Duality and operator algebras. I: Automatic weak\(^{\ast}\) continuity and applications (Q2484255)

This series of two papers [for part II, see Zbl 1083.46033 below] are devoted to investigating the weak\(^*\)-continuity of multiplication. Some remarkable theorems on automatic weak\(^*\)-continuity are proved in the first part, while the second one is, by and large, devoted to counterexamples. Recall that a left \(A\)-module (\(A\) is a \(C^*\)-algebra) \(X\) is called representable if \[ \|a_1 x_1 + a_2 x_2\|^2 \leq \|a_1 a_1^* + a_2 a_2^*\| (\|x_1\|^2 + \|x_2\|^2) \] for any \(a_1, a_2 \in A\) and \(x_1, x_2 \in X\). According to \textit{B.~Magajna} [Proc. Edinb. Math. Soc. 42, 191--208 (1999; Zbl 0932.46039)], a module is representable if and only if it can be represented isometrically as a concrete \(A\)-module. The main result of the first paper shows that, if a dual Banach space \(X\) is a representable module over a \(C^*\)-algebra \(A\), then the mapping \(x \mapsto ax\) is weak\(^*\)-continuous for every \(a \in A\). The key tool for the proof is a module analogue of Tomiyama's well-known description of conditional expectations: if \(X\) is a representable submodule over a \(C^*\)-algebra \(A\), \(Y\) is a submodule of \(X\), and \(\Phi\) is a contractive projection from \(X\) onto \(Y\), then \(\Phi\) is an \(A\)-module map. Below, we shall see that this statement is false if \(A\) is assumed to be an operator algebra, without any self-adjointness properties. In a rather straightforward manner, the main theorem implies the following Corollary. Suppose that \(X\) is a dual Banach space, \(a\) is an element of a \(C^*\)-algebra \(A\), \(\sigma :X \to A\) is a linear isometry, \(a \sigma(X) \subset \sigma(X)\), and \(a^* \sigma(X) \subset \sigma(X)\). Then \(x \mapsto \sigma^{-1}(a\sigma(x))\) is a weak\(^*\)-continuous map on \(X\). The authors use this to reprove the results of Effros--Ozawa--Ruan--Zettl on the representation of ternary rings of operators (TROs) which are dual Banach spaces as corners of von Neumann algebras, and uniqueness of preduals of such TROs. Using a ``concrete'' description of multipliers on operator spaces, one also obtains that every left multiplier on a dual operator space is weak\(^*\)-continuous. Consequently, if \(A\) is an operator algebra which is also a dual operator space, then multiplication in \(A\) is separately weak\(^*\)-continuous. Therefore, according to [\textit{D.~Blecher} and \textit{C.~Le~Merdy}, ``Operator algebras and their modules -- an operator space approach'' (Oxford University Press) (2004; Zbl 1061.47002)], for every such operator algebra \(A\) there exists a weak\(^*\)-continuous complete isometry \(J : A \rightarrow B(H)\) such that \(J(A)\) is \(\sigma\)-weakly closed, and \(J(ab) = J(a) J(b)\) for all \(a, b \in A\). Several other interesting results (not quoted here for lack of space) fit nicely with the investigation of dual operator algebras and operator modules (see the book by Blecher and Le Merdy for more information on such objects). In the ``classical'' setting, the authors show that every multiplier on a dual Banach space is weak\(^*\)-continuous. In the last section of the paper, the results described above are employed to investigate Morita equivalence between non-selfadjoint algebras. The second part of this two-part series shows that the results quoted above are sharp: if the assumptions are relaxed, then the multiplication operator no longer needs to be weak\(^*\)-continuous. First, the authors present some examples showing that the words ``operator space dual'' in the theorems quoted above cannot be replaced with the words ``Banach space dual''. Among other things, the authors prove the existence of (1) an operator algebra \(B\), which is a dual Banach space, containing an idempotent \(p\) such that \(\dim (pB) = 1\), and the operator \(b \mapsto pb\) (\(b \in B\)) is not weak\(^*\)-continuous; (2) a commutative operator algebra \(D\), which is a dual Banach space, containing a nilpotent element \(p\) such that the operator \(d \mapsto pd\) (\(d \in D\)) is not weak\(^*\)-continuous; consequently, \(D\) is not weak\(^*\) homeomorphic to a \(\sigma\)-weakly closed operator algebra. As shown in the first part, the weak\(^*\)-continuity of the multiplication operator would be automatic if either \(B\) or \(D\) had an operator (as opposed to Banach) space predual. These and other counterexamples in the paper are obtained by a very ingenious use of fairly elementary methods. In general, the self-adjointness requirement (in its various guises) cannot be eliminated, either. Without it, the analogue of Tomiyama's theorem fails. More precisely, there exists a commutative operator algebra \(A\) with a norm \(1\) identity \(e\), a contractive projection \(P\) from \(A\) onto its subalgebra \(B\) containing \(e\), and elements \(a \in A\), \(b \in B\), s.t. \(P(ab) \neq P(a)b\).