Smoothed particle magnetohydrodynamics for simulations of galaxy and cosmic structure formation
π arXiv:2608.06978 Β· π₯ PDF Β· 2026-08-07 Β· astro-ph.CO
Authors: Orestis A. Karapiperis [arXiv Β· scholar] , Matthieu Schaller [arXiv Β· scholar] , Nikyta Shchutskyi [arXiv Β· scholar] , Federico A. Stasyszyn [arXiv Β· scholar] , Maarten Elion [arXiv Β· scholar]
π Abstract
We introduce a novel formulation of cosmological smoothed particle magnetohydrodynamics (SPMHD), designed to model magnetic field physics in a vast array of nonlinear astrophysical systems, and which we have implemented in the highly-parallel, entirely modular, and open-source simulation code SWIFT. Our numerical scheme is designed to offer optimal performance at a minimal computational cost, keep a low memory footprint, and most notably couple robustly to effective sub-resolution recipes of galaxy formation. This is achieved through expressing our evolution equations in a density-energy conservative form, and augmenting them with discontinuity-capturing terms tailored to high dynamic range simulations, which are further modulated by adaptive switches that drastically improve coupling to sub-grid models and limit spurious dissipation. We moreover present novel suggestions for the two major regularisation techniques used in modern SPMHD, namely a tensile instability correction and mixed hyperbolic/parabolic divergence-cleaning scheme, to ensure code stability in highly dynamical scenarios. We evaluate the performance of our method on a series of problems of increasing complexity, culminating in three astrophysical applications which have historically proven challenging for mesh-less methods: we study jet launching from a forming proto-stellar core, dynamo amplification in a massive galaxy cluster and magnetic field evolution in a Milky Way-like disk galaxy; the latter constitutes the first reported coupling of the EAGLE galaxy formation model to a magnetohydrodynamics solver. Keeping model hyperparameters fixed across our test suite to provide a transparent picture of our method's capabilities in production, we demonstrate sound performance and convergence with resolution on standard `laboratory' numerical experiments, as well as competitive capabilities in realistic applications.