Haldane–Anderson Hamiltonian Overview

• The Haldane–Anderson Hamiltonian model is a family of frameworks that combine topological band physics or localized impurity behavior with disorder or a gapped electronic environment.
• It encompasses diverse formulations including disordered lattice models, interacting spinful systems with Hubbard interactions, and impurity models applied to semiconductor and molecular systems like vitamin B12.
• Research employs analytical methods (e.g., SCBA) and numerical techniques (e.g., transfer matrix and exact diagonalization) to map transitions between trivial, topological Anderson, and gapless phases.

Searching arXiv for the cited Haldane–Anderson model papers to ground the article. I’ll look up the most relevant arXiv records for the Haldane–Anderson model and related Haldane-with-Anderson-disorder work. Searching arXiv… The expression Haldane–Anderson Hamiltonian model does not denote a single universally fixed Hamiltonian. In the literature, it refers to several closely related constructions in which either the Haldane honeycomb Chern-insulator Hamiltonian is combined with Anderson-type disorder, or a localized impurity orbital is coupled to a gapped host in the Haldane–Anderson impurity sense. Recent work also uses the term for interacting spinful Haldane models with Hubbard and nearest-neighbor interactions in the presence of Anderson disorder, and for a single localized adsorbate orbital hybridized with a semiconductor bath in nonadiabatic gas-surface scattering (Gonçalves et al., 2018, Silva et al., 2023, Uría-Álvarez et al., 26 May 2026, Kandemir et al., 2015, Lu et al., 18 Aug 2025). Across these usages, the common structure is the coexistence of topological band physics or localized impurity physics with a disordered or gapped electronic environment.

1. Scope of the term and model classes

A compact way to organize the literature is to separate three model classes.

Model class	Degrees of freedom	Representative use
Disordered Haldane lattice model	Honeycomb-lattice fermions with complex NNN hopping and on-site disorder	Topological Anderson phases in Chern insulators
Interacting spinful Haldane–Anderson model	Spinful Haldane fermions with $U$, $V$, and Anderson disorder	Disorder-assisted $C=2$ and antiferromagnetic $C=1$ phases
Haldane–Anderson impurity model	Localized impurity orbital(s) hybridized with a gapped host	Vitamin B$_{12}$ electronic structure; semiconductor surface scattering

In the disordered lattice usage, the model is the Haldane Hamiltonian on the honeycomb lattice supplemented by random on-site potentials. In the interacting spinful usage, the same lattice structure is extended by onsite Hubbard and nearest-neighbor density-density terms, optionally with a staggered sublattice mass. In the impurity usage, the Haldane–Anderson model is a many-body framework for a localized transition-metal $d$ shell or a localized adsorbate orbital embedded in a semiconducting host with a band gap (Gonçalves et al., 2018, Silva et al., 2023, Kandemir et al., 2015).

A common misconception is that the phrase names one canonical Hamiltonian. The published record instead shows a family of models sharing a conceptual motif: hybridization, disorder, or both act in the presence of a topological mass or a host gap, and the resulting phase structure is controlled by competition among symmetry breaking, localization, and spectral reconstruction.

2. Spinless honeycomb formulation with Anderson and binary disorder

For the half-filled spinless honeycomb problem, the clean Haldane Hamiltonian is written as (Gonçalves et al., 2018)

$$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$

where $c_i$ and $c_i^\dagger$ annihilate and create a spinless fermion on site $i$, $V$0 is the nearest-neighbor hopping between sublattices $V$1 and $V$2, $V$3 is the complex next-nearest-neighbor hopping within each sublattice, and $V$4 is the staggered sublattice potential with $V$5 on $V$6 and $V$7 on $V$8. The sign structure of the NNN Peierls phases is $V$9, with $C=2$0 fixed by the orientation of the two NN bonds composing the NNN hop, while the net flux per unit cell remains zero. In the numerics, $C=2$1 is set to unity and $C=2$2.

The disorder term is purely on-site,

$C=2$3

with $C=2$4 drawn either from the Anderson distribution

$C=2$5

uniform in $C=2$6, or from the binary distribution

$C=2$7

with equal probability for $C=2$8 and $C=2$9. The full model is

$C=1$0

Linearization near the inequivalent valleys $C=1$1 and $C=1$2 gives the clean Dirac masses

$C=1$3

and the clean Chern number is

$C=1$4

A nontrivial phase with $C=1$5 therefore occurs when $C=1$6. For $C=1$7, the clean transition occurs at

$C=1$8

which is $C=1$9 for $_{12}$0.

This formulation is the minimal lattice realization of the topological Anderson-insulator problem in a Chern band. The two disorder choices are not interchangeable in their phase-diagram consequences: both support disorder-induced topology, but binary disorder additionally produces a reentrant topological regime as a function of staggered potential.

3. Weak-disorder theory: self-consistent Born approximation and mass renormalization

At half-filling and $_{12}$1, the low-energy clean Hamiltonian used in the self-consistent Born approximation is (Gonçalves et al., 2018)

$_{12}$2

with Fermi velocity $_{12}$3. Here $_{12}$4 acts on sublattice space and $_{12}$5 on valley pseudospin. The $_{12}$6 term gives the leading quadratic correction to the Haldane mass.

For scalar on-site disorder, the SCBA self-energy is momentum independent and has the decomposition

$_{12}$7

At half-filling and $_{12}$8, electron-hole symmetry implies that at $_{12}$9 only the overall energy shift $d$0 and the mass renormalization $d$1 are relevant. The SCBA equation is

$d$2

with $d$3 and disorder variance

$d$4

The effective parameters at $d$5 are

$d$6

so that at half-filling the topology is controlled by $d$7. Equivalently, in sublattice components,

$d$8

and the renormalized valley masses are

$d$9

The corresponding Chern number is

$$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$0

The central weak-disorder result is that scalar on-site disorder reduces the trivial inversion-breaking mass and can drive band inversion between valleys, thereby producing a topological Anderson insulator. In the formulation of the paper, the analytical phase boundaries at $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$1 follow from $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$2 and $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$3 computed self-consistently. This mechanism is quantitatively accurate at low disorder and captures the initial enhancement of the topological region, but it fails once higher-order scattering, nonperturbative localization, and broad spectral tails become dominant.

4. Numerical diagnostics, phase diagrams, and gapless topology

Beyond SCBA, the disordered Haldane model is characterized numerically by a combination of topological, spectral, and localization diagnostics (Gonçalves et al., 2018). The Chern number is computed with a real-space method for disordered systems, operationally expressible through the projector formula

$$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$4

where $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$5 projects onto occupied states and $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$6 are position operators on a torus of area $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$7. In practice, the calculation uses the Fukui lattice-gauge algorithm generalized to real space with broken translational invariance.

The distinction between gapped and gapless regimes is obtained from the density of states at the Fermi level, $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$8, using a recursive Green’s function method on lattices up to $$H_0 = -t \sum_{\langle i,j\rangle} c_i^\dagger c_j + t_2 \sum_{\langle\!\langle i,j\rangle\!\rangle} e^{i\phi_{ij}} c_i^\dagger c_j + \eta \sum_i \zeta_i\, c_i^\dagger c_i,$$9. The phase is called gapped if the DOS at the Fermi level is below a threshold of $c_i$0, with $c_i$1 and $c_i$2; otherwise it is called gapless. Averaging over $c_i$3 disorder realizations is used to estimate uncertainties. Localization is analyzed with the transfer matrix method through the normalized localization length $c_i$4 in strips with $c_i$5 and $c_i$6 from $c_i$7 to $c_i$8: decreasing $c_i$9 with $c_i^\dagger$0 indicates an insulating phase, increasing $c_i^\dagger$1 a metallic phase, and $c_i^\dagger$2 const a critical point.

The phase-diagram results are specific. For Anderson disorder, increasing $c_i^\dagger$3 first stretches the $c_i^\dagger$4 regions along $c_i^\dagger$5 and squeezes them along $c_i^\dagger$6; near $c_i^\dagger$7 the topological regions retract, while they persist near $c_i^\dagger$8. Along $c_i^\dagger$9, the topological phase survives up to a critical $i$0, and for $i$1 the transition to $i$2 occurs at $i$3 larger than the clean $i$4, which is the topological Anderson-insulator regime. For binary disorder, topological regions also stretch along $i$5 and squeeze along $i$6, but a reentrant behavior appears: at fixed $i$7 slightly below destruction, the last surviving topological regions occur at finite $i$8, and near $i$9 the topological region disappears earlier. In the $V$00 diagram at $V$01, one finds a single critical point at $V$02 with $V$03, and two critical points at $V$04.

An important result is that topology need not coincide with a spectral gap. At small disorder, the topological transitions are accompanied by gap closing and reopening at the Fermi level. At larger disorder, both trivial and topological regions can become gapless, with the Chern number remaining quantized because the relevant protection is a mobility gap rather than a spectral gap. The paper interprets the ultimate destruction of topology through a levitation-and-annihilation mechanism: disorder drives extended bulk states above and below the Fermi level toward one another until they annihilate at a critical point.

5. Interacting spinful extensions: Hubbard, nearest-neighbor repulsion, and antiferromagnetic topological Anderson phases

In the interacting literature, the Haldane–Anderson Hamiltonian denotes a spinful Haldane model on the honeycomb lattice with Anderson disorder and electron-electron interactions (Silva et al., 2023, Uría-Álvarez et al., 26 May 2026). One formulation writes

$V$05

with

$V$06

Another equivalent decomposition is

$V$07

where

$V$08

$V$09

$V$10

and

$V$11

with $V$12 on sublattice $V$13 and $V$14 on $V$15. The typical parameter choice in these studies is $V$16, $V$17, and $V$18, with half-filling.

Two computational strategies appear. The 2023 work treats interactions in mean field, using Hartree and Fock channels and self-consistent evaluation of spin- and charge-dependent fields. It defines the charge-density-wave and spin-density-wave order parameters

$V$19

$V$20

and computes the Chern number with the Fukui–Hatsugai–Suzuki method under twisted boundary conditions. The 2026 work uses finite-size exact diagonalization on a 12A honeycomb cluster, many-body Chern numbers under twists, and a neural network trained on single-particle density matrices at $V$21 and $V$22, with labels $V$23, $V$24, and $V$25. It reports focal loss for class imbalance and an F1 score of approximately $V$26 on the $V$27 class.

The interacting phase structure extends the noninteracting topological Anderson-insulator picture. Disorder enlarges the $V$28 topological region and can stabilize a disorder-driven $V$29 phase at $V$30 in which topology coexists with long-range spin and charge order (Silva et al., 2023). Exact diagonalization further finds an antiferromagnetic $V$31 topological Anderson phase, consistent with the antiferromagnetic quantum anomalous Hall insulator previously identified in the clean model at finite staggered mass (Uría-Álvarez et al., 26 May 2026). The central mechanism proposed in both interacting studies is that explicit charge imbalance is required to induce the $V$32 phase: in the clean problem this is supplied by a uniform Semenoff mass $V$33, whereas in the disordered problem it is generated by the spatially inhomogeneous Anderson potential. The 2026 work further argues that the disorder-induced $V$34 phase is continuously connected to the clean antiferromagnetic Chern insulator by varying $V$35 at fixed $V$36.

6. Impurity-in-a-gapped-host formulations and nonequilibrium variants

In a distinct but historically important usage, the Haldane–Anderson impurity model describes a correlated localized impurity hybridized with a semiconducting host with a band gap (Kandemir et al., 2015). For a multi-orbital single impurity, the Hamiltonian used for vitamin B$V$37 is

$V$38

Here $V$39 labels host eigenstates, $V$40 labels the five Co $V$41 natural atomic orbitals, and the impurity Green’s function contains the hybridization function

$V$42

The paper applies a double-counting shift

$V$43

with $V$44 and therefore $V$45 for cyanocobalamin.

The physical distinction from the metallic Anderson problem is explicit in the data: in a semiconductor, hybridization with a gapped bath generates discrete impurity bound states within the gap rather than a zero-energy Kondo resonance. In vitamin B$V$46, Hartree–Fock gives a host HOMO–LUMO gap of $V$47 eV, and HF+QMC finds new correlated states inside that gap, including upper-Hubbard features near the valence edge and mid-gap impurity bound states dominated by CN axial ligand and corrin-ring character with Co $V$48-like admixture. The bound-state filling controls local moments and antiferromagnetic correlations between Co $V$49 orbitals and surrounding host states.

A recent nonequilibrium extension transfers the same impurity-in-a-gapped-band logic to gas-surface scattering on $V$50 Ge(111) (Lu et al., 18 Aug 2025). The time-dependent Hamiltonian is

$V$51

with a single localized hydrogen-related orbital $V$52 coupled to a discretized semiconductor bath containing an explicit gap $V$53 eV, Fermi level $V$54 eV, bandwidth $V$55 eV, $V$56 bath states, and $V$57 electrons. No on-site Coulomb term is included in the reported simulations, so $V$58. The hybridization function is

$V$59

This surface-scattering variant is propagated with independent-electron surface hopping and Ehrenfest dynamics. The key result is threshold behavior tied to the band gap: nonadiabatic dissipation is strongly suppressed below the gap and switches on only when the projectile kinetic energy can drive transitions of at least $V$60. The reported mean nonadiabatic kinetic-energy loss at incident energy $V$61 eV is approximately $V$62–$V$63 eV, and the paper argues that independent-electron surface hopping captures the observed threshold whereas mean-field dynamics predicts weak, largely $V$64-independent energy loss.

These impurity and gapped-bath formulations show that the phrase Haldane–Anderson Hamiltonian model has broader reach than topological lattice disorder alone. It also designates a general framework in which a localized degree of freedom couples to a host with a true gap, producing impurity bound states, gap-threshold nonadiabaticity, or both. A plausible implication is that the unifying content of the name is not a particular lattice geometry, but a modeling strategy for electronic structure and dynamics in the simultaneous presence of localization, hybridization, and a gapped environment.