High accuracy reflection data from mode expansion


I’m designing integrated optical components which should have extremely low (<50 dB) reflections. I would like to simulate these structures with FDTD but I am having some trouble with the mode expansion monitors.

In the attached script I inject an optical mode into a Si waveguide and monitor the optical power in 2 micron intervals in front of the source. A mode expansion is then done on each of the monitors and the data is plotted in MATLAB. The simulation takes less than a minute to run on my PC.

Can someone explain why the mode expansion monitors show light travelling backwards in the waveguides, when light should only be propagating forward? Furthermore, why do the results (both transmission and reflection) change when the position of the monitor is changed? I’ve tried adjusting the PMLs and the simulation mesh size but neither eliminates the problem.

Any help is greatly appreciated.

Mode Injection_LumericalHelp.lsf (3.5 KB)

1 Like

I ran the simulation with a higher mesh accuracy of 3 instead of 2 and I found that the results matched up well when using the higher mesh accuracy. The discrepancies when using a mesh accuracy setting of 2 are likely due to build up of grid dispersion error due to the discretization of the mesh which is minimized as the simulation mesh is made finer.

You could also try using a mesh override region over the waveguide to make sure that the dimensions of the waveguide are accurately resolved.

1 Like

That was my original thought, but if you increase the accuracy further from 3 to 4 the problem comes back. I swept the mesh accuracy from 1 to 6 and found that a mesh accuracy of 3 was the only one which gave consistent results.

I looked into this again, and I noticed that you were using frequency domain profile monitors instead of power monitors for the expansion. The difference between these two types of monitors is the interpolation that is used - profile monitors use the “specified position” interpolation method which can involve an additional interpolation of the fields compared to “nearest mesh cell” which is used by power monitors. I found that when I switched the monitors to power monitors when using mesh accuracy 4, the difference was minimized.

Please try it out and let me know how it works.

1 Like

I applied your fix and it worked great. Thank you.