Page 1 of 1

questions about charge distribution

Posted: Tue Jul 05, 2005 9:03 am
by baki
Dear developer team
Recently I performed some charge distribution calculation on formaldehyde in 12 A box.
I got the following results.
voronoi bader
O 3.7419 3.8518 6.6390 6.3367 7.7085 64.2371 1.0012
H 2.1996 5.1075 7.0927 1.4285 0.9554 7.9616 0.3461
H 3.9413 5.8534 6.9825 1.4279 1.0053 8.3777 0.3063
C 3.2904 4.9473 6.9067 2.8069 2.3308 19.4236 0.0120



and in the BCF file:




1 3.7500 3.7500 6.6000 7.7084 1 0.1093
2 2.3250 5.1000 7.0500 0.9553 2 0.1327
3 2.9250 5.0250 6.9750 1.1910 4 0.3798
4 9.9000 8.8500 0.6000 0.0002 2 7.9261
5 3.5250 5.2500 6.9000 1.1399 4 0.3831
6 3.9000 5.7750 6.9750 1.0055 3 0.0889




My problem is the following. According to the results, the two hydrogens are not equivalent if I calculated the charge with bader. (I didn't modify the vasp according your suggestion)
Can you explain these data?
best regards
baki

Posted: Tue Jul 05, 2005 7:16 pm
by graeme
The Bader charges of 0.9554 and 1.0053 on each hydrogen, resulting in an overall charge of +0.045 and -0.005, are likely different because of a limited resolution in the fft grid. If you look at the top of the CHGCAR file, or in the OUTCAR file, you can see the number of grid points used. To make sure your Bader charges are converged, you can set NGXF,NGYF, and NGZF in the INCAR file, to twice the values you are using now. Likely, you will see less deviation on the charge on the two hydrogen atoms.

When you calculate the larger CHGCAR file, you can set NSW=1 to do one ionic iteration, and have vasp read in your old CHGCAR file. The calculation should not take very long, especially for such a small molecule.

Please post back with the more accurate charges, so we know if this solved the problem.

Thank you

Posted: Wed Jul 06, 2005 7:24 am
by baki
Dear Graeme
I try to do your suggestion, and now the charge distribution is much better.
Voronoi bader
1 3.7419 3.8518 6.6390 6.3370 7.7101 64.2506 1.0584
2 2.1996 5.1075 7.0927 1.4282 0.9607 8.0062 0.3530
3 3.9413 5.8534 6.9825 1.4278 0.9780 8.1500 0.3352
4 3.2904 4.9473 6.9067 2.8070 2.3512 19.5931 0.0120
According to this results, when I performed the calculation every time necassary to do a additioanal one with more precise grid. If you put
into the program some three dimensional spline, than this may be can help too.
Thank again
baki

Posted: Wed Jul 06, 2005 4:56 pm
by graeme
This is good to see. Yes, it is best to generate an accurate CHGCAR file before each Bader calculation. For elements with few valence electrons, additional electrons can be included at this point, by using the _pv or _sv pseudopotentials. This step, although tedious, is not overly time consuming. As you say, it would be interesting to see if an interpolation scheme could reduce the required grid density. It would be important to design a spline which conserves the normalization of total charge. This has been on our list of things to do, but since it is fairly easy to do this with vasp, it has not been the highest priority.