How bad can a Banach space with the approximation property be? (Q796043): Difference between revisions

A Banach space X is said to have the approximation property (AP) if for every compact K in X and every \(\epsilon>0\) there is a bounded linear operator of finite rank \(T:X\to X\) such that \(\| Tx-x\|<\epsilon\) for all x in K. X is said to have the bounded approximation property (BAP) if we can take all these T uniformly bounded in norm. \textit{T. Figiel} and \textit{W. B. Johnson} [Proc. Am. Math. Soc. 41, 197-200 (1973; Zbl 0289.46015)] showed that the AP does not imply the BAP. The author considers the ''relative'' bounded approximation property BAP(J) with respect to a subspace J of L(X): The identity \(I_ x\) is in the closure of some ball in the topology of compact convergence. Thus \(BAP=BAP(F),\) where F is the subspace of finite-rank operators. He shows that AP does not imply BAP(J) for some wider classes J of operators, e.g. the weakly compact operators, the strictly singular operators or the completely continuous operators. Thus showing that a space with the AP may be ''quite bad''. The Figiel-Johnson construction provides also an example of a nonnuclear operator with a nuclear adjoint. This is possible here, too. In fact, for each of the classes J above, there is a Banach space X and an operator \(T=\sum x'\!_ n\otimes x_ n\) on X, so that T is nonnuclear, \(T^*\) is nuclear, \(\sum x'\!_ n(x_ n)\) converges and \(| \sum x'\!_ n(Ux_ n)| \leq \| U\|\) for every \(U\in J\). Thus, a nonnuclear operator may be ''quite good''.