Equiangular lines via improved eigenvalue multiplicity

TL;DR

This work improves the understanding of maximal equiangular line configurations by linking the problem to refined bounds on the multiplicity of the second eigenvalue in graphs. The authors develop new lemmas controlling eigenvalue growth when modifying a graph, and use these to derive sharp bounds on $m_G(\lambda_2)$ in both subexponential and superpolynomial regimes. Consequently, they prove that for angle $\arccos{\frac{1}{2k-1}}$ and sufficiently large $r$, the maximum number of lines in $\mathbb{R}^r$ is $r-1+\left\lfloor\frac{r-1}{k-1}\right\rfloor$, improving prior doubly-exponential requirements to a polynomial-exponential threshold; they also show that $r+o(r)$ lines suffice for any fixed $\alpha$ when $r$ grows superpolynomially in $1/\alpha$. In the exponential regime, the bounds are essentially tight, and the paper highlights the construction achieving equality when $1/\alpha$ is an odd integer. These results advance the spectral-graph theoretic approach to equiangular lines, with potential implications for related extremal problems under spectral constraints.

Abstract

A family of lines passing through the origin in an inner product space is said to be equiangular if every pair of lines defines the same angle. In 1973, Lemmens and Seidel raised what has since become a central question in the study of equiangular lines in Euclidean spaces. They asked for the maximum number of equiangular lines in $\mathbb{R}^r$ with a common angle of $\arccos{\frac{1}{2k-1}}$ for any integer $k \geq 2$. We show that the answer equals $r-1+\left\lfloor\frac{r-1}{k-1}\right\rfloor,$ provided that $r$ is at least exponential in a polynomial in $k$. This improves upon a recent breakthrough of Jiang, Tidor, Yao, Zhang, and Zhao [Ann. of Math. (2) 194 (2021), no. 3, 729-743], who showed that this holds for $r$ at least doubly exponential in a polynomial in $k$. We also show that for any common angle $\arccosα$, the answer equals $r+o(r)$ already when $r$ is superpolynomial in $1/α\to \infty$. The key new ingredient underlying our results is an improved upper bound on the multiplicity of the second-largest eigenvalue of a graph. In one of the regimes, this improves and significantly extends a result of McKenzie, Rasmussen, and Srivastava [STOC 2021, pp. 396-407].

Theorems & Definitions (18)

• Theorem 1.1
• Theorem 1.2
• Corollary 1.3
• Theorem 1.4
• Theorem 1.5
• Corollary 1.6
• Lemma 2.1
• Lemma 2.2
• Lemma 2.3
• Theorem 2.4