Evidence for liquid immiscibility on the Moon has been documented in melt inclusions in minerals that crystallized from mare basalts, within mesostasis in mare basalts and in at least one sample of a lunar plutonic rock. This study focuses on trace-element partitioning between immiscible melt pairs occurring as inclusions in plagioclase. These inclusions are blade-like in shape and range in length from 2 to 150 μm. They consist of two coexisting glasses separated by a sharp meniscus. The colorless, low-index glass (felsite) is spherical and immersed in a dark-brown, high-index glass (high-Fe basalt). The Fe-rich basaltic component of the inclusion is quartz + hypersthene normative with low Mg/(Mg + Fe) and variable SiO2 content (between 34 and 45 wt%). The felsic component has a SiO2 content between 72 and 82 wt% and variable K2O (4.4 to 8.5 wt%)
In most cases, the partitioning behavior of the trace elements agrees with behavior either predicted or measured in experimental and natural systems. The high charge density elements preferentially partition into the basaltic component. High P2O5 and a wider solvus increase the Dbasalt/felsite for these elements. In contrast to many experimental studies and in agreement with studies of natural silicate liquid immiscibility, the Dbasalt/felsite for Ba indicates a preference for the felsic component. This difference in Ba behavior between experimental and natural samples has been attributed to differences in melt polymerization and compensation for charge unbalance within the polymerized melt structure. The apparent differences in Dbasalt/felsite for Sr and divalent Eu between the Apollo 11 high-Ti basalts and Apollo 12 low-Ti basalts may be a result of either subtle differences in bulk composition or the extent of plagioclase plating on inclusion walls. Based on our partitioning data from melt inclusions and the chemical characteristics of lunar felsites, the latter cannot be a product of simple fractional crystallization. Silicate liquid immiscibility can account for fractionation of Ba/ La, K/U, and Ba/U in the lunar felsites and the limited variation in Zr/La. However, it cannot account for the REE pattern of the lunar felsites or the fractionation of U/La. These characteristics must be attributed to whitlockite crystallization prior to the onset of liquid immiscibility.