>Should one first refine lattice parameters and then "fix" them before 
>proceeding with structure refinement? 

In the early stages of a Rietveld refinement you might want to fix some 
parameters, for example you might want to refine just zero point, cell 
constants, scale factor etc to get them approximately right before refining the 
structural parameters as well. But if you don't refine all parameters together, 
you neglect some of the correlations between them, and you will under-estimate 
errors. The calculated standard deviations are correct only if your model, with 
the parameters you refine, is actually capable of describing the true structure.

For example, if you fix the absorption, or the background as a function of 
angle to wrong values, you will get Uiso wrong, and even its calculated error 
will be under-estimated.

>As I see it, lattice parameters are directly observable in the xrd pattern; if 
>one allows lattice parameters to change during structure refinement, they will 
>no longer be correct.

You mean that the lattice constants are well determined by the peak positions? 
Then the lattice constants will not be much correlated with other parameters 
(except zero, wavelength etc) so they will be correctly refined, especially if 
you refine everything together at the end :-)

I am not saying that constrained refinement is always bad, but if the 
constraints are wrong, then the result will wrong. And the errors will be 
under-estimated, so you won't even know it :-)

Alan.

_____________________________________________________________
Dr Alan Hewat, ILL Grenoble, FRANCE<[EMAIL PROTECTED]>fax+33.476.20.76.48
+33.476.20.72.13 (.26 Mme Guillermet) http://www.ill.fr/dif/AlanHewat.htm 
_____________________________________________________________

Reply via email to