Trace element partitioning between ilmenite, armalcolite and anhydrous silicate melt: Implications for the formation of lunar high-Ti mare basalts

We performed a series of experiments at high pressures and temperatures to determine the partitioning of a wide range of trace elements between ilmenite (Ilm), armalcolite (Arm) and anhydrous lunar silicate melt, to constrain geochemical models of the formation of titanium-rich melts in the Moon. Experiments were performed in graphite-lined platinum capsules at pressures and temperatures ranging from 1.1 to 2.3 GPa and 1300–1400 C using a synthetic Ti-enriched Apollo ‘black glass’ composition in the CaO–FeO–MgO–Al2O3–TiO2–SiO2 system. Ilmenite–melt and armalcolite–melt partition coefficients (D) show highly incompatible values for the rare earth elements (REE) with the light REE more incompatible compared to the heavy REE (DIlm–melt La 0.0020 ± 0.0010 to DIlm–melt Lu 0.069 ± 0.010 for ilmenite; DArm–melt La 0.0048 ± 0.0023 to DArm–melt Lu 0.041 ± 0.008 for armalcolite). D values for the high field strength elements vary from highly incompatible for Th, U and to a lesser extent W (for ilmenite: DIlm–melt Th 0.0013 ± 0.0008, DIlm–melt U 0.0035 ± 0.0015 and DIlm–melt W 0.039 ± 0.005, and for armalcolite DArm–melt Th 0.008 ± 0.003, DArm–melt U 0.0048 ± 0.0022 and DArm–melt W 0.062 ± 0.03), to mildly incompatible for Nb, Ta, Zr, and Hf (e.g. DIlm–melt Hf 0.28 ± 0.05 and : DArm–melt Hf 0.76 ± 0.07). Both minerals fractionate the high field strength elements with DTa/DNb and DHf/DZr between 1.3 and 1.6 for ilmenite and 1.3 and 1.4 for armalcolite. Armalcolite is slightly more efficient at fractionating Hf from W during lunar magma ocean crystallisation, with DHf/DW = 12–13 compared to 6.7–7.5 for ilmenite. The transition metals vary from mildly incompatible to compatible, with the highest compatibilities for Cr in ilmenite (D 7.5) and V in armalcolite (D 8.1). D values show no clear variation with pressure in the small range covered. Crystal lattice strain modelling of D values for di-, tri- and tetravalent trace elements shows that in ilmenite, divalent elements prefer to substitute for Fe while armalcolite data suggest REE replacing Mg. Tetravalent cations appear to preferentially substitute for Ti in both minerals, with the exception of Th and U that likely substitute for the larger Fe or Mg cations. Crystal lattice strain modelling is also used to identify and correct for very small ( 0.3 wt.%) melt contamination of trace element concentration determinations in crystals. Our results are used to model the Lu–Hf–Ti concentrations of lunar high-Ti mare basalts. The combination of their subchondritic Lu/Hf ratios and high TiO2 contents requires preferential dissolution of ilmenite or armalcolite from late-stage, lunar magma ocean cumulates into low-Ti partial melts of deeper pyroxene-rich cumulates