Toeplitz separability, entanglement, and complete positivity using operator system duality

TL;DR

This paper provides a new proof of Gurvits's theorem on the absence of entanglement in positive block-Toeplitz matrices using operator system duality, and explores implications for infinite-dimensional operator matrices.

Contribution

It introduces a novel proof of Gurvits's separation theorem via operator system duality and characterizes the structure of extremal rays in positive Toeplitz matrix cones.

Findings

Positive block-Toeplitz matrices have no entangled elements.

Normal positive linear maps are partially completely positive on Toeplitz matrices.

A factorization theorem links Gurvits's separation approach with Ando's universality approach.

Abstract

A new proof is presented of a theorem of L.~Gurvits, which states that the cone of positive block-Toeplitz matrices with matrix entries has no entangled elements. The proof of the Gurvits separation theorem is achieved by making use of the structure of the operator system dual of the operator system $C(S^1)^{(n)}$ of $n\times n$ Toeplitz matrices over the complex field, and by determining precisely the structure of the generators of the extremal rays of the positive cones of the operator systems $C(S^1)^{(n)}\omin\B(\H)$ and $C(S^1)_{(n)}\omin\B(\H)$, where $\H$ is an arbitrary Hilbert space and $C(S^1)_{(n)}$ is the operator system dual of $C(S^1)^{(n)}$. Our approach also has the advantage of providing some new information concerning positive Toeplitz matrices whose entries are from $\B(\H)$ when $\H$ has infinite dimension. In particular, we prove that normal positive linear maps $\psi$ on $\B(\H)$ are partially completely positive in the sense that $\psi^{(n)}(x)$ is positive whenever $x$ is a positive $n\times n$ Toeplitz matrix with entries from $\B(\H)$. We also establish a certain factorisation theorem for positive Toeplitz matrices (of operators), showing an equivalence between the Gurvits approach to separation and an earlier approach of T.~Ando to universality.