A random matrix is a matrix with entries drawn from probability distributions, used to model complex systems and spectra.A random matrix is a matrix whose elements are random variables with a given joint distribution. The best-known are the Wigner Gaussian ensembles: orthogonal (GOE), unitary (GUE), and symplectic (GSE), differing in symmetry under time reversal. The main result is the Wigner distribution for spacings between eigenvalues, which shows level repulsion: the probability of finding two very close values is small.
In the 1950s, Eugene Wigner, while investigating resonance spectra of heavy nuclei, proposed replacing unknown elements of the Hamiltonian matrix with random numbers. Together with Dyson, Mehta, and others, he developed the statistical theory of spectra. Later, a deep connection emerged with eigenvalue distributions in numerical methods, number theory (distribution of zeros of the Riemann zeta function), and growth of random surfaces.
The eigenvalues of a random matrix behave like a distribution of charged particles on a wire repelling each other logarithmically. As a result, they are spaced unevenly: the probability of a small spacing s between them is proportional to s^β (β is the repulsion index). The spacing distribution is exactly described by the Wigner formula for GOE: p(s) = (πs/2) exp(−π s²/4). For comparison, a Poisson process (independent levels) would yield p(s) = exp(−s).
Random matrices are used in signal processing: they separate important data from noise in radar and medical tomographs.
The distribution of gaps between consecutive buses in some cities also resembles the Wigner distribution, because drivers tend to avoid bunching.
| a_0, a_1, ..., a_{n- | variable | First row entries of the circulant matrix |
| n | variable | Matrix size (positive integer) |
| det | operator | Determinant of a square matrix — determinant |
| Circ | operator | Constructs a circulant matrix from its first row — circulant matrix |





