null bader charges

[edward.deira]

dear all,

i'm new to this software, and i've some very weird results, i wonder if someone could throw some hints my way...

i optimized some molecules using nwchem, and generated a gaussian-format cube file using nwchem dplot function, and then computed bader charges from those cube files. major issue is:

nwchem dplot requires me to specifity limit x,y,z values and grid spacing values to generate the cube file. at first i just asked for 100x100x100 points between (-10,-10,-10) and (10,10,10) [all the molecules are within that region] and i got some nasty results, like carbon atoms [in organic compounds] with -4 partial harge [after subtracting the computed charge from the electrons of C].
then i choose much narrower regions in nwchem, just enough to include the molecule plus 1 angstrom on each side, and I got much more realistic charges BUT ACF.dat will then give me bader charges of 0.000 for some C atoms which lead to irrealistic partial charges.

note: the "bad" C atoms in the 1st case are "ok" in the 2nd case, and the "bad" C atoms in the 2nd case were "ok" in the 1st case.

in the former case i get around 139 electrons, in the latter around 108, if that matters

using the nwchem computed charges, i get reasonable partial charges - they tend to go a bit akward in some atoms, thus i went on to try new approaches to compute partial charges...

so, any hints on how these nasty partial charges arise ?

i understand different volumes lead to different charge distributions (well, it's the core of bader charge analysis), but how can that lead to such marked differences in the charge of a couple of specific atoms ?

any ideas on how to bypass this ?

thanks in advance,
all the best,
e.

[graeme]

My first guess is that your charge density file does not contain the core electrons. See this page about our solution in vasp: http://theory.cm.utexas.edu/vasp/bader/vasp.php

The basic idea is to have your code generate a total charge density (if possible) for the Bader partitioning. Then, integrate the valence charge according to this partitioning. The problem with analyzing the total charge directly is that you need a ridiculously fine grid to accurately integrate the cusp at the atomic cores. The problem with analyzing just the valence charge is that the charge density maximum at the atomic cores can be missing, which can give the zero Bader charges (which are defined as volumes around charge density maxima).