Critical Phenomena and Singularities

Core Idea

Near a critical point (T_c, P_c), thermodynamic quantities diverge: heat capacity, compressibility, and susceptibilities all diverge with characteristic power laws. The correlation length ξ → ∞, signaling long-range order correlations. These singularities cannot be predicted by mean-field theory alone and require understanding of fluctuations at all length scales.

Explainer

Near a phase transition, you've seen that free energy and thermodynamic quantities change continuously or discontinuously depending on the transition order. Critical phenomena are about what happens *exactly* at the transition between phases — and the answer is strange: quantities don't just change, they diverge or vanish following power laws that are the same across wildly different physical systems.

The central quantity is the correlation length ξ, which measures how far apart two parts of a system can be and still influence each other statistically. Far from the critical temperature Tc, ξ is finite — a local fluctuation decays exponentially with distance. As you approach Tc from either side, ξ grows: ξ ~ |T − Tc|^{−ν}. At exactly Tc, ξ diverges — the entire system is correlated at all length scales simultaneously. This scale invariance is the defining feature of a critical point and is why power laws, which have no characteristic scale, appear everywhere.

The diverging correlation length forces other thermodynamic quantities to diverge too. The heat capacity C ~ |T − Tc|^{−α} diverges because fluctuations at every scale contribute to the energy. The susceptibility (for a magnet, how easily magnetization responds to an applied field) χ ~ |T − Tc|^{−γ} diverges because the system is maximally sensitive to perturbations — a tiny field can flip correlations across the entire sample. These exponents α, γ, ν are called critical exponents. The remarkable fact of universality is that they depend only on the symmetry of the order parameter and the spatial dimensionality of the system — not on microscopic details. Water near its liquid-gas critical point and iron near its Curie temperature have the same critical exponents, even though their microscopic physics is completely different.

Mean-field theory, which you've used to approximate thermodynamic behavior, predicts specific exponent values (γ = 1, ν = 1/2) that work well in high dimensions but fail badly in two and three dimensions. The failure is physical: mean-field theory ignores fluctuations by replacing all interactions with an average field. Near the critical point, fluctuations at all scales are enormous — ξ → ∞ means there is no short-distance cutoff, no scale you can ignore. The breakdown of mean-field theory is a signal that the critical point demands a framework that handles all length scales simultaneously, which is exactly what the renormalization group provides.