Nonlinear completely positive maps and dilation theory for real involutive algebras (Q902274)

A real seminormed involutive algebra is a real associative algebra \(A\) endowed with an involutive antiautomorphism \(*\) and a submultiplicative seminorm \(p\) with \(p(a^*)=p(a)\). Then \(\text{ball}(A,p):=\{a\in A: p(a)<1\}\) is an involutive semigroup. For the purposes of this review, let us assume that \(A\) is unital, i.e., it has a multiplicative neutral element. Let \(V\) be a complex Hilbert space and \(B(V)\) be the corresponding \(C^*\)-algebra of bounded operators. For \(n\in {\mathbb N}\), let \(M_n(B(V))\cong B(V^n)\) be the \(C^*\)-algebra of \(n\times n\)-matrices with entries in \(B(V)\) and \(M_n(B(V))_+\) be its cone of positive operators. A function \(\phi: \text{ball}(A,p)\to B(V)\) is called completely positive if \((\phi(a_{jk}))_{j,k=1}^n\in M_n(B(V))_+\) for each \(n\in {\mathbb N}\) and \((a_{jk})_{j,k=1}^n\in M_n(B(V))_+\) such that \(a_{j,k}\in \text{ball}(A,p)\) for \(j,k\in \{1,\dots, n\}\). If \((\phi(s_j^*s_k))_{j,k=1}^n\in M_n(B(V))_+\) for all \(n\in {\mathbb N}\) and \(s_j,s_k\in \text{ball}(A,p)\), then \(\phi\) is called positive definite. Let \(\eta_A: A\to C^*(A,p)\) be the enveloping \(C^*\)-algebra of the real seminormed involutive algebra \(A\). For any \(C^*\)-algebra \(B\), let \(e^B\) be the \(c_0\)-direct sum of the \(C^*\)-algebras \(S^n(B)\subseteq B^{\otimes n}\), where the tensor products are constructed from the maximal \(C^*\)-cross norm. For unital \(A\) and a bounded function \(\phi: \text{ball}(A,p)\to B(V)\), the main result (Theorem 1.1) states that \(\phi\) is completely positive if and only if \(\phi\) is positive definite and real analytic with respect to some locally convex topology on \(A\) for which \(p\) is continuous, if and only if there exists a linear completely positive map \(\Phi: e^{C^*(A,p)}\to B(V)\) with \(\Phi\circ \Gamma=\phi\), where \(\Gamma(a):=\sum_{n=0}^\infty\eta_A(a)^{\otimes n}\). A similar result is obtained if \(A\) is not unital, which however is more complicated to state.