Metallic Fermi level

[vahid_askarpour]

When plotting the band structure of magnetic Mn3Ge, the bands are shifted by E_F=11.39eV by p4vasp.

However, the Fermi level I get from a converged self-consistent calculation is E_F=11.58eV.

In general, which of the above two procedures for Fermi level determination is correct for a metal?

Thanks,
Vahid

[svijay]

Dear vahid_askarpour,

Please supply VASP input and output files in accordance with the forum guidelines here: https://www.vasp.at/forum/viewtopic.php?t=17928. Additionally, please make sure to supply a minimal working example in case of specific queries: https://www.vasp.at/wiki/index.php/Mini ... le_example; thank you!

Sudarshan

[vahid_askarpour]

Dear Svijay,

The details of the calculations is part of an ongoing study and I cannot provide you with the details you requested. However, my question is rather general: For any metal, should the Fermi level be read from the band structure calculation or from a self-consistent calculation?

Thanks,
Vahid

[svijay]

Dear Vahid,

Without specific details of the calculation, unfortunately, I will not be able to provide specific comments to your question. In general, the Fermi level will be different for different parameters used in your calculation. Furthermore, typically we are interested in relative quantities to do the Fermi level (such as the workfunction: https://www.vasp.at/wiki/index.php/Comp ... k_function) and not its absolute value. If you would like to provide further context as to what you are seeing (perhaps through a simplified example: https://www.vasp.at/wiki/index.php/Mini ... le_example) we would be happy to take a closer look; thanks!

Sudarshan

[vahid_askarpour]

Dear Sudarshan,

Attached please see the self-consistent calculation and band structure calculation for Fe. The OUTCAR files are included.

The Fermi level for the self-consistent part is 5.7383eV.
The Fermi level from band structure which p4vasp uses to shift the bands is 5.9659eV.

I need the absolute value of the Fermi energy because I want to calculate the anomalous Hall conductivity using the Wannier interface (postw90). The postw90 input requires the setting of the Fermi energy.

Thanks,
Vahid

[svijay]

Dear Vahid,

Thanks for sharing a minimal reproducible example. A quick clarification - do you need to perform a non-self consistent calculation before calling the wannier90 interface? I was able to get input files for the wannier90 interface using the LWANNIER90 tag (https://www.vasp.at/wiki/index.php/LWANNIER90) and following the example here: https://www.vasp.at/wiki/index.php/Si_bandstructure (see Procedure3: VASP2WANNIER90)

Sudarshan

[vahid_askarpour]

Dear Sudarshan,

I can get the wannier90 input files with the LWANNIER90 tag using a self-consistent calculation. I followed the Procedure 3 you mentioned but that is not the issue.

The issue is that the files I sent you on Fe show two different Fermi levels. Which Fermi level is correct?

Thanks,
Vahid

[marie-therese.huebsch]

Dear Vahid,

You should use the Fermi energy from the self-consistent calculation because to determine the Fermi energy you need a regular grid.

The Fermi energy is sensitive to the k-points density. Or rather the specific set of k points because the Fermi energy is determines to be somewhere between the highest occupied and lowest unoccupied Kohn-Sham state. If the k point that is the maximum of the occupied band is not included in the k mesh, there is an error/uncertainty due to the discretization. To avoid this you can check the convergence by restarting from CHGCAR with denser and denser k mesh and make sure that the Fermi energy is converged with respect to the number of k points. Above a certain k-points density it is save to fix the charge density with ICHARG = 11, because the Fermi energy does not change because the charge density is not converged, but because the Kohn-Sham state at a different k point is determined to be the maximum occupied state.

FIY: Make sure to set LMAXMIX (=4 for d electrons and =6 for f electrons) in the SCF run when you plan to restart a calculation from CHGCAR . And p4vasp is deprecated.

Does this answer your question?

Best regards,
Marie-Therese

[vahid_askarpour]

Dear Marie-Therese,

Yes, it does. It also suggests that anyone who has plotted the band structure with p4vasp would have shifted the bands with incorrect Fermi energy.

Thank you for this clarification. I will use the Fermi energy from self-consistent calculations after ensuring that the result is converged with K-grid.

Cheers,
Vahid