Symmetric spaces with maximal projection constants (Q1874448): Difference between revisions

Let \(V\) be a linear subspace of a Banach space \(X\). Let \(P(X,V)\) be the collection of projections of \(X\) onto \(V\). This paper is about the attainment of the following projection constants: \[ \begin{aligned} \lambda(X,V)&= \inf\{Q : Q\in P(X,V)\},\\ \lambda(V) &= \sup\{\lambda(X,V): V\subseteq X\}, \end{aligned} \] and \[ \lambda_k = \sup\{\lambda(V): V\text{ has dimension } k\}. \] \noindent It is known that there is a \(k\)-dimensional space \(X^k\) such that \[ \lambda(X^k) = \lambda_k, \] and that \[ \lambda_k \leq \sqrt k \text{ and } \lim_{k\to \infty} {{\lambda_k}\over {\sqrt k}}=1. \] The maximal space \(X^k\) above is not symmetric. In fact, if all the spaces involved are symmetric and \(\lambda^S_k\) represents the corresponding constant, then it is known that \[ \lim \sup_k {{\lambda^S_k}\over {\sqrt k}}<1-{1 \over{900}}. \] This work settles two of the main questions on symmetric spaces by showing that there is a symmetric \(k\)-dimensional subspace \(Y^k \subseteq L_1\) such that \[ \lambda^S_k= \lambda(Y^k,L_1). \]