English

Extremal eigenvalues of graphs embedded on surfaces

Combinatorics 2026-01-26 v1

Abstract

Graph theory on surfaces extends classical graph structures to topological surfaces, providing a theoretical foundation for characterizing the embedding properties of complex networks in constrained spaces. The study of bounding the spectral radius ρ(G)\rho(G) of graphs on surfaces has a rich history that dates back to the 1990s. In this paper, we establish tight bounds for graphs of order nn that are embeddable on a surface with Euler genus γ\gamma. Specifically, if graph GG achieves the maximum spectral radius, then \begin{equation*} \begin{array}{ll} \frac32\!+\!\sqrt{2n\!-\!\frac{15}4}\!+\!\frac{3\gamma\!-\!1}{n}<\rho(G)<\frac32\!+\!\sqrt{2n\!-\!\frac{15}4}\!+\!\frac{3\gamma\!-\!0.95}{n}, \end{array} \end{equation*} which improves upon the earlier bound ρ(G)2+2n+8γ6\rho(G)\leq2+\sqrt{2n+8\gamma-6} by Ellingham and Zha [JCTB, 2000]. Furthermore, we prove that any extremal graph is obtained from K2Pn2K_2 \nabla P_{n-2} by adding exactly 3γ3\gamma edges, where `\nabla' means the join product. As a corollary, for γ=0\gamma = 0 and n4.5×106n \geq 4.5 \times 10^6, the graph K2Pn2K_2 \nabla P_{n-2} is the unique planar extremal graph, thereby confirming a long-standing conjecture resolved by Tait and Tobin [JCTB, 2017]. Let KrnK_r^n be the graph of order nn obtained by attaching two paths of nearly equal length to two distinct vertices of KrK_r. Integrating spectral techniques with considerable structural analysis on surface graphs, we further derive the following sharp bounds: ρ(G)ρ(K2K4n2)\rho(G) \leq \rho(K_2 \nabla K_4^{n-2}) for projective-planar graphs, and ρ(G)ρ(K2K5n2)\rho(G) \leq \rho(K_2 \nabla K_5^{n-2}) for toroidal graphs. Our study presents a novel framework for exploring the eigenvalue-extremal problem on surface graphs with high Euler genus.

Keywords

Cite

@article{arxiv.2601.16380,
  title  = {Extremal eigenvalues of graphs embedded on surfaces},
  author = {Mingqing Zhai and Longfei Fang and Huiqiu Lin},
  journal= {arXiv preprint arXiv:2601.16380},
  year   = {2026}
}
R2 v1 2026-07-01T09:16:40.344Z