Distributional Limits for Eigenvalues of Graphon Kernel Matrices

Abstract

We study eigenvalue fluctuations of kernel and adjacency matrices in dense graphon-based random graph models, for modes associated with simple, isolated nonzero eigenvalues of the graphon integral operator. Our results require no smoothness or Lipschitz continuity. They include degenerate non-leading modes whose normalized eigenfunctions have constant modulus and change sign between $-1$ and $+1$ on measurable sets. Continuous graphons exclude this possibility: their nonzero eigenfunctions have continuous representatives, so constant modulus forces a constant eigenfunction, which corresponds to the largest eigenvalue of a nonnegative graphon. Nevertheless, non-leading degeneracy already occurs at the second eigenvalue of a two-block model with equal blocks. The limiting behavior is governed by the variance of the squared population eigenfunction. When this variance is positive, the empirical eigenvalue, divided by $n-1$ and centered at its population value, has a Gaussian limit under $\sqrt{n}$ scaling. When the variance vanishes, the first-order fluctuation disappears, and the centered unnormalized kernel eigenvalue converges to a weighted series of centered chi-square variables. For adjacency matrices, Bernoulli edge noise is negligible at the $\sqrt{n}$ scale of the unnormalized eigenvalues. In the degenerate regime, after the same deterministic centering used for the kernel eigenvalue, the adjacency eigenvalue converges to the same weighted chi-square series plus a deterministic shift and a Gaussian term independent of that series. The proofs combine spectral localization, exact Schur identities, finite-mode approximations, and Hoeffding decompositions.