Different charges using vasp 5.2.11 and vasp 5.2.12

[michelle]

I have just calculated the Bader charges for bulk SiC using VASP 5.2.11 and VASP 5.2.12 and I get different results (see below).

I used identical POSCAR, KPOINTS and INCAR files. I am using PAWs and I added LAECHG=.TRUE to the INCAR file and followed the instructions at http://theory.cm.utexas.edu/vtsttools/bader/vasp.php. So the only difference is the version of the code I used.


INCAR:

System = SiC bulk

NPAR = 8
IALGO = 48
ISPIN = 2
PREC = ACCURATE
ENCUT = 400
ISMEAR = -5
ADDGRID = .TRUE.
LAECHG=.TRUE.
__________________________________________________________________________________________________________________
ACF.dat file: vasp/5.2.11

# X Y Z CHARGE MIN DIST ATOMIC VOL VALENCE CHARGE
--------------------------------------------------------------------------------
1 -0.0001 3.0887 10.1122 2.4223 0.4998 9.4087 4 1.58
2 1.7833 0 7.5833 2.6037 0.5064 9.7507 4 1.40
3 0.8916 1.5443 5.0556 2.5121 0.5068 9.4699 4 1.49
4 1.7833 0 2.5267 2.3754 0.495 8.9969 4 1.62
5 0 0 1.9007 5.6453 0.5523 11.9415 4 -1.65
6 1.7833 0 4.4228 5.4941 0.5378 11.4182 4 -1.49
7 0.8916 1.5443 6.9574 5.4298 0.6005 11.276 4 -1.43
8 1.7833 -0.0001 9.4795 5.5172 0.5487 11.2947 4 -1.52
--------------------------------------------------------------------------------
NUMBER OF ELECTRONS: 32 31.9999

_____________________________________________________________________________________________________
ACF.dat: vasp/5.2.12

# X Y Z CHARGE MIN DIST ATOMIC VOL VALENCE CHARGE
--------------------------------------------------------------------------------
1 -0.0001 3.0887 10.1122 1.0674 0.4998 4.1816 4 2.93
2 1.7833 0 7.5833 1.2417 0.5064 4.7485 4 2.76
3 0.8916 1.5443 5.0556 1.1856 0.5068 4.6381 4 2.81
4 1.7833 0 2.5267 1.2422 0.495 4.7642 4 2.76
5 0 0 1.9007 6.7361 0.5013 16.0768 4 -2.74
6 1.7833 0 4.4228 6.8331 0.5378 16.2835 4 -2.83
7 0.8916 1.5443 6.9574 6.7385 0.5321 16.1029 4 -2.74
8 1.7833 -0.0001 9.4795 6.9553 0.5487 16.7609 4 -2.96
--------------------------------------------------------------------------------
NUMBER OF ELECTRONS: 32 31.9999


Are you aware of any changes between the 2 versions of the code that would results in the different charges?

[graeme]

This is very puzzling; I would like to get to the bottom of this. If you are willing to make the charge density files available, I'll take a look.

Also, I'm a little curious about the output of the ACF file, which contains some extra data from what we print. Is this done manually, or has the code been modified to generate it?

[michelle]

Thanks for the prompt reply!

Re the extra data in the ACF files- i.e. last 2 columns- I actually added this manually after the calculation finished. Apologies for the confusion!

I am happy for you to look at the CHGCAR files- should I upload them here? They are about 8.2MB each.

[graeme]

Yes, you should be able to upload them directly here. It would be helpful to have the AECCAR0 and 2 files, as well as the CHGCAR files. If you can tar/zip them, that would be best.

[michelle]

The files you requested are attached.

Thanks.

[graeme]

Can you post the INCAR, POSCAR, KPOINTS, POTCAR and OUTCAR files as well? There is something strange about the AECCAR0 files, but I can't tell what without the details of the calculation.

It is very puzzling!

[michelle]

Here are the files. Thanks!

[graeme]

Try setting:

LADDGRID = .FALSE.
LELF = .FALSE.
NSW = 0

I'm not sure why yet, but the AECCAR0 files that are generated with LADDGRID=.TRUE. do not make sense. When set to .FALSE., things look ok. See if you still find a discrepancy between versions with these settings. Also, please do not trust any of the previous results.

[michelle]

I tried running the calculation with ADDGRID = .FALSE. and both versions of the code give exactly the same numbers (see below). Note I have manually added the 2 columns on the end to calculate the charges:

# X Y Z CHARGE MIN DIST ATOMIC VOL
--------------------------------------------------------------------------------
1 -0.0001 3.0887 10.1122 1.377 0.6199 5.185 4 2.62
2 1.7833 0 7.5833 1.3801 0.613 5.1234 4 2.62
3 0.8916 1.5443 5.0556 1.3837 0.6133 5.1962 4 2.62
4 1.7833 0 2.5267 1.3752 0.6182 5.0977 4 2.62
5 0 0 1.9007 6.6232 0.9956 15.7899 4 -2.62
6 1.7833 0 4.4228 6.6176 0.9973 15.6907 4 -2.62
7 0.8916 1.5443 6.9574 6.6189 0.9956 15.7702 4 -2.62
8 1.7833 -0.0001 9.4795 6.6243 1.0011 15.7034 4 -2.62
--------------------------------------------------------------------------------
NUMBER OF ELECTRONS: 32

That seems to have resolved the problem- thanks!

I also tried doubling the NG(X,Y,Z)F grid values and get identical charges as with the smaller grid, i.e. Si = 2.62 and C = -2.62. Do you think I can trust these values?

[graeme]

Yes, things are working as they should!

[michelle]

Thanks very much! :)