Quasi-multipliers and algebrizations of an operator space (Q2460023)

The paper may be viewed as a development of the theory of quasi-multipliers introduced by the author and \textit{V.\,I.\thinspace Paulsen} [J.~Funct.\ Anal.\ 217, No.\,2, 347--365 (2004; Zbl 1067.46050)]. Given an operator space \(X\), there is an injective envelope \(I(X)\), an (injective) operator space containing (a copy of) \(X\) and which has the structure of a TRO (ternary ring of operators). Many authors use the Paulsen operator system \({\mathcal S}_X\) associated with \(X\), and view \(I(X)\) as the upper right corner of the \(C^*\)-algebra \(I({\mathcal S}_X)\). An element \(z \in I(X)^*\) is called a quasimultiplier when \(X z X \subset X\) and the set of such \(z\) is denoted \({\mathcal Q \mathcal M}(X)\). Given an operator algebra \(\mathcal A\) and a completely isometric embedding \(\pi:X\to{\mathcal A}\), there is a relative quasimultiplier space \({\mathcal Q \mathcal M}^\pi(X)=\{a\in{\mathcal A}:\pi(X)a\pi(X)\subset\pi(X)\}\). Then, from the earlier work, \({\mathcal Q \mathcal M}(X)\) is maximal in the sense that there exists a unique completely contractive \(\sigma:{\mathcal Q \mathcal M}^\pi(X)\to{\mathcal Q \mathcal M}(X)\) with \(\pi(x_1)a\pi(x_2)=\pi(x_1\sigma(a)x_2)\). One section of the paper is devoted to a concept of quasi-centralizer for an operator algebra \(\mathcal A\) that is assumed to have a contractive approximate identity (c.a.i.). The results characterise the space \({\mathcal Q \mathcal C}({\mathcal A})\) of quasi-centralizers in terms of quasi-multipliers and show that for a \(C^*\)-algebra \(\mathcal A\), \({\mathcal Q \mathcal C}({\mathcal A})\) agrees with a standard notion. The final section characterises operator algebras in operator space terms. Given an operator space \(X\) and a bilinear map \(\phi:X \times X \to X\), (i) there is a complete isometry from \(X\) into an operator algebra such that \(\phi\) becomes the algebra product if and only if (ii) there is \(z \in {\mathcal Q \mathcal M}(X)\), \(\| z\| \leq 1\), \(\phi(x_1, x_2) = x_1 z x_2\) if and only if (iii) a certain map \(\Gamma_\phi\) defined using \(\phi\) on a certain upper triangular subspace of \(M_2( I({\mathcal S}_X) \otimes_h I({\mathcal S}_X))\) is completely contractive. The map \(\Gamma_\phi\) takes values in upper triangular \(2 \times 2\) matrices over \(X\), and its top right corner is the map \(X \otimes_h X \to X\) linearising \(\phi\) on the Haagerup tensor product. The equivalence of (i) and (ii) is already in the earlier work. This result is used to deduce (a non-unital version of) the Blecher--Ruan--Sinclair theorem for \(X\) an operator space, \(\phi\) a completely contractive bilinear map on \(X\), provided that there are left and right versions of a \(\phi\)-c.a.i.\ (showing that \(X\) is completely isometric to an operator algebra so that \(\phi\) corresponds to the algebra product).