Chromatic numbers from edge ideals: Graph classes with vanishing syzygies are polynomially $χ$-bounded
Alexander Engström
TL;DR
The paper connects graph coloring to commutative algebra by studying edge ideals and the vanishing of Betti numbers β_{i,j}(I_G). It identifies a hierarchy of forbidden subgraphs B_{n,d} that encode these vanishing syzygies and proves that graphs free of such obstructions are χ-bounded by explicit polynomials in the clique number, with degree 2(j−i−2). It additionally provides sharper bounds for triangle-free graphs and establishes asymptotic and algorithmic results, including polynomial-time coloring procedures. This framework creates a bridge between algebraic invariants and combinatorial colorability, offering new polynomial bounds and practical coloring algorithms for large graph classes.
Abstract
The chromatic number $χ$ of a graph is bounded from below by its clique number $ω,$ but it can be arbitrary large. Perfect graphs are defined by $χ=ω$ for all induced subgraphs. An interesting relaxation are $χ$-bounded graph classes, where $χ\leq f(ω).$ It is not always possible to achieve this with a polynomial $f.$ The edge ideal $I_G$ of a graph $G$ is generated by monomials $x_ux_v$ for each edge $uv$ of $G.$ The bi-graded betti numbers $β_{i,j}(I)$ are central algebraic geometric invariants. We study the graph classes where for some fixed $i,j$ that syzygy vanishes, that is, $β_{i,j}(I_G)=0.$ We prove that $χ\leq f(ω),$ where $f$ is a polynomial of degree $2j-2i-4.$ For the elementary special case $β_{i,2i+2}(I_G)=0,$ this amounts to that $(i+1)K_2$-free graphs are ${ω-1+2i \choose 2i}$-colorable, improving on an old combinatorial result by Wagon. We also show that triangle-free graphs with $β_{i,j}(I_G)=0$ are $(j-1)$-colorable. Complexity wise, we show that these colorings can be derived in time $O(n^3)$ for graphs on $n$ vertices. Moreover, we show that for almost all graphs with parabolic $i,j,$ there are better bounds on $χ.$