Learning Forum

[Daniel Marchant]

I am having a problem with diverging simulations in Lumerical FDTD. I am trying to simulate am asymmetric directional coupler which runs fine for longer wavelengths (~1600 nm) but when I change the wavelength to ~1000 nm, the simulation diverges.

I have tried the following:- Change to metal boundary conditions

- Reduce dt stability factor (I have tried reducing this to 0.95, 0.9, 0.5 and 0.1)

Neither of these seem to resolve the issue. Any insight would be greatly appreciated.

[Amrita Pati]

Hi Daniel,

Are you running single wavelength simulation or a broadband simulation? Based on the tests you have already run, I would guess that it is material fitting. If it is a broadband simulation, you can look at the material fitting for all the materials. Sometimes artifical peaks (typically seen in highly dispersive materials like Ag) in the material fitting can cause instabilites resuting in divergence. 

I will also recommed you to use a movie monitor when you run the simuation. It can tell you where exactly the fields are growing. Would you also be to share more information about the materials you are using? A screenshot of the geometry would also help. Thank you!

Regards,
Amrita

[Daniel Marchant]

Hi Anita, thank you for your quick reply.This is a single wavelength simulation (simulation bandwidth of 0.3 um which is the standard set by Lumerical FDTD, I also tried reducing this to 0.1 um but that did not resolve the issue). The materials used are Silicon Nitride (for the waveguide) and silicon dioxide (for the cladding).

I think you may be correct about the issue being related to the material fitting, but I am not sure exactly how this is causing the issue. As I mentioned, the simulation runs fine for a longer wavelength and I use similar values for the fit tolerance, number of fitting coefficients and the fitting range. 

I have attached some images from the movie monitor that show the issue.

Kind regards,

Daniel

[Amrita Pati]

Hi Daniel,

Thanks for sharing additional information! If it is a single wavelength simulation, I wouldn't expect the fitting to be very problematic. In broadband simulations, the fitting can have diiferent features in different wavelength ranges and can lead to instability, even with the same fitting parameters. Would you be able to share a screenshot of the fitting from the Material Explorer of the current simulation (the one that is diverging)?

If it is a single wavelength/narrowband simulation, what happens if you run the simulation with a fixed refractive index. For example, instead of using SiN you set the material as Object Defined Material and set the index field as the corresposnding value at that wavelength?

Also, in this region where you see this growth of electric field, I suppose there is no unique feature? 

Regards,
Amrita