The complexity of convexity number and percolation time in the cycle convexity
Carlos V. G. C. Lima, Thiago Marcilon, Pedro Paulo de Medeiros
TL;DR
This work studies cycle convexity in graphs, detailing how the interval function $I(S) = S \cup \{u \mid G[S \cup {u}]$ has a cycle containing $u$\}$ governs the hull operator and associated parameters $con(G)$ and $pn(G)$. It establishes that deciding $con(G) \ge k$ is NP-complete and [1]-hard parameterized by $k$ on thick spiders, yet solvable in polynomial time for extended $P_4$-laden graphs; for percolation time, it shows NP-completeness for fixed $k \ge 9$ while giving polynomial-time algorithms for fixed $k \le 2$ and for cactus graphs. The authors provide a linear-time algorithm for cacti via a cycle-tree construction and a cubic-time method for fixed $k \le 2$, plus a perpetuation gadget-driven reduction from $3$-SAT to prove NP-hardness for $k \ge 9$, with extensions to larger fixed $k$. Overall, the paper delineates clear computational boundaries across graph classes and demonstrates decomposition-based approaches to compute $con(G)$ and $pn(G)$ in structured families such as extended $P_4$-laden graphs and cacti.
Abstract
The subject of graph convexity is well explored in the literature, the so-called interval convexities above all. In this work, we explore the cycle convexity, an interval convexity whose interval function is $I(S) = S \cup \{u \mid G[S \cup \{u\}]$ has a cycle containing $u\}$. In this convexity, we prove that determine whether the convexity number of a graph $G$ is at least $k$ is \NP-complete and \W[1]-hard when parameterized by the size of the solution when $G$ is a thick spider, but polynomial when $G$ is an extended $P_4$-laden graph. We also prove that determining whether the percolation time of a graph is at least $k$ is \NP-complete even for fixed $k \geq 9$, but polynomial for cacti or for fixed $k\leq2$.
