LOW ORDER TRUNCATION ANALYSIS OF X-RAY INTENSITY DATA AS A NOVEL TOOL FOR ACCURATE SITE REFINEMENT. ORTHOPYROXENE AS A EXAMPLE. H. Schlenz, H. Kroll, T. Lueder, A. Fischer, A. Kirfel^*, T. Vad^*, Institut f. Mineralogie, Westf. Wilhelms-Universität, D-48149 Münster, Germany, *Institut f. Mineralogie, Universität Würzburg, D-97074 Wurzburg, Germany

Extremely precise and accurate Fe²⁺,Mg distributions are required in order to derive meaningful cooling rates from the orthopyroxene geospeedometer. Comparison of Mö[[beta]]bauer and X-ray site refinement studies shows that X-ray diffraction usually gives a more ordered distribution than Mö[[beta]]bauer. This challenges the crystallographer to test X-ray site refinement procedures. Atomic site occupancies are expected to be influenced by a variety of factors: The part of reciprocal space from which observations are collected; the structure model, i.e. assumed valence states and type of extinction correction; criteria for excluding ,,unobserved" reflections and outliers from calculations; correlations between site occupancies and vibrational parameters; weighting of the observations. The discussion of X-ray data collection and refinement strategies, using new experimental data from various orthopyroxenes, discloses that the standard practice of single X-ray refinements is inadequate. A novel bivariate technique is suggested that is based on stepwise truncation of low order data (LOT analysis). Thereby, extinction and chemical bonding effects are reduced and correlations between site occupancies and thermal vibration parameters are circumvented, when a B(M2)/B(M1) ratio is chosen that provides invariance of site occupancies against the choice of subsets of data. This technique overcomes the main obstacles in deriving accurate site occupancies from X-ray intensities. As such, it is a generally applicable tool, not being restricted to orthopyroxene.