Divergence in a slot WG (Si-SiO2-Si)


I am trying simulate a simple slot WG. The WG is composed of 70 nm SiO2 topped with 300 nm of Si, the width of the WG is 400 nm. The WG is formed over a 300 nm Si layer deposited all over the substrate. My simulations are at 1550 nm wavelength, where materials are well defined.

The cross-sectional view of the WG:

The excited mode:

The design is simple, but there is a problem in this simulation. The electric field exceeds 1.

I tried to change the PML types. I followed the steps on knowledge base for diverging simulations. I tried to replace the PML with metal boundary conditions in all directions. The results got worse. I tried to use different PML: stretched coordinate PML and uniaxial anisotropic PML. I tried choosing custom PML and changing the parameters manually; also, I set the PML to stabilized. I tried to reduce the FDTD mesh and adding override mesh to reduce the reflections. None of these trials helped reducing the divergence problem.

The results I get:

Really I don’t know where is the problem. If any one could help I would be very grateful. This is the first step in my project and I am stuck in it for more than two weeks.

The file:
str8 WG.fsp (1.4 MB)



Hi @mohamed.yousef,

It is not really a divergence issue (otherwise the fields would increase exponentially during the simulation and you would get a specific error message). I think the issue might be the PMLs are too close to the waveguide, and they interact with the tail of the mode when propagating (especially laterally, in the y direction). You can try and increase the distance between the waveguide and the PMLs and see how it affects the simulation.

As a test, I modelled the waveguide with MODE Solutions, even with increasing the simulation region, the fields seem still high near the boundaries:

The profile is in the log scale and I set the color bar to -9 as a minimum, to see where the fields are below 1e-9.

I hope this helps.


Hi Mohamed,
I have worked on ur file to solve the divergence and decrease the simulation time. I can see no divergence even when I check the time monitors at the input and output of the waveguide.

Here is the updated file str8 WG.fsp (1.5 MB)

The modifications in a nutshell:

  1. made the structure y symmetric and forced symmetrical y meshing
  2. used y symmetric boundary conditions
  3. used metal BC for y and z directions ==> just to accelerate the simulation. However, this is not useful if u r considering a bent waveguide.

I obtained the following E intensity in the oxide layer. To avoid getting E intensity greater than one, I advise you to increase the simulation time so that the time pulse reaches the other end of the simulation region.

Hope this helps! let me know if u have any questions.