A characterization of bounded convex approximation properties (Q730014)

Let \(X\) be a Banach space and \(A\) an arbitrary subset of \(\mathcal L(X)\), the space of all linear and continuous operators from \(X\) into itself. The Banach space \(X\) has the \(A\)-approximation property if there exists a net in \(A\) that converges to the identity uniformly on compact sets. This definition generalizes different types of approximation properties. For instance, if \(A\) is the set of all finite rank operators, we get the classical approximation property, and, if \(X\) is a Banach lattice and \(A\) is the set of all positive finite rank operators, we get the positive approximation property. Denoting by \(A^{\circ}\) the polar set of \(A\), and by \(A^{\text{conj}}=\{S^{*}: S\in A\} \subset \mathcal L(X^{*})\), the main Theorem of the article is the following: Theorem 2.2. Let \(X\) be a Banach space, and let \(A\) be a bounded convex subset of \(\mathcal L(X)\) containing \(0\). Then {\parindent=6mm \begin{itemize}\item[(1)] \(X\) has the \(A\)-approximation property if and only if there exists \(\phi \in A^{\circ \circ}\subset (\mathcal L(X) )^{**}\) such that \[ \phi(x^{*}\otimes x)=x^{*}(x) \quad \quad \forall x^{*} \in X^{*},\;\forall x \in \;X; \] \item[(2)] \(X^{*}\) has the \(A^{\text{conj}}\)-approximation property if and only if there exists \(\phi \in A^{\circ \circ}\subset (\mathcal L(X) )^{**}\) such that \[ \phi(x^{*}\otimes x^{**})=x^{**}(x^{*}) \quad \quad \forall x^{*} \in X^{*},\;\forall x^{**} \in \;X^{**}; \] \end{itemize}} This is an extension of results by \textit{E. Oja} [J. Math. Anal. Appl. 323, No. 1, 666--679 (2006; Zbl 1112.46016)] where this was proved in the case of \(A=E\cap\lambda B_{\mathcal L(X)}\) with \(E\) a linear subspace of \(\mathcal L(X)\) and \(B_{\mathcal L(X)}\) denoting all operators of norm less than or equal to \(1\). Finally, the above Theorem is used to ``lift'' the \(A\)-approximation property of \(X\) to the \(A^{\text{conj}}\)-approximation property of \(X^{*}\) when \(A\) is a convex subset of \(B_{\mathcal L(X)}\) containing \(0\) and \(X\) has the unique extension property.