[Ifeffit] Re: ? about l_scmt etc.

[Bruce Ravel]

> For EXAFS and EXAFS fitting, the FMS and XANES calculations should
> not be done.  For EXAFS, these values don't matter (I think!).
> For what it's worth, L_fms basically sets how high in k the FMS
> approach works well enough to use. It's typically a number <~ 5.
> I'm not sure what l_scmt is...

I agree that the FMS and XANES calculations need not be done if you are
interested in doing EXAFS analysis.

>> (i.e. in the past I have used the following for 2 Au's in the surface)
>> POTENTIALS
>> *     ipot     z     element     l_scmt     l_fms     stoichiometry
>>        0        79    Au            3             3           0.001
>>        1        28    Ni             2             2           4
>>        2        79    Au2           3            3           4
>
> Oh, I think those might be
>        ipot     z    element     lmin   lmax
>
> and that lmin=-1, lmax=3 are reasonable default values.  Hopefully
> someone can clarify this....

That's not correct.  "scmt" = "self consistent muffin tin" and "fms" =
"full multiple scattering".  Elsewhere in the feff-iverse, scmt is
referred to as "scf"  or self-consistent field.

The l_scmt value is used in the calculation of the self-consistent
potentials.  The l_fms value is used in the calculation of the final
densities of state and xanes.  Each one limits the size of the angular
momentum potion of the basis used in the matrix algebra required to do the
feff8 calculation.  The reason there are two is to allow the possibility
of considering different sized bases while determining the potentials than
while computing the final xanes spectrum.  There are some reasons why one
might want that freedom.

In practice, one can usually use the same values for l_scmt and l_fms. 
For first row elements, l=1 is often fine.  If one has the time and
computer power, l=2 might result in a better calculation.  For transition
metals, l=2 is a good choice.  For heavier things, l=3 should be used. 
Heavier elements are likely to have f-electrons in the valence band or
low-lying f-states, thus a value of 3 is necessary.

An l-value of -1 tells feff to use its default for that element.  The
defaults are more-or-less as I explained in the last paragraph.  Thus, the
easiest way to use feff8 is to use l=-1 for every atom and for both scmt
and fms.

B