The short version
- Almost every property is a recipe of two or three runs: relax, a careful static run, then the property run reading its CHGCAR or WAVECAR.
- Use the same ENCUT, POTCARs, functional and k-point density in every run you compare.
- Check your functional first: with r²SCAN or a hybrid, LEPSILON and VASP's own perturbation-theory phonons (IBRION = 7, 8) are not available; use LCALCEPS and finite differences (IBRION = 6).
- Many properties need an external tool: phonopy for phonon bands, LOBSTER for bonding, the Bader program for charges, pymatgen for convex hulls.
Find a property
What VASP can calculate
Native means VASP computes it; an external tool is named where one is needed.
| Property | Key tags | Needs first | Analysis |
|---|---|---|---|
| Formation energy, convex hull | none special; identical settings | relaxed compound and elemental references | pymatgen PhaseDiagram |
| Cohesive energy | isolated atom: ISPIN = 2, ISMEAR = 0, non-cubic box, Γ only | bulk energy | by hand |
| Surface energy | ISIF = 2, k-mesh n × m × 1, LDIPOL for asymmetric slabs | bulk reference with the same in-plane cell | γ = (Eslab − N Ebulk) / 2A |
| Adsorption energy | same cutoff in every run; dipole correction | clean slab, molecule in a box | by hand |
| Charged point defect | NELECT, ISPIN = 2 | bulk, its band edges and dielectric constant | doped, sxdefectalign (finite-size corrections) |
| Equation of state, bulk modulus | ISIF = 4 at fixed volumes | relaxed cell | pymatgen EOS, ASE |
| Elastic constants | IBRION = 6, ISIF = 3 | cell relaxed with the same settings | pymatgen ElasticTensor, VASPKIT |
| DOS, projected DOS | ICHARG = 11, ISMEAR = −5, LORBIT = 11, NEDOS | CHGCAR | py4vasp, sumo, pymatgen |
| Band structure | line-mode KPOINTS or KPOINTS_OPT | CHGCAR (GGA) | py4vasp, sumo, pymatgen |
| Band gap | BANDGAP (6.4.3 and later) | static run | OUTCAR, py4vasp, pymatgen |
| Effective mass | fine band path near the extrema | band structure | sumo-bandstats (external) |
| Work function, band alignment | LVHAR, LDIPOL, IDIPOL = 3, LVACPOTAV | slab with vacuum | LOCPOT plane average, py4vasp |
| Bader charges | LAECHG | static run | bader, chgsum.pl (external) |
| Electron localization function | LELF with NPAR = 1 | static run | VESTA |
| COHP, bonding analysis | LWAVE, ISYM = 0, NBANDS set | static run | LOBSTER, LobsterPy (external) |
| Partial charge, STM image | LPARD, IBAND or EINT | WAVECAR | VESTA; py4vasp STM (6.5.0, HDF5) |
| Fermi surface | dense uniform non-SCF mesh | CHGCAR | IFermi (external) |
| Band unfolding (supercells, alloys) | non-SCF on generated k-points | WAVECAR | easyunfold (external) |
| Wannier functions | LWANNIER90, or VASP's own LOCPROJ / SCDM | converged run | Wannier90 |
| HSE06 gap and bands | LHFCALC, HFSCREEN = 0.2 | PBE WAVECAR | as for bands |
| GW quasiparticles, BSE optics | ALGO = EVGW0, BSE … | DFT run with many empty bands | py4vasp |
| RPA total energies | ALGO = RPA (and its preparation runs) | exact diagonalization | OUTCAR |
| van der Waals corrections | IVDW (DFT-D3, …) or vdW-DF functionals | — | — |
| Optical absorption | LOPTICS, CSHIFT, many bands | static run | py4vasp, vasprun.xml |
| Dielectric tensor, Born charges, piezoelectric tensor | LEPSILON or LCALCEPS; + IBRION = 6 for the ionic part | relaxed structure | OUTCAR |
| Γ-point phonons, force constants | IBRION = 6 (or 8 for GGA) | tightly relaxed structure | OUTCAR (THz, f/i = imaginary) |
| Phonon bands, DOS, thermal properties | supercell forces with phonopy; VASP's own LPHON_DISPERSION gives the bands only | relaxed cell | phonopy |
| Thermal expansion (quasi-harmonic) | phonons at several volumes | E–V curve | phonopy-qha |
| Lattice thermal conductivity | third-order force constants | phonons | phono3py (external) |
| IR and Raman spectra | Born charges and phonons (IR); Raman needs external scripts | dielectric run | Phonopy-Spectroscopy (external) |
| Electron–phonon coupling, mobility | ELPH tags (native from 6.5.0) | supercell phonon potentials | py4vasp |
| Electronic transport (Seebeck, conductivity) | dense uniform mesh | CHGCAR | BoltzTraP2 (external) |
| Magnetic moments, orderings, anisotropy, exchange | ISPIN, MAGMOM, LSORBIT … | see the magnetism sheet | — |
| NMR chemical shielding | LCHIMAG | relaxed structure | OUTCAR |
| Electric field gradients | LEFG | — | OUTCAR |
| Hyperfine coupling | LHYPERFINE, ISPIN = 2 | — | OUTCAR |
| Core-level shifts, XAS | ICORELEVEL (core hole in a supercell); BSE-based XAS from 6.6.0 | supercell | OUTCAR, vasprun.xml |
| Ferroelectric polarization | LCALCPOL (Berry phase) | polar and reference structures | by hand |
| Finite electric field | EFIELD_PEAD | gapped system | OUTCAR |
| Implicit solvation | LSOL (VASPsol patch) | a patched VASP | — |
| Migration and reaction barriers | IMAGES, SPRING | relaxed end points | see the NEB sheet |
| Diffusion, melting, finite-temperature structure | IBRION = 0, MDALGO | — | see the AIMD sheet |
Which methods work with which functional
| Method | LDA, GGA | r²SCAN | Hybrids |
|---|---|---|---|
| LEPSILON (perturbation theory) | yes | no | no |
| LCALCEPS (finite field) | yes (gapped) | yes (gapped) | yes (gapped) |
| Phonons IBRION = 7, 8 | yes | no | no |
| Phonons IBRION = 6, phonopy | yes | yes | yes (expensive) |
| LOPTICS | yes | 6.6.0 and later; with spin-orbit set LPEAD = .TRUE. | yes |
| Band structure with ICHARG = 11 | yes | 6.6.0 and later | never |
| Band structure with KPOINTS_OPT | 6.3.0 and later | 6.3.0 and later | 6.3.0 and later (HFRCUT = −1 for gapped systems) |
Recipes
Density of states and projected DOS
Run a static calculation with LCHARG = .TRUE., then the run below with its CHGCAR on a denser Γ-centred mesh (the INCAR sheet has both). The gap appears in OUTCAR with VASP 6.4.3 and later.
ICHARG = 11
ISMEAR = -5
NEDOS = 2001
LORBIT = 11
NSW = 0
LMAXMIX = 4 ! same as the static runBader charges, ELF and charge-density differences
NSW = 0 ; IBRION = -1
LAECHG = .TRUE. ! AECCAR0 (core) and AECCAR2 (valence)
LELF = .TRUE. ! ELFCAR
NPAR = 1 ! required with LELF
LCHARG = .TRUE.chgsum.pl AECCAR0 AECCAR2 # reference density CHGCAR_sum
bader CHGCAR -ref CHGCAR_sum # charges in ACF.dat: electrons per atomThe net charge of an atom is its ZVAL minus the Bader electron count. If the total charge is off, refine the FFT grid (NGXF, NGYF, NGZF). For a density difference ρ(AB) − ρ(A) − ρ(B), compute all three at the same positions, cell and FFT grid.
Work function of a slab
NSW = 0 ; IBRION = -1
LDIPOL = .TRUE.
IDIPOL = 3 ! correction along c, the vacuum direction
DIPOL = 0.5 0.5 0.5 ! the slab's centre of mass; this line assumes a centred slab
LVHAR = .TRUE. ! electrostatic potential (no exchange-correlation) to LOCPOT
LVACPOTAV = .TRUE. ! vacuum levels on both sides in OUTCARWork function Φ = Vvacuum − EF. Leave 8–12 Å of vacuum on each side, relax first without the correction if it destabilizes the SCF, and take EF from the same run.
Elastic constants
IBRION = 6 ! finite differences with symmetry
ISIF = 3 ! needed for the elastic tensor
NFREE = 2 ! or 4
POTIM = 0.015
PREC = Accurate
EDIFF = 1E-7
NSW = 1Use a cutoff about 30% above ENMAX and start from a cell relaxed with the same settings. OUTCAR gives “TOTAL ELASTIC MODULI” in kBar (divide by 10 for GPa). VASP orders the columns XX YY ZZ XY YZ ZX. For the standard Voigt order (XX YY ZZ YZ ZX XY), move the 4th row and column (XY) to the end, that is, reorder them as 1 2 3 5 6 4. Perturbation theory (IBRION = 7, 8) does not give elastic constants.
Dielectric tensor, Born charges, piezoelectric tensor
LEPSILON = .TRUE.
IBRION = 8 ! adds the ionic contribution (GGA only)
EDIFF = 1E-8
NSW = 1LCALCEPS = .TRUE. ! gapped systems only
IBRION = 6 ! adds the ionic contribution
NFREE = 2
POTIM = 0.015
EDIFF = 1E-8
NSW = 1OUTCAR prints the electronic dielectric tensor, the Born effective charges and the piezoelectric tensor; the ionic parts appear under their own headings. LCALCEPS needs a regular Γ-centred mesh. Don't combine LEPSILON with LOPTICS in one run.
Optical absorption
LOPTICS = .TRUE.
NBANDS = 96 ! 2-3 x the default: empty bands are needed
NEDOS = 2001
CSHIFT = 0.1 ! broadening in eV (the default)
ISMEAR = 0 ; SIGMA = 0.05
NSW = 0 ; IBRION = -1This is the independent-particle spectrum; excitonic effects need BSE. The real and imaginary parts are in vasprun.xml and vaspout.h5.
Phonons
IBRION = -1 ; NSW = 0 ! one static run per displaced supercell
PREC = Accurate
EDIFF = 1E-8
ISMEAR = 0 ; SIGMA = 0.05
LREAL = .FALSE.
LWAVE = .FALSE. ; LCHARG = .FALSE.Relax first with tight forces (EDIFFG ≈ −0.001). phonopy generates the displaced supercells, collects their forces and gives bands, DOS and thermal properties; imaginary frequencies at Γ or elsewhere signal an instability or an under-converged relaxation. For Γ-point modes only, VASP alone with IBRION = 6 is enough.
COHP with LOBSTER
NSW = 0 ; IBRION = -1
ISYM = 0 ! LOBSTER needs the full (time-reversal-reduced) mesh
LWAVE = .TRUE.
NBANDS = 128 ! at least the number of LOBSTER basis functions
LORBIT = 11LOBSTER reads WAVECAR and writes COHPCAR.lobster and ICOHPLIST.lobster; LobsterPy automates the analysis.
Partial charges and STM images
LPARD = .TRUE.
EINT = -0.5 0 ! window in eV
NBMOD = -3 ! window relative to the Fermi energy
NSW = 0 ; IBRION = -1Read the WAVECAR of a converged run with the same ENCUT and symmetry settings. VASP writes PARCHG (or, from 6.5.0, to vaspout.h5 with LPARDH5 for py4vasp's STM plots).
Formation energy and convex hull
from pymatgen.analysis.phase_diagram import PhaseDiagram, PDEntry
from pymatgen.core import Composition
entries = [PDEntry(Composition("Fe"), -8.47), # energies per formula unit from your runs
PDEntry(Composition("O2"), -9.86), # same POTCARs, ENCUT, functional
PDEntry(Composition("Fe2O3"), -38.0)]
pd = PhaseDiagram(entries)
for e in entries:
print(e.composition, pd.get_form_energy_per_atom(e), pd.get_e_above_hull(e))The numbers above are placeholders, not results. Compute every reference yourself with the same settings and its own magnetic ground state. Mixing your energies with Materials Project data, or GGA with GGA+U, needs their correction scheme.
Common mistakes
- LEPSILON or IBRION = 8 with r²SCAN or a hybrid: not supported. Use LCALCEPS and IBRION = 6.
- Elastic constants read in VASP's column order as if it were Voigt order.
- Comparing energies from runs with different POTCARs, cutoffs, U values or functionals.
- A DOS or optical spectrum with too few empty bands, too coarse a mesh or the default 301 NEDOS points.
- LELF without NPAR = 1; Bader charges on a coarse grid; COHP without enough bands.
- Work functions with LVTOT (it includes the exchange-correlation potential) instead of LVHAR.
- Phonons from a structure relaxed only to 0.02 eV/Å: small imaginary modes that are not physical.
Other tags on this sheet
The physics behind this sheet
- Lecture 22: Equations of state and elastic constants (in preparation)
- Lecture 29: Band structure, DOS and projected DOS (in preparation)
- Lecture 31: Phonons and dynamical stability
Further reading
- VASP Wiki, “Category: Linear response” (dielectric tensors, Born charges, piezoelectric tensors). · www.vasp.at
- VASP Wiki, “Phonons from finite differences”. · www.vasp.at
- VASP Wiki, “Computing the work function”. · www.vasp.at
- The official VASP tutorials, with worked examples for most properties. · www.vasp.at