Bader charge analysis

[bensg]

Dear All,

I use the bader charge analysis tool in Virtual Nanolab to analyze VASP calculations. Would you know what version of the bader software is implemented in VNL ?
When using the bader software directly (version 0.95 11/17/15, as technical issues prevented me from using the version 1.00 01/11/17), I obtain different charges than the ones given by VNL.
Also, I face an issue concerning periodic boundary conditions: for some calculations, the bader charges given by VNL are completely incoherent for some of the atoms on the edge of the simulation cell. For instance, with a hexagonal boron nitride single layer, although most of the atomic charges makes sense, on the edge of the cell I could have a nitrogen with a +5e charge and a boron with a -5e charge. Would you know a possible origin for this issue and a possible fix ?

Thank you very much,

Benoit

[Daniele Stradi]

Dear Benoit,

We do not use the Bader software by the Henkelman group, we have an original implementation. Could it be possible for you to send us or attach an small example of the calculations which gives you problems for debugging, including the output files? It should also include the results that you get with the original Bader code.

Regards,
Daniele.

[bensg]

Dear Daniele,

Thank you very much for your fast answer. By which mean would you like me to send you an example of calculations results (I have a 372 Mo zip file to provide you)?

Best wishes,

Benoit

[Daniele Stradi]

Dear Benoit,

I have replied you in a private message. We will investigate if this is an issue with the bader analysis tool for VASP implemented in ATK.

Regards,
Daniele.

[Kim_W]

How to know the result is resonable? I done it with FHI-aims.  Besides, how to get the Bader charges transferred from the Al atom to N atom with below txt scprit.

[Ulrik G. Vej-Hansen]

Please provide the script you have used to get the output you have posted.

[Daniele Stradi]

Dear Wang,

How to know the result is resonable?
The reliability of the result is based on chemical intuition. If there is an error, you would normally see an unphysically large charge on some atoms, which is not the case in your calculation.

Besides, how to get the Bader charges transferred from the Al atom to N atom with below txt scprit
To get the charge transferred to a certain atom, you should compare the charge calculated using Bader analysis with the atomic number of the corresponding isolated atom. In your case, for example, you can see that 9.4122 - 7 = 2.4122 electrons have been donated to N. However notice the two WARNING messages at the beginning of the log file. I would suggest you to increase the fineness of the grid to obtain more converged results.

Daniele

[Kim_W]

How to increase the fineness of the grid? Is that k-point or Accuracy or else? I calculated with FHI-aims calculator.  Thank you very much.

[Ulrik G. Vej-Hansen]

Again, please provide the script, it makes it much easier to help you. The text-file you have posted indicates some obvious mistakes, but we cannot help you with correcting them before you have provided the scripts.

[Kim_W]

 :)Thank you!

[Ulrik G. Vej-Hansen]

Please provide also the script which actually gave the output you posted previously. This script will not give the output you posted before.

[Kim_W]

Firstly, I calculated the ED with this script. Then, I dealed with "Bader Charge Analysis" tool.

[Ulrik G. Vej-Hansen]

Thanks for the update. I cannot reproduce the error with my version of VNL, which is 2017.1. Which version did you use?

I think the problem is due to a previous bug in our FHI-AIMS interface, which gave incorrect electron densities, and was fixed in 2017.1. However, it appears that the grid is much too coarse, so I suggest that you try increasing it quite a bit, if you want accurate results.

[Kim_W]

The question is: How to increase the fineness of the grid? Is that k-point or Accuracy or else? Thank you a lot.

[Ulrik G. Vej-Hansen]

When we talk about the grid, we usually mean the real-space grid which is used to represent the density. The relevant parameter is called density_mesh_cutoff. See more here: https://docs.quantumwise.com/manuals/Types/NumericalAccuracyParameters/NumericalAccuracyParameters.html