Introduction

Water is critical for life, plays a fundamental role in the evolution of planets1, moons2,3, and asteroids4, and is one of the most precious resources for planetary utilisation5,6. One of the biggest advances in lunar and planetary sciences in the past two decades was the discovery of water on the Moon7,8,9,10,11,12,13,14,15,16,17,18, which has inspired the potential for in situ water utilisation in future lunar base construction (Artemis) and lunar research stations (CE7 and post-CE missions)19. Planetary water, generally defined, includes H/H2, OH, and H2O in various forms9,10,20. Lacking an atmosphere, the Moon on its surface is a typical airless celestial body, characterised by a high vacuum (10−12–10−10 torr; ref. 21.), strong irradiation22, large diurnal temperature variations (about – 150 to + 150 °C, ref. 23.), and numerous impact craters24. Therefore, the existence of water on the Moon under such harsh environmental conditions is surprising and can inform our understanding of the presence of water on other airless bodies in the Solar System25,26.

Water on the Moon has been detected to have concentrations ranging from ~10–1000 μg.g−1 (ref. 7), which generally increase from the equator to the polar regions and exhibits temporal variations due to fluctuating surface temperature as inferred from the near-infrared spectroscopy observations of the Chandrayaan-1 probe27,28,29. Larger amounts of water (up to 56 ± 29 mg.g−1), probably in the form of water ice, were detected in the permanently shadowed region within Cabeus crater located close to the South Pole as predicted by impact experiments8 and later confirmed by neutron spectroscopy observations15. Molecular water was also detected on the lunar midlatitude surface9,17. The spatiotemporal variations of lunar surface water27,28,29 and the detection of exospheric water release to space30 suggest that surface water on the Moon displays dynamic retention, release, and replenishment processes, inspiring hypothesis of a lunar surface water cycle30,31.

A hydrated layer at depth in lunar soils has been proposed on the basis of such a hypothesised lunar surface water cycle30. Impact glass beads (IGBs)—a common component in lunar soils produced by asteroid impacts32—collected by the Chang’e-5 (CE5) mission from the lunar nearside have been demonstrated to contain abundant solar wind-derived water (SW-H2O)16,17. The CE5 IGBs have SW-H2O hydration profiles typically up to 40 μm deep, suggesting that IGBs in lunar soils may be one of the most important components sustaining the lunar surface water cycle16. It is currently unknown, however, if IGBs in the lunar soils from different geographic locations on the Moon record similar or different SW-H2O hydration features. Such a geographic and geologic context for IGBs is critical to understanding their potential role in the lunar surface water cycle and deepen our knowledge of the origin, storage, and distribution of water on the Moon and other airless bodies. Thus, newly discovered IGBs in the first lunar soils from the farside of the Moon recently collected by the Chang’e-6 (CE6) mission, located thousands of kilometers away from previous collections, can be used to investigate the homogeneity or heterogeneity of the distribution of surface water on the Moon.

Here, we perform systematic analyses of water abundances, hydrogen isotopes, and chemical compositions of CE6 IGBs to unravel the distribution and mechanism of SW-H2O storage in lunar farside soils, finding that SW-H2O storage of IGBs exhibits a dichotomy distribution in the lunar soils.

Results and discussion

Distribution of water in CE6 impact glass beads

Among the 263 hand-picked particles from two scooped lunar soils (CE6C0200YJFM001, ~5 g and CE6C0300YJFM002, ~6 g), a total of 77 grains were identified as impact glass beads (IGBs) (Supplementary Fig. 1 and Data 1). About 36% (28) of the CE6 IGBs exhibiting smooth surfaces without exposed Fe-Ni metals, mineral clasts, and bubbles (Fig. 1 and Supplementary Figs. 1, 2, 3) were selected for in situ ion probe measurements of water abundances and hydrogen isotopic compositions. The electron microprobe results suggest that each CE6 IGB has homogeneous chemical compositions (hereafter referred to as Homo-IGBs; Supplementary Data 2). Most CE6 Homo-IGBs plot in the classification diagram of lunar glasses in the highlands (15) to mare (12) domains, except for WGP09,G31, which plots in the picritic domain in chemistry, suggesting an impact origin (Fig. 2).

Fig. 1: Representative petrographic textures of mare and highlands Chang’e-6 (CE6) homogeneous impact glass beads (Homo-IGBs).
figure 1

a WGP15,G25 and (b) WGP17,G23 are mare Homo-IGBs. c WGP09,G19 and (d) WGP15,G09 are two highlands Homo-IGBs. In this study, 28 individual Homo-IGBs were found (Supplementary Figs. 1–3). Among them, one is picritic, 12 are mare, and 15 are highlands (Supplementary Data 2). The white squares with numbers are the NanoSIMS analytical positions. Three NanoSIMS mapping locations are outlined with dashed squares at the margins of the beads.

Full size image
Fig. 2: MgO/Al2O3 versus CaO/Al2O3 for the Chang’e-6 (CE6) homogeneous impact glass beads (Homo-IGBs).
figure 2

Most of the CE6 Homo-IGBs are plotted within both highlands and mare domains, except one (WGP09,G31) in the picritic domain. The classification criteria (dash lines) of lunar glasses are referred to refs. 65,66. The datasets of the lunar volcanic glasses and impact glass beads in the nearside are from ref. 48 (Source Data). The observed SW-H2O hydration profiles are exclusively identified in the mare Homo-IGBs. The source EPMA data of CE6 Homo-IGBs are listed in Supplementary Data 2. The average values of CE6 and CE5 Homo-IGBs are compiled in Source Data. IGBs, impact glass beads.

Full size image

The CE6 Homo-IGBs were measured to have equivalent water abundances of ~10–1070 μg.g−1 with H isotope compositions (in notation of δD, δD = 1000 × (D/Hsample/D/HSMOW-1), where D/HSMOW is the D/H ratio of the standard mean ocean water; Methods) ranging from – 988 ± 61 to 3718 ± 535‰ after spallation correction at a cosmic ray exposure age of 108 million years ago (Fig. 3 and Supplementary Data 3; Methods). The highlands Homo-IGBs contain ~ 10–150 μg.g−1 H2O, with δD of – 951‰ to 3718‰. The H2O content in mare IGBs ranges from ~10 to 1070 μg.g−1, with δD values between – 988‰ and 1778‰. The picritic impact glass bead has 18–161 μg.g−1 H2O and δD values ranging from 32‰ to 2509‰ (Supplementary Data 3). Each CE6 Homo-IGB was carefully measured at the margins to the centres to investigate the potential hydration profiles of SW-H2O (Supplementary Data 3). Eight profiles were conducted on 7 Homo-IGBs to determine the distribution of water and its isotopic compositions through these targets (Fig. 4 and Supplementary Fig. 4). Two mare Homo-IGBs (WGP15,G25 and WGP17,G23) have hydration profiles showing elevated H2O (309–1070 μg.g−1) at the outermost margins and sharply decrease to tens of μg.g−1 level (Figs. 4a, c). The NanoSIMS mapping results show that the H2O-enriched layers at the margins of WGP15,G25 and WGP17,G23 are < 4 μm in their widths (Fig. 5 and Supplementary Fig. 5). The other IGBs were not identified to have SW-H2O hydration profiles like WGP15,G25 and WGP17,G23 (Fig. 4 and Supplementary Figs. 4and 6). Those CE6 Homo-IGBs without hydration profiles have relatively limited intra-bead variations in water abundances (from 34 ± 8 to 102 ± 28 μg.g−1, 1 SD) and δD values (– 348 ± 305 to 2115 ± 606‰, 1 SD) estimated on the profile analyses (Fig. 4, Supplementary Fig. 4 and Data 3). Most CE6 Homo-IGBs contain water abundances <150 μg.g−1 (Figs. 3, 6 Supplementary Data 3). All CE6 Homo-IGBs display a notably negative correlation between water abundances and δD values (Fig. 3and Supplementary Data 3). It is interesting to note that the water abundances in the CE6 Homo-IGBs display a dependence on chemical compositions (Figs. 2, 6, 7 and Supplementary Figs. 7, 8). The mare Homo-IGBs have a higher water abundance and a lower δD value than that of highlands and picritic ones (Figs. 6, 7and Supplementary Data 3). Meanwhile, SW-H2O hydration profiles were only observed in mare Homo-IGBs (Figs. 4, 5 and Supplementary Figs. 4–6).

Fig. 3: Water abundances and δD values of Chang’e-6 (CE6) homogeneous impact glass beads (Homo-IGBs).
figure 3

a The water abundances and hydrogen isotope compositions (δD) of the CE6 Homo-IGBs exhibit a negative correlation, similar to the trend observed in lunar nearside Homo-IGBs. This negative correlation between H2O and δD can be explained by a binary endmember mixing model16. One endmember represents solar wind-derived water, with a concentration of 2000 μg.g−1 and a δD value of – 990‰. The other endmember corresponds to the initial water derived from the precursor materials of the impact glass beads, with H2O of 2–30 μg.g−1 and a δD value of ~ 3000‰. The inset is an amplified plot with H2O from 0 to 200 μg.g−1. The data of CE6 Homo-IGBs are listed in Supplementary Data 3. The error bars are 2σ. b The average water abundance and δD of individual Homo-IGBs. The H2O error bars are 1 SD of the measured values, and the δD errors are 1 SD of the measured D/H ratios. Literature data for nearside glasses in (a) include: CE5 mare impact glass16, volcanic glass beads68,70, and agglutinates71, listed in Source Data. Average values of farside and nearside impact glass beads (IGBs) in (b) are listed in Source Data.

Full size image
Fig. 4: Typical H2O and δD profiles measured on Chang’e-6 (CE6) homogeneous impact glass beads (Homo-IGBs).
figure 4

a–d are profiles measured on two CE6 mare Homo-IGBs (WGP15,G25 and WGP17,G23). e–h are profiles measured on two CE6 highlands Homo-IGBs (WGP09,G19 and WGP07,G1b). Analytical positions of the outmost rims are measured to be about 3–5 μm to the edges of the impact glass beads (IGBs). All the analytical positions are shown in Fig. 1 and Supplementary Fig. 3. For the mare IGBs, the water abundances in the mantle-centre regions of the IGBs are slightly higher than the instrument H2O background (a, b), yielding the large uncertainties on the measured δD. Thus, we treated their δD values as 0‰ for the profile plot. The mare IGBs show SW-H2O hydration profiles, with H2O abundance in the rims up to 309 μg.g−1 and 1070 μg.g−1, the δD of – 990‰ and – 988‰. The highlands Homo-IGBs do not show SW-H2O hydration profiles and have relatively limited intra-bead H2O variations (23–53 μg.g−1 and 79–93 μg.g−1, respectively). Meanwhile, these beads have notably higher δD values than that of typical SW-H2O (about – 990‰) with inter-bead variations of 988 to 2946‰ and – 510 to 200‰, respectively. The error bars are 2σ. The source data are provided in Source Data.

Full size image
Fig. 5: NanoSIMS mapping of the distribution of SW-H2O in WGP17,G23 homogeneous impact glass beads (Homo-IGBs).
figure 5

Distribution of 1H (a), 2D (d), 12C (b), and 18O (e) and the derived 1H/18O (c) and 12C /18O (f) images in the analytical target. The mapping size is 10 μm × 10 μm outlined with a dashed white square in Fig. 1b. The width of the SW-H2O hydration band on WGP17,G23 is ~ 3.2 μm.

Full size image
Fig. 6: Distribution patterns of H2O and δD of homogeneous impact glass beads (Homo-IGBs) from the farside and nearside of the Moon.
figure 6

a Histogram of H2O from the farside mare Homo-IGBs. b Histogram of δD from the farside mare Homo-IGBs. c, d are histograms of H2O and δD from the nearside non-mare (highlands and picritic) Homo-IGBs. e, f are histograms of H2O and δD from the nearside mare Homo-IGBs. The farside non-mare Homo-IGBs have an H2O peak at ~50 μg.g−1 lower than that (~80 μg.g−1) of mare Homo-IGBs, and a δD peak at ~800‰ higher than that (about –200‰) of mare Homo-IGBs. In addition, the water abundance of the mare Homo-IGBs from the nearside has another H2O peak at ~550 μg.g−1 with notable SW-H2O contribution (δD peaked at about – 750‰). Gaussian fit curves are displayed. Nearside datasets are from ref. 16. The source data are provided in Source Data. IGBs, impact glass beads.

Full size image
Fig. 7: Water abundances versus chemical compositions of Chang’e-6 (CE6) homogeneous impact glass beads (Homo-IGBs).
figure 7

The relationship between H2O abundances with SiO2 (a), Al2O3 (b), TiO2 (c), and FeO (d) of CE6 Homo-IGBs. The chemical compositions are the average values measured by electron probe microanalyser for individual impact glass beads (IGBs). Water abundances are referred to the values measured at the edges of the individual IGBs, thus representing the apparent contributions of SW-H2O. The highlands Homo-IGBs host remarkable less water than mare ones. Meanwhile, water abundances in Homo-IGBs exhibit a logarithmic positive correlation with TiO2 and FeO and a logarithmic negative correlation with SiO2 and Al2O3. The light grey rectangles are the regions with notable SW-H2O hydration profiles. The source data are provided in Source Data.

Full size image

Universal role of mare IGBs in storage of surface water

The CE6 mare Homo-IGBs collected from the farside exhibit a clear negative correlation in H2O–δD space, both individual analytical spots and individual IGBs, plotting along the same trend defined by the mare Homo-IGBs collected in the nearside hemisphere (Fig. 3). This negative correlation (Fig. 3) is likely due to a binary endmember mixing rather than the degassing of water under high temperature conditions during the formation of the IGBs16. The H2O-rich spots measured in CE6 mare Homo-IGBs are extremely depleted in deuterium (δD = –988‰; Fig. 3 and Supplementary Data 3). This distinct hydrogen isotopic signature is associated with a solar wind origin33,34. These features suggest that SW-H2O can be globally documented by the mare Homo-IGBs independent of geographic locations. Thus, mare Homo-IGBs, a common component in lunar soils, may have played an important role in the recharge, release, and recycling of water on the Moon16,31,35.

Such a prediction is further supported by the hydration profiles of CE6 mare Homo-IGBs, which depict the pathways for how the SW-H2O has been preserved in the IGBs (Fig. 4a–d). Theoretically, the maximum penetration depth of solar wind into the lunar surface minerals is ~ 50–100 nm (refs. 18,36.), which is two orders of magnitude lower than the H2O-rich and δD-poor bands (~2.2–3.2 μm) observed in CE6 mare Homo-IGBs (Fig. 5and Supplementary Fig. 5). It requires additional post-implantation processes to drive the implanted SW-H2O moving inwards as well as outwards to form a thicker hydrated layer of SW-H2O in the IGBs than the theoretical calculations and ground observations18,36. There is another possibility that water vapour created by asteroid impacts may have driven the ingress of water in the molten silicate droplets which finally quenched to form IGBs. Taken-up of intermediate volatiles (e.g., Na, K, and Cu) from a transit plume37 during eruption of volcanism on the Moon has been observed in pyroclastic glass beads 74220 (ref. 38). However, the profile analyses of these volcanic glass beads did not exhibit hydration profiles instead of typical degassing profiles of H2O, as well as other highly volatile elements F, Cl, and S (ref. 3). Moreover, the asteroid impacts on the Moon could result in melting of various components in the target soils/rocks, yielding an average δD value of ~0‰ (ref. 39) for the water in the transit plume. Such a higher δD value (~0‰) of the plume is in conflict with the almost pure solar wind component (δD = –990‰) preserved in the margins of CE6 mare Homo-IGBs. Therefore, the first scenario, post-implantation diffusion processes of solar wind, is deemed more plausible than the ingress model of H2O from a transit plume to explain the observations from the CE6 mare Homo-IGBs.

Less SW-H2O stored in lunar farside mare IGBs

The CE6 mare Homo-IGBs, collected from the lunar farside, display contrasting distribution features of water compared with those collected from the Moon’s nearside (Figs. 5, 6, 8, Supplementary Fig. 5). One distinct difference is that the widths (< 4 μm) of the SW-H2O hydration profiles preserved in the farside CE6 mare Homo-IGBs are notably narrower than those observed (up to 40 μm) in the nearside CE5 mare Homo-IGBs (Fig. 5and Supplementary Fig. 5). We carefully investigated the 28 individual CE6 impact glass bead, which may have helped reduce sampling bias (Supplementary Data 3). In comparison, 12 of 33 measured CE5 Homo-IGBs display notable solar wind hydration profiles16. Thus, we assume that the datasets measured in the lunar Homo-IGBs from both hemispheres should reflect the intrinsic characteristics.

Fig. 8: Comparison of the distribution of SW-H2O in farside and nearside homogeneous impact glass beads (Homo-IGBs).
figure 8

a, b The hydration layer of SW-H2O in farside Homo-IGBs with an observed maximum depth < 4 μm (CE6, this study; Figs. 4 and 5) compared to the distribution of SW-H2O in the nearside Homo-IGBs (CE5; ref. 16). c Comparison of the deepest hydration profiles measured in the nearside (CE5; ref. 16) and the farside impact glass beads (IGBs) (CE6; Figs. 4 and 5, Supplementary Fig. 5). The source data for the plots are provided in Source Data. CE5, Chang’e-5; CE6, Chang’e-6.

Full size image

The narrow widths of the hydration profiles of SW-H2O found in the CE6 mare Homo-IGBs may be associated with local temperature23, flux of solar wind on the farside40, local magnetic anomalies41,42,43, and/or their geologic exposure history44. In terms of local temperature, the landing sites of CE5 (43.058°N; ref. 45) and CE6 (41.625°S; ref. 46) missions have comparable geographic latitudinal positions, indicating the diurnal temperature variations at both sites should have a comparable range23, eliminating it as the main reason for the distinct distribution of water in the CE6 mare Homo-IGBs (Figs. 6, 8). Regarding the flux of solar wind protons on the lunar farside, the lunar surface flux is modelled to be higher on the farside than on the nearside40, indicating the potential for more contributions of SW-H2O in the lunar farside materials, which is opposite to our observations, thereby also ruling it out as a potential explanation. Regions of magnetic anomalies have been recently discovered on the Moon43, potentially shielding or reflecting solar wind from implantation into surface materials41. However, no magnetic anomaly has been found at the landing site of CE6 (ref. 43). Its exposure history, constrained by the isotopic dating of the return mare basalt fragments of CE6 is ca. 2.8 billion years ago44, older than that the ca. 2.0 billion-year-old CE5 (ref. 47), indicating that CE6 mare Homo-IGBs may have had more opportunity to receive SW-H2O during the formation and gardening of lunar soils because of its longer exposure time32, which again is opposite to our observations, ruling out an exposure age explanation. Therefore, the contrasting hydration profiles depths between the nearside CE5 and farside CE6 remains an enigma worth of further study.

Dichotomic SW-H2O storage between mare and highlands IGBs

Beyond the differences between them, the CE6 and CE5 Homo-IGBs share a notable similarity in that their SW-H2O hydration profiles are only identified in the mare Homo-IGBs from both hemispheres (Figs. 4, 5 and Supplementary Figs. 4–6), suggesting that recharge of SW-H2O in Homo-IGBs is likely composition dependent. CE6 lunar soils contain notably more highlands Homo-IGBs (~54%; Supplementary Data 1) than CE5 soils (~8%; refs. 16,48), providing adequate targets to investigate the storage and release of SW-H2O in mare and highlands Homo-IGBs. The highlands Homo-IGBs identified in CE6 would have been delivered from non-mare units in the vicinity of the CE6 landing site or quenched from highlands ejecta accompanying with the formation of mare Homo-IGBs. Both scenarios share a co-exposure history in a time span of ca. 2.8 billion years ago (Ga; refs. 44,49), providing equivalent opportunities to receive SW-H2O and to release of SW-H2O caused by mild heating events on the near surface of the Moon. Therefore, the CE6 highlands Homo-IGBs would have experienced similar recharge and release of SW-H2O as the mare ones in the past 2.8 Ga, yielding similar hydration and dehydration profiles of SW-H2O, contrary to the observations (Fig. 4and Supplementary Fig. 4). In another aspect, provided the CE6 highlands Homo-IGBs had been recharged with SW-H2O like mare Homo-IGBs but later altered by some heating events, we should expect to observe the hydration and dehydration signatures of SW-H2O in highlands IGBs. However, most CE6 highlands Homo-IGBs measured have notably higher δD values (500–3000‰; Supplementary Data 3) than that of SW-H2O (about –990‰; ref. 16). Combining these considerations, the CE6 highlands Homo-IGBs seem to be more resistant to diffusion of SW-H2O than the mare ones (Figs. 3, 4; ref. 35). The diffusion of water in matter is mainly relying on the activation energy, diffusion coefficient, and temperature50. As discussed above, the local temperatures between the landing sites of CE5 and CE6 are comparable23. We therefore infer that the contrast H2O differences between mare and highlands Homo-IGBs observed in this study (Figs. 3 and 4) are likely composition dependent because the activation energy and diffusion coefficient are relying on chemical compositions51,52.

To explore the potential relationships between SW-H2O recharge capacity and Homo-IGBs chemical compositions, we made a comparison between the measured H2O and chemical compositions (SiO2, Al2O3, TiO2, and FeO) of Homo-IGBs from both hemispheres (Fig. 7). The H2O values refer to the maximum value measured in the rims of Homo-IGBs, therefore representing the relative contribution of SW-H2O. It is clear that the highlands Homo-IGBs host notably lower SW-H2O than most mare ones (Fig. 7). Meanwhile, SW-H2O storage of Homo-IGBs exhibits notable dependency on chemical compositions: the Homo-IGBs, with SiO2 < 43 wt%, Al2O3 < 15 wt%, TiO2 > 3 wt%, and FeO > 17 wt%, hold more SW-H2O than the rest of the beads (Fig. 7). The regression relationship between chemical composition and SW-H2O of Homo-IGBs is not linear, indicating that multiple elements may affect the recharge efficiency of SW-H2O in Homo-IGBs (Fig. 7). Anyhow, the overall trend is obvious and the correlation coefficient (R2) ranging from ~ 0.6 to 0.8 is strong to very strong (Fig. 7). To further unravel the relationship between SW-H2O storage and chemical composition in Homo-IGBs, a principal component analysis was carried out. The statistics suggest that FeO and TiO2 are positively correlated with SW-H2O, and Al2O3, CaO, SiO2, and MgO are negatively correlated with SW-H2O, and other elements are insensitive due to their low contents (Supplementary Fig. 8).

The dependence of recharge efficiency of SW-H2O in Homo-IGBs with chemical compositions carries implications for in situ water utilisation. The highlands Homo-IGBs, characterised by enrichment of Si, Al, and Ca, are more resistant to retain SW-H2O than mare Homo-IGBs, thus functionally driving release and transportation of SW-H2O to space and the polar regions31. In comparison, the mare Homo-IGBs are capable of storing SW-H2O in lunar soils, thus functionally buffering the lunar surface water cycle16. The lunar surface is characterised by two distinct lithologies, mare basalt and highlands rock53, the former of which dominantly occurs on the lunar nearside, suggesting that storage of SW-H2O in Homo-IGBs may exhibit a dichotomy feature coupled with the surface rock types. These features suggest that mare basalt units on the Moon should have higher priorities than the highlands units for in situ water utilisation. Furthermore, within the favourable Homo-IGB mare group, the Fe- and Ti-rich mare units should be prioritised over those Fe- and Ti-poor ones. Therefore, we propose that lunar base construction and water harvesting plants should ideally target lunar nearside Fe- and Ti-rich mare units.

The detailed studies on the Homo-IGBs from both the near and far hemispheres of the Moon suggest that these small beads may have played a big role in storing the solar wind-derived water for the Moon and carry abundant information about the origin, storage, and recycling of mission-critical lunar surface water. These findings provide fundamental information for figuring out the strategies of in situ water utilisation5. The recharge of SW-H2O in Homo-IGBs is thus apparently more complex than previously thought, requiring further studies to shed more light to unravelling the mysteries of these putative targets for sustaining the surface water cycle on the Moon, as well as on other airless bodies16,30,31,54.

Methods

Sample preparation

Two scooped lunar soil samples CE6C0300YJFM002 (~ 6 g) and CE6C0200YJFM001 (~5 g) were used in this study. The glass beads, along with some other components, were handpicked under a binocular microscope in the ultraclean room at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS) in Beijing. The sample preparation methods can be referred to He et al. 55. The picked particles were placed in a tunnel of silica glass cut by low-speed diamond saw and then mounted by epoxy and prepared into polished thick sections labelled as WGP01 to WGP18 (Supplementary Fig. 1). The glass beads in the sample mounts, WGP07, 09, 10, 15, and 17 were extensively measured in this work (Supplementary Data 1). The prepared sections were cleaned up using anhydrous ethanol prior to drying at 50 °C in a baking oven. All the mounts were coated with carbon for petrographic observation and chemical analysis. The coating was replaced with Au for water abundance and hydrogen isotopic analysis.

Scanning electron microscopy

Petrographic observations were carried out using field emission scanning electron microscopy (FE-SEM) on a FEI Nova NanoSEM 450 at IGGCAS, using electron beam currents of 6.4 nA and an acceleration voltage of 15 kV. The prepared sections were initially coated with carbon for petrographic observations, and then recoated with Au for in situ NanoSIMS measurement of water content and hydrogen isotopes. After NanoSIMS analyses, the original coating was removed, and the samples were cleaned and recoated with carbon for SEM observation to confirm the positions of the NanoSIMS spots (Supplementary Fig. 3).

Electron probe microanalysis

The major elements abundances of the impact glass beads were measured using a JEOL JXA-8100 electron probe micro-analyser (EPMA) at IGGCAS. The accelerating voltage was 15 kV and the beam current was 10 nA, with a beam size of 10 μm in diameter. The standards 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, and a natural obsidian glass was used as a monitor. The detection limits were 0.02 wt.% for MgO and NiO, < 0.01 wt.% for Na2O, Cr2O3, Al2O3, MnO, K2O, SiO2, FeO, CaO, and TiO2. The analytical results are listed in Supplementary Data 2.

In situ water abundance and hydrogen isotope composition analysis

In situ water abundance and hydrogen isotope analysis were conducted using a CAMECA NanoSIMS 50 L at IGGCAS, following a well-used analytical protocol described in Hu et al.56 and Barnes et al.57. The prepared samples and standards were placed in the same holder and stored in the airlock of the NanoSIMS 50 L. They were baked at ~60 °C in the airlock more than 12 h to reduce the hydrogen background during analysis. Then, the sampled were stored in the vessel chamber under high vacuum conditions prior to measurements of water abundances and hydrogen isotopic compositions. The vacuum pressure in the analysis chamber was ~ 2.0 × 10−10 mbar during analysis, facilitated by a liquid nitrogen trap58. Each 10 μm × 10 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 5 μm × 5 μm region of interest on targets was selected for analysis. Secondary anions 1H−, 2D− 12C−, and 18O− were simultaneously counted by electron multipliers for 400 cycles from the central 50% area (3 μm × 3 μm) using the electronic gate technique of NanoSIMS. The acquisition time is ~4 min for each measurement. The primary beam current (Fco) was ~0.5 nA for analysis with a beam size of ~500 nm in diameter. The charging effect on the samples surface was compensated by an electron-gun during analysis. The real time images (RTI) of 1H− and 12C− were used to monitor the analytical locations and potential contaminations.

The calibration line for water abundance was established using a series of glass and apatite standards, including SWIFT MORB glass (H2O = 2580 μg.g−1), Kovdor apatite (H2O = 9800 ± 700 μg.g−1)59, Durango apatite (H2O = 480 μg.g−1)60,61, and two basaltic glasses, ALV-519-4-1 (H2O = 1,700 μg.g−1)62 and ALV-1833-11 (H2O = 12,000 μg.g−1)62 (Supplementary Data 4;and Fig. 9). Instrument mass fractionation (IMF) on hydrogen isotope composition was corrected using the SWIFT MORB glass (δD = – 73 ± 2‰), and monitored by Kovdor apatite (δD = –66 ± 21‰)59 and Durango apatite (δD = – 120 ± 5‰)60,61, as the matrix effect on hydrogen isotope composition are same between silicate glass and apatite within analytical uncertainties56. The basaltic glasses ALV-519-4-1 and ALV-1833-11 were also used to monitor the stability of instrumental mass fractionation of hydrogen isotopes throughout the analytical session, yielding comparable δD values to previous values2 within analytical uncertainties (Supplementary Data 4and Fig. 10). Hydrogen isotopic compositions are given using the delta notation, δD = ((D/H)sample/(D/H)SMOW)− 1) × 1000 ‰, where SMOW is the standard mean ocean water with a D/H ratio of 1.5576 × 10−4.

The instrument H background was detected by a San Carlos olivine63 (H2O ~ 1.4 μg.g−1) and a synthetic anhydrous silica glass Suprasil 3001. The H background correction was referred to ref. 64., following the relationship: H/Obg = (Hcounts - Hbg) / Ocounts and D/Hmeasured = (1-f) × D/Htrue + f × D/Hbg, where f is the proportion of H emitted from the instrumental background. The instrument background has gradually decreased as the sample stored in the instrument for longer time. In the first four days of the analysis session, a slightly higher instrument background at 15 ± 5 μg.g−1 (D/Hbg average was (1.49 ± 0.75) × 10−4 and Hbg was 516 ± 164 counts per second (cps)). For the following four days, the instrument H2O background was 7 ± 1 μg.g−1 (D/Hbg average was (2.71 ± 1.67) × 10−4 and Hbg was 259 ± 50 cps) (Supplementary Data 4). After background subtraction, the water abundances of the IGBs were quantitatively calculated based on the H2O content calibration line established on the standards (Supplementary Data 3).

For notable hydration profiles, we also used a low Fco ( ~ 100 pA) to carry out high spatial resolution mapping analyses for acquiring the distribution of SW-H2O from the edge to the inner side of the IGBs after the spot measurements. The mapping results suggest that the SW-H2O hydration depths are ~ 2.2 μm for WGP15,G25 (Supplementary Fig. 5) and ~ 3.2 μm for WGP17,G23 (Fig. 5). But the SW-H2O hydration band was not observed on WGP15,G09 (Supplementary Fig. 6).

Petrography of CE6 IGBs

A total of 263 particles were mounted in five sample mounts (WGP07, 09, 10, 15, and 17). Seventy-seven impact glass beads (IGBs) were identified among these particles (Supplementary Data 1). These IGBs exhibit various textures (Supplementary Fig. 1). Twenty-eight of them have smooth exposed surface without any inclusions or vesicles (Fig. 1 and Supplementary Fig. 3), hereafter referred as Homo-IGBs. The other IGBs are characterised by the presence of vesicles, bubbles, mineral clasts, sulfide droplets, and spherical Fe-Ni metals (Supplementary Fig. 2). The Homo-IGBs were selected to investigate the potential SW-H2O hydration process in the farside IGBs and to find out the common processes may have happened in both nearside and farside IGBs. CE6 Homo-IGBs exhibit round to elliptical shapes with diameters ranging from ~50 μm to ~300 μm (Supplementary Fig. 3). Two of the Homo-IGBs are coated by some finer dusts (Supplementary Fig. 3). These Homo-IGBs display inter-bead contrasts, indicative of varied chemical compositions (Fig. 1and Supplementary Fig. 3).

Chemistry

The chemical composition of CE6 Homo-IGBs were carefully measured using EPMA (Supplementary Data 2). Each Homo-IGB has relatively homogeneous chemical compositions compared with the large inter-bead variations (Supplementary Data 2), consistent with the contrast difference observed using SEM (Fig. 1 and Supplementary Fig. 3). The studied CE6 Homo-IGBs are plotted in three domains on the classification diagram of lunar glasses65,66, highlands (15), mare (12), and picritic (1) (Fig. 2 and Supplementary Data 2). CE6 Homo-IGBs have notably more highlands beads than CE5 Homo-IGBs16,48, consistent with the characterisation of the CE6 soil samples46. The CE6 picritic glass (WGP09,G31) contains notably lower Al2O3 (6.3–6.9 wt%) and CaO (6.4–6.9 wt%) and higher MgO (18.7–20.2 wt%) than highlands and mare Homo-IGBs (Supplementary Data 2, Fig. 7). The CE6 highlands Homo-IGBs contain lower TiO2 (0.6 ± 0.6 wt%) and FeO (9.7 ± 5.4 wt%) and higher Al2O3 (19.9 ± 5.4 wt%) (Supplementary Data 2 and Fig. 7). The CE6 mare Homo-IGBs have comparable chemical compositions with CE5 mare Homo-IGBs except slightly lower TiO2 (0.82–3.38 wt%) and FeO (16.0–19.8 wt%) and higher SiO2 (40.4–47.6 wt%) (Supplementary Data 2and Fig. 7). All Homo-IGBs contain minor Cr2O3, Na2O, and MnO (Supplementary Data 2). It is observed that the highlands and picritic Homo-IGBs are plotted along the same trend in the diagram of CaO versus Al2O3, exhibiting a lower slope than the mare Homo-IGBs (Supplementary Fig. 7f). Three highlands Homo-GBs (WGP09,G19, WGP15,G06 and WGP17,G02) were observed to have sodium oxide profiles decrease from the rims towards the centres (Supplementary Fig. 11).

Correction of water abundances and D/H ratio for spallation effect

The measured D/H ratios have been corrected for cosmic-ray spallation effect, using a deuterium production rate of 2.17 × 10−12 mol.g−1.Myr−1 (Myr, million years; ref. 67). The correction introduces error of comes from the 50% uncertainty of the H and D produce rate (Supplementary Data 3; ref. 68). A cosmic ray exposure age of 108 Myr used for CE6 (ref. 55) was applied in this study to ensure the maximum correction for the extremely elevated D/H ratios.

Degassing modelling

The change in the hydrogen isotopic composition during volatile degassing into a vacuum can be modelled using Rayleigh distillation, expressed as:

$${\delta }_{{{\rm{final}}}}={\delta }_{{{\rm{initial}}}}+(1,000\,+{\delta }_{{{\rm{initial}}}})\times ({{{\rm{F}}}}^{\left({{\rm{\alpha }}}-1\right)}-1)$$
(1)

where \({\delta }_{{{\rm{initial}}}}\) and \({\delta }_{{{\rm{final}}}}\) are the initial and instantaneous isotopic compositions of the material formed IGBs, F is the fraction of the element remaining, and α is the fractionation coefficient. The fractionation coefficient α is determined by the volatile species losing into a vacuum, governed by Graham’s law and given by α2 = Μ2/Μ1, where M1 is the molar mass of the heavy isotopologue, and M2 is the molar mass of the volatilising light isotopologue. For the degassing trend shown in Fig. 2, we assume that the IGBs have an the initial water abundance of 2000 μg.g−1, with a δD value of −990‰ based on the measured datasets (Supplementary Data 3), and degas in the form of H2 (M1 = 2 for H2 and M2 = 3 for HD), yielding an α value of ~ 0.8165 (ref. 69).