Main

The 146Sm–142Nd short-lived radiometric pair records the first few hundred million years (Myr) of the Solar System and provides a powerful geochemical tool for tracing the early silicate differentiation of planetary objects. However, its use as a precise chronometer has become more problematic in the past few years, owing to uncertainties in the half-life of 146Sm (ref. 7), in the initial abundance of 146Sm (ref. 8) and in the bulk Sm/Nd ratios of planetary bodies9. In addition, 142Nd/144Nd ratios in chondrites deviate by up to −40 p.p.m. from the terrestrial value2,3,4, and model 146Sm/144Nd–142Nd/144Nd ages for silicate differentiation depend on the initial 142Nd/144Nd composition of the planets from which internal reservoirs subsequently evolved. The deviations of sample compositions from the Nd isotopic standard (and similarly for Sm normalized to 152Sm) for a given isotope of mass i are expressed (in parts per million) as μiNd = [(iNd/144Nd)sample/(iNd/144Nd)reference − 1] × 106.

Isotopic measurements of planetary materials suggest that the solar nebula was not completely homogenized during the period of planetary accretion10. Large isotope anomalies have been detected at the microscale for presolar grains11, and smaller anomalies are found for many refractory elements at the whole-rock scale (for example, Cr, Ti (ref. 12), Sr (ref. 13), Sm and Nd (ref. 3)). Among the different chondrite groups, the carbonaceous chondrites have stable isotopic compositions for several refractory elements furthest from the terrestrial composition, and so must constitute a small mass fraction of Earth’s building blocks14. The enstatite chondrites and a few ungrouped chondrites and achondrite meteorites are the only meteorite groups so far that share identical oxygen isotopic compositions15 with Earth and the Moon, suggesting a common reservoir for their accretion. The ordinary chondrites have an average 142Nd/144Nd deficit of −19 ± 5 p.p.m. (where the error indicates 2 standard deviations (s.d.); ref. 4) relative to modern terrestrial samples, whereas 142Nd/144Nd ratios in enstatite chondrites vary between those found in ordinary chondrites and those found in Earth2,3,4.

We report high-precision Sm and Nd isotope measurements and Sm/Nd ratios of some of the earliest-formed objects in the Solar System. CAIs condensed from the solar nebula gas are the oldest dated materials in the Solar System16. The internal mineral isochron method has been applied to individual CAIs to determine the initial Solar System abundances of short-lived radionuclides for several radiogenic systems17, including 146Sm–142Nd (ref. 18). In the Allende CV3 carbonaceous chondrite, CAIs can be affected by thermal metamorphism18,19. To avoid isotopic disturbances from metamorphism and to evaluate the extent of isotopic heterogeneities in CAIs, we selected inclusions that have previously been chemically and/or petrographically characterized and that have different crystallization histories (Methods). We investigated mineral separates (melilite and fassaite—a Ca, Al, Mg-rich silicate and an Al-rich pyroxene, respectively) and bulk sample powders of 1-cm-sized CAIs from the CV3-chondrite meteorites Northwest Africa (NWA) 2364 (the ‘crucible’, type B) and NWA 6991 (B4, compact type A; Extended Data Fig. 1) to constrain their internal Sm–Nd isotopic evolution. We also analysed as bulk powders two Allende fine-grained CAIs (Extended Data Fig. 1) that, on the basis of their fine-grained mineral textures, have not been melted since they condensed from the nebular gas. Primitive enstatite chondrites from petrologic type 3 of the EH (ALHA77295) and EL (MAC 02837 and MAC 02839) subgroups were also selected. The sulfide assemblages of these three enstatite chondrites indicate that they are the most pristine and unmetamorphosed enstatite chondrites available from the two subgroups4,20, suggesting that they are the best candidates to preserve both their Sm and Nd elemental abundances and any isotopic heterogeneities inherited from their formation region in the solar nebula. For each sample, the abundances of Sm and Nd stable isotopes, including the minor proton-rich isotope 144Sm (3.1% of total Sm), were measured with a precision of a few parts per million, along with the corresponding Sm/Nd elemental ratios (Supplementary Tables 1–3). Any modifications due to exposure to galactic cosmic rays were monitored, but this effect is small and does not induce any substantial modification of Nd isotope ratios (maximum 1 p.p.m.; Methods).

The two CAIs from NWA 2364 and NWA 6991 have 147Sm–143Nd ages in agreement with those obtained by U–Pb radiometric dating (ref. 16), showing that the Sm–Nd chronometers have not been disturbed by secondary processes (Extended Data Fig. 7a, b). Regression of 147Sm/144Nd–142Nd/144Nd (Fig. 1) for NWA 6991 fractions yields a μ142Nd value of −8.4 ± 5.1 p.p.m. at a chondritic Sm/Nd ratio (mean square weighted deviation, MSWD = 0.76)—a value that is indistinguishable within errors from the terrestrial equivalent (0 ± 5 p.p.m.; 2 s.d., measured on JNdi-1 Nd isotopic standard). The NWA 2364 CAI regression is less well constrained (MSWD = 2.8), yielding a μ142Nd value of −2.9 ± 25.4 p.p.m. (2 s.d.) at a chondritic Sm/Nd ratio. Therefore, internal isochron regression lines of two individual CAIs intersect the composition of the modern accessible Earth at a chondritic Sm/Nd ratio, which is the same as the Sm/Nd ratio of enstatite chondrites measured in this study (Fig. 1). These similar μ142Nd values at chondritic Sm/Nd ratios for two CAIs are in contrast with the literature average values for enstatite, ordinary and carbonaceous chondrites, which are, in that order, increasingly deficient in 142Nd (refs 2, 3, 4; Fig. 1). The CAIs have Sm isotope patterns with deficits in 144Sm, 149Sm and 154Sm, and excesses in 148Sm and 150Sm (Fig. 2a). The large deficits in 144Sm of up to −290 p.p.m. are clearly dominated by a proton-rich process (p-process) deficiency. The lack of correlation between 142Nd and 144Sm variations between Solar System objects characterized by different 144Sm signatures indicates that the contribution of p-processes to 142Nd is negligible, contrary to previous suggestions4,21 (Extended Data Fig. 5). The positive μ148Sm–μ150Sm correlation (Extended Data Fig. 3) shows that CAIs are the carrier of a slow-neutron-capture-process (s-process) excess and a rapid-neutron-capture-process (r-process) deficit in CV chondrites. The stable Nd isotope compositions of the bulk CAIs and their mineral fractions are similar to the terrestrial values within analytical error, except for the NWA 2364 CAI, which has deficits in 145Nd, 148Nd and 150Nd (Fig. 2b). We identified a CAI in CV3 NWA 6991 that is petrogenetically primitive (Methods) and shares an identical Nd stable isotopic composition with Earth. Because different mineral fractions have been separated, the radiogenic 142Nd isotope of NWA 6991 can be compared to the value for Earth without the need for any correction for neutron capture or nucleosynthetic anomalies. Its μ142Nd value at a chondritic Sm/Nd ratio is −8.4 ± 5.1 p.p.m., which is indistinguishable within errors from the terrestrial value (0 ± 5 p.p.m.). By contrast, the s-process excesses and r-process deficits identified in the Nd stable isotopes of NWA 2364 CAIs should substantially increase the 142Nd/144Nd ratios (Fig. 2a, b). The correction for 142Nd is complex and its accuracy is difficult to evaluate. Isotopes from the same element should be used to correct for nucleosynthetic effects3. A correction of −25 p.p.m. on 142Nd is calculated using different approaches: (i) by using the 142Nd–148Nd relationship calculated from a stellar model (μ148Nd = −1.01 × μ142Nd; ref. 22); (ii) by using the isotope composition measured in SiC as representative of a pure s-process component (μ148Nd = −1.15 × μ142Nd; ref. 23); and (iii) by considering the correlation in leachates data obtained on carbonaceous chondrites (μ148Nd = −0.96 × μ142Nd; ref. 24). A smaller correction of 14 p.p.m. to the 142Nd/144Nd ratio was calculated for the Allende Al3S4 CAI18, whereas the separated mineral fractions of Al3S4 were characterized by the highest deficits in both 148Nd and 150Nd measured in CAIs. We note that the mineral separates from Allende Al3S4 CAI have different μ148Nd and μ150Nd values18 that do not fall on the correction line between CAIs and carbonaceous chondrites (Fig. 3). In contrast, carbonaceous chondrites and bulk CAIs from this study and from ref. 25 plot on the μ148Nd–μ150Nd mixing line calculated from isotope production in stellar models22,26 represented in Fig. 3. These CAIs do not fall on the regression line defined by Nd isotopic data of leachates of the CM2 chondrite Murchison24, which instead plot towards the composition of presolar SiC grains. Ultimately, after correcting for nucleosynthetic anomalies and radiogenic decay, CAIs show a range of 142Nd/144Nd ratios similar to enstatite chondrites, with values between the modern terrestrial average composition and those of ordinary chondrites. Furthermore, our data on enstatite chondrites have a range of 147Sm/144Nd ratios from 0.1948 to 0.1958, within 1% of the chondritic uniform reservoir (CHUR) value of 0.1960 (ref. 27), and do not show any correlation between μ142Nd and μ144Sm anomalies, in contrast to a previous study (Extended Data Fig. 5). We can therefore calculate the initial and modern compositions at a common chondritic 147Sm/144Nd without any model dependence and without introducing large errors on initial 142Nd abundance. We calculate μ142Nd values from −3 p.p.m. to −9 p.p.m. when normalized at the CHUR 147Sm/144Nd ratio (Supplementary Table 1). From our data, we obtain an average enstatite chondrite composition with μ142Nd = −7 ± 6 p.p.m. (2. s.d.; n = 3), which is indistinguishable within errors from the composition of Earth’s mantle of 0 ± 5 p.p.m. (Fig. 1). The dataset presented in ref. 4 had more scattered Sm/Nd ratios, which introduced uncertainties when correcting for 142Nd produced by radiogenic decay and when normalizing 142Nd/144Nd ratios of individual whole-rock chondrites to a common CHUR Sm/Nd ratio. Nevertheless, we calculate a combined average for EL and EH chondrites (data from this study and ref. 4) of μ142Nd = −9 ± 17 p.p.m. (2 s.d.; n = 13 individual enstatite chondrites), which is less precise, but remains consistent with Earth’s composition.

Figure 1: 146Sm–142Nd mineral and bulk isochron of the Allende, NWA 2364 and NWA 6991 CAIs compared with the Allende Al3S4 CAI, the modern terrestrial composition and bulk-rock chondrite averages for enstatite, ordinary and carbonaceous chondrites.
figure 1

The data for Allende Al3S4 CAI are from ref. 18 and those for ordinary and carbonaceous chondrites are from ref. 4; all other data are from this study. The modern terrestrial composition is indicated by the blue rectangle labelled Earth, and is also represented by the grey band for the average 142Nd/144Nd ratio, which corresponds to μ142Nd of the JNdi-1 standard (0 ± 5 p.p.m.; 2 s.d.). Regressions of NWA 6991 (black line) and NWA 2364 (not shown) yield 146Sm/144Sm ratios of 0.0072 ± 0.0024 (MSWD = 0.76) and 0.0069 ± 0.0056 (MSWD = 2.8), respectively, at the age of Sm–Nd isotopic closure. Considering data from all CAIs from this study, the slope of the regression line (not shown) is 0.0073 ± 0.0022 (MSWD = 3.3). The NWA 6991 CAI does not have any stable isotopic variations in Nd relative to Earth, removing the need for potential corrections from nucleosynthetic or neutron capture effects. The internal isochron regression of NWA 6991 (black line; grey lines indicate the 95% confidence interval) intersects the 142Nd/144Nd compositions of Earth and enstatite chondrites at a common 147Sm/144Nd ratio, which we find matches the chondritic ratio of 0.196 (ref. 27). All errors bars are two standard errors (2 s.e.) for individual data points. Respective means and errors for chondrite groups are given in the text; those for samples and standards from this study are provided in Supplementary Table 1. The left y axis shows the 142Nd/144Nd ratios; the right y axis shows the μ142Nd values (in parts per million; deviation relative to the mean obtained on the JNdi-1 terrestrial standard). The heights of the rectangles for the chondrite groups indicate 2 s.d. on the μ142Nd averages (horizontal bars); the widths indicate 2 s.e. on the CHUR 147Sm/144Nd ratio27.

PowerPoint slide

Full size image
Figure 2: Sm and Nd isotope compositions of bulk CAIs (leached and unleached) and CV3 whole-rocks.
figure 2

a, b, μiSm (a) and μiNd (b) represent the isotopic composition measured in meteorites relative to the isotope ratio measured in the terrestrial standard, and is given in parts per million. Sm and Nd data were corrected for instrumental mass fractionation using the exponential law and 147Sm/152Sm = 0.56081 and 146Nd/144Nd = 0.7219, respectively (Supplementary Tables 1–3). Grey boxes show the external reproducibility (2 s.d.) obtained on the standards. The relative contributions of the p-, s- and r-processes for stable and radiogenic (142Nd and 143Nd) isotopes are indicated (as percentages) below the plots22. The now-extinct 146Sm is not represented, but is a pure p-process isotope. The two radiogenic Nd isotopes, 142Nd and 143Nd, are not represented because their deviations reflect mostly the radiogenic contribution.

PowerPoint slide

Full size image
Figure 3: 150Nd/144Nd versus 148Nd/144Nd measured in CAIs and carbonaceous chondrites.
figure 3

The data for carbonaceous chondrites are from ref. 9 and those for Allende CAIs are from refs 18 and 25. Error bars show the analytical uncertainty (2 s.e.); the grey boxes represent the external reproducibility (2 s.d.) obtained on the JNdi-1 standard. The internal errors for Allende bulk CAIs are not reported in ref. 25. The data plot on the mixing line calculated from isotope production in stellar models (solid black line)22,26, but not on the regression defined by leachate Nd isotopic data of the CM2 chondrite Murchison (dashed black line)24. The signatures in 148Nd and 150Nd of the Allende 322 and 323 and NWA 6991 CAIs are not different from the terrestrial standard considering the analytical uncertainties.

PowerPoint slide

Full size image

We find that several early Solar System objects, including CAIs and enstatite chondrites, share a common Nd stable isotope signature with modern terrestrial samples, even though these materials sample spatially or temporally distinct regions of the protoplanetary disk. The CAIs formed under reduced conditions, closest to the proto-star, before being transported outward to the accretion regions of the chondrite groups28. The reduced enstatite chondrites are suggested to have formed within the inner fringe of the asteroid belt, and the more oxidized ordinary chondrites and carbonaceous chondrites in the outer parts29. Our results therefore suggest a relationship between 142Nd abundance and heliocentric distance. Materials formed in the inner region of the Solar System have the highest 146Sm decay-normalized 142Nd/144Nd ratios, followed by whole-rock ordinary chondrites and carbonaceous chondrites. The amount of material required to have been hidden within or lost from Earth was previously constrained using 142Nd abundances measured in ordinary chondrites (see, for example, refs 5, 9). Although the building blocks of Earth are difficult to identify on the basis of meteorites that escaped accretion, enstatite chondrites are closest to Earth’s composition when looking at the abundance of 142Nd and of many other refractory elements14. Using the enstatite chondrite group as an isotope analogue of the accreting material that formed Earth, the calculated 147Sm/144Nd ratio of the silicate Earth would be 0.200, which is 2% higher than the CHUR 147Sm/144Nd ratio and within the range of values previously measured for whole-rock chondrites27. An external constraint on the Sm/Nd ratio for Earth arises from the combined Sm–Nd and Lu–Hf systematics of lunar samples, suggesting a chondritic Sm/Nd evolution of the bulk Moon30. Therefore, our results on CAIs together with those obtained on enstatite chondrites and lunar samples indicate that the Earth–Moon system evolved from a common isotopic reservoir in 142Nd, with Sm/Nd proportions within 2% of the average of unequilibrated whole-rock chondrites27. The increasing number of measurements of the 146Sm–142Nd systematics of Solar System objects calls into question the notion of a super-chondritic Earth for refractory elements. Instead, our results indicate that variations in 142Nd abundances found in planetary materials compared to the modern Earth are caused by nucleosynthetic anomalies, and eliminate the need for untapped or missing planetary reservoirs to explain the 146Sm–142Nd isotope systematics of terrestrial rocks5,6.

Methods

Samples and analytical methods

Calcium–aluminium-rich inclusions (CAIs) were obtained from the Center for Meteorite Studies at Arizona State University for Allende (named 322 and 323) and Northwest Africa (NWA) 6991 (named B4, fragment mass of about 150 mg), and from the American Museum of Natural History in New York for NWA 2364 (named the ‘crucible’, fragment mass of about 150 mg). For Allende, the two small (50–100 mg), fine-grained (unmelted) CAIs were processed as bulk powders. These four CAIs were selected on the basis of their petrogenetic differences, to assess potential Sm and Nd stable isotopic heterogeneities linked to different CAI groups and formation histories. Geochemical studies of fine-grained CAIs in CV3 chondrites indicate that they did not melt and that they belong to group II CAIs, on the basis of their rare-earth elemental abundances31. These characteristics point towards formation as a gas–solid condensate from the nebular gas with minor reprocessing31. The Sm and Nd concentrations normalized to CI chondrites are SmN = 19–20 and NdN = 20–22. The lower, normalized Lu concentrations of LuN = 0.3–3 (concentrations in Extended Data Table 1) that were measured in these two CAIs by isotopic dilution point towards a fractionated rare-earth element (REE) pattern and a group II classification for both Allende CAIs 322 and 323 (Extended Data Fig. 1). The inclusion B4 of NWA 6991 was petrographically characterized2 as a compact type A CAI. The CAIs B4 and crucible were processed for mineral separation. Mineral separates (melilite and fassaite) were concentrated from a 30–63-μm fraction using bromoform and methylene iodide heavy liquids, and hand picking as described in ref. 16. The same CAI B4 has been dated using the Al–Mg and U–Pb radiogenic decay systems at 4,567.94 ± 0.31 Myr (ref. 32). Bulk samples were processed at Arizona State University (Allende) and the University of Minnesota (NWA 6991 and NWA 2364) as unleached and as leached with 10% HF and, subsequently, with 20% HCl for 20 min each at room temperature. Mineral separates were similarly leached before dissolution. The leachates were not analysed for 146Sm–142Nd systematics owing to the limited leached sample masses.

Enstatite chondrites were obtained from the US Antarctic meteorite collection. Chips of about 1.2–1.4 g of enstatite chondrites MacAlpine Hills (MAC) 02837, 26, MAC 02839, 12 and Allan Hills A77295 (ALHA77295 or ALH 77295), 74 were crushed in an agate mortar and pestle into fine whole-rock powders by separating and remixing the metal to preserve all components together before acid digestion.

New PFA Savillex beakers were used for these samples to avoid memory effects. Parr bombs with 14-ml insert PTFE vials were not previously exposed to enriched spikes and were used for only meteorite dissolution. Only MilliQ Millipore water, BDH Aristar Ultra hydrofluoric and hydrochloric acids, and triple sub-boiled quartz-distilled nitric acid were used for all steps.

For enstatite chondrites, the whole-rock powders were dissolved using a first step in concentrated HF-HNO3 (10:1) at 120 °C on a hot plate for 2 days to liberate hydrogen sulfide and silicate tetrafluoride gases before transferring the residues into Parr pressured vials. For CAIs and enstatite chondrites, samples were placed into individual Parr bomb pressured vials in concentrated HF-HNO3 (10:1) for 7 days at 155 °C. After drying down the samples, perchloric acid was added and evaporated at 200 °C on a hot plate to break down the fluorides, and samples were taken back into solution in Parr bomb pressured vials in 6 M HCl for 2 days at 155 °C. Solutions were transferred into Savillex PFA beakers and split into an approximately 90% fraction for isotopic composition, an approximately 5%–10% fraction spiked with an enriched 149Sm–150Nd mixed spike for isotopic dilution to determine the Sm/Nd ratio (using the measured stable Sm and Nd isotopic compositions), and an approximately 5% fraction for inductively coupled plasma mass spectrometry (ICPMS) analysis for trace element abundances for the NWA 6991 CAI bulk sample (Extended Data Fig. 1). The CAI B4 has a flat REE pattern normalized to CI chondrites ((17–20) × CI chondrites) with slight positive Eu anomaly (EuN = 23) corresponding to a chemical group I CAI7. The same group I is suggested for NWA 2364 unleached bulk CAI with SmN = 17, NdN = 16 and LuN = 20 (CI-chondrite normalized REE concentrations shown in Extended Data Fig. 1 and concentrations given in Extended Data Table 1).

All Sm and Nd concentrations obtained by isotope dilution are reported in Extended Data Table 1. Sm and Nd in the spiked fractions were purified and separated using a two-stage chemistry procedure. The REEs were first separated from the matrix using a 2-ml or 8-ml cation resin AG50W-X8 column. The REE fraction was then loaded in 0.14 M HCl onto a 1-ml Eichrom Ln-spec resin bed to separate the Nd and Sm fractions in 0.14 M HCl and 0.40 M HCl, respectively. The chemistry procedure was different for the unspiked fractions. After the first 8-ml cation columns, the REE fractions were processed twice through a cation column using 2-methylactic acid (0.2 M, pH 4.7) with a small amount of H2O2. The Sm fraction was collected before the Nd fraction during the first chemistry, but at this stage heavier REEs were still present within the two collected fractions. The last step consists of one pass through Ln-spec resin in weak HCl. This procedure ensures perfect separation of Nd from Ce and Sm by reducing the effect of interferences on mass 142 from Ce and on masses 144, 148 and 150 from Sm, and the purification of the Sm fraction (Gd interferes at masses 152 and 154). Organic residues are then completely removed using hydrogen peroxide and the sample is ready to be loaded on a Re filament. Total procedural blanks for Sm and Nd were 3 pg and 10 pg, respectively. The blank contribution to each sample was negligible, with Sm and Nd fractions of at least 80 ng for the smallest analysed sample.

Spiked Sm and Nd isotopic fractions for the isotopic dilution method were analysed in static mode using a MC-ICPMS Neptune at Arizona State University for Allende CAIs, and Neptune Plus at the University of Minnesota and the Laboratoire Magmas et Volcans (Clermont-Ferrand, France) for the NWA 6991 and NWA 2364 CAIs using methods described in ref. 27. AMES Sm and JNdi-1 Nd standards were used to control mass bias and cup efficiency during all the sessions. The calculated Sm and Nd concentrations and 147Sm/144Nd ratio of samples are reported in Supplementary Table 1. Additionally, a BCR-2 rock standard was processed and analysed in each laboratory, and Sm and Nd concentrations and calculated 147Sm/144Nd and 143Nd/144Nd ratios are reported in Supplementary Table 1.

Sm and Nd isotopes of unspiked fractions were obtained on the Thermo-Fisher Triton thermal ionization mass spectrometer at the Laboratoire Magmas et Volcans. The purified Sm and Nd cuts were loaded in 2.5 M HCl on outgassed, zone-refined Re filaments. The Sm was measured in static mode as Sm+. The Faraday cup configuration was centred at mass 149 and Nd and Gd interferences were monitored at masses 146 (cup L3) and 156 (cup H4). Each run consisted of 18 to 27 blocks of 20 cycles using amplifier rotation with background acquisition before each block. Sample measurements were in general shorter depending on the amount loaded (see Supplementary Table 2). Signals for 152Sm ranged between 0.5 × 10−11 A and 3.5 × 10−11 A for standards and between 0.5 × 10−11 A and 2.0 × 10−11 A for samples. Sm data were corrected for instrumental mass fractionation using the exponential law and 147Sm/152Sm = 0.56081. Sm isotope compositions for standards and samples are given in Supplementary Table 2. Sm isotope compositions were measured during three different periods, called session #1 to #3 in Supplementary Table 2. The μ values correspond to deviations relative to the standard value, expressed in parts per million. In static mode, the ratios evolve with time owing to cup ageing. Consequently, we used the average of ratios measured on standards during the same analytical session for calculating the sample μ values. The external reproducibility calculated from repeated measurements of the standard (5–7 standards measured per analytical session) is 15–37 p.p.m. on 144Sm/152Sm, 5–18 p.p.m. on 148Sm/152Sm, 7–11 p.p.m. on 149Sm/152Sm, 14–27 p.p.m. on 150Sm/152Sm and 12–17 p.p.m. on 154Sm/152Sm. The Nd samples were measured as Nd+ using 9 Faraday cups. We used a dynamic procedure (axial cup centred at masses 143 and 145) to measure the 142Nd/144Nd ratios in dynamic mode; all other ratios were measured in static mode (line 1 centred at 145). Ce and Sm interferences on masses 142 and 144 were monitored using masses 140 and 147, respectively. Ce and Sm contributions on masses 142 and 144 were lower than 3 p.p.m. for standard and sample measurements, except for sample NWA 6991, in which the Sm contribution on mass 144 reaches 15 p.p.m. Nd isotope ratios were corrected for mass fractionation to 146Nd/144Nd = 0.7219 using the exponential law. The external reproducibility (2 s.d.) calculated from repeated measurements of the standard within each session is 4–6 p.p.m. on 142Nd/144Nd, 3–7 p.p.m. on 143Nd/144Nd, 4–8 p.p.m. on 145Nd/144Nd, 3–12 p.p.m. on 148Nd/144Nd and 10–12 p.p.m. on 150Nd/144Nd (Supplementary Table 3).

Sm isotopes

The Sm isotopic compositions measured for terrestrial standards and samples are presented in Supplementary Table 2. For samples measured during the first sequence, the quantity of Nd is substantially higher than that present in the Sm standard. Nevertheless, there is no correlation between Sm isotope ratios and the quantity of Nd found in our samples (Extended Data Fig. 2). The 144Sm deficits in all bulk CAI samples are in the range −236 p.p.m. to −288 p.p.m. and do not correlate with Nd/Sm (Extended Data Fig. 2). The melilite fraction of the NWA 2364 CAI has the lowest deficit in 144Sm (μ144Sm = −113 p.p.m.) and lower Sm isotope anomalies than other CAI fractions in general. This sample was measured with the lowest Sm intensity and during a very short period (60 ratios), so isotopic ratios have a precision that is three times lower than those of the other samples. All CAI fractions, except for the NWA 6991 leached bulk fraction and the melilite from NWA 2364, have excesses in 148Sm (+45 p.p.m. to +84 p.p.m.), deficits in 149Sm (−25 p.p.m. to −95 p.p.m.), excesses in 150Sm (+114 p.p.m. to +240 p.p.m.) and deficits in 154Sm (−18 p.p.m. to −54 p.p.m.). The Sm isotope patterns are presented in Fig. 2a. In this representation, the Sm isotopes at masses 147 and 152 are equal to zero because measured data are normalized using the 147Sm/152Sm ratio.

Deficits in 144Sm can reflect deficits in p-process isotopes, but they are also artificially created by s-process excess and/or r-process deficits using the 147Sm/152Sm normalization scheme. The two Sm isotopes used for mass bias correction are predominantly formed by r-processes (88% and 77% for 147Sm and 152Sm, respectively), and the remaining part is formed by s-processes. The positive μ148Sm–μ150Sm correlation (Extended Data Fig. 3) shows that CAIs are the carrier of an s-process excess and an r-process deficit. This correlation is in agreement with the general shape of the pattern, with deficits in 144Sm, 149Sm and 154Sm, and excesses in 148Sm and 150Sm (Fig. 2a). The NWA 6991 CAI has the highest deficits in 149Sm and highest excesses in 150Sm, and this sample does not fall on the μ148Sm–μ150Sm correlation line formed by CV3 chondrites and other CAIs25. This result suggests that the Sm isotope composition of the NWA 6991 CAI has been modified by neutron fluence capture, owing to long-term exposure to galactic cosmic rays following the reaction 149Sm(n, γ)150Sm. Nevertheless, this effect is small and does not induce any substantial modification of Nd isotope ratios because Nd isotopes have smaller neutron capture cross sections than does 149Sm (ref. 1). CAIs and carbonaceous chondrites plot on the same correlation line, which intersects the 148Sm terrestrial value at μ144Sm = −60 p.p.m. in the μ144Sm–μ148Sm diagram (Extended Data Fig. 3). This value is in agreement with the deficit in 144Sm of about −80 p.p.m. measured in carbonaceous chondrites that have terrestrial 148Sm abundances3,21. The CAIs represent the end-member farthest from the terrestrial composition. The three enstatite chondrites have terrestrial compositions for both 148Sm and 154Sm, indicating that the 149Sm–150Sm signature (excess in 149Sm coupled to deficit in 150Sm) reflects secondary reactions. Deficit or excess in s-processes must generate similar deviations in both 148Sm/152Sm and 150Sm/152Sm ratios. The two EL3 samples MAC 02837 and 02839 have variable 144Sm abundances with μ values of −3 ± 16 p.p.m. and 48 ± 12 p.p.m., respectively.

Nd isotopes

The Nd isotopic compositions measured for CAIs and enstatite chondrite samples are presented in Supplementary Table 3. 142Nd/144Nd ratios were measured in static and dynamic modes. Data are compared in Extended Data Fig. 4 and ratios fall on the 1:1 line. The contribution of Ce and Sm measured in JNdi-1 standard and samples is negligible, with the exception of the NWA 2364 CAI measured during the third session, in which the cerium correction on the 142Nd/144Nd ratio is 560 p.p.m. The stable Nd isotope compositions of the bulk CAIs and their mineral fractions are similar to the terrestrial values within the analytical error, except for the NWA 2364 CAI, which has deficits in 145Nd (μ = −26 ± 4 p.p.m.), 148Nd (μ = −25 ± 6 p.p.m.) and 150Nd (μ = −48 ± 9 p.p.m.) (Fig. 2b). As observed for Sm isotopes, this signature results from s-process excesses and r-process deficits. The Nd isotope patterns are presented in Fig. 2b. Using this representation, the Nd at masses 144 and 146 are equal to zero, because the measured data are normalized using the 146Nd/144Nd ratio. For Nd isotopes in enstatite chondrites, MAC 02837 has a terrestrial abundance in 142Nd (μ = −6 ± 7), whereas MAC 02839 and ALHA77295 have small deficits (μ = −10 ± 6 and μ = −11 ± 3, respectively). For stable isotopes, the abundances are identical to those measured in Nd standards and terrestrial samples, except for MAC 02837, which has a deficit of −115 ± 9 p.p.m. in 150Nd abundance. Because all other Nd isotopes are present in terrestrial abundances, the cause of this deficit in 150Nd is unclear.

Neutron fluence effect on the Sm isotope patterns of CAIs

The Sm isotope patterns for CAIs have deficits in 149Sm and excesses in 150Sm. The negative correlation between the 149Sm/152Sm and 150Sm/152Sm ratios (Extended Data Fig. 3) may reflect deficits or excesses in s- and r- process nuclides because 149Sm is dominated by r-processes (88%), whereas 150Sm is a pure s-process nuclide. This correlation can also be created by nuclear reactions induced by interaction between galactic cosmic rays and the surface of extraterrestrial material. At low energy, epithermal and thermal neutron reactions produce nuclides by neutron capture. 149Sm has the largest neutron capture cross-section of all Sm isotopes and the reaction 149Sm(n, γ)150Sm is commonly used for estimating the amplitude of this secondary nuclear reaction in lunar samples (black line in Extended Data Fig. 3a drawn from lunar data33). Nuclear reactions induced by exposure to galactic cosmic rays are the result of a surface process that becomes negligible at relatively small depths (below 2–3 m; ref. 34). The deficits in 149Sm measured in Allende CAIs (−25 p.p.m. to −31 p.p.m.) are similar to those measured in the different dissolutions of the Allende bulk sample3,21, but the excesses in 150Sm in CAIs (+147 p.p.m. to +176 p.p.m.) are higher than those measured in the whole rocks (+90 p.p.m.). The Sm isotope patterns of Allende CAIs are not affected by neutron fluence. The NWA 6991 CAI has both a higher deficit in 149Sm (−95 p.p.m.) and a higher excess in 150Sm (+235 p.p.m.) than other CAIs measured in this study, and plots close to the whole-rock CV3 Mokoia in Fig. 2a (data from ref. 3). The NWA 6991 CAI also falls far from the regression line for Allende, NWA 2364 and the CV3 whole-rock chondrites in the μ148Sm–μ150Sm diagram (except Mokoia, Extended Data Fig. 3b). Because 148Sm and 150Sm are pure s-process nuclides, the two 148Sm/152Sm and 150Sm/152Sm ratios should be positively correlated for samples that have not been affected by interaction with galactic cosmic rays. The excess observed in 150Sm for the NWA 6991 CAI is explained by secondary production of this isotope, and, using the correlation plotted in Extended Data Fig. 3b, we can estimate an increase in μ150Sm of about 100 p.p.m. Using the 149Sm–150Sm relationships established for lunar samples33, we calculate that an excess of about 100 p.p.m. in 150Sm should correspond to a decrease in μ149Sm of about 40 p.p.m. Such a variation will have a negligible effect (<1 p.p.m.) on the corrected 142Nd/144Nd ratio, considering the relationship1,35 between Sm and Nd isotopes.

P-process contribution to 142Nd production

142Nd is not a pure s-process isotope, but contains a small fraction of p-process isotope that varies between 4% and 20%, with the lower estimate obtained from stellar nucleosynthesis models36 and the upper estimate from measurements of meteorites4. By monitoring the abundance of the pure p-process 144Sm isotope in CAIs, we evaluate how this process affects the corresponding 142Nd/144Nd ratio. The Sm isotope patterns of all CAIs are similar to those measured in CV3 chondrites, but the anomalies (whether excess or deficit) are commonly larger (Fig. 2a), with two major differences. First, CAIs have large deficits in 144Sm, up to −290 p.p.m. relative to the bulk CV (−100 p.p.m. on average3,21). Second, they have small excesses in 148Sm, whereas CV3 chondrites have terrestrial 148Sm abundances. Deficits in 144Sm can reflect deficits in p-process isotopes, but apparent deficits may also arise from the 147Sm/152Sm normalization used to correct instrumental mass bias when applied to samples with s-process excesses and/or r-process deficits. The contribution of the deficits induced by the normalization is estimated using stellar models22,26, and shows that the 144Sm deficit should be about 25% of the excess in 148Sm, which corresponds to a maximum deficit of 15 p.p.m. in 144Sm. The measured deficits in μ144Sm of >200 p.p.m. in CAIs are therefore clearly dominated by a p-process deficit.

It has been shown36 that a fractionated Nd/Sm p-process isotope contribution ratio resulting from chemical separation upon condensation in a circumstellar environment could also explain the μ144Sm–μ142Nd isotope correlation found in ref. 4, with as little as 4% p-process contribution on 142Nd. Combined Sm–Nd isotope measurements of different Solar System objects show that materials with a large range of μ144Sm values have identical μ142Nd compositions, as illustrated for the measured CAIs and the FUN (fractionated and unknown nuclear effects) inclusion C1 in Extended Data Fig. 5. The CAIs and FUN C1 fall on the 1% p-process–99% s-process 142Nd line, which suggests that the contribution of p-process to 142Nd is negligible and within our measurement errors. Moreover, after including our enstatite chondrite data, the μ144Sm–μ142Nd correlation that had been previously found between the different groups of chondrites4 is no longer present. Only the carbonaceous chondrites are clearly distinct, with the highest deficits in both 144Sm and 142Nd. All the different mineral fractions and the bulk CAIs show excesses in s-process isotopes and deficits in r-process isotopes for Sm, whereas deviations in Nd isotope ratios relative to the terrestrial standard are observed for only NWA 2364 (Fig. 2b). Our data also confirm the variations in r-process deficits with a stronger effect on Sm in comparison to Nd (ref. 25).

The abundance of p-process isotopes relative to mass is shown in Extended Data Fig. 6. For isotopes formed by both p- and s-processes, the contribution of p-process isotopes are calculated using data from ref. 22 as the solar abundance minus the contribution of the s-process isotopes. The abundance of p-processes (normalized to Si) was recalculated using the total abundance of isotopes given in ref. 37. Extended Data Fig. 6 shows that the relation is well approximated by a power law. The abundance of pure p-process nuclides falls close to the curve. However, when we compare abundances calculated for ‘mixed’ isotopes (formed by mixing of p- and s-process isotopes such as 76Se, 80Kr, 142Nd, 152Gd and 164Er), the abundances fall far from the calculated curve. If the 142Nd was formed by 1% of p-process isotopes, it should fall close to the curve.

146,147Sm–142,143Nd isochrons

The 147Sm–143Nd isotopic compositions of the CAIs are shown in Extended Data Fig. 7a. We obtain an absolute 147Sm–143Nd age of 4,519 ± 140 Myr (MSWD = 0.77, initial 143Nd/144Nd = 0.50675 ± 0.00020) for all CAI fractions using a half-life value for 147Sm of (6.539 ± 0.061) × 10−12 yr−1 (ref. 38). The initial 143Nd/144Nd is in agreement with previous determinations of the initial composition of 143Nd/144Nd in the Solar System of 0.50669 ± 0.00007 (ref. 27). If we calculate individual internal CAI isochrons, we find the most precise internal age for NWA 6991 of 4,523 ± 150 Myr (MSWD = 1.7, initial 143Nd/144Nd = 0.50673 ± 0.00021). Our 147Sm–143Nd isochron ages are less precise than the Allende CAI age of 4,560 ± 34 Myr (ref. 18). A larger dispersion may come from the fact that we include four different CAIs from three different meteorites with potentially various evolution and metamorphic histories. Our mineral fractions have a limited range of Sm/Nd ratios in comparison to data presented in ref. 18. The melilite of NWA 2364 and melilite-residue of NWA 6991 with the lowest Sm/Nd ratio were too small for precise 142Nd/144Nd analysis and thus spiked for 143Nd/144Nd analysis.

In the 144Sm/144Nd versus 142Nd/144Nd diagram (Extended Data Fig. 7b), the slope of the best-fit line gives the 146Sm/144Sm ratio at the time of isotopic closure. When data obtained on the four CAIs from Allende, NWA 6991 and NWA 2364 are plotted together, they yield a 146Sm/144Sm ratio of 0.0073 ± 0.0022 (MSWD = 3.3) (Fig. 1). This ratio is indistinguishable (within errors) from the value of 0.0083 ± 0.0004 defined from the study of the Allende Al3S4 CAI18.

In conclusion, the initial 146Sm abundance deduced from CAI isochrons is inconsistent with the 146Sm/144Sm ratio obtained from the initial eucrite isochron calculated back to the age of CAIs using a half-life of 68 Myr (ref. 7). Instead, our result and those of ref. 18 support the use of a 146Sm half-life of 103 Myr (ref. 39). Having a good estimate of the 146Sm half-life is essential for using this short-lived chronometer. As shown in Extended Data Fig. 8, such variation in the 146Sm half-life produces age differences of up to 150 Myr for Hadean rocks.

Influence of CAIs on the bulk whole-rock carbonaceous chondrite Sm isotope compositions

The abundance of CAIs varies among chondrite groups, and CV chondrites contain on average about 3 vol% CAIs40. In CM chondrites, the proportion of CAIs is lower than in CV chondrites, and bulk samples are also characterized by lower deficits in 144Sm (refs 3, 21). When we compare the measured 144Sm negative anomalies measured in CAIs (this study) with enstatite, ordinary and carbonaceous whole-rock chondrites (this study and refs 3, 20) (Extended Data Fig. 9), we find that a volume abundance of CAIs in carbonaceous chondrites of approximately 3%–5% can explain the anomalies found in bulk CV3 (and CM2) chondrites, whereas enstatite chondrites and ordinary chondrites do not have any 144Sm anomalies. However, the large deficit of 144Sm in the Orgueil meteorite (CI chondrite, μ144Sm = −103 ± 46 p.p.m., ref. 3) cannot be explained by the presence of CAIs because the abundance of CAIs in CI chondrites is lower than 0.01% (ref. 40). Therefore, 144Sm deficits are inherent to the matrix of CI chondrites instead of CAIs in CV and CM chondrites. This suggests that there are p-process deficits in all carbonaceous chondrite groups, hosted within the matrix in CI chondrites or within CAIs in other carbonaceous chondrite groups.

Conversely, CAIs have s-process deficits and r-process excesses, whereas the bulk rock isotope compositions tend towards a component that has an opposite signature, and that could be similar to the FUN inclusion EK1-4-1 (refs 41, 42), which display large excesses in r-process isotopes (Extended Data Fig. 5). Deficits in 144Sm are carried by CAIs in CV chondrites, whereas deficits in 142Nd are carried within the matrix of chondrites, and are the most extreme in carbonaceous chondrites.

The CAIs measured in this study have 147Sm/144Nd ratios between 0.1755 and 0.2405. Adding 3%–5% CAIs would increase the 147Sm/144Nd ratio of the CV bulk chondrites to 0.198. Literature 147Sm–143Nd data on Allende were reported in the supporting information of ref. 4. Measured Sm/Nd ratios for CV bulk rocks range from 0.1829 (ref. 35) to 0.2262 (ref. 43). These two extreme values were measured for very small Allende samples (90 mg and 0.6 mg, respectively). The dissolution of Allende bulk-rock chips of several hundreds of milligrams to 1 g gives values closer to the average chondritic ratio of 0.1960 (ref. 8). Five CV3 whole-rock chondrites were previously measured27 to have an average 147Sm/144Nd ratio of 0.1955 ± 0.0058 (2 s.d.).