An improved bound for the linear arboricity conjecture (Q6081382)

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 \(p\).