17.3 Peak Width modes

There are three ways that peak widths can be managed.

1. If a peak width (\(\sigma ^2\) or \(\Gamma \)) has its refine flag turned on, then that value in the table will be used for an initial peak fit and will be refined if a peak fit is performed.
2. In the default mode of operation, any unrefined \(\sigma ^2\) or \(\Gamma \) values will be set from the histogram instrumental parameters, (U, V & W for \(\sigma ^2\) and X, Y & Z for \(\Gamma \)). When this is done, it is possible to refine those instrumental parameters, but make sure that there are enough peaks with widths set in that manner to over-define all the instrumental terms that will be refined.
3. Should you want to fix the \(\sigma ^2\) and \(\Gamma \) values to specific values and not refine them, there is a mode that can be used to prevent changing peak widths to the values from the There is menu command setting, “Peak Fitting”/“Gen unvaried widths” that is enabled (checked) by default. If this is unchecked, then the values of \(\sigma ^2\) and \(\Gamma \) will not be changed, even when not refined.

My recommendation is that the Gaussian broadening, \(\sigma ^2\), be generated from the instrumental values for U, V & W. If the instrument performance has been calibrated, these values do not need to be varied. If the instrument has not had its instrumental broadening calibrated, the U, V & W values may be refined here and these provide reasonable starting values for a calibration. These Gaussian values will almost never be affected by sample phenomena, so they should not be affected by any type of anisotropic broadening.

The \(\Gamma \) values will almost always arise from sample broadening and may vary from peak to peak. This can be because adjacent peaks may arise from different phases or there may be anisotropic crystallite or size broadening. None of this can be modeled with refinement of the X & Y terms. The Z term is not associated with any physical process and is not expected to be needed, but is provided to allow more freedom in Lorentzian widths generation, should that ever be needed. Thus, it does make sense to vary the individual gamma terms for each peak.