Just had a look at an old refinement. Shelx came out with a much lower wR2
(few percent) than GSAS (12/14 %). I think I used hklf 5 and shelx should
sum over peaks, not the contributing reflections as in Rietveld. Would
have to investigate further, but I notice shelx had a gof (or chi^2) of 1
and fiddles with the weights to make it so. Given the fit in question was
pretty good (for LaB6) it does seem surprising that R(F**2) from Rietveld
is suggesting that the intensities are on average wrong by more than 10%.
I would also have expected the Rietveld value to be too low, but it came
out the other way around. Maybe due to summing over more peaks. How about
your case Freidrich? Lots of people must have a silicon pattern around for
testing this out themselves...
Best,
Jon
On Tue, 17 Oct 2006, Leonid Solovyov wrote:
Dear Friedrich,
Normally, the Bragg R-factor is underestimated as it is calculated via
a single application of the Rietveld decomposition formula which
underestimates the difference between Icalc and Iobs for partly
overlapped peaks. When the Rietveld decomposition formula is applied
several times, as it is done in the Le Bail procedure, then the values
of Iobs become closer to the real observed intensities and the
difference between Icalc and Iobs may become much higher depending on
the peak overlap.
Best regards,
Leonid Solovyov
-----Original Message-----
From: Friedrich W. Karau [mailto:[EMAIL PROTECTED]
Sent: Tuesday, October 17, 2006 9:15 AM
To: rietveld_l@ill.fr
Subject: R-Value SHELX/GSAS
Dear colleagues,
may be, that I have comprehension problem. The Bragg R-value is named
R(F**2) in
GSAS and wR2 in SHELX and defined in the same manner.
when I extract the reflexion intesities using a LeBail fit and write
them into a
hklf6 formatted file, SHELX should refine the structure resulting
similar Bragg
R-values like GSAS. The fact is, the the R-values obtained by SHELX are
about 6
times bigger than the R-values obtained by GSAS. The structure is cubic
I, so
that no overlap of the reflections occurs apart from the multiplicity,
that is
incorporated in the hklf6 file. However the structure refines and
converges very
well in both cases. I have testet refinement with both SHELX and GSAS
for a
hexagonal structure, too, but the result is the same. The R-values
obtained by
SHELX are about 6 times bigger than the R-values obtained by GSAS. What
could be
the reason?
Thank you in advance
Friedrich
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++
+ Friedrich W. Karau email: [EMAIL PROTECTED]
+
+ AK-Schnick Tel.: (+49) (0)89 - 2180-77449
+
+ LMU - Muenchen mobil: 0179 6647667
+
+ Department Chemie u. Biochemie Fax: (+49) (0)89 - 2180-77440
+
+ Butenandtstrasse 5-13(D) URL:
www.cup.uni-muenchen.de/ac/schnick +
+ D-81377 Muenchen
+
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
++++
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com