🔢 mathematics

Many-body Localization and Poisson statistics in the Quantum Sun model

This paper proves that the Quantum Sun model, a genuine many-body system of interacting spins with distance-dependent coupling strengths, exhibits localization and Poissonian spectral statistics in the regime where the coupling parameter $\alpha$ is significantly smaller than the conjectured critical value of $1/\sqrt{2}$ .

Original authors: Wojciech De Roeck, Amirali Hannani

Published 2026-08-18

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Wojciech De Roeck, Amirali Hannani

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the quantum world, matter does not always behave the way we expect from everyday objects. When a particle moves through a material filled with random impurities, it can sometimes get stuck, unable to travel freely. This phenomenon, known as localization, means the particle's energy remains trapped in a small region rather than spreading out. For decades, physicists have wondered if this trapping effect could survive when many particles interact with one another. In a system with billions of particles, the constant jostling and influence between them usually act like a chaotic mixer, scrambling energy and allowing it to flow. The central question has been whether a sufficiently disordered environment can stop this mixing entirely, keeping the system frozen in a specific state forever, even as time goes on. This is not just a theoretical curiosity; understanding this boundary between frozen and fluid states is crucial for explaining why some materials conduct electricity while others do not, and for the future of quantum computing, where maintaining a stable state is essential.

A team of researchers has now provided the first rigorous proof that such a frozen state can indeed exist in a genuine many-body system, far from the edges of the energy spectrum where such effects are usually easier to find. They studied a specific mathematical model called the Quantum Sun, which imagines a small, chaotic "bath" of spins connected to a long chain of other spins. The spins in the chain are arranged so that those closer to the bath interact strongly, while those further away interact with a strength that drops off rapidly. The researchers focused on the regime where this interaction is very weak. They demonstrated that under these conditions, the system does not thermalize or mix; instead, it remains localized. Every energy state in the system is essentially a slight modification of a state where the particles were not interacting at all. The particles stay put, and the system retains a memory of its initial configuration indefinitely.

The proof relies on showing that the energy levels of the system behave in a very specific, random way. In a chaotic, mixing system, energy levels tend to repel each other, creating a regular pattern. In a localized system, however, the energy levels are independent of one another, appearing in a completely random sequence known as Poisson statistics. The authors proved that for their model, when the coupling strength is below a certain small threshold, the energy levels follow this random pattern perfectly. This statistical signature confirms that the system is truly localized and not just temporarily stuck. Their work also addresses a major concern in the field known as the "avalanche" instability. The fear was that even a tiny, chaotic region within a larger frozen system could act as a seed, eventually heating up and melting the surrounding frozen material, causing the entire system to become fluid. The researchers showed that in their one-dimensional model, if the interaction is weak enough, this avalanche never starts. The chaotic seed remains isolated, and the surrounding disorder holds firm.

To reach this conclusion, the team had to overcome significant mathematical hurdles. Unlike simpler models where particles do not interact, their system involves terms that do not commute, meaning the order in which you apply the interactions matters, making the spectrum of energy levels difficult to calculate directly. They developed a new method to track how the energy levels evolve as the system grows larger. They showed that as the chain gets longer, any small groups of energy levels that might have been close enough to interact and cause a meltdown are almost certain to be pulled apart by the randomness of the system. They proved that these potential "resonances," which could trigger the avalanche, dissolve with high probability as the system size increases. This dissolution ensures that the chaotic bath cannot influence the distant parts of the chain, preserving the localized state.

The result is a solid mathematical confirmation of a stable phase in a many-body quantum system. It proves that for a specific class of models, disorder can win over interaction, keeping the system in a non-thermal state. This finding supports the idea that one-dimensional systems with strong disorder can be stable against the avalanche mechanism, at least when the interactions are sufficiently weak. While the model is a simplified representation of reality, the rigorous nature of the proof offers a clear benchmark for what is possible in quantum mechanics. It shows that the transition between a frozen, localized world and a fluid, thermal one is not just a matter of simulation or approximation, but a real, provable feature of quantum systems. The work does not claim to solve the mystery of localization in all dimensions or all materials, but it firmly establishes that the stable, frozen side of the picture exists and can be understood with mathematical certainty.

Technical Summary: Many-body Localization and Poisson Statistics in the Quantum Sun Model

Problem Statement
The paper addresses the rigorous characterization of the Many-Body Localization (MBL) transition in a genuine many-body quantum system. While MBL is widely believed to exist in one-dimensional disordered systems, rigorous mathematical proofs have historically been scarce, often limited to spectral edges or mean-field approximations where the effective Hilbert space dimension does not grow exponentially. The specific challenge is to establish localization and Poissonian spectral statistics for a model where interactions are non-trivial and the system is considered away from spectral edges, in a regime where the Hilbert space dimension scales as $2^n$ .

The authors focus on the "Quantum Sun" model, a toy model designed to probe the stability of MBL against "avalanche" instabilities. In this model, a small ergodic bath (the "Sun") is coupled to a chain of localized spins. The coupling strength decays geometrically with distance from the bath ( $\alpha^n$ ). It is conjectured that for $\alpha > 1/\sqrt{2}$ , the system delocalizes via an avalanche mechanism, whereas for $\alpha < 1/\sqrt{2}$ , it remains localized.

Methodology
The authors employ a combination of analytic perturbation theory, probabilistic estimates, and tools from potential theory to analyze the spectrum and eigenstates of the Hamiltonian $H_n$ .

Iterative Spectral Analysis: The proof proceeds by analyzing the spectrum as the system size $n$ increases. The authors define a filtration of $\sigma$ -algebras and track the evolution of eigenvalues from scale $n$ to $n+1$ . The spectrum of $H_{n+1/2}$ (obtained by adding a random field term) is a union of shifted copies of the spectrum of $H_n$ . The full Hamiltonian $H_{n+1}$ is treated as a perturbation of $H_{n+1/2}$ .
Resonant Patches and Dissolution: A central concept is the "resonant patch," a cluster of eigenvalues separated by gaps smaller than a threshold $\alpha^{\theta n}$ . The core of the proof involves showing that these patches "dissolve" (break into singletons) as $n$ increases. This is achieved by analyzing the spectral width of a patch as a function of the random coupling variable $g_n$ .
Inequalities from Potential Theory: To prove that patches dissolve with high probability, the authors utilize the Remez inequality and the Cartan inequality. These tools bound the measure of the set of disorder realizations where the spectral width of a patch remains small (i.e., where the patch fails to dissolve). The authors construct an analytic function related to the spectral width and apply these inequalities to show that the probability of a patch surviving multiple scales decays rapidly.
Factorization of Moments: To establish Poisson statistics, the authors compute the moments of the eigenvalue counting function. They introduce a "semi-perturbed" process where the interaction is partially removed, allowing for a factorization of the joint probability density functions of eigenvalues.
- Typical vs. Atypical Configurations: They define "typical" spin configurations where the covariance between energy terms vanishes asymptotically. For typical configurations, the joint distribution factorizes into independent Gaussians (via Local Central Limit Theorem estimates and the Brascamp-Lieb inequality).
- Error Control: The contribution of "atypical" configurations (where factorization fails) and "resonant/anti-resonant" pairs is shown to be negligible. This requires proving that the number of resonant patches and anti-resonant pairs vanishes in the thermodynamic limit (Theorem 3 and Theorem 16).

Key Contributions and Results

Fock-Space Localization (Theorem 1): For sufficiently small coupling $\alpha \ll 1/\sqrt{2}$ , the authors prove that the system is localized. Specifically, they show that eigenstates are predominantly supported on a few eigenstates of the uncoupled system ( $\alpha=0$ ). This is quantified by the Inverse Participation Ratio (IPR), which is shown to be bounded by a constant $2^{\ell^*}$ (where $\ell^*$ is a random variable with an exponential tail), rather than scaling with the full Hilbert space dimension $2^n$ .
Poisson Spectral Statistics (Theorem 2): The paper proves that the spectral statistics of the Quantum Sun model converge to a Poisson point process. This is the first rigorous proof of Poisson statistics in a genuine many-body model away from spectral edges. The proof relies on the convergence of the moment generating function of the eigenvalue counting process to that of a Poisson process.
Dissolution of Resonances (Theorem 3): A crucial intermediate result establishes that the fraction of eigenvalues belonging to non-singleton resonant patches vanishes almost surely as $n \to \infty$ . This "dissolution" mechanism ensures that the interactions are effectively perturbative for most levels, allowing the spectral statistics to be governed by the non-interacting (random field) limit.
Anti-Resonance Control (Theorem 16): The authors prove that the number of "anti-resonant" pairs (eigenvalues $E_\mu, E_{\mu'}$ such that $|E_\mu + E_{\mu'}|$ is small) also vanishes relative to the total number of states. This is necessary to break the reflection symmetry that would otherwise prevent Poisson statistics at the origin.

Significance and Claims
The authors claim that this work provides the first rigorous realization of the "stable" side of the MBL phase transition picture in a genuine many-body model.

Rigorous MBL: Unlike previous results that were limited to spectral edges or mean-field models, this result holds for the bulk of the spectrum in a model with an exponentially large Hilbert space.
Avalanche Stability: The results support the heuristic picture that in one dimension, for sufficiently strong disorder (small $\alpha$ ), a local ergodic inclusion (the bath) does not trigger an avalanche that thermalizes the entire system. The localization is robust against the specific interaction structure of the Quantum Sun model.
Poisson Statistics: The demonstration of Poisson statistics confirms that the eigenstates are uncorrelated in a way consistent with localization, distinct from the Wigner-Dyson (GOE/GUE) statistics expected in the delocalized phase.

The paper explicitly avoids claiming that these results prove MBL in generic high-dimensional systems or for all interaction types. Instead, it establishes the existence of a localized phase with Poisson statistics in a specific, hierarchically structured model that serves as a rigorous testbed for the avalanche instability mechanism. The technical novelty lies in adapting tools from one-body localization (Cartan/Remez inequalities) to the many-body context and handling the combinatorial complexity of the many-body spectrum.

More like this