Open Packing in Graphs: Bounds and Complexity
M. A. Shalu, V. K. Kirubakaran
TL;DR
The paper conducts a comprehensive complexity analysis of OPEN PACKING and its dual ρ^o(G) across H-free graphs and split-graph subclasses. It delivers a sharp H-free dichotomy: Open Packing is polynomial-time solvable on H-free graphs if H ∈ {pK_1,(K_2∪(p-2)K_1),(P_3∪(p-3)K_1),(P_4∪(p-4)K_1)}; otherwise NPC, with a separate NPC result for K_{1,3}-free graphs and a polynomial tractability for (P_4∪rK_1)-free graphs for all r≥1. The work also establishes tight structural bounds such as ρ^o(G) ≤ 2r+1 and γ_t(G) ≤ 2r+2 for connected (P_4∪rK_1)-free graphs, and presents similar dichotomies in split graphs and I_r-split graphs, including NP-completeness results for certain r and polynomial-time solvability for small r. Collectively, the results illuminate the relationship between Open Packing and Total Domination (γ_t) under restricted graph classes, reveal parameterized hardness and approximation limits (e.g., W[1]-hardness on K_{1,3}-free graphs), and provide polynomial-time algorithms in key subclasses. These findings contribute precise boundaries for when efficient computation is possible and when intractability persists in graph packing/dominating problems.
Abstract
Given a graph $G(V,E)$, a vertex subset $S$ of $G$ is called an open packing in $G$ if no pair of distinct vertices in $S$ have a common neighbour in $G$. The size of a largest open packing in $G$ is called the open packing number, $ρ^o(G)$, of $G$. It would be interesting to note that the open packing number is a lower bound for the total domination number in graphs with no isolated vertices [Henning and Slater, 1999]. Given a graph $G$ and a positive integer $k$, the decision problem OPEN PACKING tests whether $G$ has an open packing of size at least $k$. The optimization problem MAX-OPEN PACKING takes a graph $G$ as input and finds the open packing number of $G$. It is known that OPEN PACKING is NP-complete on split graphs (i.e., $\{2K_2,C_4,C_5\}$-free graphs) [Ramos et al., 2014]. In this work, we complete the study on the complexity (P vs NPC) of OPEN PACKING on $H$-free graphs for every graph $H$ with at least three vertices by proving that OPEN PACKING is (i) NP-complete on $K_{1,3}$-free graphs and (ii) polynomial time solvable on $(P_4\cup rK_1)$-free graphs for every $r\geq 1$. In the course of proving (ii), we show that for every $t\in {2,3,4}$ and $r\geq 1$, if G is a $(P_t\cup rK_1)$-free graph, then $ρ^o(G)$ is bounded above by a linear function of $r$. Moreover, we show that OPEN PACKING parameterized by solution size is W[1]-complete on $K_{1,3}$-free graphs and MAX-OPEN PACKING is hard to approximate within a factor of $n^{(\frac{1}{2}-δ)}$ for any $δ>0$ on $K_{1,3}$-free graphs unless P=NP. Further, we prove that OPEN PACKING is (a) NP-complete on $K_{1,4}$-free split graphs and (b) polynomial time solvable on $K_{1,3}$-free split graphs. We prove a similar dichotomy result on split graphs with degree restrictions on the vertices in the independent set of the clique-independent set partition of the split graphs.
