[Ifeffit] switching to FEFF8 in Artemis
mamarcus at lbl.gov
Tue Mar 11 11:47:54 CDT 2008
> Is "QFS" the quick fist shell theory thingie in the Theory menu? You
> are correct -- it does not give you a choice. Using feff6 is
> hardwired into it. It seems silly to me to use feff8 for the quick
> first shell thing, but for the sake of completeness, it should be
Yes, it's the quick first shell thingie. The problem I have with it defaulting to feff6 is that if the rest of the system is set to
use feff8, it will use feff8 as well, but with the *feff6* template, which means that feffNNNN.dat files don't get written. If the
QFS system either used feff6, regardless of what version the rest used, or used the same template as the rest, then all would be
> Your other option is to run feff8 outside of artemis and import the
> results. When you "Open file", you can point the file browser at a
> feff.inp file. That will load the feff.inp file, notice that there
> are feffNNNN.dat files, and ask how many you want to import.
> That's the work-around. I would like to understand your problem a bit
> better so I can improve it.
It's not a specific problem, but wanting to use the best available calculation. A prominent user on my beamline claims that feff8
is much better for MS, but he hasn't told me what his evidence is.
>> Also, I found that when simulating Pu3Ga (a simple Cu3Au structure) out to
>> 10A, feff8 chucked up an error about too many paths, while feff6 did it
>> happily. Is there a
>> way to fix that?
> One of these might help:
Sounds like PCRITERIA may be the thing. Of course, that means editing the FEFF file or the template. Is it possible to edit the
FEFF file from within Artemis?
>> How much difference does using feff8 make in terms of proper description of
>> MS paths for k>2, anyway? Do I really win by go ing through the trouble?
> Were the question about k>4, I would say with confidence that you do
> not win. For k>2, I suspect the answer is "not much". For a highly
> asymmetric material, feff8 should significantly improve the
> description of the potential surface by virtue of moving charge around
> self-consistently. But the potentials are still spherical, so there
> really aren't bonds in feff8. As you move away from the potential
> surface by increasing energy, you become increasingly insensitive to
> those details.
So highly-MS-specific things like focusing paths aren't done much better in feff8 than feff6 in the EXAFS range? That's good to
> Is k=2 "far enough" from the potential surface? I don't know. There
> are a lot of issues down there. k=2 is a region where the details of
> the background removal are highly correlated with the fit results. Do
> you need a better feff or a better background removal? It seems like
> a tough question to answer generally.
Below k=2., you're solidly in the XANES region, where I know that feff8 doesn't work unless you play with parameters so as to make
its output look like the experiment. So far, I have not found a XANES calculator which will give, without adjustment, a
realistic-looking XANES curve for even as simple a substance as ZnO, but that's another story.
More information about the Ifeffit