Measurable extension theorems (Q761558)

The following generalization of Tietze's classical extension theorem was established by \textit{J. Dugundji} [Pac. J. Math. 1, 353-367 (1951; Zbl 0043.381)]: Let A be a closed subset of a metric space X, let L be a locally-convex linear space, and let f be a continuous function from A into L. Then, there exists \(F:X\to L\) such that F is continuous, \(F|_ A=f\), and F[X] is contained in the convex hull of \(f[A].\) In the present work, the authors establish a theorem of Dugundji type concerning random functions. Theorem 4. Let \((\Omega,{\mathcal A},\mu)\) be a \(\sigma\)-finite measure space, let X be a locally-compact, separable metric space, let A be a closed subset of X, and let \(f:\Omega \times X\to L\) be a random function such that, for each \(\omega\in \Omega\), \(f(\omega,\cdot)\) is continuous. Then, there exists a random function \(\Phi:\Omega \times X\to L\) that satisfies the following conditions: (a) \(\Phi(\omega,\cdot)\) is continuous, for each \(\omega\in \Omega\), (b) \(\Phi(\omega,\cdot)|_ A=f(\omega,\cdot)\) a.e., and (c) \(\Phi(\omega,X) \subset \overline{co}(f(\omega,A))\) a.e. The proof, using ideas from the theory of measurable selections, proceeds along the following lines. Let \(C(X,L),\) the set of all continuous functions mapping X into L, be endowed with the compact-open topology, and let the relation \(F:\Omega \to C(X,L)\) be defined by \(F(\omega) = \{g\in C(X,L):g|_ A=f(\omega,\cdot),\) and \(g[X]\subset \overline{co}(f(\omega,A))\}\). The authors observe that F is a multifunction (that is, domain\((F)=\Omega)\) by virtue of Dugundji's theorem, prove that the graph of F is measurable, and finally invoke a theorem due to \textit{C. J. Himmelberg} [Fundam. Math. 87, 53-72 (1975; Zbl 0296.28003)] to deduce the existence of a measurable function \(\phi:\Omega\to C(X,L)\) such that \(\phi(\omega)\in F(\omega)\) a.e. It follows that \(\Phi:\Omega \times X\to L\) given by \(\Phi(\omega,x) = \phi(\omega)(x)\), is a random function that satisfies (a), (b) and (c). In the balance of the article the authors give a short proof of a probabilistic Hahn-Banach theorem due to \textit{O. Hanš} [Ann. Math. Stat. 30, 1152-1157 (1959; Zbl 0094.120)], and generalize Theorem 4 in various ways. In particular, they observe that a result due to \textit{M.-F. Sainte-Beuve} [J. Funct. Anal. 17, 112-129 (1974; Zbl 0286.28005)] makes it possible to replace the requirement that (\(\Omega\),\({\mathcal A},\mu)\) be a \(\sigma\)-finite measure space by the condition that (\(\Omega\),\({\mathcal A})\) be a complete measurable space, in their theorems as well as in the Theorems of Hanš and Himmelbeg cited above.