Table of Contents
Fetching ...

Boundedness and Separation in the Graph Covering Number Framework

Miriam Goetze, Peter Stumpf, Torsten Ueckerdt

TL;DR

The paper develops a unified framework of graph covering numbers (global, union, local, folded) $cn_x^{\mathcal{G}}(H)$ to study decompositions of host graphs $H$ into guest graphs from $\mathcal{G}$. It defines binding functions to relate larger and smaller covering numbers under structural restrictions on $\mathcal{G}$ (component-closed, hereditary, monotone, sparse, bounded chi, bounded mad) and $\mathcal{H}$ (bounded chi, mad, $M$-minor-free, bounded treewidth), proving numerous positive boundedness results and constructing several separating counterexamples. Key techniques include reductions to star forests and bipartite subgraphs, tree decompositions, and subgraph restriction lemmas for hereditary/monotone classes, yielding explicit bounds such as $cn_u \le (\mathrm{tw}(H)+1) cn_l$ for component-closed $\mathcal{G}$ with bounded treewidth hosts, and $cn_u \le 2sd$ under sparse hereditary guests with mad bound $d$, $s=cn_l$. The results provide a comprehensive landscape of when the folded, local, union, and global covering numbers are functionally related, with algorithmic implications for parameters like star arboricity and potential extensions to nowhere-dense or WL-dimension settings.

Abstract

For a graph class $\mathcal G$ and a graph $H$, the four $\mathcal G$-covering numbers of $H$, namely global ${\rm cn}_{g}^{\mathcal{G}}(H)$, union ${\rm cn}_{u}^{\mathcal{G}}(H)$, local ${\rm cn}_{l}^{\mathcal{G}}(H)$, and folded ${\rm cn}_{f}^{\mathcal{G}}(H)$, each measure in a slightly different way how well $H$ can be covered with graphs from $\mathcal G$. For every $\mathcal G$ and $H$ it holds \[ {\rm cn}_{g}^{\mathcal{G}}(H) \geq {\rm cn}_{u}^{\mathcal{G}}(H) \geq {\rm cn}_{l}^{\mathcal{G}}(H) \geq {\rm cn}_{f}^{\mathcal{G}}(H) \] and in general each inequality can be arbitrarily far apart. We investigate structural properties of graph classes $\mathcal G$ and $\mathcal H$ such that for all graphs $H \in \mathcal{H}$, a larger $\mathcal G$-covering number of $H$ can be bounded in terms of a smaller $\mathcal G$-covering number of $H$. For example, we prove that if $\mathcal G$ is hereditary and the chromatic number of graphs in $\mathcal H$ is bounded, then there exists a function $f$ (called a binding function) such that for all $H \in \mathcal{H}$ it holds ${\rm cn}_{u}^{\mathcal{G}}(H) \leq f({\rm cn}_{g}^{\mathcal{G}}(H))$. For $\mathcal G$ we consider graph classes that are component-closed, hereditary, monotone, sparse, or of bounded chromatic number. For $\mathcal H$ we consider graph classes that are sparse, $M$-minor-free, of bounded chromatic number, or of bounded treewidth. For each combination and every pair of $\mathcal G$-covering numbers, we either give a binding function $f$ or provide an example of such $\mathcal{G},\mathcal{H}$ for which no binding function exists.

Boundedness and Separation in the Graph Covering Number Framework

TL;DR

The paper develops a unified framework of graph covering numbers (global, union, local, folded) to study decompositions of host graphs into guest graphs from . It defines binding functions to relate larger and smaller covering numbers under structural restrictions on (component-closed, hereditary, monotone, sparse, bounded chi, bounded mad) and (bounded chi, mad, -minor-free, bounded treewidth), proving numerous positive boundedness results and constructing several separating counterexamples. Key techniques include reductions to star forests and bipartite subgraphs, tree decompositions, and subgraph restriction lemmas for hereditary/monotone classes, yielding explicit bounds such as for component-closed with bounded treewidth hosts, and under sparse hereditary guests with mad bound , . The results provide a comprehensive landscape of when the folded, local, union, and global covering numbers are functionally related, with algorithmic implications for parameters like star arboricity and potential extensions to nowhere-dense or WL-dimension settings.

Abstract

For a graph class and a graph , the four -covering numbers of , namely global , union , local , and folded , each measure in a slightly different way how well can be covered with graphs from . For every and it holds and in general each inequality can be arbitrarily far apart. We investigate structural properties of graph classes and such that for all graphs , a larger -covering number of can be bounded in terms of a smaller -covering number of . For example, we prove that if is hereditary and the chromatic number of graphs in is bounded, then there exists a function (called a binding function) such that for all it holds . For we consider graph classes that are component-closed, hereditary, monotone, sparse, or of bounded chromatic number. For we consider graph classes that are sparse, -minor-free, of bounded chromatic number, or of bounded treewidth. For each combination and every pair of -covering numbers, we either give a binding function or provide an example of such for which no binding function exists.

Paper Structure

This paper contains 13 sections, 26 theorems, 17 equations, 1 figure, 2 tables.

Key Result

Theorem 2

Let $\mathcal{G}$ and $\mathcal{H}$ be two graph classes. Then each of the following holds.

Figures (1)

  • Figure 1: Examples of $\mathcal{G}$-covers of $H$: For $\mathcal{G} = \{K_3\}$ and $H = K_7$, we have a $7$-global, $3$-local injective $\mathcal{G}$-cover (\ref{['fig:examples_left']}), as well as a $5$-global, $5$-local $\overline{\mathcal{G}}$-cover (\ref{['fig:examples_middle']}). In fact, it holds that $\cn{g}{\{K_3\}}{K_7} = 7$ and $\cn{u}{\{K_3\}}{K_7} = 5$ and $\cn{l}{\{K_3\}}{K_7} = 3$. (\ref{['fig:examples_right']}) A $2$-local non-injective $\mathcal{G}$-cover of $K_7$ for $\mathcal{G} = {\rm Forb}(C_4)$ being the class of $C_4$-free graphs. In fact, it holds that $\cn{f}{{\rm Forb}(C_4)}{K_7} = 2$, while $\cn{l}{{\rm Forb}(C_4)}{K_7} \geq 3$ since the local Ramsey number of $C_4$ is $6$GLST87.

Theorems & Definitions (29)

  • Definition 1
  • Theorem 2: global vs. union
  • Remark
  • Theorem 3: union vs. local
  • Theorem 4: local vs. folded
  • Theorem 4: global vs. union
  • Theorem 5
  • Theorem 6
  • Theorem 7
  • Proposition 8
  • ...and 19 more