Grothendieck's theorem for operator spaces (Q1401447)

The theorem of Grothendieck in question asserts that a bounded linear operator from a \(C(K)\)-space into an \(L_1\)-space factors through a Hilbert space; equivalently, a bounded bilinear form on \(C(K_1)\times C(K_2)\) is Hilbertian. These results are known to extend to the noncommutative case by work of G. Pisier and of U. Haagerup; i.e., for \(C^*\)-algebras \(A\) and \(B\), every bounded linear operator \(T: A\to B^*\) factors through a Hilbert space. The paper under review addresses the completely bounded version of this result. There are two notions of complete boundedness for bilinear forms on operator spaces; apart from the existing notion due to \textit{E. Christensen} and \textit{A. Sinclair} [J. Funct. Anal. 72, 151--181 (1987; Zbl 0622.46040)] there is another one which is more natural from the point of view of operator spaces that is introduced here. The authors call a bilinear form \(u: E\times F \to {\mathbb C}\) on a pair of operator spaces jointly completely bounded if the associated linear map \(\widetilde{u}: E\to F^*\) is completely bounded. The main theorem of the paper can be rephrased most conveniently in terms of factorisation: If \(E\) and \(F\) are exact operator spaces, then a completely bounded operator \(T: E\to F^*\) can be decomposed into a sum \(T=T_1+T_2\) such that \(T_1\) factors completely boundedly through a row Hilbert space \(H_r\) (i.e., the operator space structure on \(H\) determined by \(B(H^*,{\mathbb C})\)) and \(T_1\) factors completely boundedly through a column Hilbert space \(K_c\) (i.e., the operator space structure on \(K\) determined by \(B({\mathbb C},K)\)). If \(E\) and \(F\) are \(C^*\)-algebras, it is enough to assume that one of them is exact or that \(T\) is well approximable by finite-rank operators. More technically, the main results are given in terms of estimates of jointly completely bounded bilinear forms with respect to the Haagerup tensor product. The proofs make use of the generalised circular elements studied by the second author [Pac. J. Math. 177, 329--368 (1997; Zbl 0882.46026)]. There are several important corollaries of the main results for an operator space \(E\): (1) If \(E\) and \(E^*\) embed completely isomorphically into the predual of a von Neumann algebra and \(E\) satisfies the completely bounded approximation property, then \(E\) is completely isomorphic to a quotient of a subspace of \(H_r \oplus K_c\) for some Hilbert spaces \(H\) and \(K\). (2) \(E\) and \(E^*\) are exact if and only if \(E\) is completely isomorphic to \(H_r \oplus K_c\). (3) Further corollaries refer to the operator Hilbert space OH. In the last section, the authors characterise the Schur multipliers that are completely bounded from the space of compact operators on \(H\) to the trace class.