Factorization of completely bounded bilinear operators and injectivity (Q1270305)

A von Neumann algebra \({\mathcal N}\) is said to be injective if, for each von Neumann algebra \({\mathcal M}\) containing \({\mathcal N}\), there exists a completely positive projection from \({\mathcal M}\) onto \({\mathcal N}\). Various authors have discovered other conditions on \({\mathcal N}\) equivalent to its being injective, usually under some more restrictive condition. In this tour de force, the authors prove in complete generality that the injectivity of \({\mathcal N}\) is equivalent to its having the property that, for each completely bounded bilinear operator \(\phi\) from \({\mathcal N}\times{\mathcal N}\) to \({\mathcal N}\), there exist families \(\{\psi_j:j\in \Lambda\}\) and \(\{\theta_j:j\in \Lambda\}\) of completely bounded linear operators from \({\mathcal N}\) to \({\mathcal N}\) such that \[ \phi= \sum_{j\in\Lambda} \psi_j\otimes \theta_j, \] where convergence is in the weak operator topology.