Local commutants and ultrainvariant subspaces (Q2235829)

If \(A\) is a bounded linear operator on a complex Banach space \(\mathcal{X}\) and \(\mathcal{M}\subseteq\mathcal{X}\) is a closed subspace, then the \textit{local commutant} of \(A\) at \(\mathcal{M}\) (denoted by \(C(A;\mathcal{M})\)) is defined as the set of all bounded operators \(T\) on \(\mathcal{X}\) for which \(ATx = TAx\) for all \(x\in\mathcal{M}\). This notion was introduced by Larson in the late nineties. It is immediate that \(C(A;\mathcal{M})\) is a linear subspace but it is not an algebra in general. The author obtains several necessary and sufficient conditions for \(C(A;\mathcal{M})\) to be an algebra. A closed subspace \(\mathcal{M}\subseteq \mathcal{X}\) is \textit{ultrainvariant} for \(A\) if and only if \(\mathcal{M}\) is invariant for all operators in \(C(A;\mathcal{M})\). It is clear that any ultrainvariant subspace for \(A\) is also hyperinvariant. The author shows that non-trivial ultrainvariant subspaces exist for several types of operators. While the closure of the linear span of any family of ultrainvariant subspaces can be proved to be ultrainvariant, an example is given to show that the intersection of two ultrainvariant subspaces might not be ultrainvariant. For algebraic operators, a complete description of all ultrainvariant subspaces is obtained, which, as a result, shows that hyperinvariant subspaces might not be ultrainvariant in general. On the other hand, it is shown that for normal operators on a separable Hilbert space, ultrainvariant subspaces are exactly the same as hyperinvariant subspaces.