You are on page 1of 10

PHYSICAL REVIEW B 95, 035414 (2017

)

Frequency-dependent magneto-optical conductivity in the generalized α-T3 model
Áron Dániel Kovács,1 Gyula Dávid,2 Balázs Dóra,3 and József Cserti1
1
Department of Physics of Complex Systems, Eötvös University, H-1117 Budapest, Pázmány Péter sétány 1/A, Hungary
2
Department of Atomic Physics, Eötvös University, H-1117 Budapest, Pázmány Péter sétány 1/A, Hungary
3
Department of Theoretical Physics and BME-MTA Exotic Quantum Phases Research Group,
Budapest University of Technology and Economics, Budapest, Hungary
(Received 31 May 2016; revised manuscript received 12 September 2016; published 12 January 2017)

We have studied a generalized three-band crossing model in 2D, the generalized α-T3 lattice, ranging from the
pseudospin-1 Dirac equation through a quadratic+flat band touching to the pseudospin-1/2 Dirac equation. A
general method is presented to determine the operator form of the Green’s function, being gauge and representation
independent. This yields the Landau level structure in a quantizing magnetic field and the longitudinal and
transversal magneto-optical conductivities of the underlying system. Although the magneto-optical selection
rules allow for many transitions between Landau levels, the dominant one stems from exciting a particle from/to
the flat band to/from a propagating band. The Hall conductivity from each valley is rational (not quantized at
all), in agreement with Berry phase considerations, though their sum is always integer quantized.

DOI: 10.1103/PhysRevB.95.035414

I. INTRODUCTION and in comparison to Eq. (1), many more new terms can be
added to this and masses can be opened in several distinct
Since the first isolation of graphene [1] in 2004 and the
ways [10]. As detailed below, Eq. (2) can be realized in the dice
theoretical prediction and experimental realization of topolog-
or T3 lattice, composed of two 2D honeycomb lattices, which
ical insulators [2,3], the Dirac equation and its variants have
share one sublattice and is sketched in Fig. 1. Experimentally,
started to attract almost unprecedented attention in condensed
the dice lattice can be realized from a trilayer structure of the
matter and related fields. The peculiar spinor structure of the
face-centred cubic lattice, grown in the [111] direction [10].
Dirac equation, which, e.g., stems from the two sublattices
Recently, a novel variant of the T3 lattice structure was
of the 2D honeycomb lattice in graphene, gives rise to many
proposed, coined as the α-T3 model, suggested first by Raoux
topology related phenomena such as a Berry phase [1] of π ,
et al. [12]. Due to the three nonequivalent lattice sites of
unusual Landau quantization in a magnetic field and the related
the T3 lattice, two nearest-neighbor hopping integrals are
unconventional quantum Hall effect [4], just to mention a few
possible, which, however, need not be equal to each other. The
immediate consequences.
generalized α-T3 model is described alternatively by a lattice
The 2D massless Dirac equation possesses the deceivingly
consisting of three layers of triangular lattices with basis atoms
simple form as
A, B, and C and with only intersublattice hoppings between
  adjacent layers shown in Fig. 1. By tuning a parameter α, which
0 p−
HS=1/2 = vF S · p = vF , (1) measures the relative strength of the two hopping integrals,
p+ 0
one can interpolate continuously from the S = 1/2 case with a
where vF is the Fermi velocity of the underlying system and completely decoupled flat band to the perfect S = 1 situation,
plays the role of the effective speed of light, p = (px ,py ) i.e., from the physics of Eq. (1) to Eq. (2). The three-band
is the 2D momentum, p± = px ± ipy and S stands for the tight-binding Hamiltonian in the basis A, B, and C is given
spin-1/2 Pauli matrices, which represent the sublattice degree by [7,10,13,14]
of freedom in this instance. Shortly after the discovery of ⎛ ⎞
graphene, this equation was generalized, still in 2D, to arbitrary 0 t1 f (k) 0
⎜ ⎟
pseudospin-S, known as the Dirac-Weyl equation with S now Hdice = ⎝t1 f ∗ (k) 0 t2 f (k)⎠, (3)
representing the (2S + 1) × (2S + 1) matrix representations
0 t2 f ∗ (k) 0
of the SU (2) algebra, and several lattices have been proposed,
hosting these Weyl fermions [5–9].
Similarly to other spin-S problems, cases with integer and where t1 and t2 are the hopping amplitudes between adjacent
half-integer spin differ from each other. The ensuing spectrum triangular lattice, and it has further been generalized by
consists of coaxial Dirac cones, crossing each other at the same adding [11,14] an on-site energy term 0 in the middle
Dirac point, and for integer spins, an additional dispersionless layer, arising from, e.g., a real
√ chemical potential, while
flat band also shows up and crosses the Dirac point. f (k) = 1 + 2 exp(i3ky a/2) cos( 3kx a/2) with k = (kx ,ky )
The simplest integer spin case is the pseudospin-1 Weyl and a is the nearest-neighbor distance in the dice lattice
equation. It has a 3 × 3 matrix structure as (the distance between sites A and B), and ∗ denotes the
⎡ ⎤ complex conjugation. In the original formulation of the model,
0 p− 0 t2 /t1 = tan(φ) = α was used for parametrization, but we
⎢ ⎥ prefer to use the two hopping amplitudes instead. Although
HS=1 = vF ⎣p+ 0 p− ⎦, (2)
the spectrum itself is independent from α, the Berry phase
0 p+ 0 depends continuously on it.

2469-9950/2017/95(3)/035414(10) 035414-1 ©2017 American Physical Society

Recently. an additional flat band appear. as sketched in Fig. as follows from Eq. DÁVID. After determining the spectrum we present 3a ⎜ ⎟ HK = ⎝t1 k+ 0 t2 k − ⎠ . 2). this is not the case.e. which possesses two touching parabola. the graphene) and contains a as bilayer graphene. The pseudospin-1 Dirac-Weyl equation also describes the Linearizing √ the function f (k) around the K = low-energy excitations in a Lieb lattice.. the Hamiltonian (3) for K  valley C can be obtained by a unitary transformation with matrix U and a replacement of the parameters as ⎛ ⎞ B 0 0 1 ⎜ ⎟ U = ⎝0 1 0⎠ and (t1 . (4). There are three atoms A. and has been realized (2π/3 3a. depending on the sign of the well as the pseudospin-1 Dirac-Weyl equation within the same local on-site energy term. Then. AND CSERTI PHYSICAL REVIEW B 95. II. (2) (see Refs. DÓRA. The on-site energy of the sixfold easily calculate any operator or Green’s function in valley K  connected site B is 0 . 035414 (2017) Similarly. (5) (see Fig. l√ B  2 (x + iy ) we have [a. 2.KOVÁCS.2π/3a) point in the Brillouin zone we have using photonic waveguides [16–18]. tuning also the hopping For 0 = 0. as was emphasized in Ref. − t1 ). and C in each from those in valley K. touching the bottom or the top this allows us to study single and bilayer graphene physics as of one of the parabolic bands [14]. we study the effect of quantizing magnetic 0 t1 k − 0 field on Eq. 1. Using the above U . where k± = kx ± iky . completely detached flat band. and A is the vector e|B| potential such that B = ∇ × A. the strength of the coupling between the touching parabolic bands. [14]. [12. and parabolic and flat band can be continuously tuned. 2). 2.2 . 2. replacing the canonical momentum by a gauge-invariant quantity k →  = k + |e|A one finds  commutation relation [x . on the other hand. (1) (i. (11). the DC Hall f (K + k) ≈ (3a/2)(kx − iky ).e. and results equation reduces to the pseudospin-1 Dirac-Weyl model of in a parabolic band touching a flat band. By introducing the bosonic creation-annihilation operators â = l√B 2 (x − iy ) and â † = FIG. In turn. separated by a band gap of size |0 |. as we do so later on in Eq.t2 ) → (−t2 . one can the red dashed and blue solid lines. position or momentum). which is independent from the 0 t2 k + 0 chosen gauge or representation (i. The touching other way round) then it corresponds to the pseudospin-1/2 parabolic and flat band realizes essentially the same physics Dirac equation of Eq. and an additional Eq. As a standard procedure. (6) 1 0 0 A y x Note that in case of t1 = t2 and 0 = 0 the contribution from the two valleys are identical. 2 4 √ where vF = 3a t12 + t22 /2. there are two special cases for this generalized model: (i) for t1 = t2 . unit cell. where lB = 2 2 the  is the magnetic length scale. which is based on the operator form of the 2 Green’s function of the system. The eigenenergies are To demonstrate the versatility of our method. The energy dispersion is sketched in the low-energy limit of the generalized α-T3 model.15]) and (ii) for t2 = 0 and t1 = 0 (or the decoupled parabolic band. however.y ] = −i / lB . (5) pseudospin-1 case with ease. one needs to calculate the Landau levels (LLs). When the on-site energy 0 = 0 (Fig. the model contains two integrals t1.a † ] = 1.. if either of these conditions FIG. B. Eq. THE OPERATOR OF THE GREEN’S FUNCTION FOR THE GENERALIZED α-T3 MODEL To obtain the magneto-optical conductivity tensor σ (ω) in magnetic field perpendicular to the plane of the dice lattice. The resulting dispersion relation is plotted in Fig. E± (k) = ± + vF2 k 2 . (4) a novel method. Then. and the Hamiltonian in 035414-2 . model. this The 0 term changes the spectrum of the model. ⎛ ⎞ In this paper. around the K = −K point we have f (K  + k) ≈ −(3a/2)(kx + iky ) and thus. we calculate the  magneto-optical response of the generalized α-T3 lattice. (5) for 0 = 0 (left) and 0 > 0. the linearized form of response and the optical conductivity without magnetic field the Hamiltonian (3) for low-energy states (around the K point) of the α-T3 lattice were studied in Ref. [15] without the local reads on-site energy term 0 . and reproduce known results along the way for graphene and the 0 02 E0 (k) = 0. The dice lattice with t1 and t2 hopping amplitude along are not met.

FREQUENCY-DEPENDENT MAGNETO-OPTICAL .N̂ ) β1 β2 z â 2 fK (z. namely the operator form of the valley but can be obtained from the above eigenvalues by the Green’s function gives an elegant way to calculate the trace following replacement (β1 .3 are coefficients to be Finally. ωk = (2k + 1)π kB T are the fermionic Matsubara fre- valley can be obtained by the transformation (6). − β1 ). T is the emphasize that fK (z. 035414 (2017) Eq. ⎛ ⎞ ⎛ ⎞ tivity tensor in the bubble approximation can be obtained from 0 β1 0 0 −β1 0 the operator of the Green’s function given by Eq. . we found for the K valley (for details conductivity. 035414-3 . operator form of the Green’s function which is independent To show how effective our method is in this work. Finally. We easily carried out the whole calculation for (z − H )−1 . [19.N̂) = z2 − 0 z − β12 N̂ − β22 (N̂ + 1) .Cζ. We should quencies (here kB is the Boltzmann constant. Now 0 β1 â 0 ⎜ ⎟ an accepted procedure is to use the position representation of H = ⎝β1 â † 0 β2 â ⎠. In contrast to the usual way where the Green’s graphene using our method and found the same results function is given in position representation. Indeed. . The sum over the fermionic Matsubara frequencies ωk in (10b) can be performed by the usual III.2. However. (8) the Fourier transform of the translation invariant part of the Green’s function for single and bilayer graphene and presented where |n is an eigenstate of the number operator N̂ = â † â a rather lengthy and complicated derivation to obtain the trace with n = 0.2. The trace poles of the Green’s function we find the same Landau levels can be obtained using the eigenstates of the Landau levels that are given in Appendix A.β2 ) → (−β2 . . ξ ∞ Then. jy = √ ⎝β1 0 −β2⎠. PHYSICAL REVIEW B 95. given in Appendix A. (4) becomes of a trace of the product of the Green’s function and other ⎛ ⎞ operators [in this work see Eq. ± 1. while the band index is denoted by in the expression of the magneto-optical conductivity tensor. αβ is the current-current correlation function (α.3 |n + 1)T .N̂ ) 1 z I + β 2 2 N̂ fK (z.i with i = 1. As we demonstrate below in contrast in Appendix A. (10a) while at the K  valley it is given by the transformation (6). (10b). where × jβ G(iωk )). The Landau levels are different at the K  to this approach our results.N̂ + 1) β1 âfK (z. using only the usual algebra of the creation and annihilation Now.N̂ − 1) ⎜ † ⎟ GK (z) = ⎝ β1 â fK (z.20] and bilayer graphene [21] in homogeneous Inspecting the Hamiltonian we assume that the eigenstate magnetic field to calculate the magneto-optical conductivity. generalized Laguerre polynomials requires further efforts to ζ The Landau levels En and the corresponding states are given obtain analytical results. this approach involves com- † 0 β2 â 0 plicated analytical calculations.ζ  = (Cζ. (9) in the elB ⎜ ⎟ ielB ⎜ ⎟ jx = √ ⎝β1 0 β2⎠. ζ = 0. We would like to emphasize that calculate the magneto-optical conductivity tensor for the the operator form of the Green’s function provides a great generalized α-T3 model. After a lengthy but straightforward involving the Green’s function such as the magneto-optical analytical calculation. . Usually.β = operator. . such quantities are expressed in terms see Appendix B): ⎛1  ⎞ z I + β12 (N̂ + 1)fK (z. and I is the identity Here. graphene [19. (10b) as an example]. MAGNETO-OPTICAL CONDUCTIVITY summation method [23].1 |n − 1.N̂ ) β2 âfK (z. the evaluation of this trace including integrals of the determined from the eigenvalue problem of Hamiltonian (7). To this end. . (10a) using the where αβ (iνm ) = Tr(jα G(iωk + iνm ) 2π lB2 k=−∞ usual analytic continuation [23] iνm → ω + iη in the current- current correlation function αβ (iν) given by Eq.2 |n. is of the form Using the Schwinger proper-time method [22] they derive |n.y). The operator of the Green’s function for the K  x.N̂ − 1)  −1 where fK (z. we give the presented in Refs. (9a) β1 β2 †2 †   z â fK (z.N̂ − 1) ⎠. (7) the Green’s function. we of any representation. (10b) η is the inverse life time of the particle. we need the operator simplification in the calculation of different physical quantities of the Green’s function. and Cζ.1. the frequency-dependent magneto-optical conduc- ikB T  tivity tensor σ (ω) can be calculated from Eq.Cζ. (11) following way: 2 2 0 β2 0 0 β2 0 αβ (ξ ) − αβ (0) σαβ (ξ ) = .N̂) is an operator but can easily be temperature and k is an integer) and νm = 2mπ kB T are calculated in the Fock representation. for example Gusynin √ √ and Sharapov recently have used the position representation where β1 = (3c/ 2) t1 / lB and β2 = (3c/ 2) t2 / lB are the of the proper-time expression for the electron propagator for rescaled hopping elements t1 and t2 . Note that studying the bosonic Matsubara frequencies (m is an integer). (9b) while N̂ = â † â is the number operator.N̂ + 1) β2 â fK (z. respectively. we derive the Green’s function defined by G(z) = operators. the current density operator j = e ∂H ∂k K with Hamiltonian (4) at the K valley is given by Using the Kubo formula [23] the magneto-optical conduc.20].N̂ + 1) zfK (z.

(13b) ηhEm+ 4m2 + 4m − 3 For the amplitudes of transitions between the cone bands ζ = −1 and ζ = +1. In summary. (14b) ηh Em+ − Em+1 − (2m + 1)(2m + 3) It is easy to check that the contributions of the transitions from flat to cone bands are significantly larger than the ones corresponding to cone to cone interband transitions when t1 = t2 (or more generally.ζ  the difference of the two Fermi functions becomes nonzero if σxx (ω) = σxx.c−c. we find from flat band (ζ = 0) to cone (ζ = 1) and at zero temperature   K.ζ K. 3 and 4.intra are the from the second and third terms in the expression of the contributions to the total longitudinal conductivity from conductivity. (12) of transitions are also allowed but much smaller than the ones mentioned above. respectively and mately the prefactor of the term (ω−E) 2 +η2 (for not too small are given in Appendix C. However. and between cones in each valley. from these results we shall establish the selection rules for the possible optical excitations between Landau levels. 4).f −c . FIG. and only cone to cone transitions remain present. we present our results for the magneto- optical conductivity. (13a) ηh Em+ − Em− which in case of β1 = β2 reads as e2 β12 4m2 + 4m − 5 Reσxx (ω = Em+ /) ≈ .c−c. when the two hoppings are sufficiently close to each other which is the case in Fig.KOVÁCS. DÓRA. the magnetic field and the Fermi energy will be discussed. not just the conductivities. 035414 (2017) IV. σxx. t2 = 0. RESULTS: THE MAGNETO-OPTICAL CONDUCTIVITY In this section. from the K  valley is given by the second term in (12) for the transitions from the flat band to the cone band this indicated by the replacement β12 ↔ β22 . in Figs. 3.ζ where σxx. After a lengthy but straightforward analytical given by Eq. The other selection rules can be obtained K. the temperature. as soon as. (14a) ηh ζ =± Em − Em Em+1 − Em+1 − Em Em+1 Em − Em+1 which in case of β1 = β2 takes the form √ + e2 β12 1 (m + 1)(2(m + 1) − (2m + 1)(2m + 3)) Reσxx (ω = (Em+1 − Em− )/) ≈ . and the intraband transition with energy E ≈ ω at frequency ω is approxi- 1 transitions (within the cones) in the K valley. Allowed transitions for different values of the Fermi energy. For finite temperature in principle other types   + β12 ↔ β22 . e.ω > 0). once the Green’s functions are calculated within our formalism. Our results show explicitly the different contributions to the conductivity corresponding to the interband and intraband transitions between the flat band and a cone.intra the Landau level indices n of the two energy levels differ ζ =± exactly by one.c−c. DÁVID. we illustrate the allowed transitions 035414-4 . the interband transitions between the flat band and a cone.f −c + σxx.inter ..g. The longitudinal conductivity The total longitudinal conductivity can be written as To see the allowed transitions between different Landau the sum of terms corresponding to intraband and interband levels we consider the three contributions to the conductivity transitions. and σxx. The contribution to the conductivity values of the magnetic field and for μ. In particular. First. In addition. one finds  −ζ 2 e2    ζ + ζ (m + 1) β12 Em + β22 Em+1 Reσxx (ω = (Em+1 − Em− )/) ≈  ζ −ζ  ζ −ζ  ζ −ζ  ζ −ζ  + β12 ↔ β22 .ζ K. The analytical calculation can be carried out in a straightforward way using the algebra of the creation and annihilation operators. Moreover. is ⎛ (m+1)β 4 mβ24 ⎞ e 2 β12 + β22 + E + E − 1 + + −  2  Reσxx (ω = Em+ /) ≈ ⎝ m+1 m+1 Em−1 Em−1 + β1 ↔ β22 ⎠.ζ K. AND CSERTI PHYSICAL REVIEW B 95.c−c. we consider the longitudinal conductivity. The magnitude of the contribution corresponding to a the interband transitions between cones. Below the dependence of the conductivity on the frequency. any response function follows straightforwardly.inter + σxx. the conventional graphene case is recovered without any flat band.ζ K. A. (12). The first term corresponds to the transition calculation.

ζ = The amplitude of the oscillations is given by Eqs. for not too small values Similarly to the case of longitudinal conductivity the of the magnetic field and for μ. [10] and terms corresponding to intraband and interband transitions. In fact. (17) (approximately) and producing a higher peak together (see Fig. (13a) 1. It is clear that for B → ∞ 1 gives the two largest peaks in the conductivity corre. However. .e. The value of Appendix C.ζ K. if E0+ < μ < E1+ (red dashed line). (13) of After a lengthy but straightforward analytical calculation. Finally.FREQUENCY-DEPENDENT MAGNETO-OPTICAL . for μ = 700 K.c−c.f −c + σxy.inter . .inter + σxy.ζ K. which implies that for large enough magnetic field transitions are positive. when t1 = t2 ) conductivity tensor σ ) can also be written as the sum of and 0 = 0. Figure 6 shows the Hall conductivity (the imaginary part Therefore main characteristics of the oscillation of the of the off-diagonal component of the conductivity tensor) as longitudinal conductivity as a function of the magnetic field a function the frequency for different chemical potential (in is mainly governed by the transitions between the flat band panel a) and for different hopping amplitudes (in panel b). Finally. the interband we consider the low-field limits since in this case the main transitions between cones. we find As far as intermediate magnetic fields are concerned. For simplicity.ζ K. (16) ηh ω transition disappears. However. we should be careful when transitions between the flat band and a cone. (21) of Ref.intra we only consider the case β1 = β2 . the height of the peaks and their positions can be determined from the results given by Eqs. 035414 (2017) for different Fermi energies. to cone levels.intra are the contribu- allowed to neglect cone-to-cone interband peaks according to tions to the total transversal conductivity from the interband the arguments above. 5) occurs at only from flat band to cone and cone to cone transitions. which is proportional to the position of the peaks Bm . So the distance between peaks decreases as the case. where γ = β1 / B independent of the magnetic field. the transition |n = 1.ζ have to deal with the single intraband transition. (15). here   K.ζ = conductivity on external field. 5). the mth When 0 < μ < E0+ in a valley (blue solid line) then there peak (starting is no intraband transition so peaks in the conductivity result −1/2 √ from the left-hand side √ in Fig. The contribution to conductivity from the K  the real part of the longitudinal conductivity tends to the low valley is given by the second term in (17) indicated by the magnetic field limit that can be determined from the integral in replacement β12 ↔ β22 . Let us consider the case when 0 < μ < E0+ so that we do not K. chemical potential μ. The height of a peak at frequency ω corresponding to a transition with energy E ≈ ω is approximately the weight B. respectively and are given in more than one transitions between the LLs. [20]. there is a negative peak (around  ≈ 400 K in the figure) difference of the square root of two neighboring odd numbers. e 2 γ 2 Bm then the aforementioned negative peak from flat band to cone Reσxx (ω = Em+ /) ≈ .c−c. It is also where σxy. and the intraband transitions (within contribution to the peaks in the conductivity results from the cones) in the K valley. All the other peaks corresponding to other flat band to cone and (13b). while another negative peak appears due to an intraband transition.ω > 0). even for very small values of the scattering rate.f −c . (15) ξ 2 − 0 −  ξ 2 ∂ ∂ 2 where tan φ = t2 /t1 . the and replacing the summation over n into a integral as Landau level indices change as n = 1 → n = 2 for K valley follows:   2        ∞ + 2 cos2 (2φ) nF 20 −  − nF 20 +  0 nF (0) − nF 20 +  2ie2 ξ σxx = d 2 + sin (2φ) 2  2 h 0 /2 2 ξ 2 − 42 ξ 2 − 20 +     2       nF (0) − nF 20 −  2 − 20 ∂nF 20 −  ∂nF 20 +  −  2 + − . we discuss the dependence of magneto-optical For μ = 50 K. While in case of μ = 500 K Landau levels tend to infinity. and σxy. For a fixed value of frequency ω. Figure 4 shows the conductivities and n = 0 → n = 1 for K  valley but the quantum number as a function of the frequency  = ω for three different ζ = 1 does not change. In this Bm = ωγ 2m + 1. σxy. Small negative peaks (around  ≈ the peaks in the longitudinal conductivity tends to 300 K in the figure) due to interband cone to cone transitions are also present. (15) transforms into Eq. The transversal conductivity factor of the term (ω−E) 1 2 +η2 (again.c−c. Eq..ζ = 1 provides the limit is obtained by introducing the variable  = En+ − 0 /2 largest peaks in the conductivity. Eq. The formula for the low-field the transition |n = 0. the pattern for ζ =± general hopping amplitudes is rather cumbersome as peak   energies corresponding to different transitions might coincide + β12 ↔ β22 .ζ K.ζ = 0 → |n = 0.ζ  σxy (ω) = σxy. Ref. in case of φ = π/2 (graphene) and 0 = 0 into Eq.ζ = 1 → |n = 1.c−c. the conductivity should vanish since the distance between the sponding to the two valleys. corresponding to the transition |n = 1. PHYSICAL REVIEW B 95. (C1). In case of φ = π/4 (i.ζ = 0 → |n = 0. The heights of positive transversal conductivity (off-diagonal component of the peaks (flat band to cone transitions) fall rapidly in both cases 035414-5 .

the corresponding transitions that one of these two peaks becomes positive as can be seen in are assigned in the form of |n. The real part of the longitudinal conductivity (in units of e2 / h) as a function of the inverse square root√ of magnetic field B (here √ B is in units of T)√for (a) β1 = β2 = 400 K B. Em−1 Em+1 Em+1 ≈ ηh Em+ − Em− (18a) which in case of β1 = β2 reads as e2 β12 2m + 1 Imσxy (ω = Em+ /) ≈ . respectively) as shown which is between the Landau levels n = 0 and n = 1 for both valleys. in Fig. 5. β1 = 350 K.ζ  → |n . Fig. η = 5 K. μ = 730 K [black line. DÓRA. The real part of the longitudinal conductivity (in units of ηhEm+ 4m2 + 4m − 3 2 e / h) as a function of the frequency ω (in units of K) for Fermi energy μ = 500 K [blue line. in terms of the frequency according to Imσxy (ω = Em+ /) ⎛ mβ24 (m+1)β14 ⎞ e2 ⎝ β2 − β1 + Em−1 − 2 2 + − + −  2  + β1 ↔ β22 ⎠. (c)]. It is also worth noting that when β1 = β2 then the (b)] which lies between the Landau levels n = 0 and n = 1 in the K first negative flat band to cone peak splits into two peaks valley and in the gap in the K  valley. β2 = 450 K. (18b) FIG. 7 exactly at  = 300 K. DÁVID.ζ  . 0 = 100 K for the red dotted line. AND CSERTI PHYSICAL REVIEW B 95. 0 = 500. μ = 50 K. 035414 (2017) FIG. and 0 = 0 in both cases. and μ = 850 K [red line.KOVÁCS. If the difference between the The parameters are T = 10 K. 4. 6(a) around  ≈ 250 K. (a)] which is in the gap. (corresponding to K and K  valleys. For the first few peaks. 035414-6 . ω = 200 K. The parameters are T = 10 K. two hopping amplitudes are large enough then it might occur and η = 5 K. (b) β1 = 300 K B and β2 = 400 K B for the blue solid line.

In this case we obtain  for the K valley. β2 = 500 K (red dashed line). when the temperature is zero and the Fermi energy lies between the flat band and first LL then the sums in Eq. K + K  contributions do agree.E0+ } corresponding to the K and K  valleys. the Hall conductivity looses its symmetry with respect to μ = 0. 035414 (2017) Similarly. e2 β12 − β22  σxy = K F (0) − 2 K Fn . for hopping amplitudes 0 = 0. β1 = 300 K. We now show that this Hall conductivity can be related to the Berry phase. (C2). (21b) h β12 + β22 ∞ 2e2   K n=0   σxy (μ. (22) h n=0 The Hall conductivity as a function of the Fermi energy at zero temperature and in DC limit is plotted in Figs. we can find the levels for the K valley. [24]. 8 and 9. The insets in this figure show the individual contributions from the two valleys to the conductivity. the conductivity has a change 2e2 / h at all the other Landau levels. PHYSICAL REVIEW B 95. . For 0 = 0. 035414-7 . Our results for 0 = 0 agree with those on Ref. The parameters are T =10 K and η=5 K in both cases. The parameters are T = 10 K and η = 5 K. β1 = 380 K. and FnK is the same as FnK with energy  the usual Hall conductivity. β2 = 420 K (blue 0. and 0 = K). (b) for μ = 50 K and for two FIG. Indeed. (21) becomes zero and F (0) = 2. although the total. The heights of peaks for interband cone-to-cone transitions (ω = Em+1 − Em− = Em+ − Em+1 ) are  2 ζ 2 −ζ 2  2  ζ (m + 1) β1 Em + β2 Em+1   + e Imσxy (ω = (Em+1 − Em− )/) ≈ −  ζ −ζ  ζ −ζ  ζ −ζ  ζ −ζ  + β12 ↔ β22 (19) ηh ζ =± Em − Em Em+1 − Em+1 −Em Em+1 Em − Em+1 and for intraband transitions  + 2 + e2 (m + 1) β12 Em+ + β22 Em+1  2  Imσxy (ω = (Em+1 − Em− )/) ≈− + + β1 ↔ β2 . it is easy to check that the peaks corresponding to cone-to-cone interband and intraband transitions are always + − negative. β1 = β2 = 423 K related to the Dirac-Weyl model for s=1 (red solid line). [15]. obtained using the Streda  ∞  formula. Our valley resolved results for 0 = contribution from the K and K  valleys as 0 differ from those in Ref.  ∞  Here the spin degeneracy is taken into account. The imaginary part of the transversal conductivity (in sets of hopping parameters: 0 = β1 = 0. 6. β2 = 423 K corresponding units of e2 / h) as a function of the frequency  = ω/kB (in units of to graphene with B = 1 T magnetic field (blue solid line). dashed line). 2 (20) ηh (Em+ − Em− )(Em+1 + − + + + − Em+1 )(Em Em+1 )(Em+1 − Em ) Finally. we consider the transversal conductivity in the dc where FnK = nF (En+ ) + nF (En− ) and En± are the energy levels limit (ω = 0) and at zero temperature.T ) = σxy K + σxy K =− Fn + FnK . From Eq. It can be shown that the conductivity is zero when μ is in the narrower energy interval {E0− . 7.FREQUENCY-DEPENDENT MAGNETO-OPTICAL . Then the FIG. For parameters used in the figure this is |μ| < β1 . The imaginary part of the transversal conductivity (in units of e2 / h) as a function of the frequency  = ω/kB (in units of K) (a) for Fermi energy μ = 50 K lying between the flat band and the first LL (blue solid line) and μ = 500 K which is between the first and second LL (red dashed line). . Thus the total e2 β22 − β12  contributions from K and K  valleys can be rewritten as K K σxy = F (0) − 2 Fn . and for 0 = 0. Moreover. (21a) Note that the Streda formula calculation also fails to reproduce h β12 + β22 n=0 the half-integer quantized Hall response of graphene though.B.

We show that the Hall conductivity at zero temperature respectively. The transversal conductivity (in units of e2 / h) in dc From our general result for the transversal conductivity we limit (ω = 0) as a function of the Fermi energy. respectively. To solve this equation. + 1 denote the Fock number and the band index.01 K.ζ . It is interesting to note that the valley ACKNOWLEDGMENTS resolved Hall response is not only fractional but can also be rational without any electron-electron interactions. DÁVID. 035414 (2017) we assumed that the on-site energy 0 of the sixfold connected site can be nonzero. in earlier works. (8). and β2 = 500 K. temper- ature and Fermi energy. we present the eigenvalues and the eigen- states of the Hamiltonian (7) (around the K valley). β1 = 300 K. CONCLUSIONS APPENDIX A: EIGENVALUES AND EIGENSTATES In this work. η = 5 K. Moreover. AND CSERTI PHYSICAL REVIEW B 95. and ζ = −1. (23) conductivity can be tested experimentally with cold atoms in h β1 + β2 2 2 h an optical lattice. agrees with that obtained from the Berry phase calculated β1 = 300 K. and n = 0. the total Hall response is always integer work is supported by The National Research. FIG. β2 = 500 K for the red A B ⎝ z − H = −β1 â † z − 0 −β2 â ≡⎠ . To this end. albeit the We would like to thank A. we partitioned the operator included!). the calculations are also carried out for graphene and it is shown that the results obtained from our method are in agreement with those known in the literature. APPENDIX B: CALCULATION OF THE GREEN’S FUNCTION FIG. external field.1. C D so that β12 + β22 has the same value for the two cases. To demonstrate the theoretical method mentioned above. 0 −β2 â † z 035414-8 . β1 = β2 = 412 K for the blue solid line. Pályi for helpful discussions. 108676.KOVÁCS. 0 = 600 K. 0 = 0.  e2 β12 − β22 e2 We believe that our predictions for the magneto-optical K. the magneto-optical conductivity tensor is calculated as functions of frequency. [12. Furthermore. respectively. we introduce a new analytical procedure to determine the Green’s function in an operator form independent of any representation. our algorithm is an efficient and where φ is given by tan φ = t2 /t1 and the spin degeneracy is universal approach and thus easily applicable to other systems. 9. 8. the Hall conductivity is integer conductivity becomes quantized and the steps of quanta depend continuously on the hopping parameters between adjacent layers. V.2. Insets (upper right and lower left) show the contributions z − H as from the K and K  valleys. The parameters are T = 0. . .ζ  = Enζ |n. (B1) dotted line and 0 = 600 K. The parameters are T = ⎛ ⎞ z −β1 â 0   0. Using the Kubo formula expressed with Green’s function. respectively. η = 5 K. Development and quantized.01 K. (A1) ζ where En and |n.K σxy = ±2 = ±2 cos(2φ).ζ  is the energy eigenvalue and the cor- responding eigenstate. 119442.0. included. No. Innovation Office under Contracts No.15]. Insets (upper right derived an analytic expression for the Hall-conductivity in dc and lower left) show the contributions from the K and K  valleys. The results are summarized in Table I. DÓRA. (7). the magneto-optical conductivity in the OF THE SYSTEMS generalized α-T3 model is calculated. When the Green’s function is given in position representation the evaluation of the trace in the Kubo formula is a quite cumbersome analytical calculation. This results is in agreement with the Berry phases obtained in Refs. . limit. In this generalized form In this section. we look for a solution of the form given by Eq. 105149 and No. The Schrödinger equation reads as H |n. This sum of the two valleys. The transversal conductivity (in units of e2 / h) in dc limit To obtain the operator of the Green’s function for the (ω = 0) as a function of the Fermi energy (spin degeneracy factor Hamiltonian H given by Eq. The advantage of our approach is that the Kubo formula can be calculated in simple way using only the algebra of the creation and annihilation operators.

f −c (ω) = ζ n+1 n+1 −ζ n−1 n−1 ζ + ζ nF (0) − nF Enζ .N̂ − 1) β1 β2 2 −A−1 BS −1 = z .En±1 |n.N̂ ) inverted. (B2). Now using (B3) and the general formula (B2) Green’s function can be obtained by the transformation (6). ζ = 0 E00 = 0 (0. (B5)   † (z − 0 ) p(z. For the case of K  valley. (B6) −1 A = . −1 we obtain the operator of the Green’s function G(z) = where p(z.26]. (B4d) where S = D − CA−1 B and the operators A and S can be β1 â † f (z.ζ (n + 1) β12 En + β22 En+1 σxx.N̂ + 1) z p(z.N̂) = [z2 − 0 z − β12 N̂ − β22 (N̂ + 1)] . PHYSICAL REVIEW B 95. (B4) into Eq. (C1b) ξ + En − En+1 ξ − En + En+1 ∞  ζ 2 ie2  ζ K. −1 −1 + A BS CA −1 −1 C D −S −1 CA−1 S −1 A     (B2) 1 I + β12 (N̂ + 1)f (z.ζ ie 1 2 E E 1 σxx.N̂ + 1) β1 âf (z. ζ = 0 En0 =0 γn0 (−β2 n + 1|n − 1.N̂) â f (z. (12) in which the different terms reads + E + E2 −  1  4 ∞ β 2 + β 2 + (n+1)β nβ 4 2  + − 1    K. and performing the Mat- â f (z. (C1a) h n=0 En − En ξ − En ξ + En ∞   −ζ 2 ie2  ζ K.0.β2 |1) n = 0.N̂ ) Finally. (B4a) z Using the operator form of the Green’s function given by   Eq.N̂) = f (z.N̂ − 1) can be calculated analytically. The normalization − 12 − 12 factors are γnζ = (β12 n + β22 (n + 1) + (Enζ )2 ) for ζ = ± and γn0 = (β12 n + β22 (n + 1)) for ζ = 0.inter (ω) =  ζ   ζ −ζ  ζ −ζ  h n=0 En − En−ζ En+1 ζ −ζ − En+1 − En En+1 En − En+1   1 1   −ζ    × ζ −ζ + ζ −ζ nF En+1 − nF Enζ .N̂) = z . (B3) β1 â † p(z. (9). . This is often called in the literature the Banachiewicz −1 inversion formula [25.N̂ + 1) β1 â p(z.FREQUENCY-DEPENDENT MAGNETO-OPTICAL . ζ Enζ |n. β2 â † f (z. ζ = ±1 En±1 = 0 +ζ + β12 n + β22 (n + 1) γn±1 (β1 n|n − 1. . Landau levels and eigenstates for valley K.N̂ ) . where f (z.0. Each levels are labeled by the Fock number n and a band index ζ . (B4b) subara summation in (10b) the magneto-optical conductivity β2 âf (z.N̂ ) = f (z.E0±1 |0. the matrix elements of the inverse of matrix in (B1) can be calculated analytically.N̂ + 1) zf (z.c−c.N̂ + 1).intra (ω) =   ζ −ζ  ζ ζ  ζ ζ h n=0 Enζ − En−ζ En+1 − En+1 En En+1 En+1 − En   1 1   ζ  ζ  × ζ ζ + ζ ζ nF En − nF En+1 .ζ (n + 1) β12 En + β22 En+1 σxx. z  −1  −1  (B4c) A B A + A−1 BS −1 CA−1 −A−1 BS −1 = .|0)T   Then we apply the general formula for the inverse of a 2×2 β1 β2 †2 partitioned matrix −S −1 CA−1 = â f (z. Then the longitudinal conduc- tivity is given by Eq. and we find APPENDIX C: EXPRESSIONS FOR THE LONGITUDINAL AND TRANSVERSAL CONDUCTIVITIES 1  S −1 = I + β22 Nf (z.N̂ − 1) .N̂) = (z2 − 0 z − β12 N̂ ) and N̂ = â † â is the (z − H )−1 as given by Eq. substituting the terms given by Eqs.c−c. (9) and the current operators (11).N̂ − 1) â.β2 n + 1|n + 1) 2 2 √ √ T n > 0. 035414 (2017) TABLE I.ζ    √ 0 2 √ T n > 0.β1 n|n + 1)  2 T n = 0.N̂ + 1) â † . (C1c) ξ + En − En+1 ξ − En + En+1 035414-9 . The inverse of operator A defined in (B1) can also be Here we have made use of the following identities: calculated from formula (B2) and after a simple algebra we find â f (z. ζ = ±1 E0±1 = 0 2 + ζ 20 + β22 γ0±1 (0. the number operator. n.

Lan. Raoux. and B. M. Cantillano. 112. Fuchs. Piéchon. R. M. Matrix Theory (Springer-Verlag. (2006). 197 (2005).ζ + − En−1 1    σxy. M. Many-Particle Physics. 045310 [26] R. R. 075104 Rev. [22] A. J. Lett. [21] E. DÓRA. Ghosh. B 84. Mosseri. [4] K.ζ (n + 1) β12 En + β22 En+1 σxy. [9] H. Rev. Mejı́a-Cortés. 109 (2009). V. Spracklen. [11] F. B 84. Ser. Choudhury. Phys. Aoki. G. B. Bandres. Mahan.-N. (C2b) ξ + En − En+1 ξ − En + En+1 ∞  ζ 2 K. UK. 82. D. [14] B. Lett. F. Lu. Biswas and T. Phys. Owen. [24] T. Matter 28. Wimmer. S. Öhberg. Kane. 245411 Rev. (17) in which the different terms are − E + E −1  1  nβ24 (n+1)β 4 e2  β2 − β1 + En−1 ∞ 2 2 K.-L. and M. Thomson. Qi and S. A. Weimann. Weimann. Moessner. and R.KOVÁCS. 2nd ed. Rev. M. A. [6] D. Rev. Green. Chamon. 5888 [25] F. J. [5] D. J. Rev. 245410 and A. Novoselov. I. 2881 (1990). V. and C. Sharapov. Santos. (1998). Hasan and C. 075414 (2012). [16] S. R. and D.ζ e2  (n + 1) β12 Enζ + β22 En+1 σxy. Herbut. A. (C2a) h n=0 En − En ξ − En ξ + En ∞  −ζ 2 e2  ζ K. Dóra. 1990). E. Dóra. Phys. Phys. 2011). Matter 19. Chodos. DÁVID. K. Phys.: Rev. Kuzmenko. Phys. C. V. Geim. V. Bermudez. A. I. A. 1057 (2011). 1985). and R. N. A. 81. 245503 (2015). H. [13] J. 035414 (2017) Similar calculations leads to the Hall conductivity given by Eq. 81. Fuchs. Morales-Inostroza. Bercioux. J. A. Montambaux. K. D 42. P. Rev. Lett. 063061 (2014). Rev. Phys. Peres. Vicencio. 115136 (2011). [17] R. Gusynin. 16. [12] A. Vicencio. S. Y. 114. B 90. Zhang. Novoselov. Grabert. 334. and H. AND CSERTI PHYSICAL REVIEW B 95. and A.: Conf. Andersson. μ is the Fermi energy. L. M. Johnson. Carbotte. Gorbar. 495302 Montambaux. Häusler. P. and S. Rechtsman. J. Carbotte. Rev. Bercioux. S. 603. B. (2015). Morozov. Phys. 3045 (2010). Phys. 195422 (2011). L. 245504 (2015). Piéchon. S. Raoux.c−c. Guinea. Phys. Phys. Goldman. [1] A. F. Nature (London) 438. Guzmán-Silva. P. Lett. and ξ = ω + iη. [8] Z. S. Phys. R. Szameit. Firsov. and W. N. D. Dubonos. Phys. [18] D. Kailasvuori. and J. D. Matrix Analysis (Cambridge (2014). K. B 92. (C2c) ξ + En − En+1 ξ − En + En+1 where nF (E) = 1/(e(E−μ)/(kB T ) + 1) is the Fermi distribution function. R. J. G.inter (ω) =   ζ −ζ  ζ −ζ  ζ −ζ  h n=0 Enζ − En−ζ En+1 − En+1 − En En+1 En − En+1   1 1   −ζ    × ζ −ζ − ζ −ζ nF En+1 − nF Enζ . Urban. Rev. Rev. B 86. A.c−c. S. Castro Neto. [10] B. 012001 (2015). Douçot. A 80.-C. F. N. Phys. A. 063603 (2009). C. Gusynin. and G. C. Phys. Condens. M. Öhberg. Mukherjee. B 73. and W. P. Rev. Urban. Phys. D.intra (ω) =   ζ −ζ  ζ ζ  ζ ζ h n=0 Enζ − En−ζ En+1 − En+1 En En+1 En+1 − En   1 1   ζ  ζ  × ζ ζ − ζ ζ nF En − nF En+1 . Szameit. Mejı́a-Cortés. Nicol. Häusler. Sharapov. [2] M. G. Everding. New J. B 82. Phys. P. Segev. K. Molina. Phys. [15] E. V. Phys. 83. Mod. 114. H. A. Moessner. W. Conf. Sharapov. 035414-10 . A. [19] V. Morigi.: Condens. M. Cambridge. Grigorieva. Rev. F. and G. 165115 (2011). J. 012044 (2011). D. Vidal. Katsnelson. [20] V. and P. Rev. University Press. L.-N. Geim. F. Mod. Z. K. Watanabe. Phys. Illes. A. (2016). J. Horn and C. Hatsugai. [3] X. I.f −c (ω) = ζ −ζ n+1 n+1 ζ − ζ nF (0) − nF Enζ . [7] D. Mod. Goldman. 026402 (2014). Rev. A. and Rev. S. Gusynin and S. Real. Zhang. New York and London. and R. (2010). Ser. 026222 (2007). J. and E. Nolte. P. S. Phys. New York. B 84.: [23] G. C. (Plenum Press. Jiang. I. Phys.