VASP comes with multiple PAW potentials for many (most) elements. Are there some simple test calculations that can be done to decide which one would be best?
I realize that the particular application may have a large role in determining this but, if there are some "rules of thumb" or simple test calcs the experts out there use, I would love to hear about them.
Hm, that depends on the system and on the purpose of calculations. You can find some basic information here http://cms.mpi.univie.ac.at/vasp/vasp/PAW_potentials.html . If you could be more specific about the system at hand, it might be easier to help.
Thank you! The system is a Rh slab on which we want to calculate CO binding energies at different adsorption sites. The link you provide shows that VASP recommends Rh_pv. That is helpful.
I once used my experiences from a lot of all-electron calculations to make up a set of rules for making a good selection - which I subsequently found to be exactly the same as the recommendations given at the VASP website. A further validation of that choice can be found here: https://molmod.ugent.be/DeltaCodesDFT where various codes/pseudopotential libraries are tested against the Wien2K code, which is considered very accurate. You may also want to check out the settings used by http://www.materialsproject.org/.
So: Rh_pv, C and O would be my choice. But of course you shouldn't take our words for it but test. For your systems, I would check the lattice constant of Rh and the CO bond length with different functionals. More accurate is in this context called "harder", which means a smaller core and more plane waves needed for convergence (but of course also a larger region of space treated without pseudization). You can see this in the POTCAR file by looking at 1) the ENMAX parameter (which is higher for more accurate PAW's) and 2) RCORE, which is the cutoff radius (smaller for more accurate PAW's).
You need harder potentials for atoms that are more spatially constrained, either because of high pressure or because of their being unusually close to other atoms. The only way to know for sure is to test, but in "normal" cases (like yours) the recommended defaults are fine. Excited state properties is a different story though, there you need to use potentials with the postfix _GW, they have extra projector functions at higher energies for better accuracy in the conduction band.
The following link might be useful:
https://www.vasp.at/wiki/index.php/Available_PAW_potentials
- Badges
- Science topic
- Similar topics
- Filing