Spectral properties and asymptotic periodicity of flows in networks (Q1769038)

The paper under review is a very valuable contribution to the modern theory about the fascinating interplay between discrete mathematics and functional analysis. In fact, networks arise in a rich variety of applications given in science, technology, social sciences and even the Internet. This paper pays attention to the aspect of processes, i.e., on dynamical graphs or flows in networks. These flows are represented by partial differential equations under initial conditions (IC) and boundary conditions (BC). After introducing a weighted outgoing incidence matrix, (BC) naturally implies the famous Kirchhoff law. The modeling by matrices like these ones and adjacency matrices of the graph and by line graph is carefully done. Then, the authors study the well-posedness of the problem, i.e., of those (flow) equations together with (IC) and (BC), whereby they establish an operator model in Lebesgue and Sobolev space, and arrive at an equivalent Cauchy problem. The operator implied here is dispersive on a Banach lattice, and it generates a positive bounded semigroup. This one's contractivity, given in the case of an identity of constants, just guarantees well-posedness. For obtaining qualitative solutions of the problem, then, the (Perron-Frobenius) spectral properties of the operator are carefully investigated. All this is done very carefully, e.g., in terms of a circular spectral mapping theorem, of periodicity and almost-periodicity. In the following, these preparations serve for proving asymptotic behaviour, e.g., uniform exponential stability of the semigroup. If the graph is strongly connected, then the semigroup is shown to be irreducible. Among a variety of results, the authors show that under a linear independence assumption of constants over the field of rationals, the semigroup asymptotically behaves as a direct sum of rotation groups on disjoint polygons. With all these deep results, never the interpretation is forgotten as to the underlying flow problem in a network with its (invariant) strongly connected components.