This section of the manual has not been updated in a long time. We strongly recommend the use of the VASP-PAW potentials described in Sec. 10.2. The use of ultrasoft pseudopotentials is at your own risk and the potentials are no longer maintained or updated.
All supplied PP's with VASP are of the ultrasoft type (with few exceptions). And for most elements only one LDA and one GGA PP is supplied. All pseudopotentials are supplied with default cutoffs (lines ENMAX and ENMIN in the POTCAR files), and information on how the PP was generated. This should make it easier to determine which version was used, and user mistakes are easier to correct. The POTCAR files also contain information on the energy of the atom in the reference configuration (i.e. the configuration for which the PP was generated). Cohesive energies calculated by VASP are with respect to this configuration. Mind that the cohesive energies written out by VASP requires a correction for the spin-polarization energies of the atoms.
For the transition metals an additional problem exists: The cohesive energies written out by VASP are with respect to a "virtual" non spin-polarized pseudo-atom having one s electron and Nvalence-1 d electrons. This is usually not the experimental ground state configuration.
The table below gives the required energy corrections (d(E)) for transition metals: i.e. it contains the difference between the "virtual" non spin-polarized pseudo-atom and a spin-polarized groundstate (GS) atom calculated with VASP. The calculations have been done consistently with VASP, using the procedure described in Sec. 9.5.
Mind that LDA/GGA is not able to predict the correct groundstate (line exp.) for all transition metals. This is not a failure of VASP but related to deficiencies of the LDA/GGA approximation. Only configuration interaction (CI) calculations are presently able to predict the groundstate of all transition metals correctly.
|exp.||3d 4s2||3d2 4s2||3d3 4s2||3d5 4s||3d5 4s2||3d6 s2||3d7 4s2||3d8 4s2||3d10 4s1|
|GS||3d 4s2||3d3 4s||3d4 4s||3d5 4s||3d5 4s2||3d6.2||3d7.7||3d9 4s||3d10 4s1|
|exp.||4d 5s2||4d2 5s2||4d4 5s||4d5 5s||4d5 5s2||4d7 5s||4d8 5s||4d10|
|GS||4d 5s2||4d3 5s||4d4 5s||4d5 5s||4d5 5s2||4d7 5s||4d8 5s||4d10|
|exp.||5d2 6s2||5d3 6s2||5d4 6s2||5d5 6s2||5d6 6s2||5d9||5d9 6s|
|GS||5d2 6s2||5d3 6s2||5d5 6s||5d5 6s2||5d6 6s2||5d8 6s1||5d9 6s|
The POTCAR file also contains information about the approximate error according to the RRKJ (Rappe, Rabe, Kaxiras and Joannopoulos) kinetic energy criterion. This approximate error is taken into account when cohesive energies are calculated, and this is the reason why cohesive energies do not decrease strictly with the energy cutoff. If you do not like this feature remove the lines after
Error from kinetic energy argument (eV)till (but not including) the line
END of PSCTR-control parametersin the POTCAR file. We want to point out, that the RRKJ kinetic energy is usually very accurate and corrects for more than of the error in the cohesive energy, but it works only if there is not a considerable charge transfer from one state to another state (sd or sp).
Two versions of PP, which one should be used
For H three POTCAR files exist. The H/POTCAR and H_200eV/POTCAR files actually contain the same PP. The only difference is that H_200eV has a lower default energy cutoff of 200 eV (the default cutoff for H is 340 eV). Up to now we have not found any difference between calculations using 200 and 340 eV, we therefore recommend to use only H_200eV (differences for the dimer are for instance less than ). If H is used together with hard elements like carbon VASP will anyway adopt the higher default cutoff of C. The third potential H_soft (generated by J. Furthmueller) should be used in conjunction with soft elements like Si, Ge, Te etc. As one can see from the data_base file H dimer length and vibrational frequencies are still quite reasonable.
For the first row elements two PP exist, we recommend the standard version, which gives very high accuracy. The second set ( B_s,C_s,O_s,N_s,F_s) is significantly softer and should be used only after careful testing. We have found that the second set is safe if a hard species is mixed with a softer one (that is for instance the case in Si-C, Si-O, or even Ti-O).
For Ga, In, Sn and Pb one should describe the 3d or 4d states as valence, corresponding PP can be found on the server in the directories
Ga_d, In_d, Sn_d, Pb_dIf one puts the 3d or 4d states in the core the results depend strongly on the location of the position of the d-reference energy. The d-reference energy for the conventional Ga, In, Sn and Pb PP (with d in the core) has been adjusted so that the equilibrium volume is within 1 percent of the equilibrium volume for the Ga_d, In_d and Sn_d PP. This is clearly a ad hoc fix, but results in reasonably accurate pseudopotentials. Mind that PP including d are currently missing for Ge, and for very accurate calculations such a PP might be required.
The following PP are currently available with p semi-core states
Li_pv Na_pv Mg_pv K_pv Ca_pv Sc_pv Ti_pv V_pv Fe_pv Rb_pv Sr_pv Y_pv Zr_pv Nb_pv Mo_pv Cs_pv Ba_pv Ta_pv W_pv
For a few elements harder NC-PP exist which can be used in calculations under pressure, for ionic systems, or for oxides:
Na_h Mg_h Al_h Si_h