Isotopic evidence for volatile loss driven by South Pole-Aitken basin–forming impact

Significance

Chang’e-6 samples from the lunar farside could help resolve the origin of lunar asymmetry. High-precision isotopic analyses reveal that Chang’e-6 basalts are isotopically heavier in Fe and K than those from Apollo and Chang’e-5 missions. These heavy isotopic enrichments cannot be solely attributed to cosmogenic effects or impactor contributions. Although magmatic processes can explain the Fe isotopic data, the K isotopes necessitate a mantle source with a heavier K isotopic composition on the farside than on the nearside. This feature most likely resulted from K evaporation caused by the SPA basin–forming impact, demonstrating the profound influence of this event on the Moon’s deep interior. This finding also implies that large-scale impacts are key drivers in shaping mantle and crustal compositions.

Abstract

Recent studies suggest that the lunar farside experienced a magma ocean evolution similar to that of the nearside. Thus, the nearside-farside dichotomy, such as volcanism and crustal thickness, is likely related to the South Pole-Aitken (SPA) basin–forming impact. Although the noritic clasts found in Chang’e-6 (CE6) samples may originate from crustal remelting induced by the SPA impact, how (and whether) the lunar mantle was modified by this event remains unclear. Here, we present the first high-precision iron (Fe) and potassium (K) isotopic measurements of CE6 low-Ti basalts, revealing higher δ⁵⁶Fe (0.13 to 0.21‰) and δ⁴¹K (0 to 0.09‰) in these basalts compared to their Apollo and Chang’e-5 (CE5) counterparts (δ⁵⁶Fe: 0 to 0.11‰; δ⁴¹K: −0.29 to −0.04‰). The heavy Fe and K isotopic signatures are unlikely to be derived from cosmogenic effects or the addition of impactor-derived materials. Instead, the heavy Fe isotopes can be explained by partial melting and fractional crystallization processes. For K isotopes, however, the data require that the mantle source beneath the SPA basin had a heavier K isotopic composition than that of the nearside mantle, most likely resulting from evaporation caused by the SPA-forming impact. Our results thus provide robust evidence for significant impact-induced modification of the lunar mantle and demonstrate that large-scale impacts may have played a key role in creating lunar asymmetry.

Orbital observations have revealed significant differences in crustal thickness, magmatic activity, and geochemical compositions between the lunar nearside and farside (1–3). However, the origin of this lunar hemispheric asymmetry remains highly debated before sample return from lunar farside. Several hypotheses have been proposed to explain this phenomenon, such as asymmetric crystallization of the lunar magma ocean (LMO) (4), early inhomogeneous tidal heating (5), and large impact events concentrated on either the nearside (6) or farside (7, 8). Recent studies demonstrate that CE6 basalts from the lunar farside within the South-Pole Aitken (SPA) basin (Fig. 1A) exhibit petrogenetic, geochemical, and Sr-Nd-Pb isotopic similarities to nearside basalts (9, 10), supporting a common LMO evolutionary model (11). From this standpoint, it suggests that the nearside-farside dichotomy may primarily relate to external geological processes. Numerical simulations indicate that the SPA-forming impact not only excavated deep crustal and potentially mantle materials (8, 12) but also generated thermal energy that likely triggered mantle convection. This process could have transported potassium, rare-earth elements, and phosphorus (KREEP)-rich materials from the farside to the nearside, contributing to the Procellarum KREEP terrane formation and the extensive formation of mare basalts (7, 8). Recently, the discovery of noritic clasts that were thought to be crystallized from impact melt sheets derived from lunar lower crust and upper mantle melting (13) confirms the SPA impact’s crustal modification effects. Nevertheless, geochemical fingerprints left by such a giant impact on the deep mantle have never been reported. If such impact-induced mantle heating and convection occurred, it should likely cause elemental evaporation and associated isotope fractionation. Characterizing the lunar mantle beneath the SPA basin could provide critical constraints for testing and refining lunar asymmetry formation models.

Fig. 1.

Lunar topographic map and the cross-sectional micro-CT images of four CE6 basalt samples. (A) The base map is a topographic product derived from Lunar Orbiter Laser Altimeter (LOLA) data (14). Locations of the 11 lunar sample-return missions are marked on the nearside and farside. (B) These lithic fragments were selected from lunar soil sample CE6C0200YJFM001. Corresponding stereomicroscopic images are provided in SI Appendix, Fig. S1. At current resolution, samples exhibit diagnostic basaltic textures with mineral assemblages dominated by plagioclase (Pl), clinopyroxene (Cpx), and ilmenite (Ilm).

Since radiogenic isotope ratios of Sr, Nd, and Pb are fundamentally controlled by parent–daughter element ratios in source regions, even subtle heterogeneities in the LMO evolution could finally result in measurable variations in these isotopic ratios. Thus, they are difficult to use for inferring mantle source characteristics. In contrast, high-precision analyses of stable isotopes of moderately volatile elements (MVEs) can overcome these limitations and serve as effective tracers for identifying asteroidal impact effects, as demonstrated in K, Zn, Cu, and Rb isotope systems (15–19). Recent investigations of sulfur (S) isotopic compositions in CE6 mare basalts indicate similarities between the far and near sides of the Moon (20), leading the authors to infer a lack of significant S loss and isotopic fractionation during the formation of SPA basin. This conclusion, however, requires careful consideration given the relatively large analytical precision (±0.5‰; 2SD) of the in-situ laser ablation multiple collector–inductively coupled plasma–mass spectrometry (MC-ICP-MS) method. In this study, we focused on the K isotope systems, which has a 50% condensation temperature of 1,006 K at 10⁻⁴ bar (21) and is one of the most abundant MVEs, with an abundance of hundreds to thousands of ppm, whereas other MVEs have abundances of only tens of ppm. This allows us to obtain high-precision isotopic ratios (better than 0.05‰) using a solution analytical method with low sample consumption. Additionally, Fe isotopes (a moderately refractory element with 50% T_c = 1,334 K; ref. 21) are also included, since the combination of two isotope systems can provide more constraints on the contributions from meteorite impactors, the nature of the lunar mantle, or kinetic processes.

Here, we present the first Fe and K isotopic measurements of four lunar basalts returned by the CE6 mission (Fig. 1B and SI Appendix, Fig. S1). The rock type of the basalt clasts was further confirmed through nondestructive X-ray micro–computed tomography (micro-CT) analysis (Materials and Methods). Our data, in combination with those from lunar nearside, are used to explore the significance of the ancient SPA basin–forming event for the Moon’s mantle and MVE depletions, as well as the formation of lunar asymmetry.

Results

Chemical Compositions of the CE6 Basalts.

The concentrations of major and trace elements for the CE6 basalts are reported in SI Appendix, Tables S1 and S2, respectively. Based on ilmenite abundance inferred from micro-CT cross-section images (Fig. 1B) and the measured TiO₂ contents (3.1 to 4.9 wt.%; SI Appendix, Table S1), these four basalts are classified as low-Ti basalts, representing the predominant volcanic type at the CE6 landing site (10, 11, 22). The MgO, FeO, CaO, and Al₂O₃ contents of these basalts align with previously published data but show a wide compositional range (SI Appendix, Fig. S2). The concentrations and patterns of rare earth elements (REEs) are similar to previously reported data for CE6 basalts (SI Appendix, Fig. S3).

Heavy Fe and K Isotopic Compositions of CE6 Basalts.

The Fe and K isotopic compositions, determined using a Sapphire collision-cell multicollector–inductively coupled plasma mass spectrometry (CC-MC-ICP-MS), are reported in SI Appendix, Tables S3 and S4, and shown in Fig. 2. The CE6 basalts exhibit δ⁵⁶Fe values ranging from 0.13 ± 0.03‰ to 0.21 ± 0.02‰ (2SD), which are slightly higher than those of CE5 low-Ti basalts (23) and nearly all Apollo low-Ti basalts (24–28) (Fig. 2A). Specifically, the four basalts present an average value of 0.16 ± 0.04‰ (2SE), which is ~0.08‰ higher than that of the Apollo low-Ti basalts (Fig. 2A). Similarly, compared to Apollo basalts, these CE6 samples also consistently exhibit high δ⁴¹K values ranging from 0.001 ± 0.028‰ to 0.093 ± 0.014‰, with an average value of 0.038 ± 0.044‰ (2SE) (Fig. 2B). This average δ⁴¹K value is approximately 0.16‰ higher than that of Apollo lunar basalts (−0.13 ± 0.06‰, 2SE; ref. 29), which are taken as the best estimate for the lunar mantle and Bulk Silicate Moon. By comparison, the K isotopic offset is roughly twice the magnitude of the δ⁵⁶Fe difference between CE6 basalts and nearside Apollo low-Ti basalts (Fig. 2B).

Fig. 2.

Iron and potassium isotopic compositions of CE6 basalts compared with other lunar samples. (A) δ⁵⁶Fe values of lunar samples. Literature data for Apollo low-Ti and high-Ti basalts are from refs. 24–28. The CE5 basalts are from ref. 23. The KREEP-rich rocks are from ref. 30. The horizontal dashed line and gray band represent the Bulk Silicate Moon (δ⁵⁶Fe = 0.05 ± 0.02‰; ref. 28). (B) δ⁴¹K values across diverse lunar rocks. Apollo mission data are from refs. 19, 29, 31. Unbrecciated lunar basaltic meteorites are also shown for comparison (29, 32), as brecciated ones may have been altered by impacts or contain exotic materials. The gray band represents the Bulk Silicate Earth (BSE; ref. 33). Notably, Apollo low-Ti basalt sample 12,005 exhibits heavier K isotopes but a submantle δ⁶⁶Zn value (−0.2‰ vs. 1.4 ± 0.2‰ for lunar mantle; refs. 29, 34). The decoupled Zn-K isotopes for this sample was attributed to volatile degassing and condensation processes at the Moon’s surface (29), which was excluded for the following discussion. Shaded bands and dashed lines for each sample group indicate mean values and 2SE uncertainties, respectively. The CE6 data from this study are provided in SI Appendix, Tables S3 and S4.

Combining our CE6 data with the Apollo mare basalts, there is no clear relationship between δ⁵⁶Fe and either Mg#, TiO₂, or Th concentrations (Fig. 3 A–C). By contrast, there appears to be a negative correlation between δ⁴¹K and Mg# (Fig. 3D), while no relationships are apparent between δ⁴¹K and K or Th concentrations (Fig. 3 E and F).

Fig. 3.

Relationships between Fe–K isotopes and elemental concentrations. (A) δ⁵⁶Fe vs. Mg#. (B) δ⁵⁶Fe vs. TiO₂ content. (C) δ⁵⁶Fe vs. Th concentration. (D) δ⁴¹K vs. Mg#. (E) δ⁴¹K vs. K content. (F) δ⁴¹K vs. Th. For Fe isotopes, the Apollo low-Ti and high-Ti data from refs. 26, 28, CE5 basalts from ref. 23, and KREEP-rich rocks from ref. 30 are shown. The gray band represents the lunar mantle (28). For K isotopes, source data are identical to those in Fig. 2B. The gray band indicates the lunar mantle (δ⁴¹K = −0.13 ± 0.06‰, 2SE), estimated from Apollo mare basalts (19, 29, 31).

Discussion

Evaluation of Contamination and Cosmogenic Effects on Fe–K Isotopes.

Lunar fine-grained soil particles are generally formed by space weathering and thus can acquire heavy Fe and K isotopic signatures due to the preferential loss of light isotopes to space (35–37). To minimize potential contamination of adhering micron-sized soil particles on the studied CE6 basalt clasts, their surfaces were ultrasonically cleaned with anhydrous ethanol. In addition, based on 3D micro-CT imaging, no impact breccias or melt glasses were observed inside the basalt clasts (Fig. 1B).

Cosmic rays may alter the relative abundances of different Fe and K isotopes through neutron capture and/or spallation reactions (38, 39), potentially creating detectable isotopic variations in lunar samples. However, the Fe isotopic data for all CE6 mare basalts fall on the mass-dependent fractionation line based on the three-isotope (δ⁵⁷Fe vs. δ⁵⁶Fe) plot (SI Appendix, Fig. S4). This result is consistent with previous works on Apollo samples (25, 35). For K isotopes, ⁴¹Ca can be generated via the low-energy neutron capture reaction: ⁴⁰Ca(n, γ)⁴¹Ca and then ⁴¹Ca can rapidly decay to ⁴¹K (39). Based on the production rate of ⁴¹Ca (39), our calculation indicates that this effect would increase the δ⁴¹K by less than 0.04‰ even at a CRE age of 1,000 Myr (SI Appendix, Fig. S5), within current analytical precision. Although the CRE age of CE6 basalts is unknown, this value represents a near upper limit for cosmic-ray effects, as most lunar samples have CRE ages ≤1,000 Myr (40). Collectively, we conclude that cosmic-ray effects are not a controlling factor for the Fe and K isotopic compositions of CE6 basalts.

Heavy Fe and K Isotopic Fingerprints of CE6 Lunar Mantle.

Magmatic processes including degassing of volatiles (41), partial melting (25, 42), fractional crystallization (28, 43), and oxygen fugacity variations (44) can induce isotope fractionation. It has been suggested that some volatiles such as Cl (45) and S (41) may degas during lava flow eruptions into vacuum. However, K isotope measurements of plutonic rocks that were formed within the lunar crust and not subjected to degassing into vacuum (e.g., unbrecciated troctolite 76535: −0.17 ± 0.08‰; norite Arguin 002: −0.21 ± 0.06‰) have δ⁴¹K values similar to the mare basalts (29). This comparison indicates that locally degassing processes do not affect the original K isotopic composition of the lunar mare basalts.

As lunar basalts are generated by partial melting of mantle cumulates, we explored whether Fe and K isotope fractionation during partial melting and crystallization could explain the heavier isotopic compositions of CE6 samples. We first quantitatively modeled Fe and K isotopic compositions of different stages of LMO differentiation (SI Appendix; Fig. 4 A and D). This model uses force constants determined by Nuclear Resonant Inelastic X-ray Scattering (NIRXS) technique and ab initio calculations (parameters detailed in SI Appendix). Given uncertainties in force constants, we also calculated mantle cumulate Fe isotopes using minimum force constants (SI Appendix, Fig. S6). For K isotopes, the average δ⁴¹K value of Apollo basalts was used as the initial LMO composition, because a recent study revealed remarkably consistent K isotope compositions across all lunar basalt types, regardless of returned samples or lunar meteorites (29). Since the cumulate minerals such as olivine, orthopyroxene, clinopyroxene, and ilmenite contain little K (46), their crystallization does not induce K isotope fractionation. When plagioclase begins to crystallize from the magma ocean, it leads to a slight decrease of δ⁴¹K in residual melts (Fig. 4D).

Fig. 4.

Calculated evolution of δ⁵⁶Fe and δ⁴¹K values during LMO differentiation and subsequent magmatic processes. (A) The Fe isotopic evolution during LMO differentiation. The light blue and light red curves illustrate the Fe isotope evolutions of residual melts and cumulates, respectively (see SI Appendix for details). Shaded areas represent the 2SD uncertainties. Noted that additional modeling results (named “Model 2”), which includes the uncertainties of Fe force constants, are shown in SI Appendix, Fig. S6. (B) Variation of δ⁵⁶Fe values with increasing mantle melting degree. The source chemical compositions, source mineral assemblages, and lunar mantle partial melts at different melting degrees are derived from ref. 47. The Fe isotopic composition of mantle cumulates (80% Ol-Opx cumulate with 20% Ilm-Cpx cumulate) was calculated based on LMO differentiation results in panel (A). The total Fe isotope fractionation factors were calculated based on mineral proportions and their individual factors. Details of the batch melting model are provided in SI Appendix. The results from two models are shown. (C) δ⁵⁶Fe vs. MgO content. Light green and red curves show modeled fractional crystallization results based on ref. 47. (D) The K isotopic evolution during LMO differentiation. Silicate minerals (olivine, orthopyroxene, clinopyroxene, and ilmenite) have negligible effects on K isotopes (46, 48), while the plagioclase-melt isotopic fractionation factor (Δ⁴¹K = 0.2‰) is based on the ab initio calculation (48). The shaded areas show measured K isotopic compositions of Apollo basalts (29). (E) δ⁴¹K vs. MgO. (F) δ⁴¹K vs. TiO₂. The primary melt δ⁴¹K is based on the LMO model in panel (D). Black and red curves show modeled fractional crystallization results.

Recently, Yin et al. (47) presented a detailed petrogenetic study and quantitatively constrained the mantle source compositions, degree of partial melting and fractional crystallization processes using major- and trace-element data of CE6 lunar basalts. They suggested that the CE6 mantle source was composed of 80 to 90% early cumulate and 10 to 20% the late-stage cumulate (clinopyroxene + ilmenite). Using this result and our model, we constrained Fe and K isotopic compositions of the mantle source and then modeled the partial melting process. For Fe isotopes, the result shows that 3 to 7% partial melting could produce melts with δ⁵⁶Fe as high as 0.11‰ when we used the minimum Fe force constants (Fig. 4B). However, although this result is higher than nearly all Apollo low-Ti basalts, it is still lower than the δ⁵⁶Fe values of CE6 basalts (0.13 to 0.21‰). For K isotopes, owing to its incompatibility in the residual mantle mineralogy, K isotopic composition of the melt is expected to be unfractionated with respect to its mantle source (Fig. 4E). Thus, partial melting of mantle alone cannot account for the observed Fe and K isotope enrichments. The involvement of KREEP material appears to be unlikely, either given that the samples do not have high Th content (Fig. 3 C and F) and the most depleted Sr–Nd–Pb isotopic characteristics for CE6 basalts (9, 22).

Since the mare basalts have experienced some degree of fractional crystallization before eruption (49), this factor should also be considered. We calculated the effect of fractional crystallization, and the results show that the δ⁵⁶Fe value after 40 to 50% fractional crystallization (olivine, clinopyroxene, and pigeonite; refs. 9, 47) reaches ~0.14‰ (Fig. 4C), where the K isotopes remained almost unchanged during the fractional crystallization process (Fig. 4 E and F). This process can match the Fe isotopic compositions of three CE6 samples, with only one sample exhibiting slightly heavier Fe isotope composition. However, given the uncertainties in the Fe isotope fractionation factors determined by the ab initio methods and spectrometric techniques (50), the isotopic signature of this sample may also be explained by magmatic processes. Nevertheless, we cannot entirely rule out the possibility that its heavy Fe isotope composition is related to the impact event (see discussion below). Future studies with more Fe isotopic data are required to test this hypothesis. By contrast, the results show that fractional crystallization plays a negligible role in controlling the K isotopic variations. Thus, another process is required to explain the heavy K isotopic compositions of the samples.

Finally, we considered the mechanism of oxygen fugacity (fO₂) since fO₂ can affect the valence state and speciation of Fe, which can influence its behavior during geological processes. Analyses of CE5 samples reveal persistently reduced fO₂ conditions on the lunar nearside from 3.6 to 2.0 Ga (51). Recent spinel and pyroxene oxybarometer on CE6 basalts further reveals a more reduced fO₂ on the farside at ~2.8 Ga (52), thereby reducing the role of fO₂ on the Fe isotope fractionation. In summary, these lines of evidence presented above suggest that magmatic processes can largely account for the Fe isotopic variations in the basalts. However, none of these processes can explain the heavy K isotopic compositions observed. This indicates that the mantle source of the basalts must have initially possessed a heavy K isotopic signature.

SPA Impact–Induced Volatile Loss and Isotope Fractionation.

Previous studies have proposed asymmetric crystallization of the LMO to account for the observed high Mg-number values on the lunar farside (4). In their model, plagioclase may crystallize relatively earlier on the farside than on the nearside (53). However, this difference in crystallization timing has limited influence on the Fe and K isotopic compositions of the lunar mantle since the mantle had already formed prior to plagioclase crystallization. The recent work of Dai et al. (29) demonstrates that the K isotope compositions of the lunar farside and nearside are broadly similar. This observation argues against a simple nearside-farside dichotomy as the origin for the features observed in our samples. Instead, given that the CE6 basalts are located within the Moon’s largest and deepest impact basin (54, 55), the heavy K isotopic signature strongly suggests that the processes linked to this giant impact event are the primary control. Numerical modeling demonstrates that the SPA impact excavated material from depths of 80 to 120 km (spanning the lower crust to upper mantle) (12, 56). This impact also induced mantle melting, with melting degree decreasing with increasing depths (12, 57). In addition, since extensive mare basalt volcanism could happen before 2.8 Ga at the CE6 landing site, the mantle source of CE6 basalt could have experienced multiple melt extractions. However, regardless of the impact-induced mantle melting or subsequent melt extraction, these processes would leave the residual solid mantle with lighter or unchanged isotopes (Fig. 4B) instead of the heavy isotopic signatures observed in CE6 basalts.

On the other hand, if the impactor itself were added into the lunar mantle, trace elemental ratios, and isotopic compositions are useful indicators to evaluate their contributions. The Fe (58–60) and K (32, 61–64) isotopic compositions for Vesta, chondrites (OC, EC, and CC), and iron meteorites have been reported (SI Appendix, Fig. S7). We find that if the impactor is a chondrite, regardless of OC, EC, or CC, it would lead to a high Ni/Co ratio (Fig. 5A) and lighter Fe–K isotopes (Fig. 5B), inconsistent with our observations. While the addition of iron meteoritic material could slightly elevate the δ⁵⁶Fe and Ni concentrations of the mantle, it would not alter its δ⁴¹K composition (Fig. 5B), due to the extremely low K abundance in iron meteorites (<1 ppm vs. ~46 ppm for the lunar mantle; refs. 65, 66). In contrast, if the impactor has Vesta-like chemical compositions, contribution from this planetary body could increase δ⁴¹K values but slightly decrease Fe isotopes (Fig. 5B). A rough mixing calculation between Vesta (K = 400 ppm; δ⁴¹K = 0.36‰; ref. 32) and lunar mantle (K = 46 ppm; δ⁴¹K = −0.13‰; refs. 29, 66) indicates at least 7% of impactor-derived residue is required to account for the δ⁴¹K values of CE6 basalts, while this contribution only slightly decreases the δ⁵⁶Fe value by 0.01‰.

Fig. 5.

Indices to identify the influence from impactor. (A) Ni/Co ratios vs. Ni concentrations. Literature data for Apollo low-Ti and high-Ti basalts, lunar basaltic meteorites, KREEP-rich rocks, and brecciated meteorites are from ref. 29. The dashed line (Ni/Co = 3) represents the upper limit of the primitive lunar mantle (67). Gray curves represent binary mixing between the mantle source of basalts and chondrite-like component that has high Ni and Co concentrations (68). (B) δ⁵⁶Fe vs. δ⁴¹K for CE6 basalts. The gray area depicts average values for Apollo mare basalts, with data sources as follows: Fe isotopes from refs. 24–28; K isotopes from refs. 19, 29, 31.

Instead, the most plausible mechanism is evaporative loss of K induced by the SPA impact. Numerical simulations suggest this event injected substantial thermal energy into the lunar interior, triggering hemispheric-scale mantle convection that transported KREEP-rich material from the farside to the nearside, contributing to the PKT formation and hemispheric asymmetry (7, 8). This impact event could elevate mantle temperature to 2,800 K (12), which is sufficient for K vaporization. For volatilization of a liquid into a vacuum, the vapor phase is enriched in the lighter isotopes, leaving the residual fraction progressively enriched in heavier isotopes. Under lunar vacuum conditions, minor K loss can induce remarkable isotope fractionation, which can be modeled using Rayleigh distillation equation. Our result shows that 2% loss of K could successfully cover the heavy K isotopes in CE6 basalts (Fig. 6). If the heavy Fe isotopes were also caused by the SPA impact, our modeling demonstrates that less than 2% Fe volatilization would be sufficient to account for the Fe isotopic composition of CE6 sample (SI Appendix, Fig. S8). This new result also implies that the SPA-forming impact was more energetic than other large basins, such as the Procellarum basin (69). Therefore, compared to the S isotopes (20), our high-precision K isotopic data provide robust evidence that the SPA impact heavily affected the lunar mantle, likely facilitating the formation of lunar dichotomy.

Fig. 6.

Modeling results of impact-induced elemental loss. Potassium isotope fractionation during evaporation follows Rayleigh distillation equation: δ⁴¹K_final = (δ⁴¹K_initial + 1,000) × F^(α−1) – 1,000. Simulated trends for three vapor-melt fractionation factors (α) are shown. The α = 0.975 corresponds to the evaporation of K in a vacuum environment. Gray bands denote the observed δ⁴¹K ranges for Apollo low-Ti basalts (Lower) and CE6 basalts (Upper). Data source for Apollo mare basalts is from ref. 29.

Materials and Methods

Sample Preparation.

The CE6 samples analyzed in this study were obtained from two lunar soil samples (CE6C0100YJFM002 and CE6C0200YJFM001), collected from shallow depths (≤3 cm) on the lunar surface (54). We first sieved the samples using a diameter of 0.9 mm sieve to isolate large clasts. This procedure was conducted in a Class-1000 clean laboratory at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). Basalt clasts were manually picked under a binocular microscope. Then, these selected clasts were ultrasonically cleaned in ethanol to remove surface-adhered fine particles. Given the small size of individual clasts, multiple fragments were combined as one sample. We finally obtained four basalt samples (designated as CE6-B1, ~19.4 mg; CE6-B2, ~15.4 mg; CE6-B4, ~14.4 mg; CE6-B5, ~15.6 mg) and the stereomicroscopic photographs are shown in SI Appendix, Fig. S1. Their lithologies were further confirmed by micro-CT scanning technique.

X-ray Micro–Computed Tomography Analysis.

To further confirm the rock type, the selected large clasts were analyzed using an FEI HeliScan micro-CT system placed at IGGCAS. The spatial resolution of the measurements was approximately 1.5 μm. Scanning parameters were set to an accelerating voltage of 80 kV and a beam current of 100 μA. Each frame had a 3.3-s exposure time, and 8 frames were averaged per two-dimensional radiograph. The total scan time per clast was approximately 2 h on average. A 2-mm aluminum X-ray beam filter was used to reduce beam hardening artifacts. This nondestructive method maximized sample utility while preserving their physical integrity. Representative image sections for each sample are displayed in Fig. 1B.

Bulk-Rock Fe Isotopic Analysis.

The samples were dissolved in sealed Savillex screw-top beakers through sequential treatment with concentrated HF-HNO₃, HCl-HNO₃, and HCl. After complete digestion, this digested solution was treated as the mother solution for major-element, trace-element, and isotopic analyses.

An aliquot was taken from the mother solution for Fe isotopic analysis. The aliquots were dried down and redissolved in 8 M HCl (+0.001% H₂O₂) for chemical purification. Iron was purified using a two-step ion exchange chromatography. For the first purification, the solution was loaded onto preconditioned 2 mL Bio-Rad AG-MP-1 M resin. Matrix elements were eluted with 37 mL of 8 M HCl. The target Fe fraction collected with 18 mL of 2 M HCl, was evaporated to dryness, and then redissolved in 1 mL of 8 M HCl (+0.001% H₂O₂) and loaded onto preconditioned 1 mL Bio-Rad AG-1-X8 resin for the second purification. Matrix elements were eluted with 9 mL of 8 M HCl, and the Fe fraction was collected with the next 10 mL of 0.4 M HCl. The Fe fraction was evaporated to dryness and diluted to a concentration of 30 ng/g in 2% HNO₃ for Fe isotopic measurements. The total procedure blank for Fe isotope analyses is <6 ng.

Iron isotopic measurements were performed on a Nu Sapphire MC-ICP-MS at the collision cell pathway in low-resolution mode using the sample-standard bracketing method (70). Data were collected in static mode, with ⁵⁴Fe and ⁵⁶Fe connected to preamplifiers fitted with 10¹¹ Ω resistors. Each analysis consisted of a block with 50 cycles of 3 s integration. A 50 s wash was performed in 2% HNO₃ between each standard and sample to avoid cross contamination. Due to significant interference from ArOH⁺ on ⁵⁷Fe and the low efficiency of the collision cell in suppressing ArOH⁺, the background of ⁵⁷Fe is approximately 2.7% (70), which may lead to inaccurate determination of δ⁵⁷Fe. Thus, only high-precision δ⁵⁶Fe values are reported. The instrumental mass bias was corrected using standard-sample bracketing against the NWU-Fe standard. The Fe isotope data are reported in standard δ-notation: δ⁵⁶Fe (‰) = [(⁵⁶Fe/⁵⁴Fe)_sample/(⁵⁶Fe/⁵⁴Fe)_IRMM-014 − 1] × 1,000. Long-term reproducibility is 0.03‰ (2SD) for δ⁵⁶Fe, based on analyses of USGS basalt standard BCR-2. Three USGS standards were measured, which yielded the following results: δ⁵⁶Fe = 0.11 ± 0.02‰ for BHVO-2 (basalt), δ⁵⁶Fe = 0.10 ± 0.02‰ for BCR-2 (basalt), and δ⁵⁶Fe = 0.15 ± 0.04‰ for AGV-2 (andesite), which are consistent with previously reported values (SI Appendix, Fig. S9). The one duplicate sample shows good reproducibility within analytical uncertainty (SI Appendix, Table S3).

Bulk-Rock K Isotopic Analysis.

An aliquot of CE6 sample was taken from the mother solution as mentioned above and the chemical procedures were adopted from refs. 71 and 72. The sample solution was loaded onto preconditioned 2 mL Bio-Rad AG50W-X8 (200 to 400 mesh) resin and then eluted with 15 mL of 0.5 mol/L HNO₃ to remove matrix elements. The K fraction, containing ~100% of the total K, was collected using 20 mL of 0.5 mol/L HNO₃ and then dried down. The purification process was repeated 2 to 4 times to ensure complete removal of matrix elements. The final K solution was redissolved in 2% HNO₃ for measurement. The total procedure blank for K isotope analysis was <3 ng K, which is negligible compared to the ~6 μg of K in the purified sample solutions. In addition to the three USGS standards (BHVO-2, BCR-2, and AGV-2), we synthesized a solution using NIST SRM3141a mixed with other elements to match the major element composition of CE6 basalts. This strategy was further used to evaluate the accuracy of CE6 sample analysis.

The signal intensities of ³⁹K and ⁴¹K were also measured using the Nu Sapphire CC-MC-ICP-MS instrument via the low-energy collision cell path (71). The hexapole collision cell utilizes He and H₂ gas to effectively remove various Ar-based polyatomic species, thus allowing K isotopic ratios to be measured in low-resolution mode. Sample introduction was performed using an autosampler SC-2DX connected to an Apex Omega desolvation nebulizer system. One Faraday cup was connected to a preamplifier fitted with a 10¹⁰ Ω resistor for collection of the ³⁹K⁺ ion beam, while the other two Faraday cups fitted with 10¹¹ Ω resistors collected ⁴¹K⁺ and mass 40 beams, respectively. The standard-sample-standard method was adopted to correct for instrumental mass fractionation. The K concentration of each sample, standard, and the synthesized solutions was matched within 10%. Each analysis consisted of one block of 50 cycles with 4-s integrations. Five repeated analyses were conducted on each of our sample solutions. The K isotopic data are reported in the delta notation relative to NIST SRM3141a. The long-term (>2.5 y) precision, based on multiple measurements of standard BCR-2, was ~0.04‰ (2SD) for δ⁴¹K (73). The results of reference materials BHVO-2, BCR-2, and AGV-2 as well as the synthesized solution were reported in SI Appendix, Table S4 and shown in SI Appendix, Fig. S10. The K isotopic compositions of the standards are consistent with previously reported values from other labs (SI Appendix, Fig. S10), ensuring the accuracy of our experimental data.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

The data are available in Figshare (74). All other data are included in the article and/or SI Appendix.