An Active Flux Method for Maxwell's equations
math.NA, cs.NA
Submitted: 2026-11-02
Updated: 2026-11-02
License: http://creativecommons.org/licenses/by-sa/4.0/
The gist: We present a novel Active Flux method and a divergence-free method using exact evolution operators for solving Maxwell's equations in three space dimensions on Cartesian grids.
Abstract
We present a novel Active Flux method and a divergence-free method using exact evolution operators for solving Maxwell's equations in three space dimensions on Cartesian grids. The proposed methods are fully discrete, third-order accurate and rely on a truly multidimensional evolution operator based on Kirchhoff's formula for the electromagnetic wave equation. The first method is a classical Active Flux method with globally continuous reconstruction while the second method is an adaptation of the Active Flux method to a locally divergence-free reconstruction by Balsara. The applied reconstruction is built from facial degrees of freedom and is no longer globally continuous. The classical method keeps the divergence error bounded. The divergence-free method with face averages initialized by a potential keeps the constraints at round-off. We compare the accuracy of both methods for smooth and discontinuous test problems.
Sources
- A fully discrete Active Flux method for the Euler equations comparing different truly multi-dimensional evolution operators
- Active flux for ideal magnetohydrodynamics: A positivity-preserving scheme with the Godunov-Powell source term
Related papers
- Do physics-informed neural networks (PINNs) need to be deep? Shallow PINNs using the Levenberg-Marquardt algorithm
- A Neural-preconditioned Poisson Solver for Mixed Dirichlet and Neumann Boundary Conditions
- Advanced Linear Algebra with Applications - Part I (Numerical linear algebra for PDEs, machine learning, and data assimilation)
- Operator learning meets inverse problems: A probabilistic perspective
- Second-order consistency for learning chaotic dynamics via randomized Jacobian matching
- Windowed thinning and query complexity for the bouncy particle and Zigzag samplers