Bounding the partition function of spin-systems

TL;DR

This paper derives an upper bound for the partition function of certain spin systems on regular bipartite graphs, extending previous bounds to more general weight configurations and models.

Contribution

It generalizes existing bounds on partition functions to broader weight collections and models, using list homomorphisms and second-moment methods.

Findings

Provides an upper bound independent of graph size for regular bipartite graphs.

Extends previous bounds to non-uniform weights and more complex models.

Utilizes a generalized approach to graph homomorphisms and second-moment calculations.

Abstract

With a graph $G=(V,E)$ we associate a collection of non-negative real weights $\cup_{v\in V}{\lambda_{i,v}:1\leq i \leq m} \cup \cup_{uv \in E} {\lambda_{ij,uv}:1\leq i \leq j \leq m}$. We consider the probability distribution on ${f:V\rightarrow{1,...,m}}$ in which each $f$ occurs with probability proportional to $\prod_{v \in V}\lambda_{f(v),v}\prod_{uv \in E}\lambda_{f(u)f(v),uv}$. Many well-known statistical physics models, including the Ising model with an external field and the hard-core model with non-uniform activities, can be framed as such a distribution. We obtain an upper bound, independent of $G$, for the partition function (the normalizing constant which turns the assignment of weights on $\{f:V\rightarrow{1,...,m\}}$ into a probability distribution) in the case when $G$ is a regular bipartite graph. This generalizes a bound obtained by Galvin and Tetali who considered the simpler weight collection $\{\lambda_i:1 \leq i \leq m\} \cup \{\lambda_{ij}:1 \leq i \leq j \leq m\}$ with each $\lambda_{ij}$ either 0 or 1 and with each $f$ chosen with probability proportional to $\prod_{v \in V}\lambda_{f(v)}\prod_{uv \in E}\lambda_{f(u)f(v)}$. Our main tools are a generalization to list homomorphisms of a result of Galvin and Tetali on graph homomorphisms and a straightforward second-moment computation.