Coupled Kinetic-Fluid Simulations of Ganymede's Magnetosphere and Hybrid Parallelization of the Magnetohydrodynamics Model

Abstract

The largest moon in the solar system, Ganymede, is the only moon known to possess a strong intrinsic magnetic field. The interaction between the Jovian plasma and Ganymede's magnetic field creates a mini-magnetosphere with periodically varying upstream conditions, which creates a perfect laboratory in nature for studying magnetic reconnection and magnetospheric physics. Using the latest version of Space Weather Modeling Framework (SWMF), we study the upstream plasma interactions and dynamics in this subsonic, sub-Alfvénic system. We have developed a coupled fluid-kinetic Hall Magnetohydrodynamics with embedded Particle-in-Cell (MHD-EPIC) model for Ganymede's magnetosphere, with a self-consistently coupled resistive body representing the electrical properties of the moon's interior, improved inner boundary conditions, and high resolution charge and energy conserved PIC scheme.

I reimplemented the boundary condition setup in SWMF for more versatile control and functionalities, and developed a new user module for Ganymede's simulation. Results from the models are validated with Galileo magnetometer data of all close encounters and compared with Plasma Subsystem (PLS) data. The energy fluxes associated with the upstream reconnection in the model is estimated to be about 10^-7 W/cm^2, which accounts for about 40% to the total peak auroral emissions observed by the Hubble Space Telescope. We find that under steady upstream conditions, magnetopause reconnection in our fluid-kinetic simulations occurs in a non-steady manner. Flux ropes with length of Ganymede's radius form on the magnetopause at a rate about 3/minute and create spatiotemporal variations in plasma and field properties. Upon reaching proper grid resolutions, the MHD-EPIC model can resolve both electron and ion kinetics at the magnetopause and show localized crescent shape distribution in both ion and electron phase space, non-gyrotropic and non-isotropic behavior inside the diffusion regions. The estimated global reconnection rate from the models is about 80 kV with 60% efficiency. There is weak evidence of $sim 1$ minute periodicity in the temporal variations of the reconnection rate due to the dynamic reconnection process.

The requirement of high fidelity results promotes the development of hybrid parallelized numerical model strategy and faster data processing techniques. The state-of-the-art finite volume/difference MHD code Block Adaptive Tree Solarwind Roe Upwind Scheme (BATS-R-US) was originally designed with pure MPI parallelization. The maximum problem size achievable was limited by the storage requirements of the block tree structure. To mitigate this limitation, we have added multithreaded OpenMP parallelization to the previous pure MPI implementation. We opt to use a coarse-grained approach by making the loops over grid blocks multithreaded and have succeeded in making BATS-R-US an efficient hybrid parallel code with modest changes in the source code while preserving the performance. Good weak scalings up to 50,0000 and 25,0000 of cores are achieved for the explicit and implicit time stepping schemes, respectively. This parallelization strategy greatly extends the possible simulation scale by an order of magnitude, and paves the way for future GPU-portable code development. To improve visualization and data processing, I have developed a whole new data processing workflow with the Julia programming language for efficient data analysis and visualization.

As a summary, 1. I build a single fluid Hall MHD-EPIC model of Ganymede's magnetosphere; 2. I did detailed analysis of the upstream reconnection; 3. I developed a MPI+OpenMP parallel MHD model with BATSRUS; 4. I wrote a package for data analysis and visualization.