Page 1 of 1

Different energies, warnings with different POTCAR verisons

Posted: Fri Dec 08, 2017 5:06 pm
by peterklaver
Hello people,

I started using VASP again for a new project after several years, hence I updated to the the latest VASP version (5.4.4) and the 5.4 POTCARs. With those, I get some warnings and results that seem possibly worrying.

Re-running an old job with the latest VASP executable gives me the exact same result as I got with VASP4.6 more than 10 years ago, so that is good. But if I then replace the old POTCAR with one from the 5.4 library, I get different total energies. I noticed that the EATOM value in the v5.4 POTCAR is slightly different from the one in the old POTCAR. And that the energy difference scales linearly with the number of atoms. Physical properties like the stress tensor and Fermi energy are very similar. Are the different total energies for different POTCAR versions purely due to different reference states used in the construction of the POTCAR (hence breaking energy compatibility with older results, but keeping energy differences and all other physics unaltered) or is this a sign that there is a problem?

Also, when using real space projection with a v5.4 POTCAR, I get a warning saying

The real-space-projection scheme for the treatment of the nonlocal
pseudopotentials has been switched on -- but on file POTCAR I have
not found any entries signalling me that you have ever performed a
real-space optimization of all nonlocal projectors! BE WARNED that
a calculation using the real-space-projection scheme together with
nonlocal projectors which were not real-space-optimized (according
to the plane-wave-cutoff used in this calculation) might give more
or less inaccurate results
I hope you know what you are doing

When running otherwise similar input with an old POTCAR, VASP does not voice any such reason for concern. So should I be concerned when VASP gives the warning in a job using a new POTCAR? Would a calculation with VASP544 with real space projectors using an older POTCAR be safe?

And finally, when using KPAR > 1 VASP still says that verification with KPAR = 1 is required. It said that in version 5.3 too. Is k-point parallelization still deemed in the testing stage or can that warning be taken more lightly now?

Kind regards,
Peter

Re: Different energies, warnings with different POTCAR veris

Posted: Wed Mar 14, 2018 3:58 pm
by admin
it cannot be expected that different POTCAR files yield comparable total energies,
therefore you should always use the same POTCAR files to compare total energies.

KPAR is now implemented correctly, I suppose that the warning still persists because
this part of the code was not programmed by members of the vasp group, but is a
contribution form a (very experienced!) colleague and long-term vasp-user