Structure analysis by reduced data. II. Increase of reproducibility of the results of diffraction studies

The second part of the paper shows the advantages of the method of interexperimental minimization (IEM) over the conventional least squares (LS) method in the crystal-structure refinement. The new IEM method minimizes the norm of the differences between the data measured in independent experiments. The parameters of the structure model refined by the LS and the IEM methods are compared for four experimental data sets from an alexandrite crystal Al2BeO4 : Cr3+. The model parameters calculated by the LS method from the initial experimental data sets showed insufficient reproducibility (the number of admissible values of the statistical-test criterion based on the normal-probability plot equals 5.6%). The use of the IEM algorithm increases the number of admissible values up to 38.9%. However, the use of fitting and statistical methods cannot provide a 100% reproducibility of all the model parameters automatically, which requires the reconstruction of the lost information on the characteristics of the experiment. (C) 2001 MAN "Nauka/Interperiodica".