A version of Chevet's theorem for stable processes (Q5966479)

Let X and Y be separable Banach spaces, and let \(Z=X{\check \otimes}Y\) denote the injective tensor product. Let \(S_ X\), \(S_ Y\) and \(S_ Z\) denote the unit spheres of these spaces. Let U and V be symmetric p- stable random variables, \(0<p\leq 2\), with values in X and Y, respectively. Let \(\eta_ U\) and \(\eta_ V\) denote the spectral measures of U and V on \(S_ X\) and \(S_ Y\). Does there exist a symmetric p-stable random variable with values in Z and with spectral measure \(\eta_ U\otimes \eta_ V\) on \(S_ Z?\) The authors answer this question in the affirmative. If \(p=2\), this is a reformulation of a theorem of \textit{S. Chevet} [C. R. Acad. Sci., Paris, Sér. A 284, 441-444 (1977; Zbl 0362.28007)], in which case U and V are Gaussian. If \(X=C(S)\), \(Y=C(T)\), \(Z=C(S\times T)\), where S and T are compact metrics, then Chevet's theorem can be restated as follows: if U and V are Gaussian random variables with values in C(S) and C(T) and with covariances \(\Gamma_ 1\) and \(\Gamma_ 2\), respectively, then there exists a Gaussian random variable with values in C(S\(\times T)\) and with covariance \(\Gamma_ 1\times \Gamma_ 2\).