Dilation theory in finite dimensions and matrix convexity (Q2066128)

Let \(S\) be an operator system contained in a unital \(C^{*}\)-algebra \(A\). Let \(\varphi:S\rightarrow B(H)\) be a unital completely positive (u.c.p.) linear map. Arveson's extension theorem combined with Stinespring's dilation theorem proves the existence of a Hilbert space \(K\supset H\) and a unital \(*\)-homomorphism \(\pi: A\rightarrow B(K)\) such that for all \(s\in S\), \[ \varphi(s) = P_{H}\pi(s)|_{H}. \] Here \(P_{H}\) is the orthogonal projection from \(K\) onto \(H\). We say that \(\varphi\) dilates to a representation of \(A\). Conversely, any linear map \(\varphi\) of the above form is a u.c.p. map. In the case \(K\) is finite dimensional, we say that \(\varphi\) dilates to a finite-dimensional representation of \(A\). The main result in the paper shows that if \(A\) is a unital FDI \(C^{*}\)-algebra and \(\dim(S)<\infty\), then any u.c.p. map \(\varphi:S\rightarrow B(H)\) (with \(\dim(H)<\infty)\) dilates to a finite-dimensional representation of \(A\). This result can be regarded as a general principle to deduce finite-dimensional dilation results from their classical infinite-dimensional counterparts. Indeed, by applying their main result, the authors recover several previously known finite-dimensional dilation theorems and establish new ones, for example, Egerváry's theorem (which is a finite-dimensional version of the well-known Sz.-Nagy's dilation theorem); M\(^{\mathrm{c}}\)Carthy-Shalit's dilations of tuples of commuting contractive matrices; a finite-dimensional version of Agler's theorem on rational dilations; Putinar-Sandberg's numerical range dilation theorem, just to name a few. An application to finite-dimensional dilations of \(q\)-commuting contractions is also discussed. As a key tool in their approach, the authors establish versions of Carathéodory's and of Minkowski's theorems for matrix convex sets, which are also of independent interest.