Ising critical behavior of inhomogeneous Curie-Weiss models and annealed random graphs

TL;DR

This paper analyzes the critical behavior of inhomogeneous Curie-Weiss models and annealed random graphs, revealing how the distribution of vertex weights influences phase transition properties and total spin fluctuations.

Contribution

It identifies critical temperatures, exponents, and limit theorems for inhomogeneous Curie-Weiss models with general weight distributions, extending classical results to inhomogeneous settings.

Findings

Critical behavior depends on the number of finite moments of the weight distribution.

When the fourth moment is finite, the model behaves like the classical Curie-Weiss model.

For power-law weights with exponent τ in (3,5), critical exponents depend on τ, and total spin scales differently.

Abstract

We study the critical behavior for inhomogeneous versions of the Curie-Weiss model, where the coupling constant $J_{ij}(\beta)$ for the edge $ij$ on the complete graph is given by $J_{ij}(\beta)=\beta w_iw_j/(\sum_{k\in[N]}w_k)$. We call the product form of these couplings the rank-1 inhomogeneous Curie-Weiss model. This model also arises (with inverse temperature $\beta$ replaced by $\sinh(\beta)$) from the annealed Ising model on the generalized random graph. We assume that the vertex weights $(w_i)_{i\in[N]}$ are regular, in the sense that their empirical distribution converges and the second moment converges as well. We identify the critical temperatures and exponents for these models, as well as a non-classical limit theorem for the total spin at the critical point. These depend sensitively on the number of finite moments of the weight distribution. When the fourth moment of the weight distribution converges, then the critical behavior is the same as on the (homogeneous) Curie-Weiss model, so that the inhomogeneity is weak. When the fourth moment of the weights converges to infinity, and the weights satisfy an asymptotic power law with exponent $\tau$ with $\tau\in(3,5)$, then the critical exponents depend sensitively on $\tau$. In addition, at criticality, the total spin $S_{N}$ satisfies that $S_{N}/N^{(\tau-1)/(\tau-2)}$ converges in law to some limiting random variable whose distribution we explicitly characterize.