The weak metric approximation property (Q2574151): Difference between revisions

The authors define the \textit{weak \(\lambda\)-bounded approximation property} (\(\lambda\)-\(wBAP\)) for a Banach space \(X\) in the following way: for every Banach space \(Y\) and every weakly compact operator \(T\colon X \to Y\), there exists a net of finite rank operators \(S_\alpha \colon X \to X\) which converges uniformly on compact subsets of \(X\) to the identity operator \(I_X\) and such that \(\| T S_\alpha\| \leq \lambda\,\| T\| \) for all \(\alpha\). For \(\lambda = 1\), one says that \(X\) has the \textit{weak metric approximation property} (\(wMAP\)). The Reinov space, which is separable as well as all its duals, and which has \(AP\), does not have \(wBAP\); the Johnson-Schechtman space, which is a closed subspace of \(c_0\), has the \(8\)-\(BAP\) but not \(wMAP\) (see below). The authors show that \(X\) has the \(\lambda\)-\(wBAP\) if and only if the trace mapping \(V \colon X \hat \otimes_\pi Y^\ast \to {\mathcal F} (X,Y)^\ast\) satisfies \(\| u\| _\pi \leq \lambda \,\| Vu\| \) for every \(u \in X \otimes_\pi Y^\ast\), for every \textit{reflexive} Banach space \(Y\) (\(\lambda\)-\(BAP\) being characterized by this condition for \textit{every} Banach space \(Y\)). It follows that if a Banach space \(X\) is complemented in its bidual by a projection \(P\) and \(X\) has \(AP\), then \(X\) has the \(\| P\| \)-\(wBAP\). In particular, every dual space which has \(AP\) has \(wMAP\). Recall that an old open problem is: does \(AP\) for a dual Banach space \(X^\ast\) imply that \(X^\ast\) has \(MAP\)? By the above result, this question is equivalent to knowing if \(wMAP\) for a dual space implies \(MAP\). The second main result states that for every Banach space \(X\) having \(wMAP\) and whose dual \(X^\ast\) is the closed linear span of the weak-star strongly exposed points of its unit ball \(B_{X^\ast}\) (in particular, if \(X^\ast\) has the Radon--Nikodým property and contains no proper norming closed subspace), this dual \(X^\ast\) has \(wMAP\). This result allows the authors to show the above quoted fact that the Johnson--Schechtman space does not have \(wMAP\). They also show that \(X^\ast\) has \(wMAP\) if and only if \(X\) has \(MAP\) with respect to every equivalent norm. It follows that the above open problem is equivalent to: does \(X^\ast\) has \(MAP\) when \(X\) have \(wMAP\) with respect to every equivalent norm? Their proofs use, in many occurences, the isometric version of Davis--Figiel--Johnson--Pełczyńki's factorization lemma, proved by \textit{A.~Lima, O.~Nygaard} and \textit{E.~Oja} [Isr.\ J.\ Math.\ 119, 325--348 (2000; Zbl 0983.46024)].