Stable 3D FDTD Method for arbitrary Fully Electric and Magnetic Anisotropic Maxwell Equations

TL;DR

This paper introduces a new fully anisotropic 3D FDTD Maxwell solver capable of handling complex media with improved accuracy and stability, validated through rigorous testing and transformation optics applications.

Contribution

The paper presents a novel second-order accurate averaged FDTD method for anisotropic media, with rigorous stability proof and extensive numerical validation.

Findings

The averaged method achieves second-order accuracy for smoothly varying materials.

The solver maintains stability over long-time simulations and extreme parameter variations.

Accuracy improves significantly with only a 20% increase in computational time.

Abstract

We have developed a new fully anisotropic 3D FDTD Maxwell solver for arbitrary electrically and magnetically anisotropic media for piecewise constant electric and magnetic materials that are co-located over the primary computational cells. Two numerical methods were developed that are called non-averaged and averaged methods, respectively. The non-averaged method is first order accurate, while the averaged method is second order accurate for smoothly-varying materials and reduces to first order for discontinuous material distributions. For the standard FDTD field locations with the co-location of the electric and magnetic materials at the primary computational cells, the averaged method require development of the different inversion algorithms of the constitutive relations for the electric and magnetic fields. We provide a mathematically rigorous stability proof followed by extensive numerical testing that includes long-time integration, eigenvalue analysis, tests with extreme, randomly placed material parameters, and various boundary conditions. For accuracy evaluation we have constructed a test case with an explicit analytic solution. Using transformation optics, we have constructed complex, spatially inhomogeneous geometrical object with fully anisotropic materials and a large dynamic range of $\underline{\epsilon}$ and $\underline{\mu}$, such that a plane wave incident on the object is perfectly reconstructed downstream. In our implementation, the considerable increase in accuracy of the averaged method only increases the computational run time by 20%.