Ag/Au coating on a fiber end



I would like to simulate reflection from 5-10 nm layer of Ag or Au at the end of a single-mode optical fiber - core ~4 um, cladding ~125 um. I have tried it both in FDTD and MODE Solutions. When I tried to decrease the mesh size (to around 10 nm for region 30x30x5 um) I was not able to calculate even the mode profiles in the fiber.
Is there a way to do it effectively? Do you have any hints I might want to try?


Hi, @mdudek !

Ir the RAM enough for calculation? For 10 nm mesh it is 4.5E9 points what needs to be operated on.
To get a hint on what can be wrong, could you attach your simulation file?


On my machine I have 96 GB of RAM.
The simulation file is here. When I try to change the mesh size to 0.01 um or add increased mesh region in the area around the end of the fiber and then try to calculate the source mode, the whole software freezes and crashes.
Ideally I would like to have the Ag/Au thickness around 5 nm, but even 10 nm for now seems not to be possible.


When I change the mesh size in all 3 dimensions from 50 nm to 10 nm, the source profile updating very slowly and sometimes it may see as freezing. However, after several minutes it starts modeling, though, demanding several hours on my laptop.

I would suggest making the mesh finer in the z direction in the metal film only, say 1/10-1/50 of the thickness, and much coarser in other dimensions (including the z dimension outside the film). Am I right that you was doing the similar ?


Dear @mdudek

I think I am on the same page as @msaygin i.e. increase the mesh size on the x, y, z direction of the fiber, and use a mesh override over metal on the z-direction. One important thing to note here is that if the mesh aspect ration is too high (say 10:1) at any point in the simulation ergion, and specillay near PML layer, your results and PML behavior might not be very reliable. See this link for example:

But other than that there is not much to do. You are dealing with a large simulations which requires fine mesh to resolve the objects. I was working with a similar case during my grad studies and I knew the difficulties of this type of simulation. They require large memory and take a long time to run.


Thank you both for your comments!

I will try to do it as you suggested - lower even more the mesh size in the z direction near the metal region and keep larger mesh in x and y directions.

One more thing. Do you think it would be better to try it in MODE Solutions with EME solver?


Dear @mdudek

I loved the idea and YES, it should be possible. I think you need 3 EME cells (one fiber, one metal, and one air). When you calculate modes in the air, you will require a larger number of modes. This is explained in this application example:

Please keep me updated with the results.