Intro	Next page	Previous page

---

GSAS/EXPGUI Alumina tutorial (part 5)
Initial Fitting: Refine Scale Factor and Background

---

At this point we are almost ready to start fitting parameters, but before we do that the POWPREF program must be run. In GSAS, each data point has a list of reflections that contribute to that data point. This assignment must be made in POWPREF before the least squares fit can be performed in the GSAS program GENLES. The program must also be rerun if a new phase or histogram is added to the refinement. POWPREF should also be rerun if the lattice constants or profile terms change significantly.

---

We want to refine the background and the scale factor to get started. The scaling parameters are shown on the Scaling panel, shown below. GSAS offers us an overall scale factor for each histogram, plus a phase fraction scale factor for each phase. These two factors have exactly the same effect for a single-phase refinement, so only one can be used. By default, the scale factor refinement flag is turned on and the phase fraction is off. This is what we will use.

---

The other change that we will make in the default refinement options is to lower the number of refinement cycles to 2. Also, make sure the "Extract F_obs" check box on the least-squares pane is selected.

---

We are now ready to start running programs. First run POWPREF by pressing the POWPREF button on the beige tool bar (or selecting the POWPREF option in the Powder menu list.) That causes a window, such as the one below to open as POWPREF runs.

When POWPREF has completed, press the ENTER key to continue and the window should close.

---

The POWPREF program makes changes to the experiment (.EXP) file and this is noted by EXPGUI by the display of the warning message to the right. At this point you do want to accept the changes made by POWPREF, so click "Load New" and EXPGUI will reread the revised file.

By default, this window is shown every time the experiment file is modified by any GSAS program. This allows the file to be "rolled back" to the previous version, in case of a disastrous refinement run, by pressing "Continue with old." However, some EXPGUI users find it annoying to be asked this question all the time. In the EXPGUI Options menu, there is a menu item called "Autoload EXP". If the "Autoload EXP" option is checked, the experiment file will always be read automatically, but then it is no longer possible to roll back erroneous steps so easily.

---

Note that the displayed history record has been updated to reflect the running of POWPREF.

---

We can now initiate the refinement, by launching the GENLES program to optimize the scale factor and background parameters. The output from this run is shown below.

After the refinement completes, press Enter to continue and then press "Load new" on the "Reload?" window.

The output from the GENLES run shows several points worth noting. First note that the number of variables (varied parameters) is 7, from the scale factor plus the 6 background terms. Note that the initial weighted R-factor (wRp or Rwp) and Chi-squared values are 99% and 258, respectively but drop to 72% and 134, respectively after a cycle of refinement. In the first cycle of refinement, there are very large shifts in the parameters, but the parameters converge in the second cycle. This is noted by the sum[(shift/esd)**2] term which is 855 in the first cycle, but 0 in the second cycle. Note that shifts become insignificant when they are small with respect to the esd (standard uncertainty), so a value of 855 means that at least some of the parameters made very large shifts in the first cycle. Note that GSAS does not compute the R-factor or Chi-squared terms after the second refinement cycle.

---

Previous: Change Background Function
Next step: Plotting the Initial Fit

---

Comments, corrections or questions: crystal@NIST.gov
Last modified 18-July-2003
$Revision: 1.2 $ $Date: 2003/04/16 15:23:34 $