Optiwave

OptiFDTD: Non-Uniform Mesh

The strength for non-uniform FDTD is to preserve memory usage and consequently save CPU time.

Why Non-Uniform Mesh?

The FDTD method requires the entire computational domain be meshed. The mesh spatial discretization must be sufficiently fine to resolve both the smallest electromagnetic wavelength and the smallest geometrical feature in the model. Very large computational domains may result to large memory usage and heavy CPU simulation time.

The strength for non-uniform FDTD is to save the memory usage and consequently to save the CPU time. In the case of a high refractive index region, a fine mesh should be used, while in the low index region, the coarse mesh can be applied.

For normal dielectric material, numerical dispersion analysis shows that to get accurate results, the mesh size should be at least 10% of the simulated smallest wavelength in medium; for surface plasma layout where metallic material is usually needed, the mesh size should be even smaller to catch the wave attenuation within the skin depth.