the Ising model

Strip magnetism down to its barest bones: a lattice of little arrows, each pointing only up or down, with each arrow preferring to align with its nearest neighbors. That is the Ising model, the fruit fly of statistical mechanics. Absurdly simple, it nonetheless captures a genuine phase transition — a temperature below which the arrows spontaneously line up into a magnet — and it has taught physicists more about collective behavior than almost any other model.

The energy of a configuration is H = -J sum over nearest-neighbor pairs of s_i s_j - h sum over sites of s_i, where each spin s_i = plus or minus 1, J > 0 favors alignment (ferromagnetic coupling), and h is an external field. The competition is between energy, which wants the spins ordered, and entropy, which at high temperature wants them randomized. In one dimension Ising showed (1925) there is no ordered phase at any finite temperature — a single domain wall costs only finite energy but gains unbounded entropy, so order is always destroyed. In two dimensions, Onsager's celebrated exact solution (1944) proves a phase transition at a nonzero T_c and yields non-mean-field exponents beta = 1/8 and gamma = 7/4. In three dimensions no exact solution exists, and the exponents come from the renormalization group and Monte Carlo simulation. The model has an exact Z2 (up-down) symmetry, s to -s, which is spontaneously broken in the ordered phase.

The Ising model is the workhorse for testing every big idea in critical phenomena — spontaneous symmetry breaking, universality, the renormalization group — and, through simple relabelings, it maps onto the lattice gas, binary alloys, and even models of neurons and social opinion. An honest caution: its most famous lesson is dimension-dependent. The 1D result (no transition at finite T) is not a failure of the model but a real theorem, and it warns that lowering the dimension can destroy order that survives in higher dimensions.