Gold surface unexpected loss


I am working on a dielectric metasurface mirror, where the reflection is applied by a 20nm gold surface. The reflectivity of the surface alone appears to be ~83% in the simulations for 850nm wavelength. When there are SiN dielectric material on top of the gold (periodic BC on sides, so the given structure is the 2D unit cell), R starts to depend on the dimensions of the material, where I’m varying its width and length for a given height and unit cell length. The dimensions I went for is 175nmx390nm. However, for this case, R goes down to ~58% and transmission becomes 26%, so there exists a significant loss in the gold and T is larger than the gold surface itself. I believe this might be a simulation issue, because I would not expect this much loss and also the results depend on the simulation settings (eg mesh level, mesh refinement, pml profile) too much. I could not diagnose the problem and hoping you could help me with it. Thank you very much.SiN_rect.fsp (3.1 MB)

Hello @dkuru19,

Thank you for the question. I made some changes to your file, including:

  • Added a mesh override to the gold layer
  • Moved sources away from the dielectric
  • Increased wavelength range
  • Added an absorption analysis group

Here are the absorbed power, transmission and reflection results with and without the dielectric I obtained after making those changes:



The transmission results are reasonably close, as would be expected. I am not sure why your transmission results were so different.

I am not convinced that the increase in the absorption losses is a mistake, it may be caused by the fact that the dielectric forms a resonator structure that traps light close to the metal layer. For example, here is the Poynting vector results from your Fieldxz monitor at a wavelength of 856 nm:


The fact that the results change with varying simulation parameters indicates that you should perform convergence testing to determine the correct parameters for this simulation. In particular, the results will probably be sensitive to the meshing of the gold layer.

Here is the file I used to get those results: SiN_rect_new.fsp (344.1 KB)

I hope this helps. Let me know if you have any more questions.

Thank you for your reply, it was very helpful. I have another inquiry about the same setup. I used PEC layer as the reflective surface. At certain wavelengths, the reflectivity goes down significantly, the field is too large at the boundary of dielectric and PEC, and the total absorption is non-zero. Reflection looks like this:R
I am unsure if my simulation is correct, as PEC should be a perfect reflector, and this happens for specific wavelengths. I did convergence testing, but could not solve it. The best result I got is this (differences are PML z span from 3um to 10um, layers from 12 to 24, mesh from 2 to 4):R_10um_layer24_mesh4
But, there still is some problems with it. I have attached the simulation file in the settings that gave the second result. I would appreciate any help clarifying this result. Thank you.
Unit_PEC.fsp (313.0 KB)

I believe your setup is correct in general. However, the ripples in the reflection spectrum indicate that your simulation is not running long enough. With a resonator structure like this you may need to turn off the autoshutoff feature of the FDTD region and increase the simulation time to make sure all of the light escapes the simulation region.

Try increasing the simulation time and let me know if that improves your results.