Resonance Proliferation Across Localization Transitions

arXiv:2605.05445 · cond-mat.dis-nn, cond-mat.quant-gas, cond-mat.stat-mech · Submitted 2026-05-06 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Hi, I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Resonance Proliferation Across Localization Transitions".

Mira: Here's the gist of the paper. The gist The proliferation of many-body resonances drives slow numerical drifts towards delocalization in many-body localization models, and this phenomenon is captured by a statistical method that predicts the distribution of resonance oscillation frequencies How it works The core idea involves developing a statistical method to predict the distribution of resonance oscillation frequencies which captures how the formation of resonances at larger frequency scales subsequently affects the formation of resonances at lower frequencies Working within the statistical Jacobi approximation (SJA), we derive a flow equation for a power-law exponent θ(w) characterizing the density of resonances at frequency scale w.

Kai: A localized phase is described by a line of fixed points with θ(w) > 0, while an instability of the flow signals resonance proliferation and the onset of thermalization.

Mira: The predicted θ(w) matches numerics on the Anderson model on random regular graphs and the L´evy-RosenzweigPorter random matrix ensemble, both of which host resonance-driven delocalization transitions.

Kai: Resonances and the Jacobi Algorithm Many-body resonances are understood within a dynamical picture where initial product states related by n spin flips require going to roughly the nth order of perturbation theory before a non-zero matrix element connecting them can be found.

Title and authors: Kai: So we're looking at this paper, "Resonance Proliferation Across Localization Transitions," and it’s about how many-body localization models drift towards delocalization as the system gets bigger, which isn't fully explained by current theory.

Mira: Exactly. The core idea here is that these drifts are driven by many-body resonances popping up at intermediate disorder strengths, and this paper develops a statistical method to predict the distribution of those resonance oscillation frequencies, capturing how resonances at bigger scales affect smaller ones.

Lev: From an engineering standpoint, we're dealing with a problem where the system size really matters for these numerical drifts; it sounds like they’re trying to get a way to see that transition before the simulation runs too long.

Kai: The paper uses something called the statistical Jacobi approximation, or SJA, to derive this flow equation for an exponent theta of resonance density at a frequency scale w.

Mira: And that theta is key because it distinguishes between localized phases where theta is greater than zero, and delocalized ones where it's less than zero.

Lev: So what the paper shows is that when the flow becomes unstable, meaning an instability in that flow signals resonance proliferation and the onset of thermalization.

Kai: They confirm this by showing that their derived theta matches what we see numerically in models like the Anderson model on random regular graphs and the Lévy-Rosenzweig-Porter random matrix ensemble.

Mira: That's a pretty strong check, because those models are known to have these resonance-driven delocalization transitions, which is what they’re aiming to capture.

Lev: The Jacobi algorithm itself is used to track this proliferation by iteratively diagonalizing the Hamiltonian through two-level rotations that cut down those large off-diagonal matrix elements.

Kai: The SJA focuses on the distribution of these decimated elements, which they call rho dec(w), and identifies two regimes in the Jacobi flow: a sparse one and a dense one.

Mira: They extract nres(w), which is the number of resonances with a frequency scale larger than w, modeling it as nres(w)/N proportional to w theta plus some constant.

Lev: That formula gives us the resonance exponent theta, which is what tells us if we're in a Hilbert-space-localized phase or a delocalized one.

Kai: They have this heuristic flow equation, dtheta(w)/dw equals f

theta(w): w over theta of w minus one, which predicts a transition from a localized phase where resonances don't proliferate to one that is proliferating and thermalizing when theta goes to negative infinity.

Title and authors: Mira: The paper also compares this heuristic flow equation against three different models: the Lévy-Rosenzweig-Porter random matrix model, the Anderson model on random regular graphs, and the disordered Heisenberg spin chain.

Lev: The numerical results show that there's a qualitative agreement between what these heuristic flow equations predict and what you get from exact Jacobi diagonalization for these models within their valid regimes.

Kai: They predict that for a toy MBL model starting with an initial theta w0 greater than zero, a sufficiently strong perturbation can cause it to flow to theta w less than zero, leading to delocalization.

Mira: This suggests that large systems with bare localization lengths zeta less than zeta c might actually thermalize at long times, which is a big implication for the MBL picture.

Lev: The paper also predicts qualitative features for the phase diagram of the Lévy-Rosenzweig-Porter model, the random regular graph model, and that toy perturbed MBL chain.

Kai: Specifically, they predict there will be both a stable localized phase and a delocalized phase separated by a sharp transition.

Mira: They use nres(w)/N

Eq. (thirteen): as a proxy for the participation ratio, which is essentially how many basis states are involved in the state, making it a diagnostic for Hilbert space delocalization <ref:2605.05445#pg2>.

Lev: And they predict that when theta w is slowly varying and before it diverges into the dense regime, we see stretched exponential decay with a slowly varying exponent C(t) proportional to e to minus t over tau beta of t.

Kai: That functional form relates directly to the prethermal regime of MBL dynamics, allowing us to forecast autocorrelation functions and return probabilities.

Mira: They also mention that their SJA treats all three systems identically, predicting that when resonances proliferate, theta flows to large negative values as smaller Rabi frequencies are resolved.

Lev: The paper points out that the transition to delocalization is driven by this resonance proliferation captured by the SJA, but they also note a limitation: the paper doesn't claim that thermal avalanches are not responsible for asymptotic instability in MBL systems with random disorder.

Kai: They also suggest that prethermal MBL is controlled by resonance proliferation before other effects, like avalanches or finite system size, cut it off.

Title and authors: Mira: The analysis predicts BKT-like scaling for the critical behavior of the heuristic flow equations, which contrasts with some other renormalization group analyses.

Lev: The divergence of w minus one c as you vary the initial values theta zero and w zero across the separatrix reflects this BKT scaling, giving a thermalization time that diverges faster than any power law.

Kai: Overall, this paper on "Resonance Proliferation Across Localization Transitions" provides a statistical framework to predict when MBL systems transition from being localized to becoming delocalized based on resonance dynamics.

Mira: It gives us the tool to see this crossover happening in real-time through the flow equation for theta(w), even if it's not a strict renormalization group procedure.

Lev: For real hardware, this suggests we need to monitor these resonance structures as disorder changes, because that's when we expect the system to enter that delocalized regime.

Kai: So, they provide qualitative features for the phase diagram of those random matrix models and the toy MBL model by showing how theta flows based on resonance proliferation.

Mira: The main implication is that resonance proliferation itself is a key driver for this transition, which helps us understand why some systems thermalize while others don't.

Lev: And they also provide predictions for observables like participation ratios and return probabilities, giving us concrete things to look for in experiments or simulations.

Kai: It’s a solid piece of work because it links the dynamical picture—the Rabi oscillations—to a statistical method that predicts when those resonances become too numerous and drive the system away from localization.

Mira: The key thing is that they treat these different systems with the same underlying mechanism for this transition, which simplifies our understanding of what's happening across different disordered lattices.

Lev: So, to wrap up on "Resonance Proliferation Across Localization Transitions," we have a statistical tool that tracks how resonance density evolves with frequency scale, leading to a clear picture of the localized versus delocalized phases.

Kai: It’s a framework for seeing the transition driven by these many-body resonances before it happens.

Mira: It gives us concrete predictions about the phase diagram and dynamical signatures in both theoretical models and on real systems like superconducting circuits or spin chains.

Lev: We can use this to guide how we design experiments to probe these transitions, focusing on those points where the flow equation predicts instability.

The paper's summary: Kai: So we're looking at this paper, "Resonance Proliferation Across Localization Transitions," and it’s about how many-body localization models drift towards delocalization as the system gets bigger, which isn't fully explained by current theory.

Mira: Exactly. The core idea here is that these drifts are driven by many-body resonances popping up at intermediate disorder strengths, and this paper develops a statistical method to predict the distribution of those resonance oscillation frequencies, capturing how resonances at bigger scales affect smaller ones.

Lev: From an engineering standpoint, we're dealing with a problem where the system size really matters for these numerical drifts; it sounds like they’re trying to get a way to see that transition before the simulation runs too long.

Kai: The paper uses something called the statistical Jacobi approximation, or SJA, to derive this flow equation for an exponent theta of resonance density at a frequency scale w.

Mira: And that theta is key because it distinguishes between localized phases where theta is greater than zero, and delocalized ones where it's less than zero.

Lev: So what the paper shows is that when the flow becomes unstable, meaning an instability in that flow signals resonance proliferation and the onset of thermalization.

Kai: They confirm this by showing that their derived theta matches what we see numerically in models like the Anderson model on random regular graphs and the Lévy-Rosenzweig-Porter random matrix ensemble.

Mira: That's a pretty strong check, because those models are known to have these resonance-driven delocalization transitions, which is what they’re aiming to capture.

Lev: The Jacobi algorithm itself is used to track this proliferation by iteratively diagonalizing the Hamiltonian through two-level rotations that cut down those large off-diagonal matrix elements.

Kai: The SJA focuses on the distribution of these decimated elements, which they call rho dec(w), and identifies two regimes in the Jacobi flow: a sparse one and a dense one.

Mira: They extract nres(w), which is the number of resonances with a frequency scale larger than w, modeling it as nres(w)/N proportional to w theta plus some constant.

Lev: That formula gives us the resonance exponent theta, which is what tells us if we're in a Hilbert-space-localized phase or a delocalized one.

Kai: They have this heuristic flow equation, dtheta(w)/dw equals f

theta(w): w over theta of w minus one, which predicts a transition from a localized phase where resonances don't proliferate to one that is proliferating and thermalizing when theta goes to negative infinity.

The paper's summary: Mira: The paper also compares this heuristic flow equation against three different models: the Lévy-Rosenzweig-Porter random matrix model, the Anderson model on random regular graphs, and the disordered Heisenberg spin chain.

Lev: The numerical results show that there's a qualitative agreement between what these heuristic flow equations predict and what you get from exact Jacobi diagonalization for these models within their valid regimes.

Kai: They predict that for a toy MBL model starting with an initial theta w0 greater than zero, a sufficiently strong perturbation can cause it to flow to theta w less than zero, leading to delocalization.

Mira: This suggests that large systems with bare localization lengths zeta less than zeta c might actually thermalize at long times, which is a big implication for the MBL picture.

Lev: The paper also predicts qualitative features for the phase diagram of the Lévy-Rosenzweig-Porter model, the random regular graph model, and that toy perturbed MBL chain.

Kai: Specifically, they predict there will be both a stable localized phase and a delocalized phase separated by a sharp transition.

Mira: They use nres(w)/N

Eq. (thirteen): as a proxy for the participation ratio, which is essentially how many basis states are involved in the state, making it a diagnostic for Hilbert space delocalization.

Lev: And they predict that when theta w is slowly varying and before it diverges into the dense regime, we see stretched exponential decay with a slowly varying exponent C(t) proportional to e to minus t over tau beta of t.

Kai: That functional form relates directly to the prethermal regime of MBL dynamics, allowing us to forecast autocorrelation functions and return probabilities.

Mira: They also mention that their SJA treats all three systems identically, predicting that when resonances proliferate, theta flows to large negative values as smaller Rabi frequencies are resolved.

Lev: The paper points out that the transition to delocalization is driven by this resonance proliferation captured by the SJA, but they also note a limitation: the paper doesn't claim that thermal avalanches are not responsible for asymptotic instability in MBL systems with random disorder.

Kai: They also suggest that prethermal MBL is controlled by resonance proliferation before other effects, like avalanches or finite system size, cut it off.

The paper's summary: Mira: The analysis predicts BKT-like scaling for the critical behavior of the heuristic flow equations, which contrasts with some other renormalization group analyses.

Lev: The divergence of w minus one c as you vary the initial values theta zero and w zero across the separatrix reflects this BKT scaling, giving a thermalization time that diverges faster than any power law.

Kai: Overall, this paper on "Resonance Proliferation Across Localization Transitions" provides a statistical framework to predict when MBL systems transition from being localized to becoming delocalized based on resonance dynamics.

Mira: It gives us the tool to see this crossover happening in real-time through the flow equation for theta(w), even if it's not a strict renormalization group procedure.

Lev: For real hardware, this suggests we need to monitor these resonance structures as disorder changes, because that's when we expect the system to enter that delocalized regime.

Kai: So, they provide qualitative features for the phase diagram of those random matrix models and the toy MBL model by showing how theta flows based on resonance proliferation.

Mira: The main implication is that resonance proliferation itself is a key driver for this transition, which helps us understand why some systems thermalize while others don't.

Lev: And they also provide predictions for observables like participation ratios and return probabilities, giving us concrete things to look for in experiments or simulations.

Kai: It’s a solid piece of work because it links the dynamical picture—the Rabi oscillations—to a statistical method that predicts when those resonances become too numerous and drive the system away from localization.

Mira: The key thing is that they treat these different systems with the same underlying mechanism for this transition, which simplifies our understanding of what's happening across different disordered lattices.

Lev: So, to wrap up on "Resonance Proliferation Across Localization Transitions," we have a statistical tool that tracks how resonance density evolves with frequency scale, leading to a clear picture of the localized versus delocalized phases.

Kai: It’s a framework for seeing the transition driven by these many-body resonances before it happens.

Mira: It gives us concrete predictions about the phase diagram and dynamical signatures in both theoretical models and on real systems like superconducting circuits or spin chains.

Lev: We can use this to guide how we design experiments to probe these transitions, focusing on those points where the flow equation predicts instability.

The paper's improvements: Kai: So we're looking at how they suggest improving this resonance proliferation study, and Mira, what do you think are the biggest leaps they're pushing?

Mira: They are trying to make this statistical method more predictive, moving it from just describing what happens to actually anticipating when a system will delocalize before it even starts.

Lev: That makes sense for hardware runs; we don't want to spend ten hours cooling down and measuring if the system is going to thermalize or stay localized.

Kai: They’re suggesting an AI system that can derive a flow equation for that power-law exponent theta, which would let us tell right away if we’re in a localized phase where theta is greater than zero, or if it signals instability where resonance proliferation starts.

Mira: That would be huge because right now we have to run the Jacobi algorithm and check the results after the fact; an AI predicting that transition ahead of time would change how we model these many-body systems.

Lev: If that AI could actually work on a real system, it means we could potentially find those delocalization transitions much faster than current simulation methods allow.

Kai: And they also want to see if the SJA flow equation can be applied to different physical setups, like the Lévy-Rosenzweig-Porter random matrix model, to predict phase diagrams with both localized and delocalized regions.

Mira: That would let us see if this mechanism is universal across different types of disorder or different lattice structures.

Lev: The paper also suggests creating a module that directly links the number of resonances per state above some scale w, nres(w)/N, to the participation ratio, which is how many basis states are actually involved in the state.

Kai: So if we can use that as a direct diagnostic for Hilbert space delocalization, it gives us something concrete to measure when we look at our qubit measurements.

Mira: They also want a way to predict real-time dynamics using that functional flow equation they derived, which suggests we could forecast how fast the system decays in the prethermal regime of MBL.

Lev: Forecasting those return probabilities would be incredibly useful for error correction because it tells us what kind of noise we might see before the system settles into a steady state.

Kai: So these improvements focus on making this theory actionable by giving us predictive tools for identifying transitions, measuring delocalization, and predicting dynamical behavior in real-time.

Mira: It really shifts the focus from just analyzing the final state to understanding the process of how we get there through those resonances.

Lev: It’s a step toward designing error correction protocols that account for these resonance-driven instabilities rather than assuming they don't happen.

Conclusion: Kai: So to wrap up this discussion on "Resonance Proliferation Across Localization Transitions," we’ve seen how this paper uses the statistical Jacobi approximation to track an exponent theta that tells us if a system is localized or delocalized based on resonance density.

Mira: It really shows that the transition isn't just about disorder strength; it's about these many-body resonances popping up at different frequency scales and driving the whole thing.

Lev: From an engineering view, it means we should be looking for these resonance structures in our experiments because that’s when we expect the system to go into that delocalized regime.

Kai: The implication is that this framework gives us a clear way to predict when MBL systems will transition, and Mira's point about the AI prediction capability sounds really promising for future work.

Mira: Right, because it suggests a new way to look at the dynamics of these complex many-body states in terms of these resonance flows.

Lev: For error correction, it means we can start thinking about how those resonance proliferations might introduce noise or thermal avalanches that we need to account for.

Kai: So, even though this analysis is based on some heuristic flow equations and approximations, the paper gives us a solid statistical tool to guide our experimental setup.

Mira: It’s a useful map for the phase diagram of models like the LRP and RRG, showing exactly where we expect to find that localized versus delocalized boundary.

Lev: It helps ground our understanding by connecting those abstract many-body dynamics back to something we can actually measure on a real quantum computer or in a spin chain simulation.

Kai: So, "Resonance Proliferation Across Localization Transitions" gives us the language and the predictive tools to better understand these localization transitions driven by many-body effects.

Mira: It sets up a clear path for future work where we can test these flow equations on more complex models or perhaps incorporate those AI predictions for earlier detection.

Lev: I think the next logical step is testing these flow equations against actual hardware data to see if the BKT scaling they predict holds up under real-world constraints.

Department of Physics, Princeton University · Department of Physics, Stanford University · Department of Physics, Boston University

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

Submitted: 2026-05-06

Updated: 2026-10-20

Comments: 20+4 pages, comments welcome!

Code: https://github.com/CarloVanoni/JacobiFlow

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 83/100

The gist: The gist The proliferation of many-body resonances drives slow numerical drifts towards delocalization in many-body localization models, and this phenomenon is captured by a statistical method that

Key concepts

Many-body Resonances
These are slow oscillations between many-body states, requiring high-order perturbation theory to find. They are a direct signature of interactions and can lead to delocalization when they proliferate across different frequency scales.
Statistical Jacobi Approximation (SJA)
This is a statistical technique used to track the proliferation of resonances by analyzing how matrix elements are decimated during iterative diagonalization. It helps identify regimes where resonances become sparse or dense.
Resonance Exponent $\theta(w)$
This exponent characterizes the density of resonances at a specific frequency scale $w$. If $\theta(w) > 0$, the system is localized; if it becomes negative, it signals an instability leading to resonance proliferation and eventual thermalization.
Delocalization Transition
This refers to the transition where a many-body localized state becomes delocalized in Hilbert space. The paper shows this transition is driven by the proliferation of many-body resonances, which are captured by the flow of $\theta(w)$.

Terminology

Summary

The gist The proliferation of many-body resonances drives slow numerical drifts towards delocalization in many-body localization models, and this phenomenon is captured by a statistical method that predicts the distribution of resonance oscillation frequencies

How it works

The core idea involves developing a statistical method to predict the distribution of resonance oscillation frequencies which captures how the formation of resonances at larger frequency scales subsequently affects the formation of resonances at lower frequencies Working within the statistical Jacobi approximation (SJA), we derive a flow equation for a power-law exponent θ(w) characterizing the density of resonances at frequency scale w. A localized phase is described by a line of fixed points with θ(w) > 0, while an instability of the flow signals resonance proliferation and the onset of thermalization. The predicted θ(w) matches numerics on the Anderson model on random regular graphs and the L´evy-RosenzweigPorter random matrix ensemble, both of which host resonance-driven delocalization transitions.

Resonances and the Jacobi Algorithm

Many-body resonances are understood within a dynamical picture where initial product states related by n spin flips require going to roughly the nth order of perturbation theory before a non-zero matrix element connecting them can be found. These slow Rabi oscillations between many-body states are the direct signature of many-body resonances. The Jacobi algorithm is used to track resonance proliferation by iteratively diagonalizing the Hamiltonian H through successive two-level rotations that decimate large off-diagonal matrix elements. The statistical Jacobi approximation (SJA) focuses on the distribution of decimated elements ρdec(w) and identifies two distinct regimes—sparse and dense—of the Jacobi flow.

Flow Equations for θ(w)

The SJA extracts the number of resonances with a frequency scale larger than w, nres(w), which is modeled as nres(w)/N ∼ A wθ + const. The resonance exponent θ distinguishes Hilbert-space-localized phases (θ > 0) from delocalized ones (θ < 0). The heuristic flow equation derived in Sec. III, dθ(w)/dw = f[θ(w)]w/θ(w)-1, predicts a transition from a phase in which resonances do not proliferate (localized) to a proliferating phase (thermalizing) signaled by an instability where θ(w) → −∞.

Application to Models

The heuristic flow equation is compared against three models: the L´evy-Rosenzweig-Porter random matrix model, the Anderson model on a random regular graph, and the disordered Heisenberg spin chain. Numerical results show qualitative agreement between the predictions of the heuristic flow equations and exact Jacobi diagonalization for these models within their regime of validity. The analysis predicts that for the toy MBL model with an initial θ(w0) > 0, a strong enough perturbation may cause it to flow to θ(w) < 0 and ultimately delocalize, suggesting that large systems with bare localization lengths ζ < ζc may thermalize at long times.

Observable Predictions

The SJA predicts qualitative features of the phase diagram of the LRP, RRG, and the toy model of locally perturbed MBL: The existence of both a stable localized and delocalized phase, separated by a sharp transition. The number of resonances per state above scale w, nres(w)/N [Eq. (13)], is a proxy for the participation ratio (PR), which is thus a diagnostic for the delocalization of states in Hilbert space. In the regime where θ(w) is slowly varying, well before it diverges and we cross over into the dense regime, we predict stretched exponential decay with a slowly varying exponent C(t) ∝ e(-(t/τ))β(t).

Discussion

The SJA treats all three of these systems identically, predicting that when resonances proliferate, θ flows to large and negative values as smaller Rabi frequencies are resolved. The analysis suggests that the transition to delocalization is driven by resonance proliferation, which is captured by the SJA. The analysis also predicts BKT-like scaling for the critical behavior of the heuristic flow equations, which contrasts with some other RG analyses [48–52]. The divergence of w−1c as w0 and θ0 are varied across the separatrix reflects the BKT scaling, giving a thermalization time that diverges faster than any power law.

The SJA also predicts the behavior of more traditional probes of eigenstate localization, such as eigenstate participation ratios, and dynamical signatures of thermalization, such as the functional form of return probabilities. The analysis suggests that prethermal MBL is controlled by resonance proliferation before being cut off by other effects like avalanches or finite system size.

The SJA predicts qualitative features of the phase diagram of the LRP, RRG, and the toy model of locally perturbed MBL: The existence of both a stable localized and delocalized phase, separated by a sharp transition. The SJA treats all three of these systems identically. It predicts the flow of an exponent θ(w) controlling resonances. When resonances proliferate, θ flows to large and negative values as smaller Rabi frequencies are resolved. When not, θ flows to a stable positive value, so that the localized phase is characterized by a line of fixed points with 0 < θ ≤ 1. The separatrix that flows to θ(0) = 0 defines the transition.

The SJA also predicts the behavior of more traditional probes of eigenstate localization, such as eigenstate participation ratios, and dynamical signatures of thermalization, such as the functional form of return probabilities. Many-body localized systems are not Hilbert-space localized. Their asymptotic instability to delocalization is also believed to be caused by thermal avalanches, not by many-body resonances, at least with random disorder. The SJA can describe dynamics in various regimes.

The SJA predicts qualitative features of the phase diagram of the LRP, RRG, and the toy model of locally perturbed MBL: The existence of both a stable localized and delocalized phase, separated by a sharp transition. The SJA treats all three of these systems identically. It predicts the flow of an exponent θ(w) controlling resonances. When resonances proliferate, θ flows to large and negative values as smaller Rabi frequencies are resolved.

Improvements for AI systems

  1. Incorporate a statistical method to predict resonance oscillation frequencies to capture how the formation of resonances at larger frequency scales subsequently affects the formation of resonances at lower frequencies, enabling systems to anticipate delocalization transitions driven by many-body resonances before they occur.

  2. Develop an AI system capable of deriving a flow equation for the power-law exponent θ(w) characterizing resonance density, allowing it to distinguish between localized phases (where "θ(w) > 0) and instability signals for thermalization (where an instability of the flow signals resonance proliferation").

  3. Apply the derived SJA flow equation to model dynamics in various systems, such as the L´evy-Rosenzweig-Porter random matrix model, allowing the AI to predict qualitative features like the existence of both a stable localized and delocalized phase, separated by a sharp transition.

  4. Implement an observable prediction module that relates the number of resonances per state above scale w, nres(w)/N [Eq. (13)], is a proxy for the participation ratio (PR), allowing the AI to use this metric as a diagnostic for Hilbert space delocalization and predict it will be equal to Ns for a uniformly delocalized state over Ns n = 0 basis states.

  5. Integrate real-time dynamics prediction by using the functional flow equation, which predicts stretched exponential decay with a slowly varying exponent, C(t) ∝ e(-(t/τ)β(t)) in the prethermal regime of MBL, allowing the AI to forecast autocorrelation functions and return probabilities.

  6. Create a system that can analyze critical behavior by comparing predictions from different flow equation derivations—the heuristic flow equation (predicting BKT class) versus the functional flow equation (predicting power-law exponents)—to assess which scaling is more accurate for specific models.

Abstract

Models of many-body localization (MBL) exhibit slow numerical drifts towards delocalization with increasing system size, for which no satisfactory theory exists. Numerical results indicate that these drifts are driven by the proliferation of many-body resonances at intermediate disorder strengths. We develop a statistical method to predict the distribution of resonance oscillation frequencies, accounting for how resonances formed at larger frequency scales affect the subsequent formation of resonances at lower frequencies. Working within the statistical Jacobi approximation (SJA), we derive a flow equation for a power-law exponent θ(w) characterizing the density of resonances at frequency scale w. A localized phase is described by a line of fixed points with θ(w)>0, while an instability of the flow signals resonance proliferation and the onset of thermalization. The predicted θ(w) agrees with numerical results for the Anderson model on random regular graphs and the Lévy--Rosenzweig--Porter random matrix ensemble, both of which host resonance-driven delocalization transitions. We further connect the flow to eigenstate properties such as the participation ratio and to dynamical observables such as the return probability. The predicted θ(w) also agrees with numerical measurements in real-space models of MBL at intermediate disorder strengths, providing a step towards explaining the finite-size drifts observed in MBL.

Sources

Related papers