English

Dual spaces of operator systems

Operator Algebras 2022-02-10 v3 Functional Analysis

Abstract

This article is to give an infinite dimensional analogue of a result of Choi and Effros. We say that an (not necessarily unital) operator system TT is \emph{dualizable} if one can find an equivalent dual matrix norm on the dual space TT^* such that under this dual matrix norm and the canonical dual matrix cone, TT^* becomes a dual operator system. We show that "a complete" operator system TT is dualizable if and only if M(T)saM_\infty(T)^\mathrm{sa} satisfies a bounded decomposition property. In this case, fd:=sup{[fi,j(xk,l)]:xMn(T)+;x1;nN},\|f\|^\mathrm{d}:= \sup \big\{\big\|[f_{i,j}(x_{k,l})]\big\|: x\in M_n(T)^+; \|x\|\leq 1; n\in \mathbb{N}\big\}, is the largest dual matrix norm that is equivalent to and dominated by the original dual matrix norm on TT^* that turns it into a dual operator system, denoted by TdT^\mathrm{d}. TdT^\mathrm{d} is again dualizable. For every completely positive completely bounded map ϕ:ST\phi:S\to T between dualizable operator systems, there is a unique weak-^*-continuous completely positive completely bounded map ϕd:TdSd\phi^\mathrm{d}:T^\mathrm{d} \to S^\mathrm{d} which is compatible with the dual map ϕ\phi^*. This gives a full and faithful functor from the category of dualizable operator systems to that of dualizable dual operator systems. Moreover, we will verify that that if SS is either a CC^*-algebra or a unital operator system, then SS is dualizable and the canonical weak-^*-homeomorphism from the unital operator system SS^{**} to the operator system (Sd)d(S^\mathrm{d})^\mathrm{d} is a completely isometric complete order isomorphism. Furthermore, the category of CC^*-algebras and that of unital "complete" operator systems can be regarded as full subcategories of the category of dual operator systems.

Keywords

Cite

@article{arxiv.2105.11112,
  title  = {Dual spaces of operator systems},
  author = {Chi-Keung Ng},
  journal= {arXiv preprint arXiv:2105.11112},
  year   = {2022}
}

Comments

The completeness assumption is missing from some statements in the published version. This completeness assumption is necessary because preduals of dual SMOS are assumed to be complete. We changed some statements in this revision. However, no change in the proof was made. All changes are marked in the color "magenta"

R2 v1 2026-06-24T02:23:46.976Z