Osman Goni Ridwan

VASP cheat sheet · Essentials

VASP INCAR cheat sheet

The INCAR tells VASP what to calculate and how carefully. This sheet lists the tags you should set yourself, with the default VASP uses if you do not, the value to use instead and the mistake each one invites. Build an INCAR for your task below, or check one you already have.

Checked against VASP 6.6.1 · Updated

The short version

  1. Set ENCUT yourself, the same in every run you compare, and at least 1.3 × the largest ENMAX of your POTCAR when the cell relaxes.
  2. Choose the smearing: ISMEAR = 0 ; SIGMA = 0.05 for insulators, ISMEAR = 1 ; SIGMA = 0.2 to relax metals, ISMEAR = -5 for final energies and DOS. Not sure? ISMEAR = 0 ; SIGMA = 0.1.
  3. Tighten EDIFF to 1E-6.
  4. Whenever NSW > 0, set IBRION (2 to relax) and a negative EDIFFG, such as -0.02 eV/Å.
  5. Switch on LASPH for d and f elements, meta-GGA and hybrids, and set LMAXMIX to 4 (d) or 6 (f).
  6. Read back what VASP understood: OUTCAR lists every value it used (commands below).

Build or check an INCAR

Choose the task and describe the system, and the builder writes a commented INCAR from the rules on this sheet. Paste an INCAR you already have into the checker to catch the common traps. Both run in your browser; nothing is uploaded.

How VASP reads the INCAR

  • Each statement is TAG = value. Put several on one line with ; between them.
  • # or ! starts a comment.
  • A \ at the end of a line continues the statement on the next line. Leave no space after it.
  • N*value repeats a value in array tags such as MAGMOM: 4*5.0 is four atoms at 5 μB.
  • Logical values are .TRUE. and .FALSE.
  • Avoid tabs: lines with tabs have broken the parsing with some compilers.
  • VASP acts only on the tags it knows. Check that every tag you set appears in OUTCAR with the value you meant (how), or run vasp_std --dry-run first (VASP 6.4 and later): it parses the input and stops before computing anything.
INCAR syntax in one example
SYSTEM = bcc Fe, two-atom cubic cell   # comments start with # or !
ISPIN  = 2
MAGMOM = 1*3.0 \
         1*3.0                         ! one value per atom, continued with a backslash
ISMEAR = 1 ; SIGMA = 0.2               ! two statements on one line

Defaults that bite

VASP runs without most tags, using its defaults. These are the defaults that most often give a wrong or wasteful result.

NSW > 0 with no IBRION

IBRION then defaults to 0, molecular dynamics, which has no default time step: VASP stops at once.

Set IBRION = 2 to relax.

ISMEAR = 1 on an insulator

Methfessel–Paxton smearing (the default) can give unphysical occupancies in a gap; phonon frequencies can be off by more than 20%.

ISMEAR = 0, or -5 for static runs.

No ENCUT

The cutoff becomes the largest ENMAX of the POTCAR, so it changes with the elements: energies of different compounds are not comparable, and it is too low for cell relaxations.

Set it in every INCAR.

EDIFFG left at its default

The default is 10 × EDIFF, positive, so the relaxation stops on a small energy change while forces can still be large.

EDIFFG = -0.02 (a force criterion).

EDIFF = 1E-4

Loose for forces, stress and anything you report.

EDIFF = 1E-6; 1E-7 or tighter for phonons.

LASPH = .FALSE.

Stays off even when you set METAGGA, though meta-GGA, hybrids, DFT+U and 3d or f elements need the non-spherical terms.

LASPH = .TRUE.

LMAXMIX = 2

A DOS or band run that reads CHGCAR (ICHARG = 11) of a d or f system can then differ from the self-consistent run.

4 for d, 6 for f elements, in both runs.

No MAGMOM

Every atom starts at 1 μB, all parallel: an antiferromagnetic or high-moment state may never be found.

Set one value per atom, in POSCAR order.

LWAVE = .TRUE.

WAVECAR is written at the end of every run, often gigabytes.

LWAVE = .FALSE. unless a later run reads it.

No KPOINTS file

VASP falls back on KSPACING = 0.5 Å⁻¹, coarse for metals; and a stray KPOINTS file silently overrides KSPACING.

Write a KPOINTS file, or set KSPACING yourself.

NPAR from an old template

If NPAR and NCORE are both set, NPAR wins.

Delete NPAR; keep NCORE.

Symmetry kept by default

A relaxation keeps the symmetry of the starting structure, so a lower-symmetry minimum is out of reach.

Displace atoms slightly in POSCAR to break it.

Choosing ISMEAR and SIGMA

Smearing decides how electrons occupy states near the Fermi level. It is the setting most often left at a bad default.

System and taskISMEARSIGMA (eV)Why
Not sure if it is a metal, or high-throughput00.03–0.1Gaussian smearing is safe for both. With EFERMI = MIDGAP (VASP 6.4 and later) the Fermi level sits mid-gap. VASP advice
Insulator or semiconductor: relaxation, phonons00.05Never ISMEAR > 0 for a gapped system. VASP advice
Insulator: static energy, DOS−5—Tetrahedron method with Blöchl corrections: no SIGMA to converge. Needs a Γ-centred mesh with at least 4 k-points, otherwise use 0.
Metal: relaxation, forces, phonons1 or 2≈ 0.2Methfessel–Paxton: as large a SIGMA as keeps “entropy T*S” in OUTCAR below 1 meV per atom. VASP advice
Metal: total energy, DOS−5—Accurate energies and DOS on a dense mesh. Not for relaxing: its forces and stress in metals can be off by 5–10%.
Band structure along a path00.05Tetrahedra need a full mesh, and the Fermi energy cannot be computed on a path: take E-fermi from the self-consistent run.
Check the smearing in a finished run
grep "entropy T\*S" OUTCAR | tail -1   # last value, in eV for the whole cell: divide by the number of atoms
grep E-fermi OUTCAR | tail -1          # Fermi energy of the last step

Recipes by task

Each block is a complete starting INCAR. Replace the ENCUT of 520 eV with your converged value, and keep it the same across a workflow. The usual chain is: relax the cell → relax again from CONTCAR until the volume stops changing → static run → DOS and band structure from its CHGCAR.

Relax atoms and cell (ISIF = 3)
ENCUT  = 520          ! at least 1.3 x the largest ENMAX: the cell changes (Pulay stress)
PREC   = Accurate
EDIFF  = 1E-6
ISMEAR = 0 ; SIGMA = 0.05   ! insulator; for a metal use ISMEAR = 1 ; SIGMA = 0.2
IBRION = 2            ! conjugate gradient; 1 (RMM-DIIS) once close to the minimum
ISIF   = 3            ! atoms, cell shape and volume
NSW    = 100
EDIFFG = -0.02        ! stop when every force is below 0.02 eV/A
NELMIN = 4            ! at least 4 SCF steps per ionic step
LCHARG = .FALSE. ; LWAVE = .FALSE.

Then: cp CONTCAR POSCAR and run again until the volume stops changing (the basis was set up for the starting cell). Finish with a static run.

Relax atoms in a fixed cell (ISIF = 2)
ENCUT  = 520
PREC   = Accurate
EDIFF  = 1E-6
ISMEAR = 0 ; SIGMA = 0.05   ! insulator; for a metal use ISMEAR = 1 ; SIGMA = 0.2
IBRION = 2
ISIF   = 2            ! atoms only: surfaces, defects, molecules in a box
NSW    = 100
EDIFFG = -0.02
NELMIN = 4            ! 8 for complex surfaces
LCHARG = .FALSE. ; LWAVE = .FALSE.
Static run (also the SCF for DOS and bands)
ENCUT   = 520         ! the same cutoff as the relaxation
PREC    = Accurate
EDIFF   = 1E-6
ISMEAR  = -5          ! tetrahedron: Gamma-centred mesh, at least 4 k-points
NSW     = 0
IBRION  = -1
LORBIT  = 11          ! projected DOS and site moments
LMAXMIX = 4           ! 4 for d, 6 for f elements; keep it in runs that read CHGCAR
LCHARG  = .TRUE.      ! CHGCAR feeds the DOS and band runs
LWAVE   = .FALSE.

For the forces or stress of a metal in a static run, use ISMEAR = 1 ; SIGMA = 0.2 instead of −5.

Density of states from CHGCAR
ENCUT   = 520         ! the same as the run that wrote CHGCAR
PREC    = Accurate
ICHARG  = 11          ! read CHGCAR and keep the density fixed
ISMEAR  = -5
NEDOS   = 2001        ! a fine energy grid
LORBIT  = 11
NSW     = 0
LMAXMIX = 4           ! must match the self-consistent run
LCHARG  = .FALSE. ; LWAVE = .FALSE.

Copy CHGCAR from the static run and use a denser Γ-centred k-mesh than it had.

Band structure from CHGCAR
ENCUT   = 520
PREC    = Accurate
ICHARG  = 11
ISMEAR  = 0 ; SIGMA = 0.05  ! take E-fermi from the self-consistent run
LORBIT  = 11          ! orbital character of each band (PROCAR)
NSW     = 0
LMAXMIX = 4
LCHARG  = .FALSE. ; LWAVE = .FALSE.

KPOINTS holds the high-symmetry path in line mode; the KPOINTS sheet covers the format.

Spin-polarized: ferromagnet and antiferromagnet
ISPIN  = 2
MAGMOM = 2*3.0              ! ferromagnet: one value per atom, in POSCAR order
# MAGMOM = 2*4.0 2*-4.0 4*0.0   ! antiferromagnet, e.g. two Mn up, two Mn down, four O
LORBIT = 11                 ! moment of each site in OUTCAR
LASPH  = .TRUE.
LMAXMIX = 4

Add these lines to any recipe above. Start the moments 1.2–1.5 × above the expected values; an antiferromagnet needs a cell that holds both sublattices, often a supercell. The magnetism sheet goes further.

r2SCAN meta-GGA
METAGGA = R2SCAN      ! VASP 6.2.0 or later; avoid 6.6.0 for spin-polarized forces
LASPH   = .TRUE.
ALGO    = All
ISEARCH = 1           ! VASP 6.5.1 or later; delete it on older versions
EDIFF   = 1E-6

Add to a recipe above, with a converged ENCUT (Materials Project uses 680 eV for r²SCAN). Check that every POTCAR has kinetic-energy data: grep kinetic POTCAR. If the SCF struggles, start from a PBE WAVECAR; with ALGO = Normal, LMIXTAU = .TRUE. can help. Before VASP 6.6.0, band structures and DOS with a meta-GGA must follow the hybrid procedure below instead of ICHARG = 11; KPOINTS_OPT works on every version from 6.3.0.

HSE06 hybrid, static run
LHFCALC  = .TRUE.
HFSCREEN = 0.2        ! HSE06; the exact-exchange fraction AEXX is 0.25 by default
ALGO     = All        ! direct optimization: the recommended choice for hybrids
ISEARCH  = 1          ! VASP 6.5.1 or later; delete it on older versions
ENCUT    = 520
PREC     = Accurate
EDIFF    = 1E-6
ISMEAR   = 0 ; SIGMA = 0.05
LASPH    = .TRUE.
NSW      = 0 ; IBRION = -1
LWAVE    = .TRUE.     ! hybrids are expensive: keep WAVECAR for restarts

Start from a converged PBE WAVECAR. Never use ICHARG = 11 with hybrids: for a band structure, put the path in a KPOINTS_OPT file (VASP 6.3.0 and later) or as zero-weight k-points, and for gapped systems set HFRCUT = -1. To relax the cell with a hybrid, set ISIF explicitly: its default is 0, which moves the atoms but computes no stress.

The tags, group by group

Each tag name opens its page on the VASP Wiki. A version badge marks tags that need a recent VASP.

Start and restart

TagWhat it doesDefaultWhat to useWatch out
ISTARTRead the orbitals from WAVECAR (1) or start from scratch (0).1 if a WAVECAR file exists, otherwise 0Leave it unset: VASP restarts from a WAVECAR in the folder. Set 0 to force a fresh start.A WAVECAR left in the folder is read automatically. That is a harmless starting guess, but delete it if it belongs to another system.
ICHARGHow the starting charge density is made.2 if ISTART = 0, otherwise 02: atomic densities (new runs). 1: read CHGCAR. 11: DOS or band structure from a converged CHGCAR, density kept fixed.11 needs CHGCAR from a self-consistent run on a full k-mesh, with the same LMAXMIX. Never use 11 with hybrid functionals.
LMAXMIXHighest angular momentum ℓ of the on-site densities that are mixed and written to CHGCAR.24 for d elements, 6 for f elements: needed for DFT+U and noncollinear runs, and for DOS or band runs from CHGCAR.Left at 2, a run that reads CHGCAR (ICHARG = 11) can differ from the self-consistent result; for DFT+U band structures 4 or 6 is strictly required.

Basis set and precision

TagWhat it doesDefaultWhat to useWatch out
ENCUTPlane-wave cutoff in eV: the size of the basis set.the largest ENMAX among the POTCAR filesAlways set it. Converge the quantity you report, raising ENCUT about 20% per step. Cell relaxations: at least 1.3 × the largest ENMAX.Never go below the default. Energies are comparable only between runs with the same ENCUT.
PRECPrecision mode: sets the default FFT grids.Normal (Medium in VASP 4)Normal for routine work; Accurate for accurate forces, stress and phonons.Low, Medium and High are deprecated. High only raises the cutoff by 30%, so set ENCUT instead.
LASPHNon-spherical gradient corrections inside the PAW spheres..FALSE..TRUE. for 3d and f elements and magnetic B–F atoms, and whenever you use DFT+U, hybrids, meta-GGA (r²SCAN) or vdW-DF.It stays .FALSE. even when METAGGA is set: switch it on yourself.
LREALEvaluate the projection operators in real space, which is faster for large cells..FALSE.Auto for more than about 30 atoms; .FALSE. for small cells and for final energies..TRUE. and On are not recommended. Compare energies only between runs with the same LREAL.
ADDGRIDAn extra, finer grid for the augmentation charges..FALSE.Leave it off. Try it only if tests show it reduces noise in the forces.Not meant as a default setting for every calculation.

Electronic minimization

TagWhat it doesDefaultWhat to useWatch out
ALGOElectronic minimization algorithm.Normal (blocked Davidson)Normal: robust default. Fast: large cells and MD. All (with ISEARCH = 1): magnetic, DFT+U, meta-GGA and hybrid runs, or when Normal fails. Damped (with TIME): metals or small-gap systems with hybrids or meta-GGA, when the others fail.VeryFast is not supported for hybrids. Fast and VeryFast were rewritten in VASP 6 ("Old Fast" selects the VASP 5 version).
ISEARCH 6.5.1+Line search used by ALGO = All.01 together with ALGO = All: it usually converges more stably.The improved line search arrived in VASP 6.5.1; the default is still 0.
EDIFFSCF stopping criterion: the energy change between steps, in eV.10−4 eV1E-6 is the best general compromise; 1E-7 or tighter for phonons and other finite-difference runs.The default is loose for anything you report.
NELMMaximum number of SCF steps per ionic step.60Keep it near the default.An SCF that has not converged in about 40 steps usually needs a different ALGO or mixing, not a larger NELM.
NELMINMinimum number of SCF steps per ionic step.24 for relaxations of simple bulk materials, 8 for complex surfaces.Too few SCF steps per ionic step give inaccurate forces.
NELMDLNon-self-consistent steps at the start, before the density is updated.−12 if ISTART = 0, INIWAV = 1 and IALGO = 48 or 50; 0 if a WAVECAR is present; −5 otherwiseLeave the default: a negative value delays only the first ionic step.
NBANDSNumber of bands (orbitals) per k-point.max(nint(NELECT + 2)/2 + max(NIONS/2, 3), int(0.6 NELECT))The default is safe for ground states. Raise it for optics, GW and f-electron systems.Before VASP 6.6.0 the default used the signed MAGMOM, so negative starting moments (as in antiferromagnets) could give too few bands. VASP rounds NBANDS up for parallel efficiency.
LMIXTAUAlso mix the kinetic-energy density (meta-GGA)..FALSE..TRUE. if a meta-GGA SCF with ALGO = Normal or Fast does not converge.

Smearing

TagWhat it doesDefaultWhat to useWatch out
ISMEARHow electrons are smeared over states near the Fermi level.1Not sure: 0. Insulators: 0, or −5 for static runs and DOS. Metals: 1 or 2 for relaxations and phonons, −5 for total energies and DOS.Never ISMEAR > 0 for insulators: phonon frequencies can be off by more than 20%. −5 needs a Γ-centred mesh with at least 4 k-points, and in metals its forces can be off by 5–10%.
SIGMASmearing width in eV.0.20.03–0.1 with ISMEAR = 0. Metals with ISMEAR = 1 or 2: as large as keeps "entropy T*S" in OUTCAR below 1 meV/atom (often 0.2).Ignored by ISMEAR = −4 and −5. Forces belong to the smeared free energy, so converge them with SIGMA too.
EFERMI 6.4+How the Fermi energy is placed.LEGACYMIDGAP puts it in the middle of the gap of a semiconductor; recommended with ISMEAR = 0.Available from VASP 6.4.

Density mixing, for SCF trouble

Touch these only when the SCF will not converge. For a stubborn magnetic system, VASP's own advice is to converge a non-magnetic density first and restart from it, to switch to nearly linear mixing with BMIX = 0.0001 ; BMIX_MAG = 0.0001, and to lower AMIX and AMIX_MAG. The errors sheet covers this in detail.

TagWhat it doesDefaultWhat to useWatch out
AMIXLinear mixing parameter for the charge density.0.4 (PAW)Lower it if the SCF oscillates.
BMIXCutoff wave vector of the Kerker mixing for the charge density.1.0A tiny value (0.0001) makes the mixing nearly linear, a step in the magnetic troubleshooting recipe.
AMIX_MAGMixing parameter for the magnetization density.1.6Lower it, together with BMIX_MAG, when a magnetic SCF will not settle.
BMIX_MAGKerker cutoff for the magnetization density.1.00.0001 for nearly linear mixing in hard magnetic cases.
MAXMIXHow many previous steps the Broyden mixer keeps.−45A positive value keeps the mixer history across ionic steps, which can save SCF steps in relaxations and MD. In hard magnetic cases, a smaller |MAXMIX| can help.Negative (the default) resets the mixer at every ionic step.

Ionic motion

TagWhat it doesDefaultWhat to useWatch out
NSWMaximum number of ionic steps.00 for static runs; about 100 for relaxations.With NSW > 0, always set IBRION too.
IBRIONHow the atoms move.−1 if NSW = −1 or 0, otherwise 0 (molecular dynamics)2 (conjugate gradient) far from the minimum; 1 (RMM-DIIS) close to it; 0 for MD; 6 or 8 for phonons.NSW > 0 without IBRION gives IBRION = 0, molecular dynamics with no time step, and VASP stops at once.
ISIFWhich of atoms, cell shape and volume relax, and whether the stress is computed.0 for molecular dynamics (IBRION = 0) or hybrid functionals (LHFCALC = .TRUE.), otherwise 22: atoms only. 3: atoms, shape and volume. 4: atoms and shape at fixed volume. 7: volume only. 8: atoms and volume (VASP 6.4.1 and later).Changing the cell gives Pulay stress: use ENCUT ≥ 1.3 × ENMAX and re-run from CONTCAR. Hybrids default to ISIF = 0, so no stress unless you set it.
EDIFFGWhen a relaxation stops.EDIFF × 10Negative = force criterion in eV/Å: −0.02 is a common choice; tighten it before phonon calculations.Positive values (including the default) stop on an energy change, not on the forces.
POTIMStep size: the time step in fs for MD, a step scaling for relaxations, the displacement in Å for finite differences.0.5 for IBRION = 1, 2 and 3; 0.015 Å for IBRION = 5 and 6; none for molecular dynamics (must be set)Keep 0.5 for relaxations. MD: it must be set, and light atoms such as H need a smaller step.IBRION = 1 is especially sensitive to POTIM.

Spin

TagWhat it doesDefaultWhat to useWatch out
ISPINSpin polarization: 1 off, 2 collinear spin-polarized.12 for anything that may be magnetic.Noncollinear runs ignore it; delete it there.
MAGMOMStarting magnetic moment of each atom in μB, in POSCAR order.1.0 per atom for ISPIN = 2; 3 × 1.0 per atom for noncollinear runsStart 1.2–1.5 × the expected moment; give antiferromagnetic sublattices opposite signs; use N*value shorthand, e.g. 4*5.0 4*-5.0.It also lowers the symmetry. Keep it on restarts, or the magnetization can be symmetrized away. The final state depends on the start.
NUPDOWNFixes the difference between spin-up and spin-down electrons, i.e. the total moment.not set (free)Only to impose a total moment, as in fixed-spin-moment scans.A wrong value forces a wrong moment: leave it unset in normal runs.
LNONCOLLINEARNoncollinear magnetism: each moment can point anywhere..FALSE. (.TRUE. if LSORBIT = .TRUE.)Needs vasp_ncl; MAGMOM then takes three numbers per atom.
LSORBITSpin–orbit coupling..FALSE.Magnetic anisotropy and heavy elements. Needs vasp_ncl and switches on LNONCOLLINEAR.Set LMAXMIX (4 for d, 6 for f elements).

Output files

TagWhat it doesDefaultWhat to useWatch out
LWAVEWrite WAVECAR at the end of the run..TRUE..FALSE. unless a later run reads it (hybrid restarts, some post-processing).WAVECAR can be gigabytes.
LCHARGWrite CHGCAR and CHG at the end of the run..TRUE. (VASP 6.0–6.4.2: .NOT. LH5).TRUE. for the self-consistent run that feeds a DOS, band-structure or charge analysis.ICHARG = 11 cannot read the density from vaspwave.h5 yet; it needs CHGCAR.
LORBITProjections onto atoms and orbitals: PROCAR, projected DOS and site moments.0 (which needs RWIGS; from LORBIT = 10 on, RWIGS is ignored)11: s, p, d (lm-resolved) projections with no RWIGS needed.Values below 10 need RWIGS. LORBIT = 12 is for qualitative work only.
NEDOSNumber of points on the DOS energy grid.3012001 or more for a smooth DOS.Too few points wash out sharp features.

Functional

TagWhat it doesDefaultWhat to useWatch out
GGAExchange–correlation functional of the GGA family.the functional given by LEXCH in the POTCAR, if METAGGA and XC are not set eitherPE = PBE, PS = PBEsol, RP = RPBE, 91 = PW91.Never edit LEXCH inside the POTCAR.
METAGGAMeta-GGA functional, e.g. R2SCAN.the functional given by LEXCH in the POTCARR2SCAN (VASP 6.2.0 and later) with LASPH = .TRUE. and ALGO = All.Check the POTCARs with grep kinetic POTCAR. Spin-polarized meta-GGA forces are wrong in VASP 6.6.0 (fixed in 6.6.1).
LHFCALCSwitches on a hybrid functional with exact exchange..FALSE.With HFSCREEN = 0.2 for HSE06, and ALGO = All with ISEARCH = 1.Expensive. ISIF then defaults to 0 (atoms move, no stress). Never ICHARG = 11.
HFSCREENRange-separation parameter of screened hybrids, in Å⁻¹.0 (no screening)0.2 for HSE06.0.3 gives HSE03.
HFRCUTCoulomb truncation for hybrid functionals.0−1 for band structures of gapped systems: it converges best.

Parallelization

TagWhat it doesDefaultWhat to useWatch out
NCORENumber of cores that work together on one band.1Cores per NUMA domain is a good start; about 4 for ~100 atoms, 12–16 above 400 atoms. Benchmark.Forced to 1 with OpenMP threads and on GPUs. Do not set NPAR as well.
NPARLegacy: the number of band groups (MPI ranks ÷ NCORE).the number of MPI ranksUse NCORE instead.If both are set, NPAR wins.
KPARNumber of k-point groups that run in parallel.1A divisor of the MPI ranks, up to the number of irreducible k-points if memory allows; on GPUs often the number of GPUs.Each group holds its own copy of the data: memory is not shared.
NSIMNumber of bands optimized at once in the blocked algorithms.4Higher on GPUs (up to 32).

Symmetry and the k-mesh from INCAR

TagWhat it doesDefaultWhat to useWatch out
ISYMHow symmetry is used.2 for PAW; 3 with hybrids (LHFCALC); 1 with ultrasoft potentialsLeave the default; 0 for molecular dynamics.A relaxation keeps the starting symmetry: to reach a lower-symmetry minimum, break the symmetry in the starting structure.
SYMPRECTolerance for finding symmetry from the POSCAR positions.1E-5Raise it only for imprecise coordinates.
KSPACINGk-mesh from the INCAR: the largest allowed spacing between k-points, in Å⁻¹ (2π included).0.5 (Γ-centred while KGAMMA = .TRUE., its default)Smaller is denser; converge it. Without a KPOINTS file it replaces one.Ignored when a KPOINTS file is present. The default 0.5 is coarse for metals.
KGAMMAΓ-centred mesh for KSPACING..TRUE.Keep .TRUE.Ignored when a KPOINTS file is present.

Check what VASP read

VASP prints your INCAR and, further down, the value it used for every parameter, defaults included. Read them before trusting a result.

What VASP understood
# your INCAR as VASP read it (printed near the top of OUTCAR by VASP 6.2 and later)
awk '/^ INCAR:/{f=1;next} f&&/^ +[A-Za-z0-9_]+ = /{print;next} f{exit}' OUTCAR

# the values used, defaults included (VASP 5 and 6)
sed -n '/Startparameter for this run/,$p' OUTCAR | grep -E '^ +(PREC|ENCUT|EDIFF|EDIFFG|IBRION|ISIF|ISMEAR|ISPIN|LASPH) +='

# the same from vasprun.xml: matches inside <incar> are your INCAR, those inside
# <parameters> the values used (there the cutoff is called ENMAX)
grep -E 'name="(ENCUT|ENMAX|ISMEAR|SIGMA|EDIFFG)"' vasprun.xml

What changed between versions

VASPChange that affects an INCAR
6.0ALGO = Fast and VeryFast rewritten; “Old Fast” and “Old VeryFast” select the VASP 5 algorithms.
6.0 – 6.4.2The LCHARG default was .NOT. LH5, the opposite of the HDF5 output switch; it is .TRUE. again afterwards.
6.2.0METAGGA = R2SCAN.
6.3.0KPOINTS_OPT files for band structures on top of a regular mesh.
6.4EFERMI = MIDGAP; the --dry-run option.
6.4.1ISIF = 8: atoms and volume relax, the cell shape stays fixed.
6.5.1ISEARCH = 1, the improved line search for ALGO = All.
6.6.0The default NBANDS uses the size of each MAGMOM; before, negative starting moments (as in antiferromagnets) could give too few bands.
6.6.1Fixes wrong forces in spin-polarized meta-GGA runs (r²SCAN with ISPIN = 2) in 6.6.0.

To see which version wrote a run: head -1 OUTCAR.

Matching Materials Project settings

To compare your energies with Materials Project data, use their settings and the same POTCARs. Their input sets are high-throughput compromises, not the careful choices above: the PBE set relaxes metals with the tetrahedron method and stops on an energy change instead of forces.

TagPBE relaxation (MPRelaxSet)r²SCAN relaxation (MPScanRelaxSet)
ENCUT520680
EDIFF5 × 10⁻⁵ eV per atom in the cell1E-5
EDIFFGnot set (energy criterion)−0.02
ALGOFastAll
ISMEAR / SIGMA−5 / 0.050 / 0.2 if the band gap is unknown; 2 / 0.2 for metals, −5 / 0.05 for insulators
k-points64 k-points per Å⁻³ of reciprocal volume (reciprocal_density 64)KSPACING 0.22 for metals or an unknown gap; coarser for insulators, set from the gap (at most 0.44)
AlsoPREC = Accurate, ISPIN = 2, LASPH = .TRUE., LREAL = Auto, LORBIT = 11, NSW = 99, ISIF = 3, NELM = 100PREC = Accurate, ISPIN = 2, LASPH = .TRUE., LREAL = Auto, LMIXTAU = .TRUE., NSW = 99, ISIF = 3, NELM = 200

The physics behind this sheet

  • Lecture 12: Self-consistent field convergence (in preparation)
  • Lecture 16: Smearing: ISMEAR and SIGMA (in preparation)
  • Lecture 17: Plane waves and the ENCUT cutoff (in preparation)
  • Lecture 19: VASP input files and convergence tests (in preparation)
  • Lecture 21: Forces, stress and structural relaxation (in preparation)

Further reading

  1. VASP Wiki, “INCAR” (file format, comments, line continuation, and a complete example). · vasp.at/wiki
  2. VASP Wiki, “Smearing technique” (how to choose ISMEAR and SIGMA for insulators, metals and unknown systems). · vasp.at/wiki
  3. VASP Wiki, “Setting up an electronic minimization” (a minimal INCAR, the dry run, and converging ENCUT and k-points). · vasp.at/wiki
  4. VASP Wiki, “Structure optimization” (choosing IBRION, POTIM and NELMIN, and the symmetry caveat). · vasp.at/wiki
  5. VASP Wiki, “Volume relaxation” (Pulay stress, the 1.3 × cutoff rule and the restart protocol). · vasp.at/wiki