Stabilized bi-cubic Hermite Bézier finite element method with application to gas-plasma interactions occurring during massive material injection in tokamaks

Ashish Bhole, Boniface Nkonga, José Tarcísio Costa, Guido Huijsmans, Stanislas Pamela, Matthias HoelzlComput. Math. Appl 2023 · DOI

TL;DR

This paper introduces a stabilized finite element method (FEM) for simulating complex gas-plasma interactions in fusion tokamaks. By combining Variational Multiscale (VMS) and shock-capturing techniques, the method significantly improves numerical stability and efficiency, enabling more accurate and cost-effective simulations of phenomena like Shattered Pellet Injection (SPI).

What Problem Does This Paper Address?

Simulating the behavior of plasma in fusion tokamaks is crucial for understanding and controlling fusion reactions. These simulations often involve complex, nonlinear equations that describe plasma dynamics. A major challenge arises when dealing with phenomena like Shattered Pellet Injection (SPI), where massive amounts of material are injected into the plasma. This process can create sharp gradients and discontinuities (shocks) in the plasma, which traditional numerical methods struggle to handle accurately and stably. Standard Galerkin FEMs can suffer from dispersion errors and oscillations near these shocks, leading to unphysical results or simulation failures. Existing methods often require prohibitively small time steps to maintain stability, making large-scale simulations computationally infeasible. This paper aims to address these stability and efficiency issues by developing a robust numerical stabilization technique for high-order FEMs used in plasma physics.

Methods

The core contribution of this paper is the development and application of a stabilized bi-cubic Hermite Bézier finite element method (FEM). This method builds upon the existing Galerkin FEM framework by incorporating two key stabilization strategies:

  • Variational Multiscale (VMS) Formulation: This approach accounts for the effects of unresolved scales (small-scale phenomena not explicitly resolved by the mesh) on the resolved scales. It introduces an upwinding mechanism into the FEM, which helps to dampen spurious oscillations that can arise in convection-dominated flows.
  • Shock-Capturing Stabilization: This strategy adaptively introduces artificial viscosity specifically in regions where shocks or discontinuities are detected. This prevents the numerical solution from developing unphysical oscillations near these sharp gradients, thereby improving stability.

These two stabilization techniques are combined and applied to the bi-cubic Hermite Bézier FEM, which is known for its C¹ continuity and suitability for plasma simulations. The authors implement these stabilized methods within the JOREK code, a well-established framework for simulating tokamak plasmas.

Datasets

Not discussed in the paper. The paper focuses on developing and validating a numerical method using simulated plasma scenarios rather than training on or evaluating against experimental datasets.

Key Insights and Ideas

  • Synergy of VMS and Shock-Capturing: The paper demonstrates that combining VMS for general upwinding and shock-capturing for localized stabilization provides a powerful approach to handle the complex physics of gas-plasma interactions, especially in the presence of shocks.
  • Improved Stability and Efficiency: The stabilized method significantly enhances numerical stability, allowing simulations to run for longer physical times and with larger time steps compared to non-stabilized approaches. This directly translates to reduced computational cost.
  • Accurate Capture of Physical Phenomena: Despite the added stabilization, the method is shown to accurately capture key physical phenomena like thermal quench (TQ) and impurity transport, which are critical for understanding plasma disruptions.
  • Adaptive Nature of Shock-Capturing: The shock-capturing method's ability to adaptively apply artificial viscosity only where needed is crucial. It avoids excessive damping of the solution in smooth regions, preserving the accuracy of the simulation.
  • Impact on MHD Dynamics: The study highlights how numerical stabilization can prevent unphysical behavior in magnetic energy cascading, leading to more realistic simulations of MHD activities.

Key Results

  • Convergence Improvement: Non-stabilized simulations often failed to converge around 0.59 ms (reduced MHD) or 1.4 ms (full MHD) due to the development of discontinuities. The stabilized simulations, particularly with shock-capturing, could run for significantly longer periods (e.g., beyond 2.5 ms).
  • Computational Speedup: Stabilized simulations with appropriate coefficients (e.g., c=10) showed a speedup of more than twice compared to non-stabilized simulations to reach a physical time of 1.5 ms. This is achieved by allowing larger time steps.
  • Accurate Representation of SPI: The method successfully simulated the complex dynamics of Shattered Pellet Injection (SPI) in JET-like plasmas, capturing the evolution of impurity density, plasma temperature, and magnetic energies.
  • Validation of Stabilization Strategies: The paper provides extensive numerical evidence showing that the stabilization techniques improve simulation robustness without significantly altering the underlying physics, as evidenced by comparable thermal energies and plasma currents in stabilized and non-stabilized runs (where convergence was possible).

Limitations and Future Work

  • Parameter Tuning: The authors acknowledge that finding optimal values for the stabilization coefficients can be challenging and often relies on experience, as these parameters depend on physical conditions and grid resolution.
  • Ohmic Heating Term: The full MHD model used in the study does not include the Ohmic heating term in the energy equation. This term can become significant during the thermal quench phase when plasma resistivity increases, and its inclusion is reserved for future work.
  • Low-Resolution Grid: Some simulations were performed on a low-resolution grid to demonstrate the local effect of stabilization. For highly resolved simulations, further investigation into the interaction between grid resolution and stabilization is needed.
  • Future Work:
    • Investigate better estimates for shock-capturing modulation parameters.
  • Include the Ohmic heating term for more accurate simulation of disruption phases like the current quench (CQ).
  • Extend the method to two-temperature reduced and full MHD models.
  • Perform parameter scans to gain deeper insights into MHD instabilities.
  • Explore adaptive mesh refinement and multi-grid methods for further computational efficiency gains, especially for large-scale applications like ITER and DEMO.
  • Implement higher-order Hermite Bézier FEMs in JOREK with these stabilization strategies.

Summary updated

BibTeX
@article{bhole2023stabilized,
  author = {Ashish Bhole and Boniface Nkonga and José Tarcísio Costa and Guido Huijsmans and Stanislas Pamela and Matthias Hoelzl},
  title = {{Stabilized bi-cubic Hermite Bézier finite element method with application to gas-plasma interactions occurring during massive material injection in tokamaks}},
  journal = {Comput. Math. Appl},
  year = {2023},
  doi = {10.1016/j.camwa.2023.04.034},
}
Figure