Pseudopotentials



Our most recent effort on pseudopotentials - the Standard Solid-State Pseudopotentials (SSSP) library - can be found in the Materials Cloud. Please use only these pseudopotentials for all your calculations - either SSSP Efficiency or SSSP Accuracy.


Note also note the large repository in the official QE pseudopotential webpage - but many have not been thoroughly tested.




Older material, kept for reference

PSLibrary 0.3.1

The following files hold the pseudopotentials in UPF format for over 60 elements in the periodic table, that are generated and generously shared by Prof. Dal Corso, SISSA, Italy. Note they are PAW-PBE, so for solids, go with PBESOL package below'''

H                He
LiBe          BCNOFNe
NaMg          AlSiPSClAr
KCaScTiVCrMnFeCoNiCuZnGaGeAsSeBrKr
RbSrYZrNbMoTcRuRhPdAgCdInSnSbTeIXe
CsBaLaHfTaWReOsIrPtAuHgTlPbBiPoAtRn

All pseudopotentials generated from PSlibrary 0.3.1

Download the compressed file for the exchange correlation type you wish to use. If you have no clue: PBE/PBEsol are standard GGA choices for molecules and solids respectively; and if you need LDA instead go with "PZ".


( Beware, these are huge files. Each XC is approx 100MB, relativistic ones are approx. 200MB. Go to the Notes section below to learn more about the subfolders in each category. )

bp.0.3.1.tgzrel-bp.0.3.1.tgzpbe.0.3.1.tgzrel-pbe.0.3.1.tgz
pbesol.0.3.1.tgzrel-pbesol.0.3.1.tgzpw91.0.3.1.tgzrel-pw91.0.3.1.tgz
revpbe.0.3.1.tgzrel-revpbe.0.3.1.tgzpz.0.3.1.tgzrel-pz.0.3.1.tgz
wc.0.3.1.tgzrel-wc.0.3.1.tgz


Notes about PSlibrary 0.3.1 tgz files above

1) Each Exchange-Correlation is a separate tgz file and holds 4 subdirectories; 3 for PAW and Ultrasoft; 1 for Norm-Conserving. The description of each subdirectory is exactly as in pslibrary project:

  • PSEUDOPOTENTIAL (PAW and US) widely tested and no error has been reported so far.
  • PSEUDOPOTENTIAL_TOT (PAW and US) tested in fewer electronic environements and in some cases (mostly in fully relativistic form) encountered an error.
  • PSEUDOPOTENTIAL_ALT (PAW and US) alternative data sets that have a different number of valence states with respect to the data set in the main distribution files in the PSEUDOPOTENTIAL or PSEUDOPOTENTIAL_TOT directories. Verification tests in a wide range of electronic environments have not been performed yet.
  • PSEUDOPOTENTIAL_NC are the norm conserving pseudopotentials. Not tested.


2) PAW files are the ones with "kjpaw" in their names while "rrkjus" is for the ultrasoft pseudopotentials.

3) Here we only host the final pseudopotential files (UPFs). The generation files for these pseudopotentials are distributed through the PSLibrary project, which is hosted at QE-Forge: http://qe-forge.org/gf/project/pslibrary/ If you want to know more about the generation scheme of a particular pseudopotential, download the project to access the generation files (ie input, output files and radial wavefunction plots) that reveal many details of the atomic properties from all electron and pseudo points of view.

4) All PAW files are compatible with GIPAW calculations using "paw_as_gipaw=.true."

How do they perform ? (PAW-PBE)

The preprint of our manuscript that attempts to answer this question can be found here: http://arxiv.org/abs/1404.3015. Here are some results from this work:


Fig.1: Equilibrium volume and bulk modulus comparison for elemental solids between pseudopotential and all electron codes. The PAW pseudopotentials listed here are noted as QE-PAW.

Fig.2: We suggest a uniform and symmetric comparison factor, Delta1', extending the works of K. Lejaeghere et. al. (arXiv:1204.2733) and F. Jollet et. al. (arXiv:1309.7274v2)

What about PSLibrary 1.0.0 ?

This new set of pseudopotentials are generated and distributed by Prof. Dal Corso to obtain a uniformly generated set of pseudopotentials. However we found that they are not necessarily better than the v0.3.1 set. As an example, here we demonstrate the performance of some of the transition metals we have tested in the new set: (Reminder: Smaller is the delta the closer are the results to the all electron calculations)

ElementDelta v.0.3.1Delta v.1.0.0
Sc0.3280.733
Ti0.2301.493
V0.2500.728
Cr1.4171.887
Mn2.0792.918
Fe3.4962.023


Ultrasoft Pseudopotentials or Projector-Augmented Wave Method?

When they are of the same quality in the atomic environment, I would always choose the PAW method over US pseudopotentials, simply because PAW datasets are more transferable by construction.

Some rather-easily-digestible references about PAW and US for beginners

  • Here is the most explanatory paper on PAW from its creator:

http://journals.aps.org/prb/abstract/10.1103/PhysRevB.50.17953 here is a talk by him: http://online.kitp.ucsb.edu/online/cem02/bloechl/)

  • Here is its relationship to Ultrasoft

http://journals.aps.org/prb/abstract/10.1103/PhysRevB.59.1758

  • A summary presentation from a very bright colleague who has also implemented a great deal of pseudopotential related parts in Quantum Espresso:

http://itp.tu-graz.ac.at/LV/ewald/TFKP/Literatur_Pseudopotentiale/Paulatto_08_PP_summ.pdf

More Pseudopotentials ?

QE community has been using/generating/converting many other pseudopotentials over the years. Prof. Giannozzi, University of Udine, Italy has been tirelessly tracking and maintaining them in the main QE website. For more pseudopotentials therefore see official QE pseudopotential webpage.

(Contributed by Emine Kucukbenli. No more maintained. )