CHEMICAL FRACTIONATIONS IN-GROUP IIIAB IRON-METEORITES - ORIGIN BY DENDRITIC CRYSTALLIZATION OF AN ASTEROIDAL CORE

We have studied the crystallization history of the asteroidal core tha t produced nearly two hundred iron meteorites of group IIIAB, the larg est group of irons. By critically reassessing the accuracy of the publ ished distribution coefficients between solid and liquid metal for sev en elements and their dependence on the concentrations of S and P usin g the formulation Of JONEs and MALVIN (1990), we have developed a noni deal fractional crystallization model that reproduces all of the major features of the overall chemical trends in group IIIAB. In particular , we can match the variations of Ni, Ga, Ge, Ir, Au, Co, and P, includ ing the Ga and Ge reversals at the IIIA-IIIB boundary. Previous author s were not able to model the Ga and Ge reversals simultaneously and us ed different initial S and P concentrations. Our models and the S conc entration of IIIAB irons suggest that the apparent distribution coeffi cient for S was much higher than its equilibrium value (less-than-or-e qual-to 0.01) and that it increased during crystallization. In our pre ferred model the apparent distribution coefficient for S increases fro m 0.6 to 0.8 while the S concentration of the liquid increases from 6 to 13 wt%. We infer that light S-rich liquid accumulates preferentiall y at the top of the core and in structural traps formed by the advanci ng solid. This process may be aided by the formation of immiscible S-r ich liquid in a boundary layer adjacent to the crystallizing solid. Ou r comparison of the chemical trends in the Cape York irons discovered by ESBENSEN et al. (1982) with the overall group IIIAB trends indicate s that the Cape York irons are not abnormal IIIAB irons that formed by an atypical solid-liquid mixing event, as these authors suggest. We f ind a positive correlation between the divergence of the Cape York and IIIAB trends on element-Ni plots and the scatter of the IIIAB irons f rom the overall IIIAB trends. This correlation and the irregularity of the histogram of Ir concentrations in group IIIAB suggest that group IIIAB contains several Cape York-like sequences and could not have cry stallized from a single well-mixed magma. Instead we suggest that the initially homogeneous magma was subdivided into numerous magma chamber s by km-sized dendrites that grew down from the core-mantle boundary a t an early stage. Although the Cape York irons crystallized from a sin gle magma chamber, we cannot model their compositional trends with clo sed-system fractional crystallization, possibly because of magma excha nge between chambers or formation of S-rich boundary layers.