Noise and error analysis and optimization in particle-based kinetic plasma simulations

TL;DR

This paper analyzes noise in particle-based plasma simulations, focusing on error sources, negative correlations due to fixed particle numbers, and optimizing kernel density estimation for improved accuracy.

Contribution

It introduces a detailed noise and error analysis framework for particle methods, including covariance analysis and bias-variance optimization, applicable to both continuous and grid-based density estimations.

Findings

Negative correlations in particle counts affect electric field noise.

Optimal kernel width balances bias and variance in density estimation.

Particle shape and kernel choices ensure charge conservation on grids.

Abstract

We analyze the noise in macro-particle methods used in plasma physics and fluid dynamics, leading to approaches for minimizing the total error, focusing on electrostatic models in one dimension. We describe kernel density estimation for continuous values of the spatial variable, $x$, and compute the covariance matrix for uniform true density. The band width of the covariance matrix is related to the width of the kernel. We find the presence of constant negative terms in the elements of the covariance matrix both on and off-diagonal. These negative correlations are related to the fact that the total number of particles is fixed at each time step. The effect of these negative correlations on the electric field, computed by Gauss's law, is that the noise in the electric field is related to a process called the Ornstein-Uhlenbeck bridge. For non-constant density still with continuous $x$, we analyze the total error in the density estimation and discuss it in terms of bias-variance optimization (BVO). For some characteristic length $l$ and kernel width $h$, having too few particles within $h$ leads to too much variance; for $h$ large relative to $l$, there is too much smoothing of the density. The optimum between these two limits is found by BVO. We repeat the analysis for $x$ discretized on a grid, connecting particle shapes and estimation kernels. If the particle shape satisfies a sum rule, the charge deposited on the grid is conserved exactly. Particle shapes expressed as the convolution of a kernel with another kernel that satisfies the sum rule also obey the sum rule. This property holds for kernels of arbitrary width, including widths that are not integer multiples of the grid spacing. We show good agreement with numerical computations.