Min Distance

Bader charge density analysis

Moderator: moderators

adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Min Distance

Post by adam2412 »

Dear Prof,

Can you please explain more on the meaning of min. distance from the ACF.dat?
I understand that it is the min. distance from the surface but what does it has to do with charges?

Is the partial charges given from the output are corresponding to the nuclear point only or it is a distribution of charges from point to the other point?

Thank you for your time.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Min Distance

Post by graeme »

The Bader charges are the total integrated charges in the Bader volumes around each atom. The minimum distance is between the atomic center and the nearest point on the surface of the Bader volume.

In fact, the minimum distance is not very important any more. At one time, when we couldn't get core charges from vasp, the Bader surfaces could come very close to the atomic centers, and this distance was a good sanity check to let you know if the analysis was failing due to missing core charge.
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Dear Prof,

So in that case, in order to check whether the analysis is a success or not from the min. distance, I have to know the radius of the atom + core charge?

Another thing is, I calculated Ru in a semiconductor material (closed shell) and I got this result. The partial charges is -3.7988 when it should be +3.. (looking from the formal charge).
# X Y Z CHARGE MIN DIST ATOMIC VOL
45 6.8454 2.5705 4.5951 11.7988 1.4902 31.1636
46 2.2818 7.7115 4.5951 11.7988 1.4902 31.1636

The total number of electron is larger than what it should be. Therefore I have increased the NG(X,Y,Z) in my next calculation.
Is this correct & sufficient or should I include other parameter?

Thank you for your comments.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Min Distance

Post by graeme »

No, just include the core charges and ignore the minimum distance information.

For the second issue, I doubt the grid size could cause such a difference. It is more likely another problem.

Make sure to follow the "Note for VASP Users" on http://theory.cm.utexas.edu/henkelman/code/bader/

Then, check the valence charge in our Ru potcar file so you know how the Bader charge of 11.7988 compares to the formal charge. And make sure to get the sign correct.
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Dear Prof,

What if the total number of electrons is smaller than what it should be?
Even if in a closed shell, the material should be optimized right?

Thank you for your comments.
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Dear Prof,

I have done everything written on the page that you have suggested.
I added AECCAR0 and AECCAR2 by chgsum.pl. Then, I run bader CHGCAR -ref CHGCAR_sum but still getting the same value.
I changed the LDA POTCAR file but still receive Ru as -3.

The problem is I'm calculating Ru in CeRu2Al10 in closed shell system.
I have added all the charge of the atom obtained from Bader; [Ce + Ru(2) + Al(10)] and the overall charge is 5 x 10e(-5): ~ 0.0
Looking at the overall charge, my analysis seems legit.
But, the charges of atom, Ce : +1.0848, Ru : -3.7988, Al : + 0.651285.
I'm trying to understand the Bader and VASP calculation that I have done. But I still couldn't find the answer why my charge is like this.
I'm stuck. I really need some help.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Min Distance

Post by graeme »

Well, at least the total charge looks reasonable.

Looking at your elements, I would expect Ru to accept charge from Cs and Al; what makes you think that Ru should be positive?
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Thank you, Prof for your comments..

Yes, the charge calculation seems legit. I assumed the charge should be positive from the electronic configuration because we still couldn't find the exact charge of Ru from any experiment.
From the experiment, we only know that Ce is +3. But my calculation gives Ce +1.

The atomic radius (calculated from the value of atomic volume) and the min. distance of all atoms < RCORE.
I have tried different POTCAR file for each atom; Ru_pv, Ce_s and Al_h but the result still gives min.distance < RCORE.

I have sum up the AECCAR0 and AECCAR2. But why the min.distance < RCORE?
I read from this thread viewtopic.php?f=1&t=59 but this thread was from 2006.
Can I still use this thread to explain my result or is there other reason for my result to behave like that?

Thank you ~
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Min Distance

Post by graeme »

If the distance is close to RCORE, then I wouldn't worry about it. Also, if the number of Bader maxima is the same as the number of atoms, you should probably be ok.
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Thank you, Prof for your comments...
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Dear Prof,

Since my result produce less than RCORE, I'm trying to understand how VASP works. Therefore, now I'm trying to understand the POTCAR file.
I look up different POTCAR file of Cerium and there are 2 different LDA POTCAR for Ce. Both of them have the same VRHFIN but different number of valency.
However for Ruthenium, the two POTCAR file has both different VRHFIN and valency.

At first, I thought VRHFIN ( the atomic configuration), determine the valency but looking from Cerium POTCAR, it seems that my understanding is wrong and I confuse again.

Could you please help me with this...

Thank you ~
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Min Distance

Post by graeme »

For elements with few valence electrons, vasp will often have another potential with the next innermost shell treated explicitly. These are labelled, for example as _pv where the inner p are treated as valence.
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Dear Prof. Graeme,

I guess I have to ask this questions after stuck for few weeks.
I'm trying to understand the Bader analysis by re-calculating using MATLAB . I read the journals and implement the same procedure to test my understanding towards Bader analysis.

Could you please correct my understanding if it's wrong...

Firstly, the min distance is the distance from charge density maxima to the zero-flux area that divide the atomic basin.
Secondly, the Bader volume is a volume of cuboid volume of X*Y = [(min.distance *2) ^2 ]. Z = length of atomic basin.

Thirdly, to calculate the partial charge, we sum up all the charge density from the grids inside the basins and we divide it with the number of grids. Is it correct? It says here "Integrate the electron density over all mesh volumes in the mesh. Let dq be the total charge contained with the volume element. Choose N random points in dV..... Once this occurs, assign dq/N to that Bader volume. End the program ".

If it is correct, why we have to divide total charge density with the number of grids? I couldn't find the answer from the journal.
If it is wrong, how exactly we can calculate the partial charge? I have sum up the total charge density only and couldn't fine the same answer.

Fourth, sometimes the journal says total charge density but sometimes it is total charge. I'm confused. But I guess, it is total charge density, right?

Thank you for your time and help.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: Min Distance

Post by graeme »

That all sounds pretty good with a few corrections.

1. The volume is not integrated in the way that you write, it is just the sum of the cell volumes assigned to an atomic basin.

2. The definition of charge density is a little funny for vasp files in that (if I remember correctly) it is actually the charge within each cell, so it is actually the charge density times the volume. But anyway, if you are looking at the code, you will have to check what the density gets multiplied by when the input density file is read in to make sense of it.

3. I don't understand the "N random points" at all; there is nothing random about the algorithm. All points are systematically assigned to atomic volumes. The charges in each cell are summed according to their atomic basin assignments to get the total charge for each atom.

4. I could believe you that we have incorrectly referred to the "total charge density" when it should really be the "total charge".
adam2412
Posts: 13
Joined: Sat Jun 21, 2014 3:15 pm

Re: Min Distance

Post by adam2412 »

Dear Prof Graeme,

Thank you for your quick reply.

1. As I check again, I'm actually using total charge not total charge density. I have divided CHGCAR with volume.

2. I went through the output plane by plane (Z = 1, 2....until end of grid) using MATLAB and summed up the total charge inside the atomic basin. If I just sum up the total charge in each plane, the answer wont be the same with partial charge Bader. But, it will only be the same if I divide the total charge in each plane with the number of grid that I sum up.

What I did was, in each X-Y plane...I trace the charge gradient density and charge density maxima. After I get the atomic basin, I sum up the charge in each grid inside the basin. Let say the plane is at grid Z = 1, after I sum up, the total charge inside the basin at that one plane is already 568.329 (it is a ridiculous number). The number of grid inside the basin is 1156 points. So, I divided the total charge with number of grids. At each plane, I do the same thing. And then, after I finish with the basin, I sum up all that I have divided in each plane. The number in the end is not far from the partial charge obtained by Bader.

I'm not sure if my method is correct. But, the answer in the end seems to be legit. In the journal, "Grid-bsed algorithm for Bader Allocation", at point no. 7 page 903, there is a statement " assign dq/N to that Bader volume". I thought N would be the number of grids and dq would be the total charge. That's why I did what I did. But, he didn't mention why it is required to do that and what does it stands for. Could you please let me know why..

Thank you...
Post Reply