Construction of Gibbs measures for 1-dimensional continuum fields (Q2502252)

This paper proves existence of Gibbs measures of \(\mathbb{R}^d\)-valued random fields \(X(s), s \in \mathbb{R}\), of Ginzburg-Landau type. The (formal) Hamiltonian consists of three terms; half the integrated squared gradient of the path, an integrated external potential of the type \(\int \varphi(X(s))\, ds\), and an interaction term of the type \(\int\int \psi(s-t,X(s),X(t))\, ds\, dt\). The external potential is decomposed as \(\varphi = \varphi_0 + \varphi_1\), and the Gibbs measure associated to the Hamiltonian is defined as a perturbation of the Gibbs measure described by the gradient term and the \(\varphi_0\)-potential, i.e.\ a perturbation of the solution to a stochastic differential equation. The essential regularity condition requires \(p_0 \geq q_0\), where \(2p_0\) is the growth order of the external potential and \(q_0+1\) is the growth order of the interaction. Under this growth condition a localization result is proved, which leads to the existence of the infinite volume Gibbs measure via the DLR equation. The proof employes moment estimates via integration by parts technique. No assumptions of convexity of the interactions are used, and arbitrary strength of the interactions in the case \(p_0 > q_0\) are allowed. In particular Nelson's scalar field model is included. The last section of the paper is devoted to the proof of existence of hard-wall Gibbs measures, i.e., with \(\varphi(x)=\infty\) for \(x < 0\) hence forcing the path to be positive. The proof employes the derived localization result.