Osman Goni Ridwan

VASP cheat sheet · Essentials

VASP POTCAR cheat sheet

POTCAR holds the PAW potential of each species. This sheet decodes the suffixes, gives VASP's current recommendations, shows how to build and check the file without printing it, and covers the rare-earth choice between f in the core and f in the valence.

Checked against VASP 6.6.1 · Updated

The short version

  1. Concatenate one potential per species, in the order of the POSCAR species line. A wrong order is the most common POTCAR mistake, and VASP only warns about it.
  2. Start from VASP's recommended potentials (potpaw.64, PBE). Since October 2026 they include Fe_pv, Co_pv, Ni_pv, Cu_pv, Eu_3 and Yb_3.
  3. The largest ENMAX of your POTCAR is the default ENCUT. Set ENCUT yourself anyway, and keep it the same for every energy you compare.
  4. Compare energies only between runs with identical POTCAR files: the same names and the same set.
  5. Lanthanide _3 potentials put the 4f electrons in the core: no 4f magnetism comes out of them.
  6. Never edit or share a POTCAR. Report the TITEL lines instead.

Plan a POTCAR

Type the species line of your POSCAR, and optionally the counts line. The planner looks up VASP's current recommendation for each element (potpaw.64, PBE), with its valence electrons and ENMAX, and writes the command that builds the POTCAR. Type a potential name such as Fe_sv to override an element. Nothing is uploaded, and no POTCAR content is used.

What the suffixes mean

SuffixMeaningWhen to use it
noneThe standard valence, e.g. Ti with 4 electrons (3d² 4s²).Most p elements, Zn, Pd, Ag, Cd, Ir to Hg.
_pvSemicore p states in the valence: Ti_pv has 10 electrons.Recommended for most 3d and some 4d/5d metals (Cr to Cu, Tc to Rh, Hf, Ta, Re, Os); safer for magnetism and DFT+U.
_svSemicore s and p states in the valence: Ti_sv has 12 electrons.Recommended for Sc, Ti, V, Y, Zr, Nb, Mo, W and most alkali and alkaline-earth metals (Na and Mg use _pv).
_dSemicore d states in the valence: Ge_d has 14 electrons.Ga, Ge, In, Sn, Tl, Pb, Bi, Po.
_sSoft: larger core radius, lower cutoff (O_s about 283 eV against 400 eV for O).Quick relaxations and large supercells. Not for hybrid functionals or Hartree–Fock.
_hHard: smaller core radius, higher cutoff, more transferable (O_h about 766 eV).Short bonds and reference accuracy. potpaw.64 adds valence-f _h lanthanides.
_GWAccurate scattering far above the Fermi level; combines with others, e.g. _sv_GW.GW, RPA, BSE, optics: anything with many empty bands.
_2, _3Lanthanides only: a fixed valence, with the remaining 4f electrons frozen in the core.When 4f magnetism does not matter and the valence is known. VASP recommends _3 for Pr to Lu, now including Eu and Yb.
_AEAll-electron and very hard (H, He, Li only).Properties of the core region.
H.5, H1.25 …Hydrogen with fractional valence.Passivating dangling bonds of a slab; set NELECT yourself.

Which set

  • potpaw.64 is the latest and recommended set. It updated the lanthanides (Pr to Yb), whose valence-f potentials in older sets were significantly flawed, and added _h versions of them.
  • potpaw.54 (2015) and potpaw.52 (2012): today's downloads give the same results as their original releases; use them only to reproduce older work. Materials Project's r²SCAN data use potpaw.54.
  • The 2010-and-earlier sets are outdated and unsupported, and can misbehave with newer features such as meta-GGA. Materials Project's PBE and PBE+U data use files of this generation, some of which VASP no longer distributes.
  • Use LDA potentials only for LDA calculations. The PBE potentials serve every GGA, meta-GGA and hybrid: choose the functional in the INCAR, never by editing LEXCH in the POTCAR.

Build a POTCAR

Concatenate in the order of the POSCAR species line
sed -n 6p POSCAR                     # the species line, e.g.  Nd Fe B
POTDIR=/path/to/potpaw_PBE.64        # your licensed copy
cat "$POTDIR/Nd_3/POTCAR" "$POTDIR/Fe_pv/POTCAR" "$POTDIR/B/POTCAR" > POTCAR

Tools such as pymatgen and ASE build the POTCAR for you once told where your licensed potentials are; check what they chose with the commands below.

Check a POTCAR without printing it

Header lines only
grep TITEL POTCAR                    # one line per species, in file order: potential name and date
grep -c "End of Dataset" POTCAR      # number of species
grep ZVAL POTCAR                     # valence electrons of each species
grep ENMAX POTCAR                    # recommended cutoff of each species; the largest is the default ENCUT
grep kinetic POTCAR                  # meta-GGA (r2SCAN): kinetic-energy data per species (H, He, Li_sv, Be_sv print less, but work)
grep SHA256 POTCAR                   # checksum line in newer files
grep -v -e COPYR -e MD5 -e SHA256 POTCAR | sha256sum   # single-element file: compare with the VASP Wiki table
Which potentials a finished run used
grep "POTCAR:" OUTCAR | awk '!seen[$0]++'   # OUTCAR lists each potential twice; keep one of each

ENMAX and ENCUT

  • Each potential carries ENMAX, its recommended cutoff, and ENMIN, the lowest usable one. VASP's default ENCUT is the largest ENMAX in the POTCAR.
  • That default changes with the elements, so energies of different compounds are not comparable unless you set ENCUT explicitly to one value for all of them.
  • For relaxations that change the cell, use at least 1.3 × the largest ENMAX. Materials Project uses 520 eV for PBE (1.3 × 400 eV) and 680 eV for r²SCAN.
  • A lanthanide _3 potential has a low ENMAX, so in a compound the cutoff is usually set by another element: in Nd₂Fe₁₄B it is boron (318.6 eV), not Nd_3 (182.6 eV).

Rare earths: f in the core or in the valence

f in the core (Nd_3, Sm_3, Dy_3)

  • Stable and cheap: no partly filled 4f shell pinned at the Fermi level.
  • The 4f moment is absent: only the 5d/6s polarization of the rare-earth site and the transition-metal moments are computed, so no 4f moment or 4f single-ion anisotropy can come out.
  • Needs the valence to be known; VASP recommends _3 unless 4f magnetism matters.

f in the valence (Nd, Nd_h)

  • Use potpaw.64: the older valence-f lanthanides were significantly flawed.
  • Plain DFT pins the 4f states at the Fermi level and overbinds; DFT+U or a similar correction is usually needed.
  • Set LMAXMIX = 6 and LASPH = .TRUE., and expect a higher cutoff (Nd 338 eV, Nd_h 402 eV).

Materials Project uses _3 for most lanthanides in its PBE and r²SCAN data and _h in its newest sets; its PBE Gd file is no longer distributed, so rare-earth energies are comparable with their hull only if you hold the identical files.

What to report

In a methods section give the set (e.g. PBE potpaw.64), the TITEL line of every potential (name and date), the valence electrons and the ENCUT. The TITEL identifies the file without sharing it. POTCAR files are licensed: never post them in a repository or supplementary information.

Common mistakes

  • POTCAR order different from the POSCAR species order (VASP warns only if POSCAR has the species line).
  • Relying on the default ENCUT, which moves with the composition.
  • Comparing energies made with different files, including the same name from a different set or date.
  • Expecting 4f magnetism from _3 or _2 potentials; using Eu_2 or Yb_2 for trivalent compounds.
  • Meta-GGA with old (2010 and earlier) potentials or with O_GW, which lack kinetic-energy data.
  • Editing LEXCH to “make” a PBEsol or SCAN potential instead of setting the functional in the INCAR.
  • Soft _s potentials with hybrids; LDA potentials with GGA.
  • Leaving out semicore states in magnetic transition-metal work: they change lattice constants, moments and energy differences.
  • Fractional hydrogen without setting NELECT.

Other tags on this sheet

ENCUT
Plane-wave cutoff in eV: the size of the basis set.Default: the largest ENMAX among the POTCAR files · VASP Wiki: ENCUT
NELECT
Total number of electrons; set it for charged cells (fewer electrons for a positive charge).Default: the valence electrons of the POTCARs (neutral cell) · VASP Wiki: NELECT

The physics behind this sheet

Further reading

  1. VASP Wiki, “Recommended pseudopotentials” (the current recommendation for each element, with valence, ENMAX and SHA256 checksums). · vasp.at/wiki
  2. VASP Wiki, “Available pseudopotentials” (the sets, the suffixes and every potential with its ENMAX). · vasp.at/wiki
  3. VASP Wiki, “Preparing a POTCAR” (building the file in POSCAR order and mixing potential sets). · vasp.at/wiki
  4. VASP Wiki, “Choosing pseudopotentials” (advice, and the table of recommended PBE potentials, potpaw.64). · vasp.at/wiki