Impactor relics of CI-like chondrites in Chang’e-6 lunar samples

Significance

We found seven olivine-bearing fragments in lunar samples returned by the Chang’e-6 mission from a huge, ancient (pre-Nectarian) basin on the far-side of the Moon. We identified these uncommon clasts as remnants of carbonaceous Ivuna-like (CI) carbonaceous chondrites similar to asteroids Ryugu and Bennu based on olivine elemental and isotopic compositions. As CI chondrites are rich in water and volatiles, this finding supports the hypothesis that asteroids played a role in delivering water and other volatiles to the lunar surface. Given the rarity of CI chondrites in Earth’s meteorite collection, our integrated methodology for identifying exogenous materials in lunar and potentially other returned samples offers a valuable tool for reassessing chondrite proportions in the inner Solar System.

Abstract

The impact history of the Moon provides the opportunity to better understand mass transfer in the Solar System. While Earth’s meteorite collection serves as a key reference for material flux in the Earth–Moon system, it suffers from profound biases arising from Earth’s orbital dynamics and atmospheric filtering. Systematic identification and classification of meteorites on the airless Moon thus provide additional critical constraints for reconstructing the primordial accretion history and impactor population of the inner Solar System. However, identifying impactors on the Moon remains challenging due to their vaporization upon colliding at high velocities with the lunar surface. In situ remote sensing has previously detected chondritic impactor materials in the South-Pole-Aitken (SPA) basin of the far side of the Moon. The first opportunity to measure materials from the SPA basin has come via the Chang’e-6 (CE-6) mission, which returned samples from the Apollo basin inside the SPA basin. In this study, we screened seven olivine-porphyritic clasts as potential impactor relics in regolith returned by the CE-6 mission. These clasts were identified, via textural characterization, olivine Fe–Mn–Zn systematics, and in-situ triple oxygen isotopes, as impact relics solidified from melted chondritic parent bodies. Intriguingly, the parent body of all the identified impactor relics in this study resemble CI-like chondrites, a volatile-rich meteorite group that is relatively rare in Earth’s meteorite collection. The detection and classification of these impactor relics impose significant constraints on the proportions of meteoritic materials in the Earth–Moon system and their potential contributions to water inventories on the lunar surface.

Large craters and basins on the Moon preserve a record of lunar impact history (1), offering insights into impact processes that may have also affected the Earth. This history goes back more than 4.0 billion years, providing a record that cannot be obtained from terrestrial samples due to erasure by Earth’s active geological processes (2, 3). The high content of highly siderophile elements in the lunar regolith confirm the addition of chondritic materials to the lunar surface (4–10), A precise quantification of carbonaceous chondrite (CC)-like materials is particularly important because these impactors contain significant amounts of life-essential volatiles including water and organic compounds (11, 12). However, relative contributions of different types of chondrites remain unclear due to similar platinum-group elements signatures for different types of chondrites (13). The chondrite collection on Earth may provide a reference for extrapolation but is seriously biased due to Earth’s orbital dynamics and atmospheric filtering (14). Consequently, direct identification and classification of chondritic relics on the airless Moon are imperative to precisely quantify the amount and portions of various impactors, carrying significant implications for deciphering the impact history of the Earth–Moon system.

Analyses of lunar meteorites (15, 16) and returned samples (17–23) have identified parent bodies of some impactors, providing “ground-truth” constraints on the chondrite collection accessible to the Moon. Recent missions have extended this understanding: Analyses of asteroid samples returned by the Hayabusa2 and OSIRIS-REx missions reveal that both the Ryugu and Bennu asteroids closely resemble the CI group of CCs (24, 25). The CI chondrites bear elemental compositions similar to that of the Sun but are rarely preserved on Earth due to their friable nature. If CI-like objects dominate kilometer-sized carbonaceous near-Earth objects (26) and are widespread among C-complex asteroids (27), they should be well-represented in the lunar regolith. However, to date, such material had not been definitively identified. This absence highlights the necessity of reassessing the meteorite collection within the Earth–Moon system. While direct spectral measurements of returned Ryugu and Bennu samples provide guides for identifying potential CI parent bodies (28), definitive laboratory identification and classification are essential for understanding inventories of CC-like objects in the asteroid belt and the transport of related materials across the Solar System.

The CE-6 mission returned 1,935.3 g of regolith samples from the floor of the Apollo basin, within the lunar far-side’s South-Pole-Aitken (SPA) basin. The in situ detection of CC components in glassy materials in the SPA basin by the Chang’e-4 rover mission (29) indicates the preservation of chondritic materials therein. Furthermore, a recent study (30) has confirmed, from the elevated Ni content, that the CE-6 regolith contains 2 to 3 wt% of meteoritic materials, making CE-6 samples promising for identifying relics of CCs and tracing their sources. We were allocated a 2-gram aliquot of the CE-6 scooped samples (CE6C0100YJFM003) by the China National Space Administration. The sample powder was first scrutinized under a petrologic microscope. Clasts with sizes larger than ~50 μm were picked out, mounted into epoxy resin, and subsequently analyzed using Scanning Electron Microscopy. In this study, we focused on seven olivine-bearing clasts with mineralogies and textures that were distinct from any previously described mare basalts and other known lunar rocks. Detailed studies on the textures of the clasts, Fe–Mn–Zn systematics, and triple oxygen isotopes of the olivines were carried out to identify the formation processes and parent bodies of these clasts.

Results and Discussion

Textures of the Olivine Porphyritic Clasts.

The two-gram aliquot used in this study is similar to previously reported CE-6 regolith samples, primarily composed of local basaltic clasts, breccias, glass beads, and leucocratic clasts (30, 31). Seven olivine-bearing clasts separated from our aliquot (SI Appendix) were identified as potential impactor relics based on their distinctive mineralogies and textures. These clasts exhibit porphyritic olivine or porphyritic olivine-pyroxene textures (Fig. 1 and SI Appendix, Fig. S1). In all the clasts, olivine is the major crystallizing phase embedded in the mesostasis. The olivine crystals are subhedral to euhedral with sizes ranging from <30 μm to 80 to 100 μm (Fig. 1 and SI Appendix, Fig. S1). The olivines show both normal and reverse zoning (SI Appendix, Figs. S2 and S3). The normal zoning was formed by olivine crystallization during cooling, while the reverse zoning was formed by crystallization of olivines from evolved melt entrapped inside the hollows of early skeletal or dendritic olivines, a process also identified in terrestrial olivines (32, 33 and SI Appendix). Tiny melt inclusions are present in several crystals but were not analyzed due to their small sizes (<2 μm). Minor minerals include troilite and spinel. The troilite grains are dispersed in the mesostasis or aggregated into globules (Fig. 1 and SI Appendix, Fig. S1). The spinels occur sporadically within olivine crystals (Fig. 1A) or the mesostasis (Fig. 1C). The mesostasis is mainly composed of plagioclase and quenched pyroxene (Fig. 1 A and C and SI Appendix, Fig. S1) or devitrified glass (Fig. 1B). The plagioclase is generally lath-like with widths <5 μm and pyroxene intergrows with plagioclase or occurs as overgrowths at the rims of olivines as shown by the high-resolution X-ray elemental intensity mapping (SI Appendix, Fig. S2).

Fig. 1.

Representative backscattered electron images of the olivine porphyritic clasts in CE-6 lunar samples. (A and B) Clasts S3552-236 and S3637-001 showing a porphyric-olivine texture. The mesostasis of S3552-236 is composed of plagioclase and quenched pyroxene and that of S3637-001 appears as devitrified glass. (C) Spherical clast S3634-026 with a porphyritic olivine–pyroxene texture. Abbreviations: Ol, olivine; Py, pyroxene; Tro, troilite; Sp, spinel.

In general, the textures of the clasts resemble the porphyritic textures in type II chondrules (12, 34) or impact melts of chondrites (35, 36). Unlike the chondrules, which are usually surrounded by chondritic matrix and spherical in shape, the clasts are not surrounded by chondritic matrix and are angular in shape except for S3634-026 (Fig. 1C) that resembles a quenched melt droplet. With their ubiquitous zoning, these olivines likely crystallized from rapidly cooled impact-induced melts. Given the lack of relict minerals in the clasts, melting must have exceeded liquidus temperature. As the melts cooled, olivine crystals formed, and the solidified melt droplets eventually fragmented into smaller pieces (Fig. 1 and SI Appendix, Fig. S1). The crystallization of olivine does not necessarily require olivine presence in the original material but reflects the bulk composition of the impact melt (37).

Nonlunar Origin of the Olivine Porphyritic Clasts.

Here, we employ the Fe–Mn–Zn systematics of olivine as a proxy for identifying exotic materials in lunar samples (15, 16, 38). The Fe/Mn ratio should remain almost unmodified during impact-induced melting as evidenced by the similar Fe/Mn of olivines from mare basalts to that from lunar impact melts (39). Therefore, the olivine Fe/Mn ratio is applicable to distinguish whether these clasts are of lunar origin or not. We analyzed the major and minor element compositions of the olivine and troilite (also pyroxene and spinel in S3634-026) with an Electron Probe Microanalyzer (EPMA) (Materials and Methods, Datasets S1–S3) (see Materials and Methods for details). The Mg# [molar Mg/(Mg+Fe) × 100] of olivines in these clasts is variable, but generally >80 (Dataset S1A), which are more magnesium-rich than those from mare basalts (Mg# < 80) (Dataset S4). The molar Fe/Mn ratios of olivines are in the range of 49 to 81 (Dataset S1A), exhibiting a general increase with decreasing Mg# (SI Appendix, Fig. S4). The increase in Fe/Mn ratios with decreasing olivine Mg# may be accounted for by the progressive re-equilibrium between the crystallizing olivine and troilite (SI Appendix, Fig. S4). It is worth noting that the Fe/Mn ratios of the olivines are comparable to nonlunar olivines previously recognized in lunar samples (SI Appendix, Fig. S5). Notwithstanding the variations, the Fe/Mn ratios of olivine from the clasts are lower than those of mare basalts, the Mg-suite, and the ferroan-anorthosites (Fig. 2A) and are unlike any indigenous lunar samples previously described in the literature. The Fe/Mn ratios are also lower than that of the experimentally simulated lunar mantle (40). Further Zn content analyses of the olivines unambiguously preclude a potential lunar mantle origin of these olivines. The Zn content in olivine from the lunar mantle is expected to be several ppm considering that the bulk silicate Moon contains 1.0 ± 0.6 ppm of Zn (41), and the partition coefficient of Zn between olivine and melt is around 1 (42). However, the Zn content of olivines in these clasts is generally >100 ppm (Dataset S1B), which is remarkably higher than lunar mantle olivines, precluding their origin in the lunar mantle. These clasts are also not terrestrial contaminants as indicated by their distinctly lower Ni and higher Cr contents compared to mid-ocean ridge basalts (SI Appendix, Fig. S6). The olivine Fe/Mn ratios are also different from those of Mars and Howardites, Eucrites, and Diogenites (HEDs) (Fig. 2B), suggesting that these clasts were most likely derived from chondrites rather than achondrites. Collectively, these lines of evidence indicate the nonlunar origin of these clasts. Taking together the olivine compositions with the textural observations, we interpret these clasts to have been formed through impact-induced bulk melting of chondritic materials, followed by olivine crystallization during cooling, and subsequent fragmentation. In the following section, we investigate the possible precursors of these impactor relics using their triple oxygen (O) isotope signatures.

Fig. 2.

The Fe vs. Mn afu (atom per formula unit) of olivines from the clasts in this study. (A) A comparison of the compositions of olivines in this study with those in lunar (basalts, Mg-suite and FAN), Earth (represented by MORB olivines), Mars, and HED samples. Olivine compositions of the experimentally simulated lunar mantle are also shown for comparison; (B) The enlargement of the light yellow area in (A), showing clearly that the olivines from this study are different in Fe vs. Mn to Mars, and HED samples. The compositions of olivines in lunar, Mars, and HED samples are from the literature and reported in Datasets S4–S6 and that in MORB are from ref. 40. The experimentally simulated lunar mantle olivines are from ref. 43. All the olivine compositions are filtered to confine the molar (Mg+Fe+Ca)/(Si+Al) ratio in the range of 1.95 to 2.05.

The Chondritic Precursors of the Impactor Relics.

In this section, we further characterize the chondritic precursors of the impactor relics using triple O isotope ratios, a canonical means of identifying the precursors of extraterrestrial materials (19, 44). Triple O isotope ratios are expressed in terms of δ¹⁷O, δ¹⁸O, and Δ¹⁷O, where δ¹⁷O and δ¹⁸O are deviations in parts per thousand of the ¹⁷O/¹⁶O and ¹⁸O/¹⁶O ratios from the Vienna Standard Mean Ocean Water, and Δ¹⁷O (= δ¹⁷O − 0.52 δ¹⁸O) denotes the deviation of δ¹⁷O and δ¹⁸O from the terrestrial fractionation line (TFL) with a mass-dependent fractionation slope of 0.52 (44). Planetary bodies throughout the Solar System exhibit varying Δ¹⁷O values as a result of mass-independent fractionation processes in the Solar Nebula. These values are not modified by mass-dependent fractionation processes such as melting, evaporation, condensation, and crystallization (44). Both the Earth and Moon possess a Δ¹⁷O value of 0‰. The CC family typically exhibits negative Δ¹⁷O values, except for the rare CI (Ivuna-type) and CY (Yamato-type) groups, which display positive Δ¹⁷O values (Figs. 3 and 4). Notably, the CI and CY groups also possess the highest δ¹⁸O values among the CCs. The CYi subgroup has experienced thermal metamorphism and is characterized by higher δ¹⁸O values (45, 46) but CI-like Δ¹⁷O compositions (45). Recent measurements of asteroid samples returned from Ryugu and Bennu display Δ¹⁷O and δ¹⁸O values similar to CI chondrites, suggesting a genetic relation between Ryugu, Bennu, and CI parent asteroids (24, 25). Due to the shared similarly positive Δ¹⁷O values (0.4 to 0.9‰) and relatively high δ¹⁸O values (>10‰) of Ryugu, Bennu, and the parent asteroids of the CIs and CYi’s, these bodies are collectively referred to as CI-like chondrites in this study.

Fig. 3.

Triple oxygen isotope of olivine grains from six impactor relics. (A) Triple oxygen isotope of olivines in the clasts compared with chondrites. Previous measurements of olivines in a clast (60255–110) returned by the Apollo mission are also shown as a gray dot (38). Reference lines shown are TFL, the ¹⁶O fractionation line (Y&R) with slope = 1 (47), the equilibrated chondrites line (ECL) (48), and the CC anhydrous mineral (CCAM) line (49). The red line is the best-fit line of triple oxygen compositions of the studied olivine crystals. (B) Summary of Δ¹⁷O of studied olivine and potentially related CC groups. Fields for Ryugu and Bennu particles, CI and CY chondrites are from previous studies (24, 25, 27, 49–52). CY chondrites show two distinct groups with similar δ¹⁸O values, but distinct Δ¹⁷O compositions, and are labeled CYi and CYm (45). The gray shadow band denotes the average and ±2 SD of Δ¹⁷O obtained from clinopyroxene in nonimpact CE-6 basaltic clasts, which are identical (within analytical errors) to the bulk silicate Moon and Earth (0‰) as expected. The error bar represents 2 SE.

Fig. 4.

A comparison of the triple oxygen isotope compositions of the studied olivine grains with chondrites (36 and references therein) and water extracted from Apollo samples (19). The shaded gray region represents a lunar water array excluding outliers (Apollo sample 10060) as discussed in ref. 19. The CI-like region shows collectively the ranges of CI and CYi chondrites, and materials from Ryugu and Bennu missions. This figure presents a broader range of δ¹⁸O and Δ¹⁷O values compared to Fig. 3B. The error bar represents 2 SE.

The O isotopes of olivine in six clasts were analyzed in situ with Secondary Ion Mass Spectroscopy (SIMS), with 3 to 14 analyses performed on each clast (SI Appendix, Fig. S7 and Dataset S7) (see Materials and Methods for details). The textures of the clasts and compositions of the olivines (Fig. 1 and SI Appendix) indicate that the olivines crystallized from rapidly cooled impact melts. Generally, crystallized olivine inherits its O isotopic signature from the parent magmas (53). Mass-dependent fractionation processes may occur during impact melting and crystallization, slightly modifying the δ¹⁷O and δ¹⁸O values of the sample along a δ¹⁷O/δ¹⁸O slope of ~0.52. This means these processes do not alter Δ¹⁷O values. Therefore, the measured Δ¹⁷O values of the olivines are expected to reflect those of their chondritic parent bodies. It is thus noteworthy that all clasts exhibit positive olivine Δ¹⁷O values ranging from 0.38 to 0.90‰ (Fig. 3B), significantly higher than the Δ¹⁷O value of the Moon (0‰). The nearly zero Δ¹⁷O value of lunar clinopyroxenes from mare basalts in the CE-6 samples, also measured in this work (−0.10 ± 0.46‰, 2 SD, n = 47), confirms that positive Δ¹⁷O values observed in olivine are not an analytical artifact (Fig. 3B and SI Appendix). Although the clast with the lowest Δ¹⁷O value shows a relatively large Δ¹⁷O variability (0.38 ± 0.42‰, 2 SE n = 5), the positive olivine Δ¹⁷O values in all other clasts are clearly resolvable statistically (Fig. 3B). Coupled with δ¹⁸O values, these olivines exhibit a CI-like isotopic compositions, indicating that these clasts originated from CI-like parent bodies (Fig. 3). Interestingly, similar isotopic compositions were observed in olivine in a clast returned by the Apollo mission (38), but the reported Δ¹⁷O value of +0.38‰ had a relatively large uncertainty (0.41‰, 2 SE) (Fig. 3) due to the limited number of data (n = 4).

The scattering of these clasts in the δ¹⁸O – Δ¹⁷O plot (Fig. 3B) represents the isotopic heterogeneity of the clasts, rather than lunar regolith contamination or vaporization loss during impact events. Bulk analyses of Ryugu and Bennu samples reveal a dispersion of δ¹⁸O values (Fig. 3), indicating a high level of isotopic heterogeneity in these asteroids at the sample scales. Such δ¹⁸O variability aligns with our observations of a considerable variability of olivine δ¹⁸O values among and within the clasts (Fig. 3 and SI Appendix). The contamination of the impact melts by lunar regolith material is minimal, as indicated by the low Fe/Mn ratio of the olivines (Fig. 2). Lunar material contamination is anticipated to elevate Fe/Mn ratios at the high Mg# end of the olivines, which was not observed in our samples. In addition, any such potential contamination would have resulted in lower Δ¹⁷O and δ¹⁸O values for the crystallized olivine compared to those of the original impactors. Therefore, the positive Δ¹⁷O and high δ¹⁸O values in olivines strongly suggest that the precursors of the impactor relics had CI-like compositions. We can exclude alternative explanations for the observed positive Δ¹⁷O values. Ordinary chondrites (OCs) have positive Δ¹⁷O values ranging from ~0.5 to ~1.5‰, but their δ¹⁸O values are significantly lower than our observations (Fig. 3). If the high δ¹⁸O of the olivines were caused by impact-induced vaporization loss of OCs, a large mass-dependent fractionation in δ¹⁸O (up to 20‰) is expected to be accompanied by similar magnitude fractionation in δ³⁰Si (54). However, no noteworthy δ³⁰Si fractionation (<3‰) was observed in our samples (SI Appendix, Fig. S8), precluding OCs origin of these clasts. To summarize, the positive Δ¹⁷O and high δ¹⁸O values of the olivines unambiguously indicate the genesis of these clasts from CI-like materials (Fig. 3).

Implications.

In this work, we developed an integrated methodology combining clast textures, and olivine Fe–Mn–Zn systematics and triple oxygen isotope analysis to identify exogenous materials in lunar samples. This approach led to the first discovery of the “rare” CI-like materials in lunar samples. A recent study (55) reported, in the CE-6 sample, an olivine-bearing clast with textures similar to those observed in our samples, but with a near-zero Δ¹⁷O value. While preliminarily classified as primitive lunar mantle olivine, the possibility that the clast originated from an enstatite chondrite cannot be excluded, as both exhibit near-zero Δ¹⁷O values. The Zn content in olivines, as proposed here, could serve as a diagnostic criterion for distinguishing the origin of the clast, but was unfortunately not measured (55). The identification of CI-like materials in this study enables us to reassess chondrite proportions in the Earth–Moon system, given the Earth’s chondrite collection biases (14). A preliminary analysis of exogenous materials found in lunar samples from both this study and previous studies (17–23, 55–57) (Dataset S9) reveals that CCs (~37%) and CIs (~30%) are substantially more abundant in lunar than in Earth’s meteorite collection (CCs: ~3%; CIs: ~0.7%; 58), exhibiting closer to the model-predicted proportion (>50%) of CCs (26). While limited sample numbers (27 clasts in total, Dataset S9) preclude definitive conclusions, future studies applying our methodology to larger sample sets may allow more robust and statistically meaningful estimates of meteoritic contributions to the Earth–Moon system.

Our findings are especially significant in the ongoing debate of the inventories of water on the lunar surface. Recent Δ¹⁷O measurements of lunar surface water (Fig. 4) reveal signatures dominated by an enstatite-like signature but mixed with multiple sources (19). The variations in both Δ¹⁷O and δ¹⁸O in lunar water are notably larger than those in the olivine-bearing clast described here. As demonstrated in Martian meteorites (19, 59), water composition is influenced by processes beyond bulk rock equilibrium. In the case of lunar water, the effects of multiple sources including indigenous water and delivered water create a source mixing array. Lunar water exhibiting positive Δ¹⁷O and negative δ¹⁸O (Fig. 4) was attributed to mixing with cometary water, whereas the positive δ¹⁸O array end-member was suggested as CI chondrites (19). Our paired δ¹⁸O – Δ¹⁷O measurements of CI-like impactor relics provide unambiguous observational evidence supporting the preservation of a CI-like endmember on the Moon (Fig. 4). Despite the unconstrained ages and abundances of these clasts, our findings establish CI-like chondrites as a key reservoir of exogenous material delivered to the lunar surface. The mixing array broadly traces a region from the EC, OC, and CI-like endmembers to δ¹⁸O = −40‰ and Δ¹⁷O = 1‰ (Fig. 4). Interestingly, the clast S3621-065 studied in this work exhibits the highest Δ¹⁷O and lowest δ¹⁸O value among all studied clasts. The fractionation direction from the CI-like field is toward OC and parallel to the lunar mixing array of the δ¹⁸O – Δ¹⁷O plot (Fig. 4). Therefore, while CI-like chondrites stand as a source, it is feasible that OC meteorites or comets would contribute to the signature. Further combined measurements of triple O isotopes in water and impactor relics could provide precise constraints on the sources of impactors and their roles in delivering water and volatiles to the Earth–Moon system.

Materials and Methods

High-Resolution X-Ray Elemental Intensity Mapping.

To check the detailed zonings of olivines, high-resolution X-ray elemental intensity mapping for Ca, Mg, and Al was performed on clast S3552-236. The X-ray mapping utilized a JEOL JXA-8230 electron microprobe at the State Key Laboratory of Deep Earth Processes and Resources, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIGCAS). The operational conditions included an accelerating voltage of 20 kV, a probe current of 100 nA, and a beam size ranging from 0.5 to 1 μm. The Mg and Al were analyzed using a TAP crystal, and Ca was analyzed with a PETJ crystal. The Kα line was selected for all elements during the analyses. The step size was equal to the beam size, and dwell time for each point was set to 80 to 100 ms.

EPMA Analyses.

In situ major oxides of olivine and troilite were acquired using a CAMECA SXFive FE EPMA at the State Key Laboratory of Deep Earth Processes and Resources, GIGCAS. Acceleration voltage was set at 20 kV for all analyses. Beam current was set at 100 nA for olivine and 40 nA for troilite. Beam diameter was 1 μm for olivine and 5 μm troilite. The peak counting times ranged from 10 s to 120 s for different elements according to the characteristic X-ray line intensity and desired precision. Calibration standards included albite (Na), olivine (Si, Mg, Fe), diopside (Si, Ca, Mg), almandine (Al, Fe), orthoclase (K), rutile (Ti), chromite (Cr), rhodonite (Mn), synthetic RbTiPO₅ (P), Co metal (Co), V metal (V), pentlandite (Ni), pyrite (Fe, S), chalcopyrite (Cu), and willemite (Zn). The PAP (Pouchou and Pichoir) procedure was used for matrix correction (60). An olivine standard MongOL Sh11-2 (61) was analyzed to monitor the precision of the olivine analysis. Calculated detection limits were 91 ppm for Si, 110 ppm for Mg, 64 ppm for Na, 28 ppm for Al, 51 ppm for P, 50 ppm for Ti, 137 ppm for Fe, 74 ppm for Mn, 97 ppm for Ni, 39 ppm for Ca, 64 ppm for Cr, 91 ppm for Co, and 133 ppm for Zn. Analytical precision for SiO2, MgO, and FeO of MongOL Sh11-2 (1σ, n = 24) was better than 1%, while that for Al₂O₃, MnO, NiO, and CaO was within 10%, and that for Na₂O, Cr₂O₃, and CoO was within 20%. Analytical precision was not satisfactory due to the low concentrations of P₂O₅ and TiO₂ in MongOL Sh11-2. However, the analytical accuracy of the weight mean value for SiO₂, MgO, and FeO was also better than 1%, and that for other oxides, except for P₂O₅ (12%), was all within 10%.

In Situ Triple O Isotopes and Si Isotopes Measurements.

In situ triple oxygen isotopes analyses of olivine grains were measured with a CAMECA IMS 1,280 h SIMS at the State Key Laboratory of Deep Earth Processes and Resources, GIGCAS. For the triple O isotopes analysis session, a Cs+ primary beam of ~2 nA with an impact energy of 20 keV was used to sputter secondary ions from a ~12 μm sample area (7 μm spot size +5 μm rastering), with a normal-incidence electron gun (NEG) providing charge compensation. A NMR controller was used to stabilize the magnetic field. The entrance slit and the field aperture were set to 40 μm and 4,000 μm, respectively, and the energy slit was set to a 30 eV bandwidth and shifted 5 eV below the maximum transmission. Negatively charged secondary ions were accelerated by a voltage of 20 kV. The ¹⁶O- and ¹⁸O- ions were detected using two off-axis Faraday cup (FC) detectors posited at L1 and H1 with resistors of 10¹⁰ Ω and 10¹¹ Ω, respectively. The ¹⁷O- was simultaneously measured using the central FC with a resistor of 10¹² Ω. For ¹⁶O- and ¹⁸O-, 500 μm collector exit slits were used to yield a ~2,500 mass resolution power (MRP), while ~173 μm exit collector slit (corresponding to a ~7,000 MRP) was used for ¹⁷O- to avoid ¹⁶O¹H- interference. In addition, degassing in the storage chamber over 3 d and using the presputtering process lasted for 180 s with a raster area of 25 × 25 μm to further minimize the ¹⁶O¹H-trailing. A standard-sample-standard bracketing protocol was adopted, i.e., measuring one or two reference materials after four/five spots on the unknown samples. Triple oxygen isotope data of olivine were normalized to the measured primary standard San Carlos olivine (SCO), which was interspersed in the analytical sequence and has a recommended δ¹⁸/¹⁷O value of 5.24‰ and 2.73‰ (62). Secondary standard 06JY06 olivine separated from mantle peridotite xenoliths yield an average δ¹⁸O and ∆¹⁷O (δ¹⁷O – 0.52 × δ¹⁸O) values of 5.27 ± 0.36‰ and 0.06 ± 0.49‰ (2 SD, n = 39), respectively, consistent with its recommended values of δ¹⁸O (5.20 ± 0.06‰, 2 SD) (63) and terrestrial nature (∆¹⁷O = 0). The good agreement indicates the high accuracy of the analysis. The obtained 2 SD for these secondary standard ∆¹⁷O measurements (0.49‰) is treated as the uncertainty in each single measurement for CE-6 samples. Clinopyroxenes from CE-6 basaltic clasts were also analyzed for comparison. The clinopyroxene oxygen isotopes were calibrated with clinopyroxene 06JY31 and 06JY29 (63), yielding an average ∆¹⁷O value of −0.10 ± 0.46‰ (2 SD, n = 40), further corroborating the high quality of the analyses. To further evaluate the external precision of sample measurements, 3 to 14 analyses were obtained on different olivines from each relic. The 2 SD for multiple measurements of each relic range from ±0.12‰ to ±0.46‰ for ∆¹⁷O. This precision is sufficiently high to resolve the nonzero ∆¹⁷O values in this study. To further estimate how much a sample mean is likely to differ from the true population mean, we report 2 SE = 2 SD/sqrt(n) as the uncertainty in the main text. Previous studies have undertaken detailed work on SIMS matrix effects showing δ¹⁸O variations in olivines as a function of Fe molar fractions and demonstrated a parabolic correlation between instrumental fractionation of ¹⁸O/¹⁶O and olivine Mg# [molar Mg/(Mg+Fe) × 100] (64). However, in olivine with Mg# of 60-100, instrumental mass fractionation (IMF) was found to be too small to be resolvable by multiple labs equipped with CAMECA IMS 1270 and 1280 series (65–67). This olivine Mg# interval covers the compositional range of the olivines studied here (Mg#: 68-89). Therefore, IMF is negligible for oxygen-isotope ratios of olivine samples analyzed in this work.

For the Si isotope analysis session, a primary mass-filtered beam of Cs⁺ ions with an impact energy of 20 kV and an intensity of ~5 nA was focused on a ~15 μm area (8 μm spot size +7 μm rastering) to sputter secondary ions. The NEG and NMR are also employed to provide charge compensation and stabilize the magnetic field. In this session, a 122 μm entrance slit, a 5,000 μm field aperture, and a 600 μm exit slit were used. Three FCs located at the L1, H1, and H’2 positions of the multicollector system were used to measure the ²⁸Si⁻, ²⁹Si⁻, and ³⁰Si⁻ signals simultaneously, corresponding to the preamplifiers with resistors of 10¹⁰, 10¹¹, and 10¹² Ω, respectively. One measurement takes ~4 min and typically comprises 30 s preanalysis sputtering with a raster area of 25 × 25 μm, and 30 cycles of data collection. Si isotope data of olivine were normalized to the measured SCO, which was also interspersed in the analytical sequence and has recommended δ²⁹Si and δ³⁰Si values of −0.16‰ and −0.30‰, respectively (68). Due to the lack of secondary standard olivine to control the accuracy of the analytical procedure, we further analyzed two zircon standard samples (Penglai and Qinghu) (69) with known Si isotope compositions under the same instrument parameters. The results showed that the Qinghu zircon calibrated with Penglai zircon had Si isotope values (δ³⁰Si, −0.14 ± 0.23‰; δ³⁰Si, −0.42 ± 0.24‰, 2 SD, n = 5 Dataset S8) consistent with their recommended values (δ³⁰Si, −0.25‰; δ³⁰Si, −0.45‰, Dataset S8) (69). Ion yield and IMF in olivine Si isotopic analysis using SIMS show complex behaviors that rely on variations in Fo values. However, similar to oxygen isotope analysis, the IMF of Si isotopic analysis can be neglected within the Mg# range of 70 to 100 (70).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.