Ultra-depleted mantle source of basalts from the South Pole–Aitken basin

Sample preparation

The studied CE6 samples (CE6C0100YJFM001, about 5,000 mg, and CE6C0100YJFM002, about 2,000 mg) were allocated by the China National Space Administration. Both samples were scooped from the lunar surface. A total of 16 basalt fragments were picked out from the soil samples for detailed petrological and geochemical analysis (Supplementary Table 1). Four large fragments (CE6C0000YJYX25101, CE6C0000YJYX48501, CE6C0000YJYX48901 and CE6C0000YJYX56201) had sufficient mass (more than 30 mg) for whole-rock major, trace and Sr–Nd isotope analyses to be performed. Thus, each of the four fragments was cut into two parts, one for scanning electron microscope (SEM) analysis and electron probe microanalysis (EPMA), and the other for whole-rock major, trace and Sr–Nd isotope analysis. The remaining 12 smaller fragments were only examined by SEM and EPMA. Before the SEM analysis and EPMA, the samples were embedded in 1-in. epoxy mounts and polished.

SEM analysis and energy dispersive spectrometer mapping

The petrography was carried out on a Zeiss Supra 55 field-emission SEM at the Key Laboratory of Lunar and Deep Space Exploration, National Astronomical Observatories, Chinese Academy of Sciences and a Zeiss Gemini 450 field-emission SEM at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS) in Beijing, China. The accelerating voltage was 15.0 kV and the probe current was 2.0 nA. In addition, a Thermo Scientific Apreo SEM equipped with an energy dispersive spectrometer was used at IGGCAS to obtain the modal abundance of each mineral and calculate the bulk major element compositions based on the elemental mapping. The results are listed in Supplementary Table 2.

Electron microprobe analysis of minerals

The major element concentrations of pyroxene, plagioclase, olivine, ilmenite, spinel, quartz, sulfide and phosphates in each sample were analysed using a JEOL JXA8230 electron probe at the National Astronomical Observatories, Chinese Academy of Sciences, and a JEOL JXA8100 electron probe at the IGGCAS. The conditions of the EPMA were as follows: accelerating voltage of 15 kV, probe current of 20 nA, focused beam and peak counting time of 10 s. Calibration of the elemental data was done using a series of natural minerals and synthetic materials. The analytical crystals and calibration standards were as follows: Na (thallium acid phthalate, natural albite), Mg (thallium acid phthalate, natural diopside), Al (thallium acid phthalate, synthetic Al₂O₃), Si (thallium acid phthalate, natural diopside), Cr (lithium fluoride, synthetic Cr₂O₃), Mn (lithium fluoride, natural bustamite), Fe (lithium fluoride, haematite), Ni (lithium fluoride, synthetic NiO), K (pentaerythritol, natural K-feldspar), Ca (pentaerythritol, natural diopside) and Ti (pentaerythritol, synthetic rutile). Based on an analysis of the internal laboratory standards, the precision for the major (more than 1.0 wt%) and minor (0.1–1.0 wt%) elements were better than 1.5 and 5.0%, respectively. The analytical data from the samples and standards are listed in Supplementary Table 3.

Whole-rock major and trace elements

Aliquots of 30 mg of each of the four samples analysed for bulk chemistry were mixed thoroughly with ultrapure lithium borate (3.0 g) in a Pt–Au crucible at a ratio of 1:100. The sample was then melted at 1,050 °C using an M4 propane gas automatic fluxer before being cast into a 27-mm-disk-shaped glass sample. The prepared disc-shaped glass sample was measured using panalytical wavelength-dispersive X-ray fluorescence spectrometry. The X-ray fluorescence spectra were calibrated after measuring the intensities of 44 international reference materials. The criteria for selecting these samples were based on the required concentration intervals. The instrument conditions were consistent with those reported in ref. ⁵⁰.

The trace elements were subsequently analysed using laser ablation inductively coupled plasma mass spectrometry on lithium borate glass discs, employing an Agilent 8900 ICP-MS instrument coupled with a high-repetition-rate Genesis GEO Femtolaser Ablation System⁵¹. Ablation was performed using spots with a diameter of 100 µm and a length of 1,000 µm, at a frequency of 1 Hz for 45 s, following a 25-s measurement of the gas blank. A 25-s washout between analyses was used. The gas flows were optimized by spot ablation of the National Institute of Standards and Technology Standard Reference Material (NIST SRM) 612 glass standard reference material to obtain maximum signal intensities while maintaining the ThO/Th ratio below 0.3% and the U/Th ratio at 0.95–1.05. During the test, Al was used as the internal standard, while NIST SRM 612 served as the external standard for sample measurement.

For the major elements, the deviations between the analytical results and the reference values range from 0.5 to 1.5%, depending on the mass fractions of the elements, while the relative standard deviation was maintained within 2%. For trace elements, the measurement bias between the analytical results and the reference values was within 10%, and the relative standard deviation was maintained within 10%. The analytical data of the samples and standards are listed in Supplementary Table 4.

Whole-rock Rb–Sr and Sm–Nd isotopes

All chemical procedures, including sample dissolution and the chromatographic separations, were conducted on and in International Organization for Standardization (ISO) class 5 clean benches or hoods in an ISO class 6 ultra-clean laboratory. Approximately 3–5 mg of CE6 basalt fragments, along with appropriate amounts of ⁸⁷Rb–⁸⁴Sr and ¹⁴⁹Sm–¹⁵⁰Nd spikes, were weighed into 2-ml Savillex perfluoro alkoxy polymer beakers. The samples were dissolved in tightly capped perfluoro alkoxy polymer vials using 0.5 ml of HF and 0.1 ml of HNO₃ at 150 °C on a hotplate for 1 day, with intermittent sonication for 1 h to enhance the dissolution. The solutions were then evaporated to dryness and redissolved in 0.2 ml of 1.5 M HCl and 0.1 M HF to obtain a clear solution with no visible residue.

The sample solutions were first loaded into pre-cleaned homemade columns packed with approximately 0.25 ml of AG 50W-X12 200–400-mesh resin to separate the matrix elements Rb, Sr and the REEs. The columns were pre-cleaned using three washes of 3 ml of 6 M HCl, followed by 1 ml of Milli-Q water, and then were equilibrated with 1 ml of 1.5 M HCl and 0.1 M HF. After loading the sample, the major matrix elements and the trace elements, such as U, Pb and Hf, were eluted with four washes of 0.25 ml of 1.5 M HCl and 0.1 M HF. Additional major matrix elements (Fe, Mg and K) were washed out with 1 ml of 1.5 M HCl. After that, Rb was stripped using 1.5 ml of 1.5 M HCl. Subsequently, Sr, Ca, Ba and the REEs were recovered using 5 ml of 6 M HCl.

In the second step, Bio-Rad Bio-Spin columns packed with 0.5 ml of Sr spec resin were used to separate the Sr and REEs. The resin columns were pre-washed using three 2-ml washes of Milli-Q water and conditioned with 2 ml of 3 M HNO₃. The Sr–REE solutions from the first column were evaporated to dryness, redissolved in 0.2 ml of 3 M HNO₃ and then loaded into the columns. The REEs were collected with the sample load and further recovered by washing with 0.5 ml of 3 M HNO₃ three times. After three rounds of washing with 2 ml of 7 M HNO₃, the Sr was recovered using 2 ml of Milli-Q water.

Next, Eichrom polypropylene columns packed with 1.7 ml of homemade 2-ethylhexyl phosphonic acid mono-2-ethylhexyl ester (HEHEHP) extraction resin (similar to Eichrom LN2 resin) were used for the Sm–Nd separation. The resin columns were pre-cleaned using 4 ml of 6 M HCl twice, followed by 4 ml of Milli-Q water, and then conditioned using 4 ml of 0.1 M HCl. The REE fractions from the Sr–resin column were dried down, redissolved in 0.15 ml of 0.1 M HCl, and loaded into the HEHEHP columns. The columns were then washed using 0.25 ml of 0.1 M HCl four times. The Ce and Pr were further removed using 4.8 ml of 0.1 M HCl. Subsequently, Nd was recovered using 2 ml of 0.2 M HCl. Finally, Sm was stripped using 2.5 ml of 0.4 M HCl.

The Rb–Sr and Sm–Nd isotopic analyses were conducted using a Thermo Scientific TRITON Plus thermal ionization mass spectrometer. The Sr isotope ratios were measured using W filaments with TaF₅ as the ion emitter⁵², and the Nd isotope ratios were measured as NdO⁺ also using W filaments with TaF₅ as the ion emitter⁵³. During the analytical sessions, the results were 0.710245 ± 0.000020 (2σ, n = 5) for National Bureau of Standards (NBS) 987 Sr and 0.512102 ± 0.000010 (2σ, n = 5) for JNdi-1 Nd, which are consistent with previously reported values (0.710248 ± 0.000011 (2σ) for NBS 987 Sr and 0.512115 ± 0.000007 (2σ) for JNdi-1 Nd)^51,54. The procedural blanks were less than 3 pg for Rb, less than 100 pg for Sr, less than 10 pg for Sm and less than 20 pg for Nd, which were negligible compared to the amounts of Sr and Nd in the analysed samples.

The US Geological Survey BCR-2 reference material, in amounts of approximately 3 mg, was analysed alongside the CE6 samples, yielding average values (±2σ, n = 2) of 0.401 ± 0.012 for ⁸⁷Rb/⁸⁶Sr, 0.705017 ± 0.000025 for ⁸⁷Sr/⁸⁶Sr, 0.1384 ± 0.0002 for ¹⁴⁷Sm/¹⁴⁴Nd and 0.512625 ± 0.000020 for ¹⁴³Nd/¹⁴⁴Nd, all of which are consistent with their reference values (0.3990 ± 0.0005 for ⁸⁷Rb/⁸⁶Sr, 0.705013 ± 0.000010 for ⁸⁷Sr/⁸⁶Sr, 0.1380 ± 0.0004 for ¹⁴⁷Sm/¹⁴⁴Nd and 0.512637 ± 0.000012 for ¹⁴³Nd/¹⁴⁴Nd)^55,56. The data are shown in Supplementary Table 5.

Petrography and mineral chemistry

The basalt fragments could be texturally subdivided into three types: porphyritic, subophitic and poikilitic (Extended Data Fig. 1). The porphyritic clasts commonly exhibit coarse-grained (50 × 300 μm) clinopyroxene phenocrysts in a fine-grained (less than 10 μm) matrix (Extended Data Fig. 1a). The matrix is composed of acicular plagioclase (An_76.3–85.2), interstitial clinopyroxene and tiny (less than 5 μm) ilmenite (Extended Data Fig. 1a). Compared with the clinopyroxene phenocrysts, those in the matrix have higher FeO (32.1–38.8 wt%) but lower MgO (4.03–19.0 wt%) and Cr₂O₃ (0.10–1.19 wt%) contents. The ilmenite needles commonly show three directions cutting the matrix plagioclase and pyroxene, representing a late-stage crystallization phase.

The subophitic clasts show various grain sizes (20–300 μm) and consist mainly of plagioclase, clinopyroxene and ilmenite, with minor Fe–Ti-spinel (ulvöspinel), troilite, olivine and cristobalite (Extended Data Fig. 1b,c). Both the clinopyroxene and olivine have compositional zoning, with Mg-rich cores and Fe-rich rims. The plagioclase shows euhedral to subhedral shape with anorthite-rich composition (An_83.6–91.9) (Supplementary Table 3). The single olivine grain with a forsterite core (Extended Data Fig. 1b) shows a large compositional range (Fo_2.7–58.5) (Supplementary Table 3).

The poikilitic clasts are mainly composed of clinopyroxene, plagioclase and ilmenite, with accessory Fe–Ti-spinel (ulvöspinel) and troilite, and a mesostasis including K-feldspar, fayalite, cristobalite, baddeleyite, tranquillityite, zirconolite and phosphates (Extended Data Fig. 1d). Clinopyroxenes of various sizes are included in the coarse-grained (larger than 100 μm) plagioclases. The plagioclase is anorthite-rich (An_81.9–94.3). The clinopyroxene shows a large compositional range (Wo_8.5–38.9En_0.2–54.9Fs_20.8–89.8) and is systematically characterized by Mg-rich cores (Mg# = 27.6–66.1) and Fe-rich rims (Mg# = 0.2–39.1). The euhedral spinel has 1.9–6.5 wt% Cr₂O₃, 61.5–64.0 wt% FeO and 30.2–32.4 wt% TiO₂. Small amounts of Fe-rich olivine (Fo_1.6), associated with cristobalite, baddeleyite, tranquillityite, zirconolite and phosphates, occur as mesostasis phases representing late-stage crystallization products.

Batch melting and fractional crystallization modelling

We used trace elements to model the batch melting and fractionation crystallization processes to reproduce the REE compositions of the CE6 basalt following the same method used in refs. ^23,25. The batch melting was modelled using the following equation: C_L/C₀ = 1/(D₀ + F(1  −D₀)), where C_L represents the weight concentration of a trace element in the melt; C₀ is the weight concentration of the trace element in the original cumulate source; F is the melt fraction; and D₀ is the bulk distribution coefficient of the solid phase.

The bulk distribution coefficient is determined by multiplying each mineral partition coefficient by its modal abundance in the source. Because the CE6 basalts have a source similar to that of the Apollo 12 basalts, we adopt the modal mineralogy calculated for the Apollo 12 samples²⁵. The REE partition coefficients for olivine⁵⁷, orthopyroxene⁵⁸, augite⁵⁸, pigeonite⁵⁹, plagioclase⁶⁰ and garnet⁶¹ are listed in Supplementary Table 6.

Four mantle sources were used for the modelling: (1) 76 PCS cumulate + 0.7% TIRL of the LMO model from ref. ²⁶; (2) 78 PCS cumulate + 1% TIRL of the LMO model from ref. ²⁷; (3) 88 PCS cumulate + 0.3% TIRL of the LMO model from ref. ²⁸; and (4) 78 PCS cumulate + 0.3% TIRL of the LMO model from ref. ²⁹. The earliest LMO cumulates that underwent plagioclase separation in each model were taken as the source of the CE6 basalt. Small amounts (0.3–1.0%) of TIRL were added to reproduce the ¹⁴⁷Sm/¹⁴⁴Nd ratio (0.262–0.272) of the CE6 basalt source. The REE concentrations of the PCS and TIRL are listed in Supplementary Table 7. Using these bulk distribution coefficients (D₀) and the solid cumulate (C₀), the REE concentrations in the melt (C_L) were calculated for increasing melt fractions (F). The results are shown in Extended Data Fig. 6.

The trace element concentrations in the remaining melt, following fractional crystallization, were calculated using the Rayleigh fractionation equation: C_L/C₀ = (1 − F)^D−1, where D is the bulk distribution coefficient (the same as in the batch melting model), F is the mass fraction of crystallized solids, and C₀ and C_L are the element concentrations in the initial and final melt, respectively. Two scenarios are proposed to reproduce the high Sm/Yb ratio of the CE6 basalt: (1) a garnet-bearing mantle source, where the initial melts are assumed to have resulted from 1–1.5% batch melting of the 78 PCS cumulate + 0.6% TIRL of the LMO model from ref. ²⁹, with 0.8% garnet in the residue; and (2) a high Sm/Yb ratio mantle source, where the initial melts are assumed to have resulted from a 0.7–1% batch melting of the 78 PCS cumulate of the LMO model from ref. ²⁷, mixed with 0.8% of a high-Ti component. These results are presented in Fig. 3 and Extended Data Fig. 7.