The cluster function of single-valued functions (Q2432519)

From MaRDI portal
scientific article
Language Label Description Also known as
English
The cluster function of single-valued functions
scientific article

    Statements

    The cluster function of single-valued functions (English)
    0 references
    0 references
    25 October 2006
    0 references
    Let \(X\) and \(Y\) be topological spaces and let \(F : X \to 2^{Y}\) be a multifunction. Then the cluster set and the reduced cluster set of \(F\) at a point \(x \in X\) are \[ C(F; x) = \bigcap_{U \in {\mathcal U}(x)} \text{cl}_{Y}(F(U)) \quad \text{and}\quad C^{r}(F; x) = \bigcap_{U \in {\mathcal U}(x)} \text{cl}_{Y}(F(U \setminus \{x\})), \] where \(\mathcal U(x)\) is the set of open neighborhoods of \(x\). For appropriate metric spaces \(X\) and \(Y\), the author characterizes the cluster functions \(C(f; \cdot)\) of single-valued \(f : X \to Y\) functions among arbitrary set-valued functions and establishes the existence of a Baire class 2 function \(h : X \to Y\) satisfying \(C(f; \cdot) = C(h; \cdot)\), as follows. If \(X\) a complete metric space, \(Y\) is a Polish space, and \(F : X \to 2^{Y} \setminus \{\emptyset\}\) is a multifunction then the following are equivalent: 1. \(\exists f: X \to Y~(F = C(f; \cdot))\). 2. \(\forall x \in X~(C^{r}(F;x)\subseteq F(x) \text{ and } \text{card}(F(x)\setminus C^{r}(F;x) \leq 1)\). Moreover, if \(F\) is lower semicontinuous then the equivalence above holds for an arbitrary metric space \(X\) and a separable metric space \(Y\) and the statements are also equivalent to 3. \(F\) has closed graph and for every isolated point \(x \in X\), \(F(x)\) is a singleton. Concerning Baire class 2 selection functions, the author obtains that for complete metric \(X\) and Polish \(\sigma\)-compact \(Y\), for every \(f : X \to Y\) there is a Baire class 2 function \(h : X \to Y\) satisfying \(C(f; \cdot) = C(h; \cdot)\). An example of \textit{G. Beer} [J.\ Math.\ Anal.\ Appl.\ 95, 416--427 (1983; Zbl 0488.54010)] shows that the \(\sigma\)-compactness assumption on \(Y\) is necessary. On the other hand, the author shows that if \(C(f; \cdot)\) is lower or upper semicontinuous then this assumption can be dropped.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    cluster set
    0 references
    reduced cluster set
    0 references
    cluster function
    0 references
    set-valued function
    0 references
    dense selection
    0 references
    Baire-2 selection
    0 references
    0 references