Local topological markers for Chern insulators in ribbon geometry Maks Repše, Tomaž Rejec, Jernej Mravlje Abstract Local topological markers are used to characterize Chern insulators in the presence of spatial inhomogeneities, such as boundaries and disorder. In this paper, we study the local Chern marker in systems with partial translational symmetry. We express the local Chern marker in the hybrid position-momentum basis for both open and periodic boundary conditions. We calculate the local Chern marker for a Haldane model ribbon. We show that the behavior at the two boundaries is qualitatively different from fully open geometries. We further compare the local Chern marker with the local Středa marker and show agreement in the bulk and small deviations at the boundaries that diminish with increasing system size. The correspondence between the two markers remains good if disorder is introduced, provided its magnitude remains below large values that cause substantial change of the Chern number due to Anderson physics. Finally, by exploiting the numerical efficiency due to partial translational symmetry, we study equilibrium critical behavior and the Kibble-Zurek mechanism in a weakly disordered Qi–Wu–Zhang Chern insulator. We extract relevant scaling exponents from the local Chern marker configuration and show that they converge to the analytically predicted values with increasing system size. I Introduction Local topological markers have been proven to be a useful tool for characterizing inhomogeneous topological systems. They were originally introduced in Chern insulators and used to recover the bulk Chern number C in systems without translational symmetry, such as disordered Prodan et al. 2010 and finite Bianco and Resta 2011 systems. The local Chern marker (LCM), introduced in Ref. Bianco and Resta 2011 , is, in a sense, a spatially resolved density of C . Since its introduction, LCM has been connected to several observable quantities, such as orbital magnetization Bianco and Resta 2013 , anomalous Hall conductivity Marrazzo and Resta 2017 ; d’Ornellas et al. 2022 and circular dichroism Tran et al. 2017 . The wide adoption of LCM and related quantities has primarily focused on systems that feature no translational symmetry. That is, systems within open boundary conditions in all directions (fully-OBC) and/or disorder that breaks translational symmetry even in systems with periodic boundary conditions (PBC). In contrast, systems with partial translational symmetry have received far less attention. Examples of such systems include ribbons (2-dimensional systems that are infinite in one and finite in the other direction) and systems with inhomogeneities or disorder oriented in stripes. Moreover, partial translational symmetry is a natural description of interfaces Irsigler et al. 2019 and domain walls between regions with different C , bulk systems in a homogeneous external magnetic field in the Landau gauge and periodically driven Floquet Chern insulators Privitera and Santoro 2016 . Here, we build on the work of Drigo and Resta Drigo and Resta 2020 and use a formula for calculating LCM in systems with partial translational symmetry using the position-momentum (hybrid) basis. In Sec. II we review the derivation of the formula and present an approach for its numerical evaluation within both periodic and open boundary conditions. In Sec. III we explore the behavior of LCM in ribbon geometry and compare it to a different topological marker, the local Středa marker Markov et al. 2024 , motivated by the experimentally measurable response of the system to an external uniform magnetic field. We show data that suggest that for moderately strongly disordered systems the two markers are equal in the thermodynamic limit. In Sec. IV we demonstrate the utility of partial translational symmetry to study slow quenches over topological phase transitions. We accurately determine how correlation length scales with the quench duration in a model Chern insulator. In Sec. V we present our conclusions. Throughout the article, we set the reduced Planck constant \hbar and the lattice constant a to 1. II Local topological markers in position-momentum basis II.1 Local Chern marker We assume spinless and non-interacting electrons and start with the expression for the LCM in the position basis Bianco and Resta 2011 ; Prodan et al. 2010 c(\mathbf{r})=2\pi i\sum_{\alpha}\expectationvalue{P\comm{-i\comm{\hat x}{P}}{-i\comm{\hat y}{P}}}{\vb r,\alpha}, (1) where \mathbf{r}=(x,y) is the lattice vector and \alpha denotes the orbital index. We denote the number of unit cells in x(y) direction as N_{x(y)} . The Fermi projector P=\sum_{E_{\psi}<\mu}\ket{\psi}\bra{\psi} projects onto the occupied subspace that is spanned by single-electron eigenstates \ket{\psi} of the system’s Hamiltonian with energies E_{\psi} below the Fermi level \mu . Assuming translational invariance in the y direction, the Fermi projector can be decomposed into N_{y} independent blocks, indexed by momenta k_{y} from the Brillouin zone P=\bigoplus_{k_{y}}P(k_{y}). (2) The first commutator in Eq. ( 1 ) decomposes into a direct sum of commutators -i\commutator{\hat x}{P}\to\bigoplus_{k_{y}}-i\commutator{\hat x}{P(k_y)} and the second commutator becomes -i\commutator{\hat y}{P}\to\bigoplus_{k_{y}}\partial_{k_{y}}P(k_{y}) Prodan 2011 . This yields the expression for the LCM in position-momentum basis Drigo and Resta 2020 \begin{split}c(x)=\frac{2\pi i}{N_{y}}\sum_{k_{y}}\bigg\{\sum_{\alpha}\bra{x,\alpha}&P(k_{y})\\ \big[-i\commutator{\hat x}{P(k_y)}\,,\,&\partial_{k_{y}}P(k_{y})\big]\ket{x,\alpha}\bigg\}.\end{split} (3) Since P(k_{y}) is periodic in k_{y} , its derivative can be calculated numerically using the spectral derivative Širca and Horvat 2018 . We use the Fourier transform \mathcal{F} to determine the series of Fourier components {\cal P}_{n} of the projector. The k_{y} derivative of P(k_{y}) can then be calculated as the inverse Fourier transform of the series \partial_{k_{y}}P(k_{y})\to\mathcal{F}^{-1}\{in{\cal P}_{n}\big\} (4) If N_{y} is even, the series runs from n=-N_{y}/2 to N_{y}/2-1 1 1 1 In this case, the component n=-N_{y}/2 should be set to 0. See Ref. Širca and Horvat 2018 for details . If N_{y} is odd, the series runs between -(N_{y}-1)/2 and (N_{y}-1)/2 . The calculation of the commutator \commutator{\hat x}{P(k_y)} depends on the boundary condition in the x direction. In order to build some intuition it is useful to first consider a system with OBC in the x direction, in which the position operator \hat{x} is well defined. The k_{y} component of the projector ( 2 ) can be written as a matrix in the position basis P(k_{y})=\sum_{x,x^{\prime}}\ket{x}P_{xx^{\prime}}(k_{y})\bra{x'}, (5) where \ket{x} is an eigenstate of the position operator \hat{x}\ket{x}=x\ket{x} and P_{xx^{\prime}}(k_{y})=\matrixelement{x}{P(k_y)}{x'} . The commutator \commutator{\hat x}{P(k_y)} acting on a position eigenstate \ket{x_0} gives \commutator{\hat x}{P(k_y)}\ket{x_0}=\sum_{x}(x-x_{0})P_{xx_{0}}(k_{y})\ket{x}. (6) Within PBC, the position operator is not well defined. However, Eq. ( 6 ) can still be used if we modify the distances x-x_{0} to account for periodicity. We define a symmetric N_{x}\times N_{x} Toeplitz matrix of distances between lattice sites \Delta_{x} with the first row \begin{pmatrix}0&1&2&\cdots&M&-M&\cdots&-2&-1\end{pmatrix} (7) where M=(N_{x}-1)/2 and N_{x}/2 for N_{x} odd and even, respectively. In the latter case, M needs to be removed from the matrix and -M set to 0. Eq. ( 6 ) reduces to an element-wise product of matrices, denoted by \odot , -i\commutator{\hat x}{P(k_y)}\to-i\Delta_{x}\odot P(k_{y}). (8) An alternative derivation of Eqs. ( 3 ) and ( 8 ) is provided in Appendix A . Eq. ( 8 ) can also be used when calculating commutators \commutator{\hat x}{P} and \commutator{\hat y}{P} in a fully local basis, assuming PBC in both directions. In that case, the formula produces results that are equivalent to those obtained by the formula presented in Ref. Prodan et al. 2010 . Furthermore, it could be useful when evaluating commutators of the type \commutator{\hat x}{P} , in order to calculate the local \mathbb{Z}_{2} topological marker in two-dimensional systems Baù and Marrazzo 2024 or local topological markers in odd spatial dimensions Hannukainen et al. 2022 within PBC. II.2 Local Středa marker While LCM is an excellent tool for diagnosing topological properties in theoretical and numerical applications, its experimental implementation remains unclear. Another local topological marker for Chern insulators that could readily be measured in an experiment, is the local Středa marker c_{S}(\mathbf{r})=\phi_{0}\partialderivative{n(\vb r)}{\phi}, (9) where n(\mathbf{r}) is the local electron density, \phi an external uniform magnetic flux per unit cell, \phi_{0}=h/e_{0} the magnetic flux quantum and e_{0} and h unit charge and Planck constant, respectively. (Unless stated otherwise, all fluxes in the remainder of this paper are given per unit cell.) The derivative must be evaluated in the limit \phi\to 0 and at a fixed chemical potential \mu . The local Středa marker has been connected to both LCM and orbital magnetization Bianco and Resta 2013 . Recently, non-equilibrium behavior of local Chern and Středa markers in systems within fully-OBC has also been investigated Markov et al. 2024 . We calculate c_{S} in the hybrid basis simply as c_{S}(x)=\phi_{0}\partialderivative{\big(\sum_{k_y}{P_{xx}(k_y)\big)}}{\phi} (10) III Haldane ribbon We first consider a Haldane model Haldane 1988 ribbon (OBC in the x and PBC in the y direction) with zigzag boundaries. The model is a 2-dimensional tight-binding Hamiltonian on a honeycomb lattice (Fig. 1 (a)) that features staggered on-site energies \pm M , real nearest neighbor hopping with amplitude t and complex next-nearest neighbor hopping t^{\prime}e^{i\varphi} . We set the hopping amplitudes to t=1 and t^{\prime}=1/3 . We implemented the model using the Python library Kwant Groth et al. 2014 . Figure 1: (a) The Haldane model ribbon: honeycomb lattice with white and black sites corresponding to on-site energies \pm M . The slice, marked with the dotted parallelogram, is the unit cell. The two ‘orbitals’ are marked by the dashed ellipse. (b) Phase diagram of the Haldane model. The Haldane model features different insulating phases at half-filling. Depending on the parameters M and \varphi , the insulator can either be trivial ( C=0 ) or topological ( C=\pm 1 ) according to the phase diagram shown in Fig 1 (b). Below, we consider the two points marked in the phase diagram. The triangle marks a point in the trivial phase with M=1.5 and \varphi=\pi/10 and the cross a point in the topological phase with M=0.2 and \varphi=\pi/2 . Configurations of the LCM, calculated using the procedure described in Sec. II.1 , are shown in Fig. 2 . Panel (a) shows the configuration in the trivial phase (point “triangle” in Fig. 1 (b)). The marker is equal to 0 in the interior and features a small deviation at each boundary. The average of LCM over the entire system approaches C with increasing system size. Figure 2: LCM configurations (a) of the system in the trivial phase (marked by the triangle in Fig. 1 (b)), (b) topological phase (cross in Fig. 1 (b)) and (c) for a heterojunction, with the system’s parameters corresponding to (a) in the left half and (b) in the right half. The two lines represent systems with different sizes N=N_{x}=N_{y} . In the topological phase (point “cross”, panel (b)) LCM equals C=+1 in the bulk and deviates at the boundary. However, in contrast with the fully-OBC case Bianco and Resta 2011 , the boundary contribution does not compensate that of the bulk. Furthermore, the amplitude of the deviations at the boundaries is independent of system size, therefore the average of LCM over the entire system approaches C as \sim 1/N_{x} , as was already demonstrated in Ref. Drigo and Resta 2020 . Panel (c) shows LCM configuration for a ribbon heterojunction. The left half of the system corresponds to the point marked by the triangle and the right half to the point marked by the cross in Fig. 1 (b). The marker is capable of clearly distinguishing between regions with different C . We compute the local Středa marker by including a homogeneous perpendicular magnetic field in the Hamiltonian via Peierls substitution. The hopping amplitude t_{ij} between sites i and j at positions \mathbf{r}_{i(j)} is multiplied by a phase factor t_{ij}\to t_{ij}\exp{-i\frac{2\pi}{\phi_0}\int_{\vb r_i}^{\vb r_j}\dd\vb r\vb A(\vb r)}, (11) where the line integral is along the shortest path between \mathbf{r}_{i} and \mathbf{r}_{j} and \mathbf{A}=\big(0,\delta\phi/(\sqrt{3}/2)[x-x_{0}]\big) is the magnetic vector potential with x_{0} chosen such that x-x_{0}=0 on the central line of the ribbon. The system is then diagonalized at zero and at a finite magnetic flux \delta\phi , keeping \mu constant. Electron densities are calculated for both cases and subtracted in order to calculate c_{S} via Eq. ( 10 ). Fig. 3 shows the behavior of the local Středa marker at different \delta\phi used in its calculation. The value of c_{S} in the bulk is independent of \delta\phi and equals the value of the bulk Chern number and LCM. This is in line with the behavior of the markers in the fully-OBC case, see Refs. Bianco and Resta 2013 ; Markov et al. 2024 . Figure 3: Local Středa marker calculated at different \delta\phi (shown in color scale). The x coordinate is shifted at each flux to show boundary behavior. Red line shows the local Chern marker. The inset shows the average electron density n_{\textrm{el}} in the ground state at each magnetic flux. Size of the system is N_{x}=N_{y}=300 . At the boundary, the value of c_{S} converges with increasing \delta\phi . The \delta\phi at which the value converges decreases with the system’s size as \delta\phi\sim 1/N_{y} (see Movie in supplemental material sup ). This means that the finite \delta\phi needed for c_{S} to converge at the boundaries is only a finite size effect. Once c_{S} converges at the boundary, its value there is similar to LCM. The inset in Fig. 3 shows the average electron density in the ground state at different \delta\phi/\phi_{0} . The dependence is linear with the slope equal to C=1 . This is in accordance with the Středa formula Středa 1982 . We also studied the correspondence between the two markers when disorder is introduced to the system. At each x , a random number \delta_{x} is sampled uniformly from the interval [-\delta/2,\delta/2] and added to the on-site energy term. Importantly, \delta_{x} does not vary along y in order to preserve partial translational symmetry. The markers are calculated as before, with \delta\phi set to 30\,\phi_{0}/N_{y} (marked by the star in the inset of Fig. 3 ). In Fig. 4 we show results for disorder amplitude \delta=1 . One sees the two markers agree closely. (Result with \delta\phi=\phi_{0}/N_{y} is shown in Fig. 9 in Appendix B .) The similarity breaks down for \delta\gtrsim 3 , when the average of LCM in the bulk deviates substantially from C=1 due to Anderson physics (see Figs. 10 – 12 in Appendix B ). Figure 4: Configuration of LCM c (blue line) and local Středa marker c_{S} (orange line) in the topological phase with on-site disorder. Disorder amplitude is \delta=1 , size of the system is N_{x}=N_{y}=300 and \delta\phi=30\,\phi_{0}/N_{y} . IV Kibble-Zurek mechanism in a disordered Chern insulator Local Chern marker has also been used to characterize non-equilibrium states after quantum quenches across topological phase transitions Ulčakar et al. 2020 ; Yuan et al. 2024 . It was found that for slow quenches characterized by a duration \tau , the non-equilibrium state features a length scale that grows as the quench becomes slower, just as expected in systems with standard Landau order parameters. Small random disorder was introduced to reveal the length scale in real space. The length scale was proposed to be governed by the Kibble-Zurek mechanism, but these predictions could be numerically tested for systems of moderate size only. Here we take advantage of partial translational symmetry to show convergence of the results with the system size. IV.1 Kibble-Zurek mechanism Kibble-Zurek mechanism (KZM) Kibble 1976 ; Zurek 1996 describes how the correlation length of a system \xi freezes-out during a slow quench across a critical point. It also predicts that \xi at the end of the quench scales with the duration of the quench (assuming a linear ramp) \tau as \xi\sim\tau^{1/(1+\nu z)}, (12) where \nu is the correlation length critical exponent and z the dynamical critical exponent Dziarmaga 2010 . As described below, the correlation length in a Chern insulator can be obtained from the LCM configuration. However, if the system is translationally invariant, the LCM configuration is homogeneous (as is any other spatially resolved quantity). The correlation length can be revealed by introducing inhomogeneities like random disorder in the system. Since the disorder is needed only to reveal the underlying length scale (present in correlation functions or the Berry curvature Chen 2016 ), it can be chosen in a way that preserves translational symmetry in one direction. This symmetry can then be used to speed up numerical calculations and allow for the study of larger systems. Figure 5: (a) LCM in the ground state of an instantaneous Hamiltonian at u=-2.5 (a1), -2.1 (a2), -2.01 (a3), -1.9 (a4) and -1.5 (a5). (b) LCM during a linear quench from u=-2.5 to u=-1.5 with \tau=20 at times t/\tau=0 (b1), 0.4 (b2), 0.49 (b3), 0.6 (b4) and 1 (b5). The size of the system is N_{x}=N_{y}=300 , although only a third is shown for clarity. IV.2 Model We use the Qi-Wu-Zhang model Qi et al. 2006 and assume PBC in both directions. The disorder \delta_{x} is invariant under translations in the y direction, making k_{y} a good quantum number. The Hamiltonian can therefore be written in block-diagonal form in the (x,k_{y}) basis as H_{u}=\bigoplus_{k_{y}}H_{u}(k_{y}), (13) where k_{y} is an element of the Brillouin zone. Each block H_{u}(k_{y}) has the form (the nearest neighbor hopping amplitude is set to 1) \begin{split}H_{u}&(k_{y})=\sum_{x=1}^{N_{x}}\left(\ket{x+1}\bra{x}\otimes\frac{\sigma_{z}+i\sigma_{x}}{2}+{\rm h.c.}\right)+\\ &\sum_{x=1}^{N_{x}}\ket{x}\bra{x}\otimes[\cos k_{y}\sigma_{z}+\sin k_{y}\sigma_{y}+(u+\delta_{x})\sigma_{z}],\end{split} (14) where x is the cell index and Pauli matrices \sigma_{x,y,z} act on a two-dimensional orbital space. The disorder \delta_{x} is uncorrelated and uniformly distributed on the interval [-\delta/2,\delta/2] . We set disorder amplitude to \delta=5\times 10^{-4} throughout. The phase diagram of the (clean) QWZ model features the following insulating phases: The system is a trivial insulator with C=0 for \absolutevalue{u}>2 and a topological insulator with C=-1 and +1 for -2