Operator-decomposability of Gaussian measures on separable Banach spaces (Q1187527)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Operator-decomposability of Gaussian measures on separable Banach spaces |
scientific article |
Statements
Operator-decomposability of Gaussian measures on separable Banach spaces (English)
0 references
22 July 1992
0 references
The author continues his investigations of strongly operator decomposable probabilities on Banach spaces [Math. Nachr. 154, 315-326 (1991; Zbl 0754.60017)]. Recall that \(\mu\in M^ 1(E)\) is strongly \(T\)-decomposable if (1) \(\mu=T\mu*\nu\) for some \(\nu\in M^ 1(E)\) and if in addition (2) \(T^ n\mu\to\varepsilon_ 0\). Let \(C(T):=\{x\in E:T^ nx\to 0\}\) be the \(T\)-contractible subspace. If \(\mu\) is supported by \(C(T)\), then \((1)\Rightarrow (2)\), and on the other hand, if the operator \(T\) is power bounded and \(\mu\) is strongly \(T\)- decomposable, then \(\mu\) is supported by \(C(T)\). The paper under review is concerned with the question if the power boundedness is a superfluous condition. Indeed, as shown finally, there exist even symmetric strongly \(T\)-decomposable Gaussian measures \(\mu\) on Hilbert spaces with \(\mu(C(T))=0\). In \S3 the author proves a decomposition of a \(T\)- decomposable measure in a convolution product of \(T\)-invariant and strongly \(T\)-decomposable factors. \S4 considers the structure of \(T\)- decomposable Gaussian measures, which are used for the construction of the above-mentioned example in \S5. \S6 gives a detailed analysis of the structure of the decomposability semigroup of Gauß-measures. [The investigation of decomposability semigroups was initiated by \textit{K. Urbanik}, see e.g. Studia Math. 63, 283-308 (1978; Zbl 0404.60010)].
0 references
\(T\)-decomposable
0 references
\(T\)-contractible
0 references
operator decomposable probabilities on Banach spaces
0 references
convolution product
0 references
decomposability semigroup of Gaussian measures
0 references