Chromatic Number Ratio in Graph Theory

• CNR is defined as the ratio of a graph's chromatic number to its clique number, indicating how far a graph's coloring requirements exceed its largest complete subgraph.
• Research using probabilistic methods and extremal constructions has refined upper and lower bounds on CNR, revealing its growth behavior in random and structured graphs.
• Spectral, topological, and convex optimization approaches offer alternate methodologies to estimate CNR, emphasizing its significance in combinatorial and graph theory.

The chromatic number ratio (CNR) is a central parameter in graph theory and combinatorics, quantifying the relationship between a graph's chromatic number $\chi(G)$ (the minimum number of colors needed for a proper coloring) and its clique number $\omega(G)$ (the size of the largest complete subgraph). Formally, $\mathrm{CNR}(G) = \chi(G)/\omega(G)$. CNR measures the maximal possible discrepancy between these two fundamental invariants and is intimately connected to the limitations of structural and topological lower bounds for chromatic number, as well as to classic problems in extremal and random graph theory.

1. Definitions and Fundamental Examples

For a finite simple graph $G=(V,E)$, the chromatic number $\chi(G)$ is the smallest $k$ admitting a proper coloring $c:V\to\{1,\ldots,k\}$ such that $c(u)\ne c(v)$ whenever $\{u,v\}\in E$. The clique number $\omega(G)$ is the maximum cardinality of a subset of pairwise adjacent vertices. The chromatic number ratio is defined as

$\mathrm{CNR}(G) = \frac{\chi(G)}{\omega(G)}.$

This parameter captures extremal coloring behavior, particularly for graphs where $\chi(G)$ far exceeds $\omega(G)$. Notably, Mycielski’s construction and classic triangle-free graphs provide templates where the chromatic number is arbitrarily large while the clique number remains bounded.

2. Extremal Growth: Maximum Chromatic Number Ratio

A foundational problem is to determine the asymptotics of $f(n) = \max_{|V(G)|=n} \chi(G)/\omega(G)$ as $n \to \infty$. In 1967, Erdős established that there exist absolute constants $c_1, c_2 > 0$ with

$$\left(\frac{1}{4} + o(1)\right) \frac{n}{(\log_2 n)^2} \le f(n) \le \left(4 + o(1)\right) \frac{n}{(\log_2 n)^2}$$

and demonstrated that this quantity is sharply controlled by Ramsey theory and the growth of diagonal Ramsey numbers $R(k,k)$. Recent advances have further improved the upper bound, yielding

$f(n) \le (c + o(1)) \frac{n}{(\log_2 n)^2}$

with $c < 3.72$ (Araujo et al., 18 Dec 2025). This utilizes refined probabilistic arguments, entropy-based analysis, and sharper diagonal Ramsey estimates. The asymptotics remain an outstanding problem: the existence and value of $\lim_{n\to\infty} f(n)/(n/(\log n)^2)$ is unresolved.

Bound type	Expression	Reference
Lower bound	$$\left(\frac{1}{4} + o(1)\right)\frac{n}{(\log_2 n)^2}$$	(Araujo et al., 18 Dec 2025)
Upper bound (Erdős)	$\left(4 + o(1)\right)\frac{n}{(\log_2 n)^2}$	(Araujo et al., 18 Dec 2025)
New upper bound	$(c + o(1))\frac{n}{(\log_2 n)^2},\ c<3.72$	(Araujo et al., 18 Dec 2025)

3. Limits of Classical and Topological Lower Bounds

Classical lower bounds for the chromatic number include the trivial $\chi(G)\geq\omega(G)$ and the Lovász topological bound, which states that if the neighborhood complex $\mathcal{N}(G)$ is $k$-connected, then $\chi(G)\geq k+3$. However, these lower bounds can be arbitrarily bad approximations for $\chi(G)$. Explicit constructions exist where both $\chi(G)/\omega(G)$ and $\chi(G)/(\mathrm{conn}(\mathcal{N}(G))+3)$ are unbounded as $n\to\infty$. For any integers $2\le p\le q$, one can construct a connected graph $G$ with

$\chi(G)=q,\quad \omega(G)=p,$

so

$\mathrm{CNR}(G)=\frac{q}{p}$

is unbounded, while the Lovász bound remains constant ($=3$) in such examples (Daneshpajouh, 2017). These constructions undermine the possibility of any universal approximation of $\chi(G)$ by $\omega(G)$ or by currently known topological invariants.

4. CNR in Random and Structured Graph Classes

The typical behavior of CNR in random graph models and special classes is a subject of significant interest. In the dense Erdős–Rényi random graph $G\sim G(n,p)$ for fixed $p\in (0,1)$, the coloring rate $\bar\alpha(G)=n/\chi(G)$ (the inverse of CNR for $n$ vertices) concentrates tightly at an explicit threshold:

$\bar\alpha(G) = \gamma - x_0 + o(1)$

where $\gamma = 2\log_b n - 2\log_b\log_b n - 2\log_b2$ and $x_0$ solves an explicit transcendental equation. This answers the CNR problem for random graphs to within $o(1)$ (Heckel, 2016). A phase transition occurs at $p=1-1/e^2$, changing the precise limiting behavior for $p$ above and below this threshold.

For geometric graphs, such as unit disk graphs with independence number two (stability two), it is proved that

$\chi(G) \leq \frac{3}{2}\omega(G)$

with equality achieved asymptotically by circulant constructions; thus, CNR attains $3/2$ as a sharp bound for this family (Bruhn, 2011).

5. Spectral and Convex Optimization Bounds for CNR

Spectral techniques, extending Hoffman's eigenvalue method, provide general lower bounds for chromatic numbers of both finite and infinite graphs and thus indirectly control CNR. For a bounded self-adjoint operator $T$ (generalizing the adjacency matrix), the lower bounds

$$\chi(T) \geq \frac{M(T)-m(T)}{-m(T)},\qquad \chi^*(T)\geq \frac{(T\mathbb{1},\mathbb{1})-m(T)}{-m(T)}$$

hold, with $m(T), M(T)$ the minimum and maximum of the numerical range. In the finite case, these reduce to classical Hoffman's bounds for graphs. For infinite graphs, e.g., on $\mathbb{R}^n$ or the unit sphere, harmonic analysis allows explicit estimation of these spectral parameters and hence CNR (Bachoc et al., 2013).

Convex optimization (semidefinite programming) generalizations, in the spirit of Lovász's $\vartheta$-function, tightly yield upper and lower bounds on both independence ratio and chromatic number, with exactness in vertex-transitive cases.

6. Open Problems and Directions

Fundamental open problems and conjectures on CNR include:

• Determining the exact leading constant in $f(n) = \max_{|V(G)|=n} \chi(G)/\omega(G)$ and establishing the existence of the limit $\lim_{n\to\infty} f(n)/(n/(\log n)^2)$.
• Identifying universal constants or structural graph classes where $\mathrm{CNR}(G)$ is uniformly bounded or sharply characterized, beyond special families such as unit disk graphs.
• Finding new topological invariants or refined complexes capable of bounding $\chi(G)$ within a constant factor for all graphs, surpassing the limitations of neighborhood complexes and the Lovász bound (Daneshpajouh, 2017).
• Understanding the relationship between CNR and diagonal Ramsey numbers, especially in light of conjectured inequalities (such as the Ramsey Diagonal Conjecture) and their extremal implications (Araujo et al., 18 Dec 2025).

Resolution of these problems would sharpen the understanding of the fundamental gap between chromatic and clique number, the power of topological and spectral lower bounds, and the extremal landscape of coloring problems.