Question about Planar + Pyramid Silicon Solar Cells

spacelover17spacelover17 Member Posts: 10

I am trying to modify the TiO2 pyramids solar cell example but first want to make sure that without the pyramids, I would get the same thing as with the planar silicon solar cell example (given identical thicknesses of the active layer).

Starting with the FDTD simulations after deactivating the pyramids, I noticed that the source only started at 0.4 nm instead of 0.3 nm (like it did in the planar cell), which was the source of a good amount of the discrepancy in predicted J_SC. I also noticed that in the planar example, the bottom 0.1 um or so had a 100x generation rate than the region directly above it, which I wasn't sure whether that was an artifact of the simulation setup or an actual effect due to the proximity of the aluminum layer below it, so I shortened the solar generation region to exclude that.

Despite these changes, I still notice a big difference in predicted J_SC between the planar silicon solar cell example and the TiO2 pyramid version of a planar solar cell (w/ deactivated pyramids). I noticed that the generation rate in the planar cell seemed to decay over longer length scales through the cell than the generation rate in the pyramid version of the planar cell, as shown in the screenshots, but I have no idea why that would be since the materials used are the same. Can you please help with this? This seems too significant a difference to write off as just "simulation error", and having reliable starting points is essential for comparing different structures.


Answers

  • kjohnsonkjohnson Posts: 175Ansys Employee

    Hello @spacelover17 ,

    The first thing I would recommend that you do is make sure the material fits are the same in the two simulations (same bandwidth, number of coefficients, fit tolerance, advanced settings, etc.). I think this would be the most likely reason the absorption is different between the two simulations.

  • spacelover17spacelover17 Posts: 18Member

    I tried your suggestions, and after adjusting the imaginary weight in the materials explorer, the # of coefficients, fit tolerance, etc. were the same. That improved the JSC in the pyramids solar cell (w/ the deactivated pyramids) by around 2 mA/cm^2, but it's still at 16.7 mA/cm^2 vs. 19.0 mA/cm^2 for the planar silicon cell. Do you have any other tips? I would not have thought to change the imaginary weight if you hadn't suggested I look at material fit, so other troubleshooting tips would be much appreciated!

  • kjohnsonkjohnson Posts: 175Ansys Employee

    The next step would be to go through the rest of the simulation settings and object properties (mesh, geometry, materials, etc.) to make sure they are the same in both simulations. However, I'm not sure how useful this would be, you would just be making both files identical. If you're concerned about the accuracy of the simulation, I think using convergence testing would be a better approach.

  • spacelover17spacelover17 Posts: 18Member

    I want them to be identical before re-enabling the pyramids, because if the version without the pyramids (which SHOULD be identical to the planar cell) is yielding a difference of 2+ mA/cm^2 in JSC, then it does not sound like I can properly compare the two in order to discuss the improvement brought about by the pyramids. Could I email you my files and ask you to look at them, since I did go through the everything I could think of.

  • kjohnsonkjohnson Posts: 175Ansys Employee

    Hello @spacelover17 ,

    If the two files are producing different results despite having the same geometry then it's true that there must be some difference in a simulation setting (mesh, bandwidth, boundary conditions, etc.) that is causing the discrepancy. However, I don't think that changing the settings in the file with the pyramids to match the planar Si file is the best approach. After all, we don't know which of these files has more accurate settings. It's possible that the TiO2 pyramid file has better settings, and by matching them to the planar Si file we are actually making the settings worse. If you believe the planar Si file has better settings, on the other hand, then it would be much easier and faster to just add the pyramids to that file rather than making adjustments to the pyramid file settings until the results match.

    I think a better approach to determine the improvement from the addition of the pyramids would be to first perform convergence testing with the file with the pyramids included. Once the convergence test has confirmed that the simulation settings are sufficient for the required level of accuracy, you can disable the pyramids and run the simulation again to get the results with no pyramids. For a proper comparison, all of the other simulation settings should be the same, the only difference between the two simulations should be disabling the pyramids. I don't think it would be necessary to perform convergence testing again for the simulation without the pyramids, since you are just replacing the pyramids with empty space.

    If you would like to verify your simulation results by comparing them to a different simulation, it is better to use a different solver or algorithm for the second simulation. For example, you could compare your FDTD results to results from the STACK or DGTD solvers. This page has an example that compares results from the STACK solver to results from the FDTD example:

    You could use the scripts from this to verify the Jsc results from the planar solar cell simulations. This wouldn't work for the simulation with the pyramids because the STACK solver requires a planar geometry, however. You might be able to replicate the FDTD simulation with the pyramids with the DGTD solver, however there isn't a solar cell analysis group in DGTD so you would have to rewrite the script from the FDTD solar cell analysis group for the Jsc calculation in DGTD.

    I hope this helps. Let me know if you have any questions.

Sign In or Register to comment.