Tensor-Product Discretization for the Spatially Inhomogeneous and Transient Boltzmann Equation in Two Dimensions

Philipp Grohs, Ralf Hiptmair, Simon PintarelliSMAI 2017 · DOI

TL;DR

This paper introduces a novel spectral-finite element method for solving the Boltzmann equation, a fundamental equation in kinetic theory describing rarefied gases. By combining specialized polynomial bases in velocity with standard finite elements in space, they achieve efficient and accurate simulations of complex gas flows, particularly for scenarios with varying densities and speeds.

What Problem Does This Paper Address?

The Boltzmann equation is crucial for understanding gas dynamics, especially in rarefied regimes where traditional fluid dynamics models break down. However, its high dimensionality (phase space: position and velocity) and the complexity of the collision operator make direct numerical solutions computationally prohibitive. Existing methods often struggle with accuracy, computational cost, or handling long-time behavior, especially for spatially inhomogeneous problems. This paper aims to develop a method that is both computationally efficient and accurate for these challenging problems, particularly for the variable hard spheres model.

Methods

The authors propose a "tensor-product ansatz" for discretizing the distribution function. This involves:

  • Velocity Discretization: Using Maxwellian-weighted, tensor-product Laguerre polynomials in velocity. This approach avoids truncating the velocity domain and naturally preserves conserved quantities, leading to better long-time behavior compared to Fourier-based methods. They also introduce Polar-Laguerre basis functions to efficiently handle the rotational symmetry of the collision operator.
  • Spatial Discretization: Employing continuous, linear finite elements in the spatial domain.
  • Advection Discretization: A Galerkin least squares technique is used for the advection part of the equation, leading to an implicit time-stepping formulation. This is advantageous as it avoids CFL-like restrictions on the time step, especially for high Knudsen numbers.
  • Collision Operator Discretization: The discrete collision operator is evaluated with an asymptotic complexity of $O(K^5)$, where $K$ is the polynomial degree in each velocity direction. They also develop techniques to exploit the translational invariance of the collision operator, reducing computational cost.
  • Conservation: A Lagrange multiplier method is used to enforce conservation of mass, momentum, and energy.
  • Time Stepping: A Lie-Trotter splitting scheme separates the advection and collision steps, with implicit treatment for advection and explicit for collisions.

Datasets

Not discussed in the paper. The paper focuses on developing and validating a numerical method using synthetic problems and established benchmarks.

Key Insights and Ideas

  • Spectral Velocity Basis for Long-Time Accuracy: Using Laguerre polynomials in velocity, rather than Fourier modes, avoids aliasing errors and allows for accurate long-time simulations, which is a significant limitation of many spectral methods.
  • Exploiting Symmetry for Efficiency: The Polar-Laguerre basis and specific transformations (to Hermite and Nodal bases) are key to efficiently handling the rotational symmetry of the collision operator and reducing computational complexity.
  • Galerkin Least Squares for Stable Advection: This technique provides a stable and implicit treatment of the advection term, crucial for avoiding restrictive time-step limitations.
  • Trade-off in Shifting for Zero Momentum: While shifting the velocity distribution to zero momentum can reduce the number of significant coefficients, the overhead of the transformation can outweigh the benefits for smaller polynomial degrees.

Key Results

  • Computational Efficiency: The collision operator evaluation has an asymptotic complexity of $O(K^5)$, and the overall method is designed for efficiency.
  • Accuracy: The method demonstrates exponential convergence in the polynomial degree $K$ for smooth solutions, as shown in the BKW solution tests.
  • Validation: The method is validated against various scenarios, including:
  • Homogeneous BKW solution: Shows exponential convergence and accurate long-time behavior.
  • Supersonic flow in a wind tunnel: Accurately reproduces shock waves and macroscopic quantities.
  • Nozzle flow: Demonstrates good convergence with respect to spatial and velocity discretization.
  • Shock tube: Shows good agreement with reference solutions, with errors decreasing for smaller Knudsen numbers.
  • Flow generated by a temperature gradient: Captures temperature profiles and mass distribution functions.
  • Numerical Blow-up Mitigation: The paper discusses and demonstrates techniques (like projecting to positive velocity values) to handle potential numerical blow-up arising from negative distribution function values near singularities.

Limitations and Future Work

  • Discontinuities: The paper notes that for initial distributions with discontinuities or discontinuities imposed by boundary conditions (e.g., hot/cold walls), convergence degrades with respect to velocity degrees of freedom.
  • Boundary Conditions: While the method can handle various boundary conditions, it is noted that for discontinuous finite elements, satisfying boundary conditions at corners can be challenging.
  • Future Work: The authors mention that details of the implementation, including parallelization strategies and specific boundary condition treatments, will be published separately. They also suggest that the method can be extended to more general collision kernels and potentially to higher dimensions. The potential for further optimization through more advanced basis transformations and quadrature rules is also implied.

Summary updated

Abstract

We consider the spatially inhomogeneous and nonlinear Boltzmann equation for the variable hard spheres model. The distribution function is discretized by a tensor-product ansatz by combining Maxwellian modulated Laguerre polynomials in velocity with continuous, linear finite elements in the spatial domain. The advection problem in phase space is discretized through a Galerkin least squares technique and yields an implicit formulation in time. The discrete collision operator can be evaluated with an asymptotic effort of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>𝒪</mml:mi> <mml:mo>(</mml:mo> <mml:msup> <mml:mi>K</mml:mi> <mml:mn>5</mml:mn> </mml:msup> <mml:mo>)</mml:mo> </mml:mrow> </mml:math> , where <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>K</mml:mi> </mml:math> is the number of velocity degrees of freedom in a single direction. Numerical results in 2D are presented for rarefied gases with different Mach and Knudsen numbers.
BibTeX
@inproceedings{grohs2017tensor,
  author = {Philipp Grohs and Ralf Hiptmair and Simon Pintarelli},
  title = {{Tensor-Product Discretization for the Spatially Inhomogeneous and Transient Boltzmann Equation in Two Dimensions}},
  booktitle = {SMAI},
  year = {2017},
  doi = {10.5802/smai-jcm.26},
}
Figure