VCAIMAGES: Difference between revisions
No edit summary |
No edit summary |
||
Line 27: | Line 27: | ||
File handling: When VASP is started it reads the file {{TAG|INCAR}} in the root directory. | File handling: When VASP is started it reads the file {{TAG|INCAR}} in the root directory. | ||
The following tags are also read from the root {{TAG|INCAR}} file: | The following tags are also read from the root {{TAG|INCAR}} file: | ||
{{TAG| | {{TAG|NCORE_IN_IMAGE1}}, {{TAG|IMAGES}}, {{TAG|KIMAGES}}, {{TAG|FOURORBIT}}, {{TAG|KPAR}}, {{TAG|NCORE}}, | ||
{{TAG|NCORES_PER_BAND}}, {{TAG|NPAR}}, {{TAG|NCSHMEM}}. Subsequently, VASP | {{TAG|NCORES_PER_BAND}}, {{TAG|NPAR}}, {{TAG|NCSHMEM}}. Subsequently, VASP | ||
splits the MPI communicator into two subgroups and continues reading from the {{TAG|INCAR}} | splits the MPI communicator into two subgroups and continues reading from the {{TAG|INCAR}} |
Revision as of 08:28, 31 March 2020
VCAIMAGES = [real]
Default: VCAIMAGES = -1
Description: The tag VCAIMAGES allows to perform thermodynamic coupling constant integrations. To achive this two molecular dynamics simulations are performed with e.g. different POTCAR or KPOINT files or different exchange-correlation functionals and force averaging between the two calculations is done.
The tag VCAIMAGES internally splits the available nodes into two groups, and each group then performs an independent VASP calculation (this implies VCAIMAGES only works in the MPI version). This behavior is implemented in the same way as the nudged elastic band method described under the tag IMAGES. As opposed to the nudged elastic band method, VASP creates always two images by setting the tag IMAGES=2 internally. The two calculations are performed in the subdirectories 01 and 02 (as opposed to the nudged elastic band method, the subdirectories 00 and 03 are not required). A precise description of the file reading and writing is given below.
The two VASP calculations are essentially performed independently in the subdirectories 01 and 02. Only the forces and energies of the two calculations are averaged according to the tag VCAIMAGES. Specificially, the value supplied in the tag VCAIMAGES determines the weight of the calculations performed in the subdirectory 01. The weight of the second image is 1-VCAIMAGES. The averaging is performed after self-consistency has been reached and affects the final total energy as well as the forces. Since the energies and forces are averaged, the trajectories generated by the two simulations will be exactly identical. Make sure though, that the initial POSCAR files in the two subdirectories are exactly identical. The averaged energies can be found in the OUTCAR file after the lines "ENERGY OF THE ELECTRON-ION-THERMOSTAT SYSTEM (eV)", as well as in the file OSZICAR (in the lines writing the free energy "F="). This makes scanning the OSZICAR file for the required energies of the individual calculations somewhat akward.
File handling: When VASP is started it reads the file INCAR in the root directory.
The following tags are also read from the root INCAR file:
NCORE_IN_IMAGE1, IMAGES, KIMAGES, FOURORBIT, KPAR, NCORE,
NCORES_PER_BAND, NPAR, NCSHMEM. Subsequently, VASP
splits the MPI communicator into two subgroups and continues reading from the INCAR
file in the subdirectories 01 and 02. If the INCAR files are not present in the subdirectories 01 or 02,
VASP will continue reading from the root INCAR file. The same logic is used from the files KPOINTS and POTCAR:
if they exist in the subdirectories, they will be read from the subdirectories 01 and 02; if they
are missing, the files are read from the root directory. The POSCAR and all other input files
are always read from the subdirectories 01 and 02. All output files (including OUTCAR and OSZICAR)
are always written to the subdirectories 01 and 02.
Special handling: In some cases it might be desirable to use a different number of cores for
the first image and the second image. For instance, this might be expedient for calculations
where thermodynmic integration is performed from a coarse to a dense k-point grid, or from a cheap
to an expensive exchange-correlation functional.
To set the number of cores in the first image the tag NCORE_IN_IMAGE1 has to be set. The second image then
contains the remaining cores.