A hereditarily indecomposable Banach space with rich spreading model structure (Q476506)

The authors present a reflexive hereditarily indecomposable (HI) Banach space with rich spreading model structure and rich -- regarding the HI setting -- family of isomorphisms. More precisely, they prove the following result. {Theorem.} There exists a reflexive Banach space \(\mathfrak{X}_{usm}\) with a basis satisfying the following: {\parindent=6mm \begin{itemize} \item[(1)] the space \(\mathfrak{X}_{usm}\) is hereditarily indecomposable, i.e., no two infinite-dimensional closed subspaces of \(\mathfrak{X}_{usm}\) form a direct sum, \item [(2)] the space \(\mathfrak{X}_{usm}\) is hereditarily unconditional spreading model universal, i.e., any infinite-dimensional subspace \(Y\subset \mathfrak{X}_{usm}\) contains a seminormalized weakly null sequence \((x_n)\) such that any spreading suppression unconditional basic sequence is equivalent to a spreading model of some subsequence of \((x_n)\), \item [(3)] the space \(\mathfrak{X}_{usm}\) is sequentially minimal, i.e., any block subspace \(Y\subset \mathfrak{X}_{usm}\) admits a block sequence \((y_n)\) so that in any infinite-dimensional subspace \(Z\subset\mathfrak{X}_{usm}\) one can find a basic sequence equivalent to some subsequence of \((y_n)\), \item [(4)] any block subspace \(Y\subset\mathfrak{X}_{usm}\) contains a seminormalized block sequence \((z_n)\) for which there is an isomorphism \(T:\mathfrak{X}_{usm}\to \mathfrak{X}_{usm}\) with \(T(z_{2n-1})=z_{2n}\) for each \(n\in\mathbb{N}\). \end{itemize}} The construction of \(\mathfrak{X}_{usm}\) uses the ``saturation under constraints'' method introduced by \textit{E. Odell} and \textit{Th. Schlumprecht} [Basel: Birkhäuser. Oper. Theory, Adv. Appl. 77, 178--198 (1995; Zbl 0827.46014); Trans. Am. Math. Soc. 352, No. 4, 1859--1888 (2000; Zbl 0947.46011)]. This method is proved to be very efficient in building spaces with hereditary heterogeneous structure. The authors partly apply the approach presented in their previous paper [Proc. Lond. Math. Soc. (3) 108, No. 6, 1381--1416 (2014; Zbl 1308.46013)], where an HI space \(\mathfrak{X}_{ISP}\) is constructed such that any operator on an infinite-dimensional subspace of \(\mathfrak{X}_{ISP}\) has a non-trivial invariant subspace. The space \(\mathfrak{X}_{ISP}\) admits only \(\ell_1\) or \(c_0\) as spreading models. The ingredients added in the construction of \(\mathfrak{X}_{usm}\) allow to find in each subspace a weakly null sequence \((x_n)\) with the following property implying (2) of the above theorem: for any Schreier set \(F\subset\mathbb{N}\), the sequences \((x_n)_{n\in F}\) and \((u_n)_{n\in F}\) are equivalent with a universal constant, where \((u_n)\) denotes Pełczyński's universal unconditional basis (see [\textit{A. Pełczyński}, Stud. Math. 32, 247--268 (1969; Zbl 0185.37401)]). As any operator in an arbitrary real HI space is a strictly singular perturbation of a scalar operator, properties (3) and (4) are highly non-trivial to achieve. The first study of existence of sequentially minimal HI spaces was made by \textit{V. Ferenczi} and \textit{Th. Schlumprecht} in [Proc. Lond. Math. Soc. (3) 106, No. 1, 163--202 (2013; Zbl 1268.46008)], who proved that a variant of the Gowers-Maurey space \(\mathcal{GM}\) satisfies a weaker version of (4). The existence of a sequentially minimal HI subspace of \(\mathcal{GM}\) follows by Gowers-type dichotomies, contrary to the discussed paper, where the authors prove properties (1) and (3) directly for the whole space \(\mathfrak{X}_{usm}\). In addition, the authors use the richness of the family of spreading models in \(\mathfrak{X}_{usm}\) to show that any block subspace of \(\mathfrak{X}_{usm}\) admits a strictly singular operator which is not polynomially compact.