A dry lunar mantle reservoir for young mare basalts of Chang’e-5

The distribution of water in the Moon’s interior carries implications for the origin of the Moon1, the crystallization of the lunar magma ocean2 and the duration of lunar volcanism2. The Chang’e-5 mission returned some of the youngest mare basalt samples reported so far, dated at 2.0 billion years ago (Ga)3, from the northwestern Procellarum KREEP Terrane, providing a probe into the spatiotemporal evolution of lunar water. Here we report the water abundances and hydrogen isotope compositions of apatite and ilmenite-hosted melt inclusions from the Chang’e-5 basalts. We derive a maximum water abundance of 283 ± 22 μg g−1 and a deuterium/hydrogen ratio of (1.06 ± 0.25) × 10–4 for the parent magma. Accounting for low-degree partial melting of the depleted mantle followed by extensive magma fractional crystallization4, we estimate a maximum mantle water abundance of 1–5 μg g−1, suggesting that the Moon’s youngest volcanism was not driven by abundant water in its mantle source. Such a modest water content for the Chang’e-5 basalt mantle source region is at the low end of the range estimated from mare basalts that erupted from around 4.0 Ga to 2.8 Ga (refs. 5,6), suggesting that the mantle source of the Chang’e-5 basalts had become dehydrated by 2.0 Ga through previous melt extraction from the Procellarum KREEP Terrane mantle during prolonged volcanic activity.

anorthosites 23,24 .The estimates of water abundances for their mantle source regions span a wide range from ~0.3 to 200 g.g - 1 25 , suggesting that the Moon's interior is not as anhydrous as previously thought.However, many questions remain regarding the origin(s) and distribution of water in the Moon's interior 25,26 .The large variation in the estimated water abundances may be indicative of geographical and/or temporal heterogeneity in water abundance as these samples were collected from different regions and crystallised between ca.4.0 to 2.8 Ga 5,6 .Hence, additional sample collections of younger basalts from new regions can have critical implications to investigate the spatiotemporal evolution of water in the Moon.This large range of estimates could also be affected by the mixing of endogenous water with various exogenic water sources, i.e. asteroids, comets, and solar wind 19,26,27 , and/or by interplay between many processes, such as volatile degassing, partial melting, fractional crystallisation, impacting, mixing with potassium (K), rare earth elements (REE) and phosphorus (P) (KREEP)-rich components, and spallation 11,15,25,[27][28][29] .It is thus crucial to combine in situ analysis of water abundances and hydrogen isotope composition with detailed contextual petrographic information.
The Chang'E-5 (CE5) mission successfully returned 1.731 kg of lunar soil samples from young mare basalt units dated at ca. 1.2-2.0Ga using crater counting chronology 30,31 .These returned samples have now been precisely dated at 2030   million years ago (Ma) using the Pb-Pb isotope isochron technique 3 .The CE5 basalts are thus much younger than the youngest lunar basalt dated so far (ca.2.8 Ga 5 ).The young basalt unit is located in northwestern Oceanus Procellarum, on the northwestern edge of the Procellarum KREEP Terrane (PKT), which is far from all landing sites of the Apollo and Luna missions (Extended Data Fig. 1).The PKT region is also thought to have enhanced concentrations of two major radioactive heat-producing elements, uranium (U) and thorium (Th), and other incompatible elements.Water behaves as a typical incompatible element during magmatic processes and thus is expected to be enriched in the PKT as well.Hence, the CE5 basalts provide a unique opportunity to constrain the water inventory of a newly sampled region of the Moon's interior, providing crucial information to account for the prolonged activity of lunar magmatism.
We studied a total of 23 basalt clasts (0.2-1.5 mm in size) from two of the CE5 soil samples (CE5C0100YJFM00103, ~1g, CE5C0400YJFM00406, ~2g) (Extended Data Table 1).These basalt clasts exhibit variable textures including subophitic, poikilitic, and equigranular, similar to those observed for other basalt clasts in CE5 soil sample 3,4 , and are mainly composed of pyroxene and plagioclase with less abundant olivine and ilmenite (Figs. 1, S1 and S2).These basalt clasts are likely representative of various locations in the same lava flow, based on their identical mineral chemistry and geochemistry 4 and their well-defined Pb-Pb isochron 3 .The textures of ilmenite in the clasts indicate that it crystallised early from the melt and continued until the last stages of melt evolution (Fig. S1).Ilmenite-hosted melt inclusions range in size of ~4-50 m in diameter and display post-entrapment crystallisation textures (~0-40%) (Figs. 1 and     S1).Apatite is the main OH-bearing phase, and is F-rich than Cl-poor, similar to those from Apollo mare basalts (Fig. S6).It is an accessory phase, comprising less than 0.4 vol% modal abundance in the CE5 basalt clasts (Table S1).The apatite occurs as euhedral grains (mostly <10 m) mainly in the fine-grained interstitial materials, with a few euhedral crystals enclosed in the margins of pyroxene (Fig. 1) and FeO-rich olivine (Fig S2, details see Supporting Information).
Eight ilmenite-hosted melt inclusions and several apatite grains were located and identified from the studied CE5 basalt clasts (Extended Data Table 1 and Figs. 1, S1 and S2), and selected for in situ analysis.The water abundance and hydrogen isotope compositions of ilmenite-hosted melt inclusions, apatite, and clinopyroxene from these CE5 basalt clasts were measured using a Nano-scale Secondary Ion Mass Spectrometer (NanoSIMS 50L) instrument (see Methods).2), which overlap with apatite water abundances and D values measured in Apollo high-Ti and low-Ti basalts 11, 15-17, 28, 29, 32-35 .Three apatite analyses yielded lower water abundances (110  13 g.g -1 to 235  19 g.g -1 ; Extended Data Table 2), with corresponding D values indistinguishable from the majority of other analyses.Because apatite is the major water-bearing phase in CE5 basalts, a water abundance of ~8 ± 4 g.g -1 for the bulk CE5 basalts was calculated from the average water content of apatite and its modal abundance of 0.4 vol% (See Supporting Information).It should be noted that this water abundance is not the original water abundance in the CE5 basalts' parent magma before eruption, but represents the residual water abundance after magma degassing at the time of apatite crystallisation 14 .Furthermore, the apatite D values reflect the signature of the last residual melt after precipitation of most constituent minerals, and the observed large D-enrichment is likely the result of degassing of H-bearing species from the melt, mostly in the form of H 2 under the reducing conditions at the Moon 15 .
The ilmenite-hosted melt inclusions contain lower water abundances of 13  g.g -1 to 661   g.g -1 , with D values ranging from -332 ‰ to 869 ‰ after correcting for the effects of cosmic ray spallation (Fig. 3 and Extended Data Table 3).Cosmic ray spallation mainly produces deuterium, and can have a large effect on D values especially for the water-poor melt inclusions (<30 g.g -1 ) 36 .The cosmic ray exposure (CRE) ages determined for various Apollo lunar samples are mostly younger than ca.200 Ma 37 , but have not yet been measured for CE5 samples.We have modeled the spallation effects on D values of the melt inclusions, using CRE ages of 10, 50, 100 and 200 Ma (see Methods and Extended Data Fig. 3).Using CRE ages from 100 to 200 Ma, the D values yield noticeable over-correction as the resulting values are even more D-depleted than the presently accepted hydrogen isotope composition of the lunar mantle (Extended Data Fig. 4).On the Moon, Apollo regolith from a depth of ~9 mm is thought to overturn at least once in approximately 10 million years 38 , suggesting that it is reasonable to assume a CRE age of ca.50 Ma for the CE5 basalt clasts.With a 50 Ma CRE age correction, the melt inclusions with the lowest H 2 O abundances yield corrected D values of 200 ± 300‰ that overlap with the lowest D value measured for apatite.Importantly, this correction does not greatly affect the D values of water-rich melt inclusions nor those of apatite grains (Extended Data Fig. 4 and Tables 2 and 3).Moreover, spallation by cosmic rays has little effect on water abundances.After correction for spallation, the melt inclusions with δD 200 ‰ display a negative correlation between water abundances (13 ± 4 g.g -1 to 367 ± 29 g.g -1 ) and D values (-332 ± 182‰ to 202 ± 390‰), except for three analyses with higher D values (271 ± 113‰ to 869 ± 224‰) that overlap with the data for the water-poor apatite grains (Fig. 2 and Extended Data Table 3).These observations provide convincing evidence that ilmenite-hosted melt inclusions recorded the progressive evolution of melts undergoing degassing of H 2 , resulting in considerable D-enrichment during crystallisation of the CE5 basalts 15,39 .Diffusion out of the melt inclusions is another process by D/H ratios can be fractionated as reported for melt inclusions enclosed in olivine and pyroxene from Apollo basalts 11 .At present there is no constraints on diffusion rate of water in ilmenite-hosted melt inclusions.The lowest D value of ~-300 ‰ measured in ilmenite-hosted melt inclusions suggests little exchange of hydrogen isotopes with the D-enriched residual melt through diffusion.5.
The water-rich melt inclusions have the lowest D values (-330 ± 164‰), which is consistent with D estimates for the lunar mantle made from analysis of various types of lunar samples 11, 16, 18-20, 22, 24 .This similarity suggests that the melt with the lowest D was trapped in the early stages of magma crystallisation before substantial degassing of water in the form of H 2 (refs. 11,15,39 ). OnOn the other hand, the melt inclusions with higher D (>270 ‰) also contain substantial water and overlap with the water-poor apatite (Fig. 3).This observation can be explained by the late crystallisation of ilmenite, when more water was concentrated in the residual melt before degassing loss and apatite became saturated in the melt.
As discussed above, the most deuterium-depleted melt inclusions likely captured the basaltic parent magma without notable degassing loss of water in the form of H 2 .Hence, the highest water abundance (~370 g.g -1 ; Extended Data Table 3) of these melt inclusions can be referred to as the maximum water content of the basaltic magma, because a fraction of the constituent minerals could have precipitated before the earliest crystallised ilmenite.With further consideration of partial post-entrapped crystallisation of nominally anhydrous minerals (Fig. 2), which enhanced the water abundance of the glassy domain analyzed by the ion probe, the maximum water abundance of the parent magma of CE5 basalts could be to some extent lower than ~370 g.g -1 .On the other hand, the water abundance of the parent magma can also be estimated from a water content of ~8 g.g -1 for the bulk CE5 basalts through calibration for degassing loss of 98-99% water in the form of H 2 based on the accompanying D increasing from ~-300‰ to the average of 578 208‰ (1) (Extended Data Table 2, details see Supporting Information).This yields an estimate for the water abundance of the parent magma of 380-760 g.g -1 , consistent with the highest melt inclusion water abundance.We thus use the highest melt inclusion water abundance (367 29 g.g -1 ; Extended Data Table 3) as the maximum water abundance estimate for the parent basaltic magma.
The parent magma of CE5 basalts were derived from a depleted lunar mantle source not associated with a KREEP-component, based on its low initial  value (680 20) ( 238 U/ 204 Pb) 3 , low initial 87 Sr/ 76 Sr ratio (0.69934 to 0.69986) and high positive  Nd (t) (7.9 to 9.3) 4 .The elevated abundances of REE and Th, and high FeO and moderate TiO 2 concentrations of CE5 basalts match a model of low degrees (2-3%) of partial melting followed by 43-78% fractional crystallisation 4 .Accordingly , the maximum water concentration in the lunar mantle source beneath the Chang'E-5 landing site can be estimated at 2-6 g.g -1 , corresponding to a maximum water abundance of ~370 g.g -1 in the derived basaltic magma (See Supporting Information).
In general, our analyses of apatite and melt inclusions outline the evolution of CE5 basalts, which can be divided into three stages.In Stage 1, the mantle source region underneath the PKT region with ~2-6 g.g -1 water experienced a low degree (2-3%) of partial melting followed by a moderate-to-high degree (43-78%) of fractional crystallisation 4 , generating a basaltic magma with ~370 g.g -1 water and a D of ~-300‰.This maximum water abundance, yielding our best estimate for the hydrogen isotopic composition of water in the parent magma, was recorded in the melt inclusions captured by the earliest-formed ilmenite analysed here.In Stage 2, H 2 degassing from the parent magma occurred during its ascent to shallower depths and on the surface of the Moon, and was accompanied by the crystallisation of ilmenite that entrapped melts at various stages of evolution.Extensive H 2 degassing 15,40 could have occurred in the reduced lunar environment 41,42 , resulting in extreme D/H fractionation from ~-300‰ up to ~300‰.In Stage 3, apatite crystallised from the residual melts that became enriched in water, halogens, and other incompatible species, after most nominally anhydrous silicates and ilmenite had formed.
The maximum water abundance of 2-6 g.g -1 estimated for the mantle source of CE5 basalts appears at the lower end of the mantle water abundance estimates derived from Apollo basalts and lunar meteorites 9-11, 15, 25 (Fig. 3).This could have important implications for understanding late volcanism on the Moon.First, such a dry mantle source for CE5 basalts excludes the possibility that a high abundance of water in the lunar mantle reservoir could be one of the main causes for the prolonged volcanic activity in this part of the PKT.Second, our observations indicate that the water abundance in the Moon's interior may have to some extent decreased from 4.0-2.8Ga to 2.0 Ga (Fig. 3).This systematic loss of water over time could be a result of prolonged magmatic activity in the PKT, probably through multiple water-bearing melt extractions episodes from the PKT mantle reservoir during extended volcanic activity.In the northwestern PKT region, in close proximity to the CE5 landing site, up to 10 basaltic units ranging in age from 3.7 to 1.2 Ga have been identified 43 , although it is difficult to be certain that all these units were derived from the same mantle source region.Nevertheless, such a dehydration partial melting process has also been observed in the Earth's mantle 44,45 .
Alternatively, the wide range of estimates for the water abundance in the mantle source regions of all studied lunar basaltic products may reflect a heterogeneous distribution of water in the Moon's interior, and/or possible contamination of some volcanic products by KREEP-components during either magma transport or in their mantle source regions during convective overturn of the lunar magma ocean 46,47 .However, CE5 basalts have not been contaminated by KREEP-components, and were derived from a depleted mantle source 4 .Our estimate of the mantle water abundance based on CE5 basalts in the PKT region is thus a surprising and critical regional constraint on the distribution of water in the Moon's interior.
The parent magma of CE5 basalts contained ~370 g.g -1 water, which is roughly comparable to but on the lower side of estimates for Apollo basalts that crystallised from ca. 4 to 2.8 Ga 11,15,25 .
Combining such water abundances with petrological evidence for low degrees of partial melting suggest that the mantle source region of CE5 was relatively depleted in water compared to the source regions of Apollo mare basalts.Additionally, the mantle source of CE5 basalts is also depleted in the heat-producing elements U, Th, and K, relative to the bulk silicate Moon 4 .Therefore, it remains an enigma to explain how mare basaltic volcanism was sustained as late as 2.0 Ga on the cooling Moon as the lunar interior should have been relatively cold by then.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at online version of the paper.

Sample preparation
Two Chang'E-5 (CE5) lunar soils (CE5C0100YJFM00103, ~1g, CE5C0400YJFM00406, ~2g) allocated by the China National Space Administration were used in this study.Both of them were scooped by the robotic arm of the CE5 lander and separated into different packages in the ultraclean room at the extraterrestrial sample curation center of the National Astronomical Observatories, Chinese Academy of Sciences.Approximately 240 soil fragments with grain sizes varying from ~100 m to ~ 1 mm were sieved and hand-picked under a binocular microscope in the ultraclean room at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS).Then, about 2/3 of the picked grains were prepared as 8 Sn-Bi metal-alloy mounts following the method of Zhang et al. ( 2018) 48 and the other 1/3 was mounted in epoxy and prepared as 3 polished thin sections.The polished metal mounts and thin sections were cleaned using ultrapure water and anhydrous ethanol prior to drying at 70 ℃ in a baking oven overnight.
The details of apatite and ilmenite-hosted melt inclusions from 23 CE5 basalt clasts and fragments are summarised in Extended Data Table 1.

Petrographic observations and elemental mapping were carried out using field emission scanning e l e c t r o n m i c r o s c o p e s ( F E -S E M ) u s i n g F E I N o v a N a n o S E M 4 5 0 a n d T h ermofisher Apreo
instruments at the IGGCAS, using electron beam currents of 2 to 3.2 nA and an acceleration voltage of 15 kV.Energy dispersive spectroscopy (EDS) X-ray maps were collected for each basaltic clasts to quickly locate P-bearing phases.The phosphates were then observed at higher magnification in back-scattered electron (BSE) images.The modal abundance of apatite from various CE5 basalt clasts were counted by the exposed surface areas (Table S1).The prepared sections were initially coated with Au to identify apatite and melt inclusions for in situ NanoSIMS measurement of water content and hydrogen isotopes.Another round of SEM observation was carried out after NanoSIMS measurement to confirm the positions of the NanoSIMS spots.

Electron probe microanalysis
We used a JEOL JXA-8100 electron probe micro-analyzer (EPMA) at the IGGCAS to quantify the major and minor elemental abundances in phosphates, melt inclusions in ilmenite, and associated mafic minerals (i.e.clinopyroxene, olivine, plagioclase, and ilmenite).The samples were coated with carbon.The operating accelerating voltage was 15 kV and the beam current was 20 nA.The EPMA analyses were carried out after the NanoSIMS measurements in order to avoid possible H loss due to bombardment by the electron beam 33 .The EPMA standards were natural albite (Na and Al), bustamite (Mn), diopside (Ca, Si, and Mg), apatite (P), K-feldspar (K), tugtupite (Cl), synthetic fluorite (F), rutile (Ti), Fe 2 O 3 (Fe), V 2 O 5 (V), NiO (Ni), and Cr 2 O 3 (Cr).Sodium, K, F, and Cl were first measured in order to minimise possible loss of volatiles by electron beam irradiation.The detection limits were (1) 0.01 wt% for Cl and S, 0.02 wt% for Na, Mg, Al, Cr, K, Si, Mn, Ca and Fe, 0.03 wt% for F, Ba, Ni and Ti, 0.04 wt% for P. A program based on the ZAF procedure was used for data correction.The EPMA data obtained for apatite, melt inclusions in ilmenite, and the coexisting silicates are listed in Table S2.

Apatite and melt inclusions
The hydrogen isotopes and water content of apatite and melt inclusions enclosed in ilmenite from the CE5 basaltic clasts were measured with a CAMECA NanoSIMS 50L at IGGCAS.The samples were coated with Au, were loaded in sample holders together with the standards, and were baked overnight at ~60 ºC in the NanoSIMS airlock.The holders were then stored in the NanoSIMS sample chamber to improve the vacuum quality and minimise the H background [49][50][51] .
The vacuum pressure in the analysis chamber was 2.8×10 -10 to 3.0×10 -10 mbar during analysis.
All data are reported with their 2 uncertainties that include reproducibility of D/H measurements on the reference materials, uncertainty of H 2 O background subtraction, and internal precision on each analysis (Extended Data Tables 2 and 3 and Table S4).The raw measured D/H ratios were corrected for the background, followed by correction for IMF.

Clinopyroxene
The water abundance of clinopyroxene from the CE5 basaltic clasts was measured with the CAMECA NanoSIMS 50L using an identical instrument setup to that described above.We used a higher Cs + primary beam current of 7 nA to improve the 1 H -counts on clinopyroxene.Each 25 μm × 25 μm analysis areas was pre-sputtered for ~ 2 mins with the same analytical beam current to remove surface coating and potential contaminations.The secondary ion signals from the central 7 μm × 7 μm areas were counted with 50% blanking of outermost regions.San Carlos olivine (H 2 O = 1.4 g.g -1 , ref. 55 ) was used for determining instrumental background of H.The analytical results are listed in Extended Data Table 4.

Correction of water abundances and D/H ratios for spallation effects
The measured D/H ratios have also been corrected for the potential effects of spallation by cosmic-ray, using a D production rate of 2.17 × 10 -12 mol D/g/Ma 56 for melt inclusions and 9.20 × 10 -13 mol D/g/Ma 57 for apatite.The correction errors induced by D spallation are around 50% on D and negligible on water content 21 .The cosmic ray exposure (CRE) ages determined for most Apollo samples are less than ~200 Ma 37 .Because no CRE age is yet available for the Chang'E-5 basaltic clasts, we modeled the effects of corrections for CRE ages of 10, 50, 100 and 200 Ma (Extended Data Table 3 and Fig. 4).The corrected D values for the melt inclusions with low water abundances appear to be over-corrected for CRE ages of 100 and 200 Ma, as indicated by unusually low D values.We thus decided to correct D values using a CRE age of 50 Ma, slightly older than that of Apollo lithic fragments (1-24 Ma) 58 , for which the corrected δD values of the melt inclusions with low H 2 O are comparable to the lowest δD values measured in apatite (~300‰), as apatite crystallization postdated that of ilmenite in which the melt inclusions are hosted.

Degassing modeling
The hydrogen isotope fractionation during volatile loss into a vacuum is given by  2 = 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 = R 0 ×f (-1) , where R 0 and R are the initial and final D/H ratios for a fraction f of remaining hydrogen 39 .Degassing of H 2 (M1 = 2 for H 2 and M2 = 3 for HD) yields an  value of ~0.8165, and degassing of H 2 O (M1 = 18 for H 2 O and M2 = 19 for HDO) yields an  value of ~0.9733 39 (Extended Data Fig. 4).

Fig. 2 |Fig. 3 |
Fig. 2 | Water abundance and D of apatite and ilmenite-hosted melt inclusions from CE5 basalts.The majority of melt inclusions display a negative correlation between the water abundance and D values, except for a few melt inclusions with higher δD values plotting close to the range of apatite.The dotted lines indicate a three-stage evolutionary path, starting with ~2-3% partial melting of the mantle source of CE5 basalts, followed by ~43-78% fractional crystallisation (Stage 1), H 2 degassing from the basaltic melts accompanied with D-enrichment (Stage 2), and crystallisation of apatite from the residual melts, possibly accompanied by further H 2 degassing (Stage 3).Apatite and melt inclusion data from Apollo samples (TableS5) are shown for comparison.The CE5 data have been corrected for a nominal cosmic-ray exposure of

Table 2 | Water abundance and hydrogen isotopes of CE5 apatite 563 CRE age = 0 Ma CRE age = 50 Ma *
D production rate of 9.20 × 10 -13 mol D/g/Ma 57 was used for cosmogenic spallation effects.CRE age: cosmic ray 564 exposure age. *

Table 5 | Summary of the water abundances estimated for the lunar mantle 573 source regions of basaltic products formed between ca. 4-2 Ga. 574 Sample name Age (Ga) * References Phase # H 2 O min (g.g -1 ) H 2 O max (g.g -1 ) References
ND=No data.575#MI-Melt inclusions, Ap-Apatite, GB-Glass bead.They were used for estimating water abundance in the lunar 576 mantle.577 *