Operadic structure of boundary conditions for two-dimensional Markov Gaussian random fields on the lattice

Abstract

The theory of Markov processes on the square lattice has been given recently by the second author a new algebraic description in terms of operads. In particular, this new approach allows for a nice description of invariant boundary conditions and infinite-volume Gibbs measures. This theory comes with new algebraic objects which have not been constructed on any non trivial model yet. In this article, the main objective is to exhibit and understand these structures in the particular case of Gaussian Markov fields on the two-dimensional square lattice. This article, in the Gaussian framework, is the first time where all the operadic constructions -- products and eigen-elements up to morphisms -- introduced by the second author are defined rigorously. We also relate these constructions to more classical approach such as the transfer matrix of statistical mechanics and the Fourier transform. The description of half-strips and corners is new and requires the introduction of new operations such as folding. From the probabilistic point of view, we also show that the operadic products on the boundaries are not easily defined and most operations are lifted to the level of parameter spaces, here quadratic forms through Schur complements.