Tiny fluctuations of the averaging process around its degenerate steady state

We analyze nonequilibrium fluctuations of the averaging process on dε, a continuous degenerate Gibbs sampler running over the edges of the discrete d-dimensional torus. We show that, if we start from a smooth deterministic nonflat interface, recenter, blow-up by a nonstandard CLT-scaling factor θε = ε−(d/2+1), and rescale diffusively, Gaussian fluctuations emerge in the limit ε→0. These fluctuations are purely dynamical, zero at times t = 0 and t =∞and nontrivial for t ∈ (0,∞). We fully determine the correlation matrix of the limiting noise, nondiagonal as soon as d ≥ 2. The main technical challenge in this stochastic homogenization procedure lies in a LLN for a weighted space-time average of squared discrete gradients. We accomplish this through a Poincaré inequality with respect to the underlying randomness of the edge updates, a tool from Malliavin calculus in Poisson space. This inequality, combined with sharp gradients’ second moment estimates, yields quantitative variance bounds without prior knowledge of the limiting mean. Our method avoids higher (e.g., fourth) moment bounds, which seem inaccessible with the present techniques.