The --emin input parameter does not work

[Shentu-jh]

Dear He Xu:
When I use wannier90 with TB2J, I tried to adjust --emin to make emin closer to the Fermi energy, but I found that the --emin input parameter does not work; it seems that the program automatically selects a value and overrides my input.
The reason I tried to modify emin is that in the material Mn3Al I am calculating, there is an approximately 0.2 eV band gap at about -5.8 eV below the Fermi energy. However, the program chooses a value of -9.37 eV, where there is a band gap of more than 30 eV.
I would like to ask why TB2J overrides my input, and whether it is because my input is considered unreasonable. Personally, I believe it is necessary to retain the user's input.
The TB2J version I am using is 0.9.12.18.
My input commend is : python3.12 /home/stjh/.local/lib/python3.12/site-packages/TB2J/scripts/wann2J.py --elements Mn --efermi 9.48181149 --kmesh 30 30 30 --emin -5.8 --cutoff 1e-8 --np 16
Appreciate your time and consideration！

vasp_band.pdf

[mailhexu]

Hello,
Indeed emin is not directly used but it is used as a hint to the code for an automatic search for the integration range.
The goal is to find a bandgap large enough, so that the states lower than that have negligible effect on the intersite exchange.

This is by design so that one does not have to always check the band structure. This is especially useful when doing high-throughput calculations. And it make sure that emin is low enough to include all the related bands and not too low so that the integration between emin and 0 (Fermi energy) do not require too many point.

It works like this:

• If emin is within a big band gap, it will search for the state above the gap, and set emin to a energy a little lower than that.
• Otherwise, it will search below emin for a large band gap. And shift emin so that it is within the gap.

Note that by a large band gap it is several eV. In your case, the 0.2 eV is not considered as a big band gap. Indeed the states below this small gap likely include atomic orbitals strongly hybridize with the states above the input emin and can affect the magnetic interactions.

Best,
HeXu