Polyhedrality in pieces

From MaRDI portal



Abstract: The aim of this paper is to present two tools, Theorems 4 and 7, that make the task of finding equivalent polyhedral norms on certain Banach spaces easier and more transparent. The hypotheses of both tools are based on countable decompositions, either of the unit sphere S_X or of certain subsets of the dual ball of a given Banach space X. The sufficient conditions of Theorem 4 are shown to be necessary in the separable case. Using Theorem 7, we can unify two known results regarding the polyhedral renorming of certain C(K) spaces, and spaces having an (uncountable) unconditional basis. New examples of spaces having equivalent polyhedral norms are given in the final section.


A finite-dimensional space is called polyhedral if its unit ball is a polytope. There are several notions of polyhedrality in infinite-dimensional spaces (cf. [\textit{V. P. Fonf} and \textit{L. Veselý}, Can. J. Math. 56, No. 3, 472--494 (2004; Zbl 1068.46007)]) and the authors consider the original notion given by \textit{V. Klee} [Acta Math. 102, 79--107 (1959; Zbl 0094.16802)]: A Banach space is said to be polyhedral when the unit balls of its finite-dimensional subspaces are polytopes. The results of the authors give conditions for replacing the norm on a given Banach space with an equivalent polyhedral norm. The main tool used is the following property: Let \(X\) be a Banach space. We say that a set \(F \subset X^*\) has property \((*)\) if, for every \(\omega^*\)-limit point \(g\) of \(F\) (i.e., any \(\omega^*\)-neighborhood of \(g\) contains infinitely many points of \(F\)), we have \(g(x) < 1\) whenever \(\sup\{f(x): f \in F\} = 1 \). \textit{A. Gleit} and \textit{R. McGuigan} [Proc. Am. Math. Soc. 33, 398--404 (1972; Zbl 0244.46019)] introduced property \((*)\) in the particular case of the set of extreme points of the dual unit ball \(\text{ext}(B_{X^*})\) and showed that \(X\) is polyhedral if \(\text{ext}(B_{X^*})\) has property \((*)\). Recall that a subset \(B \subset B_{X^*}\) is called a boundary if, for every \(x \in S_X\), there exists \(f_x \in B\) such that \(f_x(x) = 1\). Generalizing this notion, a set \(F \subset X^*\) is said to be a relative boundary if, whenever \(x \in X\) satisfies \(\sup \{ f(x): f\in F\} =1\), then there exists \(f_x \in F\) such that \(f_x(x) = 1\). Based on a countable decomposition of the unit sphere \(S_X\), the authors find the following sufficient condition for constructing an equivalent polyhedral norm: Let \(X\) be a Banach space and suppose that we have sets \(S_n \subset S_X\) and an increasing sequence \(H_n \subset B_{X^*}\) of relative boundaries, such that \(S_X = \bigcup_{n=1}^\infty S_n\) and the numbers \[ b_n = \inf\{\sup\{h(x): h \in H_n\}: x \in S_n\} \] are strictly positive and converge to 1. Then, for a suitable sequence \((a_n)_{n =1}^\infty\), the set \[ F = \bigcup_{n=1}^\infty a_n (H_n \setminus H_{n-1}) \] is a boundary of an equivalent norm \(||| \cdot |||\). Moreover, if each \(H_n\) has property \((*)\), then \(F\) has property \((*)\) and \(||| \cdot |||\) is polyhedral.











This page was built for publication: Polyhedrality in pieces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q477035)