Lipschitz structure of quasi-Banach spaces (Q839921)

A well-known open problem is whether two separable Lipschitz isomorphic Banach spaces are necessarily linearly isomorphic. For non-separable Banach spaces, examples were given by \textit{I.\,Aharoni} and \textit{J.\,Lindenstrauss} [``Uniform equivalence between Banach spaces'', Bull.\ Am.\ Math.\ Soc.\ 84, 281--283 (1978; Zbl 0401.57026)] and \textit{G.\,Godefroy} and \textit{N.\,Kalton} [``Lipschitz-free Banach spaces'', Stud.\ Math.\ 159, No.\,1, 121--141 (2003; Zbl 1059.46058)]. In the paper under review, the authors give counterexamples for separable quasi-Banach spaces. More precisely, they work with \(p\)-Banach spaces (\(0 < p \leq 1\)), i.e., it is assumed that the quasi-norm is actually a \(p\)-norm: \(\| x + y\|^p \leq \| x \|^p + \| y \|^p\). The counterexamples are \textit{Arens-Eells spaces}. Given a \(p\)-metric space \((M, d)\): \(d (x,y)^p \leq d (x, z)^p + d (z, y)^p\), with a distinguished point denoted by \(0\), let \({\mathbb R}^M\) be the space of all the functions \(f : M \to {\mathbb R}\) such that \(f (0) = 0\); the \textit{Arens-Eells \(p\)-space} \(\text{Æ}_p (M)\) is the Hausdorff completion of the linear span of all evaluation maps \(\delta_x : f \in {\mathbb R}^M \mapsto f (x) \in {\mathbb R}\), with \(x \in M\), endowed with the \(p\)-seminorm defined, for \(\mu = \sum_{j=1}^N a_j \delta_{x_j}\), by \[ \| \mu \|_{\text{Æ}_p (M)} = \sup \Big\| \sum_{j=1}^N a_j f (x_j) \Big\|_Y \,, \] where the supremum is taken over all \(p\)-normed spaces \(Y\) and all \(f : M \to Y\) such that \(f (0) = 0\) and \(\| f (x) - f (y)\|_Y \leq d (x, y)\). It is shown that for every infinite-dimensional separable Banach \(X\) (with the -- perhaps technical -- condition: \(X\) does not have the Schur property), \(\text{Æ}_p (X)\) is, for \(0 < p <1\), Lipschitz isomorphic but not linearly isomorphic to the space \(X \oplus_p \ker \beta_X\), where \(\beta_X : \text{Æ}_p (X) \to X\) is the (linear) quotient map induced by the identity map \(\text{id}_X : X \to X\). The main tool is a lifting property of Lipschitz maps (the \textit{\(p\)-Lipschitz lifting property}). The authors show that every infinite-dimensional separable Banach space without the Schur property fails the \(p\)-Lipschitz lifting property for \(0 < p <1\). To show that, they use the space \(\text{Æ}_1 (X) = {\mathcal F} (X)\) and the notion of type \(\tau\) of a Banach space \(X\): \(\tau (x) = \lim_{\mathcal U} \| x - a_n\|\), where \((a_n)\) is a bounded sequence in \(X\) and \({\mathcal U}\) is an ultrafilter on \({\mathbb N}\). The authors conjecture that the assumption that \(X\) does not have the Schur property can be removed.