Dear respected RG members,
I am doing Mulliken population analysis of Cobalt (Co) doped MgO lattice (rocksalt) using CASTEP code (ultrasoft pseudopotential). I found some strange data as follow:
for pure MgO
bond population (Mg-O) = - 0.73
for (3%) Doped MgO
bond population (Mg-O) = - 0.31
bond population (Co-O) = 0.42
Is the negative value of of Mg-O reliable?, since negative value indicate antibonding, small value close to zero indicate ionic bonding and large positive value (Co-O) indicate covalent bonding. As we know, MgO has a very strong ionic bond, therefore bond population (Mg-O) must be close to zero. All parameters that I used during optimization are correct. I also did the same calculation using ( norm-conserving ). The obtained data gave almost a good result as suggest by Segall et al. (Mg-O = 0.58) and (Co-O = 0.53).
Thanks in Advance
I wouldn't put too much faith in Mulliken analyses, they are quick and useful sometimes for qualitative trends but they are extremely dependent on the atomic states you project onto. In CASTEP the projections are done using the pseudo-atomic states, so you can always try changing the pseudopotential and seeing whether the numbers are robust.
All the best,
Phil Hasnip
(CASTEP developer)
I do appreciate your kind response Dr. Hasnip. I do agree with you that the results of Mulliken analysis are not very reliable for such crystal systems but may give a good results for others as I read in literature. At least, this method may give a rough estimation of the studied compounds.
Warm Regards.
Yes, rough estimates are just what Mulliken is good for. For more quantitative work, I find Hirshfeld charges quite useful and these are also fairly quick to compute. There's also Bader's atoms-in-molecules scheme. CASTEP has built-in subroutines to do Hirshfeld analysis; for Bader you need to use the post-processing tools from the Henkelman Group:
http://theory.cm.utexas.edu/bader/
There are a few ways to do it, but probably the simplest is just to write out a formatted charge density and convert it to a CHGCAR file (using den2vasp). Please note that this isn't strictly what Bader intended, because it only uses the valence charge density not the total electron density, but this is what everyone seems to do in the pseudopotential DFT community, and so far the results seem OK.
The Populations analyses and their different corresponding bonds types.
best regards sincerely,
BENGHIA Ali
PhD in Physic and Chemistry of Materials,
Dear Dr. Ali
I do appreciate the valuable information that you share here. You always help me Sir. I wish you good look in your future endeavors.
Regards
Mohammed
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science