F, Cl and 'water' mineral-melt partitioning in a reduced, model lunar system: what does the ‘volatile’ content of lunar rocks tell us?

The presence of ‘water’ (H-related defects), F, Cl and other volatiles in mare basalts and associated lunar volcanic glasses implies that lunar magmas were variably volatile-bearing^[1]. Mineral-melt partition coefficients for each species can, in theory, be used to calculate volatile contents of lunar mantle source regions for these magmas. However, available partitioning data is largely based on studies in model terrestrial systems. Aside from differences in composition, the lunar mantle likely had an oxygen fugacity (fO₂) at least 2 log units lower than the Earth’s upper mantle^[3] and was either at or near Fe-saturation^[1]_. fO₂ can have a fundamental influence on ‘water’ speciation in silicate melts, with H₂ and C-H related defects stabilised under reducing conditions, at the expense of (OH)^-  defects which dominate in oxidising terrestrial conditions^[4]. As such, partitioning data in model, reduced lunar systems is required to accurately interpret measured volatile contents in lunar materials.

We have developed a novel experimental design which allows us, for the first time, to constrain mineral-melt volatile partition coefficients under lunar mantle conditions. Experiments were performed at 2-3 GPa, 1350-1500°C, and at fO₂ of IW-5 to IW+2 log units, in a system based on an average Apollo green glass^[1]. Near-liquidus experiments were run to constrain olivine- and pyroxene-melt partition coefficients (D^min/melt₎ for water, F and Cl, using EPMA and SIMS data. As shown in Fig. 1, values of D^min/melt are comparable to those from experiments in more oxidised, terrestrial systems. F (and possibly Cl) incorporation in pyroxene and olivine is correlated with trivalent cation (Ti³⁺, Al³⁺, Cr³⁺) content, suggesting coupled substitution mechanisms^[5], and is also dependent on the extent of melt polymerisation. However, fO₂ has little discernible influence on partitioning behaviour. This is consistent with infrared and Raman spectra which show that (OH)^- defects are the dominant mechanism for water incorporation in both silicate melt and coexisting mineral phases, with only a small proportion of water present as H₂ defects in the most reduced, highest pressure samples. Higher pressures can stabilise a greater proportion of ‘water’ as H₂, although this is unlikely to be an important mechanism for water storage in all but the deepest parts for the lunar mantle, as noted in previous thermodynamic models ^[6].

Partitioning data can be used to re-interpret volatile contents of lunar mantle source regions. A batch melting calculation with 4-8% partial melting of a lunar mantle cumulate of 50% olivine, 25% pigeonite and 25% orthopyroxene^[7], based on volatile concentrations from olivine-hosted melt inclusions and lunar volcanic glass beads^[7,8], implies a lunar mantle source region with 2-8 ppm F, 0.07-0.3 ppm Cl, and <15 ppm H₂O. This represents 95% loss of both F and water, and 98% loss of Cl in the lunar mantle source relative to bulk silicate Earth (BSE). However, postulating a bulk lunar volatile content based on mare basalt mantle source regions is, of itself, misleading. During LMO solidification, volatiles will have been partitioned between crystallising phases (cumulates) and remaining melt, with the last dregs of magma likely solidifying as the hypothesised KREEP material. Any specific mantle source region is not necessarily representative of the bulk silicate Moon (BSE). To assess the effects of LMO solidification, we used partitioning data^{here and [10]} to simulate incremental solidification using a simplified crystallisation sequence^[11], for up to 97% solidification. Resulting volatile contents, and volatile ratios in cumulates formed at each stage of LMO solidification are shown in fig. 2, based on an arbitrary starting volatile concentration in the LMO of 1000 ppm for each. For up to 50% solidification, volatile content of cumulates remains very low. This drives volatile enrichment in the LMO, and formation of volatile-enriched, pyroxene-rich cumulates during latter stages of solidification. Differences in partitioning result in fractionation in cumulates, such that F is highly enriched relative to Cl, and slightly enriched relative to H₂O (Fig. 2B). At 97% solidification, water content of final cumulates is <60 ppm, and F just over 120 ppm. The final dregs of the LMO, as expected, are volatile enriched (>2000 ppm for each), but with no significant fractionation of volatiles (Fig 2C).

Batch melting calculations show that melt 1, formed by 10% melting of cumulate (Fig. 2A) contains 241 ppm H₂O, 428 ppm F and 36 ppm Cl (again, based on 1000 ppm of each in the original LMO), with F/Cl =12, and F/H₂O =1.8. A 10% melt 2 contains 382 ppm H₂O, 870 ppm F and 109 ppm Cl, a F/Cl =8 and F/H₂O =2.3. Therefore, partial melts of LMO cumulates partially retain the strong signatures of LMO solidification. Addition of a KREEP component will increase absolute volatile contents in lunar magmas, and dilute volatile ratios. However, our model predicts that pristine lunar mantle melts will have greatly elevated F/Cl, and high F/H₂O ratios compared to the BSM and original LMO. Back-calculating LMO volatile contents for melt 2, using published lunar magma volatile contents^[7,8], implies an original LMO with 38-88 ppm F, 16-26 ppm Cl, and 315-2441 ppm H₂O. These values overlap those of the BSE, and require only minor additional fractionation of Cl relative to F. Modelling based on new partitioning data implies, therefore, that (1) LMO solidification had a significant effect on volatile distribution and fractionation within the lunar interior; (2) that signatures of LMO are retained in later mantle melts; (3) that the high lunar F/Cl^[9] may reflect coupled effects of accretion of the Earth and later LMO solidification, and (3) that the Moon could be less volatile-depleted than previously estimated. These observations are consistent with a model where lunar volatiles are inherited from the early Earth.

^[1]Saal et al.(2008)_Nature_454:192–196.

^[2]Longhi (1992)_Geochim.Cosmochim.Acta_56:2235–2251.

^[3]Rutherford and Papale (2009)_Geology_37:219–222.

^[4]Kadik et al.(2006)_Geochem.Int._44:33–47.

^[5]Dalou et al. (2012)_Contrib.Mineral.Petrol._163:L591–609.

^[6]Hirschmann et al.(2012)_Earth.Planet.Sci.Lett._345:38–48.

^[7]Zhang et al. (2019)_Earth.Planet.Sci.Lett._522:40–47.

^[8]Chen et al.(2015)_Earth Planet.Sci.Lett._427:37–46.

^[9]Hauri et al. (2015)_Earth Planet.Sci.Lett._409:252–264.

^[10]Lin et al. (2019)_Geochemical Perspect.Lett._10:14–19.

^[11]Rapp and Draper (2018)_Meteorit.Planet.Sci._53:1432–1455.

 

Acknowledgements: Work supported by the UK National Environmental Research Council NE/M000346/1 and IMF597/0516 (to Bromiley). Brooker was funded by the NERC Thematic Grant consortium NE/M000419/1