A random matrix model for the Gaussian distribution (Q858123)

It is shown in this paper that each mixture \(\mu\) of the Wigner semicircle law admits a random matrix model, i.e., one can construct a sequence \((X^d)_{d\geq 1}\) of \(d\times d\)-random matrices such that the empirical distributions of their eigenvalues tend to \(\mu\) in some way. As the standard normal distribution can be interpreted as mixture of the Wigner semicircle law with some gamma distribution, the author in particular obtains random matrix models for normal distributions. The random matrix models belonging to such mixtures are constructed in a canonical way by using corresponding mixtures of Gaussian unitary ensembles.