Bounded degree graphs and hypergraphs with no full rainbow matchings
Ronen Wdowinski
TL;DR
This work studies full rainbow matchings in edge-colored multi-hypergraphs, showing that large color classes relative to the maximum degree do not guarantee existence of a full rainbow matching. It develops an iterative join-lemma construction to build $r$-graphs with maximum degree $oldsymbol{ riangle}$ and color classes of size at least $roldsymbol{ riangle}-1$ that lack such matchings, establishing the optimality of the Aharoni–Berger–Meshulam bound. Extending to properly edge-colored multigraphs, the paper disproves Delcourt–Postle conjectures and links rainbow matchings to list-edge-coloring, proving that a color-degree generalization of Galvin’s theorem fails. The results rely on nets, sunflowers, and blow-up techniques, and yield insights with implications for list-edge-coloring, chromatic indices, and related extremal problems in hypergraphs and graphs.
Abstract
Given a multi-hypergraph $G$ that is edge-colored into color classes $E_1, \ldots, E_n$, a full rainbow matching is a matching of $G$ that contains exactly one edge from each color class $E_i$. One way to guarantee the existence of a full rainbow matching is to have the size of each color class $E_i$ be sufficiently large compared to the maximum degree of $G$. In this paper, we apply a simple iterative method to construct edge-colored multi-hypergraphs with a given maximum degree, large color classes, and no full rainbow matchings. First, for every $r \ge 1$ and $Δ\ge 2$, we construct edge-colored $r$-uniform multi-hypergraphs with maximum degree $Δ$ such that each color class has size $|E_i| \ge rΔ- 1$ and there is no full rainbow matching, which demonstrates that a theorem of Aharoni, Berger, and Meshulam (2005) is best possible. Second, we construct properly edge-colored multigraphs with no full rainbow matchings which disprove conjectures of Delcourt and Postle (2022). Finally, we apply results on full rainbow matchings to list edge-colorings and prove that a color degree generalization of Galvin's theorem (1995) does not hold.
