TB Hamiltonian of the bilayer 2D QAHI
We begin by employing a Slater-Koster type tight-binding (TB) model to describe a single-layer ferromagnetic orbital-active honeycomb lattice, as depicted in Fig. 1a. This lattice is composed of two identical sublattices, A and B, with each site accommodating two distinct orbitals. Taking into account the nearest-neighbor (NN) hopping interactions between adjacent orbitals, we can express the Hamiltonian for this model as,
In this expression,
i and
j index the atomic sites, while
a and
b specify the atomic orbitals.
\(t_{{ij}}^{{ab}}\) represents the NN hopping energy between distinct orbitals on different sites.
\({\Delta _m}\) signifies the Zeeman splitting associated with the ferromagnetic order, and
\({\sigma _z}\) refers to the Pauli matrix in the z direction for the spin subspace. The final term represents the on-site SOC effect with
\(\lambda\)denoting its strength.
Within the C3v point group symmetry of the honeycomb lattice, three distinct orbital doublets manifest: (px, py), (dxz, dyz), and (dxy, dx2–y2), which correspond to the same 2D irreducible representation30. Consequently, they display identical electronic band structures and wavefunction forms according to Eq. (1), establishing them as symmetrical equivalents. Given the pronounced correlation interactions inherent to the d orbitals, our study primarily focuses on the (dxz, dyz) doublet. Considering the strong π-bond interaction between the dxz/dyz orbitals of different lattice points, we set \({V_{dd\pi }}\)= t and \({V_{dd\delta }}\)= 0.1t. Taking into account the ferromagnetic ground state and the strong on-site spin-orbit coupling, we set \({\Delta _m}\)= 4t and \(\lambda\)= 0.3t. The band structure, depicted in Fig. 1b, reveals Dirac points positioned at the corners of Brillouin zone (K1 and K2) in the absence of SOC. The incorporation of SOC induces a topologically nontrivial band gap at these Dirac points, accompanied by the emergence of Berry curvature monopoles11 and chiral edge states which shown in Supplementary Fig. S1. By evaluating the integral of the Berry curvature of the occupied states over the entire Brillouin zone, we can determine the Chern number (C), which corresponds to the anomalous Hall conductance of the system. As illustrated in Fig. 1c, varying the Fermi level within the band gap results in a Chern number of 1, thus characterizing the system as a 2D QAHI. Notably, reversing the magnetization of the system by altering the Zeeman splitting \(({\Delta _m})\) leads to the reversal of both the Berry curvature and anomalous Hall conductance, implying the potential of achieving layer locked Berry curvature in this 2D QAHIs.
We then consider a bilayer model comprising two 2D QAHI monolayers described by Eq. (1). The two monolayers have identical lattice structures but opposite magnetizations, as depicted in Fig. 2a. The A site in the upper layer perfectly aligns with the B site in the lower layer, showcasing the characteristic of Bernal stacking pattern. The TB Hamiltonian of the bilayer 2D QAHI model is expressed as,
Here, \({t_ \bot }\) denotes the NN interlayer hopping. Considering the weak interlayer vdW interaction, we set \({t_ \bot }\)= 0.1t. The last term accounts for the effects of electric polarization along the out-of-plane direction arising from the intrinsic electric polarization or external electric filed, with \(\delta\) being the electrostatic potential difference between the two layers.
In the absence of electric polarization (\(\delta =0\)), the Berry curvatures and Hall quantum conductance from the two antiferromagnetically stacked monolayers cancel out each other, due to the constrains of PT symmetry, in analogous to the cases of quantum spin Hall insulators (QSHIs)44. Further layer-resolved anomalous Hall conductance calculations indicate that the interlayer hopping has a minimal impact on the quantum anomalous Hall conductance of each 2D QAHI monolayer.
Interestingly, upon introducing an interlayer potential difference \(\left| \delta \right|\) = 0.3t, as depicted in Fig. 2b, the bands of the two layers undergo a relative shift, lifting the P symmetry of the system. When the Fermi level lies between the band edges of the two layers, as shown in the shaded region of Fig. 2b, the Berry curvature of the two monolayers cannot cancel out each other, leading to net layer-polarized Berry curvature and consequently nonzero Hall conductance within this energy region. Notably, the net Berry curvature exhibits dependence on the orientation of electric polarization. Specifically, the reversing the electric polarization orientation of the bilayer leads to the reversal of both the net Berry curvature and anomalous Hall conductance, accompanied by the switch of layer degree between the two layers, as shown in Fig. 2c. Considering that the out-of-plane electric polarization in a bilayer can be induced by interlayer interaction and regulated by interlayer sliding41, this intriguing scenario offers a promising approach for achieving QALHE.
It should be mentioned that the magnitude of the net Berry curvature in the bilayer depends on the distribution of the Berry curvature in the two monolayers and cannot guarantee an integral Chern number. The maximal Hall conductance given by the Hamiltonian of Eq. (2) is nearly the half of integral quantum anomalous Hall conductance corresponding to C = 1. Nevertheless, this value significantly exceeds that of classical Hall effects. In view of its quantum origination, we will continue to refer to it as QALHE.
Atomic and electronic structure of FeS monolayer
To validate the feasibility of the model described above, we explored the properties of a 2D vdW single-layer ferromagnetic material known as FeS. This material has been predicted to exhibit a correlation-enhanced substantial topological bandgap.31 The lattice structure of FeS monolayer comprises of two buckled honeycomb FeS sublayers linked by interlayer Fe-S bonds, as depicted in Fig. 3a. This configuration aligns with the previously synthesized 2D vdW MnSe material.45. The transition metal atom Fe is exposed to a triangular prism crystal field with a C3v point group. This field divides the 3d orbitals into a singlet state dz2 and two doublet states (dxz, dyz) and (dxy, dx2–y2), as shown in Fig. 3d. Given the d6 electronic configuration of Fe2+ and the on-site Coulomb interaction, the doublet state (dxz, dyz) is partially occupied, forming spin-polarized Dirac points at the Fermi level in the K1 and K2 valleys, as illustrated in Fig. 3b. The inclusion of SOC induces a substential band gap of about 0.63 eV at these Dirac points, as depicted in Fig. 3c, aligning with previously reported findings.31 The 2D QAHI aspects of the ferromagnetic FeS monolayer can be verified from the quantized anomalous Hall conductance within the band gap corresponding to C = 1, as illustrated in Fig. 3c, the Berry curvature distribution at the K1 and K2 valleys, as shown in Fig. 3e, and the presence of chiral edge states that connect the valence and conduction bands of FeS monolayer, as depicted in Fig. 3f.
Ferroelectricity and tunable QALHE in FeS bilayer
We then considered a FeS bilayer, consisting of two FeS monolayers arranged in different stacking patterns. Our calculations reveal a preference of an antiferromagnetic interlayer alignment over a ferromagnetic one, with an energy difference of approximately 130 meV/Fe. Exploring different stacking configurations for the FeS bilayer, we identified two energetically most favorable options, denoted as AB and BA, which are energetically degenerate, as depicted in Fig. 4a, similar to the scenario observed in MnSe bilayer46. In the AB stacking configuration, the Fe atoms in the upper monolayer align directly above the S atoms in the lower monolayer, whereas the Fe atoms in the lower monolayer are positioned directly beneath the hexagonal centers of the upper monolayer. In both stacking patterns, the inversion symmetry (P) is lifted, resulting in charge transfer between the monolayers, as illustrated in the inset of Fig. 4b. Consequently, an out-of-plane electric polarization of 2.9 pC/m emerges, comparable to that of the MnSe46 and BN41 bilayers. The inherent electric polarization in the FeS bilayer is manifested as an electrostatic potential difference between the monolayers, as shown in Fig. 4b.
The BA stacking configuration presents a similar scenario, albeit with a reversal in both charge transfer direction and electrostatic potential difference. Utilizing the nudge-elastic-band (NEB) method, depicted in Fig. 4c, we ascertained that the FeS bilayer can transition between the AB and BA configurations through interlayer sliding, encountering an energy barrier of approximately 3.2 meV/f.u. The observed variation in electric polarization during this sliding process demonstrates the feasibility of electrical switching between the two Bernal configurations. Furthermore, as depicted in Fig. 4d, the interlayer potential difference of about 0.13 V induces a comparable level of splitting in the monolayer bands with distinct spin polarizations. A significant band splitting is also evident at the K valley edges. Given the pronounced Berry curvature at the K valley in the monolayer FeS, which aligns with layer-locked Berry curvature, this suggests the potential for realizing an intrinsic Layer Hall effect in ferroelectrically stacked FeS bilayer.
It’s worth noting that both the AB and BA stacked FeS bilayers exhibit an indirect band gap, as shown in Fig. 4d. The the conduction band minimum (CBM) at the K (K1 and K2) valleys arises from the doublet state (dxy, dx2–y2) of the Fe atoms, whereas the valence band maximum (VBM) originates from the p orbitals of the S atoms and is located at the Γ point, as illustrated in Supplementary Fig. 3. This electronic band structure differs significantly from the electronic band structure given by the Hamiltonian of Eq. (2). Fortunately, the electronic band structure of the FeS bilayer can be effectively regulated by applying external strain. A subtle biaxial tensile strain of approximately 2% can shift the edge of the valence band at the K1 and K2 valleys, originating from the Fe atom’s doublet states (dxz, dyz), to a position above the valence band edge at the Γ point, as depicted in Supplementary Fig. 4a. Additionally, applying compressive strain along the out-of-plane direction can enhance the interlayer electric potential difference, as depicted in Supplementary Fig. 4b.
We therefore examined a strained Bernal-stacked FeS bilayer subjected to a 2% in-plane biaxial tensile strain and a 2% compressive strain along the out-of-plane direction, as showcased in Fig. 5a. The electronic band structures of both AB and BA stacked FeS bilayers near the K1 and K2 valleys are consistent with the results of our bilayer model. Specifically, both the VBM and CBM, stemming from the Fe atom’s doublet states (dxz, dyz), are situated at the K1 and K2 valleys, accompanied by the presence of topologically nontrivial band gap due to SOC, as highlighted in Supplementary Fig. 5. Further computations unveiled that when the Fermi level is regulated to the region between the band edges of the two FeS monolayers, anomalous Hall conductance approaching quantized values emerges, as depicted in the shaded region of Fig. 5a. When the Fermi level falls within the band gap of the FeS bilayer, the Berry curvature of the two FeS monolayers cancels each other out, resulting zero anomalous Hall conductance. Changing the orientation of the intrinsic electric polarization by interlayer sliding, which transforms the stacking pattern from AB to BA, reverses both the net Berry curvature and Hall conductance signs, as shown in Fig. 5b. Notably, the layer contributing to the layer-polarized net Berry curvature also switches, indicating that the intrinsic QALHE in the FeS bilayer system can be modulated by ferroelectric polarization. These findings align well with the predictions of our TB model.
Finally, it is important to highlighted that the quantum anomalous Hall conductance exhibits two distinct peaks with opposite signs in the valence and conduction band regions, as depicted in Fig. 2 and Fig. 5. These dual peaks correspond to the opposite spin polarization and distinct electron layer degrees, presenting an alternative method for regulating the QALHE.