On asymptotic models in Banach spaces (Q1885596)

Instead of using sequences and the basic Ramsey theorem to construct spreading models of Banach spaces (in the sense of A. Brunel and L. Sucheston), in this paper the method of using infinite arrays of vectors (of which the rows are basic sequences) and a refined version of Ramsey's theorem leads to the investigation of what the authors call asymptotic models of a Banach space (strong asymptotic models are also defined, under stricter assumptions). While losing some of the original force (every spreading model is an asymptotic model), the resulting theory gains in symmetry under certain conditions. For instance, if the array consists of weakly null sequences, or if each row is a block basis of some basis in the Banach space \(X\), then normalized block bases of the corresponding asymptotic model again are asymptotic models: this is not generally true for spreading models. Some questions that are still open for spreading models allow for definite answers in this new setting. For instance: if all asymptotic models of all block basis arrays of \(X\) are equivalent, then all spreading models of \(X\) are equivalent to the unit vector basis of \(c_0\) or \(\ell_p\) for some \(1\leq p<\infty\). As another example for the difference between the spreading and asymptotic theories, while Androulakis, Odell, Schlumprecht and Tomczak--Jaegermann have proved that there exists a reflexive Banach space for which no spreading model is reflexive, isomorphic to \(c_0\), or isomorphic to \(\ell_1\), we have the striking statement: every \(X\) admits an asymptotic model which is either reflexive, or isomorphic to \(c_0\) or \(\ell_1\). On the negative side, just as it is known that spreading models of spreading models do not need to be spreading models of the first space, the same negative statement is true for asymptotic models. In fact, the authors prove that there exist \(X\) and \(Y\) such that \(Y\) is a spreading model of \(X\), \(\ell_1\) is a spreading model of \(Y\), and yet \(\ell_1\) is not isomorphic to any asymptotic model of \(X\). In the words of the authors, ``asymptotic models provide a somewhat fuller theory than spreading models, although some of the same deficiencies remain.'' The article also contains a section on asymptotic models under renormings, as well as an ``odds and ends'' section in which further observations are made, and several open problems are listed.