Speaker
Description
Magnetized turbulence is ubiquitous in many astrophysical and terrestrial plasmas but no universal theory exists. Even in the simplest plasma approximation, magnetohydrodynamics (MHD), the detailed energy dynamics are still not well understood. In this talk, I present a suite of idealized MHD turbulence simulations that only vary in their dynamical range, i.e., in their separation between the large-scale forcing and dissipation scales. From a practical point of view, I show how numerical dissipation can be estimated using an energy transfer analysis framework and that implicit large eddy simulations match direct numerical simulations. From a theoretical point of view, I use the same framework to demonstrate that – contrary to hydrodynamic turbulence – the cross-scale energy fluxes are not constant in MHD turbulence. This applies both to different mediators (such as energy cascade processes) for a given dynamical range as well as to a dependence on the dynamical range itself. Moreover, there exists no indication of convergence even at the highest resolution simulation at $2048^3$ cells. This raises the question on whether an asymptotic regime in MHD turbulence exists, and, if yes, what it looks like. Finally, to tackle this question in the future I introduce Parthenon (a performance portable adaptive mesh refinement framework to solve partial differential equations) and AthenaPK (the MHD application code on top), which recently reached 92% weak scaling parallel efficiency on 73,728 GPUs on Frontier (the first TOP500 exascale supercomputer).
