Saved in:
Bibliographic Details
Main Authors: Cheung, Corwin, Johnson-Noya, Marcos, Xiang, Michael, Chang, Dominic, Guevara, Alfredo
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.23057
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917173489827840
author Cheung, Corwin
Johnson-Noya, Marcos
Xiang, Michael
Chang, Dominic
Guevara, Alfredo
author_facet Cheung, Corwin
Johnson-Noya, Marcos
Xiang, Michael
Chang, Dominic
Guevara, Alfredo
contents We construct the first physics-informed neural-network (PINN) surrogates for relativistic magnetohydrodynamics (RMHD) using a hybrid PDE and data-driven workflow. Instead of training for the conservative form of the equations, we work with Jacobians or PDE characteristics directly in terms of primitive variables. We further add to the trainable system the divergence-free condition, without the need of cleaning modes. Using a novel MUON optimizer implementation, we show that a baseline PINN trained on early-time snapshots can extrapolate RMHD dynamics in one and two spatial dimensions, and that posterior residual-guided networks can systematically reduce PDE violations.
format Preprint
id arxiv_https___arxiv_org_abs_2512_23057
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reconstructing Relativistic Magnetohydrodynamics with Physics-Informed Neural Networks
Cheung, Corwin
Johnson-Noya, Marcos
Xiang, Michael
Chang, Dominic
Guevara, Alfredo
Computational Physics
High Energy Astrophysical Phenomena
General Relativity and Quantum Cosmology
We construct the first physics-informed neural-network (PINN) surrogates for relativistic magnetohydrodynamics (RMHD) using a hybrid PDE and data-driven workflow. Instead of training for the conservative form of the equations, we work with Jacobians or PDE characteristics directly in terms of primitive variables. We further add to the trainable system the divergence-free condition, without the need of cleaning modes. Using a novel MUON optimizer implementation, we show that a baseline PINN trained on early-time snapshots can extrapolate RMHD dynamics in one and two spatial dimensions, and that posterior residual-guided networks can systematically reduce PDE violations.
title Reconstructing Relativistic Magnetohydrodynamics with Physics-Informed Neural Networks
topic Computational Physics
High Energy Astrophysical Phenomena
General Relativity and Quantum Cosmology
url https://arxiv.org/abs/2512.23057