Page 1 of 1

bader charges

Posted: Thu Sep 28, 2006 4:10 am
by achar002
Hi

I downloaded the Bader version in cvs couple of days ago. I ran the code on a neutral C42B8 (carbon and boron) graphite unit cell using the bader -b ongrid CHGCAR command. I have used USPP-PW91 POTCAR from VASP. I get the following charges on carbon and boron.

# X Y Z CHARGE MIN DIST
----------------------------------------------------------------
1 -6.2895 10.9028 5.0869 4.9818 0.4038
2 1.2435 0.6870 5.0869 4.0903 0.4078
3 -3.8639 10.9279 5.0869 4.8432 0.2722
4 -1.1936 10.9202 5.0869 4.9120 0.3911
5 7.5701 0.0159 5.0869 4.0147 0.3142
6 3.7654 10.9245 5.0869 4.8887 0.2915
7 6.3195 0.7283 5.0869 4.1775 0.3940
8 8.8594 0.7038 5.0869 4.0396 0.4131
9 -1.2354 2.2663 5.0869 4.8591 0.2930
10 10.0411 4.3619 5.0869 4.9712 0.2859
11 -3.7619 6.5807 5.0869 3.9263 0.3181
12 -5.0545 8.7660 5.0869 4.0589 0.3263
13 0.0354 2.9215 5.0869 4.0163 0.3326
14 -2.5080 7.2757 5.0869 5.0967 0.3512
15 -3.8664 9.5289 5.0869 4.9677 0.4214
16 1.2664 2.1472 5.0869 4.1724 0.3902
17 0.0488 4.3666 5.0869 5.1067 0.4708
18 -1.2504 6.6323 5.0869 4.9900 0.3220
19 3.7730 2.2186 5.0869 5.0269 0.2958
20 1.2296 6.5727 5.0869 4.0183 0.3667
21 0.0215 8.8072 5.0869 4.0700 0.3383
22 6.2532 2.1658 5.0869 4.9363 0.3249
23 5.0884 4.3798 5.0869 4.0043 0.3525
24 3.7916 6.5748 5.0869 3.9860 0.3575
25 8.9053 2.1282 5.0869 4.9418 0.4113
26 7.5757 4.4131 5.0869 5.0155 0.3484
27 5.0556 8.7905 5.0869 4.0129 0.3412
28 2.5154 2.8619 5.0869 4.9045 0.2948
29 1.2162 5.1277 5.0869 4.9567 0.3896
30 -0.0014 7.3470 5.0869 3.9865 0.3561
31 -1.1961 9.5213 5.0869 5.0259 0.3340
32 5.0269 2.9135 5.0869 4.0207 0.3796
33 3.8739 5.1324 5.0869 5.0347 0.3414
34 2.5004 7.2279 5.0869 4.9392 0.3537
35 1.2295 9.5463 5.0869 4.8758 0.3833
36 6.3393 5.0811 5.0869 4.9539 0.3158
37 5.0097 7.3661 5.0869 5.0406 0.4398
38 3.8246 9.5247 5.0869 5.0355 0.3560
39 10.1234 2.9195 5.0869 4.0855 0.3375
40 8.8266 5.1145 5.0869 4.0911 0.3204
41 7.6618 7.3285 5.0869 4.8863 0.2942
42 6.3449 9.4784 5.0869 4.0380 0.3794
43 -2.5331 8.7893 5.0869 0.0000 17.7657
44 2.5203 8.7542 5.0869 0.0000 21.1859
45 6.3273 6.5992 5.0869 0.0000 20.3020
46 2.5422 4.3848 5.0869 0.0000 20.3006
47 -1.2772 5.1094 5.0869 0.0000 20.3004
48 7.5877 2.8950 5.0869 0.0000 20.3004
49 3.7981 0.7049 5.0869 0.0000 17.7671
50 11.3947 0.7401 5.0869 0.0000 21.1861



I get zero charges on boron. It is supposed to have a positive charge. Shouldn't the sum of all charges be equal to zero? I have used a high precision fft grid. Could you give some suggestions to correct the results?

Posted: Thu Sep 28, 2006 10:20 am
by andri
This is not an uncommon problem when using pseudo potentials, though there seems to something else going on as well here. At least the minimum distance to the Bader surface for the B atoms is calculated to be rather large.

First thing to try is to add electron density around the B atoms in the hope that the maximum charged density will be shifted away from the bond to the core region. There are at least two simple ways to do that. First would be to use a psp for B that has more electrons (POTCAR_pv or _sv files). I can't find one for B so the second thing would be to use the trick suggested on the website, http://theory.cm.utexas.edu/vtsttools/bader/vasp.php . That requires recompiling VASP.

Also, can you try to run one of the pre-compilied executables on offered on the website. The CVS versions are always being tussled about.

But sometimes these problems simply can not be avoided, it is just an inherent difficulty with the pseudo potential formalism.

Posted: Thu Sep 28, 2006 4:30 pm
by achar002
Hi

If I add the lines given in http://theory.cm.utexas.edu/vtsttools/bader/vasp.php, will it affect rest of the VASP code or is it just an additional capability that I can use? Is there a tag that I will have to use to generate the Bader charges when running a VASP job or will it be generated automatically with the completion of a VASP job?

Thanks

Posted: Thu Sep 28, 2006 4:41 pm
by andri
I don't think it affects the rest of the code, forces and energies should remain the same. But just to be on the safe side then I always build this executables (which for historical reason we call vasp.TOTAL) and use it exclusively when generating CHGCARs with added partial number of core electrons.

One thing I just remember is that additional number of electrons is not added for all elements. Why that is I am not sure (this is some VASP black magic), but I remember that O gets no extra electrons while Si has 5.57 instead of 4. Somehow I recall that B doesn't get any. Maybe Graeme will remember it better from his SiB calculations.

Posted: Thu Sep 28, 2006 5:27 pm
by graeme
I think that's right - we didn't get any extra charge for B. But that was for the ultrasoft potentials. I guess it's possible that it could help for the PAW potentials, but I wouldn't be too hopeful.

The charges sum to the total number of electrons. To get the actual charge on the ions, you have to add in the positive nuclear (valance) charge your self.

If you are using the code from cvs, it would be better to use the default (neargrid) algorithm, instead of the original (ongrid) algorithm. If you just run 'bader CHGCAR', the results should be similar to what you have found, but somewhat more accurate.