Extreme points in polyhedral Banach spaces (Q2408019)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Extreme points in polyhedral Banach spaces
scientific article

    Statements

    Extreme points in polyhedral Banach spaces (English)
    0 references
    9 October 2017
    0 references
    Polyhedral Banach spaces were introduced by \textit{V. Klee} at the end of the paper [Acta Math. 103, 243--267 (1960; Zbl 0148.16203)] in 1960 as those real spaces where the unit balls of all subspaces are polygons. \(c_0\) serves as the basic example of such a space, and Klee proved in the last theorem of that paper the non-trivial fact that \(c_0\) is polyhedral and asked whether polyhedral spaces can be reflexive. After Klee, many other brilliant minds have hunted for the secrets of polyhedral spaces. In [Isr. J. Math. 4, 235--242 (1966; Zbl 0148.11301)] \textit{J. Lindenstrauss} showed that polyhedral spaces are never dual, in particular, they are never reflexive. Lindenstrauss knew that in reflexive spaces we have \(B_X=\overline{\text{conv}} \,{\text{ext}}\, B_X\), and asked whether this property still could be shared by polyhedral spaces. Lindenstrauss himself constructed a polyhedral space with infinitely many extreme points in the unit ball (\(c_0\) has none). The question of Lindenstrauss has remained open for 50 years, so the result of the paper under review deserves attention: Yes, there exists a polyhedral renorming of \(c_0\) such that \(B_X=\overline{\text{conv}} \,{\text{ext}}\, B_X\)! After \textit{V. P. Fonf} and \textit{L. VeselĂ˝}'s paper [Can. J. Math. 56, No. 3, 472--494 (2004; Zbl 1068.46007)], one operates with eight successively smaller classes of polyhedral spaces. The four inner classes cannot have extreme points in their unit ball at all. In Theorem 3.7 in the paper under review, it is shown that the constructed space is of type 5, so the constructed space is optimal relative to the Fonf-VeselĂ˝ classification. Let us end by describing how the new norm on \(c_0\) is obtained. Start with \((\varepsilon_n)\in \frac{1}{2}B_{c_0}\) and define the set \(A_n\) by \[ A_n=\{(1-2\varepsilon_1)\theta_1 e_1+(1-2\varepsilon_2)\theta_2 e_2+\ldots+(1-2\varepsilon_n)\theta_n e_n+(1+\varepsilon_n)\theta_{n+1} e_{n+1}\}, \] where \(\theta_1,\theta_2,\dots,\theta_n\) ranges over all possible combinations of signs. Let \(A=\bigcup_n A_n\) and \(B=\overline{\text{conv}}\,A\). Take \(X\) to be \(c_0\) with \(B\) as the unit ball.
    0 references
    0 references
    0 references
    polyhedral space
    0 references
    renorming
    0 references
    extreme points in the unit ball
    0 references
    0 references