Sensitivity to perturbations in the three-dimensional Anderson model

arXiv:2605.25594 · cond-mat.dis-nn, cond-mat.stat-mech, quant-ph · Submitted 2026-05-25 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "Sensitivity to perturbations in the three-dimensional Anderson model".

Paper summary: Mira: The core thesis is that these two peaks in the fidelity susceptibility tell us something about a crossover from ergodic to nonergodic behavior. It’s basically using this sensitivity measure to map out how systems transition between being chaotic and being localized.

Kai: They specifically look at how the peak of this susceptibility reveals information about that crossover point, which is why they are focusing on these two different types of perturbations they introduce in the model.

Lev: From an engineering standpoint, if we can track where those peaks are in terms of disorder strength W, it gives us a roadmap for what kind of noise or structural changes would cause a system to suddenly become highly sensitive.

Kai: The paper claims that the peak position matters—the first peak shifts towards zero disorder in the thermodynamic limit, which is a big piece of information because it ties into how we expect things to behave when you have infinitely large systems.

Mira: They also look at how these peaks scale with system size or frequency cutoff, and they find that the first peak has maximal scaling, diverging as the square of the inverse frequency c.

Lev: If that maximal scaling holds up, it suggests that this crossover is a very fundamental property of the Anderson model in three dimensions. It’s not just some artifact of a small system size.

Kai: What this means for us is that understanding these two distinct peaks gives us a way to characterize the nonergodic behavior in the localized phase above the Anderson transition, which we saw mentioned on page two.

Conclusion: Mira: So, looking at this work, "Sensitivity to perturbations in the three-dimensional Anderson model," it’s really about using fidelity susceptibility as a tool to diagnose how complex quantum states react when you change the disorder strength or introduce specific types of external changes.

Kai: The authors found that this sensitivity measure has two critical points—one at weak disorder and one right at the localization transition—which helps us map out different physical regimes within a localized phase.

Lev: For practical applications, this means if we are designing systems, knowing where these peaks are helps us predict when our system will suddenly become extremely sensitive to small changes in its environment or its internal parameters.

Mira: It suggests there are two distinct ways the localized phase can behave above the Anderson transition, which they call a trivial insulator and a nontrivial Anderson insulator depending on the disorder level.

Kai: The implication is that by studying these sensitivity peaks, we get a better handle on the dynamics of these insulators, especially when you look at how long things take to relax compared to fundamental timescales like the Heisenberg time.

Piotr Tokarczyk, Lev Vidmar, Anatoli Polkovnikov, Patrycja ŁŻba

Institute of Theoretical Physics, Wroclaw University of Science and Technology · J. Stefan Institute · University of Ljubljana · Boston University

cond-mat.dis-nn, cond-mat.stat-mech, quant-ph

Submitted: 2026-05-25

Updated: 2026-11-17

DOI: 10.1103/svth-58y2

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 80/100

The gist: The paper investigates the fidelity susceptibility in the three-dimensional Anderson model to quantify how sensitive single-particle eigenstates are to perturbations.

Key concepts

Fidelity Susceptibility
This measures how much many-body eigenstates change when a small perturbation is applied. A peak in this quantity signals a crossover between different behaviors, specifically from ergodic (random) to nonergodic (localized) states.
Anderson Localization Transition
This is the critical disorder strength where single-particle states transition from being extended across the system to being localized. The paper finds that sensitivity peaks occur near this point, linking it to quantum chaos and spectral function behavior.
Nonergodic Behavior Regimes
Above the Anderson transition, there are two types of nonergodic behavior: a trivial insulator for very strong disorder and a nontrivial Anderson insulator near the transition. These regimes dictate how relaxation times behave in the localized phase.
Fading Ergodicity
This describes how matrix elements of observables change with increasing disorder. It suggests that as disorder increases, the system loses its ability to explore all states uniformly, which contributes to sensitivity peaks at weak disorder.

Terminology

Summary

The paper investigates the fidelity susceptibility in the three-dimensional Anderson model to quantify how sensitive single-particle eigenstates are to perturbations. p. 0 The main result is that the peak of this susceptibility emerges at both weak disorder strength and the critical disorder strength associated with Anderson localization. p. 0

Fidelity Susceptibility Concepts

The fidelity susceptibility quantifies the sensitivity of manybody eigenstates to perturbations. It was proposed that the peak of the fidelity susceptibility encodes information about the crossover from ergodic to nonergodic behavior. p. 0 The authors consider two types of perturbations, one breaking underlying localization or integrability and another introduced by Oˆ in Eq. p. 1 (1) at fixed W. p. 1

Peak Locations and Scaling

The main result is that the peak of the fidelity susceptibility for several one-body observables emerges, to high numerical accuracy, at W = W∗2. For certain observables, such as the sublattice kinetic energy, the fidelity susceptibility also exhibits a peak at weak disorder strength W∗1. p. 0 The peak at W = W∗1 corresponds to a crossover from localization in quasimomentum space to single-particle quantum chaos. p. 0 The scaling of the fidelity susceptibility peaks with system size or frequency cutoff reveals that the peak at W = W∗1 exhibits maximal scaling, diverging as µ−2, whereas the peak at W = W∗2 shows submaximal scaling. p. 0

Scenarios for Peaks

Two distinct scenarios give rise to the peaks in the fidelity susceptibilities. The peak at weak disorder emerges due to the relaxation rate given by the Fermi golden rule and the enhancement of matrix elements of observables can be described using fading ergodicity. p. 0 In contrast, the peak at the localization transition point can be explained via the power-law decay of the spectral function with an exponent given by the wavefunction fractal dimension at the transition point. p. 0

Localized Phase Dynamics

Studying typical fidelity susceptibilities above the Anderson transition provides evidence of two distinct regimes of nonergodic behavior. The crossover between them emerges at the disorder strength W∗3 > W∗2. p. 0 The average and typical fidelity susceptibilities differ significantly above the Anderson localization transition, with the typical value exhibiting a peak at a disorder strength close to, but lower than W∗2. p. 8

Conclusion on Regimes

The paper concludes that there are two distinct regimes of nonergodic behavior within the localized phase. These may be referred to as a trivial insulator at W > W∗3 and a nontrivial Anderson insulator at W∗2 < W W∗2. p. 9 The behavior of the spectral function f(ω)2 across the localization transition shows a plateau in the low-ω regime, indicating that the longest relaxation time becomes comparable to the Heisenberg time. p. 2

Appendix Details

The phenomenological model proposed in Ref. [78] can merge two scenarios into a single framework for standard integrability-breaking transitions. p. 0 The spectral functions from this model correspond to the two limiting scenarios: fading ergodicity and slowing down of polynomial relaxation of observables. p. 0 The key assumption is that LIOMs acquire finite relaxation rates upon the introduction of weak integrability breaking, leading to the broadening of Drude-like peaks. p. 3

Numerical Results Summary

The numerical results show that W∗1 ∼ 1/√V for the weak-disorder crossover, in agreement with the Fermi’s golden rule. The position of the second peak recovers the well-established prediction for the localization transition, W∗2 ≈ 16. p. 05.

Spectral Function Analysis

The spectral functions develop a plateau in the low-ω regime, which disappears around W ≈ 16.5. This indicates that the longest relaxation time becomes comparable to the Heisenberg time. p. 2 p. 2 The exponent 'a' for the operator nˆ is related to the fractal dimension d2 through a = 1−d2 ≈ 0.58.

Finite-Size Scaling

The critical disorder strength W∗3 appears stable in the large system-size limit, as it does not seem to flow toward W∗2 ≈ 16.5 or toward infinite disorder strength. p. 9 The scaling collapse for the Anderson localization transition is observed for all considered µ. p. 6

Average vs Typical Behavior

The differences between the average and typical fidelity susceptibilities become significant near the critical disorder strength W∗2. The lack of self-averaging of spectral functions is not only related to the lowest frequency scales associated with ωtyp. p. 9

Appendix F Details

The off-diagonal matrix element of interest is given by ⟨+ nˆ −⟩ ∼ rb∗ + rc∗2. p. 4 Single-particle energy eigenstates with nearby energies, ωtyp ≤ ϵn(b∗) − ϵm(c∗) ≪ ∆ϵ, have localization centers that are typically separated by large distances, d(b∗, c∗) ∼ V. p. 9 This spatial separation leads to an exponential suppression of the off-diagonal matrix element. p. 1

Appendix G Details

The distribution of fidelity susceptibilities χn = χ r n(µ = 0) for the sublattice kinetic energy Tˆs, broadens with increasing disorder strength W close to and above the Anderson localization transition. The developing tail of this distribution suggests that while more fidelity susceptibilities χn become close to zero, larger fidelity susceptibilities also emerge.

Appendix E Details

The exponent 'a' for the operator nˆ is related to the fractal dimension d2 through a = 1−d2 ≈ 0.58. The minimal value of-a for this operator is in very good agreement with the above prediction. p. 5 The minimal values for the other operators deviate more substantially, see the solid horizontal lines in Figs. p. 4 15(d) and 15(f).

Appendix B Details

The behavior of the autocorrelation function ⟨nˆ(t)n⟩ is expected to be consistent with that studied in the transport context of particle-number-conserving systems. This follows from the relation. p. 1

Appendix C Details

The boundary term does not affect the fidelity susceptibilities for W > 0.3. It leads to a more consistent system-size dependence of the peak marking the weak-disorder crossover at W = W∗1. p. 4

Appendix D Details

The average ratio ⟨r˜⟩ agrees with the prediction of the Gaussian Orthogonal Ensemble in chaotic regimes, i.e., r˜GOE ≈ 0.5307. In non-chaotic regimes, this average ratio usually agrees with the Poisson ensemble prediction, i. p. 4e., r˜P ≈ 0.386.

Appendix A Details

The phenomenological model assumes that LIOMs acquire finite relaxation rates upon the introduction of weak integrability breaking, leading to the broadening of Drude-like peaks. The envelope function behaves as a polynomial, ∝ ω−(2−ζ) in the range Γmin ≲ ω ≲ Γmax. p. 1 The transition point can be defined as the condensation point where the slowest relaxation rates approach the Heisenberg scale. p. 0

The spectral function develops a plateau in the low-ω regime, which disappears around W ≈ 16.5, i.e., close to the Anderson localization transition. p. 8 This indicates that the longest relaxation time becomes comparable to the Heisenberg time. The exponent 'a' for the operator nˆ is related to the fractal dimension d2 through a = 1−d2 ≈ 0.58.

The sites b∗ and c∗ are in resonance when tb∗c∗ ≥ δε, its eigenstates become bonding and anti-bonding states separated by a gap ∆± ≈ tb∗c∗. These pairs form a Mott pair. p. 9 The off-diagonal matrix element of interest is given by ⟨+ nˆ −⟩ ∼ rb∗ + rc∗2.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved system can identify two distinct peaks in fidelity susceptibility as a function of disorder strength W, distinguishing between a crossover at weak disorder (Wstar 1) and a peak at the Anderson localization transition (Wstar 2).

  2. The improved system can determine that the position of the first peak scales as Wstar 1 proportional to one over the square root of system size, which is consistent with Fermi's golden rule.

  3. The improved system can calculate that at the localization transition point, the scaling exponent for typical fidelity susceptibility is sub-maximal and related to the fractal dimension d2 of critical wavefunctions.

  4. The improved system can distinguish between two distinct regimes of nonergodic behavior within the localized phase, with a crossover occurring at a disorder strength Wstar 3 greater than Wstar 2.

  5. The improved system can calculate that the average and typical fidelity susceptibilities exhibit significant differences above the Anderson localization transition, which is useful for characterizing nonergodic behavior in this phase.

  6. The improved system can analyze how the scaling of typical fidelity susceptibility with frequency cutoff mu simplifies from a two-parameter scaling to a one-parameter scaling under specific conditions.

Sources

Related papers