Pressing the Limits of Rietveld Refinement

Abstract

Two examples are given, one with X-ray data and one with netltron data, of the determination of structural detail which appear to be at the edge of current possibility for the Rietveld structure-refinement method. In the first example, 2·2 wt% Sb substituted in CalO(P04)6F2 was located. X-ray powder diffraction data collected with special attention to intensity precision and scale constancy were used. The problem was solved through comparison of intra-sample site-occupancy ratios between Sb-doped and undoped samples. In the second example, high quality, high resolution neutron powder diffraction data were required. The problem was to distinguish between two subtly different models of kaolinite for which the R-weighted-pattern values differed only by 2 or 3 units in the third digit and, particularly, to understand the basis for the consistent programmatic choice of one of the models (PI) over the other. The answer was found in the calculated and 'observed' intensities for (h+ k)-odd reflections; although they were very small, less than 1% of the intensities of the main reflections, many of them were distinctly nonzero. Even though these reflections were not separately observable, because of overlap and small size, they nonetheless correlated with one model sufficiently better than the other to produce the consistent choice.