It is not possible to exclude a single reflection in Rietveld since one
minimizes against the observed pattern not an Fobs table. It is possible
to exclude a region containing one or more peaks from the pattern, but
you would have to specify each region manually. Also, in the Rietveld
method, sigma(Fobs) cannot be directly determined -- except for well
separated reflections -- so (Fobs-Fcal)/sigma(Fobs) is not available.
Having said this, why would you want to do what you ask? In powder
refinements, if you have reflections that compute poorly, your model is
probably invalid, either due to an inaccurate structure or mistreatment
of an experimental artifact, such as preferred orientation. In single
crystal work, when strong reflections compute badly, one invokes the
argument of extinction (though the former conclusion may well apply) to
justify exclusion of a few very strong reflections. Personally, I have
seen a few cases where extinction can been seen in powders, but they are
rare and the experimenters worked hard to get powders that had such
large relatively perfect crystals. Extra care needs to be taken with
such a sample to be sure that enough crystallites are in the beam to be
sure that the measured intensities obey a powder average. Usually, doing
a bit more grinding is the easy way out.
Brian
Colin Lobban wrote:
>
> Are there any Rietveld programs with the capabilty to reject reflections
> for which (Fobs - Fcal)/sigma(Fobs) is greater than a pre-stated value? I
> can't find any such feature in GSAS nor Fullprof.
>
> Thanks in advance,
>
> Colin Lobban.
