Symbolic dynamics from signed matrices (Q1884239)

Given a graph, we define its adjacency matrix \(A=\{a_{ij}\}\) by setting \(a_{ij}= 1\) if \(v_j\to v_i\) and \(a_{ij}=0\) otherwise. We give each vertex \(v_u\) a label \(L(v_i)\). The sofic shift associated to a labelled directed graph consists of all infinite sequences of labels \(\{L(v_{i(0)})\), \(L(v_{i(l)}),\dots\}\) such that \(v_{i(j)}\to v_{i(j+1)}\) for all \(j\geq 0\), along the shift map \(\sigma\) defined by \(\sigma ((v_{i(0)},v_{i(1)},\dots))=(v_{i(1)},v_{i(2)},\dots)\). Let \(V\) be an \(n\)-dimensional vector space with basis \(\{v_1,\dots,v_n\}\), and \(M\) an \(n\times n\)-matrix considered as representing a map on \(V\) with the given basis. The matrix \(M\) can be now interpreted as representing a directed graph with weighted edges. The vertices of the graph are the basis elements. If \(M_g\neq 0\) then there is an edge running from \(v_i\) to \(v_j\); its weight is \(M_{ij}\). The shift associated to \(M\) is that for the adjacency matrix constructed by replacing each nonzero entry of \(M\) by 1. There are considered the sofic shift got by sorting the vertices into two distinct classes. Using the method of assigning a soft shift to a matrix, the author detects positive entropy for such sofic shift by comparing the characteristic polynomial of the original matrix to those for matrices for the restrictions of the shifts to each piece (1 and 2). The main application of that is to the use of Conley index to detect symbolic dynamics in isolated invariant sets (extending a result of \textit{M. C. Carbinatto}, \textit{J. Kwapisz} and \textit{K. Mischaikow} [Ergodic Theory Dyn. Syst. 20, 365--377 (2000; Zbl 0971.37005)].