Criticality

The Ising Model

A grid of two-state spins and one temperature knob — the simplest system in physics with a genuine phase transition, exactly solved by Onsager in 1944.

Drag the temperature slider through ≈2.27 (press set T = Tc to snap there) and watch order appear or dissolve. Amber and teal are the two spin directions; click or drag to paint a domain. Start cold and heat up, or start hot and cool down to watch domains coarsen and compete.

What you're seeing

A grid of tiny magnets — spins — each pointing one of just two ways: up (amber) or down (teal). Every spin would "prefer" to point the same way as its four immediate neighbours, because that arrangement has the lowest energy. Fighting that preference is temperature: heat jostles spins and randomly flips them. The whole picture is the tug-of-war between these two tendencies, order versus agitation, playing out one spin at a time.

The striking thing is what happens as you turn the temperature knob. Cool the grid and large same-coloured domains sweep across it — the system has spontaneously chosen a direction to magnetize. Heat it and the domains shatter into a fizzing salt-and-pepper of no net direction. And in a narrow window around one special temperature, Tc ≈ 2.27, you see neither: domains of every size, from a few pixels to nearly the whole screen, churning and reforming. That scale-free restlessness at the critical point is the heart of this entry.

The rule

Each spin s is either +1 or −1. The energy of the whole lattice is a sum over neighbouring pairs, E = −J · Σ s_i·s_j with coupling J = 1: aligned neighbours (s_i·s_j = +1) lower the energy, opposed ones raise it. The system seeks low energy but is bathed at temperature T, so it samples configurations with the Boltzmann weight e^(−E/T) (Boltzmann's constant set to 1). High T flattens that weight — every configuration is roughly equally likely, so order washes out; low T sharply favours the aligned ground states.

We sample with Metropolis Monte Carlo. Repeatedly: pick a random site, compute the energy change ΔE = 2·J·s_i·(sum of its 4 neighbours) that flipping it would cause, and flip it with probability min(1, e^(−ΔE/T)). Energetically favourable flips (ΔE ≤ 0) always happen; unfavourable ones happen sometimes, more often when it is hot. One sweep is N such attempts (N = number of cells). The grid wraps at its edges (a torus), so there are no walls, and only five distinct values of ΔE are possible, which we precompute per temperature for speed.

Why it matters

The Ising model is the simplest system that has a real phase transition. Below the critical temperature it is magnetized; above it is not; and the change between these is genuinely sharp in the infinite-lattice limit. Crucially, the magnetized state has to pick a direction — up or down — even though the rule treats the two perfectly symmetrically. This is spontaneous symmetry breaking, the same idea that runs through superconductivity, the Higgs mechanism, and the early universe, in its most strippeddown form.

Lars Onsager exactly solved the 2D model in 1944, deriving the critical temperature Tc = 2 / ln(1 + √2) ≈ 2.269 in closed form — one of the landmark calculations of twentieth century physics. And the model is the cleanest illustration of universality: near Tc the microscopic details stop mattering, and the behaviour is governed by a handful of critical exponents shared across systems as different as magnets, fluids, and alloys. Wildly unlike materials fall into the same "universality class" purely by their dimensionality and symmetry. Few models in physics have been more important or more studied.

In the wild

The Ising transition is not just a metaphor — it is the textbook account of a real effect:

Try this

References

  1. Ising, E. (1925). "Beitrag zur Theorie des Ferromagnetismus." Zeitschrift für Physik 31, 253–258. (The original 1D model, which famously has no phase transition.)
  2. Onsager, L. (1944). "Crystal Statistics. I. A Two-Dimensional Model with an Order-Disorder Transition." Physical Review 65, 117–149. (The exact solution and the critical temperature Tc.)
  3. Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., Teller, A. H., Teller, E. (1953). "Equation of State Calculations by Fast Computing Machines." Journal of Chemical Physics 21(6), 1087–1092. (The Monte Carlo algorithm used by this simulation.)
  4. Brush, S. G. (1967). "History of the Lenz–Ising Model." Reviews of Modern Physics 39(4), 883–893. (How the model came to be and why it bears Ising's name.)