A space with no unconditional basis that satisfies the Johnson-Lindenstrauss lemma (Q2311976)

The Johnson-Lindenstrauss (JL) lemma in the title means that, for each finite set of points \(x_1, \dots, x_n \in X\), there is a linear subspace \(F\subset X\) of dimension \(O(\log n)\) and a linear map \(L: X\to F\) such that \[\|x_i-x_j\| \leq \|L(x_i) - L(x_j)\|\leq C\|x_i - x_j\|.\] This is a technical result proved by \textit{W. B. Johnson} and \textit{J. Lindenstrauss} [Contemp. Math. 26, 189--206 (1984; Zbl 0539.46017)] for Hilbert spaces and quite useful in applications, which fostered the interest in finding non-Hilbert spaces satisfying the JL lemma. So far, it seems that the only positive result was obtained by \textit{W. B. Johnson} and \textit{A. Naor} [Discrete Comput. Geom. 43, No. 3, 542--553 (2010; Zbl 1196.46013)], showing that Tsirelson's \(2\)-convexification space \(\mathscr T_2\) also satisfies the JL lemma. On the line of spaces ``close to'' Hilbert spaces, one encounters twisted Hilbert spaces: Banach spaces \(X\) that admit a Hilbert subspace \(H\) so that \(X/H\) is also a Hilbert space. Twisted Hilbert (non-Hilbert) spaces exist [\textit{P. Enflo} et al., Math. Scand. 36, 199--210 (1975; Zbl 0314.46015)] and can be generated in different ways. In this paper, the author shows that the twisted Hilbert space \(Z(\mathscr T_2)\), generated by complex interpolation with respect to the pair \((\mathscr T_2, \mathscr T_2^*)\), following \textit{N. J. Kalton} [Trans. Am. Math. Soc. 333, No. 2, 479--529 (1992; Zbl 0776.46033)], also satisfies the JL lemma, which is surprising, although not as much as the fact, proved by the author [``A weak Hilbert space that is a twisted Hilbert space'', J. Inst. Math. Jussieu (to appear), \url{https://doi.org/10.1017/S1474748018000221}], that \(Z(\mathscr T_2)\) is a weak Hilbert space. As any nontrivial twisted Hilbert space, \(Z(\mathscr T_2)\) cannot have an unconditional basis [\textit{N. J. Kalton}, J. Inst. Math. Jussieu 2, No. 3, 401--408 (2003; Zbl 1035.46007)], and this explains the title of the paper. All this ranks \(Z(\mathscr T_2)\) as a quite exotic exotic Banach space it is the first known \begin{itemize} \item twisted Hilbert space that is weak Hilbert, \item space without unconditional basis and satisfying the JL lemma. \end{itemize} The author also struggles in the paper with the idea that \(Z(\mathscr T_2)\) could also be the first non-Hilbert space that is locally self dual (LSD), in the language of \textit{V. Mascioni} [Houston J. Math. 19, No. 1, 27--38 (1993; Zbl 0785.46010)], namely, that all finite-dimensional subspaces are \(C\)-isomorphic to their duals for some prefixed \(C\). What the author actually shows is that \(Z(\mathscr T_2)\) is ``asymptotically'' LSD.