A note on metric projections (Q1899353)

Let \(E\) denote a normed vector space, and \(M\hookrightarrow E\) a vector subspace. If \(P_M: E\to M\) denotes the metric projection, then trivially \(|P_M|\leq 2\). Moreover, \(|P_M|\leq 1\) for every vector subspace \(M\hookrightarrow E\) if and only if \(E\) is a pre-Hilbert space. In the paper under review, the authors establish that for a Banach space the property of being uniformly non-squared is a necessary and sufficient condition for the existence of a constant \(\lambda< 2\) such that the estimate \(|P_M|\leq \lambda\) holds for every vector subspace \(M\hookrightarrow E\). Moreover, an example of a reflexive, strictly convex, smooth Banach space \(E\) consisting of \(\ell_p\) type sequences \((p> 1)\) is presented such that there exists a Chebychev subspace \(S\hookrightarrow E\), the metric projection \(P_S: E\to S\) of which is a linear operator, and verifies the norm condition \(|P_S|= 2\).