Tensor products and independent sums of β„’_{𝓅}-spaces, 1&𝓁𝓉;𝓅&𝓁𝓉;∞

From MaRDI portal
Publication:4243755



Abstract: Two methods of constructing infinitely many isomorphically distinct CalLp-spaces have been published. In this article we show that these constructions yield very different spaces and in the process develop methods for dealing with these spaces from the isomorphic viewpoint. We use these methods to give a complete isomorphic classification of the spaces Rpalpha constructed by Bourgain, Rosenthal, and Schechtman and to show that XpotimesXp is not isomorphic to a complemented subspace of any Rpalpha.


Let \(1<p<\infty\), \(p\neq 2\). An \({\mathcal L}_{p}\)-space is a Banach space isomorphic to some complemented subspace of a function space \(L_{p}(\mu)\), yet not isomorphic to a Hilbert space. (This is actually not the original definition due to Lindenstrauss and PeΕ‚czyΕ„ski, but an equivalent reformulation.) The problem of studying the different types of separable \({\mathcal L}_{p}\)-spaces has been the subject of many papers, and this memoir makes a significant contribution to those questions. When the theory of \({\mathcal L}_{p}\)-spaces was developed, up to isomorphism only four infinite-dimensional separable \({\mathcal L}_{p}\)-spaces were known: \(L_{p}[0,1]\), \(\ell_{p}\), \(\ell_{p}\oplus \ell_{2}\), and \(\ell_{p}(\ell_{2})\). Then, in a seminal work from 1970 Rosenthal added some more spaces to this list, notably a space \(X_{p}\) of \(\{0, 1, -1\}\)-valued random variables on \([0,1]\); here \(p>2\). The question whether there exist in fact infinitely many isomorphically distinct \({\mathcal L}_{p}\)-spaces was settled by \textit{G. Schechtman} who considered tensor products \(X_{p}^{n}= X_{p} \otimes \cdots \otimes X_{p}\) where the norm on the tensor product is the one canonically inherited from \(L_{p}([0,1]^{n})\). He showed that these spaces are pairwise non-isomorphic \({\mathcal L}_{p}\)-spaces [Isr. J. Math. 22, 138-147 (1975; Zbl 0316.46018)]. Another approach was launched by \textit{J. Bourgain, H. P. Rosenthal} and \textit{G. Schechtman} [Ann. Math., II. Ser. 114, 193-228 (1981; Zbl 0496.46010)] who exhibited even uncountably many different \({\mathcal L}_{p}\)-spaces. They defined an \({\mathcal L}_{p}\)-space \(R_{p}^{\alpha}\) for each ordinal \(\alpha<\omega_{1}\) inductively; the inductive step involves embedding a sequence of subspaces of \(L_{p}[0,1]\) onto independent coordinates in \(L_{p}([0,1]^{\mathbb N})\). They went on to show that uncountably many of the spaces \(R_{p}^{\alpha}\) are pairwise non-isomorphic although it remained open precisely which ones are isomorphic and which ones aren't. In the present memoir the relationship between these two methods of constructing infinitely many \({\mathcal L}_{p}\)-spaces is investigated, and it is shown that they are essentially incompatible. More precisely, the author proves that \(X_{p}\otimes X_{p}\) is not isomorphic to a complemented subspace of any \(R_{p}^{\alpha}\). A great deal of technical work is required to reach this point; for instance, a method for dealing with gliding hump arguments in spaces with bases which have many subsequences spanning an isomorph of the space is developed. Another major result of this paper is the complete isomorphic classification of the spaces \(R_{p}^{\alpha}\): \(R_{p}^{\alpha}\) is isomorphic to \(R_{p}^{\beta}\), \(\beta > \alpha\), if and only if \(\beta = \alpha + n\) for some \(n\in {\mathbb N}\).











This page was built for publication: Tensor products and independent sums of β„’_{𝓅}-spaces, 1&𝓁𝓉;𝓅&𝓁𝓉;∞

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4243755)