Xiaolong Chen wrote:
>
>I am a rietveld user of DBW3.2S-9005. I found that spontaneous release
>of zero point, sample displacement and lattice constants often leads to
>unreasonable results of these three parameters in refinements.
>Does anyone know how to deal with this problem ? Thanks.
>
The "problem" is quite simple. I don't know how it is for other RR programs
(must be the same- it's pure geometry) but for DBWS the 0-point and sample
displacement are strongly coupled i.e the effects of the two on the peak
positions are almost the same, the difference become significant above 100
deg 2-theta. Coupling to latt. constants is weaker but still strong. I
remember only one case where refinement of samp. dispacement was possible
i.e. gave a real value. That was pure silicon sample measured with
extremely, extremely, extremely good statistics and up to 165 deg 2-theta.
In other words, if you don't have strong and narrow peaks around 150 deg.
(has anyone ever seen that?) don't touch the sample displacement. If you
see anything that makes you think that you need samp. displ. try to set it
by hand to some small value (of the order of the 0-point value) and see if it
improves your RW-P.
If you still insist or refining samp. displ. there is a trick (suggested by
Prof. Young) in DBWS that lets the user to apply damping factors to
individual variables in order to prevent them from going away from
reasonable values. You must set the codeword value after the parameter
number to less than 1.0 e.g assuming that we deal with the 8-th refinable
parameter use something like "80.1" instead of "81.0" for the codeword. This
will force DBWS to change the parameter by only 1/10 of the calculated
shift. Under usual circumstances this feature is used for dependable
parameters, but will also work for each parameter independently. The
experience is, however, that DBWS often gives diverging refinements when
treated this way.
Greetings for everybody
Stan