Coagulation kinetics beyond mean field theory using an optimised Poisson representation

TL;DR

This paper develops an advanced Poisson representation method to model coagulation kinetics beyond mean field theory, addressing small population effects and numerical instabilities with optimized gauge transformations.

Contribution

It introduces an optimized gauge transformation within the Poisson representation framework to improve numerical stability and efficiency in coagulation kinetics modeling.

Findings

Analytical results match numerical simulations.

Gauge transformation reduces numerical instabilities.

Optimized gauge minimizes statistical noise.

Abstract

Binary particle coagulation can be modelled as the repeated random process of the combination of two particles to form a third. The kinetics can be represented by population rate equations based on a mean field assumption, according to which the rate of aggregation is taken to be proportional to the product of the mean populations of the two participants. This can be a poor approximation when the mean populations are small. However, using the Poisson representation it is possible to derive a set of rate equations that go beyond mean field theory, describing pseudo-populations that are continuous, noisy and complex, but where averaging over the noise and initial conditions gives the mean of the physical population. Such an approach is explored for the simple case of a size-independent rate of coagulation between particles. Analytical results are compared with numerical computations and with results derived by other means. In the numerical work we encounter instabilities that can be eliminated using a suitable 'gauge' transformation of the problem [P. D. Drummond, Eur. Phys. J. B38, 617 (2004)] which we show to be equivalent to the application of the Cameron-Martin-Girsanov formula describing a shift in a probability measure. The cost of such a procedure is to introduce additional statistical noise into the numerical results, but we identify an optimised gauge transformation where this difficulty is minimal for the main properties of interest. For more complicated systems, such an approach is likely to be computationally cheaper than Monte Carlo simulation.