On projectional skeletons and the Plichko property in Lipschitz-free Banach spaces (Q6052760)

The paper is motivated by the open problem whether every Lipschitz-free Banach space is a Plichko space. Recall that a Banach space \(X\) is \textit{\(\lambda\)-Plichko} if there are a linearly dense subset \(\Delta\) of \(X\) and a \(\lambda\)-norming subspace \(N\) of \(X^*\) such that for all \(\varphi\in N\) the set \[ \{x\in \Delta\colon \langle\varphi, x\rangle\neq 0\} \] is countable. In this case the authors say that the set \(\Delta\) is a \textit{witness} of the Plichko property of \(X\). A Banach space is \textit{Plichko} if it is \(\lambda\)-Plichko for some \(\lambda\geq1\). The main theme of the paper is to find conditions on a metric space \(M\) such that the Lipschitz-free space \(\mathcal{F}(M)\) is Plichko and admits a simple witness. The first main result (Theorem 2.1) characterises these Lipschitz-free spaces that admit a witness consisting of Dirac measures. The result gives several equivalent conditions, in terms of the metric space \(M\), in terms of the norming subspace \(N\) in \(\mathcal{F}(M)^*\), and in terms of projectional skeletons in \(\mathcal{F}(M)\). More precisely, the metric condition is that for every \(p\in M\) and \(r<1/\lambda\) the ball \(B(p, r\cdot d(p,0))\) should be separable. Concerning skeletons, the equivalence claims that \(\mathcal{F}(M)\) admits a commutative \(\lambda\)-projectional skeleton \(\{P_s\}_{s\in \Gamma}\) such that \(P_s(\delta_p)\in \{0,\delta_p\}\) for all \(p\in M\). In other words, the skeleton preserves the Dirac measures, which then implies the existence of a Lipschitz retractional skeleton in the metric space \(M\). The second main result is motivated by \textit{A.~Godard}'s theorem [Proc. Am. Math. Soc. 138, No.~12, 4311--4320 (2010; Zbl 1222.46010)] that \(\mathcal{F}(M)\) is isometric to an \(L_1(\mu)\) space when \(M\) is an \(\mathbb{R}\)-tree. Since every \(L_1(\mu)\) is \(1\)-Plichko, it is natural to ask for some explicit witnesses. In this context the authors prove that when \(M\) is an \(\mathbb{R}\)-tree, \(\mathcal{F}(M)\) is \(1\)-Plichko with a witness given by elementary molecules, namely measures of the form \(\frac{\delta_p- \delta_q}{d(p,q)}\) (Theorem 3.4). Notice that this result is optimal, since in general \(\mathbb{R}\)-trees do not satisfy the characterisation from the previous paragraph. Finally, recall that a natural candidate for a counterexample to the problem whether all Lipschitz-free spaces are Plichko is \(\mathcal{F}(\ell_\infty)\). In this direction, the authors obtain the partial result that \(\mathcal{F}(\ell_\infty)\) is not \(\lambda\)-Plichko for \(\lambda<2\).