Electron-phonon interactions from Monte-Carlo sampling: Difference between revisions

From VASP Wiki
No edit summary
No edit summary
Line 4: Line 4:
It also involves phonon calculations for the <math>\Gamma</math> point (see {{TAG|Phonons from finite differences}}). So many tags in the {{TAG|INCAR}} will be used from the phonon calculations.
It also involves phonon calculations for the <math>\Gamma</math> point (see {{TAG|Phonons from finite differences}}). So many tags in the {{TAG|INCAR}} will be used from the phonon calculations.


To enable electron-phonon interactions from MC methods {{TAG|PHON_LMC}}=''.TRUE.'' has to be set in the {{TAG|INCAR}} file. Also {{TAG|IBRION}}=6 has to be selected (the MC method is currently only implemented for {{TAG|IBRION}}=6).
To enable electron-phonon interactions from MC methods {{TAG|PHON_LMC}}=''.TRUE.'' has to be set in the {{TAG|INCAR}} file. Also {{TAG|IBRION}}=6 has to be selected (the Monte-Carlo (MC) method is currently only implemented for {{TAG|IBRION}}=6).


== Full MC sampling ==
== Full MC sampling ==
Line 23: Line 23:
  {{TAGBL|PHON_NSTRUCT}} = 100
  {{TAGBL|PHON_NSTRUCT}} = 100
  {{TAGBL|TEBEG}} = 0.0
  {{TAGBL|TEBEG}} = 0.0


The MC sampling code produces many {{TAG|POSCAR}} files with different distorted Wycoff positions but unchanged Brillouin matrix. The files are labeled as
POSCAR.TEBEG.NUMBER
where the NUMBER runs from 1 to {{TAGBL|PHON_NSTRUCT}}. 
== ZG configuration (one-shot sampling) ==
Alternatively to the full MC sampling, a one shot method, introduced by M. Zacharias and F. Giustino{{cite|zacharias:prb:2016}} (named ZG configuration after the authors), is available. This method only uses a single distorted structure and hence it is orders of magnitude faster than the full MC sampling, while it retains an accuracy very close to the full MC sampling for converged super cell sizes. For example we showed that for the zero-point renormalization of the band gap the accuracy is within 5 meV between the ZG configurations and the full MC sampling. Hence we suggest to use this method preferably, where convergence of the super cell size is hard to achieve or the 5 meV accuracy is no enough.
To select the ZG configuration {{TAG|PHON_NSTRUCT}}=0 has to be set in the {{TAG|INCAR}} file.
To make the simultaneous calculation of the ZG configuration at several temperatures possible, the temperatures (in K) have to be provided as a list in the {{TAG|INCAR}} file:
PHON_TLIST = 0.0 100.0 200.0 ...


== ZG configuration (one-shot sampling)




Line 34: Line 45:




 
== References ==
<references/>
----
----
[[Category:Lattice Vibrations]][[Category:Electron-phonon interactions]][[Category:Howto]][[Category:VASP6]]
[[Category:Lattice Vibrations]][[Category:Electron-phonon interactions]][[Category:Howto]][[Category:VASP6]]

Revision as of 10:33, 9 April 2019

Important: This feature will be only available from VASP6.0 or higher.

First of all this method needs a sufficiently large super cell. It also involves phonon calculations for the point (see Phonons from finite differences). So many tags in the INCAR will be used from the phonon calculations.

To enable electron-phonon interactions from MC methods PHON_LMC=.TRUE. has to be set in the INCAR file. Also IBRION=6 has to be selected (the Monte-Carlo (MC) method is currently only implemented for IBRION=6).

Full MC sampling

The tag PHON_NSTRUCT sets the number of structures generated due to the MC sampling. Convergence of the observable with respect to this number should be monitored.

The tag TEBEG=0 is also needed to choose the temperature at which the sampling is run.

Additionally the PHON_LBOSE can be set .TRUE. or .FALSE. (default PHON_LBOSE=.TRUE.), which selects Bose-Einstein or Maxwell-Boltzmann statistics, respectively.

A sample INCAR file for 0 K looks like the following:

System = DEFAULT
PREC = Accurate
ISMEAR = 0; SIGMA = 0.1;
IBRION = 6

PHON_LMC = .TRUE.
PHON_NSTRUCT = 100
TEBEG = 0.0

The MC sampling code produces many POSCAR files with different distorted Wycoff positions but unchanged Brillouin matrix. The files are labeled as

POSCAR.TEBEG.NUMBER

where the NUMBER runs from 1 to PHON_NSTRUCT.


ZG configuration (one-shot sampling)

Alternatively to the full MC sampling, a one shot method, introduced by M. Zacharias and F. Giustino[1] (named ZG configuration after the authors), is available. This method only uses a single distorted structure and hence it is orders of magnitude faster than the full MC sampling, while it retains an accuracy very close to the full MC sampling for converged super cell sizes. For example we showed that for the zero-point renormalization of the band gap the accuracy is within 5 meV between the ZG configurations and the full MC sampling. Hence we suggest to use this method preferably, where convergence of the super cell size is hard to achieve or the 5 meV accuracy is no enough.

To select the ZG configuration PHON_NSTRUCT=0 has to be set in the INCAR file.

To make the simultaneous calculation of the ZG configuration at several temperatures possible, the temperatures (in K) have to be provided as a list in the INCAR file:

PHON_TLIST = 0.0 100.0 200.0 ...



A sample INCAR file for the


References