Introduction

As a natural satellite of the Earth, the Moon is rich in geological information, especially on the farside, which has been less explored and studied due to the tidal locking effect of the Earth. So far, China has successfully conducted four lunar surface landing missions. The Chang’E-3 (CE-3) and Chang’E-5 (CE-5) missions landed on the near side of the Moon, while the Chang’E-4 (CE-4) and Chang’E-6 (CE-6) missions landed on the farside1,2,3,4. The landers/rovers of the four missions were equipped with subsurface penetrating radars, providing invaluable scientific data for comparing the shallow subsurface structures and surface processes at the lunar nearside and farside. Moreover, the CE-6 mission represents a critical step in China’s lunar exploration program’s “orbit, land, and return” strategy1,2,3,5,6, since it is the first mission to return samples from the farside of the Moon to Earth. The CE-6 mission successfully landed on a flat mare unit (153.978°W, 41.625°S; Fig. 1a) in the southwestern Apollo basin, located within the South Pole–Aitken (SPA) basin. The SPA basin is among the oldest, largest, and deepest known impact structures in the solar system, predating most lunar surface features and providing a crucial record of the early formation and evolution of the Moon’s crust. The Apollo basin, located in the north-central SPA region, exhibits complex internal structures shaped by multiple impact events and possible volcanic activity7. The landing site is located right on the rim of a crater called C1, which has a diameter of ~ 51 m, and is surrounded by several smaller impact craters ranging from a few meters to several dozen meters in diameter (see Fig. 1b). High-resolution descent camera data indicates that the landing site is relatively flat, with a sparse distribution of rocks and no large-scale boulders observed (see Fig. 1b, c), which allowed a safe landing of the lander. Based on remote sensing topographic observations and mineralogical spectroscopic data, the landing area is characterized by signs of multiple phases of basaltic volcanic activity6,8,9. Moreover, recent impact crater dating results suggest that the landing region has a geological age of 2.79 Ga4.

Fig. 1: Geological context of the CE-6 landing site (The red cross in all the images indicates the position of the CE-6 lander).
figure 1

a Geological context of the CE-6 landing region within the Apollo basin. The base map is CE-2 DOM data (Image ID: I2-01~I2-07, J1-01~J1-06, J1-40, J2-01~J2-06, J2-40, K1-01~K1-06, K1-36). b High-resolution DEM of the CE-6 landing region, observed by the CE-6 landing camera (Image ID: CE6_LCAM_DEM_0.01m_LANDINGSITE_A). The white dashed lines show five local impact craters (C1-C5). The lander, which is marked by the red cross, is located on the rim of crater C1. b High-resolution image of the CE-6 landing region, observed by the landing camera. The red cross marks the location of the CE-6 lander. The line AB is connecting the center of crater C1 and the CE-6 lander. c Elevation profile along the AB line. The red cross denotes the centroid position of the lander, while the red dot marks the floor of crater C1. d Illustration of the drilling device and LRPR onboard CE-6 lander. The star marks the sampling spot. The LRPR antennas are painted in red and positioned on both sides of the drill. e High-resolution image of the CE-6 landing region, observed by the landing camera. The red cross marks the center of mass of the CE-6 lander, the bright blue line represents the coverage area of the 12 LRPR antennas (1.56 m long), and the red star indicates the drill-sampling spot.

Full size image

The main methods for investigating the internal structure of lunar regolith are core sampling and in-situ subsurface radar probing. The CE-6 lander touched down in Apollo Basin on June 2, 2024 (UTC + 8), successfully collected 1,935.3 grams of both surface and drilled samples by June 4, and after ascended from the lunar surface on June 21. The lunar samples returned by previous Apollo and Luna missions provide crucial geological context for interpreting our GPR results. These missions obtained multiple drill cores from nearside mare regions, with core diameters up to 4 cm10,11. Notably, Apollo 17 achieved the deepest sampling at 2.92 m depth10,11, though still not penetrating the complete regolith column. These sampling results provide comparative regolith baselines for our GPR data while demonstrating the deep-sampling challenges that justify radar investigations.

The CE-6 lander was equipped with a Lunar Regolith Penetrating Radar (LRPR)12, which was an antenna array radar to study the shallow regolith structure at the landing and sampling site. The antenna configuration is the same as the one onboard CE-5 lander, featuring 12 antennas mounted around the drilling device (see Fig. 1d, e). The LRPR operates at a central frequency of 2 GHz and has a working band 1-3 GHz12, providing a vertical resolution of ~5 cm and penetrating depth of ~3 m12,13. It offers a notable performance enhancement in terms of resolution when compared to its predecessors adopted for CE-3 and CE-4 missions. Indeed, the CE-3 and CE-4 penetrating radars successfully imaged subsurface structures ranging from a few meters to several tens of meters14,15 with ~30 cm vertical resolution14,15. Compared to these systems, the LRPR’s 5 cm vertical resolution allows the CE-5 and CE-6 mission to observe unprecedented details of the shallower part of lunar subsurface. Therefore, CE-6 LRPR fills a critical gap by enabling investigation of fine-scale internal structure and dielectric properties within the upper 3 m of farside lunar regolith. These radar observations also provide geological context for the drilled samples, revealing the stratigraphic framework and material properties of their source environment.

The present study advances the characterization of shallow farside lunar subsurface structures through novel processing of CE-6 radar data and dielectric property estimation algorithms, combined with systematic comparisons with datasets from the Apollo, Luna, and previous Chang’E missions. By synergistically integrating the metal plate reflection method with the frequency decay-based method and the energy attenuation-based estimation method in CE-6 radar observations, we derive dielectric parameters of the regolith: relative permittivity (\({\epsilon }_{r}\)) = 2.22 ± 0.72 and loss tangent (tanδ) = 0.010 ± 0.002. These findings establish the first in-situ reference framework for regolith properties in the South Pole–Aitken Basin. They further provide critical support for the calibration of orbital remote sensing data, the interpretation of the subsurface structure at the CE-6 landing and sampling site, and the strategic selection of future sample return locations. The derived dielectric parameters dataset also serves as a benchmark for upcoming lunar missions targeting similar geological units.

Results

Physical parameters estimation

We firstly estimated the real part and loss tangent of the dielectric permittivity of the lunar regolith at the CE-6 landing site. To estimate the real part of the dielectric permittivity, we used the reflection coefficient at the air-regolith interface as a proxy12,16. The reflection coefficient is calculated by exploiting the reference reflection signal provided by the LRPR ground calibration test and the reflection signal from the lunar surface measured during the CE-6 mission. This method has been already applied to the LRPR data from the CE-5 mission and has been validated against laboratory measurements of the CE-5 samples12,16.

As shown in Fig. 2a, the estimated value of the relative permittivity at the CE-6 landing site is 2.22 ± 0.72 (see “Methods” for details). By exploiting the empirical formula relating the dielectric constant and the bulk density for lunar regolith based on Apollo samples11,17, we estimate that the surface density of the regolith at the landing site is equal to 1.22 ± 0.50 g/cm3 (see “Methods”).

Fig. 2: Dielectric constant estimation, loss tangent estimation, and subsurface images at the CE-6 landing site based on LRPR data.
figure 2

a Histogram of the permittivity estimations using CE-6 and CE-5 LRPR datasets. The red bars show the results (\({\epsilon }_{r}=2.22\pm 0.72\)) from the CE-6 LRPR, and the blue ones represent the CE-5 LRPR results (\({\epsilon }_{r}=2.48\pm 0.79\)). b Loss tangent estimation using the central frequency decay method. The black dashed line indicates the fast time-dependent decay of the central frequency in the single-trace data from the CE-6 LRPR. The red solid line represents the best fit across all data (\(\tan \delta =0.010\pm 0.002\)), and the red dashed lines shows the upper bound and lower bounds of the estimation. The blue curve represents the averaged central frequency decay from all observations. c Pictorial image of subsurface structure at CE-6 landing and sampling site. The illustration emphasizes the rock size distribution based on the CE-6 LRPR image shown in (d). The rocks are represented by the shapes filled with solid color. The gray background represents fine-grained regolith and the brown background represents coarse regolith. d LRPR image from the CE-6 mission (Data ID: CE6-L_GRAS_LRPR-A_SCI_N_20240602012333_20240602014101_0001.01). The red regions indicate areas of high reflectivity, while the blue regions indicate areas with low reflectivity. The horizontal axis is centered at the drill position, and the vertical axis represents the depth. (e) LRPR image from the CE-5 mission for comparison (Data ID: CE5-L_GRAS_LRPR-A_SCI_N_20201201171559_20201201173326_0001.01).

Full size image

To estimate the loss tangent, we use two methods: a central frequency decay-based method and an energy attenuation-based estimation method18,19,20,21. Both methods yield consistent results (see “Methods”). The result indicates that the loss tangent at the CE-6 landing site is 0.010 ± 0.002, as illustrated in Fig. 2b. Studies of Apollo lunar samples have shown that the loss tangent is closely related to the FeO+TiO2 content of the regolith11,17. Based on the estimated loss tangent and real part of the dielectric permittivity, we evaluate the mass fraction of FeO+TiO2 at the landing site to be 23.5 ± 4.9 wt.%.

Laboratory measurements4 of two CE-6 soil samples (CE6C0000YJFM00102 and CE6C0000YJFM00103) and one subophitic basalt sample (CE6C0000YJYX41301) yielded FeO + TiO2 contents of 20.0 wt.%, 19.8 wt.%, and 27.1 wt.%, respectively. Therefore, the mass fraction of FeO+TiO2 estimated from the LRPR data (23.5 ± 4.9 wt.%) fall within the range of values associated with soil and subophitic basalt samples. Given that the radar-observed zone includes both the fine-grained regolith and rocks, this result is consistent with the laboratory measurements of the CE-6 samples.

Radar observations of the shallow subsurface at Chang’E-6 site

During the sampling mission of CE-6, the LRPR made its four in-situ observations at the landing and sampling site. The first of the four in-situ datasets was collected before the drilling process and the others were collected afterward the drilling. The LRPR operates under a Multi-Input-Multi-Output configuration; the 12 antennas will transmit electromagnetic waves one at a time. For each of 12 transmissions, the signals scattered by subsurface targets, having electromagnetic properties from the ones of the background regolith, are gathered by the other 11 antennas that operate as receivers. Therefore, each dataset is made of 12 × 11 = 132 measurements. The processing of LRPR data12,22 allows us to get a 2D image representing information about the subsurface structure and features of the lunar regolith. Herein, the synthesis imaging method12,23 is used to process the LRPR data and to obtain high-resolution radar images of the lunar regolith within the depth of 3 m. Moreover, to compensate for the attenuation caused by the spherical spreading of electromagnetic waves and the electromagnetic losses of the lunar regolith, we apply a spherical exponential compensation (SEC) gain to the radar signal (see “Methods”). For the processing, as background values we exploit the soil dielectric parameters estimated previously (relative permittivity equal to 2.22 ± 0.72 and a loss tangent of 0.010 ± 0.002). Specifically, we adopt a 2.22 relative permittivity and a 0.010 loss tangent.

Figure 2d, e show the radar imaging results for the in-situ LRPR data at the CE-6 and CE-5 landing sites, respectively. These two images were obtained by using the same processing workflow and display the echo intensity from reflectors encountered by the radar waves. Here, the blue areas are weak reflections, and red areas represent strong reflections. The 0–20 cm region of Fig. 2d corresponds to the echo from the air-regolith interface, while the 20–170 cm region represents the shallow subsurface. This section is mostly blue, indicating weak scattering, which suggests that the material distribution within the top 1.7 meters is relatively uniform and consists mainly of fine-grained regolith containing a few small rocks. Between the depths of 1.7 and 3 m, several strong 30 cm-size scattering areas are observed, indicating the presence of heterogeneities with higher dielectric constants. Based on the interpretation of the CE-6 radar image, we additionally sketch the rock distribution of the subsurface, shown by Fig. 2c.

Figure 2e presents the radar image at the CE-5 landing site for comparison. Both CE-6 and CE-5 radar images show a trend of increasing size and frequency of high-scattering regions with depth. Additionally, the comparison of echo intensities in the 0–20 cm region between Fig. 2d, e is noteworthy. The echoes in this region are primarily from the air-regolith interface, and the reflection intensity is influenced by the surface permittivity. The greater the permittivity is, the stronger the echo intensity is. The weaker surface echo at the CE-6 landing site compared to the one at the CE-5 site suggests that the surface permittivity at the CE-6 site is lower, which is consistent with the values derived from the permittivity analysis, as shown in Fig. 2a.

Discussion

The connection between lunar regolith dielectric characteristics and the weathering process

By comparing the CE‑6 dielectric properties with those from other lunar missions, we can potentially correlate the regolith’s dielectric characteristics to its geological context. Figure 3a shows the distribution of Chang’E and Apollo mission landing sites on the Moon. Besides, Fig. 3b, c provide a comparative overview of the loss tangent values derived from the radar data for the landing sites of CE-3, CE-4, CE-5, and CE-6. We additionally included the data from Apollo 17’s in-situ electromagnetic experiments24 and laboratory measurements of Apollo lunar samples17,25 for reference. Comparative analysis shows that the loss tangent at the CE-6 landing site is smaller compared to the one of the lunar nearside sites observed by CE-3 and CE-5; also, the CE-6 loss tangent is larger than the value observed at the CE-4 landing site on the farside. It is worth noting that the study by Boivin et al.26 tested the complex dielectric constant of lunar regolith simulants at different frequencies and showed that the loss tangent exhibited a slight dependence on frequency. Accordingly, the frequency behavior of the lunar regolith loss tangent values was not considered in our analysis.

Fig. 3: Distribution of Apollo and Chang’E Missions Landing Sites and Physical Property Inversion of Lunar Regolith Parameters.
figure 3

a Distribution of Apollo and Chang’E mission landing sites on the Moon. The base map is a global lunar DEM observed by CE-1 (Image ID: CE1_GRAS_DEM_500m_001_GLOBAL_A). The landing sites of CE-3 to CE-6 missions are marked with stars, while the landing sites of Apollo missions are labelled with triangles. b The relationship between the loss tangent derived from radar inversions and Apollo 12 to Apollo 17 sample tests versus the geological age of the lunar surface. This plot shows the normalized loss tangent, FeO+TiO2 mass fraction, and the corresponding geological ages. The loss tangent has been normalized using Eq. (1) to remove the effect of different bulk densities. The stars represent regolith parameters obtained from in-situ radar observations during the Chang’E missions, the cross is the Apollo 17 mission in-situ radar measurements, and the triangles denote the measurements from Apollo regolith samples. The colors of the symbols represent different geological ages, with red, green, and blue and indicate young, medium, and old geological ages, respectively. c Fitted results of loss tangent values versus the age of the landing site based on radar observations from CE-3, CE-4, CE-5, and CE-6. In this plot, the radar-derived results from the Chang’E missions are shown in stars, while the red line is the fitted relationship. Additionally, Apollo regolith sample measurements and in-situ measurement presented in panel (b) are included, represented by triangles and a cross, respectively.

Full size image

The radar data from CE-6, combined with data from previous Chang’E missions, provide a unique opportunity to investigate the different degrees of weathering of the regolith. The landing sites of the Chang’E missions have different geological ages and characteristics, offering a range of study subjects. The loss tangent of the upper 3 meters of regolith at the CE-6 landing site is constrained to ~0.010 ± 0.002, which is comparable to the value observed for the shallow regolith at the CE-3 landing site (~0.0124 ± 0.002)27. The CE-3 landing site is situated on the ejecta blanket of the Ziwei crater, a 450-meter diameter impact crater. The impact likely excavated and disrupted the underlying basalt, which dates back to the Eratosthenian period, around 2.5 Ga14. The geological age of the CE-6 landing site is estimated to be 2.79 Ga4, suggesting that the geological ages of the CE-3 and CE-6 landing sites are relatively similar. The shallow regolith at the CE-6 site could have been produced and developed on top of ejecta material from a local impact crater. This situation is similar to the regolith at the CE-3 site, where the upper fine regolith is constrained to be ~1 m and its underlaying material is interpreted as the ejecta from the nearby Ziwei crater14. This similarity in regolith structure and physical properties suggests that both sites were shaped by comparable processes of ejecta blanket deposition, prolonged space weathering, and localized impact-induced regolith reworking.

However, although both CE-4 and CE-6 mission were landed on the farside, the loss tangent values of the lunar regolith at the CE-6 landing site (0.010 ± 0.002) are significantly higher than those at the CE-4 landing site (approximately 0.005 ± 0.00215). Besides, the CE-6 loss tangent is still lower than those at the CE-3 landing site (0.0124 ± 0.00227) and the CE-5 landing site (approximately 0.014 ± 0.00212). Based on statistical analyses of Apollo lunar soil samples, the loss tangent of lunar regolith is closely related to bulk density and the mass fraction of FeO + TiO211,17. To ensure that the explanation is clear and accessible, Table 1 compares the key factors associated with the loss tangent. Spectral measurements indicate that the mass fraction of FeO + TiO2 in the lunar surface at the CE-3 landing site is 27.8 wt.%28. At the CE-4 landing site, the mass fraction of FeO and TiO2 are 7.42–18.82 wt.% and 1.48–2.1 wt.%, respectively29. Experiments on sampled materials shows that the mass fraction of FeO + TiO2 in the lunar soil at the CE-5 and CE-6 landing sites are 27.5 wt.% and 19.9 wt.%, respectively4,30. The relationship between the loss tangent values at the four Chang’E landing sites and their respective FeO + TiO2 content is consistent overall. However, despite of the similarity of FeO + TiO2 content and bulk density at the CE-3 and CE-5 landing sites, their loss tangent values differ by 19.4%. This might be indicative that there are additional factors contributing to variations in the loss tangent of lunar regolith.

Table 1 Measured properties of lunar regolith at the Chang’E landing site, focusing on loss tangent and related characteristics
Full size table

As the bulk densities at the four sites are similar, this is not a determining factor for the loss tangent. We have chosen to exclude the influence of the different bulk densities and therefore focus on the FeO + TiO2 content and other potential factors. We achieved it by normalizing the loss tangent using Apollo samples derived empirical formula11,17 as follows

$$\begin{array}{c}{\tan \delta }_{{norm}}=\tan \delta \cdot {10}^{0.312\rho }={10}^{0.038\left({FeO}+{TiO}2\right) \% -3.26}.\end{array}$$
(1)

Based on this, Fig. 3b presents the normalized loss tangent values of lunar regolith obtained from both the Chang’E and Apollo missions, along with the corresponding mass fractions of FeO + TiO2 and geological ages4,11,12,15,17,26,31,32,33,34,35. The figure classifies the data into red areas (young), green areas (moderately aged), and blue areas (old) based on relative geological age. It can be noted that as long as the mass fraction of FeO + TiO2 increases, the loss tangent of the lunar soil also tends to increase. However, a comparison of the red (young) and blue (old) regions shows that, despite of their similar FeO + TiO2 content, the loss tangent in the older region is significantly lower than that in the younger region. Additionally, when examining the green (moderately aged) and blue (old) regions, although the older region exhibits a higher FeO + TiO2 content, its loss tangent is close to the relatively younger region with a lower FeO + TiO2 mass fraction. These data suggest that regions of older geological age typically show a decrease in the loss tangent. One possible reason for this is the particle size evolution of the lunar regolith. As lunar soil evolves, its particle size tends to decrease. For instance, lunar soil samples returned by the CE-6 mission feature smaller particle sizes compared to those from the CE-5 mission4,30, indicating a higher degree of interior evolution in the lunar soil. Previous studies have also shown that an decrease in the porosity or particle size of a medium can lead to a decrease in its loss tangent36,37. This relation suggests that, beyond the FeO + TiO2 mass fraction, the electromagnetic losses of lunar soil are also closely related to its geological age and/or weathering processes. The dielectric properties of the shallow lunar regolith (0–5 m depth) show a clear dependence on surface exposure age, as demonstrated by a progressive decrease in the loss tangent (tan δ) with increasing regolith maturity. This relationship reflects the cumulative effects of space weathering processes: prolonged micrometeorite bombardment and thermal cycling preferentially alter the upper regolith layers, changing their physical and electromagnetic characteristics over time. As the surface ages, the regolith tends to develop finer grain size distributions and lower bulk porosity compared to younger terrains. These internal structural characteristics collectively contribute to reduced dielectric losses. It is noteworthy that the correlation between surface age and loss tangent persists despite vertical stratigraphic complexity38, because impact gardening primarily disturbs the upper 1–2 meters39, creating a weathering-dominated zone where electromagnetic properties reliably record cumulative exposure duration. We also fitted a linear relationship between the loss tangent values obtained from the radar measurements of CE-3, CE-4, CE-5, and CE-6 and their respective geological ages, yielding the following equation:

$$\begin{array}{c}\tan \delta =-0.0055\cdot G+0.0255;{R}^{2}=0.98\end{array}$$
(2)

where G means the geological ages in Ga.

This trend indicates a strong correlation between the electromagnetic loss of lunar soil and its geological age and/or weathering processes. Combining the data, we suggest that older geological regions and higher degree of internal evolution in the lunar soil might result in a decrease in electromagnetic wave loss.

The formation of the regolith at Chang’E-6 landing area

Chang’E-6 landed within the Apollo crater in the SPA basin, marking the first mission to return samples from the lunar farside. This site serves as a key record of the early impact history of the solar system. Based on in-situ radar measurements of the shallow regolith at the landing site, we identified a layered subsurface structure comprising an upper fine-grained, strongly weathered layer and a coarse-grained layer. This finding not only improves our understanding of the formation and evolution of the Apollo basin’s subsurface, but also provides new observational evidence of multi-phase impacts and material redistribution within the SPA basin.

As said the basaltic bedrock in the CE-6 landing area has an age of 2.79 Ga4. Notably, 76% of the region is characterized by a total slope of less than 8°, indicating that the landing site has a relatively flat surface6. According to observations from the CE-6 LRPR, till the depth of 3 m, the subsurface exhibits a two-layer structure. The first layer, with a thickness of ~1.7 m, can be interpreted as mature fine-grained lunar regolith. The second layer, located beyond the depth of 1.7 m, can be interpreted as coarse-grained lunar soil or unweathered impact crater ejecta, as illustrated in Fig. 4a. The CE-6 radar instrument may be revealing a subsurface structure that is more complex than previously suggested by present models. This highlights the need for substantial follow-up investigations to refine and advance current models based on these new observations.

Fig. 4: Geological Interpretation of Radar Results from the CE-6 and CE-5 Landing Sites.
figure 4

a DEM of the area provided by CE-6 landing camera. b Illustration of the excavation and ejecta processes at the CE-5 landing site and the Xu Guangqi impact crater. c Illustration of the excavation and ejecta processes at the CE-6 landing site and the C1 impact crater.

Full size image

From the geological profiles of the CE-6 and CE-5 radar data (Fig. 4b, c), the abundance of rocks within the shallow lunar soil at the CE-5 landing site is higher than the one at the CE-6 site. This circumstance may be due to the shallow subsurface material detected by the radar at the CE-5 landing area, which is mainly derived from the ejecta of the ~463 metre diameter Xu Guangqi impact crater. Based on impact crater excavation models40, the excavation depth of the Xu Guangqi crater is ~35.6 m. Prior to the formation of the Xu Guangqi crater, the local lunar regolith thickness at the CE-5 landing site was estimated to be about 7 m41. Therefore, the crater has penetrated the lunar regolith layer and excavated the underlying basalt, resulting in a higher rock content in the ejecta, as shown in Fig. 4b.

The CE-6 landing site are surrounded by several small impact craters, namely C1, C2, C3, C4, and C5 (see Fig. 1b). The uppermost layer of fine-grained lunar regolith may have developed from the ejecta originated from these craters. Based on impact ejecta models42, the total thickness of the ejecta from craters C1, C2, C3, C4, and C5 reaching the landing area is constrained to be about 0.47 m. Here we assume that the crater C1 is the primary crater in the study area. The ejecta from crater C1, with a diameter of about 50.9 m, contributes the most material to the landing site. The thickness of the ejecta is about 0.46 m. Thus, crater C1 had the most significant influence on the internal evolution of the recent lunar regolith at the CE-6 site. Here, we assume that the final diameter of crater C1 is ~1.3 times to its transient diameter, and that the excavation depth is about one-tenth of its transient diameter40. As a result, the excavation depth of crater C1 is estimated as ~3.9 m. This depth is unlikely to reach the bedrock beneath the CE-6 landing site, but only penetrates the background lunar regolith.

Regarding the formation of crater C1, the CE-6 landing site has not received considerable additional material from external sources. Instead, local lunar soil has been cultivated and subjected to internal grinding and space weathering processes. Since the crater C1 did not excavate bedrock, its ejecta mainly originated from the gardening of in-situ regolith. This gardening process significantly reduces the grain size within the regolith, resulting in fine-grained regolith. After the C1 ejecta blanketed the landing area, multiple smaller impact craters (e.g., C2, C3) were observed on its ejecta deposit. These smaller craters also did not reach the bedrock, so their primary effect was to further garden the surface regolith, increasing its maturity. This continuous gardening process further refines the regolith particles. Numerical simulations by Costello et al.39 suggest that micrometeorite bombardment contributes significantly to the mixing depth of shallow regolith. For example, on a billion-year timescale, micrometeorite impacts can overturn regolith to a depth of approximately 1 m39. Additionally, solar radiation penetrates the lunar surface to a depth of about 1 m43, and the resulting diurnal thermal cycling causes further fragmentation of regolith particles within this depth range, thereby promoting regolith maturity. Thus, after the C1 ejecta covered the landing area, the combined effects of continuous gardening by smaller craters and thermal cycling by solar radiation led to the development of regolith in the upper 1.7 m. As a result, the shallow lunar soil of the first 1.7 m has gradually developed into fine-grained lunar regolith, showing a high degree of uniformity in its internal medium. Conversely, radar echoes from below 1.7 m exhibit more variable features, indicating a considerable inhomogeneity within the medium of this region. This can be interpreted as unweathered ejecta from unknown sources of impact craters, which contain a relatively high abundance of rock fragments. The origin and thickness of this unweathered ejecta requires further investigation.

Conclusions

Based on high-resolution in-situ radar data from Chang’E-6, this study presents the first detailed investigation of the internal structure and dielectric properties of lunar regolith for the top ~3 meters of lunar regolith at the Moon’s farside. The revealed subsurface stratigraphy provides new constraints on the subsurface evolution of the Apollo basin. It can offer valuable insights into the post-impact reworking history of the broader SPA basin and enable the formulation of a working hypothesis regarding the potential geological structure and formation processes of the regolith in the landing area. Furthermore, by comparing data from the Chang’E and Apollo series landing sites, we have identified a possible correlation between the geological age of the lunar surface and the loss tangent of the lunar regolith. This correlation suggests that the radar signal attenuation characteristics of the shallow lunar regolith are closely related to its internal degree of evolution. In particular, a higher degree of regolith evolution or weathering may lead to lower attenuation for the electromagnetic waves propagating through the regolith. However, these results cannot be generalized to the landing area, due to the limited detection depth of the CE-6 LRPR and the constraints of its fixed detection spot. It is important to recognize that point-based radar measurements and derived information cannot be fully generalized to represent the entire lunar subsurface due to the structural and compositional heterogeneity of the Moon. Despite these limitations in detection coverage, the CE-6 detector has made an important contribution to the study of the internal structure of the lunar regolith on the lunar farside.

Methods

LRPR operation principle, data collections and processing

The CE-6 LRPR shares the same design of the CE-5 LRPR’s, which features a bandwidth of 1–3 GHz and an 1.56-meter long antenna array13. The array is composed of 12 Vivaldi antennas, which are located at the bottom of the lander, approximately 0.88 m above the lunar surface. The bandwidth and aperture provide a range resolution of ~5 cm and a horizontal resolution of ~8 cm when the depth is 1 meter, respectively12.

The radar works under a Multi-Input-Multi-Output configuration. To conduct a radar survey, the scene is illuminated by a transmitting antenna while the other 11 antennas record the subsurface echoes. The same process is repeated sequentially for each transmitter. The measurement phase lasts 18 min and produces 132 traces of data (12 × 11 measurements).

During the orbiting stage on May 30th, 2024, the CE-6 LRPR measured its system clutter by executing three radar observations in free space with the same in-situ observation parameters.

The CE-6 LRPR data is calibrated and processed according to the CE-5 LRPR methods, including the following steps12,22:

  1. 1.

    Sample time calibration. LRPR uses an equivalent sampling method, which produces non-uniform samples in time domain13. We resample the radar A-scans to eliminate the nonuniformity.

  2. 2.

    Bandpass filtering. According to the designed bandwidth of LRPR, a bandpass filter between 1000 to 3000 MHz is applied to remove the system noise from the radar data.

  3. 3.

    Cable delay calibration23. The Multi-Input-Multi-Output design of LRPR introduces a variable cable time delay for the traces associated with different transmitting and receiving antennas. We adopted a cable delay estimate and calibrated the hardware delay accordingly, based on the CE-6 LRPR in-situ observation23.

  4. 4.

    Background removal12,22. The CE-6 LRPR free-space observations represent the system background signal, including the direct coupling signals and multiple reflections from the lander. We subtract the systematic clutter from the in-situ datasets accordingly.

  5. 5.

    LRPR imaging12,23. We adopt the synthesis imaging method that has been validated by CE-5 LRPR in-situ datasets12,23. We have utilized the CE-6 LRPR derived permittivity and loss tangent (2.22 ± 0.72 and 0.010 ± 0.002, respectively) as electromamgnetic properties of the background soil in order to achieve the radar image.

The produced CE-6 LRPR radar image is shown in Fig. 2d.

Estimation of permittivity, loss tangent and derivation of other parameters

The estimation of the relative dielectric permittivity \({\epsilon }_{r}\) is accomplished by measuring the reflection coefficient of the interface between the lunar surface and air. The method used was successfully applied to CE-5 ground and in-situ observations and provides results compatible with laboratory measurements12,16. The method is based on two measurements of the reflection coefficient: the first one regards a reference target (i.e. a metal plate), while the measurement is referred to the unknown target.

The model provides the surface permittivity value as

$${\epsilon }_{r}={\left(\frac{{A}_{m}+{A}_{s}}{{A}_{m}-{A}_{s}}\right)}^{2}$$
(3)

where \({A}_{m}\) refers to the ground reflection amplitude of metal plate, and \({A}_{s}\) is the amplitude of lunar surface reflection. The achieved results are shown in Figs. 2a and 5a.

Fig. 5: Supplementary details of relative permittivity and loss tangent estimation.
figure 5

a Distribution of the estimated relative permittivity for each pair of transmitting and receiving antennas. The red stars represent results from the CE-6 LRPR, while the blue triangles represent results from the CE-5 LRPR. b Loss tangent estimation by energy attenuation-based estimation with CE-6 and CE-5 data. The blue and gray dots are energy-time observations of CE-6 and CE-5 LRPR, respectively. The red line and red dashed line represent the fitted results of CE-6 and CE-5 LRPR, respectively. The signals from the first 1 m are dominated by the surface echo, so they are excluded from the fitting window.

Full size image

Two separate approaches known as the energy loss method18,21 and the centroid frequency method19,20 are exploited for the estimation of loss tangent \(\tan \delta\). The energy loss model18, we used here, is based on the equation

$${P}_{r}\left(R\right)={P}_{t}\frac{{{G}^{2}\lambda }^{2}\xi }{{{4\pi }^{3}R}^{4}}\exp \left(-2\alpha \cdot R\right)$$
(4)

where R is the round-trip-distance of LRPR, and

$$\alpha =\omega \sqrt{\frac{{\epsilon }_{r}}{{2c}_{0}^{2}}\left(\sqrt{1+{\tan }^{2}\delta }-1\right)},$$
(5)

which refers to the attenuation factor related to loss tangent \(\tan \delta\). \({P}_{t},{P}_{r},G,\lambda ,\xi\) are the transmitted and received power of LRPR, the system gain, the centroid wavelength, and the backscattering factor. The backscattering factor ξ is dependent on the extent and the geometry of the scatterer. For a planar target, ξ equals to \(\pi {R}^{2}\Gamma\), where \(\Gamma\) is the power reflection coefficient. This results in \({{P}_{r}/{P}_{t}\propto R}^{-2}\). The other situations are sub-wavelength sphere-like targets (\(\xi ={\pi }^{5}{D}^{6}\Gamma /{\lambda }^{4}\)) and the intermediate case (\(\xi =\pi \lambda R\Gamma /2\)), which yield \({{P}_{r}/{P}_{t}\propto R}^{-4}\) and \({{P}_{r}/{P}_{t}\propto R}^{-3}\), respectively. The LRPR features extremely wide operation bandwidth ranging from 1 to 3 GHz. As a result, the scatterers could vary from sub-wavelength targets at 1 GHz and multi-wavelength targets at 3 GHz. Based on the complex situation and previous studies on LRPR12, we choose the intermediate case \(({{P}_{r}/{P}_{t}\propto R}^{-3})\) according to the model resulting from the geometric spreading and backscattering.

For the centroid frequency method, the spectrum alternation model is20

$$X\left(\nu ,\tau \right)=\left|{\rho X}_{0}\left(\nu \right)\right|\cdot \exp \left(-\pi \nu \cdot \tan \delta \cdot \tau \right)$$
(6)

where \({X}_{0}\left(\nu \right)\) is the transmitted radar signal spectrum, ρ is the Fresnel reflection coefficient, \(\nu\) is the frequency of the signal and \(\tau\) is the round-trip time for the radar signal to propagate in the regolith. Here, the spectrum of ground reflection is assumed as \({X}_{0}(\nu )\). The centroid frequency is then calculated as

$${\nu }_{c}\left(\tau \right)=\frac{{\int }_{0}^{{bw}}\nu X\left(\nu ,\tau \right)d\nu }{{\int }_{0}^{{bw}}X\left(\nu ,\tau \right)d\nu }.$$
(7)

The least square regression is used in both cases to estimate the loss tangent, and the results are provided in Figs. 2b and 5b. The loss tangent values of CE-6 data estimated with the energy loss method and the centroid frequency method are 0.010 ± 0.002 and 0.011 ± 0.002, respectively. By averaging the two results, we can estimate a loss tangent of 0.010 ± 0.002. In comparison, the final loss tangent estimation by CE-5 data is 0.013 ± 0.002.

Based on the estimated permittivity and loss tangent, several empirical formulas can be used to evaluate the bulk density and FeO + TiO2 content of the CE-6 landing site. Based on laboratory measurements and regression analysis on Apollo mission samples, an empirical model between the regolith permittivity and the regolith bulk density has been proposed as11,17

$${\epsilon }_{r}=1.{919}^{\rho }$$
(8)

Where ρ (\(g/{{cm}}^{3}\)) refers to the bulk density of regolith. On the other hand, the regolith loss tangent is found to depend on the FeO + TiO2 abundance and bulk density11,17. The empirical relationship is

$$\begin{array}{c}\tan \delta ={10}^{0.038\left({FeO}+{TiO}2\right) \% +0.312\rho -3.26}.\end{array}$$
(9)

By using Eq. (8) and considering a value of the relative dielectric permittivity equal to 2.2, we estimated the bulk density equal to 1.22 ± 0.50 g/cm3. Finally, with the estimated loss tangent (0.010) and bulk density (1.22 g/cm3), Eq. (9) permits us to derive the corresponding FeO + TiO2 content as 23.5 ± 4.9 wt%.