Introduction

Water on the Moon is of paramount importance for understanding the formation and evolution of the lunar interior and for future in-situ resource utilization (ISRU)1,2. The abundance of water in the lunar mantle provides constraints on the Moon-forming giant impact3, the subsequent crystallization of the lunar magma ocean4, and the longevity of volcanism on the Moon5. By contrast, the lunar surface is covered by a layer of fine-grained regolith with a thickness up to ~12-15 meters6. This regolith layer can contain abundant water7,8,9, and the polar lunar regolith could contain water ice, which are essential resources for building a lunar base and long-term life support10. The ISRU will drastically reduce the reliance on supplies from Earth. The water resources can be further used to produce oxygen/hydrogen for fuel11. Infrared spectrometry data provide direct evidence for the presence of OH/H2O on the lunar surface12,13,14, with both species (OH and H2O) commonly referred to as water. This finding was further confirmed by the Stratospheric Observatory for Infrared Astronomy (SOFIA), which detected molecular water at high latitudes on the Moon15. The laboratory analyses of Apollo and Chang’e-5 (CE5) lunar samples suggested that both indigenous source16,17,18,19 and exogenous sources such as solar wind (SW) implantation7,8,9,20, carbonaceous chondrites21,22, and comets23 have contributed to the distribution of OH/H2O on the lunar surface. On a global scale, the distribution of OH/H2O on the lunar surface, determined from the Moon Mineralogy Mapper (M3) reflectance spectra, is observed to correlate with latitude or temperature, with higher concentrations at higher latitudes24. This phenomenon was interpreted to be mainly caused by SW implantation, which is later verified by laboratory analyses of returned lunar samples. The data show that the OH/H2O concentrations in the space-weather rims of CE5 silicate grains from the middle latitude are indeed higher than the values measured in Apollo soil samples from low latitudes8,9,25, supporting the latitude-dependent conclusion drawn from remote sensing data. On the other hand, the time-of-day-dependent variations in OH/H2O were also observed by spectral observations, which are also attributed to the SW implantation24,26,27. These implanted SW-H is considered to combine with O to form OH or H2O, which is subsequently lost from soil grains through temperature-dependent diffusion and photolysis26,28.

However, it remains controversial whether the variations of OH/H2O content on the lunar surface depend on the chemical compositions. Some studies show that the low-Ti regions commonly have higher water contents than high-Ti mare basalts29,30. A relatively larger time-of-day-dependent variation of surficial OH/H2O in Ti-rich areas has been displayed by comparing with other low-Ti mare regions29,30. Other studies suggest that there are some areas with water anomalies in the mare regions of the Moon, but their TiO2 content does not differ significantly from the surrounding areas31. The lunar maria are mainly composed of plagioclase, pyroxene, olivine, and ilmenite, and the Ti-rich regions usually represent a high abundance of ilmenite32. However, previous studies on H isotopes in Apollo and CE5 lunar soil grains have focused only on silicate minerals, agglutinates, and glasses, but these different silicate phases do not show significant differences in the abundance of H (refs. 7,8,9,20,23,25,33,34). By contrast, only a few studies were taken to investigate the water contents and H isotopes in the returned lunar ilmenite grains’ rims34,35,36. One ilmenite from sample 71501, measured for H isotopes and H-depth profiles with SIMS, exhibited slight enrichment in deuterium (D), indicating a non-solar H component36. The atom probe tomography measurements on an Apollo 17 ilmenite revealed H-bearing molecular ions, such as OH+ and H2O+, in the space-weathered rims, which are likely sourced from the SW35. The latest nanoscale secondary ion mass spectrometry (NanoSIMS) analyses on two ilmenite grains from the CE5 drill core samples show one ilmenite has a δD value of 110 ± 1172‰ (2σ), and the other displays a SW-like H isotopic ratio (δD = −884 ± 140‰; 2σ)34. To our knowledge, this is the only reported ilmenite grain that has a SW-like H isotopic composition. However, they have not compared the oxides and silicate phases because the data on ilmenite are limited. In addition, the microstructures on these ilmenites have not been analyzed. Further systematic analysis of ilmenite is therefore required to understand the interaction between SW implantation and ilmenite and the fate of the hydration components.

In this study, we conduct a comprehensive analysis to investigate the water contents, hydrogen isotopes, and microstructures in the space-weathered lunar ilmenite grains returned by the CE5 mission from northern Oceanus Procellarum37. The CE5 lunar soil is submature-mature based on the characteristics of the in-situ spectrum and grain size distribution38,39 and consists mainly (>90%) of local mare basalt40,41. The CE5 lunar soils consist of ~4–5% ilmenite by volume, indicative of a low-Ti basalt32. We aim to (i) measure and characterize the water distribution in the rims of lunar ilmenite, (ii) compare the differences between ilmenite and silicate minerals, and (iii) discuss the effects of ilmenite on the water cycle in the Procellarum KREEP Terrane (PKT).

Results and Discussion

The measured H and D concentrations were used to estimate the water contents and isotopic compositions of ilmenite rims. Moreover, the H isotopes can be used to trace the origins of lunar surface water. After NanoSIMS analysis, the microstructures were studied to understand the effects of space weathering on oxides under different conditions and to compare with the silicate minerals.

Water content and hydrogen isotopic composition of ilmenite rims

The hydrogen concentrations with depth and isotopes in CE5 ilmenite grains were measured using NanoSIMS (Supplementary Fig. 1). It should be noted that the NanoSIMS used here cannot differentiate between H species such as H2, OH, H2O, or other compounds. Conventionally, the H concentrations measured with NanoSIMS are converted to H2O equivalents by weight for discussion. To characterize the depth-profiling data, we determined the depths of NanoSIMS-generated craters in ilmenite grains from two polished basalts (Supplementary Fig. 2) and calculated the sputtering rate to be 0.84 ± 0.08 nm/cycle (1 SD), which is slightly lower than the sputtering rate in silicate minerals (~1 nm/cycle)8.

The results show that, similar to previously reported CE5 silicate minerals, the water content in ilmenite grains decreases from the surface towards inside, but the average water content of which is about 2–4 times lower than that in silicate phases (Fig. 1). The maximum water contents within the 100-nm surface layer vary mainly from approximate 1500–5000 ppm, with only two ilmenite grains show slightly higher water contents of 5600 and 6400 ppm, respectively (Fig. 1a).

Fig. 1: Variations of measured H2O content with depth in lunar soils.
figure 1

a CE5 ilmenite. Two ilmenites from CE5 core soils34 are also shown. b–d The Apollo and CE5 olivine, pyroxene, and plagioclase grains. The silicate minerals that have been analyzed by ion microprobe are from previous studies8,25,34. Some depth profiles with abnormal water contents at some depths are not included in the diagram. Only the distribution of H2O concentration in the first 300 nm of the depth profiles is shown due to the low and constant H2O abundance within grains. The measured H2O concentrations were corrected for the background using the San Carlos olivine standard.

Full size image

Based on the simultaneously measured D and H abundances, we calculated the δD (defined as [(D/H)sample/(D/H)SMOW − 1] × 1000, see “Methods” section) in surface layers of ilmenite (~100 nm depth). In addition, we also corrected the water contents and δD values using a cosmic-ray exposure (CRE) age of 50 Myr previously used by Hu et al.5 and He et al.7 and the reported D and H production rate by cosmic rays42. This correction has a negligible effect on the water content (less than 1 ppm; Supplementary Table 1). In addition, since a ~10-nm-thick vapor-deposited layer is sometimes observed on ilmenite surfaces (Supplementary Fig. 3), we also subtracted the water content associated with this layer. The results show that the average water content at 10–100 nm depth decreased by ~10% compared to the unsubtracted data, while the H isotopic composition remained consistent within the uncertainty (Supplementary Table 1). Therefore, the reported water contents in the 0–100 nm layer likely represent the upper limit of the ilmenite rims. These values, along with their corresponding H isotopes, were still used in the subsequent discussions. With the exception of one ilmenite from CE5 core soils with mantle-like H isotopes34, ilmenite grains have corrected δD values that range from −884 ± 70‰ to −482 ± 38‰ (1σ) (Supplementary Table 1). Interestingly, almost all δD values from the thirteen ilmenite grains are consistently lower than that of the lunar mantle, which represents the endmember purely from endogenic hydrogen. In contrast, comparative analysis of SW-derived rims (δD < −400‰) reveals that while the δD values for ilmenites fall within the range of silicate minerals (olivine, pyroxene, and plagioclase), they systematically exceed those of ~60% of silicate minerals, which display much lower δD signatures (Fig. 2). This difference becomes more evident when combined with their water contents (as shown in Fig. 2), resulting in two distinct clusters in the δD vs. H2O content diagram.

Fig. 2: Relationships of Spallation-corrected δD values with H2O contents for the lunar soils.
figure 2

The results for ilmenite rims in this work are plotted to compare with literature data for olivine, pyroxene, plagioclase, and ilmenite returned by the Apollo and CE5 missions8,9,25,34. The lunar mantle81 is also shown for comparison. The SW endmember is inferred to have extremely low δD value (~−997‰; ref. 53). The H isotopic variation can be classified into two categories34, i.e. samples with δD values < −400‰ contains major contribution from SW implantation, where those with δD values > −400‰ are associated with other reservoirs, such as lunar interior, carbonaceous chondrites (CCs) and comets21,23. All analyzed ilmenite samples in this work, as well as one ilmenite from the literature, show the δD values < −400‰. The black line represents the binary mixing between the pure solar component (H2O = 12,000 ppm, δD = −950‰) and the indigenous water reservoir of the Moon (H2O = 100 ppm, δD = 0‰) reservoir of the Moon. Outgassing of H-bearing phases from the minerals into space is also modeled and shown in dashed lines with different colors, i.e., pink for H, green for H2, and blue for H2O (see “Methods” section for details). The error bars represent 1σ uncertainty, and some error bars for samples are within the size of the symbol. The silicate minerals and ilmenites are outlined with different colors, and although they overlap with each other, the silicate phases can have a much higher water content and a lower δD range, forming two distinct clusters. It is noted that the seemingly slight positive correlation observed in the ilmenite data appears to be dominated by two (outlier) data points. The data from this study is provided in Supplementary Table 1.

Full size image

Abundant npFe0 particles and vesicles in the rims of ilmenite

Space-weathered rims on four representative ilmenites (S404-09, S404-14, S404-21, and S503-04) are enriched in nanophase metallic iron (npFe0) particles and vesicles (Fig. 3 and Supplementary Fig. 4), consistent with the characteristics of previously studied space-weathered lunar ilmenites43,44,45,46. In sample S404-09, the SW-damaged rim, about 150 − 165 nm thick, is abundant in elongate npFe0 particles averaging ~13.01 ± 6.01 nm in length and ~5.94 ± 2.23 nm in width (N = 108, number of npFe0 particles), which are randomly distributed within the top ~70 nm surface layer (Fig. 3a–c). Some large vesicles (~10 nm in diameter) are also present in the rim of S404-09 (Fig. 3b, c). In sample S404-21, the space-weathered rim is about 85–100 nm thick, and the npFe0 particles in this grain are relatively small and mainly distributed in the uppermost ~40 nm layer. These npFe0 particles have an average length of ~9.79 ± 5.61 nm and an average width of ~4.92 ± 1.86 nm (N = 45) (Fig. 3f, g). Large spherical or subspherical vesicles, with an average diameter of 9.14 ± 5.25 nm (N = 71), are observed on the uppermost layer (within the top 40 nm) of this grain. Similarly, the other two ilmenites (S404-14 and S503-04) also have a >100 nm-thick SW-damaged zone and contain a ~50-nm-thick upper layer enriched in npFe0 and planar defects (Supplementary Fig. 4). These characteristics indicate a degree of deeply damaged rim based on the classification by Guo et al.46. In addition, we observed a Si-rich vapor-deposited layer (~10–15 nm thick) on two ilmenite grains (Supplementary Fig. 3), consistent with previous results from Apollo and Chang’e-5 lunar soils43,44,46.

Fig. 3: Microstructures of two representative space-weathered ilmenite rims.
figure 3

a Bright-field transmission electron microscopy (BF-TEM) image of the focused ion beam (FIB) section extracted from the ilmenite grain S404-09. b High-angle annular dark-field (HAADF) image of the region indicated in (a). c High magnification HAADF image of the region indicated in (b). d The corresponding energy-dispersive X-ray spectroscopy (EDS) map of Ti and Fe as shown in (c). e EELS spectra of the three regions indicated in (c). f BF-TEM image of the ilmenite grain S404-21. g The helium bubble observed in space-weathered rims in a bright-field TEM image. h HAADF image of the region marked in (f). i EELS spectra of the three regions indicated in (h).

Full size image

Numerous slender and nano-scaled vesicles are also present in the subsurface layer (~40–100 nm) (Fig. 3g). EELS spectra acquired from the substrate ilmenite, npFe0, and the large vesicle in sample S404-21 did not show a diagnostic peak of the H K-edge at the low energy-loss of about 13 eV, while a helium signal (~22 eV) was detected in both vesicles and npFe0 particles within the space-weathered rim (Fig. 3i). By contrast, no helium was detected in the vesicles of sample S404-09, probably due to the perforation of the vesicles by ion beam cutting during sample preparation or electron beam irradiation (Fig. 3e). It should be noted that the space-weathered rims of silicate minerals (olivine and pyroxene) contain npFe0 particles but no vesicles (Supplementary Figs. 5 and 6). Neither hydrogen nor helium signals were detected in the weathered rims of pyroxene and olivine (Supplementary Figs. 5d and 6d). When integrated with previous results from Gu et al.44 and Guo et al.46, the data reveal clear microstructural distinctions between ilmenite and silicate minerals, with ilmenite exhibiting significantly more vesicles and npFe0 particles.

Solar flare tracks are observed in all analyzed minerals, including ilmenite, olivine and pyroxene (Supplementary Fig. 7). The track densities, estimated by counting the tracks within the uppermost zone of the four ilmenite grains, were 6.93 × 1010 (S404-09), 5.02 × 109 (S404-14), 1.32 × 109 (S404-21), and 2.19 × 1010 (S503-04) tracks/cm2, respectively, which are comparable to those of silicate minerals (5.3 × 109 – 8.5 × 1010; Supplementary Table 2). Since there is no experimentally-determined track production rate for ilmenite and pyroxene, we assumed that both have the same rate as olivine and anorthite reported by Keller et al.47. The estimated ages are provided in Supplementary Table 2. It should be noted that as ilmenite is more resistant to amorphization by ion irradiation than silicates due to its close-packed crystal structure48, the actual solar-flare-track exposure time for ilmenite may be similar to or even longer than our estimated values.

Origin of water and H isotopes in ilmenite rims

The corrected δD values of the uppermost H-enriched zone (~100 nm) in our ilmenites are significantly lower than those of the lunar mantle (Fig. 2). To produce such a signature, a low-δD source is required. The reservoirs in the Solar System such as water-rich carbonaceous chondrites (−200 to +800‰)21,49,50 and comets (0 to +2000‰)51,52 generally have positive δD values10, while only the SW is measured to have extremely low δD value (~−997‰; ref. 53). From this standpoint, it suggests that SW is probably the main cause for the H isotopic ratios in ilmenites. This interpretation is also supported by the presence of abundant npFe0 particles and vesicles in the weathered rims of ilmenites, which were thought to be the diagnostic characteristics of SW irradiation46. In fact, the low-δD values (<−400‰) found in CE5 and Apollo silicate minerals and glasses have previously been explained by the contribution from SW implantation8,9,33,34. The ilmenite, whose H isotopes are similar to those of the lunar mantle, was probably only exposed to the SW for a very short time or not at all34. Therefore, we conclude that SW irradiation is a critical source for lunar surface water.

More importantly, we note that although ilmenite and silicate minerals overlap in the δD vs. H2O plot, almost all ilmenites fall above those of silicates with SW-like δD values (Fig. 2). The observed difference is further evidenced by their distinct microstructural characteristics (as detailed in the next section). These distinct H isotopic signatures in ilmenite grains likely stem from varying exposure durations to SW irradiation and/or differential outgassing of H-bearing species after implantation34. However, the exposure time estimated based on solar flare tracks for ilmenites (~0.03–1.57 Myr; Supplementary Table 2) is comparable to those of the silicate minerals (~0.12–1.95 Myr; ref. 8 and this study). Furthermore, we find that both the water content and δD values of the analyzed ilmenite rims remain almost unchanged with increasing exposure time (Supplementary Fig. 8). Specifically, the water contents in ilmenite rims are consistently lower and the δD values are consistently higher, approximately paralleling those of the silicate minerals (Supplementary Fig. 8). These observations therefore rule out the exposure time as the main cause. In contrast, an early study performed ion irradiation experiments on ilmenites and silicate materials (plagioclase and simulated glass) and found that ilmenite stored about twice as much deuterium as other target materials under the same condition and exhibited a higher loss of implanted ions than other target materials after heating54. These results suggest ilmenite grains can contain comparable concentrations of SW-implanted hydrogen to silicates, leading us to infer that lunar ilmenite rims could form similar water contents to those in silicate materials.

Our further modeling shows that the degassing of H-bearing phases from a silicate-like endmember with low δD and high water content could significantly decrease the water content but increase the H isotopes (Fig. 2). Most of the ilmenite data fall within the range defined by the degassing in the form H2, H2O and pure H (Fig. 2). Given these observations, why do ilmenite soil grains have systematically lower water content together with heavier H isotopes in their topmost ~100-nm-depth layer compared to most of the silicate minerals? We speculate that this difference is most likely related to the nature of the minerals, meaning that ilmenite water may not be retained as readily in the weathered rims as in silicate minerals.

Microstructural differences between ilmenite and silicate modulate the formation and retention of SW-derived water

The observed differences between ilmenite and silicate minerals in terms of H isotopes and water contents closely correlate with their crystal structures. Ilmenite, with lower binding energies for Fe-O (407) and Ti-O (667) than Si-O (800) in silicates (numbers in brackets are the bond dissociation energies in kJ/mol)55, can be more easily reduced with H2 in experimental settings to produce water more efficiently and yield a metal product of iron [FeTiO3(s) + H2(g) ⇌ Fe(s) + TiO2(s) + H2O(g)]56,57,58. This mineral is also a potential material for the effective extraction of oxygen and has been extensively studied for ISRU purposes59. TEM measurements reveal abundant npFe0 particles in the space-weathered rims of ilmenites (Fig. 3), indicating that a large amount of SW-implanted H should be present to involve the formation of npFe0 particles. In addition, vesicles are one of the features that can form when a mineral grain is exposed to the space weathering environment, and it has been suggested that the SW is particularly important in the development of these features, implanting H and He ions that concentrate in these void spaces43,60. The trapped H may react with O to form OH and H2O. Recently, Kling et al.61,62 detected H, OH, and H2O in the weathered rims of Apollo olivine and clinopyroxene grains based on the results of EELS and APT analyses, and they suggested that vesicles and other complex microstructural features like defects may be sites where water can be trapped and stored in higher concentrations. Similarly, Bradley et al.60 detected molecular H2O in the vesicles of an interplanetary dust particle (IDP) and in experimentally irradiated samples using the EELS detector. From this standpoint, we infer that the abundant and large vesicles in the weathered rims of ilmenite should have generated more water compared to the silicate minerals. A recent study also proposed that ilmenite is a highly efficient material for producing water63. However, the solar component in the space-weathered rims of ilmenite detected by EELS and NanoSIMS is predominantly helium, with rare H species (Fig. 3), consistent with the results from other previously reported lunar ilmenites35,43,45. Combined with the H isotopic results, we suggest that the SW-induced OH/H2O in the space-weathered rim of ilmenite is more readily lost. A schematic model is then proposed to illustrate the behaviors of SW-implanted H in ilmenite rims, in comparison with silicate minerals (Fig. 4).

Fig. 4: Schematic diagram showing the implantation and outgassing of H-bearing species in weathered ilmenite rims, compared with the silicate phases.
figure 4

a Due to SW implantation, the Fe-O and Ti-O bonds on ilmenite surfaces are destroyed, resulting in the formation of defects, vesicles, and npFe0 particles. The SW-implanted H can be trapped in the defects and vesicles, and then H combines with O to form OH/H2O. We speculate that random impacts and diurnal temperature variations lead to a significant loss of H-bearing species, leaving tiny OH/H2O in the weathered rims. b The space-weathered silicate rims contain fewer vesicles, npFe0 particles, and defects.

Full size image

Multiple factors may trigger the loss of OH/H2O in weathered rims of ilmenites. The water formed in the weathered rims could be released as the temperature rises26,64,65,66, with the lunar surface temperature of up to ~380 K (ref. 67). When the gas pressure (e.g., noble gases, H2, H2O) within vesicles exceeds their percolation threshold, as shown in the experimental study on H-irradiation of alumina68, this feature could accelerate the outgassing of H-containing species into space. Losses of volatile gases can also occur through micrometeorite impacts and SW sputtering. Thus, although ilmenite grains could form more water, they struggle to efficiently retain it. Future high-energy ion implantation experiments on oxide and silicate minerals can test our hypothesis and may help interpret the remote sensing observations.

Implications for the diurnal variation of water in the lunar mare regions

The new data of H isotopes and water contents measured in the rims of lunar ilmenites have important implications for understanding the lunar surface water cycle. Previous studies using remote sensing data showed diurnal OH/H2O variations across most of the lunar surface, particularly for the lunar mare regions, such as in Mare Crisium, Mare Serenitatis, Mare Nectaris, Mare Tranquillitatis, and western Oceanus Procellarum29,30. They further found that the high-Ti regions generally have a weaker OH/H2O absorption band than the low-Ti basalts, especially at lunar midday, and that high-Ti mare areas show much stronger variations in 3-µm band depth throughout the lunar day compared to low-Ti maria29,30. The amplitude of OH/H2O adsorption bands in the lunar highlands, which have even lower TiO2 content, is generally lower. Though the mechanism for these variations remains uncertain, it has been proposed that TiO2-rich regions tend to adsorb SW-implanted H primarily into low-energy binding states, facilitating a periodically occurring time-of-day-dependent desorption and reabsorption of OH/H2O (ref. 29).

As shown in Fig. 5, we compared the maps of mineral abundances (olivine, clinopyroxene, plagioclase, and TiO2 content) with the IBD3μm maps over a lunar day in the PKT. The IBD3μm index quantifies the amount of OH/H2O, derived from the M3 spectral data calibrated in a refined way by a newly developed physical thermal model69. Our extended comparison of ten regions in the PKT reveals that the diurnal variations in IBD3μm (morning-noon and evening-noon) positively correlate with both TiO2 and clinopyroxene, but negatively with plagioclase (Fig. 5g). The inverse relationship between plagioclase and water diurnal variation indicates that the presence of plagioclase likely inhibits water diurnal variation, which aligns with experimental results: experiments on SW implantation suggested that plagioclase can capture more H+ than olivine or pyroxene to form OH/H2O65,70. By contrast, water diurnal variations show positive correlations with both TiO2 (predominantly from ilmenites as shown in Supplementary Fig. 9) and clinopyroxene, indicating that both minerals probably influence water variations. Since clinopyroxene generally has comparable water concentrations to olivine and plagioclase in lunar soils71, we infer that ilmenite—the dominant Ti-rich oxide, constituting up to 20% of lunar maria32—may significantly contribute to the observed diurnal water variations in the PKT. However, factors such as latitude and regolith maturity have also been shown to influence lunar surface water cycling24, which must be considered when interpreting localized diurnal water variations.

Fig. 5: Comparison between mineral abundance and time-of-day-dependent variations of lunar hydration maps in the PKT.
figure 5

a TiO2 content. b Olivine. c Clinopyroxene. d Plagioclase. The PKT is outlined by a white dashed line. e Variations of hydration by subtracting the noon (12:00–14:00) map from the evening (16:00–17:00) map. f Variations of hydration by subtracting the noon (12:00–14:00) map from the morning (7:00–8:00) map. The IBD3µm parameter was defined and used to quantify the average relative absorption strength across the wavelength interval of 2.7–2.9 µm (ref. 69). The ten ellipses in these panels indicate regions with different mineral abundances and variations in OH/H2O adsorptions. The TiO2 abundance map is generated from the WAC ultra-violet and visible images82, and the mineral abundance maps are produced from the Kaguya Multiband Imager (MI) data83. Some areas lack data for morning or midday values. g Temporal variations in water content versus the abundances of TiO2, olivine, clinopyroxene, and plagioclase. Ten regions, delineated by black ellipses, were selected and quantified. Gray lines represent the possible correlations.

Full size image

Our laboratory analyses of CE5 ilmenite demonstrate that these minerals can form additional OH/H2O through SW implantation. However, most of this hydrogen is subsequently removed via diffusion/thermal evaporation and photolysis. This observation likely supports the view that SW-induced hydrogen is adsorbed into binding states of low energy29, though future ilmenite-specific experiments are required to investigate OH/H2O formation and release mechanisms. In other words, compared to other candidate water reservoirs such as lunar glasses and plagioclases7,65, our findings suggest that the widespread ilmenites, particularly in lunar maria, do not serve as potential water reservoirs, despite their well-documented capacity to retain He and noble gases72,73. Therefore, our results shed light on the dynamic interactions between SW and oxide minerals and underscore the imperative to reassess ilmenite’s viability as a strategic resource for future lunar exploration.

Methods

Samples

Two CE5 lunar soil fractions (CE5C1000YJFM00404 and CE5C0600YJFM00503; hereafter abbreviated as S404 and S503, respectively) from the China National Space Administration were used in this study. The preservation and preparation of samples were carried out in a Class 1000 clean laboratory under a bench at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). Seven ilmenite grains from sample S503 and four ilmenite grains from sample S404 were manually selected under a binocular microscope. These ilmenite grains have the size ranging from 112 × 114 μm to 368 × 370 μm (Supplementary Fig. 1). For comparison, two silicate grains were also collected from sample S404 and used for microstructural analysis: one pyroxene grain with a size of 282 μm × 326 μm and one olivine grain with a size of 274 μm × 442 μm (Supplementary Fig. 1). The grains from different soil fractions were pressed into two indium mounts, with their relatively flat sides facing upwards. Furthermore, we selected two basalt clasts from sample S404, and then embedded them in epoxy resin and polished them. The ilmenites from these two basalts were used to determine the sputtering rate of NanoSIMS analysis.

Scanning electron microscopy

The surface tomography of the ilmenite grains from sample S503 was performed with a Nova NanoSEM 450, and the grains from sample S404 were observed using a Thermo Scientific Apreo S scanning electron microscope. Both instruments are located at IGGCAS. There are no differences between these two devices when it comes to observing sample surface tomography. Sample mounts were coated with a ca. 10-nm-thick layer of gold before being placed in the instrumental chamber. We used a CBS detector for both instruments to capture the surface topography. The working parameters were an accelerating voltage of 10–15 kV and an electron beam current of 1.6 nA, as shown at the bottom of the BSE images (Supplementary Fig. 1). The optimal working distance was 9–10 mm.

NanoSIMS analysis

D/H ratios and H concentrations of the standards, including San Carlos olivine, SWIFT MORB glass, Durango apatite, Kovdor apatite, and CE5 lunar soil grains, were measured by a CAMECA NanoSIMS 50L installed at IGGCAS. The vacuum in the analysis chamber and the conditions for instrumental analysis are largely similar to our previous studies8,34. During the sample measurements, the vacuum in the analysis chamber was better than 3 × 10-10 torr. We first found a flat area using a low Cs+ ion beam. Then, each 7 μm × 7 μm area was rastered by a Cs+ primary beam of ~500 pA with a beam size of 0.8–1 μm. The secondary anions 1H-, 2H-, and 18O- were simultaneously collected by electron multipliers. An electron gun was applied to compensate for the sample surface charge. The counting time was set to 0.541 s/cycle, and the measurements on the standards and samples consisted of 500–1000 cycles. Two ilmenites from CE5 basalt clasts were used to determine the sputtering rate. Using an atomic force microscope, we determined the depth of the craters (Supplementary Fig. 2) and estimated the average sputtering rate for ilmenite (Supplementary Table 3).

Terrestrial San Carlos olivine grains were used to monitor the background and the potential surface contamination (Supplementary Fig. 10). The 90% of oxygen intensity, when stable, was used as the starting position for the depth-profiling surface (Supplementary Fig. 10). Then, the H counts of samples were calculated by subtracting the H counts of San Carlos olivine. The H counts on the surface of San Carlos olivine is less than 6% of that of the ilmenite soil grains (Supplementary Fig. 11). In addition, the H counts were converted to H2O content using the corrected-H/O ratios multiplied by the slope of the calibration line (Supplementary Fig. 12a). The calibration line was determined from measuring the standards San Carlos olivine (H2O = 1.4 ppm)74, SWIFT MORB glass (H2O = 0.258 wt.%, δD = −73 ± 2‰)75, Durango apatite (H2O = 0.0478 wt.%, δD = −120 ± 5‰)23, and Kovdor apatite (H2O = 0.98 ± 0.07 wt.%, δD = −66 ± 21‰)76. Corrections for instrumental mass fractionation (IMF) on D/H ratios were performed on the Kovdor apatite standard using αIMF = (D/H)measured/(D/H)recommended (Supplementary Fig. 12b). We calculated the D/H ratios in the uppermost ~100 nm (~120 cycles), since D counts of the ilmenite rims are too low to determine. The H isotopes are expressed in delta notation: δD (‰) = [(D/H)sample/(D/H)SMOW − 1] × 1000, where SMOW refers to the Vienna Standard Mean Ocean Water, which has a D/H ratio of 1.5576 ± 0.0001 × 10−4 (ref. 77). The δD values are reported with 1σ, which is derived from counting statistics. More technical details can be traced in Xu et al.8. The results are given in Supplementary Table 1.

Raman spectroscopic analysis

Mineral phases were identified using a LabRam HR Raman microscope at IGGCAS. A solid-state continuous-wave laser emitting at 532 nm was connected to the instrument via an optical fiber. A laser power of ~1 mW was focused on the bottom of the craters produced by NanoSIMS, and we observed the craters using an optical microscope with a ×50 magnification lens. We used a piece of single-crystal silicon as the reference to calibrate the Raman shifts. The Raman spectra from 200 to 1200 cm−1 were shown, and the monospectral collection time was ~10 s and the total accumulation time was 30 s. The diagnostic spectra made it possible to identify different phases (Supplementary Fig. 13).

Transmission electron microscopy (TEM)

The microstructures of space-weathered rims of one olivine, one pyroxene, and four ilmenite grains were examined by TEM. The ultrathin foils were cut and thinned using a Zeiss Auriga Compact focused ion beam (FIB) microscopy with a high voltage range of 5–30 kV and a beam current of 100 pA–2 nA. The slices are ~100 nm in thickness, ~5 μm in width, and ~10 μm in length. These ultrathin foils were observed using a JEOL JEM-2100 TEM. High-resolution TEM imaging (HRTEM) and selected area electron diffraction (SAED) were used to investigate crystallographic structures.

The foils were cleaned with Ar-plasma for 5 min and then analyzed using a scanning transmission electron microscope (STEM)78. High-angle annular dark-field (HAADF) imaging, electron energy-loss spectroscopy (EELS), and energy-dispersive X-ray spectroscopy (EDS) analyses were performed using a FEI Titan Cubed Themis G2 300 spherical aberration-corrected STEM at Tianjin University of Technology. This TEM is equipped with a field emission gun (FEG) and integrated four Super-EDS detectors, which operated at 300 kV and ~90 pA with a convergent semiangle of 21.4 mrad. To detect the helium and H species, EELS acquisitions were performed with a collection semiangle of 56.5 mrad. The energy resolution for EELS is ~0.4 eV, as determined by the full-width at half maximum (FWHM) of the zero-loss peak (ZLP). Gatan Digital Micrograph software was used to perform peak alignment based on shifts in the ZLP, which was used to compensate for systematic energy drift during the test.

Calibration of spallation-produced D and H

The NanoSIMS-measured D/H ratios were corrected by the effects of cosmic-ray-induced spallation, with a D production rate of 2.17 × 10-12 mol/g/Myr (ref. 42) and an H production rate of 4 × 10-10 mol/g/Myr (ref. 79). In this work, we reported the corrected water contents and δD values of CE5 soil grains assuming a cosmic-ray exposure age of 50 Myr, which was adopted from a previous work on water content of CE5 basalts5,7. According to our estimates, the influence of spallation on water content is negligible when compared with the high water content on the surfaces of the soil grains.

Outgassing model

The SW-implanted H in the mineral rims can be lost by diffusion and/or thermal evaporation, which can be modeled by using the Rayleigh fractionation equation, expressed as \(R={R}_{0}\times {f}^{(\sqrt{M1/M2}-1)}\). Here, R0 and R represent the initial and final D/H ratio, respectively; f denotes the fraction of H in the residual phase; M1 and M2 are the masses of the degassed phase. Two scenarios proposed by Sharp et al.80 were considered here to model the changes in H concentrations and isotopes. It is assumed that the degassing of H-containing phases occurs in the form of H2 (M1 = 2.016 for H2 and M2 = 3.022 for HD), and H2O (M1 = 18.015 for H2O and M2 = 19.021 for HDO)80. In addition, a special case was also modelled in which we only consider the loss of H-bearing phase in the form of pure H, but do not include D. Based on the reported values for silicate minerals, the initial endmember is set to be 12,000 ppm for H2O and −950‰ for δD.