Structure optimization with Landau's U correction

[luigi_cavallo1]

Dear All,

We are doing N2 adsorption calculations on a surface of Ca3CrN3H, which never seem to reach the required structural energy minimisation. The energy value seems to have converged to the sixth decimal digit for many iterations but the OUTCAR doesn't print the "reached required accuracy" message. This is even after running the calculation with recopying CONTCAR to POSCAR for two weeks.

Something like this is printed when we extract the TOTEN value from the OUTCAR:

free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -973.14761603 eV
free energy TOTEN = -1000.35217434 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -973.14761604 eV
free energy TOTEN = -1000.35217434 eV

Please could you shed some light on what we are doing wrong? I am attaching u.tgz with the INCAR KPOINTS POTCAR and POSCAR files. We are running vasp/5.4.4 version with vasp_std

Thank you,
Rafia

[henrique_miranda]

We recently compiled a page on our wiki that tries to help with these difficult-to-converge systems.
wiki/index.php/Diffcult_to_converge_systems

Could you have a look and let me know if following these tips you are able to solve your issue?
If this does not help let me know in which of the steps mentioned the calculation fails and why.

[alex]

Hello Rafia,

I don't think it's an actual convergence issue, more like a spin flip.
Have you checked the total magnetic moment and (on top) the magnetic moment on every single chromium atom wrt. cycle number? I'd guess electrons 'move' from one to another Cr in a different environment and hence the differenct total energy.

Hth,

alex