betaFeedback

WorksPhysics and Astronomy

Switching off a high-harmonic polarization channel by anti-unitary symmetry

Abstract

We study polarization-resolved high-harmonic generation in a laser-driven two-leg Su–Schrieffer–Heeger ladder with asymmetric dimerization under periodic boundary conditions. A field linearly polarized along the ladder can generate both longitudinal and transverse radiation, whose harmonic content is controlled by lattice symmetry and electronic filling. Using symmetry analysis of the time-dependent density matrix, we show that reflection symmetry enforces odd-order harmonics in the longitudinal response and, when the transverse response is nonzero, even-order harmonics in the transverse direction. At half filling, an additional anti-unitary symmetry relating the occupied and empty bands suppresses the transverse current completely. A unitary chiral symmetry in the clean ladder relates the same bands but fails to enforce this cancellation, showing that band swapping alone is insufficient: the suppression requires both a density-matrix constraint that persists during the driven evolution and the appropriate transformation of the transverse-current operator. Moving away from half filling lifts this cancellation and allows transverse emission, while the reflection selection rule remains whenever reflection symmetry is preserved. Numerical calculations at quarter and half filling, combined with onsite potentials that selectively break the relevant symmetries, confirm these two distinct mechanisms. Thus, reflection symmetry determines which harmonic orders occur, whereas the interplay of anti-unitary symmetry and filling controls whether the transverse HHG channel is present at all.

Introduction

High-harmonic generation (HHG) in solids has been a powerful technique for investigating electron dynamics and the microscopic structure of materials [1, 2, 3]. The field has rapidly expanded since its first observation in bulk ZnO in 2011 [1]. When a strong laser field drives a crystalline solid, the resulting nonlinear electronic motion can emit radiation at integer multiples of the driving frequency [1, 4, 5, 6, 7]. The properties of this emission, such as its cutoff energy, intensity, and polarization, are highly sensitive to the band structure, symmetries, and scattering mechanisms of the material [8, 9, 10, 11].

In gases, HHG is commonly understood in terms of electrons that tunnel from the parent ion, gain energy from the laser field, and recombine while emitting high-frequency radiation. In crystalline solids, the electronic motion is shaped by the band structure: laser-driven electrons move through crystal momentum, can be transferred between filled and empty bands, and generate a time-dependent microscopic current whose Fourier spectrum gives the emitted harmonics. Different theoretical descriptions emphasize different aspects of this process, but in the present work we focus on symmetry constraints on the total current. For such a process, selection rules determine which harmonic orders appear and are dictated by the dynamical symmetries of the system [12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22]. Notably, even-order harmonics are generally forbidden in centrosymmetric systems, since inversion symmetry combined with the half-period anti-periodicity of the drive enforces odd-only emission. More generally, dynamical symmetries can constrain not only the harmonic order, but also the direction and polarization of the emitted radiation. This classification of dynamical symmetries and their associated selection rules was developed in Refs. [12, 13, 17]; here we derive the corresponding reflection-based rule explicitly for the two-leg ladder geometry, while identifying a distinct anti-unitary mechanism, tied to electronic filling, that switches an entire polarization channel on or off rather than merely constraining its harmonic content.

A simple yet rich platform for exploring these nonlinear phenomena is the Su–Schrieffer–Heeger (SSH) model, originally developed to describe polyacetylene [23]. While traditionally studied for its dimerized ground state, its simple two-site unit cell and well-understood band structure make it an ideal testbed for investigating laser-driven electron dynamics. Recent studies have shown that the SSH model exhibits characteristic HHG signatures, with harmonic spectra that reflect the underlying lattice geometry [24, 25, 26]. Bulk tight-binding formulations of HHG in SSH-type and related two-band models have also clarified the importance of Brillouin-zone integration, topology, and current operators consistent with the chosen Bloch representation [27, 28]. Beyond the standard SSH chain, numerous generalized versions have been explored in the context of HHG. These include extended SSH models with longer-range hoppings that host multiple topological phases [29, 30], SSH-Hubbard models incorporating electron-electron interactions [31, 32], Holstein-SSH models with electron-phonon coupling leading to polaron formation [33], and disordered SSH chains displaying multiple-plateau structures arising from spectral interference [34].

Extending the SSH model to more complex geometries, such as ladder configurations, introduces additional degrees of freedom and richer band structure. In a two-leg ladder SSH model, inter-chain hopping terms couple two SSH chains, resulting in a system that can host multiple topological phases depending on the interplay between intra-chain dimerization and inter-chain coupling [35]. Here we use this ladder geometry for a different purpose: it provides a minimal four-band bulk system in which the longitudinal and transverse HHG currents are constrained differently by symmetry. In addition to lattice symmetries dictating the allowed harmonics in solid-state HHG through dynamical selection rules [15, 17], the electronic filling level or doping of a system has been shown to qualitatively affect HHG spectra. For example, theoretical studies using a Hubbard chain demonstrate that moving away from half filling modifies the HHG yield and spectral features, highlighting that filling is a critical parameter influencing harmonic emission in driven lattice systems [36]. The filling effect considered below is more specific. At half filling of the clean ladder, the two initially filled bands and the two initially empty bands are mapped into each other by a combined reflection–particle-hole anti-unitary symmetry. This gives an additional constraint on the time-dependent density matrix and can cancel a whole current component, rather than merely changing the relative strength of harmonic peaks. Despite these advances, HHG in extended SSH models, such as two-leg ladder configurations, remains relatively unexplored.

In this work, we study high-harmonic generation in both the longitudinal and transverse directions of a laser-driven two-leg SSH ladder under periodic boundary conditions (PBCs) in the velocity gauge, focusing exclusively on bulk dynamics. Using symmetry analysis of the time-dependent density matrix, we derive symmetry-based selection rules for the polarization-resolved HHG spectra. Two distinct symmetry consequences emerge. A reflection that reverses the coordinate along the ladder, combined with the half-period sign reversal of the drive, enforces odd-order harmonic emission in the longitudinal direction and, whenever the transverse current is nonzero, even-order emission in the transverse direction. At half filling, an additional symmetry relating the occupied and empty bands forces the transverse current to vanish identically. Away from half filling, this cancellation is lifted, while the reflection-based selection rule remains whenever the corresponding reflection symmetry is preserved. The remainder of this paper develops these two mechanisms in turn: a reflection-based selection rule on harmonic order, and a distinct, filling-dependent anti-unitary constraint on whether the transverse channel is active at all. These analytical predictions are confirmed by numerical time evolution of the density matrix under periodic boundary conditions.

Model and Velocity Gauge

We consider a two-leg ladder system composed of two coupled SSH chains, where each lattice site carries an onsite potential VsLV_{s_L}. Each leg, labeled by L=u,lL = u,l, consists of NN unit cells with two sublattice sites s≡A,Bs \equiv A,B per unit cell. We use the ordered Bloch basis Ψ=(Au,Bu,Al,Bl)\Psi= (A_u,B_u,A_l,B_l). Using the tensor-product convention Ψ=L⊗s\Psi= L \otimes s, the first factor acts in the leg space (u,l)(u,l), while the second acts in the sublattice space (A,B)(A,B). The position-aware field-free Hamiltonian in this basis is [28]

h(k)=(VAugu(k)rA(k)0gu∗(k)VBu0rB(k)rA∗(k)0VAlgl(k)0rB∗(k)gl∗(k)VBl).(1)h(k)=\begin{pmatrix} V_{A_u} & g_u(k) & r_A(k) & 0 \\ g_u^*(k) & V_{B_u} & 0 & r_B(k) \\ r_A^*(k) & 0 & V_{A_l} & g_l(k) \\ 0 & r_B^*(k) & g_l^*(k) & V_{B_l} \end{pmatrix}. \tag*{(1)}

The diagonal elements VsLV_{s_L} are the onsite potentials of sublattice s=A,Bs = A,B on leg L=u,lL = u,l. The off-diagonal elements gL(k)g_L(k) describe hopping along each leg, whereas rA(k)r_A(k) and rB(k)r_B(k) describe hopping between the two legs on the AA and BB sublattices, respectively. Explicitly,

gL(k)=vLeikdv,L+wLe−ikdw,L,g_L(k)=v_L e^{ikd_{v,L}}+w_L e^{-ikd_{w,L}},
rA(k)=tre−ikΔ,rB(k)=treikΔ=rA∗(k),r_A(k)=t_r e^{-ik\Delta},\qquad r_B(k)=t_r e^{ik\Delta}=r_A^*(k),

where vL=−e−dv,Lv_L=-e^{-d_{v,L}} and wL=−e−dw,Lw_L=-e^{-d_{w,L}} are the intracell and intercell hopping amplitudes along leg LL, respectively, and trt_r is the interleg hopping amplitude. The phases in these matrix elements encode the longitudinal displacements of the corresponding bonds. The longitudinal bond lengths along leg LL are

dv,L=a2−2δL,dw,L=a2+2δL,dv,L+dw,L=a,(2)d_{v,L}=\frac{a}{2}-2\delta_L,\qquad d_{w,L}=\frac{a}{2}+2\delta_L,\qquad d_{v,L}+d_{w,L}=a, \tag*{(2)}

where aa is the unit-cell length shown in Figure 1, δL\delta_L is the signed dimerization of leg LL, and we set a=2a=2 for numerical calculations. The geometric and hopping parameters are chosen for illustrative purposes and are not intended to model a specific material quantitatively. The relative longitudinal displacement of the two legs is characterized by

Schematic representation of the two-leg SSH ladder model

Figure 1. Schematic representation of the two-leg SSH ladder model. The upper and lower legs have unequal dimerizations, δu≠δl\delta_u \ne\delta_l. The blue double-headed arrow denotes the unit-cell length aa, while the dashed vertical lines mark the reference positions nan a and na+a/2n a + a/2 of the undimerized lattice. The labels δu\delta_u and δl\delta_l denote the signed dimerizations of the upper and lower legs, respectively. Each unit cell contains four sites with intraleg hoppings vu,lv_{u,l}, intercell hoppings wu,lw_{u,l}, and interleg hoppings trt_r.

Δ=δu−δl,\Delta= \delta_u - \delta_l,

so that the AA- and BB-rung displacements are +Δ+\Delta and −Δ-\Delta, respectively.

Atomic units are used throughout (ℏ=∣e∣=me=1\hbar= |e| = m_e = 1). With this convention, ω0=0.024\omega_0 = 0.024 corresponds to a photon energy of 0.653 eV, i.e., a driving wavelength λ≃1.90\lambda\simeq1.90 μm. Spin is not included. Since the Hamiltonian is spin-independent and the dynamics are noninteracting, including spin would only multiply the currents by a factor of two and leaves all selection rules and symmetry-based cancellations unchanged. Quarter and half filling correspond to one and two occupied bands, respectively, out of the four bands of the ladder. Thus, δL>0\delta_L > 0 shortens the intracell bond and lengthens the intercell bond, while δL<0\delta_L < 0 does the opposite. We assume ∣δL∣<a/4|\delta_L| < a/4, so that both bond lengths remain positive.

For the clean ladder, i.e., with all onsite potentials set to zero, let R^x\hat{R}_x denote the real-space reflection

x→a/2−xx \to a/2-x

about the axis x=a/4x = a/4. It maps the lattice sites according to

(n,AL)↔(−n,BL),L=u,l,(n,A_L) \leftrightarrow(-n,B_L), \qquad L = u,l,

while leaving the leg index unchanged.

In the Bloch representation, the same spatial reflection maps k→−kk \to-k. Its action on the four orbital components (Au,Bu,Al,Bl)(A_u,B_u,A_l,B_l) is represented by

Rx=I⊗σx,R_x = I \otimes\sigma_x,

which exchanges A↔BA \leftrightarrow B within each leg and therefore maps the AA-rung onto the BB-rung. Consequently, the field-free Bloch Hamiltonian satisfies

Rxh(k)Rx−1=h(−k).R_x h(k) R_x^{-1} = h(-k).

The clean ladder also has particle-hole symmetry,

Ph(k)P−1=−h(−k),P=PK=σz⊗σzK.\mathcal{P}h(k)\mathcal{P}^{-1} = -h(-k), \qquad\mathcal{P} = P\mathcal{K} = \sigma_z \otimes\sigma_z\mathcal{K}.

Here, II is the identity matrix, σx\sigma_x, σy\sigma_y, and σz\sigma_z are Pauli matrices, and K\mathcal{K} denotes complex conjugation. In addition, Hamiltonian (1) exhibits the combined anti-unitary symmetry S=RxP=SK=σz⊗σyK\mathcal{S} = R_x\mathcal{P} = S\mathcal{K} = \sigma_z \otimes\sigma_y\mathcal{K}, obtained by composing reflection and particle-hole symmetry, such that

Sh(k)S−1=−h(k).\mathcal{S}h(k)\mathcal{S}^{-1} = -h(k).

Both P\mathcal{P} and S\mathcal{S} are anti-unitary. Their unitary parts satisfy

Ph∗(k)P−1=−h(−k),Sh∗(k)S−1=−h(k).P h^*(k)P^{-1} = -h(-k), \qquad S h^*(k)S^{-1} = -h(k).

Note that the clean ladder is bipartite, with sublattices {Au,Bl}\{A_u,B_l\} and {Bu,Al}\{B_u,A_l\}, since the rungs connect AA–AA and BB–BB sites. It therefore also possesses a unitary chiral symmetry

Γ=σz⊗σz,Γh(k,t)Γ−1=−h(k,t),\Gamma= \sigma_z \otimes\sigma_z, \qquad\Gamma h(k,t)\Gamma^{-1} = -h(k,t),

which is the unitary part of P\mathcal{P}. Although Γ\Gamma anticommutes with hh just as S\mathcal{S} does, the anti-unitarity of S\mathcal{S}, rather than the unitarity of Γ\Gamma, turns out to be essential: We return to this point in Sec. III, once the density-matrix formalism is introduced.

For comparison, full spatial inversion reverses both the longitudinal and transverse coordinates. In Bloch space it also maps k→−kk \to-k, while its action on the four orbital components is represented by

Π=σx⊗σx,\Pi= \sigma_x \otimes\sigma_x,

which exchanges both A↔BA \leftrightarrow B and u↔lu \leftrightarrow l. Inversion therefore requires vu=vlv_u = v_l and wu=wlw_u = w_l, equivalently δu=δl\delta_u = \delta_l. The asymmetric ladder considered here breaks full inversion while preserving the reflection R^x\hat{R}_x.

Throughout this paper, we find that both RxR_x and S\mathcal{S} symmetries play crucial roles in determining the selection rules, and therefore we restrict our analysis to these two symmetries. To examine the role of symmetry, we apply three types of onsite potentials.

  • Potential I: breaks S\mathcal{S} and preserves RxR_x: VAu=VBu=−VAl=−VBl=VV_{A_u} = V_{B_u} = -V_{A_l} = -V_{B_l} = V.

  • Potential II: breaks RxR_x and preserves S\mathcal{S}: VAu=VAl=−VBu=−VBl=VV_{A_u} = V_{A_l} = -V_{B_u} = -V_{B_l} = V.

  • Potential III: breaks both RxR_x and S\mathcal{S}: VAu=−VAl=−VBu=−VBl=VV_{A_u} = -V_{A_l} = -V_{B_u} = -V_{B_l} = V.

where potential I is a static transverse (inter-leg) bias between the two legs, while Potential II is a staggered, Rice-Mele-type sublattice potential [37], and Potential III is a leg-asymmetric staggered potential.

The band structures with and without onsite potentials are shown in Figure 2. The clean ladder, shown by the black solid lines in all three panels, has bands symmetric about zero energy and preserves both RxR_x and S\mathcal{S}. Potential I [panel (a)], shown by green dashed lines, breaks particle-hole symmetry and S\mathcal{S} while preserving RxR_x. Potential II [panel (b)], shown by orange dashed lines, breaks particle-hole symmetry and RxR_x while preserving S\mathcal{S}. Finally, Potential III, shown by the cyan dashed lines in panel (c), breaks both RxR_x and S\mathcal{S}.

Band structures of the clean ladder compared with those obtained with three onsite potentials

Figure 2. Band structures of the clean ladder, shown by black solid lines, compared with those obtained with (a) Potential I (b) Potential II, (c) Potential III, shown by dashed lines. Here, δu=0.05\delta_u = 0.05, δl=−0.2\delta_l = -0.2, tr=vut_r = v_u, and V=0.1V = 0.1.

Velocity-gauge formulation and current operators

The interaction between the laser field and the lattice system is described within the velocity gauge. We consider a spatially uniform laser field linearly polarized along the ladder direction xx, with vector potential

A(t)=Ax(t)x).^\mathbf{A}(t) = A_x(t)\hat{\mathbf{x}).}

Since the system is periodic only along xx, the field enters the Bloch Hamiltonian through the minimal-coupling prescription

k→q(t)=k+Ax(t).k \to q(t) = k + A_x(t).

We consider a 10-cycle flat-top pulse [7], where

Ax(t)=A0sin⁡(ω0t)f(t),(3)A_x(t) = A_0 \sin(\omega_0 t) f(t), \tag*{(3)}
f(t)={t/2T,0≤t≤2T1,2T<t≤8T(10T−t)/2T,8T<t≤10T(4)f(t) = \begin{cases} t/2T, & 0 \le t \le2T \\ 1, & 2T < t \le8T \\ (10T-t)/2T, & 8T < t \le10T \end{cases} \tag*{(4)}

Here, f(t)f(t) is the dimensionless pulse envelope, rising linearly from 0 to 1 over the first two cycles, remaining constant at the plateau value 1 over the following six cycles, and decreasing linearly back to 0 over the final two cycles. Thus, A0A_0 is directly the plateau vector-potential amplitude. Here, ω0\omega_0 is the driving frequency, and T=2π/ω0T = 2\pi/\omega_0 is the optical period. For A0=1.1A_0 = 1.1 and ω0=0.024\omega_0 = 0.024, the peak electric-field amplitude on the flat-top part of the pulse is E0=A0ω0=0.0264E_0 = A_0\omega_0 = 0.0264 a.u., corresponding to an intensity of approximately 2.5×10132.5 \times10^{13} W/cm2^2, which lies within the range used in strong-field solid-state HHG experiments.

For a hopping amplitude tijt_{ij} between sites ii and jj, the effect of the driving field is incorporated through the Peierls substitution,

tij→tijexp⁡[−i∫rirjA(t)⋅dl],(5)t_{ij} \to t_{ij} \exp\left[-i \int_{\mathbf{r}_i}^{\mathbf{r}_j} \mathbf{A}(t) \cdot d\mathbf{l}\right], \tag*{(5)}

where ri\mathbf{r}_i and rj\mathbf{r}_j denote the lattice-site positions. Here, tijt_{ij} multiplies c^j†c^i\hat{c}_j^\dagger\hat{c}_i, the operator that hops an electron from site ii to site jj, so that the phase in Eq. (5) is attached to hopping from ii to $j. This convention is used consistently in Appendix A. Accordingly, only bonds with a finite xx projection acquire Peierls phases. Consequently, the slanted interleg hoppings become

rA(q)=tre−iqΔ,rB(q)=treiqΔ,r_A(q) = t_r e^{-iq\Delta}, \qquad r_B(q) = t_r e^{iq\Delta},

with q=k+Ax(t)q = k + A_x(t).

The longitudinal current operator is obtained from the driven Bloch Hamiltonian as

J^x(k,t)=−∂h(q)∂q∣q=k+Ax(t).\hat{J}_x(k,t) = -\left.\frac{\partial h(q)}{\partial q}\right|_{q=k+A_x(t)}.

Since translational invariance is absent in the transverse direction, the transverse current operator J^y(k,t)\hat{J}_y(k,t) is derived from the continuity equation rather than from a momentum derivative; its explicit form is given in Appendix A 4. The explicit matrix representation of J^x(k,t)\hat{J}_x(k,t) is given in Appendix A 3. We set the interleg separation to unity. For identical legs, δu=δl\delta_u = \delta_l, with VAu=VAlV_{Au} = V_{Al} and VBu=VBlV_{Bu} = V_{Bl}, leg-exchange symmetry forces ⟨Jy(t)⟩=0\langle J_y(t)\rangle= 0. We therefore focus on the asymmetric case δu≠δl\delta_u \ne\delta_l.

The density matrix evolves according to

iρ˙(k,t)=[h(k+Ax(t)),ρ(k,t)],i\dot{\rho}(k,t) = [h(k+A_x(t)),\rho(k,t)],

and the longitudinal and transverse currents are evaluated as

⟨Jx(t)⟩=1Nk∑kTr⁡[ρ(k,t)J^x(k,t)],\langle J_x(t)\rangle= \frac{1}{N_k}\sum_k \operatorname{Tr}\left[\rho(k,t)\hat{J}_x(k,t)\right],
⟨Jy(t)⟩=1Nk∑kTr⁡[ρ(k,t)J^y(k,t)],(6)\langle J_y(t)\rangle= \frac{1}{N_k}\sum_k \operatorname{Tr}\left[\rho(k,t)\hat{J}_y(k,t)\right], \tag*{(6)}

where NkN_k is the number of sampled crystal momenta. The polarization-resolved HHG spectra are obtained from the Fourier transforms of the current components,

Px(y)(ω)=∣∫dt e−iωt⟨Jx(y)(t)⟩∣2.P_{x(y)}(\omega) = \left|\int dt\,e^{-i\omega t}\langle J_{x(y)}(t)\rangle\right|^2.

A Hanning window is applied to ⟨Jx,y(t)⟩\langle J_{x,y}(t)\rangle before the Fourier transform to suppress spectral leakage arising from the finite pulse.

Density Matrix and Symmetry Constraints

The density operator describes the quantum state of the system ρ(k,t)\rho(k,t), which evolves from an initial equilibrium state ρ(k,0)\rho(k,0) according to the von Neumann equation,

idρ(k,t)dt=[h(k,t),ρ(k,t)],i\frac{d\rho(k,t)}{dt} = [h(k,t),\rho(k,t)],

where we set ℏ=1\hbar= 1, and the initial density matrix is constructed from the eigenstates of the undriven Hamiltonian h(k)h(k). Denoting its eigenvalues and eigenvectors by En(k)E_n(k) and ∣n,k⟩\lvert n,k\rangle, respectively, the equilibrium density matrix is given by

geschniegeltgeschniegelt
ρ(k,0)=∑nfn(k)∣n,k⟩⟨n,k∣.\rho(k,0)=\sum_n f_n(k)|n,k\rangle\langle n,k|.

In the numerical calculations, fn(k)f_n(k) is a zero-temperature band occupation: at quarter filling only the lowest band is occupied, whereas at half filling the two lowest bands are occupied.

We now use the symmetries identified above to derive constraints on the time-dependent density matrix and, consequently, on the harmonic response.

Under periodic driving, the reflection symmetry RxR_x of the static system acquires a dynamical character. For a monochromatic drive, the vector potential satisfies Ax(t+T/2)=−Ax(t)A_x(t+T/2)=-A_x(t), and the driven Hamiltonian obeys

Rxh(k,t)Rx−1=h(−k,t+T/2).(7)R_x h(k,t)R_x^{-1}=h(-k,t+T/2). \tag*{(7)}

Thus, Eq. (7) combines the static reflection k→−kk\to-k with a half-period translation of the drive. For the finite pulse used in the numerical calculations, this relation holds within the constant-envelope part of the pulse. The ramp regions break the ideal half-period dynamical symmetry, so the corresponding selection rules are expected to hold approximately, with symmetry-forbidden harmonics strongly suppressed rather than identically absent.

For the ideal periodic response, and for an occupied Floquet subspace closed under reflection, Eq. (7) yields the density-matrix relation derived in Appendix B,

Rxρ(k,t)Rx−1=ρ(−k,t+T/2).(8)R_x\rho(k,t)R_x^{-1}=\rho(-k,t+T/2). \tag*{(8)}

The combined anti-unitary symmetry SS, when preserved, is an exact local-in-time symmetry of the driven Hamiltonian,

Sh(k,t)S−1=−h(k,t).(9)S h(k,t)S^{-1}=-h(k,t). \tag*{(9)}

At half filling, SS maps the two occupied bands onto the two unoccupied bands, and the initial density matrix therefore satisfies

Sρ(k,0)S−1=I4−ρ(k,0),(10)S\rho(k,0)S^{-1}=\mathbb{I}_4-\rho(k,0), \tag*{(10)}

where I4\mathbb{I}_4 is the four-dimensional identity. As shown in Appendix B, the equation of motion preserves this relation, yielding

Sρ(k,t)S−1=I4−ρ(k,t).(11)S\rho(k,t)S^{-1}=\mathbb{I}_4-\rho(k,t). \tag*{(11)}

It is worth asking why the anti-unitary SS, rather than the unitary chiral symmetry Γ=σz⊗σz\Gamma=\sigma_z\otimes\sigma_z noted in Sec. II, is required to obtain Eq. (11). At t=0t=0, Γ\Gamma likewise maps the two occupied bands onto the two empty bands, so

Γρ(k,0)Γ−1=I4−ρ(k,0)\Gamma\rho(k,0)\Gamma^{-1}=\mathbb{I}_4-\rho(k,0)

holds equally well for the clean system. This relation is not preserved by the dynamics, however, because Γ\Gamma is unitary, Γρ(k,t)Γ−1\Gamma\rho(k,t)\Gamma^{-1} evolves under the generator −h(k,t)-h(k,t), while I4−ρ(k,t)\mathbb{I}_4-\rho(k,t) evolves under +h(k,t)+h(k,t), so the two coincide only at t=0t=0. It is precisely this failure to persist in time, a consequence of Γ\Gamma being unitary rather than anti-unitary, that makes the combined symmetry S=RxPS=R_xP essential for Eq. (11) to hold at all tt.

Equations (8) and (11) are the central density-matrix constraints used in the selection-rule analysis below; their detailed derivations are given in Appendix B. The two constraints are summarized in Table I.

SymmetryConstraint on ρ(k,t)\rho(k,t)Time shiftFilling
RxR_xRxρ(k,t)Rx−1=ρ(−k,t+T/2)R_x\rho(k,t)R_x^{-1}=\rho(-k,t+T/2)T/2T/2Any
SSSρ(k,t)S−1=I4−ρ(k,t)S\rho(k,t)S^{-1}=\mathbb{I}_4-\rho(k,t)NoneHalf

Table I. Density-matrix constraints used in the selection-rule analysis. For the ideal periodic response with a reflection-symmetric occupied set, RxR_x gives the half-period relation independently of the number of occupied bands, whereas the local-in-time SS relation requires half filling.

Selection Rules

The density-matrix constraints derived above translate directly into selection rules for the harmonic response. For the ideal periodic response, reflection symmetry determines the allowed harmonic orders of the two current components: JxJ_x is odd under a half-period shift, whereas JyJ_y is even under the same shift. At half filling, preservation of SS imposes the additional constraint that the transverse current vanishes identically.

a. Reflection symmetry RxR_x: The current operators transform under reflection as

RxJ^y(k,t)Rx−1=+J^y(−k,t+T/2),R_x\hat{J}_y(k,t)R_x^{-1}=+\hat{J}_y(-k,t+T/2),
RxJ^x(k,t)Rx−1=−J^x(−k,t+T/2).R_x\hat{J}_x(k,t)R_x^{-1}=-\hat{J}_x(-k,t+T/2).

Combining these relations with the density-matrix constraint in Eq. (8), the current definition in Eq. (6), cyclicity of the trace, and the k→−kk\to-k symmetry of the Brillouin-zone sum gives

⟨Jy(t+T/2)⟩=+⟨Jy(t)⟩,\langle J_y(t+T/2)\rangle=+\langle J_y(t)\rangle,
⟨Jx(t+T/2)⟩=−⟨Jx(t)⟩.\langle J_x(t+T/2)\rangle=-\langle J_x(t)\rangle.

The detailed derivation is given in Appendix § 1.9.1. These relations directly imply that only even (odd) harmonics of the driving frequency are present in the transverse (longitudinal) spectrum: Py(nω0)=0P_y(n\omega_0)=0 for odd nn, Px(nω0)=0P_x(n\omega_0)=0 for even nn.

Equivalently, the reflection selection rule can be summarized as

Rx⟹Px: odd n,Py: even n.R_x \Longrightarrow P_x:\ \text{odd }n,\qquad P_y:\ \text{even }n.

This rule constrains the allowed harmonic orders of the two current components but does not by itself require either component to be nonzero. Since no filling-specific property enters the argument, the reflection selection rule applies at any filling for which the reflection constraint holds.

b. Anti-unitary symmetry SS at half filling: When SS is preserved, at half filling the anti-unitary symmetry imposes the density-matrix constraint in Eq. (11). The transverse-current operator satisfies

SJ^y(k,t)S−1=+J^y(k,t),(12)S\hat{J}_y(k,t)S^{-1}=+\hat{J}_y(k,t), \tag*{(12)}

and is traceless for every kk. This sign is specific to SS in Eq. (12): the unitary chiral operator Γ\Gamma discussed in Secs. II and III instead gives

ΓJ^y(k,t)Γ−1=−J^y(k,t).(13)\Gamma\hat{J}_y(k,t)\Gamma^{-1}=-\hat{J}_y(k,t). \tag*{(13)}

At t=0t=0, where

Γρ(k,0)Γ−1=I4−ρ(k,0)(14)\Gamma\rho(k,0)\Gamma^{-1}=I_4-\rho(k,0) \tag*{(14)}

holds, the minus sign from Eq. (13) cancels the minus sign associated with Eq. (14), and one recovers only the trivial identity

⟨Jy(0)⟩=⟨Jy(0)⟩\langle J_y(0)\rangle=\langle J_y(0)\rangle

rather than a vanishing condition. Thus, band swapping alone does not enforce the transverse-current cancellation. In addition to the density-matrix constraint, the cancellation requires the current operator to be even under the relevant symmetry, as is the case for SS, for which SJ^yS−1=+J^yS\hat{J}_yS^{-1}=+\hat{J}_y, but not for Γ\Gamma, for which ΓJ^yΓ−1=−J^y\Gamma\hat{J}_y\Gamma^{-1}=-\hat{J}_y. As shown in Appendix § 1.9 2, these properties imply

⟨Jy(t)⟩=0for all t,\langle J_y(t)\rangle=0\quad\text{for all }t,

and therefore

Py(nω0)=0for all n.P_y(n\omega_0)=0\quad\text{for all }n.

Thus, the anti-unitary half-filling constraint can be summarized as

S+ half filling⟹Py=0.S+\text{ half filling}\Longrightarrow P_y=0.

When both RxR_x and SS are preserved at half filling, the two constraints act simultaneously: reflection fixes the allowed harmonic orders, while the half-filled SS constraint suppresses the transverse channel. The combined result is

Rx+S+ half filling⟹Px: odd n,Py=0.R_x+S+\text{ half filling}\Longrightarrow P_x:\ \text{odd }n,\qquad P_y=0.

Eq. (11) can be restated in terms of the leg-population operator

M=σz⊗I,M=\sigma_z\otimes I,

which is traceless and satisfies

SMS−1=+M.SMS^{-1}=+M.

It implies

⟨M(t)⟩=0\langle M(t)\rangle=0

for all tt, i.e.,

Nu(t)=Nl(t)N_u(t)=N_l(t)

throughout the pulse. The two legs therefore remain exactly equally populated, so there is no net transverse charge transfer to radiate. The formal statement

⟨Jy(t)⟩=0\langle J_y(t)\rangle=0

is equivalent to this simpler picture through the relation

J^y=−N˙u.\hat{J}_y=-\dot{N}_u.

For the longitudinal current,

SJ^x(k,t)S−1=−J^x(k,t).S\hat{J}_x(k,t)S^{-1}=-\hat{J}_x(k,t).

As shown in Appendix § 1.9 2, this transformation yields only an identity for ⟨Jx(t)⟩\langle J_x(t)\rangle and therefore imposes no additional selection rule on the longitudinal current. If RxR_x is also preserved, the reflection-derived odd-harmonic selection rule remains valid at any filling.

Numerical Results

To verify the symmetry-based selection rules derived above, we calculate the HHG spectra in the transverse (yy) and longitudinal (xx) directions by numerically solving the time-dependent von Neumann equation. In all calculations, δu=0.05\delta_u=0.05, δl=−0.2\delta_l=-0.2, and vr=vuv_r=v_u. Labeling the four bands as E1(k)≤E2(k)≤E3(k)≤E4(k)E_1(k)\leq E_2(k)\leq E_3(k)\leq E_4(k), the minimum direct gap min⁡k(E2−E1)\min_k(E_2-E_1), relevant at quarter filling, and min⁡k(E3−E2)\min_k(E_3-E_2), relevant at half filling, remain open at every kk for all four cases considered. min⁡k(E2−E1)=0.23\min_k(E_2-E_1)=0.23, 0.31, 0.29, 0.28 and min⁡k(E3−E2)=0.30\min_k(E_3-E_2)=0.30, 0.29, 0.31, 0.32, for the clean ladder and Potentials I, II, and III, respectively. We consider two insulating fillings of the four-band ladder: quarter filling, where only the lowest band is initially occupied, and half filling, where the two lowest bands are initially occupied. Because quarter and half filling involve different relevant band pairs, the field-driven trajectories sample different interband dispersions. Since the drive enters through q(t)=k+Ax(t)q(t) = k+A_x(t), the vector-potential amplitude controls the momentum-space excursion and therefore the range of interband energy separations explored during the pulse. We use A0=1.6A_0 = 1.6 at quarter filling and A0=1.1A_0 = 1.1 at half filling to obtain well-developed spectra over comparable cutoff ranges, while keeping the driving frequency fixed at ω0=0.024\omega_0 = 0.024. Within each filling, the same driving parameters are used for the clean ladder and all three onsite potentials. The Brillouin zone is sampled with Nk=200N_k = 200 points on a mesh symmetric under k→−kk \to-k, as required for the derivation in Appendix § 1.9.1 to apply to the discretized momentum sum. The von Neumann equation is propagated using the unitary Crank–Nicolson (Cayley) scheme [38, 39]

ρ(t+Δt)=Uρ(t)U†\rho(t+\Delta t) = U\rho(t)U^\dagger

with

U=(I+i2HΔt)−1(I−i2HΔt),U = \left(I+\frac{i}{2}H\Delta t\right)^{-1}\left(I-\frac{i}{2}H\Delta t\right),

which preserves Hermiticity and the trace of ρ\rho to machine precision at every step. The time step is Δt=0.05\Delta t = 0.05. The reported spectra were checked to be converged with respect to both NkN_k and Δt\Delta t.

Figure 3 shows the half-filling spectra and Figure 4 the quarter-filling spectra for all four cases considered: the clean ladder [(a,b)], Potential I [(c,d)], Potential II [(e,f)], and Potential III [(g,h)(g,h)], with the transverse spectrum PyP_y in the left column and the longitudinal spectrum PxP_x in the right column of each row.

HHG power spectrum in the transverse and longitudinal directions at half filling

Figure 3. HHG power spectrum in the transverse (left column) and longitudinal (right column) directions at half filling. Each row represents the clean ladder (V=0V = 0) [(a,b)], Potential I [(c,d)], Potential II [(e,f)], and Potential III [(g,h)] (V=0.1V = 0.1). Red (blue) dashed lines indicate even (odd) harmonics. Here, δu=0.05\delta_u = 0.05, δl=−0.2\delta_l = -0.2, tr=vut_r = v_u, ω0=0.024\omega_0 = 0.024, and A0=1.1A_0 = 1.1. The inset shows the band structure, with the dashed line marking the Fermi level.

HHG power spectrum in the transverse and longitudinal directions at quarter filling

Figure 4. HHG power spectrum in the transverse (left column) and longitudinal (right column) directions at quarter filling. Each row represents the clean ladder (V=0V = 0) [(a,b)], Potential I [(c,d)], Potential II [(e,f)], and Potential III [(g,h)] (V=0.1V = 0.1). Red (blue) dashed lines indicate even (odd) harmonics. Here, δu=0.05\delta_u = 0.05, δl=−0.2\delta_l = -0.2, tr=vut_r = v_u, ω0=0.024\omega_0 = 0.024, and A0=1.6A_0 = 1.6. The inset shows the band structure, with the dashed line marking the Fermi level.

We first consider the clean ladder. Figure 3(a,Figure 3b) shows the spectra at half filling. The transverse signal [Figure 3(a)] is numerically negligible, while the longitudinal spectrum [Figure 3(b)] is dominated by odd-order harmonics. This agrees with the selection rules derived above: at half filling, SS suppresses the transverse current, while RxR_x restricts the longitudinal response to odd harmonics.

We next consider Potential I, which preserves RxR_x but breaks SS. Because the SS-based cancellation no longer applies, a nonzero transverse response appears at half filling [Figure 3(c)], in contrast to the clean ladder. The transverse spectrum is dominated by even harmonics, while the longitudinal response remains dominated by odd harmonics [Figure 3(d)], as expected from the preserved reflection symmetry. This shows that the suppression of the transverse response in the clean half-filled system originates from SS, rather than from half filling alone.

We then consider Potential II, which preserves SS but breaks RxR_x. At half filling, the transverse HHG signal remains numerically negligible [Figure 3(e)] despite the absence of reflection symmetry, consistent with the SS-based suppression of the transverse current. In contrast, the longitudinal spectrum is no longer restricted to odd harmonics and contains both even and odd orders [Figure 3(f)]. We note further that Γ\Gamma is not a symmetry of the Potential-II Hamiltonian. For the chiral operator defined above, any nonzero diagonal onsite potential breaks the chiral symmetry, whereas SS requires only VBL=−VALV_{BL}=-V_{AL} on each leg. Potential II satisfies the latter condition with V≠0V\ne0, and therefore preserves SS while breaking Γ\Gamma. The fact that the transverse signal remains suppressed at half filling despite the complete absence of chiral symmetry provides direct numerical confirmation that SS, rather than chirality, controls the suppression. Thus, Potential II provides a direct numerical separation of the two symmetries: the suppression survives when SS is preserved even though Γ\Gamma is broken.

Finally, we consider Potential III, which breaks both RxR_x and SS. At half filling, the transverse HHG signal is no longer suppressed [Figure 3(g)], since neither symmetry-based mechanism remains to protect it. The spectrum contains both even- and odd-order harmonics. The longitudinal spectrum likewise exhibits both even- and odd-order harmonics [Figure 3(h)], reflecting the breaking of RxR_x.

At quarter filling, the half-filling constraint associated with SS no longer applies. As shown in Figure 4(a,Figure 4b), a nonzero transverse HHG signal appears, with dominant peaks at even harmonic orders [Figure 4(a)]. The longitudinal spectrum remains dominated by odd harmonics [Figure 4(b)]. Thus, changing the filling removes the suppression of the transverse current while leaving the RxR_x selection rule intact. For Potential I, Figure 4(c,Figure 4d) shows the same reflection-controlled harmonic structure: the transverse response is nonzero and dominated by even harmonics, while the longitudinal spectrum is dominated by odd harmonics. These results are again consistent with the preserved RxR_x symmetry. For Potential II, the half-filling SS constraint no longer applies and the transverse response becomes nonzero [Figure 4(e)]. Since RxR_x is broken, neither current component is restricted to a single set of harmonic orders, and both even and odd harmonics appear in the transverse and longitudinal spectra [Figure 4(f)]. Finally, for Potential III, the same generic structure persists [Figure 4(g,Figure 4h)], with both current components containing even- and odd-order harmonics. This is the expected behavior when both symmetry mechanisms are absent: neither the reflection selection rule nor the anti-unitary suppression constrains the spectra, leaving no filling- or symmetry-based restriction on the harmonic content.

tbl:2 summarizes the harmonic content obtained in all eight numerical cases. The results agree with the selection rules derived above: when RxR_x is preserved, the longitudinal response is dominated by odd harmonics and a nonzero transverse response by even harmonics; when RxR_x is broken, both even and odd harmonic orders can occur; and when SS is preserved at half filling, the transverse signal is suppressed.

TABLE II. Observed harmonic content of the numerically calculated spectra for the asymmetric ladder, δu≠δl\delta_u\ne\delta_l. The symmetries shown in parentheses indicate which of RxR_x and SS are preserved in each case. “Odd nn” and “Even nn” denote spectra dominated by the corresponding harmonic orders for the finite pulse, “Even/odd nn” indicates visible contributions from both even and odd orders, and “Suppressed” denotes a numerically negligible harmonic signal.

System (symmetry)FillingPx(nω0)P_x(n\omega_0)Py(nω0)P_y(n\omega_0)
Clean (Rx,SR_x,S)QuarterOdd nnEven nn
Clean (Rx,SR_x,S)HalfOdd nnSuppressed
Pot. I (RxR_x)QuarterOdd nnEven nn
Pot. I (RxR_x)HalfOdd nnEven nn
Pot. II (SS)QuarterEven/odd nnEven/odd nn
Pot. II (SS)HalfEven/odd nnSuppressed
Pot. III (none)QuarterEven/odd nnEven/odd nn
Pot. III (none)HalfEven/odd nnEven/odd nn

Table II. Observed harmonic content of the numerically calculated spectra for the asymmetric ladder, δu≠δl\delta_u\ne\delta_l. The symmetries shown in parentheses indicate which of RxR_x and SS are preserved in each case. “Odd nn” and “Even nn” denote spectra dominated by the corresponding harmonic orders for the finite pulse, “Even/odd nn” indicates visible contributions from both even and odd orders, and “Suppressed” denotes a numerically negligible harmonic signal.

Summary

We have investigated HHG in the longitudinal (xx) and transverse (yy) directions of a laser-driven two-leg SSH ladder with asymmetric dimerization under periodic boundary conditions. Since edge-state contributions are excluded by construction, the resulting selection rules originate entirely from bulk symmetries. On an open chain, zero-energy edge modes are not necessarily paired by SS with an unoccupied partner, so the pairing argument underlying Eq. (10) may fail and the transverse suppression could be lifted. This provides a testable prediction for finite ladders, which we leave for future work.

Two distinct symmetry mechanisms govern the harmonic response. Reflection symmetry RxR_x, together with the half-period sign reversal of the drive, restricts the longitudinal response to odd harmonic orders and the transverse response to even harmonic orders. This selection rule constrains the allowed harmonic orders but does not by itself require a nonzero transverse current. The asymmetric dimerization considered here removes the leg-exchange symmetry that, for identical legs, would force the transverse current to vanish.

At half filling, preservation of the combined anti-unitary symmetry S=RxPS = R_xP imposes an additional constraint, ⟨Jy(t)⟩=0\langle J_y(t)\rangle= 0, and therefore suppresses transverse HHG altogether. Away from half filling, this cancellation no longer applies, allowing a transverse response while the reflection selection rule remains whenever RxR_x is preserved.

Numerical calculations at quarter and half filling, together with onsite potentials that selectively break RxR_x or SS, confirm these predictions. These results confirm the two distinct symmetry mechanisms outlined in the Introduction.

The symmetry SS relies on strictly bipartite hopping and is broken by any next-nearest-neighbor hopping along a leg or by a diagonal rung term. Thus, the on/off switch described here is a statement about an idealized model rather than a generic material; possible realizations include cold-atom, photonic, or acoustic ladder lattices, where such couplings can be engineered to vanish. We note also that a relaxation channel respecting SS would leave the suppression argument intact, whereas a symmetry-breaking bath would not.

Acknowledgments

We thank M. V. Hosseini for useful comments.

M.J. gratefully acknowledges funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through IRTG 2676/1 ‘Imaging of Quantum Systems’, project no. 437567992. F.N. and D.B. gratefully acknowledge funding from the German Research Foundation (DFG), SFB 1477 ‘Light-Matter Interactions at Interfaces’, project number 441234705.

Bloch convention and current operators

In this appendix, we construct the driven Hamiltonian and both current operators using the position-aware Bloch convention adopted in the main text. We first compare it with the conventional Bloch convention and then derive the longitudinal and transverse current operators.

Bloch conventions in the field-free and driven problems

In the ordered basis

Ψk=(Au,Bu,Al,Bl)T\Psi_k = (A_u, B_u, A_l, B_l)^T

the field-free Hamiltonian in the conventional Bloch basis is

hˉ(k)=(VAugˉu(k)tr0gˉu∗(k)VBu0trtr0VAlgˉl(k)0trgˉl∗(k)VBl)\bar{h}(k) = \begin{pmatrix} V_{A_u} & \bar{g}_u(k) & t_r & 0 \\ \bar{g}_u^*(k) & V_{B_u} & 0 & t_r \\ t_r & 0 & V_{A_l} & \bar{g}_l(k) \\ 0 & t_r & \bar{g}_l^*(k) & V_{B_l} \end{pmatrix}

with

gˉL(k)=vL+wLe−ika,L=u,l,(15)\bar{g}_L(k) = v_L + w_L e^{-ika}, \qquad L = u,l, \tag*{(15)}

where aa is the translation length of one unit cell, as indicated in Figure 1. In this convention, the intracell orbital positions are omitted from the Bloch phases.

For the geometry used in the main text, the longitudinal orbital positions inside the unit cell are

τALx=δL,τBLx=a2−δL,L=u,l.\tau^x_{A_L} = \delta_L, \qquad\tau^x_{B_L} = \frac{a}{2} - \delta_L, \qquad L = u,l.

These positions reproduce the bond lengths in Eq. (2). We then define

U(k)=diag⁡(eikτAux,eikτBux,eikτAlx,eikτBlx).U(k) = \operatorname{diag}\left(e^{ik\tau^x_{A_u}}, e^{ik\tau^x_{B_u}}, e^{ik\tau^x_{A_l}}, e^{ik\tau^x_{B_l}}\right).

The position-aware Hamiltonian is

h(k)=U†(k)hˉ(k)U(k).h(k) = U^\dagger(k)\bar{h}(k)U(k).

Therefore the two field-free Hamiltonians have identical eigenvalues, provided the same numerical values of vLv_L, wLw_L, and trt_r are used:

Enpos(k)=Enconv(k).E_n^{\mathrm{pos}}(k) = E_n^{\mathrm{conv}}(k).

With these orbital positions and the bond lengths in Eq. (2), the directed displacements entering the Bloch phases are +dv,L+d_{v,L} for the intracell hopping and −dw,L-d_{w,L} for the intercell hopping. The position-aware matrix element is then

gL(k)=vLeikdv,L+wLe−ikdw,L=eikdv,LgˉL(k).(16)g_L(k) = v_L e^{ikd_{v,L}} + w_L e^{-ikd_{w,L}} = e^{ikd_{v,L}}\bar{g}_L(k). \tag*{(16)}

Equation (16) makes the field-free equivalence explicit. In the position-aware convention, a longitudinal vector potential is introduced directly by

q(t)=k+Ax(t),q(t) = k + A_x(t),

so that

gL(q)=vLeiqdv,L+wLe−iqdw,L.g_L(q) = v_L e^{iqd_{v,L}} + w_L e^{-iqd_{w,L}}.

By contrast, the naive replacement in Eq. (15),

gˉL(k+Ax)=vL+wLe−i(k+Ax)a,\bar{g}_L(k + A_x) = v_L + w_L e^{-i(k+A_x)a},

couples only the intercell term to the field. The conventional Bloch basis remains valid, but the correctly Peierls-coupled matrix element is

gˉL(k,Ax)=vLeiAxdv,L+wLe−ikae−iAxdw,L,\bar{g}_L(k,A_x) = v_L e^{iA_xd_{v,L}} + w_L e^{-ika}e^{-iA_xd_{w,L}},

which satisfies

gL(k+Ax)=eikdv,LgˉL(k,Ax).g_L(k + A_x) = e^{ikd_{v,L}}\bar{g}_L(k,A_x).

In addition, the rung elements are affected in the same way. In the conventional Bloch basis, the field-free rung hopping is simply trt_r, independent of kk, since Eq. (15) does not carry any phase for the rung term. The physical rungs, however, carry longitudinal displacements ±Δ\pm\Delta. The Peierls phase must therefore be attached to the rung hopping directly in real space, before Fourier transforming, giving the driven rung amplitude tre∓iAxΔt_r e^{\mp iA_x\Delta}; only after this substitution is the Hamiltonian transformed to kk-space. This phase cannot instead be introduced afterward by any substitution in kk applied to the field-free hˉ(k)\bar{h}(k), since the rung hopping trt_r in hˉ(k)\bar{h}(k) carries no kk-dependence into which the field could be substituted.

In the position-aware convention, this gives the simple prescription

h(k,t)=h(k+Ax(t)).h(k,t) = h\bigl(k + A_x(t)\bigr).

We follow the Bloch convention used in Appendix A of [28], in which the orbital positions are included in the Bloch phases; here we refer to it as the position-aware Bloch convention.

Real-space and position-aware Bloch Hamiltonians

Using the geometry and hopping amplitudes defined in Eq. (2), the field-free real-space Hamiltonian is

H^0=∑n,L[VALn^nAL+VBLn^nBL+(vLc^nAL†c^nBL+wLc^n+1,AL†c^nBL+H.c.)]+tr∑n(c^nAu†c^nAl+c^nBu†c^nBl+H.c.).\hat{H}_0=\sum_{n,L}\left[V_{A_L}\hat{n}_{nA_L}+V_{B_L}\hat{n}_{nB_L}+\left(v_L\hat{c}^{\dagger}_{nA_L}\hat{c}_{nB_L}+w_L\hat{c}^{\dagger}_{n+1,A_L}\hat{c}_{nB_L}+\mathrm{H.c.}\right)\right]+t_r\sum_n\left(\hat{c}^{\dagger}_{nA_u}\hat{c}_{nA_l}+\hat{c}^{\dagger}_{nB_u}\hat{c}_{nB_l}+\mathrm{H.c.}\right).

where L=u,lL=u,l, nn labels the unit cell, and n^nα=c^nα†c^nα\hat{n}_{n\alpha}=\hat{c}^{\dagger}_{n\alpha}\hat{c}_{n\alpha}. This equation fixes unambiguously which hopping is intracell, which hopping is intercell, and which sites are connected by the rungs. The conventional and position-aware Bloch Hamiltonians below are two representations of this same real-space operator.

Using the position-aware Bloch transformation

c^nα=1N∑keik(na+ταx)c^kα,\hat{c}_{n\alpha}=\frac{1}{\sqrt{N}}\sum_k e^{ik(na+\tau_\alpha^x)}\hat{c}_{k\alpha},

the leg matrix elements are

gL(k)=vLeikdv,L+wLe−ikdw,L,g_L(k)=v_L e^{ikd_{v,L}}+w_L e^{-ikd_{w,L}},

with the bond lengths defined in Eq. (2). The position-aware rung matrix elements are

rA(k)=tre−ikΔ,rB(k)=treikΔ=rA∗(k),r_A(k)=t_r e^{-ik\Delta},\qquad r_B(k)=t_r e^{ik\Delta}=r_A^*(k),

where Δ=δu−δl\Delta=\delta_u-\delta_l is defined in the main text and trt_r is taken to be real. The complete field-free Hamiltonian is

h(k)=(VAugu(k)rA(k)0gu∗(k)VBu0rB(k)rA∗(k)0VAlgl(k)0rB∗(k)gl∗(k)VBl).h(k)=\begin{pmatrix} V_{A_u} & g_u(k) & r_A(k) & 0\\ g_u^*(k) & V_{B_u} & 0 & r_B(k)\\ r_A^*(k) & 0 & V_{A_l} & g_l(k)\\ 0 & r_B^*(k) & g_l^*(k) & V_{B_l} \end{pmatrix}.

For a field polarized along xx, the driven Hamiltonian used in the time evolution is

h(k,t)=h(q(t)),q(t)=k+Ax(t).h(k,t)=h\left(q(t)\right),\qquad q(t)=k+A_x(t).

Longitudinal current

In the position-aware convention, every longitudinal bond displacement is contained in the qq-dependence of the Hamiltonian. The complete longitudinal electronic-current operator is

Jx(k,t)=−∂h(q)∂q∣q=k+Ax(t).J_x(k,t)=-\left.\frac{\partial h(q)}{\partial q}\right|_{q=k+A_x(t)}.

The overall sign corresponds to electron charge −1-1. For the leg matrix elements,

jLx(q)=−idv,LvLeiqdv,L+idw,LwLe−iqdw,L.j_L^x(q)=-i d_{v,L}v_L e^{iqd_{v,L}}+i d_{w,L}w_L e^{-iqd_{w,L}}.

For the interleg bonds,

jAx(q)=iΔrA(q),jBx(q)=−iΔrB(q).j_A^x(q)=i\Delta r_A(q),\qquad j_B^x(q)=-i\Delta r_B(q).

The complete current matrix is

Jx(k,t)=(0jux(q)jAx(q)0[jux(q)]∗00jBx(q)[jAx(q)]∗00jlx(q)0[jBx(q)]∗[jlx(q)]∗0).J_x(k,t)=\begin{pmatrix} 0 & j_u^x(q) & j_A^x(q) & 0\\ [j_u^x(q)]^* & 0 & 0 & j_B^x(q)\\ [j_A^x(q)]^* & 0 & 0 & j_l^x(q)\\ 0 & [j_B^x(q)]^* & [j_l^x(q)]^* & 0 \end{pmatrix}.

Here q=k+Ax(t)q=k+A_x(t). This matrix contains the intracell, intercell, and longitudinal projections of the slanted interleg currents.

Transverse current from the continuity equation

Because translational invariance is absent in the transverse direction, there is no transverse crystal momentum kyk_y. We derive the transverse current directly from the continuity equation.

Define the directed hopping operator from the lower leg to the upper leg in position space as

h^rung(t)=tr∑n[e−iAx(t)Δc^nAu†c^nAl+eiAx(t)Δc^nBu†c^nBl].\hat{h}_{\mathrm{rung}}(t)=t_r\sum_n\left[e^{-iA_x(t)\Delta}\hat{c}^{\dagger}_{nA_u}\hat{c}_{nA_l}+e^{iA_x(t)\Delta}\hat{c}^{\dagger}_{nB_u}\hat{c}_{nB_l}\right].

The Hermitian rung Hamiltonian is

H^rung(t)=h^rung(t)+h^rung†(t).\hat{H}_{\mathrm{rung}}(t)=\hat{h}_{\mathrm{rung}}(t)+\hat{h}^{\dagger}_{\mathrm{rung}}(t).

The total upper-leg occupation is

N^u=∑n(n^nAu+n^nBu).\hat{N}_u=\sum_n\left(\hat{n}_{nA_u}+\hat{n}_{nB_u}\right).

Only the rung Hamiltonian changes N^u\hat{N}_u. Using

N^˙u=−i[N^u,H^rung]\dot{\hat{N}}_u=-i[\hat{N}_u,\hat{H}_{\mathrm{rung}}]

and

[N^u,h^rung]=h^rung,[N^u,h^rung†]=−h^rung†.[\hat{N}_u,\hat{h}_{\mathrm{rung}}]=\hat{h}_{\mathrm{rung}},\qquad[\hat{N}_u,\hat{h}_{\mathrm{rung}}^\dagger]=-\hat{h}_{\mathrm{rung}}^\dagger.

one obtains

N^˙u=−ih^rung+ih^rung†.\dot{\hat{N}}_u=-i\hat{h}_{\mathrm{rung}}+i\hat{h}_{\mathrm{rung}}^\dagger.

For electron charge −1-1, and using the interleg separation as the unit of transverse length as in the main text, the transverse charge current in the positive yy direction is

J^y=−N^˙u=ih^rung−ih^rung†.\hat{J}_y=-\dot{\hat{N}}_u=i\hat{h}_{\mathrm{rung}}-i\hat{h}_{\mathrm{rung}}^\dagger.

Applying the position-aware Bloch transformation gives

h^rung(t)=∑k[rA(q)c^kAu†c^kAl+rB(q)c^kBu†c^kBl],\hat{h}_{\mathrm{rung}}(t)=\sum_k\left[r_A(q)\hat{c}_{kA_u}^\dagger\hat{c}_{kA_l}+r_B(q)\hat{c}_{kB_u}^\dagger\hat{c}_{kB_l}\right],
q=k+Ax(t).q=k+A_x(t).

Therefore,

Jy(k,t)=(00irA(q)0000irB(q)−irA∗(q)0000−irB∗(q)00).J_y(k,t)=\begin{pmatrix} 0 & 0 & i r_A(q) & 0 \\ 0 & 0 & 0 & i r_B(q) \\ -i r_A^*(q) & 0 & 0 & 0 \\ 0 & -i r_B^*(q) & 0 & 0 \end{pmatrix}.

Here q=k+Ax(t)q=k+A_x(t). The explicit kk dependence follows from the physical longitudinal offsets of the slanted rungs and must be retained together with the position-aware density matrix.

In the identical-leg limit δu=δl\delta_u=\delta_l, Eq. (2) gives identical bond lengths and hoppings on the two legs, and therefore

gu(q)=gl(q).g_u(q)=g_l(q).

At the same time, Δ=0\Delta=0, so that

rA(q)=rB(q)=tr.r_A(q)=r_B(q)=t_r.

If, in addition, VAu=VAlV_{A_u}=V_{A_l} and VBu=VBlV_{B_u}=V_{B_l}, the Hamiltonian is invariant under interchange of the upper and lower legs. Such an interchange reverses the transverse direction and therefore changes the sign of J^y\hat{J}_y. For an initial state respecting the same symmetry, the time evolution preserves it, and consequently ⟨Jy(t)⟩=0\langle J_y(t)\rangle=0.

Symmetry constraints on the density matrix

In this appendix we derive the density-matrix constraints used in the main text. We separate the reflection-based dynamical symmetry, which relies on the half-period relation of a periodic driver, from the local-in-time anti-unitary symmetry SS, which gives the half filling cancellation.

Reflection Symmetry RxR_x

For the field-free initial state, reflection symmetry implies

Rxρ(k,0)Rx−1=ρ(−k,0).R_x\rho(k,0)R_x^{-1}=\rho(-k,0).

For the ideal periodic response considered below, we assume that the occupied Floquet subspace is continuously connected to the occupied field-free subspace. Since the latter is closed under reflection, the corresponding occupied Floquet subspace is chosen to be closed under the dynamical reflection symmetry derived below.

For a strictly periodic monochromatic drive, the vector potential satisfies

Ax(t+T/2)=−Ax(t).A_x(t+T/2)=-A_x(t).

Together with reflection symmetry, this gives

Rxh(k,t)Rx−1=h(−k,t+T/2).(17)R_xh(k,t)R_x^{-1}=h(-k,t+T/2). \tag*{(17)}

Let

∣ψmk(t)⟩=e−iϵmkt∣umk(t)⟩,∣umk(t+T)⟩=∣umk(t)⟩.|\psi_{mk}(t)\rangle=e^{-i\epsilon_{mk}t}|u_{mk}(t)\rangle,\qquad |u_{mk}(t+T)\rangle=|u_{mk}(t)\rangle.

Here mm labels a Floquet quasienergy branch. The periodic part obeys

[h(k,t)−i∂t]∣umk(t)⟩=ϵmk∣umk(t)⟩.[h(k,t)-i\partial_t]|u_{mk}(t)\rangle=\epsilon_{mk}|u_{mk}(t)\rangle.

Applying RxR_x and using Eq. (17),

[h(−k,t+T/2)−i∂t]Rx∣umk(t)⟩=ϵmkRx∣umk(t)⟩.[h(-k,t+T/2)-i\partial_t]R_x|u_{mk}(t)\rangle=\epsilon_{mk}R_x|u_{mk}(t)\rangle.

Hence, for a nondegenerate quasienergy branch and consistent branch labels at kk and −k-k,

Rx∣umk(t)⟩=eiχmk∣um,−k(t+T/2)⟩.(18)R_x|u_{mk}(t)\rangle=e^{i\chi_{mk}}|u_{m,-k}(t+T/2)\rangle. \tag*{(18)}

Here, χmk\chi_{mk} is an arbitrary relative phase between the two symmetry-related Floquet states and therefore drops out of the corresponding projector. This Floquet–Bloch symmetry relation follows the constructions of Refs. [13, 17]. At a quasienergy degeneracy, the corresponding statement applies to a symmetry-adapted projector onto the degenerate subspace. For each occupied Floquet–Bloch orbital, define

ρmk(t)=∣ψmk(t)⟩⟨ψmk(t)∣=∣umk(t)⟩⟨umk(t)∣.\rho_{mk}(t)=|\psi_{mk}(t)\rangle\langle\psi_{mk}(t)|=|u_{mk}(t)\rangle\langle u_{mk}(t)|.

Using Eq. (18),

Rxρmk(t)Rx−1=Rx∣umk(t)⟩⟨umk(t)∣Rx−1=eiχmk∣um,−k(t+T/2)⟩×⟨um,−k(t+T/2)∣e−iχmk=ρm,−k(t+T/2).R_x\rho_{mk}(t)R_x^{-1}=R_x|u_{mk}(t)\rangle\langle u_{mk}(t)|R_x^{-1} = e^{i\chi_{mk}}|u_{m,-k}(t+T/2)\rangle \times\langle u_{m,-k}(t+T/2)|e^{-i\chi_{mk}} = \rho_{m,-k}(t+T/2).

The one-particle density matrix is

ρ(k,t)=∑m∈occρmk(t).\rho(k,t)=\sum_{m\in\mathrm{occ}}\rho_{mk}(t).

For an occupied set mapped onto itself by reflection,

Rxρ(k,t)Rx−1=∑m∈occRxρmk(t)Rx−1=∑m∈occρm,−k(t+T/2).R_x\rho(k,t)R_x^{-1} =\sum_{m\in\mathrm{occ}}R_x\rho_{mk}(t)R_x^{-1} =\sum_{m\in\mathrm{occ}}\rho_{m,-k}(t+T/2).

Thus, summing over the occupied Floquet branches proves the density-matrix constraint stated in Eq. (8). Equation (8) is exact for the ideal periodic Floquet response provided the occupied set is closed under reflection; it does not depend on the number of occupied branches. For the finite pulse, the ramp can generate superpositions of Floquet branches, so the relation is approximate during the plateau and the forbidden harmonic orders are strongly suppressed rather than identically zero.

We next consider the combined local-in-time anti-unitary symmetry S=RxP\mathcal{S}=R_x\mathcal{P}.

Anti-Unitary Symmetry S\mathcal{S}

The combined anti-unitary symmetry S=σz⊗σyK\mathcal{S}=\sigma_z\otimes\sigma_y\mathcal{K} acts locally in time. It satisfies

Sh(k,t)S−1=−h(k,t).(19)\mathcal{S}h(k,t)\mathcal{S}^{-1}=-h(k,t). \tag*{(19)}

Because S\mathcal{S} is anti-unitary, applying it to the von Neumann equation changes the sign of ii. Defining

ρˉS(k,t)=Sρ(k,t)S−1,\bar{\rho}_{\mathcal{S}}(k,t)=\mathcal{S}\rho(k,t)\mathcal{S}^{-1},

we obtain

−i∂tρˉS(k,t)=[Sh(k,t)S−1,ρˉS(k,t)]=[−h(k,t),ρˉS(k,t)].-i\partial_t\bar{\rho}_{\mathcal{S}}(k,t) =\left[\mathcal{S}h(k,t)\mathcal{S}^{-1},\bar{\rho}_{\mathcal{S}}(k,t)\right] =\left[-h(k,t),\bar{\rho}_{\mathcal{S}}(k,t)\right].

Equivalently,

i∂tρˉS(k,t)=[h(k,t),ρˉS(k,t)].(20)i\partial_t\bar{\rho}_{\mathcal{S}}(k,t) =\left[h(k,t),\bar{\rho}_{\mathcal{S}}(k,t)\right]. \tag*{(20)}

Hence Sρ(k,t)S−1\mathcal{S}\rho(k,t)\mathcal{S}^{-1} evolves with the same Liouvillian as ρ(k,t)\rho(k,t).

The remaining input is the initial state. At half filling, the two lowest bands are occupied and the two upper bands are empty. Since Eq. (19) maps an eigenstate of energy εn(k)\varepsilon_n(k) to one of energy −εn(k)-\varepsilon_n(k), S\mathcal{S} maps the initially filled bands onto the initially empty bands. This pairing requires the spectrum to be gapped at E=0E=0 for all kk, so that the occupied and unoccupied subspaces are unambiguously separated and mapped onto one another by S\mathcal{S}. Therefore

Sρ(k,0)S−1=I4−ρ(k,0).(21)\mathcal{S}\rho(k,0)\mathcal{S}^{-1}=\mathbb{I}_4-\rho(k,0). \tag*{(21)}

The matrix I4−ρ(k,t)\mathbb{I}_4-\rho(k,t) obeys the same equation of motion as ρ(k,t)\rho(k,t), because

i∂t[I4−ρ(k,t)]=[h(k,t),I4−ρ(k,t)].i\partial_t\left[\mathbb{I}_4-\rho(k,t)\right] =\left[h(k,t),\mathbb{I}_4-\rho(k,t)\right].

Combining this with Eqs. (B4) and (B5), uniqueness of the solution proves the density-matrix constraint stated in Eq. (11). This relation is exact for any pulse shape, as long as the driven Hamiltonian preserves S\mathcal{S} and the initial state satisfies Eq. (21). Away from half filling, the initially filled and empty sectors are not completely paired by S\mathcal{S}, so Eq. (21) fails and the constraint (11) no longer follows.

Selection Rules from Symmetry: Detailed Derivations

In this appendix, we derive the harmonic selection rules for the transverse (JyJ_y) and longitudinal (JxJ_x) currents. The current expectation values are defined in Eq. (6). The key symmetry actors are the reflection symmetry RxR_x, which acts at any filling, and the combined anti-unitary symmetry S=RxP\mathcal{S}=R_x\mathcal{P}, whose density-matrix constraint is operative only at half filling.

For a strictly periodic drive, each current component can be expanded in a Fourier series,

Jx,y(t)=∑n=−∞∞Jne−inω0t,J_{x,y}(t)=\sum_{n=-\infty}^{\infty}J_n e^{-in\omega_0t},
Jn=1T∫0Tdt Jx,y(t)einω0t.J_n=\frac{1}{T}\int_0^Tdt\,J_{x,y}(t)e^{in\omega_0t}.

Here ω0=2π/T\omega_0=2\pi/T is the driving frequency, and the selection rules correspond to identifying which Fourier coefficients JnJ_n must vanish. For the finite pulse used in the numerical calculations, the corresponding rules appear as suppression of the symmetry-forbidden harmonic components, up to spectral broadening from the envelope.

Dynamical reflection symmetry RxR_x at any filling: odd harmonics in JxJ_x, even harmonics in JyJ_y

We now use the reflection constraint in Eq. (8) to determine the allowed harmonic orders of the two current components.

a. Transverse current JyJ_y. The operator J^y\hat{J}_y transforms as

RxJ^y(k,t)Rx−1=+J^y(−k,t+T/2)(22)R_x\hat{J}_y(k,t)R_x^{-1}=+\hat{J}_y(-k,t+T/2) \tag*{(22)}

Using Eqs. (8) and (22) together with the cyclic property of the trace and the k→−kk\to-k symmetry of the Brillouin-zone sum,

⟨Jy(t+T/2)⟩=1Nk∑kTr⁡[ρ(k,t+T/2)J^y(k,t+T/2)]=1Nk∑kTr⁡[Rxρ(−k,t)J^y(−k,t)Rx−1]=1Nk∑kTr⁡[ρ(−k,t)J^y(−k,t)]=+⟨Jy(t)⟩.\begin{aligned} \langle J_y(t+T/2)\rangle &=\frac{1}{N_k}\sum_k\operatorname{Tr}\left[\rho(k,t+T/2)\hat{J}_y(k,t+T/2)\right]\\ &=\frac{1}{N_k}\sum_k\operatorname{Tr}\left[R_x\rho(-k,t)\hat{J}_y(-k,t)R_x^{-1}\right]\\ &=\frac{1}{N_k}\sum_k\operatorname{Tr}\left[\rho(-k,t)\hat{J}_y(-k,t)\right]\\ &=+\langle J_y(t)\rangle. \end{aligned}

The current is therefore periodic with half-period T/2T/2. Substituting into the Fourier series,

Jy(t+T/2)=∑nJne−inω0(t+T/2)=∑nJn(−1)ne−inω0t=!∑nJne−inω0t.\begin{aligned} J_y(t+T/2)&=\sum_n J_n e^{-in\omega_0(t+T/2)}\\ &=\sum_n J_n(-1)^n e^{-in\omega_0t}\mathrel{\overset{!}{=}}\sum_n J_n e^{-in\omega_0t}. \end{aligned}

Matching coefficients gives Jn(−1)n=JnJ_n(-1)^n=J_n, which requires (−1)n=1(-1)^n=1, i.e., nn must be even. Therefore

Jn=0for odd n,J_n=0\quad\text{for odd }n,

and the transverse current contains only even harmonics whenever it is nonzero and RxR_x is preserved.

b. Longitudinal current JxJ_x. The operator J^x\hat{J}_x transforms with the opposite sign,

RxJ^x(k,t)Rx−1=−J^x(−k,t+T/2).R_x\hat{J}_x(k,t)R_x^{-1}=-\hat{J}_x(-k,t+T/2).

The identical calculation now gives an overall minus sign,

⟨Jx(t+T/2)⟩=−1Nk∑kTr⁡[Rxρ(−k,t)J^x(−k,t)Rx−1]=−1Nk∑kTr⁡[ρ(−k,t)J^x(−k,t)]=−⟨Jx(t)⟩.\begin{aligned} \langle J_x(t+T/2)\rangle &=-\frac{1}{N_k}\sum_k\operatorname{Tr}\left[R_x\rho(-k,t)\hat{J}_x(-k,t)R_x^{-1}\right]\\ &=-\frac{1}{N_k}\sum_k\operatorname{Tr}\left[\rho(-k,t)\hat{J}_x(-k,t)\right]\\ &=-\langle J_x(t)\rangle. \end{aligned}

The longitudinal current is anti-periodic with half-period T/2T/2. Substituting into the Fourier series gives Jn(−1)n=−JnJ_n(-1)^n=-J_n, requiring (−1)n=−1(-1)^n=-1, i.e. nn odd. Therefore

Jn=0for even n,J_n=0\quad\text{for even }n,

and, when RxR_x is preserved, the longitudinal current contains only odd harmonics at any filling. This is the standard result for systems with inversion-like symmetry combined with a half-period shift.

Anti-unitary symmetry S=RxPS=R_xP at half filling: complete suppression of JyJ_y

At half filling, we use the density-matrix constraint in Eq. (11). For the zero-temperature band projector used here, this relation holds because SS maps the occupied two-band subspace onto its unoccupied complement. Half filling alone would not be sufficient for an arbitrary density matrix. Away from half filling, the initially occupied and empty subspaces are not completely paired by SS and the constraint does not follow.

The transverse current operator transforms under SS as

SJ^y(k,t)S−1=+J^y(k,t).(23)S\hat{J}_y(k,t)S^{-1}=+\hat{J}_y(k,t). \tag*{(23)}

We now evaluate ⟨Jy(t)⟩\langle J_y(t)\rangle using the transformed density matrix. Writing the anti-unitary operator as S=SKS=S_K, with SS unitary, one has SAS−1=SA∗S†SAS^{-1}=SA^*S^\dagger and therefore Tr⁡[SAS−1]=Tr⁡[A]∗\operatorname{Tr}[SAS^{-1}]=\operatorname{Tr}[A]^*. Using Eqs. (11) and (23),

⟨Jy(t)⟩=1Nk∑kTr⁡[ρ(k,t)J^y(k,t)]={1Nk∑kTr⁡[Sρ(k,t)S−1SJ^y(k,t)S−1]}∗={1Nk∑kTr⁡[(I4−ρ(k,t))J^y(k,t)]}∗.\begin{aligned} \langle J_y(t)\rangle &=\frac{1}{N_k}\sum_k\operatorname{Tr}\left[\rho(k,t)\hat{J}_y(k,t)\right]\\ &=\left\{\frac{1}{N_k}\sum_k\operatorname{Tr}\left[S\rho(k,t)S^{-1}S\hat{J}_y(k,t)S^{-1}\right]\right\}^{*}\\ &=\left\{\frac{1}{N_k}\sum_k\operatorname{Tr}\left[(\mathbb{I}_4-\rho(k,t))\hat{J}_y(k,t)\right]\right\}^{*}. \end{aligned}

Because both I4−ρ(k,t)\mathbb{I}_4-\rho(k,t) and J^y(k,t)\hat{J}_y(k,t) are Hermitian, the trace of their product is real: Tr⁡[AB]∗=Tr⁡[(AB)†]=Tr⁡[BA]=Tr⁡[AB]\operatorname{Tr}[AB]^*=\operatorname{Tr}[(AB)^\dagger]=\operatorname{Tr}[BA]=\operatorname{Tr}[AB]. Hence,

⟨Jy(t)⟩=1Nk∑k(Tr⁡[J^y(k,t)]−Tr⁡[ρ(k,t)J^y(k,t)])=1Nk∑kTr⁡[J^y(k,t)]−⟨Jy(t)⟩.\begin{aligned} \langle J_y(t)\rangle &=\frac{1}{N_k}\sum_k\left(\operatorname{Tr}\left[\hat{J}_y(k,t)\right]-\operatorname{Tr}\left[\rho(k,t)\hat{J}_y(k,t)\right]\right)\\ &=\frac{1}{N_k}\sum_k\operatorname{Tr}\left[\hat{J}_y(k,t)\right]-\langle J_y(t)\rangle. \end{aligned}

From the explicit matrix representation in Appendix § 1.1.1.§ 1.7.4,

Tr⁡[J^y(k,t)]=0for every k.\operatorname{Tr}\left[\hat{J}_y(k,t)\right]=0\quad\text{for every }k.

Therefore,

⟨Jy(t)⟩=−⟨Jy(t)⟩⟹⟨Jy(t)⟩=0for all t,\langle J_y(t)\rangle= -\langle J_y(t)\rangle\Longrightarrow\langle J_y(t)\rangle= 0 \quad\text{for all } t,

and consequently

Py(ω)=0for all harmonic orders.P_y(\omega)=0 \quad\text{for all harmonic orders.}

Thus, when SS is preserved at half filling, the transverse HHG signal is completely suppressed. a. Why SS does not constrain JxJ_x. For the longitudinal current, SS transforms with the opposite sign,

SJ^x(k,t)S−1=−J^x(k,t).S\hat{J}_x(k,t)S^{-1}=-\hat{J}_x(k,t).

Repeating the same calculation gives

⟨Jx(t)⟩=−1Nk∑kTr⁡[J^x(k,t)]+⟨Jx(t)⟩.\langle J_x(t)\rangle=-\frac{1}{N_k}\sum_k \operatorname{Tr}\left[\hat{J}_x(k,t)\right]+\langle J_x(t)\rangle.

Since

Tr⁡[J^x(k,t)]=0for every k,\operatorname{Tr}\left[\hat{J}_x(k,t)\right]=0 \quad\text{for every } k,

this reduces to the identity

⟨Jx(t)⟩=⟨Jx(t)⟩.\langle J_x(t)\rangle=\langle J_x(t)\rangle.

Thus, SS imposes no constraint on the longitudinal current. If RxR_x is also preserved, the reflection-derived odd-harmonic selection rule remains valid at any filling; when RxR_x is broken, no restriction to even or odd harmonic orders in JxJ_x follows from these symmetries.

References

References

  1. [1]S. Ghimire, A. D. DiChiara, E. Sistrunk, P. Agostini, L. F. DiMauro, and D. A. Reis, Observation of high-order harmonic generation in a bulk crystal, Nature Physics 7, 138 (2011).DOI
  2. [2]M. Hohenleutner, F. Langer, O. Schubert, M. Knorr, U. Huttner, S. W. Koch, M. Kira, and R. Huber, Real-time observation of interfering crystal electrons in high-harmonic generation, Nature 523, 572 (2015).DOI
  3. [3]M. Garg, M. Zhan, T. T. Luu, H. Lakhotia, T. Klostermann, A. Guggenmos, and E. Goulielmakis, Multi-petahertz electronic motion, Nature 538, 359 (2016).DOI
  4. [4]O. Schubert, M. Hohenleutner, F. Langer, B. Urbaszek, C. Lange, U. Huttner, D. Golde, T. Meier, M. Kira, S. W. Koch, and R. Huber, Sub-cycle control of terahertz high-harmonic generation by dynamical Bloch oscillations, Nature Photonics 8, 119 (2014).DOI
  5. [5]T. T. Luu, M. Garg, S. Y. Kruchinin, A. Moulet, M. T. Hassan, and E. Goulielmakis, Extreme ultraviolet high-harmonic spectroscopy of solids, Nature 521, 498 (2015).DOI
  6. [6]M. F. Ciappina, Solid-state high-order harmonic generation: emerging frontiers in ultrafast and quantum light science, Advances in Physics 74, 1 (2025).
  7. [7]F. Navarrete, M. F. Ciappina, and U. Thumm, Crystal-momentum-resolved contributions to high-order harmonic generation in solids, Physical Review A 100, 033405 (2019).DOI
  8. [8]G. Vampa, C. R. McDonald, G. Orlando, D. D. Klug, P. B. Corkum, and T. Brabec, Theoretical analysis of high-harmonic generation in solids, Physical Review Letters 113, 073901 (2014).pubmed.ncbi.nlm.nih.gov/25170708
  9. [9]G. Vampa, T. J. Hammond, N. Thiré, B. E. Schmidt, F. Légaré, C. R. McDonald, T. Brabec, and P. B. Corkum, Linking high harmonics from gases and solids, Nature 522, 462 (2015).DOI
  10. [10]B. Lorenzo, C. Batista, M. Jangjan, D. Kim, J. Menotti, F. Liu, W. Gao, S. Ghimire, C. Granados, and A. Chacón, High harmonic spectroscopy from lower-order to higher-order topological insulators, Physical Review B 114, 084307 (2026).arxiv.org/abs/2508.15631
  11. [11]T. T. Luu and H. J. Wörner, Measurement of the Berry curvature of solids using high-harmonic spectroscopy, Nature Communications 9, 916 (2018).DOI
  12. [12]O. E. Alon, V. Averbukh, and N. Moiseyev, Selection rules for the high harmonic generation spectra, Physical Review Letters 80, 3743 (1998).DOI
  13. [13]O. Neufeld, D. Podolsky, and O. Cohen, Floquet group theory and its application to selection rules in harmonic generation, Nature Communications 10, 405 (2019).DOI
  14. [14]M. Even Tzur, O. Neufeld, A. Fleischer, and O. Cohen, Selection rules for breaking selection rules, New Journal of Physics 23, 103039 (2021).DOI
  15. [15]X. Liu, X. Zhu, L. Li, Y. Li, Q. Zhang, P. Lan, and P. Lu, Selection rules of high-order harmonic generation: Symmetries of molecules and laser fields, Phys. Rev. A 94, 033410 (2016).DOI
  16. [16]M. Jangjan, A. Kashef, S. Pooyan, S. Scheel, and D. Bauer, Even-harmonic generation through nonequilibrium steady-state symmetry breaking (2026), arXiv:2607.27847 [quant-ph].arxiv.org/abs/2607.27847
  17. [17]N. Moiseyev, Selection rules for harmonic generation in solids, Phys. Rev. A 91, 053811 (2015).DOI
  18. [18]F. Ceccherini, D. Bauer, and F. Cornolti, Dynamical symmetries and harmonic generation, Journal of Physics B: Atomic, Molecular and Optical Physics 34, 5017 (2001).DOI
  19. [19]M. Ciappina and P. Tzallas, eds., High-Order Harmonic Generation in Solids (World Scientific, Singapore, 2024) foreword by Shambhu Ghimire and David Reis.DOI
  20. [20]P. Peterka, F. Trojánek, P. Malý, and M. Kozák, Non-collinear phase-matching of high harmonic generation in solids, APL Photonics 9, 161110 (2024).
  21. [21]S. D. C. Roscam Abbing, N. Kuzkova, R. van der Linden, F. Campi, B. de Keijzer, C. Morice, Z.-Y. Zhang, M. L. S. van der Geest, and P. M. Kraus, Enhancing the efficiency of high-order harmonics with two-color non-collinear wave mixing in silica, Nature Communications 15, 8335 (2024).
  22. [22]N. Kuzkova, S. D. C. Roscam Abbing, R. van der Linden, F. Campi, B. de Keijzer, C. Morice, Z.-Y. Zhang, M. L. S. van der Geest, and P. M. Kraus, Boosting solid-state high-harmonic generation efficiently by non-collinear XUV wave mixing with two-color drivers, EPJ Web of Conferences 335, 10003 (2025).
  23. [23]W. P. Su, J. R. Schrieffer, and A. J. Heeger, Solitons in Polyacetylene, Physical Review Letters 42, 1698 (1979).DOI
  24. [24]D. Bauer and K. K. Hansen, High-harmonic generation in solids with and without Topological Edge States, Physical Review Letters 120, 177401 (2018).DOI
  25. [25]H. Jürß and D. Bauer, High-harmonic generation in Su-Schrieffer-Heeger chains, Physical Review B 99, 195428 (2019).
  26. [26]C. Yu, High-order harmonic generation from Su-Schrieffer-Heeger chains with edge states: Three-step processes and their interference, Physical Review A 110, 033113 (2024).DOI
  27. [27]H. Drüeke and D. Bauer, Robustness of topologically sensitive harmonic generation in laser-driven linear chains, Physical Review A 99, 053402 (2019).DOI
  28. [28]D. Moos, H. Jürß, and D. Bauer, Intense-laser-driven electron dynamics and high-order harmonic generation in solids including topological effects, Physical Review A 102, 053112 (2020).
  29. [29]M. L. Bera, J. O. d. Almeida, M. Dziurawiec, M. Płodzień, M. Maśka, M. Lewenstein, T. Grass, and U. Bhattacharya, Topological phase detection through high-harmonic spectroscopy in extended Su-Schrieffer-Heeger chains, Physical Review B 108, 214104 (2023).DOI
  30. [30]C. Ma, X.-B. Bian, and T.-Y. Du, Role of symmetry breaking in high-order harmonic generation from Su-Schrieffer-Heeger systems, Physical Review B 106, 125117 (2022).DOI
  31. [31]S. Pooyan and D. Bauer, Harmonic generation with topological edge states and electron-electron interaction, Phys. Rev. B 111, 014308 (2025).arxiv.org/abs/2406.04114
  32. [32]A. M. Marques and R. G. Dias, Multihole edge states in Su-Schrieffer-Heeger chains with interactions, Physical Review B 95, 115443 (2017).DOI
  33. [33]G. Cáceres-Aravena and D. Bauer, Polaronic effect in high-harmonic generation, arXiv preprint (2025), arXiv:2508.14633 [cond-mat.mes-hall].
  34. [34]J.-X. Chen and X.-B. Bian, Multiplateau structure caused by spectral interference in high-order harmonic generation from disordered Su-Schrieffer-Heeger chains, Physical Review A 109, 033104 (2024).
  35. [35]M. Jangjan and M. V. Hosseini, Floquet engineering of topological metal states and hybridization of edge states with bulk states in dimerized two-leg ladders, Scientific Reports 10, 14256 (2020).arxiv.org/abs/2005.07529
  36. [36]T. Hansen and L. B. Madsen, Doping effects in high-harmonic generation from correlated systems, Phys. Rev. B 106, 235142 (2022).DOI
  37. [37]M. J. Rice and E. J. Mele, Elementary excitations of a linearly conjugated diatomic polymer, Physical Review Letters 49, 1455 (1982).
  38. [38]J. Crank and P. Nicolson, A practical method for numerical evaluation of solutions of partial differential equations of the heat-conduction type, Advances in Computational Mathematics 6, 207 (1996).
  39. [39]A. Askar and A. S. Cakmak, Explicit integration method for the time-dependent Schrödinger equation for collision problems, The Journal of Chemical Physics 68, 2794 (1978).

Paper details

vlnc:W_tpzjn4r2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 9

downloads · 1

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents