Variational formulation of particle algorithms for kinetic plasma simulations

TL;DR

This paper introduces variational formulations for particle algorithms in kinetic plasma simulations, reducing drawbacks of existing methods by conserving energy and momentum, and allowing higher accuracy and flexibility.

Contribution

It presents the first Hamiltonian formulation for particle algorithms in plasma simulations, enhancing accuracy and reducing numerical noise compared to traditional methods.

Findings

Formulated time-explicit, finite-size particle algorithms using variational principles.

Developed a Hamiltonian formulation that avoids solving Poisson's equation.

Presented an algorithm conserving both energy and momentum.

Abstract

Common time-explicit numerical methods for kinetic simulations of plasmas in the low-collisions limit fall into two classes of algorithms: momentum conserving and energy conserving. Each has certain drawbacks. The PIC algorithm does not conserve total energy, which may lead to spurious numerical heating (grid heating). Its overall accuracy is at most second due to the nature of the force interpolation between grid and particle position. Energy-conserving algorithms do not exhibit grid heating, but because their formulation uses potentials, computationally undesirable matrix inversions may be necessary. In addition, compared to PIC algorithms for the same accuracy, these algorithms have higher numerical noise due to the restricted choice of particle shapes. Here we formulate time-explicit, finite-size particle algorithms using particular reductions of the particle distribution function. These reductions are used in two variational principles, a Lagrangian-based and a Hamiltonian-based in conjunction with a non-canonical Poisson bracket. The Lagrangian formulations here generalize previous such formulations. The Hamiltonian formulation is presented here for the first time. Many drawbacks of the two classes of particle methods are mitigated. For example, restrictions on particle shapes are relaxed in energy conserving algorithms, which allows to decrease the numerical noise in these methods. The Hamiltonian formulation of particle algorithms is done in terms of fields instead of potentials, thus avoiding solving Poisson's equation. An algorithm that conserves both energy and momentum is presented. Other features of the algorithms include a natural way to perform coordinate transformations, the use of various time integrating methods, and the ability to increase the overall accuracy beyond second order, including all generalizations.