Lipschitz retractions and complementation properties of Banach spaces (Q2127599)

The paper is dedicated to the study of the Lipschitz retractional structure of metric spaces and its connection to the projectional structure of the corresponding Lipschitz-free Banach spaces. As motivations for such an investigation, one could mention the analogous study of the projectional structure of Banach spaces and the retractional structure of compact topological spaces, or the open problem whether every Lipschitz-free Banach space is Plichko. For this study, the authors introduce, among other things, the following concepts. For a metric space \(M\) and a closed subset \(A\) of \(M\), \begin{itemize} \item[(A)] \(A\) is a Lipschitz retract of \(M\) if there exists a Lipschitz function \(R\colon M \to A\) such that \(R(a)=a\) for every \(a\in A\). \item[(B)] \(A\) is an \(\mathcal{F}\)-Lipschitz retract of \(M\) if there exists a Lipschitz function \(F\colon M \to \mathcal{F}(A)\) such that \(F(a)=\delta(a)\) for every \(a\in A\). \item[(C)] \(A\) is locally complemented in \(M\) if there exists a bounded linear extension operator from \(\mathrm{Lip}_0(A)\) to \(\mathrm{Lip}_0(M)\). \end{itemize} Then it is not hard to see that (A) implies (B) which implies (C). Moreover, in case \(M\) is a Banach space and \(A\) a closed subspace, (B) is equivalent to \(\mathcal{F}(A)\) being complemented in \(\mathcal{F}(M)\) and (C) reduces to the classical notion of local complementation for Banach spaces. The main results of the paper are the following. First a sufficient condition if given for \(\mathcal{F}(M)\) to be Plichko, in terms of the separability of certain balls in \(M\) (Theorem~2.4). If \(\mathcal{F}(M)\) has a projectional skeleton, then every separable subset of \(M\) is contained in a separable subset that is an \(\mathcal{F}\)-Lipschitz retract of \(M\) (the so-called separable \(\mathcal{F}\)-Lipschitz retraction property, Corollary~2.3). In the converse direction, if \(M\) has a (commutative) \(r\)-Lipschitz retractional skeleton, then \(\mathcal{F}(M)\) has a (commutative) \(r\)-projectional skeleton (Proposition~2.8). Quite interestingly, the converse of the last result is not true. Indeed, the main outcome of Section~3, and probably the most interesting result in the paper, is a metric space \(M\) such that \(\mathcal{F}(M)\) has a commutative 1-projectional skeleton, i.e., it is 1-Plichko, and yet \(M\) does not have a Lipschitz retractional skeleton. More precisely, there are two points in \(M\) such that no separable subspace of \(M\) containing the two points is a Lipschitz retract of \(M\). The main result of Section~4 is a similar result for higher cardinalities: for every cardinal \(\Lambda\), a metric space \(M\) and a subset \(A\) of \(M\) are constructed such that \(A\) has density \(\Lambda\), but it is not contained in any subset of \(M\) of density \(\Lambda\) which is a Lipschitz retract of \(M\). Finally, in Section~5 it is shown that such examples cannot exist if one considers local complementation instead of Lipschitz retracts: for every metric space \(M\) and every subset \(N\) of \(M\), there exists a subspace \(S\) of \(M\) with \(N\subseteq S\), \(\mathrm{dens}(N)=\mathrm{dens}(S)\) and such that \(S\) is locally complemented in \(M\) (Theorem~5.3).