Re: [Ifeffit] FEFF6L.02 - missing path?
Matt, Thanks for the explanation. In this case, next two questions. 1. Is there a way to say it was rejected other than re-running feff with CRITERIA 0 0? I expected it to show in Feff.run file with "neglected" flag. My "default" Feff.run file says " Curved wave chi amplitude ratio 4.00% Discard feff.dat for paths with cw ratio < 2.67%" and the "neglected" flag is present for those with <2.67%. 2. Also there is still one path left (all paths No# 10 Se-Zn-Se) with cw ratio 2.83, which passes the <2.67 criterion but is weaker than the two paths that are rejected before it with amps 3.952 and 3.999. In other words - when using "default" generated feff.inp, which paths in Feff.run are reported as neglected and which are not reported at all? Best Pawel
Message: 1 Date: Thu, 28 Jul 2011 18:25:56 -0400 From: Pawe? Zajdel
To: Subject: [Ifeffit] FEFF6L.02 - missing path? Message-ID: <5EB6FA326A714E6BAD00215CBFB556AE@NewLapek> Content-Type: text/plain; charset="iso-8859-2"; Format="flowed" Dear All,
I am facing a strange issue while playing with EXAFS for a spinel type structure. I tried two typical examples: MgAl2O4 and ZnCr2Se4 with octahedral site K edge of Al or Cr. The MgAl2O4 is fine but for ZnCr2Se4, FEFF seems to omit one scattering path. It is not reported as being rejected by <2.6% amp criterion, so I wonder, while it is not shown. This is path 3 (as seen in MgAl2O4), which goes from Cr/Al to anion, then over the "octahedron" edge to another anion and back to Cr/Al. For oxide Reff=3.2251 <cut from feff.run> path cw ratio deg nleg reff 1 100.000 6.000 2 1.9371 2 37.913 6.000 2 2.8853 3 10.687 12.000 3 3.2251 <-- It's here 4 9.664 2.000 2 3.3444 </cut> For selenide Reff=4.2354 <cut from feff.run> path cw ratio deg nleg reff 1 100.000 6.000 2 2.5248 2 47.173 6.000 2 3.7084 <-- should be after this one. 3 25.382 6.000 2 4.3485 </cut> I am attaching two atoms.inp files, which may help. It is space grup 227, standard setting (2), atom list is correct. I am using Ifeffit 1.2.11c with Feff 6L.02. It s present under WinVista and Linux on two different machines. Am I missing something? I would appreciate any suggestions.
Best Pawel
Message: 2 Date: Thu, 28 Jul 2011 21:27:57 -0500 From: Matt Newville
To: XAFS Analysis using Ifeffit Subject: Re: [Ifeffit] FEFF6L.02 - missing path? Message-ID: Content-Type: text/plain; charset=UTF-8 Pawel,
For these two structures, the location of the anions (Se for ZnCr2Se4 and O for MgAl2O4) are not in in exactly the same location in the unit cell. The nearly identical double scattering paths: Al->O->O->Al and Cr->Se->Se->Cr each have small estimated scattering amplitudes ("Zabinsky Curved Wave Importance Factors"). The Cr->Se->Se->Cr path is filtered out with the default settings, and so not included in the path list, while the Al->O->O->Al path is not filtered out.
You can guarantee to see all the paths if you put CRITERIA 0 0
in the feff.inp file.
For a more detailed reason of why Al->O->O->Al has an estimated amplitude of 10.7 (and so not filtered out) while the Cr->Se->Se->Cr has an estimated amplitude of 3.9 (and so is filtered out), there are a few points to consider:
1. The amplitude reported is relative to the strongest scattering path, which always has amplitude 100. In both cases, this is the first shell Al->O->Al and Cr->Se->Cr. Se is a stronger back-scatterer than O.
2. Both absorber and scatterer are heavier in the Zn-Cr-Se case than the Mg-Al-O case, and the the heavier atoms are more point-like in their electron density.
3. The Se atoms are closer to 0.25, 0.25, 0.25 in the unit cell than the O atoms. This makes the Cr->Se->Se->Cr path have a scattering angle beta a little closer to 90 degrees, which gives less scattering amplitude.
I think all three of these conspire to give the Cr->Se->Se->Cr path less amplitude than the Al->O->O->Al path.
Hope that helps,
--Matt
-- dr Pawel Zajdel Division of Physics of Crystals Institute of Physics University of Silesia Uniwersytecka 4 40-007 Katowice Poland [48]323591978
Hi Pawel,
On Fri, Jul 29, 2011 at 7:13 AM, Pawel Zajdel
Matt,
Thanks for the explanation. In this case, next two questions.
1. Is there a way to say it was rejected other than re-running feff with CRITERIA 0 0? I expected it to show in Feff.run file with "neglected" flag. My "default" Feff.run file says " Curved wave chi amplitude ratio 4.00% Discard feff.dat for paths with cw ratio < 2.67%" and the "neglected" flag is present for those with <2.67%.
Well, you don't need to run all the Feff modules, but it would probably take more time to edit the input file than to run the program to select the modules.
2. Also there is still one path left (all paths No# 10 Se-Zn-Se) with cw ratio 2.83, which passes the <2.67 criterion but is weaker than the two paths that are rejected before it with amps 3.952 and 3.999.
In other words - when using "default" generated feff.inp, which paths in Feff.run are reported as neglected and which are not reported at all?
As you've probably read in the Feff documentation, Feff has a couple of filters that it uses: one fast and less accurate to eliminate really unimportant scattering paths to make the path-finding mechanism more efficient, and one slightly more accurate one to decide later which paths to calculate the EXAFS for. At this point, both of these optimizations seems less-than-critical, and just turning them off completely (CRITERIA 0 0) is not so bad, especially for moderately ordered systems like your crystal structure. --Matt
participants (2)
-
Matt Newville
-
Pawel Zajdel