Sample path properties of stochastic processes represented as multiple stable integrals (Q1174801)

Let \(T\) be an abstract set, and \(V\) be a vector space of real functions on \(T\). Let \(M(t)\) be a symmetric \(\alpha\)-stable \((S\alpha S)\) Lévy process on \(T\) with a control measure \(m(dt)\). Consider a multiple stochastic integral process \[ X(t)=\int^ \infty_{-\infty}\dots\int^ \infty_{-\infty} f_ t(x_ 1,\dots,x_ n)M(dx_ 1)\dots M(dx_ n),\quad t\in T,\quad(t\mapsto f_ t(x_ 1,\dots,x_ n))\in V. \] The paper studies the problem of the existence of a version of the process \(X(t)\) with sample paths in the space \(V\). Necessary conditions, and sufficient conditions as well, are presented for a wide class of vector spaces \(V\). Many classes of interest from the view point of applications [cf. \textit{M. Taqqu}, Random processes with long-range dependence and high variability, J. Geophys. Res. 92, 9683-9686 (1987)] satisfy the suitable conditions (including spaces of bounded, or continuous, or differentiable functions, and numerous alterations of the aforementioned spaces). Besides use of a large variety of techniques, pertinent to multiple integration, proofs are based on subtle measure theoretical arguments. A widely understood separability is essential although the nonseparable case (e.g., related to the space of bounded functions) is treated as well. Assuming the fulfillment of a certain separability condition, the authors prove a zero-one law for the process \(X(t)\), \(P(\{X(t), t\in T\}\in V)=0\) or 1, in the two-dimensional case (recently, the authors have proved the 0-1 law for an arbitrary dimension).