Distinguishing graphs with zeta functions and generalized spectra

TL;DR

This paper explores the use of Ihara and Bartholdi zeta functions to distinguish graphs, showing they can differentiate many cospectral graphs and proposing they may determine almost all graphs not distinguished by traditional spectra.

Contribution

It introduces zeta functions as tools for graph identification, demonstrating their effectiveness and proposing they could resolve the graph isomorphism problem for broader classes of graphs.

Findings

Zeta functions distinguish large classes of cospectral graphs.

Enumeration of graphs distinguished by zeta functions up to 11 vertices.

Disproof of a conjecture relating Ihara zeta functions to degree sequences.

Abstract

Conjecturally, almost all graphs are determined by their spectra. This problem has also been studied for variants such as the spectra of the Laplacian and signless Laplacian. Here we consider the problem of determining graphs with Ihara and Bartholdi zeta functions, which are also computable in polynomial time. These zeta functions are geometrically motivated, but can be viewed as certain generalizations of characteristic polynomials. After discussing some graph properties determined by zeta functions, we show that large classes of cospectral graphs can be distinguished with zeta functions and enumerate graphs distinguished by zeta functions on $\le 11$ vertices. This leads us to conjecture that almost all graphs which are not determined by their spectrum are determined by zeta functions. Along the way, we make some observations about the usual types of spectra and disprove a conjecture of Setyadi and Storm about Ihara zeta functions determining degree sequences.