WorksMathematics
Convergence rate of observables in the Marcus-Lushnikov process toward Smoluchowski with singular homogeneous kernels
Abstract
We study the approximation of the Smoluchowski coagulation equation by the Marcus–Lushnikov process with initial particles for the singular family of kernels . For regular mean-field particle systems, pathwise fluctuations are naturally expected on the scale , whereas sufficiently smooth observables may exhibit a weak error of order after averaging. The singularity of at the origin makes it unclear whether this second-order mechanism survives. We prove that it does: for a class of sufficiently regular nonlinear functionals of the empirical measure, we establish a weak error of order on finite time intervals. The main mechanism is a regularization produced by the singular dynamics itself. Small clusters coagulate at increasingly large rates, which generates exponential moments in the singular variable and yields a rapid escape from the singularity at zero. We exploit this phenomenon through a Wasserstein-type distance adapted to , together with weighted total variation estimates used to construct and control the derivative of the Smoluchowski flow.