arXiv++ Combinatorics

We completely characterize the conditions under which a complex unitary number is an eigenvalue of the non-backtracking matrix of an undirected graph. Further, we provide a closed formula to compute its geometric multiplicity and describe an algorithm to compute this multiplicity without making a single matrix computation. The algorithm has time complexity that is linear in the size of the graph.

Consider a finite undirected unweighted graph G and add a new node to it arbitrarily connecting it to pre-existing nodes. We study the behavior of the Perron eigenvalue of the non-backtracking matrix of G before and after such a node addition. We prove an interlacing-type result for said eigenvalue, namely, the Perron eigenvalue never decreases after node addition. Furthermore, our methods lead to bounds for the difference between the eigenvalue before and after node addition. These are the first known bounds that have been established in full rigor. Our results depend on the assumption of diagonalizability of the non-backtracking matrix. Practical experience says that this assumption is fairly mild in many families of graphs, though necessary and sufficient conditions for it remain an open question.

Much effort has been spent on characterizing the spectrum of the non-backtracking matrix of certain classes of graphs, with special emphasis on the leading eigenvalue or the second eigenvector. Much less attention has been paid to the eigenvalues of small magnitude; here, we fully characterize the eigenvalues with magnitude equal to one. We relate the multiplicities of such eigenvalues to the existence of specific subgraphs. We formulate a conjecture on necessary and sufficient conditions for the diagonalizability of the non backtracking matrix. As an application, we establish an interlacing-type result for the Perron eigenvalue.