Hi everyone, apologies for whats going to be the length of this email.I have a couple of fairly basic questions concerning the calculation of XANES of fairly complex crystalline materials using antiferromagnetic parameters. As an example I'll use the normal spinel magnesium chromite.
To carry out this calculation I use the p1=true card in the atoms.inp file to generate the 54 atom primitive unit cell positions in the file p1.inp. There I change the labels to designate my spin-up and spin-down chromium atoms, save the p1 file as the new atoms.inp and run atoms again. I get the same feff.inp file as before, but this time I can add in a new (spin-down) potential, and change the potentials in the atomic list using a global edit. I then recompile the source code to ensure I have the correct relative spin arrangement of my potentials, and run the calculation using the SPIN card. Any comments on this methodology are more than welcome.
I've included a sample feff.inp (you'll see I like to keep it simple) and a document containing the XANES result against experiment.
I hope its fairly clear that the AFM calculation shows a better agreement with experiment than the non-spin polarised calculation, but the peaks in the first 5eV appear to be compressed closer together than their counterparts in the experimental data. A little bit of broadening would help fit the data, but the narrow white line at the edge onset will disappear, so we're restricted a bit in what we can do. I have noticed a similar effect in other calculations I've done on ferrite spinels and perovskites.
Is this 'compression' of the first few peaks something people have seen before? If so, what was the origin? How was the problem resolved? What would tend to be the reason that happens?
Any advice gratefully received, or indeed, if someone is of the opinion that the fit is as good as we're gonna get with the current FEFF incarnation, please let me know. User errors should almost certainly be taken into account!!
My second query (more of a point, really) is to do with LDOS calculations and Fermi level positions - when I carry out AFM calculations, I see a good Stoner separation of the spin-up and spin-down DOS - but the Fermi level is still seen to cut through a band when I would expect it to reside at a minimum in the DOS, indicating that we may be at the lowest energy case for our material.
I can get round the small Fermi level inaccuracy by integrating the DOS plot to get the electron occupation below the Fermi level - this gives me the number of valence electrons in my material. If it is higher or lower than what it should be, then I can simply shift the Fermi level to the point where the electron occupation is correct and - voila! - I can get the Fermi level sitting in a nice minimum (certainly for the materials I'm looking at). It works out quite elegantly, and I was wondering if this was established practice, or if I'm doing something illegal (!). If there are any problems with this, please let me know. Otherwise leave me with my elegant solution!
I should mention that I'm integrating the Total DOS for the material obtained by summing the DOS from each unique atom potential, and since FEFF gives LDOS per atom, the integration gives the total number of valence electrons, e.g ZnCr2O4 = 12e for Zn + 6e for Cr + 6e for O = 24e. Obvious an AFM calculation involves electron transfer between spin-up and spin-down states, but you can add both states and shoot for, say 48e (since you're talking about two atoms (one spin-up, one spin-down)). I hope thats clear. Any problems, please let me know.
Apologies once again for the length of this email and the waffly tone (thats the chemist in me), especially on a Monday when brains are running slow!
Regards, David Eustace Postgraduate researcher Dept. of Physics and Astronomy University of Glasgow Glasgow, Scotland G12 8QQ
Attachment:
Sample feff.inp.doc
Description: MS-Word document
Attachment:
MgCr2O4 edge.doc
Description: MS-Word document