Adaptive particle management in a particle-in-cell code

Abstract

In particle-based plasma simulation, when dealing with source terms such as ionization, emission from boundaries, etc., the total number of particles can grow, at times, exponentially. Problems involving the spatial expansion of dynamic plasmas can result in statistical under representation of particle distributions in critical regions. Furthermore, when considering code optimization for massively parallel operation, it is useful to maintain a uniform number of particles per cell. Accordingly, we have developed an algorithm for coalescing or fissioning particles on 2D and 3D orthogonal grids that is based on a method of Assous et al. [F. Assous, T. Pougeard Dulimbert, J. Segre, J. Comput. Phys. 187 (2003) 550]. We present the algorithm and describe in detail its application to particle-in-cell simulations of gas ionization/streamer formation and dynamic, expanding plasmas.

Towards a fully kinetic 3D electromagnetic particle-in-cell model of streamer formation and dynamics in high-pressure electronegative gases

“…We have found that when applied to the management of kinetic particles this algorithm can be somewhat more efficient in controlling the rapid growth of macro-particle numbers than the original particle coalescence model. 14 The kinetic remap closely follows the Lsp particle coalescence model, 14 in which a bilinear weighting scheme is used to combine the particle momenta and charge weights in a cell in order to create a momentum probability distribution function (pdf) for that cell. Replacement particles are generated from the distribution in a way that conserves charge, energy, and momentum globally and locally.…”

“…Critical to the application of this fully kinetic particle treatment is use of an algorithm that manages the total particle count by species while preserving the local momentum distribution functions and conserving charge. 14 Previous computational models of streamer/leader formation and dynamics have included electrostatic kinetic, 15,16 and fluid descriptions including adaptive mesh refinement techniques and moving meshes. [17][18][19] For example, detailed simulations in electronegative gases have been carried out in 2D (Refs.…”

“…IV. A brief description of a new implementation of the adaptive particle management algorithm 14 used in the simulations is given in the Appendix. The new algorithm improves the overall efficiency of the adaptive particle management algorithm over the original, which in turn can decrease the 3D simulation run times by providing somewhat better control over the global particle number.…”

Particle-in-cell modeling of magnetized argon plasma flow through small mechanical apertures

“…The aspect ratios of the grid are allowed to be relatively large because the electromagnetic fields are solved implicitly. Approximately 12 and 36 particles per cell are adaptively maintained [15] for the Ar + and e − particles (both injected and ionized populations), respectively.…”

Moment Preserving Adaptive Particle Weights using Octree Velocity Distributions for PIC Simulations

“…Hewett used computational particles with internal energy in which the error could be accumulated [2]. Assous and then Welch designed methods of merging values to grid nodes and redistributing the moments to particles [3,4]. Though exact energy and momentum conservation is possible with these methods, they are considerably more complicated than the original naive 2:1 merge.…”