Wigner Semicircle
Generate a large N × N symmetric matrix with random Gaussian entries. Diagonalize it. The eigenvalues, properly rescaled, distribute themselves into a perfect semicircle. Eugene Wigner discovered this in 1955 studying heavy nuclei. It's the universal law of random matrix theory — the same shape appears in nuclear physics, quantum chaos, number theory, financial correlations, and the eigenvalues of social networks.
why a semicircle (and what it tells us about nuclei and primes)
Take a symmetric N × N matrix with entries drawn iid from N(0, σ²) (zero mean, variance σ²). Diagonalize. The eigenvalue density, rescaled by √N, satisfies:
ρ(λ) = (1 / (2π σ²)) √(4Nσ² − λ²) for |λ| ≤ 2σ√N
This is the Wigner semicircle law. Note what's not here: nothing about the specific distribution of entries (beyond mean and variance). The shape is universal.
Wigner's motivation: nuclei are systems of ~100 nucleons interacting via complicated forces. The exact Hamiltonian is intractable. Wigner asked: what if we replace it with a random Hamiltonian and ask about eigenvalue statistics? The semicircle emerged. The local eigenvalue spacings — not the overall density — turn out to match measured nuclear energy level spacings stunningly well.
Where the same statistics appear:
• Nuclear physics: heavy nuclei eigenvalue spacings
• Quantum chaos: spectra of chaotic billiards (e.g., Sinai billiard) follow GOE statistics
• Riemann zeros: numerical evidence (Odlyzko 1987) that ζ-zero spacings on the critical line match the GUE random-matrix distribution. This is the strongest hint of a deep, undiscovered connection between number theory and quantum physics — the Hilbert-Pólya conjecture suggests the zeros are eigenvalues of some unknown self-adjoint operator.
• Wireless networks: capacity in multi-antenna systems
• Financial correlations: most eigenvalues of correlation matrices match the random null hypothesis — only a few "signal" eigenvalues correspond to real market modes
Random matrix theory is one of the most surprising universality classes in mathematics: statistics first studied for nuclear energy levels turn out to govern the local distribution of the Riemann zeta zeros. No one knows why.