Random Matrix Theory
Random Matrix Theory studies the statistical properties of eigenvalues and eigenvectors of matrices with random entries, revealing universal patterns that appear across mathematics, physics, and beyond—from nuclear physics to the zeros of the Riemann zeta function.
Fundamental Ensembles
GUE
Gaussian Unitary Ensemble
β = 2 (Complex Hermitian)
Time-reversal breaking
GOE
Gaussian Orthogonal Ensemble
β = 1 (Real Symmetric)
Time-reversal invariant
GSE
Gaussian Symplectic Ensemble
β = 4 (Quaternionic)
Spin-orbit coupling
Joint Probability Density
For the β-ensemble, the joint eigenvalue density is:
$$P(\lambda_1, ..., \lambda_n) \propto \prod_{iGaussian Unitary Ensemble (GUE)
Wigner Semicircle Law
Eigenvalue Density
Empirical vs Theoretical
Wigner's Theorem
For large N, the eigenvalue density converges to:
$$\rho(x) = \frac{1}{2\pi}\sqrt{4-x^2}, \quad |x| \leq 2$$Tracy-Widom Distribution
Largest Eigenvalue Statistics
The rescaled largest eigenvalue follows the Tracy-Widom distribution, a universal limit appearing in diverse contexts from growth processes to traffic flow.
Connection to Number Theory
Montgomery-Odlyzko Conjecture
The pair correlation of Riemann zeta zeros matches the pair correlation of GUE eigenvalues!
Remarkable Connection
The statistical distribution of gaps between consecutive zeros of ζ(s) on the critical line Re(s) = 1/2 follows the same law as eigenvalue spacings in GUE matrices.
Level Spacing Distributions
P(s) = e^{-s}
P(s) ≈ (πs/2)e^{-πs²/4}
Marchenko-Pastur Law
For rectangular matrices with aspect ratio γ = n/m:
Correlation Functions
The n-point correlation functions exhibit determinantal structure through the Christoffel-Darboux kernel.
Revolutionary Applications
Quantum Chaos
Energy level statistics in chaotic quantum systems follow RMT predictions universally.
Wireless Communications
MIMO channel capacity analysis relies heavily on RMT for large antenna arrays.
Finance
Portfolio optimization and risk assessment using eigenvalue cleaning techniques.
Machine Learning
Understanding neural network training dynamics through the lens of RMT.
Foundational Papers
Classification of random matrix ensembles by symmetry.
Discovery of the Tracy-Widom distribution.
Numerical evidence for GUE statistics in Riemann zeros.
Modern developments in free probability approach to RMT.
Leading Research Centers
- MIT - Alan Edelman's group, numerical RMT
- UC Davis - Craig Tracy and Harold Widom
- University of Michigan - Jinho Baik, integrable systems
- IHES Paris - Mathematical physics connections
- KTH Stockholm - Kurt Johansson's group