Spin Kitaev Models

• Spin Kitaev models are quantum spin Hamiltonians characterized by bond-dependent anisotropic exchanges that produce emergent gauge fields and fractionalized excitations.
• They extend the exactly solvable spin-1/2 honeycomb case to higher-spin systems, one-dimensional chains, Z_n rotor models, and material-specific Hamiltonians.
• These models exhibit rich phase diagrams, topological order, and unconventional dynamic responses accessible via both synthetic realizations and advanced computational techniques.

Spin Kitaev models comprise a family of quantum spin Hamiltonians with strongly anisotropic, bond- or direction-dependent exchange interactions. The canonical case is the spin-$1/2$ Kitaev model on the honeycomb lattice, where the interaction component is selected by the bond orientation; this model is exactly solvable and realizes a quantum spin liquid with emergent $\mathbb{Z}_2$ gauge structure, spin fractionalization, topological degeneracy, and highly constrained correlation functions. The same organizing idea has since been extended to higher-spin honeycomb systems, one-dimensional compass-like chains, generalized $\mathbb{Z}_n$ rotor models, material-specific Kitaev-Heisenberg and $KJ\Gamma\Gamma'$ Hamiltonians, and synthetic platforms in which “spin Kitaev models” are engineered directly.

1. Canonical honeycomb construction

The standard spin-$1/2$ Kitaev Hamiltonian on the honeycomb lattice is

$$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$

Its defining feature is bond selectivity: on an $x$ bond only the $x$ spin components interact, and analogously for $y$ and $z$ bonds. This directional anisotropy is accompanied by a macroscopic set of commuting conserved quantities, most notably one plaquette operator for each hexagon. In Kitaev’s Majorana fermionization, each spin is represented by four Majorana operators, $\mathbb{Z}_2$0, and the Hamiltonian reduces, for fixed bond variables $\mathbb{Z}_2$1, to free Majorana fermions moving in a static $\mathbb{Z}_2$2 gauge background. The physical ground state lies in the zero-flux sector, and on the torus the model has a fourfold ground-state degeneracy associated with non-contractible Wilson loops (Mandal et al., 2020).

The exact solution yields a phase diagram with gapless and gapped regions. The gapless regime occurs inside the triangle defined by the inequalities $\mathbb{Z}_2$3 and cyclic permutations, where the matter sector has Dirac-point Majorana excitations; outside this region the spectrum is gapped. Local magnetic order is absent, magnetization vanishes, and the equal-time two-spin correlators are nonzero only for nearest-neighbor pairs whose bond type matches the spin component. Long-range structure instead appears in nonlocal string operators and in the topological loop algebra, reflecting long-range entanglement rather than symmetry breaking (Mandal et al., 2020).

The exact solvability has also been reformulated in broader dual frameworks. A higher-dimensional duality between staggered Majorana, or “dumbbell,” fermions and Pauli spins expresses Kitaev’s honeycomb solution as one special case in a class of soluble honeycomb spin models with bilinear nearest-neighbor couplings and external magnetic fields. In that formulation, each fixed $\mathbb{Z}_2$4 flux sector yields a single-particle fermionic Schrödinger problem, and the full spin spectrum is assembled from the union of gauge sectors (Banks, 13 Feb 2025).

2. Fractionalization, selection rules, and dynamical response

In the exact spin-$\mathbb{Z}_2$5 model, a local spin operator does not create a conventional magnon. In the Majorana representation,

$\mathbb{Z}_2$6

so a spin insertion changes both the matter sector and the adjacent plaquette fluxes. This yields the characteristic selection rules of the Kitaev spin liquid: off-diagonal correlators $\mathbb{Z}_2$7 with $\mathbb{Z}_2$8 vanish, and equal-time two-spin correlations beyond nearest neighbors vanish as well. A Green’s-function equation-of-motion treatment that preserves SU(2) symmetry reproduces these constraints and gives a nearest-neighbor correlator $\mathbb{Z}_2$9 as $\mathbb{Z}_n$0, close to the exact value $\mathbb{Z}_n$1, while also recovering the correct high-temperature expansion $\mathbb{Z}_n$2 (Takegami et al., 2024).

These selection rules do not preclude nontrivial dynamics. In the Kitaev-Heisenberg model on finite honeycomb clusters with open boundaries, an AC magnetic field applied locally at one edge can induce a spin polarization at the opposite edge,

$\mathbb{Z}_n$3

even when the static edge-to-edge spin correlations are vanishingly small. The associated interedge spin resonance is phase-specific. In magnetically ordered regimes it has the character of conventional magnetic resonance, whereas in the Kitaev quantum spin liquid it appears only in a geometry-selected spin component—$\mathbb{Z}_n$4 for the armchair edge and $\mathbb{Z}_n$5 for the zigzag edge—and its peak tracks the input frequency over a broad range. The dominant response is governed by itinerant Majorana fermions with a broad continuum excitation spectrum, while weakly induced orthogonal components show an almost input-frequency-independent feature near the excitation gap of localized fluxes. In the pure Kitaev limit the interedge resonance vanishes because accidental degeneracy and flux-sector selection rules suppress the required dynamical spin correlations; weak Heisenberg exchange lifts this obstruction and activates the effect (Misawa et al., 2023).

A recurring implication is that static short-range spin correlations and long-range dynamical transport are not mutually exclusive in Kitaev systems. The data suggest that spin transport experiments can selectively access matter-Majorana propagation and flux thresholds even when equal-time spin observables appear featureless.

3. Higher-spin honeycomb Kitaev systems

For $\mathbb{Z}_n$6 on the honeycomb lattice, the Kitaev interaction remains bond dependent but exact solvability is lost. The $\mathbb{Z}_n$7 model,

$\mathbb{Z}_n$8

nevertheless retains conserved plaquette operators

$\mathbb{Z}_n$9

and Wilson-loop operators that indicate a $KJ\Gamma\Gamma'$0 gauge structure. DMRG on cylindrical geometries with up to 250 sites concludes that the ground state is a quantum spin liquid for both ferromagnetic and antiferromagnetic Kitaev couplings. The excitation gap is bounded above by $KJ\Gamma\Gamma'$1 on the largest accessible $KJ\Gamma\Gamma'$2 antiferromagnetic system, but the thermodynamic limit is not settled conclusively. In a $KJ\Gamma\Gamma'$3 field, the ferromagnetic model exhibits a direct transition to a polarized phase at $KJ\Gamma\Gamma'$4, while the antiferromagnetic model shows two transitions, at $KJ\Gamma\Gamma'$5 and $KJ\Gamma\Gamma'$6, with an intermediate likely gapless phase (Khait et al., 2020).

Tensor-network constructions sharpen the topological interpretation of the $KJ\Gamma\Gamma'$7 case. Loop-gas and string-gas tensor-network states written directly in terms of physical spin operators realize a vortex-free $KJ\Gamma\Gamma'$8 structure, provide access to all four global flux sectors on the torus, and yield minimally entangled states whose topological entanglement properties are consistent with Abelian $KJ\Gamma\Gamma'$9 topological order. The optimized states have finite correlation length, and the results suggest that the $1/2$0 ground state is a gapped quantum spin liquid with Abelian quasiparticles (Lee et al., 2019). Taken together with the DMRG results, this supports a $1/2$1 spin-liquid interpretation while leaving the precise status of the excitation gap as a nontrivial numerical issue.

For general spin $1/2$2, the Kitaev-Heisenberg model has been studied by extending pseudofermion functional renormalization group methods to arbitrary $1/2$3. The resulting phase diagrams retain the four ordered phases—Néel, zigzag, ferromagnetic, and stripy—and also exhibit quantum spin liquid regimes near both the antiferromagnetic and ferromagnetic Kitaev points. The crucial trend is a rapid collapse of the spin-liquid windows with increasing $1/2$4: the QSL regions are already very narrow at $1/2$5 and become vanishingly small for $1/2$6, while the ordered phase boundaries remain almost unchanged up to the classical limit (Fukui et al., 2022).

A microscopic route to higher-spin Kitaev exchange has also been derived. In two-dimensional edge-shared octahedral systems with strong Hund’s coupling on $1/2$7 transition-metal cations and strong spin-orbit coupling on ligand anions, superexchange generates an $1/2$8 Kitaev interaction together with a ferromagnetic Heisenberg term; direct $1/2$9-$$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$0 exchange adds an antiferromagnetic Heisenberg contribution, so the net Heisenberg coupling can be small relative to the Kitaev scale. Exact diagonalization then finds a finite $$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$1 spin-liquid regime near $$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$2. Proposed materials include single-layer NiI$$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$3, layered honeycomb compounds $$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$4Ni$$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$5 with $$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$6 and $$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$7, and, by extension to $$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$8, single-layer CrI$$H= \sum_{\langle ij \rangle_x} J_x S^x_i S^x_j + \sum_{\langle ij \rangle_y} J_y S^y_i S^y_j + \sum_{\langle ij \rangle_z} J_z S^z_i S^z_j.$$9 (Stavropoulos et al., 2019).

4. One-dimensional, anisotropic, and modified spin-$x$0 variants

A distinct branch of spin Kitaev models arises in one dimension. One example is the ring Hamiltonian

$x$1

defined for arbitrary spin $x$2. Each bond carries a conserved $x$3 operator $x$4 with $x$5. Integer and half-odd-integer spins are qualitatively different. For integer $x$6, the Hilbert space splits into $x$7 commuting sectors; for $x$8, the ground state lies in the sector with all $x$9, whose dimension grows asymptotically as $x$0 with $x$1. In that sector the model maps to an interacting spin-$x$2 chain with nearest-neighbor exclusion, exact diagonalization indicates a finite thermodynamic gap, and the first excitation lies in a different $x$3 sector with extrapolated gap $x$4. Adding a term $x$5 produces a gapless phase for $x$6 (Sen et al., 2010).

Recent spin-basis constructions for the one-dimensional spin-$x$7 Kitaev model make the integer–half-integer contrast explicit at the wavefunction level. For half-integer $x$8 and periodic boundaries, the ground states are superpositions of bond states with an even number of “triplets” on a background of “singlets,” leading to exponential degeneracy and a topological parity constraint. For integer $x$9, a unique periodic-boundary ground state is obtained, composed purely of triplet-like bond states. These wavefunctions are not exact, but they capture the dominant weight of the ground states and agree well with exact diagonalization and, for $y$0, Jordan-Wigner spectra (Raja et al., 26 Sep 2025).

On the two-dimensional honeycomb lattice, an anisotropic limit $y$1 reveals another parity effect. Degenerate perturbation theory gives different effective conserved plaquette operators for half-integer and integer spin. For half-integer $y$2, the low-energy Hamiltonian is the toric code,

$y$3

where $y$4 has the same four-pseudospin structure as in the $y$5 case. For integer $y$6, the effective model reduces instead to noninteracting pseudospins in a uniform field,

$y$7

so quantum fluctuations are quenched and the low-energy theory is effectively classical. Boundary geometry modifies these conclusions at finite order: armchair and zigzag edges behave differently when bulk perturbative terms cancel (Minakawa et al., 2018).

A related but distinct construction is the modified spin-$y$8 Kitaev model of Baskaran, Sen, and Shankar, built from commuting or anticommuting operators $y$9:

$z$0

For half-odd-integer spins the $z$1 obey Pauli-like algebra and the model maps, up to degeneracy, onto the spin-$z$2 Kitaev model. For integer spins the $z$3 commute, the eigenstates can be chosen as product states, and the problem becomes classical. Transfer-matrix, high-temperature-series, and Monte Carlo studies of the integer-spin cases found large residual entropies and several distinct low-temperature regimes: finite correlation length, critical behavior, or exponentially diverging correlation length. The plaquette flux $z$4 tends to $z$5 as $z$6 in all cases except the $z$7 antiferromagnet, where the mean flux remains zero (Bradley et al., 2020).

5. Generalized interactions, strain, and non-Abelian extensions

The original honeycomb model is only the simplest member of a larger family. A broad class of generalized Kitaev models replaces Pauli spins by $z$8-state $z$9 rotor variables with bond-dependent two-site operators,

$\mathbb{Z}_2$00

where $\mathbb{Z}_2$01 and $\mathbb{Z}_2$02. These models retain an infinite number of conserved plaquette operators, non-contractible loop operators on the torus, and protected topological degeneracy. In the “slave-genon” formulation, the physical Hilbert space is embedded into an enlarged space of non-Abelian twist defects, or genons, so that the spin model becomes exactly a model of genons coupled to a discrete gauge field. The $\mathbb{Z}_2$03 generalization is especially significant: in appropriate limits it maps to coupled parafermion chains and can realize a non-Abelian topological phase with chiral parafermion edge states; the anyon content includes the Fibonacci anyon (Barkeshli et al., 2014).

Material-oriented generalizations abandon exact solvability in favor of microscopic realism. The $\mathbb{Z}_2$04 Hamiltonian has long been used as a central model for candidate Kitaev materials, but realistic crystals often break the ideal $\mathbb{Z}_2$05 symmetry assumed in that form. The $\mathbb{Z}_2$06-$\mathbb{Z}_2$07 model extends it to arbitrary homogeneous lattice deformations through

$\mathbb{Z}_2$08

where $\mathbb{Z}_2$09 contains strain-renormalized $\mathbb{Z}_2$10, $\mathbb{Z}_2$11, $\mathbb{Z}_2$12, and $\mathbb{Z}_2$13 terms, and $\mathbb{Z}_2$14 contains new symmetry-allowed two-spin exchanges generated by reduced symmetry. A combined symmetry-analysis, DFT+Wannier, and strong-coupling derivation shows that in strained $\mathbb{Z}_2$15-RuCl$\mathbb{Z}_2$16 new exchange channels can become comparable in magnitude to the unstrained couplings already at $\mathbb{Z}_2$17 strain. The same framework predicts strain-driven transitions between zigzag order and the Kitaev quantum spin liquid, together with strain-controlled changes in the sign of the Majorana mass and the topological invariant $\mathbb{Z}_2$18, potentially visible as thermal Hall sign reversals. The symmetry analysis applies both to $\mathbb{Z}_2$19 systems such as $\mathbb{Z}_2$20-RuCl$\mathbb{Z}_2$21 and to $\mathbb{Z}_2$22 cobalt-based compounds such as Na$\mathbb{Z}_2$23Co$\mathbb{Z}_2$24TeO$\mathbb{Z}_2$25 and Na$\mathbb{Z}_2$26Co$\mathbb{Z}_2$27SbO$\mathbb{Z}_2$28 (Noh et al., 29 May 2025).

These generalized constructions clarify a central point: “Kitaev” does not designate a single exchange tensor, but a structural principle in which local degrees of freedom, gauge constraints, and bond-selective interactions cooperate to produce conserved fluxes, unusual low-energy matter sectors, or symmetry-controlled deformations thereof.

6. Synthetic realizations and computational approaches

Several platforms aim to realize spin Kitaev physics without relying on naturally occurring exchange paths. In ultracold polar molecules pinned in an optical lattice, effective spins are encoded in microwave-dressed rotational states of rigid rotors. A DC electric field sets the quantization axis, and dipole-dipole interactions projected into the dressed-state manifold generate highly tunable spin couplings. With sufficient microwave control, the interaction between two effective spins can be decomposed into five independently controllable rank-2 spherical-harmonic components $\mathbb{Z}_2$29, enabling direct engineering of the bond-selective honeycomb Hamiltonian. The same scheme extends formally to arbitrary spin $\mathbb{Z}_2$30 and to models with $\mathbb{Z}_2$31 symmetry (Gorshkov et al., 2013).

Topological nanowire networks provide another route. Arrays of superconducting topological nanowires realize low-energy tight-binding models of physical Majorana zero modes, and these are mathematically equivalent to the Majorana representation of exactly solvable Kitaev spin models. Junctions of three wires arranged on a decorated honeycomb lattice implement the Yao-Kivelson model, producing collective phases with Chern numbers $\mathbb{Z}_2$32. In that language, local sign changes in effective Majorana hopping correspond to vortex-like excitations, and disorder or phase fluctuations can be viewed as mechanisms for vortex proliferation (Kells et al., 2013).

More recent synthetic work uses the phrase “spin Kitaev models” in a narrower sense to denote a new class of engineered one-dimensional quantum spin chains with both flip-flop and flip-flip/flop-flop exchange,

$\mathbb{Z}_2$33

These models are generated from ordinary spin exchange by locally selective Floquet pulses. Their key dynamical signature is polarization-dependent chiral transport: in zero magnetic field, excitations polarized along $\mathbb{Z}_2$34 and $\mathbb{Z}_2$35 propagate chirally in opposite directions. In the large-$\mathbb{Z}_2$36 limit the model maps to a nonlinear Hatano-Nelson problem, and the finite-$\mathbb{Z}_2$37 equations support stable chiral solitons; with a magnetic field, oppositely polarized solitons bind into a chiral solitonic molecule whose propagation direction depends on orientation (Lv et al., 23 Aug 2025). This usage is conceptually distinct from the original honeycomb Kitaev model, even though both inherit strongly anisotropic spin-exchange structure and unconventional emergent transport.

Because most generalized spin Kitaev models are nonintegrable, numerical methods are central. Chebyshev polynomial-based iterative schemes—CPGF, FTCP, and hybrid Lanczos-Chebyshev—have been used to compute thermodynamics, phase boundaries, and dynamical susceptibilities of Kitaev-Heisenberg and Kitaev-Ising models with algorithmic cost linear in Hilbert-space dimension and truncation order; these methods reproduce the characteristic two-peak specific heat structure and identify dynamical fingerprints of phase transitions (Brito et al., 2021). Variational quantum eigensolver approaches based on the fermionic formulation exploit the free-fermion structure in the solvable limit, can represent the exact ground state there, and in fixed gauge sectors reduce the required qubit count by a factor of two relative to direct spin encodings, with applications to the honeycomb and square-octagon lattices and to non-Abelian anyon sectors (Jahin et al., 2022).

Spin Kitaev models therefore form a layered subject. At one level they denote the exactly solvable spin-$\mathbb{Z}_2$38 honeycomb model and its immediate perturbations; at another they include higher-spin, one-dimensional, $\mathbb{Z}_2$39, and material-specific extensions that preserve only parts of the original structure; and in synthetic quantum matter they may even name newly engineered anisotropic spin chains whose transport phenomena have no direct honeycomb analogue. Across these variants, the recurring themes are bond-selective exchange, local conserved flux variables or their descendants, and low-energy behavior that is often more naturally described in terms of emergent fermions, gauge sectors, or topological defects than in terms of conventional spin waves.