The set of separable states has no finite semidefinite representation except in dimension \(3\times 2\) (Q2050062)

Given integers \(n \ge m,\) let \(\mathrm{Sep}(n, m)\) be the set of {\em separable states} on the Hilbert space \(\mathbb{C}^n \otimes \mathbb{C}^m,\) i.e., \[\mathrm{Sep}(n, m) := \mathbf{conv}\{x x^\dagger \otimes y y^\dagger \ : \ x \in \mathbb{C}^n, |x| = 1, y \in \mathbb{C}^m, |y| = 1\}.\] Here \(x^\dagger\) indicates conjugate transpose, \(|x|^2 := x^\dagger x\) and \(\mathbf{conv}\) denotes the convex hull. We say that a convex set \(C \subset \mathbb{R}^d\) has a {\em semidefinite representation} (of size \(r\)) if it can be expressed as \(C = \pi(S),\) where \(\pi \colon \mathbb{R}^D \to \mathbb{R}^d\) is a linear map and \(S \subset \mathbb{R}^D\) is a convex set defined using a linear matrix inequality \[S = \{w \in \mathbb{R}^D \ : \ M_0 + w_1M_1 + \cdots + w_DM_D \succeq 0\}\] where \(M_0, \ldots, M_D\) are Hermitian matrices of size \(r \times r.\) It is known, from the earlier work of [\textit{S. L. Woronowicz}, Rep. Math. Phys. 10, 165--183 (1976; Zbl 0347.46063)] that for \(n + m \le 5,\) the set \(\mathrm{Sep}(n, m)\) is just the set of states which have a positive partial transpose, and hence it has a semidefinite representation. In the paper under review, the author shows that for \(n + m > 5,\) the set \(\mathrm{Sep}(n, m)\) has no semidefinite representation, and so this provides a new counterexample to the Helton-Nie conjecture [\textit{J. W. Helton} and \textit{J. Nie}, SIAM J. Optim. 20, No. 2, 759--791 (2009; Zbl 1190.14058)], which was recently disproved by \textit{C. Scheiderer} [SIAM J. Appl. Algebra Geom. 2, No. 1, 26--44 (2018; Zbl 1391.90462)]. The paper is very clear, well written and quite interesting.