Isomorphism theorems for Markov chains. (Q2574639)

Let \(p_ {t}(x,y)\) be a~transition density of a~symmetric Markov process on a~state space \(E\) such that the Green function \(g(x,y) = \int ^ \infty _ 0 p_ {t}(x,y)\, dt\) is finite. Let \(G = (G_ {x}\), \(x\in E)\) be a~zero-mean Gaussian random field with covariance \(g\). Dynkin's isomorphism theorem relates particular moment functionals of the local times of the Markov process to certain moment functionals of the field \(G\), see, in particular, \textit{E. B. Dynkin}'s paper [J.\ Funct.\ Anal.\ 55, 344--376 (1984; Zbl 0533.60061)]. In their seminal paper \textit{M.\ B.\ Marcus} and \textit{J.\ Rosen} [Ann.\ Probab.\ 20, No. 4, 1603--1684 (1992; Zbl 0762.60068)], showed that such isomorphism theorems may be used to derive sample path properties of the local times of the Markov process from regularity results concerning Gaussian processes, and this approach has been recently paid a~considerable attention. Motivated by these achievements, the author proves a~version of the Dynkin isomorphism theorem for discrete time Markov chains.