Learning Forum

[aberg]

Hi, I am following this linear taper example (https://support.lumerical.com/hc/en-us/articles/360042800413-Linear-waveguide-taper) but I would like to add a mode field monitor to visualize the YZ mode (X-normal) instead of the XY mode (Z-normal). I have added a new monitor (cross_section_mode_monitor) and disabled the existing one (Field_profile) accordingly, but when I run the simulation, no results are loading into the new monitor and I cannot visualize the YZ field. Will you please let me know what I am doing wrong here? This forum does not allow me to upload .lms files so I have attached a screenshot of the new YZ monitor settings. Is there a way I can upload my .lms file? I am using Lumerical through a university license.

Also, what is the best approach (EME/varFDTD/FDTD?) if I want to simulate the output only of an adiabatic waveguide taper with sub-wavelength features? I am not optimizing the taper length, etc. but only need an accurate mode profile at the narrow tip of the taper. I will then calculate the overlap between the taper mode and a fiber mode. Thank you.

[Guilin Sun]

Did you run "emepropagate"? I tested and there is no any issue showing yz profile:
For common taper with small cone angle, EME is perfect to design and optimize, Here are some examples:
HAMR - tapered waveguide
Linear waveguide taper
Polarization converter using a tapered waveguide
If the taper cone angle is large, you may try varFDTD and 3D FDTD:
Curved waveguide taper (varFDTD and FDTD)
As for how large the angle is large, it depends. It does not mean EME or FDTD cannot simulate such device, it is the simulation efficiency and accuracy. When the device angle is too large, EME will need more group cells, which leads to long simulation time. In addition, if you want very broadband result, or the mode dispersion is strong, EME may also need long simulation time. For FDTD it can theoretically handle all types of tapers and can give broadband result. However if the device is larger and mesh is fine , FDTD may need long simulation time. So it depends on several factors, and there is a trade off between simulation efficiency and accuracy.

[aberg]

Thanks for your reply! Please note that I am using the Linear taper example (https://support.lumerical.com/hc/en-us/articles/360042800413-Linear-waveguide-taper), not the MMI example in your post above. I did indeed run eme propagate. But the YZ monitor I put at x = -0.5 (middle of left narrow waveguide region) did not update with any results.
However, I noticed that when I move my YZ monitor to the right of x = 0 (for example to x = +0.5) and re-ran eme propagate, the monitor updated with results and I could visualize the mode. I do not understand why the monitor failed to update with results when it was to the left of x = 0, but updated with results when placed to the right of x=0. The location of the monitor along the waveguide taper should not matter, correct?

[Guilin Sun]

I do not see the difference since they all use the same EME solver. Here is my test for the example you specified:
Please try another example.

You may need to restart computer, or reinstall the software if it is still a problem.

[Guilin Sun]

The default EME region begins from x=0, I am not sure why do you place your monitor outside of the simulation region:

[aberg]

Thank you! The issue was that the simulation region did not include x < 0. So now I extended the simulation region to the left and the monitor shows results.

[aberg]

Also, thank you for the info regarding EME vs FDTD. What about if the taper angle is small (adiabatic) but the waveguide itself is a sub-wavelength grating instead of a solid silicon waveguide? So instead of continuous silicon, there is a series of silicon blocks surrounded by oxide as shown below. So the refractive index is periodically alternating over a short distance (200 nm) compared to the wavelength (1550 nm). Because of this alternating index profile, will EME still be accurate, or must I use FDTD?

[Guilin Sun]

Q: What about if the taper angle is small (adiabatic) but the waveguide itself is a sub-wavelength grating instead of a solid silicon waveguide
A: it is ok.
Q: So the refractive index is periodically alternating over a short distance (200 nm) compared to the wavelength (1550 nm). Because of this alternating index profile, will EME still be accurate, or must I use FDTD?
A: please note that both EME and FDTD have numerical errors as mentioned previously. If it is periodic, EME is more efficient as it DOES NOT need to repeat this segmentation but can simply use the number of periods:
Fiber Bragg gratings ÔÇô Lumerical Support

[Guilin Sun]

Please note that due to some reason, your image is not shown. Sometimes some browsers may not work well for the images pasted. You may try to use other browser.

[aberg]

Thank you very much. Here is a paper (open access) instead: https://opg.optica.org/DirectPDFAccess/4E0B835C-0276-4BF8-91609504AAD67E68_336815/oe-24-5-5026.pdf?da=1&id=336815&seq=0&mobile=no.
This is the exact same structure as in Lumerical's own edge coupler example: https://support.lumerical.com/hc/en-us/articles/360042305354
In this Lumerical example it says "In this example the sub-wavelength grating structure was not included. Instead it was assumed that a graded profile can be used where the effective index of the Si3N4 layer could be continuously varied". (I followed the paper and tried to simulate the full structure with the grating.)
But actually I would like to simulate a different coupler, where instead of silicon nitride layers forming a subwavelength grating, the inverse taper waveguide itself is the grating (Si and SiO2 layers with increasing duty cycle of silicon, so that the light is coupled into a waveguide grating which adiabatically tapers into a solid waveguide after some propagation distance). So there are no silicon nitride layers above the tapered silicon grating waveguide. In the inverse taper, the duty cycle of the silicon/SiO2 layers varies (similar to page 8 of the attached paper) , so the same unit is not repeating. So I cannot use the easy periodic method with EME, right?
I will have to model the entire structure and do either EME or FDTD. To obtain the overlap between the taper output and an SMF-28 fibre mode, should I put a field profile monitor exactly at the edge of the taper for both EME and FDTD? Or can I use FDE only (without EME or FDTD) at the edge of the taper like in the Lumerical example above, even though FDE is only supposed to be used for constant cross sections (which is not the case because the waveguide itself is a tapered grating with alternating Si/SiO2 layers)? Thanks again!

[Guilin Sun]

You are right that if the period is not fixed then you may not use the periodic method.
It is the field profile that matters whether you need to use a field profile monitor or FDE mode: if at the end of the taper it only supports one mode, it is ok to use FDE to get the mode profile, what you need is to renormalized this cascaded device. If it has higher order modes, yes, a field profile monitor can better capture the final field and then doing the overlap or other calculation. But since the profile monitor is easy to use, you can simply use it.

[aberg]

Thank you. Assuming I want to only consider one mode (the fundamental TE mode), and I wish to calculate the overlap for the non-periodic grating taper to SMF-28 fibre coupling, I will need to use a field profile monitor at some small distance in front of the taper, right, not on the exact tip of the taper? So my simulation region (EME or FDTD) has to extend beyond the taper into the air where the fibre will be?
Could you please clarify what "renormalizing the cascaded device" means for FDE? Is this process done in the Lumerical edge coupler example when calculating the mode overlap with a fibre? ( https://support.lumerical.com/hc/en-us/articles/360042305354 ) It seems that in this example, the FDE solver produces more than one mode and we just have to manually select the fundamental TE mode.

[Guilin Sun]

FDE does not need to renormalize. FDTD needs when you use the monitor data as the source for the second simulation.
The edge coupler example uses EME solver, which does not need such normalization. EME only uses S parameters which does not need normalize.
Of course it would be better to record the fields at the taper tip, as long as it is possible. I also notice that the profile monitor cannot record data if it is located at the end port location. So you may need some offset.

[aberg]

Thank you. I have a question about the mode field visualization in the Lumerical edge coupler example (https://support.lumerical.com/hc/en-us/articles/360042305354). I am looking at Step 2: Initial EME optimization of taper length without substrate. I do not understand how the mode profile at the bottom left corner of the figure in Step 2 was captured. It is the biggest mode at position x = 0 along the coupler. I downloaded the example file and see that "monitor_2" is meant to capture the YZ mode profile. It is set at x = 145 um by default, which is somewhere in the middle of the coupler, not at the edge where the fibre couples. So I changed the monitor position to x = 0 um to capture the mode at the edge because this is what is important for fiber coupling. But when I visualize the result from the monitor, I see something that looks like a plane wave, not a guided mode. Where should I put the monitor_2 along the x axis to capture the mode at the edge correctly?
Also, if the mode profile at the bottom left corner of the figure in Step 2 on the website is correct, then why does it not match the mode profile in Step 1? In Step 1, the mode at the edge was found using FDE at the x = 0 position, and the overlap with the fibre mode was calculated. But that mode in Step 1 is apparently much smaller than the mode shown in Step 2, both at x = 0?

[aberg]

Also I tried to save the mode profile from monitor_2 as a dcard named "modeprof2" using the script command
savedcard("modeprof2", "::model::monitor_2");
But I got the error message
Error: prompt line 1: in savedcard: there was an error finding the data to writeto file
even though I can clearly right click monitor_2 in the object tree and visualize the mode profile. Should I use a different command to save the dcard? Thanks!

[Guilin Sun]

the biggest mode at position x = 0: you can check mode profile at cell 1.
It is about the same as step 1, the difference is due to FDE and Cell not having the same size.
Savedcard does not work in EME. If you wish, please file a feature request : Vote new features, and file your feature request

[aberg]

Thanks, we have submitted a feature request for EME savedcard. Since savedcard does not work, how do I do an overlap calculation with field profile data from EME monitors or cells?
I see that the biggest mode at x = 0 in cell_1 corresponds to the SFM-28 fibre mode. But I would like the coupler waveguide mode, so that I can do an overlap calculation between the coupler mode and the fibre mode (just like in Step 1 with FDE, but now with EME). Is the mode in cell_2 the edge coupler mode at the end where it couples to the fibre? So cell_1 and cell_2 are the equivalent of the two modes shown Step 1 using FDE, right? (with some size differences).

[Guilin Sun]

In EME, each cell represents a section of the waveguide. therefore, the modes from each cell shows the modes at that waveguide section (center of the cell). Please zoom in and see which waveguide cross section it represents. You can also find the location from a simple calculation from EME settings, as you should know the group length and number of cells.
As for the overlap, you can extract the mode data:
getresult("EME::Cells::cell_19","mode fields");
it will give you all the mode fields for cell 19. Then you can extract a specific mode E and H components. You can see the data variables in the work space.
Than you can do overlap calculation using your own script, following the formula online, with all the field components known. Please try. If you have questions regarding to the script, please write another new post. Thank you!

[aberg]

Thanks. We will try writing a separate script for EME overlap. For now, we will use FDTD and directly save the monitor results to a dcard. You mentioned that FDTD monitor results need to be normalized before the mode overlap simulation. How should I do that? Please see the attached picture. The fibre mode from FDE is on the left, and the taper mode from FDTD monitor (saved to dcard) is on the right. The plots show the electric field intensity. I see that for the FDE fibre mode, the range is 0 to 1 but for the FDTD taper mode the range is 0 to 6e-31. The calculated overlap and power coupling are both around 20% (after optimizing position). Will these values change if the FDTD taper mode on the right is normalized? Is there a built-in function for normalization in Lumerical? Or do I need to write a script? (If I need to write a script, should it just divide the intensities at all points by the maximum intensity?) Thanks again for your help!

[Guilin Sun]

This post has changed topics a few times and quite long. Please write a new post so we can focus one-question one answer. The renormalization is for cascaded FDTD simulation to get correct transmission since the transmission is normalized to source power in each simulation.
For overlap usually no normalization is required. Of course if both modes have similar magnitude it have better numerical accuracy. In this particular case, the intensity ratio is on the order or 31, quite large. you can normalize the taper result.
Please write a new post when you have other questions. thank you!