Behaviour of boron, beryllium, and lithium during melting and crystallization: Constraints from mineral-melt partitioning experiments

In order to provide a mon substantial foundation for interpreting the behaviour of B, Be, and Li during the production and early crystallization of primitive igneous rocks, we have measured olivine-, clinopyroxene-, orthopyroxene-, and amphibole-melt partition coefficients for these elements involving broadly basaltic-andesitic melt compositions. Experiments were conducted at both one atmosphere and 1.0-1.5 GPa and employed a time-temperature history that yielded large crystals with minimal compositional zoning. Experiment temperatures ranged from 1000 to 1350 degrees C and were selected to minimize the total crystal fraction in a given experiment. Partition coefficients for olivine and clinopyroxene were found to be independent of run duration or total concentration of B, Be, or Li suggesting that crystal-liquid equilibrium was closely approached. Olivine-, orthopyroxene-, and clinopyroxene-melt partition coefficients decrease in the order: Li (0.1-0.2) much greater than Be similar to B (0.002-0.03), whereas amphibole-melt partition coefficients for Be and Li are similar (similar to 0.2) and larger than those for B (similar to 0.02). Comparison of partition coefficients measured in this study with previous determinations yields good agreement, with the exception of some of our mineral-melt values for B, which are uniformly lower (up to 10 times) than values determined at similar conditions of pressure and temperature. The latter discrepancy could be due to mineral or melt compositional effects, but this hypothesis is currently untestable owing to the absence of reported mineral compositions in previous studies. Partition coefficients for olivine and clinopyroxene have been found to vary as a function of mineral and melt composition, and with the exception of B partitioning into clinopyroxene, this variation can be modeled using simple exchange reactions involving the trace element and a substituent element, such as Na, Mg, or Al. Partition coefficients measured in this study were combined with simple models of melting and crystallization to evaluate how accurately element ratios such as B/Be, B/K, B/Nb, Be/Nd, Li/V, and Li/Yb in primitive magmas reflect that of their source. These models further confirm that the source regions of IAB magmas are enriched in B/Be, B/Nb, and Li/Yb relative to the MORE source, thus lending further support to the notion of metasomatic enrichment of the IAB sourer by slab-derived fluids. Moreover, our modeling also indicates that the low B/Be and B/Nb in primitive OIB magmas is indicative of similarly low values in OIB sources, which is consistent with the hypothesis that OIB sources contain a B-depleted component, such as subducted, dehydrated oceanic crust. Partial melting models have also been constructed to explore the possibility of using the LIN ratio in MORE and IAB as a monitor of redox conditions in their source-regions. Models indicate that this ratio does not uniquely constrain source fO(2) without a priori knowledge of the degree of melting. However, the small amount of dispersion in MORE Li/V is consistent with (1) the small variation in source-region fO(2) inferred for MORE by independent means and (2) degrees of melting close to clinopyroxene exhaustion. The very large dispersion in Li/V ratios in the IAB suite can be reconciled by melt generation under more oxidising conditions than that for MORE, in addition to variation in source composition resulting from metasomatism involving a Li-rich component. Copyright (C) 1998 Elsevier Science Ltd.