Mode mismatch loss discrepancy

To help us solve your issue as quickly as possible, please provide the following information:

1. What device are you trying to simulate? Include diagrams if available.
I am trying to simulate a bent waveguide in Lumierlcal mode

2. What results are you trying to obtain? Be as specific as possible.
I am trying to obtain the mode mismatch loss between a circular bend and a straight waveguide.

3. Description of the problem or issue.
I am trying to find the largest radius that is required for two-mode mismatch losses of 0.1%. In other words, what is the radius that will result in two mode mismatch losses (between the 90 degree bend and the two waveguides intersection) that will result in 0.999 transmission. The waveguide dimensions are 500x220 nm strip silicon waveguide buried in Silica cladding, and the mode is the fundamental TE. I did this simulation 3 years ago and the answer was 10 microns. I think even last year that answer was 10 microns. I did this simulation last week and the answer was 8.5 micron. I am wondering if something has changed in terms of the waveguide center offset in the latest version, or am I doing something wrong?

4. Lumerical product and software version.
Release: 2019b r6, Version: 7.14.2088

5. Please attach your simulation files, including any logs.
File attached.

Hello @hoshoman,

Thank you for the question. Yes, there has been a change in the definition of the center of radius of curvature in recent versions. From the Solving bent waveguides in FDE overview page:

So if the width of the boundaries is not the same on both sides in your simulation, the center of radius of curvature will be different than the value from past versions.

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

Dear Johnson,

Thank you very much for your reply. The boundary conditions used were PML with the same number of layers on both sides. I still got a different answer from what I used to get previously. Is there a way to test the same simulation in the 2018 and 2020 versions? I have the latest version now (2020a r6 7.15.2305), but not the 2018 anymore.

I am not sure what else could be causing this issue. Could you please attach your simulation file so I can take a look?

Hi Johnson,

Sorry for getting back to you late. Our edX course just started and thought to check on this again. For some reason, I can’t attach the simulation files, so I am pasting a link to the files here:
https://1drv.ms/u/s!AnDrMVCWSyAtjKQBYnk8h3n_WEYG2A?e=b229vr

I am trying to find the largest radius that is required for two-mode mismatch losses of 0.1%. In other words, what is the radius that will result in two mode mismatch losses (between the 90 degree bend and the two waveguides intersection) that will result in 0.999 transmission. The waveguide dimensions are 500x220 nm strip silicon waveguide buried in Silica cladding, and the mode is the fundamental TE. I did this simulation 3 years ago and the answer was 10 microns. I think even last year that answer was 10 microns. I did this simulation last week and the answer was 8.5 micron. I am wondering if something has changed in terms of the waveguide center offset in the latest version, or am I doing something wrong?

Hello @hoshoman,

I don’t think there has been a change in the definition of the waveguide center offset other than the change I mentioned in my previous post. I tested your file with the current version and with version 2019a R3. Here are the overlap results with different bend radii:

2020 R2.2:
(bend radius) - (overlap with straight waveguide mode)
8.5 um - 0.999524
10 um - 0.999656

2019a R3:
8.5 um - 0.999372
10 um - 0.999547

The results were the same with metal BCs. I also noticed that there were small differences in the effective index results for the bent waveguide modes. There may have been some adjustments made to the bent waveguide algorithm that cause these differences, or the default settings may have changed. I will have to look into this and get back to you.

Hello @hoshoman,

A member of our R&D team directed me toward a bug fix in version 2019b R6 for the bent waveguide solver. Here are the results with 2019b R5 and 2019b R6:

2019b R5:
(bend radius) - (overlap with straight waveguide mode)
8.5 um - 0.999372
10 um - 0.999547

2019b R6:
8.5 um - 0.999524
10 um - 0.999656

The results change in version 2019b R6, so it is very likely that this bug fix caused the change in your results. The new version should be more accurate with this bug fix, so I would recommend that you use the latest version of the software for your calculation.

Let me know if you have any questions.