Generalized Kane–Mele–Hubbard Model

• Generalized Kane–Mele–Hubbard model is a family of honeycomb-lattice Hamiltonians that extend the canonical model by including electron interactions and terms like Rashba coupling, sublattice potentials, and fluxes.
• The model uncovers diverse quantum phases—including topological insulators, Mott states, and superconductivity—via methods such as quantum Monte Carlo, DMRG, and exact diagonalization.
• Its generalizations provide practical insights into how spin–orbit coupling, doping, and lattice perturbations drive magnetic orders, higher-order topologies, and phase transitions.

Searching arXiv for recent and foundational work on generalized Kane–Mele–Hubbard models. The generalized Kane–Mele–Hubbard model denotes a family of interacting honeycomb-lattice Hamiltonians built from the Kane–Mele band structure and an onsite Hubbard term, then extended by ingredients such as finite doping, Zeeman fields, sublattice potentials, long-range hopping, Rashba spin–orbit coupling, or background fluxes. In its canonical form, the spinful Kane–Mele–Hubbard Hamiltonian combines nearest-neighbor hopping, intrinsic next-nearest-neighbor spin–orbit coupling, and local repulsion,

$$\mathcal{H}= -t \sum_{\langle i,j\rangle,\sigma} c^\dagger_{i\sigma} c_{j\sigma} +i\lambda \sum_{\langle\!\langle i,j\rangle\!\rangle,\sigma} \sigma\, c^\dagger_{i\sigma} c_{j\sigma} +U\sum_i n_{i\uparrow}n_{i\downarrow},$$

or equivalent notations using $\nu_{ij}s^z$ on next-nearest-neighbor bonds (Gupta et al., 2024). At half filling and without additional perturbations, this model interpolates between a quantum spin Hall or topological band-insulating regime and correlation-driven magnetic or Mott phases; its generalizations enlarge that phase space to include higher-order topology, antiferromagnetic Chern insulating behavior, topological Mott physics, and several superconducting regimes (Hohenadler et al., 2011).

1. Canonical structure and principal generalizations

The canonical Kane–Mele–Hubbard model is defined on the honeycomb lattice with two sublattices, spinful fermions, nearest-neighbor hopping $t$, intrinsic spin–orbit coupling $\lambda$ on next-nearest-neighbor links, and onsite Hubbard interaction $U$ (Hohenadler et al., 2011). In the absence of Rashba coupling, $S^z$ is conserved and the intrinsic spin–orbit term preserves time-reversal symmetry while reducing the spin symmetry from SU(2) to U(1) (Lessnich et al., 2023). At half filling, the noninteracting limit is a quantum spin Hall or topological band insulator with a bulk gap $\Delta = 6\sqrt{3}\,\lambda$ and quantized spin Hall conductivity $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$ at $T=0$ (Lessnich et al., 2023).

The designation “generalized” is used in several distinct but related senses. One line of work generalizes the model by moving away from half filling and working at fixed particle number, for example at hole concentration $\delta=1/12$ on a three-leg honeycomb cylinder, thereby studying superconductivity on a doped $\nu_{ij}s^z$0 topological metal rather than on a half-filled topological insulator (Gupta et al., 2024). A second line adds explicit single-particle terms, such as an in-plane Zeeman field $\nu_{ij}s^z$1, which breaks time-reversal symmetry but preserves a mirror-inversion symmetry and produces crystalline and higher-order topological phases (Zhang et al., 2024). A third introduces a staggered sublattice potential $\nu_{ij}s^z$2, which competes with intrinsic spin–orbit coupling and, with interactions, supports an antiferromagnetic Chern insulator phase (Wang et al., 24 Apr 2026). Other extensions include long-range hopping $\nu_{ij}s^z$3 between sites separated by four lattice constants (Du et al., 2018), Rashba spin–orbit coupling $\nu_{ij}s^z$4 (Laubach et al., 2013), and a $\nu_{ij}s^z$5-flux threading each honeycomb plaquette, which doubles the unit cell and changes the per-spin Chern number to $\nu_{ij}s^z$6 (Bercx et al., 2014).

A further generalization replaces repulsive $\nu_{ij}s^z$7 by attractive onsite interaction. In that case, the model supports an edge superconducting state in which superconductivity nucleates at the helical boundary modes while the bulk remains insulating, followed at larger attraction by a bulk superconducting phase (Jie et al., 2012). Mean-field work with an added real next-nearest-neighbor hopping $\nu_{ij}s^z$8 shows that even this attractive branch is sensitive to generalized single-particle structure: in the nontrivial topological region, including $\nu_{ij}s^z$9 changes the phase diagram significantly and leads to inequivalent pairing amplitudes $t$0 on the two sublattices (Koinov, 2020).

2. Topological content, symmetry classes, and band-structure mechanisms

At half filling and without auxiliary perturbations, the noninteracting Kane–Mele model is a time-reversal-invariant topological insulator characterized by a $t$1 invariant and helical edge states (Gupta et al., 2024). In the interacting but symmetry-preserving regime, the topological band insulator at finite $t$2 is adiabatically connected to the noninteracting Kane–Mele state, so long as no bulk gap closes and no symmetry-breaking order intervenes (Hohenadler et al., 2011). Finite-temperature transport calculations reinforce this viewpoint: with full vertex corrections, the interacting spin Hall conductivity approaches $t$3 as $t$4 throughout the quantum spin Hall regime, even though antiferromagnetic fluctuations strongly renormalize the gap and the finite-$t$5 response (Lessnich et al., 2023).

Generalized terms alter topology through distinct mechanisms. A sublattice potential $t$6 competes directly with the intrinsic spin–orbit mass; in the noninteracting limit, $t$7 destroys the band inversion and yields a trivial band insulator, whereas interactions can produce a $t$8 antiferromagnetic Chern insulator along the QSH–CDW boundary (Wang et al., 24 Apr 2026). A long-range hopping $t$9 preserves time-reversal symmetry but modifies the off-diagonal kinetic structure; in the noninteracting generalized model, the topological transition occurs at $\lambda$0, independent of $\lambda$1, and separates the quantum spin Hall phase from a trivial band insulator (Du et al., 2018). In the $\lambda$2-flux Kane–Mele model, each spin sector is a Chern insulator with $\lambda$3, and a fine-tuned $\lambda$4 produces a quadratic band crossing point rather than a Dirac mass inversion (Bercx et al., 2014).

Rashba coupling introduces a different kind of topological restructuring. Because it fully breaks the U(1) spin symmetry preserved by the intrinsic Kane–Mele term, the standard spin-conserving topological classification no longer applies directly, yet the $\lambda$5 topological phase survives until the direct gap closes. In the noninteracting generalized model with finite $\lambda$6, a “weak topological semiconductor” regime appears for sufficiently large intrinsic spin–orbit coupling: the direct gap remains open at each momentum, edge states remain visible, but the global indirect gap vanishes due to band bending (Laubach et al., 2013). This suggests that the generalized Kane–Mele–Hubbard landscape is not exhausted by the simple dichotomy of topological insulator versus trivial insulator; direct-gap-only topological regimes and crystalline or flux-enabled topological structures occur naturally once additional one-body terms are admitted.

3. Correlation-driven bulk phases and phase transitions

For the half-filled repulsive model, the best-established bulk phases are the semimetal at $\lambda$7 and small $\lambda$8, the topological band insulator at finite $\lambda$9 and small $U$0, the antiferromagnetic Mott insulator at sufficiently large $U$1, and, in some parameter windows, a quantum spin liquid (Hohenadler et al., 2010). Projective quantum Monte Carlo finds that the topological insulator at finite $U$2 is adiabatically connected to the noninteracting state, that the magnetic transition at finite $U$3 is to an easy-plane antiferromagnet, and that the topological band insulator–to–antiferromagnetic Mott insulator transition is in the three-dimensional XY universality class (Hohenadler et al., 2011). The transverse structure factor rather than the $U$4-component orders because intrinsic spin–orbit coupling frustrates longitudinal antiferromagnetism and favors easy-plane ordering (Hohenadler et al., 2010).

The existence and extent of a spin-liquid regime depend sensitively on method and parameterization. Early quantum Monte Carlo work identified a quantum spin-liquid region at $U$5 between the semimetal and the antiferromagnetic Mott insulator and argued that it remains robust against weak spin–orbit coupling, with a gap-closing transition to the topological band insulator (Hohenadler et al., 2010). Refined projector QMC later reported that both the spin-liquid–to–topological-insulator and spin-liquid–to–Mott-insulator transitions are numerically consistent with continuity, while reaffirming the three-dimensional XY character of the topological-insulator–to–AFM transition (Hohenadler et al., 2011). In contrast, U(1) slave-boson mean field finds a small gapped spin-liquid window at half filling that becomes superconducting upon doping, but it also produces an unphysical weak-coupling superconducting phase and is explicitly presented as qualitatively rather than quantitatively reliable in the small-$U$6 regime (Wen et al., 2011). Strong-coupling Schwinger-boson and Schwinger-fermion analyses replace the original fermion problem by an anisotropic $U$7–$U$8 spin model and identify Néel, incommensurate Néel, and several candidate spin-liquid states; within that framework, increasing intrinsic spin–orbit coupling narrows the spin-liquid region, and only the chiral spin liquid is argued to be stable against gauge fluctuations (Vaezi et al., 2011).

Generalized perturbations generate additional correlated bulk phases. With a staggered sublattice potential, exact diagonalization finds four phases on the 12A cluster: QSH, CDW, SDW, and a $U$9 antiferromagnetic Chern insulator labeled “C1,” characterized by $S^z$0, a modified many-body Chern number $S^z$1, and a spin Chern number $S^z$2 (Wang et al., 24 Apr 2026). With an in-plane Zeeman field, projector QMC and mean-field calculations identify a higher-order topological insulating phase with mirror-protected corner states, an AFM Mott insulator, and spin-polarized trivial phases; for $S^z$3, the upper Zeeman field supporting corner states is $S^z$4 (Zhang et al., 2024). With $S^z$5-flux, the quadratic band crossing at $S^z$6 leads to a magnetic phase that extends to weak coupling because the quadratic touching has finite density of states and is unstable to interactions (Bercx et al., 2014). With long-range hopping $S^z$7, slave-rotor and Hartree–Fock treatments find a correlated QSH state below the Mott transition, a topological Mott insulator above it, and a spin-density-wave region at large $S^z$8, with the topological transition controlled by the renormalized condition $S^z$9 (Du et al., 2018).

4. Doping, superconductivity, and attractive-$\Delta = 6\sqrt{3}\,\lambda$0 variants

Doping the Kane–Mele–Hubbard model introduces a qualitatively distinct sector of generalized KMH physics. On a three-leg honeycomb cylinder at hole concentration $\Delta = 6\sqrt{3}\,\lambda$1, density-matrix renormalization group finds, at $\Delta = 6\sqrt{3}\,\lambda$2, a metallic regime for $\Delta = 6\sqrt{3}\,\lambda$3, a superconducting window for $\Delta = 6\sqrt{3}\,\lambda$4, and a stronger-coupling regime in which pairing correlations no longer clearly dominate over single-particle contributions (Gupta et al., 2024). The dominant pairing is spin-singlet and bond-centered,

$\Delta = 6\sqrt{3}\,\lambda$5

with equal-time correlations $\Delta = 6\sqrt{3}\,\lambda$6 (Gupta et al., 2024). At the representative point $\Delta = 6\sqrt{3}\,\lambda$7, the extrapolated pairing correlations decay as $\Delta = 6\sqrt{3}\,\lambda$8 with $\Delta = 6\sqrt{3}\,\lambda$9, while $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$0 with $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$1, indicating quasi-long-range superconducting order that dominates the single-particle sector on the cylinder (Gupta et al., 2024). A central result is that intrinsic spin–orbit coupling strongly lowers the critical interaction for superconductivity: $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$2, $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$3, with random-phase approximation calculations attributing this trend to enhanced susceptibility peaks and stronger paramagnon-driven pairing at larger $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$4 (Gupta et al., 2024).

A different doping-driven route emerges in slave-boson theory. There the half-filled intermediate-$\sigma_{xy}^{\text{SH}}=-2\,e^2/h$5 spin liquid becomes superconducting for any infinitesimal doping because the doping constraint forces holon or doublon condensation; the resulting superconducting order is spin-singlet and extended over same-sublattice and inter-sublattice channels, with an “optimal” doping where the pairing amplitudes are maximal (Wen et al., 2011). This suggests that within certain fractionalized mean-field descriptions, doping naturally converts a gapped spin liquid into a superconducting phase, although the weak-coupling sector of that approach is explicitly unreliable (Wen et al., 2011).

Attractive-$\sigma_{xy}^{\text{SH}}=-2\,e^2/h$6 models generalize the KMH framework in another direction. For the attractive Kane–Mele–Hubbard model at half filling, self-consistent mean-field theory finds a two-step sequence: as soon as the attractive interaction is turned on, an edge superconducting state appears in which only the helical edges are paired, while the bulk remains insulating; only beyond a critical $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$7 does the whole system become superconducting (Jie et al., 2012). The mechanism is the mismatch between the finite density of states of the one-dimensional helical edge modes and the gapped or Dirac-like bulk density of states (Jie et al., 2012). In the generalized attractive model with a real next-nearest-neighbor hopping $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$8, the nontrivial topological region of the phase diagram changes significantly, and the pairing amplitudes on sublattices $\sigma_{xy}^{\text{SH}}=-2\,e^2/h$9 and $T=0$0 become inequivalent, $T=0$1 (Koinov, 2020). Close to the superfluid transition, a Bethe–Salpeter analysis of the Goldstone mode finds only a $T=0$2 difference between the sound velocity obtained with full bubble-plus-ladder diagrams and that from the T-matrix approximation, indicating that near the phase boundary bubble contributions are small (Koinov, 2020).

5. Boundary, higher-order, and momentum-space phenomena

Generalized KMH models are especially rich at boundaries because modified topology often manifests first in edge, corner, or quasi-one-dimensional observables. In the $T=0$3-flux model, the spinful system at half filling is explicitly $T=0$4-trivial, yet each edge hosts two Kramers doublets, one crossing at $T=0$5 and one at $T=0$6. These helical edge states are protected at the single-particle level by translation symmetry along the edge, not by the bulk $T=0$7 index alone (Bercx et al., 2014). Bosonization predicts that at half filling inter-channel umklapp scattering is relevant and opens an interaction-induced edge gap; continuous-time QMC on a ribbon with interactions only on one edge confirms this prediction at strong coupling, showing a clear spectral gap in the edge spectral function and corresponding suppression of charge structure-factor cusps (Bercx et al., 2014).

With a Zeeman field, the edge/boundary sector becomes genuinely higher-order. On a diamond-shaped honeycomb lattice, the in-plane field induces a crystalline topological insulator with mirror-protected armchair edge states and a higher-order topological insulator with zero-energy corner states at two horizontal corners (Zhang et al., 2024). Projector QMC shows that these corner states survive weak and moderate Hubbard interactions and disappear when the system undergoes a Mott transition to an AFM insulator that breaks the protecting mirror-inversion symmetry (Zhang et al., 2024). The same study argues that increasing Hubbard interaction acts as an effective extra in-plane Zeeman field, $T=0$8, in the higher-order topological regime (Zhang et al., 2024).

Momentum-space diagnostics provide another boundary between phases. In quasi-one-dimensional zigzag ribbons studied with many-body configuration interaction, the spin gap $T=0$9 remains exactly zero at $\delta=1/12$0 across the explored $\delta=1/12$1 range, signaling persistent topological character in the spin sector, while the many-body charge gap $\delta=1/12$2 develops a new signature of the topological band-insulator–to–antiferromagnetic Mott-insulator transition: the momentum at which $\delta=1/12$3 is minimal shifts from the Brillouin-zone boundary $\delta=1/12$4 to the Dirac point as $\delta=1/12$5 increases or $\delta=1/12$6 decreases (Roy et al., 2024). This suggests a practical momentum-space criterion for interacting topological transitions that is complementary to single-particle gap-closing arguments and may generalize to other correlated topological systems (Roy et al., 2024). In the standard half-filled model, related edge-focused quantum Monte Carlo work showed that Hubbard interactions confined to a zigzag edge strongly suppress charge currents and enhance transverse spin correlations while leaving the single-particle helical edge signatures intact, thereby establishing the helical Luttinger liquid as the natural low-energy edge description of the interacting topological band-insulating phase (Hohenadler et al., 2010).

6. Methods, diagnostics, and open directions

The generalized Kane–Mele–Hubbard literature is methodologically heterogeneous because different generalizations privilege different tools. Sign-problem-free projector auxiliary-field quantum Monte Carlo at half filling is central for the canonical model and for some Zeeman or $\delta=1/12$7-flux variants without Rashba coupling, because time-reversal symmetry and the absence of spin mixing keep the fermion determinants nonnegative (Hohenadler et al., 2010). Exact diagonalization under twisted boundary conditions is particularly useful for many-body Chern numbers and fidelity susceptibility in sublattice-potential variants (Wang et al., 24 Apr 2026). Density-matrix renormalization group is effective on doped cylinders, where it can compare pairing, spin, density, and single-particle correlations with controlled bond-dimension extrapolations (Gupta et al., 2024). Variational cluster approach is suited to Rashba-coupled models where sign-problem-free QMC is unavailable, and it can resolve both bulk spectral gaps and edge-state dispersions in interacting topological regimes (Laubach et al., 2013). Two-particle self-consistent theory gives controlled weak-to-intermediate-coupling access to finite-temperature transport, including momentum- and frequency-dependent self-energies and Maki–Thompson-type vertex corrections in the spin Hall response (Lessnich et al., 2023). Slave-particle constructions and strong-coupling spin mappings remain useful for identifying possible spin liquids, topological Mott phases, and effective exchange anisotropies, but their quantitative phase boundaries are method dependent (Vaezi et al., 2011).

The principal diagnostics likewise vary by regime. Bulk topological phases are identified through adiabatic continuity, many-body Chern or spin Chern numbers under twisted boundary conditions, edge spectral functions, and, in inversion-symmetric cases, Green’s-function parity criteria at time-reversal invariant momenta (Wang et al., 24 Apr 2026). Correlated instabilities are diagnosed through structure factors such as $\delta=1/12$8, $\delta=1/12$9, charge-density-wave correlators, fidelity susceptibility, and scaling of spin or charge gaps (Hohenadler et al., 2011). Superconductivity on quasi-one-dimensional geometries requires comparison of pair correlations not only with spin and density correlations but also with $\nu_{ij}s^z$00, since Wick-factorizable metallic contributions can mimic pairing enhancement if the single-particle sector is not explicitly subdominant (Gupta et al., 2024). Higher-order topology is diagnosed through local density of states at $\nu_{ij}s^z$01, edge spectral functions, and the spatial distribution of corner weight (Zhang et al., 2024).

Several open directions recur across the generalized KMH program. The two-dimensional limit of the doped repulsive model remains unsettled because the strongest superconducting evidence so far is on narrow cylinders and fixed $\nu_{ij}s^z$02 (Gupta et al., 2024). The topological character of the superconducting states found in doped or attractive variants is usually not computed explicitly, so the relation between KMH-based pairing and time-reversal-invariant topological superconductivity remains an open problem (Gupta et al., 2024). Rashba-coupled interacting models need methods that treat full spin mixing without small-cluster artifacts, particularly in the strongly frustrated large-$\nu_{ij}s^z$03 sector where spiral order is plausible (Laubach et al., 2013). The antiferromagnetic Chern insulator found with a sublattice potential raises a broader question: which generalized KMH perturbations can stabilize time-reversal-breaking topological order without explicit magnetic fields, and how robust are such phases beyond small exact-diagonalization clusters (Wang et al., 24 Apr 2026). Finally, moiré materials such as twisted MoTe$\nu_{ij}s^z$04–WTe$\nu_{ij}s^z$05, whose bands can emulate Haldane- and Kane–Mele-type topology with strong correlations and tunable filling, provide an experimental setting in which several of these generalized KMH mechanisms—doping, topology, interaction-tuned superconductivity, and SOC control—may become directly relevant (Gupta et al., 2024).