Quantitative Grothendieck property (Q2019184): Difference between revisions

As the author recalls in the introduction, in recent years quantifications of several Banach space properties (often related to weak compactness in the theorems of Gantmacher, Eberlein-Šmulyan, James) have been undertaken. The present paper continues this by dealing with a quantification of the notion of Grothendieck spaces. The latter are Banach spaces in whose dual each weak\(^*\)-con\-ver\-gent sequence converges weakly or, equivalently, each weak\(^*\)-Cauchy sequence is weakly Cauchy. (The argument for this equivalence is very elementary but we give it for the convenience of the reader: if \((x_n^*)\) is w\(^*\)-null, then so is \((y_n^*)\), where \(y_{2n+1}^*=2x_n^*\) and \(y_{2n}^*=x_n^*\), hence \((x_n^*)=(y_{2n+1}^*-y_{2n}^*)\) is weakly null by weak Cauchyness of \((y_n^*)\).) The two expressions \[ \delta_w(x_n^*)=\sup_{\|x^{**}\|\leq1}\inf_{n\in\mathbb{N}}\sup_{k,l\geq n}|x^{**}(x_k^*)-x^{**}(x_l^*)| \] and \[ \delta_{w^*}(x_n^*)=\sup_{\|x\|\leq1}\inf_{n\in\mathbb{N}}\sup_{k,l\geq n}|x_k^*(x)-x_l^*(x)| \] measure the diameter of the set of weak, respectively weak\(^*\) cluster points of a bounded sequence \((x_n^*)\) in the dual of a Banach space \(X\), they vanish iff the sequence is weakly Cauchy, respectively weak\(^*\)-Cauchy, and they give rise to the natural quantification of being a Grothendick space: given \(c\geq1\), a Banach space \(X\) is called \(c\)-Grothendieck if \[ \delta_w\leq c\delta_{w^*} \] on the bounded sequences of \(X^*\). Among the results we find the following: {\parindent=6mm \begin{itemize} \item[(1)] \(\ell_{\infty}\) and \(L^{\infty}(\mu)\) (\(\mu\) \(\sigma\)-finite) are \(1\)-Grothendieck; \item [(2)] any nonreflexive Banach space (Grothendieck or not) admits, given \(c\geq1\), an equivalent norm for which the space is not \(c\)-Grothendieck; \item [(3)] in general, the converse of the fact ``a \(c\)-Grothendieck space (for some \(c\geq1\)) is a Grothendieck space'' does not hold. \end{itemize}} The proofs use (amongst others) a lemma of \textit{O. F. K. Kalenda} on (I)-envelopes [Isr. J. Math. 162, 157--181 (2007; Zbl 1149.46018)] which in turn is strongly related to an inequality of \textit{S. Simons} [Pac. J. Math. 40, 703--708 (1972; Zbl 0237.46012)]. The lemma allows to show an interesting description of \(c\)-Grothendieckness: \(X\) is \(c\)-Grothendieck if and only if the (I)-envelope [see the present paper for the definition] of the unit ball of \(X\) contains the ball of radius \(1/c\) of the bidual \(X^{**}\). The author mentions the open question whether known Grothendieck spaces like the Hardy space \(H^{\infty}\) are \(c\)-Grothendieck for some \(c\).