On Bj\"{o}rner and Lov\'{a}sz's conjecture

TL;DR

This paper investigates topological bounds on graph chromatic numbers, proves certain graphs with involutive automorphisms are test graphs, and provides a combinatorial proof of Lovász's conjecture with a modified version that always holds.

Contribution

It establishes that graphs with involutive automorphisms flipping edges are test graphs and introduces a supergraph construction that makes any graph a test graph.

Findings

Graphs with involutive automorphisms flipping edges are test graphs.

Odd cycles are test graphs, providing a combinatorial proof of Lovász's conjecture.

A modified conjecture holds for a supergraph with chromatic number at most one greater than the original.

Abstract

In the way of proving Kneser's conjecture, L\'{a}szl\'{o} Lov\'{a}sz settled out a new lower bound for the chromatic number of graphs. He showed that if the hom complex $||Hom(\mathcal{K}_2, H)||$ of a graph $H$ is topologically $k$-connected, then its chromatic number, $\chi (H)$, is at least $k+3$. After that, he made the following conjecture, to provide a better lower bound on the chromatic number of graphs. If $||Hom(C_{2r+1}, H)||$ is $k$-connected, then $\chi (H)\geq k+4$, where $C_{2r+1}$ is an odd cycle of length $2r+1$. Finally, Bj\"{o}rner and Lov\'{a}sz proposed a generalization of the Lov\'{a}sz conjecture as follows. If $||Hom(T, H)||$ is $k$-connected, then $\chi (H)\geq k+\chi (T) +1$. The first conjecture was originally confirmed by Babson and Kozlov, by complicated computations with spectral sequences. But the second one was disproved by Hoory and Linial. So, after that, a graph $T$ is called a test graph if for every graph $H$, the $k$-connectedness of $||Hom(T, H)||$ implies $\chi (H)\geq k + 1 + \chi(H)$. In this paper, we prove that if a graph $F$ possess an involutive automorphism that flips some edge; then it is a test graph. As a corollary, we give a purely combinatorial proof of the Lov\'{a}sz conjecture; odd cycles are test graphs. Indeed, although the Bj\"{o}rner and Lov\'{a}sz conjecture is not true in general, we show that a slight modification of the conjecture is always true. More precisely, we show that for any graph $T$, there is a supergraph $T\subseteq\hat{T}$ with $\chi(\hat{T})\leq\chi(T)+1$, such that $\hat{T}$ is a test graph.