If one aligns a diffractometer with a glass slit (like the one our
Siemens D500 was equipped with) and a lot of patience, you can align the
target on the tube anode (or the focal point in the case of an incident beam
monochromater), the axis of the goniometer and the recieving slit
to be coplaner to within .01 degrees. If you have one of the newer machines
that allow for user definition of the angles (a D5000) you can do better
still. Use of SRM 660 or 640b to obtain a delta d curve (obs d - calc d vs.
two theta) will yield a plot with a negative slope reflecting the effects
of the divergence slit and axial divergence, both of which shift the observed
position of peaks to an angle lower than the true position. As these effects
predominate at low angle, the plot tends to zero at high angle. Ploting the
data vs. Cos theta will indicate if the problem is due to sample displacement
as slope will indicate this error. However such plots are often nonlinear
reflecting the aforementioned effects.
Use of the code GSAS allows for the refinement of tranparency, sample
shift, asymetery and zero point to correct for shifts in peak position. The
transparency term works well indicating plausable values for specimen
attenuation, the asymetery correction does not work well in that it was not
designed divergent beam x-ray diffractometers. The sample shift refines to
values which attempts to correct for the axial divergence and divergence
slit effects. This is obviously not the correct route but in the absence
of an approapriate model is not unreasonable. I don't refine a zero point
because the machines are known to be correct and because of the correlation
problems which various people have mentioned. When the improved L. Finger
asymetery model is incorporated in GSAS this situation may be improved.
Jim Cline
NIST