Mature lunar soils from Fe-rich and young mare basalts in the Chang’e-5 regolith samples

Space weathering on airless bodies produces metallic iron (Fe0) particles in the rims of mineral grains, which affect visible and near-infrared spectra and complicate the identification of surface materials. The Chang’e-5 mission provides an opportunity to couple information gained from its returned samples with in situ observations and orbital monitoring to gain insight on the details of space weathering on extremely Fe-rich basalts. By putting together all these data, we could extract a soil maturity index (Is/FeO) at the Chang’e-5 landing site of ~66 ± 3.2, indicative of a formation age for the Xu Guangqi crater, whose ejecta dominate the site, of 240–300 Myr ago. In addition, abundant large Fe0 particles were found in the sample, indicating that both the inherited Fe0 particles from late-stage mare basalts and the dense clustering of oversaturated Fe0 in extremely FeO-rich (>17 wt%) basalts contribute to observed Fe0 abundances. We suggest that space weathering of Fe-richer basalt generates Fe0 particles with a larger grain size and faster production rate. A multi-observational study including laboratory analysis of the Chang’e-5 mission samples, in situ measurements and orbital datasets determined the high level of maturity and iron content of the Chang’e-5 landing site regolith. Heavily processed by space weathering, it mostly comes from the nearby Xu Guangqi crater, formed 240–300 Myr ago.

Space weathering occurs on the surface of airless planetary bodies. Surface materials are gradually altered physically and chemically by the bombardment of the solar wind, cosmic rays and (micro)meteoroids of various sizes 1,2 . These alterations can be characterized by visible and near-infrared (VNIR) reflectance spectroscopy 3 . The Moon is just such a body with a tenuous atmosphere; thus, space-weathering products are extensively distributed on the lunar surface. Large rocks are breaking down into smaller fragments (comminution) and the finest fraction of the regolith becomes welded by impact melt glass, which serves to increases the particle size of the regolith. Lunar regolith with a thickness up to ten metres formed from lithic sources by comminution and agglutination is the best evidence and witness of space-weathering effects on the lunar surface 4 . The products of space weathering, including agglutinitic glasses and metallic iron (Fe 0 ) particles, complicate the spectroscopic properties of the lunar surface with darkening and reddening of VNIR reflectance spectra as well as weakening of diagnostic absorption characteristics of minerals 5,6 . Therefore, space weathering makes it more difficult to quantify mineralogy from VNIR reflectance spectra. The degree of space weathering (soil maturity) must be quantified to reduce the incurred errors of mineralogical retrieval. Moreover, soil maturity can also reflect how long the surface materials have been exposed to space, which indicates that estimation of soil maturity will be promising for the determination of exposure age 7,8 .
Previous research on space weathering has focused on returned samples in the laboratory and orbital remote sensing datasets while lacking in situ spectral measurements on the lunar surface. The particle Article https://doi.org/10.1038/s41550-022-01838-1 into ten parts (CE5C0100-CE5C1000) 15 . Two CE-5 soil samples (CE5C0400YJFM00501 and CE5C0600YJFM00301) are allocated to Shandong University (SDU) that enabled this study. The CE5C0600 sample (<1 mm) was used to investigate the bulk spectroscopic behaviour of the scooped CE-5 soils ( Supplementary Fig. 1a). The finest portion (<25 μm) was sieved from CE5C0400 sample to represent the upper lunar regolith ( Supplementary Fig. 1c). We measured VNIR reflectance spectra at 300-2,500 nm ( Supplementary Fig. 1b,d) using the Analytical Spectral Devices (SDU-ASD). More information about SDU-ASD measurements of CE-5 soils are described in Methods. The spectra of the fine soils (<25 μm, sieved from CE5C0400) and coarse soils (<1 mm, CE5C0600) were compared to investigate the spectroscopic effects of particle size for CE-5 soils.
The VNIR reflectance data of scooping sites on the lunar surface acquired by the lunar mineralogical spectrometer aboard the CE-5 lander (CE5-LMS) were also used in this study. In situ spectra of lunar soils were acquired before and after sampling. The CE5-LMS acquires spectral information in two modes (details in Methods): full-view scanning (multispectral) mode and full-band target (hyperspectral) mode. In total, 11 hyperspectral image cubes (256 × 256 × 100 spectra) at 480-950 nm and single-pixel spectra at 900-3,200 nm were size, porosity and bulk density are critical physical factors affecting VNIR reflectance spectra of lunar soils 9,10 . These parameters obtained in the laboratory are not representative of soils on the lunar surface because the primary state of the lunar soils has been disturbed during the sampling procedure 11 . Thus, in situ measurements of the lunar surface that maintain the primary state of the lunar soils will be more representative as 'ground truth' for lunar orbital spectral data. During the Chang'e (CE)-3 and CE-4 missions, in situ spectroscopic assessments for space weathering were achieved by Yutu rovers without disturbance and destruction of the targets 12,13 . However, there were no returned samples from CE-3 and CE-4 landing sites and thus laboratory data are not available for cross validation. Recently, the CE-5 spacecraft successfully landed in northern Oceanus Procellarum (51.92° W, 43.06° N; Fig. 1) on 1 December 2020 14 , conducted in situ spectroscopic measurements and returned regolith samples. Thus, the CE-5 mission provides an unprecedented opportunity to investigate the difference in space-weathering effects between spectra acquired on the lunar surface and measured in the laboratory.
Lunar soils with a mass of 1,731 g were returned by the CE-5 mission. The scooped samples (CE5C) were sieved to separate rock fragments and soils (<1 mm size fraction). The soils were further divided Article https://doi.org/10.1038/s41550-022-01838-1 acquired by the CE5-LMS, including 6 spectra of pre-sampled surface, 3 spectra of post-sampled surface and 2 spectra of rock (Shi Gan Dang), as shown in Fig. 2. The orbital hyperspectral data from the Moon Mineralogy Mapper aboard Chandrayaan-1 (CH1-M 3 ; ref. 16 ) and data from the Multiband Imager aboard Selenological and Engineering Explorer (SEL-MI) are used to estimate soil maturity and concentrations of Fe 0 particles at the CE-5 landing site and surrounding craters (~1.5 × 1.5 km area; Fig. 1a and Supplementary Fig. 2).
Morris 7 demonstrated that I s /FeO is an index of relative soil maturity (that is, surface exposure age). The ferromagnetic resonance signal intensity (I s ) is normalized by the total iron (FeO) content of the material. The magnetic intensity predominantly comes from fine-grained iron metal in the 5-35 μm particle size fraction. The I s /FeO can be used to reflect the surface exposure time of lunar soils, and three groups were defined: immature (0.0 ≤ I s /FeO ≤ 29.0), sub-mature (30.0 ≤ I s /FeO ≤ 59.0) and mature (I s /FeO ≥ 60.0) soils.
According to the size and origin, the Fe 0 particles were further divided into three classes 17 : (1) nanophase Fe 0 (npFe) particles with a diameter range of 4-33 nm reduced from ferrous iron by vapour deposition and irradiation effects 4 ; (2) microphase Fe 0 (mpFe) particles with diameters >33 nm derived from metallic phases of micrometeoroids; and (3) mpFe particles concentrated from parental rocks. The npFe and mpFe particles are collectively known as submicroscopic iron (SMFe) particles, which are referred to as the whole Fe 0 particles in the lunar regolith. A physical theory was derived by Hapke (2001) (ref. 18 ) to describe the optical effects of Fe 0 particles. Lucey and Riner 19 further demonstrated different space-weathering effects of Fe 0 particles with various sizes, showing that npFe darkens the reflectance and reddens the slope, while mpFe only reduces total reflectance. A modified Hapke radiative transfer model using Mie theory for modelling space-weathering effects 19 is used in this study, which is similar to that used by Trang and Lucey 20 .

Results
The soil maturity of the CE-5 landing site is first characterized by orbital datasets. The 280 × 280 m area around the landing site in CH1-M 3 data exhibits an npFe abundance of 0.47 ± 0.19 wt% (mean ± s.d.), which is consistent with the observation of the 28 × 28 m local area from SEL-MI (0.51 ± 0.03 wt% npFe) ( Table 1 and Fig. 3). The space-weathering maps (spatial resolution: 30 pixels per degree) 20 derived from SEL-MI datasets present npFe content (0.50 wt%) consistent with ours (Methods and Supplementary Fig. 9). These orbital measurements indicate that soils at the CE-5 landing site are mature with I s /FeO values of 65-70 (the calculation of the I s /FeO is described in Methods).
The soil maturities predicted from CE5-LMS in situ data are shown in Table 1. The modelled spectra are compared with the spectra measured by CE5-LMS in Supplementary Fig. 3. The average npFe abundance of the pre-sampled surface at the CE-5 landing site is 0.23 ± 0.03 wt% and the post-sampled surface exhibits similar npFe content (0.21 ± 0.02 wt%) within the uncertainties. In terms of the distribution of npFe abundances near the CE-5 lander (Fig. 4), the average npFe abundance (~0.27 ± 0.09 wt%) of the CE5-LMS multispectral scanning area is consistent with the results derived from the CE5-LMS hyperspectral data (within errors). In the scanning area, the Shi Gan Dang rock is the freshest material with obvious diagnostic absorption features and high albedo. The spectra of this rock yield an npFe abundance lower than 0.07 wt% and a maximum I s /FeO value of 9.7. The average npFe content and I s /FeO value of materials at the CE-5 landing site are 0.23 ± 0.07 wt% (in situ measurements) and 31 ± 4.2 (average value from nine hyperspectral measurements on soil targets), which is approximately half that of the values from orbital observations ( Fig. 3 and Table 1).
This disparity between orbital observations and in situ measurements is reasonable given that rocks and soils in the detection region of CE5-LMS have probably been blown by rocket exhaust during the descending of the CE-5 lander. According to previous research 1, 21 , the surface layer of lunar soils undergoes space weathering more severely than subsurface soils. Thus, the undisturbed surface before the landing of CE-5 mission ought to contain more products of space weathering. Hence, it can be inferred that primordial surface soils before the CE-5 landing are more mature than soils exposed after the landing. The calculated I s /FeO value of ~31 based on CE5-LMS datasets is 40%-50%  . The 40%-50% maturity reduction of the exposed soils caused by rocket scouring is consistent with what has been observed for the CE-3 landing site 13 . The CE-5 mission successfully brought back lunar soils and allowed us to assess the soil maturity with VNIR spectroscopy in the laboratory. The npFe content (0.11 ± 0.08 wt%) and I s /FeO (16 ± 11) of CE-5 soils derived from terrestrial spectra (Supplementary Fig. 7 and Supplementary Table 2) are lower than those results derived from in situ measurements. This difference between the laboratory and the lunar surface is possibly caused by variation in roughness and/or spot sizes of different measurements. The particle size of the coarse grains (~1 mm) in the CE-5 soils is comparable to the spot size of the SDU-ASD (0.6 mm), but much smaller than the field of view of CE5-LMS (143 × 143 mm at 2 m distance 22 ). The surface of the CE-5 soils will be rougher at the length scale of the SDU-ASD light spot than that of the CE5-LMS observations.
The grain size effects on spectral properties of lunar soils are further investigated with CE-5 samples. The fine fraction (<25 μm) of CE-5 soils presents an increase in maturity by ~76% compared with the bulk soil samples (<1 mm). This is coincident with observations in Apollo soils that npFe abundance and I s /FeO values increase with decreasing particle size 23,24 . The fine fraction (<25 μm) of lunar soils exhibits different optical properties from coarse particles and rocks 25 , such as redder slope, weaker absorptions, more space-weathering products and so on. Besides, the average ASD spectra of CE-5 bulk soil (<1 mm) and fine fraction (<25 μm) exhibit reddening and darkening effects on reflectance with decreasing particle size ( Supplementary Fig. 8). The fine fraction of CE-5 soils (Supplementary Fig. 6 and Supplementary Table 2) contains ~0.46 ± 0.04 wt% npFe (I s /FeO of 64 ± 5.6), which falls within the I s /FeO range of orbital measurements. This confirms that the VNIR reflectance spectra observed by orbital spectrometers are indeed dominated by the finest portion of lunar regolith (that is, the uppermost layer of the lunar surface). Although disturbed by rocket exhaust and sampling processes, the finest portion of the returned samples can still reproduce the spectroscopic behaviour and maturity index of soils on the lunar surface. The average I s /FeO value from the fine fraction of returned samples agrees with orbital results at the CE-5 landing site (~66 ± 3.2). The difference between the spectral based I s /FeO values for the finest fraction of the returned sample and in situ soils determined by CE5-LMS is 40-50%, which is similar to the difference measured for the soils before/after the CE-3 landing 13 . In addition, Pieters et al. 26 use detailed compositional, petrographic and spectroscopic data for lunar soils produced by the LSCC (Lunar Soil Characterization Consortium) to create statistically optimized formulations of links between spectral and mineral parameters. The maturity of the finest fraction of the returned sample is also assessed by the statistically optimized formulation (equation (21)) 26 with the derived I s /FeO value of 77 ± 18, which is well consistent with our modelled results. In summary, the soil maturity derived from the fine fraction of the CE-5 samples, in situ measurements and orbital observations of CE-5 landing site are cross-validated.

The main source crater of CE-5 soils
The Xu Guangqi crater (diameter ~0.40 km) lies ~200 m northwest of the CE-5 landing site. The ejecta from Xu Guangqi crater exhibit a spatial continuity with the CE-5 landing site (Fig. 3), which indicates a possible primary source for CE-5 soils. An ~200-m-long and ~50-m-wide ejecta ray from Xu Guangqi crater to the CE-5 landing site was observed in the SEL-MI npFe abundance map. The ejecta ray deposited on the smooth mare plain outside Xu Guangqi crater (including the CE-5 landing site) exhibits similar maturity (average npFe abundance of ~0.45 wt%) to the fine fraction of CE-5 returned soils (0.46 ± 0.04 wt%). The sunlit side of the Xu Guangqi crater wall contains a lower npFe abundance of ~0.30 wt%. Other surrounding small craters also exhibit lower npFe abundances on their walls than their ejecta. The lower npFe abundance on crater walls possibly originates from continuous mass-wasting process 27 on the sloped walls induced by gravity and seismic shaking 28 . Thus, the wall of a crater is usually fresher and younger than its ejecta region 8,29 . Alternatively, the variation of viewing geometry induced by the crater topography may also be responsible for the difference Article https://doi.org/10.1038/s41550-022-01838-1 in spectral slope and the estimated npFe content between the walls and ejecta. Assuming that the smooth ejecta area of Xu Guangqi crater is not obviously affected by subsequent foreign contamination of other impact events, the exposure time of these ejecta materials (including soils returned by CE-5) should be able to constrain the age of Xu Guangqi crater.

The age of Xu Guangqi crater
In general, the longer the exposure of the lunar soils, the more space-weathering products will be formed.  33 . This spectroscopy method can be further developed for the age estimation of other young craters with well-constrained maturity data in the future.

Abundant mpFe in Fe-rich basalts
In addition to the exposure age, the amount of Fe 0 in lunar soils is also affected by the composition (for example, FeO) and inherent mpFe content of their parent rocks 17,34,35 . Two potential explanations were proposed for the abundant mpFe: (1) the exogenic space-weathering process and (2) the endogenic magma crystallization process. The abundances of mpFe (≥0.97 wt%) in the extremely iron-rich CE-5 soils (22.5 wt%) derived from the VNIR spectral data are obviously higher than those in Apollo and Luna soil samples (~0.34 wt%, measured by a combination of static magnetic and ferromagnetic resonance techniques 17 ). The global space-weathering maps also indicate more abundant mpFe (~1.3 wt%) at mare surfaces where FeO abundance exceeds 17 wt% (ref. 20 ).
This may indicate that FeO-richer basalts (22.5 wt% FeO at the CE-5 landing site) could produce more abundant and larger Fe 0 particles. The diagram of mpFe abundance versus I s /FeO of soil samples (Fig. 5) implies the production rate (the slope) of mpFe during the maturation of regolith. The production rate of high-FeO (>18 wt%) Apollo and Luna soils is larger than that of low-FeO (<18 wt%) soils 17 . The largest production rate for mpFe is observed in the most FeO-rich soils returned by the CE-5 mission. Trang and Lucey 20 suggested that the agglutinates and the glassy rims of mineral grains cannot accommodate anymore npFe when the npFe abundance reaches the saturation limit and the npFe particles would merge to become mpFe particles, that is, the dense cluster hypothesis 36,37 .
Recent transmission electron microscope characterization of CE-5 Fe-rich olivine rims suggests that Fe 0 particles in CE-5 samples are distinct from Apollo and Luna soil samples and were produced by olivine (especially fayalite) decomposition due to impact-induced heating processes 38 . Irregular Fe 0 particles (with a length up to 300 nm) were also observed in the iron sulfide from CE-5 soil samples 39 . The simulated space-weathering experiment on the synthetic olivine samples also suggests that the Fe-richer olivine produce larger Fe 0 particles 39,40 . These sample-based studies support that the abundant mpFe in CE-5 soils could be derived from Fe-rich silicate, oxide and sulfide minerals in CE-5 basalts.
Beside the dense clustering of npFe, the mpFe in Apollo and Luna soils was also considered to be inherited from metallic phases of micrometeoroids or parent rocks 17 42,43 . According to the fitting results, CE-5 soils inherited more mpFe particles from their source materials than Apollo and Luna soils did. The coexistence of coarse-grained mpFe particles and iron sulfides observed in CE-5 samples suggests a potential origin from Fe-FeS eutectic melt 44 . Alternatively, the abundant mpFe particles in CE-5 soils imply a relatively more reduced redox condition for the parental magma of the CE-5 young mare basalts, which might be produced by degassing occurring during its ascent/eruption stages and accompanied by net loss of oxygen. Orbital observations over the past decades have provided critical datasets about the space-weathering products and maturity of lunar regolith. In this work, these old datasets are integrated with information gained from ongoing studies of newly returned CE-5 samples and the unprecedented in situ datasets. Although these datasets were collected with various (spatial and spectral) resolutions and accuracies, this effort successfully coupled information obtained by remote sensing with information obtained from soil and regolith samples and helped to improve our knowledges of the space-weathering processes on the lunar surface. The consistent results among orbital, in situ and sample datasets show that regolith at the CE-5 landing site is mature with an average I s /FeO value of ~66 ± 3.2 ( Supplementary Fig. 10). These cross-validated results also indicate that CE-5 samples mainly come from the Xu Guangqi crater, whose age is constrained. The spectroscopic method can be further developed for the age estimation of young craters with well-constrained maturity data in the future 45 . A unique space-weathering mechanism, that extremely iron-rich basalts can produce more and larger-sized Fe 0 particles, is proposed. The physical status and spectral properties of the lunar regolith would be altered by the landing and sample-return processes. The in situ characterizations by current and upcoming lunar landing and/or roving missions, which do not disturb the primary state of lunar soils, are expected to bring important insights and shed light on the formation and evolution of lunar regolith.  Article https://doi.org/10.1038/s41550-022-01838-1 between the visible and NIR channels, there is a step in the spectra in the range of 900-950 nm. The visible image cubes are averaged and then connected with the low-SNR part of the NIR spectra, shortwave infrared spectra and mid-wave infrared spectra using the high SNR part (SNR = 61.47 at 940-1,450 nm) of the NIR spectra as a reference.

SDU-ASD.
VNIR reflectance spectra at the wavelength of 300-2,500 nm for lunar soils are acquired by a FieldSpec 4 Hi-Res VNIR spectrometer. The 75 W H. B. Mug Light is used as a light source with a standard incidence angle of ~30°. The reflectance spectra with spectral resolutions of 3 nm (350-1,000 nm) and 8 nm (1,000-2,500 nm) are acquired at an emission angle of ~0°. The distance between the optical probe and lunar soils is 8 cm and the spot size on the samples is 0.6 mm. The SDU-ASD was calibrated by an approximately reflecting Lambertian surface (SRT-99-100, Spectralon) before measurements of lunar soils. The experimental conditions and parameters of the calibration process are consistent with the measurements of lunar soils. The samples were not moved or spun during the calibration and measurement. To get average spectra, 9 different locations of 1 sample were measured and 100 scans for each location were performed. CH1-M 3 . CH1-M 3 is an imaging spectrometer that operates from 420 to 3,000 nm with a spatial resolution of 70-280 m per pixel (ref. 16 ). The M 3 reflectance image (frame M3G20090516T040653) was used in this study. The M 3 level 2 product contains small deviations that are corrected with scalar 'ground truth' correction factors to align them with known smooth spectral properties of lunar soils. We used the correction factors distributed by Isaacson 46 to minimize systematic errors. In addition, the spectra exceeding 2,500 nm were eliminated, considering the thermal emission effect.

Hapke radiative transfer model
The bidirectional reflectance (r) of lunar soils can be expressed as equation (1) assuming that lunar soil is an intimate mixture and that its particle size is much larger than the wavelength 9 : where μ 0 and μ are cosines of incidence angle (i) and emission angle (e), respectively, and g is phase angle. H (x/K) is the Ambartsumian-Chandrasekhar H function. The reflectance factor (REFF) is calculated from r with incidence angle (i): .
Henyey-Greenstein function The shadow-hiding opposition effect B(g) is also considered: where g is phase angle and h s is the angular width parameter: . (6) K is the porosity coefficient: Here filling factor ϕ was set to 0.41 for lunar samples analysed in the laboratory 49 and 0.17 for the upper lunar regolith measured in situ on the lunar surface 10 .
H (x/K) is expressed as (x represents μ 0 and μ, γ = (1 -ω) 1/2 ): The single-scattering albedo (ω) of a mineral mixture can be calculated as the weighted average of single-mineral single-scattering albedo using equation (9) 9 , where D i and V i are particle sizes and volume fractions of minerals in the mixture: Single-scattering albedo of a single mineral can be expressed as 18 : where S e and S i are the integral of external and integral Fresnel reflection coefficients (equations (11) 18 and (12) 47 ) and n is the real 47 part of refraction index: Θ is the internal transmission coefficient of the thin slab model: where 〈D〉 is the average distance travelled by transmitted photons during one traverse of the particles and D is the grain size of lunar soils. α is the absorption coefficient of the mineral: where α h is the absorption coefficient of the host materials and can be calculated by the refraction indices of the minerals. α c and α g represent the additional absorption from SMFe within coatings and Article https://doi.org/10.1038/s41550-022-01838-1 grains, respectively. The absorption coefficients α h , α g and α c are expressed as: where k is the imaginary 50 part of the refraction index, q a is the absorption efficiency of a single iron particle and can be calculated using Mie theory 19 , M c and M g are the abundance of iron in the coating and grain relative to the host mineral and d Fe is the iron particle diameter in units of centimetres and is set to a constant value of 200 nm (ref. 19 ). ρ h and ρ Fe are the density of the host mineral and iron particle. Z is from Maxwell-Garnett theory 18 , where n Fe k Fe are the real and imaginary parts of refraction index of iron metal and n h is the real part of refraction index of host mineral:

Validation and application of the spectral library
Modelling the absolute reflectance of lunar surface soils is difficult due to various reasons (for example, the existence of opaque minerals and distinct responses among different spectrometers. Thus, in this study, the modelled and measured spectra are normalized at 0.75 μm and then the root mean square error (r.m.s.e) is calculated using spectral data ranging from 0.75-1.55 μm. The Fe 0 abundances were derived via minimizing the r.m.s.e. (that is, the best match) between the measured spectra and the modelled spectra library. The spectra of ten lunar soils from the LSCC (10-20 μm fractions 1,25 ) are used to test the reliability of this model in estimating the content of Fe 0 . These spectral data can be downloaded at https://sites. brown.edu/relab/lscc/. The predicted and measured Fe 0 abundance are compared in Supplementary Fig. 4 and Supplementary Table 1. The modelled and measured reflectance spectra are shown in Supplementary Fig. 5. The raw reflectance spectra without continuum-removal operations are used to characterize npFe content. The npFe content of LSCC soils is predicted well using this model with a correlation coefficient (R) of 0.85 and an uncertainty (± r.m.s.e, defined as the average value of the differences between measured and modelled values) of 0.05 wt%, which is adopted as the prediction accuracy of our model. The prediction of SMFe is relatively poor (R = 0.61, uncertainty of 0.28 wt%; Supplementary Fig. 4). The mineral modes of CE-5 soils precisely measured in the laboratory were used for spectral modelling. Thus, the prediction accuracy is improved compared with Trang and Lucey's model. mpFe is systemically underestimated by this model compared with the measured mpFe abundance. This is probably because the averaged particle size of mpFe is simplified as 200 nm, which failed to accurately reproduce the darkening effect on reflectance spectra from mpFe with varying particle size.
The model and results in this study are also compared with Trang and Lucey's work 20 , which derived the global space-weathering maps (30 pixels per degree) of the Moon. The Hapke radiative transfer model and Mie theory are both used in our study, which is similar to that used in the works of Lucey et al. 19,20 . The details of our model differ from the methodology of Lucey et al. in the following ways: model building and spectral matching. First, the work of Trang and Lucey derived space-weathering maps for lunar global surface with varying mineralogy; however, we focus on a local area with homogeneous mineralogy, that is, the CE-5 landing site. Unlike the global surface of the Moon, the CE-5 mission returned soil samples whose detailed mineral and chemical compositions can be accurately determined by the state-of-art instruments in terrestrial laboratories 15 . Thus, the mineral modes of lunar samples measured in the laboratories were directly used in our model. Hence, our model achieved better prediction capacity of metallic iron abundance compared with the Trang and Lucey model. After verification using LSCC data, their model can predict npFe and mpFe contents with uncertainties within 0.1 wt% and 0.6 wt%, while prediction accuracies of our model are improved to 0.05 wt% and 0.26 wt%, respectively. Second, the npFe particles can both redden and darken the reflectance of the lunar surface. The continuum slopes of spectra can constrain the reddening effect from npFe. Thus, we didn't perform continuum removal on the spectra. This effectively improves the accuracy and uniqueness of our results. Trang and Lucey's results show that npFe content is 0.50 wt%, mpFe content is 1.1 wt% and SMFe content is 1.6 wt% at the CE-5 landing site (43.06° N, 51.92° W; Supplementary Fig. 9), which are consistent with our results derived from the same MI datasets (npFe: 0.51 ± 0.03 wt%; mpFe: 0.97 ± 0.31 wt%; SMFe: 1.5 ± 0.34 wt%).

The uniqueness of the metallic iron abundance derived from the model
The npFe and mpFe particles exhibit different spectral behaviours. The npFe particles can both redden and darken the reflectance of the lunar surface, while mpFe particles only darken the albedo. Given a specific spectral slope, the npFe content can be uniquely determined. When the darkening and weakening effects on reflectance from npFe are constrained, the subsequently derived mpFe content will be also unique.

Maturity index (I s /FeO)
The maturity index (I s /FeO) can be estimated using the fitting formula from Morris 17 : where Fe 0 A represents npFe content (grain size between 4 nm and 33 nm) and FeO values of LSCC soils are measured from the 10-20 μm fractions 23,24 . The average FeO content of CE-5 soils (22.5 wt%) 15 is used to calculate the I s /FeO of the CE-5 landing site.
The I s /FeO can also be derived by the statistically optimized formulation from Pieters et al. 26 as follows: log (I s /FeO) = −0.200 × R 750 nm + 0.161 × R 950 nm + 0.018 × R 1,000 nm + 1.960 (21) where R 750 nm , R 950 nm and R 1,000 nm are the REFF given in % at 750, 950 and 1000 nm, respectively. The correlation coefficient of the predicted and measured data is 0.92 and the standard deviation is 18.

Data availability
Large data necessary to generate the results used for this study are available online (https://doi.org/10.6084/m9.figshare.19807858). All original CE-5 data can be found in the Lunar and Planetary Data Release System (http://moon.bao.ac.cn). Source data are provided with this paper.