A Comparison of Approaches in Fitting Continuum SEDs

TL;DR

This paper compares the effectiveness and efficiency of pre-calculated databases and simulated annealing in fitting the spectral energy distributions of dusty astrophysical objects, providing guidance for practitioners.

Contribution

It demonstrates the applicability of simulated annealing for SED fitting and suggests a hybrid approach combining both methods for optimal results.

Findings

Simulated annealing finds better solutions than pre-calculated databases.

A hybrid approach leverages the strengths of both methods.

SA offers improved accuracy in SED modeling.

Abstract

We present a detailed comparison of two approaches, the use of a pre-calculated database and simulated annealing (SA), for fitting the continuum spectral energy distribution (SED) of astrophysical objects whose appearance is dominated by surrounding dust. While pre-calculated databases are commonly used to model SED data, only few studies to date employed SA due to its unclear accuracy and convergence time for this specific problem. From a methodological point of view, different approaches lead to different fitting quality, demand on computational resources and calculation time. We compare the fitting quality and computational costs of these two approaches for the task of SED fitting to provide a guide to the practitioner to find a compromise between desired accuracy and available resources. To reduce uncertainties inherent to real datasets, we introduce a reference model resembling a typical circumstellar system with 10 free parameters. We derive the SED of the reference model with our code MC3D at 78 logarithmically distributed wavelengths in the range [0.3um, 1.3mm] and use this setup to simulate SEDs for the database and SA. Our result shows directly the applicability of SA in the field of SED modeling, since the algorithm regularly finds better solutions to the optimization problem than a pre-calculated database. As both methods have advantages and shortcomings, a hybrid approach is preferable. While the database provides an approximate fit and overall probability distributions for all parameters deduced using Bayesian analysis, SA can be used to improve upon the results returned by the model grid.