UT theoretical chemistry code forum

Greetings,
When you suggest that the total charge should be correct when running a bader analysis on a system, do you mean the total electrons or the charge assocaited with a vacuum level?
Thanks

The total number of electrons; I don't know what charge would be associated with a vacuum level.

Graeme, (others),
Thanks...got it. I am looking at 1:1 Tantalum Nitride and I have been increasing the NGX,Y,Z to converge the total electrons in the system to the correct value. My cell has 3 Ta and 3 N which gives it about 240 electrons. At about NGX,Y,Z equal to 75, 75, and 45 respecitvely (as it is a hexagonal cell), I hit a lower limit to the electrons converging of about 250 electrons in teh system. Further increasing of the mesh, I get an increase in the total number of electrons and at about NGX, Y, Z, of 90, 90, 60, respectively, I get about 320 electrons in the system.

I am wondering if this is a result of the complexity of the Tantalum or just a result of the grid not being properly aligned to the system? Ultimately, I am trying to determine the ionic radius of the elements in the system. Could I use the lower limit calculation of 250 electrons even though it is 10 electrons shy of the real system to get the approximate ionic radius? Any thoughts on converging the electron number further?

Are you analyzing only the valence electrons? If so, the total number should not change significantly with grid density. If, on the other hand, you are integrating the total charge (AECCAR0+AECCAR2), it will be very hard to use a fine enough grid to resolve the charge density peaks at the atomic centers, and get the correct total number of electrons. I do not recommend trying this; since you probably only care about the valence charge, integrating over the CHGCAR makes more sense.

I must have been off on how I interpreted the need to include the valance and core electrons in the calculations. I was up to 400^3 grid size and still missing 2 electrons. I was beginning to get suspicious that there might be an issue with the convergence algo or that there was something amiss with my system. You may want to augment then the statement:

"One finally note is that you need a fine fft grid to accurately reproduce the correct total core charge. It is essential to do a few calculations, increasing NG(X,Y,Z)F until the total charge is correct. "

As for the valance - Yes, I think I understand that and I can easily rerun w/ just the CHGCAR. I am wondering if there is any rules to getting a "good" number as I recall that you should get w/in 0.1 electrons of the system.

Good point, that statement is a little misleading. The recommendation should be to analyze the total charge, but do the integration over the valence charge, as in:

bader CHGCAR -ref CHGCAR_sum

checking for convergence of the bader charges with increasing NG(X,Y,Z)F. For any large system, it is basically impossible to make a grid fine enough to accurately converge the total (valence+core) charge.

Yeah...I was starting to think there was something wrong w/ how I was using the Bader.exe or with the VASP calucations. I was up to 400^3 for one system and about the same for 2 other systems, but still having 2 extra electrons.

Appriciate the rapid response.