Hi all,
here are my suggestions/experience from working with BGMN:
On Thu, 16 Sep 1999, L. Cranswick wrote:
>
>
> At the ILL Powder meeting earlier this year
> Juan Rodriguez-Carvajal was asked when/if Spherical
> Harmonics might be put into Fullprof.
> Juan's reply was more of the nature you should
> try and do the best sample preparation possible
> and get the best diffraction data possible rather
> than try and model/fudge corrections into the data.
>
> Seeing Spherical Harmonics in action in GSAS and Topas,
> it is easy to see why people using Spherical Harmonics
> must love this correction as it can make what looks
> like quite a bad fit become exceptionally good at the
> flick of a switch. Thus giving the impression that
> something this good must be evil!?
>
> Given Spherical Harmonics can have this affect at
> making the fit look very nice, are there recommendations
> on how to review results that have included
> Spherical Harmonics corrections for preferred
> orientation?
>
> i.e.,
>
> What is considered appropriate and worrying in terms
> of the "order" used and values of the GSAS Texture
> Index when using Spherical Harmonics?
> Should people be providing before Spherical Harmonics and
> after Spherical Harmonics goodness of fit plots for publication
> and review? (possible the same should go for any use of
> preferred orientation corrections?)
>
Using Spherical Harmonics (of higher order) in structure
refinement is, at our experience, mostly nonsens. Probably
caused by the high number of structural parameters per phase.
Therefore, the parameters of the Spherical Harmonics correlate
with the structural parameters very often. As I feel,
this is the field of most the Rietveld publications, up do now.
Otherwise, using Rietveld refinement in a routine lab for
quant analysis, the number of parameters per phase is very small
(usually all atomic co-ordinates and Debye-Waller-factors remain
constant). In this case, Spherical Harmonics are very useful.
You must not trouble with difficult sample preparations. So you
get a large number of different sample preparations per time.
Naturally, mostly of the results of such routine labs will never be
published!
For testing for correlations, computation of the parameter's e.s.d.'s
would help somewhat. In quant analysis, you have to find the balance
between low (or zero) order of Spherical Harmonics (high systematic
error, bad difference plot, but low e.s.d.'s) and high order (low
systematic error, good difference plot, but high e.s.d.'s).
J"org Bergmann
[EMAIL PROTECTED]
