On \(\alpha\)-nuclear operators with applications to vector-valued function spaces (Q495973): Difference between revisions

Let \(X, Y, Z\) be Banach spaces. A~tensor norm \(\alpha\) defines a Banach operator ideal \(\mathcal{N}_\alpha = (\mathcal{N}_\alpha,\|\cdot\|_\alpha)\); the \(\alpha\)-nuclear operators. Here, \(\mathcal{N}_\alpha(X,Y)\) is the range of the map \(J_\alpha: X^* \hat{\otimes}_\alpha Y \to \mathcal{L}(X,Y)\), where \(\mathcal{L}(X,Y)\) are the bounded linear operators from \(X\) to \(Y\). The basic result of the article is Theorem 1.2: Assume that \(Y\) is a closed subspace of \(Z\) such that there exists an extension operator \(\Phi \in \mathcal{L}(Y^*,Z^*)\), i.e., \(\Phi(y^*)(y) = y^*(y)\) for all \(y^* \in Y^*\) and \(y \in Y\). Let \(j: Y \to Z\) be the identity embedding. If \(T \in \mathcal{L}(X,Y)\) and \(jT \in \mathcal{N}_\alpha(X,Z)\), then \(T \in \mathcal{N}_\alpha(X,Y)\), whenever \(X^*\) or \(Z^*\) has the approximation property. Moreover, \[ \|\Phi\|^{-1}\|T\|_{\mathcal{N}_\alpha} \leq \|jT\|_{\mathcal{N}_\alpha} \leq \|T\|_{\mathcal{N_\alpha}}. \] For \(1 \leq p \leq \infty\), let \(\mathcal{C}_p(\Omega,X)\) be the space of continuous functions from a compact Hausdorff space \(\Omega\) to a Banach space \(X\) whose range is \(p\)-compact (in the sense of \textit{D. P. Sinha} and \textit{A. K. Karn} [Stud. Math. 150, No. 1, 17--33 (2002; Zbl 1008.46008)]). The spaces \(\mathcal{C}_p(\Omega,X)\) are studied in detail in Section~3 resulting in a characterization (whose proof uses Theorem~1.2): \(\mathcal{C}_p(\Omega,X) = C(\Omega) \hat{\otimes}_{d_p} X\), where \(d_p\) is the Chevet-Saphar tensor norm. Similarly, the space \(\mathcal{UC}_p(\Omega,X)\) of continuous functions from \(\Omega\) to \(X\) whose range is unconditionally \(p\)-compact, can be canonically identified with the Fourie-Swart tensor product \(C(\Omega) \hat{\otimes}_{w_{p'}} X\), where \(1/p + 1/p' = 1\).