A limit theorem for one-dimensional Gibbs measures under conditions on the empirical field (Q1208951)

Starting from a principle of large deviations for the \(n\)-cyclic empirical field \(L_ n\) of a stationary Gibbs measure \(P\) on \(\Omega=S^{\mathbb{Z}}\) with a finite set \(S\), the author proves a conditional limit theorem for this one-dimensional Gibbs measure. Let \(\Pi\) be an appropriate subset of stationary probability measures on \(\Omega\), such that there exists a unique entropy minimizing \(P^*\) in the closure of \(\Pi\). If this \(P^*\) is mixing, then the author proves that \(P[ \cdot\mid L_ n\in\Pi]\) converges to \(P^*\) as \(n\to\infty\). Since the conditional Gibbs distributions are not averaged and the boundary conditions are not assumed to be periodic [for results in this direction see \textit{H.-O.\ Georgii}, Large deviations and maximum entropy principle for interacting random fields on \({\mathbb{Z}}^ d\) (to appear in Ann.\ Probab.)], the author needs a sufficiently strong decay of the interaction potential.