Around finite-dimensionality in functional analysis: a personal perspective (Q380431)

The author considers various properties which characterize (1) finite-dimensional spaces among Banach spaces, (2) nuclear spaces among Fréchet spaces. He then gives results specifying the size, in any infinite-dimensional Banach space, of the set of the elements which do not satisfy one of these properties. In a last part, he defines Hilbert-Schmidt spaces as Banach spaces \(X\) for which every bounded linear map between Hilbert spaces which factorizes through \(X\) is Hilbert-Schmidt. This is equivalent to saying that every operator \(T : X \to \ell_2\) is \(2\)-summing, or also to saying that every bounded linear map from \(X\) to \(\ell_2\) extends to a bounded linear map on any superspace of \(X\). The author introduces the space \(\ell_2 \{X_0\}\) as the space of sequences \((x_n)_n\) in \(X\) such that \(\sum \| T (x_n) \|^2 < \infty\) for every \(T \in {\mathcal L} (X, \ell_2)\) and characterizes Hilbert spaces as Banach spaces \(X\) for which \(\ell_2 \{X_0\} = \ell_2 \{X\}\), the space of strongly square-summable series in \(X\), and Hilbert-Schmidt spaces as Banach spaces \(X\) for which \(\ell_2 \{X_0\} = \ell_2 [X]\), the space of weakly square-summable series in \(X\), in order to conclude that \(X\) is finite-dimensional if and only if it is both a Hilbert space and a Hilbert-Schmidt space. Other factorization properties are considered in the same way. Reviewer's remarks. It should be noted that there are a number of misprints (the first ones are at the bottom of the p.\,374, where \(\ell [X]\) should be read \(\ell_p [X]\) and \(\ell_p [X]\) should be read \(\ell_p \{X\}\)). Moreover, the discussion on p.\,386 is hard to read, due to an incorrect presentation.