Performance boost for radiative magneto-hydrodynamics simulations in astrophysics

The complex dynamics of plasma flows and their interplay with the radiation flux and magnetic fields in stars is typically investigated numerically with radiation magneto-hydrodynamic simulations. Here, we propose to explore different possible approaches to improve the performances of such tools with a twofold objective. On the one hand, we propose to upgrade the CO5BOLD simulation tool, used at IRSOL to study the solar and stellar atmospheres, in order to make it more suitable for investigating the origin and evolution of sunspots and starspots. On the other hand, under the guidance of world-leading experts at USI in the field of computational science, the activities proposed will provide the applicant with new expertise in fields that are at the forefront of scientific computing. More precisely, we propose to port CO5BOLD to GPUs, improve its numerical algorithms to reduce its numerical diffusivity, and explore the use of neural networks to accelerate the solution of the radiative transfer equation. Altogether, this will lay the foundations for an application for a SNSF Professorial Fellowships. The outcome of the proposed activities is expected to have an impact on the solar and stellar physics community and in the field of computational science.