Final differentiation of the lunar magma ocean: genesis of KREEP and the limited role of silicate liquid immiscibility

• Fractional crystallisation in the lunar magma ocean created the REE-rich KREEP signature. • Silicate liquid immiscibility appears at 1030 °C and 98.8 % crystallisation. • Up to 5 % immiscible dacitic melt can segregate without altering the KREEP signature. • The KREEP layer could form from a single basaltic melt or a basaltic-dacitic emulsion. • Partial dacite segregation may lead to reservoirs for granitic melt formation. Owing to the incompatibility of K, rare-earth elements (REE) and P in silicate minerals relative to melt, the KREEP component, found on the near-side of the Moon, is thought to have formed through protracted crystallisation of the Lunar Magma Ocean (LMO). Our fractional crystallisation experiments simulate the final stages of LMO crystallisation, from plagioclase onset to the last eutectic melt remnants. Results show the LMO liquid to remain saturated in olivine ± orthopyroxene ± Cr-spinel up to 74 % solidification (PCS), transitioning to plagioclase+clinopyroxene (cpx) from 1200 °C (74 PCS) to 1120 °C (88 PCS). The plagioclase+cpx+quartz cotectic is reached at 1080 °C (92.3 PCS), with liquid immiscibility and a crystal assemblage of plagioclase+augite+Ti-spinel+ilmenite+quartz occurring at 1030 °C (98.8 PCS), until nearly complete crystallization is reached at 1000 °C (99.5 PCS). Mineral/melt (plagioclase, pigeonite, high-Ca cpx) and melt/melt partition coefficients for K, REE, P, Zr, Hf, Nb, Th, and U were determined. They are used to model melt evolution to 99.5 PCS, showing that fractional crystallisation alone replicates KREEP’s REE profile and the above trace elements, yet, distinct Lu/Hf (and U/Pb) ratios suggest additional processes. Assuming a finite oxygen budget in the LMO and incompatible behaviour of Fe 3+ , the Eu anomaly of KREEP is best reproduced by a model in which oxygen fugacity ( f o 2 ) evolves from one log unit below to 1.5 log units above the iron-wustite buffer (IW-1 to IW+1.5) from 0 PCS to 99.4 PCS. Minor dacitic melt separation (1–5 % of the melt remaining at 1030 °C) sequestering K from REE+ P is consistent with but unnecessary for KREEP formation; nevertheless, a second-stage partial re-melting of these dacites could match observed FeO and incompatible element abundances of lunar granites.