asymmetric bader charges

[marius]

Hello,

For some symmetric molecule like a simple c2h4 i would expect to have same charges on each atom but when i align the c-c axis along one of the grid axis i got different charges on each carbon atom even though the total number of electrons is correct:
# X Y Z CHARGE MIN DIST ATOMIC VOL
--------------------------------------------------------------------------------
1 0.0000 0.0000 0.0000 6.1084 1.2590 335.1137
2 0.0000 0.0000 2.5228 6.0841 1.2439 308.7432
3 1.7822 0.0000 -1.0290 0.9518 0.6873 661.0278
4 1.7822 0.0000 3.5517 0.9521 0.6893 473.4900
5 -1.7822 0.0000 3.5517 0.9521 0.6895 473.4900
6 -1.7822 0.0000 -1.0290 0.9518 0.6876 661.0278
--------------------------------------------------------------------------------
NUMBER OF ELECTRONS: 16.00027

But when i align the molecule not along one of the grid axis i got the expected symmetric charge distribution
# X Y Z CHARGE MIN DIST ATOMIC VOL
--------------------------------------------------------------------------------
1 0.0000 0.0000 0.0000 6.0971 1.2321 329.9139
2 -2.1848 0.0000 -1.2614 6.0971 1.2321 329.9139
3 1.7822 0.0000 -1.0290 0.9511 0.6850 602.0452
4 -2.1848 0.0000 -3.3193 0.9518 0.6991 583.1358
5 -3.9670 0.0000 -0.2324 0.9511 0.6848 602.0452
6 0.0000 0.0000 2.0579 0.9518 0.6988 583.1358
--------------------------------------------------------------------------------
NUMBER OF ELECTRONS: 16.00000


This seems quite odd and i am also using the latest version (v0.26a 08/07/09). Any suggestions whats going wrong ?

marius

[graeme]

Can you check to see if the asymmetry goes away with increasing grid density? If it does not, there is a problem. Without knowing your grid density, we can't know if a 0.02 e difference is a problem.

[marius]

I think the grid density is high enough it is like this

calculation at a grid of 907* 551* 730 points

origin at 0.00000 0.00000 0.00000

normalised grid vectors:
x y z
1. vector: 0.00000 0.00000 1.00000
2. vector: 0.00000 1.00000 0.00000
3. vector: 1.00000 0.00000 0.00000

minimum increment maximum
1. vector: -9.05174 0.01998 9.05174
2. vector: -5.50000 0.02000 5.50000
3. vector: -7.28220 0.01998 7.28220

increasing the grid points does not change the charges anymore.

[Wenjie]

Hi, would you please send the charge file to twj916@mail.utexas.edu, so I can look into the problem. If the file is too large, please send me the link so that I can download it.

[yeops]

Hi,

I'm wondering if the underlying reason for this issue has been discovered? I'm facing this problem of asymmetrical bader charges on a symmetrical hydrocarbon species too. I'm using version 0.27 of the bader program. I'm willing to post the cube file for my system somewhere if someone is willing to look into this problem.

Thanks in advance!

[graeme]

Yes, I would like to take a look. You should be able to upload it to the forum.

[yeops]

Hi,

Please go to ftp://ftp.ihpc.a-star.edu.sg/outgoing/yeops/ and download the file called cube.zip. Inside contains the cube file and also a color map of the bader charges that are decidedly asymmetric.

Please let me know when you have downloaded the file so that I can remove it.

Thank you!

[graeme]

This looks to me like a problem in the charge density file, rather than a problem with the analysis. I believe that the problem is due to the fact that the total charge density is used in the cube file, so that there are peaks near the nuclei.

There are a couple of reasons that I think this. First, the total charge density sums to 974.82412, which is significantly different from the expected 975. When you try to represent the total charge density on a grid, you can get variation between atoms due to how the maximum in charge density near the nucleus falls on the charge density grid. The second reason is that you see the same variation in Voronoi charges, which says that the variation is present in simple geometry regions around each nuclei.

To avoid this problem using the vasp code, we typically calculate both the total and valence charge density. You can do the analysis on the total density and the integration on the valence:

bader chg_valence -ref chg_total

and in your case, since there is a lot of vacuum, it is more efficient to use (only in the latest version) the

-vac 1e-5

tag to avoid enumerating the 6031470 insignificant charge density maxima in the vacuum region.

[yeops]

Hi graeme,

Thanks for the suggestion! I'm not using VASP, but I added a gaussian core charge to each atomic center of my system so that the bader code will correctly identify the maximas. I tried the method of

bader chg_valence -ref chg_total

and while the results are slightly better, the charges are still not symmetrical. I have uploaded the new files at ftp://ftp.ihpc.a-star.edu.sg/outgoing/yeops/ . For my system that consists of a hexagonal network of carbon atoms, 3 out of the 6 carbons in a hexagon ring are supposed to have a slightly higher charge than the other 3. I had my doubts as to whether the bader method would be able to differentiate these subtle charge differences, but happily, from the color map, it seems like it could! But charge distribution is sadly, not symmetric for my symmetric molecule.

[graeme]

Ok, but again, I'm saying that the charge density differences are in the charge density file and not a problem with the Bader analysis. If you want, I can generate total and valence charge densities for this system using a different program (vasp) and see if the charges are symmetric. Using a periodic graphene system would be a good way to check this, since there should not be any difference between the atoms.

[yeops]

Hi graeme, I scaled my system down and tried the calculation with VASP; turns out that my system needs a VERY dense grid before the bader charges start to look symmetric...

[graeme]

Really?, even when doing the analysis on the total charge and integration on the valence:
bader chg_valence -ref chg_total

I thought that the issue was the location of the charge density maxima at the atomic maxima with respect to the grid. But when you do the integration over the valence charge, I would not expect that a dense grid is required.

[yeops]

I've attached the plots for the bader charges for my system. The numbers in the file name stands for the NGXF x NGYF x NGZF grid points. There is a marked difference for the bader charges of the hydrogens that are parallel to the x-axis in those plots especially. So even though the charge distributions for the carbons are still different from what I expect, this suggests to me that the asymmetry stems from a grid that is not dense enough?

And yes, I used bader chg_valence -ref chg_total to calculate the bader charges.