Equilibrium and nonequilibrium Gibbs states on sofic groups (Q7018136)

This paper is a contribution to the extension of quite classical notions from the thermodynamic formalism from the setting of \(\mathbb{Z}\)-actions to the (possibly completely general) setting of countable sofic group actions. For a finite alphabet \(A\) and a countable sofic group \(\Gamma\), there is an associated shift \(\Gamma\)-action defined by \((g\cdot x)_h=x_{g^{-1}h}\) for points \((x_h)\) in the compact space \(A^{\Gamma}\). A continuous function \(u\colon A^{\Gamma}\to\mathbb{R}\) is used here to encode formal notions of temperature or energy, and two notions of how a \(\Gamma\)-invariant probability measure \(\mu\) may describe the state of the system in statistical equilibrium are defined: \(\mu\) is called a `\(\Gamma\)-equilibrium state' if it maximizes the sofic pressure \(h(\mu)-\int d{\mathrm{d}}\mu\) (\(h\) denoting the sofic entropy with respect to some sofic approximation to \(\Gamma\)); \(\mu\) is called a `Gibbs measure' with respect to some potential if it satisfies a local equilibrium criterion. \par The first results in this direction going beyond the case \(\Gamma=\mathbb{Z}\) showed that for a sufficiently regular function \(u\) the set of equilibrium states coincides with the set of Gibbs states if \(\Gamma\) is amenable (see the work of \textit{H.-O. Georgii} [Gibbs measures and phase transitions. 2nd extended ed. Berlin: de Gruyter (2011; Zbl 1225.60001)] for the \(\mathbb{Z}^d\) case and \textit{A. A. Tempel'man} [Sov. Math., Dokl. 22, 363--369 (1980; Zbl 0493.60061); translation from Dokl. Akad. Nauk SSSR 254, 297--302 (1980)] in general). \textit{S. Barbieri} and \textit{T. Meyerovitch} showed that every equilibrium state is a Gibbs state if \(\Gamma\) is sofic and \(u\) satisfies a technical criterion [Trans. Am. Math. Soc. 376, No. 2, 1299--1342 (2023; Zbl 1520.37013)]. A natural question in the converse direction is to ask if every Gibbs state is an equilibrium state for a suitable type of interaction, and any reasonably ambitous notion of `suitable' would certainly include the Ising model considered here. There are relatively low-hanging examples of the Ising model on a free group are known to show that the converse cannot hold. Here a more ambitious partial converse is considered, restricting to Gibbs states with finite entropy, and shown not to be true in a very strong sense. This shows in particular that the converse also cannot hold with additional reasonable assumptions on the sofic approximation, for example. Along the way methods are developed to show that some states are equilibrium, showing \textit{inter alia} that the plus and minus Ising states on a free group have the same sofic entropy over every sofic approximation.