As Maxim says, strain-related parameters are U and Y. By the same token,
size-related parameters are P and X (both go with 1/cos(Theta). The reason is
that this Rietveld model (the so-called TCH) assumes a Voigt function for both
size and strain profile and therefore has to have both Lorentzian and Gaussian
components.
When refining the parameters, one has to keep in mind that instrumental
broadening in general depends on the same parameters. Therefore, first refine
all profile parameters on the diffraction pattern of the standard sample (as
Brian said, LaB6 or similar) and then refine ONLY four parameters (U,P,X,Y) for
the sample of interest, as far as instrumental conditions stay the same.
More details and how to calculate size and strain from U,P,X, and Y were
published in JAC 37 (2004) 911-924. The reprint and original data from seven
different instruments (including synchrotron and neutron) can be found at
http://mysite.du.edu/~balzar/s-s_rr.htm.
Davor BalzarUniversity of Denverwww.du.eduwww.du.edu/~balzar
- Original Message -
From: Maxim V. Lobanov [EMAIL PROTECTED]
Date: Sunday, November 30, 2008 10:37 pm
Subject: RE: Rietveld: U,V,W
To: rietveld_l@ill.fr
Dear colleagues,
one question on that:
U and W should be instrumental constants that will not change with sample,
while V can have both an instrumental and a residual stress component.
as far as I understand, the strain broadening term should have FWMM~theta
dependence, i.e. Lorentzian Y in standard (GSAS) notation.
For the Gaussian part you have
FWHM^2=U*tan^2(th)+V*tan(th)+W+P/cos^2(th)
and the term that varies roughly as ~theta is U, not V.
Therefore, based on that and probably erroneously, I used to refine U
parameter rather than V (+ P, X, Y, of course).
I am wondering if somebody could make this more clear?
FullProf might be a bit more stable, but I think the process there
is about
the same. (BTW, if anyone works out how to convert GSAS profile
terms to
ones used in FullProf, I'd be interested to get those relationships into
CMPR; I am not sure if the scaling is only between centidegrees**2 and
degrees**2.)
So far I noticed that Fullprof uses interchanged (v.r.t. GSAS) XY,
i.e. X
corresponds to strain term there. There might be also other
differences, of
course...
Sincerely,
Maxim.
From: Brian H. Toby [mailto:[EMAIL PROTECTED]
Sent: Monday, December 01, 2008 5:21 AM
To: May, Frank
Cc: rietveld_l@ill.fr
Subject: Re: Rietveld: U,V,W
What is the correct procedure for refining U,V,W? It is my understanding
that those parameters are a function of instrument geometry. Does
one use a
standard material to determine U,V,W and then fix their values for the
instrument you're using?or do the values of U,V,W change
depending on
the sample being examined? If so, why do the values change?
The GSAS manual covers the latter part of you questions pretty well,
though
perhaps indirectly. In theory, U and W should be instrumental
constants that
will not change with sample, while V can have both an instrumental
and a
residual stress component. However, this assumes that one also
refines a
crystallite size parameter, P when needed, which many people (myself
included) do not. In that case, U, V W will all change to
compensate for
crystallite broadening.
I do recommend using a standard with good sharp peaks (SRM LaB6 is the
ideal, though there are likely to be many other oxides handy that work
reasonably.) If you can't get a good fit to your standard, then you
do not
want to advance to an unknown until you understand the problems with
your
instrument/technique.
Where possible, I try to start a refinement with values that are
close to
correct for U, V W (+ X Y where significant) and put off refining
them
until late in the refinement, when they tend to be pretty stabile.
Initially, I usually refine U, V W together and then refine X and
then Y
solo and then finally in combinations until everything is refined together.
Look for parameters that are refining to zero and turn them off,
since GSAS
does not deal with that very well. Also look at the widths vs 2theta
(widplt) to see if the functions are reasonable.
The routine in CMPR for fitting U, V W values to a set of peak
widths has
been useful for me where I don't have good calibration information
for an
instrument.
FullProf might be a bit more stable, but I think the process there is
about
the same. (BTW, if anyone works out how to convert GSAS profile terms
to
ones used in FullProf, I'd be interested to get those relationships into
CMPR; I am not sure if the scaling is only between centidegrees**2 and