TheoremComplete

Vizing's Theorem

Theorem5.2Vizing's Theorem

For every simple graph GG, the chromatic index satisfies Ξ”(G)≀χ′(G)≀Δ(G)+1\Delta(G) \leq \chi'(G) \leq \Delta(G) + 1.


Proof

Proof

The lower bound Ο‡β€²β‰₯Ξ”\chi' \geq \Delta is trivial. For the upper bound, we prove that any simple graph GG can be properly edge-colored with Ξ”+1\Delta + 1 colors by an augmenting path argument.

Setup. Assume we have a partial proper edge-coloring Ο•\phi with Ξ”+1\Delta + 1 colors, and let e=v0v1e = v_0v_1 be an uncolored edge. We will recolor edges to accommodate ee.

Fan construction. Build a fan at v0v_0: a sequence of distinct neighbors v1,v2,…,vkv_1, v_2, \ldots, v_k of v0v_0 such that the color of edge v0vj+1v_0 v_{j+1} is a color missing at vjv_j for each jj. Formally: let cj=Ο•(v0vj)c_j = \phi(v_0 v_j) for jβ‰₯1j \geq 1. Choose vj+1v_{j+1} so that cj+1c_{j+1} is missing at vjv_j (i.e., cj+1βˆ‰Ο•(edgesΒ atΒ vj)c_{j+1} \notin \phi(\text{edges at } v_j)). Continue until no further extension is possible.

Key observations. When the fan cannot be extended, every color missing at vkv_k is present at v0v_0 (otherwise we could extend). Since deg⁑(v0)≀Δ\deg(v_0) \leq \Delta and we use Ξ”+1\Delta + 1 colors, there is a color Ξ±\alpha missing at v0v_0.

Case 1: If Ξ±\alpha is also missing at some vjv_j in the fan, then shift the fan: uncolor v0v1v_0v_1, recolor v0vjβ€²v_0v_{j'} with the color of v0vjβ€²+1v_0v_{j'+1} for jβ€²=1,…,jβˆ’1j' = 1, \ldots, j-1, and color v0vjv_0v_j with Ξ±\alpha. Then color v0v1v_0v_1 with the free color.

Case 2: Ξ±\alpha is present at all vjv_j, and there is a color Ξ²\beta missing at vkv_k but present at v0v_0. Consider the Ξ±Ξ²\alpha\beta-path starting at vkv_k: the maximal path alternating colors Ξ±\alpha and Ξ²\beta starting from vkv_k.

  • If this path does not reach v0v_0: swap Ξ±\alpha and Ξ²\beta on this path. Now Ξ±\alpha is missing at vkv_k, reducing to Case 1 (after fan shifting).

  • If the path reaches v0v_0: it arrives at v0v_0 through some vjv_j (since Ξ²=cj\beta = c_j for some jj in the fan). Now we cannot swap on this path directly. Instead, shift the fan only up to vjv_j (recolor v0vjβ€²v_0v_{j'} with cjβ€²+1c_{j'+1} for jβ€²=1,…,jβˆ’1j'=1,\ldots,j-1), making Ξ²\beta free at v0v_0. Then swap Ξ±,Ξ²\alpha,\beta on the path from vkv_k (which no longer connects to v0v_0 through the recolored edge), making Ξ±\alpha free at vkv_k. Color v0v1v_0v_1 with a free color.

In all cases, we extend the coloring to include ee. Starting from the empty coloring and adding edges one by one completes the proof. β–‘\square

β– 

ExampleThe Petersen Graph is Class 2

The Petersen graph has Ξ”=3\Delta = 3 and 15 edges. If it were 3-edge-colorable, each color class would be a perfect matching (5 edges). But the Petersen graph has no Hamiltonian cycle, and by a counting argument, three perfect matchings cannot partition its edge set. Thus Ο‡β€²=4=Ξ”+1\chi' = 4 = \Delta + 1.

RemarkKΓΆnig's Edge Coloring Theorem

For bipartite graphs: Ο‡β€²(G)=Ξ”(G)\chi'(G) = \Delta(G) (KΓΆnig, 1916). The proof is constructive: iteratively find and remove perfect matchings using Hall's theorem. This shows all bipartite graphs are Class 1, in contrast to the difficulty of classifying general graphs.