Dual waveguide arc (bent waveguide) coupler with large bend radius


I want to simulate a coupler made of two bent waveguides, with relatively large radius of curvature (pic attached), and want to calculate the transmitted/coupled power. I have made the model using EME by modifying the directional coupler comprised with straight waveguides. However, the power monitor shows that the power somehow doesn’t propagate beyond 1mm. Where is it going wrong? I am not sure if varFDTD would be useful for the same model considering the structure extends in mm scale, and we cannot characterize the structure with a constant coupling coefficient.
Please advise.double arc coupler Support20181207.lms (3.8 MB)

Best approach for simulating large ring resonators

Hi @sxd

The light gets scattered when it tries to calculate the coupling from one cell to another. This is because mode profiles at different cells are different (xy positions change as waveguide positions within cell changes). One approach that comes to my mind would be to use only group span and set the bend radius accordingly for the cell group as is explained in the link below:
Is it possible to simulate bent waveguide using EME?
I have not tested this myself, but please keep me updated with your results.

Also, I was wondering if you could simplify the case with two straight waveguides. Do you expect that the results would be different between the straight and curved cases?

You can use varFDTD, but you will need to correct for the coupling coefficient by modifying the gap. Please see the varFDTD simulation section in the link below for more details on this regard:


@bkhanaliloo : Thank you for your suggestions. I am responding to them below.

  1. Unfortunately, I cannot approximate with two straight waveguides as there are other factors besides geometry involved. The schematic is the simplest form of the structure to help me understand how to tackle the geometry effect alone.
  2. One approximation that I can probably get away with is to assume that one of the waveguides is straight, very much like the ring modulator (coupler region) example. But even in that file, the geometry suggests that the coupling is happening along the region where the 2 waveguides are parallel. If I modify the file and put coupler region length=0 (bent coupler is tangential to the straight coupler), the model doesn’t yield any coupling effects…
    Any suggestions on how to modify the code?
  3. Theoretically, if there are infinite cells and infinite number of modes in each cell, the EME solver should work. So, I tried to increase the number of cells to 50 in the 5mm coupler length and the number of modes in each cell to 20. It increases the simulation time but did give some observable results (attached). I am planning to write a code to set the cell parameters for the EME solver since individually setting the cell parameters for so many cells is not feasible. In fact, if this functionality can be included in the EME solver, it would be great.
    EME solutions 20181212.pdf (292.1 KB)
  4. VarFDTD doesnt seem very practical either. It would be difficult to correct the gap for the configuration where the coupling (/gap) itself is not constant along the direction of propagation.


Hi @sxd

Thanks for your inputs.

If the gap distance between two curved waveguide varies then you cannot use one cell for the entire geometry. This means that you need to increase the number of cells (one group span is still enough) and modes to achieve reasonable results. As always, you will need to perform convergence testing:

In the ring example, if you set the coupler region length to 0, then you need to reduce the gap size. Then you will be able to see coupling between the two waveguides, it might be small though.

I think having one group span and increasing number of cells will solve the mentioned problem on #3. However, here is a good link to help you script EME simulations:

Hope this helps.