← Back to arXiv
arXivCombinatoricsarXiv:2609.31674

Exact majority C-colourings of balanced Hamming graphs and grids

The paper studies a graph-coloring problem called "majority C-coloring." The setup is this: you have a network (graph) of vertices connected by edges, and you want to divide the vertices into groups (color classes) such that every vertex has at least half of its neighbors in the same group as itself. The central question is how many groups you can create at most while satisfying this condition. The paper focuses on two families of graphs: Hamming graphs, which are high-dimensional grid-like structures built by taking a complete graph and repeating it across multiple dimensions, and simpler two-dimensional grids formed by combining cycle graphs with path graphs.

For Hamming graphs, the authors derive an exact formula for the maximum number of color classes. They build the lower bound (showing you can achieve at least this many groups) through clever geometric constructions involving rectangular partitions and a bridging argument that stitches lower-dimensional solutions into higher-dimensional ones. The upper bound (showing you cannot do better) comes from a classical result in combinatorics about how information spreads across edges in Hamming-type structures, and the authors provide a clean, self-contained proof of this direction. The combination pins down the answer exactly, which is rare and satisfying in this kind of extremal combinatorics problem.

For the two-dimensional cycle-path grids, the authors prove an exact formula for the maximum number of color classes when the cycle length is at least 4 and the path length is even. A notable consequence is that this result directly contradicts a conjecture published in a very recent companion paper on the same topic. Specifically, the conjecture made a prediction about cylinder-shaped graphs that turns out to be wrong for a broad range of parameters. The paper also derives additional bounds for related families like tori, extending the picture further. Taken together, the results advance the understanding of how local majority constraints shape global graph structure.

Read original →