Anomalous Floquet-Anderson Phase

• Anomalous Floquet-Anderson phase is a 2D periodically driven system characterized by a fully localized bulk and robust chiral edge modes traversing the quasienergy zone.
• It exhibits unique micromotion topology where quantized charge pumping occurs despite vanishing bulk Chern numbers, emphasizing the constructive role of disorder.
• Various models—including stepped lattice drives and continuous modulation schemes—demonstrate its realization and link its critical behavior to integer quantum Hall universality.

Anomalous Floquet-Anderson phase, more commonly termed the anomalous Floquet-Anderson insulator (AFAI), is a two-dimensional periodically driven phase of symmetry class A in which all bulk Floquet eigenstates are Anderson localized while chiral edge modes traverse the entire quasienergy Brillouin zone. Its defining anomaly is that the phase can have vanishing Floquet-band Chern numbers even though the full micromotion carries a nontrivial winding, so the topology is encoded in the unitary evolution over a period rather than in a local effective Hamiltonian. The resulting transport is a non-adiabatic quantized charge pump: when the winding number is $W=1$, one electron is pumped per cycle per edge channel (Titum et al., 2015, Zhang et al., 2021).

1. Defining characteristics and distinction from equilibrium topology

The AFAI combines two properties that are mutually incompatible in static two-dimensional band insulators: a fully localized bulk and robust chiral edge transport. On a torus, the bulk Floquet spectrum is localized at all quasienergies. On a cylinder, chiral edge modes exist for every quasienergy and wind across the full Floquet zone. In static Chern insulators, by contrast, chiral edge modes are tied to nonzero Chern numbers of occupied bands, and bulk localization must fail at critical energies where the localization length diverges. Conventional Floquet topological insulators remain describable by an effective Hamiltonian $H_{\mathrm{eff}}$ through $U(T)\approx e^{-iH_{\mathrm{eff}}T}$, with edge structure again determined by band Chern numbers. The AFAI is anomalous precisely because this description is insufficient (Titum et al., 2015).

Quasienergy periodicity is essential. Since quasienergy is defined modulo $\Omega=2\pi/T$, edge states can wrap around the Floquet Brillouin zone instead of terminating in bulk states. This permits a phase with chiral edge spectral flow even when all Floquet bands have zero Chern number. In disordered settings, the band-winding relation

$$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$

implies that if all bulk states are localized, then the sum of Chern numbers between any two quasienergies vanishes and $W_\varepsilon$ becomes quasienergy-independent. In the realization analyzed via Chern-band annihilation, $W(\varepsilon)=1$ for all $\varepsilon$ across the Floquet zone (Zhang et al., 2021).

A common misconception is that chiral edge transport always requires extended bulk states carrying nonzero Chern number. The AFAI is the canonical counterexample: the edge chirality is protected by micromotion topology, while disorder localizes the bulk rather than destroying the phase. In that sense, disorder is not merely tolerated; it is structurally important.

2. Floquet formalism and topological invariants

For a $T$-periodic Hamiltonian, the one-period unitary is

$$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$

and Floquet eigenstates satisfy

$H_{\mathrm{eff}}$0

with $H_{\mathrm{eff}}$1. The AFAI invariant is a winding number of the full micromotion $H_{\mathrm{eff}}$2, not of $H_{\mathrm{eff}}$3 alone. In clean systems, a standard momentum-space expression is

$H_{\mathrm{eff}}$4

In disordered real-space formulations, momentum derivatives are replaced by commutators with position operators and a trace per unit area (Titum et al., 2015).

The same topology can be expressed through edge spectral flow. On a cylinder, after flattening the localized bulk evolution, the edge winding counts the net chirality of edge quasienergy branches under threaded flux. A scattering formulation gives an equivalent invariant through the reflection block,

$H_{\mathrm{eff}}$5

which also equals the quantized pumped charge per cycle in appropriate transport geometries (Titum et al., 2015).

The bulk-edge correspondence takes a transport form:

$H_{\mathrm{eff}}$6

For $H_{\mathrm{eff}}$7, the pumped charge per cycle is one electron per edge channel. In the large-bias regime of two-terminal transport, the dc current saturates at

$H_{\mathrm{eff}}$8

reflecting complete occupation of the chiral source-to-drain Floquet edge mode while the localized bulk remains inert (Kundu et al., 2017).

3. Microscopic realization mechanisms

The original proof-of-principle construction is a solvable five-step square-lattice drive. The period is split into five equal steps; steps $H_{\mathrm{eff}}$9–$U(T)\approx e^{-iH_{\mathrm{eff}}T}$0 sequentially activate nearest-neighbor hoppings in a clockwise cycle, and step $U(T)\approx e^{-iH_{\mathrm{eff}}T}$1 applies sublattice on-site potentials $U(T)\approx e^{-iH_{\mathrm{eff}}T}$2. With $U(T)\approx e^{-iH_{\mathrm{eff}}T}$3, the clean bulk Floquet spectrum consists of two flat bands at $U(T)\approx e^{-iH_{\mathrm{eff}}T}$4, and the winding numbers in the $U(T)\approx e^{-iH_{\mathrm{eff}}T}$5 and $U(T)\approx e^{-iH_{\mathrm{eff}}T}$6 gaps are both $U(T)\approx e^{-iH_{\mathrm{eff}}T}$7. Adding disorder during the fifth step localizes the bulk while leaving chiral edge translation intact, thereby producing the AFAI (Titum et al., 2015).

A conceptually different route is anomalous levitation and pair annihilation. In a periodically kicked Chern-insulator model, disorder can drive two distinct scenarios: conventional levitation-and-annihilation, where the topological gap shrinks and the system becomes a trivial Anderson insulator, or anomalous levitation-and-annihilation, where extended states annihilate across the quasienergy zone boundary at $U(T)\approx e^{-iH_{\mathrm{eff}}T}$8. In the anomalous case, the trivial gap closes and reopens while the topological gap grows with disorder, so after all bulk states localize the system remains topological and enters the AFAI phase. In the two-band model analyzed there, a simple rule emerges at maximal disorder $U(T)\approx e^{-iH_{\mathrm{eff}}T}$9: the winding number of the fully localized phase equals the winding number of the dominant quasienergy gap at $\Omega=2\pi/T$0 (Liu et al., 2020).

A third route is near-resonant Chern-band annihilation in a disordered quantum anomalous Hall system. The starting point is a disordered QAH insulator with two mobility edges $\Omega=2\pi/T$1 and $\Omega=2\pi/T$2 carrying opposite Chern numbers. Driving at

$\Omega=2\pi/T$3

hybridizes the critical states across adjacent Floquet harmonics, annihilates the delocalized states, and localizes the bulk at all quasienergies. In a magnetically doped topological-insulator film, this mechanism is captured in the rotating wave approximation by an effective inter-harmonic coupling $\Omega=2\pi/T$4. Exact resonance implies that arbitrarily small drive amplitude suffices, while away from resonance the minimal coupling scales as

$\Omega=2\pi/T$5

This provides a tunable route to the AFAI through “Chern band annihilation” (Zhang et al., 2021).

The phase is not restricted to step drives. In a continuously driven honeycomb optical lattice with circular modulation

$\Omega=2\pi/T$6

adding static on-site disorder produces an AFAI window. For $\Omega=2\pi/T$7 and $\Omega=2\pi/T$8, all time-averaged bulk states become localized in the interval $\Omega=2\pi/T$9, while a broad plateau of quantized pumping persists for $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$0. This establishes that anomalous Floquet-Anderson topology survives under continuous, not only piecewise-constant, driving (Dutta et al., 2023).

4. Disorder, localization, and criticality

Disorder is essential because the defining phase requires complete bulk localization across the entire quasienergy zone. In the original lattice model, level-spacing statistics evolve from extended-state values near the circular or Gaussian unitary ensemble toward the Poisson value $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$1 as bulk localization sets in. Wavepacket dynamics show that bulk-initialized states remain localized while edge-initialized states propagate chirally. The pumped charge approaches $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$2 for $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$3, and a disorder-driven transition to a trivial localized phase occurs near $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$4 (Titum et al., 2015).

The critical behavior of this transition is closely tied to integer quantum Hall physics. Numerical evidence in the original study found an inverse-participation-ratio fractal dimension $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$5, consistent with quantum Hall plateau criticality. A later field-theoretic treatment mapped the AFAI transition onto the Pruisken non-linear sigma model of class A, with action

$$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$6

For the five-step model studied there, the bare couplings are

$$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$7

with critical point at $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$8, where $$W_{\varepsilon_1}-W_{\varepsilon_2}=C_{\varepsilon_1,\varepsilon_2}$$9. Numerical calculations found $W_\varepsilon$0 at criticality, supporting the conclusion that the AFAI transition is in the IQH universality class despite the absence of magnetic field, Landau levels, and energy conservation (Kim et al., 2019).

The coupled-network analysis of resonant Chern-band annihilation supplies a complementary localization picture. Two Chalker-Coddington networks with opposite chirality represent the two QAH mobility edges, and drive-induced interlayer coupling acts as a relevant perturbation with exponent $W_\varepsilon$1, compared with the single-layer quantum Hall exponent $W_\varepsilon$2. Along the resonance plane, arbitrarily small coupling localizes the bulk and yields the AFAI (Zhang et al., 2021).

5. Transport signatures and experimental diagnostics

The AFAI is distinguished experimentally by edge-dominated transport through a fully localized bulk. In the low-bias regime, conduction resembles that of a quantum anomalous Hall edge and is quantized at approximately $W_\varepsilon$3 per channel. In the far-from-equilibrium regime, the defining signature is current saturation at $W_\varepsilon$4, independent of bias details, reflecting one electron pumped per cycle by the chiral edge mode (Kundu et al., 2017).

Two-terminal and multiterminal calculations make this structure explicit. In a driven square-lattice transport setup, the two-terminal conductance in a nontrivial quasienergy gap remains close to one spin-degenerate conductance quantum: $W_\varepsilon$5 at $W_\varepsilon$6, $W_\varepsilon$7 at $W_\varepsilon$8, and $W_\varepsilon$9 at $W(\varepsilon)=1$0. More distinctively, for probe energies that lie within pristine bulk bands, the conductance first decreases as disorder suppresses bulk transport and then returns to a near-quantized plateau for $W(\varepsilon)=1$1, demonstrating transport by edge modes that wind through the entire quasienergy zone. In a T-shaped geometry, chirality appears as $W(\varepsilon)=1$2 and $W(\varepsilon)=1$3 in the topological regime (Rodriguez-Mena et al., 2019).

The large-bias quantization can be derived through a Floquet-Landauer treatment and an extended-zone mapping. If $W(\varepsilon)=1$4 denotes the summed transmission from source to drain over all sidebands, then

$W(\varepsilon)=1$5

and the sum rule

$W(\varepsilon)=1$6

relates the driven transport problem to a static extended-zone Chern insulator. In the AFAI, $W(\varepsilon)=1$7 is quantized to the winding number, yielding $W(\varepsilon)=1$8 in physical units (Kundu et al., 2017).

Flux-threading geometries provide a complementary diagnostic. In the continuously driven honeycomb model, a Laughlin pump on a cylinder reveals a discontinuity in the charge imbalance

$W(\varepsilon)=1$9

as the threaded flux $\varepsilon$0 varies from $\varepsilon$1 to $\varepsilon$2. The extracted pump amplitude

$\varepsilon$3

satisfies $\varepsilon$4 throughout the anomalous pumping plateau, consistent with $\varepsilon$5 (Dutta et al., 2023).

6. Interacting and nonperiodic extensions

The noninteracting AFAI admits many-body extensions in the presence of many-body localization. For interacting fermionic Floquet phases, a many-body edge invariant $\varepsilon$6 generalizes the single-particle winding number to a locality-preserving edge unitary. In Fock-space systems of fundamental fermions,

$\varepsilon$7

A trivial Floquet Anderson insulator has $\varepsilon$8, while the noninteracting AFAI edge corresponds to translation by one physical fermion site and has $\varepsilon$9. The same framework also identifies a radical $T$0 component, $T$1, associated with pumping a Kitaev Majorana chain to the boundary each period; in that case $T$2 (Fidkowski et al., 2017).

The phase is also stable to certain temporal perturbations. Under quasiperiodic two-tone timing modulation of the five-step drive, robust topological pumping survives at weak noise and is more stable than in the white-noise case. For representative parameters $T$3 and $T$4, the pumped charge exhibits a long-lived quantized plateau. Bulk wavepacket spreading becomes subdiffusive rather than diffusive, with participation ratio scaling $T$5 and mean-squared displacement $T$6 with $T$7. When the quasiperiodic modulation becomes strong enough to destroy topology, the dynamics cross over to diffusion with $T$8 (Zheng et al., 2022).

A more radical generalization abandons exact time-translation symmetry altogether. In anomalous random multipolar driven insulators, no global Floquet operator exists when $T$9, yet long-lived prethermal Anderson localization and quantized boundary pumping remain possible. The topological response is characterized not by a Floquet winding number but by a block-averaged bulk orbital magnetization,

$$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$0

which equals the pumped edge charge per block,

$$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$1

For Thue-Morse driving, the prethermal lifetime scales as

$$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$2

while for $$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$3-random multipolar driving,

$$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$4

These constructions are not strictly Floquet AFAIs, but they preserve the same anomalous combination of localized bulk and quantized chiral boundary transport over parametrically long times (Zhao et al., 2022).

Clean photonic anomalous Floquet phases provide another nearby setting. In staggered helical waveguide arrays, chiral $$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$5-gap edge states occur even though the single Floquet band at $$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$6 has Chern number zero, and the transition at $$U(T)=\mathcal{T}\exp\!\left[-i\int_0^T H(t)\,dt\right],$$7 hosts an unpaired Dirac cone. The paper presents this as a clean anomalous Floquet topological insulator rather than an AFAI, but it supplies the disorder-robust anomalous edge micromotion that would be required for a photonic AFAI analogue (Leykam et al., 2016).

Open problems remain concentrated around the ultimate stability of localization under interactions, phonon coupling, and long-time heating. The available results indicate that strong disorder, many-body localization, or prethermal mechanisms can preserve anomalous pumping for substantial windows, but eventual delocalization remains a central constraint outside strictly noninteracting or localized regimes (Zhang et al., 2021).