Linear graph transformations on spaces of analytic functions (Q2344319)

Let \({\mathcal H}\) be a complex Hilbert space and \({\mathcal B}({\mathcal H})\) be the Banach algebra of all bounded linear operators on \({\mathcal H}\). A subalgebra \({\mathcal A} \subseteq {\mathcal B}({\mathcal H})\) is said to be transitive if it contains the identity operator \(I\) and has no nontrivial common invariant subspaces. It is a longstanding open question posed by Kadison whether every transitive algebra is dense in \({\mathcal B}({\mathcal H})\) in the strong operator topology. If that were the case, then it would follow that every operator \(T\in {\mathcal B}({\mathcal H})\) which is not a scalar multiple of \(I\) has a nontrivial hyperinvariant subspace. An operator \(A\) (respectively, an algebra \({\mathcal A}\)) has the transitive algebra property if every transitive algebra which contains \(A\) (respectively, \({\mathcal A}\)) is strongly dense in \({\mathcal B}({\mathcal H})\). \textit{W. B. Arveson} [Duke Math. J. 34, 635--647 (1967; Zbl 0183.42403)] showed that every m.a.s.a., as well as the unilateral shift, have this property. The paper under review was motivated by the question what multiplier algebras have the transitive algebra property. Recall that the multiplier algebra of a Hilbert space \({\mathcal H} \subseteq \text{Hol}(\Omega)\), where \(\Omega \subseteq {\mathbb C}^d\) is a nonempty open and connected set, is the algebra \({\mathcal M}({\mathcal H})\) of all bounded multiplication operators \(M_\varphi f=\varphi f\) \((f \in {\mathcal H})\) induced by \(\varphi \in \text{Hol}(\Omega)\). Recently, \textit{G.-Z. Cheng} et al. [J. Funct. Anal. 258, No. 12, 4229--4250 (2010; Zbl 1190.47087)] showed that \({\mathcal M}({\mathcal H})\) has the transitive algebra property whenever \({\mathcal H}\) has a complete Nevanlinna-Pick kernel, i.e., a reproducing kernel \(k_{\lambda}(z)\) for \({\mathcal H}\) which is of the form \(k_{\lambda}(z) =\frac{\overline{f(\lambda)}f(z)}{1-u_\lambda(z)}\), where \(f\) is an analytic function and \(u_\lambda(z)\) is positive definite and sesquianalytic. Unfortunately, in the present paper, the authors have not obtained any specific answer to the question which has motivated the research. However, the investigations lead to some interesting results related to the invariant subspace structure of \({\mathcal M}({\mathcal H})\). Let us mention the first main theorem of the paper. We need the notion of the invariant graph subspace. Let \({\mathcal A} \subseteq {\mathcal B}({\mathcal H})\) be an algebra and \(n\geq 2\) be an integer. A closed subspace \({\mathcal M} \subseteq {\mathcal H}^{(n)}\) is an invariant graph subspace for \({\mathcal A}\) if there exist a linear manifold \({\mathcal D} \subseteq {\mathcal H}\) and linear transformations \(T_j: {\mathcal D} \to {\mathcal H}\) \((j=1, \dots, n-1)\) such that \({\mathcal M}=\{(x,T_1 x, \dots, T_{n-1}x);\, x\in {\mathcal D}\}\) and \((Ax_1, \dots, Ax_n)\in {\mathcal M}\) for every \((x_1, \dots, x_n)\in {\mathcal M}\) and every \(A\in {\mathcal A}\). It is observed by the authors that it follows from Arveson's lemma that \({\mathcal M}({\mathcal H})\) has the transitive algebra property if and only if the following condition is satisfied: whenever \(n\geq 2\) and \({\mathcal M}=\{ (f, T_1 f, \dots, T_{n-1}f): f\in {\mathcal D}\} \subseteq {\mathcal H}^{(n)}\) is an invariant graph subspace of \({\mathcal M}({\mathcal H})\) such that, for each nonzero \(\alpha=(\alpha_0, \dots, \alpha_{n-1}) \in {\mathbb C}^n\), the linear transformation \(L_{\alpha}:{\mathcal D} \to {\mathcal H}\) which is given by \(L_{\alpha}=\overline{\alpha}_0 I+\sum_{i=1}^{n-1}\overline{\alpha}_i T_i\) is one-to-one and has dense range, then \({\mathcal A}_{{\mathcal M}}=\{ A\in {\mathcal B}({\mathcal H}): A {\mathcal D} \subseteq {\mathcal D},\, AT_i f=T_i Af\, \text{for all }f\in {\mathcal D}\}\) has a nontrivial invariant subspace. There are a few more interesting results and examples in the paper.