The structure of the quasi-invariant set of a linear measure (Q1175656)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | The structure of the quasi-invariant set of a linear measure |
scientific article |
Statements
The structure of the quasi-invariant set of a linear measure (English)
0 references
25 June 1992
0 references
Let \(\mu\) be a probability measure on a Hilbert space \(E\). Let \(A(\mu)\) denote the set of \(x\in E\) such that the translate \(\mu_ x\) of \(\mu\) by \(x\) is equivalent to \(\mu\). A theorem of Skorokhod [\textit{A. V. Skorokhod}, ``Integration in Hilbert space.'' Berlin etc.: Springer-Verlag (1974; Zbl 0307.28010)] states that \(A(\mu)\) is a Borel set and that (1) if \(\mu(F)=0\) for every finite-dimensional subspace \(F\) of \(E\) then \(\mu(A(\mu))=0\), (2) if \(\mu(A(\mu))=1\) then there is a sequence \((L_i)\) of finite-dimensional subspaces such that \(\mu\left(\bigcup^ \infty_{i=1}L_ i\right)=1\). A similar result has been proved by \textit{Y. Okazaki} [Mem. Fac. Sci., Kyushu Univ., Ser. A 34 79--88 (1980; Zbl 0437.43002)]. He showed that if \(\mu(A(\mu))>0\) then \(A(\mu)\) is a locally compact \(\sigma\)-compact group in the induced topology. This paper generalizes the Skorokhod result to a probability measure \(\mu\) on a locally convex Hausdorff space \(E\). In this case, under the assumption that \(\mu(A(\mu))>0\), the authors are able to show that there are a finite dimensional subspace \(L\) of \(E\), a thick subgroup \(G\) of \(L\), and a countable subgroup \((x_i)\) of \(E\) such that \(A(\mu)=\bigcup^\infty_{i=1}(G+x_i)\). They also prove that if \(E\) is a Suslin locally convex Hausdorff space then \(A(\mu)\) is a Borel set and in this case if \(\mu(A(\mu))>0\) \(A(\mu)=\bigcup^\infty_{i=1}(L+x_i)\), where \(L\) and \(x_i\) are as above. The proof of the first result proceeds by replacing the restriction of \(\mu\) to \(A(\mu)\) by an equivalent measure which is \(A(\mu)\)-invariant. Then the methods of Weil are used [see \textit{P. R. Halmos}, ``Measure Theory:'' London: Macmillan (1950; Zbl 0040.16802), p. 62] to obtain the result.
0 references
admissible translate
0 references
quasi-invariant set
0 references
linear measure
0 references
Souslin locally convex Hausdorff space
0 references