A chemical model for generating the sources of mare basalts: Combined equilibrium and fractional crystallization of the lunar magmasphere

Abstract

It is generally considered that mare basalts were generated by the melting of a cumulate mantle formed in an early Moon-wide magma ocean or magmasphere. However, the nature and chemistry of this cumulate mantle and the logistics of its origin have remained elusive. Extensive studies of terrestrial layered mafic intrusions over the past sixty years have emphasized the imperfection of fractional crystallization and attendant crystal-crystal and crystal-liquid separation in a convecting magma chamber. Crystal-liquid and crystal-crystal separations were similarly inefficient during evolution of the lunar magma ocean (LMO), allowing for the trapping of interstitial liquid and entrainment of a small proportion of less-dense plagioclase into the denser mafic cumulate mush. Indeed, petrography of lunar highlands samples demonstrates this for anorthosites (with 1–10% olivine). The residual liquid after 80–90% crystallization was very evolved (in fact KREEPy) and, even in small proportions (1–5%), would have a noticeable effect on the trace-element chemistry of melts generated from these cumulates. This trapped residual liquid would elevate total REE abundances in the cumulate pile, while synchronously deepening the already negative Eu anomaly. Essentially, this trapped liquid will make the cumulate more fertile for melting to generate both KREEP basalt and mare basalt magmas. Plagioclase entrained in the mafic cumulate pile adds an essential Al component to the high-Ti basalt source and will moderate the requisite negative Eu anomaly in the cumulate. Early in the evolution of the lunar mantle, when the LMO still was largely liquid, it is likely that vigorous convection was an important factor in crystallization. Such convection would allow crystals to remain suspended and in equilibrium with the LMO liquid for relatively long periods of time. This extended period of equilibrium crystallization would then have been followed by fractional crystallization once plagioclase became a liquidus phase and began to float to form the lunar highlands crust. Previous authors have proposed a three-component model for the evolution of high-Ti mare basalt source regions. This model includes KREEP, early (olivine-rich, high Mg#) cumulates, and late (ilmeniterich, low Mg#) cumulates in various proportions. However, we propose a model for high-Ti basalt parent magmas which is in accord with studies of terrestrial layered intrusions. This model for the high-Ti source includes trapped instantaneous residual liquid (TIRL; 1–3%) and entrainment of a small (2–5%) proportion of plagioclase into the late-stage cumulate pile in order to account for both the observed Al compositions and trace-element characteristics of high-Ti mare basalts. Melting of this relatively shallow, ilmenite- and clinopyroxene-bearing, late-stage cumulate can generate high-Ti mare basalt magmas. Furthermore, we are in agreement with other workers that only through a process of nonmodal melting will the high Ti values for the parent magmas be realized. Large-scale convective overturn of the cumulate pile and mixing of KREEP with early- and late-stage cumulates is not required. However, localized overturn of the upper tenth of the cumulate pile is likely and, in fact, required to achieve an appropriate major-element balance for the high-Ti mare basalt source region.