Mineralogical and chemical properties inversed from 21-lunar-day VNIS observations taken during the Chang’E-4 mission

We report on the mineralogical and chemical properties of materials investigated by the lunar rover Yutu-2, which landed on the Von Kármán crater in the pre-Nectarian South Pole–Aitken (SPA) basin. Yutu-2 carried several scientific payloads, including the Visible and Near-infrared Imaging Spectrometer (VNIS), which is used for mineral identification, offering insights into lunar evolution. We used 86 valid VNIS data for 21 lunar days, with mineral abundance obtained using the Hapke radiative transfer model and sparse unmixing algorithm and chemical compositions empirically estimated. The mineralogical properties of the materials at the Chang’E-4 (CE-4) site referred to as norite/gabbro, based on findings of mineral abundance, indicate that they may be SPA impact melt components excavated by a surrounding impact crater. We find that CE-4 materials are dominated by plagioclase and pyroxene and feature little olivine, with 50 of 86 observations showing higher LCP than HCP in pyroxene. In view of the effects of space weathering, olivine content may be underestimated, with FeO and TiO2 content estimated using the maturity-corrected method. Estimates of chemical content are 7.42–18.82 wt% FeO and 1.48–2.1 wt% TiO2, with a low-medium Mg number (Mg # ~ 55). Olivine-rich materials are not present at the CE-4 landing site, based on the low-medium Mg #. Multi-origin materials at the CE-4 landing site were analyzed with regard to concentrations of FeO and TiO2 content, supporting our conclusion that the materials at CE-4 do not have a single source but rather are likely a mixture of SPA impact melt components excavated by surrounding impact crater and volcanic product ejecta.

Through October 25, 2020, the Yutu-2 rover of the Chang'E-4 (CE-4) mission had toured 519.29 m up to 600 days (for 21 lunar days), with the Visible and Near-Infrared Image Spectrometer (VNIS) on board the rover acquiring 106 in situ reflectance measurements (Fig. 1). This in situ scientific dataset provides a unique perspective from which to understand the mineralogical and chemical properties of the Von Kármán crater inside the South Pole-Aitken (SPA) basin on the far side of the moon 1,2 . A range of remote-sensing and in situ data, some based on spectral observations, analysis of geomorphology, and the subsurface structure within Von Kármán crater, have helped develop understandings of the moon's history and evolution.
After the CE-4's safe landing, researchers have reported initial results by interpreting the first lunar-day VNIS data showing that the materials along the rover route are dominated by olivine and low-calcium pyroxene (LCP)-suggesting that the Yutu-2 rover may have detected the moon's mantle material 3,4 . These results are consistent with orbital data, confirming that the floor of the SPA is rich in mafic materials 5 .
However, additional data from the Yutu-2 rover have not indicated the presence of expected mantle materials at the CE-4 landing site, where materials are dominated by plagioclase (Plag) and pyroxene but include little olivine (OL), for the first three lunar days 6 . As exploration continues to produce more scientific data, we can arrive at new and different understandings. Several inversion methods could cause the controversial components  9 and other empirical approaches 6 ) may produce unexpected results. Accordingly, more lunar-days' worth of data are needed to research mineralogical properties along the rover's route and support use of an objective unmixing method based on radiative transfer models. By contrast, the lunar surface material is affected mainly by space weathering, a natural process that affects spectral band depths 10 , hindering precise modeling of lunar mixtures 10,11 . Notably, various mature materials such as rock and regolith were found along the Yutu-2 rover's route 6,7,12 . The regolith is more mature than the surrounding rocks 13 . Olivine and pyroxene, the major mafic minerals of lunar material, have distinctive spectral characteristics in the visible and near-infrared (VIS/NIS) and short-wave infrared (SWIR) regions 8,14 : for example, pyroxene exhibits absorption at ~ 1000 nm and ~ 2000 nm, whereas olivine has three overlaps in absorption near 1000 nm but no absorption at 2000 nm. Olivine and pyroxene differ in their sensitivity to space weathering 15 , making it difficult to ascertain the extent to which space weathering and mineral composition influence the spectrum. Thus chemical characteristics (such as iron and titanium content) of the lunar surface are needed to be concerned except for lunar minerals.
Iron and titanium are two elements of particular interest on the moon and are useful for understanding its origin and geological evolution. The major rock types can be distinguished by concentrations of FeO and TiO 2 content 16 supplemented and compared with the results of mineral inversion. Although optical maturity strongly influences spectral reflectance 10,[17][18][19] , there are ways of reducing the effect of space weathering when inversion of FeO and TiO 2 content. Generally speaking, two types of models are used, based on spectral data. One is based on the statistical relation of reflectance spectra and compositions, such as principal component analysis 20,21 , the support vector machine model 21 , and partial least squares regression 22 , mathematical methods that show potential for estimating element content by spectra. However, purely mathematical statistics are limited by mature differences in the training data, producing artifacts when applied to other regions. The other algorithms were maturity-corrected by parameterizing spectral properties of sensitivity to iron or titanium and soil maturity (e.g., Lucey et al. 1995Lucey et al. 18 , 1998Lucey et al. 16 , 2000 ; Le Mouélic et al. 2000 23 ), an approach that can normalize the effects of space weathering, optimized to remove mature effects. Thus FeO and TiO 2 content and the optical maturity parameter (OMAT) of Yutu-2 spectral data can be estimated using Lucey et al. 's method 19 to help explain the moon's geological background.
In this study, we analyze in situ spectra observed by VNIS spectral data during the first 21 synodic days, with a view to reporting mineralogical and chemical information for the region of the Yutu-2 rover and offering insights into the nature and origin of the SPA basin. Based on analysis of the spectral data, we attempt to determine the mineralogical and chemical properties of the materials at the landing site.

Results
Mineralogy properties observed by Yutu-2 rover. The VNIS spectra measured by the Yutu-2 rover show absorption regions around 1000 nm and 2000 nm (Fig. 2b). These absorption characteristics can be used to distinguish the primary minerals on the lunar surface, including pyroxene (PYX), olivine (OL), and plagioclase (Plag), because their absorption positions vary with Fe/Ca/Mg content [24][25][26] . As Fig. 2 shows, similarities in www.nature.com/scientificreports/ absorption peak position notwithstanding, the band depth of the rock (LE00303) is larger than for other nearby regolith, indicating similarities of components but differing degrees of space weathering.
To quantitatively infer the abundance of major minerals from the VNIS spectra measured by the Yutu-2 rover, we use the Hapke radiative model, in combination with a sparse unmixing algorithm, to estimate their relative content. In view of the space weathering effect, we add the agglutinates as the products of space weathering to estimate the relative contents of the primary minerals 27 . Our estimates show an average modeled composition of 50.5% agglutinates (AGG), 16.3% low-calcium pyroxene (LCP), 14.8% high-calcium pyroxene (HCP), 6.1% olivine (OL), and 12.3% plagioclase (PLG) (Fig. 3a). The modeled mineral abundance of the CE-4 observations suggests that their main sources are norite/gabbro and olivine norite/gabbro. Furthermore, we see higher LCP than HCP at most observation sites, consistent with the findings of the recent study 7 . Figure 3b shows detailed abundance information for every VNIS detection. Most observations (50 of 86) show higher LCP concentrations than HCP, indicating that composition variation within a small area of the Yutu-2 route is consistent with that suggested by Huang et al. 7 for the CE-4 landing region, based on multiple sources. Chemistry properties observed by Yutu-2 rover. From VNIS spectral data, adopting the method used by Luccy et al. 19 , we derived FeO and TiO 2 content for Yutu-2 in situ measurements, with estimated content of 7.42-18.82 wt% FeO and 1.48-2.1 wt% TiO 2 Fig. 4). FeO + TiO 2 is consistent with estimates (11-19 wt%) based on remote sensing data 19,28,29 . The results show that surface materials have higher FeO content but lower TiO 2 content.
Elaborating on the chemical properties of the CE-4 landing site, Fig. 4 plots CE-4 FeO versus TiO 2 for major lunar rock type fields and CE-3 landing site materials. CE-3 data are from Ling et al. 30 , and lunar rock (mare basalt and nonmare rock) data are reported by Lucey et al. 16 , Lindstrom et al. 31 , and Papike et al. 32 LKFM (low-K Fra Mauro), with basaltic impact melts thought to compose the lower lunar crust 33 . Figure 4 shows that CE-4 composition is unlike that of any major lunar rock type, but with the range of CE-4 spanning a variety of rock types-suggesting a complex source for CE-4. (For further discussion, see the next section.) What's more, CE-4 composition differs from CE-3, with materials at the CE-4 landing site having lower FeO content than CE-3, indicating that materials at the CE-4 landing site are more mature, consistent with Is/FeO maturity indexes of ~ 82 for the CE-4 landing site 13 and ~ 53 for CE-3 34 .
Mg number (Mg # = mole percent MgO/(MgO + FeO)) is one of the chemical indexes used to discuss the origin of lunar samples. The Mg number (Mg # = mole percent MgO/(MgO + FeO)) at the CE-4 site was empirically estimated by the 1000-nm and 2000-nm band centers, as Fig. 5 shows. We seek to identify mafic compositional trends and assess whether they are related to band centers, with the magnesium ratio classified as high (Mg # > 75), medium (Mg # 50-75), or low (Mg # < 50). The materials at the CE-4 site have a relatively low-medium Mg # (Fig. 5), consistent with the Mg # of ~ 55 derived from the Chang'E-1 Imaging Interferometer (IIM) data 29 .

Discussions
Space weathering introduces changes to visible and near-infrared reflectance spectra 35 : because the moon is an airless body, its surface is altered through irradiation by solar wind and galactic cosmic rays 36 , a process that reduces the strength of mineralogical absorption bands, including the mafic minerals pyroxene and olivine. Various types of pits are on the path of the Yutu-2, and the fragments that fill them show varying degrees of space weathering (Fig. S1), providing a unique perspective on space weathering's influence on mineral abundance, based on in situ spectra on the far side of the moon. Optical maturity (OMAT) is a spectral parameter used to estimate the degree of space weathering for lunar soils, based on reflectance properties. We calculated the optical maturity parameter 19 and produced a scatterplot (Fig. 6) of the relationship between OMAT and lunar mafic materials (HCP, LCP, and OL), showing that the three minerals have different weathering resistances. OL has decreased tendency with increasing maturity, whereas OL has weak resistance to weathering. LCP has better weathering resistance, whereas HCP shows relative increases with maturity, because they are in the relative content system, indicating that HCP has better weathering resistance. Because of the effects of space weathering, olivine may be underestimated, so further discussion of the sources of materials at CE-4 should be combined with the chemical properties.
The CE-4 landing site is in the Von Kármán crater, which formed in the pre-Nectarian era and subsequently underwent a series of complex geological evolutions [37][38][39][40] . The materials at the CE-4 landing site measured by Yutu-2 are dominated by pyroxenes and plagioclase, with little olivine seen-suggesting that the rock type is norite/gabbro. The norite-like surface materials at the CE-4 landing site may contain materials of the noritic layer during SPA melt pool formation 41,42 , excavated by the surrounding impact craters (e.g., Finsen crater and Alder crater). Furthermore, no olivine-rich materials were found at the CE-4 landing site with low-medium Mg #, suggesting that the source might not be the lunar mantle, although the possibility of the mixture's coming from the lower crust cannot be ruled out.
By contrast, elevated CE-4 FeO content (Fig. 4) may result from exposure from lunar mantle 16 , but TiO 2 content is significantly higher than the expected mantle composition. The correlation of FeO with TiO 2 of CE-4 shown in Fig. 4 could result from the differentiation of the large melt sheet, with the differentiation materials perhaps containing parts of LKFM (lower crustal composition).
The CE-4 materials have a distinctive chemical characteristic compared with other various mare basalt (Fig. 4), and the chemical composition at CE-4 may offer insights into the mixing effects of impact melt ejecta and volcanic products.
We note multiple lava-infilling events 45 within the Von Kármán crater, ~ 70 km west of the CE-4 landing site. Volcanic activity may have peaked in the Late Imbrium and could have continued into the Eratosthenian epoch 45 , with multiple lava flooding products excavated by the fresh Zhinyu crater (~ 4 km diameter), ~ 30 km west of the CE-4 landing site 12 . The orbital data indicate the presence of a slight impact ray to the CE-4 landing site from the Zhinyu crater, based on color ratio imaging from Clementine UVVIS data (Fig. 7).
As already discussed, the CE-4 landing site has been affected by the mixing effects of ejecta and volcanic products. Accordingly, multiple lavas were probably infilled in the bottom of Zhinyu crater and the CE-4 landing site before the Zhinyu impact event, containing the ejecta and volcanic products from the Zhinyu crater brought to the surface of the CE-4 landing site.
This paper investigates the mineralogical and chemical properties by in situ spectra as measured by Yutu-2, finding that the materials detected by Yutu-2 were norite/gabbro. No olivine-rich mantle materials 3   pixel VIS/NIR imager and a short-wave infrared (SWIR) single-pixel non-imaging spectrometer 3,14 . The VNIS is fixed on the front of the rover, offering a 45° viewing angle at a height of 0.69 m. Its spectral range is 450-2395 nm, with 400 spectral channels at a default 5-nm resolution. The VIS/NIR imager works from 450 to 945 nm, with 300 channels, and the 100-channel SWIR spectrometer works from 900 to 2395 nm, with 10 overlap channels 1,14 . The SWIR spectrometer's field of view is circular, with a center of (96, 128) in the VIS/NIR image and a diameter of 107 pixels.
VNIS data processing. This section describes VNIS data processing. VNIS spectral data processing began with level 2B data, which underwent dark current, flat-field, temperature corrections and radiometric and geometric calibration 3,46-48 . L2B radiance data were converted to reflectance data by solar irradiance, for an airless planetary surface, with bidirectional reflectance commonly given as radiance factor (RADF) I/F, where J( ) is the solar irradiance 49 , S( ) the spectral response of the VNIS sensor, and I s ( i ) the radiance measured by Yutu-2 at wavelength i . To reduce the influence of the geometric circumstances, the photometric correction described in Lin et al. 6 was used to correct the data to fit common viewing geometry (incidence angle = 30°, emission angle = 0°, phase angle = 30°).
Owing to differences in signal-to-noise ratios (SNR) between the VIS/NIR spectrum (37.4 dB) and SWIR spectrum (44.0 dB) 8 , a gap is seen at the boundary of the two spectra. With no obvious slope difference at the overlapping region (900-950 nm), and taking into account SWIR's higher SNR, we multiplied the VIS/NIR spectrum by a factor (reflectance ratio of SWIR 900-nm to the VNIS 900-nm) to connect to the SWIR spectrum and obtain a continuous spectrum. Finally, Savitzky-Golay method was performed during the spectral data smoothing, shown in Fig. 2.

Estimation of absorption band center.
In lunar mineralogical studies, the absorption band center is an important spectral parameter 50 , especially for mafic minerals (olivine and pyroxene), which have two obvious absorption bands, band1 (~ 1000 nm) and band2 (~ 2000 nm). The properties of absorption band centers are related to mineral structure and chemical composition 51 .
To calculate the band center, we first used the straight-line continuum method 52 to obtain the continuumremoved spectra. We then used a sixth-order polynomial to fit the bottom of the continuum-removed absorption feature around band1 (~ 1000 nm) and band2 (~ 2000 nm) 48 . The band center is defined as the wavelength at the minimum point on the polynomial fitted curve. We set three different spectral ranges for band1 and band2 when polynomial fitting, and take the average center values as the final results.
Estimation of the mineral abundances from the spectra. In this study, we used the radiative transfer model 53,54 (see Supplementary Text S1) and the sparse unmixing algorithm 55,56 to estimate mineral abundance from the VNIS spectra. To avoid any thermal emission effects, we limited our analysis to the wavelength range 950-1500 nm. First we calculated the imaginary part (k) of the endmembers (shown in Table S1) using the Hapke model. Next we computed the single scattering (SSA) for each endmember with different particle sizes Figure 7. The color ratio image 43 www.nature.com/scientificreports/ (ranging from 5 to 200 nm, with an interval of 5 nm) using the Hapke slab model, constructed as the endmember library. Finally, we used the sparse unmixing algorithm to unmix the SSA of the Yutu-2 spectra derived by the Hapke model.

Derivation of iron and titanium abundance.
To evaluate the chemical properties of the materials detected by Yutu-2, we calculated FeO and TiO 2 content using the algorithms described by Luccy et al. 19 . These approaches were based on returned Apollo and Luna samples. FeO and TiO 2 content was calculated from the best fit curve, where the origin (x 0Fe , y 0Fe ) and (x 0Ti , y 0Ti ) were set to (-0.1, 1.39) and (-1.08, 0.208), optimized to maximize the correlation coefficient between content and θ Fe using the Apollo lunar samples 10  www.nature.com/scientificreports/ Reprints and permissions information is available at www.nature.com/reprints.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.