The Prediction of the Saturated Activity of 26Al in NonAntarctic Stony Meteorites from their Chemical Compositions
Abstract
We have assembled from the literature a database of over 300 nonAntarctic stony meteorites, containing information about their chemical composition, date of fall, total mass, and gas exposure age, etc. We have developed an iterative algorithm using weighted linear multivariate regression, which surveys all the independent variables in the database, recommends the best 26 models (combinations of variables) for the prediction of Al activity, and using those models, performs weighted linear multivariate regressions. By requiring that the residuals be normally distributed, under and supersaturated meteorites are discovered and eliminated. This process is iterated until a stable solution is obtained. As a result, we obtained a set of 128 saturated, 50 unsaturated, and 10 supersaturated meteorites. We find that the expression: ^26Al = (5.28+0.81) . Al + (2.59+ 0.06) . Si + (1.57+0.39) . S + (1.52+0.59) . Ca, Chi^2(sub)nu = 2.59, where the elemental concentration is given in weight 26%, is the best predictor of the saturated ^26Al content of a stony meteorite. We find no evidence of bias or crippling multicollinearity in this expression. About one half of the remaining variability cannot be attributed to uncertainties in the determination of the ^26Al content and thus must be attributed to variations in orbit, shielding, etc. We compare our results (see figure) with those of other workers (1,2,3,4), and examine the probable causes of the disagreements displayed there. We show that saturated ^26Al is distributed among all classes of meteorites in about the same way, with the exception of the carbonaceous chondrites and the eucrites, which both have about the same excess proportion of unsaturation. We examine the question of the convergence of expressions derived from regressions on chemical composition to the predictive expressions derived from integrals of particle fluxes and nuclear reaction cross sections and show that they need not converge. We examine the methods of calculating the uncertainties of predictions from regression expressions and show that previous methods can lead to significant errors. REFERENCES: 1) Fuse, K. and Anders, E. (1969) Geochim. Cosmochim. Acta 33 653670. 2) Cressy, P. J. Jr. (1971) Geochim. Cosmochm. Acta 35 12831296. 3) Hampel, W. et al. (1980) Geochim. Cosmochim. Acta 44 539547. 4) Honda, M. (1988) Meteoritics 23 312.
 Publication:

Meteoritics
 Pub Date:
 July 1992
 Bibcode:
 1992Metic..27Q.241K