Completeness in the Mackey topology by norming subspaces (Q2320041): Difference between revisions

Given a Banach space \(X\) and a \(w^*\)-dense subspace \(Y\subset X^*\), the Mackey topology \(\mu(X,Y)\) on \(X\) is the topology of uniform convergence on absolutely convex \(w^*\)-compact subsets of \(Y\). We say that \(X\) is \textit{fully Mackey complete} (resp. \textit{fully Mazur}) if \((X,\mu(X,Y))\) is complete (resp. every \(w^*\)-sequentially continuous linear functional \(f:Y\to\mathbb{R}\) is \(w^*\)-continuous) for every norming and norm-closed subspace \(Y\subset X^*\). In general, we have \[X\text{ is fully Mazur}\Rightarrow X\text{ is fully Mackey complete}\] and it is an open problem whether those two properties are equivalent. In this paper, the authors find new characterizations of both properties and they deduce some nice consequences. Let us mention some of the most interesting ones: \begin{itemize} \item[--] both properties (\(X\) being fully Mazur or fully Mackey complete) are inherited by closed subspaces, \item[--] the spaces \(\ell_1(\omega_1)\) and \(C([0,\omega_1])\) are not fully Mackey complete, \item[--] if \(X\) is separable, then \(X^*\) is fully Mackey complete if and only if \(\ell_1\not\hookrightarrow X\) and \((B_{X^{**}},w^*)\) is Fréchet-Urysohn, \item[--] if \(X\) is Asplund, then \(X^*\) is fully Mackey complete if and only if \(X^*\) is weakly Lindelöf determined. \end{itemize} Moreover, a connection with properties \((\mathcal {E})\) and \((\mathcal {E}')\) is examined -- we refer to the paper for a definition of \((\mathcal {E})\) and \((\mathcal {E}')\). Note that, in general, we have \[(B_{X^*},w^*)\text{ is Fréchet-Urysohn} \Rightarrow X\text{ has }(\mathcal {E})\Rightarrow X\text{ has }(\mathcal {E}')\; ,\] and examples separating the three conditions are not known in ZFC (but they are known under CH). The connection with the above notions found by the authors is: \begin{itemize} \item[--] every Banach space which has \((\mathcal {E})\) is fully Mazur, \item[--] if \(X\) has \((\mathcal {E}')\), then it is fully Mazur if and only if it is fully Mackey complete, \item[--] under (CH), there exists a Banach space which has \((\mathcal {E}')\) but is not fully Mazur. \end{itemize}