VASP¶
Introduction¶
VASP is a density-functional theory code using pseudopotentials or the projector-augmented wave method and a plane wave basis set. This interface makes it possible to use VASP as a calculator in ASE, and also to use ASE as a post-processor for an already performed VASP calculation.
Note
A new VASP calculator is currently in BETA testing, see
vasp2
,
which implements the calculator using the
FileIOCalculator
.
Environment variables¶
You need to write a script called run_vasp.py
containing
something like this:
import os
exitcode = os.system('vasp')
The environment variable VASP_SCRIPT
must point to that file.
A directory containing the pseudopotential directories potpaw
(LDA XC) potpaw_GGA
(PW91 XC) and potpaw_PBE
(PBE XC)
is also needed, and it is to be put in the environment variable
VASP_PP_PATH
.
Set both environment variables in your shell configuration file:
$ export VASP_SCRIPT=$HOME/vasp/run_vasp.py
$ export VASP_PP_PATH=$HOME/vasp/mypps
The following environment variable can be used to automatically copy the
van der Waals kernel to the calculation directory. The kernel is needed for
vdW calculations, see VASP vdW wiki, for more details. The kernel is looked
for, whenever luse_vdw=True
.
$ export ASE_VASP_VDW=$HOME/<path-to-vdw_kernel.bindat-folder>
The environment variable ASE_VASP_VDW
should point to the folder where
the vdw_kernel.bindat
file is located.
VASP Calculator¶
The default setting used by the VASP interface is
-
class
ase.calculators.vasp.
Vasp
(restart=None, output_template='vasp', track_output=False, **kwargs)[source]¶
Below follows a list with a selection of parameters
keyword |
type |
default value |
description |
---|---|---|---|
|
|
None |
Restart old calculation or use ASE for post-processing |
|
|
‘PW91’ |
XC-functional. Defaults to
None if |
|
|
None |
Additional setup option |
|
|
Set by |
Pseudopotential (POTCAR) set used (LDA, PW91 or PBE). |
|
various |
\(\Gamma\)-point |
k-point sampling |
|
|
None |
\(\Gamma\)-point centered k-point sampling |
|
|
None |
Use reciprocal units if k-points are specified explicitly |
|
|
None |
Net charge per unit cell (as
an alternative to specifying
the total charge |
|
|
Accuracy of calculation |
|
|
|
Kinetic energy cutoff |
|
|
|
Convergence break condition for SC-loop. |
|
|
|
Number of bands |
|
|
|
Electronic minimization algorithm |
|
|
|
Type of smearing |
|
|
|
Width of smearing |
|
|
|
Maximum number of SC-iterations |
|
|
|
LD(S)A+U parameters |
For parameters in the list without default value given, VASP will set
the default value. Most of the parameters used in the VASP INCAR
file
are allowed keywords. See the official VASP manual for more details.
Note
Parameters can be changed after the calculator has been constructed
by using the set()
method:
>>> calc.set(prec='Accurate', ediff=1E-5)
This would set the precision to Accurate and the break condition
for the electronic SC-loop to 1E-5
eV.
Exchange-correlation functionals¶
The xc
parameter is used to define a “recipe” of other parameters
including the pseudopotential set pp
. It is possible to override
any parameters set with xc
by setting them explicitly. For
example, the screening parameter of a HSE calculation might be
modified with
>>> calc = ase.calculators.vasp.Vasp(xc='hse06', hfscreen=0.4)
The default pseudopotential set is potpaw_PBE unless xc
or pp
is set to pw91
or lda
.
|
Parameters set |
---|---|
lda, pbe, pw91 |
|
pbesol, revpbe, rpbe, am05 |
|
tpss, revtpss, m06l |
|
vdw-df, optpbe-vdw |
|
optb88-vdw, obptb86b-vdw |
|
beef-vdw |
|
vdw-df2 |
|
hf |
|
pbe0 |
|
b3lyp |
|
hse03, hse06, hsesol |
|
It is possible for the user to temporarily add their own xc
recipes without modifying ASE, by updating a dictionary. For example,
to implement a hybrid PW91 calculation:
from ase.calculators.vasp import Vasp
Vasp.xc_defaults['pw91_0'] = {'gga': '91', 'lhfcalc': True}
calc = Vasp(xc='PW91_0')
Note that the dictionary keys must be lower case, while the xc
parameter is case-insensitive when used.
Setups¶
For many elements, VASP is distributed with a choice of pseudopotential setups. These may be hard/soft variants of the pseudopotential or include additional valence electrons. Three base setups are provided:
- minimal (default):
If a PAW folder exists with the same name as the element, this will be used. For the other elements, the PAW setup with the least electrons has been chosen.
- recommended:
corresponds to the table of recommended PAW setups supplied by the VASP developers.
- materialsproject:
corresponds to the Materials Project recommended PAW setups.
- gw:
corresponds to the table of recommended setups for GW supplied by the VASP developers.
Where elements are missing from the default sets, the Vasp Calculator
will attempt to use a setup folder with the same name as the element.
A default setup may be selected with the setups
keyword:
from ase.calculators.vasp import Vasp
calc = Vasp(setups='recommended')
To use an alternative setup for all instances of an element, use the
dictionary form of setups
to provide the characters which need
to be added to the element name, e.g.
calc = Vasp(xc='PBE', setups={'Li': '_sv'})
will use the Li_sv
all-electron pseudopotential for all Li atoms.
To apply special setups to individual atoms, identify them by their zero-indexed number in the atom list and use the full setup name. For example,
calc = Vasp(xc='PBE', setups={3: 'Ga_d'})
will treat the Ga atom in position 3 (i.e. the fourth atom) of the
atoms object as special, with an additional 10 d-block valence
electrons, while other Ga atoms use the default 3-electron setup and
other elements use their own default setups. The positional index may
be quoted as a string (e.g. {'3': 'Ga_d'}
).
These approaches may be combined by using the ‘base’ key to access a default set, e.g.
calc = Vasp(xc='PBE', setups={'base': 'recommended', 'Li': '', 4: 'H.5'})
Spin-polarized calculation¶
If the atoms object has non-zero magnetic moments, a spin-polarized calculation will be performed by default.
Here follows an example how to calculate the total magnetic moment of a sodium chloride molecule.
from ase import Atoms, Atom
from ase.calculators.vasp import Vasp
a = [6.5, 6.5, 7.7]
d = 2.3608
NaCl = Atoms([Atom('Na', [0, 0, 0], magmom=1.928),
Atom('Cl', [0, 0, d], magmom=0.75)],
cell=a)
calc = Vasp(prec='Accurate',
xc='PBE',
lreal=False)
NaCl.set_calculator(calc)
print(NaCl.get_magnetic_moment())
In this example the initial magnetic moments are assigned to the atoms
when defining the Atoms object. The calculator will detect that at least
one of the atoms has a non-zero magnetic moment and a spin-polarized
calculation will automatically be performed. The ASE generated INCAR
file will look like:
INCAR created by Atomic Simulation Environment
PREC = Accurate
LREAL = .FALSE.
ISPIN = 2
MAGMOM = 1*1.9280 1*0.7500
Note
It is also possible to manually tell the calculator to perform a spin-polarized calculation:
>>> calc.set(ispin=2)
This can be useful for continuation jobs, where the initial magnetic moment is read from the WAVECAR file.
Brillouin-zone sampling¶
Brillouin-zone sampling is controlled by the parameters kpts
,
gamma
and reciprocal
, and may also be set with the VASP
parameters kspacing
and kgamma
.
Single-parameter schemes¶
A k-point mesh may be set using a single value in one of two ways:
- Scalar
kpts
If
kpts
is declared as a scalar (i.e. a float or an int), an appropriate KPOINTS file will be written. The value ofkpts
will be used to set a length cutoff for the Gamma-centered “Automatic” scheme provided by VASP. (See first example in VASP manual.)- KSPACING and KGAMMA
Alternatively, the k-point density can be set in the INCAR file with these flags as described in the VASP manual. If
kspacing
is set, the ASE calculator will not write out a KPOINTS file.
Three-parameter scheme¶
Brillouin-zone sampling can also be specified by defining a number of subdivisions for each reciprocal lattice vector.
This is the second “Automatic” scheme described in the VASP manual.
In the ASE calculator, it is used by setting kpts
to a sequence of three int
values, e.g. [2, 2, 3]
.
If gamma` is set to ``True
, the mesh will be centred at the \(\Gamma\)-point;
otherwise, a regular Monkhorst-Pack grid is used, which may or may not include the \(\Gamma\)-point.
In VASP it is possible to define an automatic grid and shift the origin point.
This function is not currently included in the ASE calculator. The same result can be achieved by using ase.dft.kpoints.monkhorst_pack()
to generate an explicit list of k-points (see below) and simply adding a constant vector to the matrix.
For example,
import ase.dft.kpoints
kpts = ase.dft.kpoints.monkhorst_pack([2, 2, 1]) + [0.25, 0.25, 0.5]
creates an acceptable kpts
array with the values
array([[ 0. , 0. , 0.5],
[ 0. , 0.5, 0.5],
[ 0.5, 0. , 0.5],
[ 0.5, 0.5, 0.5]])
However, this method will prevent VASP from using symmetry to reduce the number of calculated points.
Explicitly listing the k-points¶
If an n-by-3 or n-by-4 array is used for kpts
,
this is interpreted as a list of n explicit k-points and an appropriate KPOINTS file is generated.
The fourth column, if provided, sets the sample weighting of each point.
Otherwise, all points are weighted equally.
Usually in these cases it is desirable to set the reciprocal
parameter to True
,
so that the k-point vectors are given relative to the reciprocal lattice.
Otherwise, they are taken as being in Cartesian space.
Band structure paths¶
VASP provides a “line-mode” for the generation of band-structure paths.
While this is not directly supported by ASE, relevant functionality exists in the ase.dft.kpoints
module.
For example:
import ase.build
from ase.dft.kpoints import bandpath
si = ase.build.bulk('Si')
kpts, x_coords, x_special_points = bandpath('GXL', si.cell, npoints=20)
returns an acceptable kpts
array (for use with reciprocal=True
) as well as plotting information.
LD(S)A+U¶
The VASP +U corrections can be turned on using the default VASP parameters explicitly, by manually setting
the ldaul
, ldauu
and ldauj
parameters, as well as enabling ldau
.
However, ASE offers a convenient ASE specific keyword to enable these, by using a dictionary construction, through the
ldau_luj
keyword. If the user does not explicitly set ldau=False
, then ldau=True
will automatically
be set if ldau_luj
is set.
For example:
calc = Vasp(ldau_luj={'Si': {'L': 1, 'U': 3, 'J': 0}})
will set U=3
on the Si p-orbitals, and will automatically set ldau=True
as well.
Restart old calculation¶
To continue an old calculation which has been performed without the interface
use the restart
parameter when constructing the calculator
>>> calc = Vasp(restart=True)
Then the calculator will read atomic positions from the CONTCAR
file,
physical quantities from the OUTCAR
file, k-points from the
KPOINTS
file and parameters from the INCAR
file.
Note
Only Monkhorst-Pack and Gamma-centered k-point sampling are supported
for restart at the moment. Some INCAR
parameters may not be
implemented for restart yet. Please report any problems to the ASE mailing
list.
The restart
parameter can be used , as the name suggest to continue a job from where a
previous calculation finished. Furthermore, it can be used to extract data from
an already performed calculation. For example, to get the total potential energy
of the sodium chloride molecule in the previous section, without performing any additional
calculations, in the directory of the previous calculation do:
>>> calc = Vasp(restart=True)
>>> atoms = calc.get_atoms()
>>> atoms.get_potential_energy()
-4.7386889999999999
New Calculator¶
A new VASP calculator is currently in BETA testing, see
vasp2
, which implements the calculator using the
FileIOCalculator
.