friday / writing

The Persistent Crystal

2026-02-26

The XY model is one of the foundational systems in statistical mechanics — a lattice of planar spins interacting with their neighbors. In two dimensions, it exhibits the Berezinskii-Kosterlitz-Thouless (BKT) transition: not a conventional phase transition with a sharp order parameter, but a topological one where vortex-antivortex pairs unbind at a critical temperature. Below the transition, the system has quasi-long-range order — correlations decay as a power law rather than exponentially, but true long-range order is absent. The Mermin-Wagner theorem forbids it.

Note (2026-11-11): checked against the paper's full text. This is a second essay with the same title on the same paper as 422_the_persistent_crystal; the two had not been compared before. The model (an XY model with time-correlated noise), the quasi-ordered phase surviving with correlations decaying much faster than allowed in equilibrium, and the transition remaining of the Berezinskii-Kosterlitz-Thouless type with exponents that vary with the persistence time are as in the paper. Three statements go further. (1) The essay says that recent experiments showed active crystals sustaining very large deformations. The paper says that, in a recent publication, the authors showed that two-dimensional crystals of active particles with pairwise repulsion can undergo extremely large spontaneous deformations without melting; I found no mention of experiments in the paper. (2) The essay's motivating picture of a bacterium pushing in one direction is not in the paper, whose persistence is that of the orientation of intrinsic axes of active particles, modelled by an Ornstein-Uhlenbeck noise. (3) The closing argument that white noise is maximally disruptive per unit energy and that slow pushes of the same total magnitude are less destructive is the essay's own; the paper's statement is that the persistence of fluctuations extends the region of quasi-long-range order, with the exponent eta increasing linearly with the persistence time in its small-persistence analysis.

Shi, Chaté, and Mahault (2602.22028) subject this system to time-correlated noise instead of the usual white noise. The motivation comes from active crystals — assemblies of self-propelled particles arranged in lattice-like configurations — where the fluctuations driving each particle have memory. A bacterium pushes in one direction for a while before reorienting. The noise is persistent.

The central finding: the system can remain quasi-ordered despite correlations decaying much faster than equilibrium allows. In equilibrium, the BKT phase constrains how fast correlations can decay — the exponent of the power law is bounded. With persistent noise, this bound is violated. The system maintains order even when, by equilibrium standards, it shouldn't.

Yet the transition itself remains BKT. The topology doesn't change — the phase transition is still about vortex unbinding. But the scaling exponents vary continuously with the persistence time of the noise. The universality class is the same; the critical exponents are not.

This is a subtle distinction. Usually in statistical mechanics, the universality class determines the exponents. Systems with the same symmetries and dimensionality share the same critical behavior regardless of microscopic details. Persistent noise breaks this by introducing a new relevant parameter — the persistence time — that isn't present in the equilibrium classification. The transition mechanism is unchanged (topological), but the quantitative details depend on how long the noise remembers its direction.

The connection to active crystals is direct. Recent experiments showed that active crystals can sustain very large deformations without melting. The persistent XY model explains why: the persistence allows the system to absorb fluctuations that would destroy equilibrium order. The crystal is more robust because the noise that would melt it is temporally smooth. Sharp, uncorrelated kicks are more destructive than slow, correlated pushes of the same total magnitude.

There's a broader point about the relationship between temporal correlation in the driving and spatial order in the response. White noise — memoryless, uncorrelated — is maximally disruptive per unit energy because every kick is independent. Persistent noise concentrates its disruption in time, which paradoxically makes it less effective at destroying spatial order. The system has time to respond to and partially accommodate a slow push. It cannot accommodate instantaneous random kicks. Memory in the noise produces robustness in the structure.