Quantcast

Strengthening some complexity results on toughness of graphs

Research paper by Gyula Y Katona, Kitti Varga

Indexed on: 22 Oct '19Published on: 19 Oct '19Published in: arXiv - Computer Science - Discrete Mathematics



Abstract

Let $t$ be a positive real number. A graph is called $t$-tough if the removal of any vertex set $S$ that disconnects the graph leaves at most $|S|/t$ components. The toughness of a graph is the largest $t$ for which the graph is $t$-tough. The main results of this paper are the following. For any positive rational number $t \le 1$ and for any $k \ge 2$ and $r \ge 6$ integers recognizing $t$-tough bipartite graphs is coNP-complete (the case $t=1$ was already known), and this problem remains coNP-complete for $k$-connected bipartite graphs, and so does the problem of recognizing 1-tough r-regular bipartite graphs. To prove these statements we also deal with other related complexity problems on toughness. % In this paper we prove the following. For any positive rational number $t$, deciding whether $\tau(G)=t$ is DP-complete and if $t < 1$, this problem remains DP-complete for bipartite graphs. For any integer $k \ge 2$ and positive rational number $t \le 1$, recognizing $t$-tough $k$-connected bipartite graphs is coNP-complete. For any integer $r \ge 5$, recognizing $1/2$-tough $r$-regular graphs is coNP-complete. For any integer $r \ge 6$, recognizing 1-tough $r$-regular bipartite graphs is coNP-complete. For any positive rational number $t < 2/3$ we give a polynomial time algorithm for recognizing 3-regular graphs with toughness $t$. Finally, we prove that every connected 4-regular graph is 1/2-tough.