Characterizing graphs with the second largest distance eigenvalue less than -1/2
Miriam Abdón, Lilian Markenzon, Cybele T. M. Vinagre
TL;DR
The paper resolves the problem of characterizing connected graphs $G$ with $\lambda_2(G)<-1/2$ for their distance matrix by showing such graphs are exactly the chordal graphs that are not block graphs and fall into two structural regimes: diameter $2$ subgraphs of relaxed block stars and diameter $3$ subgraphs of Pt1 or Pt2(p,q) with $p,q\ge2$. It then proves that the identified graph families—relaxed block stars, Pt1, and Pt2(p,q)—indeed satisfy $\lambda_2<-1/2$ using distance-matrix spectral tools, including equitable partitions and rigorous root-bounds via Descartes’ Rule of Signs and Sturm’s Theorem. The results hinge on a tight interplay between minimal vertex separators and graph classes (Ptolemaic, split, block, and distance-hereditary structures) to achieve a complete, finite description. This work advances the understanding of distance matrices in graphs, providing a precise taxonomy that can guide rapid identification of graphs with strong distance-spectral constraints in chordal graph theory.
Abstract
Let $G$ be a connected graph with vertex set $V$. The distance, $d_G(u, v)$, between vertices $u$ and $v$ of $G$ is defined as the length of a shortest path between $u$ and $v$ in $G$. The distance matrix of $G$ is the matrix $\mathbf{D}(G) =[d_G(u, v)]_{u,v\in V}$. The second largest distance eigenvalue $λ_2(G)$ of $G$ is the second largest one in the spectrum of $\mathbf{D}(G)$. In this work, we completely characterize the connected graphs $G$ for which $λ_2(G)<-1/2$ through approaches both spectral and structural.
