Critical Phenomena and Universality

The most surprising result in the study of phase transitions is that systems with completely different microscopic physics share identical critical behavior. A ferromagnet near its Curie temperature, a fluid near its liquid-gas critical point, and a binary alloy near its order-disorder transition all exhibit the same power-law exponents, the same scaling functions, the same divergences. These systems have nothing in common at the microscopic level. At the critical point, they are the same system.

The Critical Exponents

A phase transition is characterized by a set of power-law exponents that describe how thermodynamic quantities behave near the critical point. For the Ising model at h = 0:

alpha (specific heat): C ~ |T - T_c|^(-alpha). In 2D, alpha = 0 (logarithmic divergence). In 3D, alpha approximately equal to 0.110.

beta (order parameter): M ~ (T_c - T)^beta for T < T_c. In 2D, beta = 1/8. In 3D, beta approximately equal to 0.3265.

gamma (susceptibility): chi ~ |T - T_c|^(-gamma). In 2D, gamma = 7/4. In 3D, gamma approximately equal to 1.237.

nu (correlation length): xi ~ |T - T_c|^(-nu). In 2D, nu = 1. In 3D, nu approximately equal to 0.6300.

eta (correlation function at T_c): G(r) ~ r^(-(d-2+eta)). In 2D, eta = 1/4. In 3D, eta approximately equal to 0.0363.

delta (order parameter vs. field at T_c): M ~ h^(1/delta). In 2D, delta = 15. In 3D, delta approximately equal to 4.789.

These six exponents are not independent. They are constrained by four scaling relations, proven rigorously within the renormalization group framework:

  • Rushbrooke: alpha + 2beta + gamma = 2
  • Widom: gamma = beta(delta - 1)
  • Fisher: gamma = nu(2 - eta)
  • Josephson (hyperscaling): 2 - alpha = d * nu

These relations reduce six exponents to two independent ones. In two dimensions, both independent exponents are known exactly. In three dimensions, they are known to six significant figures from conformal bootstrap calculations (El-Showk et al., 2014; Kos, Poland, Simmons-Duffin, and Vichi, 2016).

The Universality Hypothesis

The universality hypothesis, articulated by Griffiths (1970) and Kadanoff (1971), states: the critical exponents of a phase transition depend only on the spatial dimension d and the symmetry of the order parameter, not on microscopic details such as the lattice geometry, the specific values of coupling constants, or the chemical identity of the system.

The evidence is overwhelming. The liquid-gas critical point of water, xenon, and carbon dioxide --- three chemically distinct substances --- shares the 3D Ising critical exponents to five decimal places. The ferromagnetic transitions of iron, nickel, and cobalt share them. The binary alloy ordering transition in beta-brass shares them. The superfluid transition of helium-4 has different exponents (because the order parameter has a different symmetry --- complex scalar rather than real scalar) but shares them with the superconducting transition in metals, which has the same symmetry.

This is not an approximate coincidence. The exponent matching between the liquid-gas critical point of SF6 and the ferromagnetic transition of a uniaxial magnet is exact within experimental precision of five significant figures (Sengers and Levelt Sengers, 1986). These are systems that share nothing at the molecular level: one is a gas of fluorinated sulfur molecules, the other is a crystal of aligned magnetic moments.

A universality class is defined by (d, symmetry). The major classes relevant to the models on this site:

  • Ising class (d-dimensional, scalar order parameter with Z_2 symmetry): ferromagnets, liquid-gas, binary alloys, Schelling segregation at the tipping point.
  • XY class (d-dimensional, two-component order parameter with O(2) symmetry): superfluids, superconductors, the 2D XY model with its Kosterlitz-Thouless transition.
  • Heisenberg class (d-dimensional, three-component order parameter with O(3) symmetry): isotropic ferromagnets.
  • Directed percolation (non-equilibrium): epidemic spreading, catalytic reactions, some traffic transitions.

The Renormalization Group Explanation

Kenneth Wilson developed the renormalization group (RG) between 1971 and 1974, providing the theoretical explanation for universality. He received the Nobel Prize in Physics in 1982.

The core idea is coarse-graining. Start with the Ising model at some microscopic coupling J and temperature T. Now “zoom out” --- replace each block of spins with a single effective spin representing the block’s net magnetization. The coarse-grained system is described by new effective couplings J’ and T’. Repeat the process, zooming out further each time.

Under this coarse-graining, the effective couplings flow through parameter space. The flow has fixed points: parameter values where the effective couplings are unchanged by further coarse-graining. The critical point of the original system corresponds to a fixed point of the RG flow.

Near a fixed point, the flow is characterized by eigenvalues of the linearized RG transformation. Some directions in parameter space are relevant: perturbations along these directions grow under coarse-graining, driving the system away from the fixed point. These correspond to the control parameters that must be tuned to reach the critical point (temperature, magnetic field). Other directions are irrelevant: perturbations shrink under coarse-graining and become negligible at large scales. These correspond to the microscopic details --- lattice geometry, exact coupling values, chemical identity --- that do not affect critical behavior.

Universality follows directly. Two systems that flow to the same fixed point under coarse-graining share the same critical exponents, because the exponents are properties of the fixed point, not of the starting system. The microscopic details that distinguish iron from carbon dioxide are irrelevant perturbations that vanish under the RG flow. What remains at the fixed point is the dimension and the symmetry.

Wilson and Fisher’s epsilon expansion (1972) provided a practical calculation method: treat the spatial dimension as a continuous variable d = 4 - epsilon, and compute critical exponents as perturbative series in epsilon. At epsilon = 0 (d = 4), mean-field theory is exact. For epsilon = 1 (d = 3) and epsilon = 2 (d = 2), the series gives good approximations to the exact (2D) and numerically known (3D) exponents.

The upper critical dimension d_c = 4 for the Ising model is the dimension at and above which mean-field theory is exact. Below d_c, fluctuations are important and the RG gives non-mean-field exponents. This is a proven result (Aizenman, 1982, for d >= 4).

Universality Classes Beyond Physics

The concept of universality class has been applied, with varying degrees of rigor, to non-physical systems.

Equilibrium models with Ising symmetry. The Schelling segregation model at its tipping point is in the Ising universality class in the mean-field sense: the bifurcation structure (symmetric state becomes unstable, two asymmetric states emerge) has the same topology as the Ising transition. Whether the spatial Schelling model on a 2D lattice has exactly the 2D Ising critical exponents has not been proven and is unlikely, because the Schelling dynamics are non-equilibrium (there is no Hamiltonian and no detailed balance).

Non-equilibrium transitions. Many non-equilibrium systems with absorbing states (states from which the system cannot escape) exhibit critical behavior in the directed percolation universality class. This includes certain epidemic models (the contact process), catalytic surface reactions, and traffic models at the free-flow to congested transition. Hinrichsen (2000) conjectured that directed percolation is the “default” universality class for non-equilibrium absorbing-state transitions, analogous to Ising universality for equilibrium transitions with Z_2 symmetry.

The practical value. Identifying a system’s universality class tells you its critical exponents, its scaling functions, and its finite-size scaling behavior --- without solving the system’s specific equations. If you can establish that a complex social, biological, or economic system undergoes a phase transition in a known universality class, you import a large body of quantitative predictions. The challenge is establishing the universality class, which requires either exact analysis (rare) or careful measurement of critical exponents (difficult in non-physical systems where data is limited and control parameters cannot be tuned precisely).

Further Reading