Fitting questions of varFDTD


#1

Hi,

I have two questions regarding the fitting in var FDTD.

  1. Please have a look of the following lms files.
    https://drive.google.com/file/d/0BxkLVH0fBO9tOHpXeHU0MUgxZnc/view?usp=sharing

The only difference between case1 and case2 in the above zip files is the imaginary weight, which is 10,000 in case 1 and 100,000 in case 2. I have checked and found out that it does not influence the fitting curve at every component, but the results from the two files differ a lot.Below are the results of optical power before and after passing through the ring (obtained from the “before” and “after” monitors), and we can see that the transmission is much lower in latter case, which is unreasonable.

So just wan to know what is the reason for that phenomenon, and which part should we check before running the simulation to prevent simulation situations happening again?

  1. Please have a look of the following input files.
    https://drive.google.com/file/d/0BxkLVH0fBO9tWWpPUUlfU0lpX0U/view?usp=sharing

The structures and the mesh point for the core material are exactly the same in those two cases. However, one of them fits the core material with a straight line, as expected. But the other’s (trial3) data point on the core material imaginary part fitting has a weird curve behavior shown below. Therefore no matter what I do to adjust the fit, it will always mimic the curve.

Again, what is the reason for the phenomenon and how do we treat it?

Just post here to let me know if additional information is needed.

Thanks,

Pufan Liu


#2

Dear @pufanliu2021

Can you please explain what you are trying to simulate? specifically why you are using a Gaussian source?

Basically, I doubt that the behavior that you are seeing is the result of material fit.There are a few concerns regarding your simulation file which without addressing them making any conclusion will be hard. You need to adjust your simulation region so that it does not truncate your source and choose core slab mode, represented by green cross, properly (on the waveguide). I guess these this 1 and this 2 KB examples will be a good start.

Thanks


#3

@bkhanaliloo

Thanks for your response. I am using a Guassian source so that I can adjust the polarization angle. This is used to understand current experimental data.

I see your concerns about the simulation structures, this might be one factor influencing the results. However my question is on the behavior of the fit. Basically, what is the reason that the results are different in question 1, and the fitting data points are different in question 2? Given that the structures, source, slab mode position are exactly the same in both cases, I doubt this behavior has something to do with the structure or source.


#4

Hi @pufanliu2021

Can you please explain your experimental setup? Since you are using waveguide to route the light and couple it to ring resonator, I am not convinced that Gaussian source is a proper choise.

Observed transmission is less than 1% while we expect ~100% at some wavelength. This might be noise or light that is not coming from waveguides. As I mentioned, further decision and analysis requires proper simulation file.

Thanks


#5

@bkhanaliloo

First thing first, the 2nd problem has been resolved. By changing the boundary condition in z direction from PML to metal solves the mystery. Perhaps the effective index method used by Lumerical found it difficult to calculate the effective index in PML boundary conditoin. However, the 1st problem still remains.

Can you please explain your experimental setup? Since you are using waveguide to route the light and couple it to ring resonator, I am not convinced that Gaussian source is a proper choise.

A polarizer is placed between the laser and the waveguide to adjust the polarization angle of the input light source. Yes you are right, since it is guided through the waveguide, it is perhaps better to use mode source than Guassian source. However, in mode source it is not quite easy to study the results of, say, 75 degree polarization angle, where both TE and TM are excited but with different magnitudes.

Observed transmission is less than 1% while we expect ~100% at some wavelength. This might be noise or light that is not coming from waveguides. As I mentioned, further decision and analysis requires proper simulation file.

That is because my monitor setting is a point not plane. So of course only a small amount of power is passing through the monitor. Guassian source may be another problem, as not all of the power are entering into the waveguide. But I am more interested in the relative transmission before and after passing the ring than the absolute transmission so this is fine.

I see you put a lot of efforts in pointing out some potential problems in the simulation design. Firstly thanks for that, but let me restate my 1st problem, as my problem is kind of independent of design.

Two identical simulations are performed, with the only difference to be the imaginary index weight in the varFDTD effective index tab. I have checked, however, that the fitted effective index is insensitive to that change, i.e., the fitted effective index at every point is exactly the same under those two simulations. So I am confused to see their simulation results differ a lot. Are there still some hidden variables present? If so, then what should we do to check for them to ensure the consistency of the simulation?