no. 14 · Friday 13 November 2026
Solving the BGK Model and Boltzmann equation by Fourier Neural Operator with conservative constraints
Abstract
The numerical approximation of the Boltzmann collision operator presents significant challenges because of its high dimensionality, quadratic nonlinearity, and nonlocal integral structure. In this work, we develop a Fourier Neural Operator (FNO)-based framework for learning finite time evolution operators associated with the spatially homogeneous Boltzmann equation and its simplified BGK model in different velocity space dimensions. The framework learns mappings between distribution functions in either sequence-to-sequence or point-to-point form and, once trained, enables rapid evaluation without repeated classical time evolution. To improve physical consistency, we introduce a conservative FNO (C-FNO) by augmenting the training loss with soft penalties for deviations in mass, momentum, and energy from the conservation law. Several numerical experiments are presented to demonstrate that the learned operators accurately reproduce the principal collision dynamics, while C-FNO generally reduces macroscopic drift relative to the standard FNO, highlighting its potential as a promising framework for physics-constrained operator learning in kinetic theory and other nonlinear integro-differential equations.
- comments
- Updated version accepted by J. Comput. Appl. Math
- fields
- Numerical Analysis