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

This paper, recently published by researchers from Oak Ridge National Laboratory and the University of Tennessee, presents a more efficient way to simulate gas dynamics in regimes where traditional methods struggle. By combining two sophisticated numerical strategies—High-Order/Low-Order (HOLO) acceleration and Micro-Macro (MM) decomposition—the authors have developed a solver for the BGK kinetic model that is both faster and more memory-efficient than standard approaches.

TL;DR

The authors combine High-Order/Low-Order (HOLO) acceleration with Micro-Macro (MM) decomposition to create a robust, fully implicit solver for the BGK kinetic equation. This hybrid approach significantly reduces iteration counts in highly collisional regimes and enables substantial data compression by focusing computational effort on the non-equilibrium perturbation of the gas.

What Problem Does This Paper Address?

Simulating gases across different scales—from the "free-streaming" vacuum of space to the "collision-dominated" air at sea level—is a major challenge in computational fluid dynamics. The BGK model is a popular simplification of the Boltzmann equation used for these "rarefied" gas simulations.

However, when collisions become frequent (the fluid limit), standard numerical solvers like Source Iteration (SI) become incredibly slow, requiring thousands of iterations to converge. While semi-implicit methods exist, certain scenarios—like steady-state solutions or simulations with very fine spatial meshes—require fully implicit time-stepping, which is computationally expensive. The authors sought a way to make these fully implicit solves faster without sacrificing accuracy or requiring massive amounts of memory.

Methods

The paper introduces a novel combination of three key numerical ingredients:

  1. HOLO Acceleration: This technique uses a "low-order" system (simplified fluid-like equations) to provide a high-quality guess for the "high-order" (full kinetic) distribution. This drastically cuts down the number of expensive kinetic "sweeps" needed.
  2. Micro-Macro (MM) Decomposition: Instead of solving for the entire gas distribution $f$, the authors split it into two parts: $f = M + g$. Here, $M$ is the Maxwellian (the local equilibrium state) and $g$ is the micro-perturbation (the departure from equilibrium).
  3. Discontinuous Galerkin (DG) Discretization: The authors implement these models using DG methods in phase space. A major advantage of DG here is that it naturally maintains mathematical consistency between the fluid and kinetic levels, which is often a significant hurdle in hybrid modeling.

The resulting MM-HOLO method uses the fluid-like moment equations to accelerate the evolution of both the equilibrium state and the perturbation.

Datasets

As a theoretical and algorithmic study, the paper evaluates the method using two classic computational benchmarks rather than external datasets:

  • Sod Shock Tube: A standard test for gas dynamics involving a high-pressure gas expanding into a low-pressure region, creating a shock wave, a contact discontinuity, and an expansion fan.
  • Sudden Wall Heating: A 1D-1V boundary layer problem where a wall's temperature is suddenly increased. This is a particularly tough test because it creates a sharp transition from kinetic to fluid behavior right at the wall.

Key Insights and Ideas

  • The Power of $g$: In highly collisional regimes, the perturbation $g$ becomes very small ($O(\nu^{-1})$). The authors realized that because $g$ is small and often smoother than the full distribution $f$, it can be represented on a much coarser velocity grid or compressed using low-rank techniques (like SVD) without losing accuracy.
  • Automatic Consistency: Previous attempts at HOLO often required complex "consistency terms" to keep the high-order and low-order models from drifting apart. By using DG discretization, the authors show that these terms are satisfied automatically by the underlying math of the finite element space.
  • Stability Trade-offs: While standard Source Iteration is "unconditionally stable" (it won't blow up, even if it's slow), HOLO methods are conditionally stable. Their stability depends on the ratio of the time-step to the spatial mesh size ($\Delta t/h_x$), a trade-off for their superior speed.

Key Results

  • Iteration Speedup: In the Sod shock tube test, as the collision frequency increased, the standard SI method's iteration count skyrocketed (from 3 to over 900). In contrast, the HOLO and MM-HOLO methods remained stable and efficient, typically converging in just 5 to 8 iterations.
  • Compression Benefits: The MM-HOLO method proved much more "compressible." For a fixed error level in a high-collision scenario, MM-HOLO achieved the same accuracy with a velocity resolution of $N_v=6$ that standard HOLO required $N_v=10$ to reach.
  • Robustness: The MM-HOLO approach successfully captured the complex boundary layer physics in the wall-heating problem, outperforming SI in terms of the time-step sizes it could handle while maintaining accuracy.

Limitations and Future Work

  • Stiffness at the Boundary: The authors noted that in the wall-heating problem, the method failed to converge when the time-step was extremely large (1000x the explicit limit). This was attributed to "stiffness" in the lagged fluid flux.
  • Future Directions: To solve this, the authors suggest moving to more complex low-order models, such as Navier-Stokes approximations, which include viscosity and heat conduction to better handle those stiff regimes. They also aim to extend the method to 3D geometries and more complex collision models (like the ES-BGK model).

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},
}

Loading related content...

Figure