An improved bound for the linear arboricity conjecture

DOI10.1007/S00493-023-00024-9MaRDI QIDQ6081382zbMATH OpenOpenAlexWikidataFDO

Authors Richard Lang, Luke Postle

Publication date 4 October 2023

Published in Combinatorica (Search for Journal in Brave)

Copyright license Creative Commons Attribution 4.0 International

Full work available at URL https://arxiv.org/abs/2008.04251

zbMATH Keywords

edge colouring graph colouring linear arboricity list colouring

Mathematics Subject Classification ID

Coloring of graphs and hypergraphs (05C15) Edge subsets with special properties (factorization, matching, partitioning, covering and packing, etc.) (05C70)

Abstract: In 1980, Akiyama, Exoo and Harary posited the Linear Arboricity Conjecture which states that any graph

G

of maximum degree

D e l t a

can be decomposed into at most

l e f t l c e i l f r a c D e l t a 2 i g h t c e i l

linear forests. (A forest is linear if all of its components are paths.) In 1988, Alon proved the conjecture holds asymptotically. The current best bound is due to Ferber, Fox and Jain from 2020 who showed that

f r a c D e l t a 2 + O (D e l t a^{.661})

suffices for large enough

D e l t a

. Here, we show that

G

admits a decomposition into at most

f r a c D e l t a 2 + 3 s q r t D e l t a l o g^{4} D e l t a

linear forests provided

D e l t a

is large enough. Moreover, our result also holds in the more general list setting, where edges have (possibly different) sets of permissible linear forests. Thus our bound also holds for the List Linear Arboricity Conjecture which was only recently shown to hold asymptotically by Kim and the second author. Indeed, our proof method ties together the Linear Arboricity Conjecture and the well-known List Colouring Conjecture; consequently, our error term for the Linear Arboricity Conjecture matches the best known error-term for the List Colouring Conjecture due to Molloy and Reed from 2000. This follows as we make two copies of every colour and then seek a proper edge colouring where we avoid bicoloured cycles between a colour and its copy; we achieve this via a clever modification of the nibble method.

Review text

This article is mainly a result about list edge colourings of graphs, though there is much emphasis on a particular application. Recall that if \(G\) is a graph with edge-colouring number \(\chi^{\prime}(G)\) (which by Vizing's theorem is either the maximum degree \(\Delta(G)\) or \(\Delta(G)+1\)) then \(\chi_{\ell}^{\prime}(G)\) is the smallest \(k\) such that, whenever every edge \(e\) has a list \(L(e)\) of at least \(k\) colours, then there is a proper edge-colouring of \(G\) in which every edge takes a colour from its list. One easily sees that \(\chi_{\ell}^{\prime}(G)\geq \chi^{\prime}(G)\): the list edge-colouring conjecture is that \(\chi_{\ell}^{\prime}(G)=\chi^{\prime}(G)\). Thus, though in general, the (vertex) list chromatic number of a graph can be a lot higher than the ordinary (vertex) chromatic number - order of magnitude \(\log(n)\), where \(n\) is the number of vertices, times as high - edge-colourings are conjectured to behave much better. \textit{J. Kahn} [J. Comb. Theory, Ser. A 73, No. 1, 1--59 (1996; Zbl 0851.05081)] showed, using the Rődl nibble technique, that the conjecture is asymptotically true in that \(\chi_{\ell}^{\prime}(G)\leq \Delta(1+o(1))\), and \textit{M. Molloy} and \textit{B. Reed} [Random Struct. Algorithms 17, No. 3--4, 376--402 (2000; Zbl 0971.05047)] sharpened the result to \(\chi_{\ell}^{\prime}(G)\leq \Delta(G)+C\Delta^{1/2}\log^{4}(\Delta)\) for some constant \(C\).\par A \(t\)-edge colouring of a graph is a colouring of the edges in which each edge receives a colour from its list, but we now allow an edge to be incident with at most \(t\) other edges of that colour. The main result of the paper under review is now that if \(G\) is a graph with sufficiently large maximum degree \(\Delta\) and each edge \(e\) receives a list \(L(e)\) of colours, with \(\vert L(e) \vert \geq \Delta/2+3\sqrt{\Delta}\log^{4}(\Delta)\) then there is a 2-edge colouring of \(G\), with each edge taking a colour from its list, in which there are no monochromatic cycles. This result is modelled on Molloy and Reed's arguments. The key new idea is to remove colours which would create monochromatic cycles during the colouring process and to show by careful analysis that this does not affect the nibble stage too much.\par The key application is the fact that any graph \(G\) with maximum degree \(\Delta(G)\) has a decomposition of its edges into at most \(\Delta(G) /2+3\Delta\ast G)^{1/2} \log^{4}(\Delta)\) linear forests, where a linear forest is a forest whose components are paths. This is progress towards the conjecture of \textit{J. Akiyama} et al. [Math. Slovaca 30, 405--417 (1980; Zbl 0458.05050)] that in fact \(\lceil \Delta(G)/2 \rceil\) linear forests should suffice. The previous best result had been \(\Delta/2+O(\Delta^{0.661})\) by \textit{A. Ferber} et al. [J. Comb. Theory, Ser. B 142, 56--79 (2020; Zbl 1436.05047)]. This argument also works in the list colouring context. Since this paper was written, a preprint by \textit{N. Draganić} et al. [``Optimal Hamilton covers and linear arboricity for random graphs, Preprint, arXiv:2310.11580 [math.CO] (2023)] has appeared which (Corollary 1.3) proves the conjecture of Akiyama et al. [loc. cit.] for random graphs \(G(n,p)\) for a broad range of value of \(p\).

Recommendations

Cited in

(11)

This page was built for publication: An improved bound for the linear arboricity conjecture

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6081382)