Solver-in-the-Loop Applications in Astrophysical (Magneto)hydrodynamics

TL;DR

This paper demonstrates how machine learning models integrated into astrophysical (magneto)hydrodynamics simulators can improve simulation accuracy and convergence, enabling more efficient and precise modeling of complex astrophysical phenomena.

Contribution

It introduces two novel applications of ML inside astrophysical simulators: a learned cooling function for better low-res simulations and a CNN for correction of MHD simulations.

Findings

Learned cooling function recovers high-res dynamics in low-res simulations.

CNN effectively corrects 2D MHD simulation errors.

Open-source code provided for reproducibility.

Abstract

We present two promising applications of training machine learning models inside a differentiable astrophysical (magneto)hydrodynamics simulator. First, we address the problem of slow convergence in hydrodynamical simulations of wind-blown bubbles with radiative cooling. We demonstrate that a learned cooling function can recover high-resolution dynamics in low-resolution simulations. Secondly, we train a convolutional neural network to correct 2D magnetohydrodynamics simulations of a specific blast wave problem. These case studies pave the way for the principled application of more general machine learning models inside astrophysical simulators. The code is available open source under https://github.com/leo1200/eurips25corr.