Testing the conservative character of particle simulations: II. Spurious heating of guiding centers and full orbits subject to fluctuations expressed in terms of ${\bf E}$ and ${\bf B}$

TL;DR

This paper investigates how numerical inaccuracies and inconsistencies in electromagnetic field representations can cause artificial heating and acceleration in particle simulations of tokamak plasmas, highlighting issues with common approximation methods.

Contribution

It demonstrates that using ${f E}$ and ${f B}$ fields directly can lead to spurious effects, and shows that N-point gyroaveraging may worsen simulation accuracy, providing insights into simulation fidelity.

Findings

Numerical noise causes partial heating, reducible by smaller time steps.

Inconsistencies between ${f E}$ and ${f B}$ induce secular acceleration.

N-point gyroaveraging can worsen simulation deviations.

Abstract

For an axisymmetric tokamak plasma, Hamiltonian theory predicts that the orbits of charged particles must stay on invariant tori of conserved energy in the moving frame of reference of a wave that propagates along the torus with a fixed angular phase velocity. In principle, this is true for arbitrary mode structures in the poloidal plane, but only if the fluctuations are expressed in terms of potentials $\Phi$ and ${\bf A}$, which satisfy Faraday's law by definition. Here, we use the physical fields ${\bf E}$ and ${\bf B}$, where Faraday's law may be violated by errors introduced during the process of computing or designing the wave field through numerical inaccuracies, approximations, or gross negligence. Numerical heating caused by noise-like artifacts on the grid scale can to some extent be reduced via shorter time steps. In contrast, coherent inconsistencies between ${\bf E}$ and ${\bf B}$ cause spurious acceleration that is independent of time steps or numerical methods, but can be sensitive to geometry. In particular, we show that secular acceleration is enhanced when one imposes nonnormal modes that possess strong up-down asymmetry instead of the usual in-out asymmetry of normal toroidal (eigen)modes. Our tests are performed for full gyroorbit and guiding center (GC) models, which give similar results. In addition, we show that $N$-point gyroaveraging is not a recommendable method to enhance the realism of GC simulations. Besides breaking conservation laws, $N$-point gyroaveraging in our example makes the GC results deviate further from the full orbit results, showing that this method can even give the wrong trend.