Table of Contents
Fetching ...

On the second-largest modulus among the eigenvalues of a power hypergraph

Changjiang Bu, Lixiang Chen, Yongtang Shi

TL;DR

This work studies the second-largest eigenvalue modulus $\Lambda$ of the $k$-power hypergraph $G^{(k)}$ and ties it to the weakest edges of the base graph $G$. By leveraging signed subgraphs and the projective eigenvariety framework, it obtains a sharp formula $\Lambda^k=\rho_E(G)^2$ for $k\ge4$ (with a complete case analysis for $k=3$) and proves that $\mathbb{V}_{\Lambda}(G^{(k)})$ is zero-dimensional, i.e., there are finitely many eigenvectors up to scalar. The authors then give explicit expressions for the algebraic multiplicity $\mathrm{am}_{\Lambda}(G^{(k)})$ and the total multiplicity $\#\mathbb{V}_{\Lambda}(G^{(k)})}$, showing $\mathrm{am}_{\Lambda}(G^{(k)})=\#\mathbb{V}_{\Lambda}(G^{(k)})$ and representing both in terms of the number of weakest edges. These results crystallize how edge fragility in $G$ governs the secondary spectral structure of its power hypergraph and connect hypergraph tensor spectra to signed-subgraph dynamics.

Abstract

It is well known that the algebraic multiplicity of an eigenvalue of a graph (or real symmetric matrix) is equal to the dimension of its corresponding linear eigen-subspace, also known as the geometric multiplicity. However, for hypergraphs, the relationship between these two multiplicities remains an open problem. For a graph $G=(V,E)$ and $k \geq 3$, the $k$-power hypergraph $G^{(k)}$ is a $k$-uniform hypergraph obtained by adding $k-2$ new vertices to each edge of $G$, who always has non-real eigenvalues. In this paper, we determine the second-largest modulus $Λ$ among the eigenvalues of $G^{(k)}$, which is indeed an eigenvalue of $G^{(k)}$. The projective eigenvariety $\mathbb{V}_Λ$ associated with $Λ$ is the set of the eigenvectors of $G^{(k)}$ corresponding to $Λ$ considered in the complex projective space. We show that the dimension of $\mathbb{V}_Λ$ is zero, i.e, there are finitely many eigenvectors corresponding to $Λ$ up to a scalar. We give both the algebraic multiplicity of $Λ$ and the total multiplicity of the eigenvector in $\mathbb{V}_Λ$ in terms of the number of the weakest edges of $G$. Our result show that these two multiplicities are equal.

On the second-largest modulus among the eigenvalues of a power hypergraph

TL;DR

This work studies the second-largest eigenvalue modulus of the -power hypergraph and ties it to the weakest edges of the base graph . By leveraging signed subgraphs and the projective eigenvariety framework, it obtains a sharp formula for (with a complete case analysis for ) and proves that is zero-dimensional, i.e., there are finitely many eigenvectors up to scalar. The authors then give explicit expressions for the algebraic multiplicity and the total multiplicity , showing and representing both in terms of the number of weakest edges. These results crystallize how edge fragility in governs the secondary spectral structure of its power hypergraph and connect hypergraph tensor spectra to signed-subgraph dynamics.

Abstract

It is well known that the algebraic multiplicity of an eigenvalue of a graph (or real symmetric matrix) is equal to the dimension of its corresponding linear eigen-subspace, also known as the geometric multiplicity. However, for hypergraphs, the relationship between these two multiplicities remains an open problem. For a graph and , the -power hypergraph is a -uniform hypergraph obtained by adding new vertices to each edge of , who always has non-real eigenvalues. In this paper, we determine the second-largest modulus among the eigenvalues of , which is indeed an eigenvalue of . The projective eigenvariety associated with is the set of the eigenvectors of corresponding to considered in the complex projective space. We show that the dimension of is zero, i.e, there are finitely many eigenvectors corresponding to up to a scalar. We give both the algebraic multiplicity of and the total multiplicity of the eigenvector in in terms of the number of the weakest edges of . Our result show that these two multiplicities are equal.

Paper Structure

This paper contains 6 sections, 20 theorems, 40 equations, 1 figure.

Key Result

Lemma 2.1

CHEN2024105909 Let $G$ be a connected graph. Then $\rho(G_{\pi}) \leq \rho(G)$, with equality if and only if $G_{\pi}$ is switching equivalent to $G_{+}$ or $G_{-}$.

Figures (1)

  • Figure .1: Examples for Theorem \ref{['thm4']}

Theorems & Definitions (29)

  • Lemma 2.1
  • Lemma 2.2
  • Lemma 2.3
  • Proposition 2.4
  • proof
  • Proposition 2.5
  • proof
  • Lemma 2.6
  • Lemma 2.7
  • Theorem 2.8
  • ...and 19 more