The low-temperature phase of Kac-Ising models (Q1285044): Difference between revisions

The authors consider the Ising model in dimension \(d\geq 2\) with the so-called Kac interaction, where spins \(\sigma_i\) and \(\sigma_j\), \(i,j\in{\mathbb Z}^d\), interact as \(J_\gamma(i-j)\sigma_i\sigma_j\), \(J_\gamma(i)\equiv\gamma^dJ(\gamma i)\) being a real-valued function satisfying \(\int_{{\mathbb R}^d}J(x) dx=1\). The models of this type were introduced in 1963 by \textit{M. Kac, G. Uhlenbeck}, and \textit{P. C. Hemmer} as microscopic models for the van der Waals theory of the liquid-gas phase transition. In 1966 \textit{J. Lebowitz} and \textit{O. Penrose} proved rigorously that the \(\nu\)-dimensional version of the Kac-Uhlenbeck-Hemmer model indeed yields, in the limit \(\gamma\to 0\), the van der Waals-Maxwell thermodynamic theory. The present paper develops the statistical mechanical theory of local mesoscopic observables (i.e., all functions expressed in terms of the averaged magnetizations \(m_x(\sigma)=l^{-d}{\displaystyle\sum_{i:| i-lx| \leq l/2}\sigma_i}\), \(l\) being a mesoscopic scale \(1\ll l\ll \gamma^{-1}\)) for small positive values of \(\gamma\). A one-dimensional case was studied previously by \textit{M. Cassandro, E. Orlandi} and \textit{E. Presutti} [Probab. Theory Relat. Fields 96, No. 1, 57-96 (1993; Zbl 0791.60096)]. The authors develop the Pirogov-Sinai theory for the long range interactions based on a properly defined notion of a coarse-grained Peierls contour. The latter is constructed via a two step blocking procedure, the first on the Kac scale \(l\) and the second on the interaction scale \(\gamma^{-1}\). Such a definition results in a contour that is thick enough to decouple its core from its complement and thin enough to give a good control on the energy (positivity of the Peierls constant). Using this approach, the authors obtain the following upper bound for the critical temperature \(\beta_c(\gamma)\): \(\beta_c(\gamma)\leq 1+\gamma^\kappa\) with \(\kappa\geq(1-\varepsilon)/(2d+2)(1+1/d)\) for any \(\varepsilon>0\). This together with the known inequality \(\beta_c(\gamma)\geq 1\) proves that \(\beta_c(\gamma)\to 1\) as \(\gamma\to 0\).