Bures contractive channels on operator algebras (Q1747834)

The authors have extended a number of central concepts of quantum information, computing and optics theory from their usual formulation for systems with a finite number of degrees of freedom to the setting of \(C^*\)-algebras \(\mathbf{A}\) (always unital and equipped with a faithful finite trace \(\tau\)) and von Neumann algebras \(\mathbf{N}\) (always finite and possessing a faithful normal trace also denoted \(\tau\)); see also [\textit{D. Farenick} et al., J. Math. Phys. 57, No.~10, 102202, 15 p. (2016; Zbl 1349.81058)]. The treatment in the present paper is by functional analysis, but the above physics theories determine the choices of topics. In the usual terminology, a positive operator \(\rho\in \mathbf{A}\) satisfying \(\tau(\rho)=1\) is referred to as a density matrix. The set of density matrices is denoted by \(\mathcal{D}_\tau(\mathbf{A})\). This space has a central role in the paper. At this point, the authors introduce the notion of a channel. For \(C^*\)-algebras, a positive linear map \(\mathcal{E}\) on \(\mathbf{A}\) is a channel if it is trace preserving, \(\tau\circ\mathcal{E}=\tau\). Special channels are those which are \(k\)-positive, completely positive, or a Schwarz map, meaning that \(\mathcal{E}(a)^* \mathcal{E}(a) \leq \mathcal{E}(a^*a)\), recognizable as a version of the Bunyakovskii-Cauchy-Schwarz inequality. For von Neumann algebras \(\mathbf{N}\), a channel is a \(C^*\)-algebra channel which, in addition, is normal. It is noted that the bidual \(\mathbf{A}^{**}\) of \(\mathbf{A}\) is a von Neumann algebra with trace \(\tau^{**}\) and \(\mathcal{E}^{**}\) is a channel on this algebra for every channel \(\mathcal E\) on \(\mathbf{A}\). We now come to one of the main results of the paper, that the Bures distance function for \(\mathbf{A}\), patterned after the Bures metric on the space of \(n\times n\) matrices, is a bounded metric on \(\mathcal{D}_\tau\). The metric in question is given by \[ d_B^\tau : \mathcal{D}_\tau(\mathbf{A}) \times \mathcal{D}_\tau(\mathbf{A}) \to [0,1], \qquad d_B^\tau(\sigma,\rho)=\sqrt{1- F_\tau(\sigma,\rho)}, \] where \(F: \mathcal{D}_\tau(\mathbf{A}) \times \mathcal{D}_\tau(\mathbf{A})\to [0,1]\), \(F_\tau(\sigma,\rho) =\tau (| \sigma^{1/2} \rho^{1/2} |)\), is the appropriate fidelity function (Uhlmann's transition probability) for \(\mathbf{A}\). The proof includes first proving the result for von Neumann algebras and then extending it. The authors then go on to consider the properties of channels accordingly. Given a channel \(\mathcal E\), \(\rho\mapsto f_{\mathcal{E}}(\rho)\) is a continuous non-expansive affine function on the metric space. If \(f_{\mathcal{E}}\) is locally contractive, the authors define it to be a Bures contraction. They give as an example of a completely positive Bures contraction the completely depolarising channel \(\Omega(a)=(\tau(a)/\tau(1))1\) and go on to carefully analyse the mathematical theory, structural and particular, for channels in the sense they have defined. This work is interesting and offers a number of useful results. For example, the set of fixed points of a Bures contractive Schwarz channel is \(\mathbb{C} 1\). Another is the Section (\S 4) on the Frobenius theory of irreducible channels; for example, a Bures contractive Schwarz channel on \(\mathbf{A}\) is irreducible. They go on to develop a Perron-Frobenius spectral theory for channels, cf. Proposition 4.12 in the paper. The final result in the paper, Proposition (5.1), asserts that Bures constructive channels are not correctible on any set of codes of cardinality larger than \(1\). The possibility of applying the results to other physics systems suggests itself. For example, specializing \(\mathbf{A}\) to local or quasilocal algebras would be of interest in mathematical physics as would the inclusion of symmetries. Another class of algebras of interest would be those carrying representations of the canonical anti/commutation relations; this would allow application to statistical mechanics. Reviewer's remarks. My recommendation is that this paper should be read by mathematical physicists, particularly those working in the topics mentioned at the beginning of the review. Finally, I wish to note with approval that the paper was published in an electronically published, peer reviewed open access journal.