A new procedure for separating a multi-component powder diffraction pattern into individual component patterns without utilizing the pattern model for one of the components

Abstract

A new procedure has been developed for separating a multi-component powder diffraction pattern into its individual component patterns without utilizing the pattern model for one of the components. In currently used whole-powder-pattern fitting (WPPF) techniques, pattern models are required for all components in a target material that are not in negligible amounts. When the pattern model is unavailable for one of the components due to, for example, low crystallinity, the straightforward application of the WPPF procedure becomes difficult. In the present procedure, WPPF can be conducted even when the pattern model is not available for one of the components. The pattern model for that component can be automatically generated from a seed pattern by filling in the difference between the observed intensity and the sum of intensities calculated for the remaining components at each step in 2θ. The WPPF is conducted by iterative applications of linear least squares to scale the individual component patterns and of direct search (DS) to optimize the parameters, such as the peak positions and profile widths, of the modeled components. The procedure has been incorporated into a computer program for Rietveld refinement, and the simplex method is used for the DS. It has been tested by using intensity data sets of artificially mixed materials with different degrees of complexity in their powder diffraction patterns. The reliability of the procedure has been tested in quantitative phase analysis using separated intensity data sets. The errors in the derived weight fractions are of the same order (0.5 to 2.0 wt%) as those obtained by conventional techniques. Intensity data extracted as unknown patterns were also tested by Rietveld refinement.