Main

Remote observations and analyses of hydrogen isotopes in returned lunar samples indicate that solar wind implantation is a large contributor to the lunar surface water1,2,3,4,5,6,7,8. Laboratory simulations have demonstrated that water could be efficiently generated and released through rapid energetic heating, such as by micrometeorite impacts, into anhydrous silicates implanted with solar wind protons9. The measured deuterium to hydrogen (D/H) isotope ratios of lunar samples could hint at a solar wind origin, but several processes could alter the original D/H ratio, including the production of deuterium through cosmic and galactic ray spallation reactions6,10. So the processes by which solar wind implantation contributes to water formation on airless bodies like the Moon, particularly its variation with radiation flux and temperature across different latitudes and local times, remain poorly understood due to limited observations. Additionally, impact gardening complicates the variation of water content with depth, further obscuring our understanding of the formation and dynamics of water on these bodies. Various remote observations and modelling work indicate that the lunar surface water is enriched towards higher latitudes, varies with time and can escape or migrate, indicating a dynamic cycle1,2,5,11,12,13,14,15. However, other studies argue that the observed spatial and temporal variations of lunar surface water are the result of inappropriate thermal corrections of reflectance data near 3 μm where water absorption occurs16,17 or due to photometry effects18. The Lunar Atmosphere and Dust Environment Explorer observations near the equator (23° N to 23° S) indicate that the lunar surface may contain no more than tens of parts per million of water and that there is a water reservoir in the subsurface at a depth of 10 cm to 3 m (ref. 14). This observation has been questioned by Hodges and Farrell (2022)19, who suggested that the upper limit for water in the lunar exosphere is only ~3 molecules per centimetre cubed. However, the new CHACE-2 observations indicate a much denser water exosphere20. Overall, the observations contradict one another, which further complicates our understanding of how solar wind implantation contributes water to the lunar surface. The subsurface water content and its variation with depth remain critical scientific questions.

Hyperspectral remote sensing and in situ detection can obtain the water content of only the topmost surface (approximately tens of micrometres to millimetres) and cannot reveal the water characteristics of the subsurface (over a few millimetres in depth). However, rocket exhaust is an effective way to expose the shallow subsurface (millimetres to centimetres in depth). Clegg et al.21 suggested that the lander plume in the descent stage could disturb the lunar surface by blowing away the fine regolith from a few millimetres to centimetres in depth. Through a numerical simulation, Zhang et al.22 pointed out that the Chang’e-5 lander plume denuded the subsurface fine regolith, with a maximum denudation depth of ~1.2 cm. A very interesting phenomenon is that finer regolith at the surface contains more water than coarser lunar soils due to their higher area-to-volume ratio23,24, which agrees well with the hypothesis that the lunar surface water could be dominantly sourced from the solar wind implantation of protons. Simulations of temperature profiles with depth, based on the Apollo regolith samples and Diviner measurements, show that there is a strong thermal gradient from the lunar surface to ~1 cm into the subsurface such that the temperature of the regolith decreases by about 30–50 K (refs. 25,26). Therefore, studying the variation of water content and the temperature of regolith disturbed by the Chang’e-6 lander exhaust from different depths will improve our understanding of the processes that contribute to water formation (especially subsurface water) on the Moon.

The first ever lunar farside sampling and return mission, Chang’e-6 successfully landed on the southern mare plain of the Apollo basin. By investigating the high-resolution topographic data recorded by Chang’e-6 (‘Spectral and topographic data’ in Methods), we found that the lander landed on the rim of a crater with a diameter of ~50 m. The images captured by the landing camera (LCAM) and panoramic camera during the descent confirmed that the Chang’e-6 landing site was heavily disturbed (Supplementary Figs. 1 and 2). The fine regolith of the top surface that was disturbed by the exhaust might have moved and been mixed on the topographic high terrain and then deposited onto the downslope terrain. The topographic profile from the landing point to the crater centre first rises and then falls (Fig. 1a,b). The lunar mineralogical spectrometer (LMS) acquired 14 hyperspectral reflectance datasets (0.48 μm to 3.2 μm) from nine locations (Fig. 1a–c), of which eight spectra were repeat collections from three regions of interest (ROIs) at different local times (Supplementary Table 1). The radiance data were calibrated in a similar way as those of Chang’e-5, which included thermal correction, photometric calibration and estimating the water contents (‘Processing of LMS data’, ‘Thermal and photometric corrections’ and ‘Estimating the water content’ in Methods). We conducted a comprehensive examination of the spatial and temporal variations of the water contents at the Chang’e-6 landing site. In regions where the top few centimetres of fine regolith had been swept away by the rocket exhaust, the water content was correlated with depth, thus providing critical information for gaining an understanding of the formation and dynamics of lunar surface and subsurface water.

Fig. 1: Topographic and spectral characteristics of in situ detection by Chang’e-6.
figure 1

a, Three-dimensional topographic view of regions where the LMS conducted spectral measurements. The base map is based on the topographic products of LCAM. Yellow shading represents areas with a relatively higher elevation. and blue shading represents lower regions. The red line marks the full view seen by LMS, and the red triangle is one location of LMS detection. The four-digit numbers correspond to various spectral acquisitions. b, Topographic profile along the sites where reflectance spectra were acquired by LMS, the distance and elevation data of which were extracted from the topographic data recorded by LCAM. The four-digit numbers also correspond to spectral acquisitions. c, All the acquired reflectance spectra before thermal correction. The dashed line marks an absorption near 2.85 μm. The red lines represent spectral group 1, which has a strong upwards tilt at long wavelengths. The blue lines represent spectral group 2, which has little or no prominent thermal contamination. d, Comparison of ~2.85 μm absorption depths of Chang’e-5 and Chang’e-6 spectra after removal of the continuum. Chang’e-5 spectra are shown in orange, and Chang’e-6 spectra are in blue. The dashed black line marks absorption near ~2.85 μm. CE5, Chang’e-5; CE6, Chang’e-6.

Full size image

Results and discussion

Spectral characteristics and water content at the landing site

Figure 1c shows the reflectance spectra for locations near the Chang’e-6 landing site (Supplementary Data 1). These data can be grouped into two groups based on the thermal contamination at wavelengths greater than ~2 μm. One group exhibits a strong upwards tilt in reflectance at long wavelengths (red lines), particularly beyond 2.6 μm, such as the spectra labelled 0003, 0006, 0013, 0014, 0018, 0019, 0020 and 0023. By contrast, the other group shows no prominent thermal contamination (for example, no strong upwards tilts in reflectance at wavelengths >2.6 μm), such as spectra 0008, 0009, 0010, 0012, 0015 and 0021 (blue lines).

Figure 1d demonstrates that Chang’e-6 reflectance spectra have more prevalent water absorption features near 2.85 μm than most of the spectra recorded by Chang’e-5 (except for the deepest orange absorption of 0012, where residual fine particles may persist due to shielding by the CE-5 rock; Supplementary Fig. 5 in ref. 27), and the absorption depths and widths of the former are also notably stronger.

The temperature of the lunar surface should have been very high (340–360 K) at the observation time of LMS, which was close to the local noon (~09:56–11:02)28. It was surprising to see that some of the Chang’e-6 reflectance spectra, such as 0008 and 0021, were not or very little affected by thermal effects. We also found that the spectral shape of the two thermal-free spectra is like the laboratory spectra of Apollo samples (Supplementary Fig. 3). The repeated LMS measurements of the same locations at different local times confirm that the two thermal-free spectra (0008 and 0021) are robust measurements that have high signal-to-noise ratios (Supplementary Fig. 4). The two thermal-free spectra, which may have been recorded at a temperature lower than ~330 K, could be used as the ground truth in the thermal correction of the remote-sensing spectra.

The mean water content of the Chang’e-6 landing site (~75 ppm) of all the 14 spectra at the local time of 10–11 a.m. is notably higher than that for Chang’e-5 spectra (~28.5–46 ppm)7,27,28, which were obtained at nearly the same local time (10–11 a.m.) and the same latitude (Chang’e-5 was at 43.1° N and Chang’e-6 at 41.6° S). These results are consistent with orbital observations, such as those made by the Moon Mineralogy Mapper (M3) (Chang’e-6 found ~82 ppm and Chang’e-5 ~44 ppm)5. Because the spectrometer onboard Chang’e-6 was a duplicate of that on Chang’e-5 and the observation conditions, such as latitude (~42° S/42° N) and local times (both at 10–11 a.m.) between the two spectrometers, were very similar, we are confident that the results from Chang’e-6 reflectance spectra are as robust as those from Chang’e-5. The higher water content observed by Chang’e-6 could be attributed to the older age or distinct surface composition compared with the Chang’e-5 landing site.

The ages of the local basalts at the Chang’e-5 and Chang’e-6 landing sites were dated, respectively, to be ~2 billion years and ~2.8 billion years old29,30,31,32. The mare unit where Chang’e-6 landed was dated around 0.8 billion years older than that of Chang’e-5. Both landed on regions dominated by mare basalts, but the composition of the Chang’e-6 regolith determined by X-ray diffraction (XRD) (‘XRD analysis of lunar soils’ in Methods) indicates that the Chang’e-6 regolith contains a much higher content of glasses (29.4 vol% versus 20.7 vol%)33,34. Although the accuracy of quantifying amorphous glass phases using XRD is lower than that for crystalline mineral phases35, the relative estimate of glass contents remains robust. Given that the mare unit of Chang’e-6 is older and the glass content is higher than that at the Chang’e-5 landing site, it is probable that the soil is more mature (with more defects) at the Chang’e-6 site and, thus, water formation by solar wind implantation is more efficient (‘Difference in water contents between the Chang’e-5 and Chang’e-6 landing sites’ in Methods).

Evidence for subsurface water

The temperatures and water contents derived from the LMS reflectance spectra (Supplementary Table 1) are shown in Fig. 2. According to our analysis of the thermal influence of the lander rocket engine and the lander itself (‘Analysis of the thermal influence of the lander rocket engine and the lander itself’ in Methods), the heat radiated from the engine and the lander itself had a negligible effect in elevating the surface temperature. Our thermal correction model (‘Thermal and photometric corrections’ in Methods) indicates that no thermal correction was needed for the two spectra 0008 and 0021. The high-temperature and low-water-content regions are mostly close to the lander, whereas those with low temperatures and high water contents are further away from the lander.

Fig. 2: Derived surface temperatures and water contents including at ROIs over time.
figure 2

a,b, Derived surface temperatures (a) and water contents (b) from the LMS reflectance spectra overlaid onto a map of the landing site captured by the LCAM. The yellow dashed lines mark the boundary between the erosion zone and the deposition zone. Numbers and ROIs mark where the reflectance spectra were collected. The base map was created based on the Chang’e-6 LCAM Digital Orthophoto Map. c, Our modelled temperatures at ROI-1 and ROI-2 compared with Diviner observations. The temperatures at ROI-3 are not plotted because little or no thermal correction was required. The red curve shows the variation of the lunar surface temperature with local time fitted from the Diviner data for the same latitude with an offset of 7 K. d, Temporal variation of water contents at the three ROIs. The blue curve shows the temporal trend for water content in the range 30–60° N/S in the northern and southern hemispheres observed by M3 (ref. 51) with an offset of −18 ppm.

Full size image

A possible scenario (Fig. 3) could be that the exhaust plume of the lander first swept the fine-grained mature regolith of the top surface in all directions during the descent. As it approached closer to the surface, the rocket exhaust may have dissipated more gases and pressure to the surface, which may have continued to blow the fine-grained regolith from the shallow subsurface (centimetres in depth)22,36,37,38. Note that the subsurface regolith may be much cooler than the regolith on the surface due to the strong thermal insulation by the fine regolith on the top surface25,26. The grains from the erosion zone in region I that were disturbed by the exhaust when the lander was still high above the ground moved at low speeds along the surface and were more often deposited in the nearby downslope deposition zone (region II), such as 0009, 0010, 0012 and 0015. The mixing of hot surface regolith with cold subsurface particles could have resulted in notably lower temperatures than at the hot surface before the redistribution of fine particles. Pictures captured by the camera during the descent (Fig. 3b,c) indicate that only the fine regolith was blown away, rather than the entire surface layer, by the rocket exhaust (for example, 0003, 0006, 0014 and 0013), and thus, most of the hot surface before the redistribution of the fine particles could have been preserved. During the very late descent stage, the rocket exhaust could have dissipated more gases and pressure to the relatively deeper and colder subsurface regolith, which enabled the regolith to move faster and further, resulting in more deposition in topographically low regions farther away (region III) from the lander, such as 0008 and 0021. Note that the topographic variations at the Chang’e-5 landing site were very small before and after landing, and all the in situ reflectance spectra were obtained in the erosion zone where the top fine regolith layer had been blown away (Supplementary Fig. 2). The differences in the erosion zones of the two landing sites indicate that the erosion zones could have been affected by both the rocket exhaust and the local topography. Modelling results combined with measurements from Apollo regolith samples and the Diviner mission indicate that the regolith temperature decreased by around 30–50 K from the surface to around 1 cm in depth25,26,39. Moreover, 0009, 0010, 0012, and 0015 in region II were around 20 K colder than 0003, 0006, 0013 and 0014 in region I, which could be attributed to the mixing of surface and subsurface regolith. By contrast, 0008 in region III was at least 30 K colder than region I, which might indicate that most of the regolith in region III could be from the deep subsurface (depth of a few centimetres).

Fig. 3: Disturbance of the fine regolith by the rocket exhaust from the surface to around a depth of a centimetre during the descent.
figure 3

a Schematic of the landing. The red solid line represents the topography of the landing area derived from the LCAM data. The light blue circles represent the subsurface fine regolith with low temperature and moderate water contents. The dark blue circles represent the surface fine regolith with high temperature and high-water contents. b,c, Single images captured by the Chang’e-6 LCAM at altitudes of ~7 m (b) and ~5 m (c) when approaching the landing site. The effect of the rocket exhaust on the surface surrounding the landing site can be seen. As the lander descended closer to the surface, the lower-reflectance zone became increasingly rougher, whereas the higher-reflectance zone became smoother. d, Double-layer water-content model for the surface and subsurface fine regolith before the disturbance by the exhaust plume.

Full size image

It has been well documented that many volatiles, including water, exhibit a strong correlation with the surface-area-to-volume ratio (the so-called surface correlation)23,24. Therefore, region II, where the surface fine regolith was deposited, shows much higher water contents at around 100–150 ppm, such as 0009, 0010, 0012, and 0015. By contrast, region I, where many fine regolith grains were blown away, exhibits relatively low water contents at around 50–60 ppm, such as 0003, 0006, 0013 and 0014. It is very interesting to see that region III, which is covered with fine regolith that could be from the deeper subsurface (depth of a few centimetres), showed very low surface temperatures and moderate amounts of water at around 60–70 ppm, such as 0008 and 0021.

We established a double-layer water-content model for the surface and subsurface fine regolith of the Chang’e-6 landing area (Fig. 3d). According to the model, the average water content in regions I and II near the lander represents the surface water content, and the average water content in region III represents the water content of the subsurface fine regolith. Taking the local time of 10:00 a.m. ± 3 min on the Moon as an example, the average surface water content is ~105 ppm (the average water content of 0003, 0006, 0009 and 0010), whereas the average water content of the subsurface fine regolith is ~76 ppm (the water content of 0008).

Temporal variation of water content

The blue curve in Fig. 2d shows the diurnal variation of water content observed by M3, which demonstrates that the water content begins to decrease as the local time moves off the morning terminator. It has a minimum value near the local noon and returns to high values by evening. Figure 2d also shows the strong temporal variation of the water content in repeat measurements by the LMS of the same regions but at different local times. Although the LMS and M3 have comparable spectral ranges (LMS: 480–3,200 nm and M3: 420–3,000 nm) and similar detection depths (tens of micrometres), their spatial resolutions differ drastically (LMS: tens of centimetres and M3: ~140 m). Consequently, we did not expect the observed temporal hydration trends to be consistent. Despite the large difference in spatial resolution, both instruments reveal consistent temporal trends for the water content. This indicates that solar wind-induced hydration processes are generally homogeneous in terms of water formation from the submetre to ~100 m spatial scales, which can be attributed to the gardening and mixing processes due to micrometeoroid bombardment that have operated on the surface at the Chang’e-6 landing site for billions of years. It is also interesting that the temporal variations observed at ROI-1 and ROI-2, which are closer to the lander than ROI-3, showed a slightly faster rate of variation in the water content. Thus, the slope of the water content decreased with the local time faster than observed by M3 (ref. 39). By contrast, ROI-3 exhibited a slightly lower rate of variation of the water content compared with that observed by M3.

Plume impingement and redistribution of the low-temperature fine-grained regolith may be the main cause for the different rates for the temporal variation of the water content at the three ROIs (Fig. 2d). ROI-1 and ROI-2 are both in the erosion zone, where most of the surface fine regolith could have been blown away by the exhaust, whereas ROI-3 is in the deposition zone, which could have been covered by fine regolith mostly from the subsurface. The fine-grained and water-rich lunar regolith on the top surface in the erosion zone was blown away, resulting in a decrease in water content. The consolidated regolith left in the erosion zone continuously underwent pitting, clump-lifting and local scouring by the lander plume, which may have resulted in an increase in the interparticle compaction and an increase in thermal conductivity. The high temperature and increased thermal conductivity of the consolidated regolith led to a higher rate of water loss but a slight increase in the surface temperature (the solar energy was conducted to the subsurface). By contrast, ROI-3, where low-temperature, fine-grained subsurface fine regolith was deposited, water loss was naturally slowed down due to the low temperature.

Conclusion

Chang’e-6 reflectance spectra show more prevalent water absorption features near 2.85 μm, and the absorption depths and widths are also notably stronger than most of those recorded by Chang’e-5. The average water content at the Chang’e-6 landing site is notably higher than that at the Chang’e-5 landing site. Areas affected by the rocket exhaust showed distinct temperature and water-content patterns. The water content of the very top surface is higher than that of the fine regolith exposed in the subsurface. The double-layer water-content model of the surface and subsurface could explain these observations. The repeated measurements made by the LMS of the same regions but at different local times exhibit a temporal variation of the water content.

The above difference in the water contents could be interpreted as being due to correlations with the glass contents, particle sizes, depths and local times, although more observations are needed to untangle these parameter dependencies. Our findings highlight the roles of solar wind implantation and impact gardening in water formation and its distribution in the lunar surface and subsurface, with implications for other airless bodies like Mercury, Vesta and the near-Earth objects. The fine regolith from the lunar surface to subsurface depths of millimetres to centimetres or deeper will probably be an important source of lunar water.

Methods

Spectral and topographic data

The LMS of Chang’e-6, which was inherited from Chang’e-5, consists of a visible and near-infrared imaging detector (480–950 nm) and three infrared single-point detectors (for the near-infrared, 900–1,450 nm; short-wavelength infrared, 1,400–2,450 nm; and mid-wavelength infrared (MWIR), 2,400–3,200 nm). Like Chang’e-5, the Chang’e-6 LMS has three operating modes: (1) full-band observation, (2) full-view scanning and multispectral observation and (3) in-flight calibration28. The main characteristics of the LMS on Chang’e-6 are the same as those of Chang’e-5 (refs. 27,28,40), and it can detect OH/H2O absorption at ~3 μm.

To observe the spatial and temporal variation of the water contents, we used two LMS investigation strategies. In the first, we investigated different locations from near to far from the lander. In the second, we observed the same regions at different local times. The LMS first began to observe the lunar surface at 05:35 utc on 2 June 2024, and observations ended at 14:22 utc on 3 June 2024. The working time span was ~33 h, corresponding to ~1 h on the lunar surface (local time: 09:56–11:02) (Supplementary Table 2). During this period, the LMS collected 21 spectra, including 1 self-check, 4 in situ calibrations and 16 lunar surface targets. The spectra of the 16 targets included one detection of the sampling site before sampling (0003) and two detections after sampling (0017 and 0022, which were not analysed in this paper due to the large number of shadows in the field of view), as well as eight observations of the three regions of interest (ROI-1, ROI-2 and ROI-3) at different local times. The nearest to the landing centre was ROI-1 (0006 and 0018). ROI-2 was at an intermediate distance (0014, 0019, 0020 and 0023), and the farthest was ROI-3 (0008 and 0021). In addition, there were five detections of other regions (0009, 0010, 0012, 0013 and 0015).

The Chang’e-2 global lunar topography data (CE2TMap2015)41,42, which have a spatial resolution of 7 m, was selected as the geographic reference datum. Using 142 sequential LCAM images and 120 pairs of stereo images recorded by the panoramic camera, the terrain data for the Chang’e-6 landing area was reconstructed based on photogrammetry principles43. The LCAM terrain data, which have a spatial resolution of 1 cm, cover an area of 2.4 km (east–west direction) times 3.0 km (north–south direction) around the landing site. Terrain data recorded by the panoramic camera have spatial resolutions of 1 mm or 3 mm and extend 18 m north of the landing site (Supplementary Fig. 2). The topographic map indicates that there is relatively high terrain close to the Chang’e-6 lander, and a downslope begins at a distance of ~2.8 m from the landing centre (Fig. 1).

Processing of LMS data

As with the Chang’e-5 data, the level 2B radiance data of LMS were used in this study, which underwent dark-current subtraction, scattering-background correction, flat-fielding, instrumental temperature correction, radiometric calibration (including in situ radiometric calibration) and geometric correction. The radiance factors for the lunar surface were calculated using the solar irradiance44, and we used in this study data from the three infrared detectors in the range from 1,000 nm to 3,100 nm. Note that we also corrected for the spectral jitter caused by the different responses of the three detectors in the connection bands. Considering the high signal-to-noise ratio of the MWIR spectra, we took the MWIR data as the standard and adjusted the data from the other two detectors to the MWIR data to obtain one continuous spectrum28. Altogether, 14 continuous hyperspectral datasets (except 0017 and 0022) recorded by LMS were used for subsequent data processing, including thermal correction, photometric correction and estimating the water content.

Thermal and photometric corrections

The radiance measured by the spectrometers at wavelengths beyond ~2 μm typically encompasses both solar-reflected and thermally emitted components originating from the lunar surface. The thermal effects induce an upward spectral curvature at longer wavelengths, which can alter or obscure the spectral characteristics of OH or H2O. Consequently, this phenomenon impedes the precise identification and quantitative analysis of OH and H2O absorption features. So, the thermal effect of the lunar surface after 2,000 nm cannot be ignored. We used an empirical model developed by Li and Milliken39 to remove the thermal effects from the measured LMS spectra. The model is based on laboratory reflectance data for Apollo and Luna samples, which have a wide range of compositions and maturities. Two spectral bands, 1.55 μm and 2.54 μm, which are hardly absorbed by lunar materials, were selected to establish the empirical relation. The ‘true’ radiance at 2.54 μm could then be predicted. Any excess radiance in the LMS measurements was attributed to thermal emissions, and the temperatures were then derived from the Planck function. Using the calculated temperatures and the Planck function, the influence of thermal radiation the in other bands was removed. This empirical model has been successfully applied to M3 and Chang’e-5 spectral data5,27,28.

The Hapke model was used for the photometric correction of all the LMS spectra after the thermal correction. In this paper, a two-term Henyey–Greenstein function was used to fit the phase function, with parameters b (describes the angular width of the backward and forward scattering lobe) and c (describes the relative proportions of the backward and forward scattering) taking the global mean values of 0.3 and 0.2, as obtained by the wide angle camera onboard the Lunar Reconnaissance Orbiter Camera45. The final product of our calibration pipeline was thermally corrected reflectance (the radiance factor) spectra at a standard viewing geometry (i = 30, e = 0 and g = 30), where i is the incidence angle, e is the emission angle and g is the phase angle.

Estimating the water content

The method used to estimate the water contents from Chang’e-5 spectra was also used for Chang’e-6 spectra5,27,28. The 2.85-μm band was chosen as the absorption centre of water, and the Hapke model was selected to calculate the single-scattering albedo and the effective single-particle absorption thickness (ESPAT). The water contents were then estimated, as shown in Supplementary Table 1.

The detection limits of water content were also estimated from the ESPAT parameters using the same method used for Chang’e-5 data5,28. According to the in situ calibration data for LMS, the signal-to-noise ratio for Chang’e-6 LMS data is ~60, and the error for the water-content estimation model is about 20%. We assumed that noise has been added to the true signal at 2.85 μm after continuum removal, and this noise was subtracted from the true signal to provide an upper limit for the water-content error that could have been introduced by noise. We then estimated the ESPAT values for the pseudo-OH/H2O absorption, and the water content was derived as the detection limit. The detection limit of the water content in the fine-grained regolith was about 60 ppm. Therefore, we choose a detection limit of 60 ppm as a conservative value.

The volatile substances generated by the lander plume primarily include water and carbon dioxide. These exhaust volatiles were typically hot and had high initial velocities. The high-speed exhaust disturbed the surface dust and underwent thermalization upon interaction with the lunar regolith. Although water from the plume may have temporarily been adsorbed onto the lunar surface before the LMS observation, the surface experienced extremely high temperatures (~360 K). Under these conditions, water adsorbed from the plume would have rapidly volatilized to near 0 ppm within approximately 30 min. As the LMS began spectral acquisition ~7 h after landing, all plume-adsorbed water would have already dissipated. Consequently, the spectral data obtained by the LMS are unlikely to be affected by residual plume-derived adsorbed water28,46. Additionally, isotopic studies of returned Chang’e-5 and Chang’e-6 samples did not find signatures of water from the rocket exhausts.

XRD analysis of lunar soils

Three lunar soil samples taken by Chang’e-6, numbered CE6C0000YJFM00107, CE6C0000YJFM00108 and CE6C0000YJFM00109, were prepared for the XRD analysis. The preparation and measurement procedures for the samples were like those used for lunar soils returned by Chang’e-533. The XRD patterns and the Rietveld full pattern fitting results produced by the software Jade for the three soil samples are shown in Supplementary Fig. 5. The fitting errors (weighted residual error) of the three samples were 5.15%, 5.83% and 5.14%, respectively, with the maximum error not exceeding 5.9%. The phases identified and involved in fitting the whole pattern Rietveld included plagioclase, augite, pigeonite, enstatite, ilmenite, olivine, other minerals and amorphous glasses. The standard diffraction patterns of each mineral phase are from the International Center for Diffraction Data, and the corresponding card number for each mineral phase is listed in Supplementary Table 3. The phases, abundances and measurement uncertainties of Chang’e-6 lunar soil samples are shown in Supplementary Table 4 (ref. 34).

Difference in water contents between the landing sites

The Chang’e-5 and Chang’e-6 data were acquired at nearly identical lunar local times (10–11 a.m.) and latitudes (Chang’e-5: 43.1° N and Chang’e-6: 41.6° S). However, there are distinct differences in the water contents and distribution characteristics between the two landing sites. The Chang’e-5 site has an overall lower water content with greater variability, whereas the Chang’e-6 site has a higher water content with less variability. This disparity may arise from two factors: (1) Supplementary Fig. 2 reveals that the Chang’e-5 landing area is relatively flat, with the entire LMS survey region situated within an erosion zone, where water-rich fine-grained regolith was entirely removed by the plumes, resulting in a low bulk water content (except for sample 0012, within which residual fine particles may have persisted due to shielding by the CE-5 rock; Supplementary Fig. 5 in ref. 27). By contrast, Chang’e-6 landed near a crater rim. The LMS survey area includes slopes, thus creating both erosion and deposition zones. The plume redistributed water-rich fine particles from the erosion zone to the deposition zone, whereas the cooler subsurface temperatures facilitated the distal settlement of moderate-water-content fine particles. (2) Laboratory XRD analyses of samples returned by Chang’e-5 and Chang’e-6 indicate a higher glass content in the Chang’e-6 samples. As lunar glass is an important reservoir for surface water8, the differences in glass contents probably contributes to the systematically higher water contents found by Chang’e-6.

Analysis of the thermal influence of the rocket engine and lander

We conducted a quantitative assessment of thermal contributions from both the rocket engine and the lander itself. After the rocket engine on the lander had shut down on the lunar surface, the nozzle cooled primarily through thermal radiation. We assumed that the nozzle cooled from 1,500 K to 500 K, and we compared the total radiated energy with the energy received from the Sun to estimate the heat radiated from the lander itself. We assumed that the initial nozzle temperature Ti = 1,500 K, final nozzle temperature Tf = 500 K, nozzle mass m = 50 kg, specific heat capacity C = 460 J kg−1 K−1 and solar irradiance PSun = 1,361 J s−1 m−2. The total thermal energy loss, assumed to be entirely radiated away, was calculated as:

$${{{E}}}_{{\rm{nozzle}}}={{m}}{{C}}({{{T}}}_{{\rm{i}}}-{{{T}}}_{{\rm{f}}}).$$
(1)

Thus, the nozzle radiated approximately 23 MJ as it cooled from 1,500 K to 500 K. We assumed a cutoff temperature at 500 K, as the heat radiated below 500 K was negligible. Then, we assumed that the total energy was distributed uniformly over a sphere of radius r = 2 m. The surface area of the sphere Asphere = 4πr2 ≈ 50 m2. The energy received per unit area Ereceived = Enozzle/Asphere ≈ 457,700 J m−2. This energy is equivalent to around 6 min of solar illumination:

$${{t}}={{{E}}}_{{\rm{received}}}/{{{P}}}_{{\rm{Sun}}}\approx 336\,{\rm{s}}\approx 6\,{\rm{min}}.$$
(2)

The thermal diffusion modelling results26,39 indicate that the lunar surface temperature may have increased by only around 1 K after around 1 Earth hour of solar illumination at the Chang’e-6 landing latitude (~43.1° S) and the landing local time at around 10 a.m. So, the heat radiated from the lander nozzle as it cooled had a negligible effect in elevating the surface temperature at the Chang’e-6 landing site.

For Chang’e-5, studies indicate that the temperature of the lander itself was around 370–380 K (ref. 47), which was very close to the surface temperature at the landing site and so would not heat up the lunar surface.

About 7 Earth hours passed between the landing and the measurements made by the LMS, so the lunar surface temperature could have increased by around 7 K before the LMS measurements, but the temperature difference between 0008 and 0013 could have reached >30 K. Why had the temperature in the top tens to hundreds of micrometres depth in the lunar regolith not equilibrated? The difference in emission angles may be a possible reason. We checked the emission angles between patches 0008 and 0013, and the difference could be as high as ~18° (the emission angle of 0008 was 63°, whereas that of 0013 was ~45°). However, according to the observation results of Bandfield et al.48 (Fig. 7 in ref. 48), the temperature difference caused by an ~18° difference in emission angles was mostly less than ~10 K (between ~40° and 60° emission angles), so it is still hard to explain an ~30 K temperature difference observed by CE-6. Besides, we also compared the temperatures of 0009, 0010, 0012 and 0013, which have similar emission angles (45°–51°), and the temperature differences could still have been as high as ~30 K (Supplementary Table 1).

According to the lunar thermal model of Vasavada et al. (1999, 2012)25,26, which has been validated by the Diviner data, the surface regolith density (ρ) varies with depth and the thermal conductivity (K) varies with depth and temperature. So, for a silicate surface regolith (ρ ≈ 1200 kg m−3 and K ≈ 9.22 × 10−4 W m−1 K−1), it may take ~56 Earth hours for a heat pulse to penetrate 2 cm. In this case, the temperature in the top tens to hundreds of micrometres of lunar regolith would not yet have equilibrated.