Lévy-type processes and Besov spaces (Q1370915)

Let \(\{X_t\}_{t\geq 0}\) be any Lévy process with values in \(\mathbb{R}\) and Blumenthal-Getoor (upper) index \(1< \beta< 2\). It is shown that locally the sample paths \(t\mapsto X_t(\omega)\) are almost surely contained in the Besov spaces \(B_{p,\infty}^s (\mathbb{R})\) with parameters \(1\leq p<\beta\) and \(s\beta< 1\). The paper extends previous results on \(\alpha\)-stable Lévy processes by \textit{Z. Ciesielski, G. Kerkyacharian}, and \textit{B. Roynette} [Stud. Math. 107, No. 2, 171-204 (1993; Zbl 0809.60004)]. The proof is based on an (explicit) atomic decomposition of Besov spaces [due to \textit{M. Frazier} and \textit{B. Jawerth}, Indiana Univ. Math. J. 34, 777-799 (1985; Zbl 0551.46018)] using the Schauder functions. The idea is to approximate the sample paths by piecewise linear arcs, to control the interpolation weights, and to obtain thus equivalent norms for certain Besov spaces. The estimate of the weights is done by pathwise decomposition of the Lévy process and standard moment estimates for the truncated process. In the last section, this method is applied to comparable processes; comparable means that the transition probabilities of the process under consideration (not necessarily a Lévy process) can be pointwise bounded by multiples of the transition probabilities of some fixed Lévy process [cf. the reviewer, Stochastics Stochastics Rep. 57, No. 1/2, 89-110 (1996; Zbl 0888.60035)]. The embedding results are then the same as for the dominating Lévy process.