Uniqueness and tube property for the Swiss cheese large deviations
Dirk Erhard
Universidade Federal da Bahia, Salvador, BrazilJulien Poisat
Université Paris-Dauphine, France

Abstract
We consider the simple random walk on the Euclidean lattice, in three dimensions and higher, conditioned to visit fewer sites than expected, when the deviation from the mean scales like the mean. The associated large deviation principle was first derived by van den Berg, Bolthausen and den Hollander [Ann. of Math. 153 (2001)] in the continuous setting, that is, for the volume of a Wiener sausage, and later taken up by Phetpradap (2011) in the discrete setting. One of the key ideas in their work is to condition the range of the random walk on a certain skeleton, that is, a subsequence of the random walk path along an appropriate intermediate scale. In this paper we answer two questions that were left open. First, we prove using analytical tools for quasilinear elliptic equations that the rate function obtained by van den Berg, Bolthausen and den Hollander has a unique minimizer over the set of probability measures modulo shifts, for almost every deviation in dimension three and for deviations below a certain threshold in dimensions four and higher. Secondly, we adapt a compactification method for the space of probability measures that was introduced by Mukherjee and Varadhan (2016) in order to (i) state a strong large deviation principle for the skeleton occupation measure under the small range constraint and (ii) prove that the latter converges, under the conditioned law, to the minimizer of the rate function modulo shifts.
Cite this article
Dirk Erhard, Julien Poisat, Uniqueness and tube property for the Swiss cheese large deviations. J. Eur. Math. Soc. (2026), published online first
DOI 10.4171/JEMS/1836