http://arxiv.org/abs/2210.01609
Context. The magnetic field in the solar atmosphere continually reconnects and accelerates charged particles to high energies. Simulations of the atmosphere in three dimensions that include the effects of accelerated particles can aid our understanding of the interplay between energetic particle beams and the environment where they emerge and propagate. We presented the first attempt at such a simulation in a previous paper, emphasising the physical model of particle beams. However, the numerical implementation of this model is not straightforward due to the diverse conditions in the atmosphere and the way we must distribute computation between multiple CPU cores. Aims. Here, we describe and verify our numerical implementation of energy transport by electron beams in a 3D magnetohydrodynamics code parallelised by domain decomposition. Methods. We trace beam trajectories using a Runge-Kutta scheme with adaptive step length control and integrate deposited beam energy along the trajectories with a hybrid analytical and numerical approach. To parallelise this, we coordinate beam transport across subdomains owned by separate processes using a buffering system designed to optimise data flow. Results. Using an ad hoc magnetic field with analytical field lines as a test scenario, we show that our parallel implementation of adaptive tracing efficiently follows a challenging trajectory with high precision. By timing executions of electron beam transport with different numbers of processes, we found that the processes communicate with minimal overhead but that the parallel scalability is still sublinear due to workload imbalance caused by the uneven spatial distribution of beams.
L. Frogner and B. Gudiksen
Wed, 5 Oct 22
53/73
Comments: Submitted to Astronomy & Astrophysics