Large deviations of the largest eigenvalue of supercritical sparse Wigner matrices

Abstract

Consider a random symmetric matrix with i.i.d. entries on and above its diagonal that are products of Bernoulli random variables and random variables with sub-Gaussian tails. Such a matrix will be called a sparse Wigner matrix and can be viewed as the adjacency matrix of a random network with sub-Gaussian weights on its edges. In the regime where the mean degree is at least logarithmic in dimension, the edge eigenvalues of an appropriately scaled sparse Wigner matrix stick to the edges of the support of the semicircle law. We show that, in this sparsity regime, the large deviations upper tail event of the largest eigenvalue of a sparse Wigner matrix with sub-Gaussian entries is generated by either the emergence of a high degree vertex with a large vertex weight or that of a clique with large edge weights. Interestingly, the rate function obtained is discontinuous at the typical value of the largest eigenvalue, which accounts for the fact that its large deviation behaviour is generated by finite rank perturbations. This complements the results of Ganguly and Nam (Probab. Theory Related Fields 184 (2022) 613–679), and Ganguly, Hiesmayr, and Nam (J. Lond. Math. Soc. (2) 110 (2024) Paper No. e12954, 64), which considered the case where the mean degree is constant.