Scalable Simulations of 3D Turbulence Fine Structure in Nanoflare Using a Novel Plasma Statistical Algorithm
Bojing Zhu, Yan Li, Zhikuo Ma, Hui Yan, Ying Zhong, Wu Wang, Yufeng Guo, Zheng Li, David A. Yuen · Proceedings of the International Conference on Statistics, Theory and Applications (ICSTA ...) · 2023
Based on algorithms of solving directly MHD partial differential equation algorithm (e.g., ATHENA [1], NIRVANA [2], ZEUS [3], FLASH [4], PIC [5-8], HPIC [9-11]), conventional simulation methods cannot attain the extreme range of scale for 3D turbulence fine-structure (Geometry and Physics) in flare-CME phenomena, especially for nanoflare heating problems.Here we present a parallel lattice Boltzmann algorithm based on plasma statistical physics, which allows us to reach even the relativistic regime necessary for modeling 3D turbulence fine-structure evolution.This innovative approach can simultaneously describe the continuous features of plasma at the macro-spatial-temporal scale, particle features of plasma at the micro-spatial-temporal scale, and particle features enforced by magnetic fields.However, the novel algorithm brings several challenges for fine-structure large-scale simulation, such as the gargantuan memory and storage requirements due to high dimensions and output data, and long simulation time, because each run takes a week.We propose optimization technologies for data access, communication, I/O, etc.These optimizations make it possible to achieve scalable and robust simulation (using up to 100,000 cores on the Tianhe-2 supercomputer) and the most extensive system ever run.For the first time, we can analyze the 3D turbulence fine structure by interacting with plasmas and magnetic fields in nanoflare.