On the classification of positions and complex structures in Banach spaces (Q517167)

This important article investigates the actual complexity of some natural operations which can be performed on separable Banach spaces. Typically, in how many different ways can a Banach space \(Y\) embed into a Banach space \(X\)? How many different complex structures can be defined on a given real Banach space? For answering such questions, one, of course, need explain the meaning of the word ``different''. For instance, two embeddings \(T\) and \(U\) are different if there is no isomorphism \(A\) of \(X\) such that \(T=AU\). Moreover, lower bounds on complexity go beyond the mere counting of the corresponding equivalence classes. It relies on the much more precise Borel reducibility of equivalence relations, and thus, in this case, on the usual frame (defined by \textit{B. Bossard} [Fundam. Math. 172, No.~2, 117--152 (2002; Zbl 1029.46009)]) which permits to speak about the descriptive complexity of classes of Banach spaces. Hence, ``how many'' must be understood in this sense. The main results of this article are: unless \(X\) is uniformly finitely extensible (a quite restrictive condition), there is a space \(Y\) such that \(E_0\) is Borel-reducible to the embeddings from \(Y\) into \(X\). If \(1<p<\infty\) and \(p\not=2\), the embeddings from \(l_p\) into itself are not Borel reducible to the orbit relation induced by the action of a Polish group. If \(1\leq p <2\), the space \(L_p\) contains a real subspace \(X\) such that \(E_1\) is Borel reducible to linear isomorphisms between complex structures on \(X\). In particular, the space \(X\) has many non-equivalent complex structures, in a strong meaning of the word ``many''.