betaFeedback

WorksPhysics and Astronomy

Calculations of Chern number: equivalence of real-space and twisted-boundary-condition formulae

Abstract

Chern number is a crucial invariant for characterizing topological feature of two-dimensional quantum systems. Real-space Chern number allows us to extract topological properties of systems without involving translational symmetry, and hence plays an important role in investigating topological systems with disorder or impurity. On the other hand, the twisted boundary condition (TBC) can also be used to define the Chern number in the absence of translational symmetry. Based on the perturbative nature of the TBC under appropriate gauges, we derive the two real-space formulae of Chern number (namely the non-commutative Chern number and the Bott index formula), which are numerically confirmed for the Chern insulator and the quantum spin Hall insulator. Our results not only establish the equivalence between the real-space and TBC formula of the Chern number, but also provide concrete and instructive examples for deriving the real-space topological invariant through the twisted boundary condition.

Introduction

Topological band theory has achieved great success in characterizing the topological nature of quantum matters in the past decades [1, 2, 3, 4]. Beyond the well-known spontaneous breaking theory, topological quantum states are classified by intrinsic topological invariants [4, 5]. In the context of band theory, topological invariants are usually defined on the Bloch manifold. A fascinating nature of topological quantum state is that it is immune to perturbations, i.e. disorder and impurity, provided that the spectral gap and the underlying symmetry are preserved. However, the translational symmetry will be broken in a spatially disordered system, invalidating the band structure. In this situation, the topological band theory fails to give the topological invariant in systems without translational symmetry.

To characterize the topological nature in the presence of disorder, one may use the twisted boundary condition (TBC) to formulate topological invariants [6, 7, 8, 9, 10, 11]. One of the prominent features of the TBC is that it operates in real space. This method works in quasi-periodic boundary condition and does not require any translational symmetry. Physically, topological invariants are closely related to the system’s response of current to the external perturbation. The TBC is equivalent to piercing magnetic flux through the system [7, 12, 13], which results in adiabatic currents in the system [14]. Hence, one is able to compute the corresponding response of current in the system to obtain the topological invariant, e.g. the quantum Hall conductance [7, 8, 15] and polarization [6, 9, 16, 17, 18]. The twist angle is related to the quasi momentum of the eigenstate under certain gauge choice in disorder-free system, connecting the topological invariant defined through quasi momenta and twist angles [19].

Recently, the real-space formula of topological invariant draws many attentions. Unlike the momentum-space topological invariant that are defined on Brillouin manifold, one can construct topological invariants directly in real space, allowing computation without translational symmetry. Roughly speaking, the real-space formula of topological invariant can be classified into two types, (i) the non-commutative form [20, 21, 22, 23, 24] and (ii) the Bott index form [25, 26, 27, 28, 29]. In 1D topological insulators with chiral symmetry, the winding number can be well expressed by these two methods in real space [11, 30]. Moreover, in 1D systems, they are proved to be equivalent to the winding number defined through the TBC in the thermodynamic limit, and hence these two real-space formulae are also equivalent [11]. In 2D Chern insulator, the real-space Chern number is also found to possess the non-commutative form and the Bott index form. Both methods have been extensively used in studying disorder effect to topological insulators. However, it is still unclear whether the real-space formulae of Chern number can be related to the Chern number defined via the TBC.

In this article, we show that the Chern number defined via the TBC and the two real-space formulae (the non-commutative method and the Bott index method) are equivalent in the thermodynamic limit. The key point of the proof is to make use of the perturbative nature of the twist angles in TBC and expand operators and eigenstates up to the first order of θj/Lj\theta_j/L_j with LjL_j being the number of cells in j=x,yj=x,y direction. With the perturbative analysis, the Chern number defined via TBC is reduced to the real-space formulae of Chern number.

The rest of this paper is organized as follows: In Sec. II, we briefly introduce the Chern number defined via the TBC. We consider a non-interacting system and target gapped eigenstates. In Sec. III, we perform numerical calculations for the Haldane model and Kane-Mele model to verify our arguments, including the flatness of Berry curvature and the calculation of Chern number in real space. In Sec. IV, we briefly summarize and discuss the results.

Chern number: from twisted boundary condition to real-space formulae

Chern number defined via twisted boundary condition

In this section, we introduce the TBC and the Chern number defined through TBC. Under the TBC, the lattice’s edges are “glued” together pairwise as a 2D torus, just like the periodic boundary condition (PBC). The core is that we enforce particles to gain an extra phase θ\theta after they tunnel through the boundary, as demonstrated in Figure 1 (a). Apparently, θ=0\theta= 0 corresponds exactly to the PBC.

Schematic illustration of the TBC in a 2D simple lattice

Figure 1. Schematic illustration of the TBC in a 2D simple lattice. In both configurations, edges are glued together in the same manner as the periodic boundary condition. (a) is a general form of the TBC. The gauge field only appears at the boundary, and the particle gains extra phase when it tunnels through the boundary, as indicated by color arrows. (b) is transformed from (a) via a large gauge transformation. The gauge field distributes uniformly throughout the bulk, and the field strength is diluted.

Next, we consider a non-interacting 2D system under TBC. The system’s Hamiltonian satisfies H^(θx+2π,θy)=H^(θx,θy+2π)=H^(θx,θy)\hat{H}(\theta_x + 2\pi,\theta_y) = \hat{H}(\theta_x,\theta_y + 2\pi) = \hat{H}(\theta_x,\theta_y). Then, we target a set of single-particle eigenstates that are gapped to other eigenstates. Note that, if translation symmetry is preserved, these eigenstates correspond to gapped bands. For simplicity, we collect these eigenstates and introduce the following notation

Ψθ=(∣ψ1(θ)⟩ ∣ψ2(θ)⟩ ⋯ ∣ψN(θ)⟩),(1)\mathbf{\Psi}_{\theta} = \left( \lvert\psi_1(\theta) \rangle\ \lvert\psi_2(\theta) \rangle\ \cdots\ \lvert\psi_{\mathcal{N}}(\theta) \rangle\right), \tag*{(1)}

where ∣ψμ(θ)⟩\lvert\psi_\mu(\theta) \rangle is the μ\mu-th eigenstate of the Hamiltonian under TBC, and Nf\mathcal{N}_f is the total number of these targeted states. There is Ψθ†Ψθ=IN\mathbf{\Psi}_{\theta}^{\dagger}\mathbf{\Psi}_{\theta}=I_{\mathcal{N}} with INI_{\mathcal{N}} being the N×N\mathcal{N}\times\mathcal{N} identity matrix, and the projector of this subspace can be expressed as ΨθΨθ†=∑μ∣ψμ(θ)⟩⟨ψμ(θ)∣\mathbf{\Psi}_{\theta}\mathbf{\Psi}_{\theta}^{\dagger}=\sum_\mu\lvert\psi_\mu(\theta)\rangle\langle\psi_\mu(\theta)\rvert. These targeted states are assumed to remain gapped for θx,θy∈[0,2π]\theta_x,\theta_y\in[0,2\pi], which is true in the thermodynamic limit. The Chern number of these targeted states can be expressed through the twist angles (θx,θy)(\theta_x,\theta_y) [8]:

CTBC=12π∫02π∫02πTr⁡[F(θ)] d2θ,(2)C_{\mathrm{TBC}}=\frac{1}{2\pi}\int_{0}^{2\pi}\int_{0}^{2\pi}\operatorname{Tr}\left[\mathcal{F}(\theta)\right]\,\mathrm{d}^{2}\theta, \tag*{(2)}

in which F(θ)\mathcal{F}(\theta) is the non-Abelian Berry curvature

F(θ)=∂θxAy(θ)−∂θyAx(θ)+[Ax(θ),Ay(θ)],\mathcal{F}(\theta)=\partial_{\theta_x}\mathcal{A}_y(\theta)-\partial_{\theta_y}\mathcal{A}_x(\theta)+[\mathcal{A}_x(\theta),\mathcal{A}_y(\theta)],
Aj(θ)=−iΨθ†∂θjΨθ,j=x,y.(3)\mathcal{A}_j(\theta)=-\mathrm{i}\mathbf{\Psi}_{\theta}^{\dagger}\partial_{\theta_j}\mathbf{\Psi}_{\theta},\qquad j=x,y. \tag*{(3)}

Note that Aj(θ)\mathcal{A}_j(\theta) is a N\mathcal{N}-by-N\mathcal{N} matrix. The minus sign in Aj(θ)\mathcal{A}_j(\theta) is related to the sign of twist angles defined in the TBC. Eq. (2) can well capture the topological property (quantum Hall conductance) of the system [8].

Alternatively, by introducing the projector that projects onto the targeted states P^θ=ΨθΨθ†\hat{P}_{\theta}=\mathbf{\Psi}_{\theta}\mathbf{\Psi}_{\theta}^{\dagger}, the Chern number can be equivalently expressed as [31]

CTBC=12πi∫02π∫02πTr⁡(P^θ[∂θxP^θ,∂θyP^θ]) d2θ.(4)C_{\mathrm{TBC}}=\frac{1}{2\pi\mathrm{i}}\int_{0}^{2\pi}\int_{0}^{2\pi}\operatorname{Tr}\left(\hat{P}_{\theta}\left[\partial_{\theta_x}\hat{P}_{\theta},\partial_{\theta_y}\hat{P}_{\theta}\right]\right)\,\mathrm{d}^{2}\theta. \tag*{(4)}

Given that the targeted states are gapped, the projector will finally return to itself after the twist angle changes 2π2\pi, that is: P^θ+2πej=P^θ\hat{P}_{\theta+2\pi\mathbf{e}_j}=\hat{P}_{\theta}, with ej\mathbf{e}_j the unit vector along j=x,yj=x,y direction. In other words, the projector is a periodic function of twist angles.

Large gauge transformation and uniform gauge

Under the TBC, the extra phase gained by particle at the boundary can be regarded as a result of gauge field, and therefore there exists gauge freedom. It is beneficial to transform the gauge field at the boundary to a uniformly distributed gauge field throughout the whole system. By introducing the twist operator [32, 33]

U^θ=exp⁡(iθxLxr^x+iθyLyr^y),(5)\hat{U}_{\theta}=\exp\left(i\frac{\theta_x}{L_x}\hat{r}_x+i\frac{\theta_y}{L_y}\hat{r}_y\right), \tag*{(5)}

where r^j=∑rrjn^r\hat{r}_j=\sum_r r_j\hat{n}_r is the position operator in j=x,yj=x,y direction, we can bring the original system under TBC to a unitarily equivalent system via the large gauge transformation

H~(θ)=U^θH^(θ)U^θ−1.(6)\tilde{H}(\theta)=\hat{U}_{\theta}\hat{H}(\theta)\hat{U}_{\theta}^{-1}. \tag*{(6)}

Here, H~(θ)\tilde{H}(\theta) indicates the transformed Hamiltonian. The two unitarily equivalent systems are illustrated in Figure 1. To distinguish these two gauges, we use the tilde notation to stress that the operator (eigenstate) is under the gauge that the twist angles uniformly distribute. For convenience, we will call this gauge the uniform gauge, while the original TBC is called boundary gauge. In this case, the Chern number can still be written as the same form

C~TBC=12π∫02π∫02πTr⁡[F~(θ)]d2θ,(7)\tilde{C}_{\mathrm{TBC}}=\frac{1}{2\pi}\int_0^{2\pi}\int_0^{2\pi}\operatorname{Tr}\left[\tilde{F}(\theta)\right]\mathrm{d}^2\theta, \tag*{(7)}

and

F~(θ)=∂θxA~y(θ)−∂θyA~x(θ)+[A~x(θ),A~y(θ)],\tilde{F}(\theta)=\partial_{\theta_x}\tilde{A}_y(\theta)-\partial_{\theta_y}\tilde{A}_x(\theta)+\left[\tilde{A}_x(\theta),\tilde{A}_y(\theta)\right],
A~j(θ)=−iΨ~θ†∂θjΨ~θ,j=x,y.(8)\tilde{A}_j(\theta)=-i\tilde{\Psi}_{\theta}^{\dagger}\partial_{\theta_j}\tilde{\Psi}_{\theta},\qquad j=x,y. \tag*{(8)}

Note that Ψ~θ=U^θΨθ\tilde{\Psi}_{\theta}=\hat{U}_{\theta}\Psi_{\theta}. It can be proved that Eq. (3) and Eq. (8) are equivalent: C~TBC=CTBC\tilde{C}_{\mathrm{TBC}}=C_{\mathrm{TBC}}, see the proof in A. Similar to Eq. (4), we can formulate the Chern number through the projector P~θ=Ψ~θΨ~θ†\tilde{P}_{\theta}=\tilde{\Psi}_{\theta}\tilde{\Psi}_{\theta}^{\dagger}

C~TBC=12πi∫02π∫02πTr⁡(P~θ[∂θxP~θ,∂θyP~θ])d2θ,(9)\tilde{C}_{\mathrm{TBC}}=\frac{1}{2\pi i}\int_0^{2\pi}\int_0^{2\pi}\operatorname{Tr}\left(\tilde{P}_{\theta}\left[\partial_{\theta_x}\tilde{P}_{\theta},\partial_{\theta_y}\tilde{P}_{\theta}\right]\right)\mathrm{d}^2\theta, \tag*{(9)}

in which the projector satisfies

P~θ=U^θP^θU^θ−1.(10)\tilde{P}_{\theta}=\hat{U}_{\theta}\hat{P}_{\theta}\hat{U}_{\theta}^{-1}. \tag*{(10)}

In the transformed Hamiltonian (6), the twist angle θ\theta is transformed to distribute in all tunneling terms of the system. Particles gain a “diluted” phase factor during the tunneling process, that is

U^θc^r†U^θ−1=ei(θxLxdx+θyLydy)c^r+d†c^r.(11)\hat{U}_{\theta}\hat{c}_{r}^{\dagger}\hat{U}_{\theta}^{-1}=e^{i\left(\frac{\theta_x}{L_x}d_x+\frac{\theta_y}{L_y}d_y\right)}\hat{c}_{r+d}^{\dagger}\hat{c}_{r}. \tag*{(11)}

When the tunneling range is finite dx,y≪Lx,yd_{x,y}\ll L_{x,y}, the twist angle appears as θj/Lj\theta_j/L_j, j=x,yj=x,y, which can be considered as a perturbation. Hence, it is desirable to expand operators or eigenstates up to the first order of 1/Lj1/L_j for sufficiently large system away from the critical point. This also explains why the gapped eigenstate will remain gapped for arbitrary θx,y\theta_{x,y} in the thermodynamic limit Lx,y→∞L_{x,y}\to\infty, as the energy gap is only weakly perturbed. In next subsections, we shall make use of the perturbative nature of the twist angle to derive the two real-space formulae of the Chern number.

Real-space Chern number via non-commutative method

In this subsection, we derive the the non-commutative real-space formula of the Chern number. This formula is firstly proposed in Ref. [20, 23], which is based on the momentum-space formula of the Chern number [34]

C=12πi∫B.Z.Tr⁡{P^(k)[∂kxP^(k),∂kyP^(k)]}d2k,(12)C=\frac{1}{2\pi i}\int_{\mathrm{B.Z.}}\operatorname{Tr}\left\{\hat{P}(\mathbf{k})\left[\partial_{k_x}\hat{P}(\mathbf{k}),\partial_{k_y}\hat{P}(\mathbf{k})\right]\right\}\mathrm{d}^2\mathbf{k}, \tag*{(12)}

where P^(k)=∑n∈occ.∣ukn⟩⟨ukn∣\hat{P}(\mathbf{k})=\sum_{n\in\mathrm{occ.}}|u_k^n\rangle\langle u_k^n| is the projector onto Bloch states of occupied bands with momentum k\mathbf{k}. To derive the real-space formula, their idea is to transform the partial derivative and the integral in Brillouin zone to real space [20, 23], which leads to the following non-commutative form of Chern number

C=−2πi∑α⟨r0,α∣P^[[x^,P^],[y^,P^]]∣r0,α⟩,(13)C=-2\pi i\sum_{\alpha}\langle r_0,\alpha|\hat{P}\left[\left[\hat{x},\hat{P}\right],\left[\hat{y},\hat{P}\right]\right]|r_0,\alpha\rangle, \tag*{(13)}

where P^\hat{P} is the projector onto eigenstates in occupied bands, and ∣r0,α⟩|r_0,\alpha\rangle is the real-space basis denoting the cell at r0r_0 and α\alpha is the label of internal orbits within the cell. This formula has been successfully applied to various systems [20, 35, 36].

Next, we would like to show that the non-commutative Chern number can be equivalently derived through the TBC Chern number [Eq. (9)]. For this purpose, we will make use of the perturbative nature of the twist angle under the uniform gauge. Firstly, we expand the projector in terms of θj/Lj\theta_j/L_j up to the first order near θ=0\theta=0:

P~θ=P^+θxLx∂P~θ∂(θx/Lx)∣θ=0+θyLy∂P~θ∂(θy/Ly)∣θ=0+O(1Lx2+1Ly2),(14)\tilde{P}_{\theta}=\hat{P}+\left.\frac{\theta_x}{L_x}\frac{\partial\tilde{P}_{\theta}}{\partial(\theta_x/L_x)}\right|_{\theta=0}+\left.\frac{\theta_y}{L_y}\frac{\partial\tilde{P}_{\theta}}{\partial(\theta_y/L_y)}\right|_{\theta=0}+O\left(\frac{1}{L_x^2}+\frac{1}{L_y^2}\right), \tag*{(14)}

where we use the fact that P^=P~θ=0\hat{P}=\tilde{P}_{\theta=0} since θ=0\theta=0 corresponds to general systems under PBC. Substituting Eq. (14) into Eq. (9) and keeping up to the leading order lead to:

P~θ[∂θxP~θ,∂θyP~θ]≈P^[(∂θxP~θ)∣θ=0,(∂θyP~θ)∣θ=0](15)\widetilde{P}_{\theta}\left[\partial_{\theta_x}\widetilde{P}_{\theta},\partial_{\theta_y}\widetilde{P}_{\theta}\right]\approx\widehat{P}\left[\left.\left(\partial_{\theta_x}\widetilde{P}_{\theta}\right)\right|_{\theta=0},\left.\left(\partial_{\theta_y}\widetilde{P}_{\theta}\right)\right|_{\theta=0}\right] \tag*{(15)}

which is independent on the twist angle. This fact implies that the trace of the Berry curvature defined through TBC should be also independent on the twist angle Tr⁡[F~(θ)]≈constant\operatorname{Tr}[\widetilde{F}(\theta)]\approx\text{constant} in the thermodynamic limit Lx,Ly→∞L_x,L_y\to\infty. We shall examine this point in Sec. III.

Since the trace of the Berry curvature is a constant for sufficiently large systems, the Chern number [Eq. (9)] can be approximated without implementing integrations [37]:

C~TBC≈−2πiTr⁡(P^[(∂θxP~θ)∣θ=0,(∂θyP~θ)∣θ=0]).(16)\widetilde{C}_{\mathrm{TBC}}\approx-2\pi i\operatorname{Tr}\left(\widehat{P}\left[\left.\left(\partial_{\theta_x}\widetilde{P}_{\theta}\right)\right|_{\theta=0},\left.\left(\partial_{\theta_y}\widetilde{P}_{\theta}\right)\right|_{\theta=0}\right]\right). \tag*{(16)}

Next, we let θx=2π,θy=0\theta_x=2\pi,\theta_y=0, which maintains the perturbative condition since 2π/Lx≪12\pi/L_x\ll1. According to Eq. (10), we find the following useful relation

P~(2π,0)=P^+2πLx∂P~θ∂(θx/Lx)∣θ=0+O(1Lx2)=U^2πxP^(U^2πx)−1,(17)\begin{aligned} \widetilde{P}_{(2\pi,0)}&=\widehat{P}+\left.\frac{2\pi}{L_x}\frac{\partial\widetilde{P}_{\theta}}{\partial(\theta_x/L_x)}\right|_{\theta=0}+O\left(\frac{1}{L_x^2}\right)\\ &=\widehat{U}_{2\pi}^{x}\widehat{P}\left(\widehat{U}_{2\pi}^{x}\right)^{-1}, \tag*{(17)} \end{aligned}

where we introduce the twist operator along single direction U^θjj≡exp⁡(iθjr^j/Lj)\widehat{U}_{\theta_j}^{j}\equiv\exp(i\theta_j\widehat{r}_j/L_j), j=x,yj=x,y for simplicity. A similar result can be obtained by letting θx=0,θy=2π\theta_x=0,\theta_y=2\pi. Up to the first order of 1/Lj1/L_j, j=x,yj=x,y, Eq. (17) provides a very convenient approach to approximate the partial derivative w.r.t. the twist angle:

2πLj∂P~θ∂(θj/Lj)∣θ=0=2π∂P~θ∂θj∣θ=0≈P~2πej−P~=U^2πjP^(U^2πj)−1−P^.(18)\begin{aligned} \left.\frac{2\pi}{L_j}\frac{\partial\widetilde{P}_{\theta}}{\partial(\theta_j/L_j)}\right|_{\theta=0}&=2\pi\left.\frac{\partial\widetilde{P}_{\theta}}{\partial\theta_j}\right|_{\theta=0}\\ &\approx\widetilde{P}_{2\pi\mathbf{e}_j}-\widetilde{P}\\ &=\widehat{U}_{2\pi}^{j}\widehat{P}\left(\widehat{U}_{2\pi}^{j}\right)^{-1}-\widehat{P}. \tag*{(18)} \end{aligned}

Furthermore, by using the Baker-Campbell-Hausdorff formula, we have

ei2πLjr^jP^e−i2πLjr^j=P^+i2πLj[r^j,P^]+O(1Lj2),(19)e^{i\frac{2\pi}{L_j}\widehat{r}_j}\widehat{P}e^{-i\frac{2\pi}{L_j}\widehat{r}_j}=\widehat{P}+i\frac{2\pi}{L_j}\left[\widehat{r}_j,\widehat{P}\right]+O\left(\frac{1}{L_j^2}\right), \tag*{(19)}

and therefore the partial derivative of the projector can be further written as

∂P~θ∂θj∣θ=0≈iLj[r^j,P~],j=x,y.(20)\left.\frac{\partial\widetilde{P}_{\theta}}{\partial\theta_j}\right|_{\theta=0}\approx\frac{i}{L_j}\left[\widehat{r}_j,\widetilde{P}\right],\qquad j=x,y. \tag*{(20)}

By combining Eqs. (9), (15), (20), we consequently obtain the following real-space form of the Chern number

C~TBC≈2πiLxLyTr⁡(P^[[r^x,P^],[r^y,P^]])=2πiLxLy∑r,α⟨r,α∣P^[[r^x,P^],[r^y,P^]]∣r,α⟩.(21)\begin{aligned} \widetilde{C}_{\mathrm{TBC}}&\approx\frac{2\pi i}{L_xL_y}\operatorname{Tr}\left(\widehat{P}\left[\left[\widehat{r}_x,\widehat{P}\right],\left[\widehat{r}_y,\widehat{P}\right]\right]\right)\\ &=\frac{2\pi i}{L_xL_y}\sum_{r,\alpha}\langle r,\alpha|\widehat{P}\left[\left[\widehat{r}_x,\widehat{P}\right],\left[\widehat{r}_y,\widehat{P}\right]\right]|r,\alpha\rangle. \tag*{(21)} \end{aligned}

It can be found that this formula is quite similar to the non-commutative Chern number [Eq. (13)] proposed before (up to a minus sign, depending on the convention of the coordinate). A minor difference is that Eq. (13) performs trace only within a single cell, while Eq. (21) traces for all cells and then averages them. For clean systems with translation symmetry, one can easily deduce that these two approaches are well consistent. As for disordered system, they are also consistent if we average over many random realizations. If the system is inhomogeneous, i.e. a harmonic trap is imposed, it is more reasonable to consider an average over all sites. We also point out that the approximation in Eq. (18) is relatively rough in finite systems. One can use a higher-order finite difference method to gain a more accurate result, see details in B.

Real-space Chern number via Bott index method

In this section, we derive the Bott-index form of the real-space Chern number through the TBC. The Bott index form of the Chern number is proposed to efficiently calculate the intrinsic topological nature in real space [25, 26, 27, 28, 29, 38], and has been applied widely in various systems [39, 40, 41, 42, 43, 44]. It is pointed out that the Bott index is equivalent to the quantum Hall conductance [45]. The Bott index is considered as a winding number index of matrices that almost represents the disk or annulus [27, 45]. For Chern insulators, the Bott index reads as

Bott⁡(Ux,Uy)=12πitr⁡[log⁡(UxUyUx†Uy†)],(22)\operatorname{Bott}(U_x,U_y)=\frac{1}{2\pi i}\operatorname{tr}\left[\log\left(U_xU_yU_x^\dagger U_y^\dagger\right)\right], \tag*{(22)}

where Uj=Ψ†U^jjΨ,j=x,yU_j=\Psi^\dagger\widehat{U}_j^j\Psi,j=x,y. Here, the notation “tr⁡()\operatorname{tr}()” means the trace operation of the matrix, which implies the trace in the subspace spanned by targeted states.

Next, we show that the Bott index Chern number can be derived from the Chern number defined through TBC [Eq. (7)]. The idea is also to use the perturbative nature of the twist angle. However, to derive a Bott index form, we shall take another route. Firstly, it can be proved that the Chern number can be written as a winding of Berry phase (line integral) [31]

C~TBC=12π∫02πdθx∂∂θxϕBerryy(θx)(23)\widetilde{C}_{\mathrm{TBC}}=\frac{1}{2\pi}\int_{0}^{2\pi}d\theta_x\frac{\partial}{\partial\theta_x}\phi_{\mathrm{Berry}}^{y}(\theta_x) \tag*{(23)}

in which

ϕBerryy(θx)=∫02πdθytr⁡[Ay(θ)](24)\phi_{\mathrm{Berry}}^{y}(\theta_x)=\int_{0}^{2\pi}d\theta_y\operatorname{tr}\left[\mathcal{A}_y(\theta)\right] \tag*{(24)}

is the Berry phase along the yy direction. Using the perturbative nature of the twist angle, one can obtain the

Berry phase in real space via the following expression [19]

ϕBerry(θx)=−itr⁡[exp⁡(∫02πA~x(θ) dθy)]≈−itr⁡[log⁡(Ψ~(θx,0)†Ψ~(θx,2π))]=−itr⁡[log⁡(Ψ(θx,0)†U^2πyΨ(θx,0))].(25)\phi_{\mathrm{Berry}}(\theta_x)=-i\operatorname{tr}\left[\exp\left(\int_{0}^{2\pi}\widetilde{\mathcal{A}}_x(\boldsymbol{\theta})\,\mathrm{d}\theta_y\right)\right] \approx-i\operatorname{tr}\left[\log\left(\widetilde{\Psi}_{(\theta_x,0)}^\dagger\widetilde{\Psi}_{(\theta_x,2\pi)}\right)\right] =-i\operatorname{tr}\left[\log\left(\Psi_{(\theta_x,0)}^\dagger\widehat{U}^{y}_{2\pi}\Psi_{(\theta_x,0)}\right)\right]. \tag*{(25)}

where we have used the fact that Ψ~(θx,2π)=U^2πyΨ~(θx,0)\widetilde{\Psi}_{(\theta_x,2\pi)}=\widehat{U}^{y}_{2\pi}\widetilde{\Psi}_{(\theta_x,0)}. The above formula can be equivalently formulated via the projected position operator P^y(θx)≡P~(θx,0)U^2πyP~(θx,0)\widehat{P}_y(\theta_x)\equiv\widetilde{P}_{(\theta_x,0)}\widehat{U}^{y}_{2\pi}\widetilde{P}_{(\theta_x,0)} such that

ϕBerry(θx)≈−iTr⁡[log⁡(P^y(θx))].(26)\phi_{\mathrm{Berry}}(\theta_x)\approx-i\operatorname{Tr}\left[\log\left(\widehat{P}_y(\theta_x)\right)\right]. \tag*{(26)}

By differentiating Eq. (26) w.r.t. the twist angle θx\theta_x, we find (see detailed derivation in D)

∂∂θxϕBerry(θx)=−iTr⁡[P^y(θx)†∂∂θxP^y(θx)].(27)\frac{\partial}{\partial\theta_x}\phi_{\mathrm{Berry}}(\theta_x)=-i\operatorname{Tr}\left[\widehat{P}_y(\theta_x)^\dagger\frac{\partial}{\partial\theta_x}\widehat{P}_y(\theta_x)\right]. \tag*{(27)}

Together with Eq. (23), one may find the Chern number is the winding number of projected position operator.

Next, we further treat the twist angle θx/Lx\theta_x/L_x as a perturbation term, and expand the projected position operator P^y(θx)\widehat{P}_y(\theta_x) up to the first order

P^y(θx)=P^y(0)+θx[∂∂θxP^y(θx)]θx=0+O(1Lx2).(28)\widehat{P}_y(\theta_x)=\widehat{P}_y(0)+\theta_x\left[\frac{\partial}{\partial\theta_x}\widehat{P}_y(\theta_x)\right]_{\theta_x=0}+O\left(\frac{1}{L_x^2}\right). \tag*{(28)}

This fact means, up to the first order of 1/Lx1/L_x, the Berry phase (polarization) along yy direction is a linear function of θx\theta_x, which is consistent with previous studies [46, 47, 48]. By substituting the expansion in Eq. (28) into Eq. (26) and integrating the twist angle θx\theta_x, we find the Chern number can be approximated to

C~TBC=−i2πTr⁡{P^y(0)†[∂∂θxP^y(θx)]θx=0}+O(1Lx2).(29)\widetilde{C}_{\mathrm{TBC}}=-\frac{i}{2\pi}\operatorname{Tr}\left\{\widehat{P}_y(0)^\dagger\left[\frac{\partial}{\partial\theta_x}\widehat{P}_y(\theta_x)\right]_{\theta_x=0}\right\}+O\left(\frac{1}{L_x^2}\right). \tag*{(29)}

Then, we can make use of the approximation

P^y(2π)−P^y(0)≈2π[∂∂θxP^y(θx)]θx=0,(30)\widehat{P}_y(2\pi)-\widehat{P}_y(0)\approx2\pi\left[\frac{\partial}{\partial\theta_x}\widehat{P}_y(\theta_x)\right]_{\theta_x=0}, \tag*{(30)}

and then approximately we have

C~TBC≈12πiTr⁡[P^y(0)†P^y(2π)−P^y(0)†P^y(0)].(31)\widetilde{C}_{\mathrm{TBC}}\approx\frac{1}{2\pi i}\operatorname{Tr}\left[\widehat{P}_y(0)^\dagger\widehat{P}_y(2\pi)-\widehat{P}_y(0)^\dagger\widehat{P}_y(0)\right]. \tag*{(31)}

To obtain the Bott index form, we convert the trace operation (Tr) to the matrix trace (tr) in the subspace spanned by the targeted states

Tr⁡[P^y(0)†P^y(2π)−P^y(0)†P^y(0)]=tr⁡(Uy†UxUyUx†−Uy†Uy).(32)\operatorname{Tr}\left[\widehat{P}_y(0)^\dagger\widehat{P}_y(2\pi)-\widehat{P}_y(0)^\dagger\widehat{P}_y(0)\right] =\operatorname{tr}\left(U_y^\dagger U_xU_yU_x^\dagger-U_y^\dagger U_y\right). \tag*{(32)}

In Eq. (32), we have used the fact that P^y(2π)=U^2πxP^y(0)(U^2πx)−1\widehat{P}_y(2\pi)=\widehat{U}^{x}_{2\pi}\widehat{P}_y(0)(\widehat{U}^{x}_{2\pi})^{-1} and [U^2πx,U^2πy]=0[\widehat{U}^{x}_{2\pi},\widehat{U}^{y}_{2\pi}]=0. For gapped targeted states, matrices Ux,yU_{x,y} are quasi-unitary Uy†Uy≈INU_y^\dagger U_y\approx I_N (see proof in C), in which INI_N is the N×NN\times N identity matrix in the subspace spanned by the targeted states. In practical calculations with finite length, one can perform the singular value decomposition (SVD) for these two matrices: Ux,y=UΣV−1U_{x,y}=U\Sigma V^{-1}, in which U,V∈U(N)U,V\in U(N) are unitary matrices, and Σ\Sigma is a diagonal matrix. The singular values are non-zero and close to identity as long as the targeted states are gapped. Then, one can enforce all singularity to be identity Σ=IN\Sigma=I_N, and then the unitarity of the matrix can be guaranteed: Ux,y=UV−1∈U(N)U_{x,y}=UV^{-1}\in U(N).

Finally, given that Uy†UxUyUx†U_y^\dagger U_xU_yU_x^\dagger is close to the identity matrix, we can utilize the matrix logarithm and approximate Eq. (31) as:

C~TBC=12πitr⁡[log⁡(Uy†UxUyUx†)],(33)\widetilde{C}_{\mathrm{TBC}}=\frac{1}{2\pi i}\operatorname{tr}\left[\log\left(U_y^\dagger U_xU_yU_x^\dagger\right)\right], \tag*{(33)}

which is consistent with Eq. (22). Note that the order of these matrices can be circularly permuted in 3232 due to the trace operation. The derivation is quite similar to the real-space representation of the winding number in Ref. [11]. Owing to the unitarity of the matrices Ux,yU_{x,y}, one can conclude that det⁡(Uy†UxUyUx†)∈R\det\left(U_y^\dagger U_xU_yU_x^\dagger\right)\in\mathbb{R}, and therefore the Chern number (Bott index) here should be strictly an integer.

Real-space formulae of the spin Chern number in quantum spin Hall insulator

In this subsection, we focus on the QSH insulator [49, 50] and derive the corresponding real-space topological invariant. Unlike the Chern insulator, the QSH insulator obeys the time-reversal symmetry, resulting in a vanishing total Chern number for the gapped band. Generally speaking, the QSH insulator cannot classified using the conventional Chern number approach. However, in Ref. [51], Kane and Mele demonstrate that it can be classified by a Z2\mathbb{Z}_2 quantity instead. For a spin-1/21/2 system with well conserved szs_z, one can faithfully define a spin Chern number. In this case, the QSH insulator can be simply seen as a two copies of Chern insulators that preserving the time-reversal symmetry. The spin Chern number can be easily computed in the two different spin sectors in momentum space.

When the conservation of szs_z is violated, the two spin sectors become mixed and a well-defined formula for computing the spin Chern number in momentum space is not readily available. However, the topological property of the QSH insulator should remain as long as the spectral gap is not closed. To properly capture the topological invariant of QSH insulators, one of the effective methods is to split the projector onto the occupied band into two sectors: P^(k)=P^+(k)⊕P^−(k)\widehat{P}(\boldsymbol{k})=\widehat{P}_{+}(\boldsymbol{k})\oplus\widehat{P}_{-}(\boldsymbol{k}) [21, 22]. Subsequently, the Chern number can be computed in each sector using Eq. (12) individually. With this approach, one can derive the real-space formula of the spin Chern number for QSH as the Chern insulator, including the non-commutative method [21, 22]:

C±=−2πi∑α⟨r0,α∣P^±[[r^x,P^±],[r^y,P^±]]∣r0,α⟩,(34)C_{\pm}=-2\pi i\sum_{\alpha}\langle r_{0},\alpha|\hat{P}_{\pm}\left[[\hat{r}_{x},\hat{P}_{\pm}],[\hat{r}_{y},\hat{P}_{\pm}]\right]|r_{0},\alpha\rangle, \tag*{(34)}

where P^±\hat{P}_{\pm} is the spectral projector onto the sector with positive (++) or negative (−-) eigenvalues of P^σ^zP^\hat{P}\hat{\sigma}_{z}\hat{P}. The Bott index method can be also constructed through this spectral decomposition [41, 42]:

C±=12πi{tr⁡[log⁡(V±yV±xV±y†V±x†)]},(35)C_{\pm}=\frac{1}{2\pi i}\left\{\operatorname{tr}\left[\log\left(V_{\pm}^{y}V_{\pm}^{x}V_{\pm}^{y\dagger}V_{\pm}^{x\dagger}\right)\right]\right\}, \tag*{(35)}

with V±j=P^±e±i2πLr^jP^±+(I−P^±)V_{\pm}^{j}=\hat{P}_{\pm}e^{\pm i\frac{2\pi}{L}\hat{r}_{j}}\hat{P}_{\pm}+\left(I-\hat{P}_{\pm}\right), j=x,yj=x,y. The spin-Chern number is then identified as Cs=(C+−C−)/2C_{\mathrm{s}}=(C_{+}-C_{-})/2.

On the other hand, it is still possible to define a generalized TBC for the spinful system [21, 52]:

c^ri†trirjc^rj→{c^ri†trirjeiθxΓxc^rj,rj→ri accros boundaryc^ri†trirjeiθyΓyc^rj,(36)\hat{c}_{r_{i}}^{\dagger}t_{r_{i}r_{j}}\hat{c}_{r_{j}}\rightarrow \begin{cases} \hat{c}_{r_{i}}^{\dagger}t_{r_{i}r_{j}}e^{i\theta_{x}\Gamma_{x}}\hat{c}_{r_{j}}, & r_{j}\rightarrow r_{i}\text{ accros boundary}\\ \hat{c}_{r_{i}}^{\dagger}t_{r_{i}r_{j}}e^{i\theta_{y}\Gamma_{y}}\hat{c}_{r_{j}}, \end{cases} \tag*{(36)}

in which c^rj†=(c^↑,rj†,c^↓,rj†)\hat{c}_{r_{j}}^{\dagger}=(\hat{c}_{\uparrow,r_{j}}^{\dagger},\hat{c}_{\downarrow,r_{j}}^{\dagger}), tijt_{ij} is 2×22\times2 tunneling matrix, θx,y\theta_{x,y} are twist angles, and Γx,y\Gamma_{x,y} are 2×22\times2 matrices called internal generators. Such kind of boundary condition is a generalization from the conventional TBC mentioned above. Then, one can use the TBC formula of Chern number to compute the spin Chern number:

Cs=14π∫02π∫02πTr⁡[F(θ)]d2θ,(37)C_{\mathrm{s}}=\frac{1}{4\pi}\int_{0}^{2\pi}\int_{0}^{2\pi}\operatorname{Tr}\left[\mathcal{F}(\theta)\right]\mathrm{d}^{2}\theta, \tag*{(37)}

in which there is an extra 1/21/2 factor compared to the general formula of Chern number in Eq. (2). Similarly, we can introduce the generalized twist operator

U^θΓ=exp⁡(iθxLx∑rrxc^r†Γxc^r+iθyLy∑rryc^r†Γyc^r)(38)\hat{U}_{\theta}^{\Gamma}=\exp\left(i\frac{\theta_{x}}{L_{x}}\sum_{r}r_{x}\hat{c}_{r}^{\dagger}\Gamma_{x}\hat{c}_{r}+i\frac{\theta_{y}}{L_{y}}\sum_{r}r_{y}\hat{c}_{r}^{\dagger}\Gamma_{y}\hat{c}_{r}\right) \tag*{(38)}

to transform the boundary term to the bulk

H~(θ)=U^θΓH^(θ)(U^θΓ)−1.(39)\widetilde{H}(\theta)=\hat{U}_{\theta}^{\Gamma}\hat{H}(\theta)\left(\hat{U}_{\theta}^{\Gamma}\right)^{-1}. \tag*{(39)}

According to Ref. [52], we can chose Γx=S^\Gamma_{x}=\hat{S} and Γy=1\Gamma_{y}=1 with S^\hat{S} being the spin operator. In this case, the system obeys the general TBC along yy direction, and the generalized TBC along xx direction. If the system has conserved spin SS, then Eq. (39) will leads to a uniformly distributed gauge field (twist angle). When SS is not conserved, the gauge field is not uniform in xx direction. More precisely, the tunneling matrix will be slightly rotated along xx direction, see Appendix E for details. However, as long as the tunneling of particles is finite-range, the gauge field can still be treated as a perturbation.

Following the same route in Sec. II, we can replace the position operator by a generalized position operator along xx:

r^x→r^xΓ=∑rrxc^r†Γxc^r,(40)\hat{r}_{x}\rightarrow\hat{r}_{x}^{\Gamma}=\sum_{r}r_{x}\hat{c}_{r}^{\dagger}\Gamma_{x}\hat{c}_{r}, \tag*{(40)}

and we leave the yy direction unchanged. Then, we can obtain the non-commutative formula of the spin Chern number:

Cs=πiLxLy∑r,α⟨r,α∣P^[[r^xΓ,P^],[r^y,P^]]∣r,α⟩(41)C_{\mathrm{s}}=\frac{\pi i}{L_{x}L_{y}}\sum_{r,\alpha}\langle r,\alpha|\hat{P}\left[[\hat{r}_{x}^{\Gamma},\hat{P}],[\hat{r}_{y},\hat{P}]\right]|r,\alpha\rangle \tag*{(41)}

and the Bott index form:

Cs=14πitr⁡[log⁡(UyΓUxΓUyΓ†UxΓ†)],(42)C_{\mathrm{s}}=\frac{1}{4\pi i}\operatorname{tr}\left[\log\left(U_{y}^{\Gamma}U_{x}^{\Gamma}U_{y}^{\Gamma\dagger}U_{x}^{\Gamma\dagger}\right)\right], \tag*{(42)}

with UxΓ=Ψ†exp⁡(i2πLxr^xΓ)ΨU_{x}^{\Gamma}=\Psi^{\dagger}\exp\left(i\frac{2\pi}{L_{x}}\hat{r}_{x}^{\Gamma}\right)\Psi. It can be noted that these two formulae are slightly different from that in Eq. (34) and (39). In fact, the choice of the generalized TBC is not unique. For example, Refs. [53, 54] introduce the concept of Chern number matrix based on the TBC. This will also allow us to establish a real-space formula of the Chern number matrix. For spin-1/2 QSH insulator, the twist angles, dubbed θx,y,σ=↑,↓\theta_{x,y,\sigma}=\uparrow,\downarrow, are addressed respectively to the two spinful particles when they tunnel through the boundary. This allows one to define the 2×22\times2 Chern number matrix [54, 55]:

Cσ,σ′=12πi∫02π∫02πTr⁡(P^θ[∂θxσP^θ,∂θyσ′P^θ])dθxσdθyσ′,(43)C_{\sigma,\sigma'}=\frac{1}{2\pi i}\int_{0}^{2\pi}\int_{0}^{2\pi}\operatorname{Tr}\left(\hat{P}_{\theta}\left[\partial_{\theta_{x}^{\sigma}}\hat{P}_{\theta},\partial_{\theta_{y}^{\sigma'}}\hat{P}_{\theta}\right]\right)\mathrm{d}\theta_{x}^{\sigma}\mathrm{d}\theta_{y}^{\sigma'}, \tag*{(43)}

and the TBC spin Chern number can be obtained via [54]

Cs=12∑σ,σ′sgn⁡(σ)Cσ,σ′,(44)C_{\mathrm{s}}=\frac{1}{2}\sum_{\sigma,\sigma'}\operatorname{sgn}(\sigma)C_{\sigma,\sigma'}, \tag*{(44)}

where we denote sgn⁡(σ)=+1\operatorname{sgn}(\sigma)=+1 if σ=↑\sigma=\uparrow and vice versa. The Chern number matrix method has been proven valid [54, 55, 56].

Eq. (43) allows us to follow the strategy in Sec. II to derive the real-space formula of the spin Chern number. Without loss of generality, we shall consider a spin-1/2 system below. Firstly, we need to introduce position operators in each spin sector: r^jσ=∑rrjn^rσ\hat{r}_{j}^{\sigma}=\sum_{r}r_{j}\hat{n}_{r}^{\sigma}, j=x,yj=x,y; σ=↑,↓\sigma=\uparrow,\downarrow, and the associated twist operator along j=x,yj=x,y direction:

U^θjσσ=exp⁡(iθjσLjr^jσ).(45)\hat{U}_{\theta_{j}^{\sigma}}^{\sigma}=\exp\left(i\frac{\theta_{j}^{\sigma}}{L_{j}}\hat{r}_{j}^{\sigma}\right). \tag*{(45)}

Under the large gauge transformation, which is similar to Eq. (39) with this twist operator, one can also transform the gauge field at the boundary to the bulk. Although such kind of transformation can not make the gauge field distributed uniformly, particles will only feel a slowly changing gauge field, and the change rate is in the order of 1/Lx,y1/L_{x,y}. Hence, the perturbative analysis in § 1.1 B can be applied. With some similar calculations, the real-space version of the spin Chern number matrix reads as

Cσ,σ′=πiLxLy∑r,α⟨r,α∣[[x^,P^],[y^,P^]]∣r,α⟩,(46)C_{\sigma,\sigma'}=\frac{\pi i}{L_xL_y}\sum_{\mathbf{r},\alpha}\left\langle\mathbf{r},\alpha\left|\left[\left[\hat{x},\hat{P}\right],\left[\hat{y},\hat{P}\right]\right]\right|\mathbf{r},\alpha\right\rangle, \tag*{(46)}

and

Cσ,σ′=14πitr⁡[log⁡(Uyσ′†Uxσ′†UyσUxσ)],(47)C_{\sigma,\sigma'}=\frac{1}{4\pi i}\operatorname{tr}\left[\log\left(U_y^{\sigma'\dagger}U_x^{\sigma'\dagger}U_y^\sigma U_x^\sigma\right)\right], \tag*{(47)}
Ujσ=Ψ†exp⁡(i2πLjσ^j^)Ψ,j=x,y; σ=↑,↓.U_j^\sigma=\Psi^\dagger\exp\left(i\frac{2\pi}{L_j}\hat{\sigma}\hat{j}\right)\Psi,\quad j=x,y;\ \sigma=\uparrow,\downarrow.

Numerical verifications in the Haldane model and the Kane-Mele model

In this section, we present some numerical calculations to verify our results, including the flatness of the Berry curvature defined through TBC and applications of the two real-space formulae of the Chern number. We shall use the Haldane model to verify our result for the Chern insulator, and use the Kane-Mele model for the QSH insulator.

The Haldane model

Here, we consider the celebrated Haldane model [57], which realizes the Chern insulator without requiring net magnetic field. The Haldane model has been experimentally simulated via ultracold atom platform [58, 59]. The Hamiltonian of Haldane model reads as

H^0=−∑⟨r,r′⟩t1c^r†c^r′−∑⟨⟨r,r′⟩⟩t2eiΦr,r′c^r†c^r′+H.c.+Δ02∑rξrc^r†c^r(48)\hat{H}_0=-\sum_{\langle r,r'\rangle}t_1\hat{c}_r^\dagger\hat{c}_{r'}-\sum_{\langle\langle r,r'\rangle\rangle}t_2e^{i\Phi_{r,r'}}\hat{c}_r^\dagger\hat{c}_{r'}+\mathrm{H.c.}+\frac{\Delta_0}{2}\sum_r\xi_r\hat{c}_r^\dagger\hat{c}_r \tag*{(48)}

in which t1t_1 is the nearest-neighbor (NN) tunneling strength, t2t_2 is the next-nearest-neighbor (NNN) tunneling strength, and Φr,r′\Phi_{r,r'} is the gauge field addressed to the NNN tunneling, as demonstrated in Figure 2, and ξr=±1\xi_r=\pm1 is the energy bias for the two sublattices. Here ⟨r,r′⟩\langle r,r'\rangle and ⟨⟨r,r′⟩⟩\langle\langle r,r'\rangle\rangle respectively represent the NN sites and NNN sites.

Schematic illustration of tunneling relations within a single hexagon in the Haldane model and typical band structure

Figure 2. (a) Schematic illustration of the tunneling relations within a single hexagon in Haldane model and (b) the typical band structure. The lattice constant is set to unity for convenience.

The phase diagram of the system under PBC is computed through the two real-space formulae, as shown in Figure 3. The non-commutative method produces relatively smooth results near the phase boundary, while the Bott index method presents a sharp boundary due to the quantized nature. The phase boundary of the non-commutative method is expected to be sharp in the thermodynamic limit. Both two real-space formulae can well capture the topological property of the Chern insulator.

Phase diagram of the Haldane model

Figure 3. Phase diagram of the Haldane model. (a) and (b) are respectively computed via the non-commutative method [Eq. (21)] and the Bott index method [Eq. (33)]. The system’s size is chosen as Lx=Ly=11L_x=L_y=11. Other parameters: t1=1,t2=0.2t_1=1,t_2=0.2.

Flatness of the Berry curvature

In § 1.1 B, we have shown that the trace of non-Abelian Berry curvature defined through the TBC should tends to be a constant in the thermodynamic limit. The flat Berry curvature allows us to compute the Chern number without integration, which is important to derive the real-space Chern number. In this subsection, we shall verify this point numerically.

To compute the Berry curvature numerically through TBC, one should discretize the twist angle and use appropriate numerical method. Generally, one can adopt the link variable method [60]

Tr⁡[F~(θ)]δθxδθy=ln⁡det⁡[Ψ~(θx,θy)†Ψ~(θx+δθx,θy)]+ln⁡det⁡[Ψ~(θx+δθx,θy)†Ψ~(θx+δθx,θy+δθy)]+ln⁡det⁡[Ψ~(θx+δθx,θy+δθy)†Ψ~(θx,θy+δθy)]+ln⁡det⁡[Ψ~(θx,θy+δθy)†Ψ~(θx,θy)],(49)\begin{aligned} \operatorname{Tr}\left[\widetilde{F}(\theta)\right]\delta\theta_x\delta\theta_y ={}&\ln\det\left[\widetilde{\Psi}^{\dagger}_{(\theta_x,\theta_y)}\widetilde{\Psi}_{(\theta_x+\delta\theta_x,\theta_y)}\right]\\ &+\ln\det\left[\widetilde{\Psi}^{\dagger}_{(\theta_x+\delta\theta_x,\theta_y)}\widetilde{\Psi}_{(\theta_x+\delta\theta_x,\theta_y+\delta\theta_y)}\right]\\ &+\ln\det\left[\widetilde{\Psi}^{\dagger}_{(\theta_x+\delta\theta_x,\theta_y+\delta\theta_y)}\widetilde{\Psi}_{(\theta_x,\theta_y+\delta\theta_y)}\right]\\ &+\ln\det\left[\widetilde{\Psi}^{\dagger}_{(\theta_x,\theta_y+\delta\theta_y)}\widetilde{\Psi}_{(\theta_x,\theta_y)}\right], \tag*{(49)} \end{aligned}

which produces a gauge-invariant result.

Alternatively, based on Eq. (9), one can use the finite difference method to approximate the partial derivative

∂θjP~θ≈P~θ+δθj−P~θ∣δθj∣,j=x,y,(50)\partial_{\theta_j}\widetilde{P}_{\theta}\approx\frac{\widetilde{P}_{\theta+\delta\theta_j}-\widetilde{P}_{\theta}}{|\delta\theta_j|},\qquad j=x,y, \tag*{(50)}

and then we have

Tr⁡[F~(θ)]≈Tr⁡{P~θ[P~θ+δθx−P~θ∣δθx∣,P~θ+δθy−P~θ∣δθy∣]}.(51)\operatorname{Tr}\left[\widetilde{F}(\theta)\right]\approx\operatorname{Tr}\left\{\widetilde{P}_{\theta}\left[\frac{\widetilde{P}_{\theta+\delta\theta_x}-\widetilde{P}_{\theta}}{|\delta\theta_x|},\frac{\widetilde{P}_{\theta+\delta\theta_y}-\widetilde{P}_{\theta}}{|\delta\theta_y|}\right]\right\}. \tag*{(51)}

It should be noted that the projector P~θ\widetilde{P}_{\theta} is gauge-invariant, and therefore Eq. (51) is gauge-invariant as well.

We then numerically compute the Berry curvature using these two methods, as shown in Figure 4 (a-b). As expected, the Berry curvature defined through the TBC is very flat. By summing over all discretized terms, one will obtain the Chern number. It can be seen that the link variable method [Eq. (49)] strictly gives a integer Chern number, which is accurate, while the finite difference method [Eq. (51)] only produces a Chern number close to integer. The deviation in the finite difference numerical method is brought by finite grid points.

Berry curvature computed using two numerical methods and flatness under different system sizes and sublattice bias

Figure 4. Numerical computation of the Berry curvature defined through the TBC [Eq. (8)] with t2/t1=0.5t_2/t_1 = 0.5. The Chern number is obtained by summing the Berry curvature C=12π∑θTr⁡[F~(θ)]δθxδθyC = \frac{1}{2\pi} \sum_{\theta} \operatorname{Tr}[\widetilde{F}(\theta)]\delta\theta_x\delta\theta_y. We have used two different numerical methods to compute the Berry curvature: (a) the link variable method [Eq. (49)], and (b) the finite difference method [Eq. (51)] with 30×3030 \times30 grid points. (c) scaling of the flatness of Berry curvature [Eq. (52)] under different system sizes Lx=Ly=LL_x = L_y = L. (d) Flatness versus the different sublattice bias energy Δ/t1\Delta/t_1 crossing the phase boundary with t2/t1=0.2t_2/t_1 = 0.2. The topological phase transition occurs at Δ0/t1≈2\Delta_0/t_1 \approx2 here. The system’s size is chosen as Lx=Ly=11L_x = L_y = 11. Other parameters are set to t1=1t_1 = 1, Φ=π/2\Phi= \pi/2.

According to the perturbative expansion in Eq. (15), the Berry curvature (after tracing) is expected to be a constant in thermodynamic limit. Next, we examine the flatness of the Berry curvature under different system sizes. We define the following quantity to reflect the degree of flatness

f=max⁡{Tr⁡[F~(θ)]}−min⁡{Tr⁡[F~(θ)]}.(52)f = \max\left\{\operatorname{Tr}\left[\widetilde{F}(\theta)\right]\right\}-\min\left\{\operatorname{Tr}\left[\widetilde{F}(\theta)\right]\right\}. \tag*{(52)}

The numerical results are shown in Figure 4 (c). It can be observed that the flatness of the Berry curvature tends to zero when L→∞L \to\infty, which is in accordance with our argument.

We then further investigate the flatness of the TBC Berry curvature across topological phase boundary. As illustrated in Figure 4 (d), the flatness remains nearly unchanged for various parameter values except at the phase point. Near the phase boundary, the TBC Berry curvature strongly fluctuates. This observation aligns with our findings, where we demonstrate that the flatness of the Berry curvature solely relies on the system’s size. Near the phase boundary, the spectral gap is extremely small, and the perturbative expansion w.r.t. twist angles up to the first order loses accuracy here, which explains why the flatness becomes parameter-dependent in this regime.

Real-space Chern number in finite systems

In Sec. II B, we have seen that the real-space Chern number is based on the flatness of the Berry curvature defined through TBC. Nevertheless, the flat Berry curvature is only true in the thermodynamic limit. In finite systems, the Berry curvature may be dependent on twist angle, and therefore the real-space Chern number may deviate from the correct result. To see this effect, we compute the Chern number using the real-space formulae [Eq. (21) and Eq. (33)] with different system sizes. For finite systems, we follow the idea proposed in Ref. [20, 23] to better approximate the finite difference of P~θ\widetilde{P}_{\theta} for the non-commutative Chern number. Detailed illustration of this higher-order finite difference method is presented in B. Numerical results are shown in Figure 5. It can be seen that both methods tends to produce an integer number when L→∞L \rightarrow\infty.

Chern number as a function of system size calculated through two real-space formulae

Figure 5. Chern number as a function of system size calculated through the two real-space formulae. Other parameters are set to: t1=1t_1 = 1, t2=0.5t_2 = 0.5, Δ0=0\Delta_0 = 0 and Φ=π/2\Phi= \pi/2.

It can be observed that the non-commutative method [Eq. (21)] is strongly affected by the system’s size. This is because the non-commutative method relies on the flatness of Berry curvature. In Eq. (16), we only keep up to the linear terms of 1/Lj1/L_j, and regard the Berry curvature as a constant. Higher-order terms become considerable in finite systems, and the Berry curvature is not necessarily a constant, as already shown in previous subsection. To gain more accurate results through this method, one may take into account higher terms in the expansion of the projector P~θ\widetilde{P}_{\theta}, or increase the system’s size.

On the other hand, it is surprising to observe that the Bott index method presents a very exact result C=1C = 1 even for small size. In other words, the non-flatness of the Berry curvature will not affect the result of the Bott index. In fact, the Bott index in Eq. (33) is approximately related to the following path integral

Ux†UyUxUy†≈exp⁡(∮DA~(θ)⋅dθ)(53)U_x^\dagger U_y U_x U_y^\dagger\approx\exp\left(\oint_{\mathcal{D}} \widetilde{\mathcal{A}}(\theta)\cdot\mathrm{d}\theta\right) \tag*{(53)}

where A~(θ)=(A~x(θ),A~y(θ))\widetilde{\mathcal{A}}(\theta) = (\widetilde{\mathcal{A}}_x(\theta),\widetilde{\mathcal{A}}_y(\theta)), and D:(0,0)→(0,−2π)→(2π,−2π)→(2π,0)→(0,0)\mathcal{D} : (0,0) \rightarrow(0,-2\pi) \rightarrow(2\pi,-2\pi) \rightarrow(2\pi,0) \rightarrow(0,0) is a closed path in (θx,θy)(\theta_x,\theta_y) plane. With Stokes theorem, this integral is equivalent to the sum of the Berry curvature in Eq. (49), as depicted in Figure 6. This means the Berry curvature is not necessarily flat when applying the Bott index method to calculate the Chern number in real space. In addition, one may recall that the Stokes theorem fails when Chern number is non-zero in topological band theory. This is because one can not find a globally smooth gauge for the Bloch state on the Brillouin manifold if the Chern number is non-zero. Here, we instead consider gapped states Ψθ\mathbf{\Psi}_{\theta} parameterized by twist angle and use the non-Abelian form of the Berry connection. There is no obstruction in applying the Stokes theorem, which explains why the Bott index produces an exactly quantized Chern number.

Diagram showing equivalence between a closed path and small plaquettes on the $(\theta_x,\theta_y)$ plane

Figure 6. The integral of Berry connection (indicated by blue arrows) defined through the TBC along the closed path D\mathcal{D} is equivalent to the sum of integral in each small plaquette on the (θx,θy)(\theta_x,\theta_y) plane.

Chern number in the presence of disorder

As mentioned in previous sections, the real-space formula allows us to compute the Chern number without requiring the translation symmetry. To examine this fact, we consider on-site energy disorder on each site’s in Haldane model

H^=H^0+W∑rϵrn^r(54)\hat{H}=\hat{H}_0+W\sum_r\epsilon_r\hat{n}_r \tag*{(54)}

where WW is the strength of the disorder, and ϵr∈[−0.5,0.5]\epsilon_r \in[-0.5,0.5] is a uniformly distributed random number. We numerically diagonalize the Hamiltonian under PBC and focus on the lowest band. The parameter is chosen to be t1=1t_1=1, t2=0.5t_2=0.5, Φ=π/2\Phi=\pi/2, corresponding to the C=1C=1 topological phase in the clean system. Then, we increase the strength of the disorder, and the topological band theory fails in this situation. By employing the real-space formula (the non-commutative method [Eq. (21)] and the Bott index method [Eq. (33)]), we compute the Chern number as a function of disorder strength WW, see results in Figure 7. The result is averaged over 500 random realizations. It can be seen that the Chern number remains C=1C=1 for moderate disorder and then there appears a disorder-induced topological transition for stronger disorder. One can find that results from the two real-space Chern number are well consistent, despite some differences due to the finite-size effect. As expected, the Bott index method will exactly produce an integer for all realizations, while the non-commutative method will give a non-quantized result near the phase transition point. These two methods are expected to be equivalent in the thermodynamic limit.

Numerical computations of the real-space Chern number

Figure 7. Numerical computations of the real-space Chern number as a function of disorder strength WW on a Lx×Ly=30×30L_x \times L_y=30\times30 system. Grey points are from each random realization, while circles are the averaged results. We have performed computations over 500 random realizations. The errorbar represents the standard deviations of Chern numbers among these random realizations. Other parameters are set to t1=1t_1=1, t2=0.5t_2=0.5, Δ0=0\Delta_0=0, Φ=π/2\Phi=\pi/2.

Open boundary condition

Previously, we only focus on the PBC or TBC. In both cases, the system is in a torus geometry and there is no boundary state. As for the open boundary condition (OBC), there may exist gapless boundary states, bringing problems to numerical computations. Since these boundary states only localize near the boundary, one can change the definition of the positions operators such that the boundary area is excluded: r^j=\hat{r}_j=

∑δL<rj<Lj−δLrjn^r,j=x,y.\sum_{\delta_L<r_j<L_j-\delta_L} r_j\hat{n}_r,\quad j=x,y.

This trick can circumvent the influence of the boundary state. In numerical computation, it is sufficient to set δL\delta_L as a length of few unit cells. The corresponding system’s size shrinks to Lj=Lj−2δL,j=x,yL_j=L_j-2\delta_L,j=x,y. We then compute the Chern number under OBC for different parameters using the reduced position operators, see results in Figure 8. The Fermi energy is set to EF=0E_F=0 here, and we target those states below it. It can be clearly seen that the topological phase transition occurs at about Δ0/t1≈2\Delta_0/t_1\approx2, where the in-gap boundary states become absent. The two real-space formulae can still correctly capture the Chern number.

Energy spectrum and Chern number as a function of sublattice bias

Figure 8. (a) Enlarged energy spectrum near EF=0E_F=0 as a function of sublattice bias Δ0\Delta_0 under OBC. (b) Chern number as a function of sublattice bias Δ0\Delta_0 under OBC. We set δL=3\delta_L=3 to truncate the range of the position operator. System’s size is chosen to: Lx=Ly=30L_x=L_y=30. Other parameters: t1=1,t2=0.2,Φ=π/2t_1=1,t_2=0.2,\Phi=\pi/2.

The Kane-Mele model

Now, we briefly discuss the application of the real-space formula of Chern number derived from the generalized TBC for the QSH insulator. One of the famous model realizing the QSH effect is proposed by Kane and Mele [49, 51]. The tight-binding Hamiltonian of Kane-Mele model reads as

H^=−t1∑⟨ij⟩c^i†c^j+iλSO∑⟨⟨ij⟩⟩νijc^i†σzc^j+iλR∑⟨ij⟩c^i†(σ×e^ij)zc^j+Δ02∑iξic^i†c^i,(55)\hat{H}=-t_1\sum_{\langle ij\rangle}\hat{c}_i^\dagger\hat{c}_j+i\lambda_{\mathrm{SO}}\sum_{\langle\langle ij\rangle\rangle}\nu_{ij}\hat{c}_i^\dagger\sigma_z\hat{c}_j+i\lambda_R\sum_{\langle ij\rangle}\hat{c}_i^\dagger(\boldsymbol{\sigma}\times\hat{e}_{ij})_z\hat{c}_j+\frac{\Delta_0}{2}\sum_i\xi_i\hat{c}_i^\dagger\hat{c}_i, \tag*{(55)}

in which c^j=(c^↑,j,c^↓,j)\hat{c}_j=(\hat{c}_{\uparrow,j},\hat{c}_{\downarrow,j}), λSO\lambda_{\mathrm{SO}} is the intrinsic spin-orbit (SO) coupling strength, and λR\lambda_R is the Rashba SO coupling strength. νij=+1(−1)\nu_{ij}=+1(-1) if the particle tunnels in

clockwise (anticlockwise) direction, e^ij\hat{e}_{ij} is the normalized vector pointing from ii to jj, and Δ0\Delta_0 is the on-site potential bias. If λR=0\lambda_R=0, this model simply consists of two copies of Haldane model with opposite NNN tunneling phase Φ=±π/2\Phi=\pm\pi/2, and szs_z is conserved. In this case, one can directly compute the Chern number in each spin sector to attain the spin Chern number. For λR≠0\lambda_R\ne0, the two subspaces are coupled and it becomes a obstacle to compute the spin Chern number in a conventional momentum-space method.

We numerically compute the spin Chern number and the Chern number matrix using the non-commutative method and the Bott index method derived in § 1.1 E. As shown in Figure 9, the spin Chern number and the Chern number matrix are both capable to produce a correct spin Chern number for the Kane-Mele model. The non-commutative and the Bott index formulae derived from the generalized TBC are also confirmed to be valid.

Figure 9. (a) Spin Chern number and (b) elements of spin Chern number matrix calculated through real-space formulae. The results of non-commutative method and the Bott index introduced in § 1.1 E are respectively marked by red and blue colors. System size is chosen to: Lx=Ly=40L_x=L_y=40. Other parameters: t=1t=1, λSO=0.06\lambda_{\mathrm{SO}}=0.06, λR=0.015\lambda_{\mathrm{R}}=0.015.

Conclusions and discussions

In summary, in the thermodynamic limit, we have shown that the non-commutative Chern number and the Bott index form of Chern number can be both derived from the Chern number defined via TBC. Thus, these two real-space formulae and the TBC formula for the Chern number are equivalent. Our derivation is based on the perturbative nature of the twist angles under the uniform gauge of TBC. The key point is to expand the eigenstate and operator to the linear terms of 1/Lx1/L_x and 1/Ly1/L_y. The trace of non-Abelian Berry curvature is shown to be a constant in the thermodynamic limit. Hence, the integration w.r.t. twist angles in the TBC Chern number will be reduced, providing feasibility to deriving the real-space Chern number. We further derive the real-space formula of spin Chern number through the generalized TBC. For different generalized TBCs in QSH insulator, one may obtain different real-space formulae. Later, the flatness of the trace of Berry curvature defined via TBC is later verified numerically. The non-flatness effect to the real-space formula of Chern number is discussed. We also show that the real-space Chern number can be applied to disordered system. We would like to stress that using the formula of topological invariants defined via TBC to derive the real-space topological invariant may be potentially generalized to other topological systems, such as higher-order topological insulators [18, 61, 62, 63] and periodically-driven systems in 2D [64]. This will benefit the search of real-space formula of topological invariants in unconventional topological insulators.

Throughout this paper, we restrict ourself to the single-particle system. In fact, the TBC method can be well applied to gapped many-body eigenstates. Since the real-space formulae of Chern number can be derived from the TBC Chern number, we expect the real-space Chern number can be extended to multi-particle systems. However, it seems that the Bott index can not work for unique many-body ground state ∣ΨGS⟩|\Psi_{\mathrm{GS}}\rangle because Ux,yU_{x,y} is only a U(1) number in this case. Hence, Eq. (33) will only produce a trivial result due to the multi-value nature of the exponential function in complex space. In other words, one can not numerically distinguish trivial and non-trivial topological insulators through the Bott index formula if the ground state is unique. To circumvent this problem, we recall that the Bott index can be considered as a integral of Berry connection, as mentioned in Figure 6. Hence, it is possible to implement the integral for half of the area under TBC with uniform gauge, that is the triangle region D′D': (0,0)→(0,−2π)→(2π,−2π)→(0,0)(0,0)\to(0,-2\pi)\to(2\pi,-2\pi)\to(0,0). This leads to the following expression

C~TBC≈1iπlog⁡⟨Ψ~GS(0,0)∣Ψ~GS(0,−2π)⟩×⟨Ψ~GS(0,−2π)∣Ψ~GS(2π,−2π)⟩×⟨Ψ~GS(2π,−2π)∣Ψ~GS(0,0)⟩=1iπlog⁡⟨ΨGS∣(U^2πy)−1∣ΨGS⟩⟨ΨGS∣U^2πx∣ΨGS⟩×⟨ΨGS∣(U^2πx)−1U^2πy∣ΨGS⟩(56)\widetilde{C}_{\mathrm{TBC}}\approx\frac{1}{i\pi}\log\langle\widetilde{\Psi}_{\mathrm{GS}}(0,0)|\widetilde{\Psi}_{\mathrm{GS}}(0,-2\pi)\rangle \times\langle\widetilde{\Psi}_{\mathrm{GS}}(0,-2\pi)|\widetilde{\Psi}_{\mathrm{GS}}(2\pi,-2\pi)\rangle \times\langle\widetilde{\Psi}_{\mathrm{GS}}(2\pi,-2\pi)|\widetilde{\Psi}_{\mathrm{GS}}(0,0)\rangle = \frac{1}{i\pi}\log\langle\Psi_{\mathrm{GS}}|(\widehat{U}_{2\pi}^{y})^{-1}|\Psi_{\mathrm{GS}}\rangle\langle\Psi_{\mathrm{GS}}|\widehat{U}_{2\pi}^{x}|\Psi_{\mathrm{GS}}\rangle \times\langle\Psi_{\mathrm{GS}}|(\widehat{U}_{2\pi}^{x})^{-1}\widehat{U}_{2\pi}^{y}|\Psi_{\mathrm{GS}}\rangle \tag*{(56)}

which is quite similar to the real-space marker of the Chern insulator proposed recently in Refs. [65, 66]. In addition, it should be noted that this quantity also relies on the flatness of the Berry curvature, and it is not necessarily quantized in finite systems. We only expect it to be quantized in the thermodynamic limit.

As for the non-commutative method [Eq. (21)], we can consider a more fundamental formula described in Eq. (16) for better approximation, which yields

C~TBC≈−2πi⟨ΨGS∣[(∂θxP~θ)θ=0,(∂θyP~θ)θ=0]∣ΨGS⟩.(57)\widetilde{C}_{\mathrm{TBC}}\approx-2\pi i\langle\Psi_{\mathrm{GS}}|\left[\left(\partial_{\theta_x}\widetilde{P}_{\theta}\right)_{\theta=0},\left(\partial_{\theta_y}\widetilde{P}_{\theta}\right)_{\theta=0}\right]|\Psi_{\mathrm{GS}}\rangle. \tag*{(57)}

By using finite difference to approximate the partial derivative, one is also able to obtain an approximated Chern number.

We also note the half-quantized Hall effect has drawn many attentions recently [67, 68, 69]. This is related to gapless surface states, in which the formula of the Chern number [Eq. (2)] generally fails. How to capture the half-quantized nature of Hall conductance through the TBC is challenging. In such scenario, it is possible to modify the computation method of the Hall conductance. Then, following the idea in § 1.1 B, one may derive a real-space formula for this system by perturbatively expanding the system up to the first order w.r.t. twist angles. We believe this could be an intriguing topic in the future.

Acknowledgments

This work is supported by the National Key Research and Development Program of China (Grant No. 2022YFA1404104), and the National Natural Science Foundation of China (Grant No. 12025509, 12247134).

A Invariance of Chern number under different gauges

We would like to prove that the formulae of TBC Chern number under the boundary gauge and uniform gauge have the same form. According to Eq. (6), there is Ψ~θ=U^θΨθ\tilde{\Psi}_\theta=\hat{U}_\theta\Psi_\theta. Then, we find the Berry connection is transformed as

A~j(θ)=Ψ~θ†∂θjΨ~θ=Ψθ†U^θ−1∂θj(U^θΨθ)=Ψθ†∂θjΨθ+Ψθ†(U^θ−1∂θjU^θ)Ψθ=Aj(θ)+Ψθ†r^jLjΨθ,j=x,y.(58)\tilde{\mathcal{A}}_j(\theta)=\tilde{\Psi}_\theta^\dagger\partial_{\theta_j}\tilde{\Psi}_\theta =\Psi_\theta^\dagger\hat{U}_\theta^{-1}\partial_{\theta_j}\left(\hat{U}_\theta\Psi_\theta\right) =\Psi_\theta^\dagger\partial_{\theta_j}\Psi_\theta+\Psi_\theta^\dagger\left(\hat{U}_\theta^{-1}\partial_{\theta_j}\hat{U}_\theta\right)\Psi_\theta =\mathcal{A}_j(\theta)+\Psi_\theta^\dagger\frac{\hat{r}_j}{L_j}\Psi_\theta,\quad j=x,y. \tag*{(58)}

Hence, it can be found that the Berry curvature satisfies

F~(θ)=F(θ)+[∂θyrˉx(θ)−∂θxrˉy(θ)],(59)\tilde{\mathcal{F}}(\theta)=\mathcal{F}(\theta)+\left[\partial_{\theta_y}\bar{r}_x(\theta)-\partial_{\theta_x}\bar{r}_y(\theta)\right], \tag*{(59)}

where we denote rˉj(θ)≡Ψθ†r^jLjΨθ\bar{r}_j(\theta)\equiv\Psi_\theta^\dagger\frac{\hat{r}_j}{L_j}\Psi_\theta for simplicity. It can be seen that generally F~(θ)≠F(θ)\tilde{\mathcal{F}}(\theta)\ne\mathcal{F}(\theta). Note that, when the twist angle changes a flux quanta, which is 2π2\pi here, the system returns to the origin. Meanwhile, it can be immediately checked that rˉj(θ)\bar{r}_j(\theta) is a single-value and gauge-invariant periodic function of twist angle. Hence, there should be

Tr⁡[rˉj(θ+2πej′)]=Tr⁡[rˉj(θ)],j,j′=x,y,(60)\operatorname{Tr}\left[\bar{r}_j(\theta+2\pi e_{j'})\right]=\operatorname{Tr}\left[\bar{r}_j(\theta)\right],\quad j,j'=x,y, \tag*{(60)}

where ej′e_{j'} represents the unit vector along j′=x,yj'=x,y direction. After integrating over twist angles, there is

∫02π∫02πdθxdθyTr⁡[∂θj′rˉj(θ)]=0,j,j′=x,y.(61)\int_0^{2\pi}\int_0^{2\pi}d\theta_xd\theta_y\operatorname{Tr}\left[\partial_{\theta_{j'}}\bar{r}_j(\theta)\right]=0,\quad j,j'=x,y. \tag*{(61)}

Thus, we can conclude that the Chern number defined through TBC is the same under the boundary gauge and the uniform gauge. In other words, the large gauge transformation in Eq. (6) will not affect the Chern number.

B Approximation for finite difference

To better approximate the partial derivative of the projector matrix numerically, one can use a higher-order finite difference method for this first-order derivative introduced in [20, 23]. To see how it works, we write down series expansion of the projector with respect to θ/L\theta/L:

Pθ/L=P0+∑k=1∞(θL)k1k![∂k∂(θ/L)kPθ/L]θ=0.(62)P_{\theta/L}=P_0+\sum_{k=1}^{\infty}\left(\frac{\theta}{L}\right)^k\frac{1}{k!}\left[\frac{\partial^k}{\partial(\theta/L)^k}P_{\theta/L}\right]_{\theta=0}. \tag*{(62)}

As a demonstration, we assume that the projector is only parameterized by θ\theta. The result can be readily extended to the 2D system discussed in the main text. By construction, there is

Pθ/L−P−θ/L2=∑m=1∞(θL)2m−11(2m−1)!×[∂2m−1∂(θ/L)2m−1Pθ/L]θ=0.(63)\frac{P_{\theta/L}-P_{-\theta/L}}{2}=\sum_{m=1}^{\infty}\left(\frac{\theta}{L}\right)^{2m-1}\frac{1}{(2m-1)!}\times\left[\frac{\partial^{2m-1}}{\partial(\theta/L)^{2m-1}}P_{\theta/L}\right]_{\theta=0}. \tag*{(63)}

Thus, by choosing a set of parameters {cn}\{c_n\} to eliminate higher-order terms, the first-order derivative can be written as

θL[∂∂(θ/L)Pθ/L]θ=0≈∑n∈Z+cn2(Pnθ/L−P−nθ/L)(64)\frac{\theta}{L}\left[\frac{\partial}{\partial(\theta/L)}P_{\theta/L}\right]_{\theta=0}\approx\sum_{n\in\mathbb{Z}_+}\frac{c_n}{2}\left(P_{n\theta/L}-P_{-n\theta/L}\right) \tag*{(64)}

The coefficient {cn}\{c_n\} can be solved through a set of linear equations according to the series expansion. By truncating the expansion to QQ terms with Q<L/2Q<L/2 (to ensure the convergence of the expansion), it can be written in a matrix form

(c1  c2  ⋯  cQ)(Pθ/L−P−θ/LP2θ/L−P−2θ/L⋮PQθ/L−P−Qθ/L)×12=θL[∂∂(θ/L)Pθ/L]θ=0.(65)\left(c_1\ \ c_2\ \ \cdots\ \ c_Q\right) \begin{pmatrix} P_{\theta/L}-P_{-\theta/L}\\ P_{2\theta/L}-P_{-2\theta/L}\\ \vdots\\ P_{Q\theta/L}-P_{-Q\theta/L} \end{pmatrix} \times\frac{1}{2} = \frac{\theta}{L}\left[\frac{\partial}{\partial(\theta/L)}P_{\theta/L}\right]_{\theta=0}. \tag*{(65)}

Then, we can convert this problem to solving a set of linear equations, which can be expressed as

(12⋯Q1323⋯Q3⋮⋮⋱⋮12Q−122Q−1⋯Q2Q−1)(c1c2⋮cQ)=(10⋮0)(66)\begin{pmatrix} 1 & 2 & \cdots& Q\\ 1^3 & 2^3 & \cdots& Q^3\\ \vdots& \vdots& \ddots& \vdots\\ 1^{2Q-1} & 2^{2Q-1} & \cdots& Q^{2Q-1} \end{pmatrix} \begin{pmatrix} c_1\\ c_2\\ \vdots\\ c_Q \end{pmatrix} = \begin{pmatrix} 1\\ 0\\ \vdots\\ 0 \tag*{(66)} \end{pmatrix}

Hence, by taking the matrix inverse, {cn}\{c_n\} can be solved numerically through:

\begin{pmatrix} c_1\\ c_2\\ \vdots\\ c_Q \end{pmatrix} = \begin{pmatrix} 1 & 2 & \cdots& Q\\ 1^3 & 2^3 & \cdots& Q^3\\ \vdots& \vdots& \ddots& \vdots\\ 1^{2Q-1} & 2^{2Q-1} & \cdots& Q^{2Q-1} \end{pmatrix}^{-1} \begin{pmatrix} 1\\ 0\\ \vdots\\ 0 \end{pmatrix}. \tag*{(67)}\end{pmatrix}

Next, by using the fact that

P2nπ/L=(U^2π)nP0(U^−2π)n(68)P_{2n\pi/L}=\left(\hat{U}_{2\pi}\right)^nP_0\left(\hat{U}_{-2\pi}\right)^n \tag*{(68)}

we can let θ=2π\theta=2\pi and then:

2πL[∂∂(θ/L)Pθ/L]θ=0≈∑n=1Qcn2[(U^2π)nP0(U^−2π)n−(U^−2π)nP0(U^2π)n],(69)\frac{2\pi}{L}\left[\frac{\partial}{\partial(\theta/L)}P_{\theta/L}\right]_{\theta=0} \approx\sum_{n=1}^{Q}\frac{c_n}{2}\left[\left(\hat{U}_{2\pi}\right)^nP_0\left(\hat{U}_{-2\pi}\right)^n-\left(\hat{U}_{-2\pi}\right)^nP_0\left(\hat{U}_{2\pi}\right)^n\right], \tag*{(69)}

which can be evaluated numerically and produces a good approximation for the partial derivative.

C Quasi-unitarity of the matrices Ux,yU_{x,y}

Below, we show that the matrix Uj=Ψ0†U^jΨ0U_j=\Psi_0^\dagger\hat{U}_j\Psi_0, j=x,yj=x,y mentioned in the main text is a unitary matrix in the thermodynamic limit Lj→∞L_j \to\infty through the TBC. Firstly, note that

Uj=Ψ0†U^2πjΨ0,Uj†=Ψ0†(U^2πj)−1Ψ0.(70)U_j=\Psi_0^\dagger\hat{U}_{2\pi}^j\Psi_0,\qquad U_j^\dagger=\Psi_0^\dagger(\hat{U}_{2\pi}^j)^{-1}\Psi_0. \tag*{(70)}

The vector is normalized: Ψ0†Ψ0=IN\Psi_0^\dagger\Psi_0=I_N, and N\mathcal{N} is the number of targeted states. Meanwhile, we have Ψ0Ψ0†=∑μ∈target∣ψμ(0)⟩⟨ψμ(0)∣=1\Psi_0\Psi_0^\dagger=\sum_{\mu\in\mathrm{target}}|\psi_\mu(0)\rangle\langle\psi_\mu(0)|=1 in the subspace spanned by targeted states without the TBC. There is

U2πj(U2πj)†=Ψ0†U^2πjΨ0Ψ0†(U^2πj)−1Ψ0=Ψ0†U^2πj(∑μ∣ψμ(0)⟩⟨ψμ(0)∣)(U^2πj)−1Ψ0=Ψ0†(∑μ∣ψ~μ(2πej)⟩⟨ψ~μ(2πej)∣)Ψ0.(71)\begin{aligned} U_{2\pi}^j(U_{2\pi}^j)^\dagger=\Psi_0^\dagger\hat{U}_{2\pi}^j\Psi_0\Psi_0^\dagger(\hat{U}_{2\pi}^j)^{-1}\Psi_0 \\ &=\Psi_0^\dagger\hat{U}_{2\pi}^j\left(\sum_\mu|\psi_\mu(0)\rangle\langle\psi_\mu(0)|\right)(\hat{U}_{2\pi}^j)^{-1}\Psi_0 \\ &=\Psi_0^\dagger\left(\sum_\mu|\widetilde{\psi}_\mu(2\pi\mathbf{e}_j)\rangle\langle\widetilde{\psi}_\mu(2\pi\mathbf{e}_j)|\right)\Psi_0. \tag*{(71)} \end{aligned}

Approximately, there is

∑μ∣ψμ(θ)⟩⟨ψμ(θ)∣=∑μ∣ψμ⟩⟨ψμ∣+O(1L)=1+O(1L).(72)\begin{aligned} \sum_\mu|\psi_\mu(\theta)\rangle\langle\psi_\mu(\theta)|=\sum_\mu|\psi_\mu\rangle\langle\psi_\mu|+O\left(\frac{1}{L}\right) \\ &=1+O\left(\frac{1}{L}\right). \tag*{(72)} \end{aligned}

which implies that ∑μ∈target∣ψμ(2π)⟩⟨ψμ(2π)∣\sum_{\mu\in\mathrm{target}}|\psi_\mu(2\pi)\rangle\langle\psi_\mu(2\pi)| is close to the identity matrix in this subspace. Hence, we find MM†=IN\mathcal{M}\mathcal{M}^\dagger=I_N in the thermodynamic limit. One can also prove that M†M=IN\mathcal{M}^\dagger\mathcal{M}=I_N using the same analysis. In summary, we have shown that the matrix M\mathcal{M} is approximately a unitary matrix in the thermodynamic limit. Similar conclusions can be found in Refs. [26, 27].

Derivations of Eq. (27)

Here we show that how to derive the partial derivative of the Berry phase formulated by projected position operator. The projected position operator reads as

P~(θx,0)U^2πyP~(θx,0)=∑μ,μ′∈target[Uy(θx)]μ,μ′∣ψμ(θx)⟩⟨ψμ′(θx)∣.(73)\widetilde{P}_{(\theta_x,0)}\hat{U}_{2\pi}^y\widetilde{P}_{(\theta_x,0)} =\sum_{\mu,\mu'\in\mathrm{target}}[U_y(\theta_x)]_{\mu,\mu'}|\psi_\mu(\theta_x)\rangle\langle\psi_{\mu'}(\theta_x)|. \tag*{(73)}

One can diagonalize the matrix Uy(θx)U_y(\theta_x) to find

P~(θx,0)U^2πyP~(θx,0)=∑neiΦn(θx)∣Φn(θx)⟩⟨Φn(θx)∣,(74)\widetilde{P}_{(\theta_x,0)}\hat{U}_{2\pi}^y\widetilde{P}_{(\theta_x,0)} =\sum_n e^{i\Phi_n(\theta_x)}|\Phi_n(\theta_x)\rangle\langle\Phi_n(\theta_x)|, \tag*{(74)}

where we assumed Uy(θx)U_y(\theta_x) is unitary and hence its eigenvalues are U(1) numbers. Then, the logarithm of the projected position operator can be equivalently expressed as

log⁡[P~(θx,0)U^2πyP~(θx,0)]=i∑nΦn(θx)∣Φn(θx)⟩⟨Φn(θx)∣.(75)\log\left[\widetilde{P}_{(\theta_x,0)}\hat{U}_{2\pi}^y\widetilde{P}_{(\theta_x,0)}\right] =i\sum_n\Phi_n(\theta_x)|\Phi_n(\theta_x)\rangle\langle\Phi_n(\theta_x)|. \tag*{(75)}

Taking the partial derivative yields

∂∂θxlog⁡[P~(θx,0)U^2πxP~(θx,0)]=i∑n∂∂θxΦn(θx)∣Φn(θx)⟩⟨Φn(θx)∣+i∑nΦn(θx)∂∂θx[∣Φn(θx)⟩⟨Φn(θx)∣].(76)\frac{\partial}{\partial\theta_x}\log\left[\widetilde{P}_{(\theta_x,0)}\hat{U}_{2\pi}^x\widetilde{P}_{(\theta_x,0)}\right] =i\sum_n\frac{\partial}{\partial\theta_x}\Phi_n(\theta_x)|\Phi_n(\theta_x)\rangle\langle\Phi_n(\theta_x)| +i\sum_n\Phi_n(\theta_x)\frac{\partial}{\partial\theta_x}\left[|\Phi_n(\theta_x)\rangle\langle\Phi_n(\theta_x)|\right]. \tag*{(76)}

The second term will vanish under the trace operation, and then we have

Tr⁡{∂∂θxlog⁡[P~(0,θx)U^2πxP~(0,θx)]}=i∑n∂∂θxΦn(θx).(77)\operatorname{Tr}\left\{\frac{\partial}{\partial\theta_x}\log\left[\widetilde{P}_{(0,\theta_x)}\hat{U}_{2\pi}^x\widetilde{P}_{(0,\theta_x)}\right]\right\} =i\sum_n\frac{\partial}{\partial\theta_x}\Phi_n(\theta_x). \tag*{(77)}

Note that, in band insulator, Φn(θx)\Phi_n(\theta_x) is proportional to the center of Wannier function. In the same vein, one can find that

Tr⁡{[P~(θx,0)(U^2πy)−1P~(θx,0)]∂∂θx[P~(0,θx)U^2πyP~(θx,0)]}=i∑n∂∂θxΦn(θx)=Tr⁡{∂∂θxlog⁡[P~(θx,0)U^2πyP~(θx,0)]},(78)\operatorname{Tr}\left\{\left[\widetilde{P}_{(\theta_x,0)}(\hat{U}_{2\pi}^y)^{-1}\widetilde{P}_{(\theta_x,0)}\right]\frac{\partial}{\partial\theta_x}\left[\widetilde{P}_{(0,\theta_x)}\hat{U}_{2\pi}^y\widetilde{P}_{(\theta_x,0)}\right]\right\} =i\sum_n\frac{\partial}{\partial\theta_x}\Phi_n(\theta_x) =\operatorname{Tr}\left\{\frac{\partial}{\partial\theta_x}\log\left[\widetilde{P}_{(\theta_x,0)}\hat{U}_{2\pi}^y\widetilde{P}_{(\theta_x,0)}\right]\right\}, \tag*{(78)}

which gives rise to Eq. (27).

Gauge transformation for the generalized TBC

Note that the twist operator manifests that

U^θΓc^i†(U^θΓ)−1=c^i†ei(rixθxLxΓx+riyθyLyΓy).(79)\hat{U}_\theta^\Gamma\hat{c}_i^\dagger(\hat{U}_\theta^\Gamma)^{-1} =\hat{c}_i^\dagger e^{i\left(\frac{r_i^x\theta_x}{L_x}\Gamma_x+\frac{r_i^y\theta_y}{L_y}\Gamma_y\right)}. \tag*{(79)}

Therefore, the tunneling term will be transformed to

U^θΓc^ri†tijc^rj(U^θΓ)−1=c^ri†t~ij(θ)c^rj(80)\hat{U}_\theta^\Gamma\hat{c}_{r_i}^\dagger t_{ij}\hat{c}_{r_j}(\hat{U}_\theta^\Gamma)^{-1} =\hat{c}_{r_i}^\dagger\widetilde{t}_{ij}(\theta)\hat{c}_{r_j} \tag*{(80)}

where the tunneling matrix is rotated

t~ij(θ)=ei(rixθxLxΓx+riyθyLyΓy)tije−i(rjxθxLxΓx+rjyθyLyΓy).(81)\widetilde{t}_{ij}(\theta)=e^{i\left(\frac{r_i^x\theta_x}{L_x}\Gamma_x+\frac{r_i^y\theta_y}{L_y}\Gamma_y\right)}t_{ij}e^{-i\left(\frac{r_j^x\theta_x}{L_x}\Gamma_x+\frac{r_j^y\theta_y}{L_y}\Gamma_y\right)}. \tag*{(81)}

For Γx=S\Gamma_x=S, Γy=1\Gamma_y=1, the tunneling matrix becomes uniform along yy direction and spatially modulated along xx direction if [S,tij]≠0[S,t_{ij}]\ne0. The rotation degree is in order of 1/Lx1/L_x between two adjacent sites provided the tunneling is finite-range.

References

References

  1. [1]M. Z. Hasan and C. L. Kane, Colloquium: Topological insulators, Rev. Mod. Phys. 82, 3045 (2010).
  2. [2]X.-L. Qi and S.-C. Zhang, Topological insulators and superconductors, Rev. Mod. Phys. 83, 1057 (2011).DOI
  3. [3]C. L. Kane, Topological band theory and the Z2 invariant, Contemporary Concepts of Condensed Matter Science, 6, 3 (2013).DOI
  4. [4]C.-K. Chiu, J. C. Y. Teo, A. P. Schnyder, and S. Ryu, Classification of topological quantum matter with symmetries, Rev. Mod. Phys. 88, 035005 (2016).DOI
  5. [5]D. J. Thouless, M. Kohmoto, M. P. Nightingale, and M. den Nijs, Quantized Hall Conductance in a Two-Dimensional Periodic Potential, Phys. Rev. Lett. 49, 405 (1982).DOI
  6. [6]Q. Niu and D. J. Thouless, Quantised adiabatic charge transport in the presence of substrate disorder and many-body interaction, Journal of Physics A: Mathematical and General 17, 2453 (1984).
  7. [7]R. Tao and Y.-S. Wu, Gauge invariance and fractional quantum hall effect, Phys. Rev. B 30, 1097 (1984).DOI
  8. [8]Q. Niu, D. J. Thouless, and Y.-S. Wu, Quantized Hall conductance as a topological invariant, Phys. Rev. B 31, 3372 (1985).
  9. [9]D. Xiao, M.-C. Chang, and Q. Niu, Berry phase effects on electronic properties, Rev. Mod. Phys. 82, 1959 (2010).DOI
  10. [10]Z. Gong, Y. Ashida, K. Kawabata, K. Takasan, S. Higashikawa, and M. Ueda, Topological Phases of Non-Hermitian Systems, Phys. Rev. X 8, 031079 (2018).DOI
  11. [11]L. Lin, Y. Ke, and C. Lee, Real-space representation of the winding number for a one-dimensional chiral-symmetric topological insulator, Phys. Rev. B 103, 224208 (2021).DOI
  12. [12]R. B. Laughlin, Quantized hall conductivity in two dimensions, Phys. Rev. B 23, 5632 (1981).
  13. [13]M. Oshikawa and T. Senthil, Fractionalization, topological order, and quasiparticle statistics, Phys. Rev. Lett. 96, 060601 (2006).pubmed.ncbi.nlm.nih.gov/16605977
  14. [14]Y. Huo and R. N. Bhatt, Current carrying states in the lowest landau level, Phys. Rev. Lett. 68, 1375 (1992).pubmed.ncbi.nlm.nih.gov/10046150
  15. [15]J. E. Avron and R. Seiler, Quantization of the hall conductance for general, multiparticulate schrödinger hamiltonians, Phys. Rev. Lett. 54, 259 (1985).pubmed.ncbi.nlm.nih.gov/10031462
  16. [16]Q. Niu, Theory of the quantized adiabatic particle transport, Modern Physics Letters B 5, 923 (1991).DOI
  17. [17]H. Watanabe and M. Oshikawa, Inequivalent berry phases for the bulk polarization, Phys. Rev. X 8, 021065 (2018).
  18. [18]J. F. Wienand, F. Horn, M. Aidelsburger, J. Bibo, and F. Grusdt, Thouless pumps and bulk-boundary correspondence in higher-order symmetry-protected topological phases, Phys. Rev. Lett. 128, 246602 (2022).DOI
  19. [19]L. Lin, Y. Ke, and C. Lee, Topological invariants for interacting systems: From twisted boundary conditions to center-of-mass momentum, Phys. Rev. B 107, 125161 (2023).DOI
  20. [20]E. Prodan, T. L. Hughes, and B. A. Bernevig, Entanglement Spectrum of a Disordered Topological Chern Insulator, Phys. Rev. Lett. 105, 115501 (2010).arxiv.org/abs/1005.5148
  21. [21]E. Prodan, Robustness of the spin-chern number, Phys. Rev. B 80, 125327 (2009).DOI
  22. [22]E. Prodan, Non-commutative tools for topological insulators, New Journal of Physics 12, 065003 (2010).DOI
  23. [23]E. Prodan, Disordered topological insulators: a non-commutative geometry perspective, Journal of Physics A: Mathematical and Theoretical 44, 113001 (2011).
  24. [24]E. Prodan, A computational non-commutative geometry program for disordered topological insulators, Vol. 23 (Springer, 2017).DOI
  25. [25]R. Exel and T. A. Loring, Invariants of almost commuting unitaries, Journal of Functional Analysis 95, 364 (1991).DOI
  26. [26]T. A. Loring and M. B. Hastings, Disordered topological insulators via C*-algebras, EPL (Europhysics Letters) 92, 67004 (2010).DOI
  27. [27]M. B. Hastings and T. A. Loring, Almost commuting matrices, localized Wannier functions, and the quantum Hall effect, Journal of mathematical physics 51, 015214 (2010).
  28. [28]M. B. Hastings and T. A. Loring, Topological insulators and C*-algebras: Theory and numerical practice, Annals of Physics 326, 1699 (2011), july 2011 Special Issue.DOI
  29. [29]T. A. Loring, K-theory and pseudospectra for topological insulators, Annals of Physics 356, 383 (2015).DOI
  30. [30]I. Mondragon-Shem, T. L. Hughes, J. Song, and E. Prodan, Topological Criticality in the Chiral-Symmetric AIII Class at Strong Disorder, Phys. Rev. Lett. 113, 046802 (2014).
  31. [31]J. E. Avron, R. Seiler, and B. Simon, Homotopy and quantization in condensed matter physics, Phys. Rev. Lett. 51, 51 (1983).DOI
  32. [32]M. Yamanaka, M. Oshikawa, and I. Affleck, Nonperturbative approach to luttinger’s theorem in one dimension, Phys. Rev. Lett. 79, 1110 (1997).DOI
  33. [33]H. Watanabe, Insensitivity of bulk properties to the twisted boundary condition, Phys. Rev. B 98, 155137 (2018).DOI
  34. [34]J. Bellissard, Noncommutative geometry and quantum hall effect, in Proceedings of the International Congress of Mathematicians. August 3–11, 1994 Zürich, Switzerland (Springer, 1995) pp. 1238–1246.DOI
  35. [35]E. Prodan, B. Leung, and J. Bellissard, The non-commutative nth-chern number (n ≥ 1), Journal of Physics A: Mathematical and Theoretical 46, 485202 (2013).
  36. [36]C. Bourne and E. Prodan, Non-commutative chern numbers for generic aperiodic discrete systems, Journal of Physics A: Mathematical and Theoretical 51, 235202 (2018).
  37. [37]K. Kudo, H. Watanabe, T. Kariyado, and Y. Hatsugai, Many-body chern number without integration, Phys. Rev. Lett. 122, 146601 (2019).DOI
  38. [38]T. A. Loring, A guide to the bott index and localizer index, arXiv preprint arXiv:1907.11791 10.48550/arXiv.1907.11791 (2019).arxiv.org/abs/1907.11791
  39. [39]M. A. Bandres, M. C. Rechtsman, and M. Segev, Topological Photonic Quasicrystals: Fractal Topological Spectrum and Protected Transport, Phys. Rev. X 6, 011016 (2016).DOI
  40. [40]A. Agarwala and V. B. Shenoy, Topological insulators in amorphous systems, Phys. Rev. Lett. 118, 236402 (2017).
  41. [41]H. Huang and F. Liu, Quantum spin hall effect and spin bott index in a quasicrystal lattice, Phys. Rev. Lett. 121, 126401 (2018).DOI
  42. [42]H. Huang and F. Liu, Theory of spin bott index for quantum spin hall states in nonperiodic systems, Phys. Rev. B 98, 125130 (2018).DOI
  43. [43]X. S. Wang, A. Brataas, and R. E. Troncoso, Bosonic bott index and disorder-induced topological transitions of magnons, Phys. Rev. Lett. 125, 217202 (2020).DOI
  44. [44]P. Titum, N. H. Lindner, M. C. Rechtsman, and G. Refael, Disorder-induced floquet topological insulators, Phys. Rev. Lett. 114, 056801 (2015).DOI
  45. [45]D. Toniolo, On the bott index of unitary matrices on a finite torus, Letters in Mathematical Physics 112, 126 (2022).DOI
  46. [46]H. Dehghani, Z.-P. Cian, M. Hafezi, and M. Barkeshli, Extraction of the many-body chern number from a single wave function, Phys. Rev. B 103, 075102 (2021).DOI
  47. [47]Z.-P. Cian, H. Dehghani, A. Elben, B. Vermersch, G. Zhu, M. Barkeshli, P. Zoller, and M. Hafezi, Many-body chern number from statistical correlations of randomized measurements, Phys. Rev. Lett. 126, 050501 (2021).DOI
  48. [48]B. Kang, W. Lee, and G. Y. Cho, Many-body invariants for chern and chiral hinge insulators, Phys. Rev. Lett. 126, 016402 (2021).DOI
  49. [49]C. L. Kane and E. J. Mele, Z₂ topological order and the quantum spin hall effect, Phys. Rev. Lett. 95, 146802 (2005).pubmed.ncbi.nlm.nih.gov/16241681
  50. [50]B. A. Bernevig, T. L. Hughes, and S.-C. Zhang, Quantum spin Hall effect and topological phase transition in HgTe quantum wells, Science 314, 1757 (2006).
  51. [51]C. L. Kane and E. J. Mele, Quantum Spin Hall Effect in Graphene, Phys. Rev. Lett. 95, 226801 (2005).pubmed.ncbi.nlm.nih.gov/16384250
  52. [52]X.-L. Qi, Y.-S. Wu, and S.-C. Zhang, General theorem relating the bulk topological number to edge states in two-dimensional insulators, Phys. Rev. B 74, 045125 (2006).DOI
  53. [53]L. Sheng, D. N. Sheng, C. S. Ting, and F. D. M. Haldane, Nondissipative Spin Hall Effect via Quantized Edge Transport, Phys. Rev. Lett. 95, 136602 (2005).DOI
  54. [54]D. N. Sheng, Z. Y. Weng, L. Sheng, and F. D. M. Haldane, Quantum spin-hall effect and topologically invariant chern numbers, Phys. Rev. Lett. 97, 036808 (2006).DOI
  55. [55]D. N. Sheng, L. Balents, and Z. Wang, Phase diagram for quantum hall bilayers at ν = 1, Phys. Rev. Lett. 91, 116802 (2003).pubmed.ncbi.nlm.nih.gov/14525450
  56. [56]T.-S. Zeng and W. Zhu, Chern-number matrix of the non-abelian spin-singlet fractional quantum hall effect, Phys. Rev. B 105, 125128 (2022).DOI
  57. [57]F. D. M. Haldane, Model for a Quantum Hall Effect without Landau Levels: Condensed-Matter Realization of the ”Parity Anomaly”, Phys. Rev. Lett. 61, 2015 (1988).pubmed.ncbi.nlm.nih.gov/10038961
  58. [58]G. Jotzu, M. Messer, R. Desbuquois, M. Lebrat, T. Uehlinger, D. Greif, and T. Esslinger, Experimental realization of the topological haldane model with ultracold fermions, Nature 515, 237 (2014).DOI
  59. [59]N. Fläschner, D. Vogel, M. Tarnowski, B. Rem, D.-S. Lühmann, M. Heyl, J. Budich, L. Mathey, K. Sengstock, and C. Weitenberg, Observation of dynamical vortices after quenches in a system with topology, Nature Physics 14, 265 (2018).DOI
  60. [60]T. Fukui, Y. Hatsugai, and H. Suzuki, Chern numbers in discretized brillouin zone: Efficient method of computing (spin) hall conductances, Journal of the Physical Society of Japan 74, 1674 (2005), https://doi.org/10.1143/JPSJ.74.1674.
  61. [61]H. Araki, T. Mizoguchi, and Y. Hatsugai, F_Q berry phase for higher-order symmetry-protected topological phases, Phys. Rev. Res. 2, 012009 (2020).DOI
  62. [62]T. Mizoguchi, H. Araki, and Y. Hatsugai, Higher-order topological phase in a honeycomb-lattice model with anti-kekulé distortion, Journal of the Physical Society of Japan 88, 104703 (2019), https://doi.org/10.7566/JPSJ.88.104703.
  63. [63]W. A. Benalcazar and A. Cerjan, Chiral-symmetric higher-order topological phases of matter, Phys. Rev. Lett. 128, 127601 (2022).DOI
  64. [64]M. S. Rudner, N. H. Lindner, E. Berg, and M. Levin, Anomalous edge states and the bulk-edge correspondence for periodically driven two-dimensional systems, Phys. Rev. X 3, 031005 (2013).arxiv.org/abs/1212.3324
  65. [65]I. Gilardoni, F. Becca, A. Marrazzo, and A. Parola, Real-space many-body marker for correlated F₂ topological insulators, Phys. Rev. B 106, L161106 (2022).DOI
  66. [66]Y.-C. Tzeng, P.-Y. Chang, and M.-F. Yang, Interaction-induced metal to topological insulator transition, Phys. Rev. B 107, 155106 (2023).DOI
  67. [67]J.-Y. Zou, B. Fu, H.-W. Wang, Z.-A. Hu, and S.-Q. Shen, Half-quantized hall effect and power law decay of edge-current distribution, Phys. Rev. B 105, L201106 (2022).DOI
  68. [68]Z.-A. Hu, H.-W. Wang, B. Fu, J.-Y. Zou, and S.-Q. Shen, Signature of parity anomaly in the measurement of optical hall conductivity in quantum anomalous hall systems, Phys. Rev. B 106, 035149 (2022).DOI
  69. [69]H.-W. Wang, B. Fu, J.-Y. Zou, Z.-A. Hu, and S.-Q. Shen, Fractional electromagnetic response in a three-dimensional chiral anomalous semimetal, Phys. Rev. B 106, 045111 (2022).DOI

Paper details

vlnc:W_zt4kbkdp
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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