Avoidable Vertices and Edges in Graphs

TL;DR

This paper introduces the concept of avoidable vertices and paths in graphs, proving key cases, exploring their relationships with graph triangulations, and applying these ideas to solve problems in graph theory and algorithms.

Contribution

It defines avoidable vertices and paths, proves the existence of avoidable edges, and links these concepts to minimal triangulations and polynomial-time algorithms for maximum weight clique problems.

Findings

Every graph with an edge has an avoidable edge.

Avoidable vertices relate closely to minimal triangulations.

New polynomial algorithms for maximum weight clique in certain graph classes.

Abstract

A vertex in a graph is simplicial if its neighborhood forms a clique. We consider three generalizations of the concept of simplicial vertices: avoidable vertices (also known as \textit{OCF}-vertices), simplicial paths, and their common generalization avoidable paths, introduced here. We present a general conjecture on the existence of avoidable paths. If true, the conjecture would imply a result due to Ohtsuki, Cheung, and Fujisawa from 1976 on the existence of avoidable vertices, and a result due to Chv\'atal, Sritharan, and Rusu from 2002 the existence of simplicial paths. In turn, both of these results generalize Dirac's classical result on the existence of simplicial vertices in chordal graphs. We prove that every graph with an edge has an avoidable edge, which settles the first open case of the conjecture. We point out a close relationship between avoidable vertices in a graph and its minimal triangulations, and identify new algorithmic uses of avoidable vertices, leading to new polynomially solvable cases of the maximum weight clique problem in classes of graphs simultaneously generalizing chordal graphs and circular-arc graphs. Finally, we observe that the proved cases of the conjecture have interesting consequences for highly symmetric graphs: in a vertex-transitive graph every induced two-edge path closes to an induced cycle, while in an edge-transitive graph every three-edge path closes to a cycle and every induced three-edge path closes to an induced cycle.