4
$\begingroup$

In certain charge densities I am getting negative values, in the place where the atomic cores would be. I do not have access to the vasp.xml or OUTCAR files, so there is no way to determine whether the ECD actually did converge. Are the negative charges just an artefact of DFT computations?

I speculate that the DFT simulation might have hit the maximum number of iterations and exited with the resultant Electron Charge Density. If the previous speculation is true, then the CHGCAR is generated from an un-converged VASP simulation.

P.S.- The DFT simulations were performed in VASP with the optb88vdw functional

$\endgroup$
2

1 Answer 1

5
$\begingroup$

Yes, it is possible to get regions of negative electronic charge density when using PAW or USP pseudopotentials. The reason is that these pseudopotentials take the "soft" charge density (the density calculated from the Kohn-Sham wavefunctions) and add an additional "correction" to the charge density, usually referred to as "augmentation". In much of the literature this "augmentation" is described as adding charge, and it usually does in total, but it can also lower the charge density in some regions. If the soft charge density in a given region is lower than the amount that the augmentation removes, then the result is a negative charge density.

This issue can arise for several reasons, but one common cause is Fourier aliasing, which happens when the augmentation grid is not fine enough to represent the augmentation charge density accurately. I recommend increasing the number of points in the augmented density grid - keeping the plane-wave cut-off energy the same - and see whether that helps.

$\endgroup$
4
  • $\begingroup$ So what would be an effective way to remove these negative charges from the final charge density, zeroing them out would affect the total number of electrons in the system right? $\endgroup$
    – Pranoy Ray
    Apr 22 at 19:41
  • 1
    $\begingroup$ @PranoyRay yes, zeroing them is usually a very bad idea. Increasing the augmentation grid resolution is usually the simplest method (fine_grid_scale in CASTEP, I think ecutrho for QE; I've never used VASP, but possibly enaug?). $\endgroup$ Apr 25 at 1:26
  • $\begingroup$ What if I systematically adjust the negative charges with the positive charges? $\endgroup$
    – Pranoy Ray
    Apr 25 at 13:25
  • $\begingroup$ @PranoyRay I'm not sure how you would do that, but in any case I don't recommend it. There isn't much point going to all the effort of computing the self-consistent ground state density if you're then going to change it arbitrarily. $\endgroup$ Apr 25 at 21:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .