WorksMathematics
Comparing Lower and Upper Spectral Gaps of Nonnegative Symmetric Matrices
Abstract
We establish sharp estimates comparing the lower and upper spectral gaps of nonnegative symmetric matrices. Combining these comparisons with a multiplicity argument in the spirit of Li and Liu, we improve and extend earlier spectral gap comparison bounds of Saha and Li and Liu, which provide quantitative refinements of an observation of Breuillard, Green, Guralnick and Tao relating combinatorial expansion to two-sided spectral expansion. Our graph estimates hold for nonbipartite even-walk-regular graphs, which covers the classes of graphs, including vertex-transitive graphs, (twisted) Cayley graphs, and (twisted) Cayley sum graphs, investigated in previous work. Under an additional eigenmap assumption, we also establish quantitative two-sided comparisons between the lower and upper spectral gaps. In particular, these comparisons apply to walk-regular graphs.