A practical approach to Rietveld analysis. Comparison of some programs running on personal computers

Abstract

Three computer programs for Rietveld analysis DBWS-9411, HILL-93.06, and FULLPROF-3.1 have been tested and compared using data for two samples of different complexity, spinel, and anglesite. The investigation shows that results are “program dependent.” The obtained R indices for spinel are in the ranges 10.60%–13.26%(Rwp) and 3.15%–5.25%(RB). Similarly, the ranges for anglesite are 9.76%–14.06%(Rwp) and 2.15%–5.06%(RB). Some atom and displacement parameters are significantly different, too. In attempt to define the standard procedure for Rietveld analysis, three parameters, n, BKPOS, and RLIM were examined. It was found that the most appropriate values for them are: n=8–10, BKPOS=90°, and RLIM=40° 2θ. Using Fourier filtering for background modeling, significantly lower R indices were obtained in comparison to polynomial and interpolated background. At the same time the great numbers of atom and cell parameters agree within ±3σmax and e.s.d.'s were identical or lower than those achieved for polynomial and interpolated background. It was found that the function given by Bérar and Baldinozzi (1993) (J. Appl. Crystallogr. 26, 128–129) much better described asymmetric peak profiles at low 2θ angles. This function and Fourier filtering were implemented only in FULLPROF, which has more possibilities and some advantages over the other two programs. In addition, the peak shift parameters (sample displacement and transparency) were tested. It was shown that under present circumstances these parameters do not have much effect on atom parameters and R indices. However, differences in unit cell parameters were considerable greater, most probably because of the large correlation between zero-point, lattice, and peak shift parameters.