Many-Body Localization: Evidence and Asymptotic Claims

Many-body localization (MBL) asks whether an interacting, disordered, isolated quantum system can retain local memory indefinitely instead of thermalizing. Long-lived memory at accessible size and time is well supported in many settings. The stronger claim of a stable phase in asymptotic limits requires additional assumptions and remains a central point of debate.

Helpful background. Many-body localization owns the stable conceptual framework; ETH supplies the thermal comparison.

The order of limits is the research problem

A density-wave experiment often measures an imbalance

$$I(t)=\frac{N_{\rm even}(t)-N_{\rm odd}(t)} {N_{\rm even}(t)+N_{\rm odd}(t)}.$$

A nonzero plateau over a finite observation window establishes persistent memory on that window. It does not alone determine $\lim_{t\to\infty}\lim_{L\to\infty} I(t)$, nor establish the same result after weak coupling to an environment.

Established: strong disorder can produce very slow relaxation, anomalous entanglement growth, and localized-regime phenomenology. Abanin et al. (2019) review that framework. Active and debated: when this finite-scale regime represents a stable thermodynamic phase, especially in relation to rare thermal regions and finite-size drifts. Sierant et al. (2025) emphasize why naive one-parameter scaling can be misleading. Different models, dimensions, disorder distributions, and interaction ranges need separate assessments.

Competing explanations for a plateau

A localized description invokes emergent quasi-local conserved quantities. A slow-thermalization description may instead involve rare regions, bottlenecks, or long crossover scales. Quasiperiodic potentials remove some randomness-induced rare-region mechanisms but introduce a different model; they should not be pooled with independently random disorder without justification.

Level-spacing ratios, eigenstate entanglement, transport, and imbalance constrain different aspects of these pictures. Each requires energy-window and symmetry-sector control. Long-time tensor-network calculations can artificially suppress entanglement if their bond dimension is insufficient, while exact diagonalization can overemphasize small-size crossings.

Starter project: expose a drifting transition estimate

Study a random-field spin chain at fixed total magnetization. Compute a spectral statistic and an eigenstate entanglement diagnostic over several small sizes and many disorder samples. Keep the energy-density window fixed in a documented way. Compare crossing locations from neighboring size pairs rather than fitting all data immediately to one critical point.

A workstation supports an initial exact-diagonalization study; dense diagonalization becomes expensive rapidly. Deliver sample-level data, bootstrap uncertainty across disorder realizations, crossing-drift plots, and sensitivity to excluding the smallest sizes. Compare a random potential with a quasiperiodic potential only as two distinct model families.

The failure test is a precise critical point that shifts outside its quoted uncertainty when a size is omitted. A positive outcome may be a demonstrated inability to distinguish two asymptotic scenarios. That is a useful result if the study identifies the additional sizes, times, or observables required.

Research extensions and boundaries

A focused extension couples a small thermal region to a disordered chain and studies how relaxation changes with its size and coupling. Another examines how weak external dephasing alters a finite-time memory diagnostic. Both require an explicit bath model and a distinction between bath-induced relaxation and intrinsic instability.

Do not announce that all MBL is disproved from a drifting small-system crossing, or that an observed plateau proves an infinite-system phase. A strong paper narrows the model and the claim enough that the evidence discriminates between mechanisms.

Entry points and evidence horizon

Read the 2019 review for the localized description and the 2025 review for numerical limitations and unresolved asymptotics. This deliberately includes differing emphases. Literature links were checked on 2026-10-11; the cited evidence horizon is 2025.

References

• Abanin, D. A., Altman, E., Bloch, I., and Serbyn, M., “Colloquium: Many-body localization, thermalization, and entanglement,” Reviews of Modern Physics 91, 021001 (2019), doi:10.1103/RevModPhys.91.021001.
• Sierant, P., Lewenstein, M., Scardicchio, A., Vidmar, L., and Zakrzewski, J., “Many-body localization in the age of classical computing,” Reports on Progress in Physics 88, 026502 (2025), arXiv:2403.07111.