Rietveld patterns in the PDF
X-ray X-press ( SNYDER@xray.alfred.edu )
Sat, 11 Nov 1995 9:58:56 GMT
Armel Le Bail brings up a complicated question that has been the subject of
extensive debate at the ICDD. The fundamental problem is as follows:
The primary use of the PDF is phase ID. If the PDF reference
pattern lists more lines as being present than a normal resolution
diffraction pattern would observe, then a user might reject the reference
as not matching his unknown. So the problem with a Rietveld refinement is
that it "knows" a lot more about the pattern than a "blind" second
derivative peak picker would "know". The intensities of lines in a,
Rietveld refined, unresolved cluster are not really "observed" - they are
more "inferred" from the model and consistent with the refinement.
Reporting them as a normal observed intensity is not only incorrect it
could lead to a failure to correctly identify the phase. Fundamentally the
ds and Is obtained from a Rietveld refinement are some kind of hybrid between
observed and calculated.
Another problem is that preferred orientation and other pattern
distortions have been modeled in a way that the refined pattern will not
even agree with a simple calculated pattern (from structure factors only).
So, some have argued to simply not include structurally refined
patterns in the file. However, this policy would not only lose important
patterns - leaving no possibility of a phase identification by a PDF user,
it would reject some of the best powder patterns flowing into the literature.
Others argue that the Rietveld pattern be run through a
conventional peak picker and the resolvable peaks be listed in the PDF.
This approach would mean that most Rietveld patterns would never make it to
the PDF because many authors would not care (or have the facilities) to
cooperate.
The solution adopted by the ICDD has been to include the Rietveld
pattern in the PDF with a special "R" quality mark on it, to warn users of
the PDF of its origin. This leave open questions as to how lines are to be
extracted from the observed pattern to be reported. Are lines too close
together to ever be observable on a diffractometer to be listed as separate
"observed" lines? Did the authors of the article allow for some kind of
resolution function in preparing their list? etc. In fact, the ICDD as
usual takes the authors data as they report it and prepare the PDF entry.
These solutions are not perfect but it was the best that the debate
in the ICDD technical committee could come up with. We would welcome other
ideas and I think Armel has started a particular good example of the use of
this mail server.
Bob Snyder
NYS College of Ceramics
Alfred University
Alfred NY 14802