The super operator system structures and their applications in quantum entanglement theory (Q665482)

Given a *-ordered vector space \((V,V^+,e)\) with an Archimedean order unit \(e\), \textit{V. I. Paulsen} et al. [Proc. Lond. Math. Soc. (3) 102, No. 1, 25--49 (2011; Zbl 1213.46050)] defined two extremal operator system structures on \(V\), denoted \(\text{OMIN}(V)\) and \(\text{OMAX}(V)\), whose underlying positive cones both coincide with \(V^+\), and which are characterised by the universal properties that, for every operator system \(\mathcal{T}\), every positive map from \(\mathcal{T}\) into \(V\) (resp. from \(V\) into \(\mathcal{T}\)) is automatically completely positive as a map from \(\mathcal{T}\) into \(\text{OMIN}(V)\) (resp. from \(\text{OMAX}(V)\) into \(\mathcal{T}\)). In the article under review, the author develops \(k\)-versions of the aforementioned operator system structures, where \(k\in \mathbb{N}\). For every operator system \(\mathcal{S}\) with unit \(1\) and positive cone \(\mathcal{S}^+\), and every \(k\geq 1\), she defines operator system structures \(\text{OMIN}_k(\mathcal{S})\) and \(\text{OMAX}_k(\mathcal{S})\) whose underlying *-ordered vector space is \((\mathcal{S}, \mathcal{S}^+,1)\) and which are characterised by the universal properties that for every operator system \(\mathcal{T}\), every \(k\)-positive map from \(\mathcal{T}\) into \(\mathcal{S}\) (resp. from \(\mathcal{S}\) into \(\mathcal{T}\)) is automatically completely positive as a map from \(\mathcal{T}\) into \(\text{OMIN}_k(\mathcal{S})\) (resp. from \(\text{OMAX}_k(\mathcal{S})\) into \(\mathcal{T}\)). The matricial cones of \(\text{OMIN}_k(\mathcal{S})\) and \(\text{OMAX}_k(\mathcal{S})\) are defined explicitly and a \(k\)-version of Kadison's characterisation theorem for function systems is given; the operator system \(\text{OMIN}_k(\mathcal{S})\) can thus be realised as an operator subsystem of an operator system of the form \(M_k(C(X))\), where \(C(X)\) is the operator system of all continuous functions on a compact Hausdorff space \(X\) and \(M_k(V)\) is the space of all \(k\) by \(k\) matrices with entries in the vector space \(V\). The dual matrix spaces of \(\text{OMIN}_k(\mathcal{S})\) and \(\text{OMAX}_k(\mathcal{S})\) are identified explicitly. This identifiction is then used to characterise the \(k\)-partially entanglement breaking maps. Namely, it is shown that the following statements are equivalent, for a linear map \(\phi : M_p \to M_m\): {\parindent=8mm \begin{itemize}\item[(i)] for every \(n\in \mathbb{N}\) and every state \(s : M_n\otimes M_m\to \mathbb{C}\), the state \(s\circ \phi^{(n)} : M_n\otimes M_p\to \mathbb{C}\) is \(k\)-separable (here, \(\phi^{(n)} = \text{id}_{M_n}\otimes\phi\)); \item[(ii)] \(\phi : \text{OMIN}_k(M_p) \to M_m\) is completely positive; \item[(iii)] \(\phi : M_p \to \text{OMAX}_k (M_m)\) is completely positive; \item[(iv)] \(\phi : \text{OMIN}_k(M_p) \to \text{OMAX}_k (M_m)\) is completely positive; \item[(v)] there exist matrices \(A_l\in M_{p,m}\), \(l = 1,\dots,r\), of rank at most \(k\), such that \(\phi(X) = \sum_{l=1}^r A_l^* X A_l\), \(X\in M_p\); \item[(vi)] there exist completely positive maps \(\psi_l : M_p\to M_k\) and matrices \(B_l\in M_{k,m}\), \(l = 1,\dots,s\), such that \(\phi(X) = \sum_{l=1}^s B_l^* \psi_l(X) B_l\), \(X\in M_p\). \end{itemize}}