Abstract
This paper introduces Poisson-EQMOM, a novel quadrature-based moment method designed to solve 2D kinetic equations where particles maintain a constant speed (angular distributions). By employing the Poisson kernel as a basis, the authors develop a system that guarantees the positivity of the distribution function, remains computationally efficient, and is mathematically proven to converge to the exact solution as the number of moments increases. This method addresses long-standing issues in kinetic modeling, such as the unphysical negative densities found in $P_N$ methods and the high computational costs of $M_N$ methods.
Introduction
Kinetic equations are essential for describing the evolution of particle distributions in high-dimensional phase spaces, with applications ranging from gas dynamics to the collective motion of swimming bacteria. However, direct numerical solutions are often computationally prohibitive. The Method of Moments reduces this complexity by tracking only the averages (moments) of the distribution, such as density and momentum.
For equations with constant velocity magnitude (KE-VC)—prevalent in active matter physics and radiative transfer—existing moment methods face significant hurdles. Fourier-based $P_N$ methods frequently produce unphysical negative values for particle density. Conversely, Maximum Entropy ($M_N$) methods, while positive-preserving, are notoriously difficult and expensive to solve. Standard Extended Quadrature Method of Moments (EQMOM) variants often lack global existence proofs for their closure systems. Poisson-EQMOM bridges this gap by providing a physically consistent, mathematically robust, and numerically stable framework.
The Poisson-EQMOM Framework
The core innovation of this work is the use of the Poisson kernel as the basis for approximating the angular distribution $f(\theta)$. Instead of Gaussian kernels, which are standard for infinite linear velocity spaces, the authors approximate the distribution as a convex combination of Poisson kernels:
$$f_N(\theta) = \sum_{\alpha=1}^{N} \rho_\alpha P_r(\phi_\alpha - \theta)$$
Where $P_r$ is the Poisson kernel, $\rho_\alpha$ are weights, $\phi_\alpha$ are centers, and $r \in [0, 1)$ represents the concentration or "width" of the kernel. As $r$ approaches 1, the kernel becomes increasingly peaked, eventually approaching a Dirac delta distribution.
Key Technical Components
- Moment Inversion: The parameters $(\rho, \phi, r)$ are determined from known moments using the theory of Toeplitz matrices and orthogonal polynomials on the unit circle.
- The "Lifting" Regularization: To handle numerical ill-conditioning, the authors introduce a "lifting" trick. By adding a small constant $\ell$ to the zeroth moment ($m_0$) before inversion, they prevent the kernels from becoming excessively narrow. This regularization drastically improves accuracy and stability for highly "peaked" distributions.
- Numerical Splitting: The system is solved using a splitting method that separates transport (movement) and collision (interaction) terms, utilizing analytical expressions for kinetic-based fluxes.
Key Insights and Results
Poisson-EQMOM is uniquely suited for periodic angular spaces. Unlike many quadrature methods that rely on heuristic closures, the authors provide a technical proof showing that the reconstructed distribution $f_N$ converges to the true distribution $f$ as the number of moments $N \to \infty$.
Validation and Performance
- Positivity and Stability: The method successfully maintained positive distributions across all benchmarks, avoiding the "ringing" artifacts typical of $P_N$ methods.
- Efficiency: The "lifting" algorithm reduced the runtime of moment inversion by orders of magnitude compared to iterative root-finding methods. For example, a case with $N=32$ moments was solved in 36 seconds, whereas standard methods might fail to converge.
- Complex Physics: In 2D tests with reflexive walls, Poisson-EQMOM captured emergent behaviors like density waves and stable vortices, matching the results of high-fidelity microscopic particle simulations (using 10,000 particles) at a fraction of the cost.
- Hydrodynamic Limits: In 1D Riemann problems, the method correctly converged to macroscopic hydrodynamic solutions as the scaling parameter $\epsilon$ approached zero.
Limitations and Future Work
While Poisson-EQMOM represents a significant advancement for 2D kinetic equations, several challenges remain. Extending the method to 3D (velocity on a sphere) requires a different kernel and more complex spherical harmonic theory. Additionally, while the system is proven to be strictly hyperbolic for $N=1$, a general proof for $N > 1$ remains an open mathematical challenge. Future research will also focus on developing Asymptotic-Preserving (AP) schemes to ensure efficiency in the "stiff" limit where collisions occur almost instantaneously.
Summary updated