Table of Contents
Fetching ...

Fractional list packing for layered graphs

Stijn Cambie, Wouter Cames van Batenburg

Abstract

The fractional list packing number $χ_{\ell}^{\bullet}(G)$ of a graph $G$ is a graph invariant that has recently arisen from the study of disjoint list-colourings. It measures how large the lists of a list-assignment $L:V(G)\rightarrow 2^{\mathbb{N}}$ need to be to ensure the existence of a `perfectly balanced' probability distribution on proper $L$-colourings, i.e., such that at every vertex $v$, every colour appears with equal probability $1/|L(v)|$. In this work we give various bounds on $χ_{\ell}^{\bullet}(G)$, which admit strengthenings for correspondence and local-degree versions. As a corollary, we improve theorems on the related notion of flexible list colouring. In particular we study Cartesian products and $d$-degenerate graphs, and we prove that $χ_{\ell}^{\bullet}(G)$ is bounded from above by the pathwidth of $G$ plus one. The correspondence analogue of the latter is false for treewidth instead of pathwidth.

Fractional list packing for layered graphs

Abstract

The fractional list packing number of a graph is a graph invariant that has recently arisen from the study of disjoint list-colourings. It measures how large the lists of a list-assignment need to be to ensure the existence of a `perfectly balanced' probability distribution on proper -colourings, i.e., such that at every vertex , every colour appears with equal probability . In this work we give various bounds on , which admit strengthenings for correspondence and local-degree versions. As a corollary, we improve theorems on the related notion of flexible list colouring. In particular we study Cartesian products and -degenerate graphs, and we prove that is bounded from above by the pathwidth of plus one. The correspondence analogue of the latter is false for treewidth instead of pathwidth.
Paper Structure (9 sections, 23 theorems, 13 equations, 5 figures, 1 table)

This paper contains 9 sections, 23 theorems, 13 equations, 5 figures, 1 table.

Key Result

Lemma 1.1

Every weighted $\epsilon$-flexibly $k$-choosable graph is $\epsilon$-flexibly $k$-choosable. The converse is false.

Figures (5)

  • Figure 1: Local extension of six correspondence-colourings of a $2$-tree, from a balanced edge $ab$ to $d$, such that $ad$ or $bd$ becomes balanced as well. The bold edges indicate full identity matchings in the cover.
  • Figure 2: A graph with pathwidth $3$.
  • Figure 3: A $2$-fold correspondence-cover $(H,L)$ of the hypercube $Q_3$, demonstrating that $\chi_{c}^{\bullet}(Q_3) >3$. For each vertex $v$ of $Q_3$, its list $L(v)=\{1_v,2_v\}$ is indicated by a grey ellipse surrounding a white dot representing $1_v$ and a black dot representing $2_v$. For most edges $uv$ of $Q_3$, the matching between $L(u)$ and $L(v)$ is the 'identity matching', consisting of edges $1_u 1_v$ and $2_u 2_v$ in $H$. The exceptions are formed by a matching of $Q_3$ that consists of three special edges$uv$ for which the matching between $L(u)$ and $L(v)$ is 'crossing', i.e., consisting of $1_u 2_v$ and $2_u 1_v$. The special edges are chosen such that every cycle $C$ of $Q_3$ traverses an odd number of them. Therefore the restriction of $(H,L)$ to a cycle $C$ of $Q_3$ cannot contain any independent set on $|V(C)|$ vertices. From this it easily follows that every independent set of the $16$-vertex graph $H$ has size at most $5$. Hence the fractional chromatic number of $H$ is at least $16/5$, which is strictly larger than $3$.
  • Figure 4: A graph $G$ with treewidth $3$ and $\chi_c^{\bullet}(G) =5$.
  • Figure 5: A correspondence-cover $(L,H)$ of the graph $G$ in \ref{['fig:treewidth3_chicbullet5']}. The covergraph $H$ has fractional chromatic number $>4$ and all lists of size $\leqslant 4$, thus certifying that $\chi_c^{\bullet}(G) >4$. The grey ellipses indicate the lists. To avoid clutter, only the matchings on the central $K_4$ of $G$ are depicted.

Theorems & Definitions (46)

  • Lemma 1.1: DNP19
  • Theorem 2.1: BMS22
  • Theorem 2.2
  • Theorem 2.3
  • Theorem 2.4: BJKM06
  • Theorem 2.5
  • Theorem 2.6: KMMP22
  • Theorem 2.7
  • Definition 3.1
  • Definition 3.2
  • ...and 36 more