Problems in charge calculations

mane001 · Post by **mane001** » Thu Mar 22, 2007 3:48 pm

Hi all,

After reading in the forum that the newest version has a fix for non-orthogonal unit cells, I downloaded the source code from cvs and compiled it on our local computer.

I'm using the .cube files generated from a dacapo netCDF file (as described in https://wiki.fysik.dtu.dk/ase/Analysis#bader-analysis). I start with a simple (111) slab of a transition metal (the PSP for which treats 10 valence electrons). When I execute 'bader TMslab.nc' I get the following charges for the surface TM atoms --
# X Y Z CHARGE MIN DIST
----------------------------------------------------------------
1 0.0000 0.0000 0.0075 8.7264 0.0075
2 2.6625 4.6115 0.0075 20.0584 1.2420
3 5.3249 9.2230 0.0075 1.3035 0.0484

So, as you see here, although the total number of electrons on the surface atoms is almost retained, the charges for the three identical surface atoms are different from each other by a LARGE margin.

I found an earlier post on the forum, where somebody was having the same problem for 2 H atoms in their setup (although the difference in the charges there was much smaller <0.1 e). They seemed to have solved the problem by increasing the FFT grid, but it doesn't help in my case, unfortunately.

Any clues? Am I doing something grossly wrong/overlooking some important aspect?

Thank you in advance,
Mane

Wenjie · Post by **Wenjie** » Thu Mar 22, 2007 7:13 pm

Can you send the cube file to me (twj916@mail.utexas.edu)? We'll see if something is going wrong.

mane001 · Post by **mane001** » Sat Mar 24, 2007 12:00 am

Hi,

Thanks to Wenjie for pointing out the mistake in my cube file. I was creating a cube file from a set of atoms, some of which were out of the box -- and Dacapo doesn't really compain about it since the pbcs take care of this issue - and was getting weird charge numbers.

Thanks again,
Mane.