Main

Coupled 146,147Sm–142,143Nd systematics is a powerful tool to constrain the timescales and processes involved in the early differentiation of Earth, the Moon and Mars6,7,11,12,13,14. However, the interpretation of 142Nd signatures is complicated by the presence of nucleosynthetic isotope variations between the terrestrial planets and meteorites. Such isotope anomalies arise from the heterogeneous distribution of presolar matter at the planetary scale, and have been documented for several elements15,16,17,18. Because different Nd isotopes have varying contributions from the proton process (p-process), the rapid neutron capture process (r-process), and the s-process of stellar nucleosynthesis (Extended Data Fig. 1), the observed 142Nd deficits in chondrites relative to the accessible Earth could, in principle, be nucleosynthetic in origin and hence unrelated to 146Sm decay8,16,19. Previous studies have identified nucleosynthetic Nd (and Sm) isotope anomalies in chondrites15,17 and their components20,21,22,23, but these effects do not seem to fully account for the observed 142Nd deficits in chondrites. For instance, although the 142Nd composition of carbonaceous chondrites can partly be attributed to s- or p-process deficits15,17, corrections for these effects still leave an 142Nd deficit of approximately 20 p.p.m. compared with the accessible silicate Earth. This would be consistent with Nd isotope data for bulk ordinary chondrites, which exhibit a deficit of a similar magnitude, but do not seem to show resolvable nucleosynthetic Nd isotope anomalies15,17,24. Enstatite chondrites have 142Nd deficits of approximately 10 p.p.m. and also do not show clearly resolved nucleosynthetic Nd isotope anomalies24. Thus, previous studies concluded that the 142Nd difference between chondrites and the accessible Earth largely reflects 146Sm decay and early Sm/Nd fractionation in the silicate Earth15,17,24. However, this interpretation remains uncertain because the available bulk chondrite data are of insufficient precision to detect the collateral effects of nucleosynthetic heterogeneities on non-radiogenic Nd isotopes and therefore do not permit the reliable quantification of nucleosynthetic 142Nd variations (Fig. 1).

Figure 1: Nd isotope compositions of enstatite and ordinary chondrites.
figure 1

Data from this study (solid symbols) show less scatter and more precisely defined mean values (the grey bars represent the 95% CI of the means from Student’s t-values) than data from previous studies6,15,17,24 (open symbols), and thus reveal systematic correlated anomalies in all Nd isotopes. The uncertainties shown for the individual data points are 2 standard errors (s.e. = standard deviation/; where n is the number of cycles per measurement) of the individual measurements. The origin of the different μ150Nd values for the ordinary chondrites analysed in this study and previous studies is unclear. We note, however, that our processed standards are indistinguishable from the unprocessed JNdi-1 standard within the uncertainty, rendering it unlikely that an analytical effect in our study is responsible. Furthermore, our μ150Nd data for ordinary chondrites are correlated with anomalies in other Nd isotopes, as expected.

PowerPoint slide

Full size image

Here we use high-precision Nd and Sm isotope measurements to better quantify the nucleosynthetic Nd isotope variations between chondrites and the Earth, with the ultimate goal of determining the magnitude of any radiogenic 142Nd difference. We digested larger sample sizes (around 2 g) than in most previous studies, allowing us to obtain higher-precision Nd and Sm isotope data for a comprehensive set of meteorites that includes 18 chondrites, the ungrouped brachinite-like achondrite NWA 5363 and the Ca–Al-rich inclusion (CAI) A-ZH-5 from the Allende chondrite (Table 1). To evaluate the accuracy of our data we also processed the JNdi-1 standard and the terrestrial basalts BHVO-2 and BIR-1 using the same analytical procedures. Within uncertainty, the Nd and Sm isotope compositions of the processed and unprocessed standards (JNdi-1, AMES) are indistinguishable (Table 1; Figs 2, 3).

Table 1 Sm/Nd ratios and Nd and Sm isotope compositions of meteoritic and terrestrial samples
Full size table
Figure 2: Nd and Sm isotope variations among meteoritic and terrestrial samples.
figure 2

a, Anomalies in non-radiogenic Nd isotopes μ145Nd and μ148Nd are consistent with a heterogeneous distribution of s-process Nd. Solid, dotted and dashed lines are the mixing lines between terrestrial Nd and s-process Nd, calculated using modelled s-process compositions27, Nd data for presolar SiC grains26 and Nd data for chondrite leachates20,21, respectively. The isotopic composition measured for bulk Allende can be accounted for by the admixture of CAIs in a CAI-free carbonaceous chondrite (CC) source reservoir (the ‘Allende without CAI (calculated)’ data point) that is characterized by an s-process deficit. b, Same as a but for μ148Nd and μ150Nd. c, The p-deficit observed for bulk Allende in μ144Sm can also be attributed to the admixture of CAIs. The grey dashed CC − CAI line represents a mixing line calculated by subtracting CAIs from the isotopic composition measured for bulk Allende. Error bars indicate the 95% CI.

PowerPoint slide

Full size image
Figure 3: Nd and Sm isotope variations among meteoritic and terrestrial samples.
figure 3

a, For enstatite chondrites, ordinary chondrites and NWA 5363, the μ142Nd anomalies are correlated with the non-radiogenic μ145Nd anomalies, as expected for a heterogeneous distribution of s-process Nd. The Allende carbonaceous chondrite plots off this correlation owing to the admixture of CAIs. Mass balance calculations indicate that a CAI-free carbonaceous chondrite source reservoir was characterized by an s-process deficit. b, Same as a but in μ142Nd versus μ148Nd space. c, Same as a but in μ142Nd versus μ150Nd space. d, Same as a but in μ142Nd versus μ144Sm space. Error bars indicate the 95% CI.

PowerPoint slide

Full size image

Most of the chondrites investigated cluster tightly around a 147Sm–143Nd isochron at 4.568 billion years ago (Ga) (Extended Data Fig. 2a). Only the EL6 chondrites Atlanta and Blithfield plot away from the isochron, probably reflecting disturbance by late-stage impact events25; the 142Nd data of these samples are therefore excluded from the following discussion. After correction of measured μ142Nd (for definition of μiNd and μiSm see Table 1) values for 146Sm decay to the average chondritic 147Sm/144Nd = 0.1960 (ref. 1; Extended Data Table 1), the μ142Nd values are tightly clustered for each chondrite group: the enstatite chondrites define a mean μ142Nd = −9 ± 5 (95% confidence interval; CI), the ordinary chondrites a mean μ142Nd = −17 ± 2 (95% CI) and the Allende CV3 chondrite a mean μ142Nd = −31 ± 1 (95% CI). NWA 5363 exhibits a decay-corrected μ142Nd of −16 ± 7, similar to ordinary chondrites, whereas CAI A-ZH-5 has a decay-corrected μ142Nd of −15 ± 8, consistent with the data for other Allende CAIs22.

In addition to variations in μ142Nd, we find resolved systematic variations in non-radiogenic Sm and Nd isotopes (Table 1, Figs 1, 2, 3). Compared with previous studies we observe less scatter for each chondrite group, reflecting the long duration and high beam intensity of our measurements, which result in more precisely defined average values for each group (Fig. 1). Plots of μ145Nd and μ150Nd versus μ148Nd reveal positively correlated anomalies, with the enstatite chondrites being closest to the terrestrial value, followed by carbonaceous and ordinary chondrites, and then NWA 5363 (Fig. 2a, b). The meteorite samples plot along mixing lines between terrestrial Nd (that is, μiNd = 0) and pure s-process Nd, regardless of whether the s-process composition is derived from presolar SiC grains26, nucleosynthesis models27 or data for acid leachates of primitive chondrites20,21. Thus, the variability in non-radiogenic Nd isotopes among the meteorites reflects variable s-deficits relative to the Earth, consistent with inferences from other elements16,28,29.

The μ145Nd, μ148Nd and μ150Nd anomalies of Allende are similar to those of ordinary and enstatite chondrites, although for most other elements nucleosynthetic anomalies are typically largest in carbonaceous chondrites16,18,28,29,30. The reason for the subdued Nd isotopic anomalies in Allende is the presence of CAIs, which host about half of the Nd and Sm in Allende31, and which, for these elements, are characterized by an s-excess and a p-deficit (Figs 2, 3). Mass balance calculations (see Methods and Extended Data Table 2) indicate that the composition of a CAI-free Allende would have μ145Nd, μ148Nd and μ150Nd values of 27 ± 14, 39 ± 28, and 56 ± 41, respectively; these anomalies are larger than those of ordinary and enstatite chondrites and thus imply that before addition of CAIs, the Allende chondrite had a substantial s-deficit (Fig. 2a, b). This interpretation is consistent with Sm isotope data for Allende and other carbonaceous chondrites, because the calculated CAI-free composition of these chondrites also shows an s-deficit (Fig. 2c, Extended Data Fig. 3). Thus, the displacement of the carbonaceous chondrites from the s-deficit line that is defined by ordinary and enstatite chondrites reflects the admixture of CAIs to carbonaceous chondrites. Note that, for ordinary and enstatite chondrites the effects of admixing CAIs are estimated to be no larger than 2 p.p.m. for Nd isotope ratios and 5 p.p.m. for Sm isotope ratios (Methods, Extended Data Table 2), and that the expected s-process Sm isotope anomalies (<10μ144Sm and >−20μ148Sm) for these two groups of chondrites are too small to be resolvable with the analytical precision of our Sm isotope measurements.

Using the information gained from the non-radiogenic isotopes, we can now assess the effect of nucleosynthetic anomalies on μ142Nd. The bulk meteorite data show inverse correlations between μ142Nd and μ145Nd, μ148Nd, μ150Nd and μ144Sm (Fig. 3), which are consistent with the covariations expected from a heterogeneous distribution of s-process isotopes. Enstatite and ordinary chondrites, as well as NWA 5363, plot on mixing lines between terrestrial and s-process Nd. The Allende CV3 chondrite is displaced from these correlations owing to the admixture of CAIs, and a calculated CAI-free carbonaceous chondrite composition plots on the s-mixing line defined by the other meteorites (Fig. 3).

The slopes obtained from linear regressions of the bulk meteorites (excluding Allende) are in good agreement with those calculated for mixing lines between terrestrial and s-process Nd, regardless of which estimate for the s-process composition is used20,21,26,27 and whether or not the calculated CAI-free carbonaceous chondrite composition and the processed standards are included in the regressions (Extended Data Fig. 4). The intercept values obtained from the regressions can thus be used to determine μ142Nd values corrected for s-process heterogeneity. For all regressions the intercept values are indistinguishable from each other, with an average value of approximately −5 p.p.m. relative to the JNdi-1 standard (Extended Data Table 3). Alternatively, μ142Nd values corrected for nucleosynthetic anomalies can be calculated for each meteorite group separately, using their measured μ145Nd, μ148Nd and μ150Nd values combined with the slopes of the s-mixing lines. Regardless of which s-process mixing relationships are applied, the calculated μ142Nds-corrected values are all mutually consistent and indistinguishable from each other (Extended Data Table 3), resulting in an average μ142Nds-corrected = −5 ± 2 p.p.m. Although this value is slightly negative, it is within the approximate long-term ±5 p.p.m. reproducibility of the JNdi-1 standard. When the regressions and corrections are calculated relative to the mean Nd isotope composition measured for the processed terrestrial standards, μ142Nds-corrected reduces to −2 ± 2 p.p.m. (Extended Data Table 3). We conclude that after correction for nucleosynthetic Nd isotope heterogeneity, the 142Nd compositions of chondrites and the accessible silicate Earth are indistinguishable at the current level of analytical precision (approximately 5 p.p.m.).

The lack of a resolved radiogenic 142Nd difference between chondrites and the accessible silicate Earth supports the long-standing paradigm of a chondritic Sm/Nd for the bulk Earth and requires the revision of the conclusions from several previous studies about the early differentiation, composition, structure and heat budget of the Earth. These studies interpreted the 142Nd offset between chondrites and terrestrial samples as resulting from 146Sm-decay and an early global Sm/Nd fractionation in the Earth’s mantle3,6,7,9,10. Our results instead demonstrate that chondrites and the accessible Earth have indistinguishable radiogenic 142Nd compositions, refuting the evidence for an early global silicate differentiation of the Earth and indicating that the hidden, enriched reservoir hypothesized in earlier studies3,6,9,10 does not exist. Moreover, our results rule out the extensive loss of early-formed crust by collisional erosion3,7,9, because otherwise the bulk silicate Earth would not have a chondritic Sm/Nd ratio. Finally, the evidence for chondritic Sm/Nd ratios in the bulk Earth implies chondritic abundances of other refractory elements, including the heat-producing elements U and Th. Thus, the total radiogenic heat generated over Earth’s history is almost a factor of two higher than was estimated recently for a non-chondritic composition of the Earth9.

Our results demonstrate that chondrites are the most appropriate proxy for the elemental composition of the Earth. However, they also highlight that chondrites cannot be the actual building blocks of the Earth because they are deficient in a presolar component that contains s-process matter. The s-process deficit increases from enstatite via ordinary to carbonaceous chondrites, indicating that the distribution of presolar matter in the solar protoplanetary disk either varied as a function of heliocentric distance or over time. For instance, the nucleosynthetic isotope heterogeneity within the disk may reflect differences in the thermal processing of stellar-derived dust, imparting isotopic heterogeneity on an initially homogeneous disk, but it could also reflect distinct compositions of material added to the disk from the molecular cloud at different times18,28,29,30. Either way, the increasing deficit in s-process matter with increasing heliocentric distance provides a new means for identifying genetic relationships among planetary bodies. For instance, Mars formed at a greater heliocentric distance than Earth and should, therefore, be characterized by an s-process deficit that may be similar to those observed for enstatite and ordinary chondrites. Thus, high-precision Nd isotopic data for Martian meteorites will make it possible to determine the distinct sources of the building materials of Earth and Mars. This information is not only critical for dating the differentiation of Mars13, but also for testing models of terrestrial planet formation.

Methods

Samples

To avoid the artefacts that can be associated with incomplete dissolution of refractory presolar components and to minimize potential disturbances through terrestrial alteration, only equilibrated chondrites (petrologic classes 4–6; except the CV3 Allende) from observed falls were selected for this study. Equilibrated chondrites are devoid of presolar grains because these components were destroyed during thermal metamorphism on the meteorite parent body32; for Allende (metamorphic grade of 3.2 to >3.6), which may contain trace amounts of presolar grains32, no difference in Nd isotopic composition was observed between table-top acid-digested, bomb-digested and alkali-fused samples6,17, indicating that for this meteorite all Nd carriers are accessed by standard acid digestion. Our sample set includes eleven ordinary chondrites (six H, two L and three LL), six enstatite chondrites (three EL and three EH), the carbonaceous chondrite Allende and the brachinite-like achondrite NWA 5363, which is a melt-depleted ultramafic sample from a partially differentiated asteroid33. This brachinite-like sample was added to the study because of its unique isotope anomalies: while the oxygen and nickel isotopic compositions of NWA 5363 are indistinguishable from terrestrial values, it exhibits nucleosynthetic isotope anomalies in Ti, Ca, Mo and Ru that are closer to ordinary chondrites34. In addition to bulk meteorites, we analysed the CAI A-ZH-5 from the Allende chondrite and, to evaluate the accuracy of our analytical methods, we also processed the JNdi-1 standard, as well as the terrestrial basalt standards BHVO-2 and BIR-1.

Sample preparation and chemical separation of Nd and Sm

Meteorite pieces were cleaned with abrasive paper, ultrasonicated in methanol and subsequently crushed to a fine powder in an acid-cleaned agate mortar exclusively used for meteorite work at the Origins Laboratory, Chicago. For each analysis about 2 g of meteorite powder was digested in a HF–HNO3–HClO4 mixture and aqua regia in 90 ml Savillex teflon vials for about 10 d on a hotplate at 170 °C. After several dry-downs, ultrasonication and redissolution steps in aqua regia and HCl, the samples were redissolved in HCl and, once a clear solution was obtained, an aliquot of approximately 5% was taken for Sm and Nd concentration measurements by isotope dilution.

Chemical procedures for Sm and Nd concentration measurements

The 5% aliquots were sent from the Origins Laboratory to the Lawrence Livermore National Laboratory (LLNL), where they were equilibrated with a 149Sm–150Nd mixed isotopic tracer. Rare earth elements (REE) were purified from the matrix of these aliquots using 2 ml BioRad columns filled with AG50-X8 (200–400 mesh) resin and 2 M and 6 M HCl. The REE were further purified using 150 μl Teflon columns with RE-Spec resin and 1 M and 0.05 M HNO3. Sm and Nd were purified from the other REE using 15 cm glass columns, Ln-Spec resin and 0.25 M and 0.60 M HCl. Total blanks of the isotope dilution procedures were 25 pg of Nd and 8 pg of Sm, resulting in Nd and Sm sample-to-blank ratios greater than 1,500 for all but one sample. The blank corrections resulted in shifts in the 147Sm/144Nd ratios that were less than 0.003% and thus substantially smaller than the typical uncertainty of 0.1% associated with the isotope dilution measurements. For NWA 5363, the Nd and Sm sample-to-blank ratios were 751 and 760, respectively, and thus required a blank correction of 0.13% for the Nd and Sm concentrations (for example, the reported 0.112 p.p.m. Nd abundance was corrected by 0.00015 p.p.m.). The blank correction is reflected in the larger uncertainty of 0.2% on the 147Sm/144Nd of NWA 5363.

Chemical procedures for Sm and Nd isotope composition measurements

After aliquoting, the remaining 95% of the sample solution was reduced and HNO3 was added. The REE cut of CAI A-ZH-5 that was obtained in a previous study35 (where the digested sample was processed through an anion exchange chromatography to separate Ti, Zr, Hf, W and Mo from the matrix; for details see ref. 35) was added to the project at this point. After additional dry-downs in aqua regia and HNO3, samples were redissolved in approximately 35 ml of 3 M HNO3 and 350 mg of H3BO3 was added before the solutions were centrifuged. A fine-grained, black, low-density residue (probably carbon-based) was present for some of the chondrites at this point and was discarded; note that because we analysed equilibrated chondrites, this C-bearing phase does not contain presolar material and therefore does not influence the Nd isotopic composition of the non-radiogenic isotopes. Furthermore, changes in the Sm/Nd ratios or the radiogenic Nd isotopic signatures of the samples by this material is also excluded, given the very good agreement of our decay-corrected 142Nd and 143Nd data with previous studies (Fig. 1; Extended Data Fig. 2). After centrifugation, the REE were separated from the matrix elements by loading the solutions onto two 2 ml Eichrom TODGA ion exchange columns stacked on top of each other36. To further purify the REE cut, the separation was repeated using a single 2 ml TODGA column. Separation of Sm and Nd from interfering REEs was accomplished with 0.2 cm × 25 cm quartz columns with AG50W-X8 (NH4+ form, with a pH of around 7) as the stationary phase and 0.2 M alpha-hydroxyisobutyric acid (pH adjusted to 4.6) as the fluid phase. The Sm and Nd cuts were passed twice over this column at the University of Chicago and were then sent to LLNL. The Nd was further purified at LLNL using 0.2 M alpha-hydroxyisobutyric acid adjusted to a pH of 4.40 on pressurized quartz glass columns loaded with AG50W-X8 (NH4+ form) resin. Neodymium was separated from the alpha-hydroxyisobutyric acid using 2 ml columns loaded with AG50W-X8 (200-400 mesh) resin using water, 2 M HCl, and 6 M HCl. The yields of the chemical procedure were determined by ICP-MS on small aliquots of the processed Nd and Sm cuts and ranged between 62% and 95% for Nd (with a mean yield of 80%) and 56% and 98% for Sm (with a mean yield of 75%). The variable yields do not have any noticeable influence on the measured Nd and Sm isotopic compositions. This is indicated both by the fact that several samples processed multiple times displayed variable yields, but had very homogeneous isotopic compositions, and that the terrestrial rock standards passed through the chemistry have indistinguishable compositions from the unprocessed standard. These observations further suggest that either the exponential law is well-suited to correct any yield-related induced mass-dependent isotope variations, or that the sample loss is associated with processes that do not induce mass-dependent fractionation effects, for example, pipetting of the samples on the columns or the loss of dry sample material from the beakers by static effects. The latter erratic losses seem to be the most likely explanation for the variable yields, which vary in a non-systematic way within a chemical campaign and among multiple digestions of the same meteorites. The procedural blanks associated with Nd and Sm isotope composition measurements were 50 pg and 12 pg respectively, and thus contributed negligibly (<0.03% of total analyte) to the isotope compositions of the samples, requiring no corrections to be made.

Procedures of Nd and Sm isotope measurement by TIMS

The Nd isotope compositions were analysed using a ThermoScientific Triton thermal ionization mass spectrometer (TIMS) at LLNL. Neodymium was loaded on zone-refined Re filaments in 2 M HCl and analysed as Nd+ using a second Re ionization filament. Isotope ratios were measured using a two-mass-step procedure that calculates 142Nd/144Nd and 148Nd/144Nd ratios dynamically, while measuring the other Nd isotopes statically following a modified version of previously established procedures17. The cup configuration of lines 1 and 2 are: L3 = 142Nd, L2 = 143Nd, L1 = 144Nd, C = 145Nd, H1 = 146Nd, H2 = 148Nd, H3 = 149Sm and H4 = 150Nd, and L3 = 140Ce, L2 = 141Pr, L1 = 142Nd, C = 143Nd, H1 = 144Nd, H2 = 146Nd, H3 = 147Sm and H4 = 148Nd, respectively. Individual mass spectrometer runs consisted of 540 ratios of 8 s integrations. The dynamic 142Nd/144Nd ratio is calculated from 142Nd/144Nd measured in cycle 2 normalized to 146Nd/144Nd measured in cycle 1, whereas the dynamic 148Nd/144Nd ratio is calculated from the 148Nd/146Nd ratio measured in cycle 1 normalized to 146Nd/144Nd measured in cycle 2. The 143Nd/144Nd ratio is calculated from the average of the 1,080 ratios of data collected in cycles 1 and 2. The 145Nd/144Nd ratio represents the average of 540 ratios collected in cycle 1. Most samples were run at least twice from the same filaments. Signal sizes varied from 144Nd = 3.2 × 10−11 A to 5.4 × 10−11 A, with most averaging in excess of 4.3 × 10−11 A. Fractionation was corrected by assuming 146Nd/144Nd = 0.7219 using the exponential law. The Nd isotope data were acquired in three measurement campaigns that were separated by a cup exchange and maintenance work on the Triton. To avoid any bias that might have been introduced by these events, the data obtained in each of the campaigns were normalized to the mean JNdi-1 composition measured in the respective campaign (see Supplementary Information). The external reproducibility of the standard (2 s.d.) for 142Nd/144Nd, 145Nd/144Nd, 148Nd/144Nd and 150Nd/144Nd are: 5 p.p.m., 9 p.p.m., 3 p.p.m. and 24 p.p.m. in campaign 1; 6 p.p.m., 6 p.p.m., 7 p.p.m. and 24 p.p.m. in campaign 2; and 8 p.p.m., 13 p.p.m., 15 p.p.m. and 31 p.p.m. in campaign 3. Table 1 presents the average values of multiple measurements from the same filament. The associated uncertainties represent the external reproducibility (2 s.d.) of the standard during that campaign, or the uncertainty of the sample measurements (2 s.e.), which were larger than the external reproducibility of the standard (3 p.p.m.) for some of the 148Nd/144Nd sample runs in campaign 1. Interferences from Ce and Sm are monitored at 140Ce and 149Sm and are presented in Supplementary Table 1.

Samarium was loaded in 2 M HCl onto a zone-refined Re filament and analysed as Sm+ using double Re filaments. All Sm isotopes, along with the interferences from Nd (measured as 146Nd) were measured statically for 200 ratios of 8 s integration each. Instrument fractionation was corrected by assuming 147Sm/152Sm = 0.56803 using the exponential law. The cup configuration for Sm isotope composition measurements is: L4 = 144Sm, L3 = 146Nd, L2 = 147Sm, L1 = 148Sm, C = 149Sm, H1 = 150Sm, H2 = 152Sm, H3 = 154Sm and H4 = 155Gd. Sample measurements consisted of one to three static runs from the same filament, depending on the amount of Sm available, and were obtained at (1–2) × 10−11 A 149Sm. The data were acquired in three campaigns and are given in the Supplementary Information. Samarium isotope anomalies were calculated relative to the mean composition of the AMES Sm standard analysed in each campaign (see Supplementary Information). The external reproducibility of the standard for 144Sm/152Sm, 148Sm/152Sm, 149Sm/152Sm, 150Sm/152Sm and 154Sm/152Sm are: 22 p.p.m., 12 p.p.m., 14 p.p.m., 12 p.p.m. and 18 p.p.m. in campaign 1; 43 p.p.m., 10 p.p.m., 10 p.p.m., 18 p.p.m. and 13 p.p.m. in campaign 2; and 38 p.p.m., 10 p.p.m., 12 p.p.m., 13 p.p.m. and 11 p.p.m. in campaign 3. Table 1 presents average values of the multiple measurements run from the same filament, and the reported uncertainties are 2 s.d. of the standard.

The Nd and Sm concentrations were determined using a ThermoScientific TIMS in static mode. Measurements consisted of 200 cycles with 8 s integration time each. Concentration data and 147Sm/144Nd ratios are given in Supplementary Table 3. Note that the nucleosynthetic anomalies measured here have no noticeable effect on the accuracy and precision of the Sm and Nd concentration measurement (the minimum variation in the Sm and Nd isotopic compositions that would be required to shift the 147Sm/144Nd ratios beyond uncertainty are 270 μiSm and 560 μiNd units, respectively; and thus substantially larger than the deviations we observed).

Isotopic mass balance calculations between CAIs and Allende

CAIs found in carbonaceous chondrites are considered to be the oldest surviving objects to have formed in the solar nebula, presumably by condensation from nebular gas. They often exhibit isotopic anomalies that are substantially different from their chondrite host rocks16,18,22,23,37, strongly suggesting that they are not genetically related to the reservoir from which the other chondrite components (namely chondrules and matrix) originated. The Nd and Sm isotopic composition of bulk carbonaceous chondrites is thus most likely to be influenced by CAIs, especially since the (light) REEs in these objects are enriched relative to the host rocks (for example, up to around 20× for CAIs from the CV (Vigarano-like) chondrite group, up to around 100× for CAIs from the CM (Mighei-like) chondrite group).

Indeed, our measurements imply that CAI material exert a strong control on the Nd and Sm isotope composition of bulk carbonaceous chondrites, because our Allende data (as well as literature data of carbonaceous chondrites) are displaced towards the CAI composition in μiNd versus μjNd, μiNd versus μjSm and μiSm versus μjSm (with i and j representing different isotopes of the Nd and Sm diagrams (Figs 2, 3; Extended Data Fig. 3).

To quantify the effect of CAIs on the Allende composition and characterize the composition of the CAI-free carbonaceous chondrite source reservoir we performed an isotopic mass balance calculation. For Nd this has the form

where NdAllende is the concentration of Nd in Allende, which is given by the sum of Nd in the carbonaceous chondrite source reservoir (Ndsource) and the Nd contributed by the CAIs (NdCAI) and X is the fraction of non-CAI material in Allende.

For the isotopic composition we can likewise write

Using the isotopic compositions measured for Allende (this study), Allende CAIs (the mean value of 11 CAIs reported in ref. 22) and 3 wt% CAIs in Allende38, and mean Nd concentrations of 0.967 p.p.m. and 14 p.p.m. for Allende and Allende CAIs31, we can solve for the unknown concentration and isotopic composition of the CAI-free material according to

and

The uncertainty in μiNdsource is mainly determined by the uncertainties in the measured isotopic compositions of Allende and the CAIs and was calculated by propagating them according to

Where F refers to the function given in Equation (4). Equivalent equations can be written for Sm. The mass balance calculation was performed using mean Sm concentrations of 0.313 p.p.m. and 4.54 p.p.m. for Allende and the CAIs, respectively (that is, with chondritic Sm/Nd ratios for both objects). All input parameters and the resulting composition of the carbonaceous chondrite source reservoir are also given in Extended Data Table 2.

The Nd and Sm mass balance calculations indicate that the CAI-free carbonaceous chondrite source reservoir is characterized by an s-deficit relative to the Earth and the other chondrites, in both Nd and Sm isotopes. This is consistent with information derived from other isotope systems (for example, Sr, Zr, Mo, Ru) where carbonaceous chondrites are characterized by the largest s-deficits relative to the Earth, followed by ordinary and enstatite chondrites16,28,29,39. We note that carbonaceous chondrite data obtained in previous studies15,17 also plot along the mass balance mixing relation between CAIs and a CAI-free carbonaceous chondrite source. This implies that the isotopic compositions of the other carbonaceous chondrites are also influenced by CAI-like material, and that they derive from a common s-depleted reservoir. The fact that some of the other carbonaceous chondrites also plot on the mixing line close to the bulk Allende values, despite containing fewer CAIs than CV chondrites, might be due to the higher REE enrichments in these non-CV CAIs (for example, hibonites in CM chondrites) or the fact that CAI-like material is not present in the form of well-defined inclusions but could be dispersed in the matrix in the form of small dust grains that are partially altered by parent-body metamorphism. Because no Sm and Nd isotope data for non-CV carbonaceous chondrite CAIs are available, one can only speculate on whether or not these CAIs also might carry larger nucleosynthetic Sm and Nd anomalies than Allende CAIs.

In principle, the Nd and Sm isotope compositions observed in ordinary and enstatite chondrites could also be influenced by CAIs. However, petrographic and chemical investigations imply that CAI-like material in these chondrite types is extremely rare38,40,41,42; and no Sm and Nd isotope data of these objects are available. Nevertheless, the effect of CAIs on the measured bulk Nd and Sm isotope composition of enstatite and ordinary chondrites is estimated to be no larger than 2 p.p.m. for Nd and 5 p.p.m. for Sm, respectively (Extended Data Table 2). This calculation assumes that the CAI-like material in ordinary and enstatite chondrites has a maximum REE enrichment of 50 times the concentrations in the CI (Ivuna-like) group of chondrites and an isotopic composition similar to normal Allende CAIs, and that the maximum CAI abundance in these chondrites is 0.05 wt.%. Given the small effects, we have omitted any correction of our measured data. However, we note that any such correction would result in slightly larger anomalies in non-radiogenic Nd isotopes and thus a higher μ142Nds-corrected; that is, an even better agreement between the nucleosynthetic anomaly-corrected μ142Nd values of meteorites and the accessible Earth.

CAIs exhibit isotope anomalies in Nd and Sm, but also in many other elements16,18,22,30,37. To explore the collateral effects of the mass balance between CAIs and Allende defined above for Nd and Sm on other isotope systems, we also applied it to Ca, Ti, Cr, Ni, Sr, Zr, Mo and Ba. The input parameters and results are given in Extended Data Table 4. Compared with the results from Nd and Sm, the isotopic compositions calculated for the CAI-free carbonaceous chondrite source reservoir for Ca, Ti, Cr, Ni, Sr, Zr, Mo and Ba do not differ greatly from the bulk Allende values (the most noticeable change is the reduction of the μ50Ti anomaly from 365 ± 34 for bulk Allende to 221 ± 46 for the CAI-free component, consistent with the measured value (189 ± 6) of a CAI-free Allende sample18). This is explained by the fact that the chemical enrichment of these elements in the CAIs relative to the host rock are not as strong as for Nd and Sm, and that the anomalies in the CAIs and bulk Allende are less disparate than for Nd and Sm. In other words, the CAIs have less influence on the bulk Allende isotopic composition for Ca, Ti, Cr, Ni, Sr, Zr, Mo and Ba than they have for Nd and Sm. We note, however, that the calculated CAI-free Allende compositions for the Sr, Zr, and Mo isotope anomalies are fully consistent with the inferences made above from Nd and Sm—that is, the formation of the carbonaceous chondrites from a nebular reservoir that is depleted in s-process material relative to Earth.