> Basically, you therefore mean that obtaining an electron density
>map (that is of course if we are using X-ray diffraction) is the "ab
>initio" part? What are the procedures to do this for powder data?
Have a look inside the tutorial :
http://fluo.univ-lemans.fr:8001/iniref/tutorial/indexa.html
or read some review papers you will find a list of 32 (!) at :
http://fluo.univ-lemans.fr:8001/iniref/revpap.html
Note that this is again an indication that ab initio is a quite hot
topic : 32 review papers for ~320 solved experimental cases,
90% in the last 10 years !
> I'm also interested if anyone has any expertise in this to
>discuss determining disordered or amorphous structures (e.g. of
>glasses) from powder patterns. It seems that radial distribution
>functions would just allow one to confirm for example the SiO2
>tetrahedral coordination for pure silicate glasses. Anyone that can
>discuss Monte Carlo methods and reverse Monte Carlo methods...I would
>be very appreciative.
Amorphous structures are essentially "undetermined" in my opinion.
Small mean models have been shown to be able to reproduce the powder
patterns by a "RDM" method (Rietveld for Disordered Materials).
Nevertheless, such applications have not found a public as large as RMC
(Reverse Monte Carlo).
Most RDM applications are authored by me :-( with less than 10
papers (conventional Rietveld users are generally skeptics on this
minor deviation of their method) ; RMC applications are now >100
so that R. McGreevy should consider the creation of a RMC mailing
list which would have really no common point with the Rietveld one...
Mmmm, nevertheless I think I will publish soon a study showing that
RMC may start from good RDM models, enlarged by x3 or x4 in all
space directions and provide same "good" fits after small atomic
displacements, keeping strict coordination constraints (this applies
to glasses based on well defined units as SiO4, FeF6... of course). This
means that RDM and RMC may provide essentially the same result
(and let me tell you that RDM is really faster than RMC : 20 mn
on a Pentium II 266MHz and you know if one model is wrong).
Depending on constraints in your model and of its size, you will
need days or months with RMC. Quite not ab initio : you should
know composition and density, at least, more will help.
Have a look at the RMC 1997 Internet Conference site :
http://www.studsvik.uu.se/rmc/confa1.htm
and best wishes,
Armel Le Bail - Universite du Maine - Laboratoire des Fluorures
CNRS UPRESA 6010 - Avenue O. Messiaen - 72085 Le Mans cedex 9 - France
Web : http://fluo.univ-lemans.fr:8001/