Charged Supercell Calculations / Dipoles

[aemly]

Hello,
I have a question about performing calculations on positively charged (1 less electron) cubic supercells and determining if there is a dipole effect on differently sized supercells.

I've seen here ( http://cms.mpi.univie.ac.at/vasp-forum/ ... .php?4.909 ) that IDIPOL = 4 should ONLY be used for isolated molecules, however, this value is apparently the default since when I do not set IDIPOL, it defaults to 4. This contradicts the suggestion here ( http://cms.mpi.univie.ac.at/vasp-forum/ ... .php?4.951 ) which does not say to set IDIPOL and instead to just set DIPOL. The link that the admin posted to reference in this suggestion is also dead.

If I set the DIPOL tag, ironically, the coordinates set for DIPOL aren't even found in my OUTCAR, the only things under "Dipole corrections" are:
Dipole corrections
LMONO = F monopole corrections only (constant potential shift)
LDIPOL = T correct potential (dipole corrections)
IDIPOL = 2 1-x, 2-y, 3-z, 4-all directions
EPSILON= 1.0000000 bulk dielectric constant

For this case, my IDIPOL and LDIPOL are flagged in my INCAR file.

I have tried a variety of different parameters to figure out what might actually be the right way to do this since there's a smattering of different possibilities I found through the VASP archives. These include:
Setting the dipole direction specifically to either X, Y, Z or all directions using IDIPOL, setting the dipole position using DIPOL, switching between the LMONO and LDIPOL flag, and then all combinations of those three flags.

I recently saw something about the LCACLPOL tag here ( http://cms.mpi.univie.ac.at/vasp/vasp/L ... again.html ), and by setting this, it claims that I should have a line that says "Ionic dipole moment:" in my OUTCAR, but I do not yet (the calculation is still running though, so there might be some hope).

My NELECT is set to one electron less than what the original run has for electrons without setting the NELECT tag.

For a 135 atom supercell (lattice constant of ~12A), the majority of the methods to converge at all. My dE jumps all over the place (from e+05 and then gets down to e+0 but then jumps back up to e+6) during the electronic SC steps. I have had better luck with getting convergence for a smaller supercell, which makes sense.

I'm using the tetrahedron method for smearing (ISMEAR = -5 ) since I'm interested in energy only, and this compound is an insulator. My dipole should be perfectly in the y direction (which was confirmed when I tried to correct the dipole in the X and Z directions and my calculation never converged). I'm running VASP 5.2.12.

I'd appreciate any guidance as to what exactly should be done for calculations on charged cubic super cells to determine dipole corrections.


Thanks so much,
Alex


<span class='smallblacktext'>[ Edited ]</span>

[tlchan]

[quote author= 4 should ONLY be used for isolated molecules, however, this value is apparently the default since when I do not set IDIPOL, it defaults to 4.
[/quote]
For a 135 atom supercell (lattice constant of ~12A), the majority of the methods to converge at all. My dE jumps all over the place (from e+05 and then gets down to e+0 but then jumps back up to e+6) during the electronic SC steps. I have had better luck with getting convergence for a smaller supercell, which makes sense.
[/quote]
This is probably related to the potential correction. From the VASP manual, it is mentioned that turning on potential correction may lead to convergence issues. I think you can decide at a later stage whether you need such correction after you have a pre-converged calculation.