scientific article; zbMATH DE number 1302540

commuting von Neumann algebras generalized conditional mean values Markov state Markov triple modular automorphisms noncommutative Markov chain Tomita theory

Mathematics Subject Classification ID

General theory of von Neumann algebras (46L10) Noncommutative probability and statistics (46L53)

Review text

The paper starts by introducing a situation involving two von Neumann algebras \(M_{1}\), \(M_{2} \subset B(H)\), \(H\) a Hilbert space and \(\Omega \in H\), separating for the von Neumann algebra generated by \(M_{1} \cup M_{2}\). Namely, the author supposes that \(M_{2} \subset(M_{1})'\), considers the linear map \(\lambda_{1, 2}: M_{1} \rightarrow M_{2}\) defined by \(E_{2} a_{1} \xi =\) J\(_{2} \lambda_{1, 2}(a_{1}^{+})J_{2} \xi\), where \(\xi \in H_{2}\) (the closure of \(M_{2} \Omega\)), \(E_{i}\) is the projector on \(H_{i}\), and imposes the condition \(\lambda_{1, 2}(\sigma_{1}^{t}(a_{1})) = \sigma_{2}^{t}(\lambda_{1, 2}(a_{1}))\); \(\sigma_{i}^{t}\) and \(J_{i}\) are the modular automorphisms and involutions in Tomita theory for \(M_{i}\) and \(\Omega\). Let \(\varepsilon_{i, j}\) be the generalized conditional expectations from M\(_{i}\) to M\(_{j}\) in the sense of the author and \textit{L. Accardi} [J. Funct. Anal. 45, 245-273 (1982; Zbl 0483.46043)]. This part of the paper uses concepts and results to appear in a paper of the author. Then a triple \(M_{1}, M_{2}, M_{3} \subset B(H)\) of von Neumann algebras and a \(\Omega \in H\), separating for \(M_{\{1, 2, 3 \}}\), are considered, every pair of them fulfilling the above condition, and the equivalence of 5 conditions is established: \(\lambda_{3, \{1, 2 \}}(M_{3})\subset M_{2}\), \(\lambda_{3, \{1, 2 \}} = \lambda_{3, 2}\), \(E_{\{1, 2 \}} E_{3} = E_{2} E_{3}\), all \(a \in E_{3, \{1, 2 \}}(M_{3})\) are fixed for \(\varepsilon_{\{1, 2 \}, 2}\) and \(\lambda_{\{1, 3 \}, 2}(a_1 a_3)= \lambda_{1, 2}(a_{1})\lambda_{3, 2}(a_{3})\) for \(a_{i} \in M_{i}\); \( \{1, 2 \} \) refers to the von Neumann algebra \(M_{\{1, 2 \}}\) generated by \(M_{1} \cup M_{2}\). In this situation, \(\Omega\) is said to be Markov, \(\lambda_{1, 2}(M_{1})\) commutes with \(\lambda_{3, 2}(M_{3})\) and \(\Omega\) is Markov also for \(M_{3}\),\(M_{2}\),\(M_{1}\). Also \(\sigma_{t}^{\{1, 2, 3 \}}(M_{1})\subset M_{\{1, 2 \}}\). Then, if \(\Omega\) is Markov for every \(M_{\{1,\dots, k- 2 \}}\),\(M_{k- 1}\),\(M_{k}\) it is proved that it is Markov for every \(M_{A}\),\(M_{B}\),\(M_{C}\) where \(A\),\(B\),\(C \subset \{1, 2, \dots \} \), \(\max A < \min B\), \(\max B < \min C\). In the last section the author proves the following structure theorem. Consider, in a Markov triple, the von Neumann algebras \(M_{2, 3}\), \(M_{2, 1}\), \(M_{1, 2}\), \(M_{3, 2}\), and \(N_{1}\), \(N_{2}\), \(N_{3}\), \(N\) generated by the images of \(\lambda_{3, 2}\), \(\lambda_{1, 2}\), \(\lambda_{2, 1}\), \(\lambda_{2, 3}\), and by \(M_{2,1} \cup M_{1, 2}\), \(M_{2, 1} \cup M_{2, 3}\), \(M_{2, 3} \cup M_{3, 2}\), \(M_{2, 1} \cup M_{1, 2} \cup M_{2, 3} \cup M_{3, 2} \). Then \(N_{1} \subset(N_{3})'\) and there exist projections of norm one, preserving the state defined by \(\Omega\), \(\varepsilon:M \rightarrow N\), \(\varepsilon_{1}:M_{1} \rightarrow M_{1, 2}\), \(\varepsilon_{2}:M_{2} \rightarrow N_{2}\) and \(\varepsilon_{3}:M_{3} \rightarrow N_{3, 2}\) such that \(\varepsilon(a_{1} a_{2} a_{3})=\) \(\varepsilon_{1}(a_{1})\varepsilon_{2}(a_{2})\varepsilon_{3}(a_{3})\). If one considers the \(\lambda\)'s \(\lambda_{3, 1}'\), \(\lambda_{1, 3}'\) for \(N_{1}, N_{3}\) then \(\lambda_{3, 1};(N_{3})\) and \(\lambda_{1, 3}'(N_{1})\) are contained in the centers of \(M_{2, 1}\), \(M_{2, 3} \). All these results appear as a noncommutative generalization, going farther than before, of the corresponding ones in the Kolmogorov probability theory. For the entire collection see [Zbl 0903.00097].

Recommendations

Cited in

(8)

This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4249071)