Arxiv:2002.05124V5 [Cond-Mat.Mes-Hall] 4 Jun 2020 [45]

Eﬀective Floquet Hamiltonians for periodically-driven twisted bilayer graphene

Michael Vogl∗ Department of Physics, The University of Texas at Austin, Austin, TX 78712, USA

Martin Rodriguez-Vega∗ Department of Physics, The University of Texas at Austin, Austin, TX 78712, USA and Department of Physics, Northeastern University, Boston, MA 02115, USA

Gregory A. Fiete Department of Physics, Northeastern University, Boston, MA 02115, USA and Department of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA (Dated: June 5, 2020) We derive effective Floquet Hamiltonians for twisted bilayer graphene driven by circularly polarized light in two different regimes beyond the weak-drive, high frequency regime. First, we consider a driving protocol relevant for experiments with frequencies smaller than the bandwidth and weak amplitudes and derive an effective Hamiltonian, which through a symmetry analysis, provides analytical insight into the rich effects of the drive. We find that circularly polarized light at low frequencies can selectively decrease the strength of AA-type interlayer hopping while leaving the AB-type unaffected. Then, we consider the intermediate frequency, and intermediate-strength drive regime. We provide a compact and accurate effective Hamiltonian which we compare with the Van Vleck expansion and demonstrate that it provides a significantly improved representation of the exact quasienergies. Finally, we discuss the effect of the drive on the symmetries, Fermi velocity and the gap of the Floquet flat bands.

I. INTRODUCTION an exponentially-long pre-thermal time regime [59–64] in driven interacting quantum systems allows one to intro- The recent discovery of strong-correlation effects in duce the notion of effective time-independent theories. twisted bilayer graphene (TBG) generated great inter- The development of several techniques to derive effec- est in moiréheterostructures [1–31] and ways to sim- tive Hamiltonians in different drive regimes led to rapid ulate them [23]. Similar to the behavior in cuprates evolution of the Floquet engineering field [61, 65–86]. [32, 33] at different filling factors superconductivity, For instance, the prediction of an anomalous Hall effect Mott-insulating [4, 16, 34–36] and ferromagnetic be- in single-layer graphene driven by circularly polarized haviour [37, 38] has been observed in TBG. The ex- light [87] has been recently confirmed in experiments [88]. perimental observations were followed by several theo- More generally, there has been an increased interest in retical proposals to explain the observations based on the study of topological transitions induced by periodic the existence of flat bands which appear at special twist drives [87, 89–107]. angles [1, 9, 39]. These flat bands play an essential More recently, the fields of twistronics and Floquet en- role for the emergence of strong correlations because gineering crossed paths in twisted bilayer graphene driven the interaction terms become relatively dominant [40] by circularly polarized light in free space [108–110]. In- over the kinetic energy contributions of the dispersive teresting effects like topological transitions at large twist bands [12, 24, 36, 39–41]. angles using high-frequency drives [108] and the induc- In TBG, the flat bands depend strongly on the twist tion of flat bands using near-infrared light in a wide range angle between the graphene layers, which is experimen- of twist angles [110] were found. These studies are mainly tally difficult to set to a precise values. This challenge has numerical, and only provide analytical descriptions in the lead to several studies proposing different mechanisms to high drive frequency regime, which we will define rigor- correct for deviations from the magic angle. For example, ously in the next section. via pressure [5, 42–44] or light confined in a waveguide The aim of this work is to derive analytical effective arXiv:2002.05124v5 [cond-mat.mes-hall] 4 Jun 2020 [45]. Floquet Hamiltonians that allow us to gain insight into In parallel to the developments on moirélattices, there twisted bilayer graphene subjected to circularly polarized has been a rapid progress in our understanding of non- light away from the conventional weak drive, high fre- equilibrium systems, both experimentally and theoreti- quency regime of Van Vleck [111, 112], Floquet-Magnus cally, particularly for the case of periodic drives, which or Brillouin-Wigner approximations [74, 113]. Our effec- may be induced by a laser. [46–58]. The existence of tive Floquet Hamiltonians allow us to elucidate the effects of the interplay of moirélattices and Floquet drives. Particularly, we consider two complementary regimes: i) a regime characterized by weak drive and low frequencies; ∗ These two authors contributed equally. and ii) a regime characterized by intermediate frequen- 2 cies and strong drives. The remainder of the manuscript (a) (b) is organized as follows: in Sec. II we describe the system we consider; in Sec.III we examine the low-frequency, weak drive limit; and in Sec.IV we address the intermediate frequency and intermediate strength drive regime. Finally, in Sec.V we present our conclusions and outlook.

II. SYSTEM DESCRIPTION

A. Static Hamiltonian (c)

The starting point of our discussion is the eﬀec- tive Hamiltonian that describes twisted bilayer graphene [1, 25, 39, 114–116] h(−θ/2, k − κ−) T (x) Hk(x) = † , (1) T (x) h(θ/2, k − κ+) which describes two stacked graphene layers that are ro- tated with respect to each other by an angle θ, as shown in the sketch of ﬁgure 1(a). Here,

0 f(R(θ)k) h(θ, k) = γ , (2) f ∗(R(θ)k) 0 FIG. 1. (Color online) a) Sketch of twisted bilayer graphene is the single-layer graphene Hamiltonian, f(k) = irradiated by circularly polarized light. (b) moiréBrilloiun ia k − 2 ia k 0 y a k π e 3 0 y +2e 3 sin √0 x − describes the intralayer zone. (c) Band structure for twisted bilayer graphene for w0 = 3 6 ◦ w1 = 110 meV, and θ = 1.05 . The low-energy flat bands are hopping amplitude between nearest-neighbor sites, and highlighted in red. γ = vF /a0, where we use natural units ~ = c = e = 1. The inclusion of the full structure of f(k) means that this Hamiltonian is valid in the full Brillouin zone and not just near a K point. The interlayer hopping matrix would incorporate the two graphene valleys. However, 1 we only consider perturbations induced by light, which X −ibix T (x) = e Ti, (3) cannot induce processes that mix the two valleys. The i=−1 Hamiltonian in the other valley is connected by a C2 ro- 2πn 2πn tation [119]. The symmetries of the continuum model Eq. Ti = w012 + w1 cos σ1 + sin σ2 , (4) (1) include C3 rotational symmetry about the center of a 3 3 AA region, C2T symmetry (taking into account both val- describes tunneling between the two graphene layers and leys, the TBG presents time-reversal symmetry T ), and encodes a hexagonal pattern that has its origin in that My : y, ky → −y, −ky mirror symmetry [16, 119, 120]. the two superimposed graphene lattices which develop a In the small-rotation limit, the angle dependence of the moirépattern (see Fig. 1(b)), where b0 = (0, 0), and graphene sectors can be neglected, leading to an approx- √ imate particle-hole symmetry C [120]. b±1 = kθ ± 3/2, 3/2 are the reciprocal lattice vectors. Following Refs. 45, 109, and 110 we introduced an additional parameter w1 into the tunneling term to model relaxation effects, since AB/BA stacking configurations are B. Driven twisted bilayer graphene energetically favoured over AA configurations [115, 117]. Furthermore, there are indications that AA and AB For the driven system, we assume that circularly po- regions have different interlayer-lattice constants [118]. larized light is applied in a direction normal to the TBG Throughout this work, we fixed γ = vF /a0 = 2.36 eV, plane as sketched in Fig. 1(a). Then, the light enters and a = 2.46 A.˚ For a detailed description of the band 0 via minimal substitution as k → k˜ (t) = k −A cos(Ωt), structure numerical implementation, see the appendix x x x k → k˜ k − A t of [45]. In figure 1(c) we show the band structure for and y y = y sin(Ω ) leaving the tunneling sec- w = w = 110 meV, and θ = 1.05◦, value near the tor almost unaltered. The reason for this is simple. The 0 1 inclusion of light in a tight binding model can be done magic angle. R Rj i R Adr The Hamiltonian in Eq. (1) describes only one val- via a Peirls substution for hoppings tij → e i tij. ley degree of freedom. A full description of the system The interlayer hopping is dominated by hopping between 3 atoms that are almost exactly on top of each other - as |ψ(t)i = eit|φ(t)i, where |φ(t + 2π/Ω)i = |φ(t)i afterall other atoms are further away and the overlap and is the quasienergy. Replacing |ψ(t)i into the between orbitals is smaller. Therefore for interlayer cou- Schrödinger equation leads to [H(x, t) − i∂t]|φ(t)i = plings mostly longitudinal components of A contribute in |φ(t)i, which governs the dynamics of the periodic sys- R the line integral R j Adr. Circularly polarized light only tem. The exact solution can be generically obtained Ri has transverse components and therefore has little effect either by constructing the Floquet evolution operator 2π/Ω R −iHF T on interlayer couplings. The time-dependent Hamilto- UF = T exp{−i 0 H(s)ds} = e or by employ- nian is ing the extended-state picture. In the extended-state pic- P inΩt ture, we use the Fourier series |φ(t)i = n e |φni, which leads to P H(n−m) + δ Ωm |φ i = |φ i, m n,m m n h(−θ/2, k˜(t) − κ−) T (x) defined in the infinite-dimensional Floquet-Hilbert space H(x, t) = † , T (x) h(θ/2, k˜(t) − κ+) spanned by the direct product of the Hilbert space of the (5) static system and the space spanned by a complete set with H(x, t + 2π/Ω) = H(x, t). The Floquet theo- of periodic functions. The Hamiltonian Fourier modes (n) R 2π −iτn rem [69, 113, 121] exploits the discrete time-translational are given by H = 0 dτ/(2π)H(τ)e , which can symmetry and allows one to write the wavefunctions be derived by making the replacements

1 a k πn π Aa (n) − i(2a0ky +3(θ−π)n) ia0ky 0 x 2 2 0 f(k) → f (k) = e 3 1 + 2e sin √ + − Jn 3 3 6 3 , (6) (n) T (x) → T (x) = δn,0T (x)

in Eq. (1). driven system in the weak-drive limit, since they couple The two exact approaches outlined above are challeng- weakly to the low-energy bands. ing to use in practice, and one usually has to employ ap- The time-dependent Hamiltonian within these approx- proximations. In the following sections, we will employ imations has the form a recently developed [79] approach valid in the weak- −iΩt † iΩt drive limit and for arbitrary frequencies. Also, we will H(t) = HL + Pe + P e , (7) introduce improved methods to study the intermediate- where the monochromatic operator P = amplitude drive regime valid in the high and intermediate T −1 R T dsH x, s eiΩs frequency regimes. 0 ( ) is given by   0 eiθ/2 0 0 0 0 0 0  P = −Aγa0   , (8) III. WEAK DRIVE REGIME 0 0 0 e−iθ/2 0 0 0 0 Thus far, most discussions of twisted bilayer graphene irradiated by circularly polarized light have focused on and HL is the same as Eq. (1) just with f(k) → fL(k) the high frequency limit. This is for practical reasons linearized momentum dependence. because the lower frequency regime, while it is more in- For weak driving amplitudes A and arbitrary frequency teresting and relevant for experiments, is also harder to Ω, the periodically driven systems can be described by treat using the existing theoretical tools. In Ref. 79, we the eﬀective self-consistent time-independent Hamilto- developed a method to address this issue in the weak nian [79] driving limit. Here, we apply our method using a series 1 † † 1 of approximations necessary to make progress and gain Heﬀ ≈ HL + P P + P P, (9) some analytical insights into the low frequency regime. − HL − Ω − HL + Ω

If we are interested on the effects of the drive on where H0 is the time-averaged Hamiltonian, and are the the low-energy bands, small angles, and weak drives our quasienergies. For large frequency drives, we can apply original Hamiltonian can be approximated with f(k) ≈ a Van Vleck expansion and obtain the effective Hamilto- −i θ fL(k) = a0e 2 (kx −iky), in the vicinity of the graphene nian [111, 112] Heff = HL + HΩ, where the leading order 2 K point. The reason we may Taylor expand for small mo- correction is given by HΩ = −∆τ0 ⊗σ3, ∆ = (Aγa0) /Ω, menta when the twist angle θ is small is because the moiré and σi, τi are the Pauli matrices in pseudo-spin and Brillouin zone is very small i.e. kθ kD. Non-linear cor- layer space, respectively. To keep the notation simple, in rections only become important for higher-energy bands. the remainder of the text we refer to the approximation † These higher-energy bands are in turn not relevant for the Heff ≈ HL + [P ,P ] as the Van Vleck approximation. 4

Therefore, in the high-frequency limit, the main effect if w0 ≈ w1. (Physically, this is a regime where the in- is the addition of the gap ∆ in the quasienergy spectrum terlayer coupling are essential to the physics.) There- originating from the breaking of time-reversal symme- fore, for small enough angles and momenta that ful- try T . This gap is topologically non-trivial, and leads fill this inequality we may introduce the approximation −1 −1 to topological Floquet flat bands with Chern number ( − HL ± Ω) ≈ ( − HT ± Ω) where C = 4 [109, 110] which could serve as platforms to realize Floquet fractional Chern insulators [109, 122]. The 0 T (x) relatively large Chern number originates from spin and HT = . (11) T †(x) 0 valley degeneracy [109, 110]. In order to evaluate the effective Hamiltonian Heff for arbitrary frequency, we notice that the Brillouin zone has Replacing this approximation in the second and third dimensions kθ ∝ sin(θ/2) and therefore the correspond- terms of equation (9), we find the effective Hamiltonian ing energy obeys vF kθ w0,1 for sufficiently low angles. 3 2 ~ A A kθ Heff = H0 + HΩ + O , , where the This is, for small angles, T (x) introduces the dominant kD kD kD energy scale i.e. minkT (x)k kh(k)k, where k.k is a ma- neglected terms that are third order in small parame- trix norm. This estimate can be written more precisely 4 ters. We find that terms of order O A vanish. The as kD leading-order correction to the Hamiltonian HΩ has the p 2 2 w1 (k − κ+) + (k − κ−) 3 (10) following form ~vF

1 1 HΩ(x) = V (x, Ω)τ0 ⊗ σ0 + U(x, Ω)τ3 ⊗ σ0 + ∆1(x, Ω)(τ0 + τ3) ⊗ σ3 + ∆2(x, Ω)(τ0 − τ3) ⊗ σ3 2 2 + ∗ − + ∗ − + δw0(x, Ω)τ ⊗ σ0 + δw0(x, Ω)τ ⊗ σ0 + β(x, Ω)τ ⊗ σ3 + β (x, Ω)τ ⊗ σ3. (12)

Equation (12) is the first main result of our work. In Bernal-stacked bilayer graphene, a interlayer bias The full expressions for each of the terms appearing in U opens up a gap in the energy spectrum around the Eq. (12) are given in appendix A, and depend on the K points [123, 124]. If we introduce a region in space quasienergy, which was omitted explicitly for brevity. A where the sign of the interlayer bias changes, U → −U, perturbative approach generically generates long-range a domain wall forms where the gap inverts, leading to hopping terms as the frequency is arbitrarily decreased. topologically protected helical (TPH) modes [125–127]. The method here employed leads to the closed form in In twisted bilayers, even though U does not gap the spec- Eq. (12), which contains all the possible terms that trum, the moirépattern alternating AB/BA regions leads can be generated by the drive, even in the low-frequency to the formation of topological boundary modes even for regime, defined as driving frequency Ω . W with W ∼ spatially-homogeneous interlayer bias U [128]. Here, we maxt kH(t)k. Conversely, we define the high-frequency obtained that circularly polarized light induces an inter- regime for the moirésystem as Ω > W ∼ maxt kH(t)k. layer potential U(x, Ω) in the low-frequency limit, which Now, we discuss the origin and implications of each could induce the formation of topologically protected he- the new terms on the symmetries of the system. Due lical modes. to the assumed approximations, the corrections to the Next, the terms ∆1/2(x, Ω)(τ0 ± τ3) ⊗ σ3 with Hamiltonian HΩ(x) presents no momentum dependence ∆1/2(x, −y) = ∆1/2(x, y) and ∆2(x, y) = ∆1(−x, y) and does not commute at different points in space, break My, and C2T symmetry, which protects the lin- 0 [HΩ(x),HΩ(x )] =6 0. ear band crossing, leading to the opening of a gap at the −2 The first term, V (x, Ω)σ0⊗τ0, with V (x, Ω) ∝ O(Ω ), κ± points in the mBZ. The ∆1/2(x, Ω) position depen- corresponds to an overall position-dependent potential dence is relevant at order O(Ω−3), and the asymmetry −4 which does not introduce new physics. The second term, ∆1 =6 ∆2 is relevant at order O(Ω ). When both TBG U(x, Ω)σ0 ⊗ τ3, is a position-dependent interlayer bias valleys are taken into account, this term breaks time- with U(x, y) = U(x, −y), U(x, y) = −U(−x, y), and reversal symmetry T and leads to the formation of topo- U(x, y) ∝ O(Ω−3). This term breaks mirror symmetry logically non-trivial Floquet flat bands [109, 110]. The My and allows a relative shift in quasienergy between asymmetry ∆1 =6 ∆2 leads to asymmetric gaps at the the Dirac crossings at κ±, as shown schematically in fig- κ± points in the mBZ, as sketched in figure 2(b), where ure 2(a) for a spatially-uniform constant U. Because we plot the bands for TBG with a constant term of the the U(x, Ω) is odd in the x-coordinate, C2T and C3 are form ∆1(τ0 ± τ3) ⊗ σ3 added. The ∆1/2(x, Ω) position- also broken when taking the position dependence into dependence leads to breaking of C3 symmetry. + account. The term δw0(x, Ω)τ ⊗ σ0 (and its hermitian conju- 5

(a) (b) curate description of the quasienergies near the Floquet zone center is challenging to achieve with high-frequency expansions such as the Magnus expansion, which high- lights the strength of our approach. Crucially, the physics of the Floquet bands near the Floquet zone center is not obfuscated by negligible contributions from static high- energy bands which do not hybridize due to the weak drives considered here. Finally, the correction to the interlayer tunneling δw0(x, Ω) is bears resemblence to ef- (c) (d) fects one would expect from the relaxation of the driven lattice. Particularly, this term only affects the AA-type interlayer coupling w0, which reduces. One could ob- serve a similar effect if the size of AA-type patches were to shrink, which would also lead to the reduction in w0. Secondly if the interlayer distance in AA stacked regions increased, this would also lead to a similar reduction of w0. Therefore the periodic drive is able to mimic these effects. Finally, we address the term β(x, Ω)τ + ⊗ σ (and FIG. 2. (Color online) Sketch of the individual effects of the 3 new term generated by low-frequency and low-intensity cir- its hermitian conjugate) with real-space transformation cularly polarized light on the TBG quasienergies. The pa- properties β(−x, y) = β(x, y), Reβ(x, −y) = Reβ(x, y), ◦ β x, −y − β x, y rameters used are w0 = w1 = 110 meV, and θ = 1.2 . The and Im ( ) = Im ( ). To leading order, 2 −3 gray dashed curves correspond to the static case, while the β(x, Ω) = i(Aγa0/Ω) T11(x) sin θ + O(Ω ). Neglecting red curve indicates the effect introduced by the non-zero per- its position dependence, β preserves C2T and My. Tak- turbation introduced by light. ing the position dependence into account, β(x) breaks both C2T and My. Physically β(x, Ω) can be interpreted as a pseudo-spin dependent tunneling term. Therefore, in the weak-drive, small angle and low-frequency regime, ± gate) where τ = 1/2 (τ1 ± iτ2), Re δw0(x, −y) = circularly polarized light can introduce a collection of Re δw0(x, y), Im δw0(x, −y) = −Im δw0(x, y), symmetry-breaking processes beyond the reach of the δw0(−x, y) = δw0(x, y) introduces a correction to high-frequency limit. the tunneling amplitude w0, consistent with the sym- In addition to the small angle limit where Eq.(10) is metries of the static system, except C3. δw0(x, Ω) fulfilled let us also consider the opposite limit effectively renormalizes the Fermi velocity at the κ± points and can modify the position of the magic angles. p 2 2 w1 δw x, ≈ − Aγa / 2T x θ (k − κ+) + (k − κ−) 3 (13) To leading order, 0( Ω) ( 0 Ω) 11( ) cos( ), ~vF where T11(x, Ω) corresponds to the diagonal entry of the tunneling matrix Eq. (3). −1 −1 where ( − HL ± Ω) ≈ ( − Hg ± Ω) with In figure 2(d), we schematically show the effect of this term in the Floquet bands. Controlled drive pro- h(−θ/2, k − κ ) 0 H = − . (14) tocols to tune the Fermi velocity of the Floquet zone g 0 h(θ/2, k − κ ) center flat quasienergy bands have previously been pro- + posed [45]. For small angles, large drive frequency Ω In this case we find that Heff = H0 + HΩ + and small quasienergies Ω, this term constitutes the 3 2 second most relevant correction after ∆ (x, Ω). An ac- O A , A w1,2 1/2 kD kD γ

1 1 HΩ(k) = V (k, Ω)τ0 ⊗ σ0 + U(k, Ω)τ3 ⊗ σ0 + ∆1(k, Ω)(τ0 + τ3) ⊗ σ3 + ∆2(k, Ω)(τ0 − τ3) ⊗ σ3 (15) 2 2

The gaps are given as the interlayer bias is

2 Ω 2 − Ω2 + f (k)) 2 2 2 2 ∆1/2(k, Ω) 1/2 U(k, Ω) X m − Ω − |fm(k))| = , (16) = − (−1) (17) 2 2 2 A2a2γ2 2 Λ (, k, Ω) A a0γ Λ1/2(, k, Ω) 0 m=1 m 6

C2T C3 My quencies is applying a rotating frame transformation U X X x before the use of a high frequency Magnus expansion U(x) x x x [69, 77, 78]. To accomplish when a Hamiltonian has the U(k) X X X form H(t) = H0 + λV (t), one applies the unitary trans- ∆ x x R X U t e−iλ dtV (t) V t ∆(x) x x x formation ( ) = to remove ( ) to lowest ∆(k) x x order. A large term λV (t) in the Hamiltonian can be X traded this way for strongly oscillating terms [69]. This δω0 X x X approach allows treating regimes where λ is too large for δω0(x) X x X β X x X a Magnus approximation to be applicable, and is known β(x) x x x to give results that are more reliable than the Magnus expansion [69, 77, 78]. TABLE I. This table lists the symmetries that are broken First, we consider the simpler driven Dirac model for the different terms that can be generated for the case of −iΩt position dependence, momentum dependence, or if the term 0 kx − iky + λe HD = iΩt , (20) is constant. A checkmark means that the symmetry is pre- kx + iky + λe 0 served, while a cross that symmetry is broken. which also describes the upper layer of twisted bilayer graphene near the K point for w1 = w0 = 0, γ = a0 = 1, and κ± = 0 and very small θ. 2 Application of the unitary transformation U(t) = 2 2 2 R V (k, Ω) X − Ω − |fm(k))| −iλ dtV (t) = . (18) e followed by a zeroth order Magnus approxi- A2a2γ2 2 Λ (, k, Ω) 0 m=1 m mation leads to a Hamiltonian of the form T where f1/2(k) = f(R(∓θ/2)(k − κ∓)), with the prop- Heff,D = (B + Ry(τ)(κ1, κ2, 0) ) · σ, (21) 2 2 erty |f1/2(kx, −ky)| = |f2/1(k)| , and where Ry(τ) is a rotation matrix around the y-axis by an 2 angle Y 2 m 2 Λ1/2(, k, Ω) = f1/2(k)) − ( + (−1) Ω) , sJ (s) m=1 −1 1 τ = tan s , (22) (19) sJ0 (s) − J1 (s) + 2 with Λ1/2(, (kx, −ky), Ω) = Λ2/1(, k, Ω). This 4λ property implies that V ((kx, −ky), Ω) = V (k, Ω), s = Ω , and Jn(x) is the n-th Bessel function of the first U((kx, −ky), Ω) = −U(k, Ω), and ∆1/2((kx, −ky), Ω) = kind. The Hamiltonian has a constant field-like part with ∆2/1(k, Ω). Furthermore, U(k, Ω), V (k, Ω), and ∆1/2(k, Ω) are invariant under a C3 rotation of the mo- λ J s 2 2 1 ( ) mentum, since |f (C {k})| = |f (k)| . We find that Bx = − λ J0 (s) − , (23) 1/2 3 1/2 2 s not all terms appearing in Eq.(12) valid in the limit Eq.(10) are generated, and that they are momentum- By = 0, (24) dependent rather than position-dependent. A summary 1 B = λ (J (s) − 1) − λJ (s) , (25) of the results for what symmetries get broken by the dif- z s 0 1 ferent terms is given in Table I. The more general case, where neither condition Eq.(10) and momenta given by nor the opposite Eq.(13) are fulfilled, we can use the k¯x = kx(s + 2sJ0(s) − 2J1(s)) general form of P to find that an effective Hamilto- q 2 2 2 2 2s J0(s)(1−2J2(s))+s (1−2J2(s))+4(s +1)J1(s) nian has the same structure as Eq.(12). However, all (s+2sJ (s)−2J (s))2 terms have an additional momentum dependence (e.g. × 0 1 , (26) 2s ∆1,2(x, Ω) → ∆1,2(x, k, Ω) etc.). While it is possible to 1 2J1(s) determine that HΩ has this structure generally, the coef- k¯y = ky + 1 . (27) ficients are too cumbersome to compute and are therefore 2 s not discussed. By inspecting k¯x and k¯y, we realize that kx and ky are not treated on equal grounds in this approximation. IV. INTERMEDIATE DRIVE REGIME Specifically, the Fermi velocity has become anisotropic. The quasi-energy spectrum is not rotationally symmet- λ k¯ ≈ A. Issues with the usual form of the rotating frame ric for large driving . Specifically if we expand x,y s2 s4 transformation kx,y 1 − 16 ∓ 384 we see that the anisotropic behaviour appears at fourth order in s- that is for relatively large λ. A standard approach for treating systems subjected This is in qualitative disagreement with an exact numer- to intermediately strong drives and intermediate fre- ical calculations, which present rotationally-symmetric 7

−iΩt quasi-energies. Since the problem already appears in the a0(kx + iky)J1(2a0A/3)e . For 2a0A/3 not too large Dirac case, we can therefore expect the rotating frame ap- compared with unit, J1(2a0A/3) 1. Therefore, terms proximation to also produce unphysical artifacts for the like kiJ1(2a0A/3) are higher order and can be neglected. more complicated problem of twisted bilayer graphene. We will thus work with the approximation It is important to note that the same type of unphysical anisotropy already appears on the level of a first order 2a0A 2a0A iΩt f(k − A) ≈ a0(kx − iky)J0 − 3J1 e . Magnus expansion [65]. Therefore, a more careful partial 3 3 resummation of the Magnus expansion is needed. (28) This type of approximation is reasonable for small angles and 2a0A/3 . 1. B. A better choice of unitary transformation After application of this approximation we can read- ily improve on the Van Vleck approximation, which In order to avoid introducing unphysical terms in we will use to compare our results from the rotating the effective Floquet Hamiltonian, we write the time- wave approximation. The effective Floquet Hamilto- dependent Hamiltonian as H(t) = H0 + λV1(t) + λV2(t) nian keeps the same structure as previously obtained, vV with [Vi(t),Vi(t1)] = 0 and apply the modified uni- H = HL − ∆τ0 ⊗ σ3 with gap R R eff tary transformation U(t) = e−i dtV1(t)e−i dtV2(t) with 2 2 V1(t) = λ cos(Ωt)σ1 and V2(t) = λ sin(Ωt)σ2. There is 9γ 2Aa0 ∆ = J1 (29) an associated arbitrariness in the exact form of this uni- Ω 3 tary transformation arising from the choice of V1 and V2. ∗ However, given our implicit Floquet gauge choice t = 0, and a renormalized Fermi velocity in a time-ordered exponential that removes all of V1 + V2 we make the smaller error by removing a V1 first, that Aa ∗ 2 0 is the larger of the two at t = 0. In the Dirac model v˜F = vF J0 . (30) 3 this choice can be justified even better better apostiori by realizing that it restores the rotational invariance in momentum space. We will make an analogous choice of unitary transfor- D. Rotating frame Hamiltonian mations for the TBG case in section IV D, where we will explicitly demonstrate that the anisotropy in the Fermi In this section, we will derive an effective Floquet R velocity is not present. Hamiltonian using a rotating frame approach, Heff, with an improved unitary transformation. Then, we compare the quasienergies obtained with the ones derived from C. Improved Van Vleck approximation vV the Van Vleck Hamiltonian Heff . We write the time-dependent Hamiltonian for twisted In this section, we identify a procedure to improve the bilayer graphene as H(t) = HL +V1(t)+V2(t), where the Van Vleck expansion used to obtain an effective Floquet time dependent potentials are given as Hamiltonian which we will use as a baseline to compare our improved rotating frame effective Hamiltonian. (−θ/2) ! σ1 0 For small twist angles θ it is sensible to treat k as V1(t) = −3J1(2a0A/3) cos(Ωt) (31) 0 σ(θ/2) a small parameter because the dimensions of the moiré 1 Brillouin zone are proportional to sin(θ/2). Therefore, we (−θ/2) ! σ2 0 V2(t) = −3J1(2a0A/3) sin(Ωt) , (32) may approximate f(k−A) ≈ f(−A)+(kx(∂kx f)(−A)+ (θ/2) 0 σ2 ky(∂ky f)(−A)). In the weak-strength drive regime, A a 0, we employed a simple Taylor expansion. How- θ θ θ −i σ3 i σ3 ever, in order to capture the effect of stronger drives, where σi = e 2 σie 2 . After applying the unitary R R we need to improve our approach. For this, we per- transformation U(t) = e−i dtV1(t)e−i dtV2(t) and after form a Fourier series in terms of eiΩnt instead. The re- taking an average over one period 2π/Ω we find the fol- sult to first order in Fourier components has the form lowing effective Hamiltonian for twisted bilayer graphene iΩt f(k − A) ≈ a0(kx − iky)J0(2a0A/3) − 3J1(2a0A/3)e − that is subjected to circularly polarized light

−i θ R (e 2 v˜F (k − κ−) + ∆êz) · σ T˜(r) † Heff = R † i θ R , (33) T˜ (r) (e 2 v˜F (k − κ+) + ∆êz) · σ

wheree ˆz is a unit vector in z-direction and σ is a vector of Pauli matrices. The unitary transformation 2Aa 3γJ 0  1( 3 ) (θ/2)  iσ e Ω 2 0 R 2Aa , =  3γJ 0  (34) 1( 3 ) iσ(−θ/2) 0 e Ω y 8 allows us to cast the Hamiltonian in more readable form. From this unitary transformation, one can directly identify the origin of the spurious anisotropy in momentum 1.00 that one would ﬁnd in a Magnus expansion approach. Particularly, an expansion of R for large frequencies un- 0.95 avoidably leads to such issue. 0.90 We ﬁnd that the Fermi velocity has been renormalized to 0.85

0.80 2Aa0 ! 2Aa0 6γJ1 3 v˜F = vF J0 J0 . (35) 0.2 0.4 0.6 0.8 1.0 3 Ω FIG. 4. Ratio ∆˜ /∆ of the renormalized gap ∆˜ and the gap of from the Van Vleck ∆ for driving frequency Ω = 2γ. In figure 3, we show a plot of the Fermi velocity and compare with to the Fermi velocity from the improved Van We find that also in this case that there is considerable vV Vleck approximation Heff . We find that the renormal- difference (∼ 10% in some regions in parameter space) ization of the Fermi velocity is ∼ 10% in some regions even for relatively large driving frequencies Ω = 2γ. even for relatively high frequencies. vV R The most striking difference between Heff and Heff ap- R pears in the tunneling sector, where Heff contains renormalized interlayer hopping

1 ˜ X −iblx ˜ 1.00 T (x) = e (Tl − iβσ3) l=−1 0.95 2πn 2πn T˜n =w ˜012 +w ˜1 cos σ1 + sin σ2 , 0.90 3 3 (37) 0.85 with improved van Vleck 0.80 2Aa0 ! 6γJ1 3 improved rot. frame w˜1 = w1J0 (38) 0.75 Ω √ " 2Aa0 ! !# 2 θ 6 2γJ1 3 0.2 0.4 0.6 0.8 1.0 w˜0 = w0 1 + sin J0 − 1 , 2 Ω (39) FIG. 3. Fermi velocity normalized to bare vF for relatively γ 1 high frequency drives Ω = 2 . In blue, we show the Van Vleck result and in orange the renormalized result employing our and a new imaginary term in the AA interlayer coupling improved rotating frame approximation. √ 2Aa0 !! 1 6 2γJ1 3 β = sin(θ) 1 − J0 . (40) 2 ω

R In the notation of the previous section III the new cou- Furthermore, in H , the quasienergy gap that is renor- + eff pling term enters as −iβτ ⊗ σ3 and is position depen- malized to dent. The new dynamically-generated tunneling compo- nent β breaks C3, the approximate particle-hole symme- C C T M √ try , 2 and reflection y symmetries. 2Aa0 ! R Aa 6 2γJ In figure 5, we compare our results using Heff in Eq. ˜ 3 2 0 1 3 ∆ = √ γJ1 J1 . (36) (33) to exact numeric results obtained employed an ex- 2 3 Ω tended space approach [65]. We use the improved Van vV Vleck approximation Heff = HL − ∆τ0 ⊗ σ3 as a bench- mark. We find that the Van Vleck approximation is only valid until a0A ≈ 0.4, while the new approximation works A comparison with the Van Vleck result is shown in well until a0A ≈ 0.8. The approximation therefore has figure 4. double the range of validity and therefore is more reliable. 9

FIG. 5. Quasi-energy band structure. The dashed red curves correspond to the exact result, in blue the improved Van Vleck approximation and in black the rotation frame transformation. The parameters used are Ω = 2γ, w1 = w0 = 110 meV, γ = 2364 meV and θ = 1.05◦.

The same observation can be made a bit more lucidly - 0.30 Van Vleck albeit losing much information- if we compute the relative error of the gap at the K point (gexact − gapprox)/gexact, 0.25 Rotating Frame where gexact is the “exact” numerical gap at the K point 0.20 and gapprox is the gap for an approximation. For both approximations the result is shown in Fig. 6 below. 0.15 0.10 0.05 0.00 0.0 0.2 0.4 0.6 0.8 1.0

FIG. 6. Plot of the relative error for the gap at the K point of the moir´eBrillouin zone. In blue we present our results from the Van Vleck approximation and in orange the results from our improved rotating frame approximation. 10

It is clear from both plots that the rotating frame ap- that only appears in AA regions. proximation derived in this paper is far more reliable than The rotating frame Hamiltonian, valid for strong drives the Van Vleck approximation. and intermediate drive frequencies reveals that the gap at the Floquet zone center, the Fermi velocity, and the interlayer-coupling strengths are renormalized. This effective Hamiltonian is useful for numerical implementa- V. CONCLUSION AND OUTLOOK tion of the quasienergy band structure and posses a wide range of validity. This would make it useful for applica- We have introduced two new effective Floquet Hamil- tions where an extended space calculation may be too ex- tonians that describe twisted bilayer graphene under the pensive. For instance if one studies the effect of disorder influence of circularly polarized light. The Hamiltoni- additional disorder averages make calculations expensive ans are applicable in the regimes where the ordinary Van and therefore it might be more feasible to do these cal- Vleck approximation fails. We found that the weak drive culations using the effective Hamiltonian we presented strength Hamiltonian, valid even in the low-frequency rather than resorting to a full treatment in an extended regime, gives insight into which new terms a periodic space picture. drive can generate well beyond the regime of validity of any other approximation scheme. The usefulness of these scheme is limited by the challenge imposed by the com- VI. ACKNOWLEDGEMENTS plexity of the terms derived and the self-consistent nature of the low-frequency regime. An important physical We thank Fengcheng Wu for useful discussions. This effect of the drive in these regime is a renormalization research was primarily supported by the National Sci- of the interlayer-coupling of the AA-type. This makes ence Foundation through the Center for Dynamics and it possible to mimic the effects of some otherwise diffi- Control of Materials: an NSF MRSEC under Coopera- cult to achieve structural reorganizations - for instance a tive Agreement No. DMR-1720595. Partial support was change in the distance between the two graphene layers from NSF Grant No. DMR-1949701.

[1] Fengcheng Wu, A.H. MacDonald, and Ivar Mar- lator in twisted bilayer graphene,” Phys. Rev. B 100 tin, “Theory of phonon-mediated superconductivity in (2019). twisted bilayer graphene,” Phys. Rev. Letters 121 [8] Francisco Guinea and Niels R. Walet, “Electrostatic (2018). effects, band distortions, and superconductivity in [2] Yuan Cao, Valla Fatemi, Shiang Fang, Kenji Watan- twisted graphene bilayers,” Proceedings of the National abe, Takashi Taniguchi, Efthimios Kaxiras, and Pablo Academy of Sciences 115, 13174–13179 (2018). Jarillo-Herrero, “Unconventional superconductivity in [9] Biao Lian, Zhijun Wang, and B. Andrei Bernevig, magic-angle graphene superlattices,” Nature 556, 43– “Twisted bilayer graphene: A phonon-driven supercon- 50 (2018), article. ductor,” Phys. Rev. Lett. 122, 257002 (2019). [3] Kan-Ting Tsai, Xi Zhang, Ziyan Zhu, Yujie Luo, [10] Sujay Ray, Jeil Jung, and Tanmoy Das, “Wannier pairs Stephen Carr, Mitchell Luskin, Efthimios Kaxiras, and in superconducting twisted bilayer graphene and related Ke Wang, “Correlated superconducting and insulating systems,” Phys. Rev. B 99, 134515 (2019). states in twisted trilayer graphene moire of moire su- [11] M. J. Calderon and E. Bascones, “Correlated states in perlattices,” (2019), arXiv:1912.03375 [cond-mat.mes- magic angle twisted bilayer graphene under the optical hall]. conductivity scrutiny,” (2019), arXiv:1912.09935 [cond- [4] Emilio Codecido, Qiyue Wang, Ryan Koester, Shi Che, mat.str-el]. Haidong Tian, Rui Lv, Son Tran, Kenji Watanabe, [12] Yu Saito, Jingyuan Ge, Kenji Watanabe, Takashi Takashi Taniguchi, Fan Zhang, Marc Bockrath, and Taniguchi, and Andrea F. Young, “Decoupling super- Chun Ning Lau, “Correlated insulating and supercon- conductivity and correlated insulators in twisted bilayer ducting states in twisted bilayer graphene below the graphene,” (2019), arXiv:1911.13302 [cond-mat.mes- magic angle,” Science Advances 5 (2019). hall]. [5] Matthew Yankowitz, Shaowen Chen, Hryhoriy Polshyn, [13] Petr Stepanov, Ipsita Das, Xiaobo Lu, Ali Fahimniya, Yuxuan Zhang, K. Watanabe, T. Taniguchi, David Kenji Watanabe, Takashi Taniguchi, Frank H. L. Graf, Andrea F. Young, and Cory R. Dean, “Tuning Koppens, Johannes Lischner, Leonid Levitov, and superconductivity in twisted bilayer graphene,” Science Dmitri K. Efetov, “The interplay of insulating and su- 363, 1059–1064 (2019). perconducting orders in magic-angle graphene bilayers,” [6] Dmitry V. Chichinadze, Laura Classen, and An- (2019), arXiv:1911.09198 [cond-mat.supr-con]. drey V. Chubukov, “Nematic superconductivity in [14] Jian Kang and Oskar Vafek, “Strong coupling phases of twisted bilayer graphene,” (2019), arXiv:1910.07379 partially filled twisted bilayer graphene narrow bands,” [cond-mat.supr-con]. Phys. Rev. Letters 122 (2019). [7] Yang-Zhi Chou, Yu-Ping Lin, Sankar Das Sarma, and [15] G. E. Volovik, “Graphite, graphene, and the flat band Rahul M. Nandkishore, “Superconductor versus insu- superconductivity,” JETP Letters 107, 516–517 (2018). 11

[16] Hoi Chun Po, Liujun Zou, Ashvin Vishwanath, and waals domain walls in bilayer graphene,” Journal of T. Senthil, “Origin of mott insulating behavior and Physics: Condensed Matter 29, 425303 (2017). superconductivity in twisted bilayer graphene,” Phys. [32] Yonglong Xie, Biao Lian, Berthold Jck, Xiaomeng Liu, Rev. X 8, 031089 (2018). Cheng-Li Chiu, Kenji Watanabe, Takashi Taniguchi, [17] Masayuki Ochi, Mikito Koshino, and Kazuhiko Kuroki, B. Andrei Bernevig, and Ali Yazdani, “Spectroscopic “Possible correlated insulating states in magic-angle signatures of many-body correlations in magic-angle twisted bilayer graphene under strongly competing in- twisted bilayer graphene,” Nature 572, 101105 (2019). teractions,” Phys. Rev. B 98 (2018). [33] Patrick A. Lee, Naoto Nagaosa, and Xiao-Gang Wen, [18] J. Gonzlez and T. Stauber, “Kohn-luttinger supercon- “Doping a mott insulator: Physics of high-temperature ductivity in twisted bilayer graphene,” Phys. Rev. Let- superconductivity,” Rev. Mod. Phys. 78, 17–85 (2006). ters 122 (2019). [34] Yuan Cao, Valla Fatemi, Ahmet Demir, Shiang Fang, [19] Yury Sherkunov and Joseph J. Betouras, “Electronic Spencer L. Tomarken, Jason Y. Luo, Javier D. Sanchez- phases in twisted bilayer graphene at magic angles as a Yamagishi, Kenji Watanabe, Takashi Taniguchi, result of van hove singularities and interactions,” Phys. Efthimios Kaxiras, Ray C. Ashoori, and Pablo Jarillo- Rev. B 98 (2018). Herrero, “Correlated insulator behaviour at half-filling [20] Evan Laksono, Jia Ning Leaw, Alexander Reaves, Man- in magic-angle graphene superlattices,” Nature 556, 80– raaj Singh, Xinyun Wang, Shaffique Adam, and Xingyu 84 (2018). Gu, “Singlet superconductivity enhanced by charge or- [35] Yi Zhang, Kun Jiang, Ziqiang Wang, and Fuchun der in nested twisted bilayer graphene fermi surfaces,” Zhang, “Correlated insulating phases of twisted bilayer Solid State Communications 282, 38–44 (2018). graphene at commensurate filling fractions: a hartree- [21] JörnW. F. Venderbos and Rafael M. Fernandes, “Corre- fock study,” (2020), arXiv:2001.02476 [cond-mat.str-el]. lations and electronic order in a two-orbital honeycomb [36] Dillon Wong, Kevin P. Nuckolls, Myungchul Oh, Biao lattice model for twisted bilayer graphene,” Phys. Rev. Lian, Yonglong Xie, Sangjun Jeon, Kenji Watanabe, B 98 (2018). Takashi Taniguchi, B. Andrei Bernevig, and Ali [22] S. Shallcross, S. Sharma, E. Kandelaki, and Yazdani, “Cascade of transitions between the corre- O. A. Pankratov, “Electronic structure of turbostratic lated electronic states of magic-angle twisted bilayer graphene,” Phys. Rev. B 81, 165105 (2010). graphene,” (2019), arXiv:1912.06145 [cond-mat.mes- [23] Tymoteusz Salamon, Alessio Celi, Ravindra W. Chha- hall]. jlany, Irne Frrot, Maciej Lewenstein, Leticia Tarruell, [37] Aaron L. Sharpe, Eli J. Fox, Arthur W. Barnard, Joe and Debraj Rakshit, “Simulating twistronics without a Finney, Kenji Watanabe, Takashi Taniguchi, M. A. twist,” (2019), arXiv:1912.12736 [cond-mat.quant-gas]. Kastner, and David Goldhaber-Gordon, “Emergent fer- [24] D. Weckbecker, S. Shallcross, M. Fleischmann, N. Ray, romagnetism near three-quarters filling in twisted bi- S. Sharma, and O. Pankratov, “Low-energy theory for layer graphene,” Science 365, 605–608 (2019). the graphene twist bilayer,” Phys. Rev. B 93, 035452 [38] Kangjun Seo, Valeri N. Kotov, and Bruno Uchoa, “Fer- (2016). romagnetic mott state in twisted graphene bilayers at [25] F. Rost, R. Gupta, M. Fleischmann, D. Weckbecker, the magic angle,” Phys. Rev. Letters 122 (2019). N. Ray, J. Olivares, M. Vogl, S. Sharma, O. Pankratov, [39] Rafi Bistritzer and Allan H. MacDonald, “Moirébands and S. Shallcross, “Nonperturbative theory of effective in twisted double-layer graphene,” Proceedings of the hamiltonians for deformations in two-dimensional ma- National Academy of Sciences 108, 12233–12237 (2011). terials: Moire systems and dislocations,” Phys. Rev. B [40] Kyounghwan Kim, Ashley DaSilva, Shengqiang Huang, 100 (2019). Babak Fallahazad, Stefano Larentis, Takashi Taniguchi, [26] M. Vogl, O. Pankratov, and S. Shallcross, “Semi- Kenji Watanabe, Brian J. LeRoy, Allan H. MacDonald, classics for matrix hamiltonians: The gutzwiller trace and Emanuel Tutuc, “Tunable moirébands and strong formula with applications to graphene-type systems,” correlations in small-twist-angle bilayer graphene,” Pro- Phys. Rev. B 96, 035442 (2017). ceedings of the National Academy of Sciences 114, [27] Yang Cheng, Chen Huang, Hao Hong, Zixun Zhao, and 3364–3369 (2017). Kaihui Liu, “Emerging properties of two-dimensional [41] M. Iqbal Bakti Utama, Roland J. Koch, Kyunghoon twisted bilayer materials,” Chinese Physics B 28, Lee, Nicolas Leconte, Hongyuan Li, Sihan Zhao, Lili 107304 (2019). Jiang, Jiayi Zhu, Kenji Watanabe, Takashi Taniguchi, [28] Kaihui Liu, Liming Zhang, Ting Cao, Chenhao Jin, Di- Paul D. Ashby, Alexander Weber-Bargioni, Alex Zettl, ana Qiu, Qin Zhou, Alex Zettl, Peidong Yang, Steve G. Chris Jozwiak, Jeil Jung, Eli Rotenberg, Aaron Bost- Louie, and Feng Wang, “Evolution of interlayer cou- wick, and Feng Wang, “Visualization of the flat elec- pling in twisted molybdenum disulfide bilayers,” Nature tronic band in twisted bilayer graphene near the magic Communications 5 (2014). angle twist,” (2019), arXiv:1912.00587 [cond-mat.mes- [29] Jhao-Ying Wu, Wu-Pei Su, and Godfrey Gumbs, hall]. “Anomalous magneto-transport properties of bilayer [42] Stephen Carr, Shiang Fang, Pablo Jarillo-Herrero, and phosphorene,” (2019), arXiv:1912.10219 [cond- Efthimios Kaxiras, “Pressure dependence of the magic mat.mes-hall]. twist angle in graphene superlattices,” Phys. Rev. B 98, [30] Ce Shang, Adel Abbout, Xiaoning Zang, Udo Schwin- 085144 (2018). genschlogl, and Aurelien Manchon, “Artificial gauge [43] Bheema Lingam Chittari, Nicolas Leconte, Srivani Jav- fields and topological insulators in moire superlattices,” vaji, and Jeil Jung, “Pressure induced compression of (2019), arXiv:1912.00447 [cond-mat.quant-gas]. flatbands in twisted bilayer graphene,” Electronic Struc- [31] H M Abdullah, B Van Duppen, M Zarenia, H Bahlouli, ture 1, 015001 (2018). and F M Peeters, “Quantum transport across van der 12

[44] Matthew Yankowitz, Jeil Jung, Evan Laksono, Nico- [61] Dmitry A. Abanin, Wojciech De Roeck, Wen Wei las Leconte, Bheema L. Chittari, K. Watanabe, Ho, and Franois Huveneers, “Effective hamiltonians, T. Taniguchi, Shaffique Adam, David Graf, and prethermalization, and slow energy absorption in peri- Cory R. Dean, “Dynamic band-structure tuning of odically driven many-body systems,” Phys. Rev. B 95, graphene moire superlattices with pressure,” Nature 014112 (2017). 557, 404–408 (2018). [62] Takashi Mori, Tomotaka Kuwahara, and Keiji Saito, [45] Michael Vogl, Martin Rodriguez-Vega, and Gre- “Rigorous bound on energy absorption and generic re- gory A. Fiete, “Tuning the magic angle of twisted bi- laxation in periodically driven quantum systems,” Phys. layer graphene at the exit of a waveguide,” (2020), Rev. Lett. 116, 120401 (2016). arXiv:2001.04416 [cond-mat.mes-hall]. [63] Dominic V. Else, Bela Bauer, and Chetan Nayak, [46] Anatoli Polkovnikov, Krishnendu Sengupta, Alessan- “Prethermal phases of matter protected by time- dro Silva, and Mukund Vengalattore, “Colloquium: translation symmetry,” Phys. Rev. X 7, 011026 (2017). Nonequilibrium dynamics of closed interacting quantum [64] Elena Canovi, Marcus Kollar, and Martin Eckstein, systems,” Rev. Mod. Phys. 83, 863–883 (2011). “Stroboscopic prethermalization in weakly interacting [47] AndréEckardt, “Colloquium: Atomic quantum gases periodically driven systems,” Phys. Rev. E 93, 012130 in periodically driven optical lattices,” Rev. Mod. Phys. (2016). 89, 011004 (2017). [65] André Eckardt and Egidijus Anisimovas, “High- [48] Immanuel Bloch, Jean Dalibard, and Wilhelm Zwerger, frequency approximation for periodically driven quan- “Many-body physics with ultracold gases,” Rev. Mod. tum systems from a floquet-space perspective,” New Phys. 80, 885–964 (2008). Journal of Physics 17, 093039 (2015). [49] Jean Dalibard, Fabrice Gerbier, Gediminas Juzeli¯unas, [66] S. Blanes, F. Casas, J.A. Oteo, and J. Ros, “The Mag- and Patrik Ohberg,¨ “Colloquium: Artificial gauge po- nus expansion and some of its applications,” Phys. Rep. tentials for neutral atoms,” Rev. Mod. Phys. 83, 1523– 470, 151 – 238 (2009). 1543 (2011). [67] E.B. Fel’dman, “On the convergence of the Magnus ex- [50] D. N. Basov, R. D. Averitt, and D. Hsieh, “Towards pansion for spin systems in periodic magnetic fields,” properties on demand in quantum materials,” Nat. Mat. Phys. Lett. A 104, 479 – 481 (1984). 16, 1077 (2017). [68] Wilhelm Magnus, “On the exponential solution of differ- [51] J. Zhang and R.D. Averitt, “Dynamics and control in ential equations for a linear operator,” Commun. Pure complex transition metal oxides,” Annu. Rev. Mater. Appl. Math. 7, 649–673 (1954). Res. 44, 19–43 (2014). [69] Marin Bukov, Luca DAlessio, and Anatoli Polkovnikov, [52] D. N. Basov, Richard D. Averitt, Dirk van der Marel, “Universal high-frequency behavior of periodically Martin Dressel, and Kristjan Haule, “Electrodynamics driven systems: from dynamical stabilization to floquet of correlated electron materials,” Rev. Mod. Phys. 83, engineering,” Advances in Physics 64, 139226 (2015). 471–541 (2011). [70] Saar Rahav, Ido Gilary, and Shmuel Fishman, “Ef- [53] Claudio Giannetti, Massimo Capone, Daniele Fausti, fective Hamiltonians for periodically driven systems,” Michele Fabrizio, Fulvio Parmigiani, and Dragan Mi- Phys. Rev. A 68, 013820 (2003). hailovic, “Ultrafast optical spectroscopy of strongly cor- [71] N. Goldman and J. Dalibard, “Periodically driven quan- related materials and high-temperature superconduc- tum systems: Effective hamiltonians and engineered tors: a non-equilibrium approach,” Adv. Phys. 65, 58– gauge fields,” Phys. Rev. X 4, 031027 (2014). 238 (2016). [72] A. P. Itin and M. I. Katsnelson, “Effective Hamilto- [54] M Gandolfi, G L Celardo, F Borgonovi, G Ferrini, nians for rapidly driven many-body lattice systems: A Avella, F Banfi, and C Giannetti, “Emergent ultra- Induced exchange interactions and density-dependent fast phenomena in correlated oxides and heterostruc- hoppings,” Phys. Rev. Lett. 115, 075301 (2015). tures,” Phys. Scr. 92, 034004 (2017). [73] Takahiro Mikami, Sota Kitamura, Kenji Yasuda, Naoto [55] O. V. Kibis, “Metal-insulator transition in graphene in- Tsuji, Takashi Oka, and Hideo Aoki, “Brillouin-wigner duced by circularly polarized photons,” Phys. Rev. B theory for high-frequency expansion in periodically 81, 165433 (2010). driven systems: Application to floquet topological in- [56] O. V. Kibis, S. Morina, K. Dini, and I. A. Shelykh, sulators,” Phys. Rev. B 93, 144307 (2016). “Magnetoelectronic properties of graphene dressed by a [74] Priyanka Mohan, Ruchi Saxena, Arijit Kundu, and high-frequency field,” Phys. Rev. B 93, 115420 (2016). Sumathi Rao, “Brillouin-Wigner theory for Floquet [57] O. V. Kibis, K. Dini, I. V. Iorsh, and I. A. Shelykh, topological phase transitions in spin-orbit-coupled ma- “All-optical band engineering of gapped dirac materi- terials,” Phys. Rev. B 94, 235419 (2016). als,” Phys. Rev. B 95, 125401 (2017). [75] Marin Bukov, Michael Kolodrubetz, and Anatoli [58] I. V. Iorsh, K. Dini, O. V. Kibis, and I. A. She- Polkovnikov, “Schrieffer-Wolff transformation for peri- lykh, “Optically induced lifshitz transition in bilayer odically driven systems: Strongly correlated systems graphene,” Phys. Rev. B 96 (2017). with artificial gauge fields,” Phys. Rev. Lett. 116, [59] R. Moessner and S. L. Sondhi, “Equilibration and order 125301 (2016). in quantum Floquet matter,” Nat. Phys. 13, 424 (2017). [76] M. Matti Maricq, “Application of average Hamiltonian [60] Dmitry A. Abanin, Wojciech De Roeck, and Fran ¸cois theory to the NMR of solids,” Phys. Rev. B 25, 6622– Huveneers, “Exponentially slow heating in periodically 6632 (1982). driven many-body systems,” Phys. Rev. Lett. 115, [77] Michael Vogl, Pontus Laurell, Aaron D. Barr, and 256803 (2015). Gregory A. Fiete, “Flow equation approach to periodically driven quantum systems,” Phys. Rev. X 9, 021037 (2019). 13

[78] Michael Vogl, Pontus Laurell, Aaron D. Barr, and Gre- [95] Liang Jiang et al., “Majorana fermions in equilibrium gory A. Fiete, “Analog of hamilton-jacobi theory for the and in driven cold-atom quantum wires,” Phys. Rev. time-evolution operator,” Phys. Rev. A 100 (2019). Lett. 106, 220402 (2011). [79] Michael Vogl, Martin Rodriguez-Vega, and Gre- [96] Zhenghao Gu, H. A. Fertig, Daniel P. Arovas, and Assa gory A. Fiete, “Effective floquet hamiltonian in the low- Auerbach, “Floquet spectrum and transport through frequency regime,” Phys. Rev. B 101, 024303 (2020). an irradiated graphene ribbon,” Phys. Rev. Lett. 107, [80] M Rodriguez-Vega, M Lentz, and B Seradjeh, “Floquet 216601 (2011). perturbation theory: formalism and application to low- [97] P. M. Perez-Piskunow, Gonzalo Usaj, C. A. Balseiro, frequency limit,” New Journal of Physics 20, 093022 and L. E. F. Foa Torres, “Floquet chiral edge states in (2018). graphene,” Phys. Rev. B 89, 121401 (2014). [81] Hanna Martiskainen and Nimrod Moiseyev, “Perturba- [98] Gonzalo Usaj, P. M. Perez-Piskunow, L. E. F. Foa Tor- tion theory for quasienergy floquet solutions in the low- res, and C. A. Balseiro, “Irradiated graphene as a tun- frequency regime of the oscillating electric field,” Phys. able Floquet topological insulator,” Phys. Rev. B 90, Rev. A 91, 023416 (2015). 115423 (2014). [82] Gustavo Rigolin, Gerardo Ortiz, and Victor Hugo [99] P. M. Perez-Piskunow, L. E. F. Foa Torres, and Gon- Ponce, “Beyond the quantum adiabatic approxima- zalo Usaj, “Hierarchy of floquet gaps and edge states for tion: Adiabatic perturbation theory,” Phys. Rev. A 78, driven honeycomb lattices,” Phys. Rev. A 91, 043625 052508 (2008). (2015). [83] M. Weinberg, C. Olschläger,¨ C. Sträter, S. Prelle, [100] H. L. Calvo, L. E. F. Foa Torres, P. M. Perez-Piskunow, A. Eckardt, K. Sengstock, and J. Simonet, “Multipho- C. A. Balseiro, and Gonzalo Usaj, “Floquet interface ton interband excitations of quantum gases in driven states in illuminated three-dimensional topological in- optical lattices,” Phys. Rev. A 92, 043621 (2015). sulators,” Phys. Rev. B 91, 241404 (2015). [84] Li Jia-Ming, He Kun-Huan, Shi Zhong-Feng, Gao Hui- [101] Bhaskar Mukherjee, Priyanka Mohan, Diptiman Sen, Yuan, and Jiang Yi-Min, “Synthesis, crystal struc- and K. Sengupta, “Low-frequency phase diagram of ir- 1 tures, and thermal and spectroscopic properties of two radiated graphene and a periodically driven spin- 2 xy cd(ii) metal-organic frameworks with a versatile ligand.” chain,” Phys. Rev. B 97, 205415 (2018). Zeitschrift fr Naturforschung B 71, 909–917 (2016). [102] Iliya Esin, Mark S. Rudner, Gil Refael, and Netanel H. [85] Albert Verdeny, Andreas Mielke, and Florian Mintert, Lindner, “Quantized transport and steady states of flo- “Accurate effective hamiltonians via unitary flow in floquet topological insulators,” Phys. Rev. B 97, 245401 quet space,” Phys. Rev. Lett. 111, 175301 (2013). (2018). [86] Juan Carlos Sandoval-Santana, Victor Guadalupe [103] Mark S. Rudner and Netanel H. Lindner, “Floquet topo- Ibarra-Sierra, Jos Luis Cardoso, Alejandro logical insulators: from band structure engineering to Kunold, Pedro Roman-Taboada, and Gerardo novel non-equilibrium quantum phenomena,” arXiv e- Naumis, “Method for finding the exact effec- prints , arXiv:1909.02008 (2019). tive hamiltonian of time-driven quantum sys- [104] Hossein Dehghani, Takashi Oka, and Aditi Mitra, “Out- tems,” Annalen der Physik 531, 1900035 (2019), of-equilibrium electrons and the hall conductance of a https://www.onlinelibrary.wiley.com/doi/pdf/10.1002/andp.201900035.floquet topological insulator,” Phys. Rev. B 91, 155422 [87] Takashi Oka and Hideo Aoki, “Photovoltaic hall effect (2015). in graphene,” Phys. Rev. B 79, 081406 (2009). [105] Hossein Dehghani and Aditi Mitra, “Optical hall con- [88] J. W. McIver, B. Schulte, F.-U. Stein, T. Matsuyama, ductivity of a floquet topological insulator,” Phys. Rev. G. Jotzu, G. Meier, and A. Cavalleri, “Light-induced B 92, 165111 (2015). anomalous hall effect in graphene,” Nature Physics 16, [106] Hossein Dehghani and Aditi Mitra, “Occupation proba- 38–41 (2020). bilities and current densities of bulk and edge states of a [89] Mark S. Rudner, Netanel H. Lindner, Erez Berg, floquet topological insulator,” Phys. Rev. B 93, 205437 and Michael Levin, “Anomalous edge states and the (2016). bulk-edge correspondence for periodically driven two- [107] Jelena Klinovaja, Peter Stano, and Daniel Loss, dimensional systems,” Phys. Rev. X 3, 031005 (2013). “Topological Floquet phases in driven coupled rashba [90] Netanel H. Lindner, Gil Refael, and Victor Galitski, nanowires,” Phys. Rev. Lett. 116, 176401 (2016). “Floquet topological insulator in semiconductor quan- [108] Gabriel E. Topp, Gregor Jotzu, James W. McIver, Lede tum wells,” Nature Physics 7, 490–495 (2011). Xian, Angel Rubio, and Michael A. Sentef, “Topological [91] Qing-Jun Tong, Jun-Hong An, Jiangbin Gong, Hong- floquet engineering of twisted bilayer graphene,” Phys. Gang Luo, and C. H. Oh, “Generating many majorana Rev. Research 1, 023031 (2019). modes via periodic driving: A superconductor model,” [109] Yantao Li, H. A. Fertig, and Babak Seradjeh, “Floquet- Phys. Rev. B 87, 201109 (2013). engineered topological flat bands in irradiated twisted [92] Manisha Thakurathi, Aavishkar A. Patel, Diptiman bilayer graphene,” (2019), arXiv:1910.04711 [cond- Sen, and Amit Dutta, “Floquet generation of majo- mat.str-el]. rana end modes and topological invariants,” Phys. Rev. [110] Or Katz, Gil Refael, and Netanel H. Lindner, “Flo- B 88, 155133 (2013). quet flat-band engineering of twisted bilayer graphene,” [93] Arijit Kundu and Babak Seradjeh, “Transport signa- (2019), arXiv:1910.13510 [cond-mat.str-el]. tures of floquet majorana fermions in driven topological [111] E. Perfetto and G. Stefanucci, “Some exact properties superconductors,” Phys. Rev. Lett. 111, 136402 (2013). of the nonequilibrium response function for transient [94] Mikael C. Rechtsman et al., “Photonic floquet topolog- photoabsorption,” Phys. Rev. A 91, 033416 (2015). ical insulators,” Nature 496, 196–200 (2013). [112] Umberto De Giovannini and Hannes Huebener, “Flo- quet analysis of excitations in materials,” Journal of 14

Physics: Materials (2019). Appendix A: The low frequency Hamiltonian [113] Takahiro Mikami, Sota Kitamura, Kenji Yasuda, Naoto Tsuji, Takashi Oka, and Hideo Aoki, “Brillouin-wigner theory for high-frequency expansion in periodically The precise form of the effective low frequency Hamil- driven systems: Application to floquet topological in- tonian is given as sulators,” Phys. Rev. B 93, 144307 (2016). 4 2 ! [114] Maximilian Fleischmann, Reena Gupta, Florian A A kθ Wullschlger, Dominik Weckbecker, Velimir Meded, Heff = H0 + HΩ + O , k k k Sangeeta Sharma, Bernd Meyer, and Sam Shallcross, D D D “Perfect and controllable nesting in the small angle twist  −  W1 0 F− 0 bilayer graphene,” (2019), arXiv:1908.08318 [cond- + 2 2 2  0 W2 0 F+  mat.mtrl-sci]. HΩ = A γ a0  ∗ − ,   F− 0 W2 0  [115] M. Fleischmann, R. Gupta, S. Sharma, and S. Shall- ∗ + cross, “Moire quantum well states in tiny angle two di- 0 F+ 0 W1 mensional semi-conductors,” (2019), arXiv:1901.04679 ∓iθ 2 ∗ e ( ± Ω) T11(x) − det(T (x))T11(x) [cond-mat.mes-hall]. F± = , [116] Ming Xie and Allan H. MacDonald, “On the nature D( ± Ω) of the correlated insulator states in twisted bilayer ( ± Ω)[( ± Ω)2 − w2λ − w2τ ] W ± = − 0 1 n , graphene,” (2018), arXiv:1812.04213 [cond-mat.str-el]. n D( ± Ω) [117] Nguyen N. T. Nam and Mikito Koshino, “Lattice re- √ ! √ ! ! laxation and energy band modulation in twisted bilayer 3xθ 3xθ 3xθ λ = 1 + 4 cos 1 cos 1 + cos 2 , graphene,” Phys. Rev. B 96, 075311 (2017). 2 2 2 [118] Francisco Guinea and Niels R. Walet, “Continuum mod- √ ! els for twisted bilayer graphene: Effect of lattice de- θ 3x2 π 3 n θ τn = 3 − 4 cos sin − (−1) x , formation and hopping parameters,” Phys. Rev. B 99, 2 6 2 1 205134 (2019). √ [119] Leon Balents, “General continuum model for twisted n θ π θ − 2 sin (−1) 3x + ; x = xikθ, bilayer graphene and arbitrary smooth deformations,” 1 6 i SciPost Phys. 7, 48 (2019). D() = −4 + 2Tr(T †(x)T (x)) − |detT|2 , [120] Kasra Hejazi, Chunxiao Liu, Hassan Shapourian, Xiao Chen, and Leon Balents, “Multiple topological transi- (A1) tions in twisted bilayer graphene near the first magic The quantities in the main text can be derived from here. angle,” Phys. Rev. B 99, 035111 (2019). We find that the intralayer gaps are given as ∆1(x) = 1 − + 1 − + [121] André Eckardt and Egidijus Anisimovas, “High- 2 (W1 − W2 ), ∆2(x) = 2 (W2 − W1 ). A Taylor series frequency approximation for periodically driven quan- reveals tum systems from a floquet-space perspective,” New 2 † 2 2 τ1+τ2 Journal of Physics 17, 093039 (2015). ∆n(x) 1 + Tr(T T ) − w0λ − w1 2 ´ = − − [122] Adolfo G. Grushin, Alvaro Gómez-León, and Titus Ne- A2γ2a2 Ω Ω3 upert, “Floquet fractional chern insulators,” Phys. Rev. 0 . 3w2(τ − τ ) Lett. 112, 156801 (2014). − (−1)n 1 1 2 + O(Ω−5) [123] Yuanbo Zhang, Tsung-Ta Tang, Caglar Girit, Zhao 2Ω4 Hao, Michael C. Martin, Alex Zettl, Michael F. Crom- (A2) 1 − + mie, Y. Ron Shen, and Feng Wang, “Direct observation The interlayer bias is given as U(x) = 4 (W1 − W1 − of a widely tunable bandgap in bilayer graphene,” Na- − + W2 + W2 ). A series expansion is ture 459, 820–823 (2009). [124] Kin Fai Mak, Chun Hung Lui, Jie Shan, and Tony F. U(x) (τ − τ )w2 Heinz, “Observation of an electric-field-induced band 1 2 1 −5 2 2 2 = 3 + O(Ω ). (A3) gap in bilayer graphene by infrared spectroscopy,” Phys. A γ a0 2Ω Rev. Lett. 102, 256405 (2009). [125] Zhenhua Qiao, Jeil Jung, Qian Niu, and Allan H. As last term from the diagonal block we find the overall MacDonald, “Electronic highways in bilayer graphene,” 1 − + − + potential of form V (x) = 4 (W1 + W1 + W2 + W2 ), Nano Letters 11, 3453–3459 (2011). which is expanded as [126] Ivar Martin, Ya. M. Blanter, and A. F. Morpurgo, “Topological confinement in bilayer graphene,” Phys. V (x) Rev. Lett. 100, 036804 (2008). − O(Ω−5) = − A2γ2a2 Ω2 [127] Jonathan S. Alden, Adam W. Tsen, Pinshane Y. Huang, 0 (A4) Robert Hovden, Lola Brown, Jiwoong Park, David A. 2 † 2 2 2 + 6(Tr(T T ) − λw0) − 3w1(τ1 + τ2) Muller, and Paul L. McEuen, “Strain solitons and topo- − 4 . logical defects in bilayer graphene,” Proceedings of the 2Ω National Academy of Sciences 110, 11256–11260 (2013). [128] Pablo San-Jose and Elsa Prada, “Helical networks in Notably the lowest order term is just a constant shift in twisted bilayer graphene under interlayer bias,” Phys. quasi-energy. Rev. B 88, 121408 (2013). On the off-diagonal blocks we find the interlayer hop- 1 ping strength δw0(x) = 2 (F− +F+), which to lower order 15 in Ω−1 is Furthermore we find that the interlayer hopping has a 1 bias β = 2 (F− − F+), which to low orders has the form

β(x) iT11 sin(θ) 2T11 cos(θ) δw0(x) −5 T11 cos(θ) 2iT11 sin(θ) − O(Ω−5) = − − 2 − O(Ω ) = − − 2 2 2 2 3 A2γ2a Ω2 Ω3 A γ a0 Ω Ω 0 . . 2 † ∗ i sin(θ) T 32 + Tr(T †T ) − T ∗ det(T ) cos(θ) T11(3 + tr(T T )) − T11det(T ) 11 11 − − 4 Ω4 Ω (A5) (A6)