On the infinite divisibility of squared Gaussian processes (Q1400827)

The aim of the present paper is to investigate the infinite divsibility of squared Gaussian processes and to establish connections between Gaussian processes and Markov processes via their symmetric (and non-symmetric) Green function. Section 2 of the paper proves the following result: A squared fractional Brownian motion with an index \(\beta\) is infinitely divisible if and only if \(\beta\in (0,1]\). Section 3 strengthens this settlement by showing that actually if the Green function of a transient Markov process is symmetric, then this function represents the covariance of a Gaussian process with an infinitely divisible square. Section 4 gives evidence that, similarly to Brownian motion, a Gaussian process with an infinitely divisible square can generate a two-parameter family of processes having the additivity property of the squared Bessel processes. The obtained processes are Markovian if and only if the corresponding Gaussian process is itself also Markovian. Section 5 tries to relate a Gaussian process to a Markov process with a non-symmetric Green function. As an instance of such a connection, the author shows that Barlow's necessary and sufficient condition for the continuity of the local time process of a Lévy process can be translated as a condition for the continuity of a Gaussian process even when the Lévy process is not symmetric.