Orbit measures, random matrix theory and interlaced determinantal processes (Q974776): Difference between revisions

This paper is concerned with the study of eigenvalue distribution for a large class of random matrix ensembles (including the classical Gaussian and Laguerre unitary ensembles). The main restriction is that the ensemble is required to be invariant under unitary conjugation, as this allows the author to view the matrix measure as a so-called orbit measure, i.e. the image of the Haar measure on orbits under the adjoint action. Using Kirillov's orbit method, this orbit measure is then obtained as a limit of weighted sums of irreducible representations of high dimension. While this paper makes heavy use of representation theory, every effort has been made by the author to first motivate its use in a random matrix theory context. Indeed, the first part of the paper reframes a series of (separate) random matrix problems into the algebraic language that is more thoroughly discussed in the second part. For instance, the joint distribution of the main minors of Hermitian matrices in a unitary-invariant ensemble is studied via the interlacing that occurs among the eigenvalues of successive minors, which leads to the combinatorial structure of Gelfand-Tsetlin polytopes and crystal graphs.