Does Yambopy support interpolation for GW-SOC eigenvalues of ferromagnets?

[Zimmi]

Dear Developers and Community,

I’m attempting to use Yambopy to interpolate GW eigenvalues of a ferromagnet that includes spin–orbit coupling (SOC), but I encountered an error during the interpolation step. The SKW interpolator reports a large mean absolute error (~143 meV), and the script crashes with a KeyError when trying to color the output.

The terminal output includes warnings about an ill-conditioned matrix and ends with:

Mean Absolute Error = 143.7 meV
KeyError: 143.69641327008011

This seems to come from the monty.termcolor module. Also, attaching the text file of the error message.

Before debugging further, I’d like to ask:
Does Yambopy support GW interpolation with SOC?
Or is this a known limitation?

Any help or suggestions would be greatly appreciated.

[palful]

Dear Zimmi,

Yambopy supports interpolation for calculations including spin-orbit coupling.
In your case, the main problem is that the interpolation is failing due to bad conditioning, resulting in a large error in the results. This triggers a warning that in turn triggers a small bug in the color printing function.

I should have fixed the small bug in the updated version of yambopy on the repository (so that it doesn't crash), but the main problem in your case is the bad interpolation. I would try to increase the value of the lpratio variable in the interpolation call, and check if the results are more accurate.

Best,
Fulvio

[Zimmi]

Dear Dr. Fulvio,
Thank you for your response.

I’m unsure how to adjust the lpratio parameter to improve interpolation accuracy, as you suggested. Could you please let me know where and how to change it in my Yambopy installation? Is it set in a specific function or file? Any tips on recommended values would also be greatly appreciated.

Just to let you know, I am attaching two band structures below, one DFT-spin polarized with SOC and the second GW-SOC without YAMBOPY interpolations.

Thanks again for your help!

[palful]

Dear Zimmi,

You can find a mention of how to tune the interpolation parameter in this tutorial.
You just need to call the interpolating function adding the value of lpratio as an argument.

Fulvio

[Zimmi]

Dear Dr. Fulvio,
Thank you for your time and suggestions.

The interpolation completes, but with significantly large errors. Below is a brief summary of the output:
Number of k-points: 64
Symmetries found: 12
MAE from interpolation: ~200–300 meV
Warning: LinAlgWarning: Ill-conditioned matrix (rcond=9.04857e-34): result may not be accurate.
Despite the calculation finishing, the Mean Absolute Error is too large for meaningful interpolations of GW eigenvalues.

I am attaching all the outputs, the drive link: https://drive.google.com/drive/folders/ ... p=sharing

I would appreciate your guidance on how to reduce this error.

[palful]

Dear Zimmi,

Indeed there was a problem with the interpolation of GW band structures for systems where time reversal symmetry is broken. In the newest version of yambopy (0.4.6) everything should be working fine.

Best,
Fulvio