Small Independent Sets versus Small Separator in Geometric Intersection Graphs

TL;DR

This paper explores the complexity of certain problems on geometric intersection graphs, identifying a new class of problems with subexponential algorithms and establishing structural properties and parameters that enable these algorithms.

Contribution

It introduces the concept of weak square-root phenomenon, develops algorithms and lower bounds for this class, and presents a new structural theorem and graph parameter for such problems.

Findings

Problems like 2-Subcoloring and Two Sets Cut-Uncut exhibit the weak square-root phenomenon.

Every such graph admits a sublinear separator with components of sublinear independence number.

Introduces the $eta$-modulator number, generalizing independence and vertex cover numbers.

Abstract

While most classical NP-hard graph problems cannot be solved in time $2^{o(n)}$ on general graphs under the Exponential Time Hypothesis (ETH), many exhibit the square-root phenomenon and admit optimal algorithms running in time $2^{O(\sqrt{n})}$ on certain geometric intersection graphs, such as planar graphs or unit disk graphs. In 2018, de Berg et al. developed a general algorithmic framework for such problems on intersection graphs of similarly sized fat objects in $\mathbb{R}^d$, achieving running times of the form $2^{O(n^{1-1/d})}$, along with matching lower bounds under ETH. In this paper, we identify problems that do not exhibit the square-root phenomenon, yet still admit subexponential algorithms on intersection graphs of similarly sized fat objects in $\mathbb{R}^d$, for every fixed dimension $d \geqslant 2$. We introduce the notion of a weak square-root phenomenon: problems that can be solved in time $2^{\tilde{O}(n^{1-1/(d+1)})}$, and for which matching lower bounds hold under ETH. We develop both an algorithmic framework and a corresponding lower bound framework. As concrete examples, we show that the problems 2-Subcoloring and Two Sets Cut-Uncut exhibit this behavior. Our algorithms rely on a new win-win structural theorem, which can be informally stated as follows: every such graph admits a sublinear separator whose removal leaves connected components with sublinear independence number. To facilitate the design of these algorithms, we introduce a new graph parameter, the $\alpha$-modulator number, which generalizes both the independence number and the vertex cover number.