Learning Forum

[Elam Day-Friedland]

 

Hi! I am trying to simulate a 3d SOI device in CHARGE across DC bias points with the thermo-electric coupled solver and am running into some issues in mesh generation (see below screenshot). Note that meshing completes (or at least the solver considers it complete).

 

Here is the error (embedding links doesn't seem to work)

https://i.ibb.co/ZBmzGZM/Screenshot-2024-04-08-at-8-13-20-AM.jpg

 

 

Here is the device. This is a 3d simulation, and the cladding material is SiO2.

https://i.ibb.co/vLqnZ51/Screenshot-2024-04-08-at-8-16-51-AM.jpg

 

It seems that the error is related to the contact (blue, aluminium) silicon (red) junction but I’m not sure how to confirm this.  One contact is constant at 0v and the other is swept from 0v to 4v

 

For reference, the si is around 150nm thick and the contact is ~1um by ~2um.

 

what I've tried:

- decreasing the triangle quality (to ~15% above ‘none’) as well as the max edge length (down to .1um, min edge length at .01um).

- enabling PLC edge splitting but I’m not quite sure how to go about setting its value

- moving the contacts down (so they overlap with the silicon) and decreasing the mesh order, this just makes tessellation run for many minutes and then the simulation fails right after tessellation without error.

- I've tried this on Lumerical versions 2024R1 and 2021R2.3

 

It seems like there isn't an actual error, just these warnings- I'm not sure if there's a way to increase verbosity or something? How should I go about debugging this? Thanks!

 

note that this error appears this issue only occurs when the temperature dependence type is set to 'coupled' (i.e. thermo-eletric), which my simulation must account for. The simulation runs correctly in isothermal mode.

 

edit: looks like the device-engine is segfaulting after the mesh is generated e.g.

```
MESH_FINISH
 
    initializing 
 

 
    0 

Segmentation fault (core dumped)```

 

I am able to run 'coupled' simulations with simpler devices

[Guilin Sun]

This message

means the mesh needs further refining however it exceeds the number of iteration. Also it suggest that some Finite Element volume is too small. It seems this device does not have curved surface on a flat surface or very thin layer.  The mesh will automatically adjusted depending on the optical generation and other factors. Please first try to use smaller optical power by reducing the scale factor. you can also increase the limit for mesh iteration, reduce triangle quality and sensitivity. Please refer to https://optics.ansys.com/hc/en-us/articles/360034918633-CHARGE-solver-Simulation-object

When the coupled simulation is running, maybe one of the solvers needs finer mesh. You can try to simulate HEAT separately and see if it works.

In addition, you can use smaller-step applied votage if zero voltage works.

 

 

[Elam Day-Friedland]

 

Thanks for the tips! It turns out the issue was not with the mesh: the bug was that the materials only had defined electrical properties (addmaterialproperties("CT",...)) and not thermal ones (addmaterialproperties("HT",)) (this is because CHARGE doesn’t specify materials by default and I had written a script to add them). The HEAT (co-)solver was silently crashing with no error reporting. I discovered this error by setting up a discrete HEAT solver and trying to run it.

 

[Guilin Sun]

I am glad that you found the root cause. A separate simulation is a direct way to validate its own settings.