Main

Following the giant impact formation of the Moon1, the lunar precursor material remained in a molten state for a period of between 20 and 200 Myr (ref. 2). Intense heating during the collision followed by large-scale evaporation during the lunar magma ocean (LMO) phase is thought to have depleted the Moon of volatile elements (for example, H, S, Cl) and triggered substantial fractionation of their stable isotopes (see, for example, refs. 2,3). The connected history of volatile elements on the Moon and Earth highlights their crucial importance to understanding the history of volatiles on Earth4. As a consequence of the lack of an atmosphere and terrestrial weathering, the Moon provides a well preserved record of events, processes and volatile composition and distribution in the inner solar system since its formation (>4.4 billion years ago, Ga)5.

Towards the end of an extended LMO stage, the fractional crystallization of the Moon produced a dense cumulate pile comprised mainly of pyroxene and olivine and a plagioclase flotation crust, which prevented further loss of volatiles6,7. This flotation crust is mainly represented in the Apollo sample suite by FAN material, with later intrusive lithologies such as the Mg- and alkali suites forming in response to cumulate overturn in the lunar mantle2,8,9,10 and melting at the crust–mantle boundary, respectively11,12,13. Ferroan anorthosites (FANs) are the only lithology on the Moon thought to be a primary product of the LMO5,14. Towards the end of LMO crystallization, the late-stage residual melt, enriched in K, rare-earth elements (REEs) and P (termed urKREEP), resided between the crust and mantle—although its geographic extent on the Moon is unclear5. A large proportion of the Apollo suite samples have a notable KREEP component, and hence observations of their volatile abundances and compositions may largely reflect the KREEP reservoir15.

F, Cl and OH form essential structural constituents (ESCs) of lunar apatite, Ca5(PO4)3(F, Cl, OH)—which is the only mineral group on the Moon identified with OH as an ESC that has been detected and studied in various lunar samples16. Apatite has thus far been identified in all lunar sample types except FANs and volcanic glass beads17, as the former are thought to have formed before apatite saturation in the LMO and apatite has not been reported as inclusions in volcanic glass beads. This has limited the study of the volatile inventory of the earliest lunar crust to nominally anhydrous minerals (for example, plagioclase)18 and bulk-rock samples19. The extremely low volatile abundances of nominally anhydrous minerals make stable isotope analysis challenging and increase the effects of cosmic-ray spallation on δD values, while the relationship between bulk-rock and apatite δ37Cl values is poorly understood, and thus bulk-rock isotopic analyses may not be entirely representative20. These studies of FAN material have reported that Cl and H are moderately to heavily fractionated (δ37Cl = +11–30‰ (ref. 19); δD = +280–310‰ (ref. 18)). In this Article we report on observations of apatite in a FAN clast in lunar meteorite Arabian Peninsula (AP) 007 and the abundance and isotopic composition of Cl and H, as well as U–Pb and Pb–Pb chronology and Ce abundance in this apatite.

Arabian Peninsula 007

AP 007 was discovered by meteorite hunters in Al Jouf, Saudi Arabia, in 2015, as a single weathered stone lacking a fusion crust21. Its texture, mineral chemistry and oxygen isotope composition confirm a lunar origin (Supplementary Fig. 1). AP 007 is a clast-rich fragmental breccia, containing granulitic and recrystallized impact melt clasts enclosed in a dark, fragmental matrix. A coarse-grained crystalline clast, bearing plagioclase (An95–97), olivine (Fo44–63) and pyroxene (En23–67Wo6–44Fs21–38), appears to be of the FAN suite (Supplementary Fig. 2) and contains apatite, ilmenite and ulvöspinel (Fig. 1). The modal abundances of plagioclase (∼76%), olivine (6%) and pyroxene (17%) are similar to those of the mafic magnesian FAN subgroup22. The higher abundance of mafic silicates in AP 007 (and some other FAN samples) could indicate that they originate from shallower levels in lunar crustal stratigraphy14. Plagioclase in this clast shows strong zoning in cathodoluminescence (CL) imaging (Supplementary Fig. 3), indicating a prolonged cooling period consistent with a crustal origin and previously observed FAN material (see, for example, ref. 23). The texture of apatite is consistent with a primary magmatic origin (Supplementary Section 1.4).

Fig. 1: Texture of the FAN clast and associated apatite.
figure 1

a, Backscattered electron (BSE) map of the FAN clast in AP 007. b,c, BSE (b) and CL (c) images of apatite in the FAN clast. ap, apatite; pl, plagioclase; px, pyroxene; ol, olivine. The white dashed lines in a denote the region within the FAN clast in which the apatite is located, and yellow dashed lines in b,c outline the grain boundaries of the apatite.

Full size image

The FAN suite is representative of the early lunar crust, and these samples have previously been reported to have ancient ages (Supplementary Fig. 4). Analysis of this clast yields a terrestrial common Pb-corrected 207Pb/206Pb age between 4,467 ± 43 and 4,540 ± 34 Myr (the latter value is data acquired on apatite; Supplementary Section 1.3 and Supplementary Fig. 5). This date is older than has been observed in most lunar material so far (with a similar age to Apollo FAN sample 67016; Supplementary Fig. 4), although there has been some evidence for a pre-4.5 Ga formation for the Moon24. FAN samples have radiometric ages of ∼4.3–4.5 Gyr (refs. 24,25). It should be noted that the trend towards a radiogenic lunar Pb component introduces some dating ambiguities (Supplementary Fig. 6). Papanastassiou and Wasserburg26,27 measured Rb–Sr ages of 4.55 ± 0.10 Gyr for Apollo dunite 72417 and 4.54 ± 0.07 Gyr in troctolite 76535, while Norman et al.14 found a Sm–Nd age for the lunar crust of 4,456 ± 40 Myr. An age of 4,467–4,540 Myr supports previous work, which proposed crystallization of the lunar crust within ∼60 Myr of calcium–aluminium-rich inclusions14,28,29. This evidence places the formation of this clast in AP 007 very early in the Moon’s history and among the earliest FAN samples. Analysis of the volatile systematics of this early crustal material may provide greater insight into the evolution of volatiles at this stage in the early Solar System history.

Results

The δ37Cl measured in this apatite is heavily fractionated (∼+44.2 ± 2.2‰, 2σ), with a high Cl abundance (∼1.64 ± 0.09 wt%; Supplementary Fig. 7). This is significantly heavier than bulk data observed in Apollo FAN samples 60015 (+10.5–11.4‰)19 and 60025 (δ37Cl = +24.5–30.2‰ (ref. 19); Fig. 2a) and in most other lunar material (Supplementary Fig. 8a and discussion in Supplementary Section 1.5). The δ37Cl values in 60025 were attributed to significant degassing during the slow cooling of the FAN material19. Through back-calculation of the Ce abundance of the parental melt (∼250 ppm) from that of apatite (∼540 ppm), there is some similarity to more evolved endmembers of the FAN subgroups (Supplementary Section 1.7), which have been shown to not represent KREEP contamination14.

Fig. 2: δ37Cl and δD values on AP 007.
figure 2

a, δ37Cl. b, δD. The values are compared with literature bulk and plagioclase in situ data on FANs18,19, KREEP15,42,46,47,48 and the Mg suite18,37,46,49,50,51. Standard error 2σ uncertainties on AP 007 data are smaller than the symbols, and are not reported for 60015 and 6002519. Data on AP 007 are presented as exact values ± 2σ. Data on 15415, 60015 and 60025 are presented as mean values ± 2 s.d. Data on Mg-suite and KREEP are presented as a range.

Full size image

The δD measured in this apatite is ∼−45 ± 45‰ with a moderately high equivalent H2O content (∼2,300 ± 100 ppm). This δD value is comparable to those of Apollo highlands and KREEP samples (Supplementary Fig. 8b and discussion in Supplementary Section 1.5). This is lighter than observed in plagioclase within Apollo FAN samples 15415 (δD = +280‰) and 60015 (δD = +310‰; Fig. 2b), although the result for 15415 is subject to very large uncertainties (500‰), in part owing to the low H2O content in plagioclase18.

Discussion

It is conceivable that the Cl signature in AP 007 reflects degassing before apatite formation; this would be expected to cause a depletion in Cl in its parental melt. Rayleigh fractionation modelling of chlorine isotopes during degassing from an initial δ37Cl of 0‰ requires >92% degassing of main metal chlorides (for example, FeCl2, NaCl and ZnCl2) to reach the δ37Cl values observed here. Wang et al.30 proposed that substantial degassing of the LMO before lunar crust formation caused a heavy 37Cl enrichment. The δ37Cl in AP 007 is greater than KREEP values (Fig. 2)—indicating that the fractionation observed in this sample cannot be primarily attributed to KREEP contamination, and thus other processes must be explored.

A similar enrichment in 37Cl (δ37Cl = +40.0 ± 2.9‰) has been attributed in Apollo 14 basalts to vapour phase metasomatism following an impact31, and in bulk FAN samples 60015 and 60025 to condensation/metasomatism of a HCl vapour following LMO degassing19. Metasomatism via an evolved liquid would be expected to enrich these crustal anorthosites in REEs32. The observed apatite and reconstructed parental melt Ce abundances (<300 ppm) do not support metasomatism (Supplementary Section 1.7). The relatively light H isotopic composition further challenges a metasomatic origin for this apatite grain, due to the predominance of the heavier isotope in aqueous fluids and vapour phases. A lack of disturbance to the strong zoning observed in plagioclase of this clast (Supplementary Fig. 3) also indicates minimal impact interference to the rock after it solidified. This indicates that alteration of the Cl and H isotopes must have occurred before the crystallization of the rock: for example, via magmatic degassing.

Assuming a primordial δD value of –280‰, which is consistent with observations in other isotopic systems (for example, Ti and Cr, which both show an enstatite chondrite contribution to the early Moon33) and comparable to the proto-Earth’s mantle (see, for example, ref. 34) and CI, CO and CM chondrites35, >96% degassing of H2 and >99% degassing of H2O are required to raise the δD of LMO residual magmas to +310‰ (reported δD value of Apollo FAN 60015)18. In comparison, ∼80% degassing of H2 and >99% of H2O is required to produce a δD of –45‰. Hui et al.18 concluded that H2 degassing during LMO crystallization controlled the hydrogen isotope composition of the residual melt, in agreement with our results. Degassing of H2 followed by metal chlorides is consistent with reported volatile abundances for the LMO in the literature (Supplementary Fig. 9). It is also possible that HCl was the dominant vapour phase, in which case the aforementioned condensation and metasomatism could produce the fractionated δD values observed in AP 007. However, this is not consistent with the observed REE abundances in AP 007, and therefore should be discounted (Supplementary Section 1.6). The H2O content of this apatite is significantly higher than estimates of the bulk silicate Moon (∼130–300 ppm H2O36), which may be a function of partitioning of volatiles into apatite37. This FAN clast is of a comparable age to other FAN suite samples (Supplementary Fig. 2), indicating that H fractionation was induced early in the lunar evolution. Through modelling of volatile contents in the LMO (Supplementary Section 1.6), the H and Cl isotopic compositions seen in AP 007 are most consistent with an initial H2O content of <10 ppm. The results for AP 007 suggest that a significant fractionation of volatiles preceded the complete crystallization of the dry LMO. This could prompt further investigation of the Cl and H isotopes of younger lunar material to recontextualize lunar volatile history.

In this Article we have presented the in situ abundance and isotopic composition of Cl and H in apatite within a FAN clast in AP 007. The identification of apatite in this rock suite provides an insight into the volatile systematics of primary products of the LMO. It also highlights the importance of lunar meteorites in sampling lunar geology/volatiles. Isotopic measurements are supplemented by U–Pb and Pb–Pb chronology and Ce abundance measurements. An ancient 207Pb/206Pb age for this clast indicates that fractionation of volatiles occurred very early in lunar history, likely prior to the complete crystallization of the LMO.

Methods

We conducted a systematic search for apatite grains in thin sections of AP 007, which allowed for the identification of grains sufficiently large (at least ∼10 × 10 µm2) for nanoscale secondary-ion mass spectrometry (NanoSIMS) analysis at a high spatial resolution, and high-precision U–Pb chronology using in situ SIMS dating methods. Apatite grains were identified using BSE images and X-ray spectra using a FEI Quanta 200 three-dimensional scanning electron microscope at The Open University at accelerating voltages of 15–20 kV and a beam current of 0.60 nA. One apatite grain was found to be associated with a FAN clast with an anhedral habit and appearing interstitially between larger silicate minerals (mainly plagioclase; Fig. 1). Quantitative geochemistry of both the silicate and phosphate phases was collected using a CAMECA SX100 electron microprobe at The Open University, at accelerating voltages of 10–20 kV and beam currents of 4–20 nA. Phosphate phases were examined at higher beam currents (∼10–20 nA) following NanoSIMS data collection, to remove the possibility of high probe beam currents affecting the accuracy of SIMS measurements (as electron-probe microanalysis currents would potentially mobilize the volatiles in the apatite grain)38.

The Cl content and isotopic composition and Ce abundance of apatite grains were measured using the CAMECA NanoSIMS 50L at The Open University, following a modified protocol by Stephant et al.39 (Supplementary Table 1). A 16 keV primary Cs+ beam of ∼50 pA was used to pre-sputter areas of either 8 × 8 µm2 or 10 × 10 µm2 on apatite. Negative secondary ions of 13C, 18O, 35Cl, 37Cl, 40Ca19F and 140Ce16O2 were collected simultaneously on electron multipliers in scanning ion imaging mode, with a mass-resolving power of ∼8,000 (CAMECA definition). Drift correction and data processing were carried out using the L’image software40 developed by L. Nittler. Regions of interest over the apatite grains were selected primarily within 35Cl/18O images (Supplementary Fig. 7), allowing for the removal of signal contaminated by epoxy present in cracks and voids.

The water content (reported in this Article as equivalent H2O; Supplementary Table 1) and hydrogen isotopes were acquired on top of existing Cl isotope image pits using the CAMECA NanoSIMS 50L in multicollection spot mode (allowing for lateral collocation of H and Cl results), refining an established protocol41. A 16 keV primary Cs+ beam of ∼200 pA was used to pre-sputter areas of 10 × 10 µm2 on apatite. Negative secondary ions of 13C, 1H, 2H and 18O were collected simultaneously on electron multipliers using this primary beam, with a mass-resolving power of ∼4,300 (refs. 37,42). Cosmic-ray exposure ages have not been reported for AP 007—and a whole-rock cosmic-ray exposure age would not be representative for all of the components of this complex fragmental breccia. Moreover, at the high H2O content of apatite in AP 007, spallogenic H would not cause significant alteration of its H isotope composition (Supplementary Fig. 10). The D/H ratios and H2O abundance of these results should, therefore, be considered as a maximum.

Bulk oxygen isotope analysis was undertaken at The Open University using infrared laser fluorination (see ref. 43 for a breakdown of this method). A 2 mg aliquot of AP 007 and standards were loaded into a Ni sample block, which was then placed in a two-part chamber, and heated under vacuum for 24 h to temperatures of at least 70 °C to ensure removal of adsorbed atmospheric moisture44. Further flushing and heating of the chamber using BrF5 ensured the reduction of retained adsorbed moisture, and then the released O2 was purified via liquid nitrogen ‘cold traps’43,44. The O isotope composition was measured using a Thermo Fisher MAT 253 dual-inlet mass spectrometer43,44 (see Supplementary Table 2 for results).

Statistical analyses run on the compositional data are described in Supplementary Section 1.2. Two-tailed Student t-tests assuming unequal variance were carried out on the electron-probe microanalysis data from AP 007 (Supplementary Tables 3 and 4).

Pb isotopes of apatite, plagioclase, olivine and pyroxene in the section (Supplementary Table 5) were measured using a CAMECA IMS 1280 ion microprobe at the NordSIMS facility in the Swedish Museum of Natural History. An Oregon Physics H201 RF plasma source generated an O2− Gaussian beam of about 2 nA (apatite) and 10 nA (plagioclase, olivine and pyroxene) with an impact energy of 23 kV, rastered over 5 × 5 µm2 (apatite) or 10 × 10 µm2 (plagioclase, olivine and pyroxene) during the analysis. Four Pb isotopes (204Pb, 206Pb, 207Pb and 208Pb) were collected simultaneously using low-noise multichannel ion counting detection, at a nominal mass resolution of 4,860 (M/ΔM). Isochrons were constructed from the reported U and Pb isotope values using the Isoplot 4.1 add-in for Microsoft Excel (Berkley Geochronology Center; Supplementary Fig. 5). The overall analytical and data process procedures followed an established protocol45.

Degassing modelling was carried out to contextualize the Cl and H isotopic compositions and abundances observed in AP 007. Rayleigh and closed system degassing modelling were carried out using the fractionation factors outlined in Supplementary Table 6. Supplementary Section 1.6 provides a detailed discussion on the effects of these degassing scenarios and the implication for the estimated H2O and Cl contents of the LMO at the time of FAN crystallization.

The Ce abundance of the parental melt of AP 007 was reconstructed using known partition coefficient values for Ce in apatite. This was then compared with various FAN subgroups to evaluate the possible subgroup with which AP 007 may be associated (Supplementary Table 7).