H Chondrite Volatile Trace Element Compositional Structure

Chemometric data analysis techniques have proven useful in interpreting volatile trace element composition data in H chondrites. Recently, multivariate linear discriminant analysis and model-independent randomization-simulation have demonstrated that H4, H5, and H6 chondrites are distinguishable from each other with differences being consistent with the onion shell model for the H chondrite parent body [1]. These techniques, however, do not identify which specific elements discriminate between populations. In this work, we examine those volatile trace element patterns that lead to discrimination by utilizing two techniques that help interpret volatile trace element composition structure: principal component analysis (PCA) and canonical details analysis. Volatile trace elements are extremely sensitive to subtle differences in thermal history. Data exists for 58 H chondrite falls; the complete dataset includes Co, Rb, Ag, Se, Cs, Te, Zn, Cd, Bi, Tl, and In (listed in increasing order of volatility) [2,3]. This dataset includes 13 H4, 32 H5, and 13 H6 chondrites, spanning shock facies from a through f. PCA is a technique for analyzing the structure of multivariate datasets. Original measurement variables are transformed into new uncorrelated variables called principal components. Each principal component is a linear combination of the original measurement variables. The first principal component is the axis that represents the direction of maximum variance. Each sucessive principal component is an orthogonal axis in the direction of maximum remaining variance. Dimension of multivariate datasets with highly correlated variables can be minimized, bringing out underlying order and facilitating interpretation [4]. Application of PCA to our 11-dimensional volatile trace element dataset reveals that 8 dimensions are required to account for 90% of the variance of in the dataset. The relatively large number of principal components required to account for the total variance indicates a minimum of correlation between these 11 volatile trace elements. This large number of factors serves as a measure of the true complexity of the volatile trace element compositional structure. These findings indicate the notion that this suite of 11 elements act as a highly correlated group is a great oversimplification. Each element contributes unique information about the thermal history of each meteorite. A method that allows visualization of each element's contribution to discrimination is a canonical details plot. Figure 1 graphically illustrates the centroid of each population in two-dimensional canonical space. The centroids of each population appear as circles corresponding to the 95% confidence region [5]. Elemental rays indicate the loading that these variables have on each dimension in this test space. Figure 1 suggests that all 11 volatile trace elements affect discrimination between these populations to varying degrees. Elements that load the highest on the first canonical axis are Cs, Tl, In, and Co. These four elements span the entire range of volatility in our element suite. These findings are consistent with the onion shell model of the H chondrite parent body. However, a more complex relationship between volatile trace element composition and thermal history is emerging. This complex volatile trace element pattern results from variations during condensation of H chondrite parent material. Subsequent thermal events have only subtly affected volatile trace element composition. References: [1] Wolf S. F. and Lipschutz M. E. (1993) Meteoritics, 28, 460-461. [2] Dodd R. T. et al. (1993) JGR, 98, 15, 105-15,118. [3] Lingner D. W. et al. (1987) GCA, 51, 727-739. [4] Malinowski E. R. (1991) In Factor Analysis in Chemistry, Wiley. [5] Mardia K. V. et al. (1979) In Multivariate Analysis, Academic. Fig. 1., which appears here in the hard copy, shows a canonical details plot of volatile trace element composition of 13 H4, 32 H5, and 13 H6 modern fall H chondrites.