On the spectra of certain matrices and the iteration of quadratic maps (Q2353268)

Let \({A_k}(c)\) be a sequence of matrices depending on a parameter \(c \in {\mathbb C}\), with \(k \in {\mathbb N}\), which is generated recursively according to \[ {A_{k + 1}}(c) = \left[ {\begin{matrix} {{A_k}(c)} & {{U_k}} \\ {c{U_k}} & {{A_k}(c)} \\ \end{matrix} } \right],\text{ with }{A_1} = \left[ {\begin{matrix} 0 & 1 \\ c & 0 \\ \end{matrix} } \right]\text{ and }{U_k} = \left[ {\begin{matrix} 0 & 0 & \ldots & 0 \\ \vdots & \vdots & {} & \vdots \\ 0 & 0 & \ldots & 0 \\ 1 & 0 & \ldots & 0 \\ \end{matrix} } \right]. \] Consequently, the characteristic polynomial is \( - f_c^k(x)\), where \(f_c^k(x) = {f_c} \circ {f_c} \circ \dots \circ {f_c}\), and \({f_c}(x) = c - {x^2}\), and the spectrum of each \({A_k}(c)\) corresponds to the pre-images of 0 under iteration of \({f_c}(x)\). Obtaining the block structure and a recursion for the eigenvectors of \({A_k}(c)\), \(k = 0,1,\dots\), it follows that the structure of the eigenspaces of \({A_k}(c)\) changes dramatically with the parameter \(c\), in particular their dimension as real linear spaces. To study this problem for \(c \in {\mathbb R}\), the symbolic dynamics for unimodal maps of an interval is extensively used in this paper. Otherwise, this sequence of matrices models a growing system where at each time step some parts are added. The system duplicates itself and the two copies maintain interaction through two specific parts, which is modeled by the matrix \({U_k}\). Further it is shown that the logarithm of the growth rate of the real eigenspace dimensions is equal to the topological entropy of \({f_c}(x)\). For \(c \in {\mathbb C}\), the relation between the sequence \({A_k}(c)\) and the iteration of \({f_c}(x)\) yields an interesting interpretation of the spectrum of \({A_k}(c)\) as the Julia set of \({f_c}(x)\).