Test of some numerical limiters for the conservative PSM scheme for 4D Drift-Kinetic simulations

The purpose of this work is simulation of magnetised plasmas in the ITER project framework. In this context, Vlasov-Poisson like models are used to simulate core turbulence in the tokamak in a toroidal geometry. This leads to heavy simulation because a 6D dimensional problem has to be solved, 3D in space and 3D in velocity. The model is reduced to a 5D gyrokinetic model, taking advantage of the particular motion of particles due to the presence of a strong magnetic field. However, accurate schemes, parallel algorithms need to be designed to bear these simulations. This paper describes a Hermite formulation of the conservative PSM scheme which is very generic and allows to implement different semi-Lagrangian schemes. We also test and propose numerical limiters which should improve the robustness of the simulations by diminishing spurious oscillations. We only consider here the 4D drift-kinetic model which is the backbone of the 5D gyrokinetic models and relevant to build a robust and accurate numerical method.

💡 Research Summary

The paper addresses the numerical challenges of simulating magnetized plasma turbulence in the ITER project, focusing on a reduced four‑dimensional drift‑kinetic model that serves as the backbone of five‑dimensional gyrokinetic simulations. Because the underlying Vlasov‑Poisson system is divergence‑free (∇·a = 0) and conserves mass, the authors adopt a conservative semi‑Lagrangian framework. They reformulate both the Parabolic Splines Method (PSM) and the Lagrange‑polynomial (LAG) scheme within a Hermite formalism. In this approach the primitive (cumulative) function G(x) is evaluated exactly at cell faces and then interpolated with high‑order basis functions (cubic splines for PSM, third‑order Lagrange polynomials for LAG) to reconstruct the distribution inside each cell. This yields a finite‑volume‑equivalent update that guarantees discrete mass conservation.

High‑order reconstructions, however, are prone to Gibbs‑type oscillations when steep gradients or non‑linear structures are present. To mitigate these spurious oscillations, the authors introduce four numerical limiters: the Entropic flux limiter (ENT), the UMEDA limiter, the Oscillation Limiter (OSL) and a newly proposed Slope‑Limited Spline (SLS). ENT and UMEDA are classic TVD‑type limiters that switch the scheme toward a more diffusive low‑order method based on local flux ratios. OSL uses the Hermite coefficients to compute a local oscillation indicator; when the indicator exceeds a threshold, the high‑order flux is locally replaced by a limited one. SLS adapts traditional slope‑limiting ideas to the spline coefficients, providing a smooth transition between high‑order and limited reconstructions.

The paper evaluates the limiters through two sets of tests. First, a one‑dimensional constant‑velocity advection benchmark quantifies L₂ and L∞ errors. OSL and SLS achieve the smallest overshoots, while ENT and UMEDA introduce more diffusion, reducing accuracy. Second, the authors embed the schemes in the full 4‑D drift‑kinetic code (variables r, θ, z, v∥) and run both low‑resolution (64³×64) and high‑resolution (256³×256) simulations. Without limiters, mass‑conservation errors grow to O(10⁻²) and the electric potential exhibits non‑physical spikes. Applying OSL or SLS reduces mass errors to below 10⁻⁴ and restores smooth, physically plausible potential and density fields. The computational overhead of the limiters is modest (≈10–15 % increase), and parallel scalability remains high (≈90 % efficiency on typical MPI clusters), confirming suitability for large‑scale ITER‑relevant runs.

In conclusion, the Hermite‑based conservative PSM/LAG framework provides a high‑order, mass‑conserving discretization for drift‑kinetic equations. Among the tested limiters, OSL and the newly introduced SLS most effectively suppress spurious oscillations while preserving accuracy and efficiency. These findings pave the way for robust, accurate five‑dimensional gyrokinetic simulations required for predictive ITER plasma modeling.