Magnetohydrodynamics with Physics Informed Neural Operators

We present the first application of physics informed neural operators, which use tensor Fourier neural operators as their backbone, to model 2D incompressible magnetohydrodynamics simulations. Our results indicate that physics informed AI can accurately model the physics of magnetohydrodynamics simulations that describe laminar flows with Reynolds numbers $Re\leq250$. We also quantify the applicability of our AI surrogates for turbulent flows, and explore how magnetohydrodynamics simulations and AI surrogates store magnetic and kinetic energy across wavenumbers. Based on these studies, we propose a variety of approaches to create AI surrogates that provide a computationally efficient and high fidelity description of magnetohydrodynamics simulations for a broad range of Reynolds numbers. Neural operators and scientific software to produce simulation data to train, validate and test our physics informed neural operators are released with this manuscript.