The ergodicity of Orlicz sequence spaces

From MaRDI portal





The classical Banach homogeneous space problem, asking whether a Banach space isomorphic to all its infinite-dimensional subspaces is Hilbertian (namely, isomorphic to a Hilbert space), was notoriously solved in the positive by \textit{R.~A. Komorowski} and \textit{N.~Tomczak-Jaegermann} [Isr. J. Math. 89, No.~1--3, 205--226 (1995; Zbl 0830.46008)], and \textit{W.~T. Gowers} [Ann. Math. (2) 156, No.~3, 797--833 (2002; Zbl 1030.46005)]. This result triggered the natural question, originally asked by Godefroy, if there are Banach spaces with only countably many infinite-dimensional subspaces, up to isomorphism. Ferenczi and Rosendal formulated the stronger conjecture that every non-Hilbertian separable Banach space is ergodic. Recall that a Banach space \(X\) is \textit{ergodic} if the equivalence relation \(\mathbb{E}_0\) Borel reduces to the isomorphism relation between subspaces of \(X\). Here, \(\mathbb{E}_0\) is the equivalence relation on \(2^\mathbb{N}\) of eventual equality between sequences. By a result of Harrington, Kechris and Louveau [\textit{L.~A. Harrington} et al., J. Am. Math. Soc. 3, No.~4, 903--928 (1990; Zbl 0778.28011)], \(\mathbb{E}_0\) does not reduce to the equivalence relation of equality of real numbers. As a consequence, if a Banach space is ergodic, not only has it continuum-many mutually non-isomorphic subspaces, but these subspaces cannot be classified by using the real numbers as invariants.\N\NWhile the conjecture is still open, there are several results that support its validity, in particular the conjecture is true for all spaces that are not near-Hilbert, [\textit{W.~Cuellar Carrera}, Trans. Am. Math. Soc. 370, No.~12, 8691--8707 (2018; Zbl 1457.46010)]. Among the spaces that are not covered by this result are, for example, some Orlicz spaces, or all non-Hilbertian twisted Hilbert spaces, in particular the Kalton-Peck space \(Z_2\). The contribution of this paper is to handle the case of Orlicz sequence spaces and to deduce the result for the Kalton-Peck space, among others. More precisely, the main result of the paper (Theorem~1.4) is that every Orlicz sequence space is either Hilbertian, or ergodic. As a consequence of this result, it is proved that, for every Lipschitz function \(\phi\colon \mathbb{R} \to \mathbb{R}\), the space \(\ell_2(\phi)\) is either Hilbertian or ergodic. In particular, taking as \(\phi\) the identity, the Kalton-Peck space \(Z_2\) is ergodic. Another consequence is that all twisted Hilbert spaces generated by complex interpolation of Orlicz sequence spaces are ergodic.











This page was built for publication: The ergodicity of Orlicz sequence spaces

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6879597)