An all-unbounded-operator version of the Fuglede-Putnam theorem (Q371750)

The classical Fuglede-Putnam theorem is the assertion that if \(A\) is a bounded operator and \(M,N\) are two normal operators such that \(AN \subset MA\), then we have \(AN^\ast \subset M^\ast A.\) \textit{B. Fuglede} [Proc. Natl. Acad. Sci. USA 36, 35--40 (1950; Zbl 0035.35804)] proved this result in the special case that \(M=N\) and \textit{C. R. Putnam} [Am. J. Math. 73, 357--362 (1951; Zbl 0042.34501)] proved the theorem as it stands. \textit{M. Rosenblum} [J. Lond. Math. Soc. 33, 376--377 (1958; Zbl 0081.11902)] gave a short proof of the Fuglede-Putnam theorem in two steps. First, he showed that as a consequence of Liouville's theorem, the result holds when both \(M\) and \(N\) are bounded. Next, he approximated the operators \(M\) and \(N\) by two bounded normal operators and he approximated \(A\) by a bounded operator that intertwines them, and he obtained the general result from the bounded case. The main result in the paper under review is the following generalization of the Fuglede-Putnam theorem. Let \(A\) be a closed operator, and let \(M\) and \(N\) be two normal operators such that \(D(N) \subset D(AN) \subset D(A)\). If \(AN \subset MA\) then \(AN^\ast \subset M^\ast A\). The proof is an adaptation of Rosemblum's argument trying to get back to the all-bounded-operator version of the result.