Radon, Baire, and Borel measures on compact spaces. I (Q910665)

In the paper under review, the author's stated intention is ``to make these connections and relations (between continuous, Baire, and Borel functions or between the corresponding types of measures) as transparent as possible'', something which, it is argued, the Riesz-Markov theorem, for all its depth, does not do. Given a compact space X, let \(\tilde B_ 0=C(X)\), the space of continuous real-valued functions on X. For each ordinal \(\beta <\omega_ 1\), the smallest uncountable ordinal, define \(\tilde B_{\beta}\) as the set of all (real-valued) functions on X that are pointwise limits of uniformly bounded sequences in \(\cup \{\tilde B_{\alpha}:\) \(\alpha <\beta \}\). The space \(\tilde B=\cup \{\tilde B_{\beta}:\) \(\beta <\omega_ 1\}\) is called the space of bounded Baire functions on X. Analogously, define the space B of bounded Borel functions on X as follows. Let \(L_ 0\) denote the convex conical hull of the set of characteristic functions \(\{\chi_ G:\) G open in \(X\}\) and let \(B_ 0\) be the uniform closure of the \(\{\) f-g: \(f,g\in L_ 0\}\). The author shows (Proposition 2.2) that \(B_ 0\) is the uniform closure of the linear span of the set \(\{\chi_ A\}\) where A belongs to the \(\sigma\)- field of subsets of X generated by the open sets. Then, for each ordinal \(\beta <\omega_ 1\), define \(B_{\beta}\) to be the set of all (real- valued) functions on X that are pointwise limits of uniformly bounded sequences in \(\cup \{B_{\alpha}:\) \(\alpha <\beta \}\). Finally, we have \(B=\cup \{B_{\beta}:\) \(\beta <\omega_ 1\}\). The basic distinction between \(\hat B\) and B, to quote from the author, ``lies exactly in the fact that in general, there exist open sets \(G\subset X\) whose characteristic function \(\chi_ G\) is not the (pointwise) upper envelope of a countable family of continuous functions.'' The main result of the article is that the spaces \(\hat B\) and B can both be identified with Banach sublattices of the second dual space \(C(X)''\) of C(X). In both cases, the identification is achieved via a linear extension of the mapping \(\phi^*\) which we now define. Let M be the set of all non-negative bounded (real-valued) functions on X and L the set of all functions in M that are lower semicontinuous. Let \(\phi\) denote the canonical imbedding of C(X) into \(C(X)''\). Then \(\phi^*: M\to \{f\in C(X)'':\) \(f\geq 0\}\) is given by \[ \phi^*(h)=\inf \{\sup \{\phi (f):\quad 0\leq f\leq g,\quad f\in C(X)\}:\quad 0\leq h\leq g,\quad g\in L\}. \] Numerous lemmas lead to the results that \(\phi^*\) is additive on the positive cones of \(\tilde B\) and B and therefore, by linear extension, defines isometric Riesz isomorphisms of \(\hat B\) and B onto suitable subspaces of \(C(X)''\).