[Paper Review] Adaptive Multi-Dimensional Particle In Cell

Giovanni Lapenta|ArXiv.org|Jun 4, 2008

Electrohydrodynamics and Fluid Dynamics3 citations

TL;DR

This paper presents an adaptive multi-dimensional Particle-In-Cell (PIC) method that combines spatial grid adaptation and particle rezoning to efficiently simulate multi-scale plasma phenomena. By using a posteriori error indicators for grid refinement and dynamic particle loading, the method maintains accuracy in regions of interest—such as collisionless shocks and dust charging—while reducing computational cost, with demonstrated improvements in numerical stability and statistical accuracy over non-adaptive simulations.

ABSTRACT

Kinetic Particle In Cell (PIC) methods can extend greatly their range of applicability if implicit time differencing and spatial adaption are used to address the wide range of time and length scales typical of plasmas. For implicit differencing, we refer the reader to our recent summary of the implicit moment PIC method implemented in our CELESTE3D code [G. Lapenta, Phys. Plasmas, 13, 055904 (2006)]. Instead, the present document deals with the issue of PIC spatial adaptation. Adapting a kinetic PIC code requires two tasks: adapting the grid description of the fields and moments and adapting the particle description of the distribution function. Below we address both issues. First, we describe how grid adaptation can be guided by appropriate measures of the local accuracy of the solution. Based on such information, grid adaptation can be obtained by moving grid points from regions of lesser interest to regions of higher interest or by adding and removing points. We discuss both strategies. Second, we describe how to adapt the local number of particles to reach the required statistical variance in the description of the particle population. Finally two typical applications of adaptive PIC are shown: collisionless shocks and charging of small bodies immersed in a plasma.

Motivation & Objective

• To address the computational inefficiency of standard explicit PIC methods in resolving wide-ranging time and length scales in plasmas.
• To overcome the limitations of fixed-grid, fixed-particle PIC simulations in capturing localized gradients and dynamic particle distributions.
• To develop a framework that integrates grid adaptation and particle rezoning for implicit PIC methods to enhance accuracy and reduce noise.
• To enable accurate simulation of complex plasma systems such as collisionless shocks and dust charging in space environments.
• To ensure energy conservation and load balancing in parallel implementations through adaptive resolution and particle control.

Proposed method

• Uses a posteriori error indicators to guide grid adaptation based on local solution accuracy, identifying regions requiring higher resolution.
• Employs two grid adaptation strategies: grid motion (moving points to high-interest regions) and grid refinement (adding points where needed).
• Applies particle rezoning by dynamically adjusting the number of particles per cell to maintain statistical accuracy and reduce noise.
• Utilizes the immersed boundary method with susceptibility and damping terms to model dust particle boundary conditions in plasma simulations.
• Implements a moving mesh approach (MMA) and adaptive mesh refinement (AMR) to balance accuracy and computational cost.
• Integrates implicit time differencing to relax stability constraints, allowing larger time steps and coarser grids where appropriate.

Experimental results

Research questions

• RQ1 How can grid adaptation be guided by local accuracy measures in multi-dimensional PIC simulations?
• RQ2 What are the most effective strategies for grid motion versus grid refinement in adaptive PIC methods?
• RQ3 How can particle rezoning maintain statistical accuracy in regions with evolving particle density?
• RQ4 What impact does adaptive grid and particle control have on energy conservation and load balancing in parallel PIC codes?
• RQ5 How does adaptive PIC improve the simulation of complex plasma phenomena such as collisionless shocks and dust charging?

Key findings

• Grid adaptation guided by a posteriori error indicators significantly improves resolution in regions of high gradients, such as shock fronts and sheaths.
• Particle rezoning maintains a consistent number of particles per cell, preventing statistical noise from increasing in low-density regions.
• The combination of adaptive grids and particle control enables stable, accurate simulations of dust charging in plasma environments, achieving steady-state charge accumulation.
• Without particle rezoning, the number of particles per cell decreases over time around the dust, degrading accuracy and preventing convergence to equilibrium.
• The adaptive PIC method achieves better computational efficiency and accuracy compared to fixed-grid, fixed-particle simulations, especially in systems with strong spatial and temporal scale separation.