Characterisations of von Neumann algebras (Q2360012)

A good number of outstanding results in \(C^*\)-algebra theory are due to S. Sakai. One of the most transcendental results is a characterization of \(W^\ast\)-algebras as those \(C^*\)-algebras that have an isometric predual [\textit{S. Sakai}, Pac. J. Math. 6, 763--773 (1956; Zbl 0072.12404)]. Let us recall that a predual of a Banach space \(\mathcal{A}\) is a pair \((E, \kappa)\) consisting of a Banach space \(E\) and a linear homeomorphism \(\kappa : \mathcal{A} \to E^*\). A predual \((E, \kappa)\) is called isometric if the mapping \(\kappa\) is isometric. Two preduals \((E_1, \kappa_1)\) and \((E_2, \kappa_2)\) are identified if there exists a linear homeomorphism \(T: E_1\to E_2\) satisfying \(\kappa_1= T^* \circ \kappa_2\), or equivalently, the weak\(^*\)-topologies on \(\mathcal{A}\) associated with \((E_1, \kappa_1)\) and \((E_2, \kappa_2)\) are identical. It should also be recalled that a von Neumann algebra is a weak\(^*\)-closed unital self-adjoint algebra of operators on a complex Hilbert space, and that a \(W^\ast\)-algebra is a \(C^*\)-algebra that has a faithful \(^*\)-representation as a von Neumann algebra on some Hilbert space. Other than by Sakai, the question of determining an intrinsic characterization of \(W^*\)-algebras has been treated by authors like \textit{R. V. Kadison} [Ann. Math. (2) 64, 175--181 (1956; Zbl 0071.11501); Expo. Math. 3, 193--227 (1985; Zbl 0576.46043)], \textit{G. K. Pedersen} [Bull. Lond. Math. Soc. 4, 171--175 (1972; Zbl 0252.46071)], and \textit{J. Dixmier} [Bull. Soc. Math. Fr. 81, 9--39 (1953; Zbl 0050.11501)]. The first aim of the interesting paper under review is an attempt to relax the (iso)metric requirement of Sakai's theorem to find a characterization of \(W^*\)-algebras as those \(C^*\)-algebras having a predual (in the sense above) that is compatible with (a priori only part of) the algebraic structure. Theorem 2.1 establishes the following characterization: Let \(\mathcal{A}\) be a \(C^*\)-algebra and let \(\tau\) be a locally convex Hausdorff topology on \(\mathcal{A}\) weaker than the norm topology which satisfies the following conditions:{\parindent=0.7cm\begin{itemize}\item[(i)] the mapping \(x\to axa\) is \(\tau\)-continuous on the closed unit ball \(\mathbf{S}_{\mathcal{A}}\) of \(\mathcal{A}\) for every self-adjoint element \(a\) in \(A\); \item[(ii)] \(\mathbf{S}_{\mathcal{C}}\) is relatively \(\tau\)-compact for every maximal commutative \(C^*\)-subalgebra \(\mathcal{C}\) of \(\mathcal{A}\). \end{itemize}} Then \(\mathcal{A}\) is a \(W^*\)-algebra and the space of \(\tau\)-continuous linear functionals is dense in \(\mathcal{A}_*\). If (ii) is replaced with the stronger condition that \(\mathbf{S}_{\mathcal{A}}\) is relatively \(\tau\)-compact, then the topology on \(\mathbf{S}_{\mathcal{A}}\) induced by \(\tau\) is identical to that by the canonical weak\(^*\)-topology. The author employs an argument similar to those developed by Kadison and Sakai based on the construction of a faithful \(^*\)-representation of \(\mathcal{A}\) from suitably chosen positive linear functionals, in this case, those which are \(\tau\)-continuous. The approach requires ingenious arguments to find sufficiently many positive linear functionals. It is also shown in Corollary 2.3 that, if \(\mathcal{A}\) is a \(C^*\)-algebra endowed with a locally convex Hausdorff topology \(\tau\) weaker than the norm topology which satisfies the following conditions:{\parindent=0.7cm\begin{itemize}\item[(i)] the Jordan product is separately \(\tau\)-continuous on the closed unit ball \(\mathbf{S}_{\mathcal{A}}\) of \(\mathcal{A}\), \item[(ii)] \(\mathbf{S}_{\mathcal{A}}\) is relatively \(\tau\)-compact, \end{itemize}} then \(\mathcal{A}\) is a \(W^*\)-algebra, and the topology on \(\mathbf{S}_{\mathcal{A}}\) induced by \(\tau\) is identical to that by the canonical weak\(^*\)-topology. This applies when \(\mathcal{A}\) is a \(C^*\)-algebra admitting a predual \((E, \kappa)\) such that the multiplication on \(A\) is separately weak\(^*\)-continuous. In Section 3, the author explores the possibility to remove the algebraic conditions in the previous statements (i) by strengthening the topological requirement from relative \(\tau\)-compactness to that of (full) \(\tau\)-compactness. Theorem 3.1 proves that, if \(\mathcal{A}\) is a \(C^*\)-algebra endowed with a locally convex Hausdorff topology \(\tau\) weaker than the norm topology such that \(\mathbf{S}_{\mathcal{A}}\) is \(\tau\)-compact, then \(\mathcal{A}\) is a \(W^*\)-algebra, and the topology on \(\mathbf{S}_{\mathcal{A}}\) induced by \(\tau\) is identical to that by the canonical weak\(^*\)-topology. Sakai's theorem is derived as a consequence of this result. Actually, if \(\mathcal{A}\) is a \(C^*\)-algebra endowed with a locally convex Hausdorff topology \(\tau\) weaker than the norm topology such that \(\mathbf{S}_{\mathcal{C}}\) is \(\tau\)-compact for every maximal commutative \(C^*\)-subalgebra \(\mathcal{C}\) of \(\mathcal{A}\), then \(\mathcal{A}\) is a \(W^*\)-algebra, and the space of \(\tau\)-continuous linear functionals is dense in \(\mathcal{A}_*\) (see Theorem 3.3). This result is applied to obtain a strengthened version of a result due to Pedersen by showing that, if an \(AW^*\)-algebra \(\mathcal{A}\) admits a separating set \(E\) of completely additive linear functionals, then \(\mathcal{A}\) is a \(W^*\)-algebra. This is a very well motivated paper on a fundamental topic in the theory of \(C^*\)-algebras. The reader will appreciate the detailed presentation and the motivating results established by the author.