Abstract
In this article, by introducing artificial anisotropic (AA) parameters in the four-step hybrid implicit-explicit finite-difference time domain (HIE-FDTD) method, an AA four-step HIE-FDTD method is proposed, which can reduce the numerical dispersion error and improve the computational accuracy. Firstly, the Maxwell's matrix equation is decomposed into four sub-matrix equations by using the split-step scheme, and artificial anisotropy parameters are introduced, then the proposed AA four-step HIE-FDTD is obtained. Furthermore, the stability analysis of the AA four-step HIE-FDTD method shows that the stability condition of the proposed method is closed to the original four-step HIE-FDTD method. Next, the numerical dispersion characteristics of the proposed method are analyzed and compared with other HIE-FDTD methods. The results show that the numerical dispersion error of the proposed method is significantly reduced compared with the four-step HIE-FDTD method. Finally, the performance of the proposed method is further verified by the numerical simulation. Numerical results show that the proposed method has lower numerical dispersion error and higher computational accuracy than that of the four-step HIE-FDTD method.
Introduction
In computational electromagnetics, the Finite-Difference Time-Domain (FDTD) method is a cornerstone for simulation. However, it is constrained by the Courant-Friedrichs-Lewy (CFL) stability condition, which necessitates extremely small time steps ($\Delta t$) when using fine spatial grids. This leads to prohibitive simulation times for complex, miniaturized devices. Hybrid Implicit-Explicit (HIE) methods were developed to bypass this limit, allowing for larger time steps. However, increasing the time step typically introduces significant "numerical dispersion error," where simulated waves travel at incorrect speeds, compromising accuracy. This paper introduces an Artificial Anisotropy (AA) Four-Step HIE-FDTD method designed to maintain high computational efficiency while drastically reducing these dispersion errors.
Methodology
The proposed AA Four-Step HIE-FDTD method utilizes two primary innovations to enhance accuracy:
- Matrix Decomposition: Maxwell’s equations are decomposed into four sub-matrix equations using a split-step scheme. This allows the algorithm to treat specific directions implicitly for stability while keeping others explicit for efficiency.
- Artificial Anisotropy (AA) Parameters: The researchers introduced artificial parameters ($\epsilon_x, \epsilon_y, \epsilon_z$ and $\mu_x, \mu_y, \mu_z$) that function as mathematical correction knobs. By deriving an optimized dispersion relation, these parameters can be tuned to ensure the numerical phase velocity remains as close as possible to the physical speed of light, even at large time steps.
The resulting method is "weakly conditionally stable," providing a more relaxed stability limit than standard FDTD and enabling significantly faster computations without the high dispersion typical of unconditionally stable methods.
Numerical Dispersion Analysis
A critical part of the study involved analyzing the Maximum Normalized Numerical Phase Velocity Error (NNPVE) across different propagation angles and grid configurations.
As illustrated in the figure above, the dispersion error is sensitive to the propagation angle and the ratio of grid sizes ($\Delta x, \Delta y, \Delta z$). The AA parameters are specifically designed to "flatten" these error curves, minimizing the directional dependence of wave speed in the simulation.
Key Results and Performance
The AA four-step method was benchmarked against standard FDTD, ADI-FDTD, and previous HIE-FDTD iterations using standardized scenarios like 3D rectangular cavities.
Key performance metrics include:
- Dispersion Reduction: In a uniform grid, the Maximum NNPVE dropped from 0.1458% in the original four-step HIE-FDTD to just 0.02486% with the new AA method.
- Resonant Frequency Accuracy: In cavity simulations, the relative error for the $TE_{101}$ mode was reduced to 0.015%, compared to 0.245% for standard HIE-FDTD and 0.097% for traditional FDTD.
- Computational Efficiency: In non-uniform grid tests mimicking real-world fine structures, the AA method achieved FDTD-level accuracy while being over 25 times faster, reducing CPU time from 216 seconds to just 8 seconds.
Conclusion
The Artificial Anisotropy Four-Step HIE-FDTD method represents a significant advancement in balancing simulation speed and precision. By intentionally introducing counter-acting anisotropy to cancel out discretization errors, the method provides a robust solution for modeling fine-scale electromagnetic structures. While the implementation is more complex than standard FDTD, the dramatic reduction in dispersion error and computational time makes it a powerful tool for modern engineering challenges.
Summary updated