The evolution of unavoidable bi-chromatic patterns and extremal cases of balanceability

TL;DR

This paper investigates unavoidable edge color patterns in large complete graphs under various restrictions, analyzing their evolution and implications for the balancing number of graphs, with new results on pattern emergence and extremal cases.

Contribution

It characterizes how unavoidable patterns evolve from no restriction to near-perfect balance and explores their impact on the balancing number, including extremal and characterized cases.

Findings

Unavoidable patterns emerge at weaker restrictions when forbidding certain sub-structures.

The balancing number can grow as large as $c n^{2- ext{epsilon}}$, showing high extremal behavior.

Graphs with bounded balancing number are characterized based on their structure.

Abstract

We study the color patterns that, for $n$ sufficiently large, are unavoidable in $2$-colorings of the edges of a complete graph $K_n$ with respect to $\min \{e(R), e(B)\}$, where $e(R)$ and $e(B)$ are the numbers of red and, respectively, blue edges. More precisely, we determine how such unavoidable patterns evolve from the case without restriction in the coloring, namely that $\min \{e(R), e(B)\} \ge 0$ (given by Ramsey's theorem), to the highest possible restriction, namely that $|e(R) - e(B)| \le 1$. We also investigate the effect of forbidding certain sub-structures in each color. In particular, we show that, in $2$-colorings whose graphs induced by each of the colors are both free from an induced matching on $r$ edges, the appearance of the unavoidable patterns is already granted with a much weaker restriction on $\min \{e(R), e(B)\}$. We finish analyzing the consequences of these results to the balancing number $bal(n,G)$ of a graph $G$ (i.e. the minimum $k$ such that every $2$-edge coloring of $K_n$ with $\min \{e(R), e(B)\} > k$ contains a copy of $G$ with half the edges in each color), and show that, for every $\varepsilon > 0$, there are graphs $G$ with $bal(n,G) \ge c n^{2-\varepsilon}$, which is the highest order of magnitude that is possible to achieve, as well as graphs where $bal(n,G) \le c(G)$, where $c(G)$ is a constant that depends only $G$. We characterize the latter ones.