Acceleration of the iterative solver in the discrete dipole approximation: Application to the orientation variation of irregularly shaped particles

We have applied a method of reducing the number of iterations required to solve a system of linear equations in the discrete dipole approximation. This method obtains an initial guess of dipole polarization from those with similar particle characteristics (e.g., the size parameter and refractive index) calculated a priori. If the initial guess is closer to the solution, the number of iterations of the linear equation solution becomes smaller than that calculated with an arbitrary initial value.

This method was applied to various particle orientations using spline interpolation of the initial guess of dipole polarization from orientations calculated a priori.

We studied three types of particle model: an aggregate, a deformed sphere with moderate surface roughness, and a particle with a large number of edges. For the particle with a large number of edges, we propose a new model called the overlapping mixture of multiple tetrahedra (OMMT).

The proposed method is most advantageous for particles with moderate surface roughness (e.g., a deformed sphere), for which the calculation time was reduced to 20–40% of the original calculation time. For OMMT and an aggregate, the computation time was reduced to 30–60% and 40–90%, respectively. The differences in the scattering coefficient, absorption coefficient, intensity and polarization introduced by our method were less than 0.008%, 0.03%, 0.1%, and 0.08%, respectively.

If the light scattering properties vary slowly with the orientation variation, interpolation of the results is more efficient than the proposed method and produces only a small difference in the results. However, the interpolation of the results fails for particles such as BCCA64, for which our proposed method produces more accurate results.