One-dimensional discrete Gaussian Markov processes: Harmonic decomposition of invariant boundary conditions

Abstract

We study invariant boundary conditions for one dimensional discrete Gaussian Markov processes, basic toy models of spatial Markov processes in statistical mechanics. More precisely, we give a decomposition of boundary objects in a non trivial basis from the study of a meromorphic matrix-valued function $\Phi$ (inherent to the model) and its singularities. This provides a simple algorithm for the explicit computation of invariant measures. As an application, we give an "eigen" version of Szegő limit theorem for matrix valued trigonometric polynomials.