Exit distributions for symmetric Markov processes via Gaussian techniques (Q1201757): Difference between revisions

Let \(X_ u\), \(u\geq 0\), be a symmetric homogeneous right Markov process with state space \(T\subset{\mathbf R}\) and everywhere finite Green's function \(g(t,s)\). Let \(\Phi_ t\), \(t\in T\), be a Gaussian process whose parameter space is the state space of \(X\), with mean zero and covariance function \(g(t,s)\) [see \textit{E. B. Dynkin}, Bull. Am. Math. Soc., New Ser. 3, 975-999 (1980; Zbl 0519.60046)]. If \(\tau_ s\) denotes the first exit time from \(S^ c\) for the process \(X\) and \(\mu^ S_ t\) the corresponding exit distribution starting at \(t\), Dynkin showed that \[ E(\Phi_ t|\Phi_ s, s\in S)=\int\mu^ S_ t(ds)\Phi_ s. \] In this paper the author uses the above formula to compute the first exit distributions. This is done for processes with continuous paths and symmetric Lévy processes. In the latter case it is shown that the associated Gaussian process is stationary and by using Wiener's prediction theory the Fourier transform of \(\mu^ S_ t\) is obtained.