Table of Contents
Fetching ...

A Spectral Lower Bound on Chromatic Numbers using $p$-Energy

Clive Elphick, Quanyu Tang, Shengtong Zhang

TL;DR

This work introduces a unified framework of $p$-energies, decomposing a graph's adjacency spectrum into positive and negative parts to derive a family of lower bounds for $\chi(G)$ and its relaxations. By combining stochastic block decompositions, Haar-measure averaging, and majorization techniques, the authors prove that for all $p\ge 0$, $\xi_f(G) \ge 1 + \max\left\{ \mathcal{E}_p^+(G)/\mathcal{E}_p^-(G), \mathcal{E}_p^-(G)/\mathcal{E}_p^+(G) \right\}^{1/|p-1|}$, which tightens and unifies the classical bounds (Hoffman, Ando–Lin, Elphick–Wocjan) at $p=\infty,2,0$ respectively. The approach uses a novel synthesis of linear algebra and measure theory, including stochastic block decompositions, majorization, and the Haar measure, to extend the analysis to non-integer $p$, yielding sharper bounds in several graphs such as the Tilley graph and a class of graphs where the optimal bound occurs at non-integer $p$. The results also resolve two conjectures by Elphick and Wocjan and imply tightness for balanced Turán graphs, highlighting the practical impact for estimating chromatic parameters in spectral graph theory.

Abstract

Let $A_G $ be the adjacency matrix of a simple graph $ G $, and let $ χ(G) $, $ χ_f(G) $, $ χ_q(G) $, $ ξ(G) $ and $ ξ_f(G) $ denote its chromatic number, fractional chromatic number, quantum chromatic number, orthogonal rank and projective rank, respectively. For $ p \geq 0 $, we define the positive and negative $ p $-energies of $ G $ by $$ \mathcal{E}_p^+(G) = \sum_{λ_i > 0} λ_i^p, \quad \mathcal{E}_p^-(G) = \sum_{λ_i < 0} |λ_i|^p, $$ where $ λ_1 \geq \cdots \geq λ_n $ are the eigenvalues of $A_G $. We prove that for all $ p \geq 0 $, $$ χ(G) \geq \left\{χ_f(G), χ_q(G), ξ(G) \right\} \geq ξ_f(G) \geq 1 + \max\left\{ \frac{\mathcal{E}_p^+(G)}{\mathcal{E}_p^-(G)}, \frac{\mathcal{E}_p^-(G)}{\mathcal{E}_p^+(G)} \right\}^{\frac{1}{|p - 1|}}. $$ This result unifies and strengthens a series of existing bounds corresponding to the cases $ p \in \{0, 2, \infty\} $. In particular, the case $ p = 0 $ yields the inertia bound $$ χ_f(G) \geq ξ_f(G) \geq1 + \max\left\{\frac{n^+}{n^-}, \frac{n^-}{n^+}\right\}, $$ where $ n^+ $ and $ n^- $ denote the number of positive and negative eigenvalues of $ A_G $, respectively. This resolves two conjectures of Elphick and Wocjan. We also demonstrate that for certain graphs, non-integer values of $ p $ provide sharper lower bounds than existing spectral bounds. As an example, we determine $ χ_q $ for the Tilley graph, which cannot be achieved using existing (unweighted) $p$-energy bounds. Our proof employs a novel synthesis of linear algebra and measure-theoretic tools, which allows us to surpass existing spectral bounds.

A Spectral Lower Bound on Chromatic Numbers using $p$-Energy

TL;DR

This work introduces a unified framework of -energies, decomposing a graph's adjacency spectrum into positive and negative parts to derive a family of lower bounds for and its relaxations. By combining stochastic block decompositions, Haar-measure averaging, and majorization techniques, the authors prove that for all , , which tightens and unifies the classical bounds (Hoffman, Ando–Lin, Elphick–Wocjan) at respectively. The approach uses a novel synthesis of linear algebra and measure theory, including stochastic block decompositions, majorization, and the Haar measure, to extend the analysis to non-integer , yielding sharper bounds in several graphs such as the Tilley graph and a class of graphs where the optimal bound occurs at non-integer . The results also resolve two conjectures by Elphick and Wocjan and imply tightness for balanced Turán graphs, highlighting the practical impact for estimating chromatic parameters in spectral graph theory.

Abstract

Let be the adjacency matrix of a simple graph , and let , , , and denote its chromatic number, fractional chromatic number, quantum chromatic number, orthogonal rank and projective rank, respectively. For , we define the positive and negative -energies of by where are the eigenvalues of . We prove that for all , This result unifies and strengthens a series of existing bounds corresponding to the cases . In particular, the case yields the inertia bound where and denote the number of positive and negative eigenvalues of , respectively. This resolves two conjectures of Elphick and Wocjan. We also demonstrate that for certain graphs, non-integer values of provide sharper lower bounds than existing spectral bounds. As an example, we determine for the Tilley graph, which cannot be achieved using existing (unweighted) -energy bounds. Our proof employs a novel synthesis of linear algebra and measure-theoretic tools, which allows us to surpass existing spectral bounds.

Paper Structure

This paper contains 13 sections, 15 theorems, 123 equations, 4 figures, 1 table.

Key Result

Theorem 1.5

Let $\chi(G)$ be the chromatic number of a graph $G$. Then

Figures (4)

  • Figure 1: Partial order of the five graph parameters.
  • Figure 2: A plot of the function $f(p;H_1)$ over $p \in [0, 25]$.
  • Figure 3: The 7-vertex graph $H_2$
  • Figure 4: A plot of the function $f(p;H_2)$ over $p \in [0, 10]$.

Theorems & Definitions (42)

  • Definition 1.1
  • Definition 1.2
  • Definition 1.3
  • Definition 1.4
  • Theorem 1.5: Ando2015
  • Theorem 1.6: Elphick2017
  • Theorem 1.8
  • Remark 1.9
  • Corollary 1.10
  • Definition 2.1
  • ...and 32 more