this discussion seems to be going into the basics, i.e. I learned by now,
that Lachlan as well as Linda register the importance of a profund
cristallographic education. As I talk from a chemist's point of view, I
combine it with my structural chemistry knowledge and almost soon have an
idea, what phase could be my impurity (while there are still samples, I did
not make it at all to identify any of the present phases).
But - things are rather easy for me, because it's me who mixes the initial
compounds, grinds them and performs the reaction. It's much more
complicated, I assume, to have a mineralogical sample and perform a phase
analysis.
As I stated yesterday, I think, computers are too often thought to be
"perfect" (as Armel showed, the databases are not at all perfect - they
cover a little fraction of the work that was really done). An example might
clearify what I want to say:
During a discussion with a colleague working on GaN (he tries to make it
from Li3N and Ga) the guy told me he had performed a Hanawalt search on the
JCPDS data base for impurity-search. The impurity phase he had in his
spectra matched pretty well with a file of Li2O. Additionally, this was the
only file matching fine, so he really rushed when I told him,that there was
no place in whole Europe (except perhaps the polar circle in wintertime, or
the spanish Nevada) where I would expect this phase to be stable in air
(because it is very hygroscopic and will almost immediately decompose to
LiOH with humidity).
His (computer-induced) belief was really that he had this phase on the
sample holder! He did not even have the idea to check the computer's
statement by looking into an inorganic chemistry book.
---
I don't believe it will be possible, to have a 100% matching expert systemfor phase analysis. Not even with the Rietveld method applied to such asystem (allthough this would probably improve the actual system deeply).There will sure be some cases left, where you have to think seriously andperhaps perform a different (i.e. non-crystallographic) experiment todecide.So, you will have to add knowledge of another kind than scattering angle,intensity data and all the rest to a further improved expert system forphase identification.
Carsten
Carsten Schinzer ---------------------------------------
Institut de Chimie de la Matiere Condensee de Bordeaux 87, Avenue du Docteur A. Schweitzer F-33608 PESSAC fon +33 556 84 2650 (poste 7397) fax +33 556 84 2640--------------------------------------------------------