Kane–Mele–Hubbard Model: Correlated Topology

• The Kane–Mele–Hubbard model is a correlated topological lattice model that combines intrinsic spin–orbit coupling with on-site Hubbard interactions to bridge quantum spin Hall and antiferromagnetic Mott insulator regimes.
• It employs advanced numerical and analytical methods—such as determinant QMC, DMRG, and mean-field theories—to analyze phase transitions, edge instabilities, and spin-liquid behavior.
• The model provides a versatile framework for investigating symmetry-protected edge states, interaction-driven magnetic instabilities, and superconducting channels in correlated electron systems.

The Kane–Mele–Hubbard (KMH) model is the interacting extension of the Kane–Mele honeycomb-lattice topological band model by an on-site Hubbard term, and it has become a standard setting for studying the interplay of intrinsic spin-orbit coupling, Mott physics, antiferromagnetism, helical boundary modes, and correlation-driven topology. In its most commonly studied form, the model interpolates between a weak-coupling quantum spin Hall or topological band-insulating regime and a strong-coupling antiferromagnetic Mott insulator, while also supporting a substantial literature on spin-liquid behavior, edge instabilities, doped superconducting tendencies, Rashba coupling, magnetic flux, Zeeman fields, and symmetry-broken topological descendants (Zheng et al., 2010).

1. Microscopic formulation and symmetries

A standard half-filled KMH Hamiltonian on the honeycomb lattice consists of nearest-neighbor hopping, intrinsic next-nearest-neighbor spin-orbit hopping, and an on-site Hubbard interaction. In the notation used in a sign-free half-filled formulation, the noninteracting part is

$$H_0=-t\sum_{\langle i,j\rangle,\sigma}c_{i\sigma}^{\dag}c_{j\sigma} +i\lambda\sum_{\langle\langle i,i^\prime \rangle\rangle\alpha,\beta} \Big\{ c^\dagger_{i\alpha}\sigma_{z,\alpha\beta}c_{i^\prime\beta} -c^\dagger_{i^\prime\alpha}\sigma_{z,\alpha\beta}c_{i\beta} \Big\} -\mu\sum_{i,\sigma}c^\dagger_{i\sigma}c_{i\sigma},$$

with

$$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$

Here $t$ is the nearest-neighbor hopping, $\lambda$ the intrinsic spin-orbit coupling, $\sigma_z$ acts in spin space, and $\mu=0$ corresponds to half-filling in the particle-hole-symmetric formulation (Zheng et al., 2010).

The intrinsic spin-orbit term preserves time-reversal symmetry and conserves $S^z$, so full spin $SU(2)$ is reduced to $U(1)$. This residual symmetry is central to the easy-plane character of the antiferromagnetic phase and to helical-edge bosonization analyses. When Rashba coupling is added,

$$i \lambda_R \sum_{\langle ij \rangle\,\alpha\beta} c_{i\alpha}^\dag (\boldsymbol{\sigma}_{\alpha\beta} \times \mathbf{d})_z \,c_{j\beta},$$

the remaining $$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$0 spin symmetry is broken completely, inversion symmetry is explicitly broken, and the interacting topological classification must be inferred from spectral and edge-state structure rather than from the simpler $$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$1-conserving picture (Laubach et al., 2013).

A particularly important special case is the half-filled model with purely imaginary next-nearest-neighbor hopping. In that case the particle-hole transformation

$$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$2

is an exact symmetry. This symmetry underlies several nontrivial consequences: it fixes the average density of each spin on every site to $$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$3, forbids equilibrium bond currents even with open edges, and removes the determinant-QMC sign problem at half-filling (Zheng et al., 2010).

2. Canonical half-filled phase diagram

At weak coupling, the half-filled KMH model is a topological band insulator with a spin-orbit-induced bulk gap and helical edge modes. At strong coupling, it becomes an antiferromagnetic Mott insulator with easy-plane order. The strong-coupling anisotropy follows from the next-nearest-neighbor exchange

$$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$4

which frustrates $$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$5-axis Néel order and favors $$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$6-plane order (Zheng et al., 2010).

Projective determinant QMC at $$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$7 reported a bulk magnetic transition at

$$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$8

while an independent QMC study found

$$H_{int}=U\sum_i\Big(n_{i\uparrow}-\frac12\Big)\Big(n_{i\downarrow}-\frac12\Big).$$9

and identified the transition from topological band insulator to antiferromagnetic Mott insulator as consistent with the universality class of the three-dimensional XY model [(Zheng et al., 2010); (Hohenadler et al., 2011)]. These results are mutually consistent at the quoted resolution and support the now-standard weak-coupling TBI to strong-coupling easy-plane AFMI picture.

The phase structure at small spin-orbit coupling is more intricate. A refined QMC phase diagram found a semimetal at $t$0 and weak $t$1, an antiferromagnetic Mott insulator at large $t$2, and a quantum spin liquid in between, with a multicritical region near $t$3. The same work reported that the spin-liquid–topological-insulator and spin-liquid–Mott-insulator transitions appear continuous within numerical resolution (Hohenadler et al., 2011). This does not settle the spin-liquid question universally, but it establishes that the half-filled KMH model cannot be reduced to a featureless direct TBI–AFMI interpolation in the small-$t$4 regime.

An analytic stochastic functional approach reformulated the half-filled Mott transition as a single variational condition,

$t$5

and derived the ordered phase as an $t$6-$t$7 antiferromagnet with

$t$8

Within that treatment, the bulk single-particle gap remains finite at the Mott transition because the interaction-generated order parameter enters in quadrature with the noninteracting gap scale (Hutchinson et al., 2021). A plausible implication is that the half-filled KMH transition is most naturally viewed as a correlation-driven magnetic instability of a topological insulator, rather than as a simple band inversion.

Finite-temperature weak-to-intermediate-coupling theory reaches the same qualitative endpoint from another direction. A spin-orbit-coupled extension of the two-particle self-consistent approach found a quantum spin Hall regime at small and intermediate $t$9, exponentially growing transverse antiferromagnetic correlation lengths near the transition, and a critical interaction increasing with $\lambda$0, consistent with the statement that spin-orbit coupling stabilizes the topological phase against magnetic order (Lessnich et al., 2023).

3. Edge physics, currents, and topological diagnostics

The half-filled particle-hole-symmetric KMH model has a distinctive edge phenomenology. Although the model supports helical boundary modes in the weak-coupling TBI regime, exact particle-hole symmetry forces both charge and spin currents to vanish on every bond, including along open edges. The corresponding bond-current operators are odd under the particle-hole transformation, so their expectation values must vanish in any symmetric state. This establishes that equilibrium edge current expectation values are not reliable diagnostics of topological order in this model (Zheng et al., 2010).

The same work sharpened the distinction between the existence of edge states in the bare Hamiltonian and the stability of those edge states against generic symmetry-allowed perturbations. In a zigzag ribbon geometry, the edge spin correlations decay as power laws, and the extracted helical Luttinger parameter $\lambda$1 controls whether two-particle backscattering becomes relevant. For $\lambda$2, the reported estimates were

$\lambda$3

so the crossover from stable to unstable helical edges occurs near $\lambda$4, well below the bulk antiferromagnetic transition (Zheng et al., 2010). In this intermediate regime the bulk remains paramagnetic, but the edge is unstable to symmetry-allowed two-particle perturbations once $\lambda$5. The authors explicitly stressed that this “bulk paramagnetic phase with unstable edges” is inferred from Luttinger-liquid diagnostics rather than observed directly in the strict $\lambda$6-conserving KMH Hamiltonian.

The interacting topological characterization of the half-filled KMH model has itself become a methodological subject. A QMC review emphasized three principal diagnostics: $\lambda$7-flux insertion, the zero-frequency Green’s function and parity-based $\lambda$8 invariant, and the spin Chern number. In inversion-symmetric cases, the zero-frequency Green’s function yields a practical interacting topological Hamiltonian, while $\lambda$9-flux insertion produces localized spin-fluxon states in the topological phase (Meng et al., 2013). The same review also highlighted a central limitation: for the interaction-driven TBI-to-AFMI transition, the single-particle gap does not close, whereas the spin gap does. Consequently, single-particle Green’s-function topological indices may remain unchanged across a transition driven by collective two-particle physics (Meng et al., 2013).

That point matches the direct QMC phase-diagram studies. A key result is that the interaction-driven TBI-to-AFMI transition does not resemble a conventional noninteracting topological transition: the single-particle gap stays finite, but the spin gap closes (Hohenadler et al., 2011). This has become one of the standard lessons drawn from the KMH benchmark.

At finite temperature, transport adds further structure. In a TPSC treatment, the spin Hall conductivity decreases with increasing $\sigma_z$0 because antiferromagnetic spin fluctuations renormalize the gap downward. However, momentum-dependent vertex corrections, interpreted as analogues of Maki–Thompson terms, are essential near the transition and are required to recover the zero-temperature quantized value

$\sigma_z$1

throughout the interacting topological phase (Lessnich et al., 2023).

4. Numerical and analytical approaches

The KMH model is unusual among correlated topological lattice models because several nonperturbative methods become available in symmetry-restricted regimes. In the half-filled particle-hole-symmetric case with purely imaginary next-nearest-neighbor hopping, determinant QMC is sign-problem-free because the spin-up and spin-down determinants become complex conjugates of each other for every Hubbard–Stratonovich configuration, so their product is positive definite. This permits high-precision projective-QMC calculations of bulk and edge observables at zero temperature (Zheng et al., 2010).

Beyond that benchmark regime, the literature uses a broad methodological spectrum. Variational cluster approach studies including Rashba coupling resolved correlated topological-insulator, metallic, and direct-gap-only topological-semiconductor regimes, as well as the strong suppression of easy-plane antiferromagnetism by Rashba-induced frustration (Laubach et al., 2013). Spin-orbit-coupled TPSC provides a weak-to-intermediate-coupling description of finite-temperature self-energy, susceptibilities, and transport (Lessnich et al., 2023). Slave-boson mean-field theory addresses half filling and doping within a fractionalized parton language and is particularly used to discuss spin-liquid and superconducting tendencies (Wen et al., 2011). Exact diagonalization with twisted boundary conditions has been used to analyze antiferromagnetic Chern-insulator behavior in the KMH model with staggered potential, including cases where standard many-body Chern-number algorithms fail because adiabatic continuity breaks down in twist-angle space (Wang et al., 24 Apr 2026). Edge-specific interaction effects can also be isolated with CT-INT by treating only boundary modes as interacting degrees of freedom coupled to a noninteracting bulk bath (Bercx et al., 2014).

This range of methods has a clear conceptual consequence. The KMH model is not a single numerical problem but a family of correlated topological problems whose accessible observables depend strongly on symmetry: sign-free QMC is strongest at half filling and with conserved $\sigma_z$2; DMRG is effective in narrow doped cylinders; VCA and mean field reach symmetry-broken or Rashba regimes more easily; and ED with twisted boundaries is especially useful when quantized Hall responses or many-body Chern numbers are at issue. This suggests that no single method resolves the full KMH landscape.

5. Doping, superconducting channels, and attractive interactions

Away from half filling, the KMH model supports several distinct superconducting scenarios, depending on method, geometry, and filling regime. In a constrained-phase QMC study of the doped KMH model, singlet $\sigma_z$3 pairing correlations dominate close to half filling, whereas triplet $\sigma_z$4 correlations dominate below the van Hove singularity associated with three-quarter filling. The same work correlated these trends with the topology of the noninteracting Fermi surface: a nested higher-density Fermi surface favors $\sigma_z$5, while a small $\sigma_z$6-centered pocket favors $\sigma_z$7 (Ma et al., 2014). The effective $\sigma_z$8 vertex contribution grows strongly with $\sigma_z$9, which the authors interpreted as an interaction-generated attraction in that triplet channel.

A more recent DMRG study of light hole doping $\mu=0$0 on $\mu=0$1 cylinders focused on quasi-one-dimensional superconducting tendencies rather than a definitive two-dimensional phase diagram. At $\mu=0$2, it reported a metallic regime for $\mu=0$3 and a possible superconducting regime for $\mu=0$4 or $\mu=0$5, with a strong SOC dependence of the onset scale: $\mu=0$6 The paper explicitly described this as evidence for a “possible SC phase” on a narrow cylinder, with dominant nearest-neighbor spin-singlet pairing correlations rather than a fully established two-dimensional superconducting order parameter (Gupta et al., 2024).

Slave-boson mean-field work addressed the same general question from a different perspective. There, a narrow gapped spin-liquid window was found at half filling, and doping that spin liquid led immediately to superconductivity with a nonmonotonic “optimal” doping dependence of the singlet pairing amplitude (Wen et al., 2011). Because the same work also predicts an unphysical weak-coupling superconducting phase already at half filling, this result is best read as a mean-field tendency rather than a settled phase boundary.

For attractive $\mu=0$7, the KMH problem changes character completely. A self-consistent mean-field study of the attractive-$\mu=0$8 KMH model identified an edge superconducting state in which superconductivity appears immediately near helical edges for any nonzero attraction, while the bulk remains insulating until $\mu=0$9 exceeds a finite critical value $S^z$0. In that picture the topological-insulator–to–superconductor transition proceeds in two steps,

$S^z$1

because the gapless edge modes have finite low-energy density of states whereas the bulk does not (Jie et al., 2012). A later half-filled attractive-KMH study that included real next-nearest-neighbor hopping found that the phase diagram in the topological region changes significantly when that hopping is retained, and that the Goldstone-mode sound velocity differs by about $S^z$2 between a T-matrix and Bethe–Salpeter treatment, indicating that bubble-diagram contributions are small near the superfluid transition boundary considered (Koinov, 2020).

6. Generalized KMH settings and descendant phases

Several extensions of the KMH model reveal how sensitive its correlated topology is to symmetry, filling, and background fields. Restoring Rashba spin-orbit coupling yields a correlated phase diagram containing a topological insulator, a metallic regime, and a “weak topological semiconductor” or direct-gap-only topological phase, while strong Rashba coupling frustrates easy-plane antiferromagnetism and likely promotes incommensurate magnetism (Laubach et al., 2013).

A $S^z$3-flux version of the KMH model doubles the unit cell and changes each spin sector from $S^z$4 to $S^z$5. At half filling the spinful system is $S^z$6-trivial but still hosts two helical edge-state pairs per edge, protected at the single-particle level by translation symmetry. Bosonization predicts, and QMC confirms at strong coupling, that half-filled edge modes are gapped by umklapp scattering (Bercx et al., 2014).

At quarter filling, the KMH model has been proposed as a unifying framework for intrinsic Dirac half-metals. In that interpretation, exchange splitting generated by Hubbard $S^z$7 removes one spin channel near the Fermi level, leaving an effectively spinless topological Dirac sector that becomes a ferromagnetic Chern insulator once spin-orbit coupling gaps it. The quarter-filled model therefore supports a very different phase structure from the half-filled one, centered on ferromagnetic metals, ferromagnetic Chern metals, ferromagnetic Chern insulators, and a strong-coupling trivial Mott phase (Mellaerts et al., 2021).

Adding orbital magnetic flux and nearest-neighbor repulsion leads to an extended Hofstadter KMH problem. At flux $S^z$8 per plaquette, unrestricted Hartree–Fock on the enlarged magnetic unit cell yields a large family of symmetry-broken normal and Chern insulators, including coplanar magnetic order at half filling and “topological multiferroic” phases with coexisting Chern number, magnetic order, and electric multipole order at other integer fillings (Mishra et al., 2018).

Two recent symmetry-breaking extensions show how interactions can generate higher-order or Chern topological descendants. With an in-plane Zeeman field, projector QMC and mean field found a higher-order topological insulator with mirror-inversion-symmetry-protected corner states on a diamond-shaped honeycomb lattice, and a Mott transition to an antiferromagnetic insulator as $S^z$9 increases. In the noninteracting Kane–Mele limit, the upper Zeeman field for corner states was reported as

$SU(2)$0

The same study argued that, within the HOTI regime, Hubbard interaction effectively contributes an additional in-plane Zeeman field (Zhang et al., 2024). With a staggered sublattice potential, exact diagonalization has been used to argue for an antiferromagnetic Chern-insulator phase with $SU(2)$1, although the topological diagnosis is subtle because standard many-body Chern-number evaluation fails when exact level crossings destroy adiabatic continuity in twist-angle space (Wang et al., 24 Apr 2026).

A broad inference from these descendants is that the KMH model is best understood not as a single fixed phase diagram but as a generative interacting topological framework. Depending on filling, lattice geometry, spin-orbit content, and external fields, it supports quantum spin Hall insulation, easy-plane antiferromagnetism, spin-liquid behavior, unstable helical edges, superconducting tendencies, ferromagnetic Chern phases, higher-order topology, and antiferromagnetic Chern insulating states. The half-filled sign-free model remains the benchmark case, but the full KMH literature shows that its significance lies equally in how naturally it deforms into correlated topological problems beyond that benchmark.