Maximal vectors in Hilbert space and quantum entanglement (Q1011436)

Let \(H\) be a separable Hilbert space and \(V\) be a norm-closed subset of the unit sphere of \(H\) that is stable under multiplication by scalars of absolute value \(1\) and has the property that \(\langle \xi, V\rangle=\{0\}\) implies \(\xi=0\) for any \(\xi \in H\). A maximal vector (for \(V\)) is a unit vector \(\xi \in H\) whose distance to \(V\) is maximal, i.e., \(d(\xi,V) = \sup_{\|\eta\|}d(\eta,V)\), where \(d(\eta,V)\) denotes the distance from \(\eta\) to the set \(V\). A normal state \(\rho\) of \(B(H)\) is called \(V\)-correlated if, for every \(\varepsilon>0\), there are vectors \(\xi_1, \dots, \xi_n \in V\) and positive numbers \(t_1, \dots, t_n\) with sum \(1\) such that \(\sup_{\|x\| \leq 1}\left|\rho(x)-\sum_{k=1}^nt_k\langle x\xi_k,\xi_k\rangle\right| \leq \varepsilon\). A normal state that is not \(V\)-correlated is called (\(V\)-)entangled. The paper under review is a sequel in a series of the author's papers related to entangled states on matrix algebras [cf.\ ``Noncommutative dynamics and E-semigroups'' (Springer Monographs in Mathematics) (2003; Zbl 1032.46001); ``The probability of entanglement'' (preprint) (2007; arXiv:0712.4163)]. The author introduces a numerical invariant \(r(V)\) of \(V\) (the inner radius) and shows that when \(r(V) > 0\), there is a uniquely determined ``entanglement measuring norm'' \(\|\cdot\|^V\) on \(H\) with the property that \(\xi \in V\) if and only if \(\|\xi\|^V=1\) and \(\xi\) is maximal if and only if \(\|\xi\|^V=r(V)^{-1}\). In addition, he shows that there is a corresponding entanglement-measuring norm on the predual of \(B(H)\) that faithfully detects entanglement of normal states. He gives an analysis of entanglement in multipartite tensor products \(H = H_1 \otimes \dots \otimes H_N\) and calculates both entanglement-measuring norms. In the cases for which \(\dim H_N\) is relatively large with respect to the others, he describes the set of maximal vectors in explicit terms as well.