A fixed-point characterization of weakly compact sets in \(L_1(\mu)\) spaces (Q2195157)

Norm compactness of convex subsets of Banach spaces can be characterized using the existence of fixed points of certain classes of mappings. For example, a closed bounded convex subset \(C\) of a Banach space is norm compact if and only if every continuous (or every Lipschitz) mapping \(T:C\to C\) admits a fixed point [\textit{V. L. Klee jun.}, Trans. Am. Math. Soc. 78, 30--45 (1955; Zbl 0064.10505); Proc. Am. Math. Soc. 93, 633--639 (1985; Zbl 0566.47039)]. Whether similar fixed point characterizations of weakly compact convex sets in a Banach space are possible remains an open question and results to date have depended on analytic and geometric properties of specific Banach spaces. In the article under review, the authors characterize the weakly compact convex subsets of \(L^1(\mu)\), where \(\mu\) is a \(\sigma\)-finite measure, in terms of fixed points of mappings in certain sub-classes of the Lipschitz functions. Let \(C\) be a closed convex subset of a Banach space and let \(T:C\to C\). Define \(C_0=C\) and \(C_n =\overline{ {co}} \,(T(C_{n-1}))\) for \(n\in \mathbb{N}\). Then \((C_n)\) is a nonincreasing sequence of closed convex subsets of \(C\) with \(T(C_n)\subset C_n\). The mapping \(T\) is \textit{eventually affine} if there exists \(n\in\mathbb{N}\) such that \(T:C_n\to C_n\) is affine; and \(T\) is \textit{cascading nonexpansive} if, for every \(n\in\mathbb{N}\cup\{0\}\), the mapping \(T:C_n\to C_n\) is Lipschitzian with Lipschitz constant \(\lambda_n\geq 1\) and \(\lim_n \lambda_n=1\). The main result of the article includes: Let \(C\) be a closed bounded convex subset of \(L^1(\mu)\) where \((\Omega, \Sigma, \mu)\) is a \(\sigma\)-finite measure space, and fix \(L>2\). Then the following are equivalent: (1) \(C\) is weakly compact; (2) every uniformly \(L\)-Lipschitzian mapping \(T:C\to C\) which is eventually affine has a fixed point; (3) every cascading nonexpansive mapping \(T:C\to C\) which is eventually affine has a fixed point. Since there exists a nonaffine isometry mapping a weakly compact convex subset of \(L^1[0,1]\) into itself that fails to have a fixed point [\textit{D. E. Alspach}, Proc. Am. Math. Soc. 82, 423--424 (1981; Zbl 0468.47036)], the affine assumptions in (2) and (3) cannot be removed.