How to get propagate loss and coupling loss?

In this trident edge coupler structure, I want to calculate propagate loss, coupling loss and simulate fabrication tolerance of this coupler, can FDTD solutions directly get information about loss? Or Should I export the data to Matlab to calculate by hand?

Dear @mux17

FDTD can calculate the total loss that includes both propagation and coupling loss.

We have the bent waveguide application example in which we separate the coupling loss from bend/propagation loss:

For complex geometries separating the loss factors might not be trivial. You may try to simulate the tapered waveguides individually to calculate the propagation loss, and then calculate the coupling loss. For these structures, EME is the most efficient software to use:

Hope this was helpful.

Low Polarization-Dependent-Loss Silicon Photonic Trident Edge Coupler Fabricated by 248 nm Optical Lithography.pdf (340.0 KB)
trident edge coupler 3.fsp (313.3 KB)
I want to simulate an edge coupler in FDTD(we just buy FDTD, no MODE solutions) according to this paper(you can see the attachment). I met two problems, one is Gaussian beam decrease too fast and the loss is to large; the other problem is that it seems no coupling between the middle taper with the other two side taper.
This simulation profile may take 5 hours to run, because the structure is large(300um in length), but I sincerely hope that you can help me with it! Thanks!

Hi @mux17

For exciting the waveguide modes we use mode source or port object:

We can do overlap calculations to calculate how much light gets coupled from a Gaussian beam to any waveguide mode:

However, if you want to simulate Gaussian beam in FDTD:

  1. Make sure that its location, polarization, and beam properties is set accordingly. Currently, Gaussian beam is located inside waveguide, however, I was expecting it to be injected in free space.

  2. Since the beam waist is 5um and distance from focal point is set to zero, you need a much larger y- and z-FDTD span, of at least 5-10um. A good approach would be to increase the y- and z-span of FDTD and source and visualize the (Gaussian beam) source fields (select the source, then visualize->fields before running the simulations). Fields need to decay few orders of magnitude at the edge of FDTD boundaries. Please visit Advanced tips section of the link below for more details:
    Brewster's law

As I said, an easier approach would be to inject a waveguide mode using MODE source or Port object.

The second problem might be due to improper Gaussian beam setting mentioned above, which we can discuss it if problem was unresolved after fixing the first part.

Please keep me updated with your decisions on what type of source you want to choose, and we can discuss it further if you had any questions.


Hi, I still chose Gaussian beam as my source since I want to find where is wrong compared to the structure in the related paper, because my goal is to repeat the simulation result in this paper. I wonder why simulation results differ so much with the same structure.
trident edge coupler 5.fsp (312.7 KB)

I put Gaussian source in free space as you recommend and enlarge y-span and z-span of FDTD region to 15um and make sure that fields decay few orders of magnitude at edge of simulation region(As shown in following figure)

But the simulation result goes wrong. From the simulation report log file"Early termination of simulation, the autoshutoff criteria are satisfied" ,where the simulation run only 2%.
trident edge coupler 4_p0.log (2.7 KB)
Besides, all visualizers of every monitor become blank. The intensity of electric field become zero. Acctually, I don’t understand why I should put light source in free space, doing so will lead to strong reflection and scattering on the boundary, right?
Sincere thanks!

Hi @mux17

Currently Gaussian beam source is located outside the simulation region. Thus, it will be ignored and no light will be injected into simulation region. You can increase the x-max of FDTD to include the Gaussian beam.

Another thing to note is about is Gaussian beam setting. Please make sure that waist radius and distance from waist are updated based on your experimental setup. This will have a large impact on how efficiently you can couple light into your waveguides:

Also, after you adjusted the source parameters, plot the Gaussian beam in log scale using visualizer settings to make sure that field has decayed at least 3 orders of magnitude at the source edges.


Thank you!
I don’t quite understand the parameter “distance from waist”. It is the distance d between injection plane and beam waist, but in this paper the author don’t describe how the light source is placed. i don’t know how to set these parameters of Gaussian source.

Besides, curve viewed through visualizer of time monitor is zero, and time monitor is placed inside the FDTD region. How it comes up?

This time I use MODE Solutions as the tool to simulate the same structure and apply mode source as the light source. But the electric field distribution is still strange.
1trident edge coupler.lms (339.2 KB)
But it is much faster than simulation using FDTD solutions. Thanks!!

Hi @mux17

I have modified your simulation file. Specifically the simulation region and source span:
trident edge coupler.lms (363.0 KB)

and here are the results that I obtained:

For fundamental TE mode:


For fundamental TM mode:


One thing to note here is that varFDTD cannot simulate vertical coupling. Thus, you will need to modify the gap distance between the waveguides manually. This is explained in details in the application example below:

You can use EME solver for fast simulations instead. The link below has a few spplication example for your review:

We also have online free EME course for you to enrol:
New EME Solver online course now available at Lumerical University!


varFDTD trident edge coupler 3.lms (413.9 KB)
Thanks for your attention! I enlarged simulation region and y span of the monitor and improve mesh accuracy to 4 to repeat this simulation as in attached lms file.
I got xy view TE power distribution as in this figure
What I want to get is like this picture appearing in the reference paper. But there are difference between this two pictures. In my simualtion result, power intensity inside the tapers is lower than that outside near the taper,which is strange(I think power inside the taper should be stronger than power outside, as shown in the paper). Besides, in my picture, it seems that taper didn’t confine light well since light beam is so large outside the taper, light diffuses and spreads . As we can see, the shape of light propagation is just the shape of the taper in the reference paper, without much transverse spreading. Thanks!

Hi @mux17

Thank you for elaborating on your inquiry.

varFDTD cannot capture the vertical coupling:
varFDTD vs EME

As you can see in the evanescent waveguide coupling example, the gap between the two waveguides is modified to compensate for this limitation. In your case, you may need to modify the gap distance between waveguides and see how the results are modified. It is possible that you are not getting the same field profile as the ones reported in the paper because the coupling coefficient is not the same.

EME can provide you with a much more precise results. Have you tried running simulations in EME? We have an online course in EME if you were new into the product:
New EME Solver online course now available at Lumerical University!

Hello @bkhanaliloo,

Even I want to calculate the coupling loss for a microring resonator in MODE solutions. I tried to open the link, “”, but this link isn’t working anymore. Could you help me on how should coupling losses be calculated?


Hi @aniruth

We have made some changes into our documentations and now we have KB and Application Gallery. For old links, simply search the key words in our website like this:

The link to the requested page is:


Hey @bkhanaliloo,
How can I calculate coupled power between two waveguides? Is there any equation that lumerical follow. I specifically want to know the equation.


Hi @mborhan.mia

We use ovelap script command which returns overlap and power coupling between the two waveguides:

Hey @bkhanaliloo,
Thanks for your reply. I have one more inquiry:
What are the use of coefficients in EME analysis? Is there any relation of coefficients and internal s matrix (or user s matrix )?

Hi @mborhan.mia

The coefficients result gives forward and backward propagating coefficients for each mode in each cell for the given source excitation, which allows you to see which modes power gets coupled into as it propagates.

Please see the below EME course link:

Thanks @bkhanaliloo.

A post was merged into an existing topic: Distance from waist while setting gaussian source?