Simulating Z_2 topological insulators with cold atoms in a one-dimensional optical lattice
Feng Mei, Shi-Liang Zhu, Zhi-Ming Zhang, C. H. Oh, N. Goldman
SSimulating Z topological insulators with cold atoms in a one-dimensional optical lattice Feng Mei,
1, 2
Shi-Liang Zhu, Zhi-Ming Zhang, ∗ C. H. Oh,
2, † and N. Goldman
4, ‡ Laboratory of Photonic Information Technology, LQIT & SIPSE,South China Normal University, Guangzhou 510006, China Centre for Quantum Technologies and Department of Physics,National University of Singapore, 3 Science Drive 2, Singapore 117543, Singapore Laboratory of Quantum Information Technology and SPTE, South China Normal University, Guangzhou, China Center for Nonlinear Phenomena and Complex Systems - Universit´e Libre de Bruxelles , 231, Campus Plaine, B-1050 Brussels, Belgium (Dated: September 5, 2018)We propose an experimental scheme to simulate and detect the properties of time-reversal invariant topo-logical insulators, using cold atoms trapped in one-dimensional bichromatic optical lattices. This system isdescribed by a one-dimensional Aubry-Andre model with an additional SU(2) gauge structure, which capturesthe essential properties of a two-dimensional Z topological insulator. We demonstrate that topologically pro-tected edge states, with opposite spin orientations, can be pumped across the lattice by sweeping a laser phaseadiabatically. This process constitutes an elegant way to transfer topologically protected quantum states in ahighly controllable environment. We discuss how density measurements could provide clear signatures of thetopological phases emanating from our one-dimensional system. PACS numbers: 37.10.Jk, 73.43.-f, 67.85.Lm
The quantum Hall (QH) effect, discovered in 1980, pro-vided the first example of a quantum phase that has no sponta-neously broken symmetry. Besides, its universal character andremarkable robustness have been shown to be related to theexistence of topological invariants [1–3]. The recent discov-ery of the quantum spin Hall (QSH) effect, in materials dis-playing strong spin-orbit coupling, has opened the path for anew family of topological states: the Z topological insulators[1, 2]. Since then, the search for topological phases of matterhas become a forefront topic in condensed matter physics [3].In general, topological insulators are insulating in the bulk butthey feature gapless edge or surface states at their boundary.These edge modes are very robust and therefore persist in thepresence of impurities. The delicate control over these edgemodes has attracted considerable interest for the realization ofquantum spintronic and magnetoelectric devices [3]. Further-more, in proximity of superconductors, topological insulatorslead to non-Abelian excitations that could lead to a new archi-tecture for topological quantum computation [4].Nowadays, cold atoms trapped in optical lattices are widelyrecognized as powerful experimental tools to mimic a widerange of systems originally stemming from condensed matterphysics [5, 6]. Recently, many experimental efforts have beenfocused on the experimental realization of synthetic mag-netic fields and spin-orbit coupling for ultracold atoms [7–11],which set the stage for the simulation of topological insulatorsand fractional quantum Hall states. In particular, several pro-posals have been suggested to realize the QH and QSH statesusing these technologies [12, 13].The QH and QSH phases are realized in 2D systems sub-jected, respectively, to strong magnetic or spin-orbit cou-plings. Surprisingly, several properties associated to thesetopological states can already be probed through a one-dimensional reduction of these systems. This idea has beenexplored theoretically and experimentally in 1D quasicrystals [14], which reproduce the Hofstadter-Aubry-Andre QH model[15, 16]. This elegant discovery opens the possibility to inves-tigate QH physics using one-dimensional optical lattices [17].In this paper, we propose an experimental scheme to simu-late and detect Z topological states with cold atoms trappedin a 1D optical lattice. We show that these topological phasescan be described by a generalized 1D Harper equation with anadditional SU(2) gauge structure, which could be simulatedwith a two-component atomic gas trapped in a 1D bichromaticoptical lattice. By adjusting the corresponding laser configu-ration, one is able to probe several properties of Z topologicalstates. In particular, one could transfer the spin-resolved edgestates from one edge to the other, and measure these statesthrough density measurements. One also discusses the pos-sibility to define a Z topological invariant in this 1D frame-work, allowing to distinguish between trivial and non-trivialtopological states. Finally, we describe how density measure-ments could provide an efficient tool to measure these invari-ants in the present context.Let us start by presenting a specific 2D tight-binding model,which has been introduced in Ref. [13] to simulate a Z topo-logical insulator with two-component fermions in an opticalsquare lattice. The corresponding second-quantized Hamilto-nian reads H = t ∑ m , n c † m + , n e i θ x c m , n + c † m , n + e i θ y ( m ) c m , n + h . c + λ stag ∑ m , n ( − ) m c † m , n c m , n , (1)where c m , n is a two-component (spin 1/2) field operator de-fined on the lattice site ( x = ma , y = na ), a = t is the nearest-neighbor hopping amplitude.Here, the spin-1/2 structure derives from the fact that eachsite hosts atoms in two internal states [11]. The second lineof Eq. (1) describes an on-site staggered potential with am-plitude λ stag , along the x direction, which has been introduced a r X i v : . [ c ond - m a t . qu a n t - g a s ] J a n to drive transitions between different topological phases. ThePeierls phases θ x and θ y ( m ) , which accompany the hoppingalong the x and y directions, are engineered within this tight-binding model to simulate the analog of spin-orbit couplings,and are expressed in terms of the Pauli matrices σ x , y , z . Thespace-dependent operator θ y ( m ) = π m ασ z reproduces theeffect of the intrinsic spin-orbit coupling [18]: it correspondsto opposite “magnetic” fluxes ± α for each spin componentand generates QSH phases. The constant operator θ x = πγσ x corresponds to a spin-mixing perturbation, and thus simulatesa Rashba spin-orbit coupling term [18].The Hamiltonian (1) being translationally invariant alongthe y direction, one can explore its properties by imposing pe-riodic boundary conditions along this direction. Consideringthis cylindrical geometry, the single-particle wave function isexpressed as ψ ( m , n ) = exp ( ik y n ) Ψ ( m ) , where k y is the quasi-momentum along the periodic coordinate, and where the two-component wave function Ψ m = (cid:16) Ψ ↑ m , Ψ ↓ m (cid:17) T satisfies a gen-eralized 1D Harper equation [15, 16] E Ψ m ( k y ) = t ( e i θ x Ψ m + ( k y ) + e − i θ x Ψ m − ( k y ))+ R ( m , k y ) Ψ m ( k y ) . (2)We have introduced the onsite 2 × R ( m , k y ) = t diag ( cos ( πα m + k y ) , cos ( πα m − k y ))+ λ stag ( − ) m I , (3)where I is the identity matrix. The Harper Eq. (2) there-fore describes the dimensional reduction of the initial system(1), but still captures the essential properties of its topologicalphases. The energy spectrum E = E ( k y ) obtained by solvingEq. (2) displays several energy bands that describe the bulk,but also several gapless states which constitute clear signa-tures of QSH topological phases [13, 18]. We note that Eq. (2)generalizes the spinless Harper equation obtained by Aubry-Andre, in their study of Anderson localization [16]. Recently,the Aubry-Andre model has been simulated in a 1D bichro-matic optical lattice [19], which motivates us to propose ageneralization of their setup to investigate the physics stem-ming from our spin-1/2 Eq. (2).Let us first focus on the case γ = λ stag =
0, where thespin-mixing perturbation θ x and the staggered potential areabsent. The realization and implication of these terms willbe discussed at the end of this work. In this case, the spin-1/2 model described by Eq. (2) can be realized by trappinga two-component atomic gas in a primary 1D optical lattice V ( x ) = V cos ( k x ) , with wave number k . In this configu-ration, and for a sufficiently deep potential, the atomic systemis governed by the tight-binding Hamiltonian H = t ∑ m c † m + c m + c † m − c m , (4)where c m is a two-component (spin 1/2) field operator definedon the lattice site ( x = ma ) and m = , . . . , L . Then, the on-site term characterized by the 2 × R ( m ) can be real-ized by two weak state-dependent lattices, with wave number k , which act independently on the two atomic states [20].For our purpose, the optical potentials associated to thesestate-dependent lattices have the form V ↑ , ↓ ( x ) = V s cos ( k x ± φ / ) , which can be produced by two counterpropagating laserbeams with linear polarization vectors forming an angle φ [5, 21, 22]. These additional lattices interfere with the pri-mary lattice, and supposing that V s (cid:28) V , simply lead to theonsite perturbation term [19, 23, 24] H = Λ ∑ m c † m ↑ c m ↑ cos ( πβ m + φ ) + c † m ↓ c m ↓ cos ( πβ m − φ ) , (5)where Λ ∼ t (cid:28) V , β = k / k . Although the phase φ couldbe affected by an overall shift of V ( x ) with respect to V ↑ , ↓ ( x ) ,this parameter can be monitored through various technics (cf.the experimental methods in Ref. [24], but also the studiesof Refs. [25] on the effects of uncontrolled phases in cold-atom Aubry-Andre models). Besides, we note that the topo-logical properties described in this work remain constant forsmall variations of the parameter φ (cf. below). In this con-figuration, the single-particle equation associated to the totalHamiltonian H tot = H + H reads E Ψ m ( φ ) = t ( Ψ m + ( φ ) + Ψ m − ( φ )) + S ( m , φ ) Ψ m ( φ ) , (6)where S ( m , φ ) = Λ diag ( cos ( πβ m + φ ) , cos ( πβ m − φ )) .Therefore, a direct mapping from this 1D Aubry-Andre sys-tem to the 2D setup of Ref. [13] is obtained by associ-ating the commensurability parameter β to the “magnetic”flux α , the potential strength Λ to twice the tunneling am-plitude 2 t , and the phase φ to the quasimomentum k y . In otherwords, our spin-1/2 generalization of the Aubry-Andre model,which could be simulated using 1D state-dependent lattices[19], could already reveal the topological properties emanat-ing from Eqs. (1)-(2). It is worth emphasizing that, altougha direct mapping exists between the energy spectra E ( k y ) and E ( φ ) , that respectively correspond to the 2D model of Ref.[13] and the present Aubry-Andre-type model, the parameter φ is fixed in the latter experimental scheme. Therefore, theenergy “bands” depicted by the spectrum E ( φ ) involves the union of different configurations of the system, obtained bycontinuously varying the parameter φ .Let us investigate the spectral properties of Eq. (6): fora fixed value of the parameters φ and Λ (in the following Λ = t = φ ∈ [ , π ] ), and for a rational value of thecommensurability parameter β = p / q , the spectrum splits into q continua of states (cf. Fig. 1 (a)). These states are delocal-ized and describe the bulk of our 1D system. In the exampleillustrated in Fig. 1 (a), one has set β = /
3, which leads tothree “bulk subbands”. Between these continua of bulk states,and within certain ranges of the parameter φ , one finds twodegenerate states with opposite spin, whose amplitudes arelocalized at the two edges of the system (cf. the states in Fig.1 (a), which are highlighted by a dot (resp. a star) for φ = φ = π − φ =
1, one finds a spin-up (resp. down) state at E ≈ − t , which (cid:113) m (cid:115) (cid:115) (cid:115) (cid:108) (m) (cid:113) =1 (cid:113) = (cid:47) (cid:113) = 2 (cid:47) -1(a) (b)(c) Figure 1: (Color online) (a) Energy spectrum as a function of thephase φ for a 1D lattice with L =
38 sites. Here β = / Λ = t .(b) The amplitudes | Ψ m ↑ | (resp. | Ψ m ↓ | ) are represented in blue(resp red) as a function of the site index m , and correspond to thethree states highlighted in (a) at φ = , π , π −
1. (c) The spin-resolved particle densities ρ ↑ ( m ) and ρ ↓ ( m ) , depicted in blue andred respectively, for E Fermi = − t and φ = , π , π −
1. These den-sities have been computed for infinitely sharp boundaries. Note theinversion of the spin structure at the opposite edges, as φ is varied. is localized at m = L (resp. m = m = m = L ), occurs at thesame energy by setting φ = π − helical edge states, which is a hallmark of the QSHeffect [18]: when the Fermi energy is set in the first bulk gap E Fermi ≈ − t , each edge is populated by opposite spins travel-ing in opposite directions. In contrast, in the present contextof a 1D lattice, the parameter φ is fixed, and therefore, eachedge is populated by a single spin species. However, by adi-abatically varying the phase φ between [ , π ] , following agapless edge state within the lowest “bulk gap”, one drives aninteresting transition between a non-trivial edge-state config-uration (e.g. spin up at m = m = L ) to theopposite configuration (e.g. spin up at m = L and spin downat m = spin pumping . We note thatthe edge states remain localized during the whole process, ex-cept at singular points (e.g. φ = π in the example presentedin Fig. 1), where these states connect to the bulk. Therefore,the edge states survive for small variations of the parameter φ , a fact which is in agreement with the topological argumentdiscussed below.In standard two-dimensional QSH systems, the existenceof gapless helical edge states inside a bulk gap is guaranteedby a topological invariant, the so-called Z index ( ν = , ν =
1, an odd number of helical edge state pairsare located at each edge, in which case the system realizesthe QSH effect. As long as the bulk gap remains open, theindex ν remains constant, which guarantees the robustness ofthe edge states against external perturbations. Now, if onemaps the example presented in Fig. 1 (a)-(b) to the analogous2D system (i.e. φ → k y , where k y takes all the values withinthe range k y ∈ [ , π ] ), one observes that each bulk gap of thespectrum E ( k y ) hosts two pairs of helical edge states (i.e. onepair for each edge). Therefore, the bulk gaps presented inFig. 1 would both correspond to the Z index ν = k y → φ , with φ fixed), this result indicatesthat, as long as the bulk gap remains open, there will alwaysbe a range φ ∈ [ φ , φ ] between which such edge states will bedetected.Obviously, this topological argument is based on theanalogy between our 1D system and its analogous two-dimensional QSH system, whose dimensionality allows toproperly define the topological invariant ν [18]. However,one can show that this topological invariant can also be rig-orously defined in the 1D framework. Indeed, when γ = Z index is simply related to the spin Chern number ν = SChNmod 2, where SChN = (
ChN ↑ − ChN ↓ ) / ↑ , ↓ are the Chern numbers associated to the up and downspin respectively [26]. When γ =
0, the spin components aredecoupled and the Chern numbers ChN ↑ , ↓ can be evaluatedindividually from the standard TKNN expression [27]ChN ↑ = π i ∑ E λ ≤ E Fermi (cid:90) T dk x d φ F ( | Ψ ↑ λ ( k x , φ ) (cid:105) ) , (7)where F is the Berry curvature associated to the single-particle state | Ψ ↑ λ ( k x , φ ) (cid:105) , which is caracterized by the bandindex λ situated below the Fermi energy E Fermi , and where k x is the quasimomentum. In this expression, one supposesthat the parameter φ evolves continuously along the interval [ , π ] : namely, the definition of the Chern number (7) re-quires the union of all the Hamiltonian operators H ( φ ) . Infact, it was recently shown that such a Chern number couldbe rigorously defined for each φ [14], and that it remains con-stant for all φ ∈ [ , π ] . This result, which is in agreementwith the argument based on the 2D analogy (cf. above), guar-antees the existence of edge states for certain ranges of theparameter φ [14]. Finally, we stress that for the general situ-ation, where spin-mixing is present γ (cid:54) =
0, the topological Z index and the SChN are no longer expressed in terms of theindividual Chern numbers ChN ↑ , ↓ . However, their values ob-tained in the limit γ → γ , as longas their associated bulk gap remains open [26]. We note that,for the example illustrated in Fig. 1 (a), the Chern numbersare given by ChN ↑ , ↓ = ± ↑ , ↓ = ∓ Z index ν = φ , does not necessarily (cid:113) (cid:115) (cid:115) (cid:115) (cid:115) Figure 2: (Color online) (a) Energy spectrum as a function of thephase φ for a 1D lattice with L =
41 sites. Here β = / Λ = t .(b) The amplitudes | Ψ m ↑ | (resp. | Ψ m ↓ | ) are represented in blue(resp. red) as a function of the site index m , and correspond to thefour states highlighted in (a), i.e. the states labeled by a dot, a rect-angle, a star and a hexagon respectively. mean that the Z index is non-trivial ( ν = φ ∈ [ , π ] to determine whether the number of such helicaledge state pairs is odd ( ν =
1) or even ( ν =
0) at each edge.This counting procedure, which could be performed experi-mentally by continuously varying the phase φ , allows to rig-orously classify the phases of our 1D model in terms of the Z topological index. To illustrate a configuration displayingtrivial and non-trivial Z phases, one has computed the energyspectrum and edge state structures for the case β = / Z index ν =
1, as for the previous example discussedabove for β = /
3. However, in the second bulk gap around E ≈ − t , one would observe two helical edge state pairs at eachedge, by varying φ ∈ [ , π ] (cf. Fig. 2 (b)): the second gapis therefore associated to the trivial phase ν =
0. This latterresult is in agreement with the value SChN =
2, which can becomputed from Eq. (7) for E Fermi = − t and β = / Z topological phasescould already be obtained from density measurements. Thefirst strategy would be to directly detect the edge states, whichis a realistic task for systems featuring infinitely sharp bound-aries and low Fermi energy. Indeed, in this configuration,the edge states will contribute to the particle density in a de-tectable way. We illustrate in Fig. 1 (c) the spin densities,defined by ρ ↑ , ↓ ( m ) = ∑ E λ ≤ E Fermi | Ψ ↑ , ↓ , λ ( m ) | , (8)for φ = , π and 2 π − E Fermi = − t . The wave-functions Ψ ↑ , ↓ , λ ( m ) are computed from a numerical diagonalization ofEq. (6). One clearly observes sharp peaks in the spin densi-ties, which correspond to opposite spins at the two edges. As φ is progressively varied from φ = π = π − t by laser-assisted tunnel-ing methods, and create synthetic walls within the confinedsystem by abruptly changing the tunneling amplitude in thecentral region [13]. Then, the sharp peaks illustrated in Fig.1 (c) and corresponding to the edge states will be observed atthese synthetic walls. Consequently, by varying the parameter φ , and performing in situ spin-resolved density measurements[28], one could directly detect the spin-pumping process illus-trated in Fig. 1 (b)-(c).Another method would be to measure the spin Chern num-ber SChN, which could also be evaluated from density mea-surements. This method is based on the fact that the Chernnumber of individual spin species, ChN ↑ , ↓ , can be computedfrom the density through the Streda formula [29]ChN ↑ ( E Fermi ) = ∆ ρ ↑ ∆ β , (9)Here, the (local) Fermi energy is supposed to lie in a bulkgap, which is associated to a plateau in the density profiles,and ∆ β = β − β (cid:48) , ∆ ρ ↑ = ρ ↑ ( β ) − ρ ↑ ( β (cid:48) ) . This equation ex-presses the fact that the Chern number associated to a bulkgap can be evaluated by comparing the density plateaus ob-tained from two configurations of the system (i.e. with β and β (cid:48) ). In Fig. 3, one shows the density profiles for β = / β (cid:48) = /
4, in a system confined by a harmonic poten-tial V trap ( m ) = V conf × ( m − c ) , where c = L /
2. Consider-ing a local-density approximation, one defines the Fermi en-ergy locally as E fermi ( m ) = E − V trap ( m ) : in this regime,where the confining potential is considered to vary smoothly,the density profiles depict several plateaus that correspond tothe bulk gaps located below the chemical potential E . InFig. 3, the Fermi energy is set at E =
0, thus the densityplateaus correspond to the lowest bulk gap of Fig. 1 (a). Theformula (9) yields ChN ↑ ( ) =
1, and therefore ν = φ , which highlights the factthat the topological invariants (i.e. ChN ↑ , ↓ , SChN and ν ) canindeed be defined at each value of the parameter φ (cf. dis-cussion above and Ref. [14]). Consequently, a density mea-surement at φ fixed allows to directly determine the Z classof our 1D system in the presence of an external confining trap.Finally, we note that this detection scheme is also suited forfinite spin-mixing perturbations, i.e. γ (cid:54) =
0, as long as the bulkgaps remain open.It was shown in Ref. [13], that the combination of a spin-mixing perturbation γ (cid:54) = λ stag (cid:54) = Z phases. These transitions occur in-dividually in the different bulk gaps and can be obtained bysolving the Harper Eq. (2). Exploring these topological phasetransitions with our 1D model requires to engineer the Peierlsoperator θ x as well as the staggered potential, which both actalong the x direction (i.e. the direction of our 1D system). Therealization of the hopping operator θ x demands to control thetunneling in a spin-dependent manner, which can be achieved (cid:108) (m)/ (cid:54)(cid:96) Figure 3: Spin-up density ρ ↑ ( m ) / ∆ β for a 1D lattice with L = β = / β (cid:48) = / ∆ β = β − β (cid:48) , the Fermi energy E = φ = π /
2. The sys-tem is confined by a harmonic trap V conf ( m − c ) , with V conf = . t .Comparing the two plateaus indicates that ChN ↑ = Z index ν = with several Raman transitions that act individually on thetwo atomic internal states (cf. Refs. [11, 13]). On the otherhand, the staggered potential could be easily produced by aweak lattice V stag ∼ t , with wave number k stag = k /
2. Usingthis configuration, one could directly detect topological phasetransitions by varying the staggered potential strength and per-forming spin-resolved density measurements, since the latterprovide sufficient informations to classify our system in termsof the Z index ν (cf. above). For example, the trivial phase ν = β = / E Fermi = − t and λ stag = ν = λ stag > . t (cf. Ref. [13]).In summary, we have proposed an experiment scheme tosimulate Z topological phases with cold atoms trapped in a1D bichromatic optical lattice. Our scheme is based on the di-mensional reduction of a 2D model exhibiting Z topologicalinsulating phases, and which captures its essential properties.Our 1D atomic system is described by a generalized Harperequation, which can be simulated by a spin-1/2 generaliza-tion of the Aubry-Andre system recently realized with coldatoms [19]. The latter can be practically engineered using thecurrent technology offered by optical lattices, exploiting theinterferences of a primary lattice with weak state-dependentlattices. Interestingly, our simple scheme is able to trans-fer spin-resolved edge states, with opposite spin components,from one edge to the other. This manipulation, which can beeasily performed by varying the secondary lattices configura-tion [22], constitutes an elegant manner for transporting topo-logically protected quantum states [30]. Besides, we have dis-cussed the possibility to drive topological phase transitions,by varying an additional staggered potential. Furthermore, wehave shown that the spin Chern number, which allows to clas- sify the topological phases of our 1D system, can be eval-uated using spin-resolved atomic density measurements. Inthe presence of sharp boundaries, we have shown that spin-resolved density profiles would already present clear signa-tures of edge states, with opposite spin components at the twoedges. Therefore, the Z phases emanating from our 1D sys-tem could be probed with the current technologies, such as in situ imaging techniques. Let us mention that additionalsignatures could be obtained through cyclotron-Bloch dynam-ics [31]. We note that state-dependent lattices generally leadto large spontaneous emission rates for fermionic species, adrawback which could be avoided by considering an atom-chip realization of our model [13]. Finally, we stress thatour dimensional reduction approach may be generalized toexplore three-dimensional topological phases [32] using 2Doptical lattice.The authors thank J.K. Pachos and A. Bermudez for manyhelpful discussions and comments. NG thanks FRS-FNRSfor financial support. This work was supported by theNSFC (Nos. 60978009, 10974059, and 11125417), theMajor Research Plan of the NSFC ( No. 91121023), theSKPBR of China (Nos.2009CB929604?2011CB922104 and2011CBA00200), and NUS Academic Research (No.WBS:R-710-000-008-271). ∗ Electronic address: zmzhangATscnu.edu.cn † Electronic address: phyohchATnus.edu.sg ‡ Electronic address: ngoldmanATulb.ac.be[1] M.Z. Hasan and C.L. Kane, Rev. Mod. Phys. , 3045 (2010).[2] X.L. Qi and S.C. Zhang, Rev. Mod. Phys. , 1057 (2011).[3] J. E. Moore, Nature (London) , 194 (2010).[4] C. Nayak et al. , Rev. Mod. Phys. , 1083 (2008).[5] I. Bloch, J. Dalibard, and W. Zwerger, Rev. Mod. Phys. , 885(2008).[6] M. Lewenstein et al. , Adv. Phys. , 243 (2007).[7] J. Dalibard, F. Gerbier, G. Juzeliunas, and P. Ohberg, Rev. Mod.Phys.83,1523 (2011);[8] Y.J. Lin et al. , Phys. Rev. Lett. , 130401 (2009); Nature(London) , 628 (2009).[9] M. Aidelsburger et al. , Phys. Rev. Lett. et al. , Nature (London) , 83 (2011).[11] D. Jaksch and P. Zoller, New J. Phys. , 56 (2003); K. Oster-loh et al. , Phys. Rev. Lett. , 010403 (2005); F. Gerbier and J.Dalibard, New J. Phys. , 033007 (2010); N. Goldman et al. ,Phys. Rev. Lett. , 035301 (2009);S.L. Zhu et al. , Phys. Rev.Lett. 97, 240401 (2006).[12] N. Goldman and P. Gaspard, Europhys. Lett. et al. , Phys. Rev. Lett. , 246810 (2008); C. Wu,Phys. Rev. Lett. , 186807 (2008); T.D. Stanescu et al. , Phys.Rev. A , 053639 (2009); X.-J. Liu et al. , Phys. Rev. A ,033622 (2010); E. Alba et al. , Phys. Rev. Lett. , 235301(2011);T.D. Stanescu et al. , Phys. Rev. A , 013608 (2010); N.Goldman, W. Beugeling and C. Morais Smith, arXiv:1104.0643(to appear in Europhys. Lett.). B. Beri and N. Cooper, Phys.Rev. Lett. et al. , Phys. Rev. Lett. , 255302 (2010).[14] Y.E. Kraus et al. , arXiv:1109.5983v2. [15] D.R. Hofstadter, Phys. Rev. B , 2239 (1976).[16] S. Aubry and G. Andre, Ann. Isr. Phys. Soc. , 133 (1980).[17] L.J. Lang, X.M. Cai and S. Chen, arXiv:1110.6120v1.[18] C. L. Kane and E. J. Mele, Phys. Rev. Lett. , 146802 (2005); ibid et al. , Nature (London) , 895 (2008); B.Deissler etal. , Nat. Phys. , 354 (2010); E. Lucioni et al. , Phys. Rev. Lett. et al. , Phys. Rev. A, etal. New J. Phys.
107 (2007).[24] T. Salger, C. Geckeler, S. Kling, and M. Weitz, Phys. Rev. Lett. , 053606 (2009); PRA , 013624(2011). [26] D.N. Sheng, Z.Y. Weng, L. Sheng, and F.D. M. Haldane, Phys.Rev. Lett. , 036808 (2006); T. Fukui and Y. Hatsugai, Phys.Rev. B , 121403(R) (2007).[27] D. J. Thouless et al. , Phys. Rev. Lett. , 405 (1982); M.Kohmoto, Ann. Phys. , 343 (1985).[28] W. S. Bakr et al. , Science , 547 (2010); J. F. Sherson et al. ,Nature , 68 (2010); C. Weitenberg et al. , Nature , 319(2011).[29] R.O. Umucalılar et al ., Phys. Rev. Lett. , 070402 (2008); ABermudez et al ., New J. Phys. , 033041 (2010); F. Mei et al .,Phys. Rev. A , 023622 (2011); S.L. Zhu et al. , Phys. Rev. Lett98, 260402 (2007).[30] N.Y. Yao et al. , arXiv:1110.3788.[31] A. R. Kolovsky, arXiv:1106.0945v1.[32] A. Bermudez et al. , Phys. Rev. Lett. , 190404 (2010); L.Mazza et al.et al.