Table of Contents
Fetching ...

Catching Rats in $H$-minor-free Graphs

Maximilian Gorsky, Giannos Stamoulis, Dimitrios M. Thilikos, Sebastian Wiederrecht

TL;DR

This work addresses the structure of $H$-minor-free graphs when a $(k\times k)$-grid minor is also excluded, establishing a polynomial-in-$t$ and linear-in-$k$ bound for both treewidth and branchwidth, with a leading term depending on the Euler genus $g_H$ of the excluded graph: $\mathsf{tw}(G),\mathsf{bw}(G)=\mathcal{O}(g_H\cdot\mathsf{bg}(G)+t^{2304})$. The authors develop a variant of the Graph Minor Structure Theorem that preserves minor relations on the torsos, integrate sphere-cut decompositions and near-embeddings, and provide algorithmic, $2^{\mathrm{poly}(t)}\cdot\mathrm{poly}(n)$-time methods to either certify grid minors or construct branch-decompositions with width bounded by the stated function. These results yield polynomial-time, $\mathcal{O}(g_H)$-type approximations for branchwidth and related width parameters in $H$-minor-free graphs, and they yield concrete consequences for bidimensional problems and subexponential algorithms on these graph classes. Overall, the paper advances both the structural and algorithmic understanding of minor-closed graph classes by tying width parameters to topological complexity via Euler genus, with practical consequences for approximation and grid-minor certifications.

Abstract

We show that every $H$-minor-free graph that also excludes a $(k \times k)$-grid as a minor has treewidth/branchwidth bounded from above by a function $f(t,k)$ that is linear in $k$ and polynomial in $t := |V(H)|$. Such a result was proven originally by [Demaine & Hajiaghayi, Combinatorica, 2008], where $f$ was indeed linear in $k$. However the dependency in $t$ in this result was non-explicit (and huge). Later, [Kawarabayashi & Kobayashi, JCTB, 2020] showed that this bound can be estimated to be $f(t,k)\in 2^{\mathcal{O}(t\log t)} \cdot k$. Wood recently asked whether $f$ can be pushed further to be polynomial, while maintaining the linearity on $k$. We answer this in a particularly strong sense, by showing that the treewidth/branchwidth of $G$ is in $\mathcal{O}(gk + t^{2304}),$ where $g$ is the Euler genus of $H$. This directly yields $f(t,k)= \mathcal{O}(t^2k + t^{2304})$. Our methods build on techniques for branchwidth and on new bounds and insights for the Graph Minor Structure Theorem (GMST) due to [Gorsky, Seweryn & Wiederrecht, 2025, arXiv:2504.02532]. In particular, we prove a variant of the GMST that ensures some helpful properties for the minor relation. We further employ our methods to provide approximation algorithms for the treewidth/branchwidth of $H$-minor-free graphs. In particular, for every $\varepsilon > 0$ and every $t$-vertex graph $H$ with Euler genus $g$, we give a $(g + \varepsilon)$-approximation algorithm for the branchwidth of $H$-minor-free graphs running in $2^{\mathsf{poly}(t) / \varepsilon} \cdot \mathsf{poly}(n)$-time. Our algorithms explicitly return either an appropriate branch-decomposition or a grid-minor certifying a negative answer.

Catching Rats in $H$-minor-free Graphs

TL;DR

This work addresses the structure of -minor-free graphs when a -grid minor is also excluded, establishing a polynomial-in- and linear-in- bound for both treewidth and branchwidth, with a leading term depending on the Euler genus of the excluded graph: . The authors develop a variant of the Graph Minor Structure Theorem that preserves minor relations on the torsos, integrate sphere-cut decompositions and near-embeddings, and provide algorithmic, -time methods to either certify grid minors or construct branch-decompositions with width bounded by the stated function. These results yield polynomial-time, -type approximations for branchwidth and related width parameters in -minor-free graphs, and they yield concrete consequences for bidimensional problems and subexponential algorithms on these graph classes. Overall, the paper advances both the structural and algorithmic understanding of minor-closed graph classes by tying width parameters to topological complexity via Euler genus, with practical consequences for approximation and grid-minor certifications.

Abstract

We show that every -minor-free graph that also excludes a -grid as a minor has treewidth/branchwidth bounded from above by a function that is linear in and polynomial in . Such a result was proven originally by [Demaine & Hajiaghayi, Combinatorica, 2008], where was indeed linear in . However the dependency in in this result was non-explicit (and huge). Later, [Kawarabayashi & Kobayashi, JCTB, 2020] showed that this bound can be estimated to be . Wood recently asked whether can be pushed further to be polynomial, while maintaining the linearity on . We answer this in a particularly strong sense, by showing that the treewidth/branchwidth of is in where is the Euler genus of . This directly yields . Our methods build on techniques for branchwidth and on new bounds and insights for the Graph Minor Structure Theorem (GMST) due to [Gorsky, Seweryn & Wiederrecht, 2025, arXiv:2504.02532]. In particular, we prove a variant of the GMST that ensures some helpful properties for the minor relation. We further employ our methods to provide approximation algorithms for the treewidth/branchwidth of -minor-free graphs. In particular, for every and every -vertex graph with Euler genus , we give a -approximation algorithm for the branchwidth of -minor-free graphs running in -time. Our algorithms explicitly return either an appropriate branch-decomposition or a grid-minor certifying a negative answer.

Paper Structure

This paper contains 46 sections, 48 theorems, 25 equations.

Key Result

Theorem 1.1

For every $t$-vertex graph $H$, every $H$-minor-free graph $G$ that excludes a $(k\times k)$-grid as a minor, has treewidth/branchwidth in $\mathcal{O}(t^2k + t^{2304})$.

Theorems & Definitions (57)

  • Theorem 1.1
  • Theorem 1.2
  • Theorem 1.3
  • Theorem 1.4
  • Theorem 1.5
  • Proposition 2.1: Cabello, Colin de Verdière, and Lazarus CabelloCL2012Algorithms
  • Proposition 2.2: Kawarabayashi, Mohar and Reed KawarabayashiMR2008Simpler
  • Lemma 2.3
  • proof
  • Proposition 2.4: Bodlaender and Thilikos BodlaenderT1997Constructive
  • ...and 47 more