Water abundance in the lunar farside mantle

Sample preparation

Five grams of scooped Chang’e-6 (CE6) lunar soil (CE6C0200YJFM001) allocated by the China National Space Administration (CNSA) were used in this study. The allocated samples were sieved into two dividers in an ultraclean room at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS) in Beijing, China. One separate was >355 μm and the other was <355 μm. Eighty-eight particles >355 μm were selected and prepared into 7 Sn–Bi polished mounts following the protocol in ref. ⁴⁶. The mounts were labelled as CE6C02,MGP01–MGP07 (Extended Data Table 1 and Supplementary Table 1). Another 490 particles with grain sizes ranging from 100 μm to 350 μm were hand-picked from the sieved soils. These particles were placed in a tunnel (depth about 300 μm, width about 400 μm and length about 1.8 mm) of silica glass cut by low-speed diamond saw (Supplementary Fig. 5). Then the tunnel was filled up with epoxy. These particles were finally prepared into 14 polished thick sections labelled as CE6C02,WGP01–WGP14 (Extended Data Table 1 and Supplementary Table 1). The prepared sections were cleaned up using anhydrous ethanol before drying at 50 °C in a baking oven. All the mounts were coated with carbon for petrographic observation, chemical analysis and isotopic analysis. Statistically, the particles used in this study consist of 163 mare basalt fragments (28%), 97 monomineralic fragments (17%), 50 agglutinates (9%), 101 breccias (18%), 52 highlands fragments (9%), 76 impact melt fragments (13%) and 39 glass beads and fragments (7%) (Extended Data Table 1). Representative petrography of these components in the CE6 lunar soils is shown in Supplementary Fig. 6.

Scanning electron microscopy

Petrographic observations and elemental mapping were carried out using field emission scanning electron microscopes using the FEI Nova NanoSEM 450 and the Thermo Fisher Apreo instruments at IGGCAS, using electron beam currents of 2–3.2 nA and an acceleration voltage of 15 kV. Energy dispersive spectroscopy (EDS) X-ray maps were collected for each basaltic clast to locate P-bearing phases, identical to the protocol used in the study of CE5 lunar soils^9,47,48. The phosphates were then observed at higher magnification in back-scattered electron (BSE) images. The modal abundance of apatite from various CE6 basalt fragments was counted by the exposed surface areas (Supplementary Table 2). The apatite in the studied samples has a modal abundance of 0.067 vol% calculated from 163 basalt fragments using Image J software (Supplementary Table 2 and Supplementary Fig. 7).

Electron probe microanalysis

The major and minor elemental abundances in phosphates, glassy melt inclusions and associated minerals (olivine, pyroxene, plagioclase and ilmenite) were measured using a JEOL JXA-8100 electron probe microanalyser at IGGCAS after the NanoSIMS measurements to avoid possible H loss due to bombardment by the electron beam⁴⁹. For silicate, ilmenite and apatite analysis, the operating accelerating voltage was 15 kV and the beam current was 20 nA. The standards used were albite for Na and Al, diopside for Si, Mg and Ca, magnetite for Fe, bustamite for Mn, sanidine for K, rutile for Ti, Cr₂O₃ for Cr, NiO for Ni, apatite for P, tugtupite for Cl and synthetic fluorite for F. The F Kα X-rays were counted with a LDE1 crystal to avoid drift with time (ref. ⁵⁰). F and Cl were measured first to minimize the loss of volatiles by electron beam irradiation. The counting time for F was 20 s. The detection limits (1σ) were 0.01 wt% for K₂O, P, F, and Cl; 0.02 wt% for Cr₂O₃, Al₂O₃, MnO, CaO, FeO, TiO₂ and NiO; 0.03 wt% for SiO₂ and Na₂O. For analysis of melt inclusion glass, the accelerating voltage was 15 kV and the beam current was 10 nA, with a beam size of 5 μm in diameter to obtain the composition of the glass within a melt inclusion. A natural obsidian glass was used as a monitor. The detection limits for MI were 0.02 wt% for K₂O, CaO, FeO and NiO; 0.03 wt% for SiO₂, MgO, Cr₂O₃, Al₂O₃, MnO and TiO₂; and 0.04 wt% for Na₂O. The EPMA data obtained for apatite, melt inclusions and the coexisting silicates are listed in Supplementary Table 3.

In situ water abundance and hydrogen isotope analysis

A CAMECA NanoSIMS 50L at IGGCAS was used to measure the hydrogen isotopes and water abundance of apatite and melt inclusions from the CE6 basaltic fragments. The samples were loaded in sample holders together with the standards, and were baked overnight at about 60 °C in the airlock chamber. The holders were then stored in the NanoSIMS vessel chamber to improve the vacuum quality and minimize the H background^51,52. The vacuum pressure in the analysis chamber was 2.1 × 10⁻¹⁰ torr to 3.0 × 10⁻¹⁰ torr during analysis facilitated by a liquid nitrogen cold trap⁵³. Each 15 μm × 15 μm analysis area was pre-sputtered for 2 min with a Cs⁺ ion beam current of 5 nA to remove the surface coating and potential contamination. A 4 μm × 4 μm region of interest on targets was selected for analysis. The secondary anions ¹H⁻, ²D⁻, ¹²C⁻ and ¹⁸O⁻ were simultaneously counted by electron multipliers from the central 3 μm × 3 μm areas using the electronic gate technique of NanoSIMS (about 50% blanking). The ion beam current (Fco) was around 0.5 nA for analysis with a beam size of about 500 nm in diameter. The charging effect on the surface of the samples was compensated by an electron gun during analysis. More details are provided in refs. ^52,54.

A series of apatite and glass standards were used for calibrating the water abundances and hydrogen isotopes of the samples: Durango apatite (H₂O = 0.0478 wt% and δD = −120 ± 5‰) (refs. ^16,55) and Kovdor apatite (H₂O = 0.98 ± 0.07 wt% and δD = −66 ± 21‰) (ref. ⁵⁶), the SWIFT MORB glass (H₂O = 0.258 wt% and δD = −73 ± 2‰), and two basaltic glasses, ALV-519-4-1 (H₂O = 0.17 wt%) (ref. ⁵¹) and ALV-1833-11 (H₂O = 1.2 wt%) (ref. ⁵¹) (Supplementary Table 4). The water content calibration lines established on apatite and glass standards were used for calibrating the samples (Extended Data Fig. 7). Instrument mass fractionation (IMF) on hydrogen isotopic composition are the same between apatite and silicate glass within analytical uncertainties (Extended Data Fig. 7). Instrumental mass fractionation on H isotopic compositions of both apatite and melt inclusions were conducted using the Kovdor apatite standard and monitored by analysing both the Durango apatite and SWIFT MORB glass standards during the whole analytical session (Extended Data Fig. 7). Hydrogen isotopic compositions are given using the delta notation, δD = ((D/H)_sample/(D/H)_SMOW) − 1) × 1,000‰, where SMOW is the standard mean ocean water with a D/H ratio of 1.5576 × 10⁻⁴ (ref. ⁵⁷).

The instrument H background was monitored by standards of San Carlos olivine⁵⁸ (H₂O = 1.4 μg g⁻¹) and a synthetic anhydrous quartz glass Suprasil 3001 (personal communication with Erik Hauri). The H background correction of ref. ⁵⁹ was used, which follows the relationship: H/O_bg = (H_counts − H_bg)/O_counts and D/H_measured = (1 − f) × D/H_true + f × D/H_bg, where f is the proportion of H emitted from the instrumental background. Owing to the inadequate baking and storing time in the instrument, the first 2 days have a slightly higher instrument background at 26–53 μg g⁻¹ (D/H_bg was (1.40 ± 0.25) × 10⁻⁴ to (1.13 ± 0.41) × 10⁻⁴ and H_bg was 1546 ± 266 counts per second (cps) to 786 ± 81 cps (Supplementary Table 4). The instrument H₂O background was about 10 μg g⁻¹ (D/H_bg was (1.19 ± 0.26) × 10⁻⁴ to (1.14 ± 0.49) × 10⁻⁴ and H_bg was 374 ± 170 cps to 311 ± 37 cps) in the following days (Supplementary Table 4). After background subtraction, the water abundances of apatite and melt inclusions were quantified by the slope of the calibration line (Extended Data Fig. 7), which was determined by measuring apatite and glass standards.

All data are reported with their 2σ uncertainties that include reproducibility of D/H measurements on the reference materials, uncertainty of H₂O background subtraction, internal precision on each analysis and uncertainty of corrections of spallation effects (Supplementary Table 5). The raw measured D/H ratios were corrected for the background, followed by IMF and spallation in sequence (Supplementary Table 5).

Correction of water abundances and D/H ratios for spallation effects

The measured D/H ratios have been corrected for the potential effects of spallation by CRE, using a D production rate of 2.17 × 10⁻¹² mol D/g/Ma (ref. ⁶⁰) for melt inclusions and 9.20 × 10⁻¹³ mol D/g/Ma (ref. ⁶¹) for apatite, where Ma signifies millions of years ago. The correction errors induced by D spallation are around 50% for δD and negligible for water content⁴². The CRE age for the CE6 lunar soils derives from a cosmogenic ²¹Ne age of 146 ± 27 Ma (T₂₁; Zhang, X. et al., manuscript in preparation), which was converted to an exposure age (T_CRE) of 108 ± 20 Ma using an empirical relationship (T₂₁ = 1.35 × T_CRE) between T₂₁ and T_CRE based on the study of Apollo samples⁶⁰. An uncertainty of 50% (2σ) in D and H production rates raised by the spallation process was combined to calculate its contribution to the δD errors (ref. ⁴²; Supplementary Table 5).

Petrography of CE6 mare basalt

Approximately 28% of the lithic clasts in a total of 578 fragments studied in this work are mare basalt fragments (Supplementary Table 1), which display subophitic, poikilitic and porphyritic textures (Supplementary Fig. 4), similar to those in Zhang, Q. W. L. et al. (manuscript in preparation) and Zhou, Q. et al. (manuscript in preparation). Lath-shaped plagioclase in the basalts exhibits sharp boundaries, whereas pyroxene is anhedral with compositional zonation in BSE images (Supplementary Fig. 4). These basalt fragments are mainly composed of pyroxene, plagioclase and ilmenite with minor Si–K-rich mesostasis, fayalite and troilite, and trace apatite, merrillite, tranquillityite and baddeleyite (Supplementary Fig. 4). Olivine is very rare and absent in most CE6 mare basalt except two basalt fragments (CE6C02,WGP09,G17 and CE6C02,WGP14,G3), as shown in this study and other parallel studies (Zhang, Q. W. L. et al., manuscript in preparation; Zhou, Q. et al., manuscript in preparation). These two olivine grains have several melt inclusions with diameters of about 5–50 μm (Extended Data Fig. 1 and Supplementary Fig. 1). Melt inclusions have experienced post-entrapment crystallization (31–52%) (Extended Data Fig. 1, Extended Data Table 2 and Supplementary Fig. 1). Ilmenite intergrows with plagioclase but has a smaller size and irregular edge, occurring as laths partially enclosed by pyroxene (Supplementary Fig. 1). Nine melt inclusions in ilmenite were identified in 7 CE6 mare basalt fragments with diameters mostly less than 30 μm (Supplementary Fig. 2). Ilmenite-hosted melt inclusions have experienced post-entrapment crystallization (0–31%) (Extended Data Table 2 and Supplementary Fig. 2). Phosphorus- and zirconium-bearing minerals mainly occur in the fine-grained interstitial areas. Several apatite grains are observable in the margins of plagioclase and pyroxene (Supplementary Fig. 3). Most apatite grains have euhedral shape and rare ones are lath-shaped (Supplementary Fig. 3). CE6 mare basalts have notably less apatite (0.067 vol%; Supplementary Table 2) than CE5 basalts (0.4 vol%) (ref. ⁹). Meanwhile, apatite grains in CE6 mare basalt fragments have smaller sizes (mostly <5 μm; Supplementary Fig. 3).

Mineral chemistry of CE6 mare basalt

Plagioclase from CE6 basalt clasts is relatively homogeneous with a composition of An_75.2–92.5Ab_5.3–20.1Or_0.1–5.0 (Extended Data Fig. 3 and Supplementary Table 3). The pyroxene displays compositional zonation from Mg-rich cores (En_42.1–56.1Wo_9.1–38.1Fs_20.7–49.5) to Fe-rich rims (En_0.3–25.9Wo_11.6–47.3Fs_47.8–87.6) (Extended Data Fig. 2 and Supplementary Table 3). Ilmenite shows no chemical difference between various basalt fragments, containing about 52.9 wt% TiO₂ and about 46.2 wt% FeO, with minor other elements (Supplementary Table 3). Most CE6 basalt fragments contain trace fayalite (Mg# < 1). Two relatively large olivine grains having melt inclusions identified in this study are quite homogeneous in chemistry with average Mg# of 49.3 ± 2.7 and 59.1 ± 2.2 (1σ, n = 5; Extended Data Fig. 3 and Supplementary Table 3). Interstitial mesostasis is SiO₂-rich (65.2–76.9 wt%) and K₂O-rich (2.40–7.65 wt%; Supplementary Table 3). Olivine- and ilmenite-hosted melt inclusions have 38.0–79.9 wt% SiO₂, 0.61–7.13 wt% TiO₂, 5.43–17.5 wt% Al₂O₃, 0.68–15.6 wt% CaO and 1.44–33.1 wt% FeO (Extended Data Fig. 5 and Extended Data Table 3). Concentrations of MgO in olivine-hosted melt inclusions range from 3.76 wt% to 16.8 wt%, whereas those of melt inclusions in ilmenite are less than 0.30 wt%. Melt inclusions in olivine have lower SiO₂ contents (38.0–60.6 wt%) compared with ilmenite-hosted ones (44.5–79.9 wt%), indicating an earlier entrapment for the former. Apatite grains in this study contain 2.54–4.08 wt% F and 0.09–0.43 wt% Cl (Supplementary Table 3), close to the fluorapatite end-member in the F–Cl–OH ternary diagram (Extended Data Fig. 4). The evaluated OH contents of apatite range from 0 wt% to 0.22 wt%, assuming the volatile site only contains F, Cl and OH. Apatite grains in CE6 mare basalt fragments are usually less than 5 μm, rendering it unavoidable to sample some surrounding minerals during EPMA and yielding an EPMA total usually less than 98 wt% (Supplementary Table 3).

Estimate of water abundance for the parent melt of CE6 mare basalt

Melt inclusions

Melt inclusions are trapped melt in the minerals during crystallization of the parent melt. Three melt inclusions were identified in olivine from two CE6 mare basalt fragments (CE6C02,WGP09,G17 and CE6C02,WGP14,G3; Extended Data Table 2 and Supplementary Fig. 1). These olivine-hosted melt inclusions have a water abundance of 28–46 μg g⁻¹, slightly higher than the instrument background, with large variation in δD values (−358 ± 610‰ to 100 ± 474‰) after correction of spallation effects (Extended Data Table 2). The correction of spallation can markedly affect δD for those samples with water abundances <50 μg g⁻¹ (ref. ⁴²). The hydrogen isotopic compositions for analyses with H₂O < 20 μg g⁻¹ may not be reliable because of the extremely low measured D counts¹. Thus, those spots with H₂O < 20 μg g⁻¹ (that is, two times the instrument background) are not discussed or used in deriving the water abundance for the parent magma and for the mantle source (Extended Data Table 2). An average δD of −96 ± 230‰ (1 s.d.) in olivine-hosted MIs is comparable to some early trapped melt hosted by ilmenite, which have water abundances of 153–184 μg g⁻¹ and δD values of −183‰ to −140‰ with an average value of −158 ± 23‰ (1 s.d.; Fig. 2 and Extended Data Table 2). The low-δD nature of the melt inclusions hosted by olivine and by early-formed ilmenite are comparable to the mantle δD value (−200 ± 200‰) recommended in previous studies^9,23,40,62, indicating that these MIs plausibly represent the parent melt in the terms of water abundance and hydrogen isotope signature. This prediction is also supported by the high MgO content in olivine-hosted melt inclusions compared with previous work^23,30,32.

The host olivine grains of melt inclusions have forsterite content of 45.0–61.6 atom% (Supplementary Table 3), indicative of early precipitation phases crystallized from the parent melt of the CE6 mare basalt. In comparison, CE6 mare basalt has a slightly lower bulk Ti abundance (4.2 wt%) (Zhou, Q. et al., manuscript in preparation) than that of CE5 (5.3 wt%) (ref. ⁶³), which would result in the late precipitation of ilmenite from the parent melt as shown by modelling²⁸. A gap in water abundance between olivine- and ilmenite-hosted melt inclusions could be the result of the crystallization sequence (Fig. 2). It is also noted that some ilmenite-hosted melt inclusions have comparable values to apatite, suggesting a late entrapment probably accompanying the onset of apatite at a quite late crystallization stage. Therefore, we provide an illustration diagram depicting the potential evolutionary pathways of magmatic water for CE6 mare basalt (Fig. 3). Finally, the δD-poor spots of olivine-hosted melt inclusions are regarded as a target for estimating the water abundance of the parent melt. We estimate a water abundance of 15–22 μg g⁻¹ for the parent melt of the CE6 mare basalt based on the olivine-hosted melt inclusions (Extended Data Table 2). Excluding the notable high δD ilmenite-hosted melt inclusions probably formed accompanying apatite in the late crystallization stages, we estimate the parent melt of the CE6 mare basalt as having a δD value of −123 ± 167‰ (1 s.d.) (Extended Data Table 2).

Apatite

Apatite is the major OH-bearing phase in lunar igneous rocks¹. A modal abundance and an average water abundance of apatite were used to quantify water in bulk CE6 basalts. Statistically assessed using the surface areas of apatite grains in all basaltic clast, the modal abundance of apatite in CE6 basalts is found to be less than 0.07 vol% (Supplementary Table 2). The average water content and δD value of the CE6 apatite measured by NanoSIMS 50L are 1,511 ± 748 μg g⁻¹ and 826 ± 94‰ (1 s.d., n = 16), respectively (Extended Data Table 3). Hence, the water abundance of the bulk CE6 basalt is 1.1 ± 0.5 μg g⁻¹ (1 s.d.). Apatite is a late-crystallizing mineral in mare basalts; its notably higher δD values (826 ± 94‰) than the low-δD (−200 ± 200‰) nature of lunar mantle^9,23,40,62 and low δD of CE6 melt inclusions indicate strong degassing and loss of water in the form of H₂ (refs. ^24,25). The hydrogen isotope fractionation during volatile loss into a vacuum is given by α² = M₁/M₂, where M₁ and M₂ are the masses of the volatile phase isotopologues. The change of the isotopic composition of H during volatile loss by Rayleigh fractionation is given by R = R₀ × f^(α−1), where R₀ and R are the initial and final D/H ratios for a fraction f of remaining hydrogen. Degassing of H₂ (M₁ = 2 for H₂ and M₂ = 3 for HD) yields an α value of about 0.8165 (ref. ²⁵). Degassing modelling indicates that the crystallization of apatite grains in CE6 mare basalt started at the time when 97.99–99.35% of initial water in the parent melt has been degassed. After degassing loss correction, the water abundance of the parent melt is about 55 ± 25 μg g⁻¹ to 168 ± 77 μg g⁻¹ (1 s.d.).

In another viable approach^1,26,27, the authors have developed apatite-based melt hygrometry methodologies, to estimate parent magma water abundance using apatite. These methods are designed to estimate parent melt water abundances from apatite regardless of how much H-degassing has occurred after apatite crystallization. The measured F and H₂O abundance in apatite from each fragment was used to derive the parental melt H₂O, using a 0.07 vol% modal abundance of apatite (Supplementary Table 2). The results yield H₂O abundance of 59–99 μg g⁻¹ for the parental magma (Supplementary Table 6), plotting within the approximately 55–168 μg g⁻¹ range of the other method presented earlier. Considering the early entrapment of olivine-hosted melt inclusions¹⁷, these targets were used to estimate the water abundance for the mantle source.