Unique Bernoulli g-measures

TL;DR

This paper establishes new, improved conditions for the uniqueness of g-measures and proves that these measures have Bernoulli properties, extending previous results and providing convergence in the Wasserstein metric.

Contribution

The paper introduces new sufficient conditions for the uniqueness and Bernoulli property of g-measures, generalizing and strengthening prior results in the field.

Findings

Established new conditions for g-measure uniqueness.

Proved Bernoulli property under these conditions.

Showed convergence of transfer operator iterates in Wasserstein metric.

Abstract

We improve and subsume the conditions of Johansson and \"Oberg [18] and Berbee [2] for uniqueness of a g-measure, i.e., a stationary distribution for chains with complete connections. In addition, we prove that these unique g-measures have Bernoulli natural extensions. In particular, we obtain a unique g-measure that has the Bernoulli property for the full shift on finitely many states under any one of the following additional assumptions. (1) $$\sum_{n=1}^\infty (\var_n \log g)^2<\infty,$$ (2) For any fixed $\epsilon>0$, $$\sum_{n=1}^\infty e^{-(\{1}{2}+\epsilon) (\var_1 \log g+...+\var_n \log g)}=\infty,$$ (3) $$\var_n \log g=\ordo{\{1}{\sqrt{n}}}, \quad n\to \infty.$$ That the measure is Bernoulli in the case of (1) is new. In (2) we have an improved version of Berbee's condition (concerning uniqueness and Bernoullicity) [2], allowing the variations of log g to be essentially twice as large. Finally, (3) is an example that our main result is new both for uniqueness and for the Bernoulli property. We also conclude that we have convergence in the Wasserstein metric of the iterates of the adjoint transfer operator to the g-measure.