A characterization of completely 1-complemented subspaces of noncommutative \(L_ 1\)-spaces. (Q701286): Difference between revisions

The present paper is devoted to two intertwined topics: (1) completely contractively complemented subspaces in the preduals of ternary rings of operators (TROs); and (2) the geometric structure of the preduals of von Neumann algebras. A ternary ring of operators (TRO) is a closed subspace \(X\) of \(B(H)\) satisfying \(X X^* X \subset X\). It is well-known that every TRO \(X\) is a ``corner'' of a \(C^*\)-algebra. In other words, \(X = pAq\), where \(p\) and \(q\) are mutually orthogonal projections in a \(C^*\)-algebra \(A\). In [\textit{E.~Effros}, \textit{N.~Ozawa} and \textit{Z.-J.~Ruan}, Duke Math. J. 110, 489--521 (2001; Zbl 1010.46060)], it was shown that, if a TRO \(X\) is a dual Banach space, then \(X = pRq\), where \(p\) and \(q\) are mutually orthogonal projections in a von Neumann algebra \(R\). Moreover, such an \(X\) has a unique predual. In the present paper, the results quoted above are put to a good use. Generalizing the classical results on contractively complemented subspaces of \(L_1\)-spaces, the authors show that, if \(Y\) is a predual to some TRO and \(Z\) is a subspace of \(Y\), then \(Z\) is completely contractively complemented in \(Y\) if and only if \(Z\) is completely isometric to a predual to a TRO. In particular, this result holds if \(Y\) is a predual to a von Neumann algebra. The authors describe the structure of finite-dimensional completely contractively complemented subspaces of von Neumann algebras, as well as projections upon them, in great detail. Last but not least, the paper deals with the local structure of preduals of injective von Neumann algebras. It is proved that, for an operator space \(X\), the following three conditions are equivalent: (1) \(X\) is completely isometric to a predual of an injective von Neumann algebra; (2) \(X\) is an \(OL_{1,1+}\)-space; and (3) \(X\) is a rigid \(OL_{1,1+}\)-space. Recall that \(X\) is called an \(OL_{1,1+}\)-space if, for every finite-dimensional subspace \(E \subset X\) and \(\varepsilon > 0\), there exist a finite-dimensional subspace \(E \subset F \subset X\) and positive integers \(n_1, \ldots , n_k\) for which \[ d_{cb} (F, T_{n_1} \oplus_1 \ldots \oplus T_{n_k}) < 1 + \varepsilon , \] where \(T_n\) is the \(n \times n\) trace class space. The condition of \(X\) being a rigid \(OL_{1,1+}\)-space is formally stronger: it means that for any finite-dimensional \(E \subset X\) and \(\varepsilon > 0\), there exists a finite-dimensional subspace \(F \subset X\), isometric to \(T_{n_1} \oplus_1 \ldots \oplus_1 T_{n_k}\) for some positive integers \(n_1, \ldots , n_k\), such that \[ \sup_{e \in E, \| e\| \leq 1} \inf_{f \in F} \| e-f\| < \varepsilon . \] The proofs presented are very ingenious and interesting in their own right. One result is worth mentioning: suppose that \(W\) and \(V\) are weak\(^*\)-closed TROs and \(\phi : W \rightarrow V\) is a weak\(^*\)-continuous complete metric surjection. Then there exists a sub-TRO \(W_0\) of \(W\) such that \((\phi(x)) (\phi(y))^* (\phi(z)) = \phi(xy^*z)\) whenever \(x, y, z\) are in \(W\), and at least two of them are in \(W_0\). Moreover, there exists a weak\(^*\)-continuous complete contraction \(\psi : V \rightarrow W\) such that \(\phi \circ \psi = \text{Id}_V\).