Solving radiative transfer with line overlaps using Gauss Seidel algorithms

TL;DR

This paper enhances radiative transfer modeling by implementing a Gauss-Seidel algorithm in spherical geometry that accounts for hyperfine line overlaps, significantly improving convergence speed and accuracy in molecular line intensity predictions.

Contribution

It introduces a Gauss-Seidel iterative scheme for radiative transfer in spherical geometry that includes hyperfine line overlaps, improving computational efficiency and modeling accuracy.

Findings

Gauss-Seidel algorithm converges 2-4 times faster than Jacobi.

Line overlap significantly affects line intensity predictions.

Including line overlap improves model accuracy for molecular lines.

Abstract

The improvement in observational facilities requires refining the modelling of the geometrical structures of astrophysical objects. Nevertheless, for complex problems such as line overlap in molecules showing hyperfine structure, a detailed analysis still requires a large amount of computing time and thus, misinterpretation cannot be dismissed due to an undersampling of the whole space of parameters. We extend the discussion of the implementation of the Gauss--Seidel algorithm in spherical geometry and include the case of hyperfine line overlap. We first review the basics of the short characteristics method that is used to solve the radiative transfer equations. Details are given on the determination of the Lambda operator in spherical geometry. The Gauss--Seidel algorithm is then described and, by analogy to the plan--parallel case, we see how to introduce it in spherical geometry. Doing so requires some approximations in order to keep the algorithm competitive. Finally, line overlap effects are included. The convergence speed of the algorithm is compared to the usual Jacobi iterative schemes. The gain in the number of iterations is typically factors of 2 and 4 for the two implementations made of the Gauss--Seidel algorithm. This is obtained despite the introduction of approximations in the algorithm. A comparison of results obtained with and without line overlaps for N2H+, HCN, and HNC shows that the J=3-2 line intensities are significantly underestimated in models where line overlap is neglected.