Complex interpolation and twisted twisted Hilbert spaces (Q2516650)

The existence of nontrivial twisted sums of Hilbert spaces is now well known; a prime example is the Kalton-Peck space \(Z_2\). The existence of a nontrivial twisted sum of \(Z_2\) with itself follows from earlier work of the first two authors [Houston J. Math. 30, No. 2, 523--536 (2004; Zbl 1063.46012)], but the proof is rather abstract. Here a concrete construction is given. Let \(U\) be an open subset of the complex plane (e.g., the strip \(\{ 0< \operatorname{re}z<1\}\)), fix \(v \in U\) (e.g., \(v=\frac{1}{2}\)), and let \(W\) be a complex Banach space (e.g., the interpolation space \([\ell_\infty, \ell_1]_v\)). In the spirit of \textit{R. Rochberg} [Pac. J. Math. 174, No. 1, 247--267 (1996; Zbl 0866.46047)], the authors work with analytic functions from \(U\) to \(W\), and their truncated sequences of Taylor coefficients \[ \left(\frac{f^{(n-1)}(v)}{(n-1)!},\frac{f^{(n-2)}(v)}{(n-2)!},\dots,f'(v), f(v)\right). \] Modulo the obvious equivalence relationship, the collection of all such \(n\)-tuples is a Banach space \(\mathcal{H}(n)\). It is shown that, for any \(m,n\), the sequence \[ 0 \to \mathcal{H}(n) \to \mathcal{H}(m+n) \to \mathcal{H}(m) \to 0 \] is exact. Moreover, if this sequence is nontrivial (respectively, has strictly singular quotient map, or has strictly cosingular embedding map) in the case \( m =n = 1\), then the same statement is true for all \(m\) and \(n\). For the choices of \(U,v,W\) considered above, it is known that \(\mathcal{H}(1) \cong\ell_2\) and \(\mathcal{H}(2)\cong Z_2\); thus \(\mathcal{H}(4)\) furnishes a nontrivial twisted sum of \(Z_2\) with itself. The authors also show that \(\mathcal{H}(n)\) is not a twisted sum of two Hilbert spaces if \(n\geq3\), and that \(\mathcal{H}(n)\) does not contain complemented copies of \(\ell_2\) if \(n\geq2\). On the other hand, \(\ell_2\oplus\mathcal{H}(3)\) is shown to be another example of a nontrivial twisted sum of \(Z_2\) with itself.