Hi Matt and Bruce, Prof. Matsuura, Matt is right, configurational average in feff8 is not reliable. However if you use as Matt suggest, and average even by looking you can see that results as quite OK. For example in one case Tl-1234, there are two Cu sites, and when one generates the XANES with both Cu1 and Cu2 as core, the average of these agree with experimental result...... For Bruce, just a little request, I prefer that Tkatoms is distributed as a separate package, at least there should be a choice so that if one wants to use only Tkatoms one can.... thanks sher

Dear Prof Matsuura,

Feff8 does claim to allow configurational averages for EXAFS. Unfortunately, the results with CFAVERAGE are clearly wrong even in simple test cases. At the most basic level, in order to describe a configurational averaged environment, fractional degeneracies would be necessary. The feffnnn.dat files *never* report fractional degeneracy, even when using CFAVERAGE. This should immediately make you very suspicious that something is not right.

I recommend not using the CFAVERAGE card, and instead making separate feff runs for each candidate central atom, and then averaging these sets of calculations, weighted by the fraction for each candidate central atom. It's unfortunate that the tools for modeling EXAFS of disordered crystals are so crude... there's probably a lot of room for improvement in this area!

Hope that helps,

--Matt

On Tue, 16 Mar 2004, Makoto Matsuura wrote:

...

My question is how the path files (feffxxx.dat) are defined when "configurational average" is used in feff calculation. And how we apply path data avaraged configurationally to the fitting calculation. For example in case of CaCu5 structure two Cu sites are there, i.e. 3g and 2c site. If we use control card of "CFAVERAGE" in feff calculation with setting Cu 3g as a core. Then configurational average will be done over the 2c sites. The below shows the two log files of feff8 ( list.log ) in the case of the configurationally averaged (core Cu 3g site) and without configurationally averaged. I calculate Fourier transform (RSF) using each path (feffxxxx.dat) deduced by feff calculation applying configurationally averaged and compared with that of the total one. I fail to coincide sum of RSF's for each path with the total RSF. I like to know the reason of it.

Next question is how we use the paths (feffxxxx.dat) in the feffit calculation when we use "configurational average" in feff calculation. Can we use the feffxxxx.dat files as it is for the feffit?

list of list.log when configurationally average card is used. //////////////////////////////////////////////////////////////////////////// /////////////////////////////////// Cu K-edge for Cu 3g of CaCu5:output for feff8 absorber Feff 8.10 Cu configurational avaraged with Cu2c site DEBYE temp =300K at 20K Abs Z=28 Rmt= 1.347 Rnm= 1.394 K shell Pot 1 Z=28 Rmt= 1.354 Rnm= 1.403 Pot 2 Z=57 Rmt= 1.773 Rnm= 1.830 Gam_ch=1.576E+00 H-L exch Mu=-1.168E+01 kf=1.918E+00 Vint=-2.026E+01 Rs_int= 1.891 ----------------------------------------------------------------------- pathindex sig2 amp ratio deg nlegs r effective 1 0.00000 100.000 2.000 2 2.4634 2 0.00000 93.844 2.000 2 3.2025 3 0.00000 16.979 1.000 2 3.9873 4 0.00000 13.047 8.000 3 4.2800 5 0.00000 55.603 4.000 2 4.3163 6 0.00000 44.329 4.000 2 4.7095 7 0.00000 36.814 2.000 2 4.7769 8 0.00000 19.574 8.000 3 4.8182 9 0.00000 8.950 8.000 3 4.8369 10 0.00000 4.441 2.000 4 4.9269 11 0.00000 3.423 4.000 4 4.9268 12 0.00000 8.603 4.000 3 5.0671 13 0.00000 13.297 8.000 3 5.2064 14 0.00000 11.402 8.000 3 5.5573 15 0.00000 27.903 4.000 2 5.5851 16 0.00000 3.001 8.000 4 5.6659 17 0.00000 31.325 8.000 3 5.8408 18 0.00000 23.977 4.000 2 5.8943 19 0.00000 8.327 4.000 4 6.0965 20 0.00000 25.802 8.000 3 6.1496

list of list.log without using configurational average //////////////////////////////////////////////////////////////////////////// ////////// Cu K-edge for Cu 3g of CaCu5:output for feff8 absorber Feff 8.10 Cu No configurational avaraged DEBYE temp =180K at 20K Abs Z=28 Rmt= 1.347 Rnm= 1.394 K shell Pot 1 Z=28 Rmt= 1.354 Rnm= 1.403 Pot 2 Z=57 Rmt= 1.773 Rnm= 1.830 Gam_ch=1.576E+00 H-L exch Mu=-1.168E+01 kf=1.918E+00 Vint=-2.026E+01 Rs_int= 1.891 ----------------------------------------------------------------------- pathindex sig2 amp ratio deg nlegs r effective 1 0.00000 100.000 4.000 2 2.4634 2 0.00000 96.591 4.000 2 2.5063 3 0.00000 93.844 4.000 2 3.2025 4 0.00000 14.579 16.000 3 3.7165 5 0.00000 16.979 2.000 2 3.9873 6 0.00000 13.047 16.000 3 4.2800 7 0.00000 55.603 8.000 2 4.3163 8 0.00000 27.399 4.000 2 4.3410 9 0.00000 8.758 16.000 3 4.4556 10 0.00000 12.606 16.000 3 4.6430 11 0.00000 12.963 16.000 3 4.6430 12 0.00000 6.132 8.000 3 4.6767 13 0.00000 44.329 8.000 2 4.7095 14 0.00000 36.814 4.000 2 4.7769 15 0.00000 19.574 16.000 3 4.8182 16 0.00000 8.950 16.000 3 4.8369 17 0.00000 10.192 4.000 3 4.9269 18 0.00000 9.147 4.000 4 4.9269 19 0.00000 4.441 4.000 4 4.9269 20 0.00000 3.423 8.000 4 4.9268 ******************************************

Miyagi National College of Technology

Prof. Makoto Matsuura

email: matsuura@miyagi-ct.ac.jp TEL&FAX 022-381-0270

******************************************

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit

_______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit