Columnar order in random packings of $2\times2$ squares on the square lattice

TL;DR

This paper proves that large-scale random packings of 2x2 squares on a lattice tend to form columnar order, breaking certain symmetries, and introduces a new method for analyzing such measures.

Contribution

It establishes the existence of four extremal Gibbs measures with columnar order and extends the chessboard estimate to infinite-volume measures.

Findings

Large $\lambda$ packings exhibit columnar order.

Four extremal Gibbs measures break lattice symmetry.

Correlation decay rates differ along symmetry-preserving and breaking directions.

Abstract

We study random packings of $2\times2$ squares with centers on the square lattice $\mathbb{Z}^{2}$, in which the probability of a packing is proportional to $\lambda$ to the number of squares. We prove that for large $\lambda$, typical packings exhibit columnar order, in which either the centers of most tiles agree on the parity of their $x$-coordinate or the centers of most tiles agree on the parity of their $y$-coordinate. This manifests in the existence of four extremal and periodic Gibbs measures in which the rotational symmetry of the lattice is broken while the translational symmetry is only broken along a single axis. We further quantify the decay of correlations in these measures, obtaining a slow rate of exponential decay in the direction of preserved translational symmetry and a fast rate in the direction of broken translational symmetry. Lastly, we prove that every periodic Gibbs measure is a mixture of these four measures. Additionally, our proof introduces an apparently novel extension of the chessboard estimate, from finite-volume torus measures to all infinite-volume periodic Gibbs measures.