A 3D radiative transfer framework: II. line transfer problems

TL;DR

This paper develops a comprehensive 3D radiative transfer framework that includes line and continuum scattering, demonstrating its accuracy and scalability for complex astrophysical simulations.

Contribution

It extends previous methods to include both line and continuum transfer in 3D, with an operator splitting technique and parallelization for large-scale computations.

Findings

3D results agree well with 1D tests despite lower interior resolution.

The code scales efficiently to many processors, enabling large-scale simulations.

The framework is suitable for realistic 3D radiative transfer in astrophysics.

Abstract

Higher resolution telescopes as well as 3D numerical simulations will require the development of detailed 3D radiative transfer calculations. Building upon our previous work we extend our method to include both continuum and line transfer. We present a general method to calculate radiative transfer including scattering in the continuum as well as in lines in 3D static atmospheres. The scattering problem for line transfer is solved via means of an operator splitting (OS) technique. The formal solution is based on a long-characteristics method. The approximate $\Lambda$ operator is constructed considering nearest neighbors {\em exactly}. The code is parallelized over both wavelength and solid angle using the MPI library. We present the results of several test cases with different values of the thermalization parameter and two choices for the temperature structure. The results are directly compared to 1D spherical tests. With our current grid setup the interior resolution is much lower in 3D than in 1D, nevertheless the 3D results agree very well with the well-tested 1D calculations. We show that with relatively simple parallelization that the code scales to very large number of processors which is mandatory for practical applications. Advances in modern computers will make realistic 3D radiative transfer calculations possible in the near future. Our current code scales to very large numbers of processors, but requires larger memory per processor at high spatial resolution.