Topological Anderson Insulators

Updated 28 January 2026

TAIs are quantum phases where strong disorder renormalizes band parameters to induce topological order even when the clean system is trivial.
Experiments across 1D, 2D, and 3D platforms, including HgTe quantum wells and photonic lattices, validate disorder-driven band inversion and quantized boundary states.
TAIs showcase robust, quantized transport amid bulk Anderson localization, with phase transitions characterized by real-space topological invariants and mobility gaps.

A Topological Anderson Insulator (TAI) is a quantum phase in which strong spatial disorder, rather than destroying topological order, fundamentally induces it: a system trivial or metallic in the clean limit becomes topological due to disorder-driven renormalization, manifesting quantized edge, surface, or higher-order boundary modes protected by a nontrivial topological invariant. The canonical mechanism is that random fluctuations, often in a mass-like or hopping parameter, drive a band inversion that reopens a mobility gap, inside which the bulk remains Anderson localized, but boundary states persist and are protected by symmetry and topology. TAIs exist across symmetry classes, dimensionalities, and physical platforms, and include disorder-induced analogues of quantum spin Hall, quantum Hall, Chern, and higher-order multipole insulators.

1. Disorder-Driven Topological Phase Transitions

The essential mechanism behind TAIs is disorder-induced renormalization of band parameters, leading to a topological transition even when the clean system is trivial. In Dirac-type models, nonmagnetic disorder in a “mass” term $m$ produces a self-energy correction $\Sigma$ such that the effective mass is $m_{eff}=m+\Re\Sigma$ . The self-consistent Born approximation (SCBA) provides a quantitative criterion for the topological transition, predicting critical disorder $W_c$ at which $m_{eff}$ crosses the value required for band inversion. Explicit demonstrations occur in HgTe/CdTe quantum wells (2D) (Zhang et al., 2011), 3D cubic lattice Dirac models (Guo et al., 2010), the Kane-Mele honeycomb model (Orth et al., 2015), and higher-order Benalcazar-Bernevig-Hughes (BBH) models (Lóio et al., 2023). The renormalization mechanism repeats in 1D SSH chains (Liu et al., 2022, Hsu et al., 2020), quasi-periodic SSH models (Sircar, 2024, Tang et al., 2022), circuit/Haldane models (Zhang et al., 2019), as well as in photonic and acoustic lattices (Cui et al., 2021, Liu et al., 2021).

A universal feature is the existence of a finite window of disorder strength: at low $W$ , the system remains trivial or metallic; at intermediate $W_{c_1}<W<W_{c_2}$ , the system is topological (TAI); at strong $W>W_{c_2}$ , all states localize and the system reverts to a trivial Anderson insulator. In higher-order TAI phases (e.g., third-order in 3D), the lower bound $W_{c_1}$ marks the disorder-induced band inversion and gap reopening, while the upper bound $W_{c_2}$ marks the closing of the mobility gap or the localization of boundary modes (Lóio et al., 2023, Guo et al., 2010).

2. Classification and Topological Invariants

TAI phases are characterized by the same topological invariants as their clean analogues, but computed in a real-space or disorder-adapted framework. In 1D SSH-type models with chiral (AIII) symmetry, the real-space winding number is used (Liu et al., 2022, Hsu et al., 2020, Lin et al., 3 Jan 2025). For 2D quantum spin Hall (QSH) or Chern insulators, the $\mathbb{Z}_2$ index or Chern number is obtained via twisted boundary conditions, noncommutative geometry, or the Bott index (Zhang et al., 2011, Zhang et al., 2019, Liu et al., 2021, Skipetrov et al., 2022). Higher-order TAIs utilize nested Wilson loops, quantized multipole moments (quadrupole in 2D, octupole in 3D), or boundary polarization markers (Lóio et al., 2023).

For non-periodic or random systems, approaches include:

Real-space winding number (SSH, BBH, Dirac chains) (Liu et al., 2022, Lóio et al., 2023, Lin et al., 3 Jan 2025).
Noncommutative Chern number (disordered 2D/3D insulators, circuits) (Zhang et al., 2019).
Bott index (photonic/atomic lattices) (Skipetrov et al., 2022, Liu et al., 2021).
$\mathbb{Z}_2$ index via scattering matrices (disordered TIs) (Zhang et al., 2011, Cui et al., 2021).

TAIs appear in all ten Altland-Zirnbauer symmetry classes depending on the disorder type and symmetry content: in preserving time-reversal symmetry (class AII), one gets 2D/3D QSH/STI-type TAIs (Guo et al., 2010, Khudaiberdiev et al., 2024); breaking TRS yields Chern TAIs with numerous disorder-induced chiral phases (Su et al., 2016).

3. Bulk, Boundary, and Higher-Order Modes

The protection of edge, surface, or higher-order “corner” states in TAIs follows the bulk-boundary correspondence, but with key nuances:

In the TAI regime, the bulk is Anderson localized (mobility gap) rather than band gapped, yet robust boundary states persist (Zhang et al., 2011, Khudaiberdiev et al., 2024, Lóio et al., 2023).
Edge/surface conductance remains quantized within the mobility gap, despite a finite density of localized bulk states (Zhang et al., 2011, Zhang et al., 2019).
In higher-order TAIs, spatially localized corner or hinge modes emerge—a direct manifestation of disorder-induced quantized multipole moments (e.g., octupole) (Lóio et al., 2023).
Real-space mapping reveals sharp transition between bulk-localized and boundary-localized states upon crossing TAI phase boundaries (Liu et al., 2022, Zhang et al., 2019, Skipetrov et al., 2022).
In systems with broken time-reversal symmetry, disorder can sequentially induce multiple chiral edge modes, leading to a hierarchy of conductance plateaus (e.g., $G=e^2/h, 2e^2/h, \ldots$ ) as the Chern number increases (Su et al., 2016).

A canonical diagnostic is the divergence of localization length of boundary states or the corresponding vanishing of their inverse participation ratio (IPR) at the TAI transition (Liu et al., 2022, Lóio et al., 2023, Zhang et al., 2011, Khudaiberdiev et al., 2024).

4. Experimental Realizations and Platforms

TAIs have been reported or predicted in multiple physical settings:

2D quantum wells: HgTe/CdTe and InAs/GaSb, with disorder arising from alloy fluctuations, interface roughness, or gate voltage inhomogeneity (Khudaiberdiev et al., 2024, Zhang et al., 2011).
Photonic lattices: Helical waveguide arrays (Floquet systems), photonic crystals with gyrotropic/dielectric structures, and atomic honeycomb arrays support both Chern and $\mathbb{Z}_2$ TAIs (Stützer et al., 2021, Cui et al., 2021, Skipetrov et al., 2022).
Acoustic and mechanical crystals: Bilayer phononic Lieb lattices as TRS-preserving TAIs, mechanical SSH chains with quasiperiodic modulations (Liu et al., 2021, Sircar, 2024).
Electric circuits: Haldane-model analogues in LC networks, enabling direct impedance measurement of edge states and the noncommutative Chern number (Zhang et al., 2019).
Synthetic lattices/cold atoms: SSH-type systems with engineered random and quasiperiodic disorder, observable via mean chiral displacement or real-space topology (Liu et al., 2022, Tang et al., 2022). TAIs are also realized in models with spatially correlated or binary disorder, and in Dirac-type systems with spatially rapid oscillations (homogenization theory) (Bal et al., 2023).

5. Mobility Gaps and Localization Structure

Unlike clean topological insulators, TAIs are characterized by a mobility gap rather than a true band gap (Zhang et al., 2011, Skipetrov et al., 2022). Within this mobility gap:

Bulk states are exponentially localized (typical DOS vanishes, $\rho_{typ}\rightarrow 0$ ), but boundary states remain extended and topologically protected.
The density of states (DOS) in the TAI is typically finite due to localized subbands, but these contribute negligibly to transport (Zhang et al., 2011, Skipetrov et al., 2022).
The mobility gap is determined via the divergence of the bulk localization length or the collapse of the typical DOS (Zhang et al., 2011, Girschik et al., 2015).
In quasiperiodic and non-Hermitian extensions, TAIs are not tied to Anderson transitions: one observes gapped TAIs with extended, partially localized (with mobility edges), or localized bulk states (Tang et al., 2022, Sircar, 2024).

6. Extensions: Higher-Order, Latent Symmetries, and Non-Hermitian Effects

TAIs generalize to higher-order topological phases:

Second-order TAIs in 2D induce protected corner modes (quantized quadrupole) (Lóio et al., 2023).
Third-order TAIs (TOTAI) in 3D are induced by disorder, supporting octupole-corner states, and their phase diagram is quantitatively captured by SCBA (Lóio et al., 2023).
Systems with latent (hidden) chiral or inversion symmetry support disorder-induced topological phases not visible in the bare Hamiltonian, classifiable in the tenfold way (Lin et al., 3 Jan 2025).
Non-Hermitian generalizations of TAIs can exhibit real-complex spectral transitions, mobility edges, and nontrivial spectral windings—these effects modify but do not necessarily destroy disorder-induced topology (Tang et al., 2022).
In photonic and acoustic systems, TAIs are insensitive to the presence or absence of strict spin conservation: sTRS ( $\mathcal{T}_f^2=-1$ ) persists even in systems lacking Kramers pairs (Cui et al., 2021, Liu et al., 2021).

7. Phase Diagrams, Percolation, and Criticality

The phase diagram of TAIs generally contains four regimes: trivial insulator, TAI, diffusive metal, and Anderson insulator. Critical points correspond to gap closings or diverging localization length (delocalization transition). Recent work identifies percolation transitions as a mechanism for the destruction of the TAI regime at strong disorder, connecting the emergence of percolating bulk states with the collapse of edge protection (Girschik et al., 2015).

In certain models (e.g., the SSH chain with correlated binary disorder), the TAI window is analytically locatable and tunable via disorder parameters such as the binary probability (Liu et al., 2022).

Table: Prototypical TAI Systems and their Signatures

Model / System	Disorder Type	Topological Invariant	Protected Boundary Modes
1D SSH / dimerized	random or binary hopping	Winding number ( $\mathbb{Z}$ )	Edge zero modes
2D BHZ/HgTe	random on-site	$\mathbb{Z}_2$ , Chern	Helical / chiral edge states
3D Dirac	random on-site	Strong $\mathbb{Z}_2$	Surface Dirac cones
BBH (2D/3D multipole)	random hopping	Quadrupole/octupole ( $\mathbb{Z}_2$ )	Corner states (2D/3D)
Photonic/Acoustic/Atomic	on-site or positional	Bott index, Chern, $\mathbb{Z}_2$	Edge or corner optical/phononic modes
Electric circuits (Haldane)	random inductance	Noncommutative Chern	Voltage edge modes

Topological Anderson Insulators constitute a unifying paradigm in which disorder, instead of destroying quantum topology, generates new robust transport phenomena. The TAI mechanism—disorder-driven band inversion captured by SCBA, bulk mobility-gap protection, and real-space topological invariants—extends to higher-order, non-Hermitian, and quasi-periodic systems and crosses the boundary between condensed matter, photonics, acoustics, cold atoms, and meta-materials (Lóio et al., 2023, Guo et al., 2010, Girschik et al., 2015, Zhang et al., 2019, Liu et al., 2022, Sircar, 2024, Liu et al., 2021, Skipetrov et al., 2022, Cui et al., 2021, Khudaiberdiev et al., 2024, Bal et al., 2023, Lin et al., 3 Jan 2025, Tang et al., 2022, Hsu et al., 2020).