mml:vasp

VASP was developed by Georg Kresse, Martijn Marsman, and Jürgen Furthmüller Computational Physics, Faculty of Physics, Universität Wien, Austria

VASP is is a DFT code using plane waves as the basis set. Both the projector augmented waves (PAW) and the ultrasoft pseudo-potential methods are implemented in VASP.

There is a lot of documentation for VASP available on the internet. A few useful links:

- VASP manual: http://cms.mpi.univie.ac.at/vasp/vasp/vasp.html
- NSC information (quite old): http://www.nsc.liu.se/software/physics/vasp/
- Peter Larsson @ NSC (useful benchmarks): http://www.nsc.liu.se/~pla/

The basic VASP calculation consists of three steps.

INCAR is the central input file of VASP. It determines 'what to do and how to do it', and can contain a relatively large number of parameters. Most of these parameters have convenient defaults, and a user unaware of their meaning should not change any of the default values. Be very careful in dealing with the INCAR file, it is the main source of errors and false results!

- INCAR
System = system description # A description of the calculation, not important ISTART = 0 # Read old wavefunctions? 0 no, ISMEAR = -5 # -5 tetrahedron method with Blöchl corrections, good for semiconductors ISIF = 2 # 2 = relax ionic positions only IBRION = 2 # 2 = use CG algorithm PREC = ACCURATE # The general level of accuracy in the calculate. # Possible values: Low | Medium | High | Normal | Single | Accurate NSW = 20 # Defines the maximum number of relaxation steps. NELMIN = 6 # Do a minimum of 6 SCF iterations NELM = 100 # Do a maximum of 100 SCF iterations EDIFF = 5E-6 # Convergence criteria for the SCF iterations in eV ENCUT = 300 # Cut-off energy, determines the number of plane waves in the basis set, system dependent. # Defaults to the largest value of ENMAX in the POTCAR. NBANDS = 50 # Number of bands in the calculation, system and temperature dependent. # There should always be some empty bands. One can check the occupancy in the OUTCAR. IALGO = 48 # The algorithm used for the SCF iterations # 38 = Blocked-Davidson algorithm (DAV), default, most stable method. # 48 = Residual minimization method direct inversion in the iterative subspace (RMM-DIIS), faster then (DAV) LCHARG = .FALSE. # Whether or not the charge density output file CHGCAR should be written, often a quite large file. LWAVE = .FALSE. # Whether or not the wave function output file WAVECAR should be written, often a quite large file.

When doing calculations it is important to check that VASP reaches a solution before “NELM” SCF iterations or “NSW” ionic steps has been reached.

A list of INCAR parameters http://cms.mpi.univie.ac.at/wiki/index.php/Category:INCAR

The POSCAR input file determines the geometry to be simulated. It consists of the following lines

- An arbitrary comment line describing the system
- A scaling factor for the lattice vectors, usually the lattice constant. If it is given with a minus sign it's interpreted as the unit cell volume instead
- The first lattice vector
- The second lattice vector
- The third lattice vector
- A list of the elements in the system i.e. “Ti” or “Ti N”. This line is new for version 5 of VASP.
- A list of the number of elements if each species in the above line
- Determines the the atomic coordinates are given in “direct coordinates” i.e. as a fraction of the lattice vectors, or in cartesian coordinates in Ångström. Possible values: Direct | Cartesian
- The following lines contain the atomic coordinates for each atom.

- POSCAR
Cubic BN 3.57 0.0 0.5 0.5 0.5 0.0 0.5 0.5 0.5 0.0 B N 1 1 Direct 0.00 0.00 0.00 0.25 0.25 0.25

A detailed description of the format: http://cms.mpi.univie.ac.at/vasp/vasp/POSCAR_file.html

The KPOINTS file specify which points to use in the integration over the Brillouin zone. It can look in many different way but a common way it the following:

- A comment line describing the KPOINT mesh
- 0 = An automatic generation scheme
- The automatic generation scheme either “Monkhorst Pack” (default) or “Gamma”, for hexagonal lattices Gamma should be used.
- The number of k-points in each direction of the reciprocal lattice. The total number of points is usually reduced by using the lattice symmetry.

- KPOINTS
Comment 0 Monkhorst Pack 35 35 35 0 0 0

A detailed description of the format: http://cms.mpi.univie.ac.at/vasp/vasp/KPOINTS_file.html

The POTCAR file contains the parametrization of the pseudo-potential for each element, and hence the number of electrons (ZVAL) that are threaten for each species. The number of electrons in the system is of importance when choosing the number of bands (NBAND in INCAR file) to use among other things. It also specifies which functional that's used (LDA or PBE). The POTCAR file contains the individual species POTCARs concatenated after each others in the same order as in the POSCAR.

On NSC machines the POTCAR files can be found in

- /software/apps/vasp/Potentials/5.2/PAW/LDA
- /software/apps/vasp/Potentials/5.2/PAW/PBE

To generate the POTCAR the “cat” tool can be used:

cat POTCAR_1 POTCAR_2 ... POTCAR_n > POTCAR

First lines of Iron POTCAR:

PAW_PBE Fe 06Sep2000 8.00000000000000000 parameters from PSCTR are: VRHFIN =Fe: d7 s1 LEXCH = PE EATOM = 594.4687 eV, 43.6922 Ry TITEL = PAW_PBE Fe 06Sep2000 LULTRA = F use ultrasoft PP ? IUNSCR = 1 unscreen: 0-lin 1-nonlin 2-no RPACOR = 2.000 partial core radius POMASS = 55.847; ZVAL = 8.000 mass and valenz RCORE = 2.300 outmost cutoff radius RWIGS = 2.460; RWIGS = 1.302 wigner-seitz radius (au A) ENMAX = 267.883; ENMIN = 200.912 eV RCLOC = 1.701 cutoff for local pot LCOR = T correct aug charges LPAW = T paw PP EAUG = 511.368 DEXC = -.022 RMAX = 2.817 core radius for proj-oper RAUG = 1.300 factor for augmentation sphere RDEP = 2.442 radius for radial grids QCUT = -4.437; QGAM = 8.874 optimization parameters Description l E TYP RCUT TYP RCUT 2 .000 23 2.300 2 .000 23 2.300 0 .000 23 2.300 0 .000 23 2.300 1 -.200 23 2.300 1 .000 23 2.300 3 .000 7 .000 Error from kinetic energy argument (eV) NDATA = 100 STEP = 20.000 1.050 97.7 96.4 95.6 94.1 93.3 91.6 89.9 88.9 87.0 85.0 84.0 81.9 79.7 77.4 75.1 72.7 70.3 67.9 65.4 61.7 59.2 56.6 52.9 50.4 47.9 44.3 40.8 38.4 35.1 31.9 28.8 25.9 23.2 20.6 18.2 16.0 13.3 11.5 9.41 8.01 6.38 5.01 3.88 2.94 2.20 1.60 1.15 .727 .490 .288 .162 .101 .573E-01 .367E-01 .282E-01 .268E-01 .268E-01 .258E-01 .231E-01 .190E-01 .144E-01 .102E-01 .691E-02 .438E-02 .323E-02 .271E-02 .262E-02 .260E-02 .239E-02 .203E-02 .152E-02 .106E-02 .729E-03 .564E-03 .513E-03 .508E-03 .481E-03 .409E-03 .300E-03 .215E-03 .160E-03 .144E-03 .142E-03 .131E-03 .105E-03 .764E-04 .571E-04 .504E-04 .494E-04 .447E-04 .342E-04 .254E-04 .210E-04 .204E-04 .186E-04 .144E-04 .107E-04 .960E-05 .924E-05 .776E-05 END of PSCTR-controll parameters local part 81.2582175088937078 .86360772E+02 .86347328E+02 .86306975E+02 .86239767E+02 .86145770E+02 .86025081E+02 .85877820E+02 .85704132E+02 .85504188E+02 .85278179E+02 .85026322E+02 .84748853E+02 .84446029E+02 .84118129E+02 .83765451E+02 ... ... ...

To carry out the actual simulation a job script is usually created and submitted to the cluster for execution. The job file and how to submit it is described in the Basics section.

- job
#SBATCH -A liu1 #SBATCH -p green #SBATCH -J MY_FIRST_RUN #SBATCH -N 1 #SBATCH -t 10:00:00 cd /path/to/simulation/dir mpprun /software/apps/vasp/5.3.2-13Sep12/default/vasp-half

The last line calls VASP using the mpprun utility to manage the MPI parallelization. In the “/software/apps/vasp/5.3.2” directory, on NSCs clusters, there are many different version of VASP for different porpoises.

The VASP calculation will generate a number of different output files:

The OSZICAR and standard out (redirected to the slurm-<job number>.out file) contains important information about how the simulation has proceeded. A detailed description can be found at http://cms.mpi.univie.ac.at/vasp/vasp/stdout_OSZICAR_file.html.

mpprun INFO: Starting openmpi run on 1 nodes (8 ranks)... running on 8 nodes distr: one band on 1 nodes, 8 groups vasp.5.2.8 07Jul10 complex POSCAR found type information on POSCAR Ti N POSCAR found : 2 types and 8 ions LDA part: xc-table for Ceperly-Alder, standard interpolation POSCAR, INCAR and KPOINTS ok, starting setup WARNING: small aliasing (wrap around) errors must be expected FFT: planning ...( 1 ) reading WAVECAR entering main loop N E dE d eps ncg rms rms(c) DAV: 1 0.303618115885E+03 0.30362E+03 -0.12398E+04 960 0.118E+03 DAV: 2 -0.457474106077E+00 -0.30408E+03 -0.29460E+03 1000 0.309E+02 DAV: 3 -0.937031702450E+02 -0.93246E+02 -0.91504E+02 1568 0.163E+02 DAV: 4 -0.987826784141E+02 -0.50795E+01 -0.50670E+01 1232 0.443E+01 DAV: 5 -0.989208737402E+02 -0.13820E+00 -0.13815E+00 1376 0.736E+00 0.242E+01 DAV: 6 -0.797499355133E+02 0.19171E+02 -0.17124E+02 1392 0.760E+01 0.115E+01 DAV: 7 -0.778572721217E+02 0.18927E+01 -0.21580E+01 1496 0.321E+01 0.410E+00 DAV: 8 -0.780775434854E+02 -0.22027E+00 -0.87662E+00 1128 0.234E+01 0.421E+00 DAV: 9 -0.775039116726E+02 0.57363E+00 -0.21855E+00 1104 0.119E+01 0.848E-01 DAV: 10 -0.774891689351E+02 0.14743E-01 -0.10357E-01 1328 0.218E+00 0.462E-01 DAV: 11 -0.774831413887E+02 0.60275E-02 -0.31284E-02 1136 0.123E+00 0.149E-01 DAV: 12 -0.774830500968E+02 0.91292E-04 -0.10262E-02 1528 0.620E-01 0.135E-01 DAV: 13 -0.774826562241E+02 0.39387E-03 -0.20010E-03 1488 0.280E-01 0.259E-02 DAV: 14 -0.774826542175E+02 0.20066E-05 -0.27251E-05 1128 0.311E-02 1 F= -.77482654E+02 E0= -.77479348E+02 d E =-.992006E-02

The OUTCAR file contains most of the important results that you will want to look at although it is very long and things can often be a bit hard to find in the beginning. The format of the OUTCAR can vary a lot depending on the set off supplied parameters in the INCAR but it often consist of the following major elements:

- A small header containing the VASP version
- A description of the POTCARs that are used
- A description of the POSCAR
- A description of the reciprocal lattice with k-points, here is also the number of k-points that are left after the symmetry reduction.
- A long description with the parameters from the INCAR, and also most other parameters that have default values.
- A few short sections containing details on the simulation that is to be performed.
- After this the calculation starts a block of information is written with information about each SCF step.
- When SCF convergence is reached a block with results is written containing energies and various information. This is usually the part you are looking for.
- If more ionic steps are to be performed the last to steps are repeated. In this case you usually want the results from the last one.
- A final section containing some statistics from the simulation. Here it is possible to see how much memory and time the simulation consumed.

Part of the OUTCAR result block after SCF convergence has been reached:

FREE ENERGIE OF THE ION-ELECTRON SYSTEM (eV) --------------------------------------------------- free energy TOTEN = -77.482654 eV energy without entropy= -77.472734 energy(sigma->0) = -77.479348 CHARGE: cpu time 0.04: real time 0.04 FORLOC: cpu time 0.00: real time 0.00 FORNL : cpu time 0.30: real time 0.30 STRESS: cpu time 0.85: real time 0.85 FORCOR: cpu time 0.02: real time 0.02 FORHAR: cpu time 0.00: real time 0.00 MIXING: cpu time 0.00: real time 0.00 FORCE on cell =-STRESS in cart. coord. units (eV): Direction XX YY ZZ XY YZ ZX -------------------------------------------------------------------------------------- Alpha Z 184.02489 184.02489 184.02489 Ewald -505.23404 -505.23404 -505.23404 0.00000 0.00000 0.00000 Hartree 80.47130 80.47130 80.47130 0.00000 0.00000 0.00000 E(xc) -148.55374 -148.55374 -148.55374 0.00000 0.00000 0.00000 Local -87.15080 -87.15080 -87.15080 0.00000 0.00000 0.00000 n-local -58.89710 -58.41662 -52.53501 -0.37058 -0.19271 -0.14650 augment 30.71091 30.71091 30.71091 0.00000 0.00000 0.00000 Kinetic 468.22941 493.33300 502.86392 1.76093 1.26206 0.50026 Fock 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 ------------------------------------------------------------------------------------- Total -14.20563 -14.20563 -14.20563 0.00000 0.00000 0.00000 in kB -267.18553 -267.18553 -267.18553 0.00000 0.00000 0.00000 external pressure = -267.19 kB Pullay stress = 0.00 kB

In an VASP calculation, when the Kohn-Sham equations are solved, the equation system are expanded in a basis set of plane waves. And the infinite lattice is replaced by a unit cell and a sum over reciprocal lattice vectors (k-points) using the Block theorem. In principle the number of k-points and the size of the basis could be infinite, but in practice it has to be limited to a finite number.

The energy cut off (ENCUT in the INCAR) determines how many plane wave to use in the expansion of the wave functions. The more wave the better the expansion and hence the result, but the larger the equation system to solve. Hence a compromise has to be found. Usually there is a number of planes wave where the result will not change in any significant way when more waves are added. So this is the number of plane wave will give good results while resulting in a as small as possible equation system. The POTCAR contains default values for the ENCUT for each elements. But in some cases and depending on what information that are to be calculated this value might need to be adjusted by specifying it in the INCAR.

When the Bloch theorem is used, the infinite lattice is reduced to a sum of unit cell calculation for each reciprocal lattice vector. In principle there are an infinite number of reciprocal lattice vectors, but it turns out we can get very good results just by considering a small sample of them. The size if this sample is specified in the K-POINTS file. The result of the simulation should not depend on the selection of K-POINTS, hence the sample size needs to be increased until the result is invariant of the number of K-POINTS.

Example of the energy dependence on k-points and energy cut off

mml/vasp.txt · Last modified: 2012-10-24 11:10 by sergeis