GPU-accelerated repulsive particle swarm optimization for parameter retrieval of single sphere from angularly distributed scattered light

Abstract

Multi-angle light scattering has been proved to be a versatile approach to characterizing a single particle in a non-contact manner. A fast and accurate inverse scattering process is required to precisely retrieve particle parameters from scattering patterns. In this paper, we report a repulsive particle swarm optimization algorithm to solve this inverse scattering problem for both non-absorbing and weakly absorbing particle characterization. Parallel computing of Mie scattering by different particles is conducted on a graphic processing unit (GPU) to speed up the optimization. To reduce the computational complexity in a single thread, the special functions in the Mie scattering formula are evaluated by iteration from lower to higher orders. Simulation results based on modeled data are provided, demonstrating that the proposed evolutionary algorithm is capable of precisely retrieving particle parameters. Further analysis on noise influences based on the proposed algorithm is presented, which gives indications to establishment of the metrology system.