Unique Bernoulli Gibbs states and g-measures
Paul Hulse
TL;DR
The work provides a sufficient condition under which Gibbs states of a one-dimensional, shift-invariant specification are $g$-chains for a continuous $g$, and uses this to derive criteria guaranteeing a unique, shift-invariant, Bernoulli Gibbs measure. The methodology centers on a ratio-bound framework that translates Gibbs-state structure into a $g$-measure problem via a limit $g_{\mathcal{V}}$, combined with lacunary-variation controls that ensure Bernoullicity through results of Johansson–Öberg–Pollicott. For interaction potentials $\Phi$ with summable variation, the paper yields explicit conditions implying uniqueness and Bernoulli mixing, strengthening classical results (Dobrushin, Ruelle, Coelho–Quas). It also discusses weaker, non-Bernoulli uniqueness results, clarifying the trade-offs between uniqueness and strong mixing in low-dimensional lattice systems. Overall, the results provide practically checkable criteria for uniqueness and Bernoulli behavior in 1D Gibbsian systems, with implications for models like the Ising chain.
Abstract
A sufficient condition for the Gibbs states of a shift-invariant specification on a one-dimensional lattice to be the $g$-chains for some continuous function $g$ is obtained. This is then used to derive criteria under which there is a unique Gibbs state, which is also shift-invariant and Bernoulli.
