Stationary determinantal processes: phase multiplicity, Bernoullicity, entropy, and domination (Q1430446)

This paper focuses on stationary processes indexed by \(\mathbb{Z}^d\) defined via determinant of Toeplitz matrices called discrete stationary determinantal processes. Determinantal probability measures can be found in several areas of mathematics and theoretical physics like fermionic point processes, random matrix theory and part of probability theory. The author give here a detailed analysis of the discrete stationary case. They describes relations with the theory of Toeplitz determinant and insist on the important role of the geometric mean of a parameter function \(f\) defining the measures via some related determinant. They also detail a wide family of examples in combinatorial models (uniform spanning trees, dimer models) and a property of negative association that is usually difficult to study directly. They prove that all discrete stationary determinantal processes is a Bernoulli shift, although only measurability rather than smoothness of the parameter \(f\) is required, that have the usual K-property for \(d>1\) and a strong full K-property in some one-dimensional models. This strong full K-regime appears to be analogous to phase uniqueness in statistical mechanics. The author also gives a characterization of this property where both the rate and the position at which \(f\) approaches 0 and 1 matters. The proofs use sharp stochastic domination of determinantal probability measures between product measures with the geometric mean of \(f\) as deterministic rate, and estimations of entropies. The results are illustrated by several examples from different fields, i.i.d. Bernoulli measures, uniform spanning trees, random cluster model and renewal processes among others.