Page 1 of 1

Hydroxyl Problem

Posted: Mon Mar 05, 2007 10:24 pm
by aarondesk
I'm trying to calculate the charges of several hydroxyl species, but always get the same answer. Zero charge (+1 net) on the hydroxyl. Something doesn't appear right. Here's the output of methanol (1 O, 2 C, 3-6 H).

# X Y Z VORONOI BADER % MIN DIST
------------------------------------------------------------------------------------------
1 0.0154 0.1134 14.2319 5.8175 7.7225 55.1607 0.9073
2 0.0016 0.0286 0.6597 2.4659 3.2121 22.9434 0.1014
3 14.9900 1.0599 1.0344 1.4326 0.9798 6.9986 0.3249
4 14.1031 14.5109 1.0391 1.4509 1.0393 7.4238 0.3180
5 0.9000 14.5247 1.0570 1.4494 1.0462 7.4732 0.2894
6 0.0345 14.2100 13.8732 1.3836 0.0001 0.0004 2.2102

As can be seen, no charge was assigned to the hydrogen attached to the O. The MIN DIST is also fairly large. I've looked at the Bader volumes using VASPVIEW, and one isosurface of O actually extends over the adjacent H. Is it possible that the O is too large a sink, or this this an algorithm problem?

I'm using the PAW PBE potentials. I tested the H_h potential, but received the same answer. Any ideas?

Aaron

Posted: Mon Mar 05, 2007 11:13 pm
by graeme
This is expected with the pseudopotential calculation. The core charge around H is smoothed out so that there is no maximum at the H nucleus when it forms an OH bond.

Fortunately this can be fixed with the latest version of vasp (4.6.31). Adding the LAECHG=.TRUE. tag to your INCAR file will result in the AECCAR0, AECCAR1, and AECCAR2 files. The core charge is in AECCAR0 and the converged valance is in AECCAR2. Add these together using our chgsum.pl script: chgsum.pl AECCAR0 AECCAR2
The output of this script will be a charge density file (CHGCAR_sum) containing the total charge.

Then, the bader analysis can be run on your CHGCAR file using this total charge as a reference:

bader CHGCAR -ref CHGCAR_sum

Note: the instructions have been updated at http://theory.cm.utexas.edu/bader/vasp.php

Posted: Tue Mar 06, 2007 4:22 pm
by aarondesk
Thank you for the information!

Posted: Wed Mar 07, 2007 5:20 pm
by vozny
Graeme, I don't get how the added core charge can help the situation.
You say yourself that there is no maximum on H due to too smooth pseudopotential.
Core charge would add only core to oxygen, while H will still stay without maximum, so for Bader analysis nothing would changed.

Or, you want to say that partial charge would be added to H as well?

Posted: Wed Mar 07, 2007 10:38 pm
by graeme
It's a little confusing to talk about missing core charge for H. In the pseudopotential calculations, there is charge missing at the H nucleus, and this results in a missing Bader maximum. But, the charge is not missing from the calculation (there is still 1 valance electron in the pseudopotential calculation), is it just smoothed out to other regions in the H core.

The fix for vasp is to construct the true core charge and add it to the true valence charge (as described above). This is not the same as adding charge to H, instead, it results in redistributing charge in a way that corrects for the errors made by the pseudopotential. I don't know exactly what is involved in this correction, but I can see that it gives a sensible density in the H core, that there is a charge density maximum in O-H and C-H bonds, and that the Bader analysis gives a sensible (non-zero) charge to H atoms in molecules with these bonds.