From: "Radaelli, PG (Paolo) " [P.G.Radaelli@rl.ac.uk]
To: "'rietveld_l@ill.fr'" [rietveld_l@ill.fr]
Subject: RE: Combined XRD and ND refinements using GSAS
Date: Thu, 30 Nov 2000 09:55:50 -0000

Dear Caroline,

in general, it is very difficult to calibrate a neutron diffractometer well
enough to trust the absolute lattice parameters (the relative ones are
usually excellent).  This is especially true for TOF diffractometers.  There
are many reasons for this, including sample position and transparency
effects.  The latter are particularly annoying, because they are
wavelength-dependent, and introduce a non-linear relation between TOF and
d-spacing.  Also, the asymmetric pulse shape does not help.  En passant (are
you listening Bob?), I notice that there is currently no Rietveld code I'm
aware of capable of correcting neutron data for transparency.  Ideally, the
refineable parameter should be mu*R, exactly as for the absorption
correction, and it should be possible to set a constraint between the two
values.  Also, a correction for sample position errors would be nice.  Note
that, for modern scintillator-based diffractometers, a 0.5mm positioning
error (forward or backwards) is equivalent to a peak shift of 10% of the
FWHM at 90 degrees.
My usual advice to my users is that if they want to know the absolute
lattice parameters, they should measure them on their x-ray diffractometer.
Having done that, they should fix the lattice constants and refine the
neutron instrument parameter constants for all banks on the room temperature
data.  These constants should than be fixed for the other temperatures.
This advice should apply equally well to your case, especially if you have
synchrotron data, which are usually very precise.  

The situation is not that bad for constant-wavelength diffractometers,
especially of the multi-soller type (like D2B), provided that they are
recalibrated every time the monochromator is moved.  
You should also consider the possibility that the difference you observe is
due to a compositional gradient from the inside to the outside of the grain.

Paolo Radaelli
GEM instrument scientist

From: "Hyatt, Neil" [NHyatt@lsm.co.uk]
To: "'rietveld_l@ill.fr'" [rietveld_l@ill.fr]
Subject: RE: Combined XRD and ND refinements using GSAS
Date: Thu, 30 Nov 2000 11:13:38 -0000

Dear Caroline,

I had a similar problem to yours caused by highly absorbing samples
(particularly when using large diameter vanadium cans).  Effectively the
problem equates to a sample displacement error (I can give a suitably
hand-wavy explanation if you like!) which translates into a shift in
calculated unit-cell parameters.  I employed the same strategy as Paolo
Radaelli suggested... and refined the structure on the basis of "high
resolution" Cu-Ka1 Lab data.  I then fixed the cell constants in the neutron
refinement and refined on DIFC in GSAS which essentially serves to get the
tick marks in the right place.  I then fixed DIFC and allowed the unit-cell
constants to vary and proceeded with the refinement (I was doing high
pressure stuff and so kept the refined value of DIFC fixed for non ambient
pressures).  This should also work for your joint neutron-xrd refinent.  

All the best,

Neil Hyatt.

Date: Tue, 05 Dec 2000 17:13:57 -0700
To: rietveld_l@ill.fr
From: vondreele@lanl.gov (Bob Von Dreele)
Subject: Re: Combined XRD and ND refinements using GSAS

Dear Caroline (& others),
Our practice here at LANSCE with TOF powder diffraction Rietveld 
refinements & multiple detectors is to use the highest resolution detector 
bank as the "reference" and then refine the lattice parameters and the 
diffractometer constants for all the other banks in a multihistogram GSAS 
refinement. The reference detector bank is in backscattering and is thus 
insensitive to some (but not all) sample positioning  & absorption errors. 
This at least minimizes the systematic error. Usually only "DIFC" need be 
adjusted in these refinements. If you are also including an x-ray pattern 
with a well characterized wavelength (CuKa?) then use that for the 
reference & refine all the DIFC's for the neutron TOF data along with the 
lattice parameters. Don't forget to refine the usual systematic effect 
parameters for the x-ray data - most notably "shft"! Don't worry about 
instability - these things are pretty orthogonal to each other. I'd not 
refine DIFA or ZERO though, that will cause problems. The combined 
x-ray/neutron refinements do require that the experiments be done at the 
same temperature, otherwise the thermal expansion will mess up the combined 
refinement. This is real serious for noncubic materials. You probably don't 
have this problem as your c/a ratios were identical to about 5 parts in 
300,000. By the way one can use the same trick for multiwavelength 
experiments done at a synchrotron; fix one wavelength (the lab CuKa one is 
best for this) and then refine all the other wavelengths and the lattice 
parameters together.
Bob Von Dreele

At 09:31 AM 11/30/00 +0000, you wrote:
>Dear Rietvelders,
>
>I have a question concerning joint XRD and ND refinements using
>the prog. GSAS.
>
>The profile fit for the XRD part of the joint ref I am currently working
>on is quite bad and it seems to be due to a two theta error.  The unit
>cell parameters from the joint refinement are only very slightly
>different to those I obtained from a refinement of the XRD data only:
>                 JOINT   a=12.40855      c=3.93371
>                 XRDonly a=12.41279      c=3.93499
>I assume the values for the unit cell parameters obtained from the
>joint refinement are an average of those obtained from the XRD
>and ND datasets: is the weightin equal?
>
>Will such a small difference in the unit cell parameters be enough to
>cause the poor fit of the data?
>
>I calculated the two theta positions for both the unit cells and the first
>few reflections have differences of only, on average, 0.005.
>
>Any comments would be welcomed as well as suggestions on how
>to proceed.
>
>Thanks
>Caroline