Studying magnetic reconnection and turbulence in magnetized plasmas using GX

The study of inverse transfer of magnetic energy mediated by reconnection in strongly magnetized plasmas is of significant importance due to the growing interests in understanding the emergence of large-scale coherent magnetic structures ubiquitous in the universe. Many of these plasmas where reconnection occurs are such that collisions are very rare, and kinetic effects are conjectured to play a key role in determining the properties of reconnection, and thus, energy transfer. However, this investigation poses a challenging multi-scale problem that requires advanced numerical techniques.

We report on the implementation of the  rigorous reduced low-beta gyro-kinetic model KREHM (Zocco & Schekochihin, 2011) in GX (Mandell et al, 2022), a GPU-native code specifically designed and optimized in CUDA/C++. By conducting numerical benchmarks on phenomena such as the tearing mode, kinetic Alfvén waves, and kinetic plasma turbulence, we demonstrate that running GX on a single GPU outperforms conventional, CPU-based solvers by several orders of magnitude.

These findings highlight the significant computational advantages offered by GX, providing researchers with a powerful tool for investigating magnetic reconnection and turbulence in magnetized plasmas. We exemplify this with simulations of the inverse transfer of magnetic energy at sub-ion scales.