Local picture and level-set percolation of the Gaussian free field on a large discrete torus

TL;DR

This paper studies the level-set percolation of the Gaussian free field on large discrete tori, approximating it by the infinite lattice case to analyze connectivity properties and phase transitions.

Contribution

It provides a new approximation of the Gaussian free field on large tori by the field on b Z^d, enabling detailed analysis of level-set percolation behavior.

Findings

Level sets above the critical level contain no large components with high probability.

Level sets below the critical level contain a giant component of diameter comparable to the torus.

Results extend understanding of phase transitions in Gaussian free field level sets.

Abstract

For $d \geq 3$ we obtain an approximation of the zero-average Gaussian free field on the discrete $d$-dimensional torus of large side length $N$ by the Gaussian free field on $\mathbb Z^d$, valid in boxes of roughly side length $N - N^\delta$ with $\delta \in (\frac12,1)$. As an implication, the level sets of the zero-average Gaussian free field on the torus can be approximated by the level sets of the Gaussian free field on $\mathbb Z^d$. This leads to a series of applications related to level-set percolation. We show that level sets of the zero-average Gaussian free field on the torus for levels $h > h_\star$ (where $h_\star$ denotes the critical value for level-set percolation of the Gaussian free field on $\mathbb Z^d$) with high probability contain no connected component of volume comparable to the total volume of the torus. Moreover, level sets with $h < h_\star$ with high probability contain a connected component of (extrinsic) diameter comparable to the torus diameter $N$. We also show that level sets of the zero-average Gaussian free field on the torus for levels $h$ above a second critical parameter $h_{\star\star}(\geq h_\star)$, again defined via the Gaussian free field on $\mathbb Z^d$, with high probability only contain connected components negligible in their size when compared to the size of the torus. Similar results have been obtained by A. Teixeira and D. Windisch in [Comm. Pure Appl. Math., 64(12):1599-1646, 2011] and J. \v{C}ern\'y and A. Teixeira in [Ann. Appl. Probab., 26(5):2883-2914, 2016] for the vacant set of simple random walk on a large discrete torus with the help of random interlacements on $\mathbb Z^d$, introduced by A.-S. Sznitman in [Ann. of Math. (2), 171(3):2039-2087, 2010].