On high-order/low-order and micro-macro methods for implicit time-stepping of the BGK model

Cory D. Hauck, M. Paul Laiu, Stefan Schnake

Abstract

In this paper, a high-order/low-order (HOLO) method is combined with a micro-macro (MM) decomposition to accelerate iterative solvers in fully implicit time-stepping of the BGK equation for gas dynamics. The MM formulation represents a kinetic distribution as the sum of a local Maxwellian and a perturbation. In highly collisional regimes, the perturbation away from initial and boundary layers is small and can be compressed to reduce the overall storage cost of the distribution. The convergence behavior of the MM methods, the usual HOLO method, and the standard source iteration method is analyzed on a linear BGK model. Both the HOLO and MM methods are implemented using a discontinuous Galerkin (DG) discretization in phase space, which naturally preserves the consistency between high- and low-order models required by the HOLO approach. The accuracy and performance of these methods are compared on the Sod shock tube problem and a sudden wall heating boundary layer problem. Overall, the results demonstrate the robustness of the MM and HOLO approaches and illustrate the compression benefits enabled by the MM formulation when the kinetic distribution is near equilibrium.

Introduction

Simulating gas dynamics across multiscale regimes—ranging from the rarefied "free-streaming" limit to the collision-dominated fluid limit—remains a significant challenge in computational fluid dynamics. The BGK model serves as a popular simplification of the Boltzmann equation for these rarefied gas simulations. However, as the gas approaches the fluid limit (highly collisional regimes), standard numerical solvers like Source Iteration (SI) suffer from severe convergence degradation, often requiring thousands of iterations. While semi-implicit methods are available, certain applications, such as steady-state solutions or simulations with high spatial resolution, necessitate fully implicit time-stepping. This research introduces a novel framework that integrates High-Order/Low-Order (HOLO) acceleration with Micro-Macro (MM) decomposition to create a robust, memory-efficient, and fast implicit solver.

Methodology

1. Micro-Macro (MM) Decomposition

The core of the proposed approach is the decomposition of the kinetic distribution function $f$ into two components: $$f = M + g$$ where $M$ represents the local Maxwellian (the equilibrium state) and $g$ is the micro-perturbation (the departure from equilibrium). In highly collisional regimes, $g$ becomes small, scaling as $O(\nu^{-1})$. This decomposition allows the solver to focus computational effort on the perturbation, which is often smoother and more compressible than the full distribution.

2. HOLO Acceleration

HOLO methods utilize a "low-order" system—typically a set of simplified fluid-like moment equations—to provide an algebraic acceleration for the "high-order" full kinetic distribution. By solving the low-order system, the method generates a high-quality guess for the kinetic state, drastically reducing the number of expensive kinetic "sweeps" required for convergence.

3. Discontinuous Galerkin (DG) Discretization

A critical aspect of the implementation is the use of DG methods in phase space. A recurring difficulty in hybrid modeling is maintaining mathematical consistency between the fluid and kinetic levels. The authors demonstrate that the DG discretization naturally satisfies the necessary consistency terms, ensuring that the high-order and low-order models remain synchronized without the need for complex ad-hoc corrections.

Numerical Experiments and Results

The performance of the MM-HOLO method was evaluated using two primary benchmarks: the Sod shock tube and a sudden wall heating problem.

Sod Shock Tube

In the Sod shock tube test, the collision frequency was varied to test the solver's efficiency in the fluid limit. As collisions increased, the standard SI method's iteration count grew from 3 to over 900. In contrast, the HOLO and MM-HOLO methods exhibited remarkable stability, consistently converging in just 5 to 8 iterations. Furthermore, the MM-HOLO method demonstrated superior data compression; it achieved the same accuracy with a velocity resolution of $N_v=6$ that the standard HOLO method required $N_v=10$ to reach.

Sudden Wall Heating

This 1D-1V boundary layer problem involves a sharp transition from kinetic to fluid behavior at the wall. The MM-HOLO approach successfully captured the complex boundary layer physics, outperforming SI in terms of the allowable time-step size while maintaining high accuracy. This robustness is particularly valuable for resolving the stiff gradients inherent in boundary layer flows.

Discussion and Insights

  • Efficiency of the Perturbation: Because the perturbation $g$ is small and relatively smooth in collisional regimes, it can be represented on coarser velocity grids or compressed via low-rank techniques like Singular Value Decomposition (SVD) without significant loss of accuracy.
  • Stability Trade-offs: While Source Iteration is unconditionally stable, HOLO methods are conditionally stable based on the ratio of the time-step to the spatial mesh size ($\Delta t/h_x$). This is a necessary trade-off for the order-of-magnitude speedup in convergence.
  • Consistency: The use of DG discretization simplifies the implementation of HOLO methods by providing automatic consistency between the moment equations and the kinetic sweeps.

Conclusion

The combination of MM decomposition and HOLO acceleration provides a powerful framework for implicit time-stepping of the BGK model. The approach significantly reduces iteration counts in collisional regimes and enables substantial memory savings through the compression of the non-equilibrium perturbation.

Future work will focus on addressing the stiffness observed at extremely large time-steps in boundary layer problems. The authors suggest that incorporating more complex low-order models, such as Navier-Stokes approximations that include viscosity and heat conduction, may further enhance stability. Additionally, extending the framework to 3D geometries and more sophisticated collision models, such as the ES-BGK model, remains a priority for future development.

Summary updated

Abstract

In this paper, a high-order/low-order (HOLO) method is combined with a micro-macro (MM) decomposition to accelerate iterative solvers in fully implicit time-stepping of the BGK equation for gas dynamics. The MM formulation represents a kinetic distribution as the sum of a local Maxwellian and a perturbation. In highly collisional regimes, the perturbation away from initial and boundary layers is small and can be compressed to reduce the overall storage cost of the distribution. The convergence behavior of the MM methods, the usual HOLO method, and the standard source iteration method is analyzed on a linear BGK model. Both the HOLO and MM methods are implemented using a discontinuous Galerkin (DG) discretization in phase space, which naturally preserves the consistency between high- and low-order models required by the HOLO approach. The accuracy and performance of these methods are compared on the Sod shock tube problem and a sudden wall heating boundary layer problem. Overall, the results demonstrate the robustness of the MM and HOLO approaches and illustrate the compression benefits enabled by the MM formulation when the kinetic distribution is near equilibrium.
BibTeX
@inproceedings{hauck2025high,
  author = {Cory D. Hauck and M. Paul Laiu and Stefan Schnake},
  title = {{On high-order/low-order and micro-macro methods for implicit time-stepping of the BGK model}},
  booktitle = {SIAM},
  year = {2025},
  doi = {10.1137/24m1698298},
}
Figure