formation energy difference using ChGCAR or LOCPOT

[asiyeh_shokri2]

Hello

I am using Vasp 6.3.2 for formation energy calculations with FNV correction which needs LOCPOT file at the end. I see there are differences between my results and publishd articles in transition level. In one case I test to insteed of LOCPOT use CHGCAR and I saw I get results like the published ones! It is wierd to me as like article I am also using spin polarised, HSE with almost the same HSE parameters, and same POTCARS. They also mention they use LOCPOT! and in their case there is not a big difference between LOCPOT and CHGCAR!
Can anyone help me find the reason for these differences?

Best regards,
Asiyeh

[christopher_sheldon1]

Hi Asiyeh,

Could you provide some more details of your calculations and literature reference? The POSCAR, POTCAR, INCAR, KPOINTS, and OUTCAR files may be attached as a .zip file. Is there a specific literature reference that you are thinking of?

[asiyeh_shokri2]

Thanks for your response.
I am sending you the DOI of the article. they have used 120 atoms but I use 160 atom supercell. But correction should get almost the same results. I checked chemical potential they used for N, O and Ga and we had almost same numbers. But when they used my INCAR file the difference between total energies of N_OI at 0 charge and N_O at -1 charge in their case was 3.2 eV but mine is 3.9 eV.

doi: 10.1063/1.5063807

I am also sening you my files.

[christopher_sheldon1]

Thanks for the files. Yes, I don't think that the number of atoms should make too big a difference.

Your INCAR settings seem reasonable, although I'd suggest keeping only the INCAR tags used in your specific calculation, and comment out the rest. This makes it a bit clearer to read. I'm presuming that the structure was relaxed with the settings given in the INCAR file? We recommend converging with respect to forces EDIFFG = -0.01 eV/A for example, rather than EDIFFG = 3E-02. However, if you say that the chemical potentials are almost the same, this may not be the issue.

I'd first check the amount of exact exchange used (AEXX). In the paper that you reference, they use 35 % (AEXX = 0.35), while you use 32 % (AEXX = 0.32). This will make a difference in the total energies, and could account for your discrepancy with their results.

[asiyeh_shokri2]

Thank you so much for checking everything. My idea is if AEXX is going to affect the total energy so it affects for all charge states. And it may shift the formation energy but it should not make a big difference in the energy transition level. am I right?

Best regards,
Asiyeh

[christopher_sheldon1]

Not necessarily. By changing AEXX, you create a different density functional and so you should expect to get different results to the literature for whatever property you're interested in. If they remain the same, then it is coincidence.

[asiyeh_shokri2]

You're right, as we also get different transision levels of energy when we use GGA and othe approximations.
Thank you so much for your time and help.

Best regards,
Asiyeh

[christopher_sheldon1]

Glad that I could help. Hopefully this will solve the problem!