Introduction

The redox conditions, determined by oxygen fugacity (fO2), for basaltic magmas originating from the melting of planetary mantles provide significant insights for determining the compositions of the precursor materials that contributed to planetary formation and the subsequent evolution of their interiors1,2,3. For example, the oxygen fugacity is crucial for ascertaining the stability and composition of crystallizing phases, the oxidation state of multivalent elements, and the composition and stability of volatile phases in a magma4,5. Therefore, the oxygen fugacity of the planetary mantle is crucial for comprehending its origin and evolution. The Moon displays a global asymmetry encompassing basalt distribution6, geomorphology7, chemical composition8,9 and crustal thickness10. However, the question of whether there is an asymmetry of oxygen fugacity between the lunar nearside and farside remains unanswered. This is attributable to the fact that all studied lunar samples were collected from the nearside of the Moon11,12. The Apollo lunar mare basalts and pyroclastic glasses yielded an oxygen fugacity of ΔIW –1 using olivine/spinel V oxybarometers4,13,14. A recent study indicates that the lunar mantle fO2 derived from CE5 basalts is ΔIW –0.84 using the same oxybarometers15. These results indicate that the fO2 of the lunar nearside mantle is nearly identical15.

China’s Chang’e-6 (CE6) mission, which successfully returned the first lunar farside samples (1935.3 g) from the SPA basin (41.625 °S, 153.978 °W), provides a unique opportunity to investigate the potential variation in fO2 between the lunar nearside and farside. Recent studies have revealed that the majority of the basaltic fragments in the CE6 soil are low-Ti basalt formed 2.8 billion years ago (Ga) (refs. 9,16,17), which are characterized by ultra-depleted Sr–Nd isotopic compositions18 and low water content19. However, the fO2 of this basalt has not been determined.

In this work, we conducted comprehensive analyses of the spinel V oxybarometer and pyroxene Eu oxybarometer on 23 CE6 basaltic fragments, with the objective of constraining the oxygen fugacity of the lunar farside mantle. The result indicates that the fO2 of CE6 basalts is more reduced than that of the nearside Apollo and CE5 basalts. The findings will facilitate a more comprehensive understanding of the oxygen fugacity between the nearside and farside and offer new insights for the origin of lunar asymmetry.

Results

Petrography and mineralogy of the CE6 basalts

The CE6 basaltic fragments examined in this study exhibit a diverse range of textures, including poikilitic, subophitic, and coarse-grained textures (Supplementary Fig. 1 and Supplementary Table 1). The primary minerals include clinopyroxene, plagioclase, and ilmenite, while the secondary minerals consist of olivine, silica, spinel, troilite, and apatite. The clinopyroxene exbibits notable chemical variability, with compositions of En0.2–60.7Fs20.8–91.1Wo7.4–39.5 (Supplementary Fig. 2 and Supplementary Data 1). The rare earth element (REE) concentrations of clinopyroxene are significantly lower than those of the CE5 basalts (Supplementary Fig. 3 and Supplementary Data 2; ref. 15). The plagioclase has a relatively more homogeneous composition (An82.2–92.5) and lower TiO2 contents than those in the CE5 basalts (Supplementary Fig. 4 and Supplementary Data 1; ref. 15). The modal abundance of olivine in the CE6 basalts is less than that observed in the CE5 basalts15,20. The majority of olivine grains are fayalite and occur in mesostasis, exhibiting a Fo [molar (Mg/Mg+Fe)] ranging from 0.3–2.1, indicative of a late-stage crystallization product. The mineral chemistry of the CE6 basaltic fragments studied here aligns with the low-Ti basaltic fragments previously reported18, which yielded an Pb-Pb age of ca. 2.8 Ga (refs. 9,16,17).

The spinel in the CE6 basalts manifests in two forms, which are Cr-spinel and Ti-spinel. The Cr-spinel that crystallized in the early stage is less abundant than the Ti-spinel (Supplementary Fig. 5). The V contents in the spinel (2372–6621 μg g−1; Supplementary Data 3) are significantly higher than those of the CE5 basalts15. The occurrence of Fe-Ni metal grains in the CE6 basalts is commonly observed. They occur at the grain boundaries of Cr-spinel in the early phases, at the boundaries of late-formed plagioclase and pyroxene, or within the troilite formed in the late-stage fractional crystallization (Supplementary Fig. 6). The Ni and Co concentrations of the Fe-Ni metal vary from 0.36–12.24 wt% and 0.82–2.38 wt% (Supplementary Data 4), respectively, which fall within the field of endogenous Fe-Ni metals21. Minor quantities of troilite ( ~0.37 wt%) have been observed in the CE6 basalts (Supplementary Table 1). They occur sporadically as inclusions within pyroxene or plagioclase, or as individual grains in the late-stage mesostasis associated with fayalitic olivine and apatite (Supplementary Fig. 7). The δ34S of the troilite varies between –2.6‰ and 1.7‰, with the inclusions having higher δ34S values than the grains in the mesostasis (Supplementary Data 5). This trend was also observed in the CE5 basalts and has been considered to be the result of magma degassing22,23.

The fO2 of the CE6 basalts

The fO2 of the CE6 basalts was determined using the clinopyroxene-melt Eu partition coefficient (DEu) and the spinel-melt V partition coefficient (DV). These coefficients can be calculated using the Eu concentrations in clinopyroxene and the V concentrations in spinel measured in this study, as well as previously reported whole-rock Eu and V concentrations18. The whole-rock Mg# of the CE6 basalts is ~30, with average Eu and V concentrations of 1.99 μg g−1 and 80.53 μg g−1, respectively18. According to the equilibrium Fe-Mg partition coefficient between pyroxene and melt (KD(Fe-Mg) cpx-liq = 0.28 ± 0.08; ref. 24), the clinopyroxene in equilibrium with the CE6 basalts is expected to have Mg# values ranging from 54.3–60.5. The fO2 of the CE6 basalts, as determined by the clinopyroxene Eu oxybarometer25, varies from ΔIW –2.62 ± 0.22 to ΔIW –1.25 ± 0.54, with an average of ΔIW –1.97 ± 0.60 (2σ) (Fig. 1 and Supplementary Data 6). The fO2 of the CE6 basalts derived from the spinel V oxybarometer26 ranges from ΔIW –1.82 ± 0.08 to ΔIW –1.63 ± 0.08, with an average of ΔIW –1.73 ± 0.04 (2σ) (Fig. 1 and Supplementary Data 7). The average fO2 of the CE6 basalts is then estimated to be ΔIW –1.93 ± 0.58 (2σ). Given that the Mg-rich clinopyroxene and Cr-spinel are the earliest crystallizing phases of the basalts20, this fO2 estimation can be considered to represent the redox state of the original melt of the CE6 basalts.

Fig. 1: Oxygen fugacity of the CE6 basalts compared to mare basalts from the lunar nearside.
figure 1

The fO2 of the CE6 basaltic fragments were determined by the spinel V oxybarometer and pyroxene Eu oxybarometer. The red and blue bars represent the average fO2 of the farside and nearside mantles. The average fO2 of the lunar nearside mantle is derived from the Apollo 12, 14, and 15 basalts27 and CE5 basalts15. The average fO2 of the lunar farside mantle is derived from the CE6 basaltic fragments. The error bar represents 2 standard deviations for each sample. Px pyroxene, Spl Spinel.

Full size image

Discussion

A comparison of our results with published results on the CE5 and Apollo basalts enables us to constrain the difference in fO2 between the lunar nearside and farside. The fO2 of the CE6 basalts (ΔIW –1.93 ± 0.58) with an age of ca. 2.8 Ga from the lunar farside is lower than the fO2 (ΔIW –0.80 ± 0.64) derived from the lunar nearside CE5 and Apollo lunar basalts with ages ranging from 3.9 Ga to 2.0 Ga (refs. 15,27; Fig. 2). This result implies that the fO2 of the lunar farside mantle is more reduced than that of the lunar nearside mantle (Fig. 2). The fO2 of CE5 and Apollo lunar basalts from the lunar nearside has remained relatively stable from 3.9 Ga to 2.0 Ga (ΔIW –0.84 to ΔIW –0.52) (Fig. 2) despite the heterogeneous volatile contents of their mantle sources and the varying degrees of partial melting and fractional crystallization processes they have experienced28,29 (Supplementary Fig. 8). This observation suggests that the oxygen fugacity of the lunar mantle may not be associated with local mantle heterogeneity or processes. Only major events may have been capable of inducing the more reduced oxygen fugacity of the farside mantle. Two potential candidates are the asymmetric crystallization of the lunar magma ocean30 (trend ① in Fig. 2a), and/or the SPA-forming massive impact31 (trend ② in Fig. 2a).

Fig. 2: The asymmetric oxygen fugacity and volcanism between the lunar nearside and farside.
figure 2

a The temporal evolution of mantle fO2 between the nearside and farside. The purple vertical bar denotes the initial fO2 of the lunar mantle (from ΔIW –1 to ΔIW –2; refs. 67,68). The blue solid line represents the temporal fO2 evolution of the lunar nearside mantle. The red lines represent the potentially temporal fO2 evolution of the lunar farside mantle. The more reduced lunar farside mantle may be attributable to two mechanisms: (i) The asymmetric crystallization of the lunar magma ocean has induced a more reduced lunar farside mantle in comparison to the nearside mantle. This initial fO2 (ΔIW –2) has remained unchanged since then (corresponding to the trend ①); (ii) The initial fO2 of the lunar farside mantle was analogous to that of the nearside mantle (ΔIW –1) and was subsequently reduced by the SPA massive impact (4.25 Ga; ref. 37) (corresponding to the trend ②). The ages for CE5, CE6, and Apollo basalts are from refs. 17,69,70,71. b Lunar volcanism asymmetry between the nearside and farside. Lunar sampling sites are indicated in yellow.

Full size image

One potential mechanism for the observed asymmetry in oxygen fugacity is the asymmetric crystallization of the lunar magma ocean, which may have resulted in oxidation of the nearside mantle while maintaining an initial fO2 on the farside. This scenario requires a greater amount of Fe metal sinking into the core from the nearside than the farside during the crystallization of the lunar magma ocean. Gravity and laser altimetry measurements have long indicated a 2 km displacement of the lunar center of mass relative to its center of figure in the nearside direction30,32. In light of this offset, a model for the early evolution of the Moon was proposed that involved an asymmetric accumulation of a liquid iron alloy, which displaced the cold, primordial, undifferentiated core upward30. The sinking of Fe metal from the nearside mantle to the core would elevate the Fe3+/∑Fe of the nearside mantle, resulting in an increase in its fO2 (Fig. 3). Although this proposed cold primordial core is not consistent with the giant impact origin of the Moon which would predict that the early Moon was uniformly heated33, there remains a possibility for asymmetric crystallization of lunar magma ocean (LMO) with a hot core34,35,36. However, the current models do not consider the potential variation in oxygen fugacity during asymmetric crystallization. Further research is required to explore this possibility.

Fig. 3: Mechanisms for mantle fO2 difference between the lunar nearside and farside.
figure 3

The black dot represents the lunar center of figure; the red dot represents the lunar center of mass. In the mechanism ①, the Fe metal sank to the core from the nearside mantle and elevated its Fe3+/∑Fe, resulting in a relatively oxidized nearside mantle than the farside mantle. This process should result in an asymmetry between the nearside and the farside. In the mechanism ②, the abundant S2 and CO degassing induced by the SPA massive impact occurred and caused Fe2+ → Fe0, and consequently, a relatively reduced mantle beneath SPA basin. This process should lead to an asymmetry between the SPA basin and PKT. Future comparative studies on the lunar northern hemisphere of the farside and southern hemisphere of the nearside are required to validate these two mechanisms. SPA South Pole–Aitken, PKT Procellarum KREEP Terrane.

Full size image

An alternative mechanism is the reduction of the farside mantle by the SPA massive impact. The SPA basin is the most extensive and ancient impact basin on the Moon, with a diameter of ~2400 km, a depth of ~6.2–8.2 km, and an age of ca. 4.25Ga (refs. 11,37). Massive impacts are expected to generate a substantial impact-heating anomaly and induce mantle convection38,39. Two potential processes may have been responsible for the reduction of the mantle beneath the SPA basin. One is the delivery of Fe metal from the impactor into the mantle, which can lead to a decrease in fO2. However, the Ni/Co values of metal grains in the CE6 basalts ranging from 0.44–5.45 are consistent with those of the Apollo low-Ti basalts (Supplementary Fig. 9), indicating an endogenous origin21. This observation suggests that if the delivery of Fe metal into the mantle did occur, the amount involved would have been insignificant.

Another consequence of the massive impact is the degassing of S2 and CO caused by shock wave heating. The passage of the shock wave caused by the impact leads to a significant increase in pressure and temperature40, resulting in volatile degassing41, which can alter magma fO2 (refs. 42,43,44). The loss of sulfur, existing as S−2 ion in the melt and as S2 gas, facilitates Fe2+ to Fe0 through the reaction of FeS → Fe0 + S2, which may contribute to the reduction of the lunar mantle43. Such a reduction mechanism has been observed on Earth and Mars that about 3500 μg g−1 or 4000–7000 μg g−1 S loss led to a decrease of ~2 log units in fO2, respectively45,46. The δ34S of the CE6 primitive magma prior to degassing was determined to be ~0.2‰ ± 0.5‰ (Supplementary Fig. 10). This result implies that the mantle source of the CE6 basalts might have undergone a certain degree of sulfur degassing ( ~45%), potentially reducing the fO2 of the lunar farside mantle. Carbon monoxide degassing is another plausible reduction mechanism42, following the reaction of FeO + C → Fe0 + CO (ref. 47). This degassing pathway is supported by the fact that CO is the most abundant gas released from lunar basalts at high temperature48. However, the process remains unverifiable due to technical challenges in determining the indigenous carbon in lunar samples. This necessitates the development of high-precision techniques, the enhancement of better resolution stepwise combustion with minimal sample loss, and the establishment of a relationship between exposure ages of samples and their carbon content and isotopic composition49. Nevertheless, if the SPA massive impact resulted in a decrease in fO2 of the underlying mantle (Fig. 3), it would indicate that the massive impact event had a profound effect on the deep interior31,50.

While the verification of the above two mechanisms remains challenging, the asymmetric crystallization of the lunar magma ocean and/or the SPA-forming massive impact. As illustrated in Fig.3, the asymmetric crystallization of the lunar magma ocean should induce an asymmetry between the nearside and farside. However, the SPA massive impact is likely to result in an asymmetry between the SPA basin and PKT (Procellarum KREEP Terrane). It is therefore imperative that future comparative studies should be conducted on the northern hemisphere of the farside and the southern hemisphere of the nearside, in order to decipher these two mechanisms. In addition, the analysis of the CE6 basalts for volatile isotopes, such as K, Cl, and Zn, has the potential to elucidate the influence of degassing processes resulting from the SPA massive impact. Nevertheless, the reduced nature of the underlying mantle beneath the SPA basin provides another aspect of lunar asymmetry, which should be considered by the models for the lunar asymmetry origin.

Methods

Sample preparation and properties

This study utilized 23 basaltic fragments sourced from several Chang’e-6 (CE6) lunar soil samples (including CE6C0200YJFM001EGP01,CE6C0200YJFM001EGP02,CE6C0200YJFM001EGP03,CE6C0200YJFM001EGP04, CE6C0200YJFM001EGP05, CE6C0200YJFM001EGP06, CE6C0200YJFM001EGP07, CE6C0200YJFM001EGP08, CE6C0200YJFM001EGP09 and CE6C0200YJFM001EGP10) provided by the China National Space Administration. They were embedded in one-inch epoxy mounts and polished prior to examination. These basaltic fragments exhibit different textures, such as typical subophitic, poikilitic, and coarse-grained textures are also found in these basaltic fragments. Coarse-grained texture denotes to the basaltic fragments >100 μm, characterized by big and few mineral grains (2–5), which do not distinctly align with the aforementioned textures. The sizes of these basaltic fragments range from a minimum of 0.3 mm to a maximum of 2.7 mm. All basaltic fragments comprise principal minerals such as clinopyroxene (20.22–79.89 vol%), plagioclase (13.75–59.31 vol%), and ilmenite (0.16–23.34 vol%), along with auxiliary minerals including olivine (0–27.01 vol%), silica (0–7.77 vol%), spinel (0–4.47 vol%), sulfide (0.04–3.46 vol%), phosphate (0–2.09 vol%) (Supplementary Table 1) and other minor minerals (Fe-Ni metal, whitlockite, baddeleyite, zirconolite, tranquillityite, etc.).

Scanning electron microscopy

Petrographic analyses, encompassing back-scattered electron (BSE) imaging, textural assessments, and mineralogical abundance modeling of the CE6 lunar basalts, were conducted utilizing a Thermofisher Apreo field emission scanning electron microscope (FE-SEM) equipped with a Bruker XFlash 60 energy dispersive spectrometer (EDS) detector in the Electron Microscopy Laboratory, Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS)51. The EDS data were obtained using the Maps-Nanomin automated mineralogy software. This study utilized electron beam currents ranging from 6.4 nA–13 nA and acceleration voltages between 15 kV and 20 kV. An X-ray for mineral analysis was conducted with a step size of 2 μm and a dwell period of 8 ms.

Electron probe micro-analyzer (EPMA) major and trace elemental analysis

EPMA can effectively identify the principal constituents of diverse minerals with excellent spatial resolution (1–5 μm), and high detection limits of up to several hundred μg g−1. The major elements in olivine, clinopyroxene, plagioclase, and ilmenite in the CE6 lunar basalts were conducted using the JEOL JXA8100 electron probe at IGGCAS. The experiment utilized an acceleration voltage of 15 kV, a beam current of 20 nA, a beam diameter of 3 μm, and a counting duration of 10–30 s. The calibration of data using various natural minerals and synthetic oxides aimed to enhance the accuracy and precision of these major elements in the primary minerals of the CE6 basalts. The analytical errors for major elements are below 1.5%.

Recent advancements in EPMA have facilitated the detection of trace elements, effectively addressing the limitation of laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS) in detecting trace concentrations due to the diminutive size of minerals. Recent research introduced a high-precision EPMA technique capable of concurrently identifying major and trace elements in spinel, demonstrating accuracy in the spinel of CE5 basalts52. On this basis, we also employed the JEOL JXA8100 electron probe micro-analyzer at IGGCAS to estimate the concentrations of major and trace elements in the spinel of the CE6 basalts. Spinel was examined using an acceleration voltage of 25 kV and a beam diameter of 1 μm. The beam currents for trace elements and major elements were 200 nA and 60 nA, respectively. The major and trace elements were identified utilizing five analytical crystals as follows: two TAP for Mg (Kα), Al (Kα), and Si (Kα), one LIF for Cr (Kα), Ni (Kα), and Mn (Kα), one LPET for Ti (Kα) and V (Kα), and one LLIF for Fe (Kα), Co (Kα), and Zn (Kα). Corrections for peak overlap of Cr (Kß) to Mn (Kα) and Ti (Kß) to V (Kα) were implemented. Natural minerals and synthetic oxides employed for high-precision elemental analysis included rhodonite (Si and Mn), rutile (Ti), FeS2 (Fe), ZnS (Zn), Cr2O3 (Cr), CoO (Co), NiO (Ni), MgO (Mg), and V2O5 (V). The peak counting durations were as follows: 10 s for Al, Mg, and Fe; 20 s for Cr; 80 s for Ti, V, Co, and Zn; 120 s for Mn and Ni; and 240 s for Si. The homogeneous spinel standard LBS13-04 was examined to assess machine drift, hence enhancing the accuracy and precision of these trace elements of spinel derived from the CE6 basalts. The mean detection limits (3σ) for trace elements ranged from 16 μg g−1 to 55 μg g−1, with a specific value of 18 μg g−1 for vanadium (V). The estimated accuracies (1σ) for the major elements and trace elements were superior to ±2 and ±6% ( ~2% for V), respectively.

The compositions of troilite in the CE6 basalts were examined with a JEOL JXA8100 electron probe micro-analyzer at IGGCAS. The experiment employed a 40 nA beam current, a 20 kV accelerating voltage, and a beam diameter of 2 μm. The analytical standards included FeS2 (S and Fe), apatite (P), rutile (Ti), pure Ni, Co and Cu metals (Ni, Co and Cu), Ca-Al silicate glass (Si), PbCrO4 (Cr), ZnS (Zn). The peak counting times were 10 s for Fe, S and Ni; 20 s for Ti, Si, P, Cu and Zn; and 40 s for Cr and Co. The detection limits (μg g−1) are as follows: 150 for Fe, 70 for S, 180 for Ni, 50 for Ti, 45 for Si, 40 for P, 60 for Cr, 75 for Co, 200 for Cu and 250 for Zn, respectively. Troilite in the Mangui ordinary chondrite (L6) was also analyzed to monitor the analysis quality.

The compositions of metal grains in the CE6 basalts were analyzed using the JEOL JXA8100 electron probe micro-analyzer at IGGCAS. The experiment utilized a 40 nA beam current, a 20 kV accelerating voltage, and a 2 μm diameter beam. The EMPA standards comprised pure Fe, Ni, Co, Cr, Ti, and Si materials, FeS2 (S), and apatite (P). The designated peak counting intervals are as follows: 10 s for Fe, S, and Ni; 20 s for Ti, Si, and P; and 40 s for Cr and Co. The detection limits (μg g−1) are as follows: 240 for Fe, 180 for Ni, 90 for Co, 85 for Cr, 70 for Ti, 70 for Si, 80 for S, and 50 for P, respectively53. Schreibersite in the Canyon Diablo iron meteorite was examined to verify the analytical accuracy.

Laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) trace elemental analysis

The trace element analyses for pyroxene in the CE6 basalts were conducted using LA-ICP-MS with an Element XRHR-ICP-MS apparatus linked to a 193-nm argon-fluoride excimer laser system (Geolas HD) at the State Key Laboratory of Lithospheric Evolution, IGGCAS. The analytical and testing protocols were derived from a previously published study54. The laser diameter is roughly 44 um with a repetition rate of 5 Hz. The laser energy density is approximately 5 J/cm2. Helium was used as a carrier gas to enhance the signal. An acoustic smoothing instrument (ASI) was employed to enhance signal accuracy. Each spot analysis started with a 20 s background collection (gas blank), succeeded by a cleaning process including two laser pulses, prior to the 35 s ablation data capture. The NIST SRM 610 and ARM-1 glass were utilized for external calibration. The GOR128-G and GOR132-G standards were employed for quality control assessment. The NIST SRM 610, ARM-1, GOR128-G, and GOR132-G standards were evaluated for accuracy, with each set comprising ten measurements. Data reduction was implemented by bulk normalization as a 100 wt% strategy utilizing the Iolite 3.4 software package with a self-developed data reduction scheme code. The mean detection limits (2σ) for most trace elements are better than 0.05 μg g−1. The analytical errors for the majority of trace elements (>0.05 μg g−1) are superior to 10% (σ). Additionally, the LA-ICP-MS approach may concurrently measure the concentrations of major elements at identical locations, with an accuracy above 5% (σ).

NanoSIMS S isotopic analysis

The Sulfur isotopic ratios of troilite were conducted using a CAMECA nano-scaled secondary ion mass spectroscopy (NanoSIMS 50 L) at IGGCAS. The samples were coated with carbon and pre-loaded in the sample chamber two days ahead in advance to enhance vacuum quality and reduce interference from 32S1H2− and 33S1H−. The vacuum in the analysis chamber was approximately ~5 × 10−10 mbar. The samples were pre-sputtered using a high primary beam of around 500 pA across a region of 10 × 10 μm2 area for 2–3 min to eliminate the carbon coating and to stabilize the secondary ions yield. For sample analysis, the raster area was diminished to 5 × 5 μm2 to 7 × 7 μm2, contingent upon the size of troilite, while maintaining the same beam current. In the static multicollection mode with nuclear magnetic resonance-based magnet regulation, the secondary ion species of 32S, 34S were detected with FC1 and FC2, respectively. The FC preamplifier boards were set with 1011 Ω resistors for 34S, 1010 Ω resistors for 32S, and temperature regulation was implemented to stabilize the FC preamplifiers. The equipment was calibrated to attain a mass resolving power of around 5000. Under these operational conditions, the counting rate of troilite ranged from the counting rate of troilite was 5 × 108 to 6 × 108 cps for 32S−. Each analysis has 200 cycles with a counting duration of 1 s per cycle. And the background of FCs was measured 20 s before and after each analysis to correct the base line. Two troilites Canyon Diablo and Nandan were examined to assess the precision of our measurements. The external reproducibility (2 SD) was better than 1.0‰ for the measurement of 34S/32S. The measured 34S/32S ratios are expressed as δ34S in the standard per mil notation (‰): δ34S = [(34S/ 32Ssample)/ (34S/32Sstandard – 1)] × 1000, where the standard adopted is VCDT (Vienna Canyon Diablo Troilite) with a 34S/32S value of 0.044162 (ref. 55).

Spinel V oxybarometer

Experimental research shows that the partitioning behavior of V between minerals and melt is sensitive to redox conditions, shown by the increasing DVspinel/melt with decreasing oxygen fugacity56,57. Previous results show that V oxybarometers have enormous potential because these methods allow for applicability over a range of redox conditions from the most reduced materials in the solar system, to the most oxidized terrestrial magmas13,56,57. The oxygen fugacity of volcanic pyroclastic glasses and lunar basalts has been determined by V oxybarometers14,15. This not only validates the efficacy of V oxybarometers but also aids in comprehending the oxygen fugacity of the lunar mantle. Given the similarity in whole rock MgO content18,58 and pyroxene composition15,20 (Supplementary Fig. 2) exhibited by the CE6 and CE5 basalts, the temperature of both basalts was estimated to be ~1200 °C. The experimental data of DVspinel/melt and fO2 at approximately 1200 °C (ref. 26) were compiled, and the following equation was obtained by means of fitting: fO2(ΔFMQ) = 3.193( ± 0.228) – 1.930( ± 0.115) × ln(Dvspinel/melt) (Supplementary Fig. 11a). In this equation, ΔFMQ signifies the fO2 disparity between the experimental fO2 and FMQ buffer. The V concentration in this study was ascertained from three CE6 basaltic fragments, with an average value of 80.53 μg g−1 (ref. 18). ΔIW is the fO2 disparity between experimental fO2 and IW buffer, with ΔIW is roughly 3.5 log units lower than ΔFMQ (ΔIW = ΔFMQ–3.5) at a temperature of 1200 °C (ref. 15).

The calculation of the fO2 uncertainties was conducted by propagation law, utilizing the uncertainties of the equation and the V analysis by EPMA. As the SE values for V analysis are almost identical ( ~2%), the uncertainties of fO2 for each point are also identical ( ±0.08). The applicable lower limit of the above equation is approximately ΔIW –0.5, whereas the oxygen fugacity of CE6 basalts is estimated at approximately ΔIW –1.73, a value for which a model extrapolation is required. However, the magnitude of this model extrapolation is minimal (from ΔIW –0.5 to ΔIW –1.73). In principle, Dvspinel/melt and fO2 should remain negatively correlated (Supplementary Fig. 11a). Therefore, the uncertainty engendered by this model extrapolation is deemed to be minimal.

Pyroxene Eu oxybarometer

The Eu mineral/melt partition coefficient (DEu) depending on the Eu3+/Eu2+ ratio in the magma is predominantly a function of the prevailing redox conditions during crystallization25,59,60,61. This reliance of the Eu2+/Eu3+ ratio on fO2 is exemplified by the reaction: Eu3+ + 1/2O2 → Eu2+ + 1/4O2 (ref. 62). The partitioning of Eu between pyroxene and melt as a function of fO2 has proven to be an effective oxybarometer, especially for basalt samples from the Mars and Moon as their pyroxenes display zonation in REE concentrations that aligns with fractional crystallization from their parent magma20,25,61,62. This implies that they have preserved their initial magmatic DEu ratio between pyroxenes and melt, being unaltered by subsequent processes. Moreover, pyroxenes are the first REE-bearing minerals to crystallize from the parent melt of lunar basalts28,63. Therefore, their DEu ratios remain unaffected by the previous crystallization of plagioclase (which scavenges Eu) and are primarily a function of magmatic redox conditions. Eu2+ exhibits lower compatibility in pyroxene than Eu3+. The greater the negative anomaly in pyroxene, the lower the oxygen fugacity during its crystallization. Research indicates that the distribution coefficient (DEu) between pyroxene and melt constantly increases with elevated fO2 (ref. 25). We compiled four partition coefficients (DEu = 0.086, fO2 = ΔIW − 1; DEu = 0.157, fO2 = ΔIW; DEu = 0.182, fO2 = ΔIW + 1; DEu = 0.274, fO2 = ΔIW + 3.5) to construct a function relating fO2 and DEu, expressed as fO2 (ΔIW) = 24.503( ± 2.990) × DEu – 3.407( ± 0.560) (Supplementary Fig. 11b). The R² of this function is 0.971, aligning with ref. 25.

The calculation of the fO2 uncertainties was conducted by propagation law, utilizing the uncertainties of the equation and the Eu analysis by LA-ICP-MS. The Eu content of pyroxene in the CE6 basalts ranges from 0.055 μg g−1 to 0.247 μg g−1 and the 2SE values range from 0.018 μg g−1 to 0.082 μg g−1. The calculated fO2 uncertainties range from 0.22 to 0.78. The lower limit of the equation that is applicable is approximately ΔIW –1, while the oxygen fugacity of CE6 samples is estimated at approximately ΔIW –1.97. This also requires a modal extrapolation. However, the magnitude of this model extrapolation is relatively small (from ΔIW –1 to ΔIW –1.97). In principle, the positive correlation between DEu and fO2 should not change (Supplementary Fig. 11b). Therefore, the uncertainty engendered by this model extrapolation is also deemed to be minimal.

Sulfur degassing and primitive S content of CE6 basalts

The δ34S of the CE6 primitive magma prior to degassing and the bulk CE6 magma was determined to be approximately 0.2‰ ± 0.5‰ and −1.6‰ ± 0.5‰ derived from the maximum δ34S (1.7‰ ± 0.5‰) and the average δ34S ( −0.1‰ ± 0.5‰) of the troilite (Supplementary Data 5), respectively, utilizing the equilibrium fractionation factor of 1.0015 between troilite and melt64. The degassing loss of sulfur was determined to be ~45% according to the equilibrium Rayleigh fractionation model. Equations and variables employed: δ34Sm = (δ34Si + 1000)f (ɑ-1) – 1000; Cm = Ci(1–F)D-1; α = 34S/32Sgas/34S/32Smelt; D = Sgas/Smelt; f = the fraction of S remained in the melts; F = fraction of degassing removed; m = residual melt; i = initial melt. Here the average αgas−melt value (1.003) was chosen under reduced conditions based on ref. 65. The estimated average modal abundance of sulfides in the CE6 basalts is approximately 0.37 wt% based on EDS analysis (Supplementary Table 1). The sulfur abundance of the bulk CE6 basalts was estimated to be approximately 1343 ± 672 ppm derived from the average sulfur content of troilite (36.3 wt%; Supplementary Data 8) and its modal abundance ( ~0.37 wt%), accounting for a 50% error in the statistical modal abundance. The primordial sulfur abundance of the CE6 magma is estimated to be approximately 2442 ± 1221 ppm predicated on a degassing rate of about 45%. Data of CE5 and Apollo samples are from refs. 22,66, respectively.