English

Perturbing eigenvalues of non-negative matrices

Spectral Theory 2014-02-06 v1

Abstract

Let AA be an irreducible (entrywise) nonnegative n×nn\times n matrix with eigenvalues ρ,b+ic,bic,λ4,,λn,\rho, b+ic,b-ic, \lambda_4,\cdots,\lambda_n, where ρ\rho is the Perron eigenvalue. It is shown that for any t[0,)t \in [0, \infty) there is a nonnegative matrix with eigenvalues ρ+t~,λ2+t,λ3+t,λ4,λn,\rho+ \tilde t,\lambda_2+t,\lambda_3+t, \lambda_4 \cdots,\lambda_n, whenever t~γnt\tilde t \ge \gamma_n t with γ3=1,γ4=2,γ5=5\gamma_3=1, \gamma_4 = 2, \gamma_5=\sqrt 5 and γn=2.25\gamma_n = 2.25 for n6n \ge 6. The result improves that of Guo et al. Our proof depends on an auxiliary result in geometry asserting that the area of an nn-sided convex polygon is bounded by γn\gamma_n times the maximum area of the triangle lying inside the polygon.

Keywords

Cite

@article{arxiv.1402.0917,
  title  = {Perturbing eigenvalues of non-negative matrices},
  author = {Chi-Kwong Li and Yiu-Tung Poon and Xuefeng Wang},
  journal= {arXiv preprint arXiv:1402.0917},
  year   = {2014}
}

Comments

17 pages

R2 v1 2026-06-22T03:01:34.995Z