A connection between Gaussian processes and Markov processes (Q1767553)

A centered Gaussian process \((\eta_x,x\in R)\) with covariance \(G\) is considered. Connection of the function \(G\) with the Green function (the kernel of a potential operator) \(g_{T_a}\) of some recurrent Markov process killed at the first hitting time of \(a\), where \(\eta_a=0\), is investigated. The property: \(G=g_{T_a}\) if and only if for any \(c\in R\) the process \((\eta_x-c)^2\) is infinitely divisible, is proved. Another criterion of the infinite divisibility is obtained in terms of independent Gaussian random values. The results are generalized for the case of Gaussian random fields on a separable locally compact metric space.