---
title: AB-Stacked Bilayer Haldane Lattice
url: https://www.emergentmind.com/topics/ab-stacked-bilayer-haldane-lattice
type: topic
---

# AB-Stacked Bilayer Haldane Lattice

The **AB-stacked bilayer Haldane lattice** denotes a family of Bernal-stacked honeycomb bilayer systems in which the Haldane mechanism—complex next-nearest-neighbor hopping with zero net flux through the unit cell—is combined with interlayer hybridization. In the electronic versions, the basis is typically \(\{A_l,B_l,A_u,B_u\}\), with the \(B\) sublattice of the upper layer directly above the \(A\) sublattice of the lower layer and interlayer tunneling retained only on that vertical pair. Relative to the monolayer Haldane model, the bilayer introduces a four-band structure, interlayer splitting, higher-Chern phases such as \(C=\pm 2\), and additional critical phenomena associated with band touchings at \(\mathbf K\), \(\mathbf K'\), and the semi-Dirac merger point \(\mathbf M\) [2304.02880, 1205.6266, 2208.02491, 2603.24551].

## 1. Structural definition and model space

AB stacking, also called Bernal stacking, is the defining geometric ingredient. In one common convention, the **\(B\) sublattice of the upper layer** lies directly above the **\(A\) sublattice of the lower layer**, and the only retained interlayer hopping is \(t_\perp\) on the vertical bond \(B_u\leftrightarrow A_l\) [2304.02880, 2603.24551]. In another notation used for the bilayer generalization of Haldane’s model, the first layer contains sites \(A,B\), the second layer contains \(C,D\), and **\(A\) sites couple vertically to the \(D\) sites**; this is again Bernal stacking in the manner of bilayer graphene [1205.6266]. The common content of these notations is an asymmetric dimerization pattern that distinguishes AB from AA stacking.

The intralayer Haldane ingredient is the complex next-nearest-neighbor term \(t_2 e^{\pm i\phi}\), which breaks time-reversal symmetry while maintaining **zero total flux per unit cell**. In the anisotropic bilayer electronic model, the real-space Hamiltonian is written as
\[
H = \sum_{p\in l,u}\left[\sum_{\langle ij\rangle} t_{ij} c_i^{p\dagger}c_j^p + t_2 \sum_{\langle\langle im\rangle\rangle} e^{i\phi_p^{im}} c_i^{p\dagger}c_m^p + \text{h.c.}\right] + \left[t_\perp \sum_{\langle q,r\rangle_\perp} c_q^{l\dagger} c_r^u + \text{h.c.}\right],
\]
with \(t_{ij}=t_1\) on one nearest-neighbor bond direction \(\boldsymbol{\delta}_1\) and \(t_{ij}=t\) on \(\boldsymbol{\delta}_{2,3}\) [2304.02880]. The anisotropy \(t_1\neq t\) is the “band engineering” that moves the Dirac points through the Brillouin zone.

A broader usage of the term encompasses closely related constructions. One line of work studies a **bilayer of the modified Haldane model**, where each isolated layer is semimetallic and AB stacking alone opens a chiral insulating gap with \(C=\pm 2\) [2208.02491]. Another studies **AB-stacked moiré bilayers** in which the honeycomb sublattices reside in different layers and a generalized Kane-Mele Hamiltonian realizes a Haldane Chern insulator under a small magnetic field [2207.02312]. A further extension treats **AB-stacked bilayer honeycomb quantum magnets**, where alternating next-nearest-neighbour Dzyaloshinsky-Moriya interaction plays the role of the Haldane term for magnons [1604.05292].

## 2. Canonical Hamiltonians and low-energy descriptions

In momentum space, the anisotropic electronic bilayer Haldane model is expressed in the basis \(\{\mathrm{A}_l,\mathrm{B}_l,\mathrm{A}_u,\mathrm{B}_u\}\) as
\[
H(\mathbf{k}) = \begin{pmatrix}
h^+_z(\mathbf{k},\phi_l) & h_{xy}(\mathbf{k},t_1) & 0 & t_\perp \\
h_{xy}^*(\mathbf{k},t_1) & h^-_z(\mathbf{k},\phi_l) & 0 & 0 \\
0 & 0 & h^+_z(\mathbf{k},\phi_u) & h_{xy}(\mathbf{k},t_1) \\
t_\perp & 0 & h_{xy}^*(\mathbf{k},t_1) & h^-_z(\mathbf{k},\phi_u)
\end{pmatrix},
\]
with
\[
h^+_z(\mathbf{k},\phi_p)=h_0(\mathbf{k},\phi_p)+h_z(\mathbf{k},\phi_p),\qquad
h^-_z(\mathbf{k},\phi_p)=h_0(\mathbf{k},\phi_p)-h_z(\mathbf{k},\phi_p),
\]
and
\[
h_{xy}(\mathbf{k},t_1)=h_x(\mathbf{k},t_1)-i h_y(\mathbf{k},t_1).
\]
The explicit functions \(h_0,h_z,h_x,h_y\) contain the Haldane phase \(\phi_p\), the anisotropic nearest-neighbor amplitudes \(t_1,t\), and the honeycomb trigonometric structure factors [2304.02880].

For equal fluxes \(\phi_l=\phi_u\), the four bands are two conduction bands and two valence bands,
\[
E^c_\pm = h_0 + \sqrt{\frac{t_\perp^2}{2} + |h_{xy}|^2 + h_z^2 \pm \frac{t_\perp}{2}\sqrt{t_\perp^2+4h_{xy}^2}},
\]
\[
E^v_\pm = h_0 - \sqrt{\frac{t_\perp^2}{2} + |h_{xy}|^2 + h_z^2 \pm \frac{t_\perp}{2}\sqrt{t_\perp^2+4h_{xy}^2}},
\]
showing explicitly how interlayer hybridization splits the monolayer spectrum into \(c_1,c_2,v_1,v_2\) sectors [2304.02880].

The bilayer generalization analyzed from the viewpoint of edge states, entanglement spectra, and Wannier functions adopts a block form
\[
H(\vec k)=
\begin{pmatrix}
H_{AB}(\vec k) & \tilde T \\
\tilde T^{\rm T} & H_{CD}(\vec k)
\end{pmatrix},
\qquad
H_{AB}=H(\phi,m),\qquad
H_{CD}=H(\chi,-m),
\]
where each layer is a Haldane monolayer and \(\tilde T\) encodes the Bernal interlayer hopping [1205.6266]. In the special case \(\phi=-\chi\), the Hamiltonian satisfies an extended inversion-related symmetry
\[
\mathcal{I}\, H(\vec k)\,\mathcal{I}=H^*(\vec k),
\]
denoted \(I^*\), which is not ordinary inversion and becomes central for the bilayer’s non-edge topological signatures [1205.6266].

In the stacking-induced Chern-insulator construction, the starting point is instead the **modified Haldane model**, where the Haldane mass is replaced by a **valley-dependent scalar potential**. Two time-reversed copies with \(\Phi_1=-\Phi_2\) are stacked in AB geometry, and the interlayer dimer coupling \(2t_\perp\) removes the nodal-line degeneracy by band repulsion [2208.02491]. This mechanism differs microscopically from the ordinary bilayer Haldane lattice, but it produces a closely related AB-stacked chiral phase.

## 3. Topological invariants and Chern-sector organization

The defining topological data are the band Chern numbers. In the band-engineered AB bilayer Haldane model, the valence band closer to the Fermi level, \(v2\), carries
\[
C=\pm 2 \quad \text{or} \quad C=\pm 1,
\]
whereas the valence band farther from the Fermi level, \(v1\), typically carries
\[
C=\pm 1 \quad \text{or} \quad 0.
\]
The conduction bands have the same magnitudes with opposite signs. A notable feature is that the higher-Chern phases \((\pm2)\) occur only in the band \(v2\), the valence band nearest \(E_F\) [2304.02880].

Multiple topological phase transitions occur when gaps close and reopen. Reported jumps include
\[
\pm 2 \leftrightarrow \mp 2,\qquad
\pm 1 \leftrightarrow \mp 1,\qquad
\pm 1 \leftrightarrow 0 \leftrightarrow \pm 2,\qquad
\pm 2 \leftrightarrow \pm 1,
\]
with the relevant band touchings occurring at \(\mathbf K\), \(\mathbf K'\), and in the anisotropic regime near the \(\mathbf M\)-related merger [2304.02880]. These transitions are the bilayer analogue of mass-sign reversals in a Dirac description, but the four-band structure permits a larger set of gap-closing channels than in the monolayer.

The stacking-induced modified-Haldane bilayer provides a more explicit higher-Chern formula. Near valley \(\xi=\pm\), the low-energy effective Hamiltonian takes the form
\[
H_\text{eff}(\mathbf{q}) = -\frac{\hbar^2 v^2}{t_\perp}(q_x^2-q_y^2)\sigma_x
+2\xi\frac{\hbar^2 v^2}{t_\perp}q_xq_y\,\sigma_y
+\left[ \frac12(M_1+M_2)-\operatorname{sgn}(\Phi_1)\,3\sqrt{3}\,t_2\,\xi \right]\sigma_z
+\frac12(M_2-M_1)\sigma_0,
\]
and the lowest-band Chern number is
\[
C=\sum_{\xi}\frac12\,\chi\,\operatorname{sgn}(m_\xi),
\qquad
m_\xi=\frac12(M_1+M_2)-\operatorname{sgn}(\Phi_1)\,3\sqrt{3}\,t_2\,\xi,
\qquad
\chi=-2\xi.
\]
For \(M_1=M_2=0\), this yields
\[
C=\operatorname{sgn}(\Phi_1)\,2,
\]
so the AB bilayer realizes a Chern insulator with \(|C|=2\) [2208.02491].

A more conventional bilayer rule appears when two Haldane monolayers are coupled without closing the bulk gap: the **total Chern number is the sum of the two monolayer Chern numbers**,
\[
C_{\text{bilayer}} = C_{AB}+C_{CD}.
\]
This additive statement holds in the bilayer generalization as long as the central gap remains open under interlayer coupling [1205.6266]. The result is conceptually important because it distinguishes ordinary Chern addition from stacking-induced topology, where the uncoupled layers may be semimetallic rather than individually topological [2208.02491].

## 4. Anisotropy, semi-Dirac criticality, and Floquet mass competition

A characteristic feature of several AB bilayer Haldane models is the motion and merger of Dirac points under **anisotropic nearest-neighbor hopping**. As \(t_1\) increases from \(t\), the two inequivalent Dirac points move toward each other, and at
\[
t_1=2t
\]
they merge at the Brillouin-zone \(\mathbf M\) point. At this **semi-Dirac limit**, the dispersion is linear along one direction and quadratic along the other [2304.02880, 2603.24551].

The Floquet-driven study makes the low-energy structure explicit. Expanding around the merger point,
\[
f(\mathbf{q}) \approx \hbar v_F q_x + i\hbar \alpha q_y^2,
\qquad
\hbar v_F = \frac{\sqrt{3}t}{a_0}, \qquad
\hbar\alpha = \frac{t}{2a_0^2},
\]
so that
\[
H(\mathbf{q}) = \hbar v_F q_x \sigma_x + \hbar \alpha q_y^2 \sigma_y,
\qquad
E(\mathbf{q})=\pm \hbar\sqrt{v_F^2 q_x^2+\alpha^2 q_y^4}.
\]
The semi-Dirac point is therefore an anisotropic critical point rather than an ordinary Dirac crossing [2603.24551].

In the static band-engineered bilayer, increasing \(t_1\) causes the Berry-curvature “Chern lobes” in the \((\phi_l,\phi_u)\) plane to shrink; at \(t_1=2t\), the gap closes and the Chern numbers collapse to zero; for \(t_1>2t\), the gap reopens but the phases are topologically trivial [2304.02880]. The Floquet-driven bilayer introduces an additional tunable mass,
\[
\lambda_\omega = \frac{[H_{-1},H_{+1}]}{\hbar\omega_0}
= \gamma\,\frac{(e v_F A_0)^2}{\hbar\omega_0},
\]
through off-resonant circularly polarized light with vector potential
\[
\mathbf{A}(t)=A_0\left(\gamma \sin\omega_0 t,\ \cos\omega_0 t\right),
\qquad \gamma=\pm1.
\]
The effective mass entering the low-energy description becomes
\[
m_{\mathrm{eff}}=m_H+\lambda_\omega,
\]
so topology is controlled by the competition between the intrinsic Haldane mass and the Floquet-induced mass [2603.24551].

Interlayer coupling is not a passive perturbation in this regime. The Floquet study reports that interlayer hybridization reshapes the band topology by inducing **helicity-dependent and valley-selective band inversions** at \(K\) and \(K'\), stabilizing higher-Chern phases in the valence bands and redistributing the Berry curvature over the Brillouin zone [2603.24551]. This suggests that the post-merger behavior depends sensitively on whether the system is treated as a purely static band-engineered bilayer or as a driven Floquet bilayer with an additional mass channel.

## 5. Boundary spectra, Wilson loops, and transport signatures

The usual bulk-boundary correspondence survives in many AB bilayer Haldane regimes. In the band-engineered bilayer nanoribbon geometry, for \(t_1<2t\) there are **chiral edge states** traversing the bulk gap, and in \(|C|=2\) phases there are **two edge modes per boundary**, with opposite boundaries carrying currents in opposite directions [2304.02880]. In the AB-stacked modified-Haldane ribbons, finite \(t_\perp\) opens a bulk gap and **two chiral edge states appear at each boundary**; reversing the sign of the phases reverses their propagation and changes \(C\to -C\) [2208.02491].

The anomalous Hall response provides a complementary bulk diagnostic. For the static band-engineered bilayer,
\[
\sigma_{xy} = \frac{\sigma_0}{2\pi}\sum_\lambda \int \frac{dk_x\,dk_y}{(2\pi)^2}\, f(E^\lambda_{k_x,k_y})\,\Omega(k_x,k_y),
\qquad
\sigma_0=\frac{e^2}{h},
\]
and when the Fermi energy lies in the bulk gap the conductivity is quantized at
\[
2\sigma_0.
\]
The plateau width equals the bulk-gap width, narrows as \(t_1\to 2t\), disappears at the semi-Dirac point, and vanishes in the trivial regime [2304.02880]. The Floquet-driven bilayer likewise exhibits quantized anomalous Hall plateaus at
\[
\sigma_{xy}=2\frac{e^2}{h},
\]
together with plateau collapse or sign reversal under changes in anisotropy, helicity, and interlayer coupling [2603.24551].

A distinct bilayer phenomenon is the **breakdown of the monolayer edge–entanglement–Wannier correspondence**. In the \(I^*\)-symmetric opposite-Chern case \((\phi=-\chi)\), the open-boundary edge spectrum is gapped and shows **no protected crossing**, yet the entanglement spectrum contains **protected half-occupancy modes** and the non-Abelian Wilson loop exhibits oppositely winding eigenphases crossing at \(\gamma=0,\pi\) [1205.6266]. The Wilson loop is
\[
\hat{\mathcal W} = \mathcal{P}\exp\left(i\int_0^{2\pi} dk\,\hat A(k)\right),
\]
and in the \(I^*\)-symmetric bilayer it reduces to
\[
\hat{\mathcal W}(k_1)=e^{i\tilde\gamma(k_1)\sigma_y},
\qquad
\gamma_w=\pm\tilde\gamma.
\]
This is not the standard inversion-protected \(\mathbb Z_2\) mechanism; the protecting symmetry is \(I^*\), not ordinary inversion, and the protected entanglement feature is not pinned to inversion-invariant momenta [1205.6266].

## 6. Realizations, analogues, and conceptual boundaries

The most direct experimental solid-state realization in the supplied literature is the **AB-stacked MoTe\(_2\)/WSe\(_2\) moiré heterobilayer**, whose moiré potential forms a honeycomb lattice with the two sublattices residing in different layers [2207.02312]. At filling \(\nu=2\), the moiré bilayer is a **quantum spin Hall insulator** with a tunable charge gap. Under a small out-of-plane magnetic field it becomes a Chern insulator with **Chern number \(c=1\)**, described by a generalized Kane-Mele tight-binding Hamiltonian containing nearest-neighbor inter-sublattice hopping, complex next-nearest-neighbor hopping, a tunable layer potential difference \(\Delta\), and a Zeeman term \(V_z\) [2207.02312]. Experimentally, the Hall plateau approaches
\[
R_{xy}\approx \frac{h}{e^2}
\]
within about \(2\%\) for fields between \(1\) T and \(3\) T, while \(R_{xx}\) drops below \(1\,\mathrm{k}\Omega\), and the line in \((\nu,B)\) space follows
\[
\frac{d\nu}{dB} \approx 1,
\]
consistent with the Středa relation and \(c=1.00\pm0.05\) [2207.02312].

The AB-stacked bilayer Haldane lattice also has a **bosonic analogue** in honeycomb quantum magnets. There, alternating next-nearest-neighbour Dzyaloshinsky-Moriya interaction is the Haldane ingredient, producing topological magnon bands in an AB-stacked bilayer geometry with sublattices \(A_1,B_1,A_2,B_2\) and interlayer exchanges \(J_0,J_1,J_2\) [1604.05292]. For ferromagnetic interlayer coupling, the magnon Chern numbers are
\[
\mathcal{C}_\mu=[2,-2,0,0],
\]
whereas for antiferromagnetic interlayer coupling they become
\[
\mathcal{C}_\mu=[2,-2,2,-2].
\]
Edge-state propagation is correspondingly layer-co-propagating in the ferromagnetic case and layer-counter-propagating in the antiferromagnetic case [1604.05292].

A further conceptual extension replaces explicit Haldane hopping by **toroidal magnetic order**. A simplified two-band excitonic-insulator model with toroidal order is unitarily equivalent, at low energy, to the Haldane model on the honeycomb lattice with zero net flux through the unit cell [1010.2369]. The toroidal moment density is
\[
\mathbf T(\mathbf r)=\mathbf r\times[\mathbf r\times \mathbf j(\mathbf r)],
\]
and the ordered phase yields a Haldane-type insulating state with broken time-reversal symmetry, zero total flux per unit cell, nonzero topological index, and chiral edge modes [1010.2369]. This is not an AB-stacked bilayer construction, but it establishes a symmetry-based route to Haldane physics in real magnetic materials.

Several misconceptions are excluded by these results. **Zero net flux does not imply topological triviality**: the Haldane mechanism, the toroidal-order analogue, and the magnon DMI construction all rely on staggered or periodic internal flux patterns with zero total flux per unit cell [1010.2369, 1604.05292]. **AB stacking is not interchangeable with AA stacking**: in the stacking-induced modified-Haldane bilayer, AA stacking remains gapless when \(M_1=M_2=0\) and becomes a trivial insulator when Semenoff masses are present, with no Chern-insulating phase induced by AA stacking [2208.02491]. The moiré realization likewise states that the mechanism is not expected in AA-stacked bilayers [2207.02312]. A plausible implication is that the defining content of the AB-stacked bilayer Haldane lattice is not merely “two Haldane layers plus hopping,” but a specific interplay of Bernal geometry, layer-selective hybridization, and Haldane-type mass generation.

Source: https://www.emergentmind.com/topics/ab-stacked-bilayer-haldane-lattice