fluctuation fields

If the hydrodynamic limit is the law of large numbers for the density of an interacting particle system, the fluctuation field is its central limit theorem: it describes the small, random deviations of the empirical density from the deterministic hydrodynamic profile, correctly rescaled so they neither vanish nor blow up. The question it answers is the natural next one after hydrodynamics: how big are the fluctuations around the macroscopic profile, what is their law, and what equation do they satisfy? Just as the CLT refines the LLN for sums, the fluctuation field refines the hydrodynamic limit for densities.

The fluctuation field is the centred, rescaled empirical measure. Around equilibrium nu_rho one defines Y^N_t (test function phi) = (1/N^(d/2)) sum_x phi(x/N) ( eta_t(x) - rho ), the order-N^(d/2) normalisation being exactly the square-root scaling that keeps the field nondegenerate (it is a sum of roughly N^d weakly dependent terms). The theorem is that Y^N_t converges, as a distribution-valued (generalized-function) process, to a Gaussian field — an Ornstein-Uhlenbeck process taking values in a space of distributions — solving a linear stochastic PDE of the form dY = A Y dt + B dW, where A is the linearization of the hydrodynamic operator (for SSEP, A = (1/2) Laplacian) and dW is a space-time white noise whose intensity is fixed by the compressibility / variance of the equilibrium measure (a fluctuation-dissipation relation: the noise strength is tied to chi(rho) = rho(1 - rho) for exclusion). So the equilibrium fluctuations of SSEP are described by the linear stochastic heat equation, a generalized Ornstein-Uhlenbeck process. Establishing this requires tightness of the fields in a suitable distribution space (Prokhorov on path space) plus identification of the limit via a martingale problem — the Holley-Stroock approach — and the Boltzmann-Gibbs principle, which says that fluctuations of any local function can be replaced, to leading order, by a multiple of the density fluctuation.

Fluctuation fields matter because they carry the physical information the deterministic limit discards — variances, correlation lengths, transport coefficients, response to perturbation — and they are where universality first becomes visible. The crucial honesty point is the dependence on symmetry and scale. In the symmetric (diffusive) case the equilibrium fluctuations are Gaussian and governed by a LINEAR stochastic PDE, even though the original dynamics is interacting; this is the generic 'Edwards-Wilkinson' behaviour. But for asymmetric dynamics at the right (KPZ) scaling the fluctuations are emphatically NOT Gaussian and NOT described by a linear equation: the relevant object is the nonlinear KPZ / stochastic Burgers equation, with t^(1/3) scaling and Tracy-Widom statistics. So 'the fluctuation field is a Gaussian Ornstein-Uhlenbeck process' is true for equilibrium symmetric systems but FAILS for the asymmetric KPZ regime — the breakdown of the Gaussian, linear picture is precisely the entry point to KPZ universality. Note also these are distribution-valued (white-noise-driven) objects: they live in negative Sobolev spaces, not as ordinary functions.