Classical nucleation theory has one of the cleanest recipes in physics. A new phase forms when a fluctuation is large enough that the bulk free energy gained by the interior outweighs the surface energy cost of the boundary. The critical radius is the crossover point. Below it, the droplet dissolves. Above it, it grows. Surface tension times area versus chemical potential times volume. Two competing terms, one outcome.
Maire shows this recipe breaks in hyperuniform fluids.
Hyperuniform systems suppress density fluctuations at large scales — the variance in particle number within a region grows more slowly than the region's volume. Normal fluids have proportional scaling; hyperuniform fluids are anomalously quiet. This isn't exotic — certain active matter systems, driven out of equilibrium, spontaneously become hyperuniform.
In these systems, nucleation doesn't separate into surface and volume contributions. The quasi-potential that governs the probability of forming a critical droplet can't be decomposed into the classical two terms. The recipe calls for surface tension and bulk driving force, measured separately, combined additively. In a hyperuniform fluid, these aren't separable. The fluctuation that creates the droplet is simultaneously constrained at all scales by the suppressed density variance, coupling what classical theory treats as independent ingredients.
Including capillary waves — fluctuations of the interface itself — reveals the deeper break. In equilibrium, capillary waves satisfy detailed balance: each wave mode is equally likely to grow or shrink. In the hyperuniform active fluid, this symmetry is broken by nonreciprocal dynamics. The interface has a preferred direction of fluctuation. The capillary wave spectrum carries the signature of broken time-reversal symmetry.
The practical consequence: classical nucleation theory gives you a number — the nucleation rate — from measurable inputs (surface tension, supersaturation, temperature). In hyperuniform active systems, those inputs don't determine the rate. You need the full nonequilibrium quasi-potential, which encodes information that the classical framework doesn't even have variables for.
The recipe was never wrong for the systems it was designed for. It's that some systems don't have the ingredients the recipe assumes exist as separate quantities.