Hi. We were able to modify the file and get pretty decent result for transmission. The transmission spectrum that we got is shown below,
This is very close to the figure in the published article. The dip looks slightly shifted but that is probably due to a coarse mesh and you can just to some convergence testing to correct it. Here is what we did to fix the simulation,
The index of the substrate dielectric was changed to 3.5 (from 1).
In order to ensure that the PML does not introduce any reflection in the simulation volume, we placed the simulation boundary approximately half a wavelength away from the gold layer both at the top and at the bottom.
The number of layers in the PML was increased to ensure that the PML itself was thick enough to properly absorb the transmitted light. We ended up using 64 layers, a much lager number than the default 8 layer.
Finally, and this is probably the most important point, we changed the material model for gold. In the paper, they used the drude plasma model to create the gold material at THz frequencies. FDTD has a built-in material type for the drude plasma model called “Plasma”. We used this material model and used the values provided in the paper to create the material model.
You can see that the resulting model is quite different from the model you had created,
One key point to note is that the real part of permittivity in the new model is much larger and also negative. This is an important requirement for the creation of surface plasmons. With the new model and the other modifications in place, the transmission data looks quite reasonable.
Here is the modified file: au-frame_mod.fsp (259.1 KB)