On measures in locally convex spaces (Q4473002)

In this paper, \(E\), \(X\), \(Y\), \(Z\) are Hausdorff locally convex spaces, \( \ell_{1}[E] \) and \( \ell_{1} \{ E \} \) are the spaces of weakly summable and absolutely summable sequences in \(E\) with the topology generated by the semi-norms \( \sigma_{U}( \{ e_{n} \} )= \sup \{ \sum |\langle e_{n}, a\rangle| : a \in U^{\circ} \} \) and \( q_{U}( \{ e_{n} \} )= \sum_{n} p_{U}( e_{n}) \), as \(U\) varies over the closed, absolutely convex 0-neighbourhoods of \(E\). An operator \( S \in L(E, X) \) is called absolutely summing if \( S( \ell_{1}[E] ) \subset \ell_{1} \{ X \} \); the set of all absolutely summing operators in \( L(E, X) \) is denoted by \( L_{1}(E, X) \). For an abstract measurable space \( (\Omega, \Sigma )\) \((\Omega \in \Sigma) \), \( M( \Omega , E) \) denotes all countably additive, \(E\)-valued measures on \(\Sigma\); out of these, the set of those having finite variation, with respect to every continuous semi-norm on \(E\), is denoted by \( MB( \Omega , E) \); if \( E = \mathbb R \), then \( M( \Omega , E) \) is denoted by \( M( \Omega ) \). A locally convex space \(X\) is said to have the B-property if for any bounded set \( A \subset \ell_{1} \{ X \} \), there is a bounded set \( C \subset X \) such that \( \sum_{n} p_{C}( x_{n}) \leq 1\), for all \(\{ x_{n} \} \subset A \). Throughout the paper, it is assumed that \(X\) is sequentially complete and has the B-property. The main result of the paper is: For a Hilbert space \(H\), an operator \( S \in L(H, X) \) is absolutely summing if and only if there is a Hilbert space \(G\) and a Hilbert-Schmidt operator \( T \in L(H, G) \), and an operator \( R \in L(G, X) \) such that \( S=RT\). The proof depends on the B-property; by using the B-property of \(X\), one gets a linear subspace of \(X\), which can be made into a Banach space, continuously enclosed in \(X\), and so the problem gets reduced to the case when \(X\) is a Banach space. By using this result, a generalization of Sazonov's theorem is proved. Then, starting with a \( \mu \in MB( \Omega , X) \), and using the B-property of \(X\), one gets a Banach space \(F \subset X\), continuously enclosed in \(X\), such that the measure \(\mu \) is \(F\)-valued and is of finite variation with respect to the Banach space \(F\). Now, assuming \(Z\) to be sequentially complete and starting with a bilinear mapping \(Y \times X \to Z \), with some continuity properties, the author develops the theory of integrable functions \( f: \Omega \to Y \) with respect to this measure, with \( \int f d \mu \in Z \), and proves a dominated convergence type theorem. The major tool is the availability of the Banach space \(F\). Another important result proved is: For a Hilbert space \(H\) and a \( \mu \in M( \Omega , H) \) and \( S \in L_{1}(H, X)\), we get: (1) \( S \mu \in MB( \Omega , X) \); (2) there exists a \(\nu \in M(\Omega) \) such that \( S \mu\ll \nu \); (3) for any \( \lambda \in M(\Omega) \), such that \( S \mu\ll \lambda \), \( \frac{ d S \mu }{ d \lambda} \) exists. Some additional results about differentiable measures on \( M(X)\) and some results about surface measures are also proved. For the entire collection see [Zbl 1029.00032].