Chern Antiferromagnet (CAF): Quantum Anomalous Hall

Updated 7 July 2026

CAF is a two-dimensional antiferromagnetic insulator with zero net magnetization and a nonzero integer Chern number, leading to quantized Hall conductance.
CAF emerges from spontaneous time-reversal symmetry breaking through antiferromagnetic order combined with spin–orbit coupling and symmetry lowering, resulting in uncompensated Berry curvature.
Experimental diagnostics of CAF include quantized edge transport, chiral edge modes, and techniques like Kerr rotation and ARPES to probe gap openings and magnetic order.

Chern antiferromagnet, often used synonymously with antiferromagnetic Chern insulator (AFCI), denotes a two-dimensional antiferromagnetic insulator with vanishing net magnetization and a nonzero integer Chern number, so that it exhibits the quantum anomalous Hall effect at zero external magnetic field with $\sigma_{xy}=C e^2/h$ . In the canonical interaction-driven realization, spontaneous time-reversal breaking by antiferromagnetic order combines with spin–orbit coupling and symmetry lowering such that the integrated Berry curvature need not vanish even though the magnetic unit cell carries zero total spin (Jiang et al., 2017, Ebrahimkhas et al., 2022).

1. Definition and terminological scope

In the strict band-topological sense, a Chern antiferromagnet is an insulating antiferromagnetic phase with $C\neq 0$ and zero net magnetization. Its defining transport signature is quantized Hall conductance without an external magnetic field, while its magnetic signature is strictly staggered order rather than ferromagnetic polarization (Jiang et al., 2017). This distinguishes it from a quantum spin Hall insulator, for which $C=0$ but opposite spin sectors carry opposite topological indices, and from a trivial antiferromagnetic insulator, for which time-reversal symmetry is broken yet the total Chern number remains zero (Jiang et al., 2017, Ebrahimkhas et al., 2022).

The acronym “CAF” is not unique in the literature. In work on frustrated square-lattice magnets it often denotes a collinear or columnar stripe antiferromagnet rather than a Chern insulator; in graphene it denotes a canted antiferromagnet; and in some metallic transport studies it denotes a compensated antiferromagnet. A separate field-theoretic usage refers to easy-plane antiferromagnets with an emergent Chern–Simons term. Unless otherwise specified, the present usage concerns the antiferromagnetic Chern insulator.

Usage of “CAF”	Meaning	Representative source
Chern antiferromagnet / AFCI	AF insulator with $C\neq 0$ and QAHE	(Jiang et al., 2017)
Collinear or columnar AF	Stripe order with $\mathbf Q=(\pi,0)$ or $(0,\pi)$	(Mabelini et al., 2012, Takeda et al., 2020)
Compensated AF	Nearly zero-net-moment metal with spontaneous Hall/Nernst response	(Khanh et al., 2024)
Canted AF	Quantum Hall antiferromagnet in graphene at $\nu=0$	(Stepanov et al., 2018)
Chern–Simons AF	Easy-plane CP $^1$ antiferromagnet with CS term	(Shyta et al., 2020)

A recurrent misconception is that nonzero Chern number in a magnetic insulator requires ferromagnetism. The AFCI literature explicitly disproves that equivalence: a staggered antiferromagnetic state can generate $C\neq 0$ provided no antiunitary or crystalline symmetry enforces Berry-curvature cancellation (Jiang et al., 2017, Ebrahimkhas et al., 2022).

2. Symmetry logic and topological mechanism

The canonical mechanism identified for interaction-driven AFCIs has three ingredients: spontaneous breaking of time-reversal symmetry $\mathcal T$ by antiferromagnetic order, strong spin–orbit interaction with a conserved spin axis, and the absence of inversion symmetry or, more generally, the absence of any space-group operation that compensates the action of $C\neq 0$ 0 on the electronic state (Jiang et al., 2017, Ebrahimkhas et al., 2022). In non-centrosymmetric honeycomb systems, the antiunitary symmetry $C\neq 0$ 1 is absent, so spin- and valley-locked Weyl points can appear on the boundary between a quantum spin Hall phase and a trivial band insulator. Once easy-axis antiferromagnetic order develops, it contributes a $C\neq 0$ 2-breaking mass $C\neq 0$ 3 that gaps those Weyl points. Because inversion is already broken, no residual combined symmetry forces the total Chern number to vanish, and the resulting antiferromagnetic insulator can have $C\neq 0$ 4 despite zero net magnetization (Jiang et al., 2017).

Near the valleys $C\neq 0$ 5 and $C\neq 0$ 6, the honeycomb theory is expressed as

$C\neq 0$ 7

with

$C\neq 0$ 8

Each massive Dirac cone contributes

$C\neq 0$ 9

so that

$C=0$ 0

In the easy-axis state, one spin–valley sector remains topological while the other becomes trivial, giving $C=0$ 1 near the noninteracting critical line $C=0$ 2. By contrast, an easy-plane collinear antiferromagnet does not generate the $C=0$ 3 mass and remains topologically trivial with $C=0$ 4 (Jiang et al., 2017).

The same general principle survives beyond non-centrosymmetric honeycomb models. In the centrosymmetric Harper–Hofstadter–Hubbard construction, an AFCI appears when a checkerboard potential prevents any space-group operation from restoring time reversal in the easy-axis antiferromagnetic state (Ebrahimkhas et al., 2022). The paper formulates the criterion directly: an AFCI can exist if the effect of the time-reversal transformation on the electronic state cannot be compensated by a space-group operation. This is why centrosymmetry itself does not preclude an AFCI, whereas combined symmetries such as $C=0$ 5, glide, screw, or analogous antiunitary operations generally do (Ebrahimkhas et al., 2022).

A distinct symmetry program appears in the magnetic-layer-point-group classification of antiferromagnetic Chern insulators. There the relevant question is which magnetic layer point groups connect antiferromagnetic sublattices while avoiding symmetries that reverse $C=0$ 6 throughout the Brillouin zone. In that setting the forbidden set is

$C=0$ 7

and the allowed antiferromagnetic Chern insulators require in-plane magnetic configurations (Liu et al., 2022). This does not contradict the easy-axis honeycomb and square-lattice AFCIs; it instead reflects a different symmetry setting, namely antiferromagnets whose sublattices are connected by specific magnetic layer point-group operations.

3. Canonical Hamiltonians and theoretical formulations

The archetypal microscopic model is the Kane–Mele honeycomb lattice with an inversion-breaking ionic potential and onsite Hubbard interaction,

$C=0$ 8

with $C=0$ 9 and $C\neq 0$ 0 in the paper’s notation (Jiang et al., 2017). Mean-field decoupling yields a Hartree renormalization of the inversion-breaking mass,

$C\neq 0$ 1

and an exchange-generated antiferromagnetic mass $C\neq 0$ 2 for easy-axis order. This furnishes the minimal interaction-driven AFCI mechanism.

A second important family replaces the Hubbard-driven magnetic order by exchange to localized moments. In the Kane–Mele–Kondo model,

$C\neq 0$ 3

a staggered potential $C\neq 0$ 4 and local exchange produce a spin-dependent effective sublattice potential

$C\neq 0$ 5

in dynamical mean-field theory (Hafez-Torbati, 2024). In heavy-fermion formulations this is the relevant quantity for the interacting topological Hamiltonian, while in the strong-coupling Hund-coupled variant it reduces approximately to $C\neq 0$ 6 and the valley masses become

$C\neq 0$ 7

The AFCI criterion is then spin-selective:

$C\neq 0$ 8

for one spin component but not the other (Hafez-Torbati, 2024, Hafez-Torbati et al., 2024).

A centrosymmetric route is provided by the time-reversal-invariant Harper–Hofstadter–Hubbard model with spin-dependent flux, next-nearest-neighbor spin–orbit coupling, checkerboard potential, and Hubbard repulsion. At half filling and for $C\neq 0$ 9, the topological transition in the easy-axis phase is controlled by the renormalized staggered potential and occurs at $\mathbf Q=(\pi,0)$ 0 in the four-band model (Ebrahimkhas et al., 2022). This construction shows explicitly that AFCIs are not confined to non-centrosymmetric lattices.

Not all CAFs are SOC-driven. In the kagome $\mathbf Q=(\pi,0)$ 1-electron model with canted noncollinear $\mathbf Q=(\pi,0)$ 2 order, no explicit spin–orbit coupling is included. There the crucial object is the scalar spin chirality

$\mathbf Q=(\pi,0)$ 3

which vanishes for coplanar order and becomes finite for noncoplanar canting. The noncoplanar texture generates intrinsic Berry curvature and quantum anomalous Hall gaps, yielding a record plateau $\mathbf Q=(\pi,0)$ 4 in the isotropic five-orbital limit (Ahmed et al., 13 Sep 2025).

Across these models, the interacting topological invariant is computed either from the full Berry curvature of the mean-field or topological Hamiltonian,

$\mathbf Q=(\pi,0)$ 5

or, in many-body exact diagonalization, from twisted-boundary-condition Berry curvature in flux space (Jiang et al., 2017, Wang et al., 24 Apr 2026).

4. Phase diagrams, bifurcations, and topological transitions

The honeycomb Kane–Mele–Hubbard phase diagram establishes the standard AFCI sequence. At fixed spin–orbit coupling one finds a quantum spin Hall insulator at small $\mathbf Q=(\pi,0)$ 6 and $\mathbf Q=(\pi,0)$ 7, a trivial band insulator at large $\mathbf Q=(\pi,0)$ 8, a $\mathbf Q=(\pi,0)$ 9 easy-axis AFCI near and beyond the Weyl boundary, and a topologically trivial easy-plane antiferromagnetic insulator at large $(0,\pi)$ 0 and smaller $(0,\pi)$ 1 (Jiang et al., 2017). The noninteracting critical line is

$(0,\pi)$ 2

For $(0,\pi)$ 3 and half filling, this Weyl line terminates at a tricritical point $(0,\pi)$ 4; beyond it the boundary bifurcates and creates an intervening AFCI region. At $(0,\pi)$ 5, the representative mean-field transition points are AFI–AFCI at $(0,\pi)$ 6 and AFCI–BI at $(0,\pi)$ 7 (Jiang et al., 2017).

Spin anisotropy produces a genuine topological spin-flop. In the same honeycomb model, random-phase-approximation susceptibilities yield an anisotropy crossover at $(0,\pi)$ 8: $(0,\pi)$ 9 for $\nu=0$ 0 and $\nu=0$ 1 for $\nu=0$ 2. The resulting transition from the $\nu=0$ 3-axis AFCI to the $\nu=0$ 4-plane AFI is first order because charge fluctuations are involved (Jiang et al., 2017).

Exact diagonalization of the Kane–Mele–Hubbard model with staggered potential provides independent many-body evidence. On the 12A honeycomb cluster at $\nu=0$ 5, the phase diagram contains CDW, QSH, SDW, and a $\nu=0$ 6 “C1” phase running along the QSH–CDW boundary (Wang et al., 24 Apr 2026). At fixed $\nu=0$ 7, the C1 window is $\nu=0$ 8, accompanied by enhanced $\nu=0$ 9 relative to $^1$ 0. In that region the standard lattice-gauge Chern-number algorithm fails because adiabatic continuity in twist-angle space breaks down: paired singularity-like values of the lattice field strength appear with magnitude near $^1$ 1, and true level crossings occur at isolated twist angles. A modified scheme that swaps the ground state with the first excited state inside a small twist-space patch restores a smooth branch and yields a robust $^1$ 2 (Wang et al., 24 Apr 2026). The same study explicitly notes that larger-scale and complementary methods remain desirable for the thermodynamic-limit robustness of the phase.

The centrosymmetric square-lattice model displays a closely analogous structure. At $^1$ 3, $^1$ 4, and $^1$ 5, dynamical mean-field theory finds QSHI, BI, easy-plane AF Mott insulator, and a $^1$ 6-axis AFCI between BI and AF Mott regimes for sufficiently large checkerboard potential (Ebrahimkhas et al., 2022). For $^1$ 7, the AFCI occurs for roughly

$^1$ 8

The paper emphasizes the qualitative similarity to the non-centrosymmetric Kane–Mele–Hubbard result and states that this close similarity suggests the AFCI as a generic consequence of spin–orbit coupling and strong electronic correlation beyond a specific model or lattice structure (Ebrahimkhas et al., 2022).

In the heavy-fermion Kane–Mele–Kondo model, the phase sequence depends strongly on the alternating sublattice potential. For $^1$ 9 and $C\neq 0$ 0, the $C\neq 0$ 1-AFQSHI changes to $C\neq 0$ 2-AFCI at $C\neq 0$ 3, then undergoes a spin-flop to $C\neq 0$ 4-AFI at $C\neq 0$ 5, followed by $C\neq 0$ 6-AFI $C\neq 0$ 7 KI at $C\neq 0$ 8. For $C\neq 0$ 9, the sequence is trivial $\mathcal T$ 0-AFI $\mathcal T$ 1-AFCI at $\mathcal T$ 2, then first-order $\mathcal T$ 3-AFCI $\mathcal T$ 4 KI at $\mathcal T$ 5 (Hafez-Torbati, 2024). In the related heavy transition-metal model with ferromagnetic Hund coupling, the AFCI is stabilized near $\mathcal T$ 6 once $\mathcal T$ 7 (Hafez-Torbati et al., 2024).

5. Edge physics, Hall response, and experimental diagnostics

The immediate consequence of $\mathcal T$ 8 is the presence of chiral edge transport. In the honeycomb AFCI with $\mathcal T$ 9, each boundary carries a single chiral edge mode and the Hall conductance is quantized at

$C\neq 0$ 00

without external magnetic field, despite zero bulk magnetization (Jiang et al., 2017). In the transition-metal and kagome generalizations, the number of co-propagating edge modes equals $C\neq 0$ 01, so larger Chern numbers imply multiple chiral channels (Hafez-Torbati et al., 2024, Ahmed et al., 13 Sep 2025). When Rashba spin–orbit coupling is introduced in the honeycomb model, the AFCI is not eliminated, but $C\neq 0$ 02 is no longer conserved and small ferrimagnetic moments can appear (Jiang et al., 2017).

The most direct experimental diagnostics follow from this conjunction of topology and staggered magnetism. The proposed signatures include quantized Hall transport at zero field coexisting with zero bulk magnetization, chiral edge conduction in nonlocal or scanning-probe measurements, Kerr or Faraday rotation as probes of broken time reversal, neutron or resonant x-ray scattering to confirm antiferromagnetic order, and ARPES or STM to resolve gap opening and valley-resolved band inversions (Jiang et al., 2017). The simultaneous observation of quantized $C\neq 0$ 03 and vanishing net $C\neq 0$ 04 is the essential hallmark.

A concrete first-principles candidate is monolayer NiRuCl $C\neq 0$ 05. In that system the antiferromagnetic phase is lower in energy than the ferromagnetic phase by $C\neq 0$ 06 meV per unit cell, and under $C\neq 0$ 07 compressive strain with spin–orbit coupling the system develops a global gap of approximately $C\neq 0$ 08 meV, a quantized Hall plateau over an energy window of about $C\neq 0$ 09 meV, and total Chern number $C\neq 0$ 10 with two chiral edge channels in a zigzag nanoribbon calculation (Zhou et al., 2016). An Ising-model Monte Carlo estimate gives $C\neq 0$ 11 K. The microscopic origin is sublattice inequivalence of Ni and Ru together with broken inversion symmetry and strong Ru $C\neq 0$ 12 spin–orbit coupling, which prevent Berry-curvature cancellation even though the total magnetic moment vanishes (Zhou et al., 2016).

The most aggressive gap estimates come from heavy transition-metal AFCIs. In the Kane–Mele–Kondo analysis with ferromagnetic Hund coupling, the AFCI charge gap grows rapidly with spin–orbit coupling according to

$C\neq 0$ 13

for $C\neq 0$ 14, and then saturates near $C\neq 0$ 15. The study states that heavy transition-metal compounds can realize an AFCI with a charge gap as large as $C\neq 0$ 16 meV, and it uses this framework to rationalize previously reported collinear AFCIs in strained monolayer CrO and MoO with charge gaps of $C\neq 0$ 17 meV and $C\neq 0$ 18 meV, respectively (Hafez-Torbati et al., 2024). By contrast, the heavy-fermion Kane–Mele–Kondo model yields smaller AFCI gaps than its transition-metal analogue (Hafez-Torbati, 2024).

The AFCI concept has expanded beyond the original easy-axis honeycomb mechanism. A symmetry-based materials search for antiferromagnetic Chern insulators identified two-dimensional in-plane antiferromagnets whose magnetic layer point groups connect AF sublattices without forcing $C\neq 0$ 19 to cancel. First-principles examples are monolayer RbCr $C\neq 0$ 20S $C\neq 0$ 21, a collinear in-plane AFM Chern insulator with $C\neq 0$ 22 and a nontrivial indirect gap $C\neq 0$ 23 meV above $C\neq 0$ 24, and bilayer Mn $C\neq 0$ 25Sn, a coplanar noncollinear AFM Chern insulator with $C\neq 0$ 26 and a $C\neq 0$ 27 meV gap roughly $C\neq 0$ 28 eV below $C\neq 0$ 29 (Liu et al., 2022). In bilayer Mn $C\neq 0$ 30Sn $C\neq 0$ 31-4.5^{\circ $C\neq 0$ 32-8.9^{\circ $C\neq 0$ 33C=-3 $</sup></sup> (<a href="/papers/2210.16507" title="" rel="nofollow" data-turbo="false" class="assistant-link" x-data x-tooltip.raw="">Liu et al., 2022</a>).</p> <p>The SOC-free kagome realization extends the notion still further. In the five-orbital kagome model with canted noncoplanar$ C\neq 0$34 order, isotropic Slater–Koster hopping $C\neq 0$35 eV, degenerate onsite energies, and moment parameters $C\neq 0$36 eV and $C\neq 0$37, nontrivial gaps of $C\neq 0$38–$C\neq 0$39 meV appear and the Chern number reaches $C\neq 0$40 (Ahmed et al., 13 Sep 2025). The same work shows that splitting the onsite energies resolves the fivefold plateau into lower plateaus: a “1–2–2” series with $C\neq 0$41 and $C\neq 0$42, and a “3–2” series with $C\neq 0$43 and $C\neq 0$44 (Ahmed et al., 13 Sep 2025). This is a different microscopic route from the easy-axis Kane–Mele mechanism, because the topological mass arises from noncoplanar scalar chirality rather than explicit spin–orbit coupling.}}

At the same time, several neighboring literatures use “Chern antiferromagnet” in a broader sense than a quantized 2D insulator. CoNb$C\neq 0$45S$C\neq 0$46 is described as a compensated antiferromagnet with near-zero net magnetization and large spontaneous Hall and Nernst responses generated by slightly gapped nodal planes at $C\neq 0$47; it is therefore a metallic Berry-curvature antiferromagnet rather than an AFCI in the strict quantized-Hall sense (Khanh et al., 2024). A moiré-motivated sign-problem-free QMC model uses “CAF” for an insulating state at $C\neq 0$48 with intra-valley ferromagnetic coherence and inter-valley antiferromagnetism in valley-contrasting Chern bands. That state preserves global time-reversal symmetry, has vanishing net anomalous Hall conductivity,

$C\neq 0$49

but carries a nonzero valley Hall response,

$C\neq 0$50

and upon doping develops superconductivity mediated by spin-polaron or skyrmion excitations (Wang et al., 30 Jul 2025). These usages mark a genuine expansion of the term beyond the original AFCI definition.

In the strictest usage, however, the Chern antiferromagnet remains the antiferromagnetic Chern insulator: a zero-net-moment magnetic insulator with nonzero charge Chern number, quantized anomalous Hall response, and chiral edge modes. Its modern theory now spans non-centrosymmetric Hubbard models, centrosymmetric Hofstadter–Hubbard systems, Kondo- and Hund-coupled honeycomb lattices, symmetry-classified in-plane antiferromagnets, and noncoplanar kagome textures, but the central condition is unchanged: antiferromagnetic order must break time reversal in a manner that no symmetry operation can undo, thereby permitting a nonvanishing integrated Berry curvature and a quantized Hall state without ferromagnetism (Jiang et al., 2017, Ebrahimkhas et al., 2022).