Abstract
The propagation of photons in a Monte Carlo (MC) simulation can be broken down in a few standard steps: propagation (hop), absorption (drop), scattering (spin), and the Fresnel phenomena of transmission and reflection at boundaries, where the refractive index changes.1 When using a rectangular grid structure, the representation of curved, smooth, or slanted boundaries is discretized into voxels, also referred to as a Cartesian grid.2 The drawbacks are not only aesthetic in nature but also represent important deviations from the modeled structure. The use of a finer grid can partially minimize the differences between the voxel-based representation and the modeled geometry. This works particularly well to improve the precision of the propagation, absorption, and scattering steps; the trade-off being higher memory requirements to store the information and slower simulation performances due to the larger number of elements involved. The deviation addressed in this work is the representation of the surface boundary between mismatched materials, having different indices of refractivity, being simplified to the facets of the voxels located at the boundary. No matter how small the voxels, the normal angles to those boundary facets are intrinsically limited to being one of the facets of these voxels. This is of particular importance when dealing with reflection and transmission using Fresnel equations because the normal angle to the surface not only plays a key role in determining the amount of light that is to be reflected or transmitted but also determines the altered angles for the trajectories of these photons during the propagation process.

This publication has 25 references indexed in Scilit: