[Ifeffit] Structures with multiple crystallographic sites

[Matt Newville]

Hi Vadim,

> Thanks, Matt! Yes, that's exactly what I tried doing. It's a simple cubic lattice with
> 20 atoms per unit cell, 8 of one type (Co1) and 12 of the other (Co2). So I did two
> FEFF runs in the same Artemis project, with Co1 and Co2 as the core atom, and
> multiplied their respective amplitudes by 0.4 and 0.6. Then I noticed that if I calculate
> a sum of all included paths I get a different FT spectrum than if the sums are calculated
> separately for the two FEFF runs and averaged (with correct weighting) outside of
> Artemis. Of course, it is a bit late here and I could be missing some obvious mistake.
> On the other hand, I am trying for 1.6-5.5 angstrom fitting range, which involves 70+
> paths for Co1 core and ~120 for Co2. Is it possible that Artemis does not like to deal
> with too many paths and/or variables? I better go check the manual... :o)
> Regards.

The "if the sums are calculated .... outside of Artemis" might be the issues:

When the sums are done inside Artemis, the set of Path Parameters is
applied to each path during the sum  That is, E0, delR, S02*N, and
sigma2 will be applied.    In addition, Artemis probably has helped
you pick a subset of the paths to include in the sum.

Depending on how you do the sums outside of Artemis, the Path
Parameters may or may not be applied, and the set of paths used might
be different.  If, for example, you're using the chi.dat file from a
Feff calculation done by Artemis, then it almost certainly has S02*N=1
and sigma2=0, delR=0, and E0=0 for all paths.

Hope that helps,

--Matt
PS:  There is a limit on the number of paths you can use (1024 paths),
but you're not hitting it.