Finite order spreading models (Q1935196)

Let \((x_k)\) be a sequence in a Banach space \(X\). A Banach sequence space \(S(x_k)\) in which the unit vector sequence \((e_k)\) is fundamental is said to be a spreading model generated by \((x_k)\) if, for a given \(\varepsilon>0\) and \(n=1,2,\dots\), there is an \(i\) such that \[ (1+\varepsilon)^{-1}\left\|\sum\nolimits_{k=1}^na_ke_k\right\|\leq \left\|\sum\nolimits_{k=1}^na_kx_{i_k}\right\|\leq (1+\varepsilon)^{-1}\left\|\sum\nolimits_{k=1}^na_ke_k\right\| \] for all scalars \(a_k\) and \(i<i_1<\cdots<i_n\). One could define a 2-iterated spreading model as a spreading model of \(S(x_k)\) generated by some other sequence. Further iteration yields the \(k\)-iterated spreading models of \(X\). \textit{B.~Beauzamy} and \textit{B.~Maurey} [Ark. Mat. 17, 193--198 (1979; Zbl 0477.46018)], answering a problem of H. P.~Rosenthal, showed that the class of iterated spreading models does not coincide with the corresponding one of spreading models. \textit{E.~Odell} and \textit{Th.~Schlumprecht} [Oper. Theory, Adv. Appl. 77, 178--198 (1995; Zbl 0827.46014)] asked whether or not every Banach space admits a \(k\)-iterated spreading model isomorphic to some \(\ell_p\) for some \(1\leq p<\infty\), or \(c_0\). The authors give a negative answer to this question. To this end, they extend the notion of the spreading model to the \(k\)-spreading model. The definition is based on \(k\)-sequences and plegma families. Most of the results of the classical theory are stated and proved in the higher order setting. Moreover, new phenomena like the universality of the class of the 2-spreading models of \(c_0\) and the composition property are established.