Entangleability of cones (Q2041861)

Given two finite-dimensional cones \(\mathcal{C}_{1}, \mathcal{C}_{2},\) one can define ``from the inside'' their minimal tensor product \(\mathcal{C}_{1} \bigodot\mathcal{C}_{2}\) or ``from outside'' their maximal tensor product \(\mathcal{C}_{1}\circledast \mathcal{C}_{2}.\) Notice that \(\mathcal{C}_{1} \bigodot\mathcal{C}_{2}\subseteq \mathcal{C}_{1}\circledast\mathcal{C}_{2}\) and \(\mathcal{C}_{1}\circledast\mathcal{C}_{2}= (\mathcal{C}^{*}_{1}\bigodot\mathcal{C}^{*}_{2})^{*}.\) Here, \(\mathcal{C}^{*}\) is the dual cone to a cone \(\mathcal{C}.\) Along the paper, the authors focus the attention on proper cones i.e. closed convex cones \(\mathcal{C}\) in a finite-dimensional real linear space \(V\) such that both \(\mathcal{C}\bigcap(-\mathcal{C})= {0}, \mathcal{C}-\mathcal{C}= V.\) A pair of cones \((\mathcal{C}_{1}, \mathcal{C}_{2})\) is said to be nuclear if \(\mathcal{C}_{1} \bigodot\mathcal{C}_{2}= \mathcal{C}_{1}\circledast\mathcal{C}_{2}\) while the pair is said to be entangleable if \(\mathcal{C}_{1} \bigodot\mathcal{C}_{2}\neq \mathcal{C}_{1}\circledast\mathcal{C}_{2}.\) The terminology ``nuclear'' is borrowed from the analogous notion in \(C^{*}\)-algebras, while the concept of entangleability comes from the interpretation of cones in the context of general probabilistic theories (quantum mechanics is a special case). Cones corresponding to quantum mechanics belong to the family \((PSD_{n})_{n\geq1}\), where \(PSD_{n}\) denotes the cone of positive semi-definite matrices of size \(n\times n\) with complex entries. Quantum entanglement is connected with the fact that \((PSD_{m}, PSD_{n}), m, n\geq 2,\) is an entangleable pair. In the paper, the authors prove a simple and nice characterization of nuclearity conjectured in [\textit{G. P. Barker}, Linear Algebra Appl. 39, 263--291 (1981; Zbl 0467.15002)]. Indeed, a pair of proper cones is nuclear if and only if on of them is classical (see Theorem A). Here, classical means that the cone is isomorphic to \(\mathbb R^{n}_{+},\) or equivalently a cone whose bases are simplices. The proof is based on a geometric property, the kite-square sandwiching, that characterises non-classical cones (see Theorem B). This geometric property involves cones based on two specific planar shapes: the \emph{kite} and the\emph{ blunt square}. In the next step, they show that kite-square sandwichings can be used to produce a certificate of entangleability. Indeed, two auxiliary results deal with the fact that a proper cone is non-classical if and only if it admits a kite-square sandwiching (Theorem B) and if a pair of proper cones admit a kite square-sandwiching, then the pair is entangleable (Theorem C). A generalisation of Theorem A for a finite number of proper cones is also deduced. As a byproduct of Theorem A, the authors answer a question raised in [\textit{B. Passer} et al., J. Funct. Anal. 274, No. 11, 3197--3253 (2018; Zbl 1422.47021)] in the framework of matrix convex sets.