VCA: Difference between revisions
No edit summary |
No edit summary |
||
| Line 3: | Line 3: | ||
Description: {{TAG|VCA}} is short for the virtual crystal approximation; the tag allows to "weight" each species found in the POTCAR file. | Description: {{TAG|VCA}} is short for the virtual crystal approximation; the tag allows to "weight" each species found in the POTCAR file. | ||
---- | ---- | ||
The tag {{TAG|VCA}} ''has'' to be supplied for each atom type or species found in the {{TAG|POTCAR}} | The tag {{TAG|VCA}} ''has'' to be supplied for each atom type or species found in the {{TAG|POTCAR}} {{TAG|POSCAR}} file, respectively . | ||
It weights the corresponding POTCAR files according to the values given in the {{TAG|INCAR}} file, with the default obviously | It weights the corresponding POTCAR files according to the values given in the {{TAG|INCAR}} file, with the default obviously | ||
being 1. For instance the formal valency found in the POTCAR files is multiplied by the supplied values, and likewise | being 1. For instance the formal valency found in the POTCAR files is multiplied by the supplied values, and likewise | ||
| Line 30: | Line 30: | ||
The implementation in VASP closely follows the methotology suggested by Bellaiche and Vanderbilt.<ref name="bellaiche2000VCA"/> | The implementation in VASP closely follows the methotology suggested by Bellaiche and Vanderbilt.<ref name="bellaiche2000VCA"/> | ||
Caveats: Unfortunately results of such VCA calculations are often not very reliable. | Caveats: Unfortunately results of such VCA calculations are often not very reliable, and even apparent in the original publucations by <ref name="bellaiche2000VCA"/>. The | ||
key point is that the used PAW potentials need to be constructed in such a manner that the pseudo atomic waves are very similar for the potentials | |||
that are "mixed" (in the example above this would be the Ge and Sn potentials). This can be achieved by carefully optimizing the radial cutoffs. Furthermore, | |||
the local potential needs to be very similar. This means that results for many of the standard potentials are not accurate. For instance, Vegard's law is | |||
often not even approximately observed (instead the volume is way too large at 50 %). The problem is particularly severe, if semi-core states | |||
are treated as valence states. For instance, for Ge and Sn the d electrons need to be treated as core electrons to obtain reasonable results. | |||
Revision as of 11:42, 16 April 2020
VCA = [real array]
Default: VCA = read from the POTCAR file
Description: VCA is short for the virtual crystal approximation; the tag allows to "weight" each species found in the POTCAR file.
The tag VCA has to be supplied for each atom type or species found in the POTCAR POSCAR file, respectively . It weights the corresponding POTCAR files according to the values given in the INCAR file, with the default obviously being 1. For instance the formal valency found in the POTCAR files is multiplied by the supplied values, and likewise the augmentation charges, and the non-local pseudopotential strenght paramters are scaled by the supplies values.
It is possible to use this flag to perform calculations in the framework of the virtual crystal approximation. Say you want to simulate Sn doping in a Ge lattice. Using a POTCAR file with a Ge and Sn data set and the following POSCAR file, this can be achieved:
cd:
1.00000000000000
2.82173 2.82173 0.00000
0.00000 2.82173 2.82173
2.82173 0.00000 2.82173
Ge Sn
2 2
Direct
0.00 0.00 0.00
0.25 0.25 0.25
0.00 0.00 0.00
0.25 0.25 0.25
If LVCA is set to
VCA = 0.99 0.01
the Ge atoms are weighted with a weight of 0.99, whereas the Sn atoms are weighted by 0.01 (see [1]). The implementation in VASP closely follows the methotology suggested by Bellaiche and Vanderbilt.[2]
Caveats: Unfortunately results of such VCA calculations are often not very reliable, and even apparent in the original publucations by [2]. The key point is that the used PAW potentials need to be constructed in such a manner that the pseudo atomic waves are very similar for the potentials that are "mixed" (in the example above this would be the Ge and Sn potentials). This can be achieved by carefully optimizing the radial cutoffs. Furthermore, the local potential needs to be very similar. This means that results for many of the standard potentials are not accurate. For instance, Vegard's law is often not even approximately observed (instead the volume is way too large at 50 %). The problem is particularly severe, if semi-core states are treated as valence states. For instance, for Ge and Sn the d electrons need to be treated as core electrons to obtain reasonable results.