EXPGUI Tutorial: Garnet (part 6)

6. Additional Least-Squares Refinement

The initial fit obtained in the previous runs is not particularly good. The "first derivative" appearance in the difference plot peaks indicates that peaks are not correctly positioned, so refining the lattice parameters and zero point correction are the first order of business.

While normally I would recommend refining the lattice first and then in a subsequent refinement cycle also refining the zero point correction in a later cycle of refinement, with a cubic cell and a small zero-point error it is possible to refine both together. Set the refinement flag for the lattice (shown above) and the zero point correction (shown below).

and then start GENLES, as was done before.

A dramatic improvement is seen both through the agreement factors (above) and as seen in the plot (below).

At this point it is a good idea to run POWPREF and GENLES again, since peak positions have moved. This causes the Reduced Chi**2 to drop from 14.9 to 13.5.

(.EXP Snapshot: garnet6.exp )

---

Previous step: Plot data using POWPLOT or LIVEPLOT
Next step: Define Constraints

---

Comments, corrections or questions: crystal@NIST.gov
Last modified 02-August-2007 by website owner: NCNR (attn: Craig Brown)
$Revision: 1.2 $ $Date: 2002/04/12 21:34:08 $