The paper studies a large random matrix with a small but deliberate perturbation added to it. The starting point is a square matrix filled with complex random numbers, and the perturbation is a simple rank-one spike, meaning you add a term that points in one fixed direction with some strength. A central question in random matrix theory is whether such a spike is strong enough to pull an eigenvalue out of the bulk, changing the statistical behavior of the system. The threshold where this happens is called the BBP transition, named after Baik, Ben Arous, and Péché, and here it is studied in a non-Hermitian setting where the matrix is not required to be symmetric.
The main findings concern the largest real part among all eigenvalues, which roughly measures how far the spectrum extends to the right in the complex plane. The authors show that as long as the spike strength is below the critical value of one, this extreme eigenvalue behaves exactly as if there were no spike at all. Specifically, after appropriate centering and rescaling, its distribution converges to a Gumbel law, the same limiting distribution that describes the maximum of many independent random variables. This is a spike irrelevance result: the perturbation is too weak to affect the edge statistics. The authors also pin down how fast this convergence happens for the unperturbed case, finding a correction term of order one over the logarithm of the matrix size.
At the critical spike strength of exactly one, the situation is more delicate because this is precisely the boundary where the spike begins to have an effect. There the authors prove an upper bound showing that the rightmost eigenvalue stays within a window of size roughly the fourth root of one over the matrix dimension from the edge of the bulk spectrum. This is established using technical tools from the analysis of resolvents, which are matrices built from the original matrix that encode spectral information, combined with a discretization argument that reduces an infinite geometric problem to a finite one. Together these results give a detailed picture of how the spectrum responds to a weak perturbation near a phase transition.