varFDTD accuracy for a ring coupler simulation


I am trying to simulate the ring coupler which is the coupler part of a ring resonator. I have applied varFDTD to compute the cross-coupled power (the power coupled from the bus to the ring or vice-versa) but the result is different from the theoretical method I have applied.

I wonder how accurate and reliable the varFDTD results for the coupled power would be in a ring coupler?

Hello @hghor044,

Thank you for the question. The FDTD algorithm makes few assumptions beyond discretizing time and space, so it is, in general, fairly accurate. Theoretical models usually make several assumptions and approximations in order to obtain analytic solutions so they tend to be less accurate than a full 3D FDTD simulation.

The varFDTD solver essentially makes an effective index approximation in the z direction. This is accurate for planar structures where there is negligible coupling between slab modes, for example in SOI structures. The varFDTD solver is mostly intended to be used for quick initial simulations to narrow down your range of parameters. Generally, if you are simulating a planar structure, you should perform your initial parameter sweeps and optimizations with varFDTD to save time, then perform your final simulations with the full 3D FDTD solver for maximum accuracy.

Calculating the coupled power into a ring resonator is a common application for the FDTD method. It should be more accurate than any theoretical model that I am aware of for couplers. You might be interested in these posts that discuss extracting the coupling coefficient from FDTD simulations:

In summary, the varFDTD simulation should be fairly accurate for this application, but you should perform your final simulations with the 3D FDTD solver to get the most accurate results.

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

Hello and thank you for your response.

I just simulated a straight waveguide directional coupler but they were so close (1 um gap) such that I had to reduce the size of mode source as well as power monitors. The coupled power amount was quite different from EME method that I believe was accurate enough.

What could be the reason for this error? Is it due to the very tiny width of source and monitor windows?

Thanks in advance

Yes, this is probbably due to the span of the monitors. It can be difficult to place monitors and sources for closely spaced coupled waveguides. You want the span of your mode sources and field monitors to be large enough that the mode field decays to close to zero at the boundaries.

For a directional coupler, you could try abruptly introducing the second waveguide after you have placed your mode source on the first waveguide:
Or you could try gradually introducing and removing the second waveguide like the image below, which is a more realistic geometry:

Both of these geometries would give you more space for your sources and monitors. You might be interested in this example, which discusss simulating directional couplers using various methods:

Let me know if you have any questions.