MLody -- Deep Learning Generated Polarized Synchrotron Coefficients

TL;DR

This paper introduces MLody, a deep neural network that accurately computes polarized synchrotron coefficients across diverse plasma conditions, improving radiative transfer modeling in high-energy astrophysics.

Contribution

MLody is a novel deep learning model that enhances the accuracy of polarized synchrotron coefficient calculations for astrophysical simulations.

Findings

MLody achieves high accuracy across various plasma parameters.

Synthetic images show up to twofold differences in polarization compared to traditional methods.

Implications for improved parameter estimation in black hole observations.

Abstract

Polarized synchrotron emission is a fundamental process in high-energy astrophysics, particularly in the environments around black holes and pulsars. Accurate modeling of this emission requires precise computation of the emission, absorption, rotation, and conversion coefficients, which are critical for radiative transfer simulations. Traditionally, these coefficients are derived using fit functions based on precomputed ground truth values. However, these fit functions often lack accuracy, particularly in specific plasma conditions not well represented in the datasets used to generate them. In this work, we introduce ${\tt MLody}$, a deep neural network designed to compute polarized synchrotron coefficients with high accuracy across a wide range of plasma parameters. We demonstrate ${\tt MLody}$'s capabilities by integrating it with a radiative transfer code to generate synthetic polarized synchrotron images for an accreting black hole simulation. Our results reveal significant differences, up to a factor of two, in both linear and circular polarization compared to traditional methods. These differences could have important implications for parameter estimation in Event Horizon Telescope observations, suggesting that ${\tt MLody}$ could enhance the accuracy of future astrophysical analyses.