Characterising Completely Positive Elementary Operators

From MaRDI portal





A linear mapping on a \(C^*\)-algebra \(A\) of the form \[ S\colon x\mapsto\sum_{j=1}^n a_jxb_j\qquad(x\in A) \] where \(a_j,b_j\) are finitely many elements in the multiplier algebra \(M(A)\) of \(A\) is called an elementary operator on \(A\), and \(n\) is said to be the length of the representation of \(S\) as a sum of two-sided multiplications \(x\mapsto axb\). The minimum of all \(n\in\mathbb{N}\) which appear as lengths of representations of \(S\) is called the minimal length \(\ell(S)\) of \(S\). It emerged that it is this number that plays the decisive role in determining the positivity properties of \(S\). It has been known for some time [\textit{M. Mathieu}, Math. Jap. 29, 375-382 (1984; Zbl 0546.46048)] that, if \(\ell(S)=1\), positivity of \(S\) implies complete positivity, and also that, for arbitrary \(S\), \(n\)-positivity with \(n=\ell(S)\) yields complete positivity. However, the question to what extent one could decrease \(n\) and still have complete positivity has been open for more than 10 years. Theorem 1. Let \(S\) be an elementary operator on a \(C^*\)-algebra \(A\) with minimal length \(\ell(S)\geq 1\). If \(S\) is \(n\)-positive for some natural number \(n\geq{\ell(S)-1\over 2}\) then \(S\) is completely positive. On the technical side, the existence of sufficiently many finite rank operators plays an important role. Hence it is interesting, and surprising, that once all finite rank operators are erased (which happens for instance in the Calkin algebra) then there is no more obstacle towards complete positivity left, as the second result shows. Theorem 2. Every positive elementary operator on a \(C^*\)-algebra with the property that every non-zero quotient is antiliminal is completely positive.











This page was built for publication: Characterising Completely Positive Elementary Operators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3839832)