Remarks on a characterization of nuclearity (Q792577)

It is well-known that a locally convex space E is nuclear iff \(\ell_ p\{E\}=\ell_ p(E)\) holds, algebraically and topologically, for some (all) \(1\leq p<\infty\). Here \(\ell_ p\{E\}\) resp. \(\ell_ p(E)\) denotes the space of all E-valued \(\ell_ p\)-sequences resp. weak \(\ell_ p\)- sequences, topologized in the usual fashion. This paper presents classes of non-nuclear locally convex spaces E such that \(\ell_ p\{E\}\) and \(\ell_ p(E)\) yet coincide as linear spaces. In fact, let E be the locally convex space obtained from supplying a given Banach space X with the locally convex topology generated by all seminorms \(x\mapsto \| Tx\|\), T ranging over all bounded operators with domain X and range in some Hilbert space. Then X verifies Grothendieck's theorem [cf. \textit{G. Pisier}: Ann. Inst. Fourier 28, No.1, 69-90 (1978; Zbl 0363.46019)] iff \(\ell_ p\{E\}=\ell_ p(E)\) algebraically for all \(1\leq p<\infty\) (equivalently, for some \(1\leq p<2)\). Similarly, X is a Hilbert-Schmidt space [cf. the author in Rend. Circ. Mat. Palermo, II. Ser. Suppl. 2, 153-160 (1982; Zbl 0503.46014)] iff \(\ell_ p\{E\}=\ell_ p(E)\) algebraically, for all (some) \(2\leq p<\infty\). But for any Banach space X, with E as above, \(\ell_ p\{E\}=\ell_ p(E)\) as locally convex spaces iff dim X\(<\infty\), \(\forall 1\leq p<\infty\).