Bimodal distributions of profile-broadening effects in Rietveld refinement

It is demonstrated with computer simulations that a bimodal distribution (narrow and broad) of reflection profile broadening effects, such as crystallite size and microstrain, can lead to several spurious results when the profiles are modeled by single pseudo-Voigt functions in Rietveld refinement (RR), e.g. mixing parameter η > 1 (`super-Lorentzian'), a few observed data points lying consistently above the calculated peak maxima for some reflections, misrepresentation of the background level and shape, and in special cases well defined small observed peaks not seen in the calculated pattern as though they belonged to an additional minor phase. With a bimodal model (two pseudo-Voigt functions) the RR's accurately returned the profile, mixing, and background parameters used in the data synthesis both without and with counting statistical fluctuations simulated in the data. The RR method thus demonstrates remarkable ability to sort out complex specimen characteristics when accurate modeling functions are used.