The program is unlikely to crash during refinement of a single phase, especially if you follow the instructions given here. But it may crash during analysis of complicated, multiphase mixtures, even if you follow these instructions. If you are unable to proceed with a refinement, it may be that a window has opened for input but is hidden by other windows and the system is waiting for your response, so it is always worth checking for any such window before terminating the task.
Regularly viewing the .LST file in the text window is useful for observing progressive changes in chi square. The direction of the change in chi square tells you if the refinement is progressing, digressing, or oscillating. The latter indicates some strong correlation issues that need to be constrained or at least monitored.
Viewing the .LST file is also useful for checking that the number of variables being refined corresponds to the number set. Sometimes more are being refined because you have inadvertently attempted to refine highly correlated variables.
Finally, the .LST file provides values such as density, lattice parameters, mass fractions, and so on.
If it is difficult to get an acceptable background, an alternative approach is to manually select points to use in the background fit. Run powder, bkgedit to bring up the background editing screen. Select background points by pointing and clicking with the mouse and using the add, zoom, and delete options. Then fit a GSAS background function to these points, as above.
If you find that you have mucked up the refinement and the parameters have gone to unreasonable values, the best strategy is usually to go to a previous version of the .EXP file. In EXPGUI, go into file open menu and then select include archived files; this lists the previous archived files by time and date. Just select a previous .EXP file you expect to be all right. If this turns out to be too late or too early an .EXP file, just select another one.
For quantitative XRD, it is generally not necessary to define the site occupancy in each phase. One exception to this recommendation is the ferrite phase, where aluminum and iron (and other elements like Mn and Ti) substitute in tetrahedral and octahedral sites [10]. The site occupancy can be refined in the SAM residue, setting the constraint that their sum is unity. While refining site occupancy improves the pattern fit, it generally affects the calculated mass fraction only a small amount. With XRD, it is both impractical and unnecessary to refine the site occupancy for O or H.
With multiple phases, set the shift to the same value for all phases by constraining the shift parameter for all phases. From the Constraints tab in EXPGUI, select the Profile Constraints tab in the lower portion of the window, select Add Constraint and a new window titled New Profile Constraint will appear. Select the profile term for shift (labeled shft), select Continue. A new window titled New Constraint for Terms will appear with a list of phases included in the refinement on top. Highlight all the phases, be sure the diffraction histogram is selected and select Save to create the constraint. The scale factors of each phase will continue to be refned individually.
Additional crystal structure parameters for minerals are available, for example in [11].