Connected greedy coloring $H$-free graphs

TL;DR

This paper investigates the connected greedy coloring problem in $H$-free graphs, establishing complexity dichotomies and conditions under which the connected chromatic number equals the chromatic number.

Contribution

It provides a complexity classification for deciding whether the connected chromatic number equals the chromatic number in $H$-free graphs, extending the understanding of connected greedy coloring.

Findings

Problem (2) is polynomial or NP-complete depending on $H$.

When $G$ is an induced subgraph of $P_5$ or $P_4+K_1$, $ ext{chi}_c(G)= ext{chi}(G)$.

Deciding $ ext{chi}_c(G)= ext{chi}(G)$ is NP-hard for certain $H$.

Abstract

A connected ordering $(v_1, v_2, \ldots, v_n)$ of $V(G)$ is an ordering of the vertices such that $v_i$ has at least one neighbour in $\{v_1, \ldots, v_{i - 1}\}$ for every $i \in \{2, \ldots, n\}$. A connected greedy coloring (CGC for short) is a coloring obtained by applying the greedy algorithm to a connected ordering. This has been first introduced in 1989 by Hertz and de Werra, but still very little is known about this problem. An interesting aspect is that, contrary to the traditional greedy coloring, it is not always true that a graph has a connected ordering that produces an optimal coloring; this motivates the definition of the connected chromatic number of $G$, which is the smallest value $\chi_c(G)$ such that there exists a CGC of $G$ with $\chi_c(G)$ colors. An even more interesting fact is that $\chi_c(G) \le \chi(G)+1$ for every graph $G$ (Benevides et. al. 2014). In this paper, in the light of the dichotomy for the coloring problem restricted to $H$-free graphs given by Kr\'al et.al. in 2001, we are interested in investigating the problems of, given an $H$-free graph $G$: (1). deciding whether $\chi_c(G)=\chi(G)$; and (2). given also a positive integer $k$, deciding whether $\chi_c(G)\le k$. We have proved that Problem (2) has the same dichotomy as the coloring problem (i.e., it is polynomial when $H$ is an induced subgraph of $P_4$ or of $P_3+K_1$, and it is NP-complete otherwise). As for Problem (1), we have proved that $\chi_c(G) = \chi(G)$ always hold when $G$ is an induced subgraph of $P_5$ or of $P_4+K_1$, and that it is NP-hard to decide whether $\chi_c(G)=\chi(G)$ when $H$ is not a linear forest or contains an induced $P_9$. We mention that some of the results actually involve fixed $k$ and fixed $\chi(G)$.