(t) → k(t + δt) (S80) We can now derive the semiclassical equations of mo- tion near a Weyl node. For simplicity, we take N = 1 for now. Our strategy, following Refs. [39] and [40], will be to take the path integral representation of the propagator (cid:90) K = DxDkei(cid:82) dt(k+A)· x−Φ−H we find (cid:32) (cid:33) ∂ x ∂x + ∂ k ∂k (S73) dV0(t + dt) = 1 + dV0(t). (S81) for a single particle (i.e. positive energy) excitation, and then perform the stationary phase approximation, with background electromagnetic vector potential A and scalar potential Φ. For this to work, we must make sure With the aid of the equations of motion (S77) and (S79), this simplifies to dV0(t + dt) − dV0(t) = − dV0(t) 1 + B · Ω ∂(B · Ω) ∂t , (S82) which allows us to identify Anomalous conductivity of point nodes 19 dV = dV0(1 + B · Ω) (S83) as the time-independent volume element. Let ni(x, k, t) be the distribution function for particles near the ith node, defined with respect to this volume element. Then conservation of particle number implies, dni dt = ∂ni ∂t + x · ∂ni ∂x + ki · ∂ni ∂k = I i(ni). (S84) Here I i is a collision integral accounting for both inter- and intra-node scattering. The density of states ρi near each node, defined with respect to the invariant volume, dV , is given by (cid:90) ρi() = = + d3k (cid:18) (2π)3 δ(+(k) − )(1 + B · Ωi) (cid:33) det Ai(1 − ui2)2 −1ui) (ui − tanh 1 4π2 ci Bi · ui 22 u3 (S85) (S86) , (S87) i = (A−1 i where we have defined Bµ ν Bν. If we assume )µ that the distribution functions ni depend on the mo- mentum only through the energy (which is true in equi- librium, and is a good assumption when the intra-node scattering time is the shortest timescale in the problem; see Ref. [31] for more detailed discussion on this point) and that the electromagnetic field is space-independent, then we can isolate the effects of inter-node scattering by multiplying Eq. (S84) by d3k/(2π)3δ(+ − )(1 + B · Ωi) on each side and integrating. Doing so yields ∂ ∂ni() (cid:16) ci (cid:17) 4π2 E · B (cid:90) d3k (2π)3 (1 + B · Ω)I i(ni) = I i(), ∂ni() ∂t + 1 ρi() I i() = 1 ρi() where ρi 0 is given by Eq. (S86), and (S88) (S89) (cid:90) N (i) = + ∇ · j = i we find dN i dt rent as(cid:88) (cid:88) (cid:90) d3ki (cid:88) Recall that(cid:80) j = = ji i i (2π)3 (S92) d3k (2π)3 (1 + B · Ωi)ni(k) (cid:90) (cid:88) d3k (2π)3 (1 + B · Ωi)I i(k) (S93) i (cid:2)vi + E × Ωi + (Ωi · vi)B(cid:3) ni (S94) (S95) We will now use the distribution function Eq. (S91) to calculate the contribution of internode scattering to the conductivity. To do so, we must first back up and derive an expression for the current density. We can do so quite generally from Eq. (S84). We know from the density of states that the number of electrons near each node is given by Using this and the kinetic equation, and defining the cur- i ci = 0 by the Nielsen-Ninomiya theorem. We can substitute Eq. (S91) into Eq. (S95) to find the internode contribution to the current at zero tempera- ture. Using the zero-temperature result, yields ∂ni 0 ∂ = −δ( − µ), ji,µ = τ 16π4ρi(µ) BµBνEν. (S96) (S97) Summing over all nodes, we find for the anomalous con- tribution to the current jµ = σµν a Eν τ 16π4 BµBν(cid:88) σµν = i ρi(µ)−1 (S98) (S99) now contains only internode scattering. For weak intern- ode scattering, we may make the relaxation time approx- imation I i() = − 1 τ (ni() − ni 0()), (S90) where ni 0 is the equilibrium distribution function for node i, and τ = τ (B) is the internode scattering time, which for field-tuned Weyls may be a strong function of mag- netic field. In the weak-scattering regime we are consid- ering, it is computable in perturbation theory. We can solve Eq. (S88) in this approximation to linear order in the electric field to find 0() − τ ρi() (cid:16) ci (cid:17) 4π2 E · B ni() ≈ ni ∂ni 0() ∂ (S91) This is the generalization of the magnetoconductance from Ref. [30], although the magnetic field dependence is now significantly more complicated, due to the magnetic field dependence of ρi. This depends on the field explic- ity, and also implicitly through the field dependence of ui and (Ai)µ ν . For example, for the exactly solvable point nodes in GdPtBi with B(001) and B(111), we have ui ∝ B√ B . Oddly enough, in one of these cases B is parallel to ui as well (recall that this vector enters the density of states), though in another case the two are ac- tually perpendicular. Note also that, for intrinsic Weyl points, in a two-point measurement (j measured parallel to E), the measured current will depend on cos2 θ, where θ is the angle between E and B, while for field-created Weyls, higher powers of cos θ can appear. Simple model for field-induced Weyls Applying the above formalism, we consider the sim- plest two-band model for field induced Weyls. We take for the Hamiltonian (ignoring temporarily the orbital magnetic field) H = k2 zσz + kxσx + kyσy − B · σ (S100) At zero field, the spectrum of this Hamiltonian has a quadratic two-band touching at k = 0. As the Zeeman field is increased, two Weyl nodes develop at Bz). The linearized Hamiltonians H± k0± = (Bx, By,±√ around each of these points are H± = δkT±A±σ  1 0 0 √ 0 0 1 0 0 ±2  , B (S101) (S102) A± = where δk± = k − k0±. We can now apply our analysis from the previous subsections to calculate the anoma- lous conductivity near each of these Weyl nodes. (This restores the orbital field, at least in the semiclassical ap- proximation). Using Eq. (S86), and the fact that u = 0, we find that (cid:112) σµν a = τ (B) 2π2µ2 BzBµBν (S103) Note the additional field dependence of the current, stem- ming from the density of states. Furthermore, at fixed density the chemical potential will also be B-dependent. Due to the simplicity of this model, we can also solve for the conductivity in the ultraquantum limit, at least for B = (0, 0, Bz). We can then choose a gauge in which the Hamiltonian (including the orbital field) is transla- tionally invariant along the z direction. In this case, the most convenient representation for the Hamiltonian is (cid:32) (cid:33) H = z − Bz k2 b b† Bz − k2 z , (S104) where b is the two-dimensional Landau-level lowering op- erator, corresponding to the Landau levels in the x − y plane. The energies of this Hamiltonian are given by z − Bz)2 + 2Bzn, n = 1, 2, . . . ± = ±(cid:112)(k2 0 = k2 z − Bz 20 (S105) (S106) Each of these states is BzAxs/2π - fold degenerate, com- ing from the degeneracy of the 2d Landau levels (Axs is the cross-sectional area in the x − y plane). Note that around each Weyl point, the zeroth Landau level is chiral with velocity √ = ±2 B B v± z = ∂0 ∂kz If we apply an external electric field Ez to this system, it will shift all of the k(cid:48) zs as a function of time. If we as- sume, however, that there is a mechanism for internodal scattering with relaxation time τ , then in the long-time limit, kz of the occupied states will be shifted at each node by the finite amount (S107) (cid:12)(cid:12)(cid:12)(cid:12)kz=±√ ∆kz = τ Ez. (S108) This results in a depletion of right-moving states from the negative chirality node and an increase in left-moving states at the positive chirality node. Counting the states involved in this process[33], we find a net current density jz = vzBz 2π2 ∆kz = z τ B3/2 π2 Ez (S109) Note that the scaling of Bz here is one power lower than in the semiclassical case (consistent with Ref [30]). Comparison to experiment As noted in the main text, the applicability of our the- ory to the existing experiments in GdPtBi is limited by the separation of the Weyl points in momentum space. Nonetheless, in order to push our theory as far as it can go, we now show that the experimental data in Ref 18 is better described by including higher powers of mag- netic field in the magnetoconductance than by the semi- classical theory of Ref 30. Below we plot magnetocon- ductance as a function of B2 using the data from Fig 1d in Ref 18. For temperatures below 50K, the magnetocon- ductance has positive curvature for small B, indicating that it scales like a greater power than B2. At higher fields, the magnetoconductance approaches the B2 scal- ing. Thus, the low-field data goes beyond the theory of Ref 30. 21 FIG. S4. Magnetoconductance as a function of B2 for data from Fig 1d of Ref 18. The curvature at low fields and low temperatures shows that σxx ∼ Bp, where p > 2. At higher fields, σxx ∼ B2.
1704.06417	1	1704	2017-04-21T07:03:19	Quantum dynamics of a domain wall in the presence of dephasing	[ "cond-mat.mes-hall", "quant-ph" ]	We compare quantum dynamics in the presence of Markovian dephasing for a particle hopping on a chain and for an Ising domain wall whose motion leaves behind a string of flipped spins. Exact solutions show that on an infinite chain, the transport responses of the models are nearly identical. However, on finite-length chains, the broadening of discrete spectral lines is much more noticeable in the case of a domain wall.	cond-mat.mes-hall	cond-mat	25th Int. Symp. "Nanostructures: Physics and Technology" Saint Petersburg, Russia, June 26 -- July 1, 2017 c(cid:13) 2017 St Petersburg Academic University Quantum dynamics of a domain wall in the presence of dephasing Claudio Castelnovo1, Mark I. Dykman2, Vadim N. Smelyanskiy3, Roderich Moessner4, Leonid P. Pryadko5 1 T.C.M. Group, Cavendish Laboratory, University of Cambridge, J. J. Thomson Avenue, Cambridge CB3 0HE, U.K. 2 Department of Physics and Astronomy, Michigan State University, East Lansing, Michigan 48824, USA 3 Google Inc., Venice, California 90291, USA 4 Max-Planck-Institut fur Physik komplexer Systeme, 01187 Dresden, Germany 5 Department of Physics & Astronomy, University of California, Riverside, California 92521, USA Abstract. We compare quantum dynamics in the presence of Markovian dephasing for a particle hopping on a chain and for an Ising domain wall whose motion leaves behind a string of flipped spins. Exact solutions show that on an infinite chain, the transport responses of the models are nearly identical. However, on finite-length chains, the broadening of discrete spectral lines is much more noticeable in the case of a domain wall. the case of a domain wall (DW), s = γ (1 − δs,0), Γ(P) Γ(DW) s = γ s. In Eq. (1), H = H0 with the matrix elements (H0)ab = − ∆ 2 (δa,b+1 + δa+1,b) (2) (3) is the usual hopping Hamiltonian, δa,b is the Kronecker sym- bol, and the parameters ∆ and γ respectively denote the half band width and the dephasing rate. s Formally, Eqs. (1) with Γ(P) of Eq. (2) can be considered a Lindblad equation[6] for particle hopping, where each site has its own bath coupled to its occupation number. It de- scribes universal long-time dephasing physics valid in the limit where both the bath cutoff frequency ωc (maximum frequency of a bath mode) and the bath temperature β−1 are large compared to the hopping band width 2∆. s Similarly, with Γ(DW) of Eq. (2), these equations describe dynamics in a single-DW sector of an Ising spin chain in the presence of the transverse field ∆, and independent fluctu- ating longitudinal magnetic fields. Moving the domain wall by a − b positions requires flipping a − b spins, which increases the dephasing rate for the matrix element ρab. 2. Results On an infinite chain, we use the translational symmetry and define the Fourier transformation for the center of mass, ρab = eiKReiπs/2φs(K, t), s ≡ a − b, (4) where R ≡ (a + b)/2 and the phase factor eiπs/2 makes explicit the reflection symmetry, s → −s. The densities φs(K, t) satisfy a 1D Schrodinger equation with hopping uK = ∆ sin(K/2) and an imaginary on-site potential −iΓs. (i) With K = 0, densities φs(0, t) at different s are inde- pendent from each other and, for s (cid:54)= 0, decay to zero with rates Γs. It is also easy to see that stationary solutions of Eq. (1) with the diagonal density ns ≡ ρss in the form of a polynomial of degree (cid:96) ≥ 0 have all off-diagonal matrix elements with a− b > (cid:96) zero. In particular, with ns linear, the stationary solution ρab is tri-diagonal, which makes the linear diffusive transport of particles and DWs identical. (ii) We obtained explicit solutions for Laplace-transformed densities ψs ≡ ψs(K, p), with the initial condition φs(K, t = (cid:90) dK 2π 7 1 0 2 r p A 1 2 ] l l a h - s e m . t a m - d n o c [ 1 v 7 1 4 6 0 . 4 0 7 1 : v i X r a Introduction The role of dephasing on the time evolution of a quantum mechanical system is a fundamental issue in the study of open quantum systems. We pose the question what happens if the object mov- ing around is not a simple pointlike particle, but rather an emergent quasiparticle which acts as a source of an observ- able emergent gauge field. One example is the monopole and Dirac string dynamics[1, 2] in spin ice[3]. As the full motion in a disordered spin background is beyond the scope of a first pass at this problem, we consider a simplified setting where the motion takes place in one dimension; we contrast the cases of a free particle and one with a string attached, in the form of a domain wall in an Ising system. Experi- mentally, this question corresponds, e.g., to the observation of a Villain mode in a quasi -- one-dimensional magnet[4, 5]. Our central results are the following. The first is a tech- nical one, namely that we can solve both cases (particle and domain wall motion) for a one-dimensional model subject to a locally uncorrelated Markovian dephasing bath. Secondly, this solution demonstrates that, for unstructured motion in one dimension, the two cases differ only weakly, in the sense that the difference between the two is considerably smaller than the difference between either and the fully coherent time evolution. In particular, linear transport responses in the presence of a linear density gradient or a uniform field are identical for the two cases. Thirdly, we notice that this is no longer the case when considering finite-length chains. Here, the discrete energy spectrum is broadened consider- ably more strongly for the case of domain walls; this can be qualitatively understood as the enhanced fragility of the interference of a domain wall with itself as it does a round trip on the finite lattice to establish the standing wave. 1. Models The two models we solve both describe single-body one- dimensional hopping in the presence of Markovian dephas- ing uncorrelated between the sites, written in terms of the density matrix with components ρab, ρab = −i[H, ρ]ab − Γa−b ρab, (no summation!) (1) where dephasing rates for all off-diagonal elements of the density matrix are equal in the case of a particle (P), while they grow linearly with the distance from the diagonal in 1 2 0) = δs,0 of a classical purely-diagonal density matrix: e−isπ/2(cid:2)y(cid:0) p+γ (cid:1)(cid:3)s K]1/2 − γ , 2uK ψ(P) s = ψ(DW) s [(p + γ)2 + 4u2 = e−isπ/2 Ip/γ+s(z) p/γ(z) γz I(cid:48) , y(t) ≡ (1 + t2)1/2 − t, z ≡ 2uK/γ, where Iν(z) is the modified Bessel function of the first kind, and I(cid:48) ν(z) = Iν+1(z) + (ν/z)Iν(z) is its derivative. At s = 0, these quantities are related to the dynamic structure factor S(K, ω) = ψ0(K, + iω), given by the spatial and tempo- ral Fourier transform of the probability P (s, t) to travel s sites in time t. While functional forms differ for the two cases, plots of S(K, ω) at γ/∆ < 1/10 are remarkably sim- ilar (not shown due to space constraint). Also similar are the probabilities P (s, t), see Fig. 1. Fig. 1. Time dependence of the probabilities P (s, t), s = 0, 1, 2, for a particle (red solid lines) and for a DW (blue dashed lines) with ∆ = 2 and dephasing γ = 0.2. Thin black lines show the corresponding results at γ = 0, P (s, t) = J 2 s (∆ t), where Js(z) is the Bessel function of order s. Even for such a relatively large γ, there is little difference between a particle and a DW. (iii) In spite of these similarities, finite-frequency re- sponses on finite chains look different in the two cases. In- deed, a bound state is formed when a wave function inter- feres with itself; such an interference requires off-diagonal matrix elements of ρ. Solving linearized versions of Eqs. (1) with a harmonically modulated linear potential, and assum- ing the unperturbed thermal density matrix e−βH0 , we an- alyzed the average conductance χ(ω) [Fig. 2]. At small γ, the corresponding real part χ(cid:48)(ω) has a series of resonant peaks at ωmn = Em − En, where symmetry requires m − n to be odd, and Em = −∆ cos km, km = πm/(L + 1), are the energy levels of the chain (3) of length L, m = 1, 2, . . . , L. In the case of a particle on such a chain, we find the width of each peak to be γ(L − 1)/(L + 1), while for a DW on a chain with L allowed positions, peak widths scale as γ O(L), as would be expected on general grounds. 3. Conclusions Quantum mechanics teaches us that particles can behave as waves, and waves as particles. For weakly-interacting parti- cles, this is described by second quantization. A superficially similar correspondence also exists outside of the perturba- tive sector where a topological defect can be often viewed as a particle, its motion described by the Schrodinger equation. Some of the examples are dislocations in lattices, vortices in Fig. 2. Real part of the frequency-dependent susceptibility, χ(cid:48)(ω), for a particle (red solid lines) and a DW (blue dashed lines) on chains of length L = 7 and L = 15 as indicated, with half band width ∆ = 2 and dephasing γ = 0.01. Vertical grid lines mark allowed transitions between the discrete energy levels in the absence of dephasing (only one level pair is shown for each line). While discrete lines for a particle are well resolved in both cases, they are suppressed entirely for a DW on the longer chain. 2D superfluids or superconductors, and various soliton-like defects in 1D systems. Experimentally observed quantum manifestations of such objects include position uncertainty and quantum delocalization of lattice defects, quantum tun- neling of vortices and magnetic domain walls, and quantum transport in conducting polymers. Our main conclusion is that this analogy between collec- tive excitations and particles is not universal. Environment can severely limit the quantum behavior of such excitations. Acknowledgements This work was supported in part by the ARO grant W911NF- 14-1-0272, the NSF grant PHY-1416578, and EPSRC grants EP/K028960/1 and EP/M007065/1. References [1] L. D. C. Jaubert and P. C. W. Holdsworth, Nature Physics 5, 258 (2009). [2] Y. Wan and O. Tchernyshyov, Phys. Rev. Lett. 108, 247210 (2012), 1201.5314. [3] C. Castelnovo, R. Moessner, and S. Sondhi, Annual Re- view of Condensed Matter Physics 3(1), 35 (2012). [4] J. Villain, Physica B+C 79(1), 1 (1975). [5] S. E. Nagler, W. J. L. Buyers, R. L. Armstrong, and B. Briat, Phys. Rev. Lett. 49, 590 (1982). [6] G. Lindblad, Commun. Math. Phys. 48, 119 (1976). 0.00.51.01.52.02.53.00.00.20.40.60.81.0tP(s,t)L=25γ=0.2s=0s=1s=20.00.51.01.50.00.10.20.30.40.51-22-33-41-4ωχ′(ω),a.u.L=7γ=0.01T=1000.00.51.01.50.000.050.100.150.200.250.301-22-33-44-55-66-77-81-42-53-64-75-86-91-62-73-84-95-10ωχ′(ω),a.u.L=15γ=0.01T=100
1612.02871	1	1612	2016-12-08T23:36:53	Topological Phonon Modes in A Two-Dimensional Wigner Crystal	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	We investigate the spin-orbit coupling effect in a two-dimensional Wigner crystal. We show that sufficiently strong spin-orbit coupling and an appropriate sign of g-factor could transform the Wigner crystal to a topological phonon system. We demonstrate the existence of chiral phonon edge modes in finite size samples, as well as the robustness of the modes in the topological phase. We explore the possibility of realizing the topological phonon system in two-dimensional Wigner crystals confined in semiconductor quantum wells/heterostructure. We find that the spin-orbit coupling is too weak for driving a topological phase transition in these systems. We argue that one may look for the topological phonon system in correlated Wigner crystals with emergent effective spin-orbit coupling.	cond-mat.mes-hall	cond-mat	Topological Phonon Modes in A Two-Dimensional Wigner Crystal Wencheng Ji1 and Junren Shi1, 2, ∗ 1International Center for Quantum Materials, Peking University, Beijing 100871, China 2Collaborative Innovation Center of Quantum Matter, Beijing 100871, China We investigate the spin-orbit coupling effect in a two-dimensional Wigner crystal. We show that sufficiently strong spin-orbit coupling and an appropriate sign of g-factor could transform the Wigner crystal to a topological phonon system. We demonstrate the existence of chiral phonon edge modes in finite size samples, as well as the robustness of the modes in the topological phase. We explore the possibility of realizing the topological phonon system in two-dimensional Wigner crystals confined in semiconductor quantum wells/heterostructure. We find that the spin-orbit coupling is too weak for driving a topological phase transition in these systems. We argue that one may look for the topological phonon system in correlated Wigner crystals with emergent effective spin-orbit coupling. PACS numbers: 63.20.-e, 73.21.-b The concept of topology attracts great interest in re- cent investigations of condensed matter physics. The quantum Hall effect demonstrates a new state of mat- ter which carries robust chiral edge modes protected by non-trivial topology of electron states. Recent research reveals various kind of topological insulators dictated by electron state topology in the presence of the time rever- sal symmetry or other symmetries [1 -- 4]. Up to now, the studies are mainly focused on topological effects of the charge degree of freedom of electrons. On the other hand, the interest is recently extended to topological effects as- sociated with collective excitations such as phonons and magnons [5 -- 10]. Particularly, a topological phonon sys- tem (TPS) is predicted to possess robust chiral vibra- tional (phonon) modes at its edge [5]. These topological phonon modes (TPMs) are believed to be contributing to quantized thermal hall conductivity at low tempera- ture [8]. It is even postulated that TPMs are ubiqui- tous in biological systems and essential for a variety of processes in living organisms [5]. While these theoreti- cal considerations are interesting and intriguing, we have yet to know an experimentally accessible and controllable way to construct a TPS in laboratory. In this paper, we explore the possibility of realizing a TPS in a Wigner crystal (WC) of electrons. We in- vestigate the effect of spin-orbit coupling (SOC) in a two-dimensional WC [11]. We show that sufficiently strong SOC and an appropriate sign of g-factor could transform the WC to a TPS. We demonstrate the ex- istence of chiral edge phonon modes in finite size sam- ples, as well as the robustness of the chiral edge modes in the topological phase. We check the possibility in two-dimensional WCs confined in semiconductor quan- tum wells/heterostructure [12]. We find that the SOC is too weak for driving a topological phase transition in these systems. We argue that one may look for the TPS in correlated WCs with emergent effective SOC. It is well known that electrons form a WC in suffi- ciently low density when the Coulomb interaction domi- nates over the kinetic energy. Recent state-of-art Monte Carlo calculations show that two-dimensional electrons undergo a series of phase transitions when lowering den- sity, ultimately stabilized to a WC phase of a triangular lattice with ferromagnetic order of electron spins when rs > 38 [13 -- 15], where rs ≡ 1/√πneaB is a dimension- less density parameter with electron density ne and ef- fective Bohr radius aB. Experimentally, two-dimensional WCs had been observed in ultra-clean 2D samples of GaAs/AlGaAs heterostructure [16]. We investigate phonon modes of a 2D WC of electrons. To model such a system, we consider a set of electrons with an effective mass m∗ arranged in a triangular lattice vibrating near their equilibrium positions. The Hamilto- nian can be written as: (cid:88) l g∗µB σlzB, (1) (cid:88) l H = P (l)2 2m∗ + Φ[{uα(l)}] + where the first term is the kinetic energy, the second term is the potential energy due to the Coulomb interaction between electrons, and the third term is the Zeeman en- ergy due to a uniform magnetic field B perpendicular to the 2D plane, with µB being Bohr magneton, g∗ being effective Land factor for the specific material hosting the 2D system, and σlz being the z-component of the Pauli spin matrices σl for an electron at site l. For an electron subjected to both the magnetic field B and SOC, the kinetic momentum should be written as, P (l) = −i∇u(l) + 1 2 u(l) × eB + α0 σl × F (l), (2) which involves two extra terms besides the canonical mo- mentum operator, consisting of a vector potential term induced by the perpendicular magnetic field B and a SOC term induced by the electrostatic force F (l) acting on the electron by all other electrons in the system, re- spectively, and u(l) is the displacement of an electron rel- ative to its equilibrium position. The strength of the SOC is determined by a coefficient α0, which is a material- specific parameter [17]. Because all electrons are local- ized near their equilibrium positions in a WC, it is suffi- cient to make a harmonic approximation to the potential energy: Φ[{uα(l)}] ≈ Φ0 + 1 2 Φl−l(cid:48) αβ uα(l)uβ(l(cid:48)), (3) (cid:88) ll(cid:48),α,β where α,β = x,y, and the detailed form of the dynamic coefficients Φl−l(cid:48) for a triangular lattice is presented in Ref. [18]. Correspondingly, the force F (l), which dictates the SOC, is related to the potential energy by: αβ Fα(l) = −∇uα(l)Φ = − Φl−l(cid:48) αβ uβ(l(cid:48)). (4) (cid:88) l(cid:48),β To proceed, we need to make a further approximation. Equation (1) defines a system involving both phonons (vibration modes) and magnons (spin waves). However, since the WC has ferromagnetic order of spins and a per- pendicular magnetic field will align and quench the spin degree of freedom, we can replace the Pauli spin matrices in the Hamiltonian with their mean field expectation val- ues: (cid:104) σl(cid:105) = (0,0,σ) with σ = −sgn(g∗B), because the di- rection of the electron spins will be parallel (anti-parallel) to the external magnetic field for g∗ < 0 (g∗ > 0). We can then obtain equations of motion: (cid:104) (cid:88) Pα(l) uα(l) = m∗ Pα(l) = − , l(cid:48)β Φl−l(cid:48) αβ uβ(l(cid:48)) + Gl−l(cid:48)αβ (cid:105) Pβ(l(cid:48)) , where αβ is antisymmetric tensor with xy = 1, and (cid:104) (cid:105) Gl−l(cid:48) =ωcδl,l(cid:48) − α0σ m∗ Φl−l(cid:48) yy + Φl−l(cid:48) xx (5) (6) (7) where ωc = eB/m∗ is the magnetic cyclotron frequency of electrons. A Fourier transformation recasts the equations of mo- tion to an eigenvalue equation for phonon modes in a system breaking time-reversal symmetry [8]: (cid:20) (cid:21) 0 i −iΦ(k) G(k)σ2 ψn(k) = ωn(k)ψn(k), (8) where k is defined in the Brillouin zone for the triangular lattice, ψn(k) = (u(k),P (k))T , σ2 is the second Pauli matrix, and Φ(k), G(k), u(k), and P (k) are Fourier transformations of Φ(l−l(cid:48))/m∗, Gl−l(cid:48), m∗u(l), and P (l), respectively. We have: G(k) = ωc − α0(Φxx(k) + Φyy(k))σ. (9) We note that Eq. (8) will give rise to four modes for each k with two positive and two negative frequency branches. However, the negative frequency branches can be related to the positive frequency branches by 2 ω−n (k) = −ω+ show the positive frequency branches. n (−k) [8]. In the following, we will only Figure 1 shows the evolution of phonon dispersion along high symmetry lines of the Brillouin zone for differ- ent strengths of SOC in the presence of a magnetic field. We observe that the uniform magnetic field induces gaps between the two branches of phonon modes at both Γ and K (K(cid:48)) points of the Brillouin zone, as shown in Fig. 1(a). In increasing the strength of SOC, the gap at K (K(cid:48)) point is gradually closed and reopened, while the gap at Γ-point is unaffected, as shown in Fig. 1(b) -- (d). Such behavior is similar to what happens in a topologi- cal insulator, in which a closing and a reopening of a gap indicates a band inversion which transforms a normal in- sulator to a topological insulator. We thus expect that the evolution observed here may also drive a topological phase transition, albeit for the phonon bands. FIG. 1. Phonon dispersion along high symmetry lines of the Brillouin zone for different strengths of SOC. We set ωc = 0.1ω0. The strengths of SOC are: (a) α0 = 0; (b) α0ω0 = 0.04; (c) α0ω0 = 0.05; (d) α0ω0 = 0.1. The effective Land factor is assumed to be negative: g < 0. To confirm the topological change of the phonon bands, we calculate the phonon Berry curvature that is defined as: Ωn(k) = −Im ∂ψn(k) ∂k , (10) (cid:21) (cid:20) ∂ ¯ψn(k) (cid:20) Φ(k) ∂k × (cid:21) 0 I2×2 0 [8]. Figure 2 shows where ¯ψn(k) ≡ ψ†n(k) the distribution of the Berry curvatures for the upper phonon bands before and after the band inversion. It can be clearly seen that the Berry curvature peaks at Γ and K (K(cid:48)). For weak SOC, the Berry curvatures at Γ and K (K(cid:48)) have opposite signs, resulting in cancellation of the Chern number that is proportional to an integration of the Berry curvature over the whole Brillouin zone. On the other hand, the Berry curvatures at Γ and K (K(cid:48)) have the same signs after the band inversion, indicating 00.20.40.60.811.2 ω/ω0ΓKMΓ(a)00.20.40.60.811.2 ω/ω0ΓKMΓ(cid:26)(b)00.20.40.60.811.2 ω/ω0ΓKMΓ(c)00.20.40.60.811.2 ω/ω0ΓKMΓ(d)3 FIG. 3. (color online) Phase diagram for the phonon bands of a two-dimensional ferromagnetic WC. The regions with non- zero Chern-number is topologically nontrivial. modes at each of the edges for both the phases. How- ever, only one of the modes survives in the gap regime near K (K(cid:48)) point. The topological phase transition is accompanied by a change of the propagating direction of the surviving edge mode. The topological difference becomes apparent when we increase the strength of the magnetic field to open a full gap in the phonon spectrum (c, d). In this case, the edge modes completely disappear in the gap regime for the topologically trivial phase, while for the topologically non-trivial phase, there are two chiral edge modes prop- agating along the same direction. The two chiral edge modes are consistent to the Chern number C = −2. One can also clearly see that both the edge modes emerge from Γ point, and end near K (K(cid:48)) point. The topological dif- ference between the two phases lies in the different ways that the edge modes connect the bulk phonon bands. For the topological trivial phase, the edges modes connect the same phonon band, while for the topological non-trivial phase, the edge modes make inter-band connections. We explore the possibility of realizing the TPS in WCs of real materials. We first check WCs formed in two-dimensional electron gases confined in semiconduc- tor quantum wells/heterostructure. Some of the host- ing semiconductors, such as AlSb, AlxIn1−xAs, InP, and ZnSe, can be ruled out because their parameters α0 and g have the same sign. On the other hand, other semicon- ductors such as GaAs, AlAs, InAs, InSb do have opposite signs for α0 and g. Table I shows relevant parameters for them. Unfortunately, we find that the SOC is too weak for all of these materials. For a WC stabilized purely by the electron-electron interaction, SOC induced gaps ∆K are of the order of 10−7 meV, and the magnetic field must be weaker than 10−7 T for a TPS. This is appar- ently impossible for real world experimental conditions. We also explore the possibility in hole systems. In this case, the SOC has a different form due to the band symmetry [17]. the cou- pling between the momentum and spin is proportional to [px(l)Ey(l) + py(l)Ex(l)]σ, instead of [px(l)Ey(l) − py(l)Ex(l)]σ for an electron system. It gives rise to a dif- As a result, FIG. 2. (color online) Distribution of the Berry curvatures for the upper phonon bands before and after the band inver- sion. The parameters are the same as those in Fig. 1(b) and Fig. 1(d), respectively. the topological change of the phonon band. We calculate the Chern numbers for both cases, and obtain C = 0 and C = −2 for bands before and after the inversion, respectively. The behavior can be easily understood. To see that, we recast the eigenvalue equation (8) to an equivalent 2 × 2 hermitian eigenvalue problem: h(k)u(k) ≡ (Φ(k) + ωG(k)σ2)u(k) = ω2u(k). (11) 0 (12) Near the K point, we can expand h(k) ≈ h0 +λ1(cid:52)kyσ1− λ2(cid:52)kxσ3 + M σ2, where h0, λ1, λ2 are constants, and M = ωG(K) = ω(cid:0)ωc − 2α0σω2 (cid:112) (cid:1), s Φxx(K) is the phonon frequency at K where ω0 = point in the absence of both the magnetic field and SOC, and is related to material parameters by ω0 = 1.967r−3/2 Ry∗, where Ry∗ is the effective Rydberg of the hosting semiconductor [17]. It is easy to see that the effective hamiltonian has the same form as the 2D massive Dirac hamiltonian with the mass M . The in- version of the bands and the topological phase transition occur when M changes sign. This is possible only when g∗α0 < 0. In this case, M changes sign when 2α0ω0 = −sgn(g∗) ωc ω0 . (13) On the other hand, the Berry curvature near the Γ point is only determined by the external magnetic field and is not affected by the SOC, because Φαβ(Γ) = 0 and G(Γ) = ωc. Figure 3 shows the corresponding phase diagram. For a TPS, there exist TPMs in a finite-size sample. To show that, we consider a strip of WC along the x di- rection and calculate phonon dispersion as a function of kx. Figure 4 shows the edge states for a few representa- tive cases. We find that there exist edge states for both the topologically trivial phase (a, c) and the topologically non-trivial phase (b, d). The difference between the two phases is obscured when the magnetic field is not strong enough for opening a full gap in the phonon spectrum (a, b). In this case, we can find two counter-propagating −101 −0.500.5(a)−1−0.50 −1−0.8−0.6−0.4−0.2(b)sgn(g⇤)↵0!20−1−0.500.5100.250.5C=0 C=+2 C=+2 C=-2 !c/!04 tron density with rs ∼ 1, we find that (see Table I) for GaAs, ∆K ∼ 12µeV and Bc ∼ 7mT. Other semicon- ductors have parameters in similar orders of magnitude. They are still too small to provide observable physical effects. All summarized, we conclude that WCs in real semi- conductor materials cannot support a TPS. This is not surprising because as a relativistic effect, the SOC is al- ways weak. From Table I, we see that the strength of the SOC must be enhanced at least four orders of mag- nitude to reach α0ω0 ∼ 1, a magnitude necessary for a clear manifestation of the topological effect. This is un- fortunately impossible in real world. We argue that one may look for the TPS in WCs with emergent effective SOC. Actually, SOC is only one of many possible forms of coupling between orbital motion and internal degrees of freedom. The strengths of other forms of the coupling are not necessarily constrained by the relativistic principle, and could be potentially very strong. An interesting and potentially relevant case could be found in the fractional quantum Hall systems, in which a new species of WCs, i.e., Wigner crystals of composite fermions, may form [22]. In these systems, the orbital motion of electrons is strongly entangled with degrees of freedom of all other electrons in the system due to the strong-correlation nature of the state. One would expect that the entanglement serves as effective SOC, and gives rise to similar effects as those predicted in this paper. The effective SOC emerges from the strong correlation, and its strength is not constrained by the relativistic prin- ciple. This possibility will be left for further investiga- tions. This work is supported by National Basic Re- search Program of China (973 Program) Grant No. 2015CB921101 and National Science Foundation of China Grant No. 11325416. ∗ [email protected] [1] C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005). [2] C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 146802 (2005). [3] B. A. Bernevig, T. L. Hughes, and S. C. Zhang, Science 314, 1757 (2006). [4] M. Knig, S. Wiedmann, C. Brne, A. Roth, H. Buhmann, L. Molenkamp, X.-L. Qi, and S.-C. Zhang, Science 318, 766 (2007). [5] E. Prodan and C. Prodan, Phys. Rev. Lett. 103, 248101 (2009). [6] L. Zhang, J. Ren, J.-S. Wang, and B. Li, Phys. Rev. Lett. 105, 225901 (2010). [7] N. Berg, K. Joel, M. Koolyk, E. Prodan, Phys. Rev. E 83, 021913 (2011). [8] T. Qin, J. Zhou, and J. Shi, Phys. Rev. B 86, 104305 (2012). FIG. 4. (color online) Phonon dispersion of a finite size sam- ple with different values of αω0 and ωc/ω0. The finite size sample is extended along the x direction, and has a width of 100 sites in the y-direction. The red (blue) dots represent modes localized at the top (bottom) edge of the sample. The size of a dot indicates the degree of localization that is propor- iψn(i)4, a larger value of which tional to a quantity en =(cid:80) corresponds to a more localized edge mode. g∗ Ry∗ (meV) α0/ (eV−1) GaAs 0.0455 AlAs + 20.157 −0.00479 InAs InSb 1.461 0.590 0.352 0.955 5.882 - - - ωrs=38 (meV) 0 K ∆rs=38 (neV) Brs=38 c (µT) 0.221 0.274 0.106 0.127 0.0494 0.356 0.169 0.0123 0.0210 0.00495 0.0469 0.00563 41 TABLE I. Parameters calculated for a number of semiconduc- tors. Material parameters are adopted from Ref. [17], with α0 ≡ (m∗/e)r6c6c . Ry∗ is the effective Rydberg for the material, ∆K ≡ 2α0ω2 0 is the gap induced by SOC at K- point, and Bc is the critical strength of the magnetic field for the topological phase transition. The values of ω0, ∆K and Bc at rs = 38 are shown, as indicated by the superscripts. The values of these quantities at other density can be deter- mined by: ω0 = ωrs=38 (38/rs)3, Bc = Brs=38 (38/rs)3/2, ∆K = ∆rs=38 0 (38/rs)3. K c ferent G(k) = eB/m + α0(Dyy(k) − Dxx(k))σ, in which the SOC contribution vanishes at K point. Therefore, the SOC in a hole system cannot drive a topological phase transition of phonon. We also explore the possibility in WCs stabilized by a strong magnetic field, which quenches the kinetic energy of electrons and favors the formation of WCs [19, 20]. These WCs could be stabilized in quantum Hall systems at much higher electron densities [21]. For a typical elec- −10100.51kxa/πω/ω0ωc/ω0=0.3;αω0=−0.1(a)−10100.51kxa/πω/ω0ωc/ω0=0.1;αω0=−0.1(b)−10100.51kxa/πω/ω0ωc/ω0=1;αω0=−0.3(c)−10100.511.5kxa/πω/ω0ωc/ω0=1;αω0=−1(d)5 [9] C. Strohm, G. L. J. A. Rikken, and P. Wyder, Phys. Rev. Rev Lett 82, 1744 (1999). Lett. 95, 155901 (2005). [10] A. V. Inyushkin and A. N. Taldenkov, JETP Lett. 86, 379 (2007). [17] R. Winkler, Spin-Orbit Coupling Effects Dimensional Electron and Hole Systems 2010). in Two- (Springer, [11] Yu. P. Monarkha and V. E. Syvokon, Low Temperature [18] L. Bonsall and A. A. Maradudin, Phys. Rev. B 15, 1959 Physics 38, 1067 (2012). (1979). [12] T. Ando, A. B. Fowler, and F. Stern, Rev. Mod. Phys. [19] H. Fukuyama, P. M. Platzman, and P. W. Anderson, 54, 437 (1982). [13] B. Tanatar and D. M. Ceperley, Phys. Rev. B 39, 5005 (1989). [14] C. Attaccalite, S. Moroni, P. Gori-Giorgi, and G. B. Bachelet, Phys. Rev. Lett. 88, 256601 (2002). Phys. Rev. B 19, 5211 (1979). [20] K. Maki and X. Zotos, Phys. Rev. B 28, 4349 (1983). [21] H. Zhu, Y. P. Chen, P. Jiang, L. W. Engel, D. C. Tsui, L. N. Pfeiffer, and K. W. West, Phys. Rev. Lett. 105, 126803 (2010). [15] N. D. Drummond and R. J. Needs, Phys. Rev. Lett. 102, [22] A. C. Archer, K. Park, and J. K. Jain, Phys. Rev. Lett. 126402 (2009). 111, 146804 (2013). [16] J. Yoon, C. Li, D. Shahar, D. C. Tsui, M. Shayegan, Phys
1208.0238	1	1208	2012-08-01T14:31:01	Gold assisted molecular beam epitaxy of Ge nanostructures on Ge(100) Surface	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	We report on the gold assisted epitaxial growth of Ge nanostructures under ultra high vacuum (UHV) conditions (\approx3\times 10-10 mbar) on clean Ge (100) surfaces. For this study, \approx2.0 nm thick Au samples were grown on the substrate surface by molecular beam epitaxy (MBE). Thermal annealing was carried out inside the UHV chamber at temperature \approx500{\deg}C and following this, well ordered gold nanostructures placed on pedestal Ge were formed. A \approx 2 nm Ge film was further deposited on the above surface while the substrate was kept at a temperature of \approx 500{\deg}C. The height of the Ge (pedestal) underneath gold increased along with the formation of small Ge islands. The effect of substrate temperature and role of gold on the formation of above structures has been discussed in detail. Electron microscopy (TEM, SEM) studies were carried out to determine the structure of Au - Ge nano systems.	cond-mat.mes-hall	cond-mat	Gold assisted molecular beam epitaxy of Ge nanostructures on Ge (100) Surface A. Rath1, J. K. Dash1, R. R. Juluri1, A. Ghosh1 and P. V. Satyam1,* 1 Inst itute of Phys ics, Sachivalaya Marg, Bhubaneswar - 751005, India Abstract We report on the go ld assisted epitaxial growth of Ge nanostructures under ultra high vacuum (UHV) condit ions (≈3× 10-10 mbar) on clean Ge (100) surfaces. For this study, ≈2.0 nm thick Au samples were grown on the substrate surface by molecular beam epitaxy (MBE). Thermal annealing was carried out inside the UHV chamber at temperature ≈ 500°C and fo llowing this, well ordered gold nanostructures placed on pedestal Ge were fo rmed. A ≈ 2 nm Ge film was further deposited on the above surface while the substrate was kept at a temperature of ≈ 500°C. The height of the Ge (pedestal) underneath gold increased along with the format ion of small Ge islands. The effect of substrate temperature and ro le o f go ld on the format ion o f above structures has been discussed in detail. E lectron microscopy (TEM, SEM) studies were carried out to determine the structure of Au – Ge nano systems. PACS: 81.15. Hi, 68.37.-d, 64.75.St, 68.55.-a, 81.15.Aa Keyword: Au-Ge nanostructures, Electron Microscopy, MBE, UHV annealing * Corresponding Author: [email protected]. in, [email protected] 1 1. Introduction Epitaxy o f thin films is a key process in modern nano electronic industries. It allows abrupt doping profiles in homoepitaxy and the format ion of sharp interfaces in heteroepitaxy to be realized. More than 90 % o f semiconductor devices consist of silicon. Knowledge of surface properties of Si, Ge and Au nanostructures are crucial for rapidly developing area of nanotechno logy. Successfu l implementation of germanium techno logy will however require an understanding of the so lid-state interact ion in metal-germanium systems 1 -3. As the feature size becomes smaller the transistor density increases and with it the heat production. This is a problem that can be compensated by using lower driving vo ltages. To allow this, a higher electron and ho le mobility, relat ing the drift current to the applied electric field, is required from the semiconductor. For this, Ge is the right material to achieve this goal because it has high mobility for both electrons and ho les 4. Also due to the low vapor pressure, Au-Ge alloys are used as catalyst for the growth of Ge nano wires inside the UHV chamber 5. The effect of size on phase stability and phase transformat ions is of both fundamental and applied interest. For example, during the nucleat ion and growth of self-assembled nanowires from nanoscale metal catalysts, the phase o f the catalyst determines properties such as the growth rate and the structure of the nanowire. Any size-dependent or growth-rate dependent changes in the catalyst may thus have strong effects on the structures that formed 6–11. In last seven years, interest ing works have been carried out in understanding growth of silicon and germanium nanowire in presence o f Au as catalyst and deviat ion in phase diagram o f Au-Si and Au-Ge systems by using in-situ TEM methods 6 – 11. Epitaxially grown self organized nanostructures on silicon have been studied in great detail 12-15. Several works has also been done using the Ge substrate instead of Si 16. Data concerning the behavior of metal thin films on germanium upon thermal treatment is 2 relat ively scarce. The thin film react ions of 20 transit ion metals, excluding gold, with germanium substrates have been reported by Gaudet et al 17. The main focus o f this work is on so lid source molecular beam epitaxy o f both Au and Ge on Ge (100) substrate. It is similar to the vapor-liquid-solid (VLS) growth of Ge with Au as transport medium, except mo lecular-beam source is used instead of a chemical compound vapor like CVD growth. This avo ids many problems, including the high temperature required to decompose the vapor compound and possible contaminat ion o f the growing layer. Also, the deposit ion will be atom by atom which ensures uniform growth of the metal and the semiconductor material at the first stage of the process. Hence, adatom deposit ion is not selective and growth is driven by adatom diffusion on the substrate. The so lid so lubility o f Ge in Au is 3.1 %, but that of Au in Ge is very low (<10-5 %) 18. Systemat ic results reported by Sutter et al. clearly point out the deviat ion o f Ge solubility from bulk for nano-scale systems (typically less than 50 nm size) 19. The systems dealt by Sutter et al., are basically for VLS type growth (free standing out-of-plane NW growth) and the results showed that the phase diagram would deviate from the bulk 19. Cheuh et al reported on the post growth engineering o f the nanowire (NW) structures and composit ion through the alloying and phase segregation that is induced by thermal annealing in Au-Ge system 20. A direct observat ion o f the VLS growth of Ge nanowires was reported by Wu and Yang 21, who ident ified the various growth stages in correlat ion to the Au–Ge binary phase diagram. In our earlier work, we demonstrated that, gold-silicide nano alloy format ion at the substrate (Si) surface is necessary for forming phase separated Au-Ge bilobed nanostructures. Our results also indicate that Si-Ge bonding is more favorable to have than having Au-Ge bond. It would be very interest ing to study the structural changes in the above Au-Ge nanostructures, when Si substrate will be replaced by Ge substrate under the same ambient 22. 3 In this art icle, we present the experimental observation o f growth of Au-Ge nano-structures during in-situ thermal treatment inside UHV chamber for go ld deposited on Ge(100) surfaces without oxide layer at the interface fo llowed by further deposit ion o f Ge on it. Ex-situ electron microscopy measurements (both SEM and TEM) confirm the presence o f Au and Ge in these structures. The format ion o f Au-Ge nanostructures via nanoscale phase separat ion is demonstrated experimentally. Also, the role of substrate temperature in format ion o f these structures has been clearly demonstrated in this work. 2. Experimental For this present study, we have prepared five types of samples. For the first case (sample A), a thin Au film o f thickness o f about ≈2.0 nm on n-type Ge(100) , by MBE method under ultra high vacuum condit ions 23. Ge (100) substrates were loaded into the MBE chamber and degassed at ≈ 400°C for about 12 hours inside the chamber and fo llowed by flashing for about 1 minute by direct heat ing at a temperature of ≈800°C. In this process, nat ive oxide was removed and a clean Ge(100) surface was obtained. On such ultra clean surfaces, ≈2.0 nm thick go ld films were grown epitaxially by evaporating Au from a Knudsen cell. Deposition rate was kept constant at ≈ 0.14 nm min-1. Thermal annealing o f the as-deposited sample (sample A) was carried out inside the UHV chamber at temperature of 500°C with a ramp rate of 7°C min-1(sample B). Also, we deposited a ≈ 2.0 nm gold on Ge(100) sample in the MBE chamber as explained above (like sample A) and about 2.0 nm Ge was further deposited at a typical deposit ion rate of 0.6 ML/min (where one ML corresponds to 6.78 × 1014 atoms cm–2 for a Si(100) surface) with various substrate temperatures: 500°C (sample C) and 600°C(sample D). It should be noted that Au deposit ion and annealing and Ge deposit ion after Au deposition are done in sequent ially without any break in the vacuum. In addit ion to this, we deposited 2.0 nm Ge on Ge((100) substrate fo llowed by annealing upto 500°C (sample E). During 4 the growth, chamber vacuum was ≈ 6.2 ×10-10 mbar. The post-growth characterization of the above samples were done with the field emission gun based scanning electron microscopy (FEGSEM) measurements with 20 kV electrons using a Neon 40 cross-beam system (M/S Carl Zeiss GmbH). Cross-sectional TEM specimens were prepared from the above samples in which electron transparency was achieved through low energy Ar+ ion milling. TEM measurements were performed with 200 keV electrons (2010, JEOL HRTEM) under cross-sectional geometries. 3. Results and discussions Fig. 1(a) depicts a SEM micrograph of as deposited 2.0 nm Au Film on Ge(100) using MBE (sample A) with surface coverage o f ~21%. Irregular and iso lated gold nanostructures of typica l size ~ 27 nm were formed. Fig.1 (b) shows the bright field XTEM image o f the interface o f go ld film and the Ge substrate and corresponding high resolut ion image(HR) shows the d-spacing o f Au(200) and Ge(111) plane (inset Fig.). The as-deposited sample was annealed in-situ inside the UHV chamber at 500°C for 30 minutes (sample B). Following this annealing in UHV, ex-situ SEM measurements (at room temperature) showed well distributed gold nanostructures placed on pedestal Ge (Fig. 2 (a)). Fig. 2(b) and 2(c) depicts the XTEM image o f the Au nanostructures and cross-sectional HRTEM image of single Au nanostructures respectively. The contrast shows the inter diffusion of go ld into the Ge substrate. In our earlier work 22, we had shown that for Au-S i system, go ld nano rectangles were formed without having pedestal Si. The reason behind the format ion is st ill not well understood. When a ≈ 2 nm Ge was further deposited (sample C) using MBE system on the above Au- patterned surface at substrate temperature 500°C inside the UHV chamber, Au nanostructures with more prominent Ge pedestal were formed (Fig. 3 (a)). This means that Au-Ge nanalloys are acted as 5 seeding posit ions for the growth of above structures. Because o f the incorporation o f larger amount of Ge in to the seed particle, height of the pedestal increases. Kodambaka et al. have shown using in situ microscopy for the Ge/Au system that catalysts can be either liquid or solid below eutectic temperature, depending on thermal history. They have shown that with a pre-annealing at 400 °C, the Ge/Au catalyst remains liquid down to 255°C 24, this phenomenon has also been reported very recent ly by Gamalski et al. 25. This means that with our growth conditions the catalyst should be still liquid. Sample was then exposed to molecular beam of Ge. In Fig. 3(a), SEM image taken at 54° tilt clearly shows the three dimensional nature of the Au-Ge nanostructures, where gold is on the top of the Ge. The bright contrast is the go ld and other one is for the Ge. As Au is higher Z material, more secondary electrons are emitted from the Au region than from Ge region and this causes the contrast difference between gold and germanium in SEM images. In addit ion to these structures, one can see the format ion o f Ge nanostructures having typical height ~ 5.0 nm (Fig. 3(b)). In Fig. 3(b), bright field XTEM images of the Au-Ge nanostructures along with Ge nano islands has been shown and corresponding high resolut ion XTEM of the epitaxially grown Ge nano structures is shown in the inset. Fig. 3(c) depicted HRXTEM micrograph o f one of the Au-Ge structures. The typical height of the pedestal Ge is ~ 8.0 nm. Furthermore, one can see that the height of the pedestal Ge is several t imes larger than the effect ive thickness of the deposited Ge. The filtered image o f the square marked region is demonstrated in Fig. 3(d). It shows the format ion of Au-Ge composit ion at the interface o f top gold and pedestal Ge. The d-spacing has been found to be of about 0.332 nm which do not match with the Au or Ge but matches quite well with the 26. (224) plane of tetragonal phase of Au0.6Ge0.4 To do the ex-situ temperature dependent study, we deposited 2.0 nm Ge on Au patterned surface at 600°C (sample D) instead o f 500°C. Formation of the above Au-Ge structures having 6 typical size o f the pedestal ~ 20 nm has been observed (Fig. 4(a)). It should be noted that, there is no format ion o f small Ge islands in this case and the height of the pedestal is more than 500°C case. In Fig. 4(b), one can clearly see the only formation of Au-Ge nanostructures. At higher enough temperature (600°C), atoms are arriving into the seeding particle not only from the mo lecular beam but also from the Ge is lands due to self diffusion. This resulted in increase in height of the pedestal Ge. Fig. 4(c) depicted the HRXTEM micrograph of one of such structures. The filtered image o f the squared marked region is shown in Fig. 4(d). The d-spacing shows the format ion of Au0.6Ge0.4 26 composit ion at the junction of top gold and pedestal Ge. In both the cases, the Au-Ge composit ion was found at the interface. It has been studied that the composit ion is not only depend on the annealing temperature but also on the cooling rate 16. Thus, there will be some composit ion gradient at the interface. From the above studies, it was found that the heated Au-Ge alloy structures may play a major role in formation o f such Au-Ge structures. During the deposit ion o f Ge, gold germanide particles act as nucleation center for the growth of Ge. As the deposit ion cont inues, the smaller liquid droplets provide a rapid diffusion path for the incoming Ge atoms, which find it energetically favorable to join already-nucleated crystals rather than to nucleate new ones. As substrate is also Ge and the Ge-Ge bonding is more favorable than Au-Ge bonding, Ge pedestal were formed underneath go ld. Another interesting aspect of this study is the format ion of Ge islands. In sample C, well distributed Ge nano islands were formed along with Au-Ge structures. To study the effect of gold over layers on format ion o f islands, 2 nm Ge was deposited on cleaned Ge (100) surface fo llowed by annealing at 500°C (sample E). Formation o f Ge nano islands were not seen (Fig. 5). In Fig. 5, the HRXTEM shows the homo epitaxial growth of Ge layer on the Ge(100) surface without islanding. The dotted white line is the sharp interface between the Ge layer and the Ge substrate. Thus presence of go ld layer resulted in format ion of Ge islands. 7 Numerous interest ing aspects of Au-Ge nanostructures format ion can st ill be studied. For instance, it would be both fascinat ing and beneficial to study the real t ime measurements of phase separation in Au-Ge system. Better understanding of growth of Au-Ge nanostructures via phase separation could be achieved by performing more in-situ experiments. This observat ion o f phase separation at nanoscale would be very useful for proper understanding o f go ld contacts on Si-Ge based devices. 4. Conclusions We have reported the out of plane growth of epitaxial Au-Ge nanostructures by using mo lecular beam epitaxy. The interest ing part of this work is to get the growth parameters in UHV condit ion to enhance the in plane material transport which leads to format ion of out of plane Au-Ge nanostructures. The effect of substrate temperature, on the format ion o f above structures was studied in detail. References: 1 Y. F. Hsieh, L. J. Chen, E. D. Marshall and S. S. Lau, Appl. Phys. Lett. 51, 1588(1987). 2 O. Thomas, S. Delage, F. M. d’Heurle and G. Scilla, Appl. Phys. Lett. 54, 228(1989). 3 S. Q. Hong, C. M. Comrie, S. W. Russell and J. W. Mayer, Appl. Phys. 70, 3655(1991). 4 J. J. Rosenberg, IEEE Trans. Electron. Devices ED-30, 1602 (1983). 5 H. Adhikari, A. F. Marshall, I. A. Goldthorpe, C.E. D. Chidsey and P. C. McIntyre ACS Nano, 1, 415(2007) 6 R. Kohler, W. Neumann, M. Schmidbauer, M. Hanke, D. Grigoriev, P. Schäfer, H. Kirmse, I. Häusler, and R. Schneider, NanoScience and Techno logy,Semiconductor Nanostructures ( D. Bimberg (ed.) chapter 5) Springer, 2008, 97-121, DOI: 10.1007/978-3-540-77899-8_5 7 E. A. Sutter and P. W. Sutter, ACS Nano 4, 4943 (2010). 8 8 W. Neumann, H. Kirmse, I. Häusler, I. Otto and I. Hähnert, J. Alloys and Compounds 382, 2 (2004). 9 I. Goldfarb, Phys. Rev.Lett. 95, 025501 (2005) 10 B. Kim, J. Tersoff, S. Kodambaka, M.C. Reuter and F. M. Ross, Science 322, 1070(2008). 11 E. Sutter and P. Sutter, Nanotechno logy 22, 295605(2011). 12 A.Shibata, surface science, 303, 161 (1994) . 13 A. Rath, J. K. Dash, R. R. Juluri, A. Rosenauer and P.V. Satyam, J. Phys D: Appl Phys, 44,115301 (2011) 14 S. H. Brongersma et al. Phys. Rev. Letts. 80, 3795–3798 (1998) 15 M. Mundschau, E. Bauer, W. Telieps Surface science 213 381-392 (1989). 16 S Hajjar et al, Phys. Rev. B, 84, 125325 (2011) and references there in. 17 S. Gaudet, C. Detavernier, A.J. Kellock, P. Desjardins, C. Lavo ie, J. Vac. Sci. Techno l. A 24(3), 474 (2006). 18 G. V. Samsonov and V. N. Bondarev, Germanides, Published by consultants Bureau, New York, 52 (1969). 19 E. A. Sutter and P. W. Sutter, ACS Nano 4, 4943 (2010). 20 Y. L. Chueh et al, Nano Lett. 10, 393(2010). 21 Y. Wu, P. Yang, J. Am. Chem. Soc., 123, 3165(2001) 22 A. Rath, J. K. Dash, R. R. Juluri, M. Schowalter, K. Mueller, A. Rosenauer and P. V. Satyam, J. Appl. Phys. 111, 104319 (2012) 23 D. K. Goswami, B. Satpati, P. V. Satyam and B. N. Dev, Curr. Sci. 84, 903(2003). 24 S. Kodambaka, J. Tersoff, M.C. Reuter, F.M. Ross, Science 316, 729 (2007). 25 A. D. Gamalski, J. Tersoff, R. Sharma, C. Ducati and S. Hofmann, Nano Lett. 10, 2972 (2010). 26 Au0.6Ge0.4 , JCPDS 18-0551 9 Figure captions: Fig 1: For as deposited MBE sample (sample A), (a) SEM micrograph showing go ld nanostructures with typical size o f about 27 nm (b) XTEM image showing the interface of Au thin film and Ge substrate and corresponding high resolut ion image (inset Fig.) shows d-spacing o f Au(200) and Ge(111). FIG 2: (a) SEM image taken at 54º tilt for the sample that was annealed at 500ºC in UHV chamber (sample B) (b) corresponding bright field XTEM image o f the islands and (c) high resolut ion XTEM image of one of the nanostructure (sample B) showing inter-diffusion o f Au into Ge substrate. FIG 3: (a) SEM micrograph taken at RT after 2.0 nm Ge deposited on Au patterned substrate at 500ºC (sample C), (b) corresponding XTEM image shows the Au -Ge nanostructures along with the Ge nano islands (the High reso lut ion o f the square marked region is shown in the inset), (c) HRXTEM image o f one of the Au-Ge islands depicts the pedestal Ge underneath the Au and (d) filtered image o f the square marked region shows the format ion o f Au-Ge composit ion at the interface. FIG 4: (a) SEM micrograph taken at RT after 2.0 nm Ge deposited on Au patterned substrate at 600ºC (sample D), (b) corresponding XTEM image shows the Au-Ge nanostructures, (c) HRXTEM image of one of the Au-Ge islands depicts the pedestal Ge underneath the Au and (d) filtered image of the square marked region shows the format ion of Au-Ge composit ion at the interface. FIG 5: High reso lution XTEM of the sample taken at RT after 2nm Ge was deposited on Ge (100) fo llowed by UHV annealing at 500ºC (sample E), shows the homo epitaxial growth of Ge layer on Ge (100) 10 FIG 1 : Rath et al FIG 2 : Rath et al 11 FIG 3 : Rath et al 12 FIG 4 : Rath et al FIG 5 : Rath et al 13
1903.00400	1	1903	2019-03-01T16:46:42	Indirect observation of phase conjugate magnons from non-degenerate four-wave mixing	[ "cond-mat.mes-hall" ]	A phase conjugate mirror utilising four-wave mixing in a magnetic system is experimentally realised for the first time. Indirect evidence of continuous-wave phase conjugation has been observed experimentally and is supported by simulations. The experiment utilizes a pump-probe method to excite a four-wave mixing process. Two antennae are used to pump a region of a thin-film yttrium iron garnet waveguide with magnons of frequency $f_{1}$ to create a spatio-temporally periodic potential. As the probe magnons of $f_{\mathrm{p}}$ impinge on the pumped region, a signal with frequency $f_{\mathrm{c}} = 2f_{1}-f_{\mathrm{p}}$ is observed. The amplitude of the nonlinear signal was highly dependent on the applied magnetic field $H$. Width modes of the probe magnons and standing wave modes of the pump magnons were shown to affect the amplitude of the signal at $f_{\mathrm{c}}$. Experimental data is compared with simulations and theory to suggest that $f_{\mathrm{c}}$ is a phase conjugate of $f_{\mathrm{p}}$.	cond-mat.mes-hall	cond-mat	Noname manuscript No. (will be inserted by the editor) Indirect observation of phase conjugate magnons from non-degenerate four-wave mixing Alistair Inglis · Calvin J. Tock · John F. Gregg March 4, 2019 Abstract A phase conjugate mirror utilising four-wave mixing in a magnetic system is experimentally realised for the first time. Indirect evidence of continuous-wave phase conjugation has been observed experimentally and is supported by simulations. The experiment uti- lizes a pump-probe method to excite a four-wave mixing process. Two antennae are used to pump a region of a thin-film yttrium iron garnet waveguide with magnons of frequency f1 to create a spatio-temporally periodic potential. As the probe magnons of fp impinge on the pumped region, a signal with frequency fc = 2f1 − fp is observed. The amplitude of the nonlinear signal was highly dependent on the applied magnetic field H. Width modes of the probe magnons and standing wave modes of the pump magnons were shown to affect the amplitude of the signal at fc. Experimental data is com- pared with simulations and theory to suggest that fc is a phase conjugate of fp. Keywords Magnons · Spin Waves · Phase Conjugate · Four-wave mixing Alistair Inglis Department of Physics, University of Oxford, Clarendon Labo- ratory, Parks Road, OX1 3PU ORCID: 0000-0001-6371-6047 E-mail: [email protected] Calvin J. Tock Department of Physics, University of Oxford, Clarendon Labo- ratory, Parks Road, OX1 3PU ORCID: 0000-0001-5277-3839 John F. Gregg Department of Physics, University of Oxford, Clarendon Labo- ratory, Parks Road, OX1 3PU 1 Introduction The insatiable appetite for smaller, more powerful com- puting devices is leading to the inevitable breakdown of Moore's Law [1]. A considerable deviation from Den- nard's scaling is already underway [2] opening the door for alternative computational paradigms. In recent years there has been increasing interest in magnon- based computing, or magnonics, [3 -- 9] as a solution to particular problems [10] facing the future of con- ventional complementary metal-oxide-semiconductor (CMOS) computing. An advantage of wave-computing not restricted to magnonics is the ability to encode information in two variables: amplitude and phase. The ability to perform operations on the phase of a spin wave is therefore of fundamental importance to the future of magnon- ics. One such useful operation is phase conjugation; a process that exactly reverses the propagation direction and phase factor for every plane wave in an arbitrary wave front [11]. The result is the creation of a phase conjugate mirror (PCM) that reflects any beam along the same path by which it arrived at the mirror, ir- respective of incident angle. This remarkable property can lead to aberration correction of waves after pass- ing through a nonuniform distorting medium; a process with many useful applications including image process- ing, encryption and spectroscopy [12 -- 14]. First experimentally realised in the 1970s, phase conjugate mirrors are a well established phenomenon in the optical community. A preferred method for cre- ating a PCM is by way of the third-order nonlinear process: four-wave mixing (FWM)[15]. Experiments of this nature have a general form wherein the confluence of two 'pump' beams and a 'probe' beam in a region of nonlinear medium causes the appearance of a fourth 9 1 0 2 r a M 1 ] l l a h - s e m . t a m - d n o c [ 1 v 0 0 4 0 0 . 3 0 9 1 : v i X r a 2 Alistair Inglis et al. beam which is the phase conjugate of the probe signal [16]. Until now, phase conjugation of spin waves has only been achieved using methods of parametric pumping and second-order nonlinear processes [17 -- 19]. In this work we report on the creation of a PCM in a magnon waveguide. Our experiments differ from previous work for three reasons: 1) We achieve a phase conjugate with the process of FWM, rather than parametric pumping. 2) Our experiments are performed in the continuous- wave (CW) regime, in contrast to the pulsed experi- ments that utilise spin wave bullets. 3) Our experiments fully utilise the isotropic dispersion of the spin waves by exciting pump and probe magnons perpendicularly. In general, for FWM to occur certain physical con- ditions must be met. Energy conservation dictates that ω1 + ω2 = ωp + ωc, (1) where ω1,2 are the angular frequencies of the two pump waves, ωp is the probe wave angular frequency, and ωc is the conjugate wave angular frequency. Simi- larly, the condition k1 + k2 = kp + kc, (2) arises from momentum conservation, where k is a magnon wave vector and the subscripts follow the same convention as the angular frequencies. These constraints are met by wave vectors depicted by Fig. 1(b). It is clear that if k1 and k2 serve as counter- propagating pumps of same frequency in an isotropic medium and kp serves as the probe then by equation (2), the condition kp = −kc must be satisfied. a perturbation expansion of the magnetisation, it has been shown that there exists a driving term for a third- order spin wave that depends on the product of m3 [20, 21], where m is a component of the transverse mag- netisation. To understand the origin of the expected phase con- jugate magnon, consider the m3 term in the location of the pumped region with the probe magnons also present. This term may be expanded into its constituent parts. Following a mathematics analogous to the deriva- tion of optical phase conjugation [22], we expand m3 and express it as a real quantity: (cid:16)m1ei(ω1t−k1 ·r) + m∗ 1e−i(ω1t−k1 ·r)+ m2ei(ω2t−k2 ·r)+m∗ 2e−i(ω2t−k2 ·r)+ (3) mpei(ωpt−kp ·r) + m∗ pe−i(ωpt−kp ·r)(cid:17)3 . Here m1,2 and mp represent the amplitude of the transverse magnetisation of the pumps and probe re- spectively. Since mi is a complex amplitude, it also con- tains the phase information. Upon expansion of equa- tion (3), we obtain 56 cross terms. Terms with phase factors that have combinations of ω and k that are for- bidden by the dispersion relation may be neglected. Of the remaining terms, there is one of particular signifi- cance: m1m2m∗ =m1m2m∗ pei([ω1+ω2−ωp]t−[k1+k2−kp]·r) + c.c. pei(ωct−kc ·r) + c.c.. (4) There are a number of points to note about this p and is therefore term. Firstly, it is proportional to m∗ the phase conjugate of mp. Secondly, this is the only possible term possessing a wavevector antiparallel to the original probe beam [22, 23]. These two properties define the resulting spin wave as a phase conjugate re- flection. Finally, we note that this term is proportional 1 assuming the pumps are of equal magnitude, that to m2 is m1 = m2. Fig. 1 (a) Experimental configuration: yttrium iron garnet (YIG) film with 45◦ edges placed on PCB with antennae. Counter-propagating pumps excite a standing wave which cre- ates a periodic potential. The probe antenna transmits excita- tion magnons and receives reflections from the pumped region. (b) Illustration of the conservation of momentum condition for four-wave mixing where k1 and k2 are the pumps, kp is the probe, and kc is the phase conjugate We now consider FWM in a nonlinear magnetic sys- tem. By examining the Landau-Lifshitz equation with 2 Experiment For the magnon waveguide we utilised a yttrium iron garnet (YIG) film of thickness 7.8 µm on a gallium gadolinium garnet (GGG) substrate, 2.1 mm wide and 18 mm long, with corners cut at 45◦ to minimise reflec- tions. A schematic of the set-up is shown in Fig. 1(a). The waveguide was mounted on a printed circuit board (PCB) with three antennae. The antennae were in the formation of a meander structure to suppress the cou- pling to the ferromagnetic resonance (FMR) mode. The Indirect observation of phase conjugate magnons from non-degenerate four-wave mixing 3 ) m B d ( P -20 -40 -60 -80 -20 -40 -60 ) m B d ( P -80 ) s l e b i c e d ( P 0 -5 -10 -15 -20 3.9 3.906 3.912 3.918 3.924 3.93 Frequency (GHz) Fig. 3 Measured and calculated frequency spectra. (a) Spec- trum measured with no magnetic field applied. Both peaks are artefacts of the experimental electronics. (b) Spectrum mea- sured with applied magnetic field of 3077 Oe. (c) Spectrum from simulated pump-probe system at 3077 Oe. Wide peak at fp is due to limited computational resolution. Both measured and calculated spectra show signals at expected phase conju- gate frequency fc 3 Results and Discussion Figure 3(a-b) shows the measured spectra for two dif- ferent field configurations. When there is no external field applied as in (a), the only signals measured are the input signals. The large signal at fp is due to the impedance mismatch between the microwave transmis- sion line and the probe antenna causing electrical reflec- tions to be detected by the spectrum analyser, while the peak at f1 is due to direct coupling from the pump an- tennae to the probe antenna. As the magnetic field was increased to 3077 Oe the spectrum 3(b) was observed. Of note is the signal at the expected phase conjugate fre- quency fc = 2f1 − fp. This is not inconsistent with the notion that at the correct field strength, phase conju- gate reflections are occurring from the pumped region. Also present in the spectrum is a term at frequency f = 2fp − f1 which is due to a third-order process resulting from reflections of probe magnons from the waveguide edge farthest from the probe antenna. Fig. 2 Schematic of experimental set-up. Magnetic field is per- pendicular to the plane of the YIG waveguide antennae were 4 mm long comprising three legs 40 µm wide, spaced 50 µm apart. The two pump antennae were parallel to the long edge of the waveguide. The probe antenna was placed 3 mm from the pumps, across the width as shown in Fig. 1(a). The probe antenna is used for both transmission and detection of magnons inter- acting with the pumped region, while the pump anten- nae excite counter-propagating spin waves that exploit the strong intrinsic nonlinearity of the magnon system to generate a periodic mesoscopic texture. In our experiment, spin waves travelling in both the x and y directions were utilised. For this reason an isotropic magnon dispersion was required. Such a condition is offered by exciting forward volume magne- tostatic spin waves (FVMSWs) which have a uniquely isotropic dispersion, compared to the highly anisop- tropic dispersion relations for other magnetostatic spin wave modes [24]. An electromagnet was used to apply the external magnetic field perpendicular to the plane of the film, a field geometry that is necessary for the excitation of FVMSWs. The pump antennae were excited at f1 = f2 = 3.915 GHz by a Hewlett Packard HP8672A microwave source. A second source (HP8671A) was used to excite the probe antenna at fp = 3.91825 GHz. This detun- ing ∆f = 3.25 MHz was required to use the spectrum analyser to discriminate between input and output sig- nals. Introducing ∆f and ensuring Equations (1) & (2) are satisfied leads to a small phase mismatch ∆k which manifests itself experimentally as a reduced effi- ciency [11, 23]. The circulator shown in Fig. 2 allows the probe to act as a transmitter-receiver antenna. Re- flections from the pumped region will propagate back towards the probe antenna and be detected by the spec- trum analyser (ZHL Rhode & Schwarz). The magnetic field was measured with a Hall probe which was con- nected to a data acquisition computer. 4 Alistair Inglis et al. ) m B d ( P -70 -75 -80 -85 3040 3050 3060 3070 H (Oe) 3080 3090 3100 Fig. 5 Magnetic field dependence of the fc = 3.91175 GHz signal returning to the probe antenna. Fast oscillation shows the dependence on the standing wave mode of the pump. The lower amplitude at H1 = 3051 Oe compared to H2 = 3082 Oe is attributed to width modes, simulations of which are shown in Fig. 6 4 Simulations To aid with the interpretation of our results, simulations of the experiment were carried out using MuMax3 [26], a micromagnetic simulation software package. Physi- cal parameters were set such that saturation magneti- sation, Msat = 197 kA/m, exchange stiffness, Aex = 3.5 × 10−12J/m, and Gilbert damping, α = 5 × 10−5 [27]. As in the experiment, the probe antenna was 3 mm from the pumps, while a separate detection region was defined 1 mm from the pumped region. The pumps were driven at 3.915 GHz and the probe driven at 3.91825 GHz with each field configuration simulated for 2 µs be- fore performing a Fourier transform to investigate the frequency response of each system. The secondary effects caused by reflections of spin waves from the short ends of the waveguide were con- trolled for by increasing the value for α in these regions by factor of 300. The regions of waveguide edge directly in contact with the pump antennae were also assigned the increased damping value. By suppressing these edge- reflections, any output signals may be attributed purely to the nonlinear interaction of the excited pump and probe magnons. Figure 6 shows amplitudes of pump and probe magnons for different magnetic fields. The white regions represent the simulated waveguides, while the red and blue represent the amplitude of the x component of the transverse magnetisation, m. In the simulations, the probe magnons are launched from an antenna placed at the leftmost edge of the waveguide. At every point along the waveguide, the magnons excited at H1 = 3051 Oe are weaker than those excited at H2 = 3082 Oe. It is clear from the many more nodes across the width of the waveguide that (a) shows a higher width mode than in (c), and that its propagation efficiency is reduced. This phenomenon manifests itself experimentally as highlighted in Fig. 5, Fig. 4 Output of probe antenna as a function of magnetic field and frequency. Dashed yellow line corresponds to spectrum in Fig. 3(b). Orange boxed region highlights signal at fc. The magnetic field dependence of boxed region is shown in more detail in Fig. 5 Investigating further, a sweep of the applied mag- netic field was performed, the results of which are shown in Fig. 4. The dashed yellow line marks the field at which the spectrum in Fig. 3(b) was measured. The strong lines at fp and f1 appear to have minimal field dependence since the field dependent contribution to the signal is small compared to the electrical response described above. More interesting however, is the white line at fc = 3.91175 GHz showing an obvious field dependence. The nature of the dependence of the boxed region can be seen in more detail in Fig. 5. The amplitude reaches the noise floor at a field of approximately 3093 Oe. This is due to the excited FVMSWs approaching the FMR above which no more spin waves are excited. Below this field however, the signal oscillates as a function of field, with a period of approximately 5 Oe. This oscillation corresponds to different standing wave modes across the width of the pumped region. That is, when there are an integer number of half-wavelengths across the width of the waveguide, a standing wave is present and the pump amplitude is larger. This in turn, amplifies the phase conjugate signal, since it scales with m1m2 as shown in equation (4). In addition to this fast oscillation, Fig. 5 also shows a field dependence on a larger scale. This may be ex- plained by considering width modes [25] of the probe magnons. At approximately 3040 Oe, the width modes excited by the probe antenna destructively interfere, in- hibiting the propagation of the probe spin wave, thus diminishing the interaction with the pumped region. Indirect observation of phase conjugate magnons from non-degenerate four-wave mixing 5 fp - probe magnons f1 - pump magnons H 1 (a) (c) H 2 (b) (d) mx 0 -mx l d e i f d e i l p p a Fig. 6 Simulations. Examples of mx amplitudes for applied magnetic field H = 3051 Oe and 3082 Oe. The images isolate the different amplitudes of magnons with fp = 3.91825 GHz and f1 = 3.915 GHz. Subfigures (a) and (c) compare how transmission efficiency of probe magnons is affected by specific width modes. Both(b) and (d) show a standing wave created by the pump antennae. The standing wave at field H1 has 18 nodes, compared to the 12 nodes at H2 where a reduced phase conjugate signal is evident at H1 compared to H2. was due to a third-order nonlinear effect caused by re- flections from the waveguide edge. The simulation also illuminated the pump be- haviour. Figures 6(b) and (d) both show a standing spin wave across the width of the waveguide. At lower field, there are 18 nodes across the width, compared to the 12 nodes present at H2. As the field increases from H1 to H2, the intensity of the standing wave will oscillate with every node that is removed. Given equation (4) this phenomenon explains the 5 Oe oscillation in Fig. 5, with the simulation matching the experiment well. Indeed, increasing from H1 to H2 the calculation shows a difference of 6 nodes, while for the same measured fields, the intensity goes through 6 oscillations. A time-domain Fourier transform was performed on the simulated data. A typical example is shown in Fig. 3(c) where the applied field is 3077 Oe as it was for the data measured in (b). The large peak at fp is due to the detection region being placed between the probe and pump antennae, therefore picking up the original probe signal. The large width of this peak is an artefact of the limited computational resolution. Also of note is the comparatively small power of the pump frequency at f1 which is due to leakage of pump magnons, which are generally well confined between the antennae as seen in Fig. 6(b) and (d). We also observe a small bump at 3.925 GHz which arises from a higher order mixing term. As expected, there is a significant peak at fc. Be- cause the simulated system elminates reflections from the ends of the waveguide, this signal must necessarily be reflecting from the pumped region of periodic po- tential. Since this signal was generated by FWM, and has angular frequency ωc = ω1 + ω2 − ωp, it must also have wavevector kc in order to satisfy equation (4) con- firming that it is indeed a phase conjugate of the probe signal. Furthermore, the absence of the peak at 2fp − f1 in the simulated spectrum supports the notion that it 5 Conclusion In summary, this work demonstrates through experi- ments and simulations the generation of a phase con- jugate magnon from a non-degenerate four-wave mix- ing process. The phase conjugate signal is enhanced when the applied magnetic field strength is such that the pumps form a standing wave across the width of the waveguide. This standing wave causes a large pump am- plitude which significantly increases the nonlinearity of region. The geometry of the experiment and simulation ensure that any return signal at fc must be a phase conjugate signal. This new phase conjugate differs from previous ob- servations in magnonic systems in three ways: 1) It uti- lizes CW signals rather than spin wave bullets or pulsed signals 2) It uses a third-order FWM process in con- trast to three-wave parametric pumping. 3) We utilise the 2D nature of the waveguide with perpendicularly travelling spin waves. Future work would involve the creation of degenerate FWM, though discriminating be- tween monochromatic signals poses different challenges. Our work opens the door for this type of phase conjuga- tion as yet another process that is exploitable in novel magnon-based computational paradigms. Acknowledgements We would like to extend our gratitude towards Prof. Paul Ewart for his invaluable insight in the nu- ances of this experiment. This research was partially funded by Magdalen College, Oxford. On behalf of all authors, the corresponding author states that there is no conflict of interest. Alistair Inglis et al. dens. Matter Mater. Phys. 79(13), 1 (2009). DOI 10.1103/PhysRevB.79.134411 19. A.A. Serga, B. Hillebrands, S.O. Deraokritov, A.N. Slavin, P. Wierzbicki, V. Vasyuchka, O. Dzyapko, A. Chumak, Phys. Rev. Lett. (2005). DOI 10.1103/ PhysRevLett.94.167202 20. Y. Khivintsev, J. Marsh, V. Zagorodnii, I. Harward, J. Lovejoy, P. Krivosik, R.E. Camley, Z. Celinski, Appl. Phys. Lett. 98(4) (2011). DOI 10.1063/1. 3541787 21. J. Marsh, R.E. Camley, Phys. Rev. B - Condens. Matter Mater. Phys. 86(22), 1 (2012). DOI 10.1103/PhysRevB.86.224405 22. A. Yariv, D.M. Pepper, Opt. Lett. (1977). DOI 10.1364/OL.1.000016 23. D.M. Pepper, R.L. Abrams, Opt. Lett. 3(6), 212 (1978) 24. A. Prabhakar, D.D. Stancil, Spin Waves (Springer US, Boston, MA, 2009). 10.1007/978-0-387-77865-5 DOI 25. B.A. Kalinikos, A.N. Slavin, J. Phys. C Solid State Phys. 19(35), 7013 (1986). DOI 10.1088/ 0022-3719/19/35/014 26. A. Vansteenkiste, J. Leliaert, M. Dvornik, M. Helsen, F. Garcia-Sanchez, B. Van Waeyen- berge, AIP Adv. (2014). DOI 10.1063/1.4899186 27. A.V. Chumak, A.A. Serga, B. Hillebrands, J. Phys. D. Appl. Phys. 50(24), 244001 (2017). DOI 10.1088/1361-6463/aa6a65 6 References 1. M.M. Waldrop, Nature 530(7589), 144 (2016). DOI 10.1038/530144a 2. S. Manipatruni, D.E. Nikonov, I.A. Young, Nat. Phys. (2018). DOI 10.1038/s41567-018-0101-4 3. A.V. Sadovnikov, S.A. Odintsov, E.N. Beginin, A.A. Grachev, V.A. Gubanov, S.E. Sheshukova, Y.P. Sharaevskii, S.A. Nikitov, JETP Lett. (2018). DOI 10.1134/S0021364018010113 4. L.J. Cornelissen, J. Liu, B.J. Van Wees, R.A. Duine, Phys. Rev. Lett. (2018). DOI 10.1103/ PhysRevLett.120.097702 5. A.V. Chumak, V. Vasyuchka, A.A. Serga, B. Hille- brands, Nat. Phys. 11(June), 453 (2015). DOI DOI:10.1038/NPHYS3347 6. P. Gruszecki, M. Kasprzak, A.E. Serebryannikov, M. Krawczyk, W. Smigaj, Sci. Rep. (2016). DOI 10.1038/srep22367 7. A. Haldar, C. Tian, A.O. Adeyeye, Sci. Adv. (2017). DOI 10.1126/sciadv.1700638 8. J. Stigloher, M. Decker, H.S. Körner, K. Tanabe, T. Moriyama, T. Taniguchi, H. Hata, M. Madami, G. Gubbiotti, K. Kobayashi, T. Ono, C.H. Back, Phys. Rev. Lett. (2016). DOI 10.1103/ PhysRevLett.117.037204 9. G. Csaba, Á. Papp, W. Porod, Phys. Lett. A 381(17), 1471 (2017). DOI 10.1016/j.physleta.2017. 02.042 10. O. Zografos, B. Soree, A. Vaysset, S. Cosemans, L. Amaru, P.E. Gaillardon, G. De Micheli, R. Lauw- ereins, S. Sayan, P. Raghavan, I.P. Radu, A. Thean, IEEE-NANO 2015 - 15th Int. Conf. Nanotechnol. pp. 686 -- 689 (2016). DOI 10.1109/NANO.2015. 7388699 11. R.A. Fisher, B.R. Suydam, D. Yevick, Opt. Lett. (1983). DOI 10.1364/OL.8.000611 12. A.E. Chiou, Proc. IEEE 87(12), 2074 (1999). DOI 10.1109/5.805922 13. G. Unnikrishnan, J. Joseph, K. Singh, Appl. Opt. (2008). DOI 10.1364/ao.37.008181 14. P. Ewart, S.V. O'Leary, Opt. Lett. 11(5), 279 (1986). DOI 10.1364/OL.11.000279 15. D.M. Bloom, G.C. Bjorklund, Appl. Phys. Lett. (1977). DOI 10.1063/1.89791 16. R.W. Boyd, Nonlinear Optics, Third Edition, 3rd edn. (Academic Press, Inc., Orlando, FL, USA, 2008) 17. G.A. Melkov, V.I. Vasyuchka, A.V. Chumak, A.N. Slavin, J. Appl. Phys. 98(7) (2005). DOI 10.1063/ 1.2077842 18. G.A. Melkov, Y.V. Koblyanskiy, R.A. Slipets, A.V. Talalaevskij, A.N. Slavin, Phys. Rev. B - Con-
1702.01859	1	1702	2017-02-07T02:58:51	Ferromagnetism in chiral multilayer 2D semimetals	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	We calculate the temperature dependent long-range magnetic coupling in the presence of dilute concentrations of random magnetic impurities in chiral multilayer two-dimensional semimetals, i.e., undoped intrinsic multilayer graphene. Assuming a carrier-mediated indirect RKKY exchange interaction among the well-separated magnetic impurities with the itinerant carriers mediating the magnetic interaction between the impurities, we investigate the magnetic properties of intrinsic multilayer graphene using an effective chiral Hamiltonian model. We find that due to the enhanced density of states in the rhombohedral stacking sequence of graphene layers, the magnetic ordering of multilayer graphene is ferromagnetic in the continuum limit. The ferromagnetic transition temperature is calculated using a finite-temperature self-consistent field approximation and found to be within the experimentally accessible range for reasonable values of the impurity-carrier coupling.	cond-mat.mes-hall	cond-mat	Ferromagnetism in chiral multilayer 2D semimetals Hongki Min1,2, E. H. Hwang1,3, and S. Das Sarma1 1Condensed Matter Theory Center and Joint Quantum Institute, Department of Physics, University of Maryland, College Park, Maryland 20742-4111 2Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea 3SKKU Advanced Institute of Nanotechnology and Department of Physics, Sungkyunkwan University, Suwon, 16419, Korea We calculate the temperature dependent long-range magnetic coupling in the presence of dilute concentrations of random magnetic impurities in chiral multilayer two-dimensional semimetals, i.e., undoped intrinsic multilayer graphene. Assuming a carrier-mediated indirect RKKY exchange in- teraction among the well-separated magnetic impurities with the itinerant carriers mediating the magnetic interaction between the impurities, we investigate the magnetic properties of intrinsic multilayer graphene using an effective chiral Hamiltonian model. We find that due to the enhanced density of states in the rhombohedral stacking sequence of graphene layers, the magnetic ordering of multilayer graphene is ferromagnetic in the continuum limit. The ferromagnetic transition tem- perature is calculated using a finite-temperature self-consistent field approximation and found to be within the experimentally accessible range for reasonable values of the impurity-carrier coupling. I. INTRODUCTION Multilayer graphene with an additional layer degree of freedom (in addition to the spin and pseudospin in- tralayer index) has recently attracted a great deal of at- tention for its fundamental properties as well as for its potential applications.1 -- 7 Multilayer graphene is not a simple extension of monolayer graphene (since it has its own characteristic layer-number-dependent band struc- ture and symmetry properties), and could open the pos- sibility of engineering electronic properties by tuning the stacking arrangement. One important salient feature of multilayer graphene is the enhancement of the electronic density of states (DOS) as the number of graphene layers increases. As a consequence, the electronic screening be- comes more important with increasing layers.6 Since the energy band structure of multilayer graphene is very sen- sitive to its stacking sequence, the screening properties depend strongly on the stacking arrangements in multi- layer graphene.6,7 Each type of rhombohedral multilayer graphene (i.e., J-graphene) with the layer number index J = 1, 2, 3, 4, 5... is a distinct 2D material tuned by J except that all of them are 2D gapless semimetals with the chemical potential precisely pinned at the touching point of conduction and valence bands. The most well- known J-grahenes are monolayer graphene (MLG) with J = 1 and bilayer graphene (BLG) with J = 2, but tri- layer (J = 3) and even higher-layer (J > 3) graphenes have also been studied in the laboratory.5 In fact, the J going to infinity limit (i.e., infinite-layer graphene) is graphite. The current theoretical work is on the J-dependent magnetic properties of intrinsic (i.e., undoped) multi- layer graphene with no free carriers in the conduction or valence band at T = 0. We study finite temperature response (screening) functions of multilayer graphene, and their consequences for the J-dependent magnetic properties induced by magnetic impurities through the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction.8,9 In the presence of a dilute concentration of magnetic im- purities in nonmagnetic metals the effective exchange in- teraction between the impurities is induced as the second- order perturbation with respect to the direct exchange interaction between the magnetic impurity and the itin- erant electrons of the host (i.e., the magnetic impurities experience a long-range indirect exchange interaction me- diated by conduction electrons, known as the RKKY in- teraction). Such an indirect nonlocal carrier-mediated RKKY interaction between two impurities could be fer- romagnetic or antiferromagnetic depending on their spa- tial separation because the interaction is oscillatory due to the sharpness of the Fermi surface. This indirect RKKY interaction exists in addition to any possible di- rect exchange interaction among the magnetic impuri- ties which may arise due to their direct wave function overlap. Because the RKKY interaction is mediated by host electrons (or holes), the enhanced DOS of multilayer graphene can lead to an increase of the effective inter- impurity magnetic coupling with increasing J, which may induce robust magnetic ordering for larger values of the layer number. The goal of the current paper is to the- oretically predict such magnetic ordering in multilayer graphene as a function of layer number. In general, intrinsic graphene does not have any per- manent magnetic moments in the bulk, but local mag- netic moments can be introduced by extrinsic doping. Doping by suitable magnetic impurities could introduce such local moments, but these local moments cannot or- der spontaneously unless there is an inter-impurity mag- netic interaction. If the impurities are far apart, i.e., the doping is dilute (which is the only situation con- sidered in the current work), then the direct exchange interaction among the impurities is basically zero since their wave function overlap is exponentially small. It is also now well-accepted that graphene (or generally, mul- tilayer graphene) is not intrinsically (i.e., in the absence of any doping) magnetic, i.e., there is no spontaneous graphene magnetic ordering of any kind unless local mag- netic moments are extrinsically introduced by magnetic impurities, vacancies, or edges.10 The important question addressed in this paper is whether a dilute (i.e., well- separated impurities) concentration of magnetic dopants could induce magnetism in graphene through the RKKY mechanism with direct exchange playing no role what- soever. We emphasize that the dilute limit is very dif- ferent from the dense (or Kondo lattice) limit where the magnetic impurities themselves form a lattice (or are in almost every unit cell of the host lattice) since direct interaction among the impurities as well as any modi- fication of the graphene band structure induced by the impurities can be neglected. The only interaction to be considered in this dilute limit is the indirect RKKY in- teraction, and a continuum approximation should suffice. The dilute approximation with well-separated dopants makes the graphene situation here very similar to the extensively studied11,12 diluted magnetic semiconductor (DMS) materials where also the primary mechanism driv- ing ferromagnetism in the semiconductor is thought to be the indirect RKKY interaction. This is the physics we investigate in the current work. The main qualitative difference between diluted-magnetic-graphene (DMG) we consider here and the well-studied DMS is that we must deal with an undoped gapless semimetal (and not a doped semiconductor) and we must incorporate the layer num- ber for multilayer graphene. It is known that the effective RKKY interaction in graphene induced by the exchange coupling between local magnetic moments and conduction electrons (or holes) behaves differently from the ordinary two dimensional (2D) systems because of the chiral gapless nature of graphene.13,14 In particular, the strength of RKKY cou- pling in monolayer graphene decays faster spatially than in ordinary 2D electron systems due to chirality (i.e., the suppression of the 2kF scattering, where kF is the Fermi wave vector) in graphene.13,14 One may wonder if RKKY interaction, with its long- is capable of induc- range spatial Friedel oscillations, ing magnetic ordering since the inter-impurity interac- tion may be of random sign (ferromagnetic or antifer- romagnetic) depending on their spatial locations. This is certainly the case in ordinary metals or semiconduc- tors (either 2D or 3D)9 in the presence of a high concen- tration of magnetic impurities where the effective inter- impurity interaction will be randomly ferromagnetic and antiferromagnetic, leading to considerable frustration in the Hamiltonian (and perhaps therefore a glassy ground state with no obvious long-range order). But for a low or dilute impurity concentration (as in DMS) where the impurities are on the average far from each other, the effective inter-impurity RKKY interaction is on the av- erage mostly ferromagnetic, and then the possibility of magnetic ordering arises, albeit with perhaps a low tran- sition temperature (as in DMS) because of the generally weakened average inter-impurity RKKY interaction.11,12 In multilayer graphene, layer index and chirality intro- duce novel RKKY physics necessitating a specific analy- sis to search for possible magnetic ordering of DMG. In particular, multilayer graphene RKKY interaction does 2 not change sign (in spite of Friedel oscillations) for J ≥ 3 indicating a strong tendency toward a long-range order- ing of the magnetic impurities induced by RKKY cou- pling. Another fundamental difference between intrinsic graphene and ordinary metals or doped semiconductors is the fact that undoped graphene is a gapless semicon- ductor or a semimetal with no free carriers at T = 0 since the chemical potential separates a filled valence band touching an empty conduction band. Taken together, all these differences between chiral intrinsic DMG and regu- lar DMS imply that our intuition based on the substantial body of DMS literature is a poor guide to understanding how graphene magnetism may arise from RKKY physics. In this paper, we provide a complete picture based on a continuum mean field theory, which should be qualita- tively and semiquantitaively valid in the dilute impurity density limit. Magnetic properties of graphene have been studied, and in particular, there have been several studies of RKKY interaction in graphene focusing on the possibil- ity of magnetic ordering of dopant magnetic impurities at zero temperature.13 -- 22 However, a systematic study of RKKY interaction in multilayer graphene as a func- tion of layer index has not been undertaken, and, in addition, finite temperature, disorder and finite carrier mean-free path effects on RKKY interaction have not yet been studied theoretically. In this paper, we cal- culate the magnetic properties of multilayer graphene with the magnetic impurities located at the interface between graphene and substrate without breaking any symmetry in the graphene layer. We calculate the tem- perature dependence of the RKKY interaction in multi- layer graphene in order to develop a self-consistent field theory to study long-range finite-temperature magnetic ordering. We show that the enhanced DOS in rhom- bohedral stacking allows ferromagnetic ordering of the magnetic impurities at experimentally accessible temper- atures, particularly for higher values of J. Our results in- dicate that the magnetic impurity induced ferromagnetic ordering is possible in semimetallic multilayer graphene arising from the RKKY indirect interaction in the dilute impurity limit. Ferromagnetism in DMG as predicted in our theory, particularly for larger layer numbers, could in principle lead to graphene spintronics if our predictions are validated experimentally. This paper is organized as follows. In Sec. II, we de- scribe our model and theoretical approaches based on chiral 2D electron systems. In Sec. III, we give the cal- culated results for RKKY interaction and the ferromag- netic transition temperatures of DMG. We conclude in Sec. IV with a discussion of the momentum cutoff effects on the effective coupling. II. MODEL To study indirect magnetic interaction between quenched local moments in multilayer graphene, we con- sider the indirect exchange interaction between magnetic impurities to be of the RKKY form, i.e., carrier-mediated effective magnetic interaction. Indirect exchange inter- action between magnetic moments is determined by the electronic structure of the relevant system. To describe electron states in multilayer graphene, we consider the Hamiltonian of noninteracting electrons in the form of a two-band pseudospin Hamiltonian for 2D chiral quasipar- ticles. Thus, multilayer graphene near the band-touching Dirac point can be described by a set of chiral 2D electron systems (C2DESs) and the Hamiltonian with the chiral- ity index J (which also represents the number of layers in rhombohedral multilayer graphene) is of the form6,7 HJ (k) = t⊥  0 t⊥ (cid:17)J (cid:16) v0k+ t⊥ (cid:17)J (cid:16) v0k− 0   , (1) and the corresponding eigenfunctions are where k± = kx ±iky, v0 is the effective in-plane Fermi ve- locity, and t⊥ is the nearest-neighbor interlayer hopping. The Hamiltonian has an energy spectrum given by ελ,k = t⊥ (cid:17)J λt⊥(cid:16) v0k λ, ki = 1√2(cid:0)λ, eiJφk(cid:1), where φk = tan−1(ky/kx) and λ = ±1 for conduction (valence) band energy states, re- spectively. We consider intrinsic multilayer semimetallic graphene with the Fermi energy precisely at the Dirac point which we take to be the zero of energy. The carrier mediated RKKY indirect exchange interac- tion describing the effective magnetic interaction between local moments induced by the free carrier spin polariza- tion is proportional to the static carrier susceptibility. The finite temperature static susceptibility can be cal- culated using the finite temperature Fermi distribution function as1,23 χ(q, T ) = −gXλ,λ′Z d2k (2π)2 fλ,k − fλ′,k′ ελ,k − ελ′,k′ Fλ,λ′ (k, k′), (2) where g = gsgv is the total degeneracy factor (gs = gv = 2 are spin and valley degeneracy factors, respectively), fλ,k = 1/[exp(ελ,k/kBT ) + 1] is the finite temperature Fermi distribution function for the band index λ = ±1 and wave vector k, Fλ,λ′ (k, k′) is the square of the wave- function overlap between λ, ki and λ′, k′i states, and k′ = k + q. For the chiral electron with the chirality index J, Fλ,λ′ (k, k′) = 1 2 [1 + λλ′ cos J(φk − φ′k)]. For the undoped intrinsic case, in which the chemical potential is located at the Dirac point for all tempera- tures, Eq. (2) can be expressed as χ(q, T ) = DJ (q)[I + J (q, T ) + I−J (q, T )] = χ+ J (q, T ) + χ−J (q, T ), (3) where DJ (q) corresponds to the DOS with a wave number q which is given by DJ (q) = gq2−J 2πJt⊥(v0/t⊥)J , (4) and 3 dφ 2π 1 ± cos(Jθ) xJ ± (x′)J I±J (q, T ) = JZ ∞ × (cid:20)tanh 0 xdxZ 2π TJ (qax)J 0 T T TJ (qax′)J ± tanh (cid:21) , (5) where x′ = p1 + 2x cos φ + x2, cos θ = (x + cos φ)/x′, TJ = (t⊥/kB)(v0/t⊥a)J , and a is the lattice constant of graphene. At zero temperature (T = 0), I−J , which corre- sponds to the intraband transition of electrons, vanishes for all q (since the conduction band is completely empty and the valence band completely full) and only interband transition, I + J , contributes to the susceptibility, and we have χ(q, T = 0) = χ+ J (q)6. As q → 0, we have χ+ J (q, T = 0) ∝ q2−J and therefore, for J ≥ 3, the static susceptibility diverges at long wavelength. J (q) = DJ (q)I + At finite temperatures (T > 0), it is interesting to notice that the interband susceptibility behaves like χ+ J (q, T ) ∝ q2 as q → 0 for all J. Thus, the q = 0 sin- gular behavior of the interband susceptibility at T = 0 for J ≥ 3, i.e., χ+ J (q → 0, T = 0) → ∞, disappears at finite temperatures. In addition, intraband transitions contribute to the susceptibility at finite temperatures due to thermal particle-hole excitations in the semimetal, and thus, at small q the total susceptibility comes entirely from the intraband transition at finite T , χ−J , due to the vanishing of long wavelength interband contribution. Es- pecially, we find that χ−J (q = 0, T ) ∝ T 2−J for J ≤ 2 and χ−J (q = 0, T ) ∝ 1/T for J ≥ 3. Thus, the total suscepti- bility at q = 0, χ(0, T ), increases with temperature only for J = 1. For J = 2, χ(0, T ) is constant for all temper- atures, and it decreases with increasing temperature for J ≥ 3. We will discuss the implications of these temper- ature dependences for J-dependent long-range magnetic ordering. It is not possible to obtain the susceptibility function analytically for all q at finite temperatures. Thus, we cal- culate the finite temperature static susceptibility numer- ically. Figure 1 shows the calculated static susceptibility χ(q, T ) as a function of wave vector for several tempera- tures. For comparison, we normalize the susceptibility for different J by the J = 1 DOS at q = 1/a, DJ=1(q = a−1), where a is the lattice constant of graphene. As shown in Fig. 1(a) the J = 1 susceptibility increases linearly with temperature at small q. For J = 2 we find that χ(0, T = 0) 6= χ(0, T → 0) and χ(0, T = 0)/χ(0, T 6= 0) = ln 4 for all finite temperatures. The finite temperature suscep- tibility at q = 0 is independent of the temperature for J = 2 as discussed above. It is interesting to compare this behavior with the asymptotic form for the corre- sponding nonchiral regular 2D electron gas susceptibility which exponentially decreases from its zero temperature value, χ(q = 0, T ) ≈ χ(q = 0, T = 0)[1 − e−TF /T ], where TF is the Fermi temperature of the system.1,24 For J ≥ 3, even though the zero temperature susceptibility is infin- ity at q = 0, the finite temperature susceptibility is finite and χ(0, T ) decreases inverse linearly with temperature, J=1 (a) T=0K J=2 (b) T=10 K3 T=10 K3 0.1 ) 1 − a ( 1 D / ) T q ( χ , 0 0 0.3 T=0K 0.1 qa (c) J=3 0.2 0 0.1 qa J=4 (d) T=0K T=0K 0.2 (e) J=5 T=0K ) 1 − a ( 1 D / ) T q ( χ , 0.2 0.1 0 0 T=10 K3 T=10 K3 T=10 K3 0.1 qa 0.2 0 0.1 qa 0.2 0 0.1 qa 0.2 FIG. 1: The calculated finite temperature static polarizability χ(q, T ) as a function of wave vector for various temperatures T = 0, 200, 400, 600, 800, and 1000 K, and for different chiralities (a) J = 1, (b) J = 2, (c) J = 3, (d) J = 4, and (e) J = 5. Here D1(a−1) = g/(2πt⊥a2)(t⊥a/v0) is given in Eq. (4) with J = 1 and q = q−1, and a is the length scale of the system (i.e., lattice constant) and a = 2.46A is used. 1/T . Overall χ(q, T ) for J ≥ 3 manifests a similar be- havior. III. RKKY INTERACTION AND EFFECTIVE MAGNETIC COUPLING The interaction between a localized spin Si located at ri and an itinerant electron spin s at r, i.e., V (r) = JexSi · s δ(ri − r) with an exchange coupling Jex, ac- counts for the interaction between magnetic impurities. In general, Jex is an unknown parameter in our theory which must be determined experimentally or from a sep- arate first principles calculation beyond the scope of the current work. Then the effective Hamiltonian describ- ing magnetic interaction between two classical Heisen- berg spins Si and Sj located at ri and rj, respectively, is given by26,30 H = −Xi,j where JRKKY(ri − rj)Si · Sj, (6) JRKKY(r, T ) = (cid:2)Jexa2(cid:3)2 4 χ(r, T ). (7) 4 Note that the 'classical moment' approximation here is justified by the large moments of the magnetic impu- rities typically used for magnetic doping and quantum fluctuations in the magnetic impurity are neglected as being small. We ignore all complications associated with Kondo physics and other quantum strong correlation as- pects assuming that the long-range Heisenberg model is the appropriate model for describing magnetic or- dering for DMG. We essentially assume that the rele- vant Kondo temperature is much less than the RKKY temperature scale in the system.25 The RKKY interac- tion is related to the range function which is defined by the Fourier transform of the static susceptibility, i.e., χ(r, T ) =Pq χ(q, T ) and in 2D it is given by χ(r, T ) =Z ∞ 0 qdq 2π J0(qr)χ(q, T ), (8) where J0(x) is the Bessel functions of the first kind. Note that even though the Fourier transform of Eq. (8) is well defined for J = 2 and 3, it requires an ultraviolet cut- off for J = 1 and an infrared cutoff for J ≥ 4. These large and small momenta regularizations are necessary for obtaining meaningful nonsingular results. In the fol- lowing results, we set the infrared momentum cutoff as q(l) c = 0.01/a and the ultraviolet momentum cutoff as q(h) c = 1/a, where a is the typical length scale of the system, i.e., a lattice constant of graphene, and we use a = 0.246 nm in our numerical calculations. While the large momentum ultraviolet cutoff (inverse lattice con- stant) for J = 1 is natural in a continuum theory, the infrared low momentum cutoff is not usual in solid state physics. We provide the details on these regularizations and their possible effects in the discussion section (see Sec. IV). Figure 2 shows the RKKY range functions χ(r, T ) for J = 1, 2, 3, 4 and for different temperatures. For J = 1 the range function oscillates for all temperatures, and its magnitude increases with temperature. For J = 2 the range function is almost independent of temperature. For J ≥ 3 the magnitude of the range function decreases with temperature. These behaviors for different J can be understood from the temperature dependence of the susceptibility shown in Fig. 1. More importantly, we find that for J = 1, 2 the range functions alternate between positive and negative values while, for J ≥ 3, it always remains positive at T = 0. Due to the suppression of large q contribution for higher values of layer number J, the range functions for J ≥ 3 do not have oscillations. Hence, there are no competing ferro- and antiferromag- netic couplings for J ≥ 3, and the magnetic impurity moments are expected to be ferromagnetically aligned since there is no frustration in the RKKY coupling. The carrier-mediated RKKY interaction induced in- direct exchange interaction [Eqs. (6) and (7)] describes the effective magnetic interaction between local mag- netic moments induced by the free carrier spin polar- ization. The effective temperature-dependent coupling is J=1 (a) 2 J=2 (b) mean-free path26 -- 29. Thus the effective coupling can be modified as 5 4 2 ) 2 − 0 1 ( 0 D / ) T , r ( χ 0 -1 0 2 ) 3 − 0 1 ( 0 D 1 / ) T , r ( χ 0 0 1 ) 3 − 0 1 ( 0 D 0 / ) T , r ( χ 5 10 r/a 15 20 J=3 (c) -1 0 1.0 5 10 r/a 15 20 J=4 (d) ) 3 − 0 1 ( 0 0.5 D / ) T , r ( χ 5 10 r/a 15 20 0.0 0 10 r/a 20 FIG. 2: The RKKY range function χ(r) as a function of distance for different chiralities (a) J = 1, (b) J = 2, (c) J = 3, (d) J = 4. In each figure the different curves represent different temperatures T = 0, 10, 100, 1000 K [from bottom to top in (a), and from top to bottom in (b), (c), (d)]. Here D0 = D1(a−1)/a2. In this calculation the infrared momentum cutoff q(l) c = 0.01/a for J = 4 and the ultraviolet momentum cutoff q(h) c = 1/a for J = 1 are used. then given by the spatial average of the JRKKY: Jeff (T ) = 1 Ωunit Z d2rJRKKY(r, T ), (9) where Ωunit is the area of a unit cell. The effective tem- perature dependent coupling can be expressed in the di- mensionless form Jeff (T ) J (0) eff = 1 D1(a−1)Z rdrχ(r, T ) (10) eff =(cid:2)Jexa2(cid:3)2 where J (0) × 2πD1(a−1)/4Ωunit is a magnetic coupling constant, which is proportional to the square of the local exchange coupling, but independent of the chirality index J and temperature T . All the interesting physics of chirality index J and temperature T enters through the integral in Eq. (10) which depends nontriv- ially and nonlinearly on both J and T . We also include disorder effects phenomenologically through a finite car- rier mean-free path by including an exponential cutoff in the range of the RKKY interaction, which allows us to take into account the dependence of the magnetic be- havior of multilayer graphene on the carrier transport properties. In the presence of (nonmagnetic) impurity scattering the RKKY interaction range is cut off at long distances, and we include this physics through an expo- nential spatial damping at distances larger than a char- acteristic length scale of the order of the carrier transport Jeff =( 1 1 Ωunit R d2rJRKKY(r), Ωunit R d2rJRKKY(r)e− r−R R . (r < R) (r > R) (11) Here the exponential cutoff R is introduced to take into account the finite mean-free path due to scattering by nonmagnetic disorder in the semimetal. In the calcula- tion, we use R = 100a = 24.6 nm, which is a charac- teristic scale of the mean-free path, and the choice of R (< 300a) does not change our results qualitatively. We note that the appropriate mean free path (R) here is the one corresponding to undoped intrinsic multilayer graphene near the Dirac point, which depends on J and T (and should be taken from transport data). We use the length cutoff parameter R just as an adjustable pa- rameter in the theory since making R large (small) pro- vides a convenient way to study the qualitative effects of long (short)-range RKKY interaction on graphene mag- netism. Obviously, magnetism is strongly suppressed when R is small. If experimental results on multilayer graphene magnetism become available in the future, it is straightforward to include quantitative effects of a J and T dependent mean-free path in our theory. Figure 3 shows the calculated temperature dependence of the effective coupling for C2DESs with chiralities J = 1, 2, 3, 4, 5. As shown in Fig. 3, the effective cou- pling for J ≥ 2 decreases with temperature as in ordi- nary non-chiral 2DES,29 but for J = 1 it increases with temperature, which is the direct consequence of the tem- perature dependence of the static susceptibility as shown in Fig. 1. The effective coupling also increases with in- creasing chiral index J (or number of layers in rhombo- hedral multilayer graphene) because of the susceptibility behavior at long wavelength limit as shown in Fig. 1. Obviously the T and J dependence of the effective mag- netic coupling shown in Fig. 3 determines the magnetic transition temperature in DMG as discussed below. From the calculated effective coupling we obtain the critical temperature for the magnetic transition in mul- tilayer graphene. For the Heisenberg classical spins the mean-field transition temperature Tc is given by30,31 kBTc = S(S + 1) 3 xJeff , (12) where S is the impurity spin, x = nimpa2 is the concen- tration of the local moments with nimp being the effective 2D magnetic impurity doping density. In the absence of any other information, we assume the magnetic dopants to be randomly distributed, but it is easy to include any correlations among the dopant positions if such dopant clustering effects are important. We note that the fer- romagnetic transition temperature is proportional to J 2 ex and x, but its dependence on the layer index J is highly nontrivial and cannot be simply inferred using dimen- sional analysis since the layer index J enters the DOS in a highly nonlinear manner [see Eq. (4)]. We note that 100 10-1 10-2 10-3 ) 0 ( f f e J / ) T ( f f e J 10-4 10-5 0 J=5 J=4 J=3 J=2 J=1 50 100 150 T (K) 200 250 300 c = 0.01/a, q(h) FIG. 3: The calculated effective coupling (solid lines) as a function of temperature for various chiralities J = 1, 2, 3, 4, 5. In this calculation, q(l) c = 1/a, and exponen- tial cutoff R = 100a are used. Here the normalization factor eff = (cid:2)Jexa2(cid:3)2 J (0) × 2πD1(a−1)/4Ωunit is independent of the chirality index J and temperature T . The dashed line repre- sents 3kBT /[S(S + 1)xJ (0) eff ] and the intersections with Jeff (T ) indicate the transition temperatures solved self-consistently. in our finite temperature RKKY model the ferromagnetic transition temperature is obtained from Eq. (12) by solv- ing it self-consistently because Jeff itself is also strongly temperature dependent.31 We emphasize that the strong temperature-dependence of the RKKY interaction in in- trinsic graphene is the key physics determining the DMG ferromagnetic transition temperature in the theory. If one makes the simplistic (and incorrect) assumption that Jeff is a temperature-independent coupling given by its T = 0 value Jeff (0), then the transition temperatures are going to be unrealistically high. The self-consistent so- lution of Eq. (12) using the full temperature dependence of the magnetic coupling as shown in Fig. 3 is crucial in the theory to obtain the correct magnitude as well as the correct J-dependence of the transition temperature Tc.31 In Fig. 4 we show the calculated self-consistent tran- sition temperature Tc as a function of layer index J. With the temperature dependent Jeff (T ) and typical val- ues of Jex = 1 eV, S = 5/2, x = 0.05, Ωunit/a2 = 1, the self-consistent results show the ferromagnetic tran- sition temperatures Tc ≈ 0.05, 2.6, 11, 18 and 23 K for J = 1, 2, 3, 4, 5, respectively, [see Fig. 4(a)]. In Fig. 4(b) we compare them with the results calculated in the non- consistent method with Jeff (T = 0) value noting that the non-selfconsistent Tc is unreasonably high. Fig. 4(b) shows that the zeroth order mean field results assuming Jeff to be given by its T = 0 value overestimate Tc by an order of magnitude (or more) for J ≥ 3 compared with the self-consistent results. We note that Tc is pro- portional to xJ 2 exS(S + 1) and the results in Fig. 4 are for very specific values of x, Jex, and S, but one can scale the results to obtain Tc for other values of Jex, S, 6 and x. We do emphasize, however, that x cannot be too large so that one is in the dilute moment regime for the validity of our continuum theory. For large impurity concentration (x > 0.1 or so) DMG physics is different since graphene band structure itself may be affected. In addition, the results obviously depend also on the basic graphene band parameters g, v0, and t⊥ [see Eq. (4)] and this dependence is complex. We choose the standard pa- rameter values: g = 4, v0 = 106 m/s, and t⊥ = 0.3 eV in our calculations. We discuss the dependence on various cutoff parameters in the next section. Our results apply to the rhombohedral stacking of graphene layers because the rhombohedral stacking sequence with J layers is de- scribed by C2DES with the chirality index J. Thus, with high values of layer index J the ferromagnetic ordering of magnetic impurities can be experimentally accessible in rhombohedral multilayer graphene provided suitable magnetic dopants are used with reasonable (∼ 1 eV or so) local exchange coupling. We note that the calculated J dependence of the self- consistent Tc in Fig. 4 is roughly linear whereas the cor- responding dependence in the non-self-consistent mean field theory is nonlinear with a high power of J. We do not believe that there is a generic unique power law be- havior of Tc on J, and the linear dependence in Fig. 4 applies only for our calculated results although it should be approximately valid for higher J values. Of course, Tc is strongly suppressed for short mean free path due to disorder effects, which can only be discussed quanti- tatively for specific experimental situations. IV. DISCUSSION AND CONCLUSION We have studied the temperature dependence of the RKKY interaction and effective magnetic ordering as a function of layer number index for the C2DES of rhombo- hedrally stacked multilayer graphene. The chiral effective Hamiltonian used in this work is obtained from a pertur- (a) 102 101 100 10-1 ) K ( c T (b) 104 102 100 ) K ( c T 1 2 3 J 4 5 10-2 1 2 4 5 3 J FIG. 4: The calculated ferromagnetic transition temperature as a function of layer (chiral) index J. We use the param- eters Jex = 1 eV, S = 5/2, x = 0.05, and Ωunit/a2 = 1 in this calculations. In (a) the self-consistent results are shown for different J. In (b), the self-consistent results (circles) are compared with the results (squares) obtained in the non- consistent method with Jeff (T = 0). 104 102 100 ) 0 ( f f e J / f f e J 10-2 J=5 J=4 J=3 J=2 J=1 10-4 10-4 (-) (+) (-) 10-3 (l) qc a 10-2 10-1 FIG. 5: The infrared momentum-cutoff dependence of the effective coupling for various chiralities J = 1, 2, 3, 4, 5 with fixed high momentum cutoff q(h) c = 1/a and exponential cutoff R = 100a. bation theory taking into account only nearest-neighbor intralayer and interlayer hoppings7, which is valid when we neglect the contributions from the trigonal warping terms which are much smaller than the terms kept in the effective Hamiltonian. Our theory is valid only when quantum fluctuations and direct exchange coupling be- tween the impurity moments are negligible, which should be valid for large impurity spins and dilute impurity con- centrations. The valid energy scale for the chiral effective model is given by 0.03 eV -- 0.3 eV32, which corresponds to the momentum scale of 0.01/a -- 0.1/a. Thus it is natural to introduce the infrared low momentum cutoff and the ultraviolet high momentum cutoff, denoted by q(l) and c q(h) , respectively. These regularizations are necessary for c obtaining meaningful results in graphene. In our model, the calculated effective coupling Jeff is insensitive to the ultraviolet cutoff q(h) for J ≥ 2, while for J = 1, Jeff shows the well-known logarithmic ultravio- let divergence at high momenta. Note that for monolayer graphene (J = 1 C2DES) there is no interlayer hopping c 1 S. Das Sarma, S. Adam, E. H. Hwang, and E. Rossi, Rev. Mod. Phys. 83, 407 (2011). 2 H. Raza, Graphene Nanoelectronics: Metrology, Synthesis, Properties and Applications (Springer, Berlin, 2012). 3 A. Yacoby, Nat. Phys. 7, 925 (2011). 4 P. H. Tan, W. P. Han, W. J. Zhao, Z. H. Wu, K. Chang, H. Wang, Y. F. Wang, N. Bonini, N. Marzari, N. Pugno, G. Savini, A. Lombardo, and A. C. Ferrari, Nat. Mat. 11, 294 (2012). 5 W. Bao, L. Jing, J. Velasco Jr, Y. Lee, G. Liu, D. Tran, B. Standley, M. Aykol, S. B. Cronin, D. Smirnov, M. Koshino, E. McCann, M. Bockrath, and C. N. Lau, Nat. Phys. 7, 7 and the valid momentum scale is restricted only by the inverse lattice constant, beyond which the linear disper- sion is no longer valid. Even though our results are inde- pendent of the high momentum cutoff, they are affected by the low momentum cutoff q(l) c . As shown in Fig. 5, the calculated effective couplings are consistent for small values of the cutoff q(l) c < 10−2/a. However, for large values of cutoff the sign of Jeff oscillates with q(l) c . For a typical value of q(l) c = 0.01/a, Jeff > 0, and thus the ordering is ferromagnetic for all C2DESs. Note that the calculated results also depend on the exponential disor- der cutoff R for R > 500a, and a larger R gives more oscillating behavior in Jeff . The finite mobility of multi- layer graphene, however, restricts the size of R and for a typical scale of mean-free path the results do not change qualitatively. In addition, the finite mean free path cut- off prevents the system from becoming an interaction- induced ordered state with a non-zero energy gap at the Dirac point, thus we can use a chiral gas model of a gap- less semimetal even at zero carrier density. In summary, we study the magnetic properties of mul- tilayer graphene (chiral 2D electron systems) in the pres- ence of magnetic impurities as a function of layer index number in the intrinsic semimetallic situation. By calcu- lating the temperature dependent susceptibility of mul- tilayer graphene we investigate the temperature depen- dence of the RKKY interaction and the associated car- rier induced effective magnetic coupling using the effec- tive chiral model of multilayer graphene. We show that due to the enhanced DOS in rhombohedral stacking the ferromagnetic ordering between magnetic impurities is possible at experimentally accessible temperatures. Our results indicate that the magnetic impurity induced ferro- magnetic order in multilayer graphene should be observ- able experimentally for layer number 3 or above in mul- tilayer graphene system, perhaps ushering in the physics of spintronics based on diluted magnetic graphene. ACKNOWLEDGMENTS This work is supported by LPS-MPO-CMTC. 948 (2011); Chun Hung Lui, Zhiqiang Li, Kin Fai Mak, Emmanuele Cappelluti, and Tony F. Heinz, ibid 7, 944 (2011). 6 Hongki Min, E. H. Hwang, and S. Das Sarma, Phys. Rev. B 86, 081402(R) (2012); Yunsu Jang, E. H. Hwang, A. H. MacDonald, and Hongki Min, Phys. Rev. B 92, 041411(R) (2015). 7 Hongki Min and A. H. MacDonald, Phys. Rev. B 77, 155416 (2008); Hongki Min and A. H. MacDonald, Prog. Theor. Phys. Suppl. 176, 227 (2008). 8 M. A. Ruderman and C. Kittel, Phys. Rev. 96, 99 (1954); T. Kasuya, Prog. Theor. Phys. 16, 45 (1956); K. Yosida, Phys. Rev. 106, 893 (1957). 9 C. Kittel, in Solid State Physics, edited by F. Seitz, D. Turnbull, and H. Ehrenreich (Academic, New York, 1968), Vol. 22. 10 Oleg V. Yazyev, Rep. Prog. Phys. 73, 056501 (2010). 11 S. Das Sarma, E. H. Hwang, A. Kaminski, Solid State Commun. 127, 99 (2003). 12 T. Jungwirth, Jairo Sinova, J. Masek, J. Kucera, and A. H. MacDonald Rev. Mod. Phys. 78, 809 (2006). 13 L. Brey, H. A. Fertig, and S. Das Sarma, Phys. Rev. Lett. 99, 116802 (2007). 14 B. Wunsch, T. Stauber, F. Sols, and F. Guinea, New J. Phys. 8, 318 (2006). 15 M. A. H. Vozmediano, M. P. Lopez-Sancho, T. Stauber, and F. Guinea, Phys. Rev. B 72, 155121 (2005). 16 V. K. Dugaev, V. I. Litvinov, and J. Barnas, Phys. Rev. B 74, 224438 (2006). 17 M. Sherafati and S. Satpathy, Phys. Rev. B 83, 165425 (2011); ibid 84, 125416 (2011). 18 Annica M. Black-Schaffer, Phys. Rev. B 81, 205416 (2010). 19 Bruno Uchoa, T. G. Rappoport, and A. H. Castro Neto, Phys. Rev. Lett. 106, 016801 (2011). 20 S. R. Power and M. S. Ferreira, Phys. Rev. B 83, 155432 (2011); ibid 94, 235439 (2016). 8 21 E. Kogan, Phys. Rev. B 84, 115119 (2011). 22 L. Jiang, X. Li, W. Gao, G. Yu, Z. Liu, and Y. Zheng, J. Phys.: Condens. Matter 24, 206003 (2012). 23 E. H. Hwang and S. Das Sarma, Phys. Rev. B 75, 205418 (2007); Phys. Rev. Lett. 101, 156802 (2008). 24 E. H. Hwang and S. Das Sarma, Phys. Rev. B 79, 165404 (2009). 25 A. Allerdt, A. E. Feiguin, and S. Das Sarma, arXiv:1604.06109 (2016). 26 D. J. Priour, Jr., E. H. Hwang, and S. Das Sarma, Phys. Rev. Lett 92, 117201 (2004); ibid 95, 037201 (2005). 27 P. J. T. Eggenkamp, H. J. M. Swagten, T. Story, V. I. Litvinov, C. H. W. Swuste, and W. J. M. de Jonge, Phys. Rev. B 51, 15250 (1995). 28 S. Das Sarma, Phys. Rev. Lett. 50, 211 (1983). 29 S. Das Sarma, Phys. Rev. B 33, 5401 (1986). 30 C. Kittel, Introduction to Solid State Physics (Wiley, New York, 2005). 31 S. Das Sarma, E. H. Hwang, and A. Kaminski, Phys. Rev. B 67, 155201 (2003). 32 Fan Zhang, Hongki Min, Marco Polini, and A. H. Mac- Donald, Phys. Rev. B 81, 041402(R) (2010).
1903.00989	1	1903	2019-03-03T21:04:42	Hot electrons modulation of third harmonic generation in graphene	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	Hot electrons dominate the ultrafast ($\sim$fs-ps) optical and electronic properties of metals and semiconductors and they are exploited in a variety of applications including photovoltaics and photodetection. We perform power-dependent third harmonic generation measurements on gated single-layer graphene and detect a significant deviation from the cubic power-law expected for a third harmonic generation process. We assign this to the presence of hot electrons. Our results indicate that the performance of nonlinear photonics devices based on graphene, such as optical modulators and frequency converters, can be affected by changes in the electronic temperature, which might occur due to increase of absorbed optical power or Joule heating.	cond-mat.mes-hall	cond-mat	Hot electrons modulation of third harmonic generation in graphene G. Soavi1∗ , G. Wang1, H. Rostami2, A. Tomadin3, O. Balci1, I. Paradisanos1, E.A.A. Pogna4, G. Cerullo4, E. Lidorikis5, M. Polini3, A. C. Ferrari1† 1 Cambridge Graphene Centre, University of Cambridge, Cambridge CB3 0FA, UK 2 Nordic Institute for Theoretical Physics, Roslagstullsbacken 23 SE-106 91 Stockholm, Sweden 3 Istituto Italiano di Tecnologia, Graphene Labs, Via Morego 30, I-16163 Genova, Italy 4IFN-CNR, Dipartimento di Fisica, Politecnico di Milano, P.zza L. da Vinci 32, 20133 Milano, Italy , 5 Department of Materials Science and Engineering, University of Ioannina, Ioannina 45110, Greece Hot electrons dominate the ultrafast (∼fs-ps) optical and electronic properties of metals and semi- conductors and they are exploited in a variety of applications including photovoltaics and photode- tection. We perform power-dependent third harmonic generation measurements on gated single-layer graphene and detect a significant deviation from the cubic power-law expected for a third harmonic generation process. We assign this to the presence of hot electrons. Our results indicate that the performance of nonlinear photonic devices based on graphene, such as optical modulators and fre- quency converters, can be affected by changes in the electronic temperature, which might occur due to increase of absorbed optical power or Joule heating. For a free electron gas at thermal equilibrium, the av- erage occupation number at energy E is described by the Fermi-Dirac distribution f (E)[1]: f (E) = 1 e(E−µ)/kB T0 + 1 , (1) where µ is the chemical potential and kB is the Boltz- mann constant. At zero temperature, µ equals the Fermi energy (EF ). At thermal equilibrium Te = Tl = T0, with Te the electronic temperature, Tl the lattice tem- perature and T0 the ambient temperature. Photoexcita- tion of a sample with ultrashort (∼fs-ps) pulses creates a non-thermal regime, i.e. a condition where the elec- tron population cannot be defined by f (E) and Te, which rapidly evolves through electron-electron (e-e) scattering into a hot-carrier distribution, with Te > Tl[2 -- 6]. Elec- trons then transfer energy to the lattice through scatter- ing with phonons (ph) until Te = Tl[4 -- 8]. Equilibrium with the surrounding environment is then reached via ph-ph scattering[4 -- 6, 9 -- 13]. The timescale of these scat- tering processes depends on the system under investiga- tion and the excitation energy. Typical values for metals (e.g. Au, Ag, Cu, Ni[3, 9 -- 11]) and semiconductors (e.g. Si[7]) are∼10fs-1ps for e-e scattering[3],∼1-100ps for e-ph scattering[7, 8], and>100ps for ph-ph scattering[9 -- 11]. Hot electrons (HEs) can be exploited to enhance the ef- ficiency of photocatalysis[14], photovoltaic devices[15, 16] and photodetectors[13]. The efficiency of photovoltaic devices can be enhanced if HEs are collected before re- laxation with ph[16], when the absorbed light energy is transferred to the lattice instead of being converted into an electrical signal. Photodetectors based on the See- beck effect[17] and Schottky junctions[18] both exploit HEs. These also play a key role in nonlinear effects, e.g. in Second Harmonic Generation (SHG)[19] and in Third Harmonic Generation (THG)[20]. Following interaction with photons with energy ¯hω0, where ¯h is the reduced Planck constant and ω0 is the photon angular frequency, new photons can be generated inside a nonlinear material at energies 2¯hω0 for SHG[19] or 3¯hω0 for THG[20]. In the scalar form, the SHG and THG optical electric field Emω0 can be written as[21, 22]: Emω0 = gχ(m)(ω0, EF , Te)E m ω0 , (2) where Eω0 is the incident electric field, m = 2 for SHG and m = 3 for THG, g is a function of the material's refractive index (n) and ω0, and χ(m) is the material's nonlinear susceptibility. g and χ(m) depend on material, angle and polarization of the incident light and on m[21 -- 23]. E.g., the THG field for a bulk sample for normal incidence and constant incident power is[21 -- 23]: E3ω0 = 1 4 i3ω0 2n3ω0c dχ(3)E 3 ω0 , (3) where d is the material's thickness. The light intensity (Imω0 in units of W/m2) is related to the optical elec- tric field by Imω0 = nmω0ǫ0cEmω02/2[21 -- 23]. Eq.(2) highlights two aspects of harmonic generation: (i) the SHG/THG electric field scales with the square/cube of Eω0 and, as a consequence, one would expect Imω0 ∝ I m ω0 ; (ii) SHG/THG intensities depend on the linear (e.g. ab- sorption) and nonlinear (through the nonlinear suscepti- bilities χ(2) and χ(3)) properties of the material[24, 25]. Both g and χ(m) are functions of Te, and thus modify the power-law relation between Imω0 and I m ω0 . The role of HEs in nonlinear optics was investigated for SHG in metals[10, 25 -- 29] and semiconductors[30, 31] but, to the best of our knowledge, has not been considered thus far for THG in any material. HEs play also a key role in the ultrafast (fs-ps)[5, 6, 32 -- 34] and nonlinear[35 -- 37] properties of single-layer graphene (SLG). In SLG e-e scattering occurs within few tens of fs after photoexcitation[5], while e-ph scat- tering takes place on a ∼ps timescale[5, 34, 38]. HEs can be exploited for the development of optoelectronic devices based on graphene[39 -- 41]. E.g., a SLG p-n junc- tion can be used as a photothermal detector because, following optical excitation, the photo-thermoelectric (or Seebeck) effect (PTE) will produce a voltage VP T E = (S1 − S2)∆Te, where S1,2 (in V K −1) are the thermoelec- tric powers (or Seebeck coefficients) and ∆Te is the HEs temperature difference in the two SLG regions[42, 43]. HEs in SLG can recombine radiatively to give broad- band emission[44 -- 49] and the timescale/mechanism of the HEs relaxation has implications for the use of SLG in mode-locked lasers[39, 41, 50]. SLG can be used to fabricate broadband and gate-tuneable optical fre- quency converters[40, 51 -- 53]. However, in these de- vices the high Te (∼ 103K) induced by the optical excitation[5, 46, 51] can significantly modify (e.g. by re- ducing the THG efficiency, THGE, defined as the ratio between the THG and incident intensities) the SLG non- linear optical response[51]. Here we demonstrate that for THG in SLG the cubic dependence I3ω0 ∝ I 3 ω0 [21, 22] fails when Te >> Tl is taken into account. We show that, more generally, THG follows a power-law I3ω0 ∝ I x ω0 , with the exponent x de- pendent on EF . This strong dependence of Te, and thus of THGE, over both EF and Iω0 has strong impact on the performance of nonlinear photonic devices based on SLG, such as optical switches and frequency converters. We use Chemical Vapor Deposition (CVD) SLG trans- ferred on Fused Silica (FS) and gated by ionic liquid (IL), Fig.1a. SLG is grown on Cu (99.8% pure, 25µm thick), as for Ref.[54]. This is then transferred on FS by polymer-assisted Cu wet etching[55], using polymethyl methacrylate (PMMA). SLG is characterized by Ra- man spectroscopy with a Renishaw inVia spectrometer. The 514nm Raman spectrum of SLG after transfer is shown in Fig.1b. The 2D peak is a single Lorentzian with full width at half maximum FWHM(2D)∼36cm−1, a signature of SLG[56]. The position of the G peak, Pos(G), is∼1599cm−1, with FWHM(G)∼13cm−1. The 2D peak position is Pos(2D)∼2696cm−1, while the 2D to G peak intensity and area ratios, I(2D)/I(G) and A(2D)/A(G), are ∼1.7 and ∼4.67, indicating a p- doping∼250-300meV[57, 58]. The absence of the D peak shows that there are no significant defects. In order to gate the SLG, we fabricate source and drain contacts by evaporating 7nm/70nm Cr/Au. Cr is used to improve Au adhesion. We etch the SLG outside the channel using an oxygen plasma. As gate electrode we use 7nm/70nm Cr/Au on a 1mm thick microscope slide. During evap- oration, we cover part of the slide to have a trans- parent region∼1cm2 for optical measurements. We use 50µm double-sided tape as a spacer between gate elec- trode and SLG. We then align the SLG channel and the non-evaporated window on the gate electrode and place the IL, Diethylmethyl(2-methoxyethyl)ammoniumbis- (triflouromethylsulfonyl)imide (C6H20F6N2O5S2), be- tween SLG and the gate electrode. 2 (a) (b) FIG. 1. a) Schematic of THG device. EF tuning is obtained by IL top-gating. Measurements are performed in transmis- sion. b) 514nm Raman spectra of SLG after transfer on FS (blue), SLG top-gated device (red) and IL (black) FWHM(2D)=32cm−1, The 514nm Raman spectra of IL and of SLG at a gate voltage VG=0V are shown in Fig.1b. For SLG, Pos(G) is∼1587cm−1, with FWHM(G)∼14cm−1. Pos(2D)∼2691cm−1, with I(2D)/I(G) and A(2D)/A(G)∼2.9 and ∼5.9, respec- tively, indicating a p-doping∼200meV[57]. Figs.2a,b plot the Raman and transmission spectra as a function of VG from 0.5 to -1.5V with steps of 0.1V for a source-drain voltage VSD=0.2V. From the Raman spectra at different VG we estimate EF . This is done by monitoring the evolution of Pos(G) as a function of VG, as shown in Fig.3[57, 58]. The relation between EF and VG can also -1.4V -1.2V -1V -0.8V -0.6V -0.4V -0.2V 0V 0.2V 0.4V 1600 2600 2700 Raman Shift (cm-1) ) 1 - m c ( ) G ( s o P 1620 1615 1610 1605 1600 1595 1590 1585 1580 (a) 2800 background (b) -651 -620 -611 -569 -528 -501 -456 EF (meV) -244 -266 -200 -329 -354 -411 -180 -168 -137 -110 -66-70 3 0 9 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 VG (V) . ) . u a ( y t i s n e n t I 1500 100 % ) ( T 99 -0.6V-0.8V -1V -1.2V -1.4V 98 0.8 1.0 1.2 1.4 Photon Energy (eV) FIG. 2. a) Raman; b) transmission spectra of SLG top-gated device at different VG. The background (100%) for the trans- mission spectra is defined as the transmission of the device without SLG. be derived from the transmission measurements. For each VG, we measure both transmission, Fig.2b, and source-drain current ISD, Fig.4(b) (red circles). The transmission of the gated device never reaches 100%, this being defined as the transmission of the device without SLG. This non-saturable residual absorption (αres) of SLG[59] originates from intra-band electronic transitions, enabled by disorder[59]. From Fig.2b we get αres ∼ 0.2 − 0.4%, by taking the difference between the background (grey curve) and the SLG transmission at 0.8eV for VG = −1.4V . The transition from intra- to inter-band absorption, at Te = 0, occurs when the energy of the photons is ¯hωT = 2EF . We thus estimate ¯hωT from the half-maximum of each transmission curve and calculate EF = ¯hωT /2, as in Fig.4(a) (red circles). This estimate is in good agreement with that derived from the Raman analysis (blue circles in Fig.4a). THG measurements are then performed at room tem- perature (RT). We excite the sample with the idler beam FIG. 3. Pos(G) as a function of VG from the Raman mea- surements in Fig.2a. EF (top horizontal axis) is obtained as detailed in Ref.[57]. of an Optical Parametric Oscillator (OPO, Coherent) at 0.69eV (∼1.8µm) pumped by a mode-locked Ti:Sa laser (Coherent) with 150fs pulse duration, 80MHz repetition rate and 4W average power at 1.55eV. The OPO idler is focused by a 40X reflective objective (Ag coating, numer- ical aperture NA=0.5) to avoid chromatic aberrations. The THG signal is collimated by an 8mm lens and deliv- ered to a spectrometer (Horiba iHR550) equipped with a nitrogen cooled Si charged-coupled-device (CCD). The idler spot-size is∼4.7µm, the pulse duration∼300fs and the polarization is linear. We use a Keithley 2612B dual channel Source Measure Unit both to apply VG and VSD and to read ISD. VG is tuned between -1.5 and +0.5V while VSD is kept at 0.2V. For THG measurements we proceed we tune VG (10 points between -1.5 and +0.5V) and scan the power (7 points between 1 and 4 mW). The incident excitation power is estimated at the sam- ple position by considering the losses of the objective, by measuring the power before and after the objective when the sample is removed. For each power (at a fixed VG), we measure the THG signal by using 10s acquisi- tions and 3 accumulations. Thus, SLG is kept at a given VG for 210s before moving to next VG. During THG experiments we also measure ISD. By comparing the transconductance (ISD as a function of VG) during the transmission, Fig.4(b) (red curve), and THG measure- ments, Fig.4(b) (black curve), we observe an increase in SLG doping. We thus estimate EF during THG exper- iments based on ISD, Fig.4a (black curve). In order to estimate the emitted THG power, we take into consider- ation the losses of the system. The major ones are the absorption of the device without SLG (FS substrate and 100 90 ) % ( T 80 70 (a) 0.4 (b) 4 FS substrate top-gated device 0.8 1.0 1.2 1.4 Photon Energy (eV) -1.6 -1.2 -0.8 -0.4 VG (V) 0.0 ) V e ( F E ) A m ( D S I 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 0.3 0.2 0.1 0.0 FIG. 5. UV-VIS transmission curves for the device substrate (FS) and the final device (IL and FS) without SLG. The IL shows an absorption peak at ∼0.6eV. cal conductivity tensor, calculated through a diagram- matic technique, with the light-matter interaction in the scalar potential gauge in order to capture all intra-, in- terband and mixed transitions[36, 51]. According to the C6v point group symmetry of SLG on a substrate, the relative angle between laser polarization and the SLG lattice is not important for the third-order response[51]. Thus, we assume the incident polarization, ℓ, to lie along the zigzag direction of the lattice, x, without loss of generality[51]. For IL we use n1(ω0) ∼1.44[61] and for FS n2(3ω0) ∼1.42[61]. At first sight, Eq.(4) predicts a cubic dependence I3ω0 ∝ I 3 ω0 . However, I3ω0 is modu- lated also by σ(3) ℓℓℓℓ, which is a function of ω0, EF and Te. The first two parameters, ω0 and EF , can be controlled by tuning the excitation photon energy and by applying an external VG. On the other hand, Te cannot be di- rectly controlled by an external input, and its value is affected by the amount of energy that is transferred from light to the SLG electrons. Te can be calculated from the Boltzmann equation, taking into account the role of intra- and inter-band e-e scattering and the population of the optical phonon modes[6]. An estimate can also be obtained with the following approach[51]. When a pulse of duration ∆t and fluence F [Jm−2] photoexcites SLG, an average power per unit area P/A = (α + αres)F /∆t is absorbed by the electronic system, where α is the sat- urable SLG absorption, due to inter-band electronic tran- sitions. α is a function of ω0, the chemical potentials in the conduction and valence bands (µc and µv) and Te. The variation dU of the energy density in a time inter- val dt is dU = (P/A)dt. The corresponding Te increase is dTe = dU/cv, where cv is the electronic heat capacity -1.5 -1.0 -0.5 0.0 VG (V) 0.5 1.0 FIG. 4. a) EF as a function of VG obtained from Raman analysis as in Fig.3 (blue dots) and from transmission mea- surements in Fig.2b (red dots). The black dots are EF during THG experiments calculated from ISD. b) ISD as a func- tion of VG before (red dots) and during (black dots) THG experiments. IL), the grating efficiency, and the CCD quantum effi- ciency. We also consider the CCD gain. The transmission of the FS substrate is∼93%, Fig.5. The IL transmission is frequency dependent, Fig.5 (red curve). We use the spectrometer specs[60] to estimate losses due to grating and CCD efficiencies. We account for the∼7 CCD gain, i.e. the number of electrons necessary for 1 count[60]. The THG intensity I3ω0 under normal incidence can be written as[51]: I3ω0 = f (ω0) I 3 ω0 4ǫ4 0c4 (cid:12) (cid:12) (cid:12) (cid:12) σ(3) ℓℓℓℓ(ω0, EF , Te) (cid:12) (cid:12) 2 , (4) where ǫ0 ∼ 8.85 × 10−12C(Vm)−1 and c = 3 × 108m/s are the vacuum permittivity and the speed of light; f (ω0) = n−3 1 (ω0)n2(3ω0)[n1(3ω0) + n2(3ω0)]−2 in which ni=1,2(ω) is the IL refractive index (i = 1) and substrate (i = 2). σ(3) ℓℓℓℓ is the SLG third-order nonlinear opti- ) K ( e T (a) 1600 1200 800 400 0 1000 1 2 Power 3 4 (mW) 2 ) 3 ( 100 0 15 12 ) W f ( 3 r e w o P 9 6 3 0 -0.2eV -0.3eV -0.5eV -0.7eV -0.8eV 5 (b) 6 5 (a) -0.25eV -0.3eV -0.53eV -0.7eV -0.8eV 1.0 2.0 1.5 Power 2.5 (mW) 3.0 3.2 t 0.2% 0.4% / ) 3 ( 10 1 0 500 1000 Te (K) 1500 2000 FIG. 6. a) Te from Eq.(6) as a function of the incident power for ¯hω0=0.69eV and different EF . b) σ(3)/σ(3) 0 2 as a function of Te for ¯hω0=0.69eV and different EF in a single-chemical potential model of the photoexcited SLG. When the pulse is off, Te re- laxes towards T0 on a time-scale τ . This reduces Te by dTe = −(Te/τ )dt in a time interval dt. Thus[51]: dTe dt = α + αres cv F ∆t − Te − T0 τ , (5) If the pulse duration is: (i) much longer than∼20fs, i.e. the time-scale for the e distribution to relax to the Fermi- Dirac profile in both bands[5, 62]; (ii) comparable to the time-scaleτ ∼ 100 − 200f s needed to heat the optical ph modes[4, 5, 62], the electronic system reaches a steady- state during the pulse, with Te obtained from Eq.(5): Te = T0 + τ α + αres cv F ∆t . (6) Fig.6(a) plots Te from Eq.(6) for our experimental con- ditions: excitation power∼0.5mW to 5mW, EF ∼-0.8 to -0.2eV, ¯hω0=0.69eV, T0=300K, τ =100fs, αres = 0.4% and ∆t=300fs. An increase of excitation power induces an increase of Te, thus a modulation of σ(3). The increase in Te is also modulated by changes in EF , as this affects α of SLG (Fig.2b). Fig.6 shows the Te dependence of n e n o p x e G H T 2.8 2.4 2.0 -1.0 -0.8 (b) -0.6 EF (eV) -0.4 -0.2 FIG. 7. a) 3¯hω0 power as a function of incident power for different EF . The dotted lines are obtained from the power- law y = a · xb, with a and b fitting parameters. b) THG exponent from fitting (y = a · xb) the power-dependent THG (black points) and from theory (dotted-lines) for different αres and ¯hω0 = 0.69eV . 0 = Nf e4¯hv2 0 2, with σ(3) σ(3) in the 0-2000K range and for different EF . Fig.6b plots σ(3)/σ(3) F /[32π(1eV)4][51]. The quantity Nf = 4 is the number of fermion flavors in SLG and vF ∼ 106m/s is the Fermi velocity, thus σ(3) 0 ∼ 4.2 × 10−24Am2/V 3[51]. Fig.6b shows that, de- pending on EF , σ(3) will either increase (e.g. EF =-0.2eV in Fig.6b) or decrease (e.g. EF =-0.5 to -0.8eV in Fig.6b) with increasing Te. This results in a deviation from the cubic dependence I3ω0 ∝ I 3 ω0 . Fig.7(a) plots the experimental THG power depen- dence for ¯hω0=0.69eV. For the same values of incident power we do not detect any THG signal from FS/IL, i.e. outside the area covered by SLG. For a fixed incident power, the THG power increases as we go to more neg- ative values of EF . This EF dependent enhancement of the THG signal arises from logarithmic resonances in the imaginary part of the nonlinear conductivity of SLG due to resonant multiphoton transitions[51]. As seen in Fig.6(b), this leads to a non-monotonic depen- dence of the nonlinear conductivity on Te for different EF . We fit the experimental data relative to our THG power-dependent measurements (circles in Fig.7a) with the power law y = a · xb (dotted lines in Fig.7a), where y is the 3¯hω0 power, x is the incident power and a, b are fitting parameters. Fig.7a shows that the power-law approximation gives excellent fits to the data, if we allow b to depend on EF . Fig.7b plots b (i.e. the THG ex- ponent) from this fit (black circles) as a function of EF . The dotted lines in Fig.7b are the theoretical b (THG ex- ponent) calculated as follows: (i) Te and corresponding chemical potentials in conduction and valence bands as a function of incident power are derived from Eq.(6), for ¯hω0 = 0.69eV and different EF ; (ii) we use these to cal- culate σ(3) as a function of incident power. To this end, we first calculate the Te = 0 expression of the third-order nonlinear conductivity[36] and then utilize the response function in Ref.[6], to express the conductivity at finite T as a weighted integral over EF of the SLG conductivity at Te = 0; (iii) we substitute the calculated σ(3) into Eq.(4) to obtain the theoretical THG intensity; (iv) we fit the THG intensity with y = a · xb. For the estimate of Te we use αres=0.2% and 0.4%, as derived from Fig.2b. We find that the THG exponent varies between∼2 and 3.4, with a non-monotonic dependence on EF and a minimum at EF ∼0.6eV for ¯hω0 = 0.69eV . An increase of the in- cident power affects Te and σ(3). This induces deviations from the cubic power law. To the best of our knowledge, this non-cubic behavior of the THG signal was not re- ported before in SLG or any other material. Most exper- iments on SLG and layered materials took the observa- tion of a cubic power law as a proof of THG[23, 63 -- 65]. In SLG, this cubic dependence was also used to calcu- late χ(3)[23, 64, 65]. This approach has two limitations: 1) the nonlinear susceptibilities are well defined only in three-dimensional materials, since they involve a polar- ization per unit volume[51], thus χ(3) should not be used for SLG; 2) a power-law fit of THG in SLG must take into account EF and Te under the specific experimental con- ditions. In other words, χ(3) in SLG must be calculated as a function of both EF and Te. In summary, hot electrons strongly affect the third- order nonlinear optical response of single-layer graphene and alter the cubic dependence of the third harmonic gen- eration signal and its efficiency. Upon ultrafast (∼100fs) excitation, Te in single-layer graphene can be as high as 103K also when EF > 2¯hω0, due to thermal broaden- ing of the Fermi-Dirac distribution and residual absorp- tion αres. Thus, Te is affected by both EF and Iω0 . Changes in Te can modify the third harmonic generation efficiency and thus the performances of nonlinear pho- tonic and optoelectronic devices, such as optical switches and frequency converters. We acknowledge funding from EU Graphene Flagship, ERC Grant Hetero2D, EPSRC Grants EP/K01711X/1, EP/K017144/1, EP/N010345/1, EP/L016087/1 and the Swedish Research Council (VR 2018-04252). 6 ∗ Present Address: Institut Friedrich-Schiller-Universitat, 07743 Jena † [email protected] fur Festkorperphysik, Max-Wien-Platz 1, [1] C. Kittel, Introduction to Solid State Physics (Wiley New York, New York, 1996). [2] S. Farm, R. Storz, H. K. Tom, and J. Bokor, Phys. Rev. Lett. 68, 2834 (1992). [3] G. Della Valle, M. Conforti, S. Longhi, G. Cerullo, and D. Brida, Phys. Rev. B 86, 155139 (2012). [4] M. Lazzeri, S. Piscanec, F. Mauri, A. C. Ferrari, and J. Robertson, Phys. Rev. Lett. 95, 236802 (2005). [5] D. Brida, A. Tomadin, C. Manzoni, Y. J. Kim, A. Lom- bardo, S. Milana, R. R. Nair, K. S. Novoselov, A. C. Ferrari, G. Cerullo, M. Polini, Nature Commun. 4, 1987 (2013). [6] A. Tomadin, D. Brida, G. Cerullo, A. C. Ferrari, and M. Polini, Phys. Rev. B 88, 035430 (2013). [7] C. V. Shank, R. Yen, C. Hirlimann, Phys. Rev. Lett.50, 454 (1983). [8] R. W. Schoenlein, W. Z. Lin, J. G. Fujimoto, and G. L. Eesley, Phys. Rev. Lett. 58, 1680 (1987). [9] F. Vall´ee Acad. Sci. Paris 2, 1469 (2001). [10] J. Hohlfeld, S.-S. Wellershoff, J. Gudde, U. Conrad, V. Jahnke, and E. Matthias, Chem. Phys. 251, 237 (2000). [11] J. Shah, Ultrafast Spectroscopy of Semiconductors and Semiconductor Nanostructures (Springer Science & Busi- ness Media, New York, 2013). [12] G. Soavi, F. Scotognella, G. Lanzani, and G. Cerullo, Adv. Opt. Mat. 4, 1670 (2016). [13] L. Wei and J. G. Valentine, Nanophot.6, 177 (2017). [14] S. Mukherjee, F. Libisch, N. Large, O. Neumann, L. V. Brown, J. Cheng, J. B. Lassiter, E. A. Carter, P. Nord- lander, N. J. Halas, Nano Lett. 13, 240 (2013). [15] R. T. Ross, A. J. Nozik, J. Appl. Phys. 53, 3813 (1982). [16] W. A. Tisdale, K. J. Williams, B. A. Timp, D. J. Norris, E. S. Aydil, X. Y. Zhu, Science 328, 1543 (2010). [17] J. Stiens, G. Shkerdin, V. Kotov, W. Vandermeiren, C. D. Tandt, G. Borghs, R. Vounckx, Proc. SPIE 6189, 61890Y (2006). [18] F. D. Shepherd, A. C. Yang, R. W. Taylor, Proc. IEEE 58, 1160 (1970). [19] P. A. Franken, A. E. Hill, C. W. Peters, G. Weinreich, Phys. Rev. Lett. 7, 118 (1961). [20] R. W. Terhune, P. D. Maker, C. M. Savage, Phys. Rev. Lett. 8, 404 (1962) [21] Y. R. Shen, The Principles of Nonlinear Optics (John Wiley & Son, New York, 1984). [22] R. W. Boyd, Nonlinear Optics (Academic Press, New York, 2003). [23] N. Kumar, J. Kumar, C. Gerstenkorn, R. Wang, H.-Y. Chiu, A. L. Smirl, and H. Zhao, Phys. Rev. B 87, 121406 (2013). [24] W. K. Burns, N. Bloembergen, Phys. Rev. B 4, 3437 (1971). [25] J. Hohlfeld, D. Grosenick, U. Conrad, E. Matthias, Appl. 7 Phys. A 60, 137 (1995). [26] C. Guo, G. Rodriguez, A. J. Taylor, Phys. Rev. Lett. 86, 1638 (2001). [47] C. F. Chen, C. H. Park, B. W. Boudouris, J. Horng, B. Geng, C. Girit, A. Zettl, M. F. Crommie, R. A. Segalman, S. G. Louie, F. Wang, Nature 471, 617 (2011). [27] J. Hohlfeld, U. Conrad, E. Matthias, Appl. Phys. B 63, [48] R. J. Stoehr, R. Kolesov, J. Pflaum, J. Wrachtrup, Phys. 541 (1996). Rev. B 82, 121408(R) (2010). [28] K. L. Moore, T. D. Donnelly, Opt. Lett. 24, 990 (1999). [29] N. Papadogiannis, S. Moustaizis, Opt. Commun. 137, [49] W.-T. Liu, W. Wu, P. J. Schuck, M. Salmeron, Y. R. Shen, F. Wang, Phys. Rev. B 82, 081408(R) (2010). 174 (1997). [30] H. W. K. Tom, G. D. Aumiller, C. H. Brito-Cruz, Phys. Rev. Lett. 60, 1438 (1988). [31] P. Saeta, J. K. Wang, Y. Siegal, N. Bloembergen, E. Mazur, Phys. Rev. Lett. 67, 1023 (1991). [32] R. Bistritzer, A. H. MacDonald, Phys. Rev. Lett. 102, 206410 (2009). [33] A. C. Betz, F. Vialla, D. Brunel, C. Voisin, M. Picher, A. Cavanna, A. Madouri, G. F´eve, J. M. Berroir et al., Phys. Rev. Lett. 109, 056805 (2012). [34] K.J. Tielrooij, J. Song, S. A. Jensen, A. Centeno, A. Pesquera, A. Z. Elorza, M. Bonn, L. Levitov, F. Koppens, Nature Phys. 9, 248 (2013). [35] S. A. Mikhailov, Phys. Rev. B 93, 085403 (2016). [36] H. Rostami, M. Polini, Phys. Rev. B 93, 161411 (2016). [37] J. L. Cheng, N. Vermeulen, J. E. Sipe, New J. Phys. 16, 053014 (2014). [38] M. Lazzeri, S. Piscanec, F. Mauri, A. C. Ferrari, J. Robertson, Phys. Rev. B 73, 155426 (2006). [39] F. Bonaccorso, Z. Sun, T. Hasan, A. C. Ferrari, Nature Photon. 4, 611 (2010). [40] M. Romagnoli, V. Sorianello, M. Midrio, F.H.L. Kop- pens, C. Huyghebaert, D. Neumaier, P. Galli, W. Templ, A. D'Errico, A.C. Ferrari, Nature Reviews Materials, 3, 392-414, (2018) [41] A. C. Ferrari, F. Bonaccorso, V. Fal'ko, K. S. Novoselov, S. Roche, P. Boggild, S. Borini, F. H. L. Koppens, V. Palermo, N. Pugno, et al., Nanoscale 7, 4598 (2015). [42] N. M. Gabor, J. C. W. Song, Q. Ma, N. L. Nair, T. Tay- chatanapat, K. Watanabe, T. Taniguchi, L. S. Levitov, P. Jarillo-Herrero, Science 334, 648 (2011). [43] F. H. L. Koppens, T. Mueller, P. Avouris, A. Ferrari, M. Vitiello, M. Polini, Nature Nanotech. 9, 780 (2014). [44] M. Freitag, H.Y. Chiu, M. Steiner, V. Perebeinos, P. Avouris, Nature Nanotech. 5, 497 (2010). [45] Y. D. Kim, H. Kim, Y. Cho, J. H. Ryoo, C.-H. Park, P. Kim, Y. S. Kim, S. Lee, Y. Li et al. Nature Nanotech. 10, 676 (2015). [46] C. H. Lui, K. F. Mak, J. Shan, and T. F. Heinz, Phys. Rev. Lett. 105, 127404 (2010). [50] Z. Sun, T. Hasan, F. Torrisi, D. Popa, G. Privitera, F. Wang, F. Bonaccorso, D. M. Basko, A. C. Ferrari, ACS Nano 4, 803 (2010). [51] G. Soavi, G. Wang, H. Rostami, D. G. Purdie, D. De Fazio, T. Ma, B. Luo, J. Wang, A. K. Ott et al., Nature Nanotech. 13, 583 (2018). [52] T. Jiang, D. Huang, J. Cheng, X. Fan, Z. Zhang, Y. Shan, Y. Yi, Y. Dai, L. Shi et al., Nature Phot. 12, 430 (2018). [53] K. Alexander, N. A. Savostianova, S. A. Mikhailov, B. Kuyken, D. Van Thourhout, ACS Phot. 4, 3039 (2017). [54] X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Velamakanni, I. Jung et al., Science 324, 1312 (2009). [55] F. Bonaccorso, A. Lombardo, T. Hasan, Z. Sun, L. Colombo, A. C. Ferrari, Materials Today 15, 564 (2012). [56] A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov et al., Phys. Rev. Lett. 97, 187401 (2006). [57] A. Das, S. Pisana, B. Chakraborty, S. Piscanec, S. K. Saha, U. V. Waghmare, K. S. Novoselov, H. R. Krish- namurthy, A. K. Geim et al., Nature Nanotech. 3, 210 (2008). [58] D. M. Basko, S. Piscanec, A. C. Ferrari, Phys. Rev. B 80, 165413 (2009). [59] K. F. Mak, L. Ju, F. Wang, T. F. Heinz, Solid State Comms 152, 1341 (2012). [60] Symphony II 1024 x 256 Cryogenic Open-Electrode CCD Detector Quantum Efficiency and grating (300 gr/mm, Blazed 600 nm 510 19 140) relative efficiency from www.horiba.com. [61] https://www.sigmaaldrich.com. [62] M. Breusing, S. Kuehn, T. Winzer, E. Mali´c, F. Milde, N. Severin, J. P. Rabe, C. Ropers, A. Knorr, T. Elsaesser, Phys. Rev. B 83, 153410 (2011). [63] R. Wang, H. C. Chien, J. Kumar, N. Kumar, H.-Y. Chiu, H. Zhao, ACS Appl. Mat. Int. 6, 314 (2014). [64] S. Y. Hong, J. I. Dadap, N. Petrone, P. C. Yeh, J. Hone, R. M. Osgood, Phys. Rev. X 3, 021014 (2013). [65] R. I. Woodward, R. T. Murray, C. F. Phelan, R. E. P. de Oliveira, T. H. Runcorn, E. J. R. Kelleher, S. Li, E. C. de Oliveira, G. J. M. Fechine et al. 2D Materials 4, 011006 (2017).
1610.05526	1	1610	2016-10-18T10:44:35	Controlled generation of a pn-junction in a waveguide integrated graphene photodetector	[ "cond-mat.mes-hall" ]	With its electrically tunable light absorption and ultrafast photoresponse, graphene is a promising candidate for high-speed chip-integrated photonics. The generation mechanisms of photosignals in graphene photodetectors have been studied extensively in the past years. However, the knowledge about efficient light conversion at graphene pn-junctions has not yet been translated into high-performance devices. Here, we present a graphene photodetector integrated on a silicon slot-waveguide, acting as a dual-gate to create a pn-junction in the optical absorption region of the device. While at zero bias the photo-thermoelectric effect is the dominant conversion process, an additional photoconductive contribution is identified in a biased configuration. Extrinsic responsivities of 35 mA/W, or 3.5 V/W, at zero bias and 76 mA/W at 300 mV bias voltage are achieved. The device exhibits a 3 dB-bandwidth of 65 GHz, which is the highest value reported for a graphene-based photodetector.	cond-mat.mes-hall	cond-mat	Controlled generation of a pn-junction in a waveguide integrated graphene photodetector Simone Schuler1,, Daniel Schall2, Daniel Neumaier2, Lukas Dobusch1, Ole Bethge3, Benedikt Schwarz3, Michael Krall1, and Thomas Mueller1, 1 Vienna University of Technology, Institute of Photonics, Gusshausstrasse 27-29, 1040 Vienna, Austria 2 AMO GmbH, Otto-Blumenthal-Strasse 25, 52074 Aachen, Germany 3 Vienna University of Technology, Institute of Solid State Electronics, Floragasse 7, 1040 Vienna, Austria *Corresponding authors: [email protected], [email protected] With its electrically tunable light absorption and ultrafast photoresponse, graphene is a promising candidate for high-speed chip-integrated photonics. The generation mechanisms of photosignals in graphene photodetectors have been studied extensively in the past years. However, the knowledge about efficient light conversion at graphene pn-junctions has not yet been translated into high-performance devices. Here, we present a graphene photodetector integrated on a silicon slot-waveguide, acting as a dual-gate to create a pn-junction in the optical absorption region of the device. While at zero bias the photo-thermoelectric effect is the dominant conversion process, an additional photoconductive contribution is identified in a biased configuration. Extrinsic responsivities of 35 mA/W, or 3.5 V/W, at zero bias and 76 mA/W at 300 mV bias voltage are achieved. The device exhibits a 3 dB-bandwidth of 65 GHz, which is the highest value reported for a graphene-based photodetector. Keywords: graphene, photodetector, photo-thermoelectric effect, integrated photonics The dense integration of photonic components on a silicon (Si) chip allows for a dramatic increase of the performance of optical communication systems at reduced cost. Photodetectors convert light into electrical signals and are at the heart of any optical link. In silicon photonics, traditionally germanium (Ge) [1,2] or III-V compound semiconductors [3,4] are used for photodetection and both technologies have reached a high level of maturity. Nevertheless, the direct monolithic integration of III-V photodetectors on Si wafers remains a challenge because of the large lattice constant mismatch and different thermal coefficients. Ge can be directly grown on crystalline Si, but pushing the bandwidth of Ge photodetectors to higher and higher frequencies [5,6,7] becomes increasingly difficult because of the material's poor electrical quality. One of the most promising routes to a new era of chip performance is the monolithic 3D integration of electronic and photonic components on the same chip, where a promising material for the photonic layers is SiN [8]. On these amorphous SiN layers, crystalline Ge (the prerequisite for realizing high performance photodetectors) cannot be grown, leaving graphene as the currently only SiN-compatible material that has the potential to enable high-speed photodetection [9]. light absorption Recently, graphene [10] has emerged as an attractive material in photonics because of its ultra-broadband [11,12,13], high carrier mobility [14], and gate-tunability of the optical absorption [15,16,17,18,19]. Moreover, it can be integrated onto virtually any waveguide material, including Si, SiN or AlN. Already early studies have revealed an ultrafast photoresponse in graphene with an intrinsic bandwidth of 260 GHz [20,21,22], showing the promise of graphene for high-speed photodetection applications. Since then, several graphene-based photodetectors on Si waveguides [23,24,25,26,27] have been realized. Devices with a 3-dB cutoff at 42 GHz [26] and detection of a 50 Gbit/s data stream [27] have been demonstrated, on par with state-of-the-art Ge detectors, but the performance of graphene based detectors suffer from a limited controllability of the Fermi level and thus of the photoresponse. In order to further improve graphene-based photodetectors especially in terms of efficiency, improved device concepts are necessary which allow exploiting graphene's remarkable properties. Figure 1. (a) Sketch of the graphene photodetector based on a slot-waveguide. Inset: SEM image and dimensions of the slot waveguide. (b) Colored SEM picture of a device on Al2O3 and mode coupler (inset). (c) Contact pads for the Si slabs, which are used as gate electrodes. Scale bar, 12 (cid:80)m. (d) FEM simulation result for the mode coupler, which couples light from a strip waveguide into the slot-waveguide. (e) Electric field distribution of the TE-mode in the slot-waveguide. In this letter, we present a Si slot-waveguide integrated graphene photodetector relying on the photo-thermoelectric (PTE) effect. The PTE effect plays an important role in graphene-based photodetectors because of the large Seebeck coefficient [28]. In addition the large optical phonon energy (~0.2 eV) [29] and the low scattering rate via acoustic phonons [30,31] give rise to an increased temperature of the photoexcited carriers for some picoseconds, while the lattice stays close to room temperature. As the electronic response is generated from hot electrons, large bandwidths can be achieved [32]. A photovoltage is generated from hot carriers, if the Seebeck coefficient, governed by the doping, as well as the temperature varies in the graphene sheet. The PTE effect [28] has been shown to be dominant in monolayer-bilayer graphene junctions [33], metal-graphene interfaces [34,35,36,37] and partially suspended graphene [38] without external bias. Under bias, photoconductive and bolometric effects also contribute to the photoresponse [39]. The slot-waveguide geometry employed in this work has a twofold function. First, the two silicon strips of the slot-waveguide are used as local gate electrodes to create a controlled and tunable pn-junction in the graphene absorption region. These static gates do not contribute to the radio-frequency (RF) capacitance and therefore leave the bandwidth of the photodetector unaffected. Second, the light is strongly confined in the slot and is hence absorbed by the graphene exactly in between the p-doped and n-doped regions, maximizing the photoresponse due to the PTE effect. A sketch of the device concept is shown in Figure 1(a). We used finite element (FEM) simulations to design the passive photonic structures. The waveguide consists of two strips of a high refractive index material (Si), separated by a subwavelength low refractive (air) slot [40]. The optical mode profile in the slot-waveguide is depicted in Figure 1(e). To efficiently couple light from an optical fiber, the light is first coupled via a grating coupler into a conventional strip waveguide and then adiabatically transferred with high efficiency into the slot-waveguide via a mode converter (Figure 1(d)). Additional information on the grating coupler and the mode converter are provided in the Supporting Information. The waveguides were fabricated on a silicon-on-insulator (SOI) wafer with 220 nm thick Si device layer and 3 (cid:80)m buried oxide. 𝐿𝐺= 250 nm wide Si strips with a gap of 𝐿𝑆= 80 nm were defined using and 1(b), respectively. Electrical contacts with 𝐿= 4 (cid:80)m drain-source separation were electron-beam lithography and reactive ion etching. Scanning electron microscopy (SEM) pictures of the waveguide and the mode converter are shown in the insets of Figures 1(a) fabricated using electron-beam lithography and metallization (5 nm titanium (Ti), 60 nm gold (Au)). The waveguides were contacted with Ti/Au pads, with additional test pads to verify the ohmic contacts to the Si strips (Figure 1(c)). The silicon waveguides were p-type doped with a resistivity of 14–22 (cid:58)cm. 𝑑𝐺= 10 nm thick layer of alumina (Al2O3) was grown by means of atomic layer To prevent electrical contact between the graphene and the waveguide strips, either a deposition (ALD) or hexagonal boron nitride (hBN) of similar thickness was transferred onto the Si waveguide structures. Graphene of proper size and monolayer thickness was prepared by mechanical exfoliation on a stack of polymers on a sacrificial Si wafer. The polymer stack consisted of PAA (poly acrylic acid) and PMMA (poly methyl methacrylate acid) with a thickness chosen such that flakes could be identified by optical microscopy. The monolayer thickness of the graphene was verified using Raman spectroscopy [41] (see Supporting Information). The PAA-PMMA stack was then put into water to dissolve the PAA, thus the PMMA layer with the graphene on top was released from the Si wafer. Afterwards, the PMMA film was positioned on a PDMS (poly-dimethyl-siloxane) stamp, turned upside down and placed with micrometer precision over the slot-waveguide. Care was taken to avoid additional placement of graphitic chunks on top of the photonic structures. Therefore, an aperture was defined from PMMA in a previous step. The PMMA carrier substrate, used for the transfer, was simultaneously used to fabricate electrical contacts to the graphene via electron-beam lithography. Critical point drying was employed to avoid damage of the suspended graphene sheet over the slot. In Figure 1(b) a (colored) SEM image of a typical device on Al2O3 is shown. Figure 2. Results for Sample A. (a) Resistance map for varying gate voltages 𝑉𝐺1 and 𝑉𝐺2 applied to drain contacts. (e) Seebeck coefficient as a function of gate voltage (𝑉𝐺1=𝑉𝐺2=𝑉𝐺) as calculated the slot-waveguide. Four characteristic regions can be identified; n-n, n-p, p-p and p-n, indicating the gate-tunability of the carriers. The resistance peak indicates the charge neutrality point (Dirac point) of the device. (b) Measured photovoltage map at zero bias. (c) Sketch of the modeled structure. (d) Calculated electron temperature profile over the distance between the source and from the model and the measured data, plotted as solid and dashed lines, respectively. (f) Calculated resistance map depending on the gate voltages. (g) Corresponding photovoltage map, calculated using the Mott formula. The measured and calculated photovoltage are in good agreement. For electrical and optical characterization, the two silicon gates were wire-bonded while drain and source were contacted using an RF probe in signal-ground (SG) configuration. The devices were characterized electrically by varying the two gate voltages, 𝑉𝐺1 and 𝑉𝐺2, at a fixed drain-source voltage 𝑉𝐷𝑆 and recording the device current 𝐼𝐷𝑆. The resulting resistance map of a 𝑊= 30 µm long monolayer device on 𝑑𝐺 telecom laser diode with a wavelength of 1560 nm (𝐸𝑝ℎ≈ 0.8 eV photon energy), which While varying the gate voltages 𝑉𝐺1 and 𝑉𝐺2, the photovoltage was recorded using a = 13 nm of hBN (sample A) is shown in Figure 2(a). Four characteristic regions can be identified: n-n, n-p, p-p and p-n, which indicate the gate tunability of the carrier density in the graphene sheet. For electro-optical characterization we used chopped light from a was coupled into an optical fiber and further coupled via the grating coupler into the chip. lock-in amplifier. The resulting photovoltage map, presented in Figure 2(b), shows the typical six-fold pattern that serves as a fingerprint for the PTE effect [28,35] (see also Supporting Information). expression for the Seebeck coefficient for further modeling, we approximate the graphene As our data suggest the PTE effect to be the dominant conversion process, we performed simulations based on the model depicted in Figure 2(c). In general, the photovoltage generated from the PTE effect can be calculated by integrating the optically induced temperature gradient ∇𝑇𝑒, with the locally varying Seebeck coefficient, 𝑉𝑃𝑇𝐸= ∫𝑆(𝑥) ∇𝑇𝑒(𝑥) d𝑥. The Seebeck coefficient 𝑆 is related to the electrical conductivity 𝜎 via the Mott equation, 𝑆=−𝜋2𝑘𝐵2𝑇𝑒/(3𝑞) 𝜎−1 𝜕𝜎/𝜕ℇ, where 𝑞 is the electron charge, 𝑘𝐵 denotes the Boltzmann constant and ℇ the Fermi energy [42]. To obtain an analytical conductivity by 𝜎= √𝜎𝑚𝑖𝑛2 +[𝜇𝐶𝐺(𝑉𝐺−𝑉𝐷𝑖𝑟𝑎𝑐)]2 , where 𝜎𝑚𝑖𝑛 denotes the minimum conductivity, 𝜇 is the carrier mobility, 𝐶𝐺=𝜖𝐺/𝑑𝐺 is the gate capacitance and 𝑉𝐷𝑖𝑟𝑎𝑐 the charge neutrality point voltage. Care has to be taken to include the contact resistance 𝑅𝐶 and the excess resistance 𝑅𝑈 of the ungated regions to both sides of the waveguide. The Setting 𝑅𝐶+𝑅𝑈 = 254 Ω results in the expected linear relation between 𝜎 and 𝑉𝐺, as graphene conductivity and obtain 𝜎𝑚𝑖𝑛≈ 0.23 mS, 𝜇≈ 2000 cm2/Vs and 𝑉𝐷𝑖𝑟𝑎𝑐≈ 0.7 V. 𝜎-model, used for further simulations, and the dashed line shows the Seebeck coefficient Based on these values, the Seebeck coefficient was calculated. The result is illustrated in Figure 2(e). The solid line depicts the Seebeck coefficient obtained from the analytical shown in the Supporting Information. From this we can then estimate the intrinsic Figure 2(f) depicts the calculated resistance map as a function of the two gate voltages. overall device conductance can be written 𝐺= 𝑅𝐶+𝑅𝑈+(𝐿/𝑊) 𝜎−1 . 1 directly calculated from the experimental data. The electron temperature 𝑇𝑒 is obtained by solving the heat equation [26] 𝜕2𝑇𝑒𝜕𝑥2+𝛼𝑃′𝜅𝑒−𝑇𝑒−𝑇𝑙 𝐿𝐶2 =0 for the given geometry, where 𝛼𝑃′ denotes the absorbed power density in the slot, 𝑃′= ⁄ , 𝑃𝑖𝑛 is the incident optical power in the waveguide, 𝐿𝐶≈ 1 µm [37] is the 𝑃𝑖𝑛 (𝐿𝑆𝑊) characteristic cooling length in graphene, and 𝑇𝑙 = 300 K the lattice temperature. 𝜅𝑒 is the relation 𝜅𝑒=𝜎𝐿0𝑇 [42], where, for simplicity, we use the standard Lorenz number 𝐿0= 2.4 × 10-8 (V/K)2 and set 𝑇 to 300 K. In order to estimate the absorption 𝛼, we electronic thermal conductivity that can be determined from the Wiedemann-Franz performed FEM simulations (see Supporting Information). The calculated temperature profile is shown in Figure 2(d). We assumed the Seebeck coefficient to be constant over the gated waveguide regions and zero in the ungated regions. The resulting photovoltage map, presented in Figure 2(g), is in good agreement with the measurement, indicating that the PTE effect is indeed the dominant mechanism. Given an incident optical power in the fiber of 𝑃𝑓𝑖𝑏𝑒𝑟≈ 0.8 mW, a 𝜂𝐶≈ 0.4 coupling 𝑅𝑉=𝑉𝑃𝑇𝐸/(𝜂𝐶𝑃𝑓𝑖𝑏𝑒𝑟)≈ 0.9 V/W under zero-bias operation. We also conducted independent measurements of the photocurrent 𝐼𝑃𝑇𝐸 under same conditions using a presented in the Supporting Information), which resulted in a responsivity of 𝑅𝐼≈ 3 mA/W. From the relation 𝑉𝑃𝑇𝐸=𝐼𝑃𝑇𝐸𝑅 we estimate a device resistance of 𝑅≈ 300 (cid:58)(cid:15) in efficiency of the grating coupler (see Supporting Information), and assuming no losses in the 2 mm long waveguides and the mode coupler, we estimate a photoresponsivity of current amplifier with low input impedance (the corresponding photocurrent map is agreement with the electrical I-V measurements. Figure 3. (a) Photocurrent map (Sample A) at 𝑉𝐷𝑆= -400 mV. (b) Responsivity map of our best neutrality point. (c) Responsivity map at 𝑉𝐷𝑆= 300 mV (Sample B). The responsivity increases to 76 device (Sample B) at zero bias, where we achieved a peak responsivity of 35 mA/W. CNP; charge mA/W. Figure 3(a) shows a photocurrent map recorded with a drain-source bias of 𝑉𝐷𝑆= -400 mV. The photoresponse shows a complex behavior with varying 𝑉𝐺1 and 𝑉𝐺2 and is maximum responsivity increases to ~18 mA/W. In order to get an understanding of the dominated by a cross-shaped feature around the Dirac point (see dashed lines). The 𝐼𝑃𝐶≈(𝐿𝑆𝐿)𝑊∆𝜎𝐸 underlying photocurrent generation mechanism, we first performed simulations of the PTE effect that showed a negligible influence of the drain-source voltage. Previous work [39], however, has demonstrated that bolometric (BL) and photoconductive (PC) effects play an important role under bias. We find the BL effect to be inconsistent with the experimental results, both in sign and gate voltage dependence. The PC effect can be modeled by where the prefactor (𝐿𝑆/𝐿) accounts for the fact that only part of the graphene sheet is optically excited. 𝐸 denotes the lateral electric field at the waveguide and ∆𝜎 is the photoexcited conductivity. The latter can be expressed as ∆𝜎=𝑞𝜇(∆𝑛𝑒+∆𝑛ℎ), with ∆𝑛𝑒=∆𝑛ℎ ≈ 𝛼𝑃𝑖𝑛𝜏𝐿/(𝐸𝑝ℎ𝑊𝐿𝑆) being the photo-generated electron and hole densities, respectively, and 𝜏𝐿≈ 2 ps the carrier lifetime in graphene [20]. It is instructive to consider first the case of a homogeneous graphene channel (i.e. 𝑉𝐺1 ~ 𝑉𝐺2 ~ 0) and estimate the expected PC current magnitude. At 𝑉𝐷𝑆= 400 mV, we obtain 𝐸≈ 𝑉𝐷𝑆/𝐿 ≈ 1 kV/cm and 𝑅𝐼≈ 10-2 A/W, in agreement with the experiment (~18 mA/W), which leads us to conclude that the PC effect is indeed dominant. The drop of the PC response with increasing carrier concentration/gate voltage is expected [39] and can be attributed to several factors, such as screening of the lateral electric field, shortening of the carrier lifetime, and enhanced electron-electron scattering. The sign of the photocurrent is (mostly) positive, also consistent with the PC effect. the electron heat is converted into a photovoltage. In devices with Al2O3, on the contrary, the sidewalls of the waveguide contribute to the gating as well, which approximately In an improved set of devices, we replaced the hBN gate dielectric by Al2O3, with a twofold motivation: (i) SEM pictures (see Supporting Information) revealed that the hBN layer does not conformally cover the waveguide. As a result, only a fraction ∝𝐿𝐺/𝐿𝐶 of doubles 𝑉𝑃𝑇𝐸. Additionally, less cooling in the freely suspended graphene sheet leads to higher electron temperatures [43]. (ii) The large series resistance 𝑅𝑈 in devices with hBN severely reduces 𝐼𝑃𝑇𝐸, according to 𝐼𝑃𝑇𝐸=𝑉𝑃𝑇𝐸/𝑅. Using Al2O3, instead, charge transfer 𝑅𝑈. With these modifications, the overall device resistance dropped to ~110 (cid:58) and the (zero-bias) responsivities increased to 𝑅𝑉≈ 3.5 V/W and 𝑅𝐼≈ 35 mA/W, respectively (sample B; Figure 3(b)). From electrical characterization we extracted a mobility of ~1000 between graphene and the oxide leads to p-doping of the ungated regions, which reduces high-responsivity p-n and n-p regimes can still be reached by applying gate voltages cm2/Vs (see Supporting Information). Although 𝑉𝐷𝑖𝑟𝑎𝑐 shifted to ~ 6.2 V, the without breaking the Al2O3 dielectric. Under a moderate drain-source voltage of 𝑉𝐷𝑆 = 300 mV (Figure 3(c)), a peak responsivity of 𝑅𝐼≈ 76 mA/W is achieved. performed impulse response measurements, where ~1 ps long optical pulses, generated (FWHM) of ∆𝑡≈ 30 ps. The electrical bandwidth can then be derived using the time-bandwidth product. Assuming a Gaussian impulse response, 𝑓3𝑑𝐵≈0.44/∆𝑡≈ 15 by a mode-locked erbium fiber laser with a wavelength of 1550 nm, were coupled into the device (sample B) and the impulse response was monitored with an oscilloscope (Figure 4(a)). From this measurement we extract a pulse duration at full-width at half-maximum The electrical bandwidth of a photodetector is a key indicator of its performance. We GHz is obtained, which is the limit of our measurement setup. The Fourier-transform of the pulse, shown in the inset of Figure 4(a), yields the same bandwidth. Figure 4. (a) Measured impulse response of Sample B using a train of ~1 ps pulses. The pulse duration is ∆𝑡≈ 30 ps (FWHM), which corresponds to a bandwidth of ~15 GHz, limited by the frequency response (𝑉𝐷𝑆= 0.5 V) using a heterodyne technique. From this we find a cut-off bandwidth of 𝑓3𝑑𝐵≈ 65 GHz. measurement setup. The Fourier- transformation of the pulse is shown in the inset. (b) Measured In order to determine the cut-off frequency of our detector, we used a heterodyne measurement technique similar to the one used in reference [27]. Two laser sources with different frequencies were multiplexed, causing amplitude beating at the difference frequency. Keeping the frequency of one laser constant while tuning the other, the difference frequency can be varied between 1 and 100 GHz. The laser light was amplified with an erbium-doped fiber amplifier and coupled into the graphene photodetector. A 67 GHz SG-probe was used to contact the device and the output power was detected with an RF power meter. A sketch of the measurement setup is depicted in the Supporting Information. The frequency response of the detector is shown in Figure 4(b). From this measurement we obtain a 3-dB cut-off frequency of 𝑓3𝑑𝐵≈ 65 GHz, independent of bias voltage, which translates into a potential bit rate of ~90 Gbit/s (for a single wavelength channel and on-off keying), and is the highest value reported for a graphene photodetector. The maximum output power measured at 1 GHz was -31 dBm at 19.4 dBm optical input power and 1.2 V bias and defines the highest RF output power delivered by a graphene detector. In summary, we have presented an ultrafast graphene-based photodetector on a Si slot-waveguide, where the Si strips serve as local back gate electrodes to create a pn-junction for efficient photodetection. A responsivity of 35 mA/W, or 3.5 V/W, at zero-bias conditions was achieved, while under a moderate drain-source bias of 300 mV, the responsivity increased to 76 mA/W. The photodetector has shown a record high 3-dB cutoff frequency of 65 GHz. To further improve the responsivity in terms of V/W the electrical gating could be extended closer to the contacts, for example, by using thin Si slabs to both sides of the waveguide. Reducing the electrode spacing would result in lower device resistance and thus improved responsivity in A/W. Supporting Information: Design of mode converter and grating coupler, measurement of waveguide and coupling losses, graphene Raman spectrum, scanning electron microscopy images of devices with hBN and Al2O3 gate dielectrics, details on electrical and optical characterization methods, additional photocurrent and photovoltage maps, extracted intrinsic graphene conductivity, calculated optical absorption in the graphene sheet, sketch of the frequency response measurement setup. Acknowledgment: We thank Andreas Pospischil and Marco Furchi for technical assistance and Frank Koppens, Ilya Goykhman and Marco Romagnoli for helpful discussions. Financial support by the European Union (grant agreement No. 696656 Graphene Flagship) and the Austrian Science Fund FWF (START Y 539-N16) is acknowledged. Conflict of Interest: The authors declare no conflict of interest REFERENCES 1. Liu, J., Cannon, D. D., Wada, K., Ishikawa, Y., Jongthammanurak, S., Danielson, D. T., Michel, Ito, H., Furuta, T., Kodama, S., Watanabe, N., Ishibashi, T.; El. Lett. 1999, 35, 1556-1557. J., Kimerling, L. C.; Appl. Phys. Lett. 2005, 87, 103501. 2. Chen, L., Lipson, M.; Opt. Exp. 2009 , 17, 7901-7906. 3. 4. Liang, D., Roelkens, G., Baets, R., Bowers, J. E.; Materials 2010, 3, 1782-1802. 5. Vivien, L., Rouvière, M., Fédéli, J. M., Marris-Morini, D., Damlencourt, J. F., Mangeney, J., Crozat,P., El Melhaoui, L., Cassan, E., Le Roux, X., Pascal, D., Laval, S.; Opt. Express 2007, 15, 9843-9848. 6. Chen, L., Lipson, M.; Opt. Express 2009, 17, 7901-7906. 7. Vivien, L., Polzer, A., Marris-Morini, D., Osmond, J., Hartmann, J.M., Crozat, P., Cassan, E., Kopp, C., Zimmermann, H., Fédéli, J.M.; Opt. Express 2012, 20, 1096-1101. Sherwood-Droz, N., Lipson, M.; Opt. Express 2011 , 19, 17758-17765. 8. 9. Wang, J., Cheng, Z., Chen, Z., Xu, J. B., Tsang, H. K., Shu, C.; J. Appl. Phys. 2015, 117, 144504. 10. Novoselov, K.S., Geim, A.K., Morozov, S.V., Jiang, D., Zhang, Y., Dubonos, S.V., Grigorieva, I.V., Firsov, A.A.; Science 2004, 306, 666-669. 11. Dawlaty, J. M., Shivaraman, S., Strait, J., George, P., Chandrashekhar, M., Rana, F., Spencer, M.G., Veksler, D., Chen, Y.; Appl. Phys. Lett. 2008, 93, 131905. 12. Kuzmenko, A. B., Van Heumen, E., Carbone, F., Van Der Marel, D.; Phys. Rev. Lett. 2008, 100, 117401. 13. Nair, R. R., Blake, P., Grigorenko, A. N., Novoselov, K. S., Booth, T. J., Stauber, T., Peres, N. M. R., Geim, A. K.; Science 2008, 320, 1308-1308. 14. Bolotin, K.I., Sikes, K.J., Jiang, Z., Klima, M., Fudenberg, M., Hone, J., Kim, P., Stormer, H.L.; Solid State Commun. 2008, 146, 351-355. 15. Liu, M., Yin, X., Ulin-Avila, E., Geng, B., Zentgraf, T., Ju, L., Wang, F., Zhang, X.; Nature 2011, 474, 64-67. 16. Liu, M., Yin, X., Zhang, X. ; Nano Lett. 2012, 12, 1482-1485. 17. Phare, C. T., Lee, Y. H. D., Cardenas, J., Lipson, M.; Nat. Photonics 2015, 9, 511-514. 18. Mohsin, M., Schall, D., Otto, M., Noculak, A., Neumaier, D., Kurz, H.; Opt. Expr. 2014, 22, 15292-15297. 19. Hu, Y. T., Pantouvaki, M., Brems, S., Asselberghs, I., Huyghebaert, C., Geisler, M., Alessandri, C., Baets, R., Absil, P., Van Thourhout, D., Van Campenhout, J.; IEDM 2014, San Francisco, CA. 20. Urich, A., Unterrainer, K., Mueller, T.; Nano Lett. 2011, 11, 2804-2808. 21. Xia, F., Mueller, T., Lin, Y. M., Valdes-Garcia, A., Avouris, P.; Nat. Nanotechnol. 2009, 4, 839-843. 22. Mueller, T., Xia, F., Avouris, P.; Nat. Photonics 2010, 4, 297-301. 23. Pospischil, A., Humer, M., Furchi, M. M., Bachmann, D., Guider, R., Fromherz, T., Mueller, T.; Nat. Photonics 2013, 7, 892-896. 24. Gan, X., Shiue, R. J., Gao, Y., Meric, I., Heinz, T. F., Shepard, K., Hone, J., Assefa, S., Englund, D.; Nat. Photonics 2013, 7, 883-887. 25. Wang, X., Cheng, Z., Xu, K., Tsang, H. K., Xu, J. B.; Nat. Photonics 2013, 7, 888-891. 26. Shiue, R. J., Gao, Y., Wang, Y., Peng, C., Robertson, A. D., Efetov, D. K., Assefa, S., Koppens, F.H., Hone, J., Englund, D. (2015). ; Nano Lett. 2015, 15, 7288-7293. 27. Schall, D., Neumaier, D., Mohsin, M., Chmielak, B., Bolten, J., Porschatis, C., Prinzen, A., Matheisen, C., Kuebart, W., Junginger, B., Templ, W., Giesecke, A.L., Kurz, H.; ACS Photonics 2014, 1, 781-784. 28. Song, J. C., Rudner, M. S., Marcus, C. M., Levitov, L. S.; Nano Lett. 2011., 11, 4688-4692. 29. Piscanec, S., Lazzeri, M., Mauri, F., Ferrari, A. C., Robertson, J.; Phys. Rev. Lett. 2004, 93, 185503. 30. Bistritzer, R., MacDonald, A. H. ; Phys. Rev. Lett. 2009, 102, 206410. 31. Sun, D., Aivazian, G., Jones, A. M., Ross, J. S., Yao, W., Cobden, D., Xu, X.; Nat. Nanotechnol. 2012, 7, 114-118. 32. Tielrooij, K. J., Piatkowski, L., Massicotte, M., Woessner, A., Ma, Q., Lee, Y., Myhro, K. S., Lau, C. N., Jarillo-Herrero, P., van Hulst, N. F., Koppens, F. H.; Nat. Nanotechnol. 2015, 10, 437-443. 33. Xu, X., Gabor, N. M., Alden, J. S., van der Zande, A. M., McEuen, P. L.; Nano Lett. 2009, 10, 562-566. 34. Grosse, K. L., Bae, M. H., Lian, F., Pop, E., King, W. P. ; Nat. Nanotechnol. 2011, 6, 287-290. 35. Gabor, N. M., Song, J. C., Ma, Q., Nair, N. L., Taychatanapat, T., Watanabe, K., Taniguchi, T., Levitov, L. S., Jarillo-Herrero, P.; Science 2011, 334, 648-652. 36. Lemme, M. C., Koppens, F. H., Falk, A. L., Rudner, M. S., Park, H., Levitov, L. S., Marcus, C. M.; Nano Lett. 2011, 11, 4134-4137. 37. Ma, Q., Gabor, N. M., Andersen, T. I., Nair, N. L., Watanabe, K., Taniguchi, T., Jarillo-Herrero, P.; Phys. Rev. Lett. 2014, 112, 247401. 38. Freitag, M., Low, T., Avouris, P.; Nano Lett. 2013, 13, 1644-1648. 39. Freitag, M., Low, T., Xia, F., Avouris, P. ; Nat. Photonics 2013 , 7, 53-59. 40. Almeida, V. R., Xu, Q., Barrios, C. A., Lipson, M.; Opt. Lett. 2004, 29, 1209-1211. 41. 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.; Phys. Rev. Lett. 2006, 97, 187401. 42. Ashcroft, N. W., Mermin, N. D.; Solid State Phys. Saunders 1976, Philadelphia, 293. 43. Wang, J., Cheng, Z., Chen, Z., Wan, X., Zhu, B., Tsang, H. K., Shu, C., Xu, J. ; Nanoscale 2016, 8, 13206- 13211. Supporting Information for Controlled generation of a pn-junction in a waveguide integrated graphene photodetector Simone Schuler1, Daniel Schall2, Daniel Neumaier2, Lukas Dobusch1, Ole Bethge3, Benedikt Schwarz3, Michael Krall1, and Thomas Mueller1 1 Vienna University of Technology, Institute of Photonics, Gusshausstrasse 27-29, 1040 Vienna, Austria 2 AMO GmbH, Otto-Blumenthal-Strasse 25, 52074 Aachen, Germany 3 Vienna University of Technology, Institute of Solid State Electronics, Floragasse 7, 1040 Vienna, Austria 1.) Optical components Comsol Multiphysics was used to perform FEM simulations to design the slot-waveguide as well as the mode converter, which couples the optical mode from a strip waveguide into the slot-waveguide. The guided mode in the strip waveguide couples via the evanescent field into the slot-waveguide. The coupling efficiency dependence on the length of the coupler is illustrated in Figure S1. Figure S1. Calculated coupling efficiency, depending on the length of the mode converter. To efficiently couple light from an optical fiber into the strip waveguide, a grating coupler was used as shown in Figure S2a. To determine the responsivity, the losses of the coupler, the waveguide and the mode converter have to be taken into account. Figure S2b shows the dependence of the losses on the length of the waveguide. From the same plot the coupling efficiency of the grating coupler is determined. 1 Figure S2. (a) SEM image of the grating coupler. (b) Waveguide and grating coupler losses. 2.) Device fabrication The thicknesses of the graphene flakes were defined using Raman spectroscopy. The flakes were characterized on the PAA/PMMA stack before transfer. Figure S3 shows a Raman spectrum of a monolayer of graphene. Figure S3. Raman spectrum of a graphene monolayer on a PAA/PMMA stack. Either hexagonal boron nitride or aluminum oxide was used as gate dielectric to prevent electrical contact between the waveguide and graphene. Comparing the performance of devices based either on one or the other, we found a better response in case of aluminum oxide. As discussed in the main text, this is attributed to the non-conformal covering of the waveguide in case of boron nitride. To verify this presumption, we took SEM images of both types of devices. Indeed, the 13 nm boron nitride layer does not conformably cover the 2 waveguide (Figure S4a). Using aluminum oxide as gate dielectric, which is deposited in an ALD-process, the graphene layer covers the waveguide as well as the sidewalls (Figure 4b). Figure S4. (a) SEM picture of a 13 nm thick hexagonal boron nitride flake on a slot waveguide. (b) SEM image of a graphene flake placed on top a slot waveguide covered with 10 nm of aluminum oxide. As the graphene layer conformally covers the waveguide, the gating length is increased compared to devices with boron nitride. 3.) Electrical characterization Electrical characterization was performed using a Semiconductor Parameter Analyzer (Agilent 4155C). The devices were contacted using a RF-probe needle in signal-ground configuration (Picoprobe, GGB Industries). 4.) Optical characterization For optical characterization, 1560 nm light from a laserdiode was coupled via polarizers and a lens into an optical fiber (SMF-28), further coupled into the strip waveguide, via the grating coupler, and then into the slot-waveguide using the mode converter. The polarizers were adjusted such that a maximum photocurrent signal was detected. The signal was recorded using a transimpedance amplifier and a lock-in amplifier. We measured both the open- circuit photovoltage as well as the short-circuit photocurrent response of our detectors. In Figure S5a the photovoltage map of sample A is presented and Figure S5b shows the corresponding photocurrent map. Our best device (sample B) exhibited a responsivity of 3.5 V/W, respectively 35 mA/W, under zero-bias conditions. Figure 6 shows the resistance map of this device. The photovoltage and photocurrent maps for sample B are presented in Figures S7a and S7b, respectively. 3 Figure S5. (a) Photovoltage map of sample A, illustrating the six-fold photocurrent generation pattern (same Figure as in the main paper, but plotted on a different color scale). (b) Corresponding photocurrent map. Figure S6. Resistance map of sample B. 4 Figure S7. (a) Photovoltage map of sample B. (b) Photocurrent map. 5.) Modeling of the photo-thermoelectric effect The photo-thermoelectric effect of sample A was modeled based on the transfer characteristic of the device, as shown in Figure S8a. The Seebeck coefficient was calculated form the intrinsic conductivity, which is shown in Figure S8b. In order to estimate the absorbed power, we used FEM simulation to estimate the absorption depending on the length of the flake for mono- and bi-layer graphene on the slot-waveguide geometry as shown in Figure S9. 5 Figure S8. (a) Measured transfer characteristic of the presented device with VG1 = VG2 = VG. (b) Corresponding intrinsic conductivity and model as dashed and solid lines, respectively. Figure S9. Calculated absorption coefficient of a mono- and bi-layer flake placed on top of our slot- waveguide geometry, red and blue line, respectively. 6 Figure S10. (a) Measured transfer characteristic of Sample B with VG1 = VG2 = VG . (b) Corresponding intrinsic conductivity data and model, respectively and (c) calculated Seebeck coefficient. 7 6.) Photoconductive effect Figure S11. (a) Photocurrent map (Sample A) at 𝑉𝐷𝑆= -400 mV. (b) Corresponding photocurrent at the diagonal (dashed line) where the PTE effect is negligible, indicating an additional photoconductive (PC) contribution when a bias is applied. 7.) Frequency response measurement Figure S12. Sketch of the RF measurement setup. GPD = graphene photodetector, SA = spectrum analyzer, EDFA = erbium doped fiber amplifier. Laser 1 is kept at constant wavelength, the wavelength of Laser 2 is tuned. For each frequency the measured power is recorded. The values in % give the coupling ratio of the fiber couplers. 8
1811.00916	3	1811	2019-02-21T14:57:03	Supercurrent carried by non-equlibrium quasiparticles in a multiterminal Josephson junction	[ "cond-mat.mes-hall" ]	We theoretically study coherent multiple Andreev reflections in a biased three-terminal Josephson junction. We demonstrate that the direct current flowing through the junction consists of supercurrent components when the bias voltages are commensurate. This dissipationless current depends on the phase in the superconducting leads and stems from the Cooper pair transfer processes induced by non-local Andreev reflections of the quasiparticles originating from the superconducting leads. We identify supercurrent-enhanced lines in the current and conductance maps of the recent measurement [Y. Cohen, et al., PNAS 115, 6991 (2018)] on a nanowire Josephson junction and show that the magnitude of the phase-dependent current components is proportional to the junction transparency with the power corresponding to the component order.	cond-mat.mes-hall	cond-mat	Supercurrent carried by non-equlibrium quasiparticles in a multiterminal Josephson junction M. P. Nowak,1 M. Wimmer,2, 3 and A. R. Akhmerov3 1AGH University of Science and Technology, Academic Centre for Materials and Nanotechnology, al. A. Mickiewicza 30, 30-059 Krakow, Poland 2QuTech, Delft University of Technology, P.O. Box 4056, 2600 GA Delft, The Netherlands 3Kavli Institute of Nanoscience, Delft University of Technology, P.O. Box 4056, 2600 GA Delft, The Netherlands (Dated: February 22, 2019) We theoretically study coherent multiple Andreev reflections in a biased three-terminal Josephson junction. We demonstrate that the direct current flowing through the junction consists of supercur- rent components when the bias voltages are commensurate. This dissipationless current depends on the phase in the superconducting leads and stems from the Cooper pair transfer processes induced by non-local Andreev reflections of the quasiparticles originating from the superconducting leads. We identify supercurrent-enhanced lines in the current and conductance maps of the recent mea- surement [Y. Cohen, et al., PNAS 115, 6991 (2018)] on a nanowire Josephson junction and show that the magnitude of the phase-dependent current components is proportional to the junction transparency with the power corresponding to the component order. I. INTRODUCTION of a Upon biasing superconductor-normal- superconductor (SNS) junction the supercurrent carried by Andreev bound states changes into quasiparticle- driven dissipative current carried in a process of multiple Andreev reflections (MAR). This phenomenon was first encountered in superconducting niobium contacts [1], later in microbridges [2 and 3] and tunnel junctions [4 -- 6]. Klapwijk, Blonder, and Tinkham explained [7] that n successive Andreev reflections of electrons and holes propagating through the normal part of the junction results in charge transfer of (n + 1)e, with n/2 Cooper pairs and a single quasiparticle between the superconductors. As a result the current-voltage (I-V) characteristic of the junction is imprinted with a subgap structure with features appearing for the bias voltages V = 2∆/ne with ∆ the superconducting gap. The original MAR theory of Ref. 7 was soon extended allowing for description of non-transparent junctions [8], coherent regime [9 and 10] -- with the prediction of appearance of DC and AC current components -- , and finally the case of several current-carrying modes [11]. The latter allowed to estimate transmission probabilities and number of the quantized modes of atomic-thick break junctions [12 and 13] -- the pincode of the struc- ture. Nowadays probing of MAR features became a standard technique for estimating the superconduct- ing gap, evaluation of the number and transmission probability of conducting modes in the state-of-the-art semiconductor-superconductor hybrids that base on nanowires [14 -- 18] and two-dimensional electron gas [19]. Recent progress in fabrication of hybrid nanodevices driven by the pursuit for creation topological quantum gates [20 -- 22] led to creation of tunable multiterminal systems based on crossed nanowires [23 -- 26] or gated graphene [27]. Termination of such structures by super- conducting leads allows to form multiterminal Josephson junctions that serve as superconducting beam splitters that entangle Cooper pairs [28 and 29], allow to obtain Shapiro steps [30] due to voltage-induced supercurrent and where the transconductance due to AC Josephson effect is quantized in units of 4e2/h [31 and 32]. Exten- sive work on the description of the phase dependence of Andreev bound states in multiterminal junctions [33 -- 35] was done demonstrating that the superconducting phases can lift Kramers degeneracy [36] or even that such junc- tions can be considered as an effective topological mate- rials themselves [37 and 38]. In a Josephson junction, where a central superconduct- ing electrode is connected with two outer superconduc- tors solely through two separate normal regions (Joseph- son bi-junction), the transport between the outer S leads is possible only via Andreev reflection at the central su- perconductor. In this geometry the nonlocal Andreev reflections correlate four particles by the exchange of two Cooper pairs between the superconducting leads which results in a quartet supercurrent [39]. Measurements on such a junction realized in a recent experiment [40] found current amplification for commensurate bias voltages in line with the above theoretical prediction. The same experiment studied transport through a three-terminal structure, where all three superconducting leads are con- nected by a common semiconducting part [40 and 41], extending the previous measurement of metallic junction of the same geometry [42]. The experimental maps of the current in the three-terminal nanowire-device [41] reveal both: lines of amplified current due to MAR processes and lines analogous to those obtained in the bi-junction geometry in-line with the prediction of current spikes as a counterpart of voltage induced Shapiro steps [30]. De- spite the experimental progress the theoretical descrip- tion of biased multi-terminal Josephson junctions was so far limited to incoherent [43] and diffusive [30] regimes or to a single bias voltage [32, 44, and 45] hindering the 9 1 0 2 b e F 1 2 ] l l a h - s e m . t a m - d n o c [ 3 v 6 1 9 0 0 . 1 1 8 1 : v i X r a 2 Figure 1. Schematics of the considered device with three semi- infinite superconducting leads (blue) and normal scattering region (grey). correspond to electron (hole) amplitudes and x points in the direction opposite to the scattering region for all three leads. The scattering properties of the normal part of the junction are contained within the scattering matrix S0 that is used to setup the matching conditions for the wave-functions ΨL. For the electron part we have: and for the hole part: n,m n+1,m n,m+1 AII AIII  AI  DI n,m n−1,m n,m−1 DII DIII  = S0  = S∗ 0 C II C III  C I  BI n,m n+1,m n,m+1 n,m n−1,m n,m−1 BII BIII  ,  , (2) (3) interpretation of the experimental data. Here we develop a theory for coherent MAR in a three- terminal Josephson junction of arbitrary transparency whose leads can be biased independently. Our approach accounts for competing processes of local and non-local MAR between all the leads which allows to capture both MAR and supercurrent features in the DC response of a multi-terminal junction. We show that when the voltages are commensurate a phase-dependent current may flow without a change in voltage reminiscent of the supercur- rent. This happens despite the fact that when all the bias voltages are unequal, the junction has no conserved quantities, and therefore every state in the junction has a finite lifetime and dissipates energy. We show that MAR DC current consists of a voltage-dependent, dis- sipative component, and series of phase-dependent, dis- sipationless contributions. As a result the DC current maps versus the bias voltages consist of pronounced lines of enhanced current for commensurate voltages as ob- served in the experiment Ref. [40 and 41]. We find that supercurrent components have oscillatory dependence on the superconducting phase with a period and amplitude inversely proportional to the component order. Further- more, we present how the visibility of the supercurrent features in the conductance maps depends on the trans- parency of the normal part which is applicable for analy- sis of the transport properties of nanoscale multiterminal devices [23, 24, and 46] This paper is organized in the following way. The the- ory is given in the second section. Section III contains results of the model. Discussion of the results along with the summary is provided in sections IV and V respec- tively. II. THEORY A. Calculation of the current We consider a junction that consists of three semi- infinite superconducting electrodes (S) connected by a normal region (N) depicted schematically in Fig 1. We assume that the first lead is kept at voltage V1 = 0, while the second and third leads are biased by V2 and V3 volt- ages respectively. To calculate the current running through the junc- tion we generalize the approach previously applied for a two-terminal case [9] and consider quasiparticle wave- functions (ΨL) in a form of a linear combination of plane waves propagating in the normal region, adjacent to the L'th superconducting lead [see Fig. 1], (cid:20)(cid:18) AL (cid:19) (cid:88) n,m ΨL = n,m BL n,m eikx + (cid:19) (cid:18) C L n,m DL n,m (cid:21) e−ikx × e−i[E+neV2+meV3]t/ . The time dependence accounts for the voltage applied to the superconducting leads, AL n,m) n,m (BL n,m, DL n,m, C L with the shifts of the indices that account for the particles gaining and loosing energy due to the bias voltages. At each SN interface we account for Andreev reflection and acquisition of the phase present at the superconduct- ing lead, hence: (cid:18) an,m 0 (cid:19)(cid:18) DL (cid:19) (cid:19) (cid:18) C L n,m BL n,m = σL 0 an,m n,m AL n,m , (4) where the Andreev-reflection amplitudes are given by an,m ≡ a(E + neV2 + meV3), with, √ √ E − i (cid:26) E − sgn(E) (cid:18) e−iφL ∆2 − E2 σL = 0 E2 − ∆2 (cid:19) , 0 eiφL E > ∆ E ≤ ∆ , (5) (6) (1) a(E) = 1 ∆ and where V1=0V2V3IIIIIIIIIS0accounts for the phase shift at the L'th SN interface. An electron/hole excitation in the normal part of the junction is created by incoming quasiparticles from the nearby superconductors. We assume that there is no Fermi wavelength difference between the normal and su- perconducting parts and that the chemical potential is much higher than the energy gap. In the superconductor for the energies exceeding the superconducting gap there are two modes propagating towards the normal region [47] with the wave-functions: Ψqe inc = e−ikx, Ψqh inc = eikx, (7) (cid:19) (cid:18) v u (cid:19) (cid:18) u v corresponds to the (hole-like) quasi- char- corresponding am- The electron-like (cid:104)(cid:16) 1 +(cid:112)1 − (∆/E)2 (cid:105)1/2 (cid:105)1/2 (cid:17) /2 /2 respectively. (cid:104)(cid:16) (Ψqh inc) has where Ψqe inc that particle E (cid:29) ∆. acter are u, v plitudes for (cid:17) 1 −(cid:112)1 − (∆/E)2 (cid:19) (cid:18) u ΨS(x) = v Assuming that the incoming quasiparticle has electron- like character we write the wave-functions at each side of the SN interface. At the superconducting part we have: (cid:19) (cid:18) u v (cid:19) (cid:18) v u e−ikx +a eikx +b e−ikx, (8) where a, b stand for the amplitudes of the reflected quasi- particles. On the normal side we have: (cid:19) (cid:18) 1 0 (cid:19) (cid:18) 0 1 ΨN (x) = c e−ikx + d eikx. (9) u Matching the wave-functions and their derivatives at the interface between the materials (x = 0) one finds a = d = 0, c = u2−v2 √ . Taking into account the super- u2 − v2) the electron- conducting density of states (1/ like quasiparticle with energy E, creates an electron in the normal part with the wave-function amplitude J = that determines the filling of the electron band. Following the same procedure one finds that hole-like quasiparticle creates a hole in the normal part with the same ampli- tude. (cid:112)FD(E) with FD(E) the Fermi distribution u2−v2 u √ The above relations allow us to write Eq. 4 including the source terms: (cid:19) (cid:18) C L n,m BL n,m (cid:19) (cid:18) an,m 0 (cid:18) J(E + eVL) (cid:18) an,m 0 0 J(E − eVL) 0 AL (cid:19)(cid:18) DL (cid:19) 1√ (cid:19) 1√ 2 2 n,m n,m δp,eδs,2κ+ L δp,hδs,2κ− L , =σL + + 3 Imax(cid:88) We calculate the electric current I in the L'th lead from the probability current taking into account one- dimensional density of states: I L = ı,je(ıV2+jV3)eit/ I L , (11) where, ı,j I L ı,j = e π (cid:88) (cid:88) s=1,2,3 p=e,h (cid:90) ∞ −∞ dE Nmax(cid:88) (cid:1)T n,m (UL∗ ı+n,j+mUL n,m − VL∗ ı+n,j+mVL n,m), (12) (cid:1)T (cid:0)AL n,m = (cid:0)C L n,m, BL are Fourier components of the current. UL n,m = stand for vectors that consist of electron and hole amplitudes of wave-functions that carry positive and negative current, respectively. n,m, DL and VL n,m n,m The integral is evaluated numerically, where at each point of the integration we solve system of equations built with Eqs. (2,3,10). We assume zero temperature and take Imax = Nmax = 8 for the calculations when both voltages are varied which sets the limit to the convergence to voltages larger than eV > ∆/8. Treatment of the regime with V2, V3 close to zero -- with the current below the critical current -- is beyond the scope of the present work. The code used for the calculations is available in Ref. [48]. B. Scattering matrix We assume that the central part of the junction is a symmetric beam splitter [grey region in Fig. 1] and that the normal part of the junction is shorter than the su- perconducting coherence length. In other words it has an energy-independent scattering matrix:  ,  α β β −eia/(cid:0)2eia − e−ia(cid:1) β α β β β α S0 = (13) 2i sin(a)/(cid:0)2eia − e−ia(cid:1) with = α transparency of probability D = 2β2. β [49], where a controls = the the splitter with the transmission and (10) III. RESULTS A. Current versus the bias voltages where p controls the incoming quasiparticle type, s the position of the source term, and κ± 2 = δn,±1δm,0, κ± 3 = δn,0δm,±1 account for the shifts in the chemical potential introduced by the bias voltage. 1 = δn,0δm,0, κ± Let us start by inspecting the DC response of the junc- tion due to biasing of the second and third leads. Figure 2 (a) shows the current in the first lead while (b) presents 4 Figure 2. a) DC current in the first lead for D = 0.5. b) Differential conductance obtained from the map of (a). the differential conductance obtained from the current map. In the maps we include components of the current that fulfill ıeV2+jeV3 < 5·10−3 meV such the current fea- tures at commensurate voltages have comparable width to the ones observed experimentally [40 and 41]. We find two types of features in the conductance map: the cen- trifugal lines at commensurate voltages (ıV2 + jV3 = 0) that are distinctively sharper than other subharmonic features that appear at (ıV2 + jV3 = 2∆/e) due to non- local MAR. Accordingly, we observe pronounced lines of altered current at commensurate voltages V2 = V3 and V2 = −V3 in the map of Fig. 2 (a). Note, that sole time-reversal symmetry present in the considered system is not enough to guarantee the symmetry of the current with respect to the change of the sign of the voltages as application of this symmetry leads to different electron occupation in the leads. We plot the current running through the first lead in the regime V2 = V3 = V with the blue curve in Fig. 3 (a). In this regime the three-terminal junction reduces to a two-terminal one and the current reproduces exactly the MAR response obtained in the work of Ref. 9. More- over, due to symmetry of the second and the third lead we find exactly the same current in both equally biased Figure 3. DC current (solid and dotted curves) in the leads versus the bias energy eV with V2 = V3 = V and D = 0.5. Dashed curves present DC components I I ı,j of the current in the first lead. GT = e2D/π. b) Cross-section of the current map of Fig. 2 (a) for eV3 = 0.7∆. conservation law (cid:80) leads depicted with the green crosses overlapping the or- ange curve. The magnitude of that current is half of the current running through the first lead and hence current L I L = 0 is fulfilled. Note however that when the voltages fulfill ıV2 + jV3 = 0 the current consists of multiple DC components, obtained for ı = −j in Eq. (11). We plot the individual current components in Fig. 3 (a) with dashed curves and observe that the increase of the component order ı,j results in a de- crease of the current amplitude. Already the sixth-order component ı = −j = 6 is nearly zero. The components of odd order are zero, as the particles outgoing from and reaching the leads are shifted in energy by even multiples of the bias voltages times e. In Fig. 3 (b) we present cross-section of the map of Fig. 2 (a) obtained for eV3 = 0.7∆. The sharp features appearing at commensurate voltages are marked by black vertical lines while the non-local MAR features at (ıV2 + jV3 = 2∆/e) are marked with arrows. Upon applying of the phase difference at the biased lead the current for the commensurate voltages changes -- as explicitly seen in the peak heights dependence for different values of φII in −2.0−1.5−1.0−0.50.00.51.01.52.0eV2/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0eV3/Δ−1.00−0.75−0.50−0.250.000.250.500.751.00I/(Δe/πħ)a)b)−2.0−1.5−1.0−0.50.00.51.01.52.0eV2/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0eV3/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0dI/dV3[2e2/h]a)b)212/32/4eV/Δ−2−10123−I/(ΔGT/e)totalL=ItotalL=IItotalL=IIIı=j=0ı=−j=2ı=−j=4ı=−j=6ı=−j=8212/32/4eV/Δ0.00.51.01.52.0−I/(ΔGT/e)ϕII=0ϕII=π/2ϕII=π/4V2= 2V32V2= 3V3V2= V3Fig. 3 (b). In the following section we will discuss the origin of this phenomenon. B. Phase dependence 5 Figure 5. Same as Fig. 4 but for V2 = −V3 = e∆/2. the current into the separate DC components we see that again the zeroth order component is phase independent and it corresponds to the ohmic current that is propor- tional to the voltage difference between the biased leads. The higher order components are phase-dependent and oscillate around zero with a small amplitude following the same rule of the decrease of period as the currents of Fig. 4 (b). Those components then constitute the supercurrent that is carried between commensurately bi- ased leads for commensurate voltages. The small phase offsets of the higher order current components are the re- sult of phase shifts introduced by complex transmission and reflection amplitudes in the scattering matrix when the scattering involves non-local processes. C. Quasiparticle driven supercurrent versus the scattering paths To understand the origin of the supercurrent let us inspect the quasiparticle scattering processes that stand behind one of the higher order components of the cur- rent running between second and the third lead, namely ı = −j = 2 for V2 = V3 = ∆/2e. In the first step we isolate the source terms that that initiate the ma- jority of the current flowing between the leads. In Fig. 6 we display the current broken down into contribution from all possible source terms: s corresponds to the lead where we include the source and p to the quasiparti- cle type -- see Eq. (12). We see that actually half of the current between equally biased leads results from an electron-like quasiparticle incoming from the first lead. Figure 4. (a) DC current running between the second and the third lead I II − I III for three values of the central region transparency D for V2 = V3 = e∆/2 versus the phase on the second lead. (b) Components of the current for D = 0.75. Let us now inspect the phase dependence of the cur- rent carried between the second and third lead depicted in Fig. 4 (a). We set φI = φIII = 0 and vary φII. In agreement with the equal currents in the second and third leads of Fig. 3 (a) we see zero current for φII = 0. Upon introduction of the phase difference a nonzero current arises and its phase dependence resembles the supercur- rent phase-relation of a diffusive Josephson junction [50] for all the transparencies D considered here. Focusing on D = 0.75 we decompose the current into the DC compo- nents. Remarkably, we see that the component ı = j = 0 is zero [the blue curve in Fig. 4 (b)] as the voltage dif- ference between the second and third lead is also zero and as the modification of the amplitudes by the super- conducting phase does not contribute to the calculated current for ı = j = 0. The subsequent components of ı'th order oscillate with the period inversely proportional to the component order with ∼ sin(ı/2φII). Similar analysis performed for another commensurate voltage configuration, i.e. V2 = −V3 = ∆/2e also shows that the current oscillation amplitude decreases when the transparency is reduced [see Fig. 5 (a)]. By decomposing −3−2−10123I/(ΔG0/e)D=0.25D=0.5D=0.75a)b)0π/8π/4π/2π3π/22πϕII−2.0−1.5−1.0−0.50.00.51.01.52.0I/(ΔG0/e)ı=j=0ı=−j=2or−2ı=−j=4or−4ı=−j=6or−6−0.50.00.51.01.52.02.5I/(ΔG0/e)D=0.25D=0.5D=0.75a)b)0π/8π/4π/2π3π/22πϕII0.00.51.01.52.0I/(ΔG0/e)ı=j=0ı=j=2or−2ı=j=4or−46 n,m, BII n,m, C II n,m, DII lead. We set the injection energy just below the gap, i.e. multiply the integrand in Eq. (12) by Dirac delta located at E = −1.0001∆. In Fig. 7 we plot the squared absolute values of amplitudes AII n,m of the wave- functions adjacent the second superconducting lead. This way we are able to quantify the probability of finding electron and hole components of the wave-functions with a given propagation direction at certain energies. We observe that overall the highest probabilities correspond to the wave-functions of the electron propagating towards the scattering region [top right panel on Fig. 7] -- with the probabilities C II n,m2 -- and to its time reversed partner -- a hole propagating in the opposite direction [bottom right panel on Fig. 7] -- with probabilities DII n,m2. Focusing on the electron part we see that the most important con- tributions to the current stem from wave-function ampli- tudes with indexes: n = 1, m = 0 and n = −1, m = 2, which both give the wave-function shifted in energy by eV2 = eV3. The C II n=1,m=0 coefficient is populated in a process where an electron coming from the first lead is Andreev reflected in the second lead, scatters back to this lead and finally is Andreev reflected again -- see the top inset to Fig. 7. This process is insensitive to the phase on the second lead and we approximate its probability as D(1 − D). The finite C II n=−1,m=2 amplitude stems from a process where the electron enters the scattering region from the first lead and propagates to the third lead. Due to nonlocal Andreev reflection, the retro-reflected hole propagates to the second lead where it is Andreev re- flected again into an electron with the energy shifted by 2eV3 − eV2 = ∆/2 -- see bottom inset to Fig. 7. As a result we can estimate the corresponding probability as ∼ D2. The process that populates the C II n=−3,m=4 ampli- tude can be drawn analogically, only now the hole visits the second lead two times and by that gains twice the phase shift. Figure 6. I II − I III DC current component ı = −j = 2 calcu- lated for different source terms (s -- the source position, p -- quasiparticle type). Another source term that induces a considerable current correspond to hole-like quasiparticles injected from the second and third lead, which are Andreev reflected at the first lead forming an electron propagating towards the scattering region as in the case of the major compo- nent. As a result in such symmetric bias configuration the current component due to source term (s : 1, p : e) is equal to the sum of components due to (s : 2, p : h) and (s : 3, p : h). Similarly the sum of the current com- ponents due to source term (s : 2, p : e) and (s : 3, p : e) correspond to the current component due to (s : 1, p : h). Figure 7. Probabilities of the wave-function components ad- jacent to the second lead for electron-like quasiparticle as the source term at energy −1.0001∆ injected from the first lead. Transparency is D = 0.75 and eV2 = eV3 = ∆/2. The insets on the right-hand-side we show two scattering processes con- tributing to the second DC component (quasiparticle induced supercurrent) that result from an electron (red filled circle) propagating towards the scattering region. Next, let us trace the scattering events that happen af- ter the electron-like quasiparticle incomes from the first According to Eq. (12) and Eq. (11) the main DC com- ponent of the current ı = j = 0 is obtained from the sum of the probabilities set by the wave-function amplitudes and hence it cannot be directly dependent on phase of the amplitudes. On the other hand, the higher order DC current components are obtained as a sum of the prod- ucts of wave-functions amplitudes with indexes n and m shifted by ı and j respectively, such ıV2 + jV3 = 0 and by that they are phase-dependent. The higher order compo- nents result from the scattering processes with increased number of Andreev reflection at the phase-biased lead, resulting in a decrease of period of oscillation in phase as seen in Fig. 4 (b). This means that the supercurrent (i.e. the higher order components of the DC current) originate from quasiparticles entering from the supercon- ducting leads but also that it involves non-local Andreev reflection process of non-equlibrium electron/holes at the biased superconductor leads. 0π/2π3π/22πϕII−0.75−0.50−0.250.000.250.500.75I/(ΔG0/e)s:1, p:es:1, p:hs:2, p:es:2, p:hs:3, p:es:3, p:h−6−4−20246njAIIn;mj2jCIIn;mj2−6−4−20246m−6−4−20246njBIIn;mj2−6−4−20246m0.0000.0030.0060.0090.0120.0150.0180.0210.024jDIIn;mj2IIIIIIIIIIII7 transparency [Fig. 9 (b)] and high transparency [Fig. 9 (c). Comparing the maps we observe a distinct change in the visibility of both: the non-local MAR lines and centrifugal supercurrent-enhanced lines. Making the sys- tem more transparent results in smoothing out the non- local MAR features and significant amplification of the visibility of the current consisting of multiple DC com- ponents. The latter results from the presence of higher order DC components that require many sequential scat- tering events through the junction and are favored in transparent junctions. On the other hand the usual MAR contribution to the map lacks higher order components and hence resembles the usual MAR response to the transparency -- the lower the transparency the sharper the features. IV. DISCUSSION The conductance map obtained within this work shows a similar subharmonic gap pattern as the map Fig. 1 of Ref. 43. Here however, we are able to recover the full spectrum of the features -- in particular with those at commensurate voltages -- missing in the modeling of Ref. 43 due to incoherent nature of the transport. On the other hand the obtained supercurrent features are com- patible with the voltage induced Shapiro steps that were demonstrated for a diffusive Josephson junction with a superconducting tunnel probe [30]. By that we are able to recover the full spectrum of features as found in the experiment on the nanowire Josephson junction. The maps of the second derivative of the current Fig. 9 obtained within the presented model exhibit the fea- tures compatible with those in the experimental maps of Ref. 41 [see Fig. 4 and S6 therein]. Specifically, we ob- tain subharmonic MAR structure but most importantly also the rapid amplification of the current for the com- mensurate voltages which in the experimental paper was attributed to the supercurrent driven by Andreev bound states in the junction. Here, however we see that all the states in the junction have a finite life-time due to un- equal bias voltages. The current amplification appear due the higher order DC current components in which the Cooper pairs are transported between commensu- rately biased leads by non-equilibrium electrons and holes through MAR processes. When the bias voltages are de- tuned from the commensurate condition ıV2+jV3 (cid:54)= 0 the higher order components become rapidly oscillating and drop out of the sum for the current Eq. (11) resulting in a change in the DC current. V. SUMMARY Figure 8. Critical current between second and third lead ver- sus the junction transparency D for eV2 = eV3 = 1.4∆ (a) and eV2 = −eV3 = 1.4∆ (b). The open circles present the numerical results and the curves are analytical dependencies as described in the text. D. Transparency dependence √ 2 3 D2 ·(cid:112)D(1 − D) ac- (cid:112)D2 · D(1 − D) (cid:112)D2 · D(1 − D) To elucidate the supercurrent dependence on the junc- tion transparency we now turn our attention to the de- pendence of the maximum of the supercurrent compo- nents on D. In Fig. 8 (a) with the open circles we show the critical current Ic = max[I(φII ) − I(φIII )] for V2 = V3, while the solid curves are estimates of the criti- cal current inferred from the junction transparency. For ı = −j = 2 we have Ic (cid:39) a1 cording to the analysis of the amplitudes of the wave- functions in the second lead. For ı = −j = 4 we have Ic (cid:39) a2 and finally for ı = −j = 6 Ic (cid:39) a3 with the free parameters ai chosen for the best fit. We see that the higher order components of the current require more scattering events between the biased and unbiased electrodes, so that the wave-functions acquire multiple gains in phase -- hence the decreased oscillation period in current-phase rela- tion. At the same time the higher order components are more sensitive to the transparency of the normal part -- resulting in the increased power of Ic dependence on D. We also calculate the critical current for V2 = −V3 as a function of D and find that it follows the same rule of increasing power in D for increasing component number [see Fig. 8 (b)]. It is important to note here that the de- tailed shape of the curve -- such as the presence of a local maximum -- depends on the voltage for which we obtain the curve. Nevertheless the qualitative dependence on the power of D is independent on that choice. Finally, let us go back to the analysis of junction re- sponse when both bias voltages are varied. To high- light the subharmonic features we now present the sec- ond derivative of the current in the first lead versus the bias voltages for low transparency [Fig. 9 (a)], average We studied coherent multiple Andreev reflections in a multiterminal Josephson junction with two bias volt- ages. We analyzed maps of the current running through the junction and identified pronounced lines of enhanced 0.20.40.60.8D0.00.10.20.30.40.50.60.70.8Ic/(ΔG0/e)ı=−j=2ı=−j=4ı=−j=60.20.40.60.8D0.000.020.040.060.080.100.120.14ı=j=2ı=j=4a)b)Ic/(ΔG0/e)8 Figure 9. Second derivative of the current in the first lead for D = 0.25 (a) D = 0.5 (b) and D = 0.75 (c). The centrifugal lines are enhanced by the supercurrent. current compatible with the ones measured in the recent experiment [40 and 41]. This enhancement is the result of multiple DC components contributing to the current for commensurate bias voltages. We found that the principal DC component of the current is voltage-dependent and corresponds to the dissipative current, while the higher order ones are non-dissipative and depend on the time- independent phase applied to the superconducting leads. This happens despite the absence of bound states that may carry dissipationless current. We explained that the latter components result from nonlocal Andreev reflec- tions of non-equlibrium particles propagating through the junction. We found that the DC current compo- nent order n determines both: the periodicity of the cur- rent/phase relation with the period 4π/n and the mag- nitude of the supercurrent with the main trend given by ∼ Dn. Finally, we identified the transparency depen- dence of the enhanced current and MAR lines in current maps and point out the open systems as the preferable one for observation of the supercurrent-enhanced lines. ACKNOWLEDGMENTS We are thankful to D. C. Sticlet, M. Irfan and T. O. Rosdahl for helpful discussions. This work was sup- ported by National Science Centre, Poland (NCN) ac- cording to decision number DEC-2016/23/D/ST3/00394, by the Netherlands Organization for Scientific Research (NWO/OCW) and by the European Research Council (ERC). The calculations were performed on PL-Grid In- frastructure. 1 L. J. Barnes, Phys. Rev. 184, 434 (1969). 2 P. E. Gregers-Hansen, E. Hendricks, M. T. Levinsen, and G. R. Pickett, Phys. Rev. Lett. 31, 524 (1973). 3 K. Flensberg and J. B. Hansen, Phys. Rev. B 40, 8693 (1989). 4 J. M. Rowell and W. L. Feldmann, Phys. Rev. 172, 393 (1968). 13 J. J. Riquelme, L. d. l. Vega, A. L. Yeyati, N. Agraıt, A. Martin-Rodero, and G. Rubio-Bollinger, EPL 70, 663 (2005). 14 H. Y. Gunel, I. E. Batov, H. Hardtdegen, K. Sladek, A. Winden, K. Weis, G. Panaitov, D. Grutzmacher, and T. Schapers, J. Appl. Phys. 112, 034316 (2012). 15 H. A. Nilsson, P. Samuelsson, P. Caroff, and H. Q. Xu, 5 A. W. Kleinsasser, R. E. Miller, W. H. Mallison, and G. B. Nano Lett. 12, 228 (2012). Arnold, Phys. Rev. Lett. 72, 1738 (1994). 6 N. van der Post, E. T. Peters, I. K. Yanson, and J. M. van Ruitenbeek, Phys. Rev. Lett. 73, 2611 (1994). 7 T. M. Klapwijk, G. E. Blonder, and M. Tinkham, Physica B+C 109-110, 1657 (1982). 8 M. Octavio, M. Tinkham, G. E. Blonder, and T. M. Klap- wijk, Phys. Rev. B 27, 6739 (1983). 9 D. Averin and A. Bardas, Phys. Rev. Lett. 75, 1831 (1995). 10 E. N. Bratus', V. S. Shumeiko, and G. Wendin, Phys. Rev. Lett. 74, 2110 (1995). 11 A. Bardas and D. V. Averin, Phys. Rev. B 56, R8518 (1997). 12 E. Scheer, P. Joyez, D. Esteve, C. Urbina, and M. H. Devoret, Phys. Rev. Lett. 78, 3535 (1997). 16 M. F. Goffman, C. Urbina, H. Pothier, J. Nygard, C. M. and P. Krogstrup, New J. Phys. 19, 092002 Marcus, (2017). 17 O. Gul, H. Zhang, F. K. de Vries, J. van Veen, K. Zuo, V. Mourik, S. Conesa-Boj, M. P. Nowak, D. J. van Wo- erkom, M. Quintero-P´erez, M. C. Cassidy, A. Geresdi, S. Koelling, D. Car, S. R. Plissard, E. P. A. M. Bakkers, and L. P. Kouwenhoven, Nano Lett. 17, 2690 (2017). 18 F. K. de Vries, J. Shen, R. J. Skolasinski, M. P. Nowak, D. Varjas, L. Wang, M. Wimmer, J. Ridderbos, F. A. Zwa- nenburg, A. Li, S. Koelling, M. A. Verheijen, E. P. A. M. Bakkers, and L. P. Kouwenhoven, Nano Lett. 18, 6483 (2018). −2.0−1.5−1.0−0.50.00.51.01.52.0eV2/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0eV3/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0eV2/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0−2.0−1.5−1.0−0.50.00.51.01.52.0eV2/Δ−2.0−1.5−1.0−0.50.00.51.01.52.0−1.00−0.75−0.50−0.250.000.250.500.751.00dG/dV3[au]a)b)c)9 19 M. Kjaergaard, H. J. Suominen, M. P. Nowak, A. R. Akhmerov, J. Shabani, C. J. Palmstrøm, F. Nichele, and C. M. Marcus, Phys. Rev. Applied 7, 034029 (2017). 20 J. Alicea, Y. Oreg, G. Refael, F. von Oppen, and M. P. A. 35 F. Vischi, M. Carrega, E. Strambini, S. D'Ambrosio, F. S. Bergeret, Y. V. Nazarov, and F. Giazotto, Phys. Rev. B 95, 054504 (2017). 36 B. van Heck, S. Mi, and A. R. Akhmerov, Phys. Rev. B Fisher, Nat. Phys. 7, 412 (2011). 90, 155450 (2014). 21 B. v. Heck, A. R. Akhmerov, F. Hassler, M. Burrello, and 37 R.-P. Riwar, M. Houzet, J. S. Meyer, and Y. V. Nazarov, C. W. J. Beenakker, New J. Phys. 14, 035019 (2012). 22 T. Hyart, B. van Heck, I. C. Fulga, M. Burrello, A. R. Akhmerov, and C. W. J. Beenakker, Phys. Rev. B 88, 035121 (2013). 23 S. R. Plissard, I. van Weperen, D. Car, M. A. Verheijen, G. W. G. Immink, J. Kammhuber, L. J. Cornelissen, D. B. Szombati, A. Geresdi, S. M. Frolov, L. P. Kouwenhoven, and E. P. A. M. Bakkers, Nat. Nano. 8, 859 (2013). 24 E. M. T. Fadaly, H. Zhang, S. Conesa-Boj, D. Car, O. Gul, S. R. Plissard, R. L. M. Op het Veld, S. Kolling, L. P. Kouwenhoven, and E. P. A. M. Bakkers, Nano Lett. 17, 6511 (2017). 25 F. Krizek, J. E. Sestoft, P. Aseev, S. Marti-Sanchez, S. Vaitiekenas, L. Casparis, S. A. Khan, Y. Liu, T. Stankevic, A. M. Whiticar, A. Fursina, F. Boekhout, R. Koops, E. Uccelli, L. P. Kouwenhoven, C. M. Mar- cus, J. Arbiol, and P. Krogstrup, Phys. Rev. Materials 2, 093401 (2018). 26 S. Vaitiekenas, A. M. Whiticar, M.-T. Deng, F. Krizek, J. E. Sestoft, C. J. Palmstrøm, S. Marti-Sanchez, J. Arbiol, P. Krogstrup, L. Casparis, and C. M. Marcus, Phys. Rev. Lett. 121, 147701 (2018). 27 A. W. Draelos, M.-T. Wei, A. Seredinski, H. Li, Y. Mehta, K. Watanabe, T. Taniguchi, I. V. Borzenets, F. Amet, and G. Finkelstein, arXiv:1810.11632 (2018). 28 A. Freyn, B. Dou¸cot, D. Feinberg, and R. M´elin, Phys. Rev. Lett. 106, 257005 (2011). 29 T. Jonckheere, J. Rech, T. Martin, B. Dou¸cot, D. Feinberg, and R. M´elin, Phys. Rev. B 87, 214501 (2013). 30 J. C. Cuevas and H. Pothier, Phys. Rev. B 75, 174513 (2007). 31 E. Eriksson, R.-P. Riwar, M. Houzet, J. S. Meyer, and Y. V. Nazarov, Phys. Rev. B 95, 075417 (2017). 32 O. Deb, K. Sengupta, and D. Sen, Phys. Rev. B 97, 174518 (2018). 33 P. Virtanen, J. Zou, I. Sosnin, V. T. Petrashov, and T. T. Heikkila, Phys. Rev. Lett. 99, 217003 (2007). 34 C. Padurariu, T. Jonckheere, J. Rech, R. M´elin, D. Fein- berg, T. Martin, and Y. V. Nazarov, Phys. Rev. B 92, 205409 (2015). Nat. Commun. 7, 11167 (2016). 38 E. Strambini, S. D'Ambrosio, F. Vischi, F. S. Bergeret, Y. V. Nazarov, and F. Giazotto, Nature Nanotechnology 11, 1055 (2016). 39 R. M´elin, D. Feinberg, and B. Dou¸cot, Eur. Phys. J. B 89, 67 (2016). 40 Y. Cohen, Y. Ronen, J.-H. Kang, M. Heiblum, D. Feinberg, R. M´elin, and H. Shtrikman, PNAS 115, 6991 (2018). 41 Y. Cohen, Y. Ronen, J.-H. Kang, M. Heiblum, D. Feinberg, R. M´elin, and H. Shtrikman, arXiv:1606.08436v2 (2016), this version includes more extensive study of the three ter- minal junction as compared to the published version Ref. [36]. 42 A. H. Pfeffer, J. E. Duvauchelle, H. Courtois, R. M´elin, D. Feinberg, and F. Lefloch, Phys. Rev. B 90, 075401 (2014). 43 M. Houzet and P. Samuelsson, Phys. Rev. B 82, 060517 (2010). 44 J. Lantz, V. S. Shumeiko, E. Bratus, and G. Wendin, Phys. Rev. B 65, 134523 (2002). 45 R.-P. Riwar, D. M. Badiane, M. Houzet, J. S. Meyer, and Y. V. Nazarov, Physica E: Low-dimensional Systems and Nanostructures 76, 231 (2016). 46 S. Gazibegovic, D. Car, H. Zhang, S. C. Balk, J. A. Logan, M. W. A. de Moor, M. C. Cassidy, R. Schmits, D. Xu, G. Wang, P. Krogstrup, R. L. M. Op het Veld, K. Zuo, Y. Vos, J. Shen, D. Bouman, B. Shojaei, D. Pennachio, J. S. Lee, P. J. van Veldhoven, S. Koelling, M. A. Verheijen, L. P. Kouwenhoven, C. J. Palmstrøm, and E. P. A. M. Bakkers, Nature 548, 434 (2017). 47 S. Datta, P. Bagwell, and M. Anantram, Department of Electrical and Computer Engineering Technical Reports (1996). 48 M. P. Nowak, M. Wimmer, and A. R. Akhmerov, Zenodo (2019), Multiple Andreev Reflections in a three terminal Josephson junction -- numerical current calculation, doi: 10.5281/zenodo.2552156. 49 T. Itoh, Phys. Rev. B 52, 1508 (1995). 50 C. W. J. Beenakker, Phys. Rev. Lett. 67, 3836 (1991).
1703.05875	1	1703	2017-03-17T03:20:13	Negative Differential Resistance and Steep Switching in Chevron Graphene Nanoribbon Field Effect Transistors	[ "cond-mat.mes-hall" ]	Ballistic quantum transport calculations based on the non-equilbrium Green's function formalism show that field-effect transistor devices made from chevron-type graphene nanoribbons (CGNRs) could exhibit negative differential resistance with peak-to-valley ratios in excess of 4800 at room temperature as well as steep-slope switching with 6 mV/decade subtheshold swing over five orders of magnitude and ON-currents of 88$\mu$A/$\mu$m. This is enabled by the superlattice-like structure of these ribbons that have large periodic unit cells with regions of different effective bandgap, resulting in minibands and gaps in the density of states above the conduction band edge. The CGNR ribbon used in our proposed device has been previously fabricated with bottom-up chemical synthesis techniques and could be incorporated into an experimentally-realizable structure.	cond-mat.mes-hall	cond-mat	a Negative Differential Resistance and Steep Switching in Chevron Graphene Nanoribbon Field Effect Transistors Samuel Smith, Juan-Pablo Llin´as, Jeffrey Bokor, Sayeef Salahuddin1 University of California, Berkeley, Department of Electrical Engineering and Computer Sciences, Berkeley, CA 94720 Ballistic quantum transport calculations based on the non-equilbrium Green's function formalism show that field-effect transistor devices made from chevron-type graphene nanoribbons (CGNRs) could exhibit negative differential resistance with peak-to-valley ratios in excess of 4800 at room temperature as well as steep- slope switching with 6 mV/decade subtheshold swing over five orders of magnitude and ON-currents of 88 µA µm−1. This is enabled by the superlattice-like structure of these ribbons that have large periodic unit cells with regions of different effective bandgap, resulting in minibands and gaps in the density of states above the conduction band edge. The CGNR ribbon used in our proposed device has been previously fabricated with bottom-up chemical synthesis techniques and could be incorporated into an experimentally-realizable structure. In 1970, L.Esaki and R. Tsu predicted1 that in an ap- propriately made superlattice, it should be possible to obtain very narrow width bands, which could then lead to negative differential resistance. The remarkable prop- erty of these superlattices is in the fact that, unlike the Esaki diodes, this negative differential resistance does not need any tunneling, rather it comes from the direct con- duction of electrons. Nonetheless, significant difficulty in synthesizing atomically precise, eptiaxial heterostruc- tures has made it very challenging to realize such su- perlattice structures2–8. Much work has been done on modeling graphene nanoribbon heterostructures and su- perlattices which could exhibit NDR9–15. Other work has been done on steep slope devices based on GNR and CNT heterojunctions16,17. Gnani et al. showed how su- perlattices could be used in a III-V nanowire FET to achieve steep slope behavior by using the superlattice gap to filter high energy electrons in the OFF state18. Here, we show that the recently synthesized chevron nanoribbons19 provides a natural, monolithic material system where narrow-width energy bands and negative differential resistance (NDR) can be achieved. Our atom- istic calculations predict that the NDR behavior should manifest at room temperature along with sub-thermal steepness (<60 mV/decade at room temperature). Such NDR behavior could lead to completely novel devices for next generation electronics. Unlike a graphene sheet, a narrow strip etched out of graphene, often called a graphene nanoribbon (GNR), can provide a sizeable bandgap. As a result, GNRs could lead to devices with good ON/OFF ratio at the nanoscale. However, a number of studies have also shown the deleterious effect of edge roughness on the device performance20,21. Recent breakthroughs in bottom-up chemical synthesis can produce GNRs with atomistically pristine edge states and overcome this shortcoming19. In fact, a recent experimental work demonstrated work- ing transistors with 9- and 13-AGNRs made with these techniques22. The methods used to synthesize these ribbons can also be used to generate complex pe- riodic structures beyond simply armchair and zigzag nanoribbons23,24. In this work, we will consider one of those structures, the chevron graphene nanoribbon (CGNR). Fig. 1 shows both the atomic structure of the 6-9 CGNR originally fabricated by Cai et al. and the elec- tronic structure calculated through a pz orbital-based tight-binding method19. A key feature of the band struc- ture is the presence of minibands with regions of forbid- den energy above the conduction band edge, such as those seen in superlattices of III-V semiconductors. Analogous to III-V superlattices, the CGNR contains regions of dif- ferent effective bandgap. When we look at the CGNR in Fig. 1, we see that its narrowest segment is 6 car- bon atoms across and its widest segment is 9 carbon atoms across, with both segments having armchair-type edges. Using a pz-basis set (GW25), the bandgap, Eg, of a 6-AGNR is 1.33 eV (2.7 eV) and the bandgap of a 9- AGNR is 0.95 eV (2.0 eV). However, given the very short length scale over which the width changes in our struc- ture (∼ 1 nm), one would not expect the system to behave as though the local effective potential oscillates between the bulk values of Eg for the isolated AGNRs. In fact, our chevron structure has an overall bandgap of 1.59 eV. This value is consistent with the 1.62 eV bandgap from LDA DFT calculations, but significantly smaller than the 3.74 eV value from calculations incorporating the GW correction26. Both LDA and GW calculations show the presence of minibands and gaps above the conduction band edge26. The structure of the simulated device is shown in in Fig. 2. Like a typical MOSFET, our superlattice field- effect transistor (SLFET) can be turned ON and OFF with a gate voltage at low drain biases. Operation dif- fers from a MOSFET in two key ways. The first is that the device shows NDR with respect to the drain volt- age. At some value of Vds determined by the width of the first miniband, Id decreases substantially when the conduction band at the source becomes aligned with the superlattice gap at the drain. At higher drain bias, cur- rent increases again when the conduction band at the source is aligned with the second miniband at the drain. The second feature of the SLFET is that the superlat- tice gap at the drain filters out higher energy electrons from the first miniband at the source when a source-drain bias is applied. This cuts off the higher energy portion of the thermionic tail at the source, which would contribute to leakage current in a traditional MOSFET. This filter- ing does not, however, affect the low-energy electrons, which carry most of the ON state current as they are in the window where the first minibands at the source and drain overlap. Transport in an SLFET is entirely intra- band like a MOSFET, whereas a TFET relies on band- to-band tunneling. This could possibly allow higher ON current than a TFET. The CGNR used in our simulation has a width of 1.9 nm. The simulation domain is approximately 70 nm long, and the gate has a length of 15 nm. The source and drain are doped with ND = 1.0 × 1013 cm−2 donors. An ef- fective oxide thickness of 1.0 nm is used for both the top and bottom gates. The gate contacts are extended 30 nm in the direction perpendicular the channel to capture the effects of fringing gate fields. While our simulation uses an effective doping density to align the source and drain Fermi levels to the CGNR conduction band, the same effect could be achieved in an experiment through elec- trostatics alone. Simulations are performed using the non-equilibrium Greens function (NEGF) formalism27. A simple pz-basis is used for the Hamiltonian for the chevron graphene nanoribbon with the hopping parameter set to t0 = 2.7 eV. Charge and current are calculated with the recur- sive Greens function algorithm28, and contact self-energy is computed with the Sancho-Rubio iteration scheme29. The NEGF equations are solved self-consistently with FIG. 1. Band structure of a chevron graphene nanoribbon based on a pz orbital basis set. The width (and thus quantum confinement) varies across the unit cell, giving a superlattice- like band structure. Forbidden energies are highlighted in red. The bandgap of the ribbon is 1.59 eV, the first con- duction band has a bandwidth of 0.272 eV, and the first gap between minibands is 0.178 eV. Inset: Molecular structure of the chevron nanoribbon. 2 FIG. 2. Artistic rendering of double-gate CGNR on insulator SLFET. Parts of the top gate and oxide region have been cut away so that the channel is visible. When a gate voltage is applied to turn the device ON, current conduction occurs at low values of Vds where the first miniband at the source is aligned with the first miniband at the drain. As the drain voltage is increased beyond qEmb1, the bandwidth of the first miniband, transmission from source to drain is cut off and the device exhibits negative differential resistance. the Poisson equation for electrostatics in three dimen- sions. Our simulator solves the nonlinear Poisson equa- tion using the predictor-corrector scheme described by Trellakis et al. using a semiclassical approximation for the charge density30. The geometry of the system is mod- eled using a tetrahedral finite element mesh generated with the SALOME package31. The solution of the final sparse matrix form of the Poisson equation discretized the with finite element method is performed using the conjugate gradient solver from the Eigen library32. The local density of states for the CGNR MOSFET is shown in Fig. 3 for several biasing conditions. Fig. 3b shows the case for peak current for the device when a large enough drain bias has been applied to generate enough splitting between the source and drain Fermi lev- els to allow significant current to flow, but not a high enough bias to move the the first miniband outside of the current conduction window. For higher bias as in Fig. 3c, intraband conduction from the first miniband is com- pletely cut off. As the drain bias is further increased, cur- rent can only flow due to a band-to-band tunneling from the first miniband at the source to the second miniband at the drain. Note that, due to the minibands, there will be regions of operation for both gate and drain voltages where current flow is abruptly turned on or off, as the overlap between source minibands and drain minibands is modified. This leads to a steep subthreshold swing (< 60 mV/decade at room temperature) in the Id − Vg 1.00.50.00.51.0k (normalized)21012E (eV)EnergyTransmission (Vs = 0) Transmission (Vds= 0) 2121Transmission (Vds> qEmb1) Current at low VdsReflection at higher VdsSourceDrainEmb13 (a) Vgs = 0.55 V, Vds = 0.10 V (b) Vgs = 0.70 V, Vds = 0.10 V (c) Vgs = 0.70 V, Vds = 0.3 V FIG. 3. Local density of states for several different biasing conditions. Fig. (a) shows the OFF state, in which leakage current is substantially reduced because the superlattice gap in the drain region filters higher energy electrons, which could otherwise travel over the source-side barrier. Fig. (b) shows the ON state, in which current is primarily carried by lower energy carriers, which are not blocked through the density of states filtering at the drain. Fig. (c) shows the ON state for a higher value of Vds. Significant ballistic transport from source to drain is no longer possible when the drain voltage is greater than the width of the first miniband minus the height of the source-side barrier. The colormap is based on a logarithmic scale. it should be possible to achieve reasonable ON current with a low supply voltage in devices of this type. The origin of the steep-slope behavior can be understood from Fig. 3. In the OFF state shown in Fig. 3a, the super- lattice gap at the drain prevents leakage current from flowing over the source-side injection barrier. The states near the top of the barrier are seen to decay rapidly in the drain region. Fig. 3b shows the ON state, in which low-energy electrons, which make up virtually all of the ON current, can flow unimpeded from source to drain. The Id − Vds curves from the results of our simulation are shown in Fig. 5. Considering the case when Vgs = 0.7 V, we see an increase in current up to Vds = 0.10 V. As the drain bias is further increased, we see a decrease in current as the drain miniband goes out of alignment with the source miniband. The current beigins to pick up again as the second miniband at the drain starts to come in alignment with the source miniband again. The peak- to-valley ratio (PVR) at this gate voltage is 4.88 × 103. At Vgs = 0.60 V, the calculated PVR is 1.71× 108. Note, however, that this value is expected to much smaller in a practical device due to the electron-phonon scattering mechanisms28 that were not taken into account in our ballistic simulations. In summary, we have shown that chevron graphene nanoribbon devices can exhibit both steep-slope sub- threshold behavior and negative differential resistance. Both properties are the result of the superlattice-like elec- tronic structure of the ribbon. CGNR SLFETs could be promising for a number of applications ranging from low-power logic transistors to high speed oscillators. A major obstacle to building a real device is making con- tacts with appropriate Schottky barrier heights to be able to match the band alignment conditions achieved in this work through a simple doping model. The per- formance of a real device would also likely be impacted by scattering mechanisms we have not considered here, FIG. 4. Id − Vgs plot for different values of Vds. Steep-slope behavior is observed with a subtheshold swing of around 6 mV/decade over five orders of magnitude around Vgs = 0.6 V for the case when Vds = 0.1 V. Negative differential resistance is evident in that the peak value of Id is lower for higher values of Vds. characteristic and a negative differential resistance in the Id − Vd characteristic. We shall first discuss the current vs gate voltage (Id − Vg) characteristics. Fig. 4 shows Id vs. Vgs for several values of Vds. While steep slope behavior is ex- hibited at some point for all values of Vds, the highest ON current is obverved for Vds = 0.1 V. At this drain bias, an ON current of 168 nA (88 µA µm−1) is achieved at a gate bias of Vgs = 0.75 V. In the steep slope region of this curve, the subthreshold swing is 6 mV/decade when aver- aged over five orders of magnitude of Id. With gate work function engineering and additional device optimization, 010203040506070x (nm)0.40.20.00.20.40.60.8E (eV)010203040506070x (nm)0.40.20.00.20.40.60.8E (eV)010203040506070x (nm)0.40.20.00.20.40.60.8E (eV)0.50.60.7Vgs (V)10-1210-910-610-3100103Id (nA)Vds=0.10 VVds=0.20 VVds=0.30 VVds=0.40 V4 5G. Bernstein and D. Ferry, Journal of Vacuum Science & Tech- nology B: Microelectronics Processing and Phenomena 5, 964 (1987). 6H. Grahn, R. Haug, W. Muller, and K. Ploog, Physical review letters 67, 1618 (1991). 7J. Kastrup, H. Grahn, K. Ploog, F. Prengel, A. Wacker, and E. Scholl, Applied physics letters 65, 1808 (1994). 8A. Warren, D. Antoniadis, H. I. Smith, and J. Melngailis, IEEE electron device letters 6, 294 (1985). 9G. J. Ferreira, M. N. Leuenberger, D. Loss, and J. C. Egues, Physical Review B 84, 125453 (2011). 10S. Li, C. K. Gan, Y.-W. Son, Y. P. Feng, and S. Y. Quek, Applied Physics Letters 106, 013302 (2015). 11H. Teong, K.-T. Lam, S. B. Khalid, and G. Liang, Journal of Applied Physics 105, 084317 (2009). 12H. Sevin¸cli, M. Topsakal, and S. Ciraci, Physical Review B 78, 245402 (2008). 13M. Sharifi, E. Akhoundi, and H. Esmaili, Journal of Computa- tional Electronics , 1 (2016). 14G. Saha, A. K. Saha, and A. H.-u. Rashid, in Nanotechnology (IEEE-NANO), 2015 IEEE 15th International Conference on (IEEE, 2015) pp. 440–443. 15E. C. Girao, E. Cruz-Silva, and V. Meunier, ACS nano 6, 6483 (2012). 16S. Kim, M. Luisier, T. B. Boykin, and G. Klimeck, Applied Physics Letters 104, 243113 (2014). 17Y. Yoon and S. Salahuddin, Applied Physics Letters 97, 033102 (2010). 18E. Gnani, P. Maiorano, S. Reggiani, A. Gnudi, and G. Baccarani, in Electron Devices Meeting (IEDM), 2011 IEEE International (IEEE, 2011) pp. 5–1. 19J. Cai, P. Ruffieux, R. Jaafar, M. Bieri, T. Braun, S. Blanken- burg, M. Muoth, A. P. Seitsonen, M. Saleh, X. Feng, K. Mullen, and R. Fasel, Nature 466, 470 (2010). 20G. Fiori and G. Iannaccone, IEEE Electron Device Letters 28, 760 (2007). 21Y. Yoon and J. Guo, Applied Physics Letters 91, 073103 (2007). 22J. P. Llin´as, A. Fairbrother, G. Barin, P. Ruffieux, W. Shi, K. Lee, B. Y. Choi, R. Braganza, N. Kau, W. Choi, Z. Pedram- razi, T. Dumslaff, A. Narita, X. Feng, K. Mullen, F. Fischer, A. Zettl, P. Ruffieux, E. Yablonovitch, M. Crommie, R. Fasel, and J. Bokor, arXiv preprint arXiv:1605.06730 (2016). 23J. Cai, C. A. Pignedoli, L. Talirz, P. Ruffieux, H. Sode, L. Liang, V. Meunier, R. Berger, R. Li, X. Feng, K. Mullen, and R. Fasel, Nature nanotechnology 9, 896 (2014). 24Y.-C. Chen, T. Cao, C. Chen, Z. Pedramrazi, D. Haberer, D. G. de Oteyza, F. R. Fischer, S. G. Louie, and M. F. Crommie, Nature nanotechnology 10, 156 (2015). 25L. Yang, C.-H. Park, Y.-W. Son, M. L. Cohen, and S. G. Louie, Physical Review Letters 99, 186801 (2007). 26S. Wang and J. Wang, The Journal of Physical Chemistry C 116, 10193 (2012). 27S. Datta, Quantum transport: atom to transistor (Cambridge University Press, 2005). 28R. Lake, G. Klimeck, R. C. Bowen, and D. Jovanovic, Journal of Applied Physics 81, 7845 (1997). 29M. L. Sancho, J. L. Sancho, J. L. Sancho, and J. Rubio, Journal of Physics F: Metal Physics 15, 851 (1985). 30A. Trellakis, A. Galick, A. Pacelli, and U. Ravaioli, Journal of Applied Physics 81, 7880 (1997). 31A. Ribes and C. Caremoli, in Computer Software and Applica- tions Conference, 2007. COMPSAC 2007. 31st Annual Interna- tional, Vol. 2 (IEEE, 2007) pp. 553–564. 32G. Guennebaud B. http://eigen.tuxfamily.org (2010). and Jacob, "Eigen v3," FIG. 5. Id − Vds plot for different values of Vgs. For the case, when Vgs = 0.7 V, a peak-to-valley ratio of 4.88 × 103 is achieved. though the ability to synthesize ribbons with virtually no defects may minimize these effects. Additional op- timization will also likely be necessary to make a func- tioning device. DFT+GW calculations predict a much higher bandgap for the CGNR in vacuum than the tight- binding model used in this work. While surface screening may reduce the bangap somewhat, a wider ribbon with a narrower bandgap may be required. Co-optimization of the bandgap with the bandwidths of the minibands and the gaps between minibands is also a necessary topic for future work. ACKNOWLEDGEMENTS This work was supported by NSF CAREER grant CISE-1149804. Work by JPL and JB was supported in part by the Office of Naval Research BRC program under Grant N00014-16-1-2229. The authors would like to acknowledge useful discussions with Felix Fischer and Michael Crommie. 1L. Esaki and R. Tsu, IBM Journal of Research and Development 14, 61 (1970). 2L. Esaki and L. Chang, Physical Review Letters 33, 495 (1974). 3K. Choi, B. Levine, R. Malik, J. Walker, and C. Bethea, Physical Review B 35, 4172 (1987). 4K. Ismail, W. Chu, A. Yen, D. Antoniadis, and H. I. Smith, Applied physics letters 54, 460 (1989).
1009.1475	2	1009	2010-09-20T13:17:20	Distribution of Persistent Currents in a Multi-Arm Mesoscopic Ring	[ "cond-mat.mes-hall" ]	We propose an idea to investigate persistent current in individual arms of a multi-arm mesoscopic ring. Following a brief description of persistent current in a traditional Aharonov-Bohm (AB) ring, we examine the behavior of persistent currents in separate arms of a three-arm mesoscopic ring. Our analysis may be helpful in studying magnetic response of any complicated quantum network.	cond-mat.mes-hall	cond-mat	Distribution of Persistent Currents in a Multi-Arm Mesoscopic Ring Santanu K. Maiti,1, 2 Srilekha Saha,1 and S. N. Karmakar1 1Theoretical Condensed Matter Physics Division, Saha Institute of Nuclear Physics, 2Department of Physics, Narasinha Dutt College, 129 Belilious Road, Howrah-711 101, India Sector-I, Block-AF, Bidhannagar, Kolkata-700 064, India We propose an idea to investigate persistent current in individual arms of a multi-arm mesoscopic ring. Following a brief description of persistent current in a traditional Aharonov-Bohm (AB) ring, we examine the behavior of persistent currents in separate arms of a three-arm mesoscopic ring. Our analysis may be helpful in studying magnetic response of any complicated quantum network. PACS numbers: 73.23.-b, 73.23.Ra. I. INTRODUCTION Generation of persistent current in a normal metal mesoscopic ring threaded by an AB flux φ has been pro- posed over a number of decades. The appearance of dis- crete energy levels and large phase coherence length allow a non-decaying current in presence of an AB flux φ. Fol- lowing the pioneering work of Buttiker et al. [1], various efforts have been made to explore the basic mechanisms of persistent current in mesoscopic rings and cylinders [2 -- 10]. Later, the existence of non-decaying current in these systems has also been verified through some nice exper- iments [11 -- 14]. The behavior of persistent current in a mesoscopic ring/cylinder can be studied theoretically by several ways as available in the literature [1 -- 10]. In all these available procedures response of the entire sys- tem is achieved only, but no information about individual branches of the system can be explored though it is highly significant to elucidate the actual mechanism of electron transport in a more transparent way. This is the main motivation behind this work. With a brief description of persistent current in a single-channel mesoscopic ring, we discuss elaborately the behavior of persistent currents in individual arms of a three-arm mesoscopic ring. In this three-arm ring system we address an unconventional feature of persistent cur- rent when impurities are introduced only in the middle arm, keeping the other two arms free from any impu- rity. It shows that the current amplitude of the system increases with the increase of impurity strength in the strong impurity regime, while it decreases in the weak impurity regime. This phenomenon is completely differ- ent from traditional disordered systems, and recently, few anomalous features of electron dynamics have also been reported in some other nano materials [15, 16]. In what follows, we present the results. Section II is devoted to present the theory. The results are discussed in Section III. We provide a summary in Section IV. of the elementary flux quantum φ0 = ch/e). A tight- binding (TB) formalism is given for the description of the ring. Within the non-interacting picture the TB Hamil- tonian for a N -site ring looks in the form, H = X i ǫic† i ci + X i v (cid:16)c† i+1ciejθ + c† i ci+1e−jθ(cid:17) (1) where, ǫi is the site energy and v gives the nearest- neighbor hopping integral. Due to the presence of AB flux φ, a phase factor θ = 2πφ/N appears in the Hamil- Φ FIG. 1: A mesoscopic ring threaded by an AB flux φ. tonian. c† i (ci) is the creation (annihilation) operator. The energy Ek corresponding to k-th energy eigenstate ψki can be obtained from the relation Ek = hψkHψki, where ψki = Pp appi. Here pi's are the Wannier states and ap's are the coefficients. Simplifying the above rela- tion we get the expression of the energy Ek as, Ek = X i ǫia∗ i ai + X i v (cid:2)a∗ i+1aiejθ + a∗ i ai+1e−jθ(cid:3) (2) where, a∗ i is the complex conjugate of ai. Here, the sum- mation index i runs over all the atomic sites in the ring. Now, the current carried by the k-th energy eigenstate ψki can be determined from the relation, Ik = − ∂Ek ∂φ = 2πjv N X i (cid:2)a∗ i ai+1e−jθ − a∗ i+1aiejθ(cid:3) . (3) II. THEORETICAL FORMULATION A simple ring: Let us start with Fig. 1, where a meso- scopic ring is subject to an AB flux φ (measured in unit At absolute zero temperature (T = 0K), the net persis- tent current for a ring described with Ne electrons can be determined by taking the sum of individual contributions from the lowest Ne energy eigenstates. Therefore, for Ne electron system the net current becomes I = PNe k=1 Ik. A three-arm ring: Following the above prescription, the distribution of persistent currents in individual arms, viz, upper, middle and lower, of a three-arm mesoscopic ring shown in Fig. 2 can be established. The TB Hamiltonian for this system is in the form, H = X i + X l ǫic† i ci + X i ǫlc† l cl + X l v1 (cid:16)c† i+1ciejθ1 + c† i ci+1e−jθ1(cid:17) v2 (cid:16)c† l+1clejθ2 + c† l cl+1e−jθ2(cid:17)(4) where, the summation index i is used to refer the atomic sites in the upper and lower arms of the ring i.e., in the outer ring, while the index l describes the atomic sites in the middle arm (filled black circles in the framed re- gion). v1 and v2 describe the nearest-neighbor hopping integrals in the outer ring and middle arm, respectively. Φ 1 Φ 2 FIG. 2: A three-arm ring, where the upper and lower sub- rings are subject to AB fluxes φ1 and φ2, respectively. The phase factor θ1 is associated with the hopping of an electron in the upper/lower arm, while in the middle arm it is described by the term θ2. They are represented as, θ1 = (φ1 + φ2)/(NU + NL) and θ2 = (φ1 − φ2)/2NM , where NU , NM and NL represent the total number of single bonds (a bond is formed by connecting two neigh- boring atoms through a line) in the upper, middle and lower arms of the ring, respectively. For this geometry Ek can be calculated according to the same prescription as mentioned above, and, with this energy expression we can determine the currents in individual arms. The ex- pressions for the currents are as follows. For the upper arm: 2 here, the individual contributions of NL bonds are added. Using these relations (Eqs. 5-7), the net persistent cur- rents at T = 0K in individual arms of a three-arm meso- scopic ring containing Ne electrons can be obtained in the same fashion as mentioned earlier. In the present work we perform all the characteristics of persistent current at T = 0K and use the units where c = h = e = 1. Throughout our numerical work we set the nearest-neighbor hopping strengths (v, v1 and v2) to −1 and the energy scale is measured in unit of v. III. NUMERICAL RESULTS AND DISCUSSION A simple ring: As illustrative examples in Fig. 3 we show the current-flux characteristics for a single-channel impurity-free (ǫi = 0 for all i) mesoscopic ring consider- ing N = 20, where (a) and (b) correspond to the results for odd and even number of electrons, respectively. Both for the cases of odd and even Ne, current provides saw- tooth like nature with sharp jumps at half-integer or inte- ger multiples of flux-quantum φ0 depending on whether t n e r r u C ® t n e r r u C ® .7 0 -.7 -1 .7 0 -.7 -1 a -0.5 0 ® Flux 0.5 1 b -0.5 0 ® Flux 0.5 1 FIG. 3: I-φ spectra of an ordered mesoscopic ring (N = 20) with (a) odd Ne and (b) even Ne. The red, green and blue curves in (a) correspond to Ne = 3, 5 and 9, respectively, while in (b) they represent Ne = 4, 6 and 10, respectively. I U k = 2πjv1 NU + NL X i (cid:2)a∗ i ai+1e−jθ1 − a∗ i+1aiejθ1(cid:3) where, contributions from the NU bonds are added. For the middle arm: I M k = πjv2 NM X l (cid:2)a∗ l al+1e−jθ2 − a∗ l+1alejθ2(cid:3) here, net contribution comes from NM bonds. For the lower arm: I L k = 2πjv1 NU + NL X i (cid:2)a∗ i ai+1e−jθ1 − a∗ i+1aiejθ1(cid:3) (5) (6) (7) the ring is described by odd or even number of elec- trons. These currents are periodic in φ exhibiting φ0 flux- quantum periodicity. The presented current-flux charac- teristics for this single-channel ring exactly match with the previous theoretical studies where persistent currents have been calculated by other approaches [1 -- 10]. Thus, making sure with the results for a single-channel ring now we can safely use this procedure to illustrate the current-flux characteristics in individual arms of a three- arm mesoscopic ring. A three-arm ring: Ring without any impurity: In Fig. 4 we present current- flux characteristics for an ordered (ǫi = ǫl = 0 for all i and l) three-arm mesoscopic ring, keeping φ2 constant. These currents are evaluated for a fixed number of elec- trons Ne = 10 and they provide a complex spectra. It is noticed that the responses of the individual branches are quite different from each other, especially, a signif- icant change in amplitude takes place among the cur- rents IU , IM and IL. By adding these three currents we get the net current (IT ) for the entire ring and it be- comes exactly identical to the current determined from the other conventional methods available in the litera- ture [1 -- 10]. This emphasizes the current conservation relation IT = IU + IM + IL. Other important feature -.04 a .24 b U I M I .5 1 -.18 -1 -.5 0 Φ1 .18 c -.04 -1 -.5 0 Φ1 .5 1 .30 d L I T I -.18 -1 -.5 0 Φ1 .5 1 -.30 -1 -.5 0 Φ1 .5 1 3 NM = 11. From the spectra it is observed that when W = 1 (weak), the current amplitude gets reduced in all the three arms compared to the perfect case. The situa- tion becomes completely different for the case of strong disorder i.e., W = 10. In this case the current in the middle arm almost disappears, while for the other arms .18 U I M I L I -.18 -1 -.5 .06 -.06 -1 -.5 .18 -.18 -1 -.5 0 Φ1 0 Φ1 0 Φ1 a .5 1 b .5 1 c .5 1 FIG. 4: Persistent current as a function of φ1 for an ordered three-arm mesoscopic ring considering NU = NL = 10, NM = 9, φ2 = 0.3 and Ne = 10. (a), (b) and (c) correspond to the currents in the upper, middle and lower arms, respectively, while (d) gives the total current for the entire three-arm ring. FIG. 5: Current-flux characteristics for the three different arms of a three-arm mesoscopic ring considering NU = NL = 15, NM = 11, φ2 = 0.5 and Ne = 15, where (a), (b) and (c) correspond to the currents in the upper, middle and lower arms, respectively. The red, green and blue curves in each panel represent W = 0, 1 and 10, respectively. is that the sum of the currents carried by the upper and middle arms becomes exactly identical to the current car- ried by the lower arm i.e., IL = IU + IM which gives an- other relation of current conservation. All these currents exhibit φ0 flux-quantum periodicity. Now, instead of φ1 if we plot the currents as a function of φ2, considering φ1 constant, exactly similar features are observed, like above, satisfying the current conserva- tion relations. Ring with impurities in the middle arm: To introduce impurities in the middle arm we choose the site energies of the atomic sites, filled black circles in the framed region of Fig. 2, randomly from a "Box" distribution function of width W . Here, we determine the currents by taking the average over 50 random disordered configuration in each case to achieve much accurate results. As the middle arm is subject to impurities, while the others are free, we call this system as an ordered- disordered separated three-arm mesoscopic ring. In such a mesoscopic ring an unconventional feature of persis- tent current is observed when we tune the strength of disorder W . To emphasize it, in Fig. 5 we show the vari- ation of persistent current in individual arms for a three- arm mesoscopic ring considering NU = NL = 15 and it achieves a high value. Thus a dramatic change in cur- rent amplitude takes place for the two different regimes of impurity strength. To explore this phenomenon more clearly, in Fig. 6, we show the variation of typical current amplitude as a function of W , where (a), (b) and (c) give the results for the upper, middle and lower arms, respectively, and (d) corresponds to the variation for the entire ring. The typical current amplitude is evaluated from the relation, Ityp = phI 2iφ1,W , where the averaging is done over a complete period of φ1 and 50 random disordered config- urations. From the Ityp-W spectra the effect of disorder is clearly visible. It shows that the current amplitude in the middle arm sharply drops to zero as W is increased. While, for the other two impurity-free arms the current amplitude initially decreases, and reaching to a minimum at a critical value W = Wc (say), it again increases with W . Thus, beyond the critical disorder strength Wc, the anomalous behavior is observed and this phenomenon can be implemented as follows. We consider the three- arm mesoscopic ring as a coupled system combining two sub-systems, one is ordered and other is disordered. In the absence of any coupling among the ordered and disor- dered regions, we can think the entire system as a simple combination of two independent sub-systems. Therefore, we get all the extended states in the ordered region, while the localized states are obtained in the disordered region. In this situation, the motion of an electron in any one re- gion is not affected by the other. But for the coupled system, the motion of the electron is no more indepen- dent, and we have to take the combined effects coming from both the two regions. With the increase of disorder, p y T I ® p y T I ® .12 .06 0 .12 .06 0 a 0 2 4 6 8 10 ® W c 0 2 4 6 8 10 ® W p y T I ® p y T I ® .04 .02 0 .24 .12 0 b 0 2 4 6 8 10 ® W d 0 2 4 6 8 10 ® W FIG. 6: Ityp-W characteristics for a three-arm mesoscopic ring considering NU = NL = 15, NM = 11, φ2 = 0.5 and Ne = 15. (a), (b) and (c) represent the responses for the upper, middle and lower arms, respectively, while (d) corre- sponds to the variation for the entire ring. the scattering effect becomes dominated more and thus the reduction of the current is expected. This scattering is due to the existence of the localized eigenstates in the disordered region. Now in the limit of weak disorder the coupling between the two sub-systems becomes strong and the motion of the electron in the ordered region is highly influenced by the disordered region. Therefore, 4 the scattering effect from both the two regions is quite significant and the current amplitude gets reduced. On the other hand, in the strong disorder regime the coupling between the two sub-systems becomes weak and the scat- tering effect from the ordered region is less significant, and it decreases with W . At the critical value Wc, we get a separation between the much weaker and the strongly localized states. Beyond this value, the weaker localized states become more extended and the strongly localized states become more localized with the increase of W . In this situation, the current is obtained mainly from these nearly extended states which provide the larger current with W in the higher disorder regime. From the above analysis, behavior of the typical cur- rent amplitude for the entire system (Fig. 6(d)) can be clearly understood and it shows the similar variation like the upper and lower arms. IV. CONCLUDING REMARKS To summarize, we have explored an idea to investigate the nature of persistent currents in individual branches of a multi-arm mesoscopic ring. Starting with a brief description of persistent current in a traditional single- channel mesoscopic ring, pierced by an AB flux φ, we have examined the characteristic features of persistent currents in separate arms of a three-arm mesoscopic ring where the upper and lower sub-rings are subject to AB fluxes φ1 and φ2, respectively. Our analysis may provide a basic framework to address magnetic response in indi- vidual branches of any complicated quantum network. In the present paper we have done all the calculations by ignoring the effects of temperature, electron-electron correlation, electron-phonon interaction, etc. We need further study by incorporating all these effects. We are thankful to Prof. S. Sil, Prof. B. Bhat- tacharyya, M. Dey and P. Dutta for useful discussion. [1] M. Buttiker, Y. Imry, and R. Landauer, Phys. Lett. 96A, Peeters, Phys. Rev. B 78, 195315 (2008). 365 (1983). [11] L. P. Levy, G. Dolan, J. Dunsmuir, and H. Bouchiat, [2] H. F. Cheung, Y. Gefen, E. K. Riedel, and W. H. Shih, Phys. Rev. Lett. 64, 2074 (1990). Phys. Rev. B 37, 6050 (1988). [3] B. L. Altshuler, Y. Gefen, and Y. Imry, Phys. Rev. Lett. 66, 88 (1991). [4] A. Schmid, Phys. Rev. Lett. 66, 80 (1991). [5] V. Ambegaokar and U. Eckern, Phys. Rev. Lett. 65, 381 (1990). [6] S. K. Maiti, Physica E 31, 117 (2006). [7] S. K. Maiti, Solid State Phenomena 155, 87 (2009). [8] S. Bellucci and P. Onorato, Physica E 41, 1393 (2009). [9] P. A. Orellana and M. Pacheco, Phys. Rev. B 71, 235330 (2005). [10] L. K. Castelano, G.-Q. Hai, B. Partoens, and F. M. [12] V. Chandrasekhar, R. A. Webb, M. J. Brady, M. B. Ketchen, W. J. Gallagher, and A. Kleinsasser, Phys. Rev. Lett. 67, 3578 (1991). [13] E. M. Q. Jariwala, P. Mohanty, M. B. Ketchen, and R. A. Webb, Phys. Rev. Lett. 86, 1594 (2001). [14] R. Deblock, R. Bel, B. Reulet, H. Bouchiat, and D. Mailly, Phys. Rev. Lett. 89, 206803 (2002). [15] J. X. Zhong and G. M. Stocks, Nano. Lett. 6, 128 (2006). [16] C. Y. Yang, J. W. Ding, and N. Xu, Physica B 394, 69 (2007).
1506.05409	2	1506	2015-12-04T21:01:12	Cooperative phenomenon in a rippled graphene: Chiral spin guide	[ "cond-mat.mes-hall" ]	We analyze spin scattering in ballistic transport of electrons through a ripple at a normal incidence of an electron flow. The model of a ripple consists of a curved graphene surface in the form of an arc of a circle connected from the left-hand and right-hand sides to two flat graphene sheets. At certain conditions the curvature induced spin-orbit coupling creates a transparent window for incoming electrons with one spin polarization simultaneously with a backscattering of those with opposite polarization. This window is equally likely transparent for electrons with spin up and spin down that move in opposite directions. The spin filtering effect being small in one ripple becomes prominent with the increase of N consequently connected ripples that create a graphene sheet of the sinusoidal type. We present the analytical expressions for spin up (down) transmission probabilities as a function of N connected ripples.	cond-mat.mes-hall	cond-mat	Cooperative phenomenon in a rippled graphene: Chiral spin guide M. Pudlak,1 K.N. Pichugin,2 and R.G. Nazmitdinov3,4 1Institute of Experimental Physics, 04001 Kosice, Slovak Republic 2Kirensky Institute of Physics, 660036 Krasnoyarsk, Russia 3Departament de F´ısica, Universitat de les Illes Balears, E-07122 Palma de Mallorca, Spain 4Bogoliubov Laboratory of Theoretical Physics, Joint Institute for Nuclear Research, 141980 Dubna, Russia (Dated: February 1, 2021) We analyze spin scattering in ballistic transport of electrons through a ripple at a normal incidence of an electron flow. The model of a ripple consists of a curved graphene surface in the form of an arc of a circle connected from the left-hand and right-hand sides to two flat graphene sheets. At certain conditions the curvature induced spin-orbit coupling creates a transparent window for incoming electrons with one spin polarization simultaneously with a backscattering of those with opposite polarization. This window is equally likely transparent for electrons with spin up and spin down that move in opposite directions. The spin filtering effect being small in one ripple becomes prominent with the increase of N consequently connected ripples that create a graphene sheet of the sinusoidal type. We present the analytical expressions for spin up (down) transmission probabilities as a function of N connected ripples. PACS numbers: 72.25.-b,71.70.Ej,73.23.Ad I. INTRODUCTION The extraordinary properties of graphene have at- tracted enormous experimental and theoretical attention for a decade (see e.g. Refs.1,2). Graphene being a zero- gap semiconductor has a band structure described by a linear dispersion relation at low energy, similar to mass- less Dirac-Weyl fermions. Such a band structure leads to exceptionally high mobility of charged carriers. A ques- tion of possible mechanisms that would allow us to throt- tle the mobility and, consequently, to control a conduc- tivity is a topical subject in graphene physics, due to its fundamental as well as technological significance. Among various mechanisms that might affect the mo- bility, the scattering that could be induced by a ripple (see, for example, discussion in Ref.3) appears to be the most natural one, since graphene sheets are not perfectly flat. Moreover, periodic ripples can be created and con- trolled in suspended graphene, in particular, by thermal treatment4 and by placing graphene in a especially pre- pared substrate. Indeed, curvature of the surface affects the π orbitals that determine the electronic properties of graphene. It results in enhancement of spin-orbit cou- pling that could serve as a source of spin scattering. We recall that the intrinsic (intraatomic) spin-orbit interac- tion in flat graphene is weak1,2,5. It makes spin deco- herence in such a material weak as well, i.e., scattering due to disorder is supposed to be unimportant. In order to get deep insight into the nature of curvature induced scattering, it is desirable to elucidate among many ques- tions the basic one: What are the distinctive features of curvature induced spin-orbit coupling ? One can further ask how to employ these features to guide an electron transport in a graphene-based system at the theoretical, and, quite likely, practical levels. i.e., physics of graphene have been developed by Ando6 and by others7 -- 9 in the framework of effective mass and tight- binding approximations. Recent measurements in ul- tra clean carbon nanotubes (CNTs)10, in an ex- treme form of curved graphene, revealed the energy splitting that can be associated with spin-orbit cou- pling. The measured shifts are compatible with theo- retical predictions6,9, while some features regarding the contribution of different spin-orbit terms in metallic and non-metallic CNTs are still debatable (see, for example, discussion in11 -- 16). Nowadays, nevertheless, there is a consensus that for armchair CNTs one obtains two SOC terms: one preserves the spin symmetry (a spin projec- tion on the CNT symmetry axis), while the second one breaks this symmetry6,14,15,17. Thus, we have a reliable answer to the first question, at least, for armchair CNTs. In some previous studies6,9,14,15 the role played by the second term was underestimated. In this paper we will attempt to answer how full curvature induced SOC, in- cluding the second term, could be used to create a po- larized spin current with a high efficiency in a rippled graphene system. The structure of this paper is as follows. In Sec. II we briefly discuss the explicit expressions for the eigen spectrum and eigenfunctions of an armchair nanotube with a full curvature induced spin-orbit coupling. By means of these results we introduce a scattering model for one ripple and extend this model for N continuously connected ripples. III we provide a discussion of our results in terms of simple estimates. The main conclusions are summarized in Sec. IV. In Sec. II. SCATTERING PROBLEM A consistent approach to introduce the curvature induced spin-orbit coupling (SOC) in the low energy In order to model a scattering problem on a ripple we consider a curved surface in the form of an arc of a circle connected from the left-hand and right-hand sides to two flat graphene sheets. The solution for flat graphene is well known1,2. The solution for a curved graphene sur- face can be expressed in terms of the results obtained for armchair CNTs in an effective mass approximation, when only the interaction between nearest neighbor atoms is taken into account17. frame, and obtains H ′ = U (θ) H U −1(θ) = Hkin + HSOC , ∂θ(cid:19) , Hkin = −iγ(cid:18)τy ⊗ I∂y + τx ⊗ I 1 R HSOC = −λy τy ⊗ σx − λx τx ⊗ σy . 2 (6) A. Low energy spectrum of the armchair nanotube Let us recapitulate the major results17 in the vicinity of the Fermi level E = 0 for a point K in the presence of the curvature induced spin-orbit interaction in an arm- chair CNT. The y axis is chosen as the symmetry and the quantizations axis. In this case the eigenvalue problem is defined as HΨ = (cid:18) 0 f † 0 (cid:19)(cid:18) F K A F K B (cid:19) = E(cid:18) F K A F K B (cid:19) , f with the following definitions: f = γ(kx − iky) + i δγ ′ 4R σx(~r) − 2δγp R σy , R∂θ , ky = −i ∂ kx = −i ∂ ∂y , σx(~r) = σx cos θ − σz sin θ . (1) (2) Here, σx,y,z are standard Pauli matrices, and the spinors of two sub-lattices are (3) F K A,↑ F K B,↑ F K A = (cid:18) F K B = (cid:18) F K A,↓ (cid:19) , F K B,↓ (cid:19) . The following notations are used: γ = −√3V π ppa/2 = γ0a, γ′ = √3(V σ pp)a/2 = γ1a, p = 1 − 3γ′/8γ (see e.g. Ref.6). The quantities V σ pp are the trans- fer integrals for σ and π orbitals, respectively in a flat graphene; a = √3d ≃ 2.46 A is the length of the prim- itive translation vector, where d is the distance between atoms in the unit cell. pp − V π pp and V π The intrinsic source of the SOC δ = ∆/(3ǫπσ) is de- fined as ∆ = i 3¯h ec2hxl 4m2 ∂V ∂x py − ∂V ∂y pxyli , (4) where V is the atomic potential and ǫπσ = ǫπ energy ǫσ carbon atoms. The energy ǫπ directed perpendicular to the curved surface. 2p. The 2p is the energy of σ-orbitals, localized between 2p is the energy of π-orbitals, 2p− ǫσ By means of the unitary transformation U (θ) = exp(i θ 2 σy) ⊗ I , (5) where I is 2 × 2 unity matrix, one removes the θ depen- dence in Hamiltonian (1), transformed in the intrinsic Here the operators τx,y,z are the Pauli matrices that act on the wave functions of A- and B-sub-lattices (a pseudo- spin space), and λx = γ (1 + 4δp) /(2R) , λy = δγ′/(4R) (7) are the strengths of the SOC terms. In the Hamiltonian (6) the term (∼ λx) conserves, while the other one (∼ λy) breaks the spin symmetry in the armchair CNT. The operator Jy, being an integral of motionh H, Jyi = 0, is defined in the laboratory frame as Jy = I ⊗(cid:18) Ly + σy 2 (cid:19) = I ⊗(cid:18)−i∂θ + σy 2 (cid:19) , while in the intrinsic frame it is Jy → J ′ y = U Jy U −1 = I ⊗ (−i∂θ) . This integral allows to present the wave functions as (8) (9) F ′(θ, y) = eimθeiky yΨ = eimθeiky y  A B C D   . (10) These wavefunctions are also the eigenfunctions of the y ≡ ky. Here, other integral of motion, the operator k′ m = ±1/2,±3/2, .... is an eigenvalue of the angular mo- mentum operator J ′ y. For the components of the eigen- vector F ′(θ, y) the relations A = D and B = C are fulfilled at real values of m and ky. Solving the eigenvalue problem H ′F ′ = EF ′, one ob- tains the eigen spectrum E = ±Em,q , Em,q = qt2 m + λ2 Dm,q = qqλ2 x(cid:0)t2 y + λ2 m + t2 y(cid:1) + t2 yλ2 y , q = ±1 , (11) y + λ2 x + 2Dm,q , where tm = γm/R, ty = γky. B. Scattering model for one ripple Keeping in mind a discussion that will be given here- after, we analyze the following geometry (see the con- struction profile on Fig.1). It consists of one arc of a circle that is connected from the left-hand side to a flat graphene sheet. This (direct) arc is continuously con- nected to the inverse arc of the same radius that is con- nected to the right-hand flat graphene sheet. We put the L I II R z θ φ R θ0 x FIG. 1: The rippled graphene system. origin of the coordinate at the center of the direct arc of the circle. To give better insight into the scattering phenomenon in our model of a ripple, we study first only the direct arc of the circle connected to two flat surfaces. Two flat surfaces are: i)the region L, defined in the intervals −∞ < x < −R cos θ0; the region R, defined in the inter- vals R cos θ0 < x < ∞. The region I is a part of a nan- otube of radius R, defined as −R cos θ0 < x < R cos θ0. At θ0 = 0, the ripple is a half of the nanotube, while at θ0 = π/2 the ripple does not exist. For the sake of anal- ysis we introduce the angle φ = π − 2θ0. To describe the scattering phenomenon one has to define wave functions in different regions: flat (L,R) and curved (I) graphene surfaces. Regions L and R are described by the Hamiltonian H0 = γ (τxkx + τyky) ⊗ I (12) that does not mix spin components. For the sake of sim- plicity, we consider the electron motion at the normal in- cidence, with the electron wave vector ~k = (kx, 0). One solves the stationary Schrodinger equation H0Ψ = E0Ψ and obtains the corresponding eigenstates 0 (kx) , E0 = ±γkx , Ψ = exp (ikxx)Ψσ 2 (cid:18) sign(γkxE0)Φσ 0 , Φσ (cid:19) , iσ (cid:19) , σ = ±1. 0 = (cid:18) 1 Φσ 0 0 Ψσ 0 (kx) = 1 σyΦσ 0 = σΦσ (13) (14) (15) (16) Evidently, the wave functions in regions L, R, can be written as a superposition of all possible solutions for flat graphene. To proceed, with the aid of eigenspinors (15), (16), we introduce an auxiliary matrix M0 (4 × 4) for a given value of energy at the normal incidence 0 (Kx) , Ψ−1 0 (Kx) , Ψ+1 0 (−Kx) , Ψ−1 M0 = (cid:0)Ψ+1 0 (−Kx)(cid:1) . (17) Here, we define the variable Kx = sign(γE0)kx to ensure that the first two columns of the matrix M0 correspond to eigenstates that move in a positive x-direction, while the last two columns correspond to eigenstates that move in a negative x-direction. 3 ) V e ( y g r e n E 0.8 0.6 0.4 0.2 0 -1 m1 0 m m −1 1 FIG. 2: (Color online) The spectrum (19) at ky = 0 as a function of the quantum number m. Dashed (red) and solid (blue) lines are associated with states characterized by the quantum number ms=−1 and ms=+1, respectively. The val- ues of ±ms at the energy EF = 0.2eV (solid horizontal line) are indicated by vertical dashed lines. The parameters of cal- culations are R = 10A, δ = 0.01, p = 0.1, γ = 9 2 1.42AeV, γ′ = γ 8 4R = 0.0043eV. The arrow indicates the gap displayed on the insert. R (1/2 + 2δp) = 0.32eV, λy = δγ ′ 3 , λx = γ The matrix M0 is unitary, i.e., M −1 0 . It allows us to define a general form of the wave function ΨL,R for a flat graphene 0 = M † ΨL,R(x) = M0 exp(cid:16)i K(x − xL,R)(cid:17) CL,R , (18) where K = diag(Kx, Kx,−Kx,−Kx) is a diagonal ma- trix, xL,R are x-coordinates where flat and curved sur- faces are connected, and CL,R are corresponding vectors with four unknown yet, normalized coefficients in each region. Note that we do not consider inelastic scattering. Therefore, since the electron energy is conserved, we use the same vector ~k = (kx, 0) for the left and right flat graphene surfaces. For the curved surface we use eigenspinors of the Hamiltonian (6). The general form of these eigenspinors is defined in the intrinsic frame17. Therefore, we apply the inverse transformation (5) to these eigenspinors in or- der to analyze the scattering problem in the laboratory frame. At ky = 0 the spectrum (11) and the eigenspinors are particularly simple EA = ±qt2 m + λ2 y + sλx , s = ±1 , Ψ = exp (imθ) Ψs (19) (20) A(tm) , A(tm) Ψs σyΦs A(tm) = (cid:18) −sΦs 2 σy(cid:1)(cid:18) A(tm) (cid:19) , λy − (sEA − λx) (cid:19) . (22) −itm (21) Φs A(tm) = exp(cid:0)−i θ Note that energies in flat graphene and in a curved sur- face are different E0 = EA+A( a R )2 (see details in Ref.18). This effect is caused by different hybridizations of π elec- trons in flat graphene and a graphene -- based system with curvature. In particular, A = 5/6, 7/12 (eV) in the arm- chair and zig-zag nanotubes, respectively. At a fixed energy of the electron flow E0 ⇔ EA, Eq.(19) yields four possible values of the quantum num- ber m m ⇒ ms = ± R γ q(sEA − λx)2 − λ2 y, s = ±1. (23) Since the angular momentum is not longer the integral of motion, we have to consider the mixture of eigenfunctions with all possible values of m at a given energy. As an example of the spectrum (19), a few positive energy branches as a function of the quantum number m are shown in Fig.2. The branches are distinguished by the index s = ±1. There is an anti-crossing effect between energy states characterized by the same ms=+1 quantum number. This anticrossing is brought about by the interaction (∼ λy) that breaks the spin symmetry (see Sect.IIA) in the curved graphene surface. It results in a gap of = 2λy near EA = λx indicated by the ar- row (see the inset on Fig.2). Similar gap occurs near EA = −λx for the m-states with index s = −1. As a consequence of these gaps, evanescent modes arise at en- ergies λx − λy < E < λx + λy in our system. For the sake of illustration the positive spectrum (19) of m-states is crossed by the horizontal line that mimics the Fermi energy. The crossing points determine quantum num- bers m that have non-quantized values when the curved surface (arc of circle) is connected to the flat one. With the aid of eigenspinors (21), (22), and the unitary transformation (5), we introduce an auxiliary matrix for a given value of energy at a curved surface MA(θ) = A(tm1 ) , Ψ−1 A (tm−1 ) , Ψ1 (cid:0)Ψ1 A(−tm1 ) , Ψ−1 A (−tm−1 )(cid:1) = (24) = U (−θ) MA(0). is a vector of As a result, in region I the wave function can be writ- ten as a superposition of all solutions for a curved sur- face in the form ΨI(θ) = MA(θ) exp(i mθ)CI . Here, CI four unknown coefficients, m = diag(m1, m−1,−m1,−m−1) is a diagonal matrix. The overlap of eigenspinors of the flat and bended re- gions can be readily calculated with the aid of Eqs.(15), (21), which results in (Ψσ 0 )† (Ψs × (Φσ A) ≃ (−sign(γkxE0)s + σ) × 0 )† exp(cid:0)−i θ 2 σy(cid:1) (Φs A) . (25) Evidently, the overlap is zero at σ = sign(γkxE0)s. Note that already this result implies that some of the four channels between the flat and curved regions could be closed. Matching the wave functions at the boundaries of re- gions L, I, and R, for an incoming electron flow from the left-hand side, leads us to the following equations r (cid:19) = MA(−φ/2)CI , ΨL(xL) = ΨI (−φ/2) ⇒ M0(cid:18) Φin ΨR(xR) = ΨI (+φ/2) ⇒ M0(cid:18) t 0 (cid:19) = MA(+φ/2) exp(i mφ)CI . 4 (26) (27) We recall that the angles θ0 and φ determine the xL,R coordinates: xL = R cos(θ0 + φ), xR = R cos θ0. Here, reflection coefficients, respectively, for incoming electron ↑ t(L)in ↑ r(L)in ↓ (cid:19) and r = (cid:18) r(L)in t = (cid:18) t(L)in either with a spin up ↑i ≡ (cid:18) 1 ↓i ≡ (cid:18) 1 −i (cid:19). ↓ (cid:19) are transmission and i (cid:19) or with a spin down Solutions of Eqs.(26) (and similar equations for an in- coming electron flow from the right-hand side) yield the following probabilities ↑2 = t(R)↓ ↓2 = t(R)↑ t(L)↑ t(L)↓ ↓2 = r(R)↓ ↑2 = r(R)↑ ↓2 = ↑2 = ↑2 = 1 − ↓2 = 1 − r(L)↑ r(L)↓ 1 1 + (z−1)2 1 1 + (z+1)2 1 1 + (z−1)2 1 1 + (z+1)2 . (28) (29) (30) (31) Here, we have also introduced the variable zs zs = λy sin(msφ) tms = λyR γ × sin(msφ) ms , s = ±1 , (32) related to the characteristics of the curved surface (see Sec.IIA). ↑2 = r(L)↓ Evidently, there is no backscattering for incoming elec- trons, if λy = 0 [see Eqs.(30)-(32)]. However, at λy 6= 0 backscattering with a spin inversion takes place. The reflection probabilities without the spin inversion are r(L)↑ ↓2 = 0. The same is true for the trans- mission probabilities with a spin inversion, i.e., t(L)↑ ↓2 = t(L)↓ ↑2 = 0. Thus, backscattering with a spin inversion is nonzero in the ripple due to the curvature induced SOC produced by the λy-term. In addition, incoming electrons with different spin orientations choose different channels (different ms, s = ±1). takes place at the condition The maximum transmission probability t(L)↑ ↑2 = 1 m−1φc = πn , n = 1, 2, . . . , (33) [see Eqs.(28),(32)]. Evidently, this probability becomes dominant at the minimum transmission t(L)↓ ↓2. The lowest minimum of the transmission t(L)↓ ↓2 occurs at the condition EA = λx, when m+1 becomes imaginary [see Eq.(23)]. In other words, the propagating mode trans- forms to the evanescent mode for the channel m+1. Tak- ing into account the condition EA = λx in Eq.(23), one obtains the critical angle of the curved surface (in form of the arc) for a maximum of spin up filter efficiency φc = πn m−1 = πnγ Rq4λ2 x − λ2 y , (34) where the SOC strengths λx,y are defined by Eq.(7). For parameters listed in the caption of Fig.2 we have φc ≈ 0.996π(n = 1). For the same critical angle φc and EA = −λx we obtain a maximum for the spin down filter efficiency, when m−1 becomes imaginary. Thus, there are different channels for the spin up and spin down electron (hole) flows. Note that the deviation from the energy value EA = ±λx could produce the equal transmission for spin up and spin down electrons (see Fig.3.a). Therefore, it is important to choose the energy EA to be in the close vicinity of the energy value ≃ λx. For the considered parameters the filter efficiency is, however, small. So far this result has met with only limited success. C. Scattering model for N ripples To increase the efficiency we suggest connecting the bent parts sequentially, as shown in Fig. 1. In particular, the construction with the direct+inverse arcs (with the same angle φ) transforms Eqs.(26)(27), to the forms r (cid:19) = MA(−φ/2)CI , 0 (cid:19) = MA(π − φ/2) exp(−i mφ) × (35) M0(cid:18) Φin M0(cid:18) t M −1 A (φ/2 − π) MA(+φ/2) exp(+i mφ)CI . (36) Since in the inverse arc the phases, accumulated from the point of connection with the direct arc to the point of connection with a straight line ( flat graphene), have a sign opposite that of the first one, we use exp(±i mφ). Matching the wave functions at the boundaries of re- gions L, I, II, and R, for electron coming from the left- hand (L) and right-hand (R) sides of the construction, leads us to the following nonzero probabilities 5 Thus, the transmissions through one and two (di- rect+inverse) arcs are accompanied by the inverse backscattering. The considered cases imply that the larger the number of arcs is, the stronger the inverse backscattering is for one of the spin components. Following the recipe described in Ref.19 (interfering Feynman paths), and combining S-matrices for N con- nected arcs, we obtain ↑2 =   ↓2 =   2 N   N   2 = t(R)↓ ↓2 , (41) = t(R)↑ ↑2 . (42) t(L)↑ N 2 + C(−) −1 C(+) −1 t(L)↓ C(+) +1 Here, the variable C(±) s N 2 + C(−) +1 is defined as (43) C(±) s = p1 + (zs)2 ± zs , s = ±1. Evidently, at zs = 0 the transmission probability is 1 for any number of arcs, while zs 6= 0 leads to the decrease of the transmission probability with the increase of the number of arcs. The suppression is, however, different for various transmission probabilities due to their different dependence on the quantum number ms. As shown above, conditions (33), (34), determine the dominance, in particular, of the transmission probability of spin up incoming electrons at EA > 0. Indeed, a set N ≫ 1 of an exact replica of the consistently connected ripples (see Fig.1) does not affect this dominance (= 1) for the ms=−1 channel. However, this set suppresses the spin down transmission probability for the ms=+1 chan- nel that is proportional to x < 1 ⇒ xN → 0. We would like to point out that Eqs.(41),(42), are valid for odd and even number of consistently connected rip- ples. In our model the only requirement is that the di- rect ripple has to be connected to the inverse one, the inverse ripple to the direct one etc. From our consider- ation it follows that, if at a certain energy, for example, EA > 0 there is a high transmission probability for spin- up electrons from the left side of our system, one obtains the same magnitude for the transmission probability for spin-down electrons from the right side. III. DISCUSSION A. N-factor = t(R)↓ ↓2 , 1 1 1 + 2(z−1)2(cid:21)2 1 + 2(z+1)2(cid:21)2 ↑2 = (cid:20) ↓2 = (cid:20) ↓2 = 1 − t(L)↑ ↑2 = 1 − t(L)↓ = t(R)↑ ↑2 , ↓2 . ↑2 = r(R)↓ ↓2 = r(R)↑ ↑2 , t(L)↑ t(L)↓ r(L)↑ r(L)↓ (37) (38) (39) (40) To obtain a simple picture of the physics behind the enhancement of the spin filtering effect, let us consider the transmission at the energy EA ≃ λx, when m+1 be- comes imaginary [see Eq.(23)] and the propagating mode transforms to the evanescent mode for the channel m+1. In light of Eqs.(23),(7), one obtains m+1 = i R γ λy = ix , x = δγ′ 4γ ≈ 0.007 (44) 1 n o i s s i m s n a r T 0.9999 0.9998 0.9997 0.9996 0.9995 -1 1 n o i s s i m s n a r T 0.95 0.9 0.85 (a) 1 0 EA(eV ) 0.8 -0.5 0 EA(eV ) (b) 0.5 1 0.8 0.6 0.4 0.2 n o i s s i m s n a r T -0.5 0 EA(eV ) 1 0.8 0.6 0.4 0.2 n o i s s i m s n a r T 0 -0.5 0 EA(eV ) (c) 0.5 (d) 0.5 ↑2 (blue, dashed lines) and t(L)↓ FIG. 3: (Color online) Dependence of transmission probabili- ties t(L)↑ ↓2 (red, solid lines) on the energy EA at ky = 0 for 1 (a), 20 (b), 100 (c) and 200 (d) sequentially connected ripples (π-arcs). The parameters are the same as in Fig.2. 6 As a result, the variable z+ (Eq.(32)) transforms in the form z+ ≃ xφ ≪ 1 . (45) Taking into account Eqs.(43-45), one can readily estimate that at N ≫ 1 C(+) +1 + C(−) +1 (46) N (47) N ≃ 2 + (N xφ)2 ⇒ 2 + (N xφ)2(cid:21)2 ↓2 ≈ (cid:20) 2 ⇒ t(L)↓ With our choice of parameters and φ ≃ π, this result yields t(L)↓ ↓2 ≪ 1 ⇐⇒ N ≫ 1 xπ ≈ 45 . (48) The illustration of this phenomenon is displayed in Fig.3 for the transmission probabilities through 1, 20, 100 and 200 sequentially connected ripples (π-arcs). Here, we consider the transmission as a function of the curved sur- face energy EA of the incoming electrons (holes). A small difference between spin up and spin down transmission probabilities for one ripple (Fig.3a) at EA > 0 evolves to ≃ 100% efficiency for the spin up transmission prob- abilities for the left-side incoming electron at N = 200 ripples (Fig.3d). The opposite picture takes place for the spin down transmission probabilities at EA < 0. To re- alize such a situation one might use the SiO2 substrate as a gate of the curved surface, which helps control the concentration of charge carriers in graphene. As a result, one can change the charge carrier type from electron to hole20. B. Spin filtering and ripple parameters R and φ In light of the above analysis, without loss of gener- ality, we can consider msφ < 1 in order to observe the suppression effect [see Eq.(48)]. With the aid of Eq.(23), taking into account that λx ≫ λy, this requirement leads to the following inequality λx − γ Rφ < EA < λx + γ Rφ . (49) To remain at the maximum, for example, the transmis- sion probability t(L)↑ ↑2 = 1, it is necessary to fulfill con- dition (33). As a result, in light of Eq.(23) and the con- dition λx ≫ λy, taking into account Eq.(7), one obtains R ≃ γ EA ( π φ − β) , β = 1 + 4δp 2 . (50) Combining this equation with Eq.(49), we have π − 1 2β < φ < π + 1 2β . (51) Thus, Eqs.(50),(51) determine the region of feasibility of the parameters R and φ, where the spin filtering ef- fects could exist at fixed system (graphene) parameters such as γ, δ, and the electron energy EA. From this ob- servation, two arguments follow in favor of our findings. First, even at φ 6= φc (see Eq.(34)) one of the spin com- ponent in the incoming electron (hole) flow is suppressed for a large enough number of ripples at some particular energy region. Second, we assume that all ripples are identical. Practically, the graphene surface is randomly curved, and it is a real challenge to create identical, con- sequently connected ripples. However, it is our belief that modern technology will allow us to realize this situation soon or later. Whatever the case, the spin filtering effect should survive if small variations of radii and angles of consequently connected ripples are subject to conditions (50),(51), at a fixed value EA of the electron energy flow. C. Effect of a finite ky-momentum In our model a single ripple is modeled as part of a nanotube that is infinite in the y direction. Evidently, realistic ripples are limited in space in both the x− and y− directions. In particular, graphene nanoribbons are considered prominent candidates to control the electronic properties of graphene based devices. This issue requires, however, a dedicated study, and is the subject of a forth- coming paper. In order to have some idea of what should be expected in graphene nanoribbons, we analyze the case with a fi- nite ky 6= 0. Nonzero ky could mimic the case of a ripple limited in the y-direction. Indeed, a finite width in the y-direction introduces the quantization of the ky momen- tum on the curved surface. As a result, the eigenspinors at the curved surface would depend on the mixture ±ky values for s = ±1, i.e., altogether four momentum ky (see details in Ref.17). In this case analytical expres- sions are too cumbersome, even in a simple case of one conserved momentum ky on the curved surface. There- fore, we proceed with a numerical analysis that provides a vivid presentation of a simple case with a single value of the ky momentum on the curved surface. Let us suppose that the incoming electron flow pos- sesses a momentum ~k = (kx, ky) in regions L (R). Evi- dently, in this case E0 = ±γqk2 y. For simplicity, we consider E0 > 0, and obtain for the momentum on the curved surface x + k2 ky = ty/γ = E0/γ × sin(α) . (52) The results of the calculations exhibit a degradation of the spin filter ability of our system. At a fixed value of the energy E0 = λx the transmission probability t(L)↑ ↑2 decreases drastically at α ≥ π/8 [see Fig.4(a)]. It seems that the spin filtering effects would survive at α < π/8. Note, however, that this estimation depends on the sys- tem parameters, such as γ and E0. 7 (a) 1 0.8 0.6 0.4 0.2 n o i s s i m s n a r T 0 -π/2 -π/4 π/4 π/2 0 α 1 0.8 0.6 0.4 0.2 n o i s s i m s n a r T (b) 0 -0.5 0 EA(eV ) 0.5 (Color online) Transmission probabilities t(L)↑ ↑2 FIG. 4: (blue, dash lines) and t(L)↓ ↓2 (red, solid lines): (a) as a func- tion of the incidence angle α at E0 = EA = λx = 0.32eV; (b)as a function of the energy EA at ky = 0.01A−1. The calculations are done for 200 sequentially connected ripples (π-arcs) The other parameters are the same as in Fig.2. At a fixed value of the momentum ky the effectivity of spin filtering is reduced by ∼ 10% [see Fig.4(b)]. At the same time, our systems manifests a zero transmis- sion for all spin orientation for charge carriers at energies −0.06eV < EA < 0.06eV due to our choice of the value ky. D. The graphene purity We restricted our consideration to a ballistic regime. This approximation is well justified due to the following factors. The remarkable strength of the carbon honey- comb lattice makes it quite difficult to introduce any de- fects into the lattice itself. Charge impurities that could limit electron mobility in graphene are still an open prob- lem from both experimental and theoretical points of view (see, for example, discussion in Ref.1). It is also well known that the difference in conductivity in graphene be- tween T ≈ 0 and room temperature is no more than a few percent. In other words, the electron-phonon scattering plays a minor role. We recall that a typical ripple size is ∼ 7 nm (see Ref.24). In our paper the ripple is modeled as the curved surface in the form of an arc of a circle with a radius R = 1nm. As a result, our system length is πR × 200 ≈ 640 nm. Taking into account that a typical mean free path of electrons in single-wall nanotubes is ℓ ≈ 1 µm (see, e.g., Ref.2), it seems our consideration is on a reasonable basis. Thus, in our model the basic mechanism that is respon- sible for spin filtering effects is an attenuation of one of the transmitting modes. It transforms to the evanescent mode in the energy gap created by the SOC in the curved surface. The multiplicative action of a large enough num- ber of ripples suppresses this transmitting mode at cer- tain conditions that provide a high efficiency for the other one. IV. SUMMARY We have analysed the transmission and reflection of ballistic electron flow through a ripple in an effective mass approximation, when only the interaction between near- est neighbor atoms is taken into account. In our consid- eration a ripple consists of the curved surface in the form of an arc of a circle connected from the left-hand and the right-hand sides to two semi-infinite flat graphene sheets. Considering the curved surface as a part of the armchair nanotube, we have shown that the curvature induced spin-orbit coupling yields a backscattering [see Eqs.(30,31)] with spin inversion. This spin inversion is caused by the spin-orbit term that breaks spin symmetry (a spin projection on the symmetry axis) in the effective Hamiltonian of the armchair CNT. In the energy gap created by the curvature induced spin-orbit coupling there is a preference for one spin ori- entation, depending on the direction of the electron flow at normal incidence. The width of the energy gap de- pends in inverse proportion on the radius of the ripple. At this energy range the ripple acts as a semipermeable membrane which is more transparent for the incoming electrons with spin up from the left-hand side and with spin down from the right-hand side, and vise versa for the holes. In other words, there is a precursor of chiral transmission of spin components of the incoming electron (hole) flow at a fixed energy. For one ripple system this effect is, however, small. In order to enforce this effect, we extended our consideration to a curved surface of the sinusoidal wave type with N arcs. This step is of crucial importance to suppress one of the spin components and to support the spin inversion symmetry for the transmis- sion probability. The larger the number of consistently connected ripples (arcs) is, the stronger the dominance of a specific spin component is in comparison with the other in the transmission from the same direction. There is a cooperative effect of chiral spin transmissions pro- 8 duced by a large number of ripples. To trace the N - dependence we have derived a formula for a composite transmission probability for well-polarized spin compo- nents: i) Eq.(41) for spin up electrons; ii) Eq.(42) for spin down electrons. Based on these results, we predict a strong spin filtering effect for a sufficiently large number of arcs in the rippled graphene system. In contrast to the usual waveguide that guides optical or sound waves of a chosen frequency in a well -- defined direction, our system guides spin electron (hole) waves with a well -- defined po- larization in one or another direction at a certain energy. It seems, therefore, natural to name this system chiral spinguide. We have considered only a curved surface that owes its origin to an armchair nanotube. Evidently, our model can be extended to other types of origins. However, the corresponding analysis requires a separate study. We also neglected the effective magnetic field that arises from the dependence of the hopping parameter γ on the curva- ture (see discussion in3). This effect influences the local density of states21. It can cause the localization of the electrons on the boundary between flat graphene and the curved surface, similar to the boundary state for some types of carbon nanoparticles22. As a result, it might affect the efficiency of the spinguide. Last, but not least, many body effects such as electron-electron interaction should be incorporated and analyzed as well. It is espe- cially noteworthy that electron-electron interaction, de- signed in the form of a specific potential barrier on the graphene sheet23, leads to separation of spin-polarized states. In fact, this result is in close agreement with our finding, obtained for one ripple. As mentioned above, the curvature induced SOC simulates a penetrable bar- rier preferable for transmission of only one of two spin components, depending on the direction and energy of the incoming electron (hole) flow. It would be interest- ing to study the interplay between the SOC and electron- electron interaction on the electron transport in our sys- tem. Evidently, this consideration would allow us to study in more detail the effect of impurities on the elec- tron mobility in our system. In conclusion, the transparency and the mathemati- cal rigor of our results provide good grounds to believe that spin filtering effects found in this paper, giving rise to a chiral spinguide phenomenon, will be observable in experiment. Acknowledgments M.P. and K.N.P. are grateful for the warm hospitality and creative atmosphere at UIB and JINR. This work was supported in part by RFBR Grant 14-02-00723 and Slovak Grant Agency VEGA Grant No. 2/0037/13. 1 M.I. Katsnelson, Graphene: Carbon in Two Dimensions 2 L.E.F. Foa Torres, S. Roche, and J.-C. Charlier, Introduc- (Cambridge University Press, Cambridge, 2012). 9 tion to Graphene-Based Nanomaterials (Cambridge Uni- versity Press, New York, 2014). 14 W. Izumida, K. Sato, and R. Saito, J. Phys. Soc. Jpn. 78, 074707 (2009). 3 M. Katsnelson and A. Geim, Philos.Trans. R. Soc. A366, 15 M. del Valle, M. Marga´nska, and M. Grifoni, Phys. Rev. B 195 (2008) 84, 165427 (2011). 4 W. Bao, F. Miao, Z. Chen, H. Zhang, W. Jang, C. Dames, 16 J. Klinovaja, M. J. Schmidt, B. Braunecker, and D. Loss, and C. Lau, Nature Nanotechnol. 4, 562 (2009). Phys. Rev. B 84, 085452 (2011). 5 H. Min, J. E. Hill, N. A. Sinitsyn, B. R. Sahu, L. Kleinman, 17 K.N. Pichugin, M. Pudlak, and R.G. Nazmitdinov, Eur. and A. H. MacDonald, Phys. Rev. B 74, 165310 (2006). 6 T. Ando, J.Phys.Soc.Jpn. 69, 1757 (2000). 7 M.V. Entin and L.I. Magarill, Phys. Rev. B 64, 085330 (2001). Phys J. B 87, 124 (2014). 18 M. Pudlak and R.Pincak, Eur. Phys J. B 67, 565 (2009). 19 S. Datta, Electronic Transport in Mesoscopic Systems (Cambridge University Press, Cambridge, 2007). 8 A. De Martino, R. Egger, K. Hallberg, and C.A. Balseiro, 20 S.V. Morozov, K.S. Novoselov, and A.K. Geim, Phys. Usp. Phys. Rev. Lett. 88, 206402 (2002). 51, 744 (2008). 9 D.Huertas-Hernando, F. Guinea, and A. Brataas, Phys. 21 F. de Juan, A.Cortijo, and M.A.H.Vozmediano, Phys. Rev. Rev. B 74, 155426 (2006) . B 76, 165409 (2007). 10 F. Kuemmeth, S.Ilani, D.C. Ralph, and P.L. McEuen, Na- 22 T. Fujita, M.B.A. Jalil, S.G.Tan, and S. Murakami, ture (London) 452, 448 (2008). J.Appl. Phys. 110, 121301 (2011). 11 D.V. Bulaev, B. Trauzettel, and D. Loss, Phys. Rev. B 77, 23 L. Ying, G. Wang, L. Huang, and Y.C. Lai, Phys. Rev. B 235301 (2008). 90, 224301 (2014). 12 L. Chico, M.P. L´opez-Sancho, and M. C. Munoz, Phys. 24 A. Fasolino, J.H. Los, and M.I. Katsnelson, Nat. Mater. 6, Rev. B 79, 235423 (2009). 13 J.-S. Jeong and H.-W. Lee, Phys. Rev. B 80 075409, 858 (2007). (2009).
1804.01692	1	1804	2018-04-05T07:05:00	Surface Step Defect in 3D Topological Insulators: Electric Manipulation of Spin and Quantum Spin Hall Effect	[ "cond-mat.mes-hall" ]	We study the influence of step defect on surface states in three-dimensional topological insulators subject to a perpendicular magnetic field. By calculating the energy spectrum of the surface states, we find that Landau levels (LLs) can form on flat regions of the surface and are distant from the step defect, and several subbands emerge at side surface of the step defect. The subband which connects to the two zeroth LLs is spin-polarized and chiral. In particular, when the electron transports along the side surface, the electron spin direction can be manipulated arbitrarily by gate voltage. And no reflection occurs even if the electron spin direction is changed. This provides a fascinating avenue to control the electron spin easily and coherently. In addition, regarding the subbands with high LL index, there exist spin-momentum locking helical states and the quantum spin Hall effect can appear.	cond-mat.mes-hall	cond-mat	Surface Step Defect in 3D Topological Insulators: Electric Manipulation of Spin and Quantum Spin Hall Effect Yan-Feng Zhou,1, 2 Ai-Min Guo,3 and Qing-Feng Sun1, 2, ∗ 1International Center for Quantum Materials, School of Physics, Peking University, Beijing 100871, China 2Collaborative Innovation Center of Quantum Matter, Beijing 100871, China 3Department of Physics, Harbin Institute of Technology, Harbin 150001, China (Dated: October 15, 2018) We study the influence of step defect on surface states in three-dimensional topological insulators subject to a perpendicular magnetic field. By calculating the energy spectrum of the surface states, we find that Landau levels (LLs) can form on flat regions of the surface and are distant from the step defect, and several subbands emerge at side surface of the step defect. The subband which connects to the two zeroth LLs is spin-polarized and chiral. In particular, when the electron transports along the side surface, the electron spin direction can be manipulated arbitrarily by gate voltage. And no reflection occurs even if the electron spin direction is changed. This provides a fascinating avenue to control the electron spin easily and coherently. In addition, regarding the subbands with high LL index, there exist spin-momentum locking helical states and the quantum spin Hall effect can appear. PACS numbers: 73.20.-r, 71.10.Pm, 72.25.-b, 73.43.-f I. INTRODUCTION the 3D TIs. In recent years, many effects have been achieved on topological insulators (TIs)1,2, which are an intriguing insulating phase of electronic materials and can be clas- sified by different topological invariants3,4. The elec- tronic states at surface, namely surface states, of three- dimensional (3D) TIs are conducting and are protected by time-reversal symmetry5 -- 10, which arises from the bulk topology through bulk-boundary correspondence7. For the 3D TIs, the excitations of the surface states are characterized by 2D Weyl fermions and present an odd number of gapless Dirac cones in the Brillouin zone. In contrast to conventional two-dimensional (2D) conduc- tors, the conducting surface states are robust against nonmagnetic disorders due to the prohibited backscatter- ing and could induce the half-quantum Hall effect under magnetic field. Several experimental11 -- 13 and theoretical studies14 -- 16 have demonstrated the Friedel oscillations of local density of states, which is attributed to scattering from surface step in the 3D TIs and provides evidence for suppression of backscattering. The Friedel oscillations present power- law decay behavior with exponent ranging from -3/2 to -1/2, by either changing the Fermi energy or extending step defect along different directions, due to strong warp- ing effect in the 3D TIs13,16. In these theoretical studies, the step defect is regarded as potential barrier of either one-dimensional δ function15,16 or step function13, where the incident electron is perpendicular to the step defect. If the electron transports parallel to the step defect, this method cannot capture all the physics, since all the sur- faces construct a closed one17,18. Recently, the match- ing conditions are derived for the surface states of two perpendicularly connected surfaces of the 3D TIs19. By using these matching conditions, one can study the elec- tronic structure of the surface states of the step defect in Since the surface states possess a novel helical spin tex- ture, they can be used in the field of spintronics1,20 -- 22. The experiments have realized the injection and detec- tion of spin-polarized current in the 3D TIs by electric methods23 -- 28. Another important theme of spintronics is to coherently manipulate the spin degree of freedom. The spin polarization of the surface states can be tuned by polarized illumination29 -- 31. In principle, the spin polar- ization direction can be modulated by external magnetic field and by optical methods, but it is difficult to produce integrated circuits with low power consumption32,33. So, it is important to control the electron spin by using the electric methods, just as the Datta-Das transistor34 -- 36, which is a spin field effect transistor and the spin di- rection is tuned by gate voltage with the aid of Rashba spin-orbit coupling. In this paper, we study the influence of the step defect on the surface states in the 3D TIs subject to a perpen- dicular magnetic field (z-axis), as illustrated in Fig.1(a). Under the magnetic field, Landau levels (LLs) can form in both upper and lower surfaces, and are distant from the step defect (side surface). Although the side surface is parallel to the magnetic field, its surface states can de- velop into subbands due to the size quantization effect. And the expectation of the electron spin is calculated for these subbands. Our results reveal that the subband which connects to the two zeroth LLs is spin polarized and chiral. When the electron transmits within the side surface, the spin polarized direction can be tuned by gate voltage, and no reflection occurs by changing the electron spin direction. These unique features provide a new av- enue to control the electron spin easily and coherently. Besides, the subbands with nonzero indices are parabolic and present helical property, which generates the quan- tum spin Hall effect and is similar to the helical edge states in the 2D TIs. (a) (b) 200 150 100 50 0 -50 -100 -150 -200 ) V e m ( y g r e n E (a) 2 (b) -0.6 -0.3 0.0 0.3 0.6 -0.6 -0.3 0.0 0.3 0.6 ky (1/nm) ky (1/nm) -d d FIG. 2: (Color online). Dispersion relation of the 2D model with (a) V0 = 0 meV and (b) V0 = 25 meV. The other parameters are 2d = 30 nm and B = 15 T. (Color online) (a) Schematic of upper and lower FIG. 1: surfaces connected by side surface (step defect) in the 3D TI under perpendicular magnetic field. (b) Mapping of the above surfaces in the regions of x > d, −d < x < d, and x < −d of 2D model. The potential energies of the upper and lower surfaces can be controlled by the gate voltage and no magnetic field threads into the side surface. The rest of the paper is organized as follows. In Sec. II, the theoretical model and the methods for studying the influence of the step defect are presented. In Sec. III, we propose an electric strategy to manipulate the electron spin. In Sec. IV, we show a quantum spin Hall effect in a convex platform on the TI's surface at zero gate voltage. Finally, a brief summary is given in Sec. V. II. MODEL AND METHODS The 3D TI is composed of the upper surface, the lower one, and the side one (step defect), as shown in Fig.1(a). Then, the effective Hamiltonian of the surface states can be expressed as37,38 H(k) = ¯hνF (σ × k) · n + U, (1) where νF is the Fermi velocity, σ ≡ (σx, σy, σz) with σx,y,z the Pauli matrices, k = (kx, ky, kz) is the momen- tum, n is the normal vector of a specific surface, and U is the potential energy which can be modulated in the experiments by the gate voltage. Here, we assume that νF is the same for all the three surfaces mentioned above. For the side surface, we make a local rotation by fixing the y-axis so that the original x-axis is changed into the z-axis, and the effective Hamiltonian of the side surface can be obtained from a unitary transformation. By com- bining the matching conditions between the side surface and the +z surface, the effective 2D model can be ob- tained and then the three surfaces can be described by the Hamiltonian H = ¯hνF (−kxσy + kyσx) + U (x).39,40 In this effective model, the upper, lower, and side surfaces locate in the regions of x > d, x < −d, and −d < x < d, respectively, and are infinite along the y-axis [Fig.1(b)]. Under the Landau gauge, the vector potential can be written as A = (0, Ay), with Ay being B(x − d), 0, and B(x + d) for the upper, side, and lower surfaces, respectively, where the parameter B is the strength of the magnetic field. We consider that the gate electrodes only locate on the top of the upper and lower surfaces, and the potential energies are U (x) = V0, 0, and −V0, respectively, in the upper, side, and lower surfaces. As the model is invariant by translating along the y-axis, the momentum ky is a good quantum number. To calculate the band structure of the 2D model, the effective Hamiltonian can be discretized along the x-axis by performing dΨ(x) dx → Ψi+1−Ψi−1 :41 2a H = X i T0ci + (c† [c† i Txci+1 + H.c.)], i T0 = U (xi)I + (W/a)σz + νF (¯hky + eAy)σx, Tx = −(W/2a)σz + (i¯hνF /2a)σy, (2) where ci and c† i are, respectively, the annihilation and creation operators at site i. I is the 2 × 2 unit matrix and W/a is the Wilson mass term which is introduced to avoid the fermion double problem, with a the lattice constant and W being 0.1¯hνF .42 Here, the Zeeman effect is not considered, because some experiments have shown that the g-factor is quite small and the Zeeman effect may not be important.43,44 III. MANIPULATION OF ELECTRON SPIN Firstly, we calculate the band structure of the 2D ef- fective model, where the Fermi velocity is set to νF = 5 × 105 m/s,11,45 the lattice constant is a = 0.375 nm, and the truncation length along the x-axis is taken as 0.01 0.00 0.01 0.00 -50 0 50 -50 0 50 (b) Sx Sy Sz (d) 0.01 0.00 0.01 0.00 -50 0 50 -50 0 50 ) 2 / 0.01 (a) ( z S , y S , x S ) 2 / ( z S , y S , x S 0.00 -0.01 0.01 (c) 0.00 -0.01 -60 -40 -20 0 20 40 60 80 -40 -20 0 20 40 60 80 Position x (nm) Position x (nm) (Color online) Spatial distribution sx,y,z of the elec- FIG. 3: tron spin for the chiral state, which connects to the two zeroth LLs of the upper and lower surfaces, with the potential energy (a) V0 = 0 meV, (b) V0 = 10 meV, (c) V0 = 25 meV, and (d) V0 = 50 meV. The insets display probability density ρ(x). Here, we choose ky = 0.01 nm−1, 2d = 30 nm, and B = 15 T. 180 nm, which is sufficiently long that the wavefunc- tions of the LLs are localized. For V0 = 0 meV, the 2D model has C2 symmetry with respect to the z- axis. As a result, the band structure is symmetric, i.e., E(ky) = E(−ky), as indicated in Fig.2(a). It is clear that except for the zero energy band which is flat, each band consists of two flat parts and a parabola-like part. In the flat parts, the states are the LLs with the en- ergy εN = sgn(N )¯hωcpN , and correspond to the up- per surface for ky < 0 and to the lower surface for ky > 0, where the integer number N is the LL index and ωc = νF p2eB/¯h is the cyclotron frequency. While in the parabola-like parts, the states are confined to the side surface and form subbands owing to the size quan- tization effect. When the gate voltage is implemented, the band structure becomes asymmetric and the energies of the LLs for the upper (lower) surface are shifted by V0 (−V0) [see Fig.2(b)]. And the band connecting to the two zeroth LLs becomes chiral, similar to the chiral edge state in the topological p-n junction46. N,ky We then study the property of the chiral state which connects to the two zeroth LLs of the upper and lower surfaces. Fig.3 plots spatial distribution sx,y,z of the electron spin for the chiral state with different values of V0, and the insets show the corresponding proba- bility density ρ(x) versus position x. Here, sx,y,z = Ψ† (x)sx,y,zΨN,ky (x), ρ(x) = ΨN,ky (x)2, and the ex- pectation of the electron spin can be obtained by the inte- gral ¯sx,y,z = R sx,y,zdx, where ΨN,ky (x) is the eigenfunc- tion of Eq.(2) and N is the band index. As mentioned above, the 3D model [Fig.1(a)] can be mapped into the 2D effective model [Fig.1(b)] by performing local rota- tion on the side surface, where the x-z plane is rotated by 90◦ with the y-axis fixed. Because of this local rota- tion, the spin operator is expressed as s = ¯h 2 (σx, σy, σz) (a) (b) 0 0 10 (d) 100 50 Sz Sx 20 30 V0 (mV) 20 nm 25 nm 30 nm 35 nm 40 nm 3 10T 11T 14T 17T 20T (c) 150 100 50 40 50 (e) 100 50 0 10 30 20 V0 (mV) 40 50 0 nm-1 0.03 nm-1 0.05 nm-1 0.07 nm-1 0.1 nm-1 0 10 20 30 V0 (mV) 40 50 0 10 40 50 20 30 V0 (mV) FIG. 4: (Color online) (a) A setup for manipulation of the electron spin under nonuniform gate voltage. (b) Spin texture in the sx-sz plane as a function of the gate voltage V0. Tilt- ing angle θ versus V0 for (c) different magnetic fields B, for (d) various heights 2d of the side surface, and for (e) several momentums ky. The unmentioned parameters are B = 15 T, 2d = 30 nm, and ky = 0.01 nm−1. when x > d and x < −d, and s = ¯h −d < x < d. 2 (−σz, σy, σx) when When ky is close to 0, the chiral state mainly locates in the side surface (see the insets of Fig.3). It can be seen from Fig.3(a) that in the region of the side surface, the chiral state is spin-polarized along the +x direction, although the magnetic field is parallel to the side surface and has no direct effect on it. While beyond the side surface, the electron spin is polarized along the −z direc- tion. It is worth noting that sy is always zero for all the bands because the Hamiltonian is real. When the gate voltage is implemented, a negative sz occurs in the side surface [see Figs.3(b)-3(d)]. In the region of the side sur- face, sx decreases with V0 and even becomes negative for large V0, and the absolute value of sz is enhanced. This implies that the electron spin direction can be changed from sx-up to sx-down. The underlying physics for the change of the spin direction is that the spin-orbit cou- pling can modify the two components of the eigenstates in the surface when the gate voltage is changed. Besides, the chiral state still resides in the side surface. Therefore, this provides a strategy to manipulate the electron spin direction coherently by tuning the gate voltage. To further demonstrate the details of this strategy, we study the spin polarization of the chiral state under the gate voltage for different system parameters. Fig.4(b) shows spin texture of the chiral state versus the gate voltage V0. The arrow in the spin texture denotes the spin polarization vector of an eigenstate at a specific gate voltage. The spin polarization vector is defined as P = (¯sx, ¯sy, ¯sz), where the ¯sx,y,z are the expectations of the x + ¯s2 spin operator as mentioned above. It clearly appears that the electron spin direction can be successively tuned from sx-up to sx-down by increasing V0. To quantitatively analyze the electron spin, a tilting angle is defined as θ = arccos[¯sx/(¯s2 z)1/2], which denotes the relative orientation of the electron spin direction with respect to the x-axis. Figs.4(c) and 4(d) present the tilting angle θ for different magnetic fields B and for different heights 2d of the side surface, respectively, as a function of V0. It can be seen that θ increases almost linearly with V0 for different magnetic fields B and heights 2d. For a specific height 2d, the weaker the magnetic field is, the faster θ changes with V0 [Fig.4(c)]. While for a certain magnetic field, the slope of the θ−V0 curve is only slightly modified by changing 2d [Fig.4(d)]. The tilting angle θ is slightly increased by decreasing 2d, which is irrespective of V0. The range of the tilting angle θ becomes small by decreasing 2d. When the height 2d is declined to zero, the system becomes a topological p-n junction and the range of θ will be less than 65◦ by changing V0 from 0 to 50 meV. This implies that the existence of the side surface is important to manipulate the spin direction, because the spin direction in the side surface is very different from that in the upper and lower surfaces. Do all the chiral states in Fig.2(b) possess these characters? By inspecting Fig.4(e), one can see that for a wide range of ky in which the band is chiral, the tilting angle θ rises monotonously as V0 increases. There exists an obvious deviation from the linear behavior when the states approach the LLs. Thus, it is feasible and effective to manipulate the electron spin of the chiral states by the gate voltage. Based on the above results, we can envision a setup [see Fig.4(a)], in which several independent gates are placed on the top of both upper and lower surfaces, and the electron transports along the side surface. By tuning the gate voltage, the potential energies can be set arbitrarily in space and the electron spin direction can be modulated in any way. Furthermore, there is not any reflection dur- ing the transport process because of the unidirectionality of the chiral state in the side surface. In a word, this is a high controllable, no reflection, low energy consumption, and electric strategy to manipulate the electron spin di- rection. IV. QUANTUM SPIN HALL EFFECT Finally, we consider the 3D TI with a convex platform on its top surface, as depicted in the inset of Fig.5, and study the bands which connect to the nonzero LLs of the upper and lower surfaces. In this 3D TI, the front and back side surfaces appear naturally at the bound- ary of the convex platform and are of identical height. The distance between the two side surfaces is sufficiently long so that the localized states in the two side surfaces cannot mix together under the perpendicular magnetic field. The magnetic field is set to B = 15 T, the height of the side surfaces is 2d = 30 nm, and no gate voltage is (b) 4 Sx Sy Sz (d) ) (a) 2 0.01 / ( z S , y S , x S ) 2 / ( z S , y S , x S 0.00 -0.01 0.01 (c) 0.00 -0.01 -60 -40 -20 0 20 40 -40 -20 0 20 40 60 Position x (nm) Position x (nm) FIG. 5: (Color online) (a) and (c) Spatial distribution sx,y,z for the electronic state in the front side surface, which con- nects the two LLs of N = 1, with (a) ky = 0.1 nm−1 and (c) ky = −0.1 nm−1. (b) and (d) show the corresponding sx,y,z for the back side surface. The parameters are B = 15 T, 2d = 30 nm, and V0 = 0 meV. In the calculation of (b) and (d), we make local rotation of the x and z coordinates around the y-axis by −90◦, and the region of −d < x < d is still the side surface. The inset shows schematic view of a convex platform in the surface of the 3D TI, and the front and back side surfaces form naturally. employed. The electronic states connecting to the nonzero LLs are non-chiral near the front side surface, and there are a couple of states with the same energy but propagating oppositely in the side surface [see Fig.2(a)]. Figs.5(a) and 5(c) display the spatial distribution sx,y,z for the states connecting to the N = 1 LLs with ky = ±0.1 nm−1. It is evident that the state with ky = 0.1 nm−1 is propagat- ing forward and is almost polarized in the +z direction [Fig.5(a)], while the state with ky = −0.1 nm−1 is prop- agating backward and is polarized in the −z direction [Fig.5(c)]. The spatial distribution sy is exactly zero ev- erywhere and ¯sx is very small. Therefore, these states in the front side surface are spin-momentum locking helical states, which are similar to the helical edge states in the 2D TI4,47. While for the back side surface, the forward state is polarized in the −z direction [Fig.5(b)] and the back- ward state is polarized in the +z direction [Fig.5(d)]. Therefore, the spin-momentum locking persists for these electronic states. And the quantum spin Hall effect can be observed in the convex platform of the 3D TI un- der the magnetic field, and the longitudinal conductance will be quantized when the Fermi energy lies within the gap between the two LLs of the upper and lower sur- faces. Besides, the bands of higher LLs are also spin- momentum locked (data not shown). Thus, by tuning the Fermi energy, the quantum spin Hall effect can be achieved with N spin-up states propagating forward and N spin-down states propagating backward. And some 5 quantum plateaus can be detected in the longitudinal re- sistance and the spin Hall resistance47,48. V. CONCLUSIONS In summary, we have investigated the effect of step defect on surface states in three-dimensional topological insulators under a perpendicular magnetic field. By cal- culating the energy spectrum and the expectation of the electron spin of the surface states in the zeroth band, a chiral subband exists in the side surface and the electron spin direction can be manipulated in any way by gate voltages on the flat regions. Besides, no reflection occurs when the electron spin direction of the chiral state is changed. This provides a low power and electric scheme to control the electron spin coherently. In addition, for each subband with high indices, there is a pair of spin- momentum locking helical states in the side surface, and these helical states can induce quantum spin Hall effect. Acknowledgments We gratefully acknowledge the financial support from NBRP of China (2012CB921303 and 2015CB921102), NSF-China under Grants No. 11274364, 11574007, and FRFCU under Grant No. and AUGA5710013615. Y.-F. Zhou gratefully acknowl- edges T. Morimoto for many helpful discussions. 11504066, ∗ [email protected] 1 M.Z. Hasan and C.L. Kane, Rev. Mod. Phys. 82, 3045 Lett. 102, 166801 (2009). 22 I. Garate and M. Franz, Phys. Rev. Lett. 104, 146802 (2010). (2010). 2 X.-L. Qi and S.-C. Zhang, Rev. Mod. Phys. 83, 1057 23 C.H. Li, O.M.J. van 't Erve, J.T. Robinson, Y. Liu, L. Li, (2011). 3 D.J. Thouless, M. Kohmoto, M.P. Nightingale, and M. den Nijs, Phys. Rev. Lett 49, 405(1982). 4 C.L. Kane and E.J. Mele, Phys. Rev. Lett. 95, 146802 (2005). 5 L. Fu, C.L. Kane, and E.J. Mele, Phys. Rev. Lett. 98, 106803 (2007). 6 J.E. Moore and L. Balents, Phys. Rev. B 75, 121306(R) (2007). 7 L. Fu and C.L. Kane, Phys. Rev. B 74, 195312(2006). 8 D. Hsieh, Y. Xia, L. Wray, D. Qian, A. Pal, J.H. Dil, J. Osterwalder, F. Meier, G. Bihlmayer, C.L. Kane, Y.S. Hor, R.J. Cava and M.Z. Hasan, Science 323, 919(2009). 9 Y.L. Chen, J.G. Analytis, J.-H. Chu, Z.K. Liu, S.-K. Mo, X.L. Qi, H.J. Zhang, D.H. Lu, X. Dai, Z. Fang, S.C. Zhang, I.R. Fisher, Z. Hussain, Z.-X. Shen, Science 325, 178 (2009). 10 Y. Xia, D. Qian, D. Hsieh, L. Wray, A. Pal, H. Lin, A. Bansil, D. Grauer, Y.S. Hor, R.J. Cava and M.Z. Hasan, Nat. Phys. 5, 398 (2009). 11 T. Zhang, P. Cheng, X. Chen, J.-F. Jia, X. Ma, K. He, L. Wang, H. Zhang, X. Dai, Z. Fang, X. Xie and Q.-K. Xue, Phys. Rev. Lett. 103, 266803 (2009). 12 Z. Alpichshev, J.G. Analytis, J.-H. Chu, I.R. Fisher, Y.L. Chen, Z.X. Shen, A. Fang, and A. Kapitulnik, Phys. Rev. Lett. 104, 016401 (2010). 13 J. Wang, W. Li, P. Cheng, C. Song, T. Zhang, P. Deng, X. Chen, X. Ma, K. He, J.-F. Jia, Q.-K. Xue, and B.-F. Zhu, Phys. Rev. B 84, 235447 (2011). 14 R.R. Biswas and A.V. Balatsky, Phys. Rev. B 83, 075439 (2011). 15 D. Zhang and C.S. Ting, Phys. Rev. B 85, 115434 (2012) 16 J. An and C.S. Ting, Phys. Rev. B 86, 165313 (2012). 17 D.-H. Lee, Phys. Rev. Lett. 103, 196804 (2009). 18 O. Vafek, Phys. Rev. B 84, 245417 (2011). 19 L. Brey and H.A. Fertig, Phys. Rev. B 89, 085305 (2014). 20 Q. Liu, C.-X. Liu, C. Xu, X.-L. Qi, and S.-C. Zhang, Phys. Rev. Lett. 102, 156603 (2009). and B.T. Jonker Nat. Nanotechnol. 9, 218 (2014). 24 J. Tang, L.-T. Chang, X. Kou, K. Murata, E.S. Choi, M. Lang, Y. Fan, Y. Jiang, M. Montazeri, W. Jiang, Y. Wang, L. He, and K.L. Wang, Nano Lett. 14, 5423 (2014). 25 Y. Ando, T. Hamasaki, T. Kurokawa, K. Ichiba, F. Yang, M. Novak, S. Sasaki, K. Segawa, Y. Ando, and M. Shi- raishi, Nano Lett. 14, 6226 (2014). 26 L. Liu, A. Richardella, I. Garate, Y. Zhu, N. Samarth, and C.-T. Chen, Phys. Rev. B 91, 235437 (2015). 27 J. Tian, I. Miotkowski, S. Hong, and Y.P. Chen, Sci. Rep. 5, 14293 (2015). 28 A. Dankert, J. Geurs, M.V. Kamalakar, S. Charpentier, and S.P. Dash, Nano Lett. 15, 7976 (2015). 29 T. Misawa, T. Yokoyama, and S. Murakami, Phys. Rev. B 84, 165407 (2011) 30 C. Jozwiak, C.-H. Park, K. Gotlieb, C. Hwang, D.-H. Lee, S.G. Louie, J.D. Denlinger, C.R. Rotundu, R.J. Birgeneau, Z. Hussain, and A. Lanzara, Nat. Phys. 9, 293 (2013). 31 D. Yudin, O.V. Kibis, and I.A. Shelykh, arXiv: 1512.00798 (2015). 32 I. Zuti´c, J. Fabian, and S.D. Sarma, Rev. Mod. Phys. 76, 323 (2004). 33 D.D. Awschalom and M.E. Flatt´e, Nat. Phys. 3, 153 (2007). 34 S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990). 35 H.C. Koo, J.H. Kwon, J. Eom, J. Chang, S.H. Han, and M. Johnson, Science 325, 1515 (2009). 36 P. Chuang, S.-C. Ho, L.W. Smith, F. Sfigakis, M. Pep- per, C.-H. Chen, J.-C. Fan, J.P. Griffiths, I. Farrer, H.E. Beere, G.A.C. Jones, D.A. Ritchie, and T.-M. Chen, Nat. Nanotechnol. 10, 35 (2015). 37 H.-Z. Lu, W.-Y. Shan, W. Yao, Q. Niu, and S.-Q. Shen, Phys. Rev. B 81, 115407 (2010). 38 H. Liu, H. Jiang, Q.-F. Sun, and X.C. Xie, Phys. Rev. Lett. 113, 046805 (2014). 39 T. Morimoto, A. Furusaki, and N. Nagaosa, Phys. Rev. Lett. 114, 146803 (2015). 40 T. Morimoto, A. Furusaki, and N. Nagaosa, Phys. Rev. B 21 T. Yokoyama, Y. Tanaka, and N. Nagaosa, Phys. Rev. 92, 085113 (2015). 41 S. Datta, electronic transport in mesoscopic system, (Cam- bridge, 1995). 42 J.B. Kogut, Rev. Mod. Phys. 55, 775 (1983). 43 J. Xiong, Y. Luo, Y.H. Khoo, S. Jia, R.J. Cava, and N.P. Ong, Phys. Rev. B 86, 045314 (2012). 44 Z. Zhang, W. Wei, F. Yang, Z. Zhu, M. Guo, Y. Feng, D. Yu, M. Yao, N. Harrison, R. McDonald, Y. Zhang, D. Guan, D. Qian, J. Jia, and Y. Wang, Phys. Rev. B 92, 235402 (2015). 45 R. Yoshimi, A. Tsukazaki, Y. Kozuka, J. Falson, K.S. 6 Takahashi, J.G. Checkelsky, N. Nagaosa, M. Kawasaki, and Y. Tokura, Nat. Commun. 6, 6627 (2015) 46 J. Wang, X. Chen, B.-F. Zhu, and S.-C. Zhang, Phys. Rev. B 85, 235131 (2012). 47 H. Jiang, S. Cheng, Q.-F. Sun, and X.C. Xie, Phys. Rev. Lett. 103, 036803 (2009). 48 Q.-F. Sun and X.C. Xie, Phys. Rev. Lett. 104, 066805 (2010).
1512.00642	1	1512	2015-12-02T10:42:08	Enhanced spin-orbit coupling in dilute fluorinated graphene	[ "cond-mat.mes-hall" ]	The preservation and manipulation of a spin state mainly depends on the strength of the spin-orbit interaction. For pristine graphene, the intrinsic spin-orbit coupling (SOC) is only in the order of few ueV, which makes it almost impossible to be used as an active element in future electric field controlled spintronics devices. This stimulates the development of a systematic method for extrinsically enhancing the SOC of graphene. In this letter, we study the strength of SOC in weakly fluorinated graphene devices. We observe high non-local signals even without applying any external magnetic field. The magnitude of the signal increases with increasing fluorine adatom coverage. From the length dependence of the non-local transport measurements, we obtain SOC values of ~ 5.1 meV and ~ 9.1 meV for the devices with ~ 0.005% and ~ 0.06% fluorination, respectively. Such a large enhancement, together with the high charge mobility of fluorinated samples (u~4300 cm2/Vs - 2700 cm2/Vs), enables the detection of the spin Hall effect even at room temperature.	cond-mat.mes-hall	cond-mat	Enhanced spin-orbit coupling in dilute fluorinated graphene Ahmet Avsar1,2,, Jong Hak Lee1,2,, Gavin Kok Wai Koon1,2, and Barbaros Özyilmaz1,2,3,† 1Centre for Advanced 2D Materials, National University of Singapore, 117542, Singapore, 2Department of Physics, National University of Singapore, 117542, Singapore, 3NanoCore, National University of Singapore, 117576, Singapore. The preservation and manipulation of a spin state mainly depends on the strength of the spin-orbit interaction. For pristine graphene, the intrinsic spin-orbit coupling (SOC) is only in the order of few μeV, which makes it almost impossible to be used as an active element in future electric field controlled spintronics devices. This stimulates the development of a systematic method for extrinsically enhancing the SOC of graphene. In this letter, we study the strength of SOC in weakly fluorinated graphene devices. We observe high non-local signals even without applying any external magnetic field. The magnitude of the signal increases with increasing fluorine adatom coverage. From the length dependence of the non-local transport measurements, we obtain SOC values of ~ 5.1 meV and ~ 9.1 meV for the devices with ~ 0.005% and ~ 0.06% fluorination, respectively. Such a large enhancement, together with the high charge mobility of fluorinated samples (μ~4300 cm2/Vs - 2700 cm2/Vs), enables the detection of the spin Hall effect even at room temperature. †Address correspondence to: [email protected] * These authors contributed equally to this work. For pristine graphene, the hopping of π electrons between the two next nearest neighbor carbon atoms is the only source of the SOC. This is a second order process and gives rise to SOC in the order of only a few μeV [1–3]. Such a weak coupling limits the prospect of potential graphene-based spin field effect transistors [4]. Different approaches have been suggested to enhance the SOC of graphene. For example, the creation of a curvature in flat graphene is expected to significantly enhance the intrinsic SOC [1,3,5] as in the carbon nanotube case [6]. However, there is no well-developed method for the fabrication of such devices. Recently, the proximity effect at the interface between graphene and WS2 layers has been shown to result in a SOC enhancement [7]. Also the hydrogenation of graphene has been shown to significantly enhance the SOC of graphene [8,9]. Unfortunately, chemisorbed hydrogen atoms can be easily detached at moderate temperatures and this makes devices less stable at ambient conditions [10,11]. The stability of devices can be improved with choose of other types of adatoms [12]. On this subject, fluorinated graphene is known to be thermodynamically more stable. Here, the carbon forms strong covalent bonds with fluorine, which is known as the most electronegative element [13]. This chemical bonding depends strongly on graphene doping and hence allows a new route to tailor the electronic and spintronics properties with a local gate similar to the one can be achieved with magnetic adatoms [14–16]. Furthermore, and unlike hydrogen, fluorine acts only as a weak resonant scatterer [17]. Therefore, higher charge mobility is expected in such weakly functionalized graphene. And lastly but most importantly, contrary to the hydrogen atoms, fluorine’s own SOC is not negligible and is expected to cause a large SOC in even weakly fluorinated graphene system [2,17]. Irmer et al., has predicted a SOC strength larger than 10 meV in weakly fluorinated graphene which is ten-fold higher what is expected in hydrogenated graphene and comparable to the atomic spin-orbit interaction in carbon itself [17]. In this letter, we study the strength of spin-orbit interaction in dilute fluorinated graphene by measuring the non-local signals of Hall bar devices. Prior to transport measurements, Raman spectroscopy is utilized to determine the fluorine adatom coverage of graphene. We observe a large non-local signals which increases with increasing the coverage. By fitting the length dependence of the non-local signal, we estimate a SOC of up to 9.1 meV. The observation of such larger SOC strength in fluorinated graphene is in a good agreement with recent studies where a large SOC-induced band splitting has been predicted [2,17]. The device fabrication starts with employing the well-known micromechanical exfoliation method to obtain single layer fluorinated graphene on Si/SiO2 wafers. However instead of using pristine HOPG graphite, we use ClF3 treated graphite. By controlling the temperature and duration of the treatment, we can achieve the transition from insulating to conducting graphene with a non-destructive recovery (See Supplementary Information). This way we can prepare weakly fluorinated graphene flakes with different fluorine concentrations. The details of the synthesis is discussed in detail elsewhere [18]. Subsequently devices are fabricated using electron beam lithography technique followed by Au/Cr electrode deposition. Following the lift off process, a second electron beam step is performed to etch the graphene into Hall bar structures. Figure 1-a shows a typical optical image of a completed device. The longitudinal spacing between the electrodes (l) is varied from 2 μm to 4 μm while keeping the width of graphene channel fixed to 1 μm. Prior to the transport measurements, the homogeneity of the fluorine coverage over the graphene surface is checked with Raman mapping (Figure 1-b). All transport measurements are performed with a four terminal ac lock-in technique under room temperature and vacuum environment (pressure ~ 1 x 10-7 mbar). Experiments are performed in two different measurement configurations (Figure 1c). In the conventional local four-terminal measurement configuration (Hall bar), a current of 1 µA flows between electrode 1 and electrode 6 and a local voltage drop is measured between electrode 2 (3) and electrode 4 (5). In the non- local measurement configuration (H bar), the current I flows between the pair of electrode 2 and electrode 3, and a non-local voltage V is recorded across the neighboring pair of electrode 4 and electrode 5. In total, we have characterized 5 dilute fluorinated graphene devices. Here we discuss two representative fluorinated devices at the maximum (device FG2) and minimum (device FG1) fluorination limits and one pristine graphene device (PG) as a control. Figure 2-a shows the Raman spectra of the flakes. Having larger 2D intensity (I2D) peak compared to G intensity (IG) peak shows that all flakes are single layer [19]. The intensity of the defect-associated graphene D band (ID) is absent in pristine graphene. The observation of higher ID and an emerging D΄ peak in FG2 compared to FG1 indicates a higher fluorine coverage in the former [18]. The Raman intensity of the ID which is normalized to the IG allows us to determine the spacing between the fluorine atoms (LD) and hence, the fluorine concentration (nimp) by using the relation [20] (cid:1838)(cid:3005)(cid:2870)(cid:4666)(cid:1866)(cid:1865)(cid:2870)(cid:4667)(cid:3404)(cid:4666)1.8(cid:3399)0.5(cid:4667)(cid:1876)10(cid:2879)(cid:2877)(cid:2019)(cid:3013)(cid:2872)(cid:4666)(cid:1835)(cid:3008) (cid:1835)(cid:3005)(cid:4667)⁄ and (cid:1866)(cid:3036)(cid:3040)(cid:3043)(cid:4666)(cid:1855)(cid:1865)(cid:2879)(cid:2870)(cid:4667)(cid:3404)10(cid:2869)(cid:2872) (cid:4666)(cid:2024)(cid:1838)(cid:3005)(cid:2870)(cid:4667) ⁄ , where λL is the wavelength of the Raman laser which is 532 nm. The ID/IG ratios for FG1 and FG2 give LD ~28 nm and 8 nm and nimp = 4 x 1011/cm2 and 4.6 x 1012/cm2 respectively. From the extracted LD values, fractions of the fluorination are estimated to be 0.005% and 0.06% for FG1 and FG2, respectively from(cid:2871)√(cid:2871)(cid:3095) (cid:4672)(cid:3028)(cid:3013)(cid:3279)(cid:4673)(cid:2870)(cid:1876)100. These imply that our devices are very weakly fluorinated and FG2 has one order of magnitude higher fluorine coverage compared to FG1. These devices are first characterized by using local charge transport experiments to confirm their homogeneity. Local charge transports are performed for top and bottom electrodes in a single Hall bar junction. Only junctions which show similar top and bottom local contributions are further used for non-local spin Hall effect (SHE) measurements. Figure 2-b shows the carrier concentration (n) dependence of the local resistivity (ρ) for FG1, FG2 and pristine graphene devices. The resistivity of FG2 is highest, followed by FG1 and pristine devices. This is in a good agreement with the impurity concentration of the devices. The full width at half maximum is largest in FG2 indicating that the sample has the lowest charge mobility. The field effect mobility of 2700 cm2/Vs, 4300 cm2/Vs and 7350 cm2/Vs are extracted for FG2, FG1 and pristine graphene devices, respectively, by using (cid:2020)(cid:3404)(cid:2869)(cid:3032) (cid:3031)(cid:3097)(cid:3031)(cid:3041) where (cid:2026)(cid:3404)(cid:2869)(cid:3096) and (cid:1866)(cid:3404)(cid:4666)7.2 (cid:1876) 10(cid:2869)(cid:2868) (cid:1855)(cid:1865)(cid:2879)(cid:2870) (cid:1848)(cid:2879)(cid:2869)(cid:4667)(cid:4666)(cid:1848)(cid:3008)(cid:3398)(cid:1848)(cid:3008)(cid:2868)(cid:4667). We note that the mobility of FG2 device is even higher than that of the hydrogenated graphene device despite the larger adatom coverage in the former [8]. In fact, it is comparable to the pristine graphene-based spin valve devices with tunnel barrier fabricated on Si/SiO2 substrate [21–23]. We now turn our attention to non-local transport measurements. We first characterize PG which does not have any functionalization treatment. The obtained non-local signal (RNL) has comparable magnitude with the Ohmic leakage contribution,(cid:1844)(cid:3016)(cid:3035)(cid:3040)(cid:3036)(cid:3030) ~(cid:2872)(cid:3095)(cid:2025)exp (cid:4666)(cid:3398)(cid:2024)(cid:1838)(cid:1875)⁄ (cid:4667), and thus there is no indication of the SHE (Inset Figure 3-a and see supplementary information) [8,24,25]. However, we observe a RNL signal approximately ~ 12 times higher than the estimated ROhmic in the FG2 device for l = 2.5 µm junction. Figure 3-b shows the n dependence of RNL at different length values for device FG2. While the non-local signal decreases as the device length is increased, all the measured signals are almost an order of magnitude larger than the expected Ohmic contributions. Thus the presence of this non-local signal at room temperature and zero magnetic field suggests the enhancement of SOC [7,8,26]. The large enhancement in SOC gives rise to the generation and detection of spin currents via the SHE and inverse SHE respectively. In order to determine the important spin parameters in our devices, we study the length dependence in FG2 compared to less fluorinated FG1 device is a direct indication of the larger spin-orbit of the non-local signal. Figure 3-c shows the RNL normalized by the local ρ at n=1x1012cm-2. At where (cid:2011) is the spin Hall zero applied field, the behavior can be fitted with R(cid:2898)(cid:2896)(cid:3404)(cid:2869)(cid:2870)(cid:2011)(cid:2870)(cid:2025)(cid:3050)(cid:3090)(cid:3268)(cid:1857)(cid:2879)(cid:3013) (cid:3090)(cid:3268)⁄ coefficient and (cid:2019)(cid:3020) is the spin relaxation length [24]. With this, we found (cid:2011) to be ~ 0.2 and ~ 0.92 and (cid:2019)(cid:3020) to be ~ 0.8 and ~ 0.34 μm for FG1 and FG2 respectively. The enhanced (cid:2011) and reduced (cid:2019)(cid:3020) strength in the former. Such small (cid:2019)(cid:3020) values are in the expected range for low mobility graphene devices. However, unlike the case of hydrogen, large (cid:2011) value especially for FG2 is surprisingly above to estimate the SOC strength assumes that le<<w<<(cid:2019)s where le (~36 nm in FG2) is the electron mean free path, and (cid:2019)s (~ 340 nm in FG2) is the electron spin relaxation length. Since it large and doubtful. While we ruled out all the possible sources of leakage currents in our measurements, such large spin Hall angle might be due to the analysis. The equation we used is very difficult to satisfy this condition experimentally, spin Hall angle values might not be truly extracted with the existing theory. Similar to our case, a recent study using the same theory for extracting the spin Hall angle found much larger values than the theoretically predicted values[27]. Further theoretical studies that satisfy the experimental conditions are required for the more accurate values. In order to further quantity the devices, we now calculate the strength of SOC in fluorinated graphene devices. Following similar arguments used in hydrogenated graphene devices, we assume Elliott-Yafet type spin scattering mechanism as the dominant dephasing mechanism [8,9,23,28]. With this, we have ∆(cid:3020)(cid:3016)(cid:3004)(cid:3404)(cid:1831)(cid:3007)(cid:3493)(cid:2028)(cid:3017) (cid:2028)(cid:3020)⁄ , where EF is the Fermi energy, (cid:2028)(cid:3017) and (cid:2028)(cid:3020) are the momentum and spin relaxation times, respectively. We obtain a SOC of 5.1 and 9.1 meV for FG1 and FG2 samples at n = 1 x 1012 cm-2, respectively. The observation of larger SOC in FG2 compared to FG1 is likely due to the presence of higher fluorination coverage. We note that the SOCs extracted for FG1 and FG2 device are nearly two and four times of what has been extracted for hydrogenated graphene devices at similar carrier concentrations, respectively [8]. In order to study the origin of this large SOC in fluorinated graphene compared to its hydrogenated counterparts, we estimate the out of plane distortion angle by assuming all SOC is caused by the sp3 hybridization of carbon atoms. The distortion angles for FG1 and FG2 devices are extracted to be 9.9° and 19.2° for FG1 and FG2 samples, respectively by using ∅(cid:3404) (cid:1827)(cid:1870)(cid:1855)tan(cid:4676)(cid:3427)14(cid:3398)(cid:4666)9(cid:3398)8(cid:1870)(cid:2868)(cid:2870)(cid:4667)(cid:2869)(cid:2870)⁄/12 (cid:3431)(cid:2869)(cid:2870)⁄(cid:4677) [9]. The obtained 19.2° in our weakly fluorinated ⁄ devices is surprising since the distortion angle for even a full sp3 hybridization is 19.5° [29]. This implies that such large SOC cannot be explained only by the lattice deformation and hence there are additional contributions to the SOC in fluorinated graphene. In fact, recently it has been discussed independently by few studies that the fluorinated graphene should have larger SOC compared to the hydrogenated graphene due to the intrinsic SOC of the fluorine atoms [2,17]. Since the obtained SOC values in this study are comparable to the atomic spin orbit interaction in carbon and much higher than the hydrogenated graphene case[9], SOC in fluorinated graphene is not only from the graphene lattice but likely from the fluorine’s own SOC[17]. The extracted spin parameters in these devices are summarized in Table1. We also extracted the spin parameters in a reference pristine-graphene based spin valve device for comparison. In summary, we observe SHE in devices made from fluorinated graphene due to the enhancement of spin-orbit interaction in this defected system while, preserving high mobility of graphene. Large non-local signals are detected even without applying any external magnetic field. Observation of very strong spin orbit interactions in the order of 9 meV cannot be explained only with the out of plane distortion of the carbon bonds which is the main SOC source in hydrogenated graphene. Based on the recent theories, we believe that the spin orbit strength of fluorine adatom itself is the main source of the enhancement in our devices. The elimination of ferromagnetic contacts in these devices is a major advantage for the development of graphene-based spintronics applications. B. Ö. would like to acknowledge support by the National Research Foundation, Prime Minister’s Office and the SMF-NUS Research Horizons Award 2009-Phase II. References [1] Gmitra M, Konschuh S, Ertler C, Ambrosch-Draxl C and Fabian J 2009 Band-structure topologies of graphene: Spin-orbit coupling effects from first principles Phys. Rev. B 80 235431 [2] Zhou J, Liang Q and Dong J 2010 Enhanced spin–orbit coupling in hydrogenated and fluorinated graphene Carbon N. Y. 48 1405–9 [3] Huertas-Hernando D, Guinea F and Brataas A 2006 Spin-orbit coupling in curved graphene, fullerenes, nanotubes, and nanotube caps Phys. Rev. B 74 155426 [4] Datta S and Das B 1990 Electronic analog of the electro-optic modulator Appl. Phys. Lett. 56 665 [5] Kato T, Onari S and Inoue J 2010 Spin Hall effect in a curved graphene with spin–orbit interaction Phys. E Low-dimensional Syst. Nanostructures 42 729–31 [6] Kuemmeth F, Ilani S, Ralph D C and McEuen P L 2008 Coupling of spin and orbital motion of electrons in carbon nanotubes. Nature 452 448–52 [7] Avsar A, Tan J Y, Taychatanapat T, Balakrishnan J, Koon G K W, Yeo Y, Lahiri J, Carvalho A, Rodin A S, O’Farrell E C T, Eda G, Castro Neto A H and Ozyilmaz B 2014 Spin-orbit proximity effect in graphene. Nat. Commun. 5 4875 [8] Balakrishnan J, Kok Wai Koon G, Jaiswal M, Castro Neto A H and Özyilmaz B 2013 Colossal enhancement of spin–orbit coupling in weakly hydrogenated graphene Nat. Phys. 9 284–7 [9] Castro Neto A and Guinea F 2009 Impurity-Induced Spin-Orbit Coupling in Graphene Phys. Rev. Lett. 103 026804 [10] Nair R R, Ren W, Jalil R, Riaz I, Kravets V G, Britnell L, Blake P, Schedin F, Mayorov A S, Yuan S, Katsnelson M I, Cheng H-M, Strupinski W, Bulusheva L G, Okotrub A V, Grigorieva I V, Grigorenko A N, Novoselov K S and Geim A K 2010 Fluorographene: a two-dimensional counterpart of Teflon. Small 6 2877–84 [11] Ryu S, Han M Y, Maultzsch J, Heinz T F, Kim P, Steigerwald M L and Brus L E 2008 Reversible basal plane hydrogenation of graphene. Nano Lett. 8 4597–602 [12] Suarez A M, Radovic L R, Bar-Ziv E and Sofo J O 2011 Gate-Voltage Control of Oxygen Diffusion on Graphene Phys. Rev. Lett. 106 146802 [13] Zbořil R, Karlický F, Bourlinos A B, Steriotis T A, Stubos A K, Georgakilas V, Šafářová K, Jančík D, Trapalis C and Otyepka M 2010 Graphene fluoride: a stable stoichiometric graphene derivative and its chemical conversion to graphene. Small 6 2885–91 [14] Sofo J O, Suarez A M, Usaj G, Cornaglia P S, Hernández-Nieves A D and Balseiro C A 2011 Electrical control of the chemical bonding of fluorine on graphene Phys. Rev. B 83 081411 [15] Guzmán-Arellano R M, Hernández-Nieves A D, Balseiro C A and Usaj G 2014 Diffusion of fluorine adatoms on doped graphene Appl. Phys. Lett. 105 121606 [16] Ming W and Liu F 2014 Tuning nucleation density of metal island with charge doping of graphene substrate Appl. Phys. Lett. 105 071609 [17] Irmer S, Frank T, Putz S, Gmitra M, Kochan D and Fabian J 2015 Spin-orbit coupling in fluorinated graphene Phys. Rev. B 91 115141 [18] Lee J H, Koon G K W, Shin D W, Fedorov V E, Choi J-Y, Yoo J-B and Özyilmaz B 2013 Property Control of Graphene by Employing “Semi-Ionic” Liquid Fluorination Adv. Funct. Mater. 23 3329–34 [19] Ferrari A C, Meyer J C, Scardaci V, Casiraghi C, Lazzeri M, Mauri F, Piscanec S, Jiang D, Novoselov K S, Roth S and Geim A K 2006 Raman Spectrum of Graphene and Graphene Layers Phys. Rev. Lett. 97 187401 [20] Cançado L G, Jorio A, Ferreira E H M, Stavale F, Achete C A, Capaz R B, Moutinho M V O, Lombardo A, Kulmala T S and Ferrari A C 2011 Quantifying defects in graphene via Raman spectroscopy at different excitation energies. Nano Lett. 11 3190–6 [21] Tombros N, Jozsa C, Popinciuc M, Jonkman H T and van Wees B J 2007 Electronic spin transport and spin precession in single graphene layers at room temperature. Nature 448 571–4 [22] Han W and Kawakami R K 2011 Spin Relaxation in Single-Layer and Bilayer Graphene Phys. Rev. Lett. 107 047207 [23] Avsar A, Yang T-Y, Bae S, Balakrishnan J, Volmer F, Jaiswal M, Yi Z, Ali S R, Güntherodt G, Hong B H, Beschoten B and Özyilmaz B 2011 Toward wafer scale fabrication of graphene based spin valve devices. Nano Lett. 11 2363–8 [24] Abanin D, Shytov A, Levitov L and Halperin B 2009 Nonlocal charge transport mediated by spin diffusion in the spin Hall effect regime Phys. Rev. B 79 035304 [25] Mihajlović G, Pearson J, Garcia M, Bader S and Hoffmann A 2009 Negative Nonlocal Resistance in Mesoscopic Gold Hall Bars: Absence of the Giant Spin Hall Effect Phys. Rev. Lett. 103 166601 [26] Balakrishnan J, Koon G K W, Avsar A, Ho Y, Lee J H, Jaiswal M, Baeck S-J, Ahn J-H, Ferreira A, Cazalilla M A, Castro Neto A H and Özyilmaz B 2014 Giant spin Hall effect in graphene grown by chemical vapour deposition. Nat. Commun. 5 4748 [27] Ehlert M, Song C, Ciorga M, Hupfauer T, Shiogai J, Utz M, Schuh D, Bougeard D and Weiss D 2014 All-electrical detection of spin Hall effect in semiconductors Phys. status solidi 251 1725–35 [28] Soriano D, Tuan D Van, Dubois S M-M, Gmitra M, Cummings A W, Kochan D, Ortmann F, Charlier J-C, Fabian J and Roche S 2015 Spin transport in hydrogenated graphene 2D Mater. 2 022002 [29] Haruyama J 2013 Graphene and Graphene Nanomesh Spintronics Electronics 2 368–86 Figure and Table Captions FIG. 1. (a) Optical image of a completed device with multiple Hall bar structures. (b) Raman intensity maps of the D, G and 2D bands for the device shown in (a). (c) Device schematics for the local and non-local measurement configurations FIG. 2. (a) Raman spectrum study of pristine graphene, FG1 and FG2 devices. (b) Resistivity of pristine graphene, FG1 and FG2 devices as a function of carrier concentration at room temperature FIG. 3. (a) Non-local resistance measurement of FG2 device for l/w = 2.5 as a function of carrier concentration. Inset: Non-local measurement of pristine graphene device. l/w is 2 in this device (b) Non-local resistance FG2 device measured as a function of carrier concentration at different l/w. (c) Non-local resistance at n=1x1012cm-2as a function of length for FG1 and FG2. TABLE1. Extracted spin parameters of FG1 and FG2 devices at carrier concentration of 1 x 1012 cm-2. Spin parameters of a pristine-graphene measured with ferromagnetic electrodes are also included as reference. Figures (a) 5 μm (c) SiO2 Cr/Au Si (b) D Peak G Peak 2D Peak Figure 1 (a) 3 2 1 0 2D PG FG1 FG2 D G D’ (b) )  k (  5 4 3 2 1 PG FG1 FG2 1200 1400 1600 2400 2800 0 -3 -2 0 1 -1 n (x1012 cm-2) 2 3 Figure 2 RNL ROhmic (b) 70 FG2 -1 0 1 2 3 n (1012 cm-2) )  ( L N R RNL ROhmic PG 12 )  ( L N R 8 4 0 -2 (a) )  ( L N R 25 20 15 10 5 0 FG1 FG2 FG2 (c) ) 3 - 0 1 (   / L N R 14 12 10 8 6 4 2 0 2 2.0 2.5 3.5 4.0 3.0 L (m) 60 50 40 30 20 10 0 -3 l/w: 2 2.5 3 3.5 4 -2 0 -1 n (x1012 cm-2) 1 Figure 3 -3 -2 0 -1 1 n (x1012 cm-2) 2 3 Table 1 Sample µ (cm2/Vs) FG1 FG2 Pristine‐SV 4,600 2,700 2,900 τS (ps) 25.8 5.9 135 λS (μm) ΔSOC (meV) θ (degree) 0.8 0.34 1.5 5.1 9.1 9.9 19.2
1812.11523	2	1812	2019-04-23T16:49:17	Optical coherence of diamond nitrogen-vacancy centers formed by ion implantation and annealing	[ "cond-mat.mes-hall", "physics.optics", "quant-ph" ]	The advancement of quantum optical science and technology with solid-state emitters such as nitrogen-vacancy (NV) centers in diamond critically relies on the coherence of the emitters' optical transitions. A widely employed strategy to create NV centers at precisely controlled locations is nitrogen ion implantation followed by a high-temperature annealing process. We report on experimental data directly correlating the NV center optical coherence to the origin of the nitrogen atom. These studies reveal low-strain, narrow-optical-linewidth ($<500$ MHz) NV centers formed from naturally-occurring $^{14}$N atoms. In contrast, NV centers formed from implanted $^{15}$N atoms exhibit significantly broadened optical transitions ($>1$ GHz) and higher strain. The data show that the poor optical coherence of the NV centers formed from implanted nitrogen is not due to an intrinsic effect related to the diamond or isotope. These results have immediate implications for the positioning accuracy of current NV center creation protocols and point to the need to further investigate the influence of lattice damage on the coherence of NV centers from implanted ions.	cond-mat.mes-hall	cond-mat	Optical coherence of diamond nitrogen-vacancy centers formed by ion implantation and annealing S. B. van Dam,1, 2, ∗ M. Walsh,3, † M. J. Degen,1, 2 E. Bersin,3 S. L. Mouradian,3, ‡ A. Galiullin,1, 2 M. Ruf,1, 2 M. IJspeert,1, 2 T. H. Taminiau,1, 2 R. Hanson,1, 2, § and D. R. Englund3 1QuTech, Delft University of Technology, PO Box 5046, 2600 GA Delft, The Netherlands 2Kavli Institute of Nanoscience, Delft University of Technology, PO Box 5046, 2600 GA Delft, The Netherlands 3Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA The advancement of quantum optical science and technology with solid-state emitters such as nitrogen-vacancy (NV) centers in diamond critically relies on the coherence of the emitters' opti- cal transitions. A widely employed strategy to create NV centers at precisely controlled locations is nitrogen ion implantation followed by a high-temperature annealing process. We report on ex- perimental data directly correlating the NV center optical coherence to the origin of the nitrogen atom. These studies reveal low-strain, narrow-optical-linewidth (< 500 MHz) NV centers formed from naturally-occurring 14N atoms. In contrast, NV centers formed from implanted 15N atoms exhibit significantly broadened optical transitions (> 1 GHz) and higher strain. The data show that the poor optical coherence of the NV centers formed from implanted nitrogen is not due to an intrinsic effect related to the diamond or isotope. These results have immediate implications for the positioning accuracy of current NV center creation protocols and point to the need to further investigate the influence of lattice damage on the coherence of NV centers from implanted ions. Coherent optical control over solid-state quantum emitters has enabled new advances in quantum science [1 -- 3] and may lead to technologies such as quantum net- works [4]. A quantum network crucially relies on en- tanglement connections that can be established through a coherent spin-photon interface. The nitrogen-vacancy (NV) defect center in diamond is a well-suited candidate owing to a spin ground state with a long coherence time [5, 6], nearby nuclear spins for quantum memories [7] or algorithms [8 -- 10], and spin-selective optical transitions allowing for efficient optical spin initialization and read- out [11]. Moreover, at low strain and low temperature (< 10 K), effects from phonon mixing in the excited state are small [12, 13], and the optical transition can be coherent. Indeed, narrow-linewidth, coherent optical transitions [14 -- 16] have been used for the generation of indistinguishable photons suited for two photon quan- tum interference [17, 18] and entanglement generation between remote NV centers [19]. To date, all experiments employing coherent photons from NV centers have been performed with NV centers that were formed during diamond growth. Key to their optical coherence is that these NV centers experience an environment with few defects, since the stability of opti- cal transitions (as with many solid-state systems) suffers from unwanted interactions with nearby bulk and surface defects leading to changes in the strain and electric field environment [20 -- 24]. For NV centers with a broadened ∗ These authors contributed equally. † These authors contributed equally.; [email protected] ‡ Present Address: Department of Physics, University of California Berkeley, California 94720, USA § [email protected] linewidth below ≈ 200 MHz dominated by slow spectral diffusion, protocols using resonant charge repumping [25] and real-time monitoring of the transition frequency [26] have been used to reduce the broadened linewidth to an effective linewidth of below 50 MHz, suitable for quan- tum optical experiments. However, such protocols are challenging for NV centers with greater spectral diffu- sion. Instead of being limited to NV centers formed during diamond growth, they can be created, for example by nitrogen ion implantation [27]. Nitrogen ion implanta- tion provides an NV positioning accuracy that enables integration with on-chip photonics [28, 29] and coupling between nearby NV centers [30 -- 32]. Precise positioning of NV centers or accurately registering their location is also a prerequisite for optimal overlap of the dipole with the electric field mode of diamond optical cavities, for en- gineering and enhancing light-matter interaction [24, 33 -- 38]. Moreover, ion implantation allows for the creation of single NV centers in high purity diamond, providing a potentially low-defect environment [22]. However, the bombardment of the diamond with nitro- gen ions creates crystal damage that can deteriorate spin and optical coherence properties of NV centers [21, 22]. High-temperature annealing can mitigate some of these issues by repairing the diamond lattice [39 -- 42]. A proce- dure including a low implantation dose, careful cleaning, and high-temperature annealing was reported by Chu et al. [43], leading to the creation of narrow-linewidth NV centers. These narrow-linewidth NV centers can result from implanted nitrogen atoms, or from native nitrogen atoms, combined with for example implantation-induced vacancies (Figure 1a). In principle, the source of nitro- gen can be verified by implanting 15N isotopes (natural abundance 0.37%) and resolving the hyperfine structure 2 simulations to predict the stopping point of implanted 15N atoms, in addition to the locations of vacancies cre- ated along the trajectory (Figure 1b). At temperatures > 600 ◦C, vacancies become mobile [57]. These vacancies can form an NV center, recombining with the implanted 15N that created the damage or with a native 14N in the lattice. The resulting 15NV and 14NV formation yields can vary significantly [30, 45 -- 48] depending on several factors, including the initial nitrogen concentration, the implantation fluence and energy, the number of vacancies created during the implantation process, and the dura- tion and temperature of annealing. A representative confocal fluorescence map at the im- plantation depth in sample A is shown in Figure 2a. Confocal fluorescence scans at foci deeper into the di- amond show a significantly lower density of fluorescent spots (Figure 2a, inset), indicating that the emitters near the surface were predominantly created by the implanta- tion and annealing process [51]. We identified emitters using different protocols in the two samples. In sam- ple A, automated spot-recognition was performed on a fluorescence scan. For each detected spot we identified an NV center based on its characteristic zero-phonon line (ZPL) emission around 637 nm using a spectrograph from a photoluminescence measurement at 4 K under 532 nm excitation. This protocol identified 120 fluorescent spots as NV centers in a ≈ 400 µm2 area. In sample B, spots in a fluorescence scan were detected visually, after which an automated protocol identified NV centers based on the presence of a resonance in an optically detected magnetic resonance (ODMR) spectrum around the characteristic NV center zero-field splitting of 2.88 GHz. In this way, 52 out of a total 57 inspected spots in a ≈ 75 µm2 area in the implantation layer were identified as NV centers. We next determined the nitrogen isotope of each NV center by observing the hyperfine structure of the ODMR spectra. A weak external magnetic field (B(cid:107) ≈ 5-10 G) was applied to separate the ms = −1 and ms = +1 electron spin transitions. We found NV centers with the characteristic triplet splitting of the 14NV (with hyper- fine splitting, A = 2.2 MHz) as well as with the 15NV doublet (A = 3.1 MHz) [58], as indicated in Figure 2b. Of the 120 NVs identified on sample A, an ODMR signal was detected in 50, out of which 18 were 15NV, 18 14NV, and there were 14 in which the isotope could not be re- liably determined from the ODMR spectra. Similarly, of the 52 NVs identified on sample B, 34 were 15NV, 3 were 14NV, and the isotope could not be determined in 15 NVs. We attribute the different isotope occurrence ratios in sample A and B to different native 14N content and different implantation fluence. Subsequently, we measured the linewidth of optical transitions of identified NV centers, recording photolu- minescence excitation (PLE) spectra at low temperature (≈ 4 K). A tunable laser with a wavelength near 637 nm was scanned over the optical transition while detecting emitted photons in the phonon-sideband. We performed two types of measurements. First, a scan was made in FIG. 1. NV creation via nitrogen-ion implantation. a, Schematic showing implanted 15N+ ions (orange) leaving a trail of vacancies (purple) until settling into a final posi- tion. Naturally abundant 14N ions (green) are shown ran- domly distributed throughout the diamond lattice. Vacancies (implantation-induced or native) mobilized by annealing can bind to a nitrogen atom (implanted or native). b, A SRIM simulation using the parameters in sample A show the distri- bution of implanted nitrogen (orange) and created vacancies (purple). The shaded green area indicates the range of the estimated natural 14N concentration reported by Element 6. of the NV magnetic spectrum [44], as done in studies of the spin coherence [41, 45] and creation efficiency [30, 46 -- 49] of NVs formed from implanted nitrogen. However, in Chu et al. [43] the isotope of the narrow-linewidth NV centers was not investigated [50]. In a later study with similar results [38], 14N isotopes were implanted, so that the origin of the NV center's nitrogen atom could not be determined. Here we report on a study that enables us to directly correlate the optical linewidth of NV transitions to the NV formation mechanism. To distinguish NVs formed by implanted nitrogen atoms from those formed by native nitrogen atoms, we implanted 15N isotopes [44]. We then experimentally cor- related the optical linewidth to the nitrogen isotope. The study was carried out on two separate samples [51]. Sam- ple A (processed at MIT) is a bulk (cid:104)100(cid:105) CVD grown diamond (Element 6), prepared with the same implan- tation and annealing procedure as presented in Chu et it was implanted with 15N+ at 85 keV with al. a fluence of 109 N/cm2 and subsequently annealed at a maximum temperature of 1200 ◦C. Sample B (Delft) is a membrane (thickness ≈ 14 µm) obtained from a bulk (cid:104)100(cid:105) CVD grown diamond (Element 6), implanted with 15N+ at 400 keV (fluence 108 N/cm2), and subsequently annealed at a maximum temperature of 1100 ◦C. [43]: During implantation, nitrogen ions penetrate the dia- mond to a depth determined by the implantation energy (Figure 1). As implanted nitrogen atoms track through the crystal, they displace carbon atoms from their lattice sites creating vacancies. The nitrogen atoms create dam- age along the entire trajectory, but the damage is greatest near the stopping point [55]. We performed SRIM [56] vacanciesnative 14Nimplanted 15N14NV creation15N+15NV creation3 FIG. 2. Isotope characterization and optical measurements of NV centers from 14N and 15N. a, A fluorescent confocal scan of sample A taken at 4 K, with labels indicating NV centers characterized as 14NV, 15NV, and a set with unresolvable hyperfine lines, labeled as ?NV. A scan a few microns below the implanted layer (inset) shows a lower NV density. b-d, Pulse sequences (top row) used for isotope characterisation and optical measurements, and representative measurement results for each isotope (green, middle row: 14NV, orange, bottom row: 15NV). b, Continuous wave (CW) ODMR measurements reveal the NV isotope. The 14NV is characterized by S = 1 hyperfine transitions; the 15NV by S = 1/2 hyperfine transitions. c, Interleaved red and green excitation probe the combined effect of short-timescale fluctuations and laser-induced spectral diffusion. The Ex and Ey ZPL transitions are visible for both isotopes; the 14NV linewidths are narrower and show a smaller strain splitting than the 15NV. d, Individual line scans of the ZPL in sample B reveal the linewidth free from laser-induced spectral diffusion. The summation of many repeated scans is broadened as a result of repump-laser-induced spectral diffusion. which resonant excitation (637 nm) and green illumina- tion (532 nm) were rapidly interleaved at each data point. The red excitation causes rapid optical spin pumping and ionization of the NV center. The green excitation pro- vides repumping into the negative charge state and the ms = 0 spin state. This measurement reveals the com- bined effect of short-time scale fluctuations and repump- laser-induced spectral diffusion in broadening the transi- tion linewidth. Examples of the resulting traces are seen in Figure 2c. Second, in sample B, an additional scan was performed to isolate the effect of short timescale fluctuations from repump-laser-induced diffusion. A single off-resonant re- pump was applied before sweeping the resonant laser at low power, as seen in Figure 2d. We applied microwaves on the spin resonances to prevent optical pumping into a dark spin state during the sweeps [59]. Remaining traces in which the NV center ionized were excluded by apply- ing a second scan over the resonance to check the charge state. If no resonance was observed, the preceding trace was disregarded. This scanning protocol was repeated many times to probe spectral diffusion through the re- sulting spread of the observed lines [12]. To extract the linewidth free from repump-laser-induced spectral diffu- sion, we performed a weighted average of linewidth values found from Lorentzian fits to each individual scan. Figures 2c and d display representative resonant opti- cal scans for the 14NV and 15NV centers, each showing two resonances corresponding to the two ms = 0 orbital To correlate transitions Ex and Ey. Notably, while the 14NV center (green, top row) exhibits a narrow optical linewidth with a full-width-at-half-maximum (FWHM) of 64 ± 4 MHz, the 15NV linewidth (orange, bottom row) is broad, with a FWHM of 860± 236 MHz. The dynamics in the second scan type (Figure 2d) indicate that both repump-induced fluctuations and a short-timescale mechanism broaden the 15NV linewidth, but that repump-induced fluctua- tions are dominant in broadening beyond 200 MHz [51]. the occurrence of narrow optical linewidths with the N isotope of the NV centers, we ac- quired an extensive data set using the data accumulation procedures described above. The resulting distributions of optical linewidths for both N isotopes are shown in Fig- ure 3. Narrow optical linewidths in both samples can be attributed almost exclusively to NVs with a native 14N host. In contrast, 15NV centers exhibiting narrow optical linewidths are extremely rare, with a median linewidth for 15NV centers of 3.1 GHz in sample A and 4.1 GHz in sample B. Notably, in both datasets one NV center with a 15N host was found that showed narrow optical linewidths (< 100 MHz). Given their low occurrence and the non- zero natural abundance of 15N, the creation mechanism of these narrow-linewidth 15NVs cannot be conclusively determined. Nevertheless, their presence demonstrates that 15NV centers can exhibit coherent optical transi- tions. Therefore, we conclude that the difference in dis- tribution of optical linewidths between 14NVs and 15NVs 1 m14NV15NV ?NV2.862.8652.87Frequency (GHz)0.940.960.981Contrast2.8452.852.855Frequency (GHz)0.940.960.981Contrast-101Detuning (GHz)024Counts (kHz)-505Detuning (GHz)012Counts (kHz)-101Detuning (GHz)Counts-101Detuning (GHz)Counts2 mIterationIterationStep !Δ=0.87GHz72 MHz64 MHzΔ=2.7GHz860 MHz974 MHz96 MHz630 MHz(a)(b)(c)(d))=3.1MHz)=2.2MHz)))532 nmMWNM532 nm637 nmNStep !M532 nm637 nmNStep !MΔΔΔΔMW4 FIG. 3. Optical linewidths per isotope. a-b, A summary of the optical linewidths identified in sample A (a) and sample B (b) from scans at the implantation depth. For sample B, that has comparatively few 14NV centers at the implantation depth, we included three 14NV centers found deeper in the diamond to enable a comparison between NVs formed from implanted versus native nitrogen. The distribution is represented as a cumulative distribution function (CDF, top), with the corresponding histogram shown below. The shaded region in the CDF indicates a 95% confidence interval calculated using Greenwood's formula. These data show that both diamonds supported narrow-linewidth NV centers, the majority of which originated from 14NVs. While 15NV do exhibit narrow lines, their median linewidth (M) is higher than for the 14NV centers in both samples. We evaluate the probability to obtain the observed linewidths for 14NV centers and 15NV centers if the samples are drawn from the same distribution with a Wilcoxon Rank Sum test, finding a p-value of 2.5 × 10−4 in sample A and 1.7 × 10−3 in sample B. c-d, A magnification of the histograms shown in a and b. is not due to an intrinsic effect related to the isotope it- self, but due to differences in the local environment re- sulting from the implantation process. Damage due to implantation may cause local strain fields. Axial strain results in an overall shift of the optical transition, while transverse strain will split the Ex and Ey transitions [58]. The distributions characterizing the strain for both NV isotopes are shown in Figure 4. The spread of the distribution in ZPL detuning representing axial strain for 15NVs (44 GHz for sample A, 60 GHz for sample B) is wider than for 14NVs (13 GHz for sample A, 5.6 GHz for sample B). Further, we found that 15NVs exhibit higher transverse strain, manifested by greater splitting with a median of 5.4 GHz (10 GHz) compared to 1.7 GHz (4.3 GHz) for 14NVs in sample A (B). Assum- ing a similar strain susceptibility for both isotopes, these results indicate that local damage around the implanted 15NVs creates a more strained environment, providing further evidence that implantation-induced local damage is responsible for the broadened 15NV linewidth. In ad- dition, in both samples we observed a shift of the average ZPL frequency for 15NV compared to 14NV ZPLs, pos- sibly due to an intrinsic dependency of the energy levels on the isotope as observed in other color centers [60, 61]. The data show indications of an increase in 14NV den- sity in the implantation layer in both samples [51]. We hypothesize that these 14NV centers can be formed from naturally occurring nitrogen combining with vacancies created during implantation. Since they can be at greater distance from the main damage center near the stopping point of the nitrogen, these NV centers may be coher- ent and useable for quantum information purposes; more work is needed for a statistically significant correlation. However, these NVs would have worse positioning ac- curacy as their spatial distribution is set by arbitrarily positioned naturally occurring nitrogen in combination with the diffusion length of the vacancies generated dur- ing implantation. In summary, the implanted nitrogen atoms yield NV centers with predominantly broad optical lines (> 1 GHz) and substantially higher strain than NV centers formed from native nitrogen. These results indicate that im- 00.51CDFSample Ap = 2.510-401234567Linewidth (GHz)051015OccurrencesM14 = 0.33 GHzM15 = 3.1 GHz00.511.5Linewidth (GHz)0246Occurrences00.51CDFSample Bp = 1.710-305101520Linewidth (GHz)0510OccurrencesM14 = 0.14 GHzM15 = 4.1 GHz00.511.522.53Linewidth (GHz)0123Occurrences14NV15NV(a)(c)(b)(d)15NV at >20GHz5 FIG. 4. Strain analysis. a-b, The distribution of axial strain (measured by absolute average ZPL frequency) in NVs acquired from analysis of sample A (a) and sample B (b). The 15NV ZPLs exhibit a larger spread in axial strain (standard deviation, σ) than the 14NV ZPLs. c-d, The distribution of transverse strain (measured by half the splitting between Ex and Ey frequencies) in NV centers of sample A (c) and sample B (d). The 15NV ZPLs show a greater median splitting (M) in both samples. planted nitrogen atoms combined with an annealing pro- cess at high temperatures do not routinely produce NV centers with narrow optical linewidths. Vacancies pro- duced in the implantation process may combine with ex- isting nitrogen atoms to produce narrow NVs, but more work is needed for a statistically significant correlation. It is clear from this work that recipes for generating im- planted NV centers should be re-investigated, addressing local lattice damage associated with implanted nitrogen. In addition, other approaches for precisely controlling the NV centers' positions while causing minimal local damage can be further explored, such as employing 2D nitrogen-doped diamond layers combined with electron irradiation or ion implantation for vacancy production [62, 63] or laser writing strategies for creating vacancies with 3D accuracy [64, 65]. ACKNOWLEDGMENTS The authors would like to thank L. Childress, Y. Chu, M. Loncar, P. Maletinsky, M. Markham, M. Trusheim, and J. Wrachtrup for helpful discussions, S. Meesala, and P. Latawiec for sample annealing at Harvard Un- versity, and S. Bogdanovic and M. Liddy for help with sample preparation. This work was supported in part by the AFOSR MURI for Optimal Measurements for Scalable Quantum Technologies (FA9550-14-1-0052) and by the AFOSR program FA9550-16-1-0391, supervised by Gernot Pomrenke. Further, we acknowledge support by the Netherlands Organization for Scientific Research (NWO) through a VIDI grant, a VICI grant, and through the Frontiers of Nanoscience program, the European Re- search Council through a Consolidator Grant (QNET- WORK) and a Synergy Grant (QC-LAB), and the Royal Netherlands Academy of Arts and Sciences (KNAW) and Ammodo through an Ammodo science award. M.W. was supported in part by the STC Center for Integrated Quantum Materials (CIQM), NSF Grant No. DMR- 1231319, in part by the Army Research Laboratory Cen- ter for Distributed Quantum Information (CDQI), and in part by Master Dynamic Limited. E.B. was supported by a NASA Space Technology Research Fellowship and the NSF Center for Ultracold Atoms (CUA). S.L.M. was supported in part by the NSF EFRI-ACQUIRE pro- gram Scalable Quantum Communications with Error- Corrected Semiconductor Qubits and in part by the AFOSR Quantum Memories MURI. -150-100-50050100150Absolute Frequency (GHz)0246810OccurrencesSample A14 = 13 GHz15 = 44 GHz(a)+470.5 THz0510152025Splitting (GHz)0246810OccurrencesM14 = 1.7 GHzM15 = 5.4 GHz(c)-150-100-50050100150Absolute Frequency (GHz)02468OccurrencesSample B14 = 5.6 GHz15 = 61 GHz(b)+470.5 THz14NV15NV010203040506070Splitting (GHz)01234OccurrencesM14 = 4.3 GHzM15 = 10 GHz(d)6 [1] I. Aharonovich, D. Englund, and M. Toth, Nat. Photon- ics 10, 631 (2016). [2] D. D. Awschalom, R. Hanson, J. Wrachtrup, and B. B. Zhou, Nat. Photonics 12, 516 (2018). [23] C. Santori, P. E. Barclay, K.-M. C. Fu, R. G. Beausoleil, S. Spillane, and M. Fisch, Nanotechnology 21, 274008 (2010). [24] A. Faraon, C. Santori, Z. Huang, V. M. Acosta, and [3] P. Lodahl, Quantum Sci. Technol. 3, 13001 (2018). [4] S. Wehner, D. Elkouss, and R. Hanson, Science 362, 303 R. G. Beausoleil, Phys. Rev. Lett. 109, 033604 (2012). [25] P. Siyushev, H. Pinto, M. Voros, A. Gali, F. Jelezko, and (2018). [5] N. Bar-Gill, L. M. Pham, A. Jarmola, D. Budker, and R. L. Walsworth, Nat. Commun. 4, 1743 (2013). [6] M. H. Abobeih, J. Cramer, M. A. Bakker, N. Kalb, M. Markham, D. J. Twitchen, and T. H. Taminiau, Nat. Commun. 9, 2552 (2018). [7] N. Kalb, A. A. Reiserer, P. C. Humphreys, J. J. W. Bak- ermans, S. J. Kamerling, N. H. Nickerson, S. C. Ben- jamin, D. J. Twitchen, M. Markham, and R. Hanson, Science 356, 928 (2017). [8] G. Waldherr, Y. Wang, S. Zaiser, M. Jamali, T. Schulte- Herbruggen, H. Abe, T. Ohshima, J. Isoya, J. F. Du, P. Neumann, and J. Wrachtrup, Nature 506, 204 (2014). [9] J. Cramer, N. Kalb, M. A. Rol, B. Hensen, M. S. Blok, M. Markham, D. J. Twitchen, R. Hanson, and T. H. Taminiau, Nat. Commun. 7, 11526 (2016). [10] I. Lovchinsky, A. O. Sushkov, E. Urbach, N. P. D. Leon, S. Choi, K. D. Greve, R. Evans, R. Gertner, C. Muller, L. Mcguinness, F. Jelezko, R. L. Walsworth, H. Park, and M. D. Lukin, Science 351, 836 (2016). [11] L. Robledo, L. Childress, H. Bernien, B. Hensen, P. F. A. Alkemade, and R. Hanson, Nature 477, 574 (2011). [12] K.-M. C. Fu, C. Santori, P. E. Barclay, L. J. Rogers, N. B. Manson, and R. G. Beausoleil, Phys. Rev. Lett. 103, 256404 (2009). [13] M. L. Goldman, M. W. Doherty, A. Sipahigil, N. Y. Yao, S. D. Bennett, N. B. Manson, A. Kubanek, and M. D. Lukin, Phys. Rev. B 91, 165201 (2015). [14] P. Tamarat, T. Gaebel, J. R. Rabeau, M. Khan, A. D. Greentree, H. Wilson, L. C. L. Hollenberg, S. Prawer, P. Hemmer, F. Jelezko, and J. Wrachtrup, Phys. Rev. Lett. 97, 083002 (2006). [15] A. Batalov, C. Zierl, T. Gaebel, P. Neumann, I. Y. Chan, G. Balasubramanian, P. R. Hemmer, F. Jelezko, and J. Wrachtrup, Phys. Rev. Lett. 100, 077401 (2008). [16] L. Robledo, H. Bernien, I. van Weperen, and R. Hanson, Phys. Rev. Lett. 105, 177403 (2010). [17] H. Bernien, L. Childress, L. Robledo, M. Markham, and R. Hanson, Phys. Rev. Lett. 108, D. Twitchen, 043604 (2012). [18] A. Sipahigil, M. L. Goldman, E. Togan, Y. Chu, M. Markham, D. J. Twitchen, A. S. Zibrov, A. Kubanek, and M. D. Lukin, Phys. Rev. Lett. 108, 143601 (2012). [19] H. Bernien, B. Hensen, W. Pfaff, G. Koolstra, M. S. Blok, L. Robledo, T. H. Taminiau, M. Markham, D. J. Twitchen, L. Childress, and R. Hanson, Nature 497, 86 (2013). J. Wrachtrup, Phys. Rev. Lett. 110, 167402 (2013). [26] B. Hensen, H. Bernien, A. E. Dr´eau, A. Reiserer, N. Kalb, M. S. Blok, J. Ruitenberg, R. F. L. Vermeulen, R. N. Schouten, C. Abell´an, W. Amaya, V. Pruneri, M. W. Mitchell, M. Markham, D. J. Twitchen, D. Elkouss, S. Wehner, T. H. Taminiau, and R. Hanson, Nature 526, 682 (2015). [27] S. Pezzagna, D. Rogalla, D. Wildanger, J. Meijer, and A. Zaitsev, New J. Phys. 13, 035024 (2011). [28] S. L. Mouradian, T. Schroder, C. B. Poitras, L. Li, J. Goldstein, E. H. Chen, M. Walsh, J. Cardenas, M. L. Markham, D. J. Twitchen, M. Lipson, and D. Englund, Phys. Rev. X 5, 031009 (2015). [29] T. Schroder, S. Mouradian, J. Zheng, M. E. Trusheim, M. Walsh, E. H. Chen, L. Li, I. Bayn, and D. Englund, J. Opt. Soc. Am. B 33, 65 (2016). [30] T. Gaebel, M. Domhan, I. Popa, C. Wittmann, P. Neu- mann, F. Jelezko, J. R. Rabeau, N. Stavrias, A. D. Greentree, S. Prawer, J. Meijer, J. Twamley, P. R. Hem- mer, and J. Wrachtrup, Nat. Phys. 2, 408 (2006). [31] F. Dolde, I. Jakobi, B. Naydenov, N. Zhao, S. Pezzagna, C. Trautmann, J. Meijer, P. Neumann, F. Jelezko, and J. Wrachtrup, Nat. Phys. 9, 139 (2013). [32] T. Yamamoto, C. Muller, L. P. McGuinness, T. Ter- aji, B. Naydenov, S. Onoda, T. Ohshima, J. Wrachtrup, F. Jelezko, and J. Isoya, Phys. Rev. B 88, 201201 (2013). [33] A. Faraon, P. E. Barclay, C. Santori, K.-M. C. Fu, and R. G. Beausoleil, Nat. Photonics 5, 301 (2011). [34] B. J. M. Hausmann, B. J. Shields, Q. Quan, Y. Chu, N. P. De Leon, R. Evans, M. J. Burek, A. S. Zibrov, M. Markham, D. J. Twitchen, H. Park, M. D. Lukin, and M. Loncar, Nano Lett. 13, 5791 (2013). [35] J. C. Lee, D. O. Bracher, S. Cui, K. Ohno, C. A. McLel- lan, X. Zhang, P. Andrich, B. Alem´an, K. J. Russell, A. P. Magyar, I. Aharonovich, A. Bleszynski Jayich, D. Awschalom, and E. L. Hu, Appl. Phys. Lett. 105, 261101 (2014). [36] L. Li, T. Schroder, E. H. Chen, M. Walsh, I. Bayn, J. Goldstein, O. Gaathon, M. E. Trusheim, M. Lu, J. Mower, M. Cotlet, M. L. Markham, D. J. Twitchen, and D. Englund, Nat. Comm. 6, 6173 (2015). [37] J. Riedrich-Moller, S. Pezzagna, J. Meijer, C. Pauly, and F. Mucklich, M. Markham, A. M. Edmonds, C. Becher, Appl. Phys. Lett. 106, 221103 (2015). [38] D. Riedel, I. Sollner, B. J. Shields, S. Starosielec, P. Ap- pel, E. Neu, P. Maletinsky, and R. J. Warburton, Phys. Rev. X 7, 031040 (2017). [20] N. B. Manson, J. P. Harrison, and M. J. Sellars, Phys. [39] M. A. Lea-Wilson, J. N. Lomer, and J. A. Van Wyk, Rev. B 74, 104303 (2006). Philos. Mag. B 72, 81 (1995). [21] K.-M. C. Fu, C. Santori, P. E. Barclay, and R. G. Beau- soleil, Appl. Phys. Lett. 96, 121907 (2010). [22] J. O. Orwa, C. Santori, K. M. C. Fu, B. Gibson, D. Simp- son, I. Aharonovich, A. Stacey, A. Cimmino, P. Balog, M. Markham, D. Twitchen, A. D. Greentree, R. G. Beau- soleil, and S. Prawer, J. Appl. Phys. 109, 083530 (2011). [40] B. Naydenov, F. Reinhard, A. Lammle, V. Richter, R. Kalish, U. F. S. D'Haenens-Johansson, M. Newton, F. Jelezko, and J. Wrachtrup, Appl. Phys. Lett. 97, 242511 (2010). [41] T. Yamamoto, T. Umeda, K. Watanabe, S. Onoda, M. L. Markham, D. J. Twitchen, B. Naydenov, L. P. McGuin- ness, T. Teraji, S. Koizumi, F. Dolde, H. Fedder, J. Hon- ert, J. Wrachtrup, T. Ohshima, F. Jelezko, and J. Isoya, Phys. Rev. B 88, 075206 (2013). [42] P. De´ak, B. Aradi, M. Kaviani, T. Frauenheim, and A. Gali, Phys. Rev. B 89, 075203 (2014). [43] Y. Chu, N. P. de Leon, B. J. Shields, B. Haus- mann, R. Evans, E. Togan, M. J. Burek, M. Markham, A. Stacey, A. S. Zibrov, A. Yacoby, D. J. Twitchen, M. Loncar, H. Park, P. Maletinsky, and M. D. Lukin, Nano Lett. 14, 1982 (2014). [44] J. R. Rabeau, P. Reichart, G. Tamanyan, D. N. Jamieson, S. Prawer, F. Jelezko, T. Gaebel, I. Popa, M. Domhan, and J. Wrachtrup, Appl. Phys. Lett. 88, 023113 (2006). [45] B. K. Ofori-Okai, S. Pezzagna, K. Chang, M. Loretz, R. Schirhagl, Y. Tao, B. A. Moores, K. Groot-Berning, J. Meijer, and C. L. Degen, Phys. Rev. B 86, 081406 (2012). [46] B. Naydenov, V. Richter, J. Beck, M. Steiner, P. Neu- mann, G. Balasubramanian, J. Achard, F. Jelezko, J. Wrachtrup, and R. Kalish, Appl. Phys. Lett. 96, 163108 (2010). [47] D. M. Toyli, C. D. Weis, G. D. Fuchs, T. Schenkel, and D. D. Awschalom, Nano Lett. 10, 3168 (2010). [48] T. Yamamoto, S. Onoda, T. Ohshima, T. Teraji, K. Watanabe, S. Koizumi, T. Umeda, L. P. McGuinness, C. Muller, B. Naydenov, F. Dolde, H. Fedder, J. Hon- ert, M. L. Markham, D. J. Twitchen, J. Wrachtrup, F. Jelezko, and J. Isoya, Phys. Rev. B 90, 081117 (2014). [49] S. Becker, N. Raatz, S. Jankuhn, R. John, and J. Meijer, Sci. Rep. 8, 1 (2018). [50] Y. Chu, private communication. [51] See Supplemental Material [url] for details on sample preparation, measurement methods, NV densities, and additional data, which contains Refs. [52 -- 54]. [52] S. Bogdanovic, M. S. Z. Liddy, S. B. van Dam, L. C. Coenen, T. Fink, M. Loncar, and R. Hanson, APL pho- tonics 2, 126101 (2017). [53] M. S. Blok, C. Bonato, M. L. Markham, D. J. Twitchen, V. V. Dobrovitski, and R. Hanson, Nat. Phys. 10, 189 (2014). [54] S. Pezzagna, B. Naydenov, F. Jelezko, J. Wrachtrup, and 7 J. Meijer, New J. Phys. 12, 065017 (2010). [55] F. F. De Oliveira, D. Antonov, Y. Wang, P. Neumann, S. A. Momenzadeh, T. Haussermann, A. Pasquarelli, A. Denisenko, and J. Wrachtrup, Nat. Commun. 8, 15409 (2017). [56] J. F. Ziegler, M. Ziegler, and J. Biersack, The Stopping and Range of Ions in Matter (SRIM-2013, http://www.srim.org/). [57] G. Davies, S. C. Lawson, A. T. Collins, A. Mainwood, and S. J. Sharp, Phys. Rev. B 46, 13157 (1992). [58] M. W. Doherty, N. B. Manson, P. Delaney, F. Jelezko, J. Wrachtrup, and L. C. Hollenberg, Phys. Rep. 528, 1 (2013). [59] P. Tamarat, N. B. Manson, J. P. Harrison, R. L. Mc- Murtrie, A. Nizovtsev, C. Santori, R. G. Beausoleil, P. Neumann, T. Gaebel, F. Jelezko, P. Hemmer, and J. Wrachtrup, New J. Phys. 10, 045004 (2008). [60] A. Dietrich, K. D. Jahnke, J. M. Binder, T. Teraji, J. Isoya, L. J. Rogers, and F. Jelezko, New J. Phys. 16, 113019 (2014). [61] E. A. Ekimov, S. G. Lyapin, K. N. Boldyrev, M. V. Kon- drin, R. Khmelnitskiy, V. A. Gavva, T. V. Kotereva, and M. N. Popova, JETP Lett. 102, 701 (2015). [62] K. Ohno, F. Joseph Heremans, L. C. Bassett, B. A. My- ers, D. M. Toyli, A. C. Bleszynski Jayich, C. J. Palm- strøm, and D. D. Awschalom, Appl. Phys. Lett. 101, 82413 (2012). [63] K. Ohno, F. J. Heremans, C. F. D. Casas, B. A. Myers, B. J. Alem´an, C. Ania, B. Jayich, and D. D. Awschalom, Appl. Phys. Lett. 105, 052406 (2014). [64] Y. C. Chen, P. S. Salter, S. Knauer, L. Weng, A. C. Frangeskou, C. J. Stephen, S. N. Ishmael, P. R. Dolan, S. Johnson, B. L. Green, G. W. Morley, M. E. Newton, J. G. Rarity, M. J. Booth, and J. M. Smith, Nat. Pho- tonics 11, 77 (2017). [65] Y.-C. Chen, B. Griffiths, L. Weng, S. Nicley, S. N. Ish- mael, Y. N. D. Lekhai, S. Johnson, C. J. Stephen, B. L. Green, G. W. Morley, M. E. Newton, M. J. Booth, P. S. Salter, and J. M. Smith, arXiv:1807.04028 (2018) .
1509.08984	6	1509	2017-07-11T06:35:18	The intrinsic inter-band optical conductivity of a C$_{2v}$ symmetric topological insulator	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	In this work we analytically investigate the longitudinal optical conductivity of the C$_{2v}$ symmetric topological insulator. The conductivity expressions at $ T = 0 $ are derived using the Kubo formula and expressed as a function of the ratio of the Dresselhaus and Rashba parameters that characterize the low-energy Hamiltonian. We find that the longitudinal inter-band conductivity vanishes when Dresselhaus and Rashba parameters are equal in strength, also called the persistent spin helix state. The calculations are extended to obtain the frequency-dependent real and imaginary components of the optical conductivity for the topological Kondo insulator SmB$_{6}$ which exhibits C$_{2v}$ symmetric and anisotropic Dirac cones hosting topological states at the $ \bar{X} $ point on the surface Brillouin zone.	cond-mat.mes-hall	cond-mat	The intrinsic inter-band optical conductivity of a C2v symmetric topological insulator Parijat Sengupta1 Photonics Center Boston University, Boston, MA, 02215. In this work we analytically investigate the longitudinal optical conductivity of the C2v symmetric topological insulator. The conductivity expressions at T = 0 are derived using the Kubo formula and expressed as a function of the ratio of the Dresselhaus and Rashba parameters that characterize the low-energy Hamiltonian. We find that the longitudinal inter-band conductivity vanishes when Dresselhaus and Rashba parameters are equal in strength, also called the persistent spin helix state. The calculations are extended to obtain the frequency-dependent real and imaginary components of the optical conductivity for the topological Kondo insulator SmB6 which exhibits C2v symmetric and anisotropic Dirac cones hosting topological states at X point on the surface Brillouin zone. I. Introduction Spectroscopy techniques serve as an important tool- box to probe the microscopic excitation of matter, the response of the electron ensemble to an external pertur- bation, and ground-state correlation functions that link to specific measurements. Spectroscopy measurements are broadly classified in to categories [1] identified by the study of the spectrum of the microscopic variable within the matter; for example, angle-resolved photo- emission spectroscopy reveals the arrangement of elec- tronic energy surface states while spin provides the basis for nuclear magnetic resonance, vital in determination of the structure of compounds. Similarly, the response of matter to light is routinely gauged using optical spec- troscopy that utilizes the variation of the microscopic current density response to examine the the frequency- dependent electrical conductivity. The optical response of matter, in principle, is conveniently gauged from reflec- tivity data obtained from elipsometry techniques which provide information about the phase and amplitude of the reflected ray. In this regard we note, that governed by Maxwell's laws, the passage (and reflection) of light incident on matter and its eventual coupling to the in- trinsic charge density has been studied for a wide class of materials spanning the entire gamut of classification. However, the emergence of topological states of mat- ter modify Maxwell's laws through the introduction of a non-trivial θ term [2] that profoundly influences the final reflectivity pattern embodied in the optical conduc- tivity behaviour. In this paper, we analytically derive expressions for inter-band optical conductivity arising from transitions between energy levels located on mul- tiple bands in a topological insulator (TI) whose surface states are marked by the C2v symmetry. In principle, the surface dispersion of a topological in- sulator in its simplest form can be described by a Rashba- like linear Hamiltonian; however, the loss of intrinsic bulk symmetry (a crystal attribute characterized by the bulk inversion asymmetry parameter) in crystals necessitates the inclusion of an additional Dresselhaus-like Hamil- tonian term. [3, 4] While the Dresselhaus contribution is usually a small effect and ignored in most calcula- tions, we show that the overall ratio of the Rashba- and Dresselhaus-coupling coefficients can indeed have a sig- nificant role in modulating the electronic spectrum with important implications for light-matter interaction on the topological insulator surface. Recent developments in the fabrication of spin-based devices show that this ratio can be easily altered [5] by adjusting the Rashba coefficient which is a direct indicator of the structural inversion asymmetry (SIA). In fact, it has been experi- mentally confirmed that the Rashba parameter can be adjusted through an external gate bias [6, 7] to val- ues as large as 2× 10−11eV m in InAs-based heterostruc- tures [8]. Remarkably, for equal strength of the Rashba and Dresselhaus coupling coefficients (the persistent spin helix state), the inter-band optical conductivity vanishes. In this work, we utilize the Kubo formalism from lin- ear response theory [9] to establish a functional depen- dence between the ratio (κ) of the Dresselhaus (αD) and Rashba (αR) spin coupling coefficients (κ = αD/αR) and the longitudinal static components of charge and spin Hall conductivity. The examination of the conductivity in these systems yields useful information about the rel- ative strength of the Rashba and Dresselhaus spin-orbit coupling coefficients; conversely, their relative strength provides key insight to the character of charge and spin conductance. This also allows us to note that while the sign of inter-band optical conductivity does not change as κ varies (the flow of charge to an external field is not reg- ulated by spin), the overall spin texture and the sign of spin Hall conductivity (SHC) can be adjusted. Precisely, as κ takes on values greater than unity, the sign of SHC switches, a result also borne out by an identical transfor- mation of the ± (2n + 1)) π Berry phase of such a system. We further extend these calculations to obtain in the long wave length limit, the dynamic (frequency-dependent) real and imaginary parts of optical conductivity for the topological Kondo insulator [10, 11] SmB6. The topo- logical surface states of SmB6 possess anisotropic Dirac cones [12] at the X point and characterized by C2v and time reversal symmetry. All calculations are performed at T = 0 K. 7 1 0 2 l u J 1 1 ] l l a h - s e m . t a m - d n o c [ 6 v 4 8 9 8 0 . 9 0 5 1 : v i X r a II. Model Hamiltonians To describe the electronic structure of the material, we employ a two-band k.p model that describes surface states in proximity of the Dirac cone. In the low energy region, the linear Hamiltonian that describes the surface states is given by HT I = αR (σykx − σxky)+αD (σxkx − σyky)+∆σz, (1) where ∆ is the symmetry-breaking potential that in- duces a finite gap between the surface bands while αR and αD are the Rashba- and Dresselhaus-like parameters, The dispersion relation- ship obtained by diagonalization of the Hamiltonian in Eq. 1 is ε (k) = (cid:112) β 2 k2 + ∆2, where β2 = (cid:0)α2 D + 2αR αD sin 2φ(cid:1). For later use, we also write (cid:18)λ± exp (iθ) (cid:19) the analytic expressions for the 2 × 1 wave functions corresponding to the Hamiltonian in Eq. 1 respectively. R + α2 (cid:115) ; λ± = 1 ± ∆(cid:112)∆2 + β2 Ψ± = 1√ 2 ± λ∓ . (2) The polar angle in Eq. 2 is θ = tan−1 kx + κ ky ky + κ kx = tan−1 cos φ + κ sin φ sin φ + κ cos φ . (3) We have introduced the additional notation κ = αD/αR as the ratio of the Dresselhaus and Rashba coupling coefficients and set kx = k cosφ and ky = k sinφ in writing Eq. 3. We also derive the corresponding veloc- ity components vx,y by evaluating the standard expres- sion vi = (1/i) [r,H]. The velocity components along x - and y-axes in operator notation are therefore vx = (1/) (αD σx − αR σy) and vy = (1/) (αR σx − αD σy), re- spectively. Finally, note that the k.p Hamiltonian written as an expansion of the states around the Γ point is only accurate in low-energy regions in its vicinity. We can now using the model Hamiltonian (around the Γ point) and the corresponding wave functions derive the longitu- dinal intra- and inter-band conductivity for TIs with C2v symmetry as a function of the Rashba and Dresselhaus coefficients. III. Intrinsic inter-band conductivity The conductivity calculations are carried out by a di- rect application of the Kubo formalism within the linear response theory. For a non-interacting sample, Kubo ex- pression for conductivity is written as σx,y = −i e2 L2 f (εn) − f (εn εn − εn (cid:48) (cid:48) ) (cid:104) n vα n (cid:48)(cid:105)(cid:104) n εn − εn (cid:48) vα n(cid:105) (cid:48) + i η , (cid:88) (cid:48) n,n asymmetry described by Eq. 1; these calculations with suitable amendment to the defining Hamiltonian are re- peated for the topological Kondo insulator SmB6. In de- riving these expressions, we tacitly assume that the wave functions in presence of impurities and small external perturbations retain their original form given in Eq. 2 and the topological insulator sample area is A = L2. The inter-band conductivity expression (the intra-band expression is derived in Appendix A) can be compactly written as xx = −i σinter e2 L2 Minter2 f (εc) − f (εv) (εc − εv) (εc − εv + iη) (cid:20) (cid:88) (cid:48) n,n 2 (cid:21) . (5) Note that the matrix element Minter in Eq. 5 for the inter-band case and a zero band gap (in Eq. 2, ∆ = 0) is defined as Minter = (cid:104)Ψ± vx Ψ∓(cid:105) = i (αD sin θ − αR cos θ) , (6) where the valence and conduction band wave functions are Ψ−(cid:105) and Ψ+(cid:105), respectively and vx is the velocity op- erator along x -axis. For a zero-temperature case (T = 0), above the Fermi level, the conduction states are devoid of carriers while below the valence band is completely filled. The Fermi distribution functions f (εc) and f (εv) are therefore Heaviside step functions. We therefore set f (εc) and f (εv) to zero and unity, respectively. With this in mind, expanding and changing the sum in Eq. 5 to an integral, the final conductivity expression leads to: (cid:90) ∞ (cid:90) 2π (cid:20) Θ (εf − εc) − Θ (εf − εv) xx = −i σinter e2 4π2 k dk kf 0 (2ε) (2ε + iη) dφ (αDf2 (φ) − αRf1 (φ))2 × Θ (εf − εv) − Θ (εf − εc) (cid:21) (−2ε) (−2ε + iη) . (7) + The functions f1 (φ) and f2 (φ) are cos (θ) and sin (θ), respectively expressed in terms of φ using Eq. 3. While writing Eq. 7, for q → 0 (in the long wavelength limit), we have set εc = −εv = ε and Θ (·) represents the Heaviside step function. Simplifying the integrals in Eq. 7, σinter normalized to e2/ can be written as xx (cid:90) 2π (cid:90) ∞ η 2π 0 σinter xx = × 1 4β2k2 + η2 kdk. kf (αDf2 (φ) − αRf1 (φ))2 β2 dφ (8) (4) (cid:48)(cid:105) are eigen functions of the defining where n(cid:105) and n Hamiltonian and η represents a finite broadening of the eigen-states resulting from surface imperfections and em- bedded impurities. We first take up the generic topo- logical insulator with C2v symmetry and bulk inversion Note that to change the variable of integration from k → , we use the dispersion relation = β k. The Fermi energy is εf and kf is the corresponding wave vector. The integral in Eq. 8 can be numerically evaluated to obtain σxx. When expressed in terms of ratio of the Dresselhaus and Rashba coefficients, Eq. 8 is recast as (cid:18) (cid:20) (cid:19) (cid:0)κ2 − 1(cid:1) sinφ π/2 − tan−1 2εf η (κ2 + 2κ sin2φ + 1) σinter xx = × 1 8π2 (cid:90) 2π 0 3 (cid:21)2 dφ. (9) We immediately observe from Eq. 9 that for κ = 1, which defines a system with identical magnitude for the Rashba and Dresselhaus coefficients - a condition known as the persistent spin helix (PSH) state with SU (2) symmetry occurs [13, 14]- the longitudinal static inter-band con- ductivity vanishes. The disappearance of the static inter- band conductivity can be simply explained by noting that the matrix element in the Kubo expression ceases to ex- ist. Li et al. obtained an identical result in Ref. 15 The inter-band optical conductivity, following a numerical in- tegration of Eq. 9, is shown in Fig. 1 for two values of η, the broadening parameter. (cid:0)kxk2 (cid:1) which leads Finally, we wish to point that the Hamiltonian in Eq. 1 is purely linear and, as a result, for the PSH state we obtained a vanishing longitudinal inter-band conductiv- ity. However, there is always a cubic Dresselhaus con- y σx − kyk2 (cid:48) x σy tribution of the form α D (cid:48) D is the third to a finite conductivity. In this case, α order Dresselhaus coefficient and is the SU (2) violating term. The vanishing inter-band optical conductivity also manifests as a zero inter-band absorption of light in the PSH state; a result which was also derived by the au- thors through an explicit calculation of the inter-band matrix element (the inter-band matrix element is zero in the PSH state) in connection to examining the circular dichroism (η) pattern in TIs with C2v symmetry. Circu- lar dichroism, which is the differential absorption of left- and right-circularly polarized light, for such a case has been derived in Ref. 16. A note about dc conductivity is in order here: We have tacitly assumed T = 0K; how- ever, the case of a finite temperature can also be easily handled by rewriting the Kubo expression in terms of the Matsubara Greens function (or the imaginary time for- malism). The conductivity expression derived in Eq. 9 is now a function of frequency and momentum σ (q, ω), with q → 0 in the long wavelength limit. The dc conductivity can be extracted from the general frequency-dependent expression by letting ω → 0. Further, notice from Eq. 9 that the inter-band con- ductivity expression is always positive regardless of the relative strength of κ = αD/αR, an observation easily reconcilable since charge conductivity does not depend on the orientation of the spin polarization brought about by the spin-orbit coupling Hamiltonian terms in Eq. 1. However, there are quantities of interest, for instance, the Berry phase [17] and the spin Hall conductivity [18] that do exhibit a direct dependence on the strength of κ. We examine the Berry phase in the following sub-section. FIG. 1. The numerically calculated static inter-band optical conductivity for a range of κ = αD/αR and a pair of transit times (τ ). At κ = 1, the inter-band optical conductivity van- ishes. The static inter-band conductivity diminishes as the strength of the broadening parameter η = /τ is reduced (for an increase in τ ) eventually ceasing to exist for an infinite transit time. The Fermi level for this calculation was set to 5.0 meV . A. The Berry phase k - parameter space is defined as γ = (cid:72) dk · The Berry phase is closely linked to the electron trans- port coefficients [19]. We evaluate the Berry phase −→ around the Γ point. The Berry phase [20] in the closed (cid:104)ψ±, θ i ∂ ∂ k Eq. 2 in the Berry phase (γ) expression and evaluating ∂θ ∂kν where ν = {x, y}, one obtains Inserting the wave function from ψ±, θ(cid:105). (cid:90) 2π γ = κ2 + 2κ sin2φ + 1 0 κ2 − 1 dφ = κ2 − 1 κ2 − 1 π, (10) where once more κ = αD/αR. Let us now consider the two cases: 1) κ < 1 for which we have the Berry phase as −π using Eq. 10 while for 2) κ > 1, one obtains γ = π. The switching of κ between the two aforesaid intervals, as we remarked before, does not change the optical conductivity; the significance of it, however, lies in its manifestation in the spin Hall conductivity. The zero-frequency spin Hall conductivity within the Kubo formalism is e L2 (cid:48) vy n(cid:105) (cid:48) + iξ f (εn) − f (εn σSH = −i α n (cid:48)(cid:105)(cid:104) n εn − εn εn − εn (cid:88) , (cid:104) n jz (cid:48) ) (cid:48) (cid:48) n,n (11) (cid:48)(cid:105) are eigen functions of the Hamilto- where n(cid:105) and n nian given in Eq. 1. The Kubo expression in Eq. 11 4 {vx, σz}. when evaluated for α = x yields the spin Hall conductiv- ity. Note that the spin current operator [21] is defined x = as jz In this definition of the spin cur- rent, the electron velocity is directed along the y-axis due to an aligned external electric field and an out-of- plane z−polarized spin current flows along the perpen- dicular x -axis. The spin Hall effect (SHE) refers to a transverse spin current induced by an external electric field in absence of a magnetic field. In this case, the SHE is produced by the intrinsic spin-orbit coupling as opposed to the extrinsic SHE driven by spin-orbit scat- tering impurities. The SHE leading to accumulation of spin in a preferred direction, as is easy to understand, depends on the strength of the two contributions to the Hamiltonian (Eq. 1) which transform the spin compo- nents differently. It is therefore reasonable to believe that the path of spin accumulation in an intrinsic SHE setup guided by the spin-orbit coupling can be formulated in terms of the Berry phase. In the Kubo expression for spin conductivity (Eq. 11), inserting the desired eigen values and eigen functions and following exactly the same set of steps carried out for optical conductivity, we arrive at the following expression for the spin Hall conductivity3 σz xy = 1 − κ2 1 − κ2 . e 8π (12) α = (cid:0)2kα/2m(cid:1) σz, where α = {x, y}. The xy = ±(cid:0)e/8π2(cid:1) γ. As the In deriving the above expression, we added a particle- hole asymmetric quadratic term p2/2m to the Hamil- tonian, following which the spin current operator eval- uates to jz connection between the geometric Berry phase and the spin Hall and diagonal conductivity can be easily seen by rewriting Eq. 12 as σz contribution of the two terms in the Hamiltonian re- flected in the ratio κ toggles between the two intervals, κ < 1 and κ > 1, the spin Hall conductivity switches sign identically to the Berry phase. Furthermore, for the two limiting cases, αR = 0, when the system pos- sesses structural inversion symmetry and a finite bulk inversion asymmetry, (αD (cid:54)= 0), and rewriting Eq. 12 as R, yields the universal σz spin Hall conductance as e/8 π. Conversely, for αR (cid:54)= 0 and αD = 0, the spin Hall conductance retains the same magnitude but switches sign. This result connecting the Berry phase to spin Hall conductivity was first obtained by S. Shen in Ref. 22. xy = e/8 π(cid:0)α2 (cid:1) /α2 D − α2 D − α2 R We again underscore the case of κ = 1 which is the condition for PSH and recognize that spins are aligned parallel [23] and there is no "effective" spin-orbit coupling that bends the trajectory. However, as we stated before, 3The spin current operator definition does not hold along the x and y-axes since the in-plane spin vectors are mixed as evident from Eq. 1. The spin Hall conductivity in this case must be computed by evaluating the quantity σς,↑ xy, where ς = x, y. xy − σς↓ 4 FIG. 2. The bulk unit cell for SmB6 which has a CsCl-type crystal structure. The samarium (red) and boron (green) atoms arranged at the vertices of an octahedron are located at the corner and centre of a cubic lattice respectively. The structure visualization was done with the VESTA [28] soft- ware. in the case of vanishing inter-band conductivity, non-zero higher-order Dresselhaus terms could lead to an inexact cancellation of the Rashba and Dresselhaus linear spin- orbit Hamiltonians. IV. Application to SmB6 A noteworthy instance of a material whose surface states have non-trivial topology with C2v symmetry is the topological Kondo insulator SmB6 (see Fig. 2 for the unit cell structure). Briefly, Kondo insulators which are marked by resistivity that has a minimum at a low temperature but increases as the temperature is lowered are highly electron-correlated systems that can exhibit the Z2 topological insulator behaviour. Experimental demonstrations of Kondo insulators with topologically non-trivial states have been carried out [24] with SmB6 confirming their robust spin-polarized [25, 26] surface states. SmB6, however, unlike other Kondo insulators has a resistivity which has a minimum at room temper- ature (T ) and a plateau-like profile [27] for (T (cid:54) 5K). The low temperature constant resistivity is attributed to topologically protected surface states within the Kondo band gap (approximately 17.7 meV) and form three Fermi surfaces. Of these three Fermi surfaces, angle- resolved photoemission spectroscopy (ARPES) reveals that they are centred at the Γ and doubly at the X points of the surface Brillouin zone. The genesis of the topologically protected surface states at Γ and X lies in the band inversion [29] that occurs between the 4f and 5d orbitals of samarium at X points (see Fig. 3) in the bulk Brillouin zone (BZ). These inverted X points in the bulk BZ when projected on the surface BZ of a [001] grown SmB6 crystal gives rise to the Dirac-like surface states (see Fig. 4) with a helical spin texture, the defining hallmark of topological insulators. A similar set of calculations was reported in Ref. 30 where 5 Hamiltonian is then adapted for a slab by carrying out the standard transformation ki = −∂/∂ki; for our case, the dispersion of the slab around the Γ and X are ob- tained by making the simple replacement kz = −i∂/∂z and kx = −i∂/∂x in the k.p Hamiltonian. Note that the set of k-vectors for each case, (kx, ky), around the Γ and (ky, kz) for X continue to be good quantum numbers. Discretizing the derivative operator on a finite difference grid, we arrive at an effective slab Hamiltonian which can be diagonalized to obtain the dispersion as shown in Fig. 5. For numerical details about discretization and other steps to construct the slab Hamiltonian, the reader is referred to Ref. 36. The anisotropic character of the Dirac cones around the X point can be further reduced to an effective mini- mal k.p surface Hamiltonian using symmetry arguments derived in Ref. 35. The band parameters for this model are obtained through a fitting procedure by a direct com- parison with first principles calculation (Fig. 4). The Hamiltonian at X has a C2v symmetry which we repro- duce in Eq. 13 and use as a starting point for further conductivity calculations with the topological Kondo in- sulator SmB6. H SS X = 1 + a0k2 + [i (a1k+ + a2k−) σ+ + h.c] . (13) In Eq. 13, the constants [35] (in units of alc ∗ eV A) a0, a1, a2 are 0.011276, 0.003059, and -0.02322, respec- tively. The other terms are defined as k± = kx ± iky, σ± = σx ± iσy, and k2 = k2 y. The Pauli spin ma- trices are σx, σy, and σz. Further, for low energy states the quadratic term in Eq. 13 can be dropped to yield a linear Dirac-like equation. Expanding, we obtain X = (a1 + a2) kxσx − (a1 − a2) kyσy. H SS x + k2 (14) Note that this represents an anisotropic (tilted) Dirac crossing at the X point with unequal and x - and y-axes directed Fermi velocities. To arrive at optical conductivity expressions for SmB6 (Section IV A), we begin (using the Hamiltonian in Eq. 14) by writing the wave functions for energy states in the vicinity of the anisotropic Dirac crossing. The conduction and valence state wave functions are (cid:18)χ± exp (iθ) (cid:19) ±χ∓ , (15a) where χ± is Ψ± = 1√ 2 χ± = ∆ (cid:113) (cid:118)(cid:117)(cid:117)(cid:116)1 ± ∆2 +(cid:0)Ak2 ε± = ±(cid:113) ∆2 +(cid:0)A2k2 (cid:1) . (cid:1). (15b) x + Bk2 y The conduction (+) and valence (-) eigen states are de- fined as x + B2k2 y (16) The conduction state wave function Ψc = Ψ+ and the corresponding valence state wave function is identified FIG. 3. The bulk band structure of SmB6 with CsCl-type crystal structure calculated from first-principles using the VASP software. The inversion in the bulk band structure happens at the X point which upon projection to the surface manifests as three topologically protected states. Tay-Rong Chang et al. computed the electronic structure of SmB6 using the GGA and GGA + U schemes; remark- ably, they observed little change in the overall character of bands as the electrostatic Coulomb energy U was in- cremented from zero to a large value of 8.0 eV . They con- cluded by noting this functional non-dependence on the Coulomb energy as a sufficient proof of the band topol- ogy and the Kondo insulator attribute of SmB6. The bulk and slab band structures were obtained using the VASP code [31, 32] within Perdew-Burke-Ernzerhof [33] exchange-correlation functionals. The spin orbit coupling was self-consistently included in our calculation. Ad- ditionally, the plane wave energy cutoff value was set to 320 eV and the Brillouin zone was sampled with a 12 × 12 × 12 Γ centered k-point mesh. We set the lat- tice constant [34] to alc = 4.1327A and the number of electrons with up and down spins were identical through- out the calculation. Note that the dispersion of the slab in Fig. 4 uses three colors to indicate the origin of the bands: The red colored bands arise because of a domi- nant surface contribution while the black bands denote the bulk dispersion. The green bands are an admixture of surface and bulk dispersion. It is, however, pertinent to remember that only those surface bands that connect the conduction and valence bands (a closing of the band gap) are topological in nature; in this case they occur at the X and Γ points of the surface Brillouin zone and marked by blue rectangular contours. To obtain the surface con- tribution, we selected atoms that lie at the interface of the slab and vacuum. For greater clarity, and to clearly identify the topo- logical surface bands from the multiple states in a first- principles calculation, a simplified eight-band k.p Hamil- tonian that selectively describes the dispersion around the X point (derived from the theory of invariants in Ref. 35) is of utility in our context here. The bulk 6 frequency-dependent Kubo expression for the inter-band conductivity is e2 L2 (cid:104) n vα n (cid:48)(cid:105)(cid:104) n ω + εn − εn f (εn) − f (εn (cid:48) vβ n(cid:105) (cid:48) + iη αβ = −i σinter εn − εn (cid:88) (cid:48) ) (cid:48) (cid:48) n,n , (17) where the symbols have meaning identical to Eq. 4. For a less cumbersome notation, the superscript 'inter' will be dropped from now. Further, the ket vector n(cid:105)(cid:16) n (cid:48)(cid:105)(cid:17) denotes the conduction (valence) state wave function Ψc(cid:105) ( Ψv(cid:105)). The corresponding conduction (valence) eigen state is εn=c (εn =v). For longitudinal optical con- ductivity along the x -axis, the velocity operators vα and vβ are identical and equal to vx = (A/) σx. Inserting the velocity operator in the Kubo expression, the matrix elements can be straightforwardly computed (cid:48) (cid:104) Ψc vx Ψv(cid:105) = − A (i sin θ + γ cos θ) , (18) (cid:113) A2k2 x + B2k2 y. The Fermi distribution where γ = ∆/ functions in Eq. 17, as before, are set under the tacit as- sumption that the Fermi energy is positioned at the top of the valence band; this effectively ensures that f (εc) = 0 and f (εv) = 1. Putting all of them together and expand- ing the real and imaginary part of the conductivity (in units of e2/) yields (cid:20)(cid:90) kc (cid:90) 2π 0 (cid:0)sin2 θ + γ2 cos2 θ(cid:1) Ω (cid:90) 2π (cid:0)sin2 θ + γ2 cos2 θ(cid:1) Ω (ω − 2ε)2 + η2 (cid:21) dθ 0 dk + (ω + 2ε)2 + η2 , (19) σR xx = η (cid:90) kc 0 where Ω ≈ A2 8π2 dk (cid:112) dθ 0 1 A2 cos2 θ + B2 sin2 θ and the upper limit of the k -space integral, kc, corresponds to the momentum vector for a given energy cut-off. The approximation to Ω is reasonably accurate for small values of ∆, the band gap opening. We set kc = 0.12 A−1 for a numerical eval- uation of all conductivity expressions in this work. The imaginary part of the longitudinal optical conductivity is likewise, (cid:0)sin2 θ + γ2 cos2 θ(cid:1) Ω (cid:90) 2π (cid:0)sin2 θ + γ2 cos2 θ(cid:1) Ω (ω − 2ε)2 + η2 (cid:21) dθ 0 (cid:48) + dk (cid:48) − + dk dθ (ω + 2ε)2 + η2 , (20) 0 (cid:20)(cid:90) kc (cid:90) 2π (cid:112) 0 σIm xx = A2 8π2 (cid:90) kc 0 where Ω (cid:48) ± ≈ ω ± 2ε A2 cos2 θ + B2 sin2 θ . The anisotropy (tilted Dirac cone) at the X point sug- gests that the y-axis directed longitudinal conductivity (σyy) is unequal to σxx. Retracing the set of steps in the calculation of σxx, we write down the result (in units of FIG. 4. The first-principles (with VASP) calculated disper- sion of an SmB6 slab of thickness 6.612 nm is plotted with an inverted bulk character at X. The inversion at bulk X is ≈ 17.7 meV (see inset of Fig. 3). The two topological surface states for the slab structure at Γ and X are boxed in blue. The slab configuration used in the VASP software package had 17 samarium atoms and 16 boron layers. The significance of the different colors on the plot is explained in the text. as Ψv = Ψ−. For brevity, in Eq. 15b, A = a1 + a2 and B = a1−a2. Note that for the sake of completeness a gap opening term of the form ∆σz appears in the Hamiltonian (Eq. 14). A. Conductivity of the X point The conductivity calculations again begin from the Kubo expression in Eq. 4; however, to obtain a frequency dependence of the inter-band conductivity we must re- work some of the expressions derived heretofore. The FIG. 5. The 8-band k.p calculated dispersion of an SmB6 slab of thickness 7.0 nm is plotted in the vicinity of Γ and the X point (right panel). The red boxed area in the right panel is the region of interest and we compute the inter-band conduc- tivity by evaluation of the momentum matrix element (and their insertion in the Kubo expression) between states lying on the surface conduction and valence bands. The anisotropy of the Dirac cone around the X point is clearly noticeable in contrast to the the isotropic Dirac cone centred at Γ. The Fermi level is aligned to top of the surface valence band. e2/) for σyy. Note that the appropriate matrix element in the Kubo expression for σyy is 7 (cid:104) Ψc vy Ψv(cid:105) = − iB (cos θ + iγ sin θ) , The real part of σyy is therefore while the corresponding imaginary part is (21) + (cid:20)(cid:90) kc (cid:90) 2π 0 (cid:20)(cid:90) kc (cid:90) 2π 0 (cid:0)cos2 θ + γ2 sin2 θ(cid:1) Ω (ω − 2ε)2 + η2 (cid:21) dk 0 dθ (cid:90) 2π (cid:0)cos2 θ + γ2 sin2 θ(cid:1) Ω (cid:90) 2π (cid:0)cos2 θ + γ2 sin2 θ(cid:1) Ω (ω + 2ε)2 + η2 dθ 0 (ω + 2ε)2 + η2 dk (cid:0)cos2 θ + γ2 sin2 θ(cid:1) Ω (ω − 2ε)2 + η2 (cid:48) − + (cid:21) (cid:48) + . (23) , (22) B2 8π2 σR yy = η (cid:90) kc dk dθ 0 0 σIm yy = B2 8π2 (cid:90) kc dk dθ 0 0 For all numerical calculations that are presented, we set the γ term in the conductivity expressions to zero indi- cating a vanishing band gap at the X point on the sur- face BZ of SmB6. The real and imaginary parts of the anisotropic in-plane longitudinal conductivity is plotted in Fig. 6. We have chosen the energy scale for the con- ductivity plot to in the range of topological surface states for the slab structure shown in Fig. 4. The inequality of the conductivity components along the x - and y-axes is clearly visible for the chosen band parameters in the defining Hamiltonian (Eq. 13. Note that the coefficients a1 + a2 and a1 − a2 represent the Fermi velocities along the x- and y-axes, respectively. FIG. 6. The numerically calculated longitudinal optical con- ductivity of SmB6 along the x - and y-axes. The transit time (τ ) for surface electrons around the X point on the surface Brillouin zone was taken to be 0.1 ns. The η in the conduc- tivity equations above is given by η = /τ . The anisotropy ratio manifested in the unequal surface Fermi velocities is vX f /vY f = 0.7672. chiral surface plasmon polaritons on the 2D anisotropic surface. V. Summary Acknowledgments In conclusion, we have employed the Kubo formalism from linear response theory to compute the inter-band, and spin conductivity point for a topological insulator with C2v and time reversal symmetry. The Hamiltonian for such a TI to a first order has contributions from the Rashba- and Dresselhaus-like spin-orbit terms. We first show that the longitudinal inter-band conductivities van- ishes when the Rashba and Dresselhaus components are of equal strength (the PSH state). At PSH, the zero inter-band longitudinal conductivity is significant since it correlates to a vanishing inter-band light absorption. We also calculated the dynamic longitudinal conductiv- ity of the C2v symmetric surface states of the topological Kondo insulator SmB6. The surface states located at the X point of the Brillouin zone host anisotropic Dirac cones; the anisotropy distinguished by unequal Fermi ve- locities along the x - and y-axes varies as we progres- sively move away from the crossing [37]. This asymme- try could be potentially modulated through embedded impurity dopants or inducing strain through a substrate that leads to a controllable optical conductivity with sig- nificant implications for production and transmission of This work at Boston University was supported in part by the BU Photonics Center and U.S. Army Research Laboratory through the Collaborative Research Alliance. 1P. Coleman, Introduction to many-body physics (Cambridge Uni- versity Press, 2015). 2X.-L. Qi, T. L. Hughes, and S.-C. Zhang, Physical Review B 78, 195424 (2008). 3R. Winkler, Spin-orbit coupling effects in two-dimensional elec- tron and hole systems (Springer, 2003). 4I. Zuti´c, J. Fabian, and S. D. Sarma, Reviews of modern physics 76, 323 (2004). 5H. Yuan, M. S. Bahramy, K. Morimoto, S. Wu, K. Nomura, B.-J. Yang, H. Shimotani, R. Suzuki, M. Toh, C. Kloc, et al., Nature Physics 9, 563 (2013). 6D. Khomitsky, Physical Review B 79, 205401 (2009). 7A. Larionov and L. Golub, Physical Review B 78, 033302 (2008). 8D. Grundler, Physical review letters 84, 6074 (2000). 9R. A. Jishi, Feynman diagram techniques in condensed matter physics (Cambridge University Press, 2013). 10M. Dzero, K. Sun, V. Galitski, and P. Coleman, Physical review letters 104, 106408 (2010). 11M. Neupane et al., Nature communications 4 (2013). 12R. Yu et al., New Journal of Physics 17, 023012 (2015). 13B. A. Bernevig, J. Orenstein, and S.-C. Zhang, Physical review letters 97, 236601 (2006). 14J. D. Koralek et al., Nature 458, 610 (2009). 15Z. Li, F. Marsiglio, and J. Carbotte, Scientific reports 3 (2013). 16P. Sengupta and E. Bellotti, Applied Physics Letters 108, 031101 (2016). 17P. Sengupta and G. Klimeck, Semiconductor Science and Tech- nology 30, 045004 (2015). 18P. Sengupta and E. Bellotti, Applied Physics Letters 108, 211104 (2016). 19G. Sundaram and Q. Niu, Physical Review B 59, 14915 (1999). 20G. Grosso and P. Giuseppe, Solid state physics (Elsevier, 2014). 21S. Murakami, in Advances in Solid State Physics (Springer, 2006), pp. 197 -- 209. 22S.-Q. Shen, Physical Review B 70, 081311 (2004). 23J. Schliemann, J. C. Egues, and D. Loss, Physical review letters 90, 146801 (2003). 24S. Wolgast, C¸ . Kurdak, K. Sun, J. Allen, D.-J. Kim, and Z. Fisk, Physical Review B 88, 180405 (2013). 25D. Kim, S. Thomas, T. Grant, J. Botimer, Z. Fisk, and J. Xia, Scientific reports 3 (2013). 26N. Xu, P. Biswas, J. Dil, R. Dhaka, G. Landolt, S. Muff, C. Matt, X. Shi, N. Plumb, M. Radovi´c, et al., Nature communications 5 (2014). 27J. Allen, B. Batlogg, and P. Wachter, Physical Review B 20, 4807 (1979). 28K. Momma and F. Izumi, Journal of Applied Crystallography 41, 653 (2008). 29F. Lu, J. Zhao, H. Weng, Z. Fang, and X. Dai, Physical review letters 110, 096401 (2013). 30T.-R. Chang, T. Das, P.-J. Chen, M. Neupane, S.-Y. Xu, M. Z. Hasan, H. Lin, H.-T. Jeng, and A. Bansil, Physical Review B 91, 155151 (2015). 31G. Kresse and J. Furthmuller, Physical review B 54, 11169 (1996). 32G. Kresse and D. Joubert, Physical Review B 59, 1758 (1999). 33J. P. Perdew, K. Burke, and M. Ernzerhof, Physical review letters 77, 3865 (1996). 34S. Funahashi, K. Tanaka, and F. Iga, Acta Crystallographica Section B: Structural Science 66, 292 (2010). 35R. Yu, H. Weng, X. Hu, Z. Fang, and X. Dai, New Journal of Physics 17, 023012 (2015). 36P. Sengupta, H. Ryu, S. Lee, Y. Tan, and G. Klimeck, Journal of Computational Electronics 15, 115 (2016). 37B. Roy, J. D. Sau, M. Dzero, and V. Galitski, Physical Review B 90, 155314 (2014). 38P. Sengupta, T. Kubis, Y. Tan, M. Povolotskyi, and G. Klimeck, Journal of Applied Physics 114, 043702 (2013). 39F. Kirtschig, J. v. d. Brink, and C. Ortix, arXiv preprint arXiv:1605.00293 (2016). Appendix A Intra-band optical conductivity 8 To evaluate the intra-band conductivity, we assume (cid:48) are close in mag- that the two energy states εn and εn nitude allowing us to Taylor expand the difference be- tween the Fermi functions in Eq. 4. One then obtains, f (εn) − f (εn . Evaluating the intra-band matrix element Mintra = (cid:104)Ψ± vx Ψ±(cid:105) in Eq. 4 we arrive at the following expression for σintra : (cid:48) ) = (εn − εn ∂ f (εn) ∂ εn (cid:48) ) xx σintra xx = i 1 A e2 ∂ f (εn) ∂ εn (εn − εn (cid:48) ) (εn − εn (cid:48) ) (cid:88) n (cid:90) 2π × (λ+ λ−)2 (αDcos θ + αRsin θ)2 Using the dispersion relation ε (k) = (cid:112) β 2 k2 + ∆2 to εn − ε (cid:48) n + iη (A1a) . make the change from k to E gives σintra xx = × 1 e2 (cid:90) εf 4π2η ε2 − ∆2 0 0 ε (αDf1 (φ) + αRf2 (φ))2 β2 δ (εf − ε) dε, dφ (A1b) where f1 (φ) and f2 (φ) are cos (θ) and sin (θ), respec- tively expressed in terms of φ using Eq. 3. Further, in Eq. A1a, we have set − ∂ f (εn) = δ (εn − ) at T = 0 to ∂ εn rewrite it in the form shown in Eq. A1b. The intra-band conductivity normalized to e2/ is (cid:90) 2π (cid:21)2 (cid:20)(cid:0)κ2 + 1(cid:1) cos φ + 2κ sin φ (cid:16) f − ∆2(cid:17) (κ2 + 2κ sin2φ + 1) ε2 /εf , and αR (cid:54)= 0. dφ, (A1c) σintra xx = ζ 4π2 η 0 where κ = αD/αR, ζ =
1812.06702	2	1812	2018-12-18T08:44:02	Domain wall-based spin-Hall nano-oscillators	[ "cond-mat.mes-hall" ]	In the last decade, two revolutionary concepts in nano magnetism emerged from research for storage technologies and advanced information processing. The first suggests the use of magnetic domain walls (DWs) in ferromagnetic nanowires to permanently store information in DW racetrack memories. The second proposes a hardware realisation of neuromorphic computing in nanomagnets using nonlinear magnetic oscillations in the GHz range. Both ideas originate from the transfer of angular momentum from conduction electrons to localised spins in ferromagnets, either to push data encoded in DWs along nanowires or to sustain magnetic oscillations in artificial neurones. Even though both concepts share a common ground, they live on very different time scales which rendered them incompatible so far. Here, we bridge both ideas by demonstrating the excitation of magnetic auto-oscillations inside nano-scale DWs using pure spin currents.	cond-mat.mes-hall	cond-mat	Domain wall-based spin-Hall nano-oscillators N. Sato,1 K. Schultheiss,1 L. Korber,1, 2 N. Puwenberg,3 T. Muhl,3 A.A. Awad,4 S.S.P.K. Arekapudi,5 O. Hellwig,1, 5 J. Fassbender,1, 2 and H. Schultheiss1, 2 1)Helmholtz-Zentrum Dresden -- Rossendorf, Institut fur Ionenstrahlphysik und Materialforschung, D-01328 Dresden, Germany 2)Technische Universitat Dresden, 01062 Dresden, Germany 3)Leibniz Institute for Solid State and Materials Research (IFW) Dresden, 01069 Dresden, Germany 4)Department of Physics, University of Gothenburg, 412 96 Gothenburg, Sweden 5)Institut fur Physik, Technische Universitat Chemnitz, D-09107 Chemnitz (Dated: 19 December 2018) In the last decade, two revolutionary concepts in nano magnetism emerged from research for storage technologies and advanced information processing. The first suggests the use of magnetic domain walls (DWs) in ferromagnetic nanowires to permanently store information in DW race- track memories1. The second proposes a hard- ware realisation of neuromorphic computing in nanomagnets using nonlinear magnetic oscilla- tions in the GHz range2,3. Both ideas origi- nate from the transfer of angular momentum from conduction electrons to localised spins in ferromagnets4,5, either to push data encoded in DWs along nanowires or to sustain magnetic os- cillations in artificial neurones. Even though both concepts share a common ground, they live on very different time scales which rendered them in- compatible so far. Here, we bridge both ideas by demonstrating the excitation of magnetic auto- oscillations inside nano-scale DWs using pure spin currents. The spin-tranfer-torque (STT) effect discovered in 1996 by Slonczewski and Berger4,5 allows the manipu- lation of localised magnetic moments in a ferromagnet by the transfer of spin angular momentum from spin po- larised conduction electrons. The direction of the mag- netisation can either be switched permanently6 or can be forced to oscillate at radio frequencies7. Quite soon thereafter it was recognised that a charge current in a fer- romagnet, which is intrinsically spin polarised, can move magnetic domain walls (DWs) in nanowires8. This gave rise to the idea of the magnetic racetrack memory1 and, quite recently, current induced skyrmion motion9 -- 11. While these schemes target nonvolatile, long term data storage, the STT effect in spin-torque nano-oscillators can be exploited to drive magnetic auto-oscillations12 and, eventually, to radiate spin waves13 by compensat- ing the intrinsic magnetic damping. Another leap was the development of spin-Hall nano-oscillators (SHNO)26,27 in which pure spin currents are generated via the spin-Hall effect (SHE)14 -- 17 and by which even propagating spin waves are excited18,19. This puts SHNOs at the heart of magnonics20 -- 22 which proposes a novel type of low energy, non boolean computing based on magnons, the quanta of spin waves, as carriers of information23 or even neuromorphic computing3 based on the nonlinear char- acter of magnonic auto-oscillations. In a previous work, we demonstrated that DWs can channel magnons in an effective magnetic potential well24. This raised the ques- tion if a magnetic DW can potentially be a self-organized, movable SHNO. In order for the STT effect to counteract the intrin- sic magnetic damping, the magnetisation (cid:126)M has to have a component antiparallel to the polarisation (cid:126)P of the injected spin current Is with density (cid:126)Js. Due to spin- dependent scattering associated with the SHE, a charge current Ic with density (cid:126)Jc flowing in a heavy metal is con- verted into a transverse pure spin current with (cid:126)Js⊥(cid:126)Jc⊥(cid:126)P (Fig. 1a). In a conventional nanowire, therefore, an ex- FIG. 1. Sample layout and working principle. a Schematic of a SHNO based on a domain wall. The func- tional layers of Pt and Co40Fe40B20 are separated only for illustration purposes. b, SEM image of the investigated 370- nm wide nanowire that exhibits two 90◦ bends at positions A and B and is connected to a current source by two Cu contacts. 8 1 0 2 c e D 8 1 ] l l a h - s e m . t a m - d n o c [ 2 v 2 0 7 6 0 . 2 1 8 1 : v i X r a ab4 µmCr(1.5)/Pt(7)/Co40Fe40B20(5)/Cr(1.5)Cu contactCu contactyxSi/SiO2 substrateHext370 nmAB+-Jc2 ternal magnetic field is needed to overcome the shape anisotropy and to align (cid:126)M⊥(cid:126)Jc 25. Additionally, Is needs to be high enough to exert sufficient torque and compen- sate the magnetic damping. Typically, this is achieved by lithographically patterning devices with active areas of only few 10 − 100 nanometres28. However, the sizes are limited by lithography and the position of the active area is fixed. To overcome these limitations, we investigated the excitation of auto- oscillations inside magnetic DWs which bring multiple benefits: they can be created and destroyed dynamically, are very small, inherently exhibit (cid:126)M⊥(cid:126)Jc and can easily be moved by electric currents, lasers, and small mag- netic fields. This Letter demonstrates the possibility to drive strongly localised magnon auto-oscillations inside 90◦ DWs before they are moved by current induced DW motion. (Fig. 1). in direct In our At the heart of any SHNO is a bilayer stack to a ferro- contact of a heavy metal experiments, we use magnet Cr(1.5)/Pt(7)/Co40Fe40B20(5)/Cr(1.5) multilayers, with Pt and CoFeB as the functional materials. Two Ta(5)/Cu(150)/Ta(5) contacts allow for the application of a direct charge current Ic with density (cid:126)Jc. All thick- nesses are given in nanometers. As is depicted in Fig. 1a, the SHE in Pt converts (cid:126)Jc into a transverse pure spin current Is with density (cid:126)Js and polarity (cid:126)P. For a parallel (anti-parallel) orientation of (cid:126)P and (cid:126)M, the STT leads to an increased (decreased) effective damping of the mag- netisation dynamics in the CoFeB. To reproducibly con- trol the formation of a DW, we pattern the functional stack to a 370-nm wide zig-zag shaped nanowire with two 90◦ bends, as highlighted in light green in the scanning electron microscopy (SEM) image in Fig. 1b. In order to define the magnetic ground state, the sample can be sat- urated by an external magnetic field (cid:126)Hext applied along the y-axis. FIG. 2. Auto-oscillations in the nanowire with ap- plied field. Micromagnetic simulation of a, the x- and b, the y-component of the equilibrium magnetisation under an applied field of µ0Hext = 86 mT. c,d, BLS spectra recorded at positions A and B, respectively, with an external field of µ0Hext = 86 mT covering the current range between −0.1 to −5 mA and 0.1 to 5 mA, respectively. Insets e,f show the two- dimensional intensity distributions of auto-oscillations excited at 9.3 and 9.9 GHz with Ic = 4 mA. Please note that the dc sweeps and the two-dimensional maps were recorded on dif- ferent samples. Due to minor differences in the lithographic patterning, there is slight mismatch in the auto-oscillation frequencies recorded at 4 mA. First, we saturated the structure at µ0Hext = 500 mT and then reduced the value to µ0Hext = 86 mT. In Fig. 2a,b, we plot the simulated x- and y-component of the magnetisation in the resulting equilibrium state. As can be seen from the My-component (Fig. 2b), the magnetisation in the centre of the nanowire width aligns along the external field. Only at the boundaries, the magnetic moments line up with the edges to reduce the demagnetisation field, as evidenced by the non-vanishing Mx-component in Fig. 2a. This state results in (cid:126)M⊥(cid:126)Jc only inside the 90◦ bends at positions A and B. In the straight parts of the nanowire, (cid:126)Jc flows at a 45◦ angle with respect to (cid:126)M, which strongly reduces the efficiency of the STT and, thus, the excitation of auto-oscillations. To observe these auto-oscillations, we use space- resolved Brillouin light scattering (BLS) microscopy29. In Fig. 2c,d, we plot the BLS intensity that was mea- sured at positions A and B, respectively, as a function of the BLS frequency and the applied charge current which was swept consecutively from −0.1 to −5 mA and from 0.1 to 5 mA. For negative currents, no auto-oscillations are excited because (cid:126)P(cid:107) (cid:126)M which leads to an increased ef- fective damping. For positive currents, however, (cid:126)P is an- tiparallel to (cid:126)M, and at Ic > 2.2 mA one auto-oscillation mode is detected at a frequency of about 9 GHz. For Ic > 3.4 mA, even two modes are observed, starting at 9.6 GHz and 10.2 GHz. All modes show a negative fre- quency shift with increasing Ic which is expected for in- plane magnetised SHNOs12. In order to illustrate the spatial character of the excited auto-oscillations, we picked a direct current of Ic = 4 mA at µ0Hext = 86 mT. At this specific current, two auto- oscillations are excited at 9.3 and 9.9 GHz. The spatial intensity distributions in Fig. 2e,f show a strong localisa- tion of the 9.3 GHz mode at the apex whereas the mode at 9.9 GHz is much more diverged. The localisation is related to the changing angle between (cid:126)Jc and (cid:126)M as well as the slightly increased current density at the 90◦ bends. Up to now, we always applied an external mag- 24681012BLS frequency (GHz)BLS intensity0550measured at position A-5-4-3-2-1012345Direct current (mA)24681012BLS frequency (GHz)measured at position B9.3 GHz9.9 GHzyxµ0Hext = 86 mTabcdefABABMy-maxmaxMx-maxmax3 In the straight parts of the nanowire, the charge cur- rent flows along the magnetisation ((cid:126)Jc(cid:107) (cid:126)M) so that no auto-oscillations can be excited. Only inside the DWs at positions A and B, the direct current and the magneti- sation are aligned perpendicularly so that we expect to observe auto-oscillations only in these narrow regions. When we apply currents between −0.1 and −5 mA and measure the BLS intensities inside the DWs at positions A and B (see Fig. 3c,d respectively), we again see no auto-oscillations due to increased damping in this di- rection. For positive currents, however, we observe two auto-oscillation modes at 2.6 and 3.5 GHz. These modes do not show a pronounced negative frequency shift and one of them is in fact strongly localised to the DW, as is confirmed by spatially resolved BLS microscopy (see insets in Fig. 3d). The experimental results are further corroborated by micromagnetic simulations in Fig. 3e, which show the localisation of both modes and a node in the dynamic magnetisation profile of the higher fre- quency mode. We would like to stress the fact that for Ic > 4.5 mA the measured BLS intensity at positions A and B abruptly disappears. Indeed, this is still the case when we decrease the current again. Only a repeated satura- tion and relaxation of the sample, as described before, restores the auto-oscillations. This illustrates that for Ic > 4.5 mA the DWs are removed from the apexes due to the current-induced DW motion. For lower currents, this motion is suppressed by the dipolar pinning of the DWs at the apexes. One can exploit this motion with a precise current pulse, e. g., to shift one DW from position A to position B. In conclusion, we have demonstrated that a pinned transversal DW is able to work as a nano-sized spin- torque oscillator. Once prepared, this oscillator is able to perform at zero external magnetic field. Moreover, the auto-oscillations appear at current densities at which the current-induced DW motion not yet sets in. In other words, the DW oscillates at GHz frequencies before it moves. In principle, the ratio of these two thresholds can be tuned by changing the relative thicknesses or materi- als of the functional layers. If desired, the DW motion could also be completely shut off by replacing the metallic ferromagnet with an insulating one. On the other hand, the combination of auto-oscillations and current-induced DW motion yields various possibilities for research in fun- damental DW physics as well as for applications. One could think of using moveable DW spin-torque oscillators in networks to achieve more flexibility in synchronisation or frequency locking or even further advance concepts for magnetic racetrack memories or neuromorphic comput- ing. METHODS Sample Preparation. A zigzag structure was fabri- cated from Cr(1.5)/Pt(7)/Co40Fe40B20(5)/Cr(1.5) mul- FIG. 3. Auto-oscillations inside magnetic domain walls. a, Micromagnetic simulation and b, MFM measure- ment of the remanent magnetisation after saturating the sam- ple with µ0Hext = 500 mT and reducing the field back to zero. c,d, BLS spectra measured as a function of the applied Ic on the DWs at positions A and B, respectively. The direct cur- rent was swept from −0.1 to −5 mA and from 0.1 to 5 mA, respectively. In between the consecutive current sweeps, the sample was saturated to ensure the proper domain configura- tion. Insets in d show the two-dimensional BLS intensity dis- tributions of auto-oscillations excited in the remanent state at 2.6 and 3.5 GHz for Ic = 4 mA. e, Micromagnetic simulations of the current sweep (left) and the temporal evolution (right) of the amplitudes of the simulated auto-oscillation modes ex- cited at 2.6 and 3.5 GHz for a half cycle with Ic = 4 mA. netic field in order to drive auto-oscillations within the nanowire. In the next step, we prepared a magnetic transverse DW at the 90 bends. Therefore, we again satu- rated the nanowire at µ0Hext = 500 mT but then reduced the external magnetic field to zero. Figure 3a depicts the resulting remanent state of the nanowire as derived from micromagnetic simulations. The shape anisotropy directs the magnetisation to align along the nanowire, which cre- ates a head-to-head (tail-to-tail) DW at position A (B), respectively. The domain configuration in the remanent state is also confirmed by magnetic force microscopy (MFM) in Fig. 3b. The bright (dark) contrast at position A (B) in the dFz/dz-signal confirms the formation of DWs where Fz is the z component of the magnetostatic tip-sample interaction. 2.6 GHzt = 0.065 ns0.115 ns0.165 ns0.214 ns0.264 ns3.5 GHzt = 0.044 ns0.082 ns0.119 ns0.156 ns0.192 nsaHext = 0yxAB-5-4-3-2-1012345Direct current (mA)23452345BLS intensity15180cdposition Ahead-to-head DWposition Btail-to-tail DW2.6 GHz3.5 GHzBLS frequency (GHz)onset of current-induced DW motionb∆Mx-maxmax1 µmdFz /dzmaxminMy-maxmaxMx-maxmaxe370 nm0max∆Mx0123456I (mA)2345f (GHz)cMagnetic force microscopy. tilayer on Si/SiO2 substrate using electron beam lithog- raphy, magnetron sputtering and lift off. On the zigzag structure, a pair of electric contacts was made from a Ta(5)/Cu(150)/Ta(5) multilayer to inject a direct current to the Pt layer. All thicknesses are given in nanometres. In order to prepare the remanent state, an external field of µ0Hext = 500 mT was applied and slowly reduced to zero before starting MFM measurements. The measurements were performed in a Nanoscan high resolution magnetic force microscope in high vacuum using a high aspect ratio MFM probe (Team Nanotec HR-MFM45 with ML1 coating and a spring constant of 0.7 N/m). The local tip-sample dis- tance was 82 nm. Micromagnetic simulation The micromagnetic ground states as well as the auto-oscillation modes were obtained using the code from the GPU-based finite- difference micromagnetic package MuMax3.30 The ma- terial parameters used in the simulations include the sat- uration magnetization µ0Ms = 1.28 T, exchange stiff- ness of Aex = 20 pJ/m, and damping parameter α = 0.001. We ran the simulations for a geometry with a lateral size of 4000×3000×5 nm3, and unit cell size 3.90625×3.90625×5 nm3. Absorbing boundary condi- tions in the form of a smooth increase of the damping profile are applied to the open edges of the nanowire to avoid artifacts from spin-wave reflection. The direct charge current transport were simulated using COMSOL Multiphysics (www.comsol.com) using layer resistivities of 90 µΩcm and 11.2 µΩcm for CoFeB and Pt, respec- tively. The current profile and the Oersted field are then supplied to MuMax3, taking into account a spin-Hall an- gle of θSH =0.08 and the calculated spatial distribution of the spin current polarization. The auto-oscillation spec- tra were obtained by applying a Fast Fourier Transform (FFT) to the net x component of the total magnetization, simulated over 100 ns. The spatial profiles of the auto- oscillation modes are constructed via an FFT for each simulation cell31. The auto-oscillation modes show very similar profiles and frequencies as the linear modes ex- cited with a field pulse at zero current, in other words the DW auto-oscillations nucleate from linear modes, similar to the high frequency modes in other SHNOs32. Brillouin light scattering microscopy. All mea- surements were performed at room temperature. The auto-oscillation intensity is locally recorded by means of BLS microscopy. This method is based on the inelastic scattering of light and magnetic oscillations. Light from a continuous wave, single-frequency 532-nm solid-state laser is focused on the sample surface using a high numer- ical aperture microscope lens giving a spatial resolution of 340 nm. The laser power on the sample surface is typi- cally about 1 mW. The frequency shift of the inelastically scattered light is analysed using a six-pass Fabry-Perot interferometer TFP-2 (JRS Scientific Instruments). To record two-dimensional maps of the intensity distribu- tion, the sample is moved via a high-precision translation stage (10-nm step size, Newport). The sample position 4 is continuously monitored with a CCD camera using the same microscope lens. A home-built active stabilisation algorithm based on picture recognition allows for control- ling the sample position with respect to the laser focus with a precision better than 20 nm. REFERENCES 1Parkin, S. S. P., Hayashi, M. & Thomas, L. Magnetic domain- wall racetrack memory. Science 320, 190-194 (2008). 2Locatelli, N., Cros, V. & Grollier, J. Spin-torque building blocks. Nat. Publ. Group 13, 11-20 (2014). 3Torrejon, J. et al. Neuromorphic computing with nanoscale spin- tronic oscillators. Nature 547, 428-431 (2017). 4Slonczewski, J. C. Current-driven excitation of magnetic multi- layers. J. Magn. Magn. Mater. 159, L1L7 (1996). 5Berger, L. Emission of spin waves by a magnetic multilayer tra- versed by a current. Phys. Rev. B 54, 93539358 (1996). 6Myers, E. et al. Current-induced switching of domains in mag- netic multilayer devices. Science 285, 867 (1999). 7Kiselev, S. I. et al. Microwave oscillations of a nanomagnet driven by a spin-polarized current Nature 425, 380 (2003). 8Beach, G., Tsoi, M., & Erskine J. L. Current-induced domain wall motion. Journal of Magnetism and Magnetic Materials 320, 1272 (2008) 9Skyrme, T. H. R. A unified field theory of mesons and baryons. Nucl. Phys. 31, 556569 (1962). 10Muhlbauer, S. et al. Skyrmion lattice in a chiral magnet. Science 323, 915919 (2009). 11Fert, A., Cros, V. & Sampaio, J. Skyrmions on the track. Nature Nano. 8, 152-156 (2013). 12Slavin, A. N. & Tiberkevich, V. Nonlinear auto-oscillator theory of microwave generation by spin-polarized current, IEEE Trans- actions on Magnetics, 45, 1875 (2009). 13Madami, M. et al., Direct observation of a propagating spin wave induced by spin-transfer torque. Nature Nanotech. 6, 635 (2011). 14Dyakonov, M. I. & Perel, V. I. Possibility of orienting electron spins with current. Sov. Phys. JETP Lett. 13, 467469 (1971). 15Hirsch, J. E. Spin Hall effect. Phys. Rev. Lett. 83, 18341837 (1999). 16Ando, K. et al. Electric manipulation of spin relaxation using the spin Hall effect. Phys. Rev. Lett. 101, 036601 (2008). 17Hoffmann, A. IEEE Trans. Magn. 49, 5172 (2013). 18Urazhdin, S., et al. Excitation of coherent propagating spin waves by pure spin currents. Nat. Commun. 7, 1-6 (2016). 19Divinskiy, B. et al. Excitation and amplification of spin waves by spin-orbit torque. Advanced Materials 30, 1802837 (2018). 20Neusser, S. & Grundler, D. Magnonics: spin waves on the nanoscale. Adv. Mater. 21, 29272932 (2009). 21Kruglyak, V. V., Demokritov, S. O. & Grundler, D. Magnonics. J. Phys. D 43, 264001 (2010). 22Lenk, B., Ulrichs, H., Garbs, F. & Munzenberg, M. The building blocks of magnonics. Phys. Rep. 507, 107136 (2011). 23Chumak, A. V., Vasyuchka, V. I., Serga, A. A. & Hillbrands, B. Magnon spintronics. Nature Physics 11, 453461 (2015). 24Wagner, K. et al. Magnetic domain walls as reconfigurable spin- wave nanochannels. Nature Nanotech. 11, 432 (2016). 25Duan, Z., et al. Nanowire spin torque oscillator driven by spin orbit torques. Nat. Commun. 5, 5616 (2014). 26Demidov, V. E., et al. Magnetic nano-oscillator driven by pure spin current. Nature Materials 11, 1028-1031 (2012). 27Demidov, V. E., Urazhdin, S., Zholud, A., Sadovnikov, A. V., & Demokritov, S. O. Nanoconstriction-based spin-Hall nano- oscillator Appl. Phys. Lett. 105, 172410 (2014). 28Durrenfeld, P., Awad, A. A., Houshang, A., Dumas, R. K., & Akerman, J. A., 20 nm spin Hall nano-oscillator Nanoscale 9, 1285 (2017). 29Sebastian, T., Schultheiss, K., Obry, B., Hillebrands, B. & Schultheiss, H. Micro-focused Brillouin light scattering: imaging spin waves at the nanoscale. Front. Phys. 3, 35 (2015). 30Vansteenkiste, A. et al. The design and verification of Mumax3. AIP Advances. 10, 107133 (2014). 31McMichael, R. D. & Stiles, M. D., Magnetic normal modes of nanoelements. J. Appl. Phys.. 10, 10J901 (2005). 32Dvornik, M., Awad, A. A. & Akerman, J., Origin of Magneti- zation Auto-Oscillations in Constriction-Based Spin Hall Nano- Oscillators. Phys. Rev. Applied. 9, 014017 (2018). ACKNOWLEDGMENTS Financial support by the Deutsche Forschungsgemein- schaft within programme SCHU2922/1-1 is gratefully acknowledged. N.S. acknowledges funding from the Alexander von Humboldt Foundation. K.S. acknowl- edges funding from the Helmholtz Postdoc Programme. Samples were fabricated at the Nanofabrication Facili- ties (NanoFaRo) at the Institute of Ion Beam Physics and Materials Research at HZDR. We thank Dr. Ingolf Monch for deposition of the Ta/Cu/Ta multilayer. 5 AUTHOR CONTRIBUTIONS H.S., N.S. conceived and designed the experiments. N.S. and S.S.P.K.A performed lithographic processing and fabricated the thin film. L.K. performed static micromagnetic simulations, A.A.A. performed COM- SOL modeling and dynamical micromagnetic simula- tions. N.S. performed BLS experiments. N.P. and T.M. performed MFM measurements. H.S., K.S. and L.K. wrote the manuscript. All authors discussed the results and commented on the manuscript. ADDITIONAL INFORMATION The authors declare no competing financial interests. Reprints and permission information is available online at http://www.nature.com/reprints.
1302.6307	1	1302	2013-02-26T03:58:28	Current partition at topological zero-line intersections	[ "cond-mat.mes-hall" ]	An intersection between one-dimensional chiral acts as a topological current splitter. We find that the splitting of a chiral zero-line mode obeys very simple, yet highly counterintuitive, partition laws which relate current paths to the geometry of the intersection. Our results have far reaching implications for device proposals based on chiral zero-line transport in the design of electron beam splitters and interferometers, and for understanding transport properties in systems where multiple topological domains lead to a statistical network of chiral channels.	cond-mat.mes-hall	cond-mat	Current partition at topological zero-line intersections Zhenhua Qiao,1 Jeil Jung,1 Chungwei Lin,1 Allan H. MacDonald,1 and Qian Niu1, 2 1Department of Physics, The University of Texas at Austin, Austin, Texas 78712, USA 2International Center for Quantum Materials, Peking University, Beijing 100871, China (Dated: June 17, 2018) An intersection between one-dimensional chiral acts as a topological current splitter. We find that the splitting of a chiral zero-line mode obeys very simple, yet highly counterintuitive, partition laws which relate current paths to the geometry of the intersection. Our results have far reaching implications for device proposals based on chiral zero-line transport in the design of electron beam splitters and interferometers, and for understanding transport properties in systems where multiple topological domains lead to a statistical network of chiral channels. chiral A massive two-dimensional electron gas (C2DEG) has a valley Hall conductivity that has the same sign as its mass. The valley Hall effect leads to conducting edge states and also, when the mass param- eter varies spatially, to conducting states localized along mass zero-lines. 1 -- 3 Provided that inter-valley scattering is weak, zero-line state properties are closely analogous1 -- 3 to edge state properties of quantum spin-Hall insula- tors and include both chiral propagation and suppressed backscattering.1 Metallic zero-line modes (ZLMs), or topological 1D kink states, provide a two dimensional realization of Dirac zero energy modes,4,5 and their ex- istence has been proposed in a wide variety of systems including graphene mono and bilayers,1 -- 3,6 -- 8 topologi- cal insulators with lattice dislocations,9 boron nitride crystals with grain boundaries,10 superfluid 3He,11 and photonic crystals.12,13 In the present Letter we examine current partition properties at zero-lines intersections,1,3 which are expected to be ubiquitous in systems in which the mass term results from a disorder potential or from spontaneous symmetry breaking. ZLMs in C2DEGs are centered on zero-lines of the mass1 -- 3,6,7, i.e. on lines along which the mass changes sign as illustrated in Figure 1a. A mass term leading to a valley Hall effect14,15 can be produced by a sublattice staggered external potential in single layer graphene,6,7 and more practically by a gate controlled interlayer po- tential difference in Bernal bilayer and ABC stacked mul- tilayer graphene.1 -- 3,8 Mass terms can also be generated by spin-orbit coupling16 -- 18 and by electron-electron inter- actions.19 -- 21 In this last case ZLMs22 appear naturally at domain walls separating regions with different local anomalous, spin, or valley Hall conductivities. Chiral propagation implies that ZLMs can travel only in the direction which places negative masses either on their left, or depending on valley, on their right. It fol- lows, as illustrated in Figure 1c, that there is no for- ward propagation at a zero-lines intersection; a propa- gating mode is split between a portion that turns clock- wise and a portion that turns counterclockwise. These unusual transport properties are potentially valuable for new types of electronic devices. We have therefore carried out quantum transport calculations for an explicit model of intersecting ZLMs in order to discover rules for current partitioning at such a ZLM splitter. The system we study is a π-band tight-binding model for single-layer graphene with a position-dependent sublattice-staggered potential constructed to form intersecting zero lines which enable propagation to four ZLMs labeled left (L), right (R), up (U) and down (D) in Figure 1b. For simplicity, we consider the case where the U and D ZLMs are fixed along the vertical direction, and we define the angles between R and D ZLMs to be α and between L and U ZLMs to be β. The blue and orange lines in Figure 1b indicate the allowed chiral propagation paths. i ci. Here c† −tPhiji c† i cj + UAPi,A c† i ci + UBPi,B c† The numerical results reported on below are for a π-orbital tight-binding Hamiltonian with nearest neigh- bor hopping and a sublattice-staggered potential: H = i (ci) is a creation (annihilation) operator for an electron at site i, and t = 2.6 eV is the nearest neighbor hopping ampli- tude. For a sublattice staggered potential the A and B sublattice energies are opposite, i.e. UA = −UB = λU0, where U0 measures the potential strength and λ="±" determines the sign of the valley Hall effect in each quad- rant. In all our simulations, the potential amplitude was chosen to be U0/t = 0.05. The ZLMs appear confined along zero-lines where the mass becomes zero and the resulting wave-function tails spread into the bulk with a depth proportional to the inverse of the mass1,3. Although our study has been carried out in single layer graphene for computational convenience, we ex- pect that similar conclusions apply to ZLMs in other systems, and in particular in bilayer graphene where the spatially patterned mass term can in principle be gener- ated externally using gates. Our transport calculations are based on the Landauer-Buttiker formalism23 and re- cursively constructed Green functions.24,25 The conduc- tance from lead q to lead p is numerically evaluated from Gpq = (e2/h)Tr[ΓpGrΓqGa], where e is the elec- tron charge, h is the Planck's constant, Gr,a are the re- tarded and advanced Green's functions,23 and Γp is a linewidth function describing the coupling between lead p and the central region. The propagation of ZLMs, or kink states, incoming from lead p is effectively illus- trated by plotting a map of its contribution to the lo- cal density of states (LDOS) at an energy ε in the gap: ρp(r, ε) = 1/2π[GrΓpGa]rr. Here r is the real space co- a b + −− + + − + + − 2 + − + − + − ) m c n ( y 20 ) m n ( y 0 +− +− +− -20 -8 0 8 -8 -8 0 0 8 x (nm) -8 0 8 FIG. 1: Current partition at zero-line intersections: a. Staggered sub lattice-potential, and hence valley Hall conductivity, spatial distribution pattern which defines the zero-line paths. Blue dots indicate positive site energies and green dots negative site energies. b. Schematic of four-terminal graphene samples with different staggered potential distributions. The Left (L), right (R), up (U), and down (D) leads are extend indefinitely from the plotted central scattering region. The U and D ZLMs are fixed along the vertical direction and the angles α and β specify the R and L ZLM directions. The solid and dashed lines in black denote zero lines. The blue and orange lines represent allowed chiral propagation paths. c. ZLM LDOS distribution for modes incident from lead R for α = 90◦, 60◦, 120◦ at fixed β = 90◦. ordinate. The central scattering region in our calculations is rect- angular with size nx = 94 and ny = 432 as explained fully in the Supplementary Information. The valley label of a state is of course not a good quantum number for ZLMs and valleys are most strongly mixed when their wave vec- tor projections in the propagation direction are identical. For energies inside the gap this coincidence happens only for propagation in the armchair direction1,2,6,10. How- ever, numerical calculations have shown a remarkable absence of back-scattering at sharp turns in the zero- line or at the encounter of a ZLM splitter1 except in a narrow energy range very close to the avoided crossing gap centered on ε/t = 0.00 between modes with opposite propagation directions. For the results shown below we have chosen ε/t = 0.01 to avoid this energy range; the chirality of the ZLM modes is then very well defined. In a four terminal ZLM splitter device [see Figures 1b], there are in total twelve distinct inter-terminal conduc- tance values. The number of independent conductances is reduced to six in time-reversal symmetric systems since Gpq = Gqp. For chiral transport forward scat- tering and back scattering are absent at a ZLM inter- section, reducing the number of independent parame- ters further. The current conservation then implies that Gpr + Gqr = G0 = e2/h for any value of r and p, q the labels of the two neighboring leads. It follows that GLU = GRD & GRU = GLD, (1) and that GRU + GRD = e2/h, leaving only one indepen- dent parameter for the entire four terminal systems. In a ZLM splitter with zero backscattering and perfect chi- ral current filtering transport is completely characterized by specifying how incoming current at an intersection is partitioned between clockwise and counterclockwise ro- tation outgoing directions. The partition law must be the same for all incoming channels. In the following, we fo- cus on the conductances GUR and GDR corresponding to the currents incoming from lead R. The above relations were numerically verified for a ZLM current splitter with α = β = 90◦ in Ref. [1] and we have now numerically verified that they are true for arbitrary values of the lead angles α and β. (See the Supplementary Information for further details.) For a ZLM splitter with a fixed vertical pair of leads U and D and rotatable R and L leads as shown in Figure 1b, we have numerically discovered a rather simple law which describes the dependence of the current partition on the angles α and β with surprising accuracy: (See Figure 2 for a summary of the data which supports this law and the Supplementary Information for further details.) GUR = GDR = G0 2 G0 2 [1 − sin (α + β)] , [1 + sin (α + β)] , (2) for 90◦ ≤ α + β ≤ 270◦. Outside of this angle range the current follows the path with the larger rotation angle: GDR = GUL = G0 and GUR = GDL = 0 when 0◦ < α + β < 90◦, or GUR = GDL = G0 and GDR = GUL = 0 when 270◦ < α + β < 360◦. For the special case of β = 90◦ the current partition law simplifies to GUR = G0 sin2(α/2) and GDR = G0 cos2(α/2). And for the special case of α = β, corresponding to zero lines that are straight at the intersection point and therefore to a mass pattern defined by a smooth external potential, GUR = G0 sin2(α − 45◦) and GDR = G0 cos2(α − 45◦). Contrary to intuition, as illustrated in Figure 1c for a current incoming from lead R, the current partition ) h / 2 e ( e c n a t c u d n o C 1.0 0.5 0.0 G DR G UR 150 131 120 90 60 49 16 60 120 a +b 180 (degree) 240 300 FIG. 2: Current partition law: Current as a function of angles α and β for the zero-line intersection defined in Fig. 1c. The current partition follows a simple relation given by Eq. 2 whenever 90◦ ≤ α + β ≤ 270◦ and completely follows the sharper rotation path outside this range. The solid and dashed lines are the fitting functions defined in Equation (2). law implies that more current always follows the path re- quiring a larger rotation angle. This makes sense since this pair of incoming and outgoing modes interact over a longer distance, providing more opportunity for interac- tions which lead to inter-mode scattering (The zero-line mode decay length is ∼ mvD/ where m is the local mass and vD is the Dirac point velocity for m ≡ 0.6). It is noteworthy that the current partition depends only on the combination α + β and indeed it is surprising that the angle β, which specifies the orientation of a lead that does not carry any current, influences the partition law. This law evidently results, however, from the interfer- ence between ZLMs close to the ZLM intersection point, allowing inactive leads to play a role in the scattering. Note that the completely unbalanced current partition favored by a very small value of α (a sharp turn from R to D) is mitigated if β (the L to U turn angle) is larger than 90◦. We emphasize we have not derived the form which fits our current partition numerical results analytically and it is likely only approximate. It applies accurately only at energies close to the middle of the bulk energy gaps. The expression can be rationalized by the following ar- gument. Let us consider a ZLM splitter, where the chiral current filtering rule restricts current incoming from R or L to scatter into the U and D leads. By denoting the asymptotic amplitudes of the ZLM at the leads as AR, AL, AU , and AD, current conservation implies that AD (cid:21) = (cid:20) cos(τ )eiu − sin(τ )e−iv (cid:20) AU sin(τ )eiv cos(τ )e−iu (cid:21)(cid:20) AL AR (cid:21) . (3) Equation (3) relates the amplitudes of incoming and outgoing waves via a general SU(2) unitary transforma- tion matrix with parameters u, v, and τ . When we set AL to be zero, then for any AR, AU 2 = AR2 sin2 τ , 3 a 1.0 > R D G < 0.5 0.0 1.0 b > R U G < 0.5 30o 60o 90o 120o 120o 90o 60o 30o 0.0 0.0 0.5 1.0 Disorder strength W/t FIG. 3: Influence of disorder on the current partition law: Averaged conductances hGDRi (in panel a) and hGURi (in panel b) from lead R to leads D and U as a function of disorder strength W at fixed β = 90◦. Four representative an- gles of incidence α = 30◦, 60◦, 90◦, and 120◦ are considered. Over 50 samples are collected at each point. The error bar is used to indicate the strength of fluctuations as a function of disorder realization. and AD2 = AR2 cos2 τ . The scattering phases u and v are irrelevant for the scattering probability and will henceforth be dropped. The conductances from R can thus be written as GUR = G0 sin2 τ (α, β) and GDR = G0 cos2 τ (α, β), where G0 = e2/h. Equipartition for (α, β) = (α, 180◦ − α) implies that τ (α, 180◦ − α) = 45◦. For the current coming from R into D lead, we get from symmetry considerations the relation GDR(α, β) = GDR(β, α). If we assume that τ is a smooth function of α and β and additionally require that the conductances are invariant under (α, β) → (α + N × 360◦, β + M × 360◦) we can conclude that τ (α, β) = c(α + β)/2 − 45◦, where c = ±1. Then we impose a second condition for the cur- rent partition saturation when α+β = 90◦, for instance in the limit when β = 90◦ and α = 0◦ where R and D ZLMs merge together, that further restricts c = +1. This ex- pression reproduces Equation (2) for GDR and GUR, out- side the saturation region which occurs for α + β < 90◦ and α + β > 270◦. Now we examine the robustness of our results in the presence of disorder. Long-range disorder is not effec- tive in producing inter valley scattering1 so we focus on the potentially more important short-range disor- der which we model as a random potential at each lat- i ci with ωi being uniformly distributed in the interval of [−W/2, +W/2], where W characterizes the disorder strength. In our simulations we considered over 50 real- tice site. Specifically we add a term Hdis = Pi ωic† b izations of the random disorder potential for each value of the strength. Figure 3 illustrates our results for the average conductances hGURi and hGDRi as a function of disorder strength W . We see that the current parti- tion law remains accurate up to disorder strengths larger than the bulk band gap ∆/t = 0.1. When the disorder strength is further increased, the averaged conductances hGURi and hGDRi are slightly reduced and disorder fluc- tuations grow. For example, when the disorder strength reaches W/t = 1.0, 10 times larger than the bulk band gap, the averaged conductances still retain over 80% of their original values. The lost current takes advantage of the intervalley scattering to access the outgoing modes of the L lead or to backscatter in the R lead. All these findings strongly indicate that the current partition law is very robust to disorder, suggesting that ideal zero-line transport properties can be approximated in real devices. Since there are presently no practical techniques for imposing staggered sublattice potentials in single layer graphene, other closely related systems may ultimately be of greater experimental interest. In Bernal stacked bilayer graphene, for example, a ZLM splitter can be re- alized by using gates to achieve perpendicular electric fields which vary in sign spatially. Another possibility is 2D honeycomb photonic crystals, in which the Dirac points have been experimentally observed and sublattice staggered potentials can be realized by choosing different diameters for the cylinders which form the structure or by varying the dielectric material used. In summary, when intervalley scattering can be ne- glected, transport along the zero lines of a sublattice- staggered potential in graphene is chiral, requiring travel in a direction which keeps positive masses on either the 4 left or the right, depending on valley. We have used the Landauer-Buttiker formula and recursively constructed Green's functions to examine how chiral currents are par- titioned between available outgoing leads at a ZLM inter- section. We find that at energies near the middle of the bulk gap our numerical results for the dependence of cur- rent on ZLM geometry are accurately described by a sim- ple partition law specified in Eq. (2), and that the influ- ence of disorder on this law is weak. The helicity of ZLM provides a new mechanism for allowing or blocking cur- rents and may find applications in alternative designs for nanoelectronic devices or in enabling electron quantum interferometry26 in a new setting. We have explored, for the first time to our knowledge, the geometry-dependent current partition laws at the intersection of two zero- lines. It will be interesting to extend our present studies to more general parameter spaces and to look for similar- ities and differences with respect to other systems with chiral 1D transport channels including photonic crystals, quantum anomalous Hall and quantum Hall effect sys- tems, and chiral superconductors. Acknowledgements. -- We are grateful to Wang- Kong Tse, Wang Yao, Jian Li and Yue Yu for useful discussions and P. J. Coulchinsky for his help with fig- ures. This work was financially supported by the Welch Foundation (F-1255 and TBF1473), and by DOE (DE- FG03-02ER45958, Division of Materials Science and En- gineering) grant. Additional support was provided by grants NBRPC (2012CB-921300) and NSFC (91121004). The Computer Center of The University of Hong Kong is gratefully acknowledged for high-performance comput- ing assistance [supported in part by a Hong Kong UGC Special Equipment Grant (SEG HKU09)]. 1 Qiao, Z. H., Jung, J., Niu, Q. & MacDonald, A. H. Elec- tronic highways in bilayer graphene. Nano Lett. 11, 3453- 3459 (2011). 2 Jung, J., Zhang, F., Qiao, Z. H. & MacDonald, A. H. Valley-Hall kink and edge states in multilayer graphene. Phys. Rev. B 84, 075418 (2011). 3 Martin, I., Blanter, Ya. M. & Morpurgo, A. F. Topologi- cal confinement in bilayer graphene. Phys. Rev. Lett. 100, 036804 (2008). 4 Jackiw, R. & Rebbi, C. Solitons with fermion number 1/2. Phys. Rev. D 13, 3398 (1976). 5 Su, W. P., Schrieffer, J. R. and Heeger, A. Solitons in poliacetylene, Phys. Rev. Lett. 42, 1698 (1979). 6 Semenoff, G. W., Semenoff, V. & Zhou, F. Domain walls in gapped graphene. Phys. Rev. Lett. 101, 087204 (2008). 7 Yao, W., Yang, S. A. & Niu, Q. Edge states in graphene: from gapped flat-band to gapless chiral modes. Phys. Rev. Lett. 102, 096801 (2009). 8 Killi, M., Wu, S. & Paramekanti, A. Band structures of bi- layer graphene superlattices. Phys. Rev. Lett. 107, 086801 (2011). 9 Ran, Y., Zhang, Y. & Vishwanath, A. One-dimensional insulators topologically protected modes in topological with lattice dislocations. Nature Phys. 5, 298-303 (2009). 10 Jung, J., Qiao, Z. H., Niu, Q. & MacDonald, A. H. Trans- port properties of graphene nanoroads in boron nitride sheets. Nano Lett. 12, 2936-2940 (2012). 11 Volovik, G. E. The Universe in a Helium Droplet (Oxf. Univ. Press, Oxford, U.K., 2003). 12 Ochiai, T. & Onoda, M. Photonic analog of graphene model and its extension: Dirac cone, symmetry, and edge states. Phys. Rev. B 80, 155103 (2009). 13 Khanikaev, Alexander B., Mousavi, S. Hossein, Tse, Wang- Kong, Kargarian, Mehdi, MacDonald, Allan H., & Shvets, Gennady Photonic topological insulators. N ature Materi- als doi:10.1038/nmat3520. 14 Xiao, D., Yao, W. & Niu, Q. Valley-contrasting physics in graphene: magnetic moment and topological transport. Phys. Rev. Lett. 99, 236809 (2007). 15 Li, J., Martin, I., Buttiker, M. & Morpurgo, A. F. Topo- logical origin of subgap conductance in insulating bilayer graphene. Nature Phys. 7, 38 (2011). 16 Qiao, Z. H., Yang, S. A., Feng, W., Tse, W.-K., Ding, J., Yao, Y., Wang, J. & Niu, Q. Quantum anomalous Hall effect in graphene from Rashba and exchange effects. Phys. Rev. B 82, 161414 (2010). 5 17 Kane, C. L. & Mele, E. J. Quantum Spin Hall Effect in Graphene. Phys. Rev. Lett. 95, 226801 (2005) 18 Qiao, Z. H., Tse, W.-K., Jiang, H., Yao, Y. & Niu, Q. Two-dimensional topological insulator state and topologi- cal phase transition in bilayer graphene. Phys. Rev. Lett. 107, 256801 (2011). 19 Zhang, F., Jung, J., Fiete, G. A., Niu, Q. & MacDonald, A. H. Spontaneous quantum Hall states in chirally stacked few-layer graphene systems. Phys. Rev. Lett. 106, 156801 (2011). 20 Jung, J., Zhang, F., MacDonald, A. H. Lattice theory of pseudospin ferromagnetism in bilayer graphene: Compet- ing interaction-induced quantum Hall states. Phys. Rev. B 83, 115408 (2011). 21 Nandkishore, R. & Levitov, L. Quantum anomalous Hall state in bilayer graphene. Phys. Rev. B 82, 115124 (2010). 22 Zhang, F., Jung, J. & MacDonald, A. H. Spontaneous quantum Hall states and novel luttinger liquids in chi- ral graphene, Journal of Physics: Conf. Ser. 334, 012002 (2011). 23 Datta, S. Electronic Transport in Mesoscopic Systems (Camb. Univ. Press, Cambridge, 1995). 24 Lopez Sancho, M. P., Lopez Sancho, J. M. & Rubio, J. Quick iterative scheme for the calculation of transfer ma- trices: application to Mo (100). J. Phys. F: Met. Phys. 14 1205 (1984). 25 Wang, J. & Guo, H. Relation between nonequilibrium Green's function and Lippmann-Schwinger formalism in the first-principles quantum transport theory. Phys. Rev. B 79, 045119 (2009). 26 Ji, Y., Chung, Y., Sprinzak, D., Heiblum, M., Mahalu, D. & Shtrikman, H. An electronic Mach-Zehnder interferom- eter. Nature 422, 415-418 (2003).
1303.0039	1	1303	2013-02-28T23:01:28	Atomistic Boron-Doped Graphene Field Effect Transistors: A Route towards Unipolar Characteristics	[ "cond-mat.mes-hall" ]	We report fully quantum simulations of realistic models of boron-doped graphene-based field effect transistors, including atomistic details based on DFT calculations. We show that the self-consistent solution of the three-dimensional (3D) Poisson and Schr\"odinger equations with a representation in terms of a tight-binding Hamiltonian manages to accurately reproduce the DFT results for an isolated boron-doped graphene nanoribbon. Using a 3D Poisson/Schr\"odinger solver within the Non-Equilibrium Green's Functions (NEGF) formalism, self-consistent calculations of the gate-screened scattering potentials induced by the boron impurities have been performed, allowing the theoretical exploration of the tunability of transistor characteristics. The boron-doped graphene transistors are found to approach unipolar behavior as the boron concentration is increased, and by tuning the density of chemical dopants the electron-hole transport asymmetry can be finely adjusted. Correspondingly, the onset of a mobility gap in the device is observed. Although the computed asymmetries are not sufficient to warrant proper device operation, our results represent an initial step in the direction of improved transfer characteristics and, in particular, the developed simulation strategy is a powerful new tool for modeling doped graphene nanostructures.	cond-mat.mes-hall	cond-mat	Atomistic Boron-Doped Graphene Field Effect Transistors: A Route towards Unipolar Characteristics Dipartimento di Ingegneria dell'Informazione, Universit`a di Pisa, Via Caruso 16, 56122 Pisa, Italy Paolo Marconcini IMEP-LAHC, UMR 5130 (Grenoble INP/UJF/CNRS/Universit´e de Savoie), Minatec, 3 Parvis Louis N´eel, 38016 Grenoble, France Alessandro Cresti CEA LETI-MINATEC, 17 rue des Martyrs, 38054 Grenoble, France Fran¸cois Triozon Dipartimento di Ingegneria dell'Informazione, Universit`a di Pisa, Via Caruso 16, 56122 Pisa, Italy Gianluca Fiori Dpto. Electr´onica y Tecnolog´ıa de Computadores, Universidad de Granada, Facultad de Ciencias, Campus de Fuente Nueva, and CITIC, Campus de Aynadamar, Universidad de Granada E-18071 Granada, Spain Blanca Biel L Sim, SP2M, UMR-E CEA/UJF-Grenoble 1, INAC, Grenoble, France Yann-Michel Niquet Dipartimento di Ingegneria dell'Informazione, Universit`a di Pisa, Via Caruso 16, 56122 Pisa, Italy∗ Massimo Macucci Stephan Roche CIN2 (ICN-CSIC) and Universitat Aut´onoma de Barcelona, Catalan Institute of Nanotechnology, Campus UAB, 08193 Bellaterra (Barcelona), Spain, and ICREA, Instituci´o Catalana de Recerca i Estudis Avan¸cats, 08070 Barcelona, Spain We report fully quantum simulations of realistic models of boron-doped graphene-based field effect transistors, including atomistic details based on DFT calculations. We show that the self-consistent solution of the three-dimensional (3D) Poisson and Schrodinger equations with a representation in terms of a tight-binding Hamiltonian manages to accurately reproduce the DFT results for an isolated boron-doped graphene nanoribbon. Using a 3D Poisson/Schrodinger solver within the Non- Equilibrium Green's Functions (NEGF) formalism, self-consistent calculations of the gate-screened scattering potentials induced by the boron impurities have been performed, allowing the theoretical exploration of the tunability of transistor characteristics. The boron-doped graphene transistors are found to approach unipolar behavior as the boron concentration is increased, and by tuning the density of chemical dopants the electron-hole transport asymmetry can be finely adjusted. Correspondingly, the onset of a mobility gap in the device is observed. Although the computed asymmetries are not sufficient to warrant proper device operation, our results represent an initial step in the direction of improved transfer characteristics and, in particular, the developed simulation strategy is a powerful new tool for modeling doped graphene nanostructures. 3 1 0 2 b e F 8 2 ] l l a h - s e m . t a m - d n o c [ 1 v 9 3 0 0 . 3 0 3 1 : v i X r a The discovery of graphene has opened a promising al- ternative to silicon-based electronics.1–3 Reported charge mobilities in undoped graphene layers are actually or- ders of magnitude larger than those measured in silicon, but the unfortunate zero-gap semiconductor nature of this material severely limits the achievable Ion/Ioff ra- tio4 (ratio -infinite for an ideal switch- of the current flowing through the device in the conducting state to that flowing in the nonconducting state) for graphene transis- tors, making them not yet able to compete with main- stream silicon technologies.5 Another adverse feature of graphene-based devices is their ambipolar electrical be- havior, which precludes the development of a comple- mentary logic architecture. A way to induce an energy gap in graphene is to reduce its lateral dimension.6–9 By using e-beam litho- graphic techniques and oxygen plasma etching processes, graphene nanoribbons (GNRs) down to a width of 10 nanometers can be successfully fabricated.10,11 The re- sulting energy gap, however, is not sufficient for the achievement of a practically useful Ion/Ioff ratio in a transistor structure. Chemical, in-solution techniques have been developed that allow one to obtain nanorib- bons with a smaller width12,13 but are hard to integrate into a large-scale device production technology. A mixed approach has been followed by Wang and Dai,14 who use lithography to define the nanoribbons, whose width is then shrunk by means of gas-phase chemical etching. Al- though quite interesting, this procedure could have some limitation in terms of the achievable device density. Fur- thermore, energy bandgaps are very unstable with re- gards to edge reconstruction and defects,15 thus prevent- ing the fabrication of graphene transistors with perfor- mances comparable with their silicon counterparts. In- deed, almost ideal graphene nanoribbons have been ob- tained by means of careful chemical synthesis,16 however such an approach cannot be applied to the fabrication of complex circuits on insulating substrates in a straightfor- ward way. Strong chemical damage of graphene could in principle be used to tune the current flow, but with the drawback of a significant decay of the conductance that leads to poor performance in terms of current drive.17–21 Some of us have recently proposed a new class of de- vices based on the chemical doping of graphene by boron or nitrogen.22–24 Using first-principles and mesoscopic quantum transport calculations (within the Landauer- Buttiker formalism) some indications of different behav- iors for electrons and holes were predicted in ribbons with lateral sizes as large as 10 nm, with mobility gaps in the order of 1 eV.22,23 Here the mobility gaps are typ- ically defined by the corresponding energy window for which the conductance is smaller than a typical value of G = 10−2G0 (G0 = 2e2/h) (see Ref.25 for details). Re- cent experimental findings confirmed the possibility to include boron and nitrogen impurities in graphene,26–28 therefore suggesting doped graphene transistors as a gen- uine alternative to ultranarrow GNRs with a more con- trolled tuning of the conductance features. However, fully self-consistent simulations of doped graphene tran- sistors are mandatory to ascertain the real impact of graphene doping. We explore the effect, at room temperature, of boron doping in graphene-based field effect transistors using the device geometry illustrated in Fig. 1. Self-consistent NEGF calculations are performed, allowing for a proper treatment of the screened impurity potential and a more robust assessment of the doping effects on the electron- hole transport asymmetry and on the onset of mobility gaps. I. RESULTS AND DISCUSSION Parameters for our calculations have been obtained from a first-principle approach, as we detail in the fol- lowing. As an initial step, first-principle density functional theory (DFT) calculations were undertaken for a single substitutional boron impurity in two-dimensional (2D) graphene and in GNRs,22,23 by means of the SIESTA 2 code,29 within the local density approximation and us- ing a double-ζ basis set. Following the methodology in Refs.23,30, the onsite energies of the pz orbitals for the (bulk) 2D case were extracted. Their variation with the distance from the boron atom is shown in Fig. 2 (inset). The boron atom acts as an electron acceptor: the re- sulting impurity potential is repulsive for electrons and the screening length is found to be of the order of a few angstroms (the screening length31 is defined as the dis- tance from the charged impurity at which the surround- ing mobile charges, rearranging under the electrostatic action of the impurity, reduce its potential by a factor equal to Euler's number e). The transport properties of "quasi-metallic" and semi- conducting armchair GNRs (following Ref.32, we refer to an armchair GNR with N dimers contained in its unit cell as an "N-aGNR") with a single boron impurity at dif- ferent sites across the ribbon width are then computed, using the full ab initio Hamiltonian, in the Landauer- Buttiker framework.22 Boron impurities actually induce a strong hole backscattering, with much weaker effect on the electron side. A similar result was already obtained in boron-doped carbon nanotubes,33,34 and attributed to strong backscattering at the resonance energies of the quasibound states localized around the boron impurity. Due to the lack of the rotational symmetry present in carbon nanotubes, in GNRs the energy of the conduc- tance dip on the hole side depends on the position of the impurity with respect to the ribbon edge (as illustrated in Fig. 2 and Fig. 3). In Ref.23 it has been shown that these ab initio re- sults can be fairly well reproduced with a simpler tight- binding model with a single orbital per carbon and boron atom. For each atom, the onsite energy is taken as the impurity potential extracted from the ab initio calcula- tion. The onsite energy of the boron atom is the only parameter that has to be adjusted in order to achieve the best match between tight-binding and ab initio transmis- sion profiles. The success of such a simple tight-binding model in describing boron impurities originates from the preservation of the sp2 hybridization of graphene in the presence of substitutional boron. By using this tight-binding model, transport studies in large armchair GNRs with a random distribution of boron atoms, with concentrations between 0.02% and 0.2%, have been performed in Ref.23. The addition of the transmission dips induced by all impurities leads to a mobility gap in the valence band, i.e. an energy inter- val where conductance is nearly suppressed by scatter- ing, while transport in the conduction band is preserved (the opposite behavior would be found in the case of ni- trogen doping). This occurs for semiconducting as well as for "quasi-metallic" GNRs, independently of ribbon width, as long as the potential can be assumed to be constant along the channel. The mobility gap increases with boron concentration and ribbon length. Ref.23 sug- gested a possible exploitation of this phenomenon to ob- tain field-effect transistors (FET) with wide GNRs (> 10 nm), within the reach of conventional lithography. The absence of a significant bandgap in these ribbons could be compensated for by the mobility gap. However, the above mentioned study has been performed for ribbons without any gate-induced charge in the channel, since the impurity potential considered was the one extracted directly from an ab initio calculation without taking into account the effects of screening when a gate is applied. section) is performed for an isolated graphene layer with a single impurity, as in the ab initio calculation. As shown in Fig. 2 (inset), the obtained self-consistent potential around the impurity is very close to the ab initio result. Also the ab initio transmission spectra are accurately re- produced for all positions of the impurity, keeping the boron onsite energy equal to zero and with no further pa- rameter adjustment. Transmission spectra comparisons are shown in Fig. 2. 3 x y SiO2 source tox 1 nm L gate gate 1 nm tox drain FIG. 1. Sketch of the simulated double gate GNR field ef- fect transistor (FET): red dots represent the substitutional boron atoms. In the inset, the longitudinal cross-section of the transistor is shown (the scale is different along the x and y directions, to make the drawing clearer). In order to simulate the behavior of a realistic tran- sistor structure, we need to perform a self-consistent cal- culation to include the effects of an applied gate voltage and of a non-zero drain-source bias on the charge density. This can be done, at a reasonable level of computational complexity, with a tight-binding model set up with pa- rameters properly extracted from the ab initio results. However, differently from Ref.23, the impurity potential obtained from ab initio approaches cannot be directly used in a self-consistent simulation, since it would be overscreened, as a result of the Poisson/Schrodinger iter- ations. The approach chosen here is to mimic DFT self- consistency with the tight-binding model and a proper distribution of fixed charges; then the final impurity po- tential arises from the tight-binding Poisson/Schrodinger self-consistency. In particular, as we will detail in the Methods section, the ab initio results can be reproduced considering a fixed charge +e (where e is the elementary charge) in corre- spondence of each carbon atom, and using the following tight-binding representation: null onsite energies are con- sidered in correspondence of all atoms, while, following Ref.32, the hopping parameter is equal to tp = −2.7 eV for all the pairs of nearest neighbor atoms, with the ex- ception of the edge dimers, for which a value of 1.12 tp is used instead. To validate this approach, a self-consistent Pois- son/Schrodinger calculation (described in the Methods i i n o s s m s n a r T 3 2 1 0 ) V e ( l a i t n e o p t y t i r u p m I 5 4 3 2 1 0 0 ab initio self-consistent TB 1 Distance (nm) 2 dB-edge = 0.5a dB-edge = 8a -0.4 -0.2 0 Energy (eV) 0.2 0.4 FIG. 2. Transmission through an isolated 32-aGNR (3.81 nm wide) with one boron impurity placed at two different dis- tances from the ribbon edge: 0.5a and 8a, where a = 2.46 A is the hexagonal lattice constant. The bold lines represent the results with our self-consistent tight-binding model. The thin dotted line represents the ab initio results. Inset: com- parison between the results for the impurity potential in 2D graphene obtained with self-consistent tight-binding and ab initio approaches. Before performing full NEGF calculations, the effect of a gate with a nonzero applied voltage was analyzed on short ribbons with a single impurity. The ribbon is sur- rounded by a double gate with a dielectric thickness of 1 nm, as shown in Fig. 1. The source-drain voltage and the Fermi level are set to 0. The charge and potential on the ribbon adjust self-consistently as a function of the gate potential VG. First, the calculation was performed at VG = 0 with vacuum chosen as gate dielectric. The ob- tained potential is very close to the result discussed above for the neutral isolated ribbon. This gate configuration does not induce any significant additional screening. A more realistic model is obtained by considering a silicon dioxide dielectric between the gate plates and the ribbon, with relative dielectric permittivity εr = 3.9. However, thin vacuum layers are kept just below and above the ribbon, with a thickness of 0.3 nm, of the order of the calculated graphene/SiO2 distance.35 The impurity po- tentials and the corresponding transmission spectra are shown in Fig. 3, for VG = 0 and VG = −0.5 V. Screen- ing is enhanced at a negative gate voltage, due to the accumulation of holes in the ribbon. This has a strong effect on transmission: the conductance dip is less pro- nounced and is shifted in energy, which could compromise the presence of a clear mobility gap in a transistor con- figuration. It was verified that the effect of the gates is much larger than that of the thin air layers, in terms of the global electrostatics, and, in particular, of the screen- ing. This is the reason why in the runs for the evaluation of the transistor characteristics we will not introduce the air layers, that would make convergence of the Poisson- Schrodinger iterations much slower. i i n o s s m s n a r T 3 2 1 0 ) V e ( l a i t n e o p t y t i r u p m I 5 4 3 2 1 0 0 VG = 0 V VG = -0.5 V 0.5 Distance (nm) 1 VG = 0 V VG = -0.5 V -0.4 -0.2 0 Energy (eV) 0.2 0.4 FIG. 3. Influence of the gate voltage on the transmission spectrum of a 32-aGNR with one boron impurity placed at a distance 0.5a from the edge. The calculation was performed with silicon dioxide as gate dielectric and a thin vacuum layer surrounding the graphene sheet. Inset: influence of the gate voltage on the impurity potential. To assess the performance of complete boron-doped graphene transistors biased in a typical operating point, we have used the open-source code NanoTCAD ViDES.36 In this code, the 3D Poisson equation is solved self-consistently with the Schrodinger equation (with open-boundary conditions), within the Non-Equilibrium Green's Function (NEGF) formalism.37 Schottky con- tacts have been modeled at the GNR ends with the phe- nomenological approach described in Ref.38, including self-energies that mimic metallic contacts. The considered device structure is reported in the in- set of Fig. 1. The device is a double-gate FET and the channel is a 32-aGNR (3.81 nm wide) embedded in SiO2 dielectric. This is consistent with typical nanofabrication techniques, in which the dielectric is deposited after the creation of the drain and source contacts. Top and bot- tom field oxides are 1 nm thick, and the channel length is equal to 20 nm. In Fig. 4, we show the transfer charac- teristics of the GNR-FET for a drain-source voltage VDS = 0.1 V, for different boron doping concentrations (de- fined as the ratio of the number of boron atoms to the total number of atoms in the channel). Dopant atoms are randomly placed along the channel. Results in Fig. 4 refer to a single distribution of dopants. 4 14 12 10 ) A µ ( D I 8 6 4 2 DSV =0.1 V 0% 0.3% 0.6% 0 −1 −0.5 0 GSV (V) 0.5 1 FIG. 4. Transfer characteristics for two different boron doping concentrations (0.3% and 0.6%), for a given random distribu- tion of the dopant atoms. The transfer characteristic for the undoped device is also reported. As expected, transfer characteristics for the undoped device are symmetric with respect to VGS = VDS/2 = 0.05 V. As soon as some boron doping is considered, the p-branch of the IDS-VGS curve is suppressed, as pre- viously observed also in carbon nanotubes.33,34 The p- branch suppression increases as the doping concentration is increased. From these results, it appears that with the considered boron doping concentrations, while still not obtaining a sufficient Ion/Iof f ratio for logic appli- cations4 with the considered 32-aGNR (which, however, having a "quasi-metallic" behavior in the absence of dop- ing, represents, in this regard, a worst-case scenario), a clear onset of unipolarity occurs. Increasing boron dop- ing, while further reducing the current in the p-branch (due to holes), also leads to a degradation of the device performance above threshold. In Fig. 5(a)-(b), we show the transfer characteristics for 23 different distributions of doping atoms along the channel, for the same doping concentrations as in Fig. 4. The thick solid lines correspond to the average transfer characteristics computed over the considered statistics of the doping distribution. As can be seen, different dop- ing distributions lead to a large dispersion of the transfer characteristics, which could be reduced if several nanorib- bon devices were operated in parallel (as can be necessary in some cases in order to increase the on current). Nevertheless, we remark that the electron-hole asym- metry is clearly developing with chemical doping, and, based on the results of Ref.23, it can be strengthened by increasing the gate length up to few hundreds of nanome- ters (especially for the case with 0.3%). Indeed, longer channel lengths could result in cumulated contributions of multiple scattering effects and localization, which will even more inhomogeneously impact on transport char- acteristics (as shown in equilibrium situations23). Due to computational limitations, we are not reaching such channel lengths here. Computational hurdles, in terms of difficulties in achieving self-consistence, also limit the doping level that can currently be simulated with our method to about 0.6%. Thus, at present we cannot han- dle the extremely high doping concentrations, of the or- der of 10%, that have been reached in a recent experi- ment39 reporting unexpectedly large gap opening effects. performed for graphene nanoribbons with a width be- low 5 nm (due to the otherwise prohibitive computa- tional cost), we wish to point out that the scaling anal- ysis detailed in Refs.22,23 suggests that our main finding (electron-hole transport asymmetry) should be confirmed for ribbon widths above 10 nm, that is within the reach of conventional fabrication lithography techniques. Finally, longer ribbon lengths and low temperature are also fac- tors which are capable of widening the mobility gaps.23 5 14 a) 12 10 ) A µ ( D I 8 6 4 2 DSV =0.1 V 0% 0.3% 14 b) 12 10 8 6 4 2 DSV =0.1 V 0% 0.6% 0 −1 −0.5 0 0.5 1 0 −1 −0.5 0 0.5 1 GSV (V) GSV (V) FIG. 5. Transfer characteristics for a) 0.3% and b) 0.6% boron doping concentration, considering a statistics of 23 randomly distributed dopant atoms. The thick solid lines correspond to the average transfer characteristic. For comparison, we report also the transfer characteristic for the undoped device. II. CONCLUSION In conclusion, by coupling a DFT-based tight-binding parameterization (validated via a multi-scale approach based on ab initio techniques) with self-consistent NEGF quantum-mechanical transport simulations, we have found the onset of an electron/hole transport asymme- try and mobility gaps in boron-doped graphene nanorib- bon transistors, confirming that chemical doping could serve as a way to shape the current-voltage characteris- tics of such devices. This is just a first step, because the achieved Ion/Iof f ratio, although for a worst-case condi- tion, is still far too small for practical applications. There is however ample room for improvement and several dif- ferent approaches could be used to selectively incorporate chemical entities into graphene, thus engineering sepa- rate chemically modified areas from a single graphene layer. A doping strategy could therefore be elaborated to implement pn junctions, logic gates and complex ar- chitectures directly patterned on the same underlying graphene layer. Conventional state-of-the-art patterning methods, such as e-beam lithography, could be exploited to design superimposed material architectures (with ei- ther metallic or other semiconducting materials) onto the selectively doped and functionalized graphene-based sub- strates. Although our transistor simulations have been III. METHODS The simulation approach is based on the observation that the DFT results can be recovered performing a self- consistent Poisson/Schrodinger calculation, within the NEGF formalism, on the system, which is described us- ing a tight-binding model and a proper distribution of fixed charges. In detail, we found that, following this procedure, the tuning of a single parameter of the tight-binding model is sufficient to properly reproduce the ab initio impurity po- tential and transmission spectra. The method is inspired by what actually occurs in the DFT calculation. The to- tal charge of a carbon nucleus and of its core electrons is +4e, while it is +3e for boron. In the isolated atoms, this charge is compensated by 4 valence electrons for carbon and 3 valence electrons for boron. In graphene, 3 elec- trons per atom (including, in the case of doped graphene, the boron atoms) tend to hybridize in a "sp2" fashion, while the remaining valence electron on each carbon lies on the "π" bonding subband formed by the pz orbitals. Also the pz orbital on the boron atom is not empty, since it tends to attract an electron from the neighboring car- bon atoms, and this determines the shape of the impu- rity potential shown in Fig. 2. In the single pz orbital tight-binding approach, the "sp2" valence electrons are included in the global ionic charge of the atoms, which becomes +e for carbon and 0 for boron. Therefore this is the net charge "localized" on each atom that we have to sum to the negative charge of the pz atomic orbitals considered in the simulation. As far as the tight-binding parameters are concerned, the onsite energy for carbon atoms is 0, while the hopping energy between nearest neighbors is set to tp = −2.7 eV. The hopping energy of edge dimers is adjusted to 1.12tp in order to reproduce the small bandgap obtained in DFT for "quasi-metallic" N-aGNRs (N = 3p + 2, where p is an integer).32,40 Only the onsite energy of boron may be further ad- justed, performing a self-consistent Poisson/Schrodinger calculation for an isolated graphene ribbon with a sin- gle boron atom and comparing the obtained results with those deriving from the ab initio simulation of the same structure. For the Poisson equation the same boundary condi- tions are used in the ab initio and in the tight-binding cal- culations: Neumann boundary conditions (i.e. derivative of the potential energy, and thus electric field, constant and, in particular, zero) in the longitudinal direction, at the ends of the nanoribbon, and periodic in the transverse direction, with a separation of 2 nm between the adjacent edges of nanoribbon replicas (indeed, with a large enough separation between the replicas this is equivalent to a Neumann boundary condition). The gate stack has been treated considering a constant permittivity, the one of sil- icon oxide, between the nanoribbon and each gate, where a Dirichlet boundary condition is assumed. At each iter- ation, the eigenvalues of the Hamiltonian are computed and the Fermi level is adjusted to obtain charge neutral- ity. The local charge density is then calculated and the Poisson equation is solved. The obtained electrostatic potential is subtracted from the onsite energies and the process is repeated until self-consistency is achieved. For all of the meshes tested in this work, it was possible to fit the ab initio data simply by adjusting the boron on- site energy. A typical choice was to spread the charge of each atom in a spatial region of half-width aCC, the inter- atomic distance, and to choose a mesh interval of 0.1 nm, small enough to achieve convergence of the Poisson so- lution. This charge repartition around each atom was defined using a smooth envelope function. In this case, the best fitting was obtained for a value of the boron on- site energy of about zero, thus substantially coincident with that for carbon. 6 ACKNOWLEDGMENTS B.B. acknowledges financial support from the Juan de la Cierva Program and the FIS2008-05805 Contract of the Spanish MICINN. A.C. acknowledges the Fondation Nanosciences via the RTRA Dispograph project. We ac- knowledge the use of the software VMD41 for graphical design. This work was partly funded by the European Union under Contract No. 215752 GRAND (GRAphene- based Nanoelectronic Devices), by the French National Research Agency (ANR), in the framework of its 2009 program in Nanosciences, Nanotechnologies & Nanosys- tems (P3N2009), through the NANOSIM-GRAPHENE project no ANR-09-NANO-016-01, and by the Ital- ian Ministry for the University and Research (MIUR) through the GRANFET project (n. 2008S2CLJ9). Part of the calculations were performed with the support of the GENCI (Grand Equipment National de Calcul In- tensif) initiative. ∗ [email protected] 1 Geim, A. K.; Novoselov, K. S. The Rise of Graphene. Nat. Mater. 2007, 6, 183–191. 2 Novoselov, K. S. Nobel Lecture: Graphene: Materials in the Flatland. Rev. Mod. Phys. 2011, 83, 837–849. 3 Schwierz, F. Graphene Transistors. Nat. Nanotechnol. 2010, 5, 487–496. 4 ITRS roadmap 2009: http://www.itrs.net (accessed May 28, 2012). 5 Iannaccone, G.; Fiori G.; Macucci M.; Michetti P.; Cheli M.; Betti A.; Marconcini P. Perspectives of Graphene Nanoelectronics: Probing Technological Options with Modeling. In Proceedings of IEDM 2009, Baltimora, USA, December 7-9, 2009; IEEE Conference Proceedings: 2009; pp. 245–248. 6 Nakada, K.; Fujita, M.; Dresselhaus, G.; Dresselhaus, M. S. Edge State in Graphene Ribbons: Nanometer Size Ef- fect and Edge Shape Dependence. Phys. Rev. B 1996, 54, 17954–17961. 7 Brey, L.; Fertig, H. A. Electronic States of Graphene Nanoribbons Studied with the Dirac Equation. Phys. Rev. B 2006, 73, 235411. 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. 9 Marconcini, P.; Macucci, M. The k · p Method and Its Application to Graphene, Carbon Nanotubes and Graphene Nanoribbons: the Dirac Equation. Riv. Nuovo Cimento Soc. Ital. Fis. 2011, 34, 489–584. See also http://brahms.iet.unipi.it/supplem/reviewgraph.html (accessed May 28, 2012). 10 Han, M. Y.; Ozyilmaz, B.; Zhang, Y.; Kim, Ph. Energy Band-Gap Engineering of Graphene Nanoribbons. Phys. Rev. Lett. 2007, 98, 206805. 11 Han, M. Y.; Brant, J. C.; Kim, Ph. Electron Transport in Disordered Graphene Nanoribbons. Phys. Rev. Lett. 2010, 104, 056801. 12 Wang, X.; Ouyang, Y.; Li, X.; Wang, H.; Guo, J.; Dai, H. Room-Temperature All-Semiconducting Sub-10- nm Graphene Nanoribbon Field-Effect Transistors. Phys. Rev. Lett. 2008 100, 206803. 13 Li, X.; Wang, X.; Zhang, L.; Lee, S.; Dai, H. Chemically Derived, Ultrasmooth Graphene Nanoribbon Semiconduc- tors. Science 2008, 319, 1229–1232. 14 Wang, X.; Dai, H. Etching and Narrowing of Graphene from the Edges. Nature Chem. 2010, 2, 661–665. 15 Cresti, A.; Nemec, N.; Biel, B.; Niebler, G.; Triozon, F.; Cuniberti, G.; Roche, S. Charge Transport in Disordered Graphene-Based Low Dimensional Materials. Nano Res. 2008, 1, 361–394. 16 Cai, J.; Ruffieux, P.; Jaafar, R., Bieri, M.; Braun, T.; Blankenburg, S.; Muoth, M.; Seitsonen, A. P.; Saleh, M.; Feng, X.; et al. Atomically Precise Bottom-Up Fabrication of Graphene Nanoribbons. Nature 2010, 466, 470–473. 17 Moser, J.; Tao, H.; Roche, S.; Alzina, F.; Sotomayor Tor- res, C. M.; Bachtold, A. Magnetotransport in Disordered Graphene Exposed to Ozone: From Weak to Strong Lo- calization. Phys. Rev. B 2010, 81, 205445. 18 Leconte, N.; Moser, J.; Ordej´on, P.; Tao, H.; Lherbier, A.; Bachtold, A.; Alsina, F.; Sotomayor Torres, C. M.; Char- lier, J. C.; Roche, S. Damaging Graphene with Ozone Treatment: A Chemically Tunable Metal-Insulator Tran- sition. ACS Nano 2010, 4, 4033–4038. 19 Leconte, N.; Lherbier, A.; Varchon, F.; Ordejon, P.; Roche, S.; Charlier, J.-C. Quantum Transport in Chem- ically Modified Two-Dimensional Graphene: From Mini- mal Conductivity to Anderson Localization. Phys. Rev. B 2011, 84, 235420. 20 Cresti, A.; Lopez-Bezanilla, A.; Ordej´on, P.; Roche, S. Oxygen Surface Functionalization of Graphene Nanorib- bons for Transport Gap Engineering. ACS Nano 2011, 5, 9271–9277. 21 Bruzzone S.; Fiori G. Ab-Initio Simulations of Deformation Potentials and Electron Mobility in Chemically Modified Graphene and Two-Dimensional Hexagonal Boron-Nitride. Appl. Phys. Lett. 2011, 99, 222108. 22 Biel, B.; Blase, X.; Triozon, F.; Roche, S. Anomalous Dop- ing Effects on Charge Transport in Graphene Nanoribbons. Phys. Rev. Lett. 2009, 102, 096803. 23 Biel, B.; Triozon, F.; Blase, X.; Roche, S. Chemically In- duced Mobility Gaps in Graphene Nanoribbons: A Route for Upscaling Device Performances. Nano Lett. 2009, 9, 2725–2729. 24 Lherbier, A.; Biel, B.; Niquet, Y.-M.; Roche, S. Transport Length Scales in Disordered Graphene-Based Materials: Strong Localization Regimes and Dimensionality Effects. Phys. Rev. Lett. 2008, 100, 036803. 25 Cresti, A.; Roche, S. Range and Correlation Effects in Edge Disordered Graphene Nanoribbons. New J. Phys. 2009, 11, 095004. 26 Zhao, L.; He, R.; Rim, K. T.; Schiros, T.; Kim, K. S.; Zhou, H.; Guti´errez, C.; Chockalingam, S. P.; Ar- guello, C. J.; P´alov´a, L.; et al. Visualizing Individual Nitro- gen Dopants in Monolayer Graphene. Science 2011, 333, 999–1003. 27 Usachov, D.; Vilkov, O.; Gruneis, A.; Haberer, D.; Fe- dorov, A.; Adamchuk, V. K.; Preobrajenski, A. B.; Dudin, P.; Barinov, A.; Oehzelt, M.; et al. Nitrogen-Doped Graphene: Efficient Growth, Structure, and Electronic Properties. Nano Lett. 2011, 11, 5401–5407. 28 Lin, T.; Huang, F.; Liang, J.; Wang, Y. A Facile Prepara- tion Route for Boron-Doped Graphene, and its CdTe Solar Cell Application. Energy Environ. Sci. 2011, 4, 862–865. 29 Soler, J. M.; Artacho, E.; Gale, J. D.; Garc´ıa, A.; Jun- quera, J.; Ordej´on, P.; S´anchez-Portal, D. The SIESTA Method for Ab Initio Order-N Materials Simulation. J. Phys.: Condens. Matter 2002, 14, 2745–2779. 30 Adessi, Ch.; Roche, S.; Blase, X. Reduced Backscattering in Potassium-Doped Nanotubes: Ab Initio and Semiempir- ical Simulations. Phys. Rev. B 2006, 73, 125414. 7 31 Stern, F. Elementary Theory of the Optical Properties of Solids. In Solid State Physics. Advances in Research and Applications; Seitz F., Turnbull D., Eds.; Academic Press: New York, 1963; Vol. 15, pp 299–408. 32 Son, Y.-W.; Cohen, M. L.; Louie, S. G. Energy Gaps in Graphene Nanoribbons. Phys. Rev. Lett. 2006, 97, 216803. 33 Choi, H. J.; Ihm, J.; Louie, S. G.; Cohen, M. L. Defects, Quasibound States, and Quantum Conductance in Metal- lic Carbon Nanotubes. Phys. Rev. Lett. 2000, 84, 2917– 2920. 34 Avriller, R.; Roche, S.; Triozon, F.; Blase, X.; Latil, S. Low-Dimensional Quantum Transport Properties of Chemically-Disordered Carbon Nanotubes: from Weak to Strong Localization Regimes. Mod. Phys. Lett. B 2007, 21, 1955–1982. 35 Hossain, M. Z. Chemistry at the Graphene-SiO2 Interface. Appl. Phys. Lett. 2009, 95, 143125. 36 Fiori, G.; Iannaccone, G. Simulation of Graphene Nanoribbon Field-Effect Transistors. IEEE Electron De- vice Lett. 2007, available at "NanoTCAD ViDES," DOI: 10254/nanohub-r5116.5. http://nanohub.org/resources/vides/ (accessed May 28, 2012). 760–762. Code 28, is 37 Datta, S. Quantum transport: Atom to transistor ; Cam- bridge University Press: Cambridge, United Kingdom, 2005. 38 Guo, J.; Datta, S.; Lundstrom, M.; Anantram, M. P. To- wards Multiscale Modeling of Carbon Nanotube Transis- tors. Int. J. Multiscale Comput. Eng. 2004, 2, 257–276. 39 Tang, Y.-B.; Yin, L.-C.; Yang, Y.; Bo, X.-H.; Cao, Y.-L.; Wang, H.-E.; Zhang, W.-J.; Bello, I.; Lee, S.-T.; Cheng, H.- M.; et al. Tunable Band Gaps and p-Type Transport Prop- erties of Boron-Doped Graphenes by Controllable Ion Dop- ing Using Reactive Microwave Plasma. ACS Nano 2012, 6, 1970–1978. 40 Fujita, M.; Igami, M.; Nakada, K. Lattice Distortion in Nanographite Ribbons. J. Phys. Soc. Jpn. 1997, 66, 1864– 1867. 41 Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J. Molec. Graphics 1996, 14, 33– 38. See http://www.ks.uiuc.edu/Research/vmd/ (accessed May 28, 2012).
1807.08962	1	1807	2018-07-24T08:52:04	Non-invasive thermometer based on proximity superconductor for ultra-sensitive calorimetry	[ "cond-mat.mes-hall" ]	We present radio-frequency thermometry based on a tunnel junction between a superconductor and proximitized normal metal. It allows operation in a wide range of biasing conditions. We demonstrate that the standard finite-bias quasiparticle tunneling thermometer suffers from large dissipation and loss of sensitivity at low temperatures, whereas thermometry based on zero bias anomaly avoids both these problems. For these reasons the latter method is suitable down to lower temperatures, here to about 25 mK. Both thermometers are shown to measure the same local temperature of the electrons in the normal metal in the range of their applicability.	cond-mat.mes-hall	cond-mat	Non-invasive thermometer based on proximity superconductor for ultra-sensitive calorimetry Bayan Karimi and Jukka P. Pekola QTF Centre of Excellence, Department of Applied Physics, Aalto University School of Science, P.O. Box 13500, 00076 Aalto, Finland (Dated: July 25, 2018) We present radio-frequency thermometry based on a tunnel junction between a superconductor and proximitized normal metal. It allows operation in a wide range of biasing conditions. We demonstrate that the standard finite-bias quasiparticle tunneling thermometer suffers from large dissipation and loss of sensitivity at low temperatures, whereas thermometry based on zero bias anomaly avoids both these problems. For these reasons the latter method is suitable down to lower temperatures, here to about 25 mK. Both thermometers are shown to measure the same local temperature of the electrons in the normal metal in the range of their applicability. I. INTRODUCTION √ Thermometry forms a basis of detecting radiation quanta. As such, detection of radiant heat by a ther- mometer dates back to 1878 by S. P. Langley [1, 2]. For measuring energetic quanta, e.g., X-ray photons or ra- dioactive particles, techniques exist for a few decades [3]. As compared, for instance, to observing 6 keV X-ray photons from Mn Kα and Kβ events [4], measuring a microwave single photon with about eight orders of mag- nitude lower energy in the range of 100 µeV poses a great challenge [5 -- 7]. The energy resolution δ of a calorime- CGthST , where C denotes the heat ter reads δ = capacity of the absorber, Gth the thermal conductance to the heat bath, and ST stands for temperature noise. Among these parameters ST is directly related to the per- formance of the thermometer. The challenge is to have a √ non-invasive thermometer, operating at low enough tem- perature with noise not exceeding ST ∼ 10 µK/ Hz in order to detect typical 1 K microwave photons, e.g. in superconducting quantum circuits [5, 8]. Thermome- ter candidates for nanocalorimetry purposes include ba- sic NIS tunnel junction probes [9 -- 11], SNS Josephson junctions [12, 13], SQUIDs (superconducting quantum interference devices) [14], current noise in quantum point contacts [15, 16], Dayem bridges [17, 18], and proximity circuit QED (Quantum Electro-Dynamics) probes [19]. Here, N stands for normal metal, I for insulator barrier, and S for superconductor. The virtue of a NIS junction in a calorimeter is based on its operation in a continu- ous manner unlike a common switching detector such as a Josephson junction. In this paper we present a ther- mometer in an RF set-up with about 10 MHz bandwidth; it is a NIS tunnel junction, where by N we denote normal metal influenced by proximity superconductivity [20, 21]. The induced gap in N depends exponentially on the dis- tance from the superconductor in the normal metal. In order to make it non-invasive, we monitor the Zero Bias Anomaly (ZBA) [21 -- 24] of the junction. We present su- perior performance as compared to common QuasiParti- cle (QP) tunneling thermometer due to low dissipation and non-vanishing responsivity down to the lowest tem- FIG. 1. (a) The studied setup; schematic illustration of the transmission measurement circuit. C1 = 10.3 f F and C2 = 59.3 f F are the coupling capacitors, C = 0.2 pF and L = 100 nH the parameters of the LC resonator, and RL = 50 Ω is the transmission line impedance. (b) Colored scanning electron micrograph of sample A used in this experiment. A normal metallic island (Cu, brown) is in contact with alu- minum leads (blue) either via a clean contact or tunnel barrier (light yellow). peratures. II. DESCRIPTION OF THE SYSTEM The studied set-up in this work consists of a small prox- imitized Cu island coupled to clean superconducting Al contact and to two tunnel junctions. A schematic illus- tration of the measurement setup and a Scanning Elec- tron Microscope (SEM) image of Sample A are shown 8 1 0 2 l u J 4 2 ] l l a h - s e m . t a m - d n o c [ 1 v 2 6 9 8 0 . 7 0 8 1 : v i X r a (a)(b)injectorthermometer2 in Fig. 1a and b, respectively. The contact to the right in Fig. 1b is the thermometer junction, the one to the left is an auxiliary tunnel contact (injector), and the one in the middle is a clean superconductor contact. The structure is fabricated on top of an oxidized silicon sub- strate by Electron-Beam Lithography (EBL) combined with three-angle shadow evaporation. We present data on two samples. In Sample A (B) the resistance of the thermometer tunnel junction is RT = 8 kΩ (20 kΩ) and the clean contact is d = 500 nm (1 µm) away from it. The junction area in both samples is A (cid:39) 0.010 µm2, yielding specific resistance RTA = 80 Ωµm2 (200 Ωµm2) for Sample A (B). In both samples the thickness of the two Al layers (blue color in Fig. 1b) is 20 nm for both, and 35 nm for Cu (brown). In this paper, we focus on the SNIS configuration which is used as a RF thermometer. The SN clean con- tact acts as a heat mirror and fixes the electric poten- tial of the island; it is directly grounded at the sample stage. Importantly, this contact induces proximity su- perconductivity at the thermometer junction. In order to obtain fast temperature readout, the superconductor lead of the junction is embedded in an LC resonator, which is connected to input and output rf-lines via cou- pling capacitors C1 and C2, schematically shown in Fig. 1a. The dc voltage of the thermometer Vth is connected to a bias-tee and a small parallel resistor fixed at the printed circuit board of the sample box using a resistive thermocoax dc-line (not shown in Fig. 1) [8]. The LC- resonator is made of Al with the thickness of 100 nm and is fabricated by EBL and one angle metallization; it is placed on a separate chip. Using elementary analysis for the circuit in Fig. 1a, and assuming almost all the signal applied on the left is reflected, the ratio of voltages V1 and V2, s ≡ 2V2/V1, is given by s(ω) = − {1 − ω2[L(C + C1 + C2) + LC2RL/R]} + iω[L/R + RLC2 − ω2L(C1 + C)RLC2] i2ω3RLC1C2L , (1) where R is the inverse of the differential conductance dI/dV of the junction and RL = 50 Ω is the transmission line impedance. For RL/R (cid:28) 1, the resonance frequency 0 ≈ 1/[L(C1 + C2 + f0 = ω0/2π (cid:39) 640 MHz is given by ω2 C)]. Then at resonance 1 s(ω0) = − 2C1 C2 2 )−1. Thus, 1 + R0dI/dV , (2) for dI/dV → 0, where R0 = (ω2 0RLC 2 s(ω0) → −2C1/C2, i.e. it obtains the value given by the ratio of the input and output couplings. Measured in dBm referenced to P0 = 1 mW, we obtain the transmis- sion in the form S21(ω0) = 10 lg( 1 s(ω0)2 V 2 4RLP0 ) = S0 − 20 lg(1 + R0 dI dV ), (3) √ where S0 = 20 lg(C1V1/(C2 RLP0)) is a constant off- set which in the actual set-up includes also the attenua- tion and amplification in the lines. For low conductance, dI/dV (cid:28) R−1 0 we may linearize the relation between S21 and dI/dV into the form FIG. 2. The transmission S21 measured against the drive frequency f at T = 170 mK and at −120 dBm power. The different curves with the overall trend from top to bottom correspond to thermometer dc bias voltages Vth ranging from 0 to 170 µV in 5 µV intervals. The inset shows a zoom of a similar measurement at T = 30 mK where the dashed curves from bottom to top are for Vth = 0, 5, 10, 20 µV and the solid curves from top to bottom are for Vth = 30, 90, 105, 112, 120 µV. We attribute the tiny frequency shift to finite Josephson inductance. dI dV = ln(10) 20R0 (S0 − S21(ω0)), (4) i.e. S21 measures the negative of the differential conduc- tance of the junction. Figure 2 shows the frequency dependence of S21 of Sample A, measured between the right and middle con- tacts of Fig. 1b, around the resonance frequency f0, at the bath temperature T = 170 mK at different bias volt- ages Vth of the thermometer. (We use the symbol Vth in- terchangably to V when we discuss the actual thermome- ter junction.) In general, the bias voltage determines the differential conductance dI/dV of the junction in a way to be described below. It is obvious based on the fig- ure that at large biases the resonance line becomes wider where r = (cid:112)π∆/(2kBT )R0/RT . An estimate of R0 3 can be obtained from the measured dI/dV versus the bias V , because at V → ∞, dI/dV → RT −1, the re- sistance of the tunnel junction. On the other hand at low bias R0dI/dV (cid:28) 1. Combining these results we have R0/RT (cid:39) 10∆S/20 − 1, where ∆S = S21(V ≈ 0) − S21(V (cid:29) ∆/e). In order to have pure NIS configuration, the direct superconducting contact is placed 1 µm away from the thermometer junction in Sample B. In practice, this leads to vanishing proximity effect at the thermometer junc- tion. The transmission S21 as a function of voltage bias of the thermometer for a set of bath temperatures Tbath is presented in Fig. 3a. In all these measurements bath temperatures are obtained based on primary Coulomb Blockade Thermometer (CBT) [28]. Figure 3b shows the data points extracted from the transmission curves at fixed voltage biases for different bath temperatures. The solid lines show fits to the corresponding experimental data based on Eq. (7). It is clear that all the experi- mental sets match the calculated ones with r and S0 as fitting parameters. The QP thermometer loses its sensi- tivity at low T demonstrated by the vanishing responsiv- ity R ≡ dS21/dT. It is advantageous for calorimetry to work at as low temperature as possible. This is because the energy resolution of an ideal calorimeter limited by fundamen- kBCT ∝ tal thermal fluctuations is given by δ = T 3/2 [30]. Therefore one would hope to have a sensitive and reliable thermometer down to the lowest tempera- tures reachable by a standard dilution refrigerator. For this purpose we next present and analyze a different con- cept, which avoids the vanishing responsivity at low T . √ IV. ZERO BIAS ANOMALY (ZBA) THERMOMETER In Sample A we place the clean superconducting con- tact closer to the thermometer junction (d = 500 nm). This gives the proximity effect to the normal metal with superconducting properties extending all the way to the position of the thermometer NIS junction. In this case a zero bias anomaly (ZBA) arises as shown in Fig. 4a. This is to be compared to the measurement in the same set-up in Fig. 3a, where on Sample B we observe no structure in the low bias region. Figure 4b exhibits transmission S21 results at different RF power levels on Sample A. It is vivid that due to the narrow ZBA feature the result is sensitive to the applied power. This ZBA offers us a sensitive probe of tempera- ture here down to the base temperature of the measure- ment (25 mK) as shown in Fig. 4c, where temperature versus the zero-bias S21 is presented. This probe is non- dissipative (zero bias) and sensitive (no saturation) at low temperatures, in contrast to the QP thermometer. It is worth pointing out that the responsivity R of ZBA FIG. 3. Measured characteristics of the RF thermometer in the quasiparticle (QP) regime (Sample B). (a) The transmis- sion S21 as a function of the dc bias voltage Vth. The mea- surement is performed at temperatures 301, 263, 225, 182, 143, 113, 89, 68, 56, and 50 mK from bottom to top. (b) Temperature calibrations at Vth = 156 µV measured at -100 dBm of the RF signal in the main panel and at various bias points in the inset. The lines are based on Eq. (7). due to increasing dissipation (larger dI/dV of the junc- tion). The inset shows a zoom-up of similar data taken at T = 30 mK, demonstrating a negative frequency shift of about 200 kHz when biasing the junction away from Vth = 0. This shift is due to the Josephson inductance of the NIS junction at zero bias. The Josephson junc- tion with critical current Ic introduces a parallel to L inductance LJ = Φ0/(2πIc) leading to a frequency shift δf0/f0 = L/(2LJ ) for LJ (cid:29) L. Here, Φ0 = h/2e is the superconducting flux quantum. The measured frequency shift of about 200 kHz would then imply Ic ∼ 5 pA for Sample A. III. QP THERMOMETRY (cid:90) ∞ Measuring current carried by single electrons (QPs) in a NIS junction has been considered for measurements of power in ultra-sensitive nanobolometers [25, 26]. For an ideal low transparency junction biased at voltage V , the expression for QP current reads 1 I = 2eRT dE nS(E){fN (E − eV ) − fN (E + eV )}, −∞ (5) √ where RT is the resistance of the tunnel junction, nS(E) = E/ E2 − ∆2 the normalized superconducting density of states for E > ∆ and nS(E) = 0 for E > ∆, and fN (E) = (1 + exp(βE))−1 the Fermi distribution in the normal metal at temperature T = (kBβ)−1. Here ∆ denotes the superconducting gap. Far below the critical temperature, T (cid:28) Tc, and for low biases we have where I0 = Eqs. (3) and (6) into (1 − eV ∆ ∆ kB T = ) I ≈ I0e−(∆−eV )/kB T , √ 2πkBT ∆/(2eRT ). Then we can combine (6) 1 20 (S0−S21) − 1) ln r − ln(e ln 10 , (7) 4 FIG. 5. Measurement of electron temperature Te of the ab- sorber under non-equilibrium conditions produced by apply- ing voltage Vinj across the injector junction. (right) Voltage dependence of Te demonstrates cooling, measured with dif- ferent biasing and power levels of the thermometer. (left) Extracted minimum temperature of Te as a function of bath temperature Tbath. Dashed line denotes Tmin = Tbath. The symbols in the main figure refer to: • QP, Vth = 90 µV, -110 dBm, • QP, Vth = 90 µV, -110 dBm, • QP, Vth = 105 µV, -110 dBm, • QP, Vth = 120 µV, -110 dBm, • QP, Vth = 0, -120 dBm, • QP, Vth = 0, -125 dBm, (cid:70) ZBA, Vth = 20 µV, -120 dBm, (cid:70) ZBA, Vth = 20 µV, -125 dBm, (cid:70) ZBA, Vth = 0, -115 dBm, (cid:70) ZBA, Vth = 0, -120 dBm, (cid:70) ZBA, Vth = 0, -125 dBm. as a QP thermometer indicates close to identical tem- peratures over the whole Vinj bias range. In the main panel of Fig. 5, we have collected measurements with different bias points Vth of the thermometers, while mea- suring the minimum temperatures Tmin at Vinj ∼ ∆/e versus bath temperature Tbath. QP thermometer at var- ious bias points spans over the range from (cid:38)100 to 400 mK, whereas ZBA thermometer can be used down to the minimum temperature of the cryostat in this experiment. We observe that the measurements form a continuous set of data showing the optimum cooling around 200 mK. Based on the collapse of the different sets of data in Fig. 5 we conclude that all these thermometers measure, in a consistent way, the temperature of the electrons in the island. V. DISCUSSION AND CONCLUSIONS Here we discuss feasibility of measuring single energy quanta by ZBA thermometer, and compare its perfor- mance to QP method. We show quantitatively that the main advantages of ZBA based calorimetry are its low operation temperature and ultra low power dissipation. The challenge, on the other hand, is that only low input power levels are feasible as the ZBA peak is very narrow in Vth. Non-invasiveness of ZBA - An important key for a nanocalorimeter in order to be able to detect single quanta like photons with energy ∼ 100 µV, is to be non- FIG. 4. The transmission of the proximitized junction of Sam- ple A in the zero bias anomaly (ZBA) regime, S21 versus Vth, for (a) a set of bath temperatures Tbath in the range of 27 to 398 mK and fixed power −130 dBm (b) few power levels at T = 30 mK. (c) Temperature calibration curves of the trans- mitted power at zero bias measured at -140 dBm of the RF signal in the main panel, and at different power levels −130, −125, and −120 dBm in its inset. thermometer at T < 200 mK is (cid:39) 0.06 dB/mK, which clearly exceeds R (cid:39) 0.01 dB/mK of the QP-thermometer in the temperature range of its applicability. Tempera- ture calibrations measured with different powers, shown in the inset of Fig. 4c, demonstrate that measuring with larger power (-120 dBm, -125 dBm) leads to saturation of S21 at low temperatures. One can also see that ZBA is composed of several peaks with origin in supercurrent, and possibly in multiple Andreev reflection due to the relatively high transparency of the junction [31 -- 34]. The back-bending at T > 300 mK is due to QP current. In order to compare different thermometers in Sample A, we manipulate the electronic temperature Te of the N island by applying a bias voltage Vinj across the auxil- iary junction, denoted "injector" in Fig. 1b with tunnel resistance Rinj (cid:39) 50 kΩ. The influence of this bias is the feature depicted in the inset of Fig. 5. In all curves the electronic temperature drops at Vinj ∼ ∆/e due to the well-known quasiparticle cooling effect [27]. Operating the thermometer at three different conditions including zero with two different powers and several bias values -60-40-200204060-35-34-33-32-31-30S21 (dB)Vth (V)-10-50510-44-42-40-38-36-34-32-30-28S21 (dB)Vth (V) -140 dBm -130 dBm -125 dBm -120 dBm -115 dBm -110 dBm-44-42-40-38-36-34-32050100150200250300350400-140 dBmT (mK)S21 (dB)-35-34-33-320100200300400 -130 dBm -125 dBm -120 dBm(a)(b)(c)050100150200160180200220240260280300320Te (mK)Vinj (V) ZBA, Vth=0, -120 dBm ZBA, Vth=0, -125 dBm QP, Vth=90 V, -110 dBm QP, Vth=105 V, -110 dBm QP, Vth=120 V, -110 dBm050100150200250300350400050100150200250300350400QP ZBATmin (mK)Tbath (mK)5 (cid:113) S Q/G2 req = δi/(V√ th =(cid:112)2kB/(5ΣV)T −1, where, accord- Sample A, and 0.0005 µm3 that represents a techni- cally realistic tiny absorber (called Sample Opt), respec- tively. In this figure, the black line in both panels 6a and 6b represents fundamental temperature fluctuations NET0 = ing to fluctuation dissipation theorem, S Q = 2kBT 2Gth is the heat current noise in equilibrium. The two other lines denote the required noise-equivalent temperature 5γΣ) T −5/2 for i = e, ph, to observe NETi quanta of energy for electrons (red line) and photons (blue line). For our estimations, we use the well-known expressions for thermal conductance to the phonon bath Gth = 5ΣVT 4, and C = γVT for heat capacity of the normal metal. Here, Σ denotes the electron-phonon con- stant, V the volume of the island, and γ refers to the Sommerfeld constant for electrons in metal. The shaded areas delineate the favourable regimes for detecting these particles. The upper boundary is given by the required NETi req and the lower bound (black line) represents the fundamental temperature fluctuations NET0. The table presented in Fig. 6c shows examples of these estimates under four different conditions. The first row presents Sample B with QP thermometry around its lowest op- eration temperature (130 mK). The estimates for ZBA thermometer of Sample A at two different temperatures (50 and 25 mK) are given in the second and third rows, respectively. Here, in the absence of precise temperature calibration for ZBA, the two rows represent the conser- vative and optimistic estimates of the actual base tem- perature using ZBA. The fourth row indicates a further optimized sample, Opt, with experimentally feasible tar- get parameters. The fifth and sixth columns in the table, NETe req, demonstrate that, it is next to im- possible to detect a single electron e or photon ph by QP thermometer. Yet using ZBA at lower T , the requirement for NET is relaxed by one to two orders of magnitude. The typical NET in the present measurement is around √ ∼ 30 µK/ Hz based on ZBA thermometer (including the amplifier noise). Thus the current NET is just about to be sufficient to detect 2.5 K×kB. In the last two columns we present the expected signal to noise ratio (S/N)i for different quanta, i = e, ph, demonstrating the possibility of detecting them by ZBA at low temperatures and with small absorber. req and NETph Finally, we comment briefly on how to optimize the ZBA thermometer in future. The critical current Ic of the NIS junction is expected to increase exponentially when the distance of the clean contact d from the junc- tion decreases [35, 36]. Therefore higher responsivity is expected for smaller values of d: in practice d can be decreased down to ∼ 50 nm from the current 500 nm. Moreover, the frequency shift of the resonance due to Josephson inductance will also be enhanced in this case, giving an extra boost in the sensitivity of the ZBA ther- mometer. With the low operation temperature (∼ 10 mK) and the proposed improvements in design, the ZBA calorimeter can detect single microwave photons. FIG. 6. Requirements for detecting the heat produced by sin- gle electron δe = 2.5 K×kB and single photon δph = 1K×kB quanta. Dependence of various noise-equivalent temperatures NET on bath temperature Tbath for two different volumes of copper absorber, (a) 0.0025 µm3 and (b) 0.0005 µm3. In both panels, black line demonstrates the fundamental tem- perature fluctuations NET0, red line shows the required noise- equivalent temperature for electrons NETe req and blue lines for photons NETph req. The shaded areas indicate the feasible regimes where the corresponding quanta can be observed. In panel (c), we present concrete examples referring to the Sam- ples A, B and Opt with a small volume yet experimentally feasible. QP and ZBA thermometry are used for samples B and A, respectively, at the minimum temperature 130 mK for QP and two different temperatures for ZBA. The last two columns give the corresponding signal to noise ratio for de- tecting electrons (S/N)e and photons (S/N)ph. The parame- ters used for evaluating the present results in panels a-c are: Σ = 2 × 109 WK−5m−3 and γ = 70 JK−2m−3. invasive [29]. Using Eq. (6) and considering ∆/e = 200 µV for superconducting gap of Al, V = 0.8∆/e and T = 130 mK as parameters for Sample B, QP ther- mometer injects ∼ 15 aW heat. On the other hand, the corresponding power for ZBA thermometer in Sample A (using Eq. (4) and −140 dBm for applied RF power) is ∼ 4 zW at 25 mK, i.e., ∼ 106 times less than in the QP thermometer. √ Sensitivity at low T - We start by estimating the re- √ ST which we de- quired noise-equivalent temperature note NETreq in µK/ Hz. For energy we choose δe = 2.5 K×kB for measurements described in [29] to de- tect an electron tunneling over the superconducting gap of Al, and δph = 1 K×kB for a 20 GHz single mi- crowave photon. Figures 6a and 6b demonstrate the feasibility of measuring the energy deposited by sin- gle electrons and single microwave photons by the en- visioned calorimeter with copper absorber. Both the panels present various predicted NET-values as func- tions of the operation temperature for two different vol- umes of the absorber, 0.0025 µm3 which is the current 20406080100101001000NET (K/Hz)T (mK)102040608010010100100010000T (mK)10(a)(b)(c)ACKNOWLEDGMENTS This work was funded through Academy of Finland grants 297240, 312057 and 303677 and from the Euro- pean Union's Horizon 2020 research and innovation pro- gramme under the European Research Council (ERC) programme and Marie Sklodowska-Curie actions (grant 6 agreements 742559 and 766025). We acknowledge the fa- cilities and technical support of Otaniemi research infras- tructure for Micro and Nanotechnologies (OtaNano). We acknowledge J. T. Peltonen and E. T. Mannila for techni- cal support, and F. Brange, P. Samuelsson, D. Golubev, A. Mel'nikov, O-P. Saira, K. L. Viisanen, W. Belzig, I. Khaymovich and P. Muratore-Ginanneschi for useful dis- cussions. [1] S. P. Langley, The bolometer, Nature 57, 620-622 (1898). [2] P. L. Richards, Bolometers for infrared and millimeter waves, J. Appl. Phys. 76, 1 (1994). [3] Christian Enss (Ed), Cryogenic particle detection (Springer-Verlag Berlin, Heidelberg, 2005). [4] D. McCammon, S. H. Moseley, J. C. Mather, and R. F. Mushotzky, Experimental tests of a single-photon calorimeter for x-ray spectroscopy, J. Appl. Phys. 56, 1263 (1984). [5] J. P. Pekola, P. Solinas, A. Shnirman, and D. V. Averin, Calorimetric measurement of work in a quantum system, New J. Phys. 15, 115006 (2013). [6] Kunihiro Inomata, Zhirong Lin, Kazuki Koshino, William D. Oliver, Jaw-Shen Tsai, Tsuyoshi Yamamoto, and Yasunobu Nakamura, Single microwave-photon de- tector using an artificial Λ-type three-level system, Na- ture Communications 7, 12303 (2016). [7] G. Oelsner, C.K. Andersen, M. Reh´ak, M. Schmelz, S. Anders, M. Grajcar, U. Hubner, K. Mølmer, and E. Il'ichev, Detection of Weak Microwave Fields with an Underdamped Josephson Junction, Phys. Rev. Appl. 7, 014012 (2017). [8] K. L. Viisanen and J. P. Pekola, Anomalous electronic heat capacity of copper nanowires at sub-kelvin temper- atures, Phys. Rev. B 97, 115422 (2018). [9] D. R. Schmidt, C. S. Yung, and A. N. Cleland, Nanoscale radio-frequency thermometry, Appl. Phys. Lett. 83, 1002 (2003). [10] Francesco Giazotto and Mar´ıa Jos´e Mart´ınez-P´erez, The Josephson heat interferometer, Nature 492, 401 (2012). [11] S. Gasparinetti, K. L. Viisanen, O.-P. Saira, T. Faivre, M. Arzeo, M. Meschke, and J. P. Pekola, Fast electron thermometry towards ultra-sensitive calorimetric detec- tion. Phys. Rev. Applied 3, 014007 (2015). [12] J. Govenius, R. E. Lake, K. Y. Tan, and M. Mottonen, Detection of Zeptojoule Microwave Pulses Using Elec- trothermal Feedback in Proximity-Induced Josephson Junctions, Phys. Rev. Lett. 117, 030802 (2016). [13] L. B. Wang, O.-P. Saira, and J. P. Pekola, Fast thermom- etry with a proximity Josephson junction, Appl. Phys. Lett. 112, 013105 (2018). [14] Dorri Halbertal, Moshe Ben Shalom, Aviram Uri, Kousik Bagani, Alexander Y. Meltzer, Ido Marcus, Yuri Myasoe- dov, John Birkbeck, Leonid S. Levitov, Andre K. Geim, Eli Zeldov, Imaging resonant dissipation from individual atomic defects in graphene, Science 358, 1303 (2017). [15] Z. Iftikhar, A. Anthore, S. Jezouin, F.D. Parmentier, Y. Jin, A. Cavanna, A. Ouerghi, U. Gennser, and F. Pierre, Primary thermometry triad at 6 mK in mesoscopic cir- cuits, Nature Communications 7, 12908 (2016). [16] Mitali Banerjee, Moty Heiblum, Amir Rosenblatt, Yuval Oreg, Dima E. Feldman, Ady Stern and Vladimir Uman- sky, Observed quantization of anyonic heat flow, Nature 545, 22052 (2017). [17] M. Foltyn and M. Zgirski, Gambling with Superconduct- ing Fluctuations, Phys. Rev. Appl. 4, 024002 (2015). [18] Federico Paolucci, Giorgio De Simoni, Elia Strambini,y Paolo Solinas, and Francesco Giazotto, Ultra-efficient superconducting Dayem bridge field-effect transistor, arXiv:1803.04925 (2018). [19] O.-P. Saira, M. Zgirski, K.L. Viisanen, D.S. Golubev, and J.P. Pekola, Dispersive thermometry with a Joseph- son junction coupled to a resonator, Phys. Rev. Appl. 6, 024005 (2016). [20] B. Pannetier and H. Courtois, Andreev reflection and proximity effect, J. Low Temp. Phys. 118, 599 (2000). [21] S. Gueron, H. Pothier, Norman O. Birge, D. Esteve, and M. H. Devoret, Superconducting Proximity Effect Probed on a Mesoscopic Length Scale. Phys. Rev. Lett. 77, 3025 (1996). [22] A. Kastalsky, A. W. Kleinsasser, L. H. Greene, R. Bhat, F. P. Milliken, and J. P. Harbison, Observation of Pair Currents in Superconductor-Semiconductor Con- tacts, Phys. Rev. Lett. 67, 3026 (1991). [23] H. A. Nilsson, P. Samuelsson, P. Caroff, and H. Q. Xu, Supercurrent and Multiple Andreev Reflections in an InSb Nanowire Josephson Junction, Nano Lett. 12, 228 (2012). [24] Alberto Ronzani, Carles Altimiras, and Francesco Gi- azotto, Highly Sensitive Superconducting Quantum- Interference Proximity Transistor, Phys. Rev. Appl. 2, 024005 (2014). [25] M. Nahum and John M. Martinis, Ultrasensitive-hot- electron microbolometer, Appl. Phys. Lett. 63, 3075 (1993). [26] A. V. Gordeeva, V. O. Zbrozhek, A. L. Pankratov, L. S. Revin, V. A. Shamporov, A. A. Gunbina, and L. S. Kuzmin, Observation of photon noise by cold-electron bolometers, Appl. Phys. Lett. 110, 162603 (2017). [27] Francesco Giazotto, Tero T. Heikkila, Arttu Luuka- nen, Alexander M. Savin, and Jukka P. Pekola, Oppor- tunities for mesoscopics in thermometry and refrigera- tion: Physics and applications, Rev. Mod. Phys. 78, 217 (2006). [28] J. P. Pekola, K. P. Hirvi, J. P. Kauppinen, and M. A. Paalanen, Thermometry by Arrays of Tunnel Junctions, Phys. Rev. Lett. 73, 2903 (1994). [29] F. Brange, P. Samuelsson, B. Karimi, and J. P. Pekola, Nanoscale Quantum Calorimetry with Electronic Tem- perature Fluctuations, arXiv:1805.02728 (2018). [30] E. M. Lifshitz and L. P. Pitaevskii, Statistical Physics, Part 1, 3rd ed. (Pergamon press, Oxford 1980). [31] T. M. Klapwijk, G. E. Blonder, and M. Tinkham, Expla- nation of subharmonic energy gap structure in supercon- ducting contacts, Physica B & C 109, 1657 (1982). [32] Athanassios Bardas and Dmitri V. Averin, Elec- tron transport in mesoscopic disordered superconductor normal-metal superconductor junctions, Phys. Rev. B 56, R8518 (1997). [33] E. Scheer, P. Joyez, D. Esteve, C. Urbina, and M. H. De- voret, Conduction channel transmissions of atomic-size aluminum contacts, Phys. Rev. Lett. 78, 3535 (1997). [34] J. C. Cuevas, J. Hammer, J. Kopu, J. K. Vil- jas, and M. Eschrig, Proximity effect and multiple Andreev reflections in diffusive superconductornormal- metalsuperconductor junctions, Phys. Rev. B 73, 184505 (2006). [35] M. Tinkham, Introduction to superconductivity, 2nd ed. (Dover New York, 1996). [36] P. Dubos, H. Courtois, B. Pannetier, F. K. Wilhelm, A. D. Zaikin, and G. Schon, Josephson critical current in a long mesoscopic S-N-S junction, Phys. Rev. B 63, 064502 (2001). 7
1004.0827	1	1004	2010-04-06T11:35:00	The CNT/BCN/CNT structure (zigzag type) as a molecular switch	[ "cond-mat.mes-hall" ]	Using a tight-binding model and some well-known approaches and methods based on Green's function theory and Landauer formalism, we numerically investigate the conductance properties and I-V characteristics of zigzag single-walled BCN alloy nanotube in the CNT/BCN/CNT structure, where nanocontacts are considered as zigzag single-walled carbon nanotubes. Our calculations show that any increasing in considerably give rise to the enhancing of the conductance of the system. With our system characteristics, this system can be a possible candidate for a nanoelectronic switching device.	cond-mat.mes-hall	cond-mat	The CNT/BCN/CNT structure (zigzag type) as a molecular switch H Milani Moghaddam Department of Physics, University of Mazandaran, Babolsar, Iran E-mail: [email protected] ; [email protected] Abstract. Using a tight-binding model and some well-known approaches and methods based on Green's function theory and Landauer formalism, we numerically investigate the conductance properties and I-V characteristics of zigzag single-walled BCN alloy nanotube in the )0,( n CNT/BCN/CNT structure, where nanocontacts are considered as zigzag single-walled )0,( n carbon nanotubes. Our calculations show that any increasing in n considerably give rise to the enhancing of the conductance of the system. With our system characteristics, this system can be a possible candidate for a nanoelectronic switching device. PACS Nos: 73.63.Fg; 73.22.-f; 73.63.-b Keywords: molecular wire, carbon nanotube, boron carbonitride alloy nanotube, Green's function 1. Introduction The study and manipulation of matter on the nanometer scale is a thriving area of research, with profound implications for technology (e.g. nanoelectronics, nanostructured materials, nanobiology). So devices have been designed in such way that a molecule is sandwiched between two electrodes (metallic or organic). So far, many theoretical works have been developed and models have been proposed to represent the molecular system and the reservoirs. In most model systems, the molecular wires connected to two semi-infinite surfaces [1, 2] or to two semi-infinite ‘rods’ [3] or to clusters [4, 5]. In some recent works, simple molecular wires have been connected to two semi-infinite carbon nanotubes (CNTs) [6-9]. The significantly improved switching characteristics of short organic FETs with metallic CNT electrodes over those with metal electrodes are attributed to the excellent electrostatics attainable with a nanotube electrode geometry [10, 11]. CNTs and hetero-materials including borons (B) and nitrogens (N) have been attracting much attention both in the fundamental science and in the interests of application to nanotechnology devices [12, 13, 14]. Though CNTs have generated great interest in use of a broad range of potential nanodevices for their unique structural and electronic properties [15, 16], other nanotubes such as boron carbonitride (BCN) alloy nanotubes are interesting in their own right and may be able to offer different possibilities for technological applications that CNTs cannot provide [17]. It is calculated that heterojunctions of B–C–N nanotubes are largely independent of the radius, helicity, multiplicity, or degree of perfection of the constituting nanotubes [17], though depending on their chirality, CNTs can be metallic or semiconducting [18]. BCN alloy nanotubes have been successfully synthesized by electrical arc discharge [19-23], pyrolysis [24, 25] and laser ablation [26] methods. Of all these properties, BCN alloy nanotubes are of especially importance for nanodevice applications. 1 = . ) ; g g = In this letter, following these interests and towards the modeling of a molecular wire, we use a model in which a zigzag single-walled BCN alloy nanotube is connected to two semi-infinite zigzag single-walled CNTs which are considered as nanocontacts. In this model of BCN alloy nanotube, B atoms are replaced by C atoms in the middle unit cell of the BN nanotube. We numerically investigate the conductance properties and I-V characteristics of BCN )0,( n )0,(n alloy nanotube in the CNT/BCN/CNT structure. The zigzag CNTs (on which we will l3=n concentrate here) with ( l integer) will be metallic. The model and description of the methods for investigation of conductance properties of the molecular wire is introduced in Sec. (2). The results and discussion are presented in Sec. (3) followed by a summary in Sec. (4). 2. Methodology The most commonly used computational schemes for calculating the (coherent) conductance g are the Landauer theory [27] and the Green’s function formalism [28-30]. The conductance g at zero temperature is simply proportional to the transmission coefficient, T(E) , for injected electrons at the Fermi energy, 2 e 2 ETg ( (1) 0 0 h We model the transport problem by dividing the system in three parts (Fig. 1): Two semi- infinite leads (L) and (R) with bulk electronic structure are connected to a finite region called device (D). the D region contains the scattering region (S) where the potential landscape for the electrons deviates from that in the leads, and a finite part of each lead (-1) and (1). The leads’ parts inside D are chosen sufficiently big such that the leads only couple to that part of the device. The Hamiltonian matrix is divided into submatrices as follows H H 0     H H H   DR DL   H H 0   RD R where it has been assumed that the electrodes do not directly interact which each other H H == 0 ) as the D region is sufficiently big, or rather the electrode parts inside D as ( LR RL mentioned above . The Green’s function matrix is given by G G G   LD   G G G   DR   G G G   RL RD R We could calculate the above submatrices of the Green’s function by solving the equation EGHE ( .1) ( ) − = (4) This results in the following expression for the Green’s function in the device region 1− EG E E HE . ( ) ( ) ( ( )) Σ− − = Σ− (5) D L R D (ELΣ ) (ERΣ ) and describe the effect of the two leads on the The complex self-energies electronic structure of the device and are given by the Green’s functions of the semi-infinite 1) 1) − − Eg Eg HE HE ( ( ) ( ) ( − = = − projected into the device and isolated leads R R L L DLH and RDH region by the coupling of the leads to the device HEgH E HEgHE = ) ( ( ) ( ) ) ( Σ Σ = and DR R RD L DL (3) (2) (6) G = H = LD D L DL , . LR D L L LD R . 2 − j j ,1+ j j ,1 + j , j 1 + jt It can be shown that the Landauer transmission at a certain energy can be expressed in the Green’s function formalism by the Caroli expression [31] + EGE EGE ET Tr ( ) ) ( ( [ ( ) ) ( .)] Γ = Γ (7) L R D D (ELΓ ) (ERΓ ) and are minus the imaginary part of the leads’ self- The coupling matrices energy. +Σ− i E E E ( ( ) ( ) ( .)) Σ= Γ (8) RL RL RL ( ( ) ) ( ) Also the electronic density of states (DOS) of the device is given as [31] Im1 { })] DOS(E) = D E ( Tr[G (9) π The CNTs and BCN alloy nanotube are modeled within the tight-binding Hamiltonian with only one π-orbital per atom [32, 33]. This Hamiltonian can describe reasonably well the band structure of a nanotube especially near the Fermi level, Fε , which is zero in this case ∑ ∑ + + + c H cc c t cc ( ,) ε = − + (10) j j j j 1 + j j jε and is the annihilation(creation) operator of an electron at the j site. where ( + c j c ) j respectively, represent the on-site energy and the nearest-neighbor hopping integral. 3. Results and discussion Based on the formalism described in section 2, we have investigated the electronic conduction properties of CNT/BCN/CNT structure for several typical CNT and BCN alloy nanotube with different features. In our calculations, the onsite energy at B atoms and that of N atoms are assumed to be +2.33 eV and -2.50 eV , respectively, if being measured from the C onsite ( t t eV 7.2−=− t eV t CC eV 3−=− 81.2−=− and , . Also we shall assume energy ( )0 =Cε CB NB stands for the hopping integral). Fig. 2 illustrates the electronic density of states (DOS) of CNT/BCN/CNT system. The plots show that the presence of C atoms in BCN alloy nanotube induces the electronics states within the band gap and causes a large enhancement in the conductance of the system. Pure BN nanotubes are wide band gap semiconductors with a band gap of nearly 5 eV . Also the electronic transmission probability through the system is shown in Fig. 3, panels (a)-(c) which correspond to Fig. 2, panels (a)-(c), respectively. Our results suggest that any increasing in n considerably give rise to the enhancing of the conductance of the system. From the most experimentally observed carbon nanotube sizes, there is a tiny gap in the zigzag nanotube types which arises from curvature effects [34]. Thus any increasing in the tube diameter give rises to decrease of the gap as /1 R [34,35] and causes a large enhancement in the 2 conductance of the system. In all of above calculations no voltage drop was considered across the system. However, in order to study the behavior of the system in the presence of an applied voltage we use µ R ∫ VI he Td f f ) ( ( )/2( [)( )( )] εε ε ε = − µ L (VI ) )(εT where , , and , respectively, eV 2/ RLf )1 )] ( (exp[ )( 1 = εµ ± −+ µεβ ε − = F RL / RL / / represent the total current through the device in the V bias voltage, the transmission coefficient, the Fermi-Dirac distribution function and the chemical potential. β is equal to Tk B/1 ( Bk and are Boltzmann constant and temperature of the reservoir responsible T K o4= for injecting the electrons into the contacts, respectively). e and h correspond to the electron’s charge and Planck’s constant in that order. (11) L R 3 The results of these studies are presented in Fig. 4. The positions of the current jumps are determined by the position of the reservoir Fermi levels and by the details of the molecular structure of the wire. The appearance of jumps in the molecular Eigenvalue staircase is a conspicuous feature of the molecular wire I-V curve. With our system characteristics, one can construct a switching device. It is clear immediately from Fig. 4 that by changing the voltage (e.g. by nearly 0.49 Volt) the device acts as a switch, V Volt 49.0≈ ) in the and is turned from “OFF” to “ON”. According to Fig. 4, the first jump ( I-V characteristics is independent of the nanotubes’ diameter but the current in that as the tube radius(R) increases, considerably increases (Fig. 5). We explain the origin of the staircase shape of the I-V curve as follows. For small V, there are no molecular resonances between the Fermi levels of the two electrodes, and the current is small. As V increases, the energies of the wire orbitals decrease and eventually one of the molecular resonances crosses one of the Fermi levels. This opens a current channel, and shows as a jump in the I–V curve. This behavior is similar to the phenomenon of resonant tunneling observed in quantum well devices. One important difference, however, is the fact that in quantum wells, application of a voltage above the resonant voltage shuts down the current channel because the resonance lies in the energy gap of the reservoirs. With molecular wires, once a channel is activated, it remains open since the reservoirs are metallic and occupied states can always be found that align with the molecular resonance. 3. Summary In brief, we numerically investigate the conductance properties and I-V characteristics of zigzag single-walled BCN alloy nanotube in the CNT/BCN/CNT structure, where the )0,( n nanocontacts are considered as zigzag single-walled carbon nanotubes. We have applied )0,( n some well-known approaches and methods based on Green's function theory and Landauer formalism as well as Tight-binding Hamiltonian model to investigate the electron conduction through the CNT/BCN/CNT structure. Our results show that the presence of C atoms in BCN alloy nanotube induces the electronics states within the band gap and causes a large enhancement in the conductance of the system. Our outcomes suggest that any increasing in the nanotubes’ diameter considerably give rise to the enhancing of the conductance of the system. Besides, according to our results, this device can be a possible candidate for a nanoelectronics switching device. -2 -3… L -1 S D Fig. 1. A schematic representation of division of system into leads L and R, device D, and scattering region S as described in the text. … 1 2 3 R 4 10 8 6 4 2 0 8 6 4 2 0 8 6 4 2 0 ) t i n u . b r a ( S O D (c) (b) (a) -6 -5 -4 -3 -2 -1 0 E(eV) 1 2 3 4 5 6 Fig. 2. Panels (a)-(c) show the electronic DOS vs. the Energy for the CNT/BCN/CNT structure for n=6, 12, 15 in (n,0) BCN nanotube and (n,0) CNTs, respectively. The length of nm92.5 . BCN nanotube is selected nearly 5 2.0m 1.0m 0.0 1.0m ) E ( T 0.0 1.0m (c) (b) (a) 0.0 -6 -4 -2 0 E (eV) 2 4 6 Fig. 3. Panels (a)-(c) show the transmission probability of an electron through the wire as a function of the Energy corresponding to plots (a)-(c) in Fig. 1, respectively. 6 ) A n ( I 1 0.5 0.0 1 0.5 0.0 1 0.5 0.0 0 (c) (b) (a) 1 2 V (Volt) 3 4 Fig.4. Panels (a)-(c) show the current-voltage characteristics of CNT/BCN/CNT structure corresponding to plots (a)-(c) in Fig. 1, respectively. 7 ) A n ( I 0.566 0.565 0.564 0.563 0.562 0.561 (6,0) (12,0) Index of zigzag (n,0) (15,0) Fig. 5. The current versus zigzag nanotubes’ diameter in V= 0.4985 (Volt). References [1] C.Kergueris and et al., Phys. Rev. B 59 (1999) 12505. [2] M.Magoga and C.Joachim, Phys. Rev. B 56 (1998) 4722 and 59 (1999) 16011. [3] E.G.Emberly and G.Kirczenow, Phys. Rev. B 58 (1998) 10911; Phys. Rev. Lett. 81 (1998) 5205; Nanotechnology 10 (1999) 285 ; J. Phys.: Condens. Matter 11 (1999) 6911. [4] L.E.Hall and et al., J.Chem. Phys. 112 (2000) 1510. [5] S.Datta and et al., Phys. Rev. Lett. 79 (1997) 2530; W.Tian, Physica E (Amsterdam) 1 (1997) 304; J.Chem.Phys. 109 (1998) 2874; Y.Xue, Phys. Rev. B 59 (1999) R7852. [6] Qi.Pengfei and et al., J.Am.Chem. Soc. 126 (2004) 11774. [7] X.Guo and et al., Nature Nanotechnology 3 (2008) 163. [8] R.Gutierrez and et al., Europhys. Lett. 62 (2003) 90. [9] Miriam del Valle and et al., Nature Nanotechnology 2 (2007) 176. [10] G.C.Liang and et al., Phys. Rev. B 69 (2004) 115302. [11] J.Guo and et al., IEEE Trans. Nanotech. 2 (2003) 329. [12] Kikuo Harigaya, Physica E 29 (2005) 628. [13] R.Saito, G.Dresselhaus, M.S.Dresselhaus, Physical Properties of Carbon Nanotubes, Imperial College Press, London, 1998. [14] J.Otsuka, K.Hirose and T.Ono, Phys. Rev. B 78 (2008) 035426. [15] C.T.White, D.H.Robertson, J.W.Mintmire, Phys. Rev. B 47 (1993) 5485. [16] J.W.G.Wildoer and et al., Nature (London) 391 (1998) 59. [17] J.Yu, Chem. Phys. Lett. 323 (2000) 529 and therein. [18] G.Y.Guo and et al., Phys. Rev. B 69 (2004) 205416. [19] F.Piazza and et al., Diamond and Related Materials 14 (2005) 965. [20] O.Stephan and et al., Science 266 (1994) 1683. [21] Z.Weng-Sieh and et al., Phys. Rev. B 51 (1995) 11229. [22] P.Redlich and et al., Chem. Phys. Lett. 260 (1996) 465. [23] K.Suenaga and et al., Science 278 (1997) 653. 8 [24] M.Terrones and et al., Chem. Phys. Lett. 257 (1996) 576. [25] R.Sen and et al., Chem. Phys. Lett. 287 (1998) 671. [26] Y.Zhang and et al., Chem. Phys. Lett. 279 (1997) 264. [27] M.Buttiker and et al., Phys. Rev. B 31 (1985) 6207. [28] V.Mujica ,M.Kemp and M.A.Ratner, J. Chem. Phys. 101 (1994) 6849. [29] B.Larade and A.M.Bratkovsky, Phys. Rev. B 68 (2003) 235305. [30] Y.Calev and et al., Israel Journal of Chemistry 44 (2004) 133. [31] P.A.Khomyakov and et al., Phys. Rev. B 72 (2005) 035450. [32] W.Fa, J.Chen, H.Liu and J.Dong, Phys. Rev. B 69 (2004) 235413. [33] X.Yang and J.Dong, Phys. Lett. A 330 (2004) 238. [34] J.W.G.Wildoer and et al., Nature 391 (1998) 59. [35] P.Kim, T.W.Odom, J.L.Huang and C.M.Lieber, Phys. Rev. Lett. 82 (1999) 1225. 9
1109.3817	1	1109	2011-09-17T20:54:33	Fractal nanostructures with the Hilbert curve geometry as a SERS substrate	[ "cond-mat.mes-hall" ]	We suggest a new type of substrates for the Surface Enhanced Raman Scattering measurements with the geometry based on self-similar fractal space filling curves. As an example, we have studied theoretically the dielectric response properties of doped semiconductor nanostructures, where the conducting electrons are trapped in the effective potential having the geometry of the Hilbert curve. We have found that the system may exhibit the induced charge distribution specific for either two dimensional or one dimensional systems, depending on the frequency of the external applied field. We have demonstrated that with the increasing of the depth of the trapping potential the resonance of the system counterintuitively shifts to lower frequencies.	cond-mat.mes-hall	cond-mat	Fractal nanostructures with the Hilbert curve geometry as a SERS substrate Ilya Grigorenko Penn State (Dated: December 5, 2018) We suggest a new type of substrates for the Surface Enhanced Raman Scattering measurements with the geometry based on self-similar fractal space filling curves. As an example, we have studied theoretically the dielectric response properties of doped semiconductor nanostructures, where the conducting electrons are trapped in the effective potential having the geometry of the Hilbert curve. We have found that the system may exhibit the induced charge distribution specific for either two dimensional or one dimensional systems, depending on the frequency of the external applied field. We have demonstrated that with the increasing of the depth of the trapping potential the resonance of the system counterintuitively shifts to lower frequencies. Quick, reliable and selective single molecule detection of different molecular species is a challenging problem. The Surface Enhanced Raman Scattering (SERS) tech- nique is considered as one of the most powerful meth- ods for single molecular detection [1, 2]. The local elec- tromagnetic field enhancement on random metallic sub- strates of irregular, fractal-like geometries, or on corru- gated metal surfaces was considered as one of the pri- mary enhancement mechanisms for the SERS measure- ments since the discovery of the phenomenon. Such ge- ometry results in a local field intensity enhancement by a factor up to E2 ∝ 103 − 105 in so called "hot spots" [3, 4]. The Raman signal is proportional to E4, there- fore a molecule in a "hot spot" may emit up to 106− 1010 times stronger signal. Such kind of substrates were in- tensively studied both experimentally and theoretically [1, 2, 5 -- 7]. For example, it was shown, that the statistics of "hot spots" in such substrates may be governed by power law distributions [3]. However, random substrates have significant disadvantage of poor reproducibility of measurements for different substrate samples. Another disadvantage of these type of substrate geometries is a relatively low degree of controllability of the resonance properties of the substrate. In the last decade there have been a considerable interest in stable, reproducible SERS substrates. For example, arrays of nanoparticle pairs [8, 9] or pho- tonic/plasmonic crystal structures were considered re- cently as an alternative geometry with more degrees of control of the substrate's resonance properties [10]. Such kind of designs can generate a periodic pattern of a high intensity "hot spots", however, a relatively large fraction of the area is SERS inactive. In this paper we consider a new type of geometry for SERS substrates, which com- bines the advantages of irregular fractal substrates, re- sulting in a strong local field enhancement, and periodic plasmonic structures, characterized by relatively easy re- producibility and a high degree of controllability of the substrate's resonance properties. We suggest to use for the SERS substrate a geometry derived from a special class of fractal-like space filling curves [11]. For example, the Hilbert curve is a special case of space filling curves. It has a large number of turns and corners for a given length of a curve (please, see Figs.(1,2)). Note, that the Hilbert curve is a self similar structure, which may be built making a finite (or infinite) number of iterations. A nanoscale SERS substrate will not be a true fractal because of the lack of an infinitely fine spatial resolution. In particular, at the atomic scale the quantization of elec- tronic states will start to play a significant role. Since the local field enhancement is the most significant at surfaces with a large local curvature (at the corners), the Hilbert curve is a good candidate as a substrate for the SERS detection with a high probability for a molecule to hit a "hot spot. Another advantage of the Hilbert curve is that while it has a self-similar fractal structure, its Haus- dorff dimension is 2, i.e. the Hilbert curve uniformly covers surface without empty spots. Regular fractal-like structures have been considered for nano-photonic ap- plications before [13 -- 15]. However, in these works the authors considered fractal structures with the Hausdorff dimension 1 < D < 2, and therefore, such structures do not fill surface uniformly. Space filling substrates can be manufactured using modern lithography techniques [10]. A possible approach is to use ion beam lithography methods to etch trenches with the shape of the Hilbert curve on a metal surface or a thin film. In a case of a thin film one may vary the depth of the trenches, continuously moving from two dimensional geometry of a thin film to the one dimen- sional self-similar geometry of a Hilbert wire. Another alternative approach is to use metal nanospheres orga- nized on a surface as the Hilbert curve. In the last case one has an additional degrees of control of the resonance properties. By choosing certain spheres radii and the inter-sphere distance one may shift the resonances in the most suitable range for the single molecule detection of a given chemical specie. In order to increase the de- gree of control of the resonant properties of the system one may fabricate the nanostructure using spheres made of different materials (for example, silver and gold), or using bimetal spheres [16]. One may also use Surface Tunneling Microscope techniques [17] to assemble and measure the response of an atomic scale structure hav- ing the geometry of the Hilbert curve. In order to utilize rich dielectric properties of fractal-like space filling curves in modern nanoelectronics one may use photo lithogra- phy methods, similar to those for microchip fabrication, which will result in non-uniform doping levels in a semi- conductor substrate. Conducting electrons will be local- ized in a highly non-uniform trapping potential with the Hilbert curve geometry. Note that while the electrons will be trapped in a quasi-1D potential, the interaction with an external field will result in the induced charge distribution having also quasi one dimensional proper- ties. However, the electrons may also interact with each other via the Coulomb interaction, which will extend out of the plane of the substrate, making the system effec- tively two dimensional. As we will show in this paper, the collective excitations in such systems may exhibit the induced charge distributions, which is characteristic for either one or two dimensional systems, depending on the excitation frequency. In order to calculate the dielectric properties of frac- tal nanostructures we used linear response theory un- der Random Phase Approximation (RPA) [12]. This ap- proach allows us to take into account the non-local prop- erties of the dielectric response of the highly inhomoge- neous atomic scale system. The Schrodinger equation for non-interacting electrons e moving in a trapping potential with the effective mass m∗ V (r), is given by HΨi(r) = (cid:18)− 2 e ∇2 + V (r)(cid:19) Ψi(r) = EiΨi(r). 2m∗ (1) The eigenenergies Ei and eigenfunctions Ψi(r) are ob- tained using numerical diagonalization. The induced po- tential φind is then determined from the self-consistent integral equation φind(r, ω) = Z χ0(r′, r′′, ω)φtot(r′′, ω)VC (r − r′)dr′dr′′.(2) Here VC (r − r′) is the Coulomb potential and χ0(r′, r′′, ω) = Xi,j f (Ei) − f (Ej) Ei − Ej − ω − iγ Ψ∗ i (r′)Ψi(r′′)Ψ∗ j (r′′)Ψj(r′) (3) is the non-local density-density response function, f (Ei) is the Fermi distribution function and γ is the level broad- ening constant. Here φtot(r, ω) = φext(r, ω) + φind(r, ω) is the self-consistent total potential. The external field is assumed to be harmonic with frequency ω, linearly po- larized, and with the wavelength much larger than the characteristic system's size, therefore Eext does not de- pend on r, φind(r) = −Eextx. The integral equation Eq.(2) was discretized on a real-space cubic mesh with the lattice constant ∆, and the resulting system of linear equations was solved numerically. In our simulations we used the discretization length ∆ = 1.8nm and we set the 2 FIG. 1. The ground state electron density for a nanostructure with the trapping potential having the geometry of the Hilbert curve. The size of the system L = 60 nm. The number of electrons N = 50. FIG. 2. The ground state electron density for a nanostructure with the trapping potential having the geometry of the Hilbert curve. The size of the system L = 240 nm. The number of electrons N = 200. effective electron mass to m∗ e = 0.067me that corresponds to doped to the concentration 1018cm−3 GaAs [18]. The e∆2) natural energy scale E0 is defined by E0 = 2/(2m∗ and E0 ≈ 174 meV in this paper. The level broaden- ing constant is set to γ = 1 × 10−3 E0. In our simu- lations we assumed zero temperature T = 0K. The di- electric response is quantified using the total energy of the local field inside and outside of the nanostructure W = RV Etot(r)2dr, normalized by the energy of the applied field W0 = RV Eext2dr. First we calculated the ground state electron density for the trapping potential having a constant depth d and a geometry of the Hilbert curve. We assumed 50 electrons in the system. The potential depth is set to d = 5E0 and has geometry the Hilbert curve (Fig.1), The size of the system 60nm×60nm×7.2nm. We have also considered a larger system with the same electron density. In Fig.2 we plot the ground state electron density assuming N = 200 electrons in 240nm×240nm×7.2nm. system (please see Fig 2). The potential depth is again set to d = 5E0. We calculated the dielectric response of the fractal sys- tem shown in Fig. 1 for two electron densities, assuming N = 25 and N = 50 electrons in the system. In both 3 104 103 102 0 0.4 0.2 0.6 Frequency, hω/E0 0.8 0 W W / , y t i s n e t n i l d e F i FIG. 3. The normalised local field energy W/W0 near the nanos- tructure shown in Fig.1 as a function of the frequency of the applied field ω for different number of electrons in the system: N = 25 (solid line), N = 50 (dashed line). FIG. 6. The local field intensity log10(Etot(r)2) at the reso- nant frequency ω1, corresponding to the induced charge distribu- tion shown in Fig. 4. The direction of the external field Eext(ω) points along the y axis. The size of the system L = 60 nm. The dots mark the shape of the trapping potential. FIG. 4. The induced charge density (in arb. units) at the resonant frequency ω1. The direction of the external field Eext(ω) points along the y axis. The size of the system L = 60 nm. The dots mark the shape of the trapping potential. cases the response has a complex dependence on the fre- quency of the external field that is consistent with the complex geometry of the nanostructure. In the case of N = 25 the normalized energy of the local field W/W0 in the system has a distinct maximum at almost zero frequency (Fig.3, solid line). In the case of N = 50 the FIG. 5. The induced charge density (in arb. units) at the resonant frequency ω2. The direction of the external field Eext(ω) points along the y axis. The size of the system L = 60 nm. The dots mark the shape of the trapping potential. FIG. 7. The local field intensity log10(Etot(r)2) at the reso- nant frequency ω2, corresponding to the induced charge distribu- tion shown in Fig. 5. The direction of the external field Eext(ω) points along the y axis. The size of the system L = 60 nm. The dots mark the shape of the trapping potential. response of the system has two clear maxima as a func- tion of the frequency of the applied field, one peak is at ω1 ≈ 0.04E0, and the second is at ω2 ≈ 0.18E0 (Fig.3, dashed line). We have also calculated the dielectric re- sponse for higher electron densities: N = 100, 250 (not shown). Higher carrier concentration leads to a shorter electron screening length and reduced quantum delocal- ization effects [12]. As a result, the dielectric response in systems with higher electron densities is more close to macroscopic (bulk) systems. The local field enhancement is localized near "corners" of the Hilbert curve. The re- sponse has many resonances, which are extended over a large frequency scale, which is consistent with the com- plex self-similar geometry of the nanostructure. We have investigated the induced charge spatial distribution at resonance frequencies for N = 50. At ω1 the induced charge density has a two dimensional profile: the nanos- tructure is divided into upper and lower halves having the induced charge densities of the opposite signs (Fig.4). At the frequency of the applied field ω2 the induced charge spatial distribution has a pronounced one dimensional 0 W W / , y t i s n e t n i l d e F i 106 104 0 0.1 Frequency, hω/E0 0.2 0.3 FIG. 8. The normalized local field energy W/W0 as a function of the frequency of the external field ω for different depth of the trapping potential d. Dotted line corresponds to d = 0, solid thin line: d = 1E0, dashed-dotted line d = 2E0, dashed line d = 4E0, solid thick line d = 5E0. distribution along the path of the Hilbert curve (Fig.5). The local field intensities produced by these two charge distributions are plotted in Figs.(6,7). One can see that in the case of quasi one dimensional excitation (Fig.7) the spatial distribution of "hot spots" is much more uni- form and has less inactive area, that beneficial for more reliable SERS measurements. Note that in Figs.(6,7) we used a logarithmic scale. m∗ bulk = ρe2 We have also calculated the dielectric response of the larger nanostructure shown in Fig.2. for different depth of the trapping potential d = 0, 1, 2, 4, 5E0. Based on the results of simulations shown in Fig. 3, the maximum of the response of a larger fractal-like nanostructure should be near ω = 0. In the case of d = 0 (finite thin film) the system has its response maximum near the surface plas- mon resonance (ωpl = ωbulk/√2), where bulk plasmon frequency ω2 e ǫ0ǫ , here ρ is the carrier concentra- tion, e is the electron charge, ǫ0 is the vacuum permittiv- ity, and ǫ ≈ 11.1 is the high-frequency dielectric constant of GaAs [18]. With the increasing of the depth d of the trapping potential one naively expects that the resonant frequency will shift to a higher frequency, since the same amount of electrons will effectively occupy a smaller vol- ume, that will correspond to a higher electron density and higher resonant frequency. However, the resonance fre- quency of the Hilbert geometry has moved to lower values (please, see Fig.8), and at d = 5E0 the resonance becomes very close to ω = 0. Note that the overall magnitude of the normalized local field is decreased with the increase of the depth d. One may attribute this to a reduction of the screening in quasi-one dimensional systems. As a re- sult, the amplitude of the collective (plasmon) resonance in the quasi one dimensional system should be less than in the quasi-two dimensional system. In summary, we have studied the dielectric properties of nanoscale systems with the geometry of the Hilbert In particular, we found that while space filling curve. 4 electrons are trapped in a quasi-one dimensional poten- tial with the Hilbert curve geometry, one can observe either two dimensional or one dimensional excitations in the system, depending on the excitation frequency. We have also studied how the plasmon response depends on the depth of the trapping potential, which effectively controls the smooth transition of the system geometry from quasi-two dimensional to quasi-one dimensional. We have found that while the average electron density increases with the increasing of the depth of the trap- ping potential, the geometry of the Hilbert curve results in the overall red shift in the plasmon frequency. This opens a broad possibility in control of the resonance prop- erties of such systems, making them a robust and repro- ducible Raman active nano-structured surfaces for the SERS, and other nanoplasmonic applications. [1] V. M. Shalaev, E. Y. Poliakov, and V. A. Markel, Phys. Rev. B 53, 2437 (1996). [2] S. Nie and S. R. Emory, Science 275, 1102 (1997). [3] M.I. Stockman, V.M. Shalaev, M. Moskovits, R. Botet, T.F. George, Phys. Rev. B 46, 2821 (1992). [4] M. I. Stockman, D. J. Bergman, and T. Kobayashi, Phys. Rev. B 69, 054202 (2004). [5] W. Kim, V. P. Safonov, V. M. Shalaev, and R. L. Arm- strong, Phys. Rev. Lett. 82, 4811 (1999). [6] M. Montagna, O. Pilla, G. Viliani, V. Mazzacurati, G. Ruocco, and G. Signorelli, Phys. Rev. Lett. 65, 1136 (1990). [7] T. Keyes and T. Ohtsuki, Phys. Rev. Lett. 59, 603 (1987). [8] T. Atay, J.H. Song, and A.V. Nurmikko, Nano Lett. 4,1627 (2004). [9] I. Grigorenko, S. Haas, A. Balatsky and A. F. J. Levi, New J. Phys. 10, 043017 (2008). [10] M. E. Stewart, C. R. Anderton, L. B. Thompson, J. Maria, S. K. Gray, J. A. Rogers, and R. G. Nuzzo Chem. Rev., 108, 494 (2008). [11] H. Sagan, Space-filling curves Springer-Verlag, NY (1994). [12] I. Grigorenko, S. Haas and A.F.J. Levi, Phys. Rev. Lett. 97, 036806 (2006); [13] G. Volpe, G. Volpe, and R. Quidant, Opt. Express., 19 3612 (2011). [14] A.A. Grunin, A.G. Zhdanov and A.A. Fedyanin Proc. of SPIE 6728, 672837, (2007). [15] J. Alda, J. M Rico-Garcia, J. M Lopez-Alonso and G Boreman, Nanotechnology 16, 230 (2005). [16] C. H. Liu, M. H. Hong, H. W. Cheung, F. Zhang, Z. Q. Huang, L. S. Tan, and T. S. A. Hor, Opt. Express. 16 10701 (2008). [17] C. R. Moon, L. S. Mattos, B. K. Foster, G. Zeltzer, W. Ko and H. C. Manoharan, Science 319, 782 (2008). [18] J. R. Hayes and A. F. J. Levi, IEEE Journ. of Quant. Elect. 22, 1744 (1986).
1603.02161	1	1603	2016-03-07T17:18:49	Gyroidal nanoporous carbons - Adsorption and separation properties explored using computer simulations	[ "cond-mat.mes-hall", "cond-mat.soft" ]	Adsorption and separation properties of gyroidal nanoporous carbons (GNCs) - a new class of exotic nanocarbon materials are studied for the first time using hyper parallel tempering Monte Carlo Simulation technique. Porous structure of GNC models is evaluated by the method proposed by Bhattacharya and Gubbins. All the studied structures are strictly microporous. Next, mechanisms of Ar adsorption are described basing on the analysis of adsorption isotherms, enthalpy plots, the values of Henry's constants, $\alpha_{s}$ and adsorption potential distribution plots. It is concluded that below pore diameters ca. 0.8 nm, primary micropore filling process dominates. For structures possessing larger micropores, primary and secondary micropore filling mechanism is observed. Finally, the separation properties of GNC toward CO$_2$/CH$_4$, CO$_2$/N$_2$, and CH$_4$/N$_2$ mixtures are discussed and compared with separation properties of Virtual Porous Carbon models. GNCs may be considered as potential adsorbents for gas mixture separation, having separation efficiency similar or even higher than activated carbons with similar diameters of pores.	cond-mat.mes-hall	cond-mat	CondensedMatterPhysics,2016,Vol.19,No1,13003:1 -- 14 DOI:10.5488/CMP.19.13003 http://www.icmp.lviv.ua/journal Gyroidalnanoporouscarbons -- Adsorptionand separationpropertiesexploredusing computersimulations∗ S.Furmaniak1,P.A.Gauden1,A.P.Terzyk1,P.Kowalczyk2 1 FacultyofChemistry,PhysicochemistryofCarbonMaterialsResearchGroup, NicolausCopernicusUniversityinToruń,GagarinSt.7,87 -- 100Toruń,Poland 2 SchoolofEngineeringandInformationTechnology,MurdochUniversity, Murdoch,WesternAustralia6150,Australia ReceivedNovember13,2015,innalformDecember17,2015 Adsorptionandseparationpropertiesofgyroidalnanoporouscarbons(GNCs) -- anewclassofexoticnanocar- bon materials are studied for the rst time using hyper parallel tempering Monte Carlo Simulation technique. Porous structure of GNC models is evaluated by the method proposed by Bhattacharya and Gubbins. All the studied structures are strictly microporous. Next, mechanisms of Ar adsorption are described basing on the analysis of adsorption isotherms, enthalpy plots, the values of Henry's constants, αs and adsorption poten- tial distribution plots. It is concluded that below pore diameters ca. 0.8 nm, primary micropore lling process dominates. For structures possessing larger micropores, primary and secondary micropore lling mechanism isobserved.Finally,theseparationpropertiesofGNCtowardCO2/CH4,CO2/N2,andCH4/N2 mixturesaredis- cussed and compared with separation properties of Virtual Porous Carbon models. GNCs may be considered as potential adsorbents for gas mixture separation, having separation eciency similar or even higher than activatedcarbonswithsimilardiametersofpores. Keywords:gyroidalnanoporouscarbons,adsorption,gasmixturesseparation,MonteCarlosimulations PACS:05.10.Ln,05.70.-a,51.10.+y,51.30.+i,64.75.+g 1. Introduction Adsorption of gas mixtures on solid surfaces has been attracting great interest for many years [1 -- 7]. In the last decades there has been observed an increased interest to the development of new separation andpuricationtechniques.Itisevidentthatadsorptionusingvariousadsorbentsisstillaversatiletool for these purposes. On the other hand, the basic problem appearing in experimental studies is caused by diculties in the synthesis of nanomaterials possessing desired properties. It is still not simple to control porosity and/or the chemical nature of the surface, and the both parameters at the same time. Moreover, experimental data on mixed systems are very limited, i.e., in the case of mixtures consisting oftwovolatilecomponentstheproblemofthesurfacecoveragedeterminationhasnotbeenfullysolved yet.Predictingadsorptionbehaviourofmixturesfrompurecomponentdataisveryimportant,fromboth thetheoreticalandpracticalviewpoints[8 -- 14].Itiswellknownthatthetheoreticalcalculationsprovide additionalopportunitiesforstudiestobetterunderstandtheseparationprocesses.However,despitethe intensive experimental and theoretical studies, our knowledge of the properties and the structure of mixedadsorbedlayersisrathersparse,especially,onnewgenerationsofnanoporousadsorbents. Computer simulation is an ecient method for resolving the above mentioned problems, since it is capable of modelling the processes of interest at the required level of detail in a controllable environ- ment, providing the necessary tools for establishing the connections between the observed phenomena ∗ThispaperisdedicatedtoProf.StefanSokołowskiontheoccasionofhis65thbirthday. ©S.Furmaniak,P.A.Gauden,A.P.Terzyk,P.Kowalczyk,2016 13003-1 6 1 0 2 r a M 7 ] l l a h - s e m . t a m - d n o c [ 1 v 1 6 1 2 0 . 3 0 6 1 : v i X r a S.Furmaniaketal. and their molecular-level physical background. Prof. Sokołowski's (and co-workers) research topics of interesthavealsobeenconcernedwiththeissueofmixturesusingMonteCarlosimulations[15 -- 22],Den- sity Functional Theory [23 -- 28], and Dissipative Particle Dynamics [29]. Their inspiring articles discuss, for example, the following problems: (i) adsorption from mixtures of monomers [15], dimers [15], the chainmolecules[30,31],andevenpolymermixtures[27],(ii)adsorptionfrommixturesonhomogeneous [25, 32] and heterogeneous surfaces [15, 33, 34], (iii) layering transition, capillary condensation, wetting phenomena, and multilayer adsorption of binary ideal mixtures, systems exhibiting negative deviations from ideal mixing or positive one, binary mixture with partially miscible components, etc. [16, 17, 34 -- 38], (iv) interaction of charged chain particles and spherical counterions in contact with charged hard wall [31], (v) analysis of the properties of two-dimensional symmetrical mixtures in an external eld of squaresymmetry[39,40],(vi)demixingandfreezingintwo-dimensionalsymmetricalmixtures[21],and (vii) the behaviour of mixed two component submonolayer lms (Ar and Kr [41, 42] or Kr and Xe [22] on graphite). The majority of the analysed adsorbents have an ideal geometry of pores, for example a slit-like[25,26,32,34,43].However,morecomplexmodelshavealsobeenstudied,forexample,pillared slit-like pores [28] and slit-like pores with walls decorated by tethered polymer brushes in the form of stripes[29]. In the last decades, novel exotic porous carbon nanostructures (such as carbon nanotubes (CNTs), single-walled carbon nanohorns, graphene and graphitic nanoribbons, ordered porous carbons, worm- likenanotubesandgraphiticnanobers,stacked-cupcarbonnanobers,cubiccarbonallotropes,carbon onions, carbyne networks, and others) have been projected to be among the most useful materials for selectiveadsorptionandseparationofuidmixtures[44 -- 48].However,inthetheoreticalstudies,differ- entcarbonadsorbentsarestudied,suchas:carbonnanotubes[13,49 -- 53],carbonnanohorns[13,51,54], 2D and 3D ordered carbon networks [55], hydrophobic virtual porous carbons (VPCs) [12, 14, 56 -- 62], ox- idized VPCs [12, 14, 60, 62, 63], and triply periodic carbon minimal surfaces (Schwarz's primitive and Schoen's gyroid) [45, 59, 64 -- 70]. Recently, scientists have paid attention to the next generation of porous carbon molecular sieves materials, i.e., crystalline exotic cubic carbon allotropes: cubic carbon poly- morphs (CCPs) [45, 71 -- 73], diamond-like super structures of CNTs (super diamonds) [74], diamond-like frameworks[75],porousaromaticframeworks(PAFs)[76,77],diamond-likecarbonframeworks(i.e.,di- amondynes,alsonamedD-carbons)[78],tetrahedralnodediamondyne[79],carbonallotropesproposed by Karfunkel and Dressler [45, 80], compressed carbon nanotubes [45, 81], sodalite-like nanostructures [45, 82], folding of graphene slit-like pore walls [52, 83], gyroidal nanoporous and mesoporous carbons (GNCsandGMCs,respectively)[84 -- 86]. OneofthemostinterestingandpromisingadsorbentfromtheabovementionedisGNC.Inthecurrent study, we consider nine different GNC structures having surface built in a way ensuring connection of eachcarbonatomwithexactlythreeneighbours,similarlyas"schwarzites".Nicolaïetal.[84]conrmed that the curvature and the rigidity do not play a crucial role in the performance of GNC structures for ionicconduction.Themajorrole,however,isplayedbytheporesizeandporevolume.Indeed,thelarger the pore is, the larger is the ionic transport. Finally, the mentioned authors stated that GNC structures withtunablepropertiescanbewidelyappliedtowaterltrationorenergystorage. 2. Simulationdetails We used the structures of nine gyroidal nanocarbons (denoted as GNC-04, GNC-07, GNC-09, GNC-11, GNC-12, GNC-13, GNC-15, GNC-18 and GNC-21) published previously by Nicolaï et al. [84] (see gure 1 in [84]). In the case of the rst six systems, original boxes generated by Nicolaï et al. [84] were multiplied (eightfold) in order to obtain box size at least two times greater than the cut-off distances used during simulationsdescribedbelow.Theporosityofallthestudiedcarbonaceousadsorbentswascharacterised by a geometrical method proposed by Bhattacharya and Gubbins (BG) [87]. The implementation of the method was described in detail elsewhere [88, 89]. The BG method provided histograms of pore sizes (effective diameters -- deff). These data were also used to calculate the average sizes of pores accessible for Ar atoms (deff,acc,av) [88]. In addition, the volume of pores accessible for Ar was determined using a combinationoftheBGmethodandMonteCarlointegration[88]. Argon adsorption isotherms at its boiling point (i.e., T = 87 K) on all the studied nanocarbons were 13003-2 Gyroidalnanoporouscarbons -- Adsorptionandseparationproperties simulated using the hyper parallel tempering Monte Carlo method (HPTMC) proposed by Yan and de Pablo [90]. The simulation scheme was the same as in previous work [88]. We considered 93 replicas correspondingtodifferentrelativepressurevalues(p/ps,where p and ps areequilibriumandsaturated Ar vapour pressure, respectively) in the range 1.0× 10 −10 − 1.0. Other details of the performed HPTMC simulations are described in [88]. The average numbers of Ar atoms in the simulation box were used to calculate the adsorption amount of Ar per unit of mass of the adsorbent (a) [88]. The isosteric enthalpy of adsorption (q st) was also determined from the theory of uctuations [88, 91] to reect the energetics oftheprocess. In order to analyse the mechanism of Ar adsorption, we constructed high resolution αs-plots [92] based on simulated adsorption isotherms. We used Ar adsorption isotherm simulated in the ideal slit- like system composed of graphene sheets with effective pore width equal to 10 nm as the reference one. We also determined the values of Henry's constant (KH) from the slope of the linear part of adsorption isotherms in the low-pressure range [83]. Finally, adsorption potential distribution (APD) curves [93 -- 95] were calculated. The APD curve is the rst derivative of the so-called characteristic curve, presenting adsorptionamountasafunctionoftheadsorptionpotential(Apot)denedas: p ps , Apot = −RT ln (2.1) where R istheuniversalgasconstant.Thedifferentiationwasperformednumericallybytheapproxima- tionoftheisothermsusingsomeempiricalfunctionsandcalculatingtheirderivatives. Wealsosimulatedtheadsorptionandseparationofthreebinarygasmixtures(importantfromprac- tical point of view): CO2/CH4, CO2/N2, and CH4/N2 on all the studied GNCs. The computations were per- formed for T = 298 K using the grand canonical Monte Carlo method (GCMC) [91, 96]. The simulation schemewasthesameasinourpreviousworks[60,62].Simulationswereperformedforthetotalmixture pressure ptot = 0.1 MPa (i.e., atmospheric pressure) and for the following mole fractions of components in the gaseous phase (y): 0.0, 0.01, 0.025, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.975, 0.99, and Figure 1. (Color online) Comparison of (a) pore size histograms for all the studied nanocarbons deter- minedfromtheBGmethod(thesubsequenthistogramsareshiftedby1unitfromthepreviousones),(b) averagesizesand(c)volumeofporesaccessibleforAratoms,respectively. 13003-3 S1/2 = x1/x2 y1/y2 . S.Furmaniaketal. 1.0.Foreachpoint,themolefractionsofcomponentsintheadsorbedphase(x)werecalculatedfromthe average numbers of molecules present in the simulation box. The eciency of the process of separation of mixtures was reected by the value of equilibrium separation factor (the 1 st component over the 2 nd one): (2.2) Theadsorbedphaseisenrichedinthe1st componentif S1/2 > 1. 3. Resultsanddiscussion Figure1(a)collectshistogramsofeffectiveporesizes,determinedfromtheapplicationofBGmethod, for all the studied GNCs. All the structures are strictly microporous, i.e., the diameters of pores do not exceed 2 nm. Generally, the size of dominant pores increases in the considered series (from the GNC- 04 up to GNC-21). However, there are two groups of nanocarbons having similar diameters of the main pores: (i) GNC-11, GNC-12 and GNC-13 -- deff around 1 nm and (ii) GNC-18 and GNC-21 -- deff around 1.5nm.ItshouldbenotedthatinthecaseofGNC-21,someamountofporeswiderthaninGNC-18isalso present. These regularities are reected by the values of the average pore diameter [gure 1 (b)]. The increaseinporediametersisaccomplishedbytheincreaseinporevolumefromca.0.4cm3/gforGNC-04 uptoca.1.3cm3/gforGNC-18andGNC-21[gure1(c)]. Figure 2 shows the comparison of Ar adsorption isotherms simulated for all the studied GNCs. The Figure2.(Coloronline)ComparisonofAradsorptionisotherms(T =87K)simulatedforalltheconsidered nanocarbons. 13003-4 Gyroidalnanoporouscarbons -- Adsorptionandseparationproperties Figure3.(Coloronline)Equilibriumargoncongurationsforselectednanocarbonsandselectedvaluesof relativepressure(theframesreectthesizeofsimulationboxes,allthestructuresareinthesamescale). ItshouldbenotedthatthisgurewascreatedusingtheVMDprogram[97]. −8 − 10 changes observed in the shape of isotherms reect the differences in the properties of nanocarbons. Adsorption capacity varies from ca. 12 mmol/g for GNC-04 up to ca. 45 mmol/g for GNC-18 and GNC-21. Thesechangescorrespondtothedifferencesintheporevolume[gure1(c)].Atthesametime,theshift oftheporellingpressuretowardhighervaluesisobserved.TheporesoftheGNC-04structurearelled −6. However, the total lling of GNC-18 and GNC-21 occurs for in the relative pressure range 10 similar values of relative pressure (around 10 −2). The middle carbons of the series (i.e., GNC-11, GNC-12 andGNC-13)arelledinthesimilarrangeofrelativepressure(p/ps > 10 −4).Thedifferencesinthepore llingarealsoclearlyseenonequilibriumsnapshotsofArcongurationsinthesimulationboxesshown in gure 3. These regularities are related to the differences in diameters of dominant pores present in the individual GNCs [gure 1 (a)]. Finally, one can observe that in the case of initial structures (up to GNC-13), the pore lling is a single-step process. However, the pores of GNC-15, GNC-18 and GNC-21 are lledintwosteps.Thisisalsocausedbytheriseinporesizes.Forporeswiderthan1nm,intherststep 13003-5 S.Furmaniaketal. Figure4.(Coloronline)ComparisonofisostericenthalpyofAradsorption[thedatarelatedtogure2;for clarityadsorptionamountsarenormalisedbythemaximumobservedvalue(amax)]. amonolayerisformedandnexttheremainingfreevolumeislled. Figure4showstheplotsofisostericenthalpyofadsorptionrelatedtotheisothermsshowningure2. For low loadings, the q st values increase as the adsorption amount rises. In this range, Ar atoms are ad- sorbed mainly on high-energetic centres. The increasing adsorption amount causes other Ar atoms to appearinthevicinityoftheinitiallyadsorbedones.Thisrisesthecontributionofuid-uidinteractions tothetotalenergyofasystem.Here,theonlyexceptionisGNC-21structure.Thereisobservedadecrease in q st for relative adsorption up to ca. 0.1. This system has probably got a heterogeneous surface. Con- sequently, the subsequent Ar atoms are adsorbed on centres having lower energy and this reduces the effectsoftheincreaseinlateralAr-Arinteractions.Intheintermediaterangeofadsorption,theenthalpy rises for all the structures until the entire adsorbent surface is covered. Next, the values of q st decrease since Ar is adsorbed at the places more distant from the surface and this is connected with lower solid- uidcontributions.Inthecaseofthestructureshavingthewidestpores(especiallyGNC-15,GNC-18and GNC-21),thesecondpeakon q st isalsoobserved.Thispeakisrelatedtothetotalllingofpores. Comparing the enthalpy at zero coverage for all the systems, one can observe that GNC-04 has a re- markablyhighervalueofthisparameter(ca.17.5kJ/mol)thantheothersystems.Thisisconnectedwith thepresenceofthenarrowestporeshavinghighadsorptionenergy.Theotherstructuresmaybedivided intotwogroups.TherstoneincludesGNC-07,GNC-09,GNC-12,GNC-13andGNC-21carbons.Inthiscase, the enthalpy at zero coverage is close to 12 kJ/mol. For the second group (GNC-11, GNC-15 and GNC-18), thisenthalpyvalueisintherange9 -- 10kJ/mol.Thismaysuggestsomesimilaritiesinthesurfacenatureof the group members (for example curvature, which is the main factor determining the energy of adsorp- tion on the surface). The above-described differences in the energy of interactions with the surface of 13003-6 Gyroidalnanoporouscarbons -- Adsorptionandseparationproperties Figure5.(Coloronline)ComparisonofHenry'sconstantsrelatedtotheAradsorptionisothermspresented ongure2. Figure 6. (Color online) Comparison of αs-plots related to the Ar adsorption isotherms presented in g- ure2[forclaritytheadsorptionamountsarenormalizedbythemaximumobservedvalue(amax)].The αs isthenormalizedadsorptiononthereferencematerial,i.e.,intheidealslit-likesystemwiththeeffective porewidthequalto10nm. 13003-7 S.Furmaniaketal. Figure 7. (Color online) (a) Comparison of APD curves [for clarity, all the curves are normalized by the maximum observed value (APDmax); subsequent curves are shifted by 1.35 units]. (b) Location of the main peak on the APD curves (Apot,max). (c) Correlation between the location of the main peak and the averagesizesofporesaccessibleforAratoms. adsorbentsfullycorrespondtothevariationofHenry'sconstantsshowningure5.Thisisnotsurprising since solid-uid interactions are the main factor affecting the shape of the isotherm in the low pressure range.Hence,thevalueof KH forGNC-04systemisatleast1000timesgreaterthanfortheotherones.For three of the remaining structures (i.e., GNC-11, GNC-15 and GNC-18), lower values of KH (< 104 mmol/g) arerecorded.Thesamecarbonshavethelowestenthalpyofadsorptionatzerocoverage. Figure6presentscomparisonof αs-plotsrelatedtotheAradsorptionisotherms.Onecanseethatthe adsorption process is dominated by a FS swings (GNC-04 and GNC-07) and the FS-CS swings (remaining structures) [98]. It can be noticed that with the rise in the pore diameter, the combination of primary andsecondarymicroporellingmechanismoccurs.Theboundarybetweenthosemechanismsislocated for the structures with porediameters around ca. 0.8 nm. It is also interesting that the range on αs-plots connectingFSandCSswingsisnotlinearasitisobservedforthecaseofslit-likecarbonmicropores.This canbecausedbythesurfacecurvature. Figure 7 (a) compares APD curves for all the systems studied. The presented data are complemen- tary to adsorption isotherms shown in gure 2. On all the APD curves, at least one (dominant) peak is observed. It corresponds to the pore lling. Its location [Apot,max, gure 7 (b)] is related to the pressure of the pore lling according to equation 2.1. Hence, this parameter may be correlated with the size of pores -- see gure 7 (c). This gure quantitatively conrms the above-described qualitative differences in the pore lling process. The width of the main peak also provides some information on the process. The narrow peak means that condensation occurs in a narrow range of relative pressure. By contrast, a wide peak denotes a wide condensation range. For example, the pore lling in GNC-04 system occurs, as mentionedabove,for 10 −6 andthisisreectedbyawidepeakwiththemaximumlocated atca.12.6kJ/mol.Forthissystem,theothertwopeaksarealsoobserved(thethirdonewiththemaximum 13003-8 −8 < p/ps < 10 Gyroidalnanoporouscarbons -- Adsorptionandseparationproperties Figure 8. (Color online) Comparison of equilibrium separation factors [S1/2, equation (2.2)] for the ad- sorption of all the mixtures (T = 298 K, ptot = 0.1 MPa) on all the nanocarbons studied. The data plotted as the function of the 1st component mol fraction in the gaseous phase (y1, the 1st component is CO2 for CO2/CH4 andCO2/N2 mixturesandCH4 forCH4/N2 mixture). atca.45kJ/molisverybroad).Thesepeaksreecttheothersub-stepsoftheArdensityriseinporesofthis structure.SimilarinterpretationalsoconcernstheadditionalpeaksobservedforGNC-11andGNC-12.In the case of GNC-18 and GNC-21 structures, the observed second peak is related to the above mentioned monolayer formation. A slightly different scenario occurs for GNC-15 carbon. Here, the dominant peak isconnectedwiththemonolayerformationandthesecondlow(alsowide)peakreectsthellingofthe remaining pore volume. This fact explains why the location of the main peak for this structure deviates fromthedistincttrendvisibleforalltheotherGNCsincorrelationshowningure7(c). Figure 8 presents a comparison of equilibrium separation factors for adsorption of all three studied mixtures for different compositions of gaseous phase. In addition, gure 9 directly compares the e- ciency of separation of equimolar mixtures for all the studied systems. The separation is a consequence of differences in the adsorption of mixture components. Since the critical temperature for the studied gases decreases signicantly in the sequence CO2 > CH4 > N2, the adsorption anity decreases in the 13003-9 S.Furmaniaketal. Figure 9. (Color online) Comparison of equilibrium separation factors for adsorption of equimolar (y = 0.5)mixtures:(a)CO2/CH4,(b)CO2/N2 and(c)CH4/N2 (ptot = 0.1MPa, T = 298K) -- seegure8. same sequence (for the subsequent gases, the process occurs for increasing value of reduced tempera- ture). Consequently, for the given GNC, the highest equilibrium separation factor is observed for CO2/N2 andthesmallestoneforCH4/N2 mixture.Thequalitativedifferencesinseparationeciencybetweenthe structuresstudiedarethesameforallthemixturesstudied.Thehighestvaluesofequilibriumseparation factor(remarkablyhigherthanfortheothersystems)areobservedforGNC-04carbon.Thisisconnected Figure 10. (Color online) Comparison of equilibrium separation factors [(a) CO2/CH4, (b) CO2/N2 and (c) CH4/N2] for mixture adsorption (ptot = 0.1 MPa, T = 298 K) on GNC-04 and GNC-21 nanocrabons and on three virtual porous carbons (VPCs): d0.5, d0.9 and d1.3 [88] (the data for VPCs taken from our previous paper [60]). In addition, panel (d) presents the pore size histograms for the presented VPCs [88] (the subsequenthistogramsareshiftedby0.15unitsfromthepreviousones). 13003-10 Gyroidalnanoporouscarbons -- Adsorptionandseparationproperties with the presence of the narrowest pores. Among the other nanocarbons, GNC-07 and GNC-09 also ex- hibit higher values of S1/2. However, they are lower than in the case of GNC-04 since these structures havewiderpores.TheeciencyofseparationforthenextGNCsissimilar.Nevertheless,somesmalldif- ferences are also noticeable (see insets in gure 9). The GNC-11, GNC-15 and GNC-18 structures are less ecient in comparison with the adjacent carbons in the series. These regularities are analogous to that observed for Henry's constants shown in gure 5 and discussed above. These facts suggest that in the case of GNCs with pores wider than ca. 1 nm, the main factor affecting the eciency of the separation processistheenergeticsofuidinteractionwiththecurvedsurfaceofthenanocarbonsstudied. Finally,gure10comparestheequilibriumseparationfactors(allthreemixturesstudied)forGNC-04 and GNC-21 structures and for three virtual porous carbons (VPCs) described in detail previously [88] and having different porosity -- see gure 10 (d). As one can see, the GNC-21 structure exhibits a sep- aration eciency similar to the d0.5 carbon. The main pores of both adsorbents have a similar width. However, this VPC has also some narrower micropores which presence probably positively affects the separation eciency. Such micropores are absent in the case of GNC-21 and, nonetheless, this nanocar- bon exhibits similar values of S1/2. This is the consequence of the adsorption energetics on a curved surface of this structure. On the contrary, the GNC-04 nanocarbon exhibits the eciency of the CO2/CH4 mixture separation similar to the d1.3 carbon. This VPC has micropores distributed in the range up to ca. 1 nm [gure 10 (d)]. A large part of them has diameters similar to or lower than the GNC-04 struc- ture.InthecaseofCO2/N2 andCH4/N2 mixtures,thevaluesofequilibriumseparationfactorsforGNC-04 are higher than for d1.3 carbon. This comparison (especially for CO2/CH4 mixture -- similar eciency for both adsorbents and for CH4/N2 mixture -- higher eciency for GNC-04) may suggest that a regu- larly curved surface of gyroidal carbons exhibits higher anity to CH4 molecules than a heterogeneous surfaceofactivatedcarbons. Summingup,theGNCsstudiedmaybeconsideredaspotentialadsorbentsforgasmixtureseparation. The eciency of this process is similar to or higher than for activated carbons with similar diameters of pores. The GNC-04 or similar structures seem to be quite promising materials for this purpose since this nanocarboncontainsnarrowandquiteuniformpores(ca.0.5nm). 4. Conclusions AdsorptionandseparationpropertiesofGNCs -- anewclassofexoticnanocarbonmaterials,arestud- ied for the rst time using computer simulation technique. All the structures studied are strictly micro- porous. The mechanisms of Ar adsorption are described basing on the analysis of adsorption isotherms, enthalpyplots,thevaluesofHenry'sconstants, αs andadsorptionpotentialdistributionplots.Belowthe porediametersca.0.8nm,primarymicroporellingprocessdominates.Forstructurespossessinglarger micropores, primary and secondary micropore lling mechanisms are observed. GNCs may be consid- eredaspotentialadsorbentsforgasmixtureseparation,havingseparationeciencysimilartoorhigher thanthisforactivatedcarbonswithsimilardiametersofpores. Acknowledgements The authors acknowledge the use of the computer cluster at Poznań Supercomputing and Network- ing Centre (Poznań, Poland) as well as the Information and Communication Technology Centre of the NicolausCopernicusUniversity(Toruń,Poland). References 1. GasSeparationbyAdsorptionProcesses,SeriesonChemicalEngineeringVol.1,YangR.T.(Ed.),ImperialCollege Press,London,1997. 2. RuthvenD.M.,FarooqS.,KnaebelK.S.,PressureSwingAdsorption,Wiley-VCH,NewYork,1994. 3. SinaiskiE.G.,LapigaE.J.,SeparationofMultiphase,MulticomponentSystems,Wiley-VCH,Weinheim,1997. 4. JaroniecM.,MadeyR.,PhysicalAdsorptiononHeterogeneousSolids,Elsevier,NewYork,1988. 13003-11 doi:10.1039/c0cp02281k. doi:10.1016/j.cplett.2014.01.031. S.Furmaniaketal. 5. RudzińskiW.,EverettD.H.,AdsorptionofGasesonHeterogeneousSurfaces,AcademicPress,NewYork,1992. 6. Rouquerol F., Rouquerol J., Sing K.S.W., Llewellyn P., Maurin G., Adsorption by Powders and Porous Solids. doi:10.1103/PhysRevE.67.061603. doi:10.1103/PhysRevE.67.031202. Principles,MethodologyandApplications,2ndEdn.,Elsevier,Amsterdam,2014; doi:10.1016/B978-0-08-097035-6.09991-5. 7. BansalR.C.,GoyalM.,ActivatedCarbonAdsorption,CRCPress,BocaRaton,2005. 8. Palmer J.C., Moore M.V., Roussel T.J., Brennan J.K., Gubbins K.E., Phys. Chem. Chem. Phys., 2011, 13, 3985; 9. LiuL.,NicholsonD.,BhatiaS.K.,J.Phys.Chem.C,2015,119,407;doi:10.1021/jp5099987. 10. OzturkT.N.,KeskinS.,Ind.Eng.Chem.Res.,2013,52,17627;doi:10.1021/ie403159c. 11. LuX.,JinD.,WeiS.,WangZ.,AnC.,GuoW.,J.Mater.Chem.A,2015,3,12118;doi:10.1039/C4TA06829G. 12. BillemontP.,CoasneB.,DeWeireldG.,Adsorption,2014,20,453;doi:10.1007/s10450-013-9570-z. 13. Furmaniak S., Terzyk A.P., Kaneko K., Gauden P.A., Kowalczyk P., Ohba T., Chem. Phys. Lett., 2014, 595-596, 67; 14. DiBiaseE.,SarkisovL.,Carbon,2015,94,27;doi:10.1016/j.carbon.2015.06.056. 15. BorówkoM.,PatrykiejewA.,RżyskoW.,SokołowskiS.,Langmuir,1997,13,1073;doi:10.1021/la950940a. 16. GrabowskiK.,PatrykiejewA.,SokołowskiS.,ThinSolidFilms,1999,352,259;doi:10.1016/S0040-6090(99)00291-6. 17. GrabowskiK.,PatrykiejewA.,SokołowskiS.,ThinSolidFilms,2000,379,297;doi:10.1016/S0040-6090(00)01398-5. 18. PatrykiejewA.,SałamachaL.,SokołowskiS.,PizioO.,Phys.Rev.E,2003,67,061603; 19. PatrykiejewA.,SałamachaL.,SokołowskiS.,DominguezH.,PizioO.,Phys.Rev.E,2003,67,031202; 20. PatrykiejewA.,SokołowskiS.,PizioO.,J.Phys.Chem.B,2005,109,14227;doi:10.1021/jp048170b. 21. PatrykiejewA.,SokołowskiS.,Phys.Rev.E,2010,81,012501;doi:10.1103/PhysRevE.81.012501. 22. PatrykiejewA.,SokołowskiS.,J.Chem.Phys.,2012,136,144702;doi:10.1063/1.3699330. 23. BrykP.,SokołowskiS.,PizioO.,J.Phys.Chem.B,1999,103,3366;doi:10.1021/jp982028r. 24. NoworytaJ.P.,HendersonD.,SokołowskiS.,ChanK.-Y.,Mol.Phys.,1998,95,415;doi:10.1080/00268979809483175. 25. PatrykiejewA.,PizioO.,PusztaiL.,SokołowskiS.,Mol.Phys.,2003,101,2219;doi:10.1080/0026897031000099925. 26. PatrykiejewA.,PizioO.,SokołowskiS.,SokołowskaZ.,Phys.Rev.E,2004,69,061605; 27. BrykP.,SokołowskiS.,J.Chem.Phys.,2004,120,8299;doi:10.1063/1.1695554. 28. PizioO.,SokołowskiS.,SokołowskaZ.,J.Chem.Phys.,2011,134,214702;doi:10.1063/1.3597773. 29. IlnytskyiJ.M.,PatsahanT.,SokołowskiS.,J.Chem.Phys.,2011,134,204903;doi:10.1063/1.3592562. 30. BrykP.,PizioO.,SokołowskiS.,J.Chem.Phys.,2005,122,174906;doi:10.1063/1.1888425. 31. PizioO.,BuciorK.,PatrykiejewA.,SokołowskiS.,J.Chem.Phys.,2005,123,214902;doi:10.1063/1.2128701. 32. SokołowskiS.,RżyskoW.,PizioO.,J.ColloidInterfaceSci.,1999,218,341;doi:10.1006/jcis.1999.6392. 33. KrukM.,PatrykiejewA.,SokołowskiS.,Surf.Sci.,1995,340,179;doi:10.1016/0039-6028(95)00681-8. 34. MartinezA.,PatrykiejewA.,PizioO.,SokołowskiS.,J.Phys.:Condens.Matter,2003,15,3107; 35. GrabowskiK.,PatrykiejewA.,SokołowskiS.,ThinSolidFilms,1997,304,344;doi:10.1016/S0040-6090(97)00214-9. 36. KrukM.,PatrykiejewA.,SokołowskiS.,ThinSolidFilms,1994,238,302;doi:10.1016/0040-6090(94)90071-X. 37. BuciorK.,PatrykiejewA.,PizioO.,SokołowskiS.,J.ColloidInterfaceSci.,2003,259,209; 38. MartinezA.,PizioO.,SokołowskiS.,J.Chem.Phys.,2003,118,6008;doi:10.1063/1.1556850. 39. SałamachaL.,PatrykiejewA.,SokołowskiS.,J.Phys.Chem.B,2009,113,13687;doi:10.1021/jp901383v. 40. PatrykiejewA.,SokołowskiS.,J.Phys.Chem.B,2010,114,396;doi:10.1021/jp908710e. 41. PatrykiejewA.,SokołowskiS.,Condens.MatterPhys.,2014,17,43601;doi:10.5488/CMP.17.43601. 42. PatrykiejewA.,RżyskoW.,SokołowskiS.,J.Phys.Chem.C,2012,116,753;doi:10.1021/jp208323b. 43. PizioO.,RżyskoW.,SokołowskiS.,SokołowskaZ.,J.Chem.Phys.,2015,142,164703;doi:10.1063/1.4918640. 44. Suarez-MartinezI.,GrobertN.,EwelsC.P.,Carbon,2012,50,741;doi:10.1016/j.carbon.2011.11.002. 45. GeorgakilasV.,PermanJ.A.,TucekJ.,ZborilR.,Chem.Rev.,2015,115,4744;doi:10.1021/cr500304f. 46. TomanekD.,GuideThroughtheNanocarbonJungle,MorganandClaypoolPublishers,SanRafael,2014. 47. Carbon Nanomaterials as Adsorbents for Environmental and Biological Applications, Carbon Nanostructures doi:10.1103/PhysRevE.69.061605. doi:10.1088/0953-8984/15/19/312. doi:10.1016/S0021-9797(02)00203-5. Series,BergmannC.P.,MachadoF.(Eds.),SpringerInternationalPublishing,NewYork,2015; doi:10.1007/978-3-319-18875-1. 48. GogotsiY.,PresserV.,CarbonNanomaterials,CRCPress,BocaRaton,2013. 49. YeganegiS.,GholampourF.,ActaChim.Slov.,2012,59,888. 50. KowalczykP.,Chem.Phys.Phys.Chem.,2012,14,2784;doi:10.1039/c2cp23445a. 13003-12 doi:10.1039/c3cp52342j. doi:10.1088/0953-8984/26/48/485006. doi:10.1021/acs.jpclett.5b01545. 406208;doi:10.1088/0953-8984/19/40/406208. Chem.,2007,9,5919;doi:10.1039/b710552e. doi:10.1021/jp302776z. 7232;doi:10.1039/c4cp05498a. doi:10.1039/c0cp02281k. doi:10.1016/j.jcis.2013.01.044. doi:10.1016/j.uid.2013.10.013. Gyroidalnanoporouscarbons -- Adsorptionandseparationproperties 51. Furmaniak S., Terzyk A.P., Kowalczyk P., Kaneko K., Gauden P.A., Chem. Phys. Phys. Chem., 2013, 15, 16468; 52. Furmaniak S., Terzyk A.P., Gauden P.A., Kowalczyk P., Harris P.J.F., J. Phys.: Condens. Matter, 2014, 26, 485006; 53. Kowalczyk P., Terzyk A.P., Gauden P.A., Furmaniak S., Kaneko K., Miller T.F., J. Phys. Chem. Lett., 2015, 6, 3367; 54. TanakaH.,KanohH.,YudasakaM.,IijimaS.,KanekoK.,J.Am.Chem.Soc.,2005,127,7511;doi:10.1021/ja0502573. 55. Romo-HerreraJ.M.,TerronesM.,TerronesH.,DagS.,MeunierV.,NanoLett.,2007,7,570;doi:10.1021/nl0622202. 56. LiuL.,NicholsonD.,BhatiaS.K.,Chem.Eng.Sci.,2015,121,268;doi:10.1016/j.ces.2014.07.041. 57. Terzyk A.P., Furmaniak S., Gauden P.A., Harris P.J.F., Wloch J., Kowalczyk P., J. Phys.: Condens. Matter, 2007, 19, 58. Terzyk A.P., Furmaniak S., Harris P.J.F., Gauden P.A., Wloch J., Kowalczyk P., Rychlickia G., Chem. Phys. Phys. 59. Kowalczyk P., Gauden P.A., Terzyk A.P., Furmaniak S., Harris P.J.F., J. Phys. Chem. C, 2012, 116, 13640; 60. FurmaniakS.,KoterS.,TerzykA.P.,GaudenP.A.,KowalczykP.,RychlickiaG.,Chem.Phys.Phys.Chem.,2015,17, 61. Palmer J.C., Moore J.D., Roussel T.J., Brennan J.K., Gubbins K.E., Chem. Phys. Phys. Chem., 2011, 13, 3985; 62. Furmaniak S., Kowalczyk P., Terzyk A.P., Gauden P.A., Harris P.J.F., J. Colloid Interface Sci., 2013, 397, 144; 63. LuL.,WangS.,MullerE.A.,CaoW.,ZhuY.,LuX.,JacksonG.,FluidPhaseEquilib.,2014,362,227; 64. KowalczykP.,HołystR.,TerronesM.,TerronesH.,Phys.Chem.Chem.Phys.,2007,9,1786;doi:10.1039/b618747a. 65. MackayA.L.,Nature,1985,314,604;doi:10.1038/314604a0. 66. MackayA.L.,TerronesH.,Nature,1991,352,762;doi:10.1038/352762a0. 67. LijmaS.,IchihashiT.,AndoY.,Nature,1992,356,776;doi:10.1038/356776a0. 68. SchwartzH.A.,GesammelteMathematischeAbhandlunge,Vol.1and2,Springer,Berlin,1980. 69. TownsendS.J.,LenoskyT.J.,MullerD.A.,NicholsC.S.,ElserV.,Phys.Rev.Lett.,1992,69,921; 70. LeeH.G.,KimJ.,Int.J.Numer.MethodsEng.,2012,91,269;doi:10.1002/nme.4262. 71. HuM.,TianF.,ZhaoZ.,HuangQ.,XuB.,WangL.-M.,WangH.-T.,TianY.,HeJ.,J.Phys.Chem.C,2012,116,24233; doi:10.1021/jp3064323. 72. Kowalczyk P., He J., Hu M., Gauden P.A., Furmaniak S., Terzyk A.P., Phys. Chem. Chem. Phys., 2013, 15, 17366; doi:10.1039/c3cp52708e. 73. Kowalczyk P., Parsons D., Terzyk A.P., Gauden P.A., Furmaniak S., In: Carbon Nanomaterials Sourcebook: Nanoparticles,Nanocapsules,Nanobers,NanoporousStructures,andNanocomposites,Vol.2,SattlerK.D.(Ed.), CRCPress,London,2016,ch.6(inpress). 74. TylianakisE.,DimitrakakisG.K.,Martin-MartinezF.J.,MelchorS.,DobadoJ.A.,KlontzasE.,FroudakisG.E.,Int.J. HydrogenEnergy,2014,39,9825;doi:10.1016/j.ijhydene.2014.03.011. 75. WangH.,CaoD.,J.Phys.Chem.C,2015,119,6324;doi:10.1021/jp512275p. 76. HuangL.,YangX.,CaoD.,J.Phys.Chem.C,2015,119,3260;doi:10.1021/jp5128404. 77. YangZ.,PengX.,CaoD.,J.Phys.Chem.C,2013,117,8353;doi:10.1021/jp402488r. 78. HuangL.,XiangZ.,CaoD.,J.Mater.Chem.A,2013,1,3851;doi:10.1039/c3ta10292k. 79. HuangL.,ZengX.,CaoD.,J.Mater.Chem.A,2014,2,4899;doi:10.1039/C3TA15062C. 80. KarfunkelH.R.,DresslerT.,J.Am.Chem.Soc.,1992,114,2285;doi:10.1021/ja00033a001. 81. Zhao Z.S., Xu B., Wang L.M., Zhou X.F., He J.L., Liu Z.Y., Wang H.T., Tian Y.J., ACS Nano, 2011, 5, 7226; 82. RibeiroF.J.,TangneyP.,LouieS.G.,CohenM.L.,Phys.Rev.B,2006,74,172101;doi:10.1103/PhysRevB.74.172101. 83. FurmaniakS.,TerzykA.P.,GaudenP.A.,WlochJ.,KowalczykP.,Werengowska-CiećwierzK.,WiśniewskiM.,Har- risP.J.F.,J.Phys.:Condens.Matter,2015,28,015002;doi:10.1088/0953-8984/28/1/015002. 84. NicolaïA.,MontiJ.,DanielsC.,MeunierV.,J.Phys.Chem.C,2015,119,2896;doi:10.1021/jp511919d. 85. WernerJ.G.,HoheiselT.N.,WiesnerU.,ACSNano,2014,8,731;doi:10.1021/nn405392t. 86. ChoudhuryS.,AgrawalM.,FormanekP.,JehnichenD.,FischerD.,KrauseB.,AlbrechtV.,StammM.,IonovL.,ACS 87. BhattacharyaS.,GubbinsK.E.,Langmuir,2006,22,7726;doi:10.1021/la052651k. 88. FurmaniakS.,Comput.Meth.Sci.Technol.,2013,19,47;doi:10.12921/cmst.2013.19.01.47-57. doi:10.1103/PhysRevLett.69.921. doi:10.1021/nn202053t. Nano,2015,9,6147;doi:10.1021/acsnano.5b01406. 13003-13 015004;doi:10.1088/0953-8984/25/1/015004. S.Furmaniaketal. 89. Furmaniak S., Terzyk A.P., Gauden P.A., Kowalczyk P., Harris P.J.F., Koter S., J. Phys.: Condens. Matter, 2013, 25, 90. YanQ.,dePabloJ.J.,J.Chem.Phys.,1999,111,9509;doi:10.1063/1.480282. 91. FrenkelD.,SmitB.,UnderstandingMolecularSimulation,AcademicPress,SanDiego,1996. 92. SetoyamaN.,SuzukiT.,KanekoK.,Carbon,1998,36,1459;doi:10.1016/S0008-6223(98)00138-9. 93. KrukM.,JaroniecM.,GadkareeK.P.,Langmuir,1999,15,1442;doi:10.1021/la980789f. 94. ChomaJ.,JaroniecM.,ColloidsSurf.A,2001,189,103;doi:10.1016/S0927-7757(01)00572-6. 95. ChomaJ.,JaroniecM.,Adsorpt.Sci.Technol.,2007,25,573;doi:10.1260/0263-6174.25.8.573. 96. TylianakisE.,FroudakisG.E.,J.Comput.Theor.Nanosci.,2009,6,335;doi:10.1166/jctn.2009.1040. 97. HumphreyW.,DalkeA.,SchultenK.,J.Mol.Graphics,1996,14,33;doi:10.1016/0263-7855(96)00018-5. 98. Kaneko K., Ishii C., Kanoh H., Hanzawa Y., Setoyama N., Suzuki T., Adv. Colloid Interface Sci., 1998, 77, 295; doi:10.1016/S0001-8686(98)00050-5. Гiроїднiнанопористiвуглецi -- адсорбцiяiособливостi роздiлення,дослiдженiзвикористанням комп'ютерногомоделювання С.Фурманяк1,П.А.Гауден1,А.П.Тержик1,П.Ковальчик2 1 Хiмiчнийфакультет,Дослiдницькагрупафiзикохiмiївуглецевихматерiалiв, УнiверситетНiколаусаКопернiкусавТорунi,Торунь,Польща 2 Школаiнженерiїтаiнформацiйнихтехнологiй,УнiверситетМердока, Мердок,ЗахiднаАвстралiя,6150,Австралiя Адсорбцiя i особливостi роздiлення у гiроїдних нанопористих вуглецях (GNC), новому класi екзотичних нановуглецевих матерiалiв, вперше дослiджено, використовуючи метод моделювання Монте Карло з гi- перпаралельним темперуванням. Пориста структура GNC моделей оцiнюється методом, запропонова- нимБхатачарiяiГубiнсом.Всiдослiдженiструктуриєстрогомiкропористi.Крiмтого,механiзмиадсорбцiї Ar описуються на основi аналiзу iзотерм адсорбцiї, кривих ентальпiї, значень сталих Генрi, αs та кри- вих розподiлу потенцiалу адсорбцiї. Зроблено висновок, що при дiаметрах пор близьких або менших за 0.8 nm домiнує процес заповнення первинних мiкропор. Для структур, що мають бiльшi мiкропори, спостерiгається механiзм заповнення первинних та вторинних мiкропор. Нарештi, описано властивостi роздiленняCO2/CH4,CO2/N2 iCH4/N2 сумiшейуGNCiвиконанопорiвняннязвластивостямироздiлення моделей Вiртуального Пористого Вуглецю. Гiроїднi пористi вуглецi можна розглядати як потенцiйнi ад- сорбенти для роздiлення газових сумiшей, якi володiють аналогiчною або навiть вищою ефективнiстю роздiлення,нiжактивованийвуглецьзподiбнимдiаметромпор. Ключовiслова:гiроїднiнанопористiвуглецi,адсорбцiя,роздiленнягазовихсумiшей,моделювання МонтеКарло 13003-14
1708.05178	2	1708	2017-09-08T11:36:14	Isolation and Characterization of Few-layer Manganese Thiophosphite	[ "cond-mat.mes-hall" ]	This work reports an experimental study on an antiferromagnetic honeycomb lattice of MnPS$_3$ that couples the valley degree of freedom to a macroscopic antiferromagnetic order. The crystal structure of MnPS$_3$ is identified by high resolution scanning transmission electron microscopy. Layer dependent angle resolved polarized Raman fingerprints of the MnPS$_3$ crystal are obtained and the Raman peak at 383 cm$^{-1}$ exhibits 100% polarity. Temperature dependences of anisotropic magnetic susceptibility of MnPS$_3$ crystal are measured in superconducting quantum interference device. Magnetic parameters like effective magnetic moment, and exchange interaction are extracted from the mean field approximation mode. Ambipolar electronic transport channels in MnPS$_3$ are realized by the liquid gating technique. The conducting channel of MnPS$_3$ offers a unique platform for exploring the spin/valleytronics and magnetic orders in 2D limitation.	cond-mat.mes-hall	cond-mat	Isolation and Characterization of Few-layer Manganese Thiophosphite Gen Long1, Ting Zhang1,2, Xiangbin Cai1, Jin Hu3, Chang-woo Cho1, Shuigang Xu1,4, Junying Shen1, Zefei Wu1, Tianyi Han1, Jiangxiazi Lin1, Jingwei Wang1, Yuan Cai1, Rolf Lortz1, Zhiqiang Mao3, and Ning Wang1,* Affiliations: 1. Department of Physics and Center for Quantum Materials, The Hong Kong University of Science and Technology, Hong Kong, China 2. Institute for Advanced Study, The Hong Kong University of Science and Technology, Hong Kong, China 3. Department of Physics and Engineering Physics, Tulane University, New Orleans, LA- 70118, USA National Graphene Institute, University of Manchester, Manchester M139PL, UK 4. *Correspondence to: [email protected] Abstract: This work reports an experimental study on an antiferromagnetic honeycomb lattice of MnPS3 that couples the valley degree of freedom to a macroscopic antiferromagnetic order. The crystal structure of MnPS3 is identified by high-resolution scanning transmission electron microscopy. Layer-dependent angle-resolved polarized Raman fingerprints of the MnPS3 crystal are obtained and the Raman peak at 383 cm−1 exhibits 100% polarity. Temperature dependences of anisotropic magnetic susceptibility of MnPS3 crystal are measured in superconducting quantum interference device. Magnetic parameters like effective magnetic moment, and exchange interaction are extracted from the mean field approximation model. Ambipolar electronic transport channels in MnPS3 are realized by the liquid gating technique. The conducting channel of MnPS3 offers a unique platform for exploring the spin/valleytronics and magnetic orders in 2D limitation. Keywords: MnPS3, Raman spectroscopy, Antiferromagnetic, Magnetic susceptibility, Electronic transport 1 The magnetic order in 2D limitation is of boundless interest not only for fundamental condensed matter studies but also as a potential candidate in numerous technological applications1-7. Despite the extensive family of 2D crystals, only a few exhibit intrinsic magnetic orders8-10. Therefore, research has been mostly limited to the magnetic orders arising from extrinsic effects, such as vacancies, defects, edges or chemical dopants11-16. An emerging 2D crystal group, namely, transition-metal thiophosphite (TMT) (MPX3; M=Fe, Ni, Mn, Cd, Zn, etc.; P=P; X=S, Se, etc.), offers new possibilities because of the suitability of TMT as a platform for exploring novel intrinsic magnetic orders17-19. Different transition- metal ions in TMT accumulate different antiferromagnetic orders. FePS3 is best described by the Ising model, NiPS3 by the anisotropic Heisenberg model, and MnPS3 by the isotropic Heisenberg model5, 17. The intrinsic degrees of electronic freedom, such as charge and spin, have been broadly explored in the last few decades in electronics and spintronics20-23. In recent years, a new electron-valley freedom has drawn much attention because of its immense potential for fundamental studies on new quantum concepts and next-generation electronics24- 26. This valley freedom is predicted to couple with antiferromagnetic order in MnPS3 because of the latter's antiferromagnetic honeycomb lattice27. Coupling the micro-nature to the macro- phenomena renders MnPS3 an ideal playground for exploring novel electronic degrees of freedom. In this work, we perform a systematic optical and electronic transport study of MnPS3 in an atomically thin level. In addition to identifying the crystal structure with high- resolution scanning transmission electron microscopy (HRSTEM), Raman fingerprints of MnPS3 with different thickness were also determined through angle-resolved polarized Raman (ARPR) spectroscopy. Temperature dependence of magnetic susceptibility of MnPS3 is measured by a superconducting quantum interference device (SQUID). Critical magnetic parameters like effect magnetic momentum and exchange interactions are extracted from the mean field approximation (MFA) model. Last, we develop a liquid gating (LG) technique and fabricate MnPS3 -based electrical double-layer transistors (EDLTs) to determine its electronic transport properties. Benefiting from the high efficiency of LG28, an ambipolar conducting channel is observed with carrier mobility ranging from 1 cm2 V−1 s−1 to 3 cm2 V−1 s−1 in MnPS3. Top view of monolayer MnPS3 is illustrated in Fig. 1a (Upper left panel). Each [P2S6]4− unit is located at the center of a regular hexagon with six corners occupied by [Mn]2+. The dumbbell-shaped structure of [P2S6]4− is shown in Fig. 1b. All the [Mn]2+ are arranged in a honeycomb structure, and each [Mn]2+ is surrounded by six sulfur atoms. The zigzag direction of [Mn]2+ is defined as the a direction, whereas the armchair direction (perpendicular to a) is 2 defined as b. The point group of monolayer MnPS3 is assigned to be . The threefold inverse rotation symmetry results a valley degeneracy at the corner (K point) of the hexagonal first Brillouin zone (BZ). When stacked together along the c direction ( ), the atomic layers break the threefold inverse rotation symmetry and render the bulk MnPS3 a monoclinic structure with a point group of 2/m (Fig. 1a, right panel). Given the weak van der Waals interaction between atomic layers, the mono- and few-layered MnPS3 flakes can be mechanically exfoliated from the bulk crystal by scotch tape method29-30. Fig. 1c shows a representative micro-optical image of MnPS3, with different layers showing distinct optical contrasts due to light interference. To confirm the thickness of these flakes, we perform atomic force microscopy (AFM) measurement and a thickness h of 0.8 nm is obtained for one atomic layer (Fig. 1d), which agrees well with the layer space of 0.68 nm31. While exfoliating the bulk crystal, we find that most of (>90%) the flakes exhibit quadrilateral shapes having inner angles of ~60° or 120° (Fig. 1c). To elucidate the mechanism behind these two special angles, we compare between the selected-area electron diffraction (SAED) pattern and its defocused transmitted spot to index the edges of the exfoliated MnPS3 flakes. The MnPS3 flake is kept fixed along the c direction (Fig. 2a) and the focus is adjusted until the real space features are visible inside the transmitted spot (Fig. 2a, inset). The two edges forming 60° angle are indexed to be (110) and (020) according to the perpendicular relation between the reciprocal space and real space. The (110) and (020) directions correspond to the zigzag directions of [Mn]2+ in the MnPS3 structure. This result can be readily explained by the weakest breaking strength along the zigzag direction of [Mn]2+ (Ref: 32). To verify this conclusion, we obtain a high-angle annular dark field (HAADF) HRSTEM image of the same sample along the c direction (Fig. 2b). Each bright spot corresponding to a single atom, the three elements are indicated by different colored balls, and primary unit cell is marked by a green-dashed parallelogram. The six-membered ring composed of manganese atoms is clearly visible. The variation in intensity along the dashed lines are displayed in the right bottom inset. The lattice parameters extracted from the image (a=6.08±0.05 nm; b=10.52±0.05 nm) agree with the values measured from neutron scattering experiment31. Besides structural identification, we also examine the Raman fingerprints of MnPS3 with different layer numbers. Fig. 3a shows the Raman spectra of MnPS3 with different layers collected at room temperature. Two Raman peaks are apparent (P273: 273 cm−1; P383: 383.6 cm−1; the Raman peak at 302 cm-1 comes from silicon substrate) in the bulk samples. The 3 32/m107.5intensities of P273 and P383 decrease dramatically with decreasing layer number and disappear in the monolayer flake (Fig. 3b). The main Raman spectral features for the monolayer, bilayer, and trilayer flakes are summarized below. Monolayer flakes exhibit no observable Raman peaks. When layer number increases to two, both P273 and P383 emerge, and P383 is considerably weaker than P273. For three-layer flakes, the two peaks at 273 cm-1 and 383 cm-1 exhibit similar intensities. The clear dependence of Raman peak intensities on layer number renders the Raman spectrum a reliable criterion for determining the thickness of a few-layer MnPS3 sample. To probe further into symmetry properties, we perform ARPR spectral measurements on a 10 nm-thick MnPS3 flake at room temperature. The ARPR measurement configuration is shown in the top right inset of Fig. 4a. The a direction (green arrow) runs along one of the edges of the exfoliated MnPS3 flakes (zigzag direction of MnPS3), and the b direction (blue arrow) is perpendicular to a direction. The linear polarization direction of the incident laser (purple arrow) and scattered light (light blue arrow) was illustrated as well. The definitions of angles and are shown in the inset. Fig. 4a presents the ARPR spectra varying with when is fixed at 0°. P273 remains unchanged with increasing , whereas the peak at 383 cm- 1 vary periodically with . Moreover, P383 reaches its maximum at 0° and 180°, and disappears at around 90°. This pattern demonstrates a 100% polarity of Raman peak at 383 cm-1 (Ref: 33-34). We perform a fine (every 10°) dependent ARPR measurement and the extracted intensities of the two peaks are shown in Fig. 4b. The black solid line implies , as expected and measured in varying systems. By contrast, the red solid line confirms the depolarized feature of P273 34. The polarized and depolarized behaviors of Raman peaks demonstrate different symmetry properties of the corresponding phonon modes. Specifically, the polarized behavior of the Raman peak signifies that the peaks arise from totally symmetric variations. Fig. 4c presents the observed ARPR spectra at a few different with fixed at 0°. The independence of the ARPR spectrum on demonstrates that the crystalline orientation hardly affects the Raman spectrum of MnPS3. The phonon spectrum and corresponding Raman susceptibilities at the center of the first BZ are calculated through the density function perturbation theory (DFPT)35-38 to assign the Raman modes and phonon frequencies (supporting information). The bulk crystal of MnPS3 belongs to the C2/m symmetry group. The 30 irreducible phonon modes at the first BZ center are expressed as and confirmed by the calculation results. Polarized peak at 383 cm- 4 2()(0)cos()II8679guguAABB1 is assigned to the Bg mode, whereas the depolarized peak at 273 cm-1 is assigned to the Ag mode according to the symmetry elements of the Ag and Bg modes39. Fig. 5a reveals the temperature dependences of mass magnetic susceptibilities of MnPS3 bulk crystal for two directions, in-plane and out-plane magnetic fields, measured in SQUID. When temperature is above 200K, the isotropic and demonstrate a Heisenberg-type magnetic order in MnPS3 40. In the high temperature paramagnetic phase, follows the Curie-Weiss law , where C=2.65±0.05×10-4 m3Kkg-1 is the Curie constant and TC=-390K is the Curie temperature. In the MFA model, the effective magnetic moment of [Mn]2+ could be extracted from the Curie constant according to , where is the vacuum permeability, N is the primitive unit cell number per unit mass (also the number of [Mn]2+ pairs per unit mass), kB denotes the Boltzmann constant41-42. The derived agrees well with the total magnetic moment of 3d5 electron system in high spin state (S=5/2) 17. To confirm the obtained electronic state of Mn element in MnPS3, we performed the electron energy-loss spectroscopy (EELS) of few-layer flakes (Fig. 5c). The Mn-L2,3 edge energy-loss near-edge structures (ELNES) and the chemical shift agree well with the reported [Mn]2+ spectrum43. The value of demonstrates that manganese is in the form of ions rather than atomic form in MnPS3. As temperature cools down, the Curie-Weiss law fails to describe the behavior of . An isotropic broad peak of at 120K suggests a short range order of [Mn]2+ spins in the ab plane. Further cooling down the sample, sharply decreases to 0 while remains essentially constant singling the antiferromagnetic order in MnPS3 with TN=77K. TN is in consistence with the report value of 78K40. The contrasting behaviors between and agree well with the MFA model. The left (right) panel of Fig. 5d present the magnetic susceptibility measurement configuration with an applied magnetic field H perpendicular (parallel) to the spin orientations. When H is perpendicular to the spin orientations, the system energy density can be expressed as: , where is the Weiss constant, M=MA=MB is the magnetization strength for the [Mn]2+ with unit orientation, is the orientation of MA and MB caused by H. U reaches its minima at , hence the magnetic susceptibility with 5 m//mm//mmm/()mCCTTeff2023effBNCk05.6effB355.9dBeffmmm//mm//m22200(10.5(2))2UMHM0/2HMH perpendicular to spin orientations (in-plane magnetic field) is a constant. The observed slight increase of when cooling down (Fig. 5a) arises from the formation of spin density wave when temperature is lower than TN 17, 44. When H is parallel to the spin orientations (right panel of Fig. 5d), if MA and MB make equal angles with H, the magnetic field is not changed and the which agrees well with the measured results. The isotropic characteristics of at high temperature as well as the anisotropic characteristics of at low temperature confirms the 2D Heisenberg-type magnetic order with spins easy axis perpendicular to the ab plane in MnPS3 as shown in the inset of Fig. 5a. Another critical parameter of Heisenberg model, the main exchange interaction J is estimated from the high field . Fig. 5b presents the mass magnetization accompanied with the derivative mass magnetic susceptibility defined as with magnetic field perpendicular to the ab plane at T=5K. The mass magnetization exhibits a spin-flop transition feature at HC=3.8×106 A/m corresponding to the peak of . In a high field region above 5×106 A/m, the mass magnetization increase with increasing magnetic field linearly and =1.08×10-6 m3kg-1 is obtained. Using the MFA expression , where g=2.0 is the g-factor, z coordination number of [Mn]2+ and z=3 for honeycomb lattice, the main exchange interaction is extracted to be J/kB=-20.4 K40. Alternatively, J can also be obtained from Curie temperature TC through , where S=5/2 for a high spin state of 3d5 system. When TC=-390K, the value of J is extracted to be J/kB=-22.2K17. The negative exchange interaction signals the antiferromagnetic coupling between two adjacent [Mn]2+ ions. The universal main exchange interaction achieved from different methods demonstrates a reliable measurement of the magnetic properties of MnPS3. In other words, the universal exchange interaction demonstrates the validity of bare electron g-factor in MnPS3 system. From the practical viewpoint, the electronic transport properties of MnPS3 flakes are probed. A large MnPS3 band gap of more than 3 eV disables chemical potential tuning between the conduction and valence bands by using a field effect with SiO2 or boron nitride as a dielectric layer4. Hence, an LG technique is applied to fabricate the EDLTs, which exhibit two orders of increased efficiency in tuning the carrier density with respect to that of 6 0//2MH//m//(0)0Tmm1mddMmdHmdhmd20()2hBmdNgzJ32(1)BCkTJzSS300 nm-thick SiO2 28. The high efficiency of LG enables the observation of an ambipolar conducting channel in MnPS3. The two-terminal output curves (Fig. 6a) of a MnPS3 EDLT device of 13-nm channel thickness show large current modulations by both positive and negative LG voltages VLG. The channel current Ids exhibits super linear dependences on the excitation voltage Vds, which is likely caused by a large contact resistance. An enhanced contact quality is likely to improve the device performance. The top-left inset shows the optical image of an EDLT device, and the top-right inset presents the magnified image of the same device. Fig. 6b reveals the transfer curve along with the carrier mobility for both electrons and holes, where is the equivalent capacitance of LG28 and L and W are the length and width of the conducting channel, respectively. The dramatic increase in channel conductance G when VLG is higher (lower) than the positive (negative) threshold voltage demonstrates the ambipolar conducting operation of our EDLTs with high mobility for electrons (~1.3 cm2 V−1 s−1) and holes (~2.5 cm2 V−1 s−1). The inset displays the transfer curve in a logarithmic scale. When VLG varies from +4 V to −4 V, the channel switches from n-type "on" state to "off" state and changed to p-type "on" state with an on/off ratio more than 104. The on/off ratio is comparable with the values in widely studied TMDCs. The ambipolar performance, as well as the high on/off ratio, render MnPS3 a promising material for low-energy consuming devices with complementary logic. These desirable attributes are crucial to a superb noise margin and robust operation. The creation of an ambipolar conducting channel opens a new avenue for exploring the magnetic order in 2D antiferromagnetic systems. This channel also supplies an ideal terrace for probing valleytronics coupled to antiferromagnetic orders. In summary, few-layer MnPS3 crystal offers a productive platform for exploring antiferromagnetic order and valley degree of electron freedom in 2D limitation. Fundamental crystal parameters, such as lattice constants, layer-dependent Raman fingerprints and Raman peak anisotropy are measured by HRSTEM and ARPR techniques. Identifying the crystal structure and obtaining the Raman spectrum of MnPS3 flakes lay a solid foundation for follow-up research. The Heisenberg-type antiferromagnetic order in MnPS3 is confirmed by SQUID measurement, critical magnetic properties like effective magnetic momentum and exchange interactions are extracted from MFA model. The confirmation of ambipolar magnetoelectric transport in an antiferromagnetic semiconductor opens a new avenue for 7 1LGdGLCdVW28.4CFcmexploring fundamental correlated phenomena, i.e. spin/valleytronics coupled to an antiferromagnetic order. Experimental section Crystal synthesize The MnPS3 single crystals were prepared by a chemical vapor transport method. The stoichiometric mixture of Mn, P, and S powder was sealed in an evacuated quartz tube. Plate- like single crystals can be obtained via the vapor transport growth with a temperature gradient from 650 °C to 600 °C. The composition and structure of MnPS3 single crystals were checked by X-ray diffraction and Energy-dispersive X-ray spectrometer. TEM characterization The TEM samples are prepared by direct transfer from scotch tapes to 400-mesh copper grids after mechanical exfoliation. The SAED and EELS are carried out with a JEM 2010F (JEOL, Japan) under 200 kV while the HRSTEM is performed with a JEM ARM 200CF (JEOL, Japan), equipped with a CEOS probe corrector and a cold field-emission gun, also at an acceleration voltage of 200 kV for its highest resolution. A low probe current (less than 75 pA) is chosen to reduce electron radiation and a convergence semiangle of around 35mrad and an inner acquisition semiangle of 79 mrad were used in the HAADF STEM. The HRSTEM image is filtered through the standard Wiener deconvolution to increase its signal-noise ratio for a better display. The zero-loss peak in the EELS data is aligned to exact 0 eV and the Mn- L2,3 edge spectrum go through the power-law background subtraction after that. EDLT fabrication Thin flakes of MnPS3 are prepared by micro-mechanical exfoliation of a single bulk crystal. The thickness of MnPS3 flakes is verified by atomic force microscopy. Then the flakes are transferred on the top of another prepared boron nitride flake. Electron-beam lithography (EBL) is then applied to define Hall patterns on the MnPS3 flakes followed by electron-beam evaporation and lift-off techniques to deposit contact metals in the defined Hall patterns. Finally standard Hall devices with side gate electrodes are fabricated. The contact metal electrodes consist of Ti/Au/SiO2 (5 nm/60 nm/30 nm) while the side electrodes are not covered by SiO2. The contact metal electrodes are covered by SiO2 to isolate the electrodes 8 from direct contact with the ionic liquid. The sizes of the side-gate electrodes are much larger (area ratio >103) than those of the MnPS3 flakes. The large area ratio is designed to make sure that the voltage drop is effectively applied at the interface between MnPS3 flakes and the ionic liquid instead of between the side gate electrode and ionic liquid. Thanks to the thin electric layer (~1 nm) formed between the ionic liquid and graphene surface, an extremely strong electric field (~5×109 V/m) is generated to introduce an extremely high density of charge carriers to the samples (~5.23×1013 V-1cm-2). The ionic liquid is N, N-diethy1-N-(2- methoxyethy1)-N-methyl ammonium bis-(trifluoromethylsulfony1)-imide (DEME-TFSI). Acknowledgements Financial support from the Research Grants Council of Hong Kong (Project Nos. 16302215, 16300717, GRF16307114 and HKU9/CRF/13G) and technical support of the Raith-HKUST Nanotechnology Laboratory for the electron-beam lithography facility at MCPF are hereby acknowledged. Work at Tulane University was supported by the U.S. Department of Energy under Grant No. DE-SC0014208 (support for crystal growth). References Lin, M.-W.; Zhuang, H. L.; Yan, J.; Ward, T. Z.; Puretzky, A. A.; Rouleau, C. M.; Gai, 1. Z.; Liang, L.; Meunier, V.; Sumpter, B. G., Ultrathin nanosheets of CrSiTe 3: a semiconducting two-dimensional ferromagnetic material. Journal of Materials Chemistry C 2016, 4 (2), 315-322. 2. potential ferromagnetic 2D atomic crystal. 2D Materials 2016, 3 (2), 025035. 3. exfoliation of layered materials. Science 2013, 340 (6139), 1226419. 4. Zhang, X.; Zhao, X.; Wu, D.; Jing, Y.; Zhou, Z., MnPSe3 Monolayer: A Promising Tian, Y.; Gray, M. J.; Ji, H.; Cava, R.; Burch, K. S., Magneto-elastic coupling in a Nicolosi, V.; Chhowalla, M.; Kanatzidis, M. G.; Strano, M. S.; Coleman, J. N., Liquid Lee, J.-U.; Lee, S.; Ryoo, J. H.; Kang, S.; Kim, T. Y.; Kim, P.; Park, C.-H.; Park, J.- 2D Visible‐Light Photohydrolytic Catalyst with High Carrier Mobility. Advanced Science 2016, 3 (10). 5. G.; Cheong, H., Ising-Type Magnetic Ordering in Atomically Thin FePS3. Nano Letters 2016, 16 (12), 7433-7438. 6. Du, K.-z.; Wang, X.-z.; Liu, Y.; Hu, P.; Utama, M. I. B.; Gan, C. K.; Xiong, Q.; Kloc, C., Weak Van der Waals stacking, wide-range band gap, and Raman study on ultrathin layers of metal phosphorus trichalcogenides. ACS nano 2015, 10 (2), 1738-1743. 7. Gan, C. K.; Sengupta, P., Raman spectroscopy of atomically thin two-dimensional magnetic iron phosphorus trisulfide (FePS3) crystals. 2D Materials 2016, 3 (3), 031009. Wang, X.; Du, K.; Liu, Y. Y. F.; Hu, P.; Zhang, J.; Zhang, Q.; Owen, M. H. S.; Lu, X.; 9 Nair, R.; Sepioni, M.; Tsai, I.-L.; Lehtinen, O.; Keinonen, J.; Krasheninnikov, A.; Červenka, J.; Katsnelson, M.; Flipse, C., Room-temperature ferromagnetism in Žutić, I.; Fabian, J.; Sarma, S. D., Spintronics: Fundamentals and applications. Sandilands, L.; Shen, J.; Chugunov, G.; Zhao, S.; Ono, S.; Ando, Y.; Burch, K., de Jongh, L. J.; Miedema, A. R., Experiments on simple magnetic model systems. González-Herrero, H.; Gómez-Rodríguez, J. M.; Mallet, P.; Moaied, M.; Palacios, J. Ugeda, M. M.; Brihuega, I.; Guinea, F.; Gómez-Rodríguez, J. M., Missing atom as a 8. Stability of exfoliated Bi 2 Sr 2 Dy x Ca 1− x Cu 2 O 8+ δ studied by Raman microscopy. Physical Review B 2010, 82 (6), 064503. 9. Advances in Physics 1974, 23 (1), 1-260. 10. Willett, R. D.; Gatteschi, D.; Kahn, O., Magneto-structural correlations in exchange coupled systems. 1985. 11. J.; Salgado, C.; Ugeda, M. M.; Veuillen, J.-Y.; Yndurain, F.; Brihuega, I., Atomic-scale control of graphene magnetism by using hydrogen atoms. Science 2016, 352 (6284), 437-441. 12. Thomson, T.; Geim, A.; Grigorieva, I., Spin-half paramagnetism in graphene induced by point defects. Nature Physics 2012, 8 (3), 199-202. 13. McCreary, K. M.; Swartz, A. G.; Han, W.; Fabian, J.; Kawakami, R. K., Magnetic moment formation in graphene detected by scattering of pure spin currents. Physical review letters 2012, 109 (18), 186604. 14. graphite driven by two-dimensional networks of point defects. Nature Physics 2009, 5 (11), 840-844. 15. source of carbon magnetism. Physical Review Letters 2010, 104 (9), 096804. 16. Uchoa, B.; Kotov, V. N.; Peres, N.; Neto, A. C., Localized magnetic states in graphene. Physical review letters 2008, 101 (2), 026805. 17. PS 3 (M= Mn, Fe, and Ni). Physical Review B 1992, 46 (9), 5425. 18. Mayorga-Martinez, C. C.; Sofer, Z. k.; Sedmidubsky, D.; Huber, S. t. p. n.; Eng, A. Y. S.; Pumera, M., Layered Metal Thiophosphite Materials: Magnetic, Electrochemical, and Electronic Properties. ACS Applied Materials & Interfaces 2017. 19. Magnetoelectric MnPS 3 as a candidate for ferrotoroidicity. Physical Review B 2010, 82 (10), 100408. 20. Wolf, S.; Awschalom, D.; Buhrman, R.; Daughton, J.; Von Molnar, S.; Roukes, M.; Chtchelkanova, A. Y.; Treger, D., Spintronics: a spin-based electronics vision for the future. Science 2001, 294 (5546), 1488-1495. 21. Reviews of modern physics 2004, 76 (2), 323. 22. Nature materials 2008, 7 (3), 179-186. 23. Physics 2007, 3 (3), 153-159. 24. Mai, C.; Barrette, A.; Yu, Y.; Semenov, Y. G.; Kim, K. W.; Cao, L.; Gundogdu, K., Many-body effects in valleytronics: Direct measurement of valley lifetimes in single-layer MoS2. Nano letters 2013, 14 (1), 202-206. 25. Hall insulators and single-valley semimetals. Physical Review B 2013, 87 (15), 155415. 26. (8), 690-691. Li, X.; Cao, T.; Niu, Q.; Shi, J.; Feng, J., Coupling the valley degree of freedom to 27. antiferromagnetic order. Proceedings of the National Academy of Sciences 2013, 110 (10), 3738-3742. Bogani, L.; Wernsdorfer, W., Molecular spintronics using single-molecule magnets. Awschalom, D. D.; Flatté, M. E., Challenges for semiconductor spintronics. Nature Ezawa, M., Spin valleytronics in silicene: Quantum spin Hall–quantum anomalous Nebel, C. E., Valleytronics: Electrons dance in diamond. Nature materials 2013, 12 Joy, P.; Vasudevan, S., Magnetism in the layered transition-metal thiophosphates M Ressouche, E.; Loire, M.; Simonet, V.; Ballou, R.; Stunault, A.; Wildes, A., 10 Kurosawa, K.; Saito, S.; Yamaguchi, Y., Neutron diffraction study on mnps3 and Long, G.; Xu, S.; Shen, J.; Hou, J.; Wu, Z.; Han, T.; Lin, J.; Wong, W. K.; Cai, Y.; Claassen, H. H.; Selig, H.; Shamir, J., Raman Apparatus Using Laser Excitation and Long, G.; Maryenko, D.; Shen, J.; Xu, S.; Hou, J.; Wu, Z.; Wong, W. K.; Han, T.; Lin, 28. Long, G.; Xu, S.; Zhang, T.; Wu, Z.; Wong, W. K.; Han, T.; Lin, J.; Cai, Y.; Wang, N., Charge density wave phase transition on the surface of electrostatically doped multilayer graphene. Applied Physics Letters 2016, 109 (18), 183107. 29. Lortz, R., Type-controlled nanodevices based on encapsulated few-layer black phosphorus for quantum transport. 2D Materials 2016, 3 (3), 031001. 30. J.; Cai, Y., Achieving Ultrahigh Carrier Mobility in Two-dimensional Hole Gas of Black Phosphorus. Nano Letters 2016. 31. feps3. Journal of the Physical Society of Japan 1983, 52 (11), 3919-3926. Liu, E.; Fu, Y.; Wang, Y.; Feng, Y.; Liu, H.; Wan, X.; Zhou, W.; Wang, B.; Shao, L.; 32. Ho, C.-H., Integrated digital inverters based on two-dimensional anisotropic ReS2 field-effect transistors. Nature communications 2015, 6. 33. Polarization Measurements. Rotational Spectrum of Fluorine. Applied Spectroscopy 1969, 23 (1), 8-12. 34. Spectroscopy 1970, 24 (3), 348-353. 35. Gonze, X.; Jollet, F.; Araujo, F. A.; Adams, D.; Amadon, B.; Applencourt, T.; Audouze, C.; Beuken, J.-M.; Bieder, J.; Bokhanchuk, A., Recent developments in the ABINIT software package. Computer Physics Communications 2016, 205, 106-131. 36. Bruneval, F.; Caliste, D.; Caracas, R.; Côté, M., ABINIT: First-principles approach to material and nanosystem properties. Computer Physics Communications 2009, 180 (12), 2582-2615. 37. Kristallographie-Crystalline Materials 2005, 220 (5/6), 558-562. 38. Sindic, L.; Verstraete, M.; Zerah, G.; Jollet, F., First-principles computation of material properties: the ABINIT software project. Computational Materials Science 2002, 25 (3), 478- 492. 39. Wu, J.; Mao, N.; Xie, L.; Xu, H.; Zhang, J., Identifying the Crystalline Orientation of Allemand, C. D., Depolarization ratio measurements in Raman spectrometry. Applied Gonze, X.; Amadon, B.; Anglade, P.-M.; Beuken, J.-M.; Bottin, F.; Boulanger, P.; Gonze, X., A brief introduction to the ABINIT software package. Zeitschrift für Gonze, X.; Beuken, J.-M.; Caracas, R.; Detraux, F.; Fuchs, M.; Rignanese, G.-M.; Okuda, K.; Kurosawa, K.; Saito, S.; Honda, M.; Yu, Z.; Date, M., Magnetic properties Sales, B.; Wohlleben, D., Susceptibility of interconfiguration-fluctuation compounds. Hamaker, H.; Woolf, L.; MacKay, H.; Fisk, Z.; Maple, M., Coexistence of Black Phosphorus Using Angle‐Resolved Polarized Raman Spectroscopy. Angewandte Chemie 2015, 127 (8), 2396-2399. 40. of layered compound mnps3. Journal of the Physical Society of Japan 1986, 55 (12), 4456- 4463. 41. Physical Review Letters 1975, 35 (18), 1240. 42. superconductivity and antiferromagnetic order in SmRh4B4. Solid State Communications 1979, 32 (4), 289-294. 43. shift investigation in transition metal oxides by EELS. Ultramicroscopy 2012, 116, 24-33. Breed, D., Antiferromagnetism of K2MnF4. Physics Letters 1966, 23 (3), 181-182. 44. Tan, H.; Verbeeck, J.; Abakumov, A.; Van Tendeloo, G., Oxidation state and chemical 11 Fig. 1 Crystal structure of MnPS3. (a) Ball-and-stick model of the MnPS3 crystal structure. The left panel shows the top view of the monolayer model. The green parallelogram indicates the primary unit cell. The arrows show the crystalline orientations (red: a; blue: b; greed: c). The right panel displays the side view of the bilayer model, and the green parallelepiped indicates the primary unit cell of the MnPS3 bulk crystal. (b) The dumbbell-shaped structure of the [P2S6]4− unit is marked by the dashed light blue circle in (a). (c) Micro-optical image of an exfoliated few-layer MnPS3. The scale bar denotes 5 µm. (d) Measured thicknesses h and corresponding layer numbers. The red dashed line represents the linear fitting result. 12 Fig. 2 Transmission electron microscopy (TEM) images of MnPS3. (a) SAED pattern of MnPS3. The scale bar represents 5 nm−1. The inset shows the defocused diffraction pattern at the same site. The real space features are apparent in the defocused transmitted spot. The light green dashed lines indicate the two edges of the MnPS3 sample, whereas the yellow dashed lines show the directions perpendicular to the two edges. A magnetic declination of 3° is corrected. The two edges are indexed to be (110) and (020) according to the diffraction patterns. (b) HAADF STEM image of MnPS3 sample along c direction. The scale bar represents 0.5 nm. The colored balls indicate the different elements in the image (purple: Mn; black: P; yellow: S). The dashed green parallelogram denotes the primary unit cell. The inset indicates that the intensities vary with distance along the directions, as marked by the dashed lines with the same colors. A constant offset between different lines is introduced for a clear display. 13 Fig. 3 Thickness-dependent Raman spectrum of MnPS3. (a) Raman spectra of MnPS3 with different thicknesses (monolayer to bulk). The Raman intensities are normalized to the peak intensity of substrate Si. (b) MnPS3 Raman peak intensities change with thickness. The peak intensities of the bulk sample is regarded as one unit. Fig. 4 Angle-resolved polarized Raman spectrum of MnPS3. (a) Polarized Raman spectra of MnPS3 with . The top-right inset shows the ARPR spectrum measurement configuration. The green and blue arrows indicate the a and b directions, respectively. The purple and light blue arrows indicate the linear polarization directions of the incident and scattered lights, respectively. Angle is the angle between the incident and scattered lights, whereas angle is the angle between the scattered light and the a direction. (b) -dependent polarized Raman peak intensities with (Red: P283; green: P383). The peak intensities at is considered as one unit. The red and black solid lines display the fitting results of constant intensity and , respectively. (c) -dependent polarized Raman spectrum with . 14 0002cos()0 Fig. 5 Magnetic susceptibility of MnPS3. (a) Temperature dependences of mass magnetic susceptibility with in-plane (blue) and out-plane (red) magnetic fields of 1600 A/m. The green line shows the fitting result of Curie-Weiss formula. The purple arrow indicate the critical temperature between paramagnetic and antiferromagnetic orders. The green arrow indicate the temperature corresponding to the peak value of . The inset displays the magnetic order of MnPS3 in antiferromagnetic order regime (blue: spin up; red: spin down; or inverse). (b) Mass magnetization (red) and derivate mass susceptibility (blue) as functions of out-plane magnetic field at T=5K. The three inset patterns display the magnetic orders at magnetic field ranges (blue: spin up; red: spin down; or inverse). The green arrows indicate the peak positions of derivate mass susceptibility. (c) Electron energy loss spectrum (EELS) of the few-layer sample shows characteristic peaks from [Mn]2+. (d) Mean field approximation (MFA) model for in- plane (left) and out-plane (right) field magnetic susceptibility in antiferromagnetic regime. MA and MB indicate the magnetizations with opposite orientations, H is the applied magnetic field. 15 mm Fig. 6 Transport features of electrical double layer transistor (EDLT) based on MnPS3. (a) Output curves of MnPS3 at different LG voltages. Current of electrons (positive LG voltage) are shown with positive indices, and those of electrons (negative LG voltage) are shown with negative indices. The two insets show the micro-optical images of the EDLT in different ratios. The scale bars denote 300 µm (left) and 3 µm (right). (b) Room temperature transfer characteristics (blue line) and the extracted field-effect mobility as a function of VLG (red circles) of the MnPS3 EDLT. The green dashed lines show the linear fitting result of transfer characteristics. The inset shows the transfer characteristics in the log scale. 16 TOC: Raman spectroscopy as long as the electronic transport properties of the hexagonal antiferromagnetic lattice in MnPS3 is studied. Raman fingerprints of MnPS3 with different thicknesses are obtained. Raman spectrum evolution of when cooling down to cryogenic temperature reveals the universal magnetic orders in MnPS3. A liquid gating technique is applied to probe the transport properties of MnPS3. 17
1507.07090	2	1507	2015-09-19T10:34:06	Spatial dispersion effects upon local excitation of extrinsic plasmons in a graphene micro-disk	[ "cond-mat.mes-hall" ]	Excitation of surface plasmon waves in extrinsic graphene is studied using a full-wave electromagnetic field solver as analysis engine. Particular emphasis is placed on the role played by spatial dispersion due to the finite size of the two-dimensional material at the micro-scale. A simple instructive set up is considered where the near field of a wire antenna is held at sub-micrometric distance from a disk-shaped graphene patch. The key-input of the simulation is the graphene conductivity tensor at terahertz frequencies, being modeled by the Boltzmann transport equation for the valence and conduction electrons at the Dirac points~(where a linear wave-vector dependence of the band energies is assumed). The conductivity equation is worked out in different levels of approximations, based on the relaxation time ansatz with an additional constraint for particle number conservation. Both drift and diffusion currents are shown to significantly contribute to the spatially dispersive anisotropic features of micro-scale graphene. More generally, spatial dispersion effects are predicted to influence not only plasmon propagation free of external sources, but also typical scanning probe microscopy configurations. The paper set the focus on plasmon excitation phenomena induced by near field probes, being a central issue for the design of optical devices and photonic circuits.	cond-mat.mes-hall	cond-mat	Spatial dispersion effects upon local excitation of extrinsic plasmons in a graphene micro-disk D. Mencarelli1,2, S. Bellucci2, A. Sindona2,3, L. Pierantoni1,2 1Universit`a Politecnica delle Marche, Ancona, Italy 2Laboratori Nazionali di Frascati (LNF -- INFN), Frascati, Roma, Italy 3Dipartimento di Fisica, Universit`a della Calabria, Via P. Bucci Cubo 30C, Rende (CS), Italy Excitation of surface plasmon waves in extrinsic graphene is studied using a full-wave electro- magnetic field solver as analysis engine. Particular emphasis is placed on the role played by spatial dispersion due to the finite size of the two-dimensional material at the micro-scale. A simple instruc- tive set up is considered where the near field of a wire antenna is held at sub-micrometric distance from a disk-shaped graphene patch. The key-input of the simulation is the graphene conductivity tensor at terahertz frequencies, being modeled by the Boltzmann transport equation for the valence and conduction electrons at the Dirac points (where a linear wave-vector dependence of the band energies is assumed). The conductivity equation is worked out in different levels of approximations, based on the relaxation time ansatz with an additional constraint for particle number conservation. Both drift and diffusion currents are shown to significantly contribute to the spatially dispersive anisotropic features of micro-scale graphene. More generally, spatial dispersion effects are predicted to influence not only plasmon propagation free of external sources, but also typical scanning probe microscopy configurations. The paper set the focus on plasmon excitation phenomena induced by near field probes, being a central issue for the design of optical devices and photonic circuits. Keywords: Surface plasmon, spatially dispersive media, wire antenna, graphene conductivity, electromagnetic response at terahertz frequencies I. INTRODUCTION One of the main challenges for current device electron- ics is the full exploitation of the terahertz (THz) electro- magnetic (EM) spectrum, bridging the gap beitween the microwaves and optics. In this context, graphene and graphene-derived materials possess a number of unique EM properties such as tunable conductivity and slow- wave features that may be used to develop high perfor- mance THz devices1 -- 7. More generally, low-dimensional systems with honeycomb-like geometry and related as- semblies or hetero-structure are undergoing massive in- vestigation as materials capable of supporting plasmon propagation in the THz frequency range8 -- 12. A THz beam impinging on a metal atomic force micro- scope tip has been used to generate guided THz waves on graphene in a recent experiment8: the nanometric curvature of the tip makes light scatter on the length scale of tens of nanometers2, providing the EM field to- gether with the wave-number components required to ex- cite short wavelength surface plasmons. A similar situa- tion has been encountered at lower frequencies in scan- ning microwave microscopy (SMM)13, where a metallic tip can act as a resonant antenna whose tip-termination is directly coupled with the sample. In this paper, we focus on numerical propagation and excitation of plasmon surface waves in spatially disper- sive graphene as seen from the macroscopic EM perspec- tive. We use the fact that a momentum matching be- tween the field source and the surface plasmon can be achieved, e.g., by grating or prism couplers. In partic- ular, we present an EM scheme that ideally recalls the set-up of Ref.13 and take a resonant wire antenna placed at sub-micrometric distance from a disk-shaped graphene patch. Our analysis suggests that the correction to the con- ductivity response given by spatial dispersion is strongly required, at least in excitation phenomena induced by a strong spatial gradient of the EM field on the sample. We therefore asses that spatial dispersion is an impor- tant limiting factor to the achievable coupling between surface-plasmons and the near field of an SMM tip, be- cause it produces a broadening of the system-response over a wide range of wave-vectors. The high impact of spatial dispersion in propagation and modal analysis of plasmons has been previously scrutinized14 -- 16. However, a similar study has not been conducted so far in near field excitation problems, which are of central importance in numerous practical applications including microscopy and photonics. Here, we provide such a characterization. From the numerical point of view, we simulate surface plasmon excitation in an open computational domain avoiding periodic boundary conditions and plane-wave excitations5. The typical high charge mobility of carbon nanostructures leads to a huge increase of the macro- scopic inductance, i.e., a high reactive energy stored per unit length and, accordingly, a slow-wave propagation. Consequently, the plasmon velocity can be much smaller 5 1 0 2 p e S 9 1 ] l l a h - s e m . t a m - d n o c [ 2 v 0 9 0 7 0 . 7 0 5 1 : v i X r a than the speed of light with very small wavelengths as compared to that of the coupled feeding antenna. This also results in a high confinement of the EM field in the direction transverse to the propagation, which further increases the aspect ratio. We then apply two differ- ent numerical methods to validate the results, i.e., the method of moments (MoM) and the method of finite ele- ments (FEM). In this way, we provide a comparison of the different levels of approximation used in semi-classical approaches to the graphene conductivity, based on the Boltzmann transport equation for the Dirac electrons and the Kubo-formula17 -- 21. to zero energy. In this way, µ coincides with the Fermi level shift caused by the doping (or gating). 2 The intra-band components of σrta have the dyadic (tensor) form15 σ± rta(q) = ie2 2π2 f(cid:48) ε±(k)−µ v±(k)v±(k) ω − v±(k) · q d2k , (1) (cid:90) 1st BZ where: (i) the band energies are linearized as ε±(k) = ±vfk, with v±(k) = ∇kε±(k) representing the elec- tron velocities and vf = v±(k) the Fermi velocity; (ii) f(cid:48) ε±(k)−µ is the first energy-derivative of the FD, i.e., (cid:12)(cid:12)(cid:12)(cid:12)ε=ε±(k) = − 4kT cosh2(cid:16) ε±(k)−µ 1 kT (cid:17) ; II. THEORY (cid:48) ε±(k)−µ = f ∂fε−µ ∂ε The dispersive conductivity of graphene on the THz regime is well defined in the literature1,15,16: basically, an electric field frequency below a few THz induces a re- sponse of the valence (π) and conduction (π∗) electrons of the material with an energy close to the Fermi energy. The dominant part of the response in extrinsic (doped or gated) graphene is given by intra-band excitations, which can be treated by the Boltzmann transport equa- tion under the relaxation-time-approximation (RTA) or Bhatnagar-Gross-Krook (BGK) model. The RTA replaces the relaxation dynamics of each one- electron state by a simple exponential decay. Then, the transport relaxation time is approximated with the life- time of the state. Such a formalism provides an accurate description of spatial dispersion effects in doped graphene with extrinsic Fermi energy shifts below ∼ 0.5 eV. Ac- cordingly, the RTA conductivity is entirely determined by the drift currents that arise from one-electron transitions and collective modes within the π and π∗ bands. The BGK model is more general than the RTA, be- cause it allows for an extra degree of freedom, which enforces charge conservation and accounts properly for electron diffusion15,16. We begin by specializing to the RTA conductivity σrta(q) = σ+ rta(q) + σ− rta(q) under an applied electric field with momentum q and frequency f . σrta includes the contributions of elec- trons (with charge −e and wave-vector k) that occupy the π∗(+) and π(-) bands. The band energies ε±(k) are populated according to the Fermi Distribution (FD) fε±(k)−µ = 1 ε±(k)−µ kT 1 + e at the absolute temperature T and chemical potential µ. We adopt the widely used convention of setting the Dirac point energy of graphene, i.e. the intrinsic Fermi level, (iii) ω = ω − iγ is a complex frequency that includes the angular frequency ω = 2πf and a small shift along the imaginary axes, which corresponds to the electron damp- ing rate γ; the latter depends on the average relaxation time τ as γ = 2π/τ . (1), In Eq. the (two-dimensional) first Brilluoin Zone (BZ) integration is well defined on circular areas of the k-space centered at the Dirac points, where the linear approximation for the band energies is valid. Neverthe- less, in most practical uses, an infinite cone-structure is assumed for the valence and conduction bands; in other words, the wave-vector integral is performed over the whole k-space for analytical convenience and a factor of 2 is included to account for the inequivalent Dirac point. Then, a change of variable from wave-vector to the energy leads to the non-dispersive conductivities σ± rta(0) = σ ± intraI, in which dkkf ie2v2 f π ω ± intra = σ (cid:48) ε±(k)−µ = (cid:48) ε−µ and I denotes the 2×2 identity matrix. The total intra- band conductivity at q → 0 is then dεεf 0 0 ±ie2 π2 ω (cid:90) ∞ (cid:90) ∞ (cid:90) ∞ − intra σintra = σ+ intra + σ ie2 π2 ω = 0 dεε(f ε−µ − f (cid:48) (cid:48) −ε−µ) (2) Interestingly enough, the RTA conductivity (1) and its q → 0 form (2) can be derived from the Kubo formula in the optical limit17 -- 21. Another part of the graphene conductivity is related to inter-band processes between the π and π∗ bands, and is also included in the Kubo formulism18 -- 21, though it does not contain the dispersive term v±(k)· q. Indeed, it has been pointed out that for surface waves supported by isolated graphene sheets, and working frequencies below a few THz, the spatial dispersion effects on inter-band transitions can be neglected15,16. Due to the absence of the dispersive term, the inter-band conductivity is a scalar, which can be expressed as (cid:90) ∞ 0 fε−µ − f−ε−µ 1 −(cid:0) 2εω (cid:1)2 dε ω σinter = ie2 π . (3) we obtain the tensor components of the intra-band con- ductivity σ± rta(q) = ¯σ ± rta(q)I + ¯¯σ ± rta(q) sin 2θq sin 2θq − cos 2θq , (7) (cid:20) cos 2θq which depend on the scalar conductivities 3 (cid:21) Therefore, the total conductivity reads σ(q) = σrta(q) + σinterI. (4) In the q → 0-limit the latter tends to the non-dispersive Kubo conductivity18 -- 21: and σk = σintra + σinter. (5) Now, looking at the denominator in Eq. (1), it is clear that when v±(k)·q and ω are comparable, spatial disper- sion cannot be neglected in surface plasmon excitation. This happens particularly in problems where a resonant behavior of the EM field is concerned. It is the case of some of the examples reported in this work, where the interaction between a radiating antenna and a graphene patch takes place via near field coupling. To give an idea of the numbers involved, the spatial harmonics of the excitation field become significant in the conductivity response at an operating frequency f of the order of ∼ 1 THz, for an applied wave-vector value q larger than ∼ 1 µm−1. At the same frequency, the elec- tric field wavelength is about ∼ 300 µm. However, the near field distribution between the tip of the antenna and the graphene sample varies on a sub-micrometric scale, depending on the tip radius and distance from the sample. Then, the slow-wave effect featured by plasmon propagation implies a wavelength reduction of more than one order of magnitude, and ensures the matching and coupling with the exciting near field. Eq. (1) can be simplified and made more explicit by expressing the planar wave-vectors q and k in polar co- ordinates, say, q = (q, θq) and k = (k, ϑk). Then, as shown in the appendix, we can reduce it to the following expression σ± rta(q) = ie2v2 f 2π2 dkkf (cid:48) ε±(k)−µ (cid:20) cos 2ϑk I + × dϑk 0 sin 2ϑk sin 2ϑk − cos 2ϑk ω − vfq cos(ϑk − θq) (6) (cid:21) , (cid:90) ∞ (cid:90) 2π 0 where the BZ integral has been turned to a wave-vector integral over the whole k-space (with the factor of 2 from the inequivalent Dirac points being included). After some straightforward manipulations on the angu- lar integral in Eq. (6) that are reported in the appendix, (cid:90) ∞ ± rta(q) = ¯σ ie2v2 f 2π2 ω (cid:90) ∞ 0 ± rta(q) = ¯¯σ ie2v2 f 2π2 ω (cid:48) dkkf ε (cid:90) 2π 0 0 × ± k −µ dθk ω − (vfq/ω) cos θk dθk cos 2θk ω − (vfq/ω) cos θk . (cid:48) kf ε (cid:90) 2π ± k −µ × 0 (cid:112) These two quantites may rewritten as ± rta(q) = ¯σ σ ± intra 1 − v2 fq2/ω2 and (cid:16)(cid:112) ± rta(q) = ¯¯σ (cid:17)2 fq2/ω2 − 1 fq2/ω2 1 − v2 (cid:112) 1 − v2 Therefore, summing over the ± channels, we find ¯σrta(q) = ¯σ+ rta(q) + ¯σ − rta(q) = ± ω2σ intra v2 fq2 . (cid:112) σintra 1 − v2 fq2/ω2 and ¯¯σrta(q) = ¯¯σ+ (cid:16)(cid:112) rta(q) + ¯σ 1 − v2 (cid:112) (cid:17)2 − rta(q) fq2/ω2 − 1 fq2/ω2 1 − v2 = ω2σintra vfq2 . Then, ¯σrta and ¯¯σrta turn out to be both proportional to the non-dispersive intra-band conductivity σintra of Eq. (2), which evaluates exactly to σintra = −iξ0/ω with ξ0 = e2 π2 [µ + 2kT ln(e − µ kT + 1)]. Incidentally, we notice that the conductivity ¯¯σ± rta, being markedly anisotropic, is proportional to ω2σintra. This makes its imaginary part change sign with increasing the frequency, in contrast to the purely inductive na- ture of ¯σ± rta, whose imaginary part keeps a negative sign in the sampled frequency range. The plots of Fig. 1(A) and 1(B) illustrate such a behavior. Now, by definition, surface waves are sustained at the interface between ma- terials having different permittivities so that, in a trans- verse resonance circuit, their associated reactances cancel 4 calculation. Then, the solution for the resulting Elec- tric Field Integral Equation (EFIE) is generalized to the case of spatially dispersive material16,20,21, by using the proper constitutive relation in the spatial domain, i.e., Js(r) = (cid:48)σ(r − r(cid:48) ) · Es(r(cid:48) ). d2r (11) (cid:90) In this equation, the planar current Js, and the tangent electric field Es, are sampled over the graphene surface. The non-local behavior of the current-field relation be- comes important when rapid field variations are involved. Transforming the conductivity in real space, the input spatial response of dispersive graphene is obtained as σ(r) = I q¯σ(q)J0(qr) + Iσinter (12) (cid:90) ∞ (cid:18) cos 2φ sin 2φ dq 2π 0 sin 2φ − cos 2φ − (cid:19)(cid:90) ∞ dq 2π q¯¯σ(q)J2(qr), 0 where the scalar conductivities ¯σ and ¯¯σ include both the ±-contributions from Eq. (7) or Eq. (9). In Eq. (12) the first and second terms contributing to the conduc- tance, are weighted by the zero-order and second-order Bessel functions, respectively. This means that the first addend of the conductivity is more sensitive to slow vary- ing fields with respect to the second addend. Equally im- portantly, the second term of the conductivity, containing off-diagonal matrix elements, is characterized by an an- gular dependence related to a current response, which is locally weighted by a "quadrupole" spatial distribution of the EM field. Consequently, unless very high angular and radial variations of the EM fields are concerned, the second term can be neglected. Let us now turn to the main application of the present work (Fig. 2) that is a circular graphene disk of diameter D = 2R, with a wire antenna of length L, placed just above its center. For this system, we can safely assume cylindrical symmetry. The gap between the antenna and the disk is L/200. In absence of angular variation of the EM excitation, the off diagonal terms vanish and no an- gular current arises. In addition, with a not too small tip-sample distance, and, thus, a not too strong EM field variation, the ¯σ-term in Eq. (12) is expected to be dom- inant. Under these limiting conditions, Eq. (11) can be approximated by Jr = (cid:48) ¯σ(r − r(cid:48))Es(r(cid:48) ) d2r (13) (cid:90) where Jr is the radial component of the surface current, and ¯σ(r) is the real space representation of ¯σ(q), which gives the (spatial) impulsive response of the current af- ter the EM excitation. Note that usual assumption of a thin hollow cylinder to approximate the wire antenna may affect the near field distribution between the tip FIG. 1. Real and imaginary parts of the scalar conductivities ± rta (A) and ¯¯σ ¯σ (7), and reported vs the dimensionless frequency ω/qvf. The two quantities are normalized to ξ0/qvf. ± rta (B) calculated from the RTA model out. Therefore, the contribution of the anisotropic term ¯¯σ± rta may strongly affect the polarization of surface plas- mons22 of high rate of spatial variation. To include the diffusion currents, we introduce the fol- lowing tensor quantity −iγ ω s±(q) = (cid:90) 1st BZ f(cid:48) ε±(k)−µ v±(k)q ω − v±(k) · q , d2k (8) where (cid:48) f ε±(k)−µ = (cid:82) f(cid:48) ε±(k)−µ d2kf(cid:48) . ε±(k)−µ 1st BZ The BGK correction to the RTA conductivity can be put in the form σ± bgk(q) = [I + s±(q)] −1σ± rta(q). (9) This is equivalent to the result presented in Ref.15, and leads to correct the total conductivity as σ(q) = σbgk(q) + σinterI. (10) The numerical results reported in the following are obtained by both a full-wave solver and a semi- analytical approach, as implemented respectively by the finite element method (FEM) and the method of mo- ments (MoM). The latter employs the usual free-space Green's tensor G, in cylindrical coordinates r = (r, ϕ) and z, as the kernel of an integral operator relating the electric current density and field20,21: E(r, z) = d2r (cid:48)G(r − r(cid:48) , z − z (cid:48) ) · J(r(cid:48) (cid:48) ). , z dz (cid:90) (cid:48)(cid:90) Here, it should be noticed that the discretized currents in space are assumed as independent variables in the MoM 0.00.51.5ωqvF2.0-2-10123=0.025(B)γqvFqvFσrtaξ0ImRe0.00.51.52.0-2024(A)ωqvF=0.025γqvFReqvFσrtaξ0Imand graphene, but it does not limit the generality of the present analysis. III. NUMERICAL RESULTS 5 In the following, we present the solution for the EFIE directly in real space, focussing on both the full tensor form (12) and the scalar form (13) of the surface conduc- tivity. A. Non-Dispersive Analysis of Surface Plasmons We begin by considering an example of plasmon ex- citation without spatial dispersion. Let us define the graphene surface impedance Zs as the reciprocal of the non-dispersive Kubo conductivity σk introduced in Sec. II. We then have Zs = 1/σk or 1/Zs = σintra + σinter, where the intra-band and inter-band terms have been re- spectively given in Eqs. (2) and (3). FIG. 2. Wire antenna placed at sub-micrometric dis- tance (L/200) from a graphene patch For a fixed frequency, the effect of spatial dispersion increases with increasing the charge scattering-time. As a practical example, we take the cumulative integral of the impulsive response (cid:90) C(r) = (cid:90) r (cid:48) d2r ¯σ(r (cid:48) ) = 2π (cid:48) (cid:48) r dr (cid:48) ), ¯σ(r r(cid:48)<r 0 i.e., the current density response to a uniform unit elec- tric field within a circular area of radius r < R. Its profile, normalized to σintra, is shown in Fig. 3 for two different scattering times (τ = 1, 2 ps), at f = 10 THz. FIG. 4. Plasmon distribution (normalized electric field) from a linear antenna coupled to a graphene disk (sketched in Fig. 2) vs the radial distance from the disk center and ReZs/ImZs. The plasmon distribution for a graphene disk of diam- eter D, coupled to an antenna of length L = D ≈ c/2f , is reported in Fig. 4 vs the radial position within the disk and the inverse plasmon "quality factor ". The latter is defined as the ratio between imaginary and real parts of Zs calculated by the non-dispersive Kubo conductivity, i.e., ImZs/ReZs = −Reσk/Imσk. In Fig. 4, both the op- eration frequency and the graphene chemical potential follow from the choice of the surface reactance ImZs and plasmon quality-factor. For example, a reactance of 5 kΩ and a quality factor of 11 are associated to a chemical potential of 0.08 eV and a frequency of about 7.4 THz. We see that the spatial oscillations and propagation of the plasmon expire, at a progressively smaller radial dis- tance from the feeding tip, as ReZs/ImZs increases from 0 to 0.15. FIG. 3. Cumulative integral of the impulsive response C/σintra as a function of radial position r, normalized to the free-space wavelength λ; two different scattering times (τ = 1, 2 ps) are tested. In absence of dispersion, the response of Fig. 3 would In presence be a real constant without spatial ripples. of dispersion, the actual response is more complex, and some spatial ripples appear. These oscillations extend just to a fraction of the free-space wavelength λ, which is typically comparable with the plasmon wavelength, and increases with increasing the charge lifetime. We expect that the above behavior of the system will reflect, numer- ically, on the solution of the EM field distribution. graphene plasmon LD=2Rr/λτ=2psτ=2psτ=1psτ=1psC/σintra-0.50.51.00.000.020.02ReIm0.01RR/20R/2R01Radial positionNormalized electric fieldIncreasingLossesTIP00.050.10.15ReZs/ImZs6 of the voltage applied to the excitation-gap of the an- tenna is set to 70 Ω. It should be noted that the value of g can be scaled up or down provided that it remains much smaller than λ and that the applied voltage is scaled reversely. In addition, the internal impedance value is equal to the real part of the input impedance Z11 of the antenna in absence of graphene to have maximum power trans- fer. Fig. 5(A) shows that the antenna, which would work precisely at its resonance frequency in absence of graphene, can be more or less "detuned" by the graphene patch. The detuning depends on the graphene-antenna coupling: the higher the coupling strength, the higher the coupled reactive power, and the higher the reflection coefficient S11 at the input port of the antenna (indi- cated by arrows). S11 expresses the amount of power that is reflected back at the terminals of a voltage-source excitation, given by an infinitesimal electric dipole lo- cated at the center of the antenna. By definition, we have S11 = (Z11 − Z0)/(Z11 + Z0). The choice of the source impedance of the voltage excitation is arbitrary. However, as stated above, we have set Z0 in order to have resonance in absence of graphene, and accordingly a bet- ter visualization of the perturbing effect of the graphene patch (see Sec. III B for more information and numerical values). Fig. 5(B) shows in details what is happening to the antenna near (right picture) and far from (left pic- ture) the detuning points, i.e., the arrows of Fig. 5(A). In particular, we have taken surface reactance values of 2 and 1.7 kΩ, corresponding respectively to input matching values of about −4 and −13dB: in the former case, the radiation is drastically reduced by the plasmon coupling and a large amount of power is reflected back to the input port of the antenna. The results reported in Fig. 5(A) are derived from our MoM simulator23, whereas the plots of Fig. 5(B) are calculated by a full-wave EM solver (HFSS by Ansoft), which provides an independent validation of our implementation. B. Dispersive Characterization of Surface Plasmons In order to discuss the effect of dispersion on the strength of plasmon excitation, we focus on the two examples reported in Fig. 6. Differently from the re- sults shown in Fig. 5, here we take into account ohmic losses using the approximate complex conductivity of Eq. (13) within the RTA limit. Assuming a nominal frequency of 10 THz, we show the effect of spatial dis- persion on the input matching (S11) of the antenna in resonant [L ≈ λ/2.11, Fig. 6(A)] and non-resonant [L ≈ λ/12.5, Fig. 6(B)] conditions, respectively. All the reflec- (A) Input matching of the antenna vs surface re- FIG. 5. actance; (B) iso-surface plot of the electric field (magnitude) radiated by the wire antenna, and plasmon excitation on the underlying graphene patch. Another interesting effect is pointed out in Fig. 5, where the overlap of the plasmon distribution with the near field under the tip determines the strength of the coupling and the amount of power transferred from the antenna to the plasmon. In order to emphasize this concept- and make the effect more evident- the losses are just neglected. More specifically, the surface resistivity is taken to be an independent variable, i.e., its value is not calculated from the Kubo conductivity as in Fig. 4. In addition, Zs is considered a purely imaginary quan- tity (no losses) and the reactance ImZs is varied from about ∼ 1.3 kΩ to ∼ 2.5 kΩ. Thus, the geometric pa- rameters may be expressed in terms of the sizes of the antenna and the graphene disk relative to the vacuum wavelength λ. In particular: (i) the diameter D of the patch and the length L of the wire antenna are related by D = L = 2.11c/f = 2.11λ; (ii) the air-gap distance d between the antenna and the graphene-patch is given by d = L/200; (iii) the radius a of the antenna is fixed to a = L/1000; (iv) a value of L/200 is chosen for the excitation-gap g in the middle of the antenna, where the voltage-source is applied; (v) The internal impedance Z0 (A)(B)Strongly Perturbed RadiationIm(ZS) kΩ1.4 1.6 1.8. 2.0 2.2 2.4Reflection Coefficient S11 (dB)0-10-20-30-40Low DetuningHigh DetuningIm(ZS)=2.0 kΩIm(ZS)=1.7 kΩtion coefficients at the input terminal of the antenna are plotted as function of the chemical potential of graphene, i.e., the Fermi energy shift associated to the local doping level. 7 corresponding non-dispersive ones. These differences be- come more and more evident as the relaxation time in- creases, or, the quality of graphene gets better. In SMM, an electrostatic tuning of the charge density with a DC voltage applied to the microscope tip8 can pro- vide the local doping for near field applications. Fig. 7 shows the spatial distributions of the plasmonic wave on the graphene disk (for a normalized electric field) as func- tion of the radial position and the doping levels (values corresponding to those of Fig. 6). The distributions are computed with and without spatial dispersion, assuming a relaxation time of 2 ps. Differences between the dis- persing and non dispersing curves are clearly observable, particularly in correspondence of the resonance peaks. Looking at the plots of Fig. 6 and Fig. 7, we see that the absolute effect of spatial dispersion is higher in the non-resonant case, where the size of the graphene disk is much smaller than the free-space wavelength λ. FIG. 6. Reflection coefficient at the input terminal of the antenna with (solid lines) and without (dashed lines) spatial dispersion, for different relaxation times. The operating fre- quency is set to 10 THz. Two disk diameters are considered, namely D = L ≈ λ/2.11 (A), D = L ≈ λ/12.5 (B). Three typical values of charge-carrier relaxation times on the ps time-scale are tested (τ = 0.5, 1, 2 ps). In- deed, the relaxation time in graphene may strongly de- pend on the quality of the sample related to the fabri- cation process. In particular, real samples are affected by the presence of grain boundaries, defects, multilayer regions, etc. For the above reason, some flexibility is needed in selecting the τ parameter. As evident from Fig. 6, the larger is the relaxation time the sharper are the S11 peaks that express the maxima and minima of the antenna-plasmon coupling. The antenna is fed by a volt- age source with an impedance equal to the real part of the input impedance of the antenna, which is about 70 Ω and 2222 Ω for the resonant and non-resonant cases, respec- tively. Clearly, the minima of reflection, corresponding to surface field resonances, have lower matching levels in the non-resonant case. In either resonant or non reso- nant condition, the dispersive reflection peaks are more broadened- and have lower maxima- with respect to the FIG. 7. Spatial distribution of the surface electric field Es (normalized to 1) as function of relative position (r/R) and chemical potential µ, with (B,D) and without (A,C) spatial dispersion effects, for f=10 THz, and D = L ≈ λ/2.11 (A,B), D = L ≈ λ/12.5 (C,D) To go beyond the diagonal conductivity approximation of Eq. (13) in the RTA, we explicitly account for the an- gular variations of the EM field and, at the same time, we include the effect of diffusion currents. Thus, we next consider the full conductivity response of Eq. (12), within both the RTA and BGK approach. In Fig 8, we report the details of the input matching of the antenna in the 0.30.40.50.60.70.80.9-30-25-20-15-10-5μ (eV)S11 (dB)(A)(B)0.100.150.20.250.30.35-3-2.5-2-1.5-1-0.50S11 (dB)μ (eV) τ=2.0 ps τ=1.0 ps τ=0.5 ps τ=2.0 ps τ=1.0 ps τ=0.5 ps τ=2.0 ps τ=1.0 ps τ=0.5 psNo Disp τ=2.0 ps τ=1.0 ps τ=0.5 psNo DispD=L≈λ/2.11D=L≈λ/12.501ESES0.00.51.00.050.150.250.350.00.51.00.200.400.600.800.00.51.00.200.400.600.800.00.51.00.050.150.250.35r/Rr/Rr/Rr/R(A)(B)(C)(D)D=L≈λ/2.11D=L≈λ/12.5D=L≈λ/2.11D=L≈λ/12.5µ (eV)µ (eV)non dispersivenon dispersivedispersivedispersiveµ (eV)µ (eV)resonant case D ≈ L = λ/2.12, comparing the results obtained from the different levels of approximations dis- cussed in the present work. We find confirmation that the RTA conductivity of the graphene disk represents a sig- nificant improvement with respect to the non-dispersive conductivity of an infinite graphene sheet. In addition, we see that the contribution of the off diagonal tensor ¯¯σ leads to a non-negligible small correction to the diagonal tensor ¯σ, within the RTA. FIG. 8. Reflection coefficient (S11) at input terminal of the antenna with the different approaches to dispersion discussed here. The following parameters are used: f = 10 THz, τ = 1 ps, and D = L ≈ λ/2.11. More importantly, we notice that the effect of electron diffusion included in the BGK model plays a significant role. Indeed the RTA and BGK expressions of ¯σ and ¯¯σ are remarkably different: dispersion effects appear to be under-estimated by the RTA conductivity [or even more simplified expressions as Eq. (13)] with respect to the 8 BGK form. IV. CONCLUSIONS We have used a semiclassical model derived from the Boltzmann transport equation to investigate the effect of spatial dispersion on the linear THz response of graphene, characterized by excitation of surface plasmons. The ex- citation source has been provided by the near field of an antenna radiating in proximity of a graphene micro-disk. We have characterized the role of spatial dispersion, with respect to a cylindrical system, within both the RTA and the BGK approach, obtaining meaningful and compact expressions of the full-tensor representing the constitutive relation of the graphene patch in real space. We have shown that the surface distribution of the field on the disk, and the macroscopic response of the antenna, is significantly affected by spatial dispersion in two distinct noteworthy examples, where the antenna has been set in resonant and non-resonant conditions. Although the role played by spatial dispersion was pre- viously clarified in propagation and modal analysis of plasmons15,16, here we have provided a focus on near field excitation problems, with potential fallout in important practical applications, concerning near field imaging. ACKNOWLEDGMENTS This work has been supported by the European Project "Carbon Based Smart Systems for Wireless Applica- tions", (NANO-RF, n. 318352). We also thank NOW Srli (www.notonlywaves.com) for providing HFSS simu- lations. 1 F. H. L. Koppens, D. E. Chang, and F. J. G. de Abajo, Nano Letters 11, 3370 (2011). 2 L. Ju et al., Nature nanotechnology 6, 630 (2011). 3 T. Echtermeyer et al., Nature communications 2, 458 (2011). 4 J. Christensen, A. Manjavacas, S. Thongrattanasiri, F. H. Koppens, and F. J. Garcia de Abajo, ACS nano 6, 431 (2011). 5 D. A. Smirnova, R. E. Noskov, L. A. Smirnov, and Y. S. Kivshar, Physical Review B 91, 075409 (2015). 6 B. Diaconescu et al., Nature 448, 57 (2007). 7 R. R. Hartmann, J. Kono, and M. Portnoi, Nanotechnol- ogy 25, 322001 (2014). 8 T. Low and P. Avouris, Acs Nano 8, 1086 (2014). 9 A. N. Grigorenko, M. Polini, and K. Novoselov, Nature photonics 6, 749 (2012). 10 K. S. Novoselov et al., Nature 490, 192 (2012). 11 F. J. Garcia de Abajo, Acs Photonics 1, 135 (2014). 12 M. Pisarra, A. Sindona, P. Riccardi, V. Silkin, and J. Pitarke, New Journal of Physics 16, 083003 (2014). 13 M. Farina et al., Nature communications 4 (2013). 14 D. Correas-Serrano, J. S. Gomez-Diaz, and A. Alvarez- Melcon, Antennas and Wireless Propagation Letters, IEEE 13, 345 (2014). 15 G. Lovat, G. W. Hanson, R. Araneo, and P. Burghignoli, Physical Review B 87, 115429 (2013). 16 D. Correas-Serrano, J. S. Gomez-Diaz, J. Perruisseau- Carrier, and A. Alvarez-Melcon, Microwave Theory and Techniques, IEEE Transactions on 61, 4333 (2013). 17 Chapter 6 electron transport, in Conceptual Foundations of Materials: A Standard Model for Ground- and Excited- State Properties, edited by S. G. Louie and M. L. Cohen, 0.20.30.40.50.60.70.80.9-18-16-14-12-10Non disp.Eq. (12)Eq. (11)-BGKEq. (11)-RTAµ (eV)S11 (dB), Contemporary Concepts of Condensed Matter Science Vol. 2, pp. 165 -- 218, Elsevier, 2006. 18 V. Gusynin, S. Sharapov, and J. Carbotte, Journal of Physics: Condensed Matter 19, 026222 (2007). 19 V. Gusynin, S. Sharapov, and J. Carbotte, Physical Re- view B 75, 165407 (2007). 20 G. W. Hanson, Journal of Applied Physics 103, 064302 (2008). 21 G. W. Hanson, Antennas and Propagation, IEEE Trans- actions on 56, 747 (2008). 22 D. Mencarelli, L. Pierantoni, A. di Donato, and M. Farina, Journal of Computational Electronics 14, 214 (2015). 23 X. Y. He and R. Li, IEEE Journal of selected topics in quantum electronics 20, 62 (2014). products, 7 Ed. (Academic Press NY, 2007). APPENDIX: DISPERSIVE AND NON-DISPERSIVE RTA CONDUCTIVITY Using the polar coordinates k = k(cos ϑk, sinϑk) and q = q(cos θq, sin θq), the tensor product at the numerator of the BZ integral in Eq. (1) becomes 24 I. Gradshteyn and I. Ryzhik, Table of integrals, series, and B(z) = to Eq. 6. Now, consider the integral identity dψ cos nψ 1 + z cos ψ = 2π(−1)n √ 1 − z2 1 − z2 − 1 z (cid:90) 2π 0 (cid:32)√ 9 (cid:33)n , (cid:90) 2π which holds true for n = 0, 1, . . . and Imz (cid:54)= 0, and can be derived from the database24. As special cases, we get A(z) = 0 dψ 1 − z cos ψ = 2π√ 1 − z2 for n = 0 and (cid:90) 2π (cid:32)√ (cid:33)2 dψ cos 2ψ 1 − z cos ψ = 2π√ 1 − z2 1 − z2 − 1 z for n = 2. By a simple change of variable, we also have: 0 (cid:90) 2π (cid:90) 2π 0 0 (cid:90) 2π It follows that the ϑk-integral in Eq. (6) evaluates to dψ cos 2ψ 1 − z cos(ψ − ψ0) = B(z) cos 2ψ0, dψ sin 2ψ 1 − z cos(ψ − ψ0) = B(z) sin 2ψ0. (cid:20) cos 2ϑk I + sin 2ϑk sin 2ϑk − cos 2ϑk ω − vfq cos(ϑk − θq) (cid:21) (cid:20) cos 2θq = A(vfq/ω) I ω (cid:21) . + B(vfq/ω) ω sin 2θq sin 2θq − cos 2θq (cid:20) (cid:26) (cid:21) (cid:21)(cid:27) v±(k)v±(k) = v2 f = v2 f 2 sin ϑk cos ϑk cos2 ϑk sin ϑk cos ϑk (cid:20) cos 2ϑk sin2 ϑk sin 2ϑk sin 2ϑk − cos 2ϑk I + dϑk 0 , while the scalar products at the denominator reads: v±(k) · q = vfq cos(ϑk − θq). Hence, Eq. 1 is turned Plugging this in to Eq. (6), we obtain Eq. (7) and the scalar conductivities ¯σ ± rta, ¯¯σ ± rta.
1807.02792	1	1807	2018-07-08T09:50:36	Stable Branched Electron Flow	[ "cond-mat.mes-hall" ]	The pattern of branched electron flow revealed by scanning gate microscopy shows the distribution of ballistic electron trajectories. The details of the pattern are determined by the correlated potential of remote dopants with an amplitude far below the Fermi energy. We find that the pattern persists even if the electron density is significantly reduced such that the change in Fermi energy exceeds the background potential amplitude. The branch pattern is robust against changes in charge carrier density, but not against changes in the background potential caused by additional illumination of the sample.	cond-mat.mes-hall	cond-mat	Stable Branched Electron Flow B. A. Braem, C. Gold, S. Hennel, M. Roosli, M. Berl, W. Dietsche, W. Wegscheider, K. Ensslin, T. Ihn ETH Zurich, Solid State Physics Laboratory, Otto-Stern-Weg 1, 8093 Zurich, Switzerland E-mail: [email protected] 17 May 2021 Abstract. The pattern of branched electron flow revealed by scanning gate microscopy shows the distribution of ballistic electron trajectories. The details of the pattern are determined by the correlated potential of remote dopants with an amplitude far below the Fermi energy. We find that the pattern persists even if the electron density is significantly reduced such that the change in Fermi energy exceeds the background potential amplitude. The branch pattern is robust against changes in charge carrier density, but not against changes in the background potential caused by additional illumination of the sample. 1. Introduction to branched electron flow and scanning gate microscopy Semiconductor heterostructures of high purity allow electrons to move ballistically in a smooth potential landscape. This is crucial to achieve high charge carrier mobilities. At the same time investigating charge transport in the buried electron gas on a microscopic level is cumbersome. Therefore little is known about the electrons microscopic behavior and there are surprises like the strongly viscous behavior of charge carriers in high- mobility electron systems [2, 17, 4, 15]. Non-invasive local measurement techniques such as scanning SQUID microscopy [11, 21] are limited in their spatial resolution by the separation between the SQUID and the two-dimensional electron gas (2DEG) of the order of 100 nm. Higher resolution can be achieved by the technique of scanning gate microscopy (SGM) [7]. This invasive method creates a movable local potential hill in the plane of the 2DEG. High resolution is achieved when electrons emanating from a quantum point contact (QPC) scatter from this barrier back through the constriction. Topinka et al. used this technique to image branched electron flow in a AlGaAs heterostructure [19, 20]. These measurements are interpreted to reflect the spatial distribution of electron flow in the unperturbed case [19, 9]. The anisotropic pattern is caused by the background potential generated by remote ionized donor atoms. Therefore one expects the branch pattern to change if the kinetic Stable Branched Electron Flow 2 energy of the charge carriers changes by the average amplitude of the background potential. In contrast to this expectation our SGM experiments at different electron densities show that the pattern of branched flow is robust. Changes of the Fermi energy up to a factor of two change the branches visibility, but not their position. 2. Experimental realization As shown in Fig. 1(a), the 2DEG of our GaAs/AlGaAs sample is etched into a Hall bar shape that allows for measuring the longitudinal voltage VL and the source-drain current ISD in a four-terminal configuration. The 2DEG is buried 130 nm under the sample surface and separated by 1.13 µm from the back-gate [3]. We illuminated the sample with red light to increase the charge carrier mobility by ionizing additional donor atoms, which changes the random background potential. This so-called persistent photo conductivity in AlGaAs heterostructures is a well established effect and has been used to tune density and mobility of 2DEGs. After illumination, we can change the electron density n by applying a back-gate voltage Vbg as shown by the blue curve in Fig. 1(b). In the presented measurements, we use the back-gate to tune n in the range 1.0−2.0×1011 cm−2 where the electron mobility µ changes in the range 3−8×106 cm2/Vs. The QPC is formed by applying a gate voltage VQPC to split gates with a lithographic gap of 400 nm. We observe conductance quantization in the entire range of charge carrier density as presented in Fig. 1(c). The dips in the conductance plateaus are caused by the voltage biased SGM tip that is placed 4 µm from the QPC. At low charge carrier densities the second and third plateau are tilted and below the expected conductance values indicated by black dotted lines. To remain on the second plateau for different charge carrier densities, we apply a Vbg-dependent split gate voltage VQPC = V 2 bg/24.5 V − 0.62 × Vbg − 1.33 V in the following measurements. We measure in a home-built atomic force microscope (AFM) in a dilution refrigerator with a base temperature of 25 mK and an electronic temperature of the sample below 30 mK. To create the movable potential perturbation in the sample we apply a voltage Vtip to the metallic tip. We tune the QPC to the second conductance plateau at G = ISD/VL = 2 × 2e2/h and scan the tip close to the QPC at a distance of 30 nm above the GaAs surface. The tip voltage Vtip = −8 V is chosen such that it creates a disk of zero electron density in the 2DEG. The recorded conductance G as a function of tip position at the highest charge carrier density n = 2.03× 1011cm−2 is presented in Fig. 2(a). If the tip is close to the QPC the tails of the tip potential shift the saddle point potential in the constriction and we observe a smooth reduction of G. The color scale is chosen such that it depicts the region of strongly reduced conductance in the grey scale and the small variations on the second plateau in the orange scale. The pattern of branched electron flow is visible (examples are marked by black arrows) but obscured by the smooth background variation of G. Therefore we compute the conductance variation ∆G by subtracting the background from the measured G. This data processing is illustrated in Fig. 2(b) with Stable Branched Electron Flow 3 Figure 1. (a) Schematic of the Hall bar shaped sample (grey) with the back-gate (green). The voltage biased AFM tip is scanned 30 nm above the GaAs surface in the scan frame (red dashed) close to the split-gate defined QPC (blue). Longitudinal voltage VL and source-drain current ISD are measured as function of tip position. (b) Hall density (blue solid line) and mobility (green) of the illuminated 2DEG as function of back-gate voltage measured in our SGM setup at an electronic temperature < 30 mK. The green dotted line shows the electron mobility of our model (see section 3). (c) QPC conductance plateaus as a function of split-gate voltage for equidistant electron densities. The voltage biased SGM tip is placed in the center of the scan frame 4 µm from the QPC gap. Curves are vertically offset by 0.1×2e2/h for clarity, the expected conductance values are indicated by the dotted lines. the example of G(x, y = 5 µm) marked by the green line in Fig. 2(a). The background is calculated by a two-dimensional running average of G with a span of 300 nm in x and 100 nm in y. The conductance variation ∆G as a function of tip position is shown in Fig. 2(c) with blue points marking local minima, which will be used to compare branch positions at different electron densities. To remove outliers, only local minima with other minima in their vicinity are shown. At x ≈ 2 µm, y ≈ 5.5 µm the transition from the second to the first conductance plateau due to tip gating of the QPC is visible. Performing the same experiment on the first or third QPC plateau produces similar results. On this sample, the visibility of the branches on the first plateau was lower than on the second. The third plateau is present only in a smaller range of electron density, therefore we used the second plateau for this study. Thanks to coherent transport through the structure and the high spatial resolution of SGM we also observe interference effects of tip reflected electrons [20]. As an example, Fig. 2(d) shows an enlarged view of the area indicated by the green rectangle in Fig. 2(c) where the interference fringes are measured. The expected periodicity of half the Fermi- wavelength is indicated by the dotted lines. From earlier SGM studies with tunable Stable Branched Electron Flow 4 Figure 2. (a) Conductance as a function of tip position displays the pattern of branched electron flow (examples of branches marked by black arrows). Split-gate voltage VQPC is tuned such that QPC conductance is on second plateau when the tip is in the scan frame center. (b) Conductance variation ∆G (red) is obtained by subtracting a smoothed background (blue) from the measured conductance (green cut in (a)). (c) ∆G with minima indicated by blue dots. (d) Conductance variation due to interference effects, position indicated by the green rectangle in (c). electron density [12, 14, 13] it is known that such pattern scale with charge carrier density as expected. In Fig. 2(c) further periodic patterns are visible, for example at x = 0.8 µm, y = 2.3 µm or at x = 3.4 µm, y = 3.3 µm. A possible origin of such patterns is the presence of hard scatterers [10]. 3. Dependence on charge carrier density We repeat the measurement of G as function of tip position at four different Vbg leading to lower electron densities. To keep the diameter of the depletion disk below the tip constant at all n, the amplitude of the tip-induced potential in the 2DEG should grow proportionally to the Fermi energy. We repeat scans with varying Vtip at different n (data not shown) to find the least negative Vtip(Vbg) at which branches are visible. This critical tip voltage corresponds to the creation of a depletion disk, i.e. when the maximum of the tip induced potential equals the Fermi energy. For all presented measurements we set the tip voltage Vtip = −1.42× Vbg− 6.87 V that lies below the critical tip voltage and therefore creates a finite size depletion disk. From the measured G(x, y) at all densities we subtract a smoothed background as described above to calculate the conductance variation ∆G. The results in Figs. 3(a)-(c) show that the pattern of branched flow is only weakly modified when reducing the electron density by 25 percent. However, at even lower n and µ we see in Figs. 3(d)-(e) that the region, where branched electron flow is observed, is limited to the vicinity of the QPC. In Fig. 3(e) we observe rings of charge rearrangements in ∆G, the most prominent rings are marked by black arrows. They occur because the tails of the tip potential Stable Branched Electron Flow 5 Figure 3. (a)-(e) Conductance variation ∆G for decreasing electron densities. The pattern of branched flow does not change position as long as it can be detected. Rings of charge rearrangements are marked by black arrows in (e). (f) After additional illumination of the sample, the pattern of the branched flow is different than in (a), while electron density and mobility are unchanged. rearrange trapped discrete charges. Such a change in its electronic environment shifts the QPC conductance curve with respect to the applied top-gate voltage. Therefore G is not affected by charge rearrangements if the QPC is on a conductance plateau. At low electron densities, the QPC plateaus are tilted as seen in Fig. 1(c) and therefore the rings of charge rearrangements are visible in Fig. 3(e). To compare the branching pattern at different charge carrier densities, we track the branches by finding local minima in ∆G as shown in Fig. 2(c). The positions of the minima are plotted in Fig. 4(a) with different colors indicating their respective charge carrier density. This analysis shows that even when changing the charge carrier density by a factor of two, the branching pattern remains roughly at the same position and the detailed features of the pattern are largely independent of n. Figure 4(b) is an enlarged view of the square indicated in Fig. 4(a) to show that there are systematic minor shifts Stable Branched Electron Flow 6 Figure 4. (a) Minima of ∆G for different charge carrier densities n as described in the main text. (b) Enlarged view shows minor shifts of minimum positions with n. of the minima of the order of 100 nm whereas the overall pattern remains unchanged. For comparison, Fig. 3(f) shows the pattern of branched electron flow after additional illumination of the sample with respect to the measurements in Figs. 3(a)- (e). The additional illumination was short enough to change the charge carrier density by only 3% at the same back-gate voltage, and no measureable difference in mobility was found (data not shown). Comparing the two measurements in Figs. 3(a) and (f) demonstrates that the microscopic pattern of branched electron flow has changed completely due to a significant change in the background potential. Measurements of the macroscopic quantities n and µ overlook this modification. 4. Trajectory simulations The stability of the distribution of electron motion in a background potential has been studied within previous theoretical work. For example, Liu and Heller [16] used a fully quantum mechanical model to show that the pattern is robust against changes of the injection into the 2DEG. We will use a less involved, classical model to investigate the stability against changes of the electron density and the Fermi energy. Such trajectory simulations have been used to model branched electron flow in the past [20, 9, 18]. Can they also explain the experimentally observed stability, i.e. does a classical particle follow the same trajectory in a random potential if it's kinetic energy changes by a factor of two? To answer this question, we use a trajectory model similar to Steinacher et al. [18]. We model the potential in the 2DEG caused by SGM tip and QPC gates by calculating the charge distribution in Thomas-Fermi approximation with the finite element software COMSOL 5.0. The three-dimensional modelling of the sample and the SGM tip includes the screening effects of the top-gates on the tip potential. We add a correlated random background potential to the solution of the COMSOL Stable Branched Electron Flow 7 Figure 5. Classical modelling of electron trajectories: (a) Transmitted electron trajectories (red) and potential of QPC and remote donors indicated by the color scale. (b) Normalized number of trajectories in (a) per 100 nm×100 nm square. (c) Transmitted trajectories in a potential including the voltage biased tip at xtip = 2.56 µm, ytip = 3.5 µm. (d) Trajectory density without tip perturbation at the position of the green line in (b) for different charge carrier densities. (e) Number of electron trajectories, that are reflected at the tip and backscattered through the QPC, for tip positions and charge carrier densities as in (d). simulation. It includes Thomas-Fermi screening of the donor atoms [1], the thickness- dependence of the 2DEG on the back-gate voltage [1, 8, 5], and correlation of ionized donors [6]. The correlation parameter of the donors is used to adjust the calculated charge carrier mobility to the experimental values. For comparison, both are shown in Fig. 1(b). The sum of random background potential and COMSOL simulation is shown as color map in Fig. 5(a) without SGM tip and in Fig. 5(c) in the presence of the tip. To simulate the motion of electrons in the potential, we calculate trajectories starting at y = 7.5 µm from an equidistant grid of x-positions and angles with a kinetic energy equal to the Fermi energy. Only the transmitted trajectories are added to Figs. 5(a) and (c) as red lines. They show the uneven spatial distribution and caustics that are expected for trajectories in a correlated potential [9]. How are these simulations without tip potential related to the SGM experiment? Figure 5(b) shows the normalized number of trajectories per 100 nm×100 nm square of the simulation in (a). The features qualitatively agree with the measured ∆G. We repeat the simulation for the values of n measured in the experiment. Figure 5(d) illustrates the evolution of the trajectory density as a function of n along the green line in Fig. 5(b) (x = 2.3 − 3.0 µm, y = 3.5 µm). To directly compare the model to the experiment, we also simulate the situation with a SGM tip potential at different positions xtip, ytip for different n. Trajectories, which are transmitted through the QPC, reflected by the tip potential, and returning Stable Branched Electron Flow 8 through the QPC, will be called tip reflected trajectories. They are experimentally relevant because they contribute to the sample resistance by being backscattered through the constriction. The number of tip reflected trajectories is shown in Fig. 5(e) for the same tip coordinates xtip, ytip as the positions x, y without tip in Fig. 5(d). The ratio of tip reflected to transmitted trajectories is up to 10%, similar to the experimentally observed ratio ∆G/G. The similarity of Figs. 5(d) and (e) confirms the findings of Topinka et al.: The ratio of tip reflected trajectories to the total number of trajectories hitting the tip head-on is roughly constant. Therefore the number of tip reflected trajectories maps the local trajectory density in the absence of the tip [19, 9], even though the reflection of trajectories happens at the edge of the tip depleted region and not at its center. The experimentally observed minor shifts of branch position as a function of electron density shown in Fig. 4(b) are reproduced in both simulations with and without tip potential. In Figs. 5(d) and (e) we see this effect as the shift of the maxima with n. In Fig. 6 simulations for the same densities as in the measured ∆G of Figs. 3(a)-(e) are presented. The trajectories in Figs. 6(a)-(e) show increasingly complex behavior with decreasing n. This can be attributed to the reduced mobility, as the background potential is higher with respect to the Fermi energy. We calculate the trajectory densities in Figs. 6(f)-(j) as described above, which are related to the number of tip reflected trajectories and therefore to the reduced conductance in the experiment. A certain signal strength is required to determine the pattern experimentally, so a threshold of the relative trajectory density is chosen at an arbitrary value of 0.1. Figures 6(k)-(o) show the trajectory density above this threshold value in black for the experimentally accessible region in front of the QPC. The maps of trajectory density above threshold in Figs. 6(k)-(o) reproduce the features of the measurements in Figs. 3(a)-(e) qualitatively: The branched flow pattern remains overall constant as long as it is visible. To observe the pattern far from the QPC, a high n and µ is required, otherwise the electron trajectories are no longer focussed in the caustics but are evenly spread in the 2DEG. In the experiment, the pattern is lost because the number of electrons scattering from the tip potential is the same at every tip position. This effect can be described by choosing a threshold of 0.1 in trajectory density. Does the observed stability of trajectory density also hold for the behavior of the individual trajectories? We found an unstable behavior of the tip reflected trajectories: A small difference (below 1 ‰) of the charge carrier density may change a particular tip reflected trajectory to become a transmitted trajectory (from the same starting position and starting direction). A similar behavior is observed by small changes in tip positions which are below the experimental resolution. Both the convex saddle point potential and the background potential make two initially close trajectories to separate after a distance of a few micrometers. From this behavior we conclude that single electron trajectories fluctuate within their bundles and the observed stability is due to averaging over multiple trajectories. Stable Branched Electron Flow 9 Figure 6. Classical modelling at different electron densities: (a)-(e) Transmitted electron trajectories (red) and potential of QPC and remote donors indicated by the color scale. (f)-(j) Normalized number of trajectories in (a)-(e) per 100 nm×100 nm square. With decreasing n the regions of high density are limited to the vicinity of the QPC. (k)-(o) Experimentally accessible region of (f)-(j) drawn black if trajectory density is above the threshold value 0.1. Those pattern are robust against changes in n. 5. Conclusions We have experimentally shown that the pattern of branched electron flow is astonishingly robust against changes of the charge carrier density. On the other hand, the branched flow pattern can be changed completely by additional illumination which modifies the background potential without changing macroscopic quantities as charge carrier density or mobility. The effect of such modifications are only visible by local investigations. Additionally, we have presented a model to simulate the classical motion of electrons in a random potential background. The simulation reproduces the experimentally observed branch stability including minor shifts in branch position. Stable Branched Electron Flow Acknowledgements 10 We thank D. Weinmann, R. Jalabert, and O. Ly for fruitfull discussions. The authors acknowledge financial support from ETH Zurich and from the Swiss National Science Foundation (SNF 2-77255-14 and NCCR QSIT). References [1] Ando, T., Fowler, A. B., and Stern, F. Electronic properties of two-dimensional systems. Rev. Mod. Phys. 54, 2 (Apr. 1982), 437 -- 672. [2] Bandurin, D. A., Torre, I., Kumar, R. K., Shalom, M. B., Tomadin, A., Principi, A., Auton, G. H., Khestanova, E., Novoselov, K. S., Grigorieva, I. V., Ponomarenko, L. A., Geim, A. K., and Polini, M. Negative local resistance caused by viscous electron backflow in graphene. Science 351, 6277 (Mar. 2016), 1055 -- 1058. [3] Berl, M., Tiemann, L., Dietsche, W., Karl, H., and Wegscheider, W. Structured back gates for high-mobility two-dimensional electron systems using oxygen ion implantation. Applied Physics Letters 108 (Mar. 2016), 132102. [4] Crossno, J., Shi, J. K., Wang, K., Liu, X., Harzheim, A., Lucas, A., Sachdev, S., Kim, P., Taniguchi, T., Watanabe, K., Ohki, T. A., and Fong, K. C. Observation of the Dirac fluid and the breakdown of the Wiedemann-Franz law in graphene. Science 351, 6277 (Mar. 2016), 1058 -- 1061. [5] Davies, J. H. The Physics of Low-dimensional Semiconductors: An Introduction. Cambridge University Press, Dec. 1997. [6] Efros, A. L., Pikus, F. G., and Samsonidze, G. G. Maximum low-temperature mobility of two-dimensional electrons in heterojunctions with a thick spacer layer. Phys. Rev. B 41, 12 (Apr. 1990), 8295 -- 8301. [7] Eriksson, M. A., Beck, R. G., Topinka, M., Katine, J. A., Westervelt, R. M., Campman, K. L., and Gossard, A. C. Cryogenic scanning probe characterization of semiconductor nanostructures. Applied Physics Letters 69, 5 (July 1996), 671 -- 673. [8] Fang, F. F., and Howard, W. E. Negative Field-Effect Mobility on (100) Si Surfaces. Phys. Rev. Lett. 16, 18 (May 1966), 797 -- 799. [9] Heller, E. J., and Shaw, S. Branching and Fringing in Microstructure Electron Flow. Int. J. Mod. Phys. B 17, 22n24 (Sept. 2003), 3977 -- 3987. [10] Kolasinski, K., Szafran, B., Brun, B., and Sellier, H. Interference features in scanning gate conductance maps of quantum point contacts with disorder. Phys. Rev. B 94, 7 (Aug. 2016), 075301. [11] Koshnick, N. C., Huber, M. E., Bert, J. A., Hicks, C. W., Large, J., Edwards, H., and Moler, K. A. A terraced scanning super conducting quantum interference device susceptometer with submicron pickup loops. Appl. Phys. Lett. 93, 24 (Dec. 2008), 243101. [12] LeRoy, B., Bleszynski, A., Topinka, M. A., Westervelt, R. M., Shaw, S., Heller, E., Maranovski, K., and Gossard, A. In Physics of Semiconductors 2002 (Edinburgh, 2003), vol. 1 of A.R. Long and J.H. Davies, Institute of Physics Publishing, pp. 169 -- 176. Imaging coherent electron flow. [13] LeRoy, B. J. Imaging coherent electron flow. J. Phys.: Condens. Matter 15, 50 (2003), R1835. [14] LeRoy, B. J., Topinka, M. A., Westervelt, R. M., Maranowski, K. D., and Gossard, Imaging electron density in a two-dimensional electron gas. Appl. Phys. Lett. 80, 23 A. C. (May 2002), 4431 -- 4433. [15] Levinson, E. V., Gusev, G. M., Levin, A. D., Levinson, E. V., and Bakarov, A. K. Viscous electron flow in mesoscopic two-dimensional electron gas. AIP Advances 8, 2 (Feb. 2018), 025318. arXiv: 1802.09619. Stable Branched Electron Flow 11 [16] Liu, B., and Heller, E. J. Stability of Branched Flow from a Quantum Point Contact. Phys. Rev. Lett. 111, 23 (Dec. 2013), 236804. [17] Moll, P. J. W., Kushwaha, P., Nandi, N., Schmidt, B., and Mackenzie, A. P. Evidence for hydrodynamic electron flow in PdCoO$ 2$. Science 351, 6277 (Mar. 2016), 1061 -- 1064. arXiv: 1509.05691. [18] Steinacher, R., Kozikov, A. A., Rossler, C., Reichl, C., Wegscheider, W., Ensslin, K., and Ihn, T. Scanning gate imaging in confined geometries. Phys. Rev. B 93, 8 (Feb. 2016), 085303. [19] Topinka, M. A., LeRoy, B. J., Shaw, S. E. J., Heller, E. J., Westervelt, R. M., Maranowski, K. D., and Gossard, A. C. Imaging Coherent Electron Flow from a Quantum Point Contact. Science 289, 5488 (Sept. 2000), 2323 -- 2326. [20] Topinka, M. A., LeRoy, B. J., Westervelt, R. M., Shaw, S. E. J., Fleischmann, R., Heller, E. J., Maranowski, K. D., and Gossard, A. C. Coherent branched flow in a two-dimensional electron gas. Nature 410, 6825 (Mar. 2001), 183 -- 186. [21] Vasyukov, D., Anahory, Y., Embon, L., Halbertal, D., Cuppens, J., Neeman, L., Finkler, A., Segev, Y., Myasoedov, Y., Rappaport, M. L., Huber, M. E., and Zeldov, E. A scanning superconducting quantum interference device with single electron spin sensitivity. Nat Nano 8, 9 (Sept. 2013), 639 -- 644.
1911.04789	2	1911	2019-11-13T18:18:53	Real-time detection of every Auger recombination in a self-assembled quantum dot	[ "cond-mat.mes-hall", "quant-ph" ]	Auger recombination is a non-radiative process, where the recombination energy of an electron-hole pair is transferred to a third charge carrier. It is a common effect in colloidal quantum dots that quenches the radiative emission with an Auger recombination time below nanoseconds. In self-assembled QDs, the Auger recombination has been observed with a much longer recombination time in the order of microseconds. Here, we use two-color laser excitation on the exciton and trion transition in resonance fluorescence on a single self-assembled quantum dot to monitor in real-time every quantum event of the Auger process. Full counting statistics on the random telegraph signal give access to the cumulants and demonstrate the tunability of the Fano factor from a Poissonian to a sub-Poissonian distribution by Auger-mediated electron emission from the dot. Therefore, the Auger process can be used to tune optically the charge carrier occupation of the dot by the incident laser intensity; independently from the electron tunneling from the reservoir by the gate voltage. Our findings are not only highly relevant for the understanding of the Auger process, it also demonstrates the perspective of the Auger effect for controlling precisely the charge state in a quantum system by optical means.	cond-mat.mes-hall	cond-mat	a Real-time detection of every Auger recombination in a self-assembled quantum dot P. Lochner,† A. Kurzmann,†,¶ J. Kerski,† P. Stegmann,† J. König,† A. D. Wieck,‡ A. Ludwig,‡ A. Lorke,† and M. Geller∗,† †University of Duisburg-Essen, Faculty of Physics and CENIDE, Lotharstr. 1, 47057 Duisburg, Germany ‡Lehrstuhl für Angewandte Festkörperphysik, Ruhr-Universität Bochum, Universitätsstrasse 150, 44780 Bochum, Germany ¶Solid State Physics Laboratory, ETH Zurich, Otto-Stern-Weg 1, 8093 Zurich, Switzerland E-mail: [email protected] Abstract Auger recombination is a non-radiative process, where the recombination energy of an electron-hole pair is transferred to a third charge carrier. It is a common effect in colloidal quantum dots that quenches the radiative emission with an Auger recombi- nation time below nanoseconds. In self-assembled QDs, the Auger recombination has been observed with a much longer recombination time in the order of microseconds. Here, we use two-color laser excitation on the exciton and trion transition in resonance fluorescence on a single self-assembled quantum dot to monitor in real-time every quan- tum event of the Auger process. Full counting statistics on the random telegraph signal give access to the cumulants and demonstrate the tunability of the Fano factor from a 1 Poissonian to a sub-Poissonian distribution by Auger-mediated electron emission from the dot. Therefore, the Auger process can be used to tune optically the charge carrier occupation of the dot by the incident laser intensity; independently from the electron tunneling from the reservoir by the gate voltage. Our findings are not only highly rel- evant for the understanding of the Auger process, it also demonstrates the perspective of the Auger effect for controlling precisely the charge state in a quantum system by optical means. Keywords: Quantum dots, Resonance fluorescence, Auger recombination, Full counting statistics, Random telegraph signal The excitonic transitions in self-assembled quantum dots (QDs)1,2 realize perfectly a two- level system in a solid-state environment. These transitions can be used to generate single photon sources3,4 with high photon indistinguishability,5,6 an important prerequisite to use quantum dots as building blocks in (optical) quantum information and communication tech- nologies.7,8 Moreover, self-assembled QDs are still one of the best model systems to study in an artificial atom the carrier dynamics,9,10 the spin- and angular-momentum properties11,12 and charge carrier interactions.13 One important effect of carrier interactions is the Auger process: An electron-hole pair recombines and instead of emitting a photon, the recombina- tion energy is transferred to a third charge carrier, which is then energetically ejected from the QD.14 -- 17 This is a common effect, mostly studied in colloidal QDs, where it quenches the radiative emission with recombination times in the order of picoseconds to nanoseconds.18 -- 20 This limits the efficiency of optical devices containing QDs like LEDs21,22 or single photon sources.23 -- 25 In self-assembled QDs, Auger recombination was speculated to be absent, and only recently, it was directly observed in optical measurements on a single self-assembled QD coupled to a charge reservoir with recombination times in the order of microseconds.26 As a single Auger process is a quantum event, it is unpredictable and only the statistical evaluation of many processes gives access to the physical information of the recombination 2 process.27,28 The most in-depth evaluation - the so-called full counting statistics - becomes possible when each single quantum event in a time trace is recorded. Such real-time detec- tion in optical experiments on a single self-assembled QD have until now only been shown for the statistical process of electron tunneling between the QD and a charge reservoir, where tunneling and spin-flip rates could be tuned by the applied electric and magnetic field.29 Here, Auger recombination in a single self-assembled QD is investigated by optical real- time measurements of the random telegraph signal. With the technique of two-laser excita- tion, we are able to detect every single quantum event of the Auger recombination. These events take place in the single QD, leaving the quantum dot empty until single-electron tun- neling into the QD from the charge reservoir takes place again. This reservoir is coupled to the QD with a small tunneling rate in the order of ms−1. The laser intensity, exciting the trion transition, precisely controls the electron emission by the Auger recombination and, hence, the average occupation with an electron. It also tunes the Fano factor from a Poisso- nian to a sub-Poissonian distribution, which we observe in analyzing the random telegraph signal by methods of full counting statistics. The investigated sample was grown by molecular beam epitaxy (MBE) with a single layer of self-assembled In(Ga)As QDs embedded in a p-i-n diode (see Supporting Information for details). A highly n-doped GaAs layer acts as charge reservoir, which is coupled to the QDs via a tunneling barrier, while a highly p-doped GaAs layer defines an epitaxial gate.30 An applied gate voltage VG shifts energetically the QD states with respect to the Fermi energy in the electron reservoir and controls the charge state of the dots by electron tunneling through the tunneling barrier. The sample is integrated into a confocal microscope setup within a bath cryostat at 4.2 K for resonant fluorescence (RF) measurements (see Methods). Figure 1 shows the RF of the neutral exciton (X0) and the negatively charged exciton, called trion (X-). A RF measurement as function of gate voltage in Figure 1b shows the fine- structure split exciton31 with an average linewidth of about 1.8 µeV at low excitation intensity (1.6·10−3 µW/µm2). Please note, that this measurement was recorded at a laser energy where 3 Figure 1: Resonance fluorescence (RF) of the exciton (X0) and trion (X-) tran- sition. a At a gate voltage of VG = 0.375 V (vertical dashed line), the electron ground state is in resonance with the Fermi energy in the charge reservoir. The exciton transition vanishes and the trion transition can be excited at lower frequencies. b Gate voltage scan of the exciton at a fixed excitation frequency of 325.710 THz (laser excitation intensity of 1.6 · 10−3 µW/µm2). c Two-color laser excitation with laser 1 and laser 2 (red and blue line in panel a) shows a bright exciton fluorescence X0 at a gate voltage where, in equilibrium, an electron would occupy the QD and quenches the X0 transition. X- signal is scaled up by factor 100. Simultaneous excitation of the trion transition X- empties the dot by Auger re- combination and the exciton transition can be excited with the second laser until an electron tunnels into the dot again (see panel d for a schematic representation). 4 RF Intensity (kcounts/s)Gate VoltageVG(V)0501001500.450.55X0X- cX0X-0e-Inad050100150-0.84-0.80-0.76200RF Intensity (kcounts/s)bGate VoltageVG(V)X0aQDReservoir )zHT( ycneuqerFGate VoltageVG(V)324.509324.511325.754325.756325.758325.760325.7620.20.40.6Laser 2Laser 10.515VX0X-1.7µ2.0µx 1000InXthe exciton gets into resonance at negative gate voltages because here, the measurement conditions were the best. The quantum-confined Stark effect shifts the exciton resonance X0 for higher gate voltages to higher frequencies up to 325.760 THz, seen in Figure 1a. This quadratic Stark shift of the two exciton transitions32 is indicated by two white lines. At a voltage of about 0.375 V (dashed vertical line in Fig. 1a), the electron ground state in the dot is in resonance with the Fermi energy in the charge reservoir. An electron tunnels into the QD and the exciton transition vanishes while the trion transition can be excited at lower frequencies from 324.5095 THz to 324.5115 THz. The spectrum of the exciton (blue dots) and the trion transition (red dots) under two- laser excitation is shown in Figure 1c. The trion transition is measured at a laser frequency of 324.511 THz (corresponding to the red line, "Laser 1" in Fig. 1a) and a laser excitation intensity of 8· 10−6 µW/µm2 at a gate voltage of 0.515 V. The exciton spectrum in Figure 1c was obtained simultaneously by a second laser 2 on the exciton transition (blue line in Fig. 1a at 325.7622 THz) with a laser excitation intensity of 1.6 · 10−3 µW/µm2, as the Auger recombination with rate γa leads to an empty QD until an electron tunnels into the In. This rate comprises the tunneling into the dot from the reservoir with rate γIn = γ0 33 (see Fig. 1d for a empty dot γ0 In and the tunneling into the dot charged with an exciton γX In In + γX schematic representation). This has been explained previously in Kurzmann et al.26 with the important conclusion that the intensity ratio between trion/exciton intensity in equilibrium measurements is given by the ratio between Auger/tunneling rate γa/γIn. As the tunneling rate γIn in the sample used here is in the range of ms−1, the Auger rate γa exceeds the tunneling rate by more than two orders of magnitude (see below). As a consequence, the intensity of the trion transition in equilibrium is by more than two orders of magnitude smaller than the exciton transition. The interplay between electron tunneling and optical-driven Auger recombination can be studied in more detail by a real-time random telegraph signal of the resonance fluorescence. In these measurements, the time stamp of every detected RF photon is recorded, see Fig- 5 Figure 2: Time-resolved RF random telegraph signal. At a gate voltage of 0.515 V, the trion and exciton are excited simultaneously (like in Fig. 1c). The intensity of the exciton excitation laser 2 is held constant at 1.6· 10−3 µW/µm2, which is far below the saturation of the RF signal (see Supporting Information). The intensity of the trion excitation laser 1 is varied (from top to bottom: 1.6· 10−5, 6· 10−6 and 8· 10−7 µW/µm2). Every time, an Auger recombination takes place, the dot is emptied, and exciton RF signal turns on. After a time τOn, an electron tunnels into the QD and the exciton RF signal quenches. All intensities smaller than the threshold (red line at 7 counts/0.1 ms) are counted as "exciton off" (white areas), all intensities above the threshold are counted as "exciton on" (blue areas). 6 )sm 1.0/stnuoc( ytisnetnI FRtime (ms)5050500000510152025301.6⋅10−5µW/µm26⋅10−6µW/µm28⋅10−7µW/µm2ττThresholdAugere-tunnelingure 2, enabling the evaluation by full counting statistics. As the intensity of the trion is very weak, the random telegraph signal has been investigated in a two-color excitation scheme. The bright exciton transition with count rates exceeding 10 MCounts/s (see Supporting In- formation) is used as an optical detector for the telegraph signal of the Auger recombination. In this two-color laser excitation scheme, the "exciton off" signal corresponds to the "trion on" signal and vice versa.26 Hence, the trion statistics can directly be determined from the "inverse" exciton signal. The intensity of the exciton excitation laser 2 is held constant at 1.6· 10−3 µW/µm2. This intensity is far below the saturation of the RF signal of the exciton (see Supporting Information) and avoids the photon-induced electron capture at high exci- tation intensities.34 However, this laser intensity yields count rates above 200 kcounts/s (see Fig. 1b), sufficiently-high for recording single quantum events in a real-time measurement.29 While the intensity of the exciton detection laser 2 is kept constant, the laser intensity of the trion excitation laser 1 is increased from 1.6 · 10−7 µW/µm2 up to 1.6 · 10−5 µW/µm2. For every trion laser intensity, the time-resolved RF signal is recorded for 15 minutes using a fast (350 ps) avalanche photo diode and a bin time of 100 µs. Figure 2 shows parts of three different time traces at three different trion laser 1 intensities. As the exciton laser 2 intensity always exceeds the trion laser 1 intensity by at least nearly two orders of magnitude, the small amount of RF counts from the trion can be neglected. As a consequence, the detected RF signal of the exciton is directly related to the Auger recombination: An Auger recombination empties the dot and the exciton transition detects an empty dot (no trion transition possible) with a count rate of about 25 counts per bin time (100 µs). After a time τOn, an electron tunnels into the QD in Figure 2 and the exciton RF signal quenches for a charged dot (trion transition possible) until, after a time τOff, another Auger recombination happens. Increasing the trion laser intensity from 8 · 10−7 up to 1.6 · 10−5 µW/µm2 in Figure 2 increases the probability of an electron emission with rate γE = nγa by an Auger process, as the probability for occupation of the dot with a trion n increases with increasing laser 1 7 intensity. Therefore, the exciton transition is observed most frequently for the highest trion laser intensity. This can be observed in Figure 2, where the optical random telegraph signal is compared for three different trion excitation intensities. A threshold between exciton "on" and "off" is set for the following statistical evaluation.35,36 All exciton RF intensities smaller than this threshold (dashed red line at 7 counts/0.1 ms in Fig. 2) are counted as "exciton off" (white areas), all intensities above the threshold are counted as "exciton on" (blue areas). Figure 3: Auger and tunneling rates from the time-resolved RF random telegraph signal. Panel a shows the probability distribution of the "off" (red dots) and "on"-times (blue dots) from the measurement at 8·10−7 µW/µm2 trion laser excitation intensity (laser 1). Fitting these data yields the emission rate γE = nγa and the tunneling rate of an electron into the dot γIn. In panel b, the emission and tunneling rates are plotted as function of the trion laser intensity (red and blue dots, respectively). The tunneling rate remains constant at a mean value of 0.74 ms−1, the emission rate increases linearly with the laser intensity and, accordingly, with the occupation probability n (top axis). c Asymmetry a, calculated from the emission and tunneling rates as shown in the inset, as a function of the trion laser intensity. From these time-resolved RF data sets, the Auger and tunneling rates can be determined by analysing the probability distributions of the "off"-times τOff and the "on"-times τOn for every 15 minutes long data set.35 A representative distribution at a trion laser intensity of 8· 10−7 µW/µm2 can be seen in Figure 3a. An exponential fit to the "on"-times (blue line in Fig. 3a) yields the tunneling rate γIn into the QD, while an exponential fit to the "off"-times (red line in Fig. 3a) yields the emission rate γE = nγa for this specific trion laser 1 intensity. In the example in Figure 3a, we find γIn = 0.80 ms−1 and γE = 0.074 ms−1. As discussed above, the probability for emitting an electron by an Auger recombination process increases 8 Eventsτa0.050.101001011021038⋅10−7µW/µm2ττX-Laser Intensity(10-5 µµb(ms-1)001200.0011Occupation Probability nIn=0.74ms−1E=⋅1.7µs−10.51.5cAsymmetry aX-Laser Intensity(µµ0-1110-510-410-710-610-5Occupation Probability n=In−EIn+E10-410-310-2with the occupation probability of the QD with a trion n. The occupation probability with a trion n depends on the laser 1 excitation intensity and has been determined from a pulsed measurement of the trion RF intensity, where the highest trion intensity corresponds to an occupation probability of n = 0.537 (see Supporting Information for more details). Figure 3b shows the expected linear dependence of the electron emission rate γE = nγa on the occupation probability of the QD with a trion n; tuning the emission rate γE from almost zero to more than γE = 2 ms−1. The Auger rate is the proportional factor γa=1.7 µs−1 (red data points) and in good agreement with the value obtained before for a different QD with slightly different size.26 The tunneling rate γIn remains approximately constant at a mean value of 0.74 ms−1 (blue data points in Fig. 3b). This is in agreement with the probability for an electron to tunnel into the empty QD at a constant gate voltage: it is independent on the trion laser intensity. That means, we are able to use the Auger recombination to tune optically the electron emission rate independently from the gate voltage, influencing the emission rate without changing the rate for capturing an electron into the QD (here by the tunneling rate γIn). An independent tuning of electron emission and capture rate is usually not possible for a QD that is tunnel-coupled to one charge reservoir. Changing the coupling strength or Fermi energy by a gate voltage always changes both rates for tunneling into and out of the dot simultaneously. Using the standard methods of full counting statistics35,38 in the following, first of all the γIn+γE asymmetry a = γIn−γE between the tunneling γIn and emission rate γE has been evaluated. The asymmetry in Figure 3c can be tuned by the trion excitation laser intensity from -1 up to 0.55 at a maximum laser intensity of 1.6·10−5 µW/µm2. Important to mention here: At high trion laser intensities above 1.6·10−5 µW/µm2, the electron emission by Auger recombination after an tunneling event from the reservoir happens much faster than the bin time of 0.1 ms. Therefore, the RF intensity within the bin time is not falling below the threshold and these events are not detected, i. e. the maximum bandwidth of 10 kHz (given by the bin time) of the optical detection scheme distorts the statistical analysis at trion laser intensities above 9 1.6 · 10−5 µW/µm2. Below this laser intensity, every single Auger recombination event is detected in the real-time telegraph signal. Figure 4: Probability distribution and cumulants of the time-resolved RF random telegraph signal. Panel a and b show the probability P (N ) for a number N of Auger events in a bin time of 200 ms (blue bars) and the Poissonian distribution related to the mean value of the probability P (N ) (red curve). At a trion excitation intensity (laser 1) of 3·10−7 µW/µm2 (panel a), which corresponds to an asymmetry close to -1, the probability P (N ) is close to a Poissonian distribution. At a trion excitation intensity of 6 · 10−6 µW/µm2 (panel b), which corresponds to an asymmetry close to 0, the probability P (N ) is sub-Poissonian. Panel c shows the second (blue) and third (red) normalized cumulant as a function of the asymmetry. Symbols are measured values, lines are calculated curves for a two-state system.36 Finally, full counting statistics35,39,40 is performed on the telegraph signal: Every 15- min long telegraph signal is divided into sections with length t0. The number N of Auger events within the time interval t0 is counted. Figure 4a and b show two examples for the corresponding probability distributions P (N ) in the limit of large t0 (0.2 s). At an asymmetry close to -1 (a trion laser intensity of 3 · 10−7 µW/µm2, Fig. 4a), the probability is close to a Poissonian distribution. At an asymmetry of about 0 (laser intensity of 6 · 10−6 µW/µm2, Fig. 4b), the probability distribution is sub-Poissonian, indicating a relation between Auger recombination and electron tunneling. The Auger recombination emits an electron after an electron has tunneled from the reservoir into the dot. Vice versa, the electron can only tunnel the cumulants Cm(t0) = ∂m M(z, t0) =(cid:80) after the Auger recombination has emptied the QD. From the probability distributions, z lnM(z, t0)z=0 can be derived with the generating function N ezN P (N ).35 The first cumulant C1 corresponds to the mean value, the second cumulant C2 is the variance and the third one describes the skewness of the distribution. The 10 ProbabilityP(N)(%) Auger events in 0.2 sAuger events in 0.2 s0a102005103⋅10−7µW/µm2bProbabilityP(N)(%)6080024686⋅10−6µW/µm2cAsymmetrya01-10.00.51.0Norm. CumulantC2/C1C3/C1second and third normalized cumulant in the limit of large t0 (20 ms and 5 ms, respectively) can be seen as data points in Figure 4c. For a two-state system, theory predicts for these normalized cumulants in the long-time limit C2/C1 = (1 + a2)/2 (also called "Fano factor") and C3/C1 = (1 + 3a4)/4,36 shown as lines in Figure 4c. The data for the second and third normalized cumulant coincide perfectly with the calculated curves. We can conclude from the statistical analysis that the QD behaves like a two-state system, where one state is the QD charged with one electron (or a trion after optical excitation) and the other state is the empty dot (or charged with an exciton). The QD charged with one electron cannot be distinguished from the dot containing a trion (same for empty dot and exciton) as the optical transition times in the order of nanoseconds are orders of magnitude faster than the tunneling and emission time by the Auger recombination.41 The statistical analysis demonstrates the influence of the Auger recombination on the cumulants, especially on the Fano factor, which can be tune from F = 1 to F = 0.5 by increasing the incident laser intensity on the trion transition. In summary, we performed real-time RF random telegraph measurements and studied full counting statistics of the Auger effect in a single self-assembled QD. With this technique, we were able to measure every single Auger recombination as a quantum jump from a charged to an uncharged QD; followed by single-electron tunneling. The full counting statistics gives access to the normalized cumulants and demonstrates the tunability of the Fano factor from Possonian to sub-Poissonian distribution by the incident laser intensity on the trion transition. Comparison with theoretical prediction shows that the empty and charged QD with the Auger recombination and tunneling follows a dynamical two-state system. For future quantum state preparation, the Auger process can be used to control optically the charge state in a quantum system by optical means. 11 Methods As the same measurement technique is used, this methods section follows the Supplemental information of Kurzmann et al.29 Optical measurements Resonant optical excitation and collection of the fluorescence light is used to detect the optical response of the single self-assembled QD, where the resonance condition is achieved by applying a specific gate voltage between the gate electrode and the Ohmic back contact. The QD sample is mounted on a piezo-controlled stage under an objective lens with a numerical aperture of N A = 0.65, giving a focal spot size of about 1 µm diameter. All experiments are carried out in a liquid He confocal dark-field microscope at 4.2 K with a tunable diode laser for excitation and an avalanche photodiode (APD) for fluorescence detection. The resonant laser excitation and fluorescence detection is aligned along the same path with a microscope head that contains a 90:10 beam splitter and two polarizers. Cross-polarization enables a suppression of the spurious laser scattering into the detection path by a factor of more than 107. The counts of the APD (dead time of 21.5 ns) were binned by a QuTau time-to-digital converter with a temporal resolution of 81 ps. Acknowledgement This work was supported by the German Research Foundation (DFG) within the Collab- orative Research Centre (SFB) 1242, Project No. 278162697 (TP A01), and the individ- ual research grant No. GE2141/5-1. A. Lu. acknowledges gratefully support of the DFG by project LU2051/1-1 and together with A. D. W. support by DFG-TRR160, BMBF - Q.Link.X 16KIS0867, and the DFH/UFA CDFA-05-06. 12 Supporting Information Available The following files are available free of charge. Sample and device fabrication and Excitation laser intensity dependent resonance fluorescence. References (1) Bimberg, D.; Grundmann, M.; Ledentsov, N. N. Quantum Dot Heterostructures; John Wiley & Sons, 1999. (2) Petroff, P. M.; Lorke, A.; Imamoglu, A. Epitaxially Self-Assembled Quantum Dots. Physics Today 2001, 54, 46 -- 52. (3) Michler, P. A Quantum Dot Single-Photon Turnstile Device. Science 2000, 290, 2282 -- 2285. (4) Yuan, Z. Electrically Driven Single-Photon Source. Science 2001, 295, 102 -- 105. (5) Santori, C.; Fattal, D.; Vučković, J.; Solomon, G. S.; Yamamoto, Y. Indistinguishable photons from a single-photon device. Nature 2002, 419, 594 -- 597. (6) Matthiesen, C.; Geller, M.; Schulte, C. H.; Le Gall, C.; Hansom, J.; Li, Z.; Hugues, M.; Clarke, E.; Atatüre, M. Phase-locked indistinguishable photons with synthesized wave- forms from a solid-state source. Nat. Commun. 2013, 4, 1600. (7) Kimble, H. J. The quantum internet. Nature 2008, 453, 1023 -- 1030. (8) Ladd, T. D.; Jelezko, F.; Laflamme, R.; Nakamura, Y.; Monroe, C.; O'Brien, J. L. Quantum computers. Nature 2010, 464, 45 -- 53. (9) Kurzmann, A.; Merkel, B.; Labud, P.; Ludwig, A.; Wieck, A.; Lorke, A.; Geller, M. Optical blocking of electron tunneling into a single self-assembled quantum dot. Phys. Rev. Lett. 2016, 117, 017401. 13 (10) Geller, M. Nonequilibrium carrier dynamics in self-assembled quantum dots. Appl. Phys. Rev. 2019, 6, 031306. (11) Bayer, M.; Stern, O.; Hawrylak, P.; Fafard, S.; Forchel, A. Hidden symmetries in the energy levels of excitonic "artificial atoms". Nature 2000, 405, 923 -- 926. (12) A. Nick Vamivakas, C.-Y. L. . M. A., Yong Zhao Spin-resolved quantum-dot resonance fluorescence. Nat. Phys. 2009, 5, 198 -- 202. (13) Labud, P.; Ludwig, A.; Wieck, A.; Bester, G.; Reuter, D. Direct Quantitative Electrical Measurement of Many-Body Interactions in Exciton Complexes in InAs Quantum Dots. Phys. Rev. Lett. 2014, 112, 046803. (14) Kharchenko, V.; Rosen, M. Auger relaxation processes in semiconductor nanocrystals and quantum wells. J. Lumin. 1996, 70, 158 -- 169. (15) Efros, A. L.; Rosen, M. Random Telegraph Signal in the Photoluminescence Intensity of a Single Quantum Dot. Phys. Rev. Lett. 1997, 78, 1110 -- 1113. (16) Fisher, B.; Caruge, J.-M.; Chan, Y.-T.; Halpert, J.; Bawendi, M. G. Multiexciton fluorescence from semiconductor nanocrystals. Chem. Phys. 2005, 318, 71 -- 81. (17) Jha, P. P.; Guyot-Sionnest, P. Trion Decay in Colloidal Quantum Dots. ACS Nano 2009, 3, 1011 -- 1015. (18) Vaxenburg, R.; Rodina, A.; Shabaev, A.; Lifshitz, E.; Efros, A. L. Nonradiative Auger Recombination in Semiconductor Nanocrystals. Nano Lett. 2015, 15, 2092 -- 2098. (19) Klimov, V. I. Quantization of Multiparticle Auger Rates in Semiconductor Quantum Dots. Science 2000, 287, 1011 -- 1013. (20) Park, Y.-S.; Bae, W. K.; Pietryga, J. M.; Klimov, V. I. Auger Recombination of Biexci- tons and Negative and Positive Trions in Individual Quantum Dots. ACS Nano 2014, 8, 7288 -- 7296. 14 (21) Caruge, J. M.; Halpert, J. E.; Wood, V.; Bulović, V.; Bawendi, M. G. Colloidal quantum-dot light-emitting diodes with metal-oxide charge transport layers. Nat. Pho- tonics 2008, 2, 247 -- 250. (22) Cho, K.-S.; Lee, E. K.; Joo, W.-J.; Jang, E.; Kim, T.-H.; Lee, S. J.; Kwon, S.-J.; Han, J. Y.; Kim, B.-K.; Choi, B. L.; Kim, J. M. High-performance crosslinked colloidal quantum-dot light-emitting diodes. Nat. Photonics 2009, 3, 341 -- 345. (23) Brokmann, X.; Messin, G.; Desbiolles, P.; Giacobino, E.; Dahan, M.; Hermier, J. P. Colloidal CdSe/ZnS quantum dots as single-photon sources. New J. Phys. 2004, 6, 99. (24) Michler, P.; Imamoğlu, A.; Mason, M. D.; Carson, P. J.; Strouse, G. F.; Buratto, S. K. Quantum correlation among photons from a single quantum dot at room temperature. Nature 2000, 406, 968 -- 970. (25) Lounis, B.; Bechtel, H.; Gerion, D.; Alivisatos, P.; Moerner, W. Photon antibunching in single CdSe/ZnS quantum dot fluorescence. Chem. Phys. Lett. 2000, 329, 399 -- 404. (26) Kurzmann, A.; Ludwig, A.; Wieck, A. D.; Lorke, A.; Geller, M. Auger Recombination in Self-Assembled Quantum Dots: Quenching and Broadening of the Charged Exciton Transition. Nano Lett. 2016, 16, 3367 -- 3372. (27) Levitov, L. S.; Lee, H.; Lesovik, G. B. Electron counting statistics and coherent states of electric current. J. Math. Phys. 1996, 37, 4845 -- 4866. (28) Blanter, Y.; BÃijttiker, M. Shot noise in mesoscopic conductors. Phys. Rep. 2000, 336, 1 -- 166. (29) Kurzmann, A.; Stegmann, P.; Kerski, J.; Schott, R.; Ludwig, A.; Wieck, A.; KÃűnig, J.; Lorke, A.; Geller, M. Optical Detection of Single-Electron Tunneling into a Semicon- ductor Quantum Dot. Phys. Rev. Lett. 2019, 122, 247403. 15 (30) Ludwig, A.; Prechtel, J. H.; Kuhlmann, A. V.; Houel, J.; Valentin, S. R.; Warbur- ton, R. J.; Wieck, A. D. Ultra-low charge and spin noise in self-assembled quantum dots. J. Cryst. Growth 2017, 477, 193 -- 196. (31) HÃűgele, A.; Seidl, S.; Kroner, M.; Karrai, K.; Warburton, R. J.; Gerardot, B. D.; Petroff, P. M. Voltage-Controlled Optics of a Quantum Dot. Phys. Rev. Lett. 2004, 93, 217401. (32) Li, S.-S.; Xia, J.-B. Quantum-confined Stark effects of InAs/GaAs self-assembled quan- tum dot. J. Appl. Phys. 2000, 88, 7171 -- 7174. (33) Seidl, S.; Kroner, M.; Dalgarno, P. A.; HÃűgele, A.; Smith, J. M.; Ediger, M.; Ger- ardot, B. D.; Garcia, J. M.; Petroff, P. M.; Karrai, K.; Warburton, R. J. Absorption and photoluminescence spectroscopy on a single self-assembled charge-tunable quantum dot. Phys. Rev. B 2005, 72, 195339. (34) Kurzmann, A.; Ludwig, A.; Wieck, A. D.; Lorke, A.; Geller, M. Photoelectron gener- ation and capture in the resonance fluorescence of a quantum dot. Appl. Phys. Lett. 2016, 108, 263108. (35) Gustavsson, S.; Leturcq, R.; Studer, M.; Shorubalko, I.; Ihn, T.; Ensslin, K.; Driscoll, D.; Gossard, A. Electron counting in quantum dots. Surf. Sci. Rep. 2009, 64, 191 -- 232. (36) Gustavsson, S.; Leturcq, R.; Simovič, B.; Schleser, R.; Ihn, T.; Studerus, P.; Ensslin, K.; Driscoll, D. C.; Gossard, A. C. Counting Statistics of Single Electron Transport in a Quantum Dot. Phys. Rev. Lett. 2006, 96, 076605. (37) Loudon, R. The Quantum Theory of Light; Oxford University Press, 2000. (38) Flindt, C.; Fricke, C.; Hohls, F.; Novotny, T.; Netocny, K.; Brandes, T.; Haug, R. J. Universal oscillations in counting statistics. PNAS 2009, 106, 10116 -- 10119. 16 (39) Fricke, C.; Hohls, F.; Wegscheider, W.; Haug, R. J. Bimodal counting statistics in single-electron tunneling through a quantum dot. Phys. Rev. B 2007, 76, 155307. (40) Gorman, S.; He, Y.; House, M.; Keizer, J.; Keith, D.; Fricke, L.; Hile, S.; Broome, M.; Simmons, M. Tunneling Statistics for Analysis of Spin-Readout Fidelity. Phys. Rev. Appl. 2017, 8, 034019. (41) Zrenner, A.; Beham, E.; Stufler, S.; Findeis, F.; Bichler, M.; Abstreiter, G. Coherent properties of a two-level system based on a quantum-dot photodiode. Nature 2002, 418, 612 -- 614. 17
1104.4915	1	1104	2011-04-26T13:20:42	Circularly polarized electroluminescence from silicon nanostructures heavily doped with boron	[ "cond-mat.mes-hall" ]	The circularly polarized electroluminescence (CPEL) from silicon nanostructures which are the p-type ultra-narrow silicon quantum well (Si-QW) confined by {\delta}-barriers heavily doped with boron, 5 10^21 cm^-3, is under study as a function of temperature and excitation levels. The CPEL dependences on the forward current and temperature show the circularly polarized light emission which appears to be caused by the exciton recombination through the negative-U dipole boron centers at the Si-QW {\delta}-barriers interface.	cond-mat.mes-hall	cond-mat	The following article has been submitted to/accepted by Applied Physics Letters. After it is published, it will be found at http://apl.aip.org/ Circularly polarized electroluminescence from silicon nanostructures heavily doped with boron N.T. Bagraev, L.E. Klyachkin, R.V. Kuzmina), A.M. Malyarenko Ioffe Physical-Technical Institute of the Russian Academy of Sciences, St.Petersburg, 194021, Russia a)Electronic mail: [email protected] Abstract The circularly polarized electroluminescence (CPEL) from silicon nanostructures which are the p-type ultra-narrow silicon quantum well (Si-QW) confined by δ-barriers heavily doped with boron, 5·1021 cm-3, is under study as a function of temperature and excitation levels. The CPEL dependences on the forward current and temperature show the circularly polarized light emission which appears to be caused by the exciton recombination through the negative-U dipole boron centers at the Si-QW δ-barriers interface. Possibility to create a fully silicon-compatible optoelectronics attracts a lot of attention to the problem of silicon light emission source.1 Numerous approaches, such as porous Si, Si nanocrystals, erbium doped Si, plastically deformed monocrystalline silicon were prepared to obtain an effective luminescence from silicon-based material.2 -6 Last decade, many observations of the high intensive near band-edge electroluminescence (EL) from the silicon p-n junctions have been documented.7-10 This quite surprising result for indirect band gap semiconductors was commonly accounted for by carrier confinement in the dislocation lops introduced by the boron ion implantation process. However the high intensity of the near band-edge electroluminescence was also observed from the p-n junctions prepared by boron diffusion, which seem to avoid the formation of the dislocation loops. 11 Furthermore this emission was found to be not related directly to the presence of the dislocations which may affect only as a getter of nonradiative recombination centers.12 Nevertheless there is a common feature of the most works used either ion implantation or diffusion of boron. This feature is rather high level of boron doping, when its ordinary concentration is near the solubility limit. Under these conditions the formation of the locally enhanced boron doping spikes appears to take place.13,14 Such spikes were shown to confine excitons effectively and thus provide the intensive near band-edge electroluminescence by 1 Fig. 1. (a) A schematic diagram of the device that demonstrates a perspective view of the p-type Si-QW confined by the δ-barriers heavily doped with boron on the n-type Si (100) surface. (b) The forward current-voltage characteristic of the p+-n junction that defines the energies of the two-dimensional subbands of 2D holes in the p- type Si-QW confined by δ-barriers (c). Fig. 2. Left- (ILCP) and right-circularly polarized (IRCP) EL spectra components of the p-type silicon nanostructure on the n-type Si (100) surface detected at the forward current of 20 mA for 77 K and 300 K. 2 releasing free electron-hole pairs at elevated temperatures.15 The EL intensity is dependent on the concentration of boron reaching a maximum at N(B) ~ 4·1020 cm-3.15 The formation of the boron clusters which is unavoidable process within frameworks of the standard impurity doping techniques is able to prevent further EL intensity enhancement. However this limit appeared to be avoided if short-time vacancy enhanced diffusion of boron is used after preliminary oxidation of the n-type Si (100) surface. Under fine injection of vacancies during diffusion process the ultra-narrow Si-QW, 2 nm, has been established to be self-assembled between the δ-barriers heavily doped with boron in the concentration of 5·1021 cm-3, which is controlled by the secondary- ion mass spectroscopy (SIMS) method. Here we present the first findings of the near band-edge electroluminescence from these p-type silicon nanostructures on the n-type Si (100) surface. The electroluminescence appears to reveal rather high intensity even at room temperature and moreover the high degree of circular polarization. The CPEL mechanism seems to be started from the recombination of the excitons bounded by the negative-U dipole boron centers at the Si-QW - δ-barriers interface. We used a 350 μm thick n-Si (100) wafer of 20 Ω·cm-1 resistivity. The short-time diffusion of boron was performed at the temperature of 900°C after preliminary oxidation in the presence of dry oxygen and chlorine compounds as well as subsequent photolithography and etching. The depth of the ultra-shallow boron diffusion profile determined by SIMS method was 8 nm. For further electroluminescence measurements the gold contacts were prepared on the front surface of the sample using standard photolithography technique, whereas the back side of the substrate was covered by aluminum. The device was designed within frameworks of the Hall geometry with the doping area of size 4.7·0.1 mm which is provided by eight 200·200 μm contacts as shown in Fig. 1a. The forward current-voltage characteristic of the p+-n junction prepared is shown in Fig. 1b which reveals the system of energy levels inside the p-type Si-QW (Fig. 1c). The samples obtained were set up in a cryostat for the EL measurements in the temperature range from 4.2 to 300 K. The electroluminescence was detected with the MDR-23 monochromator using an InGaAs photodiode and a conventional lock-in technique. The EL spectra detected reveal the left- and right-hand circular polarization with high degree up to 20% (see Fig. 2). The main emission line in the low-temperature EL spectrum is observed at 1126 nm with the full-width at half-maximum (FWHM) of 18 nm and the phonon replica at 1194 nm. Under elevated temperature, this line saves the circular polarization decreasing slightly in intensity and increasing in FWHM to 70 nm as well as shifting to 1163 nm following by the band gap temperature dependence. The intensity of the main emission line was calibrated with a gauged source in the same spectral range. The spatial distribution of the electroluminescence was taken into account. The EL power obtained is shown in Fig. 3 as a 3 Fig. 3. The electroluminescence power as a function of forward current at various temperatures: 77 K (squares), 300 K (triangles). Inset: low current part of the function. Fig. 4. The temperature dependence of the integral electroluminescence intensity of the main emission line revealing many features near 150 K which correlate with the value of the critical temperature of the superconductor transition, Tc~145 K, estimated from electrical resistivity (see inset), thermo-emf and magnetic susceptibility measurements. 4 function of the forward current, which exhibits the linear character at temperatures of 77 K and 300 K except low values. It should be noted that the external quantum efficiency is difficult to be estimated due to the presence of many emission bands in very wide spectral range from visible to far infrared regions.16,17 So it is hard to know which part of the input power participates in the emission under the study. To clarify the CPEL mechanism, the structure of the boron ultra-shallow diffusion profiles should be defined. The cyclotron resonance (CR) and ESR measurements as well as the scanning tunneling microscopy (STM) technique were used. The CR measurements were performed at 3.8 K with a standard Brucker-Physik AG ESR spectrometer at X-band (9.1-9.5 GHz). The CR quenching and the angular dependence for which a characteristic 180o symmetry was observed revealed the confining potential inside p+-diffusion profile and its different arrangement in longitudinal and lateral Si-QW s, which are formed naturally between the δ-barriers heavily doped with boron.18,19 These Si-QW s were shown to contain the high mobility 2D hole gas which is characterized by long momentum relaxation time of heavy and light holes at 3.8 K.18-20 The STM technique revealed the formation of the self-interstitials microdefects inside the δ- barriers confining Si-QW . The size of the smallest microdefect obtained is 2 nm, which is consistent with the parameters expected from the tetrahedral model of the Si60 cluster.21 The STM images of larger scales demonstrated that the ratio between the dimensions of the microdefects is supported to be equal to 3.3 thereby defining the self-assembly of the self- interstitials microdefects as the self-organization of the fractal type. Thus the δ-barriers heavily doped with boron, 5·1021 cm-3, represent alternating arrays of the smallest undoped microdefects and doped dots with dimensions restricted to 2 nm. The value of the boron concentration determined by the SIMS method seems to indicate that each doped dot located between undoped microdefects contains two impurity atoms of boron. To study of the boron centers packed up in these dots the ESR technique was applied. The same Brucker-Physik AG ESR spectrometer at X-band (9.1-9.5 GHz) was used with the rotation of the magnetic field in the {110}-plane perpendicular to a {100}-interface. The angular dependences of the ESR spectra showed that the doped microdefects appear to consist of the trigonal dipole boron centers, B+ - B-, which are caused by the elastic reconstruction along the [111] crystallographic axis of the shallow acceptors as a result of the negative-U reaction: 2Bo → B+ + B-.22 The trigonal ESR spectrum seems to be evidence of the dynamic magnetic moment that is induced by the exchange interaction between the small hole bipolarons which are formed by this negative-U reconstruction of the boron acceptors.18,22 Thus the dipole boron centres are a basis of nanostructured δ - barriers confining the Si-QW. Furthermore, it was found by the electrical resistivity, thermo-emf and magnetic susceptibility measurements that the δ-barriers heavily doped with boron, N(B) = 5·1021 cm-3, reveal a phase transition at critical temperature of Tc~145 5 K which is accompanied with appearing of the superconductor properties caused by transferring of the small hole bipolarons through these negative-U dipole boron centers.17 It seems to be likely that the microdefects containing negative-U dipole centers of boron should to affect to the electroluminescence emitted from the silicon nanostructure prepared. As a proof of this assumption the temperature dependence of the electroluminescence intensity, as can be seen in Fig. 4, increases with the temperature until reaches a maximum in a vicinity of 150 K. This EL temperature dependence seems to be result from the exciton confinement caused by the chains of the negative-U dipole boron centers which collapse at elevated temperatures. The temperature of the EL intensity quenching is in good agreement with the value of the critical temperature of the superconductor transition, Tc~145 K, estimated from electrical resistivity (see inset in Fig. 4), thermo-emf and magnetic susceptibility measurements. The fact of the circular polarization of the electroluminescence can also be referred to the negative-U reconstruction of the shallow boron acceptors, because the high temperature superconductor properties for the δ- barriers was found to be a result of coherent state of the small hole bipolarons on the dipole centers of boron.17,22 In that case the high degree CPEL may be a result of breakdown of the coherent state of small hole bipolarons at these negative-U centers. Besides, Fig. 2 shows that the high degree CPEL is retained up to room temperature, which appears to be due to the conservation of the spin-lattice relaxation time as a function of temperature in ultra-narrow quantum wells, wires etc.23 The strong highly polarized electroluminescence from the p-type ultra-narrow Si-QW confined by the δ-barriers heavily doped with boron, 5·1021 cm-3, on the n-type Si (100) surface has been demonstrated. The EL studies on different temperatures and excitation levels have shown that the mechanism of the efficient circularly polarized electroluminescence seems to be started from the recombination of the excitons confined by the chains of the negative-U dipole boron centers. The work was supported by the programme of fundamental studies of the Presidium of the Russian Academy of Sciences “Quantum Physics of Condensed Matter” (grant 9.12); programme of the Swiss National Science Foundation (grant IZ73Z0_127945/1); the Federal Targeted Programme on Research and Development in Priority Areas for the Russian Science and Technology Complex in 2007–2012 (contract no. 02.514.11.4074), the SEVENTH FRAMEWORK PROGRAMME Marie Curie Actions PIRSES-GA-2009-246784 project SPINMET. 6 1 L. Pavesi, J. Phys.: Condens. Matter 15, R1169 (2003). 2 M. Stutzmann, M. Brandt, M. Rosenbauer, J. Weber, and H. Fuchs, Phys. Rev. B 47, 4806 (1993). 3 L. Pavesi, L. Dal Negro, C. Mazzoleni, G. Franzò, and F. Priolo, Nature 408, 440 (2000). 4 B. Zheng, J. Michel, F. Y. G. Ren, L. C. Kimerling, D. C. Jacobson, and J. M. Poate, Appl. Phys. Lett. 64, 2842 (1994). 5 V. Kveder, M. Badylevich, W. Schröter, M. Seibt, E. Steinman, and A. Izotov, Phys. Status Solidi (a) 202, 901 (2005). 6 V. Kveder, M . Badylevich, E. Steinman, A. Izotov, M . Seibt, and W . Schröter, Appl. Phys. Lett. 84, 2106 (2004). 7 Wai Lek Ng, M . A. Lourenco, R. M . Gwilliam, S. Ledain, G. Shao and K. P. Homewood, Nature 410, 1036 (2001). 8 M. A. Green, J. Zhao, A. Wang, P. J. Reece, and M. Gal, Nature 412, 805 (2001). 9 J. Zhao, G. Zhang, T. Trupke, A. Wang, F. Hudert, and M. A. Green, Appl. Phys. Lett. 85, 2830 (2004). 10 A. M. Emel’yanov and N. A. Sobolev, Semiconductors 42, 336 (2008). 11 Pavel Altukhov and Evgenii Kuzminov, Phys. Status Solidi (b) 245, 1181 (2008). 12 K. Fraser, D. Stowe, S. Galloway, S. Senkader, R. Falster, and P. Wilshaw, Phys. Status Solidi (c) 4, 2977 (2007). 13 Lourdes Pelaz, G. H. Gilmer, V. C. Venezia, H.-J. Gossmann, M. Jaraiz, and J. Barbolla, Appl. Phys. Lett. 74, 2017 (1999). 14 Maria Aboy, Lourdes Pelaz, Luis A. Marqués, Juan Barbolla, Ali Mokhberi, Yayoi Takamura, Peter B. Griffin, and James D. Plummer, Appl. Phys. Lett. 83, 4166 (2003). 15 J. Sun, T. Dekorsy, W. Skorupa, B. Schmidt, A. Mücklich, and M. Helm, Phys. Rev. B 70, 155316 (2004). 16 N.T. Bagraev, A.D. Bouravleuv, W. Gehlhoff, L.E. Klyachkin, and A.M. Malyarenko, Proc. Of the 26th International Conference On the Physics Of Semiconductors (ICPS-26), Edinburgh, U.K., Physics Of Semiconductors 2002, Ed. By A.R. Long and J.H. Davies, Inst. of Physics, Conference Series 171, G3.3 (2002). 17 N. T. Bagraev, L. E. Klyachkin, A. A. Koudryavtsev, A. M. Malyarenko, and V. V. Romanov, Semiconductors 43, 1441 (2009). 18 W. Gehlhoff, N.T. Bagraev, and L.E. Klyachkin, Materials Science Forum 196-201, 467-472 (1995). 19 W. Gehlhoff, N.T. Bagraev, and L.E. Klyachkin, Solid State Phenom. 47-48, 589-594 (1996). 7 20 N. T. Bagraev, A. D. Bouravlev, L. E. Klyachkin, A. M. Malyarenko, W. Gehlhoff, Yu. I. Romanov, and S. A. Rykov, Semiconductors 39, 685 (2005). 21 Bao-xing Li, Pei-lin Cao, and Duan-lin Que, Phys. Rev. B 61, 1685 (2000). 22 N. Bagraev, W. Gehlhoff, L. Klyachkin, A. Malyarenko, V. Romanov, and S. Rykov, Physica C 437-438, 21-24 (2006). 23 Alexander Khaetskii and Yuli Nazarov, Phys. Rev. B 61, 12639 (2000). 8
1005.4739	3	1005	2010-12-13T09:13:34	Thermoelectric Transport of Massive Dirac Fermions in Bilayer Graphene	[ "cond-mat.mes-hall" ]	Thermoelectric power (TEP) is measured in bilayer graphene for various temperatures and charge-carrier densities. At low temperatures, measured TEP well follows the semiclassical Mott formula with a hyperbolic dispersion relation. TEP for a high carrier density shows a linear temperature dependence, which demonstrates a weak electron-phonon interaction in the bilayer graphene. For a low carrier density, a deviation from the Mott relation is observed at high temperatures and is attributed to the low Fermi temperature in the bilayer graphene. Oscillating TEP and the Nernst effect for varying carrier density, observed in a high magnetic field, are qualitatively explained by the two dimensionality of the system.	cond-mat.mes-hall	cond-mat	PHYSICAL REVIEW B 82, 245416 共2010兲 Thermoelectric transport of massive Dirac fermions in bilayer graphene Seung-Geol Nam,1 Dong-Keun Ki,1,2 and Hu-Jong Lee1,* 1Department of Physics, Pohang University of Science and Technology, Pohang, Republic of Korea 2DPMC and GAP, University of Geneva, 24 Quai Ernest-Ansermet, CH1211 Geneva, Switzerland 共Received 19 September 2010; published 10 December 2010兲 Thermoelectric power 共TEP兲 is measured in bilayer graphene for various temperatures and charge-carrier densities. At low temperatures, measured TEP well follows the semiclassical Mott formula with a hyperbolic dispersion relation. TEP for a high carrier density shows a linear temperature dependence, which demonstrates a weak electron-phonon interaction in the bilayer graphene. For a low carrier density, a deviation from the Mott relation is observed at high temperatures and is attributed to the low Fermi temperature in the bilayer graphene. Oscillating TEP and the Nernst effect for varying carrier density, observed in a high magnetic field, are qualitatively explained by the two dimensionality of the system. PACS number共s兲: 73.63.⫺b, 72.15.Jf, 73.22.Pr, 73.50.Lw DOI: 10.1103/PhysRevB.82.245416 I. INTRODUCTION Recently, gate voltage 共VBG兲 and temperature 共T兲 depen- dences of the thermoelectric power 共TEP兲 were measured in monolayer graphene, a single atomic layer of graphite.1–3 TEP is defined as the ratio of the thermoelectric voltage 共V th兲 to the temperature difference 共⌬T兲 across a sample. Since TEP is the electrical response to the thermodynamic pertur- bation, it is sensitive to the electronic structure near the Fermi energy. TEP in monolayer graphene was found to be dominated by diffusive carriers in the temperature range of T ⬍ 300 K. Thus, by employing the semiclassical Mott relation,4 the linear-dispersion relation as well as the nature of impurity scattering in monolayer graphene has been investigated.1,2,5 two weekly coupled Bilayer graphene, a stack of graphene layers, also has attracted high interest owing to its plausible device applications6,7 as well as fundamental interests.8 Recently, the transverse thermoelectric conductiv- ity 共␣xy兲 of bilayer graphene was measured, mainly in the quantum Hall regime.9 In the report, ␣xy increases linearly at low T and saturates close to the predicted universal value at high T. Also, an anomaly in ␣xy is found near the charge neutrality point 共CNP兲, which may be related to the counter- propagating edge states at the CNP. Compared with the case of monolayer graphene, however, reports on TEP measure- ments in bilayer graphene are still scarce. Similar to mono- layer graphene, TEP measurements in bilayer graphene in zero magnetic field are also expected to provide the detailed information on the dispersion relation, the electron-phonon interaction,10 and the yet controversial scattering mechanism in the system.11–14 In addition, the transverse component of TEP 共the Nernst signal兲 in the quantum Hall regime can be used to examine the dimensionality of graphene stacks.15–17 In this study, we report the gate-voltage dependence of TEP taken in bilayer graphene for different temperatures and magnetic fields. By assuming hyperbolic dispersion relation, we confirm that measured TEP shows a good agreement with the Mott relation at low temperatures, which demonstrates the validity of the semiclassical Boltzmann theory in this system. The phonon-drag effect was not observed, indicating a weak electron-phonon coupling in bilayer graphene. For a low carrier density, however, distinct from the monolayer graphene, a deviation from the Mott relation is observed along with a saturating tendency of the TEP at high tempera- tures, which we attribute to the low Fermi temperature 共TF兲 of the bilayer graphene. In a high magnetic field, TEP and the Nernst signal showed oscillations with changing back- gate voltage, which are qualitatively explained by the two dimensionality of the system. II. SAMPLE PREPARATION AND MEASUREMENTS The inset of Fig. 1共b兲 shows the typical configuration of devices. Bilayer graphene was prepared by the mechanical exfoliation on 300 nm SiO2/highly doped Si substrate. After a piece of suitable bilayer graphene is selected under micro- scope, electrodes were attached by e-beam patterning and subsequent Ti/Au 共3/17 nm兲 evaporation. Details of the fab- rication processes are reported elsewhere.18 After the sample fabrication, quantum Hall effect was measured to confirm the number of graphene layers in a sample. Typical mobility of our sample was 2000 – 4000 cm2 / V s. To set up a tempera- ture difference ⌬T 共ⰆT兲 across a sample, a sufficient current was applied through a meandering heater line and ⌬T was determined from the four-terminal resistances of two local thermometers shown in the inset of Fig. 1共b兲. The thermo- electric voltage V th was measured by the ac lock-in tech- nique, where the second-harmonic signal was detected as a low-frequency ac current 共 f ⬍ 2 Hz兲 was driven to the mi- croheater. Similar technique of setting up ⌬T and measuring V th were employed in earlier studies.1–3,19 The TEP 共Sxx兲 and the Nernst coefficient 共Syx兲 are obtained from the relations Sxx = −Ex / 兩ⵜT兩 and Syx = Ey / 兩ⵜT兩, respectively. Here, Ex共y 兲 is the longitudinal 共transverse兲 component of electric field and ⵜT is the longitudinal temperature gradient. Measurements were made using a homemade cryostat in a temperature range of 4.2–300 K and in magnetic fields up to 7 T. III. RESULTS AND DISCUSSION Zero-field two-terminal resistance R and the longitudinal component of TEP Sxx are shown as a function of the back- gate voltage VBG in Figs. 1共a兲 and 1共b兲, respectively. R re- 1098-0121/2010/82共24兲/245416共5兲 245416-1 ©2010 The American Physical Society NAM, KI, AND LEE PHYSICAL REVIEW B 82, 245416 共2010兲 FIG. 1. 共Color online兲共a兲 Two-terminal resistance R and 共b兲 the TEP as a function of the backgate voltage VBG in zero magnetic field and at various temperatures: T = 30, 50, 70, 140, 170, and 250 K. Solid arrows indicate VBG’s where data points in Figs. 3共a兲–3共c兲 are taken. Inset: optical image of a typical device. Size of the scale bar is 30 ␮m. veals a maximum at the CNP 共VCNP ⬃ −4 V for this sample兲. Consistent with previous studies,20,21 R near the CNP in- creases with decreasing T due to the thermal excitation of carriers, which is dominant near the CNP. Different from previous report,20 however, a small but finite T dependence of the resistance is observed away from the CNP, presumably due to the presence of electron-hole puddles.22,23 As in monolayer graphene,1–3 the ambipolar nature of the carriers manifests itself as the sign change in Sxx in Fig. 1共b兲 at the CNP. For a quantitative analysis, we adopt the semiclassical Mott relation for TEP,4 冏冏 dVBG dG dVBG dE 2 T ␲2kB 3e SM ott = − 1 G , E=EF 共1兲 where kB, e, and G are the Boltzmann constant, the electron charge, and the conductance, respectively. The only experi- mentally undetermined term in Eq. 共1兲 is dVBG / dE or the dispersion relation of the system E共k兲, where the Fermi wave vector kF is related to VBG as kF = 冑␲兩ne,h兩 = 冑␲␣兩VBG兩共ne,h is the electron or hole carrier density and ␣ is a geometry- dependent capacitance of the device兲. According to a tight- binding model in first-order approximation,24 the dispersion relation of bilayer graphene is hyperbolic as 共2兲 E = ⫾ FIG. 2. 共Color online兲 The TEP measured 共filled circle兲 and calculated from the Mott relation 共solid line兲 as a function of VBG at 共a兲 T = 30 K, 共b兲 T = 50 K, 共c兲 T = 140 K, and 共d兲 T = 250 K. ␥1关冑1 + 4vF 1 2 ប2k2/␥ 2 − 1兴 , 1 2 where ␥1 is the interlayer hopping energy, vF is the Fermi velocity, and ប is the Planck’s constant. In Fig. 2, for various temperatures, we compare the mea- meas, with the ones estimated from the Mott re- sured TEP, Sxx M ott, by adopting the dispersion relation in Eq. 共2兲. lation, Sxx Here and afterward, we fix the fitting parameters as vF ⬵ 0.95 ⫻ 106 m / s and ␥1 = 0.39 eV, which are comparable to the results of earlier spectroscopic measurements.25 At low temperatures of T = 30 K and T = 50 K 关Figs. 2共a兲 and 2共b兲兴, meas is in excellent agreement with Sxx M ott, which demon- Sxx strates the validity of the semiclassical Boltzmann theory in this system. Here, we note that, even in the presence of a perpendicu- lar electric field that may induce a band gap at the CNP,6,26 the gapless hyperbolic dispersion relation reproduces the ex- perimental data very well. Moreover, a large enhancement of TEP,27 which is expected in bilayer graphene with a large band gap, was not observed. For a high carrier density, our observation is understandable as the change in the density of states 共DOS兲 at the Fermi energy due to the band-gap open- ing can be neglected.26,12 However, the signature of the band-gap opening is also absent near the CNP where the change in the DOS is no longer negligible. We suspect that it was caused by large fluctuations of the disorder potential23 compared to the size of the band gap. meas reveals a deviation from the At high temperatures, Sxx M ott near the CNP and the deviation increases with T 关Figs. Sxx 2共c兲 and 2共d兲兴. Besides, with increasing T, the range of VBG where the deviation occurs becomes wider. For detailed meas and Sxx M ott as a function of T in Figs. analysis, we plot Sxx 3共a兲–3共c兲 for different fixed values of VBG’s marked by solid arrows in Fig. 1共b兲. 245416-2 THERMOELECTRIC TRANSPORT OF MASSIVE DIRAC… PHYSICAL REVIEW B 82, 245416 共2010兲 FIG. 3. 共Color online兲关共a兲–共c兲兴 Temperature dependence of the TEP, measured and calculated from the Mott relation, at 共a兲 VBG = −55 V, 共b兲 VBG = −25 V, and 共c兲 VBG = −15 V. Dashed line is a guide to eyes and the solid arrow indicates Tⴱ, at which the mea- sured TEP starts deviating from the Mott relation. 共d兲 The ratio of Tⴱ to TF for two different samples. As seen in Fig. 3共a兲, for VBG sufficiently away from the meas well follows Sxx M ott with a linear T dependence. It CNP, Sxx indicates a weak electron-phonon interaction in bilayer graphene as a phonon-drag component generates the nonlin- ear T dependence.5 The linear T dependence of the TEP in both bilayer and monolayer graphene1–3 is in sharp contrast with that in graphite,28 where a large negative peak appears at low temperatures due to the dominant phonon-drag con- tribution. Thus, it will be interesting to study how the phonon-drag effect evolves from monolayer graphene to graphite as the number of graphene layer increases. meas deviates from Sxx M ott at a As VBG approaches the CNP, Sxx meas and certain onset temperature Tⴱ, around which both Sxx M ott also deviate from the linear T dependence 关Figs. 3共b兲 Sxx and 3共c兲兴. Moreover, as indicated by arrows in the figures, Tⴱ decreases with reducing the carrier density. This feature is in clear contrast with that of the monolayer graphene where the linear T dependence holds for T up to room temperature even for the doping levels close to the peak position of TEP.3 M ott, the nonlinear T dependence stems In the case of Sxx from the change in the conductance with T near the CNP 关Fig. 1共a兲兴 by the thermally excited carriers.20,21 On the other meas from the Mott hand, to understand the deviation of Sxx relation and the VBG dependence of Tⴱ, one needs to recall that the Mott relation is derived from the Sommerfeld expan- sion, which is valid only for low T 共ⰆTF兲. Due to the finite DOS in bilayer graphene at the CNP, TF of the system is almost an order of magnitude smaller than that of the mono- layer graphene for a given charge-carrier density.29 Thus, as observed in this study, the Mott relation is expected to fail in bilayer graphene at lower temperatures than in monolayer graphene. We take Tⴱ to be the temperature at which the difference meas becomes 3␮V / K. As Tⴱ is expected M ott and Sxx between Sxx FIG. 4. 共Color online兲 Hall conductivity ␴xy 共solid line兲 together with 共a兲 Sxx and 共b兲 Sxy as a function of VBG at temperature of T = 15 K and in magnetic field of 6.9 T. Inset: peak values of Sxx as a function of the Landau-level index n, in both hole 共open circle兲 and electron 共filled circle兲 sides. Solid line is 1 / 共n + 1 / 2兲 fits to the peak values. to vary in proportion to TF, the reduced onset temperatures Tⴱ / TF for two different devices are plotted in Fig. 3共d兲 as a function of 兩⌬V兩 = 兩VBG − VCNP兩. For most of VBG’s, Tⴱ / TF be- comes ⬃0.2, which is comparable to the theoretically pre- dicted value for the monolayer graphene.5 Near the CNP, however, Tⴱ / TF appears to show a slow increase. Although this possible upswing of Tⴱ / TF potentially leads to new physics, its character should be further clarified with cleaner graphene by overcoming the effect of electron-hole puddles.23 In Figs. 3共b兲 and 3共c兲, Sxx meas also shows a saturating ten- dency for high values of T / TF. The saturation of TEP was theoretically discussed in monolayer graphene by consider- ing different scatterings5 due to short-range disorder, acous- tic phonons, and charged impurities. As no corresponding theoretical analysis exists for bilayer graphene our results will be useful in investigating controversial scattering mechanisms in bilayer graphene.11–14 In a perpendicular magnetic field, VBG dependence of Sxx and Sxy, and the Hall conductivity ␴xy were measured 关see Figs. 4共a兲 and 4共b兲兴. ␴xy exhibits a series of plateaus at ⫾4Ne2 / h for integer N ⱖ 1, which is from the eightfold de- generacy of the zero-energy Landau level.30 As shown in Figs. 4共a兲 and 4共b兲, both Sxx and Sxy oscillate with VBG, simi- lar to the feature observed in a two-dimensional electron gas 共2DEG兲共Ref. 31兲 and in monolayer graphene1–3 in the quan- tum Hall regime. The oscillations emerge as completely 245416-3 NAM, KI, AND LEE PHYSICAL REVIEW B 82, 245416 共2010兲 filled Landau levels do not carry entropy, resulting in vanish- ing Sxx and Sxy for the chemical potential ␮ located between the Landau levels. When ␮ lies at the center of a Landau level, however, Sxx becomes maximal. The peak values of Sxx are predicted to have a series of fractional values as a function of the Landau peak = 共kB / e兲ln 2 / 兺n f 共En兲, where f 共E兲 is the level index n; Sxx Fermi distribution function and En is the energy of the nth peak becomes ln 2共kB / e兲 / 共n + 1 / 2兲 for a Landau level.32 Sxx 2DEG 共Ref. 32兲 and ln 2共kB / e兲 / n for monolayer graphene.3 For bilayer graphene, since the n = 0 Landau level is twice peak becomes iden- more degenerate than the other levels,30 Sxx tical to that of a 2DEG. The peak values of Sxx shown in the inset of Fig. 4共a兲, for both hole and electron sides, roughly follow the 1 / 共n + 1 / 2兲 dependence. However, the maximum values of Sxx are about one half of the expectation, presum- ably due to the thermal broadening of the Landau levels together with the presence of disorder.3,31 The Landau-level broadening can be different for each level, which may be the peak at n = 1 , 2 in the hole side reason for the deviation of Sxx from the 1 / 共n + 1 / 2兲 fitting. Finally, let us focus on the Nernst signal, Sxy. In measure- ments on the Nernst effect in graphite, a three-dimensional system, the band dispersion along the c axis was found to lead to a maximum of Sxy at the center of a Landau level.15 However, as in a 2DEG 共Ref. 31兲 as well as in monolayer graphene,1–3 Sxy in Fig. 4共b兲 vanishes at the center of Landau levels, which indicates that a stack of graphene consisting of up to two layers remains to be two dimensional. IV. SUMMARY We measured the thermoelectric power in bilayer graphene at various temperatures and charge-carrier densi- ties. The low-T results of TEP well follows the Mott relation with the hyperbolic band structure of bilayer graphene. The Mott relation of TEP holds well with a linear T dependence for low values of T / TF, signifying negligible phonon-drag effect in bilayer graphene. A deviation starts taking place for high T / TF along with a saturating tendency of the TEP. These features are not observed in monolayer graphene be- cause its TF is almost an order of magnitude higher than that of a bilayer. Since the crossover temperature in bilayer graphene is well within the experimentally accessible range, behavior of the TEP in a high-T limit can be conveniently investigated. In a high magnetic field, we also observed os- cillatory TEP and the Nernst signals, which are explained by the eightfold degeneracy of the zero-energy Landau level and the two dimensionality of the system, which yields the re- sults distinctive from monolayer graphene as well as graph- ite. ACKNOWLEDGMENTS Authors thank S. Kettemann and A. Hinz for private com- munications on the phonon-drag effect in graphene and criti- cal reading of the manuscript. We also thank S. Das Sarma, E. H. Hwang, and E. Rossi for valuable discussion. This work was supported by the National Research Foundation of Korea through Acceleration Research Grants No. R17-2008- 007-01001-0 and No. 2009-0083380. *Corresponding author; [email protected] 1 Y. M. Zuev, W. Chang, and P. Kim, Phys. Rev. Lett. 102, 096807 共2009兲. 2 P. Wei, W. Bao, Y. Pu, C. N. Lau, and J. Shi, Phys. Rev. Lett. 102, 166808 共2009兲. 3 J. G. Checkelsky and N. P. Ong, Phys. Rev. B 80, 081413共R兲共2009兲. 4 M. Cutler and N. F. Mott, Phys. Rev. 181, 1336 共1969兲. 5 E. H. Hwang, E. Rossi, and S. Das Sarma, Phys. Rev. B 80, 235415 共2009兲. 6 J. B. Oostinga, H. B. Heersche, X. Liu, A. F. Morpurgo, and L. M. K. Vandersypen, Nature Mater. 7, 151 共2007兲. 7 Y. Lin and P. Avouris, Nano Lett. 8, 2119 共2008兲. 8 B. E. Feldman, J. Martin, and A. Yacoby, Nat. Phys. 5, 889 共2009兲. 9 C. R. Wang, W. S. Lu, and W. L. Lee, Phys. Rev. B 82, 121406共R兲共2010兲. 10 S. S. Kubakaddi and K. S. Bhargavi, Phys. Rev. B 82, 155410 共2010兲. 11 S. Das Sarma, E. H. Hwang, and E. Rossi, Phys. Rev. B 81, 161407共R兲共2010兲. 12 S. Xiao, J. H. Chen, S. Adam, E. D. Williams, and M. S. Fuhrer, Phys. Rev. B 82, 041406共R兲共2010兲. 13 M. Monteverde, C. Ojeda-Aristizabal, R. Weil, K. Bennaceur, M. Ferrier, S. Guéron, C. Glattli, H. Bouchiat, J. N. Fuchs, and D. L. Maslov, Phys. Rev. Lett. 104, 126801 共2010兲. 14 W. Zhu, V. Perebeinos, M. Freitag, and P. Avouris, Phys. Rev. B 80, 235402 共2009兲. 15 Z. Zhu, H. Yang, B. Fauqué, Y. Kopelevich, and K. Behnia, Nat. Phys. 6, 26 共2010兲. 16 S. Ghosh, W. Bao, D. L. Nika, S. Subrina, E. P. Pokatilov, C. N. Lau, and A. A. Balandin, Nature Mater. 9, 555 共2010兲. 17 K. F. Mak, M. Y. Sfeir, J. A. Misewich, and T. F. Heinz, Proc. Natl. Acad. Sci. U.S.A. 107, 14999 共2010兲. 18 D.-K. Ki, D. Jeong, J.-H. Choi, H.-J. Lee, and K.-S. Park, Phys. Rev. B 78, 125409 共2008兲. 19 J. P. Small, K. M. Perez, and P. Kim, Phys. Rev. Lett. 91, 256801 共2003兲. 20 S. V. Morozov, K. S. Novoselov, M. I. Katsnelson, F. Schedin, D. C. Elias, J. A. Jaszczak, and A. K. Geim, Phys. Rev. Lett. 100, 016602 共2008兲. 21 S. Adam and S. Das Sarma, Phys. Rev. B 77, 115436 共2008兲. 22 E. H. Hwang and S. Das Sarma, Phys. Rev. B 82, 081409共R兲共2010兲. 23 J. Martin, N. Akerman, G. Ulbricht, T. Lohmann, J. H. Smet, K. von Klitzing, and A. Yacoby, Nat. Phys. 4, 144 共2008兲. 24 E. McCann and V. I. Fal’ko, Phys. Rev. Lett. 96, 086805 共2006兲. 25 A. B. Kuzmenko, E. van Heumen, D. van der Marel, P. Lerch, P. Blake, K. S. Novoselov, and A. K. Geim, Phys. Rev. B 79, 115441 共2009兲; Z. Q. Li, E. A. Henriksen, Z. Jiang, Z. Hao, M. C. Martin, P. Kim, H. L. Stormer, and D. N. Basov, Phys. Rev. Lett. 102, 037403 共2009兲. 245416-4 THERMOELECTRIC TRANSPORT OF MASSIVE DIRAC… PHYSICAL REVIEW B 82, 245416 共2010兲 26 E. V. Castro, K. S. Novoselov, S. V. Morozov, N. M. R. Peres, J. M. B. Lopes dos Santos, Johan Nilsson, F. Guinea, A. K. Geim, and A. H. Castro Neto, J. Phys.: Condens. Matter 22, 175503 共2010兲. 27 L. Hao and T. K. Lee, Phys. Rev. B 81, 165445 共2010兲. 28 C. Ayache, A. de Combarieu, and J. P. Jay-Gerin, Phys. Rev. B 21, 2462 共1980兲; K. Sugihara, Y. Hishiyama, and A. Ono, ibid. 34, 4298 共1986兲. 29 Y. J. Yu, Y. Zhao, S. Ryu, L. E. Brus, K. S. Kim, and P. Kim, Nano Lett. 9, 3430 共2009兲. 30 K. S. Novoselov, E. McCann, S. V. Morozov, V. I. Fal’ko, M. I. Katsnelson, U. Zeitler, D. Jiang, F. Schedin, and A. K. Geim, Nat. Phys. 2, 177 共2006兲. 31 M. Jonson and S. M. Girvin, Phys. Rev. B 29, 1939 共1984兲. 32 S. M. Girvin and M. Jonson, J. Phys. C 15, L1147 共1982兲. 245416-5
1509.08698	1	1509	2015-09-29T11:45:06	Resonant electron scattering by graphene antidot	[ "cond-mat.mes-hall" ]	The edge states which were observed on a linear edge of graphene may also persist on a curved edge. We calculate the elastic transport scattering cross section on a graphene nanohole supporting the edge states. Resonant peaks in the gate voltage dependence of conductivity of graphene with such nanoholes are obtained. Position and height of the resonances are determined by the localization depth of the quasibound edge states, and width -- by their lifetime. The scattering amplitude near the resonant energies has a strong valley asymmetry. We evaluate the effect of moderate edge rippling, inhomogeneity of boundary parameter along the edge, and Coulomb effects (charged nanohole) on the edge states and show that they do not affect the presence of the resonances, but can substantially influence their position, height and width. The local density of states near the nanohole also demonstrates a resonant dependence on gate voltage.	cond-mat.mes-hall	cond-mat	a Resonant electron scattering by graphene antidot I.V. Zagorodnev,1, 2, ∗ Zh.A. Devizorova,1, 2 and V.V. Enaldiev2 1Moscow Institute of Physics and Technology, Institutskiy per. 9, Dolgoprudny, Moscow Region, 141700 Russia 2Kotelnikov Institute of Radio-engineering and Electronics of the Russian Academy of Sciences, 11-7 Mokhovaya St, Moscow, 125009 Russia (Dated: August 8, 2021) The edge states which were observed on a linear edge of graphene may also persist on a curved edge. We calculate the elastic transport scattering cross section on a graphene nanohole supporting the edge states. Resonant peaks in the gate voltage dependence of conductivity of graphene with such nanoholes are obtained. Position and height of the resonances are determined by the localization depth of the quasibound edge states, and width -- by their lifetime. The scattering amplitude near the resonant energies has a strong valley asymmetry. We evaluate the effect of moderate edge rippling, inhomogeneity of boundary parameter along the edge, and Coulomb effects (charged nanohole) on the edge states and show that they do not affect the presence of the resonances, but can substantially influence their position, height and width. The local density of states near the nanohole also demonstrates a resonant dependence on gate voltage. I. INTRODUCTION Graphene with circular nanoholes, which are of- ten called antidots, may be used for microelectronic applications1 -- 4, as a metamaterial in the terahertz range5,6, or for the investigation of quantum coherence effects like the Aharonov-Bohm one7,8. Even impurities and defects in graphene can be treated as antidots with a very small radius9 -- 11. To investigate the electronic properties of perforated graphene structures one should first describe the edge of the sample that emerges due to perforation. Two types of graphene edges are often considered: zigzag and armchair11 -- 13. In the nearest-neighbor tight-binding ap- proximation there exists dispersionless edge states (ESs) at the ideal linear zigzag edge, and there are no ESs near the armchair one. However, it is quite challenging to control the edge orientation experimentally, and even if the edges are macroscopically smooth and oriented at some well-defined angles, they are not necessarily mi- croscopically ordered14. Nevertheless, the ESs were de- tected near the monoatomic steps on graphite surfaces15, in graphene-hexagonal boron nitride interface16, and, in graphene structures17,18. More comprehen- finally, sive investigations predict that zigzag ESs acquire a dispersion19, depend on chemical environment20, and that the ESs may exist even on the armchair edge21. Another actual problem for the nanohole is that the edge orientation changes upon go-round the hole. To avoid the difficulty, the nanohole is sometimes replaced by a hexagon with the zigzag or armchair edges22,23. Sim- ilar replacement is applied to a charged defect in contin- uous description within the framework of the two-band Dirac model10. In this model, electrons in a single valley of graphene are described by the Weyl-Dirac equation vσpψ = Eψ, (1) where v is the Fermi velocity, and ψ = (ψ1, ψ2)T is the two component wave function. The same equation may be used for the other graphene valley. To describe the edge of graphene, one should supple- ment the equation (1) by the boundary condition (BC). Usually one uses the zigzag-like BC10,11,13 ψ2 = 0 (or ψ1 = 0) or the Berry-Mondragon (infinite mass) BC24 -- 26. We consider a general BC to describe an edge which may be relaxed, reconstructed, disordered, inhomoge- neous, it also may have dangling bonds, band bending27 or impurities28. In the Dirac model such BC was dis- cussed in Refs. [29,30,31,32] and in the single-valley ap- proximation it has the following form19,33 (cid:0)ψ1 + iae−iϕψ2(cid:1)(cid:12)(cid:12)at edge = 0. (2) Here ϕ is the angle between the x-axes and the normal to the edge point, a is a phenomenological boundary pa- rameter characterizing the edge structure. The BC stems from the fundamental physical require- ments, namely, the absence of the normal component of the current through the edge and the time reversal sym- metry, and an additional assumption of the absence of intervalley interaction (for generalization see Appendix A). Therefore, it describes a general type of edge with- out specifying its microscopic structure. The universality is paid off by the unknown boundary parameter. Another way to describe the edge of graphene is to add to the Weyl-Dirac equation (1) the effective poten- tial V (r), which is a combination of electrostatic and staggered potentials, V (r) = m(r) 2 [(1 − a2)σ0 + (1 + a2)σz], (3) If the "mass" m(r) where m(r) = 0 in the graphene. outside of graphene is much greater than the character- istic energies of electrons in the graphene (width of the conduction and valence bands), i.e. m(r) → ∞, then the wave function almost does not penetrate the outside and the matching condition for the wave function on the boundary leads to the BC (2). The sign of the parameter a is determined by the sign of m(r) outside of graphene. The infinite mass BC, which is the same as the BC (2) with a = 1, was obtained in Ref. [24] by a similar proce- dure for V (r) containing only the staggered potential. It should be mentioned here that the reformulation of the BC in terms of the effective potential V (r) takes into ac- count only two bands, while the other bands may affect the BC significantly. When a 6= 1, the electron-hole symmetry is broken. This can be easily understood if one considers the hete- rocontact of two materials, both being described by the Dirac equation with different electron affinity and band gaps34. The value of the parameter a can be found from a comparison with experiments27,35 or from more rigor- ous calculations19. It is worth noting that for any edge parameter a, the effective potential V (r) necessarily in- volves the staggered potential. Bound states cannot be produced by a scalar potential, except the state with zero energy36 or the quasistationary (or quasibound) state near the overcritical Coulomb impurity10. It is the com- bination of the scalar and staggered potential that allows to describe the bound states. The edges of graphene can be divided into two groups: those supporting ESs (the parameter a varies smoothly and a 6= 1) and those which do not (a varies greatly or a = 1 in the case of constant a). We will consider the edges of the first type, with smoothly varying (along the edge) or constant parameter a 6= 1. Motivated by the recent progress in fabrication and characterization of antidot nanostructures22,35, we are aimed to show possible manifestations of the ESs, for ex- ample, in transport and scanning tunneling microscopy (STM) measurements of graphene with nanoholes. For this purpose, we calculate the transport cross-section of electrons on the nanohole in graphene and the local den- sity of states (LDOS) near the antidot. Considerable ef- forts were devoted to the conductivity of graphene with scatterers described by scalar potentials37 -- 41. A real (im- penetrable for carriers) hole cannot be described by a scalar potential, because electrons freely penetrate inside the scalar potential region. As mentioned above, the BC (2) necessarily contains the effective staggered potential. Scattering by staggered potentials only was considered, for example, in Refs. [1,4,42]. To the best of our knowl- edge, the combination of scalar and staggered potential has not been analyzed yet. We will show that it results in the existence of the quasistationary ESs localized near the nanohole. In turn, it leads to the peculiar peaks in the dependence of transport cross section and the LDOS on the electron energy and, therefore, in conductivity and tunnel current vs. gate voltage. The resonances in the transport cross-section correspond to the elastic scattering of electron by the quasistationary ESs. We will analyze the peaks and their robustness against inho- mogeneity of the parameter a and Coulomb effects. The paper is organized as follows. Firstly, we recall the spectrum of the ESs on a graphene half-plane (Sec. II) and discuss its robustness. Then, we analyze the qua- sistationary states on the graphene antidot (Sec. III). In 2 Sec. IV and V, we calculate the transport cross section and the LDOS respectively, which are our main results, represented in Figs. 2 and 3. In Sec. VI, we consider the scattering by a charged antidot. Conclusions are made in the final section VII. II. EDGE STATES ON GRAPHENE HALF-PLANE First, we recall the solutions of the Weyl-Dirac equa- tion (1) on the graphene half-plane x ≥ 0 supplemented by the BC (2) with a constant parameter a19,33. Momen- tum along the boundary ¯hky measured from the projec- tions of the valley centers on the edge (±¯hK0y) is a good quantum number. The bulk wave function is a sum of incident and reflected plane waves. The bulk spectrum is located in the energy domain E ≥ v¯hky, while the ES spectrum is outside this region. The ES band in the E(ky)-plane represents the rays starting from ±¯hK0y and described by the dispersion equations Es = s¯hv 2a 1 + a2 ky, sky(1 − a2) > 0. (4) s = ±1 is the valley index. The ES wave func- tion exponentially decay away from the edge ψ(x) ∼ e−sx(1−a2)ky/(1+a2). It is important that the ES band is "chiral" (not sym- metric about the center of a valley), but it is symmetric about the center of the edge Brillouin zone in agreement with the time reversal symmetry of spinless system. This fact will lead to important consequences for scattering by the antidot. Inclusion of the intervalley interaction at the edge leads to qualitatively the same results, see Appendix A. Be- sides, for a wide class of the translationally invariant edges, for example zigzag one, the distance between the projections of the valley centers 2K0y is large in com- parison with the electron momentum, hence, the edge intervalley interaction is negligible19. For large electron energy, the edge intervalley interaction is important and results in connection between the rays of the ESs from the different valleys, see Fig. 1b. Moderate bending of graphene sheet near the edge, which probably took place in Ref. [18], can be included by means of renormalization of the parameter a, see Ap- pendix B. The feature of the graphene band structure is that the ESs always coexist with the bulk states. It means that they must be quasistationary with a finite lifetime, due to the probability of decay into the bulk. The edge rough- ness is one of the possible reasons of the decay. For small wave vectors, we evaluate this probability using the Fer- mis Golden Rule as (Appendix C) w ∝ k4 y, (5) which implies that for small wave vectors the ESs near a linear edge are quasistationary. Besides, it is expected that any deviation from linearity of the edge results in finite life time of the ESs. In the next section we will directly show it for a circular hole. III. QUASISTATIONARY EDGE STATE AT GRAPHENE NANOHOLE We consider the ESs at a circular hole, depicted schematically in Fig. 1c. The total angular momentum j = l + 1/2 = ±1/2,±3/2... is conserved (we will call l orbital angular momentum), therefore, in the polar co- ordinate system (r, ϕ) the wave function ψ ∝ exp(ijϕ). From here, without loss of generality, we will consider negative electron energy E < 0. Introducing the wave vector k = −E/(¯hv), one can obtain the Bessel equation for each component of the wave function. For example, the equation for the first component is (6) r2ψ′′ 1 (r) + rψ′ 1(r) +(cid:2)(kr)2 − l2(cid:3) ψ1 = 0. The solutions corresponding to the bulk states are merely the Bessel functions or their combinations. There are no localized stationary states. However, the quasistationary states can exist. To show this, we consider complex en- ergy E = E′ + iE′′ assuming E′′ ≪ E′, and E′′ > 0. The imaginary part of the energy E′′ determines the life- time of the quasistationary state with energy E′. Since the wave function of the state is an outgoing wave, we choose it in the following form ψ2(r, ϕ) (cid:19) = C (cid:18) ψ1(r, ϕ) eilϕH (2) l −iei(l+1)ϕH (2) (kr) l+1(kr) ! , (7) l where H (2) (x) is the Hankel function of the second kind, and C is the normalization constant. For positive energy E′ > 0, the Hankel function of the first kind must be used. Substituting the wave function (7) into the BC (2), we obtain the dispersion equation 3 FIG. 1. (a) Sketch of the graphene half-plane (b) Energy spectrum E(ky) of the graphene half-plane. Here ky is the momentum along the edge measured from the center of the edge Brillouin zone. The shaded region corresponds to the bulk states, the rays starting in the valley centers -- to the edge states. In the absence of intervalley interaction, the rays are infinite while it inclusion results in the "interaction" of the rays near the Brilluoin zone center shown by the dashed line. (c) Sketch of the graphene antidot (d) Energy spectrum of the quasistationary edge states localized near the antidot (only the real part of energy is shown). The sign of angular momentum j is coupled to the valley index. The energy of the "ground" ES with l = 0 is deter- mined by the transcendental equation a = −kR ln kR 2 . (10) However, the ground state has a long lifetime only for kR ≪ 0.1, otherwise, E′′ approximately equals E′. We will directly show it in the next section. The spectrum of the ESs including both graphene valleys is shown in Fig. 1d. The ESs can be detected in measurements of conduc- tivity. One of the ways to take into account the contribu- tion of the ESs is to calculate the transport cross section and the LDOS, which will be our foregoing goal. H (2) l (kR) = −aH (2) l+1(kR). (8) IV. SCATTERING BY NEUTRAL ANTIDOT Expanding the Hankel functions into the Laurent series about kR = 0, for l 6= 0 we find the low energy spectrum (la ≪ 1/2) of the qusistationary ESs: 2πa(la)2l [(l − 1)!]2 , kR ≈ −2sla + i sla < 0. (9) The equation describes the electron spectrum for both valleys. The spectrum of the ESs in each valley is equidis- tant. We stress that the ESs are chiral again, as the sign of the ES total angular momentum j depends on the val- ley index s, see second part of the equation (9). It is positive in one valley, and negative in the other one. For small parameter a, the real part of the energy can be obtained quasiclassically from (4) by replacing the mo- mentum ky with l/R. We first consider the scattering of a plane wave by the nanohole. We represent the wave function as a combina- tion of the normalized plane wave propagating in the x direction and the scattered cylindrical waves ψscat =(cid:18) 1 1(cid:19) e−ikx √2 + ∞ Xl=−∞ Cl eilϕ √2 H (2) −ieiϕH (2) (kr) l+1(kr) ! . l In the case of E > 0 one must choose the H (1) function. l The behaviour of the scattered wave function at large distances determines the scattering amplitude f (ϕ) =r 2 πk ∞ Xl=−∞ Cl exp(cid:18)ilϕ + i lπ 2 + i π 4(cid:19) . (12) (11) (x) Hankel kR=0.1 kR=0.26 kR=0.57 kR FIG. 2. Dependence of the transport cross section σtr on the electron wave vector k = E/¯hv for scattering by the nanohole with raidus R and the constant boundary parameter a = −0.15 (red solid line) and a(ϕ) = −0.15±0.04 cos ϕ (black dashed line). The resonances emerge due to the scattering by the qusistationary edge states. Inset: (left) Polar plot of the scattering amplitude in a valley for a = −0.15 and different energies. The energies are highlighted on the main figure by the vertical dashed lines. An electron is incident on the antidot from the left. (right) Total scattering amplitude, taking into account the contribution of two valleys. 4 reason of this asymmetry is the lack of the time reversal symmetry in one valley, which was mentioned in Secs. II and III. The ES in a given valley has positive j and rotates counterclockwise, in the other one it has negative j and rotates clockwise. This chirality of the ES with respect to the valley index leads to the asymmetry of the scattering amplitude. It somehow resembles the classical Magnus effect and this skew-symmetric scattering must lead, in turn, to the valley Hall effect. Moreover, this valley Hall effect should be resonant, because the mag- nitude of the asymmetry strongly depends on the energy and has maximum value in the vicinity of the ES en- ergy. At the same time, the total scattering amplitude including the contributions from both valleys is symmet- ric, because the time reversal symmetry is restored, see inset in Fig. 2. It is also important to note, that the backscattering is suppressed for the scattering by scalar potentials. How- ever, the antidot is described by the combination of scalar and staggered potentials thus allowing the backscatter- ing. To show a possible manifestation of the ESs in the resistivity, we calculate the transport cross section on the antidot σtr = 2π Z0 (1 − cos ϕ)f (ϕ)2 dϕ = 4 k ∞ Xl=−∞(cid:0)Cl2 − Im(ClC∗ l+1)(cid:1) . (14) The energy dependence of the transport cross section is shown in Fig. 2 and inset in Fig. 3. The resonant peaks in these figures emerge because of the resonant electron scattering by the quasistationary ESs and, therofore, are almost equidistant in accordance with the ESs spectrum obtained in the previous section. Let us start analysis of the energy dependence of the cross section with the low energy scattering kR ≪ 2a and for small parameter a ≪ 1. In this case, the scatter- ing is symmetrical (s-scattering) and it is the ground ES that mainly contributes to the scattering cross section. Expanding the Bessel functions, we find C0 ≈ −(−i)l 1 + i 2 1 kR − γ − ln(cid:0) kR π (cid:2) a 2 (cid:1)(cid:3) , (15) where γ is the Euler-Mascheroni constant. The cross section has the form 4π2kR2 σtr ≈ π2k2R2 + 4(cid:2)kR ln(cid:0) kR 2 (cid:1) − a(cid:3)2 . (16) Formula (16) resembles equation (21) of Ref. a general length" R/a. [9] for low energy scattering with the "scattering The cross section (16) tends to zero for small energies if a 6= 0. To clarify this, one can use the scattering phase δ1/2, which is proportional to k for small energies as in a general theory of non-relativistic scattering43. Then the elastic cross section σel ∝ 1 − e2iδ1/22/k ∝ k, which FIG. 3. Dependence of the LDOS at different distances r from the center of the circular hole with radius R on electron energy (k = E/¯hv) for the constant boundary parameter a = −0.3. Resonant peaks are clearly visible on the free LDOS trend 2k/(π¯hv). Inset: The transport cross section σtr vs. electron energy for the same value of the parameter a. Expanding the incident plane wave into the series of cylindrical waves and substituting the full wave function into the BC (2), we obtain Cl = −(−i)l Jl(kR) + aJl+1(kR) l+1(kR) (kR) + aH (2) H (2) l , (13) where Jl(x) is the Bessel function of the first kind. It follows from (13) that Cl+1 6= C−l. Therefore, intro- ducing the scattering phase e2iδj = 1+2ilCl (j = l+1/2), one can show that δj 6= δ−j. This results in the asym- metry of the scattering amplitude with respect to the re- placement ϕ → −ϕ, i.e f (ϕ) 6= f (−ϕ) in one valley. The implies that inelastic scattering dominates for small k. The inelastic scattering phases are complex, and one should expect that δ1/2 still will be proportional to k with some complex coefficient43. The total cross section which includes elastic and inelastic channels will be finite, σtot ∝ [1 − Re(e2iδ1/2 )]/k ≈ const. The energy of the first cross section maximum approx- imately corresponds to the ground ES energy (10), the height of the peak is σtr = 4/k0, the full width at half maximum ∆ = πa/[π2 + 4 ln2(k0/2)], where k0 is the root of the equation (10). It follows from these equations that the ground state with l = 0 has long lifetime only if kR ≪ 2 exp(−π) ≈ 0.1. To analyze other resonances in the transport cross- section we consider the energies in the vicinity of the qusistationary ES with an orbital angular momentum l0, i.e. kR ≈ 2al0 ≪ 1. Then, it is the coefficient Cl0 that has a resonant energy dependence Cl0 ≈ − 1 − il0!(l0−1)! 2π (kR + 2al0) ≈ −1, 1 kR(cid:1)2l0+1 (cid:0) 2 (17) while other Cl's behave smoothly. The height of the peak (measured from the background determined by other Cl's) approximately equals 2R/al0. Thus, for suffi- ciently small al0, it can be several times larger than 2R, the geometrical cross section of the hole. The peak has the Lorentzian shape, while its width is determined by the lifetime of the ES. Let us now estimate the contribution of the scattering by antidots to the net resistivity of a sample, using the parameters from experiments, Refs. [35,44]: the Fermi energy EF ≈ 20 meV≫ kT , the concentration of the antidots N ≈ 1010 cm−2. We use the Drude formula (if the Fermi energy is sufficiently far away from the Dirac point) with the scattering time τ = 1/(N σtrv) and the transport cross section σtr ≈ 2R ≈ 20 nm and obtain the 2D resistivity ρ ≈ 2RN · πv¯h2 EF e2 ∼ 102 Ohm, (18) which is a measurable value. Since the height of the transport cross section resonances can greatly exceed 2R, the resonant resistivity can be much higher than 100 Ohm and the ESs can be detected. Such measurements allow to estimate the characteristics of the ESs and the value of the boundary parameter. Thus far we considered constant boundary parameter a, though it may depend on the orientation of the edge which varies upon go-round the circular nanohole. There- fore, now we simulate this situation consider inhomogene- ity of the parameter a along the edge, i.e. a = a(ϕ). Expanding it in the Fourier series a(ϕ) = P aneinϕ and substituting the wave function (11) into the BC (2), we 5 find Xn anh(−i)l−nJl−n+1(kR) + H (2) l−n+1(kR)Cl−ni + ∀l. + H (2) l (kR)Cl + (−i)lJl(kR) = 0, It is merely an infinite matrix equation on the coeffi- cients Cl. In the simplest case of harmonically modu- lated parameter a(ϕ) = a0 + 2a1 cos ϕ (a1 ≪ a0) one can cut off the resulting tridiagonal matrix for some suf- ficiently large l and solve the reduced system. Numerical results for a0 = −0.15 and a1 = 0.04, presented in Fig. 2, demonstrate a significant change in the position and magnitude of the resonances away from the Dirac point and, at the same time, the robustness of the resonance effect against moderate inhomogeneity of the boundary parameter a. V. LOCAL DENSITY OF STATES NEAR NANOHOLE We have shown that the antidot ESs manifest them- selves in the energy dependence of the transport cross- section, and, therefore can be detected by measuring the conductivity of graphene sample with such antidots. However, high identity of the antidots is required for such an experiment and other types of scatterers could mask the effect. Another possible way to detect the ESs free of these drawbacks is to measure the LDOS near a hole. To this end, we calculate the LDOS near the nanohole. In the following calculations, we impose non restrictions on the partial (with total angular momentum j) wave function at the infinity ψk,l = Al(k)eilϕ(cid:18) Jl(kr) + Bl(k)Yl(kr) −ieiϕ [Jl+1(kr) + Bl(k)Yl+1(kr)] (cid:19) , (19) where Yl(x) is the Bessel function of the second kind, Al(k) is the normalization coefficient, and Bl(k) is ob- tained by substituting the wave function into the BC (2): Bl(k) = − aJl+1(kR) + Jl(kR) Yl(kR) + aYl+1(kR) . (20) for ∞ To find Al(k), we use eigenfunctions of a continuous the normalization rule the spectrum ψ+ k′,j ′ (r, ϕ)ψk,j (r, ϕ)d2r = gδ(k − k′)δjj ′ , where g is di- mensional parameter determined from the free (without a hole) DOS per unit area. Then, the LDOS including spin and valley degeneracy is RR ρ(k, r) = k π¯hv ∞ 1 1 + Bl(k)2 hJl(kr) + Bl(k)Yl(kr)2 Xl=−∞ + Jl+1(kr) + Bl(k)Yl+1(kr)2i . (21) The energy dependence of the LDOS is shown in Fig. 3 for a = −0.3. Again, it demonstrates the resonances for the energies in the vicinity of the ES energies. If kR tends to zero, then all coefficients Cl → 0, and we obtain the free DOS per unit area, ρf ree(k) = 2k/(π¯hv). The spatial dependence of the LDOS for any energy demonstrates a power low decrease at small distances and the Friedel oscillations far away from the nanohole with- out any resonant peaks, however, the magnitude of the LDOS near the nanohole significantly (resonantly) de- pends on energy. For this reason, we analyze below only the energy dependence of the LDOS near the nanohole. For energies below the ground state energy, k < k0, for a ≪ 1 and r ∼ R, the main contribution to the LDOS comes from the terms with l = 0,−1 (j = ±1/2) ρ(k, r) ≃ ρf ree(k)(cid:18)1 + R2 2a2r2(cid:19) . (22) It is considerably greater then the free DOS, increases with decreasing of the parameter a and diverges when a → 0, which is the typical for the zigzag case. For the linear "reczag" edge, the LDOS demonstrates qualita- tively the same behaviour19. If the energy is close to the ground state energy k ≃ k0, then the height of corresponding peak is ρ(k0, r) ≃ 4 π3¯hvk0r2 . (23) In the vicinity of the ES energy with l0, kR ≃ 2al0 ≪ 1, the coefficient Bl0 is resonantly large, while the other Bl's are small. Then, the height of the l0th peak in the LDOS is 2 π3¯hvR ρ ≃ l0!2 (al0)2l0+1 (cid:18) R r(cid:19)2l0+2 . (24) According to Eqs. (23, 24) the height of all resonances decreases while moving away from the nanohole with a power law profile. The greater the resonances number, l0's, the stronger is the spatial decay of the LDOS. For a fixed distance from the antidot each subsequent en- ergy peak is higher than the previous one. This behavior of resonances in the LDOS obtained in low energy limit kR ≪ 1 is in a qualitative agreement with the results for wider energy range represented in Fig. 3. VI. SCATTERING BY CHARGED ANTIDOT In this section we consider scattering by a charged an- tidot. This charge could appear naturally, for example, in graphene sample deposited onto a substrate which pos- sesses a large number of charged impurities45, or artifi- cially by deposition of dimers46. The antidot may also become charged while electrons occupy the ESs. Our aim is to calculate the renormalization of the ESs energy spectrum and modification of the transport cross section due to the Coulomb effects. 6 The presence of the charge Q in the center of the an- tidot leads to the presence of an extra Coulomb term −eQ/r in the Weyl-Dirac equation (1). The edge of the antidot is still described by the BC (2) with the constant parameter a. The spectrum of quasistationary ESs is determined by the vanishing of incoming wave. Introducing dimension- less charge q = eQ/¯hv, in the low-energy limit (kR ≪ 1) and under conditions q ≪ 1, la < 0 for l 6= 0 we obtain the spectrum l q+ kR ≃ −2al + l + 1/2 + iΓ(l + 1/2 − iq)2 8lΓ2(2l) e−πq(cid:18)−4al + 2l l + 1/2 q(cid:19)2l+1 . (25) For l = 0, the real part of k is determined by kR ≃ k0R − q 1 + a/k0R , (26) where k0 is the solution of the equation (10). The details of the calculations are presented in Appendix D. According to Eqs. (25, 26), the wave vector k of quasis- tationary ESs decreases (increases) for negatively (posi- tively) charged antidot when the absolute value of charge increases. At the same time, the energy of quasistation- ary ESs increases (decreases), because E = −¯hvk. This result is intuitively clear, because negatively (positively) charged antidot repels (attracts) electons, and, therefore, the energy of the stationary states increases (decreases). Similarly to the case of uncharged antidot, peculiar peaks corresponding to the resonant scattering by qua- sistationary ESs appear in the energy dependence of the transport cross section, Fig. 4. For negatively (pos- itively) charged antidot these peaks shift to the left (right), when the absolute value of charge increases, in agreement with the renormalization of ESs spectrum dis- cussed above. The width of the peaks determined by the imaginary part of the energy is also in qualitative agree- ment with (25): at negative q the imaginary part of the energy decreases when q increases and the peaks become narrower, which corresponds to more stationary states; at positive q, the imaginary part of the energy increases when q increases, and the peaks become wider. VII. CONCLUSION We have described the electronic properties of graphene with impenetrable (for carriers) nanoholes sup- porting the localized (edge) states. The energies of these states are almost equidistant. We have demonstrated a strong asymmetry of the scattering amplitude near the resonances in a given valley. It must lead to the resonant valley Hall effect. The resonant scattering of graphene electrons occurs when its energy coincides with the edge R r t Σ , n o i t c e s - s s o r c t r o p s n a r T R r t Σ , n o i t c e s - s s o r c t r o p s n a r T 14 12 10 8 6 4 35 30 25 20 15 10 5 0 0.4 0.6 0.8 1.0 1.2 kR q=-0.1 q=-0.15 q=-0.2 400 300 200 100 0 0.00 0.01 0.02 0.03 0.04 0.05 0.1 0.2 0.3 0.4 kR 0.5 0.6 0.7 0.8 FIG. 4. Dependence of the transport cross section σtr on en- ergy for the scattering by positively (upper panel, q > 0) and negatively (lower panel, q < 0) charged antidot and a = −0.15. When q increases, the resonant peaks, which position is determined by the ESs spectrum, shift to the right for positively charged antidot, and to the left if charge is neg- ative in agreement with Eqs. (25,26). The width of the peaks depends on the ESs lifetime and decreases when the antidot is negatively charged. state energy. Moderate edge rippling, inhomogeneity of the boundary parameter and charge of the nanohole gen- erally do not influence the very presence of the effect, but can substantially influence the position of the res- onant energies as well as the width and height of the resonances. One way to detect these resonanses is to measure the conductivity of graphene sample with an array of identi- cal nanoholes, another one is to measure the LDOS near an antidot. We have calculated the LDOS and shown that its dependence on the electron energy near the hole also demonstrates resonances emerging due to the edge states. The intervalley interaction at the curved edge of the nanohole and strong inhomogeneity of the edge param- eter a (or parameters if edge intervalley interaction is included) are the subject of the further studies ACKNOWLEDGMENT 7 0.05 0.10 0.15 0.20 0.25 We are grateful to Prof. V.A. Volkov for the fruitful discussions. This work was supported by the Russian Foundation for Basic Research (projects 14-02-31592 and 14-02-01166), and the Dynasty Foundation. q=0.1 q=0.15 q=0.2 50 40 30 20 10 0 0.00 Appendix A: Edge intervalley interaction on graphene half-plane Edge intervalley interaction mixes the wave functions from different valleys in BC (ψ + igψ′)at edge = 0. (A1) Here ψ and ψ′ are two component wave functions from the two valleys. The intervalley distance should be ex- plicitly included either in the BC or in the bulk (Weyl- Dirac) equation. We choose the second way and consider modified Hamiltonian, describing a valley, H = σ(p − ¯hK0) (A2) Then the requirement of absence of normal to the edge current (hermiticity) and the time reversal symmetry30 restricts the matrix g as eiφ g = sin β (cid:0)eiγ~σ~n + σ3 cos β(cid:1) , (A3) where γ, β, φ are real phenomenological parameters, which can be associated with the parameters from Ref. [30]. The parameter γ is connected with the parameter a as sin γ = 2a/(1 + a2). Solving the Schrodinger equation with Hamiltonian (A2) supplemented by the BC (A1) for the graphene half-plane, we obtain the electronic spectrum, see Fig. 1b. For any β and near the projections of the valley cen- ters, ky + sK0y ≪ K0y, the ESs spectrum is expressed as Es = s¯hv sin θ(ky + sK0y), s(ky + sK0y) cos θ < 0. (A4) Here cot θ/2 = (cos β − cos γ)/ sin γ. Thus, "rays" of the ESs start in the ±K0y points even if intervalley in- teraction is taken into account. Near the center of the Brillioun zone the rays are transform to each other. Appendix B: Rippling near edge It is well known that ripples in graphene can be treated by means of the effective electrical potential47 xh)2 + (∂2 y h)2(cid:3) and the vector-potential y h)2(cid:3), Ay = x,yh, κ1 ≈ 40eV·A2, κ2 ≈ 6.2 eV·A2. Position of a graphene list in three-dimensional space are Φ(r) = −κ1(cid:2)(∂2 A = (Ax, Ay), where Ax = −κ2(cid:2)(∂2 y h(cid:1) ∂2 2κ2(cid:0)∂2 xh)2 − (∂2 xh + ∂2 8 x′ = x − δx(y), y′ = y the Hamiltonian (1) with the BC (ψ1 + iaψ2)x=δx(y) = 0 can be transformed to the stan- dart BC (ψ1 + iaψ2)x=0 = 0 (dashes are skipped for brevity) and the same Hamiltonian with additional part δ H = −v ∂δx(y) ∂y σy px. (C1) The probability of transition from ES with momentum along the edge ¯hky, wave function ψs and energy Es (for definitness we consider the valey s = +1) to bulk state with wave function ψb and energy Eb = ±v¯hk′ due to the edge roughness can be evaluated using the Fermi golden- rule: dw = 2π ¯h hψsδ Hψbi2δ(Eb − Es), where ψs =s 2a2(1 − a2) Ly(1 + a2)2 ky(cid:18) 1 p2LxLy (cid:18) 1 ia−1 (cid:19) e− 1−a2 ±eiφk′ (cid:19) eik′ ψb = 1 1+a2 ky xeiky y, xxeik′ y y. (C2) (C3) (C4) For simplicity, here we neglect second bulk solution xx, because it doesn't sugnificantly affect on final ∼ e−ik′ result. Integrating over k′ we obtain the probability of ES decay for this realization of the boundary w = 2v πLy a21 − a23k2 (1 + a2)3sgn[a]× y Z 2π 0 dφ(cid:18) 1 + a2 2a where + 2sgn[a] sin φ(cid:19) F (ky − 2a 1+a2 ky sin φ)2 (1 − a2)2 + 4a2 cos2 φ , (C5) determined by the function h(x, y) that determines dis- tance of the list from the plane z = 0. In this section we study how smooth ripples affect on the ESs spectrum. First, we consider half-plane x ≥ 0 when the graphene list is bended in perpendicular to the edge direction (i.e. ∂yh = 0). Therefore ky is a good quantum number. The sfystem of equations for the electron wave function in a valley reads as follows (cid:26) −κ1hψ1 + (−i∂x − iky)ψ2 − κ2hψ2 = Eψ1 (−i∂x + ipy)ψ1 − κ2hψ1 − κ1hψ2 = Eψ2 , (B1) where h = (∂2 xh)2. After multiplying the first equation by ψ1, the second one by ψ2 and subtracting them we get the differential equation for the function η = ψ1/ψ2: x − 2kyη − i(κ1h + E)(1 − η2) = 0. η′ We solve the equation with the BC (2) ηx=0 + ia = 0. (B2) (B3) We will perturbatevly treat h. Hence we look for the wave function by means of series in powers of h, η = η0 + η1 + . . . , where η1 ∝ h etc. In the zero order η0 = −i(ky +qk2 η1 = −iκ1(1−η2 y − E2)/E, in the first order we get 0)Z +∞ h(x′) expn−2qk2 (B4) Substitution of the first-order wave function in the BC (B3) results in a dispersion equation with the ES spec- trum x y − E2(x′ − x)o dx′. E = 2a0ky 1+a2 0 + 2κ1ky (1 − a2 0)ky ≥ 0. h(x′)e−2ky(1−a2 0)x′ dx′, 1−a2 0 1+a2 0 0 R +∞ (B5) Thus, small edge rippling leads to the renormalization of the parameter a. We can obtain spectrum of the ESs in the antidot geometry via quasiclassical quantization of parallel momentum ky = l/R. F (k) =Z ∞ −∞ eiky ∂δx(y) ∂y dy. (C6) Then we average w over all possible realizations of the boundary using following relation Appendix C: Life time of edge states on half-plane with rough boundary hF (k)2i = lc∆2Lyk2e−l2k2/4. (C7) Finally, we obtain Let us consider a rough linear edge that can be de- scribed by the BC (2) determined on the curve x = δx(y), where δx(y) is a random deviation of the edge from its average position hδxi = 0 in the point y, Ref. [48]. The edge roughness is assumed to be smooth and δx(y) has the Gaussian correlation function hδx(y1)δx(y2)i = ∆2 exp((y1 − y2)2/l2 c ). We also assume electron wave- length to be much greater than the correlation length lc. Firstly, we consider some definite realization of the boundary x = δx(y). Using coordinates transformation 0 k2 2vlc∆2a21 − a23 π(1 + a2)3sgn[a] hwi = (ky − 2a 1+a2 ky sin φ)2 (1 − a2)2 + 4a2 cos2 φ dφ"(cid:18) 1 + a2 1 + a2 ky sin φ(cid:19)2(cid:19)#, 2a In the limit a ≪ 1 the probability of decay (C8) has yZ 2π 4 (cid:18)ky − exp(cid:18)− + 2sgn[a] sin φ(cid:19) (C8) 2a l2 c simple form: hwi = vlc∆2ak4 ye−l2 ck2 y/4 (C9) Appendix D: Scattering by charged antidot where s =q(l + 1 2 )2 + q2. The functions φ(r) and χ(r) are determined by the In the polar coordinates the Weyl-Dirac equation with following expressions: 9 Coulomb potential can be represented as: 0 −ieϕ∂r + eiϕ r ∂ϕ −ie−ϕ∂r − e−iϕ r ∂ϕ 0 ! ψ =(cid:16)−k + q r(cid:17) ψ. (D1) Following Refs. [38,10], we seek the solution with total angular momentum j = l + 1/2 for energy E < 0 in the following form: χ(r) = AM (s+iq, 2s+1,−2ikr)+BU (s+iq, 2s+1,−2ikr), (D3) φ(r) = A(s + iq) (l + 1 2 ) 1 − B(cid:18)l + M (s + iq + 1, 2s + 1,−2ikr)− 2(cid:19) U (s + iq + 1, 2s + 1,−2ikr), (D4) ψ(l)(r, ϕ) =(cid:18) φ(r) + χ(r) (φ(r) − χ(r))eiϕ (cid:19) rs− 1 2 eilϕeikr, where M (a, b, z) and U (a, b, z) are the Confluent hyperge- ometric function of the first and second kind respectively. The coefficients A and B are related by the BC: (D2) B A = (1 − ia) s+iq (1 + ia)(l + 1 l+ 1 2 M (s + iq + 1, 2s + 1,−2ikR0) + (1 − ia)M (s + iq, 2s + 1,−2ikR0) 2 )U (s + iq, 2s + 1,−2ikR0) − (1 − ia)U (s + iq, 2s + 1,−2ikR0) . (D5) To find the renormalization of the quasistationary ES spectrum, we eliminate the incoming wave and obtain the dispersion equation manner: χ φr→∞ = e2ikr+2iβln(2kr)−πi(l+ 1 2 )+2iδl(k). (D8) (cid:20)(1 + ia) s + iq l + 1/2 M (s + iq + 1, 2s + 1, z)+ + (1 − ia)M (s + iq, 2s + 1, z)(cid:21)+ +(cid:20)(1 + ia)(l + 1/2)U (s + iq + 1, 2s + 1, z)− − (1 − ia)U (s + iq, 2s + 1, z)(cid:21) Γ(2s + 1)eiπ(−s−iq) Γ(s − iq + 1) = 0. In the low-energy limit (kR ≪ 1) and under condition q ≪ 1 for l 6= 0 we obtain (25). Now we turn to the solution of scattering problem. The transport cross section can be expressed in terms of the Coulomb scattering phases δl, σtr = 2 k ∞ Xl=−∞ sin2(δl − δl+1), (D7) To find the Coulomb scattering phases we write our solution for χ(r) and φ(r) in the limit r → ∞: χ(r) ∼ A Γ(2s + 1) Γ(s − iq + 1) + Be−i π ei π 2 (s+iq)(2kr)−se−iqln(2kr)+ 2 (s+iq)(2kr)−se−iqln(2kr), (D9) (D6) φ(r) ∼ A(s + iq)Γ(2s + 1) (l + 1 2 )Γ(s + iq + 1) ei π 2 (s−iq)−2ikr+iqln(2kr)(2kr)−s. (D10) Comparing (D8) with (D9) and (D10) we find e2iδl(k) = eiπ((l+ 1 2 )−s)+ l + 1 2 s + iq + B A Γ(s + iq + 1) Γ(s − iq + 1) l + 1 2 s + iq Γ(2s + 1) Γ(s + iq + 1) eiπ((l+ 1 2 )+iq) (D11) The results for the transport cross section are pre- where the scattering phases is defined in conventional sented in Fig. 4. ∗ [email protected] 1 M. R. Thomsen, S. J. Brun and T. G. Pedersen, J. Phys.: (2013). 3 C.-H. Park, L. Yang, Y.-W. Son, M.L. Cohen, S.G. Louie, Condens. Matter 26, 335301 (2014). Nat. Phys. 4, 213 (2008). 2 M. Dvorak, W. Oswald, Z. Wu, Scientific Reports 3, 2289 4 S.R. Power and A.-P. Jauho, Phys. Rev. B 90, 115408 10 (2014). 5 A. Yu. Nikitin, F. Guinea, L. Martin-Moreno, Appl. Phys. Lett. 101, 151119 (2012). 6 D. Svintsov, T. Otsuji, V. Mitin, M. S. Shur and V. Ryzhii, Appl. Phys. Lett. 106, 113501 (2015). 7 T. Shen, Y. Q. Wu, M. A. Capano, L. P. Rokhinson, L. W. Engel and P. D. Ye, Appl. Phys. Lett. 93, 122102 (2008). 8 S. Russo, J.B. Oostinga, D. Wehenkel, H.B. Heersche, S.S. Sobhani, L.M.K. Vandersypen, A.F. Morpurgo, Phys. Rev. B 77, 085413 (2008). 9 D. M. Basko Phys. Rev. B 78, 115432 (2008). 10 A. V. Shytov, M. I. Katsnelson, and L. S. Levitov, Phys. Rev. Lett. 99, 236801 (2007). 11 V. N. Kotov, B. Uchoa, V. M. Pereira, F. Guinea, and A. arXiv:1407.0615 (accepted to Theoretical and Mathemati- cal Physics, 2015). 27 M.T. Allen, O. Shtanko, I.C. Fulga, A. Akhmerov, K. Watanabi, T. Taniguchi, P. Jarillo-Herrero, L.S. Levitov, A. Yacoby arXiv:1504.07630 (2015). 28 M. Wimmer, A. R. Akhmerov, F. Guinea, Phys. Rev. B 82, 045409 (2010). 29 E. McCann and V. I. Fal'ko, J. Phys.: Condens. Matter 16, 2371 (2004). 30 A. R. Akhmerov and C. W. J. Beenakker, Phys. Rev. B 77, 085423 (2008). 31 D. M. Basko, Phys. Rev. B 79, 205428 (2009). 32 G. Tkachov, Phys. Rev. B 79, 045429 (2009). 33 V. A. Volkov and I. V. Zagorodnev, Low Temp. Phys. 35, H. Castro Neto, Rev. Mod. Phys. 84, 1067 (2012). 2 (2009). 12 K. Nakada, M. Fujita, G. Dresselhaus, M. S. Dresselhaus, 34 B.A. Volkov, B.G. Idlis, M.Sh. Usmanov, PhysicsUspekhi Phys. Rev. B 54, 17954 (1996). 13 C. W. J. Beenakker, Rev. Mod. Phys. 80, 1337 (2008). 14 C. Casiraghi, A.Hartshu, H.Qian, S.Pisanec, C.Georgi, K. S. Novoselov, D. M. Basko, A. C. Ferrari, Nano Lett. 9, 1433 (2009). 15 Y. Niimi, T. Matsui, H. Kambara, K. Tagami, M. Tsukada, and Hiroshi Fukuyama Phys. Rev. B 73, 085421 (2006). 16 J. Park, J. Lee, L. Liu, K.W. Clark, C. Durand, C Park, B.G. Sumpter, A.P. Baddorf, A. Mohsin, M. Yoon, G. Gu, A.-P. Li, Nature Communications 5, 5403 (2014). 17 K.A. Ritter, J.W. Lyding, Nature Materials 8, 235 (2009). 18 C. Tao, L. Jiao, O.V. Yazyev, Y.-C. Chen, J. Feng, X. Zhang, R.B. Capaz, J.M. Tour, A. Zettl, S.G. Louie, H. Dai, M.F. Crommie, Nat. Phys. 7, 616 (2011). 19 J.A.M. van Ostaay, A.R. Akhmerov, C.W.J. Beenakker, M. Wimmer, Phys. Rev. B 84, 195434 (2011). 20 S. Fujii, M. Ziatdinov, M. Ohtsuka, K. Kusakabe, M. Kiguchia and Toshiaki Enoki, Farad. Discuss. 173, 173 (2014). 21 P.A. Maksimov, A.V. Rozhkov, A.O. Sboychakov, Phys. Rev. B 88, 245421 (2013). 22 F. Oberhuber, S. Blien, S. Heydrich, F. Yaghobian, T. Korn, C. Schuller, C. Strunk, D. Weiss, and J. Eroms, Appl. Phys. Lett. 103, 143111 (2013). 23 S.J. Brun, M. R. Thomsen, T. G. Pedersen, J. Phys.: Con- 38, 761 (1995). 35 Yu.I. Latyshev, A.P. Orlov, V.A. Volkov, V.V. Enaldiev I.V. Zagorodnev, O.F. Vyvenko, Yu.V. Petrov, P. Mon- ceau, Scientific Reports 4, 7578 (2014). 36 C.A. Downing, A.R. Pearce, R.J. Churchill, M. E. Portnoi arXiv:1503.08200 (2015). 37 S. Das Sarma, S. Adam, E.H. Hwang, and E. Rossi, Rev. Mod. Phys. 83, 407 (2011). 38 D.S. Novikov, Phys. Rev. B 76, 245435 (2007). 39 M. Titov, P. M. Ostrovsky, I.V. Gornyi, A. Schuessler, and A. D. Mirlin, Phys. Rev. Lett. 104, 076802 (2010). 40 R.L. Heinisch, F.X. Bronold, H. Fehske, Phys. Rev. B 87, 155409 (2013). 41 J.-S. Wu, M.M. Fogler, Phys. Rev. B 90, 235402 (2014). 42 M.R. Masir, A. Matulis, F.M. Peeters, Phys. Rev. B 84, 245413 (2011). 43 L. D. Landau, E.M. Lifshitz, Quantum Mechanics (Perga- mon, London, 1977), 3rd ed., Chap. XVIII, §142. 44 Yu.I. Latyshev, A.P. Orlov, A.V. Frolov, V.A. Volkov, I.V. Zagorodnev, V.A. Skuratov, Yu.V. Petrov, O.F. Vyvenko, D.Yu. Ivanov, M. Konczykowski, P. Monceau, JETP Let- ters 98, 214 (2013). 45 J.-H. Chen, C. Jang, S. Adam, M. S. Fuhrer, E.D. Williams, M. Ishigami, Nature Physics 4, 377 (2008). 46 Y. Wang, D. Wong, and A.V. Shytov, Science 340, 734 dens. Matter 26, 265301 (2014). (2013). 24 M.V. Berry and R.J. Mondragon, Proc. Roy. Soc. Lond. A 47 Eun-Ah Kim and A. H. Castro Neto, Europhys. Lett. 84, 412, 53 (1987). 57007 (2008). 25 P. Rakyta, M. Vigh, A. Csordas, and J. Cserti, Phys. Rev. 48 F.T. Vasko, O.E. Raichev, Quantum Kinetic Theory and B 91, 125412 (2015). 26 C.G. Beneventano, I.V. Fialkovsky, E.M. Santangelo, Applications (Springer, New York, 2004), Chap. 9, §44.
1009.0497	3	1009	2010-09-29T15:24:05	Quantum Anomalous Hall State in Bilayer Graphene	[ "cond-mat.mes-hall", "cond-mat.str-el" ]	We present a symmetry-based analysis of competition between different gapped states that have been proposed in bilayer graphene (BLG), which are all degenerate on a mean field level. We classify the states in terms of a hidden SU(4) symmetry, and distinguish symmetry protected degeneracies from accidental degeneracies. One of the states, which spontaneously breaks discrete time reversal symmetry but no continuous symmetry, is identified as a Quantum Anomalous Hall (QAH) state, which exhibits quantum Hall effect at zero magnetic field. We investigate the lifting of the accidental degeneracies by thermal and zero point fluctuations, taking account of the modes softened under RG. Working in a 'saddle point plus quadratic fluctuations' approximation, we identify two types of RG- soft modes which have competing effects. Zero point fluctuations, dominated by 'transverse' modes which are unique to BLG, favor the QAH state. Thermal fluctuations, dominated by 'longitudinal' modes, favor a SU(4) symmetry breaking multiplet of states. We discuss the phenomenology and experimental signatures of the QAH state in BLG, and also propose a way to induce the QAH state using weak external magnetic fields.	cond-mat.mes-hall	cond-mat	Quantum Anomalous Hall State in Bilayer Graphene Department of Physics, Massachusetts Institute of Technology, Cambridge, MA02139 Rahul Nandkishore and Leonid Levitov We present a symmetry-based analysis of competition between different gapped states that have been proposed in bilayer graphene (BLG), which are all degenerate on a mean field level. We classify the states in terms of a hidden SU(4) symmetry, and distinguish symmetry protected degeneracies from accidental degeneracies. One of the states, which spontaneously breaks discrete time reversal symmetry but no continuous symmetry, is identified as a Quantum Anomalous Hall (QAH) state, which exhibits quantum Hall effect at zero magnetic field. We investigate the lifting of the accidental degeneracies by thermal and zero point fluctuations, taking account of the modes softened under a renormalisation group procedure (RG). Working in a 'saddle point plus quadratic fluctuations' approximation, we identify two types of RG-soft modes which have competing effects. Zero point fluctuations, dominated by 'transverse' modes which are unique to BLG, favor the QAH state. Ther- mal fluctuations, dominated by 'longitudinal' modes, favor a SU(4) symmetry breaking multiplet of states. We discuss the phenomenology and experimental signatures of the QAH state in BLG, and also propose a way to induce the QAH state using weak external magnetic fields. PACS numbers: I. INTRODUCTION The Quantum Anomalous Hall (QAH) insulator is a state of matter where spontaneous breaking of time re- versal symmetry produces (integer) quantum Hall effect in the absence of any external magnetic field. First pre- dicted in 19881, the QAH state has never yet been ob- served. In the recent literature on interaction driven topological insulators2,3, the elusiveness of the QAH state has been ascribed to fluctuations, which typically disfa- vor the QAH state with respect to a Quantum Spin Hall (QSH) state, which is degenerate with the QAH state on a mean field level. Here we point out that the fluctua- tions which govern the competition of different gapped phases proposed in bilayer graphene (BLG)4 -- 6 are domi- nated by the modes not present in the models2,3, leaving open the door to formation of a QAH state at zero field in BLG. Also, we will propose a mechanism for inducing the QAH state using external fields. The theoretical literature on BLG predicts instabili- ties to numerous strongly correlated states, which are gapped4 -- 6 or gapless7 depending on the way the electron- electron interaction is modeled. The numerous gapped states predicted in the literature are all degenerate at the level of mean field theory4,5, and have the same instabil- ity threshold under one loop RG6. The relation between these different states, and their experimental signatures, have not yet been understood. Meanwhile, indicate that recent experiments the gapped state observed in charge neutral BLG in quan- tizing magnetic fields8 persists down to low fields, cross- ing over to another gapped state at zero field9. However, the nature of the gapped state at zero field is unknown. Hence, clarifying the relation between different gapped states and understanding their physical properties is an interesting and timely task. Here we present a unifying symmetry based analysis of strongly correlated states in BLG. The states predicted in Refs.[4 -- 6] are classified according to a hidden SU(4) flavor symmetry into symmetry breaking multiplets and an SU(4) invariant singlet. The SU(4) singlet is a QAH state. The degeneracy of the multiplets and the singlet is an artefact of the approximations made in the analy- sis, and will be lifted upon taking fluctuation effects into account. Our analysis of fluctuations in BLG focuses on the ef- fect of the modes softened under RG. Those include the 'longitudinal' fluctuation modes (L-modes) analogous to those discussed in Ref.[2,3], and also 'transverse' fluctu- ation modes (T-modes) which are unique to BLG. We find that these two types of modes have competing ef- fects: while the L-modes favor the symmetry breaking multiplets, the T-modes favor the SU(4) invariant QAH state. The zero-point fluctuations are dominated by the T-modes, and hence appear to favor a QAH state at zero temperature. Meanwhile, thermal fluctuations are domi- nated by the L-modes, and favor the symmetry breaking multiplets. We speculate that thermal fluctuations may drive a phase transition from the QAH state at low tem- peratures to a SU(4) symmetry breaking state at higher temperatures, and estimate the transition temperature. We also discuss the phenomenology of the QAH state, its possible experimental signatures, and propose a way to further stabilize it using external magnetic fields. II. SU(4) SYMMETRY In this section we show that within the often-used ap- proximation where the difference between interlayer and intra-layer interactions is neglected4 -- 6,13,14, the interact- ing Hamiltonian is invariant under rotations in a suitably defined four-dimensional flavor subspace. Specifically, we perform a unitary transformation by exchanging the sub- lattices A and B in one of the valleys, upon which the single particle Hamiltonian becomes identical for all spin and valley species, while the layer and sublattice blind interactions are left unchanged. term, which is isotropic in flavor space and thus is SU(4) invariant, dominates because d is small compared to 2 Before entering the discussion of the SU(4) invariance in BLG, we recall that electronic states in BLG at low en- ergy are described by wave-functions on the A and B sub- lattices of the upper and lower layers11, and are fourfold degenerate in spin and valley. To analyze the structure of the Hamiltonian, it will be convenient to combine the spin and valley components in a single eight-component wavefunction ψα,s,v(x), where α is the sublattice (layer) index. We shall use the Pauli matrices in sublattice, spin and valley space, denoted below by τi, σi and ηi, respec- tively. The low energy non-interacting Hamiltonian may then be written as H0 = (px + ipyη3)2 2m τ− + (px − ipyη3)2 2m τ+, (1) where τ± = τ1 ± iτ2. Here m = 0.05me is the effective mass. Because of the presence of η3 in Eq.(1), the sin- gle particle Hamiltonian is not invariant under rotations of valley components. To bring it to an SU(4) invariant form, we perform a unitary transformation on all opera- tors O = U OU †, U = 1 + η3 2 + 1 − η3 2 τ1. (2) This transformation does not act on the spin space, how- ever it mixes the layer and valley indices of the wavefunc- tion ψα,s,v(x) by interchanging the τ pseudospin compo- nent (layer) in one of the valleys. As a result, τ+ and τ− are interchanged and τ3 changes sign in the η3 = −1 valley, after which the free-particle Hamiltonian, Eq.(1), becomes identical in both valleys. Defining p± = px ± ipy, the transformed non- interacting Hamiltonian takes the compact form H0 = p2 + 2m τ− + p2 − 2m τ+, (3) where τ+ and τ− are obtained by transforming τ+ and τ− according to Eq.(2). This single particle Hamiltonian is manifestly invariant under SU (4) rotations in the trans- formed spin/valley flavor space. Meanwhile, electron interactions can be described by a many-body Hamiltonian written in terms of ρq = pψp+q (the density summed over layers) and λq = p τ3 η3ψp+q (the density difference between layers). The interacting Hamiltonian, which incorporates a differ- ence between interlayer and intralayer interaction14, can be written as Pp ψ† Pp ψ† ψ† pH0ψp + H =Xp 1 2Xq V+(q)ρqρ−q + V−λqλ−q, (4) where V+(q) = 2πe2/κq is the Coulomb interaction, and V− = πe2d/κ accounts for the layer polarization energy (here d = 3.5A is the BLG layer separation). The ρρ a0 = 2κ/me2 = 10κ A, (5) the characteristic lengthscale set by interactions5. We therefore approximate by neglecting V−, an approxima- tion that becomes exact in the weak coupling limit, where d/a0 → 0. Under this approximation, the Hamiltonian is invariant under SU(4) flavor rotations, generated by the operators ηi and σi. We will henceforth drop the ∼ symbols for notational convenience, and will refer to the operators τ , η and σ as τ , η and σ respectively. All op- erators are assumed to be transformed operators unless specified otherwise. III. CLASSIFICATION OF STATES AND TOPOLOGICAL PROPERTIES In the transformed basis, the mean field Hamiltonian for the gapped states described in4 -- 6 may be written as H = +τ− + p2 p2 −τ+ 2m + ∆τ3Q, (6) where m = 0.05me is the effective mass. Here the Pauli matrices τi act on the transformed sublattice space, and Q is a 4 × 4 hermitian matrix in the transformed spin- valley space (flavor space), satisfying Q2 = 1. Since unitary hermitian matrices have eigenvalues ±1, all gapped states can be classified as (M>, M<), where M> and M< are the numbers of +1 and −1 eigenval- ues of Q respectively. There are three general types of states: (2, 2), (3, 1), and (4, 0). There is an additional Z2 symmetry associated with the overall sign of Q which is absorbed into the sign of ∆. Following Refs.[1,15], the Hall conductance of a state (M>, M<) can be written as σxy = (M> − M<) e2 h , (7) where we took into account an additional factor of 2 due to the Berry phase 2π in BLG11. The (4,0) and (3,1) states, which have M> 6= M<, thus exhibit a quantized Hall conductance at zero magnetic field -- the hallmark of a QAH state. Because these states have σxy 6= 0, they must spontaneously break time reversal symmetry. We will henceforth focus on comparing the (4,0) and (2,2) states, since the (3,1) states are intermediate between the two. We will refer to the (4,0) state as the QAH state, but it should be remembered that the (3,1) states are also QAH states. In contrast, the (2,2) states have σxy = 0, and preserve time reversal symmetry, but instead exhibit quantum flavor Hall effect. If we parameterize the flavor space by Pauli matrices ηi and σi in transformed valley and spin space respectively, then the Q = σ3 state is a QSH state (Ref.[16]), while the Q = η3 state is a Quan- tum Valley Hall (QVH) state (Ref.[17]). 3 transformations, and are hence degenerate within the ap- proximations leading to SU(4) symmetry. In contrast, the degeneracy of the different manifolds (M>, M<) is purely accidental, and may be lifted in the presence of a weak SU(4) invariant perturbation. As an example, we consider application of a weak trans- verse magnetic field B. Incorporated in the Hamilto- nian (6) through the replacement p → p − eA, it pre- serves the SU(4) symmetry, and causes the spectrum to split into Landau levels19 with an energy spacing of order ωc, where ωc = eB/mc. The Zeeman energy 2µBBσ3 is not SU (4) invariant, but may be neglected since ωc ≫ 2µBB. When ∆ = 0 and Zeeman terms are neglected, the Landau level spectrum is particle-hole symmetric and is fourfold degenerate in flavors. Crucially, the T non-invariance of the mass term ∆τ3 means that the Landau level spectrum for the (4,0) state is not invariant under B → −B and is not particle-hole symmetric15. In particular, the zeroth Landau level, which has an additional two-fold orbital degeneracy19, forms at energy ∆signB only, and has no counterpart at −∆signB (see Fig.1a,b). This breaking of particle hole symmetry can be exploited to induce the (4,0) state using magnetic fields. We illsutrate this by comparing the energies of the (4, 0) QAH state and the (2, 2) QSH state in external magnetic field at filling factor ν = 4. In a magnetic field, these states are no longer degenerate, because of the anomalous Landau level. It is clear from Fig.1 that the QAH state with the appropriate sign of ∆ (such that B∆ < 0) is favored over the QSH state: F4,0 − F2,2 = 4∆B Φ0 < 0, (8) where Φ0 is the flux quantum and FM>,M< is the free en- ergy per unit area for a state (M>, M<). This mechanism for lifting the degeneracy between QAH and QSH states in favor of the QAH state applies to all systems where there is such a degeneracy, including the models studied in Refs.[2,3]. Of course, at finite B, there is no time re- versal symmetry, so the state realized is not a true QAH state, but rather is a state showing quantum Hall effect at anomalously low magnetic fields, which is smoothly connected to a QAH state at B = 0. The analysis above is valid only for sufficiently small B, when BLG at ν = 4 is not far from charge neutrality. This is because the excitonic instability that generates the gap ∆ (Refs.[4 -- 6]) is suppressed by detuning away from charge neutrality. V. SADDLE POINT ANALYSIS We now investigate the energy splitting between the different manifolds at B = 0 by going beyond a mean field approximation, and including the effect of fluctuations. We consider BLG in the presence of screened Coulomb in- teractions between electrons. A static screening approx- FIG. 1: Landau level spectrum of the QAH and QSH states. Note an anomalous Landau level in the QAH state that has no particle-hole-symmetric counterpart. Occupation of this anomalous Landau level allows the QAH state (a) to lower its energy relative to the states (b,c) at filling factor ν = 4. These states are analogs of the 'topological Mott insu- lators' discussed in Refs.[2,3], and as such host topolog- ically protected edge states. The counter-propagating valley modes for the QVH state were worked out in Ref.[17], the co-propagating charge modes and the counter-propagating spin modes for the QAH and QSH states follow similarly. The protection of edge modes is strongest for the (4,0) state due to the unidirec- tional, chiral character of these modes. The counter- propagating spin currents in the QSH state are protected in the absence of spin-flip scattering, while the counter- propagating valley currents in the QVH state are pro- tected in the absence of intervalley scattering (e.g. by short range disorder). We note that the above classification of states superfi- cially resembles that arising in an entirely different prob- lem, namely the Quantum Hall Ferromagnet (QHF) in graphene in quantizing magnetic field18. In the latter case, however, the integers M> and M< are fixed by the electron density, i.e. by filling of the four-fold degenerate zeroth Landau level. In the QHF problem, spontaneous time-reversal symmetry breaking cannot occur: the ana- log of the (4,0) QAH state is a fully filled zeroth Landau level exhibiting quantized Hall conductance 2e2/h. Fur- thermore, in the QHF problem there is no competition between states with different M> and M< values, which is the main question of interest for us here. IV. LIFTING ACCIDENTAL DEGENERACIES USING EXTERNAL FIELDS The SU(4) symmetry of the Hamiltonian guarantees the degeneracy of all states within a given manifold (M>, M<), even when the states involved have very different physical properties. For example, the QVH state is a ferroelectric state which polarizes the layers by charge, while the QSH state polarizes the layers by spin and val- ley. Nonetheless, the two states are related by SU(4) imation, ignoring the effects of dynamical screening5, is sufficient to understand the main features. In this ap- proximation, the interaction is short range, and we can write the partition function as a functional field integral in Euclidean time, Z =Z Dψ†Dψ exp(cid:18)−Z dxL[ψ†(x), ψ(x)](cid:19) , (9) where x = (t, r), dx = dtd2r, the ψ fields are fermionic fields, with the Lagrangian L = ψ†(∂t + H0)ψ + Xj,k=1...8 λ 2 ψ† j ψ† kψkψj. (10) Here j, k are combined sublattice and flavor indices, and H0 is the non-interacting Hamiltonian (given by Eq.(6) at ∆ = 0). The coupling constant λ represents the statically screened Coulomb interaction, which in the RPA model takes value20 λ = 1/(4ν0 ln 4), where ν0 = m/2π is the non-interacting single species density of states. fermion four We the now decouple term via tion formation in a Hubbard-Stratonovich to exchange channel, the Z =R Dψ†DψDh exp[−R dxL(ψ†, ψ, h)], where interac- trans- obtain L = ψ†(cid:2)∂t + H0 + h(cid:3)ψ + 1 2λ Tr[hh†]. Here, h is an 8 × 8 hermitian matrix, which we write as h = M ⊗ Q, where M is a 2 × 2 hermitian matrix in sublattice space and Q is a 4 × 4 hermitian matrix in flavor space. The gapped states (Ref.[4 -- 6]) correspond to taking M = ∆τ3. Integrating out the fermions yields Tr[Q2]. (12) Z =R D(Q)D(∆) exp(−R dxL[∆(x), Q(x)], where L(∆, Q) = −Tr ln(cid:2)∂t + H0 + ∆τ3Q(cid:3) + ∆2 λ The SU(4) flavor invariance manifests itself in an exact SU(4) flavor degeneracy of the many body states. Upon minimizing the action (12) in a saddle point approxima- tion, we find Q2 = 1, and ∆ = Λ exp(−2/λν0), where Λ ≈ 0.4 eV is the bandwidth for the two band Hamilto- nian. This gives the mean field Hamiltonian, Eq.(6). We note that instead of decoupling the interaction in the excitonic channel h = τ3 ⊗ Q, we could have chosen the channel h = τ1,2 ⊗ Q. This choice would lead us to the nematic state of Ref.7, which is gapless, but breaks lattice rotation symmetry. However, the nematic state is higher in energy than the gapped states at the saddle point level, so we will concentrate on the gapped states, and specifically on the lifting of the accidental degenera- cies by thermal and zero-point fluctuations. Our analysis, symmetry involving multiplets (M>, M<) for different matrices Q, could also be applied to the nematic state7. However, the fluctuation analysis cannot be perfomed because the τ3δQ mode has negative rigidity, i.e. the nematic mean field is unstable. K αβ ijkl(ω, k) = δilδjk(cid:18) δαβ λ + Παβ ij (ω, k)(cid:19) , where we have introduced the polarization operator (11) (14) (15) 4 VI. LIFTING THE DEGENERACY: ZERO-POINT FLUCTUATIONS We first analyze the case of zero temperature, when the degeneracy is lifted by zero point fluctuations. The most important fluctuation modes are those that are soft- ened under RG. In BLG, this means the 'L' modes τ3δQ, which describe fluctuations longitudinal with respect to the order parameter in sublattice space, and also the 'T' modes τ1,2δQ, which describe fluctuations transverse to the order parameter in sublattice space. In that, δQ is an arbitrary 4 × 4 hermitian matrix. We therefore expand the action in Eq.(12) to quadratic order in the fluctuation modes τα ⊗ δQα, α = 1, 2, 3, to obtain δ2S = XijklαβXω,q δQα ij,ωqK αβ ijkl(ω, q)δQβ kl,−ω,−q. (13) Here, Latin indices i, j, k = 1...N refer to fermion flavor, whereas Greek indices α, β = 1, 2, 3 refer to the Pauli ma- trices τα that parameterize the fluctuations in sublattice space. The matrix K is defined by Παβ ij (ω, q) =Z d2pdε 2(2π)3 Tr (ταGi(p+)τβGj (p−)) . It is convenient to choose a diagonal background state Q = ζiδij, where ζi = ±1, so that the Greens function takes a form diagonal in the flavor space, Gi(p±) = 1 i(ε ± 1 2 ω) − H0(p ± 1 2 q) − ζi∆τ3 . (16) The trace in Eq.(15) goes over sublattice indices, but not over flavors. The matrix K is positive definite, so we may integrate out fluctuations to obtain an expression for the fluctua- tion contribution to the free energy, Ffluct = 1 2Xαij Xωk ln K αα ijji(ω, k), (17) where we took into account that the only contribution comes from the diagonal terms, α = β, i = l, j = k. We now subtract the fluctuation energy of the (4, 0) QAH state from that of the (2, 2) state, to obtain δF = Ffluct,(4,0) − Ffluct,(2,2) = 4 ln 1 λ + Παα >> 1 λ + Παα >< 3 Xα=1 , (18) where Παα >> and Παα >< are defined by Eq.(15), with (ζi, ζj) = (1, 1) and (1, −1) respectively: Παβ (2π)3 >>(ω, q) =Z d2pdε ><(ω, q) =Z d2pdε (2π)3 Παβ G>(<)(ε, p) = 1 2 1 2 Tr (ταG>(p+)ταG>(p−)) ,(19) Tr (ταG>(p+)ταG<(p−)) ,(20) 1 iε − H0(p) ∓ ∆τ3 , where we used a shorthand notation p± = (ε ± 1 2 ω, p ± 1 2 q). To analyze the effect of competition of different modes in full detail, below we compare the fluctuation energy for the states of different type (M>, M<). To evaluate the difference of fluctuation energies, given by Eq.(18), it is convenient to rewrite it as δF = 4 3 Xα=1 ln(cid:18)1 + >> − Παα Παα >< 1 λ + Παα >< (cid:19) (21) >> − Παα In particular, we find that Π33 Below we evaluate the differences of polarization func- tions Παα ><, and find that different modes, L and T, yield contributions of opposite sign. >> − Π33 >< is positive, i.e. the L-modes favor the (2,2) state. This effect of longi- tudinal modes is well known in the topological insula- >> − Παα tor literature2. In contrast, the differences Παα >< with α = 1, 2 are negative. Thus, the T-modes, which are unique to BLG, favor the (4,0) state. We evaluate Eq.(18), and find that the T-modes dominate the free energy, favoring the QAH state. To proceed with the analysis of the quantities Παα >> − ><, it is convenient to define ε± = ε ± ω/2, and z± = 2 q2/2m. In this compact notation, we have Παα p ± 1 Παα >> − Παα >< =Z Fαα(ε, p) (ε2 + + z2 + + ∆2)(ε2 − + z2 − + ∆2) ,(22) Fαα(ε, p) = ∆2Tr(τατ3τατ3) + ∆Tr(τατατ3)(iε+) +∆Tr(ταH0(p + 1 2 q)τατ3), where R ... = R d2pdε (2π)3 .... Terms in Eq.(22) linear in ∆ must vanish, since the fluctuation energy should be invariant under sign changing ∆ → −∆. Techni- cally, the vanishing of terms linear in ∆ follows because Tr(τατ3τα) = 0, and Tr(ταH0τατ3) = 0. As a result, the first term in Fαα(ε, p) (at order ∆2) is the only term that survives. We can substitute the expression in Eq.(22) into Eq.(18) and expand the logarithm in small ∆2, to obtain δF = 4Z Xα = 4Z (cid:18)− Dαα(ω, q)(ε2 1 D11(ω, q) − + 1 D33(ω, q)(cid:19) ∆2Tr(τατ3τατ3) + + z2 + + ∆2)(ε2 1 D22(ω, q) 2∆2 + + ∆2)(ε2 − + z2 − + ∆2) (23) ,(24) (ε2 + + z2 − + z2 − + ∆2) where R ... = R dεdωd2pd2q λ + ><(ω, q). The integral over ε may be performed ex- Παα actly by the method of residues, to give ... and Dαα(ω, q) = 1 (2π)6 5 (2π)5 (cid:18)− δF = 4∆2Z dωd2qd2p D33(ω, q)(cid:19) 1 ξ+ + 1 1 D11(ω, q) + 1 ξ− ω2 + (ξ+ + ξ−)2 , − 1 D22(ω, q) where ξ± = qz2 ± + ∆2. The integral over p may now be performed with logarithmic accuracy. The dominant contributions come from ξ± ≈ 0, and may be evaluated as δF = 8∆2ν0Z dωd2q (2π)3 (cid:18)− D33(ω, q)(cid:19) ln(r/∆) r2 + 1 1 D11(ω, q) − 1 D22(ω, q) , (25) where we have used the pseudo-polar coordinates r2 = ω2 + (q2/2m)2 and have assumed that r ≫ ∆. We now have to calculate the various functions Dαα. We will calculate these quantities analytically with log- arithmic accuracy. We begin with the definition Dαα = 1 λ + Παα ><, where the polarization functions are defined in Eqs.(19),(20). We note that the polarization func- tions Παα ij are logarithmically divergent at small ω, small q2/2m and ∆ = 0. The coefficient of the logarithm can be extracted by setting ω, q, ∆ = 0 in the integral in Eqs.(19),(20), and introducing an IR cutoff r, where , and we assume r & ∆. In this r2 = ω2 +(cid:0)q2/2m(cid:1)2 manner, we obtain D11 = D22 = 1 λ D33 = 1 λ −Z Λ r −Z Λ r dεd2p (2π)3 ε2 − z2e2iθp (ε2 + z2)2 , , dεd2p (2π)3 ε2 + z2 (ε2 + z2)2 , (26) (27) where we have introduced the notation z = p2 2m and peiθp = px + ipy. The integrals may be straightfor- wardly performed by changing to the pseudopolar coor- dinates (ρ, ϕ, θp), where ε = ρ cos ϕ, z = ρ sin ϕ, and θp was defined above. The integral goes over 0 < θp < 2π, 0 < ϕ < π and r < ρ < Λ. Integrating in turn over θp, ϕ and ρ, we find D11 = D22 = 1 λ − ν0 4 ln Λ r , D33 = 1 λ − ν0 2 ln Λ r . (28) We now recall the relation λ−1 = 1 equation), and substitute it into Eq.(25), to obtain 2 ν0 ln Λ/∆ (the gap δF = 8∆2ν0Z Λ ∆ dr r − 4 ln Λ ∆ + ln r ∆ + 1 ln r ∆! ln r ∆ . (29) This integral can be evaluated using the substitution x = ln r ∆ , giving into account L-modes only, is given by a sum over Mat- subara frequencies, δF = 8∆2ν0Z ln(Λ/∆) 0 ln(Λ/∆) − 3x ln(Λ/∆) + x dx. (30) Ffluct = 1 2 T Xωn,kXi,j ln(cid:18) 1 λ + Π33 ij (ωn, k)(cid:19) , (34) Evaluating the integral, we obtain a negative value where ωn = 2πnT . 6 δF = 8(−3 + 4 ln 2)∆2ν0 ln Λ ∆ ≈ −1.82∆2ν0 ln Λ ∆ , (31) which favors the QAH state. It should be noted that the difference in energies be- tween the (4,0) and (2,2) manifolds is of the same order as the mean field energy, so the mean field plus fluctu- ations analysis is ill controlled. However, it provides us with an intuition about the splitting between manifolds of different signatures, and we believe the qualitative de- tails of the fluctuation splitting are reproduced correctly by this analysis. We note that our fluctuation analysis included only those modes that correspond to weak coupling instability in BLG. We could also have included Stoner modes in our fluctuation analysis. These would produce an additional contribution δFStoner = 8∆2ν0Z dωd2q (2π)3 D00 = + Π00 ><, 1 λ 1 ln(r/∆) D00(ω, q) r2 ,(32) (33) where Π00 >< is defined by Eq.(20) with α = β = 0, i.e. with τα = τβ = 1. Now, since Π00 >< is not log diver- gent, we can take D00 = 1/λ with logarithmic accuracy. We then obtain a contribution δFStoner = 4∆2ν0 ln Λ/∆, which is sufficiently large to change the sign of the result Eq.(31). However, this calculation, which neglects corre- lation effects, is likely to strongly overestimate the effect of Stoner modes, and therefore we believe that Stoner modes should be left out of the fluctuation analysis. VII. LIFTING THE DEGENERACY: THERMAL FLUCTUATIONS Thermal fluctuations are dominated by gapless Gold- stone modes, which are present only in the states that break SU(4) symmetry. In a state (M>, M<), there are M>M< Goldstone modes. Thermal fluctuations due to Goldstone modes allow a state to gain entropy, and since the (2,2) states have the most Goldstone modes, they have the highest entropy. It may thus be expected that the (2,2) states dominate at sufficiently high tempera- ture. Below we present an analysis showing that this expec- tation is correct. Since gapless fluctuation modes appear only in the L-mode channel δh ∝ τ3δQ, it is sufficient to restrict our attention to the L-modes. The general ex- pression for the fluctuation part of the free energy, taking We will perform a long wavelength expansion of ij (ω, k). At zeroth order, we note that at ω, k = 0 Π33 the values of Π33 >< and Π33 >> are given by Π33 >>(ω, k = 0) = ε2 + z2 − ∆2 (ε2 + z2 + ∆2)2 1 2Z d2pdε (2π)3 tr (τ3G>τ3G>) d2pdε (2π)3 , 2Z d2pdε (2π)3 tr (τ3G>τ3G<) d2pdε (2π)3 , 1 = −Z = −Z Π33 ><(ω, k = 0) = ε2 + z2 + ∆2 (ε2 + z2 + ∆2)2 where G>(<) = 1/(iε − H0(p) ∓ ∆τ3). To distinguish Goldstone modes from gapped modes, it is convenient to recall the gap equation 1 λ =Z 1 ε2 + z2 + ∆2 d2pdε (2π)3 . (35) Hence, we have 1 Goldstone mode, whereas in the case of Π33 λ + Π33 ><(0) = 0, which corresponds to a >> we have 1 λ + Π33 >>(0) =Z 2∆2 (ε2 + z2 + ∆2)2 d2pdε (2π)3 , (36) which is manifestly positive. Thus, Goldstone modes ex- ist only in states (M>, M<), where M> 6= 0 and M< 6= 0. The free energy, Eq.(34), evaluated at leading order in a long wavelength expansion around ω, k = 0, is given by a sum M>M< ln(aω2 n + bk2) (37) Ffluct = T Xωn,k 1 2 + (M 2 > + M 2 <) ln(a′ω2 n + b′k2 + c), (38) where the first term is the contribution of the gapless modes (originating from Π33 ><), while the second term is the contribution of the gapped modes (originating from >>). The coefficients a, a′, b, b′ are obtained by Taylor Π33 expanding Π33 ij (ω, k) in small ω and k, while c is given by Eq.(36). To simplify the sum over Matsubara frequencies, it is n+ u2). We can evaluate f (u) by first taking the derivative convenient to define the quantity f (u) = T Pωn,k ln(ω2 df du = T Xωn,k 1 iωn + u + c.c. = coth u 2T , and then integrating it over u to obtain f (u) = 2T ln sinh u 2T = 2T ln(cid:16)1 − e−u/T(cid:17)+u−(2 ln 2)T. Plugging this identity into the sum (37), we see that the contribution of the gapped modes is exponentially small at low temperatures, T ≪ pc/a′ ∼ ∆, while the sum over gapless modes gives a negative contribution of a power law form, Ffluct =Xk 2M>M<T ln(cid:16)1 − e−vq/T(cid:17) , (39) where v = b/a ∼ p∆/m. Evaluating the integral, we obtain an estimate Ffluct ∼ −M>M<(ν0/∆)T 3, (40) which describes the free energy gain due to thermal fluc- tuations of Goldstone modes. We see that the gapless Goldstone modes dominate the finite-temperature fluctuation contribution to the free energy. These modes lower the free energy (by increasing entropy). Since the number of gapless modes M>M< is maximal for the (2,2) states, these states are entropically favored by thermal fluctuations. What is the outcome of competition between the zero- point fluctuations and thermal fluctuations? In Sec.VI we found that at zero temperature the (0,4) QAH state is energetically favored by zero point fluctuations of the modes "softened" under RG. At the same time, the zero- point fluctuations of other modes, such as the Stoner modes, may have an opposite effect, favoring the (2,2) state. In the event the zero-point fluctuation energy is dominated by such non-soft modes, the (2,2) state will be realized in the entire temperature interval where the system is unstable to gap formation. A more interesting situation may arise if the zero- point fluctuation energy is dominated by the RG-softened modes, favoring the QAH state at zero temperature. In this case, given the opposite effect of zero-point and ther- mal fluctuations, we have to consider the competition between the QAH and (2,2) states. Since the thermal fluctuation energy (40) vanishes at T = 0, we expect that zero point fluctuations will dominate below a cer- tain temperature T∗, above which thermal fluctuations will dominate. If T∗ < Tc, where Tc ≈ ∆(T = 0) is the critical temperature for gap opening, then a QAH state will be realized at low temperatures 0 < T < T∗, whereas a (2,2) gapped state will be realized in the in- terval T∗ < T < Tc. In contrast, if T∗ > Tc, then the QAH state will transition directly to an ungapped state at T = Tc via a second order phase transition, and the (2,2) state will not be realized. A rough estimate of the temperature T∗ can be ob- tained by comparing the free energies (40) and (31), δFfluct,(2,2) ∼ −4ν0 T 3 ∆ , δFfluct,(0,4) ∼ −1.82ν0∆2 ln , Λ ∆ (41) 7 indicating that the scale for T∗ is comparable to the tem- perature Tc at which the gapped state forms. A more de- tailed analysis of temperature-driven transition between the QAH state and (2,2) state is beyond the scope of this work. VIII. EXPERIMENTAL SIGNATURES OF THE QAH STATE We now discuss experimental tests of the QAH state. The clearest experimental signature would be detection of the quantum Hall effect at zero external magnetic field. However, detection of this effect requires four probe mea- surements performed on a sample of BLG that is suffi- ciently clean and at sufficiently low temperatures as to exhibit spontaneous gap opening5. Such measurements have not yet been performed. Moreover, detection of this effect could be complicated by the formation of domains with opposite signs of ∆. Different domains will have opposite σxy, so the Hall conductance of a macroscopic sample will average to a value near zero. However, if there is percolation of edges, there will be a non-vanishing two-terminal conductance of order e2/h. Alternative experimental tests of the QAH state may be performed by examining the electronic compressibility in weak magnetic fields. When the chemical potential sits near the missing Landau level in Fig.(1), there should be a gap that extrapolates to a non-zero value as B → 0. This effect will be seen at either ν = 4 or ν = −4 if there is only one domain, and at ν = ±4 if there are multiple domains. The gap at ν = 4 will be strengthened by the mecha- nism outlined around Eq.(8), however, a signal at ν = −4 will be seen only if the QAH state is intrinsic, rather than field induced. An incompressible region at ν = −4 com- bined with a gapped state at B = 0 can thus be taken as a diagnostic for a QAH state at B = 0. The filling factors ν = ±4 are not equivalent because the QAH state breaks particle-hole symmetry in magnetic field. Another experimental signature is a phase transition at filling factor ν = 0 and finite B from a QAH state to the Quantum Hall Ferromagnet (QHF) states that are expected to form at large magnetic fields13. Such a phase transition would not be seen if the dominant state at small B was of (2,2) type, since the (2,2) states are smoothly connected to the QHF state. An incompressible region at ν = ±4 that occurs at anomalously low magnetic fields, such that the features in compressibility at other integer ν values are washed out, was found in recent experiments that employed a capaci- tance scanning probe to study suspended BLG samples9. In transport measurements10 performed on the same sys- tem, a state with finite two-terminal conductance of or- der e2/h was found at zero field, which at a finite B field undergoes a transition to an insulating state. These measurements are all compatible with the QAH state, however, since there is as yet no four-terminal measure- ment, it is not possible to say for certain whether a QAH state has been observed. In summary, our symmetry classification of the vari- ous gapped states proposed for BLG singles out the QAH state as the only gapped state not breaking any contin- uous symmetry. We have investigated the fluctuation- induced splitting of the gapped states, and concluded that at zero temperature and zero field, the leading in- stability is to the QAH state. We have discussed the phenomenology and experimental signatures of this state, and have shown that it can be stabilized by weak external magnetic field. We thank M. Allen, B. Feldman, J. Martin, T. Weitz, and A. Yacoby for sharing with us unpublished data. We also acknowledge useful discussions with V. Desh- pande, A. Young, S. E. Korshunov and P.Kim. This work was supported by Office of Naval Research Grant No. N00014-09-1-0724. 8 1 F.D.M. Haldane, Phys. Rev. Lett. 61, 2015 (1988). 2 S. Raghu, X.L. Qi, C. Honerkamp, S.C. Zhang, Phys. Rev. Lett. 100, 156401 (2008). 12 I. Martin and C. D. Batista, Phys. Rev. Lett. 101, 156402 (2008). 13 Y. Barlas, R. Cote, K. Nomura and A. H. MacDonald, 3 K. Sun, H. Yao, E. Fradkin and S. Kivelson, Phys. Rev. Phys Rev Lett. 101 (2008). Lett. 103, 046811 (2009). 14 J. Nilsson, A.H.Castro Neto, N. Peres and F. Guinea, Phys. 4 H. Min, G. Borghi, M. Polini and A.H. MacDonald, Phys. Rev. B 73, 214418 (2006). Rev. B 77, 041407(R) (2008). 5 R. Nandkishore and L. Levitov. Phys. Rev. Lett. 104, 15 R. Jackiw, Phys. Rev. D 27, 2375 (1984). 16 C. L. Kane and E. J. Mele, Phys. Rev. Lett. 95, 226801 156803 (2010). (2005). 6 F. Zhang, H. Min, M. Polini, and A. H. MacDonald Phys. 17 I. Martin, Y. Blanter and A. Morpurgo, Phys. Rev. Lett. Rev. B 81, 041402(R) (2010). 7 O. Vafek and K. Yang, Phys. Rev. B 81, 041401(R) (2010). 8 B. Feldman, J. Martin and A. Yacoby, Nature Physics 5, 889 (2009). 100, 036804 (2008). 18 K. Yang, S. Das Sarma, A. H. MacDonald, Phys. Rev. B 74, 075423 (2006). 19 E. McCann and V. Fal'ko, Phys. Rev. Lett. 96, 086805 9 J. Martin, B.E.Feldman, R.T.Weitz, M.T.Allen and (2006). A.Yacoby, arXiv: 1009.2069 (2010) 20 E. H. Hwang and S. Das Sarma, Phys. Rev. Lett. 101, 10 R.T.Weitz, M.T.Allen, B.E.Feldman, J. Martin and A. Ya- 156802 (2008). coby, to be published. 11 K.S. Novoselov et al, Nature Physics 2, 177 (2006).
1510.06109	1	1510	2015-10-21T02:21:40	Experimental Observation of Surface States and Landau Levels Bending in Bilayer Graphene	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	We report on microscopic measurements of the low-energy electronic structures both at zigzag and armchair edges of bilayer graphene using scanning tunneling microscopy and spectroscopy (STM and STS). We have found that, both in the absence and in the presence of a magnetic field, an almost zero-energy peak in density of states was localized at zigzag edges, as expected for the surface states at zigzag edges of bilayer graphene. In the quantum Hall regime, we have observed clearly Landau levels bending away from the charge neutrality point near both the zigzag and armchair edges. Such a result is a direct evidence for the evolution of Landau levels into the quantum Hall edge states in graphene bilayers. Our experiment indicates that it is possible to explore rich quantum Hall physics in graphene systems using STM and STS.	cond-mat.mes-hall	cond-mat	Experimental Observation of Surface States and Landau Levels Bending in Bilayer Graphene Long-Jing Yin§, Yu Zhang§, Jia-Bin Qiao, Si-Yu Li, and Lin He* Center for Advanced Quantum Studies, Department of Physics, Beijing Normal University, Beijing, 100875, People’s Republic of China §These authors contributed equally to this work. * Email: [email protected] We report on microscopic measurements of the low-energy electronic structures both at zigzag and armchair edges of bilayer graphene using scanning tunneling microscopy and spectroscopy (STM and STS). We have found that, both in the absence and in the presence of a magnetic field, an almost zero-energy peak in density of states was localized at zigzag edges, as expected for the surface states at zigzag edges of bilayer graphene. In the quantum Hall regime, we have observed clearly Landau levels bending away from the charge neutrality point near both the zigzag and armchair edges. Such a result is a direct evidence for the evolution of Landau levels into the quantum Hall edge states in graphene bilayers. Our experiment indicates that it is possible to explore rich quantum Hall physics in graphene systems using STM and STS. 1 There are two possible (perfect) edge terminations, i.e., zigzag and armchair, in graphene monolayer and the edge orientations affect the electronic structures of graphene sheet strongly [1,2]. Very recently, graphene system with zigzag edge termination has attracted much attention because that its surface states are believed to be closely related to the magnetic order and exceptional ballistic transport [3-5]. In the quantum Hall regime, both the zigzag and armchair edges could bend Landau levels (LLs) to produce dispersive edge states [6-8], which carry the chiral Dirac fermions responsible for the quantum Hall effect in graphene monolayer [9,10]. In graphene bilayers, the zigzag edge is also predicted to host surface states, but with an enhanced penetration into the bulk comparing to that of graphene monolayer [11]. In the presence of high magnetic fields, unconventional quantum Hall effect and a wealth of exotic electronic behavior have been observed in graphene bilayers [12-18]. Though many great successes have been achieved in the study of electronic properties of graphene bilayer, considerable work is still necessary to address some fundamental problems in this system. For example, a direct experimental observation of the surface states at the zigzag edges of graphene bilayer and its LLs bending at the edge terminations is still lacking up to now. In this Letter, we present scanning tunneling microscopy (STM) and spectroscopy (STS) measurements of bilayer graphene on graphite substrate both in the absence and in the presence of magnetic fields. The high-quality bilayer sample and the ultra-low random potential fluctuations due to substrate imperfections ensure us to direct probe the surface states at the zigzag edges and to measure the LLs bending at both the zigzag and armchair edges. We performed STM measurements in an UNISOKU (USM-1500S) instrument with the magnetic fields up to 8 T. The STS spectra, i.e., differential conductance dI/dV curves, were measured with a lock-in detection (modulation voltage: 5-10 mV, frequency: 793 Hz). The STM tips were obtained by chemical etching from a wire of Pt0.8Ir0.2 alloys. Lateral dimensions observed in the STM images were calibrated using a standard graphene lattice and Ag (111) surface. All the STM and STS measurements were performed in the ultrahigh vacuum chamber (~ 10-11 Torr) at ~ 4.4 K. Bilayer 2 graphene samples used in our experiments were prepared on ZYA grade (NT-MDT) Highly Oriented Pyrolytic Graphite (HOPG) substrates. The HOPG samples were surface cleaved by adhesive tape prior to experiments immediately. The bilayer graphene flakes deposited on the substrate during the process of mechanical exfoliation and, very importantly, these graphene sheets may decouple from the graphite surface, as demonstrated in previous studies [19-25]. Figure 1(a) shows a representative STM image of a Bernal graphene bilayer on graphite surface. The triangular contrast in the atomic image arises from the A/B atoms’ asymmetry generated by the adjacent two layers. To further identify the bilayer graphene region, we carried out STS measurement in various magnetic fields, as shown in Fig. 1(b). In zero magnetic field, there are two peaks located at about 0 mV and 25 mV in the spectrum, which are attributed to the density of states (DOS) peak generated at the valence-band edge (VBE) and conduction-band edge (CBE) of a gapped bilayer respectively [22,26]. The finite gap in the low-energy bands is generated by inversion symmetry breaking of the adjacent two layers induced by the substrate [22,26-33]. The spectra recorded in high magnetic fields [Fig. 1(b)] exhibit Landau quantization of massive Dirac fermions, as expected for the gapped graphene bilayers [13,22,26]. The LL sequences of gapped graphene bilayers can be described by En = EC ± [( ωc)2(n(n-1)) + (U/2)]1/2 – ξzU/4, n = 2,3,4…  E0 = EC + ξ U/2, E1 = EC + ξ (U/2)(1- z), (1) where EC is the energy of charge neutrality point (CNP), ωc = eB/m* is the cyclotron frequency, m* is the effective mass of charge carriers, and ξ = ± are the valley indices. We have z = 2 tω ⊥  /c << 1 for B ≤ 8 T and U ≈ Eg (gap energy) when the interlayer bias U < t⊥ . According to the fitting, as shown in Fig. 1(c), we obtain Eg ≈ 25 meV and m* = (0.035 ± 0.002)me (me is the free-electron mass). Both the values of Eg and m* agree well with the range of values reported in Bernal bilayers previously [22,26]. Note that the two lowest levels LL(0,1,+) and LL(0,1,-) are a couple of layer-polarized 3 its edges are carefully studied. Fig. 2(a) and 2(b) show quartets, and they are mainly localized on the first and second graphene layers, respectively. Therefore, the signal of LL(0,1,+) is much stronger than that of LL(0,1,-) in the spectra since that the STS predominantly probes the DOS on the top layer. The said measurements demonstrate explicitly that the topmost two layers are Bernal stacking and they are completely decoupled from the substrate. Once a high quality bilayer graphene region is identified, the structures and electronic properties around typical atomic-resolution images of a zigzag edge and an armchair edge of the graphene bilayer, respectively. Away from the edges, the STM images exhibit triangular contrasting, as expected to be observed in Bernal bilayer. The types of the terminative edges can be determined by the arrangement of the triangular dots, as schematically shown in Fig. 2(a-c). Around both the zigzag and armchair edges, clear interference patterns are observed. Similar interference patterns have also been observed around edges of graphene monolayer on graphite substrate [34-36] and are attributed to the interference between the incident and scattered electron waves in two Dirac cones at the atomically sharped boundaries (see Fig. S1 in Supplementary Information [37] for more experimental data). The graphene bilayers with zigzag and armchair edges are expected to exhibit quite different electronic band structures: there are localized surface states at zigzag edges but not at armchair edges [11,38], as shown in Fig. 2(d)-2(f). In a gapped graphene bilayer with zigzag edges, the surface states may be layer-polarized [38] and they are predicted to have a much larger penetration length into the bulk than that in graphene monolayer [11]. To study the effect of edges on the electronic properties of bilayer graphene, we measured the spatial-resolved dI/dV spectra near both the zigzag and armchair edges under zero magnetic field. Fig. 3 shows a representative result obtained around a zigzag edge (see Fig. S2 in Supplementary Information [37] for experimental data recorded around an armchair edge). Typical tunneling spectra recorded at different distances away from the edge are shown in Fig. 3(a) [here we define zero-position at the zigzag edge, as shown in Fig. 3(b)]. With approaching the edge, the signal of the DOS peaks at the VBE and CBE [dashed lines in Fig. 3(a)] becomes weak because 4 that the VBE and CBE of a gapped bilayer graphene are exactly valid in the bulk and they are less well defined around the edges. The energy spacing between the VBE and the CBE increases about 10 meV [Fig. 3(a)], which may arise from a slightly enhanced band gap around the edge. Beside the said result, another notable feature of the spectra is the emergence of a new DOS peak around the zigzag edge and the signal of the peak increases with approaching the edge. Such a peak, which is absent around the armchair edge (Figure S2), is attributed to a layer-polarized surface state at the zigzag edge of the gapped graphene bilayer [38]. The presence of this DOS peak (surface state) is a fundamental result, although anticipated in many theoretical works, had never been experimentally observed before in graphene bilayer. Fig. 3(c) plots the measured peak height of the surface state as a function of the distance from the edge. The surface state shows a decreasing intensity with increasing distance and extends over 10 nm away from the edge, consistent with the expected decaying behavior of the surface states in graphene bilayer. Here we should point out that the decaying length of the surface states in graphene bilayer is much larger than that in graphene monolayer [Fig. 3(c)]. Additionally, the surface state can also be detected even in the quantum Hall regime [Fig. 4(a) and 4(b)]. two-dimensional electron systems, low-energy band structures of quasi-particles develop into dispersionless LLs in the presence of a high magnetic field and give raise to the insulating behavior in the bulk. While the confining potential at the edges of the system bends the discrete LLs to form dispersive edge states that carry charge carries in the quantum Hall effect. The high-quality bilayer sample with crystallographically perfect edges and the ultra-low random potential fluctuations induced by the substrate, as demonstrated in Fig. 1-3, allow us to direct probe the LLs bending at the edges. In the Figure 4 summarizes the measured result of bilayer graphene in the quantum Hall regime and we observe clearly LLs bending at both the zigzag and armchair edges (see Fig. S3 in Supplementary Information [37] for more experimental data). Away from the edges, the well-defined LL spectra, as shown in Fig. 4(a)-4(d), follow the sequence of massive Dirac fermions in gapped graphene bilayers (here we use 5 =  Bl eB to define the distance from the edge). With approaching the edges, the DOS peaks for the LLs become weak and the LLs are shifted away from the charge neutrality point, as shown in Fig. 4(e) and 4(f). At a fixed energy, the measured local DOS at position r is determined by the wavefunctions according to r r ( ) ψ∝ 2 r ( ) , while the wavefunctions of LLs have their spatial extent, ~ 2 BNl . It indicates that there is an important contribution from the bulk states even for the recorded LL spectra near the edges. The wavefunctions of LLs with higher indices have greater spatial extents, as shown in the inset of Fig. 4(f). Consequently, the amplitude of high-index LL peaks decreases slower than that of low-index LL peaks [Fig. 4(e)] and the bending of low-index LLs seems stronger than that of high-index LLs [Fig. 4(f)] (due to greater contribution from the bulk states to higher LLs). Theoretically, the shift length of the LLs bending around the edges is predicted to be of the magnetic length [7,8]. In Fig. 4(g), we summarized the measured shift length at different magnetic fields around both the zigzag and armchair edges. We find that the shift length depends on neither the magnetic fields nor the edge types, and it is of the magnetic length (see Fig. S4 in Supplementary Information [37] for more experimental data). Additionally, the shift length seems to be dependent on the LL index: the estimated shift lengths for the LL(0,1,+) and LL2 are about ~ 1.4 lB and ~ 2.0 lB, respectively. In conclusion, we measured the surface state and its spatial evolution around the zigzag edges of bilayer graphene. Our result demonstrated an enhanced penetration length of the surface states in bilayer graphene comparing to that in graphene monolayer. In the quantum Hall regime, we provided direct evidence for the LLs bending around both the zigzag and armchair edges of bilayer graphene, which may open the door to explore exotic quantum Hall physics in graphene bilayers using scanned probe techniques. 6 Acknowledgments in University of This work was supported by the National Basic Research Program of China (Grants Nos. 2014CB920903, 2013CBA01603), the National Natural Science Foundation of China (Grant Nos. 11422430, 11374035), the program for New Century Excellent Talents the Ministry of Education of China (Grant No. NCET-13-0054), Beijing Higher Education Young Elite Teacher Project (Grant No. YETP0238). L.H. also acknowledges support from the National Program for Support of Top-notch Young Professionals. 7 Reference: [1] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, A. K. Geim, The electronic properties of graphene. Rev. Mod. Phys. 81, 109-162 (2009). [2] M. O. Goerbig, Electronic properties of graphene in a strong magnetic field. Rev. Mod. Phys. 83, 1193-1243 (2011). [3] Y.-W. Son, M. L. Cohen, S. G. Louie, Half-metallic graphene nanoribbons. Nature 444, 347 (2006). [4] G. Z. Magda, X. Jin, I. Hagymasi, P. Vancso, Z. Osvath, P. Nemes-Incze, C. Hwang, L. P. Biro, L. Tapaszto, Room-temperature magnetic order on zigzag edges of narrow graphene nanoribbons. Nature 514, 608 (2014). [5] J. Baringhaus, M. Ruan, F. Edler, A. Tejeda, M. Sicot, A. Taleb-Ibrahimi, A.-P. Li, Z. Jiang, E. H. Conrad, C. Berger, C. Tegenkamp, W. A. de Heer, Exceptional ballistic transport in epitaxial graphene nanoribbons. Nature 506, 349 (2014). [6] V. P. Gusynin, S. G. Sharapov, Unconventional integer quantum Hall effect in graphene. Phys. Rev. Lett. 95, 146801 (2005). [7] D. A. Abanin, P. A. Lee, L. S. Levitov, Spin-filtered edge states and quantum Hall effect in graphene. Phys. Rev. Lett. 96, 176803 (2006). [8] G. Li, A. Luican-Mayer, D. Abanin, L. Levitov, and E. Y. Andrei, Evolution of Landau levels into edge states in graphene. Nature Commun. 4, 1744 (2013). [9] 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, 197 (2005). [10] Y. Zhang, Y.-W. Tan, H. L. Stormer, P. Kim, Experimental observation of quantum Hall effect and Berry’s phase in graphene. Nature 438, 201 (2005). [11] E. V. Castro, N. M. R. Peres, J. M. B. Lopes dos Santos, A. H. Castro Neto, and F. Guinea, Localized states at zigzag edges of bilayer graphene. Phys. Rev. Lett. 100, 026802 (2008). [12] K. S. Novoselov, E. McCann, S. V. Morozov, V. I. Fal’ko, M. I. Katsnelson, U. Zeitler, D. Jiang, F. Schedin, A. K. Geim, Unconventional quantum Hall effect and Berry’s phase of 2π in bilayer graphene. Nature Phys. 2, 177 (2006). 8 [13] E. McCann, V. Fal’ko, Landau-Level Degeneracy and Quantum Hall Effect in a Graphite Bilayer. Phys. Rev. Lett. 96, 086805 (2006). [14] F. Zhang, A. H. MacDonald, Distinguishing Spontaneous Quantum Hall States in Bilayer Graphene. Phys. Rev. Lett. 108, 186804 (2012). [15] B. J. LeRoy, M. Yankowitz, Emergent complex states in bilayer graphene. Science 345, 31-32 (2014). [16] A. Kou, et al. Electron-hole asymmetric integer and fractional quantum Hall effect in bilayer graphene. Science 345, 55-57 (2014). [17] K. Lee, et al. Chemical potential and quantum Hall ferromagnetism in bilayer graphene. Science 345, 58-61 (2014). [18] P. Maher, et al. Tunable fractional quantum Hall phases in bilayer graphene. Science 345, 61-64 (2014). [19] G. Li, A. Luican, E. Y. Andrei, Scanning Tunneling Spectroscopy of Graphene on Graphite. Phys. Rev. Lett. 102, 176804 (2009). [20] G. Li, E. Y. Andrei, Observation of Landau levels of Dirac fermions in graphite. Nat. Phys. 3, 623-627 (2007). [21] L.-J. Yin, et al. Tuning structures and electronic spectra of graphene layers with tilt grain boundaries. Phys. Rev. B 89, 205410 (2014). [22] L.-J. Yin, S.-Y. Li, J.-B. Qiao, J.-C. Nie, and L. He, Landau quantization in graphene monolayer, Bernal bilayer, and Bernal trilayer on graphite surface. Phys. Rev. B 91, 115405 (2015). [23] L.-J. Yin, J.-B. Qiao, W.-J. Zuo, W.-T. Li, and L. He, Experimental evidence for non-Abelian gauge potentials in twisted graphene bilayers. Phys. Rev. B 92, 081406(R) (2015). [24] G. Li, A. Luican-Mayer, D. Abanin, L. Levitov, E. Y. Andrei, Evolution of Landau levels into edge states in graphene. Nature Commun. 4, 1744 (2013). [25] W.-X. Wang, et al. Atomic resolution imaging of the two-components Dirac Landau levels in graphene monolayer. Phys. Rev. B 92, 165420 (2015). [26] G. M. Rutter, et al. Microscopic polarization in bilayer graphene. Nat. Phys. 7, 649-655 (2011). 9 [27] E. McCann, Asymmetry gap in the electronic band structure of bilayer graphene. Phys. Rev. B 74, 161403(R) (2006). [28] Y. Zhang, et al. Direct observation of a widely tunable bandgap in bilayer graphene. Nature 459, 20-823 (2009). [29] E. Castro, et al. Biased Bilayer Graphene: Semiconductor with a Gap Tunable by the Electric Field Effect. Phys. Rev. Lett. 99, 216802 (2007). [30] T. Ohta, A. Bostwick, T. Seyller, K. Horn, E. Rotenberg, Controlling the electronic structure of bilayer graphene. Science 313, 951-954 (2006). [31] K. F. Mak, C. H. Lui, J. Shan, T. F. Heinz, Observation of an Electric-Field-Induced Band Gap in Bilayer Graphene by Infrared Spectroscopy. Phys. Rev. Lett. 102, 256405 (2009). [32] Z. Q. Li, et al. Band Structure Asymmetry of Bilayer Graphene Revealed by Infrared Spectroscopy. Phys. Rev. Lett. 102, 037403 (2009). [33] J. B. Oostinga, H. B. Heersche, X. Liu, A. F. Morpurgo, L. M. Vandersypen, Gate-induced insulating state in bilayer graphene devices. Nature Mater 7, 151-157 (2008). [34] Y. Niimi, T. Matsui, H. Kambara, K. Tagami, M. Tsukada, and H. Fukuyama, Scanning tunneling microscopy and spectroscopy of the electronic local density of states of graphite surfaces near monoatomic step edges. Phys. Rev. B 73, 085421 (2006). [35] Y. Kobayashi, K.-i. Fukui, and T. Enoki, Edge state on hydrogen-terminated graphite edges investigated by scanning tunneling microscopy. Phys. Rev. B 73, 125415 (2006). [36] Y. Kobayashi, K.-i. Fukui, T. Enoki, K. Kusakabe, and Y. Kaburagi, Observation of zigzag and armchair edges of graphite using scanning tunneling microscopy and spectroscopy. Phys. Rev. B 71, 193406 (2005). [37] See Supplemental Material for more STM images, STS spectra, and the detail of the analysis. [38] W. Yao, S. Yang, and Q. Niu, Edge states in graphene: from gapped flat-band to gapless chiral modes. Phys. Rev. Lett. 102, 096801 (2009). 10 Figure captions: FIG. 1 (color online). (a) 100 nm × 100 nm STM topographic image of a bilayer graphene region on graphite surface (Vb = 0.2 V, I = 0.2 nA). Inset: (upper) Atomic resolution image of the graphene bilayer showing the triangular contrasting, which reflects only one of the two sublattices of the topmost graphene due to the inversion symmetry breaking in Bernal (AB-stacked) bilayers; (lower) Height profile along the black line shows the height difference of two steps ~ (0.70 ± 0.01) nm, which is slightly larger than the equilibrium spacing of the bilayer step (~ 0.67 nm). (b) Tunneling spectra of the graphene bilayers recorded away from the edges under various magnetic fields. LL peak indices are labeled (+/- are valley indices) and the data are offset in Y-axis for clarity. (c) The LL peaks energies extracted from (b) plotted versus the magnetic fields B. The solid curves are the fitting of the data with Eq. (1). 11 FIG. 2 (color online). Atomic resolution images of a zigzag bilayer edge (a) and an armchair bilayer edge (b). The insets are the the fast Fourier transforms (FFT) of the STM images. The outer hexangular spots and inner bright spots correspond to the reciprocal lattice of the graphene lattice and the interference of the scattering, respectively. (c) Schematic of the Bernal bilayer graphene with the zigzag and armchair edges. The green dots, representing a set of sublattices imaged in STM topography, can be used to determine the type of graphene edge. Energy spectrum of an unbiased (d) and biased (e) bilayer graphene with zigzag edges. The green curves correspond to the quasi-localized surface states of the two zigzag edges. (f) Energy spectrum of an unbiased (solid lines) and biased (dotted lines) bilayer graphene ribbon with armchair edges. 12 FIG. 3 (color online). (a) Spatial-resolved STS spectra recorded along the line drawn in panel (b) around the zigzag edge of bilayer graphene under 0 T. The DOS peaks marked by green arrows correspond to the quasi-localized surface states of the zigzag edge in graphene bilayer. The blue and red dashed lines label the positions of the VBE and the CBE in the bulk graphene bilayer. (c) The decay of the DOS peak height of the surface state obtained in panel (a). The green and gray curves correspond to the expected decaying behavior of the surface states in graphene bilayer and monolayer, respectively. 13 FIG. 4 (color online). (a) and (c) show spatial variation of the LL spectra measured at 7 T around the zigzag and armchair edges of bilayer graphene, respectively. The dashed lines indicate the energy positions of the LL(0,1,-) and LL(0,1,+) in the bulk of bilayer graphene. The blue and red arrows mark the spatially evolution of the LL(0,1,-) and LL(0,1,+) peaks. (b) and (d) show LL spectra maps at 7 T recorded around the zigzag and armchair edges, respectively. In panel (a) and (b), the peaks marked by green arrows correspond to the quasi-localized surface states of the zigzag edge. (e) Evolution of the peak positions and heights at 7 T with distance from the armchair edge on the conduction-band side. Inset shows LL peak heights extracted from (e) as a function of the distance from the edge. (f) LL bending as a function of distance around the armchair edge measured at 8 T. It shows an explicit shift of the energy positions for the LL(0,1,+) , LL2 and LL3 toward high energy with approaching the edge. The insets show calculated probability densities for the wave functions of the LL(0,1,+) , LL2 and LL3 at 8 T. (g) Shift length of the LL(0,1,+) and LL2 bending from the bilayer edges taken at different magnetic fields. The solid dots (empty dots) correspond to the data of armchair edges (zigzag edges). The dashed lines are the average values of ~ 14 1.4 lB and ~ 2.0 lB for the LL(0,1,+) and LL2, respectively. 15
1304.4969	1	1304	2013-04-17T20:53:08	Adiabatic quantum motors	[ "cond-mat.mes-hall" ]	When parameters are varied periodically, charge can be pumped through a mesoscopic conductor without applied bias. Here, we consider the inverse effect in which a transport current drives a periodic variation of an adiabatic degree of freedom. This provides a general operating principle for adiabatic quantum motors, for which we develop a comprehensive theory. We relate the work performed per cycle on the motor degree of freedom to characteristics of the underlying quantum pump and discuss the motor's efficiency. Quantum motors based on chaotic quantum dots operate solely due to quantum interference, motors based on Thouless pumps have ideal efficiency.	cond-mat.mes-hall	cond-mat	Adiabatic quantum motors Ra´ul Bustos-Mar´un,1 Gil Refael,2, 1 and Felix von Oppen1, 2 1Dahlem Center for Complex Quantum Systems and Fachbereich Physik, Freie Universitat Berlin, 14195 Berlin, Germany 2Department of Physics, California Institute of Technology, Pasadena, CA 91125, USA (Dated: October 30, 2018) When parameters are varied periodically, charge can be pumped through a mesoscopic conductor without applied bias. Here, we consider the inverse effect in which a transport current drives a periodic variation of an adiabatic degree of freedom. This provides a general operating principle for adiabatic quantum motors, for which we develop a comprehensive theory. We relate the work performed per cycle on the motor degree of freedom to characteristics of the underlying quantum pump and discuss the motors' efficiency. Quantum motors based on chaotic quantum dots operate solely due to quantum interference, motors based on Thouless pumps have ideal efficiency. Introduction. -- Popular culture has long been fasci- nated with microscopic and nanoscopic motors. Perhaps best known is the contest announced by Richard Feyn- man, who promised a $1000 prize to the developer of an engine that fits a cube of sides 1/64" [1]. While this feat was carried out shortly thereafter, in 1960, and did not produce an intellectual breakthrough, Feynman's contest has continued to provide tremendous inspiration to the field of nanotechnology. A prototypical nanomotor was unveiled in 2003, using tiny gold leaves mounted on multi- walled carbon nanotubes, with the carbon layers them- selves carrying out the motion [2]. The motor was driven through AC actuation, and basically relied on classical physics for their operation. As the dimensions of mo- tors are reduced, however, it is natural to expect that quantum mechanics could be used to operate and to op- timize nanomotors. In fact, cold-atom-based AC-driven quantum motors have been explored in Refs. [3, 4]. Nanomotors can also be actuated by DC driving [5 -- 7]. A general strategy towards realizing a DC nanoscale motor is based on operating an electron pump in reverse. Consider an electron pump in which the periodic varia- tion of parameters (such as shape, gate voltage, or tun- neling strength) originates from the adiabatic motion of, say, a mechanical rotor degree of freedom. To operate this pump as a motor, an applied bias voltage produces a charge current through the pump which, in turn, exerts a force on the mechanical rotor. The existence of quan- tum pumps [8, 9] suggests that by this operating princi- ple, quantum mechanics can be put to work in DC-driven nanomotors. Here, we develop a theory of such adiabatic quantum motors, expressing the work performed per cy- cle in terms of characteristics of the pump on which the motor relies and discussing the efficiency of quantum mo- tors in general terms. Our theory relies on progress in the understanding of adiabatic reaction (or current-induced) forces [6, 10 -- 13] which applies when the mechanical motor degree of free- dom is slow compared to electronic time scales and can be treated as classical. Conventionally, adiabatic reaction forces acting on the slow degree of freedom are considered for closed quantum systems [14]. This has recently been FIG. 1. Generic adiabatic quantum motors building on (a) a quantum pump based on a chaotic quantum dot and (b) a Thouless pump. When a voltage is applied to the pump, the current 'turns the wheel' and makes the phase angle θ wind. extended to situations where the fast degrees of freedom constitute a quantum mechanical scattering problem and thus to mesoscopic conductors [11 -- 13]. The resulting ex- pressions for the reaction forces in terms of the scattering matrix of the mesoscopic conductor allow one to explore the relations to quantum pumping in general terms. Before developing our general theory, we sketch two conceptual examples of adiabatic quantum motors in Fig. 1. One motor is based on a chaotic quantum dot operated as a pump [8, 9], as illustrated in Fig. 1(a). In this motor, the time-dependent gate voltages varying the shape of the quantum dot are provided by a periodic set of charges situated around the rim of a wheel which approach and modify the quantum dot in two locations. A current flowing through the quantum dot will then produce a rotation of the wheel. Alternatively, we could base a quantum motor on a Thouless pump. A schematic of such a motor is shown in Fig. 1(b). A single-channel quantum wire is located next to a conveyor belt with periodic attached charges (alternatively, a cogwheel with periodically spaced and electrically charged teeth). The charges induce a periodic potential in the quantum wire which slides as the conveyor belt or cogwheel turns. It is well-known from the seminal works of Thouless [15] that when the Fermi energy lies in an energy gap, such pumps transport integer amounts of charge per cycle (i.e., when the periodic potential slides by one period). An alternative physical realization of a Thouless motor is based on a helical wire in an electric field [7]. +-+-+-θ-+-+-++--+-+-+-+eθOutput power of adiabatic quantum motors. -- We start by deriving a general expression for the output power of an adiabatic quantum motor. The motor consists of a mesoscopic conductor with left (L) and right (R) lead, described within the independent-electron approximation by an electronic scattering matrix. In the adiabatic quan- tum motors of Fig. 1, the mesoscopic conductor is cou- pled to a single (classical) angle degree of freedom θ, as described through the dependence S(θ) of the S-matrix on the motor coordinate θ. More generally, the meso- scopic conductor could be coupled to several mechanical motor degrees of freedom Xν (ν = 1, 2 . . . N ) so that S = S(X). Retaining the dependence on several mode coordinates X for generality, the adiabatic reaction force F(X) on the motor degrees of freedom can be expressed in terms of the S-matrix of the mesoscopic conductor [11 -- 13], (cid:90) d (cid:88) 2πi α (cid:18) Fν(X) = fαTr ΠαS† ∂S ∂Xν (cid:19) Here, fα() denotes the Fermi distribution function in lead α = L, R with chemical potential µα and Πα is a projector onto the scattering channels in lead α. It has been shown [6, 10 -- 13] that this adiabatic reaction force need not be conservative when the electronic conductor is out of equilibrium. Thus, the work per cycle performed by this force is nonzero and given by Wout = dX · F(X). (2) and(cid:80) Note that in the absence of an applied bias, Wout = 0 (i.e., the force is conservative). In this case, fα() = f () α Πα = 1, and inserting Eq. (1) into Eq. (2) yields Wout = dX · ∇X f () Tr ln S() = 0. (3) (cid:73) The work performed by the adiabatic quantum motor per cycle is nonzero when a finite bias V is applied. In linear response, Eq. (1) yields dX · df(cid:48)()Tr[(ΠL − ΠR)S† ∂S ∂X Wout = ieV 4π (cid:73) (cid:90) ],(4) where we used that Wout = 0 in equilibrium and ex- panded Eq. (1) to linear order in the applied bias V . Using Brouwer's formula [8] which expresses the charge Qp pumped during one cycle of X in terms of the elec- tronic S-matrix S(X), the right-hand side of Eq. (4) can be identified as Wout = QpV. (5) Remarkably, the output of the nonequilibrium device is described by Qp, which characterizes the underlying quantum pump in equilibrium. Eq. (5) shows that the mechanical output of the motor per cycle originates from (cid:73) (cid:90) d 2πi . (1) η = Pout/Pin = Qp/Iτ. (7) 2 the fact that a charge Qp is pumped through the system with every revolution of the motor, and that this pumped charge gains an electrical energy QpV due to the applied bias. Thus, the average output power of the motor is Pout = QpV /τ, (6) where τ denotes the motor's cycle period. We also em- phasize that Eq. (5) identifies quantum pumping as the physical origin of the nonconservative nature of the adi- abatic reaction force in Eq. (1). Efficiency of adiabatic quantum motors. -- The applied bias V induces a slowly-varying DC charge current I in the adiabatic quantum motor. Thus, on average, opera- tion of the motor requires an input power of Pin = IV . (The overline denotes an average over a single cycle). The efficiency η of the adiabatic quantum motor is then naturally defined as the ratio of output to input power, Here, we have used Eq. (6) in the second equality. For adiabatic motor degrees of freedom, the current I is made up of two contributions: the pumped charge and the transport current induced by the applied bias V . If G(X) denotes the conductance of the device for fixed X, the linear-response current averaged over one cycle is I = G(X)V + Qp τ . (8) Note that the pumping current also depends on voltage through the motor's operating frequency (as character- ized by τ ). We note in passing that this expression can be obtained more formally, see Ref. [11]. With Eq. (8), the quantum motor's efficiency becomes η = 1 1 + GV τ /Qp . (9) Interesting conclusions can be drawn directly from this expression: (i) Quantum motors can operate entirely on the basis of quantum interference and become inef- fective due to phase-breaking processes, justifying the term quantum motor. A conceptually interesting exam- ple is the motor in Fig. 1(a) which is based on a chaotic quantum dot. It is well-known that the charge pumped through chaotic quantum dots (and hence the output power of the corresponding quantum motor) vanishes with increasing phase breaking. (ii) Quantum motors can have ideal efficiency η = 1, implying perfect conversion of electrical into mechanical energy. Indeed, this can be realized by motors based on Thouless pumps; when the Fermi energy lies in the gap, the conductance vanishes while the pumped charge is quantized to integer multi- ples of e. Thus, Eq. (9) yields η = 1, making Thouless pumps ideal adiabatic quantum motors. Motor dynamics. -- The output power of a quantum motor depends on its dynamics through the cycle period τ . Here, we discuss this for the simplest case, in which both the driving force and the load Fload acting on the angular motor degree of freedom θ are independent of the state of the motor. (This is realized for Thouless motors, but typically not for motors based on chaotic quantum dots.) If the motor degree of freedom is subject to damp- ing with damping coefficient γ, the steady-state velocity of the motor follows from the (classical) condition γ θ = QpV 2π − Fload. (10) Thus, we obtain for the cycle period of the motor τ = 2π/ θ = (2π)2γ/(QpV − 2πFload). We can use Eq. (8) to eliminate V in favor of the current I. This yields 1 τ = QpI − 2πFloadG Q2 p + (2π)2γG . (11) For an ideal Thouless motor with G = 0 [7], this yields the relation 1/τ = I/Qp. This is a direct consequence of the fact that in this case, the entire current passing the device must be due to pumping. More generally, this re- mains a good approximation as long as G (cid:28) Q2 p/(2π)2γ. This result also implies that the maximum load on the motor is given by F max load = QpI/2πG. Thouless motor. -- Thouless motors provide an instruc- tive example not only because they realize ideal quantum motors but also because they allow for a thorough analyt- ical discussion. Consider a single-channel quantum wire subject to a periodic potential of period a, as described by the Hamiltonian H = p2/2m + 2∆ cos(2πx/a + θ)Θ(L/2 − x) (12) The periodic potential of strength 2∆ acting for −L/2 < x < L/2 arises, e.g., from a periodic set of charges situ- ated along a conveyor belt or cogwheel so that the nearby electrons in the wire experience an electrostatic potential [cf. Fig. 1(b)]. This potential slides as the cogwheel turns and the mechanical variable θ varies by 2π as the teeth of the cogwheel advance by one spacing a. When the chemical potential µ is chosen such that the Fermi wavevector kF = (2mµ/2)1/2 is close to k0 = π/a, one can linearize the Hamiltonian for momenta close to ±k0. This results in an effective Hamiltonian H with counterpropagating linear channels and backscattering due to the periodic potential. Measuring momenta from ±k0 and energies from 2k2 H = vF pσz + ∆ (σx cos θ + σy sin θ) Θ(L/2 − x). (13) 0/2m, one has Here, the σi denote the Pauli matrices in the space of the counterpropagating channels. We do not include the real electron spin for simplicity. Within the linearized model, the adiabatic S-matrix S(θ) can be readily obtained analytically. We start with the transfer matrix M from x = L/2 to x = −L/2. Since the model is linear in momentum p, this can be done by analogy with the time-evolution operator in quantum mechanics which yields (cid:26) 3 (cid:27) (cid:33) . (cid:32) −ieiθ M = exp − iL vF σz [E − ∆(σx cos θ + σy sin θ)] . (14) This can be rewritten as M = cos λL− iσeff sin λL, where σeff = [Eσz − i∆ cos θσy + i∆ sin θσx]/[E2 − ∆2]1/2 and λL = (L/vF )[E2 − ∆2]1/2. Note that σ2 eff = 1. To obtain the S-matrix from the transfer matrix M , we first assume that there is only an outgoing wave on the right. Then, the wavefunction on the left is (iL, oL)T = M (oR, iR)T = (M11oR, M21oR)T , where i and o refer to the in- and outgoing waves, respectively. This immedi- ately implies that the transmission S21 is 1/M11, and the reflection S11 is M21/M11. Repeating the same argu- ments with only an outgoing wave on the left, we also find S22 = (M−1)12/(M−1)22 and S12 = 1/(M−1)22. With M11 = (M−1)22 = cos λL − i E√ E2−∆2 sin λL, this yields S = 1 M11 ∆√ E2−∆2 sin λL 1 −ie−iθ 1 ∆√ E2−∆2 sin λL (15) We can now use this S-matrix to obtain explicit expres- sions for the efficiency of the Thouless motor. Using the Landauer formula, the conductance for a Fermi energy EF takes the form G = G = e2 h ∆2 − E2 F ∆2 − E2 F cos2 λL + E2 F sin λL2 (16) In accord with the fact that the periodic potential opens a gap, the conductance is exponentially small in L for FIG. 2. Efficiency of the Thouless motor vs Fermi energy for L = 0.75µm and v = 105m/s. From top to bottom, the curves correspond to dissipative loads γ/ = 1/2π, 0.5, 1. The mo- tor has ideal efficiency (η = 1) when the Fermi energy lies in the gap, EF < ∆, and the length is taken to infinity. The inset shows the cycle frequency for a current-biased Thouless motor vs Fermi energy. 01230.00.51.0 0.51.00 1 2 3EF / ∆ 1/τ ( I / e )0.0ηEF / ∆EF < ∆ and becomes oscillatory and finite in L for EF > ∆. Similarly, we can obtain the pumped charge in the standard way from Brouwer's formula [8] (evalu- ated at zero temperature and for an angular degree of freedom), Qp = ∆2 − E2 e∆2 sin λL2 F cos2 λL + E2 F sin λL2 . (17) For Fermi energies in the gap, the charge pumped is quantized to e with exponential precision. When the Fermi energy is outside the gap, EF > ∆, the charge is no longer quantized and smaller than e. We can combine these results to obtain an explicit ex- pression for the efficiency of the Thouless motor. To do so, we note that the force acting on the motor is inde- pendent of θ. Thus, we can combine Eqs. (9), (11), (16), and (17) to obtain (for zero load, Fload = 0) 1 η = F −∆2 E2 ∆4 sin λL4 [E2 1 + 2πγ F − ∆2 cos2 λL + E2 F sin λL2] (18) In Fig. 2, we plot the efficiency of the Thouless motor as a function of the Fermi energy. As can be seen from Eq. (18), the efficiency is exponentially close to unity when the Fermi energy is within the gap. For this range of Fermi energies, the Thouless motor is an ideal adiabatic quantum motor. When the Fermi energy moves out of the energy gap, the efficiency is oscillatory with an al- gebraically dropping amplitude. In this regime, Fabry- Perot interference alone produces peaks in the efficiency, which appear when the reflection coefficients are maxi- mal. The inset of Fig. 2 also shows the cycle frequency of the Thouless motor, for a given current and zero load, as a function of Fermi energy, cf. Eq. (11). Intrinsic damping. -- So far, we have treated the damp- ing coefficient γ of the motor degree of freedom as phe- nomenological. However, in addition to extrinsic, purely mechanical friction, there is a contribution to γ which arises intrinsically from the coupling to the electronic system. As shown recently, this intrinsic damping γint can also be obtained from the electronic S-matrix [11 -- 13]. Restricting attention to small bias voltages, we can approximate γint by its equilibrium value, γint = (/4π)tr[(∂S†/∂θ)(∂S/∂θ)]. This is readily evaluated for the Thouless motor when the Fermi energy is in the vicin- ity of the fundamental gap. We find that the intrinsic damping can be expressed in terms of the pumped charge, γint = (/2πe)Qp. Quite surprisingly, the electronic sys- tem induces finite mechanical damping even when the Fermi energy lies in the gap (and Qp = e). We interpret this damping as arising from forming plasmon excita- tions in the leads when pumped charge enters or leaves. When the Fermi energy of a current-biased Thouless mo- tor lies inside the fundamental gap, the motor (without load) rotates at angular frequency ω = 2πI/e, which, 4 from Eq. (10), gives a friction-induced voltage drop of V = I(2π)2γ/e2. The existence of the intrinsic friction implies that for a given current, there is a minimal volt- age of V = (h/e2)I at which the Thouless motor de- scribed by Eq. (13) can operate. At first sight, the intrinsic damping may seem to negate the possibility of an ideal quantum motor when the motor is subject to a load. Indeed, the electrical input power is then split between the power consumed θ, and the power dissipated by the load, Pload = Fload by damping, Pγ = γ θ2. Nevertheless, for a quantized θ = 2πI/Qp, so that Pload ∝ I while Thouless pump, Pγ ∝ I 2. Hence, the power dissipated by damping be- comes negligible at small currents, and the load efficiency ηload = Pload/Pin can be made arbitrarily close to unity by operating the motor at low currents. Conclusions. -- Motion at the nanoscale tends to be dominated by fluctuations. It is an important chal- lenge to develop schemes to generate directed motion in nanoscale devices [16 -- 18]. Here, we investigated a gen- eral strategy to this effect which is based on operating quantum pumps in reverse. We developed a correspond- ing theory which expresses the output power and the effi- ciency of such adiabatic quantum motors to characteris- tics of the pumps on which they are based. The concept of adiabatic quantum motors offers numerous possibilities for future research. Interesting directions include motors based on electron pumps which involve electron-electron interactions as well as systems in which the motor degree of freedom is itself quantum mechanical. We acknowledge discussions with P. Brouwer as well as support by the Deutsche Forschungsgemeinschaft through SFB 658, the Humboldt Foundation through a Bessel Award, the Packard Foundation, and the Institute for Quantum Information and Matter, an NSF Physics Frontiers Center with support of the Gordon and Betty Moore Foundation. [1] R.P. Feynman, Engineering and Science 23, 22 (1960). [2] A.M. Fennimore, T.D. Yuzvinsky, W.-Q. Han, M.S. Fuhrer, J. Cumings, and A. Zettl, Nature 424, 408 (2003). [3] A.V. Ponomarev, S. Denisov, and P. Hanggi, Phys. Rev. Lett. 102, 230601 (2009). [4] T. Salger, S. Kling, T. Hecking, C. Geckeler, L. Morales- Molina, and M. Weitz, Science 326, 1241 (2009). [5] S.W.D. Bailey, I. Amanatidis, and C.J. Lambert, Phys. Rev. Lett. 100, 256802 (2008). [6] D. Dundas, E.J. McEniry, and T.N. Todorov, Nature Nanotech. 4, 99 (2009). [7] X.-L. Qi and S.C. Zhang, Phys. Rev. B 79, 235442 (2009). [8] P.W. Brouwer, Phys. Rev. B 58, R10135 (1998). [9] M. Switkes, C.M. Marcus, K. Campman, A.C. Gossard, Science 283, 1905 (1999). [10] J.T. Lu, M. Brandbyge, and P. Hedegard, Nano Lett. 10, 1657 (2010). [11] N. Bode, S. Viola Kusminskiy, R. Egger, F. von Oppen, Phys. Rev. Lett. 107, 036804 (2011). [12] N. Bode, S. Viola Kusminskiy, R. Egger, F. von Oppen, Beilstein J. Nanotechnol. 3, 144 (2012). [13] M. Thomas, T. Karzig, S. Viola Kusminskiy, G. Zarand, F. von Oppen, Phys. Rev. B 86, 195419 (2012). [14] M. V. Berry, in Geometric Phases in Physics, edited by A. Shapere and F. Wilczek (World Scientific, Singapore, 1989). [15] D.J. Thouless, Phys. Rev. B 27, 6083 (1983). 5 [16] T. Kudernac, N. Ruangsupapichat, M. Parschau, B. Ma- cia, N. Katsonis, S.R. Harutyunyan, K.-H. Ernst, and B.L. Feringa, Nature 479, 208 (2011). [17] H.L. Tierney, C.J. Murphy, A.D. Jewell, A.E. Baber, E.V. Iski, H.Y. Khodaverdian, A.F. McGuire, N. Kle- banov, and E.C.H. Sykes, Nature Nanotech. 6, 625 (2011). [18] U.G.E. Perera, F. Ample, H. Kersell, Y. Zhang, G. Vives, J. Echeverria, M. Grisolia, G. Rapenne, C. Joachim, and S.-W. Hla Nature Nanotech. 8, 46 (2013).
1106.2982	2	1106	2011-10-21T08:02:02	Transport Properties of Clean Quantum Point Contacts	[ "cond-mat.mes-hall" ]	Quantum point contacts are fundamental building blocks for mesoscopic transport experiments and play an important role in recent interference- and fractional quantum Hall experiments. However, it is not clear how electron-electron interactions and the random disorder potential influence the confinement potential and give rise to phenomena like the mysterious 0.7 anomaly. Novel growth techniques of GaAs/AlGaAs heterostructures for high-mobility two-dimensional electron gases enable us to investigate quantum point contacts with a strongly suppressed disorder potential. These clean quantum point contacts indeed show transport features that are obscured by disorder in standard samples. From this transport data, we are able to extract the parameters of the confinement potential which describe its shape in longitudinal and transverse direction. Knowing the shape (and hence the slope) of the confinement potential might be crucial to predict which interaction-induced states can form in quantum point contacts.	cond-mat.mes-hall	cond-mat	Transport Properties of Clean Quantum Point Contacts C. Rossler Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland E-mail: [email protected] S. Baer Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland E. de Wiljes Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland P.-L. Ardelt Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland T. Ihn Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland K. Ensslin Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland C. Reichl Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland W. Wegscheider Solid State Physics Laboratory, ETH Zurich, 8093 Zurich, Switzerland Abstract. Quantum point contacts are fundamental building blocks for mesoscopic transport experiments and play an important role in recent interference- and fractional quantum Hall experiments. However, it is not clear how electron-electron interactions and the random disorder potential influence the confinement potential and give rise to phenomena like the mysterious 0.7 anomaly. Novel growth techniques of AlXGa1−XAs heterostructures for high-mobility two-dimensional electron gases enable us to investigate quantum point contacts with a strongly suppressed disorder potential. These clean quantum point contacts indeed show transport features that are obscured by disorder in standard samples. From this transport data, we are able to extract the parameters of the confinement potential which describe its shape in longitudinal and transverse direction. Knowing the shape (and hence the slope) of the confinement Transport Properties of Clean Quantum Point Contacts 2 potential might be crucial to predict which interaction-induced states can form best in quantum point contacts. PACS numbers: 72.20.-i, 72.25.Dc, 73., 75.76.+j Transport Properties of Clean Quantum Point Contacts 3 1. Introduction Quantum devices on semiconductor nanostructures rely on quantum point contacts (QPCs) as basic building blocks. Quantized conductance has been observed early on [1, 2] and it has been used as a signature of the quality of a QPC. With ever improving sample quality and the perspective for the detection of non-abelian anyons in the ν = 5/2 fractional quantum Hall state, several experiments [3, 4, 5] have recently used the properties of QPCs fabricated on ultra high-mobility two-dimensional electron gases (2DEGs). In view of proposals to investigate fractional quantum Hall states in confined geometries and interferometer-like setups, the detailed understanding and control of QPCs are essential. Here we present experimental data which go beyond previously published data by demonstrating experiments that profit from the extraordinary cleanliness of the high-mobility 2DEG. In contrast to standard 2DEGs, we do not observe defect-induced resonances when the QPCs are shifted laterally. Higher order half plateaus are observed in the finite-bias differential conductance (at magnetic field B⊥ = 0 T) as well as spin-split half plateaus at B⊥ = 2 T. Finally, the 0.7-anomaly is investigated as a function of temperature and in perpendicular magnetic field. 2. Experiment The samples are fabricated on a high-mobility wafer with a two-dimensional electron gas (2DEG) residing z = 160 nm beneath the surface. The high mobility is achieved by placing Si dopants in a narrow GaAs layer sandwiched by AlAs layers [6, 7, 8]. The population of the X band in AlAs results in hardly mobile electrons which screen the static disorder potential but do not cause a measurable parallel conductance. Optical lithography is employed to define Hall bars via mesa etch and deposition of Au/Ge Ohmic contacts. Processed Hall bars have an electron density of nS = 3.5 × 1015 m−2 and Drude mobilities in the range of µ = 1000...2000 m2/Vs. The characterization as well as the following experiments are carried out at a temperature of T = 1.3 K, if not stated otherwise. Schottky electrodes are defined via electron beam lithography and subsequent deposition of Ti/Au. AFM micrographs of two QPCs are shown in the insets of Figs. 1a) and b). The gates appear bright with the gap between them being w = 200 nm (a) and w = 500 nm (b). Applying a voltage of VG (cid:46) −1.1 V to the gates depletes the underlying 2DEG and creates a constriction between source and drain. The source-drain current ISD and the voltage drop across the QPC VD are measured in four-terminal configuration while applying a small lock in amplitude of VAC = 100 µV at a frequency of fAC = 33 Hz to source and drain. A DC source-drain voltage VSD can be added to VAC with both voltages being applied symmetrically with respect to the common reference potential of source, drain and the gates. Most transport properties of the employed high-mobility heterostructures are hysteretic as a function of gate bias [9]. Therefore all traces are recorded in the same sweep direction by sweeping towards more negative values of gate voltage. Transport Properties of Clean Quantum Point Contacts 4 Figure 1. a) Inset: Atomic force micrograph of the sample surface. Two Schottky- gates appear bright, the GaAs surface appears dark. The distance between the gates is w = 200 nm. When the gates are negatively biased, free electrons reside only in the electron gas underneath the dark areas. Main graph: differential conductance of QPC1, measured as a function of the voltage applied to gates G1 and G2. Quantized conductance in multiples of G = 2e2/h indicates the formation of discrete subbands between the tips of the gates. b) Differential conductance of QPC2, which is w = 500 nm wide and l ≈ 1 µm long. Figure 1a) shows the differential conductance G = dISD/dVD (VSD = 0 mV) of QPC1, plotted as a function of the voltage applied to gates G1 and G2. From the Fermi wavelength of the 2DEG λF =(cid:112)2π/nS = 42 nm and the distance of the gates, it would be expected that n ≈ w/(λF/2) = 9...10 modes can be observed due to confinement transversal to the electron flow [2]. Indeed, the number of quantized plateaus in Fig. 1a) agrees with this estimation, indicating that the largest electronic width of the QPC matches the lithographic gap of the Schottky split-gates. Due to the larger gate-spacing of QPC2, correspondingly more modes are observed in Fig. 1b) and a significantly larger gate bias has to be applied in order to pinch off. We observe irreversible charging of the sample typically at VG ∼ −5 V which limits the QPCs' range of operation. 2.1. Lateral Shifting of the QPC QPC1 can be further characterized by varying the voltages applied to each of the gates, which is not possible for QPC2 due to its extreme pinch-off voltage. Figure 2 shows the transconductance GTC = dG/dVG1&G2 of QPC1 in grayscale, plotted as a function of VG1 and VG2. Black areas correspond to pinch-off (bottom left) and successive conductance plateaus (marked by 1, 2, 3). Such a plot reveals scattering centers in the channel, since changing the ratio of gate voltages causes the position of the channel a) b) QPC1 T = 1.3 K VSD = 0 mV VG1&G2 (V) G (2e2/h) -1.2 -1.4 -1.6 -1.8 -2.0 -2.2 0 5 10 15 QPC2 T = 1.3 K VSD = 0 mV VG1&G2 (V) G (2e2/h) -1.5 -2.0 -2.5 -3.0 -3.5 -4.0 0 10 15 25 -4.5 20 5 G1 G2 200 nm G1 G2 500 nm ISD Transport Properties of Clean Quantum Point Contacts 5 Figure 2. (color online) Transconductance GTC = dG/dVG1&G2 of QPC1, plotted in false colors as a function of the voltages applied to gates G1 and G2. Integer conductance values of G = 0, 1, 2, 3, ... × 2e2/h result in GTC = 0 (black), steps in-between integer conductance values appear bright. A faint red stripe with G ≈ 0.7 × 2e2/h indicates the presence of a 0.7-anomaly in the QPC. The strong increase of the conductance at the right and top border (white) marks the gate-pinch-off, where the 2DEG underneath gates G1 (right) and G2 (top) starts to be depleted. If both gates are biased identically, 12-13 plateaus are observed between QPC-pinch- off and gate-pinch-off. When the ratio of gate-voltages is varied by following the lowest transconductance-stripe, the number of plateaus in dependence of either VG1 or VG2 can be varied, indicating that the QPC is shifted laterally between the gates. Scattering centers in-between the gates would appear as straight lines with the slope corresponding to their capacitance to gates G1 and G2. No such defects are visible in this scan. to shift laterally between the gates [10]. The shift can be approximately determined by counting the number of steps (nG1, nG2) that can be observed as a function of each gate [11]: ∆y = λF/2 × (nG1 − nG2)/2. In the situation marked by white arrows in Fig. 2 this amounts to ∆y = 21 nm × (9 − 4)/2 = 53 nm. The largest observed shift is ∆y = ±21 nm × (11 − 2)/2 = ±95 nm, which corresponds to the lithographic distance of the gates. When the QPC is shifted from one gate to the other, its potential would naturally change if localized impurities in the channel or the static disorder potential created by the dopants were relevant. Since these defects are fixed in space, they should appear as lines intersecting the QPC steps, with their slope given by the capacitance to gates G1 and G2. The absence of such defect-induced lines confirms the cleanliness of the sample and the effectiveness of the screening layers in suppressing the charged dopants' disorder potential. 2.2. The 0.7-Anomaly in Perpendicular Magnetic Field It is noteworthy that the 0.7 anomaly [12, 13, 14], an additional plateau with a conductance of G ≈ 0.7 × 2e2/h, appears as a weak shoulder close to pinch-off of VG1 (V) VG2 (V) -4 -3 -2 -1 -1 -4 -3 -2 dG/dVG1&G2 (a.u.) 0 1 QPC1 T = 1.3 K 0 1 0.7 × 2e2/h 2 3 nG1 nG2 Transport Properties of Clean Quantum Point Contacts 6 Figure 3. (color online) Pinch-off traces for different positions of the QPC. The traces have been shifted and linearly scaled along the gate-axis for better comparison. From left to right: simultaneously sweeping VG1 = VG2; sweeping VG2 while VG1 = −1.7 V; sweeping VG1 while VG2 = −1.7 V. The 0.7-anomaly appears as a shoulder (change of slope dG/dVG1&G2 - see inset) at G ≈ 0.65 × 2e2/h. Switching events in the two rightmost traces arise because the more negative gate voltage has to be biased with VG (cid:46) −3 V in order to pinch off. In the transconductance-plot in Fig. 2, the 0.7 anomaly is visible as both QPCs. an asymmetry of the pinch-off line, giving rise to a gray (red) stripe adjacent to the G = 1 × 2e2/h plateau. The 0.7-stripe is continuous and reaches all the way to the extreme QPC shifts, emphasizing that the 0.7-anomaly is an intrinsic property of the QPC. Cuts along the diagonal (where VG1 = VG2 as well as strongly shifted configurations (either VG1 or VG2 being fixed at −1.7 V) are shown in Fig. 3. When the QPC is defined centrally in-between the gates (VG1 = VG2, leftmost trace), the 0.7- anomaly manifests itself as a weak shoulder below the G = 2e2/h-plateau. The top left inset shows the numerically derived slope dG/dVG1&G2 which exhibits a clear change of slope at the position marked by an arrow. The corresponding conductance at this gate voltage is G = 0.65 × 2e2/h. For comparison, two configurations with the QPC being defined close to gate G2 (central trace) or gate G1 (rightmost trace) are plotted on the same gate axis. We find that both asymmetrically measured traces resemble the shape of the symmetric case. The conductance value of the 0.7-anomaly does not change when shifting the QPC laterally, however our accuracy of determining it is limited to G = 0.65±0.05×2e2/h due to switching events caused by the more negative gate voltages required for pinch-off in an asymmetric gate configuration. In agreement with previous studies [12, 13, 14], we find that the 0.7-anomaly is less pronounced in 2DEGs with high density (here: nS = 3.5 × 1015 m−2) compared to samples with electron densities in the range of nS ∼ 1 × 1015 m−2. In order to compare the results obtained on low-density 2DEGs to the behavior of dG/dVG1&G2(a.u.) 1 0 G (2e2/h) VG1+VG2 (shifted and scaled) QPC1 T = 1.3 K VSD = 0 mV Transport Properties of Clean Quantum Point Contacts 7 Figure 4. (color online) Transmission νQPC through QPC1 measured as a function of symmetrically applied gate bias at different temperatures (from left to right: T = 1.3 K/2.5 K/5 K/10 K/15 K). The traces are plotted with horizontal offset for clarity. A dashed horizontal line marks the transmission value of νQPC = 1.3 associated with the 0.7-anomaly. a) The 0.7-anomaly becomes more pronounced with increasing temperature. At T = 10 K the subband quantization is completely washed out but the 0.7-anomaly is still clearly visible. b) In a magnetic field of B⊥ = 2 T applied perpendicular to the plane of the 2DEG, the shoulder at νQPC = 1.3 is well developed at low temperature and shifts to lower transmission with increased temperature. c) At B⊥ = 3 T, the transmission shows a non-monotonic behavior as a function of temperature. The transmission of the 0.7-anomaly first recovers almost to its zero- field value, then decreases again for T > 5 K. d) At B⊥ = 5 T, various plateaus related to the transmission of (fractional) edge channels are observed. These features wash out when the temperature is increased. our system, the temperature- and magnetic-field-dependence is investigated in detail. However, similar to recent work performed on comparable high-mobility 2DEGs [15], we find a suppression of the Hall mobility in parallel magnetic field which is in our devices accompanied by a suppression of the QPCs' spin splitting (data not shown). We are hence limited to applying a magnetic field perpendicular to the 2DEG, which should also weaken the 0.7-anomaly by lifting the spin degeneracy. Since the differential conductance is strongly modified by the presence of edge channels in the quantum Hall regime, the filling factor νQPC is obtained from the diagonal voltage drop across QPC1. Figure 4a) shows the lower part of the pinch-off trace for different temperatures without a magnetic field being applied. As expected, the 0.7-anomaly evolves into a more pronounced shoulder when the temperature is increased from T = 1.3 K (left) to T = 15 K (right). The marked value of νQPC = 1.3 (dashed line) corresponds to the a) b) c) d) QPC1 B┴ = 0 T VSD = 0 mV QPC1 B┴ = 3 T VSD = 0 mV QPC1 B┴ = 5 T VSD = 0 mV VG1&G2 (V) νQPC 2 0 1 νQPC 2 0 1 T = 1.3 K T = 15 K QPC1 B┴ = 2 T VSD = 0 mV T = 1.3 K T = 15 K T = 1.3 K T = 15 K T = 1.3 K T = 15 K -2.3 -2.2 VG1&G2 (V) -2.2 -2.0 VG1&G2 (V) -2.2 -1.8 VG1&G2 (V) -2.0 -1.5 -2.0 Transport Properties of Clean Quantum Point Contacts conductance of G = 0.65 × 2e2/h extracted from Fig. 3. 8 By applying a magnetic field perpendicular to the 2DEG, we expect the 0.7-anomaly to be influenced by the increase of both the energetic and spatial separation of the lowest two spin channels. This idea of "mimicking the 0.7 scenario" was previously investigated in [16], but the interpretation of the data proved difficult due to additional resonances in the pinch-off traces. Figure 4b) is recorded at B⊥ = 2 T, where spin-resolved edge channels begin to form at T = 1.3 K. The data shows a well pronounced plateau at νQPC = 1.3 which is weakened when the temperature is increased to T > 5 K. Figure 4c) shows the same measurement at B⊥ = 3 T, where the edge channels are further separated energetically as well as spatially. Now, the temperature dependence is non-monotonic with the 0.7-anomaly first rising almost to the expected transmission of νQPC = 1.3, then decaying to lower transmission. Data taken at B⊥ = 5 T is shown in Fig. 4d). The structure is more complex now due to the formation of fractional edge channels and does not show a feature which is unambiguously related to the 0.7-anomaly. The observed shoulders and plateaus wash out, perhaps with the plateau at νQPC ≈ 0.7 being more resilient than all other features at νQPC < 2. The overall dependence of the 0.7- anomaly on magnetic field is in contrast to observations in two-dimensional hole gases where the 0.7-anomaly was found to evolve into a resonance for strong perpendicular magnetic field [17]. There, the appearance of a resonance was discussed in view of a quasi-localized state in combination with Kondo effect. One possible interpretation of our magnetic field dependence follows the idea of two spin-polarized channels leading to the 0.7 scenario [16]. Moderate magnetic field (B⊥ = 2 T) increases the spin polarization, thereby enhancing the 0.7-anomaly. Stronger fields increase the spatial separation between the edge channels, thereby reducing interactions and weakening the 0.7-anomaly. At B⊥ = 3 T, the spatial separation can be overcome by increasing the temperature to a value where thermal energy and B-field- induced spin splitting become comparable in magnitude. At even stronger magnetic field, other many-body effects besides the 0.7-anomaly might become relevant which makes a detailed interpretation difficult. For future studies it might prove worthwhile to investigate the zero-bias anomaly in perpendicular magnetic field in order to check if the interpretation of spatially separated edge channels is consistent with other experimental findings. 2.3. Finite Bias Spectroscopy Further characterization of QPC1 requires finite-bias measurements, because employing VSD as an energy reference gives access to the QPC's subband spacings [18, 19, 20, 21]. Three exemplary gate traces are depicted in Fig. 5. Figure 5a) shows the linear-response regime VSD = 0 mV which is identical to the trace shown in Fig. 1a). A sketch of the energy landscape is shown at the right hand side. The parabolic electron dispersions are energetically separated by the subband spacing ∆SB due to transversal confinement. In the depicted situation, three subband-bottoms reside below the chemical potentials of Transport Properties of Clean Quantum Point Contacts 9 Figure 5. (color online) Pinch-off trace (left) and schematic view (right) for different values of the source-drain bias VSD. a) Linear-response regime VSD = 0 mV. The measured trace displays plateaus at the expected conductance values. The conductance value of G = 3 × 2e2/h corresponding to the sketched situation (right) is marked by a dashed line. In the depicted situation the gate voltage is set such that three (spin degenerate) subband bottoms lie below the chemical potential of source (µS) and drain (µD). The energies of the subband-bottoms are labeled E1, E2,... b) Finite bias measurement with one subband bottom in-between µS and µD. The expected half- plateau conductance of G = 2.5× 2e2/h is marked by a dashed line in the experimental trace. c) Two subband bottoms reside in-between µS and µD. Integer plateau values are expected and can be observed as shoulders in the pinch-off curve. source and drain, giving rise to a conductance of G = 3 × 2e2/h. Figure 5b) shows the pinch-off trace for VSD = −2.6 mV, where plateaus appear at half-integer values of the conductance. The sketch corresponding to a conductance of G = 2.5 × 2e2/h is shown on the right hand side: two subbands contribute fully and one subband contributes half to the overall conductance. So-called half-plateaus can only be observed in clean samples, presumably because scattering events inside the QPC become more likely when more unoccupied subband states are energetically available at larger source-drain bias. At even higher bias, the conductance is usually obscured by noise [22] or it increases/decreases due to various self- gating effects [13, 20]. In QPC1 however, the return of integer conductance quantization for two subband bottoms residing in-between µS and µD is observable in Fig. 5c) at VSD = −5.5 mV. We interpret this observation as another result of the cleanliness of the QPC which reduces the probability for backscattering. In order to retrieve the full information about the confinement potential, the transconductance of QPC1 is plotted in Fig. 6a) as a function of VSD and VG1&G2. Integer conductance plateaus without subband minima between µS and µD (labeled 1, 2, 3), appear as black diamonds around VSD = 0 mV. Half plateaus (1.5, 2.5, 3.5) and second order integer plateaus (2, 3, 4) appear in a regular pattern at finite source-drain a) b) µS µD QPC1 4 0 G (2e2/h) G = 3×2e2/h G = 2.5×2e2/h VSD = -2.6 mV VG1&G2(V) -1.25 -2.4 G = 2×2e2/h VSD = -5.5 mV 4 0 G (2e2/h) 4 0 G (2e2/h) µS µD µS c) ΔSB VSD = 0 mV E1 E2 E kx -eVSD µD Transport Properties of Clean Quantum Point Contacts 10 Figure 6. (color online) a) Transconductance GTC = dG/dVG1&G2 of QPC1, plotted as a function of source-drain bias VSD and voltage VG1&G2 applied to gates G1 and G2. Plateaus in the differential conductance of G = 1, 2, 3, ... × 2e2/h appear as black diamonds centered around VSD = 0 mV. Their extent in VSD corresponds to the subband-spacing ∆SB. Higher order half-plateaus (G = 1.5, 2.5, 3.5, ... × 2e2/h) and second-order integer-plateaus (G = 2, 3, 4, ... × 2e2/h) appear as black diamonds at finite source-drain bias. b) Transconductance of QPC2. Integer plateaus around VSD = 0 mV are resolved. For gate voltages VG1&G2 (cid:46) −2.5 V higher order plateaus are obscured by noise. bias. In our experience, the higher order plateaus can not be observed in samples with mobility µ (cid:46) 10 m2/Vs (cf. [22]). Comparing our data to data of a defect-free QPC [21] defined in a 2DEG with mobility µ = 150 m2/Vs, we find subtle differences in the transconductance pattern. Although the QPC in reference [21] is measured at a lower temperature (T = 90 mK) than our device (T = 1.3 K), second order integer plateaus seem to be suppressed as long as no magnetic field is applied perpendicular to the 2DEG. a) VG1&G2 (V) VSD (mV) -12 -6 0 6 -1.3 -2.3 -2.1 -1.9 -1.7 -1.5 dG/dVG1&G2 (a.u.) 0 1 QPC1 T = 1.3 K VG1&G2 (V) VSD (mV) -12 -6 0 6 -1.5 -2.5 -3.5 -4.5 QPC2 T = 1.3 K b) VSD= ΔSB dG/dVG1&G2 (a.u.) 0 1 1 2 1.5 2.5 2 3 3 3.5 4 Transport Properties of Clean Quantum Point Contacts 11 The authors state that "Due to the suppression of backscattering in the presence of a small magnetic field the reappearance of the integer plateaus at high VSD can be clearly observed". Our QPC is defined in a 2DEG with mobility µ (cid:38) 1000 m2/Vs and second order integer plateaus are clearly resolved at B⊥ = 0 T. These observations indicate that even though the mean free path of the electrons is much larger than the length of the QPC in both cases, a higher electron mobility still manifests itself as reduced backscattering in the regime of nonlinear conductance. As seen from the sketches in Fig. 5, the maximum extent of the diamonds in VSD corresponds to the energy spacing ∆SB of the involved QPC subbands. The whole pattern of transconductance diamonds is sheared with features at positive VSD shifted by about 5 % to more positive gate voltage than in a perfectly symmetric configuration. This asymmetry could hint at a slight asymmetry of the QPC's coupling to source and drain but might also be explained by a gradual drift of the local potential over the measurement time of 21 hours. The effect of VSD onto the confinement potential, the so-called self-gating, manifests itself as a deviation from a pattern of straight lines [13]. From Fig. 6a) it appears that self-gating plays an important role mainly close to pinch-off (white dashed line) and perhaps at very large VSD, where no clear quantization is observed any more. Since self-gating appears not to dominate the shape of the transconductance pattern it is possible to learn more about the confinement potential by comparing the position of transconductance nodes in Fig. 6a). Three exemplary nodes are highlighted by white circles. They correspond to the resonance conditions (from top to bottom) (µS = E4, µD = E6), (µS = E5 = µD) and (µS = E6, µD = E4), respectively. The fact that these resonance conditions occur at almost the same gate voltage (along a straight line) means that the subband-spacings ∆45 = E5 − E4 and ∆56 = E6 − E5 are very similar at this gate voltage (compare sketches in Fig. 5). Therefore the transversal confinement can be well described by a harmonic potential. If the confinement potential were for example a square well, the subband spacings would increase with higher mode number and hence the higher order modes would occur at more positive gate voltage than the linear response node. As a comparison to the clean and regular pattern of QPC1, Figure 6b) shows the transconductance of QPC2. Higher order plateaus are visible for VG1&G2 (cid:38) −2.5 V but are obscured by noise at more negative bias which is usually related to tunneling events from the gates into the doping layer [23]. Furthermore, the shape of the five leftmost diamonds is distorted with the upper and lower tip being shifted to more negative gate bias. This shift as well as the curvature of the plateau borders follows the dependence observed in quantum wires and arises from the requirement of satisfying charge neutrality with a 1-dimensional density of states while applying a finite source- drain voltage [24]. The quantum wire-like characteristic is consistent with the geometry of the gates which should create a channel that is longer than the screening length of the 2DEG (compare inset of Fig. 1b)). It is noteworthy that also in the quantum wire-like QPC2, we observe a well pronounced half-plateau related to the 0.7-anomaly Transport Properties of Clean Quantum Point Contacts 12 which resembles the features observed in [24]. Since we do not observe defect-related resonances in QPC2, the design might be extended to even longer gate-defined quantum wires [25, 26, 27] in order to study the length-dependence of the 0.7-anomaly. However, the observation of diffusive transport [26] in LQWR ≥ 5 µm long quantum wires (with the mean free path in the 2DEG being LMF ∼ 40 µm) suggests that gate-defined quantum wires might not profit from an increased free electron mobility at least if split-gate technology is used for confinement. 2.4. Extracting the QPCs' Shape Parameters As discussed earlier, the transconductance plot can now be used to reconstruct the confinement potential. The subband spacing can be determined from the VSD position of the borders of the transconductance diamonds [20]. Due to the finite resistance of the leads RS = 400 Ω, a fraction of the applied DC-bias VSD does not drop at the QPC. Using RS and the measured four-terminal conductance G, this is taken into account via ∆SB = VSD/(1 + GRS). The thereby determined subband spacings are plotted as a function of VG1&G2 in Fig. 7a) for QPC1 (left) and QPC2 (right). The subband spacings increase monotonically with more negative gate bias, indicating that the confinement potential becomes narrower and steeper while approaching pinch-off. This trend has been observed before [20] and can be explained by the reduced influence of screening on the confinement potential when the local electron density is reduced [28]. Since the higher order plateaus indicate that the confinement potential of QPC1 has a close-to harmonic shape, we can now use the measured subband spacings to apply Buttiker's saddle-point model [29] to our linear response data and extract all parameters of the potential profile at the constriction. Since QPC2 shows quantum wire-like transport characteristics, the model is not expected to reflect the exact potential shape of QPC2, but should still give qualitatively meaningful results. Temperature broadening is not accounted for in this model since the subband-spacings ∆SB > 1 meV are much larger than the thermal broadening kBT ≈ 0.1meV. Neglecting inter-mode scattering and including spin degeneracy, the transmission of the n-th subband is given by Tn = 2/(1 + exp(−πεn/ωX)) with the energy of the n-th subband εn = 2(ωY (n + 1/2) − ECB). (1) (2) The gate dependence ωY (VG1&G2) is known from Fig.7a), the gate dependence of the conduction band bottom ECB(VG1&G2) is approximated by ECB(VG1&G2) = E0 + α × VG1&G2 for each conductance step with the lever arm α converting gate voltage to energy. Usually, the lever arm can be determined by taking the source-drain bias as an energy reference and comparing it to the gate dependence of a given transport resonance [30]. But as shown in Fig.7a), the gate voltage not only lifts ECB but also increases the subband spacing, giving rise to a seemingly increased lever arm. Knowing that the Transport Properties of Clean Quantum Point Contacts 13 Figure 7. (color online) a) Subband spacings ∆SB of QPC1 (left) and QPC2 (right), as determined from finite bias transport measurements. With more negative gate voltage, the subband spacings increase. b) Differential conductance, plotted as a function of gate voltage. The measured curve (black) can be fit by assuming a saddle- point potential and calculating the transversal harmonic confinement potential from the subband spacing in a). The resulting fit for each step is plotted at the bottom. c) Subband spacing of the transversal (squares) and longitudinal (circles) confinement potential, as extracted from the fits in b). Approaching pinch-off, QPC1 (left) becomes much narrower but only slightly shorter. The longitudinal curvature of QPC2 (right) is smaller than that of QPC1, which is in agreement with the lithographic dimensions of the gates. confinement potential is harmonic enables an alternative way to determine α. At the position of the conductance steps in Fig.7a) (at G = (n − 0.5) × 2e2/h), the conduction band bottom is ECB = (n − 0.5) ωY below the Fermi energy. Hence, the lever arm in-between two successive steps is given by α = (n + 1 − 1/2) ωY,n+1 − (n − 1/2) ωY,n (3) with ωY,n being the confinement at the n-th step as extracted from Fig.7a). Now the only fitting parameters are the longitudinal curvature ωX and the energy offset E0. Figure 7b) shows the pinch off curve of QPC1 (left) and the fits resulting from the described procedure. All fits are plotted below G = 2e2/h for clarity. The same procedure applied to QPC2 is shown on the right hand side. Figure 7c) shows the extracted values for the longitudinal curvature (circles) of QPC1 (left) and QPC2 (right). The transversal curvature is plotted as squares for comparison. In both QPCs, ωX a) 4 3 2 1 0 ΔSB(meV) VG1&G2(V) -1.5 -2 G (2e2/h) 0 2 4 6 8 10 4 3 2 1 0 ħω (meV) QPC1 measured fit ħωy ħωx b) c) 4 3 2 1 0 ΔSB(meV) VG1&G2(V) -2 -3 G (2e2/h) 0 5 10 15 4 3 2 1 0 ħω (meV) QPC2 measured fit ħωy ħωx 5 5 -4 Transport Properties of Clean Quantum Point Contacts 14 is smaller than ωY , as required for the observation of conductance quantization [29]. Approaching pinch-off, QPC1 becomes much narrower (strong increase of ωY ) and slightly shorter (increase of ωX). QPC2 also becomes much narrower but there is no strong increase of ωX. Although the saddle-point model might not be the ideal model for a quantum wire, this fits the intuitive picture of a 1-dimensional channel with a lithography-defined length and voltage-controlled width. Comparing the parameters obtained from our analysis to those from earlier investigations, we find surprising discrepancies despite similar 2DEG density and gate- spacings. The data analyzed in [31] is well described by ωY = 0.9 meV, ωX = 0.3 meV with both values being independent of gate voltage. In our devices, the shape of the confinement potential changes dramatically as a function of gate voltage and reaches oscillator strengths of ωY > 4 meV, ωX > 1 meV close to pinch-off. We interpret this observation as the result of the screening properties of the screening layers that are grown into the heterostructure in order to achieve ultra-high electron mobilities [6, 7, 8]. Screening should reduce the range of the gate-induced potential and thereby increase the slope of the confinement potential. 2.5. Spin-Resolved Transport at Low Temperatures Additional changes in the confinement can be created by applying a magnetic field B⊥ perpendicular to the plane of the 2DEG which lifts the spin degeneracy and increases the subband spacing [1, 32, 21, 12, 9]. QPC3 was not equipped with a 2DEG terminal that could be used to measure the diagonal voltage, so the filling factor of the QPC νQPC is calculated from the longitudinal four-terminal resistance: In analogy to magnetotransport through a barrier [33], νQPC relates via RQPC×e2/h = 1/νQPC+1/νBulk to the number of occupied Landau levels in the bulk νBulk = nSh/eB⊥. Knowing the electron density nS and Planck's constant h, the measured four-terminal resistance RQPC can be directly converted to νQPC. Figure 8a) shows data of QPC3 (split-gate gap w = 250 nm), measured at a temperature of T = 0.1 K with B⊥ = 2 T. Plotting νQPC as a function of VG1&G2 reveals integer filling factors related to the magneto-electric subband spacing (νQPC = 2, νQPC = 4) but also smaller plateaus due to the lifted spin degeneracy (νQPC = 3, νQPC = 5). The energy diagram corresponding to νQPC = 3 is shown on the right hand side. The lowest spin-split plateau is obscured by the 0.7 anomaly [12]. The same trace repeated at finite source-drain bias VSD = 2 mV is shown in Fig. 8b). Additional half-plateaus with νQPC = 2.5 and νQPC = 3.5 appear which correspond to a situation as depicted to the right: one spin-split mode is situated in- between µS and µD. The 0.7 anomaly has evolved into a plateau with νQPC = 1.8. Figure 8c) shows a false color plot of the transconductance GTC = dG/dVG1&G2, plotted as a function of VSD and VG1&G2. Similar to the transconductance plots in Figs. 6, regions of integer filling factor appear as black diamonds. Due to the different sizes of the subband-split levels (labeled 2, 4 and 6) and the spin-split levels (3 and 5), the half-plateaus at finite VSD appear as black stripes. A clear deviation from the regular Transport Properties of Clean Quantum Point Contacts 15 Figure 8. (color online) a) Filling factor νQPC of QPC3 (split-gate gap w = 250 nm), plotted as a function of VG1&G2. Measured at a temperature of T = 100 mK, the spin degeneracy is lifted by a magnetic field B⊥ = 2 T applied perpendicular to the plane of the 2DEG. The situation corresponding to νQPC = 3 is sketched on the right hand side: three non-degenerate subbands reside below the chemical potentials of source and drain. b) Filling factor of QPC3 as a function of VG1&G2 while a source-drain bias of VSD = 2 mV is applied. Filling factors νQPC = 2.5 and νQPC = 3.5 are observed when one spin-split level lies in-between µS and µD (sketched on the right hand side). c) False color plot of the transconductance GTC = dG/dVG1&G2 of QPC3 (inset), plotted as a function of VSD and VG1&G2. Regions of integer filling factor νQPC = 2, 3, 4, 5, 6 appear as black diamonds centered around VSD = 0 mV, half plateaus with νQPC = 2.5, 3.5, 4.5 appear as black stripes at finite source-drain bias. The 0.7 anomaly creates a plateau with νQPC = 1.8. even-odd pattern is observed at low filling factors, where νQPC = 0.5, 1, 1.5 are replaced by a νQPC = 1.8 plateau related to the 0.7 anomaly. From the extent of the spin-split plateaus the exchange-enhanced g-factor g∗ can be extracted via VSD = ∆Spin = g∗µBB⊥, where g∗ is the effective g-factor and µB denotes the Bohr magneton. Compared to the bare g-factor of GaAs (g = −0.44), we find a strongly enhanced g∗ = 4.4 at νQPC = 3 and g∗ = 3.8 at νQPC = 5. Similar to findings of Thomas at al. [12], g∗ increases with lower mode number. However, the VG1&G2 (V) VSD (mV) -12 -6 0 6 -1.3 -2.7 -2.5 -2.2 -1.9 -1.6 dG/dVG1&G2 (a.u.) 0 1 QPC3 T = 0.1 K B┴ = 2 T 2 4 0 3 5 2.5 4.5 3.5 1.8 6 12 b) a) c) VG1&G2 (V) -1.5 -2.5 -2.0 νQPC 0 2 4 νQPC 0 2 4 B┴ = 2 T VSD= 0 mV B┴ = 2 T VSD= 2 mV µS µD g*µBB µS µD νQPC= 3 QPC3 νQPC= 3.5 250 nm E kx Transport Properties of Clean Quantum Point Contacts 16 magnitude of the exchange enhancement is different: 0.4 < g∗ < 1.3 was reported by Thomas at al., while our results indicate a much stronger enhancement. Assuming that disorder reduces the effectiveness of the exchange enhancement, the observation of strongly enhanced g-factors can be interpreted as another manifestation of the good sample quality. 3. Conclusion In conclusion, we investigated the transport properties of two differently shaped constrictions that were defined within a high-mobility 2DEG. Transport spectroscopy in the linear response regime demonstrates that conductance quantization is observed and that no scattering centers are found when shifting QPC1 between the gates. The 0.7-anomaly is investigated by varying the temperature and by applying a magnetic field perpendicular to the 2DEG. Depending on the ratio of these two parameters we observe either a weakening or an enhancement of the 0.7-anomaly which is discussed in view of spin-polarized edge channels mimicking the 0.7 scenario. Measurements at finite source-drain bias give access to the subband spacings of the QPC and reveal that QPC1 is described best by a short constriction whereas QPC2 shows quantum wire-like characteristics. In addition to QPCs defined in standard 2DEGs, resonances at large bias are observed in QPC1 that correspond to higher order transport conditions. These higher order resonance conditions give valuable information about the shape of the confinement potential and enable the reconstruction of the full saddle point potential in the constriction. Knowing the shape (and hence the slope) might prove important for the investigation of many body states like the 0.7 anomaly or the transmission of fractional quantum Hall states. At finite magnetic field and lower temperature, a strongly exchange-enhanced g-factor is observed which we interpret as the result of a very smooth confinement potential. While the measurements on QPCs are possible with high signal to noise ratio if the gate voltages are always swept in the same regime and direction, it was not possible to form a stable quantum dot with the same technique on the same wafer. For future interferometer experiments on high-mobility samples in the fractional quantum Hall regime it is desirable to prepare split-gate electrodes on high-mobility wafers that do not rely on screening electrons at the X-valley and thereby allow stable gate operation. 4. Acknowledgments We acknowledge the support of the ETH FIRST laboratory and financial support of the Swiss Science Foundation (Schweizerischer Nationalfonds, NCCR Nanoscience). References [1] Wharam D A, Thornton T J, Newbury R, Pepper M, Ahmed H, Frost J E F, Hasko D G, Peacock D C, Ritchie D A and Jones G A C 1988 J. Phys. C.: Solid State Phys. 21 L209 Transport Properties of Clean Quantum Point Contacts 17 [2] van Wees B J, van Houten H, Beenakker C W J, Williamson J G, Kouwenhoven L P, van der Marel D and Foxon C T 1988 Phys. Rev. Lett. 60 848 [3] Willett R, Eisenstein J P, Stormer H L, Tsui D C, Gossard A C and English J H 1987 Phys. Rev. Lett. 59 1776 [4] Miller J B, Radu I P, Zumbuhl D M, Levenson-Falk E M, Kastner M A, Marcus C M, Pfeiffer L N and West K W 2007 Nature Physics 3 561 [5] Dolev M, Heiblum M, Umansky V, Stern A and Mahalu D 2008 Nature 452 829 [6] Friedland K-J, Hey R, Kostial H, Klann R and Ploog K 1996 Phys. Rev. Lett. 77 4616 [7] Umansky V, de Picciotto R and Heiblum M 1997 Appl. Phys. Lett. 71 683 [8] Hwang E H and Sarma S D 2008 Phys. Rev. B. 77, 235437 [9] Rossler C, Feil T, Mensch P, Ihn T, Ensslin K, Schuh D and Wegscheider W 2010 N. J. Phys. 12 043007 [10] Williamson J G, Timmering C E, Harmans C J P M, Harris J J and Foxon C T 1990 Phys. Rev. B. 42 7675 (R) [11] The magnitude of the QPC-shift has been independently verified via low-temperature scanning gate microscopy; Schnez S et al., arXiv:1109.1544v1 [12] Thomas K J, Nicholls J T, Simmons M Y, Pepper M, Mace D R and Ritchie D A 1996 Phys. Rev. Lett. 77 135 [13] Kristensen A, Bruus H, Hansen A E, Jensen J B, Lindelof P E, Marckmann J C, Nygard J, Sorensen C B, Beuscher F, Forchel A and Michel M 2000 Phys. Rev. B. 62 10950 [14] Cronenwett S M, Lynch H J, Goldhaber-Gordon D, Kouwenhoven L P, Marcus C M, Hirose K, Wingreen N S and Umansky V 2002 Phys. Rev. Lett. 88, 226805 [15] Hatke A T, Zudov M A, Pfeiffer L N and West K W 2011 Phys. Rev. B 83, 081301(R) [16] Shailos A, Bird J P, Lilly M P, Reno J L and Simmons J A 2006 J. Phys. Cond. Mat. 18, 3277 [17] Komijani Y, Csontos M, Shorubalko I, Ihn T, Ensslin K, Meir Y, Reuter D and Wieck A. D. 2010 EPL 91, 67010 [18] Glazman L I and Khaetskii A V 1989 Europhys. Lett. 9 263 [19] Patel N K, Martin-Moreno L, Pepper M, Newbury R, Frost J E F, Ritchie D A, Jones G A C, Janssen J T M B, Singleton J and Perenboom J A A J 1990 J. Phys. C 2, 7247 [20] Patel N K, Nicholls J T, Martin-Moreno L, Pepper M, Frost J E F, Ritchie D A and Jones G A C 1991 Phys. Rev. B 44 13549 [21] Thomas K J, Simmons M Y, Nicholls J T, Mace D R, Pepper M and Ritchie D A 1995 Appl. Phys. Lett. 67, 109 [22] Rossler C, Herz M, Bichler M and Ludwig S 2010 Solid State Comm. 150 861 [23] Pioro-Ladriere M, Davies J H, Long A R, Sachrajda A R, Gaudreau L, Zawadzki P, Lapointe J, Gupta J, Wasilewski Z and Studenikin S 2005 Phys. Rev. B 72 115331 [24] de Picciotto R, Pfeiffer L N, Baldwin K W and West K W 2004 Phys. Rev. Lett. 93 36805 [25] Tarucha S, Honda T and Saku T 1995 Solid State Comm. 94 413 [26] Liang C-T, Simmons M Y, Smith C G, Ritchie D A and Pepper M 1999 Appl. Phys. Lett. 75 2975 [27] Morimoto T, Henmi M, Naito R, Tsubaki K, Aoki N, Bird J P and Ochiai Y 2006 Phys. Rev. Lett. 97 096801 [28] Laux S E, Frank D J and Stern F 1988 Surf. Sci. 196 101 [29] Buttiker M 1990 Phys. Rev. B 41 7906 [30] Kouwenhoven L P, Austing D G and Tarucha S 2001 Rep. Prog. Phys. 64 701 [31] Taboryski R, Kristensen A, Sorensen C B and Lindelof P E 1995 Phys. Rev. B 51 2282 [32] van Wees B J, Kouwenhoven L P, van Houten H, Beenakker C W J, Mooij J E, Foxon C T and Harris J J 1988 Phys. Rev. B 38 3625 (R) [33] Haug R J, MacDonald A H, Streda P and von Klitzing K 1988 Phys. Rev. Lett. 61 2797
1104.0147	1	1104	2011-04-01T11:42:34	Edge states and flat bands in graphene nanoribbons with arbitrary geometries	[ "cond-mat.mes-hall" ]	We prescribe general rules to predict the existence of edge states and zero-energy flat bands in graphene nanoribbons and graphene edges of arbitrary shape. No calculations are needed. For the so-called {\it{minimal}} edges, the projection of the edge translation vector into the zigzag direction of graphene uniquely determines the edge bands. By adding extra nodes to minimal edges, arbitrary modified edges can be obtained. The edge bands of modified graphene edges can be found by applying hybridization rules of the extra atoms with the ones belonging to the original edge. Our prescription correctly predicts the localization and degeneracy of the zero-energy bands at one of the graphene sublattices, confirmed by tight-binding and first-principle calculations. It also allows us to qualitatively predict the existence of $E\ne 0$ bands appearing in the energy gap of certain edges and nanoribbons.	cond-mat.mes-hall	cond-mat	Edge states and flat bands in graphene nanoribbons with arbitrary geometries W. Jaskolski1,∗ A. Ayuela2, M. Pelc1, H. Santos3, and L. Chico3 1 Instytut Fizyki UMK, Grudzi¸adzka 5, 87-100 Toru´n, Poland 2 Centro de F´ısica de Materiales CFM-CPM CSIC-UPV/EHU, Departamento de F´ısica de Materiales(Facultad de Qu´ımica, UPV) and Donostia International Physics Center, 20080 San Sebastian/Donostia, Spain 3 Instituto de Ciencias de Materiales de Madrid, CSIC, Cantoblanco, 28049 Madrid, Spain (Dated: July 5, 2018) We prescribe general rules to predict the existence of edge states and zero-energy flat bands in graphene nanoribbons and graphene edges of arbitrary shape. No calculations are needed. For the so-called minimal edges, the projection of the edge translation vector into the zigzag direction of graphene uniquely determines the edge bands. By adding extra nodes to minimal edges, arbitrary modified edges can be obtained. The edge bands of modified graphene edges can be found by applying hybridization rules of the extra atoms with the ones belonging to the original edge. Our prescription correctly predicts the localization and degeneracy of the zero-energy bands at one of the graphene sublattices, confirmed by tight-binding and first-principle calculations. It also allows us to qualitatively predict the existence of E 6= 0 bands appearing in the energy gap of certain edges and nanoribbons. PACS numbers: 73.20.-r, 73.22.-f, 73.22.Pr I. INTRODUCTION Graphene is presently one of the most studied ma- terials in condensed matter and materials science. It presents a plethora of interesting physical phenomena due to the fact that its elementary electronic excita- tions behave as two-dimensional chiral Dirac fermions.1 Graphene nanoribbons (GNR), stripes of nanometric widths cut from graphene, are also the subject of a grow- ing interest. They exhibit edge-localized states, which may have potential practical applications and are the key ingredient in many of the fascinating properties of graphene and its nanometric derivatives. Edge states play an important role in transport and magnetic prop- erties of GNRs, such as the quantum Hall effect2 and the quantum spin Hall effect.3 The magnetic properties of nanoribbons are directly related to the existence of localized edge states.4 The appearance of the edge states in GNR have been investigated long before5,6 graphene sheets were ex- perimentally achieved.7 A large number of theoretical works based on continuum Dirac-like,8 tight-binding,9 and density-functional10 approaches have been applied to study GNRs with different edges, such as zigzag,8,10 armchair,8 -- 10 or mixed with cove and with Klein nodes.11 All these edge terminations have been experimentally identified by different techniques, such as scanning tun- neling microscopy,12,13 high resolution transmission elec- tron microscopy14 or atom-by-atom spectroscopy.15 From the theoretical viewpoint, it is important to iden- tify general edges and nanoribbons which present local- ized edge states, as well as their degeneracy and charac- teristics. Although the boundary conditions for an im- portant subset of edge terminations has been studied,17 as well as certain modifications19 with experimental interest,14 until now identifying general ribbons with edge states and their band structure characterization is still an open question. In this paper we solve this problem by giving a simple prescription which allows to predict the existence of the edge states and their degeneracies in a given graphene edge or nanoribbon. We show that no calculations are needed to find out whether the edge states and flat bands exist at the Fermi energy (EF ) for any kind of periodic graphene edge or wide enough nanoribbon with noninter- acting edges, at least at the level of π-electron approxi- mation. We consider periodic edges defined by a translation vector T. Our approach follows two steps: First, we characterize minimal edges,17 i.e., those with a minimum number of edge nodes and dangling bonds per translation period. For minimal edges, the spectrum of E = 0 flat bands is determined by the zigzag edge component of T, i.e., the projection of T along the zigzag direction, which poses a folding rule to the graphene zigzag edge band. Next, we show that any other edge can be ob- tained from a minimal one by adding extra nodes. We call these modified edges. The extra nodes provide ex- tra bands at the Fermi energy that may hybridize with the edge orbitals. If hybridization takes place, the extra bands couple with the existing E = 0 edge bands and split in energy, moving towards the bulk bands. Such splitting depends on whether the extra nodes belong to the same sublattice as that where the edge zero-energy bands are localized. We find an extremely simple rule to determine, without performing any calculations, the ex- istence, origin and localization properties of edge states and flat bands of any modified edges, thus allowing the complete characterization of the low-energy properties of any GNR with general edges. This prescription for the identification of edge states and flat bands in graphene edges and nanoribbons was found after performing calculations for a large number of different GNRs. The calculations were performed in the tight-binding (TB) π-electron approximations using hop- ping parameter t0 = 2.66 eV. We have collected a huge amount of data, but only some of them are presented here for illustration purposes. The rest of the paper is organized as follows. In Sec. II we give the geometrical description of the edges and nanoribbons as well as the edge band folding rule with an example regarding a minimal edge. Sec. III describes modified (i.e., non-minimal) edges, starting with zigzag edges with Klein defects and cape structures, which al- lows us to set the rules to find out the zero-energy edge bands. We introduce simple diagrams, which determine their localization and degeneracy. The Section concludes discussing edge bands away from the Fermi energy, fo- cusing on armchair modified and chiral edges. Finally, In Sec. IV we summarize our results. II. CHARACTERIZATION OF GRAPHENE EDGES. FOLDING RULE AND MINIMAL EDGES A graphene edge consists of a set of lattice sites with only one or two neighbors, i.e., having two or one dan- gling bonds, respectively. In this work we assume that the edge atoms are arranged periodically. Therefore, they form a one-dimensional superlattice with a translation period defined by T = nR1 + mR2, where R1 and R2 are the primitive vectors of the honeycomb lattice, as seen in Fig. 1. For a given period T with indices (n, m), the number of edge sites Ne and the number of dangling bonds Nd can be arbitrarily large, but neither of them can be smaller than n + m.17 Following Ref. 17, when the number of edge atoms equals that of dangling bonds and both are equal to n + m (Ne = Nd = n + m), we call the edge minimal. For a minimal edge, the number of nodes in one sublattice equals n and the number of nodes in the other sublattice equals m. TZ TA T R1 R2 FIG. 1. (Color online) Graphene lattice and the translation vector T (n, m) = (8, 1). The vectors TZ and TA are pro- jections of T into the zigzag and armchair directions. The primitive lattice vectors R1 and R1 are also shown so that T= nR1 + mR2. The dark-grey (red) line shows the minimal edge; the light-grey (blue) line shows a modified edge con- structed by attaching two Klein nodes to the minimal case. 2 Any minimal edge can be modified by adding extra edge nodes. To identify a modified edge one has just to add information on the extra nodes added to the corre- sponding minimal edge. As an example, Fig. 1 presents two edges associated to the translation vector T (8, 1). The minimal edge is marked by a red line. Another pos- sible edge, with two additional nodes, constituting the so-called Klein defects, is marked by a blue line. The translation vector T = nR1 + mR2 defining any graphene edge can be decomposed into two important directions in the honeycomb lattice (see Fig. 1), namely the armchair and zigzag: T = TA + TZ . Note that TZ = (n − m)R1 and TA = m(R1 + R2). This decompo- sition is crucial in our analysis, since it is well-known that an armchair edge does not have E = 0 localized states, while the zigzag termination reveals a flat edge band at Fermi energy (in the π electron approximation) for the wavevector k > 2/3π, as shown in previous nanoribbon band structure calculations5,6. Also, it is easy to see that a minimal edge corresponding to T is a simple combina- tion of the zigzag edge along TZ and the armchair edge along TA. The schematic spectrum (close to EF ) of the zigzag edge defined by the smallest TZ = R1, i.e., the edge (1,0), is shown in the upper left panel of Fig. 2. The spectrum of a zigzag (n − m, 0) = (S, 0) edge defined by TZ = SR1 is obtained by folding this spectrum S-times. The (2,0) case is shown explicitly in the upper right panel of Fig. 2. Both, the (1,0) and the (2,0) edge, have degen- eracy 1. By repeated folding of the minimum (1,0) zigzag edge one can easily find the bandstructure and edge band degeneracies of a general (S, 0) edge. Any integer number S > 0 can be written as S = I + 3M , where I = 1, 2, 3 and M = 0, 1, 2, .... For I = 1 and 2 the folded spectrum has always the Dirac point at 2/3π while for I = 3 the Dirac point is at k = 0, as illustrated in Fig. 2. The schematic band structures of all zigzag edges obtained from this folding are shown in Fig. 2. The degeneracies of the zero-energy band are also indicated in the lower panels (M and M + 1) close to the corresponding edge bands. Since the armchair component does not provide any edge states, one can expect that the spectrum of a minimal-edge (n, m) will be similar to the spectrum of the (n − m, 0) zigzag edge at least close to EF . We have performed tight-binding calculations for a large number of different minimal-edge GNRs, and verified that the presented above folding rule holds in all the cases con- sidered. Notice that for a graphene nanoribbon with two equal edges, the degeneracy of the flat band is twice that of the isolated edge, provided that the ribbon is wide enough to neglect interaction between the edges. In order to construct ribbons corresponding to a par- ticular edge T, we first define the vector H as the smallest graphene lattice vector perpendicular to T. The width of the ribbons studied here is spanned by a vector W given by an integer multiple of H, W = N H as it is shown in Fig. 3 for the case of 2(7,1) GNR. For a fixed (1,0) I=1, M=0 (2,0) I=2, M=0 M M+1 M M+1 I=1 I=2 I=3 FIG. 2. Schematic band structures of zigzag (1,0), (2,0) and general (S, 0) edges after folding the (1,0) zigzag edge band, where S = I + 3M as described in the text. The shaded areas represent the band continuum of states. Degeneracies of the zero-energy bands (M , M + 1) are indicated in the lower pan- els close to the corresponding edge bands at the Fermi energy. They correspond to semi-infinite graphene sheet with only one edge. In case of a GNR with equal edges the degeneracies are doubled. T, H is uniquely determined up to a global plus or minus sign; therefore, for our purposes, the ribbons with min- imal edges are labeled by N (n, m), where N states the ribbon width, and (n, m) indicates the minimal edge. 3 As a particular example, the spectrum of a nanoribbon with minimal-edge 3(8,1) is presented in the left panel of Fig. 4. One can easily see that degeneracies of the flat E = 0 bands are exactly the same as for the 20(7,0) zigzag GNR (right panel of Fig. 4). Also the gaps at k = 0 and π follow the folding rule to a large extent. In general, the minimal-edge (n, m) GNR reveals the same zero-energy bands as the (n − m, 0) zigzag GNR, which in turn has the same spectrum as the (n − m) times folded spectrum of the (1,0) zigzag nanoribbon of the same width. GNR 20(7,0) GNR 3(8,1) 3 2 1 3 2 1 4 0 ) V e ( E 6 ) V e ( E 4 0 6 −1 −2 −3 0 −1 −2 0.22 k 0.44 −3 0 0.18 k 0.36 FIG. 4. Energy spectra of the 20(7,0) GNR (left) and the 3(8,1) GNR (right) close to the Fermi energy. The unit cells of both GNRs contain similar number of nodes. The degen- eracies of the zero-energy edge bands (4 and 6) are indicated close to extreme k values. III. MODIFIED EDGES A. Coupling of edge defects and band splitting FIG. 3. (Color online) Geometric structure of the 2(7,1) GNR highlighted in dark grey (green) on a graphene sheet, indicat- ing the translation vector T = TZ + TA and the width vector W = 2H, where H is the smallest nanoribbon width vector belonging to the graphene lattice. TZ = 6T(1, 0). The unit cells spanned by T and H or W are marked. Of course, there exist ribbons with minimal edge ge- ometries, that may require a semi-integer N , such as the so-called antizigzag ribbons. As our main goal here is to study edge bands, we restrict ourselves to integer N , using values which yield non-interacting GNR edges. Here we study a couple of modified zigzag edges. We start with the Klein defects,18 which consist of atoms with only one neighbor, as depicted in Fig. 5(a). Then we proceed with other modified edges related to the so-called cove structures, which can be constructed by adding ex- tra atoms to the Klein-edge zigzag nanoribbon. The basic unit of such modified edges is what we call a cape. This can be built by bonding two Klein defects to one ex- tra atom, and the resulting structure is depicted in Fig. 5(b). When extra atoms are added every two Klein de- fects, one gets the cove edge, shown in Fig. 5(c). We will consider modified edges where capes are more separated 4 than in the cove edge structure, as in Fig. 5(d). Based on these examples we discuss how such modifications in- fluence the mixing and splitting of states at E = 0. In order to perform numerical calculations, we choose wide ribbons with equal edges. As discussed in the previous Section, if the ribbons are wide enough, the interaction between edges is reduced; and as they have equal edges, the degeneracies are obviously twice those stated in Sec. II. FIG. 5. Geometries of several modified zigzag graphene edges. (a) Bearded zigzag edge, composed of Klein defects; (b) a cape structure on a zigzag edge, obtained by bonding one extra atom to two adjacent Klein defects; (c) a cove edge; (d) a periodic modified edge with a cape. 1. Zigzag edge with Klein defects We consider first a simple zigzag nanoribbon modi- fied by Klein defects. The spectrum of a zigzag nanorib- bon of width 40 and minimal edge, i.e., a 40(1,0) GNR, is presented in Fig. 6(a). The two zero-energy edge bands localized at opposite edges6 of the GNR extend for k > 2/3π. The condition to get E = 0 requires a non-vanishing amplitude of the wavefunction in one of the two graphene sublattices16. For k = π the corre- sponding wavefunctions are localized at the edge nodes. Opposite zigzag edges have atoms belonging to different sublattices. For k closer to 2/3π the wavefunctions cor- responding to these edge bands penetrate more into the inner part of the GNR; therefore, the edge states interact more for these lower k values, mixing and splitting into bonding and antibonding bands. When Klein defects are added to both sides of the GNR, the flat bands appear from k = 0 to 2/3π. In order to understand this change on the flat E = 0 bands with respect to k values, we gradually modify the cou- pling of the extra atoms conforming the Klein bearded edge. We first add a Klein node at each side of the GNR unit cell, but setting the hoppings equal to zero. As the on-site energies of these extra atoms are set to zero, an additional doubly degenerated zero-energy band appears, as shown in the spectrum of Fig. 6(b). The double degeneracy is due to the contribution of the two edges. Let us now switch on the hopping, setting t = 0.5t0. The resulting spectrum is shown in Fig. 6(c). The hopping connects atoms that belong to different sublattices; this allows for FIG. 6. Evolution of the spectrum of 40(1,0) zigzag GNR (a) to the spectrum of 40(1,0) GNR with Klein nodes attached at both sides (d). In (b), two extra non-attached nodes introduce a doubly degenerate E = 0 flat band. In (c), connecting the extra nodes with t = 0.5t0 couples and split the flat bands in the range of k > 2/3π. The bottom panel illustrates how the extra nodes are attached to the upper and lower part of the unit cell of the zigzag GNR. the interaction of the corresponding flat bands, which hybridize and split. Such splitting is more pronounced for k = π, since their localization and overlap is stronger than for any other k, and increases gradually from k = 2/3π to the edge of the Brillouin zone. Finally, when t = t0, see Fig. 6d, the splitting is so strong that the bonding and antibonding bands interact from k = 2/3π to k = π and reach the continuum of bands. We end up with the spectrum of a zigzag GNR with Klein edges, which has a zero-energy flat band for k < 2/3π. A closer inspection of the analytical TB solution for the E = 0 band of zigzag GNR with Klein edges reveals that, contrary to the zigzag edge, the wavefunction never localizes only at the Klein nodes. The wavefunction pen- etrates into GNR even for k = 0: the damping factor equals 1/2 in this case and rises up to 1 for k = 2/3π. It is noteworthy that if we modify only one edge by adding Klein nodes, the extra zero-energy band will be non-degenerate and will mix and split with only one of the edge bands in the range of k > 2/3π. This mixed band is obviously localized at the edge where the Klein nodes are added. 5 quadrupled unit cell. Two extra adjacent Klein nodes (open circles) are next added. Finally, the pair of Klein nodes is connected via another extra node (black circle), to form the cape structure. The spectrum around the Fermi level of the zigzag 40(4,0) GNR is presented in Fig. 7(a). It is obtained by folding four times the spectrum of Fig. 6(a). The edge band degeneracies indicated in the Figure result from this folding. The two non-connected Klein nodes at each side of the GNR unit cell (a total of four extra nodes) yields the addition of a four-fold degenerate flat band at zero energy. The degeneracies sum up to six and eight at k = 0 and k = π, respectively. This is shown in Fig. 7(b). We begin by connecting these extra nodes with a small hopping t = 0.2t0 and plotting the bands in Fig. 7(c). For k > 2/3π all the flat bands mix and split; they are still doubly degenerated since both edges of the GNR are equal. For k < 2/3π only two bands split and a doubly degenerate flat band survives at E = 0. When t = t0, see Fig. 7(d), all the split bands merge into the states continuum. The flat bands that survived at E = 0 reveals that they are mainly localized at the extra Klein nodes and enter into GNR nodes of the same sublattice. Their spreading into the GNR is similar to that found in the flat bands of zigzag GNRs with Klein edges (see Fig. 6(d)). To form the (4,0) GNR with a single cape structure we must add yet another extra node and connect it to the existing Klein nodes. Note that the extra node and the Klein nodes belong to different sublattices. A non- connected node adds a doubly degenerate zero-energy band to the spectrum of Fig. 7(d). When connecting this extra atom to the previous Klein nodes with t = 0.5t0, the flat bands at k < 2/3π, which are due to the Klein de- fects, hybridize with these added bands arising from the extra connected nodes and split. This splitting is shown in Fig. 7(e). For t = t0, as seen in Fig. 7(f). Due to the strong coupling, the bands split so much that they merge into the continuum of states. This doubly degenerate flat band at k > 2/3π appears due to the introduction of the outermost extra atoms, and as there were no localized bands at that k range which could mix with them, they remain localized in the extra node, spreading into nodes belonging to the same sublattice. This localization is confirmed by numerical calcula- tions performed within the tight-binding model, as well as using first-principles DFT approach20. The obtained wavefunctions are shown in the right panel of Fig. 8. Their localization in the (4,0) GNR with a single cape is different from the case of the cove edge,19 which can be considered as built from a (2,0) GNR with a cape structure (see Fig. 5c). In Ref. 19 it was shown that in the case of cove edge the wavefunctions of the zero-energy bands are not localized at the outermost edge atoms, but at the neighboring nodes which belong to the other sub- lattice. We confirm this finding, both in the tight-binding and the DFT calculations shown in the left panel of Fig. 821. FIG. 7. Evolution of the spectrum of a (4,0) zigzag GNR when extra edge nodes are added to form a single cape structure. The upper panel shows the edge of GNR where connections to extra nodes are marked by dotted lines. The four unit cells are marked by dashed lines. The steps are the following: (a) pure (4,0) GNR, (b) with two disconnected extra nodes at each side of unit cell, (c) with two extra nodes (i.e. Klein defects) connected by t = 0.2t0, (d) with two Klein nodes fully connected, (e) with another extra node connected to the Klein nodes by t = 0.2t0 forming the cape structure, and (f) with a cape structure added to the edge and all extra nodes fully connected. Diagrams illustrate the mixing and splitting of flat bands, to be explained in the text. 2. Zigzag edge modified with a cape structure Here we consider a more complex modification of the zigzag edge. It helps to illustrate and refine in detail the presented prescription, which allows us to predict the oc- currence and degeneracies of edge states for any nanorib- bon. The top panel of Fig. 7 shows the construction of the cape at the upper edge only. In a similar way the cape is formed at the lower edge, but the black and open circels are reversed there. We have considered the (4,0) zigzag GNR and a se- quential addition of nodes to get the edges with a cape structure, as shown in Fig. 7. The procedure follows the next sequence. One starts from a zigzag ribbon with One would naively expect that the wave function should be more localized at the outermost atoms, which seems more exposed to a chemical attack. However, it is easy to check that in a cove edge, the majority of the edge atoms are not the outermost ones, but rather their nearest neighbors. The edge state is therefore localized in the atoms closest to the outermost ones, which belong to the opposite sublattice, whereas the wavefunction weight of the outermost node is zero in the tight-binding or neg- ligible in the DFT approach. For the same reason, in a larger zigzag edge with a single cape structure such as that shown in Fig. 7, the majority of edge atoms are from the original zigzag edge, which belong to the same sublattice as the outermost atoms of the cape structures. In this case the weight in the edge state is thus in the sublattice of the majority of the zigzag edge atoms, with a nonzero value in the outermost cape node, as illustrated in the right part of Fig. 8. B. Mixing and splitting diagrams As discussed in the previous Sections, the mixing and splitting of the flat bands wavefunctions occur due to the hybridization of orbitals between neighboring nodes, which belong to different sublattices. We introduce sim- ple diagrams that help to understand such hybridization. We can explain how such bands at E = 0 split and their wavefunctions localize. Each diagram is built of two rows containing square boxes where each box represents a non-degenerate band at zero energy. The upper and lower rows correspond to the upper and lower GNR edges, respectively. Empty and filled boxes represent bands that localize at differ- ent sublattices. Each extra node added to a given edge is represented by a box added to the corresponding row. The added box is empty or filled depending on the sub- lattice it belongs to. A pair of empty and filled boxes in a given row represents now two interacting and hybridizing bands which must split and move away from zero energy. Their corresponding pair of empty and filled boxes anni- hilate and disappear from the row. The remaining boxes represent the flat bands which survive such hybridization process. Their filling determines the sublattice at which they localize. We describe next how the diagrams explain the mixing and splitting of the flat bands in a practical case. We consider extra nodes added to the edge of (4,0) GNR to form a ribbon with a cape structure, as shown previously in Fig. 7. The diagram corresponding to Fig. 7(a), i.e. to the (4,0) zigzag GNR, has for k < 2/3π one filled box in the upper row (representing an E = 0 state localized at the upper edge, on the sublattice marked by filled circles) and one empty box in the lower row (representing an E = 0 state localized at the lower edge, on the sublattice marked by empty circles). For k > 2/3π we have two filled boxes in the upper row and two empty boxes in 6 FIG. 8. (Color online) Localization of the wavefunctions cor- responding to the E = 0 band at k = π for 40(2,0) (left) and 40(4,0) (right) GNR with a cape structure at the edges. The corresponding edges are shown in Fig. 5 (c) and (d), re- spectively. Only an edge and a few neighboring nodes in the GNRs unit cells are shown. Upper panel: Results obtained using tight-binding method. Bottom panel: Results of first- principles calculations. The dot diameter in the upper panel reflects the TB density at the nodes. No dot means that the wavefunction is exactly zero at this node. the lower row. We have already described the process of adding two extra Klein nodes in the previous subsection. The corresponding diagrams are shown under the bands depicted in Fig. 7(b) and (c). The two Klein nodes at each side of the GNR unit cell add two empty/filled boxes in the upper/lower rows (Fig. 7(b)). The pairs of empty/filled boxes in each row annihilate (Fig. 7(c)) and the corresponding bands split. Two boxes are only left at the Fermi energy for k < 2/3π as shown in Fig. 7(d). The (4,0) GNR with a cape structure has an extra node connected to the existing Klein nodes at each side of the GNR. When these extra nodes are not connected, two additional E = 0 bands appear in the spectrum. They are represented by two extra boxes added to the exist- ing diagram: filled box in the upper row, and an empty box in the lower row. Now, for k < 2/3π the pair of filled/empty boxes in a given row annihilates, leading to the hybridization and splitting of the zero-energy bands, as shown in Fig. 7(e) and (f). For k > 2/3π the remain- ing two boxes do not have any partner state to hybridize, therefore they origin the states at E = 0 which are lo- calized in the external cape node and its sublattice. Our prescription and diagram analysis confirms the degen- eracy and localization of bands as seen in the previous Subsection, but without performing any calculations. C. Gap states away from the Fermi energy We have already observed (see Fig. 7(d)) that in some cases the split bands do not reach the bulk continuum 7 2. Chiral edges In this Subsection we present the results corresponding to general (n, m) edges, i.e., those which are not purely armchair or zigzag, either minimal or modified. Their borders with Klein defects and the corresponding spec- tra can be explained by the folding rule and diagrams presented above. As example, we investigate a 3(8,1) GNR with a min- imal edge and two different modifications. Studying dif- ferent edges for a given translation vector T is important, since different edge modifications are required sometimes to form junctions between graphene edges, as it happens when constructing junctions between carbon nanotubes. Such studies suggest whether localized or resonance in- terface states appear at the junctions and even allow to estimate their energies23. FIG. 9. Spectra of armchair GNRs with Klein nodes added to the edge. Left: 40(1,1) armchair GNR with one Klein node per unit cell. Right: the same GNR but with two Klein nodes per unit cell. and for some range of k they form gap states with E 6= 0. Similar bands occur in the bearded edges investigated in Ref. 19, where zigzag edges and Klein defects appear alternatively.26 Gap states with E 6= 0 are usually related to edges of mixed character. In this section we consider only two examples of such edges and explain their origin in more detail as an illustration of how our prescription applies to any kind of edges of mixed character. 1. Klein defects in armchair edges Let us consider the case of an armchair ribbon with one edge modified by attaching a Klein node. The cor- responding spectrum is shown in the upper left panel of Fig. 9. When both edges are modified the flat band is doubly degenerate.19,22 It is noteworthy that the same flat band appears no matter whether the Klein node is disconnected or attached. Our prescription explains this fact in a simple way: such a flat band has no partner band to hybridize and split. When a second Klein node is attached (see the bottom right panel of Fig. 9), the flat bands mix and split because they belong to different sub- lattices. However, they do not reach the band continuum and become E 6= 0 gap states. When we additionally connect the previous two Klein nodes we obtain again an armchair GNR. In this case the bands split even farther and merge into the continuum; we recover the spectrum of armchair GNR. FIG. 10. Spectra of 3(8,1) GNR close to the Fermi energy with different edge terminations, ranging from the minimal edge (a) to the modified edge with one Klein node, marked as A, added (b-c) or two Klein nodes A and B attached (d). In (b) t = 0.1t0, while in (c) and (d) t = t0. The top panels show the minimal and modified edges. The top panel of Fig. 10 shows the different edges in- vestigated: minimal (left) and modified (right) by attach- ing one, A, or two, A and B, Klein nodes. The spectrum of the GNR with minimal edges close to EF is shown in Fig. 10(a). Note that it is the same spectrum as in Fig. 4(right). The degeneracy of the E = 0 band results from the folding rules corresponding to the (7,0) GNR. When an extra non-connected node (marked as A in the Fig- ure) is added at each side of the GNR unit cell, a doubly degenerated E = 0 band appears in the spectrum. Thus, the degeneracies increase to six and eight for k < 2/3π and k > 2/3π, respectively. When we connect this extra node by t = 0.1t0, two flat bands hybridize and split, see Fig. 10(b). The hybridization and splitting can be explained by the attached diagrams. For t = t0 the split bands merge into the ribbon continuum of states and disappear from the gap region; the spectrum and the corresponding diagrams are shown in Fig. 10(c). If the other Klein node marked B is added to this edge, first as a non-connected node, the degeneracy of the E = 0 ribbon band increases by two. If the connection of the atom B is turned on, the bands mix and split. Fig. 10(d) shows the bands for t = t0. All the boxes in the diagram corresponding to k < 2/3π annihilate. Only a doubly degenerate zero-energy band survives for k > 2/3π. In the corresponding diagram two pairs of empty and filled boxes annihilate, leaving only one box in each row that corresponds to a doubly degenerate band at zero energy. The attached diagrams also indicate that the flat bands in both cases (c) and (d) localize in the sublattices corre- sponding to the zigzag edges. Our numerical calculations fully confirm again these predictions. A comment is required to explain why the addition of the second Klein node B yields weaker splitting of the flat bands than in the case where only a node A is attached. A closer inspection of the two wavefunctions localized at one edge at k = 0 in Fig. 10a reveals that one of them is nearest to a step in the edge, see upper panel. Such step- localized wavefunction is thus nearer to the Klein defects. The second wavefunction at B localizes away from the step. Attaching the first Klein node A yields a strong hybridization with the step-localized wavefunction. The wavefunction of the second Klein node B must hybridize with the remaining function, which localizes away from the Klein node. This explains why the mixing of B is weaker and its splitting smaller. IV. SUMMARY We have presented a simple prescription to predict, without performing any calculations, the existence of zero-energy bands in graphene nanoribbons and graphene edges of arbitrary shape. Our prescription is based on two observations: (a) any edge can be created from a 8 minimal edge by adding extra nodes, (b) the zero-energy spectra of graphene minimal edges are uniquely defined by the edge band structure of its zigzag (n, 0) component, which in turn is obtained by folding n times the spectrum of the zigzag (1,0) edge. Extra but disconnected nodes provide zero-energy states which, after connecting them to the graphene edge, may hybridize with the existing flat bands and split. The splitting occurs only when the extra node belongs to a different sublattice as that where the edge zero-energy bands are localized. We have intro- duced simple rules and diagrams allowing to precisely de- termine not only the existence of the flat bands, but also their degeneracies and localization of their wavefunctions on the graphene sublattices. The folding rules allow also to estimate the energy gaps which open in certain regions of k. Our prescription and diagrams hold for graphene edges and nanoribbons with arbitrary geomeries. How- ever, one has to remember that for extremely narrow rib- bons the interaction between the GNR edges may lead to the edge bands splitting and the loss of their flatness. We have considered a number of GNRs with differ- ent edges, some of them with attached Klein nodes or cape structures. They are also important in the study of graphene-based complex systems, since connecting differ- ent portions of graphene requires sometimes the modifi- cation of their edges. Finally, we have shown that our prescription predicts the localization properties of edge states. We have stud- ied a couple of cases by performing tight-binding and first-principles DFT calculations. The correspondence between the ab-initio and the tight-binding results back our prescriptions, with the caveat that within the DFT approach the edge bands are no longer exactly flat and with zero-energy. Our method can be widely used to foresee the existence of edge states and flat bands in any graphene edges and ribbons. V. ACKNOWLEDGMENTS We acknowledge financial support from Basque Depar- tamento de Educaci´on and the UPV/EHU (Grant No. IT-366-07), the Spanish Ministerio de Innovaci´on, Cien- cia y Tecnolog´ıa (Grants No. FIS2007-66711-C02-02 and FIS2009-08744) and the ETORTEK research program funded by the Basque Departamento de Industria and the Diputaci´on Foral de Guip´uzcoa. W.J. thanks Donostia International Physics Center, and A.A. and L.C. thank Nicolaus Copernicus University in Toru´n for hospitality. ∗ [email protected] 1 A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). 2 Y. Zhang, Y.-W. Tan, H. L. Stormer, and P. Kim, Nature (London) 438, 201 (2006). 3 C. L. Kane, and E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005) 4 Y. W. Son, M. L. Cohen, and S. G. Louie, Nature (London) 444, 347 (2006). 5 M. Fujita, K. Wakabayashi, K. Nakada, and K. Kusakabe, J. Phys. Soc. Jpn. 65, 1920 (1996). 6 K. Nakada, M. Fujita, G. Dresselhaus, and M.S. Dressel- haus, Phys. Rev. B 54, 17954 (1996). 7 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, Science 306, 5296 (2004). 8 L. Brey, and H. A. Fertig, Phys. Rev. B 73, 235411 (2006). 9 H. Zheng, Z. F. Wang, T. Luo, Q. W. Shi, and J. Chen, Phys. Rev. B 75, 165414 (2007). 10 Y. W. Son, M. L. Cohen, and S. G. Louie, Phys. Rev. Lett. 97, 216803 (2006). 11 A. Cresti and S. Roche, Phys. Rev. B 79, 233404 (2009). 12 Y. Niimi, T. Matsui, H. Kambara, K. Tagami, M. Tsukada, and Hiroshi Fukuyama, Phys. Rev. B 73, 085421 (2006). 13 Y. Kobayashi, K.-I. Fukui, T. Enoki, and K. Kusakabe, Phys. Rev. B 73, 125415 (2006). 14 Z. Liu, K. Suenaga, P. J. F. Harris, and S. Iijima, Phys. Rev. Lett. 102, 015501 (2009). 15 K. Suenaga and M. Koshino, Nature (London) 468, 1088 (2010). 16 K. Wakabayashi, Phys. Rev. B 64, 125428 (2001). 17 A. R. Akhmerov and C. W. Beenakker, Phys. Rev. B 77, 085423 (2008). T . 18 D. J. Klein, Chem. Phys. Lett. 217, 261 (1994). 19 K. Wakabayashi, S. Okada, R. Tomita, S. Fujimoto, and Y. Natsume, J. Phys. Soc. Jpn. 79, 034706 (2010). 20 For a width of 10 zigzag unit cells, we use SIESTA method to solve the Kohn-Sham equations using for the exchange- correlation functional the generalized gradient approach of 9 Perdew-Burke-Ernzerhof. J. M. Soler, E. Artacho, J. D. Gale, A. Garc´ıa, J. Junquera, P. Ordej´on, and D. S´anchez- Portal, J. Physics: Condensed Matter 14, 2745 (2002); J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996). The number of K points were converged and the atoms were relaxed until the forces were smaller than 0.02 eV/A. 21 The energy bands at E=0 of TB method split in DFT calculations. The small splitting arises from the spin po- larization at the edges [Ref. 4; A. Mananes, F. Duque, A. Ayuela, M. J. L´opez, and J. A. Alonso, Phys. Rev. B 78, 035432 (2008)], but it does not change our main TB con- clusions. For the cape studies, the inclusion of spin polar- ization removes the double degeneracy of E=0 band, but they remain close to the Fermi energy. 22 When both Klein nodes belong to the same sublattice the degeneracy is exact for all k, otherwise the bands will split for narrow ribbons. 23 H. Santos, A. Ayuela, W. Jaskolski, M. Pelc, and L. Chico, Phys. Rev. B. 80, 035436 (2009). 24 W. Yao, S. A. Yang, and Q. Niu, Phys. Rev. Lett. 102, 096801 (2009). 25 D. Gunlycke, D. A. Areshkin, and C. T. White, Appl. Phys. Lett 90, 142104 (2007). 26 The band structure of these ribbons with bearded edges can be easily explained with our prescription. It is sufficient to consider a (2,0) GNR with only one Klein defect added at each side of the GNR unit cell.
1812.00286	2	1812	2019-05-27T07:16:15	Rectification in Spin-Orbit Materials Using Low Energy Barrier Magnets	[ "cond-mat.mes-hall" ]	The coupling of spin-orbit materials to high energy barrier ($\sim$40-60 $k_BT$) nano-magnets has attracted growing interest for exciting new physics and various spintronic applications. We predict that a coupling between the spin-momentum locking (SML) observed in spin-orbit materials and low-energy barrier magnets (LBM) should exhibit a unique multi-terminal rectification for arbitrarily small amplitude channel currents. The basic idea is to measure the charge current induced spin accumulation in the SML channel in the form of a magnetization dependent voltage using an LBM, either with an in-plane or perpendicular anisotropy (IMA or PMA). The LBM feels an instantaneous spin-orbit torque due to the accumulated spins in the channel which causes the average magnetization to follow the current, leading to the non-linear rectification. We discuss the frequency band of this multi-terminal rectification which can be understood in terms of the angular momentum conservation in the LBM. For a fixed spin-current from the SML channel, the frequency band is same for LBMs with IMA and PMA, as long as they have the same total magnetic moment in a given volume. The proposed all-metallic structure could find application as highly sensitive passive rf detectors and as energy harvesters from weak ambient sources where standard technologies may not operate.	cond-mat.mes-hall	cond-mat	a Rectification in Spin-Orbit Materials Using Low Energy Barrier Magnets Shehrin Sayed,1, 2, ∗ Kerem Y. Camsari,1 Rafatul Faria,1 and Supriyo Datta1, † 1Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907, USA 2Electrical Engineering and Computer Science, University of California, Berkeley, CA 94720, USA The coupling of spin-orbit materials to high energy barrier (∼40-60 kBT ) nano-magnets has attracted growing interest for exciting new physics and various spintronic applications. We predict that a coupling between the spin-momentum locking (SML) observed in spin-orbit materials and low- energy barrier magnets (LBM) should exhibit a unique multi-terminal rectification for arbitrarily small amplitude channel currents. The basic idea is to measure the charge current induced spin accumulation in the SML channel in the form of a magnetization dependent voltage using an LBM, either with an in-plane or perpendicular anisotropy (IMA or PMA). The LBM feels an instantaneous spin-orbit torque due to the accumulated spins in the channel which causes the average magnetization to follow the current, leading to the non-linear rectification. We discuss the frequency band of this multi-terminal rectification which can be understood in terms of the angular momentum conservation in the LBM. For a fixed spin-current from the SML channel, the frequency band is same for LBMs with IMA and PMA, as long as they have the same total magnetic moment in a given volume. The proposed all-metallic structure could find application as highly sensitive passive rf detectors and as energy harvesters from weak ambient sources where standard technologies may not operate. I. INTRODUCTION The interplay between spin-orbit materials and nano- magnetism has attracted much attention for interesting phenomena e.g. spin-orbit torque switching [1, 2], prob- ing the spin-momentum locking [3 -- 8], spin amplification [9], spin battery [10], skyrmion dynamics [11, 12], among other examples. In this paper, we predict that the spin- momentum locking (SML) observed in spin-orbit mate- rials when coupled to a nano-magnet with low-energy barrier, will rectify the channel current in the form of a voltage in a multi-terminal structure. We start our arguments with the spin-potentiometric measurements well-established in diverse classes of spin-orbit materi- als (see, for example, [3 -- 8]) where a high-energy-barrier stable ferromagnet (FM) is used to measure the charge current induced spin potential in the SML channel. We show that such spin-potential measurement on a metallic SML channel using a low-energy barrier magnet (LBM) will result in a rectified voltage, even for arbitrarily small channel current. The discussions on the multi-terminal rectification is limited in the linear response regime of transport in the SML channel and the non-linearity occurs due to the spin-orbit torque (SOT) driven magnetization dynamics of the LBM. We further show that the multi-terminal rec- tification is limited by a characteristic frequency of the LBM that can be understood in terms of angular momen- tum conservation between the spins injected from spin- orbit materials and the spins absorbed by the LBM. We argue that, for a fixed spin-current from the SML chan- nel, the characteristic frequency is the same for LBMs with in-plane and perpendicular magnetic anisotropies ∗ [email protected] † [email protected] (IMA and PMA), as long as they have the same total magnetic moment in a given volume. We analyze the rectification in the proposed all- metallic structure (see Fig. 1(a)) considering both IMA and PMA LBMs and provide simple models to under- stand the underlying mechanisms of (i) the spin-orbit torque (SOT) induced magnetization pinning and (ii) the frequency band of the rectification. We compare the simple models with detailed numerical simulations using an experimentally benchmarked multi-physics framework [13]. The simulations are carried out using a transmis- sion line model for the SML channel [14] and a stochas- tic Landau-Lifshitz-Gilbert (s-LLG) model for LBM [15], considering thermal noise within the magnet. We con- sider the spin-polarization axis to be in-plane of the SML channel and orthogonal to the current flow direction. Hence, in the present discussion, pinning for a IMA or a PMA magnet occurs along the easy-axis or hard-axis, respectively. We argue that such wideband rectification in an all- metallic structure (Fig. 1(a)) could be used for 'passive' radio frequency (rf) detection. Recently, Magnetic Tun- nel Junction (MTJ) diodes with stable magnet as free layer and under an external dc current bias have demon- strated orders of magnitude higher sensitivity compared to the state-of-the-art Schottky diodes [16 -- 18]. However, the reported no-bias sensitivity is lower or comparable to that of semiconductor diodes. The low-barrier nature of the magnet in the proposed structure should exhibit no- bias sensitivity as high as those observed using state-of- the-art technologies under external biases [16 -- 18]. Fur- thermore, we discuss the possibility to harvest energy from weak ambient sources where standard technologies may not operate. The paper is organized as follows. In Section II, we establish the concept of the multi-terminal rectification in the SML channel using LBM, starting from the well- established spin-potentiometric measurements typically done with high-energy-barrier stable magnets. In Section III, we discuss the frequency bandwidth of the rectifica- tion and provide a simple model that applies to LBMs with both IMA and PMA. We argue using detailed sim- ulaiton results that such bandwidth arises due to the principles of angular momentum conservation between the spins injected from the SML channel and the spins absorbed by the LBM. In Section IV, we discuss possi- ble applications of the proposed all-metallic structure in 'passive' rf detection and energy harvesting. We argue that the no-bias sensitivity of the proposed rectifier can be as high as those observed in state-of-the-art technolo- gies under external bias. Finally, in Section V, we end with a brief conclusion. II. MULTI-TERMINAL RECTIFICATION We start our arguments with the well-established spin- potentiometric measurements [3 -- 8] where the charge cur- rent induced spin potential in the SML channel is mea- sured in the form of a magnetization dependent voltage using a stable ferromagnet (FM). The voltage at the FM with respect to a reference normal metal (NM) contact, placed at the same position along the current path as the FM (see Fig. 1(a)), is given by [19, 20] V34 ( (cid:126)m) = (s · (cid:126)m) αξp0pf RB 2 I12, (1) which shows opposite signs for the two magnetic states of the FM under a fixed channel current I12 flowing along n-direction (see Fig. 1(a)). Here s is the spin polariza- tion axis in the SML channel defined by y × n with y being the out-of-plane direction [14], (cid:126)m is the magneti- zation vector, pf is the FM polarization, 0 ≤ ξ ≤ 1 is the current shunting factor [21] of the contact with 0 and 1 indicating very high and very low shunting respectively, p0 is the degree of SML in the channel [14], α ≈ 2/π is an angular averaging factor [14], and RB = (h/q2)(1/Mt) is the ballistic resistance of the channel with total number of modes Mt (q : electron charge, h : Planck's constant). Note that Eq. (1) is valid all the way from ballistic to diffusive regime of operation [14, 19]. We restrict our dis- cussion to linear response where V34 ( (cid:126)m) in Eq. (1) scales linearly with I12 and satisfies the Onsager reciprocity re- lation [21, 22] Rij,kl ( (cid:126)m) = Rkl,ij (− (cid:126)m) , with Rij,kl = Vkl/Iij. The Onsager reciprocity does not require any specific relation between Rij,kl ( (cid:126)m) and Rij,kl (− (cid:126)m) in linear response and the phenomenon de- scribed by Eq. (1) has been observed on diverse spin- orbit materials e.g. insulator (TI) [3 -- 6], Kondo insulators [23], transition metals [7], semimetals [24], and semiconductors [8]. topological To measure Eq. (1) from a highly resistive SML chan- nel (e.g. TI [3 -- 6], semiconductor [8], etc.) using a metal- lic FM, usually a thin tunnel barrier is inserted at the FIG. 1. (a) Multi-terminal structure with a low barrier magnet (LBM) on top of a channel with spin-momentum locking (SML). LBM can be of in-plane and perpendicular anisotropies (IMA or PMA). (b) Average magnetization (cid:104)s· (cid:126)m(cid:105) of the LBM. (c) Average voltage (cid:104)V34(cid:105) between the LBM and a reference contact as a function of the input current I12. Inset shows zoomed version of (cid:104)V34(cid:105) for very small input current, which exhibits a parabolic nature. Simulations are compared with Eqs. (1) and (3). I0 for IMA and PMA are 80 µA and 1.6 µA respectively. Here, V0 = I0RB. interface. This tunnel barrier effectively enhances V34 by improving ξ [21], however, degrades the spin injec- tion into the FM from the SML channel. It has been recently demonstrated [7] that V34 can be measured with metallic FM in direct contact with metallic SML chan- nels (e.g. Pt, Ta, W, etc.), which indicates the possibility of spin-voltage reading (e.g. [7]) and spin-orbit torque (SOT) writing (e.g. [1, 2]) of the nano-magnet within same setup with different current magnitudes [20]. The energy barrier of a mono-domain magnet is given by ∆B = 1 2 HkMsΩ [25] where Hk is the anisotropy field, Ms is the saturation magnetization, and Ω is the FM volume. For a stable FM, ∆B ≈ 40 ∼ 60 kBT and exhibit very long retention time τ ∝ exp (∆B/kBT ) of the magnetization state (kB : Boltzmann constant, T : temperature). LBMs have very small τ and the s · (cid:126)m component becomes random within the range {+1,−1} driven by the thermal noise. Experimentally, LBMs have been achieved by lowering the total moment (MsΩ) [26] or by lowering the anisotropy field (Hk) either by increas- ing the thickness of a PMA [27], or by making a circular IMA with no shape anisotropy [28]. At equilibrium (I12 = 0), the time-averaged (cid:104)s· (cid:126)m(cid:105) = 0 for an LBM. For I12 (cid:54)= 0, induced non-equilibrium spins in the channel apply SOT on the LBM and (cid:104)s· (cid:126)m(cid:105) follows the accumulated spins, which can be calculated using (cid:90) θ=π (cid:90) φ=π (cid:90) θ=π (cid:90) φ=π φ=−π θ=0 φ=−π θ=0 (cid:104)s · (cid:126)m(cid:105) = (s · (cid:126)m) ρ sin θ dθdφ . (2) ρ sin θ dθdφ where ρ is the probability distribution function of the magnetization of the LBM under a particular I12, which can be obtained from the Fokker-Planck equation [29, 30]. The dependence of (cid:104)s· (cid:126)m(cid:105) on I12 deduced from Eq. (2) for a particular LBM, can in-principle be any saturating odd- functions e.g. Langevin function for low-barrier PMA (see Appendix A). We approximate (cid:104)s · (cid:126)m(cid:105) in Eq. (2) with a tanh func- tional dependence on I12, given by (cid:104)s · (cid:126)m(cid:105) ≈ tanh , (3) (cid:18) I12 (cid:19) I0 which is in good agreement with the detailed numeri- cal simulations for both IMA and PMA, as shown in Fig. 1(b). The simulations are carried out within a multi-physics framework [13] using our experimentally benchmarked transmission line model for SML [14] and stochastic Landau-Lifshitz-Gilbert (s-LLG) model for LBM [15] which considers thermal noise. The details of the simulation setup is discussed in Appendix B. Here, I0 is a parameter that determines the SOT in- duced magnetization pinning of the LBM. I0 depends on the temperature, geometry, and material parameters and much larger for an IMA as compared to a PMA due to the demagnetization field. I12 along ∓n-direction causes the magnetization pinning along ±s-direction. In the present discussion, easy axis for PMA is along y-direction, hence, the pinning occurs along the hard axis. The easy axis of IMA, in principle, can be in any direction on the plane spanned by n and s and the magnetization pinning along s-direction should be described by Eq. (3) with a modi- fied I0. However, we set the easy axis along ±s-direction in our IMA simulations for simplicity. For a given structure, I0 can be determined directly from experiments using a characteristic curve similar to that in Figs. 1(b) or (c). We provide a simple expression using Eq. (2) and considering easy-axis pinning of a PMA magnet (see Appendix A for the derivation), as given by I0 ≈ 6q kBT αg β , (4) where αg is the Gilbert damping and β is the charge to spin current conversion ratio. Eq. (4) is reasonably valid up to ∆B ≈ kBT and provides the correct order of magnitude up to several kBT (see Appendix A for details). In this discussion, we consider very low energy barrier (≤ 1kBT ) nano-magnets that do not have bistable states. A higher barrier magnet that exhibits bistable states, in principle could exhibit effects like stochastic For I12 (cid:29) I0 in Eq. resonance [31], which is not the subject of the present discussion. (3), we have tanh (I12/I0) ≈ +1 or −1 when I12 > 0 or I12 < 0 respectively. Hence, I12×tanh (I12/I0) ≈ I12. On the other hand, for I12 (cid:28) I0 in Eq. (3), we have I12×tanh (I12/I0) ≈ I 2 12/I0. Thus, the time-average of the voltage in Eq. (1) is given by (cid:18) αξp0pf RB (cid:18) αξp0pf RB 2I0 (cid:19) (cid:19)  2 (cid:104)V34(cid:105) = I 2 12, I12. for I12 (cid:28) I0 for I12 (cid:29) I0 (5) Note that (cid:104)V34(cid:105) represents the steady-state voltage of the capacitor CL placed between contacts 3 and 4. The rela- tive position between contacts 3 and 4 along s-direction do not affect Eq. (1), however, a shift in the n-direction creates an offset due to Ohmic drop [20], that should cancel out over averaging in Eq. (5) when an ac I12 is applied. For arbitrary I12, (cid:104)V34(cid:105) is always of the same sign leading to a multi-terminal rectification. This ob- servation agrees well with simulation results for IMA and PMA, as shown in Fig. 1(c). For I12 (cid:28) I0, (cid:104)V34(cid:105) ex- hibits a parabolic nature (see zoomed inset of Fig. 1(c)), as suggested by Eq. (5). All simulation results pre- sented in this paper are normalized by I0, V0 = I0RB, and f0 = I0/q for current, voltage, and frequency, re- spectively. In all simulations, Mt = 100 which yields RB = 259Ω. III. FREQUENCY BANDWIDTH The frequency bandwidth of the the multi-terminal rectification is limited by a characteristic frequency fc that is determined by the angular momentum conserva- tion between the spins injected from the SML channel and the spins absorbed by the LBM. We plot the (cid:104)V34(cid:105) as a function of the frequency f of the ac I12 = ic0 sin (2πf t) (see Fig. 2(a)) while other parameters are kept constant in our simulations. Note that (cid:104)V34(cid:105) is relatively constant in the low frequency region and degrades significantly for f > fc. We have defined fc as the frequency where (cid:104)V34(cid:105) degrades by an order of magnitude compared to the re- gion where (cid:104)V34(cid:105) vs. f is relatively flat. We observe the time-dynamics of s · (cid:126)m and V34 for the two cases indicated with red-dots in Fig. 2(a): (I) f < fc and (II) f > fc. For the first case, I12 is slow enough that the injected spins from SML channel into the LBM satisfies the angular momentum conservation and s · (cid:126)m follows the I12 at the same frequency, as shown in Fig. 2(b). This leads to a rectified voltage V34 that charges up the capacitor CL to the steady-state value (cid:104)V34(cid:105). The ripples observed in V34 is similar to those in conventional rectifiers and gets attenuated for increased CL. For the latter case, s · (cid:126)m struggles to follow I12 (see Fig. 2(c)) since the spins injected from the SML channel to the LBM is fast enough that they do not satisfy the angular FIG. 2. (a) Rectified voltage (cid:104)V34(cid:105) as a function of the input ac frequency f , showing the frequency bandwidth fc. We observe the time dynamics of the LBM under input current I12 for (b) f < fc, where s· (cid:126)m on average follows I12 leading to rectification and (c) f > fc, where s · (cid:126)m struggles to follow I12 and yields no net rectification. Results apply to both IMA and PMA. Here, we consider 49 nm × 61 nm × 5 nm LBM with Ms = 900 emu/cc. is0 = βic0 = 2 mA. momentum conservation. s · (cid:126)m has no correlation with I12, as a result, there is no rectification that charges up CL to a steady dc voltage. We obtain an empirical expression for fc from the de- tailed s-LLG simulations using a broad range of param- eter values, given by 2πfc = is0 2qNs , (6) where the injected spin current amplitude is0 = βic0, Ns = MsΩ/µB is the total number of spins in LBM and µB is the Bohr magneton. The functional dependence of fc on is0 and Ns is very similar to the switching delay for stable magnets [32] that also arises from the principles of angular momentum conservation. Note that Eq. (6) is valid for both IMA and PMA. We show comparison between Eq. (6) and simulation results in Fig. 3. The simulation data points shown on Fig. 3 are extracted from a plot similar to Fig. 2(a). Eq. (6) shows good agreement with the simulation for LBMs having IMA with easy-axis pinning and PMA with hard-axis pinning (see Figs. 3(a)-(b)). Fig. 3(a) shows that fc scales linearly with is0. A similar scenario has been reported [33] for a stochastic MTJ oscillator made with relatively lower-barrier free magnetic layer. Accord- ing to Eq. (6) and detailed simulations, the conclusion that fc ∝ is0 seems valid even if is0 changes by orders of magnitude. Moreover, Fig. 3(b) shows that fc scales inversely proportional to the Ns which depends only on MsΩ of the magnet and independent of the magnetic anisotropy. Eq. (6) could be useful for recent interest on LBM based applications e.g. stochastic oscillators [33], random number generators [26, 34], probabilistic spin logic [15, 35], etc. IV. APPLICATIONS: RF DETECTION AND ENERGY HARVESTING The proposed all-metallic structure can find useful ap- plications like rf detection and energy harvesting. In this section, we show that the low-barrier nature of the mag- net can lead to very high rf detection sensitivity without any external bias, comparable to those observed in state- of-the-art technologies under an external bias. We pro- vide a simple model for no-bias sensitivity which provides insight into the design of a high-sensitivity device. This could be of interest for rf detection from weak sources typically proposed to sense with quantum sensors (see, e.g., [36, 37]).We further argue that the proposed struc- ture can extract useful energy from the ambient rf en- ergy, especially from the weak sources where standard It can be seen from Eq. (5) that (cid:104)V34(cid:105) scales ∝ √ technologies may not operate. Pin when max (I12) (cid:29) I0, where Pin = 1 12 (t) R12 dt, T = 1/f , and R12 is the channel resistance. However, for max (I12) (cid:28) I0, (cid:104)V34(cid:105) scales ∝ Pin with a constant slope given by (cid:82) T 0 I 2 T d(cid:104)V34(cid:105) dPin = αξp0pf 2 RB I0R12 . (7) The derivation is given in Appendix C. The quantity in Eq. (7) is often considered as the sensitivity of rf detec- tors [16 -- 18]. Recently, Magnetic Tunnel Junction (MTJ) diodes with stable magnet as free layer and under an ex- ternal dc current bias have demonstrated orders of mag- nitude higher sensitivity compared to the state-of-the-art Schottky diodes [16 -- 18]. However, the reported no-bias sensitivity is lower or comparable to that of semicon- ductor diodes. Eq. (7) indicates that the no-external- bias sensitivity can be very high within the all-metallic structure in Fig. 1(a) when designed to have very low I0, enabling highly sensitive 'passive' rf detection. With αg = 0.01 and T = 300 K we have I0 ≈ 0.37µA/β from Eq. (4). For a Py LBM of dimension of 49 nm × 61 nm MTJ diodes recently demonstrated rf energy harvesting with similar efficiency [40], however, the input power was in the µW range. Such MTJs should achieve reasonable efficiency at lower input power if the stable free layer is replaced with an LBM. V. CONCLUSION In conclusion, we predict multi-terminal rectification in an all-metallic structure that comprises a spin-orbit material exhibiting spin-momentum locking (SML) and a low-energy barrier magnet (LBM) having either in-plane or perpendicular anisotropy (IMA or PMA). The discus- sion of such multi-terminal rectification was limited in the linear response regime of transport and the non-linearity occurs due to the spin-orbit torque driven magnetiza- tion dynamics of the LBM. We draw attention to a fre- quency band of the rectification which can be understood in terms of angular momentum conservation within the LBM. For a fixed spin-current from the SML channel, the frequency band is same for LBMs with IMA and PMA, as long as they have the same total magnetic moment for a given volume. We further discuss possible applications of the wideband rectification as highly sensitive passive rf detectors and as energy harvesters from ambient sources. ACKNOWLEDGMENTS This work was supported by ASCENT, one of six cen- ters in JUMP, a SRC program sponsored by DARPA. Appendix A: Average Magnetization of Low-Barrier Magnets and Magnetization Pinning Current This section discusses the pinning current of a LBM and derives Eqs. (3)-(4), starting from the steady-state solution of the Fokker-Planck Equation. (cid:20) (cid:18) (cid:18) Hext Hk (cid:19) (cid:19)(cid:21) mz , We start from the steady-state solution of probabil- ity distribution from Fokker-Planck Equation assuming a magnet with perpendicular magnetic anisotropy (PMA) (see Eq. (4.3) in Ref. [30]), given by ρ (mz) = 1 Z exp − ∆B kBT 1 − m2 z + 2 + is Is0 (A1) where Z is a normalizing factor, mz is the magnetiza- tion along easy-axis (s · (cid:126)m in the present discussion), ∆B = HkMsΩ/2 is the energy barrier of a magnet with anisotropy field Hk, saturation magnetization Ms, and volume Ω, kB is the Boltzmann constant, T is the tem- perature, Hext is the external magnetic field along the easy-axis, is is the z-polarized spin current injected into the magnet, and Is0 is the critical spin current for magne- tization switching [20, 25] for a magnet with PMA, given FIG. 3. Characteristic frequency fc is (a) proportional to ic0 (Ns = 104 and 106 for PMA and IMA respectively) and (b) inversely proportional to Ns. Here, f0 = I0/q. × 5 nm (see Ref. [27]) and 2 nm thick Pt channel, β can be ∼ 2 as estimated from the charge to spin conversion [2], yielding I0 ≈ 0.18 µA. Note ratio reported in Ref. that β can be much higher based on the geometry and the choice of the SML material. For Bi2Se3 and Pt, we roughly estimate the sensitivity as 21,000 and 860 mV/mW respectively, assuming 2D SML channel of width w = 210 nm and length L = 500 nm. These estimations were done based on Eq. (7) us- ing: (i) RB = 259 Ω (Bi2Se3) and 58 Ω (Pt), (ii) Rch ≈ 6.5 kΩ (Bi2Se3) and ∼ 3 kΩ (Pt), (iii) p0 ≈ 0.6 (Bi2Se3) and 0.05 (Pt) (see Ref. [20]), and (iv) pf ≈ 0.5 [8]. We have assumed ξ ≈ 1 and the quoted estimations will be lower for higher shunting. RB has been estimated using Mt = kF w/π, where kF = 1.5 nm−1 (Bi2Se3) and 6.7 nm−1 (Pt) [20]. The channel resistance has been esti- mated using Rch = RB(L + λ)/λ with mean free path λ of 20 nm (Bi2Se3 [38]) and 10 nm (Pt [39]), respectively. More detailed analysis and performance evaluation con- sidering signal-to-noise ratio we leave for future work. With proper materials and geometry, it may be pos- sible to extract usable energy from such rectification of rf signals, especially from weak ambient sources. The dc power PL = VLIL extracted by an arbitrary load RL is limited by the equivalent resistance R34 between contacts 3 and 4. The maximum efficiency of such rf to dc power conversion occurs for max (I12) (cid:29) I0, given by (cid:19)2 RB (cid:18) αξp0pf√ 2π ηmax = PL,max Pin = RB R34 . (8) R12 The derivation is given in Appendix D. Note that the maximum efficiency is independent of I0. Assuming Req = 10RB for enhanced ξ, we estimate the maximum efficiency to be 0.001% for Bi2Se3 and 3×10−6% for Pt even with Pin in the ∼pW range given I0 ≤ 0.18µA. by Is0 = 4q ∆Bαg, (A2) where = h/(2π) and αg is the Gilbert damping con- stant. We consider the case with no external field i.e. Hext = 0 which from Eqs. (A1) and (A2) gives − ∆B kBT (cid:0)1 − m2 (cid:1) − z   mz  . is 2q kBT αg ρ (mz) = 1 Z exp (A3) We consider very low energy barrier magnet i.e. ∆B kB T → 0, which in Eq. (A3) yields −  . ρ (mz) = 1 Z exp is 2q kBT αg mz (A4) The steady-state average (cid:104)mz(cid:105) is defined as (see Eq. (2)) (cid:104)mz(cid:105) = (cid:82) φ=π (cid:82) φ=π φ=−π (cid:82) θ=π (cid:82) θ=π θ=0 mz ρ(mz) sin θ dθdφ φ=−π θ=0 ρ(mz) sin θ dθdφ   − with (mz, mx, my) ≡ (cos θ, sin θ cos φ, sin θ sin φ). Com- bining Eq. (A5) with Eq. (A4) we get the long time av- eraged magnetization (cid:104)mz(cid:105) for a very low barrier PMA without external magnetic field as 2q kBT αg (cid:104)mz(cid:105) = coth is , (A6) is 2q kBT αg Note that Eq. (A6) was derived assuming ∆B which is a Langevin function L(x) of x ≡ is/(cid:0) 2q kBT αg (cid:1). kB T → 0, however, the expression remains reasonably valid up to ∆B ≈ kBT . We have compared Eq. (A6) with numer- (A3) and (A5) for ical calculations directly from Eqs. ∆B = 0.1kBT (see Fig. 4(a)) and kBT (see Fig. 4(b)) re- spectively, which shows reasonably good agreement. For ∆B > kBT , the simple expression in Eq. (A6) deviates from Eqs. (A3) and (A5). can approximate the Langevin function L(x) ≈ tanh hence For an estimation of the pinning spin current we x , 3   . (cid:104)mz(cid:105) ≈ tanh is 6q kBT αg (A7) Note that is from SML materials are related to input charge current ic with a conversion factor β given by is = βic. (A8) , (A5) Comparing Eq. (A9) with Eq. (3) yields FIG. 4. Comparison of simple expression in Eq. (A6) which assumes ∆B → 0 with the numerical calculation from Eqs. (A3) and (A5) for (a) ∆B = 0.1kBT and (b) ∆B = 1kBT . This comparison indicates that Eq. (A6) is reasonably valid for 0 ≤ ∆B ≤ 1kBT . Combining Eq. (A8) with Eq. (A7) yields   . (cid:104)mz(cid:105) ≈ tanh ic 6qkBT αg β (A9) I0 ≈ 6qkBT αg β , which gives the expression in Eq. (4). Appendix B: Simulation Setup This section provides the details of the simulation setup in SPICE that was used to analyze the proposed rectifier. We have discretized the structure in Fig. 1(a) into 100 small sections and represented each of the small sections with the corresponding circuit model. Note that each of the nodes in Fig. 5 are two component: charge (c) and z-component of spin (s). We have connected the charge and spin terminals of the models for all the small sec- tions in a modular fashion using standard circuit rules as shown in Fig. 5. The models are connected in a se- ries to reconstruct the structure along length direction. We have two of such parallel chains to take into account the structure along width direction and the two chains represent the area under the LBM and the reference NM respectively. The SML block with LBM is connected to a s-LLG block which takes the spin current from the SML block as input and self-consistently solves for mz and feeds back to the SML block. The contacts (1, 2, 3, and 4) in this discussion are point contacts. The polarization of contacts 1, 2, and 4 are pf = 0 since they represent normal metals. Polar- ization of contact 3 is 0.8 which represents an LBM. We set the total number of modes M + N in the channel to erage as (cid:104)V34(cid:105) = = 1 T 1 T (cid:90) T 0 V34 dt (cid:90) T 0 (cid:18) I12(t) (cid:19) I0 tanh (C1) I12(t) dt. αξp0pf 2GB Note that the timed average of the random fluctuation in LBM is zero. Here, GB = 1/RB. We apply an alternating current as input, given by I12(t) = ic0 sin , (C2) The average ac input power applied to the channel with resistance R12 is given by (cid:18) 2πt (cid:19) T 1 T i2 c0R12 (cid:90) T (cid:18) ic0√ T 0 2 (cid:90) T (cid:19)2 0 R12. Pin = = = I 2 12(t)R12 dt (cid:18) 2πt (cid:19) dt (C3) sin2 T 1. Case I: ic0 (cid:29) I0 For ic0 (cid:29) I0, we get tanh (I12(t)/I0) ≈ +1 when I12(t) > 0 and tanh (I12(t)/I0) ≈ −1 when I12(t) < 0. Thus we have tanh × I12(t) ≈ I12(t), (cid:18) I12(t) I0 and from Eq. (C1), we get (cid:19) (cid:90) T (cid:90) T (cid:90) T (cid:90) T 0 0 0 2 ic0 ic0 T 2 ic0. (cid:104)V34(cid:105) = = = + = 1 T 1 T αξp0pf 2GB αξp0pf 2GB 1 T 1 T αξp0pf 2GB αξp0pf 2GB αξp0pf 2GB ic0 × 2 π I12(t) dt (cid:19)(cid:12)(cid:12)(cid:12)(cid:12) dt (cid:12)(cid:12)(cid:12)(cid:12)sin (cid:18) 2πt (cid:19) (cid:18) 2πt (cid:18) 2πt (cid:19) sin dt T T − sin dt T FIG. 5. SPICE simulation setup for the structure shown in Fig. 1(a). SML channel is modeled by connecting SPICE compatible transmission line model [14] in a distributed man- ner. The LBM is modeled with stochastic Landau-Lifshitz- Gilbert (s-LLG) SPICE model [15]. be 100. We have assumed that the reflection with spin- flip scattering mechanism is dominant in the channel i.e. rs1,2 (cid:29) r, ts. The scattering rate per unit mode was set to 0.04 per lattice point. We apply the charge open and spin ground boundary condition at the two boundaries given by (cid:27) (cid:26) ic vs = L (cid:27) (cid:26) 0 0 (cid:27) (cid:26) ic vs = R (cid:27) (cid:26) 0 0 , and . (B1) Here, ic and vs indicates boundary charge current and boundary spin voltage. Indices L and R indicate left and right boundaries respectively. Both charge and spin terminals of contact 1 and 2 and the two boundaries of the two parallel model chains are connected together. We apply a current ic at the charge terminal of contact 1 and make the spin terminal grounded to take into account the spin relaxation process in the contact. We ground both charge and spin termi- nals of contact 2. The boundary conditions of contacts 1 and 2 are given by (cid:27) (cid:26) ic vs = 1 (cid:27) (cid:26) ic 0 (cid:27) (cid:26) vc vs = 2 (cid:27) (cid:26) 0 0 , and . (B2) We place a capacitor CL and load RL across the charge terminals of contacts 3 and 4. The spin terminals of contacts 3 and 4 are grounded. The boundary conditions of the contacts 3 and 4 are given by (cid:27) (cid:26) ic vs = 3 (cid:27) (cid:26) 0 0 , and (cid:27) (cid:26) ic vs = 4 (cid:27) (cid:26) 0 0 (C4) (C5) . (B3) We write Eq. (C4) as (cid:104)V34(cid:105) = αξp0pf πGB × ×(cid:112) Pin, √ 2√ R12 Appendix C: Sensitivity This section discusses the detailed derivation of sensitivity model in Eq. (7). the We start from Eq. (1) with (cid:126)m(t) being the instanta- neous magnetization of the LBM. and calculate the av- d(cid:104)V34(cid:105) dPin √ and the sensitivity is given by × = αξp0pf πGB 1√ 2R12 × 1√ Pin (C6) The sensitivity for ic0 (cid:29) I0 decreases inversely pro- portional to Pin. Sensitivity increases for decreasing Pin and eventually saturates to a maximum value for ic0 (cid:28) I0. . 2. Case II: ic0 (cid:28) I0 Under the no load condition (RL → ∞), we have the open circuit dc voltage from Eq. (C8) for ic0 (cid:28) I0 For ic0 (cid:28) I0, we get tanh (I12(t)/I0) ≈ I12(t)/I0. Thus from Eq. (C1), we get (cid:104)V34(cid:105) = = = 1 T 1 T αξp0pf 2GBI0 αξp0pf 2GB αξp0pf 2GB × We write Eq. (C7) as (cid:90) T (cid:90) T (cid:0)ic0/ 0 i2 c0 I0 0 √ I0 I 2 12(t) dt sin2 2(cid:1)2 . (cid:18) 2πt (cid:19) dt (C7) T (cid:104)V34(cid:105) = αξp0pf 2GBR12I0 Pin, (C8) and the sensitivity is given as d(cid:104)V34(cid:105) dPin = αξp0pf 2GBR12I0 , which gives the maximum sensitivity in Eq. (7). η = dPL,max dPin Appendix D: Power Conversion Efficiency This section discusses the ac to dc power conversion efficiency and provides the details of derivation of Eq. (8). ηmax = (cid:104)V34(cid:105) = αξp0pf 2GBR12I0 Pin, and from Eq. (C5) we know that for ic0 (cid:29) I0 (cid:104)V34(cid:105) = αξp0pf πGB × ×(cid:112) Pin. √ 2√ R12 Under the short circuit condition (RL → 0), we have the short circuit dc current ILRL→0 = (cid:104)V34(cid:105)/R34, where R34 is the equivalent resistance between the LBM and the reference NM. The maximum power transferred to the load is given by PL,max = 1 4 × VLRL→∞ × ILRL→0 = (cid:104)V34(cid:105)2 4R34 . (D1) in 4R34 (cid:19)2 P 2 2GBR12I0 (cid:18) αξp0pf (cid:19)2 (cid:18) αξp0pf (cid:18) αξp0pf (cid:18) αξp0pf (cid:19)2 2GBR12I0 πGB = Pin 2R12R34 (cid:19)2 Pin 2R34 for ic0 (cid:28) I0 for ic0 (cid:29) I0. (D2) for ic0 (cid:28) I0 The ac to dc power conversion efficiency is given by = for ic0 (cid:29) I0. (D3) Note that η increases with input ac power Pin and reaches a maximum when ic0 (cid:29) I0 given by 2R12R34 πGB 1 (cid:18) αξp0pf (cid:19)2 1 . πGB 2R12R34 which yields PL,max = (C9) = [1] L. Liu, C.-F. Pai, Y. Li, H. W. Tseng, D. C. Ralph, and R. A. Buhrman, "Spin-torque switching with the gi- ant spin hall effect of tantalum," Science 336, 555 -- 558 (2012). [2] L. Liu, O. J. Lee, T. J. Gudmundsen, D. C. Ralph, and R. A. Buhrman, "Current-induced switching of perpen- dicularly magnetized magnetic layers using spin torque from the spin hall effect," Phys. Rev. Lett. 109, 096602 (2012). [3] C. H. Li, O. M. van 't Erve, J. T. Robinson, Y. Liu, L. Li, and J. B. T., "Electrical detection of charge- current-induced spin polarization due to spin-momentum locking in bi2se3," Nature Nanotechnol. 9, 20325 (2014). [4] S. M. Hus, X.-G. Zhang, G. D. Nguyen, W. Ko, A. P. Baddorf, Y. P. Chen, and A.-P. Li, "Detection of the spin-chemical potential in topological insulators using spin-polarized four-probe stm," Phys. Rev. Lett. 119, 137202 (2017). [5] L. Liu, A. Richardella, I. Garate, Y. Zhu, N. Samarth, and C.-T. Chen, "Spin-polarized tunneling study of spin- momentum locking in topological insulators," Phys. Rev. B 91, 235437 (2015). [6] A. Dankert, J. Geurs, M. V. Kamalakar, S. Charpentier, and S. P. Dash, "Room temperature electrical detection of spin polarized currents in topological insulators," Nano Lett. 15, 7976 -- 7981 (2015). [7] V. T. Pham, L. Vila, G. Zahnd, A. Marty, W. Savero- Torres, M. Jamet, and J.-P. Attan, "Ferromag- netic/nonmagnetic nanostructures for the electrical mea- surement of the spin hall effect," Nano Lett. 16, 6755 -- 6760 (2016). [8] J.-H. Lee, H.-J. Kim, J. Chang, S. H. Han, H.-C. Koo, S. Sayed, S. Hong, and S. Datta, "Multi-terminal spin valve in a strong rashba channel exhibiting three resis- tance states," Scientific Reports 8, 3397 (2018). [9] K. M. M. Habib, R. N. Sajjad, and A. W. Ghosh, "Chi- ral tunneling of topological states: Towards the efficient generation of spin current using spin-momentum lock- ing," Phys. Rev. Lett. 114, 176801 (2015). [10] J. Tian, S. Hong, I. Miotkowski, S. Datta, and Y. P. Chen, "Observation of current-induced, long-lived per- sistent spin polarization in a topological insulator: A rechargeable spin battery," Science Advances 3, e1602531 (2017). [11] S. Woo, K. M. Song, H.-S. Han, M.-S. Jung, M.-Y. Im, K.-S. Lee, K. S. Song, P. Fischer, J.-I. Hong, J. W. Choi, B.-C. Min, H. C. Koo, and J. Chang, "Spin-orbit torque- driven skyrmion dynamics revealed by time-resolved x- ray microscopy," Nature Commun. 8, 15573 (2017). [12] G. Yu, P. Upadhyaya, X. Li, W. Li, S. K. Kim, Y. Fan, K. L. Wong, Y. Tserkovnyak, P. K. Amiri, and K. L. Wang, "Room-temperature creation and spinorbit torque manipulation of skyrmions in thin films with engineered asymmetry," Nano Letters 16, 1981 -- 1988 (2016). [13] K. Y. Camsari, S. Ganguly, and S. Datta, "Modular tion of candidate topological kondo insulator smb6," arXiv:1809.04977 [cond-mat.str-el] (2018). [24] P. Li, W. Wu, Y. Wen, C. Zhang, J. Zhang, S. Zhang, Z. Yu, S. A. Yang, A. Manchon, and X.-x. Zhang, "Spin- momentum locking and spin-orbit torques in magnetic nano-heterojunctions composed of weyl semimetal wte2," Nature Communications 9, 3990 (2018). [25] J. Z. Sun, "Spin-current interaction with a monodomain magnetic body: A model study," Phys. Rev. B 62, 570 -- 578 (2000). [26] D. Vodenicarevic, N. Locatelli, A. Mizrahi, J. S. Fried- man, A. F. Vincent, M. Romera, A. Fukushima, K. Yakushiji, H. Kubota, S. Yuasa, S. Tiwari, J. Grollier, and D. Querlioz, "Low-energy truly random number gen- eration with superparamagnetic tunnel junctions for un- conventional computing," Phys. Rev. Applied 8, 054045 (2017). [27] P. Debashis, R. Faria, K. Y. Camsari, and Z. Chen, "Design of stochastic nanomagnets for probabilistic spin logic," IEEE Magnetics Letters 9, 1 -- 5 (2018). [28] P. Debashis, R. Faria, K. Y. Camsari, J. Appenzeller, S. Datta, and Z. Chen, "Experimental demonstration of nanomagnet networks as hardware for ising comput- ing," in 2016 IEEE International Electron Devices Meet- ing (IEDM) (2016) pp. 34.3.1 -- 34.3.4. approach to spintronics," Sci. Rep. 5, 10571 (2015). [29] W. F. Brown, "Thermal fluctuations of a single-domain [14] S. Sayed, S. Hong, and S. Datta, "Transmission-line model for materials with spin-momentum locking," Phys. Rev. Applied 10, 054044 (2018). [15] K. Y. Camsari, R. Faria, B. M. Sutton, and S. Datta, "Stochastic p-bits for invertible logic," Phys. Rev. X 7, 031014 (2017). [16] S. Miwa, S. Ishibashi, H. Tomita, T. Nozaki, E. Tamura, K. Ando, N. Mizuochi, T. Saruya, H. Kubota, K. Yakushiji, T. Taniguchi, H. Imamura, A. Fukushima, S. Yuasa, and Y. Suzuki, "Highly sensitive nanoscale spin-torque diode," Nature Materials 13, 50 -- 56 (2013). [17] B. Fang, M. Carpentieri, X. Hao, H. Jiang, J. A. Ka- tine, I. N. Krivorotov, B. Ocker, J. Langer, K. L. Wang, B. Zhang, B. Azzerboni, P. K. Amiri, G. Finocchio, and Z. Zeng, "Giant spin-torque diode sensitivity in the ab- sence of bias magnetic field," Nature Communications 7, 11259 (2016). [18] L. Zhang, B. Fang, J. Cai, M. Carpentieri, V. Puliafito, F. Garesc, P. K. Amiri, G. Finocchio, and Z. Zeng, "Ultrahigh detection sensitivity exceeding 105 v/w in spin-torque diode," Applied Physics Letters 113, 102401 (2018). [19] S. Hong, V. Diep, S. Datta, and Y. P. Chen, "Modeling potentiometric measurements in topological insulators including parallel channels," Phys. Rev. B 86, 085131 (2012). [20] S. Sayed, S. Hong, E. E. Marinero, and S. Datta, "Pro- posal of a single nano-magnet memory device," IEEE Electron Device Letters 38, 1665 -- 1668 (2017). [21] S. Sayed, S. Hong, and S. Datta, "Multi-terminal spin valve on channels with spin-momentum locking," Sci. Rep. 6, 35658 (2016). [22] P. Jacquod, R. S. Whitney, J. Meair, and M. Buttiker, "Onsager relations in coupled electric, thermoelectric, and spin transport: The tenfold way," Phys. Rev. B 86, 155118 (2012). [23] J. Kim, C. Jang, X. Wang, J. Paglione, S. Hong, and D. Kim, "Electrical detection of surface spin polariza- particle," Phys. Rev. 130, 1677 -- 1686 (1963). [30] W. H. Butler, T. Mewes, C. K. A. Mewes, P. B. Visscher, W. H. Rippard, S. E. Russek, and R. Heindl, "Switch- ing distributions for perpendicular spin-torque devices within the macrospin approximation," IEEE Transac- tions on Magnetics 48, 4684 -- 4700 (2012). [31] X. Cheng, C. T. Boone, J. Zhu, and I. N. Krivoro- tov, "Nonadiabatic stochastic resonance of a nanomagnet excited by spin torque," Phys. Rev. Lett. 105, 047202 (2010). [32] B. Behin-Aein, A. Sarkar, S. Srinivasan, and S. Datta, "Switching energy-delay of all spin logic devices," Appl. Phys. Lett. 98, 123510 (2011). [33] N. Locatelli, A. Mizrahi, A. Accioly, R. Matsumoto, A. Fukushima, H. Kubota, S. Yuasa, V. Cros, L. G. Pereira, D. Querlioz, J.-V. Kim, and J. Grollier, "Noise- enhanced synchronization of stochastic magnetic oscilla- tors," Phys. Rev. Applied 2, 034009 (2014). [34] B. Parks, M. Bapna, J. Igbokwe, H. Almasi, W. Wang, and S. A. Majetich, "Superparamagnetic perpendicular magnetic tunnel junctions for true random number gen- erators," AIP Advances 8, 055903 (2018). [35] A. Mizrahi, T. Hirtzlin, A. Fukushima, H. Kubota, S. Yuasa, J. Grollier, and D. Querlioz, "Neural-like com- puting with populations of superparamagnetic basis func- tions," Nature Commun. 9, 1533 (2018). [36] M. G. House, T. Kobayashi, B. Weber, S. J. Hile, T. F. Watson, J. van der Heijden, S. Rogge, and M. Y. Sim- mons, "Radio frequency measurements of tunnel cou- plings and singlet-triplet spin states in si:p quantum dots," Nat. Commun. 6, 8848 (2015). [37] M. F. Gely, M. Kounalakis, C. Dickel, J. Dalle, R. Vatr´e, B. Baker, M. D. Jenkins, and G. A. Steele, "Observation and stabilization of photonic fock states in a hot radio- frequency resonator," Science 363, 1072 -- 1075 (2019), https://science.sciencemag.org/content/363/6431/1072.full.pdf. [38] W. J. Wang, K. H. Gao, dependent transport channels in topological and Z. Q. Li, "Thickness- insulator bi2se3 thin films grown by magnetron sputtering," Sci- entific Reports 6, 25291 (2016). [39] G. Fischer, H. Hoffmann, and J. Vancea, "Mean free path and density of conductance electrons in platinum determined by the size effect in extremely thin films," Phys. Rev. B 22, 6065 -- 6073 (1980). [40] B. Fang, M. Carpentieri, S. Louis, V. Tiberkevich, A. Slavin, I. N. Krivorotov, R. Tomasello, A. Gior- dano, H. Jiang, J. Cai, Y. Fan, Z. Zhang, B. Zhang, J. A. Katine, K. L. Wang, P. K. Amiri, G. Finocchio, and Z. Zeng, "Experimental demonstration of spintronic broadband microwave detectors and their capability for powering nanodevices," Phys. Rev. Applied 11, 014022 (2019).
1010.1794	1	1010	2010-10-08T23:14:26	Noiseless nonreciprocity in a parametric active device	[ "cond-mat.mes-hall", "quant-ph" ]	Nonreciprocal devices such as circulators and isolators belong to an important class of microwave components employed in applications like the measurement of mesoscopic circuits at cryogenic temperatures. The measurement protocols usually involve an amplification chain which relies on circulators to separate input and output channels and to suppress backaction from different stages on the sample under test. In these devices the usual reciprocal symmetry of circuits is broken by the phenomenon of Faraday rotation based on magnetic materials and fields. However, magnets are averse to on-chip integration, and magnetic fields are deleterious to delicate superconducting devices. Here we present a new proposal combining two stages of parametric modulation emulating the action of a circulator. It is devoid of magnetic components and suitable for on-chip integration. As the design is free of any dissipative elements and based on reversible operation, the device operates noiselessly, giving it an important advantage over other nonreciprocal active devices for quantum information processing applications.	cond-mat.mes-hall	cond-mat	Noiseless nonreciprocity in a parametric active device Archana Kamal,1 John Clarke,2 M. H. Devoret1∗ 1Department of Physics and Applied Physics, Yale University, 15 Prospect Street, New Haven, CT 06520, USA 2Department of Physics, University of California, and Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA ∗Correspondence to: M. H. Devoret1 email: [email protected]. Nonreciprocal devices such as circulators and isolators belong to an impor- tant class of microwave components employed in applications like the mea- surement of mesoscopic circuits at cryogenic temperatures [1, 2, 3, 4]. The measurement protocols usually involve an amplification chain which relies on circulators to separate input and output channels and to suppress backaction from different stages on the sample under test. In these devices the usual recip- rocal symmetry of circuits is broken by the phenomenon of Faraday rotation based on magnetic materials and fields [5]. However, magnets are averse to on-chip integration, and magnetic fields are deleterious to delicate supercon- ducting devices [6, 7, 8]. Here we present a new proposal combining two stages of parametric modulation emulating the action of a circulator. It is devoid of magnetic components and suitable for on-chip integration. As the design is free of any dissipative elements and based on reversible operation, the device operates noiselessly, giving it an important advantage over other nonrecipro- cal active devices for quantum information processing applications. 1 arXiv:1010.1794v1 [cond-mat.mes-hall] 8 Oct 2010 Reciprocity is one of the fundamental symmetries frequently encountered in electrical cir- cuits. It is equivalent to the more familiar notion of the principle of reversibility in optics which states that any experiment is symmetric under an exchange of source and image [9]. Reci- procity can, however, be violated, for example, by the magneto-optic effect of Faraday rotation [5] which leads to rotation of the polarization vector of light resulting from different propaga- tion velocities of left- and right-circularly polarized waves in the presence of an applied mag- netic field B parallel to the direction of propagation (Fig. 1). The nonreciprocal phenomenon of Faraday rotation should be contrasted with the superficially similar, though reciprocal, ef- fect of optical activity where the polarization vector of light is rotated on passage through a non-centrosymmetric (chiral) medium. This change in the sense of rotation of polarization for counterpropagating waves in a Faraday medium (as seen by the observer receiving the light) has led some physicists to refer to this effect as a form of time-reversal-symmetry-breaking, a use of words that we prefer to avoid here [10, 11]. The phenomenon of nonreciprocity has propelled numerous theoretical investigations [[12] and refs. therein]; furthermore it offers immediate practical applications. Recent progress in solid state superconducting qubits, that provide some of the most promising architectures for scalable quantum computers [13, 14], has generated a huge incentive to integrate the compo- nents required for qubit operations and readout on-chip for incorporation in future quantum mechanical processors. A large variety of qubit readout protocols involve microwave reflec- tion based measurements and rely on nonreciprocal devices like circulators (or isolators) for separation of input and output channels [1, 2]. These devices also play a strategic role in mea- surements based on low-noise microwave parametric amplifiers which, with the exception of designs based on the current-biased dc SQUID (Superconducting Quantum Interference De- vice), are also operated in a reflection mode with both the input and output signals collected on the same spatial channel [15, 16]. However, circulators (and isolators) routinely use bulk 2 components made from ferrites to achieve nonreciprocal phase shifts (Fig. 1c) through Faraday rotation, making them unamenable to chip fabrication. Moreover, to bias the magnetic field in the ferrite, most of these devices use a permanent magnet which may channel flux into the superconducting device under test. In this letter, we present the full analysis of a model for a four-port circulator based on parametric active devices with no magnetic components. In active devices the energy source -- provided by the pump -- acts as the external "bias" field and sets the reference phase for the system, in analogy with the role played by the magnetic field in a Faraday medium. We exploit this effect in a cascade of active devices with pump phases at each stage tuned appropriately to obtain nonreciprocal transmission. The main building block of our design is a reversible IQ (in-phase/quadrature) modulator capable of performing noiseless frequency up- and down-conversion. A convenient analytical model capturing the fundamental properties of the device is shown in Fig. 2a. The device com- prises two low frequency LC resonators (addressed by two semi -infinite transmission lines A and B) coupled to a high frequency resonator (addressed by the transmission line C) through time-varying couplings M1, M2 that emulate the role of the pump drive in active nonlinear devices and transfer energy from the tone at ωc to the signal modes propagating on the trans- mission lines. It operates in a manner analogous to the IQ modulation schemes routinely used in radiofrequency (RF) communication systems and microwave pulse engineering (hence the name) and converts two orthogonal spatial modes travelling on two distinct spatial ports (A, B) at same frequency (here ω0) into two orthogonal temporal modes travelling on the same spa- tial line (C) at different frequencies (ω+, ω−). In view of the reversible frequency conversion performed by this device (Fig. 2a), we will henceforth refer to it as the up/down-converter (UDC). In practice, such a device can be implemented on-chip using a ring modulator based on Josephson junctions, along the lines of the recently demonstrated experiment with Josephson 3 parametric converter [17, 18]. The complete design for the active circulator (Fig. 3a) consists of a UDC functioning as a frequency up-converter, a phase-shifter and a second UDC functioning as a frequency down- converter. A concise representation of the dynamics at each of the three stages in the cascade is pro- vided by the scattering matrix S which relates the outgoing wave amplitudes to the incoming wave amplitudes as seen from the ports of a network. We start by deriving the scattering matrix of the UDC stage. This is done by evaluating the impedance matrix Z of the UDC, as seen from its ports, and using the identity [5] where S = (Z + Z0)−1 × (Z − Z0) Z0 = diag(ZA, ZB, ZC, ZC), (1) (2) with ZA = ZB and ZC denoting the characteristic impedances of the semi-infinite transmis- sion lines serving as low and high frequency ports respectively. We obtain (see supplementary −q0 r0 r0 q0 −itueiφ tueiφ −sue−iφ −isue−iφ tde−iφ sdeiφ itde−iφ −isdeiφ r+ 0 0 r−           a0in 0 a00in 0 bin + b†in −      . (3) information, Fig. S1) a0out 0 a00out 0 bout + b†out −           = Here a and b denote the (reduced) amplitudes or the annihilation operators for the waves travel- ling on left and right transmission lines respectively (see supplementary information for details). These satisfy bosonic commutation relations of the form [19] [ai, a†j] = δ(ωi − ωj). (4) In writing Eqs. (3) and (4), we have set a0 = a[ω0], a+ = a[ω+], a− = a[ω−] (see Fig. 2b). Similarly, the reflection coefficients at various ports are denoted by r0, r+ and r−. The cross 4 reflection between the low frequency signal ports is denoted by q0. The transmission coefficients are written as t (transmission without conjugation) and s (transmission with conjugation) with subscripts (u, d) indicating the up-conversion and down-conversion respectively. It is useful to note that the phase of the carrier, denoted by φ, affects only the transmitted amplitudes and rotates the two sidebands in opposite directions as can be seen from the corresponding scattering coefficients s and t in Eq. (3). The invariance of reflection amplitudes to the phase of the coupling will be important in understanding total reflections of the cascade, as we describe later. Further, we note that the matrix obtained in Eq. (3) is non-unitary, that is S†S 6= 1, which implies nonconservation of photon number as is natural for an active device. The matrix re- covers its unitary form as we turn off the couplings M1, M2 responsible for energy transfer between the pump and the signal modes. The full 8× 8 matrix [supplementary information, Eq. (S18)], describing the device operation for all modes and their respective conjugates, fulfils the fundamental requirement of symplecticity[17]. We can similarly describe the action of the frequency-independent phase shifting (PS) stage using a scattering matrix of the form      aout + a†out − bout + b†out −      = 0 0 e−iθ 0 0 0 0 eiθ e−iθ 0 0 0 0 eiθ 0 0           ain + a†in − bin + b†in −      . (5) For each stage of the cascade, we now go from the scattering matrix representation to the transfer matrix representation [5],      = T ain + aout + a†in − a†out −      , (6) bout in + b†out †in − −b +b      aout + a†out − bout + b†out −      = S ain + a†in − bin + b†in −     7→       since it is straightforward to calculate the total transfer matrix of the device by multiplying the 5 respective transfer matrices of different stages [20], Ttotal = TDCR × TP S × TU CL. (7) Here the subscripts L, R index the left hand upconversion (U C) and right hand downconversion (DC) stage (Fig. 3a). The scattering matrix of the whole device is then obtained from Ttotal using the inverse of the transformation in Eq. (6) (see supplementary information). We also note that TDC = F −1 × T −1 U C × F = F × T −1 U C × F, (F −1 = F ) (8) where F = σX ⊗ I2, (σX is the 2D pauli spin matrix and I2 is the 2D unity matrix). This matrix F is required to flip the indices, thus maintaining consistency in labelling the 'in' and 'out' amplitudes along a given direction of propagation. In our analysis we consider the operation at resonance, that is, when the input signal fre- quency coincides with the band center of the input resonators. Setting the phase of the pump at the first UDC stage φL = 0 for calculational simplicity, we observe a transmission resonance for θ = ±π/2 (phase rotation by the PS stage), φR = π/4 (phase of the pump at the second UDC stage), δ± = 1/√2 (detuning of the sidebands from the carrier in units of linewidth i.e. half width at full maximum of the resonance lineshape) of the high frequency resonator), and αL = αR = M0/qLA,BLC = 2−3/4 (strength of the parametric coupling). For this choice of parameters, we obtain the scattering matrix of the complete device as Stotal = 0 0 i 0 0 0 0 −i 0 0 i 0 i 0 0 0      . (9) This is the matrix of a perfect four-port circulator. The analogy between a conventional circula- tor and the active circulator design proposed in this paper is made apparent from the respective 6 wave propagation diagrams in Figs. 1c and 3b (see supplementary information for details on the calculation of coefficients on different arms in 3b). Nonetheless there are important differ- ences between the two designs despite the identity of the final S matrix. The coefficients on the forward (green) and backward (red) propagating arms of the active circulator design (Fig. 3b) involve deamplification followed by amplification, unlike the passive splitters (90 or 180 degree hybrids) employed in Faraday rotation schemes. This can be observed by squaring the ampli- tudes on each of the two arms originating from (or terminating into) a port and calculating the net power output, for each isolated UDC stage. It is straightforward to observe that, unlike the case of Fig. 1c, they do not add up to unity. Nonetheless, the overall transmission is unity due to an exact cancellation of the reduction and gain in amplitudes. The wave propagation diagrams in Fig. 3b reveal another important difference of this design from that of a conventional circulator. The non-reciprocal action of the active circulator is not based upon any non-reciprocal phase shifters; instead it relies on the active stages used for frequency up- and down-conversion. The phase matching condition in the forward direction is met by tuning the phase of the coupling at the input and output UDC stages. In the reverse direction the phase mismatch leads to unity transmission in the spatially orthogonal port instead, leading to complete isolation between the incident signal port and its corresponding output port. Fig. 3c shows a convenient method to visualize this circulator action geometrically by mapping the device dynamics at different stages using a modulation ellipse. This approach is inspired by the polarization ellipse used to represent of state of polarization of an electromag- netic wave (linear, circular or elliptical), which involves recording the trajectory traced out by the tip of the polarization vector of light (defined by the instantaneous direction of the electric field vector E) in a plane perpendicular to the direction of propagation. Equivalently, a two- dimensional representation of the components EX and EY of the electric field in the complex plane can be used to obtain a geometric description of the polarization of the light wave. In 7 the case of a modulation ellipse representation of the dynamics of the proposed device, we ex- 0, y = a00 tend this idea to map two distinct orthogonal modes (x,y) at each stage of the device [spatial: (x = a0 0) or temporal: (x = a+, y = a−)] as an ellipse in the plane defined by the coordinates I = Re[x + y], Q = Im[x− y∗]. This exercise shows that the final ellipses obtained at the output in case of forward and backward propagation through the device are rotated by 90 degrees with respect to each other. This indicates that in the case of reverse propagation the orthogonal spatial port, relative to the forward propagation, receives the transmitted energy leading to a circulator action (see supplementary information and Fig. S2 for more details). Furthermore, as seen from Eq. (3), the reflection coefficients at the UDC stages are non- zero for all modes. However, for the whole cascade, the total reflection is identically zero at every port [sii = 0 for all i in Eq. (9)]. This remarkable cancellation of total reflections for the cascade can be understood in analogy with a Fabry-Perot resonance where a cavity flanked by two identical reflecting mirrors displays unity transmission at resonance. The total phase shift between the active "mirrors" in our device: (π/2)a+ − (−π/2)a† resonance condition when a half-wavelength of the incident radiation equals the length of the = π, is akin to the − Fabry-Perot cavity. Also the reflections at the two UDC stages are identical [as the reflection coefficients are independent of the phase angle φ, cf. Eq. (3)], fulfilling the second condition for the transmission resonance and net cancellation of reflections [21]. We show the dependence of circulator action on different parameters in the device in Fig. 4. Since the isolation achieved is robust to reasonable deviations of parameters from their ideal values (φL = 0, φR = π/4, θ = π/2, δ± = 1/√2, αL = αR = 2−3/4), the active circulator design holds promise for use in practical circuits. Another interesting feature of this device is the reversal of transmission characteristics with phase of the pumps (φL,R 7→ −φL,R) (Fig. 4c). In the classic circulators based on passive Faraday rotation, this can be accomplished by changing the polarity of the magnetic bias field. Thus the clock ("pump phase") in an active 8 device indeed plays a role equivalent to the magnetic field in a Faraday medium. In conclusion, we have described a scheme for achieving nonreciprocal wave propagation using a protocol involving up-conversion followed by down-conversion mediated by an appro- priate phase shift [22]. The proposed design performs noiselessly as it consists of purely disper- sive components with no dissipation, making it attractive for quantum information applications using superconducting circuits [23]. Besides microwave applications, the architecture of the protocol described in this paper can be adapted to optical frequencies, where it can complement the recently proposed designs of nonreciprocal light propagation based on dynamical modula- tion of the refractive index of photonic structures [24] and the use of a surface waveguide on photonic crystals [25]. In addition to the practical applications outlined above, the treatment de- scribed in this letter may also give theoretical insights into the inherently directional dynamics of devices like the dc SQUID [26], when additional active stages are included in the chain. This can be useful in tackling unanswered questions pertaining to the quantum noise of dc SQUID amplifiers [27]. References [1] Wallraff, A. et al. Strong coupling of a single photon to a superconducting qubit using circuit quantum electrodynamics. Nature, 431:162, 2004. [2] Mallet, F. et al. Single-shot qubit readout in circuit quantum electrodynamics. Nat Phys, 5:791 -- 795, 2009. [3] Johnson, B. R. et. al. Quantum non-demolition detection of single microwave photons in a circuit. Nat. Phys., 6:1745 -- 2473, 2010. 9 [4] Naaman, O., Aumentado, J., Friedland, L., Wurtele, J. S. & Siddiqi, I. Phase-locking tran- sition in a chirped superconducting josephson resonator. Phys. Rev. Lett., 101(11):117005, 2008. [5] Pozar, David M. Microwave Engineering, pages 471 -- 482. Wiley, ed. 3, 2005. [6] Van Harlingen, D. J. et. al. Decoherence in Josephson-junction qubits due to critical- current fluctuations. Phys. Rev. B, 70(6):064517, 2004. [7] Kakuyanagi, K. et. al. Dephasing of a superconducting flux qubit. Phys. Rev. Lett., 98(4):047004, 2007. [8] Choi, S., Lee, D.-H., Louie, S. G. & Clarke, J. Localization of metal-induced gap states at the metal-insulator interface: Origin of flux noise in squids and superconducting qubits. Phys. Rev. Lett., 103(19):197001, 2009. [9] Schuster, A. An Introduction to the Theory of Optics, pages 41 -- 45. Edward Arnold, London, 1904. [10] Barron, L. D. Parity and Optical Activity. Nature, 238:17 -- 19, 1972. [11] This confusion can be resolved by considering the symmetries of E and B fields under parity (P ) and time reversal (T ) operations. The electric field E described by a polar vector that is symmetric under time reversal and the magnetic field B described by an axial vector (more appropriately as an antisymmetric tensor of rank two) that is anti- symmetric under time reversal lead to the invariance of Maxwell's equations describing the electromagnetic field under both P and T operations. Thus any physical process involving only electromagnetic interactions also preserves these two symmetries. In other words, if P and T are applied to a complete experiment, the resulting situation should also be 10 physically realizable as a solution of Maxwell's equations. It can be checked that Faraday rotation under both P and T operations leads to feasible outcomes, thus preserving both the symmetries. [12] Potton, R. J. Reciprocity in optics. Rep. Prog. Phys., 67:717 -- 754, 2004. [13] Ladd, T. D. et al. Quantum computers. Nature, 464:45 -- 53, 2010. [14] Clarke, J. & Wilhelm, F. K. Superconducting quantum bits. Nature, 453:1031 -- 1042, 2008. [15] Castellanos-Beltran, M. A. & Lehnert, K. W. Widely tunable parametric amplifier based on a superconducting quantum interference device array resonator. Appl. Phys. Lett., 91(8):083509, 2007. [16] Yamamoto, T. et al. Flux-driven josephson parametric amplifier. Appl. Phys. Lett., 93(4):042510, 2008. [17] Bergeal, N et al. Analog information processing at the quantum limit with a josephson ring modulator. Nat Phys, 6:296 -- 302, 2010. [18] Bergeal, N. et al. Phase preserving amplification near the quantum limit with a Josephson Ring Modulator. available at http://arxiv.org/abs/arXiv:0912.3407. [19] Yurke, B. Input output theory. In Drummond, P.D. & Ficek, Z., editor, Quantum Squeez- ing, pages 53 -- 95. Springer, 2004. [20] We note that for the case of IQ coupling considered in Fig. 2(A), the terms coupling the two sidebands, s34 and s43, are zero [Eq. (3)]. If the phase difference between the two couplings M1 and M2 deviates from 90 degrees, crosstalk appears between the two 11 sidebands generated at port C. Furthermore, if M1 and M2 would be completely in phase, the two sidebands would be maximally coupled while the cross reflections q0 between the low frequency input ports would reduce to zero. In such a case, the transformation defining the transfer matrix is singular (refer to methods under supplementary information for details). [21] A point of disticntion between the two pictures is that the symmetry of the active circulator design is described by a sub-group of the SU (4) group (the group formed by 4×4 complex matrices of unit determinant, Eq. 9]), and not by the SU (2) group that describes passive lossless two-port devices like the Fabry-Perot resonator. [22] A recent theoretical work [28] showed the existence of a non-reciprocal effect using non- linear circuit that involves microwave resonators coupled through Josephson junctions. While the proposal of this reference is based on a passive Josephson circuit, it involves, unlike our proposal, very small JJ susceptible to offset charges. Furthermore, due to its small characteristic energy, that device will handle the qubit readout signals with a lesser throughput than our proposed device. [23] DiCarlo, L. et. al. Demonstration of two-qubit algorithms with a superconducting quantum processor. Nature, 460:240 -- 244, 2009. [24] Yu, Z. & Fan, S. Complete optical isolation created by indirect interband photonic transi- tions. Nature Photonics, 3:91 -- 94, 2009. [25] Liu, A. Q., Khoo, E. H., Cheng, T. H., Li, E. P. & Li, J. A frequency-selective circu- lator via mode coupling between surface waveguide and resonators. Appl. Phys. Lett., 92(2):021119, 2008. 12 [26] Clarke, J. & Braginsky, A. I., editor. The SQUID Handbook Vol. I Fundamentals and Technology of SQUIDs and SQUID Systems. Wiley-VCH Verlag GmbH and Co. KGaA, 2004. [27] Clarke, J. & Braginsky, A. I., editor. The SQUID Handbook Vol. II Applications of SQUIDs and SQUID Systems. Wiley-VCH Verlag GmbH and Co. KGaA, 2006. [28] Koch, J., Houck, A. A., Le Hur, K. & Girvin, S. M. Time-reversal symmetry breaking in circuit-QED based photon lattices. arXiv:1006.0762, 2010. Acknowledgements We acknowledge useful discussions with S. M. Girvin, Jens Koch and R. J. Schoelkopf. This research was supported by the US National Security Agency through the US Army Research Office grant W911NF-05-01-0365, the W. M. Keck Foundation, the US National Science Foun- dation through grant DMR-032-5580 (A.K. and M.H.D.) as well as by the Director, Office of Science, Office of Basic Energy Sciences, Materials Sciences and Engineering Division, of the U.S. Department of Energy under Contract No. DE-AC02- 05CH11231 (J.C.). M.H.D. also ac- knowledges partial support from the College de France and from the French Agence Nationale de la Recherche. Author Contributions M.H.D. proposed the original idea for the device. A.K. developed the ideas and performed the theoretical analysis. J.C. contributed extensively to the discussions of the results. All authors contributed in writing the manuscript. 13 Figure 1: Faraday rotation and circulator action. a, Faraday rotation for a wave travelling from left to right in a Faraday-active medium, followed by a reflection back into the medium leading to a reversal of the direction of propagation. The rotation of the light polarization is fixed to a rotation-like property of the medium (shown by the arrows), set by an external magnetic field oriented along the propagation axis. The sense of light rotation as seen with respect to the direction of propagation remains the same, leading to the doubling of the rotation angle on reversing the ray through the medium. b, Rotation of the polarization vector of light on passage through an optically active medium, on the other hand, cancels out on reversing the direction of propagation. This occurs because optical rotation depends on the chirality of the medium (represented as a helix) which also reverses with the direction of propagation. c, Representation and schematic design of a conventional four-port circulator. The device consists of two 90 degree hybrids (equivalent to optical beam splitters) separated by a non-reciprocal phase shifter based on Faraday rotation. Solid black arrows indicate an amplitude split with no phase change while, open arrows indicate an amplitude split with a 90 degree phase change. The non-reciprocal phase shift is effective only for the propagation direction indicated by the arrow on the phase shifter box. 14 B L to R β β B 2β R to L β β 3 4 i/ 2 1/ 2 1/ 2 i/ 2 i π/2 −π/2 -i 1/ 2 i/ 2 i/ 2 1 2 1/ 2 3 2 L to R R to L a b c 1 4 Figure 2: Description of an active reversible (information-conserving) IQ modulator per- forming frequency up- and down-conversion (UDC). a, Circuit schematic of the UDC containing only dispersive components. The two low frequency series LC resonators (with LA = LB and CA = CB), are fed by two input semi-infinite transmission lines, A and B, and parametrically coupled to a third high frequency series LC resonator leading to an out- put line C. The parametric coupling is achieved by varying the mutual inductances M1 and M2 between the left and right resonators at the carrier frequency ωc which, for optimal fre- quency conversion, is set at the band center of the right resonator. When operated from left to right, the circuit performs the modulation of low frequency signals of frequency ω0 travel- ling on ports A and B to generate sidebands at ωc ± ω0 travelling on the high frequency line C. It performs the inverse operation of demodulation when operated in reverse from right to left. b, Spectral density/response landscape for different spatial channels (or ports) of the UDC circuit in a as a function of frequency. The dotted lines represent the couplings between dif- ferent ports. The solid and the dashed arrows represent different frequencies and respective conjugates. The resonance lineshapes of the two spatially distinct ports A and B are centered at ωA = ωB = 1/qLA,BCA,B. Here we show the case when the incoming signal at ω0 is resonant with the center frequency (ω0 = ωA,B). The two sidebands generated by the UDC on channel C are detuned from the carrier ωc by equal amounts. 15 a b Port A Port C Port B Pump A B eiω0t a0 ' ZA ZB R1 a0 '' eiω0t M1=M0 cos(ωct + φ) CC LC CA LA LB CB ei(ωC + ω0)t b+ ZC C b− e−i(ωC − ω0)t M2=M0 sin(ωct + φ) b+ b− −ω+ −ω_ ' a0 ' a0 −ω0 ω0 '' a0 '' a0 −ω0 ω0 b_ b+ ω_ ω+ −ωc 0 ωc ω ω ω ω Figure 3: Description of the active circulator. a, Circuit schematic of the active circulator design: the first UDC stage acts as a frequency up-converter (UC) (also indicated by a gradation in the color of the relevant box) with a parametric coupling modulated at the carrier frequency ωc = ω+ − ω0 = ω− + ω0 and a phase φL = 0. This is followed by a phase shifter (PS) that phase shifts both the sidebands by π/2, in opposite directions. They are then demodulated by the final UDC stage acting as a frequency down-converter (DC), with the carrier phase φR = π/4. b, Forward (green) and backward (red) propagation diagrams calculated using transfer matrix method for a with appropriate choice of detuning (δ± = 1/√2) and coupling strengths (αL = αR = 2−3/4) for maximum isolation. c, Representation of the device operation using modulation ellipses at each stage in the cascade. The top panel shows the forward propagation for the case when two distinct signals enter ports 1 and 2 respectively (a0 0 = 0.5eiη; η = π/24) while the bottom panel shows the backward propagation dynamics when the same signals enter ports 3 and 4 respectively (b0 0 = 0.5eiη). The relative phase and amplitudes are chosen to represent the most general case of two input signals which differ in both amplitude and phase. The relative phase difference between the two signals is encoded as the tilt of the modulation ellipse in the IQ plane while their average phase is represented as the color along the perimeter of the ellipse with yellow indicating zero phase (also see supplementary information for more details). 0 = 1, a00 0 = 1, b00 16 −i/21/4 1 21/4 -i 21/4 −1/21/4 −i3/2/23/4 −i1/2 23/4 i1/2 23/4 i3/2/23/4 3 4 3 4 −i π/2 −π/2 i −i π/2 −π/2 i DC 1 2 1 2 12 -1 -2 12 -1 -2 1 2 -2 -1 UC 1 2 -2 -1 12 -1 -2 12 -1 -2 b −i/23/4 i 23/4 -1 23/4 −1/23/4 i−1/2/21/4 −i1/2 21/4 i1/2 21/4 −i−1/2/21/4 1 2 1 2 PS 3 4 I b0 ' b0' ' ω c Q UDC φR = π/4 ω c Q I a0 ' a0' ' UDC φL= 0 b + a + b + a + PS θ=π/2 b− a− b− a− 12 UC 12 Forward a 1 2 c -2 -1 1 2 -2 -1 1 2 -2 -1 PS -1 -2 12 DC 1 2 2 1 1 2 -2 -1 21 -1 -2 12 -1 -2 Backward -2 -1 Figure 4: Variation of the difference between forward and backward transmission coeffi- cients (s31 − s13). Asymmetry in transmission, calculated for coupling angles φL = 0 and φR = π/4, as a function of a, strength of the coupling α and phase rotation θ performed by the second phase shifting stage, b, detuning δ± of the sidebands from the carrier and phase rotation θ, and c, detuning δ± and the phase of the pump at second UDC stage φR. The points of maxima correspond to the ideal values reported in the text. The plot in b, also shows the periodicity of the response of the device as a function of θ. In c, the variation with respect to the pump phase shows the reversal of transmission characteristics with φR → −φR. As in b, the response is periodic in φR with a period equal to π. It can be seen that the design continues to work for moderate deviations from the preferred phase angle θ = π/2, coupling αL,R = 2−3/4, detuning δ± = 1/√2 and pump phase φR = π/4 (values indicated with dashed arrows along the axis). 17 s31-s13 φ R π s31-s13 s31-s13 α δ+ − δ+− a b 1 θπ 0 θ π 0 c 1 -1 0 1 Supplementary Information for Noiseless Nonreciprocity in a Parametric Active Device Archana Kamal,1 John Clarke,2 Michel Devoret1∗ 1Department of Physics and Applied Physics, Yale University, 15 Prospect Street, New Haven, CT 06520, USA 2Department of Physics, University of California, and Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA Methods Derivation of scattering matrix. In this section, we present the details of the derivation of the scattering matrix S for the U DC stage. A scattering matrix has the general form a[ω]out = S[ω, ω0]~ain[ω0], ~ (S1) where ~ain/out is a column vector formed by the reduced mode amplitudes (in terms of photon number at the relevant frequencies) of the network. This description extends to all the spatial and temporal modes of the network. A convenient and simple way of deriving the S matrix involves calculating the impedance matrix Z using usual circuit theory, and then using the identity [1] S = (Z + Z0)−1.(Z − Z0) (S2) to obtain the S matrix. Here Z0 denotes the characteristic impedance of the transmission lines acting as channels for propagation of the incoming and outgoing signals at various ports. It is 1 arXiv:1010.1794v1 [cond-mat.mes-hall] 8 Oct 2010 Figure S1: Comparison of two IQ modulation schemes: a Detailed circuit schematic of the U DC performing reversible IQ modulation. The fluxes across different inductances denoted as Φi and the respective mode currents denoted as Ii are shown for each series LC resonator. The mutual inductances M1, M2 lead to off-diagonal coupling (mixing) between the current of the left resonators and the flux across the inductance of the right resonator and vice versa. b, Block diagram of a non-reversible circuit performing the operation described in a. The circuit employs dissipative components like the mixers (represented by the cross in circle symbol). The CW tone from the generator imposing a carrier frequency ωc is split into two copies, one phase shifted by π/2 with respect to the other. The carrier is modulated by the signal at ω0 and encoded on two separate channels: the 'in phase" (or I) component and the "quadrature" (or Q) component. The I and Q channels are then combined to propagate onto a single spatial channel. of the form Z0 = diag(ZA, ZB, ZC, ZC), (S3) where ZA = ZB and ZC are characteristic impedances of the semi-infinite transmission lines addressing the low and high frequency ports respectively. For the parametrically coupled series LC oscillators forming the UDC stage, Fig. S1a, we obtain the total impedance matrix Z by adding the inductive (Zind) and capacitive (Zcap) contri- butions respectively. The inductance matrix L defines the constitutive relationship between the currents and fluxes for different inductances of the circuit:      Φ0 0(t) Φ00 0(t) Φ+(t) Φ−(t)      =  LA 0 0 LB M0eiφ −iM0eiφ iM0e−iφ M0e−iφ 2 M0e−iφ M0eiφ iM0e−iφ −iM0eiφ LC 0 0 LC           I0 0(t) I00 0 (t) I+(t) I−(t)      (S4) a A B M1=M0 cos(ωct + φ) ZA ZB R1 CA ' I0 LA ' Φ0 LB '' I0 CB '' Φ0 Φ+ − CC I+ LC V I− M2=M0 sin(ωct + φ) b ωc ZC C π/2 I input LO IF RF Q input LO IF RF RF output = L I0 0(t) I00 0 (t) I+(t) I−(t)      . (S5) In writing the above matrix, we ignore the fluxes at higher harmonics of the modes at ω0 and ω±. The inductive contribution to the impedance is then calculated by using the identity      V 0 0 V 00 0 V+ V−      = Φ0 0 Φ00 0 Φ+ Φ−      = L d dt      I0 0 I00 0 I+ I−      ⇒ V [ω] = ZindI[ω] (S6) where the subscripts denote the relevant frequency modes. It is straightforward to define the capacitance matrix for the circuit in the same manner:      V 0 0 V 00 0 V+ V−      = (−jω)−1 C−1 A 0 0 0 0 C−1 B 0 0 0 0 C−1 C 0 V [ω] = ZcapI[ω] 0 0 0 −C−1 C           I0 0 I00 0 I+ I−      (S7) ⇒ The extra negative sign in s44 of Eq. (S7) accounts for the generation of the conjugate wave amplitude a†[ω−] as a result of mixing of carrier ωc and signal ω0. Hence, on taking the Fourier transform of the current and voltage vectors, an extra negative sign appears for the correspond- ing coefficient in Zcap. The total Z matrix can then be written as Z = Zind + Zcap, which gives Z = 0 0 −iδ0ZA,B −iδ0ZA,B −iαZA,Beiφ −αZA,Beiφ iαZA,Be−iφ −αZA,Be−iφ −iαZCe−iφ −iαZCeiφ αZCe−iφ −αZCeiφ −iδ±ZC −iδ±ZC 0 0 Here we have introduced the symbols (S8) (S9) .      δ0 = ω0 − ωA,B ΓA,B ; ΓA,B = ZA,B 2LA,B 3 ; ΓC = ZC 2LC ΓC δ± = ω± − ωC M0qLA,BLC α = , with Γi denoting the linewidth of the ith resonator. On using Eqs. (S9) and (S3) in Eq. (S2), we obtain      a0out 0 a00out 0 bout + b†out −      =  = S −q0 r0 q0 r0 tueiφ −itueiφ −sue−iφ −isue−iφ a0in 0 a00in 0 bin + b†in −      , tde−iφ sdeiφ itde−iφ −isdeiφ r+ 0 0 r−           a0in 0 a00in 0 bin + b†in −      (S10) (S11) where ain/out i denotes the shorthand notation for a[ωi]in/out [cf. Eq. (S1)]. The symbols a, b denote the wave amplitudes for left and right ports respectively. The detailed expressions for scattering coefficients are listed below: 4α2(δ± + i) (δ0 + i)2(δ± + i)2 − 4α4 ; (δ0 + i)(δ± − i) + 2α2 (δ0 + i)(δ± + i) + 2α2 ; r+ = 2α q0 = r0 = r− = (δ2 0 + 1)(δ± + i)2 − 4α4 (δ0 + i)2(δ± + i)2 − 4α4 ; (δ0 + i)(δ± − i) − 2α2 (δ0 + i)(δ± + i) − 2α2 ; tu = i ZC ZA,B! su = i ZC ZA,B! ZC (cid:19) td = i(cid:18) ZA,B ZC (cid:19) sd = i(cid:18) ZA,B (δ0 + i)(δ± + i) − 2α2! ; (δ0 + i)(δ± + i) + 2α2! ; (δ0 + i)(δ± + i) − 2α2! ; (δ0 + i)(δ± + i) + 2α2! . 2α 2α 2α (S12) (S13) (S14) (S15) (S16) (S17) Equations (S12)-(S17) show that the effective coupling strength α plays the role of the paramet- ric drive ("pump") in the UDC. In the limit α = 0, transmission coefficients ti and si become identically zero while the reflection coefficients ri reduce to those for three independent series LCR circuits with resonance frequencies ω0 and ωc respectively. 4 aout 1 a†out 1 aout 2 a†out 2...                 = S       s11 0 . . . s2N,1 0 s∗11 . . . . . . . . . . . . ... . . . . . . . . . . . . s2N,2N             ain 1 a†in 1 . . . a†in 2N       . (S18) We note that the matrix in Eq. (S11) is one of the block diagonals of the full 8 × 8 scat- tering matrix that describes the interaction of both the positive and negative frequency wave amplitudes for all participating modes: This is especially important in the case of an active network where a conjugation operation is possible i.e. ai 7→ a†i = a[−ωi]. The full scattering matrix S satisfies the following general properties: 1. det( S) =1 (as det (S) =1) 2. P4 j=1 sij2 = 1 3. ST J S = J, where J represents a symplectic structure defined on the 2N × 2N phase space (N = number of degrees of freedom), J = iσY ⊗ IN . (S19) The last condition of symplecticity follows from the fact that a transformation of the modes as performed by the scattering matrix needs to be a canonical transformation. This requirement translates into the condition for preservation of phase space volume (information) or the number of participating modes (degrees of freedom) in the system. Derivation of transfer matrix. In this section, we derive the transfer matrix for the UDC stage. This description is equivalent to the usual ABCD matrix of the circuit theory defined in terms of the voltages and currents for a two-port network [1], (cid:18) Vb Ib(cid:19) =(cid:18) A B C D(cid:19)(cid:18) Va Ia(cid:19) . 5 (S20) (S23) , (S24) As there exists a straightforward mapping between the reduced wave amplitudes ai introduced earlier, and the currents and voltages at the ports [2], aout i = ain i = Vi + Z0Ii √2Z0¯hωi Vi − Z0Ii √2Z0¯hωi , (S21) (S22) we can easily adapt the concept to the above choice of variables. For the cascaded chain, we first evaluate the transfer matrix of each stage and then multiply them to obtain the total transfer matrix of the cascade. For the UDC stage performing up- conversion, we can find the relevant transfer matrix by solving Eq. (S11) to obtain sideband The reversal of 'in' and 'out' in the column vectors on left and right hand sides of Eq. (S23) is required to maintain a consistent sense of propagation through the device as the output of the (N − 1)th stage acts as the input for the N th stage in the chain. On doing the above transforma- tion, we obtain 0, a00 0) . −) in terms of low frequency amplitudes (a0      = T a0in 0 a0out 0 a00in 0 a00out 0           bout in + b†out †in − −b +b amplitudes (a+, a† TU C = with      t+,LReiφ t+,RLeiφ t−,LRe−iφ t−,RLe−iφ t+,LR = i(cid:18) ZA,B t+,RL = i(cid:18) ZA,B t−,LR = i(cid:18) ZA,B t−,RL = i(cid:18) ZA,B t∗+,RLeiφ −it+,LReiφ t∗+,LReiφ −it+,RLeiφ it−,LRe−iφ −,RLe−iφ t∗ t∗ −,LRe−iφ it−,RLe−iφ ZC (cid:19) (δ0 − i)(δ± − i) − 2α2 ZC (cid:19) (δ0 − i)(δ± + i) − 2α2 ZC (cid:19) (δ0 − i)(δ± − i) + 2α2 ZC (cid:19) (δ0 − i)(δ± + i) + 2α2 it∗+,RLeiφ it∗+,LReiφ −,RLe−iφ it∗ it∗ −,LRe−iφ ! ! ! ! . 4α 4α 4α 4α 6 The subscripts (+,-) refer to the resultant sideband generated at the output (ω+ or ω∗ − ) while LR(RL) indicates the relevant direction of propagation as left-to-right (right-to-left). We note that the condition for the transformation describing the mapping between S and T matrices to be non-singular is This condition is violated when the couplings M1, M2 (Fig. S1) are in phase and hence a s13s24 − s23s14 6= 0. (S25) transfer matrix cannot be defined in such a case. Similarly, the scattering matrix of the phase shifting stage, yields a transfer matrix of the form P S = TP S = 0 0 e−iθ 0 0 0 0 eiθ e−iθ 0 0 0 0 eiθ 0 0 e−iθ 0 0 0 0 eiθ 0 0 0 0 eiθ 0 0 0 0 e−iθ (S26) . (S27) ,           Finally, by exploiting the fact that down-conversion is just the inverse operation of the phe- nomenon of up-conversion we obtain the transfer matrix of the second UDC stage as TDC = F × T −1 U C × F, (S28) where F is the flip matrix of the form F = iσX ⊗ I2 required to preserve the consistency in labelling the input and output amplitudes. The forward and backward propagation diagrams in Fig. 3b of the main text have been calculated using the matrices TU C [Eq. (S24)] and TDC [Eq. (S28)]. Each arm of the propagation diagram notes the net forward going amplitude at that frequency: for instance, the amplitudes for first UDC stage performing upconversion in the forward propagation diagram (green) can be evaluated using the first and third rows of the matrix in Eq. (S24) with φ = 0. The net forward going amplitude contributed by a0 0 to a+ 7 can be calculated as t11 − t12 = (t+,LR − t∗+,RL) = −i/23/4 while that due to a00 0 is given by t13 − t14 = (−it+,LR − it∗+,RL) = −1/23/4, evaluated for the optimal parameter values δ0 = 0, δ± = 1/√2, α = 1/23/4 reported in the main text. For the backward propagation, a similar exercise using the second and fourth rows of TU C (and TDC) leads to the propagation diagram shown in the lower panel of Fig. 3b in red. The total transfer matrix Ttotal = TDC2(φ = π/4) × TP S(θ = π/2) × TU C1(φ = 0) (S29) establishes the relationship between the low frequency signals on the left and the right ports of the cascade as           a0in 0 a0out 0 a00in 0 a00out 0      (S30) (S31) 0 i 0 0 . i 0 0 0 0 0 0 −i =  = Ttotal 0 0 i 0 a0in 0 a0out 0 a00in 0 a00out 0           b0out 0in 0 b00out 00in 0 0b 0b      The matrix in Eq. (S30) has been evaluated for the resonant case δ0 = 0 and the parameter (S32) 1 23/4 . ; α = 2 1√ δ± = On evaluating the total scattering matrix Stotal from Ttotal, we obtain the four port circulator reported in Eq. (9) of the main text. The device behaves nonreciprocally since ST 6= S [1]. values Modulation ellipse. This scheme, providing a geometrical visualization of the action of our device, represents the superposition of two sinusoidal signals as an ellipse in the plane defined by the quadratures I and Q. In general for any two complex phasors ~a and ~b rotating in opposite 8 directions aeiωt + ~be−iωt = Re[(~a + ~b∗)eiωt] ~ . (S33) I {z + i Im[(~a − ~b∗)eiωt] } {z Q } The magnitudes and phases of the two complex signals (4 quantities in total) are encoded as different properties of a colored ellipse in the IQ plane: the semi-major axis of the ellipse equals ρ+ = a + b while the semi minor axis equals ρ− = a − b, the angle with the I axis equals (θa − θb)/2 and the location of the colors on the ellipse represents the phase angle (θa + θb)/2. onal modes (a0 The output at each stage in the active circulator comprises two modes -- the spatially orthog- 0) and the sidebands (a+, a−) that can be represented as phasors with opposite sense of rotations in a frame rotating at the carrier frequency ωc. Thus we can faithfully map the 0, a00 detailed dynamics of the device after each stage using the modulation ellipse representation and plot the combined output signal in the IQ plane with I = Re[(a+b)eiωt] and Q = Im[(a−b)eiωt] (a and b representing the relevant modes at each stage). Now we present examples of two different kinds of phase rotations and the resultant trans- formations ("rotations") of the modulation ellipse (see Fig. S3 for additional examples on modulation ellipse representation). 1. Phase shift: The action of a phase shifter which performs frequency independent phase rotations of both the phasors can be described using the transformations: a 7→ aeiθ and b 7→ beiθ. (S34) On using the above and performing the analysis in the IQ plane, we obtain the expressions for new coordinates as I = Re[aei(ωt+θ) + b∗ei(ωt−θ)] (S35) 9 Q = Im[aei(ωt+θ) − b∗ei(ωt−θ)]. (S36) The action of such an operation can be easily visualized using the modulation ellipse as shown in Fig. S2(e). 2. Free evolution: In contrast to the transformation described above, we now consider the rotation preformed by a mere time evolution of the two counter-rotating phasors (say by passage through a transmission line). In such a case, the phases of the two signals continue to evolve in opposite directions collecting a phase δ in time t (δ = ωt), a 7→ ae+iδ and b 7→ be−iδ. The IQ coordinates are calculated as I = Re[(a + b∗)ei(ωt+δ)] Q = Im[(a − b∗)ei(ωt+δ)]. (S37) (S38) (S39) It is immediately evident, that under time evolution, there is only a trivial phase accumu- lation leading to change of relative positions of the two phasors along the circumference of the modulation ellipse with no rigid rotation of the ellipse, Fig. S2(f). In Fig. 3C of the main text, we have used modulation ellipses to represent the dynamics at each stage of the active circulator. In the forward propagation direction (L to R), the output ellipse representing ports 3 and 4 has the same orientation as the input ellipse representing ports 1 and 2; only the average phase of the total output changes as indicated by the change of color along the circumference of the output ellipse. In the backward propagation direction (R to L), however, the ellipse obtained at ports 1 and 2 is rotated by π/2 with respect to the input ellipse of ports 3 and 4, in addition to the trivial average phase change. This indicates the swapping of 10 Figure S2: Modulation ellipse representation of two phasors. In each of the panels, the first column describes the phasors under consideration, the second column gives a precise mathemat- ical formula for them and the third column shows the corresponding modulation ellipse. In a-d, we show both the input modes and the resultant modulation ellipse. The tilt of the ellipse with respect to the I axis represents the relative phase between the two modes [(θa− θb)/2] while the color along the ellipse represents the average initial phase of the two modes [(θa + θb)/2], with yellow representing the position of 0 (or 2π). Figs. (e) and (f) represent the resulting ellipses on performing the indicated transformations on the ellipse in (d). 11 1 2 2 4 2 1 -1 -2 24 -2 -4 24 = 1 2 -2 -1 = 1 2 -4 -2 2 1 -1 -2 12 -1 -2 13 1 2 + -2 -1 1 2 + -2 -1 -2 -4 24 1 = 3 -4 -2 2 4 1 = 3 -4 -2 2 4 -2 -4 13 -3 -1 -1 -3 -3 -1 -1 -3 1 2 + 1 2 + 2 1 -1 -2 2 1 -1 -2 2 1 -1 -2 2 1 -1 -2 a Single mode a = e iπ/4 b = 0 -2 -1 b Two modes with equal amplitudes, phase shifted by π/2 c Two modes with unequal amplitudes d Two modes with unequal amplitudes, phase shifted by π/2 e a = e iπ/4 b = e −iπ/4 -2 -1 a = e iπ/4 b = 2eiπ/4 -2 -1 a1 = e iπ/4 b1 = 2e −iπ/4 -2 -1 2 4 2 4 24 -2 -4 24 -2 -4 2 4 -4 -2 2 4 -4 -2 24 -2 -4 24 -2 -4 Differential phase shift: Rigid body rotation of modulation ellipse a2= a1eiπ/4 b2= b1e −iπ/4 -4 -2 f Common phase shift: Rotation of the phase along modulation ellipse a2= a1eiπ/4 b2= b1eiπ/4 -4 -2 the transmitted amplitudes -- port 1 (port 2) receiving the input at port 4 (port 3) -- that leads to nonreciprocal transmission characteristics of the device. References [1] Pozar, David M. Microwave Engineering (Wiley, ed. 3, 2005), pp.471-482. [2] Kamal, A., Marblestone, A. & Devoret, M. H. Phys. Rev. B 79, 184301 (2009). 12
1506.01548	2	1506	2015-10-24T06:03:07	Topological charges of three-dimensional Dirac semimetals with rotation symmetry	[ "cond-mat.mes-hall", "cond-mat.mtrl-sci" ]	In general, the stability of a band crossing point indicates the presence of a quantized topological number associated with it. In particular, the recent discovery of three-dimensional Dirac semimetals in Na$_{3}$Bi and Cd$_{3}$As$_{2}$ demonstrates that a Dirac point with four-fold degeneracy can be stable as long as certain crystalline symmetries are supplemented in addition to the time-reversal and inversion symmetries. However, the topological charges associated with Na$_{3}$Bi and Cd$_{3}$As$_{2}$ are not clarified yet. In this work, we identify the topological charge of three-dimensional Dirac points. It is found that although the simultaneous presence of the time-reversal and inversion symmetries forces the net chiral charge to vanish, a Dirac point can carry another quantized topological charge when an additional rotation symmetry is considered. Two different classes of Dirac semimetals are identified depending on the nature of the rotation symmetries. First, the conventional symmorphic rotational symmetry which commutes with the inversion gives rise to the class I Dirac semimetals having a pair of Dirac points on the rotation axes. Since the topological charges of each pair of Dirac points have the opposite sign, a pair-creation or a pair-annihilation is required to change the number of Dirac points in the momentum space. On the other hand, the class II Dirac semimetals possess a single isolated Dirac point at a time-reversal invariant momentum, which is protected by a screw rotation. The non-symmorphic nature of screw rotations allows the anti-commutation relation between the rotation and inversion symmetries, which enables to circumvent the doubling of the number of Dirac points and create a single Dirac point at the Brillouin zone boundary.	cond-mat.mes-hall	cond-mat	Topological charges of three-dimensional Dirac semimetals with rotation symmetry Bohm-Jung Yang,1, 2, 3 Takahiro Morimoto,1, 4 and Akira Furusaki1, 4 1RIKEN, Center for Emergent Matter Science, Wako, Saitama, 351-0198, Japan 2Department of Physics and Astronomy, Seoul National University, Seoul 151-747, Korea 3Center for Correlated Electron Systems, Institute for Basic Science (IBS), Seoul 151-747, Korea 4Condensed Matter Theory Laboratory, RIKEN, Wako, Saitama, 351-0198, Japan 5 1 0 2 t c O 4 2 ] l l a h - s e m . t a m - d n o c [ 2 v 8 4 5 1 0 . 6 0 5 1 : v i X r a (Dated: October 27, 2015) In general, the stability of a band crossing point indicates the presence of a quantized topological number associated with it. In particular, the recent discovery of three-dimensional Dirac semimetals in Na3Bi and Cd3As2 demonstrates that a Dirac point with four-fold degeneracy can be stable as long as certain crystalline symmetries are supplemented in addition to the time-reversal and inversion symmetries. However, the topological charges associated with Na3Bi and Cd3As2 are not clarified yet. In this work, we identify the topological charge of three-dimensional Dirac points. It is found that although the simultaneous presence of the time-reversal and inversion symmetries forces the net chiral charge to vanish, a Dirac point can carry another quantized topological charge when an additional rotation symmetry is considered. Two different classes of Dirac semimetals are identified depending on the nature of the rotation symmetries. First, the conventional symmorphic rotational symmetry which commutes with the inversion gives rise to the class I Dirac semimetals having a pair of Dirac points on the rotation axes. Since the topological charges of each pair of Dirac points have the opposite sign, a pair-creation or a pair-annihilation is required to change the number of Dirac points in the momentum space. On the other hand, the class II Dirac semimetals possess a single isolated Dirac point at a time-reversal invariant momentum, which is protected by a screw rotation. The non-symmorphic nature of screw rotations allows the anti-commutation relation between the rotation and inversion symmetries, which enables to circumvent the doubling of the number of Dirac points and create a single Dirac point at the Brillouin zone boundary. I. INTRODUCTION After the discovery of graphene, a class of materi- als, dubbed Dirac semimetals, have come to the fore of condensed matter research. In general, a Dirac semimetal has several Fermi points around which pseudo- relativistic linear dispersion relation is realized. This pseudo-relativistic energy dispersion forces the density of states on the Fermi level to vanish without opening of an energy gap, which is the unique property of Dirac semimetals distinct from ordinary metals or insulators1. In particular, the recent theoretical prediction2,3 and the experimental confirmation4–10 of three-dimensional (3D) Dirac semimetals in Na3Bi and Cd3As2 demonstrate that there are a variety of materials realizing Dirac semimetals in both two dimensions and three dimensions. Such a di- versity of Dirac materials requires us to find a systematic way to characterize and classify them. In general, the stability of nodal points in a Dirac semimetal has topological origin. This is because there is no characteristic energy scale, such as the Fermi en- ergy or the energy gap, characterizing the perturbative stability of the system. For instance, a nodal point in graphene carries a quantized pseudo-spin winding num- ber that is defined on a loop encircling the Dirac point1. On the other hand, a nodal point in 3D Weyl semimet- als is endowed with a Chern number defined on a two- dimensional (2D) closed surface surrounding the Weyl point11. The presence of such a quantized topological charge carried by a nodal point guarantees its stability, hence a nodal point can be annihilated only by collid- ing with another nodal point with the opposite topo- logical charge as long as the symmetry of the system is preserved. Recently, there have been several theoretical studies which attempt to classify the topological invari- ants of nodal points, and to extend the concept of topo- logical band theory to gapless systems such as semimetals and nodal superconductors12–28. However, in our opin- ion, a proper definition of the topological charge of Dirac points in Na3Bi or Cd3As2 has not been given so far. Dirac points in Na3Bi or Cd3As2 are protected by the time-reversal (T ), the inversion (P ), and the rota- tion symmetries2,3,22. The experimental observation of Dirac points in these systems demonstrates their stabil- ity, hence the presence of topological invariants associ- ated with them. Moreover, the theoretical observation of pair-annihilation and pair-creation of Dirac points22,33 indicates that the topological charges of the two Dirac points should have the opposite sign. Then the ques- tion is what the nature of the topological charge is as- sociated with the Dirac points. Considering the two- dimensionality of the sphere surrounding a Dirac point, the natural candidate is either a Chern number similar to the case of Weyl semimetals, or a Z2 invariant associ- ated with T symmetry satisfying T 2 = −1. However, the simultaneous presence of the time-reversal and the inver- sion symmetries forces the Berry curvature to be zero at each momentum, hence the Chern number of the Dirac point, which is basically the integral of the Berry curva- ture, also vanishes. Moreover, a Dirac point can carry a Z2 topological charge only in the presence of SU(2) spin rotation symmetry together with time-reversal and inversion symmetries satisfying (T P )2 = 1 as shown in Ref. 23. These indicate that a special care is required to find the topological charge of a Dirac point in Na3Bi or Cd3As2, which should obviously be distinct from the monopole charge of Weyl semimetals. In this work, we will show that the Dirac points in Na3Bi or Cd3As2 are characterized by topological in- variants of zero-dimensional subsystems defined on the rotation axis. Since the rotation eigenvalue is a good quantum number on the rotation axis, a zero-dimensional topological invariant can be defined by comparing the ro- tation eigenvalues of the valence and conduction bands at two points enclosing a Dirac point. We find that the nature of 3D Dirac semimetals strongly depends on the nature of the rotation symmetry. Namely, the ordinary symmorphic rotation symmetry commuting with the in- version symmetry always creates a pair of Dirac points having the opposite topological charges, and generates class I Dirac semimetals. Both Na3Bi and Cd3As2 be- long to this class. On the other hand, we find a different class of Dirac semimetals when the system has a screw rotation sym- metry. In general, the presence of non-symmorphic sym- metries, such as screw rotations and glide mirror sym- metries, guarantees a nontrivial band connection at the Brillouin zone boundary29–32. Also, it is proposed that when the double space group of non-symmorphic crystals satisfies certain conditions, a Dirac point can be realized at the Brillouin zone boundary25. Consistent with these results, our theoretical study shows that when the band degeneracy at the zone boundary is compatible with the time-reversal and inversion symmetries, a single isolated Dirac point can be created on the rotation axis. Based on this observation, we define a class II Dirac semimetal which is protected by a screw rotation symmetry and the inversion, which are mutually anti-commuting, in addi- tion to the time-reversal symmetry. The partial transla- tion associated with a screw rotation adds a U(1) phase to the rotation eigenvalue, which varies on the rotation axis. This projective nature of a screw rotation enables to circumvent the doubling of Dirac points, and create a single isolated Dirac point at a time-reversal invariant momentum at the Brillouin zone boundary. The rest of the paper is organized in the following way. We describe the general idea to define a topologi- cal charge in systems with rotation symmetry in Sec. II. Based on this general idea, class I Dirac semimetals are defined and systematically classified in Sec. III. In particular, we show that the doubling of Dirac points is unavoidable in class I Dirac semimetals. Sec. IV is about the nontrivial band connection generated by non- symmorphic screw rotation symmetries. Here we show that the partial translation associated with a screw rota- tion induces a momentum dependent U(1) phase factor to the rotation eigenvalue, which enables to circumvent the fermion number doubling and protects a single Dirac point at the Brillouin zone boundary. Based on the dis- cussion in Sec. IV, class II Dirac semimetals are defined and systematically classified in Sec. V. We present the 2 conclusion and discussion in Sec. VI. In the Appendix, we prove that there is no stable Dirac semimetal in sys- tems only with the time-reversal and inversion symme- tries based on K theory approach. The classification of Dirac semimetals in C2 invariant systems shown in the main text is also confirmed by using the K theory ap- proach. Finally, we present a short discussion about the stability of 2D Dirac semimetals protected by two-fold screw rotations. II. GENERAL IDEA: ROLE OF ROTATIONAL SYMMETRY IN SYMMORPHIC CRYSTALS In general, electronic systems having only the time- reversal (T ) and inversion (P ) symmetries cannot sup- port a stable Dirac point with a quantized topological charge11,34,35. In Appendix A, we have revisited this known fact in a different perspective and proved it by us- ing K theory approach. Thus additional crystalline sym- metries play a crucial role to stabilize Dirac semimetals realized in Na3Bi and Cd3As2. Here we consider the role of the additional rotation symmetry (CN ) in addition to P and T . For convenience, we first focus on 3D crystals with a symmorphic space group symmetry in which the point group can be completely separable from pure trans- lation operations. Also we choose the z axis as the axis for CN rotation with N indicating the discrete rotation angle of 2π/N (N = 2, 3, 4, 6). Under the operation of the CN symmetry, the Hamiltonian satisfies CN H(kx, ky, kz)(CN )−1 = H(kx, ky, kz), (1) where (kx, ky) is obtained from 2π/N rotation of (kx, ky), i.e., (kx + iky) = (kx + iky)e2πi/N . The symmetry oper- ators satisfy T 2 = −1, (CN )N = −1, P 2 = 1, (2) and [T, P ] = 0, [T, CN ] = 0, [P, CN ] = 0, (3) where we have considered the fact that an electron is a spin-1/2 particle. Now let us explain the general idea of how to deter- mine the topological charge of a Dirac point locating at a generic point k0 = (0, 0, k0 z) on the rotation axis. To determine the topological charge, we first consider a sphere in the momentum space surrounding the Dirac point at k = k0 in a CN symmetric way as shown in Fig. 1. Namely, the center of the sphere sits on the rota- tion axis. At every point on the sphere, the Hamiltonian is invariant under the compound antiunitary symmetry P T satisfying (P T )2 = −1. Moreover, the intersection of the kz axis and the sphere consists of a north pole z ) and a south pole at kS = (0, 0, kS at kN = (0, 0, kN z ), which are invariant under the rotation. We define the topological charge of the Dirac point from the topologi- cal numbers associated with these two points. Since the CN kN k0 kS FIG. 1. Local geometry around a Dirac point at k = k0 sitting on the rotation axis. The sphere surrounds a Dirac point at the center. kN and kS mark the points on the sphere crossing the rotation axis. Hamiltonian commutes with the rotation operator CN at these points, [H(kN/S), CN ] = 0, (4) hence H(kN/S) eigenspace of CN with the eigenvalues Jm given by can be block-diagonalized in the Jm = exp(cid:18)i 2m + 1 N π(cid:19) , 3 Next let us consider possible constraints to the al- lowed νm values. For a gapped system defined on the sphere, the number of conduction bands and that of va- lence bands are constants independent of the momentum on the sphere, which leads to the following constraint Xm m(cid:1) m − nS νm =Xm (cid:0)nN =Xm (cid:8)[Nc(Jm, kN ) − Nc(Jm, kS)] − [Nv(Jm, kN ) − Nv(Jm, kS)](cid:9) = 0. (8) Moreover, since P T symmetry imposes additional con- straints between different νm values, the number of in- dependent topological invariants depends on the details of the symmetry as shown in Sec. III and V. However, as long as a Dirac point possesses a nonzero νm value, it guarantees the stability of the relevant Dirac point. Hence the set of nonzero νm can be considered as a topo- logical invariant characterizing a stable Dirac point. III. CLASS I DIRAC SEMIMETALS Class I Dirac semimetals are protected by the ordinary rotation symmetry commuting with an inversion symme- try, i.e., [P, CN ] = 0. (9) m = 0, . . . , N − 1. (5) Let us consider the eigenstate ψmi of the CN operator with the eigenvalue Jm. The P T symmetry requires Since the P T symmetry is not satisfied in each CN eigenspace with a given Jm in general, (one exceptional case is shown in Sec. III B), each diagonal block of H(kN,S) belongs to the symmetry class A in terms of the Altland-Zirnbauer classification scheme38, hence car- m or nS ries an integer topological number nN,S m indicates the topological invariant of a zero-dimensional system belonging to the symmetry class A, which is de- fined as m . Here nN 1 nN m ≡ nS m ≡ 1 2(cid:2)Nc(Jm, kN ) − Nv(Jm, kN )(cid:3) , 2(cid:2)Nc(Jm, kS) − Nv(Jm, kS)(cid:3) , (6a) (6b) where Nc/v(Jm, k) denotes the number of conduction bands (c) or valence bands (v) with the eigenvalue Jm at the momentum k. It is worth to note that a triv- ial conduction or valence band with a constant energy can always be added to each Jm sector, so that the sum nN m + nS m can be changed freely. Therefore the nontrivial topological number in the Jm sector is determined by the difference νm ≡ nN m − nS m. (7) CN P Tψmi = P T CNψmi = P T Jmψmi = J∗mP Tψmi = JN−m−1P Tψmi, (10) from which we find P Tψmi is an eigenstate of CN with the eigenvalue JN−m−1. Therefore when a state with the eigenvalue Jm is occupied (or unoccupied), there should be another occupied (or unoccupied) state with the eigen- value JN−m−1, which leads to the constraint νm = νN−m−1. (11) For further analysis, we distinguish two cases based on the parity of N as shown below. A. CN symmetric systems with even N Due to the P T symmetry, the CN eigenspaces with the eigenvalues Jm and JN−m−1 can be paired as {Jm, JN−m−1} with m = 0, . . . , N/2− 1. [See Fig. 2 (a).] Since P T interchanges eigenspaces within each pair, P T is not a symmetry in each eigenspace separately, and each (a) N = even (b) N = odd J N 2 -2 J1 J N 2 -1 J N 2 J N 2 +1 JN-2 J0 JN-1 J N-3 2 J N-1 2 JN+1 2 J1 JN-2 J0 JN-1 FIG. 2. Constraints on the rotation eigenvalues (a) for even N , (b) for odd N . The (black) solid circle indicates a unit circle in the complex plane, and each (red) dot on the circle denotes Jm. Two dots connected by a dotted line are related by the P T symmetry, hence only one of them is independent. eigenspace belongs to the symmetry class A. Therefore an integer topological invariant defined in Eq. (6a) can be computed in each eigenspace with Jm. Considering the constraints shown in Eq. (8) and Eq. (11), we conclude that the topological charge of the Dirac point is given by (ν0, . . . , ν N 2 −2) ∈ Z N 2 −1. (12) Therefore, the topological charge of a Dirac point with C4 or C6 symmetry is an element of Z or Z2, respectively. At the same time, it implies that a C2 invariant system cannot support a stable Dirac point. B. CN symmetric systems with odd N When N is odd, the P T symmetry pairs the eigenspaces of CN in a slightly different way as compared to even N cases as shown in Fig. 2 (b). At first, we find 1 2 (N − 1) pairs of eigenspaces {Jm, JN−m−1}, m = 0, . . . , N − 3 2 , (13) On the other hand, the remaining eigenspace with the eigenvalue J(N−1)/2 is invariant under the P T symmetry, hence belongs to class AII. Thus, in a block-diagonalized Hamiltonian H(kN/S), there are 1 2 (N − 1) blocks be- longing to class A and an extra block with the eigen- value J(N−1)/2 belonging to class AII. In both symmetry classes, zero dimensional systems have an integer topo- logical invariant, which is defined as the difference in the number of conduction bands and that of valence bands. Hence as in the even N case, the topological charge can be defined as νm = nN m − nS m, (14) in each eigenspace with the eigenvalue Jm. The con- straints to the topological numbers νm are 4 (i) νm = νN−m−1, m = 0, . . . , (ii) Xm νm = 0. N − 3 2 , (15) (16) Thus the independent topological charge for a Dirac point is given by (ν0, . . . , ν(N−3)/2) ∈ Z N −1 2 . (17) Therefore a Dirac point with C3 symmetry has an integer (Z) topological charge. CN Topological charge C2 C3 C4 C6 Not allowed Z Z Z × Z TABLE I. Summary of topological charges of class I Dirac semimetals. C. Applications: classification of stable Dirac points in 4-band systems Let us apply the general theory developed above to minimal 4-band models, and classify stable Dirac points. In a 4-band model, a pair of doubly degenerate bands cross at a Dirac point which we assume to sit on the Fermi level. On the rotation axis k = (0, 0, kz), each band is assigned with a quantum number Jm. Since the pair of degenerate bands should have different rotation eigenval- ues to generate a stable Dirac point, each band with the rotation eigenvalue Jm satisfies nN m. Namely, a band which is below (above) the Fermi level at the mo- mentum kN should be above (below) the Fermi level at the momentum kS to have a Dirac point in between. m = −nS 1. C2 symmetric systems Jm=0 = exp(iπ/2) = i and Jm=1 = exp(3iπ/2) = −i = Jm=−1 are the only allowed C2 eigenval- ues. Due to the P T symmetry, a pair of eigenstates {ψm=0(k)i,ψm=1(k)i} are always degenerate locally at each momentum k, hence nN,S m=1 and Nm=0 = Nm=1. Then a 4-band model can be constructed by intro- ducing two pairs of eigenstates {ψA m=1(k)i} and {ψB m=1(k)i}, where A, B indicates the valence band (v) or the conduction band (c), respectively. It is straightforward to show that nN,S m = 0 (m = 0, 1), because if one state is occupied, among {ψA m(k)i, ψB m(k)i}, the other state is unoccupied. Therefore m=0(k)i,ψB m=0(k)i,ψA m=0 = nN,S νm=0,1 = 0 and there is no stable Dirac point with a nontrivial topological invariant in systems with C2 sym- metry. 2. C3 symmetric systems Possible C3 eigenvalues are Jm=0 = exp(iπ/3), Jm=1 = exp(iπ), Jm=2 = exp(5iπ/3). Due to the P T symmetry, {ψm=0(k)i,ψm=2(k)i} and {ψm=1(k)i, ψm=1(k)i} form degenerate pairs, where ψm=1(k)i = P Tψm=1(k)i. Thus νm=0 = νm=2. Since νm=0 + νm=1 + νm=2 = 0, we have ν1 ≡ νm=0 = νm=2 = − 1 2 νm=1. (18) Hence there is only one independent topological num- ber, ν1 ∈ Z. Since the topological charge of a Dirac point can be nonzero only when the valence and conduction bands have different rotation eigenvalues, a 4-band model can be constructed by using a ba- m=1(k)i, ψB sis {ψA m=1(k)i} where A = v (c) and B = c (v). Since nN,S m=0 = nN,S m=2 = 2 nN,S − 1 m for 4-band models, the Dirac point has a nonzero topological invariant m=0(k)i,ψA m=1 = ± 1 m=2(k)i,ψB m = −nS 2 and nN ν 1 = ±1. (19) 3. C4 symmetric systems Possible C4 eigenvalues are Jm=0 = exp(iπ/4), Jm=1 = exp(3iπ/4), Jm=2 = exp(5iπ/4), Jm=3 = exp(7iπ/4). Due to the P T symmetry, {ψm=0(k)i,ψm=3(k)i} and {ψm=1(k)i,ψm=2(k)i} form degenerate pairs at each momentum, thus νm=0 = νm=3 and νm=1 = νm=2. Since Pm νm = 0, 5 {ψm=2(k)i,ψm=3(k)i} form degenerate pairs. Thus νm=0 = νm=5, νm=1 = νm=4, νm=2 = νm=3. Consid- eringPm νm = 0, we can find two independent topolog- ical numbers (ν1, ν2) ∈ Z2, which, for instance, can be defined as, ν1 ≡ νm=0 = νm=5, ν2 ≡ νm=1 = νm=4. we convenience, for (22) However, use (νm=0, νm=1, νm=2) to indicate the topological charge in which νm=0 + νm=1 + νm=2 = 0. A 4-band model can be constructed by choosing two different pairs of eigenstates such as also can {ψm=0i,ψm=5iψm=1i,ψm=4i}, {ψm=0i,ψm=5iψm=2i,ψm=3i}, {ψm=1i,ψm=4iψm=2i,ψm=3i}. (23) For a given 4-band model, a nonzero topological number νm = ±1 can be assigned if Jm is the eigenvalue of one of the four bands. Whereas νm = 0 if Jm is the eigenvalue of the other two states which are not included in the 4-band model. Therefore the topological charges of the system are in the form of (νm=0, νm=1, νm=2) = (±1,∓1, 0), (νm=0, νm=1, νm=2) = (±1, 0,∓1), (νm=0, νm=1, νm=2) = (0,±1,∓1), (24) for each case shown in Eq. (23), respectively. Then the corresponding (ν 1, ν2) are (ν 1, ν2) = (±1,∓1), (ν 1, ν2) = (±1, 0), (ν 1, ν2) = (0,±1), (25) respectively. ν1 ≡ νm=0 = νm=3 = −νm=1 = −νm=2, (20) CN 4-band model Materials hence there is only one independent topological num- ber, ν1. Since the topological charge of a Dirac point can be nonzero only when the valence and conduction bands have different rotation eigenvalues, a 4-band model with Dirac points can be constructed by using a ba- sis {ψA m=1(k)i,ψB m=2(k)i} where A = v (c) and B = c (v). Since nN,S m=0 = nN,S m=3 = −nN,S m = −nS m for 4-band models, the Dirac point has a nonzero topological invari- ant m=0(k)i,ψA m=1 = −nN,S m=3(k)i,ψB m=2 = ± 1 2 and nN ν 1 = ±1. (21) 4. C6 symmetric systems In the presence rotation symmetry, {ψm=0(k)i,ψm=5(k)i}, {ψm=1(k)i,ψm=4(k)i}, and a C6 of C2 C3 C4 C6 C6 C6 Not allowed ν 1 = ±1 ν 1 = ±1 (ν1, ν 2) = (±1, ∓1) (ν 1, ν2) = (±1, 0) (ν 1, ν2) = (0, ±1) Na3Bi [2], strained TlN [39] Cd3As2 [3] TABLE II. Topological charges of class I Dirac semimetals for 4-band models and relevant materials. D. Fermion number doubling in class I Dirac semimetals Up to now, we have described how to determine the topological charge of a single Dirac point. Now let us compare the topological charges of two Dirac points at the momenta k0 and −k0. Due to the inversion symme- try satisfying [P, CN ] = 0, the eigenstates at k and −k satisfy the following relationship, (a) CN Pψm(k)i = P CNψm(k)i = P Jmψm(k)i = JmPψm(k)i, (26) which means that if there is an eigenstate with the eigen- value Jm at k, there should be a degenerate eigenstate with the same eigenvalue Jm at −k. This imposes the following condition of nN,S m (k0) = nS,N m (−k0). (27) It is to be noted that the north (south) pole at k0 and the south (north) pole at −k0 are interchanged under the inversion symmetry. Thus we obtain νm(k0) = −νm(−k0). (28) Since the net topological charge of the two Dirac points related by the inversion symmetry is zero, we obtain the following conclusions. • The total topological charge of two Dirac points within the first Brillouin zone should be zero in each angular momentum channel (Jm), i.e., ν(iD ) m = 0, XiD (29) where iD labels Dirac points. It is worth to note that this is nothing but the Nielsen-Ninomiya the- orem36,37 for three-dimensional Dirac semimetals. • A stable Dirac point with a nontrivial topological charge cannot exist at a time-reversal invariant mo- mentum (TRIM) where k0 = −k0 modulo a recip- rocal lattice vector due to the relationship νm(k0) = −νm(−k0) = −νm(k0) = 0. (30) IV. SCREW ROTATIONS AND A SINGLE DIRAC POINT A. Projective symmetry and circumventing fermion number doubling It is worth to note that the doubling of the number of Dirac points in class I Dirac semimetals results from the commutation relation [P, CN ] = 0 as discussed in Sec. III D. This means that it may be possible to avoid the doubling of the Dirac points, once the commutation relation is violated, i.e., [P, CN ] 6= 0. However, the pres- ence of a single Dirac point on the rotation axis brings about a more fundamental problem, when the periodic structure of the system is considered. This is because the band crossing (or a nonzero topological charge of a 6 2π kz 2π kz J 1,2 J 3,4 J 1,2 J 3,4 π π 0 0 (b) FIG. 3. Band structure along the rotation axis (z axis) of (a) a class I Dirac semimetal and (b) a class II Dirac semimetal. J1,2 and J3,4 are the rotation eigenvalues of each doubly de- generate band. A band crossing requires J1,2 6= J3,4. In class I (II) Dirac semimetals, the band crossing condition and the periodicity of the eigenstates are compatible (incompatible) when J1,2,3,4 are constant on the rotation axis. Dirac point) requires that the valence band and the con- duction band should have distinct eigenvalues, whereas the lattice periodicity requires the continuity of the eigen- state and its relevant eigenvalues as described in Fig. 3. Therefore the presence of a single Dirac point or an odd number of Dirac points on the rotation axis sounds un- physical, when the rotation symmetry exists along a line satisfying the periodic boundary condition. One possible way to circumvent the contradiction is when the rotation symmetry is realized projectively. Namely, if the rotation eigenvalue is well-defined only up to an additional phase factor, it is possible to create a single Dirac point compatible with the lattice periodicity by adjusting the phase degrees of freedom on the rota- tion axis. In fact, a screw rotation is such an example of projective symmetry, which can support a single isolated Dirac point as discussed in detail below. A screw rotation (eCN,p) is a non-symmorphic symme- try operation composed of an ordinary rotation (CN ) followed by a partial lattice translation τp = p N z (p = 1, ..., N − 1) parallel to the rotation axis. Here z is the unit lattice translation along the z axis assuming that the screw axis is parallel to it. Schematic figures de- scribing all possible screw rotations in 3D crystals are shown in Fig. 4. Let us note that, in many crystals, the screw rotation axis does not pass the reference point of the point group symmetry, which is invariant under point group operations of the lattice. In this case, the partial (a) (b) 7 also includes in-plane translation components perpendic- translation τp associated with the screw rotation eCN,p ular to the screw axis direction. Generally, eCN,p can be compactly represented as eCN,p = {CNτp}, τp =(cid:16)τp,x, τp,y, τp,z = p N(cid:17). (31) (32) where in the following way, In the real space, eCN,p transforms the spatial coordinates eCN,p : (x, y, z) → (x′ + τp,x, y′ + τp,y, z + τp,z), (33) where x′ = x cos y′ = x sin 2π N − y sin 2π + y cos N 2π N 2π N , . (34) Now we consider the combination of a screw rotation and the inversion symmetry. At first, we see PeCN,p : (x, y, z) → (−x′ − τp,x,−y′ − τp,y,−z − τp,z), thus Similarly, PeCN,p = {P CN − τp}. (35) eCN,pP : (x, y, z) → (−x′ + τp,z,−y′ + τp,z,−z + τp,z), thus eCN,pP = {P CNτp}. Let us note that [P, CN ] = 0 in general. Equations (35) the partial translation τp, thus there is a chance to avoid the doubling of the Dirac points. and (36) clearly show that generally [P, eCN,p] 6= 0 due to In the presence of the screw rotation symmetry eCN,p, the Bloch Hamiltonian H(kz) on the rotation axis (kx = ky = 0) satisfies Here the minus sign stems from the spin 1/2 nature of electrons. Therefore all bands on the kz axis can be la- where N,p(kz) = H(kz), eCN,p(kz)H(kz)eC−1 [eCN,p(kz)]N = − exp(−ipkz). beled by the eigenvalues of eCN,p(kz) given by i exp(−i kz(cid:17), eJm(kz) = exphiπ = Jm exp(cid:16) − i (2m + 1) p N p N N kz) (39) (36) (37) (38) 1 2 t 1 4 t (c) (d) 1 3 t 2 4 t 2 3 t 3 4 t 1 6 t 2 6 t 3 6 t 4 6 t 5 6 t Schematic figure describing all possible screw ro- FIG. 4. tations eCN,p in 3D crystals. In each figure, the system is periodic under the lattice translation t along the vertical di- rection. eCN,p indicates a 2π N counterclockwise rotation com- t. (a) eC2,1 symmetry. (b) bined with a partial translation p N eC3,p symmetry (p = 1, 2). (c) eC4,p symmetry (p = 1, 2, 3). (d) eC6,p symmetry (p = 1, 2, 3, 4, 5). eCN,p is not Jm but Jm exp(−i p where Jm is an eigenvalue of CN defined in Eq. (5). It is worth to note that the eigenvalue of the screw rotation N kz) which varies along the rotation axis. Therefore through the variation of this additional phase factor, it may be possible to satisfy the condition for the band crossing to create a Dirac point and the periodicity (or the continuity) of the eigenval- ues, simultaneously, even in the presence of a single Dirac point. In fact, the assignment of non-quantized quantum numbers to fermions, such as eJm(kz) varying in the mo- mentum space, is one way to get around the fermion dou- bling problem, as pointed out by Nielsen and Ninomiya in their seminal work36,37. B. Screw rotations and band connections at the zone boundary The momentum dependence of screw rotation eigen- connections between different eigenstates at the Brillouin zone boundary. For instance, if the system has 2π peri- odicity along the kz axis, we find that values eJm(kz) shown in Eq. (39) induces nontrivial band eJm(kz + 2π) = exphiπ kz(cid:17) = eJm−p(kz), thus the eigenstate with the eigenvalue eJm(kz) should be smoothly connected to the other eigenstate with the i exp(cid:16) − i (2m − 2p + 1) p N (40) N ~ J (k ) m z ~ J (k -4π) m-2p z −π π 3π 5π kz 7π ~ J (k -2π) m-p z ~ J (k -6π) m-3p z FIG. 5. The band connection required by a screw rotation eCN,p when the system is 2π periodic along the kz direction. eigenvalue eJm−p(kz) (eJm+p(kz)) at the Brillouin zone boundary kz = π (kz = −π) as shown in Fig. 5. This naturally gives rise to a band crossing point at the Brillouin zone boundary. If this band connection is compatible with the T and P symmetries, a single Dirac point can be realized at the Brillouin zone bound- ary. The P T symmetry requires that the state with N kz) should be lo- cally degenerate with the other state with the eigen- N kz) at each kz. Sim- ilarly, we can expect the degeneracy between two states N kz) and N kz), respectively. Here the important point is that the screw rotation requires the eigenvalue eJm(kz) = Jm exp(−i p value eJN−m−1(kz) = J∗m exp(−i p with the eigenvalues eJm−p(kz) = Jm−p exp(−i p eJN−m+p−1(kz) = J∗m−p exp(−i p a nontrivial band connection between eJN−m−1(kz) and eJN−m+p−1(kz), similar to the relation shown in Eq. (40). Namely, eJN−m−1(kz + 2π) = eJN−m−1−p(kz) = eJN−m+p−1(kz), which gives (41) N − m − 1 − p = N − m + p − 1 (mod N ), (42) thus p = N 2 . (43) Since p is an integer, this condition can be satisfied only in systems with eC2,1, eC4,2, eC6,3 symmetries. Let us note that, in 3D crystals, the periodicity along the kz direction can be longer than 2π/az although the system is periodic under the translation by az along the z direction, unless az is a primitive lattice vector. (For instance, it happens in the face centered cubic lattice.) Generally, when the system is 2nπ periodic along the kz axis with an integer 1 < n < N , (2m − 2np + 1) eJm(kz + 2nπ) = exphiπ kz(cid:17) = eJm−np(kz), thus the eigenstate with the eigenvalue eJm(kz) should be smoothly connected to the other eigenstate with the i exp(cid:16) − i p N (44) N 8 eigenvalue eJm−np(kz) (eJm+np(kz)) at the Brillouin zone boundary kz = nπ (kz = −nπ). Considering the P T symmetry and following the same procedure that we have used to derive Eq. (43), we obtain 2np = 0 (mod N ). (45) For example, when the system is 4π periodic (n = 2), equivalent to a 2π shift, which is already considered be- symmetries, a 6π shift connects an eigenstate with it- state into itself, hence nontrivial band connection at the Brillouin zone boundary is not expected. On the other hand, when the system is 6π periodic (n = 3), Eq. (45) Eq. (45) can also be satisfied in systems with eC4,1, eC4,2, eC4,3, eC6,3 symmetries. However, in systems with eC4,2 and eC6,3 symmetries, a 4π shift merely maps an eigen- can be satisfied in systems with eC6,p symmetry where p = 1, 2, 3, 4, 5. However, in the case of eC6,2 and eC6,4 self. Also in the eC6,3 symmetric case, a 6π shift is simply fore. Hence only the systems with eC6,1 and eC6,5 can sup- in a eCN,p symmetric system satisfy- eCN,p eigenstates should be connected to each other at the eigenstate with the eigenvalue eJm(kz) should be eigenvalue eJm−N/2(kz) at the Brillouin zone boundary eJm(kz + N′π) = exphiπ kz(cid:17) port a nontrivial band connection at the Brillouin zone boundary kz = ±3π. ing p/N = p′/N′ with two co-prime numbers p′ (an odd integer) and N′ (an even integer), two distinct the Brillouin zone boundary kz = ±N′π/2. Namely, smoothly connected to the other eigenstate with the kz = ±(N′π)/2 in the following way, i exp(cid:16) − i (2m − N p′ + 1) To sum up, p′ N′ N (46) = eJm−N p′/2(kz) = eJm−N/2(kz), where we have used the fact that p′ is an odd integer and m is well-defined modulo N . It is interesting to note that the eigenvalue eJm(kz) at the zone boundary kz = ±N′π/2 becomes eJm(kz = ±N′π/2) = Jm exp(cid:16) ∓ i π(cid:17) = ∓iJm(−1)(p′−1)/2. p′ 2 (47) at the zone boundary. This additional factor ±i gives rise to the following relations between the screw rotation Namely, the eigenvalue eJm(kz) is simply given by ±iJm eCN,p and the inversion P at the zone boundary k± = (0, 0,±N′π/2), PeCN,p(cid:12)(cid:12)(cid:12)k±E = ∓i(−1)(p′−1)/2P CN(cid:12)(cid:12)(cid:12)k±E, eCN,pP(cid:12)(cid:12)(cid:12)k±E = ±i(−1)(p′−1)/2P CN(cid:12)(cid:12)(cid:12)k±E, (48) (a) ~ J (k ) 0 z ~ J (k ) 0 z ~ J (k -4π) 1 z 9 kz 7π ~ J (k -6π) 0 z −π 0 π 2π (b) −π ~ J (k ) 1 z 0 ~ J (k ) 0 z (c) kz π −π 0 kz 3π ~ J (k -2π) 1 z ~ J (k ) 0 z kz π ~ J (k ) 1 z FIG. 6. (a) An example of the band connection required by the screw rotation in Eq. (51). For each band, the eigenvalue of the screw rotation eC2,1 is marked in the figure. (b,c) Pos- sible band structure protected by a two-fold screw rotation eC2,1 with a Dirac point at kz = 0 and at kz = π, respectively. which can be easily derived from Eq. (35) and (36). Hence P and eCN,p anticommute when the Bloch state k± = (0, 0,±N′π/2)i is used as a basis for the repre- sentation. From this, we obtain the following general principle to create a stable Dirac semimetal with a single Dirac point. Namely, • The Dirac point should be located at a TRIM at the Brillouin zone boundary (kz = ±N′π/2) where the screw rotation symmetry anti-commutes with the inversion symmetry. C. Applications In the following, we examine the possible Dirac semimetals with a single Dirac point by considering var- ious screw rotation symmetries explicitly. −π π 3π 5π ~ J (k -2π) 2 z FIG. 7. An example of the band connection required by eC3,1 symmetry shown in Eq.(54). Two bands with the eigenvalues eJ0(kz) and eJ1(kz) ( eJ0(kz) and eJ1(kz)) form a conduction (va- lence) band. However, each degenerate pair violate the P T symmetry. Now we prepare two bands Ψ0(kz) and Ψ1(kz) with an eigenvalue eJ0(kz) and eJ1(kz), respectively, and construct a band structure with a Dirac point. Here the cru- cial point is that the band Ψ0(kz) (Ψ1(kz)) should be smoothly connected to the other band Ψ1(kz) (Ψ0(kz)) at the Brillouin zone boundary (kz = ±π) to satisfy Eq. (50) and (51) as shown in Fig. 6. This naturally gives rise to a band structure with a single band crossing point at a TRIM. Considering the P or T symmetry, there are two possible band structures having a single band crossing point as shown in Fig. 6 (b) and (c). In each case, the band crossing point locates at a TRIM either at kz = 0 (Fig. 6 (b)) or at kz = π (Fig. 6 (c)). However, let us note that, due to the T P symmetry, the state with the eigen- each momentum kz. This requires that both conduction band and valence band should have the same eigenvalues value eJ0(kz) (eJ1(kz)) should be locally degenerate with the other state with the eigenvalue eJ1(kz) (eJ0(kz)) at of eJ0(kz) and eJ1(kz), hence a stable Dirac point cannot be created due to the finite hybridization between the valence and conduction bands. 1. Two-fold screw rotation eC2,1 2. Three-fold screw rotation lowing two eigenvalues A two-fold screw rotation symmetry eC2,1 has the fol- (kz − π)i, eJ0(kz) = J0 exp(cid:16) − i (kz − 3π)i, eJ1(kz) = J1 exp(cid:16) − i kz(cid:17) = exph − i kz(cid:17) = exph − i 1 2 1 2 1 2 1 2 (49) where for m = 0, 1 and eJm(kz = −π) 6= eJm(kz = π), eJ0(kz) = eJ1(kz + 2π), eJ1(kz) = eJ0(kz + 2π). (50) (51) are three possible eigenvalues given by In the case of a three-fold screw rotation eC3,1, there eJ0(kz) = J0 exp(cid:16) − i (kz − π)i, (kz − 3π)i, eJ1(kz) = J1 exp(cid:16) − i (kz − 5π)i, eJ2(kz) = J1 exp(cid:16) − i kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i 1 3 1 3 1 3 1 3 1 3 1 3 (52) where eJm(kz = −π) 6= eJm(kz = π), (53) for m = 0, 1, 2 and ~ J (k ) 0,3 z 10 (54) −2π 0 2π 4π kz 6π eJ0(kz) = eJ1(kz + 2π), eJ1(kz) = eJ2(kz + 2π), eJ2(kz) = eJ0(kz + 2π), Now we prepare three bands Ψ0,1,2(kz) with an eigen- ture with a Dirac point. To satisfy Eq. (53) and (54), each value eJ0,1,2(kz), respectively, and construct a band struc- band with a given eC3,1 eigenvalue should be connected to the other two bands with different eC3,1 eigenvalues at each Brillouin zone boundary. An example of the band connection satisfying Eq. (54) is shown in Fig. 7. Ac- cording to Fig. 7, when the band structure is drawn for a reduced Brillouin zone with kz ∈ [−π, π], two bands with would form a degenerate band. the eigenvalues eJ0(kz) and eJ1(kz) (or eJ0(kz) and eJ2(kz)) quires the band with an eigenvalue eJ0(kz) to be locally degenerate with the other band with an eigenvalue eJ2(kz) whereas the band with an eigenvalue eJ1(kz) to be locally eJ1(kz). Namely, the four bands However, this band structure is generally incompatible with the T and P symmetries. The P T symmetry re- degenerate with the other band with the same eigenvalue {Ψ0(kz), Ψ2(kz); Ψ1(kz), Ψ′1(kz)} (55) would form a basis to create a Dirac point. In the paren- thesis, the first two bands form a conduction (valence) band whereas the last two bands form a valence (con- duction) band. This basis is obviously incompatible with the band connection described in Fig. 7. It is straight- forward to show that the same problem happens for eC3,2 symmetric systems. Therefore a system with a three- fold screw rotation cannot satisfy the P T symmetry at the same time, hence cannot have a stable Dirac point. In fact, every screw symmetric system which does not satisfy Eq. (43) or Eq. (45) has the same problem, thus a Dirac semimetal with a single Dirac point cannot be created. 3. Four-fold screw rotation eC4,1 are four possible eigenvalues given by In the case of the four-fold screw rotation eC4,1, there eJ0(kz) = J0 exp(cid:16) − i (kz − π)i, (kz − 3π)i, eJ1(kz) = J1 exp(cid:16) − i (kz − 5π)i, eJ2(kz) = J2 exp(cid:16) − i eJ3(kz) = J3 exp(cid:16) − i (kz − 7π)i. kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i 1 4 1 4 1 4 1 4 1 4 1 4 1 4 1 4 (56) ~ J (k -4π) 2,1 z FIG. 8. An example of the band connection required by eC4,1 symmetry shown in Eq.(60). It is assumed that the system is 4π periodic. The band with the eigenvalue eJ0(kz) ( eJ3(kz)) should be connected with the other hand with the eigenvalue eJ2(kz) ( eJ1(kz)) at the zone boundary. Moreover, due to the P T symmetry, two bands with the eigenvalues eJ0(kz) and eJ3(kz) (or, eJ2(kz) and eJ1(kz)) should be locally degenerate at each momentum. In this case, the band connection required by the screw rotation is compatible with the P T symmetry. If the system is 2π periodic along the kz direction, we obtain eJm(kz = −π) 6= eJm(kz = π), for m = 0, 1, 2, 3 and eJm(kz) = eJm+1(kz + 2π), Brillouin zone boundary (kz = ±π). However, this band connection is not compatible with the P T symmetry of hence the state with the eigenvalue eJm(kz) should be con- nected to the state with the eigenvalue eJm+1(kz) at the the system, which requires the state with eJ0(kz) (eJ1(kz)) to be degenerate with the state with eJ3(kz) (eJ2(kz)). On the other hand, if the system is 4π periodic along the kz direction, eJm=0,1,2,3(kz = −2π) 6= eJm=0,1,2,3(kz = 2π), and eJm(kz) = eJm+2(kz + 4π). This band connection is compatible with the P T symme- try. The basis for creation of a Dirac semimetal with a single Dirac point at the Brillouin zone boundary is given by {Ψ0(kz), Ψ3(kz); Ψ1(kz), Ψ2(kz)}. (61) (57) (58) (59) (60) 4. Four-fold screw rotation eC4,2 6. Six-fold screw rotation eC6,3 11 are four possible eigenvalues given by In the case of the four-fold screw rotation eC4,2, there π)i, eJ0(kz) = J0 exp(cid:16) − i π)i, eJ1(kz) = J1 exp(cid:16) − i π)i, eJ2(kz) = J2 exp(cid:16) − i π)i, eJ3(kz) = J3 exp(cid:16) − i kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i (kz − (kz − (kz − (kz − 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2 3 2 5 2 7 2 (62) where for m = 0, 1, 2, 3 and eJm(kz = −π) 6= eJm(kz = π), eJm(kz) = eJm+2(kz + 2π), should be connected with the state with the eigenvalue hence the state with the eigenvalue eJ0(kz) (eJ1(kz)) eJ2(kz) (eJ3(kz)) at the Brillouin zone boundary (kz = ±π). This band connection is compatible with the P T symmetry. The basis for creation of a Dirac semimetal with a single Dirac point at the Brillouin zone boundary is given by and (63) are six possible eigenvalues given by In the case of the six-fold screw rotation eC6,3, there eJm(kz) = Jm exp(cid:16) − i kz(cid:17) 2(cid:16)kz − = exph − i 1 2 1 2m + 1 3 π(cid:17)i, where m = 1, 2, ..., 6. When the system is 2π periodic along the kz direction, these eigenvalues satisfy eJm(kz = −π) 6= eJm(kz = π), eJm(kz) = eJm+3(kz + 2π), the Brillouin zone boundary (kz = ±π), respectively. Due to the T and the P symmetries, the pair of states hence the state with the eigenvalue eJ0,1,2(kz) should be connected to the state with the eigenvalue eJ3,4,5(kz) at with the eigenvalues {eJ0(kz), eJ5(kz)}, {eJ1(kz), eJ4(kz)}, {eJ2(kz), eJ3(kz)} should be locally degenerate at each mo- mentum kz. Considering the band connection described in Eq. (71), we find the following basis {Ψ0(kz), Ψ5(kz); Ψ3(kz), Ψ2(kz)}, (72) (69) (70) (71) {Ψ0(kz), Ψ3(kz); Ψ1(kz), Ψ2(kz)}. (64) which can create a Dirac semimetal with a single Dirac point at the Brillouin zone boundary. 5. Four-fold screw rotation eC4,3 7. Six-fold screw rotation eC6,p6=3 are four possible eigenvalues given by In the case of the four-fold screw rotation eC4,3, there eJ0(kz) = J0 exp(cid:16) − i π)i, π)i, eJ1(kz) = J1 exp(cid:16) − i π)i, eJ2(kz) = J2 exp(cid:16) − i π)i. eJ3(kz) = J3 exp(cid:16) − i Similar to the case of eC4,1 symmetric systems, a Dirac kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i kz(cid:17) = exph − i semimetal with a single Dirac point can be created only if the system is 4π periodic along the kz direction. Then (kz − (kz − (kz − (kz − 3 4 3 4 3 4 3 4 3 4 3 4 3 4 3 4 1 3 3 3 5 3 7 3 (65) for m = 0, 1, 2, 3 and eJm(kz = −2π) 6= eJm(kz = 2π), eJm(kz) = eJm−2(kz + 4π). The basis for creation of a Dirac semimetal with a single Dirac point at the Brillouin zone boundary is again {Ψ0(kz), Ψ3(kz); Ψ1(kz), Ψ2(kz)}. (68) and (66) (67) gle Dirac point can be created only if the system is 6π periodic along the kz direction. In the case of the six-fold screw rotation eC6,p6=3, a sin- In eC6,1 symmetric systems, there are six possible eigen- values given by where m = 1, 2, ..., 6. These eigenvalues satisfy 1 6 1 6 kz(cid:17) [kz − (2m + 1)π]o, eJm(kz) = Jm exp(cid:16) − i = expn − i eJm(kz = −3π) 6= eJm(kz = 3π), eJm(kz) = eJm+3(kz + 6π), values given by In eC6,5 symmetric systems there are six possible eigen- kz(cid:17) eJm(kz) = Jm exp(cid:16) − i 6(cid:16)kz − = exph − i 5 6 5 2m + 1 5 π(cid:17)i, (76) (73) (74) (75) where m = 1, 2, ..., 6. These eigenvalues satisfy and eJm(kz = −3π) 6= eJm(kz = 3π), eJm(kz) = eJm+3(kz + 6π). (77) (78) {P,eCN} = 0. In both eC6,1 and eC6,5 symmetric cases, we find the fol- lowing basis {Ψ0(kz), Ψ5(kz); Ψ3(kz), Ψ2(kz)}, (79) which can create a Dirac semimetal with a single Dirac point at the Brillouin zone boundary. Finally, it is straightforward to show that eC6,2 and eC6,4 symmetric cases, which are similar to the system with a three-fold screw symmetry, cannot support a Dirac semimetal with a single Dirac point. V. CLASS II DIRAC SEMIMETALS Based on the discussion in Sec. IV, we define class II Dirac semimetals in the following way. Class II Dirac semimetals are associated with a special type of rotation symmetry eCN which anti-commutes with the inversion symmetry, i.e., (80) Considering that i) a rotation operator C generally has a form of C = exp(iθJ) with the angular momentum operator J and the rotation angle θ, and ii) J is a pseu- dovector which is even under the inversion symmetry P , i.e., P JP −1 = J, the anti-commutation relation between P and eCN looks quite unusual. However, as discussed in the previous section, non-symmorphic screw rotation symmetries can generally satisfy such anti-commutation relations at the Brillouin zone boundary27,28, when Bloch states are used as a basis for the representation of P and eCN . We first describe the physical consequence resulting from the relation {P, eCN} = 0 and the topological charge of the associated Dirac point. After that we describe how the Dirac point protected by a screw rotation symmetry leads to a class II Dirac semimetal, and its associated topological charge can be determined. A. Symmetry constraint and band connection Let us consider the constraints to the rotation eigen- values eJm due to the symmetries P , T , eCN satisfying [T, P ] = 0, (81) [T,eCN ] = 0, {P,eCN} = 0, T 2 = −1, P 2 = 1. (82) 12 Moreover, considering Eq. (47), we assume the eCN eigen- value eJm to have the following form, eJm = iJm, where Jm = exp[iπ(2m + 1)/N ]. The action of the P T (83) on an eigenvector ψm(k)i of eCN with an eigenvalue eJm gives eCN P Tψm(k)i = −P TeCNψm(k)i = −P TeJmψm(k)i = −eJ∗mP Tψm(k)i = eJN−m−1P Tψm(k)i ≡ eJN−m−1ψN−m−1(k)i, from which we find P Tψmi is a degenerate eigenvector T transform the eigenvector ψmi at k to another eigen- vector at −k as with the eigenvalue eJN−m−1. On the other hand, P and (84) (85) (86) eCN Tψm(k)i = TeCNψm(k)i = TeJmψm(k)i = eJ∗mTψm(k)i = eJ N ≡ eJ N 2 −m−1Tψm(k)i 2 −m−1ψ N 2 −m−1(−k)i, and eCN Pψm(k)i = −PeCNψm(k)i = −PeJmψm(k)i = −eJmPψm(k)i = eJ N ≡ eJ N 2 − m − 1 and N 2 +mPψm(k)i 2 +mψ N 2 +m(−k)i. Since N 2 + m should be an integer, we find that N should be an even number. Hence the class 2 −m−1(−k)i and ψ N II Dirac semimetal cannot exist in systems with eC3 sym- metry. From Eqs. (84), (85), and (86), we find that if ψm(k)i is an eigenstate at k, ψN−m−1(k)i is also a de- generate eigenstate at the same momentum k, whereas 2 +m(−k)i are degenerate at −k ψ N with the same energy as ψm(k)i. From this, we can infer the band connection near a TRIM (kTRIM) where k and −k are equivalent. The question is whether two doubly- degenerate states at k and −k are crossing or smoothly connected at kTRIM. A smooth connection of degenerate bands requires the rotation eigenvalues of the states at k and −k to be identical, which is obviously not satisfied in this case. Therefore there should be a band crossing at kTRIM where the Dirac point locates. (a) N = 4n (b) ~ J 0 ~ N-1J ~ J 1 ~ -2J N 2 ~ J N 2 -1 ~ J N 2 ~ J N-2 ~ J N 2 +1 ~ J1 ~ JN-2 4 ~ J N 2 -2 N = 4n+2 ~ ~ JN-1 J0 ~ J N 2 -1 ~ J N 2 ~ JN-2 ~ J N 2 +1 ~ J3N-2 4 FIG. 9. Constraints on the rotation eigenvalues (a) for N = 4n, (b) for N = 4n + 2 with an integer n. The (black) solid circle indicates a unit circle in the complex plane, and each (red) dot on the circle denotes eJm. Four dots connected by a dotted line are related by the P , T , and P T symmetries, hence only one of them is independent. In the case of (b), the topological charge associated with eJ(N −2)/4 and eJ(3N −2)/4 is zero. B. Topological charge From Eqs. (84), (85), and (86), we can easily find the constraints to the topological charge νm of the Dirac point locating at a TRIM. At first, the P T symmetry requires that νm = νN−m−1 2 +m. ν N 2 −m−1 = ν N (87) Moreover, since P or T symmetry interchanges the north and south pole surrounding a Dirac point at a TRIM, we find νm = νN−m−1 = −ν N 2 −m−1 = −ν N 2 +m. (88) This constraint reduces the number of independent topo- logical numbers νm. Considering that N is an even in- teger, we distinguish two cases, i.e., when N = 4n and when N = 4n+2 with an integer n. In each case, the rela- tion between different eigenstates are described in Fig. 9, from which we find the topological invariant (ν0, ..., ν N (ν0, ..., ν N −6 4 if N = 4n + 2. (89) N 4 if N = 4n, 4 −1) ∈ Z ) ∈ Z 4 N −2 Hence the topological charge of the system with eC4 or eC6 symmetry is an element of Z whereas eC2 or eC3 symmet- ric system cannot support a Dirac point with a nonzero topological charge. 13 eCN Topological charge eC2 eC3 eC4 eC6 Not allowed Not allowed Z Z TABLE III. Summary of topological charges of class II Dirac semimetals. pairs at each momentum k, and these four states cross at a TRIM, and create a Dirac point. A 4-band model can be constructed by using these four states. 1. eC2 symmetric systems Possible eJm values are eJm=0 = i exp(i 1 eJm=1 = i exp(i 3 2 π) and 2 π). Due to the P T symmetry, and {ψ′m=0(k)i,ψ′m=1(k)i} {ψm=0(k)i,ψm=1(k)i} form degenerate pairs. The symmetry constraint in Eq. (88) requires νm=0 = νm=1 = 0. (90) Hence a eC2 invariant system cannot support a stable Dirac point at a TRIM. 2. eC3 symmetric systems 3 π). 3 π), eJm=1 = Possible eJm values are eJm=0 = i exp(i 1 i exp(iπ) = −i, and eJm=2 = i exp(i 5 of ψm=1i, the P symmetry requires that eC3Pψm=1i = −PeC3ψm=1i = iPψm=1i. Thus Pψm=1i should be an eigenstate of eC3 with the eigenvalue +i, which is not allowed. Hence a eC3 invariant system cannot support a stable Dirac point at a TRIM. In the case 3. eC4 symmetric systems Due to the P T symmetry, {ψm=0i,ψm=3i} and {ψm=1i,ψm=2i} form degenerate pairs. The constraint in Eq. (88) requires that ν 1 ≡ νm=0 = νm=3 = −νm=1 = −νm=2, (91) thus there is only one independent topological invariant ν1 = ±1. C. Applications: classification of stable Dirac points in 4-band systems 4. eC6 symmetric systems Due to the P T symmetry, {ψm(k)i,ψN−m−1(k)i} and {ψN/2−m−1(k)i,ψN/2+m(k)i} form degenerate Due to the P and T symmetries, we find a ba- sis {ψm=0i,ψm=5i;ψm=2i,ψm=3i}. The constraint in Eq. (88) requires ary as eψm(δkz)i =(Ψm(N′π/2 + δkz)i, Ψm− N ′ 2 p(−N′π/2 + δkz)i, 14 (δkz ≤ 0) (δkz > 0) (93) ν1 ≡ νm=0 = νm=5 = −νm=2 = −νm=3, (92) thus there is only one independent topological invariant ν1 = ±1. eCN 4-band model Materials eC2 Not allowed eC3 Not allowed eC4 eC6 ν1 = ±1 ν1 = ±1 β-BiO2 [25] TABLE IV. Topological charges of class II Dirac semimetals for 4-band models. D. Topological charge of Dirac points protected by screw rotations Here we show that the Dirac semimetals protected by screw rotations belong to the class II, thus the topological charge of the relevant Dirac point can be determined by following the prescription described in Sec. V B. In partic- ular, we resolve a subtle issue associated with the multi- valued nature of the screw rotation eigenvalues, which we encounter when we define the topological charge of the Dirac point at the Brillouin zone boundary. To under- stand this, let us again introduce a sphere in the momen- tum space surrounding the Dirac point, and consider the two points kN and kS on the sphere passing the rotation axis. To compare the zero-dimensional topological num- bers at these two points, we need a single-valued wave function which varies smoothly around the Dirac point. However, since the eigenvalue of a screw rotation eJm(kz) is multi-valued, the relevant eigenstates also change dis- continuously at the Brillouin zone boundary. To remedy this problem, we propose a way to construct a smooth function which is single-valued around the Dirac point by modifying eigenvectors of eCN,p. Moreover, we show that such a smooth single-valued function satisfies the al- gebraic relations shown in Eq. (84), (85), and (86), thus we prove that a Dirac semimetal protected by a screw rotation belongs to the class II. First, we suppose that p/N = p′/N′ holds with an even integer N′ and an odd integer p′ that are coprime, and the Brillouin zone boundary is located at kz = N′π/2. To construct a smooth single-valued wave function around the Brillouin zone boundary at kz = N′π/2, we prepare two eigenstates Ψm(kz)i and Ψm− N ′ 2 p(kz), respectively. Then we define a hybrid wave function around the zone bound- eigenvalues eJm(kz) and eJm− N ′ 2 p(kz)i with eCN,p which is smooth and single-valued around the zone It is straightfor- boundary at δkz = 0; see Eq. (46). where eCN,peψm(δkz)i = J′m(δkz)eψm(δkz)i, 2(cid:19) π(cid:21) exp(cid:18)−i ward to show that eψm(δkz)i is also an eigenvector of eCN,p satisfying J′m(δkz) = exp(cid:20)i(cid:18) 2m + 1 δkz(cid:19) . Now we determine the eCN,p eigenvalue of Teψm(δkz)i. At first, if δkz ≤ 0, we obtain N − p′ N′ (94) (95) p′ Teψm(δkz)i = TΨm(N′π/2 + δkz)i ∝ ΨN−m−1(−N′π/2 − δkz)i = eψN−m−1+ N ′ 2 p(−δkz)i, where we have used the following relation, J∗m = JN−m−1. On the other hand, if δkz > 0, Teψm(δkz)i = TΨm− N ′ = eψN−m−1+ N ′ ∝ ΨN−m−1+ N ′ 2 p(−N′π/2 + δkz)i 2 p(N′π/2 − δkz)i 2 p(−δkz)i. From Eqs. (96) and (98), we obtain (96) (97) (98) eCN,pTeψm(δkz)i = J′ N−m−1+ N ′ 2 p = J′N−m−1+ N = J′N (−δkz)Teψm(δkz)i (−δkz)Teψm(δkz)i, 2 −m−1(−δkz)Teψm(δkz)i, 2 (99) (mod N ), (100) where we have used N 2 in which p′ is an odd integer. p′ = N 2 p = N′ 2 tained similarly from the following two relations, The eCN,p eigenvalue of Peψm(δkz)i can also be ob- Peψm(δkz ≤ 0)i = PΨm(N′π/2 + δkz)i ∝ Ψm(−N′π/2 − δkz)i = eψm+ N ′ = eψm+ N 2 p(−δkz)i (−δkz)i, 2 (101) and ∝ Ψm− N ′ Peψm(δkz > 0)i = PΨm− N ′ = eψm− N ′ = eψm+ N 2 2 p(−N′π/2 + δkz)i 2 p(N′π/2 − δkz)i 2 p(−δkz)i (−δkz)i, where we have used the fact that m is well-defined mod- ulo N . From Eqs. (101) and (102), we obtain (102) 2 (103) be obtained by following similar steps, which give (−δkz)Peψm(δkz)i. eCN,pPeψm(δkz)i = J′m+ N Finally, the eCN,p eigenvalue of P Teψm(δkz)i can also eCN,pP Teψm(δkz)i = J′N−m−1(δkz)P Teψm(δkz)i. (104) These transformation laws should be compared with Eqs. (84), (85), and (86), which differ from Eqs. (99), (103), and (104) merely due to the momentum-dependent phase factor in J′m(δkz). However, since the wave func- tion eψm(δkz)i and its associated the eigenvalue J′m(δkz) are smooth and single-valued around δkz = 0, i.e., near the Brillouin zone boundary, the topological charge of the Dirac point can be defined by using J′m(δkz = 0) at the two momenta kN and kS near the Dirac point. Once δkz is fixed to be δkz = 0, Eqs. (84), (85), and (86) are identical to Eqs. (99), (103), and (104), which shows that the Dirac point at the Brillouin zone boundary protected by a screw rotation belongs to the class II. E. Example 1: a class II Dirac semimetal on a hcp lattice To illustrate the role of screw rotations on the pro- tection of a Dirac point, let us consider a tight-binding Hamiltonian on a hexagonal close-packed (hcp) lattice, which corresponds to the space group P 63/mmc (no. 194). The hcp lattice is generated by the primitive lat- √3 tice vectors a1 = ax, a2 = a( 1 2 y), a3 = cz, and two sites in a unit cell located at r1 = 0 and r2 = 1 2 a3, respectively. The crystal has 3 a1 + 1 3 a2 + 1 2 x + 3 a1 + 2 a 6-fold screw rotation symmetry eC6,3 about the z axis centered at 2 3 a2 accompanied by a partial transla- tion 1 2 a3 as shown in Fig. 10. To confirm the presence of a single Dirac point at the zone boundary, corresponding to the A point in Fig. 11 (b), we construct the following tight binding model, c†i τxcj c†i cj − t2Xhiji H = − t1Xhiji SOXhiji + iλ(1) ij c†i σzτzcj + iλ(2) ν(1) SO Xhhijii ν(2) ij c†i σzτzcj, (105) 15 1/2 0 0 1/2 1/2 0 (b) y (c) x (a) z y 1 1 1 1/2 1/2 0 1/2 0 x 0 FIG. 10. (a) Hexagonal close-packed (hcp) lattice structure and 6-fold screw rotation. Two sublattice sites are marked with different colors. The arrows indicate a 6-fold screw rota- tion about the z axis ( eC6,3). The number on a lattice site sym- bol indicates its z-coordinate in the unit of the vertical lattice spacing c. (b) Projection of the lattice to the xy plane. (c) Spin-dependent complex hopping process between the near- est neighbor sites (λ(1) SO) and the next nearest neighbor sites (λ(2) SO). When a spin-up electron on the A sublattice hops par- allel (anti-parallel) to the arrow direction, the corresponding νij = +1 (νij = −1). where t1 (t2) indicates the nearest-neighbor hopping be- tween the same (different) sublattice sites and λ(1) SO (λ(2) SO) denotes the spin-orbit interaction between the nearest- = −ν(1,2) neighbor (next-nearest-neighbor) sites. ν(1,2) is +1 (−1) if the bond ij is parallel (anti-parallel) to the arrow direction on the bond as shown in Fig. 10 (c). σx,y,z (τx,y,z) are Pauli matrices indicating the spin (sub- lattice) degrees of freedom. ji ij In the momentum space, the Hamiltonian becomes H(k) = F0 +" (F (1) SO + F (2) SO )σz F ∗1 12 F112 −(F (1) SO + F (2) SO )σz # , where 12 indicates a 2 × 2 identity matrix and F0,1 and F (1,2) SO are given by F0 = −2t1 {cos(k · a1) + cos(k · a2) + cos[k · (a1 − a2)]} , F1 = −2t2 cos(cid:18) ckz SO = 2λ(1) F (1) F (2) SO = −2λ(2) 2 (cid:19)(cid:0)eik·b1 + eik·b2 + eik·b3(cid:1) , SO {sin(k · a1) − sin(k · a2) − sin[k · (a1 − a2)]} , SO [sin(3k · b1) + sin(3k · b2) + sin(3k · b3)] , (106) where b1 = a − a√3 2 x + a 2√3 2 x + a 2√3 SO = λ(2) y. We choose t1 = 1, t2 = 5, and λ(1) y, b2 = − a y, and b3 = SO = 5. (a) 30 E(k) 0 -30 (b) k x kz A L Γ M ky H K Γ M K Γ A L H A L M K H (a) Band structure of the tight binding model on a FIG. 11. hcp lattice. There are two Dirac points at the A and L points. (b) First Brillouin zone of the hcp lattice. thus 16 where eC6,3(kz) means the representation of eC6,3 in a Bloch basis, in which the momentum dependent phase factor results from the partial lattice translation along the z direction. Finally, for ky → −ky, we find F0(kx,−ky, kz) = F0(kx, ky, kz), ReF1(kx,−ky, kz) = ReF1(kx, ky, kz), ImF1(kx,−ky, kz) = −ImF1(kx, ky, kz), F (1) SO (kx,−ky, kz) = F (1) SO (kx,−ky, kz) = −F (2) F (2) SO (kx, ky, kz), SO (kx, ky, kz), The resulting band structure is shown in Fig. 11. We can clearly see two Dirac points at A and L, respectively. Between these two points, the Dirac point at A is the one protected by the 6-fold screw rotation, hence is located at the boundary of the rotation axis (z-axis). To confirm the symmetry protection of the Dirac point at A and its characteristic dispersion, let us examine the symmetry of the Hamiltonian. The symmetries, which are important for the protection of Dirac points, are the time-reversal T , the inversion P , the 6-fold screw rotation 2 z} where My transforms the spatial coordinate (x, y, z) to (x,−y, z). To find the matrix representation of each sym- metry operator, we can use the following information. At first, for k → −k, we find 2 z}, and the glide symmetry fMy = {My c eC6,3 = {C6 c F0(−k) = F0(k), ReF1(−k) = ReF1(k), ImF1(−k) = −ImF1(k), F (1,2) SO (−k) = −F (1,2) SO (k), (107) which gives P = τx, T = iσyK, (108) where K is a complex conjugation operator. Moreover, under π/3 rotation about the z axis, we obtain (kx + iky) → (k′x + ik′y) = (kx + iky) exp(cid:16)i kz → k′z = kz π 3(cid:17) , (109) and thus F0(k′) = F0(k), ReF1(k′) = ReF1(k), ImF1(k′) = −ImF1(k), F (1,2) SO (k′) = −F (1,2) SO (k), eC6,3(kz) = τx exp(cid:16)i π 6 σz(cid:17) exp(cid:18)−i ckz 2 (cid:19) , (110) (111) (112) (113) fMy(kz) = iτxσy exp(cid:18)−i ckz 2 (cid:19) . Let us note that F (2) On the kz axis with kx = ky = 0, the Hamiltonian SO term breaks the glide mirror fMy. becomes H(kz) = F0 + F1(kz)τx, (114) from which we find two degenerate eigenstates, , , 1 0 1 0 0 1 0 1 , , 1 √2 1 √2 ψ−1i = ψ+2i = ψ+1i =       with the eigenvalue E+(kz) = F0 + F1(kz), and the other two degenerate eigenstates,   with the eigenvalue E−(kz) = F0 − F1(kz). Let us note that ψ+1i, ψ+2i, ψ−1i, ψ−2i are also the eigenstates of eC6,3 with the corresponding eigenvalues exp(−i ckz 2 + 6 ), exp(−i ckz 6 ), exp(−i ckz i π 2 + 6 ), respectively. They are exactly the eC6,3 eigenstates i 5π [See Eq. (72)], which can support a single Dirac point at the zone boundary on the rotation axis, i.e., at the A point with the momentum k = (0, 0, π c ). The low-energy Hamiltonian near the A point is given by 6 ), exp(−i ckz 0 1 0 −1 1 0 −1 0 ψ−2i = 2 + i 11π 2 + i 7π 1 √2 1 √2 HA(q) ≈ 3t2qz 1 − y − q3 λ(1) SO(3qxq2 4 +h 1 q2 x + q2 y 12 ! τx + 3√3 4 x) + t2 24√3 (3q2 λ(2) SO(3q2 xqy − q3 y)qzτy xqy − q3 y)iτzσz, (115) where the momentum q are measured relative to the A point assuming a = c = 1 and the constant term F0 is dropped. It is interesting to note that the dispersion on On the other hand, thus thus c c 2 2(cid:17) , PfMy : (x, y, z) →(cid:16)−x, y,−z − zo . PfMy =nP My − fMyP : (x, y, z) →(cid:16)−x, y,−z + 2(cid:17) , zo . fMyP =nP My c 2 c (116) (117) (118) (119) the (qx, qy) plane is cubic whereas it is linear along the qz direction. The cubic dispersion arises due to the an- gular momentum difference between the conduction and valence bands, as indicated in Eq. (72). Thus we obtain a cubic Dirac point at the A point, which is protected by Finally, the 6-fold screw rotation eC6,3. let us briefly explain the physical origin of the Dirac point at the L point with the momentum k = (0, 2π√3a c ). At the L point, the system is invariant under a set of point group symmetry operations, which is so-called the little co-group at L, G . The little co- group G is generated by three symmetry operations, a , π L L glide mirror fMy, the inversion P and the two-fold ro- 2 z involved in fMy, fMy and P do not tation about the y axis C2y. Due to the partial lattice translation t = c commute. Namely, we find that Therefore fMy and P anti-commute at the zone boundary SO term breaking the fMy with kz = π/c, which guarantees the four-fold degener- acy. Since two-fold rotation symmetry cannot support a stable Dirac point, C2y symmetry cannot play an impor- tant role here. Moreover, the F (2) vanishes at the L point. To establish a general theory about the protection of a Dirac point by a glide mirror symmetry and its associated topological charge is an in- teresting research topic, which we leave for future study. F. Example 2: a class II Dirac semimetal on a diamond lattice As a second example of class II Dirac semimetal, let us consider the Fu-Kane-Mele Hamiltonian on a diamond lattice40, H = tXhiji c†i cj + 8i λSO a2 Xhhijii c†i σ · (d1 ij × d2 ij )cj,(120) (a) 4 E(k) 0 - 4 Γ X W K L Γ 17 kz X W K L Γ ky (b) kx FIG. 12. (a) Band structure of the Fu-Kane-Mele model on a diamond lattice. There is a Dirac point at the X point. (b) First Brillouin zone of a diamond lattice. (a) z (b) 3/4 1/4 1/2 1 0 y 1/4 1/2 0 1 y 3/4 x x FIG. 13. Structure of a diamond lattice. Two sublattice sites are marked by using different colors. The arrows indicate a 4-fold screw rotation about the z axis. spin degrees of freedom. a denotes the cubic lattice con- stant. In the momentum space, the Hamiltonian becomes H(k) = P3 i=1 Fiσi F012 F ∗0 12 i=1 Fiσi! , −P3 where 12 indicates a 2× 2 identity matrix and F0,1,2,3 are given by 4 (kx+ky+kz ) + e 4 (−kx+ky−kz ) + e ia 4 (kx−ky−kz) ia ia +e F0 = t(cid:2)e F1 = 4λSO sin(cid:16) akx F2 = 4λSO sin(cid:16) aky F3 = 4λSO sin(cid:16) akz ia 4 (−kx−ky+kz )(cid:3), 2 (cid:1)i, 2 (cid:1) − cos(cid:0) akz 2 (cid:1)i, 2 (cid:1) − cos(cid:0) akx 2 (cid:1)i.(121) 2 (cid:1) − cos(cid:0) aky 2 (cid:17)h cos(cid:0) aky 2 (cid:17)h cos(cid:0) akz 2 (cid:17)h cos(cid:0) akx This Hamiltonian exhibits 3D bulk Dirac points at three inequivalent X points X r = 2πr/a where r = x, y, z. Each Dirac point at X r is protected by the 4-fold screw rotation about r axis. where the first term indicates the nearest-neighbor hop- ping and the second term connects the second-nearest- neighbors with a spin dependent amplitude. d1,2 ij are the two nearest-neighbor bond vectors traversed between sites i and j, and σx,y,z are Pauli matrices indicating the To understand the role of the screw rotation, let us de- scribe the symmetry of the system. The space group of the diamond lattice is F d3m(O7 h), which contains the 24 symmorphic elements of tetrahedral point group ¯43m(Td) and 24 non-symmorphic elements. The non-symmorphic 4 , a elements are obtained by compounding each symmor- phic symmetry operation of Td with a translation along td = ( a 4 ), which takes one sublattice site to another inequivalent sublattice site. On the other hand, the sym- morphic symmetry operation connects sites belonging to the same sublattice. 4 , a At a generic point ∆ = (0, 0, kz) on the kz axis, the system has C4v symmetry that is composed of 5 different symmetry classes with elements {E0}, {C2 40}, 2{C4td}, 2{iC2 4td}, 2{iC′20}, respectively. It is worth to note that some symmetry elements contain a partial lattice transla- tion td = ( a 4 ) which is the characteristic property of the non-symmorphic nature of the diamond lattice space group. This contrasts with the case of conventional sym- morphic cubic lattices such as simple cubic (sc), face cen- tered cubic (fcc), body centered cubic (bcc) lattices where the system on the kz axis has the ordinary C4v group containing only symmorphic point group operations41. 4 , a 4 , a tation eC4,1 ≡ {C4ztd} = exp(−ik·td)τx exp(i π By considering the symmetry of the Hamiltonian, one can easily find the matrix representation of the screw ro- 4 σz) where exp(−ik · td) represents the translation td of the Bloch state with the momentum k, τx indicates the sublattice change due to partial translation, and exp(i π 4 σz) repre- sents the 4-fold rotation. Then it is straightforward to confirm that Since F1,2,3 = ImF0 = 0 on the kz axis, the Hamilto- 4,1 = H(−ky, kx, kz). (122) eC4,1H(kx, ky, kz)eC−1 nian on the kz axis becomes H(kz) = 4t cos( akz 4 )τx, (123) from which we obtain one 4t cos( 1 4 akz) with the corresponding eigenvectors eigenvalue E+(k) = 1 0 1 0 0 1 0 1 , 1 √2 ψ+1i = 1 √2 , ψ+2i = 4 akz) with     and the other eigenvalue E−(k) = −4t cos( 1 the eigenvectors     Let us note that ψ+1i, ψ+2i, ψ−1i, ψ−2i are also the eigenstates of eC4,1 with the eigenvalues exp(−i akz 4 + i π 4 ), exp(−i akz 4 ), exp(−i akz 4 + i 5π 4 − i 5π 4 ), respectively. Since the system is 4π a periodic along the kz direction, a single Dirac point protected by eC4,1 can be created at the Brillouin zone boundary. It is straightforward to see that these four eigenvalues have 4 ), exp(−i akz 0 1 0 −1 1 0 −1 0 4 − i π , ψ−2i = ψ−1i = 1 √2 1 √2 . 18 X the same form as Eq. (61), hence satisfy the condition for the band crossing at the Brillouin zone boundary with kz = ±2π/a. Also the band structure of the system along the kz axis is consistent with Fig. 8. Before we close this subsection, let us briefly perform a group theoretical analysis at the X point. The lit- tle co-group G at the X point with the momentum k = (0, 0, 2π a ) is generated by three symmetry operators, i.e., the four-fold screw rotation eC4,1 about the z axis, the inversion P , the two-fold rotation about the x axis 27,43. One interesting property of the diamond lat- C2x tice is that the inversion symmetry P also accompanies a partial translation td. Thus, it is more suitable to use lation associated with the inversion. Here we describe two interesting physical consequences resulting from the the notation eP = {Ptd} to indicate the partial trans- non-symmorphic nature of eP . Firstly, the partial translation involved in eP does not affect the commutation relation between eP and a screw rotation eCN,q on the rotation axis. This fact can be mation under the combination of eCN,q = {CNτq} and eP = {PtP}. Assuming the z axis is the screw rotation axis, ePeCN,q transforms the coordinate (x, y, z) to (−x′ − τq,x + tP,x,−y′ − τq,y + tP,y,−z − τq,z + tP,z), (124) easily understood by considering the coordinate tranfor- where x′ and y′ are rotated coordinates satisfying x′ + iy′ = (x + iy) exp(i2π/N ). Thus we obtain ePeCN,q = {P CN − τq + tP}. On the other hand, eCN,qeP transforms (x, y, z) to (−x′ + τq,x + t′P,x,−y′ + τq,y + t′P,y,−z + τq,z + tP,z), (126) (125) where t′P,x + it′P,y = (tP,x + itP,y) exp(i2π/N ). Thus we obtain eCN,qeP = {P CNτq + t′P}. Now let us consider a Bloch state kzi on the rotation axis with the momentum k = (0, 0, kz). We find (127) ePeCN,qkzi = exp [ikz(−τq,z + tP,z)] P CNkzi, eCN,qePkzi = exp [ikz(τq,z + tP,z)] P CNkzi, which shows that the partial translation tP,z associated with the inversion only provides an overall phase factor, and does not affect the commutation relation whereas τq,z does. Therefore our theory can also be applied to systems (128) with the inversion eP accompanying a partial translation, as long as the Dirac point is located on the rotation axis. However, when the Dirac point is located away from the rotation axis, the translation tP can cause nontrivial physical consequence as well. For instance, in the kz = 0 plane, perpendicular to the rotation axis, a Bloch state kx, kyi satisfies, ePeCN,qkx, kyi = exp(cid:2) − ik′x(τq,x − tP,x) − ik′y(τq,y − tP,y)(cid:3)P CNkx, kyi, (129) and eCN,qePkx, kyi = exp(cid:2)ik′x(τq,x + t′P,x) + ik′y(τq,y + t′P,y)(cid:3)P CNkx, kyi. (130) where k′x +ik′y = (kx +iky) exp(i2π/N ) and we have used the relation k′xt′P,x + k′yt′P,y = kxtP,x + kytP,y. The point is that since (tP,x, tP,y) 6= (t′P,x, t′P,y) due to the rota- tion, the partial translation tP associated with the inver- sion can also modify the commutation relation between the inversion and the rotation symmetries. Because of this, even a symmorphic rotation symmetry, which is not accompanied by a translation, can create a Dirac point away from the rotation axis when it is combined with the non-symmorphic inversion symmetry. For illustration, let us consider the commutation re- ( a 4 , a 4 , a and 4 ). The Bloch state kx, kyi satisfies, lation between C2z and eP = {PtP} where tP = td = eP C2zkx, kyi = exp(cid:2)ikx(−tP,x) + iky(−tP,y)(cid:3)P C2zkx, kyi, C2zePkx, kyi = exp(cid:2)ikx(tP,x) + iky(tP,y)(cid:3)P C2zkx, kyi. a , 0) where exp(cid:0)ikxtP,x + ikytP,y(cid:1) = i, we obtain {eP , C2z} = 0. This anti- commutation relation can create a stable Dirac point at k1 and k2, which is again confirmed by K-theory analysis in the Appendix B 3. Therefore, although the symmor- phic C2 symmetry cannot support a stable Dirac point on the rotation axis, it can create a stable Dirac point at the zone boundary in the plane perpendicular to the rotation axis, when the C2 is combined with a non-symmorphic Hence at the high symmetry momentum such as k1 = ( 2π a , 0, 0) or k2 = (0, 2π (131) (132) In fact, as noted before, there are three symmetry gen- a ) in the diamond lattice. According to the previous discus- sion, the Dirac point at X can be protected not only by inversion symmetry eP . erators, eP , eC4,1, C2x, at the X point with k = (0, 0, 2π eP and eC4,1 satisfying {eP ,eC4,1} = 0 but also by eP and C2x, which are also anti-commuting {eP , C2x} = 0. Be- cause of the high crystalline symmetry, the Dirac point in the diamond lattice is protected by multiple pairs of symmetry operators27. VI. DISCUSSION 19 To sum up, we have studied the topological charge of 3D Dirac semimetals protected by the time-reversal, the inversion, and the rotation symmetries. Consideration of topological charges naturally leads to two different classes of Dirac semimetals, which is consistent with the previ- ous observation based on the symmetry constrained min- imal Hamiltonian analysis22. Class I Dirac semimetals are protected by an ordinary symmorphic rotation sym- metry which commutes with the inversion. Since each eigenstate carries a quantized rotation eigenvalue on the rotation axis, Dirac points should form a pair having the opposite topological charges when the system is periodic along the rotation axis. On the other hand, class II Dirac semimetals are associated with non-symmorphic screw rotation symmetries. The eigenvalue of a screw rotation is not quantized on the rotation axis due to the phase fac- tor induced by a partial lattice translation, which enables to create a single isolated Dirac point at the Brillouin zone boundary. The nonzero topological charge of a Dirac point not only guarantees the stability of the gap-closing point, but can trigger new types of Lifshitz transitions. For in- stance, when two Dirac points merge at the same momen- tum, the topological charge of the merging point is given by the summation of their topological charges. Since the energy dispersion around the gap-closing point strongly depends on its topological charge, such a merging tran- sition can generate intriguing nodal quasi-particles with novel physical properties44–47. Moreover, the presence of a quantized topological charge can be a source of new topological responses. For instance, it is well-known that the nonzero monopole charge of Weyl points induces novel topological responses in Weyl semimetals11. Recent theoretical studies of interesting topological responses in Dirac semimetals48,49 may imply nontrivial role of topo- logical charges in these systems. Up to now, two materials (Na3Bi and Cd3As2) belong- ing to the class I are discovered and extensively studied whereas class II Dirac semimetals are not uncovered yet. Though there are some hypothetical candidate materi- als proposed theoretically25,26, all of them are chemically unstable because the metallic ion in each candidate com- pound is required to have a lone-pair valence electron to locate the Fermi level at the Dirac point42. In this re- spect, synthesizing class II Dirac semimetals is a challeng- ing problem in material science which should be properly addressed in near future. We believe that class II Dirac semimetals are as impor- tant as class I Dirac semimetals in the following sense. In the case of class I Dirac semimetals, the Dirac points are created by a band inversion, hence the intrinsic proper- ties of 3D Dirac particles can be observed only within the narrow energy scale associated with the band inversion9. Because of this, if the competing energy scales, such as the Fermi energy due to doped carriers or the disorder- induced broadening, are comparable to the band inver- sion energy, the intrinsic properties of 3D Dirac particles can be easily washed out. However, in the case of class II Dirac semimetals, the energy scale of the Dirac dispersion is simply given by the bandwidth of the system (roughly in the order of a few eV), which obviously provides a better playground to study the intrinsic properties of 3D Dirac particles. Secondly, we would like to draw attention to class II Dirac semimetals as potential novel topological states. The fact that the Fu-Kane-Mele model is a canonical model to construct a 3D Z2 topological insulator implies the intrinsic topological nature of the associated Dirac semimetal state. In particular, in the present paper, we have demonstrated that the presence of a single Dirac point on the rotation axis is unnatural in consideration of the Nielsen-Ninomiya theorem, and, in fact, the pro- jective nature of the screw rotation symmetry plays an essential role to circumvent the doubling of Dirac points. Although the discrete nature of the rotation symme- try should be distinct from the continuous U(1) symme- try associated with the original Nielsen-Ninomiya theo- rem, the mechanism leading to circumventing the fermion number doubling shares the common origin, i.e., assign- ing a non-quantized quantum number to fermions. To re- veal the topological properties of class II Dirac semimet- als would definitely be an exciting theoretical problem which we leave for future studies. Finally, we would like to point out that there are a class of Dirac semimetals which are not completely treated in our classification scheme. Let us note that, in both class I and II Dirac semimetals considered in the present work, Dirac points are located on the rotation axis. However, rotation symmetries can also protect a Dirac point which is away from the rotation axis. For instance, we have shown in Sec. V F that the symmorphic C2 rotation can protect a Dirac point which is not on the rotation axis, when it is combined with a non-symmorphic inversion symmetry. Moreover, the tight-binding model on a hcp lattice considered in Sec. V E indicates that glide mir- ror symmetries can also give rise to symmetry-protected Dirac points. To find a systematic way to classify these different types of Dirac semimetals would also be an im- portant problem for future research. ACKNOWLEDGMENTS BJY is supported from the Japan Society for the Pro- motion of Science (JSPS) through the ‘Funding Program for World-Leading Innovative R&D on Science and Tech- nology (FIRST Program), and Grant-in-Aids for Scien- tific Research (Kiban (S), No. 24224009) from the Min- istry of Education, Culture, Sports, Science and Technol- ogy (MEXT). AF is grateful for support by Grants-in-Aid from the Japan Society for Promotion of Science (Grant No.15K05141) and by the RIKEN iTHES Project. 20 Appendix A: Absence of a stable Dirac point in systems with time-reversal and inversion symmetries only Here we prove that stable Dirac points do not exist in systems having only the time-reversal (T ) and inver- sion (P ) symmetries. For this, we distinguish two cases: one is when the Dirac point is located at a generic mo- mentum point, and the other is when the Dirac point is located at a time-reversal invariant momentum (TRIM). This distinction is necessary because the symmetry as- sociated with the Dirac point differs depending on the position of the Dirac point in the momentum space. In each case, the stability of a Dirac point is determined by using K-theory50,51. 1. A Dirac point located at a generic momentum When the Dirac point locates at a generic momentum, the combination of T and P is the only symmetry satis- fied around the Dirac point. In general, a P T symmetric system satisfies (P T )H(k)(P T )−1 = H(k), (A1) where the anti-unitary P T symmetry satisfies (P T )2 = −1, which is coming from P 2 = 1, T 2 = −1, [P, T ] = 0, (A2) in electronic systems. x, k0 y, k0 The stability of the Dirac point can be understood by using K theory approach. We consider a Dirac point lo- cating at a generic momentum k0 = (k0 z ). The effective Hamiltonian describing the low energy excita- tion around the Dirac point is given by HD = (kx − k0 where γ0,x,y,z are gamma matrices satisfying the an- ticommutation relations {γi, γj} = 2δi,j, and m indi- cates a possible symmetry-preserving Dirac mass term. The presence (absence) of the symmetry-preserving Dirac mass term m indicates the instability (stability) of the Dirac point. From Eq. (A1), we obtain x)γx + (ky − k0 y)γy + (kz − k0 z )γz + mγ0, [γx, P T ] = [γy, P T ] = [γz, P T ] = [γ0, P T ] = 0. (A3) To confirm the existence or absence of the Dirac mass term m, let us define a real Clifford algebra generated by P T and γ0,x,y,z, which is given by Cl0,4 ⊗ Cl2,0 = {; γx, γy, γz, γ0} ⊗ {P T, JP T ;},(A4) where Clp,q indicates a real Clifford algebra with p + q generators {e1, e2, ..., ep; ep+1, ep+2, ..., ep+q} satisfying {ei, ej} = 0, i 6= j i =(−1, +1, e2 1 ≤ i ≤ p, p + 1 ≤ i ≤ p + q. 21 The algebra in Eq. (A4) can be easily obtained by con- sidering the following relations: where the matrices γ0,x,y,z satisfy {γi, γj} = 2δi,j. T and P symmetries require (i) [γx, P T ] = [γy, P T ] = [γz, P T ] = [γ0, P T ] = 0, (ii) [γx, J] = [γy, J] = [γz, J] = [γ0, J] = 0, (iii){γi, γj} = 2δi,j, (iv){P T, JP T} = 0, (P T )2 = (JP T )2 = −1, (A5) where the symbol J indicating the pure imaginary num- ber i is introduced to construct a real Clifford algebra. The existence or absence of the Dirac mass mγ0 can be judged by considering the following extension problem: i.e., {; γx, γy} ⊗ {P T, JP T ;} → {; γx, γy, γz} ⊗ {P T, JP T ;}, Cl0,2 ⊗ Cl2,0 → Cl0,3 ⊗ Cl2,0. (A6) (A7) Namely, the topological classification of γz determines the topological nature of the Dirac point. This is because the topologically trivial classification of γz implies the existence of another gamma matrix such as γ0 which an- ticommutes with the three generators γx,y,z (i.e., a mass term exists) whereas the topologically nontrivial classifi- cation of γz implies the absence of γ0, thus the topolog- ically nontrivial nature of the Dirac point. Generally, in the classification scheme with Clifford algebera, the exis- tence condition of a particular generator ei (Dirac mass term) is equivalent to the classification of another gen- erator of the same type as ei in the Clifford algebra in which ei is removed.23 The extension problem Cl0,2 ⊗ Cl2,0 → Cl0,3 ⊗ Cl2,0 is equivalent to due to the relation Cl4,0 → Cl5,0 Clp,q ⊗ Cl2,0 ≃ Clq+2,p. (A8) (A9) (A10) Since the classifying space for the extension Clp,q → Clp+1,q is given by Rp+2−q, the classifying space for the extension Cl4,0 → Cl5,0 is R6 = Sp(n)/U (n) with a sufficiently large integer n. Since π0(R6) = 0, the space of possible representation for γz is singly connected. Namely, a Dirac mass term always exists, hence the Dirac point is unstable. 2. A Dirac point locating at a TRIM On the other hand, when the Dirac point locates at a TRIM, both P and T are the symmetry of the Dirac point. To understand the stability of the Dirac point, we consider the following Dirac Hamiltonian HD = kxγx + kyγy + kzγz + mγ0, (A11) {γx, T} = {γy, T} = {γz, T} = [γ0, T ] = 0, {γx, P} = {γy, P} = {γz, P} = [γ0, P ] = 0, (A12) where P 2 = 1, T 2 = −1, [T, P ] = 0. (A13) The Clifford algebra generated by γ0,x,y,z, T , P , J is given by {T, JT, Jγ0; γx, γy, γz, P γxγyγz}, (A14) The existence or absence of the Dirac mass mγ0 can be judged by considering the following extension problem: {JT ; γx, γy, γz, P γxγyγz} → {T, JT ; γx, γy, γz, P γxγyγz} i.e., Cl1,4 → Cl2,4, (A15) (A16) the corresponding classifying space is Rp+2−q = R−1 ≃ R7. Since π0(R7) = 0, a Dirac mass term always exists, hence the Dirac point is unstable. Therefore, indepen- dent of the location of the Dirac point in the momentum space, the system with only T and P symmetries cannot support a stable Dirac point. Appendix B: The stability of Dirac points in C2 symmetric systems In C2 invariant systems, since all the symmetry opera- tors squared become ±1, the classification scheme based on Clifford algebras can be applied. 1. When [C2, P ] = 0 In systems with the symmorphic C2 rotation satisfying (C2)2 = −1, the Dirac point can be located at a generic momentum k0 = (0, 0, k0 z) on the rotation axis (z axis). We consider the following massive Dirac Hamiltonian, H(k) = kxγx + kyγy + (kz − k0 which satisfy the following relations, z )γz + mγ0, (P T )H(k)(P T )−1 = H(k), C2H(kx, ky, kz)(C2)−1 = H(−kx,−ky, kz). (B1) (B2) (B3) From this, we obtain [γx, P T ] = [γy, P T ] = [γz, P T ] = [γ0, P T ] = 0, {γx, C2} = {γy, C2} = [γz, C2] = [γ0, C2] = 0. (B4) (B5) Then we can determine the Clifford algebra generated by the gamma matrices in the Dirac Hamiltonian and the relevant symmetry operators. The resulting Clifford algebra is rotation axis. The commutation relations relevant to this problem are as follows. 22 Cl6,0 ⊗ Cl0,1 = {P T, JP T, Jγx, Jγy, Jγz, Jγ0;} ⊗ {; γxγyC2}. (B6) The relevant extension problem is Cl4,0 ⊗ Cl0,1 → Cl5,0 ⊗ Cl0,1, (B7) for which the classifying space is given by R6 × R6. From its zeroth homotopy group, we find the topological charge as π0(R6×R6) = 0. Therefore a Dirac point cannot carry a nontrivial topological charge, which is consistent with the absence of a topological invariant found before. (See Table I.) 2. When { eC2, P } = 0 Now we consider a two-fold screw rotation eC2 satisfy- ing (eC2)2 = 1. Since the Dirac point locates at a TRIM, both the P and the T are the symmetry of the Dirac point. The commutation relations relevant to this prob- lem are as follows. {γx, T} = {γy, T} = {γz, T} = [γ0, T ] = 0, {γx, P} = {γy, P} = {γz, P} = [γ0, P ] = 0, {γx,eC2} = {γy,eC2} = [γz,eC2] = [γ0, eC2] = 0, [T, P ] = 0, [eC2, T ] = 0, {eC2, P} = 0, P 2 = 1, T 2 = −1, (eC2)2 = 1. and in which The relevant Clifford algebra of gamma matrices and symmetry operators is given by (B8) (B9) (B10) Cl3,5 = {T, JT, Jγ0; γx, γy, γz, P γxγyγz,eC2P γz}. (B11) The existence condition of the Dirac mass term is deter- mined by the extension problem Cl1,5 → Cl2,5, (B12) for which the classifying space is R6. From π0(R6) = 0, we see that the eC2 symmetry cannot protect a Dirac point consistent with Table III. 3. When {C2, eP } = 0 This is relevant to the case when the Dirac point is located at a TRIM in the plane perpendicular to the {γx, T} = {γy, T} = {γz, T} = [γ0, T ] = 0, {γx, C2} = {γy, C2} = [γz, C2] = [γ0, C2] = 0, {γx, eP} = {γy,eP} = {γz, eP} = [γ0,eP ] = 0, [T,eP ] = 0, [C2, T ] = 0, {C2,eP} = 0, eP 2 = −1, T 2 = −1, (C2)2 = −1. and in which Then the relevant Clifford algebra of gamma matrices and symmetry operators is given by (B13) (B14) (B15) Cl3,5 = {T, JT, Jγ0, eP γxγyγz; γx, γy, γz, C2eP γz}. The existence condition of the Dirac mass term is deter- mined by the extension problem (B16) Cl2,4 → Cl3,4, (B17) for which the classifying space is R0. From π0(R0) = Z, we see that the symmorphic C2 symmetry can pro- tect a Dirac point when it is combined with the non- of the Dirac point is not on the rotation axis but at a TRIM on the plane perpendicular to the rotation axis symmorphic inversion symmetry eP . Here the location because the anti-commutation relation {C2,eP} = 0 can be satisfied only away from the rotation axis. Appendix C: The stability of 2D Dirac points in systems with two-fold screw rotations Recently, Young and Kane proposed a theory52 about 2D Dirac points located at a TRIM on the Brilluoin zone boundary. One interesting finding in their work is that a two-fold screw rotation can protect a Dirac point at a TRIM on the rotation axis, which is forbidden in the case of 3D systems. To confirm their claim, let us check the stability of the Dirac point by using K theory. Let us consider a Dirac Hamiltonian at the zone bound- ary. H(k) = kxγx + kyγy + mγ0, (C1) where γ0,x,y are mutually anti-commuting. Under P , T , eC2y symmetry satisfying the gamma matrices satisfy [P, T ] = [eC2y, T ] = 0, {eC2y, P} = 0, P 2 = 1, T 2 = −1, eC2 2y = 1, {γx, T} = {γy, T} = [γ0, T ] = 0, {γx, P} = {γy, P} = [γ0, P ] = 0, {γx, eC2y} = [γy,eC2y] = [γ0, eC2y] = 0, (C2) (C3) 23 The relevant Clifford algebra of gamma matrices and symmetry operators is given by Cl4,2 ⊗ Cl1,0 = {T, T J, Jγ0,eC2yγx; γx, γy} ⊗ {P γxγy;}(C4) The existence of the Dirac mass term mγ0 is determined by the extension problem Cl2,2 ⊗ Cl1,0 → Cl3,2 ⊗ Cl1,0. (C5) Since the extra generator P γxγy commutes with all the other generator, and satisfies (P γxγy)2 = −1, the above extension problem is rearranged in the following way, Cl4 → Cl5, (C6) for which the classifying space is C4 ∼= C0. Namely, the extra generator P γxγy deforms the original real Clifford algebra extension problem to a complex Clifford algebra extension problem53. From π0(C0) = Z, we see that eC2 can protect a 2D Dirac point with topological charge Z on the rotation axis. 1 A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). 2 Z. Wang, Y. Sun, X. -Q. Chen, C. Franchini, G. Xu, H. Weng, X. Dai, and Z. Fang, Phys. Rev. B 85, 195320 (2012). 3 Z. Wang, H. Weng, Q. Wu, X. Dai, and Z. Fang, Phys. Rev. B 88, 125427 (2013). 4 Z. K. Liu, B. Zhou, Y. Zhang, Z. J. Wang, H. M. Weng, D. Prabhakaran, S.-K. Mo, Z. X. Shen, Z. Fang, X. Dai, Z. Hussain, and Y. L. Chen, Science 343, 864 (2014). 5 S.-Y. Xu, C. Liu, S. K. Kushwaha, T.-R. Chang, J. W. Krizan, R. Sankar, C. M. Polley, J. Adell, T. Balasubra- manian, K. Miyamoto, N. Alidoust, G. Bian, M. Neupane, I. Belopolski, H.-T. Jeng, C.-Y. Huang, W.-F. Tsai, H. Lin, F. C. Chou, T. Okuda, A. Bansil, R. J. Cava, and M. Z. Hasan, arXiv:1312.7624. 6 S.-Y. Xu, C. Liu, S. K. Kushwaha, R. Sankar, J. W. Krizan, I. Belopolski, M. Neupane, G. Bian, N. Alidoust, T.-R. Chang, H.-T. Jeng, C.-Y. Huang, W.-F. Tsai, H. Lin, P. P. Shibayev, F.-C. Chou, R. J. Cava, M. Z. Hasan, Science, 347, 294 (2014). 7 S. Borisenko, Q. Gibson, D. Evtushinsky, V. Zabolotnyy, B. Buchner, and R. J. Cava, Phys. Rev. Lett. 113, 027603 (2014). 8 M. Neupane, Su-Yang Xu, R. Sankar, N. Alidoust, G. Bian, C. Liu, I. Belopolski, T.-R. Chang, H.-T. Jeng, H. Lin, A. Bansil, F. Chou, and M. Zahid Hasan, Nat. Commun. 5, 3786 (2014). 9 S. Jeon, B. B. Zhou, A. Gyenis, B. E. Feldman, I. Kimchi, A. C. Potter, Q. D. Gibson, R. J. Cava, A. Vishwanath, and A. Yazdani, Nat. Mater. 13, 851 (2014). 10 Z. K. Liu, J. Jiang, B. Zhou, Z. J. Wang, Y. Zhang, H. M. Weng, D. Prabhakaran, S-K. Mo, H. Peng, P. Dudin, T. Kim, M. Hoesch, Z. Fang, X. Dai, Z. X. Shen, D. L. Feng, Z. Hussain, Y. L. Chen, Nat. Mater. 13, 677 (2014). 11 X.Wan, A.M. Turner, A. Vishwanath, and S. Y. Savrasov, Phys. Rev. B 83, 205101 (2011). 12 P. Horava, Phys. Rev. Lett. 95, 016405 (2005). 13 B. B´eri, Phys. Rev. B 81, 134515 (2010). 14 A. P. Schnyder and S. Ryu, Phys. Rev. B 84, 060504 (2011). 15 P. M. R. Brydon, A. P. Schnyder, and C. Timm, Phys. Rev. B 84, 020501 (2011). 16 A. P. Schnyder, P. M. R. Brydon, and C. Timm, Phys. Rev. B 85, 024522 (2012). 17 J. L. Manes, Phys. Rev. B 85, 155118 (2012). 18 S. Matsuura, P. Y. Chang, A. P. Schnyder and S. Ryu, New J. Phys. 15, 065001 (2013). 19 Y. X. Zhao and Z. D. Wang, Phys. Rev. Lett. 110, 240404 (2013). 20 Y. X. Zhao and Z. D. Wang, Phys. Rev. B 89, 075111 (2014). 21 C. K. Chiu and A. P. Schnyder, Phys. Rev. B 90, 205136 (2014). 22 B. -J. Yang and N. Nagaosa, Nat. Commun. 5, 4898 (2014). 23 T. Morimoto and A. Furusaki, Phys. Rev. B 89, 235127 (2014). 24 S. A. Yang, H. Pan, F. Zhang, Phys. Rev. Lett. 113, 046401 (2014). 25 S. M. Young, S. Zaheer, J. C. Y. Teo, C. L. Kane, E. J. Mele, and A. M. Rappe, Phys. Rev. Lett. 108, 140405 (2012). 26 J. A. Steinberg, S. M. Young, S. Zaheer, C. L. Kane, E. J. Mele, and A. M. Rappe, Phys. Rev. Lett. 112, 036403 (2014). 27 T. Bzdusek, A. Ruegg, M. Sigrist, Phys. Rev. B 91, 165105 (2015). 28 C. -X. Liu, R. -X. Zhang, and B. K. VanLeeuwen, Phys. Rev. B 90, 085304 (2014). 29 L. Michel and J. Zak, Phys. Rev. B 59, 5998 (1999). 30 L. Michel and J. Zak, Europhys. Lett. 50, 519 (2000). 31 L. Michel and J. Zak, Physics Reports 341, 377 (2001). 32 S. A. Parameswaran, A. M. Turner, D. P Arovas, and A. Vishwanath Nat. Phys. 9, 299 (2013). 33 A. Narayan, D. DiSante, S. Picozzi, and S. Sanvito, Phys. Rev. Lett. 113, 256403 (2014). 34 S. Murakami, New J. Phys., 9, 356 (2007). 35 A. A. Burkov and L. Balents, Phys. Rev. Lett. 107, 127205 (2011). 36 H. B. Nielsen and M. Ninomiya, Nucl. Phys. B 185 20 (1981). 37 H. B. Nielsen and M. Ninomiya, Nucl. Phys. B 193 173 (1981). 38 A. Altland and M. R. Zirnbauer, Phys. Rev. B 55, 1142 (1997). 39 X. -L. Sheng, Z. Wang, R. Yu, H. Weng, Z. Fang, and X. Dai, Phys. Rev. B 90, 245308 (2014). 40 L. Fu, C. L. Kane, and E. J. Mele, Phys. Rev. Lett. 98, 106803 (2007). 41 M. S. Dresselhaus, G. Dresslhaus, and A. Jorio, Group Theory: Application to the Physics of Condensed Matter (Springer-Verlag, Berlin, Heidelberg, 2008). 42 Q. D. Gibson, L. M. Schoop, L. Muechler, L. S. Xie, M. Hirschberger, N. P. Ong, R. Car, R. J. Cava, Phys. Rev. B 91, 205128 (2015). 43 C. J. Bradley and A. P. Cracknell, The Mathematical Theory of Symmetry in Solids (Clarendon Press, Oxford, 1972). 44 E.- G. Moon, C. Xu, Y. B. Kim, and L. Balents, Phys. Rev. Lett.111, 206401 (2013). 45 B. -J. Yang, E.- G. Moon, H. Isobe, N. Nagaosa, Nat. Phys. 10, 774 (2014). 46 S. -K. Jian and H. Yao, Phys. Rev. B 92, 045121 (2015). 24 47 H. -H. Lai, Phys. Rev. B 91, 235131 (2015). 48 S. T. Ramamurthy and T. L. Hughes, Phys. Rev. B 92, 085105 (2015). 49 R. -X. Zhang, J. A. Hutasoit, Y. Sun, B. Yan, C. Xu, C.-X. Liu, arXiv:1503.00358 (unpublished). 50 A. Kitaev, AIP Conf. Proc. 1134, 22 (2009). 51 T. Morimoto and A. Furusaki, Phys. Rev. B 88, 125129 (2013). 52 S. M. Young and C. L. Kane, arXiv:1504.07977. 53 Y. -M. Lu and D. -H. Lee, arXiv:1403.5558.
1501.07189	5	1501	2015-09-22T03:03:28	Manipulation of a single magnetic atom using polarized single electron transport in a double quantum dot	[ "cond-mat.mes-hall" ]	We consider theoretically a magnetic impurity spin driven by polarized electrons tunneling through a double quantum dot system. Spin blockade effect and spin conservation in the system make the magnetic impurity sufficiently interact with each transferring electron. As a results, a single collected electron carries information about spin change of the magnetic impurity. The scheme may develop all electrical manipulation of magnetic atoms by means of single electrons, which is significant for the implementation of scalable logical gates in information processing systems.	cond-mat.mes-hall	cond-mat	a Manipulation of a single magnetic atom using polarized single electron transport in a double quantum dot Wenxi Lai and Wen Yang Beijing Computational Science Research Center, Beijing 100094, China We consider theoretically a magnetic impurity spin driven by polarized electrons tunneling through a double quantum dot system. Spin blockade effect and spin conservation in the system make the magnetic impurity sufficiently interact with each transferring electron. As a results, a single collected electron carries information about spin change of the magnetic impurity. The scheme may develop all electrical manipulation of magnetic atoms by means of single electrons, which is significant for the implementation of scalable logical gates in information processing systems. PACS numbers: 76.30.Da, 72.25.Pn, 42.50.Dv, 73.23.Hk I. Introduction Magnetic atoms are critical spin systems which have potential applications in data storage and quantum in- formation processing.1 -- 3 In particular, using electrons to manipulate magnetic atoms is a natural step towards the implementation of scalable memory units for future inte- grated circuits. In dilute II-VI semiconductor quantum dots (QD), interaction between a manganese (Mn) atom and a carrier can be effectively described with the sp − d exchange interaction.4 -- 6 Based on the impurity-carrier coupling, electrical control of the single magnetic atom is feasible by injecting different charges in the magnetic atom doped QD.6 As shown both in experiment2 and theory,7 electrons can directly tunnel through a Mn atom by taking its spin states. The ground state and excited states of the Mn atom can be identified with the current, since they sup- port different conductance. However, in the tunneling from a STM tip to Cu2N surface through an individual Mn atom, the Mn spin spontaneous relaxation is more frequent than excitation by tunneling electrons.2 Com- pared with the case in high dimensional bulk material, the life time of the Mn spin is much longer in a QD.8 Therefore, coherent electrical manipulation of the Mn spin is possible in low dimensional nanostructure. In a QD doped with a single Mn atom, charge and conduc- tance of the single electron tunneling can be related to the spin state of the Mn atom.9 Whereas, it is hard to exactly connect a quantum state of the magnetic atom to a single electron in the above QD system, which should be a problem required to be solved for future quantum information processing. In this paper, we propose a scheme for all electrical manipulation of the magnetic impurity spin, scaling the number of driving electrons down to one. To this end, we consider two inter-coupled semiconductor QDs, in which one QD contains a single magnetic impurity and couples to a spin polarized electron source. The other QD is lo- calized in a homogeneous magnetic field and connected to a normal conductance, playing a role of a spin filter. In previous study,2,6,7,9 electrons transport through the magnetic atom without any qualification to their spins, as a result, interaction between the magnetic atom and electrons are very weak. In contrast, the QD spin filter induces spin dependent tunneling, which makes sure each electron would be completely flipped by the magnetic atom before it passes through the filter. Therefore, each collected electron can be correlated with the change of spin state in the magnetic atom. There are several facts that are very beneficial for the realization of our scheme. First, coupled two QDs can be fabricated, doping a single magnetic ion in one of them.10 Second, spin life time of an electron in II-VI semiconductor quantum dots can be long enough. Electron relaxation time in a similar system was reported to be 50 ns in a previous work.11 In a QD imbedded a magnetic atom, a longer life time of around 1 µs was predicted.12 Third, Magnetic atom such as Mn impurity with relaxation time from 1 µs to 0.4 ms was observed in experiment.10,13 The time is longer than driv- ing time of the magnetic atom, which will be shown later in the present work. Forth, the technique for real time detection of single electron tunneling has been well de- veloped recently.14 -- 17 Fifth, polarized electron current is available from several sources, for instance, QD spin split- ter under the premise of local magnetic field,18,19 ferro- magnetic leads,20,21 graphene or carbon atom wires.22,23 Recently, nuclear spins of donor atoms such as phospho- rus and 29Si have been coherently controlled and read using bounded electrons in these donors and the spin to charge conversion. Electromagnetic field was applied to initialize these nuclear spins.24 -- 26 Comparing with these early works, the main advance in our present protocol is that the magnetic atom would be manipulated all elec- trically. In other words, a magnetic atom can be con- trolled only using a single electron transistor without the application of any electromagnetic field. Even the initial- ization of the impurity spin can be progressed using the single electron tunneling in principle. II. MODEL AND THEORY Our model is illustrated in Fig. 1. The two QDs are denoted by dot 1 and dot 2 with ground orbital levels ε1 and ε2, respectively. Both electron polarization in the left lead and the external magnetic field ~B that applied on the dot 2 are assumed to be parallel to the QD growth direction z. Electrons injected from the left lead into the 2 of the magnetic atom. In this case, due to spin conserva- tion, if spin of the magnetic atom is changed from h Mzi to h Mzi + n↓, then the number of electrons detected in the right lead should equal to n↓ and their spins are down polarized. The magnetic atom can also be driven reversely, which is presented in Figs. 1(e)-(h). Since the magnetic field is turned to be in z direction here, the spin filter only allows spin up electrons tunnel through the dot 2. The input electrons are required to be in spin down state and out- put electrons are expected to be in spin up state. Each transported electron contributes to the spin of the mag- netic atom with the value −. It yields the impurity magnetization orient in the reversal way, which means spin of the magnetic atom would be changed from h Mzi to h Mzi − n↑ with collection of n↑ spin up electrons. To give a quantitative description to the model, we use the following Hamiltonian5,27 H = Hd1 + Hd2 + Hd12 + Hlead + Htun, (1) where the Hamiltonian for dot 1 is Hd1 = ε1n1 + U1n1↑n1↓ − je ~M · ~S1, and the Hamiltonian for dot 2 is Hd2 = ε2n2 + U2n2↑n2↓ + g∗µB ~B · ~S2. The inter-dot tunneling Hamiltonian reads Hd12 = Ω(n12 + n† 12) + U n1n2. (2) (3) (4) i↑ci↑, ni↓ = c† 1↑c2↑, n12↓ = c† Here, ni = ni↑ + ni↓, ni↑ = c† i↓ci↓, n12 = n12↑ + n12↓, n12↑ = c† 1↓c2↓. ci↑ (ci↓) is the annihilation operator of spin up (down) electron in dot i (i = 1, 2). S1 and S2 represent electron spin in the dot 1 and the dot 2, respectively. U is the inter- dot Coulomb potential. The exchange coupling strength between the electron in dot 1 and the magnetic atom is given by je = Jψ0(rM )2 with exchange integral J and the electron ground state wave function ψ0 at the magnetic impurity position rM . The left and right electronic leads are described by the free electron baths with the Hamiltonian Hlead = Xk,σ;α=L,R ǫαkc† αkσ cαkσ. (5) The dot 1 is coupled to the left leads and the dot 2 is coupled to the right lead by Htun = Xk,σ VLc† Lkσc1σ + Xk,σ VRc† Rkσ c2σ + H.c., (6) with the left and right tunneling amplitudes VL and VR, respectively. FIG. 1: (Color on line) Schematic illustration of the principle in our model. In (a) - (d) the applied magnetic field is in down direction and up polarized electrons are injected into the double QD. In (e)-(h), the external magnetic field is changed to be up direction and down polarized electrons are injected. dot 1 are coupled to the magnetic atom by the ferro- magnetic Heisenberg type spin exchange interaction. We describe spin of the magnetic atom with mean value of the spin along z direction h Mzi, where M is the mag- netic atom spin operator and the bracket indicates aver- age over quantum states of the system. Bias voltage and the magnetic field is tuned that just the lowest levels of the two QDs fall within the bias window µL > ε1, ε↓ (or ε↑)> µR, where the indexes ↓, ↑ indicate electron states with spin up and down, respectively. In addition, the intra-dot Coulomb blockade energies U1, U2 correspond- ing to the dot 1 and the dot 2 are assumed to be much larger than other energy scales, which yields only single electron occupation is involved in either of the QDs. It remarkably simplifies our model and calculation. In Figs. 1(a)-(d), the magnetic field in this config- uration is applied along the −z direction with value ~B = (0, 0, −B). It means, in dot 2, the ground state level is ε↓ = ε2 − g∗µBB/2 and the first excited level is ε↑ = ε2 + g∗µBB/2, where g∗ is the Lande g-factor of electron in the QD, µB is the Bohr magneton. The sys- tem requires spin up electrons that injected from the left lead into the dot 1. We take energy levels that satisfy ε↑ > µL and ε↑ − ε1 ≫ Ω, where Ω is inter-dot cou- pling strength. This energy structure forms a spin condi- tioned repulsive potential which ensures that the spin up electron is forbidden to enter the dot 2 until its spin is flipped to be upside down due to its coupling to the mag- netic atom. As soon as the electron spin changes to be down, it would be allowed to pass through the dot 2 and collected in the right lead. At the same time, the spin of magnetic atom would change from h Mzi to h Mzi + . If a spin down electron is injected from the left lead, it directly transports through the double dots without any change in the spin of the magnetic atom. However, if the left lead is a fully up polarized electron source, each passed electron would be connected to the spin change Time evolution of electron transport through the dou- ble QD system is described by a quantum master equa- tion which is derived based on the Hamiltonian (1) and the Liouville-von Neumann equation in the Born-Markov approximation. Since the transport in our system is a process of single electron sequential tunneling through QDs and works in the weak tunneling regime, the mas- ter equation is an effective approach to describe our model.28 -- 30 Equation of motion is given in terms of the reduced density matrix ρ of the system, ∂ ∂t ρ = 1 i [ Hd1 + Hd2 + Hd12, ρ] + LL ρ + LR ρ. (7) The Liouville super-operators, LL and LR, acting on the density matrix ρ describe tunneling on the left and right side of the double dots, respectively. They are writ- ten as 3 and LL ρ = 1 2 Xσ LR ρ = 1 2 Xσ L( fL,σ(c† Γσ 1σ ρc1σ − c1σc† 1σ ρ) + (1 − fL,σ)(c1σ ρc† 1σ − c† 1σc1σ ρ) + H.c.), R( fR,σ(c† Γσ 2σ ρc2σ − c2σc† 2σ ρ) + (1 − fR,σ)(c2σ ρc† 2σ − c† 2σc2σ ρ) + H.c.), (8) (9) where the Fermi distribution function in the left lead is fL,σ = (exp[(ε1 + U1n1¯σ − µL)/kBT ] + 1)−1 and in the right lead is fR,σ, = (exp[(εσ + U2n2¯σ − µR)/kBT ] + 1)−1, depending on the charging energy Ui conditioned by the occupation, ni¯σ = c† 1¯σc1¯σ, of electron with spin ¯σ. Here, the spin index ¯σ is defined that if σ =↑ (↓), then ¯σ =↓ (↑). The tunneling rates are spin dependent α = Γα(1 + Pα)/2 and Γ↓ given by Γ↑ α = Γα(1 − Pα)/2 with the current polarization Pα = (I ↑ α − I ↓ α). Here, I σ α is the current with pure spin σ on the side of α (α = L, R). The bare tunneling rates can be expressed as Γα = 2π tα2 Nα(ǫ) with the density of states Nα(ǫ) of electrons at energy ǫ. α)/(I ↑ α + I ↓ III. MANIPULATION OF A MAGNETIC ATOM WITH SPIN M=5/2 The Hilbert space of the double QD system is gen- erated by the basic vectors ii1mi1ji2, where ii1 and ji2 represent electronic state in dot 1 and dot 2, re- spectively. Here, i, j = 0 denote the empty state, i, j =↓, ↑ indicate occupation states of single electron with spin down and spin up, respectively. mi1 is eigen- state of the impurity spin operator Mz with eigenvalues m = −M, −m + 1, ..., M . First, we consider a typical single magnetic atom which displays a spin of M = 5/2 in the QD, such as Fe or Mn. These magnetic atoms have six quan- tized spin states mi, with corresponding eigenvalues m = −5/2,−3/2,−1/2,1/2,3/2,5/2.4 In Fig.2(a), initial state ψM (0)i of the magnetic atom is set to be any of the six quantized states. In all cases the magnetic atom is driven to the final state 5/2i. In average value of the Mn spin hMzi = T r[ Mz ρ] trace is taken over the basic vectors (a) 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 1.0 0.8 0.6 0.4 0.2 0.0 0 P-5/2 P-3/2 P-1/2 P1/2 P3/2 P5/2 10 20 30 40 50 60 70 t / 0 10 20 30 40 50 60 70 80 t / (b) 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 0 P-5/2 P-3/2 P-1/2 P1/2 P3/2 P5/2 30 40 50 1.0 0.8 0.6 0.4 0.2 0.0 0 10 20 t / 10 20 30 40 50 60 70 80 t / FIG. 2: (Color on line) Time evolution of the single impurity magnetization as a function of time. (a) The applied external magnetic field is ~B = (0, 0, −B). (b) The applied external field is ~B = (0, 0, B). The Parameters are µL = 75Γ, µR = −75Γ, ε1 = 0, ε2 = 62.5Γ, g ∗µBB = 135Γ, je = 3Γ, kBT = 12.5Γ, ΓL = ΓR = Γ, PR = 0, Ω = 5Γ, U = 10Γ. ii1mi1ji2. When the impurity spin changes to 5/2i, it becomes parallel to the injected electron spin and there will be no spin flip in the later time. Then the electron tunneling should be switched off and the spin down polar- ized current decreases to be zero as plotted in Fig. 3 (b). The figures also show that orientation time of the Mn spin is a few tens of ns when one takes a tunneling char- acteristic time of Γ−1 ∼ 1ns. It is comparable to the time scale of a previously reported optical control.13 Since each electron contributes to the spin change of the magnetic (a) n 5 4 3 2 1 0 (b) 0.25 0.20 0.15 0.10 0.05 0.00 I I 0 10 20 30 40 t / 50 60 70 80 0 10 20 40 50 60 30 t / FIG. 3: (Color on line) (a) The number of spin down electrons collected in the right side of the system. (b) Current for spin up and down electron s as a function of time. The parameters are the same as that in Fig. 2 atom with momentum , the magnetic atom initialized in the states ψM (0)i = − 5/2i, − 3/2i, − 1/2i, 1/2i, 3/2i, 5/2i leads to finite electrons collected in the right lead with definite numbers hn↓i = 5, 4, 3, 2, 1, 0, respec- tively. It can be seen in Fig. 3(a) for corresponding initial spin states. The collected electron number is calculated using the formula hnσi = 1 e Z ∞ 0 Iσ(t)dt, (10) where Iσ(t) is current on the right side of dot 2. The current is derived from the charge fluctuating in the two R)/e with the R. The reversal time evolution of dots d(hn1i + hn2i)/dt = Pσ=↑,↓(I σ replacement Iσ(t) = I σ the magnetization is shown in Fig. 2 (b). L − I σ IV. MANIPULATION OF A MAGNETIC ATOM WITH SPIN M=1/2 Now, we consider a magnetic ion with spin M = 1/2, such as Cu2+. This kind of impurity has particular mean- ing that its maximum magnetization difference is , from −/2 to /2 or by inverse. In this case, only a sin- gle electron is involved in the tunneling. As shown in Fig.4(a), when we set initial state of the magnetic atom to be − 1/2i, it is transformed into 1/2i. At the same time, the up polarized current is changed to be down po- larized current after interacting with the magnetic atom. The down polarized current is expected to be collected in the right side, which is characterized by a singe elec- tron with spin down as shown in Fig.4(b). To show the collected charge is really limited, average current as a function of time is plotted in the inset of Fig.4(b). The down polarized current sharply increases and then disap- pears slowly with a small fluctuation. The up polarized current is negligible weak. From the point of application, (a) 1.0 0.8 0.6 0.4 0.2 0.0 P-1/2 P1/2 (b) 1.0 0.8 0.6 0.4 0.2 0.0 4 n n I /e I /e 0.12 0.10 0.08 0.06 0.04 0.02 0.00 0 10 20 40 30 t / 50 60 0 10 20 30 t / 40 50 60 0 10 20 40 30 t / 50 60 FIG. 4: (Color on line) (a) Probabilities of the magnetic atom in its spin states − 1/2i and 1/2i. The magnetic atom is initially in the state − 1/2i. (b) The number of collected electrons with spin up and spin down corresponding to the case (a). The parameters are µL = 70Γ, µR = −70Γ, ε1 = 0, ε2 = 57.5Γ, g ∗µBB = 125Γ, je = 2Γ, kBT = 10Γ, ΓL = ΓR = Γ, PL = 1, PR = 0, Ω = 3Γ, U = 10Γ. it is a very good system for date storage, where only sin- gle electron of current and the two states of the magnetic atom is correlated. The small fluctuation in the current is occurred due to the back action from the system. Since charge transfer through the double dots requires electron spin flip, the amplitude of current is proportional to rate of electron spin change. The rate of spin change is determined by the strength of spin coupling between the magnetic atom and electrons. Therefore, we can deduce that the oscilla- tions observed in Fig.4 is contributed from the coherent coupling between the magnetic atom and an individual electron. Indeed, character of the small oscillation can be tuned by changing the spin coupling strength je. Ac- tually, the back action effect can also be seen in the for- mer situation for M = 5/2. However, considering six spin states of the magnetic atom are involved in the ex- change interaction and the larger couplings je and Ω are taken, the small oscillation during the evolution of the spin states and current is too smooth to be observed. The above results are described in a relatively ideal situation. At the end of this section, let us talk about some more practical cases. Fig. 5(a) reveals that increase of the inter-dot tunneling strength excites a spin up elec- tron from the dot 1 into the excited level ε↑ of the dot 2. It leads to the unexpected spin up electron leakage from the double dot system into the right lead. To restrain the electron leakage through the excited level ε↑, we suggest that one can take a relatively small inter-dot coupling Ω which is required to be much smaller than the decoupling ε↑ − ε1. The system is also sensitive to polarization of the right lead. In Fig. 5(b), it is illustrated that when the polar- ization is not pure, the number of electrons collected in the right lead is larger than the expected value. Since (a) = = = = = = n 5 4 3 2 1 0 n n 5 4 3 2 1 0 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 0 (b) 50 100 150 PL=1.0 PL=0.4 200 PL=0.8 PL=0.2 300 250 PL=0.6 PL=0.0 0 50 100 150 200 250 300 (c) kBT=10 kBT=30 kBT=50 kBT=20 kBT=40 kBT=60 0 10 20 30 40 t / 50 60 70 80 90 FIG. 5: (Color on line) (a) Number of spin up electrons col- lected in the right side of the double QD with different inter- dot coupling strength, kBT = 10Γ, PL = 1. (b) Number of spin down electrons for different spin polarization in the left lead, kBT = 10Γ, Ω = 3Γ. (c) Number of collected spin down electrons at different temperature, Ω = 3Γ, PL = 1. The rest parameters are the same as that in Fig. 4. non-purely polarized current contains electrons with dif- ferent spin orientations, thereby some electrons could not be blocked by dot 2. The negative number of electrons in Fig. 5(c) implies that electrons in the right lead have certain probability to flow into the QDs at the beginning of tunneling due to the thermal excitation, since the QDs is empty initially. After the system reaches steady state, the higher the tem- perature, the larger electron distribution would be in the QDs. The electron staying in the QDs comes from the leads. Therefore, net number of electrons collected in the right lead decreases when temperature increases. V. DISCUSSIONS In experiment, to realize electron transport through a magnetic atom doped II-VI semiconductor QD is still a challenge. There are some well developed experimental backgrounds, for instance, the spin dependent electron transport through a quantum well that containing dilute magnetic material,31 and resonant tunneling through a CdSe self-assembled QD with Mn ions.32 In addition, an attempt to orient a Mn spin using the charge that trans- ported from a neighboring QD has been successfully re- alized.10 To create a local magnetic field is a key technical prob- lem for the experimental implementation of our proposal. The left lead is a polarized electron source, so an exter- nal magnetic field may be required on it. In addition, the right QD needs a magnetic field to create a Zeeman splitting for an electron. At the same time, these mag- netic fields must have negligible affect to the left QD and 5 the right lead. Therefore, some local magnetic fields are necessary in the nanostructure. To this end, a proper quantity of external magnetic field can be exerted on the left lead for the generation of spin polarized electron source. As the QD is a zero dimensional nanoscale sys- tem, our suggestion is using a magnetic grain, such as Co grain,33,34 to create magnetic field on the right QD. To explore its effectiveness we estimate the field strength of a typical Co grain which consists of hundreds of atoms. We assume the Co grain is a uniformly magnetized sphere. Then, the field of the grain can be obtain from the for- mula B = 2 3 µ0M, µ0 is permeability of free space and M is magnetization of the grain.35 We take that average mag- netic moment per Co atom is 1.5µB. In real materials the local magnetic moment is always larger than this value.33 We take diameter of a Co atom is about 0.15nm. Then, magnetic field of the Co grain is estimated to be not less than B = 6.6T . By increasing size of the Co grain, more strong field can be obtained. In a CdTe semiconductor QD, the corresponding Zeeman splitting of an electron reaches g∗µBB = 0.64meV , where the g-factor is given by g∗ = −1.67 in this material.36 To guarantee the left QD and the right lead is not effectively influenced by the surrounding magnetic field, magnetic shielding may be applied here to protect the spreading field. One kind of the magnetic shielding materials is superconducting chip which expels magnetic field via the Meissner effect. An- other kind shielding material is certain high permeability metal alloy which, in contrast the superconductor, draws the field into themselves. Taking an example with the high permeability shielding with permeability µ, field in the shielded volume is 2(b3−a3)µ times smaller than the outside field.37 For the convenience of quantitative esti- mation, the shielding material here is assumed to be a spherical shell with inner radius a and outer radius b. When the permeability µ is large enough, good shield- ing of the field in the shielded area can be achieved. In practice, as the QD couples to an electronic reservoir and another dot, the shielding material may be not absolutely closed with a lower efficiency. 9b3µ0 In the Hamiltonian of our model, spin exchange inter- action between the two QDs is not considered. In fact, when each quantum dot contains one electron, the ex- change interaction does not play important role for the whole system. The reason is that the electron in dot 2 has definite spin direction. It has to gain large energy to change spin. However, it is hard for the electron in dot 1 to provide the large energy. Besides, the system is robust against the double occupancy in dot 1. Since cur- rent injected from the left lead is assumed to be pure spin polarized, two electrons with different spin states in dot 1 does not break the spin conservation as soon as dot 2 guarantees to output electrons with pure spin. Whereas, double occupancy in dot 2 induces spin leakage and, as a result, the correlation between electrons and the mag- netic atom becomes weak. Even though, considering the high Coulomb blockade effect in either of the QDs, double occupancy in any dot is negligible small in our model. 6 As an information processing system, its characteristic times are very significant. There are two kinds of critical times, one of them is manipulation time τs of the system, another is lowest life-time bound τi of these information carriers, such as the magnetic atom and electrons. The manipulation time of the system indicates a time range during which an electron is emitted from the left lead and then collected in the right lead, at the same time control of the magnetic atom is accomplished. It is clear that spin life times of the information carriers are required to be, at least, longer than the manipulation time of the system, i.e. τi > τs. As mentioned in the introduction, a single Mn spin relaxation time from 1 µs to 0.4 ms in CdTe QD10,13, and electron life time from 50 ns to 1 µs in II-VI semiconductor systems are reported. Even in a QD including certain charges, the Mn atom relaxes in a time scale of about 100 ns.38 As shown in section III and IV, manipulation time of the system is about 50 ns for a characteristic time of the electron tunneling, Γ−1 ∼ 1ns. Furthermore, there is still a space for re- ducing the manipulation time of the system. Actually, properly increasing the dot-reservoir coupling (electron tunneling rate) or inter-dot coupling would improve the rate of control process. In this case, a more lower bound of the spin life time is allowed. VI. CONCLUSIONS In the system, change of magnetic impurity spin is cor- related to spin state and number of single electrons that tunneling through the two QDs. Based on this principle, we give the following predictions: (i) Our model works as an electron source in which number of emitted elec- trons can be determined beforehand by setting an initial state of the magnetic atom or using a magnetic atom with certain spin M . In particular, a single electron emitter is available using a magnetic atom with spin 1/2. (ii) The number of polarized transferring electrons can be recorded in the spin state of the magnetic atoms. (iii) The change in spin state of the magnetic atom should be detected by counting the number of electrons that emitted from the double QDs. (iv) Spin state of the magnetic impurity can be controlled, in principle, by in- jecting suitable number of spin polarized electrons, and this controlling is reversal. Acknowledgments We appreciate the valuable discussion with Shuhui Zhang about the numerical calculation. 1 L. Thomas, F. Lionti, R. Ballou, D. Gatteschi, R. Sessoli, and B. Barbara, Nature 383, 145 (1996). 2 S. Loth, K. von Bergmann, M. Ternes, A. F. Otte, C. P. Lutz and A. J. Heinrich, Nature Phys. 6, 340 (2010). 3 R. Hanson and D. D. Awschalom, Nature 453, 1043 (2008). 4 L. Besombes, Y. L´eger, L. Maingault, D. Ferrand, H. Ma- riette, J. Cibert, Phys. Rev. Lett. 93, 207403 (2004). 5 D. E. Reiter, T. Kuhn, and V. M. Axt, Phys. Rev. B 85, 045308 (2012). 6 Y. L´eger, L. Besombes, J. Fern´andez-Rossier, L. Main- gault, and H. Mariette, Phys. Rev. Lett. 97, 107401 (2006). 7 F. Delgado, J. J. Palacios, and J. Fern´andez-Rossier, Phys. Rev. Lett. 104, 026601 (2010). 8 M. Scheibner, T. A. Kennedy, L. Worschech, A. Forchel, G. Bacher, T. Slobodskyy, G. Schmidt, and L. W. Molenkamp, Phys. Rev. B 73, 081308 (2006). 9 J. Fern´andez-Rossier and R. Aguado, Phys. Rev. Lett. 98, 106805 (2007). 10 M. Goryca, T. Kazimierczuk, M. Nawrocki, A. Golnik, J. A. Gaj, P. Kossacki, P. Wojnar, G. Karczewski, Phys. Rev. Lett. 103, 087401 (2009). 11 D. H. Feng, I. A. Akimov, and F. Henneberger, Phys. Rev. Lett. 99, 036604(2007). 12 M. D. Petrovi´c and N. Vukmirovi´c, Phys. Rev. B 85, 195311(2012). 13 C. Le Gall, L. Besombes, H. Boukari, R. Kolodka, J. Cib- ert, and H. Mariette, Phys. Rev. Lett. 102, 127402(2009). 14 J. Bylander, T. Duty, and P. Delsing, Nature 434, 361 (2005). 15 W. Lu, Z. Ji, L. Pfeiffer, K. W. West, and A. J. Rimberg, Nature 423, 422 (2003). 16 S. Gustavsson, R. Leturcq, M. Studer, I. Shorubalko, T. Ihn, K. Ensslin, D.C. Driscoll, and A.C. Gossard, Surface Science Reports 64, 191 (2009). 17 J. P. Pekola, O.-P. Saira, V. F. Maisi, A. Kemppinen, M. Mottonen, Y. A. Pashkin, D. V. Averin, Rev. Mod. Phys. 85, 1421 (2013). 18 P. Recher, E. V. Sukhorukov, and D. Loss, Phys. Rev. Lett. 85, 1962 (2000). 19 R. Hanson, L. M. K. Vandersypen, L. H. W. van Beveren, J. M. Elzerman, I. T. Vink, and L. P. Kouwenhoven, Phys. Rev. B 70, 241304 (2004). 20 P. Hyde, L. Bai, D. M. J. Kumar, B. W. Southern, C.-M. Hu, S. Y. Huang, B. F. Miao, and C. L. Chien, Phys. Rev. B 89, 180404 (2014). 21 A. Dirnaichner, M. Grifoni, A. Prufling, D. Steininger, A. K. Huttel, and C. Strunk, Phys. Rev. B 91, 195402 (2015). 22 Z. Y. Li, W. Sheng, Z. Y. Ning, Z. H. Zhang, Z. Q. Yang, and H. Guo, Phys. Rev. B 80, 115429 (2009). 23 V. M. Karpan, G. Giovannetti, P. A. Khomyakov, M. Ta- lanana, A. A. Starikov, M. Zwierzycki, J. van den Brink, G. Brocks, and P. J. Kelly, Phys. Rev. Lett. 99, 176602 (2007). 24 J. J. Pla, K Y. Tan, J. P. Dehollain, W. H. Lim, J. J. L. Morton, D. N. Jamieson, A. S. Dzurak and A. Morello, Nature 489, 541 (2012). 25 J. J. Pla, K. Y. Tan, J. P. Dehollain, W. H. Lim, J. J. L. Morton, F. A. Zwanenburg, D. N. Jamieson, A. S. Dzurak, A. Morello, Nature 496, 334 (2013). 26 J. J. Pla, F. A. Mohiyaddin, K. Y. Tan, J. P. Dehollain, R. Rahman, G. Klimeck, D. N. Jamieson, A. S. Dzurak, and A. Morello, Phys. Rev. Lett. 113, 246801 (2014). 27 D. Becker, S. Weiss, M. Thorwart, and D. Pfannkuche, 7 New J. Phys. 14, 073049 (2012). 28 S. A. Gurvitz and Y. S. Prager, Phys. Rev. B 53, 15932 (1996). 29 T. Novotn´y, A. Donarini, and A.-P. Jauho, Phys. Rev. Lett. 90, 256801 (2003). (2000). 34 L. V. Dzemiantsova, M. Hortamani, C. Hanneken, A. Ku- betzka, K. von Bergmann, and R. Wiesendanger, Phys. Rev. B 86, 094427 (2012). 35 D. J. Griffiths, Introduction to electrodynamics, 3rd ed, 30 S. Braig and P. W. Brouwer, Phys. Rev. B 71, 195324 Prentice Hall (1999). (2005). 31 A. Slobodskyy, C. Gould, T. Slobodskyy, C. R. Becker, G. Schmidt, and L.W. Molenkamp, Phys. Rev. Lett. 90, 246601 (2003). 32 C. Gould, A. Slobodskyy, D. Supp, T. Slobodskyy, P. Grabs, P. Hawrylak, F. Qu, G. Schmidt, and L.W. Molenkamp, Phys. Rev. Lett. 97, 017202 (2006). 33 R. N. Nogueira and H. M. Petrilli, Phys. Rev. B 63, 012405 36 F. Qu and P. Hawrylak, Phys. Rev. Lett. 95, 217206 (2005). 37 J. D. Jackson, Classical electrodynamics, 3rd ed, John Wi- ley and Sons (1999). 38 C. Le Gall, R. S. Kolodka, C. L. Cao, H. Boukari, H. Ma- riette, J. Fern´andez-Rossier, and L. Besombes, Phys. Rev. B 81, 245315 (2010).
1311.4295	2	1311	2015-06-17T09:23:27	Manipulation of the nuclear spin ensemble in a quantum dot with chirped magnetic resonance pulses	[ "cond-mat.mes-hall" ]	The nuclear spins in nanostructured semiconductors play a central role in quantum applications. The nuclear spins represent a useful resource for generating local magnetic fields but nuclear spin noise represents a major source of dephasing for spin qubits. Controlling the nuclear spins enhances the resource while suppressing the noise. NMR techniques are challenging: the group III and V isotopes have large spins with widely different gyromagnetic ratios; in strained material there are large atom-dependent quadrupole shifts; and nanoscale NMR is hard to detect. We report NMR on 100,000 nuclear spins of a quantum dot using chirped radiofrequency pulses. Following polarization, we demonstrate a reversal of the nuclear spin. We can flip the nuclear spin back and forth a hundred times.We demonstrate that chirped NMR is a powerful way of determining the chemical composition, the initial nuclear spin temperatures and quadrupole frequency distributions for all the main isotopes. The key observation is a plateau in the NMR signal as a function of sweep rate: we achieve inversion at the first quantum transition for all isotopes simultaneously. These experiments represent a generic technique for manipulating nanoscale inhomogeneous nuclear spin ensembles and open the way to probe the coherence of such mesoscopic systems.	cond-mat.mes-hall	cond-mat	Manipulation of the nuclear spin ensemble in a quantum dot with chirped magnetic resonance pulses Mathieu Munsch,1 Gunter Wust,1 Andreas V. Kuhlmann,1 Fei Xue,1 Arne Ludwig,1, 2 Dirk Reuter,2, 3 Andreas D. Wieck,2 Martino Poggio,1 and Richard J. Warburton1 1Department of Physics, University of Basel, Klingelbergstrasse 82, CH-4056 Basel, Switzerland 2Lehrstuhl fur Angewandte Festkorperphysik, Ruhr-Universitat Bochum, D-44780 Bochum, Germany 3Department Physik, Universitat Paderborn, Warburger Strasse 100, D-33098 Paderborn, Germany (Dated: August 20, 2018) The nuclear spins in nano-structured semiconductors play a central role in quantum applications1 -- 4. The nuclear spins represent a useful resource for generating local magnetic5 fields but nuclear spin noise represents a major source of dephasing for spin qubits2,3. Controlling the nu- clear spins enhances the resource while suppressing the noise. Nuclear magnetic resonance (NMR) techniques are challenging: the group-III and group-V isotopes have large spins with widely dif- ferent gyromagnetic-ratios; in strained material there are large atom-dependent quadrupole-shifts6; nano-scale NMR is hard to detect7,8. We report NMR on 100, 000 nuclear spins of a quantum dot using chirped radio-frequency pulses. Following polarization, we demonstrate a reversal of the nuclear spin. We can flip the nuclear spin back-and-forth a hundred times. We demonstrate that chirped-NMR is a powerful way of determining the chemical composition, the initial nuclear spin temperatures and quadrupole frequency distributions for all the main isotopes. The key observa- tion is a plateau in the NMR signal as a function of sweep-rate: we achieve inversion at the first quantum transition for all isotopes simultaneously. These experiments represent a generic technique for manipulating nano-scale inhomogeneous nuclear spin ensembles and open the way to probe the coherence of such mesoscopic systems. NMR signals can be boosted by polarizing the nuclei. This is particularly beneficial on the nano-scale where NMR signals are invariably small and hard to detect. The nuclear spins in a self-assembled quantum dot can be polarized optically by exploiting the hyperfine interaction with an electron spin3,5. Extremely long-lived polarizations4,9 -- 11 up to about 50% have been achieved. The nuclear spin polarization results in a shift of the optical resonance, the Overhauser shift, facilitating its sensitive detection5. These features have enabled the observation of isotope-selective NMR of the nuclear spins associated with strain-free GaAs quantum dots12,13. Self-assembled quantum dots, attractive for single photon generation and optically- controlled spin qubits2, have highly inhomogeneous nuclear spins5,14 -- 16. Additional side peaks appear in 2 the NMR spectra, a consequence of a strain-dependent quadrupole interaction, along with a distribution of chemical shifts6. Manipulating the nuclear spin ensemble of a single quantum dot is challenging yet important: projection of the nuclear spins into a specific state boosts the single electron spin dephasing time4; developing techniques to probe nano-sized ensembles of highly inhomogeneous nuclear spins has impact also for semiconductor nanowires17 and nanocrystals. Here we use chirped NMR pulses. The main concept is that by sweeping over a large frequency range, 2 nucleus, a 2-level system, the Hamiltonian the pulse addresses each nuclear spin at some point. For a spin- 1 in the rotating frame is, H = h∆ν(t)Iz + 1 2 hγBxIx (1) where h is the Planck constant, I the nuclear spin, γ the gyromagnetic ratio of the nuclear isotope (in frequency units) and ∆ν(t) is the time-dependent detuning between the radio frequency (RF) excitation and the Larmor frequency νL = γBz. The coupling between the RF magnetic field Bx and the spin, the second term in the Hamiltonian, leads to an avoided crossing in the eigen-energies with splitting hνRF , Fig. 1a , where νRF = γBx. On traversing the avoided crossing from large and negative ∆ν to large and positive ∆ν with a single pulse (N = 1) at sweep rate α, the probability that the final state is ↑(cid:105) for initial state ↑(cid:105), is PLZ = exp(−π2ν2 RF /α), (2) the Landau-Zener result18. In the sudden regime when PLZ (cid:39) 1, the system "tunnels" through the avoided crossing and ↑(cid:105) → ↑(cid:105), ↓(cid:105) → ↓(cid:105). Alternatively, in the limit when PLZ (cid:28) 1, the states are swapped ↑(cid:105) → ↓(cid:105), ↓(cid:105) → ↑(cid:105): this is adiabatic passage, Fig. 1a. We attempt to apply these concepts to a single nano-scale nuclear spin ensemble. The challenges are, first, each nuclear spin is more complex than a two-level system; and second, there is an inhomogeneous dis- tribution of 105 nuclear spins. Initialization and detection of the nuclear spin polarization of a single quan- tum dot is carried out optically with exquisite spectral resolution provided by resonant laser spectroscopy, representing a sensitivity to ∼ 1, 000 spins. The quantum dots for these experiments are gate-controlled InxGa1−xAs quantum dots11, Fig. 1c. The bias voltage controls both the occupation of the quantum dot (here empty) and the exact optical transition frequency via the Stark effect. Key to reaching the adiabatic limit PLZ (cid:28) 1 is the generation of RF fields with high amplitude. We use an on-chip, low-impedance, high bandwidth microwire11,19, fabricated directly above the gate: large oscillating currents in the microwire generate oscillating magnetic fields (Bx (cid:39) 5 mT11); the small impedance of the microwire enables fast pulsing. An aperture in the microwire allows optical access to the quantum dots directly underneath, Fig. 3 1b,d. The quantum dot optical resonance (X0) is driven with a coherent laser with resonance fluorescence detection20,21, the read-out after one RF pulse providing the initialization for the next, Figs. 1e and 2c. A resonance fluorescence spectrum of the quantum dot at zero applied magnetic field, Bz = 0 T, is shown in Fig. 2a: the two lines, split by the fine-structure, have linewidths of 1.2 µeV, close to the transform limit of 0.9 µeV22. At Bz ≥ 0.5 T, on sweeping through the optical resonance, the nuclear spins adjust their polarization to maintain an optical resonance of the quantum dot with the laser, the "dragging" effect23,24: the Overhauser shift OHS equals the laser detuning δL. Dragging represents a way of generating large bi-directional nuclear spin polarizations23. An example is shown in Fig. 2b: starting with the nuclei in a depolarized state11, the optical resonance is "dragged" to δL = −41 µeV. The nuclear spin polarization decays extremely slowly (timescale days for an empty quantum dot4,9 -- 11), resulting in optical memory effects. A sequence of optical sweeps is shown in Fig. 2b: the rise point of each scan is related to the polarization set by the previous scan whereas the end of the plateau sets the new polarization state. For a given laser sweep direction, the change in width of the dragging "plateau" following an NMR pulse is used to measure the change in the Overhauser field, ∆OHS in Fig. 2c. Manipulation of the nuclear spin ensemble is demonstrated in Fig. 2c. The nuclear spin polarization along z, (cid:104)Iz(cid:105), is initialized with a sweep from positive to negative δL. With the laser off, a chirped NMR pulse is applied, ν = ν1 → ν2. The laser is then turned back on and the sweep from positive to negative δL repeated. The optical signal now appears not at negative δL but at positive δL, unambiguous evidence that the RF pulse inverts the nuclear spin polarization. In this particular case, following optical polarization, (cid:104)Iz(cid:105) /I max z (cid:39) −13%11. This interpretation is backed up by applying not one but a sequence of (phase-matched) chirped pulses, ν1 → ν2 → ν1 → ν2. . . . As a function of pulse number N, (cid:104)Iz(cid:105) oscillates from positive to negative, evidence of close-to-adiabatic manipulation of (cid:104)Iz(cid:105). We can invert-restore the nuclear spin polarization ∼ 100 times before the signal is lost, Fig. 2d. z (cid:39) +32%, and after one NMR pulse, (cid:104)Iz(cid:105) /I max We explore the dependence on sweep rate α on tuning from low ν1 to high ν2 such that all nuclear spins are addressed. The signal increases with decreasing sweep rate, Fig. 3. Significantly, there is an exponential increase followed by a plateau and then another exponential increase. The step-wise transition from the sudden to the adiabatic regime is a consequence of a hierarchy of avoided crossings in the energy level structure. It arises from a quadrupole interaction of the nuclear spin with a local electric field gradient resulting in an additional term in the Hamiltonian, HQ = 1 6 hνQ (cid:104) z − I(I + 1) 3I 2 (cid:105) . (3) where hνQ is the strength of the quadrupole field11. Fig. 1a shows the eigen-energies for I = 3 2, both for νQ = 0 and for νQ (cid:29) νRF . When νQ (cid:54)= 0, a hierarchy of avoided crossings appears, large for the first 4 quantum transitions (bare states separated by ∆m = 1); intermediate at the second quantum transitions (∆m = 2); and small at the third quantum transition (∆m = 3). A similar but more complex hierarchy also arises in the In (I = 9 2) eigen-energies. Given the exponential dependence of PLZ on the energy separation at the avoided crossing, this means that the different quantum transitions satisfy the adiabaticity condition at quite different sweep rates25 -- 27. At the plateau in Fig. 3, the sweep is adiabatic for the first quantum transitions (PLZ (cid:28) 1) whereas the others are still in the sudden regime (PLZ (cid:39) 1). At first sight, it is surprising that the step signifying adiabaticity at the first quantum transitions survives the ensemble averaging. The explanation is to be found in the scaling of the energies at the avoided crossings, hνeff. In the limit νQ (cid:29) νRF , νeff ∝ νRF (νRF /νQ)∆m−1 for all I11,25 -- 27. This means that for ∆m = 1, νeff does not depend on νQ (to first order), suppressing the sensitivity of the adiabaticity criterion to the quadrupole interaction. The plateau in the sweep rate dependence is the key observation that allows both the indium concentra- tion x and the initial nuclear spin temperature T to be determined. The point is that the signal at the plateau, ∆OHS = 28.8 µeV, and the initial Overhauser shift, OHS = 27.0 µeV, are determined solely by x, T and the known nuclear parameters (nuclear spins, hyperfine coupling constants and abundances of 75As, 115In, 69Ga and 71Ga)11. We find x = (20.2 ± 5.7)% and T = (8.2 ± 0.8) mK. The composition x represents the indium concentration over the extent of the electron wave function; the temperature, much lower than the bath temperature of 4.2 K, interprets the dynamic nuclear spin polarization as a laser cooling phenomenon. Spectroscopic identification of the isotopes is presented in Fig. 4 where the NMR pulse is chirped from a fixed ν1 to a variable ν2 using a slow and constant sweep rate. The NMR signal ∆OHS increases step-wise around 44 MHz. This arises when ν2 goes above the central NMR frequency of a particular isotope, in this case 75As. Another clear step arises at 79 MHz, the 71Ga resonance. Around the central transition, the single spin satellite steps11 are broadened through atom-dependent quadrupole couplings. This is particularly visible in the In contribution because of the large number of satellites. This curve enables us to determine the average quadrupole frequency (cid:104)νQ(cid:105) and an approximate distribution p(νQ) for all the main isotopes, 75As, 115In, 69Ga and 71Ga. For a specific I, νQ and νRF , we occupy the initial nuclear states according to the known T , and integrate the Schrodinger equation numerically to determine (cid:104)Iz(cid:105) after a single NMR pulse, converting (cid:104)Iz(cid:105) to ∆OHS with the appropriate hyperfine coefficient. We find that the ν2-dependence is a strong function of both (cid:104)νQ(cid:105) and p(νQ)11 and is therefore ideal to determine them. The 75As and 71Ga are well isolated as a function of ν2 and in both cases, (cid:104)νQ(cid:105) and p(νQ) are readily determined by comparing the experimental results to the theory. The 69Ga ν2-dependence can be predicted from the 71Ga ν2-dependence simply by the known abundances and quadrupole moments11. The remaining signal at intermediate ν2 arises mostly from 115In 5 allowing us to determine the 115In quadrupole parameters. Fig. 4 shows that, first, we achieve an excellent description of the experimental results; and second, the signals from the four isotopes 75As, 115In, 69Ga and 71Ga overlap little facilitating the determination of each quadrupole distribution. We return to the sweep rate dependence. We calculate the α-dependence, adding the results from each isotope with x, T , (cid:104)νQ(cid:105) as input parameters. (Bx is adjusted within its error window to ensure that the plateau occurs at the correct α.) The same set of parameters describes both the ν2- and α-dependences. Fig. 3 shows the contribution from each isotope. 115In has the largest hνeff (on account of its large spin, I = 9 2) and inversion at the first quantum transition is achieved first of all, closely followed by inversion at the first 2 nuclei. At the smallest α, inversion at the second quantum transition is achieved for most of the In nuclei (and some of the 71Ga nuclei) but for most of the 75As and 69Ga nuclei, quantum transition for the I = 3 inversion at the first quantum transition is complete but inversion at the second quantum transition is not yet achieved. This explains the second change in gradient at the smallest α in the experiment. The combination of the ν2 and the α-dependences allows in principle an initial nuclear spin temperature to be determined for each isotope. In practice, these temperatures are not significantly different to within the random error11 and we take a common temperature for simplicity. The overall conclusion is that frequency-swept NMR enables the determination of all key parameters of the nuclear spins even at the single quantum dot level: the chemical composition, the effective temperatures and the quadrupole frequency distribution of each isotope. As an outlook, we note that a sweep adiabatic for ∆m = 1 but sudden for ∆m = 2 can be used to pro- duce highly non-thermal distributions of the spin states, boosting the NMR signal of the central transitions. Also, at an intermediate sweep rate, a superposition of the spin states is created with a chirped NMR pulse, and back-and-forth frequency sweeps result in quantum interferences, the Stuckelberg oscillations18,28 -- 31. This experiment represents the ideal springboard to explore quantum coherence in a complex nuclear spin ensemble using multiple chirped pulses. 1 Ribeiro, H. & Burkard, G. Nuclear spins keep coming back. Nature Materials 12, 469 -- 471 (2013). 2 Warburton, R. J. Single spins in self-assembled quantum dots. Nature Materials 12, 483 -- 493 (2013). 3 Chekhovich, E. A. et al. Nuclear spin effects in semiconductor quantum dots. Nature Materials 12, 494 -- 504 (2013). 4 Greilich, A. et al. Nuclei-induced frequency focusing of electron spin coherence. Science 317, 1896 -- 1899 (2007). 5 Urbaszek, B. et al. Nuclear spin physics in quantum dots: An optical investigation. Rev. Mod. Phys. 85, 79 -- 133 (2013). 6 6 Chekhovich, E. et al. Structural analysis of strained quantum dots using nuclear magnetic resonance. Nature Nanotechnology 7, 646 -- 650 (2012). 7 Staudacher, T. et al. Nuclear magnetic resonance spectroscopy on a (5-nanometer)3 sample volume. Science 339, 561 -- 563 (2013). 8 Mamin, H. J. et al. Nanoscale nuclear magnetic resonance with a nitrogen-vacancy spin sensor. Science 339, 557 -- 560 (2013). 9 Maletinsky, P., Kroner, M. & Imamoglu, A. Breakdown of the nuclear-spin-temperature approach in quantum-dot demagnetization experiments. Nature Physics 5, 407 -- 411 (2009). 10 Latta, C., Srivastava, A. & Imamoglu, A. Hyperfine interaction-dominated dynamics of nuclear spins in self- assembled ingaas quantum dots. Phys. Rev. Lett. 107, 167401 (2011). 11 Supplementary Information. 12 Gammon, D. et al. Nuclear spectroscopy in single quantum dots: Nanoscopic raman scattering and nuclear magnetic resonance. Science 277, 85 -- 88 (1997). 13 Makhonin, M. et al. Fast control of nuclear spin polarization in an optically pumped single quantum dot. Nature Materials 10, 844 -- 848 (2011). 14 Flisinski, K. et al. Optically detected magnetic resonance at the quadrupole-split nuclear states in (in,ga)as/gaas quantum dots. Phys. Rev. B 82, 081308 (2010). 15 Cherbunin, R. V. et al. Resonant nuclear spin pumping in (in,ga)as quantum dots. Phys. Rev. B 84, 041304 (2011). 16 Bulutay, C. Quadrupolar spectra of nuclear spins in strained ingaas quantum dots. Phys. Rev. B 85, 115313 (2012). 17 Peddibhotla, P. et al. Harnessing nuclear spin polarization fluctuations in a semiconductor nanowire. Nature Physics 9, 631 (2013). 18 Shevchenko, S., Ashhab, S. & Nori, F. Landau-Zener-Stuckelberg interferometry. Physics Reports 492, 1 -- 30 (2010). 19 Poggio, M., Degen, C. L., Rettner, C., Mamin, H. & Rugar, D. Nuclear magnetic resonance force microscopy with a microwire rf source. Applied Physics Letters 90, 263111 (2007). 20 Kuhlmann, A. V. et al. Charge noise and spin noise in a semiconductor quantum device. Nature Physics 9, 570 -- 575 (2013). 21 Kuhlmann, A. V. et al. A dark-field microscope for background-free detection of resonance fluorescence from sin- gle semiconductor quantum dots operating in a set-and-forget mode. Review of Scientific Instruments 84, 073905 (2013). 22 Kuhlmann, A. V. et al. Linewidth of single photons from a single quantum dot: key role of nuclear spins. arXiv:1307.7109 (2013). 23 Latta, C. et al. Confluence of resonant laser excitation and bidirectional quantum-dot nuclear-spin polarization. Nature Physics 5, 758 -- 763 (2009). 24 Hogele, A. et al. Dynamic nuclear spin polarization in the resonant laser excitation of an ingaas quantum dot. Phys. Rev. Lett. 108, 197403 (2012). 25 Vega, S. Fictitious spin 1/2 operator formalism for multiple quantum NMR. The Journal of Chemical Physics 68, 7 5518 -- 5527 (1978). 26 Haase, J., Conradi, M., Grey, C. & Vega, A. Population transfers for NMR of quadrupolar spins in solids. Journal of Magnetic Resonance, Series A 109, 90 -- 97 (1994). 27 van Veenendaal, E., Meier, B. H. & Kentgens, A. P. M. Frequency stepped adiabatic passage excitation of half- integer quadrupolar spin systems. Molecular Physics 93, 195 -- 213 (1998). 28 Stuckelberg, E. Theorie der unelastischen Stossen zwischen Atomen. Helv. Phys. Acta. 5, 369 (1932). 29 Yoakum, S., Sirko, L. & Koch, P. M. Stueckelberg oscillations in the multiphoton excitation of helium rydberg atoms: Observation with a pulse of coherent field and suppression by additive noise. Phys. Rev. Lett. 69, 1919 -- 1922 (1992). 30 Oliver, W. D. et al. Mach-Zehnder interferometry in a strongly driven superconducting qubit. Science 310, 1653 -- 1657 (2005). 31 Huang, P. et al. Landau-Zener-Stuckelberg interferometry of a single electronic spin in a noisy environment. Phys. Rev. X 1, 011003 (2011). Acknowledgements M.M., G.W., A.V.K., M.P. and R.J.W. acknowledge support from NCCR QSIT and EU ITN S3NANO; M.P. and F.X. from the SNI; A.L., D.R. and A.D.W. from Mercur Pr-2013-0001 and BMBF-Q.com-H 16KIS0109. We thank Phani Peddibhotla and Christoph Kloeffel for technical assistance and Christian Degen, Patrick Maletinsky and Hugo Ribeiro for fruitful discussions. Author contributions M.M. and G.W. carried out the experiments, the data analysis and the theoretical modeling. A.V.K. pro- vided expertise in resonance fluorescence on single quantum dots; F.X. expertise in micro-wire design and sample processing. F.X. and M.P. provided electronics and software expertise for the NMR. A.L., D.R. and A.D.W. carried out the molecular beam epitaxy. M.M., G.W., M.P. and R.J.W. took the lead in writing the paper/supplementary information. R.J.W. conceived and managed the project. Additional information Supplementary information is available in the online version of the paper. Reprints and permissions infor- mation is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to M.M. Competing financial interests The authors declare no competing financial interests. 8 2(cid:105) , ↓(cid:105) ≡ − 1 2(cid:105); middle: I = 3 2 without quadrupole interaction; bottom: I = 3 FIG. 1. The experiment: concepts and design. (a) Eigen-energies of the nuclear spin in a static magnetic field along z and oscillating (radio-frequency, RF) magnetic field along x, in the rotating frame. In black are the eigen-energies versus RF detuning in three cases. In blue are the diabatic states. Top: I = 1 2, a two-level system with avoided- crossing at ∆ν = 0, ↑(cid:105) ≡ + 1 2 with quadrupole interaction (νQ (cid:29) νRF ) showing a hierarchy of avoided crossings, the first, second and third quantum transitions (∆m = 1, ∆m = 2 and ∆m = 3, respectively)11. (b) Device for magnetic resonance experiments on the nuclear spins of a single self-assembled quantum dots. The quantum dots are embedded in a vertical tunnelling structure controlled by gate voltage Vg. A gold microwire is fabricated above the gate with a hole for optical access. Magnetic resonance is driven with an RF current passing through the microwire. A solid-immersion-lens enhances the collection efficiency of the resonance fluorescence. (c) Cross-section of a single InGaAs quantum dot (TEM image courtesy of Arne Ludwig and Jean-Michel Chauveau). (d) Top view of microwire. (e) Pulse sequence of NMR experiment. A resonance is established with a constant frequency laser. On ramping the gate voltage, the nuclear spins polarize in order to maintain the optical resonance: the Stark effect is compensated by the Overhauser shift. A RF pulse is then applied to manipulate the nuclear spin ensemble. The optical sequence is repeated to read-out the nuclear spin polarization, acting also as initialization for the next sequence. 10 nm (c) Dn I=3/2 nQ   (a) I=3/2 nQ = 0 I=1/2 20 µm (d) -1/2 -1/2 +1/2 +1/2 +1/2 -1/2 -3/2 +3/2 +1/2 -1/2 -3/2 +3/2 -1/2 +1/2 +3/2 -3/2 -1/2 +1/2 +3/2 -3/2  Laser Vg Iwire t Manipu- lation Read-out / Initialization ON OFF (e) resonant laser Vg SIL Back contact (n+ GaAs) Bz Bx Iwire Top contact (Au) (b) QD layer 1 1 1 2 2 3 SiO2 9 FIG. 2. Adiabatic passage of the nuclear spin ensemble. (a) Resonance fluorescence versus laser detuning on an empty single quantum dot (X0 transition) at Bz = 0 T and T = 4 K. (b) Resonance fluorescence versus laser detuning at Bz = 6 T on the blue X0 transition showing "dragging". The plateau-like features signify nuclear spin polarization. A sequence of sweeps shows clear memory effects. The extent of the plateaux are reproducible to within 0.6 µeV on repeating a specific cycle. In blue (red) the laser is tuned to more negative (positive) values. (c) A sequence of resonance fluorescence sweeps with N chirped RF pulses (ν1 = 32.5 MHz, ν2 = 87.5 MHz, α = 0.18 GHz/s) following nuclear spin polarization (N = 0, 1, 2, 3, 4). N = 0 reads initial (cid:104)Iz(cid:105)11; N = 1 inverts (cid:104)Iz(cid:105); N = 2 restores (cid:104)Iz(cid:105) to almost its N = 0 value, etc. The Overhauser shift (OHS) and the change in Overhauser shift ∆OHS following a chirped pulse are labelled. (d) ∆OHS versus N for large N. The decay at large N arises mostly from relaxation processes during the sweep; the residual signal at large N is presently not understood. Solid lines are guides for the eye. 0.00.10.20.00.10.00.10.00.1-50-40-30-20-10010203040500.00.1 Count rate (MHz)-50-40-30-20-10010203040500.00.10.2 L (µeV)(a)B = 0 T L(µeV)(b)B = 6 T𝜈 t 2x sweep 1x sweep 3x sweep 4x sweep 𝜈 t 𝜈 t 𝜈 t 𝜈 t 1 2 3 4 5 𝛿𝐹𝑆=3.8 µeV FWHM = 1.2 µeV 050100150200250300(d) odd evenNumber of sweepsOHS (µeV)010203040020406080Time (s)30150-15Iz / Imaxz (%)depolarization 0.00.10.20.00.10.00.10.00.1-50-40-30-20-10010203040500.00.1L(µeV)Count rate (MHz)6040200-20-40-60Iz / Imaxz (%) B = 6 T(c) OHS ΔOHS no sweep 10 FIG. 3. Nuclear spin inversion at the first quantum transition in chirped NMR. NMR signal, ∆OHS, following a single chirped RF pulse with ν1 → ν2 (ν1 = 32.5 MHz, ν2 = 87.5 MHz) as a function of sweep rate α: experimental data (open circles) along with theory (dark gray line). The theory uses x = 20.2%, T = 8.2 mK, Bx = 3.8 mT, (cid:10)νQ[75As](cid:11) = 3.0 MHz,(cid:10)νQ[115In](cid:11) = 1.5 MHz,(cid:10)νQ[69Ga](cid:11) = 3.1 MHz,(cid:10)νQ[71Ga](cid:11) = 2.1 MHz. The relative abundances are 75As (100%), 113In (4.3%), 115In (95.7%); 69Ga (60.1%) 71Ga (39.9%). ∆OHS versus α is shown for the four isotopes separately (colour plots). The plateau arises because a range of α exists in which inversion at the first quantum transition is achieved for all isotopes yet inversion at the second quantum transition is achieved for none. At the smallest α, inversion at the first and second quantum transitions is achieved for the majority of In nuclei but only inversion at the first quantum transition for the majority of I = 3 2 nuclei. 10410310210110010-101020304069Ga 75As113/115In OHS (µeV)Sweep rate  (GHz/s)71Ga11 FIG. 4. Isotope-sensitive NMR with chirped pulses. NMR signal ∆OHS following a single chirped RF pulse, ν1 → ν2 as a function of ν2: experimental data (open circles) along with theory (dark gray line). The sweep rate α = 0.09 GHz/s and ν1 = 32.5 MHz. The vertical lines show the text-book NMR frequencies of the In (I = 9 2), Ga (I = 3 2) isotopes: step-wise increases in signal occur each time ν2 crosses these particular frequencies. The theory uses x = 20.2%, T = 8.2 mK and Bx = 3.8 mT as in Fig. 3, along with Gaussian distributions for the quadrupole frequencies (inset). 2) and As (I = 3 30405060708090010203040OHS (µeV) (MHz)75As115In69Ga71Ga02468100.00.10.2 p(Q)Q (MHz)
1601.01162	2	1601	2016-02-17T07:32:18	Crystal-phase quantum dots in GaN quantum wires	[ "cond-mat.mes-hall" ]	We study the nature of excitons bound to I1 basal plane stacking faults in ensembles of ultrathin GaN nanowires by continuous-wave and time-resolved photoluminescence spectroscopy. These ultrathin nanowires, obtained by the thermal decomposition of spontaneously formed GaN nanowire ensembles, are tapered and have tip diameters down to 6 nm. With decreasing nanowire diameter, we observe a strong blue shift of the transition originating from the radiative decay of stacking fault-bound excitons. Moreover, the radiative lifetime of this transition in the ultrathin nanowires is independent of temperature up to 60 K and significantly longer than that of the corresponding transition in as-grown nanowires. These findings reveal a zero-dimensional character of the confined exciton state and thus demonstrate that I1 stacking faults in ultrathin nanowires act as genuine quantum dots.	cond-mat.mes-hall	cond-mat	Crystal-phase quantum dots in GaN quantum wires Pierre Corfdir,∗ Christian Hauswald, Oliver Marquardt, Timur Flissikowski, Johannes K. Zettler, Sergio Fern´andez-Garrido, Lutz Geelhaar, Holger T. Grahn, and Oliver Brandt Paul-Drude-Institut fur Festkorperelektronik, Hausvogteiplatz 5 -- 7, 10117 Berlin, Germany We study the nature of excitons bound to I1 basal plane stacking faults in ensembles of ultrathin GaN nanowires by continuous-wave and time-resolved photoluminescence spectroscopy. These ultrathin nanowires, obtained by the thermal decomposition of spontaneously formed GaN nanowire ensembles, are tapered and have tip diameters down to 6 nm. With decreasing nanowire diameter, we observe a strong blue shift of the transi- tion originating from the radiative decay of stacking fault-bound excitons. Moreover, the radiative lifetime of this transition in the ultrathin nanowires is independent of temperature up to 60 K and significantly longer than that of the corresponding transition in as-grown nanowires. These findings reveal a zero-dimensional character of the confined exciton state and thus demonstrate that I1 stacking faults in ultrathin nanowires act as genuine quantum dots. Spontaneously formed GaN nanowires are compara- ble in structural perfection to state-of-the-art freestanding GaN.1 The nanowire geometry inhibits the propagation of threading dislocations along the nanowire axis, resulting in dislocation-free crystals regardless of the substrate.2,3 In contrast to group-III arsenide and phosphide nanowires, which are synthesized by vapor-liquid-solid growth and are prone to a pronounced polytypism,4,5 spontaneously formed GaN nanowires exclusively crystallize in the wurtzite lattice structure with only occasional I1 basal plane stacking faults (BSFs).6,7 Consequently, the radiative transitions related to excitons bound to I1 BSFs [(I1,X)] in GaN nanowires are spectrally well resolved and distinct from other excitonic transitions in GaN. This fact has been essential for shed- ding light on the nature of the (I1,X).8,9 In particular, for nanowires with a diameter larger than 50 nm, the (I1,X) was shown to exhibit a two-dimensional density of states, i. e., I1 BSFs indeed act as quantum wells.8,10 These so-called crystal-phase quantum structures are free of strain and alloy disorder, their interfaces are atomically abrupt.11 In GaN nanowires, the decay of the (I1,X) is purely radiative up to 60 K.8,9 BSFs thus form an exceptionally well-defined model system for fundamental studies of con- fined excitons. In this context, the recent fabrication of ul- trathin GaN nanowires with a diameter down to 6 nm is of great interest.12 Due to the mismatch in dielectric constants between GaN and air, excitons in these ultrathin nanowires experience a strong radial confinement, i. e., the ultrathin GaN nanowires act as quantum wires despite the fact that their diameter still exceeds at least twice the exciton Bohr radius.12 In addition, the thermal decomposition technique used for the controlled thinning of the as-grown nanowires does not affect their high structural perfection. Finally, de- spite their extremely small diameter, these nanostructures exhibit a high radiative efficiency due to a rather slow sur- face recombination velocity at the nanowires' sidewall facets less than than 210 cm/s at 60 K.8 In this work, we use continuous-wave (cw) and time- resolved (TR) photoluminescence (PL) spectroscopy to in- vestigate the radiative decay and the dynamics of the (I1,X) in GaN quantum wire ensembles fabricated by partial ther- mal decomposition. We demonstrate that I1 BSFs in these FIG. 1. (color online) PL spectra of the as-grown nanowire and the quantum wire ensembles at 5 K acquired with an excitation density of 10 mW/cm2(the spectra have been shifted vertically). The average dB is specified for each sample. The dashed line is a guide to the eye highlighting the blueshift of the (I1,X) transition with decreasing dB. nanowires act as quantum dots. With decreasing nanowire diameter, the (I1,X) transition blueshifts as a result of ra- dial confinement. The radiative lifetime of the (I1,X) in crystal-phase quantum dots does not vary with temperature and is significantly longer than the one measured at 5 K for as-grown nanowires. Using self-consistent eight-band k· p calculations, we show that the increase in radiative lifetime with decreasing diameter results from the reduced coherence area of the (I1,X). Ultrathin GaN nanowires with a length of about 1 µm have been obtained by partial thermal decomposition at 920 °C of GaN nanowire ensembles formed during molecular beam epitaxy on a Si(111) substrate.12 The length and the diame- ter of the as-grown nanowires are 2 µm and 51 nm, respec- tively, and we estimate the density of I1 BSFs to be about 1 per nanowire (see the cathodoluminescence mappings in Ref. 8). The concurrent layer-by-layer desorption of atoms from the top surface and from the sidewalls leads to tapering. The thinnest nanowires obtained exhibit an average base di- ameter dB = 27 nm while the diameter at their tip can be 6 1 0 2 b e F 7 1 ] l l a h - s e m . t a m - d n o c [ 2 v 2 6 1 1 0 . 1 0 6 1 : v i X r a 3.403.453.503.55dB = 29 nmdB = 32 nmdB = 39 nmIntensity (arb. units)Energy (eV)as-grown(I1,X)5 KdB = 27 nm(D0,XA)2 (cid:16) 1 + aT exp kT FIG. 2. (color online) (a) (I1,X) intensity I as a func- tion of temperature T for the as-grown nanowire (tri- angles) and for the quantum wire ensembles with two different average dB. The arrows show the tempera- ture at which the emission intensity starts to quench. The solid line shows the best fit to the (I1,X) intensity using I(T ) ∝ 1/ , with a a fitting parameter and Ea = 57± 5 meV the activation energy for the thermal escape of the (I1,X) from the BSF. (b) PL spectra of the quantum wire ensembles with dB = 39 nm at 5, 60 and 120 K taken with an exci- tation density of 10 mW/cm2 (the spectra have been normalized). (cid:104)− Ea (cid:105)(cid:17) as small as 6 nm.12 Following the results in Ref. 12, ul- trathin nanowires are referred to as quantum wires in the remainder of the paper. Charging effects leads to an over- estimation of the tip diameters when measuring such thin wires by top-view and cross-sectional secondary electron microscopy.12,13 Therefore, the emission properties of GaN quantum wires were correlated with their average dB. Note, however, that the emission is very likely to originate from sections of the nanowires with diameter substantially smaller than dB. Continuous-wave (cw) PL experiments were performed using the 325 nm line of a HeCd laser for excitation. The laser was focused onto the sample to a diameter of 60 µm. The PL signal was analyzed using a monochromator fol- lowed by a charge-coupled device camera for detection. Time-resolved (TR) PL spectroscopy was carried out using the second harmonic of fs pulses obtained from an optical parametric oscillator pumped by a Ti:sapphire laser (emis- sion wavelength and repetition rate of 325 nm and 76 MHz, respectively). The energy fluence per pulse was kept below 0.3 µJ/cm2. The transient emission was spectrally dispersed by a monochromator and detected by a streak camera op- erating in single shot mode. For both cw and TR PL mea- surements, the samples were mounted in a coldfinger cryo- stat whose temperature can be varied between 5 and 300 K. For both experiments, the laser was polarized perpendicu- lar to the nanowire axis. As the average nanowire diameter for all of samples is well within the sub-wavelength range, the coupling of light into the nanowires is getting less and less efficient with decreasing diameter, thus strongly reduc- ing absorption. The photogenerated carrier density in the quantum wires should thus be lower as compared to the as- grown nanowires. Figure 1 shows the PL spectra at 5 K for an ensemble of as-grown nanowires and for ensembles of partially de- composed, quantum wires with dB between 39 and 27 nm. The spectrum for the as-grown ensemble is dominated by the recombination of A excitons bound to neutral O donors at 3.471 eV [(D0,XA)]. The lower energy band centered at 3.410 eV is related to the recombination of excitons bound to I1 BSFs.8,9 Decreasing dB from 51 to 27 nm, the energy of the (D0,XA) transition increases from 3.471 to 3.481 eV (Fig. 1), indicating a progressively stronger confinement of the ex- citons in the corresponding nanowires. This confinement is caused by the mismatch in dielectric constants between GaN and vacuum at the nanowire sidewalls.12,14,15 An even stronger blueshift (42 meV) with decreasing nanowire di- ameter is observed for the (I1,X) line, suggesting that the (I1,X) state in GaN quantum wires is radially confined. In other words, I1 BSFs in GaN quantum wires seem to act as crystal-phase quantum dots. The larger blueshift of the (I1,X) as compared to the one observed for the (D0,XA) line probably arises from the different location of the exciton states involved in these transitions: whereas donors are dis- tributed uniformly along the entire length of the nanowire, I1 BSFs may be located preferentially in the top parts of the nanowires, where the diameter is smaller and the con- finement of the exciton stronger. This result is consistent with the fact that BSFs result from the nanowire coalescence and form several hundreds of nm above the contact point between adjacent nanowires.16 We also observe that the thinnest nanowires exhibit the broadest (I1,X) and (D0,XA) lines at 5 K (Fig. 1). This finding is a direct consequence of the increase in confinement with decreasing diameter: the thinner the nanowire, the larger the impact of diameter fluc- tuations on the (D0,XA) and (I1,X) energies and hence the larger the broadening of the corresponding emission lines. As a result of the significant spectral overlap between the (I1,X) and (D0,XA) transitions for the thinnest nanowires (Fig. 1), we focus in the following on the optical properties of the quantum wires with dB = 39 and 32 nm. Figure 2(a) shows the temperature dependence of the inte- grated intensity of the (I1,X) line for the as-grown nanowire ensembles and for two quantum wire ensembles with dB = 39 and 32 nm. For the as-grown nanowire ensemble, the in- tensity of the (I1,X) transition remains constant between 4 and 60 K, indicating that the recombination of the (I1,X) is purely radiative up to 60 K.8 For temperatures above 60 K, excitons can thermally escape from the comparatively shallow crystal-phase quantum well, leading to an abrupt quenching of the (I1,X) line.17,18 The intensity of the (I1,X) transition follows an Arrhenius behavior with an activation energy Ea = (57± 5) meV [Fig. 1(b)], coinciding with the energy difference between the (I1,X) and the free exciton in fault-free segments.8,18 For the ensembles of quantum (a)10100Intensity (arb. units)Temperature (K)as-growndB = 39 nmdB = 32 nm3.353.403.453.503.555 K60 K120 KIntensity (arb. units)Energy (eV)(b)dB = 39 nm3 FIG. 3. (color online) (a) Streak camera image of an ensemble of quantum wires with dB = 39 nm recorded at 5 K with an energy fluence per pulse of 0.3 µJ/cm2. The intensity is displayed on a logarithmic scale from blue (low intensity) to red (high intensity). (b) Temporal evolution of the PL spectra of an ensemble of quantum wires with dB = 39 nm at 5 K. (c) PL intensity transients of the (I1,X) transition at 5 and 50 K (solid and open symbols, respectively) for the ensembles of as-grown nanowires (triangles) and quantum wires with dB = 39 nm (squares). The peak intensities are normalized. The transients for the quantum wires have been shifted vertically for clarity. The solid lines are exponential fits of the transients. (d) Temperature dependence of τ for the ensembles of as-grown nanowires (triangles) and quantum wires with dB = 39 nm (squares). The lifetimes τ have been obtained from the single exponential fits shown in (c). The dashed line is a guide to the eye showing the linear increase of τ for the as-grown nanowires between 15 and 60 K. wires, the (I1,X) transition also shows a constant PL inten- sity in the low temperature range before decreasing strongly at high temperatures [Fig. 2(a)]. Figure 2(b) shows normal- ized PL spectra taken at 5, 60 and 120 K on the sample with dB = 39 nm. The intensity ratio between the (I1,X) and the free exciton decreases strongly between 60 and 120 K, indi- cating that the quenching of the (I1,X) PL in the quantum wires is also due to the thermal escape of the exciton from the BSFs. The range of constant PL intensity however de- creases with decreasing dB, and the intensity quenching be- comes less abrupt than for the as-grown nanowires. Both of these findings are easily understood: since a smaller diame- ter results in a larger confinement energy for the (I1,X) state (Fig. 1), it also results in a lower value of Ea. Furthermore, the tapering of the quantum wires leads to a distribution of Ea values, which manifests itself in a gradual quenching of the (I1,X) PL intensity as compared to that observed for the as-grown, non-tapered nanowire ensemble. Figure 3(a) displays a streak camera image taken at 5 K on the sample with dB = 39 nm. Spectral profiles taken at vari- ous time delays are shown in Fig. 3(b). The (D0,XA) PL de- cays exponentially with a decay time of 390 ps. This fast de- cay has a nonradiative origin and most probably arises from exciton recombination at point defects.19 Figure 3(c) shows PL intensity transients of the (I1,X) line at 5 and 50 K for the as-grown nanowires and for the quantum wire ensemble with dB = 39 nm. The integrated intensity of the (I1,X) line has been obtained by a spectral deconvolution of the transient spectra.20 Note that due to the significant spectral overlap between the (I1,X) and the (D0,XA) transitions at the early stage of the decay [Fig. 3(b)], the values obtained for the (I1,X) PL intensities of the partially decomposed nanowire sample exhibit a comparatively large uncertainty for the ini- tial 1 ns. At 5 K, the (I1,X) state for the as-grown nanowires decays exponentially, and the decay time τ = 1.0 ns is iden- tical with the radiative lifetime τr. The (I1,X) PL decay for the quantum wires with dB = 39 nm is more complex: it is nonexponential during the first two ns after excitation, and becomes exponential with a decay time τ = 3.4 ns thereafter. Comparable results have been obtained for the sample with 101000.1110Effective lifetime (ns)Temperature (K)02468(I1,X) inquantum wiresIntensity (arb. units)Time (ns)(I1,X) in as-grown nanowires(I1,X) inas-grown nanowires(I1,X) inquantum wires(c)(d)50 K5 K5 K50 KT3.353.403.453.503.55Intensity (arb. units)Energy (eV)0 ns1 ns2 ns3 ns4 ns5 ns6 ns(b)5 K3.403.53.6Energy (eV)Time (ns)2468Intensity (arb. units)100101102103(a)5 K4 the density increases from 1.5 to 155 W/cm2). In view of the results in Figs. 3(b) and 4, it is unlikely that the non- exponential decay observed for the (I1,X) in the quantum wires [Fig. 3(c)] originates from the dynamical descreening of the built-in electric fields. Second, the origin for the nonexponential decay may be associated with the pronounced tapering of the quantum wires. Since the BSFs are likely to occur at different po- sitions along the nanowire axis, their radial dimension and thus the degree of radial confinement also varies. Since the radiative lifetime almost certainly depends on this degree of radial confinement,26,27 a multiexponential decay would be an inevitable consequence. Following this interpretation, the experimental result of longer decay times for the quantum wires implies that the radiative lifetime increases with de- creasing diameter. We will return to this issue after a dis- cussion of the transients at 50 K and the temperature depen- dence of τ. At a temperature of 50 K, for which the PL intensity of our samples is still close to that at 5 K [cf. Fig. 2(a)], the decay of the (I1,X) remains exponential for the as-grown nanowires, but τ increases from 1.0 to 4.2 ns. In contrast, the increased temperature does not affect the decay of the (I1,X) for the quantum wires. Figure 3(d) shows the evolution of τ be- tween 5 and 120 K for both samples. Up to a temperature of 50 -- 60 K, the decay of the (I1,X) is purely radiative and τ = τr. Between 15 and 40 K, τr increases nearly linearly for the as-grown nanowires, demonstrating that (i) I1 BSFs in nanowires with a diameter of 51 nm act as quantum wells,28 and (ii) the (I1,X) is free to move along the BSF plane. The deviation from a linear behavior for temperatures lower than 15 K arises from exciton localization along the BSF plane due to the presence of donors as discussed in Ref. 8. The significant reduction in τ for temperatures larger than 60 K is due to the thermal escape of excitons from the BSF plane. For the quantum wires, the long component of the nonexpo- nential decay is independent of temperature and corresponds to the radiative lifetime of the (I1,X) in the thinnest nanowire segments according to the discussion above. The constant lifetime confirms unambiguously that BSFs in GaN quan- tum wires behave as crystal-phase quantum dots. The re- duction of the dimensionality of the (I1,X) state from two to zero already when dB = 39 nm again suggests that the crystal-phase quantum dots are located in the top part of the nanowires, where the diameter is much smaller. Finally, we address the origin of the increase in τr with decreasing diameter observed above. Considering that the exciton is coherent over the entire BSF, the radiative decay rate Γr = 1/τr can be written approximately as:26,30 (cid:18) R (cid:19)2 a⊥ Γr ∝ f Ac ∝ (cid:104)χe(z)χh(z)(cid:105)2 (1) with the oscillator strength per unit area f , the coher- ence area Ac (which is assumed to be determined by the nanowire's radius R),26,27,30 the overlap integral between the electron and hole wavefunctions along the nanowire axis FIG. 4. (color online) Continuous-wave photoluminescence spectra of an ensemble of quantum wires with dB = 39 nm at 5 K as a function of the excitation power density. dB = 32 nm (not shown). Two different phenomena may, in principle, account for this initial nonexponential decay. First, the discontinuity of the polarization field at the in- terfaces of I1 BSFs induces strong electrostatic fields along the nanowire axis, spatially separating the electron and hole wavefunctions.21 -- 23 A high initial carrier density created by pulsed excitation may screen these fields, giving rise to a minimum value for the radiative lifetime τr directly after ex- citation. Since the carrier density is subsequently reduced by recombination, the electric fields are restored again with time, resulting potentially in a continuous decrease in the (I1,X) energy as well as in a continuous increase in τr.24,25 Since the (I1,X) recombination is purely radiative at low temperature [Fig. 2(a)], the latter increase could explain the nonexponential decay of the (I1,X) emission after pulsed ex- citation. However, the (I1,X) energy remains constant dur- ing the whole decay [Fig. 3(b)]. This finding suggests that the change in carrier density with time after pulsed excita- tion does not lead to strong modifications in the strength of built-in electric fields, and that the nonexponential (I1,X) PL decay seen in Fig. 3(c) is not due to the dynamical descreen- ing of these fields. To confirm this result, we have recorded excitation-density dependent cw PL spectra at 5 K on the sample with dB = 39 nm (Fig. 4). Increasing the excitation density from 0.01 to 1.5 W/cm2 does not lead to any change in the energy of the (I1,X), confirming that screening is neg- ligible in this range of excitation densities.24 Note that the small blueshift observed for larger excitation densities most probably arises from band filling and/or heating effects (see the change in the (D0,XA) and free exciton lineshapes when 3.353.403.453.503.55Intensity (arb. units)Energy (eV)0.01 W/cm20.13 W/cm21.5 W/cm219 W/cm2155 W/cm25 KdB = 39 nm5 quantum disk. With decreasing R, the electron experiences a progressively stronger confinement which results in a strong increase of f as displayed in Fig. 5. A radial separation be- tween the electron and hole wavefunctions as observed in Ref. 31 does not occur for the range of nanowire diameters considered here, a finding that still holds when considering the presence of surface potentials due to Fermi level pinning at the free sidewalls and a homogeneous background doping of 1017 cm−3 (not shown here). The increase in f with de- creasing R should result in a decrease in τr, in contradiction to our experimental observation [cf. Fig. 3(c,d)]. As shown in Fig. 5, the dependence of the radiative de- cay rate on diameter is reversed when taking into account the factor R2 in Eq. (1). This factor accounts for the fact that the exciton's radiative decay is enhanced by its coherent macroscopic polarization.26,27,30 This enhanced radiative de- cay may also be understood in the context of the arguments of Rashba and Gurgenishvili 32: the larger the coherence area of the exciton, the smaller the spread of its wavefunc- tion in k space and thus the shorter τr. As shown in Fig. 5, the total oscillator strength f R2 is indeed predicted to de- crease with decreasing R, and the radiative lifetime τr is thus expected to increase correspondingly in agreement with the experiment. To conclude, changing the diameter of GaN nanowires in a controlled fashion has allowed us to observe the transition from two- to zero-dimensional stacking-fault bound exci- tons. This transition occurs at diameters significantly larger than the exciton's Bohr radius and is induced by dielectric confinement. Because of the absence of structural imperfec- tions such as interfacial steps and alloy fluctuations, the ra- diative decay rate of these excitons scales with the nanowire diameter, which imposes a geometrical limitation of their co- herence area. We thank Alberto Hern´andez-M´ınguez for carefully read- ing our manuscript. Partial funding from the Deutsche Forschungsgemeinschaft within SFB 951 is gratefully ac- knowledged. P.C. acknowledges partial funding from the Fonds National Suisse de la Recherche Scientifique through project 161032. FIG. 5. (color online) Oscillator strength per unit area f and total oscillator strength f R2 for the (I1,X) as a function of the disk di- ameter (solid and dashed lines, respectively). The inset shows the electron (red) and hole (blue) ground state charge densities for a nanowire diameter of 20 nm. The inset has been prepared using Visual Molecular Dynamics.29 (cid:104)χe(z)χh(z)(cid:105), and the Bohr radius of the exciton in the BSF plane a⊥. For examining the change of τr with a decrease in R, we calculate the wavefunction of the (I1,X) in nanowires of diameters between 10 and 50 nm using eight-band k· p calculations.31 The I1 BSF within the GaN nanowire is represented by 3 monolayers of zincblende GaN within a wurtzite GaN segment of 20 nm length. The nanowires are considered to be undoped, i. e., surface potentials are absent. The surface is considered as an infinite potential barrier. The spontaneous polarization of wurtzite GaN induces axial elec- trostatic fields in the I1 BSF with a magnitude of 3 MV/cm.23 We average over all polarization directions, i. e., the oscilla- tor strength f is simply given by (cid:104)χe(z)χh(z)(cid:105)2/a2⊥. Figure 5 shows the dependence of f and f R2 as a function of R. As shown in the inset of Fig. 5 for a diameter of 20 nm, both the electron and the hole are located in the center of the ∗ [email protected] 1 J. K. Zettler, C. Hauswald, P. Corfdir, M. Musolino, L. Geel- haar, H. Riechert, O. Brandt, and S. Fern´andez-Garrido, Cryst. Growth Des. 15, 4104 (2015). 2 S. D. Hersee, A. K. Rishinaramangalam, M. N. Fairchild, L. Zhang, and P. Varangis, J. Mat. Res. 26, 2293 (2011). 3 K. Kishino and S. Ishizawa, Nanotechnology 26, 225602 (2015). 4 J. Bao, D. C. Bell, F. Capasso, J. B. Wagner, T. Martensson, J. Tragardh, and L. Samuelson, Nano Lett. 8, 836 (2008). 5 P. Corfdir, B. Van Hattem, E. Uccelli, S. Conesa-Boj, P. Lefeb- vre, A. Fontcuberta i Morral, and R. T. Phillips, Nano Lett. 13, 5303 (2013). 6 E. Calleja, J. Risti´c, S. Fern´andez-Garrido, L. Cerutti, M. A. S´anchez-Garc´ıa, J. Grandal, A. Trampert, U. Jahn, G. S´anchez, A. Griol, and B. S´anchez, Phys. Status Solidi B 244, 2816 (2007). 7 L. Geelhaar, C. Ch`eze, B. Jenichen, O. Brandt, C. Pfuller, S. Munch, R. Rothemund, S. Reitzenstein, A. Forchel, T. Keha- gias, P. Komninou, G. P. Dimitrakopulos, T. Karakostas, L. Lari, P. R. Chalker, M. H. Gass, and H. Riechert, IEEE J. Sel. Top. Quantum Electro. 17, 878 (2011). 8 P. Corfdir, C. Hauswald, J. K. Zettler, T. Flissikowski, J. Lahnemann, S. Fern´andez-Garrido, L. Geelhaar, H. T. Grahn, and O. Brandt, Phys. Rev. B 90, 195309 (2014). 9 K. P. Korona, A. Reszka, M. Sobanska, P. Perkowska, A. Wysmołek, K. Klosek, and Z. R. Zytkiewicz, J. Lumin. 155, 293 (2014). 10 C. Stampfl and C. G. Van de Walle, Phys. Rev. B 57, R15052 (1998). 00.20.40.60.81.0f (arb. units)fR2 (arb. units)510152025R (nm)BSFR11 N. Akopian, G. Patriarche, L. Liu, J.-C. Harmand, and V. Zwiller, Nano Lett. 10, 1198 (2010). 12 J. K. Zettler, P. Corfdir, C. Hauswald, E. Luna, U. Jahn, T. Flis- sikowski, E. Schmidt, C. Ronning, A. Trampert, L. Geelhaar, H. T. Grahn, O. Brandt, and S. Fern´andez-Garrido, Nano Lett. 16, 973 (2016). 13 B. Loitsch, D. Rudolph, S. Morkotter, M. Doblinger, G. Grimaldi, L. Hanschke, S. Matich, E. Parzinger, U. Wurst- and G. Koblmuller, Adv. bauer, G. Abstreiter, J. J. Finley, Mater. , 2195 (2015). 14 L. V. Keldysh, Pis´ma Zh. Eksp. Teor. Fiz. 29, 716 (1979), [JETP Lett. 29, 658 (1979)]. 15 M. Kumagai and T. Takagahara, Phys. Rev. B 40, 12359 (1989). 16 V. Consonni, M. Knelangen, U. Jahn, A. Trampert, L. Geelhaar, and H. Riechert, Appl. Phys. Lett. 95, 241910 (2009). 17 A. M. Graham, P. Corfdir, M. Heiss, S. Conesa-Boj, E. Uccelli, A. Fontcuberta i Morral, and R. T. Phillips, Phys. Rev. B 87, 125304 (2013). 18 D. Rudolph, L. Schweickert, S. Morkotter, L. Hanschke, S. Hertenberger, M. Bichler, G. Koblmuller, G. Abstreiter, and J. J. Finley, New J. Phys. 15, 113032 (2013). 19 C. Hauswald, P. Corfdir, J. K. Zettler, V. M. Kaganer, K. K. Sabelfeld, S. Fern´andez-Garrido, T. Flissikowski, V. Consonni, T. Gotschke, H. T. Grahn, L. Geelhaar, and O. Brandt, Phys. Rev. B 90, 165304 (2014). 20 C. Hauswald, T. Flissikowski, T. Gotschke, R. Calarco, L. Geel- haar, H. T. Grahn, and O. Brandt, Phys. Rev. B 88, 075312 (2013). 21 Y. J. Sun, O. Brandt, U. Jahn, T. Y. Liu, A. Trampert, S. Cro- 6 nenberg, S. Dhar, and K. H. Ploog, J. Appl. Phys. 92, 5714 (2002). 22 J. Lahnemann, O. Brandt, U. Jahn, C. Pfuller, C. Roder, P. Do- gan, F. Grosse, A. Belabbes, F. Bechstedt, A. Trampert, and L. Geelhaar, Phys. Rev. B 86, 081302 (2012). 23 P. Corfdir and P. Lefebvre, J. Appl. Phys. 112, 053512 (2012). 24 A. Reale, G. Massari, A. Di Carlo, P. Lugli, A. Vinattieri, and D. Alderighi, M. Colocci, F. Semond, N. Grandjean, J. Massies, J. Appl. Phys. 93, 400 (2003). 25 T. Bretagnon, P. Lefebvre, P. Valvin, R. Bardoux, T. Guillet, T. Taliercio, B. Gil, N. Grandjean, F. Semond, B. Damilano, A. Dussaigne, and J. Massies, Phys. Rev. B 73, 113304 (2006). 26 A. V. Kavokin, Phys. Rev. B 50, 8000 (1994). 27 J. Bellessa, V. Voliotis, R. Grousson, X. L. Wang, M. Ogura, and H. Matsuhata, Phys. Rev. B 58, 9933 (1998). 28 P. Corfdir, J. Levrat, A. Dussaigne, P. Lefebvre, H. Teis- seyre, I. Grzegory, T. Suski, J.-D. Gani`ere, N. Grandjean, and B. Deveaud-Pl´edran, Phys. Rev. B 83, 245326 (2011). 29 W. Humphrey, A. Dalke, and K. Schulten, J. Mol. Graph. 14, 33 (1996). 30 O. Brandt, G. C. La Rocca, A. Heberle, A. Ruiz, and K. Ploog, Phys. Rev. B 45, 3803 (1992). 31 O. Marquardt, S. Boeck, C. Freysoldt, T. Hickel, S. Schulz, J. Neugebauer, and E. P. O'Reilly, Comp. Mat. Sci. 95, 280 (2014). 32 E. I. Rashba and G. E. Gurgenishvili, Fiz. Tverd. Tela 4, 1029 (1962), [Sov. Phys. - Solid State 4, 759 (1962)].
1811.03409	1	1811	2018-11-08T13:59:27	High thermoelectric performance in the hexagonal bilayer structure consisting of light boron and phosphorus elements	[ "cond-mat.mes-hall" ]	Two-dimensional layered materials have attracted tremendous attentions due to their extraordinary physical and chemical properties. Using first-principles calculations and Boltzmann transport theory, we give an accurate prediction of the thermoelectric properties of boron phosphide (BP) bilayer, where the carrier relaxation time is treated within the framework of electron-phonon coupling. It is found that the lattice thermal conductivity of BP bilayer is much lower than that of its monolayer structure, which can be attributed to the presence of van der Waals interactions. On the other hand, the graphene-like BP bilayer shows very high carrier mobility with a moderate band gap of 0.88 eV. As a consequence, a maximum p-type ZT value of ~1.8 can be realized along the x-direction at 1200 K, which is amazingly high for systems consisting of light elements only. Moreover, we obtain almost identical p- and n-type ZT of ~1.6 along the y-direction, which is very desirable for fabrication of thermoelectric modules with comparative efficiencies. Collectively, these findings demonstrate great advantages of the layered structures containing earth-abundant elements for environment-friendly thermoelectric applications.	cond-mat.mes-hall	cond-mat	High thermoelectric performance in the hexagonal bilayer structure consisting of light boron and phosphorus elements Z. Z. Zhou, H. J. Liu, D. D. Fan, G. H. Cao, C. Y. Sheng Key Laboratory of Artificial Micro- and Nano-Structures of Ministry of Education and School of Physics and Technology, Wuhan University, Wuhan 430072, China Two-dimensional layered materials have attracted tremendous attentions due to their extraordinary physical and chemical properties. Using first-principles calculations and Boltzmann transport theory, we give an accurate prediction of the thermoelectric properties of boron phosphide (BP) bilayer, where the carrier relaxation time is treated within the framework of electron-phonon coupling. It is found that the lattice thermal conductivity of BP bilayer is much lower than that of its monolayer structure, which can be attributed to the presence of van der Waals interactions. On the other hand, the graphene-like BP bilayer shows very high carrier mobility with a moderate band gap of 0.88 eV. As a consequence, a maximum p-type ZT value of ~1.8 can be realized along the x-direction at 1200 K, which is amazingly high for systems consisting of light elements only. Moreover, we obtain almost identical p- and n-type ZT of ~1.6 along the y-direction, which is very desirable for fabrication of thermoelectric modules with comparative efficiencies. Collectively, these findings demonstrate great advantages of the layered structures containing earth-abundant elements for environment-friendly thermoelectric applications. 1. Introduction In the past decades, two-dimensional (2D) materials have attracted tremendous attentions due to their intriguing properties and great potential in various applications such as deep ultraviolet laser, optoelectronic device, and energy conversion [1−6]. In order to develop high performance thermoelectric materials which can directly convert waste heat into electrical power, 2D systems require moderate band gap, Author to whom correspondence should be addressed. Electronic mail: [email protected] 1 reasonably large carrier mobility, and relatively low thermal conductivity. Unfortunately, the thermoelectric applications of most 2D materials are more or less limited by some inherent weaknesses. For example, graphene has been investigated intensively for its ultrahigh carrier mobility and massless Dirac fermions, while the gapless energy band makes it unsuitable for the application as thermoelectric material [7]. In contrast, the graphene-like III -- V binary compounds are all predicted to exhibit direct band gaps [8], whereas the representative hexagonal boron nitride (BN) monolayer, which has been successfully isolated, behaves as an insulator with a wide band gap of about 6.0 eV [9−13]. Recently, another III -- V binary compound named hexagonal BP monolayer has also attracted considerable attentions [14−18]. Experimentally, the BP films with wurtzite structure have been synthesized by chemical vapor deposition [19]. Similar to BN monolayer, it is believed that BP layers could also adopt hexagonal lattice if prepared appropriately [ 20 ]. Theoretically [ 21 ], the phonon dispersion relations of BP monolayer show no imaginary frequency, and its in-plane Young modulus and Poisson's ratio are comparable with those of the MoS2 monolayer. All these ensure a satisfied mechanical stability of BP monolayer. Besides, Wang et al. [22] found that the B−P bond is still reserved in the Born-Oppenheimer molecular dynamics simulation for 5 ps at the temperature of 2500 K, which suggests high thermal stability of BP monolayer. On the other hand, it was reported that BP monolayer is a semiconductor with low effective mass and moderate band gap [23,24]. It is thus naturally reminiscent of the thermoelectric application of BP monolayer, whereas an inherent problem is that almost all of the hexagonal monolayer structures consisting of light elements have pretty large lattice thermal conductivities [25,26]. Fortunately, previous investigations revealed that the lattice thermal conductivities of bilayer graphene and hexagonal BN are obviously lower than those of their monolayer counterparts [27−31]. Accordingly, we believe that BP bilayer should also have lower lattice thermal conductivity and thus better thermoelectric performance compared with BP monolayer, and a comprehensive understanding on its electronic and phonon transport properties is quite necessary. 2 In this work, the thermoelectric transport properties of the hexagonal BP bilayer are investigated by using first-principles calculations and Boltzmann theory. It is demonstrated that the lattice thermal conductivity of BP bilayer is much lower than that of monolayer structure due to the van der Waals (vdW) interactions. Combined with the extremely large power factors originated from the ultrahigh carrier mobility, an unexpected high thermoelectric performance can be realized in both p- and n-type BP bilayer in the high temperature region. 2. Computational method The lattice thermal conductivity ( ) of the BP bilayer can be obtained by solving the phonon Boltzmann transport equation, as implemented in the so-called ShengBTE package [32]. The second- and third-order interatomic force constants (IFCs) are investigated by density functional theory (DFT) combined with the finite displacement method using a supercell. The ninth nearest neighbors are included for the third-order interactions to ensure convergence. The DFT calculations are implemented in the Vienna ab-initio simulation package (VASP) [33], and the harmonic and anharmonic IFCs are computed based on the PHONOPY program [34] and the THIRDORDER.PY script [32], respectively. The electronic properties of the BP bilayer are calculated within the framework of DFT, as implemented in the QUANTUM ESPRESSO package [ 35 ]. The norm-conserving scalar-relativistic pseudopotential is used to describe the core-valence interaction [36], and the exchange-correlation functional is in the form of Perdew-Burke-Ernzerhof (PBE) with the generalized gradient approximation (GGA) [37]. To eliminate the interactions between the bilayer and its periodic images, the system is modeled by adopting hexagonal supercell with the vacuum distance of 30 Å. The vdW interactions are treated by including the optB86b functional when optimizing the lattice parameters [38]. The kinetic energy cutoff is set to 80 Ry and the atomic positions are fully relaxed. To give accurate electronic transport coefficients, the band structure of BP bilayer is obtained by adopting hybrid density 3 l551functional in the form of Heyd-Scuseria-Ernzerhof (HSE) [39], which has been successfully used to predict the band gaps of bulk BP [22] and hexagonal BN monolayer [40]. Based on the calculated band structure, the Seebeck coefficient ( ), the electrical conductivity ( ), and the electronic thermal conductivity ( ) can be derived from the Boltzmann transport theory [41], where the key point is appropriate treatment of the carrier relaxation time . Using the density functional perturbation theory (DFPT) [42] and the Wannier interpolation techniques [43], the k resolved relaxation time can be obtained from the imaginary part of the electron self-energy by a complete electron-phonon coupling (EPC) calculation [44]. To ensure converged results, the EPC calculations have been performed by using coarse grids of k-points with q-points, and then interpolated to dense meshes of k-points with q-points via the maximally localized Wannier functions, as implemented in the electron-phonon Wannier (EPW) package [44]. To make a better comparison, the transport coefficients , , and are all renormalized with respect to the interlayer distance of BP bilayer. 3. Results and discussion Owing to the low cleavage energy, the BP bilayers have been proved to be easily realized in experiment [22]. Among several possible bilayer systems, the stacking configuration shown in Figure 1 is the most stable one [22,45]. As can be seen from Fig. 1(a), the bottom B1 atoms entirely overlap with the top B2 atoms in the xy plane, while the P1 atoms are positioned at the center of the hexangular ring consisting of B2 and P2 atoms. The optimized interlayer distance is 3.57 Å, and the atoms in each layer are almost completely located in the same plane. The bilayer structure keeps the same B−P bond length (1.86 Å) as that of the BP monolayer, which is in good agreement with previous theoretical results [45]. Figure 2(a) displays the phonon spectrum of the BP bilayer, where there is no imaginary frequency guaranteeing the dynamic stability of the system. As the primitive cell of the bilayer system contains four atoms, there exist 12 phonon 4 Se202011010120020011001001elbranches with three acoustic (LA, TA, ZA) and nine optical ones (LOi, TOi, ZOi, i = 1, 2, 3). It can be seen that the LA and TA modes show linear dependence on the wave vector near the Brillouin zone, while the ZA mode exhibits a quadratic dispersion, which are very similar to those found in layered systems such as graphene [25] and hexagonal BN monolayer [31]. As a comparison, Fig. 2(b) also plots the phonon dispersion relations of BP monolayer. If two independent monolayers are held together, one would expect a complete overlap of their phonon dispersions. Indeed, we see that Fig. 2(a) and Fig. 2(b) are very similar to each other and the doubly degenerate bands (TA and TO1, LA and LO1, TO2 and TO3, LO2 and LO3) remain unchanged. The only exception is that the ZA and ZO1 modes exhibit obvious band splitting around the  point, which is less pronounced for the ZO2 and ZO3 modes. Such observation suggests that the interlayer vdW interactions in the BP bilayer mainly affect the out-of-plane phonon modes. Meanwhile, we see that all the ZO modes hybridize with the acoustic modes in the BP bilayer, which means more three-phonon process and thus a lower relaxation time [46]. We will come back to this point later. Fig. 2(c) plots the lattice thermal conductivity of the BP bilayer as a function of temperature from 300 K to 1300 K. For comparison, the result of the BP monolayer is also shown in the inset. In the whole temperature range, we see that the of the BP bilayer exhibit less isotropy and the values are nearly one order of magnitude lower than those of the monolayer structure. Table 1 summarizes the contribution of different phonon modes to the lattice thermal conductivity at a prototypical temperature of 1200 K. When the system changes from the monolayer to bilayer, we see there is a significant reduction of the lattice thermal conductivity contributed by the acoustic modes, especially for the TA and ZA phonons. In another word, the heat transport by the acoustic phonons is largely limited in the BP bilayer. To have a further understanding, we first note that the phonon group velocities of BP bilayer and monolayer are almost identical to each other (see Figure S1(a) and S1(b) in the supplementary materials). This is reasonable since the phonon dispersion relations in both systems are very similar. We further observe from Fig. 2(d) and its 5 linset that the relaxation times of the acoustic phonons in the BP bilayer are at least one order of magnitude lower than those in the BP monolayer, which can be attributed to the fact that more optical branches hybridizes with acoustic branches (see Fig. 2(a)) caused by the vdW interactions discussed above. Besides, the relaxation times of the low frequency optical phonons are also obviously lower for the bilayer system. As a consequence, the remarkably lower lattice thermal conductivity of the BP bilayer can be attributed to its significantly lower phonon relaxation time caused by the presence of vdW interactions. It should be noted that the EPC may also have certain influences on the phonon transport properties [47−50], especially at high carrier concentration. However, our additional calculations (see Figure S2 in the supplementary materials) find that the phonon relaxation time from the intrinsic phonon-phonon scattering is at least two orders of magnitude lower than that originating from the EPC. It is thus reasonable to neglect the effects of EPC on the lattice thermal conductivity of the BP bilayer. Figure 3(a) displays the energy band structure of the BP bilayer calculated by adopting the HSE functional. In contrast to that of the BP monolayer (see Figure S3 in the supplementary materials), the conduction band minimum (CBM) of the bilayer system is located at the K point whereas the valence band maximum (VBM) is slightly shifted away. The HSE bands show an indirect gap of 0.88 eV, which is more suitable for thermoelectric application compared with that of the monolayer system (1.37 eV). Besides, we see there are two top valence bands degenerated at the K point, which indicates a sharp increase in the density of state (DOS) near the VBM and thus a large p-type Seebeck coefficient is expected [51]. In contrast, the steep energy dispersion around the CBM means a relatively lower n-type Seebeck coefficient but a higher group velocity. Note that the dispersions around the VBM and CBM of BP bilayer are similar to those of the monolayer, suggesting that the carrier mobility of BP bilayer should be comparable with that of the monolayer structure which has been proved to be very high in previous work [24]. We now move to the discussion of the carrier relaxation time of the BP bilayer. It should be noted that better thermoelectric performance should appear at higher 6 temperatures due to its relatively larger band gap compared with those of the Bi2Te3 [52] and SnSe [53]. In Fig. 3(b), we plot the energy dependence of the carrier relaxation time at 1200 K. It is found that the BP bilayer exhibits higher n-type carrier relaxation time, which can be attributed to less scattering channels near the CBM originating from the lower DOS. The relaxation times near the VBM and CBM are calculated to be 21 fs and 42 fs at 1200 K, respectively, which are obviously higher than those of Bi2Te3 [52] and suggest very favorable electronic transport properties of the bilayer system. Based on the calculated band structure and the carrier relaxation time, the electronic transport coefficient of the BP bilayer can be accurately predicted by solving the Boltzmann transport equation. Figure 4(a) and 4(b) show the absolute values of the Seebeck coefficient and the electrical conductivity of BP bilayer at 1200 K, respectively. In the concentration range of ~ , we find obvious higher value of p-type Seebeck coefficient caused by the band degeneracy around the VBM while larger n-type electrical conductivity due to the strong band dispersion near the CBM. As a consequence, we see from Fig. 4(c) that the BP bilayer shows similar p- and n-type power factors (S2σ) at 1200 K, especially for those along the y-direction. At the optimized electron concentration of , the y-direction Seebeck coefficient of −228 μV/K for the BP bilayer is comparable with those found in many good thermoelectric materials [52,53], where the electrical conductivity of the former is much larger (3113 S/cm) due to its ultrahigh carrier mobility (6294 cm2/Vs). Extraordinarily high power factor ( ) can thus be realized at 1200 K, suggesting the great advantage of the BP bilayer for thermoelectric application. With all the transport coefficients obtained, we can now evaluate the thermoelectric performance of the BP bilayer, which is characterized by the dimensionless figure-of-merit . Fig. 4(d) displays the calculated ZT values as a function of carrier concentration at 1200 K, where we see that both the p- and n-type BP bilayer exhibit good thermoelectric performance. The highest ZT value of 1.8 can 7 12310132310 cm122410 cm221.610 W/mK2elZTSTbe achieved along the x-direction at the hole concentration of , which is amazingly high for systems consisting of light elements only. For the case of y-direction, we see that both the p- and n-type systems exhibit almost identical ZT value of ~1.6, which is very desirable for fabrication of thermoelectric modules with comparative efficiencies. The optimized ZT values of the BP bilayer are summarized in Table 2 together with the corresponding carrier concentration and transport coefficients. On the other hand, although the best thermoelectric performance appears at 1200 K, our addition calculations (see Figure S4 in the supplementary materials) show that fairly good ZT values above 1.0 can already be achieved at 700 K, indicating that the BP bilayer can also be operated in the intermediate temperature region. Compared with traditional thermoelectric materials, the BP bilayer, only consisting of light, earth-abundant, and environment-friendly elements, may be very favorable candidate to be designed as promising thermoelectric materials. 4. Summary We present a comprehensive study of the thermoelectric properties of the BP bilayer, which shows a much lower lattice thermal conductivity compared with that of the monolayer structure caused by the presence of the vdW interactions, as well as a quite large power factor originated from the ultrahigh carrier mobility. As a consequence, excellent thermoelectric performance can be achieved in the bilayer system, where almost identical ZT values of p- and n-type systems are found in a wide temperature range, indicating great advantage of the fabrication of thermoelectric modules with comparative efficiencies. It is interesting to check whether other hexagonal layered structure in the III -- V compounds could also exhibit good thermoelectric performance by appropriate design of the constituent elements and stacking order, which will be addressed in our future work. 8 1321.210 cmAcknowledgements We thank financial support from the National Natural Science Foundation (Grant Nos. 51772220 and 11574236). The numerical calculations in this work have been done on the platform in the Supercomputing Center of Wuhan University. Table 1 The contributions of different phonon modes to the lattice thermal conductivity (in unit of W/mK) of the BP monolayer and bilayer at 1200 K. direction monolayer bilayer x y x y 12.1 30.9 0.5 0.4 21.3 16.7 1.3 0.3 13.8 14.1 1.1 1.7 8.2 9.4 4.2 5.2 Table 2 The optimized ZT values of the BP bilayer at 1200 K. The corresponding carrier concentration and transport coefficients are also listed. carriers x y hole electron hole electron n S σ S2σ κe κl (S/cm) 3001 2675 3030 3113 (10−3 W/mK2) (W/mK) (W/mK) 17 14 16 16 4.4 4.2 4.7 4.3 7.1 7.1 7.6 7.6 (1012 cm−3) (μV/K) 12 4.1 13 4.0 237 −228 233 −228 ZT 1.8 1.5 1.6 1.6 9 ZATALAOPFigure 1. Ball-and-stick model of BP bilayer: (a) top-view, and (b) side-view. 10 Figure 2. The phonon dispersion relations of the BP (a) bilayer and (b) monolayer. (c) The lattice thermal conductivity of BP bilayer as a function of temperature. (d) The phonon relaxation time of the BP bilayer as a function of frequency. The insets in (c) and (d) are the lattice thermal conductivity and the phonon relaxation time of the BP monolayer, respectively. 11 Figure 3. (a) The electronic band structures of BP bilayer calculated by adopting HSE functional. (b) The energy-dependent carrier relaxation time of BP bilayer at 1200 K. The Fermi level is at 0 eV. 12 Figure 4. (a) The absolute values of the Seebeck coefficient, (b) the electrical conductivity, (c) the power factor, and (d) the ZT values of BP bilayer, plotted as a function of carrier concentration at 1200 K. 13 References [1] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, and A. A. Firsov, Science 306, 6664 (2004). [2] K. Watanabe, T. Taniguchi, and H. Kanda, Nat. Mater. 3, 404 (2004). [3] Y. Zhang, Y. Tan, H. Stormer, and P. Kim, Nature 438, 201 (2005). [4] K. S. Novoselov, Z. Jiang, Y. Zhang, S. V. Morozov, H. L. Stormer, U. Zeitler, J. C. Maan, G. S. Boebinger, P. Kim, and A. K. Geim, Science 315, 1379 (2007). [5] D. Malko, C. Neiss, F. Viñes, and A. Görling, Phys. Rev. Lett. 108, 086804 (2012). [6] N. Ding, J, Q, Xu, Q. Zhang, J. W. Su, Y. Gao, X. Zhou, and T. Y. Zhai, ACS Appl. Mater. Interfaces 10, 10296 (2018). [7] A. H. Castro Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K. Geim, Rev. Mod. Phys. 81, 109 (2009). [8] H. Şahin, S. Cahangirov, M. Topsakal, E. Bekaroglu, E. Akturk, R. T. Senger, and S. Ciraci, Phys. Rev. B: Condens. Matter Mater. Phys. 80, 155453 (2009). [9] E. Hernández, C. Goze, P. Bernier, and A. Rubio, Phys. Rev. Lett. 80, 4502 (1998). [10] C. H. Jin, F. Lin, K. Suenaga, and S. Iijima, Phys. Rev. Lett. 102, 195505 (2009). [11] J. C. Meyer, A. Chuvilin, G. Algara-Siller, J. Biskupek, and U. Kiser, Nano Lett. 9, 2683 (2009). [12] C. Zhi, Y. Bando, C. Tang, H. Kuwahara, and D. Golberg, Adv. Mater. 21, 2889 (2009). [13] A. Nag, K. Raidongia, K. P. S. S. Hembram, R. Datta, U. V. Waghmare, and C. N. R. Rao, ACS Nano 4, 1539 (2010). [14] M. Wu, Z. Zhang, and X. C. Zeng, Appl. Phys. Lett. 97, 093109 (2010). [15] Y. Wang and S. Shi, Solid State Commun. 150, 1473 (2010). [16] Q. Lin, X. Zou, G. Zhou, R. Liu, J. Wu, J. Li, and W. Duan, Phys. Chem. Chem. Phys. 13, 12225 (2011). [17] H. Jiang, T. Zhao, L. Shi, P. Tan, and L. An, J. Phys. Chem. C 120, 6612 (2016). [18] H. R. Jiang, W. Shyy, M. Liu, L. Wei, M. C. Wu, and T. S. Zhao, J. Mater. Chem. A 5, 672 (2017). [ 19 ] B. Padavala, C. D. Frye, X. Wang, Z. Ding, R. Chen, M. Dudley, B. 14 Raghothamachar, P. Lu, B. N. Flanders, and J. H. Edgar, Cryst. Growth Des. 16, 981 (2016). [20] C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T. Taniguchi, P. Kim, K. L. Shepard, and J. Hone, Nat. Nanotechnol. 5, 722 (2010). [21] D. Çakır, D. Kecik, H. Sahin, E. Durgun, and F. M. Peeters, Phys. Chem. Chem. Phys. 17, 13013 (2015). [22] S. F. Wang and X. J. Wu, Chin. J. Chem. Phys. 28, 588 (2015). [23] H. L. Zhuang and R. G. Hennig, Appl. Phys. Lett. 101, 153109 (2012). [24] B. W. Zeng, M. J. Li, X. J. Zhang, Y. G. Yi, L. P. Fu, and M. Q. Long, J. Phys. Chem. C 120, 25037 (2016). [25] B. D. Kong, S. Paul, M. Buongiorno Nardelli, and K. W. Kim, Phys. Rev. B 80, 033406 (2009). [26] I. Jo, M. T. Pettes, J. Kim, K. Watanabe, T. Taniguchi, Z. Yao, and L. Shi, Nano Lett. 13, 550 (2013). [27] J. H. Seol, I. Jo, A. L. Moore, L. Lindsay, Z. H. Aitken, M. T. Pettes, X. Li, Z. Yao, R. Huang, D. Broido et al., Science 328, 213 (2010). [28] L. Lindsay, D. A. Broido, and N. Mingo, Phys. Rev. B 83, 235428 (2011). [29] H. Li, H. Ying, X. Chen, D. L. Nika, A. I. Cocemasov, W. Cai, A. A. Balandin, and S. Chen, Nanoscale 6, 13402 (2014). [30] R. D'Souza and S. Mukherjee, Phys. Rev. B 95, 085435 (2017). [31] R. D'Souza and S. Mukherjee, Phys. Rev. B 96, 205422 (2017). [32] W. Li, J. Carrete, N. A. Katcho, and N. Mingo, Comput. Phys. Commun. 185, 1747 (2014). [33] G. Kresse, and J. Furthműller, Comput. Mater. Sci. 6, 15 (1996). [34] A. Togo, and I. Tanaka, Scripta Mater. 108, 1 (2015). [35] P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, et al., J. Phys.: Condens. Matter 21, 395502 (2009). [36] C. Hartwigsen, S. Goedecker, and J. Hutter, Phys. Rev. B: Condens. Matter Mater. Phys. 58, 3641 (1998). [37] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 77, 3865 (1996). [38] J. Klimeš, D. R. Bowler, and A. Michaelide, Phys. Rev. B: Condens. Matter Mater. Phys. 83, 195131 (2011). [39] J. Heyd, G. E. Scuseria, and M. Ernzerhof, J. Chem. Phys. 124, 219906 (2006). 15 [40] K. Watanabe, T. Taniguchi, and H. Kanda, Nature Mater. 3, 404 (2004). [41] G. K. H. Madsen and D. J. Singh, Comput. Phys. Commun. 175, 67 (2006). [42] S. Baroni, S. de Gironcoli, A. D. Corso, and P. Giannozzi, Rev. Mod. Phys. 73, 515 (2001). [43] F. Giustino, M. L. Cohen, and S. G. Louie, Phys. Rev. B 76 165108 (2007). [44] J. Noffsinger, F. Giustino, B. D. Malone, C. H. Park, S. G. Louie, and M. L. Cohen, Comput. Phys. Commun. 181, 2140 (2010). [45] X. P. Chen, C. J. Tan, Q. Yang, R. S. Meng, Q. H. Liang, J. K. Jiang, X. Sun, D. Q. Yang, and T. L. Ren, Phys. Chem. Chem. Phys. 18, 16229 (2016). [46] D. Wee, B. Kozinsky, N. Marzari, and M. Fornari, Phys. Rev. B 81, 045204 (2010). [47] B. L. Liao, B. Qiu, J. Zhou, S. Huberman, K. Esfarjani, and G. Chen, Phys. Rev. Lett. 114, 115901 (2015). [48] T. J. Zhu, G. T. Yu, J. Xu, H. J. Wu, C. G. Fu, X. H. Liu, J. Q. He, and X. B. Zhao, Adv. Electron. Mater. 2, 1600171 (2016). [49] Y. Wang, Z. X. Lu, and X. L. Ruan, J. Appl. Phys. 119, 225109 (2016). [50] D. D. Fan, H. J. Liu, L. Cheng, J. H. Liang, and P. H. Jiang, J. Mater. Chem. A 6, 12125 (2018). [51] J. P. Heremans, V. Jovovic, E. S. Toberer, A. Saramat, K. Kurosaki, A. Charoenphakdee, S. Yamanaka, and G. J. Snyder, Science 321, 554 (2008). [52] L. Cheng, H. J. Liu, J. Zhang, J. Wei, J. H. Liang, J. Shi, and X. F. Tang, Phys. Rev. B 90, 085118 (2014). [53] L. D. Zhao, G. Tan, S. Hao, J. He, Y. Pei, H. Chi, H. Wang, S. Gong, H. Xu and V. P. Dravid, Science 351, 141 (2016). 16
1005.5138	2	1005	2010-07-23T23:12:14	Spin waves in diluted magnetic quantum wells	[ "cond-mat.mes-hall" ]	We study collective spin excitations in two-dimensional diluted magnetic semiconductors, placed into external magnetic field. Two coupled modes of the spin waves (the electron and ion modes) are found to exist in the system along with a number of the ion spin excitations decoupled from the electron system. We calculate analytically the spectrum of the waves taking into account the exchange interaction of itinerant electrons both with each other and with electrons localized on the magnetic ions. The interplay of these interactions leads to a number of intriguing phenomena including tunable anticrossing of the modes and a field-induced change in a sign of the group velocity of the ion mode.	cond-mat.mes-hall	cond-mat	Spin waves in diluted magnetic quantum wells P. M. Shmakov2, A. P. Dmitriev1,2, and V. Yu. Kachorovskii1,2 1Institut fur Nanotechnologie, Forschungszentrum Karlsruhe, 76021 Karlsruhe, Germany 2A.F.Ioffe Physico-Technical Institute, 194021 St. Petersburg, Russia (Dated: March 28, 2018) We study collective spin excitations in two-dimensional diluted magnetic semiconductors, placed into external magnetic field. Two coupled modes of the spin waves (the electron and ion modes) are found to exist in the system along with a number of the ion spin excitations decoupled from the electron system. We calculate analytically the spectrum of the waves taking into account the exchange interaction of itinerant electrons both with each other and with electrons localized on the magnetic ions. The interplay of these interactions leads to a number of intriguing phenomena including tunable anticrossing of the modes and a field-induced change in a sign of the group velocity of the ion mode. PACS numbers: 75.30.Ds, 85.75.-d, 76.50.+g Diluted magnetic semiconductors (DMS) have recently been the subject of great interest [1, 2] due to their potential in combining magnetic and semiconductor properties in a single mate- rial. The DMS are formed by replacing of cations in ordinary semiconductors with magnetic ions, typically Mn ions. Strong exchange interaction between the itinerant electrons and the electrons localized on d-shells of the magnetic ions leads to a number of remarkable features of the DMS. In particular, it results in the effective indirect interaction between the ion spins thus promis- ing for creating room-temperature ferromagnetic systems that may offer advantages of semicon- ductors. It also dramatically enhances the effec- tive coupling of the itinerant electrons with the external magnetic field. In contrast to conven- tional GaAs/GaAlAs systems, where small val- ues of g−factor prevents manipulation of the spin degree of freedom, the giant electron Zeeman splitting arising in the DMS as a manifestation of the exchange interaction can be on the order of the Fermi energy [3, 4], offering a wide range of spintronics applications. Here we discuss spin excitations in the two- dimensional DMS. Our studies are motivated by recent experiments [5 -- 7] and a theoretical dis- cussion [7 -- 9] focused on the spin dynamics in diluted magnetic Cd1−xMnxTe quantum wells placed into the magnetic field [10]. In Ref. 5, the spectrum of the spin waves, ω(k), was mea- sured. Only one excitation mode was observed. It was demonstrated that the excitations exist in a finite range of wavelengths, k < km, and their group velocity is negative: dω(k)/dk < 0. The experimental data were interpreted [5, 8] in terms of conventional spin waves in the Fermi liquids [11], while km was attributed to the edge of the Stoner continuum of the single-particle spin exci- tations. Such interpretation implies that the only effect of the magnetic ions on the electron spin waves is the strong renormalization of the elec- tron Zeeman splitting. However, more recent ex- perimental observations [6, 7] supported by the- oretical studies [7, 9] appear to be in disagree- ment with this conclusion. Indeed, in Refs. [6, 7] two modes of the collective homogeneous (k = 0) spin excitations were observed in Cd1−xMnxTe wells. The modes were identified [6, 7, 9] as the spin excitations of delocalized electrons (the electron mode) and the electrons on d-shells of Mn ions (the ion mode). The dependencies of the frequencies ω1,2(0) of observed modes on B are shown schematically in Fig. (1). The most important observation is the anticrossing of the modes which occurs at a certain "resonant" value of magnetic field B = Bres [6, 7]. As it was also shown in Ref. [7], other types of the homogeneous spin modes may exist in the system correspond- ing to excitations of the ion spins decoupled from the spins of the itinerant electrons. In this paper, we develop a theory of the spin waves in diluted magnetic quantum wells placed into magnetic field. We study analytically two collective modes which correspond to coupled propagation of the electron and ion spin excita- tions. We also discuss the ion modes decoupled from the electron system. To describe the homo- geneous spin oscillations (k = 0), it is sufficient to take into account only one type of exchange inter- action: the interaction of the itinerant electrons with electrons localized on the Mn ions. The thus obtained results coincide with those presented in Refs. [6, 7]. For k 6= 0, the exchange interaction between delocalized electrons comes into play. Our main purpose is to demonstrate that the si- multaneous presence of two types of exchange in- teraction gives rise to interesting phenomena the most remarkable one is the magnetic-field-driven anticrossing of the spin modes. In contrast to the case k = 0, the anticrossing can take place in a wide range of the magnetic fields and may be tuned by the field to occur at a certain value of k (see Fig. 2). FIG. 1: Anticrossing of the electron and ion spin pre- cession frequencies Ωe and ΩJ at "resonant" mag- netic field B = Bres; Ω0 e is the edge of the Stoner continuum. We consider the 2D degenerate electron gas in- teracting with randomly placed magnetic ions. The electrons are located in the r = (x, y) plane and occupy the lowest level in the well. The ions are distributed homogeneously with the 2D concentration nJ , which is assumed to be much higher than the electron concentration ne. The magnetic field is applied parallel to the well plane (B k ex). The field leads to the Zeeman split- ting of the electron and ion spin levels with en- ergies ωe and ΩJ , respectively, while the or- bital motion remains intact. The Hamiltonian of electron-ion exchange interaction reads HJe = Pauli matrix vector, Jk are the spin operators of the ions located at the points Rk = (rk, zk), −α/2Pk σJkδ(r − rk)Ψ(zk)2, where σ is the and Ψ(z) = p2/a sin(πz/a) is a wave func- tion of the lowest level in a rectangular well of width a. Since nJ ≫ ne, the distance be- tween the ions is much smaller than the electron wave length and the mean-field approximation is applicable. In this approximation, we first re- place σJk with hσiJk + σhJki, where averag- ing is taken over density matrix of the system ρ. After such decoupling one may search the so- lution of the quantum Liouville equation for ρ as a product of the electron and ion density matri- ces: ρ = ρe(r, r′, t)Qk ρk(t). The averaged spins e R s(r, p, t)d2p/(2π)2, Jk(t) = Tr(ρk(t)Jk), of the electrons and ions are given by s0(r, t) = n−1 where s(r, p, t) = Tr(σ f )/2 is the electron spin density and f = f (r, p, t) is the Wigner function corresponding to ρe. As a next step, we replace Jk(t) with a smooth function J(r, z, t). Doing so, one finds the electron spin precession frequency in the ion-induced exchange field ωeJ (r, t) = αnJ ¯J(r, z, t)/a, (1) [here ¯J(r, t) =R dzΨ(z)2J(r, z, t)], and the local frequency of the ion spin precession ωJe(r, z, t) = αnes0(r, t)Ψ(z)2/. (2) In addition to the electron-ion exchange inter- action, we take into account the isotropic ferro- magnetic electron-electron exchange interaction by adding the following term [11] ωee(r, t) = −2Gneν−1s0(r, t), (3) The equilibrium electron to the electron spin precession frequency. Here G < 0 is the interaction constant (we assume G < 1), ν = m/2π2, and m is the electron effective mass. den- (here and ǫ = p2/2m), is expressed via the effective Zeeman splitting Ω0 e = ωe +ωeJ +ωee is found self-consistently from Eqs. (1) and (3): spin seq(p) = [n↑(ǫ) − n↓(ǫ)] ex/2 e/2 − EF )/T ] + 1(cid:9)−1 n↑↓(ǫ) =(cid:8)exp[(ǫ ∓ Ω0 e. The frequency Ω0 sity, where Ω0 e = Ωe/(1 + G), Ωe = ωe + αnJ J eq x /a, (4) (5) x ex = 5 is the effective electron Zeeman splitting renor- malized by exchange interaction with the ions. In deriving these equations, J(r, z) was sub- stituted with the equilibrium ion polarization, Jeq = J eq 2 B5/2(ΩJ /T )ex, where BJ (x) is the Brillouin function. We also assumed that the equilibrium exchange field acting on the ions is small, αneseq ene/4EF a ≪ ΩJ , which implies that the equilibrium ion polarization is not affected by exchange interaction. In con- trast, the electron Zeeman splitting is strongly x /a ≈ αΩ0 enhanced due to high ion concentration, so that Ωe ≫ ωe (ωe < 0, because of the negative elec- tron g-factor, which explains non-monotonic de- pendence of Ωe on B shown in Fig. 1 [6, 7]). The out-of-equilibrium spin dynamics can be described by the Landau-Silin equation [12] for the electrons and the Bloch equation for the ions: ∂ f ∂t ∂J ∂t , 1 p + ∂ ε m∇ f − 2( ∂ f ∂r) + + [ΩJ ex + ωJe] × J = 0. ∂p i [ε, f ] = 0, (6) (7) Here [··· ] and {···} stand for the commuta- tor and the anticommutator, respectively, and ε = −[ωeex + ωeJ + ωee]σ/2. For Ω0 e ≪ EF , Eqs. (6) and (7) give the following system of equations for the perpendicular (with respect to B) components of the electron and ion spins + (vF n∇ + iΩ0 + iΩJ ¯J = iδ2s0. e)(s + Gs0) = δ1(i ¯J + ξn∇ ¯J ), (8) ∂s ∂t ∂ ¯J ∂t Here s = sy + isz, ¯J = ¯Jy + i ¯Jz, vF is Fermi velocity, n = (cos ϕ, sin ϕ), ϕ is the velocity an- gle in the well plane, δ1 = αnJ seq x /a, δ2 = 3αneJ eq 0 sdϕ/2π [13]. After Fourier transform of Eqs. (8), one can find the dispersion equation for the collec- tive modes e and s0 = R 2π x /2a, ξ = vF /Ω0 e , ω e)2 = ω − Ω0 v2 F k2 (ω − Ω0 s1 − δ2 + GΩ0 e(ω − ΩJ ) δ2 + Ωe(ω − ΩJ ) where δ = √δ1δ2. For k = 0, we get [7, 9], ω1,2(0) = (Ωe + ΩJ )/2 ± p(Ωe − ΩJ )2/4 + δ2. The anticrossing occurs when Ωe(B) = ΩJ (B). We see that the constant G drops out from ω1,2(0). In contrast, the dispersion of the collec- tive modes strongly depends on the relation be- tween G and the dimensionless parameter δ/Ωe. For δ/Ωe ≪ G (this was the case in the experi- ment [5]), the anti-crossing occurs for B < Bres, when Ωe > ΩJ (see Fig. 2a). To see this, one may consider the case δ = 0 (coupling between transverse components of electron and ion spins is turned off) as a first approximation. This ap- proximation was implicitly used in Ref. [8]. For δ = 0, there are two branches of the spectrum (dashed lines in Fig. 2a), corresponding respec- tively to the Fermi-liquid spin waves with nega- tive dispersion and the dispersionless excitations FIG. 2: (a) Anticrossing of the two collective spin modes for B < Bres; (b) Sign inversion of the group velocity of the ion mode for B > B. of the ion spins. Importantly, for B < Bres these two branches intersect each other. Turning on a finite coupling between modes, δ 6= 0, results in the anticrossing, which, for B close to Bres, occurs at the point (9) kres =p2GΩe(Ωe − ΩJ )/vF (1 + G). Remarkably, kres depends on B, so that the an- ticrossing position may be tuned by the exter- nal field. The splitting between the modes for k ≈ kres is given by ω1(kres) − ω2(kres) ≈ 2δ. As seen from Fig. 2a the upper branch of the spectrum at a certain wave vector km reaches the Stoner continuum (single-particle excitations), which is defined by inequality ω − Ω0 e ≤ vF k. For k > km the corresponding ion-type excita- tions slowly decay in time due to weak exchange coupling with the system of itinerant electrons. This decay is similar to the well-known Landau damping in plasma [12], so that the decay rate γ is calculated in a quite analogous way, yielding γ ≈ Ω0 e[(1 + G)2v2 m) + G2Ω2 J ] δ2ΩJ vFpk2 − k2 F (k2 − k2 m . (10) x ≈ 0.2) we find γmax ≈ 2 · 109s−1. As a function of k, γ has a maximum. The maximal value is given by γmax = δ2/2GΩe ≈ 3αneseq x /4Ga. Using data of Ref. [7] (ne = 0.7 · 1011cm−2, a = 80A, α = 1, 5 · 10−23eVcm3, G ≈ 0.2 seq Another interesting phenomenon arising due to simultaneous presence of two types of inter- action is a change in a sign of the group ve- locity of the ion mode. It can by understood by analyzing the spectrum in the limit k → 0, when ω1,2(k) ≈ ω1,2(0) + v2 F k2/2β. Here β = Ω0 e][(ω1,2(0)−Ωe)2+δ2]/ω1,2(0)δ2, so e[ω1,2(0)−Ω0 that the dispersions of the modes are controlled e and ω2(0) − Ω0 by signs of ω1(0) − Ω0 e, respec- tively. As seen from Fig 1, there is a critical filed B = B0, at which ω1(0) = Ω0 e. For B < B0, both spin-wave branches are below the Stoner continuum and have negative dispersions. While B increases, approaching B0, the ion spin-wave branch becomes shorter (km → 0) and disappears when B = B0. For B > B0, this branch appears again above the Stoner continuum and has a pos- itive dispersion as shown in Fig. 2b. The disper- sion of one of the modes can also change sign for Ωe > ΩJ , provided that δ/Ωe ≫ G. Above we discussed the coupled collective modes. Now we notice that Eqs. (8) have also a solution s = 0 and ¯J = 0. Importantly, the latter equation apart from the trivial solution J(r, z, t) = 0 has also non-zero solutions obey- ing the constraint R dzΨ(z)2J(r, z, t) = 0. Such solutions were called "decoupled" modes [7]. To find a number of such modes one should go be- yond continuous approximation and replace inte- gration over dz in all above equations with the sum over N atomic layers. This yields N − 1 decoupled modes, corresponding to independent solutions of the equation Pm JmΨ(zm)2 = 0, where m numerates layers [7]. All these modes are, indeed, decoupled from electron system pro- vided that one neglects the equilibrium electron exchange field acting on the ion spins. In this approximation, we find from Eq. (7) that the modes have no dispersion and their frequencies coincide and are equal to ΩJ . In fact, weak in- teraction with the electrons gives rise to a small splitting of the ion Zeeman energies, which be- come dependent on the atomic layer number m: ΩJm = ΩJ + αneseq x Ψ(zm)2. Taking into ac- count this splitting, one finds that in a symmetric quantum well, which we consider here, the de- coupled modes with antisymmetric distribution of ion spins Jm ∝ δm,m0 − δm,−m0 still obey the condition ¯J = 0, thus having no dispersion. For m0−th mode, the ion spin precession fre- quency is equal to ΩJm0. As for the modes with a symmetric distribution, they become weakly cou- pled to the electron collective mode. One can show, however, that the corresponding dispersion is very weak provided that ne/nJ ≪ 1. Symmet- ric modes also become weakly coupled to the sin- gle electron excitations and, consequently, slowly decay in the region of the Stoner continuum with m/ΩeΩJ N. To conclude, we have developed a theory of the characteristic rate δ2vFpk2 − k2 the spin waves in the 2D diluted magnetic semi- conductors. We have described analytically two collective modes corresponding to the coupled ex- citations of the electron and ion spins, and a large number of decoupled excitations of the ion spins. Our main finding is the tunable anticrossing of the collective modes. We have also predicted a field-induced change in a sign of the group veloc- ity of the ion mode and have calculated the decay of the waves in the Stoner continuum. We thank M.Vladimirova and D.Scalbert for valuable discussions. The work was supported by RFBR, by programmes of the RAS and by grant of RosNauka (project number 02.740.11.5072). [1] Semiconductor Spintronics and Quantum Com- putation, eds. D.D. Awschalom, D. Loss, and N. Samarth, in the series Nanoscience and tech- nology, eds. K. von Klitzing, H. Sakaki, and R. Wiesendanger (Springer, Berlin, 2002). [2] J. Cibert and D. Scalbert, Diluted Magnetic Semiconductors: Basic Physics and Optical Properties, in Spin Physics in Semiconduc- tors, chap. 13, ed. by M.I. Dyakonov (Berlin, Springer, 2008). [3] J.A. Gaj, R. Planel, and G. Fishman, Solid State Communication 29, 435 (1979). [4] A. Lemaitre, C. Testelin, C. Rigaux, T. Wojtow- icz, and G. Karczewski, Phys. Rev. B 62, 5059, (2000). [5] B. Jusserand, F. Perez, D.R. Richards, G. Kar- czewski, T. Wojtowicz, C. Testelin, D. Wolver- son, and J.J. Davies, Phys. Rev. Lett. 91, 086802 (2003). [6] F.J. Teran, M. Potemski, D.K. Maude, D. Plantier, A.K. Hassan, A. Sachrajda, Z. Wil- amowski, J. Jaroszynski, T. Wojtowicz, and G. Karczewski, Phys. Rev. Lett. 91, 077201 (2003). [7] M. Vladimirova, S. Cronenberger, P. Barate, D. Scalbert, F.J. Teran, A.P. Dmitriev, Phys. Rev. B 78, 081305(R) (2008). [8] F. Perez, Phys. Rev. B 79, 045306 (2009). [9] J. Konig and A.H. MacDonald Phys. Rev. Lett. 91, 077202 (2003). [10] The zero-field case was also discussed, see D. Frustaglia, J. Konig, and A.H. MacDonald, Phys. Rev. B 70, 045205 (2004). [11] P.M. Platzmann and E. Wolf, Waves and Inter- action in Solid State Plasmas (Academic, New York 1973). [12] A.I. Akhiezer, V.G. Baryakhtar and S.V. Pelet- minskii, Spin Waves (North-Holland, Amster- dam, 1968). [13] For low temperature, one can search the electron spin density in the following form s(p, r, t) = neν −1δ(ǫ − EF )s(r, ϕ, t).

Subsets and Splits