Second mode is TE mode (FDE to simulate mode profile)


Hello all,
I did the simulation of modes profile for different dimensions of the waveguide cross section. I used PML for boundary conditions.
For the simulation results, I expected that the first supported mode is the fundamental TE mode, and second is fundamental TM and then third supported mode is the second TE.
By looking at the mode profile and the fraction of TE polarization, the second supported mode looks like TE mode (the TE polarization fraction is 96). The field profile of the third supported mode is unlike a typical mode profile of fundamental TM mode, although the TE fraction is 4. (This simulation is done with 1u*180nm strip waveguide). The result is shown below. The used simulation file is too large to upload. It can be viewed in wa_sim dropbox link.


Dear @p.liu-2

Thank you very much for providing simulation file and a detailed explanation of the problem.

I think your results make sense. Since waveguide is wide (5 times more than its thickness), second order TE mode might possess higher effective index than TM modes. However, there are a few things that you can try and re-analyze the results as part of convergence testing:

  • Use metal BC instead of PML as simulations with metal BCs are faster compared to PML. Here is a good link that compares these two BCs:

  • When you use PML, you need to increase the FDE span large enough that electric fields are decaying enough before reaching PML. You can use log scale plots and make sure that fields are decaying 5-10 orders of magnitude before reaching the FDE boundaries.

  • For cases that FDE span is not large enough and you use PML BCs, you might see some negative loss due to electric field interference with PML. You can also use finer mesh as part of convergence testing. The converged results will not vary with simulation region or mesh size. Please see this link how for a complete list of convergence tests to examine:

Also, if you want to upload simulation file directly into the forum, you can save them in layout mode rather than in the run mode. Simulation file in the layout mode will not have data, thus will be very small in size.

I hope this was helpful.



Hello all,

I used mode solution eigenmode solver to analysis a waveguide. I’m trying to find the possible thickness range for a 1 micrometer width monomode waveguide. However, I found the number of supported mode and the mode profile excluding the fundamental mode that depend on the parameters of material fitting.

Here are the results of two simulations.
The waveguide structure and simulation settings are exactly the same except the material fitting parameters.

  1. Which one I would trust?
  2. Also the fundamental TM mode profile looks different as my expectations.
  3. Also what is the negative loss meaning?

Thank you all!


Dear @bkhanaliloo,

Thank you for all the explanations and comments.

  • I still don’t clearly understand why the second supported mode is TE that is reasonable. Could you help me to understand it in more details.

  • I used metal BC at the very first time since it saves the simulation time. And the given results for the modes are TE, TM.

  • I actually did the convergence test of FDE simulation area with PML BCs, shown below. The chosen area size is 2.5um3.5um (zy). The corresponding effective index error is at level of 10e-9. Therefore, I think the simulation area size should be large enough.

  • Could you take a look about my second question about material fittings?

Thank you so much.


Dear @p.liu-2

Thank you for providing additional info.

  • Since your waveguide is wide, you are supporting second order TE mode with higher effective index than first TM mode. For TE ™ modes, you expect to have a larger Ey (Ez) component. You can check this by plotting different component of electric field. For example, for second mode we have:

As you can see, Ey component of the second mode is larger than Ez, and mode is quasi-TE polarized.

  • You are right. I used metal BCs with FDE mesh region of y-span*z-span = 8um * 6um. The effective index that I obtained for the first mode, neff=1.64, is same as what you obtained:

Plot is in the log scale and field has decayed few orders of magnitude before reaching simulation region.

Regarding the material fitting:

Both results are very similar, but the material fits in the first case look better than second case. I am not sure how precise you want your results to be, but this is a good link to optimize the material fit. Please let me know if you had any concerns regarding the fits or if I missed any point.

Also, what does not make sense about third mode? Can you provide any references that supports your claim?