GPU accelerated electric field Monte Carlo simulation of light propagation in turbid media using a finite-size beam model

Abstract

An electric field Monte Carlo (EMC) simulation directly traces the complex electric field vectors in multiple scattering and estimates the electric field in a preferred direction. The full vectorial nature of EMC makes it a powerful and flexible tool to simulate the coherence and polarization phenomena of light. As a numerical method, EMC needs to launch a large number of photons to achieve an accurate result, making it time-consuming. Previously, EMC did not account for the beam size. Because of the stochastic character of the instantaneous electric field in the simulation, the convolution method alone is unsuitable for the Monte Carlo simulation of photon energy for a beam with a finite size. It is necessary to launch photons from all possible locations to simulate a finite-size beam, which results in a significant increase in the computational burden. In order to accelerate the simulation, a parallel implementation of the electric field Monte Carlo simulation based on the compute unified device architecture (CUDA) running on a graphics processing unit (GPU) is presented in this paper. Our program, which is optimized for Fermi architecture, is able to simulate the coherence phenomenon of a finite-size beam normally incident on turbid media. A maximum speedup of over 370x is achieved with a GTX480 GPU, compared with that obtained using an Intel i3-2120 CPU.