Beurling type representation for certain invariant subspaces of maximal subdiagonal algebras (Q6085413)

Let \(\mathcal{M}\) be a \(\sigma\)-finite (not necessarily semifinite) von Neumann algebra acting on a complex Hilbert space \(\mathcal{H}\). Let \(\Phi\) be a faithful normal conditional expectation from \(\mathcal{M}\) onto a von Neumann subalgebra \(\mathfrak{D}\). Let \(\mathfrak A\) be a \(\sigma\)-weakly closed subalgebra of \(\mathfrak{D}\) containing \(\mathfrak{D}\) such that \begin{itemize} \item[1.] \(\mathfrak{A}\cap \mathfrak{A}^* =\mathfrak{D}\), \item[2.] \(\Phi\) is multiplicative on \(\mathfrak{A}\), \item[3.] \(\mathfrak{A}+\mathfrak{A}^*\) is \(\sigma\)-weakly dense in \(\mathcal{M}\). \end{itemize} In this case, \(\mathfrak{A} \) is called a subdiagonal algebra of \(\mathcal{M}\) with respect to \(\Phi\), and the algebra \(\mathfrak{D}\) is called the diagonal of \(\mathfrak{A}\). We say that \(\mathfrak A\) is a maximal subdiagonal algebra in a \(\sigma\)-finite von Neumann algebra \(\mathcal{M}\) with respect to \(\Phi\) if \(\mathfrak{A} \) is not properly contained in any other subalgebra of \(\mathcal M\) which is a subdiagonal algebra of \(\mathcal M\) with respect to \(\Phi\). Let \(\{\sigma_t^\varphi:t\in \mathbb{R}\}\) be the modular automorphism group of \(\mathcal M\) associated with \(\varphi\) by Tomita-Takesaki theory. Let \(1\le p<\infty \) and let \(L^p(\mathcal{M})\) be a Haagerup \(L_p\)-space. Let \(h_0\) be the corresponding noncommutative Radon-Nikodym derivative. A closed subspace \(\mathfrak{M}_p\subset L^p(\mathcal{M})\) of \(L^p(\mathcal{M})\) is said to be a right (resp. left) invariant subspace of \(\mathcal{A}\) if \(\mathfrak{M}_p \mathfrak{A}\subset \mathfrak{M}_p\) (resp. \(\mathfrak{A} \mathfrak{M}_p \subset \mathfrak{M}_p\)). The paper contains several interesting results: (a) Let \(\mathfrak{A}\) be a maximal subdiagonal algebra with respect to \(\Phi\). \begin{itemize} \item[1.] If \(\mathfrak{M}_p\subset L^p(\mathcal{M})\) is a right invariant subspace of \(\mathfrak{A}\) which is \(\sigma^\varphi\)-invariant, then \(\mathfrak{M}_p\cap \{\mathcal{M} h_0^{1/p}\}\) is dense in \(\mathfrak{M}_p\). \item[2.] \(\Lambda_p\) is a lattice isomorphism from \(L_\infty ^\varphi\) onto \(L_p^\varphi\) for any \(1\le p<\infty\), where \(L_p^\varphi\) is the lattice of all right invariant subspaces of \(\mathfrak{A}\) in \(L^p(\mathcal{M})\) which are \(\sigma^\varphi\)-invaraint, for any \(\mathfrak{M}_p\in L_\infty ^\varphi\), \(\Lambda_p(\mathfrak{M}_\infty ) =[\mathcal{M}_\infty h_0^{1/p}]_p\) (the closed linear span in \(L^p(\mathcal{M})\)), \item[3.] there exists a unique pair of type \(1\) and type \(2\) right invariant subspace \(\mathfrak{M}^1_p\) and \(\mathfrak{M}_p^2\) such that \(\mathfrak{M}_p=\mathfrak{M}_p^1 \oplus ^{col}\mathfrak{M}_p^2\). \end{itemize} (b) Let \(\mathfrak{A}\) be a maximal subdiagonal algebra with respect to \(\Phi\). Then, for any \(1\le p<\infty\), \begin{itemize} \item[1.] we have \([(\bigcap _{n\ge 1} \mathfrak{A}_0^n) h_0^{1/p}]_p =\bigcap _{n\ge 1 }[\mathfrak{A}_0^n h_0^{1/p}]_p \), \item[2.] for any projection \(E\) in the center of \(\mathcal{M}\), \([EH^p]_p =FH^p \oplus ^{col} [(E-F) H^p]_p\) for a unique central projection \(F\) in \(\mathfrak{D}\) with \(F\le E\) such that \(FH^p\) is of type \(1\) while \([(E-F)H^p]_p\) is of type \(2\). \end{itemize}