Invariant subspaces of finite codimension in Banach spaces of analytic functions (Q710901)

The paper under review gives a characterization of invariant subspaces of finite codimension in Banach spaces of vector-valued analytic functions in several variables. Let \(X\) denote a given Banach space and \(\Omega\subset\mathbb{C}^n\) a domain. A function \(f:\Omega\to X\) is called analytic in \(\Omega\) if for each \(\lambda\in\Omega\) it is possible to write \(f\) in a power series expansion, namely, \[ f(\xi)=\sum_{j\in\mathbb{N}^n}\alpha(j)(\xi-\lambda)^j \] with \(\alpha(j)\in X\) and the series being absolutely convergent in some neighborhood of \(\lambda\). Let \(\text{Pol}(\Omega,X)\) denote the set of Banach space valued polynomials on \(\Omega\). Given a point \(\lambda\in\Omega\), we set \[ P_\lambda^0(f)=f(\lambda), \] pointwise evaluation of the function \(f\). For a number \(t\in\mathbb{N}\), let \(X^{\{0,\dots,t\}^n}\) denote the linear space of functions \(\alpha:\{0,\dots,t\}^n\to X\). For fixed \(\lambda\in\Omega\) and \(f\) viewed as its series expansion, we define the operator \(P^t_\lambda(f)=\alpha|_{\{0,\dots,t\}^n}\). Namely, the operator \(P^t_\lambda\) applied to \(f\) yields the coefficients of its power series expansion restricted to the set \(\{0,\dots,t\}^n\). For every \(\lambda\in\mathbb{C}^n\) and an \(n\)-tuple of functions \((f_1,\dots, f_n)\), define \[ \Theta_\lambda(f_1,\dots, f_n)=\sum_{j=1}^n(z_j-\lambda_j)f_j. \] By the choice of spaces \(\mathcal H\) the author is interested in, it will be the case that \(\Theta_\lambda:\mathcal{H}^n\to\mathcal{H}\). The author considers Banach spaces \(\mathcal{H}\) of vector-valued analytic functions on a domain \(\Omega\subset\mathbb{C}^n\) that satisfy the following conditions: (A1) \(\text{Pol}\cdot\mathcal{H}\subset\mathcal{H}\); (A2) \(P_\lambda^0\) is bounded and surjective for all \(\lambda\in\Omega\); (A3) \(\mathcal{H}= \text{closure}\left(\text{Ran} \Theta_\lambda\right)\) for all \(\lambda\in\mathbb{C}^n\setminus\Omega\); (A4) \(\text{ker}P_\lambda^0=\text{closure} (\text{Ran}\Theta_\lambda)\) for all \(\lambda\in\Omega\). The author shows under the conditions of the axioms (A1)--(A4) that any invariant subspace of finite codimension of \(\mathcal{H}\) has an especially nice form. To state the main result, we introduce slightly more notation. For each \(i\in\{1,\dots, n\}\) and any \(t\in\mathbb{N}\), let \(S_i:X^{\{0,\dots,t\}^n}\to X^{\{0,\dots,t\}^n}\) denote the \(i\)th shift, namely, the operator given by \[ S_i(\alpha)(j_1,\dots,j_n)= \begin{cases}0: & j_i=0, \\ \alpha(j_1,\dots, j_{i-1}, j_i-1,j_{i+1},\dots, j_n) : & 0<j_i\leq t. \end{cases} \] A simple computation shows that \(P_\lambda^t((z_i-\lambda_i)f)=S_iP^t_\lambda(f)\). Suppose that the space of functions \(\mathcal{H}\) satisfies (A1) and (A2), \(t,K\in\mathbb{N}\) and \(\lambda_1,\dots,\lambda_K\in\Omega\). For each \(1\leq k\leq K\), let \(\mathcal{N}_k\) be the \(S_i\)-invariant subspace of \(X^{\{0,\dots,t\}^n}\) with finite codimension, where \(i=1,\dots,n\). Set \[ \mathcal{M}=\{f\in\mathcal{H}: P_{\lambda_k}^t\in\mathcal{N}_k, k=1,\dots, K\}. \tag{1} \] Then \(\mathcal{M}\) is an invariant subspace with finite codimension. It turns out the structure of this space is what characterizes the invariant subspaces of finite codimension. This is the content of the main result of the paper. Theorem: Let \(\mathcal{H}\) satisfy (A1) and (A2). Then the subspace given by (1) has codimension \[ \operatorname{codim}\mathcal{M}=\sum_{k=1}^K \operatorname{codim}\mathcal{N}_k. \] Moreover, if the space \(\mathcal{H}\) satisfies (A3) and (A4), then any closed invariant subspace with finite codimension has the form given by (1). The author illustrates his result with examples in the Bergman spaces on domains \(\Omega\) in \(\mathbb{C}\) and \(\mathbb{C}^n\), the Besov-Sobolev spaces on the unit disc. Motivations for the axioms (A1)--(A4) are provided as well. The author also points out that conditions (A3) and (A4) are connected to questions about the Taylor spectrum of \((M_{z_1},\dots, M_{z_n})\) and the set equality between \(\text{Ran}\Theta_\lambda=\text{Ker}P_\lambda^0\).