Universality in several-matrix models via approximate transport maps (Q518167)

Many recent papers have paid attention to the question of universality of the distribution of the local statistics of Wigner matrices. The aim of this sizeable article is to provide new universality results for general perturbative several-matrix models. The authors prove that if \(X_1^N,\ldots,X_d^N\) are independent \(N\times N\) matrices in the Gaussian Unitary Ensemble (GUE), if \(B_1^N,\ldots, B_m^N\) are Hermitian deterministic matrices and if \(p\) is a polynomial in \(d+m\) variables which is a perturbation of \(x_1\), then the eigenvalues of \(p(X_1^N,\ldots,X_d^N, B_1^N,\ldots,B_m^N)\) fluctuate as the eigenvalues of \(X_1^N\). In particular, when \(p\) depends only on \(d\) variables and \(p(x_1,\ldots,x_d)=x_1+ \varepsilon Q((x_1,\ldots,x_d)\) with \(Q\) self-adjoint, then the properly normalized fluctuations of the eigenvalues of \(p(X_1^N,\ldots,X_d^N)\) follow the sine-kernel distribution inside the bulk and the Tracy-Widom law at he edges of the support of the semi-circle law. The result derives from an universality property for unitarily invariant matrices interacting via a potential, which is obtained by finding a map from the law of the eigenvalues of independent GUE or GOE matrices to a probability measure that approximates the matrix model. As it is necessary to treat quadratic statistics, a large \(N\)-expansion for integrals over the unitary and orthogonal group is included in the proof.