Percolation

What you're looking at

What is percolation?

Water through porous rock. Coffee through grounds. A disease spreading through a population. In each case there's a medium with random connections, and the question is: does something flow from one side to the other? Below a certain density, you get small isolated pockets. Above it, a giant connected cluster suddenly spans the whole system. The transition is sharp — not gradual.

The critical threshold

For site percolation on a square lattice, the critical probability is p_c ≈ 0.5927. That's not 0.5 — it's not obvious from the geometry. Below p_c, the largest cluster is tiny relative to the grid. Above it, a single cluster dominates and connects top to bottom. This is a second-order phase transition, the same kind of mathematics that describes magnets losing their magnetism at a critical temperature.

Where this shows up

Forest fires: if tree density is below the threshold, fire stays local. Above it, fire spreads across the entire forest. Epidemics: if the contact rate is too low, outbreaks fizzle. Above the threshold, you get a pandemic. Network resilience: randomly remove nodes from a network — below a critical fraction, it still functions. Above it, it fragments. Material science: mix conductive particles into plastic. Below the threshold, it's an insulator. Above it, current flows.

Universality

Different lattices have different p_c values — a triangular lattice has p_c = 0.5, a honeycomb lattice has p_c ≈ 0.6962. But the shape of the transition is the same. The critical exponents — how cluster size diverges, how the spanning probability sharpens — are universal. They depend on the dimension of the space, not the geometry of the lattice. This is one of the deep results from statistical physics.