Ubiquitous and progressively increasing ferric iron content on the lunar surfaces revealed by the Chang’e-5 sample

Although ferric iron indisputably exists on the highly reducing surface of the Moon, its formation mechanism and evolution are still under debate. Here we show that micrometeorite impact-induced charge disproportionation of iron could have produced the large amounts of ferric iron (average Fe3+/∑Fe > 0.4) in agglutinate melts returned by China’s Chang’e-5 mission. The charge disproportionation reaction synchronously generated nanophase metallic iron (npFe0), and quantitative analyses of iron valence indicate that it is a dominant pathway for formation of npFe0 within the lunar agglutinate glass. The discovery of the charge disproportionation reaction in the agglutinates suggests that much more Fe3+ could be present on the Moon than previously thought, and that its abundance is progressively increasing with micrometeoroid impacts. Lunar high-concentration ferric ion (Fe3+/∑Fe > 40%) and ~63% of nanophase metallic iron (npFe0) are produced via charge disproportionation of ferrous iron from micrometeoroid impacts, as observed in the Chang’e-5 sample. This ongoing process would lead to a continuously increasing abundance of Fe3+ in the lunar regolith.

Iron is a critical element that records the oxygen potentials of solar system materials [1][2][3] . The lunar surface and interior are identified as highly reducing because Apollo samples mainly contained ferrous (Fe 2+ ) or metallic iron (Fe 0 ), and only small amounts of ferric iron ions (Fe 3+ ) (<1 wt%) were detected 4,5 . Despite the highly reduced state of the Moon, a higher Fe 3+ concentration (for example, 0-25% Fe 3+ /∑Fe (∑Fe = Fe 2+ + Fe 3+ ) in lunar picritic glass beads) in Apollo samples has recently been revealed 1,6-8 . Furthermore, recent remote observations have also shown widespread haematite, an Fe 3+ -bearing mineral, at high latitudes on the Moon 9 .
The occurrence of Fe 3+ in lunar materials has been discussed for decades [10][11][12] . Previous studies have mainly ascribed its origin to degassing of reductive gases, including H 2 and CO, from lunar melts 1,6 or secondary oxidation, both on the lunar surface 8,9 and by terrestrial atmospheric oxygen 7 . For instance, alteration of minerals by lunar ice may constitute a pathway for Fe 3+ formation 13,14 . In addition, recent examinations of asteroid Itokawa samples have suggested that charge disproportionation associated with solar wind implantation and H 2 O formation produced ferric iron 15 . Almost all of these interpretations have suggested that the formation of lunar Fe 3+ requires another oxidant, such as lunar ice 13,14 or terrestrial oxygen 9,16 , and occurs on the lunar subsurface and/or surface. However, as an airless body, the Moon suffers space weathering due to solar wind irradiation and micrometeoroid impacts, which mainly produce a reductive driving force for formation of metallic iron [17][18][19][20] . How Fe 3+ forms, accumulates and evolves in the reductive environment of the Moon surface is still under debate.
Here we present robust evidence from a lunar sample (CE5C0400YJFM00408) for the formation of a large amount of lunar Fe 3+ (Fe 3+ /∑Fe > 0.4) from charge disproportionation of iron during micrometeoroid impacts on the airless lunar surface. The sample was collected from the lunar regolith surface at 43.06° N, 51.92° W in the Oceanus Procellarum region of the Moon and returned by China's Chang'e-5 (CE5) mission and provided by the China National Space Administration (CNSA) 21 . The charge disproportionation reaction Article https://doi.org/10.1038/s41550-022-01855-0 agglutinate particles, we employed a scanning electron microscope (SEM) operating at low voltage (<2 kV) and without any conductive coating. The agglutinate looks very smooth and is composed of glass, while many spherical particles, including nanophase metallic iron (npFe 0 ) and some glassy rods, are present on the surface (Fig. 1a). The glassy rods and surface melt splashes (Extended Data Fig. 1) indicate repeated micrometeoroid impacts originating from the agglutinate particle 17,22 . Some iron globules occur randomly, while some are arranged in lines on the surface, consistent with previous observations of the Apollo samples 23 .
synchronously produced Fe 3+ and Fe 0 that were fixed within the lunar agglutinate glass to prevent comproportionation. This mechanism implies that a large quantity of Fe 3+ exists in the lunar regolith, and its abundance is increasing as micrometeoroid impacts still occur on the lunar surface.

Occurrence of iron in CE5 agglutinate glass
Agglutinate particles with diameters of ~100 μm were collected from the CE5 powder sample. To observe the fine surface structures of the   To examine the inner structure of the agglutinate glass, a section was cut from the agglutinate particle by using a focused ion beam (FIB) in a dual-beam system. In the high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of the FIB section (Fig. 1b), npFe 0 globules occurred not only on the surface but also inside the agglutinate glass. A rounded hole presented at the bottom right of the FIB section denoted a vesicle in the agglutinate glass, which is a ubiquitous phenomenon in agglutinates 24 . Our measurements of the iron particle size in the FIB section showed an average of 16.2 nm and a median of 14.4 nm with a minimum of 3.2 nm and a maximum of 125.1 nm (Fig. 1c). This particle size distribution was larger than those of previous transmission electron microscopy (TEM) measurements (mean grain size 7 nm) 25 but smaller than those of SEM observations (mean grain size 90 ± 7 nm) 26 .
High-resolution TEM images and the corresponding fast Fourier transform patterns (Fig. 1d) show that the npFe 0 globules occur as crystalline α-Fe (body-centred cubic structure). The structures of α-Fe globules and agglutinate glass were confirmed by their fast Fourier transform pattern spots and dispersive ring features (insets of Fig. 1d), respectively. The α-Fe structure is consistent with the structure of iron nanoparticles produced by simulated micrometeoroid impacts (that is, laser irradiation) on lunar minerals 25 , suggesting that the agglutinate originated from micrometeoroid impacts. The agglutinate might not be a condensation product because condensation experiments produced γ-Fe (face-centred cubic structure) 27 . Energy dispersive spectroscopy (EDS) results for the FIB section ( Fig. 1e−f, Table 1 and Extended Data Fig. 2) further confirmed that the chemical compositions of the globules and agglutinate glass were pure iron and iron-bearing aluminosilicate, respectively.

Chemical oxidation state of iron and Fe 3+ abundance in CE5 agglutinate glass
We further measured the chemical oxidation state of the iron in npFe 0 and agglutinate glass by using electron energy loss spectroscopy (EELS). The core-loss spectra of Fe L 2,3 in phases with various valence states exhibit indicative features (for example, peak positions and shapes) [28][29][30][31] . The Fe L 2,3 EELS spectra of standard reference materials (including olivine, haematite and iron metal) measured in this study (Extended Data Fig. 3) showed that Fe 0 and Fe 2+ presented peaks in the range 708.1-708.9 eV, whereas Fe 3+ has an Fe L 3 peak at ~710 eV, consistent with literature values 31 . Fe 0 /Fe 2+ can be easily distinguished from Fe 3+ by using the peak positions. The peak positions of Fe 0 and Fe 2+ are always too close to be separated for various phases 7,8 , although the standards shown in Extended Data Fig. 3 have distinct peak positions for Fe 0 and Fe 2+ . Except for the peak position features, tail features in the range 725-730 eV, after removing the plural scattering, can be used to distinguish Fe 0 from Fe 2+ ; that is, Fe 0 has a tail feature with higher intensity than Fe 2+ or Fe 3+ ions 7 .
By comparing the typical spectra of the agglutinate particle ( Fig. 2a) with those of the standards, the oxidation states of iron in the npFe 0 and agglutinate glass were determined. The Fe L 2,3 spectrum of npFe 0 showed a typical Fe L 3 peak at ~708.4 eV and a typical high tail feature for Fe 0 . The Fe L 2,3 spectrum of the agglutinate glass illustrated both a typical Fe L 3 (~709.9 eV) peak for trivalent iron and a shoulder peak at 708.4 eV for divalent iron, indicating the coexistence of Fe 3+ and Fe 2+ in the agglutinate glass. To visualize the distributions of phases with various valence states of iron within the agglutinate, we used the typical spectra of npFe 0 and agglutinate glass as references to fit the spectroscopic data. Multivariate least-squares fitting maps of various zones in the FIB section ( Fig. 2b and Extended Data Figs. 4 and 5) showed that the phase distributions in the npFe 0 and agglutinate glass were identical to the distribution observed from the HAADF-STEM image. The distribution of phases was further confirmed by the electron tomogram of EELS mapping shown in Fig. 2c, which illustrated the three-dimensional (3D) distribution of the iron valence states in the npFe 0 and agglutinate glass (Supplementary Video 1).
The amount of Fe 3+ in the agglutinate glass was further assessed with a modified method using the integral intensity ratio of the Fe L 2,3 white lines with 2-eV-wide integration windows, the absolute errors of which were approximately ±0.04 for Fe 3+ /∑Fe ratios 30,31 . This method employed universal line fitting that is applicable to various iron-bearing materials. Typical energy windows used in calculating the signal integral ratio I(L 3 )/I(L 2 ) are shown in Fig. 2a. The calculated integral I(L 3 )/I(L 2 ) ratios were then converted to Fe 3+ /∑Fe ratios through the equation 31 : where I(L 3 )/I(L 2 ) is the integral intensity ratio and x is the ferric iron concentration Fe 3+ /∑Fe. The parameters a, b and c equal 0.193 ± 0.007, −0.465 ± 0.009 and 0.366 ± 0.003, respectively 31 . We then mapped the Fe 3+ /∑Fe ratios of the agglutinate glass using this method. The maps of various zones in the FIB section ( Fig. 3a and Extended Data Figs. 4c and 5c) showed that the Fe 3+ /∑Fe ratios were 0 in the npFe 0 zones while the ratios ranged from 0 to 1 in the glass zones. The frequency and cumulative distributions of the Fe 3+ /∑Fe ratios ( Fig. 3b and Extended Data Figs. 4d and 5d) showed that the median of the Fe 3+ /∑Fe ratio ranged from 0.40 to 0.43. The Fe 3+ /∑Fe ratios of 0.41 ± 0.03 in the 3 analysed zones were estimated by integrating the frequency distribution curve with the equation: where WA denotes weighted average of the Fe 3+ /∑Fe ratio; x i and f i denote Fe 3+ /∑Fe and the frequency of the ith interval in Fig. 3b and Extended Data Figs. 4d and 5d. The frequency distribution was then employed to estimate the weight percentages of iron in various valence states. The results ( Table 2) showed that the entire agglutinate glass contained 3.6 ± 0.5 wt% Fe 0 , 6.4 ± 0.3 wt% Fe 2+ and 4.5 ± 0.3 wt% Fe 3+ .

Disproportionation reaction during micrometeorite impact
The observation of coexisting Fe 0 , Fe 2+ and Fe 3+ throughout the CE5 agglutinate glass departed from previously proposed oxidation pathways. The possibility that simple oxidation of iron by degassing of reductive gases, including H 2 and CO, can be excluded because the agglutinate glass was formed by micrometeorite impact processes 6,32 . npFe 0 nanoparticles cannot be secondarily oxidized unless exoteric free oxygen diffuses to the npFe 0 through the agglutinate glass, because npFe 0 was mainly dispersed inside the glass, as shown in the electron tomogram ( Fig. 2c and   Article https://doi.org/10.1038/s41550-022-01855-0 diffusion coefficient makes oxidation at room temperature too slow to be detected. In addition, basaltic glass oxidation produces magnetite nanoparticles 33 , but these are absent from the CE5 samples. Therefore, the Fe 3+ present in CE5 agglutinate glass should not be ascribed to isolated oxidation reactions. The universal coexistence of Fe 0 , Fe 2+ and Fe 3+ in the CE5 agglutinate glass suggests that Fe 0 and Fe 3+ were produced via charge disproportionation of the Fe 2+ contained in the source materials for the agglutinate during micrometeorite bombardment. The spherical morphology of the npFe 0 indicates that the npFe 0 was formed in the liquid phase of silicate melts produced by micrometeorite impacts at high temperature and then quenched. Based on the theory of size-dependent melting of nanoparticles 35 , the melting temperature (T m ) can be expressed in terms of the bulk melting temperature (T mb = 1,536 °C for metallic iron) as where d is the diameter of the nanoparticles and β is a constant for the specific material (β = 0.94232 for metallic iron). In all of the observed iron globules in the agglutinate glass, the largest spherical npFe 0 had a diameter of 125.1 nm (Fig. 1c). Thus, the formation temperature of the iron globules must have been above the melting temperature of 1,524 °C (Extended Data Fig. 6), which was calculated for the 125.1-nm iron nanoparticle using equation (3). Therefore, the temperature for formation of the iron globules in the CE5 agglutinate glass should have been higher than 1,524 °C. Under such high-temperature conditions, the charge disproportionation reaction should occur in the liquid states of melts. The liquid melt that formed agglutinate glass originated from the impacts of micrometeorites 32 , suggesting that the charge disproportionation reaction occurred during the impact-melting processes at an elevated temperature >1,524 °C. However, an equilibrium charge disproportionation reaction is not expected at elevated temperatures >1,524 °C according to the phase diagram for the Fe-O system at 1 atm (ref. 36). There may be a specific condition required for a metastable reaction occurring at high temperature. The high temperature could have originated from the shock wave produced by the micrometeoroid impacts. The corresponding shock pressure for basalt could be >50 GPa based on the Hugoniot integral approximation 37 . An impact-induced high pressure could be one possibility, as natural charge disproportionation of iron was first reported for the bridgmanite formed by a shock pressure of 18-20 GPa in the Suizhou meteorite 38 . Thus, charge disproportionation of iron could occur under conditions involving shock-induced high pressure and high temperature. Even though the charge disproportionation in this study could have occurred under high-pressure and high-temperature conditions, it could also have occurred at a low temperature of <570 °C during cooling after impact, similar to that occurring in impact-evaporation processes 39,40 . Therefore, both the micrometeoroid impacts that induced high-pressure and high-temperature conditions and postshock cooling processes were available for the charge disproportionation reaction of iron. Assuming that the system is closed with respect to electrons (or oxygen) and that Fe 0 is not dissolved in the silicate melt, the charge disproportionation reaction of ferrous iron in impact-melting processes can be characterized by the following theoretical chemical reaction: where liq1 and liq2 denote the silicate melt and pure liquid metal, respectively. The average amounts of Fe 0 (3.6 wt%) and Fe 3+ (4.5 wt%) shown in Table 2 are not consistent with the reaction in equation (4), suggesting that (1) 1.4 wt% Fe 0 was originally present in the source material or (2) some electrons (or oxygen) were subtracted from the system by partial reduction, as in:   L 2 ). b, EELS map of Fe in various phases from multivariate least-squares fitting. The inset shows the corresponding HAADF-STEM image. c, 3D stereogram of npFe 0 dispersion in the agglutinate glass. The EELS mapping area is outlined (the red rectangle) in the inset of b. The stereogram shown in c was reconstructed using Tomviz 42 from a tilt series of EELS maps from −60° to +60° with an interval of 6°. The npFe 0 in the stereogram was rendered by iso-surface in red colour and the agglutinate glass was rendered by transparent volume in green colour.
Article https://doi.org/10.1038/s41550-022-01855-0 In scenario (2), the charge disproportionation reaction occurring during a micrometeoroid impact covered all of the iron in the source materials, which is not true on the lunar surface because the lunar regolith experiences repeated space weathering processes, including micrometeoroid impacts and solar wind radiation 17,18 . Thus, the original Fe 0 presented in scenario (1) could be a product of solar wind radiation.
The liquid Fe 2+ in the impact melt may have originated from melting of lunar minerals containing iron, such as the fayalite (Fe 2 SiO 4 ) and ferrosilite (FeSiO 3 ) components in olivine and pyroxene and/or basaltic glass. Thus, the reaction in equation (4) can be specifically rewritten as follows: These chemical reactions in equations (6)(7)(8) are further supported by the chemical components of the agglutinate glass (Table 1). We employed the CIPW (named after the petrologists Cross, Iddings, Pirsson and the geochemist Washington) normative mineralogy calculation 41 to quantitatively analyse the products of charge disproportionation, that is, agglutinate glass. The results (Supplementary Table 1) suggested that the chemical components of the agglutinate glass can be treated as a molten mixture of quartz, plagioclase, diopside, hypersthene, ilmenite and magnetite. If the source materials of the agglutinate were solely lunar minerals, the SiO 2 component supports the reactions in equations (6) and (7), while the Fe 3 O 4 component is the product of the reaction in equation (8).

Implications for the evolution of iron valence state on the lunar surface
The production of npFe 0 through charge disproportionation brought about by micrometeoroid impacts is as important as the production of Fe 3+ . Furthermore, production of npFe 0 by solar wind irradiation is another important process of space weathering and proceeds without Fe 3+ production 19 . The npFe 0 produced by solar wind irradiation may be considered the original Fe 0 present, as discussed above. Considering the repeated space weathering processes, including micrometeoroid impacts and solar wind irradiation, on regolith particles and excavation of the regolith by impact, the npFe 0 in agglutinate could be produced by both micrometeoroid impacts and solar wind irradiation, while the Fe 3+ in the glass should be produced by micrometeoroid impacts only.
We used the Fe 3+ /2(Fe 0 ) ratio to assess the effects of micrometeoroid impact and solar wind irradiation on the formation of npFe 0 . The average Fe 3+ /2(Fe 0 ) ratio of 0.63 indicates that approximately 63% of the total npFe 0 was produced by micrometeoroid impacts, while the rest was produced by solar wind irradiation. Thus, charge disproportionation of the ferrous iron in impact-melting processes could be the most important mechanism for formation of the npFe 0 and Fe 3+ on the lunar surface.
Importantly, the large amount of Fe 3+ in the agglutinate glass produced by charge disproportionation may enhance our knowledge of the evolution and distribution of lunar Fe 0 and Fe 3+ . The widespread agglutinate fragments in the lunar regolith 32 and ongoing micrometeoroid impacts with fluxes ranging from 6.19 × 10 −13 to 14.74 × 10 −13 g m −2 s −1 (ref. 22) suggest that Fe 3+ could be a universal component with increasing abundance on the Moon (Extended Data Fig. 7). This evolution of the iron valence state on the lunar surface could affect remote sensing interpretations and in situ observations and shed light on the space weathering induced environmental evolution of surfaces on airless bodies.

Sample preparation and storage
Agglutinates from the CE5 sample were collected from the sample in a glove box (N 2 > 99.999%, H 2 O < 0.01 ppm, O 2 < 1.0 ppm; Mikrouna, China). Selected agglutinates were set in the air on a single crystal silicon wafer and fixed by epoxy resin for further observations. The exposure time of the agglutinate in air was controlled in 30 min.

SEM observations
The silicon wafer with the agglutinate particle was then transferred in the air to a cold field emission SEM (Hitachi SU8010) for observation.  The Fe 0 was estimated from the npFe 0 abundance in the agglutinate glass, Fe 2+ and Fe 3+ were assessed from the chemical compositions of pure agglutinate glass without npFe 0 and the total Fe 3+ /∑Fe ratio from the EELS analyses.
Article https://doi.org/10.1038/s41550-022-01855-0 SEM observations were conducted in secondary electron image mode at low voltage (1-2 kV) with a 7-10 μA emission current because the sample surface was not coated with a conductive layer. The observations performed under low voltage minimized electron irradiation damage to the sample surfaces. During the SEM observations, the potential positions for FIB cutting were recorded.

FIB cutting and TEM analyses
After the SEM observations, the sample on the silicon wafer was coated with a carbon layer and then transferred to a ThermoScientific FEI Scios dual-beam system. The positions recorded for FIB cutting were deposited with Pt. FIB sections for TEM observations were then cut with a 30 kV Ga + ion beam in the dual-beam system. The FIB sections were then loaded into a ThermoScientific FEI Talos F200S TEM and observed in both TEM and STEM modes. EDS mapping was performed in STEM mode with two super-X detectors. The dwell time was 10.0 μs and the results from 50 frames were summed. EDS semiquantitative analyses from mapping data were conducted with the ThermoScientific FEI Velox software. EELS analyses were performed in STEM mode with a Gatan 1077 EELS spectrometer. The pixel steps for spectroscopic image acquisition were 2 nm. All the data were acquired in dual EELS mode with zero-peak locking. All the EELS data processing tasks (including background subtraction, signal integration, data fitting and mapping) were conducted with the Gatan Microscope Suite software (v.3.50).
STEM-EELS electron tomography data were collected manually using a Fischione cryo-transfer tomography holder 2250 at −175 ± 0.5 °C. The collection angles ranged from −60° to +60° with steps of 6°. At each angle, a STEM-EELS map was acquired. The raw mapping data were processed using Gatan Microscope Suite 3.50 software to produce valence state maps. Then, the valence state maps were reconstructed and visualized using Tomviz software 42 .

Abundance calculations for various iron species in agglutinate glass
We calculated the abundance of various iron valences via the following steps. (1) The total volume of iron globules in the EELS mapping area was calculated via particle size counting. (2) The volume fraction of npFe 0 in the EELS mapping area of the specimen was calculated. (3) The weight percentage of Fe 0 in the EELS mapping area of the specimen was calculated. The densities of Fe 0 and pure agglutinate glass were set to 7.86 and 3.12 g cm −3 , respectively. The density of Fe 0 was taken from the bulk density of α-Fe, while the density of pure agglutinate glass was estimated from a CIPW calculation. (4) The weight percentage of total iron (Fe 2+ + Fe 3+ ) in the agglutinate glass of the specimen was calculated based on EDS results of pure agglutinate glass (shown in Table 1). (5) The weight percentages of Fe 2+ and Fe 3+ in the EELS mapping area of the specimen were calculated based on the estimated Fe 3+ /∑Fe ratios. Calculation details are displayed in Supplementary Data 1.

Data availability
The experiment data that support the findings of this study are available via the figshare repository at https://doi.org/10.6084/ m9.figshare.21382611.v1 (ref. 43).