On eigenvalues of the sum of two random projections (Q691018): Difference between revisions

In the paper, a general problem of the distribution of eigenvalues is considered for a case of the matrix \(X= A + UBU^*\), where \(A\) and \(B\) are two complex Hermitian matrices and \(U\) is a uniformly-distributed unitary matrix. In more detail the universality of eigenvalue distribution of \(X\) is investigated in the case when \(A\) and \(B\) denoted as \(P\) and \(Q\) are random orthogonal projections. The behavior of eigenvalues of \(P + Q\) is much easier to analyze than the general case \(A + B\) using the relation of eigenvalues between \(P + Q\) and \(PQ\). It is shown in the large size limit of matrices that the local behavior of eigenvalues is described by the sine kernel in the bulk and by the Bessel or Airy kernel depending on the rank of \(P\) and \(Q\). The case when the ranks of \(P\) and \(Q\) are both approximately \(N/2\) and the local behavior of the eigenvalues of \(P + Q\) is not universal is also analyzed.