Acyclic edge-colouring

Medium ★★ Graph Theory » Coloring » Edge coloring

Status partial high confidence

The conjecture that every simple graph with maximum degree $\Delta$ has a proper acyclic edge-colouring using at most $\Delta+2$ colours remains open. Since the problem was posted, the general upper bound has been reduced below $4\Delta-4$, reaching approximately $3.142(\Delta-1)+1$ as of 2026, and the conjecture has been confirmed for several special classes including 3-sparse graphs; moreover, Delcourt, Li, and Postle announced in 2023 that $(1+o(1))\Delta$ colours always suffice, though the exact Fiamčík bound is still unproved for general graphs.

Cited literature (4)

Reviewer notes. An important announced result — Delcourt, Li, and Postle, 'The Acyclic Edge Coloring Conjecture holds asymptotically,' showing $(1+o(1))\Delta$ colours suffice — was presented at the Georgia Tech Graph Theory Seminar on 17 October 2023, but no arXiv preprint or journal URL was locatable; it cannot be included in since_posted. A 2022 arXiv preprint (arXiv:2202.13846, Kirousis–Livieratos) claimed the stronger bound $2\Delta-1$ but was subsequently withdrawn due to errors in key lemmas and must not be cited. The intermediate general-bound improvement by Fialho et al. (2020) — referenced as the prior best in arXiv:2602.14859 — was not independently verified and is therefore not listed.

Auto-reviewed 2026-05-08 with claude-sonnet-4-6 (web search enabled) · 193s.

Conjecture. Every simple graph with maximum degree $ \Delta $ has a proper $ (\Delta+2) $ - edge-colouring so that every cycle contains edges of at least three distinct colours.
Keywords: edge-coloring

Discussion

An edge-colouring with the property that every cycle contains edges of at least three distinct colours is called an acyclic edge-colouring. It is known (see [AMR]) that every graph of maximum degree $ \Delta $ has an acyclic edge-colouring of size $ O(\Delta ) $ . The best upper bound so far is $ 4\Delta -4 $ and is due to Esperet and Parreau [EP]. It is also known (see [ASZ]) that this conjecture is true for graphs with girth at least $ C \Delta \log(\Delta ) $ (for some fixed constant $ C $ ).

Bibliography