A necessary and sufficient condition for edge universality at the largest singular values of covariance matrices (Q1661567)

The authors prove a necessary and sufficient condition for edge universality of sample covariance random matrices. More precisely, they consider random matrices of the form \(Q:= (TX)(TX)^*\), where \(X\) is a random \(M_2\times N\)-matrix with entries \(X_{ij} = N^{-1/2} q_{ij}\) such that \(q_{ij}\) are i.i.d. random variables with zero mean and unit variance, and \(T\) is a deterministic \((M_1\times M_2)\)-matrix such that \(T^*T\) is diagonal and satisfies some additional minor assumption. The authors are interested in the asymptotic distribution of the largest eigenvalue \(\lambda_1(Q)\) when \(M:=\min \{M_1,M_2\}\) and \(N\) both tend to \(\infty\) in such a way that the ratio \(N/M\) is finite and positive. It is known that the empirical spectral distribution of \(Q\) converges to the Marchenko-Pastur law with certain rightmost edge \(m\). The authors prove that \(N^{2/3} (\lambda_1 - m)\) converges weakly to the Tracy-Widom law if and only if the entries of \(X\) satisfy the following tail condition: \[ \lim_{s\to\infty} s^4 \mathbb P[|Q_{11}|\geq s] = 0. \]